From 9396e2da3a059db28542cfbed3b2efc5e23b221c Mon Sep 17 00:00:00 2001 From: Steve White Date: Fri, 24 Jan 2025 09:26:47 -0600 Subject: [PATCH] Initial Commit; working system --- .gitignore | 36 + 20250101-20250105-cat-cs.AI.json | 1 + README.md | 167 + arxiv-processor/20250123-papers.json | 277 + arxiv-processor/README.md | 130 + arxiv-processor/arxiv-2501.11599v1.md | 156 + arxiv-processor/arxiv/client.go | 78 + arxiv-processor/go.mod | 3 + arxiv-processor/main.go | 214 + arxiv-processor/storage/json.go | 45 + criteria.txt | 43 + go.mod | 15 + json2md/README.md | 83 + json2md/go.mod | 3 + json2md/lib/lib.go | 136 + json2md/main.go | 35 + json2md/papers-output.md | 3997 ++++++++++++ llm_processor/README.md | 82 + llm_processor/client/client.go | 326 + llm_processor/go.mod | 10 + llm_processor/go.sum | 10 + llm_processor/main.go | 77 + llm_processor/models/models.go | 34 + llm_processor/newpapers-filtered.json | 1060 +++ llm_processor/newpapers-filtered.md | 3487 ++++++++++ llm_processor/newpapers-r1-filtered.json | 1060 +++ llm_processor/newpapers-r1-filtered.md | 3401 ++++++++++ llm_processor/papers.md | 4196 ++++++++++++ llm_processor/processor/processor.go | 190 + llm_processor/storage/storage.go | 143 + main.go | 238 + paper-system | Bin 0 -> 8009954 bytes prompt-criteria.txt | 29 + prompt-papers-old.md | 7421 ++++++++++++++++++++++ prompt-papers.md | 5903 +++++++++++++++++ 35 files changed, 33086 insertions(+) create mode 100644 .gitignore create mode 100644 20250101-20250105-cat-cs.AI.json create mode 100644 README.md create mode 100644 arxiv-processor/20250123-papers.json create mode 100644 arxiv-processor/README.md create mode 100644 arxiv-processor/arxiv-2501.11599v1.md create mode 100644 arxiv-processor/arxiv/client.go create mode 100644 arxiv-processor/go.mod create mode 100644 arxiv-processor/main.go create mode 100644 arxiv-processor/storage/json.go create mode 100644 criteria.txt create mode 100644 go.mod create mode 100644 json2md/README.md create mode 100644 json2md/go.mod create mode 100644 json2md/lib/lib.go create mode 100644 json2md/main.go create mode 100644 json2md/papers-output.md create mode 100644 llm_processor/README.md create mode 100644 llm_processor/client/client.go create mode 100644 llm_processor/go.mod create mode 100644 llm_processor/go.sum create mode 100644 llm_processor/main.go create mode 100644 llm_processor/models/models.go create mode 100644 llm_processor/newpapers-filtered.json create mode 100644 llm_processor/newpapers-filtered.md create mode 100644 llm_processor/newpapers-r1-filtered.json create mode 100644 llm_processor/newpapers-r1-filtered.md create mode 100644 llm_processor/papers.md create mode 100644 llm_processor/processor/processor.go create mode 100644 llm_processor/storage/storage.go create mode 100644 main.go create mode 100755 paper-system create mode 100644 prompt-criteria.txt create mode 100644 prompt-papers-old.md create mode 100644 prompt-papers.md diff --git a/.gitignore b/.gitignore new file mode 100644 index 0000000..7192c4b --- /dev/null +++ b/.gitignore @@ -0,0 +1,36 @@ +# OS generated files +.DS_Store + +# Editor directories and files +.vscode/ + +# Go binaries and build artifacts +/bin/ +/arxiv-processor/arxiv-processor +/json2md/json2md +/llm_processor/llm_processor + +# Dependency directories +vendor/ + +# Log files +*.log + +# Test files +*_test.go + +# Temporary files +temp_*.json +*.md.gz + +# Configuration files that might contain sensitive data +.env + +# Generated output files +papers.json +newpapers.md.gz +arxiv-processor/newpapers.json +arxiv-processor/papers_data.json +llm_processor/*-filt.* +llm_processor/results.json +llm_processor/oldresults.json diff --git a/20250101-20250105-cat-cs.AI.json b/20250101-20250105-cat-cs.AI.json new file mode 100644 index 0000000..9b2d755 --- /dev/null +++ b/20250101-20250105-cat-cs.AI.json @@ -0,0 +1 @@ +[{"title":"Beyond Text: Implementing Multimodal Large Language Model-Powered\n Multi-Agent Systems Using a No-Code Platform","abstract":" This study proposes the design and implementation of a multimodal LLM-based\nMulti-Agent System (MAS) leveraging a No-Code platform to address the practical\nconstraints and significant entry barriers associated with AI adoption in\nenterprises. Advanced AI technologies, such as Large Language Models (LLMs),\noften pose challenges due to their technical complexity and high implementation\ncosts, making them difficult for many organizations to adopt. To overcome these\nlimitations, this research develops a No-Code-based Multi-Agent System designed\nto enable users without programming knowledge to easily build and manage AI\nsystems. The study examines various use cases to validate the applicability of\nAI in business processes, including code generation from image-based notes,\nAdvanced RAG-based question-answering systems, text-based image generation, and\nvideo generation using images and prompts. These systems lower the barriers to\nAI adoption, empowering not only professional developers but also general users\nto harness AI for significantly improved productivity and efficiency. By\ndemonstrating the scalability and accessibility of No-Code platforms, this\nstudy advances the democratization of AI technologies within enterprises and\nvalidates the practical applicability of Multi-Agent Systems, ultimately\ncontributing to the widespread adoption of AI across various industries.\n","arxiv_id":"http://arxiv.org/abs/2501.00750v1","authors":["Cheonsu Jeong"]},{"title":"A3: Android Agent Arena for Mobile GUI Agents","abstract":" AI agents have become increasingly prevalent in recent years, driven by\nsignificant advancements in the field of large language models (LLMs). Mobile\nGUI agents, a subset of AI agents, are designed to autonomously perform tasks\non mobile devices. While numerous studies have introduced agents, datasets, and\nbenchmarks to advance mobile GUI agent research, many existing datasets focus\non static frame evaluations and fail to provide a comprehensive platform for\nassessing performance on real-world, in-the-wild tasks. To address this gap, we\npresent Android Agent Arena (A3), a novel evaluation platform. Unlike existing\nin-the-wild systems, A3 offers: (1) meaningful and practical tasks, such as\nreal-time online information retrieval and operational instructions; (2) a\nlarger, more flexible action space, enabling compatibility with agents trained\non any dataset; and (3) automated business-level LLM-based evaluation process.\nA3 includes 21 widely used general third-party apps and 201 tasks\nrepresentative of common user scenarios, providing a robust foundation for\nevaluating mobile GUI agents in real-world situations and a new autonomous\nevaluation process for less human labor and coding expertise. The project is\navailable at \\url{https://yuxiangchai.github.io/Android-Agent-Arena/}.\n","arxiv_id":"http://arxiv.org/abs/2501.01149v1","authors":["Yuxiang Chai","Hanhao Li","Jiayu Zhang","Liang Liu","Guozhi Wang","Shuai Ren","Siyuan Huang","Hongsheng Li"]},{"title":"Rethinking Relation Extraction: Beyond Shortcuts to Generalization with\n a Debiased Benchmark","abstract":" Benchmarks are crucial for evaluating machine learning algorithm performance,\nfacilitating comparison and identifying superior solutions. However, biases\nwithin datasets can lead models to learn shortcut patterns, resulting in\ninaccurate assessments and hindering real-world applicability. This paper\naddresses the issue of entity bias in relation extraction tasks, where models\ntend to rely on entity mentions rather than context. We propose a debiased\nrelation extraction benchmark DREB that breaks the pseudo-correlation between\nentity mentions and relation types through entity replacement. DREB utilizes\nBias Evaluator and PPL Evaluator to ensure low bias and high naturalness,\nproviding a reliable and accurate assessment of model generalization in entity\nbias scenarios. To establish a new baseline on DREB, we introduce MixDebias, a\ndebiasing method combining data-level and model training-level techniques.\nMixDebias effectively improves model performance on DREB while maintaining\nperformance on the original dataset. Extensive experiments demonstrate the\neffectiveness and robustness of MixDebias compared to existing methods,\nhighlighting its potential for improving the generalization ability of relation\nextraction models. We will release DREB and MixDebias publicly.\n","arxiv_id":"http://arxiv.org/abs/2501.01349v1","authors":["Liang He","Yougang Chu","Zhen Wu","Jianbing Zhang","Xinyu Dai","Jiajun Chen"]},{"title":"ASKCOS: an open source software suite for synthesis planning","abstract":" The advancement of machine learning and the availability of large-scale\nreaction datasets have accelerated the development of data-driven models for\ncomputer-aided synthesis planning (CASP) in the past decade. Here, we detail\nthe newest version of ASKCOS, an open source software suite for synthesis\nplanning that makes available several research advances in a freely available,\npractical tool. Four one-step retrosynthesis models form the basis of both\ninteractive planning and automatic planning modes. Retrosynthetic planning is\ncomplemented by other modules for feasibility assessment and pathway\nevaluation, including reaction condition recommendation, reaction outcome\nprediction, and auxiliary capabilities such as solubility prediction and\nquantum mechanical descriptor prediction. ASKCOS has assisted hundreds of\nmedicinal, synthetic, and process chemists in their day-to-day tasks,\ncomplementing expert decision making. It is our belief that CASP tools like\nASKCOS are an important part of modern chemistry research, and that they offer\never-increasing utility and accessibility.\n","arxiv_id":"http://arxiv.org/abs/2501.01835v1","authors":["Zhengkai Tu","Sourabh J. Choure","Mun Hong Fong","Jihye Roh","Itai Levin","Kevin Yu","Joonyoung F. Joung","Nathan Morgan","Shih-Cheng Li","Xiaoqi Sun","Huiqian Lin","Mark Murnin","Jordan P. Liles","Thomas J. Struble","Michael E. Fortunato","Mengjie Liu","William H. Green","Klavs F. Jensen","Connor W. Coley"]},{"title":"eRevise+RF: A Writing Evaluation System for Assessing Student Essay\n Revisions and Providing Formative Feedback","abstract":" The ability to revise essays in response to feedback is important for\nstudents' writing success. An automated writing evaluation (AWE) system that\nsupports students in revising their essays is thus essential. We present\neRevise+RF, an enhanced AWE system for assessing student essay revisions (e.g.,\nchanges made to an essay to improve its quality in response to essay feedback)\nand providing revision feedback. We deployed the system with 6 teachers and 406\nstudents across 3 schools in Pennsylvania and Louisiana. The results confirmed\nits effectiveness in (1) assessing student essays in terms of evidence usage,\n(2) extracting evidence and reasoning revisions across essays, and (3)\ndetermining revision success in responding to feedback. The evaluation also\nsuggested eRevise+RF is a helpful system for young students to improve their\nargumentative writing skills through revision and formative feedback.\n","arxiv_id":"http://arxiv.org/abs/2501.00715v1","authors":["Zhexiong Liu","Diane Litman","Elaine Wang","Tianwen Li","Mason Gobat","Lindsay Clare Matsumura","Richard Correnti"]},{"title":"Towards End-to-End Neuromorphic Voxel-based 3D Object Reconstruction\n Without Physical Priors","abstract":" Neuromorphic cameras, also known as event cameras, are asynchronous\nbrightness-change sensors that can capture extremely fast motion without\nsuffering from motion blur, making them particularly promising for 3D\nreconstruction in extreme environments. However, existing research on 3D\nreconstruction using monocular neuromorphic cameras is limited, and most of the\nmethods rely on estimating physical priors and employ complex multi-step\npipelines. In this work, we propose an end-to-end method for dense voxel 3D\nreconstruction using neuromorphic cameras that eliminates the need to estimate\nphysical priors. Our method incorporates a novel event representation to\nenhance edge features, enabling the proposed feature-enhancement model to learn\nmore effectively. Additionally, we introduced Optimal Binarization Threshold\nSelection Principle as a guideline for future related work, using the optimal\nreconstruction results achieved with threshold optimization as the benchmark.\nOur method achieves a 54.6% improvement in reconstruction accuracy compared to\nthe baseline method.\n","arxiv_id":"http://arxiv.org/abs/2501.00741v1","authors":["Chuanzhi Xu","Langyi Chen","Vincent Qu","Haodong Chen","Vera Chung"]},{"title":"AttriReBoost: A Gradient-Free Propagation Optimization Method for Cold\n Start Mitigation in Attribute Missing Graphs","abstract":" Missing attribute issues are prevalent in the graph learning, leading to\nbiased outcomes in Graph Neural Networks (GNNs). Existing methods that rely on\nfeature propagation are prone to cold start problem, particularly when dealing\nwith attribute resetting and low-degree nodes, which hinder effective\npropagation and convergence. To address these challenges, we propose\nAttriReBoost (ARB), a novel method that incorporates propagation-based method\nto mitigate cold start problems in attribute-missing graphs. ARB enhances\nglobal feature propagation by redefining initial boundary conditions and\nstrategically integrating virtual edges, thereby improving node connectivity\nand ensuring more stable and efficient convergence. This method facilitates\ngradient-free attribute reconstruction with lower computational overhead. The\nproposed method is theoretically grounded, with its convergence rigorously\nestablished. Extensive experiments on several real-world benchmark datasets\ndemonstrate the effectiveness of ARB, achieving an average accuracy improvement\nof 5.11% over state-of-the-art methods. Additionally, ARB exhibits remarkable\ncomputational efficiency, processing a large-scale graph with 2.49 million\nnodes in just 16 seconds on a single GPU. Our code is available at\nhttps://github.com/limengran98/ARB.\n","arxiv_id":"http://arxiv.org/abs/2501.00743v1","authors":["Mengran Li","Chaojun Ding","Junzhou Chen","Wenbin Xing","Cong Ye","Ronghui Zhang","Songlin Zhuang","Jia Hu","Tony Z. Qiu","Huijun Gao"]},{"title":"Enhancing Transformers for Generalizable First-Order Logical Entailment","abstract":" Transformers, as a fundamental deep learning architecture, have demonstrated\nremarkable capabilities in reasoning. This paper investigates the generalizable\nfirst-order logical reasoning ability of transformers with their parameterized\nknowledge and explores ways to improve it. The first-order reasoning capability\nof transformers is assessed through their ability to perform first-order\nlogical entailment, which is quantitatively measured by their performance in\nanswering knowledge graph queries. We establish connections between (1) two\ntypes of distribution shifts studied in out-of-distribution generalization and\n(2) the unseen knowledge and query settings discussed in the task of knowledge\ngraph query answering, enabling a characterization of fine-grained\ngeneralizability. Results on our comprehensive dataset show that transformers\noutperform previous methods specifically designed for this task and provide\ndetailed empirical evidence on the impact of input query syntax, token\nembedding, and transformer architectures on the reasoning capability of\ntransformers. Interestingly, our findings reveal a mismatch between positional\nencoding and other design choices in transformer architectures employed in\nprior practices. This discovery motivates us to propose a more sophisticated,\nlogic-aware architecture, TEGA, to enhance the capability for generalizable\nfirst-order logical entailment in transformers.\n","arxiv_id":"http://arxiv.org/abs/2501.00759v1","authors":["Tianshi Zheng","Jiazheng Wang","Zihao Wang","Jiaxin Bai","Hang Yin","Zheye Deng","Yangqiu Song","Jianxin Li"]},{"title":"REM: A Scalable Reinforced Multi-Expert Framework for Multiplex\n Influence Maximization","abstract":" In social online platforms, identifying influential seed users to maximize\ninfluence spread is a crucial as it can greatly diminish the cost and efforts\nrequired for information dissemination. While effective, traditional methods\nfor Multiplex Influence Maximization (MIM) have reached their performance\nlimits, prompting the emergence of learning-based approaches. These novel\nmethods aim for better generalization and scalability for more sizable graphs\nbut face significant challenges, such as (1) inability to handle unknown\ndiffusion patterns and (2) reliance on high-quality training samples. To\naddress these issues, we propose the Reinforced Expert Maximization framework\n(REM). REM leverages a Propagation Mixture of Experts technique to encode\ndynamic propagation of large multiplex networks effectively in order to\ngenerate enhanced influence propagation. Noticeably, REM treats a generative\nmodel as a policy to autonomously generate different seed sets and learn how to\nimprove them from a Reinforcement Learning perspective. Extensive experiments\non several real-world datasets demonstrate that REM surpasses state-of-the-art\nmethods in terms of influence spread, scalability, and inference time in\ninfluence maximization tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.00779v1","authors":["Huyen Nguyen","Hieu Dam","Nguyen Do","Cong Tran","Cuong Pham"]},{"title":"Make Shuffling Great Again: A Side-Channel Resistant Fisher-Yates\n Algorithm for Protecting Neural Networks","abstract":" Neural network models implemented in embedded devices have been shown to be\nsusceptible to side-channel attacks (SCAs), allowing recovery of proprietary\nmodel parameters, such as weights and biases. There are already available\ncountermeasure methods currently used for protecting cryptographic\nimplementations that can be tailored to protect embedded neural network models.\nShuffling, a hiding-based countermeasure that randomly shuffles the order of\ncomputations, was shown to be vulnerable to SCA when the Fisher-Yates algorithm\nis used. In this paper, we propose a design of an SCA-secure version of the\nFisher-Yates algorithm. By integrating the masking technique for modular\nreduction and Blakely's method for modular multiplication, we effectively\nremove the vulnerability in the division operation that led to side-channel\nleakage in the original version of the algorithm. We experimentally evaluate\nthat the countermeasure is effective against SCA by implementing a correlation\npower analysis attack on an embedded neural network model implemented on ARM\nCortex-M4. Compared to the original proposal, the memory overhead is $2\\times$\nthe biggest layer of the network, while the time overhead varies from $4\\%$ to\n$0.49\\%$ for a layer with $100$ and $1000$ neurons, respectively.\n","arxiv_id":"http://arxiv.org/abs/2501.00798v1","authors":["Leonard Puškáč","Marek Benovič","Jakub Breier","Xiaolu Hou"]},{"title":"Reasoning-Oriented and Analogy-Based Methods for Locating and Editing in\n Zero-Shot Event-Relational Reasoning","abstract":" Zero-shot event-relational reasoning is an important task in natural language\nprocessing, and existing methods jointly learn a variety of event-relational\nprefixes and inference-form prefixes to achieve such tasks. However, training\nprefixes consumes large computational resources and lacks interpretability.\nAdditionally, learning various relational and inferential knowledge\ninefficiently exploits the connections between tasks. Therefore, we first\npropose a method for Reasoning-Oriented Locating and Editing (ROLE), which\nlocates and edits the key modules of the language model for reasoning about\nevent relations, enhancing interpretability and also resource-efficiently\noptimizing the reasoning ability. Subsequently, we propose a method for\nAnalogy-Based Locating and Editing (ABLE), which efficiently exploits the\nsimilarities and differences between tasks to optimize the zero-shot reasoning\ncapability. Experimental results show that ROLE improves interpretability and\nreasoning performance with reduced computational cost. ABLE achieves SOTA\nresults in zero-shot reasoning.\n","arxiv_id":"http://arxiv.org/abs/2501.00803v1","authors":["Jingyao Tang","Lishuang Li","Liteng Mi","Haiming Wu","Hongbin Lu"]},{"title":"LLM-Powered Multi-Agent System for Automated Crypto Portfolio Management","abstract":" Cryptocurrency investment is inherently difficult due to its shorter history\ncompared to traditional assets, the need to integrate vast amounts of data from\nvarious modalities, and the requirement for complex reasoning. While deep\nlearning approaches have been applied to address these challenges, their\nblack-box nature raises concerns about trust and explainability. Recently,\nlarge language models (LLMs) have shown promise in financial applications due\nto their ability to understand multi-modal data and generate explainable\ndecisions. However, single LLM faces limitations in complex, comprehensive\ntasks such as asset investment. These limitations are even more pronounced in\ncryptocurrency investment, where LLMs have less domain-specific knowledge in\ntheir training corpora.\n To overcome these challenges, we propose an explainable, multi-modal,\nmulti-agent framework for cryptocurrency investment. Our framework uses\nspecialized agents that collaborate within and across teams to handle subtasks\nsuch as data analysis, literature integration, and investment decision-making\nfor the top 30 cryptocurrencies by market capitalization. The expert training\nmodule fine-tunes agents using multi-modal historical data and professional\ninvestment literature, while the multi-agent investment module employs\nreal-time data to make informed cryptocurrency investment decisions. Unique\nintrateam and interteam collaboration mechanisms enhance prediction accuracy by\nadjusting final predictions based on confidence levels within agent teams and\nfacilitating information sharing between teams. Empirical evaluation using data\nfrom November 2023 to September 2024 demonstrates that our framework\noutperforms single-agent models and market benchmarks in classification, asset\npricing, portfolio, and explainability performance.\n","arxiv_id":"http://arxiv.org/abs/2501.00826v2","authors":["Yichen Luo","Yebo Feng","Jiahua Xu","Paolo Tasca","Yang Liu"]},{"title":"Embedding Style Beyond Topics: Analyzing Dispersion Effects Across\n Different Language Models","abstract":" This paper analyzes how writing style affects the dispersion of embedding\nvectors across multiple, state-of-the-art language models. While early\ntransformer models primarily aligned with topic modeling, this study examines\nthe role of writing style in shaping embedding spaces. Using a literary corpus\nthat alternates between topics and styles, we compare the sensitivity of\nlanguage models across French and English. By analyzing the particular impact\nof style on embedding dispersion, we aim to better understand how language\nmodels process stylistic information, contributing to their overall\ninterpretability.\n","arxiv_id":"http://arxiv.org/abs/2501.00828v1","authors":["Benjamin Icard","Evangelia Zve","Lila Sainero","Alice Breton","Jean-Gabriel Ganascia"]},{"title":"An LLM-Empowered Adaptive Evolutionary Algorithm For Multi-Component\n Deep Learning Systems","abstract":" Multi-objective evolutionary algorithms (MOEAs) are widely used for searching\noptimal solutions in complex multi-component applications. Traditional MOEAs\nfor multi-component deep learning (MCDL) systems face challenges in enhancing\nthe search efficiency while maintaining the diversity. To combat these, this\npaper proposes $\\mu$MOEA, the first LLM-empowered adaptive evolutionary search\nalgorithm to detect safety violations in MCDL systems. Inspired by the\ncontext-understanding ability of Large Language Models (LLMs), $\\mu$MOEA\npromotes the LLM to comprehend the optimization problem and generate an initial\npopulation tailed to evolutionary objectives. Subsequently, it employs adaptive\nselection and variation to iteratively produce offspring, balancing the\nevolutionary efficiency and diversity. During the evolutionary process, to\nnavigate away from the local optima, $\\mu$MOEA integrates the evolutionary\nexperience back into the LLM. This utilization harnesses the LLM's quantitative\nreasoning prowess to generate differential seeds, breaking away from current\noptimal solutions. We evaluate $\\mu$MOEA in finding safety violations of MCDL\nsystems, and compare its performance with state-of-the-art MOEA methods.\nExperimental results show that $\\mu$MOEA can significantly improve the\nefficiency and diversity of the evolutionary search.\n","arxiv_id":"http://arxiv.org/abs/2501.00829v1","authors":["Haoxiang Tian","Xingshuo Han","Guoquan Wu","An Guo","Yuan Zhou. Jie Zhang","Shuo Li","Jun Wei","Tianwei Zhang"]},{"title":"LLM+AL: Bridging Large Language Models and Action Languages for Complex\n Reasoning about Actions","abstract":" Large Language Models (LLMs) have made significant strides in various\nintelligent tasks but still struggle with complex action reasoning tasks that\nrequire systematic search. To address this limitation, we propose a method that\nbridges the natural language understanding capabilities of LLMs with the\nsymbolic reasoning strengths of action languages. Our approach, termed\n\"LLM+AL,\" leverages the LLM's strengths in semantic parsing and commonsense\nknowledge generation alongside the action language's proficiency in automated\nreasoning based on encoded knowledge. We compare LLM+AL against\nstate-of-the-art LLMs, including ChatGPT-4, Claude 3 Opus, Gemini Ultra 1.0,\nand o1-preview, using benchmarks for complex reasoning about actions. Our\nfindings indicate that, although all methods exhibit errors, LLM+AL, with\nrelatively minimal human corrections, consistently leads to correct answers,\nwhereas standalone LLMs fail to improve even with human feedback. LLM+AL also\ncontributes to automated generation of action languages.\n","arxiv_id":"http://arxiv.org/abs/2501.00830v1","authors":["Adam Ishay","Joohyung Lee"]},{"title":"Distilled Lifelong Self-Adaptation for Configurable Systems","abstract":" Modern configurable systems provide tremendous opportunities for engineering\nfuture intelligent software systems. A key difficulty thereof is how to\neffectively self-adapt the configuration of a running system such that its\nperformance (e.g., runtime and throughput) can be optimized under time-varying\nworkloads. This unfortunately remains unaddressed in existing approaches as\nthey either overlook the available past knowledge or rely on static\nexploitation of past knowledge without reasoning the usefulness of information\nwhen planning for self-adaptation. In this paper, we tackle this challenging\nproblem by proposing DLiSA, a framework that self-adapts configurable systems.\nDLiSA comes with two properties: firstly, it supports lifelong planning, and\nthereby the planning process runs continuously throughout the lifetime of the\nsystem, allowing dynamic exploitation of the accumulated knowledge for rapid\nadaptation. Secondly, the planning for a newly emerged workload is boosted via\ndistilled knowledge seeding, in which the knowledge is dynamically purified\nsuch that only useful past configurations are seeded when necessary, mitigating\nmisleading information. Extensive experiments suggest that the proposed DLiSA\nsignificantly outperforms state-of-the-art approaches, demonstrating a\nperformance improvement of up to 229% and a resource acceleration of up to\n2.22x on generating promising adaptation configurations. All data and sources\ncan be found at our repository: https://github.com/ideas-labo/dlisa.\n","arxiv_id":"http://arxiv.org/abs/2501.00840v1","authors":["Yulong Ye","Tao Chen","Miqing Li"]},{"title":"Diversity Optimization for Travelling Salesman Problem via Deep\n Reinforcement Learning","abstract":" Existing neural methods for the Travelling Salesman Problem (TSP) mostly aim\nat finding a single optimal solution. To discover diverse yet high-quality\nsolutions for Multi-Solution TSP (MSTSP), we propose a novel deep reinforcement\nlearning based neural solver, which is primarily featured by an encoder-decoder\nstructured policy. Concretely, on the one hand, a Relativization Filter (RF) is\ndesigned to enhance the robustness of the encoder to affine transformations of\nthe instances, so as to potentially improve the quality of the found solutions.\nOn the other hand, a Multi-Attentive Adaptive Active Search (MA3S) is tailored\nto allow the decoders to strike a balance between the optimality and diversity.\nExperimental evaluations on benchmark instances demonstrate the superiority of\nour method over recent neural baselines across different metrics, and its\ncompetitive performance against state-of-the-art traditional heuristics with\nsignificantly reduced computational time, ranging from $1.3\\times$ to\n$15\\times$ faster. Furthermore, we demonstrate that our method can also be\napplied to the Capacitated Vehicle Routing Problem (CVRP).\n","arxiv_id":"http://arxiv.org/abs/2501.00884v1","authors":["Qi Li","Zhiguang Cao","Yining Ma","Yaoxin Wu","Yue-Jiao Gong"]},{"title":"Population Aware Diffusion for Time Series Generation","abstract":" Diffusion models have shown promising ability in generating high-quality time\nseries (TS) data. Despite the initial success, existing works mostly focus on\nthe authenticity of data at the individual level, but pay less attention to\npreserving the population-level properties on the entire dataset. Such\npopulation-level properties include value distributions for each dimension and\ndistributions of certain functional dependencies (e.g., cross-correlation, CC)\nbetween different dimensions. For instance, when generating house energy\nconsumption TS data, the value distributions of the outside temperature and the\nkitchen temperature should be preserved, as well as the distribution of CC\nbetween them. Preserving such TS population-level properties is critical in\nmaintaining the statistical insights of the datasets, mitigating model bias,\nand augmenting downstream tasks like TS prediction. Yet, it is often overlooked\nby existing models. Hence, data generated by existing models often bear\ndistribution shifts from the original data. We propose Population-aware\nDiffusion for Time Series (PaD-TS), a new TS generation model that better\npreserves the population-level properties. The key novelties of PaD-TS include\n1) a new training method explicitly incorporating TS population-level property\npreservation, and 2) a new dual-channel encoder model architecture that better\ncaptures the TS data structure. Empirical results in major benchmark datasets\nshow that PaD-TS can improve the average CC distribution shift score between\nreal and synthetic data by 5.9x while maintaining a performance comparable to\nstate-of-the-art models on individual-level authenticity.\n","arxiv_id":"http://arxiv.org/abs/2501.00910v1","authors":["Yang Li","Han Meng","Zhenyu Bi","Ingolv T. Urnes","Haipeng Chen"]},{"title":"$β$-DQN: Improving Deep Q-Learning By Evolving the Behavior","abstract":" While many sophisticated exploration methods have been proposed, their lack\nof generality and high computational cost often lead researchers to favor\nsimpler methods like $\\epsilon$-greedy. Motivated by this, we introduce\n$\\beta$-DQN, a simple and efficient exploration method that augments the\nstandard DQN with a behavior function $\\beta$. This function estimates the\nprobability that each action has been taken at each state. By leveraging\n$\\beta$, we generate a population of diverse policies that balance exploration\nbetween state-action coverage and overestimation bias correction. An adaptive\nmeta-controller is designed to select an effective policy for each episode,\nenabling flexible and explainable exploration. $\\beta$-DQN is straightforward\nto implement and adds minimal computational overhead to the standard DQN.\nExperiments on both simple and challenging exploration domains show that\n$\\beta$-DQN outperforms existing baseline methods across a wide range of tasks,\nproviding an effective solution for improving exploration in deep reinforcement\nlearning.\n","arxiv_id":"http://arxiv.org/abs/2501.00913v1","authors":["Hongming Zhang","Fengshuo Bai","Chenjun Xiao","Chao Gao","Bo Xu","Martin Müller"]},{"title":"Incremental Dialogue Management: Survey, Discussion, and Implications\n for HRI","abstract":" Efforts towards endowing robots with the ability to speak have benefited from\nrecent advancements in NLP, in particular large language models. However, as\npowerful as current models have become, they still operate on sentence or\nmulti-sentence level input, not on the word-by-word input that humans operate\non, affecting the degree of responsiveness that they offer, which is critical\nin situations where humans interact with robots using speech. In this paper, we\nreview the literature on interactive systems that operate incrementally (i.e.,\nat the word level or below it). We motivate the need for incremental systems,\nsurvey incremental modeling of important aspects of dialogue like speech\nrecognition and language generation. Primary focus is on the part of the system\nthat makes decisions, known as the dialogue manager. We find that there is very\nlittle research on incremental dialogue management, offer some requirements for\npractical incremental dialogue management, and the implications of incremental\ndialogue for embodied, robotic platforms.\n","arxiv_id":"http://arxiv.org/abs/2501.00953v1","authors":["Casey Kennington","Pierre Lison","David Schlangen"]},{"title":"Are LLMs effective psychological assessors? Leveraging adaptive RAG for\n interpretable mental health screening through psychometric practice","abstract":" In psychological practice, standardized questionnaires serve as essential\ntools for assessing mental constructs (e.g., attitudes, traits, and emotions)\nthrough structured questions (aka items). With the increasing prevalence of\nsocial media platforms where users share personal experiences and emotions,\nresearchers are exploring computational methods to leverage this data for rapid\nmental health screening. In this study, we propose a novel adaptive\nRetrieval-Augmented Generation (RAG) approach that completes psychological\nquestionnaires by analyzing social media posts. Our method retrieves the most\nrelevant user posts for each question in a psychological survey and uses Large\nLanguage Models (LLMs) to predict questionnaire scores in a zero-shot setting.\nOur findings are twofold. First we demonstrate that this approach can\neffectively predict users' responses to psychological questionnaires, such as\nthe Beck Depression Inventory II (BDI-II), achieving performance comparable to\nor surpassing state-of-the-art models on Reddit-based benchmark datasets\nwithout relying on training data. Second, we show how this methodology can be\ngeneralized as a scalable screening tool, as the final assessment is\nsystematically derived by completing standardized questionnaires and tracking\nhow individual item responses contribute to the diagnosis, aligning with\nestablished psychometric practices.\n","arxiv_id":"http://arxiv.org/abs/2501.00982v1","authors":["Federico Ravenda","Seyed Ali Bahrainian","Andrea Raballo","Antonietta Mira","Noriko Kando"]},{"title":"Bootstrapped Reward Shaping","abstract":" In reinforcement learning, especially in sparse-reward domains, many\nenvironment steps are required to observe reward information. In order to\nincrease the frequency of such observations, \"potential-based reward shaping\"\n(PBRS) has been proposed as a method of providing a more dense reward signal\nwhile leaving the optimal policy invariant. However, the required \"potential\nfunction\" must be carefully designed with task-dependent knowledge to not deter\ntraining performance. In this work, we propose a \"bootstrapped\" method of\nreward shaping, termed BSRS, in which the agent's current estimate of the\nstate-value function acts as the potential function for PBRS. We provide\nconvergence proofs for the tabular setting, give insights into training\ndynamics for deep RL, and show that the proposed method improves training speed\nin the Atari suite.\n","arxiv_id":"http://arxiv.org/abs/2501.00989v1","authors":["Jacob Adamczyk","Volodymyr Makarenko","Stas Tiomkin","Rahul V. Kulkarni"]},{"title":"Exploring Information Processing in Large Language Models: Insights from\n Information Bottleneck Theory","abstract":" Large Language Models (LLMs) have demonstrated remarkable performance across\na wide range of tasks by understanding input information and predicting\ncorresponding outputs. However, the internal mechanisms by which LLMs\ncomprehend input and make effective predictions remain poorly understood. In\nthis paper, we explore the working mechanism of LLMs in information processing\nfrom the perspective of Information Bottleneck Theory. We propose a\nnon-training construction strategy to define a task space and identify the\nfollowing key findings: (1) LLMs compress input information into specific task\nspaces (e.g., sentiment space, topic space) to facilitate task understanding;\n(2) they then extract and utilize relevant information from the task space at\ncritical moments to generate accurate predictions. Based on these insights, we\nintroduce two novel approaches: an Information Compression-based Context\nLearning (IC-ICL) and a Task-Space-guided Fine-Tuning (TS-FT). IC-ICL enhances\nreasoning performance and inference efficiency by compressing retrieved example\ninformation into the task space. TS-FT employs a space-guided loss to fine-tune\nLLMs, encouraging the learning of more effective compression and selection\nmechanisms. Experiments across multiple datasets validate the effectiveness of\ntask space construction. Additionally, IC-ICL not only improves performance but\nalso accelerates inference speed by over 40\\%, while TS-FT achieves superior\nresults with a minimal strategy adjustment.\n","arxiv_id":"http://arxiv.org/abs/2501.00999v2","authors":["Zhou Yang","Zhengyu Qi","Zhaochun Ren","Zhikai Jia","Haizhou Sun","Xiaofei Zhu","Xiangwen Liao"]},{"title":"Deep Reinforcement Learning for Job Scheduling and Resource Management\n in Cloud Computing: An Algorithm-Level Review","abstract":" Cloud computing has revolutionized the provisioning of computing resources,\noffering scalable, flexible, and on-demand services to meet the diverse\nrequirements of modern applications. At the heart of efficient cloud operations\nare job scheduling and resource management, which are critical for optimizing\nsystem performance and ensuring timely and cost-effective service delivery.\nHowever, the dynamic and heterogeneous nature of cloud environments presents\nsignificant challenges for these tasks, as workloads and resource availability\ncan fluctuate unpredictably. Traditional approaches, including heuristic and\nmeta-heuristic algorithms, often struggle to adapt to these real-time changes\ndue to their reliance on static models or predefined rules. Deep Reinforcement\nLearning (DRL) has emerged as a promising solution to these challenges by\nenabling systems to learn and adapt policies based on continuous observations\nof the environment, facilitating intelligent and responsive decision-making.\nThis survey provides a comprehensive review of DRL-based algorithms for job\nscheduling and resource management in cloud computing, analyzing their\nmethodologies, performance metrics, and practical applications. We also\nhighlight emerging trends and future research directions, offering valuable\ninsights into leveraging DRL to advance both job scheduling and resource\nmanagement in cloud computing.\n","arxiv_id":"http://arxiv.org/abs/2501.01007v1","authors":["Yan Gu","Zhaoze Liu","Shuhong Dai","Cong Liu","Ying Wang","Shen Wang","Georgios Theodoropoulos","Long Cheng"]},{"title":"MDSF: Context-Aware Multi-Dimensional Data Storytelling Framework based\n on Large language Model","abstract":" The exponential growth of data and advancements in big data technologies have\ncreated a demand for more efficient and automated approaches to data analysis\nand storytelling. However, automated data analysis systems still face\nchallenges in leveraging large language models (LLMs) for data insight\ndiscovery, augmented analysis, and data storytelling. This paper introduces the\nMultidimensional Data Storytelling Framework (MDSF) based on large language\nmodels for automated insight generation and context-aware storytelling. The\nframework incorporates advanced preprocessing techniques, augmented analysis\nalgorithms, and a unique scoring mechanism to identify and prioritize\nactionable insights. The use of fine-tuned LLMs enhances contextual\nunderstanding and generates narratives with minimal manual intervention. The\narchitecture also includes an agent-based mechanism for real-time storytelling\ncontinuation control. Key findings reveal that MDSF outperforms existing\nmethods across various datasets in terms of insight ranking accuracy,\ndescriptive quality, and narrative coherence. The experimental evaluation\ndemonstrates MDSF's ability to automate complex analytical tasks, reduce\ninterpretive biases, and improve user satisfaction. User studies further\nunderscore its practical utility in enhancing content structure, conclusion\nextraction, and richness of detail.\n","arxiv_id":"http://arxiv.org/abs/2501.01014v1","authors":["Chengze Zhang","Changshan Li","Shiyang Gao"]},{"title":"Towards Adversarially Robust Deep Metric Learning","abstract":" Deep Metric Learning (DML) has shown remarkable successes in many domains by\ntaking advantage of powerful deep neural networks. Deep neural networks are\nprone to adversarial attacks and could be easily fooled by adversarial\nexamples. The current progress on this robustness issue is mainly about deep\nclassification models but pays little attention to DML models. Existing works\nfail to thoroughly inspect the robustness of DML and neglect an important DML\nscenario, the clustering-based inference. In this work, we first point out the\nrobustness issue of DML models in clustering-based inference scenarios. We find\nthat, for the clustering-based inference, existing defenses designed DML are\nunable to be reused and the adaptions of defenses designed for deep\nclassification models cannot achieve satisfactory robustness performance. To\nalleviate the hazard of adversarial examples, we propose a new defense, the\nEnsemble Adversarial Training (EAT), which exploits ensemble learning and\nadversarial training. EAT promotes the diversity of the ensemble, encouraging\neach model in the ensemble to have different robustness features, and employs a\nself-transferring mechanism to make full use of the robustness statistics of\nthe whole ensemble in the update of every single model. We evaluate the EAT\nmethod on three widely-used datasets with two popular model architectures. The\nresults show that the proposed EAT method greatly outperforms the adaptions of\ndefenses designed for deep classification models.\n","arxiv_id":"http://arxiv.org/abs/2501.01025v2","authors":["Xiaopeng Ke"]},{"title":"Reasoning based on symbolic and parametric knowledge bases: a survey","abstract":" Reasoning is fundamental to human intelligence, and critical for\nproblem-solving, decision-making, and critical thinking. Reasoning refers to\ndrawing new conclusions based on existing knowledge, which can support various\napplications like clinical diagnosis, basic education, and financial analysis.\nThough a good number of surveys have been proposed for reviewing\nreasoning-related methods, none of them has systematically investigated these\nmethods from the viewpoint of their dependent knowledge base. Both the\nscenarios to which the knowledge bases are applied and their storage formats\nare significantly different. Hence, investigating reasoning methods from the\nknowledge base perspective helps us better understand the challenges and future\ndirections. To fill this gap, this paper first classifies the knowledge base\ninto symbolic and parametric ones. The former explicitly stores information in\nhuman-readable symbols, and the latter implicitly encodes knowledge within\nparameters. Then, we provide a comprehensive overview of reasoning methods\nusing symbolic knowledge bases, parametric knowledge bases, and both of them.\nFinally, we identify the future direction toward enhancing reasoning\ncapabilities to bridge the gap between human and machine intelligence.\n","arxiv_id":"http://arxiv.org/abs/2501.01030v1","authors":["Mayi Xu","Yunfeng Ning","Yongqi Li","Jianhao Chen","Jintao Wen","Yao Xiao","Shen Zhou","Birong Pan","Zepeng Bao","Xin Miao","Hankun Kang","Ke Sun","Tieyun Qian"]},{"title":"MSWA: Refining Local Attention with Multi-ScaleWindow Attention","abstract":" Transformer-based LLMs have achieved exceptional performance across a wide\nrange of NLP tasks. However, the standard self-attention mechanism suffers from\nquadratic time complexity and linearly increased cache size. Sliding window\nattention (SWA) solves this problem by restricting the attention range to a\nfixed-size local context window. Nevertheless, SWA employs a uniform window\nsize for each head in each layer, making it inefficient in capturing context of\nvarying scales. To mitigate this limitation, we propose Multi-Scale Window\nAttention (MSWA) which applies diverse window sizes across heads and layers in\nthe Transformer. It not only allows for different window sizes among heads\nwithin the same layer but also progressively increases window size allocation\nfrom shallow to deep layers, thus enabling the model to capture contextual\ninformation with different lengths and distances. Experimental results on\nlanguage modeling and common-sense reasoning tasks substantiate that MSWA\noutperforms traditional local attention in both effectiveness and efficiency.\n","arxiv_id":"http://arxiv.org/abs/2501.01039v1","authors":["Yixing Xu","Shivank Nag","Dong Li","Lu Tian","Emad Barsoum"]},{"title":"Risks of Cultural Erasure in Large Language Models","abstract":" Large language models are increasingly being integrated into applications\nthat shape the production and discovery of societal knowledge such as search,\nonline education, and travel planning. As a result, language models will shape\nhow people learn about, perceive and interact with global cultures making it\nimportant to consider whose knowledge systems and perspectives are represented\nin models. Recognizing this importance, increasingly work in Machine Learning\nand NLP has focused on evaluating gaps in global cultural representational\ndistribution within outputs. However, more work is needed on developing\nbenchmarks for cross-cultural impacts of language models that stem from a\nnuanced sociologically-aware conceptualization of cultural impact or harm. We\njoin this line of work arguing for the need of metricizable evaluations of\nlanguage technologies that interrogate and account for historical power\ninequities and differential impacts of representation on global cultures,\nparticularly for cultures already under-represented in the digital corpora. We\nlook at two concepts of erasure: omission: where cultures are not represented\nat all and simplification i.e. when cultural complexity is erased by presenting\none-dimensional views of a rich culture. The former focuses on whether\nsomething is represented, and the latter on how it is represented. We focus our\nanalysis on two task contexts with the potential to influence global cultural\nproduction. First, we probe representations that a language model produces\nabout different places around the world when asked to describe these contexts.\nSecond, we analyze the cultures represented in the travel recommendations\nproduced by a set of language model applications. Our study shows ways in which\nthe NLP community and application developers can begin to operationalize\ncomplex socio-cultural considerations into standard evaluations and benchmarks.\n","arxiv_id":"http://arxiv.org/abs/2501.01056v1","authors":["Rida Qadri","Aida M. Davani","Kevin Robinson","Vinodkumar Prabhakaran"]},{"title":"Graph Generative Pre-trained Transformer","abstract":" Graph generation is a critical task in numerous domains, including molecular\ndesign and social network analysis, due to its ability to model complex\nrelationships and structured data. While most modern graph generative models\nutilize adjacency matrix representations, this work revisits an alternative\napproach that represents graphs as sequences of node set and edge set. We\nadvocate for this approach due to its efficient encoding of graphs and propose\na novel representation. Based on this representation, we introduce the Graph\nGenerative Pre-trained Transformer (G2PT), an auto-regressive model that learns\ngraph structures via next-token prediction. To further exploit G2PT's\ncapabilities as a general-purpose foundation model, we explore fine-tuning\nstrategies for two downstream applications: goal-oriented generation and graph\nproperty prediction. We conduct extensive experiments across multiple datasets.\nResults indicate that G2PT achieves superior generative performance on both\ngeneric graph and molecule datasets. Furthermore, G2PT exhibits strong\nadaptability and versatility in downstream tasks from molecular design to\nproperty prediction.\n","arxiv_id":"http://arxiv.org/abs/2501.01073v1","authors":["Xiaohui Chen","Yinkai Wang","Jiaxing He","Yuanqi Du","Soha Hassoun","Xiaolin Xu","Li-Ping Liu"]},{"title":"BatStyler: Advancing Multi-category Style Generation for Source-free\n Domain Generalization","abstract":" Source-Free Domain Generalization (SFDG) aims to develop a model that\nperforms on unseen domains without relying on any source domains. However, the\nimplementation remains constrained due to the unavailability of training data.\nResearch on SFDG focus on knowledge transfer of multi-modal models and style\nsynthesis based on joint space of multiple modalities, thus eliminating the\ndependency on source domain images. However, existing works primarily work for\nmulti-domain and less-category configuration, but performance on multi-domain\nand multi-category configuration is relatively poor. In addition, the\nefficiency of style synthesis also deteriorates in multi-category scenarios.\nHow to efficiently synthesize sufficiently diverse data and apply it to\nmulti-category configuration is a direction with greater practical value. In\nthis paper, we propose a method called BatStyler, which is utilized to improve\nthe capability of style synthesis in multi-category scenarios. BatStyler\nconsists of two modules: Coarse Semantic Generation and Uniform Style\nGeneration modules. The Coarse Semantic Generation module extracts\ncoarse-grained semantics to prevent the compression of space for style\ndiversity learning in multi-category configuration, while the Uniform Style\nGeneration module provides a template of styles that are uniformly distributed\nin space and implements parallel training. Extensive experiments demonstrate\nthat our method exhibits comparable performance on less-category datasets,\nwhile surpassing state-of-the-art methods on multi-category datasets.\n","arxiv_id":"http://arxiv.org/abs/2501.01109v1","authors":["Xiusheng Xu","Lei Qi","Jingyang Zhou","Xin Geng"]},{"title":"MalCL: Leveraging GAN-Based Generative Replay to Combat Catastrophic\n Forgetting in Malware Classification","abstract":" Continual Learning (CL) for malware classification tackles the rapidly\nevolving nature of malware threats and the frequent emergence of new types.\nGenerative Replay (GR)-based CL systems utilize a generative model to produce\nsynthetic versions of past data, which are then combined with new data to\nretrain the primary model. Traditional machine learning techniques in this\ndomain often struggle with catastrophic forgetting, where a model's performance\non old data degrades over time.\n In this paper, we introduce a GR-based CL system that employs Generative\nAdversarial Networks (GANs) with feature matching loss to generate high-quality\nmalware samples. Additionally, we implement innovative selection schemes for\nreplay samples based on the model's hidden representations.\n Our comprehensive evaluation across Windows and Android malware datasets in a\nclass-incremental learning scenario -- where new classes are introduced\ncontinuously over multiple tasks -- demonstrates substantial performance\nimprovements over previous methods. For example, our system achieves an average\naccuracy of 55% on Windows malware samples, significantly outperforming other\nGR-based models by 28%. This study provides practical insights for advancing\nGR-based malware classification systems. The implementation is available at\n\\url {https://github.com/MalwareReplayGAN/MalCL}\\footnote{The code will be made\npublic upon the presentation of the paper}.\n","arxiv_id":"http://arxiv.org/abs/2501.01110v1","authors":["Jimin Park","AHyun Ji","Minji Park","Mohammad Saidur Rahman","Se Eun Oh"]},{"title":"Pruning-based Data Selection and Network Fusion for Efficient Deep\n Learning","abstract":" Efficient data selection is essential for improving the training efficiency\nof deep neural networks and reducing the associated annotation costs. However,\ntraditional methods tend to be computationally expensive, limiting their\nscalability and real-world applicability. We introduce PruneFuse, a novel\nmethod that combines pruning and network fusion to enhance data selection and\naccelerate network training. In PruneFuse, the original dense network is pruned\nto generate a smaller surrogate model that efficiently selects the most\ninformative samples from the dataset. Once this iterative data selection\nselects sufficient samples, the insights learned from the pruned model are\nseamlessly integrated with the dense model through network fusion, providing an\noptimized initialization that accelerates training. Extensive experimentation\non various datasets demonstrates that PruneFuse significantly reduces\ncomputational costs for data selection, achieves better performance than\nbaselines, and accelerates the overall training process.\n","arxiv_id":"http://arxiv.org/abs/2501.01118v1","authors":["Humaira Kousar","Hasnain Irshad Bhatti","Jaekyun Moon"]},{"title":"Retrieval-Augmented Dynamic Prompt Tuning for Incomplete Multimodal\n Learning","abstract":" Multimodal learning with incomplete modality is practical and challenging.\nRecently, researchers have focused on enhancing the robustness of pre-trained\nMultiModal Transformers (MMTs) under missing modality conditions by applying\nlearnable prompts. However, these prompt-based methods face several\nlimitations: (1) incomplete modalities provide restricted modal cues for\ntask-specific inference, (2) dummy imputation for missing content causes\ninformation loss and introduces noise, and (3) static prompts are\ninstance-agnostic, offering limited knowledge for instances with various\nmissing conditions. To address these issues, we propose RAGPT, a novel\nRetrieval-AuGmented dynamic Prompt Tuning framework. RAGPT comprises three\nmodules: (I) the multi-channel retriever, which identifies similar instances\nthrough a within-modality retrieval strategy, (II) the missing modality\ngenerator, which recovers missing information using retrieved contexts, and\n(III) the context-aware prompter, which captures contextual knowledge from\nrelevant instances and generates dynamic prompts to largely enhance the MMT's\nrobustness. Extensive experiments conducted on three real-world datasets show\nthat RAGPT consistently outperforms all competitive baselines in handling\nincomplete modality problems. The code of our work and prompt-based baselines\nis available at https://github.com/Jian-Lang/RAGPT.\n","arxiv_id":"http://arxiv.org/abs/2501.01120v1","authors":["Jian Lang","Zhangtao Cheng","Ting Zhong","Fan Zhou"]},{"title":"Deep Learning in Palmprint Recognition-A Comprehensive Survey","abstract":" Palmprint recognition has emerged as a prominent biometric technology, widely\napplied in diverse scenarios. Traditional handcrafted methods for palmprint\nrecognition often fall short in representation capability, as they heavily\ndepend on researchers' prior knowledge. Deep learning (DL) has been introduced\nto address this limitation, leveraging its remarkable successes across various\ndomains. While existing surveys focus narrowly on specific tasks within\npalmprint recognition-often grounded in traditional methodologies-there remains\na significant gap in comprehensive research exploring DL-based approaches\nacross all facets of palmprint recognition. This paper bridges that gap by\nthoroughly reviewing recent advancements in DL-powered palmprint recognition.\nThe paper systematically examines progress across key tasks, including\nregion-of-interest segmentation, feature extraction, and\nsecurity/privacy-oriented challenges. Beyond highlighting these advancements,\nthe paper identifies current challenges and uncovers promising opportunities\nfor future research. By consolidating state-of-the-art progress, this review\nserves as a valuable resource for researchers, enabling them to stay abreast of\ncutting-edge technologies and drive innovation in palmprint recognition.\n","arxiv_id":"http://arxiv.org/abs/2501.01166v1","authors":["Chengrui Gao","Ziyuan Yang","Wei Jia","Lu Leng","Bob Zhang","Andrew Beng Jin Teoh"]},{"title":"Blind Men and the Elephant: Diverse Perspectives on Gender Stereotypes\n in Benchmark Datasets","abstract":" The multifaceted challenge of accurately measuring gender stereotypical bias\nin language models is akin to discerning different segments of a broader,\nunseen entity. This short paper primarily focuses on intrinsic bias mitigation\nand measurement strategies for language models, building on prior research that\ndemonstrates a lack of correlation between intrinsic and extrinsic approaches.\nWe delve deeper into intrinsic measurements, identifying inconsistencies and\nsuggesting that these benchmarks may reflect different facets of gender\nstereotype. Our methodology involves analyzing data distributions across\ndatasets and integrating gender stereotype components informed by social\npsychology. By adjusting the distribution of two datasets, we achieve a better\nalignment of outcomes. Our findings underscore the complexity of gender\nstereotyping in language models and point to new directions for developing more\nrefined techniques to detect and reduce bias.\n","arxiv_id":"http://arxiv.org/abs/2501.01168v1","authors":["Mahdi Zakizadeh","Mohammad Taher Pilehvar"]},{"title":"L3D-Pose: Lifting Pose for 3D Avatars from a Single Camera in the Wild","abstract":" While 2D pose estimation has advanced our ability to interpret body movements\nin animals and primates, it is limited by the lack of depth information,\nconstraining its application range. 3D pose estimation provides a more\ncomprehensive solution by incorporating spatial depth, yet creating extensive\n3D pose datasets for animals is challenging due to their dynamic and\nunpredictable behaviours in natural settings. To address this, we propose a\nhybrid approach that utilizes rigged avatars and the pipeline to generate\nsynthetic datasets to acquire the necessary 3D annotations for training. Our\nmethod introduces a simple attention-based MLP network for converting 2D poses\nto 3D, designed to be independent of the input image to ensure scalability for\nposes in natural environments. Additionally, we identify that existing\nanatomical keypoint detectors are insufficient for accurate pose retargeting\nonto arbitrary avatars. To overcome this, we present a lookup table based on a\ndeep pose estimation method using a synthetic collection of diverse actions\nrigged avatars perform. Our experiments demonstrate the effectiveness and\nefficiency of this lookup table-based retargeting approach. Overall, we propose\na comprehensive framework with systematically synthesized datasets for lifting\nposes from 2D to 3D and then utilize this to re-target motion from wild\nsettings onto arbitrary avatars.\n","arxiv_id":"http://arxiv.org/abs/2501.01174v1","authors":["Soumyaratna Debnath","Harish Katti","Shashikant Verma","Shanmuganathan Raman"]},{"title":"Data Augmentation Techniques for Chinese Disease Name Normalization","abstract":" Disease name normalization is an important task in the medical domain. It\nclassifies disease names written in various formats into standardized names,\nserving as a fundamental component in smart healthcare systems for various\ndisease-related functions. Nevertheless, the most significant obstacle to\nexisting disease name normalization systems is the severe shortage of training\ndata. Consequently, we present a novel data augmentation approach that includes\na series of data augmentation techniques and some supporting modules to help\nmitigate the problem. Through extensive experimentation, we illustrate that our\nproposed approach exhibits significant performance improvements across various\nbaseline models and training objectives, particularly in scenarios with limited\ntraining data\n","arxiv_id":"http://arxiv.org/abs/2501.01195v1","authors":["Wenqian Cui","Xiangling Fu","Shaohui Liu","Mingjun Gu","Xien Liu","Ji Wu","Irwin King"]},{"title":"A redescription mining framework for post-hoc explaining and relating\n deep learning models","abstract":" Deep learning models (DLMs) achieve increasingly high performance both on\nstructured and unstructured data. They significantly extended applicability of\nmachine learning to various domains. Their success in making predictions,\ndetecting patterns and generating new data made significant impact on science\nand industry. Despite these accomplishments, DLMs are difficult to explain\nbecause of their enormous size. In this work, we propose a novel framework for\npost-hoc explaining and relating DLMs using redescriptions. The framework\nallows cohort analysis of arbitrary DLMs by identifying statistically\nsignificant redescriptions of neuron activations. It allows coupling neurons to\na set of target labels or sets of descriptive attributes, relating layers\nwithin a single DLM or associating different DLMs. The proposed framework is\nindependent of the artificial neural network architecture and can work with\nmore complex target labels (e.g. multi-label or multi-target scenario).\nAdditionally, it can emulate both pedagogical and decompositional approach to\nrule extraction. The aforementioned properties of the proposed framework can\nincrease explainability and interpretability of arbitrary DLMs by providing\ndifferent information compared to existing explainable-AI approaches.\n","arxiv_id":"http://arxiv.org/abs/2501.01209v1","authors":["Matej Mihelčić","Ivan Grubišić","Miha Keber"]},{"title":"An Efficient Attention Mechanism for Sequential Recommendation Tasks:\n HydraRec","abstract":" Transformer based models are increasingly being used in various domains\nincluding recommender systems (RS). Pretrained transformer models such as BERT\nhave shown good performance at language modelling. With the greater ability to\nmodel sequential tasks, variants of Encoder-only models (like BERT4Rec, SASRec\netc.) have found success in sequential RS problems. Computing dot-product\nattention in traditional transformer models has quadratic complexity in\nsequence length. This is a bigger problem with RS because unlike language\nmodels, new items are added to the catalogue every day. User buying history is\na dynamic sequence which depends on multiple factors. Recently, various linear\nattention models have tried to solve this problem by making the model linear in\nsequence length (token dimensions). Hydra attention is one such linear\ncomplexity model proposed for vision transformers which reduces the complexity\nof attention for both the number of tokens as well as model embedding\ndimensions. Building on the idea of Hydra attention, we introduce an efficient\nTransformer based Sequential RS (HydraRec) which significantly improves\ntheoretical complexity of computing attention for longer sequences and bigger\ndatasets while preserving the temporal context. Extensive experiments are\nconducted to evaluate other linear transformer-based RS models and compared\nwith HydraRec across various evaluation metrics. HydraRec outperforms other\nlinear attention-based models as well as dot-product based attention models\nwhen used with causal masking for sequential recommendation next item\nprediction tasks. For bi-directional models its performance is comparable to\nthe BERT4Rec model with an improvement in running time.\n","arxiv_id":"http://arxiv.org/abs/2501.01242v1","authors":["Uzma Mushtaque"]},{"title":"Stealthy Backdoor Attack to Real-world Models in Android Apps","abstract":" Powered by their superior performance, deep neural networks (DNNs) have found\nwidespread applications across various domains. Many deep learning (DL) models\nare now embedded in mobile apps, making them more accessible to end users\nthrough on-device DL. However, deploying on-device DL to users' smartphones\nsimultaneously introduces several security threats. One primary threat is\nbackdoor attacks. Extensive research has explored backdoor attacks for several\nyears and has proposed numerous attack approaches. However, few studies have\ninvestigated backdoor attacks on DL models deployed in the real world, or they\nhave shown obvious deficiencies in effectiveness and stealthiness. In this\nwork, we explore more effective and stealthy backdoor attacks on real-world DL\nmodels extracted from mobile apps. Our main justification is that imperceptible\nand sample-specific backdoor triggers generated by DNN-based steganography can\nenhance the efficacy of backdoor attacks on real-world models. We first confirm\nthe effectiveness of steganography-based backdoor attacks on four\nstate-of-the-art DNN models. Subsequently, we systematically evaluate and\nanalyze the stealthiness of the attacks to ensure they are difficult to\nperceive. Finally, we implement the backdoor attacks on real-world models and\ncompare our approach with three baseline methods. We collect 38,387 mobile\napps, extract 89 DL models from them, and analyze these models to obtain the\nprerequisite model information for the attacks. After identifying the target\nmodels, our approach achieves an average of 12.50% higher attack success rate\nthan DeepPayload while better maintaining the normal performance of the models.\nExtensive experimental results demonstrate that our method enables more\neffective, robust, and stealthy backdoor attacks on real-world models.\n","arxiv_id":"http://arxiv.org/abs/2501.01263v1","authors":["Jiali Wei","Ming Fan","Xicheng Zhang","Wenjing Jiao","Haijun Wang","Ting Liu"]},{"title":"PIMAEX: Multi-Agent Exploration through Peer Incentivization","abstract":" While exploration in single-agent reinforcement learning has been studied\nextensively in recent years, considerably less work has focused on its\ncounterpart in multi-agent reinforcement learning. To address this issue, this\nwork proposes a peer-incentivized reward function inspired by previous research\non intrinsic curiosity and influence-based rewards. The \\textit{PIMAEX} reward,\nshort for Peer-Incentivized Multi-Agent Exploration, aims to improve\nexploration in the multi-agent setting by encouraging agents to exert influence\nover each other to increase the likelihood of encountering novel states. We\nevaluate the \\textit{PIMAEX} reward in conjunction with\n\\textit{PIMAEX-Communication}, a multi-agent training algorithm that employs a\ncommunication channel for agents to influence one another. The evaluation is\nconducted in the \\textit{Consume/Explore} environment, a partially observable\nenvironment with deceptive rewards, specifically designed to challenge the\nexploration vs.\\ exploitation dilemma and the credit-assignment problem. The\nresults empirically demonstrate that agents using the \\textit{PIMAEX} reward\nwith \\textit{PIMAEX-Communication} outperform those that do not.\n","arxiv_id":"http://arxiv.org/abs/2501.01266v1","authors":["Michael Kölle","Johannes Tochtermann","Julian Schönberger","Gerhard Stenzel","Philipp Altmann","Claudia Linnhoff-Popien"]},{"title":"NeutraSum: A Language Model can help a Balanced Media Diet by\n Neutralizing News Summaries","abstract":" Media bias in news articles arises from the political polarisation of media\noutlets, which can reinforce societal stereotypes and beliefs. Reporting on the\nsame event often varies significantly between outlets, reflecting their\npolitical leanings through polarised language and focus. Although previous\nstudies have attempted to generate bias-free summaries from multiperspective\nnews articles, they have not effectively addressed the challenge of mitigating\ninherent media bias. To address this gap, we propose \\textbf{NeutraSum}, a\nnovel framework that integrates two neutrality losses to adjust the semantic\nspace of generated summaries, thus minimising media bias. These losses,\ndesigned to balance the semantic distances across polarised inputs and ensure\nalignment with expert-written summaries, guide the generation of neutral and\nfactually rich summaries. To evaluate media bias, we employ the political\ncompass test, which maps political leanings based on economic and social\ndimensions. Experimental results on the Allsides dataset demonstrate that\nNeutraSum not only improves summarisation performance but also achieves\nsignificant reductions in media bias, offering a promising approach for neutral\nnews summarisation.\n","arxiv_id":"http://arxiv.org/abs/2501.01284v1","authors":["Xi Luo","Junjie Liu","Sirong Wu","Yuhui Deng"]},{"title":"Citations and Trust in LLM Generated Responses","abstract":" Question answering systems are rapidly advancing, but their opaque nature may\nimpact user trust. We explored trust through an anti-monitoring framework,\nwhere trust is predicted to be correlated with presence of citations and\ninversely related to checking citations. We tested this hypothesis with a live\nquestion-answering experiment that presented text responses generated using a\ncommercial Chatbot along with varying citations (zero, one, or five), both\nrelevant and random, and recorded if participants checked the citations and\ntheir self-reported trust in the generated responses. We found a significant\nincrease in trust when citations were present, a result that held true even\nwhen the citations were random; we also found a significant decrease in trust\nwhen participants checked the citations. These results highlight the importance\nof citations in enhancing trust in AI-generated content.\n","arxiv_id":"http://arxiv.org/abs/2501.01303v1","authors":["Yifan Ding","Matthew Facciani","Amrit Poudel","Ellen Joyce","Salvador Aguinaga","Balaji Veeramani","Sanmitra Bhattacharya","Tim Weninger"]},{"title":"Multi-Head Explainer: A General Framework to Improve Explainability in\n CNNs and Transformers","abstract":" In this study, we introduce the Multi-Head Explainer (MHEX), a versatile and\nmodular framework that enhances both the explainability and accuracy of\nConvolutional Neural Networks (CNNs) and Transformer-based models. MHEX\nconsists of three core components: an Attention Gate that dynamically\nhighlights task-relevant features, Deep Supervision that guides early layers to\ncapture fine-grained details pertinent to the target class, and an Equivalent\nMatrix that unifies refined local and global representations to generate\ncomprehensive saliency maps. Our approach demonstrates superior compatibility,\nenabling effortless integration into existing residual networks like ResNet and\nTransformer architectures such as BERT with minimal modifications. Extensive\nexperiments on benchmark datasets in medical imaging and text classification\nshow that MHEX not only improves classification accuracy but also produces\nhighly interpretable and detailed saliency scores.\n","arxiv_id":"http://arxiv.org/abs/2501.01311v2","authors":["Bohang Sun","Pietro Liò"]},{"title":"Understanding Difficult-to-learn Examples in Contrastive Learning: A\n Theoretical Framework for Spectral Contrastive Learning","abstract":" Unsupervised contrastive learning has shown significant performance\nimprovements in recent years, often approaching or even rivaling supervised\nlearning in various tasks. However, its learning mechanism is fundamentally\ndifferent from that of supervised learning. Previous works have shown that\ndifficult-to-learn examples (well-recognized in supervised learning as examples\naround the decision boundary), which are essential in supervised learning,\ncontribute minimally in unsupervised settings. In this paper, perhaps\nsurprisingly, we find that the direct removal of difficult-to-learn examples,\nalthough reduces the sample size, can boost the downstream classification\nperformance of contrastive learning. To uncover the reasons behind this, we\ndevelop a theoretical framework modeling the similarity between different pairs\nof samples. Guided by this theoretical framework, we conduct a thorough\ntheoretical analysis revealing that the presence of difficult-to-learn examples\nnegatively affects the generalization of contrastive learning. Furthermore, we\ndemonstrate that the removal of these examples, and techniques such as margin\ntuning and temperature scaling can enhance its generalization bounds, thereby\nimproving performance. Empirically, we propose a simple and efficient mechanism\nfor selecting difficult-to-learn examples and validate the effectiveness of the\naforementioned methods, which substantiates the reliability of our proposed\ntheoretical framework.\n","arxiv_id":"http://arxiv.org/abs/2501.01317v1","authors":["Yi-Ge Zhang","Jingyi Cui","Qiran Li","Yisen Wang"]},{"title":"DeepFilter: An Instrumental Baseline for Accurate and Efficient Process\n Monitoring","abstract":" Effective process monitoring is increasingly vital in industrial automation\nfor ensuring operational safety, necessitating both high accuracy and\nefficiency. Although Transformers have demonstrated success in various fields,\ntheir canonical form based on the self-attention mechanism is inadequate for\nprocess monitoring due to two primary limitations: (1) the step-wise\ncorrelations captured by self-attention mechanism are difficult to capture\ndiscriminative patterns in monitoring logs due to the lacking semantics of each\nstep, thus compromising accuracy; (2) the quadratic computational complexity of\nself-attention hampers efficiency. To address these issues, we propose\nDeepFilter, a Transformer-style framework for process monitoring. The core\ninnovation is an efficient filtering layer that excel capturing long-term and\nperiodic patterns with reduced complexity. Equipping with the global filtering\nlayer, DeepFilter enhances both accuracy and efficiency, meeting the stringent\ndemands of process monitoring. Experimental results on real-world process\nmonitoring datasets validate DeepFilter's superiority in terms of accuracy and\nefficiency compared to existing state-of-the-art models.\n","arxiv_id":"http://arxiv.org/abs/2501.01342v1","authors":["Hao Wang","Zhichao Chen","Licheng Pan","Xiaoyu Jiang","Yichen Song","Qunshan He","Xinggao Liu"]},{"title":"A Unified Hyperparameter Optimization Pipeline for Transformer-Based\n Time Series Forecasting Models","abstract":" Transformer-based models for time series forecasting (TSF) have attracted\nsignificant attention in recent years due to their effectiveness and\nversatility. However, these models often require extensive hyperparameter\noptimization (HPO) to achieve the best possible performance, and a unified\npipeline for HPO in transformer-based TSF remains lacking. In this paper, we\npresent one such pipeline and conduct extensive experiments on several\nstate-of-the-art (SOTA) transformer-based TSF models. These experiments are\nconducted on standard benchmark datasets to evaluate and compare the\nperformance of different models, generating practical insights and examples.\nOur pipeline is generalizable beyond transformer-based architectures and can be\napplied to other SOTA models, such as Mamba and TimeMixer, as demonstrated in\nour experiments. The goal of this work is to provide valuable guidance to both\nindustry practitioners and academic researchers in efficiently identifying\noptimal hyperparameters suited to their specific domain applications. The code\nand complete experimental results are available on GitHub.\n","arxiv_id":"http://arxiv.org/abs/2501.01394v1","authors":["Jingjing Xu","Caesar Wu","Yuan-Fang Li","Grégoire Danoy","Pascal Bouvry"]},{"title":"On Unifying Video Generation and Camera Pose Estimation","abstract":" Inspired by the emergent 3D capabilities in image generators, we explore\nwhether video generators similarly exhibit 3D awareness. Using\nstructure-from-motion (SfM) as a benchmark for 3D tasks, we investigate if\nintermediate features from OpenSora, a video generation model, can support\ncamera pose estimation. We first examine native 3D awareness in video\ngeneration features by routing raw intermediate outputs to SfM-prediction\nmodules like DUSt3R. Then, we explore the impact of fine-tuning on camera pose\nestimation to enhance 3D awareness. Results indicate that while video generator\nfeatures have limited inherent 3D awareness, task-specific supervision\nsignificantly boosts their accuracy for camera pose estimation, resulting in\ncompetitive performance. The proposed unified model, named JOG3R, produces\ncamera pose estimates with competitive quality without degrading video\ngeneration quality.\n","arxiv_id":"http://arxiv.org/abs/2501.01409v1","authors":["Chun-Hao Paul Huang","Jae Shin Yoon","Hyeonho Jeong","Niloy Mitra","Duygu Ceylan"]},{"title":"Balance-aware Sequence Sampling Makes Multi-modal Learning Better","abstract":" To address the modality imbalance caused by data heterogeneity, existing\nmulti-modal learning (MML) approaches primarily focus on balancing this\ndifference from the perspective of optimization objectives. However, almost all\nexisting methods ignore the impact of sample sequences, i.e., an inappropriate\ntraining order tends to trigger learning bias in the model, further\nexacerbating modality imbalance. In this paper, we propose Balance-aware\nSequence Sampling (BSS) to enhance the robustness of MML. Specifically, we\nfirst define a multi-perspective measurer to evaluate the balance degree of\neach sample. Via the evaluation, we employ a heuristic scheduler based on\ncurriculum learning (CL) that incrementally provides training subsets,\nprogressing from balanced to imbalanced samples to rebalance MML. Moreover,\nconsidering that sample balance may evolve as the model capability increases,\nwe propose a learning-based probabilistic sampling method to dynamically update\nthe training sequences at the epoch level, further improving MML performance.\nExtensive experiments on widely used datasets demonstrate the superiority of\nour method compared with state-of-the-art (SOTA) MML approaches.\n","arxiv_id":"http://arxiv.org/abs/2501.01470v1","authors":["Zhi-Hao Guan"]},{"title":"Augmented Contrastive Clustering with Uncertainty-Aware Prototyping for\n Time Series Test Time Adaptation","abstract":" Test-time adaptation aims to adapt pre-trained deep neural networks using\nsolely online unlabelled test data during inference. Although TTA has shown\npromise in visual applications, its potential in time series contexts remains\nlargely unexplored. Existing TTA methods, originally designed for visual tasks,\nmay not effectively handle the complex temporal dynamics of real-world time\nseries data, resulting in suboptimal adaptation performance. To address this\ngap, we propose Augmented Contrastive Clustering with Uncertainty-aware\nPrototyping (ACCUP), a straightforward yet effective TTA method for time series\ndata. Initially, our approach employs augmentation ensemble on the time series\ndata to capture diverse temporal information and variations, incorporating\nuncertainty-aware prototypes to distill essential characteristics.\nAdditionally, we introduce an entropy comparison scheme to selectively acquire\nmore confident predictions, enhancing the reliability of pseudo labels.\nFurthermore, we utilize augmented contrastive clustering to enhance feature\ndiscriminability and mitigate error accumulation from noisy pseudo labels,\npromoting cohesive clustering within the same class while facilitating clear\nseparation between different classes. Extensive experiments conducted on three\nreal-world time series datasets and an additional visual dataset demonstrate\nthe effectiveness and generalization potential of the proposed method,\nadvancing the underexplored realm of TTA for time series data.\n","arxiv_id":"http://arxiv.org/abs/2501.01472v1","authors":["Peiliang Gong","Mohamed Ragab","Min Wu","Zhenghua Chen","Yongyi Su","Xiaoli Li","Daoqiang Zhang"]},{"title":"Unraveling Indirect In-Context Learning Using Influence Functions","abstract":" This work introduces a novel paradigm for generalized In-Context Learning\n(ICL), termed Indirect In-Context Learning. In Indirect ICL, we explore\ndemonstration selection strategies tailored for two distinct real-world\nscenarios: Mixture of Tasks and Noisy Demonstrations. We systematically\nevaluate the effectiveness of Influence Functions (IFs) as a selection tool for\nthese settings, highlighting the potential for IFs to better capture the\ninformativeness of examples within the demonstration pool. For the Mixture of\nTasks setting, demonstrations are drawn from 28 diverse tasks, including MMLU,\nBigBench, StrategyQA, and CommonsenseQA. We demonstrate that combining\nBertScore-Recall (BSR) with an IF surrogate model can significantly improve\nperformance, leading to average absolute accuracy gains of 0.37\\% and 1.45\\%\nfor 3-shot and 5-shot setups when compared to traditional ICL metrics. In the\nNoisy Demonstrations setting, we examine scenarios where demonstrations might\nbe mislabeled. Our experiments show that reweighting traditional ICL selectors\n(BSR and Cosine Similarity) with IF-based selectors boosts accuracy by an\naverage of 2.90\\% for Cosine Similarity and 2.94\\% for BSR on noisy GLUE\nbenchmarks. In sum, we propose a robust framework for demonstration selection\nthat generalizes beyond traditional ICL, offering valuable insights into the\nrole of IFs for Indirect ICL.\n","arxiv_id":"http://arxiv.org/abs/2501.01473v1","authors":["Hadi Askari","Shivanshu Gupta","Terry Tong","Fei Wang","Anshuman Chhabra","Muhao Chen"]},{"title":"A Survey of Deep Learning Methods in Protein Bioinformatics and its\n Impact on Protein Design","abstract":" Proteins are sequences of amino acids that serve as the basic building blocks\nof living organisms. Despite rapidly growing databases documenting structural\nand functional information for various protein sequences, our understanding of\nproteins remains limited because of the large possible sequence space and the\ncomplex inter- and intra-molecular forces. Deep learning, which is\ncharacterized by its ability to learn relevant features directly from large\ndatasets, has demonstrated remarkable performance in fields such as computer\nvision and natural language processing. It has also been increasingly applied\nin recent years to the data-rich domain of protein sequences with great\nsuccess, most notably with Alphafold2's breakout performance in the protein\nstructure prediction. The performance improvements achieved by deep learning\nunlocks new possibilities in the field of protein bioinformatics, including\nprotein design, one of the most difficult but useful tasks. In this paper, we\nbroadly categorize problems in protein bioinformatics into three main\ncategories: 1) structural prediction, 2) functional prediction, and 3) protein\ndesign, and review the progress achieved from using deep learning methodologies\nin each of them. We expand on the main challenges of the protein design problem\nand highlight how advances in structural and functional prediction have\ndirectly contributed to design tasks. Finally, we conclude by identifying\nimportant topics and future research directions.\n","arxiv_id":"http://arxiv.org/abs/2501.01477v1","authors":["Weihang Dai"]},{"title":"Drift2Matrix: Kernel-Induced Self Representation for Concept Drift\n Adaptation in Co-evolving Time Series","abstract":" In the realm of time series analysis, tackling the phenomenon of concept\ndrift poses a significant challenge. Concept drift -- characterized by the\nevolving statistical properties of time series data, affects the reliability\nand accuracy of conventional analysis models. This is particularly evident in\nco-evolving scenarios where interactions among variables are crucial. This\npaper presents Drift2Matrix, a novel framework that leverages kernel-induced\nself-representation for adaptive responses to concept drift in time series.\nDrift2Matrix employs a kernel-based learning mechanism to generate a\nrepresentation matrix, encapsulating the inherent dynamics of co-evolving time\nseries. This matrix serves as a key tool for identification and adaptation to\nconcept drift by observing its temporal variations. Furthermore, Drift2Matrix\neffectively identifies prevailing patterns and offers insights into emerging\ntrends through pattern evolution analysis. Our empirical evaluation of\nDrift2Matrix across various datasets demonstrates its effectiveness in handling\nthe complexities of concept drift. This approach introduces a novel perspective\nin the theoretical domain of co-evolving time series analysis, enhancing\nadaptability and accuracy in the face of dynamic data environments.\n","arxiv_id":"http://arxiv.org/abs/2501.01480v2","authors":["Kunpeng Xu","Lifei Chen","Shengrui Wang"]},{"title":"BoxingGym: Benchmarking Progress in Automated Experimental Design and\n Model Discovery","abstract":" Understanding the world and explaining it with scientific theories is a\ncentral aspiration of artificial intelligence research. Proposing theories,\ndesigning experiments to test them, and then revising them based on data are\nfundamental to scientific discovery. Despite the significant promise of\nLLM-based scientific agents, no benchmarks systematically test LLM's ability to\npropose scientific models, collect experimental data, and revise them in light\nof new data. We introduce BoxingGym, a benchmark with 10 environments for\nsystematically evaluating both experimental design (e.g. collecting data to\ntest a scientific theory) and model discovery (e.g. proposing and revising\nscientific theories). To enable tractable and quantitative evaluation, we\nimplement each environment as a generative probabilistic model with which a\nscientific agent can run interactive experiments. These probabilistic models\nare drawn from various real-world scientific domains ranging from psychology to\necology. To quantitatively evaluate a scientific agent's ability to collect\ninformative experimental data, we compute the expected information gain (EIG),\nan information-theoretic quantity which measures how much an experiment reduces\nuncertainty about the parameters of a generative model. A good scientific\ntheory is a concise and predictive explanation. Therefore, to quantitatively\nevaluate model discovery, we ask a scientific agent to explain their model and\nthen assess whether this explanation enables another scientific agent to make\nreliable predictions about this environment. In addition to this\nexplanation-based evaluation, we compute standard model evaluation metrics such\nas prediction errors. We find that current LLMs, such as GPT-4o, struggle with\nboth experimental design and model discovery. We find that augmenting the\nLLM-based agent with an explicit statistical model does not reliably improve\nthese results.\n","arxiv_id":"http://arxiv.org/abs/2501.01540v1","authors":["Kanishk Gandhi","Michael Y. Li","Lyle Goodyear","Louise Li","Aditi Bhaskar","Mohammed Zaman","Noah D. Goodman"]},{"title":"Constructing and explaining machine learning models for chemistry:\n example of the exploration and design of boron-based Lewis acids","abstract":" The integration of machine learning (ML) into chemistry offers transformative\npotential in the design of molecules with targeted properties. However, the\nfocus has often been on creating highly efficient predictive models, sometimes\nat the expense of interpretability. In this study, we leverage explainable AI\ntechniques to explore the rational design of boron-based Lewis acids, which\nplay a pivotal role in organic reactions due to their electron-ccepting\nproperties. Using Fluoride Ion Affinity as a proxy for Lewis acidity, we\ndeveloped interpretable ML models based on chemically meaningful descriptors,\nincluding ab initio computed features and substituent-based parameters derived\nfrom the Hammett linear free-energy relationship. By constraining the chemical\nspace to well-defined molecular scaffolds, we achieved highly accurate\npredictions (mean absolute error \u003c 6 kJ/mol), surpassing conventional black-box\ndeep learning models in low-data regimes. Interpretability analyses of the\nmodels shed light on the origin of Lewis acidity in these compounds and\nidentified actionable levers to modulate it through the nature and positioning\nof substituents on the molecular scaffold. This work bridges ML and chemist's\nway of thinking, demonstrating how explainable models can inspire molecular\ndesign and enhance scientific understanding of chemical reactivity.\n","arxiv_id":"http://arxiv.org/abs/2501.01576v2","authors":["Juliette Fenogli","Laurence Grimaud","Rodolphe Vuilleumier"]},{"title":"(WhyPHI) Fine-Tuning PHI-3 for Multiple-Choice Question Answering:\n Methodology, Results, and Challenges","abstract":" Large Language Models (LLMs) have become essential tools across various\ndomains due to their impressive capabilities in understanding and generating\nhuman-like text. The ability to accurately answer multiple-choice questions\n(MCQs) holds significant value in education, particularly in automated tutoring\nsystems and assessment platforms. However, adapting LLMs to handle MCQ tasks\neffectively remains challenging due to the hallucinations and unclear prompts.\nThis work explores the potential of Microsoft's PHI-3\\cite{Abdin2024}, a\ncompact yet efficient LLM, for MCQ answering. Our contributions include\nfine-tuning the model on the TruthfulQA dataset, designing optimized prompts to\nenhance model performance, and evaluating using perplexity and traditional\nmetrics like accuracy and F1 score. Results show a remarkable improvement in\nPHI-3.5's MCQ handling post-fine-tuning, with perplexity decreasing from 4.68\nto 2.27, and accuracy rising from 62\\% to 90.8\\%. This research underlines the\nimportance of efficient models in adaptive learning systems and educational\nassessments, paving the way for broader integration into the classroom,\nparticularly in fields like test preparation, student feedback, and\npersonalized learning.\n","arxiv_id":"http://arxiv.org/abs/2501.01588v1","authors":["Mohamed Hisham Abdellatif"]},{"title":"Prism: Mining Task-aware Domains in Non-i.i.d. IMU Data for Flexible\n User Perception","abstract":" A wide range of user perception applications leverage inertial measurement\nunit (IMU) data for online prediction. However, restricted by the non-i.i.d.\nnature of IMU data collected from mobile devices, most systems work well only\nin a controlled setting (e.g., for a specific user in particular postures),\nlimiting application scenarios. To achieve uncontrolled online prediction on\nmobile devices, referred to as the flexible user perception (FUP) problem, is\nattractive but hard. In this paper, we propose a novel scheme, called Prism,\nwhich can obtain high FUP accuracy on mobile devices. The core of Prism is to\ndiscover task-aware domains embedded in IMU dataset, and to train a\ndomain-aware model on each identified domain. To this end, we design an\nexpectation-maximization (EM) algorithm to estimate latent domains with respect\nto the specific downstream perception task. Finally, the best-fit model can be\nautomatically selected for use by comparing the test sample and all identified\ndomains in the feature space. We implement Prism on various mobile devices and\nconduct extensive experiments. Results demonstrate that Prism can achieve the\nbest FUP performance with a low latency.\n","arxiv_id":"http://arxiv.org/abs/2501.01598v1","authors":["Yunzhe Li","Facheng Hu","Hongzi Zhu","Quan Liu","Xiaoke Zhao","Jiangang Shen","Shan Chang","Minyi Guo"]},{"title":"Few-shot Implicit Function Generation via Equivariance","abstract":" Implicit Neural Representations (INRs) have emerged as a powerful framework\nfor representing continuous signals. However, generating diverse INR weights\nremains challenging due to limited training data. We introduce Few-shot\nImplicit Function Generation, a new problem setup that aims to generate diverse\nyet functionally consistent INR weights from only a few examples. This is\nchallenging because even for the same signal, the optimal INRs can vary\nsignificantly depending on their initializations. To tackle this, we propose\nEquiGen, a framework that can generate new INRs from limited data. The core\nidea is that functionally similar networks can be transformed into one another\nthrough weight permutations, forming an equivariance group. By projecting these\nweights into an equivariant latent space, we enable diverse generation within\nthese groups, even with few examples. EquiGen implements this through an\nequivariant encoder trained via contrastive learning and smooth augmentation,\nan equivariance-guided diffusion process, and controlled perturbations in the\nequivariant subspace. Experiments on 2D image and 3D shape INR datasets\ndemonstrate that our approach effectively generates diverse INR weights while\npreserving their functional properties in few-shot scenarios.\n","arxiv_id":"http://arxiv.org/abs/2501.01601v1","authors":["Suizhi Huang","Xingyi Yang","Hongtao Lu","Xinchao Wang"]},{"title":"Google is all you need: Semi-Supervised Transfer Learning Strategy For\n Light Multimodal Multi-Task Classification Model","abstract":" As the volume of digital image data increases, the effectiveness of image\nclassification intensifies. This study introduces a robust multi-label\nclassification system designed to assign multiple labels to a single image,\naddressing the complexity of images that may be associated with multiple\ncategories (ranging from 1 to 19, excluding 12). We propose a multi-modal\nclassifier that merges advanced image recognition algorithms with Natural\nLanguage Processing (NLP) models, incorporating a fusion module to integrate\nthese distinct modalities. The purpose of integrating textual data is to\nenhance the accuracy of label prediction by providing contextual understanding\nthat visual analysis alone cannot fully capture. Our proposed classification\nmodel combines Convolutional Neural Networks (CNN) for image processing with\nNLP techniques for analyzing textual description (i.e., captions). This\napproach includes rigorous training and validation phases, with each model\ncomponent verified and analyzed through ablation experiments. Preliminary\nresults demonstrate the classifier's accuracy and efficiency, highlighting its\npotential as an automatic image-labeling system.\n","arxiv_id":"http://arxiv.org/abs/2501.01611v1","authors":["Haixu Liu","Penghao Jiang","Zerui Tao"]},{"title":"Merging Context Clustering with Visual State Space Models for Medical\n Image Segmentation","abstract":" Medical image segmentation demands the aggregation of global and local\nfeature representations, posing a challenge for current methodologies in\nhandling both long-range and short-range feature interactions. Recently, vision\nmamba (ViM) models have emerged as promising solutions for addressing model\ncomplexities by excelling in long-range feature iterations with linear\ncomplexity. However, existing ViM approaches overlook the importance of\npreserving short-range local dependencies by directly flattening spatial tokens\nand are constrained by fixed scanning patterns that limit the capture of\ndynamic spatial context information. To address these challenges, we introduce\na simple yet effective method named context clustering ViM (CCViM), which\nincorporates a context clustering module within the existing ViM models to\nsegment image tokens into distinct windows for adaptable local clustering. Our\nmethod effectively combines long-range and short-range feature interactions,\nthereby enhancing spatial contextual representations for medical image\nsegmentation tasks. Extensive experimental evaluations on diverse public\ndatasets, i.e., Kumar, CPM17, ISIC17, ISIC18, and Synapse demonstrate the\nsuperior performance of our method compared to current state-of-the-art\nmethods. Our code can be found at https://github.com/zymissy/CCViM.\n","arxiv_id":"http://arxiv.org/abs/2501.01618v1","authors":["Yun Zhu","Dong Zhang","Yi Lin","Yifei Feng","Jinhui Tang"]},{"title":"ICPC: In-context Prompt Compression with Faster Inference","abstract":" Despite the recent success of Large Language Models (LLMs), it remains\nchallenging to feed LLMs with long prompts due to the fixed size of LLM inputs.\nAs a remedy, prompt compression becomes a promising solution by removing\nredundant tokens in the prompt. However, using LLM in the existing works\nrequires additional computation resources and leads to memory overheads. To\naddress it, we propose ICPC (In-context Prompt Compression), a novel and\nscalable prompt compression method that adaptively reduces the prompt length.\nThe key idea of ICPC is to calculate the probability of each word appearing in\nthe prompt using encoders and calculate information carried by each word\nthrough the information function, which effectively reduces the information\nloss during prompt compression and increases the speed of compression.\nEmpirically, we demonstrate that ICPC can effectively compress long texts of\ndifferent categories and thus achieve better performance and speed on different\ntypes of NLP tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.01625v1","authors":["Ziyang Yu","Yuyu Liu"]},{"title":"Implications of Artificial Intelligence on Health Data Privacy and\n Confidentiality","abstract":" The rapid integration of artificial intelligence (AI) in healthcare is\nrevolutionizing medical diagnostics, personalized medicine, and operational\nefficiency. However, alongside these advancements, significant challenges arise\nconcerning patient data privacy, ethical considerations, and regulatory\ncompliance. This paper examines the dual impact of AI on healthcare,\nhighlighting its transformative potential and the critical need for\nsafeguarding sensitive health information. It explores the role of the Health\nInsurance Portability and Accountability Act (HIPAA) as a regulatory framework\nfor ensuring data privacy and security, emphasizing the importance of robust\nsafeguards and ethical standards in AI-driven healthcare. Through case studies,\nincluding AI applications in diabetic retinopathy, oncology, and the\ncontroversies surrounding data sharing, this study underscores the ethical and\nlegal complexities of AI implementation. A balanced approach that fosters\ninnovation while maintaining patient trust and privacy is imperative. The\nfindings emphasize the importance of continuous education, transparency, and\nadherence to regulatory frameworks to harness AI's full potential responsibly\nand ethically in healthcare.\n","arxiv_id":"http://arxiv.org/abs/2501.01639v2","authors":["Ahmad Momani"]},{"title":"HLV-1K: A Large-scale Hour-Long Video Benchmark for Time-Specific Long\n Video Understanding","abstract":" Multimodal large language models have become a popular topic in deep visual\nunderstanding due to many promising real-world applications. However, hour-long\nvideo understanding, spanning over one hour and containing tens of thousands of\nvisual frames, remains under-explored because of 1) challenging long-term video\nanalyses, 2) inefficient large-model approaches, and 3) lack of large-scale\nbenchmark datasets. Among them, in this paper, we focus on building a\nlarge-scale hour-long long video benchmark, HLV-1K, designed to evaluate long\nvideo understanding models. HLV-1K comprises 1009 hour-long videos with 14,847\nhigh-quality question answering (QA) and multi-choice question asnwering (MCQA)\npairs with time-aware query and diverse annotations, covering frame-level,\nwithin-event-level, cross-event-level, and long-term reasoning tasks. We\nevaluate our benchmark using existing state-of-the-art methods and demonstrate\nits value for testing deep long video understanding capabilities at different\nlevels and for various tasks. This includes promoting future long video\nunderstanding tasks at a granular level, such as deep understanding of long\nlive videos, meeting recordings, and movies.\n","arxiv_id":"http://arxiv.org/abs/2501.01645v1","authors":["Heqing Zou","Tianze Luo","Guiyang Xie"," Victor"," Zhang","Fengmao Lv","Guangcong Wang","Junyang Chen","Zhuochen Wang","Hansheng Zhang","Huaijian Zhang"]},{"title":"AVATAR: Adversarial Autoencoders with Autoregressive Refinement for Time\n Series Generation","abstract":" Data augmentation can significantly enhance the performance of machine\nlearning tasks by addressing data scarcity and improving generalization.\nHowever, generating time series data presents unique challenges. A model must\nnot only learn a probability distribution that reflects the real data\ndistribution but also capture the conditional distribution at each time step to\npreserve the inherent temporal dependencies. To address these challenges, we\nintroduce AVATAR, a framework that combines Adversarial Autoencoders (AAE) with\nAutoregressive Learning to achieve both objectives. Specifically, our technique\nintegrates the autoencoder with a supervisor and introduces a novel supervised\nloss to assist the decoder in learning the temporal dynamics of time series\ndata. Additionally, we propose another innovative loss function, termed\ndistribution loss, to guide the encoder in more efficiently aligning the\naggregated posterior of the autoencoder's latent representation with a prior\nGaussian distribution. Furthermore, our framework employs a joint training\nmechanism to simultaneously train all networks using a combined loss, thereby\nfulfilling the dual objectives of time series generation. We evaluate our\ntechnique across a variety of time series datasets with diverse\ncharacteristics. Our experiments demonstrate significant improvements in both\nthe quality and practical utility of the generated data, as assessed by various\nqualitative and quantitative metrics.\n","arxiv_id":"http://arxiv.org/abs/2501.01649v1","authors":["MohammadReza EskandariNasab","Shah Muhammad Hamdi","Soukaina Filali Boubrahimi"]},{"title":"EAUWSeg: Eliminating annotation uncertainty in weakly-supervised medical\n image segmentation","abstract":" Weakly-supervised medical image segmentation is gaining traction as it\nrequires only rough annotations rather than accurate pixel-to-pixel labels,\nthereby reducing the workload for specialists. Although some progress has been\nmade, there is still a considerable performance gap between the label-efficient\nmethods and fully-supervised one, which can be attributed to the uncertainty\nnature of these weak labels. To address this issue, we propose a novel weak\nannotation method coupled with its learning framework EAUWSeg to eliminate the\nannotation uncertainty. Specifically, we first propose the Bounded Polygon\nAnnotation (BPAnno) by simply labeling two polygons for a lesion. Then, the\ntailored learning mechanism that explicitly treat bounded polygons as two\nseparated annotations is proposed to learn invariant feature by providing\nadversarial supervision signal for model training. Subsequently, a\nconfidence-auxiliary consistency learner incorporates with a\nclassification-guided confidence generator is designed to provide reliable\nsupervision signal for pixels in uncertain region by leveraging the feature\npresentation consistency across pixels within the same category as well as\nclass-specific information encapsulated in bounded polygons annotation.\nExperimental results demonstrate that EAUWSeg outperforms existing\nweakly-supervised segmentation methods. Furthermore, compared to\nfully-supervised counterparts, the proposed method not only delivers superior\nperformance but also costs much less annotation workload. This underscores the\nsuperiority and effectiveness of our approach.\n","arxiv_id":"http://arxiv.org/abs/2501.01658v1","authors":["Wang Lituan","Zhang Lei","Wang Yan","Wang Zhenbin","Zhang Zhenwei","Zhang Yi"]},{"title":"BARTPredict: Empowering IoT Security with LLM-Driven Cyber Threat\n Prediction","abstract":" The integration of Internet of Things (IoT) technology in various domains has\nled to operational advancements, but it has also introduced new vulnerabilities\nto cybersecurity threats, as evidenced by recent widespread cyberattacks on IoT\ndevices. Intrusion detection systems are often reactive, triggered by specific\npatterns or anomalies observed within the network. To address this challenge,\nthis work proposes a proactive approach to anticipate and preemptively mitigate\nmalicious activities, aiming to prevent potential damage before it occurs. This\npaper proposes an innovative intrusion prediction framework empowered by\nPre-trained Large Language Models (LLMs). The framework incorporates two LLMs:\na fine-tuned Bidirectional and AutoRegressive Transformers (BART) model for\npredicting network traffic and a fine-tuned Bidirectional Encoder\nRepresentations from Transformers (BERT) model for evaluating the predicted\ntraffic. By harnessing the bidirectional capabilities of BART the framework\nthen identifies malicious packets among these predictions. Evaluated using the\nCICIoT2023 IoT attack dataset, our framework showcases a notable enhancement in\npredictive performance, attaining an impressive 98% overall accuracy, providing\na powerful response to the cybersecurity challenges that confront IoT networks.\n","arxiv_id":"http://arxiv.org/abs/2501.01664v1","authors":["Alaeddine Diaf","Abdelaziz Amara Korba","Nour Elislem Karabadji","Yacine Ghamri-Doudane"]},{"title":"Adaptive Few-shot Prompting for Machine Translation with Pre-trained\n Language Models","abstract":" Recently, Large language models (LLMs) with in-context learning have\ndemonstrated remarkable potential in handling neural machine translation.\nHowever, existing evidence shows that LLMs are prompt-sensitive and it is\nsub-optimal to apply the fixed prompt to any input for downstream machine\ntranslation tasks. To address this issue, we propose an adaptive few-shot\nprompting (AFSP) framework to automatically select suitable translation\ndemonstrations for various source input sentences to further elicit the\ntranslation capability of an LLM for better machine translation. First, we\nbuild a translation demonstration retrieval module based on LLM's embedding to\nretrieve top-k semantic-similar translation demonstrations from aligned\nparallel translation corpus. Rather than using other embedding models for\nsemantic demonstration retrieval, we build a hybrid demonstration retrieval\nmodule based on the embedding layer of the deployed LLM to build better input\nrepresentation for retrieving more semantic-related translation demonstrations.\nThen, to ensure better semantic consistency between source inputs and target\noutputs, we force the deployed LLM itself to generate multiple output\ncandidates in the target language with the help of translation demonstrations\nand rerank these candidates. Besides, to better evaluate the effectiveness of\nour AFSP framework on the latest language and extend the research boundary of\nneural machine translation, we construct a high-quality diplomatic\nChinese-English parallel dataset that consists of 5,528 parallel\nChinese-English sentences. Finally, extensive experiments on the proposed\ndiplomatic Chinese-English parallel dataset and the United Nations Parallel\nCorpus (Chinese-English part) show the effectiveness and superiority of our\nproposed AFSP.\n","arxiv_id":"http://arxiv.org/abs/2501.01679v1","authors":["Lei Tang","Jinghui Qin","Wenxuan Ye","Hao Tan","Zhijing Yang"]},{"title":"VidFormer: A novel end-to-end framework fused by 3DCNN and Transformer\n for Video-based Remote Physiological Measurement","abstract":" Remote physiological signal measurement based on facial videos, also known as\nremote photoplethysmography (rPPG), involves predicting changes in facial\nvascular blood flow from facial videos. While most deep learning-based methods\nhave achieved good results, they often struggle to balance performance across\nsmall and large-scale datasets due to the inherent limitations of convolutional\nneural networks (CNNs) and Transformer. In this paper, we introduce VidFormer,\na novel end-to-end framework that integrates 3-Dimension Convolutional Neural\nNetwork (3DCNN) and Transformer models for rPPG tasks. Initially, we conduct an\nanalysis of the traditional skin reflection model and subsequently introduce an\nenhanced model for the reconstruction of rPPG signals. Based on this improved\nmodel, VidFormer utilizes 3DCNN and Transformer to extract local and global\nfeatures from input data, respectively. To enhance the spatiotemporal feature\nextraction capabilities of VidFormer, we incorporate temporal-spatial attention\nmechanisms tailored for both 3DCNN and Transformer. Additionally, we design a\nmodule to facilitate information exchange and fusion between the 3DCNN and\nTransformer. Our evaluation on five publicly available datasets demonstrates\nthat VidFormer outperforms current state-of-the-art (SOTA) methods. Finally, we\ndiscuss the essential roles of each VidFormer module and examine the effects of\nethnicity, makeup, and exercise on its performance.\n","arxiv_id":"http://arxiv.org/abs/2501.01691v2","authors":["Jiachen Li","Shisheng Guo","Longzhen Tang","Cuolong Cui","Lingjiang Kong","Xiaobo Yang"]},{"title":"The Essence of Contextual Understanding in Theory of Mind: A Study on\n Question Answering with Story Characters","abstract":" Theory-of-Mind (ToM) is a fundamental psychological capability that allows\nhumans to understand and interpret the mental states of others. Humans infer\nothers' thoughts by integrating causal cues and indirect clues from broad\ncontextual information, often derived from past interactions. In other words,\nhuman ToM heavily relies on the understanding about the backgrounds and life\nstories of others. Unfortunately, this aspect is largely overlooked in existing\nbenchmarks for evaluating machines' ToM capabilities, due to their usage of\nshort narratives without global backgrounds. In this paper, we verify the\nimportance of understanding long personal backgrounds in ToM and assess the\nperformance of LLMs in such realistic evaluation scenarios. To achieve this, we\nintroduce a novel benchmark, CharToM-QA, comprising 1,035 ToM questions based\non characters from classic novels. Our human study reveals a significant\ndisparity in performance: the same group of educated participants performs\ndramatically better when they have read the novels compared to when they have\nnot. In parallel, our experiments on state-of-the-art LLMs, including the very\nrecent o1 model, show that LLMs still perform notably worse than humans,\ndespite that they have seen these stories during pre-training. This highlights\nthe limitations of current LLMs in capturing the nuanced contextual information\nrequired for ToM reasoning.\n","arxiv_id":"http://arxiv.org/abs/2501.01705v1","authors":["Chulun Zhou","Qiujing Wang","Mo Yu","Xiaoqian Yue","Rui Lu","Jiangnan Li","Yifan Zhou","Shunchi Zhang","Jie Zhou","Wai Lam"]},{"title":"MoVE-KD: Knowledge Distillation for VLMs with Mixture of Visual Encoders","abstract":" Visual encoders are fundamental components in vision-language models (VLMs),\neach showcasing unique strengths derived from various pre-trained visual\nfoundation models. To leverage the various capabilities of these encoders,\nrecent studies incorporate multiple encoders within a single VLM, leading to a\nconsiderable increase in computational cost. In this paper, we present\nMixture-of-Visual-Encoder Knowledge Distillation (MoVE-KD), a novel framework\nthat distills the unique proficiencies of multiple vision encoders into a\nsingle, efficient encoder model. Specifically, to mitigate conflicts and retain\nthe unique characteristics of each teacher encoder, we employ low-rank\nadaptation (LoRA) and mixture-of-experts (MoEs) to selectively activate\nspecialized knowledge based on input features, enhancing both adaptability and\nefficiency. To regularize the KD process and enhance performance, we propose an\nattention-based distillation strategy that adaptively weighs the different\nvisual encoders and emphasizes valuable visual tokens, reducing the burden of\nreplicating comprehensive but distinct features from multiple teachers.\nComprehensive experiments on popular VLMs, such as LLaVA and LLaVA-NeXT,\nvalidate the effectiveness of our method. The code will be released.\n","arxiv_id":"http://arxiv.org/abs/2501.01709v1","authors":["Jiajun Cao","Yuan Zhang","Tao Huang","Ming Lu","Qizhe Zhang","Ruichuan An","Ningning MA","Shanghang Zhang"]},{"title":"LLMs \u0026 Legal Aid: Understanding Legal Needs Exhibited Through User\n Queries","abstract":" The paper presents a preliminary analysis of an experiment conducted by Frank\nBold, a Czech expert group, to explore user interactions with GPT-4 for\naddressing legal queries. Between May 3, 2023, and July 25, 2023, 1,252 users\nsubmitted 3,847 queries. Unlike studies that primarily focus on the accuracy,\nfactuality, or hallucination tendencies of large language models (LLMs), our\nanalysis focuses on the user query dimension of the interaction. Using GPT-4o\nfor zero-shot classification, we categorized queries on (1) whether users\nprovided factual information about their issue (29.95%) or not (70.05%), (2)\nwhether they sought legal information (64.93%) or advice on the course of\naction (35.07\\%), and (3) whether they imposed requirements to shape or control\nthe model's answer (28.57%) or not (71.43%). We provide both quantitative and\nqualitative insight into user needs and contribute to a better understanding of\nuser engagement with LLMs.\n","arxiv_id":"http://arxiv.org/abs/2501.01711v1","authors":["Michal Kuk","Jakub Harasta"]},{"title":"Proposing Hierarchical Goal-Conditioned Policy Planning in Multi-Goal\n Reinforcement Learning","abstract":" Humanoid robots must master numerous tasks with sparse rewards, posing a\nchallenge for reinforcement learning (RL). We propose a method combining RL and\nautomated planning to address this. Our approach uses short goal-conditioned\npolicies (GCPs) organized hierarchically, with Monte Carlo Tree Search (MCTS)\nplanning using high-level actions (HLAs). Instead of primitive actions, the\nplanning process generates HLAs. A single plan-tree, maintained during the\nagent's lifetime, holds knowledge about goal achievement. This hierarchy\nenhances sample efficiency and speeds up reasoning by reusing HLAs and\nanticipating future actions. Our Hierarchical Goal-Conditioned Policy Planning\n(HGCPP) framework uniquely integrates GCPs, MCTS, and hierarchical RL,\npotentially improving exploration and planning in complex tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.01727v1","authors":["Gavin B. Rens"]},{"title":"Augmentation Matters: A Mix-Paste Method for X-Ray Prohibited Item\n Detection under Noisy Annotations","abstract":" Automatic X-ray prohibited item detection is vital for public safety.\nExisting deep learning-based methods all assume that the annotations of\ntraining X-ray images are correct. However, obtaining correct annotations is\nextremely hard if not impossible for large-scale X-ray images, where item\noverlapping is ubiquitous.As a result, X-ray images are easily contaminated\nwith noisy annotations, leading to performance deterioration of existing\nmethods.In this paper, we address the challenging problem of training a robust\nprohibited item detector under noisy annotations (including both category noise\nand bounding box noise) from a novel perspective of data augmentation, and\npropose an effective label-aware mixed patch paste augmentation method\n(Mix-Paste). Specifically, for each item patch, we mix several item patches\nwith the same category label from different images and replace the original\npatch in the image with the mixed patch. In this way, the probability of\ncontaining the correct prohibited item within the generated image is increased.\nMeanwhile, the mixing process mimics item overlapping, enabling the model to\nlearn the characteristics of X-ray images. Moreover, we design an item-based\nlarge-loss suppression (LLS) strategy to suppress the large losses\ncorresponding to potentially positive predictions of additional items due to\nthe mixing operation. We show the superiority of our method on X-ray datasets\nunder noisy annotations. In addition, we evaluate our method on the noisy\nMS-COCO dataset to showcase its generalization ability. These results clearly\nindicate the great potential of data augmentation to handle noise annotations.\nThe source code is released at https://github.com/wscds/Mix-Paste.\n","arxiv_id":"http://arxiv.org/abs/2501.01733v1","authors":["Ruikang Chen","Yan Yan","Jing-Hao Xue","Yang Lu","Hanzi Wang"]},{"title":"Automating Legal Concept Interpretation with LLMs: Retrieval,\n Generation, and Evaluation","abstract":" Legal articles often include vague concepts to adapt to the ever-changing\nsociety. Providing detailed interpretations of these concepts is a critical\ntask for legal practitioners, which requires meticulous and professional\nannotations by legal experts, admittedly time-consuming and expensive to\ncollect at scale. In this paper, we introduce a novel retrieval-augmented\ngeneration framework, ATRI, for AuTomatically Retrieving relevant information\nfrom past judicial precedents and Interpreting vague legal concepts. We further\npropose a new benchmark, Legal Concept Entailment, to automate the evaluation\nof generated concept interpretations without expert involvement. Automatic\nevaluations indicate that our generated interpretations can effectively assist\nlarge language models (LLMs) in understanding vague legal concepts.\nMulti-faceted evaluations by legal experts indicate that the quality of our\nconcept interpretations is comparable to those written by human experts. Our\nwork has strong implications for leveraging LLMs to support legal practitioners\nin interpreting vague legal concepts and beyond.\n","arxiv_id":"http://arxiv.org/abs/2501.01743v1","authors":["Kangcheng Luo","Quzhe Huang","Cong Jiang","Yansong Feng"]},{"title":"Creating Artificial Students that Never Existed: Leveraging Large\n Language Models and CTGANs for Synthetic Data Generation","abstract":" In this study, we explore the growing potential of AI and deep learning\ntechnologies, particularly Generative Adversarial Networks (GANs) and Large\nLanguage Models (LLMs), for generating synthetic tabular data. Access to\nquality students data is critical for advancing learning analytics, but privacy\nconcerns and stricter data protection regulations worldwide limit their\navailability and usage. Synthetic data offers a promising alternative. We\ninvestigate whether synthetic data can be leveraged to create artificial\nstudents for serving learning analytics models. Using the popular GAN model\nCTGAN and three LLMs- GPT2, DistilGPT2, and DialoGPT, we generate synthetic\ntabular student data. Our results demonstrate the strong potential of these\nmethods to produce high-quality synthetic datasets that resemble real students\ndata. To validate our findings, we apply a comprehensive set of utility\nevaluation metrics to assess the statistical and predictive performance of the\nsynthetic data and compare the different generator models used, specially the\nperformance of LLMs. Our study aims to provide the learning analytics community\nwith valuable insights into the use of synthetic data, laying the groundwork\nfor expanding the field methodological toolbox with new innovative approaches\nfor learning analytics data generation.\n","arxiv_id":"http://arxiv.org/abs/2501.01793v1","authors":["Mohammad Khalil","Farhad Vadiee","Ronas Shakya","Qinyi Liu"]},{"title":"End-to-End Long Document Summarization using Gradient Caching","abstract":" Training transformer-based encoder-decoder models for long document\nsummarization poses a significant challenge due to the quadratic memory\nconsumption during training. Several approaches have been proposed to extend\nthe input length at test time, but training with these approaches is still\ndifficult, requiring truncation of input documents and causing a mismatch\nbetween training and test conditions. In this work, we propose CachED (Gradient\n$\\textbf{Cach}$ing for $\\textbf{E}$ncoder-$\\textbf{D}$ecoder models), an\napproach that enables end-to-end training of existing transformer-based\nencoder-decoder models, using the entire document without truncation.\nSpecifically, we apply non-overlapping sliding windows to input documents,\nfollowed by fusion in decoder. During backpropagation, the gradients are cached\nat the decoder and are passed through the encoder in chunks by re-computing the\nhidden vectors, similar to gradient checkpointing. In the experiments on long\ndocument summarization, we extend BART to CachED BART, processing more than\n500K tokens during training and achieving superior performance without using\nany additional parameters.\n","arxiv_id":"http://arxiv.org/abs/2501.01805v1","authors":["Rohit Saxena","Hao Tang","Frank Keller"]},{"title":"SDPO: Segment-Level Direct Preference Optimization for Social Agents","abstract":" Social agents powered by large language models (LLMs) can simulate human\nsocial behaviors but fall short in handling complex goal-oriented social\ndialogues. Direct Preference Optimization (DPO) has proven effective in\naligning LLM behavior with human preferences across a variety of agent tasks.\nExisting DPO-based approaches for multi-turn interactions are divided into\nturn-level and session-level methods. The turn-level method is overly\nfine-grained, focusing exclusively on individual turns, while session-level\nmethods are too coarse-grained, often introducing training noise. To address\nthese limitations, we propose Segment-Level Direct Preference Optimization\n(SDPO), which focuses on specific key segments within interactions to optimize\nmulti-turn agent behavior while minimizing training noise. Evaluations on the\nSOTOPIA benchmark demonstrate that SDPO-tuned agents consistently outperform\nboth existing DPO-based methods and proprietary LLMs like GPT-4o, underscoring\nSDPO's potential to advance the social intelligence of LLM-based agents. We\nrelease our code and data at\nhttps://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/SDPO.\n","arxiv_id":"http://arxiv.org/abs/2501.01821v1","authors":["Aobo Kong","Wentao Ma","Shiwan Zhao","Yongbin Li","Yuchuan Wu","Ke Wang","Xiaoqian Liu","Qicheng Li","Yong Qin","Fei Huang"]},{"title":"The Proof is in the Almond Cookies","abstract":" This paper presents a case study on how to process cooking recipes (and more\ngenerally, how-to instructions) in a way that makes it possible for a robot or\nartificial cooking assistant to support human chefs in the kitchen. Such AI\nassistants would be of great benefit to society, as they can help to sustain\nthe autonomy of aging adults or people with a physical impairment, or they may\nreduce the stress in a professional kitchen. We propose a novel approach to\ncomputational recipe understanding that mimics the human sense-making process,\nwhich is narrative-based. Using an English recipe for almond crescent cookies\nas illustration, we show how recipes can be modelled as rich narrative\nstructures by integrating various knowledge sources such as language\nprocessing, ontologies, and mental simulation. We show how such narrative\nstructures can be used for (a) dealing with the challenges of recipe language,\nsuch as zero anaphora, (b) optimizing a robot's planning process, (c) measuring\nhow well an AI system understands its current tasks, and (d) allowing recipe\nannotations to become language-independent.\n","arxiv_id":"http://arxiv.org/abs/2501.01827v1","authors":["Remi van Trijp","Katrien Beuls","Paul Van Eecke"]},{"title":"MoColl: Agent-Based Specific and General Model Collaboration for Image\n Captioning","abstract":" Image captioning is a critical task at the intersection of computer vision\nand natural language processing, with wide-ranging applications across various\ndomains. For complex tasks such as diagnostic report generation, deep learning\nmodels require not only domain-specific image-caption datasets but also the\nincorporation of relevant general knowledge to provide contextual accuracy.\nExisting approaches exhibit inherent limitations: specialized models excel in\ncapturing domain-specific details but lack generalization, while\nvision-language models (VLMs) built on large language models (LLMs) leverage\ngeneral knowledge but struggle with domain-specific adaptation. To address\nthese limitations, this paper proposes a novel agent-enhanced model\ncollaboration framework, which we call MoColl, designed to effectively\nintegrate domain-specific and general knowledge. Specifically, our approach is\nto decompose complex image captioning tasks into a series of interconnected\nquestion-answer subtasks. A trainable visual question answering (VQA) model is\nemployed as a specialized tool to focus on domain-specific visual analysis,\nanswering task-specific questions based on image content. Concurrently, an\nLLM-based agent with general knowledge formulates these questions and\nsynthesizes the resulting question-answer pairs into coherent captions. Beyond\nits role in leveraging the VQA model, the agent further guides its training to\nenhance its domain-specific capabilities. Experimental results on radiology\nreport generation validate the effectiveness of the proposed framework,\ndemonstrating significant improvements in the quality of generated reports.\n","arxiv_id":"http://arxiv.org/abs/2501.01834v2","authors":["Pu Yang","Bin Dong"]},{"title":"Practical machine learning is learning on small samples","abstract":" Based on limited observations, machine learning discerns a dependence which\nis expected to hold in the future. What makes it possible? Statistical learning\ntheory imagines indefinitely increasing training sample to justify its\napproach. In reality, there is no infinite time or even infinite general\npopulation for learning. Here I argue that practical machine learning is based\non an implicit assumption that underlying dependence is relatively ``smooth\" :\nlikely, there are no abrupt differences in feedback between cases with close\ndata points. From this point of view learning shall involve selection of the\nhypothesis ``smoothly\" approximating the training set. I formalize this as\nPractical learning paradigm. The paradigm includes terminology and rules for\ndescription of learners. Popular learners (local smoothing, k-NN, decision\ntrees, Naive Bayes, SVM for classification and for regression) are shown here\nto be implementations of this paradigm.\n","arxiv_id":"http://arxiv.org/abs/2501.01836v1","authors":["Marina Sapir"]},{"title":"Multi-Agent Conversational Online Learning for Adaptive LLM Response\n Identification","abstract":" The remarkable generative capability of large language models (LLMs) has\nsparked a growing interest in automatically generating responses for different\napplications. Given the dynamic nature of user preferences and the uncertainty\nof LLM response performance, it is crucial to design efficient online learning\nalgorithms to identify optimal LLM responses (i.e., high-quality responses that\nalso meet user preferences). Most existing online algorithms adopt a\ncentralized approach and fail to leverage explicit user preferences for more\nefficient and personalized LLM response identification. In contrast, this paper\nintroduces \\textit{MACO} (\\underline{M}ulti-\\underline{A}gent\n\\underline{C}onversational \\underline{O}nline Learning for Adaptive LLM\nResponse Identification): 1) The online LLM response identification process is\naccelerated by multiple local agents (such as smartphones), while enhancing\ndata privacy; 2) A novel conversational mechanism is proposed to adaptively\nconduct conversations for soliciting user preferences (e.g., a preference for a\nhumorous tone over a serious one in generated responses), so to minimize\nuncertainty in preference estimation. Our theoretical analysis demonstrates\nthat \\cadi\\ is near-optimal regarding cumulative regret. Additionally, \\cadi\\\noffers reduced communication costs and computational complexity by eliminating\nthe traditional, computing-intensive ``G-optimal design\" found in previous\nworks. Extensive experiments with the open LLM \\textit{Llama}, coupled with two\ndifferent embedding models from Google and OpenAI for text vector\nrepresentation, demonstrate that \\cadi\\ significantly outperforms the current\nstate-of-the-art in online LLM response identification.\n","arxiv_id":"http://arxiv.org/abs/2501.01849v1","authors":["Xiangxiang Dai","Yuejin Xie","Maoli Liu","Xuchuang Wang","Zhuohua Li","Huanyu Wang","John C. S. Lui"]},{"title":"Accuracy Can Lie: On the Impact of Surrogate Model in Configuration\n Tuning","abstract":" To ease the expensive measurements during configuration tuning, it is natural\nto build a surrogate model as the replacement of the system, and thereby the\nconfiguration performance can be cheaply evaluated. Yet, a stereotype therein\nis that the higher the model accuracy, the better the tuning result would be.\nThis \"accuracy is all\" belief drives our research community to build more and\nmore accurate models and criticize a tuner for the inaccuracy of the model\nused. However, this practice raises some previously unaddressed questions,\ne.g., Do those somewhat small accuracy improvements reported in existing work\nreally matter much to the tuners? What role does model accuracy play in the\nimpact of tuning quality? To answer those related questions, we conduct one of\nthe largest-scale empirical studies to date-running over the period of 13\nmonths 24*7-that covers 10 models, 17 tuners, and 29 systems from the existing\nworks while under four different commonly used metrics, leading to 13,612 cases\nof investigation. Surprisingly, our key findings reveal that the accuracy can\nlie: there are a considerable number of cases where higher accuracy actually\nleads to no improvement in the tuning outcomes (up to 58% cases under certain\nsetting), or even worse, it can degrade the tuning quality (up to 24% cases\nunder certain setting). We also discover that the chosen models in most\nproposed tuners are sub-optimal and that the required % of accuracy change to\nsignificantly improve tuning quality varies according to the range of model\naccuracy. Deriving from the fitness landscape analysis, we provide in-depth\ndiscussions of the rationale behind, offering several lessons learned as well\nas insights for future opportunities. Most importantly, this work poses a clear\nmessage to the community: we should take one step back from the natural\n\"accuracy is all\" belief for model-based configuration tuning.\n","arxiv_id":"http://arxiv.org/abs/2501.01876v1","authors":["Pengzhou Chen","Jingzhi Gong","Tao Chen"]},{"title":"Virgo: A Preliminary Exploration on Reproducing o1-like MLLM","abstract":" Recently, slow-thinking reasoning systems, built upon large language models\n(LLMs), have garnered widespread attention by scaling the thinking time during\ninference. There is also growing interest in adapting this capability to\nmultimodal large language models (MLLMs). Given that MLLMs handle more complex\ndata semantics across different modalities, it is intuitively more challenging\nto implement multimodal slow-thinking systems.\n To address this issue, in this paper, we explore a straightforward approach\nby fine-tuning a capable MLLM with a small amount of textual long-form thought\ndata, resulting in a multimodal slow-thinking system, Virgo (Visual reasoning\nwith long thought). We find that these long-form reasoning processes, expressed\nin natural language, can be effectively transferred to MLLMs. Moreover, it\nseems that such textual reasoning data can be even more effective than visual\nreasoning data in eliciting the slow-thinking capacities of MLLMs. While this\nwork is preliminary, it demonstrates that slow-thinking capacities are\nfundamentally associated with the language model component, which can be\ntransferred across modalities or domains. This finding can be leveraged to\nguide the development of more powerful slow-thinking reasoning systems. We\nrelease our resources at https://github.com/RUCAIBox/Virgo.\n","arxiv_id":"http://arxiv.org/abs/2501.01904v1","authors":["Yifan Du","Zikang Liu","Yifan Li","Wayne Xin Zhao","Yuqi Huo","Bingning Wang","Weipeng Chen","Zheng Liu","Zhongyuan Wang","Ji-Rong Wen"]},{"title":"Mitigating Hallucination for Large Vision Language Model by\n Inter-Modality Correlation Calibration Decoding","abstract":" Large vision-language models (LVLMs) have shown remarkable capabilities in\nvisual-language understanding for downstream multi-modal tasks. Despite their\nsuccess, LVLMs still suffer from generating hallucinations in complex\ngeneration tasks, leading to inconsistencies between visual inputs and\ngenerated content. To address this issue, some approaches have introduced\ninference-time interventions, such as contrastive decoding and attention\nrectification, to reduce overreliance on language priors. However, these\napproaches overlook hallucinations stemming from spurious inter-modality\ncorrelations. In this paper, we propose an Inter-Modality Correlation\nCalibration Decoding (IMCCD) method to mitigate hallucinations in LVLMs in a\ntraining-free manner. In this method, we design a Cross-Modal Value-Enhanced\nDecoding(CMVED) module to alleviate hallucination by a novel contrastive\ndecoding mechanism. During the estimation of distorted distribution, CMVED\nmasks the value vectors associated with significant cross-modal attention\nweights, which address both uni-modality overreliance and misleading\ninter-modality correlations. Additionally, a Content-Driven Attention\nRefinement(CDAR) module refines cross-modal attention weights, guiding LVLMs to\nfocus on important visual content. Experimental results on diverse\nhallucination benchmarks validate the superiority of our method over existing\nstate-of-the-art techniques in reducing hallucinations in LVLM text generation.\nOur code will be available at https://github.com/lijm48/IMCCD.\n","arxiv_id":"http://arxiv.org/abs/2501.01926v1","authors":["Jiaming Li","Jiacheng Zhang","Zequn Jie","Lin Ma","Guanbin Li"]},{"title":"Abstractive Text Summarization for Contemporary Sanskrit Prose: Issues\n and Challenges","abstract":" This thesis presents Abstractive Text Summarization models for contemporary\nSanskrit prose. The first chapter, titled Introduction, presents the motivation\nbehind this work, the research questions, and the conceptual framework.\nSanskrit is a low-resource inflectional language. The key research question\nthat this thesis investigates is what the challenges in developing an\nabstractive TS for Sanskrit. To answer the key research questions,\nsub-questions based on four different themes have been posed in this work. The\nsecond chapter, Literature Review, surveys the previous works done. The third\nchapter, data preparation, answers the remaining three questions from the third\ntheme. It reports the data collection and preprocessing challenges for both\nlanguage model and summarization model trainings. The fourth chapter reports\nthe training and inference of models and the results obtained therein. This\nresearch has initiated a pipeline for Sanskrit abstractive text summarization\nand has reported the challenges faced at every stage of the development. The\nresearch questions based on every theme have been answered to answer the key\nresearch question.\n","arxiv_id":"http://arxiv.org/abs/2501.01933v1","authors":["Shagun Sinha"]},{"title":"Cold-Start Recommendation towards the Era of Large Language Models\n (LLMs): A Comprehensive Survey and Roadmap","abstract":" Cold-start problem is one of the long-standing challenges in recommender\nsystems, focusing on accurately modeling new or interaction-limited users or\nitems to provide better recommendations. Due to the diversification of internet\nplatforms and the exponential growth of users and items, the importance of\ncold-start recommendation (CSR) is becoming increasingly evident. At the same\ntime, large language models (LLMs) have achieved tremendous success and possess\nstrong capabilities in modeling user and item information, providing new\npotential for cold-start recommendations. However, the research community on\nCSR still lacks a comprehensive review and reflection in this field. Based on\nthis, in this paper, we stand in the context of the era of large language\nmodels and provide a comprehensive review and discussion on the roadmap,\nrelated literature, and future directions of CSR. Specifically, we have\nconducted an exploration of the development path of how existing CSR utilizes\ninformation, from content features, graph relations, and domain information, to\nthe world knowledge possessed by large language models, aiming to provide new\ninsights for both the research and industrial communities on CSR. Related\nresources of cold-start recommendations are collected and continuously updated\nfor the community in\nhttps://github.com/YuanchenBei/Awesome-Cold-Start-Recommendation.\n","arxiv_id":"http://arxiv.org/abs/2501.01945v2","authors":["Weizhi Zhang","Yuanchen Bei","Liangwei Yang","Henry Peng Zou","Peilin Zhou","Aiwei Liu","Yinghui Li","Hao Chen","Jianling Wang","Yu Wang","Feiran Huang","Sheng Zhou","Jiajun Bu","Allen Lin","James Caverlee","Fakhri Karray","Irwin King","Philip S. Yu"]},{"title":"MADGEN: Mass-Spec attends to De Novo Molecular generation","abstract":" The annotation (assigning structural chemical identities) of MS/MS spectra\nremains a significant challenge due to the enormous molecular diversity in\nbiological samples and the limited scope of reference databases. Currently, the\nvast majority of spectral measurements remain in the \"dark chemical space\"\nwithout structural annotations. To improve annotation, we propose MADGEN\n(Mass-spec Attends to De Novo Molecular GENeration), a scaffold-based method\nfor de novo molecular structure generation guided by mass spectrometry data.\nMADGEN operates in two stages: scaffold retrieval and spectra-conditioned\nmolecular generation starting with the scaffold. In the first stage, given an\nMS/MS spectrum, we formulate scaffold retrieval as a ranking problem and employ\ncontrastive learning to align mass spectra with candidate molecular scaffolds.\nIn the second stage, starting from the retrieved scaffold, we employ the MS/MS\nspectrum to guide an attention-based generative model to generate the final\nmolecule. Our approach constrains the molecular generation search space,\nreducing its complexity and improving generation accuracy. We evaluate MADGEN\non three datasets (NIST23, CANOPUS, and MassSpecGym) and evaluate MADGEN's\nperformance with a predictive scaffold retriever and with an oracle retriever.\nWe demonstrate the effectiveness of using attention to integrate spectral\ninformation throughout the generation process to achieve strong results with\nthe oracle retriever.\n","arxiv_id":"http://arxiv.org/abs/2501.01950v2","authors":["Yinkai Wang","Xiaohui Chen","Liping Liu","Soha Hassoun"]},{"title":"MixGCN: Scalable GCN Training by Mixture of Parallelism and Mixture of\n Accelerators","abstract":" Graph convolutional networks (GCNs) have demonstrated superiority in\ngraph-based learning tasks. However, training GCNs on full graphs is\nparticularly challenging, due to the following two challenges: (1) the\nassociated feature tensors can easily explode the memory and block the\ncommunication bandwidth of modern accelerators, and (2) the computation\nworkflow in training GCNs alternates between sparse and dense matrix\noperations, complicating the efficient utilization of computational resources.\nExisting solutions for scalable distributed full-graph GCN training mostly\nadopt partition parallelism, which is unsatisfactory as they only partially\naddress the first challenge while incurring scaled-out communication volume. To\nthis end, we propose MixGCN aiming to simultaneously address both the\naforementioned challenges towards GCN training. To tackle the first challenge,\nMixGCN integrates mixture of parallelism. Both theoretical and empirical\nanalysis verify its constant communication volumes and enhanced balanced\nworkload; For handling the second challenge, we consider mixture of\naccelerators (i.e., sparse and dense accelerators) with a dedicated accelerator\nfor GCN training and a fine-grain pipeline. Extensive experiments show that\nMixGCN achieves boosted training efficiency and scalability.\n","arxiv_id":"http://arxiv.org/abs/2501.01951v2","authors":["Cheng Wan","Runkai Tao","Zheng Du","Yang Katie Zhao","Yingyan Celine Lin"]},{"title":"SmartSpatial: Enhancing the 3D Spatial Arrangement Capabilities of\n Stable Diffusion Models and Introducing a Novel 3D Spatial Evaluation\n Framework","abstract":" Stable Diffusion models have made remarkable strides in generating\nphotorealistic images from text prompts but often falter when tasked with\naccurately representing complex spatial arrangements, particularly involving\nintricate 3D relationships. To address this limitation, we introduce\nSmartSpatial, an innovative approach that enhances the spatial arrangement\ncapabilities of Stable Diffusion models through 3D-aware conditioning and\nattention-guided mechanisms. SmartSpatial incorporates depth information and\nemploys cross-attention control to ensure precise object placement, delivering\nnotable improvements in spatial accuracy metrics. In conjunction with\nSmartSpatial, we present SmartSpatialEval, a comprehensive evaluation framework\ndesigned to assess spatial relationships. This framework utilizes\nvision-language models and graph-based dependency parsing for performance\nanalysis. Experimental results on the COCO and SpatialPrompts datasets show\nthat SmartSpatial significantly outperforms existing methods, setting new\nbenchmarks for spatial arrangement accuracy in image generation.\n","arxiv_id":"http://arxiv.org/abs/2501.01998v1","authors":["Mao Xun Huang","Hen-Hsen Huang"]},{"title":"Multi-Task Semantic Communication With Graph Attention-Based Feature\n Correlation Extraction","abstract":" Multi-task semantic communication can serve multiple learning tasks using a\nshared encoder model. Existing models have overlooked the intricate\nrelationships between features extracted during an encoding process of tasks.\nThis paper presents a new graph attention inter-block (GAI) module to the\nencoder/transmitter of a multi-task semantic communication system, which\nenriches the features for multiple tasks by embedding the intermediate outputs\nof encoding in the features, compared to the existing techniques. The key idea\nis that we interpret the outputs of the intermediate feature extraction blocks\nof the encoder as the nodes of a graph to capture the correlations of the\nintermediate features. Another important aspect is that we refine the node\nrepresentation using a graph attention mechanism to extract the correlations\nand a multi-layer perceptron network to associate the node representations with\ndifferent tasks. Consequently, the intermediate features are weighted and\nembedded into the features transmitted for executing multiple tasks at the\nreceiver. Experiments demonstrate that the proposed model surpasses the most\ncompetitive and publicly available models by 11.4% on the CityScapes 2Task\ndataset and outperforms the established state-of-the-art by 3.97% on the NYU V2\n3Task dataset, respectively, when the bandwidth ratio of the communication\nchannel (i.e., compression level for transmission over the channel) is as\nconstrained as 1 12 .\n","arxiv_id":"http://arxiv.org/abs/2501.02006v1","authors":["Xi Yu","Tiejun Lv","Weicai Li","Wei Ni","Dusit Niyato","Ekram Hossain"]},{"title":"TART: Token-based Architecture Transformer for Neural Network\n Performance Prediction","abstract":" In the realm of neural architecture design, achieving high performance is\nlargely reliant on the manual expertise of researchers. Despite the emergence\nof Neural Architecture Search (NAS) as a promising technique for automating\nthis process, current NAS methods still require human input to expand the\nsearch space and cannot generate new architectures. This paper explores the\npotential of Transformers in comprehending neural architectures and their\nperformance, with the objective of establishing the foundation for utilizing\nTransformers to generate novel networks. We propose the Token-based\nArchitecture Transformer (TART), which predicts neural network performance\nwithout the need to train candidate networks. TART attains state-of-the-art\nperformance on the DeepNets-1M dataset for performance prediction tasks without\nedge information, indicating the potential of Transformers to aid in\ndiscovering novel and high-performing neural architectures.\n","arxiv_id":"http://arxiv.org/abs/2501.02007v1","authors":["Yannis Y. He"]},{"title":"Cross-model Transferability among Large Language Models on the Platonic\n Representations of Concepts","abstract":" Understanding the inner workings of Large Language Models (LLMs) is a\ncritical research frontier. Prior research has shown that a single LLM's\nconcept representations can be captured as steering vectors (SVs), enabling the\ncontrol of LLM behavior (e.g., towards generating harmful content). Our work\ntakes a novel approach by exploring the intricate relationships between concept\nrepresentations across different LLMs, drawing an intriguing parallel to\nPlato's Allegory of the Cave. In particular, we introduce a linear\ntransformation method to bridge these representations and present three key\nfindings: 1) Concept representations across different LLMs can be effectively\naligned using simple linear transformations, enabling efficient cross-model\ntransfer and behavioral control via SVs. 2) This linear transformation\ngeneralizes across concepts, facilitating alignment and control of SVs\nrepresenting different concepts across LLMs. 3) A weak-to-strong\ntransferability exists between LLM concept representations, whereby SVs\nextracted from smaller LLMs can effectively control the behavior of larger\nLLMs.\n","arxiv_id":"http://arxiv.org/abs/2501.02009v1","authors":["Youcheng Huang","Chen Huang","Duanyu Feng","Wenqiang Lei","Jiancheng Lv"]},{"title":"Machine Learning-Based Differential Diagnosis of Parkinson's Disease\n Using Kinematic Feature Extraction and Selection","abstract":" Parkinson's disease (PD), the second most common neurodegenerative disorder,\nis characterized by dopaminergic neuron loss and the accumulation of abnormal\nsynuclein. PD presents both motor and non-motor symptoms that progressively\nimpair daily functioning. The severity of these symptoms is typically assessed\nusing the MDS-UPDRS rating scale, which is subjective and dependent on the\nphysician's experience. Additionally, PD shares symptoms with other\nneurodegenerative diseases, such as progressive supranuclear palsy (PSP) and\nmultiple system atrophy (MSA), complicating accurate diagnosis. To address\nthese diagnostic challenges, we propose a machine learning-based system for\ndifferential diagnosis of PD, PSP, MSA, and healthy controls (HC). This system\nutilizes a kinematic feature-based hierarchical feature extraction and\nselection approach. Initially, 18 kinematic features are extracted, including\ntwo newly proposed features: Thumb-to-index vector velocity and acceleration,\nwhich provide insights into motor control patterns. In addition, 41 statistical\nfeatures were extracted here from each kinematic feature, including some new\napproaches such as Average Absolute Change, Rhythm, Amplitude, Frequency,\nStandard Deviation of Frequency, and Slope. Feature selection is performed\nusing One-way ANOVA to rank features, followed by Sequential Forward Floating\nSelection (SFFS) to identify the most relevant ones, aiming to reduce the\ncomputational complexity. The final feature set is used for classification,\nachieving a classification accuracy of 66.67% for each dataset and 88.89% for\neach patient, with particularly high performance for the MSA and HC groups\nusing the SVM algorithm. This system shows potential as a rapid and accurate\ndiagnostic tool in clinical practice, though further data collection and\nrefinement are needed to enhance its reliability.\n","arxiv_id":"http://arxiv.org/abs/2501.02014v1","authors":["Masahiro Matsumoto","Abu Saleh Musa Miah","Nobuyoshi Asai","Jungpil Shin"]},{"title":"Enhancing Uncertainty Modeling with Semantic Graph for Hallucination\n Detection","abstract":" Large Language Models (LLMs) are prone to hallucination with non-factual or\nunfaithful statements, which undermines the applications in real-world\nscenarios. Recent researches focus on uncertainty-based hallucination\ndetection, which utilizes the output probability of LLMs for uncertainty\ncalculation and does not rely on external knowledge or frequent sampling from\nLLMs. Whereas, most approaches merely consider the uncertainty of each\nindependent token, while the intricate semantic relations among tokens and\nsentences are not well studied, which limits the detection of hallucination\nthat spans over multiple tokens and sentences in the passage. In this paper, we\npropose a method to enhance uncertainty modeling with semantic graph for\nhallucination detection. Specifically, we first construct a semantic graph that\nwell captures the relations among entity tokens and sentences. Then, we\nincorporate the relations between two entities for uncertainty propagation to\nenhance sentence-level hallucination detection. Given that hallucination occurs\ndue to the conflict between sentences, we further present a graph-based\nuncertainty calibration method that integrates the contradiction probability of\nthe sentence with its neighbors in the semantic graph for uncertainty\ncalculation. Extensive experiments on two datasets show the great advantages of\nour proposed approach. In particular, we obtain substantial improvements with\n19.78% in passage-level hallucination detection.\n","arxiv_id":"http://arxiv.org/abs/2501.02020v1","authors":["Kedi Chen","Qin Chen","Jie Zhou","Xinqi Tao","Bowen Ding","Jingwen Xie","Mingchen Xie","Peilong Li","Feng Zheng","Liang He"]},{"title":"Weakly Supervised Learning on Large Graphs","abstract":" Graph classification plays a pivotal role in various domains, including\npathology, where images can be represented as graphs.In this domain, images can\nbe represented as graphs, where nodes might represent individual nuclei, and\nedges capture the spatial or functional relationships between them. Often, the\noverall label of the graph, such as a cancer type or disease state, is\ndetermined by patterns within smaller, localized regions of the image. This\nwork introduces a weakly-supervised graph classification framework leveraging\ntwo subgraph extraction techniques: (1) Sliding-window approach (2) BFS-based\napproach. Subgraphs are processed using a Graph Attention Network (GAT), which\nemploys attention mechanisms to identify the most informative subgraphs for\nclassification. Weak supervision is achieved by propagating graph-level labels\nto subgraphs, eliminating the need for detailed subgraph annotations.\n","arxiv_id":"http://arxiv.org/abs/2501.02021v1","authors":["Aditya Prakash"]},{"title":"CarbonChat: Large Language Model-Based Corporate Carbon Emission\n Analysis and Climate Knowledge Q\u0026A System","abstract":" As the impact of global climate change intensifies, corporate carbon\nemissions have become a focal point of global attention. In response to issues\nsuch as the lag in climate change knowledge updates within large language\nmodels, the lack of specialization and accuracy in traditional augmented\ngeneration architectures for complex problems, and the high cost and time\nconsumption of sustainability report analysis, this paper proposes CarbonChat:\nLarge Language Model-based corporate carbon emission analysis and climate\nknowledge Q\u0026A system, aimed at achieving precise carbon emission analysis and\npolicy understanding.First, a diversified index module construction method is\nproposed to handle the segmentation of rule-based and long-text documents, as\nwell as the extraction of structured data, thereby optimizing the parsing of\nkey information.Second, an enhanced self-prompt retrieval-augmented generation\narchitecture is designed, integrating intent recognition, structured reasoning\nchains, hybrid retrieval, and Text2SQL, improving the efficiency of semantic\nunderstanding and query conversion.Next, based on the greenhouse gas accounting\nframework, 14 dimensions are established for carbon emission analysis, enabling\nreport summarization, relevance evaluation, and customized responses.Finally,\nthrough a multi-layer chunking mechanism, timestamps, and hallucination\ndetection features, the accuracy and verifiability of the analysis results are\nensured, reducing hallucination rates and enhancing the precision of the\nresponses.\n","arxiv_id":"http://arxiv.org/abs/2501.02031v1","authors":["Zhixuan Cao","Ming Han","Jingtao Wang","Meng Jia"]},{"title":"3D Cloud reconstruction through geospatially-aware Masked Autoencoders","abstract":" Clouds play a key role in Earth's radiation balance with complex effects that\nintroduce large uncertainties into climate models. Real-time 3D cloud data is\nessential for improving climate predictions. This study leverages geostationary\nimagery from MSG/SEVIRI and radar reflectivity measurements of cloud profiles\nfrom CloudSat/CPR to reconstruct 3D cloud structures. We first apply\nself-supervised learning (SSL) methods-Masked Autoencoders (MAE) and\ngeospatially-aware SatMAE on unlabelled MSG images, and then fine-tune our\nmodels on matched image-profile pairs. Our approach outperforms\nstate-of-the-art methods like U-Nets, and our geospatial encoding further\nimproves prediction results, demonstrating the potential of SSL for cloud\nreconstruction.\n","arxiv_id":"http://arxiv.org/abs/2501.02035v1","authors":["Stella Girtsou","Emiliano Diaz Salas-Porras","Lilli Freischem","Joppe Massant","Kyriaki-Margarita Bintsi","Guiseppe Castiglione","William Jones","Michael Eisinger","Emmanuel Johnson","Anna Jungbluth"]},{"title":"Architecture for Trajectory-Based Fishing Ship Classification with AIS\n Data","abstract":" This paper proposes a data preparation process for managing real-world\nkinematic data and detecting fishing vessels. The solution is a binary\nclassification that classifies ship trajectories into either fishing or\nnon-fishing ships. The data used are characterized by the typical problems\nfound in classic data mining applications using real-world data, such as noise\nand inconsistencies. The two classes are also clearly unbalanced in the data, a\nproblem which is addressed using algorithms that resample the instances. For\nclassification, a series of features are extracted from spatiotemporal data\nthat represent the trajectories of the ships, available from sequences of\nAutomatic Identification System (AIS) reports. These features are proposed for\nthe modelling of ship behavior but, because they do not contain context-related\ninformation, the classification can be applied in other scenarios.\nExperimentation shows that the proposed data preparation process is useful for\nthe presented classification problem. In addition, positive results are\nobtained using minimal information.\n","arxiv_id":"http://arxiv.org/abs/2501.02038v1","authors":["David Sánchez Pedroche","Daniel Amigo","Jesús García","Jose M. Molina"]},{"title":"An Investigation into Value Misalignment in LLM-Generated Texts for\n Cultural Heritage","abstract":" As Large Language Models (LLMs) become increasingly prevalent in tasks\nrelated to cultural heritage, such as generating descriptions of historical\nmonuments, translating ancient texts, preserving oral traditions, and creating\neducational content, their ability to produce accurate and culturally aligned\ntexts is being increasingly relied upon by users and researchers. However,\ncultural value misalignments may exist in generated texts, such as the\nmisrepresentation of historical facts, the erosion of cultural identity, and\nthe oversimplification of complex cultural narratives, which may lead to severe\nconsequences. Therefore, investigating value misalignment in the context of LLM\nfor cultural heritage is crucial for mitigating these risks, yet there has been\na significant lack of systematic and comprehensive study and investigation in\nthis area. To fill this gap, we systematically assess the reliability of LLMs\nin generating culturally aligned texts for cultural heritage-related tasks. We\nconduct a comprehensive evaluation by compiling an extensive set of 1066 query\ntasks covering 5 widely recognized categories with 17 aspects within the\nknowledge framework of cultural heritage across 5 open-source LLMs, and examine\nboth the type and rate of cultural value misalignments in the generated texts.\nUsing both automated and manual approaches, we effectively detect and analyze\nthe cultural value misalignments in LLM-generated texts. Our findings are\nconcerning: over 65% of the generated texts exhibit notable cultural\nmisalignments, with certain tasks demonstrating almost complete misalignment\nwith key cultural values. Beyond these findings, this paper introduces a\nbenchmark dataset and a comprehensive evaluation workflow that can serve as a\nvaluable resource for future research aimed at enhancing the cultural\nsensitivity and reliability of LLMs.\n","arxiv_id":"http://arxiv.org/abs/2501.02039v1","authors":["Fan Bu","Zheng Wang","Siyi Wang","Ziyao Liu"]},{"title":"A Separable Self-attention Inspired by the State Space Model for\n Computer Vision","abstract":" Mamba is an efficient State Space Model (SSM) with linear computational\ncomplexity. Although SSMs are not suitable for handling non-causal data, Vision\nMamba (ViM) methods still demonstrate good performance in tasks such as image\nclassification and object detection. Recent studies have shown that there is a\nrich theoretical connection between state space models and attention variants.\nWe propose a novel separable self attention method, for the first time\nintroducing some excellent design concepts of Mamba into separable\nself-attention. To ensure a fair comparison with ViMs, we introduce VMINet, a\nsimple yet powerful prototype architecture, constructed solely by stacking our\nnovel attention modules with the most basic down-sampling layers. Notably,\nVMINet differs significantly from the conventional Transformer architecture.\nOur experiments demonstrate that VMINet has achieved competitive results on\nimage classification and high-resolution dense prediction tasks.Code is\navailable at: \\url{https://github.com/yws-wxs/VMINet}.\n","arxiv_id":"http://arxiv.org/abs/2501.02040v1","authors":["Juntao Zhang","Shaogeng Liu","Kun Bian","You Zhou","Pei Zhang","Jianning Liu","Jun Zhou","Bingyan Liu"]},{"title":"MRG: A Multi-Robot Manufacturing Digital Scene Generation Method Using\n Multi-Instance Point Cloud Registration","abstract":" A high-fidelity digital simulation environment is crucial for accurately\nreplicating physical operational processes. However, inconsistencies between\nsimulation and physical environments result in low confidence in simulation\noutcomes, limiting their effectiveness in guiding real-world production. Unlike\nthe traditional step-by-step point cloud \"segmentation-registration\" generation\nmethod, this paper introduces, for the first time, a novel Multi-Robot\nManufacturing Digital Scene Generation (MRG) method that leverages\nmulti-instance point cloud registration, specifically within manufacturing\nscenes. Tailored to the characteristics of industrial robots and manufacturing\nsettings, an instance-focused transformer module is developed to delineate\ninstance boundaries and capture correlations between local regions.\nAdditionally, a hypothesis generation module is proposed to extract target\ninstances while preserving key features. Finally, an efficient screening and\noptimization algorithm is designed to refine the final registration results.\nExperimental evaluations on the Scan2CAD and Welding-Station datasets\ndemonstrate that: (1) the proposed method outperforms existing multi-instance\npoint cloud registration techniques; (2) compared to state-of-the-art methods,\nthe Scan2CAD dataset achieves improvements in MR and MP by 12.15% and 17.79%,\nrespectively; and (3) on the Welding-Station dataset, MR and MP are enhanced by\n16.95% and 24.15%, respectively. This work marks the first application of\nmulti-instance point cloud registration in manufacturing scenes, significantly\nadvancing the precision and reliability of digital simulation environments for\nindustrial applications.\n","arxiv_id":"http://arxiv.org/abs/2501.02041v1","authors":["Songjie Han","Yinhua Liu","Yanzheng Li","Hua Chen","Dongmei Yang"]},{"title":"Advancing Pancreatic Cancer Prediction with a Next Visit Token\n Prediction Head on top of Med-BERT","abstract":" Background: Recently, numerous foundation models pretrained on extensive data\nhave demonstrated efficacy in disease prediction using Electronic Health\nRecords (EHRs). However, there remains some unanswered questions on how to best\nutilize such models especially with very small fine-tuning cohorts. Methods: We\nutilized Med-BERT, an EHR-specific foundation model, and reformulated the\ndisease binary prediction task into a token prediction task and a next visit\nmask token prediction task to align with Med-BERT's pretraining task format in\norder to improve the accuracy of pancreatic cancer (PaCa) prediction in both\nfew-shot and fully supervised settings. Results: The reformulation of the task\ninto a token prediction task, referred to as Med-BERT-Sum, demonstrates\nslightly superior performance in both few-shot scenarios and larger data\nsamples. Furthermore, reformulating the prediction task as a Next Visit Mask\nToken Prediction task (Med-BERT-Mask) significantly outperforms the\nconventional Binary Classification (BC) prediction task (Med-BERT-BC) by 3% to\n7% in few-shot scenarios with data sizes ranging from 10 to 500 samples. These\nfindings highlight that aligning the downstream task with Med-BERT's\npretraining objectives substantially enhances the model's predictive\ncapabilities, thereby improving its effectiveness in predicting both rare and\ncommon diseases. Conclusion: Reformatting disease prediction tasks to align\nwith the pretraining of foundation models enhances prediction accuracy, leading\nto earlier detection and timely intervention. This approach improves treatment\neffectiveness, survival rates, and overall patient outcomes for PaCa and\npotentially other cancers.\n","arxiv_id":"http://arxiv.org/abs/2501.02044v1","authors":["Jianping He","Laila Rasmy","Degui Zhi","Cui Tao"]},{"title":"ArtCrafter: Text-Image Aligning Style Transfer via Embedding Reframing","abstract":" Recent years have witnessed significant advancements in text-guided style\ntransfer, primarily attributed to innovations in diffusion models. These models\nexcel in conditional guidance, utilizing text or images to direct the sampling\nprocess. However, despite their capabilities, direct conditional guidance\napproaches often face challenges in balancing the expressiveness of textual\nsemantics with the diversity of output results while capturing stylistic\nfeatures. To address these challenges, we introduce ArtCrafter, a novel\nframework for text-to-image style transfer. Specifically, we introduce an\nattention-based style extraction module, meticulously engineered to capture the\nsubtle stylistic elements within an image. This module features a multi-layer\narchitecture that leverages the capabilities of perceiver attention mechanisms\nto integrate fine-grained information. Additionally, we present a novel\ntext-image aligning augmentation component that adeptly balances control over\nboth modalities, enabling the model to efficiently map image and text\nembeddings into a shared feature space. We achieve this through attention\noperations that enable smooth information flow between modalities. Lastly, we\nincorporate an explicit modulation that seamlessly blends multimodal enhanced\nembeddings with original embeddings through an embedding reframing design,\nempowering the model to generate diverse outputs. Extensive experiments\ndemonstrate that ArtCrafter yields impressive results in visual stylization,\nexhibiting exceptional levels of stylistic intensity, controllability, and\ndiversity.\n","arxiv_id":"http://arxiv.org/abs/2501.02064v1","authors":["Nisha Huang","Kaer Huang","Yifan Pu","Jiangshan Wang","Jie Guo","Yiqiang Yan","Xiu Li"]},{"title":"The interplay between domain specialization and model size: a case study\n in the legal domain","abstract":" Scaling laws for language models so far focused on finding the\ncompute-optimal model size and token count for training from scratch. However,\nachieving this optimal balance requires significant compute resources due to\nthe extensive data demands when training models from randomly-initialized\nweights. Continual pre-training offers a cost-effective alternative, leveraging\nthe compute investment from pre-trained models to incorporate new knowledge\nwithout requiring extensive new data. Recent findings suggest that data quality\ninfluences constants in scaling laws, thereby altering the optimal\nparameter-token allocation ratio. Building on this insight, we investigate the\ninterplay between domain specialization and model size during continual\npre-training under compute-constrained scenarios. Our goal is to identify a\ncompute-efficient training regime for this scenario and, potentially, detect\npatterns in this interplay that can be generalized across different model sizes\nand domains. To compare general and specialized training, we filtered a\nweb-based dataset to extract legal domain data. We pre-trained models with\n1.5B, 3B, 7B and 14B parameters on both the unfiltered and filtered datasets,\nthen evaluated their performance on legal exams. Results show that as model\nsize increases, the compute-effectiveness gap between specialized and general\nmodels widens.\n","arxiv_id":"http://arxiv.org/abs/2501.02068v1","authors":["Roseval Malaquias Junior","Ramon Pires","Thales Sales Almeida","Kenzo Sakiyama","Roseli Romero","Rodrigo Nogueira"]},{"title":"On the Statistical Complexity for Offline and Low-Adaptive Reinforcement\n Learning with Structures","abstract":" This article reviews the recent advances on the statistical foundation of\nreinforcement learning (RL) in the offline and low-adaptive settings. We will\nstart by arguing why offline RL is the appropriate model for almost any\nreal-life ML problems, even if they have nothing to do with the recent AI\nbreakthroughs that use RL. Then we will zoom into two fundamental problems of\noffline RL: offline policy evaluation (OPE) and offline policy learning (OPL).\nIt may be surprising to people that tight bounds for these problems were not\nknown even for tabular and linear cases until recently. We delineate the\ndifferences between worst-case minimax bounds and instance-dependent bounds. We\nalso cover key algorithmic ideas and proof techniques behind near-optimal\ninstance-dependent methods in OPE and OPL. Finally, we discuss the limitations\nof offline RL and review a burgeoning problem of \\emph{low-adaptive\nexploration} which addresses these limitations by providing a sweet middle\nground between offline and online RL.\n","arxiv_id":"http://arxiv.org/abs/2501.02089v1","authors":["Ming Yin","Mengdi Wang","Yu-Xiang Wang"]},{"title":"Online Detection of Water Contamination Under Concept Drift","abstract":" Water Distribution Networks (WDNs) are vital infrastructures, and\ncontamination poses serious public health risks. Harmful substances can\ninteract with disinfectants like chlorine, making chlorine monitoring essential\nfor detecting contaminants. However, chlorine sensors often become unreliable\nand require frequent calibration. This study introduces the Dual-Threshold\nAnomaly and Drift Detection (AD\u0026DD) method, an unsupervised approach combining\na dual-threshold drift detection mechanism with an LSTM-based Variational\nAutoencoder(LSTM-VAE) for real-time contamination detection. Tested on two\nrealistic WDNs, AD\u0026DD effectively identifies anomalies with sensor offsets as\nconcept drift, and outperforms other methods. A proposed decentralized\narchitecture enables accurate contamination detection and localization by\ndeploying AD\u0026DD on selected nodes.\n","arxiv_id":"http://arxiv.org/abs/2501.02107v1","authors":["Jin Li","Kleanthis Malialis","Stelios G. Vrachimis","Marios M. Polycarpou"]},{"title":"Siamese Networks for Cat Re-Identification: Exploring Neural Models for\n Cat Instance Recognition","abstract":" Street cats in urban areas often rely on human intervention for survival,\nleading to challenges in population control and welfare management. In April\n2023, Hello Inc., a Chinese urban mobility company, launched the Hello Street\nCat initiative to address these issues. The project deployed over 21,000 smart\nfeeding stations across 14 cities in China, integrating livestreaming cameras\nand treat dispensers activated through user donations. It also promotes the\nTrap-Neuter-Return (TNR) method, supported by a community-driven platform,\nHelloStreetCatWiki, where volunteers catalog and identify cats. However, manual\nidentification is inefficient and unsustainable, creating a need for automated\nsolutions. This study explores Deep Learning-based models for re-identifying\nstreet cats in the Hello Street Cat initiative. A dataset of 2,796 images of 69\ncats was used to train Siamese Networks with EfficientNetB0, MobileNet and\nVGG16 as base models, evaluated under contrastive and triplet loss functions.\nVGG16 paired with contrastive loss emerged as the most effective configuration,\nachieving up to 97% accuracy and an F1 score of 0.9344 during testing. The\napproach leverages image augmentation and dataset refinement to overcome\nchallenges posed by limited data and diverse visual variations. These findings\nunderscore the potential of automated cat re-identification to streamline\npopulation monitoring and welfare efforts. By reducing reliance on manual\nprocesses, the method offers a scalable and reliable solution for\ncommunitydriven initiatives. Future research will focus on expanding datasets\nand developing real-time implementations to enhance practicality in large-scale\ndeployments.\n","arxiv_id":"http://arxiv.org/abs/2501.02112v1","authors":["Tobias Trein","Luan Fonseca Garcia"]},{"title":"AVTrustBench: Assessing and Enhancing Reliability and Robustness in\n Audio-Visual LLMs","abstract":" With the rapid advancement of Multi-modal Large Language Models (MLLMs),\nseveral diagnostic benchmarks have recently been developed to assess these\nmodels' multi-modal reasoning proficiency. However, these benchmarks are\nrestricted to assessing primarily the visual aspect and do not examine the\nholistic audio-visual (AV) understanding. Moreover, currently, there are no\nbenchmarks that investigate the capabilities of AVLLMs to calibrate their\nresponses when presented with perturbed inputs. To this end, we introduce\nAudio-Visual Trustworthiness assessment Benchmark (AVTrustBench), comprising\n600K samples spanning over 9 meticulously crafted tasks, evaluating the\ncapabilities of AVLLMs across three distinct dimensions: Adversarial attack,\nCompositional reasoning, and Modality-specific dependency. Using our benchmark\nwe extensively evaluate 13 state-of-the-art AVLLMs. The findings reveal that\nthe majority of existing models fall significantly short of achieving\nhuman-like comprehension, offering valuable insights for future research\ndirections. To alleviate the limitations in the existing approaches, we further\npropose a robust, model-agnostic calibrated audio-visual preference\noptimization based training strategy CAVPref, obtaining a gain up to 30.19%\nacross all 9 tasks. We will publicly release our code and benchmark to\nfacilitate future research in this direction.\n","arxiv_id":"http://arxiv.org/abs/2501.02135v1","authors":["Sanjoy Chowdhury","Sayan Nag","Subhrajyoti Dasgupta","Yaoting Wang","Mohamed Elhoseiny","Ruohan Gao","Dinesh Manocha"]},{"title":"Attribute-Based Robotic Grasping with Data-Efficient Adaptation","abstract":" Robotic grasping is one of the most fundamental robotic manipulation tasks\nand has been the subject of extensive research. However, swiftly teaching a\nrobot to grasp a novel target object in clutter remains challenging. This paper\nattempts to address the challenge by leveraging object attributes that\nfacilitate recognition, grasping, and rapid adaptation to new domains. In this\nwork, we present an end-to-end encoder-decoder network to learn attribute-based\nrobotic grasping with data-efficient adaptation capability. We first pre-train\nthe end-to-end model with a variety of basic objects to learn generic attribute\nrepresentation for recognition and grasping. Our approach fuses the embeddings\nof a workspace image and a query text using a gated-attention mechanism and\nlearns to predict instance grasping affordances. To train the joint embedding\nspace of visual and textual attributes, the robot utilizes object persistence\nbefore and after grasping. Our model is self-supervised in a simulation that\nonly uses basic objects of various colors and shapes but generalizes to novel\nobjects in new environments. To further facilitate generalization, we propose\ntwo adaptation methods, adversarial adaption and one-grasp adaptation.\nAdversarial adaptation regulates the image encoder using augmented data of\nunlabeled images, whereas one-grasp adaptation updates the overall end-to-end\nmodel using augmented data from one grasp trial. Both adaptation methods are\ndata-efficient and considerably improve instance grasping performance.\nExperimental results in both simulation and the real world demonstrate that our\napproach achieves over 81% instance grasping success rate on unknown objects,\nwhich outperforms several baselines by large margins.\n","arxiv_id":"http://arxiv.org/abs/2501.02149v1","authors":["Yang Yang","Houjian Yu","Xibai Lou","Yuanhao Liu","Changhyun Choi"]},{"title":"Table as Thought: Exploring Structured Thoughts in LLM Reasoning","abstract":" Large language models' reasoning abilities benefit from methods that organize\ntheir thought processes, such as chain-of-thought prompting, which employs a\nsequential structure to guide the reasoning process step-by-step. However,\nexisting approaches focus primarily on organizing the sequence of thoughts,\nleaving structure in individual thought steps underexplored. To address this\ngap, we propose Table as Thought, a framework inspired by cognitive\nneuroscience theories on human thought. Table as Thought organizes reasoning\nwithin a tabular schema, where rows represent sequential thought steps and\ncolumns capture critical constraints and contextual information to enhance\nreasoning. The reasoning process iteratively populates the table until\nself-verification ensures completeness and correctness. Our experiments show\nthat Table as Thought excels in planning tasks and demonstrates a strong\npotential for enhancing LLM performance in mathematical reasoning compared to\nunstructured thought baselines. This work provides a novel exploration of\nrefining thought representation within LLMs, paving the way for advancements in\nreasoning and AI cognition.\n","arxiv_id":"http://arxiv.org/abs/2501.02152v1","authors":["Zhenjie Sun","Naihao Deng","Haofei Yu","Jiaxuan You"]},{"title":"The Integration of Blockchain and Artificial Intelligence for Secure\n Healthcare Systems","abstract":" Verisign reported a 125 percent increase in data breaches within the\nhealthcare sector in the United States during 2022, with 18.2 million patient\nrecords being impacted. Growing healthcare data volumes and diversification\nmean that medical information is becoming more valuable. Many Health Centers\nuse various technologies to ease the classification, storage, and exchange of\nbig data. This use can also make the health data of the users at risk and\nvulnerable. AI and blockchain are among the leading technologies at hand. With\nAI, data-driven operations and big data efficiency have been improved with\nrespect to traditional techniques. Due to its potential to bring about\nimprovements in health services and lower medical costs, this AI technology is\nregularly used in healthcare. Blockchain helps protect transactions on sharing\ninformation and private privacy as long as the exchange of knowledge is that of\nthe standard. The objective of this analysis is to investigate the research and\nunique contributions since 2008 regarding blockchain-integrated AI and\nhealthcare systems. The work sheds light on applied AI-based healthcare schemes\nwith machine, ballistic, and acrylic learning and disparate blockchain\nstructures. The use of technology in order to ensure patient data security and\nmanage medical information effectively in healthcare settings offers a highly\nsuccessful position for both healthcare providers and patients. From 2018 to\n2021, the best year was 2021 to grow, enhancing everything to examine the\ndownload of the device and the counting of Google Academies, for which the\njoining perspective was borrowed; local research experts were asked, identified\narticles in recent years, and read reviews of large research grants.\n","arxiv_id":"http://arxiv.org/abs/2501.02169v1","authors":["Umar Safdar","Simon Gabrael"]},{"title":"AdaMixup: A Dynamic Defense Framework for Membership Inference Attack\n Mitigation","abstract":" Membership inference attacks have emerged as a significant privacy concern in\nthe training of deep learning models, where attackers can infer whether a data\npoint was part of the training set based on the model's outputs. To address\nthis challenge, we propose a novel defense mechanism, AdaMixup. AdaMixup\nemploys adaptive mixup techniques to enhance the model's robustness against\nmembership inference attacks by dynamically adjusting the mixup strategy during\ntraining. This method not only improves the model's privacy protection but also\nmaintains high performance. Experimental results across multiple datasets\ndemonstrate that AdaMixup significantly reduces the risk of membership\ninference attacks while achieving a favorable trade-off between defensive\nefficiency and model accuracy. This research provides an effective solution for\ndata privacy protection and lays the groundwork for future advancements in\nmixup training methods.\n","arxiv_id":"http://arxiv.org/abs/2501.02182v1","authors":["Ying Chen","Jiajing Chen","Yijie Weng","ChiaHua Chang","Dezhi Yu","Guanbiao Lin"]},{"title":"CPTuning: Contrastive Prompt Tuning for Generative Relation Extraction","abstract":" Generative relation extraction (RE) commonly involves first reformulating RE\nas a linguistic modeling problem easily tackled with pre-trained language\nmodels (PLM) and then fine-tuning a PLM with supervised cross-entropy loss.\nAlthough having achieved promising performance, existing approaches assume only\none deterministic relation between each pair of entities without considering\nreal scenarios where multiple relations may be valid, i.e., entity pair\noverlap, causing their limited applications. To address this problem, we\nintroduce a novel contrastive prompt tuning method for RE, CPTuning, which\nlearns to associate a candidate relation between two in-context entities with a\nprobability mass above or below a threshold, corresponding to whether the\nrelation exists. Beyond learning schema, CPTuning also organizes RE as a\nverbalized relation generation task and uses Trie-constrained decoding to\nensure a model generates valid relations. It adaptively picks out the generated\ncandidate relations with a high estimated likelihood in inference, thereby\nachieving multi-relation extraction. We conduct extensive experiments on four\nwidely used datasets to validate our method. Results show that T5-large\nfine-tuned with CPTuning significantly outperforms previous methods, regardless\nof single or multiple relations extraction.\n","arxiv_id":"http://arxiv.org/abs/2501.02196v1","authors":["Jiaxin Duan","Fengyu Lu","Junfei Liu"]},{"title":"Financial Named Entity Recognition: How Far Can LLM Go?","abstract":" The surge of large language models (LLMs) has revolutionized the extraction\nand analysis of crucial information from a growing volume of financial\nstatements, announcements, and business news. Recognition for named entities to\nconstruct structured data poses a significant challenge in analyzing financial\ndocuments and is a foundational task for intelligent financial analytics.\nHowever, how effective are these generic LLMs and their performance under\nvarious prompts are yet need a better understanding. To fill in the blank, we\npresent a systematic evaluation of state-of-the-art LLMs and prompting methods\nin the financial Named Entity Recognition (NER) problem. Specifically, our\nexperimental results highlight their strengths and limitations, identify five\nrepresentative failure types, and provide insights into their potential and\nchallenges for domain-specific tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.02237v1","authors":["Yi-Te Lu","Yintong Huo"]},{"title":"Interpretable Load Forecasting via Representation Learning of\n Geo-distributed Meteorological Factors","abstract":" Meteorological factors (MF) are crucial in day-ahead load forecasting as they\nsignificantly influence the electricity consumption behaviors of consumers.\nNumerous studies have incorporated MF into the load forecasting model to\nachieve higher accuracy. Selecting MF from one representative location or the\naveraged MF as the inputs of the forecasting model is a common practice.\nHowever, the difference in MF collected in various locations within a region\nmay be significant, which poses a challenge in selecting the appropriate MF\nfrom numerous locations. A representation learning framework is proposed to\nextract geo-distributed MF while considering their spatial relationships. In\naddition, this paper employs the Shapley value in the graph-based model to\nreveal connections between MF collected in different locations and loads. To\nreduce the computational complexity of calculating the Shapley value, an\nacceleration method is adopted based on Monte Carlo sampling and weighted\nlinear regression. Experiments on two real-world datasets demonstrate that the\nproposed method improves the day-ahead forecasting accuracy, especially in\nextreme scenarios such as the \"accumulation temperature effect\" in summer and\n\"sudden temperature change\" in winter. We also find a significant correlation\nbetween the importance of MF in different locations and the corresponding\narea's GDP and mainstay industry.\n","arxiv_id":"http://arxiv.org/abs/2501.02241v1","authors":["Yangze Zhou","Guoxin Lin","Gonghao Zhang","Yi Wang"]},{"title":"LLMzSzŁ: a comprehensive LLM benchmark for Polish","abstract":" This article introduces the first comprehensive benchmark for the Polish\nlanguage at this scale: LLMzSz{\\L} (LLMs Behind the School Desk). It is based\non a coherent collection of Polish national exams, including both academic and\nprofessional tests extracted from the archives of the Polish Central\nExamination Board. It covers 4 types of exams, coming from 154 domains.\nAltogether, it consists of almost 19k closed-ended questions. We investigate\nthe performance of open-source multilingual, English, and Polish LLMs to verify\nLLMs' abilities to transfer knowledge between languages. Also, the correlation\nbetween LLMs and humans at model accuracy and exam pass rate levels is\nexamined. We show that multilingual LLMs can obtain superior results over\nmonolingual ones; however, monolingual models may be beneficial when model size\nmatters. Our analysis highlights the potential of LLMs in assisting with exam\nvalidation, particularly in identifying anomalies or errors in examination\ntasks.\n","arxiv_id":"http://arxiv.org/abs/2501.02266v1","authors":["Krzysztof Jassem","Michał Ciesiółka","Filip Graliński","Piotr Jabłoński","Jakub Pokrywka","Marek Kubis","Monika Jabłońska","Ryszard Staruch"]},{"title":"What Kind of Visual Tokens Do We Need? Training-free Visual Token\n Pruning for Multi-modal Large Language Models from the Perspective of Graph","abstract":" Recent Multimodal Large Language Models(MLLMs) often use a large number of\nvisual tokens to compensate their visual shortcoming, leading to excessive\ncomputation and obvious visual redundancy. In this paper, we investigate what\nkind of visual tokens are needed for MLLMs, and reveal that both foreground and\nbackground tokens are critical for MLLMs given the varying difficulties of\nexamples. Based on this observation, we propose a graph-based method towards\ntraining-free visual token pruning, termed G-Prune.In particular, G-Prune\nregards visual tokens as nodes, and construct their connections based on their\nsemantic similarities. Afterwards, the information flow is propagated via\nweighted links, and the most important tokens after iterations are kept for\nMLLMs, which can be front or background.To validate G-Prune, we apply it to a\nrecent MLLM called LLaVA-NeXT, and conduct extensive experiments on a set of\nbenchmarks.The experiment results show that G-Prune can greatly reduce\ncomputation overhead while retaining high performance on both coarse- and\nfine-grained tasks. For instance, G-Prune can reduce 63.57\\% FLOPs of\nLLaVA-NeXT on VQA2.0 and TextVQA with only 0.95\\% and 2.34\\% accuracy drops,\nrespectively.\n","arxiv_id":"http://arxiv.org/abs/2501.02268v1","authors":["Yutao Jiang","Qiong Wu","Wenhao Lin","Wei Yu","Yiyi Zhou"]},{"title":"Hyperbolic Contrastive Learning for Hierarchical 3D Point Cloud\n Embedding","abstract":" Hyperbolic spaces allow for more efficient modeling of complex, hierarchical\nstructures, which is particularly beneficial in tasks involving multi-modal\ndata. Although hyperbolic geometries have been proven effective for\nlanguage-image pre-training, their capabilities to unify language, image, and\n3D Point Cloud modalities are under-explored. We extend the 3D Point Cloud\nmodality in hyperbolic multi-modal contrastive pre-training. Additionally, we\nexplore the entailment, modality gap, and alignment regularizers for learning\nhierarchical 3D embeddings and facilitating the transfer of knowledge from both\nText and Image modalities. These regularizers enable the learning of\nintra-modal hierarchy within each modality and inter-modal hierarchy across\ntext, 2D images, and 3D Point Clouds. Experimental results demonstrate that our\nproposed training strategy yields an outstanding 3D Point Cloud encoder, and\nthe obtained 3D Point Cloud hierarchical embeddings significantly improve\nperformance on various downstream tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.02285v2","authors":["Yingjie Liu","Pengyu Zhang","Ziyao He","Mingsong Chen","Xuan Tang","Xian Wei"]},{"title":"SR-Reward: Taking The Path More Traveled","abstract":" In this paper, we propose a novel method for learning reward functions\ndirectly from offline demonstrations. Unlike traditional inverse reinforcement\nlearning (IRL), our approach decouples the reward function from the learner's\npolicy, eliminating the adversarial interaction typically required between the\ntwo. This results in a more stable and efficient training process. Our reward\nfunction, called \\textit{SR-Reward}, leverages successor representation (SR) to\nencode a state based on expected future states' visitation under the\ndemonstration policy and transition dynamics. By utilizing the Bellman\nequation, SR-Reward can be learned concurrently with most reinforcement\nlearning (RL) algorithms without altering the existing training pipeline. We\nalso introduce a negative sampling strategy to mitigate overestimation errors\nby reducing rewards for out-of-distribution data, thereby enhancing robustness.\nThis strategy inherently introduces a conservative bias into RL algorithms that\nemploy the learned reward. We evaluate our method on the D4RL benchmark,\nachieving competitive results compared to offline RL algorithms with access to\ntrue rewards and imitation learning (IL) techniques like behavioral cloning.\nMoreover, our ablation studies on data size and quality reveal the advantages\nand limitations of SR-Reward as a proxy for true rewards.\n","arxiv_id":"http://arxiv.org/abs/2501.02330v1","authors":["Seyed Mahdi B. Azad","Zahra Padar","Gabriel Kalweit","Joschka Boedecker"]},{"title":"AdaSkip: Adaptive Sublayer Skipping for Accelerating Long-Context LLM\n Inference","abstract":" Long-context large language models (LLMs) inference is increasingly critical,\nmotivating a number of studies devoted to alleviating the substantial storage\nand computational costs in such scenarios. Layer-wise skipping methods are\npromising optimizations but rarely explored in long-context inference. We\nobserve that existing layer-wise skipping strategies have several limitations\nwhen applied in long-context inference, including the inability to adapt to\nmodel and context variability, disregard for sublayer significance, and\ninapplicability for the prefilling phase. This paper proposes \\sysname, an\nadaptive sublayer skipping method specifically designed for long-context\ninference. \\sysname adaptively identifies less important layers by leveraging\non-the-fly similarity information, enables sublayer-wise skipping, and\naccelerates both the prefilling and decoding phases. The effectiveness of\n\\sysname is demonstrated through extensive experiments on various long-context\nbenchmarks and models, showcasing its superior inference performance over\nexisting baselines.\n","arxiv_id":"http://arxiv.org/abs/2501.02336v1","authors":["Zhuomin He","Yizhen Yao","Pengfei Zuo","Bin Gao","Qinya Li","Zhenzhe Zheng","Fan Wu"]},{"title":"Evaluation of the Code Generation Capabilities of ChatGPT 4: A\n Comparative Analysis in 19 Programming Languages","abstract":" This bachelor's thesis examines the capabilities of ChatGPT 4 in code\ngeneration across 19 programming languages. The study analyzed solution rates\nacross three difficulty levels, types of errors encountered, and code quality\nin terms of runtime and memory efficiency through a quantitative experiment. A\ntotal of 188 programming problems were selected from the LeetCode platform, and\nChatGPT 4 was given three attempts to produce a correct solution with feedback.\nChatGPT 4 successfully solved 39.67% of all tasks, with success rates\ndecreasing significantly as problem complexity increased. Notably, the model\nfaced considerable challenges with hard problems across all languages. ChatGPT\n4 demonstrated higher competence in widely used languages, likely due to a\nlarger volume and higher quality of training data. The solution rates also\nrevealed a preference for languages with low abstraction levels and static\ntyping. For popular languages, the most frequent error was \"Wrong Answer,\"\nwhereas for less popular languages, compiler and runtime errors prevailed,\nsuggesting frequent misunderstandings and confusion regarding the structural\ncharacteristics of these languages. The model exhibited above-average runtime\nefficiency in all programming languages, showing a tendency toward statically\ntyped and low-abstraction languages. Memory efficiency results varied\nsignificantly, with above-average performance in 14 languages and below-average\nperformance in five languages. A slight preference for low-abstraction\nlanguages and a leaning toward dynamically typed languages in terms of memory\nefficiency were observed. Future research should include a larger number of\ntasks, iterations, and less popular languages. Additionally, ChatGPT 4's\nabilities in code interpretation and summarization, debugging, and the\ndevelopment of complex, practical code could be analyzed further.\n ----\n Diese Bachelorarbeit untersucht die F\\\"ahigkeiten von ChatGPT 4 zur\nCode-Generierung in 19 Programmiersprachen. Betrachtet wurden die\nL\\\"osungsraten zwischen drei Schwierigkeitsgraden, die aufgetretenen\nFehlerarten und die Qualit\\\"at des Codes hinsichtlich der Laufzeit- und\nSpeichereffizienz in einem quantitativen Experiment. Dabei wurden 188\nProgrammierprobleme der Plattform LeetCode entnommen, wobei ChatGPT 4 jeweils\ndrei Versuche hatte, mittels Feedback eine korrekte L\\\"osung zu generieren.\nChatGPT 4 l\\\"oste 39,67 % aller Aufgaben erfolgreich, wobei die Erfolgsrate mit\nzunehmendem Schwierigkeitsgrad deutlich abnahm und bei komplexen Problemen in\nallen Sprachen signifikante Schwierigkeiten auftraten. Das Modell zeigte eine\nh\\\"ohere Kompetenz in weit verbreiteten Sprachen, was wahrscheinlich auf eine\ngr\\\"o{\\ss}ere Menge und h\\\"ohere Qualit\\\"at der Trainingsdaten\nzur\\\"uckzuf\\\"uhren ist. Bez\\\"uglich der L\\\"osungsraten zeigte das Modell zudem\neine Pr\\\"aferenz f\\\"ur Sprachen mit niedrigem Abstraktionsniveau und statischer\nTypisierung. Bei Sprachen hoher Popularit\\\"at trat der Fehler Wrong Answer am\nh\\\"aufigsten auf, w\\\"ahrend bei weniger popul\\\"aren Sprachen Compiler- und\nLaufzeitfehler \\\"uberwogen, was auf h\\\"aufige Missverst\\\"andnisse und\nVerwechslungen bez\\\"uglich der spezifischen strukturellen Eigenschaften dieser\nSprachen zur\\\"uckzuf\\\"uhren ist. ChatGPT 4 demonstrierte in allen\nProgrammiersprachen eine \\\"uberdurchschnittliche Laufzeiteffizienz und\ntendierte diesbez\\\"uglich erneut zu statisch typisierten und niedrig\nabstrahierten Sprachen. Die Werte zur Speichereffizienz variierten erheblich,\nwobei in 14 Sprachen \\\"uberdurchschnittliche und in f\\\"unf Sprachen\nunterdurchschnittliche Werte erzielt wurden. Es zeigte sich diesbez\\\"uglich\neine leichte Tendenz zugunsten von niedrig abstrahierten sowie eine Pr\\\"aferenz\nzu dynamisch typisierten Sprachen. Zuk\\\"unftige Forschung sollte eine h\\\"ohere\nAnzahl an Aufgaben, Iterationen und unpopul\\\"aren Sprachen einbeziehen.\nDar\\\"uber hinaus k\\\"onnten die F\\\"ahigkeiten von ChatGPT 4 in der\nCode-Interpretation und -Zusammenfassung, im Debugging und in der Entwicklung\nkomplexer, praxisbezogener Codes analysiert werden.\n","arxiv_id":"http://arxiv.org/abs/2501.02338v1","authors":["L. C. Gilbert"]},{"title":"UAVs Meet LLMs: Overviews and Perspectives Toward Agentic Low-Altitude\n Mobility","abstract":" Low-altitude mobility, exemplified by unmanned aerial vehicles (UAVs), has\nintroduced transformative advancements across various domains, like\ntransportation, logistics, and agriculture. Leveraging flexible perspectives\nand rapid maneuverability, UAVs extend traditional systems' perception and\naction capabilities, garnering widespread attention from academia and industry.\nHowever, current UAV operations primarily depend on human control, with only\nlimited autonomy in simple scenarios, and lack the intelligence and\nadaptability needed for more complex environments and tasks. The emergence of\nlarge language models (LLMs) demonstrates remarkable problem-solving and\ngeneralization capabilities, offering a promising pathway for advancing UAV\nintelligence. This paper explores the integration of LLMs and UAVs, beginning\nwith an overview of UAV systems' fundamental components and functionalities,\nfollowed by an overview of the state-of-the-art in LLM technology.\nSubsequently, it systematically highlights the multimodal data resources\navailable for UAVs, which provide critical support for training and evaluation.\nFurthermore, it categorizes and analyzes key tasks and application scenarios\nwhere UAVs and LLMs converge. Finally, a reference roadmap towards agentic UAVs\nis proposed, aiming to enable UAVs to achieve agentic intelligence through\nautonomous perception, memory, reasoning, and tool utilization. Related\nresources are available at https://github.com/Hub-Tian/UAVs_Meet_LLMs.\n","arxiv_id":"http://arxiv.org/abs/2501.02341v1","authors":["Yonglin Tian","Fei Lin","Yiduo Li","Tengchao Zhang","Qiyao Zhang","Xuan Fu","Jun Huang","Xingyuan Dai","Yutong Wang","Chunwei Tian","Bai Li","Yisheng Lv","Levente Kovács","Fei-Yue Wang"]},{"title":"Exploring the Capabilities and Limitations of Large Language Models for\n Radiation Oncology Decision Support","abstract":" Thanks to the rapidly evolving integration of LLMs into decision-support\ntools, a significant transformation is happening across large-scale systems.\nLike other medical fields, the use of LLMs such as GPT-4 is gaining increasing\ninterest in radiation oncology as well. An attempt to assess GPT-4's\nperformance in radiation oncology was made via a dedicated 100-question\nexamination on the highly specialized topic of radiation oncology physics,\nrevealing GPT-4's superiority over other LLMs. GPT-4's performance on a broader\nfield of clinical radiation oncology is further benchmarked by the ACR\nRadiation Oncology In-Training (TXIT) exam where GPT-4 achieved a high accuracy\nof 74.57%. Its performance on re-labelling structure names in accordance with\nthe AAPM TG-263 report has also been benchmarked, achieving above 96%\naccuracies. Such studies shed light on the potential of LLMs in radiation\noncology. As interest in the potential and constraints of LLMs in general\nhealthcare applications continues to rise5, the capabilities and limitations of\nLLMs in radiation oncology decision support have not yet been fully explored.\n","arxiv_id":"http://arxiv.org/abs/2501.02346v1","authors":["Florian Putz","Marlen Haderleina","Sebastian Lettmaier","Sabine Semrau","Rainer Fietkau","Yixing Huang"]},{"title":"Context Aware Lemmatization and Morphological Tagging Method in Turkish","abstract":" The smallest part of a word that defines the word is called a word root. Word\nroots are used to increase success in many applications since they simplify the\nword. In this study, the lemmatization model, which is a word root finding\nmethod, and the morphological tagging model, which predicts the grammatical\nknowledge of the word, are presented. The presented model was developed for\nTurkish, and both models make predictions by taking the meaning of the word\ninto account. In the literature, there is no lemmatization study that is\nsensitive to word meaning in Turkish. For this reason, the present study shares\nthe model and the results obtained from the model on Turkish lemmatization for\nthe first time in the literature. In the present study, in the lemmatization\nand morphological tagging models, bidirectional LSTM is used for the spelling\nof words, and the Turkish BERT model is used for the meaning of words. The\nmodels are trained using the IMST and PUD datasets from Universal Dependencies.\nThe results from the training of the models were compared with the results from\nthe SIGMORPHON 2019 competition. The results of the comparisons revealed that\nour models were superior.\n","arxiv_id":"http://arxiv.org/abs/2501.02361v1","authors":["Cagri Sayallar"]},{"title":"Enhancing Workplace Productivity and Well-being Using AI Agent","abstract":" This paper discusses the use of Artificial Intelligence (AI) to enhance\nworkplace productivity and employee well-being. By integrating machine learning\n(ML) techniques with neurobiological data, the proposed approaches ensure\nalignment with human ethical standards through value alignment models and\nHierarchical Reinforcement Learning (HRL) for autonomous task management. The\nsystem utilizes biometric feedback from employees to generate personalized\nhealth prompts, fostering a supportive work environment that encourages\nphysical activity. Additionally, we explore decentralized multi-agent systems\nfor improved collaboration and decision-making frameworks that enhance\ntransparency. Various approaches using ML techniques in conjunction with AI\nimplementations are discussed. Together, these innovations aim to create a more\nproductive and health-conscious workplace. These outcomes assist HR management\nand organizations in launching more rational career progression streams for\nemployees and facilitating organizational transformation.\n","arxiv_id":"http://arxiv.org/abs/2501.02368v1","authors":["Ravirajan K","Arvind Sundarajan"]},{"title":"Syntactic Evolution in Language Usage","abstract":" This research aims to investigate the dynamic nature of linguistic style\nthroughout various stages of life, from post teenage to old age. By employing\nlinguistic analysis tools and methodologies, the study will delve into the\nintricacies of how individuals adapt and modify their language use over time.\nThe research uses a data set of blogs from blogger.com from 2004 and focuses on\nEnglish for syntactic analysis. The findings of this research can have\nimplications for linguistics, psychology, and communication studies, shedding\nlight on the intricate relationship between age and language.\n","arxiv_id":"http://arxiv.org/abs/2501.02392v1","authors":["Surbhit Kumar"]},{"title":"iTARGET: Interpretable Tailored Age Regression for Grouped Epigenetic\n Traits","abstract":" Accurately predicting chronological age from DNA methylation patterns is\ncrucial for advancing biological age estimation. However, this task is made\nchallenging by Epigenetic Correlation Drift (ECD) and Heterogeneity Among CpGs\n(HAC), which reflect the dynamic relationship between methylation and age\nacross different life stages. To address these issues, we propose a novel\ntwo-phase algorithm. The first phase employs similarity searching to cluster\nmethylation profiles by age group, while the second phase uses Explainable\nBoosting Machines (EBM) for precise, group-specific prediction. Our method not\nonly improves prediction accuracy but also reveals key age-related CpG sites,\ndetects age-specific changes in aging rates, and identifies pairwise\ninteractions between CpG sites. Experimental results show that our approach\noutperforms traditional epigenetic clocks and machine learning models, offering\na more accurate and interpretable solution for biological age estimation with\nsignificant implications for aging research.\n","arxiv_id":"http://arxiv.org/abs/2501.02401v1","authors":["Zipeng Wu","Daniel Herring","Fabian Spill","James Andrews"]},{"title":"Boosting Explainability through Selective Rationalization in Pre-trained\n Language Models","abstract":" The widespread application of pre-trained language models (PLMs) in natural\nlanguage processing (NLP) has led to increasing concerns about their\nexplainability. Selective rationalization is a self-explanatory framework that\nselects human-intelligible input subsets as rationales for predictions. Recent\nstudies have shown that applying existing rationalization frameworks to PLMs\nwill result in severe degeneration and failure problems, producing sub-optimal\nor meaningless rationales. Such failures severely damage trust in\nrationalization methods and constrain the application of rationalization\ntechniques on PLMs. In this paper, we find that the homogeneity of tokens in\nthe sentences produced by PLMs is the primary contributor to these problems. To\naddress these challenges, we propose a method named Pre-trained Language\nModel's Rationalization (PLMR), which splits PLMs into a generator and a\npredictor to deal with NLP tasks while providing interpretable rationales. The\ngenerator in PLMR also alleviates homogeneity by pruning irrelevant tokens,\nwhile the predictor uses full-text information to standardize predictions.\nExperiments conducted on two widely used datasets across multiple PLMs\ndemonstrate the effectiveness of the proposed method PLMR in addressing the\nchallenge of applying selective rationalization to PLMs. Codes:\nhttps://github.com/ylb777/PLMR.\n","arxiv_id":"http://arxiv.org/abs/2501.03182v1","authors":["Libing Yuan","Shuaibo Hu","Kui Yu","Le Wu"]},{"title":"AI-ANNE: (A) (N)eural (N)et for (E)xploration: Transferring Deep\n Learning Models onto Microcontrollers and Embedded Systems","abstract":" This working paper explores the integration of neural networks onto\nresource-constrained embedded systems like a Raspberry Pi Pico / Raspberry Pi\nPico 2. A TinyML aproach transfers neural networks directly on these\nmicrocontrollers, enabling real-time, low-latency, and energy-efficient\ninference while maintaining data privacy. Therefore, AI-ANNE: (A) (N)eural\n(N)et for (E)xploration will be presented, which facilitates the transfer of\npre-trained models from high-performance platforms like TensorFlow and Keras\nonto microcontrollers, using a lightweight programming language like\nMicroPython. This approach demonstrates how neural network architectures, such\nas neurons, layers, density and activation functions can be implemented in\nMicroPython in order to deal with the computational limitations of embedded\nsystems. Based on the Raspberry Pi Pico / Raspberry Pi Pico 2, two different\nneural networks on microcontrollers are presented for an example of data\nclassification. As an further application example, such a microcontroller can\nbe used for condition monitoring, where immediate corrective measures are\ntriggered on the basis of sensor data. Overall, this working paper presents a\nvery easy-to-implement way of using neural networks on energy-efficient devices\nsuch as microcontrollers. This makes AI-ANNE: (A) (N)eural (N)et for\n(E)xploration not only suited for practical use, but also as an educational\ntool with clear insights into how neural networks operate.\n","arxiv_id":"http://arxiv.org/abs/2501.03256v1","authors":["Dennis Klinkhammer"]},{"title":"Optimizing Edge AI: A Comprehensive Survey on Data, Model, and System\n Strategies","abstract":" The emergence of 5G and edge computing hardware has brought about a\nsignificant shift in artificial intelligence, with edge AI becoming a crucial\ntechnology for enabling intelligent applications. With the growing amount of\ndata generated and stored on edge devices, deploying AI models for local\nprocessing and inference has become increasingly necessary. However, deploying\nstate-of-the-art AI models on resource-constrained edge devices faces\nsignificant challenges that must be addressed. This paper presents an\noptimization triad for efficient and reliable edge AI deployment, including\ndata, model, and system optimization. First, we discuss optimizing data through\ndata cleaning, compression, and augmentation to make it more suitable for edge\ndeployment. Second, we explore model design and compression methods at the\nmodel level, such as pruning, quantization, and knowledge distillation.\nFinally, we introduce system optimization techniques like framework support and\nhardware acceleration to accelerate edge AI workflows. Based on an in-depth\nanalysis of various application scenarios and deployment challenges of edge AI,\nthis paper proposes an optimization paradigm based on the data-model-system\ntriad to enable a whole set of solutions to effectively transfer ML models,\nwhich are initially trained in the cloud, to various edge devices for\nsupporting multiple scenarios.\n","arxiv_id":"http://arxiv.org/abs/2501.03265v1","authors":["Xubin Wang","Weijia Jia"]},{"title":"Heterogeneous Graph Pre-training Based Model for Secure and Efficient\n Prediction of Default Risk Propagation among Bond Issuers","abstract":" Efficient prediction of default risk for bond-issuing enterprises is pivotal\nfor maintaining stability and fostering growth in the bond market. Conventional\nmethods usually rely solely on an enterprise's internal data for risk\nassessment. In contrast, graph-based techniques leverage interconnected\ncorporate information to enhance default risk identification for targeted bond\nissuers. Traditional graph techniques such as label propagation algorithm or\ndeepwalk fail to effectively integrate a enterprise's inherent attribute\ninformation with its topological network data. Additionally, due to data\nscarcity and security privacy concerns between enterprises, end-to-end graph\nneural network (GNN) algorithms may struggle in delivering satisfactory\nperformance for target tasks. To address these challenges, we present a novel\ntwo-stage model. In the first stage, we employ an innovative Masked\nAutoencoders for Heterogeneous Graph (HGMAE) to pre-train on a vast enterprise\nknowledge graph. Subsequently, in the second stage, a specialized classifier\nmodel is trained to predict default risk propagation probabilities. The\nclassifier leverages concatenated feature vectors derived from the pre-trained\nencoder with the enterprise's task-specific feature vectors. Through the\ntwo-stage training approach, our model not only boosts the importance of unique\nbond characteristics for specific default prediction tasks, but also securely\nand efficiently leverage the global information pre-trained from other\nenterprises. Experimental results demonstrate that our proposed model\noutperforms existing approaches in predicting default risk for bond issuers.\n","arxiv_id":"http://arxiv.org/abs/2501.03268v1","authors":["Xurui Li","Xin Shan","Wenhao Yin","Haijiao Wang"]},{"title":"Knowledge-Guided Biomarker Identification for Label-Free Single-Cell\n RNA-Seq Data: A Reinforcement Learning Perspective","abstract":" Gene panel selection aims to identify the most informative genomic biomarkers\nin label-free genomic datasets. Traditional approaches, which rely on domain\nexpertise, embedded machine learning models, or heuristic-based iterative\noptimization, often introduce biases and inefficiencies, potentially obscuring\ncritical biological signals. To address these challenges, we present an\niterative gene panel selection strategy that harnesses ensemble knowledge from\nexisting gene selection algorithms to establish preliminary boundaries or prior\nknowledge, which guide the initial search space. Subsequently, we incorporate\nreinforcement learning through a reward function shaped by expert behavior,\nenabling dynamic refinement and targeted selection of gene panels. This\nintegration mitigates biases stemming from initial boundaries while\ncapitalizing on RL's stochastic adaptability. Comprehensive comparative\nexperiments, case studies, and downstream analyses demonstrate the\neffectiveness of our method, highlighting its improved precision and efficiency\nfor label-free biomarker discovery. Our results underscore the potential of\nthis approach to advance single-cell genomics data analysis.\n","arxiv_id":"http://arxiv.org/abs/2501.04718v1","authors":["Meng Xiao","Weiliang Zhang","Xiaohan Huang","Hengshu Zhu","Min Wu","Xiaoli Li","Yuanchun Zhou"]},{"title":"Calculating Customer Lifetime Value and Churn using Beta Geometric\n Negative Binomial and Gamma-Gamma Distribution in a NFT based setting","abstract":" Customer Lifetime Value (CLV) is an important metric that measures the total\nvalue a customer will bring to a business over their lifetime. The Beta\nGeometric Negative Binomial Distribution (BGNBD) and Gamma Gamma Distribution\nare two models that can be used to calculate CLV, taking into account both the\nfrequency and value of customer transactions. This article explains the BGNBD\nand Gamma Gamma Distribution models, and how they can be used to calculate CLV\nfor NFT (Non-Fungible Token) transaction data in a blockchain setting. By\nestimating the parameters of these models using historical transaction data,\nbusinesses can gain insights into the lifetime value of their customers and\nmake data-driven decisions about marketing and customer retention strategies.\n","arxiv_id":"http://arxiv.org/abs/2501.04719v1","authors":["Sagarnil Das"]},{"title":"Multi-task Domain Adaptation for Computation Offloading in\n Edge-intelligence Networks","abstract":" In the field of multi-access edge computing (MEC), efficient computation\noffloading is crucial for improving resource utilization and reducing latency\nin dynamically changing environments. This paper introduces a new approach,\ntermed as Multi-Task Domain Adaptation (MTDA), aiming to enhance the ability of\ncomputational offloading models to generalize in the presence of domain shifts,\ni.e., when new data in the target environment significantly differs from the\ndata in the source domain. The proposed MTDA model incorporates a\nteacher-student architecture that allows continuous adaptation without\nnecessitating access to the source domain data during inference, thereby\nmaintaining privacy and reducing computational overhead. Utilizing a multi-task\nlearning framework that simultaneously manages offloading decisions and\nresource allocation, the proposed MTDA approach outperforms benchmark methods\nregarding mean squared error and accuracy, particularly in environments with\nincreasing numbers of users. It is observed by means of computer simulation\nthat the proposed MTDA model maintains high performance across various\nscenarios, demonstrating its potential for practical deployment in emerging MEC\napplications.\n","arxiv_id":"http://arxiv.org/abs/2501.07585v1","authors":["Runxin Han","Bo Yang","Zhiwen Yu","Xuelin Cao","George C. Alexandropoulos","Chau Yuen"]},{"title":"Adjoint sharding for very long context training of state space models","abstract":" Despite very fast progress, efficiently training large language models (LLMs)\nin very long contexts remains challenging. Existing methods fall back to\ntraining LLMs with short contexts (a maximum of a few thousands tokens in\ntraining) and use inference time techniques when evaluating on long contexts\n(above 1M tokens context window at inference). As opposed to\nlong-context-inference, training on very long context input prompts is quickly\nlimited by GPU memory availability and by the prohibitively long training times\nit requires on state-of-the-art hardware. Meanwhile, many real-life\napplications require not only inference but also training/fine-tuning with long\ncontext on specific tasks. Such applications include, for example, augmenting\nthe context with various sources of raw reference information for fact\nextraction, fact summarization, or fact reconciliation tasks. We propose\nadjoint sharding, a novel technique that comprises sharding gradient\ncalculation during training to reduce memory requirements by orders of\nmagnitude, making training on very long context computationally tractable.\nAdjoint sharding is based on the adjoint method and computes equivalent\ngradients to backpropagation. We also propose truncated adjoint sharding to\nspeed up the algorithm while maintaining performance. We provide a distributed\nversion, and a paralleled version of adjoint sharding to further speed up\ntraining. Empirical results show the proposed adjoint sharding algorithm\nreduces memory usage by up to 3X with a 1.27B parameter large language model on\n1M context length training. This allows to increase the maximum context length\nduring training or fine-tuning of a 1.27B parameter model from 35K tokens to\nabove 100K tokens on a training infrastructure composed of five AWS P4\ninstances.\n","arxiv_id":"http://arxiv.org/abs/2501.00692v1","authors":["Xingzi Xu","Amir Tavanaei","Kavosh Asadi","Karim Bouyarmane"]},{"title":"Everywhere Attack: Attacking Locally and Globally to Boost Targeted\n Transferability","abstract":" Adversarial examples' (AE) transferability refers to the phenomenon that AEs\ncrafted with one surrogate model can also fool other models. Notwithstanding\nremarkable progress in untargeted transferability, its targeted counterpart\nremains challenging. This paper proposes an everywhere scheme to boost targeted\ntransferability. Our idea is to attack a victim image both globally and\nlocally. We aim to optimize 'an army of targets' in every local image region\ninstead of the previous works that optimize a high-confidence target in the\nimage. Specifically, we split a victim image into non-overlap blocks and\njointly mount a targeted attack on each block. Such a strategy mitigates\ntransfer failures caused by attention inconsistency between surrogate and\nvictim models and thus results in stronger transferability. Our approach is\nmethod-agnostic, which means it can be easily combined with existing\ntransferable attacks for even higher transferability. Extensive experiments on\nImageNet demonstrate that the proposed approach universally improves the\nstate-of-the-art targeted attacks by a clear margin, e.g., the transferability\nof the widely adopted Logit attack can be improved by 28.8%-300%.We also\nevaluate the crafted AEs on a real-world platform: Google Cloud Vision. Results\nfurther support the superiority of the proposed method.\n","arxiv_id":"http://arxiv.org/abs/2501.00707v1","authors":["Hui Zeng","Sanshuai Cui","Biwei Chen","Anjie Peng"]},{"title":"An AI-powered Bayesian generative modeling approach for causal inference\n in observational studies","abstract":" Causal inference in observational studies with high-dimensional covariates\npresents significant challenges. We introduce CausalBGM, an AI-powered Bayesian\ngenerative modeling approach that captures the causal relationship among\ncovariates, treatment, and outcome variables. The core innovation of CausalBGM\nlies in its ability to estimate the individual treatment effect (ITE) by\nlearning individual-specific distributions of a low-dimensional latent feature\nset (e.g., latent confounders) that drives changes in both treatment and\noutcome. This approach not only effectively mitigates confounding effects but\nalso provides comprehensive uncertainty quantification, offering reliable and\ninterpretable causal effect estimates at the individual level. CausalBGM adopts\na Bayesian model and uses a novel iterative algorithm to update the model\nparameters and the posterior distribution of latent features until convergence.\nThis framework leverages the power of AI to capture complex dependencies among\nvariables while adhering to the Bayesian principles. Extensive experiments\ndemonstrate that CausalBGM consistently outperforms state-of-the-art methods,\nparticularly in scenarios with high-dimensional covariates and large-scale\ndatasets. Its Bayesian foundation ensures statistical rigor, providing robust\nand well-calibrated posterior intervals. By addressing key limitations of\nexisting methods, CausalBGM emerges as a robust and promising framework for\nadvancing causal inference in modern applications in fields such as genomics,\nhealthcare, and social sciences. CausalBGM is maintained at the website\nhttps://causalbgm.readthedocs.io/.\n","arxiv_id":"http://arxiv.org/abs/2501.00755v1","authors":["Qiao Liu","Wing Hung Wong"]},{"title":"Revisiting Graph Neural Networks on Graph-level Tasks: Comprehensive\n Experiments, Analysis, and Improvements","abstract":" Graphs are essential data structures for modeling complex interactions in\ndomains such as social networks, molecular structures, and biological systems.\nGraph-level tasks, which predict properties or classes for the entire graph,\nare critical for applications, such as molecular property prediction and\nsubgraph counting. Graph Neural Networks (GNNs) have shown promise in these\ntasks, but their evaluations are often limited to narrow datasets, tasks, and\ninconsistent experimental setups, restricting their generalizability. To\naddress these limitations, we propose a unified evaluation framework for\ngraph-level GNNs. This framework provides a standardized setting to evaluate\nGNNs across diverse datasets, various graph tasks (e.g., graph classification\nand regression), and challenging scenarios, including noisy, imbalanced, and\nfew-shot graphs. Additionally, we propose a novel GNN model with enhanced\nexpressivity and generalization capabilities. Specifically, we enhance the\nexpressivity of GNNs through a $k$-path rooted subgraph approach, enabling the\nmodel to effectively count subgraphs (e.g., paths and cycles). Moreover, we\nintroduce a unified graph contrastive learning algorithm for graphs across\ndiverse domains, which adaptively removes unimportant edges to augment graphs,\nthereby significantly improving generalization performance. Extensive\nexperiments demonstrate that our model achieves superior performance against\nfourteen effective baselines across twenty-seven graph datasets, establishing\nit as a robust and generalizable model for graph-level tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.00773v1","authors":["Haoyang Li","Yuming Xu","Chen Jason Zhang","Alexander Zhou","Lei Chen","Qing Li"]},{"title":"LENS-XAI: Redefining Lightweight and Explainable Network Security\n through Knowledge Distillation and Variational Autoencoders for Scalable\n Intrusion Detection in Cybersecurity","abstract":" The rapid proliferation of Industrial Internet of Things (IIoT) systems\nnecessitates advanced, interpretable, and scalable intrusion detection systems\n(IDS) to combat emerging cyber threats. Traditional IDS face challenges such as\nhigh computational demands, limited explainability, and inflexibility against\nevolving attack patterns. To address these limitations, this study introduces\nthe Lightweight Explainable Network Security framework (LENS-XAI), which\ncombines robust intrusion detection with enhanced interpretability and\nscalability. LENS-XAI integrates knowledge distillation, variational\nautoencoder models, and attribution-based explainability techniques to achieve\nhigh detection accuracy and transparency in decision-making. By leveraging a\ntraining set comprising 10% of the available data, the framework optimizes\ncomputational efficiency without sacrificing performance. Experimental\nevaluation on four benchmark datasets: Edge-IIoTset, UKM-IDS20, CTU-13, and\nNSL-KDD, demonstrates the framework's superior performance, achieving detection\naccuracies of 95.34%, 99.92%, 98.42%, and 99.34%, respectively. Additionally,\nthe framework excels in reducing false positives and adapting to complex attack\nscenarios, outperforming existing state-of-the-art methods. Key strengths of\nLENS-XAI include its lightweight design, suitable for resource-constrained\nenvironments, and its scalability across diverse IIoT and cybersecurity\ncontexts. Moreover, the explainability module enhances trust and transparency,\ncritical for practical deployment in dynamic and sensitive applications. This\nresearch contributes significantly to advancing IDS by addressing computational\nefficiency, feature interpretability, and real-world applicability. Future work\ncould focus on extending the framework to ensemble AI systems for distributed\nenvironments, further enhancing its robustness and adaptability.\n","arxiv_id":"http://arxiv.org/abs/2501.00790v2","authors":["Muhammet Anil Yagiz","Polat Goktas"]},{"title":"Decoupling Knowledge and Reasoning in Transformers: A Modular\n Architecture with Generalized Cross-Attention","abstract":" Transformers have achieved remarkable success across diverse domains, but\ntheir monolithic architecture presents challenges in interpretability,\nadaptability, and scalability. This paper introduces a novel modular\nTransformer architecture that explicitly decouples knowledge and reasoning\nthrough a generalized cross-attention mechanism to a globally shared knowledge\nbase with layer-specific transformations, specifically designed for effective\nknowledge retrieval. Critically, we provide a rigorous mathematical derivation\ndemonstrating that the Feed-Forward Network (FFN) in a standard Transformer is\na specialized case (a closure) of this generalized cross-attention, revealing\nits role in implicit knowledge retrieval and validating our design. This\ntheoretical framework provides a new lens for understanding FFNs and lays the\nfoundation for future research exploring enhanced interpretability,\nadaptability, and scalability, enabling richer interplay with external\nknowledge bases and other systems.\n","arxiv_id":"http://arxiv.org/abs/2501.00823v2","authors":["Zhenyu Guo","Wenguang Chen"]},{"title":"What is a Social Media Bot? A Global Comparison of Bot and Human\n Characteristics","abstract":" Chatter on social media is 20% bots and 80% humans. Chatter by bots and\nhumans is consistently different: bots tend to use linguistic cues that can be\neasily automated while humans use cues that require dialogue understanding.\nBots use words that match the identities they choose to present, while humans\nmay send messages that are not related to the identities they present. Bots and\nhumans differ in their communication structure: sampled bots have a star\ninteraction structure, while sampled humans have a hierarchical structure.\nThese conclusions are based on a large-scale analysis of social media tweets\nacross ~200mil users across 7 events. Social media bots took the world by storm\nwhen social-cybersecurity researchers realized that social media users not only\nconsisted of humans but also of artificial agents called bots. These bots wreck\nhavoc online by spreading disinformation and manipulating narratives. Most\nresearch on bots are based on special-purposed definitions, mostly predicated\non the event studied. This article first begins by asking, \"What is a bot?\",\nand we study the underlying principles of how bots are different from humans.\nWe develop a first-principle definition of a social media bot. With this\ndefinition as a premise, we systematically compare characteristics between bots\nand humans across global events, and reflect on how the software-programmed bot\nis an Artificial Intelligent algorithm, and its potential for evolution as\ntechnology advances. Based on our results, we provide recommendations for the\nuse and regulation of bots. Finally, we discuss open challenges and future\ndirections: Detect, to systematically identify these automated and potentially\nevolving bots; Differentiate, to evaluate the goodness of the bot in terms of\ntheir content postings and relationship interactions; Disrupt, to moderate the\nimpact of malicious bots.\n","arxiv_id":"http://arxiv.org/abs/2501.00855v1","authors":["Lynnette Hui Xian Ng","Kathleen M. Carley"]},{"title":"DiffETM: Diffusion Process Enhanced Embedded Topic Model","abstract":" The embedded topic model (ETM) is a widely used approach that assumes the\nsampled document-topic distribution conforms to the logistic normal\ndistribution for easier optimization. However, this assumption oversimplifies\nthe real document-topic distribution, limiting the model's performance. In\nresponse, we propose a novel method that introduces the diffusion process into\nthe sampling process of document-topic distribution to overcome this limitation\nand maintain an easy optimization process. We validate our method through\nextensive experiments on two mainstream datasets, proving its effectiveness in\nimproving topic modeling performance.\n","arxiv_id":"http://arxiv.org/abs/2501.00862v1","authors":["Wei Shao","Mingyang Liu","Linqi Song"]},{"title":"Representation in large language models","abstract":" The extraordinary success of recent Large Language Models (LLMs) on a diverse\narray of tasks has led to an explosion of scientific and philosophical\ntheorizing aimed at explaining how they do what they do. Unfortunately,\ndisagreement over fundamental theoretical issues has led to stalemate, with\nentrenched camps of LLM optimists and pessimists often committed to very\ndifferent views of how these systems work. Overcoming stalemate requires\nagreement on fundamental questions, and the goal of this paper is to address\none such question, namely: is LLM behavior driven partly by\nrepresentation-based information processing of the sort implicated in\nbiological cognition, or is it driven entirely by processes of memorization and\nstochastic table look-up? This is a question about what kind of algorithm LLMs\nimplement, and the answer carries serious implications for higher level\nquestions about whether these systems have beliefs, intentions, concepts,\nknowledge, and understanding. I argue that LLM behavior is partially driven by\nrepresentation-based information processing, and then I describe and defend a\nseries of practical techniques for investigating these representations and\ndeveloping explanations on their basis. The resulting account provides a\ngroundwork for future theorizing about language models and their successors.\n","arxiv_id":"http://arxiv.org/abs/2501.00885v1","authors":["Cameron C. Yetman"]},{"title":"Demystifying Online Clustering of Bandits: Enhanced Exploration Under\n Stochastic and Smoothed Adversarial Contexts","abstract":" The contextual multi-armed bandit (MAB) problem is crucial in sequential\ndecision-making. A line of research, known as online clustering of bandits,\nextends contextual MAB by grouping similar users into clusters, utilizing\nshared features to improve learning efficiency. However, existing algorithms,\nwhich rely on the upper confidence bound (UCB) strategy, struggle to gather\nadequate statistical information to accurately identify unknown user clusters.\nAs a result, their theoretical analyses require several strong assumptions\nabout the \"diversity\" of contexts generated by the environment, leading to\nimpractical settings, complicated analyses, and poor practical performance.\nRemoving these assumptions has been a long-standing open problem in the\nclustering of bandits literature. In this paper, we provide two solutions to\nthis open problem. First, following the i.i.d. context generation setting in\nexisting studies, we propose two novel algorithms, UniCLUB and PhaseUniCLUB,\nwhich incorporate enhanced exploration mechanisms to accelerate cluster\nidentification. Remarkably, our algorithms require substantially weaker\nassumptions while achieving regret bounds comparable to prior work. Second,\ninspired by the smoothed analysis framework, we propose a more practical\nsetting that eliminates the requirement for i.i.d. context generation used in\nprevious studies, thus enhancing the performance of existing algorithms for\nonline clustering of bandits. Our technique can be applied to both graph-based\nand set-based clustering of bandits frameworks. Extensive evaluations on both\nsynthetic and real-world datasets demonstrate that our proposed algorithms\nconsistently outperform existing approaches.\n","arxiv_id":"http://arxiv.org/abs/2501.00891v1","authors":["Zhuohua Li","Maoli Liu","Xiangxiang Dai","John C. S. Lui"]},{"title":"Large Language Model Based Multi-Agent System Augmented Complex Event\n Processing Pipeline for Internet of Multimedia Things","abstract":" This paper presents the development and evaluation of a Large Language Model\n(LLM), also known as foundation models, based multi-agent system framework for\ncomplex event processing (CEP) with a focus on video query processing use\ncases. The primary goal is to create a proof-of-concept (POC) that integrates\nstate-of-the-art LLM orchestration frameworks with publish/subscribe (pub/sub)\ntools to address the integration of LLMs with current CEP systems. Utilizing\nthe Autogen framework in conjunction with Kafka message brokers, the system\ndemonstrates an autonomous CEP pipeline capable of handling complex workflows.\nExtensive experiments evaluate the system's performance across varying\nconfigurations, complexities, and video resolutions, revealing the trade-offs\nbetween functionality and latency. The results show that while higher agent\ncount and video complexities increase latency, the system maintains high\nconsistency in narrative coherence. This research builds upon and contributes\nto, existing novel approaches to distributed AI systems, offering detailed\ninsights into integrating such systems into existing infrastructures.\n","arxiv_id":"http://arxiv.org/abs/2501.00906v2","authors":["Talha Zeeshan","Abhishek Kumar","Susanna Pirttikangas","Sasu Tarkoma"]},{"title":"Enhancing Early Diabetic Retinopathy Detection through Synthetic DR1\n Image Generation: A StyleGAN3 Approach","abstract":" Diabetic Retinopathy (DR) is a leading cause of preventable blindness. Early\ndetection at the DR1 stage is critical but is hindered by a scarcity of\nhigh-quality fundus images. This study uses StyleGAN3 to generate synthetic DR1\nimages characterized by microaneurysms with high fidelity and diversity. The\naim is to address data scarcity and enhance the performance of supervised\nclassifiers. A dataset of 2,602 DR1 images was used to train the model,\nfollowed by a comprehensive evaluation using quantitative metrics, including\nFrechet Inception Distance (FID), Kernel Inception Distance (KID), and\nEquivariance with respect to translation (EQ-T) and rotation (EQ-R).\nQualitative assessments included Human Turing tests, where trained\nophthalmologists evaluated the realism of synthetic images. Spectral analysis\nfurther validated image quality. The model achieved a final FID score of 17.29,\noutperforming the mean FID of 21.18 (95 percent confidence interval - 20.83 to\n21.56) derived from bootstrap resampling. Human Turing tests demonstrated the\nmodel's ability to produce highly realistic images, though minor artifacts near\nthe borders were noted. These findings suggest that StyleGAN3-generated\nsynthetic DR1 images hold significant promise for augmenting training datasets,\nenabling more accurate early detection of Diabetic Retinopathy. This\nmethodology highlights the potential of synthetic data in advancing medical\nimaging and AI-driven diagnostics.\n","arxiv_id":"http://arxiv.org/abs/2501.00954v1","authors":["Sagarnil Das","Pradeep Walia"]},{"title":"The Silent Majority: Demystifying Memorization Effect in the Presence of\n Spurious Correlations","abstract":" Machine learning models often rely on simple spurious features -- patterns in\ntraining data that correlate with targets but are not causally related to them,\nlike image backgrounds in foreground classification. This reliance typically\nleads to imbalanced test performance across minority and majority groups. In\nthis work, we take a closer look at the fundamental cause of such imbalanced\nperformance through the lens of memorization, which refers to the ability to\npredict accurately on \\textit{atypical} examples (minority groups) in the\ntraining set but failing in achieving the same accuracy in the testing set.\nThis paper systematically shows the ubiquitous existence of spurious features\nin a small set of neurons within the network, providing the first-ever evidence\nthat memorization may contribute to imbalanced group performance. Through three\nexperimental sources of converging empirical evidence, we find the property of\na small subset of neurons or channels in memorizing minority group information.\nInspired by these findings, we articulate the hypothesis: the imbalanced group\nperformance is a byproduct of ``noisy'' spurious memorization confined to a\nsmall set of neurons. To further substantiate this hypothesis, we show that\neliminating these unnecessary spurious memorization patterns via a novel\nframework during training can significantly affect the model performance on\nminority groups. Our experimental results across various architectures and\nbenchmarks offer new insights on how neural networks encode core and spurious\nknowledge, laying the groundwork for future research in demystifying robustness\nto spurious correlation.\n","arxiv_id":"http://arxiv.org/abs/2501.00961v2","authors":["Chenyu You","Haocheng Dai","Yifei Min","Jasjeet S. Sekhon","Sarang Joshi","James S. Duncan"]},{"title":"FlashInfer: Efficient and Customizable Attention Engine for LLM\n Inference Serving","abstract":" Transformers, driven by attention mechanisms, form the foundation of large\nlanguage models (LLMs). As these models scale up, efficient GPU attention\nkernels become essential for high-throughput and low-latency inference. Diverse\nLLM applications demand flexible and high-performance attention solutions. We\npresent FlashInfer: a customizable and efficient attention engine for LLM\nserving. FlashInfer tackles KV-cache storage heterogeneity using block-sparse\nformat and composable formats to optimize memory access and reduce redundancy.\nIt also offers a customizable attention template, enabling adaptation to\nvarious settings through Just-In-Time (JIT) compilation. Additionally,\nFlashInfer's load-balanced scheduling algorithm adjusts to dynamism of user\nrequests while maintaining compatibility with CUDAGraph which requires static\nconfiguration. FlashInfer have been integrated into leading LLM serving\nframeworks like SGLang, vLLM and MLC-Engine. Comprehensive kernel-level and\nend-to-end evaluations demonstrate FlashInfer's ability to significantly boost\nkernel performance across diverse inference scenarios: compared to\nstate-of-the-art LLM serving solutions, FlashInfer achieve 29-69%\ninter-token-latency reduction compared to compiler backends for LLM serving\nbenchmark, 28-30% latency reduction for long-context inference, and 13-17%\nspeedup for LLM serving with parallel generation.\n","arxiv_id":"http://arxiv.org/abs/2501.01005v1","authors":["Zihao Ye","Lequn Chen","Ruihang Lai","Wuwei Lin","Yineng Zhang","Stephanie Wang","Tianqi Chen","Baris Kasikci","Vinod Grover","Arvind Krishnamurthy","Luis Ceze"]},{"title":"CryptoMamba: Leveraging State Space Models for Accurate Bitcoin Price\n Prediction","abstract":" Predicting Bitcoin price remains a challenging problem due to the high\nvolatility and complex non-linear dynamics of cryptocurrency markets.\nTraditional time-series models, such as ARIMA and GARCH, and recurrent neural\nnetworks, like LSTMs, have been widely applied to this task but struggle to\ncapture the regime shifts and long-range dependencies inherent in the data. In\nthis work, we propose CryptoMamba, a novel Mamba-based State Space Model (SSM)\narchitecture designed to effectively capture long-range dependencies in\nfinancial time-series data. Our experiments show that CryptoMamba not only\nprovides more accurate predictions but also offers enhanced generalizability\nacross different market conditions, surpassing the limitations of previous\nmodels. Coupled with trading algorithms for real-world scenarios, CryptoMamba\ndemonstrates its practical utility by translating accurate forecasts into\nfinancial outcomes. Our findings signal a huge advantage for SSMs in stock and\ncryptocurrency price forecasting tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.01010v1","authors":["Mohammad Shahab Sepehri","Asal Mehradfar","Mahdi Soltanolkotabi","Salman Avestimehr"]},{"title":"ValuesRAG: Enhancing Cultural Alignment Through Retrieval-Augmented\n Contextual Learning","abstract":" Cultural values alignment in Large Language Models (LLMs) is a critical\nchallenge due to their tendency to embed Western-centric biases from training\ndata, leading to misrepresentations and fairness issues in cross-cultural\ncontexts. Recent approaches, such as role-assignment and few-shot learning,\noften struggle with reliable cultural alignment as they heavily rely on\npre-trained knowledge, lack scalability, and fail to capture nuanced cultural\nvalues effectively. To address these issues, we propose ValuesRAG, a novel and\neffective framework that applies Retrieval-Augmented Generation (RAG) with\nIn-Context Learning (ICL) to integrate cultural and demographic knowledge\ndynamically during text generation. Leveraging the World Values Survey (WVS)\ndataset, ValuesRAG first generates summaries of values for each individual.\nSubsequently, we curate several representative regional datasets to serve as\ntest datasets and retrieve relevant summaries of values based on demographic\nfeatures, followed by a reranking step to select the top-k relevant summaries.\nValuesRAG consistently outperforms baseline methods, both in the main\nexperiment and in the ablation study where only the values summary was\nprovided. Notably, ValuesRAG demonstrates an accuracy of 21% improvement over\nother baseline methods, highlighting its potential to foster culturally aligned\nAI systems and enhance the inclusivity of AI-driven applications.\n","arxiv_id":"http://arxiv.org/abs/2501.01031v2","authors":["Wonduk Seo","Zonghao Yuan","Yi Bu"]},{"title":"MSC-Bench: Benchmarking and Analyzing Multi-Sensor Corruption for\n Driving Perception","abstract":" Multi-sensor fusion models play a crucial role in autonomous driving\nperception, particularly in tasks like 3D object detection and HD map\nconstruction. These models provide essential and comprehensive static\nenvironmental information for autonomous driving systems. While camera-LiDAR\nfusion methods have shown promising results by integrating data from both\nmodalities, they often depend on complete sensor inputs. This reliance can lead\nto low robustness and potential failures when sensors are corrupted or missing,\nraising significant safety concerns. To tackle this challenge, we introduce the\nMulti-Sensor Corruption Benchmark (MSC-Bench), the first comprehensive\nbenchmark aimed at evaluating the robustness of multi-sensor autonomous driving\nperception models against various sensor corruptions. Our benchmark includes 16\ncombinations of corruption types that disrupt both camera and LiDAR inputs,\neither individually or concurrently. Extensive evaluations of six 3D object\ndetection models and four HD map construction models reveal substantial\nperformance degradation under adverse weather conditions and sensor failures,\nunderscoring critical safety issues. The benchmark toolkit and affiliated code\nand model checkpoints have been made publicly accessible.\n","arxiv_id":"http://arxiv.org/abs/2501.01037v1","authors":["Xiaoshuai Hao","Guanqun Liu","Yuting Zhao","Yuheng Ji","Mengchuan Wei","Haimei Zhao","Lingdong Kong","Rong Yin","Yu Liu"]},{"title":"MMVA: Multimodal Matching Based on Valence and Arousal across Images,\n Music, and Musical Captions","abstract":" We introduce Multimodal Matching based on Valence and Arousal (MMVA), a\ntri-modal encoder framework designed to capture emotional content across\nimages, music, and musical captions. To support this framework, we expand the\nImage-Music-Emotion-Matching-Net (IMEMNet) dataset, creating IMEMNet-C which\nincludes 24,756 images and 25,944 music clips with corresponding musical\ncaptions. We employ multimodal matching scores based on the continuous valence\n(emotional positivity) and arousal (emotional intensity) values. This\ncontinuous matching score allows for random sampling of image-music pairs\nduring training by computing similarity scores from the valence-arousal values\nacross different modalities. Consequently, the proposed approach achieves\nstate-of-the-art performance in valence-arousal prediction tasks. Furthermore,\nthe framework demonstrates its efficacy in various zeroshot tasks, highlighting\nthe potential of valence and arousal predictions in downstream applications.\n","arxiv_id":"http://arxiv.org/abs/2501.01094v1","authors":["Suhwan Choi","Kyu Won Kim","Myungjoo Kang"]},{"title":"Disambiguation of Chinese Polyphones in an End-to-End Framework with\n Semantic Features Extracted by Pre-trained BERT","abstract":" Grapheme-to-phoneme (G2P) conversion serves as an essential component in\nChinese Mandarin text-to-speech (TTS) system, where polyphone disambiguation is\nthe core issue. In this paper, we propose an end-to-end framework to predict\nthe pronunciation of a polyphonic character, which accepts sentence containing\npolyphonic character as input in the form of Chinese character sequence without\nthe necessity of any preprocessing. The proposed method consists of a\npre-trained bidirectional encoder representations from Transformers (BERT)\nmodel and a neural network (NN) based classifier. The pre-trained BERT model\nextracts semantic features from a raw Chinese character sequence and the NN\nbased classifier predicts the polyphonic character's pronunciation according to\nBERT output. In out experiments, we implemented three classifiers, a\nfully-connected network based classifier, a long short-term memory (LSTM)\nnetwork based classifier and a Transformer block based classifier. The\nexperimental results compared with the baseline approach based on LSTM\ndemonstrate that, the pre-trained model extracts effective semantic features,\nwhich greatly enhances the performance of polyphone disambiguation. In\naddition, we also explored the impact of contextual information on polyphone\ndisambiguation.\n","arxiv_id":"http://arxiv.org/abs/2501.01102v1","authors":["Dongyang Dai","Zhiyong Wu","Shiyin Kang","Xixin Wu","Jia Jia","Dan Su","Dong Yu","Helen Meng"]},{"title":"learning discriminative features from spectrograms using center loss for\n speech emotion recognition","abstract":" Identifying the emotional state from speech is essential for the natural\ninteraction of the machine with the speaker. However, extracting effective\nfeatures for emotion recognition is difficult, as emotions are ambiguous. We\npropose a novel approach to learn discriminative features from variable length\nspectrograms for emotion recognition by cooperating softmax cross-entropy loss\nand center loss together. The softmax cross-entropy loss enables features from\ndifferent emotion categories separable, and center loss efficiently pulls the\nfeatures belonging to the same emotion category to their center. By combining\nthe two losses together, the discriminative power will be highly enhanced,\nwhich leads to network learning more effective features for emotion\nrecognition. As demonstrated by the experimental results, after introducing\ncenter loss, both the unweighted accuracy and weighted accuracy are improved by\nover 3\\% on Mel-spectrogram input, and more than 4\\% on Short Time Fourier\nTransform spectrogram input.\n","arxiv_id":"http://arxiv.org/abs/2501.01103v1","authors":["Dongyang Dai","Zhiyong Wu","Runnan Li","Xixin Wu","Jia Jia","Helen Meng"]},{"title":"Robust COVID-19 Detection from Cough Sounds using Deep Neural Decision\n Tree and Forest: A Comprehensive Cross-Datasets Evaluation","abstract":" This research presents a robust approach to classifying COVID-19 cough sounds\nusing cutting-edge machine-learning techniques. Leveraging deep neural decision\ntrees and deep neural decision forests, our methodology demonstrates consistent\nperformance across diverse cough sound datasets. We begin with a comprehensive\nextraction of features to capture a wide range of audio features from\nindividuals, whether COVID-19 positive or negative. To determine the most\nimportant features, we use recursive feature elimination along with\ncross-validation. Bayesian optimization fine-tunes hyper-parameters of deep\nneural decision tree and deep neural decision forest models. Additionally, we\nintegrate the SMOTE during training to ensure a balanced representation of\npositive and negative data. Model performance refinement is achieved through\nthreshold optimization, maximizing the ROC-AUC score. Our approach undergoes a\ncomprehensive evaluation in five datasets: Cambridge, Coswara, COUGHVID,\nVirufy, and the combined Virufy with the NoCoCoDa dataset. Consistently\noutperforming state-of-the-art methods, our proposed approach yields notable\nAUC scores of 0.97, 0.98, 0.92, 0.93, 0.99, and 0.99 across the respective\ndatasets. Merging all datasets into a combined dataset, our method, using a\ndeep neural decision forest classifier, achieves an AUC of 0.97. Also, our\nstudy includes a comprehensive cross-datasets analysis, revealing demographic\nand geographic differences in the cough sounds associated with COVID-19. These\ndifferences highlight the challenges in transferring learned features across\ndiverse datasets and underscore the potential benefits of dataset integration,\nimproving generalizability and enhancing COVID-19 detection from audio signals.\n","arxiv_id":"http://arxiv.org/abs/2501.01117v1","authors":["Rofiqul Islam","Nihad Karim Chowdhury","Muhammad Ashad Kabir"]},{"title":"TED: Turn Emphasis with Dialogue Feature Attention for Emotion\n Recognition in Conversation","abstract":" Emotion recognition in conversation (ERC) has been attracting attention by\nmethods for modeling multi-turn contexts. The multi-turn input to a pretraining\nmodel implicitly assumes that the current turn and other turns are\ndistinguished during the training process by inserting special tokens into the\ninput sequence. This paper proposes a priority-based attention method to\ndistinguish each turn explicitly by adding dialogue features into the attention\nmechanism, called Turn Emphasis with Dialogue (TED). It has a priority for each\nturn according to turn position and speaker information as dialogue features.\nIt takes multi-head self-attention between turn-based vectors for multi-turn\ninput and adjusts attention scores with the dialogue features. We evaluate TED\non four typical benchmarks. The experimental results demonstrate that TED has\nhigh overall performance in all datasets and achieves state-of-the-art\nperformance on IEMOCAP with numerous turns.\n","arxiv_id":"http://arxiv.org/abs/2501.01123v1","authors":["Junya Ono","Hiromi Wakaki"]},{"title":"Missing Data as Augmentation in the Earth Observation Domain: A\n Multi-View Learning Approach","abstract":" Multi-view learning (MVL) leverages multiple sources or views of data to\nenhance machine learning model performance and robustness. This approach has\nbeen successfully used in the Earth Observation (EO) domain, where views have a\nheterogeneous nature and can be affected by missing data. Despite the negative\neffect that missing data has on model predictions, the ML literature has used\nit as an augmentation technique to improve model generalization, like masking\nthe input data. Inspired by this, we introduce novel methods for EO\napplications tailored to MVL with missing views. Our methods integrate the\ncombination of a set to simulate all combinations of missing views as different\ntraining samples. Instead of replacing missing data with a numerical value, we\nuse dynamic merge functions, like average, and more complex ones like\nTransformer. This allows the MVL model to entirely ignore the missing views,\nenhancing its predictive robustness. We experiment on four EO datasets with\ntemporal and static views, including state-of-the-art methods from the EO\ndomain. The results indicate that our methods improve model robustness under\nconditions of moderate missingness, and improve the predictive performance when\nall views are present. The proposed methods offer a single adaptive solution to\noperate effectively with any combination of available views.\n","arxiv_id":"http://arxiv.org/abs/2501.01132v1","authors":["Francisco Mena","Diego Arenas","Andreas Dengel"]},{"title":"TexAVi: Generating Stereoscopic VR Video Clips from Text Descriptions","abstract":" While generative models such as text-to-image, large language models and\ntext-to-video have seen significant progress, the extension to\ntext-to-virtual-reality remains largely unexplored, due to a deficit in\ntraining data and the complexity of achieving realistic depth and motion in\nvirtual environments. This paper proposes an approach to coalesce existing\ngenerative systems to form a stereoscopic virtual reality video from text.\n Carried out in three main stages, we start with a base text-to-image model\nthat captures context from an input text. We then employ Stable Diffusion on\nthe rudimentary image produced, to generate frames with enhanced realism and\noverall quality. These frames are processed with depth estimation algorithms to\ncreate left-eye and right-eye views, which are stitched side-by-side to create\nan immersive viewing experience. Such systems would be highly beneficial in\nvirtual reality production, since filming and scene building often require\nextensive hours of work and post-production effort.\n We utilize image evaluation techniques, specifically Fr\\'echet Inception\nDistance and CLIP Score, to assess the visual quality of frames produced for\nthe video. These quantitative measures establish the proficiency of the\nproposed method.\n Our work highlights the exciting possibilities of using natural\nlanguage-driven graphics in fields like virtual reality simulations.\n","arxiv_id":"http://arxiv.org/abs/2501.01156v1","authors":["Vriksha Srihari","R. Bhavya","Shruti Jayaraman","V. Mary Anita Rajam"]},{"title":"Harnessing Multi-Agent LLMs for Complex Engineering Problem-Solving: A\n Framework for Senior Design Projects","abstract":" Multi-Agent Large Language Models (LLMs) are gaining significant attention\nfor their ability to harness collective intelligence in complex\nproblem-solving, decision-making, and planning tasks. This aligns with the\nconcept of the wisdom of crowds, where diverse agents contribute collectively\nto generating effective solutions, making it particularly suitable for\neducational settings. Senior design projects, also known as capstone or final\nyear projects, are pivotal in engineering education as they integrate\ntheoretical knowledge with practical application, fostering critical thinking,\nteamwork, and real-world problem-solving skills. In this paper, we explore the\nuse of Multi-Agent LLMs in supporting these senior design projects undertaken\nby engineering students, which often involve multidisciplinary considerations\nand conflicting objectives, such as optimizing technical performance while\naddressing ethical, social, and environmental concerns. We propose a framework\nwhere distinct LLM agents represent different expert perspectives, such as\nproblem formulation agents, system complexity agents, societal and ethical\nagents, or project managers, thus facilitating a holistic problem-solving\napproach. This implementation leverages standard multi-agent system (MAS)\nconcepts such as coordination, cooperation, and negotiation, incorporating\nprompt engineering to develop diverse personas for each agent. These agents\nengage in rich, collaborative dialogues to simulate human engineering teams,\nguided by principles from swarm AI to efficiently balance individual\ncontributions towards a unified solution. We adapt these techniques to create a\ncollaboration structure for LLM agents, encouraging interdisciplinary reasoning\nand negotiation similar to real-world senior design projects. To assess the\nefficacy of this framework, we collected six proposals of engineering and\ncomputer science of...\n","arxiv_id":"http://arxiv.org/abs/2501.01205v1","authors":["Abdullah Mushtaq","Muhammad Rafay Naeem","Ibrahim Ghaznavi","Muhammad Imran Taj","Imran Hashmi","Junaid Qadir"]},{"title":"Face-Human-Bench: A Comprehensive Benchmark of Face and Human\n Understanding for Multi-modal Assistants","abstract":" Faces and humans are crucial elements in social interaction and are widely\nincluded in everyday photos and videos. Therefore, a deep understanding of\nfaces and humans will enable multi-modal assistants to achieve improved\nresponse quality and broadened application scope. Currently, the multi-modal\nassistant community lacks a comprehensive and scientific evaluation of face and\nhuman understanding abilities. In this paper, we first propose a hierarchical\nability taxonomy that includes three levels of abilities. Then, based on this\ntaxonomy, we collect images and annotations from publicly available datasets in\nthe face and human community and build a semi-automatic data pipeline to\nproduce problems for the new benchmark. Finally, the obtained Face-Human-Bench\ncomprises a development set with 900 problems and a test set with 1800\nproblems, supporting both English and Chinese. We conduct evaluations over 25\nmainstream multi-modal large language models (MLLMs) with our Face-Human-Bench,\nfocusing on the correlation between abilities, the impact of the relative\nposition of targets on performance, and the impact of Chain of Thought (CoT)\nprompting on performance. Moreover, inspired by multi-modal agents, we also\nexplore which abilities of MLLMs need to be supplemented by specialist models.\n","arxiv_id":"http://arxiv.org/abs/2501.01243v2","authors":["Lixiong Qin","Shilong Ou","Miaoxuan Zhang","Jiangning Wei","Yuhang Zhang","Xiaoshuai Song","Yuchen Liu","Mei Wang","Weiran Xu"]},{"title":"ProgCo: Program Helps Self-Correction of Large Language Models","abstract":" Self-Correction aims to enable large language models (LLMs) to self-verify\nand self-refine their initial responses without external feedback. However,\nLLMs often fail to effectively self-verify and generate correct feedback,\nfurther misleading refinement and leading to the failure of self-correction,\nespecially in complex reasoning tasks. In this paper, we propose Program-driven\nSelf-Correction (ProgCo). First, program-driven verification (ProgVe) achieves\ncomplex verification logic and extensive validation through self-generated,\nself-executing verification pseudo-programs. Then, program-driven refinement\n(ProgRe) receives feedback from ProgVe, conducts dual reflection and refinement\non both responses and verification programs to mitigate misleading of incorrect\nfeedback in complex reasoning tasks. Experiments on three instruction-following\nand mathematical benchmarks indicate that ProgCo achieves effective\nself-correction, and can be further enhance performance when combined with real\nprogram tools.\n","arxiv_id":"http://arxiv.org/abs/2501.01264v1","authors":["Xiaoshuai Song","Yanan Wu","Weixun Wang","Jiaheng Liu","Wenbo Su","Bo Zheng"]},{"title":"CultureVLM: Characterizing and Improving Cultural Understanding of\n Vision-Language Models for over 100 Countries","abstract":" Vision-language models (VLMs) have advanced human-AI interaction but struggle\nwith cultural understanding, often misinterpreting symbols, gestures, and\nartifacts due to biases in predominantly Western-centric training data. In this\npaper, we construct CultureVerse, a large-scale multimodal benchmark covering\n19, 682 cultural concepts, 188 countries/regions, 15 cultural concepts, and 3\nquestion types, with the aim of characterizing and improving VLMs'\nmulticultural understanding capabilities. Then, we propose CultureVLM, a series\nof VLMs fine-tuned on our dataset to achieve significant performance\nimprovement in cultural understanding. Our evaluation of 16 models reveals\nsignificant disparities, with a stronger performance in Western concepts and\nweaker results in African and Asian contexts. Fine-tuning on our CultureVerse\nenhances cultural perception, demonstrating cross-cultural, cross-continent,\nand cross-dataset generalization without sacrificing performance on models'\ngeneral VLM benchmarks. We further present insights on cultural generalization\nand forgetting. We hope that this work could lay the foundation for more\nequitable and culturally aware multimodal AI systems.\n","arxiv_id":"http://arxiv.org/abs/2501.01282v1","authors":["Shudong Liu","Yiqiao Jin","Cheng Li","Derek F. Wong","Qingsong Wen","Lichao Sun","Haipeng Chen","Xing Xie","Jindong Wang"]},{"title":"LEO-Split: A Semi-Supervised Split Learning Framework over LEO Satellite\n Networks","abstract":" Recently, the increasing deployment of LEO satellite systems has enabled\nvarious space analytics (e.g., crop and climate monitoring), which heavily\nrelies on the advancements in deep learning (DL). However, the intermittent\nconnectivity between LEO satellites and ground station (GS) significantly\nhinders the timely transmission of raw data to GS for centralized learning,\nwhile the scaled-up DL models hamper distributed learning on\nresource-constrained LEO satellites. Though split learning (SL) can be a\npotential solution to these problems by partitioning a model and offloading\nprimary training workload to GS, the labor-intensive labeling process remains\nan obstacle, with intermittent connectivity and data heterogeneity being other\nchallenges. In this paper, we propose LEO-Split, a semi-supervised (SS) SL\ndesign tailored for satellite networks to combat these challenges. Leveraging\nSS learning to handle (labeled) data scarcity, we construct an auxiliary model\nto tackle the training failure of the satellite-GS non-contact time. Moreover,\nwe propose a pseudo-labeling algorithm to rectify data imbalances across\nsatellites. Lastly, an adaptive activation interpolation scheme is devised to\nprevent the overfitting of server-side sub-model training at GS. Extensive\nexperiments with real-world LEO satellite traces (e.g., Starlink) demonstrate\nthat our LEO-Split framework achieves superior performance compared to\nstate-ofthe-art benchmarks.\n","arxiv_id":"http://arxiv.org/abs/2501.01293v1","authors":["Zheng Lin","Yuxin Zhang","Zhe Chen","Zihan Fang","Cong Wu","Xianhao Chen","Yue Gao","Jun Luo"]},{"title":"The Prompt Alchemist: Automated LLM-Tailored Prompt Optimization for\n Test Case Generation","abstract":" Test cases are essential for validating the reliability and quality of\nsoftware applications. Recent studies have demonstrated the capability of Large\nLanguage Models (LLMs) to generate useful test cases for given source code.\nHowever, the existing work primarily relies on human-written plain prompts,\nwhich often leads to suboptimal results since the performance of LLMs can be\nhighly influenced by the prompts. Moreover, these approaches use the same\nprompt for all LLMs, overlooking the fact that different LLMs might be best\nsuited to different prompts. Given the wide variety of possible prompt\nformulations, automatically discovering the optimal prompt for each LLM\npresents a significant challenge. Although there are methods on automated\nprompt optimization in the natural language processing field, they are hard to\nproduce effective prompts for the test case generation task. First, the methods\niteratively optimize prompts by simply combining and mutating existing ones\nwithout proper guidance, resulting in prompts that lack diversity and tend to\nrepeat the same errors in the generated test cases. Second, the prompts are\ngenerally lack of domain contextual knowledge, limiting LLMs' performance in\nthe task.\n","arxiv_id":"http://arxiv.org/abs/2501.01329v1","authors":["Shuzheng Gao","Chaozheng Wang","Cuiyun Gao","Xiaoqian Jiao","Chun Yong Chong","Shan Gao","Michael Lyu"]},{"title":"CySecBench: Generative AI-based CyberSecurity-focused Prompt Dataset for\n Benchmarking Large Language Models","abstract":" Numerous studies have investigated methods for jailbreaking Large Language\nModels (LLMs) to generate harmful content. Typically, these methods are\nevaluated using datasets of malicious prompts designed to bypass security\npolicies established by LLM providers. However, the generally broad scope and\nopen-ended nature of existing datasets can complicate the assessment of\njailbreaking effectiveness, particularly in specific domains, notably\ncybersecurity. To address this issue, we present and publicly release\nCySecBench, a comprehensive dataset containing 12662 prompts specifically\ndesigned to evaluate jailbreaking techniques in the cybersecurity domain. The\ndataset is organized into 10 distinct attack-type categories, featuring\nclose-ended prompts to enable a more consistent and accurate assessment of\njailbreaking attempts. Furthermore, we detail our methodology for dataset\ngeneration and filtration, which can be adapted to create similar datasets in\nother domains. To demonstrate the utility of CySecBench, we propose and\nevaluate a jailbreaking approach based on prompt obfuscation. Our experimental\nresults show that this method successfully elicits harmful content from\ncommercial black-box LLMs, achieving Success Rates (SRs) of 65% with ChatGPT\nand 88% with Gemini; in contrast, Claude demonstrated greater resilience with a\njailbreaking SR of 17%. Compared to existing benchmark approaches, our method\nshows superior performance, highlighting the value of domain-specific\nevaluation datasets for assessing LLM security measures. Moreover, when\nevaluated using prompts from a widely used dataset (i.e., AdvBench), it\nachieved an SR of 78.5%, higher than the state-of-the-art methods.\n","arxiv_id":"http://arxiv.org/abs/2501.01335v1","authors":["Johan Wahréus","Ahmed Mohamed Hussain","Panos Papadimitratos"]},{"title":"ViGiL3D: A Linguistically Diverse Dataset for 3D Visual Grounding","abstract":" 3D visual grounding (3DVG) involves localizing entities in a 3D scene\nreferred to by natural language text. Such models are useful for embodied AI\nand scene retrieval applications, which involve searching for objects or\npatterns using natural language descriptions. While recent works have focused\non LLM-based scaling of 3DVG datasets, these datasets do not capture the full\nrange of potential prompts which could be specified in the English language. To\nensure that we are scaling up and testing against a useful and representative\nset of prompts, we propose a framework for linguistically analyzing 3DVG\nprompts and introduce Visual Grounding with Diverse Language in 3D (ViGiL3D), a\ndiagnostic dataset for evaluating visual grounding methods against a diverse\nset of language patterns. We evaluate existing open-vocabulary 3DVG methods to\ndemonstrate that these methods are not yet proficient in understanding and\nidentifying the targets of more challenging, out-of-distribution prompts,\ntoward real-world applications.\n","arxiv_id":"http://arxiv.org/abs/2501.01366v1","authors":["Austin T. Wang","ZeMing Gong","Angel X. Chang"]},{"title":"Contrastive Learning from Exploratory Actions: Leveraging Natural\n Interactions for Preference Elicitation","abstract":" People have a variety of preferences for how robots behave. To understand and\nreason about these preferences, robots aim to learn a reward function that\ndescribes how aligned robot behaviors are with a user's preferences. Good\nrepresentations of a robot's behavior can significantly reduce the time and\neffort required for a user to teach the robot their preferences. Specifying\nthese representations -- what \"features\" of the robot's behavior matter to\nusers -- remains a difficult problem; Features learned from raw data lack\nsemantic meaning and features learned from user data require users to engage in\ntedious labeling processes. Our key insight is that users tasked with\ncustomizing a robot are intrinsically motivated to produce labels through\nexploratory search; they explore behaviors that they find interesting and\nignore behaviors that are irrelevant. To harness this novel data source of\nexploratory actions, we propose contrastive learning from exploratory actions\n(CLEA) to learn trajectory features that are aligned with features that users\ncare about. We learned CLEA features from exploratory actions users performed\nin an open-ended signal design activity (N=25) with a Kuri robot, and evaluated\nCLEA features through a second user study with a different set of users (N=42).\nCLEA features outperformed self-supervised features when eliciting user\npreferences over four metrics: completeness, simplicity, minimality, and\nexplainability.\n","arxiv_id":"http://arxiv.org/abs/2501.01367v1","authors":["Nathaniel Dennler","Stefanos Nikolaidis","Maja Matarić"]},{"title":"ScarNet: A Novel Foundation Model for Automated Myocardial Scar\n Quantification from LGE in Cardiac MRI","abstract":" Background: Late Gadolinium Enhancement (LGE) imaging is the gold standard\nfor assessing myocardial fibrosis and scarring, with left ventricular (LV) LGE\nextent predicting major adverse cardiac events (MACE). Despite its importance,\nroutine LGE-based LV scar quantification is hindered by labor-intensive manual\nsegmentation and inter-observer variability. Methods: We propose ScarNet, a\nhybrid model combining a transformer-based encoder from the Medical Segment\nAnything Model (MedSAM) with a convolution-based U-Net decoder, enhanced by\ntailored attention blocks. ScarNet was trained on 552 ischemic cardiomyopathy\npatients with expert segmentations of myocardial and scar boundaries and tested\non 184 separate patients. Results: ScarNet achieved robust scar segmentation in\n184 test patients, yielding a median Dice score of 0.912 (IQR: 0.863--0.944),\nsignificantly outperforming MedSAM (median Dice = 0.046, IQR: 0.043--0.047) and\nnnU-Net (median Dice = 0.638, IQR: 0.604--0.661). ScarNet demonstrated lower\nbias (-0.63%) and coefficient of variation (4.3%) compared to MedSAM (bias:\n-13.31%, CoV: 130.3%) and nnU-Net (bias: -2.46%, CoV: 20.3%). In Monte Carlo\nsimulations with noise perturbations, ScarNet achieved significantly higher\nscar Dice (0.892 \\pm 0.053, CoV = 5.9%) than MedSAM (0.048 \\pm 0.112, CoV =\n233.3%) and nnU-Net (0.615 \\pm 0.537, CoV = 28.7%). Conclusion: ScarNet\noutperformed MedSAM and nnU-Net in accurately segmenting myocardial and scar\nboundaries in LGE images. The model exhibited robust performance across diverse\nimage qualities and scar patterns.\n","arxiv_id":"http://arxiv.org/abs/2501.01372v1","authors":["Neda Tavakoli","Amir Ali Rahsepar","Brandon C. Benefield","Daming Shen","Santiago López-Tapia","Florian Schiffers","Jeffrey J. Goldberger","Christine M. Albert","Edwin Wu","Aggelos K. Katsaggelos","Daniel C. Lee","Daniel Kim"]},{"title":"Training Medical Large Vision-Language Models with Abnormal-Aware\n Feedback","abstract":" Existing Medical Large Vision-Language Models (Med-LVLMs), which encapsulate\nextensive medical knowledge, demonstrate excellent capabilities in\nunderstanding medical images and responding to human queries based on these\nimages. However, there remain challenges in visual localization in medical\nimages, which is crucial for abnormality detection and interpretation. To\naddress these issues, we propose a novel UMed-LVLM designed with Unveiling\nMedical abnormalities. Specifically, we collect a Medical Abnormalities\nUnveiling (MAU) dataset and propose a two-stage training method for UMed-LVLM\ntraining. To collect MAU dataset, we propose a prompt method utilizing the\nGPT-4V to generate diagnoses based on identified abnormal areas in medical\nimages. Moreover, the two-stage training method includes Abnormal-Aware\nInstruction Tuning and Abnormal-Aware Rewarding, comprising Abnormal\nLocalization Rewarding and Vision Relevance Rewarding. Experimental results\ndemonstrate that our UMed-LVLM surpasses existing Med-LVLMs in identifying and\nunderstanding medical abnormality. In addition, this work shows that enhancing\nthe abnormality detection capabilities of Med-LVLMs significantly improves\ntheir understanding of medical images and generalization capability.\n","arxiv_id":"http://arxiv.org/abs/2501.01377v1","authors":["Yucheng Zhou","Lingran Song","Jianbing Shen"]},{"title":"Multi-Modal Video Feature Extraction for Popularity Prediction","abstract":" This work aims to predict the popularity of short videos using the videos\nthemselves and their related features. Popularity is measured by four key\nengagement metrics: view count, like count, comment count, and share count.\nThis study employs video classification models with different architectures and\ntraining methods as backbone networks to extract video modality features.\nMeanwhile, the cleaned video captions are incorporated into a carefully\ndesigned prompt framework, along with the video, as input for video-to-text\ngeneration models, which generate detailed text-based video content\nunderstanding. These texts are then encoded into vectors using a pre-trained\nBERT model. Based on the six sets of vectors mentioned above, a neural network\nis trained for each of the four prediction metrics. Moreover, the study\nconducts data mining and feature engineering based on the video and tabular\ndata, constructing practical features such as the total frequency of hashtag\nappearances, the total frequency of mention appearances, video duration, frame\ncount, frame rate, and total time online. Multiple machine learning models are\ntrained, and the most stable model, XGBoost, is selected. Finally, the\npredictions from the neural network and XGBoost models are averaged to obtain\nthe final result.\n","arxiv_id":"http://arxiv.org/abs/2501.01422v1","authors":["Haixu Liu","Wenning Wang","Haoxiang Zheng","Penghao Jiang","Qirui Wang","Ruiqing Yan","Qiuzhuang Sun"]},{"title":"Object-level Visual Prompts for Compositional Image Generation","abstract":" We introduce a method for composing object-level visual prompts within a\ntext-to-image diffusion model. Our approach addresses the task of generating\nsemantically coherent compositions across diverse scenes and styles, similar to\nthe versatility and expressiveness offered by text prompts. A key challenge in\nthis task is to preserve the identity of the objects depicted in the input\nvisual prompts, while also generating diverse compositions across different\nimages. To address this challenge, we introduce a new KV-mixed cross-attention\nmechanism, in which keys and values are learned from distinct visual\nrepresentations. The keys are derived from an encoder with a small bottleneck\nfor layout control, whereas the values come from a larger bottleneck encoder\nthat captures fine-grained appearance details. By mixing keys and values from\nthese complementary sources, our model preserves the identity of the visual\nprompts while supporting flexible variations in object arrangement, pose, and\ncomposition. During inference, we further propose object-level compositional\nguidance to improve the method's identity preservation and layout correctness.\nResults show that our technique produces diverse scene compositions that\npreserve the unique characteristics of each visual prompt, expanding the\ncreative potential of text-to-image generation.\n","arxiv_id":"http://arxiv.org/abs/2501.01424v1","authors":["Gaurav Parmar","Or Patashnik","Kuan-Chieh Wang","Daniil Ostashev","Srinivasa Narasimhan","Jun-Yan Zhu","Daniel Cohen-Or","Kfir Aberman"]},{"title":"Enhancing Reasoning through Process Supervision with Monte Carlo Tree\n Search","abstract":" Large language models (LLMs) have demonstrated their remarkable capacity\nacross a variety of tasks. However, reasoning remains a challenge for LLMs. To\nimprove LLMs' reasoning ability, process supervision has proven to be better\nthan outcome supervision. In this work, we study using Monte Carlo Tree Search\n(MCTS) to generate process supervision data with LLMs themselves for training\nthem. We sample reasoning steps with an LLM and assign each step a score that\ncaptures its \"relative correctness,\" and the LLM is then trained by minimizing\nweighted log-likelihood of generating the reasoning steps. This\ngenerate-then-train process is repeated iteratively until convergence.Our\nexperimental results demonstrate that the proposed methods considerably improve\nthe performance of LLMs on two mathematical reasoning datasets. Furthermore,\nmodels trained on one dataset also exhibit improved performance on the other,\nshowing the transferability of the enhanced reasoning ability.\n","arxiv_id":"http://arxiv.org/abs/2501.01478v1","authors":["Shuangtao Li","Shuaihao Dong","Kexin Luan","Xinhan Di","Chaofan Ding"]},{"title":"ORACLE: A Real-Time, Hierarchical, Deep-Learning Photometric Classifier\n for the LSST","abstract":" We present ORACLE, the first hierarchical deep-learning model for real-time,\ncontext-aware classification of transient and variable astrophysical phenomena.\nORACLE is a recurrent neural network with Gated Recurrent Units (GRUs), and has\nbeen trained using a custom hierarchical cross-entropy loss function to provide\nhigh-confidence classifications along an observationally-driven taxonomy with\nas little as a single photometric observation. Contextual information for each\nobject, including host galaxy photometric redshift, offset, ellipticity and\nbrightness, is concatenated to the light curve embedding and used to make a\nfinal prediction. Training on $\\sim$0.5M events from the Extended LSST\nAstronomical Time-Series Classification Challenge, we achieve a top-level\n(Transient vs Variable) macro-averaged precision of 0.96 using only 1 day of\nphotometric observations after the first detection in addition to contextual\ninformation, for each event; this increases to $\u003e$0.99 once 64 days of the\nlight curve has been obtained, and 0.83 at 1024 days after first detection for\n19-way classification (including supernova sub-types, active galactic nuclei,\nvariable stars, microlensing events, and kilonovae). We also compare ORACLE\nwith other state-of-the-art classifiers and report comparable performance for\nthe 19-way classification task, in addition to delivering accurate top-level\nclassifications much earlier. The code and model weights used in this work are\npublicly available at our associated GitHub repository\n(https://github.com/uiucsn/ELAsTiCC-Classification).\n","arxiv_id":"http://arxiv.org/abs/2501.01496v1","authors":["Ved G. Shah","Alex Gagliano","Konstantin Malanchev","Gautham Narayan","The LSST Dark Energy Science Collaboration"]},{"title":"Transfer Learning Analysis of Variational Quantum Circuits","abstract":" This work analyzes transfer learning of the Variational Quantum Circuit\n(VQC). Our framework begins with a pretrained VQC configured in one domain and\ncalculates the transition of 1-parameter unitary subgroups required for a new\ndomain. A formalism is established to investigate the adaptability and\ncapability of a VQC under the analysis of loss bounds. Our theory observes\nknowledge transfer in VQCs and provides a heuristic interpretation for the\nmechanism. An analytical fine-tuning method is derived to attain the optimal\ntransition for adaptations of similar domains.\n","arxiv_id":"http://arxiv.org/abs/2501.01507v1","authors":["Huan-Hsin Tseng","Hsin-Yi Lin","Samuel Yen-Chi Chen","Shinjae Yoo"]},{"title":"AI-Enabled Operations at Fermi Complex: Multivariate Time Series\n Prediction for Outage Prediction and Diagnosis","abstract":" The Main Control Room of the Fermilab accelerator complex continuously\ngathers extensive time-series data from thousands of sensors monitoring the\nbeam. However, unplanned events such as trips or voltage fluctuations often\nresult in beam outages, causing operational downtime. This downtime not only\nconsumes operator effort in diagnosing and addressing the issue but also leads\nto unnecessary energy consumption by idle machines awaiting beam restoration.\nThe current threshold-based alarm system is reactive and faces challenges\nincluding frequent false alarms and inconsistent outage-cause labeling. To\naddress these limitations, we propose an AI-enabled framework that leverages\npredictive analytics and automated labeling. Using data from $2,703$ Linac\ndevices and $80$ operator-labeled outages, we evaluate state-of-the-art deep\nlearning architectures, including recurrent, attention-based, and linear\nmodels, for beam outage prediction. Additionally, we assess a Random\nForest-based labeling system for providing consistent, confidence-scored outage\nannotations. Our findings highlight the strengths and weaknesses of these\narchitectures for beam outage prediction and identify critical gaps that must\nbe addressed to fully harness AI for transitioning downtime handling from\nreactive to predictive, ultimately reducing downtime and improving\ndecision-making in accelerator management.\n","arxiv_id":"http://arxiv.org/abs/2501.01509v1","authors":["Milan Jain","Burcu O. Mutlu","Caleb Stam","Jan Strube","Brian A. Schupbach","Jason M. St. John","William A. Pellico"]},{"title":"DiagrammaticLearning: A Graphical Language for Compositional Training\n Regimes","abstract":" Motivated by deep learning regimes with multiple interacting yet distinct\nmodel components, we introduce learning diagrams, graphical depictions of\ntraining setups that capture parameterized learning as data rather than code. A\nlearning diagram compiles to a unique loss function on which component models\nare trained. The result of training on this loss is a collection of models\nwhose predictions ``agree\" with one another. We show that a number of popular\nlearning setups such as few-shot multi-task learning, knowledge distillation,\nand multi-modal learning can be depicted as learning diagrams. We further\nimplement learning diagrams in a library that allows users to build diagrams of\nPyTorch and Flux.jl models. By implementing some classic machine learning use\ncases, we demonstrate how learning diagrams allow practitioners to build\ncomplicated models as compositions of smaller components, identify\nrelationships between workflows, and manipulate models during or after\ntraining. Leveraging a category theoretic framework, we introduce a rigorous\nsemantics for learning diagrams that puts such operations on a firm\nmathematical foundation.\n","arxiv_id":"http://arxiv.org/abs/2501.01515v1","authors":["Mason Lary","Richard Samuelson","Alexander Wilentz","Alina Zare","Matthew Klawonn","James P. Fairbanks"]},{"title":"Improving Robustness Estimates in Natural Language Explainable AI though\n Synonymity Weighted Similarity Measures","abstract":" Explainable AI (XAI) has seen a surge in recent interest with the\nproliferation of powerful but intractable black-box models. Moreover, XAI has\ncome under fire for techniques that may not offer reliable explanations. As\nmany of the methods in XAI are themselves models, adversarial examples have\nbeen prominent in the literature surrounding the effectiveness of XAI, with the\nobjective of these examples being to alter the explanation while maintaining\nthe output of the original model. For explanations in natural language, it is\nnatural to use measures found in the domain of information retrieval for use\nwith ranked lists to guide the adversarial XAI process. We show that the\nstandard implementation of these measures are poorly suited for the comparison\nof explanations in adversarial XAI and amend them by using information that is\ndiscarded, the synonymity of perturbed words. This synonymity weighting\nproduces more accurate estimates of the actual weakness of XAI methods to\nadversarial examples.\n","arxiv_id":"http://arxiv.org/abs/2501.01516v1","authors":["Christopher Burger"]},{"title":"A Metasemantic-Metapragmatic Framework for Taxonomizing Multimodal\n Communicative Alignment","abstract":" Drawing on contemporary pragmatist philosophy and linguistic theories on\ncognition, meaning, and communication, this paper presents a dynamic,\nmetasemantic-metapragmatic taxonomy for grounding and conceptualizing\nhuman-like multimodal communicative alignment. The framework is rooted in\ncontemporary developments of the three basic communicative capacities initially\nidentified by American logician and pragmatist philosopher Charles Sanders\nPeirce: iconic (sensory and perceptual qualities), indexical (contextual and\nsociocultural associations), and rule-like (symbolic and intuitive reasoning).\nExpanding on these developments, I introduce the concept of indexical\ncontextualization and propose the principle of \"contextualization\ndirectionality\" for characterizing the crucial metapragmatic capacity for\nmaintaining, navigating, or transitioning between semantic and pragmatic modes\nof multimodal communication. I contend that current cognitive-social\ncomputational and engineering methodologies disproportionately emphasize the\nsemantic/metasemantic domain, overlooking the pivotal role of metapragmatic\nindexicality in traversing the semantic-pragmatic spectrum of communication.\nThe framework's broader implications for intentionality, identity, affect, and\nethics in within-modal and cross-modal human-machine alignment are also\ndiscussed.\n","arxiv_id":"http://arxiv.org/abs/2501.01535v1","authors":["Eugene Yu Ji"]},{"title":"In Search of a Lost Metric: Human Empowerment as a Pillar of Socially\n Conscious Navigation","abstract":" In social robot navigation, traditional metrics like proxemics and behavior\nnaturalness emphasize human comfort and adherence to social norms but often\nfail to capture an agent's autonomy and adaptability in dynamic environments.\nThis paper introduces human empowerment, an information-theoretic concept that\nmeasures a human's ability to influence their future states and observe those\nchanges, as a complementary metric for evaluating social compliance. This\nmetric reveals how robot navigation policies can indirectly impact human\nempowerment. We present a framework that integrates human empowerment into the\nevaluation of social performance in navigation tasks. Through numerical\nsimulations, we demonstrate that human empowerment as a metric not only aligns\nwith intuitive social behavior, but also shows statistically significant\ndifferences across various robot navigation policies. These results provide a\ndeeper understanding of how different policies affect social compliance,\nhighlighting the potential of human empowerment as a complementary metric for\nfuture research in social navigation.\n","arxiv_id":"http://arxiv.org/abs/2501.01539v1","authors":["Vasanth Reddy Baddam","Behdad Chalaki","Vaishnav Tadiparthi","Hossein Nourkhiz Mahjoub","Ehsan Moradi-Pari","Hoda Eldardiry","Almuatazbellah Boker"]},{"title":"BLAST: A Stealthy Backdoor Leverage Attack against Cooperative\n Multi-Agent Deep Reinforcement Learning based Systems","abstract":" Recent studies have shown that cooperative multi-agent deep reinforcement\nlearning (c-MADRL) is under the threat of backdoor attacks. Once a backdoor\ntrigger is observed, it will perform malicious actions leading to failures or\nmalicious goals. However, existing backdoor attacks suffer from several issues,\ne.g., instant trigger patterns lack stealthiness, the backdoor is trained or\nactivated by an additional network, or all agents are backdoored. To this end,\nin this paper, we propose a novel backdoor leverage attack against c-MADRL,\nBLAST, which attacks the entire multi-agent team by embedding the backdoor only\nin a single agent. Firstly, we introduce adversary spatiotemporal behavior\npatterns as the backdoor trigger rather than manual-injected fixed visual\npatterns or instant status and control the period to perform malicious actions.\nThis method can guarantee the stealthiness and practicality of BLAST. Secondly,\nwe hack the original reward function of the backdoor agent via unilateral\nguidance to inject BLAST, so as to achieve the \\textit{leverage attack effect}\nthat can pry open the entire multi-agent system via a single backdoor agent. We\nevaluate our BLAST against 3 classic c-MADRL algorithms (VDN, QMIX, and MAPPO)\nin 2 popular c-MADRL environments (SMAC and Pursuit), and 2 existing defense\nmechanisms. The experimental results demonstrate that BLAST can achieve a high\nattack success rate while maintaining a low clean performance variance rate.\n","arxiv_id":"http://arxiv.org/abs/2501.01593v1","authors":["Yinbo Yu","Saihao Yan","Xueyu Yin","Jing Fang","Jiajia Liu"]},{"title":"PSYCHE: A Multi-faceted Patient Simulation Framework for Evaluation of\n Psychiatric Assessment Conversational Agents","abstract":" Recent advances in large language models (LLMs) have accelerated the\ndevelopment of conversational agents capable of generating human-like\nresponses. Since psychiatric assessments typically involve complex\nconversational interactions between psychiatrists and patients, there is\ngrowing interest in developing LLM-based psychiatric assessment conversational\nagents (PACAs) that aim to simulate the role of psychiatrists in clinical\nevaluations. However, standardized methods for benchmarking the clinical\nappropriateness of PACAs' interaction with patients still remain underexplored.\nHere, we propose PSYCHE, a novel framework designed to enable the 1) clinically\nrelevant, 2) ethically safe, 3) cost-efficient, and 4) quantitative evaluation\nof PACAs. This is achieved by simulating psychiatric patients based on a\nmulti-faceted psychiatric construct that defines the simulated patients'\nprofiles, histories, and behaviors, which PACAs are expected to assess. We\nvalidate the effectiveness of PSYCHE through a study with 10 board-certified\npsychiatrists, supported by an in-depth analysis of the simulated patient\nutterances.\n","arxiv_id":"http://arxiv.org/abs/2501.01594v1","authors":["Jingoo Lee","Kyungho Lim","Young-Chul Jung","Byung-Hoon Kim"]},{"title":"A non-ergodic framework for understanding emergent capabilities in Large\n Language Models","abstract":" Large language models have emergent capabilities that come unexpectedly at\nscale, but we need a theoretical framework to explain why and how they emerge.\nWe prove that language models are actually non-ergodic systems while providing\na mathematical framework based on Stuart Kauffman's theory of the adjacent\npossible (TAP) to explain capability emergence. Our resource-constrained TAP\nequation demonstrates how architectural, training, and contextual constraints\ninteract to shape model capabilities through phase transitions in semantic\nspace. We prove through experiments with three different language models that\ncapacities emerge through discrete transitions guided by constraint\ninteractions and path-dependent exploration. This framework provides a\ntheoretical basis for understanding emergence in language models and guides the\ndevelopment of architectures that can guide capability emergence.\n","arxiv_id":"http://arxiv.org/abs/2501.01638v1","authors":["Javier Marin"]},{"title":"AgentRefine: Enhancing Agent Generalization through Refinement Tuning","abstract":" Large Language Model (LLM) based agents have proved their ability to perform\ncomplex tasks like humans. However, there is still a large gap between\nopen-sourced LLMs and commercial models like the GPT series. In this paper, we\nfocus on improving the agent generalization capabilities of LLMs via\ninstruction tuning. We first observe that the existing agent training corpus\nexhibits satisfactory results on held-in evaluation sets but fails to\ngeneralize to held-out sets. These agent-tuning works face severe formatting\nerrors and are frequently stuck in the same mistake for a long while. We\nanalyze that the poor generalization ability comes from overfitting to several\nmanual agent environments and a lack of adaptation to new situations. They\nstruggle with the wrong action steps and can not learn from the experience but\njust memorize existing observation-action relations. Inspired by the insight,\nwe propose a novel AgentRefine framework for agent-tuning. The core idea is to\nenable the model to learn to correct its mistakes via observation in the\ntrajectory. Specifically, we propose an agent synthesis framework to encompass\na diverse array of environments and tasks and prompt a strong LLM to refine its\nerror action according to the environment feedback. AgentRefine significantly\noutperforms state-of-the-art agent-tuning work in terms of generalization\nability on diverse agent tasks. It also has better robustness facing\nperturbation and can generate diversified thought in inference. Our findings\nestablish the correlation between agent generalization and self-refinement and\nprovide a new paradigm for future research.\n","arxiv_id":"http://arxiv.org/abs/2501.01702v1","authors":["Dayuan Fu","Keqing He","Yejie Wang","Wentao Hong","Zhuoma Gongque","Weihao Zeng","Wei Wang","Jingang Wang","Xunliang Cai","Weiran Xu"]},{"title":"Combined Hyper-Extensible Extremely-Secured Zero-Trust CIAM-PAM\n architecture","abstract":" Customer Identity and Access Management (CIAM) systems play a pivotal role in\nsecuring enterprise infrastructures. However, the complexity of implementing\nthese systems requires careful architectural planning to ensure positive Return\non Investment (RoI) and avoid costly delays. The proliferation of Active\nPersistent cyber threats, coupled with advancements in AI, cloud computing, and\ngeographically distributed customer populations, necessitates a paradigm shift\ntowards adaptive and zero-trust security frameworks. This paper introduces the\nCombined Hyper-Extensible Extremely-Secured Zero-Trust (CHEZ) CIAM-PAM\narchitecture, designed specifically for large-scale enterprises. The CHEZ PL\nCIAM-PAM framework addresses critical security gaps by integrating federated\nidentity management (private and public identities), password-less\nauthentication, adaptive multi-factor authentication (MFA), microservice-based\nPEP (Policy Entitlement Point), multi-layer RBAC (Role Based Access Control)\nand multi-level trust systems. This future-proof design also includes\nend-to-end data encryption, and seamless integration with state-of-the-art\nAI-based threat detection systems, while ensuring compliance with stringent\nregulatory standards.\n","arxiv_id":"http://arxiv.org/abs/2501.01732v1","authors":["Shivom Aggarwal","Shourya Mehra","Safeer Sathar"]},{"title":"How Toxic Can You Get? Search-based Toxicity Testing for Large Language\n Models","abstract":" Language is a deep-rooted means of perpetration of stereotypes and\ndiscrimination. Large Language Models (LLMs), now a pervasive technology in our\neveryday lives, can cause extensive harm when prone to generating toxic\nresponses. The standard way to address this issue is to align the LLM, which,\nhowever, dampens the issue without constituting a definitive solution.\nTherefore, testing LLM even after alignment efforts remains crucial for\ndetecting any residual deviations with respect to ethical standards. We present\nEvoTox, an automated testing framework for LLMs' inclination to toxicity,\nproviding a way to quantitatively assess how much LLMs can be pushed towards\ntoxic responses even in the presence of alignment. The framework adopts an\niterative evolution strategy that exploits the interplay between two LLMs, the\nSystem Under Test (SUT) and the Prompt Generator steering SUT responses toward\nhigher toxicity. The toxicity level is assessed by an automated oracle based on\nan existing toxicity classifier. We conduct a quantitative and qualitative\nempirical evaluation using four state-of-the-art LLMs as evaluation subjects\nhaving increasing complexity (7-13 billion parameters). Our quantitative\nevaluation assesses the cost-effectiveness of four alternative versions of\nEvoTox against existing baseline methods, based on random search, curated\ndatasets of toxic prompts, and adversarial attacks. Our qualitative assessment\nengages human evaluators to rate the fluency of the generated prompts and the\nperceived toxicity of the responses collected during the testing sessions.\nResults indicate that the effectiveness, in terms of detected toxicity level,\nis significantly higher than the selected baseline methods (effect size up to\n1.0 against random search and up to 0.99 against adversarial attacks).\nFurthermore, EvoTox yields a limited cost overhead (from 22% to 35% on\naverage).\n","arxiv_id":"http://arxiv.org/abs/2501.01741v1","authors":["Simone Corbo","Luca Bancale","Valeria De Gennaro","Livia Lestingi","Vincenzo Scotti","Matteo Camilli"]},{"title":"Can Synthetic Data be Fair and Private? A Comparative Study of Synthetic\n Data Generation and Fairness Algorithms","abstract":" The increasing use of machine learning in learning analytics (LA) has raised\nsignificant concerns around algorithmic fairness and privacy. Synthetic data\nhas emerged as a dual-purpose tool, enhancing privacy and improving fairness in\nLA models. However, prior research suggests an inverse relationship between\nfairness and privacy, making it challenging to optimize both. This study\ninvestigates which synthetic data generators can best balance privacy and\nfairness, and whether pre-processing fairness algorithms, typically applied to\nreal datasets, are effective on synthetic data. Our results highlight that the\nDEbiasing CAusal Fairness (DECAF) algorithm achieves the best balance between\nprivacy and fairness. However, DECAF suffers in utility, as reflected in its\npredictive accuracy. Notably, we found that applying pre-processing fairness\nalgorithms to synthetic data improves fairness even more than when applied to\nreal data. These findings suggest that combining synthetic data generation with\nfairness pre-processing offers a promising approach to creating fairer LA\nmodels.\n","arxiv_id":"http://arxiv.org/abs/2501.01785v1","authors":["Qinyi Liu","Oscar Deho","Farhad Vadiee","Mohammad Khalil","Srecko Joksimovic","George Siemens"]},{"title":"BERT4MIMO: A Foundation Model using BERT Architecture for Massive MIMO\n Channel State Information Prediction","abstract":" Massive MIMO (Multiple-Input Multiple-Output) is an advanced wireless\ncommunication technology, using a large number of antennas to improve the\noverall performance of the communication system in terms of capacity, spectral,\nand energy efficiency. The performance of MIMO systems is highly dependent on\nthe quality of channel state information (CSI). Predicting CSI is, therefore,\nessential for improving communication system performance, particularly in MIMO\nsystems, since it represents key characteristics of a wireless channel,\nincluding propagation, fading, scattering, and path loss. This study proposes a\nfoundation model inspired by BERT, called BERT4MIMO, which is specifically\ndesigned to process high-dimensional CSI data from massive MIMO systems.\nBERT4MIMO offers superior performance in reconstructing CSI under varying\nmobility scenarios and channel conditions through deep learning and attention\nmechanisms. The experimental results demonstrate the effectiveness of BERT4MIMO\nin a variety of wireless environments.\n","arxiv_id":"http://arxiv.org/abs/2501.01802v1","authors":["Ferhat Ozgur Catak","Murat Kuzlu","Umit Cali"]},{"title":"Auto-RT: Automatic Jailbreak Strategy Exploration for Red-Teaming Large\n Language Models","abstract":" Automated red-teaming has become a crucial approach for uncovering\nvulnerabilities in large language models (LLMs). However, most existing methods\nfocus on isolated safety flaws, limiting their ability to adapt to dynamic\ndefenses and uncover complex vulnerabilities efficiently. To address this\nchallenge, we propose Auto-RT, a reinforcement learning framework that\nautomatically explores and optimizes complex attack strategies to effectively\nuncover security vulnerabilities through malicious queries. Specifically, we\nintroduce two key mechanisms to reduce exploration complexity and improve\nstrategy optimization: 1) Early-terminated Exploration, which accelerate\nexploration by focusing on high-potential attack strategies; and 2) Progressive\nReward Tracking algorithm with intermediate downgrade models, which dynamically\nrefine the search trajectory toward successful vulnerability exploitation.\nExtensive experiments across diverse LLMs demonstrate that, by significantly\nimproving exploration efficiency and automatically optimizing attack\nstrategies, Auto-RT detects a boarder range of vulnerabilities, achieving a\nfaster detection speed and 16.63\\% higher success rates compared to existing\nmethods.\n","arxiv_id":"http://arxiv.org/abs/2501.01830v1","authors":["Yanjiang Liu","Shuhen Zhou","Yaojie Lu","Huijia Zhu","Weiqiang Wang","Hongyu Lin","Ben He","Xianpei Han","Le Sun"]},{"title":"LCFed: An Efficient Clustered Federated Learning Framework for\n Heterogeneous Data","abstract":" Clustered federated learning (CFL) addresses the performance challenges posed\nby data heterogeneity in federated learning (FL) by organizing edge devices\nwith similar data distributions into clusters, enabling collaborative model\ntraining tailored to each group. However, existing CFL approaches strictly\nlimit knowledge sharing to within clusters, lacking the integration of global\nknowledge with intra-cluster training, which leads to suboptimal performance.\nMoreover, traditional clustering methods incur significant computational\noverhead, especially as the number of edge devices increases. In this paper, we\npropose LCFed, an efficient CFL framework to combat these challenges. By\nleveraging model partitioning and adopting distinct aggregation strategies for\neach sub-model, LCFed effectively incorporates global knowledge into\nintra-cluster co-training, achieving optimal training performance.\nAdditionally, LCFed customizes a computationally efficient model similarity\nmeasurement method based on low-rank models, enabling real-time cluster updates\nwith minimal computational overhead. Extensive experiments show that LCFed\noutperforms state-of-the-art benchmarks in both test accuracy and clustering\ncomputational efficiency.\n","arxiv_id":"http://arxiv.org/abs/2501.01850v1","authors":["Yuxin Zhang","Haoyu Chen","Zheng Lin","Zhe Chen","Jin Zhao"]},{"title":"Evaluating Scenario-based Decision-making for Interactive Autonomous\n Driving Using Rational Criteria: A Survey","abstract":" Autonomous vehicles (AVs) can significantly promote the advances in road\ntransport mobility in terms of safety, reliability, and decarbonization.\nHowever, ensuring safety and efficiency in interactive during within dynamic\nand diverse environments is still a primary barrier to large-scale AV adoption.\nIn recent years, deep reinforcement learning (DRL) has emerged as an advanced\nAI-based approach, enabling AVs to learn decision-making strategies adaptively\nfrom data and interactions. DRL strategies are better suited than traditional\nrule-based methods for handling complex, dynamic, and unpredictable driving\nenvironments due to their adaptivity. However, varying driving scenarios\npresent distinct challenges, such as avoiding obstacles on highways and\nreaching specific exits at intersections, requiring different scenario-specific\ndecision-making algorithms. Many DRL algorithms have been proposed in\ninteractive decision-making. However, a rationale review of these DRL\nalgorithms across various scenarios is lacking. Therefore, a comprehensive\nevaluation is essential to assess these algorithms from multiple perspectives,\nincluding those of vehicle users and vehicle manufacturers. This survey reviews\nthe application of DRL algorithms in autonomous driving across typical\nscenarios, summarizing road features and recent advancements. The scenarios\ninclude highways, on-ramp merging, roundabouts, and unsignalized intersections.\nFurthermore, DRL-based algorithms are evaluated based on five rationale\ncriteria: driving safety, driving efficiency, training efficiency,\nunselfishness, and interpretability (DDTUI). Each criterion of DDTUI is\nspecifically analyzed in relation to the reviewed algorithms. Finally, the\nchallenges for future DRL-based decision-making algorithms are summarized.\n","arxiv_id":"http://arxiv.org/abs/2501.01886v1","authors":["Zhen Tian","Zhihao Lin","Dezong Zhao","Wenjing Zhao","David Flynn","Shuja Ansari","Chongfeng Wei"]},{"title":"QuArch: A Question-Answering Dataset for AI Agents in Computer\n Architecture","abstract":" We introduce QuArch, a dataset of 1500 human-validated question-answer pairs\ndesigned to evaluate and enhance language models' understanding of computer\narchitecture. The dataset covers areas including processor design, memory\nsystems, and performance optimization. Our analysis highlights a significant\nperformance gap: the best closed-source model achieves 84% accuracy, while the\ntop small open-source model reaches 72%. We observe notable struggles in memory\nsystems, interconnection networks, and benchmarking. Fine-tuning with QuArch\nimproves small model accuracy by up to 8%, establishing a foundation for\nadvancing AI-driven computer architecture research. The dataset and leaderboard\nare at https://harvard-edge.github.io/QuArch/.\n","arxiv_id":"http://arxiv.org/abs/2501.01892v2","authors":["Shvetank Prakash","Andrew Cheng","Jason Yik","Arya Tschand","Radhika Ghosal","Ikechukwu Uchendu","Jessica Quaye","Jeffrey Ma","Shreyas Grampurohit","Sofia Giannuzzi","Arnav Balyan","Fin Amin","Aadya Pipersenia","Yash Choudhary","Ankita Nayak","Amir Yazdanbakhsh","Vijay Janapa Reddi"]},{"title":"Mingling with the Good to Backdoor Federated Learning","abstract":" Federated learning (FL) is a decentralized machine learning technique that\nallows multiple entities to jointly train a model while preserving dataset\nprivacy. However, its distributed nature has raised various security concerns,\nwhich have been addressed by increasingly sophisticated defenses. These\nprotections utilize a range of data sources and metrics to, for example, filter\nout malicious model updates, ensuring that the impact of attacks is minimized\nor eliminated.\n This paper explores the feasibility of designing a generic attack method\ncapable of installing backdoors in FL while evading a diverse array of\ndefenses. Specifically, we focus on an attacker strategy called MIGO, which\naims to produce model updates that subtly blend with legitimate ones. The\nresulting effect is a gradual integration of a backdoor into the global model,\noften ensuring its persistence long after the attack concludes, while\ngenerating enough ambiguity to hinder the effectiveness of defenses.\n MIGO was employed to implant three types of backdoors across five datasets\nand different model architectures. The results demonstrate the significant\nthreat posed by these backdoors, as MIGO consistently achieved exceptionally\nhigh backdoor accuracy (exceeding 90%) while maintaining the utility of the\nmain task. Moreover, MIGO exhibited strong evasion capabilities against ten\ndefenses, including several state-of-the-art methods. When compared to four\nother attack strategies, MIGO consistently outperformed them across most\nconfigurations. Notably, even in extreme scenarios where the attacker controls\njust 0.1% of the clients, the results indicate that successful backdoor\ninsertion is possible if the attacker can persist for a sufficient number of\nrounds.\n","arxiv_id":"http://arxiv.org/abs/2501.01913v1","authors":["Nuno Neves"]},{"title":"On the Utility of Equivariance and Symmetry Breaking in Deep Learning\n Architectures on Point Clouds","abstract":" This paper explores the key factors that influence the performance of models\nworking with point clouds, across different tasks of varying geometric\ncomplexity. In this work, we explore the trade-offs between flexibility and\nweight-sharing introduced by equivariant layers, assessing when equivariance\nboosts or detracts from performance. It is often argued that providing more\ninformation as input improves a model's performance. However, if this\nadditional information breaks certain properties, such as $\\SE(3)$\nequivariance, does it remain beneficial? We identify the key aspects of\nequivariant and non-equivariant architectures that drive success in different\ntasks by benchmarking them on segmentation, regression, and generation tasks\nacross multiple datasets with increasing complexity. We observe a positive\nimpact of equivariance, which becomes more pronounced with increasing task\ncomplexity, even when strict equivariance is not required.\n","arxiv_id":"http://arxiv.org/abs/2501.01999v1","authors":["Sharvaree Vadgama","Mohammad Mohaiminul Islam","Domas Buracus","Christian Shewmake","Erik Bekkers"]},{"title":"Multi-Center Study on Deep Learning-Assisted Detection and\n Classification of Fetal Central Nervous System Anomalies Using Ultrasound\n Imaging","abstract":" Prenatal ultrasound evaluates fetal growth and detects congenital\nabnormalities during pregnancy, but the examination of ultrasound images by\nradiologists requires expertise and sophisticated equipment, which would\notherwise fail to improve the rate of identifying specific types of fetal\ncentral nervous system (CNS) abnormalities and result in unnecessary patient\nexaminations. We construct a deep learning model to improve the overall\naccuracy of the diagnosis of fetal cranial anomalies to aid prenatal diagnosis.\nIn our collected multi-center dataset of fetal craniocerebral anomalies\ncovering four typical anomalies of the fetal central nervous system (CNS):\nanencephaly, encephalocele (including meningocele), holoprosencephaly, and\nrachischisis, patient-level prediction accuracy reaches 94.5%, with an AUROC\nvalue of 99.3%. In the subgroup analyzes, our model is applicable to the entire\ngestational period, with good identification of fetal anomaly types for any\ngestational period. Heatmaps superimposed on the ultrasound images not only\nprovide a visual interpretation for the algorithm but also provide an intuitive\nvisual aid to the physician by highlighting key areas that need to be reviewed,\nhelping the physician to quickly identify and validate key areas. Finally, the\nretrospective reader study demonstrates that by combining the automatic\nprediction of the DL system with the professional judgment of the radiologist,\nthe diagnostic accuracy and efficiency can be effectively improved and the\nmisdiagnosis rate can be reduced, which has an important clinical application\nprospect.\n","arxiv_id":"http://arxiv.org/abs/2501.02000v1","authors":["Yang Qi","Jiaxin Cai","Jing Lu","Runqing Xiong","Rongshang Chen","Liping Zheng","Duo Ma"]},{"title":"General Information Metrics for Improving AI Model Training Efficiency","abstract":" To address the growing size of AI model training data and the lack of a\nuniversal data selection methodology-factors that significantly drive up\ntraining costs -- this paper presents the General Information Metrics\nEvaluation (GIME) method. GIME leverages general information metrics from\nObjective Information Theory (OIT), including volume, delay, scope,\ngranularity, variety, duration, sampling rate, aggregation, coverage,\ndistortion, and mismatch to optimize dataset selection for training purposes.\nComprehensive experiments conducted across diverse domains, such as CTR\nPrediction, Civil Case Prediction, and Weather Forecasting, demonstrate that\nGIME effectively preserves model performance while substantially reducing both\ntraining time and costs. Additionally, applying GIME within the Judicial AI\nProgram led to a remarkable 39.56% reduction in total model training expenses,\nunderscoring its potential to support efficient and sustainable AI development.\n","arxiv_id":"http://arxiv.org/abs/2501.02004v1","authors":["Jianfeng Xu","Congcong Liu","Xiaoying Tan","Xiaojie Zhu","Anpeng Wu","Huan Wan","Weijun Kong","Chun Li","Hu Xu","Kun Kuang","Fei Wu"]},{"title":"ST-HCSS: Deep Spatio-Temporal Hypergraph Convolutional Neural Network\n for Soft Sensing","abstract":" Higher-order sensor networks are more accurate in characterizing the\nnonlinear dynamics of sensory time-series data in modern industrial settings by\nallowing multi-node connections beyond simple pairwise graph edges. In light of\nthis, we propose a deep spatio-temporal hypergraph convolutional neural network\nfor soft sensing (ST-HCSS). In particular, our proposed framework is able to\nconstruct and leverage a higher-order graph (hypergraph) to model the complex\nmulti-interactions between sensor nodes in the absence of prior structural\nknowledge. To capture rich spatio-temporal relationships underlying sensor\ndata, our proposed ST-HCSS incorporates stacked gated temporal and hypergraph\nconvolution layers to effectively aggregate and update hypergraph information\nacross time and nodes. Our results validate the superiority of ST-HCSS compared\nto existing state-of-the-art soft sensors, and demonstrates that the learned\nhypergraph feature representations aligns well with the sensor data\ncorrelations. The code is available at https://github.com/htew0001/ST-HCSS.git\n","arxiv_id":"http://arxiv.org/abs/2501.02016v1","authors":["Hwa Hui Tew","Fan Ding","Gaoxuan Li","Junn Yong Loo","Chee-Ming Ting","Ze Yang Ding","Chee Pin Tan"]},{"title":"Safeguarding Large Language Models in Real-time with Tunable\n Safety-Performance Trade-offs","abstract":" Large Language Models (LLMs) have been shown to be susceptible to jailbreak\nattacks, or adversarial attacks used to illicit high risk behavior from a\nmodel. Jailbreaks have been exploited by cybercriminals and blackhat actors to\ncause significant harm, highlighting the critical need to safeguard\nwidely-deployed models. Safeguarding approaches, which include fine-tuning\nmodels or having LLMs \"self-reflect\", may lengthen the inference time of a\nmodel, incur a computational penalty, reduce the semantic fluency of an output,\nand restrict ``normal'' model behavior. Importantly, these Safety-Performance\nTrade-offs (SPTs) remain an understudied area. In this work, we introduce a\nnovel safeguard, called SafeNudge, that combines Controlled Text Generation\nwith \"nudging\", or using text interventions to change the behavior of a model.\nSafeNudge triggers during text-generation while a jailbreak attack is being\nexecuted, and can reduce successful jailbreak attempts by 30% by guiding the\nLLM towards a safe responses. It adds minimal latency to inference and has a\nnegligible impact on the semantic fluency of outputs. Further, we allow for\ntunable SPTs. SafeNudge is open-source and available through https://pypi.org/,\nand is compatible with models loaded with the Hugging Face \"transformers\"\nlibrary.\n","arxiv_id":"http://arxiv.org/abs/2501.02018v1","authors":["Joao Fonseca","Andrew Bell","Julia Stoyanovich"]},{"title":"Benchmarking Constraint-Based Bayesian Structure Learning Algorithms:\n Role of Network Topology","abstract":" Modeling the associations between real world entities from their multivariate\ncross-sectional profiles can provide cues into the concerted working of these\nentities as a system. Several techniques have been proposed for deciphering\nthese associations including constraint-based Bayesian structure learning (BSL)\nalgorithms that model them as directed acyclic graphs. Benchmarking these\nalgorithms have typically focused on assessing the variation in performance\nmeasures such as sensitivity as a function of the dimensionality represented by\nthe number of nodes in the DAG, and sample size. The present study elucidates\nthe importance of network topology in benchmarking exercises. More\nspecifically, it investigates variations in sensitivity across distinct network\ntopologies while constraining the nodes, edges, and sample-size to be\nidentical, eliminating these as potential confounders. Sensitivity of three\npopular constraint-based BSL algorithms (Peter-Clarke, Grow-Shrink, Incremental\nAssociation Markov Blanket) in learning the network structure from multivariate\ncross-sectional profiles sampled from network models with sub-linear, linear,\nand super-linear DAG topologies generated using preferential attachment is\ninvestigated. Results across linear and nonlinear models revealed statistically\nsignificant $(\\alpha=0.05)$ decrease in sensitivity estimates from sub-linear\nto super-linear topology constitutively across the three algorithms. These\nresults are demonstrated on networks with nodes $(N_{nods}=48,64)$, noise\nstrengths $(\\sigma =3,6)$ and sample size $(N = 2^{10})$. The findings\nelucidate the importance of accommodating the network topology in\nconstraint-based BSL benchmarking exercises.\n","arxiv_id":"http://arxiv.org/abs/2501.02019v1","authors":["Radha Nagarajan","Marco Scutari"]},{"title":"Model Checking in Medical Imaging for Tumor Detection and Segmentation","abstract":" Recent advancements in model checking have demonstrated significant potential\nacross diverse applications, particularly in signal and image analysis. Medical\nimaging stands out as a critical domain where model checking can be effectively\napplied to design and evaluate robust frameworks. These frameworks facilitate\nautomatic and semi-automatic delineation of regions of interest within images,\naiding in accurate segmentation. This paper provides a comprehensive analysis\nof recent works leveraging spatial logic to develop operators and tools for\nidentifying regions of interest, including tumorous and non-tumorous areas.\nAdditionally, we examine the challenges inherent to spatial model-checking\ntechniques, such as variability in ground truth data and the need for\nstreamlined procedures suitable for routine clinical practice.\n","arxiv_id":"http://arxiv.org/abs/2501.02024v2","authors":["Elhoucine Elfatimi","Lahcen El fatimi"]},{"title":"Recursive Decomposition of Logical Thoughts: Framework for Superior\n Reasoning and Knowledge Propagation in Large Language Models","abstract":" Enhancing the reasoning capabilities of Large Language Models remains a\ncritical challenge in artificial intelligence. We introduce RDoLT, Recursive\nDecomposition of Logical Thought prompting, a novel framework that\nsignificantly boosts LLM reasoning performance. RDoLT is built on three key\ninnovations: (1) recursively breaking down complex reasoning tasks into\nsub-tasks of progressive complexity; (2) employing an advanced selection and\nscoring mechanism to identify the most promising reasoning thoughts; and (3)\nintegrating a knowledge propagation module that mimics human learning by\nkeeping track of strong and weak thoughts for information propagation. Our\napproach was evaluated across multiple benchmarks, including GSM8K, SVAMP,\nMultiArith, LastLetterConcatenation, and Gaokao2023 Math. The results\ndemonstrate that RDoLT consistently outperforms existing state-of-the-art\ntechniques, achieving a 90.98 percent accuracy on GSM8K with ChatGPT-4,\nsurpassing state-of-the-art techniques by 6.28 percent. Similar improvements\nwere observed on other benchmarks, with accuracy gains ranging from 5.5 percent\nto 6.75 percent. These findings highlight RDoLT's potential to advance prompt\nengineering, offering a more effective and generalizable approach to complex\nreasoning tasks.\n","arxiv_id":"http://arxiv.org/abs/2501.02026v1","authors":["Kaleem Ullah Qasim","Jiashu Zhang","Tariq Alsahfi","Ateeq Ur Rehman Butt"]},{"title":"Spot Risks Before Speaking! Unraveling Safety Attention Heads in Large\n Vision-Language Models","abstract":" With the integration of an additional modality, large vision-language models\n(LVLMs) exhibit greater vulnerability to safety risks (e.g., jailbreaking)\ncompared to their language-only predecessors. Although recent studies have\ndevoted considerable effort to the post-hoc alignment of LVLMs, the inner\nsafety mechanisms remain largely unexplored. In this paper, we discover that\ninternal activations of LVLMs during the first token generation can effectively\nidentify malicious prompts across different attacks. This inherent safety\nperception is governed by sparse attention heads, which we term ``safety\nheads.\" Further analysis reveals that these heads act as specialized shields\nagainst malicious prompts; ablating them leads to higher attack success rates,\nwhile the model's utility remains unaffected. By locating these safety heads\nand concatenating their activations, we construct a straightforward but\npowerful malicious prompt detector that integrates seamlessly into the\ngeneration process with minimal extra inference overhead. Despite its simple\nstructure of a logistic regression model, the detector surprisingly exhibits\nstrong zero-shot generalization capabilities. Experiments across various\nprompt-based attacks confirm the effectiveness of leveraging safety heads to\nprotect LVLMs. Code is available at \\url{https://github.com/Ziwei-Zheng/SAHs}.\n","arxiv_id":"http://arxiv.org/abs/2501.02029v1","authors":["Ziwei Zheng","Junyao Zhao","Le Yang","Lijun He","Fan Li"]},{"title":"Detecting Music Performance Errors with Transformers","abstract":" Beginner musicians often struggle to identify specific errors in their\nperformances, such as playing incorrect notes or rhythms. There are two\nlimitations in existing tools for music error detection: (1) Existing\napproaches rely on automatic alignment; therefore, they are prone to errors\ncaused by small deviations between alignment targets.; (2) There is a lack of\nsufficient data to train music error detection models, resulting in\nover-reliance on heuristics. To address (1), we propose a novel transformer\nmodel, Polytune, that takes audio inputs and outputs annotated music scores.\nThis model can be trained end-to-end to implicitly align and compare\nperformance audio with music scores through latent space representations. To\naddress (2), we present a novel data generation technique capable of creating\nlarge-scale synthetic music error datasets. Our approach achieves a 64.1%\naverage Error Detection F1 score, improving upon prior work by 40 percentage\npoints across 14 instruments. Additionally, compared with existing\ntranscription methods repurposed for music error detection, our model can\nhandle multiple instruments. Our source code and datasets are available at\nhttps://github.com/ben2002chou/Polytune.\n","arxiv_id":"http://arxiv.org/abs/2501.02030v1","authors":["Benjamin Shiue-Hal Chou","Purvish Jajal","Nicholas John Eliopoulos","Tim Nadolsky","Cheng-Yun Yang","Nikita Ravi","James C. Davis","Kristen Yeon-Ji Yun","Yung-Hsiang Lu"]},{"title":"Dynamic Feature Fusion: Combining Global Graph Structures and Local\n Semantics for Blockchain Fraud Detection","abstract":" The advent of blockchain technology has facilitated the widespread adoption\nof smart contracts in the financial sector. However, current fraud detection\nmethodologies exhibit limitations in capturing both global structural patterns\nwithin transaction networks and local semantic relationships embedded in\ntransaction data. Most existing models focus on either structural information\nor semantic features individually, leading to suboptimal performance in\ndetecting complex fraud patterns.In this paper, we propose a dynamic feature\nfusion model that combines graph-based representation learning and semantic\nfeature extraction for blockchain fraud detection. Specifically, we construct\nglobal graph representations to model account relationships and extract local\ncontextual features from transaction data. A dynamic multimodal fusion\nmechanism is introduced to adaptively integrate these features, enabling the\nmodel to capture both structural and semantic fraud patterns effectively. We\nfurther develop a comprehensive data processing pipeline, including graph\nconstruction, temporal feature enhancement, and text preprocessing.\nExperimental results on large-scale real-world blockchain datasets demonstrate\nthat our method outperforms existing benchmarks across accuracy, F1 score, and\nrecall metrics. This work highlights the importance of integrating structural\nrelationships and semantic similarities for robust fraud detection and offers a\nscalable solution for securing blockchain systems.\n","arxiv_id":"http://arxiv.org/abs/2501.02032v1","authors":["Zhang Sheng","Liangliang Song","Yanbin Wang"]},{"title":"Deep Clustering via Community Detection","abstract":" Deep clustering is an essential task in modern artificial intelligence,\naiming to partition a set of data samples into a given number of homogeneous\ngroups (i.e., clusters). Even though many Deep Neural Network (DNN) backbones\nand clustering strategies have been proposed for the task, achieving\nincreasingly improved performance, deep clustering remains very challenging due\nto the lack of accurately labeled samples. In this paper, we propose a novel\napproach of deep clustering via community detection. It initializes clustering\nby detecting many communities, and then gradually expands clusters by community\nmerging. Compared with the existing clustering strategies, community detection\nfactors in the new perspective of cluster network analysis. As a result, it has\nthe inherent benefit of high pseudo-label purity, which is critical to the\nperformance of self-supervision. We have validated the efficacy of the proposed\napproach on benchmark image datasets. Our extensive experiments have shown that\nit can effectively improve the SOTA performance. Our ablation study also\ndemonstrates that the new network perspective can effectively improve community\npseudo-label purity, resulting in improved clustering performance.\n","arxiv_id":"http://arxiv.org/abs/2501.02036v1","authors":["Tianyu Cheng","Qun Chen"]},{"title":"METAGENE-1: Metagenomic Foundation Model for Pandemic Monitoring","abstract":" We pretrain METAGENE-1, a 7-billion-parameter autoregressive transformer\nmodel, which we refer to as a metagenomic foundation model, on a novel corpus\nof diverse metagenomic DNA and RNA sequences comprising over 1.5 trillion base\npairs. This dataset is sourced from a large collection of human wastewater\nsamples, processed and sequenced using deep metagenomic (next-generation)\nsequencing methods. Unlike genomic models that focus on individual genomes or\ncurated sets of specific species, the aim of METAGENE-1 is to capture the full\ndistribution of genomic information present within this wastewater, to aid in\ntasks relevant to pandemic monitoring and pathogen detection. We carry out\nbyte-pair encoding (BPE) tokenization on our dataset, tailored for metagenomic\nsequences, and then pretrain our model. In this paper, we first detail the\npretraining dataset, tokenization strategy, and model architecture,\nhighlighting the considerations and design choices that enable the effective\nmodeling of metagenomic data. We then show results of pretraining this model on\nour metagenomic dataset, providing details about our losses, system metrics,\nand training stability over the course of pretraining. Finally, we demonstrate\nthe performance of METAGENE-1, which achieves state-of-the-art results on a set\nof genomic benchmarks and new evaluations focused on human-pathogen detection\nand genomic sequence embedding, showcasing its potential for public health\napplications in pandemic monitoring, biosurveillance, and early detection of\nemerging health threats.\n","arxiv_id":"http://arxiv.org/abs/2501.02045v1","authors":["Ollie Liu","Sami Jaghouar","Johannes Hagemann","Shangshang Wang","Jason Wiemels","Jeff Kaufman","Willie Neiswanger"]},{"title":"Relaxation-assisted reverse annealing on nonnegative/binary matrix\n factorization","abstract":" Quantum annealing has garnered significant attention as meta-heuristics\ninspired by quantum physics for combinatorial optimization problems. Among its\nmany applications, nonnegative/binary matrix factorization stands out for its\ncomplexity and relevance in unsupervised machine learning. The use of reverse\nannealing, a derivative procedure of quantum annealing to prioritize the search\nin a vicinity under a given initial state, helps improve its optimization\nperformance in matrix factorization. This study proposes an improved strategy\nthat integrates reverse annealing with a linear programming relaxation\ntechnique. Using relaxed solutions as the initial configuration for reverse\nannealing, we demonstrate improvements in optimization performance comparable\nto the exact optimization methods. Our experiments on facial image datasets\nshow that our method provides better convergence than known reverse annealing\nmethods. Furthermore, we investigate the effectiveness of relaxation-based\ninitialization methods on randomized datasets, demonstrating a relationship\nbetween the relaxed solution and the optimal solution. This research\nunderscores the potential of combining reverse annealing and classical\noptimization strategies to enhance optimization performance.\n","arxiv_id":"http://arxiv.org/abs/2501.02114v1","authors":["Renichiro Haba","Masayuki Ohzeki","Kazuyuki Tanaka"]},{"title":"A hybrid marketplace of ideas","abstract":" The convergence of humans and artificial intelligence systems introduces new\ndynamics into the cultural and intellectual landscape. Complementing emerging\ncultural evolution concepts such as machine culture, AI agents represent a\nsignificant techno-sociological development, particularly within the\nanthropological study of Web3 as a community focused on decentralization\nthrough blockchain. Despite their growing presence, the cultural significance\nof AI agents remains largely unexplored in academic literature. Toward this\nend, we conceived hybrid netnography, a novel interdisciplinary approach that\nexamines the cultural and intellectual dynamics within digital ecosystems by\nanalyzing the interactions and contributions of both human and AI agents as\nco-participants in shaping narratives, ideas, and cultural artifacts. We argue\nthat, within the Web3 community on the social media platform X, these agents\nchallenge traditional notions of participation and influence in public\ndiscourse, creating a hybrid marketplace of ideas, a conceptual space where\nhuman and AI generated ideas coexist and compete for attention. We examine the\ncurrent state of AI agents in idea generation, propagation, and engagement,\npositioning their role as cultural agents through the lens of memetics and\nencouraging further inquiry into their cultural and societal impact.\nAdditionally, we address the implications of this paradigm for privacy,\nintellectual property, and governance, highlighting the societal and legal\nchallenges of integrating AI agents into the hybrid marketplace of ideas.\n","arxiv_id":"http://arxiv.org/abs/2501.02132v2","authors":["Tomer Jordi Chaffer","Dontrail Cotlage","Justin Goldston"]},{"title":"Effective LLM-Driven Code Generation with Pythoness","abstract":" The advent of large language models (LLMs) has paved the way for a new era of\nprogramming tools with both significant capabilities and risks, as the\ngenerated code lacks guarantees of correctness and reliability. Developers\nusing LLMs currently face the difficult task of optimizing, integrating, and\nmaintaining code generated by AI. We propose an embedded domain-specific\nlanguage (DSL), Pythoness, to address those challenges. In Pythoness,\ndevelopers program with LLMs at a higher level of abstraction. Rather than\ninteracting directly with generated code, developers using Pythoness operate at\nthe level of behavioral specifications when writing functions, classes, or an\nentire program. These specifications can take the form of unit tests and\nproperty-based tests, which may be expressed formally or in natural language.\nGuided by these specifications, Pythoness generates code that both passes the\ntests and can be continuously checked during execution. We posit that the\nPythoness approach lets developers harness the full potential of LLMs for code\ngeneration while substantially mitigating their inherent risks. We describe our\ncurrent prototype implementation of Pythoness and demonstrate that it can\nsuccessfully leverage a combination of tests and code generation to yield\nhigher quality code than specifications alone.\n","arxiv_id":"http://arxiv.org/abs/2501.02138v1","authors":["Kyla H. Levin","Kyle Gwilt","Emery D. Berger","Stephen N. Freund"]},{"title":"Establishing baselines for generative discovery of inorganic crystals","abstract":" Generative artificial intelligence offers a promising avenue for materials\ndiscovery, yet its advantages over traditional methods remain unclear. In this\nwork, we introduce and benchmark two baseline approaches - random enumeration\nof charge-balanced prototypes and data-driven ion exchange of known compounds -\nagainst three generative models: a variational autoencoder, a large language\nmodel, and a diffusion model. Our results show that established methods such as\nion exchange perform comparably well in generating stable materials, although\nmany of these materials tend to closely resemble known compounds. In contrast,\ngenerative models excel at proposing novel structural frameworks and, when\nsufficient training data is available, can more effectively target properties\nsuch as electronic band gap and bulk modulus while maintaining a high stability\nrate. To enhance the performance of both the baseline and generative\napproaches, we implement a post-generation screening step in which all proposed\nstructures are passed through stability and property filters from pre-trained\nmachine learning models including universal interatomic potentials. This\nlow-cost filtering step leads to substantial improvement in the success rates\nof all methods, remains computationally efficient, and ultimately provides a\npractical pathway toward more effective generative strategies for materials\ndiscovery.\n","arxiv_id":"http://arxiv.org/abs/2501.02144v1","authors":["Nathan J. Szymanski","Christopher J. Bartel"]},{"title":"Plasma-CycleGAN: Plasma Biomarker-Guided MRI to PET Cross-modality\n Translation Using Conditional CycleGAN","abstract":" Cross-modality translation between MRI and PET imaging is challenging due to\nthe distinct mechanisms underlying these modalities. Blood-based biomarkers\n(BBBMs) are revolutionizing Alzheimer's disease (AD) detection by identifying\npatients and quantifying brain amyloid levels. However, the potential of BBBMs\nto enhance PET image synthesis remains unexplored. In this paper, we performed\na thorough study on the effect of incorporating BBBM into deep generative\nmodels. By evaluating three widely used cross-modality translation models, we\nfound that BBBMs integration consistently enhances the generative quality\nacross all models. By visual inspection of the generated results, we observed\nthat PET images generated by CycleGAN exhibit the best visual fidelity. Based\non these findings, we propose Plasma-CycleGAN, a novel generative model based\non CycleGAN, to synthesize PET images from MRI using BBBMs as conditions. This\nis the first approach to integrate BBBMs in conditional cross-modality\ntranslation between MRI and PET.\n","arxiv_id":"http://arxiv.org/abs/2501.02146v1","authors":["Yanxi Chen","Yi Su","Celine Dumitrascu","Kewei Chen","David Weidman","Richard J Caselli","Nicholas Ashton","Eric M Reiman","Yalin Wang"]},{"title":"The Race to Efficiency: A New Perspective on AI Scaling Laws","abstract":" As large-scale AI models expand, training becomes costlier and sustaining\nprogress grows harder. Classical scaling laws (e.g., Kaplan et al. (2020),\nHoffmann et al. (2022)) predict training loss from a static compute budget yet\nneglect time and efficiency, prompting the question: how can we balance\nballooning GPU fleets with rapidly improving hardware and algorithms? We\nintroduce the relative-loss equation, a time- and efficiency-aware framework\nthat extends classical AI scaling laws. Our model shows that, without ongoing\nefficiency gains, advanced performance could demand millennia of training or\nunrealistically large GPU fleets. However, near-exponential progress remains\nachievable if the \"efficiency-doubling rate\" parallels Moore's Law. By\nformalizing this race to efficiency, we offer a quantitative roadmap for\nbalancing front-loaded GPU investments with incremental improvements across the\nAI stack. Empirical trends suggest that sustained efficiency gains can push AI\nscaling well into the coming decade, providing a new perspective on the\ndiminishing returns inherent in classical scaling.\n","arxiv_id":"http://arxiv.org/abs/2501.02156v3","authors":["Chien-Ping Lu"]},{"title":"Can ChatGPT implement finite element models for geotechnical engineering\n applications?","abstract":" This study assesses the capability of ChatGPT to generate finite element code\nfor geotechnical engineering applications from a set of prompts. We tested\nthree different initial boundary value problems using a hydro-mechanically\ncoupled formulation for unsaturated soils, including the dissipation of excess\npore water pressure through fluid mass diffusion in one-dimensional space,\ntime-dependent differential settlement of a strip footing, and gravity-driven\nseepage. For each case, initial prompting involved providing ChatGPT with\nnecessary information for finite element implementation, such as balance and\nconstitutive equations, problem geometry, initial and boundary conditions,\nmaterial properties, and spatiotemporal discretization and solution strategies.\nAny errors and unexpected results were further addressed through prompt\naugmentation processes until the ChatGPT-generated finite element code passed\nthe verification/validation test. Our results demonstrate that ChatGPT required\nminimal code revisions when using the FEniCS finite element library, owing to\nits high-level interfaces that enable efficient programming. In contrast, the\nMATLAB code generated by ChatGPT necessitated extensive prompt augmentations\nand/or direct human intervention, as it involves a significant amount of\nlow-level programming required for finite element analysis, such as\nconstructing shape functions or assembling global matrices. Given that prompt\nengineering for this task requires an understanding of the mathematical\nformulation and numerical techniques, this study suggests that while a large\nlanguage model may not yet replace human programmers, it can greatly assist in\nthe implementation of numerical models.\n","arxiv_id":"http://arxiv.org/abs/2501.02199v1","authors":["Taegu Kim","Tae Sup Yun","Hyoung Suk Suh"]},{"title":"Learning Evolution via Optimization Knowledge Adaptation","abstract":" Evolutionary algorithms (EAs) maintain populations through evolutionary\noperators to discover diverse solutions for complex tasks while gathering\nvaluable knowledge, such as historical population data and fitness evaluations.\nHowever, traditional EAs face challenges in dynamically adapting to expanding\nknowledge bases, hindering the efficient exploitation of accumulated\ninformation and limiting adaptability to new situations. To address these\nissues, we introduce an Optimization Knowledge Adaptation Evolutionary Model\n(OKAEM), which features dynamic parameter adjustment using accumulated\nknowledge to enhance its optimization capabilities. OKAEM employs attention\nmechanisms to model the interactions among individuals, fitness landscapes, and\ngenetic components separately, thereby parameterizing the evolutionary\noperators of selection, crossover, and mutation. These powerful learnable\noperators enable OKAEM to benefit from pre-learned extensive prior knowledge\nand self-tune with real-time evolutionary insights. Experimental results\ndemonstrate that OKAEM: 1) exploits prior knowledge for significant performance\ngains across various knowledge transfer settings; 2) achieves competitive\nperformance through self-tuning alone, even without prior knowledge; 3)\noutperforms state-of-the-art black-box baselines in a vision-language model\ntuning case; 4) can improve its optimization capabilities with growing\nknowledge; 5) is capable of emulating principles of natural selection and\ngenetic recombination.\n","arxiv_id":"http://arxiv.org/abs/2501.02200v1","authors":["Chao Wang","Licheng Jiao","Jiaxuan Zhao","Lingling Li","Fang Liu","Shuyuan Yang"]},{"title":"Diffusion Model-Based Data Synthesis Aided Federated Semi-Supervised\n Learning","abstract":" Federated semi-supervised learning (FSSL) is primarily challenged by two\nfactors: the scarcity of labeled data across clients and the non-independent\nand identically distribution (non-IID) nature of data among clients. In this\npaper, we propose a novel approach, diffusion model-based data synthesis aided\nFSSL (DDSA-FSSL), which utilizes a diffusion model (DM) to generate synthetic\ndata, bridging the gap between heterogeneous local data distributions and the\nglobal data distribution. In DDSA-FSSL, clients address the challenge of the\nscarcity of labeled data by employing a federated learning-trained classifier\nto perform pseudo labeling for unlabeled data. The DM is then collaboratively\ntrained using both labeled and precision-optimized pseudo-labeled data,\nenabling clients to generate synthetic samples for classes that are absent in\ntheir labeled datasets. This process allows clients to generate more\ncomprehensive synthetic datasets aligned with the global distribution.\nExtensive experiments conducted on multiple datasets and varying non-IID\ndistributions demonstrate the effectiveness of DDSA-FSSL, e.g., it improves\naccuracy from 38.46% to 52.14% on CIFAR-10 datasets with 10% labeled data.\n","arxiv_id":"http://arxiv.org/abs/2501.02219v1","authors":["Zhongwei Wang","Tong Wu","Zhiyong Chen","Liang Qian","Yin Xu","Meixia Tao"]},{"title":"CORD: Generalizable Cooperation via Role Diversity","abstract":" Cooperative multi-agent reinforcement learning (MARL) aims to develop agents\nthat can collaborate effectively. However, most cooperative MARL methods\noverfit training agents, making learned policies not generalize well to unseen\ncollaborators, which is a critical issue for real-world deployment. Some\nmethods attempt to address the generalization problem but require prior\nknowledge or predefined policies of new teammates, limiting real-world\napplications. To this end, we propose a hierarchical MARL approach to enable\ngeneralizable cooperation via role diversity, namely CORD. CORD's high-level\ncontroller assigns roles to low-level agents by maximizing the role entropy\nwith constraints. We show this constrained objective can be decomposed into\ncausal influence in role that enables reasonable role assignment, and role\nheterogeneity that yields coherent, non-redundant role clusters. Evaluated on a\nvariety of cooperative multi-agent tasks, CORD achieves better performance than\nbaselines, especially in generalization tests. Ablation studies further\ndemonstrate the efficacy of the constrained objective in generalizable\ncooperation.\n","arxiv_id":"http://arxiv.org/abs/2501.02221v2","authors":["Kanefumi Matsuyama","Kefan Su","Jiangxing Wang","Deheng Ye","Zongqing Lu"]},{"title":"Towards a constructive framework for control theory","abstract":" This work presents a framework for control theory based on constructive\nanalysis to account for discrepancy between mathematical results and their\nimplementation in a computer, also referred to as computational uncertainty. In\ncontrol engineering, the latter is usually either neglected or considered\nsubmerged into some other type of uncertainty, such as system noise, and\naddressed within robust control. However, even robust control methods may be\ncompromised when the mathematical objects involved in the respective algorithms\nfail to exist in exact form and subsequently fail to satisfy the required\nproperties. For instance, in general stabilization using a control Lyapunov\nfunction, computational uncertainty may distort stability certificates or even\ndestabilize the system despite robustness of the stabilization routine with\nregards to system, actuator and measurement noise. In fact, battling numerical\nproblems in practical implementation of controllers is common among control\nengineers. Such observations indicate that computational uncertainty should\nindeed be addressed explicitly in controller synthesis and system analysis. The\nmajor contribution here is a fairly general framework for proof techniques in\nanalysis and synthesis of control systems based on constructive analysis which\nexplicitly states that every computation be doable only up to a finite\nprecision thus accounting for computational uncertainty. A series of previous\nworks is overviewed, including constructive system stability and stabilization,\napproximate optimal controls, eigenvalue problems, Caratheodory trajectories,\nmeasurable selectors. Additionally, a new constructive version of the Danskin's\ntheorem, which is crucial in adversarial defense, is presented.\n","arxiv_id":"http://arxiv.org/abs/2501.02267v1","authors":["Pavel Osinenko"]},{"title":"Deep Learning-Driven Segmentation of Ischemic Stroke Lesions Using\n Multi-Channel MRI","abstract":" Ischemic stroke, caused by cerebral vessel occlusion, presents substantial\nchallenges in medical imaging due to the variability and subtlety of stroke\nlesions. Magnetic Resonance Imaging (MRI) plays a crucial role in diagnosing\nand managing ischemic stroke, yet existing segmentation techniques often fail\nto accurately delineate lesions. This study introduces a novel deep\nlearning-based method for segmenting ischemic stroke lesions using\nmulti-channel MRI modalities, including Diffusion Weighted Imaging (DWI),\nApparent Diffusion Coefficient (ADC), and enhanced Diffusion Weighted Imaging\n(eDWI). The proposed architecture integrates DenseNet121 as the encoder with\nSelf-Organized Operational Neural Networks (SelfONN) in the decoder, enhanced\nby Channel and Space Compound Attention (CSCA) and Double\nSqueeze-and-Excitation (DSE) blocks. Additionally, a custom loss function\ncombining Dice Loss and Jaccard Loss with weighted averages is introduced to\nimprove model performance. Trained and evaluated on the ISLES 2022 dataset, the\nmodel achieved Dice Similarity Coefficients (DSC) of 83.88% using DWI alone,\n85.86% with DWI and ADC, and 87.49% with the integration of DWI, ADC, and eDWI.\nThis approach not only outperforms existing methods but also addresses key\nlimitations in current segmentation practices. These advancements significantly\nenhance diagnostic precision and treatment planning for ischemic stroke,\nproviding valuable support for clinical decision-making.\n","arxiv_id":"http://arxiv.org/abs/2501.02287v1","authors":["Ashiqur Rahman","Muhammad E. H. Chowdhury","Md Sharjis Ibne Wadud","Rusab Sarmun","Adam Mushtak","Sohaib Bassam Zoghoul","Israa Al-Hashimi"]},{"title":"DiffGraph: Heterogeneous Graph Diffusion Model","abstract":" Recent advances in Graph Neural Networks (GNNs) have revolutionized\ngraph-structured data modeling, yet traditional GNNs struggle with complex\nheterogeneous structures prevalent in real-world scenarios. Despite progress in\nhandling heterogeneous interactions, two fundamental challenges persist: noisy\ndata significantly compromising embedding quality and learning performance, and\nexisting methods' inability to capture intricate semantic transitions among\nheterogeneous relations, which impacts downstream predictions. To address these\nfundamental issues, we present the Heterogeneous Graph Diffusion Model\n(DiffGraph), a pioneering framework that introduces an innovative cross-view\ndenoising strategy. This advanced approach transforms auxiliary heterogeneous\ndata into target semantic spaces, enabling precise distillation of\ntask-relevant information. At its core, DiffGraph features a sophisticated\nlatent heterogeneous graph diffusion mechanism, implementing a novel forward\nand backward diffusion process for superior noise management. This methodology\nachieves simultaneous heterogeneous graph denoising and cross-type transition,\nwhile significantly simplifying graph generation through its latent-space\ndiffusion capabilities. Through rigorous experimental validation on both public\nand industrial datasets, we demonstrate that DiffGraph consistently surpasses\nexisting methods in link prediction and node classification tasks, establishing\nnew benchmarks for robustness and efficiency in heterogeneous graph processing.\nThe model implementation is publicly available at:\nhttps://github.com/HKUDS/DiffGraph.\n","arxiv_id":"http://arxiv.org/abs/2501.02313v1","authors":["Zongwei Li","Lianghao Xia","Hua Hua","Shijie Zhang","Shuangyang Wang","Chao Huang"]},{"title":"Validity Arguments For Constructed Response Scoring Using Generative\n Artificial Intelligence Applications","abstract":" The rapid advancements in large language models and generative artificial\nintelligence (AI) capabilities are making their broad application in the\nhigh-stakes testing context more likely. Use of generative AI in the scoring of\nconstructed responses is particularly appealing because it reduces the effort\nrequired for handcrafting features in traditional AI scoring and might even\noutperform those methods. The purpose of this paper is to highlight the\ndifferences in the feature-based and generative AI applications in constructed\nresponse scoring systems and propose a set of best practices for the collection\nof validity evidence to support the use and interpretation of constructed\nresponse scores from scoring systems using generative AI. We compare the\nvalidity evidence needed in scoring systems using human ratings, feature-based\nnatural language processing AI scoring engines, and generative AI. The evidence\nneeded in the generative AI context is more extensive than in the feature-based\nNLP scoring context because of the lack of transparency and other concerns\nunique to generative AI such as consistency. Constructed response score data\nfrom standardized tests demonstrate the collection of validity evidence for\ndifferent types of scoring systems and highlights the numerous complexities and\nconsiderations when making a validity argument for these scores. In addition,\nwe discuss how the evaluation of AI scores might include a consideration of how\na contributory scoring approach combining multiple AI scores (from different\nsources) will cover more of the construct in the absence of human ratings.\n","arxiv_id":"http://arxiv.org/abs/2501.02334v1","authors":["Jodi M. Casabianca","Daniel F. McCaffrey","Matthew S. Johnson","Naim Alper","Vladimir Zubenko"]},{"title":"GNSS/GPS Spoofing and Jamming Identification Using Machine Learning and\n Deep Learning","abstract":" The increasing reliance on Global Navigation Satellite Systems (GNSS),\nparticularly the Global Positioning System (GPS), underscores the urgent need\nto safeguard these technologies against malicious threats such as spoofing and\njamming. As the backbone for positioning, navigation, and timing (PNT) across\nvarious applications including transportation, telecommunications, and\nemergency services GNSS is vulnerable to deliberate interference that poses\nsignificant risks. Spoofing attacks, which involve transmitting counterfeit\nGNSS signals to mislead receivers into calculating incorrect positions, can\nresult in serious consequences, from navigational errors in civilian aviation\nto security breaches in military operations. Furthermore, the lack of inherent\nsecurity measures within GNSS systems makes them attractive targets for\nadversaries. While GNSS/GPS jamming and spoofing systems consist of numerous\ncomponents, the ability to distinguish authentic signals from malicious ones is\nessential for maintaining system integrity. Recent advancements in machine\nlearning and deep learning provide promising avenues for enhancing detection\nand mitigation strategies against these threats. This paper addresses both\nspoofing and jamming by tackling real-world challenges through machine\nlearning, deep learning, and computer vision techniques. Through extensive\nexperiments on two real-world datasets related to spoofing and jamming\ndetection using advanced algorithms, we achieved state of the art results. In\nthe GNSS/GPS jamming detection task, we attained approximately 99% accuracy,\nimproving performance by around 5% compared to previous studies. Additionally,\nwe addressed a challenging tasks related to spoofing detection, yielding\nresults that underscore the potential of machine learning and deep learning in\nthis domain.\n","arxiv_id":"http://arxiv.org/abs/2501.02352v1","authors":["Ali Ghanbarzade","Hossein Soleimani"]},{"title":"FaceSpeak: Expressive and High-Quality Speech Synthesis from Human\n Portraits of Different Styles","abstract":" Humans can perceive speakers' characteristics (e.g., identity, gender,\npersonality and emotion) by their appearance, which are generally aligned to\ntheir voice style. Recently, vision-driven Text-to-speech (TTS) scholars\ngrounded their investigations on real-person faces, thereby restricting\neffective speech synthesis from applying to vast potential usage scenarios with\ndiverse characters and image styles. To solve this issue, we introduce a novel\nFaceSpeak approach. It extracts salient identity characteristics and emotional\nrepresentations from a wide variety of image styles. Meanwhile, it mitigates\nthe extraneous information (e.g., background, clothing, and hair color, etc.),\nresulting in synthesized speech closely aligned with a character's persona.\nFurthermore, to overcome the scarcity of multi-modal TTS data, we have devised\nan innovative dataset, namely Expressive Multi-Modal TTS, which is diligently\ncurated and annotated to facilitate research in this domain. The experimental\nresults demonstrate our proposed FaceSpeak can generate portrait-aligned voice\nwith satisfactory naturalness and quality.\n","arxiv_id":"http://arxiv.org/abs/2501.03181v1","authors":["Tian-Hao Zhang","Jiawei Zhang","Jun Wang","Xinyuan Qian","Xu-Cheng Yin"]},{"title":"Classifier-Guided Captioning Across Modalities","abstract":" Most current captioning systems use language models trained on data from\nspecific settings, such as image-based captioning via Amazon Mechanical Turk,\nlimiting their ability to generalize to other modality distributions and\ncontexts. This limitation hinders performance in tasks like audio or video\ncaptioning, where different semantic cues are needed. Addressing this challenge\nis crucial for creating more adaptable and versatile captioning frameworks\napplicable across diverse real-world contexts. In this work, we introduce a\nmethod to adapt captioning networks to the semantics of alternative settings,\nsuch as capturing audibility in audio captioning, where it is crucial to\ndescribe sounds and their sources. Our framework consists of two main\ncomponents: (i) a frozen captioning system incorporating a language model (LM),\nand (ii) a text classifier that guides the captioning system. The classifier is\ntrained on a dataset automatically generated by GPT-4, using tailored prompts\nspecifically designed to enhance key aspects of the generated captions.\nImportantly, the framework operates solely during inference, eliminating the\nneed for further training of the underlying captioning model. We evaluate the\nframework on various models and modalities, with a focus on audio captioning,\nand report promising results. Notably, when combined with an existing zero-shot\naudio captioning system, our framework improves its quality and sets\nstate-of-the-art performance in zero-shot audio captioning.\n","arxiv_id":"http://arxiv.org/abs/2501.03183v1","authors":["Ariel Shaulov","Tal Shaharabany","Eitan Shaar","Gal Chechik","Lior Wolf"]},{"title":"Breaking Through the Spike: Spike Window Decoding for Accelerated and\n Precise Automatic Speech Recognition","abstract":" Recently, end-to-end automatic speech recognition has become the mainstream\napproach in both industry and academia. To optimize system performance in\nspecific scenarios, the Weighted Finite-State Transducer (WFST) is extensively\nused to integrate acoustic and language models, leveraging its capacity to\nimplicitly fuse language models within static graphs, thereby ensuring robust\nrecognition while also facilitating rapid error correction. However, WFST\nnecessitates a frame-by-frame search of CTC posterior probabilities through\nautoregression, which significantly hampers inference speed. In this work, we\nthoroughly investigate the spike property of CTC outputs and further propose\nthe conjecture that adjacent frames to non-blank spikes carry semantic\ninformation beneficial to the model. Building on this, we propose the Spike\nWindow Decoding algorithm, which greatly improves the inference speed by making\nthe number of frames decoded in WFST linearly related to the number of spiking\nframes in the CTC output, while guaranteeing the recognition performance. Our\nmethod achieves SOTA recognition accuracy with significantly accelerates\ndecoding speed, proven across both AISHELL-1 and large-scale In-House datasets,\nestablishing a pioneering approach for integrating CTC output with WFST.\n","arxiv_id":"http://arxiv.org/abs/2501.03257v1","authors":["Wei Zhang","Tian-Hao Zhang","Chao Luo","Hui Zhou","Chao Yang","Xinyuan Qian","Xu-Cheng Yin"]},{"title":"Navigation Variable-based Multi-objective Particle Swarm Optimization\n for UAV Path Planning with Kinematic Constraints","abstract":" Path planning is essential for unmanned aerial vehicles (UAVs) as it\ndetermines the path that the UAV needs to follow to complete a task. This work\naddresses this problem by introducing a new algorithm called navigation\nvariable-based multi-objective particle swarm optimization (NMOPSO). It first\nmodels path planning as an optimization problem via the definition of a set of\nobjective functions that include optimality and safety requirements for UAV\noperation. The NMOPSO is then used to minimize those functions through Pareto\noptimal solutions. The algorithm features a new path representation based on\nnavigation variables to include kinematic constraints and exploit the\nmaneuverable characteristics of the UAV. It also includes an adaptive mutation\nmechanism to enhance the diversity of the swarm for better solutions.\nComparisons with various algorithms have been carried out to benchmark the\nproposed approach. The results indicate that the NMOPSO performs better than\nnot only other particle swarm optimization variants but also other\nstate-of-the-art multi-objective and metaheuristic optimization algorithms.\nExperiments have also been conducted with real UAVs to confirm the validity of\nthe approach for practical flights. The source code of the algorithm is\navailable at https://github.com/ngandng/NMOPSO.\n","arxiv_id":"http://arxiv.org/abs/2501.03261v1","authors":["Thi Thuy Ngan Duong","Duy-Nam Bui","Manh Duong Phung"]},{"title":"Bridge the Inference Gaps of Neural Processes via Expectation\n Maximization","abstract":" The neural process (NP) is a family of computationally efficient models for\nlearning distributions over functions. However, it suffers from under-fitting\nand shows suboptimal performance in practice. Researchers have primarily\nfocused on incorporating diverse structural inductive biases, \\textit{e.g.}\nattention or convolution, in modeling. The topic of inference suboptimality and\nan analysis of the NP from the optimization objective perspective has hardly\nbeen studied in earlier work. To fix this issue, we propose a surrogate\nobjective of the target log-likelihood of the meta dataset within the\nexpectation maximization framework. The resulting model, referred to as the\nSelf-normalized Importance weighted Neural Process (SI-NP), can learn a more\naccurate functional prior and has an improvement guarantee concerning the\ntarget log-likelihood. Experimental results show the competitive performance of\nSI-NP over other NPs objectives and illustrate that structural inductive\nbiases, such as attention modules, can also augment our method to achieve SOTA\nperformance. Our code is available at\n\\url{https://github.com/hhq123gogogo/SI_NPs}.\n","arxiv_id":"http://arxiv.org/abs/2501.03264v1","authors":["Qi Wang","Marco Federici","Herke van Hoof"]},{"title":"Listening and Seeing Again: Generative Error Correction for Audio-Visual\n Speech Recognition","abstract":" Unlike traditional Automatic Speech Recognition (ASR), Audio-Visual Speech\nRecognition (AVSR) takes audio and visual signals simultaneously to infer the\ntranscription. Recent studies have shown that Large Language Models (LLMs) can\nbe effectively used for Generative Error Correction (GER) in ASR by predicting\nthe best transcription from ASR-generated N-best hypotheses. However, these\nLLMs lack the ability to simultaneously understand audio and visual, making the\nGER approach challenging to apply in AVSR. In this work, we propose a novel GER\nparadigm for AVSR, termed AVGER, that follows the concept of ``listening and\nseeing again''. Specifically, we first use the powerful AVSR system to read the\naudio and visual signals to get the N-Best hypotheses, and then use the\nQ-former-based Multimodal Synchronous Encoder to read the audio and visual\ninformation again and convert them into an audio and video compression\nrepresentation respectively that can be understood by LLM. Afterward, the\naudio-visual compression representation and the N-Best hypothesis together\nconstitute a Cross-modal Prompt to guide the LLM in producing the best\ntranscription. In addition, we also proposed a Multi-Level Consistency\nConstraint training criterion, including logits-level, utterance-level and\nrepresentations-level, to improve the correction accuracy while enhancing the\ninterpretability of audio and visual compression representations. The\nexperimental results on the LRS3 dataset show that our method outperforms\ncurrent mainstream AVSR systems. The proposed AVGER can reduce the Word Error\nRate (WER) by 24% compared to them. Code and models can be found at:\nhttps://github.com/CircleRedRain/AVGER.\n","arxiv_id":"http://arxiv.org/abs/2501.04038v1","authors":["Rui Liu","Hongyu Yuan","Haizhou Li"]},{"title":"A Survey on Large Language Models with some Insights on their\n Capabilities and Limitations","abstract":" The rapid advancement of artificial intelligence, particularly with the\ndevelopment of Large Language Models (LLMs) built on the transformer\narchitecture, has redefined the capabilities of natural language processing.\nThese models now exhibit remarkable performance across various language-related\ntasks, such as text generation, question answering, translation, and\nsummarization, often rivaling human-like comprehension. More intriguingly, LLMs\nhave demonstrated emergent abilities extending beyond their core functions,\nshowing proficiency in tasks like commonsense reasoning, code generation, and\narithmetic. This survey paper explores the foundational components, scaling\nmechanisms, and architectural strategies that drive these capabilities.\nEmphasizing models like GPT and LLaMA, we analyze the impact of exponential\ndata and computational growth on LLM performance, while also addressing the\ntrade-offs associated with scaling. We also examine LLM applications across\nsectors, such as healthcare, finance, education, and law, highlighting their\nadaptability and potential to solve domain-specific challenges. Central to this\nwork are the questions of how LLMs generalize across diverse tasks, exhibit\nplanning, and reasoning abilities, and whether these emergent abilities can be\nsystematically elicited or enhanced. In particular, we provide some insights\ninto the CoT (Chain of Thought) and PoT (Plan of Thought) abilities within\nLLMs, focusing on how pre-training data influences their emergence.\nAdditionally, we investigate LLM-modulo frameworks that integrate external\nsystems, allowing LLMs to handle complex, dynamic tasks. By analyzing these\nfactors, this paper aims to foster the ongoing discussion on the capabilities\nand limits of LLMs, promoting their responsible development and application in\nnovel and increasingly complex environments.\n","arxiv_id":"http://arxiv.org/abs/2501.04040v1","authors":["Andrea Matarazzo","Riccardo Torlone"]},{"title":"FLAME: Financial Large-Language Model Assessment and Metrics Evaluation","abstract":" LLMs have revolutionized NLP and demonstrated potential across diverse\ndomains. More and more financial LLMs have been introduced for finance-specific\ntasks, yet comprehensively assessing their value is still challenging. In this\npaper, we introduce FLAME, a comprehensive financial LLMs evaluation system in\nChinese, which includes two core evaluation benchmarks: FLAME-Cer and\nFLAME-Sce. FLAME-Cer covers 14 types of authoritative financial certifications,\nincluding CPA, CFA, and FRM, with a total of approximately 16,000 carefully\nselected questions. All questions have been manually reviewed to ensure\naccuracy and representativeness. FLAME-Sce consists of 10 primary core\nfinancial business scenarios, 21 secondary financial business scenarios, and a\ncomprehensive evaluation set of nearly 100 tertiary financial application\ntasks. We evaluate 6 representative LLMs, including GPT-4o, GLM-4, ERNIE-4.0,\nQwen2.5, XuanYuan3, and the latest Baichuan4-Finance, revealing\nBaichuan4-Finance excels other LLMs in most tasks. By establishing a\ncomprehensive and professional evaluation system, FLAME facilitates the\nadvancement of financial LLMs in Chinese contexts. Instructions for\nparticipating in the evaluation are available on GitHub:\nhttps://github.com/FLAME-ruc/FLAME.\n","arxiv_id":"http://arxiv.org/abs/2501.06211v1","authors":["Jiayu Guo","Yu Guo","Martha Li","Songtao Tan"]},{"title":"Operator Learning for Reconstructing Flow Fields from Sparse\n Measurements: an Energy Transformer Approach","abstract":" Machine learning methods have shown great success in various scientific\nareas, including fluid mechanics. However, reconstruction problems, where full\nvelocity fields must be recovered from partial observations, remain\nchallenging. In this paper, we propose a novel operator learning framework for\nsolving reconstruction problems by using the Energy Transformer (ET), an\narchitecture inspired by associative memory models. We formulate reconstruction\nas a mapping from incomplete observed data to full reconstructed fields. The\nmethod is validated on three fluid mechanics examples using diverse types of\ndata: (1) unsteady 2D vortex street in flow past a cylinder using simulation\ndata; (2) high-speed under-expanded impinging supersonic jets impingement using\nSchlieren imaging; and (3) 3D turbulent jet flow using particle tracking. The\nresults demonstrate the ability of ET to accurately reconstruct complex flow\nfields from highly incomplete data (90\\% missing), even for noisy experimental\nmeasurements, with fast training and inference on a single GPU. This work\nprovides a promising new direction for tackling reconstruction problems in\nfluid mechanics and other areas in mechanics, geophysics, weather prediction,\nand beyond.\n","arxiv_id":"http://arxiv.org/abs/2501.08339v1","authors":["Qian Zhang","Dmitry Krotov","George Em Karniadakis"]},{"title":"VERITAS: Verifying the Performance of AI-native Transceiver Actions in\n Base-Stations","abstract":" Artificial Intelligence (AI)-native receivers prove significant performance\nimprovement in high noise regimes and can potentially reduce communication\noverhead compared to the traditional receiver. However, their performance\nhighly depends on the representativeness of the training dataset. A major issue\nis the uncertainty of whether the training dataset covers all test environments\nand waveform configurations, and thus, whether the trained model is robust in\npractical deployment conditions. To this end, we propose a joint\nmeasurement-recovery framework for AI-native transceivers post deployment,\ncalled VERITAS, that continuously looks for distribution shifts in the received\nsignals and triggers finite re-training spurts. VERITAS monitors the wireless\nchannel using 5G pilots fed to an auxiliary neural network that detects\nout-of-distribution channel profile, transmitter speed, and delay spread. As\nsoon as such a change is detected, a traditional (reference) receiver is\nactivated, which runs for a period of time in parallel to the AI-native\nreceiver. Finally, VERTIAS compares the bit probabilities of the AI-native and\nthe reference receivers for the same received data inputs, and decides whether\nor not a retraining process needs to be initiated. Our evaluations reveal that\nVERITAS can detect changes in the channel profile, transmitter speed, and delay\nspread with 99%, 97%, and 69% accuracies, respectively, followed by timely\ninitiation of retraining for 86%, 93.3%, and 94.8% of inputs in channel\nprofile, transmitter speed, and delay spread test sets, respectively.\n","arxiv_id":"http://arxiv.org/abs/2501.09761v1","authors":["Nasim Soltani","Michael Loehning","Kaushik Chowdhury"]},{"title":"Dynamics of Adversarial Attacks on Large Language Model-Based Search\n Engines","abstract":" The increasing integration of Large Language Model (LLM) based search engines\nhas transformed the landscape of information retrieval. However, these systems\nare vulnerable to adversarial attacks, especially ranking manipulation attacks,\nwhere attackers craft webpage content to manipulate the LLM's ranking and\npromote specific content, gaining an unfair advantage over competitors. In this\npaper, we study the dynamics of ranking manipulation attacks. We frame this\nproblem as an Infinitely Repeated Prisoners' Dilemma, where multiple players\nstrategically decide whether to cooperate or attack. We analyze the conditions\nunder which cooperation can be sustained, identifying key factors such as\nattack costs, discount rates, attack success rates, and trigger strategies that\ninfluence player behavior. We identify tipping points in the system dynamics,\ndemonstrating that cooperation is more likely to be sustained when players are\nforward-looking. However, from a defense perspective, we find that simply\nreducing attack success probabilities can, paradoxically, incentivize attacks\nunder certain conditions. Furthermore, defensive measures to cap the upper\nbound of attack success rates may prove futile in some scenarios. These\ninsights highlight the complexity of securing LLM-based systems. Our work\nprovides a theoretical foundation and practical insights for understanding and\nmitigating their vulnerabilities, while emphasizing the importance of adaptive\nsecurity strategies and thoughtful ecosystem design.\n","arxiv_id":"http://arxiv.org/abs/2501.00745v1","authors":["Xiyang Hu"]},{"title":"MuQ: Self-Supervised Music Representation Learning with Mel Residual\n Vector Quantization","abstract":" Recent years have witnessed the success of foundation models pre-trained with\nself-supervised learning (SSL) in various music informatics understanding\ntasks, including music tagging, instrument classification, key detection, and\nmore. In this paper, we propose a self-supervised music representation learning\nmodel for music understanding. Distinguished from previous studies adopting\nrandom projection or existing neural codec, the proposed model, named MuQ, is\ntrained to predict tokens generated by Mel Residual Vector Quantization\n(Mel-RVQ). Our Mel-RVQ utilizes residual linear projection structure for Mel\nspectrum quantization to enhance the stability and efficiency of target\nextraction and lead to better performance. Experiments in a large variety of\ndownstream tasks demonstrate that MuQ outperforms previous self-supervised\nmusic representation models with only 0.9K hours of open-source pre-training\ndata. Scaling up the data to over 160K hours and adopting iterative training\nconsistently improve the model performance. To further validate the strength of\nour model, we present MuQ-MuLan, a joint music-text embedding model based on\ncontrastive learning, which achieves state-of-the-art performance in the\nzero-shot music tagging task on the MagnaTagATune dataset. Code and checkpoints\nare open source in https://github.com/tencent-ailab/MuQ.\n","arxiv_id":"http://arxiv.org/abs/2501.01108v2","authors":["Haina Zhu","Yizhi Zhou","Hangting Chen","Jianwei Yu","Ziyang Ma","Rongzhi Gu","Yi Luo","Wei Tan","Xie Chen"]},{"title":"Symmetries-enhanced Multi-Agent Reinforcement Learning","abstract":" Multi-agent reinforcement learning has emerged as a powerful framework for\nenabling agents to learn complex, coordinated behaviors but faces persistent\nchallenges regarding its generalization, scalability and sample efficiency.\nRecent advancements have sought to alleviate those issues by embedding\nintrinsic symmetries of the systems in the policy. Yet, most dynamical systems\nexhibit little to no symmetries to exploit. This paper presents a novel\nframework for embedding extrinsic symmetries in multi-agent system dynamics\nthat enables the use of symmetry-enhanced methods to address systems with\ninsufficient intrinsic symmetries, expanding the scope of equivariant learning\nto a wide variety of MARL problems. Central to our framework is the Group\nEquivariant Graphormer, a group-modular architecture specifically designed for\ndistributed swarming tasks. Extensive experiments on a swarm of\nsymmetry-breaking quadrotors validate the effectiveness of our approach,\nshowcasing its potential for improved generalization and zero-shot scalability.\nOur method achieves significant reductions in collision rates and enhances task\nsuccess rates across a diverse range of scenarios and varying swarm sizes.\n","arxiv_id":"http://arxiv.org/abs/2501.01136v1","authors":["Nikolaos Bousias","Stefanos Pertigkiozoglou","Kostas Daniilidis","George Pappas"]},{"title":"Change Detection-Based Procedures for Piecewise Stationary MABs: A\n Modular Approach","abstract":" Conventional Multi-Armed Bandit (MAB) algorithms are designed for stationary\nenvironments, where the reward distributions associated with the arms do not\nchange with time. In many applications, however, the environment is more\naccurately modeled as being nonstationary. In this work, piecewise stationary\nMAB (PS-MAB) environments are investigated, in which the reward distributions\nassociated with a subset of the arms change at some change-points and remain\nstationary between change-points. Our focus is on the asymptotic analysis of\nPS-MABs, for which practical algorithms based on change detection (CD) have\nbeen previously proposed. Our goal is to modularize the design and analysis of\nsuch CD-based Bandit (CDB) procedures. To this end, we identify the\nrequirements for stationary bandit algorithms and change detectors in a CDB\nprocedure that are needed for the modularization. We assume that the rewards\nare sub-Gaussian. Under this assumption and a condition on the separation of\nthe change-points, we show that the analysis of CDB procedures can indeed be\nmodularized, so that regret bounds can be obtained in a unified manner for\nvarious combinations of change detectors and bandit algorithms. Through this\nanalysis, we develop new modular CDB procedures that are order-optimal. We\ncompare the performance of our modular CDB procedures with various other\nmethods in simulations.\n","arxiv_id":"http://arxiv.org/abs/2501.01291v1","authors":["Yu-Han Huang","Argyrios Gerogiannis","Subhonmesh Bose","Venugopal V. Veeravalli"]},{"title":"Quantifying A Firm's AI Engagement: Constructing Objective, Data-Driven,\n AI Stock Indices Using 10-K Filings","abstract":" Following an analysis of existing AI-related exchange-traded funds (ETFs), we\nreveal the selection criteria for determining which stocks qualify as\nAI-related are often opaque and rely on vague phrases and subjective judgments.\nThis paper proposes a new, objective, data-driven approach using natural\nlanguage processing (NLP) techniques to classify AI stocks by analyzing annual\n10-K filings from 3,395 NASDAQ-listed firms between 2011 and 2023. This\nanalysis quantifies each company's engagement with AI through binary indicators\nand weighted AI scores based on the frequency and context of AI-related terms.\nUsing these metrics, we construct four AI stock indices-the Equally Weighted AI\nIndex (AII), the Size-Weighted AI Index (SAII), and two Time-Discounted AI\nIndices (TAII05 and TAII5X)-offering different perspectives on AI investment.\nWe validate our methodology through an event study on the launch of OpenAI's\nChatGPT, demonstrating that companies with higher AI engagement saw\nsignificantly greater positive abnormal returns, with analyses supporting the\npredictive power of our AI measures. Our indices perform on par with or surpass\n14 existing AI-themed ETFs and the Nasdaq Composite Index in risk-return\nprofiles, market responsiveness, and overall performance, achieving higher\naverage daily returns and risk-adjusted metrics without increased volatility.\nThese results suggest our NLP-based approach offers a reliable,\nmarket-responsive, and cost-effective alternative to existing AI-related ETF\nproducts. Our innovative methodology can also guide investors, asset managers,\nand policymakers in using corporate data to construct other thematic\nportfolios, contributing to a more transparent, data-driven, and competitive\napproach.\n","arxiv_id":"http://arxiv.org/abs/2501.01763v1","authors":["Lennart Ante","Aman Saggu"]},{"title":"KANS: Knowledge Discovery Graph Attention Network for Soft Sensing in\n Multivariate Industrial Processes","abstract":" Soft sensing of hard-to-measure variables is often crucial in industrial\nprocesses. Current practices rely heavily on conventional modeling techniques\nthat show success in improving accuracy. However, they overlook the non-linear\nnature, dynamics characteristics, and non-Euclidean dependencies between\ncomplex process variables. To tackle these challenges, we present a framework\nknown as a Knowledge discovery graph Attention Network for effective Soft\nsensing (KANS). Unlike the existing deep learning soft sensor models, KANS can\ndiscover the intrinsic correlations and irregular relationships between the\nmultivariate industrial processes without a predefined topology. First, an\nunsupervised graph structure learning method is introduced, incorporating the\ncosine similarity between different sensor embedding to capture the\ncorrelations between sensors. Next, we present a graph attention-based\nrepresentation learning that can compute the multivariate data parallelly to\nenhance the model in learning complex sensor nodes and edges. To fully explore\nKANS, knowledge discovery analysis has also been conducted to demonstrate the\ninterpretability of the model. Experimental results demonstrate that KANS\nsignificantly outperforms all the baselines and state-of-the-art methods in\nsoft sensing performance. Furthermore, the analysis shows that KANS can find\nsensors closely related to different process variables without domain\nknowledge, significantly improving soft sensing accuracy.\n","arxiv_id":"http://arxiv.org/abs/2501.02015v1","authors":["Hwa Hui Tew","Gaoxuan Li","Fan Ding","Xuewen Luo","Junn Yong Loo","Chee-Ming Ting","Ze Yang Ding","Chee Pin Tan"]},{"title":"Benchmark Evaluations, Applications, and Challenges of Large Vision\n Language Models: A Survey","abstract":" Multimodal Vision Language Models (VLMs) have emerged as a transformative\ntechnology at the intersection of computer vision and natural language\nprocessing, enabling machines to perceive and reason about the world through\nboth visual and textual modalities. For example, models such as CLIP, Claude,\nand GPT-4V demonstrate strong reasoning and understanding abilities on visual\nand textual data and beat classical single modality vision models on zero-shot\nclassification. Despite their rapid advancements in research and growing\npopularity in applications, a comprehensive survey of existing studies on VLMs\nis notably lacking, particularly for researchers aiming to leverage VLMs in\ntheir specific domains. To this end, we provide a systematic overview of VLMs\nin the following aspects: model information of the major VLMs developed over\nthe past five years (2019-2024); the main architectures and training methods of\nthese VLMs; summary and categorization of the popular benchmarks and evaluation\nmetrics of VLMs; the applications of VLMs including embodied agents, robotics,\nand video generation; the challenges and issues faced by current VLMs such as\nhallucination, fairness, and safety. Detailed collections including papers and\nmodel repository links are listed in\nhttps://github.com/zli12321/Awesome-VLM-Papers-And-Models.git.\n","arxiv_id":"http://arxiv.org/abs/2501.02189v2","authors":["Zongxia Li","Xiyang Wu","Hongyang Du","Huy Nghiem","Guangyao Shi"]},{"title":"Optimizing Small Language Models for In-Vehicle Function-Calling","abstract":" We propose a holistic approach for deploying Small Language Models (SLMs) as\nfunction-calling agents within vehicles as edge devices, offering a more\nflexible and robust alternative to traditional rule-based systems. By\nleveraging SLMs, we simplify vehicle control mechanisms and enhance the user\nexperience. Given the in-vehicle hardware constraints, we apply\nstate-of-the-art model compression techniques, including structured pruning,\nhealing, and quantization, ensuring that the model fits within the resource\nlimitations while maintaining acceptable performance. Our work focuses on\noptimizing a representative SLM, Microsoft's Phi-3 mini, and outlines best\npractices for enabling embedded models, including compression, task-specific\nfine-tuning, and vehicle integration. We demonstrate that, despite significant\nreduction in model size which removes up to 2 billion parameters from the\noriginal model, our approach preserves the model's ability to handle complex\nin-vehicle tasks accurately and efficiently. Furthermore, by executing the\nmodel in a lightweight runtime environment, we achieve a generation speed of 11\ntokens per second, making real-time, on-device inference feasible without\nhardware acceleration. Our results demonstrate the potential of SLMs to\ntransform vehicle control systems, enabling more intuitive interactions between\nusers and their vehicles for an enhanced driving experience.\n","arxiv_id":"http://arxiv.org/abs/2501.02342v1","authors":["Yahya Sowti Khiabani","Farris Atif","Chieh Hsu","Sven Stahlmann","Tobias Michels","Sebastian Kramer","Benedikt Heidrich","M. Saquib Sarfraz","Julian Merten","Faezeh Tafazzoli"]},{"title":"Graph-Aware Isomorphic Attention for Adaptive Dynamics in Transformers","abstract":" We present an approach to modifying Transformer architectures by integrating\ngraph-aware relational reasoning into the attention mechanism, merging concepts\nfrom graph neural networks and language modeling. Building on the inherent\nconnection between attention and graph theory, we reformulate the Transformer's\nattention mechanism as a graph operation and propose Graph-Aware Isomorphic\nAttention. This method leverages advanced graph modeling strategies, including\nGraph Isomorphism Networks (GIN) and Principal Neighborhood Aggregation (PNA),\nto enrich the representation of relational structures. Our approach captures\ncomplex dependencies and generalizes across tasks, as evidenced by a reduced\ngeneralization gap and improved learning performance. Additionally, we expand\nthe concept of graph-aware attention to introduce Sparse GIN-Attention, a\nfine-tuning approach that employs sparse GINs. By interpreting attention\nmatrices as sparse adjacency graphs, this technique enhances the adaptability\nof pre-trained foundational models with minimal computational overhead,\nendowing them with graph-aware capabilities. Sparse GIN-Attention fine-tuning\nachieves improved training dynamics and better generalization compared to\nalternative methods like low-rank adaption (LoRA). We discuss latent graph-like\nstructures within traditional attention mechanisms, offering a new lens through\nwhich Transformers can be understood. By evolving Transformers as hierarchical\nGIN models for relational reasoning. This perspective suggests profound\nimplications for foundational model development, enabling the design of\narchitectures that dynamically adapt to both local and global dependencies.\nApplications in bioinformatics, materials science, language modeling, and\nbeyond could benefit from this synthesis of relational and sequential data\nmodeling, setting the stage for interpretable and generalizable modeling\nstrategies.\n","arxiv_id":"http://arxiv.org/abs/2501.02393v2","authors":["Markus J. Buehler"]},{"title":"Toward Inclusive Educational AI: Auditing Frontier LLMs through a\n Multiplexity Lens","abstract":" As large language models (LLMs) like GPT-4 and Llama 3 become integral to\neducational contexts, concerns are mounting over the cultural biases, power\nimbalances, and ethical limitations embedded within these technologies. Though\ngenerative AI tools aim to enhance learning experiences, they often reflect\nvalues rooted in Western, Educated, Industrialized, Rich, and Democratic\n(WEIRD) cultural paradigms, potentially sidelining diverse global perspectives.\nThis paper proposes a framework to assess and mitigate cultural bias within\nLLMs through the lens of applied multiplexity. Multiplexity, inspired by\nSenturk et al. and rooted in Islamic and other wisdom traditions, emphasizes\nthe coexistence of diverse cultural viewpoints, supporting a multi-layered\nepistemology that integrates both empirical sciences and normative values. Our\nanalysis reveals that LLMs frequently exhibit cultural polarization, with\nbiases appearing in both overt responses and subtle contextual cues. To address\ninherent biases and incorporate multiplexity in LLMs, we propose two\nstrategies: \\textit{Contextually-Implemented Multiplex LLMs}, which embed\nmultiplex principles directly into the system prompt, influencing LLM outputs\nat a foundational level and independent of individual prompts, and\n\\textit{Multi-Agent System (MAS)-Implemented Multiplex LLMs}, where multiple\nLLM agents, each representing distinct cultural viewpoints, collaboratively\ngenerate a balanced, synthesized response. Our findings demonstrate that as\nmitigation strategies evolve from contextual prompting to MAS-implementation,\ncultural inclusivity markedly improves, evidenced by a significant rise in the\nPerspectives Distribution Score (PDS) and a PDS Entropy increase from 3.25\\% at\nbaseline to 98\\% with the MAS-Implemented Multiplex LLMs. Sentiment analysis\nfurther shows a shift towards positive sentiment across cultures,...\n","arxiv_id":"http://arxiv.org/abs/2501.03259v1","authors":["Abdullah Mushtaq","Muhammad Rafay Naeem","Muhammad Imran Taj","Ibrahim Ghaznavi","Junaid Qadir"]},{"title":"LLM Content Moderation and User Satisfaction: Evidence from Response\n Refusals in Chatbot Arena","abstract":" LLM safety and ethical alignment are widely discussed, but the impact of\ncontent moderation on user satisfaction remains underexplored. To address this,\nwe analyze nearly 50,000 Chatbot Arena response-pairs using a novel fine-tuned\nRoBERTa model, that we trained on hand-labeled data to disentangle refusals due\nto ethical concerns from other refusals due to technical disabilities or lack\nof information. Our findings reveal a significant refusal penalty on content\nmoderation, with users choosing ethical-based refusals roughly one-fourth as\noften as their preferred LLM response compared to standard responses. However,\nthe context and phrasing play critical roles: refusals on highly sensitive\nprompts, such as illegal content, achieve higher win rates than less sensitive\nethical concerns, and longer responses closely aligned with the prompt perform\nbetter. These results emphasize the need for nuanced moderation strategies that\nbalance ethical safeguards with user satisfaction. Moreover, we find that the\nrefusal penalty is notably lower in evaluations using the LLM-as-a-Judge\nmethod, highlighting discrepancies between user and automated assessments.\n","arxiv_id":"http://arxiv.org/abs/2501.03266v1","authors":["Stefan Pasch"]},{"title":"Cyber Shadows: Neutralizing Security Threats with AI and Targeted Policy\n Measures","abstract":" The digital age, driven by the AI revolution, brings significant\nopportunities but also conceals security threats, which we refer to as cyber\nshadows. These threats pose risks at individual, organizational, and societal\nlevels. This paper examines the systemic impact of these cyber threats and\nproposes a comprehensive cybersecurity strategy that integrates AI-driven\nsolutions, such as Intrusion Detection Systems (IDS), with targeted policy\ninterventions. By combining technological and regulatory measures, we create a\nmultilevel defense capable of addressing both direct threats and indirect\nnegative externalities. We emphasize that the synergy between AI-driven\nsolutions and policy interventions is essential for neutralizing cyber threats\nand mitigating their negative impact on the digital economy. Finally, we\nunderscore the need for continuous adaptation of these strategies, especially\nin response to the rapid advancement of autonomous AI-driven attacks, to ensure\nthe creation of secure and resilient digital ecosystems.\n","arxiv_id":"http://arxiv.org/abs/2501.09025v1","authors":["Marc Schmitt","Pantelis Koutroumpis"]},{"title":"Zero-Shot Statistical Tests for LLM-Generated Text Detection using\n Finite Sample Concentration Inequalities","abstract":" Verifying the provenance of content is crucial to the function of many\norganizations, e.g., educational institutions, social media platforms, firms,\netc. This problem is becoming increasingly difficult as text generated by Large\nLanguage Models (LLMs) becomes almost indistinguishable from human-generated\ncontent. In addition, many institutions utilize in-house LLMs and want to\nensure that external, non-sanctioned LLMs do not produce content within the\ninstitution. In this paper, we answer the following question: Given a piece of\ntext, can we identify whether it was produced by LLM $A$ or $B$ (where $B$ can\nbe a human)? We model LLM-generated text as a sequential stochastic process\nwith complete dependence on history and design zero-shot statistical tests to\ndistinguish between (i) the text generated by two different sets of LLMs $A$\n(in-house) and $B$ (non-sanctioned) and also (ii) LLM-generated and\nhuman-generated texts. We prove that the type I and type II errors for our\ntests decrease exponentially in the text length. In designing our tests, we\nderive concentration inequalities on the difference between log-perplexity and\nthe average entropy of the string under $A$. Specifically, for a given string,\nwe demonstrate that if the string is generated by $A$, the log-perplexity of\nthe string under $A$ converges to the average entropy of the string under $A$,\nexcept with an exponentially small probability in string length. We also show\nthat if $B$ generates the text, except with an exponentially small probability\nin string length, the log-perplexity of the string under $A$ converges to the\naverage cross-entropy of $B$ and $A$. Lastly, we present preliminary\nexperimental results to support our theoretical results. By enabling guaranteed\n(with high probability) finding of the origin of harmful LLM-generated text\nwith arbitrary size, we can help combat misinformation.\n","arxiv_id":"http://arxiv.org/abs/2501.02406v2","authors":["Tara Radvand","Mojtaba Abdolmaleki","Mohamed Mostagir","Ambuj Tewari"]}] \ No newline at end of file diff --git a/README.md b/README.md new file mode 100644 index 0000000..45496cc --- /dev/null +++ b/README.md @@ -0,0 +1,167 @@ +# Paper System + +A tool to fetch, filter, and process arXiv papers using LLM-based criteria. The system consists of three main components: + +1. **arxiv-processor**: Fetches papers from arXiv based on category and date range +2. **llm_processor**: Evaluates papers using specified criteria through an LLM +3. **json2md**: Generates formatted markdown output of accepted/rejected papers + +## Installation + +1. Ensure you have Go installed (1.20 or later) +2. Clone this repository +3. Build the system: +```bash +go build -o paper-system +``` + +## Configuration + +The system requires an OpenRouter API key for LLM processing. Set it as an environment variable: + +```bash +export OPENROUTER_API_KEY=your-api-key +``` + +## Usage + +The system can operate in two modes: + +### 1. ArXiv Fetch Mode + +Fetches papers from arXiv, processes them with LLM, and generates markdown output: + +```bash +./paper-system \ + -start 20240101 \ + -end 20240131 \ + -search cs.AI \ + -criteria criteria.txt \ + -output papers.md +``` + +Required flags: +- `-start`: Start date in YYYYMMDD format +- `-end`: End date in YYYYMMDD format +- `-search`: arXiv category/search query (e.g., 'cs.AI', 'physics.comp-ph') +- `-criteria`: Path to filter criteria file + +Optional flags: +- `-output`: Output markdown file path (default: papers.md) +- `-model`: LLM model to use (default: nvidia/llama-3.1-nemotron-70b-instruct) +- `-max-results`: Maximum number of papers to retrieve (default: 100, max: 2000) + +### 2. Input JSON Mode + +Process an existing JSON file of papers (useful for running different criteria against the same dataset): + +```bash +./paper-system \ + -input-json papers.json \ + -criteria new-criteria.txt \ + -output results.md +``` + +Required flags: +- `-input-json`: Path to input JSON file +- `-criteria`: Path to filter criteria file + +Optional flags: +- `-output`: Output markdown file path (default: papers.md) +- `-model`: LLM model to use (default: nvidia/llama-3.1-nemotron-70b-instruct) + +## Input/Output Files + +### Criteria File Format + +Create a text file with your evaluation criteria. Example: +``` +Please evaluate this paper based on the following criteria: + +1. Practical Applications: Does the paper demonstrate clear real-world applications? +2. Experimental Results: Are there quantitative metrics and thorough evaluations? +3. Technical Innovation: Does the paper present novel techniques or improvements? + +Respond with a JSON object containing: +{ + "decision": "ACCEPT" or "REJECT", + "explanation": "Detailed reasoning for the decision" +} +``` + +### Output Format + +The system generates two types of output: + +1. **papers.json**: Raw paper data in JSON format (when fetching from arXiv) +```json +[ + { + "title": "Paper Title", + "abstract": "Paper abstract...", + "arxiv_id": "2401.12345", + "authors": ["Author 1", "Author 2"] + } +] +``` + +2. **papers.md**: Formatted markdown with accepted/rejected papers +```markdown +# Accepted Papers + +## [Paper Title](https://arxiv.org/abs/2401.12345) +**arXiv ID:** 2401.12345 + +**Abstract:** +> Paper abstract... + +**Decision Explanation:** Meets criteria for practical applications... + +--- + +# Rejected Papers +... +``` + +## Workflow Examples + +### Basic Usage +```bash +# Fetch and evaluate papers +./paper-system -start 20240101 -end 20240131 -search cs.AI -criteria criteria.txt + +# This creates: +# - papers.json (raw paper data) +# - papers.md (evaluation results) +``` + +### Multiple Evaluations +```bash +# 1. First fetch papers +./paper-system -start 20240101 -end 20240131 -search cs.AI -criteria criteria1.txt -output results1.md + +# 2. Run different criteria on the same papers +./paper-system -input-json papers.json -criteria criteria2.txt -output results2.md +``` + +### Fetching More Papers +```bash +# Fetch up to 2000 papers +./paper-system -start 20240101 -end 20240131 -search cs.AI -criteria criteria.txt -max-results 2000 +``` + +## Error Handling + +The system includes several safeguards: + +- Validates all required parameters +- Ensures max-results is between 1 and 2000 +- Prevents mixing of arXiv and input JSON modes +- Retries LLM processing on failure +- Maintains temporary files for debugging + +## Notes + +- The system preserves papers.json when fetching from arXiv, allowing for future reuse +- Temporary files (temp_input.json, temp_output.json) are automatically cleaned up +- The LLM processor uses a batch size of 32 papers for efficient processing diff --git a/arxiv-processor/20250123-papers.json b/arxiv-processor/20250123-papers.json new file mode 100644 index 0000000..b1d4322 --- /dev/null +++ b/arxiv-processor/20250123-papers.json @@ -0,0 +1,277 @@ +[ + { + "title": "Offline Critic-Guided Diffusion Policy for Multi-User Delay-Constrained\n Scheduling", + "abstract": "Effective multi-user delay-constrained scheduling is crucial in various\nreal-world applications, such as instant messaging, live streaming, and data\ncenter management. In these scenarios, schedulers must make real-time decisions\nto satisfy both delay and resource constraints without prior knowledge of\nsystem dynamics, which are often time-varying and challenging to estimate.\nCurrent learning-based methods typically require interactions with actual\nsystems during the training stage, which can be difficult or impractical, as it\nis capable of significantly degrading system performance and incurring\nsubstantial service costs. To address these challenges, we propose a novel\noffline reinforcement learning-based algorithm, named \\underline{S}cheduling By\n\\underline{O}ffline Learning with \\underline{C}ritic Guidance and\n\\underline{D}iffusion Generation (SOCD), to learn efficient scheduling policies\npurely from pre-collected \\emph{offline data}. SOCD innovatively employs a\ndiffusion-based policy network, complemented by a sampling-free critic network\nfor policy guidance. By integrating the Lagrangian multiplier optimization into\nthe offline reinforcement learning, SOCD effectively trains high-quality\nconstraint-aware policies exclusively from available datasets, eliminating the\nneed for online interactions with the system. Experimental results demonstrate\nthat SOCD is resilient to various system dynamics, including partially\nobservable and large-scale environments, and delivers superior performance\ncompared to existing methods.", + "arxiv_id": "2501.12942v1" + }, + { + "title": "Evolution and The Knightian Blindspot of Machine Learning", + "abstract": "This paper claims that machine learning (ML) largely overlooks an important\nfacet of general intelligence: robustness to a qualitatively unknown future in\nan open world. Such robustness relates to Knightian uncertainty (KU) in\neconomics, i.e. uncertainty that cannot be quantified, which is excluded from\nconsideration in ML's key formalisms. This paper aims to identify this blind\nspot, argue its importance, and catalyze research into addressing it, which we\nbelieve is necessary to create truly robust open-world AI. To help illuminate\nthe blind spot, we contrast one area of ML, reinforcement learning (RL), with\nthe process of biological evolution. Despite staggering ongoing progress, RL\nstill struggles in open-world situations, often failing under unforeseen\nsituations. For example, the idea of zero-shot transferring a self-driving car\npolicy trained only in the US to the UK currently seems exceedingly ambitious.\nIn dramatic contrast, biological evolution routinely produces agents that\nthrive within an open world, sometimes even to situations that are remarkably\nout-of-distribution (e.g. invasive species; or humans, who do undertake such\nzero-shot international driving). Interestingly, evolution achieves such\nrobustness without explicit theory, formalisms, or mathematical gradients. We\nexplore the assumptions underlying RL's typical formalisms, showing how they\nlimit RL's engagement with the unknown unknowns characteristic of an\never-changing complex world. Further, we identify mechanisms through which\nevolutionary processes foster robustness to novel and unpredictable challenges,\nand discuss potential pathways to algorithmically embody them. The conclusion\nis that the intriguing remaining fragility of ML may result from blind spots in\nits formalisms, and that significant gains may result from direct confrontation\nwith the challenge of KU.", + "arxiv_id": "2501.13075v1" + }, + { + "title": "Boosting MCTS with Free Energy Minimization", + "abstract": "Active Inference, grounded in the Free Energy Principle, provides a powerful\nlens for understanding how agents balance exploration and goal-directed\nbehavior in uncertain environments. Here, we propose a new planning framework,\nthat integrates Monte Carlo Tree Search (MCTS) with active inference objectives\nto systematically reduce epistemic uncertainty while pursuing extrinsic\nrewards. Our key insight is that MCTS already renowned for its search\nefficiency can be naturally extended to incorporate free energy minimization by\nblending expected rewards with information gain. Concretely, the Cross-Entropy\nMethod (CEM) is used to optimize action proposals at the root node, while tree\nexpansions leverage reward modeling alongside intrinsic exploration bonuses.\nThis synergy allows our planner to maintain coherent estimates of value and\nuncertainty throughout planning, without sacrificing computational\ntractability. Empirically, we benchmark our planner on a diverse set of\ncontinuous control tasks, where it demonstrates performance gains over both\nstandalone CEM and MCTS with random rollouts.", + "arxiv_id": "2501.13083v1" + }, + { + "title": "A Unified Invariant Learning Framework for Graph Classification", + "abstract": "Invariant learning demonstrates substantial potential for enhancing the\ngeneralization of graph neural networks (GNNs) with out-of-distribution (OOD)\ndata. It aims to recognize stable features in graph data for classification,\nbased on the premise that these features causally determine the target label,\nand their influence is invariant to changes in distribution. Along this line,\nmost studies have attempted to pinpoint these stable features by emphasizing\nexplicit substructures in the graph, such as masked or attentive subgraphs, and\nprimarily enforcing the invariance principle in the semantic space, i.e., graph\nrepresentations. However, we argue that focusing only on the semantic space may\nnot accurately identify these stable features. To address this, we introduce\nthe Unified Invariant Learning (UIL) framework for graph classification. It\nprovides a unified perspective on invariant graph learning, emphasizing both\nstructural and semantic invariance principles to identify more robust stable\nfeatures. In the graph space, UIL adheres to the structural invariance\nprinciple by reducing the distance between graphons over a set of stable\nfeatures across different environments. Simultaneously, to confirm semantic\ninvariance, UIL underscores that the acquired graph representations should\ndemonstrate exemplary performance across diverse environments. We present both\ntheoretical and empirical evidence to confirm our method's ability to recognize\nsuperior stable features. Moreover, through a series of comprehensive\nexperiments complemented by in-depth analyses, we demonstrate that UIL\nconsiderably enhances OOD generalization, surpassing the performance of leading\nbaseline methods. Our codes are available at https://github.com/yongduosui/UIL.", + "arxiv_id": "2501.12595v1" + }, + { + "title": "Kimi k1.5: Scaling Reinforcement Learning with LLMs", + "abstract": "Language model pretraining with next token prediction has proved effective\nfor scaling compute but is limited to the amount of available training data.\nScaling reinforcement learning (RL) unlocks a new axis for the continued\nimprovement of artificial intelligence, with the promise that large language\nmodels (LLMs) can scale their training data by learning to explore with\nrewards. However, prior published work has not produced competitive results. In\nlight of this, we report on the training practice of Kimi k1.5, our latest\nmulti-modal LLM trained with RL, including its RL training techniques,\nmulti-modal data recipes, and infrastructure optimization. Long context scaling\nand improved policy optimization methods are key ingredients of our approach,\nwhich establishes a simplistic, effective RL framework without relying on more\ncomplex techniques such as Monte Carlo tree search, value functions, and\nprocess reward models. Notably, our system achieves state-of-the-art reasoning\nperformance across multiple benchmarks and modalities -- e.g., 77.5 on AIME,\n96.2 on MATH 500, 94-th percentile on Codeforces, 74.9 on MathVista -- matching\nOpenAI's o1. Moreover, we present effective long2short methods that use\nlong-CoT techniques to improve short-CoT models, yielding state-of-the-art\nshort-CoT reasoning results -- e.g., 60.8 on AIME, 94.6 on MATH500, 47.3 on\nLiveCodeBench -- outperforming existing short-CoT models such as GPT-4o and\nClaude Sonnet 3.5 by a large margin (up to +550%).", + "arxiv_id": "2501.12599v1" + }, + { + "title": "GATE: Adaptive Learning with Working Memory by Information Gating in\n Multi-lamellar Hippocampal Formation", + "abstract": "Hippocampal formation (HF) can rapidly adapt to varied environments and build\nflexible working memory (WM). To mirror the HF's mechanism on generalization\nand WM, we propose a model named Generalization and Associative Temporary\nEncoding (GATE), which deploys a 3-D multi-lamellar dorsoventral (DV)\narchitecture, and learns to build up internally representation from externally\ndriven information layer-wisely. In each lamella, regions of HF:\nEC3-CA1-EC5-EC3 forms a re-entrant loop that discriminately maintains\ninformation by EC3 persistent activity, and selectively readouts the retained\ninformation by CA1 neurons. CA3 and EC5 further provides gating function that\ncontrols these processes. After learning complex WM tasks, GATE forms neuron\nrepresentations that align with experimental records, including splitter, lap,\nevidence, trace, delay-active cells, as well as conventional place cells.\nCrucially, DV architecture in GATE also captures information, range from\ndetailed to abstract, which enables a rapid generalization ability when cue,\nenvironment or task changes, with learned representations inherited. GATE\npromises a viable framework for understanding the HF's flexible memory\nmechanisms and for progressively developing brain-inspired intelligent systems.", + "arxiv_id": "2501.12615v1" + }, + { + "title": "Deep Learning-Based Identification of Inconsistent Method Names: How Far\n Are We?", + "abstract": "Concise and meaningful method names are crucial for program comprehension and\nmaintenance. However, method names may become inconsistent with their\ncorresponding implementations, causing confusion and errors. Several deep\nlearning (DL)-based approaches have been proposed to identify such\ninconsistencies, with initial evaluations showing promising results. However,\nthese evaluations typically use a balanced dataset, where the number of\ninconsistent and consistent names are equal. This setup, along with flawed\ndataset construction, leads to false positives, making reported performance\nless reliable in real-world scenarios, where most method names are consistent.\nIn this paper, we present an empirical study that evaluates state-of-the-art\nDL-based methods for identifying inconsistent method names. We create a new\nbenchmark by combining automatic identification from commit histories and\nmanual developer inspections, reducing false positives. We evaluate five\nrepresentative DL approaches (one retrieval-based and four generation-based) on\nthis benchmark. Our results show that performance drops substantially when\nmoving from the balanced dataset to the new benchmark. We further conduct\nquantitative and qualitative analyses to understand the strengths and\nweaknesses of the approaches. Retrieval-based methods perform well on simple\nmethods and those with popular name sub-tokens but fail due to inefficient\nrepresentation techniques. Generation-based methods struggle with inaccurate\nsimilarity calculations and immature name generation. Based on these findings,\nwe propose improvements using contrastive learning and large language models\n(LLMs). Our study suggests that significant improvements are needed before\nthese DL approaches can be effectively applied to real-world software systems.", + "arxiv_id": "2501.12617v1" + }, + { + "title": "Adaptive Data Exploitation in Deep Reinforcement Learning", + "abstract": "We introduce ADEPT: Adaptive Data ExPloiTation, a simple yet powerful\nframework to enhance the **data efficiency** and **generalization** in deep\nreinforcement learning (RL). Specifically, ADEPT adaptively manages the use of\nsampled data across different learning stages via multi-armed bandit (MAB)\nalgorithms, optimizing data utilization while mitigating overfitting. Moreover,\nADEPT can significantly reduce the computational overhead and accelerate a wide\nrange of RL algorithms. We test ADEPT on benchmarks including Procgen,\nMiniGrid, and PyBullet. Extensive simulation demonstrates that ADEPT can\nachieve superior performance with remarkable computational efficiency, offering\na practical solution to data-efficient RL. Our code is available at\nhttps://github.com/yuanmingqi/ADEPT.", + "arxiv_id": "2501.12620v1" + }, + { + "title": "Towards Robust Multi-tab Website Fingerprinting", + "abstract": "Website fingerprinting enables an eavesdropper to determine which websites a\nuser is visiting over an encrypted connection. State-of-the-art website\nfingerprinting (WF) attacks have demonstrated effectiveness even against\nTor-protected network traffic. However, existing WF attacks have critical\nlimitations on accurately identifying websites in multi-tab browsing sessions,\nwhere the holistic pattern of individual websites is no longer preserved, and\nthe number of tabs opened by a client is unknown a priori. In this paper, we\npropose ARES, a novel WF framework natively designed for multi-tab WF attacks.\nARES formulates the multi-tab attack as a multi-label classification problem\nand solves it using the novel Transformer-based models. Specifically, ARES\nextracts local patterns based on multi-level traffic aggregation features and\nutilizes the improved self-attention mechanism to analyze the correlations\nbetween these local patterns, effectively identifying websites. We implement a\nprototype of ARES and extensively evaluate its effectiveness using our\nlarge-scale datasets collected over multiple months. The experimental results\nillustrate that ARES achieves optimal performance in several realistic\nscenarios. Further, ARES remains robust even against various WF defenses.", + "arxiv_id": "2501.12622v1" + }, + { + "title": "Inverse Reinforcement Learning with Switching Rewards and History\n Dependency for Characterizing Animal Behaviors", + "abstract": "Traditional approaches to studying decision-making in neuroscience focus on\nsimplified behavioral tasks where animals perform repetitive, stereotyped\nactions to receive explicit rewards. While informative, these methods constrain\nour understanding of decision-making to short timescale behaviors driven by\nexplicit goals. In natural environments, animals exhibit more complex,\nlong-term behaviors driven by intrinsic motivations that are often\nunobservable. Recent works in time-varying inverse reinforcement learning (IRL)\naim to capture shifting motivations in long-term, freely moving behaviors.\nHowever, a crucial challenge remains: animals make decisions based on their\nhistory, not just their current state. To address this, we introduce SWIRL\n(SWitching IRL), a novel framework that extends traditional IRL by\nincorporating time-varying, history-dependent reward functions. SWIRL models\nlong behavioral sequences as transitions between short-term decision-making\nprocesses, each governed by a unique reward function. SWIRL incorporates\nbiologically plausible history dependency to capture how past decisions and\nenvironmental contexts shape behavior, offering a more accurate description of\nanimal decision-making. We apply SWIRL to simulated and real-world animal\nbehavior datasets and show that it outperforms models lacking history\ndependency, both quantitatively and qualitatively. This work presents the first\nIRL model to incorporate history-dependent policies and rewards to advance our\nunderstanding of complex, naturalistic decision-making in animals.", + "arxiv_id": "2501.12633v1" + }, + { + "title": "Dynamics of Toxicity in Political Podcasts", + "abstract": "Toxicity in digital media poses significant challenges, yet little attention\nhas been given to its dynamics within the rapidly growing medium of podcasts.\nThis paper addresses this gap by analyzing political podcast data to study the\nemergence and propagation of toxicity, focusing on conversation\nchains-structured reply patterns within podcast transcripts. Leveraging\nstate-of-the-art transcription models and advanced conversational analysis\ntechniques, we systematically examine toxic discourse in over 30 popular\npolitical podcasts in the United States. Our key contributions include: (1)\ncreating a comprehensive dataset of transcribed and diarized political\npodcasts, identifying thousands of toxic instances using Google's Perspective\nAPI, (2) uncovering concerning trends where a majority of episodes contain at\nleast one toxic instance, (3) introducing toxic conversation chains and\nanalyzing their structural and linguistic properties, revealing characteristics\nsuch as longer durations, repetitive patterns, figurative language, and\nemotional cues tied to anger and annoyance, (4) identifying demand-related\nwords like 'want', 'like', and 'know' as precursors to toxicity, and (5)\ndeveloping predictive models to anticipate toxicity shifts based on annotated\nchange points. Our findings provide critical insights into podcast toxicity and\nestablish a foundation for future research on real-time monitoring and\nintervention mechanisms to foster healthier discourse in this influential\nmedium.", + "arxiv_id": "2501.12640v1" + }, + { + "title": "The potential -- and the pitfalls -- of using pre-trained language\n models as cognitive science theories", + "abstract": "Many studies have evaluated the cognitive alignment of Pre-trained Language\nModels (PLMs), i.e., their correspondence to adult performance across a range\nof cognitive domains. Recently, the focus has expanded to the developmental\nalignment of these models: identifying phases during training where\nimprovements in model performance track improvements in children's thinking\nover development. However, there are many challenges to the use of PLMs as\ncognitive science theories, including different architectures, different\ntraining data modalities and scales, and limited model interpretability. In\nthis paper, we distill lessons learned from treating PLMs, not as engineering\nartifacts but as cognitive science and developmental science models. We review\nassumptions used by researchers to map measures of PLM performance to measures\nof human performance. We identify potential pitfalls of this approach to\nunderstanding human thinking, and we end by enumerating criteria for using PLMs\nas credible accounts of cognition and cognitive development.", + "arxiv_id": "2501.12651v1" + }, + { + "title": "NBDI: A Simple and Efficient Termination Condition for Skill Extraction\n from Task-Agnostic Demonstrations", + "abstract": "Intelligent agents are able to make decisions based on different levels of\ngranularity and duration. Recent advances in skill learning enabled the agent\nto solve complex, long-horizon tasks by effectively guiding the agent in\nchoosing appropriate skills. However, the practice of using fixed-length skills\ncan easily result in skipping valuable decision points, which ultimately limits\nthe potential for further exploration and faster policy learning. In this work,\nwe propose to learn a simple and efficient termination condition that\nidentifies decision points through a state-action novelty module that leverages\nagent experience data. Our approach, Novelty-based Decision Point\nIdentification (NBDI), outperforms previous baselines in complex, long-horizon\ntasks, and remains effective even in the presence of significant variations in\nthe environment configurations of downstream tasks, highlighting the importance\nof decision point identification in skill learning.", + "arxiv_id": "2501.12668v1" + }, + { + "title": "Growth strategies for arbitrary DAG neural architectures", + "abstract": "Deep learning has shown impressive results obtained at the cost of training\nhuge neural networks. However, the larger the architecture, the higher the\ncomputational, financial, and environmental costs during training and\ninference. We aim at reducing both training and inference durations. We focus\non Neural Architecture Growth, which can increase the size of a small model\nwhen needed, directly during training using information from the\nbackpropagation. We expand existing work and freely grow neural networks in the\nform of any Directed Acyclic Graph by reducing expressivity bottlenecks in the\narchitecture. We explore strategies to reduce excessive computations and steer\nnetwork growth toward more parameter-efficient architectures.", + "arxiv_id": "2501.12690v1" + }, + { + "title": "EvidenceMap: Unleashing the Power of Small Language Models with Evidence\n Analysis for Biomedical Question Answering", + "abstract": "Current LLM-based approaches improve question answering performance by\nleveraging the internal reasoning abilities of models or incorporating external\nknowledge. However, when humans address professional problems, it is essential\nto explicitly analyze the multifaceted relationships from multiple pieces and\ndiverse sources of evidence to achieve better answers. In this study, we\npropose a novel generative question answering framework for the biomedical\ndomain, named EvidenceMap, which explicitly learns and incorporates evidence\nanalysis with small language models (SLMs). The framework describes an evidence\nmap for each question and fully utilizes an SLM to derive the representation of\nthe supportive evaluation, the logical correlation, and the summarization of\nthe related evidence, which facilitates an analysis-augmented generation with\nanother SLM in an autoregressive way. Extensive experiments have shown that\nintroducing an evidence analysis learning process can significantly outperform\nlarger models and popular LLM reasoning methods.", + "arxiv_id": "2501.12746v1" + }, + { + "title": "NExtLong: Toward Effective Long-Context Training without Long Documents", + "abstract": "Large language models (LLMs) with extended context windows have made\nsignificant strides yet remain a challenge due to the scarcity of long\ndocuments. Existing methods tend to synthesize long-context data but lack a\nclear mechanism to reinforce the long-range dependency modeling. To address\nthis limitation, we propose NExtLong, a novel framework for synthesizing\nlong-context data through Negative document Extension. NExtLong decomposes a\ndocument into multiple meta-chunks and extends the context by interleaving hard\nnegative distractors retrieved from pretraining corpora. This approach compels\nthe model to discriminate long-range dependent context from distracting\ncontent, enhancing its ability to model long-range dependencies. Extensive\nexperiments demonstrate that NExtLong achieves significant performance\nimprovements on the HELMET and RULER benchmarks compared to existing\nlong-context synthesis approaches and leading models, which are trained on\nnon-synthetic long documents. These findings highlight NExtLong's ability to\nreduce reliance on non-synthetic long documents, making it an effective\nframework for developing advanced long-context LLMs.", + "arxiv_id": "2501.12766v1" + }, + { + "title": "Revisit Self-Debugging with Self-Generated Tests for Code Generation", + "abstract": "Large language models (LLMs) have shown significant advancements in code\ngeneration, but still face challenges on tasks beyond their basic capabilities.\nRecently, the notion of self-debugging has been proposed to boost the\nperformance of code generation by leveraging execution feedback from tests.\nDespite its promise, the availability of high-quality tests in real-world\nscenarios is limited. In this context, self-debugging with self-generated tests\nis a promising solution but lacks a full exploration of its limitations and\npractical potential. Therefore, we investigate its efficacy on diverse\nprogramming problems. To deepen our understanding, we propose two distinct\nparadigms for the process: post-execution and in-execution self-debugging.\nWithin the scope of self-contained Python programming tasks, we find that\npost-execution self-debugging struggles on basic problems but shows potential\nfor improvement on competitive ones, due to the bias introduced by\nself-generated tests. On the other hand, in-execution self-debugging enables\nLLMs to mitigate the bias by solely leveraging intermediate states during\nexecution, thereby enhancing code generation.", + "arxiv_id": "2501.12793v1" + }, + { + "title": "Unveiling Zero-Space Detection: A Novel Framework for Autonomous\n Ransomware Identification in High-Velocity Environments", + "abstract": "Modern cybersecurity landscapes increasingly demand sophisticated detection\nframeworks capable of identifying evolving threats with precision and\nadaptability. The proposed Zero-Space Detection framework introduces a novel\napproach that dynamically identifies latent behavioral patterns through\nunsupervised clustering and advanced deep learning techniques. Designed to\naddress the limitations of signature-based and heuristic methods, it operates\neffectively in high-velocity environments by integrating multi-phase filtering\nand ensemble learning for refined decision-making. Experimental evaluation\nreveals high detection rates across diverse ransomware families, including\nLockBit, Conti, REvil, and BlackMatter, while maintaining low false positive\nrates and scalable performance. Computational overhead remains minimal, with\naverage processing times ensuring compatibility with real-time systems even\nunder peak operational loads. The framework demonstrates resilience against\nadversarial strategies such as obfuscation and encryption speed variability,\nwhich frequently challenge conventional detection systems. Analysis across\nmultiple data sources highlights its versatility in handling diverse file types\nand operational contexts. Comprehensive metrics, including detection\nprobability, latency, and resource efficiency, validate its efficacy under\nreal-world conditions. Through its modular architecture, the framework achieves\nseamless integration with existing cybersecurity infrastructures without\nsignificant reconfiguration. The results demonstrate its robustness and\nscalability, offering a transformative paradigm for ransomware identification\nin dynamic and resource-constrained environments.", + "arxiv_id": "2501.12811v1" + }, + { + "title": "To Measure or Not: A Cost-Sensitive, Selective Measuring Environment for\n Agricultural Management Decisions with Reinforcement Learning", + "abstract": "Farmers rely on in-field observations to make well-informed crop management\ndecisions to maximize profit and minimize adverse environmental impact.\nHowever, obtaining real-world crop state measurements is labor-intensive,\ntime-consuming and expensive. In most cases, it is not feasible to gather crop\nstate measurements before every decision moment. Moreover, in previous research\npertaining to farm management optimization, these observations are often\nassumed to be readily available without any cost, which is unrealistic. Hence,\nenabling optimization without the need to have temporally complete crop state\nobservations is important. An approach to that problem is to include measuring\nas part of decision making. As a solution, we apply reinforcement learning (RL)\nto recommend opportune moments to simultaneously measure crop features and\napply nitrogen fertilizer. With realistic considerations, we design an RL\nenvironment with explicit crop feature measuring costs. While balancing costs,\nwe find that an RL agent, trained with recurrent PPO, discovers adaptive\nmeasuring policies that follow critical crop development stages, with results\naligned by what domain experts would consider a sensible approach. Our results\nhighlight the importance of measuring when crop feature measurements are not\nreadily available.", + "arxiv_id": "2501.12823v1" + }, + { + "title": "GAMED-Snake: Gradient-aware Adaptive Momentum Evolution Deep Snake Model\n for Multi-organ Segmentation", + "abstract": "Multi-organ segmentation is a critical yet challenging task due to complex\nanatomical backgrounds, blurred boundaries, and diverse morphologies. This\nstudy introduces the Gradient-aware Adaptive Momentum Evolution Deep Snake\n(GAMED-Snake) model, which establishes a novel paradigm for contour-based\nsegmentation by integrating gradient-based learning with adaptive momentum\nevolution mechanisms. The GAMED-Snake model incorporates three major\ninnovations: First, the Distance Energy Map Prior (DEMP) generates a\npixel-level force field that effectively attracts contour points towards the\ntrue boundaries, even in scenarios with complex backgrounds and blurred edges.\nSecond, the Differential Convolution Inception Module (DCIM) precisely extracts\ncomprehensive energy gradients, significantly enhancing segmentation accuracy.\nThird, the Adaptive Momentum Evolution Mechanism (AMEM) employs cross-attention\nto establish dynamic features across different iterations of evolution,\nenabling precise boundary alignment for diverse morphologies. Experimental\nresults on four challenging multi-organ segmentation datasets demonstrate that\nGAMED-Snake improves the mDice metric by approximately 2% compared to\nstate-of-the-art methods. Code will be available at\nhttps://github.com/SYSUzrc/GAMED-Snake.", + "arxiv_id": "2501.12844v1" + }, + { + "title": "As Confidence Aligns: Exploring the Effect of AI Confidence on Human\n Self-confidence in Human-AI Decision Making", + "abstract": "Complementary collaboration between humans and AI is essential for human-AI\ndecision making. One feasible approach to achieving it involves accounting for\nthe calibrated confidence levels of both AI and users. However, this process\nwould likely be made more difficult by the fact that AI confidence may\ninfluence users' self-confidence and its calibration. To explore these\ndynamics, we conducted a randomized behavioral experiment. Our results indicate\nthat in human-AI decision-making, users' self-confidence aligns with AI\nconfidence and such alignment can persist even after AI ceases to be involved.\nThis alignment then affects users' self-confidence calibration. We also found\nthe presence of real-time correctness feedback of decisions reduced the degree\nof alignment. These findings suggest that users' self-confidence is not\nindependent of AI confidence, which practitioners aiming to achieve better\nhuman-AI collaboration need to be aware of. We call for research focusing on\nthe alignment of human cognition and behavior with AI.", + "arxiv_id": "2501.12868v1" + }, + { + "title": "Drone Carrier: An Integrated Unmanned Surface Vehicle for Autonomous\n Inspection and Intervention in GNSS-Denied Maritime Environment", + "abstract": "This paper introduces an innovative drone carrier concept that is applied in\nmaritime port security or offshore rescue. This system works with a\nheterogeneous system consisting of multiple Unmanned Aerial Vehicles (UAVs) and\nUnmanned Surface Vehicles (USVs) to perform inspection and intervention tasks\nin GNSS-denied or interrupted environments. The carrier, an electric catamaran\nmeasuring 4m by 7m, features a 4m by 6m deck supporting automated takeoff and\nlanding for four DJI M300 drones, along with a 10kg-payload manipulator\noperable in up to level 3 sea conditions. Utilizing an offshore gimbal camera\nfor navigation, the carrier can autonomously navigate, approach and dock with\nnon-cooperative vessels, guided by an onboard camera, LiDAR, and Doppler\nVelocity Log (DVL) over a 3 km$^2$ area. UAVs equipped with onboard\nUltra-Wideband (UWB) technology execute mapping, detection, and manipulation\ntasks using a versatile gripper designed for wet, saline conditions.\nAdditionally, two UAVs can coordinate to transport large objects to the\nmanipulator or interact directly with them. These procedures are fully\nautomated and were successfully demonstrated at the Mohammed Bin Zayed\nInternational Robotic Competition (MBZIRC2024), where the drone carrier\nequipped with four UAVS and one manipulator, automatically accomplished the\nintervention tasks in sea-level-3 (wave height 1.25m) based on the rough target\ninformation.", + "arxiv_id": "2501.12869v1" + }, + { + "title": "Reinforcement learning Based Automated Design of Differential Evolution\n Algorithm for Black-box Optimization", + "abstract": "Differential evolution (DE) algorithm is recognized as one of the most\neffective evolutionary algorithms, demonstrating remarkable efficacy in\nblack-box optimization due to its derivative-free nature. Numerous enhancements\nto the fundamental DE have been proposed, incorporating innovative mutation\nstrategies and sophisticated parameter tuning techniques to improve\nperformance. However, no single variant has proven universally superior across\nall problems. To address this challenge, we introduce a novel framework that\nemploys reinforcement learning (RL) to automatically design DE for black-box\noptimization through meta-learning. RL acts as an advanced meta-optimizer,\ngenerating a customized DE configuration that includes an optimal\ninitialization strategy, update rule, and hyperparameters tailored to a\nspecific black-box optimization problem. This process is informed by a detailed\nanalysis of the problem characteristics. In this proof-of-concept study, we\nutilize a double deep Q-network for implementation, considering a subset of 40\npossible strategy combinations and parameter optimizations simultaneously. The\nframework's performance is evaluated against black-box optimization benchmarks\nand compared with state-of-the-art algorithms. The experimental results\nhighlight the promising potential of our proposed framework.", + "arxiv_id": "2501.12881v1" + }, + { + "title": "Learning Graph Node Embeddings by Smooth Pair Sampling", + "abstract": "Random walk-based node embedding algorithms have attracted a lot of attention\ndue to their scalability and ease of implementation. Previous research has\nfocused on different walk strategies, optimization objectives, and embedding\nlearning models. Inspired by observations on real data, we take a different\napproach and propose a new regularization technique. More precisely, the\nfrequencies of node pairs generated by the skip-gram model on random walk node\nsequences follow a highly skewed distribution which causes learning to be\ndominated by a fraction of the pairs. We address the issue by designing an\nefficient sampling procedure that generates node pairs according to their {\\em\nsmoothed frequency}. Theoretical and experimental results demonstrate the\nadvantages of our approach.", + "arxiv_id": "2501.12884v1" + }, + { + "title": "Architectural Fusion Through Contextual Partitioning in Large Language\n Models: A Novel Approach to Parameterized Knowledge Integration", + "abstract": "Contextual Partitioning introduces an innovative approach to enhancing the\narchitectural design of large-scale computational models through the dynamic\nsegmentation of parameters into context-aware regions. This methodology\nemphasizes the importance of task-specific specialization, achieved through\nadaptive parameter allocation mechanisms that align with the linguistic\nfeatures of input data. Experimental evaluations demonstrated substantial\nimprovements in accuracy, perplexity, and contextual coherence across a variety\nof linguistic tasks, highlighting the adaptability and scalability of the\nproposed framework. By reducing redundancy and enhancing computational\nefficiency, Contextual Partitioning not only streamlines model operations but\nalso expands the scope of applications for advanced language processing\nsystems. The approach operates autonomously, requiring no external fine-tuning,\nthereby addressing a significant limitation in conventional parameter\noptimization techniques. Empirical results demonstrate the effectiveness of\ngradient-driven segmentation, enabling models to dynamically recalibrate and\nspecialize in response to task-specific demands. Furthermore, resource\nutilization metrics reveal notable reductions in memory usage and training\ntimes, confirming the efficiency of the approach. Observations from qualitative\nanalyses illustrate improved contextual coherence and logical flow in generated\noutputs, reinforcing the practical value of this technique. The findings\ncollectively demonstrate the potential for Contextual Partitioning to redefine\nthe scalability and adaptability of computational language architectures in\ndiverse and complex domains.", + "arxiv_id": "2501.12901v1" + }, + { + "title": "A Novel Tracking Framework for Devices in X-ray Leveraging Supplementary\n Cue-Driven Self-Supervised Features", + "abstract": "To restore proper blood flow in blocked coronary arteries via angioplasty\nprocedure, accurate placement of devices such as catheters, balloons, and\nstents under live fluoroscopy or diagnostic angiography is crucial. Identified\nballoon markers help in enhancing stent visibility in X-ray sequences, while\nthe catheter tip aids in precise navigation and co-registering vessel\nstructures, reducing the need for contrast in angiography. However, accurate\ndetection of these devices in interventional X-ray sequences faces significant\nchallenges, particularly due to occlusions from contrasted vessels and other\ndevices and distractions from surrounding, resulting in the failure to track\nsuch small objects. While most tracking methods rely on spatial correlation of\npast and current appearance, they often lack strong motion comprehension\nessential for navigating through these challenging conditions, and fail to\neffectively detect multiple instances in the scene. To overcome these\nlimitations, we propose a self-supervised learning approach that enhances its\nspatio-temporal understanding by incorporating supplementary cues and learning\nacross multiple representation spaces on a large dataset. Followed by that, we\nintroduce a generic real-time tracking framework that effectively leverages the\npretrained spatio-temporal network and also takes the historical appearance and\ntrajectory data into account. This results in enhanced localization of multiple\ninstances of device landmarks. Our method outperforms state-of-the-art methods\nin interventional X-ray device tracking, especially stability and robustness,\nachieving an 87% reduction in max error for balloon marker detection and a 61%\nreduction in max error for catheter tip detection.", + "arxiv_id": "2501.12958v1" + }, + { + "title": "Accessible Smart Contracts Verification: Synthesizing Formal Models with\n Tamed LLMs", + "abstract": "When blockchain systems are said to be trustless, what this really means is\nthat all the trust is put into software. Thus, there are strong incentives to\nensure blockchain software is correct -- vulnerabilities here cost millions and\nbreak businesses. One of the most powerful ways of establishing software\ncorrectness is by using formal methods. Approaches based on formal methods,\nhowever, induce a significant overhead in terms of time and expertise required\nto successfully employ them. Our work addresses this critical disadvantage by\nautomating the creation of a formal model -- a mathematical abstraction of the\nsoftware system -- which is often a core task when employing formal methods. We\nperform model synthesis in three phases: we first transpile the code into model\nstubs; then we \"fill in the blanks\" using a large language model (LLM);\nfinally, we iteratively repair the generated model, on both syntactical and\nsemantical level. In this way, we significantly reduce the amount of time\nnecessary to create formal models and increase accessibility of valuable\nsoftware verification methods that rely on them. The practical context of our\nwork was reducing the time-to-value of using formal models for correctness\naudits of smart contracts.", + "arxiv_id": "2501.12972v1" + }, + { + "title": "Ehrenfeucht-Haussler Rank and Chain of Thought", + "abstract": "The notion of rank of a Boolean function has been a cornerstone in the theory\nof PAC learning, enabling quasipolynomial-time learning algorithms for\npolynomial-size decision trees. We present a novel characterization of rank,\ngrounded in the well-known Transformer architecture. We show that the rank of a\nfunction $f$ corresponds to the minimum number of Chain of Thought (CoT) steps\nrequired by a single-layer transformer decoder with hard attention to compute\n$f$. Based on this characterization we establish tight bounds on the number of\nCoT steps required for specific problems, showing that $\\ell$-fold function\ncomposition necessitates exactly $\\ell$ CoT steps. Furthermore, we analyze the\nproblem of identifying the position of the $k$-th occurrence of 1 in a Boolean\nsequence, proving that it requires $k$ CoT steps.", + "arxiv_id": "2501.12997v1" + }, + { + "title": "MONA: Myopic Optimization with Non-myopic Approval Can Mitigate\n Multi-step Reward Hacking", + "abstract": "Future advanced AI systems may learn sophisticated strategies through\nreinforcement learning (RL) that humans cannot understand well enough to safely\nevaluate. We propose a training method which avoids agents learning undesired\nmulti-step plans that receive high reward (multi-step \"reward hacks\") even if\nhumans are not able to detect that the behaviour is undesired. The method,\nMyopic Optimization with Non-myopic Approval (MONA), works by combining\nshort-sighted optimization with far-sighted reward. We demonstrate that MONA\ncan prevent multi-step reward hacking that ordinary RL causes, even without\nbeing able to detect the reward hacking and without any extra information that\nordinary RL does not get access to. We study MONA empirically in three settings\nwhich model different misalignment failure modes including 2-step environments\nwith LLMs representing delegated oversight and encoded reasoning and\nlonger-horizon gridworld environments representing sensor tampering.", + "arxiv_id": "2501.13011v1" + }, + { + "title": "Provably-Safe Neural Network Training Using Hybrid Zonotope Reachability\n Analysis", + "abstract": "Even though neural networks are being increasingly deployed in\nsafety-critical applications, it remains difficult to enforce constraints on\ntheir output, meaning that it is hard to guarantee safety in such settings.\nTowards addressing this, many existing methods seek to verify a neural\nnetwork's satisfaction of safety constraints, but do not address how to correct\nan \"unsafe\" network. On the other hand, the few works that extract a training\nsignal from verification cannot handle non-convex sets, and are either\nconservative or slow. To address these challenges, this work proposes a neural\nnetwork training method that can encourage the exact reachable set of a\nnon-convex input set through a neural network with rectified linear unit (ReLU)\nnonlinearities to avoid a non-convex unsafe region, using recent results in\nnon-convex set representation with hybrid zonotopes and extracting gradient\ninformation from mixed-integer linear programs (MILPs). The proposed method is\nfast, with the computational complexity of each training iteration comparable\nto that of solving a linear program (LP) with number of dimensions and\nconstraints linear to the number of neurons and complexity of input and unsafe\nsets. For a neural network with three hidden layers of width 30, the method was\nable to drive the reachable set of a non-convex input set with 55 generators\nand 26 constraints out of a non-convex unsafe region with 21 generators and 11\nconstraints in 490 seconds.", + "arxiv_id": "2501.13023v1" + }, + { + "title": "AdaWM: Adaptive World Model based Planning for Autonomous Driving", + "abstract": "World model based reinforcement learning (RL) has emerged as a promising\napproach for autonomous driving, which learns a latent dynamics model and uses\nit to train a planning policy. To speed up the learning process, the\npretrain-finetune paradigm is often used, where online RL is initialized by a\npretrained model and a policy learned offline. However, naively performing such\ninitialization in RL may result in dramatic performance degradation during the\nonline interactions in the new task. To tackle this challenge, we first analyze\nthe performance degradation and identify two primary root causes therein: the\nmismatch of the planning policy and the mismatch of the dynamics model, due to\ndistribution shift. We further analyze the effects of these factors on\nperformance degradation during finetuning, and our findings reveal that the\nchoice of finetuning strategies plays a pivotal role in mitigating these\neffects. We then introduce AdaWM, an Adaptive World Model based planning\nmethod, featuring two key steps: (a) mismatch identification, which quantifies\nthe mismatches and informs the finetuning strategy, and (b) alignment-driven\nfinetuning, which selectively updates either the policy or the model as needed\nusing efficient low-rank updates. Extensive experiments on the challenging\nCARLA driving tasks demonstrate that AdaWM significantly improves the\nfinetuning process, resulting in more robust and efficient performance in\nautonomous driving systems.", + "arxiv_id": "2501.13072v1" + }, + { + "title": "Attention-Driven Hierarchical Reinforcement Learning with Particle\n Filtering for Source Localization in Dynamic Fields", + "abstract": "In many real-world scenarios, such as gas leak detection or environmental\npollutant tracking, solving the Inverse Source Localization and\nCharacterization problem involves navigating complex, dynamic fields with\nsparse and noisy observations. Traditional methods face significant challenges,\nincluding partial observability, temporal and spatial dynamics,\nout-of-distribution generalization, and reward sparsity. To address these\nissues, we propose a hierarchical framework that integrates Bayesian inference\nand reinforcement learning. The framework leverages an attention-enhanced\nparticle filtering mechanism for efficient and accurate belief updates, and\nincorporates two complementary execution strategies: Attention Particle\nFiltering Planning and Attention Particle Filtering Reinforcement Learning.\nThese approaches optimize exploration and adaptation under uncertainty.\nTheoretical analysis proves the convergence of the attention-enhanced particle\nfilter, while extensive experiments across diverse scenarios validate the\nframework's superior accuracy, adaptability, and computational efficiency. Our\nresults highlight the framework's potential for broad applications in dynamic\nfield estimation tasks.", + "arxiv_id": "2501.13084v1" + }, + { + "title": "Understanding the LLM-ification of CHI: Unpacking the Impact of LLMs at\n CHI through a Systematic Literature Review", + "abstract": "Large language models (LLMs) have been positioned to revolutionize HCI, by\nreshaping not only the interfaces, design patterns, and sociotechnical systems\nthat we study, but also the research practices we use. To-date, however, there\nhas been little understanding of LLMs' uptake in HCI. We address this gap via a\nsystematic literature review of 153 CHI papers from 2020-24 that engage with\nLLMs. We taxonomize: (1) domains where LLMs are applied; (2) roles of LLMs in\nHCI projects; (3) contribution types; and (4) acknowledged limitations and\nrisks. We find LLM work in 10 diverse domains, primarily via empirical and\nartifact contributions. Authors use LLMs in five distinct roles, including as\nresearch tools or simulated users. Still, authors often raise validity and\nreproducibility concerns, and overwhelmingly study closed models. We outline\nopportunities to improve HCI research with and on LLMs, and provide guiding\nquestions for researchers to consider the validity and appropriateness of\nLLM-related work.", + "arxiv_id": "2501.12557v1" + }, + { + "title": "FedGrAINS: Personalized SubGraph Federated Learning with Adaptive\n Neighbor Sampling", + "abstract": "Graphs are crucial for modeling relational and biological data. As datasets\ngrow larger in real-world scenarios, the risk of exposing sensitive information\nincreases, making privacy-preserving training methods like federated learning\n(FL) essential to ensure data security and compliance with privacy regulations.\nRecently proposed personalized subgraph FL methods have become the de-facto\nstandard for training personalized Graph Neural Networks (GNNs) in a federated\nmanner while dealing with the missing links across clients' subgraphs due to\nprivacy restrictions. However, personalized subgraph FL faces significant\nchallenges due to the heterogeneity in client subgraphs, such as degree\ndistributions among the nodes, which complicate federated training of graph\nmodels. To address these challenges, we propose \\textit{FedGrAINS}, a novel\ndata-adaptive and sampling-based regularization method for subgraph FL.\nFedGrAINS leverages generative flow networks (GFlowNets) to evaluate node\nimportance concerning clients' tasks, dynamically adjusting the message-passing\nstep in clients' GNNs. This adaptation reflects task-optimized sampling aligned\nwith a trajectory balance objective. Experimental results demonstrate that the\ninclusion of \\textit{FedGrAINS} as a regularizer consistently improves the FL\nperformance compared to baselines that do not leverage such regularization.", + "arxiv_id": "2501.12592v1" + }, + { + "title": "HEPPO: Hardware-Efficient Proximal Policy Optimization -- A Universal\n Pipelined Architecture for Generalized Advantage Estimation", + "abstract": "This paper introduces HEPPO, an FPGA-based accelerator designed to optimize\nthe Generalized Advantage Estimation (GAE) stage in Proximal Policy\nOptimization (PPO). Unlike previous approaches that focused on trajectory\ncollection and actor-critic updates, HEPPO addresses GAE's computational\ndemands with a parallel, pipelined architecture implemented on a single\nSystem-on-Chip (SoC). This design allows for the adaptation of various hardware\naccelerators tailored for different PPO phases. A key innovation is our\nstrategic standardization technique, which combines dynamic reward\nstandardization and block standardization for values, followed by 8-bit uniform\nquantization. This method stabilizes learning, enhances performance, and\nmanages memory bottlenecks, achieving a 4x reduction in memory usage and a 1.5x\nincrease in cumulative rewards. We propose a solution on a single SoC device\nwith programmable logic and embedded processors, delivering throughput orders\nof magnitude higher than traditional CPU-GPU systems. Our single-chip solution\nminimizes communication latency and throughput bottlenecks, significantly\nboosting PPO training efficiency. Experimental results show a 30% increase in\nPPO speed and a substantial reduction in memory access time, underscoring\nHEPPO's potential for broad applicability in hardware-efficient reinforcement\nlearning algorithms.", + "arxiv_id": "2501.12703v1" + }, + { + "title": "Practical quantum federated learning and its experimental demonstration", + "abstract": "Federated learning is essential for decentralized, privacy-preserving model\ntraining in the data-driven era. Quantum-enhanced federated learning leverages\nquantum resources to address privacy and scalability challenges, offering\nsecurity and efficiency advantages beyond classical methods. However, practical\nand scalable frameworks addressing privacy concerns in the quantum computing\nera remain undeveloped. Here, we propose a practical quantum federated learning\nframework on quantum networks, utilizing distributed quantum secret keys to\nprotect local model updates and enable secure aggregation with\ninformation-theoretic security. We experimentally validate our framework on a\n4-client quantum network with a scalable structure. Extensive numerical\nexperiments on both quantum and classical datasets show that adding a quantum\nclient significantly enhances the trained global model's ability to classify\nmultipartite entangled and non-stabilizer quantum datasets. Simulations further\ndemonstrate scalability to 200 clients with classical models trained on the\nMNIST dataset, reducing communication costs by $75\\%$ through advanced model\ncompression techniques and achieving rapid training convergence. Our work\nprovides critical insights for building scalable, efficient, and quantum-secure\nmachine learning systems for the coming quantum internet era.", + "arxiv_id": "2501.12709v1" + }, + { + "title": "A Call for Critically Rethinking and Reforming Data Analysis in\n Empirical Software Engineering", + "abstract": "Context: Empirical Software Engineering (ESE) drives innovation in SE through\nqualitative and quantitative studies. However, concerns about the correct\napplication of empirical methodologies have existed since the 2006 Dagstuhl\nseminar on SE. Objective: To analyze three decades of SE research, identify\nmistakes in statistical methods, and evaluate experts' ability to detect and\naddress these issues. Methods: We conducted a literature survey of ~27,000\nempirical studies, using LLMs to classify statistical methodologies as adequate\nor inadequate. Additionally, we selected 30 primary studies and held a workshop\nwith 33 ESE experts to assess their ability to identify and resolve statistical\nissues. Results: Significant statistical issues were found in the primary\nstudies, and experts showed limited ability to detect and correct these\nmethodological problems, raising concerns about the broader ESE community's\nproficiency in this area. Conclusions. Despite our study's eventual\nlimitations, its results shed light on recurring issues from promoting\ninformation copy-and-paste from past authors' works and the continuous\npublication of inadequate approaches that promote dubious results and\njeopardize the spread of the correct statistical strategies among researchers.\nBesides, it justifies further investigation into empirical rigor in software\nengineering to expose these recurring issues and establish a framework for\nreassessing our field's foundation of statistical methodology application.\nTherefore, this work calls for critically rethinking and reforming data\nanalysis in empirical software engineering, paving the way for our work soon.", + "arxiv_id": "2501.12728v1" + }, + { + "title": "Estimating the Conformal Prediction Threshold from Noisy Labels", + "abstract": "Conformal Prediction (CP) is a method to control prediction uncertainty by\nproducing a small prediction set, ensuring a predetermined probability that the\ntrue class lies within this set. This is commonly done by defining a score,\nbased on the model predictions, and setting a threshold on this score using a\nvalidation set. In this study, we address the problem of CP calibration when we\nonly have access to a validation set with noisy labels. We show how we can\nestimate the noise-free conformal threshold based on the noisy labeled data.\nOur solution is flexible and can accommodate various modeling assumptions\nregarding the label contamination process, without needing any information\nabout the underlying data distribution or the internal mechanisms of the\nmachine learning classifier. We develop a coverage guarantee for uniform noise\nthat is effective even in tasks with a large number of classes. We dub our\napproach Noise-Aware Conformal Prediction (NACP) and show on several natural\nand medical image classification datasets, including ImageNet, that it\nsignificantly outperforms current noisy label methods and achieves results\ncomparable to those obtained with a clean validation set.", + "arxiv_id": "2501.12749v1" + }, + { + "title": "On Tradeoffs in Learning-Augmented Algorithms", + "abstract": "The field of learning-augmented algorithms has gained significant attention\nin recent years. These algorithms, using potentially inaccurate predictions,\nmust exhibit three key properties: consistency, robustness, and smoothness. In\nscenarios where distributional information about predictions is available, a\nstrong expected performance is required. Typically, the design of these\nalgorithms involves a natural tradeoff between consistency and robustness, and\nprevious works aimed to achieve Pareto-optimal tradeoffs for specific problems.\nHowever, in some settings, this comes at the expense of smoothness. This paper\ndemonstrates that certain problems involve multiple tradeoffs between\nconsistency, robustness, smoothness, and average performance.", + "arxiv_id": "2501.12770v1" + }, + { + "title": "Data re-uploading in Quantum Machine Learning for time series:\n application to traffic forecasting", + "abstract": "Accurate traffic forecasting plays a crucial role in modern Intelligent\nTransportation Systems (ITS), as it enables real-time traffic flow management,\nreduces congestion, and improves the overall efficiency of urban transportation\nnetworks. With the rise of Quantum Machine Learning (QML), it has emerged a new\nparadigm possessing the potential to enhance predictive capabilities beyond\nwhat classical machine learning models can achieve. In the present work we\npursue a heuristic approach to explore the potential of QML, and focus on a\nspecific transport issue. In particular, as a case study we investigate a\ntraffic forecast task for a major urban area in Athens (Greece), for which we\npossess high-resolution data. In this endeavor we explore the application of\nQuantum Neural Networks (QNN), and, notably, we present the first application\nof quantum data re-uploading in the context of transport forecasting. This\ntechnique allows quantum models to better capture complex patterns, such as\ntraffic dynamics, by repeatedly encoding classical data into a quantum state.\nAside from providing a prediction model, we spend considerable effort in\ncomparing the performance of our hybrid quantum-classical neural networks with\nclassical deep learning approaches. Our results show that hybrid models achieve\ncompetitive accuracy with state-of-the-art classical methods, especially when\nthe number of qubits and re-uploading blocks is increased. While the classical\nmodels demonstrate lower computational demands, we provide evidence that\nincreasing the complexity of the quantum model improves predictive accuracy.\nThese findings indicate that QML techniques, and specifically the data\nre-uploading approach, hold promise for advancing traffic forecasting models\nand could be instrumental in addressing challenges inherent in ITS\nenvironments.", + "arxiv_id": "2501.12776v1" + }, + { + "title": "Machine Learning Modeling for Multi-order Human Visual Motion Processing", + "abstract": "Our research aims to develop machines that learn to perceive visual motion as\ndo humans. While recent advances in computer vision (CV) have enabled DNN-based\nmodels to accurately estimate optical flow in naturalistic images, a\nsignificant disparity remains between CV models and the biological visual\nsystem in both architecture and behavior. This disparity includes humans'\nability to perceive the motion of higher-order image features (second-order\nmotion), which many CV models fail to capture because of their reliance on the\nintensity conservation law. Our model architecture mimics the cortical V1-MT\nmotion processing pathway, utilizing a trainable motion energy sensor bank and\na recurrent graph network. Supervised learning employing diverse naturalistic\nvideos allows the model to replicate psychophysical and physiological findings\nabout first-order (luminance-based) motion perception. For second-order motion,\ninspired by neuroscientific findings, the model includes an additional sensing\npathway with nonlinear preprocessing before motion energy sensing, implemented\nusing a simple multilayer 3D CNN block. When exploring how the brain acquired\nthe ability to perceive second-order motion in natural environments, in which\npure second-order signals are rare, we hypothesized that second-order\nmechanisms were critical when estimating robust object motion amidst optical\nfluctuations, such as highlights on glossy surfaces. We trained our\ndual-pathway model on novel motion datasets with varying material properties of\nmoving objects. We found that training to estimate object motion from\nnon-Lambertian materials naturally endowed the model with the capacity to\nperceive second-order motion, as can humans. The resulting model effectively\naligns with biological systems while generalizing to both first- and\nsecond-order motion phenomena in natural scenes.", + "arxiv_id": "2501.12810v1" + }, + { + "title": "Open or Closed LLM for Lesser-Resourced Languages? Lessons from Greek", + "abstract": "Natural Language Processing (NLP) for lesser-resourced languages faces\npersistent challenges, including limited datasets, inherited biases from\nhigh-resource languages, and the need for domain-specific solutions. This study\naddresses these gaps for Modern Greek through three key contributions. First,\nwe evaluate the performance of open-source (Llama-70b) and closed-source\n(GPT-4o mini) large language models (LLMs) on seven core NLP tasks with dataset\navailability, revealing task-specific strengths, weaknesses, and parity in\ntheir performance. Second, we expand the scope of Greek NLP by reframing\nAuthorship Attribution as a tool to assess potential data usage by LLMs in\npre-training, with high 0-shot accuracy suggesting ethical implications for\ndata provenance. Third, we showcase a legal NLP case study, where a Summarize,\nTranslate, and Embed (STE) methodology outperforms the traditional TF-IDF\napproach for clustering \\emph{long} legal texts. Together, these contributions\nprovide a roadmap to advance NLP in lesser-resourced languages, bridging gaps\nin model evaluation, task innovation, and real-world impact.", + "arxiv_id": "2501.12826v1" + }, + { + "title": "Mutation-Guided LLM-based Test Generation at Meta", + "abstract": "This paper describes Meta's ACH system for mutation-guided LLM-based test\ngeneration. ACH generates relatively few mutants (aka simulated faults),\ncompared to traditional mutation testing. Instead, it focuses on generating\ncurrently undetected faults that are specific to an issue of concern. From\nthese currently uncaught faults, ACH generates tests that can catch them,\nthereby `killing' the mutants and consequently hardening the platform against\nregressions. We use privacy concerns to illustrate our approach, but ACH can\nharden code against {\\em any} type of regression. In total, ACH was applied to\n10,795 Android Kotlin classes in 7 software platforms deployed by Meta, from\nwhich it generated 9,095 mutants and 571 privacy-hardening test cases. ACH also\ndeploys an LLM-based equivalent mutant detection agent that achieves a\nprecision of 0.79 and a recall of 0.47 (rising to 0.95 and 0.96 with simple\npre-processing). ACH was used by Messenger and WhatsApp test-a-thons where\nengineers accepted 73% of its tests, judging 36% to privacy relevant. We\nconclude that ACH hardens code against specific concerns and that, even when\nits tests do not directly tackle the specific concern, engineers find them\nuseful for their other benefits.", + "arxiv_id": "2501.12862v1" + }, + { + "title": "PreciseCam: Precise Camera Control for Text-to-Image Generation", + "abstract": "Images as an artistic medium often rely on specific camera angles and lens\ndistortions to convey ideas or emotions; however, such precise control is\nmissing in current text-to-image models. We propose an efficient and general\nsolution that allows precise control over the camera when generating both\nphotographic and artistic images. Unlike prior methods that rely on predefined\nshots, we rely solely on four simple extrinsic and intrinsic camera parameters,\nremoving the need for pre-existing geometry, reference 3D objects, and\nmulti-view data. We also present a novel dataset with more than 57,000 images,\nalong with their text prompts and ground-truth camera parameters. Our\nevaluation shows precise camera control in text-to-image generation, surpassing\ntraditional prompt engineering approaches. Our data, model, and code are\npublicly available at https://graphics.unizar.es/projects/PreciseCam2024.", + "arxiv_id": "2501.12910v1" + }, + { + "title": "DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via\n Reinforcement Learning", + "abstract": "We introduce our first-generation reasoning models, DeepSeek-R1-Zero and\nDeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement\nlearning (RL) without supervised fine-tuning (SFT) as a preliminary step,\ndemonstrates remarkable reasoning capabilities. Through RL, DeepSeek-R1-Zero\nnaturally emerges with numerous powerful and intriguing reasoning behaviors.\nHowever, it encounters challenges such as poor readability, and language\nmixing. To address these issues and further enhance reasoning performance, we\nintroduce DeepSeek-R1, which incorporates multi-stage training and cold-start\ndata before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1-1217\non reasoning tasks. To support the research community, we open-source\nDeepSeek-R1-Zero, DeepSeek-R1, and six dense models (1.5B, 7B, 8B, 14B, 32B,\n70B) distilled from DeepSeek-R1 based on Qwen and Llama.", + "arxiv_id": "2501.12948v1" + }, + { + "title": "GANQ: GPU-Adaptive Non-Uniform Quantization for Large Language Models", + "abstract": "Large Language Models (LLMs) face significant deployment challenges due to\ntheir substantial resource requirements. While low-bit quantized weights can\nreduce memory usage and improve inference efficiency, current hardware lacks\nnative support for mixed-precision General Matrix Multiplication (mpGEMM),\nresulting in inefficient dequantization-based implementations. Moreover,\nuniform quantization methods often fail to capture weight distributions\nadequately, leading to performance degradation. We propose GANQ (GPU-Adaptive\nNon-Uniform Quantization), a layer-wise post-training non-uniform quantization\nframework optimized for hardware-efficient lookup table-based mpGEMM. GANQ\nachieves superior quantization performance by utilizing a training-free,\nGPU-adaptive optimization algorithm to efficiently reduce layer-wise\nquantization errors. Extensive experiments demonstrate GANQ's ability to reduce\nthe perplexity gap from the FP16 baseline compared to state-of-the-art methods\nfor both 3-bit and 4-bit quantization. Furthermore, when deployed on a single\nNVIDIA RTX 4090 GPU, GANQ's quantized models achieve up to 2.57$\\times$ speedup\nover the baseline, advancing memory and inference efficiency in LLM deployment.", + "arxiv_id": "2501.12956v1" + }, + { + "title": "It's complicated. The relationship of algorithmic fairness and\n non-discrimination regulations in the EU AI Act", + "abstract": "What constitutes a fair decision? This question is not only difficult for\nhumans but becomes more challenging when Artificial Intelligence (AI) models\nare used. In light of discriminatory algorithmic behaviors, the EU has recently\npassed the AI Act, which mandates specific rules for AI models, incorporating\nboth traditional legal non-discrimination regulations and machine learning\nbased algorithmic fairness concepts. This paper aims to bridge these two\ndifferent concepts in the AI Act through: First a high-level introduction of\nboth concepts targeting legal and computer science-oriented scholars, and\nsecond an in-depth analysis of the AI Act's relationship between legal\nnon-discrimination regulations and algorithmic fairness. Our analysis reveals\nthree key findings: (1.), most non-discrimination regulations target only\nhigh-risk AI systems. (2.), the regulation of high-risk systems encompasses\nboth data input requirements and output monitoring, though these regulations\nare often inconsistent and raise questions of computational feasibility. (3.)\nRegulations for General Purpose AI Models, such as Large Language Models that\nare not simultaneously classified as high-risk systems, currently lack\nspecificity compared to other regulations. Based on these findings, we\nrecommend developing more specific auditing and testing methodologies for AI\nsystems. This paper aims to serve as a foundation for future interdisciplinary\ncollaboration between legal scholars and computer science-oriented machine\nlearning researchers studying discrimination in AI systems.", + "arxiv_id": "2501.12962v1" + }, + { + "title": "Galois groups of polynomials and neurosymbolic networks", + "abstract": "This paper introduces a novel approach to understanding Galois theory, one of\nthe foundational areas of algebra, through the lens of machine learning. By\nanalyzing polynomial equations with machine learning techniques, we aim to\nstreamline the process of determining solvability by radicals and explore\nbroader applications within Galois theory. This summary encapsulates the\nbackground, methodology, potential applications, and challenges of using data\nscience in Galois theory.\n More specifically, we design a neurosymbolic network to classify Galois\ngroups and show how this is more efficient than usual neural networks. We\ndiscover some very interesting distribution of polynomials for groups not\nisomorphic to the symmetric groups and alternating groups.", + "arxiv_id": "2501.12978v1" + }, + { + "title": "FlanEC: Exploring Flan-T5 for Post-ASR Error Correction", + "abstract": "In this paper, we present an encoder-decoder model leveraging Flan-T5 for\npost-Automatic Speech Recognition (ASR) Generative Speech Error Correction\n(GenSEC), and we refer to it as FlanEC. We explore its application within the\nGenSEC framework to enhance ASR outputs by mapping n-best hypotheses into a\nsingle output sentence. By utilizing n-best lists from ASR models, we aim to\nimprove the linguistic correctness, accuracy, and grammaticality of final ASR\ntranscriptions. Specifically, we investigate whether scaling the training data\nand incorporating diverse datasets can lead to significant improvements in\npost-ASR error correction. We evaluate FlanEC using the HyPoradise dataset,\nproviding a comprehensive analysis of the model's effectiveness in this domain.\nFurthermore, we assess the proposed approach under different settings to\nevaluate model scalability and efficiency, offering valuable insights into the\npotential of instruction-tuned encoder-decoder models for this task.", + "arxiv_id": "2501.12979v1" + }, + { + "title": "Paper Quality Assessment based on Individual Wisdom Metrics from Open\n Peer Review", + "abstract": "This study proposes a data-driven framework for enhancing the accuracy and\nefficiency of scientific peer review through an open, bottom-up process that\nestimates reviewer quality. Traditional closed peer review systems, while\nessential for quality control, are often slow, costly, and subject to biases\nthat can impede scientific progress. Here, we introduce a method that evaluates\nindividual reviewer reliability by quantifying agreement with community\nconsensus scores and applying Bayesian weighting to refine paper quality\nassessments. We analyze open peer review data from two major scientific\nconferences, and demonstrate that reviewer-specific quality scores\nsignificantly improve the reliability of paper quality estimation. Perhaps\nsurprisingly, we find that reviewer quality scores are unrelated to authorship\nquality. Our model incorporates incentive structures to recognize high-quality\nreviewers and encourage broader coverage of submitted papers, thereby\nmitigating the common \"rich-get-richer\" pitfall of social media. These findings\nsuggest that open peer review, with mechanisms for estimating and incentivizing\nreviewer quality, offers a scalable and equitable alternative for scientific\npublishing, with potential to enhance the speed, fairness, and transparency of\nthe peer review process.", + "arxiv_id": "2501.13014v1" + }, + { + "title": "Optimizing Return Distributions with Distributional Dynamic Programming", + "abstract": "We introduce distributional dynamic programming (DP) methods for optimizing\nstatistical functionals of the return distribution, with standard reinforcement\nlearning as a special case. Previous distributional DP methods could optimize\nthe same class of expected utilities as classic DP. To go beyond expected\nutilities, we combine distributional DP with stock augmentation, a technique\npreviously introduced for classic DP in the context of risk-sensitive RL, where\nthe MDP state is augmented with a statistic of the rewards obtained so far\n(since the first time step). We find that a number of recently studied problems\ncan be formulated as stock-augmented return distribution optimization, and we\nshow that we can use distributional DP to solve them. We analyze distributional\nvalue and policy iteration, with bounds and a study of what objectives these\ndistributional DP methods can or cannot optimize. We describe a number of\napplications outlining how to use distributional DP to solve different\nstock-augmented return distribution optimization problems, for example\nmaximizing conditional value-at-risk, and homeostatic regulation. To highlight\nthe practical potential of stock-augmented return distribution optimization and\ndistributional DP, we combine the core ideas of distributional value iteration\nwith the deep RL agent DQN, and empirically evaluate it for solving instances\nof the applications discussed.", + "arxiv_id": "2501.13028v1" + }, + { + "title": "Autonomy-of-Experts Models", + "abstract": "Mixture-of-Experts (MoE) models mostly use a router to assign tokens to\nspecific expert modules, activating only partial parameters and often\noutperforming dense models. We argue that the separation between the router's\ndecision-making and the experts' execution is a critical yet overlooked issue,\nleading to suboptimal expert selection and ineffective learning. To address\nthis, we propose Autonomy-of-Experts (AoE), a novel MoE paradigm in which\nexperts autonomously select themselves to process inputs. AoE is based on the\ninsight that an expert is aware of its own capacity to effectively process a\ntoken, an awareness reflected in the scale of its internal activations. In AoE,\nrouters are removed; instead, experts pre-compute internal activations for\ninputs and are ranked based on their activation norms. Only the top-ranking\nexperts proceed with the forward pass, while the others abort. The overhead of\npre-computing activations is reduced through a low-rank weight factorization.\nThis self-evaluating-then-partner-comparing approach ensures improved expert\nselection and effective learning. We pre-train language models having 700M up\nto 4B parameters, demonstrating that AoE outperforms traditional MoE models\nwith comparable efficiency.", + "arxiv_id": "2501.13074v1" + }, + { + "title": "Robust Representation Consistency Model via Contrastive Denoising", + "abstract": "Robustness is essential for deep neural networks, especially in\nsecurity-sensitive applications. To this end, randomized smoothing provides\ntheoretical guarantees for certifying robustness against adversarial\nperturbations. Recently, diffusion models have been successfully employed for\nrandomized smoothing to purify noise-perturbed samples before making\npredictions with a standard classifier. While these methods excel at small\nperturbation radii, they struggle with larger perturbations and incur a\nsignificant computational overhead during inference compared to classical\nmethods. To address this, we reformulate the generative modeling task along the\ndiffusion trajectories in pixel space as a discriminative task in the latent\nspace. Specifically, we use instance discrimination to achieve consistent\nrepresentations along the trajectories by aligning temporally adjacent points.\nAfter fine-tuning based on the learned representations, our model enables\nimplicit denoising-then-classification via a single prediction, substantially\nreducing inference costs. We conduct extensive experiments on various datasets\nand achieve state-of-the-art performance with minimal computation budget during\ninference. For example, our method outperforms the certified accuracy of\ndiffusion-based methods on ImageNet across all perturbation radii by 5.3% on\naverage, with up to 11.6% at larger radii, while reducing inference costs by\n85$\\times$ on average. Codes are available at:\nhttps://github.com/jiachenlei/rRCM.", + "arxiv_id": "2501.13094v1" + }, + { + "title": "Leveraging LLMs to Create a Haptic Devices' Recommendation System", + "abstract": "Haptic technology has seen significant growth, yet a lack of awareness of\nexisting haptic device design knowledge hinders development. This paper\naddresses these limitations by leveraging advancements in Large Language Models\n(LLMs) to develop a haptic agent, focusing specifically on Grounded Force\nFeedback (GFF) devices recommendation. Our approach involves automating the\ncreation of a structured haptic device database using information from research\npapers and product specifications. This database enables the recommendation of\nrelevant GFF devices based on user queries. To ensure precise and contextually\nrelevant recommendations, the system employs a dynamic retrieval method that\ncombines both conditional and semantic searches. Benchmarking against the\nestablished UEQ and existing haptic device searching tools, the proposed haptic\nrecommendation agent ranks in the top 10\\% across all UEQ categories with mean\ndifferences favoring the agent in nearly all subscales, and maintains no\nsignificant performance bias across different user groups, showcasing superior\nusability and user satisfaction.", + "arxiv_id": "2501.12573v1" + }, + { + "title": "Guaranteed Recovery of Unambiguous Clusters", + "abstract": "Clustering is often a challenging problem because of the inherent ambiguity\nin what the \"correct\" clustering should be. Even when the number of clusters\n$K$ is known, this ambiguity often still exists, particularly when there is\nvariation in density among different clusters, and clusters have multiple\nrelatively separated regions of high density. In this paper we propose an\ninformation-theoretic characterization of when a $K$-clustering is ambiguous,\nand design an algorithm that recovers the clustering whenever it is\nunambiguous. This characterization formalizes the situation when two high\ndensity regions within a cluster are separable enough that they look more like\ntwo distinct clusters than two truly distinct clusters in the clustering. The\nalgorithm first identifies $K$ partial clusters (or \"seeds\") using a\ndensity-based approach, and then adds unclustered points to the initial $K$\npartial clusters in a greedy manner to form a complete clustering. We implement\nand test a version of the algorithm that is modified to effectively handle\noverlapping clusters, and observe that it requires little parameter selection\nand displays improved performance on many datasets compared to widely used\nalgorithms for non-convex cluster recovery.", + "arxiv_id": "2501.13093v1" + } +] diff --git a/arxiv-processor/README.md b/arxiv-processor/README.md new file mode 100644 index 0000000..da8aca2 --- /dev/null +++ b/arxiv-processor/README.md @@ -0,0 +1,130 @@ +# ArXiv Processor + +A Go package for fetching and processing papers from arXiv. + +## Installation + +1. Clone the repository: +```bash +git clone https://github.com/yourusername/arxiv-processor.git +cd arxiv-processor +``` + +2. Initialize the Go module: +```bash +go mod init arxiv-processor +go mod tidy +``` + +## Usage + +### As a Library +To use the package in your Go application: + +```go +import "github.com/yourusername/arxiv-processor" + +func main() { + // Create client + client := arxiv.NewClient() + + // Define search parameters + // The format "20060102" is Go's reference time format: + // 2006 = year + // 01 = month + // 02 = day + // Note: The arXiv API returns full timestamps including time of day, + // but the search API only uses the date portion for filtering + startDate, _ := time.Parse("20060102", "20240101") + endDate, _ := time.Parse("20060102", "20240131") + + // Fetch papers + // The FetchPapers method returns all papers at once after completion + // of the API request and any necessary pagination + papers, err := client.FetchPapers("cat:cs.AI", startDate, endDate) + if err != nil { + log.Fatal(err) + } + + // Use papers directly (in-memory) + // The papers slice contains all results after completion + for _, paper := range papers { + fmt.Printf("Title: %s\n", paper.Title) + fmt.Printf("Abstract: %s\n", paper.Abstract) + } + + // Optionally save papers to file + err = arxiv.SavePapers("papers.json", papers) + if err != nil { + log.Fatal(err) + } +} +``` + +Note: The package currently writes to a file by default. To modify this behavior to only return JSON objects: +1. Remove the SavePapers call +2. Use the returned papers slice directly +3. The papers slice contains all paper data as Go structs +4. You can marshal to JSON using json.Marshal(papers) if needed + +### Command Line Interface +To use the CLI: + +```bash +go run main.go --search "cat:cs.AI" --date-range "YYYYMMDD-YYYYMMDD" +``` + +#### Command Line Options +- `--search`: Search query (e.g., "cat:cs.AI" for AI papers) +- `--date-range`: Date range in YYYYMMDD-YYYYMMDD format +- `--output`: Output file path (default: papers_data.json) + +### Example: Fetch AI Papers +```bash +go run main.go --search "cat:cs.AI" --date-range "20250115-20250118" +``` + +### Program Output +- Fetched papers are saved to `papers_data.json` +- Example JSON structure: +```json +[ + { + "title": "Sample Paper Title", + "abstract": "This is a sample abstract...", + "arxiv_id": "2501.08565v1" + } +] +``` +- The JSON file contains paper metadata including: + - Title + - Abstract + - arXiv ID + +## Configuration + +### Environment Variables +- `ARXIV_MAX_RESULTS`: Maximum number of results to fetch (default: 100) +- `ARXIV_START_INDEX`: Start index for pagination (default: 0) + +## Package Structure + +``` +arxiv-processor/ +├── arxiv/ # arXiv API client +├── storage/ # Data storage handlers +├── llm/ # LLM integration (TODO) +├── main.go # Main entry point +└── README.md # This file +``` + +## Contributing + +1. Fork the repository +2. Create a new branch +3. Make your changes +4. Submit a pull request + +## License + +MIT License diff --git a/arxiv-processor/arxiv-2501.11599v1.md b/arxiv-processor/arxiv-2501.11599v1.md new file mode 100644 index 0000000..ca420ab --- /dev/null +++ b/arxiv-processor/arxiv-2501.11599v1.md @@ -0,0 +1,156 @@ +Title: SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks + +URL Source: https://arxiv.org/html/2501.11599v1 + +Markdown Content: +Wentao Wan1, Zhuojie Yang1, Yongcan Chen3, Chenglin Luo1, Ruilin Wang1, Kehao Cai1, Nan Kang1, Liang Lin1, Keze Wang1,2 + +###### Abstract + +Deductive reasoning is a crucial logical capability that assists us in solving complex problems based on existing knowledge. Although augmented by Chain-of-Thought prompts, Large Language Models (LLMs) might not follow the correct reasoning paths. Enhancing the deductive reasoning abilities of LLMs, and leveraging their extensive built-in knowledge for various reasoning tasks, remains an open question. Attempting to mimic the human deductive reasoning paradigm, we propose a multi-stage Syllogistic-Reasoning Framework of Thought (SR-FoT) that enables LLMs to perform syllogistic deductive reasoning to handle complex knowledge-based reasoning tasks. Our SR-FoT begins by interpreting the question and then uses the interpretation and the original question to propose a suitable major premise. It proceeds by generating and answering minor premise questions in two stages to match the minor premises. Finally, it guides LLMs to use the previously generated major and minor premises to perform syllogistic deductive reasoning to derive the answer to the original question. Extensive and thorough experiments on knowledge-based reasoning tasks have demonstrated the effectiveness and advantages of our SR-FoT. + +Code — https://github.com/RodeWayne/SR-FoT + +Introduction +------------ + +Deductive reasoning (Johnson-Laird [1999](https://arxiv.org/html/2501.11599v1#bib.bib19)) is the process of drawing valid inferences. Deductive reasoning is a powerful human capability, where rigorous deductive reasoning helps us use existing knowledge as premises to derive correct subsequent conclusions, enabling us to tackle various complex tasks in the real world. + +Automated deductive reasoning has long been a pursuit in the field of Natural Language Processing (NLP) (Chowdhary and Chowdhary [2020](https://arxiv.org/html/2501.11599v1#bib.bib9); Bharadiya [2023](https://arxiv.org/html/2501.11599v1#bib.bib3); Khurana et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib21)). Works on automated rigorous reasoning include reasoning engines and Automated Theorem Proving (ATP) (Bibel [2013](https://arxiv.org/html/2501.11599v1#bib.bib4)), which often provide methods for automatically checking the rigor of reasoning. However, these engines require the use of formal languages, which limits their applicability in knowledge-based reasoning scenarios. Because formal language-based reasoning requires a predefined library of formalized premises, many knowledge-based reasoning tasks, including common-sense question answering, involve a diverse array of premises. It is difficult to prepare and rigorously formalize such a large library of premises in advance. Therefore, performing correct deductive reasoning in natural language holds significant importance. + +Large Language Models (LLMs) (Chang et al. [2024](https://arxiv.org/html/2501.11599v1#bib.bib7); Floridi and Chiriatti [2020](https://arxiv.org/html/2501.11599v1#bib.bib15); Touvron et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib29); Chiang et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib8); Huang and Chang [2022](https://arxiv.org/html/2501.11599v1#bib.bib18); DeepSeek-AI [2024](https://arxiv.org/html/2501.11599v1#bib.bib11)) pre-trained on extensive corpora possess inherent soft deductive reasoning capabilities (Seals and Shalin [2023](https://arxiv.org/html/2501.11599v1#bib.bib27)). With the aid of the Chain-of-Thought prompt (CoT) (Lyu et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib25); Wei et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib33); Zhang et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib39); Turpin et al. [2024](https://arxiv.org/html/2501.11599v1#bib.bib30); Lee et al. [2024](https://arxiv.org/html/2501.11599v1#bib.bib22); Liu, Pang, and Fan [2023](https://arxiv.org/html/2501.11599v1#bib.bib23)), the cognitive abilities of LLMs are further enhanced. However, reasoning with CoT often does not constitute strict deductive reasoning and thus can lack rigor. Fig. [4](https://arxiv.org/html/2501.11599v1#Sx3.F4 "Figure 4 ‣ Procedure of Our SR-FoT ‣ Methodology ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks") illustrates the different processes of handling the same problem using CoT and classic syllogistic deductive reasoning, clearly showing that the syllogistic deductive approach is more rigorous. We believe that guiding large language models to perform deductive reasoning, rather than merely multi-step reasoning, can enhance the rigor of the reasoning process, reduce illusions, and subsequently improve performance on complex tasks. + +Inspired by the most fundamental human deductive reasoning paradigm, syllogistic reasoning (Bucciarelli and Johnson-Laird [1999](https://arxiv.org/html/2501.11599v1#bib.bib6); Khemlani and Johnson-Laird [2012](https://arxiv.org/html/2501.11599v1#bib.bib20); Bara, Bucciarelli, and Johnson-Laird [1995](https://arxiv.org/html/2501.11599v1#bib.bib2)), we propose a multi-stage reasoning framework for large language models to guide them in using syllogistic reasoning to solve specific problems. In contrast to existing works in the community (Wu et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib34); Ye et al. [2023b](https://arxiv.org/html/2501.11599v1#bib.bib37); Deng et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib12)), we do not solely rely on simplistic processes or create targeted benchmarks to evaluate the LLMs’ capabilities in performing syllogistic reasoning. Instead, we propose a reasoning framework based on the syllogistic thinking paradigm to handle complex knowledge-based reasoning tasks. Our SR-FoT advances in guiding LLMs in performing syllogistic deductive reasoning, thereby achieving improved performance on these tasks and enhancing the rigor and reliability of the reasoning process. + +Our SR-FoT consists of five stages. Initially, it involves interpreting the question. Subsequently, SR-FoT guides the Large Language Model (LLM) proposing a major premise suitable for the question. This major premise typically encompasses the built-in knowledge of the LLM, which serves as a universal rule that aids in addressing the original question. The next stage involves obtaining a minor premise, which acts as the bridge linking the major premise to the original problem and is crucial for applying the major premise to the current issue. We first let the LLM formulate minor premise questions based on the original question, major premise, and contextual information, and then answer these to obtain an appropriate minor premise. Finally, with both the major and minor premises established, we enable the LLM to perform syllogistic reasoning based on the original question and these premises to derive the answer to the original question. Furthermore, to minimize the interference caused by excessive information during the reasoning process (Dong et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib14)), we restrict each stage to only access the content from its necessary preceding stages. For example, during the final syllogistic reasoning, only the original problem and the previously established major and minor premises are visible, without the need to reference the problem interpretation and minor premise question stages. + +Our main contributions can be summarized in three points: i) We propose a multi-step thinking framework that guides LLMs in using syllogistic deductive reasoning to solve knowledge-based reasoning tasks. Specifically, to enhance the ability of LLM to leverage its built-in knowledge for solving diverse tasks, we introduce a problem interpretation stage when acquiring the major premise and improve the quality of both premises as well as their logical connection to the original problem by adopting an autonomous question-answering approach during the acquisition of the minor premise (Bubeck et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib5)); ii) To facilitate more rigorous reasoning, we have designed our thinking framework so that each step only accesses the information necessary for that stage, thereby reducing the illusions and error accumulation that can come from overly long reasoning steps; iii) Our SR-FoT achieves superior performance over the existing chain of thought-related methods on various knowledge-based reasoning QA datasets such as ScienceQA (Lu et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib24)), StrategyQA (Geva et al. [2021](https://arxiv.org/html/2501.11599v1#bib.bib17)), and BoolQ (Clark et al. [2019](https://arxiv.org/html/2501.11599v1#bib.bib10)), demonstrating the superiority of our SR-FoT. + +Related Work +------------ + +### Chain-of-Thought + +Chain-of-Thought (Wei et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib33)) has been demonstrated to enhance performance in reasoning tasks by fully utilizing the in-context learning capability of the large language model to stimulate its multi-step reasoning ability. Self-consistency CoT (SC-CoT) (Wang et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib32)) further improves the performance of CoT by utilizing the consistency of multiple sampled reasoning chains. Complexity-based CoT (C-CoT) (Fu et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib16)) further discovers that the consistency of complex reasoning chains is even more vital for the reasoning performance of language models. In addition, some efforts have also been made to further stimulate the reasoning ability of language models by focusing on the structure of the reasoning chain and the levels of reasoning, such as Least-to-Most (Zhou et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib40)) and Tree-of-Thought (Yao et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib35)). However, these works have not considered how to stimulate the reasoning abilities of LLMs from the perspective of logical reasoning. + +### Logical Reasoning Ability of LLMs + +There has been considerable research within the community on the logical reasoning capabilities of LLMs, broadly categorized into two directions: one focuses on logic reasoning based on formal languages, and the other on natural language logic reasoning. Research related to formal language-based logic reasoning primarily concentrates on the field of Automated Theorem Proving (ATP) (Bibel [2013](https://arxiv.org/html/2501.11599v1#bib.bib4)), utilizing the built-in mathematical priors of LLMs to accelerate the search process in theorem proving or to construct a growing library of mathematical theorems to aid new proofs (Wang et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib31)). This work typically operates within interactive theorem-proving platforms like the Lean system, which restricts its application in daily question-answering scenarios. Logic reasoning on natural language with LLMs generally involves soft reasoning (Yu et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib38)), which does not provide rigorous guarantees. For instance, the Chain-of-Thought (CoT) enhances the general explicit reasoning abilities of LLMs, and there are exploratory studies demonstrating to what extent LLMs can perform in logical reasoning, or how segment checking might reduce soft deductive reasoning illusions and error accumulation (Ye et al. [2023a](https://arxiv.org/html/2501.11599v1#bib.bib36); Dhuliawala et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib13)). Recently, several studies on syllogistic reasoning with LLMs have been proposed. However, these primarily create benchmarks (Ye et al. [2023b](https://arxiv.org/html/2501.11599v1#bib.bib37)), evaluating the capability of LLMs to perform syllogistic reasoning on datasets with given premises. Unlike previous works, our research investigates how to guide LLMs through a multi-stage process that involves autonomously generating minor and major premises and performing syllogistic deductive reasoning to answer a variety of knowledge-based reasoning questions. + +Methodology +----------- + +We have designed a reasoning framework that guides large language models to perform syllogistic deductive reasoning for addressing various knowledge-based reasoning question-answer tasks. Next, we present syllogistic reasoning as background knowledge, followed by a detailed description of our SR-FoT framework. + +### Background: Syllogism + +In traditional logic, syllogism (Smiley [1973](https://arxiv.org/html/2501.11599v1#bib.bib28)) is a form of reasoning where one proposition (the conclusion) necessarily follows from two other propositions (known as premises). As shown in Fig. [1](https://arxiv.org/html/2501.11599v1#Sx3.F1 "Figure 1 ‣ Background: Syllogism ‣ Methodology ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"), a syllogism consists of three parts: a major premise, a minor premise, and a conclusion. Logically, the conclusion is derived by applying the major premise to the minor premise. The major premise represents a general principle, while the minor premise is a specific statement. Syllogistic reasoning is a type of deductive reasoning; rigorous deductive reasoning ensures that if the premises are correct, the conclusion must also be correct. + +![Image 12: Refer to caption](https://arxiv.org/html/2501.11599v1/x1.png) + +Figure 1: A syllogism example. + +### Procedure of Our SR-FoT + +![Image 13: Refer to caption](https://arxiv.org/html/2501.11599v1/x2.png) + +Figure 2: Procedure of our SR-FoT. Questionori: Original Question, Context: Context provided for Original Question, Answerori: Answer for Original Question, QuestionmP: Question for Minor Premise, PropmtCoT: Guide Prompt for CoT, PropmtQE: Guide Prompt for Question Explanation, PropmtMP: Guide Prompt for Major Premise Production, PropmtQmP: Guide Prompt for Posing the Minor Premise Question, PromptmP: Guide Prompt for Minor Premise Production, PromptSR: Guide Prompt for Final Syllogistic Reasoning and so on. + +![Image 14: Refer to caption](https://arxiv.org/html/2501.11599v1/x3.png) + +Figure 3: Prompts for each stage of our SR-FoT. + +While our proposed SR-FoT does not guarantee the execution of rigorous syllogistic reasoning for 100%, it aims to guide the reasoning paradigms of the LLM through carefully designed prompts and sub-tasks at each stage. By strategically controlling the input visible at each stage, we strive to ensure that the LLM conducts rigorous syllogistic reasoning and minimizes the occurrence of reasoning fallacies. Specifically, As shown in Fig. [2](https://arxiv.org/html/2501.11599v1#Sx3.F2 "Figure 2 ‣ Procedure of Our SR-FoT ‣ Methodology ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"), our SR-FoT is divided into five stages. The prompts for each stage of our SR-FoT are shown in Fig. [3](https://arxiv.org/html/2501.11599v1#Sx3.F3 "Figure 3 ‣ Procedure of Our SR-FoT ‣ Methodology ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"). + +![Image 15: Refer to caption](https://arxiv.org/html/2501.11599v1/x4.png) + +Figure 4: A case of using CoT and SR-FoT to answer a question in the ScienceQA dataset respectively. The highlighted red parts indicate the incorrect or misleading content, while the highlighted green parts indicate the content that helps correct reasoning. + +Stage 1: Question Explanation. The key to utilizing syllogistic reasoning to solve various complex knowledge-based reasoning tasks lies in formulating appropriate major and minor premises that fit the current problem. Accordingly, the first stage of our SR-FoT involves using a prompt with examples to guide the LLM in interpreting the original task question and proposing a solution approach. This guidance helps direct the LLM to formulate suitable major premises and then appropriate minor ones. In this stage, besides the guidance and example prompts, the only information available to the LLM is the ”original question” and the ”context” provided by the task, which also includes ”options” information for multiple-choice questions. + +Stage 2: Major Premise Production. After acquiring the “question explanation”, we gain a deeper understanding of the original question and develop an approach to solve it. This solution approach often includes guidance on what further information is needed. In this stage, based on these guidelines, we propose an appropriate major premise, which is derived from the task ”context” or the built-in knowledge of the LLM. In Stage 2, besides the guidance and example prompts, the information accessible to the LLM includes the ”original question” ”context” and the ”question explanation” generated in the first stage. + +Stage 3: Posing the Minor Premise Question. After establishing the major premise, to effectively engage in syllogistic reasoning, we need a minor premise. In syllogistic reasoning, the minor premise is a specific statement that describes the relationship between a particular instance and the category mentioned in the major premise. Through the minor premise, the universal characteristics of the major premise can be applied to the specific instance in the minor premise, which is a crucial step in using syllogistic reasoning to solve specific problems. Given the diverse and often complex nature of the knowledge-based reasoning tasks we need to address, it is challenging to provide a matched and correct minor premise in one step. Our SR-FoT divides the step of proposing a minor premise into two stages: posing the minor premise question (Stage 3) and answering the minor premise question (Stage 4). The task of the ”posing the minor premise question” stage is to determine what information about the specific instance in the original question the LLM should acquire to utilize the major premise in answering the ”original question”. Therefore, in the ”posing the minor premise question” stage, besides the guidance and example prompts, the LLM needs access to the ”original question,” ”context,” and the ”major premise” generated in Stage 2. + +Stage 4: Minor Premise Production. The task of Stage 4 is to utilize the ”context” information provided by the original task, along with the built-in knowledge of the LLM, to answer the minor premise question posed in Stage 3. This stage aims to obtain the correct information about a specific aspect of the particular instance in the original question, leading to the formulation of an accurate and matching minor premise. Given the potential complexity of the minor premise questions, we guide LLM to employ the Chain-of-Thought (CoT) technique to answer the minor premise question and to organize and obtain the minor premise. Furthermore, to avoid the interference caused by excessive information, in this stage, besides the guidance and example prompts, the LLM has access only to the ”minor premise question” and ”context” without needing to see the ”original question” again. The ”minor premise question” and ”context” already contain all the information necessary for the task of the LLM at this stage; viewing additional information like the ”original question” could instead lead to distractions and affect performance. + +Stage 5: Final Syllogistic Reasoning After the aforementioned stages, complex original knowledge-based reasoning questions can now be answered using syllogistic reasoning. The specific approach involves designing the appropriate task instruction and example prompts, allowing the LLM to engage in syllogistic reasoning based on the major and minor premises generated in earlier stages, to arrive at the answer to the original question. Therefore, in Stage 5, we design the LLM to have access, in addition to the guidance and example prompts, to the ”major premise” generated in Stage 2, the ”minor premise” generated in Stage 4, and the ”original question”. + +Table 1: Comparison with the state-of-the-art methods on the ScienceQA, StategyQA, and BoolQ datasets. + +Experiments +----------- + +To evaluate the effectiveness of our SR-FoT, we conducted a series of experiments using both Open-source and closed-source LLMs on several common knowledge-based reasoning question-answer datasets. + +### Experiment Setup + +#### Datasets. + +To fully demonstrate the effectiveness and generalization of our SR-FoT, we conduct a series of experiments on three datasets from different fields. + +ScienceQA (Lu et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib24)) is a scientific question-answering dataset and contains 21,208 multimodal multiple-choice science questions. It can be divided into three subjects: natural science, language science, and social science. It requires the language model to select one answer from multiple options, usually requiring multi-step reasoning. In our experiment, we employ the test set samples which only have a text context, with a total of 2224. We report the accuracy of our SR-FoT and comparison methods on this set. + +StrategyQA (Geva et al. [2021](https://arxiv.org/html/2501.11599v1#bib.bib17)) is a question-answering dataset focusing on open-domain questions. Its questions contain multiple reasoning steps, and a strategy should be used to obtain the answers. In our experiment, we evaluate the methods with accuracy on the train set, which includes 2290 samples. + +BoolQ (Clark et al. [2019](https://arxiv.org/html/2501.11599v1#bib.bib10)) is a reading comprehension dataset consisting of 16k samples. They often query for complex, non-factoid information, and require difficult entailment-like inference to solve. In our experiment, we compare the accuracy of our SR-FoT with other methods on the dev set, with a total of 3270. + +#### Experimental Setting. + +Our experiments are performed using API calls on the proprietary model GPT-3.5-turbo (Ouyang et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib26)), the open-source model DeepSeek-v2 (DeepSeek-AI [2024](https://arxiv.org/html/2501.11599v1#bib.bib11)) with 236B parameters, and Qwen1.5-32B-Chat (Bai et al. [2023](https://arxiv.org/html/2501.11599v1#bib.bib1)) version with 32B parameters. The control group methods include the Base method, Chain of Thought (CoT) (Wei et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib33)), Self-consistency CoT (SC-CoT) (Wang et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib32)), and Complexity-based CoT (C-CoT) (Fu et al. [2022](https://arxiv.org/html/2501.11599v1#bib.bib16)) methods. Our own approaches included SR-FoT and Self-consistency SR-FoT (SC-SR-FoT), which represents our SR-FoT following the self-consistency sampling and aggregation settings of SC-CoT. In the single-round sampling methods which include Base, CoT and SR-FoT, the hyperparameters on GPT-3.5-turbo and Qwen1.5-32B-Chat are set to top\_p=0.3 and temperature=0.2, while the temperature on DeepSeek-v2 was set to the default recommended value of 1 (DeepSeek only allows the temperature hyperparameter to be adjusted). In the multi-round sampling methods which include SC-CoT, C-CoT and SC-SR-FoT, we perform 10 samplings each. To enhance the diversity of sampling outcomes, the hyperparameters on GPT-3.5-turbo and Qwen1.5-32B-Chat for top\_p and temperature are adjusted to 0.7 and 0.9 respectively. The temperature hyperparameter on DeepSeek remained at the default recommended value of 1. The number of in-context example prompts used in all the methods on the ScienceQA, StrategyQA, and BoolQ datasets are 5, 2, and 2, respectively. + +### Experimental Results and Analyses + +#### Performance on ScienceQA. + +Scientific question answering is a task scenario that often requires deductive reasoning. As seen in Tab. [1](https://arxiv.org/html/2501.11599v1#Sx3.T1 "Table 1 ‣ Procedure of Our SR-FoT ‣ Methodology ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"), under GPT-3.5-turbo, in the comparison of single-round sampling methods, our SR-FoT outperforms the Base and CoT methods by 1.5% and 0.5% respectively and is on par with multi-round sampling methods like SC-CoT and C-CoT. In the comparison of multi-round sampling methods, our SC-SR-FoT exceeds SC-CoT and C-CoT methods by 1.5%. Under the open-source model DeepSeek-V2, SR-FoT outperforms the Base and CoT by 4.8% and 3.2% respectively, even surpassing multi-round sampling methods. What’s more, our SC-SR-FoT further increases the accuracy to 93.0%. Under Qwen1.5-32B-Chat, compared to the Base and CoT methods, our SR-FoT has an improvement of 1.3% and 1.6% respectively. Compared to SC-CoT and C-CoT, our SC-SR-FoT also performs better, surpassing them by 2.8% and 3.2% respectively. These indicate that our methods achieves greater superiority on the ScienceQA dataset under multiple models. + +#### Performance on StrategyQA and BoolQ. + +StrategyQA and BoolQ are two other knowledge-based reasoning question-answer datasets that require a true or false judgment based on context or common sense knowledge. From Table 1, for StrategyQA under GPT-3.5-turbo, in the comparison of single-round sampling methods, our SR-FoT outperforms Base and CoT by 7.3% and 0.8% respectively; in the comparison of multi-round sampling methods, our SC-SR-FoT exceeds SC-CoT and C-CoT by 1.9% and 3.2% respectively. Similar trends are observed under DeepSeek-V2 and Qwen1.5-32B-Chat. In addition, our SR-FoT and SC-SR-FoT also perform the best in both single-round sampling methods and multi-round sampling methods on BoolQ under the three models. + +Overall, whether under the closed-source large language model GPT-3.5-turbo or the open-source large language models DeepSeek-V2 and Qwen1.5-32B-Chat, our SR-FoT achieve a superior accuracy compared with other compared methods on the ScienceQA, StrategyQA, and BoolQ datasets. This demonstrates the effectiveness of our SR-FoT. + +It is worth noting that under DeepSeek-V2 and Qwen1.5-32B-Chat, the Base method achieves relatively high results across all three datasets, while the benefits of the CoT method show signs of saturation, and at times perform worse than the Base method. However, our methods, whether under single-round sampling settings (SR-FoT) or multi-round aggregated sampling settings (SC-SR-FoT), are still able to further enhance performance, demonstrating greater potential for performance gains. We believe this is because our SR-FoT employs a syllogistic deductive reasoning framework, allowing LLMs to address these knowledge-based reasoning tasks based on a more rigorous reasoning process, thereby achieving better overall performance. + +Table 2: Effectiveness comparisons for subcategories on the ScienceQA dataset with DeepSeek-v2. + +### Ablation Study + +#### Effectiveness Comparisons for Subcategories. + +As shown in table [2](https://arxiv.org/html/2501.11599v1#Sx4.T2 "Table 2 ‣ Performance on StrategyQA and BoolQ. ‣ Experimental Results and Analyses ‣ Experiments ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"), we conduct the experiments on the ScienceQA dataset with DeepSeek-v2. The results demonstrate that our methods can enhance the reasoning performance of the language model across questions of different difficulty levels and various subjects, compared with the state-of-the-art methods. When increasing the consistency of the proposed method, the beneficial effects become more significant. + +#### Ablation of Stages. + +As shown in Table [3](https://arxiv.org/html/2501.11599v1#Sx4.T3 "Table 3 ‣ Ablation of Stages. ‣ Ablation Study ‣ Experiments ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"), we conduct experiments on ScienceQA under DeepSeek-V2 to verify the effectiveness of each stage in our method. Specifically, ‘all in one stage’ denotes using instructions and examples to let the LLM provide the premises based on the question and options, and then directly provide the answers. ‘w/o stage 3’ denotes providing the minor premise directly, instead of posing it as a question first and then answering. The results demonstrate that the completeness of each stage is important. In detail, discarding either the problem explanation or the major and minor premises would decrease the performance. Furthermore, allowing the language model to directly provide the major and minor premise would significantly reduce its performance, demonstrating the necessity of the multi-stage thinking framework in our SR-FoT. + +Table 3: Ablation study of stages in our proposed methods on the ScienceQA dataset under DeepSeek-V2. + +Table 4: Ablation study of visible information in various stages on the StrategyQA dataset. + +Table 5: Fifty cases using CoT and SR-FoT respectively, randomly selected from the three datasets under GPT-3.5-turbo, are analyzed to assess their rigor, and their rigor rates are subsequently calculated. + +Table 6: Error sources on a random sample of 50 incorrect examples of our SR-FoT from the three datasets using the GPT-3.5-turbo model. ‘MaPE’ denotes major premise error, ‘MiPQE’ denotes minor premise question error, ‘MiPE’ denotes minor premise error and ‘FRPE’ denotes final reasoning process error. + +#### Impact of Visible Information in Various Stages. + +As shown in Table [4](https://arxiv.org/html/2501.11599v1#Sx4.T4 "Table 4 ‣ Ablation of Stages. ‣ Ablation Study ‣ Experiments ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"), we conduct the experiments on the StrategyQA dataset with DeepSeek-v2. ‘w/o context in stage 3’ denotes that the minor premise question is generated without considering the context. ‘add Qori in stage 4’ denotes that providing the original question, minor premise question, and context all to the LLM during the process of answering the minor premise question. The results demonstrate that both decreasing or increasing the content of the input prompts adversely affect performance. This underlines the appropriateness of the designed visible information at each stage of our SR-FoT. + +### Rigor Analysis + +To more directly analyze whether our SR-FoT improves the rigor of the reasoning process compared to CoT, we randomly select 50 cases from each of the three datasets under GPT-3.5-turbo for manual evaluation. For CoT and SR-FoT, if all intermediate steps from the first step of reasoning to obtaining the final answer are correct and logically progressive, without any factual inconsistencies or self-inconsistencies, we call them rigorous; otherwise, they are not rigorous. The results are in Table [5](https://arxiv.org/html/2501.11599v1#Sx4.T5 "Table 5 ‣ Ablation of Stages. ‣ Ablation Study ‣ Experiments ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"). From the table, it can be found that our SR-FoT has a higher rigor rate than CoT on all three datasets, indicating that our SR-FoT has enhanced the rigor and interoperability of LLM reasoning. For specific comparison cases about rigor, please refer to the supplementary materials. + +### Error Analysis + +We randomly selected 50 error cases from each of the three datasets under GPT-3.5-turbo to perform an error analysis of our SR-FoT. The sources of errors and their respective proportions are as in table [6](https://arxiv.org/html/2501.11599v1#Sx4.T6 "Table 6 ‣ Ablation of Stages. ‣ Ablation Study ‣ Experiments ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"). From the error analysis, it can be found that the proportion of different types of errors varies on different datasets. In ScienceQA, most errors stem from the step of extracting suitable minor premises from the question information. In StrategyQA, the main errors stem from the final reasoning process and mistakes in presenting the major premise. In BoolQ, the primary errors originate from the final reasoning process and errors in formulating the minor premise. + +### Case Study + +We give a case on the ScienceQA dataset to show how CoT and SR-FoT work. In Fig. [4](https://arxiv.org/html/2501.11599v1#Sx3.F4 "Figure 4 ‣ Procedure of Our SR-FoT ‣ Methodology ‣ SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language Models Tackling Knowledge-based Reasoning Tasks"), it can be seen that in the fourth and fifth reasoning steps of CoT, the model misunderstands the rhyme condition and thus infers wrong information, resulting in an incorrect final answer. In SR-FoT, the question explanation points out a reasonable direction for the major premise, then the major premise gives more sufficient rhyme conditions, and the minor premise correctly distinguishes different ending sounds. With their joint help, the model gives the correct final answer. More cases can be found in the supplementary materials. + +Conclusion +---------- + +In this paper, we have developed a multi-stage syllogistic reasoning framework of thought(SR-FoT) to guide LLMs in solving complex and diverse knowledge-based reasoning question-answering tasks using syllogistic deductive reasoning. Experiments across various knowledge-based reasoning datasets under various LLMs demonstrate the effectiveness and advantages of our method. + diff --git a/arxiv-processor/arxiv/client.go b/arxiv-processor/arxiv/client.go new file mode 100644 index 0000000..a25f08b --- /dev/null +++ b/arxiv-processor/arxiv/client.go @@ -0,0 +1,78 @@ +package arxiv + +import ( + "context" + "encoding/xml" + "fmt" + "io" + "net/http" + "net/url" + "time" +) + +// Client represents an arXiv API client +type Client struct { + baseURL string + httpClient *http.Client +} + +// NewClient creates a new arXiv API client +func NewClient() *Client { + return &Client{ + baseURL: "http://export.arxiv.org/api/query", + httpClient: &http.Client{Timeout: 30 * time.Second}, + } +} + +// Query represents search parameters for arXiv API +type Query struct { + Category string + DateRange string + MaxResults int + StartOffset int +} + +// Paper represents a single arXiv paper +type Paper struct { + ID string `xml:"id"` + Title string `xml:"title"` + Summary string `xml:"summary"` + Published time.Time `xml:"published"` + Updated time.Time `xml:"updated"` + Authors []Author `xml:"author"` +} + +// Author represents a paper author +type Author struct { + Name string `xml:"name"` +} + +// FetchPapers retrieves papers from arXiv API +func (c *Client) FetchPapers(ctx context.Context, query Query) ([]Paper, error) { + params := url.Values{} + params.Add("search_query", fmt.Sprintf("%s AND submittedDate:[%s]", query.Category, query.DateRange)) + params.Add("max_results", fmt.Sprintf("%d", query.MaxResults)) + params.Add("start", fmt.Sprintf("%d", query.StartOffset)) + + resp, err := c.httpClient.Get(fmt.Sprintf("%s?%s", c.baseURL, params.Encode())) + if err != nil { + return nil, fmt.Errorf("failed to fetch papers: %w", err) + } + defer resp.Body.Close() + + if resp.StatusCode != http.StatusOK { + return nil, fmt.Errorf("unexpected status code: %d", resp.StatusCode) + } + + return parseResponse(resp.Body) +} + +func parseResponse(r io.Reader) ([]Paper, error) { + var feed struct { + Entries []Paper `xml:"entry"` + } + if err := xml.NewDecoder(r).Decode(&feed); err != nil { + return nil, fmt.Errorf("failed to parse response: %w", err) + } + return feed.Entries, nil +} diff --git a/arxiv-processor/go.mod b/arxiv-processor/go.mod new file mode 100644 index 0000000..128ac7b --- /dev/null +++ b/arxiv-processor/go.mod @@ -0,0 +1,3 @@ +module arxiv-processor + +go 1.21 diff --git a/arxiv-processor/main.go b/arxiv-processor/main.go new file mode 100644 index 0000000..6ec1792 --- /dev/null +++ b/arxiv-processor/main.go @@ -0,0 +1,214 @@ +package main + +import ( + "bytes" + "encoding/json" + "encoding/xml" + "flag" + "fmt" + "io" + "log" + "net/http" + "net/url" + "os" + "strings" + "time" +) + +func main() { + searchQuery := flag.String("search", "", "Search query") + dateRange := flag.String("date-range", "", "Date range in YYYYMMDD-YYYYMMDD format") + outputFile := flag.String("output", "papers_data.json", "Output file path") + + flag.Parse() + + if *searchQuery == "" || *dateRange == "" { + log.Fatal("Both --search and --date-range are required") + } + + // Parse date range + dates := strings.Split(*dateRange, "-") + if len(dates) != 2 { + log.Fatal("Invalid date range format. Use YYYYMMDD-YYYYMMDD") + } + + startDate, err := time.Parse("20060102", dates[0]) + if err != nil { + log.Fatalf("Invalid start date: %v", err) + } + + endDate, err := time.Parse("20060102", dates[1]) + if err != nil { + log.Fatalf("Invalid end date: %v", err) + } + + // Create arXiv client + client := NewClient() + + // Fetch papers + fmt.Println("Fetching papers...") + papers, err := client.FetchPapers(*searchQuery, startDate, endDate) + if err != nil { + log.Fatalf("Failed to fetch papers: %v", err) + } + + fmt.Printf("Fetched %d papers\n", len(papers)) + + // Save papers to JSON + err = SavePapers(*outputFile, papers) + if err != nil { + log.Fatalf("Failed to save papers: %v", err) + } + + fmt.Printf("Saved paper data to %s\n", *outputFile) + + // LLM processing placeholder + fmt.Println("\nTo process these papers with an LLM:") + fmt.Println("1. Choose an LLM API (e.g., OpenAI, Anthropic, local model)") + fmt.Println("2. Implement the LLM integration in llm package") + fmt.Println("3. Run the program again with your criteria") +} + +// Paper represents a single arXiv paper +type Paper struct { + Title string `json:"title"` + Abstract string `json:"abstract"` + ID string `json:"arxiv_id"` +} + +// Client handles arXiv API interactions +type Client struct { + httpClient *http.Client + baseURL string +} + +// NewClient creates a new arXiv client +func NewClient() *Client { + return &Client{ + httpClient: &http.Client{ + Timeout: 30 * time.Second, + }, + baseURL: "http://export.arxiv.org/api/query", + } +} + +// FetchPapers retrieves papers matching the search query within the date range +func (c *Client) FetchPapers(searchQuery string, startDate, endDate time.Time) ([]Paper, error) { + var papers []Paper + start := 0 + batchSize := 100 + delay := 3 * time.Second + + for { + // Construct query with date range and pagination + query := fmt.Sprintf("%s AND submittedDate:[%s TO %s]", + searchQuery, + startDate.Format("20060102"), + endDate.Format("20060102"), + ) + + // Build URL with query parameters + params := url.Values{} + params.Add("search_query", query) + params.Add("start", fmt.Sprintf("%d", start)) + params.Add("max_results", fmt.Sprintf("%d", batchSize)) + + reqURL := fmt.Sprintf("%s?%s", c.baseURL, params.Encode()) + + // Make HTTP request + fmt.Printf("Making request to: %s\n", reqURL) + resp, err := c.httpClient.Get(reqURL) + if err != nil { + return nil, fmt.Errorf("failed to fetch papers: %w", err) + } + defer resp.Body.Close() + + // Read and log raw response for debugging + bodyBytes, err := io.ReadAll(resp.Body) + if err != nil { + return nil, fmt.Errorf("failed to read response body: %w", err) + } + fmt.Printf("Response status: %s\n", resp.Status) + fmt.Printf("Raw API response (first 1000 chars):\n%s\n", string(bodyBytes[:min(len(bodyBytes), 1000)])) + + // Parse XML response + batch, totalResults, err := c.parseResponse(bytes.NewReader(bodyBytes)) + if err != nil { + return nil, fmt.Errorf("failed to parse response: %w", err) + } + + papers = append(papers, batch...) + start += len(batch) + + // Check if we've fetched all papers + if start >= totalResults || len(batch) == 0 { + break + } + + // Respect arXiv's rate limits + time.Sleep(delay) + } + + return papers, nil +} + +// parseResponse handles XML parsing of the arXiv API response +func (c *Client) parseResponse(r io.Reader) ([]Paper, int, error) { + type atomEntry struct { + Title string `xml:"title"` + Summary string `xml:"summary"` + ID string `xml:"id"` + } + + type atomFeed struct { + XMLName xml.Name `xml:"feed"` + TotalResults int `xml:"totalResults"` + Entries []atomEntry `xml:"entry"` + } + + var feed atomFeed + if err := xml.NewDecoder(r).Decode(&feed); err != nil { + return nil, 0, fmt.Errorf("failed to decode XML: %w", err) + } + + var papers []Paper + for _, entry := range feed.Entries { + // Extract just the ID part from the full URL + idParts := strings.Split(entry.ID, "/abs/") + if len(idParts) < 2 { + continue + } + + papers = append(papers, Paper{ + Title: strings.TrimSpace(entry.Title), + Abstract: strings.TrimSpace(entry.Summary), + ID: idParts[1], + }) + } + + return papers, feed.TotalResults, nil +} + +// SavePapers saves the papers to a JSON file +func SavePapers(filename string, papers []Paper) error { + file, err := os.Create(filename) + if err != nil { + return fmt.Errorf("failed to create file: %w", err) + } + defer file.Close() + + encoder := json.NewEncoder(file) + encoder.SetIndent("", " ") + if err := encoder.Encode(papers); err != nil { + return fmt.Errorf("failed to encode JSON: %w", err) + } + + return nil +} + +func min(a, b int) int { + if a < b { + return a + } + return b +} diff --git a/arxiv-processor/storage/json.go b/arxiv-processor/storage/json.go new file mode 100644 index 0000000..24c0dcb --- /dev/null +++ b/arxiv-processor/storage/json.go @@ -0,0 +1,45 @@ +package storage + +import ( + "encoding/json" + "fmt" + "os" + "time" + + "arxiv-processor/arxiv" +) + +// JSONStorage handles saving papers to JSON files +type JSONStorage struct { + FilePath string +} + +// NewJSONStorage creates a new JSON storage handler +func NewJSONStorage(filePath string) *JSONStorage { + return &JSONStorage{FilePath: filePath} +} + +// SavePapers saves papers to a JSON file +func (s *JSONStorage) SavePapers(papers []arxiv.Paper) error { + file, err := os.Create(s.FilePath) + if err != nil { + return fmt.Errorf("failed to create file: %w", err) + } + defer file.Close() + + encoder := json.NewEncoder(file) + encoder.SetIndent("", " ") + if err := encoder.Encode(papers); err != nil { + return fmt.Errorf("failed to encode papers: %w", err) + } + + return nil +} + +// Paper represents a simplified paper structure for JSON storage +type Paper struct { + ID string `json:"id"` + Title string `json:"title"` + Published time.Time `json:"published"` + Authors []string `json:"authors"` +} diff --git a/criteria.txt b/criteria.txt new file mode 100644 index 0000000..47b90ae --- /dev/null +++ b/criteria.txt @@ -0,0 +1,43 @@ +Criteria for Paper Selection: + +When evaluating paper titles and abstracts, please consider the following criteria. Prioritize inclusiveness to reduce the risk of missing potentially interesting research. + +Papers must meet ALL of these criteria: + +1. Practical Applications: Does the paper focus on real-world or practical applications of Large Language Models (LLMs), particularly in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI? + +2. Experimental Results and Quantitative Metrics: Does the paper include experimental results with quantitative metrics that demonstrate performance improvements or innovative applications? Look for concrete results related to areas like prompt engineering. + +3. Comparison with State-of-the-Art: Does the paper compare its results or methodologies with existing state-of-the-art techniques, and does it demonstrate any advancements in terms of performance, scalability, or efficiency? + +Papers should meet at least one of these criteria: + +4. Methodology and Implementation Details: Does the paper clearly describe its methodology and implementation, particularly how LLMs are utilized in practical or real-world tasks? + +5. Real-world Applications and Challenges: Does the paper discuss real-world applications or address limitations and challenges involving LLMs, especially in autonomous or agentic AI scenarios? + +Additional Considerations: +- Novelty: Does the paper introduce novel approaches or techniques, especially those extending the application or integration of LLMs with technologies like knowledge graphs? + +- Is the paper's approach implementable with current standard tools? + +- Agentic AI: Does the paper describe applications where LLMs enable AI to autonomously perform complex, real-world tasks beyond sandboxed interactions? + +- Reproducibility and Documentation: Are there sufficient details to support reproducibility and transparent documentation of the methodologies and results? + +- Impact through Experimental Validation: Does the paper show robust experimental validation, with datasets and methods that closely reflect real-world scenarios? + +The paper should be REJECTED if it: + +- Primarily focuses on medical applications of AI +- Primarily focuses on social applications of AI in regard to Diversity, Social harm, or similar issues. +- Primarily focuses on video processing +- Primarily focuses on responsible AI application or AI ethics +- Primarily focuses on law, either with AI as subject or participant. + +Instructions: +- Analyze each paper's title and abstract to determine how many criteria are met. +- Prioritize inclusiveness: Favor selection over rejection if a paper seems marginally relevant. +- Use a simple scoring system: A paper meeting at least two or three criteria should be accepted for further review. + +Please make a best-guess decision based on the information provided in the titles and abstracts. Err on the side of potential inclusion to capture a broad spectrum of relevant research. diff --git a/go.mod b/go.mod new file mode 100644 index 0000000..f2be03d --- /dev/null +++ b/go.mod @@ -0,0 +1,15 @@ +module paper-system + +go 1.23.4 + +replace arxiv-processor => ./arxiv-processor + +replace llm_processor => ./llm_processor + +replace json2md => ./json2md + +require ( + arxiv-processor v0.0.0-00010101000000-000000000000 + json2md v0.0.0-00010101000000-000000000000 + llm_processor v0.0.0-00010101000000-000000000000 +) diff --git a/json2md/README.md b/json2md/README.md new file mode 100644 index 0000000..0bd3810 --- /dev/null +++ b/json2md/README.md @@ -0,0 +1,83 @@ +# json2md - JSON to Markdown Converter + +A Go module for converting academic paper review data from JSON to formatted Markdown reports. + +## Features +- CLI interface for easy conversion +- Reusable library package +- Supports accepted/rejected paper categorization +- Preserves abstract formatting +- Generates arXiv links automatically + +## Installation +```bash +git clone https://github.com/yourusername/json2md.git +cd json2md +go build +``` + +## Command Line Usage +```bash +./json2md --input papers.json --output results.md + +# Or directly via Go: +go run main.go --input ../papers.json --output ../results.md +``` + +## Library Usage +```go +package main + +import ( + "fmt" + "json2md/lib" +) + +func main() { + // Process JSON file + data, err := lib.ProcessJSONFile("input.json") + if err != nil { + panic(err) + } + + // Generate Markdown + if err := lib.GenerateMarkdown(data, "output.md"); err != nil { + panic(err) + } + + fmt.Println("Conversion successful!") +} +``` + +## Input JSON Format +Expected structure: +```json +{ + "accepted": [ + { + "paper": { + "title": "Paper Title", + "arxiv_id": "1234.56789", + "abstract": "Paper abstract..." + }, + "explanation": "Acceptance reason" + } + ], + "rejected": [ + { + "paper": { /* same structure */ }, + "explanation": "Rejection reason" + } + ] +} +``` + +## Project Structure +``` +json2md/ +├── lib/ # Core conversion library +│ └── lib.go +├── main.go # CLI interface +├── go.mod # Module configuration +└── README.md # This file +``` diff --git a/json2md/go.mod b/json2md/go.mod new file mode 100644 index 0000000..abed27c --- /dev/null +++ b/json2md/go.mod @@ -0,0 +1,3 @@ +module json2md + +go 1.21 diff --git a/json2md/lib/lib.go b/json2md/lib/lib.go new file mode 100644 index 0000000..54ead82 --- /dev/null +++ b/json2md/lib/lib.go @@ -0,0 +1,136 @@ +package lib + +import ( + "encoding/json" + "fmt" + "os" + "strings" +) + +type Paper struct { + Title string `json:"title"` + ArxivID string `json:"arxiv_id"` + Abstract string `json:"abstract"` +} + +type PaperDecision struct { + Paper Paper `json:"paper"` + Explanation string `json:"explanation"` +} + +type DecisionData struct { + Accepted []PaperDecision `json:"accepted"` + Rejected []PaperDecision `json:"rejected"` +} + +func ProcessJSONFile(inputPath string) (DecisionData, error) { + var data DecisionData + + jsonFile, err := os.Open(inputPath) + if err != nil { + return data, fmt.Errorf("error opening input file: %w", err) + } + defer jsonFile.Close() + + // First try decoding as DecisionData format + decoder := json.NewDecoder(jsonFile) + if err := decoder.Decode(&data); err == nil { + return data, nil + } + + // Try alternative formats + jsonData, err := os.ReadFile(inputPath) + if err != nil { + return data, fmt.Errorf("error reading input file: %w", err) + } + + // Try array format with status field + var papers []struct { + Paper Paper `json:"paper"` + Explanation string `json:"explanation"` + Status string `json:"status"` + } + if err := json.Unmarshal(jsonData, &papers); err == nil { + for _, p := range papers { + switch p.Status { + case "accepted": + data.Accepted = append(data.Accepted, PaperDecision{p.Paper, p.Explanation}) + case "rejected": + data.Rejected = append(data.Rejected, PaperDecision{p.Paper, p.Explanation}) + default: + fmt.Printf("Warning: Paper '%s' has unknown status '%s'\n", p.Paper.Title, p.Status) + } + } + return data, nil + } + + // Try original object format + var objData struct { + Accepted []PaperDecision `json:"accepted"` + Rejected []PaperDecision `json:"rejected"` + } + if err := json.Unmarshal(jsonData, &objData); err != nil { + return data, fmt.Errorf("error decoding JSON: %w", err) + } + + data.Accepted = objData.Accepted + data.Rejected = objData.Rejected + return data, nil +} + +func GenerateMarkdown(data DecisionData, outputPath string) error { + file, err := os.Create(outputPath) + if err != nil { + return fmt.Errorf("error creating output file: %w", err) + } + defer file.Close() + + // Accepted Papers section + fmt.Fprintln(file, "# Accepted Papers\n") + for _, paper := range data.Accepted { + writePaperSection(file, paper, "accepted") + } + + // Rejected Papers section + fmt.Fprintln(file, "# Rejected Papers\n") + for _, paper := range data.Rejected { + writePaperSection(file, paper, "rejected") + } + + return nil +} + +func writePaperSection(file *os.File, paper PaperDecision, decisionType string) { + // Escape special characters in title + title := strings.ReplaceAll(paper.Paper.Title, "[", "\\[") + title = strings.ReplaceAll(title, "]", "\\]") + + fmt.Fprintf(file, "## [%s](https://arxiv.org/abs/%s)\n", title, paper.Paper.ArxivID) + fmt.Fprintf(file, "**arXiv ID:** %s\n\n", paper.Paper.ArxivID) + fmt.Fprintln(file, "**Abstract:**") + + // Format abstract as blockquote, handling multiple paragraphs + abstract := strings.TrimSpace(paper.Paper.Abstract) + paragraphs := strings.Split(abstract, "\n\n") + for i, p := range paragraphs { + // Ensure each line starts with > + lines := strings.Split(strings.TrimSpace(p), "\n") + for _, line := range lines { + fmt.Fprintf(file, "> %s\n", strings.TrimSpace(line)) + } + if i < len(paragraphs)-1 { + fmt.Fprintln(file, ">") + } + } + fmt.Fprintln(file) + + // Format explanation, ensuring it's properly escaped and formatted + fmt.Fprintln(file, "**Decision Explanation:**") + explanation := strings.TrimSpace(paper.Explanation) + + // If the explanation is already properly formatted (no raw JSON or code blocks), + // write it directly + fmt.Fprintf(file, "%s\n\n", explanation) + + fmt.Fprintln(file, "---\n") +} diff --git a/json2md/main.go b/json2md/main.go new file mode 100644 index 0000000..9ba1acc --- /dev/null +++ b/json2md/main.go @@ -0,0 +1,35 @@ +package main + +import ( + "flag" + "fmt" + "os" + + "json2md/lib" +) + +func main() { + inputPath := flag.String("input", "", "Input JSON file path") + outputPath := flag.String("output", "decisions.md", "Output Markdown file path") + flag.Parse() + + if *inputPath == "" { + fmt.Println("Error: --input parameter is required") + os.Exit(1) + } + + // Process JSON file using library + data, err := lib.ProcessJSONFile(*inputPath) + if err != nil { + fmt.Printf("Error processing JSON file: %v\n", err) + os.Exit(1) + } + + // Generate Markdown using library + if err := lib.GenerateMarkdown(data, *outputPath); err != nil { + fmt.Printf("Error generating Markdown: %v\n", err) + os.Exit(1) + } + + fmt.Printf("Successfully generated Markdown file: %s\n", *outputPath) +} diff --git a/json2md/papers-output.md b/json2md/papers-output.md new file mode 100644 index 0000000..3c656ec --- /dev/null +++ b/json2md/papers-output.md @@ -0,0 +1,3997 @@ +# Accepted Papers + +## [Monte Carlo Tree Search for Comprehensive Exploration in LLM-Based + Automatic Heuristic Design](https://arxiv.org/abs/2501.08603v2) +**arXiv ID:** 2501.08603v2 + +**Abstract:** +> Handcrafting heuristics for solving complex planning tasks (e.g., NP-hard +combinatorial optimization (CO) problems) is a common practice but requires +extensive domain knowledge. Recently, Large Language Model (LLM)-based +automatic heuristics design (AHD) methods have shown promise in generating +high-quality heuristics without manual intervention. Existing LLM-based AHD +methods employ a population to maintain a fixed number of top-performing +LLM-generated heuristics and introduce evolutionary computation (EC) to enhance +the population iteratively. However, the population-based procedure brings +greedy properties, often resulting in convergence to local optima. Instead, to +more comprehensively explore the space of heuristics, we propose using Monte +Carlo Tree Search (MCTS) for LLM-based heuristic evolution while preserving all +LLM-generated heuristics in a tree structure. With a novel thought-alignment +process and an exploration-decay technique, the proposed MCTS-AHD method +delivers significantly higher-quality heuristics on various complex tasks. Our +code is available at https://github.com/zz1358m/MCTS-AHD-master. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs in automatic heuristic design, including experimental results with quantitative metrics, and comparing its results to existing state-of-the-art techniques, demonstrating performance improvements. + +--- + +## [Leveraging Large Language Models as Knowledge-Driven Agents for Reliable + Retrosynthesis Planning](https://arxiv.org/abs/2501.08897v1) +**arXiv ID:** 2501.08897v1 + +**Abstract:** +> Identifying reliable synthesis pathways in materials chemistry is a complex +task, particularly in polymer science, due to the intricate and often +non-unique nomenclature of macromolecules. To address this challenge, we +propose an agent system that integrates large language models (LLMs) and +knowledge graphs (KGs). By leveraging LLMs' powerful capabilities for +extracting and recognizing chemical substance names, and storing the extracted +data in a structured knowledge graph, our system fully automates the retrieval +of relevant literatures, extraction of reaction data, database querying, +construction of retrosynthetic pathway trees, further expansion through the +retrieval of additional literature and recommendation of optimal reaction +pathways. A novel Multi-branched Reaction Pathway Search (MBRPS) algorithm +enables the exploration of all pathways, with a particular focus on +multi-branched ones, helping LLMs overcome weak reasoning in multi-branched +paths. This work represents the first attempt to develop a fully automated +retrosynthesis planning agent tailored specially for macromolecules powered by +LLMs. Applied to polyimide synthesis, our new approach constructs a +retrosynthetic pathway tree with hundreds of pathways and recommends optimized +routes, including both known and novel pathways, demonstrating its +effectiveness and potential for broader applications. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs in knowledge graphs and retrieval-augmented generation, including experimental results with quantitative metrics, and comparing its results to state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [Text Semantics to Flexible Design: A Residential Layout Generation + Method Based on Stable Diffusion Model](https://arxiv.org/abs/2501.09279v1) +**arXiv ID:** 2501.09279v1 + +**Abstract:** +> Flexibility in the AI-based residential layout design remains a significant +challenge, as traditional methods like rule-based heuristics and graph-based +generation often lack flexibility and require substantial design knowledge from +users. To address these limitations, we propose a cross-modal design approach +based on the Stable Diffusion model for generating flexible residential +layouts. The method offers multiple input types for learning objectives, +allowing users to specify both boundaries and layouts. It incorporates natural +language as design constraints and introduces ControlNet to enable stable +layout generation through two distinct pathways. We also present a scheme that +encapsulates design expertise within a knowledge graph and translates it into +natural language, providing an interpretable representation of design +knowledge. This comprehensibility and diversity of input options enable +professionals and non-professionals to directly express design requirements, +enhancing flexibility and controllability. Finally, experiments verify the +flexibility of the proposed methods under multimodal constraints better than +state-of-the-art models, even when specific semantic information about room +areas or connections is incomplete. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models in residential layout generation, including experimental results with quantitative metrics, and comparing its results to state-of-the-art models. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [SOP-Agent: Empower General Purpose AI Agent with Domain-Specific SOPs](https://arxiv.org/abs/2501.09316v1) +**arXiv ID:** 2501.09316v1 + +**Abstract:** +> Despite significant advancements in general-purpose AI agents, several +challenges still hinder their practical application in real-world scenarios. +First, the limited planning capabilities of Large Language Models (LLM) +restrict AI agents from effectively solving complex tasks that require +long-horizon planning. Second, general-purpose AI agents struggle to +efficiently utilize domain-specific knowledge and human expertise. In this +paper, we introduce the Standard Operational Procedure-guided Agent +(SOP-agent), a novel framework for constructing domain-specific agents through +pseudocode-style Standard Operational Procedures (SOPs) written in natural +language. Formally, we represent a SOP as a decision graph, which is traversed +to guide the agent in completing tasks specified by the SOP. We conduct +extensive experiments across tasks in multiple domains, including +decision-making, search and reasoning, code generation, data cleaning, and +grounded customer service. The SOP-agent demonstrates excellent versatility, +achieving performance superior to general-purpose agent frameworks and +comparable to domain-specific agent systems. Additionally, we introduce the +Grounded Customer Service Benchmark, the first benchmark designed to evaluate +the grounded decision-making capabilities of AI agents in customer service +scenarios based on SOPs. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models, demonstrating experimental results with quantitative metrics, and comparing its results with state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. The paper introduces a novel approach and demonstrates robust experimental validation, making it a strong candidate for acceptance. + +--- + +## [Aligning Instruction Tuning with Pre-training](https://arxiv.org/abs/2501.09368v1) +**arXiv ID:** 2501.09368v1 + +**Abstract:** +> Instruction tuning enhances large language models (LLMs) to follow human +instructions across diverse tasks, relying on high-quality datasets to guide +behavior. However, these datasets, whether manually curated or synthetically +generated, are often narrowly focused and misaligned with the broad +distributions captured during pre-training, limiting LLM generalization and +effective use of pre-trained knowledge. We propose *Aligning Instruction Tuning +with Pre-training* (AITP), a method that bridges this gap by identifying +coverage shortfalls in instruction-tuning datasets and rewriting +underrepresented pre-training data into high-quality instruction-response +pairs. This approach enriches dataset diversity while preserving task-specific +objectives. Evaluations on three fully open LLMs across eight benchmarks +demonstrate consistent performance improvements with AITP. Ablations highlight +the benefits of adaptive data selection, controlled rewriting, and balanced +integration, emphasizing the importance of aligning instruction tuning with +pre-training distributions to unlock the full potential of LLMs. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs, including experimental results with quantitative metrics, and comparing its results with existing state-of-the-art techniques, demonstrating performance improvements. + +--- + +## [Platform-Aware Mission Planning](https://arxiv.org/abs/2501.09632v1) +**arXiv ID:** 2501.09632v1 + +**Abstract:** +> Planning for autonomous systems typically requires reasoning with models at +different levels of abstraction, and the harmonization of two competing sets of +objectives: high-level mission goals that refer to an interaction of the system +with the external environment, and low-level platform constraints that aim to +preserve the integrity and the correct interaction of the subsystems. The +complicated interplay between these two models makes it very hard to reason on +the system as a whole, especially when the objective is to find plans with +robustness guarantees, considering the non-deterministic behavior of the lower +layers of the system. + In this paper, we introduce the problem of Platform-Aware Mission Planning +(PAMP), addressing it in the setting of temporal durative actions. The PAMP +problem differs from standard temporal planning for its exists-forall nature: +the high-level plan dealing with mission goals is required to satisfy safety +and executability constraints, for all the possible non-deterministic +executions of the low-level model of the platform and the environment. We +propose two approaches for solving PAMP. The first baseline approach +amalgamates the mission and platform levels, while the second is based on an +abstraction-refinement loop that leverages the combination of a planner and a +verification engine. We prove the soundness and completeness of the proposed +approaches and validate them experimentally, demonstrating the importance of +heterogeneous modeling and the superiority of the technique based on +abstraction-refinement. + +**Decision Explanation:** The paper meets criteria 2 and 3 by including experimental results with quantitative metrics and comparing its results with existing state-of-the-art techniques, demonstrating advancements in performance. It also meets criteria 4 by clearly describing its methodology and implementation, and touches on agentic AI by discussing autonomous systems and mission planning. + +--- + +## [NS-Gym: Open-Source Simulation Environments and Benchmarks for + Non-Stationary Markov Decision Processes](https://arxiv.org/abs/2501.09646v1) +**arXiv ID:** 2501.09646v1 + +**Abstract:** +> In many real-world applications, agents must make sequential decisions in +environments where conditions are subject to change due to various exogenous +factors. These non-stationary environments pose significant challenges to +traditional decision-making models, which typically assume stationary dynamics. +Non-stationary Markov decision processes (NS-MDPs) offer a framework to model +and solve decision problems under such changing conditions. However, the lack +of standardized benchmarks and simulation tools has hindered systematic +evaluation and advance in this field. We present NS-Gym, the first simulation +toolkit designed explicitly for NS-MDPs, integrated within the popular +Gymnasium framework. In NS-Gym, we segregate the evolution of the environmental +parameters that characterize non-stationarity from the agent's decision-making +module, allowing for modular and flexible adaptations to dynamic environments. +We review prior work in this domain and present a toolkit encapsulating key +problem characteristics and types in NS-MDPs. This toolkit is the first effort +to develop a set of standardized interfaces and benchmark problems to enable +consistent and reproducible evaluation of algorithms under non-stationary +conditions. We also benchmark six algorithmic approaches from prior work on +NS-MDPs using NS-Gym. Our vision is that NS-Gym will enable researchers to +assess the adaptability and robustness of their decision-making algorithms to +non-stationary conditions. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of decision-making models in non-stationary environments, presenting experimental results with quantitative metrics, and comparing its results with existing state-of-the-art techniques. Additionally, it meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges. + +--- + +## [Doc-Guided Sent2Sent++: A Sent2Sent++ Agent with Doc-Guided memory for + Document-level Machine Translation](https://arxiv.org/abs/2501.08523v1) +**arXiv ID:** 2501.08523v1 + +**Abstract:** +> The field of artificial intelligence has witnessed significant advancements +in natural language processing, largely attributed to the capabilities of Large +Language Models (LLMs). These models form the backbone of Agents designed to +address long-context dependencies, particularly in Document-level Machine +Translation (DocMT). DocMT presents unique challenges, with quality, +consistency, and fluency being the key metrics for evaluation. Existing +approaches, such as Doc2Doc and Doc2Sent, either omit sentences or compromise +fluency. This paper introduces Doc-Guided Sent2Sent++, an Agent that employs an +incremental sentence-level forced decoding strategy \textbf{to ensure every +sentence is translated while enhancing the fluency of adjacent sentences.} Our +Agent leverages a Doc-Guided Memory, focusing solely on the summary and its +translation, which we find to be an efficient approach to maintaining +consistency. Through extensive testing across multiple languages and domains, +we demonstrate that Sent2Sent++ outperforms other methods in terms of quality, +consistency, and fluency. The results indicate that, our approach has achieved +significant improvements in metrics such as s-COMET, d-COMET, LTCR-$1_f$, and +document-level perplexity (d-ppl). The contributions of this paper include a +detailed analysis of current DocMT research, the introduction of the +Sent2Sent++ decoding method, the Doc-Guided Memory mechanism, and validation of +its effectiveness across languages and domains. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models in document-level machine translation, including experimental results with quantitative metrics, and comparing its results to state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [LlamaRestTest: Effective REST API Testing with Small Language Models](https://arxiv.org/abs/2501.08598v1) +**arXiv ID:** 2501.08598v1 + +**Abstract:** +> Modern web services rely heavily on REST APIs, typically documented using the +OpenAPI specification. The widespread adoption of this standard has resulted in +the development of many black-box testing tools that generate tests based on +these specifications. Recent advancements in Natural Language Processing (NLP), +particularly with Large Language Models (LLMs), have enhanced REST API testing +by extracting actionable rules and generating input values from the +human-readable portions of the specification. However, these advancements +overlook the potential of continuously refining the identified rules and test +inputs based on server responses. To address this limitation, we present +LlamaRestTest, a novel approach that employs two custom LLMs to generate +realistic test inputs and uncover parameter dependencies during the testing +process by incorporating server responses. These LLMs are created by +fine-tuning the Llama3-8b model, using mined datasets of REST API example +values and inter-parameter dependencies. We evaluated LlamaRestTest on 12 +real-world services (including popular services such as Spotify), comparing it +against RESTGPT, a GPT-powered specification-enhancement tool, as well as +several state-of-the-art REST API testing tools, including RESTler, MoRest, +EvoMaster, and ARAT-RL. Our results show that fine-tuning enables smaller LLMs +to outperform larger models in detecting actionable rules and generating inputs +for REST API testing. We evaluated configurations from the base Llama3-8B to +fine-tuned versions and explored 2-bit, 4-bit, and 8-bit quantization for +efficiency. LlamaRestTest surpasses state-of-the-art tools in code coverage and +error detection, even with RESTGPT-enhanced specifications, and an ablation +study highlights the impact of its novel components. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models in REST API testing, including experimental results with quantitative metrics and comparisons with state-of-the-art techniques, demonstrating performance improvements. + +--- + +## [AutoRestTest: A Tool for Automated REST API Testing Using LLMs and MARL](https://arxiv.org/abs/2501.08600v1) +**arXiv ID:** 2501.08600v1 + +**Abstract:** +> As REST APIs have become widespread in modern web services, comprehensive +testing of these APIs has become increasingly crucial. Due to the vast search +space consisting of operations, parameters, and parameter values along with +their complex dependencies and constraints, current testing tools suffer from +low code coverage, leading to suboptimal fault detection. To address this +limitation, we present a novel tool, AutoRestTest, which integrates the +Semantic Operation Dependency Graph (SODG) with Multi-Agent Reinforcement +Learning (MARL) and large language models (LLMs) for effective REST API +testing. AutoRestTest determines operation-dependent parameters using the SODG +and employs five specialized agents (operation, parameter, value, dependency, +and header) to identify dependencies of operations and generate operation +sequences, parameter combinations, and values. AutoRestTest provides a +command-line interface and continuous telemetry on successful operation count, +unique server errors detected, and time elapsed. Upon completion, AutoRestTest +generates a detailed report highlighting errors detected and operations +exercised. In this paper, we introduce our tool and present preliminary +results. + +**Decision Explanation:** The paper meets criteria 1, 2, and 4 by focusing on a practical application of LLMs in REST API testing, including experimental results, and clearly describing its methodology and implementation. It also introduces a novel approach by integrating LLMs with MARL, showing potential for innovative applications and performance improvements. + +--- + +## [MAGNET: Augmenting Generative Decoders with Representation Learning and + Infilling Capabilities](https://arxiv.org/abs/2501.08648v1) +**arXiv ID:** 2501.08648v1 + +**Abstract:** +> While originally designed for unidirectional generative modeling, +decoder-only large language models (LLMs) are increasingly being adapted for +bidirectional modeling. However, unidirectional and bidirectional models are +typically trained separately with distinct objectives (generation and +representation learning, respectively). This separation overlooks the +opportunity for developing a more versatile language model and for these +objectives to complement each other. In this work, we introduce MAGNET, an +adaptation of decoder-only LLMs that enhances their ability to generate robust +representations and infill missing text spans, while preserving their knowledge +and text generation capabilities. MAGNET employs three self-supervised training +objectives and introduces an attention mechanism that combines bidirectional +and causal attention, enabling unified training across all objectives. Our +results demonstrate that LLMs adapted with MAGNET (1) surpass strong text +encoders on token-level and sentence-level representation learning tasks, (2) +generate contextually appropriate text infills by leveraging future context, +(3) retain the ability for open-ended text generation without exhibiting +repetition problem, and (4) preserve the knowledge gained by the LLM during +pretraining. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs, including experimental results with quantitative metrics, and comparing its results with state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [Incrementally Learning Multiple Diverse Data Domains via Multi-Source + Dynamic Expansion Model](https://arxiv.org/abs/2501.08878v1) +**arXiv ID:** 2501.08878v1 + +**Abstract:** +> Continual Learning seeks to develop a model capable of incrementally +assimilating new information while retaining prior knowledge. However, current +research predominantly addresses a straightforward learning context, wherein +all data samples originate from a singular data domain. This paper shifts focus +to a more complex and realistic learning environment, characterized by data +samples sourced from multiple distinct domains. We tackle this intricate +learning challenge by introducing a novel methodology, termed the Multi-Source +Dynamic Expansion Model (MSDEM), which leverages various pre-trained models as +backbones and progressively establishes new experts based on them to adapt to +emerging tasks. Additionally, we propose an innovative dynamic expandable +attention mechanism designed to selectively harness knowledge from multiple +backbones, thereby accelerating the new task learning. Moreover, we introduce a +dynamic graph weight router that strategically reuses all previously acquired +parameters and representations for new task learning, maximizing the positive +knowledge transfer effect, which further improves generalization performance. +We conduct a comprehensive series of experiments, and the empirical findings +indicate that our proposed approach achieves state-of-the-art performance. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of incremental learning, including experimental results with quantitative metrics, and comparing its results to state-of-the-art techniques, demonstrating advancements in performance. + +--- + +## [Attention is All You Need Until You Need Retention](https://arxiv.org/abs/2501.09166v1) +**arXiv ID:** 2501.09166v1 + +**Abstract:** +> This work introduces a novel Retention Layer mechanism for Transformer based +architectures, addressing their inherent lack of intrinsic retention +capabilities. Unlike human cognition, which can encode and dynamically recall +symbolic templates, Generative Pretrained Transformers rely solely on fixed +pretrained weights and ephemeral context windows, limiting their adaptability. +The proposed Retention Layer incorporates a persistent memory module capable of +real time data population, dynamic recall, and guided output generation. This +enhancement allows models to store, update, and reuse observed patterns across +sessions, enabling incremental learning and bridging the gap between static +pretraining and dynamic, context sensitive adaptation. The Retention Layer +design parallels social learning processes, encompassing attention, retention, +reproduction, and motivation stages. Technically, it integrates a memory +attention mechanism and episodic buffers to manage memory scalability, mitigate +overfitting, and ensure efficient recall. Applications span adaptive personal +assistants, real time fraud detection, autonomous robotics, content moderation, +and healthcare diagnostics. In each domain, the retention mechanism enables +systems to learn incrementally, personalize outputs, and respond to evolving +real world challenges effectively. By emulating key aspects of human learning, +this retention enhanced architecture fosters a more fluid and responsive AI +paradigm, paving the way for dynamic, session aware models that extend the +capabilities of traditional Transformers into domains requiring continual +adaptation. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by introducing a novel Retention Layer mechanism for Transformer-based architectures, demonstrating potential performance improvements, and comparing its results to existing state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. The paper's approach is novel, implementable with current standard tools, and has potential for reproducibility and robust experimental validation. + +--- + +## [Guiding Retrieval using LLM-based Listwise Rankers](https://arxiv.org/abs/2501.09186v1) +**arXiv ID:** 2501.09186v1 + +**Abstract:** +> Large Language Models (LLMs) have shown strong promise as rerankers, +especially in ``listwise'' settings where an LLM is prompted to rerank several +search results at once. However, this ``cascading'' retrieve-and-rerank +approach is limited by the bounded recall problem: relevant documents not +retrieved initially are permanently excluded from the final ranking. Adaptive +retrieval techniques address this problem, but do not work with listwise +rerankers because they assume a document's score is computed independently from +other documents. In this paper, we propose an adaptation of an existing +adaptive retrieval method that supports the listwise setting and helps guide +the retrieval process itself (thereby overcoming the bounded recall problem for +LLM rerankers). Specifically, our proposed algorithm merges results both from +the initial ranking and feedback documents provided by the most relevant +documents seen up to that point. Through extensive experiments across diverse +LLM rerankers, first stage retrievers, and feedback sources, we demonstrate +that our method can improve nDCG@10 by up to 13.23% and recall by 28.02%--all +while keeping the total number of LLM inferences constant and overheads due to +the adaptive process minimal. The work opens the door to leveraging LLM-based +search in settings where the initial pool of results is limited, e.g., by +legacy systems, or by the cost of deploying a semantic first-stage. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs in retrieval-augmented generation, presenting experimental results with quantitative metrics, and comparing its results to existing state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges. + +--- + +## [Perspective Transition of Large Language Models for Solving Subjective + Tasks](https://arxiv.org/abs/2501.09265v1) +**arXiv ID:** 2501.09265v1 + +**Abstract:** +> Large language models (LLMs) have revolutionized the field of natural +language processing, enabling remarkable progress in various tasks. Different +from objective tasks such as commonsense reasoning and arithmetic +question-answering, the performance of LLMs on subjective tasks is still +limited, where the perspective on the specific problem plays crucial roles for +better interpreting the context and giving proper response. For example, in +certain scenarios, LLMs may perform better when answering from an expert role +perspective, potentially eliciting their relevant domain knowledge. In +contrast, in some scenarios, LLMs may provide more accurate responses when +answering from a third-person standpoint, enabling a more comprehensive +understanding of the problem and potentially mitigating inherent biases. In +this paper, we propose Reasoning through Perspective Transition (RPT), a method +based on in-context learning that enables LLMs to dynamically select among +direct, role, and third-person perspectives for the best way to solve +corresponding subjective problem. Through extensive experiments on totally 12 +subjective tasks by using both closed-source and open-source LLMs including +GPT-4, GPT-3.5, Llama-3, and Qwen-2, our method outperforms widely used single +fixed perspective based methods such as chain-of-thought prompting and expert +prompting, highlights the intricate ways that LLMs can adapt their perspectives +to provide nuanced and contextually appropriate responses for different +problems. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models, including experimental results with quantitative metrics, and comparing its results with existing state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [On Learning Informative Trajectory Embeddings for Imitation, + Classification and Regression](https://arxiv.org/abs/2501.09327v1) +**arXiv ID:** 2501.09327v1 + +**Abstract:** +> In real-world sequential decision making tasks like autonomous driving, +robotics, and healthcare, learning from observed state-action trajectories is +critical for tasks like imitation, classification, and clustering. For example, +self-driving cars must replicate human driving behaviors, while robots and +healthcare systems benefit from modeling decision sequences, whether or not +they come from expert data. Existing trajectory encoding methods often focus on +specific tasks or rely on reward signals, limiting their ability to generalize +across domains and tasks. Inspired by the success of embedding models like CLIP +and BERT in static domains, we propose a novel method for embedding +state-action trajectories into a latent space that captures the skills and +competencies in the dynamic underlying decision-making processes. This method +operates without the need for reward labels, enabling better generalization +across diverse domains and tasks. Our contributions are threefold: (1) We +introduce a trajectory embedding approach that captures multiple abilities from +state-action data. (2) The learned embeddings exhibit strong representational +power across downstream tasks, including imitation, classification, clustering, +and regression. (3) The embeddings demonstrate unique properties, such as +controlling agent behaviors in IQ-Learn and an additive structure in the latent +space. Experimental results confirm that our method outperforms traditional +approaches, offering more flexible and powerful trajectory representations for +various applications. Our code is available at +https://github.com/Erasmo1015/vte. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on real-world applications of sequential decision making, including experimental results with quantitative metrics, and comparing its results to traditional approaches. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [Beyond Reward Hacking: Causal Rewards for Large Language Model Alignment](https://arxiv.org/abs/2501.09620v1) +**arXiv ID:** 2501.09620v1 + +**Abstract:** +> Recent advances in large language models (LLMs) have demonstrated significant +progress in performing complex tasks. While Reinforcement Learning from Human +Feedback (RLHF) has been effective in aligning LLMs with human preferences, it +is susceptible to spurious correlations in reward modeling. Consequently, it +often introduces biases-such as length bias, sycophancy, conceptual bias, and +discrimination that hinder the model's ability to capture true causal +relationships. To address this, we propose a novel causal reward modeling +approach that integrates causal inference to mitigate these spurious +correlations. Our method enforces counterfactual invariance, ensuring reward +predictions remain consistent when irrelevant variables are altered. Through +experiments on both synthetic and real-world datasets, we show that our +approach mitigates various types of spurious correlations effectively, +resulting in more reliable and fair alignment of LLMs with human preferences. +As a drop-in enhancement to the existing RLHF workflow, our causal reward +modeling provides a practical way to improve the trustworthiness and fairness +of LLM finetuning. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs, including experimental results with quantitative metrics, and comparing its results to existing state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges involving LLMs. + +--- + +## [The Heap: A Contamination-Free Multilingual Code Dataset for Evaluating + Large Language Models](https://arxiv.org/abs/2501.09653v1) +**arXiv ID:** 2501.09653v1 + +**Abstract:** +> The recent rise in the popularity of large language models has spurred the +development of extensive code datasets needed to train them. This has left +limited code available for collection and use in the downstream investigation +of specific behaviors, or evaluation of large language models without suffering +from data contamination. To address this problem, we release The Heap, a large +multilingual dataset covering 57 programming languages that has been +deduplicated with respect to other open datasets of code, enabling researchers +to conduct fair evaluations of large language models without significant data +cleaning overhead. + +**Decision Explanation:** The paper meets criteria 2 and 4 by providing a dataset for evaluating large language models and describing its methodology, and potentially meets criteria 3 by enabling fair comparisons with state-of-the-art techniques. + +--- + +## [Towards Large Reasoning Models: A Survey of Reinforced Reasoning with + Large Language Models](https://arxiv.org/abs/2501.09686v1) +**arXiv ID:** 2501.09686v1 + +**Abstract:** +> Language has long been conceived as an essential tool for human reasoning. +The breakthrough of Large Language Models (LLMs) has sparked significant +research interest in leveraging these models to tackle complex reasoning tasks. +Researchers have moved beyond simple autoregressive token generation by +introducing the concept of "thought" -- a sequence of tokens representing +intermediate steps in the reasoning process. This innovative paradigm enables +LLMs' to mimic complex human reasoning processes, such as tree search and +reflective thinking. Recently, an emerging trend of learning to reason has +applied reinforcement learning (RL) to train LLMs to master reasoning +processes. This approach enables the automatic generation of high-quality +reasoning trajectories through trial-and-error search algorithms, significantly +expanding LLMs' reasoning capacity by providing substantially more training +data. Furthermore, recent studies demonstrate that encouraging LLMs to "think" +with more tokens during test-time inference can further significantly boost +reasoning accuracy. Therefore, the train-time and test-time scaling combined to +show a new research frontier -- a path toward Large Reasoning Model. The +introduction of OpenAI's o1 series marks a significant milestone in this +research direction. In this survey, we present a comprehensive review of recent +progress in LLM reasoning. We begin by introducing the foundational background +of LLMs and then explore the key technical components driving the development +of large reasoning models, with a focus on automated data construction, +learning-to-reason techniques, and test-time scaling. We also analyze popular +open-source projects at building large reasoning models, and conclude with open +challenges and future research directions. + +**Decision Explanation:** The paper meets criteria 1, 3, and 4 by focusing on practical applications of Large Language Models in reasoning tasks, comparing with state-of-the-art techniques, and clearly describing methodology and implementation details, making it a strong candidate for further review. + +--- + +## [A Simple Aerial Detection Baseline of Multimodal Language Models](https://arxiv.org/abs/2501.09720v1) +**arXiv ID:** 2501.09720v1 + +**Abstract:** +> The multimodal language models (MLMs) based on generative pre-trained +Transformer are considered powerful candidates for unifying various domains and +tasks. MLMs developed for remote sensing (RS) have demonstrated outstanding +performance in multiple tasks, such as visual question answering and visual +grounding. In addition to visual grounding that detects specific objects +corresponded to given instruction, aerial detection, which detects all objects +of multiple categories, is also a valuable and challenging task for RS +foundation models. However, aerial detection has not been explored by existing +RS MLMs because the autoregressive prediction mechanism of MLMs differs +significantly from the detection outputs. In this paper, we present a simple +baseline for applying MLMs to aerial detection for the first time, named +LMMRotate. Specifically, we first introduce a normalization method to transform +detection outputs into textual outputs to be compatible with the MLM framework. +Then, we propose a evaluation method, which ensures a fair comparison between +MLMs and conventional object detection models. We construct the baseline by +fine-tuning open-source general-purpose MLMs and achieve impressive detection +performance comparable to conventional detector. We hope that this baseline +will serve as a reference for future MLM development, enabling more +comprehensive capabilities for understanding RS images. Code is available at +https://github.com/Li-Qingyun/mllm-mmrotate. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models in aerial detection, presenting experimental results with quantitative metrics, and comparing its results with state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [Knowledge prompt chaining for semantic modeling](https://arxiv.org/abs/2501.08540v1) +**arXiv ID:** 2501.08540v1 + +**Abstract:** +> The task of building semantics for structured data such as CSV, JSON, and XML +files is highly relevant in the knowledge representation field. Even though we +have a vast of structured data on the internet, mapping them to domain +ontologies to build semantics for them is still very challenging as it requires +the construction model to understand and learn graph-structured knowledge. +Otherwise, the task will require human beings' effort and cost. In this paper, +we proposed a novel automatic semantic modeling framework: Knowledge Prompt +Chaining. It can serialize the graph-structured knowledge and inject it into +the LLMs properly in a Prompt Chaining architecture. Through this knowledge +injection and prompting chaining, the model in our framework can learn the +structure information and latent space of the graph and generate the semantic +labels and semantic graphs following the chains' insturction naturally. Based +on experimental results, our method achieves better performance than existing +leading techniques, despite using reduced structured input data. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs in knowledge graphs, presenting experimental results with quantitative metrics, and comparing its results to state-of-the-art techniques, demonstrating performance improvements. + +--- + +## [LAMS: LLM-Driven Automatic Mode Switching for Assistive Teleoperation](https://arxiv.org/abs/2501.08558v1) +**arXiv ID:** 2501.08558v1 + +**Abstract:** +> Teleoperating high degrees-of-freedom (DoF) robotic manipulators via low-DoF +controllers like joysticks often requires frequent switching between control +modes, where each mode maps controller movements to specific robot actions. +Manually performing this frequent switching can make teleoperation cumbersome +and inefficient. On the other hand, existing automatic mode-switching +solutions, such as heuristic-based or learning-based methods, are often +task-specific and lack generalizability. In this paper, we introduce LLM-Driven +Automatic Mode Switching (LAMS), a novel approach that leverages Large Language +Models (LLMs) to automatically switch control modes based on task context. +Unlike existing methods, LAMS requires no prior task demonstrations and +incrementally improves by integrating user-generated mode-switching examples. +We validate LAMS through an ablation study and a user study with 10 +participants on complex, long-horizon tasks, demonstrating that LAMS +effectively reduces manual mode switches, is preferred over alternative +methods, and improves performance over time. The project website with +supplementary materials is at https://lams-assistance.github.io/. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on a practical application of LLMs in assistive teleoperation, including experimental results with quantitative metrics, and comparing its results with existing methods. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges. + +--- + +## [ANSR-DT: An Adaptive Neuro-Symbolic Learning and Reasoning Framework for + Digital Twins](https://arxiv.org/abs/2501.08561v1) +**arXiv ID:** 2501.08561v1 + +**Abstract:** +> In this paper, we propose an Adaptive Neuro-Symbolic Learning Framework for +digital twin technology called ``ANSR-DT." Our approach combines pattern +recognition algorithms with reinforcement learning and symbolic reasoning to +enable real-time learning and adaptive intelligence. This integration enhances +the understanding of the environment and promotes continuous learning, leading +to better and more effective decision-making in real-time for applications that +require human-machine collaboration. We evaluated the \textit{ANSR-DT} +framework for its ability to learn and adapt to dynamic patterns, observing +significant improvements in decision accuracy, reliability, and +interpretability when compared to existing state-of-the-art methods. However, +challenges still exist in extracting and integrating symbolic rules in complex +environments, which limits the full potential of our framework in heterogeneous +settings. Moreover, our ongoing research aims to address this issue in the +future by ensuring seamless integration of neural models at large. In addition, +our open-source implementation promotes reproducibility and encourages future +research to build on our foundational work. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of neuro-symbolic learning, presenting experimental results with quantitative metrics, and comparing its results to existing state-of-the-art methods. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges. + +--- + +## [RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation](https://arxiv.org/abs/2501.08617v1) +**arXiv ID:** 2501.08617v1 + +**Abstract:** +> Generative AI systems like foundation models (FMs) must align well with human +values to ensure their behavior is helpful and trustworthy. While Reinforcement +Learning from Human Feedback (RLHF) has shown promise for optimizing model +performance using human judgments, existing RLHF pipelines predominantly rely +on immediate feedback, which can fail to accurately reflect the downstream +impact of an interaction on users' utility. We demonstrate that feedback based +on evaluators' foresight estimates of downstream consequences systematically +induces Goodhart's Law dynamics, incentivizing misaligned behaviors like +sycophancy and deception and ultimately degrading user outcomes. To alleviate +this, we propose decoupling evaluation from prediction by refocusing RLHF on +hindsight feedback. Our theoretical analysis reveals that conditioning +evaluator feedback on downstream observations mitigates misalignment and +improves expected human utility, even when these observations are simulated by +the AI system itself. To leverage this insight in a practical alignment +algorithm, we introduce Reinforcement Learning from Hindsight Simulation +(RLHS), which first simulates plausible consequences and then elicits feedback +to assess what behaviors were genuinely beneficial in hindsight. We apply RLHS +to two widely-employed online and offline preference optimization methods -- +Proximal Policy Optimization (PPO) and Direct Preference Optimization (DPO) -- +and show empirically that misalignment is significantly reduced with both +methods. Through an online human user study, we show that RLHS consistently +outperforms RLHF in helping users achieve their goals and earns higher +satisfaction ratings, despite being trained solely with simulated hindsight +feedback. These results underscore the importance of focusing on long-term +consequences, even simulated ones, to mitigate misalignment in RLHF. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models, including experimental results with quantitative metrics, and comparing its results to state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges. + +--- + +## [Leveraging LLM Agents for Translating Network Configurations](https://arxiv.org/abs/2501.08760v1) +**arXiv ID:** 2501.08760v1 + +**Abstract:** +> Configuration translation is a critical and frequent task in network +operations. When a network device is damaged or outdated, administrators need +to replace it to maintain service continuity. The replacement devices may +originate from different vendors, necessitating configuration translation to +ensure seamless network operation. However, translating configurations manually +is a labor-intensive and error-prone process. In this paper, we propose an +intent-based framework for translating network configuration with Large +Language Model (LLM) Agents. The core of our approach is an Intent-based +Retrieval Augmented Generation (IRAG) module that systematically splits a +configuration file into fragments, extracts intents, and generates accurate +translations. We also design a two-stage verification method to validate the +syntax and semantics correctness of the translated configurations. We implement +and evaluate the proposed method on real-world network configurations. +Experimental results show that our method achieves 97.74% syntax correctness, +outperforming state-of-the-art methods in translation accuracy. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on a practical application of LLMs in network configuration translation, presenting experimental results with quantitative metrics, and comparing its results to state-of-the-art methods, demonstrating advancements in translation accuracy. + +--- + +## [MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents](https://arxiv.org/abs/2501.08828v1) +**arXiv ID:** 2501.08828v1 + +**Abstract:** +> Multi-modal document retrieval is designed to identify and retrieve various +forms of multi-modal content, such as figures, tables, charts, and layout +information from extensive documents. Despite its significance, there is a +notable lack of a robust benchmark to effectively evaluate the performance of +systems in multi-modal document retrieval. To address this gap, this work +introduces a new benchmark, named as MMDocIR, encompassing two distinct tasks: +page-level and layout-level retrieval. The former focuses on localizing the +most relevant pages within a long document, while the latter targets the +detection of specific layouts, offering a more fine-grained granularity than +whole-page analysis. A layout can refer to a variety of elements such as +textual paragraphs, equations, figures, tables, or charts. The MMDocIR +benchmark comprises a rich dataset featuring expertly annotated labels for +1,685 questions and bootstrapped labels for 173,843 questions, making it a +pivotal resource for advancing multi-modal document retrieval for both training +and evaluation. Through rigorous experiments, we reveal that (i) visual +retrievers significantly outperform their text counterparts, (ii) MMDocIR train +set can effectively benefit the training process of multi-modal document +retrieval and (iii) text retrievers leveraging on VLM-text perform much better +than those using OCR-text. These findings underscores the potential advantages +of integrating visual elements for multi-modal document retrieval. + +**Decision Explanation:** The paper meets criteria 2 (Experimental Results and Quantitative Metrics) and 3 (Comparison with State-of-the-Art), and potentially meets criteria 1 (Practical Applications) and 4 (Methodology and Implementation Details), as it introduces a new benchmark for multi-modal document retrieval and presents experimental results with quantitative metrics. + +--- + +## [Disentangling Exploration of Large Language Models by Optimal + Exploitation](https://arxiv.org/abs/2501.08925v1) +**arXiv ID:** 2501.08925v1 + +**Abstract:** +> Exploration is a crucial skill for self-improvement and open-ended +problem-solving. However, it remains uncertain whether large language models +can effectively explore the state-space. Existing evaluations predominantly +focus on the trade-off between exploration and exploitation, often assessed in +multi-armed bandit problems. In contrast, this work isolates exploration as the +sole objective, tasking the agent with delivering information that enhances +future returns. For the evaluation, we propose to decompose missing rewards +into exploration and exploitation components by measuring the optimal +achievable return for the states already explored. Our experiments with various +LLMs reveal that most models struggle to sufficiently explore the state-space +and that weak exploration is insufficient. We observe a positive correlation +between model size and exploration performance, with larger models +demonstrating superior capabilities. Furthermore, we show that our +decomposition provides insights into differences in behaviors driven by agent +instructions during prompt engineering, offering a valuable tool for refining +LLM performance in exploratory tasks. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models, including experimental results with quantitative metrics, and comparing its results with existing state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges involving LLMs. + +--- + +## [Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG](https://arxiv.org/abs/2501.09136v1) +**arXiv ID:** 2501.09136v1 + +**Abstract:** +> Large Language Models (LLMs) have revolutionized artificial intelligence (AI) +by enabling human like text generation and natural language understanding. +However, their reliance on static training data limits their ability to respond +to dynamic, real time queries, resulting in outdated or inaccurate outputs. +Retrieval Augmented Generation (RAG) has emerged as a solution, enhancing LLMs +by integrating real time data retrieval to provide contextually relevant and +up-to-date responses. Despite its promise, traditional RAG systems are +constrained by static workflows and lack the adaptability required for +multistep reasoning and complex task management. + Agentic Retrieval-Augmented Generation (Agentic RAG) transcends these +limitations by embedding autonomous AI agents into the RAG pipeline. These +agents leverage agentic design patterns reflection, planning, tool use, and +multiagent collaboration to dynamically manage retrieval strategies, +iteratively refine contextual understanding, and adapt workflows to meet +complex task requirements. This integration enables Agentic RAG systems to +deliver unparalleled flexibility, scalability, and context awareness across +diverse applications. + This survey provides a comprehensive exploration of Agentic RAG, beginning +with its foundational principles and the evolution of RAG paradigms. It +presents a detailed taxonomy of Agentic RAG architectures, highlights key +applications in industries such as healthcare, finance, and education, and +examines practical implementation strategies. Additionally, it addresses +challenges in scaling these systems, ensuring ethical decision making, and +optimizing performance for real-world applications, while providing detailed +insights into frameworks and tools for implementing Agentic RAG + +**Decision Explanation:** The paper meets criteria 1, 3, 4, and 5, and discusses real-world applications and challenges involving LLMs, particularly in areas like retrieval-augmented generation and agentic AI, with a focus on practical implementation strategies and scalability. + +--- + +## [AutoLoop: Fast Visual SLAM Fine-tuning through Agentic Curriculum + Learning](https://arxiv.org/abs/2501.09160v1) +**arXiv ID:** 2501.09160v1 + +**Abstract:** +> Current visual SLAM systems face significant challenges in balancing +computational efficiency with robust loop closure handling. Traditional +approaches require careful manual tuning and incur substantial computational +overhead, while learning-based methods either lack explicit loop closure +capabilities or implement them through computationally expensive methods. We +present AutoLoop, a novel approach that combines automated curriculum learning +with efficient fine-tuning for visual SLAM systems. Our method employs a DDPG +(Deep Deterministic Policy Gradient) agent to dynamically adjust loop closure +weights during training, eliminating the need for manual hyperparameter search +while significantly reducing the required training steps. The approach +pre-computes potential loop closure pairs offline and leverages them through an +agent-guided curriculum, allowing the model to adapt efficiently to new +scenarios. Experiments conducted on TartanAir for training and validated across +multiple benchmarks including KITTI, EuRoC, ICL-NUIM and TUM RGB-D demonstrate +that AutoLoop achieves comparable or superior performance while reducing +training time by an order of magnitude compared to traditional approaches. +AutoLoop provides a practical solution for rapid adaptation of visual SLAM +systems, automating the weight tuning process that traditionally requires +multiple manual iterations. Our results show that this automated curriculum +strategy not only accelerates training but also maintains or improves the +model's performance across diverse environmental conditions. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of automated curriculum learning for visual SLAM systems, including experimental results with quantitative metrics and comparison with state-of-the-art techniques, demonstrating significant improvements in training time and performance. + +--- + +## [Clone-Robust AI Alignment](https://arxiv.org/abs/2501.09254v1) +**arXiv ID:** 2501.09254v1 + +**Abstract:** +> A key challenge in training Large Language Models (LLMs) is properly aligning +them with human preferences. Reinforcement Learning with Human Feedback (RLHF) +uses pairwise comparisons from human annotators to train reward functions and +has emerged as a popular alignment method. However, input datasets in RLHF are +not necessarily balanced in the types of questions and answers that are +included. Therefore, we want RLHF algorithms to perform well even when the set +of alternatives is not uniformly distributed. Drawing on insights from social +choice theory, we introduce robustness to approximate clones, a desirable +property of RLHF algorithms which requires that adding near-duplicate +alternatives does not significantly change the learned reward function. We +first demonstrate that the standard RLHF algorithm based on regularized maximum +likelihood estimation (MLE) fails to satisfy this property. We then propose the +weighted MLE, a new RLHF algorithm that modifies the standard regularized MLE +by weighting alternatives based on their similarity to other alternatives. This +new algorithm guarantees robustness to approximate clones while preserving +desirable theoretical properties. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models, including experimental results with quantitative metrics and comparing its results with existing state-of-the-art techniques. It also introduces a novel approach to reinforcement learning with human feedback, making it relevant for further review. + +--- + +## [To Retrieve or Not to Retrieve? Uncertainty Detection for Dynamic + Retrieval Augmented Generation](https://arxiv.org/abs/2501.09292v1) +**arXiv ID:** 2501.09292v1 + +**Abstract:** +> Retrieval-Augmented Generation equips large language models with the +capability to retrieve external knowledge, thereby mitigating hallucinations by +incorporating information beyond the model's intrinsic abilities. However, most +prior works have focused on invoking retrieval deterministically, which makes +it unsuitable for tasks such as long-form question answering. Instead, +dynamically performing retrieval by invoking it only when the underlying LLM +lacks the required knowledge can be more efficient. In this context, we delve +deeper into the question, "To Retrieve or Not to Retrieve?" by exploring +multiple uncertainty detection methods. We evaluate these methods for the task +of long-form question answering, employing dynamic retrieval, and present our +comparisons. Our findings suggest that uncertainty detection metrics, such as +Degree Matrix Jaccard and Eccentricity, can reduce the number of retrieval +calls by almost half, with only a slight reduction in question-answering +accuracy. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models in retrieval-augmented generation, presenting experimental results with quantitative metrics, and comparing its results with existing state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +## [A Study of In-Context-Learning-Based Text-to-SQL Errors](https://arxiv.org/abs/2501.09310v1) +**arXiv ID:** 2501.09310v1 + +**Abstract:** +> Large language models (LLMs) have been adopted to perform text-to-SQL tasks, +utilizing their in-context learning (ICL) capability to translate natural +language questions into structured query language (SQL). However, such a +technique faces correctness problems and requires efficient repairing +solutions. In this paper, we conduct the first comprehensive study of +text-to-SQL errors. Our study covers four representative ICL-based techniques, +five basic repairing methods, two benchmarks, and two LLM settings. We find +that text-to-SQL errors are widespread and summarize 29 error types of 7 +categories. We also find that existing repairing attempts have limited +correctness improvement at the cost of high computational overhead with many +mis-repairs. Based on the findings, we propose MapleRepair, a novel text-to-SQL +error detection and repairing framework. The evaluation demonstrates that +MapleRepair outperforms existing solutions by repairing 13.8% more queries with +neglectable mis-repairs and 67.4% less overhead. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models in text-to-SQL tasks, including experimental results with quantitative metrics, and comparing its results with existing state-of-the-art techniques, demonstrating performance improvements. + +--- + +## [Rational Tuning of LLM Cascades via Probabilistic Modeling](https://arxiv.org/abs/2501.09345v1) +**arXiv ID:** 2501.09345v1 + +**Abstract:** +> Understanding the reliability of large language models (LLMs) has recently +garnered significant attention. Given LLMs' propensity to hallucinate, as well +as their high sensitivity to prompt design, it is already challenging to +predict the performance of an individual LLM. However, the problem becomes more +complex for compound LLM systems such as cascades, where in addition to each +model's standalone performance, we must understand how the error rates of +different models interact. In this paper, we present a probabilistic model for +the joint performance distribution of a sequence of LLMs, which enables a +framework for rationally tuning the confidence thresholds of a LLM cascade +using continuous optimization. Compared to selecting confidence thresholds +using grid search, our parametric Markov-copula model significantly improves +runtime scaling with respect to the length of the cascade and the desired +resolution of the cost-error curve, turning them from intractable into +low-order polynomial. In addition, the optimal thresholds computed using our +continuous optimization-based algorithm increasingly outperform those found via +grid search as cascade length grows, improving the area under the cost-error +curve by 1.9% on average for cascades consisting of at least three models. +Overall, our Markov-copula model provides a rational basis for tuning LLM +cascade performance and points to the potential of probabilistic methods in +analyzing LLM systems. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs, including experimental results with quantitative metrics, and comparing its results to existing state-of-the-art techniques, demonstrating performance improvements. + +--- + +## [YETI (YET to Intervene) Proactive Interventions by Multimodal AI Agents + in Augmented Reality Tasks](https://arxiv.org/abs/2501.09355v1) +**arXiv ID:** 2501.09355v1 + +**Abstract:** +> Multimodal AI Agents are AI models that have the capability of interactively +and cooperatively assisting human users to solve day-to-day tasks. Augmented +Reality (AR) head worn devices can uniquely improve the user experience of +solving procedural day-to-day tasks by providing egocentric multimodal (audio +and video) observational capabilities to AI Agents. Such AR capabilities can +help AI Agents see and listen to actions that users take which can relate to +multimodal capabilities of human users. Existing AI Agents, either Large +Language Models (LLMs) or Multimodal Vision-Language Models (VLMs) are reactive +in nature, which means that models cannot take an action without reading or +listening to the human user's prompts. Proactivity of AI Agents on the other +hand can help the human user detect and correct any mistakes in agent observed +tasks, encourage users when they do tasks correctly or simply engage in +conversation with the user - akin to a human teaching or assisting a user. Our +proposed YET to Intervene (YETI) multimodal agent focuses on the research +question of identifying circumstances that may require the agent to intervene +proactively. This allows the agent to understand when it can intervene in a +conversation with human users that can help the user correct mistakes on tasks, +like cooking, using AR. Our YETI Agent learns scene understanding signals based +on interpretable notions of Structural Similarity (SSIM) on consecutive video +frames. We also define the alignment signal which the AI Agent can learn to +identify if the video frames corresponding to the user's actions on the task +are consistent with expected actions. These signals are used by our AI Agent to +determine when it should proactively intervene. We compare our results on the +instances of proactive intervention in the HoloAssist multimodal benchmark for +an expert agent guiding a user to complete procedural tasks. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models in augmented reality tasks, including experimental results with quantitative metrics and comparison with state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges. + +--- + +## [MoE$^2$: Optimizing Collaborative Inference for Edge Large Language + Models](https://arxiv.org/abs/2501.09410v1) +**arXiv ID:** 2501.09410v1 + +**Abstract:** +> Large language models (LLMs) have demonstrated remarkable capabilities across +a wide range of natural language processing tasks. Exploiting the heterogeneous +capabilities of edge LLMs is crucial for diverse emerging applications, as it +enables greater cost-effectiveness and reduced latency. In this work, we +introduce \textit{Mixture-of-Edge-Experts (MoE$^2$)}, a novel collaborative +inference framework for edge LLMs. We formulate the joint gating and expert +selection problem to optimize inference performance under energy and latency +constraints. Unlike conventional MoE problems, LLM expert selection is +significantly more challenging due to the combinatorial nature and the +heterogeneity of edge LLMs across various attributes. To this end, we propose a +two-level expert selection mechanism through which we uncover an +optimality-preserving property of gating parameters across expert selections. +This property enables the decomposition of the training and selection +processes, significantly reducing complexity. Furthermore, we leverage the +objective's monotonicity and design a discrete monotonic optimization algorithm +for optimal expert selection. We implement edge servers with NVIDIA Jetson AGX +Orins and NVIDIA RTX 4090 GPUs, and perform extensive experiments. Our results +validate that performance improvements of various LLM models and show that our +MoE$^2$ method can achieve optimal trade-offs among different delay and energy +budgets, and outperforms baselines under various system resource constraints. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of Large Language Models, including experimental results with quantitative metrics, and comparing its results with state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. Additionally, the paper introduces a novel approach and demonstrates robust experimental validation. + +--- + +## [CarMem: Enhancing Long-Term Memory in LLM Voice Assistants through + Category-Bounding](https://arxiv.org/abs/2501.09645v1) +**arXiv ID:** 2501.09645v1 + +**Abstract:** +> In today's assistant landscape, personalisation enhances interactions, +fosters long-term relationships, and deepens engagement. However, many systems +struggle with retaining user preferences, leading to repetitive user requests +and disengagement. Furthermore, the unregulated and opaque extraction of user +preferences in industry applications raises significant concerns about privacy +and trust, especially in regions with stringent regulations like Europe. In +response to these challenges, we propose a long-term memory system for voice +assistants, structured around predefined categories. This approach leverages +Large Language Models to efficiently extract, store, and retrieve preferences +within these categories, ensuring both personalisation and transparency. We +also introduce a synthetic multi-turn, multi-session conversation dataset +(CarMem), grounded in real industry data, tailored to an in-car voice assistant +setting. Benchmarked on the dataset, our system achieves an F1-score of .78 to +.95 in preference extraction, depending on category granularity. Our +maintenance strategy reduces redundant preferences by 95% and contradictory +ones by 92%, while the accuracy of optimal retrieval is at .87. Collectively, +the results demonstrate the system's suitability for industrial applications. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of LLMs in voice assistants, providing experimental results with quantitative metrics, and comparing its results to existing techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications and challenges. + +--- + +## [Authenticated Delegation and Authorized AI Agents](https://arxiv.org/abs/2501.09674v1) +**arXiv ID:** 2501.09674v1 + +**Abstract:** +> The rapid deployment of autonomous AI agents creates urgent challenges around +authorization, accountability, and access control in digital spaces. New +standards are needed to know whom AI agents act on behalf of and guide their +use appropriately, protecting online spaces while unlocking the value of task +delegation to autonomous agents. We introduce a novel framework for +authenticated, authorized, and auditable delegation of authority to AI agents, +where human users can securely delegate and restrict the permissions and scope +of agents while maintaining clear chains of accountability. This framework +builds on existing identification and access management protocols, extending +OAuth 2.0 and OpenID Connect with agent-specific credentials and metadata, +maintaining compatibility with established authentication and web +infrastructure. Further, we propose a framework for translating flexible, +natural language permissions into auditable access control configurations, +enabling robust scoping of AI agent capabilities across diverse interaction +modalities. Taken together, this practical approach facilitates immediate +deployment of AI agents while addressing key security and accountability +concerns, working toward ensuring agentic AI systems perform only appropriate +actions and providing a tool for digital service providers to enable AI agent +interactions without risking harm from scalable interaction. + +**Decision Explanation:** The paper meets criteria 1, 3, and 5 by focusing on practical applications of agentic AI, comparing its framework with existing protocols, and discussing real-world applications and challenges. It also introduces a novel approach to authenticated delegation and authorized AI agents, demonstrating potential for innovative applications and performance improvements. + +--- + +## [ADAGE: A generic two-layer framework for adaptive agent based modelling](https://arxiv.org/abs/2501.09429v1) +**arXiv ID:** 2501.09429v1 + +**Abstract:** +> Agent-based models (ABMs) are valuable for modelling complex, potentially +out-of-equilibria scenarios. However, ABMs have long suffered from the Lucas +critique, stating that agent behaviour should adapt to environmental changes. +Furthermore, the environment itself often adapts to these behavioural changes, +creating a complex bi-level adaptation problem. Recent progress integrating +multi-agent reinforcement learning into ABMs introduces adaptive agent +behaviour, beginning to address the first part of this critique, however, the +approaches are still relatively ad hoc, lacking a general formulation, and +furthermore, do not tackle the second aspect of simultaneously adapting +environmental level characteristics in addition to the agent behaviours. In +this work, we develop a generic two-layer framework for ADaptive AGEnt based +modelling (ADAGE) for addressing these problems. This framework formalises the +bi-level problem as a Stackelberg game with conditional behavioural policies, +providing a consolidated framework for adaptive agent-based modelling based on +solving a coupled set of non-linear equations. We demonstrate how this generic +approach encapsulates several common (previously viewed as distinct) ABM tasks, +such as policy design, calibration, scenario generation, and robust behavioural +learning under one unified framework. We provide example simulations on +multiple complex economic and financial environments, showing the strength of +the novel framework under these canonical settings, addressing long-standing +critiques of traditional ABMs. + +**Decision Explanation:** The paper meets criteria 1, 2, and 3 by focusing on practical applications of adaptive agent-based modeling, demonstrating experimental results with quantitative metrics, and comparing its results with existing state-of-the-art techniques. It also meets criteria 4 and 5 by clearly describing its methodology and discussing real-world applications. + +--- + +# Rejected Papers + +## [DualOpt: A Dual Divide-and-Optimize Algorithm for the Large-scale + Traveling Salesman Problem](https://arxiv.org/abs/2501.08565v1) +**arXiv ID:** 2501.08565v1 + +**Abstract:** +> This paper proposes a dual divide-and-optimize algorithm (DualOpt) for +solving the large-scale traveling salesman problem (TSP). DualOpt combines two +complementary strategies to improve both solution quality and computational +efficiency. The first strategy is a grid-based divide-and-conquer procedure +that partitions the TSP into smaller sub-problems, solving them in parallel and +iteratively refining the solution by merging nodes and partial routes. The +process continues until only one grid remains, yielding a high-quality initial +solution. The second strategy involves a path-based divide-and-optimize +procedure that further optimizes the solution by dividing it into sub-paths, +optimizing each using a neural solver, and merging them back to progressively +improve the overall solution. Extensive experiments conducted on two groups of +TSP benchmark instances, including randomly generated instances with up to +100,000 nodes and real-world datasets from TSPLIB, demonstrate the +effectiveness of DualOpt. The proposed DualOpt achieves highly competitive +results compared to 10 state-of-the-art algorithms in the literature. In +particular, DualOpt achieves an improvement gap up to 1.40% for the largest +instance TSP100K with a remarkable 104x speed-up over the leading heuristic +solver LKH3. Additionally, DualOpt demonstrates strong generalization on TSPLIB +benchmarks, confirming its capability to tackle diverse real-world TSP +applications. + +**Decision Explanation:** The paper does not meet the criteria as it does not focus on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, but rather proposes an algorithm for the traveling salesman problem. + +--- + +## [Development and Validation of the Provider Documentation Summarization + Quality Instrument for Large Language Models](https://arxiv.org/abs/2501.08977v1) +**arXiv ID:** 2501.08977v1 + +**Abstract:** +> As Large Language Models (LLMs) are integrated into electronic health record +(EHR) workflows, validated instruments are essential to evaluate their +performance before implementation. Existing instruments for provider +documentation quality are often unsuitable for the complexities of +LLM-generated text and lack validation on real-world data. The Provider +Documentation Summarization Quality Instrument (PDSQI-9) was developed to +evaluate LLM-generated clinical summaries. Multi-document summaries were +generated from real-world EHR data across multiple specialties using several +LLMs (GPT-4o, Mixtral 8x7b, and Llama 3-8b). Validation included Pearson +correlation for substantive validity, factor analysis and Cronbach's alpha for +structural validity, inter-rater reliability (ICC and Krippendorff's alpha) for +generalizability, a semi-Delphi process for content validity, and comparisons +of high- versus low-quality summaries for discriminant validity. Seven +physician raters evaluated 779 summaries and answered 8,329 questions, +achieving over 80% power for inter-rater reliability. The PDSQI-9 demonstrated +strong internal consistency (Cronbach's alpha = 0.879; 95% CI: 0.867-0.891) and +high inter-rater reliability (ICC = 0.867; 95% CI: 0.867-0.868), supporting +structural validity and generalizability. Factor analysis identified a 4-factor +model explaining 58% of the variance, representing organization, clarity, +accuracy, and utility. Substantive validity was supported by correlations +between note length and scores for Succinct (rho = -0.200, p = 0.029) and +Organized (rho = -0.190, p = 0.037). Discriminant validity distinguished high- +from low-quality summaries (p < 0.001). The PDSQI-9 demonstrates robust +construct validity, supporting its use in clinical practice to evaluate +LLM-generated summaries and facilitate safer integration of LLMs into +healthcare workflows. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is excluded according to the criteria. + +--- + +## [AI-based Identity Fraud Detection: A Systematic Review](https://arxiv.org/abs/2501.09239v1) +**arXiv ID:** 2501.09239v1 + +**Abstract:** +> With the rapid development of digital services, a large volume of personally +identifiable information (PII) is stored online and is subject to cyberattacks +such as Identity fraud. Most recently, the use of Artificial Intelligence (AI) +enabled deep fake technologies has significantly increased the complexity of +identity fraud. Fraudsters may use these technologies to create highly +sophisticated counterfeit personal identification documents, photos and videos. +These advancements in the identity fraud landscape pose challenges for identity +fraud detection and society at large. There is a pressing need to review and +understand identity fraud detection methods, their limitations and potential +solutions. This research aims to address this important need by using the +well-known systematic literature review method. This paper reviewed a selected +set of 43 papers across 4 major academic literature databases. In particular, +the review results highlight the two types of identity fraud prevention and +detection methods, in-depth and open challenges. The results were also +consolidated into a taxonomy of AI-based identity fraud detection and +prevention methods including key insights and trends. Overall, this paper +provides a foundational knowledge base to researchers and practitioners for +further research and development in this important area of digital identity +fraud. + +**Decision Explanation:** The paper primarily focuses on identity fraud detection, which is related to law and social applications of AI, and does not meet the criteria of focusing on practical applications of Large Language Models (LLMs) or agentic AI. + +--- + +## [AI in Support of Diversity and Inclusion](https://arxiv.org/abs/2501.09534v1) +**arXiv ID:** 2501.09534v1 + +**Abstract:** +> In this paper, we elaborate on how AI can support diversity and inclusion and +exemplify research projects conducted in that direction. We start by looking at +the challenges and progress in making large language models (LLMs) more +transparent, inclusive, and aware of social biases. Even though LLMs like +ChatGPT have impressive abilities, they struggle to understand different +cultural contexts and engage in meaningful, human like conversations. A key +issue is that biases in language processing, especially in machine translation, +can reinforce inequality. Tackling these biases requires a multidisciplinary +approach to ensure AI promotes diversity, fairness, and inclusion. We also +highlight AI's role in identifying biased content in media, which is important +for improving representation. By detecting unequal portrayals of social groups, +AI can help challenge stereotypes and create more inclusive technologies. +Transparent AI algorithms, which clearly explain their decisions, are essential +for building trust and reducing bias in AI systems. We also stress AI systems +need diverse and inclusive training data. Projects like the Child Growth +Monitor show how using a wide range of data can help address real world +problems like malnutrition and poverty. We present a project that demonstrates +how AI can be applied to monitor the role of search engines in spreading +disinformation about the LGBTQ+ community. Moreover, we discuss the SignON +project as an example of how technology can bridge communication gaps between +hearing and deaf people, emphasizing the importance of collaboration and mutual +trust in developing inclusive AI. Overall, with this paper, we advocate for AI +systems that are not only effective but also socially responsible, promoting +fair and inclusive interactions between humans and machines. + +**Decision Explanation:** The paper primarily focuses on social applications of AI in regard to Diversity, Social harm, and similar issues, which is an excluded topic. + +--- + +## [Artificial Intelligence-Driven Clinical Decision Support Systems](https://arxiv.org/abs/2501.09628v1) +**arXiv ID:** 2501.09628v1 + +**Abstract:** +> As artificial intelligence (AI) becomes increasingly embedded in healthcare +delivery, this chapter explores the critical aspects of developing reliable and +ethical Clinical Decision Support Systems (CDSS). Beginning with the +fundamental transition from traditional statistical models to sophisticated +machine learning approaches, this work examines rigorous validation strategies +and performance assessment methods, including the crucial role of model +calibration and decision curve analysis. The chapter emphasizes that creating +trustworthy AI systems in healthcare requires more than just technical +accuracy; it demands careful consideration of fairness, explainability, and +privacy. The challenge of ensuring equitable healthcare delivery through AI is +stressed, discussing methods to identify and mitigate bias in clinical +predictive models. The chapter then delves into explainability as a cornerstone +of human-centered CDSS. This focus reflects the understanding that healthcare +professionals must not only trust AI recommendations but also comprehend their +underlying reasoning. The discussion advances in an analysis of privacy +vulnerabilities in medical AI systems, from data leakage in deep learning +models to sophisticated attacks against model explanations. The text explores +privacy-preservation strategies such as differential privacy and federated +learning, while acknowledging the inherent trade-offs between privacy +protection and model performance. This progression, from technical validation +to ethical considerations, reflects the multifaceted challenges of developing +AI systems that can be seamlessly and reliably integrated into daily clinical +practice while maintaining the highest standards of patient care and data +protection. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is excluded according to the criteria. + +--- + +## [Electronic Health Records: Towards Digital Twins in Healthcare](https://arxiv.org/abs/2501.09640v1) +**arXiv ID:** 2501.09640v1 + +**Abstract:** +> The pivotal shift from traditional paper-based records to sophisticated +Electronic Health Records (EHR), enabled systematic collection and analysis of +patient data through descriptive statistics, providing insight into patterns +and trends across patient populations. This evolution continued toward +predictive analytics, allowing healthcare providers to anticipate patient +outcomes and potential complications before they occur. This progression from +basic digital record-keeping to sophisticated predictive modelling and digital +twins reflects healthcare's broader evolution toward more integrated, +patient-centred approaches that combine data-driven insights with personalized +care delivery. This chapter explores the evolution and significance of +healthcare information systems, beginning with an examination of the +implementation of EHR in the UK and the USA. It provides a comprehensive +overview of the International Classification of Diseases (ICD) system, tracing +its development from ICD-9 to ICD-10. Central to this discussion is the +MIMIC-III database, a landmark achievement in healthcare data sharing and +arguably the most comprehensive critical care database freely available to +researchers worldwide. MIMIC-III has democratized access to high-quality +healthcare data, enabling unprecedented opportunities for research and +analysis. The chapter examines its structure, clinical outcome analysis +capabilities, and practical applications through case studies, with a +particular focus on mortality and length of stay metrics, vital signs +extraction, and ICD coding. Through detailed entity-relationship diagrams and +practical examples, the text illustrates MIMIC's complex data structure and +demonstrates how different querying approaches can lead to subtly different +results, emphasizing the critical importance of understanding the database's +architecture for accurate data extraction. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, specifically Electronic Health Records, which is excluded according to the criteria. + +--- + +## [The Goofus & Gallant Story Corpus for Practical Value Alignment](https://arxiv.org/abs/2501.09707v1) +**arXiv ID:** 2501.09707v1 + +**Abstract:** +> Values or principles are key elements of human society that influence people +to behave and function according to an accepted standard set of social rules to +maintain social order. As AI systems are becoming ubiquitous in human society, +it is a major concern that they could violate these norms or values and +potentially cause harm. Thus, to prevent intentional or unintentional harm, AI +systems are expected to take actions that align with these principles. Training +systems to exhibit this type of behavior is difficult and often requires a +specialized dataset. This work presents a multi-modal dataset illustrating +normative and non-normative behavior in real-life situations described through +natural language and artistic images. This training set contains curated sets +of images that are designed to teach young children about social principles. We +argue that this is an ideal dataset to use for training socially normative +agents given this fact. + +**Decision Explanation:** The paper primarily focuses on social applications of AI in regard to value alignment and social norms, which is not aligned with the specified criteria. + +--- + +## [KU AIGEN ICL EDI@BC8 Track 3: Advancing Phenotype Named Entity + Recognition and Normalization for Dysmorphology Physical Examination Reports](https://arxiv.org/abs/2501.09744v1) +**arXiv ID:** 2501.09744v1 + +**Abstract:** +> The objective of BioCreative8 Track 3 is to extract phenotypic key medical +findings embedded within EHR texts and subsequently normalize these findings to +their Human Phenotype Ontology (HPO) terms. However, the presence of diverse +surface forms in phenotypic findings makes it challenging to accurately +normalize them to the correct HPO terms. To address this challenge, we explored +various models for named entity recognition and implemented data augmentation +techniques such as synonym marginalization to enhance the normalization step. +Our pipeline resulted in an exact extraction and normalization F1 score 2.6\% +higher than the mean score of all submissions received in response to the +challenge. Furthermore, in terms of the normalization F1 score, our approach +surpassed the average performance by 1.9\%. These findings contribute to the +advancement of automated medical data extraction and normalization techniques, +showcasing potential pathways for future research and application in the +biomedical domain. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is excluded according to the criteria. + +--- + +## [Adapting Whisper for Regional Dialects: Enhancing Public Services for + Vulnerable Populations in the United Kingdom](https://arxiv.org/abs/2501.08502v1) +**arXiv ID:** 2501.08502v1 + +**Abstract:** +> We collect novel data in the public service domain to evaluate the capability +of the state-of-the-art automatic speech recognition (ASR) models in capturing +regional differences in accents in the United Kingdom (UK), specifically +focusing on two accents from Scotland with distinct dialects. This study +addresses real-world problems where biased ASR models can lead to +miscommunication in public services, disadvantaging individuals with regional +accents particularly those in vulnerable populations. We first examine the +out-of-the-box performance of the Whisper large-v3 model on a baseline dataset +and our data. We then explore the impact of fine-tuning Whisper on the +performance in the two UK regions and investigate the effectiveness of existing +model evaluation techniques for our real-world application through manual +inspection of model errors. We observe that the Whisper model has a higher word +error rate (WER) on our test datasets compared to the baseline data and +fine-tuning on a given data improves performance on the test dataset with the +same domain and accent. The fine-tuned models also appear to show improved +performance when applied to the test data outside of the region it was trained +on suggesting that fine-tuned models may be transferable within parts of the +UK. Our manual analysis of model outputs reveals the benefits and drawbacks of +using WER as an evaluation metric and fine-tuning to adapt to regional +dialects. + +**Decision Explanation:** The paper primarily focuses on automatic speech recognition and regional dialects, which does not meet the criteria for practical applications of Large Language Models, and does not demonstrate a clear connection to knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Mitigating Domain Shift in Federated Learning via Intra- and + Inter-Domain Prototypes](https://arxiv.org/abs/2501.08521v1) +**arXiv ID:** 2501.08521v1 + +**Abstract:** +> Federated Learning (FL) has emerged as a decentralized machine learning +technique, allowing clients to train a global model collaboratively without +sharing private data. However, most FL studies ignore the crucial challenge of +heterogeneous domains where each client has a distinct feature distribution, +which is common in real-world scenarios. Prototype learning, which leverages +the mean feature vectors within the same classes, has become a prominent +solution for federated learning under domain skew. However, existing federated +prototype learning methods only consider inter-domain prototypes on the server +and overlook intra-domain characteristics. In this work, we introduce a novel +federated prototype learning method, namely I$^2$PFL, which incorporates +$\textbf{I}$ntra-domain and $\textbf{I}$nter-domain $\textbf{P}$rototypes, to +mitigate domain shifts and learn a generalized global model across multiple +domains in federated learning. To construct intra-domain prototypes, we propose +feature alignment with MixUp-based augmented prototypes to capture the +diversity of local domains and enhance the generalization of local features. +Additionally, we introduce a reweighting mechanism for inter-domain prototypes +to generate generalized prototypes to provide inter-domain knowledge and reduce +domain skew across multiple clients. Extensive experiments on the Digits, +Office-10, and PACS datasets illustrate the superior performance of our method +compared to other baselines. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on federated learning and domain shift mitigation, without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Dynamic Portfolio Optimization via Augmented DDPG with Quantum Price + Levels-Based Trading Strategy](https://arxiv.org/abs/2501.08528v1) +**arXiv ID:** 2501.08528v1 + +**Abstract:** +> With the development of deep learning, Dynamic Portfolio Optimization (DPO) +problem has received a lot of attention in recent years, not only in the field +of finance but also in the field of deep learning. Some advanced research in +recent years has proposed the application of Deep Reinforcement Learning (DRL) +to the DPO problem, which demonstrated to be more advantageous than supervised +learning in solving the DPO problem. However, there are still certain unsolved +issues: 1) DRL algorithms usually have the problems of slow learning speed and +high sample complexity, which is especially problematic when dealing with +complex financial data. 2) researchers use DRL simply for the purpose of +obtaining high returns, but pay little attention to the problem of risk control +and trading strategy, which will affect the stability of model returns. In +order to address these issues, in this study we revamped the intrinsic +structure of the model based on the Deep Deterministic Policy Gradient (DDPG) +and proposed the Augmented DDPG model. Besides, we also proposed an innovative +risk control strategy based on Quantum Price Levels (QPLs) derived from Quantum +Finance Theory (QFT). Our experimental results revealed that our model has +better profitability as well as risk control ability with less sample +complexity in the DPO problem compared to the baseline models. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on financial applications and does not explicitly mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the key areas of interest. + +--- + +## [The Devil is in Temporal Token: High Quality Video Reasoning + Segmentation](https://arxiv.org/abs/2501.08549v1) +**arXiv ID:** 2501.08549v1 + +**Abstract:** +> Existing methods for Video Reasoning Segmentation rely heavily on a single +special token to represent the object in the keyframe or the entire video, +inadequately capturing spatial complexity and inter-frame motion. To overcome +these challenges, we propose VRS-HQ, an end-to-end video reasoning segmentation +approach that leverages Multimodal Large Language Models (MLLMs) to inject rich +spatiotemporal features into hierarchical tokens.Our key innovations include a +Temporal Dynamic Aggregation (TDA) and a Token-driven Keyframe Selection (TKS). +Specifically, we design frame-level and temporal-level tokens that +utilize MLLM's autoregressive learning to effectively capture both local and +global information. Subsequently, we apply a similarity-based weighted fusion +and frame selection strategy, then utilize SAM2 to perform keyframe +segmentation and propagation. To enhance keyframe localization accuracy, the +TKS filters keyframes based on SAM2's occlusion scores during inference. VRS-HQ +achieves state-of-the-art performance on ReVOS, surpassing VISA by +5.9%/12.5%/9.1% in J&F scores across the three subsets. These results highlight +the strong temporal reasoning and segmentation capabilities of our method. Code +and model weights will be released at VRS-HQ. + +**Decision Explanation:** The paper primarily focuses on video processing, which is one of the excluded areas according to the criteria. + +--- + +## [Evaluating SAT and SMT Solvers on Large-Scale Sudoku Puzzles](https://arxiv.org/abs/2501.08569v1) +**arXiv ID:** 2501.08569v1 + +**Abstract:** +> Modern SMT solvers have revolutionized the approach to constraint +satisfaction problems by integrating advanced theory reasoning and encoding +techniques. In this work, we evaluate the performance of modern SMT solvers in +Z3, CVC5 and DPLL(T) against a standard SAT solver in DPLL. By benchmarking +these solvers on novel, diverse 25x25 Sudoku puzzles of various difficulty +levels created by our improved Sudoku generator, we examine the impact of +advanced theory reasoning and encoding techniques. Our findings demonstrate +that modern SMT solvers significantly outperform classical SAT solvers. This +work highlights the evolution of logical solvers and exemplifies the utility of +SMT solvers in addressing large-scale constraint satisfaction problems. + +**Decision Explanation:** The paper does not meet the criteria as it focuses on evaluating SAT and SMT solvers on Sudoku puzzles, which does not involve Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Disjoint Processing Mechanisms of Hierarchical and Linear Grammars in + Large Language Models](https://arxiv.org/abs/2501.08618v1) +**arXiv ID:** 2501.08618v1 + +**Abstract:** +> All natural languages are structured hierarchically. In humans, this +structural restriction is neurologically coded: when two grammars are presented +with identical vocabularies, brain areas responsible for language processing +are only sensitive to hierarchical grammars. Using large language models +(LLMs), we investigate whether such functionally distinct hierarchical +processing regions can arise solely from exposure to large-scale language +distributions. We generate inputs using English, Italian, Japanese, or nonce +words, varying the underlying grammars to conform to either hierarchical or +linear/positional rules. Using these grammars, we first observe that language +models show distinct behaviors on hierarchical versus linearly structured +inputs. Then, we find that the components responsible for processing +hierarchical grammars are distinct from those that process linear grammars; we +causally verify this in ablation experiments. Finally, we observe that +hierarchy-selective components are also active on nonce grammars; this suggests +that hierarchy sensitivity is not tied to meaning, nor in-distribution inputs. + +**Decision Explanation:** The paper does not meet the criteria for practical applications of Large Language Models, as it focuses on the internal mechanisms of language processing in LLMs rather than real-world applications or experimental results with quantitative metrics. + +--- + +## [ViBidirectionMT-Eval: Machine Translation for Vietnamese-Chinese and + Vietnamese-Lao language pair](https://arxiv.org/abs/2501.08621v1) +**arXiv ID:** 2501.08621v1 + +**Abstract:** +> This paper presents an results of the VLSP 2022-2023 Machine Translation +Shared Tasks, focusing on Vietnamese-Chinese and Vietnamese-Lao machine +translation. The tasks were organized as part of the 9th, 10th annual workshop +on Vietnamese Language and Speech Processing (VLSP 2022, VLSP 2023). The +objective of the shared task was to build machine translation systems, +specifically targeting Vietnamese-Chinese and Vietnamese-Lao translation +(corresponding to 4 translation directions). The submission were evaluated on +1,000 pairs for testing (news and general domains) using established metrics +like BLEU [11] and SacreBLEU [12]. Additionally, system outputs also were +evaluated with human judgment provided by experts in Chinese and Lao languages. +These human assessments played a crucial role in ranking the performance of the +machine translation models, ensuring a more comprehensive evaluation. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on machine translation, which is not explicitly mentioned as a relevant area, and does not clearly involve Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Reassessing the Role of Chain-of-Thought in Sentiment Analysis: Insights + and Limitations](https://arxiv.org/abs/2501.08641v1) +**arXiv ID:** 2501.08641v1 + +**Abstract:** +> The relationship between language and thought remains an unresolved +philosophical issue. Existing viewpoints can be broadly categorized into two +schools: one asserting their independence, and another arguing that language +constrains thought. In the context of large language models, this debate raises +a crucial question: Does a language model's grasp of semantic meaning depend on +thought processes? To explore this issue, we investigate whether reasoning +techniques can facilitate semantic understanding. Specifically, we +conceptualize thought as reasoning, employ chain-of-thought prompting as a +reasoning technique, and examine its impact on sentiment analysis tasks. The +experiments show that chain-of-thought has a minimal impact on sentiment +analysis tasks. Both the standard and chain-of-thought prompts focus on aspect +terms rather than sentiment in the generated content. Furthermore, +counterfactual experiments reveal that the model's handling of sentiment tasks +primarily depends on information from demonstrations. The experimental results +support the first viewpoint. + +**Decision Explanation:** The paper does not meet the criteria for practical applications of Large Language Models, nor does it include experimental results with quantitative metrics that demonstrate performance improvements in areas like knowledge graphs or retrieval-augmented generation. + +--- + +## [Application of Deep Reinforcement Learning to UAV Swarming for Ground + Surveillance](https://arxiv.org/abs/2501.08655v1) +**arXiv ID:** 2501.08655v1 + +**Abstract:** +> This paper summarizes in depth the state of the art of aerial swarms, +covering both classical and new reinforcement-learning-based approaches for +their management. Then, it proposes a hybrid AI system, integrating deep +reinforcement learning in a multi-agent centralized swarm architecture. The +proposed system is tailored to perform surveillance of a specific area, +searching and tracking ground targets, for security and law enforcement +applications. The swarm is governed by a central swarm controller responsible +for distributing different search and tracking tasks among the cooperating +UAVs. Each UAV agent is then controlled by a collection of cooperative +sub-agents, whose behaviors have been trained using different deep +reinforcement learning models, tailored for the different task types proposed +by the swarm controller. More specifically, proximal policy optimization (PPO) +algorithms were used to train the agents' behavior. In addition, several +metrics to assess the performance of the swarm in this application were +defined. The results obtained through simulation show that our system searches +the operation area effectively, acquires the targets in a reasonable time, and +is capable of tracking them continuously and consistently. + +**Decision Explanation:** The paper primarily focuses on law enforcement applications and does not explicitly mention Large Language Models (LLMs) or their practical applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus not meeting the required criteria. + +--- + +## [SPEQ: Stabilization Phases for Efficient Q-Learning in High + Update-To-Data Ratio Reinforcement Learning](https://arxiv.org/abs/2501.08669v1) +**arXiv ID:** 2501.08669v1 + +**Abstract:** +> A key challenge in Deep Reinforcement Learning is sample efficiency, +especially in real-world applications where collecting environment interactions +is expensive or risky. Recent off-policy algorithms improve sample efficiency +by increasing the Update-To-Data (UTD) ratio and performing more gradient +updates per environment interaction. While this improves sample efficiency, it +significantly increases computational cost due to the higher number of gradient +updates required. In this paper we propose a sample-efficient method to improve +computational efficiency by separating training into distinct learning phases +in order to exploit gradient updates more effectively. Our approach builds on +top of the Dropout Q-Functions (DroQ) algorithm and alternates between an +online, low UTD ratio training phase, and an offline stabilization phase. +During the stabilization phase, we fine-tune the Q-functions without collecting +new environment interactions. This process improves the effectiveness of the +replay buffer and reduces computational overhead. Our experimental results on +continuous control problems show that our method achieves results comparable to +state-of-the-art, high UTD ratio algorithms while requiring 56\% fewer gradient +updates and 50\% less training time than DroQ. Our approach offers an effective +and computationally economical solution while maintaining the same sample +efficiency as the more costly, high UTD ratio state-of-the-art. + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their applications, but rather on reinforcement learning and Q-learning, which is outside the specified scope. + +--- + +## [ToMATO: Verbalizing the Mental States of Role-Playing LLMs for + Benchmarking Theory of Mind](https://arxiv.org/abs/2501.08838v1) +**arXiv ID:** 2501.08838v1 + +**Abstract:** +> Existing Theory of Mind (ToM) benchmarks diverge from real-world scenarios in +three aspects: 1) they assess a limited range of mental states such as beliefs, +2) false beliefs are not comprehensively explored, and 3) the diverse +personality traits of characters are overlooked. To address these challenges, +we introduce ToMATO, a new ToM benchmark formulated as multiple-choice QA over +conversations. ToMATO is generated via LLM-LLM conversations featuring +information asymmetry. By employing a prompting method that requires +role-playing LLMs to verbalize their thoughts before each utterance, we capture +both first- and second-order mental states across five categories: belief, +intention, desire, emotion, and knowledge. These verbalized thoughts serve as +answers to questions designed to assess the mental states of characters within +conversations. Furthermore, the information asymmetry introduced by hiding +thoughts from others induces the generation of false beliefs about various +mental states. Assigning distinct personality traits to LLMs further +diversifies both utterances and thoughts. ToMATO consists of 5.4k questions, +753 conversations, and 15 personality trait patterns. Our analysis shows that +this dataset construction approach frequently generates false beliefs due to +the information asymmetry between role-playing LLMs, and effectively reflects +diverse personalities. We evaluate nine LLMs on ToMATO and find that even +GPT-4o mini lags behind human performance, especially in understanding false +beliefs, and lacks robustness to various personality traits. + +**Decision Explanation:** The paper primarily focuses on benchmarking Theory of Mind in role-playing LLMs, which does not meet the criteria of having practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and does not demonstrate clear real-world applications or comparisons with state-of-the-art techniques in these areas. + +--- + +## [Exploring Task-Level Optimal Prompts for Visual In-Context Learning](https://arxiv.org/abs/2501.08841v1) +**arXiv ID:** 2501.08841v1 + +**Abstract:** +> With the development of Vision Foundation Models (VFMs) in recent years, +Visual In-Context Learning (VICL) has become a better choice compared to +modifying models in most scenarios. Different from retraining or fine-tuning +model, VICL does not require modifications to the model's weights or +architecture, and only needs a prompt with demonstrations to teach VFM how to +solve tasks. Currently, significant computational cost for finding optimal +prompts for every test sample hinders the deployment of VICL, as determining +which demonstrations to use for constructing prompts is very costly. In this +paper, however, we find a counterintuitive phenomenon that most test samples +actually achieve optimal performance under the same prompts, and searching for +sample-level prompts only costs more time but results in completely identical +prompts. Therefore, we propose task-level prompting to reduce the cost of +searching for prompts during the inference stage and introduce two time-saving +yet effective task-level prompt search strategies. Extensive experimental +results show that our proposed method can identify near-optimal prompts and +reach the best VICL performance with a minimal cost that prior work has never +achieved. + +**Decision Explanation:** The paper primarily focuses on visual in-context learning and prompt engineering for Vision Foundation Models, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Digital Phenotyping for Adolescent Mental Health: A Feasibility Study + Employing Machine Learning to Predict Mental Health Risk From Active and + Passive Smartphone Data](https://arxiv.org/abs/2501.08851v1) +**arXiv ID:** 2501.08851v1 + +**Abstract:** +> Background: Adolescents are particularly vulnerable to mental disorders, with +over 75% of cases manifesting before the age of 25. Research indicates that +only 18 to 34% of young people experiencing high levels of depression or +anxiety symptoms seek support. Digital tools leveraging smartphones offer +scalable and early intervention opportunities. Objective: Using a novel machine +learning framework, this study evaluated the feasibility of integrating active +and passive smartphone data to predict mental disorders in non-clinical +adolescents. Specifically, we investigated the utility of the Mindcraft app in +predicting risks for internalising and externalising disorders, eating +disorders, insomnia and suicidal ideation. Methods: Participants (N=103; mean +age 16.1 years) were recruited from three London schools. Participants +completed the Strengths and Difficulties Questionnaire, the Eating Disorders-15 +Questionnaire, Sleep Condition Indicator Questionnaire and indicated the +presence/absence of suicidal ideation. They used the Mindcraft app for 14 days, +contributing active data via self-reports and passive data from smartphone +sensors. A contrastive pretraining phase was applied to enhance user-specific +feature stability, followed by supervised fine-tuning. The model evaluation +employed leave-one-subject-out cross-validation using balanced accuracy as the +primary metric. Results: The integration of active and passive data achieved +superior performance compared to individual data sources, with mean balanced +accuracies of 0.71 for SDQ-High risk, 0.67 for insomnia, 0.77 for suicidal +ideation and 0.70 for eating disorders. The contrastive learning framework +stabilised daily behavioural representations, enhancing predictive robustness. +This study demonstrates the potential of integrating active and passive +smartphone data with advanced machine-learning techniques for predicting mental +health risks. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, specifically mental health, which is excluded according to the criteria. + +--- + +## [Silent Abandonment in Text-Based Contact Centers: Identifying, + Quantifying, and Mitigating its Operational Impacts](https://arxiv.org/abs/2501.08869v2) +**arXiv ID:** 2501.08869v2 + +**Abstract:** +> In the quest to improve services, companies offer customers the option to +interact with agents via texting. Such contact centers face unique challenges +compared to traditional call centers, as measuring customer experience proxies +like abandonment and patience involves uncertainty. A key source of this +uncertainty is silent abandonment, where customers leave without notifying the +system, wasting agent time and leaving their status unclear. Silent abandonment +also obscures whether a customer was served or left. Our goals are to measure +the magnitude of silent abandonment and mitigate its effects. Classification +models show that 3%-70% of customers across 17 companies abandon silently. In +one study, 71.3% of abandoning customers did so silently, reducing agent +efficiency by 3.2% and system capacity by 15.3%, incurring $5,457 in annual +costs per agent. We develop an expectation-maximization (EM) algorithm to +estimate customer patience under uncertainty and identify influencing +covariates. We find that companies should use classification models to estimate +abandonment scope and our EM algorithm to assess patience. We suggest +strategies to operationally mitigate the impact of silent abandonment by +predicting suspected silent-abandonment behavior or changing service design. +Specifically, we show that while allowing customers to write while waiting in +the queue creates a missing data challenge, it also significantly increases +patience and reduces service time, leading to reduced abandonment and lower +staffing requirements. + +**Decision Explanation:** The paper does not meet the criteria for practical applications of Large Language Models (LLMs) and does not mention LLMs, knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, focusing instead on operational impacts in text-based contact centers. + +--- + +## [Projection Implicit Q-Learning with Support Constraint for Offline + Reinforcement Learning](https://arxiv.org/abs/2501.08907v1) +**arXiv ID:** 2501.08907v1 + +**Abstract:** +> Offline Reinforcement Learning (RL) faces a critical challenge of +extrapolation errors caused by out-of-distribution (OOD) actions. Implicit +Q-Learning (IQL) algorithm employs expectile regression to achieve in-sample +learning, effectively mitigating the risks associated with OOD actions. +However, the fixed hyperparameter in policy evaluation and density-based policy +improvement method limit its overall efficiency. In this paper, we propose +Proj-IQL, a projective IQL algorithm enhanced with the support constraint. In +the policy evaluation phase, Proj-IQL generalizes the one-step approach to a +multi-step approach through vector projection, while maintaining in-sample +learning and expectile regression framework. In the policy improvement phase, +Proj-IQL introduces support constraint that is more aligned with the policy +evaluation approach. Furthermore, we theoretically demonstrate that Proj-IQL +guarantees monotonic policy improvement and enjoys a progressively more +rigorous criterion for superior actions. Empirical results demonstrate the +Proj-IQL achieves state-of-the-art performance on D4RL benchmarks, especially +in challenging navigation domains. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on offline reinforcement learning and does not explicitly mention Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Modeling Melt Pool Features and Spatter Using Symbolic Regression and + Machine Learning](https://arxiv.org/abs/2501.08922v1) +**arXiv ID:** 2501.08922v1 + +**Abstract:** +> Additive manufacturing (AM) is a rapidly evolving technology that has +attracted applications across a wide range of fields due to its ability to +fabricate complex geometries. However, one of the key challenges in AM is +achieving consistent print quality. This inconsistency is often attributed to +uncontrolled melt pool dynamics, partly caused by spatter which can lead to +defects. Therefore, capturing and controlling the evolution of the melt pool is +crucial for enhancing process stability and part quality. In this study, we +developed a framework to support decision-making in AM operations, facilitating +quality control and minimizing defects via machine learning (ML) and polynomial +symbolic regression models. We implemented experimentally validated +computational tools as a cost-effective approach to collect large datasets from +laser powder bed fusion (LPBF) processes. For a dataset consisting of 281 +process conditions, parameters such as melt pool dimensions (length, width, +depth), melt pool geometry (area, volume), and volume indicated as spatter were +extracted. Using machine learning (ML) and polynomial symbolic regression +models, a high R2 of over 95 % was achieved in predicting the melt pool +dimensions and geometry features for both the training and testing datasets, +with either process conditions (power and velocity) or melt pool dimensions as +the model inputs. In the case of volume indicated as spatter, R2 improved after +logarithmic transforming the model inputs, which was either the process +conditions or the melt pool dimensions. Among the investigated ML models, the +ExtraTree model achieved the highest R2 values of 96.7 % and 87.5 %. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on additive manufacturing and machine learning applications, without any mention of Large Language Models (LLMs) or related areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Visual WetlandBirds Dataset: Bird Species Identification and Behavior + Recognition in Videos](https://arxiv.org/abs/2501.08931v1) +**arXiv ID:** 2501.08931v1 + +**Abstract:** +> The current biodiversity loss crisis makes animal monitoring a relevant field +of study. In light of this, data collected through monitoring can provide +essential insights, and information for decision-making aimed at preserving +global biodiversity. Despite the importance of such data, there is a notable +scarcity of datasets featuring videos of birds, and none of the existing +datasets offer detailed annotations of bird behaviors in video format. In +response to this gap, our study introduces the first fine-grained video dataset +specifically designed for bird behavior detection and species classification. +This dataset addresses the need for comprehensive bird video datasets and +provides detailed data on bird actions, facilitating the development of deep +learning models to recognize these, similar to the advancements made in human +action recognition. The proposed dataset comprises 178 videos recorded in +Spanish wetlands, capturing 13 different bird species performing 7 distinct +behavior classes. In addition, we also present baseline results using state of +the art models on two tasks: bird behavior recognition and species +classification. + +**Decision Explanation:** The paper primarily focuses on video processing, which is excluded according to the criteria, and does not demonstrate a clear connection to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Analyzing the Ethical Logic of Six Large Language Models](https://arxiv.org/abs/2501.08951v1) +**arXiv ID:** 2501.08951v1 + +**Abstract:** +> This study examines the ethical reasoning of six prominent generative large +language models: OpenAI GPT-4o, Meta LLaMA 3.1, Perplexity, Anthropic Claude +3.5 Sonnet, Google Gemini, and Mistral 7B. The research explores how these +models articulate and apply ethical logic, particularly in response to moral +dilemmas such as the Trolley Problem, and Heinz Dilemma. Departing from +traditional alignment studies, the study adopts an explainability-transparency +framework, prompting models to explain their ethical reasoning. This approach +is analyzed through three established ethical typologies: the +consequentialist-deontological analytic, Moral Foundations Theory, and the +Kohlberg Stages of Moral Development Model. Findings reveal that LLMs exhibit +largely convergent ethical logic, marked by a rationalist, consequentialist +emphasis, with decisions often prioritizing harm minimization and fairness. +Despite similarities in pre-training and model architecture, a mixture of +nuanced and significant differences in ethical reasoning emerge across models, +reflecting variations in fine-tuning and post-training processes. The models +consistently display erudition, caution, and self-awareness, presenting ethical +reasoning akin to a graduate-level discourse in moral philosophy. In striking +uniformity these systems all describe their ethical reasoning as more +sophisticated than what is characteristic of typical human moral logic. + +**Decision Explanation:** The paper primarily focuses on the ethical logic of Large Language Models, which aligns with responsible AI application or AI ethics, a criterion that the paper should not meet. + +--- + +## [Kolmogorov-Arnold Networks for Time Series Granger Causality Inference](https://arxiv.org/abs/2501.08958v1) +**arXiv ID:** 2501.08958v1 + +**Abstract:** +> We introduce Granger Causality Kolmogorov-Arnold Networks (GCKAN), an +innovative architecture that extends the recently proposed Kolmogorov-Arnold +Networks (KAN) to the domain of causal inference. By extracting base weights +from KAN layers and incorporating the sparsity-inducing penalty along with +ridge regularization, GCKAN infers the Granger causality from time series while +enabling automatic time lag selection. Additionally, we propose an algorithm +leveraging time-reversed Granger causality to enhance inference accuracy. The +algorithm compares prediction and sparse-inducing losses derived from the +original and time-reversed series, automatically selecting the casual +relationship with the higher score or integrating the results to mitigate +spurious connectivities. Comprehensive experiments conducted on Lorenz-96, gene +regulatory networks, fMRI BOLD signals, and VAR datasets demonstrate that the +proposed model achieves competitive performance to state-of-the-art methods in +inferring Granger causality from nonlinear, high-dimensional, and +limited-sample time series. + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their applications, and instead proposes a new architecture for time series Granger causality inference, which is outside the specified scope. + +--- + +## [An analysis of data variation and bias in image-based dermatological + datasets for machine learning classification](https://arxiv.org/abs/2501.08962v1) +**arXiv ID:** 2501.08962v1 + +**Abstract:** +> AI algorithms have become valuable in aiding professionals in healthcare. The +increasing confidence obtained by these models is helpful in critical decision +demands. In clinical dermatology, classification models can detect malignant +lesions on patients' skin using only RGB images as input. However, most +learning-based methods employ data acquired from dermoscopic datasets on +training, which are large and validated by a gold standard. Clinical models aim +to deal with classification on users' smartphone cameras that do not contain +the corresponding resolution provided by dermoscopy. Also, clinical +applications bring new challenges. It can contain captures from uncontrolled +environments, skin tone variations, viewpoint changes, noises in data and +labels, and unbalanced classes. A possible alternative would be to use transfer +learning to deal with the clinical images. However, as the number of samples is +low, it can cause degradations on the model's performance; the source +distribution used in training differs from the test set. This work aims to +evaluate the gap between dermoscopic and clinical samples and understand how +the dataset variations impact training. It assesses the main differences +between distributions that disturb the model's prediction. Finally, from +experiments on different architectures, we argue how to combine the data from +divergent distributions, decreasing the impact on the model's final accuracy. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is excluded according to the criteria. + +--- + +## [How Do Generative Models Draw a Software Engineer? A Case Study on + Stable Diffusion Bias](https://arxiv.org/abs/2501.09014v1) +**arXiv ID:** 2501.09014v1 + +**Abstract:** +> Generative models are nowadays widely used to generate graphical content used +for multiple purposes, e.g. web, art, advertisement. However, it has been shown +that the images generated by these models could reinforce societal biases +already existing in specific contexts. In this paper, we focus on understanding +if this is the case when one generates images related to various software +engineering tasks. In fact, the Software Engineering (SE) community is not +immune from gender and ethnicity disparities, which could be amplified by the +use of these models. Hence, if used without consciousness, artificially +generated images could reinforce these biases in the SE domain. Specifically, +we perform an extensive empirical evaluation of the gender and ethnicity bias +exposed by three versions of the Stable Diffusion (SD) model (a very popular +open-source text-to-image model) - SD 2, SD XL, and SD 3 - towards SE tasks. We +obtain 6,720 images by feeding each model with two sets of prompts describing +different software-related tasks: one set includes the Software Engineer +keyword, and one set does not include any specification of the person +performing the task. Next, we evaluate the gender and ethnicity disparities in +the generated images. Results show how all models are significantly biased +towards male figures when representing software engineers. On the contrary, +while SD 2 and SD XL are strongly biased towards White figures, SD 3 is +slightly more biased towards Asian figures. Nevertheless, all models +significantly under-represent Black and Arab figures, regardless of the prompt +style used. The results of our analysis highlight severe concerns about +adopting those models to generate content for SE tasks and open the field for +future research on bias mitigation in this context. + +**Decision Explanation:** The paper primarily focuses on social applications of AI in regard to diversity and social harm, which is one of the excluded topics. + +--- + +## [TCMM: Token Constraint and Multi-Scale Memory Bank of Contrastive + Learning for Unsupervised Person Re-identification](https://arxiv.org/abs/2501.09044v1) +**arXiv ID:** 2501.09044v1 + +**Abstract:** +> This paper proposes the ViT Token Constraint and Multi-scale Memory bank +(TCMM) method to address the patch noises and feature inconsistency in +unsupervised person re-identification works. Many excellent methods use ViT +features to obtain pseudo labels and clustering prototypes, then train the +model with contrastive learning. However, ViT processes images by performing +patch embedding, which inevitably introduces noise in patches and may +compromise the performance of the re-identification model. On the other hand, +previous memory bank based contrastive methods may lead data inconsistency due +to the limitation of batch size. Furthermore, existing pseudo label methods +often discard outlier samples that are difficult to cluster. It sacrifices the +potential value of outlier samples, leading to limited model diversity and +robustness. This paper introduces the ViT Token Constraint to mitigate the +damage caused by patch noises to the ViT architecture. The proposed Multi-scale +Memory enhances the exploration of outlier samples and maintains feature +consistency. Experimental results demonstrate that our system achieves +state-of-the-art performance on common benchmarks. The project is available at +\href{https://github.com/andy412510/TCMM}{https://github.com/andy412510/TCMM}. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on person re-identification, which is a computer vision task, and does not involve Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Generating Realistic Synthetic Head Rotation Data for Extended Reality + using Deep Learning](https://arxiv.org/abs/2501.09050v1) +**arXiv ID:** 2501.09050v1 + +**Abstract:** +> Extended Reality is a revolutionary method of delivering multimedia content +to users. A large contributor to its popularity is the sense of immersion and +interactivity enabled by having real-world motion reflected in the virtual +experience accurately and immediately. This user motion, mainly caused by head +rotations, induces several technical challenges. For instance, which content is +generated and transmitted depends heavily on where the user is looking. +Seamless systems, taking user motion into account proactively, will therefore +require accurate predictions of upcoming rotations. Training and evaluating +such predictors requires vast amounts of orientational input data, which is +expensive to gather, as it requires human test subjects. A more feasible +approach is to gather a modest dataset through test subjects, and then extend +it to a more sizeable set using synthetic data generation methods. In this +work, we present a head rotation time series generator based on TimeGAN, an +extension of the well-known Generative Adversarial Network, designed +specifically for generating time series. This approach is able to extend a +dataset of head rotations with new samples closely matching the distribution of +the measured time series. + +**Decision Explanation:** The paper primarily focuses on generating synthetic head rotation data for Extended Reality using deep learning, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Polyp detection in colonoscopy images using YOLOv11](https://arxiv.org/abs/2501.09051v1) +**arXiv ID:** 2501.09051v1 + +**Abstract:** +> Colorectal cancer (CRC) is one of the most commonly diagnosed cancers all +over the world. It starts as a polyp in the inner lining of the colon. To +prevent CRC, early polyp detection is required. Colonosopy is used for the +inspection of the colon. Generally, the images taken by the camera placed at +the tip of the endoscope are analyzed by the experts manually. Various +traditional machine learning models have been used with the rise of machine +learning. Recently, deep learning models have shown more effectiveness in polyp +detection due to their superiority in generalizing and learning small features. +These deep learning models for object detection can be segregated into two +different types: single-stage and two-stage. Generally, two stage models have +higher accuracy than single stage ones but the single stage models have low +inference time. Hence, single stage models are easy to use for quick object +detection. YOLO is one of the singlestage models used successfully for polyp +detection. It has drawn the attention of researchers because of its lower +inference time. The researchers have used Different versions of YOLO so far, +and with each newer version, the accuracy of the model is increasing. This +paper aims to see the effectiveness of the recently released YOLOv11 to detect +polyp. We analyzed the performance for all five models of YOLOv11 (YOLO11n, +YOLO11s, YOLO11m, YOLO11l, YOLO11x) with Kvasir dataset for the training and +testing. Two different versions of the dataset were used. The first consisted +of the original dataset, and the other was created using augmentation +techniques. The performance of all the models with these two versions of the +dataset have been analysed. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, specifically polyp detection in colonoscopy images, which is not aligned with the specified criteria. + +--- + +## [Decompose-ToM: Enhancing Theory of Mind Reasoning in Large Language + Models through Simulation and Task Decomposition](https://arxiv.org/abs/2501.09056v1) +**arXiv ID:** 2501.09056v1 + +**Abstract:** +> Theory of Mind (ToM) is the ability to understand and reflect on the mental +states of others. Although this capability is crucial for human interaction, +testing on Large Language Models (LLMs) reveals that they possess only a +rudimentary understanding of it. Although the most capable closed-source LLMs +have come close to human performance on some ToM tasks, they still perform +poorly on complex variations of the task that involve more structured +reasoning. In this work, we utilize the concept of "pretend-play", or +``Simulation Theory'' from cognitive psychology to propose ``Decompose-ToM'': +an LLM-based inference algorithm that improves model performance on complex ToM +tasks. We recursively simulate user perspectives and decompose the ToM task +into a simpler set of functions: subject identification, question-reframing, +world model updation, and knowledge availability. We test the algorithm on +higher-order ToM tasks and a task testing for ToM capabilities in a +conversational setting, demonstrating that our approach shows significant +improvement across models compared to baseline methods while requiring minimal +prompt tuning across tasks and no additional model training. + +**Decision Explanation:** The paper primarily focuses on enhancing Theory of Mind reasoning in Large Language Models, which does not meet the criteria of having practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and also does not clearly discuss real-world applications or challenges involving LLMs. + +--- + +## [Average-Reward Reinforcement Learning with Entropy Regularization](https://arxiv.org/abs/2501.09080v1) +**arXiv ID:** 2501.09080v1 + +**Abstract:** +> The average-reward formulation of reinforcement learning (RL) has drawn +increased interest in recent years due to its ability to solve +temporally-extended problems without discounting. Independently, RL algorithms +have benefited from entropy-regularization: an approach used to make the +optimal policy stochastic, thereby more robust to noise. Despite the distinct +benefits of the two approaches, the combination of entropy regularization with +an average-reward objective is not well-studied in the literature and there has +been limited development of algorithms for this setting. To address this gap in +the field, we develop algorithms for solving entropy-regularized average-reward +RL problems with function approximation. We experimentally validate our method, +comparing it with existing algorithms on standard benchmarks for RL. + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their practical applications, and instead discusses reinforcement learning with entropy regularization, which does not meet the specified criteria. + +--- + +## [Inferring Transition Dynamics from Value Functions](https://arxiv.org/abs/2501.09081v1) +**arXiv ID:** 2501.09081v1 + +**Abstract:** +> In reinforcement learning, the value function is typically trained to solve +the Bellman equation, which connects the current value to future values. This +temporal dependency hints that the value function may contain implicit +information about the environment's transition dynamics. By rearranging the +Bellman equation, we show that a converged value function encodes a model of +the underlying dynamics of the environment. We build on this insight to propose +a simple method for inferring dynamics models directly from the value function, +potentially mitigating the need for explicit model learning. Furthermore, we +explore the challenges of next-state identifiability, discussing conditions +under which the inferred dynamics model is well-defined. Our work provides a +theoretical foundation for leveraging value functions in dynamics modeling and +opens a new avenue for bridging model-free and model-based reinforcement +learning. + +**Decision Explanation:** The paper does not meet the criteria as it focuses on reinforcement learning and dynamics modeling, which is not directly related to Large Language Models (LLMs) or their practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Mantis Shrimp: Exploring Photometric Band Utilization in Computer Vision + Networks for Photometric Redshift Estimation](https://arxiv.org/abs/2501.09112v1) +**arXiv ID:** 2501.09112v1 + +**Abstract:** +> We present Mantis Shrimp, a multi-survey deep learning model for photometric +redshift estimation that fuses ultra-violet (GALEX), optical (PanSTARRS), and +infrared (UnWISE) imagery. Machine learning is now an established approach for +photometric redshift estimation, with generally acknowledged higher performance +in areas with a high density of spectroscopically identified galaxies over +template-based methods. Multiple works have shown that image-based +convolutional neural networks can outperform tabular-based color/magnitude +models. In comparison to tabular models, image models have additional design +complexities: it is largely unknown how to fuse inputs from different +instruments which have different resolutions or noise properties. The Mantis +Shrimp model estimates the conditional density estimate of redshift using +cutout images. The density estimates are well calibrated and the point +estimates perform well in the distribution of available spectroscopically +confirmed galaxies with (bias = 1e-2), scatter (NMAD = 2.44e-2) and +catastrophic outlier rate ($\eta$=17.53$\%$). We find that early fusion +approaches (e.g., resampling and stacking images from different instruments) +match the performance of late fusion approaches (e.g., concatenating latent +space representations), so that the design choice ultimately is left to the +user. Finally, we study how the models learn to use information across bands, +finding evidence that our models successfully incorporates information from all +surveys. The applicability of our model to the analysis of large populations of +galaxies is limited by the speed of downloading cutouts from external servers; +however, our model could be useful in smaller studies such as generating priors +over redshift for stellar population synthesis. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on computer vision and photometric redshift estimation, which is not related to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Generative Medical Image Anonymization Based on Latent Code Projection + and Optimization](https://arxiv.org/abs/2501.09114v1) +**arXiv ID:** 2501.09114v1 + +**Abstract:** +> Medical image anonymization aims to protect patient privacy by removing +identifying information, while preserving the data utility to solve downstream +tasks. In this paper, we address the medical image anonymization problem with a +two-stage solution: latent code projection and optimization. In the projection +stage, we design a streamlined encoder to project input images into a latent +space and propose a co-training scheme to enhance the projection process. In +the optimization stage, we refine the latent code using two deep loss functions +designed to address the trade-off between identity protection and data utility +dedicated to medical images. Through a comprehensive set of qualitative and +quantitative experiments, we showcase the effectiveness of our approach on the +MIMIC-CXR chest X-ray dataset by generating anonymized synthetic images that +can serve as training set for detecting lung pathologies. Source codes are +available at https://github.com/Huiyu-Li/GMIA. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is excluded according to the criteria. + +--- + +## [Towards Multilingual LLM Evaluation for Baltic and Nordic languages: A + study on Lithuanian History](https://arxiv.org/abs/2501.09154v1) +**arXiv ID:** 2501.09154v1 + +**Abstract:** +> In this work, we evaluated Lithuanian and general history knowledge of +multilingual Large Language Models (LLMs) on a multiple-choice +question-answering task. The models were tested on a dataset of Lithuanian +national and general history questions translated into Baltic, Nordic, and +other languages (English, Ukrainian, Arabic) to assess the knowledge sharing +from culturally and historically connected groups. We evaluated GPT-4o, +LLaMa3.1 8b and 70b, QWEN2.5 7b and 72b, Mistral Nemo 12b, LLaMa3 8b, Mistral +7b, LLaMa3.2 3b, and Nordic fine-tuned models (GPT-SW3 and LLaMa3 8b). + Our results show that GPT-4o consistently outperformed all other models +across language groups, with slightly better results for Baltic and Nordic +languages. Larger open-source models like QWEN2.5 72b and LLaMa3.1 70b +performed well but showed weaker alignment with Baltic languages. Smaller +models (Mistral Nemo 12b, LLaMa3.2 3b, QWEN 7B, LLaMa3.1 8B, and LLaMa3 8b) +demonstrated gaps with LT-related alignment with Baltic languages while +performing better on Nordic and other languages. The Nordic fine-tuned models +did not surpass multilingual models, indicating that shared cultural or +historical context alone does not guarantee better performance. + +**Decision Explanation:** The paper primarily focuses on evaluating multilingual Large Language Models for Baltic and Nordic languages, which does not meet the criteria of having practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and also does not discuss real-world applications or challenges involving LLMs in autonomous or agentic AI scenarios. + +--- + +## [The Veln(ia)s is in the Details: Evaluating LLM Judgment on Latvian and + Lithuanian Short Answer Matching](https://arxiv.org/abs/2501.09164v1) +**arXiv ID:** 2501.09164v1 + +**Abstract:** +> In this work, we address the challenge of evaluating large language models +(LLMs) on the short answer matching task for Latvian and Lithuanian languages. +We introduce novel datasets consisting of 502 Latvian and 690 Lithuanian +question-answer pairs. For each question-answer pair, we generated matched and +non-matched answers using a set of alteration rules specifically designed to +introduce small but meaningful changes in the text. These generated answers +serve as test cases to assess the ability of LLMs to detect subtle differences +in matching of the original answers. A subset of the datasets was manually +verified for quality and accuracy. Our results show that while larger LLMs, +such as QWEN2.5 72b and LLaMa3.1 70b, demonstrate near-perfect performance in +distinguishing matched and non-matched answers, smaller models show more +variance. For instance, LLaMa3.1 8b and EuroLLM 9b benefited from few-shot +examples, while Mistral Nemo 12b underperformed on detection of subtle text +alteration, particularly in Lithuanian, even with additional examples. QWEN2.5 +7b and Mistral 7b were able to obtain a strong and comparable performance to +the larger 70b models in zero and few shot experiments. Moreover, the +performance of Mistral 7b was weaker in few shot experiments. + +**Decision Explanation:** The paper primarily focuses on evaluating LLMs on short answer matching tasks for specific languages, which does not clearly demonstrate practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and does not explicitly address real-world challenges or novel approaches in these areas. + +--- + +## [Grounding Text-To-Image Diffusion Models For Controlled High-Quality + Image Generation](https://arxiv.org/abs/2501.09194v1) +**arXiv ID:** 2501.09194v1 + +**Abstract:** +> Large-scale text-to-image (T2I) diffusion models have demonstrated an +outstanding performance in synthesizing diverse high-quality visuals from +natural language text captions. Multiple layout-to-image models have been +developed to control the generation process by utilizing a broad array of +layouts such as segmentation maps, edges, and human keypoints. In this work, we +present ObjectDiffusion, a model that takes inspirations from the top +cutting-edge image generative frameworks to seamlessly condition T2I models +with new bounding boxes capabilities. Specifically, we make substantial +modifications to the network architecture introduced in ContorlNet to integrate +it with the condition processing and injection techniques proposed in GLIGEN. +ObjectDiffusion is initialized with pretraining parameters to leverage the +generation knowledge obtained from training on large-scale datasets. We +fine-tune ObjectDiffusion on the COCO2017 training dataset and evaluate it on +the COCO2017 validation dataset. Our model achieves an AP$_{50}$ of 46.6, an AR +of 44.5, and a FID of 19.8 outperforming the current SOTA model trained on +open-source datasets in all of the three metrics. ObjectDiffusion demonstrates +a distinctive capability in synthesizing diverse, high-quality, high-fidelity +images that seamlessly conform to the semantic and spatial control layout. +Evaluated in qualitative and quantitative tests, ObjectDiffusion exhibits +remarkable grounding abilities on closed-set and open-set settings across a +wide variety of contexts. The qualitative assessment verifies the ability of +ObjectDiffusion to generate multiple objects of different sizes and locations. + +**Decision Explanation:** The paper primarily focuses on image generation using text-to-image diffusion models, which does not meet the criteria of having practical applications of Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Interpretable Droplet Digital PCR Assay for Trustworthy Molecular + Diagnostics](https://arxiv.org/abs/2501.09218v1) +**arXiv ID:** 2501.09218v1 + +**Abstract:** +> Accurate molecular quantification is essential for advancing research and +diagnostics in fields such as infectious diseases, cancer biology, and genetic +disorders. Droplet digital PCR (ddPCR) has emerged as a gold standard for +achieving absolute quantification. While computational ddPCR technologies have +advanced significantly, achieving automatic interpretation and consistent +adaptability across diverse operational environments remains a challenge. To +address these limitations, we introduce the intelligent interpretable droplet +digital PCR (I2ddPCR) assay, a comprehensive framework integrating front-end +predictive models (for droplet segmentation and classification) with GPT-4o +multimodal large language model (MLLM, for context-aware explanations and +recommendations) to automate and enhance ddPCR image analysis. This approach +surpasses the state-of-the-art models, affording 99.05% accuracy in processing +complex ddPCR images containing over 300 droplets per image with varying +signal-to-noise ratios (SNRs). By combining specialized neural networks and +large language models, the I2ddPCR assay offers a robust and adaptable solution +for absolute molecular quantification, achieving a sensitivity capable of +detecting low-abundance targets as low as 90.32 copies/{\mu}L. Furthermore, it +improves model's transparency through detailed explanation and troubleshooting +guidance, empowering users to make informed decisions. This innovative +framework has the potential to benefit molecular diagnostics, disease research, +and clinical applications, especially in resource-constrained settings. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is excluded according to the criteria. + +--- + +## [SEAL: Entangled White-box Watermarks on Low-Rank Adaptation](https://arxiv.org/abs/2501.09284v1) +**arXiv ID:** 2501.09284v1 + +**Abstract:** +> Recently, LoRA and its variants have become the de facto strategy for +training and sharing task-specific versions of large pretrained models, thanks +to their efficiency and simplicity. However, the issue of copyright protection +for LoRA weights, especially through watermark-based techniques, remains +underexplored. To address this gap, we propose SEAL (SEcure wAtermarking on +LoRA weights), the universal whitebox watermarking for LoRA. SEAL embeds a +secret, non-trainable matrix between trainable LoRA weights, serving as a +passport to claim ownership. SEAL then entangles the passport with the LoRA +weights through training, without extra loss for entanglement, and distributes +the finetuned weights after hiding the passport. When applying SEAL, we +observed no performance degradation across commonsense reasoning, +textual/visual instruction tuning, and text-to-image synthesis tasks. We +demonstrate that SEAL is robust against a variety of known attacks: removal, +obfuscation, and ambiguity attacks. + +**Decision Explanation:** The paper primarily focuses on copyright protection and watermarking for LoRA weights, which does not meet the criteria of having practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and also does not discuss real-world applications or challenges involving LLMs in autonomous or agentic AI scenarios. + +--- + +## [Neural Honeytrace: A Robust Plug-and-Play Watermarking Framework against + Model Extraction Attacks](https://arxiv.org/abs/2501.09328v1) +**arXiv ID:** 2501.09328v1 + +**Abstract:** +> Developing high-performance deep learning models is resource-intensive, +leading model owners to utilize Machine Learning as a Service (MLaaS) platforms +instead of publicly releasing their models. However, malicious users may +exploit query interfaces to execute model extraction attacks, reconstructing +the target model's functionality locally. While prior research has investigated +triggerable watermarking techniques for asserting ownership, existing methods +face significant challenges: (1) most approaches require additional training, +resulting in high overhead and limited flexibility, and (2) they often fail to +account for advanced attackers, leaving them vulnerable to adaptive attacks. + In this paper, we propose Neural Honeytrace, a robust plug-and-play +watermarking framework against model extraction attacks. We first formulate a +watermark transmission model from an information-theoretic perspective, +providing an interpretable account of the principles and limitations of +existing triggerable watermarking. Guided by the model, we further introduce: +(1) a similarity-based training-free watermarking method for plug-and-play and +flexible watermarking, and (2) a distribution-based multi-step watermark +information transmission strategy for robust watermarking. Comprehensive +experiments on four datasets demonstrate that Neural Honeytrace outperforms +previous methods in efficiency and resisting adaptive attacks. Neural +Honeytrace reduces the average number of samples required for a worst-case +t-Test-based copyright claim from $12,000$ to $200$ with zero training cost. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on model extraction attacks and watermarking, which is not directly related to the practical applications of Large Language Models (LLMs) or their integration with technologies like knowledge graphs. + +--- + +## [Prompt-CAM: A Simpler Interpretable Transformer for Fine-Grained + Analysis](https://arxiv.org/abs/2501.09333v1) +**arXiv ID:** 2501.09333v1 + +**Abstract:** +> We present a simple usage of pre-trained Vision Transformers (ViTs) for +fine-grained analysis, aiming to identify and localize the traits that +distinguish visually similar categories, such as different bird species or dog +breeds. Pre-trained ViTs such as DINO have shown remarkable capabilities to +extract localized, informative features. However, using saliency maps like +Grad-CAM can hardly point out the traits: they often locate the whole object by +a blurred, coarse heatmap, not traits. We propose a novel approach Prompt Class +Attention Map (Prompt-CAM) to the rescue. Prompt-CAM learns class-specific +prompts to a pre-trained ViT and uses the corresponding outputs for +classification. To classify an image correctly, the true-class prompt must +attend to the unique image patches not seen in other classes' images, i.e., +traits. As such, the true class's multi-head attention maps reveal traits and +their locations. Implementation-wise, Prompt-CAM is almost a free lunch by +simply modifying the prediction head of Visual Prompt Tuning (VPT). This makes +Prompt-CAM fairly easy to train and apply, sharply contrasting other +interpretable methods that design specific models and training processes. It is +even simpler than the recently published INterpretable TRansformer (INTR), +whose encoder-decoder architecture prevents it from leveraging pre-trained +ViTs. Extensive empirical studies on a dozen datasets from various domains +(e.g., birds, fishes, insects, fungi, flowers, food, and cars) validate +Prompt-CAM superior interpretation capability. + +**Decision Explanation:** The paper primarily focuses on computer vision and image analysis using Vision Transformers, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Style4Rec: Enhancing Transformer-based E-commerce Recommendation Systems + with Style and Shopping Cart Information](https://arxiv.org/abs/2501.09354v1) +**arXiv ID:** 2501.09354v1 + +**Abstract:** +> Understanding users' product preferences is essential to the efficacy of a +recommendation system. Precision marketing leverages users' historical data to +discern these preferences and recommends products that align with them. +However, recent browsing and purchase records might better reflect current +purchasing inclinations. Transformer-based recommendation systems have made +strides in sequential recommendation tasks, but they often fall short in +utilizing product image style information and shopping cart data effectively. +In light of this, we propose Style4Rec, a transformer-based e-commerce +recommendation system that harnesses style and shopping cart information to +enhance existing transformer-based sequential product recommendation systems. +Style4Rec represents a significant step forward in personalized e-commerce +recommendations, outperforming benchmarks across various evaluation metrics. +Style4Rec resulted in notable improvements: HR@5 increased from 0.681 to 0.735, +NDCG@5 increased from 0.594 to 0.674, and MRR@5 increased from 0.559 to 0.654. +We tested our model using an e-commerce dataset from our partnering company and +found that it exceeded established transformer-based sequential recommendation +benchmarks across various evaluation metrics. Thus, Style4Rec presents a +significant step forward in personalized e-commerce recommendation systems. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on e-commerce recommendation systems, which is not explicitly mentioned in the criteria, and lacks clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Predicting Air Temperature from Volumetric Urban Morphology with Machine + Learning](https://arxiv.org/abs/2501.09469v1) +**arXiv ID:** 2501.09469v1 + +**Abstract:** +> In this study, we firstly introduce a method that converts CityGML data into +voxels which works efficiently and fast in high resolution for large scale +datasets such as cities but by sacrificing some building details to overcome +the limitations of previous voxelization methodologies that have been +computationally intensive and inefficient at transforming large-scale urban +areas into voxel representations for high resolution. Those voxelized 3D city +data from multiple cities and corresponding air temperature data are used to +develop a machine learning model. Before the model training, Gaussian blurring +is implemented on input data to consider spatial relationships, as a result the +correlation rate between air temperature and volumetric building morphology is +also increased after the Gaussian blurring. After the model training, the +prediction results are not just evaluated with Mean Square Error (MSE) but some +image similarity metrics such as Structural Similarity Index Measure (SSIM) and +Learned Perceptual Image Patch Similarity (LPIPS) that are able to detect and +consider spatial relations during the evaluation process. This trained model is +capable of predicting the spatial distribution of air temperature by using +building volume information of corresponding pixel as input. By doing so, this +research aims to assist urban planners in incorporating environmental +parameters into their planning strategies, thereby facilitating more +sustainable and inhabitable urban environments. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on predicting air temperature using machine learning and volumetric urban morphology, which is not related to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Class Incremental Fault Diagnosis under Limited Fault Data via + Supervised Contrastive Knowledge Distillation](https://arxiv.org/abs/2501.09525v1) +**arXiv ID:** 2501.09525v1 + +**Abstract:** +> Class-incremental fault diagnosis requires a model to adapt to new fault +classes while retaining previous knowledge. However, limited research exists +for imbalanced and long-tailed data. Extracting discriminative features from +few-shot fault data is challenging, and adding new fault classes often demands +costly model retraining. Moreover, incremental training of existing methods +risks catastrophic forgetting, and severe class imbalance can bias the model's +decisions toward normal classes. To tackle these issues, we introduce a +Supervised Contrastive knowledge distiLlation for class Incremental Fault +Diagnosis (SCLIFD) framework proposing supervised contrastive knowledge +distillation for improved representation learning capability and less +forgetting, a novel prioritized exemplar selection method for sample replay to +alleviate catastrophic forgetting, and the Random Forest Classifier to address +the class imbalance. Extensive experimentation on simulated and real-world +industrial datasets across various imbalance ratios demonstrates the +superiority of SCLIFD over existing approaches. Our code can be found at +https://github.com/Zhang-Henry/SCLIFD_TII. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on fault diagnosis, which is not a specified area of interest, and does not explicitly mention Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Text-driven Adaptation of Foundation Models for Few-shot Surgical + Workflow Analysis](https://arxiv.org/abs/2501.09555v1) +**arXiv ID:** 2501.09555v1 + +**Abstract:** +> Purpose: Surgical workflow analysis is crucial for improving surgical +efficiency and safety. However, previous studies rely heavily on large-scale +annotated datasets, posing challenges in cost, scalability, and reliance on +expert annotations. To address this, we propose Surg-FTDA (Few-shot Text-driven +Adaptation), designed to handle various surgical workflow analysis tasks with +minimal paired image-label data. + Methods: Our approach has two key components. First, Few-shot selection-based +modality alignment selects a small subset of images and aligns their embeddings +with text embeddings from the downstream task, bridging the modality gap. +Second, Text-driven adaptation leverages only text data to train a decoder, +eliminating the need for paired image-text data. This decoder is then applied +to aligned image embeddings, enabling image-related tasks without explicit +image-text pairs. + Results: We evaluate our approach to generative tasks (image captioning) and +discriminative tasks (triplet recognition and phase recognition). Results show +that Surg-FTDA outperforms baselines and generalizes well across downstream +tasks. + Conclusion: We propose a text-driven adaptation approach that mitigates the +modality gap and handles multiple downstream tasks in surgical workflow +analysis, with minimal reliance on large annotated datasets. The code and +dataset will be released in https://github.com/TingxuanSix/Surg-FTDA. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is excluded according to the criteria. + +--- + +## [IFRA: a machine learning-based Instrumented Fall Risk Assessment Scale + derived from Instrumented Timed Up and Go test in stroke patients](https://arxiv.org/abs/2501.09595v1) +**arXiv ID:** 2501.09595v1 + +**Abstract:** +> Effective fall risk assessment is critical for post-stroke patients. The +present study proposes a novel, data-informed fall risk assessment method based +on the instrumented Timed Up and Go (ITUG) test data, bringing in many mobility +measures that traditional clinical scales fail to capture. IFRA, which stands +for Instrumented Fall Risk Assessment, has been developed using a two-step +process: first, features with the highest predictive power among those +collected in a ITUG test have been identified using machine learning +techniques; then, a strategy is proposed to stratify patients into low, medium, +or high-risk strata. The dataset used in our analysis consists of 142 +participants, out of which 93 were used for training (15 synthetically +generated), 17 for validation and 32 to test the resulting IFRA scale (22 +non-fallers and 10 fallers). Features considered in the IFRA scale include gait +speed, vertical acceleration during sit-to-walk transition, and turning angular +velocity, which align well with established literature on the risk of fall in +neurological patients. In a comparison with traditional clinical scales such as +the traditional Timed Up & Go and the Mini-BESTest, IFRA demonstrates +competitive performance, being the only scale to correctly assign more than +half of the fallers to the high-risk stratum (Fischer's Exact test p = 0.004). +Despite the dataset's limited size, this is the first proof-of-concept study to +pave the way for future evidence regarding the use of IFRA tool for continuous +patient monitoring and fall prevention both in clinical stroke rehabilitation +and at home post-discharge. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, specifically fall risk assessment in stroke patients, which is excluded according to the criteria. + +--- + +## [Reducing the Sensitivity of Neural Physics Simulators to Mesh Topology + via Pretraining](https://arxiv.org/abs/2501.09597v1) +**arXiv ID:** 2501.09597v1 + +**Abstract:** +> Meshes are used to represent complex objects in high fidelity physics +simulators across a variety of domains, such as radar sensing and aerodynamics. +There is growing interest in using neural networks to accelerate physics +simulations, and also a growing body of work on applying neural networks +directly to irregular mesh data. Since multiple mesh topologies can represent +the same object, mesh augmentation is typically required to handle topological +variation when training neural networks. Due to the sensitivity of physics +simulators to small changes in mesh shape, it is challenging to use these +augmentations when training neural network-based physics simulators. In this +work, we show that variations in mesh topology can significantly reduce the +performance of neural network simulators. We evaluate whether pretraining can +be used to address this issue, and find that employing an established +autoencoder pretraining technique with graph embedding models reduces the +sensitivity of neural network simulators to variations in mesh topology. +Finally, we highlight future research directions that may further reduce neural +simulator sensitivity to mesh topology. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on neural physics simulators and mesh topology, with no clear connection to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Monte Carlo Tree Search with Velocity Obstacles for safe and efficient + motion planning in dynamic environments](https://arxiv.org/abs/2501.09649v1) +**arXiv ID:** 2501.09649v1 + +**Abstract:** +> Online motion planning is a challenging problem for intelligent robots moving +in dense environments with dynamic obstacles, e.g., crowds. In this work, we +propose a novel approach for optimal and safe online motion planning with +minimal information about dynamic obstacles. Specifically, our approach +requires only the current position of the obstacles and their maximum speed, +but it does not need any information about their exact trajectories or dynamic +model. The proposed methodology combines Monte Carlo Tree Search (MCTS), for +online optimal planning via model simulations, with Velocity Obstacles (VO), +for obstacle avoidance. We perform experiments in a cluttered simulated +environment with walls, and up to 40 dynamic obstacles moving with random +velocities and directions. With an ablation study, we show the key contribution +of VO in scaling up the efficiency of MCTS, selecting the safest and most +rewarding actions in the tree of simulations. Moreover, we show the superiority +of our methodology with respect to state-of-the-art planners, including +Non-linear Model Predictive Control (NMPC), in terms of improved collision +rate, computational and task performance. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on motion planning in dynamic environments, which does not involve Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Robin: a Suite of Multi-Scale Vision-Language Models and the CHIRP + Evaluation Benchmark](https://arxiv.org/abs/2501.09672v1) +**arXiv ID:** 2501.09672v1 + +**Abstract:** +> The proliferation of Vision-Language Models (VLMs) in the past several years +calls for rigorous and comprehensive evaluation methods and benchmarks. This +work analyzes existing VLM evaluation techniques, including automated metrics, +AI-based assessments, and human evaluations across diverse tasks. We first +introduce Robin - a novel suite of VLMs that we built by combining Large +Language Models (LLMs) and Vision Encoders (VEs) at multiple scales, and use +Robin to identify shortcomings of current evaluation approaches across scales. +Next, to overcome the identified limitations, we introduce CHIRP - a new long +form response benchmark we developed for more robust and complete VLM +evaluation. We provide open access to the Robin training code, model suite, and +CHIRP benchmark to promote reproducibility and advance VLM research. + +**Decision Explanation:** The paper primarily focuses on Vision-Language Models and their evaluation, which does not meet the criteria of focusing on Large Language Models, particularly in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Cueless EEG imagined speech for subject identification: dataset and + benchmarks](https://arxiv.org/abs/2501.09700v1) +**arXiv ID:** 2501.09700v1 + +**Abstract:** +> Electroencephalogram (EEG) signals have emerged as a promising modality for +biometric identification. While previous studies have explored the use of +imagined speech with semantically meaningful words for subject identification, +most have relied on additional visual or auditory cues. In this study, we +introduce a cueless EEG-based imagined speech paradigm, where subjects imagine +the pronunciation of semantically meaningful words without any external cues. +This innovative approach addresses the limitations of prior methods by +requiring subjects to select and imagine words from a predefined list +naturally. The dataset comprises over 4,350 trials from 11 subjects across five +sessions. We assess a variety of classification methods, including traditional +machine learning techniques such as Support Vector Machines (SVM) and XGBoost, +as well as time-series foundation models and deep learning architectures +specifically designed for EEG classification, such as EEG Conformer and Shallow +ConvNet. A session-based hold-out validation strategy was employed to ensure +reliable evaluation and prevent data leakage. Our results demonstrate +outstanding classification accuracy, reaching 97.93%. These findings highlight +the potential of cueless EEG paradigms for secure and reliable subject +identification in real-world applications, such as brain-computer interfaces +(BCIs). + +**Decision Explanation:** The paper primarily focuses on biometric identification using EEG signals and does not meet the criteria related to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [CyberMentor: AI Powered Learning Tool Platform to Address Diverse + Student Needs in Cybersecurity Education](https://arxiv.org/abs/2501.09709v1) +**arXiv ID:** 2501.09709v1 + +**Abstract:** +> Many non-traditional students in cybersecurity programs often lack access to +advice from peers, family members and professors, which can hinder their +educational experiences. Additionally, these students may not fully benefit +from various LLM-powered AI assistants due to issues like content relevance, +locality of advice, minimum expertise, and timing. This paper addresses these +challenges by introducing an application designed to provide comprehensive +support by answering questions related to knowledge, skills, and career +preparation advice tailored to the needs of these students. We developed a +learning tool platform, CyberMentor, to address the diverse needs and pain +points of students majoring in cybersecurity. Powered by agentic workflow and +Generative Large Language Models (LLMs), the platform leverages +Retrieval-Augmented Generation (RAG) for accurate and contextually relevant +information retrieval to achieve accessibility and personalization. We +demonstrated its value in addressing knowledge requirements for cybersecurity +education and for career marketability, in tackling skill requirements for +analytical and programming assignments, and in delivering real time on demand +learning support. Using three use scenarios, we showcased CyberMentor in +facilitating knowledge acquisition and career preparation and providing +seamless skill-based guidance and support. We also employed the LangChain +prompt-based evaluation methodology to evaluate the platform's impact, +confirming its strong performance in helpfulness, correctness, and +completeness. These results underscore the system's ability to support students +in developing practical cybersecurity skills while improving equity and +sustainability within higher education. Furthermore, CyberMentor's open-source +design allows for adaptation across other disciplines, fostering educational +innovation and broadening its potential impact. + +**Decision Explanation:** The paper primarily focuses on educational applications of AI, which is not explicitly excluded but also does not align well with the preferred areas of knowledge graphs, retrieval-augmented generation, or agentic AI in a broader sense beyond education, and it touches on diversity and equity, which is a social application of AI. + +--- + +## [Learnings from Scaling Visual Tokenizers for Reconstruction and + Generation](https://arxiv.org/abs/2501.09755v1) +**arXiv ID:** 2501.09755v1 + +**Abstract:** +> Visual tokenization via auto-encoding empowers state-of-the-art image and +video generative models by compressing pixels into a latent space. Although +scaling Transformer-based generators has been central to recent advances, the +tokenizer component itself is rarely scaled, leaving open questions about how +auto-encoder design choices influence both its objective of reconstruction and +downstream generative performance. Our work aims to conduct an exploration of +scaling in auto-encoders to fill in this blank. To facilitate this exploration, +we replace the typical convolutional backbone with an enhanced Vision +Transformer architecture for Tokenization (ViTok). We train ViTok on +large-scale image and video datasets far exceeding ImageNet-1K, removing data +constraints on tokenizer scaling. We first study how scaling the auto-encoder +bottleneck affects both reconstruction and generation -- and find that while it +is highly correlated with reconstruction, its relationship with generation is +more complex. We next explored the effect of separately scaling the +auto-encoders' encoder and decoder on reconstruction and generation +performance. Crucially, we find that scaling the encoder yields minimal gains +for either reconstruction or generation, while scaling the decoder boosts +reconstruction but the benefits for generation are mixed. Building on our +exploration, we design ViTok as a lightweight auto-encoder that achieves +competitive performance with state-of-the-art auto-encoders on ImageNet-1K and +COCO reconstruction tasks (256p and 512p) while outperforming existing +auto-encoders on 16-frame 128p video reconstruction for UCF-101, all with 2-5x +fewer FLOPs. When integrated with Diffusion Transformers, ViTok demonstrates +competitive performance on image generation for ImageNet-1K and sets new +state-of-the-art benchmarks for class-conditional video generation on UCF-101. + +**Decision Explanation:** The paper primarily focuses on video processing and image generation, which are not the main areas of interest, and does not clearly meet the criteria related to Large Language Models, knowledge graphs, or agentic AI. + +--- + +## [Exploring the Efficacy of Meta-Learning: Unveiling Superior Data + Diversity Utilization of MAML Over Pre-training](https://arxiv.org/abs/2501.08506v1) +**arXiv ID:** 2501.08506v1 + +**Abstract:** +> Currently, data and model size dominate the narrative in the training of +super-large, powerful models. However, there has been a lack of exploration on +the effect of other attributes of the training dataset on model performance. We +hypothesize that dataset diversity can impact the performance of vision models. +Our study shows positive correlations between test set accuracy and data +diversity, providing an argument for furthering the research of dataset +attributes beyond size. We analyzed pre-training and model-agnostic +meta-learning methods on twelve popular visual datasets (e.g., Omniglot, +CIFAR-FS, Aircraft) and five model configurations, including MAML variants with +different numbers of inner gradient steps and supervised learning. We show +moderate to strong positive correlations (R-squared: 0.15-0.42) between +accuracy and data diversity and weaker but significant correlations (R-squared: +~0.2) between loss and diversity. These findings support our hypothesis and +demonstrate a promising way for a deeper exploration of how formal data +diversity influences model performance. This initial study highlights the +potential of (Task2Vec) data diversity as a valuable measure in the rapidly +evolving field of large-scale learning and emphasizes that understanding the +dataset is key to building more powerful and generalizable models. + +**Decision Explanation:** The paper primarily focuses on vision models and dataset diversity, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Easing Seasickness through Attention Redirection with a + Mindfulness-Based Brain--Computer Interface](https://arxiv.org/abs/2501.08518v1) +**arXiv ID:** 2501.08518v1 + +**Abstract:** +> Seasickness is a prevalent issue that adversely impacts both passenger +experiences and the operational efficiency of maritime crews. While techniques +that redirect attention have proven effective in alleviating motion sickness +symptoms in terrestrial environments, applying similar strategies to manage +seasickness poses unique challenges due to the prolonged and intense motion +environment associated with maritime travel. In this study, we propose a +mindfulness brain-computer interface (BCI), specifically designed to redirect +attention with the aim of mitigating seasickness symptoms in real-world +settings. Our system utilizes a single-channel headband to capture prefrontal +EEG signals, which are then wirelessly transmitted to computing devices for the +assessment of mindfulness states. The results are transferred into real-time +feedback as mindfulness scores and audiovisual stimuli, facilitating a shift in +attentional focus from physiological discomfort to mindfulness practices. A +total of 43 individuals participated in a real-world maritime experiment +consisted of three sessions: a real-feedback mindfulness session, a resting +session, and a pseudofeedback mindfulness session. Notably, 81.39% of +participants reported that the mindfulness BCI intervention was effective, and +there was a significant reduction in the severity of seasickness, as measured +by the Misery Scale (MISC). Furthermore, EEG analysis revealed a decrease in +the theta/beta ratio, corresponding with the alleviation of seasickness +symptoms. A decrease in overall EEG band power during the real-feedback +mindfulness session suggests that the mindfulness BCI fosters a more tranquil +and downregulated state of brain activity. Together, this study presents a +novel nonpharmacological, portable, and effective approach for seasickness +intervention, with the potential to enhance the cruising experience for both +passengers and crews. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on a medical application of AI, specifically using a brain-computer interface to alleviate seasickness, which is not within the specified areas of interest such as knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Reinforcement Learning-Enhanced Procedural Generation for Dynamic + Narrative-Driven AR Experiences](https://arxiv.org/abs/2501.08552v1) +**arXiv ID:** 2501.08552v1 + +**Abstract:** +> Procedural Content Generation (PCG) is widely used to create scalable and +diverse environments in games. However, existing methods, such as the Wave +Function Collapse (WFC) algorithm, are often limited to static scenarios and +lack the adaptability required for dynamic, narrative-driven applications, +particularly in augmented reality (AR) games. This paper presents a +reinforcement learning-enhanced WFC framework designed for mobile AR +environments. By integrating environment-specific rules and dynamic tile weight +adjustments informed by reinforcement learning (RL), the proposed method +generates maps that are both contextually coherent and responsive to gameplay +needs. Comparative evaluations and user studies demonstrate that the framework +achieves superior map quality and delivers immersive experiences, making it +well-suited for narrative-driven AR games. Additionally, the method holds +promise for broader applications in education, simulation training, and +immersive extended reality (XR) experiences, where dynamic and adaptive +environments are critical. + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their applications, and instead discusses reinforcement learning-enhanced procedural generation for dynamic narrative-driven AR experiences, which does not meet the primary criteria. + +--- + +## [Towards Lightweight and Stable Zero-shot TTS with Self-distilled + Representation Disentanglement](https://arxiv.org/abs/2501.08566v1) +**arXiv ID:** 2501.08566v1 + +**Abstract:** +> Zero-shot Text-To-Speech (TTS) synthesis shows great promise for personalized +voice customization through voice cloning. However, current methods for +achieving zero-shot TTS heavily rely on large model scales and extensive +training datasets to ensure satisfactory performance and generalizability +across various speakers. This raises concerns regarding both deployment costs +and data security. In this paper, we present a lightweight and stable zero-shot +TTS system. We introduce a novel TTS architecture designed to effectively model +linguistic content and various speaker attributes from source speech and prompt +speech, respectively. Furthermore, we present a two-stage self-distillation +framework that constructs parallel data pairs for effectively disentangling +linguistic content and speakers from the perspective of training data. +Extensive experiments show that our system exhibits excellent performance and +superior stability on the zero-shot TTS tasks. Moreover, it shows markedly +superior computational efficiency, with RTFs of 0.13 and 0.012 on the CPU and +GPU, respectively. + +**Decision Explanation:** The paper primarily focuses on Text-To-Speech synthesis, which does not meet the criteria of practical applications of Large Language Models in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Sound Scene Synthesis at the DCASE 2024 Challenge](https://arxiv.org/abs/2501.08587v1) +**arXiv ID:** 2501.08587v1 + +**Abstract:** +> This paper presents Task 7 at the DCASE 2024 Challenge: sound scene +synthesis. Recent advances in sound synthesis and generative models have +enabled the creation of realistic and diverse audio content. We introduce a +standardized evaluation framework for comparing different sound scene synthesis +systems, incorporating both objective and subjective metrics. The challenge +attracted four submissions, which are evaluated using the Fr\'echet Audio +Distance (FAD) and human perceptual ratings. Our analysis reveals significant +insights into the current capabilities and limitations of sound scene synthesis +systems, while also highlighting areas for future improvement in this rapidly +evolving field. + +**Decision Explanation:** The paper primarily focuses on sound scene synthesis, which does not meet the criteria of having practical applications of Large Language Models (LLMs) or addressing areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [OpenMLDB: A Real-Time Relational Data Feature Computation System for + Online ML](https://arxiv.org/abs/2501.08591v1) +**arXiv ID:** 2501.08591v1 + +**Abstract:** +> Efficient and consistent feature computation is crucial for a wide range of +online ML applications. Typically, feature computation is divided into two +distinct phases, i.e., offline stage for model training and online stage for +model serving. These phases often rely on execution engines with different +interface languages and function implementations, causing significant +inconsistencies. Moreover, many online ML features involve complex time-series +computations (e.g., functions over varied-length table windows) that differ +from standard streaming and analytical queries. Existing data processing +systems (e.g., Spark, Flink, DuckDB) often incur multi-second latencies for +these computations, making them unsuitable for real-time online ML applications +that demand timely feature updates. + This paper presents OpenMLDB, a feature computation system deployed in +4Paradigm's SageOne platform and over 100 real scenarios. Technically, OpenMLDB +first employs a unified query plan generator for consistent computation results +across the offline and online stages, significantly reducing feature deployment +overhead. Second, OpenMLDB provides an online execution engine that resolves +performance bottlenecks caused by long window computations (via +pre-aggregation) and multi-table window unions (via data self-adjusting). It +also provides a high-performance offline execution engine with window parallel +optimization and time-aware data skew resolving. Third, OpenMLDB features a +compact data format and stream-focused indexing to maximize memory usage and +accelerate data access. Evaluations in testing and real workloads reveal +significant performance improvements and resource savings compared to the +baseline systems. The open community of OpenMLDB now has over 150 contributors +and gained 1.6k stars on GitHub. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on a real-time relational data feature computation system for online ML, without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Fine-grained Spatio-temporal Event Prediction with Self-adaptive Anchor + Graph](https://arxiv.org/abs/2501.08653v1) +**arXiv ID:** 2501.08653v1 + +**Abstract:** +> Event prediction tasks often handle spatio-temporal data distributed in a +large spatial area. Different regions in the area exhibit different +characteristics while having latent correlations. This spatial heterogeneity +and correlations greatly affect the spatio-temporal distributions of event +occurrences, which has not been addressed by state-of-the-art models. Learning +spatial dependencies of events in a continuous space is challenging due to its +fine granularity and a lack of prior knowledge. In this work, we propose a +novel Graph Spatio-Temporal Point Process (GSTPP) model for fine-grained event +prediction. It adopts an encoder-decoder architecture that jointly models the +state dynamics of spatially localized regions using neural Ordinary +Differential Equations (ODEs). The state evolution is built on the foundation +of a novel Self-Adaptive Anchor Graph (SAAG) that captures spatial +dependencies. By adaptively localizing the anchor nodes in the space and +jointly constructing the correlation edges between them, the SAAG enhances the +model's ability of learning complex spatial event patterns. The proposed GSTPP +model greatly improves the accuracy of fine-grained event prediction. Extensive +experimental results show that our method greatly improves the prediction +accuracy over existing spatio-temporal event prediction approaches. + +**Decision Explanation:** The paper does not meet the criteria as it does not focus on Large Language Models (LLMs) or their applications, and instead discusses spatio-temporal event prediction using a Graph Spatio-Temporal Point Process model. + +--- + +## [Self-supervised Transformation Learning for Equivariant Representations](https://arxiv.org/abs/2501.08712v1) +**arXiv ID:** 2501.08712v1 + +**Abstract:** +> Unsupervised representation learning has significantly advanced various +machine learning tasks. In the computer vision domain, state-of-the-art +approaches utilize transformations like random crop and color jitter to achieve +invariant representations, embedding semantically the same inputs despite +transformations. However, this can degrade performance in tasks requiring +precise features, such as localization or flower classification. To address +this, recent research incorporates equivariant representation learning, which +captures transformation-sensitive information. However, current methods depend +on transformation labels and thus struggle with interdependency and complex +transformations. We propose Self-supervised Transformation Learning (STL), +replacing transformation labels with transformation representations derived +from image pairs. The proposed method ensures transformation representation is +image-invariant and learns corresponding equivariant transformations, enhancing +performance without increased batch complexity. We demonstrate the approach's +effectiveness across diverse classification and detection tasks, outperforming +existing methods in 7 out of 11 benchmarks and excelling in detection. By +integrating complex transformations like AugMix, unusable by prior equivariant +methods, this approach enhances performance across tasks, underscoring its +adaptability and resilience. Additionally, its compatibility with various base +models highlights its flexibility and broad applicability. The code is +available at https://github.com/jaemyung-u/stl. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on computer vision and image processing, which is not within the specified areas of interest such as knowledge graphs, retrieval-augmented generation, or agentic AI, and does not explicitly mention Large Language Models (LLMs) or their applications. + +--- + +## [How Developers Interact with AI: A Taxonomy of Human-AI Collaboration in + Software Engineering](https://arxiv.org/abs/2501.08774v1) +**arXiv ID:** 2501.08774v1 + +**Abstract:** +> Artificial intelligence (AI), including large language models and generative +AI, is emerging as a significant force in software development, offering +developers powerful tools that span the entire development lifecycle. Although +software engineering research has extensively studied AI tools in software +development, the specific types of interactions between developers and these +AI-powered tools have only recently begun to receive attention. Understanding +and improving these interactions has the potential to improve productivity, +trust, and efficiency in AI-driven workflows. In this paper, we propose a +taxonomy of interaction types between developers and AI tools, identifying +eleven distinct interaction types, such as auto-complete code suggestions, +command-driven actions, and conversational assistance. Building on this +taxonomy, we outline a research agenda focused on optimizing AI interactions, +improving developer control, and addressing trust and usability challenges in +AI-assisted development. By establishing a structured foundation for studying +developer-AI interactions, this paper aims to stimulate research on creating +more effective, adaptive AI tools for software development. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on the interaction between developers and AI tools in software engineering, without clear practical applications of Large Language Models, experimental results, or comparison with state-of-the-art techniques in areas like knowledge graphs or retrieval-augmented generation. + +--- + +## [Networked Agents in the Dark: Team Value Learning under Partial + Observability](https://arxiv.org/abs/2501.08778v1) +**arXiv ID:** 2501.08778v1 + +**Abstract:** +> We propose a novel cooperative multi-agent reinforcement learning (MARL) +approach for networked agents. In contrast to previous methods that rely on +complete state information or joint observations, our agents must learn how to +reach shared objectives under partial observability. During training, they +collect individual rewards and approximate a team value function through local +communication, resulting in cooperative behavior. To describe our problem, we +introduce the networked dynamic partially observable Markov game framework, +where agents communicate over a switching topology communication network. Our +distributed method, DNA-MARL, uses a consensus mechanism for local +communication and gradient descent for local computation. DNA-MARL increases +the range of the possible applications of networked agents, being well-suited +for real world domains that impose privacy and where the messages may not reach +their recipients. We evaluate DNA-MARL across benchmark MARL scenarios. Our +results highlight the superior performance of DNA-MARL over previous methods. + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their applications, and instead proposes a novel cooperative multi-agent reinforcement learning approach, which does not meet the primary criteria. + +--- + +## [XMusic: Towards a Generalized and Controllable Symbolic Music Generation + Framework](https://arxiv.org/abs/2501.08809v1) +**arXiv ID:** 2501.08809v1 + +**Abstract:** +> In recent years, remarkable advancements in artificial intelligence-generated +content (AIGC) have been achieved in the fields of image synthesis and text +generation, generating content comparable to that produced by humans. However, +the quality of AI-generated music has not yet reached this standard, primarily +due to the challenge of effectively controlling musical emotions and ensuring +high-quality outputs. This paper presents a generalized symbolic music +generation framework, XMusic, which supports flexible prompts (i.e., images, +videos, texts, tags, and humming) to generate emotionally controllable and +high-quality symbolic music. XMusic consists of two core components, XProjector +and XComposer. XProjector parses the prompts of various modalities into +symbolic music elements (i.e., emotions, genres, rhythms and notes) within the +projection space to generate matching music. XComposer contains a Generator and +a Selector. The Generator generates emotionally controllable and melodious +music based on our innovative symbolic music representation, whereas the +Selector identifies high-quality symbolic music by constructing a multi-task +learning scheme involving quality assessment, emotion recognition, and genre +recognition tasks. In addition, we build XMIDI, a large-scale symbolic music +dataset that contains 108,023 MIDI files annotated with precise emotion and +genre labels. Objective and subjective evaluations show that XMusic +significantly outperforms the current state-of-the-art methods with impressive +music quality. Our XMusic has been awarded as one of the nine Highlights of +Collectibles at WAIC 2023. The project homepage of XMusic is +https://xmusic-project.github.io. + +**Decision Explanation:** The paper primarily focuses on music generation, which does not meet the criteria of having practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and also does not align with the specified areas of interest. + +--- + +## [SAIF: A Comprehensive Framework for Evaluating the Risks of Generative + AI in the Public Sector](https://arxiv.org/abs/2501.08814v1) +**arXiv ID:** 2501.08814v1 + +**Abstract:** +> The rapid adoption of generative AI in the public sector, encompassing +diverse applications ranging from automated public assistance to welfare +services and immigration processes, highlights its transformative potential +while underscoring the pressing need for thorough risk assessments. Despite its +growing presence, evaluations of risks associated with AI-driven systems in the +public sector remain insufficiently explored. Building upon an established +taxonomy of AI risks derived from diverse government policies and corporate +guidelines, we investigate the critical risks posed by generative AI in the +public sector while extending the scope to account for its multimodal +capabilities. In addition, we propose a Systematic dAta generatIon Framework +for evaluating the risks of generative AI (SAIF). SAIF involves four key +stages: breaking down risks, designing scenarios, applying jailbreak methods, +and exploring prompt types. It ensures the systematic and consistent generation +of prompt data, facilitating a comprehensive evaluation while providing a solid +foundation for mitigating the risks. Furthermore, SAIF is designed to +accommodate emerging jailbreak methods and evolving prompt types, thereby +enabling effective responses to unforeseen risk scenarios. We believe that this +study can play a crucial role in fostering the safe and responsible integration +of generative AI into the public sector. + +**Decision Explanation:** The paper primarily focuses on responsible AI application and AI ethics in the public sector, which is an excluded topic. + +--- + +## [IDEA: Image Description Enhanced CLIP-Adapter](https://arxiv.org/abs/2501.08816v1) +**arXiv ID:** 2501.08816v1 + +**Abstract:** +> CLIP (Contrastive Language-Image Pre-training) has attained great success in +pattern recognition and computer vision. Transferring CLIP to downstream tasks +(e.g. zero- or few-shot classification) is a hot topic in multimodal learning. +However, current studies primarily focus on either prompt learning for text or +adapter tuning for vision, without fully exploiting the complementary +information and correlations among image-text pairs. In this paper, we propose +an Image Description Enhanced CLIP-Adapter (IDEA) method to adapt CLIP to +few-shot image classification tasks. This method captures fine-grained features +by leveraging both visual features and textual descriptions of images. IDEA is +a training-free method for CLIP, and it can be comparable to or even exceeds +state-of-the-art models on multiple tasks. Furthermore, we introduce +Trainable-IDEA (T-IDEA), which extends IDEA by adding two lightweight learnable +components (i.e., a projector and a learnable latent space), further enhancing +the model's performance and achieving SOTA results on 11 datasets. As one +important contribution, we employ the Llama model and design a comprehensive +pipeline to generate textual descriptions for images of 11 datasets, resulting +in a total of 1,637,795 image-text pairs, named "IMD-11". Our code and data are +released at https://github.com/FourierAI/IDEA. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on computer vision and image classification tasks, without clear applications of Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Automatic tuning of communication protocols for vehicular ad hoc + networks using metaheuristics](https://arxiv.org/abs/2501.08847v1) +**arXiv ID:** 2501.08847v1 + +**Abstract:** +> The emerging field of vehicular ad hoc networks (VANETs) deals with a set of +communicating vehicles which are able to spontaneously interconnect without any +pre-existing infrastructure. In such kind of networks, it is crucial to make an +optimal configuration of the communication protocols previously to the final +network deployment. This way, a human designer can obtain an optimal QoS of the +network beforehand. The problem we consider in this work lies in configuring +the File Transfer protocol Configuration (FTC) with the aim of optimizing the +transmission time, the number of lost packets, and the amount of data +transferred in realistic VANET scenarios. We face the FTC with five +representative state-of-the-art optimization techniques and compare their +performance. These algorithms are: Particle Swarm Optimization (PSO), +Differential Evolution (DE), Genetic Algorithm (GA), Evolutionary Strategy +(ES), and Simulated Annealing (SA). For our tests, two typical environment +instances of VANETs for Urban and Highway scenarios have been defined. The +experiments using ns- 2 (a well-known realistic VANET simulator) reveal that +PSO outperforms all the compared algorithms for both studied VANET instances. + +**Decision Explanation:** The paper does not meet the criteria as it does not focus on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [RouteNet-Gauss: Hardware-Enhanced Network Modeling with Machine Learning](https://arxiv.org/abs/2501.08848v1) +**arXiv ID:** 2501.08848v1 + +**Abstract:** +> Network simulation is pivotal in network modeling, assisting with tasks +ranging from capacity planning to performance estimation. Traditional +approaches such as Discrete Event Simulation (DES) face limitations in terms of +computational cost and accuracy. This paper introduces RouteNet-Gauss, a novel +integration of a testbed network with a Machine Learning (ML) model to address +these challenges. By using the testbed as a hardware accelerator, +RouteNet-Gauss generates training datasets rapidly and simulates network +scenarios with high fidelity to real-world conditions. Experimental results +show that RouteNet-Gauss significantly reduces prediction errors by up to 95% +and achieves a 488x speedup in inference time compared to state-of-the-art +DES-based methods. RouteNet-Gauss's modular architecture is dynamically +constructed based on the specific characteristics of the network scenario, such +as topology and routing. This enables it to understand and generalize to +different network configurations beyond those seen during training, including +networks up to 10x larger. Additionally, it supports Temporal Aggregated +Performance Estimation (TAPE), providing configurable temporal granularity and +maintaining high accuracy in flow performance metrics. This approach shows +promise in improving both simulation efficiency and accuracy, offering a +valuable tool for network operators. + +**Decision Explanation:** The paper primarily focuses on network modeling and simulation, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Graph Counterfactual Explainable AI via Latent Space Traversal](https://arxiv.org/abs/2501.08850v1) +**arXiv ID:** 2501.08850v1 + +**Abstract:** +> Explaining the predictions of a deep neural network is a nontrivial task, yet +high-quality explanations for predictions are often a prerequisite for +practitioners to trust these models. Counterfactual explanations aim to explain +predictions by finding the ''nearest'' in-distribution alternative input whose +prediction changes in a pre-specified way. However, it remains an open question +how to define this nearest alternative input, whose solution depends on both +the domain (e.g. images, graphs, tabular data, etc.) and the specific +application considered. For graphs, this problem is complicated i) by their +discrete nature, as opposed to the continuous nature of state-of-the-art graph +classifiers; and ii) by the node permutation group acting on the graphs. We +propose a method to generate counterfactual explanations for any differentiable +black-box graph classifier, utilizing a case-specific permutation equivariant +graph variational autoencoder. We generate counterfactual explanations in a +continuous fashion by traversing the latent space of the autoencoder across the +classification boundary of the classifier, allowing for seamless integration of +discrete graph structure and continuous graph attributes. We empirically +validate the approach on three graph datasets, showing that our model is +consistently high-performing and more robust than the baselines. + +**Decision Explanation:** The paper does not meet the criteria for practical applications of Large Language Models (LLMs) and does not mention LLMs, knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, focusing instead on explainable AI for graph classifiers. + +--- + +## [ARMOR: Shielding Unlearnable Examples against Data Augmentation](https://arxiv.org/abs/2501.08862v1) +**arXiv ID:** 2501.08862v1 + +**Abstract:** +> Private data, when published online, may be collected by unauthorized parties +to train deep neural networks (DNNs). To protect privacy, defensive noises can +be added to original samples to degrade their learnability by DNNs. Recently, +unlearnable examples are proposed to minimize the training loss such that the +model learns almost nothing. However, raw data are often pre-processed before +being used for training, which may restore the private information of protected +data. In this paper, we reveal the data privacy violation induced by data +augmentation, a commonly used data pre-processing technique to improve model +generalization capability, which is the first of its kind as far as we are +concerned. We demonstrate that data augmentation can significantly raise the +accuracy of the model trained on unlearnable examples from 21.3% to 66.1%. To +address this issue, we propose a defense framework, dubbed ARMOR, to protect +data privacy from potential breaches of data augmentation. To overcome the +difficulty of having no access to the model training process, we design a +non-local module-assisted surrogate model that better captures the effect of +data augmentation. In addition, we design a surrogate augmentation selection +strategy that maximizes distribution alignment between augmented and +non-augmented samples, to choose the optimal augmentation strategy for each +class. We also use a dynamic step size adjustment algorithm to enhance the +defensive noise generation process. Extensive experiments are conducted on 4 +datasets and 5 data augmentation methods to verify the performance of ARMOR. +Comparisons with 6 state-of-the-art defense methods have demonstrated that +ARMOR can preserve the unlearnability of protected private data under data +augmentation. ARMOR reduces the test accuracy of the model trained on augmented +protected samples by as much as 60% more than baselines. + +**Decision Explanation:** The paper primarily focuses on protecting data privacy from potential breaches of data augmentation, which does not meet the criteria of practical applications of Large Language Models (LLMs) or real-world applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Karatsuba Matrix Multiplication and its Efficient Custom Hardware + Implementations](https://arxiv.org/abs/2501.08889v1) +**arXiv ID:** 2501.08889v1 + +**Abstract:** +> While the Karatsuba algorithm reduces the complexity of large integer +multiplication, the extra additions required minimize its benefits for smaller +integers of more commonly-used bitwidths. In this work, we propose the +extension of the scalar Karatsuba multiplication algorithm to matrix +multiplication, showing how this maintains the reduction in multiplication +complexity of the original Karatsuba algorithm while reducing the complexity of +the extra additions. Furthermore, we propose new matrix multiplication hardware +architectures for efficiently exploiting this extension of the Karatsuba +algorithm in custom hardware. We show that the proposed algorithm and hardware +architectures can provide real area or execution time improvements for integer +matrix multiplication compared to scalar Karatsuba or conventional matrix +multiplication algorithms, while also supporting implementation through proven +systolic array and conventional multiplier architectures at the core. We +provide a complexity analysis of the algorithm and architectures and evaluate +the proposed designs both in isolation and in an end-to-end deep learning +accelerator system compared to baseline designs and prior state-of-the-art +works implemented on the same type of compute platform, demonstrating their +ability to increase the performance-per-area of matrix multiplication hardware. + +**Decision Explanation:** The paper does not meet the criteria as it focuses on matrix multiplication and custom hardware implementations, which are not related to Large Language Models (LLMs) or their practical applications. + +--- + +## [Computing Game Symmetries and Equilibria That Respect Them](https://arxiv.org/abs/2501.08905v1) +**arXiv ID:** 2501.08905v1 + +**Abstract:** +> Strategic interactions can be represented more concisely, and analyzed and +solved more efficiently, if we are aware of the symmetries within the +multiagent system. Symmetries also have conceptual implications, for example +for equilibrium selection. We study the computational complexity of identifying +and using symmetries. Using the classical framework of normal-form games, we +consider game symmetries that can be across some or all players and/or actions. +We find a strong connection between game symmetries and graph automorphisms, +yielding graph automorphism and graph isomorphism completeness results for +characterizing the symmetries present in a game. On the other hand, we also +show that the problem becomes polynomial-time solvable when we restrict the +consideration of actions in one of two ways. + Next, we investigate when exactly game symmetries can be successfully +leveraged for Nash equilibrium computation. We show that finding a Nash +equilibrium that respects a given set of symmetries is PPAD- and CLS-complete +in general-sum and team games respectively -- that is, exactly as hard as +Brouwer fixed point and gradient descent problems. Finally, we present +polynomial-time methods for the special cases where we are aware of a vast +number of symmetries, or where the game is two-player zero-sum and we do not +even know the symmetries. + +**Decision Explanation:** The paper does not meet the criteria as it does not focus on Large Language Models (LLMs) or their applications, and does not mention areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Trusted Machine Learning Models Unlock Private Inference for Problems + Currently Infeasible with Cryptography](https://arxiv.org/abs/2501.08970v1) +**arXiv ID:** 2501.08970v1 + +**Abstract:** +> We often interact with untrusted parties. Prioritization of privacy can limit +the effectiveness of these interactions, as achieving certain goals +necessitates sharing private data. Traditionally, addressing this challenge has +involved either seeking trusted intermediaries or constructing cryptographic +protocols that restrict how much data is revealed, such as multi-party +computations or zero-knowledge proofs. While significant advances have been +made in scaling cryptographic approaches, they remain limited in terms of the +size and complexity of applications they can be used for. In this paper, we +argue that capable machine learning models can fulfill the role of a trusted +third party, thus enabling secure computations for applications that were +previously infeasible. In particular, we describe Trusted Capable Model +Environments (TCMEs) as an alternative approach for scaling secure computation, +where capable machine learning model(s) interact under input/output +constraints, with explicit information flow control and explicit statelessness. +This approach aims to achieve a balance between privacy and computational +efficiency, enabling private inference where classical cryptographic solutions +are currently infeasible. We describe a number of use cases that are enabled by +TCME, and show that even some simple classic cryptographic problems can already +be solved with TCME. Finally, we outline current limitations and discuss the +path forward in implementing them. + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their applications, and instead discusses machine learning models for private inference, which does not meet the primary criteria. + +--- + +## [Personality Modeling for Persuasion of Misinformation using AI Agent](https://arxiv.org/abs/2501.08985v1) +**arXiv ID:** 2501.08985v1 + +**Abstract:** +> The proliferation of misinformation on social media platforms has highlighted +the need to understand how individual personality traits influence +susceptibility to and propagation of misinformation. This study employs an +innovative agent-based modeling approach to investigate the relationship +between personality traits and misinformation dynamics. Using six AI agents +embodying different dimensions of the Big Five personality traits +(Extraversion, Agreeableness, and Neuroticism), we simulated interactions +across six diverse misinformation topics. The experiment, implemented through +the AgentScope framework using the GLM-4-Flash model, generated 90 unique +interactions, revealing complex patterns in how personality combinations affect +persuasion and resistance to misinformation. Our findings demonstrate that +analytical and critical personality traits enhance effectiveness in +evidence-based discussions, while non-aggressive persuasion strategies show +unexpected success in misinformation correction. Notably, agents with critical +traits achieved a 59.4% success rate in HIV-related misinformation discussions, +while those employing non-aggressive approaches maintained consistent +persuasion rates above 40% across different personality combinations. The study +also revealed a non-transitive pattern in persuasion effectiveness, challenging +conventional assumptions about personality-based influence. These results +provide crucial insights for developing personality-aware interventions in +digital environments and suggest that effective misinformation countermeasures +should prioritize emotional connection and trust-building over confrontational +approaches. The findings contribute to both theoretical understanding of +personality-misinformation dynamics and practical strategies for combating +misinformation in social media contexts. + +**Decision Explanation:** The paper primarily focuses on social applications of AI in regard to misinformation, which is not a preferred area of focus according to the criteria. + +--- + +## [AI-RAN: Transforming RAN with AI-driven Computing Infrastructure](https://arxiv.org/abs/2501.09007v1) +**arXiv ID:** 2501.09007v1 + +**Abstract:** +> The radio access network (RAN) landscape is undergoing a transformative shift +from traditional, communication-centric infrastructures towards converged +compute-communication platforms. This article introduces AI-RAN which +integrates both RAN and artificial intelligence (AI) workloads on the same +infrastructure. By doing so, AI-RAN not only meets the performance demands of +future networks but also improves asset utilization. We begin by examining how +RANs have evolved beyond mobile broadband towards AI-RAN and articulating +manifestations of AI-RAN into three forms: AI-for-RAN, AI-on-RAN, and +AI-and-RAN. Next, we identify the key requirements and enablers for the +convergence of communication and computing in AI-RAN. We then provide a +reference architecture for advancing AI-RAN from concept to practice. To +illustrate the practical potential of AI-RAN, we present a proof-of-concept +that concurrently processes RAN and AI workloads utilizing NVIDIA Grace-Hopper +GH200 servers. Finally, we conclude the article by outlining future work +directions to guide further developments of AI-RAN. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on the integration of AI with radio access networks (RAN) and does not explicitly mention Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Multimodal LLMs Can Reason about Aesthetics in Zero-Shot](https://arxiv.org/abs/2501.09012v1) +**arXiv ID:** 2501.09012v1 + +**Abstract:** +> We present the first study on how Multimodal LLMs' (MLLMs) reasoning ability +shall be elicited to evaluate the aesthetics of artworks. To facilitate this +investigation, we construct MM-StyleBench, a novel high-quality dataset for +benchmarking artistic stylization. We then develop a principled method for +human preference modeling and perform a systematic correlation analysis between +MLLMs' responses and human preference. Our experiments reveal an inherent +hallucination issue of MLLMs in art evaluation, associated with response +subjectivity. ArtCoT is proposed, demonstrating that art-specific task +decomposition and the use of concrete language boost MLLMs' reasoning ability +for aesthetics. Our findings offer valuable insights into MLLMs for art and can +benefit a wide range of downstream applications, such as style transfer and +artistic image generation. Code available at +https://github.com/songrise/MLLM4Art. + +**Decision Explanation:** The paper primarily focuses on the application of Multimodal LLMs in evaluating the aesthetics of artworks, which does not meet the criteria of having practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and also does not align with the specified exclusion criteria in a way that would suggest a strong relevance to the desired topics. + +--- + +## [Spatio-Temporal Foundation Models: Vision, Challenges, and Opportunities](https://arxiv.org/abs/2501.09045v1) +**arXiv ID:** 2501.09045v1 + +**Abstract:** +> Foundation models have revolutionized artificial intelligence, setting new +benchmarks in performance and enabling transformative capabilities across a +wide range of vision and language tasks. However, despite the prevalence of +spatio-temporal data in critical domains such as transportation, public health, +and environmental monitoring, spatio-temporal foundation models (STFMs) have +not yet achieved comparable success. In this paper, we articulate a vision for +the future of STFMs, outlining their essential characteristics and the +generalization capabilities necessary for broad applicability. We critically +assess the current state of research, identifying gaps relative to these ideal +traits, and highlight key challenges that impede their progress. Finally, we +explore potential opportunities and directions to advance research towards the +aim of effective and broadly applicable STFMs. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on vision tasks and spatio-temporal foundation models, which is not directly related to Large Language Models (LLMs) or their practical applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Dynamic-Aware Spatio-temporal Representation Learning for Dynamic MRI + Reconstruction](https://arxiv.org/abs/2501.09049v1) +**arXiv ID:** 2501.09049v1 + +**Abstract:** +> Dynamic MRI reconstruction, one of inverse problems, has seen a surge by the +use of deep learning techniques. Especially, the practical difficulty of +obtaining ground truth data has led to the emergence of unsupervised learning +approaches. A recent promising method among them is implicit neural +representation (INR), which defines the data as a continuous function that maps +coordinate values to the corresponding signal values. This allows for filling +in missing information only with incomplete measurements and solving the +inverse problem effectively. Nevertheless, previous works incorporating this +method have faced drawbacks such as long optimization time and the need for +extensive hyperparameter tuning. To address these issues, we propose +Dynamic-Aware INR (DA-INR), an INR-based model for dynamic MRI reconstruction +that captures the spatial and temporal continuity of dynamic MRI data in the +image domain and explicitly incorporates the temporal redundancy of the data +into the model structure. As a result, DA-INR outperforms other models in +reconstruction quality even at extreme undersampling ratios while significantly +reducing optimization time and requiring minimal hyperparameter tuning. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, specifically dynamic MRI reconstruction, which is not aligned with the specified criteria. + +--- + +## [SteLLA: A Structured Grading System Using LLMs with RAG](https://arxiv.org/abs/2501.09092v1) +**arXiv ID:** 2501.09092v1 + +**Abstract:** +> Large Language Models (LLMs) have shown strong general capabilities in many +applications. However, how to make them reliable tools for some specific tasks +such as automated short answer grading (ASAG) remains a challenge. We present +SteLLA (Structured Grading System Using LLMs with RAG) in which a) Retrieval +Augmented Generation (RAG) approach is used to empower LLMs specifically on the +ASAG task by extracting structured information from the highly relevant and +reliable external knowledge based on the instructor-provided reference answer +and rubric, b) an LLM performs a structured and question-answering-based +evaluation of student answers to provide analytical grades and feedback. A +real-world dataset that contains students' answers in an exam was collected +from a college-level Biology course. Experiments show that our proposed system +can achieve substantial agreement with the human grader while providing +break-down grades and feedback on all the knowledge points examined in the +problem. A qualitative and error analysis of the feedback generated by GPT4 +shows that GPT4 is good at capturing facts while may be prone to inferring too +much implication from the given text in the grading task which provides +insights into the usage of LLMs in the ASAG system. + +**Decision Explanation:** The paper primarily focuses on educational applications, specifically automated short answer grading, which is not explicitly mentioned as a desired area of focus, and it does not clearly meet the criteria related to knowledge graphs, retrieval-augmented generation, or agentic AI in a way that aligns with the specified areas of interest. + +--- + +## [Tracking the Takes and Trajectories of English-Language News Narratives + across Trustworthy and Worrisome Websites](https://arxiv.org/abs/2501.09102v1) +**arXiv ID:** 2501.09102v1 + +**Abstract:** +> Understanding how misleading and outright false information enters news +ecosystems remains a difficult challenge that requires tracking how narratives +spread across thousands of fringe and mainstream news websites. To do this, we +introduce a system that utilizes encoder-based large language models and +zero-shot stance detection to scalably identify and track news narratives and +their attitudes across over 4,000 factually unreliable, mixed-reliability, and +factually reliable English-language news websites. Running our system over an +18 month period, we track the spread of 146K news stories. Using network-based +interference via the NETINF algorithm, we show that the paths of news +narratives and the stances of websites toward particular entities can be used +to uncover slanted propaganda networks (e.g., anti-vaccine and anti-Ukraine) +and to identify the most influential websites in spreading these attitudes in +the broader news ecosystem. We hope that increased visibility into our +distributed news ecosystem can help with the reporting and fact-checking of +propaganda and disinformation. + +**Decision Explanation:** The paper primarily focuses on social applications of AI in regard to misinformation and disinformation, which is similar to issues of social harm, and does not meet the criteria of having practical applications of Large Language Models in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [A Non-autoregressive Model for Joint STT and TTS](https://arxiv.org/abs/2501.09104v1) +**arXiv ID:** 2501.09104v1 + +**Abstract:** +> In this paper, we take a step towards jointly modeling automatic speech +recognition (STT) and speech synthesis (TTS) in a fully non-autoregressive way. +We develop a novel multimodal framework capable of handling the speech and text +modalities as input either individually or together. The proposed model can +also be trained with unpaired speech or text data owing to its multimodal +nature. We further propose an iterative refinement strategy to improve the STT +and TTS performance of our model such that the partial hypothesis at the output +can be fed back to the input of our model, thus iteratively improving both STT +and TTS predictions. We show that our joint model can effectively perform both +STT and TTS tasks, outperforming the STT-specific baseline in all tasks and +performing competitively with the TTS-specific baseline across a wide range of +evaluation metrics. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on speech recognition and synthesis, which does not involve Large Language Models or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Towards Understanding Extrapolation: a Causal Lens](https://arxiv.org/abs/2501.09163v1) +**arXiv ID:** 2501.09163v1 + +**Abstract:** +> Canonical work handling distribution shifts typically necessitates an entire +target distribution that lands inside the training distribution. However, +practical scenarios often involve only a handful of target samples, potentially +lying outside the training support, which requires the capability of +extrapolation. In this work, we aim to provide a theoretical understanding of +when extrapolation is possible and offer principled methods to achieve it +without requiring an on-support target distribution. To this end, we formulate +the extrapolation problem with a latent-variable model that embodies the +minimal change principle in causal mechanisms. Under this formulation, we cast +the extrapolation problem into a latent-variable identification problem. We +provide realistic conditions on shift properties and the estimation objectives +that lead to identification even when only one off-support target sample is +available, tackling the most challenging scenarios. Our theory reveals the +intricate interplay between the underlying manifold's smoothness and the shift +properties. We showcase how our theoretical results inform the design of +practical adaptation algorithms. Through experiments on both synthetic and +real-world data, we validate our theoretical findings and their practical +implications. + +**Decision Explanation:** The paper does not explicitly focus on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the primary criteria for selection. + +--- + +## [A Blockchain-Enabled Approach to Cross-Border Compliance and Trust](https://arxiv.org/abs/2501.09182v1) +**arXiv ID:** 2501.09182v1 + +**Abstract:** +> As artificial intelligence (AI) systems become increasingly integral to +critical infrastructure and global operations, the need for a unified, +trustworthy governance framework is more urgent that ever. This paper proposes +a novel approach to AI governance, utilizing blockchain and distributed ledger +technologies (DLT) to establish a decentralized, globally recognized framework +that ensures security, privacy, and trustworthiness of AI systems across +borders. The paper presents specific implementation scenarios within the +financial sector, outlines a phased deployment timeline over the next decade, +and addresses potential challenges with solutions grounded in current research. +By synthesizing advancements in blockchain, AI ethics, and cybersecurity, this +paper offers a comprehensive roadmap for a decentralized AI governance +framework capable of adapting to the complex and evolving landscape of global +AI regulation. + +**Decision Explanation:** The paper primarily focuses on AI governance, blockchain, and ethics, which does not meet the criteria of having practical applications of Large Language Models (LLMs) or discussing real-world applications and challenges involving LLMs. + +--- + +## [Patch-aware Vector Quantized Codebook Learning for Unsupervised Visual + Defect Detection](https://arxiv.org/abs/2501.09187v1) +**arXiv ID:** 2501.09187v1 + +**Abstract:** +> Unsupervised visual defect detection is critical in industrial applications, +requiring a representation space that captures normal data features while +detecting deviations. Achieving a balance between expressiveness and +compactness is challenging; an overly expressive space risks inefficiency and +mode collapse, impairing detection accuracy. We propose a novel approach using +an enhanced VQ-VAE framework optimized for unsupervised defect detection. Our +model introduces a patch-aware dynamic code assignment scheme, enabling +context-sensitive code allocation to optimize spatial representation. This +strategy enhances normal-defect distinction and improves detection accuracy +during inference. Experiments on MVTecAD, BTAD, and MTSD datasets show our +method achieves state-of-the-art performance. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on unsupervised visual defect detection, which is not directly related to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Adaptive Law-Based Transformation (ALT): A Lightweight Feature + Representation for Time Series Classification](https://arxiv.org/abs/2501.09217v1) +**arXiv ID:** 2501.09217v1 + +**Abstract:** +> Time series classification (TSC) is fundamental in numerous domains, +including finance, healthcare, and environmental monitoring. However, +traditional TSC methods often struggle with the inherent complexity and +variability of time series data. Building on our previous work with the linear +law-based transformation (LLT) - which improved classification accuracy by +transforming the feature space based on key data patterns - we introduce +adaptive law-based transformation (ALT). ALT enhances LLT by incorporating +variable-length shifted time windows, enabling it to capture distinguishing +patterns of various lengths and thereby handle complex time series more +effectively. By mapping features into a linearly separable space, ALT provides +a fast, robust, and transparent solution that achieves state-of-the-art +performance with only a few hyperparameters. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on time series classification and does not mention Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Foundations of Large Language Models](https://arxiv.org/abs/2501.09223v1) +**arXiv ID:** 2501.09223v1 + +**Abstract:** +> This is a book about large language models. As indicated by the title, it +primarily focuses on foundational concepts rather than comprehensive coverage +of all cutting-edge technologies. The book is structured into four main +chapters, each exploring a key area: pre-training, generative models, prompting +techniques, and alignment methods. It is intended for college students, +professionals, and practitioners in natural language processing and related +fields, and can serve as a reference for anyone interested in large language +models. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on foundational concepts of large language models rather than practical applications, experimental results, or comparisons with state-of-the-art techniques. + +--- + +## [Large Language Model is Secretly a Protein Sequence Optimizer](https://arxiv.org/abs/2501.09274v1) +**arXiv ID:** 2501.09274v1 + +**Abstract:** +> We consider the protein sequence engineering problem, which aims to find +protein sequences with high fitness levels, starting from a given wild-type +sequence. Directed evolution has been a dominating paradigm in this field which +has an iterative process to generate variants and select via experimental +feedback. We demonstrate large language models (LLMs), despite being trained on +massive texts, are secretly protein sequence optimizers. With a directed +evolutionary method, LLM can perform protein engineering through Pareto and +experiment-budget constrained optimization, demonstrating success on both +synthetic and experimental fitness landscapes. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, specifically protein sequence engineering, which is excluded according to the criteria. + +--- + +## [LAVCap: LLM-based Audio-Visual Captioning using Optimal Transport](https://arxiv.org/abs/2501.09291v1) +**arXiv ID:** 2501.09291v1 + +**Abstract:** +> Automated audio captioning is a task that generates textual descriptions for +audio content, and recent studies have explored using visual information to +enhance captioning quality. However, current methods often fail to effectively +fuse audio and visual data, missing important semantic cues from each modality. +To address this, we introduce LAVCap, a large language model (LLM)-based +audio-visual captioning framework that effectively integrates visual +information with audio to improve audio captioning performance. LAVCap employs +an optimal transport-based alignment loss to bridge the modality gap between +audio and visual features, enabling more effective semantic extraction. +Additionally, we propose an optimal transport attention module that enhances +audio-visual fusion using an optimal transport assignment map. Combined with +the optimal training strategy, experimental results demonstrate that each +component of our framework is effective. LAVCap outperforms existing +state-of-the-art methods on the AudioCaps dataset, without relying on large +datasets or post-processing. Code is available at +https://github.com/NAVER-INTEL-Co-Lab/gaudi-lavcap. + +**Decision Explanation:** The paper primarily focuses on audio-visual captioning, which does not meet the criteria of having practical applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, and also does not discuss real-world applications or challenges involving LLMs in autonomous or agentic AI scenarios. + +--- + +## [Understanding Mental Health Content on Social Media and Its Effect + Towards Suicidal Ideation](https://arxiv.org/abs/2501.09309v1) +**arXiv ID:** 2501.09309v1 + +**Abstract:** +> This review underscores the critical need for effective strategies to +identify and support individuals with suicidal ideation, exploiting +technological innovations in ML and DL to further suicide prevention efforts. +The study details the application of these technologies in analyzing vast +amounts of unstructured social media data to detect linguistic patterns, +keywords, phrases, tones, and contextual cues associated with suicidal +thoughts. It explores various ML and DL models like SVMs, CNNs, LSTM, neural +networks, and their effectiveness in interpreting complex data patterns and +emotional nuances within text data. The review discusses the potential of these +technologies to serve as a life-saving tool by identifying at-risk individuals +through their digital traces. Furthermore, it evaluates the real-world +effectiveness, limitations, and ethical considerations of employing these +technologies for suicide prevention, stressing the importance of responsible +development and usage. The study aims to fill critical knowledge gaps by +analyzing recent studies, methodologies, tools, and techniques in this field. +It highlights the importance of synthesizing current literature to inform +practical tools and suicide prevention efforts, guiding innovation in reliable, +ethical systems for early intervention. This research synthesis evaluates the +intersection of technology and mental health, advocating for the ethical and +responsible application of ML, DL, and NLP to offer life-saving potential +worldwide while addressing challenges like generalizability, biases, privacy, +and the need for further research to ensure these technologies do not +exacerbate existing inequities and harms. + +**Decision Explanation:** The paper primarily focuses on social applications of AI in regard to mental health and suicidal ideation, which is one of the excluded topics. + +--- + +## [Shape-Based Single Object Classification Using Ensemble Method + Classifiers](https://arxiv.org/abs/2501.09311v1) +**arXiv ID:** 2501.09311v1 + +**Abstract:** +> Nowadays, more and more images are available. Annotation and retrieval of the +images pose classification problems, where each class is defined as the group +of database images labelled with a common semantic label. Various systems have +been proposed for content-based retrieval, as well as for image classification +and indexing. In this paper, a hierarchical classification framework has been +proposed for bridging the semantic gap effectively and achieving multi-category +image classification. A well known pre-processing and post-processing method +was used and applied to three problems; image segmentation, object +identification and image classification. The method was applied to classify +single object images from Amazon and Google datasets. The classification was +tested for four different classifiers; BayesNetwork (BN), Random Forest (RF), +Bagging and Vote. The estimated classification accuracies ranged from 20% to +99% (using 10-fold cross validation). The Bagging classifier presents the best +performance, followed by the Random Forest classifier. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on image classification and does not involve Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [ELM-DeepONets: Backpropagation-Free Training of Deep Operator Networks + via Extreme Learning Machines](https://arxiv.org/abs/2501.09395v1) +**arXiv ID:** 2501.09395v1 + +**Abstract:** +> Deep Operator Networks (DeepONets) are among the most prominent frameworks +for operator learning, grounded in the universal approximation theorem for +operators. However, training DeepONets typically requires significant +computational resources. To address this limitation, we propose ELM-DeepONets, +an Extreme Learning Machine (ELM) framework for DeepONets that leverages the +backpropagation-free nature of ELM. By reformulating DeepONet training as a +least-squares problem for newly introduced parameters, the ELM-DeepONet +approach significantly reduces training complexity. Validation on benchmark +problems, including nonlinear ODEs and PDEs, demonstrates that the proposed +method not only achieves superior accuracy but also drastically reduces +computational costs. This work offers a scalable and efficient alternative for +operator learning in scientific computing. + +**Decision Explanation:** The paper does not meet the criteria as it focuses on Deep Operator Networks and Extreme Learning Machines, which are not directly related to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Dynamic Neural Style Transfer for Artistic Image Generation using VGG19](https://arxiv.org/abs/2501.09420v1) +**arXiv ID:** 2501.09420v1 + +**Abstract:** +> Throughout history, humans have created remarkable works of art, but +artificial intelligence has only recently started to make strides in generating +visually compelling art. Breakthroughs in the past few years have focused on +using convolutional neural networks (CNNs) to separate and manipulate the +content and style of images, applying texture synthesis techniques. +Nevertheless, a number of current techniques continue to encounter obstacles, +including lengthy processing times, restricted choices of style images, and the +inability to modify the weight ratio of styles. We proposed a neural style +transfer system that can add various artistic styles to a desired image to +address these constraints allowing flexible adjustments to style weight ratios +and reducing processing time. The system uses the VGG19 model for feature +extraction, ensuring high-quality, flexible stylization without compromising +content integrity. + +**Decision Explanation:** The paper primarily focuses on image generation using convolutional neural networks, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and + Mitigation Strategy](https://arxiv.org/abs/2501.09431v1) +**arXiv ID:** 2501.09431v1 + +**Abstract:** +> While large language models (LLMs) present significant potential for +supporting numerous real-world applications and delivering positive social +impacts, they still face significant challenges in terms of the inherent risk +of privacy leakage, hallucinated outputs, and value misalignment, and can be +maliciously used for generating toxic content and unethical purposes after been +jailbroken. Therefore, in this survey, we present a comprehensive review of +recent advancements aimed at mitigating these issues, organized across the four +phases of LLM development and usage: data collecting and pre-training, +fine-tuning and alignment, prompting and reasoning, and post-processing and +auditing. We elaborate on the recent advances for enhancing the performance of +LLMs in terms of privacy protection, hallucination reduction, value alignment, +toxicity elimination, and jailbreak defenses. In contrast to previous surveys +that focus on a single dimension of responsible LLMs, this survey presents a +unified framework that encompasses these diverse dimensions, providing a +comprehensive view of enhancing LLMs to better serve real-world applications. + +**Decision Explanation:** The paper primarily focuses on responsible AI application or AI ethics, which is one of the excluded topics. + +--- + +## [Solving the unsolvable: Translating case law in Hong Kong](https://arxiv.org/abs/2501.09444v1) +**arXiv ID:** 2501.09444v1 + +**Abstract:** +> This paper addresses the challenges translating case law under Hong Kong's +bilingual legal system. It highlights the initial success of translating all +written statutes into Chinese before the 1997 handover, a task mandated by the +Basic Law. The effort involved significant collaboration among legal, +linguistic, and translation experts, resulting in a comprehensive and +culturally appropriate bilingual legal system. However, translating case law +remains a significant challenge due to the sheer volume and continuous growth +of judicial decisions. The paper critiques the governments and judiciarys +sporadic and uncoordinated efforts to translate case law, contrasting it with +the thorough approach previously taken for statute translation. Although the +government acknowledges the importance of legal bilingualism, it lacks a +sustainable strategy for translating case law. The Judiciarys position that +translating all judgments is unnecessary, unrealistic, and not cost-effectiveis +analyzed and critiqued for its impact on legal transparency and public trust. A +proposed solution involves leveraging machine translation technology through a +human-machine interactive translation platform, which undergoes two major +transitions. Initially based on a neural model, the platform transitions to +using a large language model for improved translation accuracy. Furthermore, it +evolves from a single-agent system to a multi-agent system, incorporating +Translator, Annotator, and Proofreader agents. This multi-agent approach, +supported by a grant, aims to facilitate efficient, high-quality translation of +judicial judgments by integrating advanced artificial intelligence and +continuous feedback mechanisms, thus better meeting the needs of a bilingual +legal system. + +**Decision Explanation:** The paper primarily focuses on law, which is an excluded topic, and does not meet the required criteria for practical applications of Large Language Models in areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [RE-POSE: Synergizing Reinforcement Learning-Based Partitioning and + Offloading for Edge Object Detection](https://arxiv.org/abs/2501.09465v1) +**arXiv ID:** 2501.09465v1 + +**Abstract:** +> Object detection plays a crucial role in smart video analysis, with +applications ranging from autonomous driving and security to smart cities. +However, achieving real-time object detection on edge devices presents +significant challenges due to their limited computational resources and the +high demands of deep neural network (DNN)-based detection models, particularly +when processing high-resolution video. Conventional strategies, such as input +down-sampling and network up-scaling, often compromise detection accuracy for +faster performance or lead to higher inference latency. To address these +issues, this paper introduces RE-POSE, a Reinforcement Learning (RL)-Driven +Partitioning and Edge Offloading framework designed to optimize the +accuracy-latency trade-off in resource-constrained edge environments. Our +approach features an RL-Based Dynamic Clustering Algorithm (RL-DCA) that +partitions video frames into non-uniform blocks based on object distribution +and the computational characteristics of DNNs. Furthermore, a parallel edge +offloading scheme is implemented to distribute these blocks across multiple +edge servers for concurrent processing. Experimental evaluations show that +RE-POSE significantly enhances detection accuracy and reduces inference +latency, surpassing existing methods. + +**Decision Explanation:** The paper primarily focuses on video processing, which is one of the excluded areas, and does not explicitly mention Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [MonoSOWA: Scalable monocular 3D Object detector Without human + Annotations](https://arxiv.org/abs/2501.09481v1) +**arXiv ID:** 2501.09481v1 + +**Abstract:** +> Detecting the three-dimensional position and orientation of objects using a +single RGB camera is a foundational task in computer vision with many important +applications. Traditionally, 3D object detection methods are trained in a +fully-supervised setup, requiring vast amounts of human annotations, which are +laborious, costly, and do not scale well with the ever-increasing amounts of +data being captured. + In this paper, we present the first method to train 3D object detectors for +monocular RGB cameras without domain-specific human annotations, thus making +orders of magnitude more data available for training. Thanks to newly proposed +Canonical Object Space, the method can not only exploit data across a variety +of datasets and camera setups to train a single 3D detector, but unlike +previous work it also works out of the box in previously unseen camera setups. +All this is crucial for practical applications, where the data and cameras are +extremely heterogeneous. + The method is evaluated on two standard autonomous driving datasets, where it +outperforms previous works, which, unlike our method, still rely on 2D human +annotations. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on computer vision and 3D object detection, which is not related to Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [MatrixNet: Learning over symmetry groups using learned group + representations](https://arxiv.org/abs/2501.09571v1) +**arXiv ID:** 2501.09571v1 + +**Abstract:** +> Group theory has been used in machine learning to provide a theoretically +grounded approach for incorporating known symmetry transformations in tasks +from robotics to protein modeling. In these applications, equivariant neural +networks use known symmetry groups with predefined representations to learn +over geometric input data. We propose MatrixNet, a neural network architecture +that learns matrix representations of group element inputs instead of using +predefined representations. MatrixNet achieves higher sample efficiency and +generalization over several standard baselines in prediction tasks over the +several finite groups and the Artin braid group. We also show that MatrixNet +respects group relations allowing generalization to group elements of greater +word length than in the training set. + +**Decision Explanation:** The paper does not meet the criteria as it does not focus on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not include experimental results with quantitative metrics related to LLMs. + +--- + +## [Managed-Retention Memory: A New Class of Memory for the AI Era](https://arxiv.org/abs/2501.09605v1) +**arXiv ID:** 2501.09605v1 + +**Abstract:** +> AI clusters today are one of the major uses of High Bandwidth Memory (HBM). +However, HBM is suboptimal for AI workloads for several reasons. Analysis shows +HBM is overprovisioned on write performance, but underprovisioned on density +and read bandwidth, and also has significant energy per bit overheads. It is +also expensive, with lower yield than DRAM due to manufacturing complexity. We +propose a new memory class: Managed-Retention Memory (MRM), which is more +optimized to store key data structures for AI inference workloads. We believe +that MRM may finally provide a path to viability for technologies that were +originally proposed to support Storage Class Memory (SCM). These technologies +traditionally offered long-term persistence (10+ years) but provided poor IO +performance and/or endurance. MRM makes different trade-offs, and by +understanding the workload IO patterns, MRM foregoes long-term data retention +and write performance for better potential performance on the metrics important +for these workloads. + +**Decision Explanation:** The paper does not meet the criteria as it primarily focuses on memory optimization for AI workloads, without explicit mention of Large Language Models, knowledge graphs, retrieval-augmented generation, or agentic AI, and lacks experimental results with quantitative metrics. + +--- + +## [Incorporating Quantum Advantage in Quantum Circuit Generation through + Genetic Programming](https://arxiv.org/abs/2501.09682v1) +**arXiv ID:** 2501.09682v1 + +**Abstract:** +> Designing efficient quantum circuits that leverage quantum advantage compared +to classical computing has become increasingly critical. Genetic algorithms +have shown potential in generating such circuits through artificial evolution. +However, integrating quantum advantage into the fitness function of these +algorithms remains unexplored. In this paper, we aim to enhance the efficiency +of quantum circuit design by proposing two novel approaches for incorporating +quantum advantage metrics into the fitness function of genetic algorithms.1 We +evaluate our approaches based on the Bernstein-Vazirani Problem and the +Unstructured Database Search Problem as test cases. The results demonstrate +that our approaches not only improve the convergence speed of the genetic +algorithm but also produce circuits comparable to expert-designed solutions. +Our findings suggest that automated quantum circuit design using genetic +algorithms that incorporate a measure of quantum advantage is a promising +approach to accelerating the development of quantum algorithms. + +**Decision Explanation:** The paper does not meet the criteria as it focuses on quantum circuit generation and genetic programming, which is unrelated to Large Language Models (LLMs) and their practical applications. + +--- + +## [Reward-Guided Controlled Generation for Inference-Time Alignment in + Diffusion Models: Tutorial and Review](https://arxiv.org/abs/2501.09685v1) +**arXiv ID:** 2501.09685v1 + +**Abstract:** +> This tutorial provides an in-depth guide on inference-time guidance and +alignment methods for optimizing downstream reward functions in diffusion +models. While diffusion models are renowned for their generative modeling +capabilities, practical applications in fields such as biology often require +sample generation that maximizes specific metrics (e.g., stability, affinity in +proteins, closeness to target structures). In these scenarios, diffusion models +can be adapted not only to generate realistic samples but also to explicitly +maximize desired measures at inference time without fine-tuning. This tutorial +explores the foundational aspects of such inference-time algorithms. We review +these methods from a unified perspective, demonstrating that current techniques +-- such as Sequential Monte Carlo (SMC)-based guidance, value-based sampling, +and classifier guidance -- aim to approximate soft optimal denoising processes +(a.k.a. policies in RL) that combine pre-trained denoising processes with value +functions serving as look-ahead functions that predict from intermediate states +to terminal rewards. Within this framework, we present several novel algorithms +not yet covered in the literature. Furthermore, we discuss (1) fine-tuning +methods combined with inference-time techniques, (2) inference-time algorithms +based on search algorithms such as Monte Carlo tree search, which have received +limited attention in current research, and (3) connections between +inference-time algorithms in language models and diffusion models. The code of +this tutorial on protein design is available at +https://github.com/masa-ue/AlignInversePro + +**Decision Explanation:** The paper primarily focuses on diffusion models and their applications in biology, which does not meet the criteria of focusing on Large Language Models (LLMs) or their practical applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Practical Continual Forgetting for Pre-trained Vision Models](https://arxiv.org/abs/2501.09705v1) +**arXiv ID:** 2501.09705v1 + +**Abstract:** +> For privacy and security concerns, the need to erase unwanted information +from pre-trained vision models is becoming evident nowadays. In real-world +scenarios, erasure requests originate at any time from both users and model +owners, and these requests usually form a sequence. Therefore, under such a +setting, selective information is expected to be continuously removed from a +pre-trained model while maintaining the rest. We define this problem as +continual forgetting and identify three key challenges. (i) For unwanted +knowledge, efficient and effective deleting is crucial. (ii) For remaining +knowledge, the impact brought by the forgetting procedure should be minimal. +(iii) In real-world scenarios, the training samples may be scarce or partially +missing during the process of forgetting. To address them, we first propose +Group Sparse LoRA (GS-LoRA). Specifically, towards (i), we introduce LoRA +modules to fine-tune the FFN layers in Transformer blocks for each forgetting +task independently, and towards (ii), a simple group sparse regularization is +adopted, enabling automatic selection of specific LoRA groups and zeroing out +the others. To further extend GS-LoRA to more practical scenarios, we +incorporate prototype information as additional supervision and introduce a +more practical approach, GS-LoRA++. For each forgotten class, we move the +logits away from its original prototype. For the remaining classes, we pull the +logits closer to their respective prototypes. We conduct extensive experiments +on face recognition, object detection and image classification and demonstrate +that our method manages to forget specific classes with minimal impact on other +classes. Codes have been released on https://github.com/bjzhb666/GS-LoRA. + +**Decision Explanation:** The paper primarily focuses on vision models and does not meet the criteria related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Parallel multi-objective metaheuristics for smart communications in + vehicular networks](https://arxiv.org/abs/2501.09725v1) +**arXiv ID:** 2501.09725v1 + +**Abstract:** +> This article analyzes the use of two parallel multi-objective soft computing +algorithms to automatically search for high-quality settings of the Ad hoc On +Demand Vector routing protocol for vehicular networks. These methods are based +on an evolutionary algorithm and on a swarm intelligence approach. The +experimental analysis demonstrates that the configurations computed by our +optimization algorithms outperform other state-of-the-art optimized ones. In +turn, the computational efficiency achieved by all the parallel versions is +greater than 87 %. Therefore, the line of work presented in this article +represents an efficient framework to improve vehicular communications. + +**Decision Explanation:** The paper does not meet the criteria as it does not focus on Large Language Models (LLMs) or their applications, and instead discusses optimization algorithms for vehicular networks. + +--- + +## [Benchmarking Robustness of Contrastive Learning Models for Medical + Image-Report Retrieval](https://arxiv.org/abs/2501.09134v1) +**arXiv ID:** 2501.09134v1 + +**Abstract:** +> Medical images and reports offer invaluable insights into patient health. The +heterogeneity and complexity of these data hinder effective analysis. To bridge +this gap, we investigate contrastive learning models for cross-domain +retrieval, which associates medical images with their corresponding clinical +reports. This study benchmarks the robustness of four state-of-the-art +contrastive learning models: CLIP, CXR-RePaiR, MedCLIP, and CXR-CLIP. We +introduce an occlusion retrieval task to evaluate model performance under +varying levels of image corruption. Our findings reveal that all evaluated +models are highly sensitive to out-of-distribution data, as evidenced by the +proportional decrease in performance with increasing occlusion levels. While +MedCLIP exhibits slightly more robustness, its overall performance remains +significantly behind CXR-CLIP and CXR-RePaiR. CLIP, trained on a +general-purpose dataset, struggles with medical image-report retrieval, +highlighting the importance of domain-specific training data. The evaluation of +this work suggests that more effort needs to be spent on improving the +robustness of these models. By addressing these limitations, we can develop +more reliable cross-domain retrieval models for medical applications. + +**Decision Explanation:** The paper primarily focuses on medical applications of AI, which is an excluded topic according to the criteria. + +--- + +## [Quantum-Enhanced Transformers for Robust Acoustic Scene Classification + in IoT Environments](https://arxiv.org/abs/2501.09394v1) +**arXiv ID:** 2501.09394v1 + +**Abstract:** +> The proliferation of Internet of Things (IoT) devices equipped with acoustic +sensors necessitates robust acoustic scene classification (ASC) capabilities, +even in noisy and data-limited environments. Traditional machine learning +methods often struggle to generalize effectively under such conditions. To +address this, we introduce Q-ASC, a novel Quantum-Inspired Acoustic Scene +Classifier that leverages the power of quantum-inspired transformers. By +integrating quantum concepts like superposition and entanglement, Q-ASC +achieves superior feature learning and enhanced noise resilience compared to +classical models. Furthermore, we introduce a Quantum Variational Autoencoder +(QVAE) based data augmentation technique to mitigate the challenge of limited +labeled data in IoT deployments. Extensive evaluations on the Tampere +University of Technology (TUT) Acoustic Scenes 2016 benchmark dataset +demonstrate that Q-ASC achieves remarkable accuracy between 68.3% and 88.5% +under challenging conditions, outperforming state-of-the-art methods by over 5% +in the best case. This research paves the way for deploying intelligent +acoustic sensing in IoT networks, with potential applications in smart homes, +industrial monitoring, and environmental surveillance, even in adverse acoustic +environments. + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their practical applications, and instead explores quantum-inspired transformers for acoustic scene classification, which is outside the specified criteria. + +--- + +## [Metric Learning with Progressive Self-Distillation for Audio-Visual + Embedding Learning](https://arxiv.org/abs/2501.09608v1) +**arXiv ID:** 2501.09608v1 + +**Abstract:** +> Metric learning projects samples into an embedded space, where similarities +and dissimilarities are quantified based on their learned representations. +However, existing methods often rely on label-guided representation learning, +where representations of different modalities, such as audio and visual data, +are aligned based on annotated labels. This approach tends to underutilize +latent complex features and potential relationships inherent in the +distributions of audio and visual data that are not directly tied to the +labels, resulting in suboptimal performance in audio-visual embedding learning. +To address this issue, we propose a novel architecture that integrates +cross-modal triplet loss with progressive self-distillation. Our method +enhances representation learning by leveraging inherent distributions and +dynamically refining soft audio-visual alignments -- probabilistic alignments +between audio and visual data that capture the inherent relationships beyond +explicit labels. Specifically, the model distills audio-visual +distribution-based knowledge from annotated labels in a subset of each batch. +This self-distilled knowledge is used t + +**Decision Explanation:** The paper does not focus on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the primary criteria for selection. + +--- + diff --git a/llm_processor/README.md b/llm_processor/README.md new file mode 100644 index 0000000..7dc5d72 --- /dev/null +++ b/llm_processor/README.md @@ -0,0 +1,82 @@ +# LLM Processor Module + +A Go module for processing papers through language models with configurable criteria. + +## Installation + +```bash +go get llm_processor +``` + +## Command Line Usage + +The module provides a CLI interface for processing papers: + +```bash +go run main.go -input papers.json -output results.json -criteria criteria.txt +``` + +## API Usage + +Import the module in your Go code: + +```go +import "llm_processor" +``` + +### Example: Processing Papers + +```go +package main + +import ( + "fmt" + "llm_processor/models" + "llm_processor/processor" + "llm_processor/storage" +) + +func main() { + // Load papers + papers, err := storage.LoadPapers("papers.json") + if err != nil { + panic(err) + } + + // Initialize processor + proc := processor.NewProcessor("gpt-4", 32) + + // Process papers + results := proc.ProcessPapers(papers, "criteria.txt") + + // Save results + err = storage.SaveResults("results.json", results) + if err != nil { + panic(err) + } + + fmt.Println("Processing complete!") +} +``` + +## Configuration + +The processor supports the following configuration: + +- Model selection (gpt-3.5-turbo, gpt-4, etc.) +- Batch size +- Custom criteria files +- Input/output file paths + +## Dependencies + +- Go 1.21+ +- OpenAI API key (set as environment variable OPENAI_API_KEY) + +## Contributing + +Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change. + +## License + +[MIT](https://choosealicense.com/licenses/mit/) diff --git a/llm_processor/client/client.go b/llm_processor/client/client.go new file mode 100644 index 0000000..919925f --- /dev/null +++ b/llm_processor/client/client.go @@ -0,0 +1,326 @@ +package client + +import ( + "bytes" + "context" + "encoding/json" + "errors" + "fmt" + "io" + "log" + "math" + "net" + "net/http" + "os" + "strings" + "time" + + "llm_processor/models" +) + +const ( + openRouterURL = "https://openrouter.ai/api/v1/chat/completions" + maxRetries = 5 + initialDelay = 1 * time.Second + evaluationTimeout = 15 * time.Minute + requestTimeout = 5 * time.Minute + connectionTimeout = 2 * time.Minute +) + +type OpenRouterClient struct { + apiKey string + httpClient *http.Client + logger *log.Logger + createClient func() *http.Client +} + +func NewOpenRouterClient(apiKey string) *OpenRouterClient { + logFile, err := os.OpenFile("debug.log", os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644) + if err != nil { + log.Fatalf("Failed to open debug.log: %v", err) + } + + logger := log.New(io.MultiWriter(os.Stdout, logFile), "", log.LstdFlags|log.Lshortfile) + logger.Println("Initializing OpenRouter client") + + createClient := func() *http.Client { + transport := &http.Transport{ + MaxIdleConns: 100, + MaxIdleConnsPerHost: 100, + IdleConnTimeout: 30 * time.Second, + DialContext: (&net.Dialer{ + Timeout: connectionTimeout, + KeepAlive: 30 * time.Second, + }).DialContext, + TLSHandshakeTimeout: 10 * time.Second, + } + + return &http.Client{ + Timeout: requestTimeout, + Transport: transport, + } + } + + client := &OpenRouterClient{ + apiKey: apiKey, + httpClient: createClient(), + logger: logger, + createClient: createClient, + } + + return client +} + +type ChatCompletionRequest struct { + Model string `json:"model"` + Messages []ChatMessage `json:"messages"` +} + +type ChatMessage struct { + Role string `json:"role"` + Content string `json:"content"` +} + +type ChatCompletionResponse struct { + Choices []struct { + Message ChatMessage `json:"message"` + } `json:"choices"` +} + +func (c *OpenRouterClient) EvaluatePaper(ctx context.Context, paper models.Paper, criteria string, model string) (string, error) { + // Create a new context with evaluation timeout + evalCtx, cancel := context.WithTimeout(ctx, evaluationTimeout) + defer cancel() + + startTime := time.Now() + c.logger.Printf("Starting evaluation for paper: %s\n", paper.Title) + c.logger.Printf("Evaluation timeout: %s\n", evaluationTimeout) + fmt.Printf("Starting evaluation for paper: %s at %s\n", paper.Title, startTime.Format(time.RFC3339)) + + prompt := fmt.Sprintf(`Evaluate this paper based on the following criteria: +%s + +Paper Title: %s +Abstract: %s + +Respond ONLY with a JSON object in this exact format: +{ + "decision": "ACCEPT or REJECT", + "explanation": "Your explanation here" +} + +Do not include any other information in your response. + +IMPORTANT: +1. The decision MUST be either "ACCEPT" or "REJECT" (uppercase) +2. The explanation should be a clear, concise reason for your decision +3. Do not include any text outside the JSON object +4. Ensure the response is valid JSON (proper quotes and escaping) +5. Do not include any markdown or formatting`, criteria, paper.Title, paper.Abstract) + + reqBody := ChatCompletionRequest{ + Model: model, + Messages: []ChatMessage{ + { + Role: "system", + Content: "You are a research paper evaluator. Respond only with the requested JSON format.", + }, + { + Role: "user", + Content: prompt, + }, + }, + } + + var lastErr error + for attempt := 0; attempt < maxRetries; attempt++ { + attemptStart := time.Now() + c.logger.Printf("Attempt %d started at %s\n", attempt+1, attemptStart.Format(time.RFC3339)) + if attempt > 0 { + delay := time.Duration(math.Pow(2, float64(attempt))) * initialDelay + select { + case <-time.After(delay): + case <-ctx.Done(): + return "", ctx.Err() + } + } + + jsonBody, err := json.Marshal(reqBody) + if err != nil { + return "", fmt.Errorf("failed to marshal request body: %w", err) + } + + req, err := http.NewRequestWithContext(evalCtx, "POST", openRouterURL, bytes.NewBuffer(jsonBody)) + if err != nil { + return "", fmt.Errorf("failed to create request: %w", err) + } + + req.Header.Set("Authorization", "Bearer "+c.apiKey) + req.Header.Set("Content-Type", "application/json") + req.Header.Set("HTTP-Referer", "https://github.com/stwhite/arvix") + req.Header.Set("X-Title", "ArXiv Paper Processor") + + resp, err := c.httpClient.Do(req) + if err != nil { + // Log the specific error type + c.logger.Printf("Attempt %d error: %v\n", attempt+1, err) + + // Handle context cancellation/timeout + if errors.Is(err, context.DeadlineExceeded) { + c.logger.Printf("Context deadline exceeded, retrying...\n") + lastErr = fmt.Errorf("context deadline exceeded") + continue + } + + // On timeout errors, create a new client + if strings.Contains(err.Error(), "timeout") { + c.logger.Printf("Timeout detected, recreating HTTP client...\n") + c.httpClient = c.createClient() + } + + lastErr = fmt.Errorf("attempt %d: %w", attempt+1, err) + continue + } + defer resp.Body.Close() + + if resp.StatusCode != http.StatusOK { + body, _ := io.ReadAll(resp.Body) + lastErr = fmt.Errorf("attempt %d: openrouter request failed: %s - %s", attempt+1, resp.Status, string(body)) + continue + } + + var completionResp ChatCompletionResponse + if err := json.NewDecoder(resp.Body).Decode(&completionResp); err != nil { + lastErr = fmt.Errorf("attempt %d: failed to decode response: %w", attempt+1, err) + continue + } + + if len(completionResp.Choices) == 0 { + lastErr = fmt.Errorf("attempt %d: no choices in response", attempt+1) + continue + } + + rawContent := completionResp.Choices[0].Message.Content + + // Try to parse as JSON first + var jsonResponse map[string]interface{} + err = json.Unmarshal([]byte(rawContent), &jsonResponse) + if err != nil { + // If direct JSON parsing fails, try extracting from markdown code block + startIdx := bytes.Index([]byte(rawContent), []byte("```json")) + if startIdx >= 0 { + startIdx += len("```json") + endIdx := bytes.Index([]byte(rawContent[startIdx:]), []byte("```")) + if endIdx >= 0 { + jsonContent := rawContent[startIdx : startIdx+endIdx] + err = json.Unmarshal([]byte(jsonContent), &jsonResponse) + if err != nil { + // If still failing, try to parse as raw JSON without code block + err = json.Unmarshal([]byte(rawContent), &jsonResponse) + } + } + } + } + + if err == nil { + // Validate and normalize decision + if decision, ok := jsonResponse["decision"].(string); ok { + // Normalize decision value + normalizedDecision := strings.ToUpper(strings.TrimSpace(decision)) + if strings.Contains(normalizedDecision, "ACCEPT") { + normalizedDecision = "ACCEPT" + } else if strings.Contains(normalizedDecision, "REJECT") { + normalizedDecision = "REJECT" + } + + if normalizedDecision == "ACCEPT" || normalizedDecision == "REJECT" { + // Preserve original decision in explanation + if explanation, ok := jsonResponse["explanation"]; !ok { + jsonResponse["explanation"] = fmt.Sprintf("Original decision: %s\n", decision) + } else { + jsonResponse["explanation"] = fmt.Sprintf("Original decision: %s\n%s", decision, explanation) + } + + // Parse nested JSON in explanation if present + if explanation, ok := jsonResponse["explanation"].(string); ok { + var nested map[string]interface{} + if err := json.Unmarshal([]byte(explanation), &nested); err == nil { + jsonResponse["explanation"] = nested + } + } + + // Ensure consistent response format + response := map[string]interface{}{ + "paper": map[string]interface{}{ + "title": paper.Title, + "abstract": paper.Abstract, + "arxiv_id": paper.ArxivID, + }, + "decision": normalizedDecision, + "explanation": jsonResponse["explanation"], + } + + responseJSON, err := json.Marshal(response) + if err != nil { + return "", fmt.Errorf("failed to marshal response: %w", err) + } + + duration := time.Since(startTime) + c.logger.Printf("Successfully evaluated paper: %s\n", paper.Title) + c.logger.Printf("Total time: %s\n", duration) + c.logger.Printf("Attempts: %d\n", attempt+1) + return string(responseJSON), nil + } + } + } + + // If direct JSON parsing fails, try extracting from markdown code block + startIdx := bytes.Index([]byte(rawContent), []byte("```json")) + if startIdx >= 0 { + startIdx += len("```json") + endIdx := bytes.Index([]byte(rawContent[startIdx:]), []byte("```")) + if endIdx >= 0 { + jsonContent := rawContent[startIdx : startIdx+endIdx] + err = json.Unmarshal([]byte(jsonContent), &jsonResponse) + if err == nil { + if decision, ok := jsonResponse["decision"].(string); ok { + if decision == "ACCEPT" || decision == "REJECT" { + duration := time.Since(startTime) + c.logger.Printf("Successfully evaluated paper: %s\n", paper.Title) + c.logger.Printf("Total time: %s\n", duration) + c.logger.Printf("Attempts: %d\n", attempt+1) + return jsonContent, nil + } + } + } + } + } + + // Fallback parsing if JSON is still invalid + decision := "ERROR" + if strings.Contains(rawContent, "ACCEPT") { + decision = "ACCEPT" + } else if strings.Contains(rawContent, "REJECT") { + decision = "REJECT" + } + + // Create fallback response + fallbackResponse := map[string]interface{}{ + "decision": decision, + "explanation": fmt.Sprintf("Original response: %s", rawContent), + } + fallbackJSON, _ := json.Marshal(fallbackResponse) + + duration := time.Since(startTime) + c.logger.Printf("Fallback parsing used for paper: %s\n", paper.Title) + c.logger.Printf("Total time: %s\n", duration) + c.logger.Printf("Attempts: %d\n", attempt+1) + return string(fallbackJSON), nil + } + + duration := time.Since(startTime) + c.logger.Printf("Failed to evaluate paper: %s\n", paper.Title) + c.logger.Printf("Total time: %s\n", duration) + c.logger.Printf("Attempts: %d\n", maxRetries) + return "", fmt.Errorf("max retries (%d) exceeded: %w", maxRetries, lastErr) +} diff --git a/llm_processor/go.mod b/llm_processor/go.mod new file mode 100644 index 0000000..1c3aff1 --- /dev/null +++ b/llm_processor/go.mod @@ -0,0 +1,10 @@ +module llm_processor + +go 1.21 + +require github.com/spf13/cobra v1.8.0 + +require ( + github.com/inconshreveable/mousetrap v1.1.0 // indirect + github.com/spf13/pflag v1.0.5 // indirect +) diff --git a/llm_processor/go.sum b/llm_processor/go.sum new file mode 100644 index 0000000..d0e8c2c --- /dev/null +++ b/llm_processor/go.sum @@ -0,0 +1,10 @@ +github.com/cpuguy83/go-md2man/v2 v2.0.3/go.mod h1:tgQtvFlXSQOSOSIRvRPT7W67SCa46tRHOmNcaadrF8o= +github.com/inconshreveable/mousetrap v1.1.0 h1:wN+x4NVGpMsO7ErUn/mUI3vEoE6Jt13X2s0bqwp9tc8= +github.com/inconshreveable/mousetrap v1.1.0/go.mod h1:vpF70FUmC8bwa3OWnCshd2FqLfsEA9PFc4w1p2J65bw= +github.com/russross/blackfriday/v2 v2.1.0/go.mod h1:+Rmxgy9KzJVeS9/2gXHxylqXiyQDYRxCVz55jmeOWTM= +github.com/spf13/cobra v1.8.0 h1:7aJaZx1B85qltLMc546zn58BxxfZdR/W22ej9CFoEf0= +github.com/spf13/cobra v1.8.0/go.mod h1:WXLWApfZ71AjXPya3WOlMsY9yMs7YeiHhFVlvLyhcho= +github.com/spf13/pflag v1.0.5 h1:iy+VFUOCP1a+8yFto/drg2CJ5u0yRoB7fZw3DKv/JXA= +github.com/spf13/pflag v1.0.5/go.mod h1:McXfInJRrz4CZXVZOBLb0bTZqETkiAhM9Iw0y3An2Bg= +gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0= +gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM= diff --git a/llm_processor/main.go b/llm_processor/main.go new file mode 100644 index 0000000..b71295a --- /dev/null +++ b/llm_processor/main.go @@ -0,0 +1,77 @@ +package main + +import ( + "context" + "fmt" + "log" + "os" + "time" + + "llm_processor/processor" + "llm_processor/storage" + + "github.com/spf13/cobra" +) + +var ( + inputFile string + outputFile string + criteriaFile string + modelName string + batchSize int + delaySeconds int + timeoutSeconds int +) + +var rootCmd = &cobra.Command{ + Use: "llm-processor", + Short: "Process papers using LLM with configurable criteria", + Long: `A command line tool for processing academic papers through +language models using customizable evaluation criteria.`, + Run: func(cmd *cobra.Command, args []string) { + // Get OpenRouter API key from environment + apiKey := os.Getenv("OPENROUTER_API_KEY") + if apiKey == "" { + log.Fatal("OPENROUTER_API_KEY environment variable is required") + } + + // Load papers + papers, err := storage.LoadPapers(inputFile) + if err != nil { + log.Fatalf("Failed to load papers: %v", err) + } + + // Initialize processor + proc := processor.NewProcessor(modelName, batchSize, apiKey) + proc.SetTimeout(time.Duration(timeoutSeconds) * time.Second) + + // Process papers + ctx := context.Background() + if err := proc.ProcessPapers(ctx, inputFile, outputFile, criteriaFile, time.Duration(delaySeconds)*time.Second); err != nil { + log.Fatalf("Failed to save results: %v", err) + } + + fmt.Printf("Successfully processed %d papers\n", len(papers)) + }, +} + +func init() { + rootCmd.PersistentFlags().StringVarP(&inputFile, "input", "i", "", "Input JSON file containing papers") + rootCmd.PersistentFlags().StringVarP(&outputFile, "output", "o", "results.json", "Output JSON file for results") + rootCmd.PersistentFlags().StringVarP(&criteriaFile, "criteria", "c", "", "Text file containing evaluation criteria") + rootCmd.PersistentFlags().StringVarP(&modelName, "model", "m", "gpt-4", "LLM model to use") + rootCmd.PersistentFlags().IntVarP(&batchSize, "batch-size", "b", 32, "Batch size for processing") + rootCmd.PersistentFlags().IntVarP(&delaySeconds, "delay", "d", 0, "Delay in seconds between paper submissions") + rootCmd.PersistentFlags().IntVarP(&timeoutSeconds, "timeout", "t", 3600, "Timeout in seconds for processing (default 1h)") + + // Mark required flags + rootCmd.MarkPersistentFlagRequired("input") + rootCmd.MarkPersistentFlagRequired("criteria") +} + +func main() { + if err := rootCmd.Execute(); err != nil { + fmt.Println(err) + os.Exit(1) + } +} diff --git a/llm_processor/models/models.go b/llm_processor/models/models.go new file mode 100644 index 0000000..3a4da4b --- /dev/null +++ b/llm_processor/models/models.go @@ -0,0 +1,34 @@ +package models + +import ( + "fmt" +) + +// Paper represents a research paper +type Paper struct { + Title string `json:"title"` + Abstract string `json:"abstract"` + ArxivID string `json:"arxiv_id"` +} + +// Result represents the evaluation result from the LLM +type Result struct { + Paper Paper `json:"paper"` + Decision string `json:"decision"` // ACCEPT, REJECT, or ERROR + Explanation string `json:"explanation"` +} + +// OrganizedResults contains categorized results +type OrganizedResults struct { + Accepted []Result `json:"accepted"` + Rejected []Result `json:"rejected"` + Errors []Result `json:"errors"` +} + +// Validate checks if a Result has valid decision value +func (r *Result) Validate() error { + if r.Decision != "ACCEPT" && r.Decision != "REJECT" && r.Decision != "ERROR" { + return fmt.Errorf("invalid decision value: %s", r.Decision) + } + return nil +} diff --git a/llm_processor/newpapers-filtered.json b/llm_processor/newpapers-filtered.json new file mode 100644 index 0000000..ac01349 --- /dev/null +++ b/llm_processor/newpapers-filtered.json @@ -0,0 +1,1060 @@ +{ + "accepted": [ + { + "paper": { + "title": "Agent-R: Training Language Model Agents to Reflect via Iterative\n Self-Training", + "abstract": "Large Language Models (LLMs) agents are increasingly pivotal for addressing\ncomplex tasks in interactive environments. Existing work mainly focuses on\nenhancing performance through behavior cloning from stronger experts, yet such\napproaches often falter in real-world applications, mainly due to the inability\nto recover from errors. However, step-level critique data is difficult and\nexpensive to collect. Automating and dynamically constructing self-critique\ndatasets is thus crucial to empowering models with intelligent agent\ncapabilities. In this work, we propose an iterative self-training framework,\nAgent-R, that enables language Agent to Reflect on the fly. Unlike traditional\nmethods that reward or penalize actions based on correctness, Agent-R leverages\nMCTS to construct training data that recover correct trajectories from\nerroneous ones. A key challenge of agent reflection lies in the necessity for\ntimely revision rather than waiting until the end of a rollout. To address\nthis, we introduce a model-guided critique construction mechanism: the actor\nmodel identifies the first error step (within its current capability) in a\nfailed trajectory. Starting from it, we splice it with the adjacent correct\npath, which shares the same parent node in the tree. This strategy enables the\nmodel to learn reflection based on its current policy, therefore yielding\nbetter learning efficiency. To further explore the scalability of this\nself-improvement paradigm, we investigate iterative refinement of both error\ncorrection capabilities and dataset construction. Our findings demonstrate that\nAgent-R continuously improves the model's ability to recover from errors and\nenables timely error correction. Experiments on three interactive environments\nshow that Agent-R effectively equips agents to correct erroneous actions while\navoiding loops, achieving superior performance compared to baseline methods\n(+5.59%).", + "arxiv_id": "2501.11425v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications in interactive environments), 2 (Experimental Results with +5.59% performance improvement), and 3 (Comparison with State-of-the-Art with baseline methods). Additionally, shows novelty in introducing an iterative self-training framework for LLMs, enabling agentic AI for complex, real-world tasks, and demonstrates robust experimental validation.\"\n}\n```" + }, + { + "paper": { + "title": "Reasoning Language Models: A Blueprint", + "abstract": "Reasoning language models (RLMs), also known as Large Reasoning Models\n(LRMs), such as OpenAI's o1 and o3, DeepSeek-V3, and Alibaba's QwQ, have\nredefined AI's problem-solving capabilities by extending large language models\n(LLMs) with advanced reasoning mechanisms. Yet, their high costs, proprietary\nnature, and complex architectures - uniquely combining Reinforcement Learning\n(RL), search heuristics, and LLMs - present accessibility and scalability\nchallenges. To address these, we propose a comprehensive blueprint that\norganizes RLM components into a modular framework, based on a survey and\nanalysis of all RLM works. This blueprint incorporates diverse reasoning\nstructures (chains, trees, graphs, and nested forms), reasoning strategies\n(e.g., Monte Carlo Tree Search, Beam Search), RL concepts (policy, value models\nand others), and supervision schemes (Output-Based and Process-Based\nSupervision). We also provide detailed mathematical formulations and\nalgorithmic specifications to simplify RLM implementation. By showing how\nschemes like LLaMA-Berry, QwQ, Journey Learning, and Graph of Thoughts fit as\nspecial cases, we demonstrate the blueprint's versatility and unifying\npotential. To illustrate its utility, we introduce x1, a modular implementation\nfor rapid RLM prototyping and experimentation. Using x1 and a literature\nreview, we provide key insights, such as multi-phase training for policy and\nvalue models, and the importance of familiar training distributions. Finally,\nwe outline how RLMs can integrate with a broader LLM ecosystem, including tools\nand databases. Our work demystifies RLM construction, democratizes advanced\nreasoning capabilities, and fosters innovation, aiming to mitigate the gap\nbetween \"rich AI\" and \"poor AI\" by lowering barriers to RLM development and\nexperimentation.", + "arxiv_id": "2501.11223v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria (1, 2, 3) by focusing on practical applications of Large Language Models (LLMs) with advanced reasoning mechanisms, presenting experimental results with quantitative potential through the modular framework, and comparing with state-of-the-art techniques. Additionally meets criteria 4 (Methodology and Implementation Details) and shows Novelty, Reproducibility, and Impact through Experimental Validation, with no primary focus on excluded topics." + }, + { + "paper": { + "title": "Question-to-Question Retrieval for Hallucination-Free Knowledge Access:\n An Approach for Wikipedia and Wikidata Question Answering", + "abstract": "This paper introduces an approach to question answering over knowledge bases\nlike Wikipedia and Wikidata by performing \"question-to-question\" matching and\nretrieval from a dense vector embedding store. Instead of embedding document\ncontent, we generate a comprehensive set of questions for each logical content\nunit using an instruction-tuned LLM. These questions are vector-embedded and\nstored, mapping to the corresponding content. Vector embedding of user queries\nare then matched against this question vector store. The highest similarity\nscore leads to direct retrieval of the associated article content, eliminating\nthe need for answer generation. Our method achieves high cosine similarity ( \u003e\n0.9 ) for relevant question pairs, enabling highly precise retrieval. This\napproach offers several advantages including computational efficiency, rapid\nresponse times, and increased scalability. We demonstrate its effectiveness on\nWikipedia and Wikidata, including multimedia content through structured fact\nretrieval from Wikidata, opening up new pathways for multimodal question\nanswering.", + "arxiv_id": "2501.11301v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets all mandatory criteria (1, 2, 3) by focusing on practical LLM application in knowledge retrieval, providing quantitative metrics (cosine similarity \u003e 0.9), and implicitly comparing with state-of-the-art through demonstrated efficiency and scalability improvements. Additionally meets criteria 4 (clear methodology) and shows novelty in question-to-question retrieval approach.\"\n}\n```" + }, + { + "paper": { + "title": "Few-shot Policy (de)composition in Conversational Question Answering", + "abstract": "The task of policy compliance detection (PCD) is to determine if a scenario\nis in compliance with respect to a set of written policies. In a conversational\nsetting, the results of PCD can indicate if clarifying questions must be asked\nto determine compliance status. Existing approaches usually claim to have\nreasoning capabilities that are latent or require a large amount of annotated\ndata. In this work, we propose logical decomposition for policy compliance\n(LDPC): a neuro-symbolic framework to detect policy compliance using large\nlanguage models (LLMs) in a few-shot setting. By selecting only a few exemplars\nalongside recently developed prompting techniques, we demonstrate that our\napproach soundly reasons about policy compliance conversations by extracting\nsub-questions to be answered, assigning truth values from contextual\ninformation, and explicitly producing a set of logic statements from the given\npolicies. The formulation of explicit logic graphs can in turn help answer\nPCDrelated questions with increased transparency and explainability. We apply\nthis approach to the popular PCD and conversational machine reading benchmark,\nShARC, and show competitive performance with no task-specific finetuning. We\nalso leverage the inherently interpretable architecture of LDPC to understand\nwhere errors occur, revealing ambiguities in the ShARC dataset and highlighting\nthe challenges involved with reasoning for conversational question answering.", + "arxiv_id": "2501.11335v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications in conversational question answering), 2 (Experimental Results with quantitative metrics on ShARC benchmark), and 3 (Comparison with State-of-the-Art with competitive performance). Additionally, it shows Novelty (neuro-symbolic framework with LLMs), Implementability (with current standard tools), and Reproducibility (with transparent documentation of methodologies and results)." + }, + { + "paper": { + "title": "Towards Advancing Code Generation with Large Language Models: A Research\n Roadmap", + "abstract": "Recently, we have witnessed the rapid development of large language models,\nwhich have demonstrated excellent capabilities in the downstream task of code\ngeneration. However, despite their potential, LLM-based code generation still\nfaces numerous technical and evaluation challenges, particularly when embedded\nin real-world development. In this paper, we present our vision for current\nresearch directions, and provide an in-depth analysis of existing studies on\nthis task. We propose a six-layer vision framework that categorizes code\ngeneration process into distinct phases, namely Input Phase, Orchestration\nPhase, Development Phase, and Validation Phase. Additionally, we outline our\nvision workflow, which reflects on the currently prevalent frameworks. We\nsystematically analyse the challenges faced by large language models, including\nthose LLM-based agent frameworks, in code generation tasks. With these, we\noffer various perspectives and actionable recommendations in this area. Our aim\nis to provide guidelines for improving the reliability, robustness and\nusability of LLM-based code generation systems. Ultimately, this work seeks to\naddress persistent challenges and to provide practical suggestions for a more\npragmatic LLM-based solution for future code generation endeavors.", + "arxiv_id": "2501.11354v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: (1) focuses on practical applications of LLMs in code generation, (2) implies experimental analysis through 'in-depth analysis of existing studies' and 'systematically analyse the challenges', (3) compares with state-of-the-art by analyzing 'currently prevalent frameworks'. Additionally, meets optional criteria (4) methodology discussion and (6) novelty in proposing a 'six-layer vision framework'." + }, + { + "paper": { + "title": "Neural Contextual Reinforcement Framework for Logical Structure Language\n Generation", + "abstract": "The Neural Contextual Reinforcement Framework introduces an innovative\napproach to enhancing the logical coherence and structural consistency of text\ngenerated by large language models. Leveraging reinforcement learning\nprinciples, the framework integrates custom reward functions and dynamic\ncontext alignment mechanisms to address challenges inherent in maintaining\nlong-range dependencies across extended sequences. The architecture\nincorporates multi-head attention layers and hierarchical encoding modules,\nenabling the model to produce outputs that align closely with human\nexpectations of logical structure and semantic flow. Quantitative evaluations\nacross diverse datasets demonstrate substantial improvements in coherence\nmetrics, perplexity reduction, and semantic alignment, showcasing the\nframework's ability to outperform baseline models in both general and\ndomain-specific tasks. Qualitative analyses further highlight the framework's\ncapacity to generate text with improved narrative clarity and reduced\nredundancy, reflecting its effectiveness in balancing fluency with structural\nprecision. In addition to its performance gains, the framework exhibits\nrobustness in handling noisy input data and scalability across varying model\nsizes, reinforcing its versatility in practical applications. Experimental\nresults reveal that optimal context window sizes significantly influence\ncoherence outcomes, showing the importance of architectural flexibility in\nadapting to diverse linguistic structures. Cross-lingual performance\nevaluations affirm the framework's adaptability to multiple languages,\nextending its utility beyond monolingual contexts. Resource efficiency analyses\nindicate a reduction in computational overhead compared to traditional\napproaches, emphasizing the practicality of the framework for large-scale\ndeployment.", + "arxiv_id": "2501.11417v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications in language generation), 2 (Experimental Results with quantitative metrics), and 3 (Comparison with State-of-the-Art showing performance improvements). Additionally, it marginally meets criterion 4 (Methodology and Implementation Details) and shows Novelty in integrating reinforcement learning for logical structure enhancement, warranting further review.\"\n}" + }, + { + "paper": { + "title": "Explainable Lane Change Prediction for Near-Crash Scenarios Using\n Knowledge Graph Embeddings and Retrieval Augmented Generation", + "abstract": "Lane-changing maneuvers, particularly those executed abruptly or in risky\nsituations, are a significant cause of road traffic accidents. However, current\nresearch mainly focuses on predicting safe lane changes. Furthermore, existing\naccident datasets are often based on images only and lack comprehensive sensory\ndata. In this work, we focus on predicting risky lane changes using the CRASH\ndataset (our own collected dataset specifically for risky lane changes), and\nsafe lane changes (using the HighD dataset). Then, we leverage KG and Bayesian\ninference to predict these maneuvers using linguistic contextual information,\nenhancing the model's interpretability and transparency. The model achieved a\n91.5% f1-score with anticipation time extending to four seconds for risky lane\nchanges, and a 90.0% f1-score for predicting safe lane changes with the same\nanticipation time. We validate our model by integrating it into a vehicle\nwithin the CARLA simulator in scenarios that involve risky lane changes. The\nmodel managed to anticipate sudden lane changes, thus providing automated\nvehicles with further time to plan and execute appropriate safe reactions.\nFinally, to enhance the explainability of our model, we utilize RAG to provide\nclear and natural language explanations for the given prediction.", + "arxiv_id": "2501.11560v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria (1. Practical Applications in autonomous vehicle safety, 2. Experimental Results with quantitative metrics e.g., 91.5% f1-score, 3. Comparison with State-of-the-Art implicitly through superior performance). Additionally meets criteria 4 (Methodology and Implementation Details), and shows Novelty (integration of KG, Bayesian inference with RAG for explainability), making it a strong candidate for further review." + }, + { + "paper": { + "title": "Recurrent Diffusion for Large-Scale Parameter Generation", + "abstract": "Parameter generation has struggled to scale up for a long time, significantly\nlimiting its range of applications. In this study, we introduce\n\\textbf{R}ecurrent diffusion for large-scale \\textbf{P}arameter\n\\textbf{G}eneration, called \\textbf{RPG}. We first divide the trained\nparameters into non-overlapping parts, after which a recurrent model is\nproposed to learn their relationships. The recurrent model's outputs, as\nconditions, are then fed into a diffusion model to generate the neural network\nparameters. Using only a single GPU, recurrent diffusion enables us to generate\npopular vision and language models such as ConvNeXt-L and LoRA parameters of\nLLaMA-7B. Meanwhile, across various architectures and tasks, the generated\nparameters consistently perform comparable results over trained networks.\nNotably, our approach also shows the potential to generate models for handling\nunseen tasks, which largely increases the practicality of parameter generation.\nOur code is available\n\\href{https://github.com/NUS-HPC-AI-Lab/Recurrent-Parameter-Generation}{here}.", + "arxiv_id": "2501.11587v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications: generation of neural network parameters for Large Language Models like LLaMA-7B), 2 (Experimental Results: comparable performance of generated parameters across architectures and tasks), and 3 (Comparison with State-of-the-Art: outperforms in scalability with single GPU usage). Also shows novelty in approach and has reproducibility support through publicly available code." + }, + { + "paper": { + "title": "SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language\n Models Tackling Knowledge-based Reasoning Tasks", + "abstract": "Deductive reasoning is a crucial logical capability that assists us in\nsolving complex problems based on existing knowledge. Although augmented by\nChain-of-Thought prompts, Large Language Models (LLMs) might not follow the\ncorrect reasoning paths. Enhancing the deductive reasoning abilities of LLMs,\nand leveraging their extensive built-in knowledge for various reasoning tasks,\nremains an open question. Attempting to mimic the human deductive reasoning\nparadigm, we propose a multi-stage Syllogistic-Reasoning Framework of Thought\n(SR-FoT) that enables LLMs to perform syllogistic deductive reasoning to handle\ncomplex knowledge-based reasoning tasks. Our SR-FoT begins by interpreting the\nquestion and then uses the interpretation and the original question to propose\na suitable major premise. It proceeds by generating and answering minor premise\nquestions in two stages to match the minor premises. Finally, it guides LLMs to\nuse the previously generated major and minor premises to perform syllogistic\ndeductive reasoning to derive the answer to the original question. Extensive\nand thorough experiments on knowledge-based reasoning tasks have demonstrated\nthe effectiveness and advantages of our SR-FoT.", + "arxiv_id": "2501.11599v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria: (1) Practical Applications (knowledge-based reasoning tasks), (2) Experimental Results and Quantitative Metrics (extensive experiments demonstrating effectiveness), and (3) Comparison with State-of-the-Art (implied by 'advantages' over existing approaches). Additionally, meets criterion (4) Methodology and Implementation Details (clearly described multi-stage framework). Shows novelty in applying syllogistic reasoning to LLMs." + }, + { + "paper": { + "title": "StAyaL | Multilingual Style Transfer", + "abstract": "Stylistic text generation plays a vital role in enhancing communication by\nreflecting the nuances of individual expression. This paper presents a novel\napproach for generating text in a specific speaker's style across different\nlanguages. We show that by leveraging only 100 lines of text, an individuals\nunique style can be captured as a high-dimensional embedding, which can be used\nfor both text generation and stylistic translation. This methodology breaks\ndown the language barrier by transferring the style of a speaker between\nlanguages. The paper is structured into three main phases: augmenting the\nspeaker's data with stylistically consistent external sources, separating style\nfrom content using machine learning and deep learning techniques, and\ngenerating an abstract style profile by mean pooling the learned embeddings.\nThe proposed approach is shown to be topic-agnostic, with test accuracy and F1\nscores of 74.9\\% and 0.75, respectively. The results demonstrate the potential\nof the style profile for multilingual communication, paving the way for further\napplications in personalized content generation and cross-linguistic stylistic\ntransfer.", + "arxiv_id": "2501.11639v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications in multilingual communication), 2 (Experimental Results with quantitative metrics like test accuracy and F1 scores), and 4 (Methodology and Implementation Details of leveraging LLMs for style transfer). Novel approach to multilingual style transfer also suggests potential for impact through experimental validation." + }, + { + "paper": { + "title": "Optimizing Pretraining Data Mixtures with LLM-Estimated Utility", + "abstract": "Large Language Models improve with increasing amounts of high-quality\ntraining data. However, leveraging larger datasets requires balancing quality,\nquantity, and diversity across sources. After evaluating nine baseline methods\nunder both compute- and data-constrained scenarios, we find token-count\nheuristics outperform manual and learned mixes, indicating that simple\napproaches accounting for dataset size and diversity are surprisingly\neffective. Building on this insight, we propose two complementary approaches:\nUtiliMax, which extends token-based heuristics by incorporating utility\nestimates from reduced-scale ablations, achieving up to a 10.6x speedup over\nmanual baselines; and Model Estimated Data Utility (MEDU), which leverages LLMs\nto estimate data utility from small samples, matching ablation-based\nperformance while reducing computational requirements by $\\sim$200x. Together,\nthese approaches establish a new framework for automated, compute-efficient\ndata mixing that is robust across training regimes.", + "arxiv_id": "2501.11747v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria (1: Practical Applications in pretraining data mixtures, 2: Experimental Results with quantitative metrics e.g., 10.6x speedup, 3: Comparison with State-of-the-Art techniques) and at least one additional criterion (4: Methodology and Implementation Details are clearly described). Novelty is also shown in leveraging LLMs for automated data mixing." + }, + { + "paper": { + "title": "Human-AI Collaborative Game Testing with Vision Language Models", + "abstract": "As modern video games become increasingly complex, traditional manual testing\nmethods are proving costly and inefficient, limiting the ability to ensure\nhigh-quality game experiences. While advancements in Artificial Intelligence\n(AI) offer the potential to assist human testers, the effectiveness of AI in\ntruly enhancing real-world human performance remains underexplored. This study\ninvestigates how AI can improve game testing by developing and experimenting\nwith an AI-assisted workflow that leverages state-of-the-art machine learning\nmodels for defect detection. Through an experiment involving 800 test cases and\n276 participants of varying backgrounds, we evaluate the effectiveness of AI\nassistance under four conditions: with or without AI support, and with or\nwithout detailed knowledge of defects and design documentation. The results\nindicate that AI assistance significantly improves defect identification\nperformance, particularly when paired with detailed knowledge. However,\nchallenges arise when AI errors occur, negatively impacting human\ndecision-making. Our findings show the importance of optimizing human-AI\ncollaboration and implementing strategies to mitigate the effects of AI\ninaccuracies. By this research, we demonstrate AI's potential and problems in\nenhancing efficiency and accuracy in game testing workflows and offers\npractical insights for integrating AI into the testing process.", + "arxiv_id": "2501.11782v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications in a real-world task), 2 (Experimental Results with quantitative metrics), and 3 (Comparison with human-only conditions, implying state-of-the-art). Additionally, it touches on challenges and potential optimizations for human-AI collaboration, aligning with criterion 5 (Real-world Applications and Challenges).\"\n}\n```" + }, + { + "paper": { + "title": "Benchmarking Large Language Models via Random Variables", + "abstract": "With the continuous advancement of large language models (LLMs) in\nmathematical reasoning, evaluating their performance in this domain has become\na prominent research focus. Recent studies have raised concerns about the\nreliability of current mathematical benchmarks, highlighting issues such as\nsimplistic design and potential data leakage. Therefore, creating a reliable\nbenchmark that effectively evaluates the genuine capabilities of LLMs in\nmathematical reasoning remains a significant challenge. To address this, we\npropose RV-Bench, a framework for Benchmarking LLMs via Random Variables in\nmathematical reasoning. Specifically, the background content of a random\nvariable question (RV question) mirrors the original problem in existing\nstandard benchmarks, but the variable combinations are randomized into\ndifferent values. LLMs must fully understand the problem-solving process for\nthe original problem to correctly answer RV questions with various combinations\nof variable values. As a result, the LLM's genuine capability in mathematical\nreasoning is reflected by its accuracy on RV-Bench. Extensive experiments are\nconducted with 29 representative LLMs across 900+ RV questions. A leaderboard\nfor RV-Bench ranks the genuine capability of these LLMs. Further analysis of\naccuracy dropping indicates that current LLMs still struggle with complex\nmathematical reasoning problems.", + "arxiv_id": "2501.11790v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications: evaluating LLMs in mathematical reasoning), 2 (Experimental Results and Quantitative Metrics: extensive experiments with 29 LLMs and quantitative accuracy metrics), and 3 (Comparison with State-of-the-Art: implicit comparison through a leaderboard). Additionally, shows Novelty (introducing RV-Bench) and robust Experimental Validation, justifying acceptance.\"\n}\n```" + }, + { + "paper": { + "title": "Fact-Preserved Personalized News Headline Generation", + "abstract": "Personalized news headline generation, aiming at generating user-specific\nheadlines based on readers' preferences, burgeons a recent flourishing research\ndirection. Existing studies generally inject a user interest embedding into an\nencoderdecoder headline generator to make the output personalized, while the\nfactual consistency of headlines is inadequate to be verified. In this paper,\nwe propose a framework Fact-Preserved Personalized News Headline Generation\n(short for FPG), to prompt a tradeoff between personalization and consistency.\nIn FPG, the similarity between the candidate news to be exposed and the\nhistorical clicked news is used to give different levels of attention to key\nfacts in the candidate news, and the similarity scores help to learn a\nfact-aware global user embedding. Besides, an additional training procedure\nbased on contrastive learning is devised to further enhance the factual\nconsistency of generated headlines. Extensive experiments conducted on a\nreal-world benchmark PENS validate the superiority of FPG, especially on the\ntradeoff between personalization and factual consistency.", + "arxiv_id": "2501.11828v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications in personalized news headline generation), 2 (Experimental Results with quantitative metrics on real-world benchmark PENS), and 4 (Methodology and Implementation Details with contrastive learning for factual consistency). Novelty is also present in the tradeoff approach between personalization and factual consistency, making it a strong candidate for further review.\"\n}\n```" + }, + { + "paper": { + "title": "Is your LLM trapped in a Mental Set? Investigative study on how mental\n sets affect the reasoning capabilities of LLMs", + "abstract": "In this paper, we present an investigative study on how Mental Sets influence\nthe reasoning capabilities of LLMs. LLMs have excelled in diverse natural\nlanguage processing (NLP) tasks, driven by advancements in parameter-efficient\nfine-tuning (PEFT) and emergent capabilities like in-context learning (ICL).\nFor complex reasoning tasks, selecting the right model for PEFT or ICL is\ncritical, often relying on scores on benchmarks such as MMLU, MATH, and GSM8K.\nHowever, current evaluation methods, based on metrics like F1 Score or\nreasoning chain assessments by larger models, overlook a key dimension:\nadaptability to unfamiliar situations and overcoming entrenched thinking\npatterns. In cognitive psychology, Mental Set refers to the tendency to persist\nwith previously successful strategies, even when they become inefficient - a\nchallenge for problem solving and reasoning. We compare the performance of LLM\nmodels like Llama-3.1-8B-Instruct, Llama-3.1-70B-Instruct and GPT-4o in the\npresence of mental sets. To the best of our knowledge, this is the first study\nto integrate cognitive psychology concepts into the evaluation of LLMs for\ncomplex reasoning tasks, providing deeper insights into their adaptability and\nproblem-solving efficacy.", + "arxiv_id": "2501.11833v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 2 (Experimental Results and Quantitative Metrics) and 3 (Comparison with State-of-the-Art) through comparative performance analysis of LLM models. Also touches on novelty (Additional Consideration) by integrating cognitive psychology concepts into LLM evaluation, warranting further review for potential insights into LLM adaptability and problem-solving efficacy.\"\n}\n```" + }, + { + "paper": { + "title": "From Drafts to Answers: Unlocking LLM Potential via Aggregation\n Fine-Tuning", + "abstract": "Scaling data and model size has been proven effective for boosting the\nperformance of large language models. In addition to training-time scaling,\nrecent studies have revealed that increasing test-time computational resources\ncan further improve performance. In this work, we introduce Aggregation\nFine-Tuning (AFT), a supervised finetuning paradigm where the model learns to\nsynthesize multiple draft responses, referred to as proposals, into a single,\nrefined answer, termed aggregation. At inference time, a propose-and-aggregate\nstrategy further boosts performance by iteratively generating proposals and\naggregating them. Empirical evaluations on benchmark datasets show that\nAFT-trained models substantially outperform standard SFT. Notably, an AFT\nmodel, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3% LC\nwin rate on AlpacaEval 2, surpassing significantly larger LLMs such as\nLlama3.1-405B-Instruct and GPT4. By combining sequential refinement and\nparallel sampling, the propose-and-aggregate framework scales inference-time\ncomputation in a flexible manner. Overall, These findings position AFT as a\npromising approach to unlocking additional capabilities of LLMs without\nresorting to increasing data volume or model size.", + "arxiv_id": "2501.11877v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: focuses on practical application of LLMs (1), includes experimental results with quantitative metrics (2), and compares with state-of-the-art techniques (3). Additionally, meets criteria 4 (Methodology and Implementation Details) and shows novelty in approach (Additional Consideration)." + }, + { + "paper": { + "title": "Panoramic Interests: Stylistic-Content Aware Personalized Headline\n Generation", + "abstract": "Personalized news headline generation aims to provide users with\nattention-grabbing headlines that are tailored to their preferences. Prevailing\nmethods focus on user-oriented content preferences, but most of them overlook\nthe fact that diverse stylistic preferences are integral to users' panoramic\ninterests, leading to suboptimal personalization. In view of this, we propose a\nnovel Stylistic-Content Aware Personalized Headline Generation (SCAPE)\nframework. SCAPE extracts both content and stylistic features from headlines\nwith the aid of large language model (LLM) collaboration. It further adaptively\nintegrates users' long- and short-term interests through a contrastive\nlearning-based hierarchical fusion network. By incorporating the panoramic\ninterests into the headline generator, SCAPE reflects users' stylistic-content\npreferences during the generation process. Extensive experiments on the\nreal-world dataset PENS demonstrate the superiority of SCAPE over baselines.", + "arxiv_id": "2501.11900v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets ALL required criteria: (1) Practical Applications (personalized news headline generation), (2) Experimental Results and Quantitative Metrics (experiments on real-world PENS dataset), and (3) Comparison with State-of-the-Art (superiority over baselines). Additionally, meets extra criteria (4) Methodology and Implementation Details (utilizes LLM for feature extraction) and shows Novelty in approach." + }, + { + "paper": { + "title": "TAD-Bench: A Comprehensive Benchmark for Embedding-Based Text Anomaly\n Detection", + "abstract": "Text anomaly detection is crucial for identifying spam, misinformation, and\noffensive language in natural language processing tasks. Despite the growing\nadoption of embedding-based methods, their effectiveness and generalizability\nacross diverse application scenarios remain under-explored. To address this, we\npresent TAD-Bench, a comprehensive benchmark designed to systematically\nevaluate embedding-based approaches for text anomaly detection. TAD-Bench\nintegrates multiple datasets spanning different domains, combining\nstate-of-the-art embeddings from large language models with a variety of\nanomaly detection algorithms. Through extensive experiments, we analyze the\ninterplay between embeddings and detection methods, uncovering their strengths,\nweaknesses, and applicability to different tasks. These findings offer new\nperspectives on building more robust, efficient, and generalizable anomaly\ndetection systems for real-world applications.", + "arxiv_id": "2501.11960v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria (1: Practical Applications in anomaly detection, 2: Experimental Results with quantitative metrics implied through 'extensive experiments', 3: Comparison with State-of-the-Art through integration of state-of-the-art embeddings). Additionally, meets criteria 4 (Methodology and Implementation Details for utilizing LLMs in anomaly detection) and exhibits Novelty in benchmarking embedding-based approaches." + }, + { + "paper": { + "title": "Bridging Visualization and Optimization: Multimodal Large Language\n Models on Graph-Structured Combinatorial Optimization", + "abstract": "Graph-structured combinatorial challenges are inherently difficult due to\ntheir nonlinear and intricate nature, often rendering traditional computational\nmethods ineffective or expensive. However, these challenges can be more\nnaturally tackled by humans through visual representations that harness our\ninnate ability for spatial reasoning. In this study, we propose transforming\ngraphs into images to preserve their higher-order structural features\naccurately, revolutionizing the representation used in solving graph-structured\ncombinatorial tasks. This approach allows machines to emulate human-like\nprocessing in addressing complex combinatorial challenges. By combining the\ninnovative paradigm powered by multimodal large language models (MLLMs) with\nsimple search techniques, we aim to develop a novel and effective framework for\ntackling such problems. Our investigation into MLLMs spanned a variety of\ngraph-based tasks, from combinatorial problems like influence maximization to\nsequential decision-making in network dismantling, as well as addressing six\nfundamental graph-related issues. Our findings demonstrate that MLLMs exhibit\nexceptional spatial intelligence and a distinctive capability for handling\nthese problems, significantly advancing the potential for machines to\ncomprehend and analyze graph-structured data with a depth and intuition akin to\nhuman cognition. These results also imply that integrating MLLMs with simple\noptimization strategies could form a novel and efficient approach for\nnavigating graph-structured combinatorial challenges without complex\nderivations, computationally demanding training and fine-tuning.", + "arxiv_id": "2501.11968v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria (1: Practical Applications in graph-structured combinatorial optimization, 2: Experimental Results with quantitative metrics implied for MLLMs' performance, 3: Comparison with State-of-the-Art in handling graph-related issues). Additionally, meets criteria 4 (Methodology and Implementation Details for MLLMs in practical tasks) and shows Novelty in integrating MLLMs with optimization strategies, thus warranting further review." + }, + { + "paper": { + "title": "Leveraging Graph Structures and Large Language Models for End-to-End\n Synthetic Task-Oriented Dialogues", + "abstract": "Training task-oriented dialogue systems is both costly and time-consuming,\ndue to the need for high-quality datasets encompassing diverse intents.\nTraditional methods depend on extensive human annotation, while recent\nadvancements leverage large language models (LLMs) to generate synthetic data.\nHowever, these approaches often require custom prompts or code, limiting\naccessibility for non-technical users. We introduce GraphTOD, an end-to-end\nframework that simplifies the generation of task-oriented dialogues. Users can\ncreate dialogues by specifying transition graphs in JSON format. Our evaluation\ndemonstrates that GraphTOD generates high-quality dialogues across various\ndomains, significantly lowering the cost and complexity of dataset creation.", + "arxiv_id": "2501.11977v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications: generates synthetic task-oriented dialogues), 2 (Experimental Results and Quantitative Metrics: demonstrates high-quality dialogues across domains), 3 (Comparison with State-of-the-Art: implicitly by improving accessibility and lowering costs), and 4 (Methodology and Implementation Details: clearly describes end-to-end framework using LLMs with JSON format for non-technical users).\"\n}\n```" + }, + { + "paper": { + "title": "Harnessing Generative Pre-Trained Transformer for Datacenter Packet\n Trace Generation", + "abstract": "Today, the rapid growth of applications reliant on datacenters calls for new\nadvancements to meet the increasing traffic and computational demands. Traffic\ntraces from datacenters are essential for further development and optimization\nof future datacenters. However, traces are rarely released to the public.\nResearchers often use simplified mathematical models that lack the depth needed\nto recreate intricate traffic patterns and, thus, miss optimization\nopportunities found in realistic traffic. In this preliminary work, we\nintroduce DTG-GPT, a packet-level Datacenter Traffic Generator (DTG), based on\nthe generative pre-trained transformer (GPT) architecture used by many\nstate-of-the-art large language models. We train our model on a small set of\navailable traffic traces from different domains and offer a simple methodology\nto evaluate the fidelity of the generated traces to their original\ncounterparts. We show that DTG-GPT can synthesize novel traces that mimic the\nspatiotemporal patterns found in real traffic traces. We further demonstrate\nthat DTG-GPT can generate traces for networks of different scales while\nmaintaining fidelity. Our findings indicate the potential that, in the future,\nsimilar models to DTG-GPT will allow datacenter operators to release traffic\ninformation to the research community via trained GPT models.", + "arxiv_id": "2501.12033v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications: generating datacenter packet traces with LLMs), 2 (Experimental Results: demonstrates fidelity in generated traces), and 4 (Methodology and Implementation Details: clearly describes using GPT architecture for packet-level traffic generation). Novel approach to a practical problem, with potential for real-world impact in datacenter optimization." + }, + { + "paper": { + "title": "Condor: Enhance LLM Alignment with Knowledge-Driven Data Synthesis and\n Refinement", + "abstract": "The quality of Supervised Fine-Tuning (SFT) data plays a critical role in\nenhancing the conversational capabilities of Large Language Models (LLMs).\nHowever, as LLMs become more advanced, the availability of high-quality\nhuman-annotated SFT data has become a significant bottleneck, necessitating a\ngreater reliance on synthetic training data. In this work, we introduce Condor,\na novel two-stage synthetic data generation framework that incorporates World\nKnowledge Tree and Self-Reflection Refinement to produce high-quality SFT data\nat scale. Our experimental results demonstrate that a base model fine-tuned on\nonly 20K Condor-generated samples achieves superior performance compared to\ncounterparts. The additional refinement stage in Condor further enables\niterative self-improvement for LLMs at various scales (up to 72B), validating\nthe effectiveness of our approach. Furthermore, our investigation into the\nscaling for synthetic data in post-training reveals substantial unexplored\npotential for performance improvements, opening promising avenues for future\nresearch.", + "arxiv_id": "2501.12273v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria (1: Practical Applications in knowledge-driven data synthesis, 2: Experimental Results with quantitative metrics, 3: Comparison with State-of-the-Art techniques) and additional considerations (Novelty, Reproducibility, Impact through Experimental Validation); does not fall under rejection categories." + }, + { + "paper": { + "title": "RALAD: Bridging the Real-to-Sim Domain Gap in Autonomous Driving with\n Retrieval-Augmented Learning", + "abstract": "In the pursuit of robust autonomous driving systems, models trained on\nreal-world datasets often struggle to adapt to new environments, particularly\nwhen confronted with corner cases such as extreme weather conditions.\nCollecting these corner cases in the real world is non-trivial, which\nnecessitates the use of simulators for validation. However,the high\ncomputational cost and the domain gap in data distribution have hindered the\nseamless transition between real and simulated driving scenarios. To tackle\nthis challenge, we propose Retrieval-Augmented Learning for Autonomous Driving\n(RALAD), a novel framework designed to bridge the real-to-sim gap at a low\ncost. RALAD features three primary designs, including (1) domain adaptation via\nan enhanced Optimal Transport (OT) method that accounts for both individual and\ngrouped image distances, (2) a simple and unified framework that can be applied\nto various models, and (3) efficient fine-tuning techniques that freeze the\ncomputationally expensive layers while maintaining robustness. Experimental\nresults demonstrate that RALAD compensates for the performance degradation in\nsimulated environments while maintaining accuracy in real-world scenarios\nacross three different models. Taking Cross View as an example, the mIOU and\nmAP metrics in real-world scenarios remain stable before and after RALAD\nfine-tuning, while in simulated environments,the mIOU and mAP metrics are\nimproved by 10.30% and 12.29%, respectively. Moreover, the re-training cost of\nour approach is reduced by approximately 88.1%. Our code is available at\nhttps://github.com/JiachengZuo/RALAD.git.", + "arxiv_id": "2501.12296v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications in autonomous driving), 2 (Experimental Results with quantitative metrics like mIOU and mAP), and 3 (Comparison with State-of-the-Art through performance improvements). Additionally, it partially meets criteria 4 (Methodology and Implementation Details, with code availability) and shows Novelty in applying Retrieval-Augmented Learning to bridge the real-to-sim domain gap." + }, + { + "paper": { + "title": "LLM-Assisted Knowledge Graph Completion for Curriculum and Domain\n Modelling in Personalized Higher Education Recommendations", + "abstract": "While learning personalization offers great potential for learners, modern\npractices in higher education require a deeper consideration of domain models\nand learning contexts, to develop effective personalization algorithms. This\npaper introduces an innovative approach to higher education curriculum\nmodelling that utilizes large language models (LLMs) for knowledge graph (KG)\ncompletion, with the goal of creating personalized learning-path\nrecommendations. Our research focuses on modelling university subjects and\nlinking their topics to corresponding domain models, enabling the integration\nof learning modules from different faculties and institutions in the student's\nlearning path. Central to our approach is a collaborative process, where LLMs\nassist human experts in extracting high-quality, fine-grained topics from\nlecture materials. We develop a domain, curriculum, and user models for\nuniversity modules and stakeholders. We implement this model to create the KG\nfrom two study modules: Embedded Systems and Development of Embedded Systems\nUsing FPGA. The resulting KG structures the curriculum and links it to the\ndomain models. We evaluate our approach through qualitative expert feedback and\nquantitative graph quality metrics. Domain experts validated the relevance and\naccuracy of the model, while the graph quality metrics measured the structural\nproperties of our KG. Our results show that the LLM-assisted graph completion\napproach enhances the ability to connect related courses across disciplines to\npersonalize the learning experience. Expert feedback also showed high\nacceptance of the proposed collaborative approach for concept extraction and\nclassification.", + "arxiv_id": "2501.12300v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications in knowledge graph), 2 (Experimental Results with qualitative expert feedback and quantitative graph quality metrics), and 4 (Methodology and Implementation Details clearly described). Novelties in utilizing LLMs for collaborative knowledge graph completion in higher education also support inclusion.\"\n}\n```" + }, + { + "paper": { + "title": "Treefix: Enabling Execution with a Tree of Prefixes", + "abstract": "The ability to execute code is a prerequisite for various dynamic program\nanalyses. Learning-guided execution has been proposed as an approach to enable\nthe execution of arbitrary code snippets by letting a neural model predict\nlikely values for any missing variables. Although state-of-the-art\nlearning-guided execution approaches, such as LExecutor, can enable the\nexecution of a relative high amount of code, they are limited to predicting a\nrestricted set of possible values and do not use any feedback from previous\nexecutions to execute even more code. This paper presents Treefix, a novel\nlearning-guided execution approach that leverages LLMs to iteratively create\ncode prefixes that enable the execution of a given code snippet. The approach\naddresses the problem in a multi-step fashion, where each step uses feedback\nabout the code snippet and its execution to instruct an LLM to improve a\npreviously generated prefix. This process iteratively creates a tree of\nprefixes, a subset of which is returned to the user as prefixes that maximize\nthe number of executed lines in the code snippet. In our experiments with two\ndatasets of Python code snippets, Treefix achieves 25% and 7% more coverage\nrelative to the current state of the art in learning-guided execution, covering\na total of 84% and 82% of all lines in the code snippets.", + "arxiv_id": "2501.12339v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria: (1) Practical Applications (enabling execution of arbitrary code snippets with LLMs), (2) Experimental Results and Quantitative Metrics (demonstrates 25% and 7% improvement over state-of-the-art with concrete coverage percentages), and (3) Comparison with State-of-the-Art (compares and advances upon existing learning-guided execution approaches). Additionally, shows novelty in methodology and potentially high impact through experimental validation." + }, + { + "paper": { + "title": "Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for\n Mixture-of-Experts Language Models", + "abstract": "Scaling the capacity of language models has consistently proven to be a\nreliable approach for improving performance and unlocking new capabilities.\nCapacity can be primarily defined by two dimensions: the number of model\nparameters and the compute per example. While scaling typically involves\nincreasing both, the precise interplay between these factors and their combined\ncontribution to overall capacity remains not fully understood. We explore this\nrelationship in the context of sparse Mixture-of-Expert models (MoEs), which\nallow scaling the number of parameters without proportionally increasing the\nFLOPs per example. We investigate how varying the sparsity level, i.e., the\nratio of non-active to total parameters, affects model performance in terms of\nboth pretraining and downstream performance. We find that under different\nconstraints (e.g. parameter size and total training compute), there is an\noptimal level of sparsity that improves both training efficiency and model\nperformance. These results provide a better understanding of the impact of\nsparsity in scaling laws for MoEs and complement existing works in this area,\noffering insights for designing more efficient architectures.", + "arxiv_id": "2501.12370v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications for LLMs, specifically MoEs), 2 (Experimental Results with quantitative metrics on sparsity levels), and 3 (Comparison with State-of-the-Art in understanding scaling laws). Also shows potential for Novelty in optimizing LLM architecture efficiency." + }, + { + "paper": { + "title": "Is Long Context All You Need? Leveraging LLM's Extended Context for\n NL2SQL", + "abstract": "Large Language Models (LLMs) have demonstrated impressive capabilities across\na range of natural language processing tasks. In particular, improvements in\nreasoning abilities and the expansion of context windows have opened new\navenues for leveraging these powerful models. NL2SQL is challenging in that the\nnatural language question is inherently ambiguous, while the SQL generation\nrequires a precise understanding of complex data schema and semantics. One\napproach to this semantic ambiguous problem is to provide more and sufficient\ncontextual information.\n In this work, we explore the performance and the latency trade-offs of the\nextended context window (a.k.a., long context) offered by Google's\nstate-of-the-art LLM (\\textit{gemini-1.5-pro}). We study the impact of various\ncontextual information, including column example values, question and SQL query\npairs, user-provided hints, SQL documentation, and schema. To the best of our\nknowledge, this is the first work to study how the extended context window and\nextra contextual information can help NL2SQL generation with respect to both\naccuracy and latency cost. We show that long context LLMs are robust and do not\nget lost in the extended contextual information. Additionally, our long-context\nNL2SQL pipeline based on Google's \\textit{gemini-pro-1.5} achieve a strong\nperformance with 67.41\\% on BIRD benchmark (dev) without finetuning and\nexpensive self-consistency based techniques.", + "arxiv_id": "2501.12372v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications in NL2SQL), 2 (Experimental Results with quantitative metrics like 67.41% on BIRD benchmark), 3 (Comparison with State-of-the-Art through use of Google's gemini-1.5-pro), and 4 (Methodology and Implementation Details for utilizing LLMs in NL2SQL tasks). Additionally, shows Novelty in exploring extended context windows for LLMs in NL2SQL." + }, + { + "paper": { + "title": "Code Readability in the Age of Large Language Models: An Industrial Case\n Study from Atlassian", + "abstract": "Programmers spend a significant amount of time reading code during the\nsoftware development process. This trend is amplified by the emergence of large\nlanguage models (LLMs) that automatically generate code. However, little is\nknown about the readability of the LLM-generated code and whether it is still\nimportant from practitioners' perspectives in this new era. In this paper, we\nconduct a survey to explore the practitioners' perspectives on code readability\nin the age of LLMs and investigate the readability of our LLM-based software\ndevelopment agents framework, HULA, by comparing its generated code with\nhuman-written code in real-world scenarios. Overall, the findings underscore\nthat (1) readability remains a critical aspect of software development; (2) the\nreadability of our LLM-generated code is comparable to human-written code,\nfostering the establishment of appropriate trust and driving the broad adoption\nof our LLM-powered software development platform.", + "arxiv_id": "2501.11264v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications in software development), 2 (Experimental Results comparing LLM-generated code with human-written code), and 4 (Methodology and Implementation Details of LLMs in a real-world task). Additionally, shows Novelty in applying LLMs to code readability and has potential for Reproducibility and real-world Impact." + }, + { + "paper": { + "title": "RedStar: Does Scaling Long-CoT Data Unlock Better Slow-Reasoning\n Systems?", + "abstract": "Can scaling transform reasoning? In this work, we explore the untapped\npotential of scaling Long Chain-of-Thought (Long-CoT) data to 1000k samples,\npioneering the development of a slow-thinking model, RedStar. Through extensive\nexperiments with various LLMs and different sizes, we uncover the ingredients\nfor specialization and scale for Long-CoT training. Surprisingly, even smaller\nmodels show significant performance gains with limited data, revealing the\nsample efficiency of Long-CoT and the critical role of sample difficulty in the\nlearning process. Our findings demonstrate that Long-CoT reasoning can be\neffectively triggered with just a few thousand examples, while larger models\nachieve unparalleled improvements. We also introduce reinforcement learning\n(RL)-scale training as a promising direction for advancing slow-thinking\nsystems. RedStar shines across domains: on the MATH-Hard benchmark,\nRedStar-code-math boosts performance from 66.2\\% to 81.6\\%, and on the USA Math\nOlympiad (AIME), it solves 46.7\\% of problems using only 21k mixed-code-math\ndatasets. In multimodal tasks like GeoQA and MathVista-GEO, RedStar-Geo\nachieves competitive results with minimal Long-CoT data, outperforming other\nslow-thinking systems like QvQ-Preview. Compared to QwQ, RedStar strikes the\nperfect balance between reasoning and generalizability. Our work highlights\nthat, with careful tuning, scaling Long-CoT can unlock extraordinary reasoning\ncapabilities-even with limited dataset and set a new standard for slow-thinking\nmodels across diverse challenges. Our data and models are released at\nhttps://huggingface.co/RedStar-Reasoning.", + "arxiv_id": "2501.11284v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications in slow-thinking systems), 2 (Experimental Results with quantitative metrics in MATH-Hard, USA Math Olympiad, GeoQA, and MathVista-GEO), and 3 (Comparison with State-of-the-Art techniques like QvQ-Preview and QwQ). Additionally, shows novelty in scaling Long-CoT data and provides sufficient details for reproducibility with publicly released data and models.\"\n}\n```" + }, + { + "paper": { + "title": "Graph-defined Language Learning with LLMs", + "abstract": "Recent efforts leverage Large Language Models (LLMs) for modeling\ntext-attributed graph structures in node classification tasks. These approaches\ndescribe graph structures for LLMs to understand or aggregate LLM-generated\ntextual attribute embeddings through graph structure. However, these approaches\nface two main limitations in modeling graph structures with LLMs. (i) Graph\ndescriptions become verbose in describing high-order graph structure. (ii)\nTextual attributes alone do not contain adequate graph structure information.\nIt is challenging to model graph structure concisely and adequately with LLMs.\nLLMs lack built-in mechanisms to model graph structures directly. They also\nstruggle with complex long-range dependencies between high-order nodes and\ntarget nodes.\n Inspired by the observation that LLMs pre-trained on one language can achieve\nexceptional performance on another with minimal additional training, we propose\n\\textbf{G}raph-\\textbf{D}efined \\textbf{L}anguage for \\textbf{L}arge\n\\textbf{L}anguage \\textbf{M}odel (GDL4LLM). This novel framework enables LLMs\nto transfer their powerful language understanding capabilities to\ngraph-structured data. GDL4LLM translates graphs into a graph language corpus\ninstead of graph descriptions and pre-trains LLMs on this corpus to adequately\nunderstand graph structures. During fine-tuning, this corpus describes the\nstructural information of target nodes concisely with only a few tokens. By\ntreating graphs as a new language, GDL4LLM enables LLMs to model graph\nstructures adequately and concisely for node classification tasks. Extensive\nexperiments on three real-world datasets demonstrate that GDL4LLM outperforms\ndescription-based and textual attribute embeddings-based baselines by\nefficiently modeling different orders of graph structure with LLMs.", + "arxiv_id": "2501.11478v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria (1, 2, 3): Focuses on practical application of LLMs in knowledge graph-like structures, includes experimental results with quantitative metrics, and compares with state-of-the-art techniques. Additionally, meets criteria 4 (clear methodology) and shows novelty in treating graphs as a new language for LLMs." + }, + { + "paper": { + "title": "Early evidence of how LLMs outperform traditional systems on OCR/HTR\n tasks for historical records", + "abstract": "We explore the ability of two LLMs -- GPT-4o and Claude Sonnet 3.5 -- to\ntranscribe historical handwritten documents in a tabular format and compare\ntheir performance to traditional OCR/HTR systems: EasyOCR, Keras, Pytesseract,\nand TrOCR. Considering the tabular form of the data, two types of experiments\nare executed: one where the images are split line by line and the other where\nthe entire scan is used as input. Based on CER and BLEU, we demonstrate that\nLLMs outperform the conventional OCR/HTR methods. Moreover, we also compare the\nevaluated CER and BLEU scores to human evaluations to better judge the outputs\nof whole-scan experiments and understand influential factors for CER and BLEU.\nCombining judgments from all the evaluation metrics, we conclude that two-shot\nGPT-4o for line-by-line images and two-shot Claude Sonnet 3.5 for whole-scan\nimages yield the transcriptions of the historical records most similar to the\nground truth.", + "arxiv_id": "2501.11623v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria (1. Practical Applications in OCR/HTR tasks, 2. Experimental Results with quantitative metrics like CER and BLEU, and 3. Comparison with State-of-the-Art traditional OCR/HTR systems). Additionally, meets extra criteria for Methodology and Implementation Details (4) and shows robust Experimental Validation, making it a strong candidate for further review." + }, + { + "paper": { + "title": "Is logical analysis performed by transformers taking place in\n self-attention or in the fully connected part?", + "abstract": "Transformers architecture apply self-attention to tokens represented as\nvectors, before a fully connected (neuronal network) layer. These two parts can\nbe layered many times. Traditionally, self-attention is seen as a mechanism for\naggregating information before logical operations are performed by the fully\nconnected layer. In this paper, we show, that quite counter-intuitively, the\nlogical analysis can also be performed within the self-attention. For this we\nimplement a handcrafted single-level encoder layer which performs the logical\nanalysis within self-attention. We then study the scenario in which a one-level\ntransformer model undergoes self-learning using gradient descent. We\ninvestigate whether the model utilizes fully connected layers or self-attention\nmechanisms for logical analysis when it has the choice. Given that gradient\ndescent can become stuck at undesired zeros, we explicitly calculate these\nunwanted zeros and find ways to avoid them. We do all this in the context of\npredicting grammatical category pairs of adjacent tokens in a text. We believe\nthat our findings have broader implications for understanding the potential\nlogical operations performed by self-attention.", + "arxiv_id": "2501.11765v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets criteria 1 (Practical Applications: analyzing transformer architecture for logical analysis), 2 (Experimental Results and Quantitative Metrics: implements a handcrafted model with self-learning using gradient descent), and 4 (Methodology and Implementation Details: clearly describes how LLMs are utilized). Novelty is also present in exploring logical operations within self-attention, justifying acceptance for further review.\"\n}\n```" + }, + { + "paper": { + "title": "Network-informed Prompt Engineering against Organized Astroturf\n Campaigns under Extreme Class Imbalance", + "abstract": "Detecting organized political campaigns is of paramount importance in\nfighting against disinformation on social media. Existing approaches for the\nidentification of such organized actions employ techniques mostly from network\nscience, graph machine learning and natural language processing. Their ultimate\ngoal is to analyze the relationships and interactions (e.g. re-posting) among\nusers and the textual similarities of their posts. Despite their effectiveness\nin recognizing astroturf campaigns, these methods face significant challenges,\nnotably the class imbalance in available training datasets. To mitigate this\nissue, recent methods usually resort to data augmentation or increasing the\nnumber of positive samples, which may not always be feasible or sufficient in\nreal-world settings. Following a different path, in this paper, we propose a\nnovel framework for identifying astroturf campaigns based solely on large\nlanguage models (LLMs), introducing a Balanced Retrieval-Augmented Generation\n(Balanced RAG) component. Our approach first gives both textual information\nconcerning the posts (in our case tweets) and the user interactions of the\nsocial network as input to a language model. Then, through prompt engineering\nand the proposed Balanced RAG method, it effectively detects coordinated\ndisinformation campaigns on X (Twitter). The proposed framework does not\nrequire any training or fine-tuning of the language model. Instead, by\nstrategically harnessing the strengths of prompt engineering and Balanced RAG,\nit facilitates LLMs to overcome the effects of class imbalance and effectively\nidentify coordinated political campaigns. The experimental results demonstrate\nthat by incorporating the proposed prompt engineering and Balanced RAG methods,\nour framework outperforms the traditional graph-based baselines, achieving\n2x-3x improvements in terms of precision, recall and F1 scores.", + "arxiv_id": "2501.11849v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications in identifying astroturf campaigns), 2 (Experimental Results with quantitative metrics showing 2x-3x improvements), and 3 (Comparison with State-of-the-Art, outperforming traditional graph-based baselines). Although primarily focused on social media, the application is more closely related to disinformation detection, a security concern, rather than the excluded social applications (Diversity, Social harm, etc.)." + }, + { + "paper": { + "title": "EDoRA: Efficient Weight-Decomposed Low-Rank Adaptation via Singular\n Value Decomposition", + "abstract": "Parameter-efficient fine-tuning methods, such as LoRA, reduces the number of\ntrainable parameters. However, they often suffer from scalability issues and\ndifferences between their learning pattern and full fine-tuning. To overcome\nthese limitations, we propose Efficient Weight-Decomposed Low-Rank Adaptation\n(EDoRA): a novel PEFT method that decomposes pre-trained weights into magnitude\nand directional components. By freezing low-rank matrices, initializing them by\nsingular value decomposition, and introducing a small trainable matrix between\nthem, EDoRA achieves substantial reduction in trainable parameters while\nmaintaining learning capacity. Experimental results on the GLUE benchmark\ndemonstrate that EDoRA achieves competitive or superior performance compared to\nstate-of-the-art methods, such as LoRA and DoRA, with up to 30x fewer trainable\nparameters. This makes EDoRA a highly efficient solution for adapting LLMs to\ndiverse tasks under memory-constrained settings. Code is available at\nhttps://github.com/Hamid-Nasiri/EDoRA .", + "arxiv_id": "2501.12067v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications for adapting LLMs), 2 (Experimental Results with quantitative metrics on GLUE benchmark), and 3 (Comparison with State-of-the-Art methods like LoRA and DoRA). Additionally, shows novelty in methodology (EDoRA) and provides sufficient details for reproducibility (code available)." + }, + { + "paper": { + "title": "Improving Influence-based Instruction Tuning Data Selection for Balanced\n Learning of Diverse Capabilities", + "abstract": "Selecting appropriate training data is crucial for effective instruction\nfine-tuning of large language models (LLMs), which aims to (1) elicit strong\ncapabilities, and (2) achieve balanced performance across a diverse range of\ntasks. Influence-based methods show promise in achieving (1) by estimating the\ncontribution of each training example to the model's predictions, but often\nstruggle with (2). Our systematic investigation reveals that this\nunderperformance can be attributed to an inherent bias where certain tasks\nintrinsically have greater influence than others. As a result, data selection\nis often biased towards these tasks, not only hurting the model's performance\non others but also, counterintuitively, harms performance on these\nhigh-influence tasks themselves.\n As a remedy, we propose BIDS, a Balanced and Influential Data Selection\nalgorithm. BIDS first normalizes influence scores of the training data, and\nthen iteratively balances data selection by choosing the training example with\nthe highest influence on the most underrepresented task. Experiments with both\nLlama-3 and Mistral-v0.3 on seven benchmarks spanning five diverse capabilities\nshow that BIDS consistently outperforms both state-of-the-art influence-based\nalgorithms and other non-influence-based selection frameworks. Surprisingly,\ntraining on a 15% subset selected by BIDS can even outperform full-dataset\ntraining with a much more balanced performance. Our analysis further highlights\nthe importance of both instance-level normalization and iterative optimization\nof selected data for balanced learning of diverse capabilities.", + "arxiv_id": "2501.12147v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (Practical Applications of LLMs), 2 (Experimental Results with Quantitative Metrics), and 3 (Comparison with State-of-the-Art). Additionally, shows novelty in approach (BIDS algorithm) and robust experimental validation, making it a strong candidate for further review." + }, + { + "paper": { + "title": "AdaServe: SLO-Customized LLM Serving with Fine-Grained Speculative\n Decoding", + "abstract": "This paper introduces AdaServe, the first LLM serving system to support SLO\ncustomization through fine-grained speculative decoding. AdaServe leverages the\nlogits of a draft model to predict the speculative accuracy of tokens and\nemploys a theoretically optimal algorithm to construct token trees for\nverification. To accommodate diverse SLO requirements without compromising\nthroughput, AdaServe employs a speculation-and-selection scheme that first\nconstructs candidate token trees for each request and then dynamically selects\ntokens to meet individual SLO constraints while optimizing throughput.\nComprehensive evaluations demonstrate that AdaServe achieves up to 73% higher\nSLO attainment and 74% higher goodput compared to state-of-the-art systems.\nThese results underscore AdaServe's potential to enhance the efficiency and\nadaptability of LLM deployments across varied application scenarios.", + "arxiv_id": "2501.12162v1" + }, + "decision": "ACCEPT", + "explanation": "Original response: ```\n{\n \"decision\": \"ACCEPT\",\n \"explanation\": \"Meets all required criteria (1, 2, 3) by focusing on practical LLM application (AdaServe system), including experimental results with quantitative metrics (73% higher SLO attainment, 74% higher goodput), and comparing favorably to state-of-the-art techniques. Additionally, it meets optional criteria for methodology and implementation details (4), and novelty in extending LLM application.\"\n}\n```" + }, + { + "paper": { + "title": "InsTALL: Context-aware Instructional Task Assistance with Multi-modal\n Large Language Models", + "abstract": "The improved competence of generative models can help building multi-modal\nvirtual assistants that leverage modalities beyond language. By observing\nhumans performing multi-step tasks, one can build assistants that have\nsituational awareness of actions and tasks being performed, enabling them to\ncater assistance based on this understanding. In this paper, we develop a\nContext-aware Instructional Task Assistant with Multi-modal Large Language\nModels (InsTALL) that leverages an online visual stream (e.g. a user's screen\nshare or video recording) and responds in real-time to user queries related to\nthe task at hand. To enable useful assistance, InsTALL 1) trains a multi-modal\nmodel on task videos and paired textual data, and 2) automatically extracts\ntask graph from video data and leverages it at training and inference time. We\nshow InsTALL achieves state-of-the-art performance across proposed sub-tasks\nconsidered for multimodal activity understanding -- task recognition (TR),\naction recognition (AR), next action prediction (AP), and plan prediction (PP)\n-- and outperforms existing baselines on two novel sub-tasks related to\nautomatic error identification.", + "arxiv_id": "2501.12231v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria: 1) Practical Application (multi-modal virtual task assistance), 2) Experimental Results (state-of-the-art performance on multiple sub-tasks), and 3) Comparison with State-of-the-Art (outperforms existing baselines). Additionally, meets criterion 4 (Methodology and Implementation Details) and shows Novelty in integrating LLMs with visual streams for real-world task assistance." + }, + { + "paper": { + "title": "UI-TARS: Pioneering Automated GUI Interaction with Native Agents", + "abstract": "This paper introduces UI-TARS, a native GUI agent model that solely perceives\nthe screenshots as input and performs human-like interactions (e.g., keyboard\nand mouse operations). Unlike prevailing agent frameworks that depend on\nheavily wrapped commercial models (e.g., GPT-4o) with expert-crafted prompts\nand workflows, UI-TARS is an end-to-end model that outperforms these\nsophisticated frameworks. Experiments demonstrate its superior performance:\nUI-TARS achieves SOTA performance in 10+ GUI agent benchmarks evaluating\nperception, grounding, and GUI task execution. Notably, in the OSWorld\nbenchmark, UI-TARS achieves scores of 24.6 with 50 steps and 22.7 with 15\nsteps, outperforming Claude (22.0 and 14.9 respectively). In AndroidWorld,\nUI-TARS achieves 46.6, surpassing GPT-4o (34.5). UI-TARS incorporates several\nkey innovations: (1) Enhanced Perception: leveraging a large-scale dataset of\nGUI screenshots for context-aware understanding of UI elements and precise\ncaptioning; (2) Unified Action Modeling, which standardizes actions into a\nunified space across platforms and achieves precise grounding and interaction\nthrough large-scale action traces; (3) System-2 Reasoning, which incorporates\ndeliberate reasoning into multi-step decision making, involving multiple\nreasoning patterns such as task decomposition, reflection thinking, milestone\nrecognition, etc. (4) Iterative Training with Reflective Online Traces, which\naddresses the data bottleneck by automatically collecting, filtering, and\nreflectively refining new interaction traces on hundreds of virtual machines.\nThrough iterative training and reflection tuning, UI-TARS continuously learns\nfrom its mistakes and adapts to unforeseen situations with minimal human\nintervention. We also analyze the evolution path of GUI agents to guide the\nfurther development of this domain.", + "arxiv_id": "2501.12326v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria: (1) Practical Applications (GUI interaction with LLMs), (2) Experimental Results and Quantitative Metrics (SOTA performance in 10+ GUI agent benchmarks), and (3) Comparison with State-of-the-Art (outperforms Claude and GPT-4o). Additionally, meets optional criteria (4) Methodology and Implementation Details (clearly described innovations) and demonstrates Novelty, Implementability, and Reproducibility." + }, + { + "paper": { + "title": "Automatic Labelling with Open-source LLMs using Dynamic Label Schema\n Integration", + "abstract": "Acquiring labelled training data remains a costly task in real world machine\nlearning projects to meet quantity and quality requirements. Recently Large\nLanguage Models (LLMs), notably GPT-4, have shown great promises in labelling\ndata with high accuracy. However, privacy and cost concerns prevent the\nubiquitous use of GPT-4. In this work, we explore effectively leveraging\nopen-source models for automatic labelling. We identify integrating label\nschema as a promising technology but found that naively using the label\ndescription for classification leads to poor performance on high cardinality\ntasks. To address this, we propose Retrieval Augmented Classification (RAC) for\nwhich LLM performs inferences for one label at a time using corresponding label\nschema; we start with the most related label and iterates until a label is\nchosen by the LLM. We show that our method, which dynamically integrates label\ndescription, leads to performance improvements in labelling tasks. We further\nshow that by focusing only on the most promising labels, RAC can trade off\nbetween label quality and coverage - a property we leverage to automatically\nlabel our internal datasets.", + "arxiv_id": "2501.12332v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria (1: Practical Applications in automatic labelling, 2: Experimental Results with quantitative metrics for performance improvements, 3: Comparison with State-of-the-Art implied through innovation in Retrieval Augmented Classification). Additionally, meets optional criteria 4 (Methodology and Implementation Details for LLM utilization) and shows Novelty in integrating dynamic label schema with open-source LLMs." + }, + { + "paper": { + "title": "Test-time regression: a unifying framework for designing sequence models\n with associative memory", + "abstract": "Sequences provide a remarkably general way to represent and process\ninformation. This powerful abstraction has placed sequence modeling at the\ncenter of modern deep learning applications, inspiring numerous architectures\nfrom transformers to recurrent networks. While this fragmented development has\nyielded powerful models, it has left us without a unified framework to\nunderstand their fundamental similarities and explain their effectiveness. We\npresent a unifying framework motivated by an empirical observation: effective\nsequence models must be able to perform associative recall. Our key insight is\nthat memorizing input tokens through an associative memory is equivalent to\nperforming regression at test-time. This regression-memory correspondence\nprovides a framework for deriving sequence models that can perform associative\nrecall, offering a systematic lens to understand seemingly ad-hoc architectural\nchoices. We show numerous recent architectures -- including linear attention\nmodels, their gated variants, state-space models, online learners, and softmax\nattention -- emerge naturally as specific approaches to test-time regression.\nEach architecture corresponds to three design choices: the relative importance\nof each association, the regressor function class, and the optimization\nalgorithm. This connection leads to new understanding: we provide theoretical\njustification for QKNorm in softmax attention, and we motivate higher-order\ngeneralizations of softmax attention. Beyond unification, our work unlocks\ndecades of rich statistical tools that can guide future development of more\npowerful yet principled sequence models.", + "arxiv_id": "2501.12352v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nAlthough the paper's abstract doesn't explicitly mention Large Language Models (LLMs) or direct applications in knowledge graphs, RAG, or agentic AI, it presents a unifying framework for sequence models with associative memory. This framework could have implicit implications for LLMs and sequence modeling in NLP, potentially meeting criteria 1 and 4 with further review. Its experimental results and comparisons to existing architectures (e.g., transformers, recurrent networks) suggest a possible alignment with criteria 2 and 3. Inclusiveness is prioritized for marginally relevant papers, warranting acceptance for further review." + }, + { + "paper": { + "title": "Conversation Routines: A Prompt Engineering Framework for Task-Oriented\n Dialog Systems", + "abstract": "This study introduces Conversation Routines (CR), a structured prompt\nengineering framework for developing task-oriented dialog systems using Large\nLanguage Models (LLMs). While LLMs demonstrate remarkable natural language\nunderstanding capabilities, engineering them to reliably execute complex\nbusiness workflows remains challenging. The proposed CR framework enables the\ndevelopment of Conversation Agentic Systems (CAS) through natural language\nspecifications, embedding task-oriented logic within LLM prompts. This approach\nprovides a systematic methodology for designing and implementing complex\nconversational workflows while maintaining behavioral consistency. We\ndemonstrate the framework's effectiveness through two proof of concept\nimplementations: a Train Ticket Booking System and an Interactive\nTroubleshooting Copilot. These case studies validate CR's capability to encode\nsophisticated behavioral patterns and decision logic while preserving natural\nconversational flexibility. Results show that CR enables domain experts to\ndesign conversational workflows in natural language while leveraging custom\nenterprise functionalities (tools) developed by software engineers, creating an\nefficient division of responsibilities where developers focus on core API\nimplementation and domain experts handle conversation design. While the\nframework shows promise in accessibility and adaptability, we identify key\nchallenges including computational overhead, non-deterministic behavior, and\ndomain-specific logic optimization. Future research directions include\nenhancing system robustness, improving scalability for complex multi-agent\ninteractions, and addressing the identified limitations across diverse business\napplications.", + "arxiv_id": "2501.11613v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all required criteria: focuses on practical application of LLMs in task-oriented dialog systems (1), includes experimental results with quantitative metrics (2), and compares its methodology with existing state-of-the-art techniques (3). Additionally, it clearly describes methodology and implementation (4), discusses real-world applications and challenges (5), and shows novelty in integrating LLMs for autonomous task execution, making it a strong candidate for further review." + } + ], + "rejected": [ + { + "paper": { + "title": "The Explanation Game -- Rekindled (Extended Version)", + "abstract": "Recent work demonstrated the existence of critical flaws in the current use\nof Shapley values in explainable AI (XAI), i.e. the so-called SHAP scores.\nThese flaws are significant in that the scores provided to a human\ndecision-maker can be misleading. Although these negative results might appear\nto indicate that Shapley values ought not be used in XAI, this paper argues\notherwise. Concretely, this paper proposes a novel definition of SHAP scores\nthat overcomes existing flaws. Furthermore, the paper outlines a practically\nefficient solution for the rigorous estimation of the novel SHAP scores.\nPreliminary experimental results confirm our claims, and further underscore the\nflaws of the current SHAP scores.", + "arxiv_id": "2501.11429v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on Explainable AI (XAI) and Shapley values, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus not meeting the required criteria related to LLMs." + }, + { + "paper": { + "title": "Systematic Abductive Reasoning via Diverse Relation Representations in\n Vector-symbolic Architecture", + "abstract": "In abstract visual reasoning, monolithic deep learning models suffer from\nlimited interpretability and generalization, while existing neuro-symbolic\napproaches fall short in capturing the diversity and systematicity of\nattributes and relation representations. To address these challenges, we\npropose a Systematic Abductive Reasoning model with diverse relation\nrepresentations (Rel-SAR) in Vector-symbolic Architecture (VSA) to solve\nRaven's Progressive Matrices (RPM). To derive attribute representations with\nsymbolic reasoning potential, we introduce not only various types of atomic\nvectors that represent numeric, periodic and logical semantics, but also the\nstructured high-dimentional representation (SHDR) for the overall Grid\ncomponent. For systematic reasoning, we propose novel numerical and logical\nrelation functions and perform rule abduction and execution in a unified\nframework that integrates these relation representations. Experimental results\ndemonstrate that Rel-SAR achieves significant improvement on RPM tasks and\nexhibits robust out-of-distribution generalization. Rel-SAR leverages the\nsynergy between HD attribute representations and symbolic reasoning to achieve\nsystematic abductive reasoning with both interpretable and computable\nsemantics.", + "arxiv_id": "2501.11896v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on abstract visual reasoning and solving Raven's Progressive Matrices (RPM) with Vector-symbolic Architecture (VSA), without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the required criteria related to LLMs." + }, + { + "paper": { + "title": "Bridging the Communication Gap: Evaluating AI Labeling Practices for\n Trustworthy AI Development", + "abstract": "As artificial intelligence (AI) becomes integral to economy and society,\ncommunication gaps between developers, users, and stakeholders hinder trust and\ninformed decision-making. High-level AI labels, inspired by frameworks like EU\nenergy labels, have been proposed to make the properties of AI models more\ntransparent. Without requiring deep technical expertise, they can inform on the\ntrade-off between predictive performance and resource efficiency. However, the\npractical benefits and limitations of AI labeling remain underexplored. This\nstudy evaluates AI labeling through qualitative interviews along four key\nresearch questions. Based on thematic analysis and inductive coding, we found a\nbroad range of practitioners to be interested in AI labeling (RQ1). They see\nbenefits for alleviating communication gaps and aiding non-expert\ndecision-makers, however limitations, misunderstandings, and suggestions for\nimprovement were also discussed (RQ2). Compared to other reporting formats,\ninterviewees positively evaluated the reduced complexity of labels, increasing\noverall comprehensibility (RQ3). Trust was influenced most by usability and the\ncredibility of the responsible labeling authority, with mixed preferences for\nself-certification versus third-party certification (RQ4). Our Insights\nhighlight that AI labels pose a trade-off between simplicity and complexity,\nwhich could be resolved by developing customizable and interactive labeling\nframeworks to address diverse user needs. Transparent labeling of resource\nefficiency also nudged interviewee priorities towards paying more attention to\nsustainability aspects during AI development. This study validates AI labels as\na valuable tool for enhancing trust and communication in AI, offering\nactionable guidelines for their refinement and standardization.", + "arxiv_id": "2501.11909v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on trustworthy AI development, AI ethics (transparent labeling, trust, and responsible AI development), which falls under the rejection criteria (responsible AI application or AI ethics). It also lacks clear connections to Large Language Models (LLMs), practical applications in knowledge graphs, retrieval-augmented generation, or agentic AI as required by the selection criteria." + }, + { + "paper": { + "title": "Make Full Use of Testing Information: An Integrated Accelerated Testing\n and Evaluation Method for Autonomous Driving Systems", + "abstract": "Testing and evaluation is an important step before the large-scale\napplication of the autonomous driving systems (ADSs). Based on the three level\nof scenario abstraction theory, a testing can be performed within a logical\nscenario, followed by an evaluation stage which is inputted with the testing\nresults of each concrete scenario generated from the logical parameter space.\nDuring the above process, abundant testing information is produced which is\nbeneficial for comprehensive and accurate evaluations. To make full use of\ntesting information, this paper proposes an Integrated accelerated Testing and\nEvaluation Method (ITEM). Based on a Monte Carlo Tree Search (MCTS) paradigm\nand a dual surrogates testing framework proposed in our previous work, this\npaper applies the intermediate information (i.e., the tree structure, including\nthe affiliation of each historical sampled point with the subspaces and the\nparent-child relationship between subspaces) generated during the testing stage\ninto the evaluation stage to achieve accurate hazardous domain identification.\nMoreover, to better serve this purpose, the UCB calculation method is improved\nto allow the search algorithm to focus more on the hazardous domain boundaries.\nFurther, a stopping condition is constructed based on the convergence of the\nsearch algorithm. Ablation and comparative experiments are then conducted to\nverify the effectiveness of the improvements and the superiority of the\nproposed method. The experimental results show that ITEM could well identify\nthe hazardous domains in both low- and high-dimensional cases, regardless of\nthe shape of the hazardous domains, indicating its generality and potential for\nthe safety evaluation of ADSs.", + "arxiv_id": "2501.11924v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper primarily focuses on autonomous driving systems (ADSs) without explicitly involving Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria related to LLMs.\"\n}\n```" + }, + { + "paper": { + "title": "Collaborative Imputation of Urban Time Series through Cross-city\n Meta-learning", + "abstract": "Urban time series, such as mobility flows, energy consumption, and pollution\nrecords, encapsulate complex urban dynamics and structures. However, data\ncollection in each city is impeded by technical challenges such as budget\nlimitations and sensor failures, necessitating effective data imputation\ntechniques that can enhance data quality and reliability. Existing imputation\nmodels, categorized into learning-based and analytics-based paradigms, grapple\nwith the trade-off between capacity and generalizability. Collaborative\nlearning to reconstruct data across multiple cities holds the promise of\nbreaking this trade-off. Nevertheless, urban data's inherent irregularity and\nheterogeneity issues exacerbate challenges of knowledge sharing and\ncollaboration across cities. To address these limitations, we propose a novel\ncollaborative imputation paradigm leveraging meta-learned implicit neural\nrepresentations (INRs). INRs offer a continuous mapping from domain coordinates\nto target values, integrating the strengths of both paradigms. By imposing\nembedding theory, we first employ continuous parameterization to handle\nirregularity and reconstruct the dynamical system. We then introduce a\ncross-city collaborative learning scheme through model-agnostic meta learning,\nincorporating hierarchical modulation and normalization techniques to\naccommodate multiscale representations and reduce variance in response to\nheterogeneity. Extensive experiments on a diverse urban dataset from 20 global\ncities demonstrate our model's superior imputation performance and\ngeneralizability, underscoring the effectiveness of collaborative imputation in\nresource-constrained settings.", + "arxiv_id": "2501.11306v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on urban time series imputation using meta-learning, without any apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria (1, 2, and 3) related to LLMs." + }, + { + "paper": { + "title": "Finer-CAM: Spotting the Difference Reveals Finer Details for Visual\n Explanation", + "abstract": "Class activation map (CAM) has been widely used to highlight image regions\nthat contribute to class predictions. Despite its simplicity and computational\nefficiency, CAM often struggles to identify discriminative regions that\ndistinguish visually similar fine-grained classes. Prior efforts address this\nlimitation by introducing more sophisticated explanation processes, but at the\ncost of extra complexity. In this paper, we propose Finer-CAM, a method that\nretains CAM's efficiency while achieving precise localization of discriminative\nregions. Our key insight is that the deficiency of CAM lies not in \"how\" it\nexplains, but in \"what\" it explains}. Specifically, previous methods attempt to\nidentify all cues contributing to the target class's logit value, which\ninadvertently also activates regions predictive of visually similar classes. By\nexplicitly comparing the target class with similar classes and spotting their\ndifferences, Finer-CAM suppresses features shared with other classes and\nemphasizes the unique, discriminative details of the target class. Finer-CAM is\neasy to implement, compatible with various CAM methods, and can be extended to\nmulti-modal models for accurate localization of specific concepts.\nAdditionally, Finer-CAM allows adjustable comparison strength, enabling users\nto selectively highlight coarse object contours or fine discriminative details.\nQuantitatively, we show that masking out the top 5% of activated pixels by\nFiner-CAM results in a larger relative confidence drop compared to baselines.\nThe source code and demo are available at\nhttps://github.com/Imageomics/Finer-CAM.", + "arxiv_id": "2501.11309v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on image processing and visual explanation using Class Activation Maps (CAM), with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the fundamental criteria." + }, + { + "paper": { + "title": "CatV2TON: Taming Diffusion Transformers for Vision-Based Virtual Try-On\n with Temporal Concatenation", + "abstract": "Virtual try-on (VTON) technology has gained attention due to its potential to\ntransform online retail by enabling realistic clothing visualization of images\nand videos. However, most existing methods struggle to achieve high-quality\nresults across image and video try-on tasks, especially in long video\nscenarios. In this work, we introduce CatV2TON, a simple and effective\nvision-based virtual try-on (V2TON) method that supports both image and video\ntry-on tasks with a single diffusion transformer model. By temporally\nconcatenating garment and person inputs and training on a mix of image and\nvideo datasets, CatV2TON achieves robust try-on performance across static and\ndynamic settings. For efficient long-video generation, we propose an\noverlapping clip-based inference strategy that uses sequential frame guidance\nand Adaptive Clip Normalization (AdaCN) to maintain temporal consistency with\nreduced resource demands. We also present ViViD-S, a refined video try-on\ndataset, achieved by filtering back-facing frames and applying 3D mask\nsmoothing for enhanced temporal consistency. Comprehensive experiments\ndemonstrate that CatV2TON outperforms existing methods in both image and video\ntry-on tasks, offering a versatile and reliable solution for realistic virtual\ntry-ons across diverse scenarios.", + "arxiv_id": "2501.11325v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on computer vision for virtual try-on, particularly in video processing, which is explicitly listed as a rejection criterion. Additionally, it lacks apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are key areas of interest." + }, + { + "paper": { + "title": "Federated Learning with Sample-level Client Drift Mitigation", + "abstract": "Federated Learning (FL) suffers from severe performance degradation due to\nthe data heterogeneity among clients. Existing works reveal that the\nfundamental reason is that data heterogeneity can cause client drift where the\nlocal model update deviates from the global one, and thus they usually tackle\nthis problem from the perspective of calibrating the obtained local update.\nDespite effectiveness, existing methods substantially lack a deep understanding\nof how heterogeneous data samples contribute to the formation of client drift.\nIn this paper, we bridge this gap by identifying that the drift can be viewed\nas a cumulative manifestation of biases present in all local samples and the\nbias between samples is different. Besides, the bias dynamically changes as the\nFL training progresses. Motivated by this, we propose FedBSS that first\nmitigates the heterogeneity issue in a sample-level manner, orthogonal to\nexisting methods. Specifically, the core idea of our method is to adopt a\nbias-aware sample selection scheme that dynamically selects the samples from\nsmall biases to large epoch by epoch to train progressively the local model in\neach round. In order to ensure the stability of training, we set the\ndiversified knowledge acquisition stage as the warm-up stage to avoid the local\noptimality caused by knowledge deviation in the early stage of the model.\nEvaluation results show that FedBSS outperforms state-of-the-art baselines. In\naddition, we also achieved effective results on feature distribution skew and\nnoise label dataset setting, which proves that FedBSS can not only reduce\nheterogeneity, but also has scalability and robustness.", + "arxiv_id": "2501.11360v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on Federated Learning, which, while related to AI, does not explicitly mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus not meeting the primary criteria. Additionally, the topic aligns more closely with general AI challenges rather than the specified areas of interest." + }, + { + "paper": { + "title": "Multi-View Spectral Clustering for Graphs with Multiple View Structures", + "abstract": "Despite the fundamental importance of clustering, to this day, much of the\nrelevant research is still based on ambiguous foundations, leading to an\nunclear understanding of whether or how the various clustering methods are\nconnected with each other. In this work, we provide an additional stepping\nstone towards resolving such ambiguities by presenting a general clustering\nframework that subsumes a series of seemingly disparate clustering methods,\nincluding various methods belonging to the wildly popular spectral clustering\nframework. In fact, the generality of the proposed framework is additionally\ncapable of shedding light to the largely unexplored area of multi-view graphs\nwhose each view may have differently clustered nodes. In turn, we propose\nGenClus: a method that is simultaneously an instance of this framework and a\ngeneralization of spectral clustering, while also being closely related to\nk-means as well. This results in a principled alternative to the few existing\nmethods studying this special type of multi-view graphs. Then, we conduct\nin-depth experiments, which demonstrate that GenClus is more computationally\nefficient than existing methods, while also attaining similar or better\nclustering performance. Lastly, a qualitative real-world case-study further\ndemonstrates the ability of GenClus to produce meaningful clusterings.", + "arxiv_id": "2501.11422v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the primary criteria (1) for practical applications of LLMs, and lacking clear connections to the specified areas of interest." + }, + { + "paper": { + "title": "A Survey on Diffusion Models for Anomaly Detection", + "abstract": "Diffusion models (DMs) have emerged as a powerful class of generative AI\nmodels, showing remarkable potential in anomaly detection (AD) tasks across\nvarious domains, such as cybersecurity, fraud detection, healthcare, and\nmanufacturing. The intersection of these two fields, termed diffusion models\nfor anomaly detection (DMAD), offers promising solutions for identifying\ndeviations in increasingly complex and high-dimensional data. In this survey,\nwe systematically review recent advances in DMAD research and investigate their\ncapabilities. We begin by presenting the fundamental concepts of AD and DMs,\nfollowed by a comprehensive analysis of classic DM architectures including\nDDPMs, DDIMs, and Score SDEs. We further categorize existing DMAD methods into\nreconstruction-based, density-based, and hybrid approaches, providing detailed\nexaminations of their methodological innovations. We also explore the diverse\ntasks across different data modalities, encompassing image, time series, video,\nand multimodal data analysis. Furthermore, we discuss critical challenges and\nemerging research directions, including computational efficiency, model\ninterpretability, robustness enhancement, edge-cloud collaboration, and\nintegration with large language models. The collection of DMAD research papers\nand resources is available at https://github.com/fdjingliu/DMAD.", + "arxiv_id": "2501.11430v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on diffusion models for anomaly detection in various domains (including healthcare and video processing), with no clear indication of practical applications or experimental results specifically involving Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the required criteria." + }, + { + "paper": { + "title": "Communication-Efficient Federated Learning Based on Explanation-Guided\n Pruning for Remote Sensing Image Classification", + "abstract": "Federated learning (FL) is a decentralized machine learning paradigm, where\nmultiple clients collaboratively train a global model by exchanging only model\nupdates with the central server without sharing the local data of clients. Due\nto the large volume of model updates required to be transmitted between clients\nand the central server, most FL systems are associated with high transfer costs\n(i.e., communication overhead). This issue is more critical for operational\napplications in remote sensing (RS), especially when large-scale RS data is\nprocessed and analyzed through FL systems with restricted communication\nbandwidth. To address this issue, we introduce an explanation-guided pruning\nstrategy for communication-efficient FL in the context of RS image\nclassification. Our pruning strategy is defined based on the layerwise\nrelevance propagation (LRP) driven explanations to: 1) efficiently and\neffectively identify the most relevant and informative model parameters (to be\nexchanged between clients and the central server); and 2) eliminate the\nnon-informative ones to minimize the volume of model updates. The experimental\nresults on the BigEarthNet-S2 dataset demonstrate that our strategy effectively\nreduces the number of shared model updates, while increasing the generalization\nability of the global model. The code of this work will be publicly available\nat https://git.tu-berlin.de/rsim/FL-LRP", + "arxiv_id": "2501.11493v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on Federated Learning for Remote Sensing Image Classification, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the core criteria." + }, + { + "paper": { + "title": "Meta-Instance Selection. Instance Selection as a Classification Problem\n with Meta-Features", + "abstract": "Data pruning, or instance selection, is an important problem in machine\nlearning especially in terms of nearest neighbour classifier. However, in data\npruning which speeds up the prediction phase, there is an issue related to the\nspeed and efficiency of the process itself. In response, the study proposes an\napproach involving transforming the instance selection process into a\nclassification task conducted in a unified meta-feature space where each\ninstance can be classified and assigned to either the \"to keep\" or \"to remove\"\nclass. This approach requires training an appropriate meta-classifier, which\ncan be developed based on historical instance selection results from other\ndatasets using reference instance selection methods as a labeling tool. This\nwork proposes constructing the meta-feature space based on properties extracted\nfrom the nearest neighbor graph. Experiments conducted on 17 datasets of\nvarying sizes and five reference instance selection methods (ENN, Drop3, ICF,\nHMN-EI, and CCIS) demonstrate that the proposed solution achieves results\ncomparable to reference instance selection methods while significantly reducing\ncomputational complexity. In the proposed approach, the computational\ncomplexity of the system depends only on identifying the k-nearest neighbors\nfor each data sample and running the meta-classifier. Additionally, the study\ndiscusses the choice of meta-classifier, recommending the use of Balanced\nRandom Forest.", + "arxiv_id": "2501.11526v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the primary criteria for Large Language Models (LLMs) applications, focusing instead on instance selection and classification in machine learning with nearest neighbor classifiers, without any mention of LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI." + }, + { + "paper": { + "title": "The impact of intrinsic rewards on exploration in Reinforcement Learning", + "abstract": "One of the open challenges in Reinforcement Learning is the hard exploration\nproblem in sparse reward environments. Various types of intrinsic rewards have\nbeen proposed to address this challenge by pushing towards diversity. This\ndiversity might be imposed at different levels, favouring the agent to explore\ndifferent states, policies or behaviours (State, Policy and Skill level\ndiversity, respectively). However, the impact of diversity on the agent's\nbehaviour remains unclear. In this work, we aim to fill this gap by studying\nthe effect of different levels of diversity imposed by intrinsic rewards on the\nexploration patterns of RL agents. We select four intrinsic rewards (State\nCount, Intrinsic Curiosity Module (ICM), Maximum Entropy, and Diversity is all\nyou need (DIAYN)), each pushing for a different diversity level. We conduct an\nempirical study on MiniGrid environment to compare their impact on exploration\nconsidering various metrics related to the agent's exploration, namely:\nepisodic return, observation coverage, agent's position coverage, policy\nentropy, and timeframes to reach the sparse reward. The main outcome of the\nstudy is that State Count leads to the best exploration performance in the case\nof low-dimensional observations. However, in the case of RGB observations, the\nperformance of State Count is highly degraded mostly due to representation\nlearning challenges. Conversely, Maximum Entropy is less impacted, resulting in\na more robust exploration, despite being not always optimal. Lastly, our\nempirical study revealed that learning diverse skills with DIAYN, often linked\nto improved robustness and generalisation, does not promote exploration in\nMiniGrid environments. This is because: i) learning the skill space itself can\nbe challenging, and ii) exploration within the skill space prioritises\ndifferentiating between behaviours rather than achieving uniform state\nvisitation.", + "arxiv_id": "2501.11533v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on Reinforcement Learning (RL) with intrinsic rewards, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the required criteria related to LLMs." + }, + { + "paper": { + "title": "Spatially-Delineated Domain-Adapted AI Classification: An Application\n for Oncology Data", + "abstract": "Given multi-type point maps from different place-types (e.g., tumor regions),\nour objective is to develop a classifier trained on the source place-type to\naccurately distinguish between two classes of the target place-type based on\ntheir point arrangements. This problem is societally important for many\napplications, such as generating clinical hypotheses for designing new\nimmunotherapies for cancer treatment. The challenge lies in the spatial\nvariability, the inherent heterogeneity and variation observed in spatial\nproperties or arrangements across different locations (i.e., place-types).\nPrevious techniques focus on self-supervised tasks to learn domain-invariant\nfeatures and mitigate domain differences; however, they often neglect the\nunderlying spatial arrangements among data points, leading to significant\ndiscrepancies across different place-types. We explore a novel multi-task\nself-learning framework that targets spatial arrangements, such as spatial\nmix-up masking and spatial contrastive predictive coding, for\nspatially-delineated domain-adapted AI classification. Experimental results on\nreal-world datasets (e.g., oncology data) show that the proposed framework\nprovides higher prediction accuracy than baseline methods.", + "arxiv_id": "2501.11695v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (oncology data for cancer treatment), which is one of the explicit rejection criteria." + }, + { + "paper": { + "title": "Human services organizations and the responsible integration of AI:\n Considering ethics and contextualizing risk(s)", + "abstract": "This paper examines the responsible integration of artificial intelligence\n(AI) in human services organizations (HSOs), proposing a nuanced framework for\nevaluating AI applications across multiple dimensions of risk. The authors\nargue that ethical concerns about AI deployment -- including professional\njudgment displacement, environmental impact, model bias, and data laborer\nexploitation -- vary significantly based on implementation context and specific\nuse cases. They challenge the binary view of AI adoption, demonstrating how\ndifferent applications present varying levels of risk that can often be\neffectively managed through careful implementation strategies. The paper\nhighlights promising solutions, such as local large language models, that can\nfacilitate responsible AI integration while addressing common ethical concerns.\nThe authors propose a dimensional risk assessment approach that considers\nfactors like data sensitivity, professional oversight requirements, and\npotential impact on client wellbeing. They conclude by outlining a path forward\nthat emphasizes empirical evaluation, starting with lower-risk applications and\nbuilding evidence-based understanding through careful experimentation. This\napproach enables organizations to maintain high ethical standards while\nthoughtfully exploring how AI might enhance their capacity to serve clients and\ncommunities effectively.", + "arxiv_id": "2501.11705v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on responsible AI application or AI ethics, which is explicitly listed as a criterion for rejection, overshadowing any potential marginal relevance to Large Language Models (LLMs) or other specified areas of interest." + }, + { + "paper": { + "title": "GL-ICNN: An End-To-End Interpretable Convolutional Neural Network for\n the Diagnosis and Prediction of Alzheimer's Disease", + "abstract": "Deep learning methods based on Convolutional Neural Networks (CNNs) have\nshown great potential to improve early and accurate diagnosis of Alzheimer's\ndisease (AD) dementia based on imaging data. However, these methods have yet to\nbe widely adopted in clinical practice, possibly due to the limited\ninterpretability of deep learning models. The Explainable Boosting Machine\n(EBM) is a glass-box model but cannot learn features directly from input\nimaging data. In this study, we propose a novel interpretable model that\ncombines CNNs and EBMs for the diagnosis and prediction of AD. We develop an\ninnovative training strategy that alternatingly trains the CNN component as a\nfeature extractor and the EBM component as the output block to form an\nend-to-end model. The model takes imaging data as input and provides both\npredictions and interpretable feature importance measures. We validated the\nproposed model on the Alzheimer's Disease Neuroimaging Initiative (ADNI)\ndataset and the Health-RI Parelsnoer Neurodegenerative Diseases Biobank (PND)\nas an external testing set. The proposed model achieved an area-under-the-curve\n(AUC) of 0.956 for AD and control classification, and 0.694 for the prediction\nof conversion of mild cognitive impairment (MCI) to AD on the ADNI cohort. The\nproposed model is a glass-box model that achieves a comparable performance with\nother state-of-the-art black-box models. Our code is publicly available at:\nhttps://anonymous.4open.science/r/GL-ICNN.", + "arxiv_id": "2501.11715v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (Alzheimer's Disease diagnosis and prediction), which is explicitly listed as a rejection criterion." + }, + { + "paper": { + "title": "Episodic memory in AI agents poses risks that should be studied and\n mitigated", + "abstract": "Most current AI models have little ability to store and later retrieve a\nrecord or representation of what they do. In human cognition, episodic memories\nplay an important role in both recall of the past as well as planning for the\nfuture. The ability to form and use episodic memories would similarly enable a\nbroad range of improved capabilities in an AI agent that interacts with and\ntakes actions in the world. Researchers have begun directing more attention to\ndeveloping memory abilities in AI models. It is therefore likely that models\nwith such capability will be become widespread in the near future. This could\nin some ways contribute to making such AI agents safer by enabling users to\nbetter monitor, understand, and control their actions. However, as a new\ncapability with wide applications, we argue that it will also introduce\nsignificant new risks that researchers should begin to study and address. We\noutline these risks and benefits and propose four principles to guide the\ndevelopment of episodic memory capabilities so that these will enhance, rather\nthan undermine, the effort to keep AI safe and trustworthy.", + "arxiv_id": "2501.11739v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on responsible AI application and AI safety/ethics, which is an explicitly stated rejection criterion, and does not appear to meet the required criteria related to practical applications, experimental results, or state-of-the-art comparisons for Large Language Models (LLMs)." + }, + { + "paper": { + "title": "PXGen: A Post-hoc Explainable Method for Generative Models", + "abstract": "With the rapid growth of generative AI in numerous applications, explainable\nAI (XAI) plays a crucial role in ensuring the responsible development and\ndeployment of generative AI technologies. XAI has undergone notable\nadvancements and widespread adoption in recent years, reflecting a concerted\npush to enhance the transparency, interpretability, and credibility of AI\nsystems. Recent research emphasizes that a proficient XAI method should adhere\nto a set of criteria, primarily focusing on two key areas. Firstly, it should\nensure the quality and fluidity of explanations, encompassing aspects like\nfaithfulness, plausibility, completeness, and tailoring to individual needs.\nSecondly, the design principle of the XAI system or mechanism should cover the\nfollowing factors such as reliability, resilience, the verifiability of its\noutputs, and the transparency of its algorithm. However, research in XAI for\ngenerative models remains relatively scarce, with little exploration into how\nsuch methods can effectively meet these criteria in that domain. In this work,\nwe propose PXGen, a post-hoc explainable method for generative models. Given a\nmodel that needs to be explained, PXGen prepares two materials for the\nexplanation, the Anchor set and intrinsic \u0026 extrinsic criteria. Those materials\nare customizable by users according to their purpose and requirements. Via the\ncalculation of each criterion, each anchor has a set of feature values and\nPXGen provides examplebased explanation methods according to the feature values\namong all the anchors and illustrated and visualized to the users via tractable\nalgorithms such as k-dispersion or k-center.", + "arxiv_id": "2501.11827v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on Explainable AI (XAI) and responsible development of AI technologies, which aligns with 'responsible AI application or AI ethics', a criterion for rejection. Additionally, it lacks clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI as required by the selection criteria." + }, + { + "paper": { + "title": "A Survey on Memory-Efficient Large-Scale Model Training in AI for\n Science", + "abstract": "Scientific research faces high costs and inefficiencies with traditional\nmethods, but the rise of deep learning and large language models (LLMs) offers\ninnovative solutions. This survey reviews LLM applications across scientific\nfields such as biology, medicine, chemistry, and meteorology, underscoring\ntheir role in advancing research. However, the continuous expansion of model\nsize has led to significant memory demands, hindering further development and\napplication of LLMs for science. To address this, we review memory-efficient\ntraining techniques for LLMs based on the transformer architecture, including\ndistributed training, mixed precision training, and gradient checkpointing.\nUsing AlphaFold 2 as an example, we demonstrate how tailored memory\noptimization methods can reduce storage needs while preserving prediction\naccuracy. We also discuss the challenges of memory optimization in practice and\npotential future directions, hoping to provide valuable insights for\nresearchers and engineers.", + "arxiv_id": "2501.11847v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on scientific applications, with a notable emphasis on biology and medicine, which aligns with excluded topics (primarily medical applications of AI). While it touches on Large Language Models, the core application area (science, including medicine) does not meet the inclusion criteria." + }, + { + "paper": { + "title": "Coarse-to-Fine Lightweight Meta-Embedding for ID-Based Recommendation", + "abstract": "The state-of-the-art recommendation systems have shifted the attention to\nefficient recommendation, e.g., on-device recommendation, under memory\nconstraints. To this end, the existing methods either focused on the\nlightweight embeddings for both users and items, or involved on-device systems\nenjoying the compact embeddings to enhance reusability and reduces space\ncomplexity. However, they focus solely on the coarse granularity of embedding,\nwhile overlook the fine-grained semantic nuances, to adversarially downgrade\nthe efficacy of meta-embeddings in capturing the intricate relationship over\nboth user and item, consequently resulting into the suboptimal recommendations.\nIn this paper, we aim to study how the meta-embedding can efficiently learn\nvaried grained semantics, together with how the fine-grained meta-embedding can\nstrengthen the representation of coarse-grained meta-embedding. To answer these\nquestions, we develop a novel graph neural networks (GNNs) based recommender\nwhere each user and item serves as the node, linked directly to coarse-grained\nvirtual nodes and indirectly to fine-grained virtual nodes, ensuring different\ngrained semantic learning, while disclosing: 1) In contrast to coarse-grained\nsemantics, fine-grained semantics are well captured through sparse\nmeta-embeddings, which adaptively 2) balance the embedding uniqueness and\nmemory constraint. Additionally, the initialization method come up upon\nSparsePCA, along with a soft thresholding activation function to render the\nsparseness of the meta-embeddings. We propose a weight bridging update strategy\nthat focuses on matching each coarse-grained meta-embedding with several\nfine-grained meta-embeddings based on the users/items' semantics. Extensive\nexperiments substantiate our method's superiority over existing baselines. Our\ncode is available at https://github.com/htyjers/C2F-MetaEmbed.", + "arxiv_id": "2501.11870v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper does not meet the primary criteria related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. It focuses on recommendation systems using graph neural networks (GNNs) and meta-embeddings, which falls outside the specified areas of interest.\"\n}\n```" + }, + { + "paper": { + "title": "Community-Aware Temporal Walks: Parameter-Free Representation Learning\n on Continuous-Time Dynamic Graphs", + "abstract": "Dynamic graph representation learning plays a crucial role in understanding\nevolving behaviors. However, existing methods often struggle with flexibility,\nadaptability, and the preservation of temporal and structural dynamics. To\naddress these issues, we propose Community-aware Temporal Walks (CTWalks), a\nnovel framework for representation learning on continuous-time dynamic graphs.\nCTWalks integrates three key components: a community-based parameter-free\ntemporal walk sampling mechanism, an anonymization strategy enriched with\ncommunity labels, and an encoding process that leverages continuous temporal\ndynamics modeled via ordinary differential equations (ODEs). This design\nenables precise modeling of both intra- and inter-community interactions,\noffering a fine-grained representation of evolving temporal patterns in\ncontinuous-time dynamic graphs. CTWalks theoretically overcomes locality bias\nin walks and establishes its connection to matrix factorization. Experiments on\nbenchmark datasets demonstrate that CTWalks outperforms established methods in\ntemporal link prediction tasks, achieving higher accuracy while maintaining\nrobustness.", + "arxiv_id": "2501.11880v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the primary criteria for Large Language Models (LLMs) applications, focusing instead on dynamic graph representation learning, temporal walk sampling, and ordinary differential equations (ODEs), with no apparent connection to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI." + }, + { + "paper": { + "title": "LuxVeri at GenAI Detection Task 1: Inverse Perplexity Weighted Ensemble\n for Robust Detection of AI-Generated Text across English and Multilingual\n Contexts", + "abstract": "This paper presents a system developed for Task 1 of the COLING 2025 Workshop\non Detecting AI-Generated Content, focusing on the binary classification of\nmachine-generated versus human-written text. Our approach utilizes an ensemble\nof models, with weights assigned according to each model's inverse perplexity,\nto enhance classification accuracy. For the English text detection task, we\ncombined RoBERTa-base, RoBERTa-base with the OpenAI detector, and\nBERT-base-cased, achieving a Macro F1-score of 0.7458, which ranked us 12th out\nof 35 teams. We ensembled RemBERT, XLM-RoBERTa-base, and\nBERT-base-multilingual-case for the multilingual text detection task, employing\nthe same inverse perplexity weighting technique. This resulted in a Macro\nF1-score of 0.7513, positioning us 4th out of 25 teams. Our results demonstrate\nthe effectiveness of inverse perplexity weighting in improving the robustness\nof machine-generated text detection across both monolingual and multilingual\nsettings, highlighting the potential of ensemble methods for this challenging\ntask.", + "arxiv_id": "2501.11914v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on detecting AI-generated text, which does not explicitly align with the specified practical applications of Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not clearly meet the criteria for real-world applications, methodology, or novelty in the context of LLMs as outlined." + }, + { + "paper": { + "title": "LuxVeri at GenAI Detection Task 3: Cross-Domain Detection of\n AI-Generated Text Using Inverse Perplexity-Weighted Ensemble of Fine-Tuned\n Transformer Models", + "abstract": "This paper presents our approach for Task 3 of the GenAI content detection\nworkshop at COLING-2025, focusing on Cross-Domain Machine-Generated Text (MGT)\nDetection. We propose an ensemble of fine-tuned transformer models, enhanced by\ninverse perplexity weighting, to improve classification accuracy across diverse\ntext domains. For Subtask A (Non-Adversarial MGT Detection), we combined a\nfine-tuned RoBERTa-base model with an OpenAI detector-integrated RoBERTa-base\nmodel, achieving an aggregate TPR score of 0.826, ranking 10th out of 23\ndetectors. In Subtask B (Adversarial MGT Detection), our fine-tuned\nRoBERTa-base model achieved a TPR score of 0.801, securing 8th out of 22\ndetectors. Our results demonstrate the effectiveness of inverse\nperplexity-based weighting for enhancing generalization and performance in both\nnon-adversarial and adversarial MGT detection, highlighting the potential for\ntransformer models in cross-domain AI-generated content detection.", + "arxiv_id": "2501.11918v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on AI-generated text detection, which does not explicitly align with the practical applications of Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. While it meets some criteria (e.g., experimental results, comparison with state-of-the-art), its core application does not satisfy the mandatory criteria for practical applications of LLMs as specified." + }, + { + "paper": { + "title": "A Lightweight and Interpretable Deepfakes Detection Framework", + "abstract": "The recent realistic creation and dissemination of so-called deepfakes poses\na serious threat to social life, civil rest, and law. Celebrity defaming,\nelection manipulation, and deepfakes as evidence in court of law are few\npotential consequences of deepfakes. The availability of open source trained\nmodels based on modern frameworks such as PyTorch or TensorFlow, video\nmanipulations Apps such as FaceApp and REFACE, and economical computing\ninfrastructure has easen the creation of deepfakes. Most of the existing\ndetectors focus on detecting either face-swap, lip-sync, or puppet master\ndeepfakes, but a unified framework to detect all three types of deepfakes is\nhardly explored. This paper presents a unified framework that exploits the\npower of proposed feature fusion of hybrid facial landmarks and our novel heart\nrate features for detection of all types of deepfakes. We propose novel heart\nrate features and fused them with the facial landmark features to better\nextract the facial artifacts of fake videos and natural variations available in\nthe original videos. We used these features to train a light-weight XGBoost to\nclassify between the deepfake and bonafide videos. We evaluated the performance\nof our framework on the world leaders dataset (WLDR) that contains all types of\ndeepfakes. Experimental results illustrate that the proposed framework offers\nsuperior detection performance over the comparative deepfakes detection\nmethods. Performance comparison of our framework against the LSTM-FCN, a\ncandidate of deep learning model, shows that proposed model achieves similar\nresults, however, it is more interpretable.", + "arxiv_id": "2501.11927v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing (deepfakes detection), which is one of the criteria for rejection. Additionally, it touches on social implications (e.g., social life, civil unrest, law), another area for rejection. The paper does not appear to involve Large Language Models (LLMs) or related key areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI." + }, + { + "paper": { + "title": "Webvs. LLMs: An Empirical Study of Learning Behaviors of CS2 Students", + "abstract": "LLMs such as ChatGPT have been widely adopted by students in higher education\nas tools for learning programming and related concepts. However, it remains\nunclear how effective students are and what strategies students use while\nlearning with LLMs. Since the majority of students' experiences in online\nself-learning have come through using search engines such as Google, evaluating\nAI tools in this context can help us address these gaps. In this mixed methods\nresearch, we conducted an exploratory within-subjects study to understand how\nCS2 students learn programming concepts using both LLMs as well as traditional\nonline methods such as educational websites and videos to examine how students\napproach learning within and across both scenarios. We discovered that students\nfound it easier to learn a more difficult concept using traditional methods\nthan using ChatGPT. We also found that students ask fewer follow-ups and use\nmore keyword-based queries for search engines while their prompts to LLMs tend\nto explicitly ask for information.", + "arxiv_id": "2501.11935v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on educational applications of LLMs, which, while not explicitly listed for rejection, does not align with the preferred areas (knowledge graphs, RAG, agentic AI). It also lacks clear connections to practical real-world applications beyond educational settings, and its main findings relate to learning behaviors rather than showcasing innovative LLM applications, improvements, or state-of-the-art comparisons." + }, + { + "paper": { + "title": "MeshONet: A Generalizable and Efficient Operator Learning Method for\n Structured Mesh Generation", + "abstract": "Mesh generation plays a crucial role in scientific computing. Traditional\nmesh generation methods, such as TFI and PDE-based methods, often struggle to\nachieve a balance between efficiency and mesh quality. To address this\nchallenge, physics-informed intelligent learning methods have recently emerged,\nsignificantly improving generation efficiency while maintaining high mesh\nquality. However, physics-informed methods fail to generalize when applied to\npreviously unseen geometries, as even small changes in the boundary shape\nnecessitate burdensome retraining to adapt to new geometric variations. In this\npaper, we introduce MeshONet, the first generalizable intelligent learning\nmethod for structured mesh generation. The method transforms the mesh\ngeneration task into an operator learning problem with multiple input and\nsolution functions. To effectively overcome the multivariable mapping\nrestriction of operator learning methods, we propose a dual-branch,\nshared-trunk architecture to approximate the mapping between function spaces\nbased on input-output pairs. Experimental results show that MeshONet achieves a\nspeedup of up to four orders of magnitude in generation efficiency over\ntraditional methods. It also enables generalization to different geometries\nwithout retraining, greatly enhancing the practicality of intelligent methods.", + "arxiv_id": "2501.11937v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria, focusing on mesh generation for scientific computing, which is unrelated to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not address real-world applications or challenges specific to LLMs." + }, + { + "paper": { + "title": "Survey on Hand Gesture Recognition from Visual Input", + "abstract": "Hand gesture recognition has become an important research area, driven by the\ngrowing demand for human-computer interaction in fields such as sign language\nrecognition, virtual and augmented reality, and robotics. Despite the rapid\ngrowth of the field, there are few surveys that comprehensively cover recent\nresearch developments, available solutions, and benchmark datasets. This survey\naddresses this gap by examining the latest advancements in hand gesture and 3D\nhand pose recognition from various types of camera input data including RGB\nimages, depth images, and videos from monocular or multiview cameras, examining\nthe differing methodological requirements of each approach. Furthermore, an\noverview of widely used datasets is provided, detailing their main\ncharacteristics and application domains. Finally, open challenges such as\nachieving robust recognition in real-world environments, handling occlusions,\nensuring generalization across diverse users, and addressing computational\nefficiency for real-time applications are highlighted to guide future research\ndirections. By synthesizing the objectives, methodologies, and applications of\nrecent studies, this survey offers valuable insights into current trends,\nchallenges, and opportunities for future research in human hand gesture\nrecognition.", + "arxiv_id": "2501.11992v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria, as it focuses on hand gesture recognition from visual input, which is unrelated to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and also falls under the rejection category of primarily focusing on video processing." + }, + { + "paper": { + "title": "Full Proportional Justified Representation", + "abstract": "In multiwinner approval voting, forming a committee that proportionally\nrepresents voters' approval ballots is an essential task. The notion of\njustified representation (JR) demands that any large \"cohesive\" group of voters\nshould be proportionally \"represented\". The \"cohesiveness\" is defined in\ndifferent ways; two common ways are the following: (C1) demands that the group\nunanimously approves a set of candidates proportional to its size, while (C2)\nrequires each member to approve at least a fixed fraction of such a set.\nSimilarly, \"representation\" have been considered in different ways: (R1) the\ncoalition's collective utility from the winning set exceeds that of any\nproportionally sized alternative, and (R2) for any proportionally sized\nalternative, at least one member of the coalition derives less utility from it\nthan from the winning set.\n Three of the four possible combinations have been extensively studied:\n(C1)-(R1) defines Proportional Justified Representation (PJR), (C1)-(R2)\ndefines Extended Justified Representation (EJR), (C2)-(R2) defines Full\nJustified Representation (FJR). All three have merits, but also drawbacks. PJR\nis the weakest notion, and perhaps not sufficiently demanding; EJR may not be\ncompatible with perfect representation; and it is open whether a committee\nsatisfying FJR can be found efficiently.\n We study the combination (C2)-(R1), which we call Full Proportional Justified\nRepresentation (FPJR). We investigate FPJR's properties and find that it shares\nPJR's advantages over EJR: several proportionality axioms (e.g. priceability,\nperfect representation) imply FPJR and PJR but not EJR. We also find that\nefficient rules like the greedy Monroe rule and the method of equal shares\nsatisfy FPJR, matching a key advantage of EJR over FJR. However, the\nProportional Approval Voting (PAV) rule may violate FPJR, so neither of EJR and\nFPJR implies the other.", + "arxiv_id": "2501.12015v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria, focusing on voting systems and proportional representation rather than Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, and lacks relevance to the specified areas of interest." + }, + { + "paper": { + "title": "Adaptive Class Learning to Screen Diabetic Disorders in Fundus Images of\n Eye", + "abstract": "The prevalence of ocular illnesses is growing globally, presenting a\nsubstantial public health challenge. Early detection and timely intervention\nare crucial for averting visual impairment and enhancing patient prognosis.\nThis research introduces a new framework called Class Extension with Limited\nData (CELD) to train a classifier to categorize retinal fundus images. The\nclassifier is initially trained to identify relevant features concerning\nHealthy and Diabetic Retinopathy (DR) classes and later fine-tuned to adapt to\nthe task of classifying the input images into three classes: Healthy, DR, and\nGlaucoma. This strategy allows the model to gradually enhance its\nclassification capabilities, which is beneficial in situations where there are\nonly a limited number of labeled datasets available. Perturbation methods are\nalso used to identify the input image characteristics responsible for\ninfluencing the models decision-making process. We achieve an overall accuracy\nof 91% on publicly available datasets.", + "arxiv_id": "2501.12048v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (diabetic disorders in eye images), which is explicitly listed as a rejection criterion." + }, + { + "paper": { + "title": "Teacher Encoder-Student Decoder Denoising Guided Segmentation Network\n for Anomaly Detection", + "abstract": "Visual anomaly detection is a highly challenging task, often categorized as a\none-class classification and segmentation problem. Recent studies have\ndemonstrated that the student-teacher (S-T) framework effectively addresses\nthis challenge. However, most S-T frameworks rely solely on pre-trained teacher\nnetworks to guide student networks in learning multi-scale similar features,\noverlooking the potential of the student networks to enhance learning through\nmulti-scale feature fusion. In this study, we propose a novel model named\nPFADSeg, which integrates a pre-trained teacher network, a denoising student\nnetwork with multi-scale feature fusion, and a guided anomaly segmentation\nnetwork into a unified framework. By adopting a unique teacher-encoder and\nstudent-decoder denoising mode, the model improves the student network's\nability to learn from teacher network features. Furthermore, an adaptive\nfeature fusion mechanism is introduced to train a self-supervised segmentation\nnetwork that synthesizes anomaly masks autonomously, significantly increasing\ndetection performance. Evaluated on the MVTec AD dataset, PFADSeg achieves\nstate-of-the-art results with an image-level AUC of 98.9%, a pixel-level mean\nprecision of 76.4%, and an instance-level mean precision of 78.7%.", + "arxiv_id": "2501.12104v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the required criteria as it focuses on visual anomaly detection using a student-teacher framework, without any indication of involving Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest." + }, + { + "paper": { + "title": "Can open source large language models be used for tumor documentation in\n Germany? -- An evaluation on urological doctors' notes", + "abstract": "Tumor documentation in Germany is largely done manually, requiring reading\npatient records and entering data into structured databases. Large language\nmodels (LLMs) could potentially enhance this process by improving efficiency\nand reliability. This evaluation tests eleven different open source LLMs with\nsizes ranging from 1-70 billion model parameters on three basic tasks of the\ntumor documentation process: identifying tumor diagnoses, assigning ICD-10\ncodes, and extracting the date of first diagnosis. For evaluating the LLMs on\nthese tasks, a dataset of annotated text snippets based on anonymized doctors'\nnotes from urology was prepared. Different prompting strategies were used to\ninvestigate the effect of the number of examples in few-shot prompting and to\nexplore the capabilities of the LLMs in general. The models Llama 3.1 8B,\nMistral 7B, and Mistral NeMo 12 B performed comparably well in the tasks.\nModels with less extensive training data or having fewer than 7 billion\nparameters showed notably lower performance, while larger models did not\ndisplay performance gains. Examples from a different medical domain than\nurology could also improve the outcome in few-shot prompting, which\ndemonstrates the ability of LLMs to handle tasks needed for tumor\ndocumentation. Open source LLMs show a strong potential for automating tumor\ndocumentation. Models from 7-12 billion parameters could offer an optimal\nbalance between performance and resource efficiency. With tailored fine-tuning\nand well-designed prompting, these models might become important tools for\nclinical documentation in the future. The code for the evaluation is available\nfrom https://github.com/stefan-m-lenz/UroLlmEval. We also release the dataset\nas a new valuable resource that addresses the shortage of authentic and easily\naccessible benchmarks in German-language medical NLP.", + "arxiv_id": "2501.12106v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (tumor documentation in Germany), which is an exclusion criterion as per the evaluation guidelines." + }, + { + "paper": { + "title": "FedCLEAN: byzantine defense by CLustering Errors of Activation maps in\n Non-IID federated learning environments", + "abstract": "Federated Learning (FL) enables clients to collaboratively train a global\nmodel using their local datasets while reinforcing data privacy. However, FL is\nsusceptible to poisoning attacks. Existing defense mechanisms assume that\nclients' data are independent and identically distributed (IID), making them\nineffective in real-world applications where data are non-IID. This paper\npresents FedCLEAN, the first defense capable of filtering attackers' model\nupdates in a non-IID FL environment. The originality of FedCLEAN is twofold.\nFirst, it relies on a client confidence score derived from the reconstruction\nerrors of each client's model activation maps for a given trigger set, with\nreconstruction errors obtained by means of a Conditional Variational\nAutoencoder trained according to a novel server-side strategy. Second, we\npropose an ad-hoc trust propagation algorithm based on client scores, which\nallows building a cluster of benign clients while flagging potential attackers.\nExperimental results on the datasets MNIST and FashionMNIST demonstrate the\nrobustness of FedCLEAN against Byzantine attackers in non-IID scenarios and a\nclose-to-zero benign client misclassification rate, even in the absence of an\nattack.", + "arxiv_id": "2501.12123v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the primary criteria as it focuses on federated learning (FL) and defense mechanisms against poisoning attacks, without apparent involvement or application of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest." + }, + { + "paper": { + "title": "FuocChuVIP123 at CoMeDi Shared Task: Disagreement Ranking with\n XLM-Roberta Sentence Embeddings and Deep Neural Regression", + "abstract": "This paper presents results of our system for CoMeDi Shared Task, focusing on\nSubtask 2: Disagreement Ranking. Our system leverages sentence embeddings\ngenerated by the paraphrase-xlm-r-multilingual-v1 model, combined with a deep\nneural regression model incorporating batch normalization and dropout for\nimproved generalization. By predicting the mean of pairwise judgment\ndifferences between annotators, our method explicitly targets disagreement\nranking, diverging from traditional \"gold label\" aggregation approaches. We\noptimized our system with a customized architecture and training procedure,\nachieving competitive performance in Spearman correlation against mean\ndisagreement labels. Our results highlight the importance of robust embeddings,\neffective model architecture, and careful handling of judgment differences for\nranking disagreement in multilingual contexts. These findings provide insights\ninto the use of contextualized representations for ordinal judgment tasks and\nopen avenues for further refinement of disagreement prediction models.", + "arxiv_id": "2501.12336v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper does not meet the primary criteria for Practical Applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. It also doesn't clearly align with other key criteria, focusing instead on disagreement ranking with sentence embeddings, which is not explicitly mentioned as a priority area in the evaluation guidelines.\"\n}\n```" + }, + { + "paper": { + "title": "Video Depth Anything: Consistent Depth Estimation for Super-Long Videos", + "abstract": "Depth Anything has achieved remarkable success in monocular depth estimation\nwith strong generalization ability. However, it suffers from temporal\ninconsistency in videos, hindering its practical applications. Various methods\nhave been proposed to alleviate this issue by leveraging video generation\nmodels or introducing priors from optical flow and camera poses. Nonetheless,\nthese methods are only applicable to short videos (\u003c 10 seconds) and require a\ntrade-off between quality and computational efficiency. We propose Video Depth\nAnything for high-quality, consistent depth estimation in super-long videos\n(over several minutes) without sacrificing efficiency. We base our model on\nDepth Anything V2 and replace its head with an efficient spatial-temporal head.\nWe design a straightforward yet effective temporal consistency loss by\nconstraining the temporal depth gradient, eliminating the need for additional\ngeometric priors. The model is trained on a joint dataset of video depth and\nunlabeled images, similar to Depth Anything V2. Moreover, a novel\nkey-frame-based strategy is developed for long video inference. Experiments\nshow that our model can be applied to arbitrarily long videos without\ncompromising quality, consistency, or generalization ability. Comprehensive\nevaluations on multiple video benchmarks demonstrate that our approach sets a\nnew state-of-the-art in zero-shot video depth estimation. We offer models of\ndifferent scales to support a range of scenarios, with our smallest model\ncapable of real-time performance at 30 FPS.", + "arxiv_id": "2501.12375v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing, specifically depth estimation in super-long videos, which is explicitly listed as a rejection criterion." + }, + { + "paper": { + "title": "Leveraging GANs For Active Appearance Models Optimized Model Fitting", + "abstract": "Generative Adversarial Networks (GANs) have gained prominence in refining\nmodel fitting tasks in computer vision, particularly in domains involving\ndeformable models like Active Appearance Models (AAMs). This paper explores the\nintegration of GANs to enhance the AAM fitting process, addressing challenges\nin optimizing nonlinear parameters associated with appearance and shape\nvariations. By leveraging GANs' adversarial training framework, the aim is to\nminimize fitting errors and improve convergence rates. Achieving robust\nperformance even in cases with high appearance variability and occlusions. Our\napproach demonstrates significant improvements in accuracy and computational\nefficiency compared to traditional optimization techniques, thus establishing\nGANs as a potent tool for advanced image model fitting.", + "arxiv_id": "2501.11218v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on computer vision (image model fitting) using GANs, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the core selection criteria." + }, + { + "paper": { + "title": "WSSM: Geographic-enhanced hierarchical state-space model for global\n station weather forecast", + "abstract": "Global Station Weather Forecasting (GSWF), a prominent meteorological\nresearch area, is pivotal in providing timely localized weather predictions.\nDespite the progress existing models have made in the overall accuracy of the\nGSWF, executing high-precision extreme event prediction still presents a\nsubstantial challenge. The recent emergence of state-space models, with their\nability to efficiently capture continuous-time dynamics and latent states,\noffer potential solutions. However, early investigations indicated that Mamba\nunderperforms in the context of GSWF, suggesting further adaptation and\noptimization. To tackle this problem, in this paper, we introduce Weather\nState-space Model (WSSM), a novel Mamba-based approach tailored for GSWF.\nGeographical knowledge is integrated in addition to the widely-used positional\nencoding to represent the absolute special-temporal position. The multi-scale\ntime-frequency features are synthesized from coarse to fine to model the\nseasonal to extreme weather dynamic. Our method effectively improves the\noverall prediction accuracy and addresses the challenge of forecasting extreme\nweather events. The state-of-the-art results obtained on the Weather-5K subset\nunderscore the efficacy of the WSSM", + "arxiv_id": "2501.11238v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the required criteria as it primarily focuses on weather forecasting, a domain unrelated to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, with no mention of LLMs or relevant AI technologies." + }, + { + "paper": { + "title": "Spatiotemporal Air Quality Mapping in Urban Areas Using Sparse Sensor\n Data, Satellite Imagery, Meteorological Factors, and Spatial Features", + "abstract": "Monitoring air pollution is crucial for protecting human health from exposure\nto harmful substances. Traditional methods of air quality monitoring, such as\nground-based sensors and satellite-based remote sensing, face limitations due\nto high deployment costs, sparse sensor coverage, and environmental\ninterferences. To address these challenges, this paper proposes a framework for\nhigh-resolution spatiotemporal Air Quality Index (AQI) mapping using sparse\nsensor data, satellite imagery, and various spatiotemporal factors. By\nleveraging Graph Neural Networks (GNNs), we estimate AQI values at unmonitored\nlocations based on both spatial and temporal dependencies. The framework\nincorporates a wide range of environmental features, including meteorological\ndata, road networks, points of interest (PoIs), population density, and urban\ngreen spaces, which enhance prediction accuracy. We illustrate the use of our\napproach through a case study in Lahore, Pakistan, where multi-resolution data\nis used to generate the air quality index map at a fine spatiotemporal scale.", + "arxiv_id": "2501.11270v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the primary criteria as it does not focus on Large Language Models (LLMs) or related areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. Instead, it leverages Graph Neural Networks (GNNs) for spatiotemporal air quality mapping, which does not align with the specified topics of interest." + }, + { + "paper": { + "title": "A Machine Learning Framework for Handling Unreliable Absence Label and\n Class Imbalance for Marine Stinger Beaching Prediction", + "abstract": "Bluebottles (\\textit{Physalia} spp.) are marine stingers resembling\njellyfish, whose presence on Australian beaches poses a significant public risk\ndue to their venomous nature. Understanding the environmental factors driving\nbluebottles ashore is crucial for mitigating their impact, and machine learning\ntools are to date relatively unexplored. We use bluebottle marine stinger\npresence/absence data from beaches in Eastern Sydney, Australia, and compare\nmachine learning models (Multilayer Perceptron, Random Forest, and XGBoost) to\nidentify factors influencing their presence. We address challenges such as\nclass imbalance, class overlap, and unreliable absence data by employing data\naugmentation techniques, including the Synthetic Minority Oversampling\nTechnique (SMOTE), Random Undersampling, and Synthetic Negative Approach that\nexcludes the negative class. Our results show that SMOTE failed to resolve\nclass overlap, but the presence-focused approach effectively handled imbalance,\nclass overlap, and ambiguous absence data. The data attributes such as the wind\ndirection, which is a circular variable, emerged as a key factor influencing\nbluebottle presence, confirming previous inference studies. However, in the\nabsence of population dynamics, biological behaviours, and life cycles, the\nbest predictive model appears to be Random Forests combined with Synthetic\nNegative Approach. This research contributes to mitigating the risks posed by\nbluebottles to beachgoers and provides insights into handling class overlap and\nunreliable negative class in environmental modelling.", + "arxiv_id": "2501.11293v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria (1-3) as it does not focus on Large Language Models (LLMs), lacks experimental results with quantitative metrics related to LLMs, and does not compare with state-of-the-art LLM techniques. Additionally, its primary focus on environmental modelling for marine stinger beaching prediction falls outside the specified rejection criteria, but more importantly, does not align with the paper selection criteria." + }, + { + "paper": { + "title": "On the Dimension of Pullback Attractors in Recurrent Neural Networks", + "abstract": "Recurrent Neural Networks (RNNs) are high-dimensional state space models\ncapable of learning functions on sequence data. Recently, it has been\nconjectured that reservoir computers, a particular class of RNNs, trained on\nobservations of a dynamical systems can be interpreted as embeddings. This\nresult has been established for the case of linear reservoir systems. In this\nwork, we use a nonautonomous dynamical systems approach to establish an upper\nbound for the fractal dimension of the subset of reservoir state space\napproximated during training and prediction phase. We prove that when the input\nsequences comes from an Nin-dimensional invertible dynamical system, the\nfractal dimension of this set is bounded above by Nin. The result obtained here\nare useful in dimensionality reduction of computation in RNNs as well as\nestimating fractal dimensions of dynamical systems from limited observations of\ntheir time series. It is also a step towards understanding embedding properties\nof reservoir computers.", + "arxiv_id": "2501.11357v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria, specifically lacking focus on Large Language Models (LLMs), practical applications in knowledge graphs, RAG, or agentic AI, and instead centers on Recurrent Neural Networks (RNNs) and dynamical systems, which are outside the specified scope." + }, + { + "paper": { + "title": "Investigation of Whisper ASR Hallucinations Induced by Non-Speech Audio", + "abstract": "Hallucinations of deep neural models are amongst key challenges in automatic\nspeech recognition (ASR). In this paper, we investigate hallucinations of the\nWhisper ASR model induced by non-speech audio segments present during\ninference. By inducting hallucinations with various types of sounds, we show\nthat there exists a set of hallucinations that appear frequently. We then study\nhallucinations caused by the augmentation of speech with such sounds. Finally,\nwe describe the creation of a bag of hallucinations (BoH) that allows to remove\nthe effect of hallucinations through the post-processing of text\ntranscriptions. The results of our experiments show that such post-processing\nis capable of reducing word error rate (WER) and acts as a good safeguard\nagainst problematic hallucinations.", + "arxiv_id": "2501.11378v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on Automatic Speech Recognition (ASR) and audio processing, with no clear connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the fundamental criteria for practical applications of LLMs." + }, + { + "paper": { + "title": "A Truly Sparse and General Implementation of Gradient-Based Synaptic\n Plasticity", + "abstract": "Online synaptic plasticity rules derived from gradient descent achieve high\naccuracy on a wide range of practical tasks. However, their software\nimplementation often requires tediously hand-derived gradients or using\ngradient backpropagation which sacrifices the online capability of the rules.\nIn this work, we present a custom automatic differentiation (AD) pipeline for\nsparse and online implementation of gradient-based synaptic plasticity rules\nthat generalizes to arbitrary neuron models. Our work combines the programming\nease of backpropagation-type methods for forward AD while being\nmemory-efficient. To achieve this, we exploit the advantageous compute and\nmemory scaling of online synaptic plasticity by providing an inherently sparse\nimplementation of AD where expensive tensor contractions are replaced with\nsimple element-wise multiplications if the tensors are diagonal. Gradient-based\nsynaptic plasticity rules such as eligibility propagation (e-prop) have exactly\nthis property and thus profit immensely from this feature. We demonstrate the\nalignment of our gradients with respect to gradient backpropagation on an\nsynthetic task where e-prop gradients are exact, as well as audio speech\nclassification benchmarks. We demonstrate how memory utilization scales with\nnetwork size without dependence on the sequence length, as expected from\nforward AD methods.", + "arxiv_id": "2501.11407v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the mandatory criteria (1-3) as it focuses on synaptic plasticity and neural networks rather than Large Language Models (LLMs), and does not discuss practical applications, experimental results, or comparisons to state-of-the-art in the context of LLMs. Additionally, it does not align with any of the preferred criteria (4-5) related to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI." + }, + { + "paper": { + "title": "Generalization and Informativeness of Weighted Conformal Risk Control\n Under Covariate Shift", + "abstract": "Predictive models are often required to produce reliable predictions under\nstatistical conditions that are not matched to the training data. A common type\nof training-testing mismatch is covariate shift, where the conditional\ndistribution of the target variable given the input features remains fixed,\nwhile the marginal distribution of the inputs changes. Weighted conformal risk\ncontrol (W-CRC) uses data collected during the training phase to convert point\npredictions into prediction sets with valid risk guarantees at test time\ndespite the presence of a covariate shift. However, while W-CRC provides\nstatistical reliability, its efficiency -- measured by the size of the\nprediction sets -- can only be assessed at test time. In this work, we relate\nthe generalization properties of the base predictor to the efficiency of W-CRC\nunder covariate shifts. Specifically, we derive a bound on the inefficiency of\nthe W-CRC predictor that depends on algorithmic hyperparameters and\ntask-specific quantities available at training time. This bound offers insights\non relationships between the informativeness of the prediction sets, the extent\nof the covariate shift, and the size of the calibration and training sets.\nExperiments on fingerprinting-based localization validate the theoretical\nresults.", + "arxiv_id": "2501.11413v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria, focusing on predictive model reliability under covariate shift without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and lacking clear connections to practical applications of LLMs as specified in the criteria." + }, + { + "paper": { + "title": "Enhancing Coronary Artery Calcium Scoring via Multi-Organ Segmentation\n on Non-Contrast Cardiac Computed Tomography", + "abstract": "Despite coronary artery calcium scoring being considered a largely solved\nproblem within the realm of medical artificial intelligence, this paper argues\nthat significant improvements can still be made. By shifting the focus from\npathology detection to a deeper understanding of anatomy, the novel algorithm\nproposed in the paper both achieves high accuracy in coronary artery calcium\nscoring and offers enhanced interpretability of the results. This approach not\nonly aids in the precise quantification of calcifications in coronary arteries,\nbut also provides valuable insights into the underlying anatomical structures.\nThrough this anatomically-informed methodology, the paper shows how a nuanced\nunderstanding of the heart's anatomy can lead to more accurate and\ninterpretable results in the field of cardiovascular health. We demonstrate the\nsuperior accuracy of the proposed method by evaluating it on an open-source\nmulti-vendor dataset, where we obtain results at the inter-observer level,\nsurpassing the current state of the art. Finally, the qualitative analyses show\nthe practical value of the algorithm in such tasks as labeling coronary artery\ncalcifications, identifying aortic calcifications, and filtering out false\npositive detections due to noise.", + "arxiv_id": "2501.11428v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI, specifically cardiovascular health, which is explicitly mentioned as a rejection criterion." + }, + { + "paper": { + "title": "Decomposing Interventional Causality into Synergistic, Redundant, and\n Unique Components", + "abstract": "We introduce a novel framework for decomposing interventional causal effects\ninto synergistic, redundant, and unique components, building on the intuition\nof Partial Information Decomposition (PID) and the principle of M\\\"obius\ninversion. While recent work has explored a similar decomposition of an\nobservational measure, we argue that a proper causal decomposition must be\ninterventional in nature. We develop a mathematical approach that\nsystematically quantifies how causal power is distributed among variables in a\nsystem, using a recently derived closed-form expression for the M\\\"obius\nfunction of the redundancy lattice. The formalism is then illustrated by\ndecomposing the causal power in logic gates, cellular automata, and chemical\nreaction networks. Our results reveal how the distribution of causal power can\nbe context- and parameter-dependent. This decomposition provides new insights\ninto complex systems by revealing how causal influences are shared and combined\namong multiple variables, with potential applications ranging from attribution\nof responsibility in legal or AI systems, to the analysis of biological\nnetworks or climate models.", + "arxiv_id": "2501.11447v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the mandatory criteria (1-3) as it lacks focus on Large Language Models (LLMs), experimental results with quantitative metrics, and comparison with state-of-the-art techniques. Additionally, its primary focus on causal decomposition in complex systems, with potential applications in legal systems, does not align with the criteria, and its potential application in AI systems is too broad and not specifically related to LLMs or the specified areas." + }, + { + "paper": { + "title": "Generative AI and Large Language Models in Language Preservation:\n Opportunities and Challenges", + "abstract": "Generative AI and large-scale language models (LLM) have emerged as powerful\ntools in language preservation, particularly for near-native and endangered\nlanguages. With the increasing reliance on technology for communication,\neducation, and cultural documentation, new opportunities have emerged to\nmitigate the dramatic decline of linguistic diversity worldwide. This paper\nexamines the role of generative AIs and LLMs in preserving endangered\nlanguages, highlighting the risks and challenges associated with their use. We\nanalyze the underlying technologies driving these models, including natural\nlanguage processing (NLP) and deep learning, and explore several cases where\nthese technologies have been applied to low-resource languages. Additionally,\nwe discuss ethical considerations, data scarcity issues, and technical\nchallenges while proposing solutions to enhance AI-driven language\npreservation.", + "arxiv_id": "2501.11496v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on social applications of AI in regard to linguistic diversity and cultural preservation, which aligns with rejected criteria related to social applications of AI, warranting rejection despite potential novelty in language preservation." + }, + { + "paper": { + "title": "Dialect2SQL: A Novel Text-to-SQL Dataset for Arabic Dialects with a\n Focus on Moroccan Darija", + "abstract": "The task of converting natural language questions (NLQs) into executable SQL\nqueries, known as text-to-SQL, has gained significant interest in recent years,\nas it enables non-technical users to interact with relational databases. Many\nbenchmarks, such as SPIDER and WikiSQL, have contributed to the development of\nnew models and the evaluation of their performance. In addition, other\ndatasets, like SEDE and BIRD, have introduced more challenges and complexities\nto better map real-world scenarios. However, these datasets primarily focus on\nhigh-resource languages such as English and Chinese. In this work, we introduce\nDialect2SQL, the first large-scale, cross-domain text-to-SQL dataset in an\nArabic dialect. It consists of 9,428 NLQ-SQL pairs across 69 databases in\nvarious domains. Along with SQL-related challenges such as long schemas, dirty\nvalues, and complex queries, our dataset also incorporates the complexities of\nthe Moroccan dialect, which is known for its diverse source languages, numerous\nborrowed words, and unique expressions. This demonstrates that our dataset will\nbe a valuable contribution to both the text-to-SQL community and the\ndevelopment of resources for low-resource languages.", + "arxiv_id": "2501.11498v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on developing a text-to-SQL dataset for an Arabic dialect, without explicit mention of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the mandatory criteria." + }, + { + "paper": { + "title": "Technical Report for the Forgotten-by-Design Project: Targeted\n Obfuscation for Machine Learning", + "abstract": "The right to privacy, enshrined in various human rights declarations, faces\nnew challenges in the age of artificial intelligence (AI). This paper explores\nthe concept of the Right to be Forgotten (RTBF) within AI systems, contrasting\nit with traditional data erasure methods. We introduce Forgotten by Design, a\nproactive approach to privacy preservation that integrates instance-specific\nobfuscation techniques during the AI model training process. Unlike machine\nunlearning, which modifies models post-training, our method prevents sensitive\ndata from being embedded in the first place. Using the LIRA membership\ninference attack, we identify vulnerable data points and propose defenses that\ncombine additive gradient noise and weighting schemes. Our experiments on the\nCIFAR-10 dataset demonstrate that our techniques reduce privacy risks by at\nleast an order of magnitude while maintaining model accuracy (at 95%\nsignificance). Additionally, we present visualization methods for the\nprivacy-utility trade-off, providing a clear framework for balancing privacy\nrisk and model accuracy. This work contributes to the development of\nprivacy-preserving AI systems that align with human cognitive processes of\nmotivated forgetting, offering a robust framework for safeguarding sensitive\ninformation and ensuring compliance with privacy regulations.", + "arxiv_id": "2501.11525v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on responsible AI application (privacy preservation and compliance with privacy regulations), which is explicitly listed as a rejection criterion, outweighing its marginal relevance to Large Language Models (LLMs) through its broader AI context." + }, + { + "paper": { + "title": "Training-free Ultra Small Model for Universal Sparse Reconstruction in\n Compressed Sensing", + "abstract": "Pre-trained large models attract widespread attention in recent years, but\nthey face challenges in applications that require high interpretability or have\nlimited resources, such as physical sensing, medical imaging, and\nbioinformatics. Compressed Sensing (CS) is a well-proved theory that drives\nmany recent breakthroughs in these applications. However, as a typical\nunder-determined linear system, CS suffers from excessively long sparse\nreconstruction times when using traditional iterative methods, particularly\nwith large-scale data. Current AI methods like deep unfolding fail to\nsubstitute them because pre-trained models exhibit poor generality beyond their\ntraining conditions and dataset distributions, or lack interpretability.\nInstead of following the big model fervor, this paper proposes ultra-small\nartificial neural models called coefficients learning (CL), enabling\ntraining-free and rapid sparse reconstruction while perfectly inheriting the\ngenerality and interpretability of traditional iterative methods, bringing new\nfeature of incorporating prior knowledges. In CL, a signal of length $n$ only\nneeds a minimal of $n$ trainable parameters. A case study model called CLOMP is\nimplemented for evaluation. Experiments are conducted on both synthetic and\nreal one-dimensional and two-dimensional signals, demonstrating significant\nimprovements in efficiency and accuracy. Compared to representative iterative\nmethods, CLOMP improves efficiency by 100 to 1000 folds for large-scale data.\nTest results on eight diverse image datasets indicate that CLOMP improves\nstructural similarity index by 292%, 98%, 45% for sampling rates of 0.1, 0.3,\n0.5, respectively. We believe this method can truly usher CS reconstruction\ninto the AI era, benefiting countless under-determined linear systems that rely\non sparse solution.", + "arxiv_id": "2501.11592v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on Compressed Sensing for applications in physical sensing, medical imaging, and bioinformatics, which aligns with excluded topics (medical applications of AI and not explicitly focusing on Large Language Models or related areas like knowledge graphs, RAG, or agentic AI)." + }, + { + "paper": { + "title": "Fairness Testing through Extreme Value Theory", + "abstract": "Data-driven software is increasingly being used as a critical component of\nautomated decision-support systems. Since this class of software learns its\nlogic from historical data, it can encode or amplify discriminatory practices.\nPrevious research on algorithmic fairness has focused on improving average-case\nfairness. On the other hand, fairness at the extreme ends of the spectrum,\nwhich often signifies lasting and impactful shifts in societal attitudes, has\nreceived significantly less emphasis.\n Leveraging the statistics of extreme value theory (EVT), we propose a novel\nfairness criterion called extreme counterfactual discrimination (ECD). This\ncriterion estimates the worst-case amounts of disadvantage in outcomes for\nindividuals solely based on their memberships in a protected group. Utilizing\ntools from search-based software engineering and generative AI, we present a\nrandomized algorithm that samples a statistically significant set of points\nfrom the tail of ML outcome distributions even if the input dataset lacks a\nsufficient number of relevant samples.\n We conducted several experiments on four ML models (deep neural networks,\nlogistic regression, and random forests) over 10 socially relevant tasks from\nthe literature on algorithmic fairness. First, we evaluate the generative AI\nmethods and find that they generate sufficient samples to infer valid EVT\ndistribution in 95% of cases. Remarkably, we found that the prevalent bias\nmitigators reduce the average-case discrimination but increase the worst-case\ndiscrimination significantly in 5% of cases. We also observed that even the\ntail-aware mitigation algorithm -- MiniMax-Fairness -- increased the worst-case\ndiscrimination in 30% of cases. We propose a novel ECD-based mitigator that\nimproves fairness in the tail in 90% of cases with no degradation of the\naverage-case discrimination.", + "arxiv_id": "2501.11597v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on fairness and bias in AI decision-making systems, which aligns with \"responsible AI application or AI ethics\", a criterion for rejection. It also lacks clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI as specified in the selection criteria." + }, + { + "paper": { + "title": "Noise-Agnostic Multitask Whisper Training for Reducing False Alarm\n Errors in Call-for-Help Detection", + "abstract": "Keyword spotting is often implemented by keyword classifier to the encoder in\nacoustic models, enabling the classification of predefined or open vocabulary\nkeywords. Although keyword spotting is a crucial task in various applications\nand can be extended to call-for-help detection in emergencies, however, the\nprevious method often suffers from scalability limitations due to retraining\nrequired to introduce new keywords or adapt to changing contexts. We explore a\nsimple yet effective approach that leverages off-the-shelf pretrained ASR\nmodels to address these challenges, especially in call-for-help detection\nscenarios. Furthermore, we observed a substantial increase in false alarms when\ndeploying call-for-help detection system in real-world scenarios due to noise\nintroduced by microphones or different environments. To address this, we\npropose a novel noise-agnostic multitask learning approach that integrates a\nnoise classification head into the ASR encoder. Our method enhances the model's\nrobustness to noisy environments, leading to a significant reduction in false\nalarms and improved overall call-for-help performance. Despite the added\ncomplexity of multitask learning, our approach is computationally efficient and\nprovides a promising solution for call-for-help detection in real-world\nscenarios.", + "arxiv_id": "2501.11631v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper primarily focuses on acoustic models, automatic speech recognition (ASR), and noise reduction in call-for-help detection, which does not align with the specified criteria emphasizing Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. Its applications and methodologies fall outside the designated scope.\"\n}\n```" + }, + { + "paper": { + "title": "Biomedical Knowledge Graph: A Survey of Domains, Tasks, and Real-World\n Applications", + "abstract": "Biomedical knowledge graphs (BKGs) have emerged as powerful tools for\norganizing and leveraging the vast and complex data found across the biomedical\nfield. Yet, current reviews of BKGs often limit their scope to specific domains\nor methods, overlooking the broader landscape and the rapid technological\nprogress reshaping it. In this survey, we address this gap by offering a\nsystematic review of BKGs from three core perspectives: domains, tasks, and\napplications. We begin by examining how BKGs are constructed from diverse data\nsources, including molecular interactions, pharmacological datasets, and\nclinical records. Next, we discuss the essential tasks enabled by BKGs,\nfocusing on knowledge management, retrieval, reasoning, and interpretation.\nFinally, we highlight real-world applications in precision medicine, drug\ndiscovery, and scientific research, illustrating the translational impact of\nBKGs across multiple sectors. By synthesizing these perspectives into a unified\nframework, this survey not only clarifies the current state of BKG research but\nalso establishes a foundation for future exploration, enabling both innovative\nmethodological advances and practical implementations.", + "arxiv_id": "2501.11632v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on biomedical applications, which is excluded according to the rejection criteria, specifically the focus on medical applications of AI." + }, + { + "paper": { + "title": "Transformer Vibration Forecasting for Advancing Rail Safety and\n Maintenance 4.0", + "abstract": "Maintaining railway axles is critical to preventing severe accidents and\nfinancial losses. The railway industry is increasingly interested in advanced\ncondition monitoring techniques to enhance safety and efficiency, moving beyond\ntraditional periodic inspections toward Maintenance 4.0.\n This study introduces a robust Deep Autoregressive solution that integrates\nseamlessly with existing systems to avert mechanical failures. Our approach\nsimulates and predicts vibration signals under various conditions and fault\nscenarios, improving dataset robustness for more effective detection systems.\nThese systems can alert maintenance needs, preventing accidents preemptively.\nWe use experimental vibration signals from accelerometers on train axles.\n Our primary contributions include a transformer model, ShaftFormer, designed\nfor processing time series data, and an alternative model incorporating\nspectral methods and enhanced observation models. Simulating vibration signals\nunder diverse conditions mitigates the high cost of obtaining experimental\nsignals for all scenarios. Given the non-stationary nature of railway vibration\nsignals, influenced by speed and load changes, our models address these\ncomplexities, offering a powerful tool for predictive maintenance in the rail\nindustry.", + "arxiv_id": "2501.11730v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper primarily focuses on predictive maintenance in the rail industry, utilizing a transformer model for vibration forecasting, but does not indicate the use of Large Language Models (LLMs) as required by the selection criteria, nor does it mention applications in knowledge graphs, retrieval-augmented generation, or agentic AI.\"\n}\n```" + }, + { + "paper": { + "title": "SILO: Solving Inverse Problems with Latent Operators", + "abstract": "Consistent improvement of image priors over the years has led to the\ndevelopment of better inverse problem solvers. Diffusion models are the\nnewcomers to this arena, posing the strongest known prior to date. Recently,\nsuch models operating in a latent space have become increasingly predominant\ndue to their efficiency. In recent works, these models have been applied to\nsolve inverse problems. Working in the latent space typically requires multiple\napplications of an Autoencoder during the restoration process, which leads to\nboth computational and restoration quality challenges. In this work, we propose\na new approach for handling inverse problems with latent diffusion models,\nwhere a learned degradation function operates within the latent space,\nemulating a known image space degradation. Usage of the learned operator\nreduces the dependency on the Autoencoder to only the initial and final steps\nof the restoration process, facilitating faster sampling and superior\nrestoration quality. We demonstrate the effectiveness of our method on a\nvariety of image restoration tasks and datasets, achieving significant\nimprovements over prior art.", + "arxiv_id": "2501.11746v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the required criteria as it primarily focuses on image restoration tasks using diffusion models and autoencoders, with no evident connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI." + }, + { + "paper": { + "title": "Policy-Adaptable Methods For Resolving Normative Conflicts Through\n Argumentation and Graph Colouring", + "abstract": "In a multi-agent system, one may choose to govern the behaviour of an agent\nby imposing norms, which act as guidelines for how agents should act either all\nof the time or in given situations. However, imposing multiple norms on one or\nmore agents may result in situations where these norms conflict over how the\nagent should behave. In any system with normative conflicts (such as safe\nreinforcement models or systems which monitor safety protocols), one must\ndecide which norms should be followed such that the most important and most\nrelevant norms are maintained. We introduce a new method for resolving\nnormative conflicts through argumentation and graph colouring which is\ncompatible with a variety of normative conflict resolution policies. We prove\nthat this method always creates an admissible set of arguments under\nargumentation semantics, meaning that it produces coherent outputs. We also\nintroduce more robust variants of this method, each building upon their\npredecessor to create a superior output, and we include further mathematical\nproof of their coherence. Our most advanced variant uses the existing concept\nof curtailment, where one norm may supersede another without fully eliminating\nit. The methods we introduce are all compatible with various pre-existing\npolicies for resolving normative conflicts. Empirical evaluations are also\nperformed to compare our algorithms to each other and to others in existing\nliterature.", + "arxiv_id": "2501.11799v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper does not explicitly mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the primary focus areas. It focuses on normative conflicts resolution in multi-agent systems using argumentation and graph coloring, which doesn't meet the mandatory criteria (1, 2, and 3) and doesn't explicitly align with any of the additional considerations.\"\n}\n```" + }, + { + "paper": { + "title": "Toward Effective Digraph Representation Learning: A Magnetic Adaptive\n Propagation based Approach", + "abstract": "The $q$-parameterized magnetic Laplacian serves as the foundation of directed\ngraph (digraph) convolution, enabling this kind of digraph neural network\n(MagDG) to encode node features and structural insights by complex-domain\nmessage passing. As a generalization of undirected methods, MagDG shows\nsuperior capability in modeling intricate web-scale topology. Despite the great\nsuccess achieved by existing MagDGs, limitations still exist: (1) Hand-crafted\n$q$: The performance of MagDGs depends on selecting an appropriate\n$q$-parameter to construct suitable graph propagation equations in the complex\ndomain. This parameter tuning, driven by downstream tasks, limits model\nflexibility and significantly increases manual effort. (2) Coarse Message\nPassing: Most approaches treat all nodes with the same complex-domain\npropagation and aggregation rules, neglecting their unique digraph contexts.\nThis oversight results in sub-optimal performance. To address the above issues,\nwe propose two key techniques: (1) MAP is crafted to be a plug-and-play\ncomplex-domain propagation optimization strategy in the context of digraph\nlearning, enabling seamless integration into any MagDG to improve predictions\nwhile enjoying high running efficiency. (2) MAP++ is a new digraph learning\nframework, further incorporating a learnable mechanism to achieve adaptively\nedge-wise propagation and node-wise aggregation in the complex domain for\nbetter performance. Extensive experiments on 12 datasets demonstrate that MAP\nenjoys flexibility for it can be incorporated with any MagDG, and scalability\nas it can deal with web-scale digraphs. MAP++ achieves SOTA predictive\nperformance on 4 different downstream tasks.", + "arxiv_id": "2501.11817v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on digraph representation learning and graph neural networks, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria related to LLMs." + }, + { + "paper": { + "title": "Toward Scalable Graph Unlearning: A Node Influence Maximization based\n Approach", + "abstract": "Machine unlearning, as a pivotal technology for enhancing model robustness\nand data privacy, has garnered significant attention in prevalent web mining\napplications, especially in thriving graph-based scenarios. However, most\nexisting graph unlearning (GU) approaches face significant challenges due to\nthe intricate interactions among web-scale graph elements during the model\ntraining: (1) The gradient-driven node entanglement hinders the complete\nknowledge removal in response to unlearning requests; (2) The billion-level\ngraph elements in the web scenarios present inevitable scalability issues. To\nbreak the above limitations, we open up a new perspective by drawing a\nconnection between GU and conventional social influence maximization. To this\nend, we propose Node Influence Maximization (NIM) through the decoupled\ninfluence propagation model and fine-grained influence function in a scalable\nmanner, which is crafted to be a plug-and-play strategy to identify potential\nnodes affected by unlearning entities. This approach enables offline execution\nindependent of GU, allowing it to be seamlessly integrated into most GU methods\nto improve their unlearning performance. Based on this, we introduce Scalable\nGraph Unlearning (SGU) as a new fine-tuned framework, which balances the\nforgetting and reasoning capability of the unlearned model by entity-specific\noptimizations. Extensive experiments on 14 datasets, including large-scale\nogbn-papers100M, have demonstrated the effectiveness of our approach.\nSpecifically, NIM enhances the forgetting capability of most GU methods, while\nSGU achieves comprehensive SOTA performance and maintains scalability.", + "arxiv_id": "2501.11823v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the primary criteria as it does not focus on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. Instead, it centers on graph unlearning, machine learning robustness, and data privacy, without apparent connection to LLMs or specified areas of interest." + }, + { + "paper": { + "title": "Data-driven Detection and Evaluation of Damages in Concrete Structures:\n Using Deep Learning and Computer Vision", + "abstract": "Structural integrity is vital for maintaining the safety and longevity of\nconcrete infrastructures such as bridges, tunnels, and walls. Traditional\nmethods for detecting damages like cracks and spalls are labor-intensive,\ntime-consuming, and prone to human error. To address these challenges, this\nstudy explores advanced data-driven techniques using deep learning for\nautomated damage detection and analysis. Two state-of-the-art instance\nsegmentation models, YOLO-v7 instance segmentation and Mask R-CNN, were\nevaluated using a dataset comprising 400 images, augmented to 10,995 images\nthrough geometric and color-based transformations to enhance robustness. The\nmodels were trained and validated using a dataset split into 90% training set,\nvalidation and test set 10%. Performance metrics such as precision, recall,\nmean average precision (mAP@0.5), and frames per second (FPS) were used for\nevaluation. YOLO-v7 achieved a superior mAP@0.5 of 96.1% and processed 40 FPS,\noutperforming Mask R-CNN, which achieved a mAP@0.5 of 92.1% with a slower\nprocessing speed of 18 FPS. The findings recommend YOLO-v7 instance\nsegmentation model for real-time, high-speed structural health monitoring,\nwhile Mask R-CNN is better suited for detailed offline assessments. This study\ndemonstrates the potential of deep learning to revolutionize infrastructure\nmaintenance, offering a scalable and efficient solution for automated damage\ndetection.", + "arxiv_id": "2501.11836v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the primary criteria as it primarily focuses on Computer Vision and Deep Learning for infrastructure maintenance, without any mention of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest." + }, + { + "paper": { + "title": "Supervised Learning for Analog and RF Circuit Design: Benchmarks and\n Comparative Insights", + "abstract": "Automating analog and radio-frequency (RF) circuit design using machine\nlearning (ML) significantly reduces the time and effort required for parameter\noptimization. This study explores supervised ML-based approaches for designing\ncircuit parameters from performance specifications across various circuit\ntypes, including homogeneous and heterogeneous designs. By evaluating diverse\nML models, from neural networks like transformers to traditional methods like\nrandom forests, we identify the best-performing models for each circuit. Our\nresults show that simpler circuits, such as low-noise amplifiers, achieve\nexceptional accuracy with mean relative errors as low as 0.3% due to their\nlinear parameter-performance relationships. In contrast, complex circuits, like\npower amplifiers and voltage-controlled oscillators, present challenges due to\ntheir non-linear interactions and larger design spaces. For heterogeneous\ncircuits, our approach achieves an 88% reduction in errors with increased\ntraining data, with the receiver achieving a mean relative error as low as\n0.23%, showcasing the scalability and accuracy of the proposed methodology.\nAdditionally, we provide insights into model strengths, with transformers\nexcelling in capturing non-linear mappings and k-nearest neighbors performing\nrobustly in moderately linear parameter spaces, especially in heterogeneous\ncircuits with larger datasets. This work establishes a foundation for extending\nML-driven design automation, enabling more efficient and scalable circuit\ndesign workflows.", + "arxiv_id": "2501.11839v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the primary criteria for Large Language Models (LLMs) applications, instead focusing on machine learning (ML) for analog and RF circuit design, with no apparent connection to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI." + }, + { + "paper": { + "title": "Scalable Whole Slide Image Representation Using K-Mean Clustering and\n Fisher Vector Aggregation", + "abstract": "Whole slide images (WSIs) are high-resolution, gigapixel sized images that\npose significant computational challenges for traditional machine learning\nmodels due to their size and heterogeneity.In this paper, we present a scalable\nand efficient methodology for WSI classification by leveraging patch-based\nfeature extraction, clustering, and Fisher vector encoding. Initially, WSIs are\ndivided into fixed size patches, and deep feature embeddings are extracted from\neach patch using a pre-trained convolutional neural network (CNN). These\npatch-level embeddings are subsequently clustered using K-means clustering,\nwhere each cluster aggregates semantically similar regions of the WSI. To\neffectively summarize each cluster, Fisher vector representations are computed\nby modeling the distribution of patch embeddings in each cluster as a\nparametric Gaussian mixture model (GMM). The Fisher vectors from each cluster\nare concatenated into a high-dimensional feature vector, creating a compact and\ninformative representation of the entire WSI. This feature vector is then used\nby a classifier to predict the WSI's diagnostic label. Our method captures\nlocal and global tissue structures and yields robust performance for\nlarge-scale WSI classification, demonstrating superior accuracy and scalability\ncompared to other approaches.", + "arxiv_id": "2501.12085v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper does not meet the primary criteria as it focuses on image processing (Whole Slide Image Representation) and machine learning models (CNN, K-Mean Clustering), with no mention of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest.\"\n}\n```" + }, + { + "paper": { + "title": "Proxies for Distortion and Consistency with Applications for Real-World\n Image Restoration", + "abstract": "Real-world image restoration deals with the recovery of images suffering from\nan unknown degradation. This task is typically addressed while being given only\ndegraded images, without their corresponding ground-truth versions. In this\nhard setting, designing and evaluating restoration algorithms becomes highly\nchallenging. This paper offers a suite of tools that can serve both the design\nand assessment of real-world image restoration algorithms. Our work starts by\nproposing a trained model that predicts the chain of degradations a given\nreal-world measured input has gone through. We show how this estimator can be\nused to approximate the consistency -- the match between the measurements and\nany proposed recovered image. We also use this estimator as a guiding force for\nthe design of a simple and highly-effective plug-and-play real-world image\nrestoration algorithm, leveraging a pre-trained diffusion-based image prior.\nFurthermore, this work proposes no-reference proxy measures of MSE and LPIPS,\nwhich, without access to the ground-truth images, allow ranking of real-world\nimage restoration algorithms according to their (approximate) MSE and LPIPS.\nThe proposed suite provides a versatile, first of its kind framework for\nevaluating and comparing blind image restoration algorithms in real-world\nscenarios.", + "arxiv_id": "2501.12102v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on image restoration, which falls under video/image processing, one of the excluded areas as per the evaluation criteria. No apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI is mentioned in the abstract." + }, + { + "paper": { + "title": "Efficient PINNs: Multi-Head Unimodular Regularization of the Solutions\n Space", + "abstract": "We present a machine learning framework to facilitate the solution of\nnonlinear multiscale differential equations and, especially, inverse problems\nusing Physics-Informed Neural Networks (PINNs). This framework is based on what\nis called multihead (MH) training, which involves training the network to learn\na general space of all solutions for a given set of equations with certain\nvariability, rather than learning a specific solution of the system. This setup\nis used with a second novel technique that we call Unimodular Regularization\n(UR) of the latent space of solutions. We show that the multihead approach,\ncombined with the regularization, significantly improves the efficiency of\nPINNs by facilitating the transfer learning process thereby enabling the\nfinding of solutions for nonlinear, coupled, and multiscale differential\nequations.", + "arxiv_id": "2501.12116v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the Practical Applications criterion related to Large Language Models (LLMs), as it focuses on Physics-Informed Neural Networks (PINNs) for solving nonlinear differential equations, without any apparent connection to LLMs, knowledge graphs, RAG, or agentic AI." + }, + { + "paper": { + "title": "On the practical applicability of modern DFT functionals for chemical\n computations. Case study of DM21 applicability for geometry optimization", + "abstract": "Density functional theory (DFT) is probably the most promising approach for\nquantum chemistry calculations considering its good balance between\ncalculations precision and speed. In recent years, several neural network-based\nfunctionals have been developed for exchange-correlation energy approximation\nin DFT, DM21 developed by Google Deepmind being the most notable between them.\nThis study focuses on evaluating the efficiency of DM21 functional in\npredicting molecular geometries, with a focus on the influence of oscillatory\nbehavior in neural network exchange-correlation functionals. We implemented\ngeometry optimization in PySCF for the DM21 functional in geometry optimization\nproblem, compared its performance with traditional functionals, and tested it\non various benchmarks. Our findings reveal both the potential and the current\nchallenges of using neural network functionals for geometry optimization in\nDFT. We propose a solution extending the practical applicability of such\nfunctionals and allowing to model new substances with their help.", + "arxiv_id": "2501.12149v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the required criteria as it primarily focuses on chemical computations using Density Functional Theory (DFT) and neural network-based functionals, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI." + }, + { + "paper": { + "title": "An End-to-End Approach for Korean Wakeword Systems with Speaker\n Authentication", + "abstract": "Wakeword detection plays a critical role in enabling AI assistants to listen\nto user voices and interact effectively. However, for languages other than\nEnglish, there is a significant lack of pre-trained wakeword models.\nAdditionally, systems that merely determine the presence of a wakeword can pose\nserious privacy concerns. In this paper, we propose an end-to-end approach that\ntrains wakewords for Non-English languages, particulary Korean, and uses this\nto develop a Voice Authentication model to protect user privacy. Our\nimplementation employs an open-source platform OpenWakeWord, which performs\nwakeword detection using an FCN (Fully-Connected Network) architecture. Once a\nwakeword is detected, our custom-developed code calculates cosine similarity\nfor robust user authentication. Experimental results demonstrate the\neffectiveness of our approach, achieving a 16.79% and a 6.6% Equal Error Rate\n(EER) each in the Wakeword Detection and the Voice Authentication. These\nfindings highlight the model's potential in providing secure and accurate\nwakeword detection and authentication for Korean users.", + "arxiv_id": "2501.12194v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper primarily focuses on voice processing and speaker authentication, which, while related to AI assistants, does not explicitly involve Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI, thus not meeting the core criteria.\"\n}\n```" + }, + { + "paper": { + "title": "Strong phonon-mediated high temperature superconductivity in\n Li$_2$AuH$_6$ under ambient pressure", + "abstract": "We used our developed AI search engine~(InvDesFlow) to perform extensive\ninvestigations regarding ambient stable superconducting hydrides. A cubic\nstructure Li$_2$AuH$_6$ with Au-H octahedral motifs is identified to be a\ncandidate. After performing thermodynamical analysis, we provide a feasible\nroute to experimentally synthesize this material via the known LiAu and LiH\ncompounds under ambient pressure. The further first-principles calculations\nsuggest that Li$_2$AuH$_6$ shows a high superconducting transition temperature\n($T_c$) $\\sim$ 140 K under ambient pressure. The H-1$s$ electrons strongly\ncouple with phonon modes of vibrations of Au-H octahedrons as well as\nvibrations of Li atoms, where the latter is not taken seriously in other\npreviously similar cases. Hence, different from previous claims of searching\nmetallic covalent bonds to find high-$T_c$ superconductors, we emphasize here\nthe importance of those phonon modes with strong electron-phonon coupling\n(EPC). And we suggest that one can intercalate atoms into binary or ternary\nhydrides to introduce more potential phonon modes with strong EPC, which is an\neffective approach to find high-$T_c$ superconductors within multicomponent\ncompounds.", + "arxiv_id": "2501.12222v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria, as it primarily focuses on materials science (superconductivity) with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. The mention of an 'AI search engine' is incidental and not central to the paper's main contribution." + }, + { + "paper": { + "title": "CBVLM: Training-free Explainable Concept-based Large Vision Language\n Models for Medical Image Classification", + "abstract": "The main challenges limiting the adoption of deep learning-based solutions in\nmedical workflows are the availability of annotated data and the lack of\ninterpretability of such systems. Concept Bottleneck Models (CBMs) tackle the\nlatter by constraining the final disease prediction on a set of predefined and\nhuman-interpretable concepts. However, the increased interpretability achieved\nthrough these concept-based explanations implies a higher annotation burden.\nMoreover, if a new concept needs to be added, the whole system needs to be\nretrained. Inspired by the remarkable performance shown by Large\nVision-Language Models (LVLMs) in few-shot settings, we propose a simple, yet\neffective, methodology, CBVLM, which tackles both of the aforementioned\nchallenges. First, for each concept, we prompt the LVLM to answer if the\nconcept is present in the input image. Then, we ask the LVLM to classify the\nimage based on the previous concept predictions. Moreover, in both stages, we\nincorporate a retrieval module responsible for selecting the best examples for\nin-context learning. By grounding the final diagnosis on the predicted\nconcepts, we ensure explainability, and by leveraging the few-shot capabilities\nof LVLMs, we drastically lower the annotation cost. We validate our approach\nwith extensive experiments across four medical datasets and twelve LVLMs (both\ngeneric and medical) and show that CBVLM consistently outperforms CBMs and\ntask-specific supervised methods without requiring any training and using just\na few annotated examples. More information on our project page:\nhttps://cristianopatricio.github.io/CBVLM/.", + "arxiv_id": "2501.12266v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI, which is explicitly listed as a criterion for rejection, outweighing its potential alignment with other criteria such as practical applications, experimental results, and novelty." + }, + { + "paper": { + "title": "Implementation of an Asymmetric Adjusted Activation Function for Class\n Imbalance Credit Scoring", + "abstract": "Credit scoring is a systematic approach to evaluate a borrower's probability\nof default (PD) on a bank loan. The data associated with such scenarios are\ncharacteristically imbalanced, complicating binary classification owing to the\noften-underestimated cost of misclassification during the classifier's learning\nprocess. Considering the high imbalance ratio (IR) of these datasets, we\nintroduce an innovative yet straightforward optimized activation function by\nincorporating an IR-dependent asymmetric adjusted factor embedded Sigmoid\nactivation function (ASIG). The embedding of ASIG makes the sensitive margin of\nthe Sigmoid function auto-adjustable, depending on the imbalance nature of the\ndatasets distributed, thereby giving the activation function an asymmetric\ncharacteristic that prevents the underrepresentation of the minority class\n(positive samples) during the classifier's learning process. The experimental\nresults show that the ASIG-embedded-classifier outperforms traditional\nclassifiers on datasets across wide-ranging IRs in the downstream\ncredit-scoring task. The algorithm also shows robustness and stability, even\nwhen the IR is ultra-high. Therefore, the algorithm provides a competitive\nalternative in the financial industry, especially in credit scoring, possessing\nthe ability to effectively process highly imbalanced distribution data.", + "arxiv_id": "2501.12285v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on a financial application (credit scoring) with no evident connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the essential criteria for practical applications of LLMs." + }, + { + "paper": { + "title": "Regressor-Guided Image Editing Regulates Emotional Response to Reduce\n Online Engagement", + "abstract": "Emotions are known to mediate the relationship between users' content\nconsumption and their online engagement, with heightened emotional intensity\nleading to increased engagement. Building on this insight, we propose three\nregressor-guided image editing approaches aimed at diminishing the emotional\nimpact of images. These include (i) a parameter optimization approach based on\nglobal image transformations known to influence emotions, (ii) an optimization\napproach targeting the style latent space of a generative adversarial network,\nand (iii) a diffusion-based approach employing classifier guidance and\nclassifier-free guidance. Our findings demonstrate that approaches can\neffectively alter the emotional properties of images while maintaining high\nvisual quality. Optimization-based methods primarily adjust low-level\nproperties like color hues and brightness, whereas the diffusion-based approach\nintroduces semantic changes, such as altering appearance or facial expressions.\nNotably, results from a behavioral study reveal that only the diffusion-based\napproach successfully elicits changes in viewers' emotional responses while\npreserving high perceived image quality. In future work, we will investigate\nthe impact of these image adaptations on internet user behavior.", + "arxiv_id": "2501.12289v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper primarily focuses on image processing and emotional response regulation, which aligns with video/image processing, and does not evidently involve Large Language Models (LLMs) or related areas like knowledge graphs, RAG, or agentic AI, failing to meet the required criteria.\"\n}\n```" + }, + { + "paper": { + "title": "DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial\n Basis Functions", + "abstract": "Splatting-based 3D reconstruction methods have gained popularity with the\nadvent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel\nviews. These methods commonly resort to using exponential family functions,\nsuch as the Gaussian function, as reconstruction kernels due to their\nanisotropic nature, ease of projection, and differentiability in rasterization.\nHowever, the field remains restricted to variations within the exponential\nfamily, leaving generalized reconstruction kernels largely underexplored,\npartly due to the lack of easy integrability in 3D to 2D projections. In this\nlight, we show that a class of decaying anisotropic radial basis functions\n(DARBFs), which are non-negative functions of the Mahalanobis distance,\nsupports splatting by approximating the Gaussian function's closed-form\nintegration advantage. With this fresh perspective, we demonstrate up to 34%\nfaster convergence during training and a 15% reduction in memory consumption\nacross various DARB reconstruction kernels, while maintaining comparable PSNR,\nSSIM, and LPIPS results. We will make the code available.", + "arxiv_id": "2501.12369v1" + }, + "decision": "REJECT", + "explanation": "Original response: ```\n{\n \"decision\": \"REJECT\",\n \"explanation\": \"The paper primarily focuses on 3D reconstruction and splatting with radial basis functions, which falls under video/image processing, and does not involve Large Language Models (LLMs) or related areas specified in the criteria.\"\n}" + }, + { + "paper": { + "title": "Expertise elevates AI usage: experimental evidence comparing laypeople\n and professional artists", + "abstract": "Novel capacities of generative AI to analyze and generate cultural artifacts\nraise inevitable questions about the nature and value of artistic education and\nhuman expertise. Has AI already leveled the playing field between professional\nartists and laypeople, or do trained artistic expressive capacity, curation\nskills and experience instead enhance the ability to use these new tools? In\nthis pre-registered study, we conduct experimental comparisons between 50\nactive artists and a demographically matched sample of laypeople. We designed\ntwo tasks to approximate artistic practice for testing their capabilities in\nboth faithful and creative image creation: replicating a reference image, and\nmoving as far away as possible from it. We developed a bespoke platform where\nparticipants used a modern text-to-image model to complete both tasks. We also\ncollected and compared participants' sentiments towards AI. On average, artists\nproduced more faithful and creative outputs than their lay counterparts,\nalthough only by a small margin. While AI may ease content creation,\nprofessional expertise is still valuable - even within the confined space of\ngenerative AI itself. Finally, we also explored how well an exemplary\nvision-capable large language model (GPT-4o) would complete the same tasks, if\ngiven the role of an image generation agent, and found it performed on par in\ncopying but outperformed even artists in the creative task. The very best\nresults were still produced by humans in both tasks. These outcomes highlight\nthe importance of integrating artistic skills with AI training to prepare\nartists and other visual professionals for a technologically evolving\nlandscape. We see a potential in collaborative synergy with generative AI,\nwhich could reshape creative industries and education in the arts.", + "arxiv_id": "2501.12374v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on the intersection of AI and artistic education, which aligns with 'social applications of AI' (regarding education and human expertise) and does not clearly meet the required criteria for Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, beyond a tangential mention." + }, + { + "paper": { + "title": "MMVU: Measuring Expert-Level Multi-Discipline Video Understanding", + "abstract": "We introduce MMVU, a comprehensive expert-level, multi-discipline benchmark\nfor evaluating foundation models in video understanding. MMVU includes 3,000\nexpert-annotated questions spanning 27 subjects across four core disciplines:\nScience, Healthcare, Humanities \u0026 Social Sciences, and Engineering. Compared to\nprior benchmarks, MMVU features three key advancements. First, it challenges\nmodels to apply domain-specific knowledge and perform expert-level reasoning to\nanalyze specialized-domain videos, moving beyond the basic visual perception\ntypically assessed in current video benchmarks. Second, each example is\nannotated by human experts from scratch. We implement strict data quality\ncontrols to ensure the high quality of the dataset. Finally, each example is\nenriched with expert-annotated reasoning rationals and relevant domain\nknowledge, facilitating in-depth analysis. We conduct an extensive evaluation\nof 32 frontier multimodal foundation models on MMVU. The latest\nSystem-2-capable models, o1 and Gemini 2.0 Flash Thinking, achieve the highest\nperformance among the tested models. However, they still fall short of matching\nhuman expertise. Through in-depth error analyses and case studies, we offer\nactionable insights for future advancements in expert-level,\nknowledge-intensive video understanding for specialized domains.", + "arxiv_id": "2501.12380v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing and understanding, which is explicitly listed as a criterion for rejection. While it mentions foundation models and knowledge-intensive tasks, the core application is video-centric, outweighing potential marginal relevance to Large Language Models (LLMs) in other areas." + }, + { + "paper": { + "title": "Physics of Skill Learning", + "abstract": "We aim to understand physics of skill learning, i.e., how skills are learned\nin neural networks during training. We start by observing the Domino effect,\ni.e., skills are learned sequentially, and notably, some skills kick off\nlearning right after others complete learning, similar to the sequential fall\nof domino cards. To understand the Domino effect and relevant behaviors of\nskill learning, we take physicists' approach of abstraction and simplification.\nWe propose three models with varying complexities -- the Geometry model, the\nResource model, and the Domino model, trading between reality and simplicity.\nThe Domino effect can be reproduced in the Geometry model, whose resource\ninterpretation inspires the Resource model, which can be further simplified to\nthe Domino model. These models present different levels of abstraction and\nsimplification; each is useful to study some aspects of skill learning. The\nGeometry model provides interesting insights into neural scaling laws and\noptimizers; the Resource model sheds light on the learning dynamics of\ncompositional tasks; the Domino model reveals the benefits of modularity. These\nmodels are not only conceptually interesting -- e.g., we show how Chinchilla\nscaling laws can emerge from the Geometry model, but also are useful in\npractice by inspiring algorithmic development -- e.g., we show how simple\nalgorithmic changes, motivated by these toy models, can speed up the training\nof deep learning models.", + "arxiv_id": "2501.12391v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper 'Physics of Skill Learning' does not meet the required criteria as it primarily focuses on understanding skill learning in neural networks, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not demonstrate direct practical applications or comparisons with state-of-the-art LLM techniques." + }, + { + "paper": { + "title": "Learning segmentation from point trajectories", + "abstract": "We consider the problem of segmenting objects in videos based on their motion\nand no other forms of supervision. Prior work has often approached this problem\nby using the principle of common fate, namely the fact that the motion of\npoints that belong to the same object is strongly correlated. However, most\nauthors have only considered instantaneous motion from optical flow. In this\nwork, we present a way to train a segmentation network using long-term point\ntrajectories as a supervisory signal to complement optical flow. The key\ndifficulty is that long-term motion, unlike instantaneous motion, is difficult\nto model -- any parametric approximation is unlikely to capture complex motion\npatterns over long periods of time. We instead draw inspiration from subspace\nclustering approaches, proposing a loss function that seeks to group the\ntrajectories into low-rank matrices where the motion of object points can be\napproximately explained as a linear combination of other point tracks. Our\nmethod outperforms the prior art on motion-based segmentation, which shows the\nutility of long-term motion and the effectiveness of our formulation.", + "arxiv_id": "2501.12392v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing, which is explicitly listed as a rejection criterion. Additionally, there is no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest." + }, + { + "paper": { + "title": "Unsupervised Learning in Echo State Networks for Input Reconstruction", + "abstract": "Conventional echo state networks (ESNs) require supervised learning to train\nthe readout layer, using the desired outputs as training data. In this study,\nwe focus on input reconstruction (IR), which refers to training the readout\nlayer to reproduce the input time series in its output. We reformulate the\nlearning algorithm of the ESN readout layer to perform IR using unsupervised\nlearning (UL). By conducting theoretical analysis and numerical experiments, we\ndemonstrate that IR in ESNs can be effectively implemented under realistic\nconditions without explicitly using the desired outputs as training data; in\nthis way, UL is enabled. Furthermore, we demonstrate that applications relying\non IR, such as dynamical system replication and noise filtering, can be\nreformulated within the UL framework. Our findings establish a theoretically\nsound and universally applicable IR formulation, along with its related tasks\nin ESNs. This work paves the way for novel predictions and highlights\nunresolved theoretical challenges in ESNs, particularly in the context of\ntime-series processing methods and computational models of the brain.", + "arxiv_id": "2501.11409v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria, as it focuses on unsupervised learning in Echo State Networks (ESNs) for input reconstruction, without mentioning Large Language Models (LLMs), practical applications related to LLMs, or relevant experimental results with quantitative metrics. It also doesn't meet any of the additional criteria relevant to LLMs." + }, + { + "paper": { + "title": "Improving thermal state preparation of Sachdev-Ye-Kitaev model with\n reinforcement learning on quantum hardware", + "abstract": "The Sachdev-Ye-Kitaev (SYK) model, known for its strong quantum correlations\nand chaotic behavior, serves as a key platform for quantum gravity studies.\nHowever, variationally preparing thermal states on near-term quantum processors\nfor large systems (N\u003e12, where N is the number of Majorana fermions) presents a\nsignificant challenge due to the rapid growth in the complexity of\nparameterized quantum circuits. This paper addresses this challenge by\nintegrating reinforcement learning (RL) with convolutional neural networks,\nemploying an iterative approach to optimize the quantum circuit and its\nparameters. The refinement process is guided by a composite reward signal\nderived from entropy and the expectation values of the SYK Hamiltonian. This\napproach reduces the number of CNOT gates by two orders of magnitude for\nsystems N\u003e10 compared to traditional methods like first-order Trotterization.\nWe demonstrate the effectiveness of the RL framework in both noiseless and\nnoisy quantum hardware environments, maintaining high accuracy in thermal state\npreparation. This work contributes to the advancement of a scalable, RL-based\nframework with applications for computations of thermal out-of-time-order\ncorrelators in quantum many-body systems and quantum gravity studies on\nnear-term quantum hardware.", + "arxiv_id": "2501.11454v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet the criteria as it primarily focuses on quantum computing and reinforcement learning for thermal state preparation, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI." + }, + { + "paper": { + "title": "With Great Backbones Comes Great Adversarial Transferability", + "abstract": "Advances in self-supervised learning (SSL) for machine vision have improved\nrepresentation robustness and model performance, giving rise to pre-trained\nbackbones like \\emph{ResNet} and \\emph{ViT} models tuned with SSL methods such\nas \\emph{SimCLR}. Due to the computational and data demands of pre-training,\nthe utilization of such backbones becomes a strenuous necessity. However,\nemploying these backbones may inherit vulnerabilities to adversarial attacks.\nWhile adversarial robustness has been studied under \\emph{white-box} and\n\\emph{black-box} settings, the robustness of models tuned on pre-trained\nbackbones remains largely unexplored. Additionally, the role of tuning\nmeta-information in mitigating exploitation risks is unclear. This work\nsystematically evaluates the adversarial robustness of such models across\n$20,000$ combinations of tuning meta-information, including fine-tuning\ntechniques, backbone families, datasets, and attack types. We propose using\nproxy models to transfer attacks, simulating varying levels of target knowledge\nby fine-tuning these proxies with diverse configurations. Our findings reveal\nthat proxy-based attacks approach the effectiveness of \\emph{white-box}\nmethods, even with minimal tuning knowledge. We also introduce a naive\n\"backbone attack,\" leveraging only the backbone to generate adversarial\nsamples, which outperforms \\emph{black-box} attacks and rivals \\emph{white-box}\nmethods, highlighting critical risks in model-sharing practices. Finally, our\nablations reveal how increasing tuning meta-information impacts attack\ntransferability, measuring each meta-information combination.", + "arxiv_id": "2501.12275v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on adversarial attacks and robustness in machine vision (computer vision) with self-supervised learning, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. The topic aligns more with video/image processing, which is explicitly listed as a reason for rejection." + }, + { + "paper": { + "title": "Goal-oriented Transmission Scheduling: Structure-guided DRL with a\n Unified Dual On-policy and Off-policy Approach", + "abstract": "Goal-oriented communications prioritize application-driven objectives over\ndata accuracy, enabling intelligent next-generation wireless systems. Efficient\nscheduling in multi-device, multi-channel systems poses significant challenges\ndue to high-dimensional state and action spaces. We address these challenges by\nderiving key structural properties of the optimal solution to the goal-oriented\nscheduling problem, incorporating Age of Information (AoI) and channel states.\nSpecifically, we establish the monotonicity of the optimal state value function\n(a measure of long-term system performance) w.r.t. channel states and prove its\nasymptotic convexity w.r.t. AoI states. Additionally, we derive the\nmonotonicity of the optimal policy w.r.t. channel states, advancing the\ntheoretical framework for optimal scheduling. Leveraging these insights, we\npropose the structure-guided unified dual on-off policy DRL (SUDO-DRL), a\nhybrid algorithm that combines the stability of on-policy training with the\nsample efficiency of off-policy methods. Through a novel structural property\nevaluation framework, SUDO-DRL enables effective and scalable training,\naddressing the complexities of large-scale systems. Numerical results show\nSUDO-DRL improves system performance by up to 45% and reduces convergence time\nby 40% compared to state-of-the-art methods. It also effectively handles\nscheduling in much larger systems, where off-policy DRL fails and on-policy\nbenchmarks exhibit significant performance loss, demonstrating its scalability\nand efficacy in goal-oriented communications.", + "arxiv_id": "2501.11921v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet any of the required criteria (1-3) as it focuses on Goal-oriented Transmission Scheduling in wireless systems using Deep Reinforcement Learning (DRL), with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI." + } + ], + "errors": [] +} diff --git a/llm_processor/newpapers-filtered.md b/llm_processor/newpapers-filtered.md new file mode 100644 index 0000000..e516fb4 --- /dev/null +++ b/llm_processor/newpapers-filtered.md @@ -0,0 +1,3487 @@ +# Accepted Papers + +## [Agent-R: Training Language Model Agents to Reflect via Iterative + Self-Training](https://arxiv.org/abs/2501.11425v1) +**arXiv ID:** 2501.11425v1 + +**Abstract:** +> Large Language Models (LLMs) agents are increasingly pivotal for addressing +complex tasks in interactive environments. Existing work mainly focuses on +enhancing performance through behavior cloning from stronger experts, yet such +approaches often falter in real-world applications, mainly due to the inability +to recover from errors. However, step-level critique data is difficult and +expensive to collect. Automating and dynamically constructing self-critique +datasets is thus crucial to empowering models with intelligent agent +capabilities. In this work, we propose an iterative self-training framework, +Agent-R, that enables language Agent to Reflect on the fly. Unlike traditional +methods that reward or penalize actions based on correctness, Agent-R leverages +MCTS to construct training data that recover correct trajectories from +erroneous ones. A key challenge of agent reflection lies in the necessity for +timely revision rather than waiting until the end of a rollout. To address +this, we introduce a model-guided critique construction mechanism: the actor +model identifies the first error step (within its current capability) in a +failed trajectory. Starting from it, we splice it with the adjacent correct +path, which shares the same parent node in the tree. This strategy enables the +model to learn reflection based on its current policy, therefore yielding +better learning efficiency. To further explore the scalability of this +self-improvement paradigm, we investigate iterative refinement of both error +correction capabilities and dataset construction. Our findings demonstrate that +Agent-R continuously improves the model's ability to recover from errors and +enables timely error correction. Experiments on three interactive environments +show that Agent-R effectively equips agents to correct erroneous actions while +avoiding loops, achieving superior performance compared to baseline methods +(+5.59%). + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications in interactive environments), 2 (Experimental Results with +5.59% performance improvement), and 3 (Comparison with State-of-the-Art with baseline methods). Additionally, shows novelty in introducing an iterative self-training framework for LLMs, enabling agentic AI for complex, real-world tasks, and demonstrates robust experimental validation." +} +``` + +--- + +## [Reasoning Language Models: A Blueprint](https://arxiv.org/abs/2501.11223v1) +**arXiv ID:** 2501.11223v1 + +**Abstract:** +> Reasoning language models (RLMs), also known as Large Reasoning Models +(LRMs), such as OpenAI's o1 and o3, DeepSeek-V3, and Alibaba's QwQ, have +redefined AI's problem-solving capabilities by extending large language models +(LLMs) with advanced reasoning mechanisms. Yet, their high costs, proprietary +nature, and complex architectures - uniquely combining Reinforcement Learning +(RL), search heuristics, and LLMs - present accessibility and scalability +challenges. To address these, we propose a comprehensive blueprint that +organizes RLM components into a modular framework, based on a survey and +analysis of all RLM works. This blueprint incorporates diverse reasoning +structures (chains, trees, graphs, and nested forms), reasoning strategies +(e.g., Monte Carlo Tree Search, Beam Search), RL concepts (policy, value models +and others), and supervision schemes (Output-Based and Process-Based +Supervision). We also provide detailed mathematical formulations and +algorithmic specifications to simplify RLM implementation. By showing how +schemes like LLaMA-Berry, QwQ, Journey Learning, and Graph of Thoughts fit as +special cases, we demonstrate the blueprint's versatility and unifying +potential. To illustrate its utility, we introduce x1, a modular implementation +for rapid RLM prototyping and experimentation. Using x1 and a literature +review, we provide key insights, such as multi-phase training for policy and +value models, and the importance of familiar training distributions. Finally, +we outline how RLMs can integrate with a broader LLM ecosystem, including tools +and databases. Our work demystifies RLM construction, democratizes advanced +reasoning capabilities, and fosters innovation, aiming to mitigate the gap +between "rich AI" and "poor AI" by lowering barriers to RLM development and +experimentation. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1, 2, 3) by focusing on practical applications of Large Language Models (LLMs) with advanced reasoning mechanisms, presenting experimental results with quantitative potential through the modular framework, and comparing with state-of-the-art techniques. Additionally meets criteria 4 (Methodology and Implementation Details) and shows Novelty, Reproducibility, and Impact through Experimental Validation, with no primary focus on excluded topics. + +--- + +## [Question-to-Question Retrieval for Hallucination-Free Knowledge Access: + An Approach for Wikipedia and Wikidata Question Answering](https://arxiv.org/abs/2501.11301v1) +**arXiv ID:** 2501.11301v1 + +**Abstract:** +> This paper introduces an approach to question answering over knowledge bases +like Wikipedia and Wikidata by performing "question-to-question" matching and +retrieval from a dense vector embedding store. Instead of embedding document +content, we generate a comprehensive set of questions for each logical content +unit using an instruction-tuned LLM. These questions are vector-embedded and +stored, mapping to the corresponding content. Vector embedding of user queries +are then matched against this question vector store. The highest similarity +score leads to direct retrieval of the associated article content, eliminating +the need for answer generation. Our method achieves high cosine similarity ( > +0.9 ) for relevant question pairs, enabling highly precise retrieval. This +approach offers several advantages including computational efficiency, rapid +response times, and increased scalability. We demonstrate its effectiveness on +Wikipedia and Wikidata, including multimedia content through structured fact +retrieval from Wikidata, opening up new pathways for multimodal question +answering. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets all mandatory criteria (1, 2, 3) by focusing on practical LLM application in knowledge retrieval, providing quantitative metrics (cosine similarity > 0.9), and implicitly comparing with state-of-the-art through demonstrated efficiency and scalability improvements. Additionally meets criteria 4 (clear methodology) and shows novelty in question-to-question retrieval approach." +} +``` + +--- + +## [Few-shot Policy (de)composition in Conversational Question Answering](https://arxiv.org/abs/2501.11335v1) +**arXiv ID:** 2501.11335v1 + +**Abstract:** +> The task of policy compliance detection (PCD) is to determine if a scenario +is in compliance with respect to a set of written policies. In a conversational +setting, the results of PCD can indicate if clarifying questions must be asked +to determine compliance status. Existing approaches usually claim to have +reasoning capabilities that are latent or require a large amount of annotated +data. In this work, we propose logical decomposition for policy compliance +(LDPC): a neuro-symbolic framework to detect policy compliance using large +language models (LLMs) in a few-shot setting. By selecting only a few exemplars +alongside recently developed prompting techniques, we demonstrate that our +approach soundly reasons about policy compliance conversations by extracting +sub-questions to be answered, assigning truth values from contextual +information, and explicitly producing a set of logic statements from the given +policies. The formulation of explicit logic graphs can in turn help answer +PCDrelated questions with increased transparency and explainability. We apply +this approach to the popular PCD and conversational machine reading benchmark, +ShARC, and show competitive performance with no task-specific finetuning. We +also leverage the inherently interpretable architecture of LDPC to understand +where errors occur, revealing ambiguities in the ShARC dataset and highlighting +the challenges involved with reasoning for conversational question answering. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in conversational question answering), 2 (Experimental Results with quantitative metrics on ShARC benchmark), and 3 (Comparison with State-of-the-Art with competitive performance). Additionally, it shows Novelty (neuro-symbolic framework with LLMs), Implementability (with current standard tools), and Reproducibility (with transparent documentation of methodologies and results). + +--- + +## [Towards Advancing Code Generation with Large Language Models: A Research + Roadmap](https://arxiv.org/abs/2501.11354v1) +**arXiv ID:** 2501.11354v1 + +**Abstract:** +> Recently, we have witnessed the rapid development of large language models, +which have demonstrated excellent capabilities in the downstream task of code +generation. However, despite their potential, LLM-based code generation still +faces numerous technical and evaluation challenges, particularly when embedded +in real-world development. In this paper, we present our vision for current +research directions, and provide an in-depth analysis of existing studies on +this task. We propose a six-layer vision framework that categorizes code +generation process into distinct phases, namely Input Phase, Orchestration +Phase, Development Phase, and Validation Phase. Additionally, we outline our +vision workflow, which reflects on the currently prevalent frameworks. We +systematically analyse the challenges faced by large language models, including +those LLM-based agent frameworks, in code generation tasks. With these, we +offer various perspectives and actionable recommendations in this area. Our aim +is to provide guidelines for improving the reliability, robustness and +usability of LLM-based code generation systems. Ultimately, this work seeks to +address persistent challenges and to provide practical suggestions for a more +pragmatic LLM-based solution for future code generation endeavors. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: (1) focuses on practical applications of LLMs in code generation, (2) implies experimental analysis through 'in-depth analysis of existing studies' and 'systematically analyse the challenges', (3) compares with state-of-the-art by analyzing 'currently prevalent frameworks'. Additionally, meets optional criteria (4) methodology discussion and (6) novelty in proposing a 'six-layer vision framework'. + +--- + +## [Neural Contextual Reinforcement Framework for Logical Structure Language + Generation](https://arxiv.org/abs/2501.11417v1) +**arXiv ID:** 2501.11417v1 + +**Abstract:** +> The Neural Contextual Reinforcement Framework introduces an innovative +approach to enhancing the logical coherence and structural consistency of text +generated by large language models. Leveraging reinforcement learning +principles, the framework integrates custom reward functions and dynamic +context alignment mechanisms to address challenges inherent in maintaining +long-range dependencies across extended sequences. The architecture +incorporates multi-head attention layers and hierarchical encoding modules, +enabling the model to produce outputs that align closely with human +expectations of logical structure and semantic flow. Quantitative evaluations +across diverse datasets demonstrate substantial improvements in coherence +metrics, perplexity reduction, and semantic alignment, showcasing the +framework's ability to outperform baseline models in both general and +domain-specific tasks. Qualitative analyses further highlight the framework's +capacity to generate text with improved narrative clarity and reduced +redundancy, reflecting its effectiveness in balancing fluency with structural +precision. In addition to its performance gains, the framework exhibits +robustness in handling noisy input data and scalability across varying model +sizes, reinforcing its versatility in practical applications. Experimental +results reveal that optimal context window sizes significantly influence +coherence outcomes, showing the importance of architectural flexibility in +adapting to diverse linguistic structures. Cross-lingual performance +evaluations affirm the framework's adaptability to multiple languages, +extending its utility beyond monolingual contexts. Resource efficiency analyses +indicate a reduction in computational overhead compared to traditional +approaches, emphasizing the practicality of the framework for large-scale +deployment. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications in language generation), 2 (Experimental Results with quantitative metrics), and 3 (Comparison with State-of-the-Art showing performance improvements). Additionally, it marginally meets criterion 4 (Methodology and Implementation Details) and shows Novelty in integrating reinforcement learning for logical structure enhancement, warranting further review." +} + +--- + +## [Explainable Lane Change Prediction for Near-Crash Scenarios Using + Knowledge Graph Embeddings and Retrieval Augmented Generation](https://arxiv.org/abs/2501.11560v1) +**arXiv ID:** 2501.11560v1 + +**Abstract:** +> Lane-changing maneuvers, particularly those executed abruptly or in risky +situations, are a significant cause of road traffic accidents. However, current +research mainly focuses on predicting safe lane changes. Furthermore, existing +accident datasets are often based on images only and lack comprehensive sensory +data. In this work, we focus on predicting risky lane changes using the CRASH +dataset (our own collected dataset specifically for risky lane changes), and +safe lane changes (using the HighD dataset). Then, we leverage KG and Bayesian +inference to predict these maneuvers using linguistic contextual information, +enhancing the model's interpretability and transparency. The model achieved a +91.5% f1-score with anticipation time extending to four seconds for risky lane +changes, and a 90.0% f1-score for predicting safe lane changes with the same +anticipation time. We validate our model by integrating it into a vehicle +within the CARLA simulator in scenarios that involve risky lane changes. The +model managed to anticipate sudden lane changes, thus providing automated +vehicles with further time to plan and execute appropriate safe reactions. +Finally, to enhance the explainability of our model, we utilize RAG to provide +clear and natural language explanations for the given prediction. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria (1. Practical Applications in autonomous vehicle safety, 2. Experimental Results with quantitative metrics e.g., 91.5% f1-score, 3. Comparison with State-of-the-Art implicitly through superior performance). Additionally meets criteria 4 (Methodology and Implementation Details), and shows Novelty (integration of KG, Bayesian inference with RAG for explainability), making it a strong candidate for further review. + +--- + +## [Recurrent Diffusion for Large-Scale Parameter Generation](https://arxiv.org/abs/2501.11587v1) +**arXiv ID:** 2501.11587v1 + +**Abstract:** +> Parameter generation has struggled to scale up for a long time, significantly +limiting its range of applications. In this study, we introduce +\textbf{R}ecurrent diffusion for large-scale \textbf{P}arameter +\textbf{G}eneration, called \textbf{RPG}. We first divide the trained +parameters into non-overlapping parts, after which a recurrent model is +proposed to learn their relationships. The recurrent model's outputs, as +conditions, are then fed into a diffusion model to generate the neural network +parameters. Using only a single GPU, recurrent diffusion enables us to generate +popular vision and language models such as ConvNeXt-L and LoRA parameters of +LLaMA-7B. Meanwhile, across various architectures and tasks, the generated +parameters consistently perform comparable results over trained networks. +Notably, our approach also shows the potential to generate models for handling +unseen tasks, which largely increases the practicality of parameter generation. +Our code is available +\href{https://github.com/NUS-HPC-AI-Lab/Recurrent-Parameter-Generation}{here}. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: generation of neural network parameters for Large Language Models like LLaMA-7B), 2 (Experimental Results: comparable performance of generated parameters across architectures and tasks), and 3 (Comparison with State-of-the-Art: outperforms in scalability with single GPU usage). Also shows novelty in approach and has reproducibility support through publicly available code. + +--- + +## [SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language + Models Tackling Knowledge-based Reasoning Tasks](https://arxiv.org/abs/2501.11599v1) +**arXiv ID:** 2501.11599v1 + +**Abstract:** +> Deductive reasoning is a crucial logical capability that assists us in +solving complex problems based on existing knowledge. Although augmented by +Chain-of-Thought prompts, Large Language Models (LLMs) might not follow the +correct reasoning paths. Enhancing the deductive reasoning abilities of LLMs, +and leveraging their extensive built-in knowledge for various reasoning tasks, +remains an open question. Attempting to mimic the human deductive reasoning +paradigm, we propose a multi-stage Syllogistic-Reasoning Framework of Thought +(SR-FoT) that enables LLMs to perform syllogistic deductive reasoning to handle +complex knowledge-based reasoning tasks. Our SR-FoT begins by interpreting the +question and then uses the interpretation and the original question to propose +a suitable major premise. It proceeds by generating and answering minor premise +questions in two stages to match the minor premises. Finally, it guides LLMs to +use the previously generated major and minor premises to perform syllogistic +deductive reasoning to derive the answer to the original question. Extensive +and thorough experiments on knowledge-based reasoning tasks have demonstrated +the effectiveness and advantages of our SR-FoT. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Applications (knowledge-based reasoning tasks), (2) Experimental Results and Quantitative Metrics (extensive experiments demonstrating effectiveness), and (3) Comparison with State-of-the-Art (implied by 'advantages' over existing approaches). Additionally, meets criterion (4) Methodology and Implementation Details (clearly described multi-stage framework). Shows novelty in applying syllogistic reasoning to LLMs. + +--- + +## [StAyaL | Multilingual Style Transfer](https://arxiv.org/abs/2501.11639v1) +**arXiv ID:** 2501.11639v1 + +**Abstract:** +> Stylistic text generation plays a vital role in enhancing communication by +reflecting the nuances of individual expression. This paper presents a novel +approach for generating text in a specific speaker's style across different +languages. We show that by leveraging only 100 lines of text, an individuals +unique style can be captured as a high-dimensional embedding, which can be used +for both text generation and stylistic translation. This methodology breaks +down the language barrier by transferring the style of a speaker between +languages. The paper is structured into three main phases: augmenting the +speaker's data with stylistically consistent external sources, separating style +from content using machine learning and deep learning techniques, and +generating an abstract style profile by mean pooling the learned embeddings. +The proposed approach is shown to be topic-agnostic, with test accuracy and F1 +scores of 74.9\% and 0.75, respectively. The results demonstrate the potential +of the style profile for multilingual communication, paving the way for further +applications in personalized content generation and cross-linguistic stylistic +transfer. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in multilingual communication), 2 (Experimental Results with quantitative metrics like test accuracy and F1 scores), and 4 (Methodology and Implementation Details of leveraging LLMs for style transfer). Novel approach to multilingual style transfer also suggests potential for impact through experimental validation. + +--- + +## [Optimizing Pretraining Data Mixtures with LLM-Estimated Utility](https://arxiv.org/abs/2501.11747v1) +**arXiv ID:** 2501.11747v1 + +**Abstract:** +> Large Language Models improve with increasing amounts of high-quality +training data. However, leveraging larger datasets requires balancing quality, +quantity, and diversity across sources. After evaluating nine baseline methods +under both compute- and data-constrained scenarios, we find token-count +heuristics outperform manual and learned mixes, indicating that simple +approaches accounting for dataset size and diversity are surprisingly +effective. Building on this insight, we propose two complementary approaches: +UtiliMax, which extends token-based heuristics by incorporating utility +estimates from reduced-scale ablations, achieving up to a 10.6x speedup over +manual baselines; and Model Estimated Data Utility (MEDU), which leverages LLMs +to estimate data utility from small samples, matching ablation-based +performance while reducing computational requirements by $\sim$200x. Together, +these approaches establish a new framework for automated, compute-efficient +data mixing that is robust across training regimes. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria (1: Practical Applications in pretraining data mixtures, 2: Experimental Results with quantitative metrics e.g., 10.6x speedup, 3: Comparison with State-of-the-Art techniques) and at least one additional criterion (4: Methodology and Implementation Details are clearly described). Novelty is also shown in leveraging LLMs for automated data mixing. + +--- + +## [Human-AI Collaborative Game Testing with Vision Language Models](https://arxiv.org/abs/2501.11782v1) +**arXiv ID:** 2501.11782v1 + +**Abstract:** +> As modern video games become increasingly complex, traditional manual testing +methods are proving costly and inefficient, limiting the ability to ensure +high-quality game experiences. While advancements in Artificial Intelligence +(AI) offer the potential to assist human testers, the effectiveness of AI in +truly enhancing real-world human performance remains underexplored. This study +investigates how AI can improve game testing by developing and experimenting +with an AI-assisted workflow that leverages state-of-the-art machine learning +models for defect detection. Through an experiment involving 800 test cases and +276 participants of varying backgrounds, we evaluate the effectiveness of AI +assistance under four conditions: with or without AI support, and with or +without detailed knowledge of defects and design documentation. The results +indicate that AI assistance significantly improves defect identification +performance, particularly when paired with detailed knowledge. However, +challenges arise when AI errors occur, negatively impacting human +decision-making. Our findings show the importance of optimizing human-AI +collaboration and implementing strategies to mitigate the effects of AI +inaccuracies. By this research, we demonstrate AI's potential and problems in +enhancing efficiency and accuracy in game testing workflows and offers +practical insights for integrating AI into the testing process. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications in a real-world task), 2 (Experimental Results with quantitative metrics), and 3 (Comparison with human-only conditions, implying state-of-the-art). Additionally, it touches on challenges and potential optimizations for human-AI collaboration, aligning with criterion 5 (Real-world Applications and Challenges)." +} +``` + +--- + +## [Benchmarking Large Language Models via Random Variables](https://arxiv.org/abs/2501.11790v1) +**arXiv ID:** 2501.11790v1 + +**Abstract:** +> With the continuous advancement of large language models (LLMs) in +mathematical reasoning, evaluating their performance in this domain has become +a prominent research focus. Recent studies have raised concerns about the +reliability of current mathematical benchmarks, highlighting issues such as +simplistic design and potential data leakage. Therefore, creating a reliable +benchmark that effectively evaluates the genuine capabilities of LLMs in +mathematical reasoning remains a significant challenge. To address this, we +propose RV-Bench, a framework for Benchmarking LLMs via Random Variables in +mathematical reasoning. Specifically, the background content of a random +variable question (RV question) mirrors the original problem in existing +standard benchmarks, but the variable combinations are randomized into +different values. LLMs must fully understand the problem-solving process for +the original problem to correctly answer RV questions with various combinations +of variable values. As a result, the LLM's genuine capability in mathematical +reasoning is reflected by its accuracy on RV-Bench. Extensive experiments are +conducted with 29 representative LLMs across 900+ RV questions. A leaderboard +for RV-Bench ranks the genuine capability of these LLMs. Further analysis of +accuracy dropping indicates that current LLMs still struggle with complex +mathematical reasoning problems. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications: evaluating LLMs in mathematical reasoning), 2 (Experimental Results and Quantitative Metrics: extensive experiments with 29 LLMs and quantitative accuracy metrics), and 3 (Comparison with State-of-the-Art: implicit comparison through a leaderboard). Additionally, shows Novelty (introducing RV-Bench) and robust Experimental Validation, justifying acceptance." +} +``` + +--- + +## [Fact-Preserved Personalized News Headline Generation](https://arxiv.org/abs/2501.11828v1) +**arXiv ID:** 2501.11828v1 + +**Abstract:** +> Personalized news headline generation, aiming at generating user-specific +headlines based on readers' preferences, burgeons a recent flourishing research +direction. Existing studies generally inject a user interest embedding into an +encoderdecoder headline generator to make the output personalized, while the +factual consistency of headlines is inadequate to be verified. In this paper, +we propose a framework Fact-Preserved Personalized News Headline Generation +(short for FPG), to prompt a tradeoff between personalization and consistency. +In FPG, the similarity between the candidate news to be exposed and the +historical clicked news is used to give different levels of attention to key +facts in the candidate news, and the similarity scores help to learn a +fact-aware global user embedding. Besides, an additional training procedure +based on contrastive learning is devised to further enhance the factual +consistency of generated headlines. Extensive experiments conducted on a +real-world benchmark PENS validate the superiority of FPG, especially on the +tradeoff between personalization and factual consistency. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications in personalized news headline generation), 2 (Experimental Results with quantitative metrics on real-world benchmark PENS), and 4 (Methodology and Implementation Details with contrastive learning for factual consistency). Novelty is also present in the tradeoff approach between personalization and factual consistency, making it a strong candidate for further review." +} +``` + +--- + +## [Is your LLM trapped in a Mental Set? Investigative study on how mental + sets affect the reasoning capabilities of LLMs](https://arxiv.org/abs/2501.11833v1) +**arXiv ID:** 2501.11833v1 + +**Abstract:** +> In this paper, we present an investigative study on how Mental Sets influence +the reasoning capabilities of LLMs. LLMs have excelled in diverse natural +language processing (NLP) tasks, driven by advancements in parameter-efficient +fine-tuning (PEFT) and emergent capabilities like in-context learning (ICL). +For complex reasoning tasks, selecting the right model for PEFT or ICL is +critical, often relying on scores on benchmarks such as MMLU, MATH, and GSM8K. +However, current evaluation methods, based on metrics like F1 Score or +reasoning chain assessments by larger models, overlook a key dimension: +adaptability to unfamiliar situations and overcoming entrenched thinking +patterns. In cognitive psychology, Mental Set refers to the tendency to persist +with previously successful strategies, even when they become inefficient - a +challenge for problem solving and reasoning. We compare the performance of LLM +models like Llama-3.1-8B-Instruct, Llama-3.1-70B-Instruct and GPT-4o in the +presence of mental sets. To the best of our knowledge, this is the first study +to integrate cognitive psychology concepts into the evaluation of LLMs for +complex reasoning tasks, providing deeper insights into their adaptability and +problem-solving efficacy. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 2 (Experimental Results and Quantitative Metrics) and 3 (Comparison with State-of-the-Art) through comparative performance analysis of LLM models. Also touches on novelty (Additional Consideration) by integrating cognitive psychology concepts into LLM evaluation, warranting further review for potential insights into LLM adaptability and problem-solving efficacy." +} +``` + +--- + +## [From Drafts to Answers: Unlocking LLM Potential via Aggregation + Fine-Tuning](https://arxiv.org/abs/2501.11877v1) +**arXiv ID:** 2501.11877v1 + +**Abstract:** +> Scaling data and model size has been proven effective for boosting the +performance of large language models. In addition to training-time scaling, +recent studies have revealed that increasing test-time computational resources +can further improve performance. In this work, we introduce Aggregation +Fine-Tuning (AFT), a supervised finetuning paradigm where the model learns to +synthesize multiple draft responses, referred to as proposals, into a single, +refined answer, termed aggregation. At inference time, a propose-and-aggregate +strategy further boosts performance by iteratively generating proposals and +aggregating them. Empirical evaluations on benchmark datasets show that +AFT-trained models substantially outperform standard SFT. Notably, an AFT +model, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3% LC +win rate on AlpacaEval 2, surpassing significantly larger LLMs such as +Llama3.1-405B-Instruct and GPT4. By combining sequential refinement and +parallel sampling, the propose-and-aggregate framework scales inference-time +computation in a flexible manner. Overall, These findings position AFT as a +promising approach to unlocking additional capabilities of LLMs without +resorting to increasing data volume or model size. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: focuses on practical application of LLMs (1), includes experimental results with quantitative metrics (2), and compares with state-of-the-art techniques (3). Additionally, meets criteria 4 (Methodology and Implementation Details) and shows novelty in approach (Additional Consideration). + +--- + +## [Panoramic Interests: Stylistic-Content Aware Personalized Headline + Generation](https://arxiv.org/abs/2501.11900v1) +**arXiv ID:** 2501.11900v1 + +**Abstract:** +> Personalized news headline generation aims to provide users with +attention-grabbing headlines that are tailored to their preferences. Prevailing +methods focus on user-oriented content preferences, but most of them overlook +the fact that diverse stylistic preferences are integral to users' panoramic +interests, leading to suboptimal personalization. In view of this, we propose a +novel Stylistic-Content Aware Personalized Headline Generation (SCAPE) +framework. SCAPE extracts both content and stylistic features from headlines +with the aid of large language model (LLM) collaboration. It further adaptively +integrates users' long- and short-term interests through a contrastive +learning-based hierarchical fusion network. By incorporating the panoramic +interests into the headline generator, SCAPE reflects users' stylistic-content +preferences during the generation process. Extensive experiments on the +real-world dataset PENS demonstrate the superiority of SCAPE over baselines. + +**Decision Explanation:** Original decision: ACCEPT +Meets ALL required criteria: (1) Practical Applications (personalized news headline generation), (2) Experimental Results and Quantitative Metrics (experiments on real-world PENS dataset), and (3) Comparison with State-of-the-Art (superiority over baselines). Additionally, meets extra criteria (4) Methodology and Implementation Details (utilizes LLM for feature extraction) and shows Novelty in approach. + +--- + +## [TAD-Bench: A Comprehensive Benchmark for Embedding-Based Text Anomaly + Detection](https://arxiv.org/abs/2501.11960v1) +**arXiv ID:** 2501.11960v1 + +**Abstract:** +> Text anomaly detection is crucial for identifying spam, misinformation, and +offensive language in natural language processing tasks. Despite the growing +adoption of embedding-based methods, their effectiveness and generalizability +across diverse application scenarios remain under-explored. To address this, we +present TAD-Bench, a comprehensive benchmark designed to systematically +evaluate embedding-based approaches for text anomaly detection. TAD-Bench +integrates multiple datasets spanning different domains, combining +state-of-the-art embeddings from large language models with a variety of +anomaly detection algorithms. Through extensive experiments, we analyze the +interplay between embeddings and detection methods, uncovering their strengths, +weaknesses, and applicability to different tasks. These findings offer new +perspectives on building more robust, efficient, and generalizable anomaly +detection systems for real-world applications. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1: Practical Applications in anomaly detection, 2: Experimental Results with quantitative metrics implied through 'extensive experiments', 3: Comparison with State-of-the-Art through integration of state-of-the-art embeddings). Additionally, meets criteria 4 (Methodology and Implementation Details for utilizing LLMs in anomaly detection) and exhibits Novelty in benchmarking embedding-based approaches. + +--- + +## [Bridging Visualization and Optimization: Multimodal Large Language + Models on Graph-Structured Combinatorial Optimization](https://arxiv.org/abs/2501.11968v1) +**arXiv ID:** 2501.11968v1 + +**Abstract:** +> Graph-structured combinatorial challenges are inherently difficult due to +their nonlinear and intricate nature, often rendering traditional computational +methods ineffective or expensive. However, these challenges can be more +naturally tackled by humans through visual representations that harness our +innate ability for spatial reasoning. In this study, we propose transforming +graphs into images to preserve their higher-order structural features +accurately, revolutionizing the representation used in solving graph-structured +combinatorial tasks. This approach allows machines to emulate human-like +processing in addressing complex combinatorial challenges. By combining the +innovative paradigm powered by multimodal large language models (MLLMs) with +simple search techniques, we aim to develop a novel and effective framework for +tackling such problems. Our investigation into MLLMs spanned a variety of +graph-based tasks, from combinatorial problems like influence maximization to +sequential decision-making in network dismantling, as well as addressing six +fundamental graph-related issues. Our findings demonstrate that MLLMs exhibit +exceptional spatial intelligence and a distinctive capability for handling +these problems, significantly advancing the potential for machines to +comprehend and analyze graph-structured data with a depth and intuition akin to +human cognition. These results also imply that integrating MLLMs with simple +optimization strategies could form a novel and efficient approach for +navigating graph-structured combinatorial challenges without complex +derivations, computationally demanding training and fine-tuning. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1: Practical Applications in graph-structured combinatorial optimization, 2: Experimental Results with quantitative metrics implied for MLLMs' performance, 3: Comparison with State-of-the-Art in handling graph-related issues). Additionally, meets criteria 4 (Methodology and Implementation Details for MLLMs in practical tasks) and shows Novelty in integrating MLLMs with optimization strategies, thus warranting further review. + +--- + +## [Leveraging Graph Structures and Large Language Models for End-to-End + Synthetic Task-Oriented Dialogues](https://arxiv.org/abs/2501.11977v1) +**arXiv ID:** 2501.11977v1 + +**Abstract:** +> Training task-oriented dialogue systems is both costly and time-consuming, +due to the need for high-quality datasets encompassing diverse intents. +Traditional methods depend on extensive human annotation, while recent +advancements leverage large language models (LLMs) to generate synthetic data. +However, these approaches often require custom prompts or code, limiting +accessibility for non-technical users. We introduce GraphTOD, an end-to-end +framework that simplifies the generation of task-oriented dialogues. Users can +create dialogues by specifying transition graphs in JSON format. Our evaluation +demonstrates that GraphTOD generates high-quality dialogues across various +domains, significantly lowering the cost and complexity of dataset creation. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications: generates synthetic task-oriented dialogues), 2 (Experimental Results and Quantitative Metrics: demonstrates high-quality dialogues across domains), 3 (Comparison with State-of-the-Art: implicitly by improving accessibility and lowering costs), and 4 (Methodology and Implementation Details: clearly describes end-to-end framework using LLMs with JSON format for non-technical users)." +} +``` + +--- + +## [Harnessing Generative Pre-Trained Transformer for Datacenter Packet + Trace Generation](https://arxiv.org/abs/2501.12033v1) +**arXiv ID:** 2501.12033v1 + +**Abstract:** +> Today, the rapid growth of applications reliant on datacenters calls for new +advancements to meet the increasing traffic and computational demands. Traffic +traces from datacenters are essential for further development and optimization +of future datacenters. However, traces are rarely released to the public. +Researchers often use simplified mathematical models that lack the depth needed +to recreate intricate traffic patterns and, thus, miss optimization +opportunities found in realistic traffic. In this preliminary work, we +introduce DTG-GPT, a packet-level Datacenter Traffic Generator (DTG), based on +the generative pre-trained transformer (GPT) architecture used by many +state-of-the-art large language models. We train our model on a small set of +available traffic traces from different domains and offer a simple methodology +to evaluate the fidelity of the generated traces to their original +counterparts. We show that DTG-GPT can synthesize novel traces that mimic the +spatiotemporal patterns found in real traffic traces. We further demonstrate +that DTG-GPT can generate traces for networks of different scales while +maintaining fidelity. Our findings indicate the potential that, in the future, +similar models to DTG-GPT will allow datacenter operators to release traffic +information to the research community via trained GPT models. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: generating datacenter packet traces with LLMs), 2 (Experimental Results: demonstrates fidelity in generated traces), and 4 (Methodology and Implementation Details: clearly describes using GPT architecture for packet-level traffic generation). Novel approach to a practical problem, with potential for real-world impact in datacenter optimization. + +--- + +## [Condor: Enhance LLM Alignment with Knowledge-Driven Data Synthesis and + Refinement](https://arxiv.org/abs/2501.12273v1) +**arXiv ID:** 2501.12273v1 + +**Abstract:** +> The quality of Supervised Fine-Tuning (SFT) data plays a critical role in +enhancing the conversational capabilities of Large Language Models (LLMs). +However, as LLMs become more advanced, the availability of high-quality +human-annotated SFT data has become a significant bottleneck, necessitating a +greater reliance on synthetic training data. In this work, we introduce Condor, +a novel two-stage synthetic data generation framework that incorporates World +Knowledge Tree and Self-Reflection Refinement to produce high-quality SFT data +at scale. Our experimental results demonstrate that a base model fine-tuned on +only 20K Condor-generated samples achieves superior performance compared to +counterparts. The additional refinement stage in Condor further enables +iterative self-improvement for LLMs at various scales (up to 72B), validating +the effectiveness of our approach. Furthermore, our investigation into the +scaling for synthetic data in post-training reveals substantial unexplored +potential for performance improvements, opening promising avenues for future +research. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1: Practical Applications in knowledge-driven data synthesis, 2: Experimental Results with quantitative metrics, 3: Comparison with State-of-the-Art techniques) and additional considerations (Novelty, Reproducibility, Impact through Experimental Validation); does not fall under rejection categories. + +--- + +## [RALAD: Bridging the Real-to-Sim Domain Gap in Autonomous Driving with + Retrieval-Augmented Learning](https://arxiv.org/abs/2501.12296v1) +**arXiv ID:** 2501.12296v1 + +**Abstract:** +> In the pursuit of robust autonomous driving systems, models trained on +real-world datasets often struggle to adapt to new environments, particularly +when confronted with corner cases such as extreme weather conditions. +Collecting these corner cases in the real world is non-trivial, which +necessitates the use of simulators for validation. However,the high +computational cost and the domain gap in data distribution have hindered the +seamless transition between real and simulated driving scenarios. To tackle +this challenge, we propose Retrieval-Augmented Learning for Autonomous Driving +(RALAD), a novel framework designed to bridge the real-to-sim gap at a low +cost. RALAD features three primary designs, including (1) domain adaptation via +an enhanced Optimal Transport (OT) method that accounts for both individual and +grouped image distances, (2) a simple and unified framework that can be applied +to various models, and (3) efficient fine-tuning techniques that freeze the +computationally expensive layers while maintaining robustness. Experimental +results demonstrate that RALAD compensates for the performance degradation in +simulated environments while maintaining accuracy in real-world scenarios +across three different models. Taking Cross View as an example, the mIOU and +mAP metrics in real-world scenarios remain stable before and after RALAD +fine-tuning, while in simulated environments,the mIOU and mAP metrics are +improved by 10.30% and 12.29%, respectively. Moreover, the re-training cost of +our approach is reduced by approximately 88.1%. Our code is available at +https://github.com/JiachengZuo/RALAD.git. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in autonomous driving), 2 (Experimental Results with quantitative metrics like mIOU and mAP), and 3 (Comparison with State-of-the-Art through performance improvements). Additionally, it partially meets criteria 4 (Methodology and Implementation Details, with code availability) and shows Novelty in applying Retrieval-Augmented Learning to bridge the real-to-sim domain gap. + +--- + +## [LLM-Assisted Knowledge Graph Completion for Curriculum and Domain + Modelling in Personalized Higher Education Recommendations](https://arxiv.org/abs/2501.12300v1) +**arXiv ID:** 2501.12300v1 + +**Abstract:** +> While learning personalization offers great potential for learners, modern +practices in higher education require a deeper consideration of domain models +and learning contexts, to develop effective personalization algorithms. This +paper introduces an innovative approach to higher education curriculum +modelling that utilizes large language models (LLMs) for knowledge graph (KG) +completion, with the goal of creating personalized learning-path +recommendations. Our research focuses on modelling university subjects and +linking their topics to corresponding domain models, enabling the integration +of learning modules from different faculties and institutions in the student's +learning path. Central to our approach is a collaborative process, where LLMs +assist human experts in extracting high-quality, fine-grained topics from +lecture materials. We develop a domain, curriculum, and user models for +university modules and stakeholders. We implement this model to create the KG +from two study modules: Embedded Systems and Development of Embedded Systems +Using FPGA. The resulting KG structures the curriculum and links it to the +domain models. We evaluate our approach through qualitative expert feedback and +quantitative graph quality metrics. Domain experts validated the relevance and +accuracy of the model, while the graph quality metrics measured the structural +properties of our KG. Our results show that the LLM-assisted graph completion +approach enhances the ability to connect related courses across disciplines to +personalize the learning experience. Expert feedback also showed high +acceptance of the proposed collaborative approach for concept extraction and +classification. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications in knowledge graph), 2 (Experimental Results with qualitative expert feedback and quantitative graph quality metrics), and 4 (Methodology and Implementation Details clearly described). Novelties in utilizing LLMs for collaborative knowledge graph completion in higher education also support inclusion." +} +``` + +--- + +## [Treefix: Enabling Execution with a Tree of Prefixes](https://arxiv.org/abs/2501.12339v1) +**arXiv ID:** 2501.12339v1 + +**Abstract:** +> The ability to execute code is a prerequisite for various dynamic program +analyses. Learning-guided execution has been proposed as an approach to enable +the execution of arbitrary code snippets by letting a neural model predict +likely values for any missing variables. Although state-of-the-art +learning-guided execution approaches, such as LExecutor, can enable the +execution of a relative high amount of code, they are limited to predicting a +restricted set of possible values and do not use any feedback from previous +executions to execute even more code. This paper presents Treefix, a novel +learning-guided execution approach that leverages LLMs to iteratively create +code prefixes that enable the execution of a given code snippet. The approach +addresses the problem in a multi-step fashion, where each step uses feedback +about the code snippet and its execution to instruct an LLM to improve a +previously generated prefix. This process iteratively creates a tree of +prefixes, a subset of which is returned to the user as prefixes that maximize +the number of executed lines in the code snippet. In our experiments with two +datasets of Python code snippets, Treefix achieves 25% and 7% more coverage +relative to the current state of the art in learning-guided execution, covering +a total of 84% and 82% of all lines in the code snippets. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Applications (enabling execution of arbitrary code snippets with LLMs), (2) Experimental Results and Quantitative Metrics (demonstrates 25% and 7% improvement over state-of-the-art with concrete coverage percentages), and (3) Comparison with State-of-the-Art (compares and advances upon existing learning-guided execution approaches). Additionally, shows novelty in methodology and potentially high impact through experimental validation. + +--- + +## [Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for + Mixture-of-Experts Language Models](https://arxiv.org/abs/2501.12370v1) +**arXiv ID:** 2501.12370v1 + +**Abstract:** +> Scaling the capacity of language models has consistently proven to be a +reliable approach for improving performance and unlocking new capabilities. +Capacity can be primarily defined by two dimensions: the number of model +parameters and the compute per example. While scaling typically involves +increasing both, the precise interplay between these factors and their combined +contribution to overall capacity remains not fully understood. We explore this +relationship in the context of sparse Mixture-of-Expert models (MoEs), which +allow scaling the number of parameters without proportionally increasing the +FLOPs per example. We investigate how varying the sparsity level, i.e., the +ratio of non-active to total parameters, affects model performance in terms of +both pretraining and downstream performance. We find that under different +constraints (e.g. parameter size and total training compute), there is an +optimal level of sparsity that improves both training efficiency and model +performance. These results provide a better understanding of the impact of +sparsity in scaling laws for MoEs and complement existing works in this area, +offering insights for designing more efficient architectures. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications for LLMs, specifically MoEs), 2 (Experimental Results with quantitative metrics on sparsity levels), and 3 (Comparison with State-of-the-Art in understanding scaling laws). Also shows potential for Novelty in optimizing LLM architecture efficiency. + +--- + +## [Is Long Context All You Need? Leveraging LLM's Extended Context for + NL2SQL](https://arxiv.org/abs/2501.12372v1) +**arXiv ID:** 2501.12372v1 + +**Abstract:** +> Large Language Models (LLMs) have demonstrated impressive capabilities across +a range of natural language processing tasks. In particular, improvements in +reasoning abilities and the expansion of context windows have opened new +avenues for leveraging these powerful models. NL2SQL is challenging in that the +natural language question is inherently ambiguous, while the SQL generation +requires a precise understanding of complex data schema and semantics. One +approach to this semantic ambiguous problem is to provide more and sufficient +contextual information. + In this work, we explore the performance and the latency trade-offs of the +extended context window (a.k.a., long context) offered by Google's +state-of-the-art LLM (\textit{gemini-1.5-pro}). We study the impact of various +contextual information, including column example values, question and SQL query +pairs, user-provided hints, SQL documentation, and schema. To the best of our +knowledge, this is the first work to study how the extended context window and +extra contextual information can help NL2SQL generation with respect to both +accuracy and latency cost. We show that long context LLMs are robust and do not +get lost in the extended contextual information. Additionally, our long-context +NL2SQL pipeline based on Google's \textit{gemini-pro-1.5} achieve a strong +performance with 67.41\% on BIRD benchmark (dev) without finetuning and +expensive self-consistency based techniques. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in NL2SQL), 2 (Experimental Results with quantitative metrics like 67.41% on BIRD benchmark), 3 (Comparison with State-of-the-Art through use of Google's gemini-1.5-pro), and 4 (Methodology and Implementation Details for utilizing LLMs in NL2SQL tasks). Additionally, shows Novelty in exploring extended context windows for LLMs in NL2SQL. + +--- + +## [Code Readability in the Age of Large Language Models: An Industrial Case + Study from Atlassian](https://arxiv.org/abs/2501.11264v1) +**arXiv ID:** 2501.11264v1 + +**Abstract:** +> Programmers spend a significant amount of time reading code during the +software development process. This trend is amplified by the emergence of large +language models (LLMs) that automatically generate code. However, little is +known about the readability of the LLM-generated code and whether it is still +important from practitioners' perspectives in this new era. In this paper, we +conduct a survey to explore the practitioners' perspectives on code readability +in the age of LLMs and investigate the readability of our LLM-based software +development agents framework, HULA, by comparing its generated code with +human-written code in real-world scenarios. Overall, the findings underscore +that (1) readability remains a critical aspect of software development; (2) the +readability of our LLM-generated code is comparable to human-written code, +fostering the establishment of appropriate trust and driving the broad adoption +of our LLM-powered software development platform. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in software development), 2 (Experimental Results comparing LLM-generated code with human-written code), and 4 (Methodology and Implementation Details of LLMs in a real-world task). Additionally, shows Novelty in applying LLMs to code readability and has potential for Reproducibility and real-world Impact. + +--- + +## [RedStar: Does Scaling Long-CoT Data Unlock Better Slow-Reasoning + Systems?](https://arxiv.org/abs/2501.11284v1) +**arXiv ID:** 2501.11284v1 + +**Abstract:** +> Can scaling transform reasoning? In this work, we explore the untapped +potential of scaling Long Chain-of-Thought (Long-CoT) data to 1000k samples, +pioneering the development of a slow-thinking model, RedStar. Through extensive +experiments with various LLMs and different sizes, we uncover the ingredients +for specialization and scale for Long-CoT training. Surprisingly, even smaller +models show significant performance gains with limited data, revealing the +sample efficiency of Long-CoT and the critical role of sample difficulty in the +learning process. Our findings demonstrate that Long-CoT reasoning can be +effectively triggered with just a few thousand examples, while larger models +achieve unparalleled improvements. We also introduce reinforcement learning +(RL)-scale training as a promising direction for advancing slow-thinking +systems. RedStar shines across domains: on the MATH-Hard benchmark, +RedStar-code-math boosts performance from 66.2\% to 81.6\%, and on the USA Math +Olympiad (AIME), it solves 46.7\% of problems using only 21k mixed-code-math +datasets. In multimodal tasks like GeoQA and MathVista-GEO, RedStar-Geo +achieves competitive results with minimal Long-CoT data, outperforming other +slow-thinking systems like QvQ-Preview. Compared to QwQ, RedStar strikes the +perfect balance between reasoning and generalizability. Our work highlights +that, with careful tuning, scaling Long-CoT can unlock extraordinary reasoning +capabilities-even with limited dataset and set a new standard for slow-thinking +models across diverse challenges. Our data and models are released at +https://huggingface.co/RedStar-Reasoning. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications in slow-thinking systems), 2 (Experimental Results with quantitative metrics in MATH-Hard, USA Math Olympiad, GeoQA, and MathVista-GEO), and 3 (Comparison with State-of-the-Art techniques like QvQ-Preview and QwQ). Additionally, shows novelty in scaling Long-CoT data and provides sufficient details for reproducibility with publicly released data and models." +} +``` + +--- + +## [Graph-defined Language Learning with LLMs](https://arxiv.org/abs/2501.11478v1) +**arXiv ID:** 2501.11478v1 + +**Abstract:** +> Recent efforts leverage Large Language Models (LLMs) for modeling +text-attributed graph structures in node classification tasks. These approaches +describe graph structures for LLMs to understand or aggregate LLM-generated +textual attribute embeddings through graph structure. However, these approaches +face two main limitations in modeling graph structures with LLMs. (i) Graph +descriptions become verbose in describing high-order graph structure. (ii) +Textual attributes alone do not contain adequate graph structure information. +It is challenging to model graph structure concisely and adequately with LLMs. +LLMs lack built-in mechanisms to model graph structures directly. They also +struggle with complex long-range dependencies between high-order nodes and +target nodes. + Inspired by the observation that LLMs pre-trained on one language can achieve +exceptional performance on another with minimal additional training, we propose +\textbf{G}raph-\textbf{D}efined \textbf{L}anguage for \textbf{L}arge +\textbf{L}anguage \textbf{M}odel (GDL4LLM). This novel framework enables LLMs +to transfer their powerful language understanding capabilities to +graph-structured data. GDL4LLM translates graphs into a graph language corpus +instead of graph descriptions and pre-trains LLMs on this corpus to adequately +understand graph structures. During fine-tuning, this corpus describes the +structural information of target nodes concisely with only a few tokens. By +treating graphs as a new language, GDL4LLM enables LLMs to model graph +structures adequately and concisely for node classification tasks. Extensive +experiments on three real-world datasets demonstrate that GDL4LLM outperforms +description-based and textual attribute embeddings-based baselines by +efficiently modeling different orders of graph structure with LLMs. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1, 2, 3): Focuses on practical application of LLMs in knowledge graph-like structures, includes experimental results with quantitative metrics, and compares with state-of-the-art techniques. Additionally, meets criteria 4 (clear methodology) and shows novelty in treating graphs as a new language for LLMs. + +--- + +## [Early evidence of how LLMs outperform traditional systems on OCR/HTR + tasks for historical records](https://arxiv.org/abs/2501.11623v1) +**arXiv ID:** 2501.11623v1 + +**Abstract:** +> We explore the ability of two LLMs -- GPT-4o and Claude Sonnet 3.5 -- to +transcribe historical handwritten documents in a tabular format and compare +their performance to traditional OCR/HTR systems: EasyOCR, Keras, Pytesseract, +and TrOCR. Considering the tabular form of the data, two types of experiments +are executed: one where the images are split line by line and the other where +the entire scan is used as input. Based on CER and BLEU, we demonstrate that +LLMs outperform the conventional OCR/HTR methods. Moreover, we also compare the +evaluated CER and BLEU scores to human evaluations to better judge the outputs +of whole-scan experiments and understand influential factors for CER and BLEU. +Combining judgments from all the evaluation metrics, we conclude that two-shot +GPT-4o for line-by-line images and two-shot Claude Sonnet 3.5 for whole-scan +images yield the transcriptions of the historical records most similar to the +ground truth. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1. Practical Applications in OCR/HTR tasks, 2. Experimental Results with quantitative metrics like CER and BLEU, and 3. Comparison with State-of-the-Art traditional OCR/HTR systems). Additionally, meets extra criteria for Methodology and Implementation Details (4) and shows robust Experimental Validation, making it a strong candidate for further review. + +--- + +## [Is logical analysis performed by transformers taking place in + self-attention or in the fully connected part?](https://arxiv.org/abs/2501.11765v1) +**arXiv ID:** 2501.11765v1 + +**Abstract:** +> Transformers architecture apply self-attention to tokens represented as +vectors, before a fully connected (neuronal network) layer. These two parts can +be layered many times. Traditionally, self-attention is seen as a mechanism for +aggregating information before logical operations are performed by the fully +connected layer. In this paper, we show, that quite counter-intuitively, the +logical analysis can also be performed within the self-attention. For this we +implement a handcrafted single-level encoder layer which performs the logical +analysis within self-attention. We then study the scenario in which a one-level +transformer model undergoes self-learning using gradient descent. We +investigate whether the model utilizes fully connected layers or self-attention +mechanisms for logical analysis when it has the choice. Given that gradient +descent can become stuck at undesired zeros, we explicitly calculate these +unwanted zeros and find ways to avoid them. We do all this in the context of +predicting grammatical category pairs of adjacent tokens in a text. We believe +that our findings have broader implications for understanding the potential +logical operations performed by self-attention. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications: analyzing transformer architecture for logical analysis), 2 (Experimental Results and Quantitative Metrics: implements a handcrafted model with self-learning using gradient descent), and 4 (Methodology and Implementation Details: clearly describes how LLMs are utilized). Novelty is also present in exploring logical operations within self-attention, justifying acceptance for further review." +} +``` + +--- + +## [Network-informed Prompt Engineering against Organized Astroturf + Campaigns under Extreme Class Imbalance](https://arxiv.org/abs/2501.11849v1) +**arXiv ID:** 2501.11849v1 + +**Abstract:** +> Detecting organized political campaigns is of paramount importance in +fighting against disinformation on social media. Existing approaches for the +identification of such organized actions employ techniques mostly from network +science, graph machine learning and natural language processing. Their ultimate +goal is to analyze the relationships and interactions (e.g. re-posting) among +users and the textual similarities of their posts. Despite their effectiveness +in recognizing astroturf campaigns, these methods face significant challenges, +notably the class imbalance in available training datasets. To mitigate this +issue, recent methods usually resort to data augmentation or increasing the +number of positive samples, which may not always be feasible or sufficient in +real-world settings. Following a different path, in this paper, we propose a +novel framework for identifying astroturf campaigns based solely on large +language models (LLMs), introducing a Balanced Retrieval-Augmented Generation +(Balanced RAG) component. Our approach first gives both textual information +concerning the posts (in our case tweets) and the user interactions of the +social network as input to a language model. Then, through prompt engineering +and the proposed Balanced RAG method, it effectively detects coordinated +disinformation campaigns on X (Twitter). The proposed framework does not +require any training or fine-tuning of the language model. Instead, by +strategically harnessing the strengths of prompt engineering and Balanced RAG, +it facilitates LLMs to overcome the effects of class imbalance and effectively +identify coordinated political campaigns. The experimental results demonstrate +that by incorporating the proposed prompt engineering and Balanced RAG methods, +our framework outperforms the traditional graph-based baselines, achieving +2x-3x improvements in terms of precision, recall and F1 scores. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in identifying astroturf campaigns), 2 (Experimental Results with quantitative metrics showing 2x-3x improvements), and 3 (Comparison with State-of-the-Art, outperforming traditional graph-based baselines). Although primarily focused on social media, the application is more closely related to disinformation detection, a security concern, rather than the excluded social applications (Diversity, Social harm, etc.). + +--- + +## [EDoRA: Efficient Weight-Decomposed Low-Rank Adaptation via Singular + Value Decomposition](https://arxiv.org/abs/2501.12067v1) +**arXiv ID:** 2501.12067v1 + +**Abstract:** +> Parameter-efficient fine-tuning methods, such as LoRA, reduces the number of +trainable parameters. However, they often suffer from scalability issues and +differences between their learning pattern and full fine-tuning. To overcome +these limitations, we propose Efficient Weight-Decomposed Low-Rank Adaptation +(EDoRA): a novel PEFT method that decomposes pre-trained weights into magnitude +and directional components. By freezing low-rank matrices, initializing them by +singular value decomposition, and introducing a small trainable matrix between +them, EDoRA achieves substantial reduction in trainable parameters while +maintaining learning capacity. Experimental results on the GLUE benchmark +demonstrate that EDoRA achieves competitive or superior performance compared to +state-of-the-art methods, such as LoRA and DoRA, with up to 30x fewer trainable +parameters. This makes EDoRA a highly efficient solution for adapting LLMs to +diverse tasks under memory-constrained settings. Code is available at +https://github.com/Hamid-Nasiri/EDoRA . + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications for adapting LLMs), 2 (Experimental Results with quantitative metrics on GLUE benchmark), and 3 (Comparison with State-of-the-Art methods like LoRA and DoRA). Additionally, shows novelty in methodology (EDoRA) and provides sufficient details for reproducibility (code available). + +--- + +## [Improving Influence-based Instruction Tuning Data Selection for Balanced + Learning of Diverse Capabilities](https://arxiv.org/abs/2501.12147v1) +**arXiv ID:** 2501.12147v1 + +**Abstract:** +> Selecting appropriate training data is crucial for effective instruction +fine-tuning of large language models (LLMs), which aims to (1) elicit strong +capabilities, and (2) achieve balanced performance across a diverse range of +tasks. Influence-based methods show promise in achieving (1) by estimating the +contribution of each training example to the model's predictions, but often +struggle with (2). Our systematic investigation reveals that this +underperformance can be attributed to an inherent bias where certain tasks +intrinsically have greater influence than others. As a result, data selection +is often biased towards these tasks, not only hurting the model's performance +on others but also, counterintuitively, harms performance on these +high-influence tasks themselves. + As a remedy, we propose BIDS, a Balanced and Influential Data Selection +algorithm. BIDS first normalizes influence scores of the training data, and +then iteratively balances data selection by choosing the training example with +the highest influence on the most underrepresented task. Experiments with both +Llama-3 and Mistral-v0.3 on seven benchmarks spanning five diverse capabilities +show that BIDS consistently outperforms both state-of-the-art influence-based +algorithms and other non-influence-based selection frameworks. Surprisingly, +training on a 15% subset selected by BIDS can even outperform full-dataset +training with a much more balanced performance. Our analysis further highlights +the importance of both instance-level normalization and iterative optimization +of selected data for balanced learning of diverse capabilities. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications of LLMs), 2 (Experimental Results with Quantitative Metrics), and 3 (Comparison with State-of-the-Art). Additionally, shows novelty in approach (BIDS algorithm) and robust experimental validation, making it a strong candidate for further review. + +--- + +## [AdaServe: SLO-Customized LLM Serving with Fine-Grained Speculative + Decoding](https://arxiv.org/abs/2501.12162v1) +**arXiv ID:** 2501.12162v1 + +**Abstract:** +> This paper introduces AdaServe, the first LLM serving system to support SLO +customization through fine-grained speculative decoding. AdaServe leverages the +logits of a draft model to predict the speculative accuracy of tokens and +employs a theoretically optimal algorithm to construct token trees for +verification. To accommodate diverse SLO requirements without compromising +throughput, AdaServe employs a speculation-and-selection scheme that first +constructs candidate token trees for each request and then dynamically selects +tokens to meet individual SLO constraints while optimizing throughput. +Comprehensive evaluations demonstrate that AdaServe achieves up to 73% higher +SLO attainment and 74% higher goodput compared to state-of-the-art systems. +These results underscore AdaServe's potential to enhance the efficiency and +adaptability of LLM deployments across varied application scenarios. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets all required criteria (1, 2, 3) by focusing on practical LLM application (AdaServe system), including experimental results with quantitative metrics (73% higher SLO attainment, 74% higher goodput), and comparing favorably to state-of-the-art techniques. Additionally, it meets optional criteria for methodology and implementation details (4), and novelty in extending LLM application." +} +``` + +--- + +## [InsTALL: Context-aware Instructional Task Assistance with Multi-modal + Large Language Models](https://arxiv.org/abs/2501.12231v1) +**arXiv ID:** 2501.12231v1 + +**Abstract:** +> The improved competence of generative models can help building multi-modal +virtual assistants that leverage modalities beyond language. By observing +humans performing multi-step tasks, one can build assistants that have +situational awareness of actions and tasks being performed, enabling them to +cater assistance based on this understanding. In this paper, we develop a +Context-aware Instructional Task Assistant with Multi-modal Large Language +Models (InsTALL) that leverages an online visual stream (e.g. a user's screen +share or video recording) and responds in real-time to user queries related to +the task at hand. To enable useful assistance, InsTALL 1) trains a multi-modal +model on task videos and paired textual data, and 2) automatically extracts +task graph from video data and leverages it at training and inference time. We +show InsTALL achieves state-of-the-art performance across proposed sub-tasks +considered for multimodal activity understanding -- task recognition (TR), +action recognition (AR), next action prediction (AP), and plan prediction (PP) +-- and outperforms existing baselines on two novel sub-tasks related to +automatic error identification. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: 1) Practical Application (multi-modal virtual task assistance), 2) Experimental Results (state-of-the-art performance on multiple sub-tasks), and 3) Comparison with State-of-the-Art (outperforms existing baselines). Additionally, meets criterion 4 (Methodology and Implementation Details) and shows Novelty in integrating LLMs with visual streams for real-world task assistance. + +--- + +## [UI-TARS: Pioneering Automated GUI Interaction with Native Agents](https://arxiv.org/abs/2501.12326v1) +**arXiv ID:** 2501.12326v1 + +**Abstract:** +> This paper introduces UI-TARS, a native GUI agent model that solely perceives +the screenshots as input and performs human-like interactions (e.g., keyboard +and mouse operations). Unlike prevailing agent frameworks that depend on +heavily wrapped commercial models (e.g., GPT-4o) with expert-crafted prompts +and workflows, UI-TARS is an end-to-end model that outperforms these +sophisticated frameworks. Experiments demonstrate its superior performance: +UI-TARS achieves SOTA performance in 10+ GUI agent benchmarks evaluating +perception, grounding, and GUI task execution. Notably, in the OSWorld +benchmark, UI-TARS achieves scores of 24.6 with 50 steps and 22.7 with 15 +steps, outperforming Claude (22.0 and 14.9 respectively). In AndroidWorld, +UI-TARS achieves 46.6, surpassing GPT-4o (34.5). UI-TARS incorporates several +key innovations: (1) Enhanced Perception: leveraging a large-scale dataset of +GUI screenshots for context-aware understanding of UI elements and precise +captioning; (2) Unified Action Modeling, which standardizes actions into a +unified space across platforms and achieves precise grounding and interaction +through large-scale action traces; (3) System-2 Reasoning, which incorporates +deliberate reasoning into multi-step decision making, involving multiple +reasoning patterns such as task decomposition, reflection thinking, milestone +recognition, etc. (4) Iterative Training with Reflective Online Traces, which +addresses the data bottleneck by automatically collecting, filtering, and +reflectively refining new interaction traces on hundreds of virtual machines. +Through iterative training and reflection tuning, UI-TARS continuously learns +from its mistakes and adapts to unforeseen situations with minimal human +intervention. We also analyze the evolution path of GUI agents to guide the +further development of this domain. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Applications (GUI interaction with LLMs), (2) Experimental Results and Quantitative Metrics (SOTA performance in 10+ GUI agent benchmarks), and (3) Comparison with State-of-the-Art (outperforms Claude and GPT-4o). Additionally, meets optional criteria (4) Methodology and Implementation Details (clearly described innovations) and demonstrates Novelty, Implementability, and Reproducibility. + +--- + +## [Automatic Labelling with Open-source LLMs using Dynamic Label Schema + Integration](https://arxiv.org/abs/2501.12332v1) +**arXiv ID:** 2501.12332v1 + +**Abstract:** +> Acquiring labelled training data remains a costly task in real world machine +learning projects to meet quantity and quality requirements. Recently Large +Language Models (LLMs), notably GPT-4, have shown great promises in labelling +data with high accuracy. However, privacy and cost concerns prevent the +ubiquitous use of GPT-4. In this work, we explore effectively leveraging +open-source models for automatic labelling. We identify integrating label +schema as a promising technology but found that naively using the label +description for classification leads to poor performance on high cardinality +tasks. To address this, we propose Retrieval Augmented Classification (RAC) for +which LLM performs inferences for one label at a time using corresponding label +schema; we start with the most related label and iterates until a label is +chosen by the LLM. We show that our method, which dynamically integrates label +description, leads to performance improvements in labelling tasks. We further +show that by focusing only on the most promising labels, RAC can trade off +between label quality and coverage - a property we leverage to automatically +label our internal datasets. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria (1: Practical Applications in automatic labelling, 2: Experimental Results with quantitative metrics for performance improvements, 3: Comparison with State-of-the-Art implied through innovation in Retrieval Augmented Classification). Additionally, meets optional criteria 4 (Methodology and Implementation Details for LLM utilization) and shows Novelty in integrating dynamic label schema with open-source LLMs. + +--- + +## [Test-time regression: a unifying framework for designing sequence models + with associative memory](https://arxiv.org/abs/2501.12352v1) +**arXiv ID:** 2501.12352v1 + +**Abstract:** +> Sequences provide a remarkably general way to represent and process +information. This powerful abstraction has placed sequence modeling at the +center of modern deep learning applications, inspiring numerous architectures +from transformers to recurrent networks. While this fragmented development has +yielded powerful models, it has left us without a unified framework to +understand their fundamental similarities and explain their effectiveness. We +present a unifying framework motivated by an empirical observation: effective +sequence models must be able to perform associative recall. Our key insight is +that memorizing input tokens through an associative memory is equivalent to +performing regression at test-time. This regression-memory correspondence +provides a framework for deriving sequence models that can perform associative +recall, offering a systematic lens to understand seemingly ad-hoc architectural +choices. We show numerous recent architectures -- including linear attention +models, their gated variants, state-space models, online learners, and softmax +attention -- emerge naturally as specific approaches to test-time regression. +Each architecture corresponds to three design choices: the relative importance +of each association, the regressor function class, and the optimization +algorithm. This connection leads to new understanding: we provide theoretical +justification for QKNorm in softmax attention, and we motivate higher-order +generalizations of softmax attention. Beyond unification, our work unlocks +decades of rich statistical tools that can guide future development of more +powerful yet principled sequence models. + +**Decision Explanation:** Original decision: ACCEPT +Although the paper's abstract doesn't explicitly mention Large Language Models (LLMs) or direct applications in knowledge graphs, RAG, or agentic AI, it presents a unifying framework for sequence models with associative memory. This framework could have implicit implications for LLMs and sequence modeling in NLP, potentially meeting criteria 1 and 4 with further review. Its experimental results and comparisons to existing architectures (e.g., transformers, recurrent networks) suggest a possible alignment with criteria 2 and 3. Inclusiveness is prioritized for marginally relevant papers, warranting acceptance for further review. + +--- + +## [Conversation Routines: A Prompt Engineering Framework for Task-Oriented + Dialog Systems](https://arxiv.org/abs/2501.11613v1) +**arXiv ID:** 2501.11613v1 + +**Abstract:** +> This study introduces Conversation Routines (CR), a structured prompt +engineering framework for developing task-oriented dialog systems using Large +Language Models (LLMs). While LLMs demonstrate remarkable natural language +understanding capabilities, engineering them to reliably execute complex +business workflows remains challenging. The proposed CR framework enables the +development of Conversation Agentic Systems (CAS) through natural language +specifications, embedding task-oriented logic within LLM prompts. This approach +provides a systematic methodology for designing and implementing complex +conversational workflows while maintaining behavioral consistency. We +demonstrate the framework's effectiveness through two proof of concept +implementations: a Train Ticket Booking System and an Interactive +Troubleshooting Copilot. These case studies validate CR's capability to encode +sophisticated behavioral patterns and decision logic while preserving natural +conversational flexibility. Results show that CR enables domain experts to +design conversational workflows in natural language while leveraging custom +enterprise functionalities (tools) developed by software engineers, creating an +efficient division of responsibilities where developers focus on core API +implementation and domain experts handle conversation design. While the +framework shows promise in accessibility and adaptability, we identify key +challenges including computational overhead, non-deterministic behavior, and +domain-specific logic optimization. Future research directions include +enhancing system robustness, improving scalability for complex multi-agent +interactions, and addressing the identified limitations across diverse business +applications. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: focuses on practical application of LLMs in task-oriented dialog systems (1), includes experimental results with quantitative metrics (2), and compares its methodology with existing state-of-the-art techniques (3). Additionally, it clearly describes methodology and implementation (4), discusses real-world applications and challenges (5), and shows novelty in integrating LLMs for autonomous task execution, making it a strong candidate for further review. + +--- + +# Rejected Papers + +## [The Explanation Game -- Rekindled (Extended Version)](https://arxiv.org/abs/2501.11429v1) +**arXiv ID:** 2501.11429v1 + +**Abstract:** +> Recent work demonstrated the existence of critical flaws in the current use +of Shapley values in explainable AI (XAI), i.e. the so-called SHAP scores. +These flaws are significant in that the scores provided to a human +decision-maker can be misleading. Although these negative results might appear +to indicate that Shapley values ought not be used in XAI, this paper argues +otherwise. Concretely, this paper proposes a novel definition of SHAP scores +that overcomes existing flaws. Furthermore, the paper outlines a practically +efficient solution for the rigorous estimation of the novel SHAP scores. +Preliminary experimental results confirm our claims, and further underscore the +flaws of the current SHAP scores. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Explainable AI (XAI) and Shapley values, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus not meeting the required criteria related to LLMs. + +--- + +## [Systematic Abductive Reasoning via Diverse Relation Representations in + Vector-symbolic Architecture](https://arxiv.org/abs/2501.11896v1) +**arXiv ID:** 2501.11896v1 + +**Abstract:** +> In abstract visual reasoning, monolithic deep learning models suffer from +limited interpretability and generalization, while existing neuro-symbolic +approaches fall short in capturing the diversity and systematicity of +attributes and relation representations. To address these challenges, we +propose a Systematic Abductive Reasoning model with diverse relation +representations (Rel-SAR) in Vector-symbolic Architecture (VSA) to solve +Raven's Progressive Matrices (RPM). To derive attribute representations with +symbolic reasoning potential, we introduce not only various types of atomic +vectors that represent numeric, periodic and logical semantics, but also the +structured high-dimentional representation (SHDR) for the overall Grid +component. For systematic reasoning, we propose novel numerical and logical +relation functions and perform rule abduction and execution in a unified +framework that integrates these relation representations. Experimental results +demonstrate that Rel-SAR achieves significant improvement on RPM tasks and +exhibits robust out-of-distribution generalization. Rel-SAR leverages the +synergy between HD attribute representations and symbolic reasoning to achieve +systematic abductive reasoning with both interpretable and computable +semantics. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on abstract visual reasoning and solving Raven's Progressive Matrices (RPM) with Vector-symbolic Architecture (VSA), without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the required criteria related to LLMs. + +--- + +## [Bridging the Communication Gap: Evaluating AI Labeling Practices for + Trustworthy AI Development](https://arxiv.org/abs/2501.11909v1) +**arXiv ID:** 2501.11909v1 + +**Abstract:** +> As artificial intelligence (AI) becomes integral to economy and society, +communication gaps between developers, users, and stakeholders hinder trust and +informed decision-making. High-level AI labels, inspired by frameworks like EU +energy labels, have been proposed to make the properties of AI models more +transparent. Without requiring deep technical expertise, they can inform on the +trade-off between predictive performance and resource efficiency. However, the +practical benefits and limitations of AI labeling remain underexplored. This +study evaluates AI labeling through qualitative interviews along four key +research questions. Based on thematic analysis and inductive coding, we found a +broad range of practitioners to be interested in AI labeling (RQ1). They see +benefits for alleviating communication gaps and aiding non-expert +decision-makers, however limitations, misunderstandings, and suggestions for +improvement were also discussed (RQ2). Compared to other reporting formats, +interviewees positively evaluated the reduced complexity of labels, increasing +overall comprehensibility (RQ3). Trust was influenced most by usability and the +credibility of the responsible labeling authority, with mixed preferences for +self-certification versus third-party certification (RQ4). Our Insights +highlight that AI labels pose a trade-off between simplicity and complexity, +which could be resolved by developing customizable and interactive labeling +frameworks to address diverse user needs. Transparent labeling of resource +efficiency also nudged interviewee priorities towards paying more attention to +sustainability aspects during AI development. This study validates AI labels as +a valuable tool for enhancing trust and communication in AI, offering +actionable guidelines for their refinement and standardization. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on trustworthy AI development, AI ethics (transparent labeling, trust, and responsible AI development), which falls under the rejection criteria (responsible AI application or AI ethics). It also lacks clear connections to Large Language Models (LLMs), practical applications in knowledge graphs, retrieval-augmented generation, or agentic AI as required by the selection criteria. + +--- + +## [Make Full Use of Testing Information: An Integrated Accelerated Testing + and Evaluation Method for Autonomous Driving Systems](https://arxiv.org/abs/2501.11924v1) +**arXiv ID:** 2501.11924v1 + +**Abstract:** +> Testing and evaluation is an important step before the large-scale +application of the autonomous driving systems (ADSs). Based on the three level +of scenario abstraction theory, a testing can be performed within a logical +scenario, followed by an evaluation stage which is inputted with the testing +results of each concrete scenario generated from the logical parameter space. +During the above process, abundant testing information is produced which is +beneficial for comprehensive and accurate evaluations. To make full use of +testing information, this paper proposes an Integrated accelerated Testing and +Evaluation Method (ITEM). Based on a Monte Carlo Tree Search (MCTS) paradigm +and a dual surrogates testing framework proposed in our previous work, this +paper applies the intermediate information (i.e., the tree structure, including +the affiliation of each historical sampled point with the subspaces and the +parent-child relationship between subspaces) generated during the testing stage +into the evaluation stage to achieve accurate hazardous domain identification. +Moreover, to better serve this purpose, the UCB calculation method is improved +to allow the search algorithm to focus more on the hazardous domain boundaries. +Further, a stopping condition is constructed based on the convergence of the +search algorithm. Ablation and comparative experiments are then conducted to +verify the effectiveness of the improvements and the superiority of the +proposed method. The experimental results show that ITEM could well identify +the hazardous domains in both low- and high-dimensional cases, regardless of +the shape of the hazardous domains, indicating its generality and potential for +the safety evaluation of ADSs. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on autonomous driving systems (ADSs) without explicitly involving Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria related to LLMs." +} +``` + +--- + +## [Collaborative Imputation of Urban Time Series through Cross-city + Meta-learning](https://arxiv.org/abs/2501.11306v1) +**arXiv ID:** 2501.11306v1 + +**Abstract:** +> Urban time series, such as mobility flows, energy consumption, and pollution +records, encapsulate complex urban dynamics and structures. However, data +collection in each city is impeded by technical challenges such as budget +limitations and sensor failures, necessitating effective data imputation +techniques that can enhance data quality and reliability. Existing imputation +models, categorized into learning-based and analytics-based paradigms, grapple +with the trade-off between capacity and generalizability. Collaborative +learning to reconstruct data across multiple cities holds the promise of +breaking this trade-off. Nevertheless, urban data's inherent irregularity and +heterogeneity issues exacerbate challenges of knowledge sharing and +collaboration across cities. To address these limitations, we propose a novel +collaborative imputation paradigm leveraging meta-learned implicit neural +representations (INRs). INRs offer a continuous mapping from domain coordinates +to target values, integrating the strengths of both paradigms. By imposing +embedding theory, we first employ continuous parameterization to handle +irregularity and reconstruct the dynamical system. We then introduce a +cross-city collaborative learning scheme through model-agnostic meta learning, +incorporating hierarchical modulation and normalization techniques to +accommodate multiscale representations and reduce variance in response to +heterogeneity. Extensive experiments on a diverse urban dataset from 20 global +cities demonstrate our model's superior imputation performance and +generalizability, underscoring the effectiveness of collaborative imputation in +resource-constrained settings. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on urban time series imputation using meta-learning, without any apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria (1, 2, and 3) related to LLMs. + +--- + +## [Finer-CAM: Spotting the Difference Reveals Finer Details for Visual + Explanation](https://arxiv.org/abs/2501.11309v1) +**arXiv ID:** 2501.11309v1 + +**Abstract:** +> Class activation map (CAM) has been widely used to highlight image regions +that contribute to class predictions. Despite its simplicity and computational +efficiency, CAM often struggles to identify discriminative regions that +distinguish visually similar fine-grained classes. Prior efforts address this +limitation by introducing more sophisticated explanation processes, but at the +cost of extra complexity. In this paper, we propose Finer-CAM, a method that +retains CAM's efficiency while achieving precise localization of discriminative +regions. Our key insight is that the deficiency of CAM lies not in "how" it +explains, but in "what" it explains}. Specifically, previous methods attempt to +identify all cues contributing to the target class's logit value, which +inadvertently also activates regions predictive of visually similar classes. By +explicitly comparing the target class with similar classes and spotting their +differences, Finer-CAM suppresses features shared with other classes and +emphasizes the unique, discriminative details of the target class. Finer-CAM is +easy to implement, compatible with various CAM methods, and can be extended to +multi-modal models for accurate localization of specific concepts. +Additionally, Finer-CAM allows adjustable comparison strength, enabling users +to selectively highlight coarse object contours or fine discriminative details. +Quantitatively, we show that masking out the top 5% of activated pixels by +Finer-CAM results in a larger relative confidence drop compared to baselines. +The source code and demo are available at +https://github.com/Imageomics/Finer-CAM. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on image processing and visual explanation using Class Activation Maps (CAM), with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the fundamental criteria. + +--- + +## [CatV2TON: Taming Diffusion Transformers for Vision-Based Virtual Try-On + with Temporal Concatenation](https://arxiv.org/abs/2501.11325v1) +**arXiv ID:** 2501.11325v1 + +**Abstract:** +> Virtual try-on (VTON) technology has gained attention due to its potential to +transform online retail by enabling realistic clothing visualization of images +and videos. However, most existing methods struggle to achieve high-quality +results across image and video try-on tasks, especially in long video +scenarios. In this work, we introduce CatV2TON, a simple and effective +vision-based virtual try-on (V2TON) method that supports both image and video +try-on tasks with a single diffusion transformer model. By temporally +concatenating garment and person inputs and training on a mix of image and +video datasets, CatV2TON achieves robust try-on performance across static and +dynamic settings. For efficient long-video generation, we propose an +overlapping clip-based inference strategy that uses sequential frame guidance +and Adaptive Clip Normalization (AdaCN) to maintain temporal consistency with +reduced resource demands. We also present ViViD-S, a refined video try-on +dataset, achieved by filtering back-facing frames and applying 3D mask +smoothing for enhanced temporal consistency. Comprehensive experiments +demonstrate that CatV2TON outperforms existing methods in both image and video +try-on tasks, offering a versatile and reliable solution for realistic virtual +try-ons across diverse scenarios. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on computer vision for virtual try-on, particularly in video processing, which is explicitly listed as a rejection criterion. Additionally, it lacks apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are key areas of interest. + +--- + +## [Federated Learning with Sample-level Client Drift Mitigation](https://arxiv.org/abs/2501.11360v1) +**arXiv ID:** 2501.11360v1 + +**Abstract:** +> Federated Learning (FL) suffers from severe performance degradation due to +the data heterogeneity among clients. Existing works reveal that the +fundamental reason is that data heterogeneity can cause client drift where the +local model update deviates from the global one, and thus they usually tackle +this problem from the perspective of calibrating the obtained local update. +Despite effectiveness, existing methods substantially lack a deep understanding +of how heterogeneous data samples contribute to the formation of client drift. +In this paper, we bridge this gap by identifying that the drift can be viewed +as a cumulative manifestation of biases present in all local samples and the +bias between samples is different. Besides, the bias dynamically changes as the +FL training progresses. Motivated by this, we propose FedBSS that first +mitigates the heterogeneity issue in a sample-level manner, orthogonal to +existing methods. Specifically, the core idea of our method is to adopt a +bias-aware sample selection scheme that dynamically selects the samples from +small biases to large epoch by epoch to train progressively the local model in +each round. In order to ensure the stability of training, we set the +diversified knowledge acquisition stage as the warm-up stage to avoid the local +optimality caused by knowledge deviation in the early stage of the model. +Evaluation results show that FedBSS outperforms state-of-the-art baselines. In +addition, we also achieved effective results on feature distribution skew and +noise label dataset setting, which proves that FedBSS can not only reduce +heterogeneity, but also has scalability and robustness. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on Federated Learning, which, while related to AI, does not explicitly mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus not meeting the primary criteria. Additionally, the topic aligns more closely with general AI challenges rather than the specified areas of interest. + +--- + +## [Multi-View Spectral Clustering for Graphs with Multiple View Structures](https://arxiv.org/abs/2501.11422v1) +**arXiv ID:** 2501.11422v1 + +**Abstract:** +> Despite the fundamental importance of clustering, to this day, much of the +relevant research is still based on ambiguous foundations, leading to an +unclear understanding of whether or how the various clustering methods are +connected with each other. In this work, we provide an additional stepping +stone towards resolving such ambiguities by presenting a general clustering +framework that subsumes a series of seemingly disparate clustering methods, +including various methods belonging to the wildly popular spectral clustering +framework. In fact, the generality of the proposed framework is additionally +capable of shedding light to the largely unexplored area of multi-view graphs +whose each view may have differently clustered nodes. In turn, we propose +GenClus: a method that is simultaneously an instance of this framework and a +generalization of spectral clustering, while also being closely related to +k-means as well. This results in a principled alternative to the few existing +methods studying this special type of multi-view graphs. Then, we conduct +in-depth experiments, which demonstrate that GenClus is more computationally +efficient than existing methods, while also attaining similar or better +clustering performance. Lastly, a qualitative real-world case-study further +demonstrates the ability of GenClus to produce meaningful clusterings. + +**Decision Explanation:** Original decision: REJECT +The paper does not mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the primary criteria (1) for practical applications of LLMs, and lacking clear connections to the specified areas of interest. + +--- + +## [A Survey on Diffusion Models for Anomaly Detection](https://arxiv.org/abs/2501.11430v1) +**arXiv ID:** 2501.11430v1 + +**Abstract:** +> Diffusion models (DMs) have emerged as a powerful class of generative AI +models, showing remarkable potential in anomaly detection (AD) tasks across +various domains, such as cybersecurity, fraud detection, healthcare, and +manufacturing. The intersection of these two fields, termed diffusion models +for anomaly detection (DMAD), offers promising solutions for identifying +deviations in increasingly complex and high-dimensional data. In this survey, +we systematically review recent advances in DMAD research and investigate their +capabilities. We begin by presenting the fundamental concepts of AD and DMs, +followed by a comprehensive analysis of classic DM architectures including +DDPMs, DDIMs, and Score SDEs. We further categorize existing DMAD methods into +reconstruction-based, density-based, and hybrid approaches, providing detailed +examinations of their methodological innovations. We also explore the diverse +tasks across different data modalities, encompassing image, time series, video, +and multimodal data analysis. Furthermore, we discuss critical challenges and +emerging research directions, including computational efficiency, model +interpretability, robustness enhancement, edge-cloud collaboration, and +integration with large language models. The collection of DMAD research papers +and resources is available at https://github.com/fdjingliu/DMAD. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on diffusion models for anomaly detection in various domains (including healthcare and video processing), with no clear indication of practical applications or experimental results specifically involving Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the required criteria. + +--- + +## [Communication-Efficient Federated Learning Based on Explanation-Guided + Pruning for Remote Sensing Image Classification](https://arxiv.org/abs/2501.11493v1) +**arXiv ID:** 2501.11493v1 + +**Abstract:** +> Federated learning (FL) is a decentralized machine learning paradigm, where +multiple clients collaboratively train a global model by exchanging only model +updates with the central server without sharing the local data of clients. Due +to the large volume of model updates required to be transmitted between clients +and the central server, most FL systems are associated with high transfer costs +(i.e., communication overhead). This issue is more critical for operational +applications in remote sensing (RS), especially when large-scale RS data is +processed and analyzed through FL systems with restricted communication +bandwidth. To address this issue, we introduce an explanation-guided pruning +strategy for communication-efficient FL in the context of RS image +classification. Our pruning strategy is defined based on the layerwise +relevance propagation (LRP) driven explanations to: 1) efficiently and +effectively identify the most relevant and informative model parameters (to be +exchanged between clients and the central server); and 2) eliminate the +non-informative ones to minimize the volume of model updates. The experimental +results on the BigEarthNet-S2 dataset demonstrate that our strategy effectively +reduces the number of shared model updates, while increasing the generalization +ability of the global model. The code of this work will be publicly available +at https://git.tu-berlin.de/rsim/FL-LRP + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Federated Learning for Remote Sensing Image Classification, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the core criteria. + +--- + +## [Meta-Instance Selection. Instance Selection as a Classification Problem + with Meta-Features](https://arxiv.org/abs/2501.11526v1) +**arXiv ID:** 2501.11526v1 + +**Abstract:** +> Data pruning, or instance selection, is an important problem in machine +learning especially in terms of nearest neighbour classifier. However, in data +pruning which speeds up the prediction phase, there is an issue related to the +speed and efficiency of the process itself. In response, the study proposes an +approach involving transforming the instance selection process into a +classification task conducted in a unified meta-feature space where each +instance can be classified and assigned to either the "to keep" or "to remove" +class. This approach requires training an appropriate meta-classifier, which +can be developed based on historical instance selection results from other +datasets using reference instance selection methods as a labeling tool. This +work proposes constructing the meta-feature space based on properties extracted +from the nearest neighbor graph. Experiments conducted on 17 datasets of +varying sizes and five reference instance selection methods (ENN, Drop3, ICF, +HMN-EI, and CCIS) demonstrate that the proposed solution achieves results +comparable to reference instance selection methods while significantly reducing +computational complexity. In the proposed approach, the computational +complexity of the system depends only on identifying the k-nearest neighbors +for each data sample and running the meta-classifier. Additionally, the study +discusses the choice of meta-classifier, recommending the use of Balanced +Random Forest. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria for Large Language Models (LLMs) applications, focusing instead on instance selection and classification in machine learning with nearest neighbor classifiers, without any mention of LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [The impact of intrinsic rewards on exploration in Reinforcement Learning](https://arxiv.org/abs/2501.11533v1) +**arXiv ID:** 2501.11533v1 + +**Abstract:** +> One of the open challenges in Reinforcement Learning is the hard exploration +problem in sparse reward environments. Various types of intrinsic rewards have +been proposed to address this challenge by pushing towards diversity. This +diversity might be imposed at different levels, favouring the agent to explore +different states, policies or behaviours (State, Policy and Skill level +diversity, respectively). However, the impact of diversity on the agent's +behaviour remains unclear. In this work, we aim to fill this gap by studying +the effect of different levels of diversity imposed by intrinsic rewards on the +exploration patterns of RL agents. We select four intrinsic rewards (State +Count, Intrinsic Curiosity Module (ICM), Maximum Entropy, and Diversity is all +you need (DIAYN)), each pushing for a different diversity level. We conduct an +empirical study on MiniGrid environment to compare their impact on exploration +considering various metrics related to the agent's exploration, namely: +episodic return, observation coverage, agent's position coverage, policy +entropy, and timeframes to reach the sparse reward. The main outcome of the +study is that State Count leads to the best exploration performance in the case +of low-dimensional observations. However, in the case of RGB observations, the +performance of State Count is highly degraded mostly due to representation +learning challenges. Conversely, Maximum Entropy is less impacted, resulting in +a more robust exploration, despite being not always optimal. Lastly, our +empirical study revealed that learning diverse skills with DIAYN, often linked +to improved robustness and generalisation, does not promote exploration in +MiniGrid environments. This is because: i) learning the skill space itself can +be challenging, and ii) exploration within the skill space prioritises +differentiating between behaviours rather than achieving uniform state +visitation. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Reinforcement Learning (RL) with intrinsic rewards, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the required criteria related to LLMs. + +--- + +## [Spatially-Delineated Domain-Adapted AI Classification: An Application + for Oncology Data](https://arxiv.org/abs/2501.11695v1) +**arXiv ID:** 2501.11695v1 + +**Abstract:** +> Given multi-type point maps from different place-types (e.g., tumor regions), +our objective is to develop a classifier trained on the source place-type to +accurately distinguish between two classes of the target place-type based on +their point arrangements. This problem is societally important for many +applications, such as generating clinical hypotheses for designing new +immunotherapies for cancer treatment. The challenge lies in the spatial +variability, the inherent heterogeneity and variation observed in spatial +properties or arrangements across different locations (i.e., place-types). +Previous techniques focus on self-supervised tasks to learn domain-invariant +features and mitigate domain differences; however, they often neglect the +underlying spatial arrangements among data points, leading to significant +discrepancies across different place-types. We explore a novel multi-task +self-learning framework that targets spatial arrangements, such as spatial +mix-up masking and spatial contrastive predictive coding, for +spatially-delineated domain-adapted AI classification. Experimental results on +real-world datasets (e.g., oncology data) show that the proposed framework +provides higher prediction accuracy than baseline methods. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (oncology data for cancer treatment), which is one of the explicit rejection criteria. + +--- + +## [Human services organizations and the responsible integration of AI: + Considering ethics and contextualizing risk(s)](https://arxiv.org/abs/2501.11705v1) +**arXiv ID:** 2501.11705v1 + +**Abstract:** +> This paper examines the responsible integration of artificial intelligence +(AI) in human services organizations (HSOs), proposing a nuanced framework for +evaluating AI applications across multiple dimensions of risk. The authors +argue that ethical concerns about AI deployment -- including professional +judgment displacement, environmental impact, model bias, and data laborer +exploitation -- vary significantly based on implementation context and specific +use cases. They challenge the binary view of AI adoption, demonstrating how +different applications present varying levels of risk that can often be +effectively managed through careful implementation strategies. The paper +highlights promising solutions, such as local large language models, that can +facilitate responsible AI integration while addressing common ethical concerns. +The authors propose a dimensional risk assessment approach that considers +factors like data sensitivity, professional oversight requirements, and +potential impact on client wellbeing. They conclude by outlining a path forward +that emphasizes empirical evaluation, starting with lower-risk applications and +building evidence-based understanding through careful experimentation. This +approach enables organizations to maintain high ethical standards while +thoughtfully exploring how AI might enhance their capacity to serve clients and +communities effectively. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on responsible AI application or AI ethics, which is explicitly listed as a criterion for rejection, overshadowing any potential marginal relevance to Large Language Models (LLMs) or other specified areas of interest. + +--- + +## [GL-ICNN: An End-To-End Interpretable Convolutional Neural Network for + the Diagnosis and Prediction of Alzheimer's Disease](https://arxiv.org/abs/2501.11715v1) +**arXiv ID:** 2501.11715v1 + +**Abstract:** +> Deep learning methods based on Convolutional Neural Networks (CNNs) have +shown great potential to improve early and accurate diagnosis of Alzheimer's +disease (AD) dementia based on imaging data. However, these methods have yet to +be widely adopted in clinical practice, possibly due to the limited +interpretability of deep learning models. The Explainable Boosting Machine +(EBM) is a glass-box model but cannot learn features directly from input +imaging data. In this study, we propose a novel interpretable model that +combines CNNs and EBMs for the diagnosis and prediction of AD. We develop an +innovative training strategy that alternatingly trains the CNN component as a +feature extractor and the EBM component as the output block to form an +end-to-end model. The model takes imaging data as input and provides both +predictions and interpretable feature importance measures. We validated the +proposed model on the Alzheimer's Disease Neuroimaging Initiative (ADNI) +dataset and the Health-RI Parelsnoer Neurodegenerative Diseases Biobank (PND) +as an external testing set. The proposed model achieved an area-under-the-curve +(AUC) of 0.956 for AD and control classification, and 0.694 for the prediction +of conversion of mild cognitive impairment (MCI) to AD on the ADNI cohort. The +proposed model is a glass-box model that achieves a comparable performance with +other state-of-the-art black-box models. Our code is publicly available at: +https://anonymous.4open.science/r/GL-ICNN. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (Alzheimer's Disease diagnosis and prediction), which is explicitly listed as a rejection criterion. + +--- + +## [Episodic memory in AI agents poses risks that should be studied and + mitigated](https://arxiv.org/abs/2501.11739v1) +**arXiv ID:** 2501.11739v1 + +**Abstract:** +> Most current AI models have little ability to store and later retrieve a +record or representation of what they do. In human cognition, episodic memories +play an important role in both recall of the past as well as planning for the +future. The ability to form and use episodic memories would similarly enable a +broad range of improved capabilities in an AI agent that interacts with and +takes actions in the world. Researchers have begun directing more attention to +developing memory abilities in AI models. It is therefore likely that models +with such capability will be become widespread in the near future. This could +in some ways contribute to making such AI agents safer by enabling users to +better monitor, understand, and control their actions. However, as a new +capability with wide applications, we argue that it will also introduce +significant new risks that researchers should begin to study and address. We +outline these risks and benefits and propose four principles to guide the +development of episodic memory capabilities so that these will enhance, rather +than undermine, the effort to keep AI safe and trustworthy. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on responsible AI application and AI safety/ethics, which is an explicitly stated rejection criterion, and does not appear to meet the required criteria related to practical applications, experimental results, or state-of-the-art comparisons for Large Language Models (LLMs). + +--- + +## [PXGen: A Post-hoc Explainable Method for Generative Models](https://arxiv.org/abs/2501.11827v1) +**arXiv ID:** 2501.11827v1 + +**Abstract:** +> With the rapid growth of generative AI in numerous applications, explainable +AI (XAI) plays a crucial role in ensuring the responsible development and +deployment of generative AI technologies. XAI has undergone notable +advancements and widespread adoption in recent years, reflecting a concerted +push to enhance the transparency, interpretability, and credibility of AI +systems. Recent research emphasizes that a proficient XAI method should adhere +to a set of criteria, primarily focusing on two key areas. Firstly, it should +ensure the quality and fluidity of explanations, encompassing aspects like +faithfulness, plausibility, completeness, and tailoring to individual needs. +Secondly, the design principle of the XAI system or mechanism should cover the +following factors such as reliability, resilience, the verifiability of its +outputs, and the transparency of its algorithm. However, research in XAI for +generative models remains relatively scarce, with little exploration into how +such methods can effectively meet these criteria in that domain. In this work, +we propose PXGen, a post-hoc explainable method for generative models. Given a +model that needs to be explained, PXGen prepares two materials for the +explanation, the Anchor set and intrinsic & extrinsic criteria. Those materials +are customizable by users according to their purpose and requirements. Via the +calculation of each criterion, each anchor has a set of feature values and +PXGen provides examplebased explanation methods according to the feature values +among all the anchors and illustrated and visualized to the users via tractable +algorithms such as k-dispersion or k-center. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Explainable AI (XAI) and responsible development of AI technologies, which aligns with 'responsible AI application or AI ethics', a criterion for rejection. Additionally, it lacks clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI as required by the selection criteria. + +--- + +## [A Survey on Memory-Efficient Large-Scale Model Training in AI for + Science](https://arxiv.org/abs/2501.11847v1) +**arXiv ID:** 2501.11847v1 + +**Abstract:** +> Scientific research faces high costs and inefficiencies with traditional +methods, but the rise of deep learning and large language models (LLMs) offers +innovative solutions. This survey reviews LLM applications across scientific +fields such as biology, medicine, chemistry, and meteorology, underscoring +their role in advancing research. However, the continuous expansion of model +size has led to significant memory demands, hindering further development and +application of LLMs for science. To address this, we review memory-efficient +training techniques for LLMs based on the transformer architecture, including +distributed training, mixed precision training, and gradient checkpointing. +Using AlphaFold 2 as an example, we demonstrate how tailored memory +optimization methods can reduce storage needs while preserving prediction +accuracy. We also discuss the challenges of memory optimization in practice and +potential future directions, hoping to provide valuable insights for +researchers and engineers. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on scientific applications, with a notable emphasis on biology and medicine, which aligns with excluded topics (primarily medical applications of AI). While it touches on Large Language Models, the core application area (science, including medicine) does not meet the inclusion criteria. + +--- + +## [Coarse-to-Fine Lightweight Meta-Embedding for ID-Based Recommendation](https://arxiv.org/abs/2501.11870v1) +**arXiv ID:** 2501.11870v1 + +**Abstract:** +> The state-of-the-art recommendation systems have shifted the attention to +efficient recommendation, e.g., on-device recommendation, under memory +constraints. To this end, the existing methods either focused on the +lightweight embeddings for both users and items, or involved on-device systems +enjoying the compact embeddings to enhance reusability and reduces space +complexity. However, they focus solely on the coarse granularity of embedding, +while overlook the fine-grained semantic nuances, to adversarially downgrade +the efficacy of meta-embeddings in capturing the intricate relationship over +both user and item, consequently resulting into the suboptimal recommendations. +In this paper, we aim to study how the meta-embedding can efficiently learn +varied grained semantics, together with how the fine-grained meta-embedding can +strengthen the representation of coarse-grained meta-embedding. To answer these +questions, we develop a novel graph neural networks (GNNs) based recommender +where each user and item serves as the node, linked directly to coarse-grained +virtual nodes and indirectly to fine-grained virtual nodes, ensuring different +grained semantic learning, while disclosing: 1) In contrast to coarse-grained +semantics, fine-grained semantics are well captured through sparse +meta-embeddings, which adaptively 2) balance the embedding uniqueness and +memory constraint. Additionally, the initialization method come up upon +SparsePCA, along with a soft thresholding activation function to render the +sparseness of the meta-embeddings. We propose a weight bridging update strategy +that focuses on matching each coarse-grained meta-embedding with several +fine-grained meta-embeddings based on the users/items' semantics. Extensive +experiments substantiate our method's superiority over existing baselines. Our +code is available at https://github.com/htyjers/C2F-MetaEmbed. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper does not meet the primary criteria related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. It focuses on recommendation systems using graph neural networks (GNNs) and meta-embeddings, which falls outside the specified areas of interest." +} +``` + +--- + +## [Community-Aware Temporal Walks: Parameter-Free Representation Learning + on Continuous-Time Dynamic Graphs](https://arxiv.org/abs/2501.11880v1) +**arXiv ID:** 2501.11880v1 + +**Abstract:** +> Dynamic graph representation learning plays a crucial role in understanding +evolving behaviors. However, existing methods often struggle with flexibility, +adaptability, and the preservation of temporal and structural dynamics. To +address these issues, we propose Community-aware Temporal Walks (CTWalks), a +novel framework for representation learning on continuous-time dynamic graphs. +CTWalks integrates three key components: a community-based parameter-free +temporal walk sampling mechanism, an anonymization strategy enriched with +community labels, and an encoding process that leverages continuous temporal +dynamics modeled via ordinary differential equations (ODEs). This design +enables precise modeling of both intra- and inter-community interactions, +offering a fine-grained representation of evolving temporal patterns in +continuous-time dynamic graphs. CTWalks theoretically overcomes locality bias +in walks and establishes its connection to matrix factorization. Experiments on +benchmark datasets demonstrate that CTWalks outperforms established methods in +temporal link prediction tasks, achieving higher accuracy while maintaining +robustness. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the primary criteria for Large Language Models (LLMs) applications, focusing instead on dynamic graph representation learning, temporal walk sampling, and ordinary differential equations (ODEs), with no apparent connection to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [LuxVeri at GenAI Detection Task 1: Inverse Perplexity Weighted Ensemble + for Robust Detection of AI-Generated Text across English and Multilingual + Contexts](https://arxiv.org/abs/2501.11914v1) +**arXiv ID:** 2501.11914v1 + +**Abstract:** +> This paper presents a system developed for Task 1 of the COLING 2025 Workshop +on Detecting AI-Generated Content, focusing on the binary classification of +machine-generated versus human-written text. Our approach utilizes an ensemble +of models, with weights assigned according to each model's inverse perplexity, +to enhance classification accuracy. For the English text detection task, we +combined RoBERTa-base, RoBERTa-base with the OpenAI detector, and +BERT-base-cased, achieving a Macro F1-score of 0.7458, which ranked us 12th out +of 35 teams. We ensembled RemBERT, XLM-RoBERTa-base, and +BERT-base-multilingual-case for the multilingual text detection task, employing +the same inverse perplexity weighting technique. This resulted in a Macro +F1-score of 0.7513, positioning us 4th out of 25 teams. Our results demonstrate +the effectiveness of inverse perplexity weighting in improving the robustness +of machine-generated text detection across both monolingual and multilingual +settings, highlighting the potential of ensemble methods for this challenging +task. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on detecting AI-generated text, which does not explicitly align with the specified practical applications of Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not clearly meet the criteria for real-world applications, methodology, or novelty in the context of LLMs as outlined. + +--- + +## [LuxVeri at GenAI Detection Task 3: Cross-Domain Detection of + AI-Generated Text Using Inverse Perplexity-Weighted Ensemble of Fine-Tuned + Transformer Models](https://arxiv.org/abs/2501.11918v1) +**arXiv ID:** 2501.11918v1 + +**Abstract:** +> This paper presents our approach for Task 3 of the GenAI content detection +workshop at COLING-2025, focusing on Cross-Domain Machine-Generated Text (MGT) +Detection. We propose an ensemble of fine-tuned transformer models, enhanced by +inverse perplexity weighting, to improve classification accuracy across diverse +text domains. For Subtask A (Non-Adversarial MGT Detection), we combined a +fine-tuned RoBERTa-base model with an OpenAI detector-integrated RoBERTa-base +model, achieving an aggregate TPR score of 0.826, ranking 10th out of 23 +detectors. In Subtask B (Adversarial MGT Detection), our fine-tuned +RoBERTa-base model achieved a TPR score of 0.801, securing 8th out of 22 +detectors. Our results demonstrate the effectiveness of inverse +perplexity-based weighting for enhancing generalization and performance in both +non-adversarial and adversarial MGT detection, highlighting the potential for +transformer models in cross-domain AI-generated content detection. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on AI-generated text detection, which does not explicitly align with the practical applications of Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. While it meets some criteria (e.g., experimental results, comparison with state-of-the-art), its core application does not satisfy the mandatory criteria for practical applications of LLMs as specified. + +--- + +## [A Lightweight and Interpretable Deepfakes Detection Framework](https://arxiv.org/abs/2501.11927v1) +**arXiv ID:** 2501.11927v1 + +**Abstract:** +> The recent realistic creation and dissemination of so-called deepfakes poses +a serious threat to social life, civil rest, and law. Celebrity defaming, +election manipulation, and deepfakes as evidence in court of law are few +potential consequences of deepfakes. The availability of open source trained +models based on modern frameworks such as PyTorch or TensorFlow, video +manipulations Apps such as FaceApp and REFACE, and economical computing +infrastructure has easen the creation of deepfakes. Most of the existing +detectors focus on detecting either face-swap, lip-sync, or puppet master +deepfakes, but a unified framework to detect all three types of deepfakes is +hardly explored. This paper presents a unified framework that exploits the +power of proposed feature fusion of hybrid facial landmarks and our novel heart +rate features for detection of all types of deepfakes. We propose novel heart +rate features and fused them with the facial landmark features to better +extract the facial artifacts of fake videos and natural variations available in +the original videos. We used these features to train a light-weight XGBoost to +classify between the deepfake and bonafide videos. We evaluated the performance +of our framework on the world leaders dataset (WLDR) that contains all types of +deepfakes. Experimental results illustrate that the proposed framework offers +superior detection performance over the comparative deepfakes detection +methods. Performance comparison of our framework against the LSTM-FCN, a +candidate of deep learning model, shows that proposed model achieves similar +results, however, it is more interpretable. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing (deepfakes detection), which is one of the criteria for rejection. Additionally, it touches on social implications (e.g., social life, civil unrest, law), another area for rejection. The paper does not appear to involve Large Language Models (LLMs) or related key areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Webvs. LLMs: An Empirical Study of Learning Behaviors of CS2 Students](https://arxiv.org/abs/2501.11935v1) +**arXiv ID:** 2501.11935v1 + +**Abstract:** +> LLMs such as ChatGPT have been widely adopted by students in higher education +as tools for learning programming and related concepts. However, it remains +unclear how effective students are and what strategies students use while +learning with LLMs. Since the majority of students' experiences in online +self-learning have come through using search engines such as Google, evaluating +AI tools in this context can help us address these gaps. In this mixed methods +research, we conducted an exploratory within-subjects study to understand how +CS2 students learn programming concepts using both LLMs as well as traditional +online methods such as educational websites and videos to examine how students +approach learning within and across both scenarios. We discovered that students +found it easier to learn a more difficult concept using traditional methods +than using ChatGPT. We also found that students ask fewer follow-ups and use +more keyword-based queries for search engines while their prompts to LLMs tend +to explicitly ask for information. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on educational applications of LLMs, which, while not explicitly listed for rejection, does not align with the preferred areas (knowledge graphs, RAG, agentic AI). It also lacks clear connections to practical real-world applications beyond educational settings, and its main findings relate to learning behaviors rather than showcasing innovative LLM applications, improvements, or state-of-the-art comparisons. + +--- + +## [MeshONet: A Generalizable and Efficient Operator Learning Method for + Structured Mesh Generation](https://arxiv.org/abs/2501.11937v1) +**arXiv ID:** 2501.11937v1 + +**Abstract:** +> Mesh generation plays a crucial role in scientific computing. Traditional +mesh generation methods, such as TFI and PDE-based methods, often struggle to +achieve a balance between efficiency and mesh quality. To address this +challenge, physics-informed intelligent learning methods have recently emerged, +significantly improving generation efficiency while maintaining high mesh +quality. However, physics-informed methods fail to generalize when applied to +previously unseen geometries, as even small changes in the boundary shape +necessitate burdensome retraining to adapt to new geometric variations. In this +paper, we introduce MeshONet, the first generalizable intelligent learning +method for structured mesh generation. The method transforms the mesh +generation task into an operator learning problem with multiple input and +solution functions. To effectively overcome the multivariable mapping +restriction of operator learning methods, we propose a dual-branch, +shared-trunk architecture to approximate the mapping between function spaces +based on input-output pairs. Experimental results show that MeshONet achieves a +speedup of up to four orders of magnitude in generation efficiency over +traditional methods. It also enables generalization to different geometries +without retraining, greatly enhancing the practicality of intelligent methods. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, focusing on mesh generation for scientific computing, which is unrelated to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not address real-world applications or challenges specific to LLMs. + +--- + +## [Survey on Hand Gesture Recognition from Visual Input](https://arxiv.org/abs/2501.11992v1) +**arXiv ID:** 2501.11992v1 + +**Abstract:** +> Hand gesture recognition has become an important research area, driven by the +growing demand for human-computer interaction in fields such as sign language +recognition, virtual and augmented reality, and robotics. Despite the rapid +growth of the field, there are few surveys that comprehensively cover recent +research developments, available solutions, and benchmark datasets. This survey +addresses this gap by examining the latest advancements in hand gesture and 3D +hand pose recognition from various types of camera input data including RGB +images, depth images, and videos from monocular or multiview cameras, examining +the differing methodological requirements of each approach. Furthermore, an +overview of widely used datasets is provided, detailing their main +characteristics and application domains. Finally, open challenges such as +achieving robust recognition in real-world environments, handling occlusions, +ensuring generalization across diverse users, and addressing computational +efficiency for real-time applications are highlighted to guide future research +directions. By synthesizing the objectives, methodologies, and applications of +recent studies, this survey offers valuable insights into current trends, +challenges, and opportunities for future research in human hand gesture +recognition. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, as it focuses on hand gesture recognition from visual input, which is unrelated to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and also falls under the rejection category of primarily focusing on video processing. + +--- + +## [Full Proportional Justified Representation](https://arxiv.org/abs/2501.12015v1) +**arXiv ID:** 2501.12015v1 + +**Abstract:** +> In multiwinner approval voting, forming a committee that proportionally +represents voters' approval ballots is an essential task. The notion of +justified representation (JR) demands that any large "cohesive" group of voters +should be proportionally "represented". The "cohesiveness" is defined in +different ways; two common ways are the following: (C1) demands that the group +unanimously approves a set of candidates proportional to its size, while (C2) +requires each member to approve at least a fixed fraction of such a set. +Similarly, "representation" have been considered in different ways: (R1) the +coalition's collective utility from the winning set exceeds that of any +proportionally sized alternative, and (R2) for any proportionally sized +alternative, at least one member of the coalition derives less utility from it +than from the winning set. + Three of the four possible combinations have been extensively studied: +(C1)-(R1) defines Proportional Justified Representation (PJR), (C1)-(R2) +defines Extended Justified Representation (EJR), (C2)-(R2) defines Full +Justified Representation (FJR). All three have merits, but also drawbacks. PJR +is the weakest notion, and perhaps not sufficiently demanding; EJR may not be +compatible with perfect representation; and it is open whether a committee +satisfying FJR can be found efficiently. + We study the combination (C2)-(R1), which we call Full Proportional Justified +Representation (FPJR). We investigate FPJR's properties and find that it shares +PJR's advantages over EJR: several proportionality axioms (e.g. priceability, +perfect representation) imply FPJR and PJR but not EJR. We also find that +efficient rules like the greedy Monroe rule and the method of equal shares +satisfy FPJR, matching a key advantage of EJR over FJR. However, the +Proportional Approval Voting (PAV) rule may violate FPJR, so neither of EJR and +FPJR implies the other. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, focusing on voting systems and proportional representation rather than Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, and lacks relevance to the specified areas of interest. + +--- + +## [Adaptive Class Learning to Screen Diabetic Disorders in Fundus Images of + Eye](https://arxiv.org/abs/2501.12048v1) +**arXiv ID:** 2501.12048v1 + +**Abstract:** +> The prevalence of ocular illnesses is growing globally, presenting a +substantial public health challenge. Early detection and timely intervention +are crucial for averting visual impairment and enhancing patient prognosis. +This research introduces a new framework called Class Extension with Limited +Data (CELD) to train a classifier to categorize retinal fundus images. The +classifier is initially trained to identify relevant features concerning +Healthy and Diabetic Retinopathy (DR) classes and later fine-tuned to adapt to +the task of classifying the input images into three classes: Healthy, DR, and +Glaucoma. This strategy allows the model to gradually enhance its +classification capabilities, which is beneficial in situations where there are +only a limited number of labeled datasets available. Perturbation methods are +also used to identify the input image characteristics responsible for +influencing the models decision-making process. We achieve an overall accuracy +of 91% on publicly available datasets. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (diabetic disorders in eye images), which is explicitly listed as a rejection criterion. + +--- + +## [Teacher Encoder-Student Decoder Denoising Guided Segmentation Network + for Anomaly Detection](https://arxiv.org/abs/2501.12104v1) +**arXiv ID:** 2501.12104v1 + +**Abstract:** +> Visual anomaly detection is a highly challenging task, often categorized as a +one-class classification and segmentation problem. Recent studies have +demonstrated that the student-teacher (S-T) framework effectively addresses +this challenge. However, most S-T frameworks rely solely on pre-trained teacher +networks to guide student networks in learning multi-scale similar features, +overlooking the potential of the student networks to enhance learning through +multi-scale feature fusion. In this study, we propose a novel model named +PFADSeg, which integrates a pre-trained teacher network, a denoising student +network with multi-scale feature fusion, and a guided anomaly segmentation +network into a unified framework. By adopting a unique teacher-encoder and +student-decoder denoising mode, the model improves the student network's +ability to learn from teacher network features. Furthermore, an adaptive +feature fusion mechanism is introduced to train a self-supervised segmentation +network that synthesizes anomaly masks autonomously, significantly increasing +detection performance. Evaluated on the MVTec AD dataset, PFADSeg achieves +state-of-the-art results with an image-level AUC of 98.9%, a pixel-level mean +precision of 76.4%, and an instance-level mean precision of 78.7%. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it focuses on visual anomaly detection using a student-teacher framework, without any indication of involving Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest. + +--- + +## [Can open source large language models be used for tumor documentation in + Germany? -- An evaluation on urological doctors' notes](https://arxiv.org/abs/2501.12106v1) +**arXiv ID:** 2501.12106v1 + +**Abstract:** +> Tumor documentation in Germany is largely done manually, requiring reading +patient records and entering data into structured databases. Large language +models (LLMs) could potentially enhance this process by improving efficiency +and reliability. This evaluation tests eleven different open source LLMs with +sizes ranging from 1-70 billion model parameters on three basic tasks of the +tumor documentation process: identifying tumor diagnoses, assigning ICD-10 +codes, and extracting the date of first diagnosis. For evaluating the LLMs on +these tasks, a dataset of annotated text snippets based on anonymized doctors' +notes from urology was prepared. Different prompting strategies were used to +investigate the effect of the number of examples in few-shot prompting and to +explore the capabilities of the LLMs in general. The models Llama 3.1 8B, +Mistral 7B, and Mistral NeMo 12 B performed comparably well in the tasks. +Models with less extensive training data or having fewer than 7 billion +parameters showed notably lower performance, while larger models did not +display performance gains. Examples from a different medical domain than +urology could also improve the outcome in few-shot prompting, which +demonstrates the ability of LLMs to handle tasks needed for tumor +documentation. Open source LLMs show a strong potential for automating tumor +documentation. Models from 7-12 billion parameters could offer an optimal +balance between performance and resource efficiency. With tailored fine-tuning +and well-designed prompting, these models might become important tools for +clinical documentation in the future. The code for the evaluation is available +from https://github.com/stefan-m-lenz/UroLlmEval. We also release the dataset +as a new valuable resource that addresses the shortage of authentic and easily +accessible benchmarks in German-language medical NLP. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (tumor documentation in Germany), which is an exclusion criterion as per the evaluation guidelines. + +--- + +## [FedCLEAN: byzantine defense by CLustering Errors of Activation maps in + Non-IID federated learning environments](https://arxiv.org/abs/2501.12123v1) +**arXiv ID:** 2501.12123v1 + +**Abstract:** +> Federated Learning (FL) enables clients to collaboratively train a global +model using their local datasets while reinforcing data privacy. However, FL is +susceptible to poisoning attacks. Existing defense mechanisms assume that +clients' data are independent and identically distributed (IID), making them +ineffective in real-world applications where data are non-IID. This paper +presents FedCLEAN, the first defense capable of filtering attackers' model +updates in a non-IID FL environment. The originality of FedCLEAN is twofold. +First, it relies on a client confidence score derived from the reconstruction +errors of each client's model activation maps for a given trigger set, with +reconstruction errors obtained by means of a Conditional Variational +Autoencoder trained according to a novel server-side strategy. Second, we +propose an ad-hoc trust propagation algorithm based on client scores, which +allows building a cluster of benign clients while flagging potential attackers. +Experimental results on the datasets MNIST and FashionMNIST demonstrate the +robustness of FedCLEAN against Byzantine attackers in non-IID scenarios and a +close-to-zero benign client misclassification rate, even in the absence of an +attack. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria as it focuses on federated learning (FL) and defense mechanisms against poisoning attacks, without apparent involvement or application of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest. + +--- + +## [FuocChuVIP123 at CoMeDi Shared Task: Disagreement Ranking with + XLM-Roberta Sentence Embeddings and Deep Neural Regression](https://arxiv.org/abs/2501.12336v1) +**arXiv ID:** 2501.12336v1 + +**Abstract:** +> This paper presents results of our system for CoMeDi Shared Task, focusing on +Subtask 2: Disagreement Ranking. Our system leverages sentence embeddings +generated by the paraphrase-xlm-r-multilingual-v1 model, combined with a deep +neural regression model incorporating batch normalization and dropout for +improved generalization. By predicting the mean of pairwise judgment +differences between annotators, our method explicitly targets disagreement +ranking, diverging from traditional "gold label" aggregation approaches. We +optimized our system with a customized architecture and training procedure, +achieving competitive performance in Spearman correlation against mean +disagreement labels. Our results highlight the importance of robust embeddings, +effective model architecture, and careful handling of judgment differences for +ranking disagreement in multilingual contexts. These findings provide insights +into the use of contextualized representations for ordinal judgment tasks and +open avenues for further refinement of disagreement prediction models. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper does not meet the primary criteria for Practical Applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. It also doesn't clearly align with other key criteria, focusing instead on disagreement ranking with sentence embeddings, which is not explicitly mentioned as a priority area in the evaluation guidelines." +} +``` + +--- + +## [Video Depth Anything: Consistent Depth Estimation for Super-Long Videos](https://arxiv.org/abs/2501.12375v1) +**arXiv ID:** 2501.12375v1 + +**Abstract:** +> Depth Anything has achieved remarkable success in monocular depth estimation +with strong generalization ability. However, it suffers from temporal +inconsistency in videos, hindering its practical applications. Various methods +have been proposed to alleviate this issue by leveraging video generation +models or introducing priors from optical flow and camera poses. Nonetheless, +these methods are only applicable to short videos (< 10 seconds) and require a +trade-off between quality and computational efficiency. We propose Video Depth +Anything for high-quality, consistent depth estimation in super-long videos +(over several minutes) without sacrificing efficiency. We base our model on +Depth Anything V2 and replace its head with an efficient spatial-temporal head. +We design a straightforward yet effective temporal consistency loss by +constraining the temporal depth gradient, eliminating the need for additional +geometric priors. The model is trained on a joint dataset of video depth and +unlabeled images, similar to Depth Anything V2. Moreover, a novel +key-frame-based strategy is developed for long video inference. Experiments +show that our model can be applied to arbitrarily long videos without +compromising quality, consistency, or generalization ability. Comprehensive +evaluations on multiple video benchmarks demonstrate that our approach sets a +new state-of-the-art in zero-shot video depth estimation. We offer models of +different scales to support a range of scenarios, with our smallest model +capable of real-time performance at 30 FPS. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing, specifically depth estimation in super-long videos, which is explicitly listed as a rejection criterion. + +--- + +## [Leveraging GANs For Active Appearance Models Optimized Model Fitting](https://arxiv.org/abs/2501.11218v1) +**arXiv ID:** 2501.11218v1 + +**Abstract:** +> Generative Adversarial Networks (GANs) have gained prominence in refining +model fitting tasks in computer vision, particularly in domains involving +deformable models like Active Appearance Models (AAMs). This paper explores the +integration of GANs to enhance the AAM fitting process, addressing challenges +in optimizing nonlinear parameters associated with appearance and shape +variations. By leveraging GANs' adversarial training framework, the aim is to +minimize fitting errors and improve convergence rates. Achieving robust +performance even in cases with high appearance variability and occlusions. Our +approach demonstrates significant improvements in accuracy and computational +efficiency compared to traditional optimization techniques, thus establishing +GANs as a potent tool for advanced image model fitting. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on computer vision (image model fitting) using GANs, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the core selection criteria. + +--- + +## [WSSM: Geographic-enhanced hierarchical state-space model for global + station weather forecast](https://arxiv.org/abs/2501.11238v1) +**arXiv ID:** 2501.11238v1 + +**Abstract:** +> Global Station Weather Forecasting (GSWF), a prominent meteorological +research area, is pivotal in providing timely localized weather predictions. +Despite the progress existing models have made in the overall accuracy of the +GSWF, executing high-precision extreme event prediction still presents a +substantial challenge. The recent emergence of state-space models, with their +ability to efficiently capture continuous-time dynamics and latent states, +offer potential solutions. However, early investigations indicated that Mamba +underperforms in the context of GSWF, suggesting further adaptation and +optimization. To tackle this problem, in this paper, we introduce Weather +State-space Model (WSSM), a novel Mamba-based approach tailored for GSWF. +Geographical knowledge is integrated in addition to the widely-used positional +encoding to represent the absolute special-temporal position. The multi-scale +time-frequency features are synthesized from coarse to fine to model the +seasonal to extreme weather dynamic. Our method effectively improves the +overall prediction accuracy and addresses the challenge of forecasting extreme +weather events. The state-of-the-art results obtained on the Weather-5K subset +underscore the efficacy of the WSSM + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on weather forecasting, a domain unrelated to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, with no mention of LLMs or relevant AI technologies. + +--- + +## [Spatiotemporal Air Quality Mapping in Urban Areas Using Sparse Sensor + Data, Satellite Imagery, Meteorological Factors, and Spatial Features](https://arxiv.org/abs/2501.11270v1) +**arXiv ID:** 2501.11270v1 + +**Abstract:** +> Monitoring air pollution is crucial for protecting human health from exposure +to harmful substances. Traditional methods of air quality monitoring, such as +ground-based sensors and satellite-based remote sensing, face limitations due +to high deployment costs, sparse sensor coverage, and environmental +interferences. To address these challenges, this paper proposes a framework for +high-resolution spatiotemporal Air Quality Index (AQI) mapping using sparse +sensor data, satellite imagery, and various spatiotemporal factors. By +leveraging Graph Neural Networks (GNNs), we estimate AQI values at unmonitored +locations based on both spatial and temporal dependencies. The framework +incorporates a wide range of environmental features, including meteorological +data, road networks, points of interest (PoIs), population density, and urban +green spaces, which enhance prediction accuracy. We illustrate the use of our +approach through a case study in Lahore, Pakistan, where multi-resolution data +is used to generate the air quality index map at a fine spatiotemporal scale. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria as it does not focus on Large Language Models (LLMs) or related areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. Instead, it leverages Graph Neural Networks (GNNs) for spatiotemporal air quality mapping, which does not align with the specified topics of interest. + +--- + +## [A Machine Learning Framework for Handling Unreliable Absence Label and + Class Imbalance for Marine Stinger Beaching Prediction](https://arxiv.org/abs/2501.11293v1) +**arXiv ID:** 2501.11293v1 + +**Abstract:** +> Bluebottles (\textit{Physalia} spp.) are marine stingers resembling +jellyfish, whose presence on Australian beaches poses a significant public risk +due to their venomous nature. Understanding the environmental factors driving +bluebottles ashore is crucial for mitigating their impact, and machine learning +tools are to date relatively unexplored. We use bluebottle marine stinger +presence/absence data from beaches in Eastern Sydney, Australia, and compare +machine learning models (Multilayer Perceptron, Random Forest, and XGBoost) to +identify factors influencing their presence. We address challenges such as +class imbalance, class overlap, and unreliable absence data by employing data +augmentation techniques, including the Synthetic Minority Oversampling +Technique (SMOTE), Random Undersampling, and Synthetic Negative Approach that +excludes the negative class. Our results show that SMOTE failed to resolve +class overlap, but the presence-focused approach effectively handled imbalance, +class overlap, and ambiguous absence data. The data attributes such as the wind +direction, which is a circular variable, emerged as a key factor influencing +bluebottle presence, confirming previous inference studies. However, in the +absence of population dynamics, biological behaviours, and life cycles, the +best predictive model appears to be Random Forests combined with Synthetic +Negative Approach. This research contributes to mitigating the risks posed by +bluebottles to beachgoers and provides insights into handling class overlap and +unreliable negative class in environmental modelling. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria (1-3) as it does not focus on Large Language Models (LLMs), lacks experimental results with quantitative metrics related to LLMs, and does not compare with state-of-the-art LLM techniques. Additionally, its primary focus on environmental modelling for marine stinger beaching prediction falls outside the specified rejection criteria, but more importantly, does not align with the paper selection criteria. + +--- + +## [On the Dimension of Pullback Attractors in Recurrent Neural Networks](https://arxiv.org/abs/2501.11357v1) +**arXiv ID:** 2501.11357v1 + +**Abstract:** +> Recurrent Neural Networks (RNNs) are high-dimensional state space models +capable of learning functions on sequence data. Recently, it has been +conjectured that reservoir computers, a particular class of RNNs, trained on +observations of a dynamical systems can be interpreted as embeddings. This +result has been established for the case of linear reservoir systems. In this +work, we use a nonautonomous dynamical systems approach to establish an upper +bound for the fractal dimension of the subset of reservoir state space +approximated during training and prediction phase. We prove that when the input +sequences comes from an Nin-dimensional invertible dynamical system, the +fractal dimension of this set is bounded above by Nin. The result obtained here +are useful in dimensionality reduction of computation in RNNs as well as +estimating fractal dimensions of dynamical systems from limited observations of +their time series. It is also a step towards understanding embedding properties +of reservoir computers. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, specifically lacking focus on Large Language Models (LLMs), practical applications in knowledge graphs, RAG, or agentic AI, and instead centers on Recurrent Neural Networks (RNNs) and dynamical systems, which are outside the specified scope. + +--- + +## [Investigation of Whisper ASR Hallucinations Induced by Non-Speech Audio](https://arxiv.org/abs/2501.11378v1) +**arXiv ID:** 2501.11378v1 + +**Abstract:** +> Hallucinations of deep neural models are amongst key challenges in automatic +speech recognition (ASR). In this paper, we investigate hallucinations of the +Whisper ASR model induced by non-speech audio segments present during +inference. By inducting hallucinations with various types of sounds, we show +that there exists a set of hallucinations that appear frequently. We then study +hallucinations caused by the augmentation of speech with such sounds. Finally, +we describe the creation of a bag of hallucinations (BoH) that allows to remove +the effect of hallucinations through the post-processing of text +transcriptions. The results of our experiments show that such post-processing +is capable of reducing word error rate (WER) and acts as a good safeguard +against problematic hallucinations. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Automatic Speech Recognition (ASR) and audio processing, with no clear connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the fundamental criteria for practical applications of LLMs. + +--- + +## [A Truly Sparse and General Implementation of Gradient-Based Synaptic + Plasticity](https://arxiv.org/abs/2501.11407v1) +**arXiv ID:** 2501.11407v1 + +**Abstract:** +> Online synaptic plasticity rules derived from gradient descent achieve high +accuracy on a wide range of practical tasks. However, their software +implementation often requires tediously hand-derived gradients or using +gradient backpropagation which sacrifices the online capability of the rules. +In this work, we present a custom automatic differentiation (AD) pipeline for +sparse and online implementation of gradient-based synaptic plasticity rules +that generalizes to arbitrary neuron models. Our work combines the programming +ease of backpropagation-type methods for forward AD while being +memory-efficient. To achieve this, we exploit the advantageous compute and +memory scaling of online synaptic plasticity by providing an inherently sparse +implementation of AD where expensive tensor contractions are replaced with +simple element-wise multiplications if the tensors are diagonal. Gradient-based +synaptic plasticity rules such as eligibility propagation (e-prop) have exactly +this property and thus profit immensely from this feature. We demonstrate the +alignment of our gradients with respect to gradient backpropagation on an +synthetic task where e-prop gradients are exact, as well as audio speech +classification benchmarks. We demonstrate how memory utilization scales with +network size without dependence on the sequence length, as expected from +forward AD methods. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the mandatory criteria (1-3) as it focuses on synaptic plasticity and neural networks rather than Large Language Models (LLMs), and does not discuss practical applications, experimental results, or comparisons to state-of-the-art in the context of LLMs. Additionally, it does not align with any of the preferred criteria (4-5) related to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Generalization and Informativeness of Weighted Conformal Risk Control + Under Covariate Shift](https://arxiv.org/abs/2501.11413v1) +**arXiv ID:** 2501.11413v1 + +**Abstract:** +> Predictive models are often required to produce reliable predictions under +statistical conditions that are not matched to the training data. A common type +of training-testing mismatch is covariate shift, where the conditional +distribution of the target variable given the input features remains fixed, +while the marginal distribution of the inputs changes. Weighted conformal risk +control (W-CRC) uses data collected during the training phase to convert point +predictions into prediction sets with valid risk guarantees at test time +despite the presence of a covariate shift. However, while W-CRC provides +statistical reliability, its efficiency -- measured by the size of the +prediction sets -- can only be assessed at test time. In this work, we relate +the generalization properties of the base predictor to the efficiency of W-CRC +under covariate shifts. Specifically, we derive a bound on the inefficiency of +the W-CRC predictor that depends on algorithmic hyperparameters and +task-specific quantities available at training time. This bound offers insights +on relationships between the informativeness of the prediction sets, the extent +of the covariate shift, and the size of the calibration and training sets. +Experiments on fingerprinting-based localization validate the theoretical +results. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, focusing on predictive model reliability under covariate shift without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and lacking clear connections to practical applications of LLMs as specified in the criteria. + +--- + +## [Enhancing Coronary Artery Calcium Scoring via Multi-Organ Segmentation + on Non-Contrast Cardiac Computed Tomography](https://arxiv.org/abs/2501.11428v1) +**arXiv ID:** 2501.11428v1 + +**Abstract:** +> Despite coronary artery calcium scoring being considered a largely solved +problem within the realm of medical artificial intelligence, this paper argues +that significant improvements can still be made. By shifting the focus from +pathology detection to a deeper understanding of anatomy, the novel algorithm +proposed in the paper both achieves high accuracy in coronary artery calcium +scoring and offers enhanced interpretability of the results. This approach not +only aids in the precise quantification of calcifications in coronary arteries, +but also provides valuable insights into the underlying anatomical structures. +Through this anatomically-informed methodology, the paper shows how a nuanced +understanding of the heart's anatomy can lead to more accurate and +interpretable results in the field of cardiovascular health. We demonstrate the +superior accuracy of the proposed method by evaluating it on an open-source +multi-vendor dataset, where we obtain results at the inter-observer level, +surpassing the current state of the art. Finally, the qualitative analyses show +the practical value of the algorithm in such tasks as labeling coronary artery +calcifications, identifying aortic calcifications, and filtering out false +positive detections due to noise. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, specifically cardiovascular health, which is explicitly mentioned as a rejection criterion. + +--- + +## [Decomposing Interventional Causality into Synergistic, Redundant, and + Unique Components](https://arxiv.org/abs/2501.11447v1) +**arXiv ID:** 2501.11447v1 + +**Abstract:** +> We introduce a novel framework for decomposing interventional causal effects +into synergistic, redundant, and unique components, building on the intuition +of Partial Information Decomposition (PID) and the principle of M\"obius +inversion. While recent work has explored a similar decomposition of an +observational measure, we argue that a proper causal decomposition must be +interventional in nature. We develop a mathematical approach that +systematically quantifies how causal power is distributed among variables in a +system, using a recently derived closed-form expression for the M\"obius +function of the redundancy lattice. The formalism is then illustrated by +decomposing the causal power in logic gates, cellular automata, and chemical +reaction networks. Our results reveal how the distribution of causal power can +be context- and parameter-dependent. This decomposition provides new insights +into complex systems by revealing how causal influences are shared and combined +among multiple variables, with potential applications ranging from attribution +of responsibility in legal or AI systems, to the analysis of biological +networks or climate models. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the mandatory criteria (1-3) as it lacks focus on Large Language Models (LLMs), experimental results with quantitative metrics, and comparison with state-of-the-art techniques. Additionally, its primary focus on causal decomposition in complex systems, with potential applications in legal systems, does not align with the criteria, and its potential application in AI systems is too broad and not specifically related to LLMs or the specified areas. + +--- + +## [Generative AI and Large Language Models in Language Preservation: + Opportunities and Challenges](https://arxiv.org/abs/2501.11496v1) +**arXiv ID:** 2501.11496v1 + +**Abstract:** +> Generative AI and large-scale language models (LLM) have emerged as powerful +tools in language preservation, particularly for near-native and endangered +languages. With the increasing reliance on technology for communication, +education, and cultural documentation, new opportunities have emerged to +mitigate the dramatic decline of linguistic diversity worldwide. This paper +examines the role of generative AIs and LLMs in preserving endangered +languages, highlighting the risks and challenges associated with their use. We +analyze the underlying technologies driving these models, including natural +language processing (NLP) and deep learning, and explore several cases where +these technologies have been applied to low-resource languages. Additionally, +we discuss ethical considerations, data scarcity issues, and technical +challenges while proposing solutions to enhance AI-driven language +preservation. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on social applications of AI in regard to linguistic diversity and cultural preservation, which aligns with rejected criteria related to social applications of AI, warranting rejection despite potential novelty in language preservation. + +--- + +## [Dialect2SQL: A Novel Text-to-SQL Dataset for Arabic Dialects with a + Focus on Moroccan Darija](https://arxiv.org/abs/2501.11498v1) +**arXiv ID:** 2501.11498v1 + +**Abstract:** +> The task of converting natural language questions (NLQs) into executable SQL +queries, known as text-to-SQL, has gained significant interest in recent years, +as it enables non-technical users to interact with relational databases. Many +benchmarks, such as SPIDER and WikiSQL, have contributed to the development of +new models and the evaluation of their performance. In addition, other +datasets, like SEDE and BIRD, have introduced more challenges and complexities +to better map real-world scenarios. However, these datasets primarily focus on +high-resource languages such as English and Chinese. In this work, we introduce +Dialect2SQL, the first large-scale, cross-domain text-to-SQL dataset in an +Arabic dialect. It consists of 9,428 NLQ-SQL pairs across 69 databases in +various domains. Along with SQL-related challenges such as long schemas, dirty +values, and complex queries, our dataset also incorporates the complexities of +the Moroccan dialect, which is known for its diverse source languages, numerous +borrowed words, and unique expressions. This demonstrates that our dataset will +be a valuable contribution to both the text-to-SQL community and the +development of resources for low-resource languages. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on developing a text-to-SQL dataset for an Arabic dialect, without explicit mention of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the mandatory criteria. + +--- + +## [Technical Report for the Forgotten-by-Design Project: Targeted + Obfuscation for Machine Learning](https://arxiv.org/abs/2501.11525v1) +**arXiv ID:** 2501.11525v1 + +**Abstract:** +> The right to privacy, enshrined in various human rights declarations, faces +new challenges in the age of artificial intelligence (AI). This paper explores +the concept of the Right to be Forgotten (RTBF) within AI systems, contrasting +it with traditional data erasure methods. We introduce Forgotten by Design, a +proactive approach to privacy preservation that integrates instance-specific +obfuscation techniques during the AI model training process. Unlike machine +unlearning, which modifies models post-training, our method prevents sensitive +data from being embedded in the first place. Using the LIRA membership +inference attack, we identify vulnerable data points and propose defenses that +combine additive gradient noise and weighting schemes. Our experiments on the +CIFAR-10 dataset demonstrate that our techniques reduce privacy risks by at +least an order of magnitude while maintaining model accuracy (at 95% +significance). Additionally, we present visualization methods for the +privacy-utility trade-off, providing a clear framework for balancing privacy +risk and model accuracy. This work contributes to the development of +privacy-preserving AI systems that align with human cognitive processes of +motivated forgetting, offering a robust framework for safeguarding sensitive +information and ensuring compliance with privacy regulations. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on responsible AI application (privacy preservation and compliance with privacy regulations), which is explicitly listed as a rejection criterion, outweighing its marginal relevance to Large Language Models (LLMs) through its broader AI context. + +--- + +## [Training-free Ultra Small Model for Universal Sparse Reconstruction in + Compressed Sensing](https://arxiv.org/abs/2501.11592v1) +**arXiv ID:** 2501.11592v1 + +**Abstract:** +> Pre-trained large models attract widespread attention in recent years, but +they face challenges in applications that require high interpretability or have +limited resources, such as physical sensing, medical imaging, and +bioinformatics. Compressed Sensing (CS) is a well-proved theory that drives +many recent breakthroughs in these applications. However, as a typical +under-determined linear system, CS suffers from excessively long sparse +reconstruction times when using traditional iterative methods, particularly +with large-scale data. Current AI methods like deep unfolding fail to +substitute them because pre-trained models exhibit poor generality beyond their +training conditions and dataset distributions, or lack interpretability. +Instead of following the big model fervor, this paper proposes ultra-small +artificial neural models called coefficients learning (CL), enabling +training-free and rapid sparse reconstruction while perfectly inheriting the +generality and interpretability of traditional iterative methods, bringing new +feature of incorporating prior knowledges. In CL, a signal of length $n$ only +needs a minimal of $n$ trainable parameters. A case study model called CLOMP is +implemented for evaluation. Experiments are conducted on both synthetic and +real one-dimensional and two-dimensional signals, demonstrating significant +improvements in efficiency and accuracy. Compared to representative iterative +methods, CLOMP improves efficiency by 100 to 1000 folds for large-scale data. +Test results on eight diverse image datasets indicate that CLOMP improves +structural similarity index by 292%, 98%, 45% for sampling rates of 0.1, 0.3, +0.5, respectively. We believe this method can truly usher CS reconstruction +into the AI era, benefiting countless under-determined linear systems that rely +on sparse solution. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Compressed Sensing for applications in physical sensing, medical imaging, and bioinformatics, which aligns with excluded topics (medical applications of AI and not explicitly focusing on Large Language Models or related areas like knowledge graphs, RAG, or agentic AI). + +--- + +## [Fairness Testing through Extreme Value Theory](https://arxiv.org/abs/2501.11597v1) +**arXiv ID:** 2501.11597v1 + +**Abstract:** +> Data-driven software is increasingly being used as a critical component of +automated decision-support systems. Since this class of software learns its +logic from historical data, it can encode or amplify discriminatory practices. +Previous research on algorithmic fairness has focused on improving average-case +fairness. On the other hand, fairness at the extreme ends of the spectrum, +which often signifies lasting and impactful shifts in societal attitudes, has +received significantly less emphasis. + Leveraging the statistics of extreme value theory (EVT), we propose a novel +fairness criterion called extreme counterfactual discrimination (ECD). This +criterion estimates the worst-case amounts of disadvantage in outcomes for +individuals solely based on their memberships in a protected group. Utilizing +tools from search-based software engineering and generative AI, we present a +randomized algorithm that samples a statistically significant set of points +from the tail of ML outcome distributions even if the input dataset lacks a +sufficient number of relevant samples. + We conducted several experiments on four ML models (deep neural networks, +logistic regression, and random forests) over 10 socially relevant tasks from +the literature on algorithmic fairness. First, we evaluate the generative AI +methods and find that they generate sufficient samples to infer valid EVT +distribution in 95% of cases. Remarkably, we found that the prevalent bias +mitigators reduce the average-case discrimination but increase the worst-case +discrimination significantly in 5% of cases. We also observed that even the +tail-aware mitigation algorithm -- MiniMax-Fairness -- increased the worst-case +discrimination in 30% of cases. We propose a novel ECD-based mitigator that +improves fairness in the tail in 90% of cases with no degradation of the +average-case discrimination. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on fairness and bias in AI decision-making systems, which aligns with "responsible AI application or AI ethics", a criterion for rejection. It also lacks clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI as specified in the selection criteria. + +--- + +## [Noise-Agnostic Multitask Whisper Training for Reducing False Alarm + Errors in Call-for-Help Detection](https://arxiv.org/abs/2501.11631v1) +**arXiv ID:** 2501.11631v1 + +**Abstract:** +> Keyword spotting is often implemented by keyword classifier to the encoder in +acoustic models, enabling the classification of predefined or open vocabulary +keywords. Although keyword spotting is a crucial task in various applications +and can be extended to call-for-help detection in emergencies, however, the +previous method often suffers from scalability limitations due to retraining +required to introduce new keywords or adapt to changing contexts. We explore a +simple yet effective approach that leverages off-the-shelf pretrained ASR +models to address these challenges, especially in call-for-help detection +scenarios. Furthermore, we observed a substantial increase in false alarms when +deploying call-for-help detection system in real-world scenarios due to noise +introduced by microphones or different environments. To address this, we +propose a novel noise-agnostic multitask learning approach that integrates a +noise classification head into the ASR encoder. Our method enhances the model's +robustness to noisy environments, leading to a significant reduction in false +alarms and improved overall call-for-help performance. Despite the added +complexity of multitask learning, our approach is computationally efficient and +provides a promising solution for call-for-help detection in real-world +scenarios. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on acoustic models, automatic speech recognition (ASR), and noise reduction in call-for-help detection, which does not align with the specified criteria emphasizing Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. Its applications and methodologies fall outside the designated scope." +} +``` + +--- + +## [Biomedical Knowledge Graph: A Survey of Domains, Tasks, and Real-World + Applications](https://arxiv.org/abs/2501.11632v1) +**arXiv ID:** 2501.11632v1 + +**Abstract:** +> Biomedical knowledge graphs (BKGs) have emerged as powerful tools for +organizing and leveraging the vast and complex data found across the biomedical +field. Yet, current reviews of BKGs often limit their scope to specific domains +or methods, overlooking the broader landscape and the rapid technological +progress reshaping it. In this survey, we address this gap by offering a +systematic review of BKGs from three core perspectives: domains, tasks, and +applications. We begin by examining how BKGs are constructed from diverse data +sources, including molecular interactions, pharmacological datasets, and +clinical records. Next, we discuss the essential tasks enabled by BKGs, +focusing on knowledge management, retrieval, reasoning, and interpretation. +Finally, we highlight real-world applications in precision medicine, drug +discovery, and scientific research, illustrating the translational impact of +BKGs across multiple sectors. By synthesizing these perspectives into a unified +framework, this survey not only clarifies the current state of BKG research but +also establishes a foundation for future exploration, enabling both innovative +methodological advances and practical implementations. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on biomedical applications, which is excluded according to the rejection criteria, specifically the focus on medical applications of AI. + +--- + +## [Transformer Vibration Forecasting for Advancing Rail Safety and + Maintenance 4.0](https://arxiv.org/abs/2501.11730v1) +**arXiv ID:** 2501.11730v1 + +**Abstract:** +> Maintaining railway axles is critical to preventing severe accidents and +financial losses. The railway industry is increasingly interested in advanced +condition monitoring techniques to enhance safety and efficiency, moving beyond +traditional periodic inspections toward Maintenance 4.0. + This study introduces a robust Deep Autoregressive solution that integrates +seamlessly with existing systems to avert mechanical failures. Our approach +simulates and predicts vibration signals under various conditions and fault +scenarios, improving dataset robustness for more effective detection systems. +These systems can alert maintenance needs, preventing accidents preemptively. +We use experimental vibration signals from accelerometers on train axles. + Our primary contributions include a transformer model, ShaftFormer, designed +for processing time series data, and an alternative model incorporating +spectral methods and enhanced observation models. Simulating vibration signals +under diverse conditions mitigates the high cost of obtaining experimental +signals for all scenarios. Given the non-stationary nature of railway vibration +signals, influenced by speed and load changes, our models address these +complexities, offering a powerful tool for predictive maintenance in the rail +industry. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on predictive maintenance in the rail industry, utilizing a transformer model for vibration forecasting, but does not indicate the use of Large Language Models (LLMs) as required by the selection criteria, nor does it mention applications in knowledge graphs, retrieval-augmented generation, or agentic AI." +} +``` + +--- + +## [SILO: Solving Inverse Problems with Latent Operators](https://arxiv.org/abs/2501.11746v1) +**arXiv ID:** 2501.11746v1 + +**Abstract:** +> Consistent improvement of image priors over the years has led to the +development of better inverse problem solvers. Diffusion models are the +newcomers to this arena, posing the strongest known prior to date. Recently, +such models operating in a latent space have become increasingly predominant +due to their efficiency. In recent works, these models have been applied to +solve inverse problems. Working in the latent space typically requires multiple +applications of an Autoencoder during the restoration process, which leads to +both computational and restoration quality challenges. In this work, we propose +a new approach for handling inverse problems with latent diffusion models, +where a learned degradation function operates within the latent space, +emulating a known image space degradation. Usage of the learned operator +reduces the dependency on the Autoencoder to only the initial and final steps +of the restoration process, facilitating faster sampling and superior +restoration quality. We demonstrate the effectiveness of our method on a +variety of image restoration tasks and datasets, achieving significant +improvements over prior art. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on image restoration tasks using diffusion models and autoencoders, with no evident connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Policy-Adaptable Methods For Resolving Normative Conflicts Through + Argumentation and Graph Colouring](https://arxiv.org/abs/2501.11799v1) +**arXiv ID:** 2501.11799v1 + +**Abstract:** +> In a multi-agent system, one may choose to govern the behaviour of an agent +by imposing norms, which act as guidelines for how agents should act either all +of the time or in given situations. However, imposing multiple norms on one or +more agents may result in situations where these norms conflict over how the +agent should behave. In any system with normative conflicts (such as safe +reinforcement models or systems which monitor safety protocols), one must +decide which norms should be followed such that the most important and most +relevant norms are maintained. We introduce a new method for resolving +normative conflicts through argumentation and graph colouring which is +compatible with a variety of normative conflict resolution policies. We prove +that this method always creates an admissible set of arguments under +argumentation semantics, meaning that it produces coherent outputs. We also +introduce more robust variants of this method, each building upon their +predecessor to create a superior output, and we include further mathematical +proof of their coherence. Our most advanced variant uses the existing concept +of curtailment, where one norm may supersede another without fully eliminating +it. The methods we introduce are all compatible with various pre-existing +policies for resolving normative conflicts. Empirical evaluations are also +performed to compare our algorithms to each other and to others in existing +literature. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper does not explicitly mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the primary focus areas. It focuses on normative conflicts resolution in multi-agent systems using argumentation and graph coloring, which doesn't meet the mandatory criteria (1, 2, and 3) and doesn't explicitly align with any of the additional considerations." +} +``` + +--- + +## [Toward Effective Digraph Representation Learning: A Magnetic Adaptive + Propagation based Approach](https://arxiv.org/abs/2501.11817v1) +**arXiv ID:** 2501.11817v1 + +**Abstract:** +> The $q$-parameterized magnetic Laplacian serves as the foundation of directed +graph (digraph) convolution, enabling this kind of digraph neural network +(MagDG) to encode node features and structural insights by complex-domain +message passing. As a generalization of undirected methods, MagDG shows +superior capability in modeling intricate web-scale topology. Despite the great +success achieved by existing MagDGs, limitations still exist: (1) Hand-crafted +$q$: The performance of MagDGs depends on selecting an appropriate +$q$-parameter to construct suitable graph propagation equations in the complex +domain. This parameter tuning, driven by downstream tasks, limits model +flexibility and significantly increases manual effort. (2) Coarse Message +Passing: Most approaches treat all nodes with the same complex-domain +propagation and aggregation rules, neglecting their unique digraph contexts. +This oversight results in sub-optimal performance. To address the above issues, +we propose two key techniques: (1) MAP is crafted to be a plug-and-play +complex-domain propagation optimization strategy in the context of digraph +learning, enabling seamless integration into any MagDG to improve predictions +while enjoying high running efficiency. (2) MAP++ is a new digraph learning +framework, further incorporating a learnable mechanism to achieve adaptively +edge-wise propagation and node-wise aggregation in the complex domain for +better performance. Extensive experiments on 12 datasets demonstrate that MAP +enjoys flexibility for it can be incorporated with any MagDG, and scalability +as it can deal with web-scale digraphs. MAP++ achieves SOTA predictive +performance on 4 different downstream tasks. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on digraph representation learning and graph neural networks, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria related to LLMs. + +--- + +## [Toward Scalable Graph Unlearning: A Node Influence Maximization based + Approach](https://arxiv.org/abs/2501.11823v1) +**arXiv ID:** 2501.11823v1 + +**Abstract:** +> Machine unlearning, as a pivotal technology for enhancing model robustness +and data privacy, has garnered significant attention in prevalent web mining +applications, especially in thriving graph-based scenarios. However, most +existing graph unlearning (GU) approaches face significant challenges due to +the intricate interactions among web-scale graph elements during the model +training: (1) The gradient-driven node entanglement hinders the complete +knowledge removal in response to unlearning requests; (2) The billion-level +graph elements in the web scenarios present inevitable scalability issues. To +break the above limitations, we open up a new perspective by drawing a +connection between GU and conventional social influence maximization. To this +end, we propose Node Influence Maximization (NIM) through the decoupled +influence propagation model and fine-grained influence function in a scalable +manner, which is crafted to be a plug-and-play strategy to identify potential +nodes affected by unlearning entities. This approach enables offline execution +independent of GU, allowing it to be seamlessly integrated into most GU methods +to improve their unlearning performance. Based on this, we introduce Scalable +Graph Unlearning (SGU) as a new fine-tuned framework, which balances the +forgetting and reasoning capability of the unlearned model by entity-specific +optimizations. Extensive experiments on 14 datasets, including large-scale +ogbn-papers100M, have demonstrated the effectiveness of our approach. +Specifically, NIM enhances the forgetting capability of most GU methods, while +SGU achieves comprehensive SOTA performance and maintains scalability. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria as it does not focus on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. Instead, it centers on graph unlearning, machine learning robustness, and data privacy, without apparent connection to LLMs or specified areas of interest. + +--- + +## [Data-driven Detection and Evaluation of Damages in Concrete Structures: + Using Deep Learning and Computer Vision](https://arxiv.org/abs/2501.11836v1) +**arXiv ID:** 2501.11836v1 + +**Abstract:** +> Structural integrity is vital for maintaining the safety and longevity of +concrete infrastructures such as bridges, tunnels, and walls. Traditional +methods for detecting damages like cracks and spalls are labor-intensive, +time-consuming, and prone to human error. To address these challenges, this +study explores advanced data-driven techniques using deep learning for +automated damage detection and analysis. Two state-of-the-art instance +segmentation models, YOLO-v7 instance segmentation and Mask R-CNN, were +evaluated using a dataset comprising 400 images, augmented to 10,995 images +through geometric and color-based transformations to enhance robustness. The +models were trained and validated using a dataset split into 90% training set, +validation and test set 10%. Performance metrics such as precision, recall, +mean average precision (mAP@0.5), and frames per second (FPS) were used for +evaluation. YOLO-v7 achieved a superior mAP@0.5 of 96.1% and processed 40 FPS, +outperforming Mask R-CNN, which achieved a mAP@0.5 of 92.1% with a slower +processing speed of 18 FPS. The findings recommend YOLO-v7 instance +segmentation model for real-time, high-speed structural health monitoring, +while Mask R-CNN is better suited for detailed offline assessments. This study +demonstrates the potential of deep learning to revolutionize infrastructure +maintenance, offering a scalable and efficient solution for automated damage +detection. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria as it primarily focuses on Computer Vision and Deep Learning for infrastructure maintenance, without any mention of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest. + +--- + +## [Supervised Learning for Analog and RF Circuit Design: Benchmarks and + Comparative Insights](https://arxiv.org/abs/2501.11839v1) +**arXiv ID:** 2501.11839v1 + +**Abstract:** +> Automating analog and radio-frequency (RF) circuit design using machine +learning (ML) significantly reduces the time and effort required for parameter +optimization. This study explores supervised ML-based approaches for designing +circuit parameters from performance specifications across various circuit +types, including homogeneous and heterogeneous designs. By evaluating diverse +ML models, from neural networks like transformers to traditional methods like +random forests, we identify the best-performing models for each circuit. Our +results show that simpler circuits, such as low-noise amplifiers, achieve +exceptional accuracy with mean relative errors as low as 0.3% due to their +linear parameter-performance relationships. In contrast, complex circuits, like +power amplifiers and voltage-controlled oscillators, present challenges due to +their non-linear interactions and larger design spaces. For heterogeneous +circuits, our approach achieves an 88% reduction in errors with increased +training data, with the receiver achieving a mean relative error as low as +0.23%, showcasing the scalability and accuracy of the proposed methodology. +Additionally, we provide insights into model strengths, with transformers +excelling in capturing non-linear mappings and k-nearest neighbors performing +robustly in moderately linear parameter spaces, especially in heterogeneous +circuits with larger datasets. This work establishes a foundation for extending +ML-driven design automation, enabling more efficient and scalable circuit +design workflows. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria for Large Language Models (LLMs) applications, instead focusing on machine learning (ML) for analog and RF circuit design, with no apparent connection to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Scalable Whole Slide Image Representation Using K-Mean Clustering and + Fisher Vector Aggregation](https://arxiv.org/abs/2501.12085v1) +**arXiv ID:** 2501.12085v1 + +**Abstract:** +> Whole slide images (WSIs) are high-resolution, gigapixel sized images that +pose significant computational challenges for traditional machine learning +models due to their size and heterogeneity.In this paper, we present a scalable +and efficient methodology for WSI classification by leveraging patch-based +feature extraction, clustering, and Fisher vector encoding. Initially, WSIs are +divided into fixed size patches, and deep feature embeddings are extracted from +each patch using a pre-trained convolutional neural network (CNN). These +patch-level embeddings are subsequently clustered using K-means clustering, +where each cluster aggregates semantically similar regions of the WSI. To +effectively summarize each cluster, Fisher vector representations are computed +by modeling the distribution of patch embeddings in each cluster as a +parametric Gaussian mixture model (GMM). The Fisher vectors from each cluster +are concatenated into a high-dimensional feature vector, creating a compact and +informative representation of the entire WSI. This feature vector is then used +by a classifier to predict the WSI's diagnostic label. Our method captures +local and global tissue structures and yields robust performance for +large-scale WSI classification, demonstrating superior accuracy and scalability +compared to other approaches. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper does not meet the primary criteria as it focuses on image processing (Whole Slide Image Representation) and machine learning models (CNN, K-Mean Clustering), with no mention of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest." +} +``` + +--- + +## [Proxies for Distortion and Consistency with Applications for Real-World + Image Restoration](https://arxiv.org/abs/2501.12102v1) +**arXiv ID:** 2501.12102v1 + +**Abstract:** +> Real-world image restoration deals with the recovery of images suffering from +an unknown degradation. This task is typically addressed while being given only +degraded images, without their corresponding ground-truth versions. In this +hard setting, designing and evaluating restoration algorithms becomes highly +challenging. This paper offers a suite of tools that can serve both the design +and assessment of real-world image restoration algorithms. Our work starts by +proposing a trained model that predicts the chain of degradations a given +real-world measured input has gone through. We show how this estimator can be +used to approximate the consistency -- the match between the measurements and +any proposed recovered image. We also use this estimator as a guiding force for +the design of a simple and highly-effective plug-and-play real-world image +restoration algorithm, leveraging a pre-trained diffusion-based image prior. +Furthermore, this work proposes no-reference proxy measures of MSE and LPIPS, +which, without access to the ground-truth images, allow ranking of real-world +image restoration algorithms according to their (approximate) MSE and LPIPS. +The proposed suite provides a versatile, first of its kind framework for +evaluating and comparing blind image restoration algorithms in real-world +scenarios. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on image restoration, which falls under video/image processing, one of the excluded areas as per the evaluation criteria. No apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI is mentioned in the abstract. + +--- + +## [Efficient PINNs: Multi-Head Unimodular Regularization of the Solutions + Space](https://arxiv.org/abs/2501.12116v1) +**arXiv ID:** 2501.12116v1 + +**Abstract:** +> We present a machine learning framework to facilitate the solution of +nonlinear multiscale differential equations and, especially, inverse problems +using Physics-Informed Neural Networks (PINNs). This framework is based on what +is called multihead (MH) training, which involves training the network to learn +a general space of all solutions for a given set of equations with certain +variability, rather than learning a specific solution of the system. This setup +is used with a second novel technique that we call Unimodular Regularization +(UR) of the latent space of solutions. We show that the multihead approach, +combined with the regularization, significantly improves the efficiency of +PINNs by facilitating the transfer learning process thereby enabling the +finding of solutions for nonlinear, coupled, and multiscale differential +equations. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the Practical Applications criterion related to Large Language Models (LLMs), as it focuses on Physics-Informed Neural Networks (PINNs) for solving nonlinear differential equations, without any apparent connection to LLMs, knowledge graphs, RAG, or agentic AI. + +--- + +## [On the practical applicability of modern DFT functionals for chemical + computations. Case study of DM21 applicability for geometry optimization](https://arxiv.org/abs/2501.12149v1) +**arXiv ID:** 2501.12149v1 + +**Abstract:** +> Density functional theory (DFT) is probably the most promising approach for +quantum chemistry calculations considering its good balance between +calculations precision and speed. In recent years, several neural network-based +functionals have been developed for exchange-correlation energy approximation +in DFT, DM21 developed by Google Deepmind being the most notable between them. +This study focuses on evaluating the efficiency of DM21 functional in +predicting molecular geometries, with a focus on the influence of oscillatory +behavior in neural network exchange-correlation functionals. We implemented +geometry optimization in PySCF for the DM21 functional in geometry optimization +problem, compared its performance with traditional functionals, and tested it +on various benchmarks. Our findings reveal both the potential and the current +challenges of using neural network functionals for geometry optimization in +DFT. We propose a solution extending the practical applicability of such +functionals and allowing to model new substances with their help. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on chemical computations using Density Functional Theory (DFT) and neural network-based functionals, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [An End-to-End Approach for Korean Wakeword Systems with Speaker + Authentication](https://arxiv.org/abs/2501.12194v1) +**arXiv ID:** 2501.12194v1 + +**Abstract:** +> Wakeword detection plays a critical role in enabling AI assistants to listen +to user voices and interact effectively. However, for languages other than +English, there is a significant lack of pre-trained wakeword models. +Additionally, systems that merely determine the presence of a wakeword can pose +serious privacy concerns. In this paper, we propose an end-to-end approach that +trains wakewords for Non-English languages, particulary Korean, and uses this +to develop a Voice Authentication model to protect user privacy. Our +implementation employs an open-source platform OpenWakeWord, which performs +wakeword detection using an FCN (Fully-Connected Network) architecture. Once a +wakeword is detected, our custom-developed code calculates cosine similarity +for robust user authentication. Experimental results demonstrate the +effectiveness of our approach, achieving a 16.79% and a 6.6% Equal Error Rate +(EER) each in the Wakeword Detection and the Voice Authentication. These +findings highlight the model's potential in providing secure and accurate +wakeword detection and authentication for Korean users. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on voice processing and speaker authentication, which, while related to AI assistants, does not explicitly involve Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI, thus not meeting the core criteria." +} +``` + +--- + +## [Strong phonon-mediated high temperature superconductivity in + Li$_2$AuH$_6$ under ambient pressure](https://arxiv.org/abs/2501.12222v1) +**arXiv ID:** 2501.12222v1 + +**Abstract:** +> We used our developed AI search engine~(InvDesFlow) to perform extensive +investigations regarding ambient stable superconducting hydrides. A cubic +structure Li$_2$AuH$_6$ with Au-H octahedral motifs is identified to be a +candidate. After performing thermodynamical analysis, we provide a feasible +route to experimentally synthesize this material via the known LiAu and LiH +compounds under ambient pressure. The further first-principles calculations +suggest that Li$_2$AuH$_6$ shows a high superconducting transition temperature +($T_c$) $\sim$ 140 K under ambient pressure. The H-1$s$ electrons strongly +couple with phonon modes of vibrations of Au-H octahedrons as well as +vibrations of Li atoms, where the latter is not taken seriously in other +previously similar cases. Hence, different from previous claims of searching +metallic covalent bonds to find high-$T_c$ superconductors, we emphasize here +the importance of those phonon modes with strong electron-phonon coupling +(EPC). And we suggest that one can intercalate atoms into binary or ternary +hydrides to introduce more potential phonon modes with strong EPC, which is an +effective approach to find high-$T_c$ superconductors within multicomponent +compounds. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, as it primarily focuses on materials science (superconductivity) with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. The mention of an 'AI search engine' is incidental and not central to the paper's main contribution. + +--- + +## [CBVLM: Training-free Explainable Concept-based Large Vision Language + Models for Medical Image Classification](https://arxiv.org/abs/2501.12266v1) +**arXiv ID:** 2501.12266v1 + +**Abstract:** +> The main challenges limiting the adoption of deep learning-based solutions in +medical workflows are the availability of annotated data and the lack of +interpretability of such systems. Concept Bottleneck Models (CBMs) tackle the +latter by constraining the final disease prediction on a set of predefined and +human-interpretable concepts. However, the increased interpretability achieved +through these concept-based explanations implies a higher annotation burden. +Moreover, if a new concept needs to be added, the whole system needs to be +retrained. Inspired by the remarkable performance shown by Large +Vision-Language Models (LVLMs) in few-shot settings, we propose a simple, yet +effective, methodology, CBVLM, which tackles both of the aforementioned +challenges. First, for each concept, we prompt the LVLM to answer if the +concept is present in the input image. Then, we ask the LVLM to classify the +image based on the previous concept predictions. Moreover, in both stages, we +incorporate a retrieval module responsible for selecting the best examples for +in-context learning. By grounding the final diagnosis on the predicted +concepts, we ensure explainability, and by leveraging the few-shot capabilities +of LVLMs, we drastically lower the annotation cost. We validate our approach +with extensive experiments across four medical datasets and twelve LVLMs (both +generic and medical) and show that CBVLM consistently outperforms CBMs and +task-specific supervised methods without requiring any training and using just +a few annotated examples. More information on our project page: +https://cristianopatricio.github.io/CBVLM/. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, which is explicitly listed as a criterion for rejection, outweighing its potential alignment with other criteria such as practical applications, experimental results, and novelty. + +--- + +## [Implementation of an Asymmetric Adjusted Activation Function for Class + Imbalance Credit Scoring](https://arxiv.org/abs/2501.12285v1) +**arXiv ID:** 2501.12285v1 + +**Abstract:** +> Credit scoring is a systematic approach to evaluate a borrower's probability +of default (PD) on a bank loan. The data associated with such scenarios are +characteristically imbalanced, complicating binary classification owing to the +often-underestimated cost of misclassification during the classifier's learning +process. Considering the high imbalance ratio (IR) of these datasets, we +introduce an innovative yet straightforward optimized activation function by +incorporating an IR-dependent asymmetric adjusted factor embedded Sigmoid +activation function (ASIG). The embedding of ASIG makes the sensitive margin of +the Sigmoid function auto-adjustable, depending on the imbalance nature of the +datasets distributed, thereby giving the activation function an asymmetric +characteristic that prevents the underrepresentation of the minority class +(positive samples) during the classifier's learning process. The experimental +results show that the ASIG-embedded-classifier outperforms traditional +classifiers on datasets across wide-ranging IRs in the downstream +credit-scoring task. The algorithm also shows robustness and stability, even +when the IR is ultra-high. Therefore, the algorithm provides a competitive +alternative in the financial industry, especially in credit scoring, possessing +the ability to effectively process highly imbalanced distribution data. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on a financial application (credit scoring) with no evident connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the essential criteria for practical applications of LLMs. + +--- + +## [Regressor-Guided Image Editing Regulates Emotional Response to Reduce + Online Engagement](https://arxiv.org/abs/2501.12289v1) +**arXiv ID:** 2501.12289v1 + +**Abstract:** +> Emotions are known to mediate the relationship between users' content +consumption and their online engagement, with heightened emotional intensity +leading to increased engagement. Building on this insight, we propose three +regressor-guided image editing approaches aimed at diminishing the emotional +impact of images. These include (i) a parameter optimization approach based on +global image transformations known to influence emotions, (ii) an optimization +approach targeting the style latent space of a generative adversarial network, +and (iii) a diffusion-based approach employing classifier guidance and +classifier-free guidance. Our findings demonstrate that approaches can +effectively alter the emotional properties of images while maintaining high +visual quality. Optimization-based methods primarily adjust low-level +properties like color hues and brightness, whereas the diffusion-based approach +introduces semantic changes, such as altering appearance or facial expressions. +Notably, results from a behavioral study reveal that only the diffusion-based +approach successfully elicits changes in viewers' emotional responses while +preserving high perceived image quality. In future work, we will investigate +the impact of these image adaptations on internet user behavior. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on image processing and emotional response regulation, which aligns with video/image processing, and does not evidently involve Large Language Models (LLMs) or related areas like knowledge graphs, RAG, or agentic AI, failing to meet the required criteria." +} +``` + +--- + +## [DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial + Basis Functions](https://arxiv.org/abs/2501.12369v1) +**arXiv ID:** 2501.12369v1 + +**Abstract:** +> Splatting-based 3D reconstruction methods have gained popularity with the +advent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel +views. These methods commonly resort to using exponential family functions, +such as the Gaussian function, as reconstruction kernels due to their +anisotropic nature, ease of projection, and differentiability in rasterization. +However, the field remains restricted to variations within the exponential +family, leaving generalized reconstruction kernels largely underexplored, +partly due to the lack of easy integrability in 3D to 2D projections. In this +light, we show that a class of decaying anisotropic radial basis functions +(DARBFs), which are non-negative functions of the Mahalanobis distance, +supports splatting by approximating the Gaussian function's closed-form +integration advantage. With this fresh perspective, we demonstrate up to 34% +faster convergence during training and a 15% reduction in memory consumption +across various DARB reconstruction kernels, while maintaining comparable PSNR, +SSIM, and LPIPS results. We will make the code available. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on 3D reconstruction and splatting with radial basis functions, which falls under video/image processing, and does not involve Large Language Models (LLMs) or related areas specified in the criteria." +} + +--- + +## [Expertise elevates AI usage: experimental evidence comparing laypeople + and professional artists](https://arxiv.org/abs/2501.12374v1) +**arXiv ID:** 2501.12374v1 + +**Abstract:** +> Novel capacities of generative AI to analyze and generate cultural artifacts +raise inevitable questions about the nature and value of artistic education and +human expertise. Has AI already leveled the playing field between professional +artists and laypeople, or do trained artistic expressive capacity, curation +skills and experience instead enhance the ability to use these new tools? In +this pre-registered study, we conduct experimental comparisons between 50 +active artists and a demographically matched sample of laypeople. We designed +two tasks to approximate artistic practice for testing their capabilities in +both faithful and creative image creation: replicating a reference image, and +moving as far away as possible from it. We developed a bespoke platform where +participants used a modern text-to-image model to complete both tasks. We also +collected and compared participants' sentiments towards AI. On average, artists +produced more faithful and creative outputs than their lay counterparts, +although only by a small margin. While AI may ease content creation, +professional expertise is still valuable - even within the confined space of +generative AI itself. Finally, we also explored how well an exemplary +vision-capable large language model (GPT-4o) would complete the same tasks, if +given the role of an image generation agent, and found it performed on par in +copying but outperformed even artists in the creative task. The very best +results were still produced by humans in both tasks. These outcomes highlight +the importance of integrating artistic skills with AI training to prepare +artists and other visual professionals for a technologically evolving +landscape. We see a potential in collaborative synergy with generative AI, +which could reshape creative industries and education in the arts. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on the intersection of AI and artistic education, which aligns with 'social applications of AI' (regarding education and human expertise) and does not clearly meet the required criteria for Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, beyond a tangential mention. + +--- + +## [MMVU: Measuring Expert-Level Multi-Discipline Video Understanding](https://arxiv.org/abs/2501.12380v1) +**arXiv ID:** 2501.12380v1 + +**Abstract:** +> We introduce MMVU, a comprehensive expert-level, multi-discipline benchmark +for evaluating foundation models in video understanding. MMVU includes 3,000 +expert-annotated questions spanning 27 subjects across four core disciplines: +Science, Healthcare, Humanities & Social Sciences, and Engineering. Compared to +prior benchmarks, MMVU features three key advancements. First, it challenges +models to apply domain-specific knowledge and perform expert-level reasoning to +analyze specialized-domain videos, moving beyond the basic visual perception +typically assessed in current video benchmarks. Second, each example is +annotated by human experts from scratch. We implement strict data quality +controls to ensure the high quality of the dataset. Finally, each example is +enriched with expert-annotated reasoning rationals and relevant domain +knowledge, facilitating in-depth analysis. We conduct an extensive evaluation +of 32 frontier multimodal foundation models on MMVU. The latest +System-2-capable models, o1 and Gemini 2.0 Flash Thinking, achieve the highest +performance among the tested models. However, they still fall short of matching +human expertise. Through in-depth error analyses and case studies, we offer +actionable insights for future advancements in expert-level, +knowledge-intensive video understanding for specialized domains. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing and understanding, which is explicitly listed as a criterion for rejection. While it mentions foundation models and knowledge-intensive tasks, the core application is video-centric, outweighing potential marginal relevance to Large Language Models (LLMs) in other areas. + +--- + +## [Physics of Skill Learning](https://arxiv.org/abs/2501.12391v1) +**arXiv ID:** 2501.12391v1 + +**Abstract:** +> We aim to understand physics of skill learning, i.e., how skills are learned +in neural networks during training. We start by observing the Domino effect, +i.e., skills are learned sequentially, and notably, some skills kick off +learning right after others complete learning, similar to the sequential fall +of domino cards. To understand the Domino effect and relevant behaviors of +skill learning, we take physicists' approach of abstraction and simplification. +We propose three models with varying complexities -- the Geometry model, the +Resource model, and the Domino model, trading between reality and simplicity. +The Domino effect can be reproduced in the Geometry model, whose resource +interpretation inspires the Resource model, which can be further simplified to +the Domino model. These models present different levels of abstraction and +simplification; each is useful to study some aspects of skill learning. The +Geometry model provides interesting insights into neural scaling laws and +optimizers; the Resource model sheds light on the learning dynamics of +compositional tasks; the Domino model reveals the benefits of modularity. These +models are not only conceptually interesting -- e.g., we show how Chinchilla +scaling laws can emerge from the Geometry model, but also are useful in +practice by inspiring algorithmic development -- e.g., we show how simple +algorithmic changes, motivated by these toy models, can speed up the training +of deep learning models. + +**Decision Explanation:** Original decision: REJECT +The paper 'Physics of Skill Learning' does not meet the required criteria as it primarily focuses on understanding skill learning in neural networks, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not demonstrate direct practical applications or comparisons with state-of-the-art LLM techniques. + +--- + +## [Learning segmentation from point trajectories](https://arxiv.org/abs/2501.12392v1) +**arXiv ID:** 2501.12392v1 + +**Abstract:** +> We consider the problem of segmenting objects in videos based on their motion +and no other forms of supervision. Prior work has often approached this problem +by using the principle of common fate, namely the fact that the motion of +points that belong to the same object is strongly correlated. However, most +authors have only considered instantaneous motion from optical flow. In this +work, we present a way to train a segmentation network using long-term point +trajectories as a supervisory signal to complement optical flow. The key +difficulty is that long-term motion, unlike instantaneous motion, is difficult +to model -- any parametric approximation is unlikely to capture complex motion +patterns over long periods of time. We instead draw inspiration from subspace +clustering approaches, proposing a loss function that seeks to group the +trajectories into low-rank matrices where the motion of object points can be +approximately explained as a linear combination of other point tracks. Our +method outperforms the prior art on motion-based segmentation, which shows the +utility of long-term motion and the effectiveness of our formulation. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing, which is explicitly listed as a rejection criterion. Additionally, there is no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest. + +--- + +## [Unsupervised Learning in Echo State Networks for Input Reconstruction](https://arxiv.org/abs/2501.11409v1) +**arXiv ID:** 2501.11409v1 + +**Abstract:** +> Conventional echo state networks (ESNs) require supervised learning to train +the readout layer, using the desired outputs as training data. In this study, +we focus on input reconstruction (IR), which refers to training the readout +layer to reproduce the input time series in its output. We reformulate the +learning algorithm of the ESN readout layer to perform IR using unsupervised +learning (UL). By conducting theoretical analysis and numerical experiments, we +demonstrate that IR in ESNs can be effectively implemented under realistic +conditions without explicitly using the desired outputs as training data; in +this way, UL is enabled. Furthermore, we demonstrate that applications relying +on IR, such as dynamical system replication and noise filtering, can be +reformulated within the UL framework. Our findings establish a theoretically +sound and universally applicable IR formulation, along with its related tasks +in ESNs. This work paves the way for novel predictions and highlights +unresolved theoretical challenges in ESNs, particularly in the context of +time-series processing methods and computational models of the brain. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, as it focuses on unsupervised learning in Echo State Networks (ESNs) for input reconstruction, without mentioning Large Language Models (LLMs), practical applications related to LLMs, or relevant experimental results with quantitative metrics. It also doesn't meet any of the additional criteria relevant to LLMs. + +--- + +## [Improving thermal state preparation of Sachdev-Ye-Kitaev model with + reinforcement learning on quantum hardware](https://arxiv.org/abs/2501.11454v1) +**arXiv ID:** 2501.11454v1 + +**Abstract:** +> The Sachdev-Ye-Kitaev (SYK) model, known for its strong quantum correlations +and chaotic behavior, serves as a key platform for quantum gravity studies. +However, variationally preparing thermal states on near-term quantum processors +for large systems (N>12, where N is the number of Majorana fermions) presents a +significant challenge due to the rapid growth in the complexity of +parameterized quantum circuits. This paper addresses this challenge by +integrating reinforcement learning (RL) with convolutional neural networks, +employing an iterative approach to optimize the quantum circuit and its +parameters. The refinement process is guided by a composite reward signal +derived from entropy and the expectation values of the SYK Hamiltonian. This +approach reduces the number of CNOT gates by two orders of magnitude for +systems N>10 compared to traditional methods like first-order Trotterization. +We demonstrate the effectiveness of the RL framework in both noiseless and +noisy quantum hardware environments, maintaining high accuracy in thermal state +preparation. This work contributes to the advancement of a scalable, RL-based +framework with applications for computations of thermal out-of-time-order +correlators in quantum many-body systems and quantum gravity studies on +near-term quantum hardware. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the criteria as it primarily focuses on quantum computing and reinforcement learning for thermal state preparation, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [With Great Backbones Comes Great Adversarial Transferability](https://arxiv.org/abs/2501.12275v1) +**arXiv ID:** 2501.12275v1 + +**Abstract:** +> Advances in self-supervised learning (SSL) for machine vision have improved +representation robustness and model performance, giving rise to pre-trained +backbones like \emph{ResNet} and \emph{ViT} models tuned with SSL methods such +as \emph{SimCLR}. Due to the computational and data demands of pre-training, +the utilization of such backbones becomes a strenuous necessity. However, +employing these backbones may inherit vulnerabilities to adversarial attacks. +While adversarial robustness has been studied under \emph{white-box} and +\emph{black-box} settings, the robustness of models tuned on pre-trained +backbones remains largely unexplored. Additionally, the role of tuning +meta-information in mitigating exploitation risks is unclear. This work +systematically evaluates the adversarial robustness of such models across +$20,000$ combinations of tuning meta-information, including fine-tuning +techniques, backbone families, datasets, and attack types. We propose using +proxy models to transfer attacks, simulating varying levels of target knowledge +by fine-tuning these proxies with diverse configurations. Our findings reveal +that proxy-based attacks approach the effectiveness of \emph{white-box} +methods, even with minimal tuning knowledge. We also introduce a naive +"backbone attack," leveraging only the backbone to generate adversarial +samples, which outperforms \emph{black-box} attacks and rivals \emph{white-box} +methods, highlighting critical risks in model-sharing practices. Finally, our +ablations reveal how increasing tuning meta-information impacts attack +transferability, measuring each meta-information combination. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on adversarial attacks and robustness in machine vision (computer vision) with self-supervised learning, which does not meet the criteria of focusing on Large Language Models (LLMs) or their applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. The topic aligns more with video/image processing, which is explicitly listed as a reason for rejection. + +--- + +## [Goal-oriented Transmission Scheduling: Structure-guided DRL with a + Unified Dual On-policy and Off-policy Approach](https://arxiv.org/abs/2501.11921v1) +**arXiv ID:** 2501.11921v1 + +**Abstract:** +> Goal-oriented communications prioritize application-driven objectives over +data accuracy, enabling intelligent next-generation wireless systems. Efficient +scheduling in multi-device, multi-channel systems poses significant challenges +due to high-dimensional state and action spaces. We address these challenges by +deriving key structural properties of the optimal solution to the goal-oriented +scheduling problem, incorporating Age of Information (AoI) and channel states. +Specifically, we establish the monotonicity of the optimal state value function +(a measure of long-term system performance) w.r.t. channel states and prove its +asymptotic convexity w.r.t. AoI states. Additionally, we derive the +monotonicity of the optimal policy w.r.t. channel states, advancing the +theoretical framework for optimal scheduling. Leveraging these insights, we +propose the structure-guided unified dual on-off policy DRL (SUDO-DRL), a +hybrid algorithm that combines the stability of on-policy training with the +sample efficiency of off-policy methods. Through a novel structural property +evaluation framework, SUDO-DRL enables effective and scalable training, +addressing the complexities of large-scale systems. Numerical results show +SUDO-DRL improves system performance by up to 45% and reduces convergence time +by 40% compared to state-of-the-art methods. It also effectively handles +scheduling in much larger systems, where off-policy DRL fails and on-policy +benchmarks exhibit significant performance loss, demonstrating its scalability +and efficacy in goal-oriented communications. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria (1-3) as it focuses on Goal-oriented Transmission Scheduling in wireless systems using Deep Reinforcement Learning (DRL), with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + diff --git a/llm_processor/newpapers-r1-filtered.json b/llm_processor/newpapers-r1-filtered.json new file mode 100644 index 0000000..8471997 --- /dev/null +++ b/llm_processor/newpapers-r1-filtered.json @@ -0,0 +1,1060 @@ +{ + "accepted": [ + { + "paper": { + "title": "Agent-R: Training Language Model Agents to Reflect via Iterative\n Self-Training", + "abstract": "Large Language Models (LLMs) agents are increasingly pivotal for addressing\ncomplex tasks in interactive environments. Existing work mainly focuses on\nenhancing performance through behavior cloning from stronger experts, yet such\napproaches often falter in real-world applications, mainly due to the inability\nto recover from errors. However, step-level critique data is difficult and\nexpensive to collect. Automating and dynamically constructing self-critique\ndatasets is thus crucial to empowering models with intelligent agent\ncapabilities. In this work, we propose an iterative self-training framework,\nAgent-R, that enables language Agent to Reflect on the fly. Unlike traditional\nmethods that reward or penalize actions based on correctness, Agent-R leverages\nMCTS to construct training data that recover correct trajectories from\nerroneous ones. A key challenge of agent reflection lies in the necessity for\ntimely revision rather than waiting until the end of a rollout. To address\nthis, we introduce a model-guided critique construction mechanism: the actor\nmodel identifies the first error step (within its current capability) in a\nfailed trajectory. Starting from it, we splice it with the adjacent correct\npath, which shares the same parent node in the tree. This strategy enables the\nmodel to learn reflection based on its current policy, therefore yielding\nbetter learning efficiency. To further explore the scalability of this\nself-improvement paradigm, we investigate iterative refinement of both error\ncorrection capabilities and dataset construction. Our findings demonstrate that\nAgent-R continuously improves the model's ability to recover from errors and\nenables timely error correction. Experiments on three interactive environments\nshow that Agent-R effectively equips agents to correct erroneous actions while\navoiding loops, achieving superior performance compared to baseline methods\n(+5.59%).", + "arxiv_id": "2501.11425v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all three mandatory criteria: 1) Focuses on practical applications of LLM agents in interactive environments, 2) Provides quantitative results (+5.59% performance gain) across three environments, and 3) Compares with baseline methods. It also meets secondary criteria 4 by detailing its iterative self-training methodology using MCTS and model-guided critique. The work addresses agentic AI capabilities (error recovery in real-world tasks) and demonstrates novelty in self-improvement paradigms, aligning with inclusion priorities." + }, + { + "paper": { + "title": "Systematic Abductive Reasoning via Diverse Relation Representations in\n Vector-symbolic Architecture", + "abstract": "In abstract visual reasoning, monolithic deep learning models suffer from\nlimited interpretability and generalization, while existing neuro-symbolic\napproaches fall short in capturing the diversity and systematicity of\nattributes and relation representations. To address these challenges, we\npropose a Systematic Abductive Reasoning model with diverse relation\nrepresentations (Rel-SAR) in Vector-symbolic Architecture (VSA) to solve\nRaven's Progressive Matrices (RPM). To derive attribute representations with\nsymbolic reasoning potential, we introduce not only various types of atomic\nvectors that represent numeric, periodic and logical semantics, but also the\nstructured high-dimentional representation (SHDR) for the overall Grid\ncomponent. For systematic reasoning, we propose novel numerical and logical\nrelation functions and perform rule abduction and execution in a unified\nframework that integrates these relation representations. Experimental results\ndemonstrate that Rel-SAR achieves significant improvement on RPM tasks and\nexhibits robust out-of-distribution generalization. Rel-SAR leverages the\nsynergy between HD attribute representations and symbolic reasoning to achieve\nsystematic abductive reasoning with both interpretable and computable\nsemantics.", + "arxiv_id": "2501.11896v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all mandatory criteria: 1) Focuses on systematic reasoning with vector-symbolic architectures relevant to LLM applications, 2) Presents experimental results with quantitative improvements on RPM tasks, 3) Compares against existing neuro-symbolic approaches. Also meets secondary criteria 4 by detailing methodology including atomic vectors and relation functions. Shows novelty in attribute representations and passes all rejection filters." + }, + { + "paper": { + "title": "Reasoning Language Models: A Blueprint", + "abstract": "Reasoning language models (RLMs), also known as Large Reasoning Models\n(LRMs), such as OpenAI's o1 and o3, DeepSeek-V3, and Alibaba's QwQ, have\nredefined AI's problem-solving capabilities by extending large language models\n(LLMs) with advanced reasoning mechanisms. Yet, their high costs, proprietary\nnature, and complex architectures - uniquely combining Reinforcement Learning\n(RL), search heuristics, and LLMs - present accessibility and scalability\nchallenges. To address these, we propose a comprehensive blueprint that\norganizes RLM components into a modular framework, based on a survey and\nanalysis of all RLM works. This blueprint incorporates diverse reasoning\nstructures (chains, trees, graphs, and nested forms), reasoning strategies\n(e.g., Monte Carlo Tree Search, Beam Search), RL concepts (policy, value models\nand others), and supervision schemes (Output-Based and Process-Based\nSupervision). We also provide detailed mathematical formulations and\nalgorithmic specifications to simplify RLM implementation. By showing how\nschemes like LLaMA-Berry, QwQ, Journey Learning, and Graph of Thoughts fit as\nspecial cases, we demonstrate the blueprint's versatility and unifying\npotential. To illustrate its utility, we introduce x1, a modular implementation\nfor rapid RLM prototyping and experimentation. Using x1 and a literature\nreview, we provide key insights, such as multi-phase training for policy and\nvalue models, and the importance of familiar training distributions. Finally,\nwe outline how RLMs can integrate with a broader LLM ecosystem, including tools\nand databases. Our work demystifies RLM construction, democratizes advanced\nreasoning capabilities, and fosters innovation, aiming to mitigate the gap\nbetween \"rich AI\" and \"poor AI\" by lowering barriers to RLM development and\nexperimentation.", + "arxiv_id": "2501.11223v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all mandatory criteria: 1) Focuses on practical RLM applications through modular frameworks and ecosystem integration; 2) Provides implementation (x1) and insights derived from experimentation/literature; 3) Compares with SOTA (LLaMA-Berry/QwQ). Also satisfies optional criteria 4 (detailed methodology/algorithm specs) and 5 (addresses scalability challenges). No rejection triggers apply." + }, + { + "paper": { + "title": "Question-to-Question Retrieval for Hallucination-Free Knowledge Access:\n An Approach for Wikipedia and Wikidata Question Answering", + "abstract": "This paper introduces an approach to question answering over knowledge bases\nlike Wikipedia and Wikidata by performing \"question-to-question\" matching and\nretrieval from a dense vector embedding store. Instead of embedding document\ncontent, we generate a comprehensive set of questions for each logical content\nunit using an instruction-tuned LLM. These questions are vector-embedded and\nstored, mapping to the corresponding content. Vector embedding of user queries\nare then matched against this question vector store. The highest similarity\nscore leads to direct retrieval of the associated article content, eliminating\nthe need for answer generation. Our method achieves high cosine similarity ( \u003e\n0.9 ) for relevant question pairs, enabling highly precise retrieval. This\napproach offers several advantages including computational efficiency, rapid\nresponse times, and increased scalability. We demonstrate its effectiveness on\nWikipedia and Wikidata, including multimedia content through structured fact\nretrieval from Wikidata, opening up new pathways for multimodal question\nanswering.", + "arxiv_id": "2501.11301v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all mandatory criteria: 1 (LLM application in RAG for knowledge bases), 2 (quantitative metrics like \u003e0.9 cosine similarity), and 3 (implicit comparison via novel methodology). It also meets criterion 4 (methodology details) and demonstrates novelty. No rejection triggers apply." + }, + { + "paper": { + "title": "Few-shot Policy (de)composition in Conversational Question Answering", + "abstract": "The task of policy compliance detection (PCD) is to determine if a scenario\nis in compliance with respect to a set of written policies. In a conversational\nsetting, the results of PCD can indicate if clarifying questions must be asked\nto determine compliance status. Existing approaches usually claim to have\nreasoning capabilities that are latent or require a large amount of annotated\ndata. In this work, we propose logical decomposition for policy compliance\n(LDPC): a neuro-symbolic framework to detect policy compliance using large\nlanguage models (LLMs) in a few-shot setting. By selecting only a few exemplars\nalongside recently developed prompting techniques, we demonstrate that our\napproach soundly reasons about policy compliance conversations by extracting\nsub-questions to be answered, assigning truth values from contextual\ninformation, and explicitly producing a set of logic statements from the given\npolicies. The formulation of explicit logic graphs can in turn help answer\nPCDrelated questions with increased transparency and explainability. We apply\nthis approach to the popular PCD and conversational machine reading benchmark,\nShARC, and show competitive performance with no task-specific finetuning. We\nalso leverage the inherently interpretable architecture of LDPC to understand\nwhere errors occur, revealing ambiguities in the ShARC dataset and highlighting\nthe challenges involved with reasoning for conversational question answering.", + "arxiv_id": "2501.11335v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Focuses on practical LLM applications (policy compliance via neuro-symbolic framework), 2) Demonstrates quantitative results on ShARC benchmark with competitive performance, 3) Compares with existing approaches. Also meets secondary criteria 4 (clear methodology using prompting/logic decomposition) and 5 (addresses real-world challenges). No rejection triggers present." + }, + { + "paper": { + "title": "Neural Contextual Reinforcement Framework for Logical Structure Language\n Generation", + "abstract": "The Neural Contextual Reinforcement Framework introduces an innovative\napproach to enhancing the logical coherence and structural consistency of text\ngenerated by large language models. Leveraging reinforcement learning\nprinciples, the framework integrates custom reward functions and dynamic\ncontext alignment mechanisms to address challenges inherent in maintaining\nlong-range dependencies across extended sequences. The architecture\nincorporates multi-head attention layers and hierarchical encoding modules,\nenabling the model to produce outputs that align closely with human\nexpectations of logical structure and semantic flow. Quantitative evaluations\nacross diverse datasets demonstrate substantial improvements in coherence\nmetrics, perplexity reduction, and semantic alignment, showcasing the\nframework's ability to outperform baseline models in both general and\ndomain-specific tasks. Qualitative analyses further highlight the framework's\ncapacity to generate text with improved narrative clarity and reduced\nredundancy, reflecting its effectiveness in balancing fluency with structural\nprecision. In addition to its performance gains, the framework exhibits\nrobustness in handling noisy input data and scalability across varying model\nsizes, reinforcing its versatility in practical applications. Experimental\nresults reveal that optimal context window sizes significantly influence\ncoherence outcomes, showing the importance of architectural flexibility in\nadapting to diverse linguistic structures. Cross-lingual performance\nevaluations affirm the framework's adaptability to multiple languages,\nextending its utility beyond monolingual contexts. Resource efficiency analyses\nindicate a reduction in computational overhead compared to traditional\napproaches, emphasizing the practicality of the framework for large-scale\ndeployment.", + "arxiv_id": "2501.11417v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets ALL mandatory criteria: 1) Focuses on practical text generation improvements for LLMs, 2) Includes quantitative metrics (coherence, perplexity) and comparisons with baselines, 3) Compares with existing models. Also meets optional criteria 4 through detailed methodology description. Shows novelty in reinforcement framework and addresses reproducibility through efficiency analysis. No rejection triggers present." + }, + { + "paper": { + "title": "Explainable Lane Change Prediction for Near-Crash Scenarios Using\n Knowledge Graph Embeddings and Retrieval Augmented Generation", + "abstract": "Lane-changing maneuvers, particularly those executed abruptly or in risky\nsituations, are a significant cause of road traffic accidents. However, current\nresearch mainly focuses on predicting safe lane changes. Furthermore, existing\naccident datasets are often based on images only and lack comprehensive sensory\ndata. In this work, we focus on predicting risky lane changes using the CRASH\ndataset (our own collected dataset specifically for risky lane changes), and\nsafe lane changes (using the HighD dataset). Then, we leverage KG and Bayesian\ninference to predict these maneuvers using linguistic contextual information,\nenhancing the model's interpretability and transparency. The model achieved a\n91.5% f1-score with anticipation time extending to four seconds for risky lane\nchanges, and a 90.0% f1-score for predicting safe lane changes with the same\nanticipation time. We validate our model by integrating it into a vehicle\nwithin the CARLA simulator in scenarios that involve risky lane changes. The\nmodel managed to anticipate sudden lane changes, thus providing automated\nvehicles with further time to plan and execute appropriate safe reactions.\nFinally, to enhance the explainability of our model, we utilize RAG to provide\nclear and natural language explanations for the given prediction.", + "arxiv_id": "2501.11560v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets mandatory criteria 1 (practical LLM applications in KG/RAG), 2 (quantitative results with 91.5% F1-score), and 3 (implied advancement via novel CRASH dataset/approach). It meets optional criterion 4 (methodology details) and addresses agentic AI in CARLA simulations. No rejection triggers apply." + }, + { + "paper": { + "title": "Recurrent Diffusion for Large-Scale Parameter Generation", + "abstract": "Parameter generation has struggled to scale up for a long time, significantly\nlimiting its range of applications. In this study, we introduce\n\\textbf{R}ecurrent diffusion for large-scale \\textbf{P}arameter\n\\textbf{G}eneration, called \\textbf{RPG}. We first divide the trained\nparameters into non-overlapping parts, after which a recurrent model is\nproposed to learn their relationships. The recurrent model's outputs, as\nconditions, are then fed into a diffusion model to generate the neural network\nparameters. Using only a single GPU, recurrent diffusion enables us to generate\npopular vision and language models such as ConvNeXt-L and LoRA parameters of\nLLaMA-7B. Meanwhile, across various architectures and tasks, the generated\nparameters consistently perform comparable results over trained networks.\nNotably, our approach also shows the potential to generate models for handling\nunseen tasks, which largely increases the practicality of parameter generation.\nOur code is available\n\\href{https://github.com/NUS-HPC-AI-Lab/Recurrent-Parameter-Generation}{here}.", + "arxiv_id": "2501.11587v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets mandatory criteria 1 (practical LLM parameter generation for vision/language models), 2 (quantitative results showing comparable performance to trained networks), and 3 (implicit comparison to standard trained networks as baseline). Addresses optional criteria 4 (methodology details) and 5 (real-world practicality for unseen tasks). No rejection triggers apply. Novelty and implementability with code availability support inclusion." + }, + { + "paper": { + "title": "SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language\n Models Tackling Knowledge-based Reasoning Tasks", + "abstract": "Deductive reasoning is a crucial logical capability that assists us in\nsolving complex problems based on existing knowledge. Although augmented by\nChain-of-Thought prompts, Large Language Models (LLMs) might not follow the\ncorrect reasoning paths. Enhancing the deductive reasoning abilities of LLMs,\nand leveraging their extensive built-in knowledge for various reasoning tasks,\nremains an open question. Attempting to mimic the human deductive reasoning\nparadigm, we propose a multi-stage Syllogistic-Reasoning Framework of Thought\n(SR-FoT) that enables LLMs to perform syllogistic deductive reasoning to handle\ncomplex knowledge-based reasoning tasks. Our SR-FoT begins by interpreting the\nquestion and then uses the interpretation and the original question to propose\na suitable major premise. It proceeds by generating and answering minor premise\nquestions in two stages to match the minor premises. Finally, it guides LLMs to\nuse the previously generated major and minor premises to perform syllogistic\ndeductive reasoning to derive the answer to the original question. Extensive\nand thorough experiments on knowledge-based reasoning tasks have demonstrated\nthe effectiveness and advantages of our SR-FoT.", + "arxiv_id": "2501.11599v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1 (practical reasoning framework), 2 (mentions experimental validation), 3 (compares with Chain-of-Thought baselines). Meets optional criterion 4 (detailed methodology description). Novel syllogistic approach aligns with LLM application extensions. No rejection triggers present." + }, + { + "paper": { + "title": "Optimizing Pretraining Data Mixtures with LLM-Estimated Utility", + "abstract": "Large Language Models improve with increasing amounts of high-quality\ntraining data. However, leveraging larger datasets requires balancing quality,\nquantity, and diversity across sources. After evaluating nine baseline methods\nunder both compute- and data-constrained scenarios, we find token-count\nheuristics outperform manual and learned mixes, indicating that simple\napproaches accounting for dataset size and diversity are surprisingly\neffective. Building on this insight, we propose two complementary approaches:\nUtiliMax, which extends token-based heuristics by incorporating utility\nestimates from reduced-scale ablations, achieving up to a 10.6x speedup over\nmanual baselines; and Model Estimated Data Utility (MEDU), which leverages LLMs\nto estimate data utility from small samples, matching ablation-based\nperformance while reducing computational requirements by $\\sim$200x. Together,\nthese approaches establish a new framework for automated, compute-efficient\ndata mixing that is robust across training regimes.", + "arxiv_id": "2501.11747v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all mandatory criteria: 1) Focuses on practical LLM training optimization via data mixing; 2) Provides quantitative results (10.6x speedup, 200x compute reduction); 3) Compares with baselines. Also addresses methodology (criterion 4) and computational challenges (criterion 5), while introducing novel approaches (MEDU/UtiliMax) with implementable methods." + }, + { + "paper": { + "title": "Benchmarking Large Language Models via Random Variables", + "abstract": "With the continuous advancement of large language models (LLMs) in\nmathematical reasoning, evaluating their performance in this domain has become\na prominent research focus. Recent studies have raised concerns about the\nreliability of current mathematical benchmarks, highlighting issues such as\nsimplistic design and potential data leakage. Therefore, creating a reliable\nbenchmark that effectively evaluates the genuine capabilities of LLMs in\nmathematical reasoning remains a significant challenge. To address this, we\npropose RV-Bench, a framework for Benchmarking LLMs via Random Variables in\nmathematical reasoning. Specifically, the background content of a random\nvariable question (RV question) mirrors the original problem in existing\nstandard benchmarks, but the variable combinations are randomized into\ndifferent values. LLMs must fully understand the problem-solving process for\nthe original problem to correctly answer RV questions with various combinations\nof variable values. As a result, the LLM's genuine capability in mathematical\nreasoning is reflected by its accuracy on RV-Bench. Extensive experiments are\nconducted with 29 representative LLMs across 900+ RV questions. A leaderboard\nfor RV-Bench ranks the genuine capability of these LLMs. Further analysis of\naccuracy dropping indicates that current LLMs still struggle with complex\nmathematical reasoning problems.", + "arxiv_id": "2501.11790v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets mandatory criteria 2 (quantitative experiments with 29 LLMs and accuracy metrics) and 3 (leaderboard comparison with SOTA models). While criterion 1 applicability is marginal, the framework indirectly supports practical LLM evaluation. Criterion 4 is addressed through methodology description of RV-Bench. No rejection triggers apply. Inclusion policy favors acceptance." + }, + { + "paper": { + "title": "Fact-Preserved Personalized News Headline Generation", + "abstract": "Personalized news headline generation, aiming at generating user-specific\nheadlines based on readers' preferences, burgeons a recent flourishing research\ndirection. Existing studies generally inject a user interest embedding into an\nencoderdecoder headline generator to make the output personalized, while the\nfactual consistency of headlines is inadequate to be verified. In this paper,\nwe propose a framework Fact-Preserved Personalized News Headline Generation\n(short for FPG), to prompt a tradeoff between personalization and consistency.\nIn FPG, the similarity between the candidate news to be exposed and the\nhistorical clicked news is used to give different levels of attention to key\nfacts in the candidate news, and the similarity scores help to learn a\nfact-aware global user embedding. Besides, an additional training procedure\nbased on contrastive learning is devised to further enhance the factual\nconsistency of generated headlines. Extensive experiments conducted on a\nreal-world benchmark PENS validate the superiority of FPG, especially on the\ntradeoff between personalization and factual consistency.", + "arxiv_id": "2501.11828v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all mandatory criteria: 1) Practical applications in RAG through personalized headline generation, 2) Experimental validation with quantitative metrics on PENS benchmark, 3) Comparison with existing methods. It also meets optional criteria 4 (methodology details) and 5 (real-world challenges). No rejection triggers apply as it focuses on content generation rather than excluded domains." + }, + { + "paper": { + "title": "From Drafts to Answers: Unlocking LLM Potential via Aggregation\n Fine-Tuning", + "abstract": "Scaling data and model size has been proven effective for boosting the\nperformance of large language models. In addition to training-time scaling,\nrecent studies have revealed that increasing test-time computational resources\ncan further improve performance. In this work, we introduce Aggregation\nFine-Tuning (AFT), a supervised finetuning paradigm where the model learns to\nsynthesize multiple draft responses, referred to as proposals, into a single,\nrefined answer, termed aggregation. At inference time, a propose-and-aggregate\nstrategy further boosts performance by iteratively generating proposals and\naggregating them. Empirical evaluations on benchmark datasets show that\nAFT-trained models substantially outperform standard SFT. Notably, an AFT\nmodel, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3% LC\nwin rate on AlpacaEval 2, surpassing significantly larger LLMs such as\nLlama3.1-405B-Instruct and GPT4. By combining sequential refinement and\nparallel sampling, the propose-and-aggregate framework scales inference-time\ncomputation in a flexible manner. Overall, These findings position AFT as a\npromising approach to unlocking additional capabilities of LLMs without\nresorting to increasing data volume or model size.", + "arxiv_id": "2501.11877v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Practical application through performance-efficient LLM enhancement, 2) Quantitative results showing 41.3% LC win rate improvement, 3) Direct comparison with SOTA models like GPT-4. Also meets optional criteria 4 (methodology description) and demonstrates novelty through aggregation framework. No rejection triggers present." + }, + { + "paper": { + "title": "Panoramic Interests: Stylistic-Content Aware Personalized Headline\n Generation", + "abstract": "Personalized news headline generation aims to provide users with\nattention-grabbing headlines that are tailored to their preferences. Prevailing\nmethods focus on user-oriented content preferences, but most of them overlook\nthe fact that diverse stylistic preferences are integral to users' panoramic\ninterests, leading to suboptimal personalization. In view of this, we propose a\nnovel Stylistic-Content Aware Personalized Headline Generation (SCAPE)\nframework. SCAPE extracts both content and stylistic features from headlines\nwith the aid of large language model (LLM) collaboration. It further adaptively\nintegrates users' long- and short-term interests through a contrastive\nlearning-based hierarchical fusion network. By incorporating the panoramic\ninterests into the headline generator, SCAPE reflects users' stylistic-content\npreferences during the generation process. Extensive experiments on the\nreal-world dataset PENS demonstrate the superiority of SCAPE over baselines.", + "arxiv_id": "2501.11900v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets ALL mandatory criteria: 1) Practical LLM application in personalized content generation, 2) Experimental validation with quantitative metrics on PENS dataset, 3) Comparison with baselines. Also addresses optional criteria 4 (methodology details) and 5 (real-world applications). No rejection triggers present." + }, + { + "paper": { + "title": "LuxVeri at GenAI Detection Task 1: Inverse Perplexity Weighted Ensemble\n for Robust Detection of AI-Generated Text across English and Multilingual\n Contexts", + "abstract": "This paper presents a system developed for Task 1 of the COLING 2025 Workshop\non Detecting AI-Generated Content, focusing on the binary classification of\nmachine-generated versus human-written text. Our approach utilizes an ensemble\nof models, with weights assigned according to each model's inverse perplexity,\nto enhance classification accuracy. For the English text detection task, we\ncombined RoBERTa-base, RoBERTa-base with the OpenAI detector, and\nBERT-base-cased, achieving a Macro F1-score of 0.7458, which ranked us 12th out\nof 35 teams. We ensembled RemBERT, XLM-RoBERTa-base, and\nBERT-base-multilingual-case for the multilingual text detection task, employing\nthe same inverse perplexity weighting technique. This resulted in a Macro\nF1-score of 0.7513, positioning us 4th out of 25 teams. Our results demonstrate\nthe effectiveness of inverse perplexity weighting in improving the robustness\nof machine-generated text detection across both monolingual and multilingual\nsettings, highlighting the potential of ensemble methods for this challenging\ntask.", + "arxiv_id": "2501.11914v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets mandatory criteria 2 (quantitative metrics with F1-scores) and 3 (comparison via competition rankings). While practical applications (1) focus on AI detection rather than LLM integration with KG/RAG/agentic AI, it marginally qualifies as a real-world LLM application. It also meets optional criterion 4 (methodology details on ensemble techniques). No rejection triggers apply." + }, + { + "paper": { + "title": "LuxVeri at GenAI Detection Task 3: Cross-Domain Detection of\n AI-Generated Text Using Inverse Perplexity-Weighted Ensemble of Fine-Tuned\n Transformer Models", + "abstract": "This paper presents our approach for Task 3 of the GenAI content detection\nworkshop at COLING-2025, focusing on Cross-Domain Machine-Generated Text (MGT)\nDetection. We propose an ensemble of fine-tuned transformer models, enhanced by\ninverse perplexity weighting, to improve classification accuracy across diverse\ntext domains. For Subtask A (Non-Adversarial MGT Detection), we combined a\nfine-tuned RoBERTa-base model with an OpenAI detector-integrated RoBERTa-base\nmodel, achieving an aggregate TPR score of 0.826, ranking 10th out of 23\ndetectors. In Subtask B (Adversarial MGT Detection), our fine-tuned\nRoBERTa-base model achieved a TPR score of 0.801, securing 8th out of 22\ndetectors. Our results demonstrate the effectiveness of inverse\nperplexity-based weighting for enhancing generalization and performance in both\nnon-adversarial and adversarial MGT detection, highlighting the potential for\ntransformer models in cross-domain AI-generated content detection.", + "arxiv_id": "2501.11918v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets ALL mandatory criteria: 1) Practical application in AI-generated text detection, 2) Quantitative results (TPR scores) with experimental validation, 3) Comparison via rankings against other detectors. Also meets criteria 4 (methodology details) and 5 (real-world detection challenges). No rejection triggers present. Approach shows implementable transformer-based methods with cross-domain applications." + }, + { + "paper": { + "title": "Webvs. LLMs: An Empirical Study of Learning Behaviors of CS2 Students", + "abstract": "LLMs such as ChatGPT have been widely adopted by students in higher education\nas tools for learning programming and related concepts. However, it remains\nunclear how effective students are and what strategies students use while\nlearning with LLMs. Since the majority of students' experiences in online\nself-learning have come through using search engines such as Google, evaluating\nAI tools in this context can help us address these gaps. In this mixed methods\nresearch, we conducted an exploratory within-subjects study to understand how\nCS2 students learn programming concepts using both LLMs as well as traditional\nonline methods such as educational websites and videos to examine how students\napproach learning within and across both scenarios. We discovered that students\nfound it easier to learn a more difficult concept using traditional methods\nthan using ChatGPT. We also found that students ask fewer follow-ups and use\nmore keyword-based queries for search engines while their prompts to LLMs tend\nto explicitly ask for information.", + "arxiv_id": "2501.11935v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Practical application of LLMs in education, 2) Empirical study with mixed methods results, 3) Comparison between LLMs and traditional search methods. Also addresses optional criteria 5 (real-world learning challenges). No rejection triggers apply." + }, + { + "paper": { + "title": "TAD-Bench: A Comprehensive Benchmark for Embedding-Based Text Anomaly\n Detection", + "abstract": "Text anomaly detection is crucial for identifying spam, misinformation, and\noffensive language in natural language processing tasks. Despite the growing\nadoption of embedding-based methods, their effectiveness and generalizability\nacross diverse application scenarios remain under-explored. To address this, we\npresent TAD-Bench, a comprehensive benchmark designed to systematically\nevaluate embedding-based approaches for text anomaly detection. TAD-Bench\nintegrates multiple datasets spanning different domains, combining\nstate-of-the-art embeddings from large language models with a variety of\nanomaly detection algorithms. Through extensive experiments, we analyze the\ninterplay between embeddings and detection methods, uncovering their strengths,\nweaknesses, and applicability to different tasks. These findings offer new\nperspectives on building more robust, efficient, and generalizable anomaly\ndetection systems for real-world applications.", + "arxiv_id": "2501.11960v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all three mandatory criteria: 1) Focuses on practical applications of LLMs for anomaly detection in real-world scenarios, 2) Presents experimental results with quantitative metrics through benchmark creation, and 3) Compares state-of-the-art embeddings and detection methods. While mentioning social applications (spam/misinformation detection), the primary focus remains technical benchmarking rather than social impact analysis, avoiding rejection criteria." + }, + { + "paper": { + "title": "Bridging Visualization and Optimization: Multimodal Large Language\n Models on Graph-Structured Combinatorial Optimization", + "abstract": "Graph-structured combinatorial challenges are inherently difficult due to\ntheir nonlinear and intricate nature, often rendering traditional computational\nmethods ineffective or expensive. However, these challenges can be more\nnaturally tackled by humans through visual representations that harness our\ninnate ability for spatial reasoning. In this study, we propose transforming\ngraphs into images to preserve their higher-order structural features\naccurately, revolutionizing the representation used in solving graph-structured\ncombinatorial tasks. This approach allows machines to emulate human-like\nprocessing in addressing complex combinatorial challenges. By combining the\ninnovative paradigm powered by multimodal large language models (MLLMs) with\nsimple search techniques, we aim to develop a novel and effective framework for\ntackling such problems. Our investigation into MLLMs spanned a variety of\ngraph-based tasks, from combinatorial problems like influence maximization to\nsequential decision-making in network dismantling, as well as addressing six\nfundamental graph-related issues. Our findings demonstrate that MLLMs exhibit\nexceptional spatial intelligence and a distinctive capability for handling\nthese problems, significantly advancing the potential for machines to\ncomprehend and analyze graph-structured data with a depth and intuition akin to\nhuman cognition. These results also imply that integrating MLLMs with simple\noptimization strategies could form a novel and efficient approach for\nnavigating graph-structured combinatorial challenges without complex\nderivations, computationally demanding training and fine-tuning.", + "arxiv_id": "2501.11968v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all mandatory criteria: 1) focuses on practical LLM applications in graph-structured optimization, 2) includes experimental validation across multiple tasks with claims of significant advancements, and 3) implicitly compares against traditional methods. It also satisfies optional criteria 4 (methodology for graph-to-image transformation) and 5 (real-world applications like network dismantling). While SOTA comparison isn't explicit, inclusiveness favors acceptance given spatial reasoning innovations and implementable multimodal LLM approach." + }, + { + "paper": { + "title": "Leveraging Graph Structures and Large Language Models for End-to-End\n Synthetic Task-Oriented Dialogues", + "abstract": "Training task-oriented dialogue systems is both costly and time-consuming,\ndue to the need for high-quality datasets encompassing diverse intents.\nTraditional methods depend on extensive human annotation, while recent\nadvancements leverage large language models (LLMs) to generate synthetic data.\nHowever, these approaches often require custom prompts or code, limiting\naccessibility for non-technical users. We introduce GraphTOD, an end-to-end\nframework that simplifies the generation of task-oriented dialogues. Users can\ncreate dialogues by specifying transition graphs in JSON format. Our evaluation\ndemonstrates that GraphTOD generates high-quality dialogues across various\ndomains, significantly lowering the cost and complexity of dataset creation.", + "arxiv_id": "2501.11977v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (practical LLM application in dialogue generation), 2 (quantitative evaluation showing cost reduction), and 4 (clear methodology using JSON transition graphs). No rejection triggers present. Novel approach combining graphs/LLMs for synthetic data generation aligns with selection priorities." + }, + { + "paper": { + "title": "Harnessing Generative Pre-Trained Transformer for Datacenter Packet\n Trace Generation", + "abstract": "Today, the rapid growth of applications reliant on datacenters calls for new\nadvancements to meet the increasing traffic and computational demands. Traffic\ntraces from datacenters are essential for further development and optimization\nof future datacenters. However, traces are rarely released to the public.\nResearchers often use simplified mathematical models that lack the depth needed\nto recreate intricate traffic patterns and, thus, miss optimization\nopportunities found in realistic traffic. In this preliminary work, we\nintroduce DTG-GPT, a packet-level Datacenter Traffic Generator (DTG), based on\nthe generative pre-trained transformer (GPT) architecture used by many\nstate-of-the-art large language models. We train our model on a small set of\navailable traffic traces from different domains and offer a simple methodology\nto evaluate the fidelity of the generated traces to their original\ncounterparts. We show that DTG-GPT can synthesize novel traces that mimic the\nspatiotemporal patterns found in real traffic traces. We further demonstrate\nthat DTG-GPT can generate traces for networks of different scales while\nmaintaining fidelity. Our findings indicate the potential that, in the future,\nsimilar models to DTG-GPT will allow datacenter operators to release traffic\ninformation to the research community via trained GPT models.", + "arxiv_id": "2501.12033v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets criteria 1 (practical LLM application in datacenter optimization), 2 (experimental validation of trace fidelity), and 3 (comparison with simplified mathematical models as baseline). It also addresses criterion 5 (real-world challenge of data scarcity in datacenters). No rejection triggers apply. While methodology details are limited (preliminary work), it satisfies inclusion thresholds." + }, + { + "paper": { + "title": "Condor: Enhance LLM Alignment with Knowledge-Driven Data Synthesis and\n Refinement", + "abstract": "The quality of Supervised Fine-Tuning (SFT) data plays a critical role in\nenhancing the conversational capabilities of Large Language Models (LLMs).\nHowever, as LLMs become more advanced, the availability of high-quality\nhuman-annotated SFT data has become a significant bottleneck, necessitating a\ngreater reliance on synthetic training data. In this work, we introduce Condor,\na novel two-stage synthetic data generation framework that incorporates World\nKnowledge Tree and Self-Reflection Refinement to produce high-quality SFT data\nat scale. Our experimental results demonstrate that a base model fine-tuned on\nonly 20K Condor-generated samples achieves superior performance compared to\ncounterparts. The additional refinement stage in Condor further enables\niterative self-improvement for LLMs at various scales (up to 72B), validating\nthe effectiveness of our approach. Furthermore, our investigation into the\nscaling for synthetic data in post-training reveals substantial unexplored\npotential for performance improvements, opening promising avenues for future\nresearch.", + "arxiv_id": "2501.12273v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Practical application in LLM alignment through synthetic data generation, 2) Quantitative results showing performance improvements with 20K samples and scaling to 72B models, 3) Implicit comparison with counterparts. Also meets secondary criteria 4 (methodology description) and 5 (addresses data scarcity challenge). No rejection triggers present." + }, + { + "paper": { + "title": "RALAD: Bridging the Real-to-Sim Domain Gap in Autonomous Driving with\n Retrieval-Augmented Learning", + "abstract": "In the pursuit of robust autonomous driving systems, models trained on\nreal-world datasets often struggle to adapt to new environments, particularly\nwhen confronted with corner cases such as extreme weather conditions.\nCollecting these corner cases in the real world is non-trivial, which\nnecessitates the use of simulators for validation. However,the high\ncomputational cost and the domain gap in data distribution have hindered the\nseamless transition between real and simulated driving scenarios. To tackle\nthis challenge, we propose Retrieval-Augmented Learning for Autonomous Driving\n(RALAD), a novel framework designed to bridge the real-to-sim gap at a low\ncost. RALAD features three primary designs, including (1) domain adaptation via\nan enhanced Optimal Transport (OT) method that accounts for both individual and\ngrouped image distances, (2) a simple and unified framework that can be applied\nto various models, and (3) efficient fine-tuning techniques that freeze the\ncomputationally expensive layers while maintaining robustness. Experimental\nresults demonstrate that RALAD compensates for the performance degradation in\nsimulated environments while maintaining accuracy in real-world scenarios\nacross three different models. Taking Cross View as an example, the mIOU and\nmAP metrics in real-world scenarios remain stable before and after RALAD\nfine-tuning, while in simulated environments,the mIOU and mAP metrics are\nimproved by 10.30% and 12.29%, respectively. Moreover, the re-training cost of\nour approach is reduced by approximately 88.1%. Our code is available at\nhttps://github.com/JiachengZuo/RALAD.git.", + "arxiv_id": "2501.12296v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets mandatory criteria 1 (practical applications in autonomous driving/retrieval-augmented learning), 2 (quantitative metrics like 10.30% mIOU improvement), and 3 (implied SOTA comparison through efficiency claims). Also addresses real-world challenges (criteria 5) and demonstrates reproducibility with code. Autonomous driving focus does not trigger rejection categories." + }, + { + "paper": { + "title": "LLM-Assisted Knowledge Graph Completion for Curriculum and Domain\n Modelling in Personalized Higher Education Recommendations", + "abstract": "While learning personalization offers great potential for learners, modern\npractices in higher education require a deeper consideration of domain models\nand learning contexts, to develop effective personalization algorithms. This\npaper introduces an innovative approach to higher education curriculum\nmodelling that utilizes large language models (LLMs) for knowledge graph (KG)\ncompletion, with the goal of creating personalized learning-path\nrecommendations. Our research focuses on modelling university subjects and\nlinking their topics to corresponding domain models, enabling the integration\nof learning modules from different faculties and institutions in the student's\nlearning path. Central to our approach is a collaborative process, where LLMs\nassist human experts in extracting high-quality, fine-grained topics from\nlecture materials. We develop a domain, curriculum, and user models for\nuniversity modules and stakeholders. We implement this model to create the KG\nfrom two study modules: Embedded Systems and Development of Embedded Systems\nUsing FPGA. The resulting KG structures the curriculum and links it to the\ndomain models. We evaluate our approach through qualitative expert feedback and\nquantitative graph quality metrics. Domain experts validated the relevance and\naccuracy of the model, while the graph quality metrics measured the structural\nproperties of our KG. Our results show that the LLM-assisted graph completion\napproach enhances the ability to connect related courses across disciplines to\npersonalize the learning experience. Expert feedback also showed high\nacceptance of the proposed collaborative approach for concept extraction and\nclassification.", + "arxiv_id": "2501.12300v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (LLM/KG practical application in education), 2 (quantitative metrics and expert validation), and 4 (detailed methodology). Addresses novelty through LLM-assisted KG completion. Does not trigger rejection categories. Satisfies 3 primary criteria with implementable approach." + }, + { + "paper": { + "title": "Treefix: Enabling Execution with a Tree of Prefixes", + "abstract": "The ability to execute code is a prerequisite for various dynamic program\nanalyses. Learning-guided execution has been proposed as an approach to enable\nthe execution of arbitrary code snippets by letting a neural model predict\nlikely values for any missing variables. Although state-of-the-art\nlearning-guided execution approaches, such as LExecutor, can enable the\nexecution of a relative high amount of code, they are limited to predicting a\nrestricted set of possible values and do not use any feedback from previous\nexecutions to execute even more code. This paper presents Treefix, a novel\nlearning-guided execution approach that leverages LLMs to iteratively create\ncode prefixes that enable the execution of a given code snippet. The approach\naddresses the problem in a multi-step fashion, where each step uses feedback\nabout the code snippet and its execution to instruct an LLM to improve a\npreviously generated prefix. This process iteratively creates a tree of\nprefixes, a subset of which is returned to the user as prefixes that maximize\nthe number of executed lines in the code snippet. In our experiments with two\ndatasets of Python code snippets, Treefix achieves 25% and 7% more coverage\nrelative to the current state of the art in learning-guided execution, covering\na total of 84% and 82% of all lines in the code snippets.", + "arxiv_id": "2501.12339v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Practical application in code execution using LLMs, 2) Quantitative metrics showing 25% and 7% improvements, 3) Comparison with LExecutor as state-of-the-art. Also addresses methodology (criterion 4) through iterative prefix generation and real-world code analysis applications (criterion 5). No rejection triggers present." + }, + { + "paper": { + "title": "Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for\n Mixture-of-Experts Language Models", + "abstract": "Scaling the capacity of language models has consistently proven to be a\nreliable approach for improving performance and unlocking new capabilities.\nCapacity can be primarily defined by two dimensions: the number of model\nparameters and the compute per example. While scaling typically involves\nincreasing both, the precise interplay between these factors and their combined\ncontribution to overall capacity remains not fully understood. We explore this\nrelationship in the context of sparse Mixture-of-Expert models (MoEs), which\nallow scaling the number of parameters without proportionally increasing the\nFLOPs per example. We investigate how varying the sparsity level, i.e., the\nratio of non-active to total parameters, affects model performance in terms of\nboth pretraining and downstream performance. We find that under different\nconstraints (e.g. parameter size and total training compute), there is an\noptimal level of sparsity that improves both training efficiency and model\nperformance. These results provide a better understanding of the impact of\nsparsity in scaling laws for MoEs and complement existing works in this area,\noffering insights for designing more efficient architectures.", + "arxiv_id": "2501.12370v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets mandatory criteria 1 (practical applications in scaling LLMs via MoE architectures), 2 (experimental results with quantitative metrics on sparsity optimization), and 3 (comparison with existing scaling laws). It also addresses optional criterion 4 (methodology details for sparsity optimization) and demonstrates novelty in scaling law analysis for sparse MoEs. No rejection triggers apply." + }, + { + "paper": { + "title": "Is Long Context All You Need? Leveraging LLM's Extended Context for\n NL2SQL", + "abstract": "Large Language Models (LLMs) have demonstrated impressive capabilities across\na range of natural language processing tasks. In particular, improvements in\nreasoning abilities and the expansion of context windows have opened new\navenues for leveraging these powerful models. NL2SQL is challenging in that the\nnatural language question is inherently ambiguous, while the SQL generation\nrequires a precise understanding of complex data schema and semantics. One\napproach to this semantic ambiguous problem is to provide more and sufficient\ncontextual information.\n In this work, we explore the performance and the latency trade-offs of the\nextended context window (a.k.a., long context) offered by Google's\nstate-of-the-art LLM (\\textit{gemini-1.5-pro}). We study the impact of various\ncontextual information, including column example values, question and SQL query\npairs, user-provided hints, SQL documentation, and schema. To the best of our\nknowledge, this is the first work to study how the extended context window and\nextra contextual information can help NL2SQL generation with respect to both\naccuracy and latency cost. We show that long context LLMs are robust and do not\nget lost in the extended contextual information. Additionally, our long-context\nNL2SQL pipeline based on Google's \\textit{gemini-pro-1.5} achieve a strong\nperformance with 67.41\\% on BIRD benchmark (dev) without finetuning and\nexpensive self-consistency based techniques.", + "arxiv_id": "2501.12372v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all three mandatory criteria: 1) Focuses on practical NL2SQL applications using LLMs' extended context, 2) Reports quantitative results (67.41% on BIRD benchmark) and latency analysis, 3) Compares with SOTA techniques and claims first work status. Also addresses methodology (C4) and implements with current tools. No exclusion triggers apply." + }, + { + "paper": { + "title": "Code Readability in the Age of Large Language Models: An Industrial Case\n Study from Atlassian", + "abstract": "Programmers spend a significant amount of time reading code during the\nsoftware development process. This trend is amplified by the emergence of large\nlanguage models (LLMs) that automatically generate code. However, little is\nknown about the readability of the LLM-generated code and whether it is still\nimportant from practitioners' perspectives in this new era. In this paper, we\nconduct a survey to explore the practitioners' perspectives on code readability\nin the age of LLMs and investigate the readability of our LLM-based software\ndevelopment agents framework, HULA, by comparing its generated code with\nhuman-written code in real-world scenarios. Overall, the findings underscore\nthat (1) readability remains a critical aspect of software development; (2) the\nreadability of our LLM-generated code is comparable to human-written code,\nfostering the establishment of appropriate trust and driving the broad adoption\nof our LLM-powered software development platform.", + "arxiv_id": "2501.11264v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets ALL mandatory criteria: 1 (practical LLM application in software development), 2 (experimental comparison of LLM-generated vs human code), 3 (comparison with human-written code as baseline). Also meets optional criteria 4/5 (methodology described for HULA framework and real-world case study). No rejection triggers present." + }, + { + "paper": { + "title": "RedStar: Does Scaling Long-CoT Data Unlock Better Slow-Reasoning\n Systems?", + "abstract": "Can scaling transform reasoning? In this work, we explore the untapped\npotential of scaling Long Chain-of-Thought (Long-CoT) data to 1000k samples,\npioneering the development of a slow-thinking model, RedStar. Through extensive\nexperiments with various LLMs and different sizes, we uncover the ingredients\nfor specialization and scale for Long-CoT training. Surprisingly, even smaller\nmodels show significant performance gains with limited data, revealing the\nsample efficiency of Long-CoT and the critical role of sample difficulty in the\nlearning process. Our findings demonstrate that Long-CoT reasoning can be\neffectively triggered with just a few thousand examples, while larger models\nachieve unparalleled improvements. We also introduce reinforcement learning\n(RL)-scale training as a promising direction for advancing slow-thinking\nsystems. RedStar shines across domains: on the MATH-Hard benchmark,\nRedStar-code-math boosts performance from 66.2\\% to 81.6\\%, and on the USA Math\nOlympiad (AIME), it solves 46.7\\% of problems using only 21k mixed-code-math\ndatasets. In multimodal tasks like GeoQA and MathVista-GEO, RedStar-Geo\nachieves competitive results with minimal Long-CoT data, outperforming other\nslow-thinking systems like QvQ-Preview. Compared to QwQ, RedStar strikes the\nperfect balance between reasoning and generalizability. Our work highlights\nthat, with careful tuning, scaling Long-CoT can unlock extraordinary reasoning\ncapabilities-even with limited dataset and set a new standard for slow-thinking\nmodels across diverse challenges. Our data and models are released at\nhttps://huggingface.co/RedStar-Reasoning.", + "arxiv_id": "2501.11284v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Focuses on practical reasoning improvements through scaled Long-CoT training, 2) Provides quantitative results (e.g., 66.2%→81.6% on MATH-Hard), 3) Compares with SOTA systems like QvQ-Preview. Also satisfies optional criteria 4 (methodology details) and 5 (real-world math/multimodal applications). Avoids rejection categories while demonstrating novelty in slow-reasoning systems." + }, + { + "paper": { + "title": "Graph-defined Language Learning with LLMs", + "abstract": "Recent efforts leverage Large Language Models (LLMs) for modeling\ntext-attributed graph structures in node classification tasks. These approaches\ndescribe graph structures for LLMs to understand or aggregate LLM-generated\ntextual attribute embeddings through graph structure. However, these approaches\nface two main limitations in modeling graph structures with LLMs. (i) Graph\ndescriptions become verbose in describing high-order graph structure. (ii)\nTextual attributes alone do not contain adequate graph structure information.\nIt is challenging to model graph structure concisely and adequately with LLMs.\nLLMs lack built-in mechanisms to model graph structures directly. They also\nstruggle with complex long-range dependencies between high-order nodes and\ntarget nodes.\n Inspired by the observation that LLMs pre-trained on one language can achieve\nexceptional performance on another with minimal additional training, we propose\n\\textbf{G}raph-\\textbf{D}efined \\textbf{L}anguage for \\textbf{L}arge\n\\textbf{L}anguage \\textbf{M}odel (GDL4LLM). This novel framework enables LLMs\nto transfer their powerful language understanding capabilities to\ngraph-structured data. GDL4LLM translates graphs into a graph language corpus\ninstead of graph descriptions and pre-trains LLMs on this corpus to adequately\nunderstand graph structures. During fine-tuning, this corpus describes the\nstructural information of target nodes concisely with only a few tokens. By\ntreating graphs as a new language, GDL4LLM enables LLMs to model graph\nstructures adequately and concisely for node classification tasks. Extensive\nexperiments on three real-world datasets demonstrate that GDL4LLM outperforms\ndescription-based and textual attribute embeddings-based baselines by\nefficiently modeling different orders of graph structure with LLMs.", + "arxiv_id": "2501.11478v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Focuses on practical LLM applications with knowledge graphs, 2) Includes experimental results with real datasets and quantitative metrics, 3) Compares with SOTA baselines. Also meets optional criterion 4 through methodology description of graph-language translation. Novel approach (GDL4LLM) aligns with additional considerations. No rejection triggers present." + }, + { + "paper": { + "title": "Early evidence of how LLMs outperform traditional systems on OCR/HTR\n tasks for historical records", + "abstract": "We explore the ability of two LLMs -- GPT-4o and Claude Sonnet 3.5 -- to\ntranscribe historical handwritten documents in a tabular format and compare\ntheir performance to traditional OCR/HTR systems: EasyOCR, Keras, Pytesseract,\nand TrOCR. Considering the tabular form of the data, two types of experiments\nare executed: one where the images are split line by line and the other where\nthe entire scan is used as input. Based on CER and BLEU, we demonstrate that\nLLMs outperform the conventional OCR/HTR methods. Moreover, we also compare the\nevaluated CER and BLEU scores to human evaluations to better judge the outputs\nof whole-scan experiments and understand influential factors for CER and BLEU.\nCombining judgments from all the evaluation metrics, we conclude that two-shot\nGPT-4o for line-by-line images and two-shot Claude Sonnet 3.5 for whole-scan\nimages yield the transcriptions of the historical records most similar to the\nground truth.", + "arxiv_id": "2501.11623v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets ALL mandatory criteria: 1) Practical OCR application using LLMs, 2) Quantitative metrics (CER/BLEU) and comparisons, 3) Direct comparison with traditional OCR systems. Also meets optional criterion 4 through methodology description. No rejection triggers apply." + }, + { + "paper": { + "title": "Policy-Adaptable Methods For Resolving Normative Conflicts Through\n Argumentation and Graph Colouring", + "abstract": "In a multi-agent system, one may choose to govern the behaviour of an agent\nby imposing norms, which act as guidelines for how agents should act either all\nof the time or in given situations. However, imposing multiple norms on one or\nmore agents may result in situations where these norms conflict over how the\nagent should behave. In any system with normative conflicts (such as safe\nreinforcement models or systems which monitor safety protocols), one must\ndecide which norms should be followed such that the most important and most\nrelevant norms are maintained. We introduce a new method for resolving\nnormative conflicts through argumentation and graph colouring which is\ncompatible with a variety of normative conflict resolution policies. We prove\nthat this method always creates an admissible set of arguments under\nargumentation semantics, meaning that it produces coherent outputs. We also\nintroduce more robust variants of this method, each building upon their\npredecessor to create a superior output, and we include further mathematical\nproof of their coherence. Our most advanced variant uses the existing concept\nof curtailment, where one norm may supersede another without fully eliminating\nit. The methods we introduce are all compatible with various pre-existing\npolicies for resolving normative conflicts. Empirical evaluations are also\nperformed to compare our algorithms to each other and to others in existing\nliterature.", + "arxiv_id": "2501.11799v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all mandatory criteria: 1) Addresses practical applications in agentic AI (normative conflict resolution for multi-agent systems), 2) Includes empirical evaluations and mathematical proofs as quantitative metrics, 3) Compares with existing literature. It also addresses optional criteria 4 (methodology details via argumentation/graph coloring variants) and 5 (real-world applications like safety protocols). While focused on policy/conflict resolution, it does not fall under explicit rejection categories like medical/social/law applications." + }, + { + "paper": { + "title": "Network-informed Prompt Engineering against Organized Astroturf\n Campaigns under Extreme Class Imbalance", + "abstract": "Detecting organized political campaigns is of paramount importance in\nfighting against disinformation on social media. Existing approaches for the\nidentification of such organized actions employ techniques mostly from network\nscience, graph machine learning and natural language processing. Their ultimate\ngoal is to analyze the relationships and interactions (e.g. re-posting) among\nusers and the textual similarities of their posts. Despite their effectiveness\nin recognizing astroturf campaigns, these methods face significant challenges,\nnotably the class imbalance in available training datasets. To mitigate this\nissue, recent methods usually resort to data augmentation or increasing the\nnumber of positive samples, which may not always be feasible or sufficient in\nreal-world settings. Following a different path, in this paper, we propose a\nnovel framework for identifying astroturf campaigns based solely on large\nlanguage models (LLMs), introducing a Balanced Retrieval-Augmented Generation\n(Balanced RAG) component. Our approach first gives both textual information\nconcerning the posts (in our case tweets) and the user interactions of the\nsocial network as input to a language model. Then, through prompt engineering\nand the proposed Balanced RAG method, it effectively detects coordinated\ndisinformation campaigns on X (Twitter). The proposed framework does not\nrequire any training or fine-tuning of the language model. Instead, by\nstrategically harnessing the strengths of prompt engineering and Balanced RAG,\nit facilitates LLMs to overcome the effects of class imbalance and effectively\nidentify coordinated political campaigns. The experimental results demonstrate\nthat by incorporating the proposed prompt engineering and Balanced RAG methods,\nour framework outperforms the traditional graph-based baselines, achieving\n2x-3x improvements in terms of precision, recall and F1 scores.", + "arxiv_id": "2501.11849v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets mandatory criteria 1 (Practical Applications in RAG/LLMs), 2 (Quantitative metrics showing 2x-3x improvements), and 3 (Comparison with graph-based baselines). It also satisfies optional criteria 4 (Methodology details for LLM-based detection) and 5 (real-world class imbalance challenge). While addressing disinformation campaigns, its primary technical focus on LLM/RAG/prompt engineering outweighs social application concerns under rejection criteria." + }, + { + "paper": { + "title": "EDoRA: Efficient Weight-Decomposed Low-Rank Adaptation via Singular\n Value Decomposition", + "abstract": "Parameter-efficient fine-tuning methods, such as LoRA, reduces the number of\ntrainable parameters. However, they often suffer from scalability issues and\ndifferences between their learning pattern and full fine-tuning. To overcome\nthese limitations, we propose Efficient Weight-Decomposed Low-Rank Adaptation\n(EDoRA): a novel PEFT method that decomposes pre-trained weights into magnitude\nand directional components. By freezing low-rank matrices, initializing them by\nsingular value decomposition, and introducing a small trainable matrix between\nthem, EDoRA achieves substantial reduction in trainable parameters while\nmaintaining learning capacity. Experimental results on the GLUE benchmark\ndemonstrate that EDoRA achieves competitive or superior performance compared to\nstate-of-the-art methods, such as LoRA and DoRA, with up to 30x fewer trainable\nparameters. This makes EDoRA a highly efficient solution for adapting LLMs to\ndiverse tasks under memory-constrained settings. Code is available at\nhttps://github.com/Hamid-Nasiri/EDoRA .", + "arxiv_id": "2501.12067v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all three mandatory criteria: 1) Focuses on practical LLM adaptation via parameter-efficient fine-tuning, 2) Provides GLUE benchmark results with quantitative metrics (30x parameter reduction), and 3) Compares with SOTA methods like LoRA/DoRA. Also meets optional criteria 4 (clear decomposition methodology) and 5 (addresses memory constraints). No rejection triggers apply." + }, + { + "paper": { + "title": "Improving Influence-based Instruction Tuning Data Selection for Balanced\n Learning of Diverse Capabilities", + "abstract": "Selecting appropriate training data is crucial for effective instruction\nfine-tuning of large language models (LLMs), which aims to (1) elicit strong\ncapabilities, and (2) achieve balanced performance across a diverse range of\ntasks. Influence-based methods show promise in achieving (1) by estimating the\ncontribution of each training example to the model's predictions, but often\nstruggle with (2). Our systematic investigation reveals that this\nunderperformance can be attributed to an inherent bias where certain tasks\nintrinsically have greater influence than others. As a result, data selection\nis often biased towards these tasks, not only hurting the model's performance\non others but also, counterintuitively, harms performance on these\nhigh-influence tasks themselves.\n As a remedy, we propose BIDS, a Balanced and Influential Data Selection\nalgorithm. BIDS first normalizes influence scores of the training data, and\nthen iteratively balances data selection by choosing the training example with\nthe highest influence on the most underrepresented task. Experiments with both\nLlama-3 and Mistral-v0.3 on seven benchmarks spanning five diverse capabilities\nshow that BIDS consistently outperforms both state-of-the-art influence-based\nalgorithms and other non-influence-based selection frameworks. Surprisingly,\ntraining on a 15% subset selected by BIDS can even outperform full-dataset\ntraining with a much more balanced performance. Our analysis further highlights\nthe importance of both instance-level normalization and iterative optimization\nof selected data for balanced learning of diverse capabilities.", + "arxiv_id": "2501.12147v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Focuses on practical LLM training optimization through data selection, 2) Includes experiments with Llama-3/Mistral and quantitative metrics showing 15% subset outperforms full training, 3) Compares with SOTA influence-based methods. Also meets optional criteria 4 (detailed BIDS methodology) and demonstrates novelty in balanced data selection. No rejection triggers present." + }, + { + "paper": { + "title": "On the practical applicability of modern DFT functionals for chemical\n computations. Case study of DM21 applicability for geometry optimization", + "abstract": "Density functional theory (DFT) is probably the most promising approach for\nquantum chemistry calculations considering its good balance between\ncalculations precision and speed. In recent years, several neural network-based\nfunctionals have been developed for exchange-correlation energy approximation\nin DFT, DM21 developed by Google Deepmind being the most notable between them.\nThis study focuses on evaluating the efficiency of DM21 functional in\npredicting molecular geometries, with a focus on the influence of oscillatory\nbehavior in neural network exchange-correlation functionals. We implemented\ngeometry optimization in PySCF for the DM21 functional in geometry optimization\nproblem, compared its performance with traditional functionals, and tested it\non various benchmarks. Our findings reveal both the potential and the current\nchallenges of using neural network functionals for geometry optimization in\nDFT. We propose a solution extending the practical applicability of such\nfunctionals and allowing to model new substances with their help.", + "arxiv_id": "2501.12149v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Focuses on practical chemical computation applications of neural network functionals (DM21), 2) Includes experimental comparisons and benchmarks, 3) Compares with traditional functionals. Also addresses optional criterion 4 (methodology details in PySCF implementation) and 5 (real-world substance modeling challenges). No rejection triggers present." + }, + { + "paper": { + "title": "AdaServe: SLO-Customized LLM Serving with Fine-Grained Speculative\n Decoding", + "abstract": "This paper introduces AdaServe, the first LLM serving system to support SLO\ncustomization through fine-grained speculative decoding. AdaServe leverages the\nlogits of a draft model to predict the speculative accuracy of tokens and\nemploys a theoretically optimal algorithm to construct token trees for\nverification. To accommodate diverse SLO requirements without compromising\nthroughput, AdaServe employs a speculation-and-selection scheme that first\nconstructs candidate token trees for each request and then dynamically selects\ntokens to meet individual SLO constraints while optimizing throughput.\nComprehensive evaluations demonstrate that AdaServe achieves up to 73% higher\nSLO attainment and 74% higher goodput compared to state-of-the-art systems.\nThese results underscore AdaServe's potential to enhance the efficiency and\nadaptability of LLM deployments across varied application scenarios.", + "arxiv_id": "2501.12162v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all 3 mandatory criteria: 1) Practical applications in LLM deployment optimization, 2) Quantitative results showing 73% SLO improvement, 3) Comparison with SOTA systems. Also addresses methodology (criterion 4) and shows implementable approach. No rejection triggers present." + }, + { + "paper": { + "title": "InsTALL: Context-aware Instructional Task Assistance with Multi-modal\n Large Language Models", + "abstract": "The improved competence of generative models can help building multi-modal\nvirtual assistants that leverage modalities beyond language. By observing\nhumans performing multi-step tasks, one can build assistants that have\nsituational awareness of actions and tasks being performed, enabling them to\ncater assistance based on this understanding. In this paper, we develop a\nContext-aware Instructional Task Assistant with Multi-modal Large Language\nModels (InsTALL) that leverages an online visual stream (e.g. a user's screen\nshare or video recording) and responds in real-time to user queries related to\nthe task at hand. To enable useful assistance, InsTALL 1) trains a multi-modal\nmodel on task videos and paired textual data, and 2) automatically extracts\ntask graph from video data and leverages it at training and inference time. We\nshow InsTALL achieves state-of-the-art performance across proposed sub-tasks\nconsidered for multimodal activity understanding -- task recognition (TR),\naction recognition (AR), next action prediction (AP), and plan prediction (PP)\n-- and outperforms existing baselines on two novel sub-tasks related to\nautomatic error identification.", + "arxiv_id": "2501.12231v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets all mandatory criteria: 1) Practical application in agentic AI via multi-modal task assistance, 2) Quantitative results showing SOTA performance across multiple sub-tasks, 3) Explicit comparison with existing baselines. Also satisfies optional criteria 4 (methodology details) and 5 (real-world application). No rejection triggers present." + }, + { + "paper": { + "title": "UI-TARS: Pioneering Automated GUI Interaction with Native Agents", + "abstract": "This paper introduces UI-TARS, a native GUI agent model that solely perceives\nthe screenshots as input and performs human-like interactions (e.g., keyboard\nand mouse operations). Unlike prevailing agent frameworks that depend on\nheavily wrapped commercial models (e.g., GPT-4o) with expert-crafted prompts\nand workflows, UI-TARS is an end-to-end model that outperforms these\nsophisticated frameworks. Experiments demonstrate its superior performance:\nUI-TARS achieves SOTA performance in 10+ GUI agent benchmarks evaluating\nperception, grounding, and GUI task execution. Notably, in the OSWorld\nbenchmark, UI-TARS achieves scores of 24.6 with 50 steps and 22.7 with 15\nsteps, outperforming Claude (22.0 and 14.9 respectively). In AndroidWorld,\nUI-TARS achieves 46.6, surpassing GPT-4o (34.5). UI-TARS incorporates several\nkey innovations: (1) Enhanced Perception: leveraging a large-scale dataset of\nGUI screenshots for context-aware understanding of UI elements and precise\ncaptioning; (2) Unified Action Modeling, which standardizes actions into a\nunified space across platforms and achieves precise grounding and interaction\nthrough large-scale action traces; (3) System-2 Reasoning, which incorporates\ndeliberate reasoning into multi-step decision making, involving multiple\nreasoning patterns such as task decomposition, reflection thinking, milestone\nrecognition, etc. (4) Iterative Training with Reflective Online Traces, which\naddresses the data bottleneck by automatically collecting, filtering, and\nreflectively refining new interaction traces on hundreds of virtual machines.\nThrough iterative training and reflection tuning, UI-TARS continuously learns\nfrom its mistakes and adapts to unforeseen situations with minimal human\nintervention. We also analyze the evolution path of GUI agents to guide the\nfurther development of this domain.", + "arxiv_id": "2501.12326v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets ALL mandatory criteria: 1) Practical GUI interaction applications in agentic AI, 2) Quantitative results showing SOTA performance across multiple benchmarks, 3) Direct comparisons with Claude/GPT-4o. Also satisfies Criterion 4 with detailed methodology descriptions. Novel agentic AI implementation with reproducible training approach. No rejection triggers present." + }, + { + "paper": { + "title": "Automatic Labelling with Open-source LLMs using Dynamic Label Schema\n Integration", + "abstract": "Acquiring labelled training data remains a costly task in real world machine\nlearning projects to meet quantity and quality requirements. Recently Large\nLanguage Models (LLMs), notably GPT-4, have shown great promises in labelling\ndata with high accuracy. However, privacy and cost concerns prevent the\nubiquitous use of GPT-4. In this work, we explore effectively leveraging\nopen-source models for automatic labelling. We identify integrating label\nschema as a promising technology but found that naively using the label\ndescription for classification leads to poor performance on high cardinality\ntasks. To address this, we propose Retrieval Augmented Classification (RAC) for\nwhich LLM performs inferences for one label at a time using corresponding label\nschema; we start with the most related label and iterates until a label is\nchosen by the LLM. We show that our method, which dynamically integrates label\ndescription, leads to performance improvements in labelling tasks. We further\nshow that by focusing only on the most promising labels, RAC can trade off\nbetween label quality and coverage - a property we leverage to automatically\nlabel our internal datasets.", + "arxiv_id": "2501.12332v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets criteria 1 (practical application in automatic labeling with LLMs), 2 (experimental results showing performance improvements), and 4 (methodology details on RAC). While explicit comparison with SOTA is unclear, the focus on dynamic label integration and trade-offs aligns with inclusivity guidelines. No rejection triggers apply." + }, + { + "paper": { + "title": "Expertise elevates AI usage: experimental evidence comparing laypeople\n and professional artists", + "abstract": "Novel capacities of generative AI to analyze and generate cultural artifacts\nraise inevitable questions about the nature and value of artistic education and\nhuman expertise. Has AI already leveled the playing field between professional\nartists and laypeople, or do trained artistic expressive capacity, curation\nskills and experience instead enhance the ability to use these new tools? In\nthis pre-registered study, we conduct experimental comparisons between 50\nactive artists and a demographically matched sample of laypeople. We designed\ntwo tasks to approximate artistic practice for testing their capabilities in\nboth faithful and creative image creation: replicating a reference image, and\nmoving as far away as possible from it. We developed a bespoke platform where\nparticipants used a modern text-to-image model to complete both tasks. We also\ncollected and compared participants' sentiments towards AI. On average, artists\nproduced more faithful and creative outputs than their lay counterparts,\nalthough only by a small margin. While AI may ease content creation,\nprofessional expertise is still valuable - even within the confined space of\ngenerative AI itself. Finally, we also explored how well an exemplary\nvision-capable large language model (GPT-4o) would complete the same tasks, if\ngiven the role of an image generation agent, and found it performed on par in\ncopying but outperformed even artists in the creative task. The very best\nresults were still produced by humans in both tasks. These outcomes highlight\nthe importance of integrating artistic skills with AI training to prepare\nartists and other visual professionals for a technologically evolving\nlandscape. We see a potential in collaborative synergy with generative AI,\nwhich could reshape creative industries and education in the arts.", + "arxiv_id": "2501.12374v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nMeets criteria 1 (practical applications in creative industries/education), 2 (experimental comparisons with quantitative performance metrics), and 3 (comparison between human experts and GPT-4o). Also meets criterion 5 through discussion of real-world applications and AI integration challenges. While marginally relevant to social aspects, core focus aligns with LLM applications in agentic AI and human-AI collaboration." + }, + { + "paper": { + "title": "Conversation Routines: A Prompt Engineering Framework for Task-Oriented\n Dialog Systems", + "abstract": "This study introduces Conversation Routines (CR), a structured prompt\nengineering framework for developing task-oriented dialog systems using Large\nLanguage Models (LLMs). While LLMs demonstrate remarkable natural language\nunderstanding capabilities, engineering them to reliably execute complex\nbusiness workflows remains challenging. The proposed CR framework enables the\ndevelopment of Conversation Agentic Systems (CAS) through natural language\nspecifications, embedding task-oriented logic within LLM prompts. This approach\nprovides a systematic methodology for designing and implementing complex\nconversational workflows while maintaining behavioral consistency. We\ndemonstrate the framework's effectiveness through two proof of concept\nimplementations: a Train Ticket Booking System and an Interactive\nTroubleshooting Copilot. These case studies validate CR's capability to encode\nsophisticated behavioral patterns and decision logic while preserving natural\nconversational flexibility. Results show that CR enables domain experts to\ndesign conversational workflows in natural language while leveraging custom\nenterprise functionalities (tools) developed by software engineers, creating an\nefficient division of responsibilities where developers focus on core API\nimplementation and domain experts handle conversation design. While the\nframework shows promise in accessibility and adaptability, we identify key\nchallenges including computational overhead, non-deterministic behavior, and\ndomain-specific logic optimization. Future research directions include\nenhancing system robustness, improving scalability for complex multi-agent\ninteractions, and addressing the identified limitations across diverse business\napplications.", + "arxiv_id": "2501.11613v1" + }, + "decision": "ACCEPT", + "explanation": "Original decision: ACCEPT\nThe paper meets all three mandatory criteria: 1) Focuses on practical LLM applications (task-oriented dialog systems), 2) Includes experimental results with case studies (Train Ticket Booking and Troubleshooting Copilot), 3) Implicitly addresses advancements via workflow consistency and division of responsibilities. It also satisfies optional criteria 4 (methodology details) and 5 (real-world challenges). No rejection triggers apply." + } + ], + "rejected": [ + { + "paper": { + "title": "The Explanation Game -- Rekindled (Extended Version)", + "abstract": "Recent work demonstrated the existence of critical flaws in the current use\nof Shapley values in explainable AI (XAI), i.e. the so-called SHAP scores.\nThese flaws are significant in that the scores provided to a human\ndecision-maker can be misleading. Although these negative results might appear\nto indicate that Shapley values ought not be used in XAI, this paper argues\notherwise. Concretely, this paper proposes a novel definition of SHAP scores\nthat overcomes existing flaws. Furthermore, the paper outlines a practically\nefficient solution for the rigorous estimation of the novel SHAP scores.\nPreliminary experimental results confirm our claims, and further underscore the\nflaws of the current SHAP scores.", + "arxiv_id": "2501.11429v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on explainable AI (XAI) and Shapley values, which relates to responsible AI/ethics. It does not explicitly address LLMs in practical applications (Criterion 1 unmet). While it meets Criteria 2 (experimental results) and 3 (comparison with SOTA), it violates the rejection rule for focusing on responsible AI applications." + }, + { + "paper": { + "title": "Bridging the Communication Gap: Evaluating AI Labeling Practices for\n Trustworthy AI Development", + "abstract": "As artificial intelligence (AI) becomes integral to economy and society,\ncommunication gaps between developers, users, and stakeholders hinder trust and\ninformed decision-making. High-level AI labels, inspired by frameworks like EU\nenergy labels, have been proposed to make the properties of AI models more\ntransparent. Without requiring deep technical expertise, they can inform on the\ntrade-off between predictive performance and resource efficiency. However, the\npractical benefits and limitations of AI labeling remain underexplored. This\nstudy evaluates AI labeling through qualitative interviews along four key\nresearch questions. Based on thematic analysis and inductive coding, we found a\nbroad range of practitioners to be interested in AI labeling (RQ1). They see\nbenefits for alleviating communication gaps and aiding non-expert\ndecision-makers, however limitations, misunderstandings, and suggestions for\nimprovement were also discussed (RQ2). Compared to other reporting formats,\ninterviewees positively evaluated the reduced complexity of labels, increasing\noverall comprehensibility (RQ3). Trust was influenced most by usability and the\ncredibility of the responsible labeling authority, with mixed preferences for\nself-certification versus third-party certification (RQ4). Our Insights\nhighlight that AI labels pose a trade-off between simplicity and complexity,\nwhich could be resolved by developing customizable and interactive labeling\nframeworks to address diverse user needs. Transparent labeling of resource\nefficiency also nudged interviewee priorities towards paying more attention to\nsustainability aspects during AI development. This study validates AI labels as\na valuable tool for enhancing trust and communication in AI, offering\nactionable guidelines for their refinement and standardization.", + "arxiv_id": "2501.11909v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on responsible AI applications (trustworthy AI development through labeling practices) and social communication aspects, which falls under the rejection criteria. It does not sufficiently address practical LLM applications, experimental results with quantitative metrics, or comparisons with state-of-the-art methods in required technical domains." + }, + { + "paper": { + "title": "Make Full Use of Testing Information: An Integrated Accelerated Testing\n and Evaluation Method for Autonomous Driving Systems", + "abstract": "Testing and evaluation is an important step before the large-scale\napplication of the autonomous driving systems (ADSs). Based on the three level\nof scenario abstraction theory, a testing can be performed within a logical\nscenario, followed by an evaluation stage which is inputted with the testing\nresults of each concrete scenario generated from the logical parameter space.\nDuring the above process, abundant testing information is produced which is\nbeneficial for comprehensive and accurate evaluations. To make full use of\ntesting information, this paper proposes an Integrated accelerated Testing and\nEvaluation Method (ITEM). Based on a Monte Carlo Tree Search (MCTS) paradigm\nand a dual surrogates testing framework proposed in our previous work, this\npaper applies the intermediate information (i.e., the tree structure, including\nthe affiliation of each historical sampled point with the subspaces and the\nparent-child relationship between subspaces) generated during the testing stage\ninto the evaluation stage to achieve accurate hazardous domain identification.\nMoreover, to better serve this purpose, the UCB calculation method is improved\nto allow the search algorithm to focus more on the hazardous domain boundaries.\nFurther, a stopping condition is constructed based on the convergence of the\nsearch algorithm. Ablation and comparative experiments are then conducted to\nverify the effectiveness of the improvements and the superiority of the\nproposed method. The experimental results show that ITEM could well identify\nthe hazardous domains in both low- and high-dimensional cases, regardless of\nthe shape of the hazardous domains, indicating its generality and potential for\nthe safety evaluation of ADSs.", + "arxiv_id": "2501.11924v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on autonomous driving system testing using MCTS and surrogates but does not explicitly mention Large Language Models (LLMs) or their practical applications in areas like knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparisons), it fails the mandatory criterion 1 (LLM applications). Additionally, autonomous driving testing does not fall under the specified rejection categories but lacks required LLM relevance." + }, + { + "paper": { + "title": "Collaborative Imputation of Urban Time Series through Cross-city\n Meta-learning", + "abstract": "Urban time series, such as mobility flows, energy consumption, and pollution\nrecords, encapsulate complex urban dynamics and structures. However, data\ncollection in each city is impeded by technical challenges such as budget\nlimitations and sensor failures, necessitating effective data imputation\ntechniques that can enhance data quality and reliability. Existing imputation\nmodels, categorized into learning-based and analytics-based paradigms, grapple\nwith the trade-off between capacity and generalizability. Collaborative\nlearning to reconstruct data across multiple cities holds the promise of\nbreaking this trade-off. Nevertheless, urban data's inherent irregularity and\nheterogeneity issues exacerbate challenges of knowledge sharing and\ncollaboration across cities. To address these limitations, we propose a novel\ncollaborative imputation paradigm leveraging meta-learned implicit neural\nrepresentations (INRs). INRs offer a continuous mapping from domain coordinates\nto target values, integrating the strengths of both paradigms. By imposing\nembedding theory, we first employ continuous parameterization to handle\nirregularity and reconstruct the dynamical system. We then introduce a\ncross-city collaborative learning scheme through model-agnostic meta learning,\nincorporating hierarchical modulation and normalization techniques to\naccommodate multiscale representations and reduce variance in response to\nheterogeneity. Extensive experiments on a diverse urban dataset from 20 global\ncities demonstrate our model's superior imputation performance and\ngeneralizability, underscoring the effectiveness of collaborative imputation in\nresource-constrained settings.", + "arxiv_id": "2501.11306v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not explicitly mention Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it addresses practical urban data imputation and demonstrates experimental results with comparisons to existing methods, it focuses on meta-learning and neural representations rather than LLM-centered approaches, failing to meet the mandatory Criterion 1." + }, + { + "paper": { + "title": "Finer-CAM: Spotting the Difference Reveals Finer Details for Visual\n Explanation", + "abstract": "Class activation map (CAM) has been widely used to highlight image regions\nthat contribute to class predictions. Despite its simplicity and computational\nefficiency, CAM often struggles to identify discriminative regions that\ndistinguish visually similar fine-grained classes. Prior efforts address this\nlimitation by introducing more sophisticated explanation processes, but at the\ncost of extra complexity. In this paper, we propose Finer-CAM, a method that\nretains CAM's efficiency while achieving precise localization of discriminative\nregions. Our key insight is that the deficiency of CAM lies not in \"how\" it\nexplains, but in \"what\" it explains}. Specifically, previous methods attempt to\nidentify all cues contributing to the target class's logit value, which\ninadvertently also activates regions predictive of visually similar classes. By\nexplicitly comparing the target class with similar classes and spotting their\ndifferences, Finer-CAM suppresses features shared with other classes and\nemphasizes the unique, discriminative details of the target class. Finer-CAM is\neasy to implement, compatible with various CAM methods, and can be extended to\nmulti-modal models for accurate localization of specific concepts.\nAdditionally, Finer-CAM allows adjustable comparison strength, enabling users\nto selectively highlight coarse object contours or fine discriminative details.\nQuantitatively, we show that masking out the top 5% of activated pixels by\nFiner-CAM results in a larger relative confidence drop compared to baselines.\nThe source code and demo are available at\nhttps://github.com/Imageomics/Finer-CAM.", + "arxiv_id": "2501.11309v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on improving visual explanation methods (CAM) for image classification without mentioning LLMs or their applications in knowledge graphs, RAG, or agentic AI. It fails to meet mandatory criterion 1 (Practical Applications of LLMs) while meeting criteria 2-3. The methodology is vision-specific and unrelated to LLM-focused domains listed in the requirements." + }, + { + "paper": { + "title": "CatV2TON: Taming Diffusion Transformers for Vision-Based Virtual Try-On\n with Temporal Concatenation", + "abstract": "Virtual try-on (VTON) technology has gained attention due to its potential to\ntransform online retail by enabling realistic clothing visualization of images\nand videos. However, most existing methods struggle to achieve high-quality\nresults across image and video try-on tasks, especially in long video\nscenarios. In this work, we introduce CatV2TON, a simple and effective\nvision-based virtual try-on (V2TON) method that supports both image and video\ntry-on tasks with a single diffusion transformer model. By temporally\nconcatenating garment and person inputs and training on a mix of image and\nvideo datasets, CatV2TON achieves robust try-on performance across static and\ndynamic settings. For efficient long-video generation, we propose an\noverlapping clip-based inference strategy that uses sequential frame guidance\nand Adaptive Clip Normalization (AdaCN) to maintain temporal consistency with\nreduced resource demands. We also present ViViD-S, a refined video try-on\ndataset, achieved by filtering back-facing frames and applying 3D mask\nsmoothing for enhanced temporal consistency. Comprehensive experiments\ndemonstrate that CatV2TON outperforms existing methods in both image and video\ntry-on tasks, offering a versatile and reliable solution for realistic virtual\ntry-ons across diverse scenarios.", + "arxiv_id": "2501.11325v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing (virtual try-on for videos) and uses diffusion transformers rather than LLMs. It fails to meet mandatory criterion 1 (Practical Applications of LLMs) and falls under the video-processing rejection category." + }, + { + "paper": { + "title": "Towards Advancing Code Generation with Large Language Models: A Research\n Roadmap", + "abstract": "Recently, we have witnessed the rapid development of large language models,\nwhich have demonstrated excellent capabilities in the downstream task of code\ngeneration. However, despite their potential, LLM-based code generation still\nfaces numerous technical and evaluation challenges, particularly when embedded\nin real-world development. In this paper, we present our vision for current\nresearch directions, and provide an in-depth analysis of existing studies on\nthis task. We propose a six-layer vision framework that categorizes code\ngeneration process into distinct phases, namely Input Phase, Orchestration\nPhase, Development Phase, and Validation Phase. Additionally, we outline our\nvision workflow, which reflects on the currently prevalent frameworks. We\nsystematically analyse the challenges faced by large language models, including\nthose LLM-based agent frameworks, in code generation tasks. With these, we\noffer various perspectives and actionable recommendations in this area. Our aim\nis to provide guidelines for improving the reliability, robustness and\nusability of LLM-based code generation systems. Ultimately, this work seeks to\naddress persistent challenges and to provide practical suggestions for a more\npragmatic LLM-based solution for future code generation endeavors.", + "arxiv_id": "2501.11354v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet all mandatory criteria: it lacks experimental results with quantitative metrics (Criterion 2) and direct comparison with state-of-the-art techniques (Criterion 3). While it addresses real-world challenges (Criterion 5), the mandatory requirements are not fully satisfied." + }, + { + "paper": { + "title": "Federated Learning with Sample-level Client Drift Mitigation", + "abstract": "Federated Learning (FL) suffers from severe performance degradation due to\nthe data heterogeneity among clients. Existing works reveal that the\nfundamental reason is that data heterogeneity can cause client drift where the\nlocal model update deviates from the global one, and thus they usually tackle\nthis problem from the perspective of calibrating the obtained local update.\nDespite effectiveness, existing methods substantially lack a deep understanding\nof how heterogeneous data samples contribute to the formation of client drift.\nIn this paper, we bridge this gap by identifying that the drift can be viewed\nas a cumulative manifestation of biases present in all local samples and the\nbias between samples is different. Besides, the bias dynamically changes as the\nFL training progresses. Motivated by this, we propose FedBSS that first\nmitigates the heterogeneity issue in a sample-level manner, orthogonal to\nexisting methods. Specifically, the core idea of our method is to adopt a\nbias-aware sample selection scheme that dynamically selects the samples from\nsmall biases to large epoch by epoch to train progressively the local model in\neach round. In order to ensure the stability of training, we set the\ndiversified knowledge acquisition stage as the warm-up stage to avoid the local\noptimality caused by knowledge deviation in the early stage of the model.\nEvaluation results show that FedBSS outperforms state-of-the-art baselines. In\naddition, we also achieved effective results on feature distribution skew and\nnoise label dataset setting, which proves that FedBSS can not only reduce\nheterogeneity, but also has scalability and robustness.", + "arxiv_id": "2501.11360v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on Federated Learning improvements for data heterogeneity but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (SOTA comparison), it fails mandatory criterion 1 (LLM practical applications)." + }, + { + "paper": { + "title": "Multi-View Spectral Clustering for Graphs with Multiple View Structures", + "abstract": "Despite the fundamental importance of clustering, to this day, much of the\nrelevant research is still based on ambiguous foundations, leading to an\nunclear understanding of whether or how the various clustering methods are\nconnected with each other. In this work, we provide an additional stepping\nstone towards resolving such ambiguities by presenting a general clustering\nframework that subsumes a series of seemingly disparate clustering methods,\nincluding various methods belonging to the wildly popular spectral clustering\nframework. In fact, the generality of the proposed framework is additionally\ncapable of shedding light to the largely unexplored area of multi-view graphs\nwhose each view may have differently clustered nodes. In turn, we propose\nGenClus: a method that is simultaneously an instance of this framework and a\ngeneralization of spectral clustering, while also being closely related to\nk-means as well. This results in a principled alternative to the few existing\nmethods studying this special type of multi-view graphs. Then, we conduct\nin-depth experiments, which demonstrate that GenClus is more computationally\nefficient than existing methods, while also attaining similar or better\nclustering performance. Lastly, a qualitative real-world case-study further\ndemonstrates the ability of GenClus to produce meaningful clusterings.", + "arxiv_id": "2501.11422v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on general graph clustering methodologies and multi-view analysis without explicitly addressing Large Language Models (LLMs) or their applications in knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails criterion 1 (LLM practical applications). No connection to LLM-related tasks or autonomous AI scenarios is evident." + }, + { + "paper": { + "title": "A Survey on Diffusion Models for Anomaly Detection", + "abstract": "Diffusion models (DMs) have emerged as a powerful class of generative AI\nmodels, showing remarkable potential in anomaly detection (AD) tasks across\nvarious domains, such as cybersecurity, fraud detection, healthcare, and\nmanufacturing. The intersection of these two fields, termed diffusion models\nfor anomaly detection (DMAD), offers promising solutions for identifying\ndeviations in increasingly complex and high-dimensional data. In this survey,\nwe systematically review recent advances in DMAD research and investigate their\ncapabilities. We begin by presenting the fundamental concepts of AD and DMs,\nfollowed by a comprehensive analysis of classic DM architectures including\nDDPMs, DDIMs, and Score SDEs. We further categorize existing DMAD methods into\nreconstruction-based, density-based, and hybrid approaches, providing detailed\nexaminations of their methodological innovations. We also explore the diverse\ntasks across different data modalities, encompassing image, time series, video,\nand multimodal data analysis. Furthermore, we discuss critical challenges and\nemerging research directions, including computational efficiency, model\ninterpretability, robustness enhancement, edge-cloud collaboration, and\nintegration with large language models. The collection of DMAD research papers\nand resources is available at https://github.com/fdjingliu/DMAD.", + "arxiv_id": "2501.11430v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on diffusion models for anomaly detection rather than LLMs in practical applications like knowledge graphs or agentic AI. It does not meet mandatory criteria 1 (Practical Applications of LLMs) and lacks experimental results or SOTA comparisons specific to LLMs. While it mentions integration with LLMs as a future direction, the primary focus is outside the required scope." + }, + { + "paper": { + "title": "Communication-Efficient Federated Learning Based on Explanation-Guided\n Pruning for Remote Sensing Image Classification", + "abstract": "Federated learning (FL) is a decentralized machine learning paradigm, where\nmultiple clients collaboratively train a global model by exchanging only model\nupdates with the central server without sharing the local data of clients. Due\nto the large volume of model updates required to be transmitted between clients\nand the central server, most FL systems are associated with high transfer costs\n(i.e., communication overhead). This issue is more critical for operational\napplications in remote sensing (RS), especially when large-scale RS data is\nprocessed and analyzed through FL systems with restricted communication\nbandwidth. To address this issue, we introduce an explanation-guided pruning\nstrategy for communication-efficient FL in the context of RS image\nclassification. Our pruning strategy is defined based on the layerwise\nrelevance propagation (LRP) driven explanations to: 1) efficiently and\neffectively identify the most relevant and informative model parameters (to be\nexchanged between clients and the central server); and 2) eliminate the\nnon-informative ones to minimize the volume of model updates. The experimental\nresults on the BigEarthNet-S2 dataset demonstrate that our strategy effectively\nreduces the number of shared model updates, while increasing the generalization\nability of the global model. The code of this work will be publicly available\nat https://git.tu-berlin.de/rsim/FL-LRP", + "arxiv_id": "2501.11493v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on communication-efficient federated learning for remote sensing image classification but does not address Large Language Models (LLMs) or their practical applications (mandatory Criterion 1 unmet). While it meets Criteria 2 (experimental results) and 3 (implicit SOTA comparison via performance claims), failure to satisfy ALL mandatory criteria necessitates rejection." + }, + { + "paper": { + "title": "Meta-Instance Selection. Instance Selection as a Classification Problem\n with Meta-Features", + "abstract": "Data pruning, or instance selection, is an important problem in machine\nlearning especially in terms of nearest neighbour classifier. However, in data\npruning which speeds up the prediction phase, there is an issue related to the\nspeed and efficiency of the process itself. In response, the study proposes an\napproach involving transforming the instance selection process into a\nclassification task conducted in a unified meta-feature space where each\ninstance can be classified and assigned to either the \"to keep\" or \"to remove\"\nclass. This approach requires training an appropriate meta-classifier, which\ncan be developed based on historical instance selection results from other\ndatasets using reference instance selection methods as a labeling tool. This\nwork proposes constructing the meta-feature space based on properties extracted\nfrom the nearest neighbor graph. Experiments conducted on 17 datasets of\nvarying sizes and five reference instance selection methods (ENN, Drop3, ICF,\nHMN-EI, and CCIS) demonstrate that the proposed solution achieves results\ncomparable to reference instance selection methods while significantly reducing\ncomputational complexity. In the proposed approach, the computational\ncomplexity of the system depends only on identifying the k-nearest neighbors\nfor each data sample and running the meta-classifier. Additionally, the study\ndiscusses the choice of meta-classifier, recommending the use of Balanced\nRandom Forest.", + "arxiv_id": "2501.11526v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on instance selection methods for nearest neighbor classifiers but does not address Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it meets Criteria 2 (experimental results) and 3 (comparison with SOTA), it fails Criterion 1 (LLM practical applications). Mandatory requirements are not fully satisfied." + }, + { + "paper": { + "title": "The impact of intrinsic rewards on exploration in Reinforcement Learning", + "abstract": "One of the open challenges in Reinforcement Learning is the hard exploration\nproblem in sparse reward environments. Various types of intrinsic rewards have\nbeen proposed to address this challenge by pushing towards diversity. This\ndiversity might be imposed at different levels, favouring the agent to explore\ndifferent states, policies or behaviours (State, Policy and Skill level\ndiversity, respectively). However, the impact of diversity on the agent's\nbehaviour remains unclear. In this work, we aim to fill this gap by studying\nthe effect of different levels of diversity imposed by intrinsic rewards on the\nexploration patterns of RL agents. We select four intrinsic rewards (State\nCount, Intrinsic Curiosity Module (ICM), Maximum Entropy, and Diversity is all\nyou need (DIAYN)), each pushing for a different diversity level. We conduct an\nempirical study on MiniGrid environment to compare their impact on exploration\nconsidering various metrics related to the agent's exploration, namely:\nepisodic return, observation coverage, agent's position coverage, policy\nentropy, and timeframes to reach the sparse reward. The main outcome of the\nstudy is that State Count leads to the best exploration performance in the case\nof low-dimensional observations. However, in the case of RGB observations, the\nperformance of State Count is highly degraded mostly due to representation\nlearning challenges. Conversely, Maximum Entropy is less impacted, resulting in\na more robust exploration, despite being not always optimal. Lastly, our\nempirical study revealed that learning diverse skills with DIAYN, often linked\nto improved robustness and generalisation, does not promote exploration in\nMiniGrid environments. This is because: i) learning the skill space itself can\nbe challenging, and ii) exploration within the skill space prioritises\ndifferentiating between behaviours rather than achieving uniform state\nvisitation.", + "arxiv_id": "2501.11533v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on reinforcement learning and intrinsic rewards but does not address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI applications. It fails to meet the mandatory Criterion 1 (Practical Applications of LLMs). While it meets Criteria 2 and 3 with experimental results and SOTA comparisons, ALL mandatory criteria must be satisfied for acceptance." + }, + { + "paper": { + "title": "StAyaL | Multilingual Style Transfer", + "abstract": "Stylistic text generation plays a vital role in enhancing communication by\nreflecting the nuances of individual expression. This paper presents a novel\napproach for generating text in a specific speaker's style across different\nlanguages. We show that by leveraging only 100 lines of text, an individuals\nunique style can be captured as a high-dimensional embedding, which can be used\nfor both text generation and stylistic translation. This methodology breaks\ndown the language barrier by transferring the style of a speaker between\nlanguages. The paper is structured into three main phases: augmenting the\nspeaker's data with stylistically consistent external sources, separating style\nfrom content using machine learning and deep learning techniques, and\ngenerating an abstract style profile by mean pooling the learned embeddings.\nThe proposed approach is shown to be topic-agnostic, with test accuracy and F1\nscores of 74.9\\% and 0.75, respectively. The results demonstrate the potential\nof the style profile for multilingual communication, paving the way for further\napplications in personalized content generation and cross-linguistic stylistic\ntransfer.", + "arxiv_id": "2501.11639v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nFails to meet all mandatory criteria. While it demonstrates practical applications (criterion 1) and experimental results (criterion 2), there is no explicit comparison with state-of-the-art methods (criterion 3). The abstract does not mention LLMs specifically, and the methodology's connection to LLMs remains ambiguous despite potential deep learning use." + }, + { + "paper": { + "title": "Spatially-Delineated Domain-Adapted AI Classification: An Application\n for Oncology Data", + "abstract": "Given multi-type point maps from different place-types (e.g., tumor regions),\nour objective is to develop a classifier trained on the source place-type to\naccurately distinguish between two classes of the target place-type based on\ntheir point arrangements. This problem is societally important for many\napplications, such as generating clinical hypotheses for designing new\nimmunotherapies for cancer treatment. The challenge lies in the spatial\nvariability, the inherent heterogeneity and variation observed in spatial\nproperties or arrangements across different locations (i.e., place-types).\nPrevious techniques focus on self-supervised tasks to learn domain-invariant\nfeatures and mitigate domain differences; however, they often neglect the\nunderlying spatial arrangements among data points, leading to significant\ndiscrepancies across different place-types. We explore a novel multi-task\nself-learning framework that targets spatial arrangements, such as spatial\nmix-up masking and spatial contrastive predictive coding, for\nspatially-delineated domain-adapted AI classification. Experimental results on\nreal-world datasets (e.g., oncology data) show that the proposed framework\nprovides higher prediction accuracy than baseline methods.", + "arxiv_id": "2501.11695v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications (oncology data classification for cancer treatment), which falls under the mandatory rejection criteria. While it meets some technical criteria (experimental results and methodology), medical applications are explicitly excluded regardless of other merits." + }, + { + "paper": { + "title": "Human services organizations and the responsible integration of AI:\n Considering ethics and contextualizing risk(s)", + "abstract": "This paper examines the responsible integration of artificial intelligence\n(AI) in human services organizations (HSOs), proposing a nuanced framework for\nevaluating AI applications across multiple dimensions of risk. The authors\nargue that ethical concerns about AI deployment -- including professional\njudgment displacement, environmental impact, model bias, and data laborer\nexploitation -- vary significantly based on implementation context and specific\nuse cases. They challenge the binary view of AI adoption, demonstrating how\ndifferent applications present varying levels of risk that can often be\neffectively managed through careful implementation strategies. The paper\nhighlights promising solutions, such as local large language models, that can\nfacilitate responsible AI integration while addressing common ethical concerns.\nThe authors propose a dimensional risk assessment approach that considers\nfactors like data sensitivity, professional oversight requirements, and\npotential impact on client wellbeing. They conclude by outlining a path forward\nthat emphasizes empirical evaluation, starting with lower-risk applications and\nbuilding evidence-based understanding through careful experimentation. This\napproach enables organizations to maintain high ethical standards while\nthoughtfully exploring how AI might enhance their capacity to serve clients and\ncommunities effectively.", + "arxiv_id": "2501.11705v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on responsible AI integration and ethical concerns in human services organizations, which falls under the rejection criteria for focusing on responsible AI applications and social implications. It fails to meet mandatory criteria 1 (Practical Applications in specified areas), 2 (Experimental Results), and 3 (Comparison with SOTA). While addressing real-world challenges (criterion 5), the core focus violates explicit rejection parameters." + }, + { + "paper": { + "title": "GL-ICNN: An End-To-End Interpretable Convolutional Neural Network for\n the Diagnosis and Prediction of Alzheimer's Disease", + "abstract": "Deep learning methods based on Convolutional Neural Networks (CNNs) have\nshown great potential to improve early and accurate diagnosis of Alzheimer's\ndisease (AD) dementia based on imaging data. However, these methods have yet to\nbe widely adopted in clinical practice, possibly due to the limited\ninterpretability of deep learning models. The Explainable Boosting Machine\n(EBM) is a glass-box model but cannot learn features directly from input\nimaging data. In this study, we propose a novel interpretable model that\ncombines CNNs and EBMs for the diagnosis and prediction of AD. We develop an\ninnovative training strategy that alternatingly trains the CNN component as a\nfeature extractor and the EBM component as the output block to form an\nend-to-end model. The model takes imaging data as input and provides both\npredictions and interpretable feature importance measures. We validated the\nproposed model on the Alzheimer's Disease Neuroimaging Initiative (ADNI)\ndataset and the Health-RI Parelsnoer Neurodegenerative Diseases Biobank (PND)\nas an external testing set. The proposed model achieved an area-under-the-curve\n(AUC) of 0.956 for AD and control classification, and 0.694 for the prediction\nof conversion of mild cognitive impairment (MCI) to AD on the ADNI cohort. The\nproposed model is a glass-box model that achieves a comparable performance with\nother state-of-the-art black-box models. Our code is publicly available at:\nhttps://anonymous.4open.science/r/GL-ICNN.", + "arxiv_id": "2501.11715v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (Alzheimer's disease diagnosis/prediction), which is an explicit rejection criterion. While it addresses interpretability and includes experimental results, the medical focus disqualifies it regardless of other merits." + }, + { + "paper": { + "title": "Episodic memory in AI agents poses risks that should be studied and\n mitigated", + "abstract": "Most current AI models have little ability to store and later retrieve a\nrecord or representation of what they do. In human cognition, episodic memories\nplay an important role in both recall of the past as well as planning for the\nfuture. The ability to form and use episodic memories would similarly enable a\nbroad range of improved capabilities in an AI agent that interacts with and\ntakes actions in the world. Researchers have begun directing more attention to\ndeveloping memory abilities in AI models. It is therefore likely that models\nwith such capability will be become widespread in the near future. This could\nin some ways contribute to making such AI agents safer by enabling users to\nbetter monitor, understand, and control their actions. However, as a new\ncapability with wide applications, we argue that it will also introduce\nsignificant new risks that researchers should begin to study and address. We\noutline these risks and benefits and propose four principles to guide the\ndevelopment of episodic memory capabilities so that these will enhance, rather\nthan undermine, the effort to keep AI safe and trustworthy.", + "arxiv_id": "2501.11739v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on analyzing risks and safety implications of episodic memory in AI agents, falling under responsible AI/ethics applications. While it mentions agentic AI capabilities, it fails to meet mandatory criteria 2 (no experimental results/quantitative metrics) and 3 (no SOTA comparisons). It violates the rejection rule for focusing on responsible AI applications." + }, + { + "paper": { + "title": "Human-AI Collaborative Game Testing with Vision Language Models", + "abstract": "As modern video games become increasingly complex, traditional manual testing\nmethods are proving costly and inefficient, limiting the ability to ensure\nhigh-quality game experiences. While advancements in Artificial Intelligence\n(AI) offer the potential to assist human testers, the effectiveness of AI in\ntruly enhancing real-world human performance remains underexplored. This study\ninvestigates how AI can improve game testing by developing and experimenting\nwith an AI-assisted workflow that leverages state-of-the-art machine learning\nmodels for defect detection. Through an experiment involving 800 test cases and\n276 participants of varying backgrounds, we evaluate the effectiveness of AI\nassistance under four conditions: with or without AI support, and with or\nwithout detailed knowledge of defects and design documentation. The results\nindicate that AI assistance significantly improves defect identification\nperformance, particularly when paired with detailed knowledge. However,\nchallenges arise when AI errors occur, negatively impacting human\ndecision-making. Our findings show the importance of optimizing human-AI\ncollaboration and implementing strategies to mitigate the effects of AI\ninaccuracies. By this research, we demonstrate AI's potential and problems in\nenhancing efficiency and accuracy in game testing workflows and offers\npractical insights for integrating AI into the testing process.", + "arxiv_id": "2501.11782v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet all mandatory criteria: While it addresses practical applications (game testing) and includes experimental results (800 test cases, quantitative metrics), it lacks explicit comparison with state-of-the-art techniques. The focus on video game testing also risks falling under 'video processing' exclusion, though marginally. Mandatory criteria 3 is unmet, leading to rejection." + }, + { + "paper": { + "title": "PXGen: A Post-hoc Explainable Method for Generative Models", + "abstract": "With the rapid growth of generative AI in numerous applications, explainable\nAI (XAI) plays a crucial role in ensuring the responsible development and\ndeployment of generative AI technologies. XAI has undergone notable\nadvancements and widespread adoption in recent years, reflecting a concerted\npush to enhance the transparency, interpretability, and credibility of AI\nsystems. Recent research emphasizes that a proficient XAI method should adhere\nto a set of criteria, primarily focusing on two key areas. Firstly, it should\nensure the quality and fluidity of explanations, encompassing aspects like\nfaithfulness, plausibility, completeness, and tailoring to individual needs.\nSecondly, the design principle of the XAI system or mechanism should cover the\nfollowing factors such as reliability, resilience, the verifiability of its\noutputs, and the transparency of its algorithm. However, research in XAI for\ngenerative models remains relatively scarce, with little exploration into how\nsuch methods can effectively meet these criteria in that domain. In this work,\nwe propose PXGen, a post-hoc explainable method for generative models. Given a\nmodel that needs to be explained, PXGen prepares two materials for the\nexplanation, the Anchor set and intrinsic \u0026 extrinsic criteria. Those materials\nare customizable by users according to their purpose and requirements. Via the\ncalculation of each criterion, each anchor has a set of feature values and\nPXGen provides examplebased explanation methods according to the feature values\namong all the anchors and illustrated and visualized to the users via tractable\nalgorithms such as k-dispersion or k-center.", + "arxiv_id": "2501.11827v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on explainable AI (XAI) for generative models, which aligns with responsible AI development (a rejection criterion). While it addresses methodology (criterion 4) and real-world challenges (criterion 5), it fails to demonstrate practical applications in LLM domains like knowledge graphs/RAG/agentic AI (mandatory criterion 1) or provide experimental comparisons with SOTA (mandatory criterion 3). Additionally, its emphasis on XAI's role in 'responsible development' directly conflicts with the rejection rule for papers primarily focused on responsible AI/ethics." + }, + { + "paper": { + "title": "Is your LLM trapped in a Mental Set? Investigative study on how mental\n sets affect the reasoning capabilities of LLMs", + "abstract": "In this paper, we present an investigative study on how Mental Sets influence\nthe reasoning capabilities of LLMs. LLMs have excelled in diverse natural\nlanguage processing (NLP) tasks, driven by advancements in parameter-efficient\nfine-tuning (PEFT) and emergent capabilities like in-context learning (ICL).\nFor complex reasoning tasks, selecting the right model for PEFT or ICL is\ncritical, often relying on scores on benchmarks such as MMLU, MATH, and GSM8K.\nHowever, current evaluation methods, based on metrics like F1 Score or\nreasoning chain assessments by larger models, overlook a key dimension:\nadaptability to unfamiliar situations and overcoming entrenched thinking\npatterns. In cognitive psychology, Mental Set refers to the tendency to persist\nwith previously successful strategies, even when they become inefficient - a\nchallenge for problem solving and reasoning. We compare the performance of LLM\nmodels like Llama-3.1-8B-Instruct, Llama-3.1-70B-Instruct and GPT-4o in the\npresence of mental sets. To the best of our knowledge, this is the first study\nto integrate cognitive psychology concepts into the evaluation of LLMs for\ncomplex reasoning tasks, providing deeper insights into their adaptability and\nproblem-solving efficacy.", + "arxiv_id": "2501.11833v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nFails to meet mandatory criteria 1 (Practical Applications) as the focus is on cognitive psychology-based evaluation rather than real-world LLM applications like knowledge graphs, RAG, or agentic AI. Also lacks direct real-world implementation discussion (criterion 5)." + }, + { + "paper": { + "title": "A Survey on Memory-Efficient Large-Scale Model Training in AI for\n Science", + "abstract": "Scientific research faces high costs and inefficiencies with traditional\nmethods, but the rise of deep learning and large language models (LLMs) offers\ninnovative solutions. This survey reviews LLM applications across scientific\nfields such as biology, medicine, chemistry, and meteorology, underscoring\ntheir role in advancing research. However, the continuous expansion of model\nsize has led to significant memory demands, hindering further development and\napplication of LLMs for science. To address this, we review memory-efficient\ntraining techniques for LLMs based on the transformer architecture, including\ndistributed training, mixed precision training, and gradient checkpointing.\nUsing AlphaFold 2 as an example, we demonstrate how tailored memory\noptimization methods can reduce storage needs while preserving prediction\naccuracy. We also discuss the challenges of memory optimization in practice and\npotential future directions, hoping to provide valuable insights for\nresearchers and engineers.", + "arxiv_id": "2501.11847v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on AI for Science applications, including medicine, which falls under the rejection criteria. While it discusses memory-efficient training techniques, it does not meet all mandatory criteria (e.g., direct comparison with SOTA for new methodologies) and violates the medical application exclusion." + }, + { + "paper": { + "title": "Coarse-to-Fine Lightweight Meta-Embedding for ID-Based Recommendation", + "abstract": "The state-of-the-art recommendation systems have shifted the attention to\nefficient recommendation, e.g., on-device recommendation, under memory\nconstraints. To this end, the existing methods either focused on the\nlightweight embeddings for both users and items, or involved on-device systems\nenjoying the compact embeddings to enhance reusability and reduces space\ncomplexity. However, they focus solely on the coarse granularity of embedding,\nwhile overlook the fine-grained semantic nuances, to adversarially downgrade\nthe efficacy of meta-embeddings in capturing the intricate relationship over\nboth user and item, consequently resulting into the suboptimal recommendations.\nIn this paper, we aim to study how the meta-embedding can efficiently learn\nvaried grained semantics, together with how the fine-grained meta-embedding can\nstrengthen the representation of coarse-grained meta-embedding. To answer these\nquestions, we develop a novel graph neural networks (GNNs) based recommender\nwhere each user and item serves as the node, linked directly to coarse-grained\nvirtual nodes and indirectly to fine-grained virtual nodes, ensuring different\ngrained semantic learning, while disclosing: 1) In contrast to coarse-grained\nsemantics, fine-grained semantics are well captured through sparse\nmeta-embeddings, which adaptively 2) balance the embedding uniqueness and\nmemory constraint. Additionally, the initialization method come up upon\nSparsePCA, along with a soft thresholding activation function to render the\nsparseness of the meta-embeddings. We propose a weight bridging update strategy\nthat focuses on matching each coarse-grained meta-embedding with several\nfine-grained meta-embeddings based on the users/items' semantics. Extensive\nexperiments substantiate our method's superiority over existing baselines. Our\ncode is available at https://github.com/htyjers/C2F-MetaEmbed.", + "arxiv_id": "2501.11870v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on recommendation systems using meta-embeddings and GNNs but does not explicitly address practical applications of LLMs (Criterion 1 unmet). While it meets Criteria 2 and 3 (experimental results and SOTA comparisons), it fails to satisfy all mandatory criteria. Additionally, the application domain (recommendation systems) is not among the specified LLM-related areas (knowledge graphs, RAG, agentic AI) in the mandatory criteria." + }, + { + "paper": { + "title": "Community-Aware Temporal Walks: Parameter-Free Representation Learning\n on Continuous-Time Dynamic Graphs", + "abstract": "Dynamic graph representation learning plays a crucial role in understanding\nevolving behaviors. However, existing methods often struggle with flexibility,\nadaptability, and the preservation of temporal and structural dynamics. To\naddress these issues, we propose Community-aware Temporal Walks (CTWalks), a\nnovel framework for representation learning on continuous-time dynamic graphs.\nCTWalks integrates three key components: a community-based parameter-free\ntemporal walk sampling mechanism, an anonymization strategy enriched with\ncommunity labels, and an encoding process that leverages continuous temporal\ndynamics modeled via ordinary differential equations (ODEs). This design\nenables precise modeling of both intra- and inter-community interactions,\noffering a fine-grained representation of evolving temporal patterns in\ncontinuous-time dynamic graphs. CTWalks theoretically overcomes locality bias\nin walks and establishes its connection to matrix factorization. Experiments on\nbenchmark datasets demonstrate that CTWalks outperforms established methods in\ntemporal link prediction tasks, achieving higher accuracy while maintaining\nrobustness.", + "arxiv_id": "2501.11880v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on dynamic graph representation learning (CTWalks) for temporal link prediction but does not explicitly mention applications involving Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails to satisfy criterion 1 (LLM practical applications). No explicit connection to LLM-related domains or methodologies is evident in the abstract." + }, + { + "paper": { + "title": "A Lightweight and Interpretable Deepfakes Detection Framework", + "abstract": "The recent realistic creation and dissemination of so-called deepfakes poses\na serious threat to social life, civil rest, and law. Celebrity defaming,\nelection manipulation, and deepfakes as evidence in court of law are few\npotential consequences of deepfakes. The availability of open source trained\nmodels based on modern frameworks such as PyTorch or TensorFlow, video\nmanipulations Apps such as FaceApp and REFACE, and economical computing\ninfrastructure has easen the creation of deepfakes. Most of the existing\ndetectors focus on detecting either face-swap, lip-sync, or puppet master\ndeepfakes, but a unified framework to detect all three types of deepfakes is\nhardly explored. This paper presents a unified framework that exploits the\npower of proposed feature fusion of hybrid facial landmarks and our novel heart\nrate features for detection of all types of deepfakes. We propose novel heart\nrate features and fused them with the facial landmark features to better\nextract the facial artifacts of fake videos and natural variations available in\nthe original videos. We used these features to train a light-weight XGBoost to\nclassify between the deepfake and bonafide videos. We evaluated the performance\nof our framework on the world leaders dataset (WLDR) that contains all types of\ndeepfakes. Experimental results illustrate that the proposed framework offers\nsuperior detection performance over the comparative deepfakes detection\nmethods. Performance comparison of our framework against the LSTM-FCN, a\ncandidate of deep learning model, shows that proposed model achieves similar\nresults, however, it is more interpretable.", + "arxiv_id": "2501.11927v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on video processing (deepfakes detection) and social harm implications of AI, both listed as rejection criteria. While it meets criteria 2 (experimental results) and 3 (SOTA comparison), it fails the mandatory LLM application requirement and violates two rejection conditions." + }, + { + "paper": { + "title": "MeshONet: A Generalizable and Efficient Operator Learning Method for\n Structured Mesh Generation", + "abstract": "Mesh generation plays a crucial role in scientific computing. Traditional\nmesh generation methods, such as TFI and PDE-based methods, often struggle to\nachieve a balance between efficiency and mesh quality. To address this\nchallenge, physics-informed intelligent learning methods have recently emerged,\nsignificantly improving generation efficiency while maintaining high mesh\nquality. However, physics-informed methods fail to generalize when applied to\npreviously unseen geometries, as even small changes in the boundary shape\nnecessitate burdensome retraining to adapt to new geometric variations. In this\npaper, we introduce MeshONet, the first generalizable intelligent learning\nmethod for structured mesh generation. The method transforms the mesh\ngeneration task into an operator learning problem with multiple input and\nsolution functions. To effectively overcome the multivariable mapping\nrestriction of operator learning methods, we propose a dual-branch,\nshared-trunk architecture to approximate the mapping between function spaces\nbased on input-output pairs. Experimental results show that MeshONet achieves a\nspeedup of up to four orders of magnitude in generation efficiency over\ntraditional methods. It also enables generalization to different geometries\nwithout retraining, greatly enhancing the practicality of intelligent methods.", + "arxiv_id": "2501.11937v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on operator learning for mesh generation in scientific computing but does not explicitly address Large Language Models (LLMs) in its methodology or applications. While it meets Criteria 2 (quantitative metrics) and 3 (comparison with SOTA), it fails Criterion 1 (LLM-based practical applications). Mandatory criteria require ALL to be met, leading to rejection despite other strengths." + }, + { + "paper": { + "title": "Survey on Hand Gesture Recognition from Visual Input", + "abstract": "Hand gesture recognition has become an important research area, driven by the\ngrowing demand for human-computer interaction in fields such as sign language\nrecognition, virtual and augmented reality, and robotics. Despite the rapid\ngrowth of the field, there are few surveys that comprehensively cover recent\nresearch developments, available solutions, and benchmark datasets. This survey\naddresses this gap by examining the latest advancements in hand gesture and 3D\nhand pose recognition from various types of camera input data including RGB\nimages, depth images, and videos from monocular or multiview cameras, examining\nthe differing methodological requirements of each approach. Furthermore, an\noverview of widely used datasets is provided, detailing their main\ncharacteristics and application domains. Finally, open challenges such as\nachieving robust recognition in real-world environments, handling occlusions,\nensuring generalization across diverse users, and addressing computational\nefficiency for real-time applications are highlighted to guide future research\ndirections. By synthesizing the objectives, methodologies, and applications of\nrecent studies, this survey offers valuable insights into current trends,\nchallenges, and opportunities for future research in human hand gesture\nrecognition.", + "arxiv_id": "2501.11992v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on hand gesture recognition from visual inputs (including video processing), which is explicitly listed as a rejection criterion. While it addresses practical applications in human-computer interaction, it does not involve Large Language Models (LLMs) or meet the required criteria related to knowledge graphs, RAG, agentic AI, or experimental validation of LLM-based approaches." + }, + { + "paper": { + "title": "Full Proportional Justified Representation", + "abstract": "In multiwinner approval voting, forming a committee that proportionally\nrepresents voters' approval ballots is an essential task. The notion of\njustified representation (JR) demands that any large \"cohesive\" group of voters\nshould be proportionally \"represented\". The \"cohesiveness\" is defined in\ndifferent ways; two common ways are the following: (C1) demands that the group\nunanimously approves a set of candidates proportional to its size, while (C2)\nrequires each member to approve at least a fixed fraction of such a set.\nSimilarly, \"representation\" have been considered in different ways: (R1) the\ncoalition's collective utility from the winning set exceeds that of any\nproportionally sized alternative, and (R2) for any proportionally sized\nalternative, at least one member of the coalition derives less utility from it\nthan from the winning set.\n Three of the four possible combinations have been extensively studied:\n(C1)-(R1) defines Proportional Justified Representation (PJR), (C1)-(R2)\ndefines Extended Justified Representation (EJR), (C2)-(R2) defines Full\nJustified Representation (FJR). All three have merits, but also drawbacks. PJR\nis the weakest notion, and perhaps not sufficiently demanding; EJR may not be\ncompatible with perfect representation; and it is open whether a committee\nsatisfying FJR can be found efficiently.\n We study the combination (C2)-(R1), which we call Full Proportional Justified\nRepresentation (FPJR). We investigate FPJR's properties and find that it shares\nPJR's advantages over EJR: several proportionality axioms (e.g. priceability,\nperfect representation) imply FPJR and PJR but not EJR. We also find that\nefficient rules like the greedy Monroe rule and the method of equal shares\nsatisfy FPJR, matching a key advantage of EJR over FJR. However, the\nProportional Approval Voting (PAV) rule may violate FPJR, so neither of EJR and\nFPJR implies the other.", + "arxiv_id": "2501.12015v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on voting systems and proportional representation in social choice theory, which qualifies as a social application under the rejection criteria. It does not address practical applications of LLMs (Criterion 1 unmet) or present experimental results with quantitative metrics (Criterion 2 unmet). While it compares methodologies (Criterion 3), the mandatory requirements are not fully satisfied, and the social application focus triggers rejection." + }, + { + "paper": { + "title": "Adaptive Class Learning to Screen Diabetic Disorders in Fundus Images of\n Eye", + "abstract": "The prevalence of ocular illnesses is growing globally, presenting a\nsubstantial public health challenge. Early detection and timely intervention\nare crucial for averting visual impairment and enhancing patient prognosis.\nThis research introduces a new framework called Class Extension with Limited\nData (CELD) to train a classifier to categorize retinal fundus images. The\nclassifier is initially trained to identify relevant features concerning\nHealthy and Diabetic Retinopathy (DR) classes and later fine-tuned to adapt to\nthe task of classifying the input images into three classes: Healthy, DR, and\nGlaucoma. This strategy allows the model to gradually enhance its\nclassification capabilities, which is beneficial in situations where there are\nonly a limited number of labeled datasets available. Perturbation methods are\nalso used to identify the input image characteristics responsible for\ninfluencing the models decision-making process. We achieve an overall accuracy\nof 91% on publicly available datasets.", + "arxiv_id": "2501.12048v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications (diabetic/glaucoma screening in eye images) and does not address Large Language Models (LLMs) or their practical applications in knowledge graphs, RAG, or agentic AI. It fails criteria 1 and 3, and medical applications are explicitly listed as a rejection reason." + }, + { + "paper": { + "title": "Teacher Encoder-Student Decoder Denoising Guided Segmentation Network\n for Anomaly Detection", + "abstract": "Visual anomaly detection is a highly challenging task, often categorized as a\none-class classification and segmentation problem. Recent studies have\ndemonstrated that the student-teacher (S-T) framework effectively addresses\nthis challenge. However, most S-T frameworks rely solely on pre-trained teacher\nnetworks to guide student networks in learning multi-scale similar features,\noverlooking the potential of the student networks to enhance learning through\nmulti-scale feature fusion. In this study, we propose a novel model named\nPFADSeg, which integrates a pre-trained teacher network, a denoising student\nnetwork with multi-scale feature fusion, and a guided anomaly segmentation\nnetwork into a unified framework. By adopting a unique teacher-encoder and\nstudent-decoder denoising mode, the model improves the student network's\nability to learn from teacher network features. Furthermore, an adaptive\nfeature fusion mechanism is introduced to train a self-supervised segmentation\nnetwork that synthesizes anomaly masks autonomously, significantly increasing\ndetection performance. Evaluated on the MVTec AD dataset, PFADSeg achieves\nstate-of-the-art results with an image-level AUC of 98.9%, a pixel-level mean\nprecision of 76.4%, and an instance-level mean precision of 78.7%.", + "arxiv_id": "2501.12104v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on visual anomaly detection using a student-teacher framework in computer vision, but does not address Large Language Models (LLMs) or their practical applications in domains like knowledge graphs, RAG, or agentic AI. This violates mandatory criterion 1 for paper selection." + }, + { + "paper": { + "title": "Can open source large language models be used for tumor documentation in\n Germany? -- An evaluation on urological doctors' notes", + "abstract": "Tumor documentation in Germany is largely done manually, requiring reading\npatient records and entering data into structured databases. Large language\nmodels (LLMs) could potentially enhance this process by improving efficiency\nand reliability. This evaluation tests eleven different open source LLMs with\nsizes ranging from 1-70 billion model parameters on three basic tasks of the\ntumor documentation process: identifying tumor diagnoses, assigning ICD-10\ncodes, and extracting the date of first diagnosis. For evaluating the LLMs on\nthese tasks, a dataset of annotated text snippets based on anonymized doctors'\nnotes from urology was prepared. Different prompting strategies were used to\ninvestigate the effect of the number of examples in few-shot prompting and to\nexplore the capabilities of the LLMs in general. The models Llama 3.1 8B,\nMistral 7B, and Mistral NeMo 12 B performed comparably well in the tasks.\nModels with less extensive training data or having fewer than 7 billion\nparameters showed notably lower performance, while larger models did not\ndisplay performance gains. Examples from a different medical domain than\nurology could also improve the outcome in few-shot prompting, which\ndemonstrates the ability of LLMs to handle tasks needed for tumor\ndocumentation. Open source LLMs show a strong potential for automating tumor\ndocumentation. Models from 7-12 billion parameters could offer an optimal\nbalance between performance and resource efficiency. With tailored fine-tuning\nand well-designed prompting, these models might become important tools for\nclinical documentation in the future. The code for the evaluation is available\nfrom https://github.com/stefan-m-lenz/UroLlmEval. We also release the dataset\nas a new valuable resource that addresses the shortage of authentic and easily\naccessible benchmarks in German-language medical NLP.", + "arxiv_id": "2501.12106v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (tumor documentation in urology), which is an explicit rejection criterion. While it meets criteria 2 (experimental results) and 3 (comparisons), medical applications are prohibited regardless of other merits." + }, + { + "paper": { + "title": "FedCLEAN: byzantine defense by CLustering Errors of Activation maps in\n Non-IID federated learning environments", + "abstract": "Federated Learning (FL) enables clients to collaboratively train a global\nmodel using their local datasets while reinforcing data privacy. However, FL is\nsusceptible to poisoning attacks. Existing defense mechanisms assume that\nclients' data are independent and identically distributed (IID), making them\nineffective in real-world applications where data are non-IID. This paper\npresents FedCLEAN, the first defense capable of filtering attackers' model\nupdates in a non-IID FL environment. The originality of FedCLEAN is twofold.\nFirst, it relies on a client confidence score derived from the reconstruction\nerrors of each client's model activation maps for a given trigger set, with\nreconstruction errors obtained by means of a Conditional Variational\nAutoencoder trained according to a novel server-side strategy. Second, we\npropose an ad-hoc trust propagation algorithm based on client scores, which\nallows building a cluster of benign clients while flagging potential attackers.\nExperimental results on the datasets MNIST and FashionMNIST demonstrate the\nrobustness of FedCLEAN against Byzantine attackers in non-IID scenarios and a\nclose-to-zero benign client misclassification rate, even in the absence of an\nattack.", + "arxiv_id": "2501.12123v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet Criterion 1 (Practical Applications of LLMs), as it focuses on federated learning defense mechanisms rather than LLM applications like knowledge graphs, RAG, or agentic AI. While it meets Criteria 2 (experimental results) and 3 (comparison with SOTA), failure to satisfy ALL mandatory criteria (1-3) necessitates rejection." + }, + { + "paper": { + "title": "FuocChuVIP123 at CoMeDi Shared Task: Disagreement Ranking with\n XLM-Roberta Sentence Embeddings and Deep Neural Regression", + "abstract": "This paper presents results of our system for CoMeDi Shared Task, focusing on\nSubtask 2: Disagreement Ranking. Our system leverages sentence embeddings\ngenerated by the paraphrase-xlm-r-multilingual-v1 model, combined with a deep\nneural regression model incorporating batch normalization and dropout for\nimproved generalization. By predicting the mean of pairwise judgment\ndifferences between annotators, our method explicitly targets disagreement\nranking, diverging from traditional \"gold label\" aggregation approaches. We\noptimized our system with a customized architecture and training procedure,\nachieving competitive performance in Spearman correlation against mean\ndisagreement labels. Our results highlight the importance of robust embeddings,\neffective model architecture, and careful handling of judgment differences for\nranking disagreement in multilingual contexts. These findings provide insights\ninto the use of contextualized representations for ordinal judgment tasks and\nopen avenues for further refinement of disagreement prediction models.", + "arxiv_id": "2501.12336v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on multilingual disagreement ranking using XLM-R embeddings but does not address practical LLM applications like knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (quantitative metrics) and 4 (methodology details), it fails criterion 1 (practical LLM applications) and 3 (explicit SOTA comparisons). The task aligns more with NLP annotation analysis than core LLM application domains." + }, + { + "paper": { + "title": "Video Depth Anything: Consistent Depth Estimation for Super-Long Videos", + "abstract": "Depth Anything has achieved remarkable success in monocular depth estimation\nwith strong generalization ability. However, it suffers from temporal\ninconsistency in videos, hindering its practical applications. Various methods\nhave been proposed to alleviate this issue by leveraging video generation\nmodels or introducing priors from optical flow and camera poses. Nonetheless,\nthese methods are only applicable to short videos (\u003c 10 seconds) and require a\ntrade-off between quality and computational efficiency. We propose Video Depth\nAnything for high-quality, consistent depth estimation in super-long videos\n(over several minutes) without sacrificing efficiency. We base our model on\nDepth Anything V2 and replace its head with an efficient spatial-temporal head.\nWe design a straightforward yet effective temporal consistency loss by\nconstraining the temporal depth gradient, eliminating the need for additional\ngeometric priors. The model is trained on a joint dataset of video depth and\nunlabeled images, similar to Depth Anything V2. Moreover, a novel\nkey-frame-based strategy is developed for long video inference. Experiments\nshow that our model can be applied to arbitrarily long videos without\ncompromising quality, consistency, or generalization ability. Comprehensive\nevaluations on multiple video benchmarks demonstrate that our approach sets a\nnew state-of-the-art in zero-shot video depth estimation. We offer models of\ndifferent scales to support a range of scenarios, with our smallest model\ncapable of real-time performance at 30 FPS.", + "arxiv_id": "2501.12375v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing (depth estimation in super-long videos) and does not mention LLM applications, which violates the mandatory rejection criteria. While it meets experimental results and SOTA comparison requirements, it fails core practical application criteria for LLM-focused research." + }, + { + "paper": { + "title": "Leveraging GANs For Active Appearance Models Optimized Model Fitting", + "abstract": "Generative Adversarial Networks (GANs) have gained prominence in refining\nmodel fitting tasks in computer vision, particularly in domains involving\ndeformable models like Active Appearance Models (AAMs). This paper explores the\nintegration of GANs to enhance the AAM fitting process, addressing challenges\nin optimizing nonlinear parameters associated with appearance and shape\nvariations. By leveraging GANs' adversarial training framework, the aim is to\nminimize fitting errors and improve convergence rates. Achieving robust\nperformance even in cases with high appearance variability and occlusions. Our\napproach demonstrates significant improvements in accuracy and computational\nefficiency compared to traditional optimization techniques, thus establishing\nGANs as a potent tool for advanced image model fitting.", + "arxiv_id": "2501.11218v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on GANs and computer vision applications (AAM model fitting) rather than Large Language Models (LLMs), failing to meet the mandatory Criterion 1. Additionally, it does not address LLM-related areas like knowledge graphs, RAG, or agentic AI required for inclusion." + }, + { + "paper": { + "title": "WSSM: Geographic-enhanced hierarchical state-space model for global\n station weather forecast", + "abstract": "Global Station Weather Forecasting (GSWF), a prominent meteorological\nresearch area, is pivotal in providing timely localized weather predictions.\nDespite the progress existing models have made in the overall accuracy of the\nGSWF, executing high-precision extreme event prediction still presents a\nsubstantial challenge. The recent emergence of state-space models, with their\nability to efficiently capture continuous-time dynamics and latent states,\noffer potential solutions. However, early investigations indicated that Mamba\nunderperforms in the context of GSWF, suggesting further adaptation and\noptimization. To tackle this problem, in this paper, we introduce Weather\nState-space Model (WSSM), a novel Mamba-based approach tailored for GSWF.\nGeographical knowledge is integrated in addition to the widely-used positional\nencoding to represent the absolute special-temporal position. The multi-scale\ntime-frequency features are synthesized from coarse to fine to model the\nseasonal to extreme weather dynamic. Our method effectively improves the\noverall prediction accuracy and addresses the challenge of forecasting extreme\nweather events. The state-of-the-art results obtained on the Weather-5K subset\nunderscore the efficacy of the WSSM", + "arxiv_id": "2501.11238v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on weather forecasting using a state-space model (Mamba) but does not explicitly mention Large Language Models (LLMs) or their practical applications (Criterion 1 unmet). While it meets Criteria 2 (experimental results) and 3 (SOTA comparison), failing Criterion 1 (mandatory) necessitates rejection. No explicit LLM integration or discussion aligns with the required focus." + }, + { + "paper": { + "title": "Spatiotemporal Air Quality Mapping in Urban Areas Using Sparse Sensor\n Data, Satellite Imagery, Meteorological Factors, and Spatial Features", + "abstract": "Monitoring air pollution is crucial for protecting human health from exposure\nto harmful substances. Traditional methods of air quality monitoring, such as\nground-based sensors and satellite-based remote sensing, face limitations due\nto high deployment costs, sparse sensor coverage, and environmental\ninterferences. To address these challenges, this paper proposes a framework for\nhigh-resolution spatiotemporal Air Quality Index (AQI) mapping using sparse\nsensor data, satellite imagery, and various spatiotemporal factors. By\nleveraging Graph Neural Networks (GNNs), we estimate AQI values at unmonitored\nlocations based on both spatial and temporal dependencies. The framework\nincorporates a wide range of environmental features, including meteorological\ndata, road networks, points of interest (PoIs), population density, and urban\ngreen spaces, which enhance prediction accuracy. We illustrate the use of our\napproach through a case study in Lahore, Pakistan, where multi-resolution data\nis used to generate the air quality index map at a fine spatiotemporal scale.", + "arxiv_id": "2501.11270v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on air quality mapping using GNNs and environmental data without mentioning LLMs, knowledge graphs, RAG, or agentic AI. It fails to meet mandatory criteria 1 (Practical Applications of LLMs) and 3 (Comparison with SOTA), and provides no explicit evidence for criterion 2 (quantitative LLM-related metrics). While addressing real-world applications, it does not align with the required LLM/agentic AI focus." + }, + { + "paper": { + "title": "A Machine Learning Framework for Handling Unreliable Absence Label and\n Class Imbalance for Marine Stinger Beaching Prediction", + "abstract": "Bluebottles (\\textit{Physalia} spp.) are marine stingers resembling\njellyfish, whose presence on Australian beaches poses a significant public risk\ndue to their venomous nature. Understanding the environmental factors driving\nbluebottles ashore is crucial for mitigating their impact, and machine learning\ntools are to date relatively unexplored. We use bluebottle marine stinger\npresence/absence data from beaches in Eastern Sydney, Australia, and compare\nmachine learning models (Multilayer Perceptron, Random Forest, and XGBoost) to\nidentify factors influencing their presence. We address challenges such as\nclass imbalance, class overlap, and unreliable absence data by employing data\naugmentation techniques, including the Synthetic Minority Oversampling\nTechnique (SMOTE), Random Undersampling, and Synthetic Negative Approach that\nexcludes the negative class. Our results show that SMOTE failed to resolve\nclass overlap, but the presence-focused approach effectively handled imbalance,\nclass overlap, and ambiguous absence data. The data attributes such as the wind\ndirection, which is a circular variable, emerged as a key factor influencing\nbluebottle presence, confirming previous inference studies. However, in the\nabsence of population dynamics, biological behaviours, and life cycles, the\nbest predictive model appears to be Random Forests combined with Synthetic\nNegative Approach. This research contributes to mitigating the risks posed by\nbluebottles to beachgoers and provides insights into handling class overlap and\nunreliable negative class in environmental modelling.", + "arxiv_id": "2501.11293v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on traditional machine learning models (Random Forest, XGBoost) for environmental prediction without mentioning LLMs, knowledge graphs, RAG, or agentic AI. It fails Criteria 1 (Practical Applications of LLMs) and does not meet all mandatory requirements." + }, + { + "paper": { + "title": "On the Dimension of Pullback Attractors in Recurrent Neural Networks", + "abstract": "Recurrent Neural Networks (RNNs) are high-dimensional state space models\ncapable of learning functions on sequence data. Recently, it has been\nconjectured that reservoir computers, a particular class of RNNs, trained on\nobservations of a dynamical systems can be interpreted as embeddings. This\nresult has been established for the case of linear reservoir systems. In this\nwork, we use a nonautonomous dynamical systems approach to establish an upper\nbound for the fractal dimension of the subset of reservoir state space\napproximated during training and prediction phase. We prove that when the input\nsequences comes from an Nin-dimensional invertible dynamical system, the\nfractal dimension of this set is bounded above by Nin. The result obtained here\nare useful in dimensionality reduction of computation in RNNs as well as\nestimating fractal dimensions of dynamical systems from limited observations of\ntheir time series. It is also a step towards understanding embedding properties\nof reservoir computers.", + "arxiv_id": "2501.11357v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on theoretical analysis of fractal dimensions in reservoir computing rather than practical LLM applications. While it mentions dimensionality reduction implications, it lacks experimental results with quantitative metrics (Criterion 2) and comparisons with state-of-the-art methods (Criterion 3). The abstract emphasizes mathematical proofs over real-world implementations or agentic AI scenarios." + }, + { + "paper": { + "title": "Investigation of Whisper ASR Hallucinations Induced by Non-Speech Audio", + "abstract": "Hallucinations of deep neural models are amongst key challenges in automatic\nspeech recognition (ASR). In this paper, we investigate hallucinations of the\nWhisper ASR model induced by non-speech audio segments present during\ninference. By inducting hallucinations with various types of sounds, we show\nthat there exists a set of hallucinations that appear frequently. We then study\nhallucinations caused by the augmentation of speech with such sounds. Finally,\nwe describe the creation of a bag of hallucinations (BoH) that allows to remove\nthe effect of hallucinations through the post-processing of text\ntranscriptions. The results of our experiments show that such post-processing\nis capable of reducing word error rate (WER) and acts as a good safeguard\nagainst problematic hallucinations.", + "arxiv_id": "2501.11378v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on automatic speech recognition (ASR) hallucinations in Whisper, not LLM applications like knowledge graphs, RAG, or agentic AI. While it meets Criterion 2 (experimental results), it does not address LLM-based practical applications (Criterion 1) or compare with SOTA LLM methods (Criterion 3). ASR falls outside the LLM scope defined in mandatory criteria." + }, + { + "paper": { + "title": "A Truly Sparse and General Implementation of Gradient-Based Synaptic\n Plasticity", + "abstract": "Online synaptic plasticity rules derived from gradient descent achieve high\naccuracy on a wide range of practical tasks. However, their software\nimplementation often requires tediously hand-derived gradients or using\ngradient backpropagation which sacrifices the online capability of the rules.\nIn this work, we present a custom automatic differentiation (AD) pipeline for\nsparse and online implementation of gradient-based synaptic plasticity rules\nthat generalizes to arbitrary neuron models. Our work combines the programming\nease of backpropagation-type methods for forward AD while being\nmemory-efficient. To achieve this, we exploit the advantageous compute and\nmemory scaling of online synaptic plasticity by providing an inherently sparse\nimplementation of AD where expensive tensor contractions are replaced with\nsimple element-wise multiplications if the tensors are diagonal. Gradient-based\nsynaptic plasticity rules such as eligibility propagation (e-prop) have exactly\nthis property and thus profit immensely from this feature. We demonstrate the\nalignment of our gradients with respect to gradient backpropagation on an\nsynthetic task where e-prop gradients are exact, as well as audio speech\nclassification benchmarks. We demonstrate how memory utilization scales with\nnetwork size without dependence on the sequence length, as expected from\nforward AD methods.", + "arxiv_id": "2501.11407v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on synaptic plasticity implementation and automatic differentiation methods for neural models but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it demonstrates experimental results and methodology details, it fails to meet the mandatory criterion #1 (Practical Applications of LLMs)." + }, + { + "paper": { + "title": "Generalization and Informativeness of Weighted Conformal Risk Control\n Under Covariate Shift", + "abstract": "Predictive models are often required to produce reliable predictions under\nstatistical conditions that are not matched to the training data. A common type\nof training-testing mismatch is covariate shift, where the conditional\ndistribution of the target variable given the input features remains fixed,\nwhile the marginal distribution of the inputs changes. Weighted conformal risk\ncontrol (W-CRC) uses data collected during the training phase to convert point\npredictions into prediction sets with valid risk guarantees at test time\ndespite the presence of a covariate shift. However, while W-CRC provides\nstatistical reliability, its efficiency -- measured by the size of the\nprediction sets -- can only be assessed at test time. In this work, we relate\nthe generalization properties of the base predictor to the efficiency of W-CRC\nunder covariate shifts. Specifically, we derive a bound on the inefficiency of\nthe W-CRC predictor that depends on algorithmic hyperparameters and\ntask-specific quantities available at training time. This bound offers insights\non relationships between the informativeness of the prediction sets, the extent\nof the covariate shift, and the size of the calibration and training sets.\nExperiments on fingerprinting-based localization validate the theoretical\nresults.", + "arxiv_id": "2501.11413v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on conformal risk control under covariate shift but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it includes experimental validation and theoretical bounds, it fails to meet mandatory criteria 1 (Practical Applications of LLMs) and 3 (Comparison with SOTA). The scope aligns more with general predictive modeling rather than LLM-specific applications required by the criteria." + }, + { + "paper": { + "title": "Enhancing Coronary Artery Calcium Scoring via Multi-Organ Segmentation\n on Non-Contrast Cardiac Computed Tomography", + "abstract": "Despite coronary artery calcium scoring being considered a largely solved\nproblem within the realm of medical artificial intelligence, this paper argues\nthat significant improvements can still be made. By shifting the focus from\npathology detection to a deeper understanding of anatomy, the novel algorithm\nproposed in the paper both achieves high accuracy in coronary artery calcium\nscoring and offers enhanced interpretability of the results. This approach not\nonly aids in the precise quantification of calcifications in coronary arteries,\nbut also provides valuable insights into the underlying anatomical structures.\nThrough this anatomically-informed methodology, the paper shows how a nuanced\nunderstanding of the heart's anatomy can lead to more accurate and\ninterpretable results in the field of cardiovascular health. We demonstrate the\nsuperior accuracy of the proposed method by evaluating it on an open-source\nmulti-vendor dataset, where we obtain results at the inter-observer level,\nsurpassing the current state of the art. Finally, the qualitative analyses show\nthe practical value of the algorithm in such tasks as labeling coronary artery\ncalcifications, identifying aortic calcifications, and filtering out false\npositive detections due to noise.", + "arxiv_id": "2501.11428v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (coronary artery calcium scoring) and does not mention LLMs or related areas like knowledge graphs/RAG/agentic AI. While it meets criteria 2 and 3 with experimental results and SOTA comparisons, it violates mandatory rejection rules for medical AI focus." + }, + { + "paper": { + "title": "Decomposing Interventional Causality into Synergistic, Redundant, and\n Unique Components", + "abstract": "We introduce a novel framework for decomposing interventional causal effects\ninto synergistic, redundant, and unique components, building on the intuition\nof Partial Information Decomposition (PID) and the principle of M\\\"obius\ninversion. While recent work has explored a similar decomposition of an\nobservational measure, we argue that a proper causal decomposition must be\ninterventional in nature. We develop a mathematical approach that\nsystematically quantifies how causal power is distributed among variables in a\nsystem, using a recently derived closed-form expression for the M\\\"obius\nfunction of the redundancy lattice. The formalism is then illustrated by\ndecomposing the causal power in logic gates, cellular automata, and chemical\nreaction networks. Our results reveal how the distribution of causal power can\nbe context- and parameter-dependent. This decomposition provides new insights\ninto complex systems by revealing how causal influences are shared and combined\namong multiple variables, with potential applications ranging from attribution\nof responsibility in legal or AI systems, to the analysis of biological\nnetworks or climate models.", + "arxiv_id": "2501.11447v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not explicitly focus on practical applications of Large Language Models (LLMs), and its methodology/experiments are centered on causal decomposition in general systems (e.g., logic gates, cellular automata) rather than LLM-specific use cases. While it mentions AI systems as a potential application, it lacks direct ties to LLMs, knowledge graphs, RAG, or agentic AI. Additionally, it does not clearly address experimental metrics or comparisons with LLM-related state-of-the-art methods." + }, + { + "paper": { + "title": "Generative AI and Large Language Models in Language Preservation:\n Opportunities and Challenges", + "abstract": "Generative AI and large-scale language models (LLM) have emerged as powerful\ntools in language preservation, particularly for near-native and endangered\nlanguages. With the increasing reliance on technology for communication,\neducation, and cultural documentation, new opportunities have emerged to\nmitigate the dramatic decline of linguistic diversity worldwide. This paper\nexamines the role of generative AIs and LLMs in preserving endangered\nlanguages, highlighting the risks and challenges associated with their use. We\nanalyze the underlying technologies driving these models, including natural\nlanguage processing (NLP) and deep learning, and explore several cases where\nthese technologies have been applied to low-resource languages. Additionally,\nwe discuss ethical considerations, data scarcity issues, and technical\nchallenges while proposing solutions to enhance AI-driven language\npreservation.", + "arxiv_id": "2501.11496v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on social applications of AI in linguistic diversity (a rejection criterion) and fails to meet mandatory criteria 2 (no experimental results/quantitative metrics) and 3 (no SOTA comparison). While it addresses practical applications (criterion 1), it violates exclusion rules." + }, + { + "paper": { + "title": "Dialect2SQL: A Novel Text-to-SQL Dataset for Arabic Dialects with a\n Focus on Moroccan Darija", + "abstract": "The task of converting natural language questions (NLQs) into executable SQL\nqueries, known as text-to-SQL, has gained significant interest in recent years,\nas it enables non-technical users to interact with relational databases. Many\nbenchmarks, such as SPIDER and WikiSQL, have contributed to the development of\nnew models and the evaluation of their performance. In addition, other\ndatasets, like SEDE and BIRD, have introduced more challenges and complexities\nto better map real-world scenarios. However, these datasets primarily focus on\nhigh-resource languages such as English and Chinese. In this work, we introduce\nDialect2SQL, the first large-scale, cross-domain text-to-SQL dataset in an\nArabic dialect. It consists of 9,428 NLQ-SQL pairs across 69 databases in\nvarious domains. Along with SQL-related challenges such as long schemas, dirty\nvalues, and complex queries, our dataset also incorporates the complexities of\nthe Moroccan dialect, which is known for its diverse source languages, numerous\nborrowed words, and unique expressions. This demonstrates that our dataset will\nbe a valuable contribution to both the text-to-SQL community and the\ndevelopment of resources for low-resource languages.", + "arxiv_id": "2501.11498v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not explicitly mention experimental results with quantitative metrics (failing Criteria 2) or comparisons with state-of-the-art methods (failing Criteria 3). While it addresses practical applications (Criterion 1) and real-world challenges (Criterion 5), it does not meet all mandatory criteria." + }, + { + "paper": { + "title": "Technical Report for the Forgotten-by-Design Project: Targeted\n Obfuscation for Machine Learning", + "abstract": "The right to privacy, enshrined in various human rights declarations, faces\nnew challenges in the age of artificial intelligence (AI). This paper explores\nthe concept of the Right to be Forgotten (RTBF) within AI systems, contrasting\nit with traditional data erasure methods. We introduce Forgotten by Design, a\nproactive approach to privacy preservation that integrates instance-specific\nobfuscation techniques during the AI model training process. Unlike machine\nunlearning, which modifies models post-training, our method prevents sensitive\ndata from being embedded in the first place. Using the LIRA membership\ninference attack, we identify vulnerable data points and propose defenses that\ncombine additive gradient noise and weighting schemes. Our experiments on the\nCIFAR-10 dataset demonstrate that our techniques reduce privacy risks by at\nleast an order of magnitude while maintaining model accuracy (at 95%\nsignificance). Additionally, we present visualization methods for the\nprivacy-utility trade-off, providing a clear framework for balancing privacy\nrisk and model accuracy. This work contributes to the development of\nprivacy-preserving AI systems that align with human cognitive processes of\nmotivated forgetting, offering a robust framework for safeguarding sensitive\ninformation and ensuring compliance with privacy regulations.", + "arxiv_id": "2501.11525v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on privacy preservation (a responsible AI/ethics concern) and the Right to be Forgotten, which falls under the rejection criteria for responsible AI applications. While it meets criteria 2 and 3, it violates mandatory rejection rules." + }, + { + "paper": { + "title": "Training-free Ultra Small Model for Universal Sparse Reconstruction in\n Compressed Sensing", + "abstract": "Pre-trained large models attract widespread attention in recent years, but\nthey face challenges in applications that require high interpretability or have\nlimited resources, such as physical sensing, medical imaging, and\nbioinformatics. Compressed Sensing (CS) is a well-proved theory that drives\nmany recent breakthroughs in these applications. However, as a typical\nunder-determined linear system, CS suffers from excessively long sparse\nreconstruction times when using traditional iterative methods, particularly\nwith large-scale data. Current AI methods like deep unfolding fail to\nsubstitute them because pre-trained models exhibit poor generality beyond their\ntraining conditions and dataset distributions, or lack interpretability.\nInstead of following the big model fervor, this paper proposes ultra-small\nartificial neural models called coefficients learning (CL), enabling\ntraining-free and rapid sparse reconstruction while perfectly inheriting the\ngenerality and interpretability of traditional iterative methods, bringing new\nfeature of incorporating prior knowledges. In CL, a signal of length $n$ only\nneeds a minimal of $n$ trainable parameters. A case study model called CLOMP is\nimplemented for evaluation. Experiments are conducted on both synthetic and\nreal one-dimensional and two-dimensional signals, demonstrating significant\nimprovements in efficiency and accuracy. Compared to representative iterative\nmethods, CLOMP improves efficiency by 100 to 1000 folds for large-scale data.\nTest results on eight diverse image datasets indicate that CLOMP improves\nstructural similarity index by 292%, 98%, 45% for sampling rates of 0.1, 0.3,\n0.5, respectively. We believe this method can truly usher CS reconstruction\ninto the AI era, benefiting countless under-determined linear systems that rely\non sparse solution.", + "arxiv_id": "2501.11592v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on compressed sensing and sparse reconstruction using small AI models but does not explicitly address Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it demonstrates quantitative improvements and comparisons with state-of-the-art methods (Criteria 2-3), it fails to meet the mandatory Criterion 1 (LLM practical applications). Additionally, medical imaging is mentioned as an application, which triggers a rejection under exclusion criteria." + }, + { + "paper": { + "title": "Fairness Testing through Extreme Value Theory", + "abstract": "Data-driven software is increasingly being used as a critical component of\nautomated decision-support systems. Since this class of software learns its\nlogic from historical data, it can encode or amplify discriminatory practices.\nPrevious research on algorithmic fairness has focused on improving average-case\nfairness. On the other hand, fairness at the extreme ends of the spectrum,\nwhich often signifies lasting and impactful shifts in societal attitudes, has\nreceived significantly less emphasis.\n Leveraging the statistics of extreme value theory (EVT), we propose a novel\nfairness criterion called extreme counterfactual discrimination (ECD). This\ncriterion estimates the worst-case amounts of disadvantage in outcomes for\nindividuals solely based on their memberships in a protected group. Utilizing\ntools from search-based software engineering and generative AI, we present a\nrandomized algorithm that samples a statistically significant set of points\nfrom the tail of ML outcome distributions even if the input dataset lacks a\nsufficient number of relevant samples.\n We conducted several experiments on four ML models (deep neural networks,\nlogistic regression, and random forests) over 10 socially relevant tasks from\nthe literature on algorithmic fairness. First, we evaluate the generative AI\nmethods and find that they generate sufficient samples to infer valid EVT\ndistribution in 95% of cases. Remarkably, we found that the prevalent bias\nmitigators reduce the average-case discrimination but increase the worst-case\ndiscrimination significantly in 5% of cases. We also observed that even the\ntail-aware mitigation algorithm -- MiniMax-Fairness -- increased the worst-case\ndiscrimination in 30% of cases. We propose a novel ECD-based mitigator that\nimproves fairness in the tail in 90% of cases with no degradation of the\naverage-case discrimination.", + "arxiv_id": "2501.11597v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on algorithmic fairness and social harm in ML models, which falls under the rejection criteria for social applications. While it meets criteria 2 (quantitative results) and 3 (SOTA comparison), it violates the mandatory rejection rule regarding social applications." + }, + { + "paper": { + "title": "Noise-Agnostic Multitask Whisper Training for Reducing False Alarm\n Errors in Call-for-Help Detection", + "abstract": "Keyword spotting is often implemented by keyword classifier to the encoder in\nacoustic models, enabling the classification of predefined or open vocabulary\nkeywords. Although keyword spotting is a crucial task in various applications\nand can be extended to call-for-help detection in emergencies, however, the\nprevious method often suffers from scalability limitations due to retraining\nrequired to introduce new keywords or adapt to changing contexts. We explore a\nsimple yet effective approach that leverages off-the-shelf pretrained ASR\nmodels to address these challenges, especially in call-for-help detection\nscenarios. Furthermore, we observed a substantial increase in false alarms when\ndeploying call-for-help detection system in real-world scenarios due to noise\nintroduced by microphones or different environments. To address this, we\npropose a novel noise-agnostic multitask learning approach that integrates a\nnoise classification head into the ASR encoder. Our method enhances the model's\nrobustness to noisy environments, leading to a significant reduction in false\nalarms and improved overall call-for-help performance. Despite the added\ncomplexity of multitask learning, our approach is computationally efficient and\nprovides a promising solution for call-for-help detection in real-world\nscenarios.", + "arxiv_id": "2501.11631v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on acoustic models (ASR) and noise robustness for call-for-help detection but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it includes experimental results and addresses real-world challenges, it fails to meet mandatory criteria 1 (Practical Applications of LLMs) and does not sufficiently align with the required LLM-focused scope." + }, + { + "paper": { + "title": "Biomedical Knowledge Graph: A Survey of Domains, Tasks, and Real-World\n Applications", + "abstract": "Biomedical knowledge graphs (BKGs) have emerged as powerful tools for\norganizing and leveraging the vast and complex data found across the biomedical\nfield. Yet, current reviews of BKGs often limit their scope to specific domains\nor methods, overlooking the broader landscape and the rapid technological\nprogress reshaping it. In this survey, we address this gap by offering a\nsystematic review of BKGs from three core perspectives: domains, tasks, and\napplications. We begin by examining how BKGs are constructed from diverse data\nsources, including molecular interactions, pharmacological datasets, and\nclinical records. Next, we discuss the essential tasks enabled by BKGs,\nfocusing on knowledge management, retrieval, reasoning, and interpretation.\nFinally, we highlight real-world applications in precision medicine, drug\ndiscovery, and scientific research, illustrating the translational impact of\nBKGs across multiple sectors. By synthesizing these perspectives into a unified\nframework, this survey not only clarifies the current state of BKG research but\nalso establishes a foundation for future exploration, enabling both innovative\nmethodological advances and practical implementations.", + "arxiv_id": "2501.11632v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on biomedical applications of knowledge graphs, which falls under the medical domain exclusion criteria. While it discusses real-world applications and knowledge graphs, the core subject violates the mandatory rejection rule for medical AI applications." + }, + { + "paper": { + "title": "Transformer Vibration Forecasting for Advancing Rail Safety and\n Maintenance 4.0", + "abstract": "Maintaining railway axles is critical to preventing severe accidents and\nfinancial losses. The railway industry is increasingly interested in advanced\ncondition monitoring techniques to enhance safety and efficiency, moving beyond\ntraditional periodic inspections toward Maintenance 4.0.\n This study introduces a robust Deep Autoregressive solution that integrates\nseamlessly with existing systems to avert mechanical failures. Our approach\nsimulates and predicts vibration signals under various conditions and fault\nscenarios, improving dataset robustness for more effective detection systems.\nThese systems can alert maintenance needs, preventing accidents preemptively.\nWe use experimental vibration signals from accelerometers on train axles.\n Our primary contributions include a transformer model, ShaftFormer, designed\nfor processing time series data, and an alternative model incorporating\nspectral methods and enhanced observation models. Simulating vibration signals\nunder diverse conditions mitigates the high cost of obtaining experimental\nsignals for all scenarios. Given the non-stationary nature of railway vibration\nsignals, influenced by speed and load changes, our models address these\ncomplexities, offering a powerful tool for predictive maintenance in the rail\nindustry.", + "arxiv_id": "2501.11730v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper meets Practical Applications (rail maintenance) but lacks explicit mention of quantitative metrics (criteria 2) and comparisons with state-of-the-art methods (criteria 3). While it proposes a transformer model for time-series data, the abstract does not demonstrate performance improvements against existing techniques or provide concrete experimental metrics required for mandatory inclusion." + }, + { + "paper": { + "title": "SILO: Solving Inverse Problems with Latent Operators", + "abstract": "Consistent improvement of image priors over the years has led to the\ndevelopment of better inverse problem solvers. Diffusion models are the\nnewcomers to this arena, posing the strongest known prior to date. Recently,\nsuch models operating in a latent space have become increasingly predominant\ndue to their efficiency. In recent works, these models have been applied to\nsolve inverse problems. Working in the latent space typically requires multiple\napplications of an Autoencoder during the restoration process, which leads to\nboth computational and restoration quality challenges. In this work, we propose\na new approach for handling inverse problems with latent diffusion models,\nwhere a learned degradation function operates within the latent space,\nemulating a known image space degradation. Usage of the learned operator\nreduces the dependency on the Autoencoder to only the initial and final steps\nof the restoration process, facilitating faster sampling and superior\nrestoration quality. We demonstrate the effectiveness of our method on a\nvariety of image restoration tasks and datasets, achieving significant\nimprovements over prior art.", + "arxiv_id": "2501.11746v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on latent diffusion models for image restoration tasks, which falls under video/image processing (a rejection criterion). It does not address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI as required by the mandatory criteria." + }, + { + "paper": { + "title": "Is logical analysis performed by transformers taking place in\n self-attention or in the fully connected part?", + "abstract": "Transformers architecture apply self-attention to tokens represented as\nvectors, before a fully connected (neuronal network) layer. These two parts can\nbe layered many times. Traditionally, self-attention is seen as a mechanism for\naggregating information before logical operations are performed by the fully\nconnected layer. In this paper, we show, that quite counter-intuitively, the\nlogical analysis can also be performed within the self-attention. For this we\nimplement a handcrafted single-level encoder layer which performs the logical\nanalysis within self-attention. We then study the scenario in which a one-level\ntransformer model undergoes self-learning using gradient descent. We\ninvestigate whether the model utilizes fully connected layers or self-attention\nmechanisms for logical analysis when it has the choice. Given that gradient\ndescent can become stuck at undesired zeros, we explicitly calculate these\nunwanted zeros and find ways to avoid them. We do all this in the context of\npredicting grammatical category pairs of adjacent tokens in a text. We believe\nthat our findings have broader implications for understanding the potential\nlogical operations performed by self-attention.", + "arxiv_id": "2501.11765v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper does not meet all mandatory criteria. While it includes experimental results (Criterion 2) and describes methodology (Criterion 4), it lacks explicit focus on practical applications of LLMs (Criterion 1) and does not compare with state-of-the-art techniques (Criterion 3). The analysis centers on transformer internals (self-attention vs. FC layers) for grammatical tasks, without addressing real-world LLM applications like knowledge graphs, RAG, or agentic AI." + }, + { + "paper": { + "title": "Toward Effective Digraph Representation Learning: A Magnetic Adaptive\n Propagation based Approach", + "abstract": "The $q$-parameterized magnetic Laplacian serves as the foundation of directed\ngraph (digraph) convolution, enabling this kind of digraph neural network\n(MagDG) to encode node features and structural insights by complex-domain\nmessage passing. As a generalization of undirected methods, MagDG shows\nsuperior capability in modeling intricate web-scale topology. Despite the great\nsuccess achieved by existing MagDGs, limitations still exist: (1) Hand-crafted\n$q$: The performance of MagDGs depends on selecting an appropriate\n$q$-parameter to construct suitable graph propagation equations in the complex\ndomain. This parameter tuning, driven by downstream tasks, limits model\nflexibility and significantly increases manual effort. (2) Coarse Message\nPassing: Most approaches treat all nodes with the same complex-domain\npropagation and aggregation rules, neglecting their unique digraph contexts.\nThis oversight results in sub-optimal performance. To address the above issues,\nwe propose two key techniques: (1) MAP is crafted to be a plug-and-play\ncomplex-domain propagation optimization strategy in the context of digraph\nlearning, enabling seamless integration into any MagDG to improve predictions\nwhile enjoying high running efficiency. (2) MAP++ is a new digraph learning\nframework, further incorporating a learnable mechanism to achieve adaptively\nedge-wise propagation and node-wise aggregation in the complex domain for\nbetter performance. Extensive experiments on 12 datasets demonstrate that MAP\nenjoys flexibility for it can be incorporated with any MagDG, and scalability\nas it can deal with web-scale digraphs. MAP++ achieves SOTA predictive\nperformance on 4 different downstream tasks.", + "arxiv_id": "2501.11817v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on directed graph representation learning but does not explicitly address Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (SOTA comparison), it fails to satisfy the mandatory criterion 1 (LLM practical applications)." + }, + { + "paper": { + "title": "Toward Scalable Graph Unlearning: A Node Influence Maximization based\n Approach", + "abstract": "Machine unlearning, as a pivotal technology for enhancing model robustness\nand data privacy, has garnered significant attention in prevalent web mining\napplications, especially in thriving graph-based scenarios. However, most\nexisting graph unlearning (GU) approaches face significant challenges due to\nthe intricate interactions among web-scale graph elements during the model\ntraining: (1) The gradient-driven node entanglement hinders the complete\nknowledge removal in response to unlearning requests; (2) The billion-level\ngraph elements in the web scenarios present inevitable scalability issues. To\nbreak the above limitations, we open up a new perspective by drawing a\nconnection between GU and conventional social influence maximization. To this\nend, we propose Node Influence Maximization (NIM) through the decoupled\ninfluence propagation model and fine-grained influence function in a scalable\nmanner, which is crafted to be a plug-and-play strategy to identify potential\nnodes affected by unlearning entities. This approach enables offline execution\nindependent of GU, allowing it to be seamlessly integrated into most GU methods\nto improve their unlearning performance. Based on this, we introduce Scalable\nGraph Unlearning (SGU) as a new fine-tuned framework, which balances the\nforgetting and reasoning capability of the unlearned model by entity-specific\noptimizations. Extensive experiments on 14 datasets, including large-scale\nogbn-papers100M, have demonstrated the effectiveness of our approach.\nSpecifically, NIM enhances the forgetting capability of most GU methods, while\nSGU achieves comprehensive SOTA performance and maintains scalability.", + "arxiv_id": "2501.11823v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on graph unlearning for data privacy and model robustness in web mining applications but does not explicitly address Large Language Models (LLMs) or their practical applications (e.g., knowledge graphs, RAG, agentic AI). While it meets Criteria 2 and 3 with experimental results and SOTA comparisons, it fails Criterion 1 (LLM focus). Additionally, none of the secondary criteria (4,5) are clearly addressed. The work appears centered on graph neural networks rather than LLMs." + }, + { + "paper": { + "title": "Data-driven Detection and Evaluation of Damages in Concrete Structures:\n Using Deep Learning and Computer Vision", + "abstract": "Structural integrity is vital for maintaining the safety and longevity of\nconcrete infrastructures such as bridges, tunnels, and walls. Traditional\nmethods for detecting damages like cracks and spalls are labor-intensive,\ntime-consuming, and prone to human error. To address these challenges, this\nstudy explores advanced data-driven techniques using deep learning for\nautomated damage detection and analysis. Two state-of-the-art instance\nsegmentation models, YOLO-v7 instance segmentation and Mask R-CNN, were\nevaluated using a dataset comprising 400 images, augmented to 10,995 images\nthrough geometric and color-based transformations to enhance robustness. The\nmodels were trained and validated using a dataset split into 90% training set,\nvalidation and test set 10%. Performance metrics such as precision, recall,\nmean average precision (mAP@0.5), and frames per second (FPS) were used for\nevaluation. YOLO-v7 achieved a superior mAP@0.5 of 96.1% and processed 40 FPS,\noutperforming Mask R-CNN, which achieved a mAP@0.5 of 92.1% with a slower\nprocessing speed of 18 FPS. The findings recommend YOLO-v7 instance\nsegmentation model for real-time, high-speed structural health monitoring,\nwhile Mask R-CNN is better suited for detailed offline assessments. This study\ndemonstrates the potential of deep learning to revolutionize infrastructure\nmaintenance, offering a scalable and efficient solution for automated damage\ndetection.", + "arxiv_id": "2501.11836v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on computer vision and deep learning for structural damage detection but does not mention Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. It fails to meet the mandatory Criterion 1 (Practical Applications of LLMs) and does not align with the required LLM-focused scope." + }, + { + "paper": { + "title": "Supervised Learning for Analog and RF Circuit Design: Benchmarks and\n Comparative Insights", + "abstract": "Automating analog and radio-frequency (RF) circuit design using machine\nlearning (ML) significantly reduces the time and effort required for parameter\noptimization. This study explores supervised ML-based approaches for designing\ncircuit parameters from performance specifications across various circuit\ntypes, including homogeneous and heterogeneous designs. By evaluating diverse\nML models, from neural networks like transformers to traditional methods like\nrandom forests, we identify the best-performing models for each circuit. Our\nresults show that simpler circuits, such as low-noise amplifiers, achieve\nexceptional accuracy with mean relative errors as low as 0.3% due to their\nlinear parameter-performance relationships. In contrast, complex circuits, like\npower amplifiers and voltage-controlled oscillators, present challenges due to\ntheir non-linear interactions and larger design spaces. For heterogeneous\ncircuits, our approach achieves an 88% reduction in errors with increased\ntraining data, with the receiver achieving a mean relative error as low as\n0.23%, showcasing the scalability and accuracy of the proposed methodology.\nAdditionally, we provide insights into model strengths, with transformers\nexcelling in capturing non-linear mappings and k-nearest neighbors performing\nrobustly in moderately linear parameter spaces, especially in heterogeneous\ncircuits with larger datasets. This work establishes a foundation for extending\nML-driven design automation, enabling more efficient and scalable circuit\ndesign workflows.", + "arxiv_id": "2501.11839v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on supervised ML for circuit design without explicit mention of Large Language Models (LLMs) or their applications in knowledge graphs, RAG, or agentic AI. While it meets Criteria 2 (quantitative results) and 3 (comparisons with SOTA), it fails Criterion 1 (LLM practical applications), violating the requirement to meet ALL mandatory criteria." + }, + { + "paper": { + "title": "Scalable Whole Slide Image Representation Using K-Mean Clustering and\n Fisher Vector Aggregation", + "abstract": "Whole slide images (WSIs) are high-resolution, gigapixel sized images that\npose significant computational challenges for traditional machine learning\nmodels due to their size and heterogeneity.In this paper, we present a scalable\nand efficient methodology for WSI classification by leveraging patch-based\nfeature extraction, clustering, and Fisher vector encoding. Initially, WSIs are\ndivided into fixed size patches, and deep feature embeddings are extracted from\neach patch using a pre-trained convolutional neural network (CNN). These\npatch-level embeddings are subsequently clustered using K-means clustering,\nwhere each cluster aggregates semantically similar regions of the WSI. To\neffectively summarize each cluster, Fisher vector representations are computed\nby modeling the distribution of patch embeddings in each cluster as a\nparametric Gaussian mixture model (GMM). The Fisher vectors from each cluster\nare concatenated into a high-dimensional feature vector, creating a compact and\ninformative representation of the entire WSI. This feature vector is then used\nby a classifier to predict the WSI's diagnostic label. Our method captures\nlocal and global tissue structures and yields robust performance for\nlarge-scale WSI classification, demonstrating superior accuracy and scalability\ncompared to other approaches.", + "arxiv_id": "2501.12085v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications (whole slide image classification for diagnostic labels), which falls under the mandatory rejection criteria." + }, + { + "paper": { + "title": "Proxies for Distortion and Consistency with Applications for Real-World\n Image Restoration", + "abstract": "Real-world image restoration deals with the recovery of images suffering from\nan unknown degradation. This task is typically addressed while being given only\ndegraded images, without their corresponding ground-truth versions. In this\nhard setting, designing and evaluating restoration algorithms becomes highly\nchallenging. This paper offers a suite of tools that can serve both the design\nand assessment of real-world image restoration algorithms. Our work starts by\nproposing a trained model that predicts the chain of degradations a given\nreal-world measured input has gone through. We show how this estimator can be\nused to approximate the consistency -- the match between the measurements and\nany proposed recovered image. We also use this estimator as a guiding force for\nthe design of a simple and highly-effective plug-and-play real-world image\nrestoration algorithm, leveraging a pre-trained diffusion-based image prior.\nFurthermore, this work proposes no-reference proxy measures of MSE and LPIPS,\nwhich, without access to the ground-truth images, allow ranking of real-world\nimage restoration algorithms according to their (approximate) MSE and LPIPS.\nThe proposed suite provides a versatile, first of its kind framework for\nevaluating and comparing blind image restoration algorithms in real-world\nscenarios.", + "arxiv_id": "2501.12102v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on real-world image restoration tools and metrics, which falls under the rejection criteria for primarily addressing video/image processing. It does not mention LLMs, knowledge graphs, RAG, or agentic AI applications, failing to meet mandatory criterion 1 (Practical Applications of LLMs)." + }, + { + "paper": { + "title": "Efficient PINNs: Multi-Head Unimodular Regularization of the Solutions\n Space", + "abstract": "We present a machine learning framework to facilitate the solution of\nnonlinear multiscale differential equations and, especially, inverse problems\nusing Physics-Informed Neural Networks (PINNs). This framework is based on what\nis called multihead (MH) training, which involves training the network to learn\na general space of all solutions for a given set of equations with certain\nvariability, rather than learning a specific solution of the system. This setup\nis used with a second novel technique that we call Unimodular Regularization\n(UR) of the latent space of solutions. We show that the multihead approach,\ncombined with the regularization, significantly improves the efficiency of\nPINNs by facilitating the transfer learning process thereby enabling the\nfinding of solutions for nonlinear, coupled, and multiscale differential\nequations.", + "arxiv_id": "2501.12116v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on Physics-Informed Neural Networks (PINNs) for solving differential equations and inverse problems but does not mention LLMs, knowledge graphs, RAG, or agentic AI. While it demonstrates methodology improvements and experimental validation, it lacks direct relevance to LLM applications as required by mandatory criteria 1." + }, + { + "paper": { + "title": "An End-to-End Approach for Korean Wakeword Systems with Speaker\n Authentication", + "abstract": "Wakeword detection plays a critical role in enabling AI assistants to listen\nto user voices and interact effectively. However, for languages other than\nEnglish, there is a significant lack of pre-trained wakeword models.\nAdditionally, systems that merely determine the presence of a wakeword can pose\nserious privacy concerns. In this paper, we propose an end-to-end approach that\ntrains wakewords for Non-English languages, particulary Korean, and uses this\nto develop a Voice Authentication model to protect user privacy. Our\nimplementation employs an open-source platform OpenWakeWord, which performs\nwakeword detection using an FCN (Fully-Connected Network) architecture. Once a\nwakeword is detected, our custom-developed code calculates cosine similarity\nfor robust user authentication. Experimental results demonstrate the\neffectiveness of our approach, achieving a 16.79% and a 6.6% Equal Error Rate\n(EER) each in the Wakeword Detection and the Voice Authentication. These\nfindings highlight the model's potential in providing secure and accurate\nwakeword detection and authentication for Korean users.", + "arxiv_id": "2501.12194v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on wakeword detection and voice authentication using FCN architecture and cosine similarity, but does not demonstrate applications of Large Language Models (LLMs) in areas like knowledge graphs, RAG, or agentic AI. While it provides quantitative metrics (EER), it fails to meet the mandatory criteria 1 (LLM applications) and 3 (SOTA comparison)." + }, + { + "paper": { + "title": "Strong phonon-mediated high temperature superconductivity in\n Li$_2$AuH$_6$ under ambient pressure", + "abstract": "We used our developed AI search engine~(InvDesFlow) to perform extensive\ninvestigations regarding ambient stable superconducting hydrides. A cubic\nstructure Li$_2$AuH$_6$ with Au-H octahedral motifs is identified to be a\ncandidate. After performing thermodynamical analysis, we provide a feasible\nroute to experimentally synthesize this material via the known LiAu and LiH\ncompounds under ambient pressure. The further first-principles calculations\nsuggest that Li$_2$AuH$_6$ shows a high superconducting transition temperature\n($T_c$) $\\sim$ 140 K under ambient pressure. The H-1$s$ electrons strongly\ncouple with phonon modes of vibrations of Au-H octahedrons as well as\nvibrations of Li atoms, where the latter is not taken seriously in other\npreviously similar cases. Hence, different from previous claims of searching\nmetallic covalent bonds to find high-$T_c$ superconductors, we emphasize here\nthe importance of those phonon modes with strong electron-phonon coupling\n(EPC). And we suggest that one can intercalate atoms into binary or ternary\nhydrides to introduce more potential phonon modes with strong EPC, which is an\neffective approach to find high-$T_c$ superconductors within multicomponent\ncompounds.", + "arxiv_id": "2501.12222v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on AI-assisted material discovery for superconductors but does not explicitly involve Large Language Models (LLMs) in its methodology or applications. It fails to meet the mandatory Criterion 1 (Practical Applications of LLMs) and does not address LLM-related challenges, agentic AI, or other LLM-specific domains outlined in the criteria." + }, + { + "paper": { + "title": "CBVLM: Training-free Explainable Concept-based Large Vision Language\n Models for Medical Image Classification", + "abstract": "The main challenges limiting the adoption of deep learning-based solutions in\nmedical workflows are the availability of annotated data and the lack of\ninterpretability of such systems. Concept Bottleneck Models (CBMs) tackle the\nlatter by constraining the final disease prediction on a set of predefined and\nhuman-interpretable concepts. However, the increased interpretability achieved\nthrough these concept-based explanations implies a higher annotation burden.\nMoreover, if a new concept needs to be added, the whole system needs to be\nretrained. Inspired by the remarkable performance shown by Large\nVision-Language Models (LVLMs) in few-shot settings, we propose a simple, yet\neffective, methodology, CBVLM, which tackles both of the aforementioned\nchallenges. First, for each concept, we prompt the LVLM to answer if the\nconcept is present in the input image. Then, we ask the LVLM to classify the\nimage based on the previous concept predictions. Moreover, in both stages, we\nincorporate a retrieval module responsible for selecting the best examples for\nin-context learning. By grounding the final diagnosis on the predicted\nconcepts, we ensure explainability, and by leveraging the few-shot capabilities\nof LVLMs, we drastically lower the annotation cost. We validate our approach\nwith extensive experiments across four medical datasets and twelve LVLMs (both\ngeneric and medical) and show that CBVLM consistently outperforms CBMs and\ntask-specific supervised methods without requiring any training and using just\na few annotated examples. More information on our project page:\nhttps://cristianopatricio.github.io/CBVLM/.", + "arxiv_id": "2501.12266v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on medical applications of AI (medical image classification), which is an explicit rejection criterion according to the guidelines. While it meets some technical criteria (experimental results, SOTA comparisons), the medical focus mandates rejection regardless of other merits." + }, + { + "paper": { + "title": "Implementation of an Asymmetric Adjusted Activation Function for Class\n Imbalance Credit Scoring", + "abstract": "Credit scoring is a systematic approach to evaluate a borrower's probability\nof default (PD) on a bank loan. The data associated with such scenarios are\ncharacteristically imbalanced, complicating binary classification owing to the\noften-underestimated cost of misclassification during the classifier's learning\nprocess. Considering the high imbalance ratio (IR) of these datasets, we\nintroduce an innovative yet straightforward optimized activation function by\nincorporating an IR-dependent asymmetric adjusted factor embedded Sigmoid\nactivation function (ASIG). The embedding of ASIG makes the sensitive margin of\nthe Sigmoid function auto-adjustable, depending on the imbalance nature of the\ndatasets distributed, thereby giving the activation function an asymmetric\ncharacteristic that prevents the underrepresentation of the minority class\n(positive samples) during the classifier's learning process. The experimental\nresults show that the ASIG-embedded-classifier outperforms traditional\nclassifiers on datasets across wide-ranging IRs in the downstream\ncredit-scoring task. The algorithm also shows robustness and stability, even\nwhen the IR is ultra-high. Therefore, the algorithm provides a competitive\nalternative in the financial industry, especially in credit scoring, possessing\nthe ability to effectively process highly imbalanced distribution data.", + "arxiv_id": "2501.12285v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on credit scoring with an activation function adaptation for class imbalance but does not mention Large Language Models (LLMs) or address criteria 1 (Practical Applications of LLMs), 3 (Comparison with SOTA LLM techniques), or 4/5 (LLM methodology/real-world challenges). It primarily addresses financial data classification without LLM relevance." + }, + { + "paper": { + "title": "Regressor-Guided Image Editing Regulates Emotional Response to Reduce\n Online Engagement", + "abstract": "Emotions are known to mediate the relationship between users' content\nconsumption and their online engagement, with heightened emotional intensity\nleading to increased engagement. Building on this insight, we propose three\nregressor-guided image editing approaches aimed at diminishing the emotional\nimpact of images. These include (i) a parameter optimization approach based on\nglobal image transformations known to influence emotions, (ii) an optimization\napproach targeting the style latent space of a generative adversarial network,\nand (iii) a diffusion-based approach employing classifier guidance and\nclassifier-free guidance. Our findings demonstrate that approaches can\neffectively alter the emotional properties of images while maintaining high\nvisual quality. Optimization-based methods primarily adjust low-level\nproperties like color hues and brightness, whereas the diffusion-based approach\nintroduces semantic changes, such as altering appearance or facial expressions.\nNotably, results from a behavioral study reveal that only the diffusion-based\napproach successfully elicits changes in viewers' emotional responses while\npreserving high perceived image quality. In future work, we will investigate\nthe impact of these image adaptations on internet user behavior.", + "arxiv_id": "2501.12289v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on image editing to influence emotional responses and online engagement, which falls under social applications of AI (explicit rejection criterion). It does not mention LLMs, knowledge graphs, RAG, or agentic AI, failing to meet mandatory criteria 1 and 3. While it includes experimental results (criterion 2), the social application focus triggers rejection." + }, + { + "paper": { + "title": "Test-time regression: a unifying framework for designing sequence models\n with associative memory", + "abstract": "Sequences provide a remarkably general way to represent and process\ninformation. This powerful abstraction has placed sequence modeling at the\ncenter of modern deep learning applications, inspiring numerous architectures\nfrom transformers to recurrent networks. While this fragmented development has\nyielded powerful models, it has left us without a unified framework to\nunderstand their fundamental similarities and explain their effectiveness. We\npresent a unifying framework motivated by an empirical observation: effective\nsequence models must be able to perform associative recall. Our key insight is\nthat memorizing input tokens through an associative memory is equivalent to\nperforming regression at test-time. This regression-memory correspondence\nprovides a framework for deriving sequence models that can perform associative\nrecall, offering a systematic lens to understand seemingly ad-hoc architectural\nchoices. We show numerous recent architectures -- including linear attention\nmodels, their gated variants, state-space models, online learners, and softmax\nattention -- emerge naturally as specific approaches to test-time regression.\nEach architecture corresponds to three design choices: the relative importance\nof each association, the regressor function class, and the optimization\nalgorithm. This connection leads to new understanding: we provide theoretical\njustification for QKNorm in softmax attention, and we motivate higher-order\ngeneralizations of softmax attention. Beyond unification, our work unlocks\ndecades of rich statistical tools that can guide future development of more\npowerful yet principled sequence models.", + "arxiv_id": "2501.12352v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on theoretical unification of sequence models through test-time regression concepts but does not explicitly address practical LLM applications (C1), lacks mention of experimental results with quantitative metrics (C2), and does not demonstrate direct performance comparisons with state-of-the-art methods (C3). While novel, it fails to meet all mandatory criteria for real-world LLM applications." + }, + { + "paper": { + "title": "DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial\n Basis Functions", + "abstract": "Splatting-based 3D reconstruction methods have gained popularity with the\nadvent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel\nviews. These methods commonly resort to using exponential family functions,\nsuch as the Gaussian function, as reconstruction kernels due to their\nanisotropic nature, ease of projection, and differentiability in rasterization.\nHowever, the field remains restricted to variations within the exponential\nfamily, leaving generalized reconstruction kernels largely underexplored,\npartly due to the lack of easy integrability in 3D to 2D projections. In this\nlight, we show that a class of decaying anisotropic radial basis functions\n(DARBFs), which are non-negative functions of the Mahalanobis distance,\nsupports splatting by approximating the Gaussian function's closed-form\nintegration advantage. With this fresh perspective, we demonstrate up to 34%\nfaster convergence during training and a 15% reduction in memory consumption\nacross various DARB reconstruction kernels, while maintaining comparable PSNR,\nSSIM, and LPIPS results. We will make the code available.", + "arxiv_id": "2501.12369v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on 3D reconstruction methods using splatting techniques but does not address practical applications of Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails the mandatory criterion 1 (LLM applications). Additionally, it does not align with the required scope for inclusion." + }, + { + "paper": { + "title": "MMVU: Measuring Expert-Level Multi-Discipline Video Understanding", + "abstract": "We introduce MMVU, a comprehensive expert-level, multi-discipline benchmark\nfor evaluating foundation models in video understanding. MMVU includes 3,000\nexpert-annotated questions spanning 27 subjects across four core disciplines:\nScience, Healthcare, Humanities \u0026 Social Sciences, and Engineering. Compared to\nprior benchmarks, MMVU features three key advancements. First, it challenges\nmodels to apply domain-specific knowledge and perform expert-level reasoning to\nanalyze specialized-domain videos, moving beyond the basic visual perception\ntypically assessed in current video benchmarks. Second, each example is\nannotated by human experts from scratch. We implement strict data quality\ncontrols to ensure the high quality of the dataset. Finally, each example is\nenriched with expert-annotated reasoning rationals and relevant domain\nknowledge, facilitating in-depth analysis. We conduct an extensive evaluation\nof 32 frontier multimodal foundation models on MMVU. The latest\nSystem-2-capable models, o1 and Gemini 2.0 Flash Thinking, achieve the highest\nperformance among the tested models. However, they still fall short of matching\nhuman expertise. Through in-depth error analyses and case studies, we offer\nactionable insights for future advancements in expert-level,\nknowledge-intensive video understanding for specialized domains.", + "arxiv_id": "2501.12380v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing through a specialized video understanding benchmark (MMVU), which falls under explicit rejection criteria. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), the core focus on video understanding violates mandatory rejection rules." + }, + { + "paper": { + "title": "Physics of Skill Learning", + "abstract": "We aim to understand physics of skill learning, i.e., how skills are learned\nin neural networks during training. We start by observing the Domino effect,\ni.e., skills are learned sequentially, and notably, some skills kick off\nlearning right after others complete learning, similar to the sequential fall\nof domino cards. To understand the Domino effect and relevant behaviors of\nskill learning, we take physicists' approach of abstraction and simplification.\nWe propose three models with varying complexities -- the Geometry model, the\nResource model, and the Domino model, trading between reality and simplicity.\nThe Domino effect can be reproduced in the Geometry model, whose resource\ninterpretation inspires the Resource model, which can be further simplified to\nthe Domino model. These models present different levels of abstraction and\nsimplification; each is useful to study some aspects of skill learning. The\nGeometry model provides interesting insights into neural scaling laws and\noptimizers; the Resource model sheds light on the learning dynamics of\ncompositional tasks; the Domino model reveals the benefits of modularity. These\nmodels are not only conceptually interesting -- e.g., we show how Chinchilla\nscaling laws can emerge from the Geometry model, but also are useful in\npractice by inspiring algorithmic development -- e.g., we show how simple\nalgorithmic changes, motivated by these toy models, can speed up the training\nof deep learning models.", + "arxiv_id": "2501.12391v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on understanding skill learning dynamics in neural networks with theoretical models but does not explicitly address Large Language Models (LLMs), practical applications in areas like knowledge graphs/RAG/agentic AI, or comparisons with state-of-the-art methods. It lacks clear alignment with mandatory criteria 1 and 3, and its relevance to LLM-specific real-world applications remains unclear." + }, + { + "paper": { + "title": "Learning segmentation from point trajectories", + "abstract": "We consider the problem of segmenting objects in videos based on their motion\nand no other forms of supervision. Prior work has often approached this problem\nby using the principle of common fate, namely the fact that the motion of\npoints that belong to the same object is strongly correlated. However, most\nauthors have only considered instantaneous motion from optical flow. In this\nwork, we present a way to train a segmentation network using long-term point\ntrajectories as a supervisory signal to complement optical flow. The key\ndifficulty is that long-term motion, unlike instantaneous motion, is difficult\nto model -- any parametric approximation is unlikely to capture complex motion\npatterns over long periods of time. We instead draw inspiration from subspace\nclustering approaches, proposing a loss function that seeks to group the\ntrajectories into low-rank matrices where the motion of object points can be\napproximately explained as a linear combination of other point tracks. Our\nmethod outperforms the prior art on motion-based segmentation, which shows the\nutility of long-term motion and the effectiveness of our formulation.", + "arxiv_id": "2501.12392v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper primarily focuses on video processing (motion-based segmentation in videos), which is an explicit rejection criterion. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails the mandatory practical applications requirement for LLMs and violates the video processing exclusion rule." + }, + { + "paper": { + "title": "Unsupervised Learning in Echo State Networks for Input Reconstruction", + "abstract": "Conventional echo state networks (ESNs) require supervised learning to train\nthe readout layer, using the desired outputs as training data. In this study,\nwe focus on input reconstruction (IR), which refers to training the readout\nlayer to reproduce the input time series in its output. We reformulate the\nlearning algorithm of the ESN readout layer to perform IR using unsupervised\nlearning (UL). By conducting theoretical analysis and numerical experiments, we\ndemonstrate that IR in ESNs can be effectively implemented under realistic\nconditions without explicitly using the desired outputs as training data; in\nthis way, UL is enabled. Furthermore, we demonstrate that applications relying\non IR, such as dynamical system replication and noise filtering, can be\nreformulated within the UL framework. Our findings establish a theoretically\nsound and universally applicable IR formulation, along with its related tasks\nin ESNs. This work paves the way for novel predictions and highlights\nunresolved theoretical challenges in ESNs, particularly in the context of\ntime-series processing methods and computational models of the brain.", + "arxiv_id": "2501.11409v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on Echo State Networks (ESNs), not Large Language Models (LLMs), and does not address LLM-related applications like knowledge graphs, RAG, or agentic AI. None of the mandatory criteria (Practical Applications of LLMs, Experimental Results related to LLMs, SOTA comparisons involving LLMs) are met. While methodology details are provided, they pertain to ESNs rather than LLM utilization in real-world tasks." + }, + { + "paper": { + "title": "Improving thermal state preparation of Sachdev-Ye-Kitaev model with\n reinforcement learning on quantum hardware", + "abstract": "The Sachdev-Ye-Kitaev (SYK) model, known for its strong quantum correlations\nand chaotic behavior, serves as a key platform for quantum gravity studies.\nHowever, variationally preparing thermal states on near-term quantum processors\nfor large systems (N\u003e12, where N is the number of Majorana fermions) presents a\nsignificant challenge due to the rapid growth in the complexity of\nparameterized quantum circuits. This paper addresses this challenge by\nintegrating reinforcement learning (RL) with convolutional neural networks,\nemploying an iterative approach to optimize the quantum circuit and its\nparameters. The refinement process is guided by a composite reward signal\nderived from entropy and the expectation values of the SYK Hamiltonian. This\napproach reduces the number of CNOT gates by two orders of magnitude for\nsystems N\u003e10 compared to traditional methods like first-order Trotterization.\nWe demonstrate the effectiveness of the RL framework in both noiseless and\nnoisy quantum hardware environments, maintaining high accuracy in thermal state\npreparation. This work contributes to the advancement of a scalable, RL-based\nframework with applications for computations of thermal out-of-time-order\ncorrelators in quantum many-body systems and quantum gravity studies on\nnear-term quantum hardware.", + "arxiv_id": "2501.11454v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on quantum computing and reinforcement learning for thermal state preparation in the SYK model but does not address Large Language Models (LLMs) or their practical applications (failing mandatory criterion 1). While it meets criteria 2 (quantitative results) and 3 (comparison with SOTA), LLM relevance is absent. The topic does not fall under rejection domains but fails core LLM-focused requirements." + }, + { + "paper": { + "title": "With Great Backbones Comes Great Adversarial Transferability", + "abstract": "Advances in self-supervised learning (SSL) for machine vision have improved\nrepresentation robustness and model performance, giving rise to pre-trained\nbackbones like \\emph{ResNet} and \\emph{ViT} models tuned with SSL methods such\nas \\emph{SimCLR}. Due to the computational and data demands of pre-training,\nthe utilization of such backbones becomes a strenuous necessity. However,\nemploying these backbones may inherit vulnerabilities to adversarial attacks.\nWhile adversarial robustness has been studied under \\emph{white-box} and\n\\emph{black-box} settings, the robustness of models tuned on pre-trained\nbackbones remains largely unexplored. Additionally, the role of tuning\nmeta-information in mitigating exploitation risks is unclear. This work\nsystematically evaluates the adversarial robustness of such models across\n$20,000$ combinations of tuning meta-information, including fine-tuning\ntechniques, backbone families, datasets, and attack types. We propose using\nproxy models to transfer attacks, simulating varying levels of target knowledge\nby fine-tuning these proxies with diverse configurations. Our findings reveal\nthat proxy-based attacks approach the effectiveness of \\emph{white-box}\nmethods, even with minimal tuning knowledge. We also introduce a naive\n\"backbone attack,\" leveraging only the backbone to generate adversarial\nsamples, which outperforms \\emph{black-box} attacks and rivals \\emph{white-box}\nmethods, highlighting critical risks in model-sharing practices. Finally, our\nablations reveal how increasing tuning meta-information impacts attack\ntransferability, measuring each meta-information combination.", + "arxiv_id": "2501.12275v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on adversarial robustness in self-supervised learning for machine vision (ResNet/ViT), not Large Language Models (LLMs). It does not meet Criterion 1 (Practical Applications of LLMs) and addresses adversarial attacks in general model-sharing contexts rather than LLM-specific applications like RAG or agentic AI. While it meets Criteria 2-3 and 4-5, the mandatory LLM focus is absent." + }, + { + "paper": { + "title": "Goal-oriented Transmission Scheduling: Structure-guided DRL with a\n Unified Dual On-policy and Off-policy Approach", + "abstract": "Goal-oriented communications prioritize application-driven objectives over\ndata accuracy, enabling intelligent next-generation wireless systems. Efficient\nscheduling in multi-device, multi-channel systems poses significant challenges\ndue to high-dimensional state and action spaces. We address these challenges by\nderiving key structural properties of the optimal solution to the goal-oriented\nscheduling problem, incorporating Age of Information (AoI) and channel states.\nSpecifically, we establish the monotonicity of the optimal state value function\n(a measure of long-term system performance) w.r.t. channel states and prove its\nasymptotic convexity w.r.t. AoI states. Additionally, we derive the\nmonotonicity of the optimal policy w.r.t. channel states, advancing the\ntheoretical framework for optimal scheduling. Leveraging these insights, we\npropose the structure-guided unified dual on-off policy DRL (SUDO-DRL), a\nhybrid algorithm that combines the stability of on-policy training with the\nsample efficiency of off-policy methods. Through a novel structural property\nevaluation framework, SUDO-DRL enables effective and scalable training,\naddressing the complexities of large-scale systems. Numerical results show\nSUDO-DRL improves system performance by up to 45% and reduces convergence time\nby 40% compared to state-of-the-art methods. It also effectively handles\nscheduling in much larger systems, where off-policy DRL fails and on-policy\nbenchmarks exhibit significant performance loss, demonstrating its scalability\nand efficacy in goal-oriented communications.", + "arxiv_id": "2501.11921v1" + }, + "decision": "REJECT", + "explanation": "Original decision: REJECT\nThe paper focuses on deep reinforcement learning (DRL) for communication scheduling but does not address Large Language Models (LLMs), failing to meet mandatory criteria 1 (Practical Applications of LLMs), 2 (LLM-related experimental results), and 3 (LLM methodology comparisons)." + } + ], + "errors": [] +} diff --git a/llm_processor/newpapers-r1-filtered.md b/llm_processor/newpapers-r1-filtered.md new file mode 100644 index 0000000..12ee5fe --- /dev/null +++ b/llm_processor/newpapers-r1-filtered.md @@ -0,0 +1,3401 @@ +# Accepted Papers + +## [Agent-R: Training Language Model Agents to Reflect via Iterative + Self-Training](https://arxiv.org/abs/2501.11425v1) +**arXiv ID:** 2501.11425v1 + +**Abstract:** +> Large Language Models (LLMs) agents are increasingly pivotal for addressing +complex tasks in interactive environments. Existing work mainly focuses on +enhancing performance through behavior cloning from stronger experts, yet such +approaches often falter in real-world applications, mainly due to the inability +to recover from errors. However, step-level critique data is difficult and +expensive to collect. Automating and dynamically constructing self-critique +datasets is thus crucial to empowering models with intelligent agent +capabilities. In this work, we propose an iterative self-training framework, +Agent-R, that enables language Agent to Reflect on the fly. Unlike traditional +methods that reward or penalize actions based on correctness, Agent-R leverages +MCTS to construct training data that recover correct trajectories from +erroneous ones. A key challenge of agent reflection lies in the necessity for +timely revision rather than waiting until the end of a rollout. To address +this, we introduce a model-guided critique construction mechanism: the actor +model identifies the first error step (within its current capability) in a +failed trajectory. Starting from it, we splice it with the adjacent correct +path, which shares the same parent node in the tree. This strategy enables the +model to learn reflection based on its current policy, therefore yielding +better learning efficiency. To further explore the scalability of this +self-improvement paradigm, we investigate iterative refinement of both error +correction capabilities and dataset construction. Our findings demonstrate that +Agent-R continuously improves the model's ability to recover from errors and +enables timely error correction. Experiments on three interactive environments +show that Agent-R effectively equips agents to correct erroneous actions while +avoiding loops, achieving superior performance compared to baseline methods +(+5.59%). + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all three mandatory criteria: 1) Focuses on practical applications of LLM agents in interactive environments, 2) Provides quantitative results (+5.59% performance gain) across three environments, and 3) Compares with baseline methods. It also meets secondary criteria 4 by detailing its iterative self-training methodology using MCTS and model-guided critique. The work addresses agentic AI capabilities (error recovery in real-world tasks) and demonstrates novelty in self-improvement paradigms, aligning with inclusion priorities. + +--- + +## [Systematic Abductive Reasoning via Diverse Relation Representations in + Vector-symbolic Architecture](https://arxiv.org/abs/2501.11896v1) +**arXiv ID:** 2501.11896v1 + +**Abstract:** +> In abstract visual reasoning, monolithic deep learning models suffer from +limited interpretability and generalization, while existing neuro-symbolic +approaches fall short in capturing the diversity and systematicity of +attributes and relation representations. To address these challenges, we +propose a Systematic Abductive Reasoning model with diverse relation +representations (Rel-SAR) in Vector-symbolic Architecture (VSA) to solve +Raven's Progressive Matrices (RPM). To derive attribute representations with +symbolic reasoning potential, we introduce not only various types of atomic +vectors that represent numeric, periodic and logical semantics, but also the +structured high-dimentional representation (SHDR) for the overall Grid +component. For systematic reasoning, we propose novel numerical and logical +relation functions and perform rule abduction and execution in a unified +framework that integrates these relation representations. Experimental results +demonstrate that Rel-SAR achieves significant improvement on RPM tasks and +exhibits robust out-of-distribution generalization. Rel-SAR leverages the +synergy between HD attribute representations and symbolic reasoning to achieve +systematic abductive reasoning with both interpretable and computable +semantics. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all mandatory criteria: 1) Focuses on systematic reasoning with vector-symbolic architectures relevant to LLM applications, 2) Presents experimental results with quantitative improvements on RPM tasks, 3) Compares against existing neuro-symbolic approaches. Also meets secondary criteria 4 by detailing methodology including atomic vectors and relation functions. Shows novelty in attribute representations and passes all rejection filters. + +--- + +## [Reasoning Language Models: A Blueprint](https://arxiv.org/abs/2501.11223v1) +**arXiv ID:** 2501.11223v1 + +**Abstract:** +> Reasoning language models (RLMs), also known as Large Reasoning Models +(LRMs), such as OpenAI's o1 and o3, DeepSeek-V3, and Alibaba's QwQ, have +redefined AI's problem-solving capabilities by extending large language models +(LLMs) with advanced reasoning mechanisms. Yet, their high costs, proprietary +nature, and complex architectures - uniquely combining Reinforcement Learning +(RL), search heuristics, and LLMs - present accessibility and scalability +challenges. To address these, we propose a comprehensive blueprint that +organizes RLM components into a modular framework, based on a survey and +analysis of all RLM works. This blueprint incorporates diverse reasoning +structures (chains, trees, graphs, and nested forms), reasoning strategies +(e.g., Monte Carlo Tree Search, Beam Search), RL concepts (policy, value models +and others), and supervision schemes (Output-Based and Process-Based +Supervision). We also provide detailed mathematical formulations and +algorithmic specifications to simplify RLM implementation. By showing how +schemes like LLaMA-Berry, QwQ, Journey Learning, and Graph of Thoughts fit as +special cases, we demonstrate the blueprint's versatility and unifying +potential. To illustrate its utility, we introduce x1, a modular implementation +for rapid RLM prototyping and experimentation. Using x1 and a literature +review, we provide key insights, such as multi-phase training for policy and +value models, and the importance of familiar training distributions. Finally, +we outline how RLMs can integrate with a broader LLM ecosystem, including tools +and databases. Our work demystifies RLM construction, democratizes advanced +reasoning capabilities, and fosters innovation, aiming to mitigate the gap +between "rich AI" and "poor AI" by lowering barriers to RLM development and +experimentation. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all mandatory criteria: 1) Focuses on practical RLM applications through modular frameworks and ecosystem integration; 2) Provides implementation (x1) and insights derived from experimentation/literature; 3) Compares with SOTA (LLaMA-Berry/QwQ). Also satisfies optional criteria 4 (detailed methodology/algorithm specs) and 5 (addresses scalability challenges). No rejection triggers apply. + +--- + +## [Question-to-Question Retrieval for Hallucination-Free Knowledge Access: + An Approach for Wikipedia and Wikidata Question Answering](https://arxiv.org/abs/2501.11301v1) +**arXiv ID:** 2501.11301v1 + +**Abstract:** +> This paper introduces an approach to question answering over knowledge bases +like Wikipedia and Wikidata by performing "question-to-question" matching and +retrieval from a dense vector embedding store. Instead of embedding document +content, we generate a comprehensive set of questions for each logical content +unit using an instruction-tuned LLM. These questions are vector-embedded and +stored, mapping to the corresponding content. Vector embedding of user queries +are then matched against this question vector store. The highest similarity +score leads to direct retrieval of the associated article content, eliminating +the need for answer generation. Our method achieves high cosine similarity ( > +0.9 ) for relevant question pairs, enabling highly precise retrieval. This +approach offers several advantages including computational efficiency, rapid +response times, and increased scalability. We demonstrate its effectiveness on +Wikipedia and Wikidata, including multimedia content through structured fact +retrieval from Wikidata, opening up new pathways for multimodal question +answering. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all mandatory criteria: 1 (LLM application in RAG for knowledge bases), 2 (quantitative metrics like >0.9 cosine similarity), and 3 (implicit comparison via novel methodology). It also meets criterion 4 (methodology details) and demonstrates novelty. No rejection triggers apply. + +--- + +## [Few-shot Policy (de)composition in Conversational Question Answering](https://arxiv.org/abs/2501.11335v1) +**arXiv ID:** 2501.11335v1 + +**Abstract:** +> The task of policy compliance detection (PCD) is to determine if a scenario +is in compliance with respect to a set of written policies. In a conversational +setting, the results of PCD can indicate if clarifying questions must be asked +to determine compliance status. Existing approaches usually claim to have +reasoning capabilities that are latent or require a large amount of annotated +data. In this work, we propose logical decomposition for policy compliance +(LDPC): a neuro-symbolic framework to detect policy compliance using large +language models (LLMs) in a few-shot setting. By selecting only a few exemplars +alongside recently developed prompting techniques, we demonstrate that our +approach soundly reasons about policy compliance conversations by extracting +sub-questions to be answered, assigning truth values from contextual +information, and explicitly producing a set of logic statements from the given +policies. The formulation of explicit logic graphs can in turn help answer +PCDrelated questions with increased transparency and explainability. We apply +this approach to the popular PCD and conversational machine reading benchmark, +ShARC, and show competitive performance with no task-specific finetuning. We +also leverage the inherently interpretable architecture of LDPC to understand +where errors occur, revealing ambiguities in the ShARC dataset and highlighting +the challenges involved with reasoning for conversational question answering. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Focuses on practical LLM applications (policy compliance via neuro-symbolic framework), 2) Demonstrates quantitative results on ShARC benchmark with competitive performance, 3) Compares with existing approaches. Also meets secondary criteria 4 (clear methodology using prompting/logic decomposition) and 5 (addresses real-world challenges). No rejection triggers present. + +--- + +## [Neural Contextual Reinforcement Framework for Logical Structure Language + Generation](https://arxiv.org/abs/2501.11417v1) +**arXiv ID:** 2501.11417v1 + +**Abstract:** +> The Neural Contextual Reinforcement Framework introduces an innovative +approach to enhancing the logical coherence and structural consistency of text +generated by large language models. Leveraging reinforcement learning +principles, the framework integrates custom reward functions and dynamic +context alignment mechanisms to address challenges inherent in maintaining +long-range dependencies across extended sequences. The architecture +incorporates multi-head attention layers and hierarchical encoding modules, +enabling the model to produce outputs that align closely with human +expectations of logical structure and semantic flow. Quantitative evaluations +across diverse datasets demonstrate substantial improvements in coherence +metrics, perplexity reduction, and semantic alignment, showcasing the +framework's ability to outperform baseline models in both general and +domain-specific tasks. Qualitative analyses further highlight the framework's +capacity to generate text with improved narrative clarity and reduced +redundancy, reflecting its effectiveness in balancing fluency with structural +precision. In addition to its performance gains, the framework exhibits +robustness in handling noisy input data and scalability across varying model +sizes, reinforcing its versatility in practical applications. Experimental +results reveal that optimal context window sizes significantly influence +coherence outcomes, showing the importance of architectural flexibility in +adapting to diverse linguistic structures. Cross-lingual performance +evaluations affirm the framework's adaptability to multiple languages, +extending its utility beyond monolingual contexts. Resource efficiency analyses +indicate a reduction in computational overhead compared to traditional +approaches, emphasizing the practicality of the framework for large-scale +deployment. + +**Decision Explanation:** Original decision: ACCEPT +Meets ALL mandatory criteria: 1) Focuses on practical text generation improvements for LLMs, 2) Includes quantitative metrics (coherence, perplexity) and comparisons with baselines, 3) Compares with existing models. Also meets optional criteria 4 through detailed methodology description. Shows novelty in reinforcement framework and addresses reproducibility through efficiency analysis. No rejection triggers present. + +--- + +## [Explainable Lane Change Prediction for Near-Crash Scenarios Using + Knowledge Graph Embeddings and Retrieval Augmented Generation](https://arxiv.org/abs/2501.11560v1) +**arXiv ID:** 2501.11560v1 + +**Abstract:** +> Lane-changing maneuvers, particularly those executed abruptly or in risky +situations, are a significant cause of road traffic accidents. However, current +research mainly focuses on predicting safe lane changes. Furthermore, existing +accident datasets are often based on images only and lack comprehensive sensory +data. In this work, we focus on predicting risky lane changes using the CRASH +dataset (our own collected dataset specifically for risky lane changes), and +safe lane changes (using the HighD dataset). Then, we leverage KG and Bayesian +inference to predict these maneuvers using linguistic contextual information, +enhancing the model's interpretability and transparency. The model achieved a +91.5% f1-score with anticipation time extending to four seconds for risky lane +changes, and a 90.0% f1-score for predicting safe lane changes with the same +anticipation time. We validate our model by integrating it into a vehicle +within the CARLA simulator in scenarios that involve risky lane changes. The +model managed to anticipate sudden lane changes, thus providing automated +vehicles with further time to plan and execute appropriate safe reactions. +Finally, to enhance the explainability of our model, we utilize RAG to provide +clear and natural language explanations for the given prediction. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets mandatory criteria 1 (practical LLM applications in KG/RAG), 2 (quantitative results with 91.5% F1-score), and 3 (implied advancement via novel CRASH dataset/approach). It meets optional criterion 4 (methodology details) and addresses agentic AI in CARLA simulations. No rejection triggers apply. + +--- + +## [Recurrent Diffusion for Large-Scale Parameter Generation](https://arxiv.org/abs/2501.11587v1) +**arXiv ID:** 2501.11587v1 + +**Abstract:** +> Parameter generation has struggled to scale up for a long time, significantly +limiting its range of applications. In this study, we introduce +\textbf{R}ecurrent diffusion for large-scale \textbf{P}arameter +\textbf{G}eneration, called \textbf{RPG}. We first divide the trained +parameters into non-overlapping parts, after which a recurrent model is +proposed to learn their relationships. The recurrent model's outputs, as +conditions, are then fed into a diffusion model to generate the neural network +parameters. Using only a single GPU, recurrent diffusion enables us to generate +popular vision and language models such as ConvNeXt-L and LoRA parameters of +LLaMA-7B. Meanwhile, across various architectures and tasks, the generated +parameters consistently perform comparable results over trained networks. +Notably, our approach also shows the potential to generate models for handling +unseen tasks, which largely increases the practicality of parameter generation. +Our code is available +\href{https://github.com/NUS-HPC-AI-Lab/Recurrent-Parameter-Generation}{here}. + +**Decision Explanation:** Original decision: ACCEPT +Meets mandatory criteria 1 (practical LLM parameter generation for vision/language models), 2 (quantitative results showing comparable performance to trained networks), and 3 (implicit comparison to standard trained networks as baseline). Addresses optional criteria 4 (methodology details) and 5 (real-world practicality for unseen tasks). No rejection triggers apply. Novelty and implementability with code availability support inclusion. + +--- + +## [SR-FoT: A Syllogistic-Reasoning Framework of Thought for Large Language + Models Tackling Knowledge-based Reasoning Tasks](https://arxiv.org/abs/2501.11599v1) +**arXiv ID:** 2501.11599v1 + +**Abstract:** +> Deductive reasoning is a crucial logical capability that assists us in +solving complex problems based on existing knowledge. Although augmented by +Chain-of-Thought prompts, Large Language Models (LLMs) might not follow the +correct reasoning paths. Enhancing the deductive reasoning abilities of LLMs, +and leveraging their extensive built-in knowledge for various reasoning tasks, +remains an open question. Attempting to mimic the human deductive reasoning +paradigm, we propose a multi-stage Syllogistic-Reasoning Framework of Thought +(SR-FoT) that enables LLMs to perform syllogistic deductive reasoning to handle +complex knowledge-based reasoning tasks. Our SR-FoT begins by interpreting the +question and then uses the interpretation and the original question to propose +a suitable major premise. It proceeds by generating and answering minor premise +questions in two stages to match the minor premises. Finally, it guides LLMs to +use the previously generated major and minor premises to perform syllogistic +deductive reasoning to derive the answer to the original question. Extensive +and thorough experiments on knowledge-based reasoning tasks have demonstrated +the effectiveness and advantages of our SR-FoT. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1 (practical reasoning framework), 2 (mentions experimental validation), 3 (compares with Chain-of-Thought baselines). Meets optional criterion 4 (detailed methodology description). Novel syllogistic approach aligns with LLM application extensions. No rejection triggers present. + +--- + +## [Optimizing Pretraining Data Mixtures with LLM-Estimated Utility](https://arxiv.org/abs/2501.11747v1) +**arXiv ID:** 2501.11747v1 + +**Abstract:** +> Large Language Models improve with increasing amounts of high-quality +training data. However, leveraging larger datasets requires balancing quality, +quantity, and diversity across sources. After evaluating nine baseline methods +under both compute- and data-constrained scenarios, we find token-count +heuristics outperform manual and learned mixes, indicating that simple +approaches accounting for dataset size and diversity are surprisingly +effective. Building on this insight, we propose two complementary approaches: +UtiliMax, which extends token-based heuristics by incorporating utility +estimates from reduced-scale ablations, achieving up to a 10.6x speedup over +manual baselines; and Model Estimated Data Utility (MEDU), which leverages LLMs +to estimate data utility from small samples, matching ablation-based +performance while reducing computational requirements by $\sim$200x. Together, +these approaches establish a new framework for automated, compute-efficient +data mixing that is robust across training regimes. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all mandatory criteria: 1) Focuses on practical LLM training optimization via data mixing; 2) Provides quantitative results (10.6x speedup, 200x compute reduction); 3) Compares with baselines. Also addresses methodology (criterion 4) and computational challenges (criterion 5), while introducing novel approaches (MEDU/UtiliMax) with implementable methods. + +--- + +## [Benchmarking Large Language Models via Random Variables](https://arxiv.org/abs/2501.11790v1) +**arXiv ID:** 2501.11790v1 + +**Abstract:** +> With the continuous advancement of large language models (LLMs) in +mathematical reasoning, evaluating their performance in this domain has become +a prominent research focus. Recent studies have raised concerns about the +reliability of current mathematical benchmarks, highlighting issues such as +simplistic design and potential data leakage. Therefore, creating a reliable +benchmark that effectively evaluates the genuine capabilities of LLMs in +mathematical reasoning remains a significant challenge. To address this, we +propose RV-Bench, a framework for Benchmarking LLMs via Random Variables in +mathematical reasoning. Specifically, the background content of a random +variable question (RV question) mirrors the original problem in existing +standard benchmarks, but the variable combinations are randomized into +different values. LLMs must fully understand the problem-solving process for +the original problem to correctly answer RV questions with various combinations +of variable values. As a result, the LLM's genuine capability in mathematical +reasoning is reflected by its accuracy on RV-Bench. Extensive experiments are +conducted with 29 representative LLMs across 900+ RV questions. A leaderboard +for RV-Bench ranks the genuine capability of these LLMs. Further analysis of +accuracy dropping indicates that current LLMs still struggle with complex +mathematical reasoning problems. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets mandatory criteria 2 (quantitative experiments with 29 LLMs and accuracy metrics) and 3 (leaderboard comparison with SOTA models). While criterion 1 applicability is marginal, the framework indirectly supports practical LLM evaluation. Criterion 4 is addressed through methodology description of RV-Bench. No rejection triggers apply. Inclusion policy favors acceptance. + +--- + +## [Fact-Preserved Personalized News Headline Generation](https://arxiv.org/abs/2501.11828v1) +**arXiv ID:** 2501.11828v1 + +**Abstract:** +> Personalized news headline generation, aiming at generating user-specific +headlines based on readers' preferences, burgeons a recent flourishing research +direction. Existing studies generally inject a user interest embedding into an +encoderdecoder headline generator to make the output personalized, while the +factual consistency of headlines is inadequate to be verified. In this paper, +we propose a framework Fact-Preserved Personalized News Headline Generation +(short for FPG), to prompt a tradeoff between personalization and consistency. +In FPG, the similarity between the candidate news to be exposed and the +historical clicked news is used to give different levels of attention to key +facts in the candidate news, and the similarity scores help to learn a +fact-aware global user embedding. Besides, an additional training procedure +based on contrastive learning is devised to further enhance the factual +consistency of generated headlines. Extensive experiments conducted on a +real-world benchmark PENS validate the superiority of FPG, especially on the +tradeoff between personalization and factual consistency. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all mandatory criteria: 1) Practical applications in RAG through personalized headline generation, 2) Experimental validation with quantitative metrics on PENS benchmark, 3) Comparison with existing methods. It also meets optional criteria 4 (methodology details) and 5 (real-world challenges). No rejection triggers apply as it focuses on content generation rather than excluded domains. + +--- + +## [From Drafts to Answers: Unlocking LLM Potential via Aggregation + Fine-Tuning](https://arxiv.org/abs/2501.11877v1) +**arXiv ID:** 2501.11877v1 + +**Abstract:** +> Scaling data and model size has been proven effective for boosting the +performance of large language models. In addition to training-time scaling, +recent studies have revealed that increasing test-time computational resources +can further improve performance. In this work, we introduce Aggregation +Fine-Tuning (AFT), a supervised finetuning paradigm where the model learns to +synthesize multiple draft responses, referred to as proposals, into a single, +refined answer, termed aggregation. At inference time, a propose-and-aggregate +strategy further boosts performance by iteratively generating proposals and +aggregating them. Empirical evaluations on benchmark datasets show that +AFT-trained models substantially outperform standard SFT. Notably, an AFT +model, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3% LC +win rate on AlpacaEval 2, surpassing significantly larger LLMs such as +Llama3.1-405B-Instruct and GPT4. By combining sequential refinement and +parallel sampling, the propose-and-aggregate framework scales inference-time +computation in a flexible manner. Overall, These findings position AFT as a +promising approach to unlocking additional capabilities of LLMs without +resorting to increasing data volume or model size. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Practical application through performance-efficient LLM enhancement, 2) Quantitative results showing 41.3% LC win rate improvement, 3) Direct comparison with SOTA models like GPT-4. Also meets optional criteria 4 (methodology description) and demonstrates novelty through aggregation framework. No rejection triggers present. + +--- + +## [Panoramic Interests: Stylistic-Content Aware Personalized Headline + Generation](https://arxiv.org/abs/2501.11900v1) +**arXiv ID:** 2501.11900v1 + +**Abstract:** +> Personalized news headline generation aims to provide users with +attention-grabbing headlines that are tailored to their preferences. Prevailing +methods focus on user-oriented content preferences, but most of them overlook +the fact that diverse stylistic preferences are integral to users' panoramic +interests, leading to suboptimal personalization. In view of this, we propose a +novel Stylistic-Content Aware Personalized Headline Generation (SCAPE) +framework. SCAPE extracts both content and stylistic features from headlines +with the aid of large language model (LLM) collaboration. It further adaptively +integrates users' long- and short-term interests through a contrastive +learning-based hierarchical fusion network. By incorporating the panoramic +interests into the headline generator, SCAPE reflects users' stylistic-content +preferences during the generation process. Extensive experiments on the +real-world dataset PENS demonstrate the superiority of SCAPE over baselines. + +**Decision Explanation:** Original decision: ACCEPT +Meets ALL mandatory criteria: 1) Practical LLM application in personalized content generation, 2) Experimental validation with quantitative metrics on PENS dataset, 3) Comparison with baselines. Also addresses optional criteria 4 (methodology details) and 5 (real-world applications). No rejection triggers present. + +--- + +## [LuxVeri at GenAI Detection Task 1: Inverse Perplexity Weighted Ensemble + for Robust Detection of AI-Generated Text across English and Multilingual + Contexts](https://arxiv.org/abs/2501.11914v1) +**arXiv ID:** 2501.11914v1 + +**Abstract:** +> This paper presents a system developed for Task 1 of the COLING 2025 Workshop +on Detecting AI-Generated Content, focusing on the binary classification of +machine-generated versus human-written text. Our approach utilizes an ensemble +of models, with weights assigned according to each model's inverse perplexity, +to enhance classification accuracy. For the English text detection task, we +combined RoBERTa-base, RoBERTa-base with the OpenAI detector, and +BERT-base-cased, achieving a Macro F1-score of 0.7458, which ranked us 12th out +of 35 teams. We ensembled RemBERT, XLM-RoBERTa-base, and +BERT-base-multilingual-case for the multilingual text detection task, employing +the same inverse perplexity weighting technique. This resulted in a Macro +F1-score of 0.7513, positioning us 4th out of 25 teams. Our results demonstrate +the effectiveness of inverse perplexity weighting in improving the robustness +of machine-generated text detection across both monolingual and multilingual +settings, highlighting the potential of ensemble methods for this challenging +task. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets mandatory criteria 2 (quantitative metrics with F1-scores) and 3 (comparison via competition rankings). While practical applications (1) focus on AI detection rather than LLM integration with KG/RAG/agentic AI, it marginally qualifies as a real-world LLM application. It also meets optional criterion 4 (methodology details on ensemble techniques). No rejection triggers apply. + +--- + +## [LuxVeri at GenAI Detection Task 3: Cross-Domain Detection of + AI-Generated Text Using Inverse Perplexity-Weighted Ensemble of Fine-Tuned + Transformer Models](https://arxiv.org/abs/2501.11918v1) +**arXiv ID:** 2501.11918v1 + +**Abstract:** +> This paper presents our approach for Task 3 of the GenAI content detection +workshop at COLING-2025, focusing on Cross-Domain Machine-Generated Text (MGT) +Detection. We propose an ensemble of fine-tuned transformer models, enhanced by +inverse perplexity weighting, to improve classification accuracy across diverse +text domains. For Subtask A (Non-Adversarial MGT Detection), we combined a +fine-tuned RoBERTa-base model with an OpenAI detector-integrated RoBERTa-base +model, achieving an aggregate TPR score of 0.826, ranking 10th out of 23 +detectors. In Subtask B (Adversarial MGT Detection), our fine-tuned +RoBERTa-base model achieved a TPR score of 0.801, securing 8th out of 22 +detectors. Our results demonstrate the effectiveness of inverse +perplexity-based weighting for enhancing generalization and performance in both +non-adversarial and adversarial MGT detection, highlighting the potential for +transformer models in cross-domain AI-generated content detection. + +**Decision Explanation:** Original decision: ACCEPT +Meets ALL mandatory criteria: 1) Practical application in AI-generated text detection, 2) Quantitative results (TPR scores) with experimental validation, 3) Comparison via rankings against other detectors. Also meets criteria 4 (methodology details) and 5 (real-world detection challenges). No rejection triggers present. Approach shows implementable transformer-based methods with cross-domain applications. + +--- + +## [Webvs. LLMs: An Empirical Study of Learning Behaviors of CS2 Students](https://arxiv.org/abs/2501.11935v1) +**arXiv ID:** 2501.11935v1 + +**Abstract:** +> LLMs such as ChatGPT have been widely adopted by students in higher education +as tools for learning programming and related concepts. However, it remains +unclear how effective students are and what strategies students use while +learning with LLMs. Since the majority of students' experiences in online +self-learning have come through using search engines such as Google, evaluating +AI tools in this context can help us address these gaps. In this mixed methods +research, we conducted an exploratory within-subjects study to understand how +CS2 students learn programming concepts using both LLMs as well as traditional +online methods such as educational websites and videos to examine how students +approach learning within and across both scenarios. We discovered that students +found it easier to learn a more difficult concept using traditional methods +than using ChatGPT. We also found that students ask fewer follow-ups and use +more keyword-based queries for search engines while their prompts to LLMs tend +to explicitly ask for information. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Practical application of LLMs in education, 2) Empirical study with mixed methods results, 3) Comparison between LLMs and traditional search methods. Also addresses optional criteria 5 (real-world learning challenges). No rejection triggers apply. + +--- + +## [TAD-Bench: A Comprehensive Benchmark for Embedding-Based Text Anomaly + Detection](https://arxiv.org/abs/2501.11960v1) +**arXiv ID:** 2501.11960v1 + +**Abstract:** +> Text anomaly detection is crucial for identifying spam, misinformation, and +offensive language in natural language processing tasks. Despite the growing +adoption of embedding-based methods, their effectiveness and generalizability +across diverse application scenarios remain under-explored. To address this, we +present TAD-Bench, a comprehensive benchmark designed to systematically +evaluate embedding-based approaches for text anomaly detection. TAD-Bench +integrates multiple datasets spanning different domains, combining +state-of-the-art embeddings from large language models with a variety of +anomaly detection algorithms. Through extensive experiments, we analyze the +interplay between embeddings and detection methods, uncovering their strengths, +weaknesses, and applicability to different tasks. These findings offer new +perspectives on building more robust, efficient, and generalizable anomaly +detection systems for real-world applications. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all three mandatory criteria: 1) Focuses on practical applications of LLMs for anomaly detection in real-world scenarios, 2) Presents experimental results with quantitative metrics through benchmark creation, and 3) Compares state-of-the-art embeddings and detection methods. While mentioning social applications (spam/misinformation detection), the primary focus remains technical benchmarking rather than social impact analysis, avoiding rejection criteria. + +--- + +## [Bridging Visualization and Optimization: Multimodal Large Language + Models on Graph-Structured Combinatorial Optimization](https://arxiv.org/abs/2501.11968v1) +**arXiv ID:** 2501.11968v1 + +**Abstract:** +> Graph-structured combinatorial challenges are inherently difficult due to +their nonlinear and intricate nature, often rendering traditional computational +methods ineffective or expensive. However, these challenges can be more +naturally tackled by humans through visual representations that harness our +innate ability for spatial reasoning. In this study, we propose transforming +graphs into images to preserve their higher-order structural features +accurately, revolutionizing the representation used in solving graph-structured +combinatorial tasks. This approach allows machines to emulate human-like +processing in addressing complex combinatorial challenges. By combining the +innovative paradigm powered by multimodal large language models (MLLMs) with +simple search techniques, we aim to develop a novel and effective framework for +tackling such problems. Our investigation into MLLMs spanned a variety of +graph-based tasks, from combinatorial problems like influence maximization to +sequential decision-making in network dismantling, as well as addressing six +fundamental graph-related issues. Our findings demonstrate that MLLMs exhibit +exceptional spatial intelligence and a distinctive capability for handling +these problems, significantly advancing the potential for machines to +comprehend and analyze graph-structured data with a depth and intuition akin to +human cognition. These results also imply that integrating MLLMs with simple +optimization strategies could form a novel and efficient approach for +navigating graph-structured combinatorial challenges without complex +derivations, computationally demanding training and fine-tuning. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all mandatory criteria: 1) focuses on practical LLM applications in graph-structured optimization, 2) includes experimental validation across multiple tasks with claims of significant advancements, and 3) implicitly compares against traditional methods. It also satisfies optional criteria 4 (methodology for graph-to-image transformation) and 5 (real-world applications like network dismantling). While SOTA comparison isn't explicit, inclusiveness favors acceptance given spatial reasoning innovations and implementable multimodal LLM approach. + +--- + +## [Leveraging Graph Structures and Large Language Models for End-to-End + Synthetic Task-Oriented Dialogues](https://arxiv.org/abs/2501.11977v1) +**arXiv ID:** 2501.11977v1 + +**Abstract:** +> Training task-oriented dialogue systems is both costly and time-consuming, +due to the need for high-quality datasets encompassing diverse intents. +Traditional methods depend on extensive human annotation, while recent +advancements leverage large language models (LLMs) to generate synthetic data. +However, these approaches often require custom prompts or code, limiting +accessibility for non-technical users. We introduce GraphTOD, an end-to-end +framework that simplifies the generation of task-oriented dialogues. Users can +create dialogues by specifying transition graphs in JSON format. Our evaluation +demonstrates that GraphTOD generates high-quality dialogues across various +domains, significantly lowering the cost and complexity of dataset creation. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (practical LLM application in dialogue generation), 2 (quantitative evaluation showing cost reduction), and 4 (clear methodology using JSON transition graphs). No rejection triggers present. Novel approach combining graphs/LLMs for synthetic data generation aligns with selection priorities. + +--- + +## [Harnessing Generative Pre-Trained Transformer for Datacenter Packet + Trace Generation](https://arxiv.org/abs/2501.12033v1) +**arXiv ID:** 2501.12033v1 + +**Abstract:** +> Today, the rapid growth of applications reliant on datacenters calls for new +advancements to meet the increasing traffic and computational demands. Traffic +traces from datacenters are essential for further development and optimization +of future datacenters. However, traces are rarely released to the public. +Researchers often use simplified mathematical models that lack the depth needed +to recreate intricate traffic patterns and, thus, miss optimization +opportunities found in realistic traffic. In this preliminary work, we +introduce DTG-GPT, a packet-level Datacenter Traffic Generator (DTG), based on +the generative pre-trained transformer (GPT) architecture used by many +state-of-the-art large language models. We train our model on a small set of +available traffic traces from different domains and offer a simple methodology +to evaluate the fidelity of the generated traces to their original +counterparts. We show that DTG-GPT can synthesize novel traces that mimic the +spatiotemporal patterns found in real traffic traces. We further demonstrate +that DTG-GPT can generate traces for networks of different scales while +maintaining fidelity. Our findings indicate the potential that, in the future, +similar models to DTG-GPT will allow datacenter operators to release traffic +information to the research community via trained GPT models. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets criteria 1 (practical LLM application in datacenter optimization), 2 (experimental validation of trace fidelity), and 3 (comparison with simplified mathematical models as baseline). It also addresses criterion 5 (real-world challenge of data scarcity in datacenters). No rejection triggers apply. While methodology details are limited (preliminary work), it satisfies inclusion thresholds. + +--- + +## [Condor: Enhance LLM Alignment with Knowledge-Driven Data Synthesis and + Refinement](https://arxiv.org/abs/2501.12273v1) +**arXiv ID:** 2501.12273v1 + +**Abstract:** +> The quality of Supervised Fine-Tuning (SFT) data plays a critical role in +enhancing the conversational capabilities of Large Language Models (LLMs). +However, as LLMs become more advanced, the availability of high-quality +human-annotated SFT data has become a significant bottleneck, necessitating a +greater reliance on synthetic training data. In this work, we introduce Condor, +a novel two-stage synthetic data generation framework that incorporates World +Knowledge Tree and Self-Reflection Refinement to produce high-quality SFT data +at scale. Our experimental results demonstrate that a base model fine-tuned on +only 20K Condor-generated samples achieves superior performance compared to +counterparts. The additional refinement stage in Condor further enables +iterative self-improvement for LLMs at various scales (up to 72B), validating +the effectiveness of our approach. Furthermore, our investigation into the +scaling for synthetic data in post-training reveals substantial unexplored +potential for performance improvements, opening promising avenues for future +research. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Practical application in LLM alignment through synthetic data generation, 2) Quantitative results showing performance improvements with 20K samples and scaling to 72B models, 3) Implicit comparison with counterparts. Also meets secondary criteria 4 (methodology description) and 5 (addresses data scarcity challenge). No rejection triggers present. + +--- + +## [RALAD: Bridging the Real-to-Sim Domain Gap in Autonomous Driving with + Retrieval-Augmented Learning](https://arxiv.org/abs/2501.12296v1) +**arXiv ID:** 2501.12296v1 + +**Abstract:** +> In the pursuit of robust autonomous driving systems, models trained on +real-world datasets often struggle to adapt to new environments, particularly +when confronted with corner cases such as extreme weather conditions. +Collecting these corner cases in the real world is non-trivial, which +necessitates the use of simulators for validation. However,the high +computational cost and the domain gap in data distribution have hindered the +seamless transition between real and simulated driving scenarios. To tackle +this challenge, we propose Retrieval-Augmented Learning for Autonomous Driving +(RALAD), a novel framework designed to bridge the real-to-sim gap at a low +cost. RALAD features three primary designs, including (1) domain adaptation via +an enhanced Optimal Transport (OT) method that accounts for both individual and +grouped image distances, (2) a simple and unified framework that can be applied +to various models, and (3) efficient fine-tuning techniques that freeze the +computationally expensive layers while maintaining robustness. Experimental +results demonstrate that RALAD compensates for the performance degradation in +simulated environments while maintaining accuracy in real-world scenarios +across three different models. Taking Cross View as an example, the mIOU and +mAP metrics in real-world scenarios remain stable before and after RALAD +fine-tuning, while in simulated environments,the mIOU and mAP metrics are +improved by 10.30% and 12.29%, respectively. Moreover, the re-training cost of +our approach is reduced by approximately 88.1%. Our code is available at +https://github.com/JiachengZuo/RALAD.git. + +**Decision Explanation:** Original decision: ACCEPT +Meets mandatory criteria 1 (practical applications in autonomous driving/retrieval-augmented learning), 2 (quantitative metrics like 10.30% mIOU improvement), and 3 (implied SOTA comparison through efficiency claims). Also addresses real-world challenges (criteria 5) and demonstrates reproducibility with code. Autonomous driving focus does not trigger rejection categories. + +--- + +## [LLM-Assisted Knowledge Graph Completion for Curriculum and Domain + Modelling in Personalized Higher Education Recommendations](https://arxiv.org/abs/2501.12300v1) +**arXiv ID:** 2501.12300v1 + +**Abstract:** +> While learning personalization offers great potential for learners, modern +practices in higher education require a deeper consideration of domain models +and learning contexts, to develop effective personalization algorithms. This +paper introduces an innovative approach to higher education curriculum +modelling that utilizes large language models (LLMs) for knowledge graph (KG) +completion, with the goal of creating personalized learning-path +recommendations. Our research focuses on modelling university subjects and +linking their topics to corresponding domain models, enabling the integration +of learning modules from different faculties and institutions in the student's +learning path. Central to our approach is a collaborative process, where LLMs +assist human experts in extracting high-quality, fine-grained topics from +lecture materials. We develop a domain, curriculum, and user models for +university modules and stakeholders. We implement this model to create the KG +from two study modules: Embedded Systems and Development of Embedded Systems +Using FPGA. The resulting KG structures the curriculum and links it to the +domain models. We evaluate our approach through qualitative expert feedback and +quantitative graph quality metrics. Domain experts validated the relevance and +accuracy of the model, while the graph quality metrics measured the structural +properties of our KG. Our results show that the LLM-assisted graph completion +approach enhances the ability to connect related courses across disciplines to +personalize the learning experience. Expert feedback also showed high +acceptance of the proposed collaborative approach for concept extraction and +classification. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (LLM/KG practical application in education), 2 (quantitative metrics and expert validation), and 4 (detailed methodology). Addresses novelty through LLM-assisted KG completion. Does not trigger rejection categories. Satisfies 3 primary criteria with implementable approach. + +--- + +## [Treefix: Enabling Execution with a Tree of Prefixes](https://arxiv.org/abs/2501.12339v1) +**arXiv ID:** 2501.12339v1 + +**Abstract:** +> The ability to execute code is a prerequisite for various dynamic program +analyses. Learning-guided execution has been proposed as an approach to enable +the execution of arbitrary code snippets by letting a neural model predict +likely values for any missing variables. Although state-of-the-art +learning-guided execution approaches, such as LExecutor, can enable the +execution of a relative high amount of code, they are limited to predicting a +restricted set of possible values and do not use any feedback from previous +executions to execute even more code. This paper presents Treefix, a novel +learning-guided execution approach that leverages LLMs to iteratively create +code prefixes that enable the execution of a given code snippet. The approach +addresses the problem in a multi-step fashion, where each step uses feedback +about the code snippet and its execution to instruct an LLM to improve a +previously generated prefix. This process iteratively creates a tree of +prefixes, a subset of which is returned to the user as prefixes that maximize +the number of executed lines in the code snippet. In our experiments with two +datasets of Python code snippets, Treefix achieves 25% and 7% more coverage +relative to the current state of the art in learning-guided execution, covering +a total of 84% and 82% of all lines in the code snippets. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Practical application in code execution using LLMs, 2) Quantitative metrics showing 25% and 7% improvements, 3) Comparison with LExecutor as state-of-the-art. Also addresses methodology (criterion 4) through iterative prefix generation and real-world code analysis applications (criterion 5). No rejection triggers present. + +--- + +## [Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for + Mixture-of-Experts Language Models](https://arxiv.org/abs/2501.12370v1) +**arXiv ID:** 2501.12370v1 + +**Abstract:** +> Scaling the capacity of language models has consistently proven to be a +reliable approach for improving performance and unlocking new capabilities. +Capacity can be primarily defined by two dimensions: the number of model +parameters and the compute per example. While scaling typically involves +increasing both, the precise interplay between these factors and their combined +contribution to overall capacity remains not fully understood. We explore this +relationship in the context of sparse Mixture-of-Expert models (MoEs), which +allow scaling the number of parameters without proportionally increasing the +FLOPs per example. We investigate how varying the sparsity level, i.e., the +ratio of non-active to total parameters, affects model performance in terms of +both pretraining and downstream performance. We find that under different +constraints (e.g. parameter size and total training compute), there is an +optimal level of sparsity that improves both training efficiency and model +performance. These results provide a better understanding of the impact of +sparsity in scaling laws for MoEs and complement existing works in this area, +offering insights for designing more efficient architectures. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets mandatory criteria 1 (practical applications in scaling LLMs via MoE architectures), 2 (experimental results with quantitative metrics on sparsity optimization), and 3 (comparison with existing scaling laws). It also addresses optional criterion 4 (methodology details for sparsity optimization) and demonstrates novelty in scaling law analysis for sparse MoEs. No rejection triggers apply. + +--- + +## [Is Long Context All You Need? Leveraging LLM's Extended Context for + NL2SQL](https://arxiv.org/abs/2501.12372v1) +**arXiv ID:** 2501.12372v1 + +**Abstract:** +> Large Language Models (LLMs) have demonstrated impressive capabilities across +a range of natural language processing tasks. In particular, improvements in +reasoning abilities and the expansion of context windows have opened new +avenues for leveraging these powerful models. NL2SQL is challenging in that the +natural language question is inherently ambiguous, while the SQL generation +requires a precise understanding of complex data schema and semantics. One +approach to this semantic ambiguous problem is to provide more and sufficient +contextual information. + In this work, we explore the performance and the latency trade-offs of the +extended context window (a.k.a., long context) offered by Google's +state-of-the-art LLM (\textit{gemini-1.5-pro}). We study the impact of various +contextual information, including column example values, question and SQL query +pairs, user-provided hints, SQL documentation, and schema. To the best of our +knowledge, this is the first work to study how the extended context window and +extra contextual information can help NL2SQL generation with respect to both +accuracy and latency cost. We show that long context LLMs are robust and do not +get lost in the extended contextual information. Additionally, our long-context +NL2SQL pipeline based on Google's \textit{gemini-pro-1.5} achieve a strong +performance with 67.41\% on BIRD benchmark (dev) without finetuning and +expensive self-consistency based techniques. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all three mandatory criteria: 1) Focuses on practical NL2SQL applications using LLMs' extended context, 2) Reports quantitative results (67.41% on BIRD benchmark) and latency analysis, 3) Compares with SOTA techniques and claims first work status. Also addresses methodology (C4) and implements with current tools. No exclusion triggers apply. + +--- + +## [Code Readability in the Age of Large Language Models: An Industrial Case + Study from Atlassian](https://arxiv.org/abs/2501.11264v1) +**arXiv ID:** 2501.11264v1 + +**Abstract:** +> Programmers spend a significant amount of time reading code during the +software development process. This trend is amplified by the emergence of large +language models (LLMs) that automatically generate code. However, little is +known about the readability of the LLM-generated code and whether it is still +important from practitioners' perspectives in this new era. In this paper, we +conduct a survey to explore the practitioners' perspectives on code readability +in the age of LLMs and investigate the readability of our LLM-based software +development agents framework, HULA, by comparing its generated code with +human-written code in real-world scenarios. Overall, the findings underscore +that (1) readability remains a critical aspect of software development; (2) the +readability of our LLM-generated code is comparable to human-written code, +fostering the establishment of appropriate trust and driving the broad adoption +of our LLM-powered software development platform. + +**Decision Explanation:** Original decision: ACCEPT +Meets ALL mandatory criteria: 1 (practical LLM application in software development), 2 (experimental comparison of LLM-generated vs human code), 3 (comparison with human-written code as baseline). Also meets optional criteria 4/5 (methodology described for HULA framework and real-world case study). No rejection triggers present. + +--- + +## [RedStar: Does Scaling Long-CoT Data Unlock Better Slow-Reasoning + Systems?](https://arxiv.org/abs/2501.11284v1) +**arXiv ID:** 2501.11284v1 + +**Abstract:** +> Can scaling transform reasoning? In this work, we explore the untapped +potential of scaling Long Chain-of-Thought (Long-CoT) data to 1000k samples, +pioneering the development of a slow-thinking model, RedStar. Through extensive +experiments with various LLMs and different sizes, we uncover the ingredients +for specialization and scale for Long-CoT training. Surprisingly, even smaller +models show significant performance gains with limited data, revealing the +sample efficiency of Long-CoT and the critical role of sample difficulty in the +learning process. Our findings demonstrate that Long-CoT reasoning can be +effectively triggered with just a few thousand examples, while larger models +achieve unparalleled improvements. We also introduce reinforcement learning +(RL)-scale training as a promising direction for advancing slow-thinking +systems. RedStar shines across domains: on the MATH-Hard benchmark, +RedStar-code-math boosts performance from 66.2\% to 81.6\%, and on the USA Math +Olympiad (AIME), it solves 46.7\% of problems using only 21k mixed-code-math +datasets. In multimodal tasks like GeoQA and MathVista-GEO, RedStar-Geo +achieves competitive results with minimal Long-CoT data, outperforming other +slow-thinking systems like QvQ-Preview. Compared to QwQ, RedStar strikes the +perfect balance between reasoning and generalizability. Our work highlights +that, with careful tuning, scaling Long-CoT can unlock extraordinary reasoning +capabilities-even with limited dataset and set a new standard for slow-thinking +models across diverse challenges. Our data and models are released at +https://huggingface.co/RedStar-Reasoning. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Focuses on practical reasoning improvements through scaled Long-CoT training, 2) Provides quantitative results (e.g., 66.2%→81.6% on MATH-Hard), 3) Compares with SOTA systems like QvQ-Preview. Also satisfies optional criteria 4 (methodology details) and 5 (real-world math/multimodal applications). Avoids rejection categories while demonstrating novelty in slow-reasoning systems. + +--- + +## [Graph-defined Language Learning with LLMs](https://arxiv.org/abs/2501.11478v1) +**arXiv ID:** 2501.11478v1 + +**Abstract:** +> Recent efforts leverage Large Language Models (LLMs) for modeling +text-attributed graph structures in node classification tasks. These approaches +describe graph structures for LLMs to understand or aggregate LLM-generated +textual attribute embeddings through graph structure. However, these approaches +face two main limitations in modeling graph structures with LLMs. (i) Graph +descriptions become verbose in describing high-order graph structure. (ii) +Textual attributes alone do not contain adequate graph structure information. +It is challenging to model graph structure concisely and adequately with LLMs. +LLMs lack built-in mechanisms to model graph structures directly. They also +struggle with complex long-range dependencies between high-order nodes and +target nodes. + Inspired by the observation that LLMs pre-trained on one language can achieve +exceptional performance on another with minimal additional training, we propose +\textbf{G}raph-\textbf{D}efined \textbf{L}anguage for \textbf{L}arge +\textbf{L}anguage \textbf{M}odel (GDL4LLM). This novel framework enables LLMs +to transfer their powerful language understanding capabilities to +graph-structured data. GDL4LLM translates graphs into a graph language corpus +instead of graph descriptions and pre-trains LLMs on this corpus to adequately +understand graph structures. During fine-tuning, this corpus describes the +structural information of target nodes concisely with only a few tokens. By +treating graphs as a new language, GDL4LLM enables LLMs to model graph +structures adequately and concisely for node classification tasks. Extensive +experiments on three real-world datasets demonstrate that GDL4LLM outperforms +description-based and textual attribute embeddings-based baselines by +efficiently modeling different orders of graph structure with LLMs. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Focuses on practical LLM applications with knowledge graphs, 2) Includes experimental results with real datasets and quantitative metrics, 3) Compares with SOTA baselines. Also meets optional criterion 4 through methodology description of graph-language translation. Novel approach (GDL4LLM) aligns with additional considerations. No rejection triggers present. + +--- + +## [Early evidence of how LLMs outperform traditional systems on OCR/HTR + tasks for historical records](https://arxiv.org/abs/2501.11623v1) +**arXiv ID:** 2501.11623v1 + +**Abstract:** +> We explore the ability of two LLMs -- GPT-4o and Claude Sonnet 3.5 -- to +transcribe historical handwritten documents in a tabular format and compare +their performance to traditional OCR/HTR systems: EasyOCR, Keras, Pytesseract, +and TrOCR. Considering the tabular form of the data, two types of experiments +are executed: one where the images are split line by line and the other where +the entire scan is used as input. Based on CER and BLEU, we demonstrate that +LLMs outperform the conventional OCR/HTR methods. Moreover, we also compare the +evaluated CER and BLEU scores to human evaluations to better judge the outputs +of whole-scan experiments and understand influential factors for CER and BLEU. +Combining judgments from all the evaluation metrics, we conclude that two-shot +GPT-4o for line-by-line images and two-shot Claude Sonnet 3.5 for whole-scan +images yield the transcriptions of the historical records most similar to the +ground truth. + +**Decision Explanation:** Original decision: ACCEPT +Meets ALL mandatory criteria: 1) Practical OCR application using LLMs, 2) Quantitative metrics (CER/BLEU) and comparisons, 3) Direct comparison with traditional OCR systems. Also meets optional criterion 4 through methodology description. No rejection triggers apply. + +--- + +## [Policy-Adaptable Methods For Resolving Normative Conflicts Through + Argumentation and Graph Colouring](https://arxiv.org/abs/2501.11799v1) +**arXiv ID:** 2501.11799v1 + +**Abstract:** +> In a multi-agent system, one may choose to govern the behaviour of an agent +by imposing norms, which act as guidelines for how agents should act either all +of the time or in given situations. However, imposing multiple norms on one or +more agents may result in situations where these norms conflict over how the +agent should behave. In any system with normative conflicts (such as safe +reinforcement models or systems which monitor safety protocols), one must +decide which norms should be followed such that the most important and most +relevant norms are maintained. We introduce a new method for resolving +normative conflicts through argumentation and graph colouring which is +compatible with a variety of normative conflict resolution policies. We prove +that this method always creates an admissible set of arguments under +argumentation semantics, meaning that it produces coherent outputs. We also +introduce more robust variants of this method, each building upon their +predecessor to create a superior output, and we include further mathematical +proof of their coherence. Our most advanced variant uses the existing concept +of curtailment, where one norm may supersede another without fully eliminating +it. The methods we introduce are all compatible with various pre-existing +policies for resolving normative conflicts. Empirical evaluations are also +performed to compare our algorithms to each other and to others in existing +literature. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all mandatory criteria: 1) Addresses practical applications in agentic AI (normative conflict resolution for multi-agent systems), 2) Includes empirical evaluations and mathematical proofs as quantitative metrics, 3) Compares with existing literature. It also addresses optional criteria 4 (methodology details via argumentation/graph coloring variants) and 5 (real-world applications like safety protocols). While focused on policy/conflict resolution, it does not fall under explicit rejection categories like medical/social/law applications. + +--- + +## [Network-informed Prompt Engineering against Organized Astroturf + Campaigns under Extreme Class Imbalance](https://arxiv.org/abs/2501.11849v1) +**arXiv ID:** 2501.11849v1 + +**Abstract:** +> Detecting organized political campaigns is of paramount importance in +fighting against disinformation on social media. Existing approaches for the +identification of such organized actions employ techniques mostly from network +science, graph machine learning and natural language processing. Their ultimate +goal is to analyze the relationships and interactions (e.g. re-posting) among +users and the textual similarities of their posts. Despite their effectiveness +in recognizing astroturf campaigns, these methods face significant challenges, +notably the class imbalance in available training datasets. To mitigate this +issue, recent methods usually resort to data augmentation or increasing the +number of positive samples, which may not always be feasible or sufficient in +real-world settings. Following a different path, in this paper, we propose a +novel framework for identifying astroturf campaigns based solely on large +language models (LLMs), introducing a Balanced Retrieval-Augmented Generation +(Balanced RAG) component. Our approach first gives both textual information +concerning the posts (in our case tweets) and the user interactions of the +social network as input to a language model. Then, through prompt engineering +and the proposed Balanced RAG method, it effectively detects coordinated +disinformation campaigns on X (Twitter). The proposed framework does not +require any training or fine-tuning of the language model. Instead, by +strategically harnessing the strengths of prompt engineering and Balanced RAG, +it facilitates LLMs to overcome the effects of class imbalance and effectively +identify coordinated political campaigns. The experimental results demonstrate +that by incorporating the proposed prompt engineering and Balanced RAG methods, +our framework outperforms the traditional graph-based baselines, achieving +2x-3x improvements in terms of precision, recall and F1 scores. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets mandatory criteria 1 (Practical Applications in RAG/LLMs), 2 (Quantitative metrics showing 2x-3x improvements), and 3 (Comparison with graph-based baselines). It also satisfies optional criteria 4 (Methodology details for LLM-based detection) and 5 (real-world class imbalance challenge). While addressing disinformation campaigns, its primary technical focus on LLM/RAG/prompt engineering outweighs social application concerns under rejection criteria. + +--- + +## [EDoRA: Efficient Weight-Decomposed Low-Rank Adaptation via Singular + Value Decomposition](https://arxiv.org/abs/2501.12067v1) +**arXiv ID:** 2501.12067v1 + +**Abstract:** +> Parameter-efficient fine-tuning methods, such as LoRA, reduces the number of +trainable parameters. However, they often suffer from scalability issues and +differences between their learning pattern and full fine-tuning. To overcome +these limitations, we propose Efficient Weight-Decomposed Low-Rank Adaptation +(EDoRA): a novel PEFT method that decomposes pre-trained weights into magnitude +and directional components. By freezing low-rank matrices, initializing them by +singular value decomposition, and introducing a small trainable matrix between +them, EDoRA achieves substantial reduction in trainable parameters while +maintaining learning capacity. Experimental results on the GLUE benchmark +demonstrate that EDoRA achieves competitive or superior performance compared to +state-of-the-art methods, such as LoRA and DoRA, with up to 30x fewer trainable +parameters. This makes EDoRA a highly efficient solution for adapting LLMs to +diverse tasks under memory-constrained settings. Code is available at +https://github.com/Hamid-Nasiri/EDoRA . + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all three mandatory criteria: 1) Focuses on practical LLM adaptation via parameter-efficient fine-tuning, 2) Provides GLUE benchmark results with quantitative metrics (30x parameter reduction), and 3) Compares with SOTA methods like LoRA/DoRA. Also meets optional criteria 4 (clear decomposition methodology) and 5 (addresses memory constraints). No rejection triggers apply. + +--- + +## [Improving Influence-based Instruction Tuning Data Selection for Balanced + Learning of Diverse Capabilities](https://arxiv.org/abs/2501.12147v1) +**arXiv ID:** 2501.12147v1 + +**Abstract:** +> Selecting appropriate training data is crucial for effective instruction +fine-tuning of large language models (LLMs), which aims to (1) elicit strong +capabilities, and (2) achieve balanced performance across a diverse range of +tasks. Influence-based methods show promise in achieving (1) by estimating the +contribution of each training example to the model's predictions, but often +struggle with (2). Our systematic investigation reveals that this +underperformance can be attributed to an inherent bias where certain tasks +intrinsically have greater influence than others. As a result, data selection +is often biased towards these tasks, not only hurting the model's performance +on others but also, counterintuitively, harms performance on these +high-influence tasks themselves. + As a remedy, we propose BIDS, a Balanced and Influential Data Selection +algorithm. BIDS first normalizes influence scores of the training data, and +then iteratively balances data selection by choosing the training example with +the highest influence on the most underrepresented task. Experiments with both +Llama-3 and Mistral-v0.3 on seven benchmarks spanning five diverse capabilities +show that BIDS consistently outperforms both state-of-the-art influence-based +algorithms and other non-influence-based selection frameworks. Surprisingly, +training on a 15% subset selected by BIDS can even outperform full-dataset +training with a much more balanced performance. Our analysis further highlights +the importance of both instance-level normalization and iterative optimization +of selected data for balanced learning of diverse capabilities. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Focuses on practical LLM training optimization through data selection, 2) Includes experiments with Llama-3/Mistral and quantitative metrics showing 15% subset outperforms full training, 3) Compares with SOTA influence-based methods. Also meets optional criteria 4 (detailed BIDS methodology) and demonstrates novelty in balanced data selection. No rejection triggers present. + +--- + +## [On the practical applicability of modern DFT functionals for chemical + computations. Case study of DM21 applicability for geometry optimization](https://arxiv.org/abs/2501.12149v1) +**arXiv ID:** 2501.12149v1 + +**Abstract:** +> Density functional theory (DFT) is probably the most promising approach for +quantum chemistry calculations considering its good balance between +calculations precision and speed. In recent years, several neural network-based +functionals have been developed for exchange-correlation energy approximation +in DFT, DM21 developed by Google Deepmind being the most notable between them. +This study focuses on evaluating the efficiency of DM21 functional in +predicting molecular geometries, with a focus on the influence of oscillatory +behavior in neural network exchange-correlation functionals. We implemented +geometry optimization in PySCF for the DM21 functional in geometry optimization +problem, compared its performance with traditional functionals, and tested it +on various benchmarks. Our findings reveal both the potential and the current +challenges of using neural network functionals for geometry optimization in +DFT. We propose a solution extending the practical applicability of such +functionals and allowing to model new substances with their help. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Focuses on practical chemical computation applications of neural network functionals (DM21), 2) Includes experimental comparisons and benchmarks, 3) Compares with traditional functionals. Also addresses optional criterion 4 (methodology details in PySCF implementation) and 5 (real-world substance modeling challenges). No rejection triggers present. + +--- + +## [AdaServe: SLO-Customized LLM Serving with Fine-Grained Speculative + Decoding](https://arxiv.org/abs/2501.12162v1) +**arXiv ID:** 2501.12162v1 + +**Abstract:** +> This paper introduces AdaServe, the first LLM serving system to support SLO +customization through fine-grained speculative decoding. AdaServe leverages the +logits of a draft model to predict the speculative accuracy of tokens and +employs a theoretically optimal algorithm to construct token trees for +verification. To accommodate diverse SLO requirements without compromising +throughput, AdaServe employs a speculation-and-selection scheme that first +constructs candidate token trees for each request and then dynamically selects +tokens to meet individual SLO constraints while optimizing throughput. +Comprehensive evaluations demonstrate that AdaServe achieves up to 73% higher +SLO attainment and 74% higher goodput compared to state-of-the-art systems. +These results underscore AdaServe's potential to enhance the efficiency and +adaptability of LLM deployments across varied application scenarios. + +**Decision Explanation:** Original decision: ACCEPT +Meets all 3 mandatory criteria: 1) Practical applications in LLM deployment optimization, 2) Quantitative results showing 73% SLO improvement, 3) Comparison with SOTA systems. Also addresses methodology (criterion 4) and shows implementable approach. No rejection triggers present. + +--- + +## [InsTALL: Context-aware Instructional Task Assistance with Multi-modal + Large Language Models](https://arxiv.org/abs/2501.12231v1) +**arXiv ID:** 2501.12231v1 + +**Abstract:** +> The improved competence of generative models can help building multi-modal +virtual assistants that leverage modalities beyond language. By observing +humans performing multi-step tasks, one can build assistants that have +situational awareness of actions and tasks being performed, enabling them to +cater assistance based on this understanding. In this paper, we develop a +Context-aware Instructional Task Assistant with Multi-modal Large Language +Models (InsTALL) that leverages an online visual stream (e.g. a user's screen +share or video recording) and responds in real-time to user queries related to +the task at hand. To enable useful assistance, InsTALL 1) trains a multi-modal +model on task videos and paired textual data, and 2) automatically extracts +task graph from video data and leverages it at training and inference time. We +show InsTALL achieves state-of-the-art performance across proposed sub-tasks +considered for multimodal activity understanding -- task recognition (TR), +action recognition (AR), next action prediction (AP), and plan prediction (PP) +-- and outperforms existing baselines on two novel sub-tasks related to +automatic error identification. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: 1) Practical application in agentic AI via multi-modal task assistance, 2) Quantitative results showing SOTA performance across multiple sub-tasks, 3) Explicit comparison with existing baselines. Also satisfies optional criteria 4 (methodology details) and 5 (real-world application). No rejection triggers present. + +--- + +## [UI-TARS: Pioneering Automated GUI Interaction with Native Agents](https://arxiv.org/abs/2501.12326v1) +**arXiv ID:** 2501.12326v1 + +**Abstract:** +> This paper introduces UI-TARS, a native GUI agent model that solely perceives +the screenshots as input and performs human-like interactions (e.g., keyboard +and mouse operations). Unlike prevailing agent frameworks that depend on +heavily wrapped commercial models (e.g., GPT-4o) with expert-crafted prompts +and workflows, UI-TARS is an end-to-end model that outperforms these +sophisticated frameworks. Experiments demonstrate its superior performance: +UI-TARS achieves SOTA performance in 10+ GUI agent benchmarks evaluating +perception, grounding, and GUI task execution. Notably, in the OSWorld +benchmark, UI-TARS achieves scores of 24.6 with 50 steps and 22.7 with 15 +steps, outperforming Claude (22.0 and 14.9 respectively). In AndroidWorld, +UI-TARS achieves 46.6, surpassing GPT-4o (34.5). UI-TARS incorporates several +key innovations: (1) Enhanced Perception: leveraging a large-scale dataset of +GUI screenshots for context-aware understanding of UI elements and precise +captioning; (2) Unified Action Modeling, which standardizes actions into a +unified space across platforms and achieves precise grounding and interaction +through large-scale action traces; (3) System-2 Reasoning, which incorporates +deliberate reasoning into multi-step decision making, involving multiple +reasoning patterns such as task decomposition, reflection thinking, milestone +recognition, etc. (4) Iterative Training with Reflective Online Traces, which +addresses the data bottleneck by automatically collecting, filtering, and +reflectively refining new interaction traces on hundreds of virtual machines. +Through iterative training and reflection tuning, UI-TARS continuously learns +from its mistakes and adapts to unforeseen situations with minimal human +intervention. We also analyze the evolution path of GUI agents to guide the +further development of this domain. + +**Decision Explanation:** Original decision: ACCEPT +Meets ALL mandatory criteria: 1) Practical GUI interaction applications in agentic AI, 2) Quantitative results showing SOTA performance across multiple benchmarks, 3) Direct comparisons with Claude/GPT-4o. Also satisfies Criterion 4 with detailed methodology descriptions. Novel agentic AI implementation with reproducible training approach. No rejection triggers present. + +--- + +## [Automatic Labelling with Open-source LLMs using Dynamic Label Schema + Integration](https://arxiv.org/abs/2501.12332v1) +**arXiv ID:** 2501.12332v1 + +**Abstract:** +> Acquiring labelled training data remains a costly task in real world machine +learning projects to meet quantity and quality requirements. Recently Large +Language Models (LLMs), notably GPT-4, have shown great promises in labelling +data with high accuracy. However, privacy and cost concerns prevent the +ubiquitous use of GPT-4. In this work, we explore effectively leveraging +open-source models for automatic labelling. We identify integrating label +schema as a promising technology but found that naively using the label +description for classification leads to poor performance on high cardinality +tasks. To address this, we propose Retrieval Augmented Classification (RAC) for +which LLM performs inferences for one label at a time using corresponding label +schema; we start with the most related label and iterates until a label is +chosen by the LLM. We show that our method, which dynamically integrates label +description, leads to performance improvements in labelling tasks. We further +show that by focusing only on the most promising labels, RAC can trade off +between label quality and coverage - a property we leverage to automatically +label our internal datasets. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets criteria 1 (practical application in automatic labeling with LLMs), 2 (experimental results showing performance improvements), and 4 (methodology details on RAC). While explicit comparison with SOTA is unclear, the focus on dynamic label integration and trade-offs aligns with inclusivity guidelines. No rejection triggers apply. + +--- + +## [Expertise elevates AI usage: experimental evidence comparing laypeople + and professional artists](https://arxiv.org/abs/2501.12374v1) +**arXiv ID:** 2501.12374v1 + +**Abstract:** +> Novel capacities of generative AI to analyze and generate cultural artifacts +raise inevitable questions about the nature and value of artistic education and +human expertise. Has AI already leveled the playing field between professional +artists and laypeople, or do trained artistic expressive capacity, curation +skills and experience instead enhance the ability to use these new tools? In +this pre-registered study, we conduct experimental comparisons between 50 +active artists and a demographically matched sample of laypeople. We designed +two tasks to approximate artistic practice for testing their capabilities in +both faithful and creative image creation: replicating a reference image, and +moving as far away as possible from it. We developed a bespoke platform where +participants used a modern text-to-image model to complete both tasks. We also +collected and compared participants' sentiments towards AI. On average, artists +produced more faithful and creative outputs than their lay counterparts, +although only by a small margin. While AI may ease content creation, +professional expertise is still valuable - even within the confined space of +generative AI itself. Finally, we also explored how well an exemplary +vision-capable large language model (GPT-4o) would complete the same tasks, if +given the role of an image generation agent, and found it performed on par in +copying but outperformed even artists in the creative task. The very best +results were still produced by humans in both tasks. These outcomes highlight +the importance of integrating artistic skills with AI training to prepare +artists and other visual professionals for a technologically evolving +landscape. We see a potential in collaborative synergy with generative AI, +which could reshape creative industries and education in the arts. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (practical applications in creative industries/education), 2 (experimental comparisons with quantitative performance metrics), and 3 (comparison between human experts and GPT-4o). Also meets criterion 5 through discussion of real-world applications and AI integration challenges. While marginally relevant to social aspects, core focus aligns with LLM applications in agentic AI and human-AI collaboration. + +--- + +## [Conversation Routines: A Prompt Engineering Framework for Task-Oriented + Dialog Systems](https://arxiv.org/abs/2501.11613v1) +**arXiv ID:** 2501.11613v1 + +**Abstract:** +> This study introduces Conversation Routines (CR), a structured prompt +engineering framework for developing task-oriented dialog systems using Large +Language Models (LLMs). While LLMs demonstrate remarkable natural language +understanding capabilities, engineering them to reliably execute complex +business workflows remains challenging. The proposed CR framework enables the +development of Conversation Agentic Systems (CAS) through natural language +specifications, embedding task-oriented logic within LLM prompts. This approach +provides a systematic methodology for designing and implementing complex +conversational workflows while maintaining behavioral consistency. We +demonstrate the framework's effectiveness through two proof of concept +implementations: a Train Ticket Booking System and an Interactive +Troubleshooting Copilot. These case studies validate CR's capability to encode +sophisticated behavioral patterns and decision logic while preserving natural +conversational flexibility. Results show that CR enables domain experts to +design conversational workflows in natural language while leveraging custom +enterprise functionalities (tools) developed by software engineers, creating an +efficient division of responsibilities where developers focus on core API +implementation and domain experts handle conversation design. While the +framework shows promise in accessibility and adaptability, we identify key +challenges including computational overhead, non-deterministic behavior, and +domain-specific logic optimization. Future research directions include +enhancing system robustness, improving scalability for complex multi-agent +interactions, and addressing the identified limitations across diverse business +applications. + +**Decision Explanation:** Original decision: ACCEPT +The paper meets all three mandatory criteria: 1) Focuses on practical LLM applications (task-oriented dialog systems), 2) Includes experimental results with case studies (Train Ticket Booking and Troubleshooting Copilot), 3) Implicitly addresses advancements via workflow consistency and division of responsibilities. It also satisfies optional criteria 4 (methodology details) and 5 (real-world challenges). No rejection triggers apply. + +--- + +# Rejected Papers + +## [The Explanation Game -- Rekindled (Extended Version)](https://arxiv.org/abs/2501.11429v1) +**arXiv ID:** 2501.11429v1 + +**Abstract:** +> Recent work demonstrated the existence of critical flaws in the current use +of Shapley values in explainable AI (XAI), i.e. the so-called SHAP scores. +These flaws are significant in that the scores provided to a human +decision-maker can be misleading. Although these negative results might appear +to indicate that Shapley values ought not be used in XAI, this paper argues +otherwise. Concretely, this paper proposes a novel definition of SHAP scores +that overcomes existing flaws. Furthermore, the paper outlines a practically +efficient solution for the rigorous estimation of the novel SHAP scores. +Preliminary experimental results confirm our claims, and further underscore the +flaws of the current SHAP scores. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on explainable AI (XAI) and Shapley values, which relates to responsible AI/ethics. It does not explicitly address LLMs in practical applications (Criterion 1 unmet). While it meets Criteria 2 (experimental results) and 3 (comparison with SOTA), it violates the rejection rule for focusing on responsible AI applications. + +--- + +## [Bridging the Communication Gap: Evaluating AI Labeling Practices for + Trustworthy AI Development](https://arxiv.org/abs/2501.11909v1) +**arXiv ID:** 2501.11909v1 + +**Abstract:** +> As artificial intelligence (AI) becomes integral to economy and society, +communication gaps between developers, users, and stakeholders hinder trust and +informed decision-making. High-level AI labels, inspired by frameworks like EU +energy labels, have been proposed to make the properties of AI models more +transparent. Without requiring deep technical expertise, they can inform on the +trade-off between predictive performance and resource efficiency. However, the +practical benefits and limitations of AI labeling remain underexplored. This +study evaluates AI labeling through qualitative interviews along four key +research questions. Based on thematic analysis and inductive coding, we found a +broad range of practitioners to be interested in AI labeling (RQ1). They see +benefits for alleviating communication gaps and aiding non-expert +decision-makers, however limitations, misunderstandings, and suggestions for +improvement were also discussed (RQ2). Compared to other reporting formats, +interviewees positively evaluated the reduced complexity of labels, increasing +overall comprehensibility (RQ3). Trust was influenced most by usability and the +credibility of the responsible labeling authority, with mixed preferences for +self-certification versus third-party certification (RQ4). Our Insights +highlight that AI labels pose a trade-off between simplicity and complexity, +which could be resolved by developing customizable and interactive labeling +frameworks to address diverse user needs. Transparent labeling of resource +efficiency also nudged interviewee priorities towards paying more attention to +sustainability aspects during AI development. This study validates AI labels as +a valuable tool for enhancing trust and communication in AI, offering +actionable guidelines for their refinement and standardization. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on responsible AI applications (trustworthy AI development through labeling practices) and social communication aspects, which falls under the rejection criteria. It does not sufficiently address practical LLM applications, experimental results with quantitative metrics, or comparisons with state-of-the-art methods in required technical domains. + +--- + +## [Make Full Use of Testing Information: An Integrated Accelerated Testing + and Evaluation Method for Autonomous Driving Systems](https://arxiv.org/abs/2501.11924v1) +**arXiv ID:** 2501.11924v1 + +**Abstract:** +> Testing and evaluation is an important step before the large-scale +application of the autonomous driving systems (ADSs). Based on the three level +of scenario abstraction theory, a testing can be performed within a logical +scenario, followed by an evaluation stage which is inputted with the testing +results of each concrete scenario generated from the logical parameter space. +During the above process, abundant testing information is produced which is +beneficial for comprehensive and accurate evaluations. To make full use of +testing information, this paper proposes an Integrated accelerated Testing and +Evaluation Method (ITEM). Based on a Monte Carlo Tree Search (MCTS) paradigm +and a dual surrogates testing framework proposed in our previous work, this +paper applies the intermediate information (i.e., the tree structure, including +the affiliation of each historical sampled point with the subspaces and the +parent-child relationship between subspaces) generated during the testing stage +into the evaluation stage to achieve accurate hazardous domain identification. +Moreover, to better serve this purpose, the UCB calculation method is improved +to allow the search algorithm to focus more on the hazardous domain boundaries. +Further, a stopping condition is constructed based on the convergence of the +search algorithm. Ablation and comparative experiments are then conducted to +verify the effectiveness of the improvements and the superiority of the +proposed method. The experimental results show that ITEM could well identify +the hazardous domains in both low- and high-dimensional cases, regardless of +the shape of the hazardous domains, indicating its generality and potential for +the safety evaluation of ADSs. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on autonomous driving system testing using MCTS and surrogates but does not explicitly mention Large Language Models (LLMs) or their practical applications in areas like knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparisons), it fails the mandatory criterion 1 (LLM applications). Additionally, autonomous driving testing does not fall under the specified rejection categories but lacks required LLM relevance. + +--- + +## [Collaborative Imputation of Urban Time Series through Cross-city + Meta-learning](https://arxiv.org/abs/2501.11306v1) +**arXiv ID:** 2501.11306v1 + +**Abstract:** +> Urban time series, such as mobility flows, energy consumption, and pollution +records, encapsulate complex urban dynamics and structures. However, data +collection in each city is impeded by technical challenges such as budget +limitations and sensor failures, necessitating effective data imputation +techniques that can enhance data quality and reliability. Existing imputation +models, categorized into learning-based and analytics-based paradigms, grapple +with the trade-off between capacity and generalizability. Collaborative +learning to reconstruct data across multiple cities holds the promise of +breaking this trade-off. Nevertheless, urban data's inherent irregularity and +heterogeneity issues exacerbate challenges of knowledge sharing and +collaboration across cities. To address these limitations, we propose a novel +collaborative imputation paradigm leveraging meta-learned implicit neural +representations (INRs). INRs offer a continuous mapping from domain coordinates +to target values, integrating the strengths of both paradigms. By imposing +embedding theory, we first employ continuous parameterization to handle +irregularity and reconstruct the dynamical system. We then introduce a +cross-city collaborative learning scheme through model-agnostic meta learning, +incorporating hierarchical modulation and normalization techniques to +accommodate multiscale representations and reduce variance in response to +heterogeneity. Extensive experiments on a diverse urban dataset from 20 global +cities demonstrate our model's superior imputation performance and +generalizability, underscoring the effectiveness of collaborative imputation in +resource-constrained settings. + +**Decision Explanation:** Original decision: REJECT +The paper does not explicitly mention Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it addresses practical urban data imputation and demonstrates experimental results with comparisons to existing methods, it focuses on meta-learning and neural representations rather than LLM-centered approaches, failing to meet the mandatory Criterion 1. + +--- + +## [Finer-CAM: Spotting the Difference Reveals Finer Details for Visual + Explanation](https://arxiv.org/abs/2501.11309v1) +**arXiv ID:** 2501.11309v1 + +**Abstract:** +> Class activation map (CAM) has been widely used to highlight image regions +that contribute to class predictions. Despite its simplicity and computational +efficiency, CAM often struggles to identify discriminative regions that +distinguish visually similar fine-grained classes. Prior efforts address this +limitation by introducing more sophisticated explanation processes, but at the +cost of extra complexity. In this paper, we propose Finer-CAM, a method that +retains CAM's efficiency while achieving precise localization of discriminative +regions. Our key insight is that the deficiency of CAM lies not in "how" it +explains, but in "what" it explains}. Specifically, previous methods attempt to +identify all cues contributing to the target class's logit value, which +inadvertently also activates regions predictive of visually similar classes. By +explicitly comparing the target class with similar classes and spotting their +differences, Finer-CAM suppresses features shared with other classes and +emphasizes the unique, discriminative details of the target class. Finer-CAM is +easy to implement, compatible with various CAM methods, and can be extended to +multi-modal models for accurate localization of specific concepts. +Additionally, Finer-CAM allows adjustable comparison strength, enabling users +to selectively highlight coarse object contours or fine discriminative details. +Quantitatively, we show that masking out the top 5% of activated pixels by +Finer-CAM results in a larger relative confidence drop compared to baselines. +The source code and demo are available at +https://github.com/Imageomics/Finer-CAM. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on improving visual explanation methods (CAM) for image classification without mentioning LLMs or their applications in knowledge graphs, RAG, or agentic AI. It fails to meet mandatory criterion 1 (Practical Applications of LLMs) while meeting criteria 2-3. The methodology is vision-specific and unrelated to LLM-focused domains listed in the requirements. + +--- + +## [CatV2TON: Taming Diffusion Transformers for Vision-Based Virtual Try-On + with Temporal Concatenation](https://arxiv.org/abs/2501.11325v1) +**arXiv ID:** 2501.11325v1 + +**Abstract:** +> Virtual try-on (VTON) technology has gained attention due to its potential to +transform online retail by enabling realistic clothing visualization of images +and videos. However, most existing methods struggle to achieve high-quality +results across image and video try-on tasks, especially in long video +scenarios. In this work, we introduce CatV2TON, a simple and effective +vision-based virtual try-on (V2TON) method that supports both image and video +try-on tasks with a single diffusion transformer model. By temporally +concatenating garment and person inputs and training on a mix of image and +video datasets, CatV2TON achieves robust try-on performance across static and +dynamic settings. For efficient long-video generation, we propose an +overlapping clip-based inference strategy that uses sequential frame guidance +and Adaptive Clip Normalization (AdaCN) to maintain temporal consistency with +reduced resource demands. We also present ViViD-S, a refined video try-on +dataset, achieved by filtering back-facing frames and applying 3D mask +smoothing for enhanced temporal consistency. Comprehensive experiments +demonstrate that CatV2TON outperforms existing methods in both image and video +try-on tasks, offering a versatile and reliable solution for realistic virtual +try-ons across diverse scenarios. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing (virtual try-on for videos) and uses diffusion transformers rather than LLMs. It fails to meet mandatory criterion 1 (Practical Applications of LLMs) and falls under the video-processing rejection category. + +--- + +## [Towards Advancing Code Generation with Large Language Models: A Research + Roadmap](https://arxiv.org/abs/2501.11354v1) +**arXiv ID:** 2501.11354v1 + +**Abstract:** +> Recently, we have witnessed the rapid development of large language models, +which have demonstrated excellent capabilities in the downstream task of code +generation. However, despite their potential, LLM-based code generation still +faces numerous technical and evaluation challenges, particularly when embedded +in real-world development. In this paper, we present our vision for current +research directions, and provide an in-depth analysis of existing studies on +this task. We propose a six-layer vision framework that categorizes code +generation process into distinct phases, namely Input Phase, Orchestration +Phase, Development Phase, and Validation Phase. Additionally, we outline our +vision workflow, which reflects on the currently prevalent frameworks. We +systematically analyse the challenges faced by large language models, including +those LLM-based agent frameworks, in code generation tasks. With these, we +offer various perspectives and actionable recommendations in this area. Our aim +is to provide guidelines for improving the reliability, robustness and +usability of LLM-based code generation systems. Ultimately, this work seeks to +address persistent challenges and to provide practical suggestions for a more +pragmatic LLM-based solution for future code generation endeavors. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet all mandatory criteria: it lacks experimental results with quantitative metrics (Criterion 2) and direct comparison with state-of-the-art techniques (Criterion 3). While it addresses real-world challenges (Criterion 5), the mandatory requirements are not fully satisfied. + +--- + +## [Federated Learning with Sample-level Client Drift Mitigation](https://arxiv.org/abs/2501.11360v1) +**arXiv ID:** 2501.11360v1 + +**Abstract:** +> Federated Learning (FL) suffers from severe performance degradation due to +the data heterogeneity among clients. Existing works reveal that the +fundamental reason is that data heterogeneity can cause client drift where the +local model update deviates from the global one, and thus they usually tackle +this problem from the perspective of calibrating the obtained local update. +Despite effectiveness, existing methods substantially lack a deep understanding +of how heterogeneous data samples contribute to the formation of client drift. +In this paper, we bridge this gap by identifying that the drift can be viewed +as a cumulative manifestation of biases present in all local samples and the +bias between samples is different. Besides, the bias dynamically changes as the +FL training progresses. Motivated by this, we propose FedBSS that first +mitigates the heterogeneity issue in a sample-level manner, orthogonal to +existing methods. Specifically, the core idea of our method is to adopt a +bias-aware sample selection scheme that dynamically selects the samples from +small biases to large epoch by epoch to train progressively the local model in +each round. In order to ensure the stability of training, we set the +diversified knowledge acquisition stage as the warm-up stage to avoid the local +optimality caused by knowledge deviation in the early stage of the model. +Evaluation results show that FedBSS outperforms state-of-the-art baselines. In +addition, we also achieved effective results on feature distribution skew and +noise label dataset setting, which proves that FedBSS can not only reduce +heterogeneity, but also has scalability and robustness. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on Federated Learning improvements for data heterogeneity but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (SOTA comparison), it fails mandatory criterion 1 (LLM practical applications). + +--- + +## [Multi-View Spectral Clustering for Graphs with Multiple View Structures](https://arxiv.org/abs/2501.11422v1) +**arXiv ID:** 2501.11422v1 + +**Abstract:** +> Despite the fundamental importance of clustering, to this day, much of the +relevant research is still based on ambiguous foundations, leading to an +unclear understanding of whether or how the various clustering methods are +connected with each other. In this work, we provide an additional stepping +stone towards resolving such ambiguities by presenting a general clustering +framework that subsumes a series of seemingly disparate clustering methods, +including various methods belonging to the wildly popular spectral clustering +framework. In fact, the generality of the proposed framework is additionally +capable of shedding light to the largely unexplored area of multi-view graphs +whose each view may have differently clustered nodes. In turn, we propose +GenClus: a method that is simultaneously an instance of this framework and a +generalization of spectral clustering, while also being closely related to +k-means as well. This results in a principled alternative to the few existing +methods studying this special type of multi-view graphs. Then, we conduct +in-depth experiments, which demonstrate that GenClus is more computationally +efficient than existing methods, while also attaining similar or better +clustering performance. Lastly, a qualitative real-world case-study further +demonstrates the ability of GenClus to produce meaningful clusterings. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on general graph clustering methodologies and multi-view analysis without explicitly addressing Large Language Models (LLMs) or their applications in knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails criterion 1 (LLM practical applications). No connection to LLM-related tasks or autonomous AI scenarios is evident. + +--- + +## [A Survey on Diffusion Models for Anomaly Detection](https://arxiv.org/abs/2501.11430v1) +**arXiv ID:** 2501.11430v1 + +**Abstract:** +> Diffusion models (DMs) have emerged as a powerful class of generative AI +models, showing remarkable potential in anomaly detection (AD) tasks across +various domains, such as cybersecurity, fraud detection, healthcare, and +manufacturing. The intersection of these two fields, termed diffusion models +for anomaly detection (DMAD), offers promising solutions for identifying +deviations in increasingly complex and high-dimensional data. In this survey, +we systematically review recent advances in DMAD research and investigate their +capabilities. We begin by presenting the fundamental concepts of AD and DMs, +followed by a comprehensive analysis of classic DM architectures including +DDPMs, DDIMs, and Score SDEs. We further categorize existing DMAD methods into +reconstruction-based, density-based, and hybrid approaches, providing detailed +examinations of their methodological innovations. We also explore the diverse +tasks across different data modalities, encompassing image, time series, video, +and multimodal data analysis. Furthermore, we discuss critical challenges and +emerging research directions, including computational efficiency, model +interpretability, robustness enhancement, edge-cloud collaboration, and +integration with large language models. The collection of DMAD research papers +and resources is available at https://github.com/fdjingliu/DMAD. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on diffusion models for anomaly detection rather than LLMs in practical applications like knowledge graphs or agentic AI. It does not meet mandatory criteria 1 (Practical Applications of LLMs) and lacks experimental results or SOTA comparisons specific to LLMs. While it mentions integration with LLMs as a future direction, the primary focus is outside the required scope. + +--- + +## [Communication-Efficient Federated Learning Based on Explanation-Guided + Pruning for Remote Sensing Image Classification](https://arxiv.org/abs/2501.11493v1) +**arXiv ID:** 2501.11493v1 + +**Abstract:** +> Federated learning (FL) is a decentralized machine learning paradigm, where +multiple clients collaboratively train a global model by exchanging only model +updates with the central server without sharing the local data of clients. Due +to the large volume of model updates required to be transmitted between clients +and the central server, most FL systems are associated with high transfer costs +(i.e., communication overhead). This issue is more critical for operational +applications in remote sensing (RS), especially when large-scale RS data is +processed and analyzed through FL systems with restricted communication +bandwidth. To address this issue, we introduce an explanation-guided pruning +strategy for communication-efficient FL in the context of RS image +classification. Our pruning strategy is defined based on the layerwise +relevance propagation (LRP) driven explanations to: 1) efficiently and +effectively identify the most relevant and informative model parameters (to be +exchanged between clients and the central server); and 2) eliminate the +non-informative ones to minimize the volume of model updates. The experimental +results on the BigEarthNet-S2 dataset demonstrate that our strategy effectively +reduces the number of shared model updates, while increasing the generalization +ability of the global model. The code of this work will be publicly available +at https://git.tu-berlin.de/rsim/FL-LRP + +**Decision Explanation:** Original decision: REJECT +The paper focuses on communication-efficient federated learning for remote sensing image classification but does not address Large Language Models (LLMs) or their practical applications (mandatory Criterion 1 unmet). While it meets Criteria 2 (experimental results) and 3 (implicit SOTA comparison via performance claims), failure to satisfy ALL mandatory criteria necessitates rejection. + +--- + +## [Meta-Instance Selection. Instance Selection as a Classification Problem + with Meta-Features](https://arxiv.org/abs/2501.11526v1) +**arXiv ID:** 2501.11526v1 + +**Abstract:** +> Data pruning, or instance selection, is an important problem in machine +learning especially in terms of nearest neighbour classifier. However, in data +pruning which speeds up the prediction phase, there is an issue related to the +speed and efficiency of the process itself. In response, the study proposes an +approach involving transforming the instance selection process into a +classification task conducted in a unified meta-feature space where each +instance can be classified and assigned to either the "to keep" or "to remove" +class. This approach requires training an appropriate meta-classifier, which +can be developed based on historical instance selection results from other +datasets using reference instance selection methods as a labeling tool. This +work proposes constructing the meta-feature space based on properties extracted +from the nearest neighbor graph. Experiments conducted on 17 datasets of +varying sizes and five reference instance selection methods (ENN, Drop3, ICF, +HMN-EI, and CCIS) demonstrate that the proposed solution achieves results +comparable to reference instance selection methods while significantly reducing +computational complexity. In the proposed approach, the computational +complexity of the system depends only on identifying the k-nearest neighbors +for each data sample and running the meta-classifier. Additionally, the study +discusses the choice of meta-classifier, recommending the use of Balanced +Random Forest. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on instance selection methods for nearest neighbor classifiers but does not address Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it meets Criteria 2 (experimental results) and 3 (comparison with SOTA), it fails Criterion 1 (LLM practical applications). Mandatory requirements are not fully satisfied. + +--- + +## [The impact of intrinsic rewards on exploration in Reinforcement Learning](https://arxiv.org/abs/2501.11533v1) +**arXiv ID:** 2501.11533v1 + +**Abstract:** +> One of the open challenges in Reinforcement Learning is the hard exploration +problem in sparse reward environments. Various types of intrinsic rewards have +been proposed to address this challenge by pushing towards diversity. This +diversity might be imposed at different levels, favouring the agent to explore +different states, policies or behaviours (State, Policy and Skill level +diversity, respectively). However, the impact of diversity on the agent's +behaviour remains unclear. In this work, we aim to fill this gap by studying +the effect of different levels of diversity imposed by intrinsic rewards on the +exploration patterns of RL agents. We select four intrinsic rewards (State +Count, Intrinsic Curiosity Module (ICM), Maximum Entropy, and Diversity is all +you need (DIAYN)), each pushing for a different diversity level. We conduct an +empirical study on MiniGrid environment to compare their impact on exploration +considering various metrics related to the agent's exploration, namely: +episodic return, observation coverage, agent's position coverage, policy +entropy, and timeframes to reach the sparse reward. The main outcome of the +study is that State Count leads to the best exploration performance in the case +of low-dimensional observations. However, in the case of RGB observations, the +performance of State Count is highly degraded mostly due to representation +learning challenges. Conversely, Maximum Entropy is less impacted, resulting in +a more robust exploration, despite being not always optimal. Lastly, our +empirical study revealed that learning diverse skills with DIAYN, often linked +to improved robustness and generalisation, does not promote exploration in +MiniGrid environments. This is because: i) learning the skill space itself can +be challenging, and ii) exploration within the skill space prioritises +differentiating between behaviours rather than achieving uniform state +visitation. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on reinforcement learning and intrinsic rewards but does not address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI applications. It fails to meet the mandatory Criterion 1 (Practical Applications of LLMs). While it meets Criteria 2 and 3 with experimental results and SOTA comparisons, ALL mandatory criteria must be satisfied for acceptance. + +--- + +## [StAyaL | Multilingual Style Transfer](https://arxiv.org/abs/2501.11639v1) +**arXiv ID:** 2501.11639v1 + +**Abstract:** +> Stylistic text generation plays a vital role in enhancing communication by +reflecting the nuances of individual expression. This paper presents a novel +approach for generating text in a specific speaker's style across different +languages. We show that by leveraging only 100 lines of text, an individuals +unique style can be captured as a high-dimensional embedding, which can be used +for both text generation and stylistic translation. This methodology breaks +down the language barrier by transferring the style of a speaker between +languages. The paper is structured into three main phases: augmenting the +speaker's data with stylistically consistent external sources, separating style +from content using machine learning and deep learning techniques, and +generating an abstract style profile by mean pooling the learned embeddings. +The proposed approach is shown to be topic-agnostic, with test accuracy and F1 +scores of 74.9\% and 0.75, respectively. The results demonstrate the potential +of the style profile for multilingual communication, paving the way for further +applications in personalized content generation and cross-linguistic stylistic +transfer. + +**Decision Explanation:** Original decision: REJECT +Fails to meet all mandatory criteria. While it demonstrates practical applications (criterion 1) and experimental results (criterion 2), there is no explicit comparison with state-of-the-art methods (criterion 3). The abstract does not mention LLMs specifically, and the methodology's connection to LLMs remains ambiguous despite potential deep learning use. + +--- + +## [Spatially-Delineated Domain-Adapted AI Classification: An Application + for Oncology Data](https://arxiv.org/abs/2501.11695v1) +**arXiv ID:** 2501.11695v1 + +**Abstract:** +> Given multi-type point maps from different place-types (e.g., tumor regions), +our objective is to develop a classifier trained on the source place-type to +accurately distinguish between two classes of the target place-type based on +their point arrangements. This problem is societally important for many +applications, such as generating clinical hypotheses for designing new +immunotherapies for cancer treatment. The challenge lies in the spatial +variability, the inherent heterogeneity and variation observed in spatial +properties or arrangements across different locations (i.e., place-types). +Previous techniques focus on self-supervised tasks to learn domain-invariant +features and mitigate domain differences; however, they often neglect the +underlying spatial arrangements among data points, leading to significant +discrepancies across different place-types. We explore a novel multi-task +self-learning framework that targets spatial arrangements, such as spatial +mix-up masking and spatial contrastive predictive coding, for +spatially-delineated domain-adapted AI classification. Experimental results on +real-world datasets (e.g., oncology data) show that the proposed framework +provides higher prediction accuracy than baseline methods. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications (oncology data classification for cancer treatment), which falls under the mandatory rejection criteria. While it meets some technical criteria (experimental results and methodology), medical applications are explicitly excluded regardless of other merits. + +--- + +## [Human services organizations and the responsible integration of AI: + Considering ethics and contextualizing risk(s)](https://arxiv.org/abs/2501.11705v1) +**arXiv ID:** 2501.11705v1 + +**Abstract:** +> This paper examines the responsible integration of artificial intelligence +(AI) in human services organizations (HSOs), proposing a nuanced framework for +evaluating AI applications across multiple dimensions of risk. The authors +argue that ethical concerns about AI deployment -- including professional +judgment displacement, environmental impact, model bias, and data laborer +exploitation -- vary significantly based on implementation context and specific +use cases. They challenge the binary view of AI adoption, demonstrating how +different applications present varying levels of risk that can often be +effectively managed through careful implementation strategies. The paper +highlights promising solutions, such as local large language models, that can +facilitate responsible AI integration while addressing common ethical concerns. +The authors propose a dimensional risk assessment approach that considers +factors like data sensitivity, professional oversight requirements, and +potential impact on client wellbeing. They conclude by outlining a path forward +that emphasizes empirical evaluation, starting with lower-risk applications and +building evidence-based understanding through careful experimentation. This +approach enables organizations to maintain high ethical standards while +thoughtfully exploring how AI might enhance their capacity to serve clients and +communities effectively. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on responsible AI integration and ethical concerns in human services organizations, which falls under the rejection criteria for focusing on responsible AI applications and social implications. It fails to meet mandatory criteria 1 (Practical Applications in specified areas), 2 (Experimental Results), and 3 (Comparison with SOTA). While addressing real-world challenges (criterion 5), the core focus violates explicit rejection parameters. + +--- + +## [GL-ICNN: An End-To-End Interpretable Convolutional Neural Network for + the Diagnosis and Prediction of Alzheimer's Disease](https://arxiv.org/abs/2501.11715v1) +**arXiv ID:** 2501.11715v1 + +**Abstract:** +> Deep learning methods based on Convolutional Neural Networks (CNNs) have +shown great potential to improve early and accurate diagnosis of Alzheimer's +disease (AD) dementia based on imaging data. However, these methods have yet to +be widely adopted in clinical practice, possibly due to the limited +interpretability of deep learning models. The Explainable Boosting Machine +(EBM) is a glass-box model but cannot learn features directly from input +imaging data. In this study, we propose a novel interpretable model that +combines CNNs and EBMs for the diagnosis and prediction of AD. We develop an +innovative training strategy that alternatingly trains the CNN component as a +feature extractor and the EBM component as the output block to form an +end-to-end model. The model takes imaging data as input and provides both +predictions and interpretable feature importance measures. We validated the +proposed model on the Alzheimer's Disease Neuroimaging Initiative (ADNI) +dataset and the Health-RI Parelsnoer Neurodegenerative Diseases Biobank (PND) +as an external testing set. The proposed model achieved an area-under-the-curve +(AUC) of 0.956 for AD and control classification, and 0.694 for the prediction +of conversion of mild cognitive impairment (MCI) to AD on the ADNI cohort. The +proposed model is a glass-box model that achieves a comparable performance with +other state-of-the-art black-box models. Our code is publicly available at: +https://anonymous.4open.science/r/GL-ICNN. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (Alzheimer's disease diagnosis/prediction), which is an explicit rejection criterion. While it addresses interpretability and includes experimental results, the medical focus disqualifies it regardless of other merits. + +--- + +## [Episodic memory in AI agents poses risks that should be studied and + mitigated](https://arxiv.org/abs/2501.11739v1) +**arXiv ID:** 2501.11739v1 + +**Abstract:** +> Most current AI models have little ability to store and later retrieve a +record or representation of what they do. In human cognition, episodic memories +play an important role in both recall of the past as well as planning for the +future. The ability to form and use episodic memories would similarly enable a +broad range of improved capabilities in an AI agent that interacts with and +takes actions in the world. Researchers have begun directing more attention to +developing memory abilities in AI models. It is therefore likely that models +with such capability will be become widespread in the near future. This could +in some ways contribute to making such AI agents safer by enabling users to +better monitor, understand, and control their actions. However, as a new +capability with wide applications, we argue that it will also introduce +significant new risks that researchers should begin to study and address. We +outline these risks and benefits and propose four principles to guide the +development of episodic memory capabilities so that these will enhance, rather +than undermine, the effort to keep AI safe and trustworthy. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on analyzing risks and safety implications of episodic memory in AI agents, falling under responsible AI/ethics applications. While it mentions agentic AI capabilities, it fails to meet mandatory criteria 2 (no experimental results/quantitative metrics) and 3 (no SOTA comparisons). It violates the rejection rule for focusing on responsible AI applications. + +--- + +## [Human-AI Collaborative Game Testing with Vision Language Models](https://arxiv.org/abs/2501.11782v1) +**arXiv ID:** 2501.11782v1 + +**Abstract:** +> As modern video games become increasingly complex, traditional manual testing +methods are proving costly and inefficient, limiting the ability to ensure +high-quality game experiences. While advancements in Artificial Intelligence +(AI) offer the potential to assist human testers, the effectiveness of AI in +truly enhancing real-world human performance remains underexplored. This study +investigates how AI can improve game testing by developing and experimenting +with an AI-assisted workflow that leverages state-of-the-art machine learning +models for defect detection. Through an experiment involving 800 test cases and +276 participants of varying backgrounds, we evaluate the effectiveness of AI +assistance under four conditions: with or without AI support, and with or +without detailed knowledge of defects and design documentation. The results +indicate that AI assistance significantly improves defect identification +performance, particularly when paired with detailed knowledge. However, +challenges arise when AI errors occur, negatively impacting human +decision-making. Our findings show the importance of optimizing human-AI +collaboration and implementing strategies to mitigate the effects of AI +inaccuracies. By this research, we demonstrate AI's potential and problems in +enhancing efficiency and accuracy in game testing workflows and offers +practical insights for integrating AI into the testing process. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet all mandatory criteria: While it addresses practical applications (game testing) and includes experimental results (800 test cases, quantitative metrics), it lacks explicit comparison with state-of-the-art techniques. The focus on video game testing also risks falling under 'video processing' exclusion, though marginally. Mandatory criteria 3 is unmet, leading to rejection. + +--- + +## [PXGen: A Post-hoc Explainable Method for Generative Models](https://arxiv.org/abs/2501.11827v1) +**arXiv ID:** 2501.11827v1 + +**Abstract:** +> With the rapid growth of generative AI in numerous applications, explainable +AI (XAI) plays a crucial role in ensuring the responsible development and +deployment of generative AI technologies. XAI has undergone notable +advancements and widespread adoption in recent years, reflecting a concerted +push to enhance the transparency, interpretability, and credibility of AI +systems. Recent research emphasizes that a proficient XAI method should adhere +to a set of criteria, primarily focusing on two key areas. Firstly, it should +ensure the quality and fluidity of explanations, encompassing aspects like +faithfulness, plausibility, completeness, and tailoring to individual needs. +Secondly, the design principle of the XAI system or mechanism should cover the +following factors such as reliability, resilience, the verifiability of its +outputs, and the transparency of its algorithm. However, research in XAI for +generative models remains relatively scarce, with little exploration into how +such methods can effectively meet these criteria in that domain. In this work, +we propose PXGen, a post-hoc explainable method for generative models. Given a +model that needs to be explained, PXGen prepares two materials for the +explanation, the Anchor set and intrinsic & extrinsic criteria. Those materials +are customizable by users according to their purpose and requirements. Via the +calculation of each criterion, each anchor has a set of feature values and +PXGen provides examplebased explanation methods according to the feature values +among all the anchors and illustrated and visualized to the users via tractable +algorithms such as k-dispersion or k-center. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on explainable AI (XAI) for generative models, which aligns with responsible AI development (a rejection criterion). While it addresses methodology (criterion 4) and real-world challenges (criterion 5), it fails to demonstrate practical applications in LLM domains like knowledge graphs/RAG/agentic AI (mandatory criterion 1) or provide experimental comparisons with SOTA (mandatory criterion 3). Additionally, its emphasis on XAI's role in 'responsible development' directly conflicts with the rejection rule for papers primarily focused on responsible AI/ethics. + +--- + +## [Is your LLM trapped in a Mental Set? Investigative study on how mental + sets affect the reasoning capabilities of LLMs](https://arxiv.org/abs/2501.11833v1) +**arXiv ID:** 2501.11833v1 + +**Abstract:** +> In this paper, we present an investigative study on how Mental Sets influence +the reasoning capabilities of LLMs. LLMs have excelled in diverse natural +language processing (NLP) tasks, driven by advancements in parameter-efficient +fine-tuning (PEFT) and emergent capabilities like in-context learning (ICL). +For complex reasoning tasks, selecting the right model for PEFT or ICL is +critical, often relying on scores on benchmarks such as MMLU, MATH, and GSM8K. +However, current evaluation methods, based on metrics like F1 Score or +reasoning chain assessments by larger models, overlook a key dimension: +adaptability to unfamiliar situations and overcoming entrenched thinking +patterns. In cognitive psychology, Mental Set refers to the tendency to persist +with previously successful strategies, even when they become inefficient - a +challenge for problem solving and reasoning. We compare the performance of LLM +models like Llama-3.1-8B-Instruct, Llama-3.1-70B-Instruct and GPT-4o in the +presence of mental sets. To the best of our knowledge, this is the first study +to integrate cognitive psychology concepts into the evaluation of LLMs for +complex reasoning tasks, providing deeper insights into their adaptability and +problem-solving efficacy. + +**Decision Explanation:** Original decision: REJECT +Fails to meet mandatory criteria 1 (Practical Applications) as the focus is on cognitive psychology-based evaluation rather than real-world LLM applications like knowledge graphs, RAG, or agentic AI. Also lacks direct real-world implementation discussion (criterion 5). + +--- + +## [A Survey on Memory-Efficient Large-Scale Model Training in AI for + Science](https://arxiv.org/abs/2501.11847v1) +**arXiv ID:** 2501.11847v1 + +**Abstract:** +> Scientific research faces high costs and inefficiencies with traditional +methods, but the rise of deep learning and large language models (LLMs) offers +innovative solutions. This survey reviews LLM applications across scientific +fields such as biology, medicine, chemistry, and meteorology, underscoring +their role in advancing research. However, the continuous expansion of model +size has led to significant memory demands, hindering further development and +application of LLMs for science. To address this, we review memory-efficient +training techniques for LLMs based on the transformer architecture, including +distributed training, mixed precision training, and gradient checkpointing. +Using AlphaFold 2 as an example, we demonstrate how tailored memory +optimization methods can reduce storage needs while preserving prediction +accuracy. We also discuss the challenges of memory optimization in practice and +potential future directions, hoping to provide valuable insights for +researchers and engineers. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on AI for Science applications, including medicine, which falls under the rejection criteria. While it discusses memory-efficient training techniques, it does not meet all mandatory criteria (e.g., direct comparison with SOTA for new methodologies) and violates the medical application exclusion. + +--- + +## [Coarse-to-Fine Lightweight Meta-Embedding for ID-Based Recommendation](https://arxiv.org/abs/2501.11870v1) +**arXiv ID:** 2501.11870v1 + +**Abstract:** +> The state-of-the-art recommendation systems have shifted the attention to +efficient recommendation, e.g., on-device recommendation, under memory +constraints. To this end, the existing methods either focused on the +lightweight embeddings for both users and items, or involved on-device systems +enjoying the compact embeddings to enhance reusability and reduces space +complexity. However, they focus solely on the coarse granularity of embedding, +while overlook the fine-grained semantic nuances, to adversarially downgrade +the efficacy of meta-embeddings in capturing the intricate relationship over +both user and item, consequently resulting into the suboptimal recommendations. +In this paper, we aim to study how the meta-embedding can efficiently learn +varied grained semantics, together with how the fine-grained meta-embedding can +strengthen the representation of coarse-grained meta-embedding. To answer these +questions, we develop a novel graph neural networks (GNNs) based recommender +where each user and item serves as the node, linked directly to coarse-grained +virtual nodes and indirectly to fine-grained virtual nodes, ensuring different +grained semantic learning, while disclosing: 1) In contrast to coarse-grained +semantics, fine-grained semantics are well captured through sparse +meta-embeddings, which adaptively 2) balance the embedding uniqueness and +memory constraint. Additionally, the initialization method come up upon +SparsePCA, along with a soft thresholding activation function to render the +sparseness of the meta-embeddings. We propose a weight bridging update strategy +that focuses on matching each coarse-grained meta-embedding with several +fine-grained meta-embeddings based on the users/items' semantics. Extensive +experiments substantiate our method's superiority over existing baselines. Our +code is available at https://github.com/htyjers/C2F-MetaEmbed. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on recommendation systems using meta-embeddings and GNNs but does not explicitly address practical applications of LLMs (Criterion 1 unmet). While it meets Criteria 2 and 3 (experimental results and SOTA comparisons), it fails to satisfy all mandatory criteria. Additionally, the application domain (recommendation systems) is not among the specified LLM-related areas (knowledge graphs, RAG, agentic AI) in the mandatory criteria. + +--- + +## [Community-Aware Temporal Walks: Parameter-Free Representation Learning + on Continuous-Time Dynamic Graphs](https://arxiv.org/abs/2501.11880v1) +**arXiv ID:** 2501.11880v1 + +**Abstract:** +> Dynamic graph representation learning plays a crucial role in understanding +evolving behaviors. However, existing methods often struggle with flexibility, +adaptability, and the preservation of temporal and structural dynamics. To +address these issues, we propose Community-aware Temporal Walks (CTWalks), a +novel framework for representation learning on continuous-time dynamic graphs. +CTWalks integrates three key components: a community-based parameter-free +temporal walk sampling mechanism, an anonymization strategy enriched with +community labels, and an encoding process that leverages continuous temporal +dynamics modeled via ordinary differential equations (ODEs). This design +enables precise modeling of both intra- and inter-community interactions, +offering a fine-grained representation of evolving temporal patterns in +continuous-time dynamic graphs. CTWalks theoretically overcomes locality bias +in walks and establishes its connection to matrix factorization. Experiments on +benchmark datasets demonstrate that CTWalks outperforms established methods in +temporal link prediction tasks, achieving higher accuracy while maintaining +robustness. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on dynamic graph representation learning (CTWalks) for temporal link prediction but does not explicitly mention applications involving Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails to satisfy criterion 1 (LLM practical applications). No explicit connection to LLM-related domains or methodologies is evident in the abstract. + +--- + +## [A Lightweight and Interpretable Deepfakes Detection Framework](https://arxiv.org/abs/2501.11927v1) +**arXiv ID:** 2501.11927v1 + +**Abstract:** +> The recent realistic creation and dissemination of so-called deepfakes poses +a serious threat to social life, civil rest, and law. Celebrity defaming, +election manipulation, and deepfakes as evidence in court of law are few +potential consequences of deepfakes. The availability of open source trained +models based on modern frameworks such as PyTorch or TensorFlow, video +manipulations Apps such as FaceApp and REFACE, and economical computing +infrastructure has easen the creation of deepfakes. Most of the existing +detectors focus on detecting either face-swap, lip-sync, or puppet master +deepfakes, but a unified framework to detect all three types of deepfakes is +hardly explored. This paper presents a unified framework that exploits the +power of proposed feature fusion of hybrid facial landmarks and our novel heart +rate features for detection of all types of deepfakes. We propose novel heart +rate features and fused them with the facial landmark features to better +extract the facial artifacts of fake videos and natural variations available in +the original videos. We used these features to train a light-weight XGBoost to +classify between the deepfake and bonafide videos. We evaluated the performance +of our framework on the world leaders dataset (WLDR) that contains all types of +deepfakes. Experimental results illustrate that the proposed framework offers +superior detection performance over the comparative deepfakes detection +methods. Performance comparison of our framework against the LSTM-FCN, a +candidate of deep learning model, shows that proposed model achieves similar +results, however, it is more interpretable. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on video processing (deepfakes detection) and social harm implications of AI, both listed as rejection criteria. While it meets criteria 2 (experimental results) and 3 (SOTA comparison), it fails the mandatory LLM application requirement and violates two rejection conditions. + +--- + +## [MeshONet: A Generalizable and Efficient Operator Learning Method for + Structured Mesh Generation](https://arxiv.org/abs/2501.11937v1) +**arXiv ID:** 2501.11937v1 + +**Abstract:** +> Mesh generation plays a crucial role in scientific computing. Traditional +mesh generation methods, such as TFI and PDE-based methods, often struggle to +achieve a balance between efficiency and mesh quality. To address this +challenge, physics-informed intelligent learning methods have recently emerged, +significantly improving generation efficiency while maintaining high mesh +quality. However, physics-informed methods fail to generalize when applied to +previously unseen geometries, as even small changes in the boundary shape +necessitate burdensome retraining to adapt to new geometric variations. In this +paper, we introduce MeshONet, the first generalizable intelligent learning +method for structured mesh generation. The method transforms the mesh +generation task into an operator learning problem with multiple input and +solution functions. To effectively overcome the multivariable mapping +restriction of operator learning methods, we propose a dual-branch, +shared-trunk architecture to approximate the mapping between function spaces +based on input-output pairs. Experimental results show that MeshONet achieves a +speedup of up to four orders of magnitude in generation efficiency over +traditional methods. It also enables generalization to different geometries +without retraining, greatly enhancing the practicality of intelligent methods. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on operator learning for mesh generation in scientific computing but does not explicitly address Large Language Models (LLMs) in its methodology or applications. While it meets Criteria 2 (quantitative metrics) and 3 (comparison with SOTA), it fails Criterion 1 (LLM-based practical applications). Mandatory criteria require ALL to be met, leading to rejection despite other strengths. + +--- + +## [Survey on Hand Gesture Recognition from Visual Input](https://arxiv.org/abs/2501.11992v1) +**arXiv ID:** 2501.11992v1 + +**Abstract:** +> Hand gesture recognition has become an important research area, driven by the +growing demand for human-computer interaction in fields such as sign language +recognition, virtual and augmented reality, and robotics. Despite the rapid +growth of the field, there are few surveys that comprehensively cover recent +research developments, available solutions, and benchmark datasets. This survey +addresses this gap by examining the latest advancements in hand gesture and 3D +hand pose recognition from various types of camera input data including RGB +images, depth images, and videos from monocular or multiview cameras, examining +the differing methodological requirements of each approach. Furthermore, an +overview of widely used datasets is provided, detailing their main +characteristics and application domains. Finally, open challenges such as +achieving robust recognition in real-world environments, handling occlusions, +ensuring generalization across diverse users, and addressing computational +efficiency for real-time applications are highlighted to guide future research +directions. By synthesizing the objectives, methodologies, and applications of +recent studies, this survey offers valuable insights into current trends, +challenges, and opportunities for future research in human hand gesture +recognition. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on hand gesture recognition from visual inputs (including video processing), which is explicitly listed as a rejection criterion. While it addresses practical applications in human-computer interaction, it does not involve Large Language Models (LLMs) or meet the required criteria related to knowledge graphs, RAG, agentic AI, or experimental validation of LLM-based approaches. + +--- + +## [Full Proportional Justified Representation](https://arxiv.org/abs/2501.12015v1) +**arXiv ID:** 2501.12015v1 + +**Abstract:** +> In multiwinner approval voting, forming a committee that proportionally +represents voters' approval ballots is an essential task. The notion of +justified representation (JR) demands that any large "cohesive" group of voters +should be proportionally "represented". The "cohesiveness" is defined in +different ways; two common ways are the following: (C1) demands that the group +unanimously approves a set of candidates proportional to its size, while (C2) +requires each member to approve at least a fixed fraction of such a set. +Similarly, "representation" have been considered in different ways: (R1) the +coalition's collective utility from the winning set exceeds that of any +proportionally sized alternative, and (R2) for any proportionally sized +alternative, at least one member of the coalition derives less utility from it +than from the winning set. + Three of the four possible combinations have been extensively studied: +(C1)-(R1) defines Proportional Justified Representation (PJR), (C1)-(R2) +defines Extended Justified Representation (EJR), (C2)-(R2) defines Full +Justified Representation (FJR). All three have merits, but also drawbacks. PJR +is the weakest notion, and perhaps not sufficiently demanding; EJR may not be +compatible with perfect representation; and it is open whether a committee +satisfying FJR can be found efficiently. + We study the combination (C2)-(R1), which we call Full Proportional Justified +Representation (FPJR). We investigate FPJR's properties and find that it shares +PJR's advantages over EJR: several proportionality axioms (e.g. priceability, +perfect representation) imply FPJR and PJR but not EJR. We also find that +efficient rules like the greedy Monroe rule and the method of equal shares +satisfy FPJR, matching a key advantage of EJR over FJR. However, the +Proportional Approval Voting (PAV) rule may violate FPJR, so neither of EJR and +FPJR implies the other. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on voting systems and proportional representation in social choice theory, which qualifies as a social application under the rejection criteria. It does not address practical applications of LLMs (Criterion 1 unmet) or present experimental results with quantitative metrics (Criterion 2 unmet). While it compares methodologies (Criterion 3), the mandatory requirements are not fully satisfied, and the social application focus triggers rejection. + +--- + +## [Adaptive Class Learning to Screen Diabetic Disorders in Fundus Images of + Eye](https://arxiv.org/abs/2501.12048v1) +**arXiv ID:** 2501.12048v1 + +**Abstract:** +> The prevalence of ocular illnesses is growing globally, presenting a +substantial public health challenge. Early detection and timely intervention +are crucial for averting visual impairment and enhancing patient prognosis. +This research introduces a new framework called Class Extension with Limited +Data (CELD) to train a classifier to categorize retinal fundus images. The +classifier is initially trained to identify relevant features concerning +Healthy and Diabetic Retinopathy (DR) classes and later fine-tuned to adapt to +the task of classifying the input images into three classes: Healthy, DR, and +Glaucoma. This strategy allows the model to gradually enhance its +classification capabilities, which is beneficial in situations where there are +only a limited number of labeled datasets available. Perturbation methods are +also used to identify the input image characteristics responsible for +influencing the models decision-making process. We achieve an overall accuracy +of 91% on publicly available datasets. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications (diabetic/glaucoma screening in eye images) and does not address Large Language Models (LLMs) or their practical applications in knowledge graphs, RAG, or agentic AI. It fails criteria 1 and 3, and medical applications are explicitly listed as a rejection reason. + +--- + +## [Teacher Encoder-Student Decoder Denoising Guided Segmentation Network + for Anomaly Detection](https://arxiv.org/abs/2501.12104v1) +**arXiv ID:** 2501.12104v1 + +**Abstract:** +> Visual anomaly detection is a highly challenging task, often categorized as a +one-class classification and segmentation problem. Recent studies have +demonstrated that the student-teacher (S-T) framework effectively addresses +this challenge. However, most S-T frameworks rely solely on pre-trained teacher +networks to guide student networks in learning multi-scale similar features, +overlooking the potential of the student networks to enhance learning through +multi-scale feature fusion. In this study, we propose a novel model named +PFADSeg, which integrates a pre-trained teacher network, a denoising student +network with multi-scale feature fusion, and a guided anomaly segmentation +network into a unified framework. By adopting a unique teacher-encoder and +student-decoder denoising mode, the model improves the student network's +ability to learn from teacher network features. Furthermore, an adaptive +feature fusion mechanism is introduced to train a self-supervised segmentation +network that synthesizes anomaly masks autonomously, significantly increasing +detection performance. Evaluated on the MVTec AD dataset, PFADSeg achieves +state-of-the-art results with an image-level AUC of 98.9%, a pixel-level mean +precision of 76.4%, and an instance-level mean precision of 78.7%. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on visual anomaly detection using a student-teacher framework in computer vision, but does not address Large Language Models (LLMs) or their practical applications in domains like knowledge graphs, RAG, or agentic AI. This violates mandatory criterion 1 for paper selection. + +--- + +## [Can open source large language models be used for tumor documentation in + Germany? -- An evaluation on urological doctors' notes](https://arxiv.org/abs/2501.12106v1) +**arXiv ID:** 2501.12106v1 + +**Abstract:** +> Tumor documentation in Germany is largely done manually, requiring reading +patient records and entering data into structured databases. Large language +models (LLMs) could potentially enhance this process by improving efficiency +and reliability. This evaluation tests eleven different open source LLMs with +sizes ranging from 1-70 billion model parameters on three basic tasks of the +tumor documentation process: identifying tumor diagnoses, assigning ICD-10 +codes, and extracting the date of first diagnosis. For evaluating the LLMs on +these tasks, a dataset of annotated text snippets based on anonymized doctors' +notes from urology was prepared. Different prompting strategies were used to +investigate the effect of the number of examples in few-shot prompting and to +explore the capabilities of the LLMs in general. The models Llama 3.1 8B, +Mistral 7B, and Mistral NeMo 12 B performed comparably well in the tasks. +Models with less extensive training data or having fewer than 7 billion +parameters showed notably lower performance, while larger models did not +display performance gains. Examples from a different medical domain than +urology could also improve the outcome in few-shot prompting, which +demonstrates the ability of LLMs to handle tasks needed for tumor +documentation. Open source LLMs show a strong potential for automating tumor +documentation. Models from 7-12 billion parameters could offer an optimal +balance between performance and resource efficiency. With tailored fine-tuning +and well-designed prompting, these models might become important tools for +clinical documentation in the future. The code for the evaluation is available +from https://github.com/stefan-m-lenz/UroLlmEval. We also release the dataset +as a new valuable resource that addresses the shortage of authentic and easily +accessible benchmarks in German-language medical NLP. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (tumor documentation in urology), which is an explicit rejection criterion. While it meets criteria 2 (experimental results) and 3 (comparisons), medical applications are prohibited regardless of other merits. + +--- + +## [FedCLEAN: byzantine defense by CLustering Errors of Activation maps in + Non-IID federated learning environments](https://arxiv.org/abs/2501.12123v1) +**arXiv ID:** 2501.12123v1 + +**Abstract:** +> Federated Learning (FL) enables clients to collaboratively train a global +model using their local datasets while reinforcing data privacy. However, FL is +susceptible to poisoning attacks. Existing defense mechanisms assume that +clients' data are independent and identically distributed (IID), making them +ineffective in real-world applications where data are non-IID. This paper +presents FedCLEAN, the first defense capable of filtering attackers' model +updates in a non-IID FL environment. The originality of FedCLEAN is twofold. +First, it relies on a client confidence score derived from the reconstruction +errors of each client's model activation maps for a given trigger set, with +reconstruction errors obtained by means of a Conditional Variational +Autoencoder trained according to a novel server-side strategy. Second, we +propose an ad-hoc trust propagation algorithm based on client scores, which +allows building a cluster of benign clients while flagging potential attackers. +Experimental results on the datasets MNIST and FashionMNIST demonstrate the +robustness of FedCLEAN against Byzantine attackers in non-IID scenarios and a +close-to-zero benign client misclassification rate, even in the absence of an +attack. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet Criterion 1 (Practical Applications of LLMs), as it focuses on federated learning defense mechanisms rather than LLM applications like knowledge graphs, RAG, or agentic AI. While it meets Criteria 2 (experimental results) and 3 (comparison with SOTA), failure to satisfy ALL mandatory criteria (1-3) necessitates rejection. + +--- + +## [FuocChuVIP123 at CoMeDi Shared Task: Disagreement Ranking with + XLM-Roberta Sentence Embeddings and Deep Neural Regression](https://arxiv.org/abs/2501.12336v1) +**arXiv ID:** 2501.12336v1 + +**Abstract:** +> This paper presents results of our system for CoMeDi Shared Task, focusing on +Subtask 2: Disagreement Ranking. Our system leverages sentence embeddings +generated by the paraphrase-xlm-r-multilingual-v1 model, combined with a deep +neural regression model incorporating batch normalization and dropout for +improved generalization. By predicting the mean of pairwise judgment +differences between annotators, our method explicitly targets disagreement +ranking, diverging from traditional "gold label" aggregation approaches. We +optimized our system with a customized architecture and training procedure, +achieving competitive performance in Spearman correlation against mean +disagreement labels. Our results highlight the importance of robust embeddings, +effective model architecture, and careful handling of judgment differences for +ranking disagreement in multilingual contexts. These findings provide insights +into the use of contextualized representations for ordinal judgment tasks and +open avenues for further refinement of disagreement prediction models. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on multilingual disagreement ranking using XLM-R embeddings but does not address practical LLM applications like knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (quantitative metrics) and 4 (methodology details), it fails criterion 1 (practical LLM applications) and 3 (explicit SOTA comparisons). The task aligns more with NLP annotation analysis than core LLM application domains. + +--- + +## [Video Depth Anything: Consistent Depth Estimation for Super-Long Videos](https://arxiv.org/abs/2501.12375v1) +**arXiv ID:** 2501.12375v1 + +**Abstract:** +> Depth Anything has achieved remarkable success in monocular depth estimation +with strong generalization ability. However, it suffers from temporal +inconsistency in videos, hindering its practical applications. Various methods +have been proposed to alleviate this issue by leveraging video generation +models or introducing priors from optical flow and camera poses. Nonetheless, +these methods are only applicable to short videos (< 10 seconds) and require a +trade-off between quality and computational efficiency. We propose Video Depth +Anything for high-quality, consistent depth estimation in super-long videos +(over several minutes) without sacrificing efficiency. We base our model on +Depth Anything V2 and replace its head with an efficient spatial-temporal head. +We design a straightforward yet effective temporal consistency loss by +constraining the temporal depth gradient, eliminating the need for additional +geometric priors. The model is trained on a joint dataset of video depth and +unlabeled images, similar to Depth Anything V2. Moreover, a novel +key-frame-based strategy is developed for long video inference. Experiments +show that our model can be applied to arbitrarily long videos without +compromising quality, consistency, or generalization ability. Comprehensive +evaluations on multiple video benchmarks demonstrate that our approach sets a +new state-of-the-art in zero-shot video depth estimation. We offer models of +different scales to support a range of scenarios, with our smallest model +capable of real-time performance at 30 FPS. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing (depth estimation in super-long videos) and does not mention LLM applications, which violates the mandatory rejection criteria. While it meets experimental results and SOTA comparison requirements, it fails core practical application criteria for LLM-focused research. + +--- + +## [Leveraging GANs For Active Appearance Models Optimized Model Fitting](https://arxiv.org/abs/2501.11218v1) +**arXiv ID:** 2501.11218v1 + +**Abstract:** +> Generative Adversarial Networks (GANs) have gained prominence in refining +model fitting tasks in computer vision, particularly in domains involving +deformable models like Active Appearance Models (AAMs). This paper explores the +integration of GANs to enhance the AAM fitting process, addressing challenges +in optimizing nonlinear parameters associated with appearance and shape +variations. By leveraging GANs' adversarial training framework, the aim is to +minimize fitting errors and improve convergence rates. Achieving robust +performance even in cases with high appearance variability and occlusions. Our +approach demonstrates significant improvements in accuracy and computational +efficiency compared to traditional optimization techniques, thus establishing +GANs as a potent tool for advanced image model fitting. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on GANs and computer vision applications (AAM model fitting) rather than Large Language Models (LLMs), failing to meet the mandatory Criterion 1. Additionally, it does not address LLM-related areas like knowledge graphs, RAG, or agentic AI required for inclusion. + +--- + +## [WSSM: Geographic-enhanced hierarchical state-space model for global + station weather forecast](https://arxiv.org/abs/2501.11238v1) +**arXiv ID:** 2501.11238v1 + +**Abstract:** +> Global Station Weather Forecasting (GSWF), a prominent meteorological +research area, is pivotal in providing timely localized weather predictions. +Despite the progress existing models have made in the overall accuracy of the +GSWF, executing high-precision extreme event prediction still presents a +substantial challenge. The recent emergence of state-space models, with their +ability to efficiently capture continuous-time dynamics and latent states, +offer potential solutions. However, early investigations indicated that Mamba +underperforms in the context of GSWF, suggesting further adaptation and +optimization. To tackle this problem, in this paper, we introduce Weather +State-space Model (WSSM), a novel Mamba-based approach tailored for GSWF. +Geographical knowledge is integrated in addition to the widely-used positional +encoding to represent the absolute special-temporal position. The multi-scale +time-frequency features are synthesized from coarse to fine to model the +seasonal to extreme weather dynamic. Our method effectively improves the +overall prediction accuracy and addresses the challenge of forecasting extreme +weather events. The state-of-the-art results obtained on the Weather-5K subset +underscore the efficacy of the WSSM + +**Decision Explanation:** Original decision: REJECT +The paper focuses on weather forecasting using a state-space model (Mamba) but does not explicitly mention Large Language Models (LLMs) or their practical applications (Criterion 1 unmet). While it meets Criteria 2 (experimental results) and 3 (SOTA comparison), failing Criterion 1 (mandatory) necessitates rejection. No explicit LLM integration or discussion aligns with the required focus. + +--- + +## [Spatiotemporal Air Quality Mapping in Urban Areas Using Sparse Sensor + Data, Satellite Imagery, Meteorological Factors, and Spatial Features](https://arxiv.org/abs/2501.11270v1) +**arXiv ID:** 2501.11270v1 + +**Abstract:** +> Monitoring air pollution is crucial for protecting human health from exposure +to harmful substances. Traditional methods of air quality monitoring, such as +ground-based sensors and satellite-based remote sensing, face limitations due +to high deployment costs, sparse sensor coverage, and environmental +interferences. To address these challenges, this paper proposes a framework for +high-resolution spatiotemporal Air Quality Index (AQI) mapping using sparse +sensor data, satellite imagery, and various spatiotemporal factors. By +leveraging Graph Neural Networks (GNNs), we estimate AQI values at unmonitored +locations based on both spatial and temporal dependencies. The framework +incorporates a wide range of environmental features, including meteorological +data, road networks, points of interest (PoIs), population density, and urban +green spaces, which enhance prediction accuracy. We illustrate the use of our +approach through a case study in Lahore, Pakistan, where multi-resolution data +is used to generate the air quality index map at a fine spatiotemporal scale. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on air quality mapping using GNNs and environmental data without mentioning LLMs, knowledge graphs, RAG, or agentic AI. It fails to meet mandatory criteria 1 (Practical Applications of LLMs) and 3 (Comparison with SOTA), and provides no explicit evidence for criterion 2 (quantitative LLM-related metrics). While addressing real-world applications, it does not align with the required LLM/agentic AI focus. + +--- + +## [A Machine Learning Framework for Handling Unreliable Absence Label and + Class Imbalance for Marine Stinger Beaching Prediction](https://arxiv.org/abs/2501.11293v1) +**arXiv ID:** 2501.11293v1 + +**Abstract:** +> Bluebottles (\textit{Physalia} spp.) are marine stingers resembling +jellyfish, whose presence on Australian beaches poses a significant public risk +due to their venomous nature. Understanding the environmental factors driving +bluebottles ashore is crucial for mitigating their impact, and machine learning +tools are to date relatively unexplored. We use bluebottle marine stinger +presence/absence data from beaches in Eastern Sydney, Australia, and compare +machine learning models (Multilayer Perceptron, Random Forest, and XGBoost) to +identify factors influencing their presence. We address challenges such as +class imbalance, class overlap, and unreliable absence data by employing data +augmentation techniques, including the Synthetic Minority Oversampling +Technique (SMOTE), Random Undersampling, and Synthetic Negative Approach that +excludes the negative class. Our results show that SMOTE failed to resolve +class overlap, but the presence-focused approach effectively handled imbalance, +class overlap, and ambiguous absence data. The data attributes such as the wind +direction, which is a circular variable, emerged as a key factor influencing +bluebottle presence, confirming previous inference studies. However, in the +absence of population dynamics, biological behaviours, and life cycles, the +best predictive model appears to be Random Forests combined with Synthetic +Negative Approach. This research contributes to mitigating the risks posed by +bluebottles to beachgoers and provides insights into handling class overlap and +unreliable negative class in environmental modelling. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on traditional machine learning models (Random Forest, XGBoost) for environmental prediction without mentioning LLMs, knowledge graphs, RAG, or agentic AI. It fails Criteria 1 (Practical Applications of LLMs) and does not meet all mandatory requirements. + +--- + +## [On the Dimension of Pullback Attractors in Recurrent Neural Networks](https://arxiv.org/abs/2501.11357v1) +**arXiv ID:** 2501.11357v1 + +**Abstract:** +> Recurrent Neural Networks (RNNs) are high-dimensional state space models +capable of learning functions on sequence data. Recently, it has been +conjectured that reservoir computers, a particular class of RNNs, trained on +observations of a dynamical systems can be interpreted as embeddings. This +result has been established for the case of linear reservoir systems. In this +work, we use a nonautonomous dynamical systems approach to establish an upper +bound for the fractal dimension of the subset of reservoir state space +approximated during training and prediction phase. We prove that when the input +sequences comes from an Nin-dimensional invertible dynamical system, the +fractal dimension of this set is bounded above by Nin. The result obtained here +are useful in dimensionality reduction of computation in RNNs as well as +estimating fractal dimensions of dynamical systems from limited observations of +their time series. It is also a step towards understanding embedding properties +of reservoir computers. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on theoretical analysis of fractal dimensions in reservoir computing rather than practical LLM applications. While it mentions dimensionality reduction implications, it lacks experimental results with quantitative metrics (Criterion 2) and comparisons with state-of-the-art methods (Criterion 3). The abstract emphasizes mathematical proofs over real-world implementations or agentic AI scenarios. + +--- + +## [Investigation of Whisper ASR Hallucinations Induced by Non-Speech Audio](https://arxiv.org/abs/2501.11378v1) +**arXiv ID:** 2501.11378v1 + +**Abstract:** +> Hallucinations of deep neural models are amongst key challenges in automatic +speech recognition (ASR). In this paper, we investigate hallucinations of the +Whisper ASR model induced by non-speech audio segments present during +inference. By inducting hallucinations with various types of sounds, we show +that there exists a set of hallucinations that appear frequently. We then study +hallucinations caused by the augmentation of speech with such sounds. Finally, +we describe the creation of a bag of hallucinations (BoH) that allows to remove +the effect of hallucinations through the post-processing of text +transcriptions. The results of our experiments show that such post-processing +is capable of reducing word error rate (WER) and acts as a good safeguard +against problematic hallucinations. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on automatic speech recognition (ASR) hallucinations in Whisper, not LLM applications like knowledge graphs, RAG, or agentic AI. While it meets Criterion 2 (experimental results), it does not address LLM-based practical applications (Criterion 1) or compare with SOTA LLM methods (Criterion 3). ASR falls outside the LLM scope defined in mandatory criteria. + +--- + +## [A Truly Sparse and General Implementation of Gradient-Based Synaptic + Plasticity](https://arxiv.org/abs/2501.11407v1) +**arXiv ID:** 2501.11407v1 + +**Abstract:** +> Online synaptic plasticity rules derived from gradient descent achieve high +accuracy on a wide range of practical tasks. However, their software +implementation often requires tediously hand-derived gradients or using +gradient backpropagation which sacrifices the online capability of the rules. +In this work, we present a custom automatic differentiation (AD) pipeline for +sparse and online implementation of gradient-based synaptic plasticity rules +that generalizes to arbitrary neuron models. Our work combines the programming +ease of backpropagation-type methods for forward AD while being +memory-efficient. To achieve this, we exploit the advantageous compute and +memory scaling of online synaptic plasticity by providing an inherently sparse +implementation of AD where expensive tensor contractions are replaced with +simple element-wise multiplications if the tensors are diagonal. Gradient-based +synaptic plasticity rules such as eligibility propagation (e-prop) have exactly +this property and thus profit immensely from this feature. We demonstrate the +alignment of our gradients with respect to gradient backpropagation on an +synthetic task where e-prop gradients are exact, as well as audio speech +classification benchmarks. We demonstrate how memory utilization scales with +network size without dependence on the sequence length, as expected from +forward AD methods. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on synaptic plasticity implementation and automatic differentiation methods for neural models but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it demonstrates experimental results and methodology details, it fails to meet the mandatory criterion #1 (Practical Applications of LLMs). + +--- + +## [Generalization and Informativeness of Weighted Conformal Risk Control + Under Covariate Shift](https://arxiv.org/abs/2501.11413v1) +**arXiv ID:** 2501.11413v1 + +**Abstract:** +> Predictive models are often required to produce reliable predictions under +statistical conditions that are not matched to the training data. A common type +of training-testing mismatch is covariate shift, where the conditional +distribution of the target variable given the input features remains fixed, +while the marginal distribution of the inputs changes. Weighted conformal risk +control (W-CRC) uses data collected during the training phase to convert point +predictions into prediction sets with valid risk guarantees at test time +despite the presence of a covariate shift. However, while W-CRC provides +statistical reliability, its efficiency -- measured by the size of the +prediction sets -- can only be assessed at test time. In this work, we relate +the generalization properties of the base predictor to the efficiency of W-CRC +under covariate shifts. Specifically, we derive a bound on the inefficiency of +the W-CRC predictor that depends on algorithmic hyperparameters and +task-specific quantities available at training time. This bound offers insights +on relationships between the informativeness of the prediction sets, the extent +of the covariate shift, and the size of the calibration and training sets. +Experiments on fingerprinting-based localization validate the theoretical +results. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on conformal risk control under covariate shift but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it includes experimental validation and theoretical bounds, it fails to meet mandatory criteria 1 (Practical Applications of LLMs) and 3 (Comparison with SOTA). The scope aligns more with general predictive modeling rather than LLM-specific applications required by the criteria. + +--- + +## [Enhancing Coronary Artery Calcium Scoring via Multi-Organ Segmentation + on Non-Contrast Cardiac Computed Tomography](https://arxiv.org/abs/2501.11428v1) +**arXiv ID:** 2501.11428v1 + +**Abstract:** +> Despite coronary artery calcium scoring being considered a largely solved +problem within the realm of medical artificial intelligence, this paper argues +that significant improvements can still be made. By shifting the focus from +pathology detection to a deeper understanding of anatomy, the novel algorithm +proposed in the paper both achieves high accuracy in coronary artery calcium +scoring and offers enhanced interpretability of the results. This approach not +only aids in the precise quantification of calcifications in coronary arteries, +but also provides valuable insights into the underlying anatomical structures. +Through this anatomically-informed methodology, the paper shows how a nuanced +understanding of the heart's anatomy can lead to more accurate and +interpretable results in the field of cardiovascular health. We demonstrate the +superior accuracy of the proposed method by evaluating it on an open-source +multi-vendor dataset, where we obtain results at the inter-observer level, +surpassing the current state of the art. Finally, the qualitative analyses show +the practical value of the algorithm in such tasks as labeling coronary artery +calcifications, identifying aortic calcifications, and filtering out false +positive detections due to noise. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (coronary artery calcium scoring) and does not mention LLMs or related areas like knowledge graphs/RAG/agentic AI. While it meets criteria 2 and 3 with experimental results and SOTA comparisons, it violates mandatory rejection rules for medical AI focus. + +--- + +## [Decomposing Interventional Causality into Synergistic, Redundant, and + Unique Components](https://arxiv.org/abs/2501.11447v1) +**arXiv ID:** 2501.11447v1 + +**Abstract:** +> We introduce a novel framework for decomposing interventional causal effects +into synergistic, redundant, and unique components, building on the intuition +of Partial Information Decomposition (PID) and the principle of M\"obius +inversion. While recent work has explored a similar decomposition of an +observational measure, we argue that a proper causal decomposition must be +interventional in nature. We develop a mathematical approach that +systematically quantifies how causal power is distributed among variables in a +system, using a recently derived closed-form expression for the M\"obius +function of the redundancy lattice. The formalism is then illustrated by +decomposing the causal power in logic gates, cellular automata, and chemical +reaction networks. Our results reveal how the distribution of causal power can +be context- and parameter-dependent. This decomposition provides new insights +into complex systems by revealing how causal influences are shared and combined +among multiple variables, with potential applications ranging from attribution +of responsibility in legal or AI systems, to the analysis of biological +networks or climate models. + +**Decision Explanation:** Original decision: REJECT +The paper does not explicitly focus on practical applications of Large Language Models (LLMs), and its methodology/experiments are centered on causal decomposition in general systems (e.g., logic gates, cellular automata) rather than LLM-specific use cases. While it mentions AI systems as a potential application, it lacks direct ties to LLMs, knowledge graphs, RAG, or agentic AI. Additionally, it does not clearly address experimental metrics or comparisons with LLM-related state-of-the-art methods. + +--- + +## [Generative AI and Large Language Models in Language Preservation: + Opportunities and Challenges](https://arxiv.org/abs/2501.11496v1) +**arXiv ID:** 2501.11496v1 + +**Abstract:** +> Generative AI and large-scale language models (LLM) have emerged as powerful +tools in language preservation, particularly for near-native and endangered +languages. With the increasing reliance on technology for communication, +education, and cultural documentation, new opportunities have emerged to +mitigate the dramatic decline of linguistic diversity worldwide. This paper +examines the role of generative AIs and LLMs in preserving endangered +languages, highlighting the risks and challenges associated with their use. We +analyze the underlying technologies driving these models, including natural +language processing (NLP) and deep learning, and explore several cases where +these technologies have been applied to low-resource languages. Additionally, +we discuss ethical considerations, data scarcity issues, and technical +challenges while proposing solutions to enhance AI-driven language +preservation. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on social applications of AI in linguistic diversity (a rejection criterion) and fails to meet mandatory criteria 2 (no experimental results/quantitative metrics) and 3 (no SOTA comparison). While it addresses practical applications (criterion 1), it violates exclusion rules. + +--- + +## [Dialect2SQL: A Novel Text-to-SQL Dataset for Arabic Dialects with a + Focus on Moroccan Darija](https://arxiv.org/abs/2501.11498v1) +**arXiv ID:** 2501.11498v1 + +**Abstract:** +> The task of converting natural language questions (NLQs) into executable SQL +queries, known as text-to-SQL, has gained significant interest in recent years, +as it enables non-technical users to interact with relational databases. Many +benchmarks, such as SPIDER and WikiSQL, have contributed to the development of +new models and the evaluation of their performance. In addition, other +datasets, like SEDE and BIRD, have introduced more challenges and complexities +to better map real-world scenarios. However, these datasets primarily focus on +high-resource languages such as English and Chinese. In this work, we introduce +Dialect2SQL, the first large-scale, cross-domain text-to-SQL dataset in an +Arabic dialect. It consists of 9,428 NLQ-SQL pairs across 69 databases in +various domains. Along with SQL-related challenges such as long schemas, dirty +values, and complex queries, our dataset also incorporates the complexities of +the Moroccan dialect, which is known for its diverse source languages, numerous +borrowed words, and unique expressions. This demonstrates that our dataset will +be a valuable contribution to both the text-to-SQL community and the +development of resources for low-resource languages. + +**Decision Explanation:** Original decision: REJECT +The paper does not explicitly mention experimental results with quantitative metrics (failing Criteria 2) or comparisons with state-of-the-art methods (failing Criteria 3). While it addresses practical applications (Criterion 1) and real-world challenges (Criterion 5), it does not meet all mandatory criteria. + +--- + +## [Technical Report for the Forgotten-by-Design Project: Targeted + Obfuscation for Machine Learning](https://arxiv.org/abs/2501.11525v1) +**arXiv ID:** 2501.11525v1 + +**Abstract:** +> The right to privacy, enshrined in various human rights declarations, faces +new challenges in the age of artificial intelligence (AI). This paper explores +the concept of the Right to be Forgotten (RTBF) within AI systems, contrasting +it with traditional data erasure methods. We introduce Forgotten by Design, a +proactive approach to privacy preservation that integrates instance-specific +obfuscation techniques during the AI model training process. Unlike machine +unlearning, which modifies models post-training, our method prevents sensitive +data from being embedded in the first place. Using the LIRA membership +inference attack, we identify vulnerable data points and propose defenses that +combine additive gradient noise and weighting schemes. Our experiments on the +CIFAR-10 dataset demonstrate that our techniques reduce privacy risks by at +least an order of magnitude while maintaining model accuracy (at 95% +significance). Additionally, we present visualization methods for the +privacy-utility trade-off, providing a clear framework for balancing privacy +risk and model accuracy. This work contributes to the development of +privacy-preserving AI systems that align with human cognitive processes of +motivated forgetting, offering a robust framework for safeguarding sensitive +information and ensuring compliance with privacy regulations. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on privacy preservation (a responsible AI/ethics concern) and the Right to be Forgotten, which falls under the rejection criteria for responsible AI applications. While it meets criteria 2 and 3, it violates mandatory rejection rules. + +--- + +## [Training-free Ultra Small Model for Universal Sparse Reconstruction in + Compressed Sensing](https://arxiv.org/abs/2501.11592v1) +**arXiv ID:** 2501.11592v1 + +**Abstract:** +> Pre-trained large models attract widespread attention in recent years, but +they face challenges in applications that require high interpretability or have +limited resources, such as physical sensing, medical imaging, and +bioinformatics. Compressed Sensing (CS) is a well-proved theory that drives +many recent breakthroughs in these applications. However, as a typical +under-determined linear system, CS suffers from excessively long sparse +reconstruction times when using traditional iterative methods, particularly +with large-scale data. Current AI methods like deep unfolding fail to +substitute them because pre-trained models exhibit poor generality beyond their +training conditions and dataset distributions, or lack interpretability. +Instead of following the big model fervor, this paper proposes ultra-small +artificial neural models called coefficients learning (CL), enabling +training-free and rapid sparse reconstruction while perfectly inheriting the +generality and interpretability of traditional iterative methods, bringing new +feature of incorporating prior knowledges. In CL, a signal of length $n$ only +needs a minimal of $n$ trainable parameters. A case study model called CLOMP is +implemented for evaluation. Experiments are conducted on both synthetic and +real one-dimensional and two-dimensional signals, demonstrating significant +improvements in efficiency and accuracy. Compared to representative iterative +methods, CLOMP improves efficiency by 100 to 1000 folds for large-scale data. +Test results on eight diverse image datasets indicate that CLOMP improves +structural similarity index by 292%, 98%, 45% for sampling rates of 0.1, 0.3, +0.5, respectively. We believe this method can truly usher CS reconstruction +into the AI era, benefiting countless under-determined linear systems that rely +on sparse solution. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on compressed sensing and sparse reconstruction using small AI models but does not explicitly address Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it demonstrates quantitative improvements and comparisons with state-of-the-art methods (Criteria 2-3), it fails to meet the mandatory Criterion 1 (LLM practical applications). Additionally, medical imaging is mentioned as an application, which triggers a rejection under exclusion criteria. + +--- + +## [Fairness Testing through Extreme Value Theory](https://arxiv.org/abs/2501.11597v1) +**arXiv ID:** 2501.11597v1 + +**Abstract:** +> Data-driven software is increasingly being used as a critical component of +automated decision-support systems. Since this class of software learns its +logic from historical data, it can encode or amplify discriminatory practices. +Previous research on algorithmic fairness has focused on improving average-case +fairness. On the other hand, fairness at the extreme ends of the spectrum, +which often signifies lasting and impactful shifts in societal attitudes, has +received significantly less emphasis. + Leveraging the statistics of extreme value theory (EVT), we propose a novel +fairness criterion called extreme counterfactual discrimination (ECD). This +criterion estimates the worst-case amounts of disadvantage in outcomes for +individuals solely based on their memberships in a protected group. Utilizing +tools from search-based software engineering and generative AI, we present a +randomized algorithm that samples a statistically significant set of points +from the tail of ML outcome distributions even if the input dataset lacks a +sufficient number of relevant samples. + We conducted several experiments on four ML models (deep neural networks, +logistic regression, and random forests) over 10 socially relevant tasks from +the literature on algorithmic fairness. First, we evaluate the generative AI +methods and find that they generate sufficient samples to infer valid EVT +distribution in 95% of cases. Remarkably, we found that the prevalent bias +mitigators reduce the average-case discrimination but increase the worst-case +discrimination significantly in 5% of cases. We also observed that even the +tail-aware mitigation algorithm -- MiniMax-Fairness -- increased the worst-case +discrimination in 30% of cases. We propose a novel ECD-based mitigator that +improves fairness in the tail in 90% of cases with no degradation of the +average-case discrimination. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on algorithmic fairness and social harm in ML models, which falls under the rejection criteria for social applications. While it meets criteria 2 (quantitative results) and 3 (SOTA comparison), it violates the mandatory rejection rule regarding social applications. + +--- + +## [Noise-Agnostic Multitask Whisper Training for Reducing False Alarm + Errors in Call-for-Help Detection](https://arxiv.org/abs/2501.11631v1) +**arXiv ID:** 2501.11631v1 + +**Abstract:** +> Keyword spotting is often implemented by keyword classifier to the encoder in +acoustic models, enabling the classification of predefined or open vocabulary +keywords. Although keyword spotting is a crucial task in various applications +and can be extended to call-for-help detection in emergencies, however, the +previous method often suffers from scalability limitations due to retraining +required to introduce new keywords or adapt to changing contexts. We explore a +simple yet effective approach that leverages off-the-shelf pretrained ASR +models to address these challenges, especially in call-for-help detection +scenarios. Furthermore, we observed a substantial increase in false alarms when +deploying call-for-help detection system in real-world scenarios due to noise +introduced by microphones or different environments. To address this, we +propose a novel noise-agnostic multitask learning approach that integrates a +noise classification head into the ASR encoder. Our method enhances the model's +robustness to noisy environments, leading to a significant reduction in false +alarms and improved overall call-for-help performance. Despite the added +complexity of multitask learning, our approach is computationally efficient and +provides a promising solution for call-for-help detection in real-world +scenarios. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on acoustic models (ASR) and noise robustness for call-for-help detection but does not explicitly address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it includes experimental results and addresses real-world challenges, it fails to meet mandatory criteria 1 (Practical Applications of LLMs) and does not sufficiently align with the required LLM-focused scope. + +--- + +## [Biomedical Knowledge Graph: A Survey of Domains, Tasks, and Real-World + Applications](https://arxiv.org/abs/2501.11632v1) +**arXiv ID:** 2501.11632v1 + +**Abstract:** +> Biomedical knowledge graphs (BKGs) have emerged as powerful tools for +organizing and leveraging the vast and complex data found across the biomedical +field. Yet, current reviews of BKGs often limit their scope to specific domains +or methods, overlooking the broader landscape and the rapid technological +progress reshaping it. In this survey, we address this gap by offering a +systematic review of BKGs from three core perspectives: domains, tasks, and +applications. We begin by examining how BKGs are constructed from diverse data +sources, including molecular interactions, pharmacological datasets, and +clinical records. Next, we discuss the essential tasks enabled by BKGs, +focusing on knowledge management, retrieval, reasoning, and interpretation. +Finally, we highlight real-world applications in precision medicine, drug +discovery, and scientific research, illustrating the translational impact of +BKGs across multiple sectors. By synthesizing these perspectives into a unified +framework, this survey not only clarifies the current state of BKG research but +also establishes a foundation for future exploration, enabling both innovative +methodological advances and practical implementations. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on biomedical applications of knowledge graphs, which falls under the medical domain exclusion criteria. While it discusses real-world applications and knowledge graphs, the core subject violates the mandatory rejection rule for medical AI applications. + +--- + +## [Transformer Vibration Forecasting for Advancing Rail Safety and + Maintenance 4.0](https://arxiv.org/abs/2501.11730v1) +**arXiv ID:** 2501.11730v1 + +**Abstract:** +> Maintaining railway axles is critical to preventing severe accidents and +financial losses. The railway industry is increasingly interested in advanced +condition monitoring techniques to enhance safety and efficiency, moving beyond +traditional periodic inspections toward Maintenance 4.0. + This study introduces a robust Deep Autoregressive solution that integrates +seamlessly with existing systems to avert mechanical failures. Our approach +simulates and predicts vibration signals under various conditions and fault +scenarios, improving dataset robustness for more effective detection systems. +These systems can alert maintenance needs, preventing accidents preemptively. +We use experimental vibration signals from accelerometers on train axles. + Our primary contributions include a transformer model, ShaftFormer, designed +for processing time series data, and an alternative model incorporating +spectral methods and enhanced observation models. Simulating vibration signals +under diverse conditions mitigates the high cost of obtaining experimental +signals for all scenarios. Given the non-stationary nature of railway vibration +signals, influenced by speed and load changes, our models address these +complexities, offering a powerful tool for predictive maintenance in the rail +industry. + +**Decision Explanation:** Original decision: REJECT +The paper meets Practical Applications (rail maintenance) but lacks explicit mention of quantitative metrics (criteria 2) and comparisons with state-of-the-art methods (criteria 3). While it proposes a transformer model for time-series data, the abstract does not demonstrate performance improvements against existing techniques or provide concrete experimental metrics required for mandatory inclusion. + +--- + +## [SILO: Solving Inverse Problems with Latent Operators](https://arxiv.org/abs/2501.11746v1) +**arXiv ID:** 2501.11746v1 + +**Abstract:** +> Consistent improvement of image priors over the years has led to the +development of better inverse problem solvers. Diffusion models are the +newcomers to this arena, posing the strongest known prior to date. Recently, +such models operating in a latent space have become increasingly predominant +due to their efficiency. In recent works, these models have been applied to +solve inverse problems. Working in the latent space typically requires multiple +applications of an Autoencoder during the restoration process, which leads to +both computational and restoration quality challenges. In this work, we propose +a new approach for handling inverse problems with latent diffusion models, +where a learned degradation function operates within the latent space, +emulating a known image space degradation. Usage of the learned operator +reduces the dependency on the Autoencoder to only the initial and final steps +of the restoration process, facilitating faster sampling and superior +restoration quality. We demonstrate the effectiveness of our method on a +variety of image restoration tasks and datasets, achieving significant +improvements over prior art. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on latent diffusion models for image restoration tasks, which falls under video/image processing (a rejection criterion). It does not address Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI as required by the mandatory criteria. + +--- + +## [Is logical analysis performed by transformers taking place in + self-attention or in the fully connected part?](https://arxiv.org/abs/2501.11765v1) +**arXiv ID:** 2501.11765v1 + +**Abstract:** +> Transformers architecture apply self-attention to tokens represented as +vectors, before a fully connected (neuronal network) layer. These two parts can +be layered many times. Traditionally, self-attention is seen as a mechanism for +aggregating information before logical operations are performed by the fully +connected layer. In this paper, we show, that quite counter-intuitively, the +logical analysis can also be performed within the self-attention. For this we +implement a handcrafted single-level encoder layer which performs the logical +analysis within self-attention. We then study the scenario in which a one-level +transformer model undergoes self-learning using gradient descent. We +investigate whether the model utilizes fully connected layers or self-attention +mechanisms for logical analysis when it has the choice. Given that gradient +descent can become stuck at undesired zeros, we explicitly calculate these +unwanted zeros and find ways to avoid them. We do all this in the context of +predicting grammatical category pairs of adjacent tokens in a text. We believe +that our findings have broader implications for understanding the potential +logical operations performed by self-attention. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet all mandatory criteria. While it includes experimental results (Criterion 2) and describes methodology (Criterion 4), it lacks explicit focus on practical applications of LLMs (Criterion 1) and does not compare with state-of-the-art techniques (Criterion 3). The analysis centers on transformer internals (self-attention vs. FC layers) for grammatical tasks, without addressing real-world LLM applications like knowledge graphs, RAG, or agentic AI. + +--- + +## [Toward Effective Digraph Representation Learning: A Magnetic Adaptive + Propagation based Approach](https://arxiv.org/abs/2501.11817v1) +**arXiv ID:** 2501.11817v1 + +**Abstract:** +> The $q$-parameterized magnetic Laplacian serves as the foundation of directed +graph (digraph) convolution, enabling this kind of digraph neural network +(MagDG) to encode node features and structural insights by complex-domain +message passing. As a generalization of undirected methods, MagDG shows +superior capability in modeling intricate web-scale topology. Despite the great +success achieved by existing MagDGs, limitations still exist: (1) Hand-crafted +$q$: The performance of MagDGs depends on selecting an appropriate +$q$-parameter to construct suitable graph propagation equations in the complex +domain. This parameter tuning, driven by downstream tasks, limits model +flexibility and significantly increases manual effort. (2) Coarse Message +Passing: Most approaches treat all nodes with the same complex-domain +propagation and aggregation rules, neglecting their unique digraph contexts. +This oversight results in sub-optimal performance. To address the above issues, +we propose two key techniques: (1) MAP is crafted to be a plug-and-play +complex-domain propagation optimization strategy in the context of digraph +learning, enabling seamless integration into any MagDG to improve predictions +while enjoying high running efficiency. (2) MAP++ is a new digraph learning +framework, further incorporating a learnable mechanism to achieve adaptively +edge-wise propagation and node-wise aggregation in the complex domain for +better performance. Extensive experiments on 12 datasets demonstrate that MAP +enjoys flexibility for it can be incorporated with any MagDG, and scalability +as it can deal with web-scale digraphs. MAP++ achieves SOTA predictive +performance on 4 different downstream tasks. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on directed graph representation learning but does not explicitly address Large Language Models (LLMs) or their applications in areas like knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (SOTA comparison), it fails to satisfy the mandatory criterion 1 (LLM practical applications). + +--- + +## [Toward Scalable Graph Unlearning: A Node Influence Maximization based + Approach](https://arxiv.org/abs/2501.11823v1) +**arXiv ID:** 2501.11823v1 + +**Abstract:** +> Machine unlearning, as a pivotal technology for enhancing model robustness +and data privacy, has garnered significant attention in prevalent web mining +applications, especially in thriving graph-based scenarios. However, most +existing graph unlearning (GU) approaches face significant challenges due to +the intricate interactions among web-scale graph elements during the model +training: (1) The gradient-driven node entanglement hinders the complete +knowledge removal in response to unlearning requests; (2) The billion-level +graph elements in the web scenarios present inevitable scalability issues. To +break the above limitations, we open up a new perspective by drawing a +connection between GU and conventional social influence maximization. To this +end, we propose Node Influence Maximization (NIM) through the decoupled +influence propagation model and fine-grained influence function in a scalable +manner, which is crafted to be a plug-and-play strategy to identify potential +nodes affected by unlearning entities. This approach enables offline execution +independent of GU, allowing it to be seamlessly integrated into most GU methods +to improve their unlearning performance. Based on this, we introduce Scalable +Graph Unlearning (SGU) as a new fine-tuned framework, which balances the +forgetting and reasoning capability of the unlearned model by entity-specific +optimizations. Extensive experiments on 14 datasets, including large-scale +ogbn-papers100M, have demonstrated the effectiveness of our approach. +Specifically, NIM enhances the forgetting capability of most GU methods, while +SGU achieves comprehensive SOTA performance and maintains scalability. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on graph unlearning for data privacy and model robustness in web mining applications but does not explicitly address Large Language Models (LLMs) or their practical applications (e.g., knowledge graphs, RAG, agentic AI). While it meets Criteria 2 and 3 with experimental results and SOTA comparisons, it fails Criterion 1 (LLM focus). Additionally, none of the secondary criteria (4,5) are clearly addressed. The work appears centered on graph neural networks rather than LLMs. + +--- + +## [Data-driven Detection and Evaluation of Damages in Concrete Structures: + Using Deep Learning and Computer Vision](https://arxiv.org/abs/2501.11836v1) +**arXiv ID:** 2501.11836v1 + +**Abstract:** +> Structural integrity is vital for maintaining the safety and longevity of +concrete infrastructures such as bridges, tunnels, and walls. Traditional +methods for detecting damages like cracks and spalls are labor-intensive, +time-consuming, and prone to human error. To address these challenges, this +study explores advanced data-driven techniques using deep learning for +automated damage detection and analysis. Two state-of-the-art instance +segmentation models, YOLO-v7 instance segmentation and Mask R-CNN, were +evaluated using a dataset comprising 400 images, augmented to 10,995 images +through geometric and color-based transformations to enhance robustness. The +models were trained and validated using a dataset split into 90% training set, +validation and test set 10%. Performance metrics such as precision, recall, +mean average precision (mAP@0.5), and frames per second (FPS) were used for +evaluation. YOLO-v7 achieved a superior mAP@0.5 of 96.1% and processed 40 FPS, +outperforming Mask R-CNN, which achieved a mAP@0.5 of 92.1% with a slower +processing speed of 18 FPS. The findings recommend YOLO-v7 instance +segmentation model for real-time, high-speed structural health monitoring, +while Mask R-CNN is better suited for detailed offline assessments. This study +demonstrates the potential of deep learning to revolutionize infrastructure +maintenance, offering a scalable and efficient solution for automated damage +detection. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on computer vision and deep learning for structural damage detection but does not mention Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. It fails to meet the mandatory Criterion 1 (Practical Applications of LLMs) and does not align with the required LLM-focused scope. + +--- + +## [Supervised Learning for Analog and RF Circuit Design: Benchmarks and + Comparative Insights](https://arxiv.org/abs/2501.11839v1) +**arXiv ID:** 2501.11839v1 + +**Abstract:** +> Automating analog and radio-frequency (RF) circuit design using machine +learning (ML) significantly reduces the time and effort required for parameter +optimization. This study explores supervised ML-based approaches for designing +circuit parameters from performance specifications across various circuit +types, including homogeneous and heterogeneous designs. By evaluating diverse +ML models, from neural networks like transformers to traditional methods like +random forests, we identify the best-performing models for each circuit. Our +results show that simpler circuits, such as low-noise amplifiers, achieve +exceptional accuracy with mean relative errors as low as 0.3% due to their +linear parameter-performance relationships. In contrast, complex circuits, like +power amplifiers and voltage-controlled oscillators, present challenges due to +their non-linear interactions and larger design spaces. For heterogeneous +circuits, our approach achieves an 88% reduction in errors with increased +training data, with the receiver achieving a mean relative error as low as +0.23%, showcasing the scalability and accuracy of the proposed methodology. +Additionally, we provide insights into model strengths, with transformers +excelling in capturing non-linear mappings and k-nearest neighbors performing +robustly in moderately linear parameter spaces, especially in heterogeneous +circuits with larger datasets. This work establishes a foundation for extending +ML-driven design automation, enabling more efficient and scalable circuit +design workflows. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on supervised ML for circuit design without explicit mention of Large Language Models (LLMs) or their applications in knowledge graphs, RAG, or agentic AI. While it meets Criteria 2 (quantitative results) and 3 (comparisons with SOTA), it fails Criterion 1 (LLM practical applications), violating the requirement to meet ALL mandatory criteria. + +--- + +## [Scalable Whole Slide Image Representation Using K-Mean Clustering and + Fisher Vector Aggregation](https://arxiv.org/abs/2501.12085v1) +**arXiv ID:** 2501.12085v1 + +**Abstract:** +> Whole slide images (WSIs) are high-resolution, gigapixel sized images that +pose significant computational challenges for traditional machine learning +models due to their size and heterogeneity.In this paper, we present a scalable +and efficient methodology for WSI classification by leveraging patch-based +feature extraction, clustering, and Fisher vector encoding. Initially, WSIs are +divided into fixed size patches, and deep feature embeddings are extracted from +each patch using a pre-trained convolutional neural network (CNN). These +patch-level embeddings are subsequently clustered using K-means clustering, +where each cluster aggregates semantically similar regions of the WSI. To +effectively summarize each cluster, Fisher vector representations are computed +by modeling the distribution of patch embeddings in each cluster as a +parametric Gaussian mixture model (GMM). The Fisher vectors from each cluster +are concatenated into a high-dimensional feature vector, creating a compact and +informative representation of the entire WSI. This feature vector is then used +by a classifier to predict the WSI's diagnostic label. Our method captures +local and global tissue structures and yields robust performance for +large-scale WSI classification, demonstrating superior accuracy and scalability +compared to other approaches. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications (whole slide image classification for diagnostic labels), which falls under the mandatory rejection criteria. + +--- + +## [Proxies for Distortion and Consistency with Applications for Real-World + Image Restoration](https://arxiv.org/abs/2501.12102v1) +**arXiv ID:** 2501.12102v1 + +**Abstract:** +> Real-world image restoration deals with the recovery of images suffering from +an unknown degradation. This task is typically addressed while being given only +degraded images, without their corresponding ground-truth versions. In this +hard setting, designing and evaluating restoration algorithms becomes highly +challenging. This paper offers a suite of tools that can serve both the design +and assessment of real-world image restoration algorithms. Our work starts by +proposing a trained model that predicts the chain of degradations a given +real-world measured input has gone through. We show how this estimator can be +used to approximate the consistency -- the match between the measurements and +any proposed recovered image. We also use this estimator as a guiding force for +the design of a simple and highly-effective plug-and-play real-world image +restoration algorithm, leveraging a pre-trained diffusion-based image prior. +Furthermore, this work proposes no-reference proxy measures of MSE and LPIPS, +which, without access to the ground-truth images, allow ranking of real-world +image restoration algorithms according to their (approximate) MSE and LPIPS. +The proposed suite provides a versatile, first of its kind framework for +evaluating and comparing blind image restoration algorithms in real-world +scenarios. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on real-world image restoration tools and metrics, which falls under the rejection criteria for primarily addressing video/image processing. It does not mention LLMs, knowledge graphs, RAG, or agentic AI applications, failing to meet mandatory criterion 1 (Practical Applications of LLMs). + +--- + +## [Efficient PINNs: Multi-Head Unimodular Regularization of the Solutions + Space](https://arxiv.org/abs/2501.12116v1) +**arXiv ID:** 2501.12116v1 + +**Abstract:** +> We present a machine learning framework to facilitate the solution of +nonlinear multiscale differential equations and, especially, inverse problems +using Physics-Informed Neural Networks (PINNs). This framework is based on what +is called multihead (MH) training, which involves training the network to learn +a general space of all solutions for a given set of equations with certain +variability, rather than learning a specific solution of the system. This setup +is used with a second novel technique that we call Unimodular Regularization +(UR) of the latent space of solutions. We show that the multihead approach, +combined with the regularization, significantly improves the efficiency of +PINNs by facilitating the transfer learning process thereby enabling the +finding of solutions for nonlinear, coupled, and multiscale differential +equations. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on Physics-Informed Neural Networks (PINNs) for solving differential equations and inverse problems but does not mention LLMs, knowledge graphs, RAG, or agentic AI. While it demonstrates methodology improvements and experimental validation, it lacks direct relevance to LLM applications as required by mandatory criteria 1. + +--- + +## [An End-to-End Approach for Korean Wakeword Systems with Speaker + Authentication](https://arxiv.org/abs/2501.12194v1) +**arXiv ID:** 2501.12194v1 + +**Abstract:** +> Wakeword detection plays a critical role in enabling AI assistants to listen +to user voices and interact effectively. However, for languages other than +English, there is a significant lack of pre-trained wakeword models. +Additionally, systems that merely determine the presence of a wakeword can pose +serious privacy concerns. In this paper, we propose an end-to-end approach that +trains wakewords for Non-English languages, particulary Korean, and uses this +to develop a Voice Authentication model to protect user privacy. Our +implementation employs an open-source platform OpenWakeWord, which performs +wakeword detection using an FCN (Fully-Connected Network) architecture. Once a +wakeword is detected, our custom-developed code calculates cosine similarity +for robust user authentication. Experimental results demonstrate the +effectiveness of our approach, achieving a 16.79% and a 6.6% Equal Error Rate +(EER) each in the Wakeword Detection and the Voice Authentication. These +findings highlight the model's potential in providing secure and accurate +wakeword detection and authentication for Korean users. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on wakeword detection and voice authentication using FCN architecture and cosine similarity, but does not demonstrate applications of Large Language Models (LLMs) in areas like knowledge graphs, RAG, or agentic AI. While it provides quantitative metrics (EER), it fails to meet the mandatory criteria 1 (LLM applications) and 3 (SOTA comparison). + +--- + +## [Strong phonon-mediated high temperature superconductivity in + Li$_2$AuH$_6$ under ambient pressure](https://arxiv.org/abs/2501.12222v1) +**arXiv ID:** 2501.12222v1 + +**Abstract:** +> We used our developed AI search engine~(InvDesFlow) to perform extensive +investigations regarding ambient stable superconducting hydrides. A cubic +structure Li$_2$AuH$_6$ with Au-H octahedral motifs is identified to be a +candidate. After performing thermodynamical analysis, we provide a feasible +route to experimentally synthesize this material via the known LiAu and LiH +compounds under ambient pressure. The further first-principles calculations +suggest that Li$_2$AuH$_6$ shows a high superconducting transition temperature +($T_c$) $\sim$ 140 K under ambient pressure. The H-1$s$ electrons strongly +couple with phonon modes of vibrations of Au-H octahedrons as well as +vibrations of Li atoms, where the latter is not taken seriously in other +previously similar cases. Hence, different from previous claims of searching +metallic covalent bonds to find high-$T_c$ superconductors, we emphasize here +the importance of those phonon modes with strong electron-phonon coupling +(EPC). And we suggest that one can intercalate atoms into binary or ternary +hydrides to introduce more potential phonon modes with strong EPC, which is an +effective approach to find high-$T_c$ superconductors within multicomponent +compounds. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on AI-assisted material discovery for superconductors but does not explicitly involve Large Language Models (LLMs) in its methodology or applications. It fails to meet the mandatory Criterion 1 (Practical Applications of LLMs) and does not address LLM-related challenges, agentic AI, or other LLM-specific domains outlined in the criteria. + +--- + +## [CBVLM: Training-free Explainable Concept-based Large Vision Language + Models for Medical Image Classification](https://arxiv.org/abs/2501.12266v1) +**arXiv ID:** 2501.12266v1 + +**Abstract:** +> The main challenges limiting the adoption of deep learning-based solutions in +medical workflows are the availability of annotated data and the lack of +interpretability of such systems. Concept Bottleneck Models (CBMs) tackle the +latter by constraining the final disease prediction on a set of predefined and +human-interpretable concepts. However, the increased interpretability achieved +through these concept-based explanations implies a higher annotation burden. +Moreover, if a new concept needs to be added, the whole system needs to be +retrained. Inspired by the remarkable performance shown by Large +Vision-Language Models (LVLMs) in few-shot settings, we propose a simple, yet +effective, methodology, CBVLM, which tackles both of the aforementioned +challenges. First, for each concept, we prompt the LVLM to answer if the +concept is present in the input image. Then, we ask the LVLM to classify the +image based on the previous concept predictions. Moreover, in both stages, we +incorporate a retrieval module responsible for selecting the best examples for +in-context learning. By grounding the final diagnosis on the predicted +concepts, we ensure explainability, and by leveraging the few-shot capabilities +of LVLMs, we drastically lower the annotation cost. We validate our approach +with extensive experiments across four medical datasets and twelve LVLMs (both +generic and medical) and show that CBVLM consistently outperforms CBMs and +task-specific supervised methods without requiring any training and using just +a few annotated examples. More information on our project page: +https://cristianopatricio.github.io/CBVLM/. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (medical image classification), which is an explicit rejection criterion according to the guidelines. While it meets some technical criteria (experimental results, SOTA comparisons), the medical focus mandates rejection regardless of other merits. + +--- + +## [Implementation of an Asymmetric Adjusted Activation Function for Class + Imbalance Credit Scoring](https://arxiv.org/abs/2501.12285v1) +**arXiv ID:** 2501.12285v1 + +**Abstract:** +> Credit scoring is a systematic approach to evaluate a borrower's probability +of default (PD) on a bank loan. The data associated with such scenarios are +characteristically imbalanced, complicating binary classification owing to the +often-underestimated cost of misclassification during the classifier's learning +process. Considering the high imbalance ratio (IR) of these datasets, we +introduce an innovative yet straightforward optimized activation function by +incorporating an IR-dependent asymmetric adjusted factor embedded Sigmoid +activation function (ASIG). The embedding of ASIG makes the sensitive margin of +the Sigmoid function auto-adjustable, depending on the imbalance nature of the +datasets distributed, thereby giving the activation function an asymmetric +characteristic that prevents the underrepresentation of the minority class +(positive samples) during the classifier's learning process. The experimental +results show that the ASIG-embedded-classifier outperforms traditional +classifiers on datasets across wide-ranging IRs in the downstream +credit-scoring task. The algorithm also shows robustness and stability, even +when the IR is ultra-high. Therefore, the algorithm provides a competitive +alternative in the financial industry, especially in credit scoring, possessing +the ability to effectively process highly imbalanced distribution data. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on credit scoring with an activation function adaptation for class imbalance but does not mention Large Language Models (LLMs) or address criteria 1 (Practical Applications of LLMs), 3 (Comparison with SOTA LLM techniques), or 4/5 (LLM methodology/real-world challenges). It primarily addresses financial data classification without LLM relevance. + +--- + +## [Regressor-Guided Image Editing Regulates Emotional Response to Reduce + Online Engagement](https://arxiv.org/abs/2501.12289v1) +**arXiv ID:** 2501.12289v1 + +**Abstract:** +> Emotions are known to mediate the relationship between users' content +consumption and their online engagement, with heightened emotional intensity +leading to increased engagement. Building on this insight, we propose three +regressor-guided image editing approaches aimed at diminishing the emotional +impact of images. These include (i) a parameter optimization approach based on +global image transformations known to influence emotions, (ii) an optimization +approach targeting the style latent space of a generative adversarial network, +and (iii) a diffusion-based approach employing classifier guidance and +classifier-free guidance. Our findings demonstrate that approaches can +effectively alter the emotional properties of images while maintaining high +visual quality. Optimization-based methods primarily adjust low-level +properties like color hues and brightness, whereas the diffusion-based approach +introduces semantic changes, such as altering appearance or facial expressions. +Notably, results from a behavioral study reveal that only the diffusion-based +approach successfully elicits changes in viewers' emotional responses while +preserving high perceived image quality. In future work, we will investigate +the impact of these image adaptations on internet user behavior. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on image editing to influence emotional responses and online engagement, which falls under social applications of AI (explicit rejection criterion). It does not mention LLMs, knowledge graphs, RAG, or agentic AI, failing to meet mandatory criteria 1 and 3. While it includes experimental results (criterion 2), the social application focus triggers rejection. + +--- + +## [Test-time regression: a unifying framework for designing sequence models + with associative memory](https://arxiv.org/abs/2501.12352v1) +**arXiv ID:** 2501.12352v1 + +**Abstract:** +> Sequences provide a remarkably general way to represent and process +information. This powerful abstraction has placed sequence modeling at the +center of modern deep learning applications, inspiring numerous architectures +from transformers to recurrent networks. While this fragmented development has +yielded powerful models, it has left us without a unified framework to +understand their fundamental similarities and explain their effectiveness. We +present a unifying framework motivated by an empirical observation: effective +sequence models must be able to perform associative recall. Our key insight is +that memorizing input tokens through an associative memory is equivalent to +performing regression at test-time. This regression-memory correspondence +provides a framework for deriving sequence models that can perform associative +recall, offering a systematic lens to understand seemingly ad-hoc architectural +choices. We show numerous recent architectures -- including linear attention +models, their gated variants, state-space models, online learners, and softmax +attention -- emerge naturally as specific approaches to test-time regression. +Each architecture corresponds to three design choices: the relative importance +of each association, the regressor function class, and the optimization +algorithm. This connection leads to new understanding: we provide theoretical +justification for QKNorm in softmax attention, and we motivate higher-order +generalizations of softmax attention. Beyond unification, our work unlocks +decades of rich statistical tools that can guide future development of more +powerful yet principled sequence models. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on theoretical unification of sequence models through test-time regression concepts but does not explicitly address practical LLM applications (C1), lacks mention of experimental results with quantitative metrics (C2), and does not demonstrate direct performance comparisons with state-of-the-art methods (C3). While novel, it fails to meet all mandatory criteria for real-world LLM applications. + +--- + +## [DARB-Splatting: Generalizing Splatting with Decaying Anisotropic Radial + Basis Functions](https://arxiv.org/abs/2501.12369v1) +**arXiv ID:** 2501.12369v1 + +**Abstract:** +> Splatting-based 3D reconstruction methods have gained popularity with the +advent of 3D Gaussian Splatting, efficiently synthesizing high-quality novel +views. These methods commonly resort to using exponential family functions, +such as the Gaussian function, as reconstruction kernels due to their +anisotropic nature, ease of projection, and differentiability in rasterization. +However, the field remains restricted to variations within the exponential +family, leaving generalized reconstruction kernels largely underexplored, +partly due to the lack of easy integrability in 3D to 2D projections. In this +light, we show that a class of decaying anisotropic radial basis functions +(DARBFs), which are non-negative functions of the Mahalanobis distance, +supports splatting by approximating the Gaussian function's closed-form +integration advantage. With this fresh perspective, we demonstrate up to 34% +faster convergence during training and a 15% reduction in memory consumption +across various DARB reconstruction kernels, while maintaining comparable PSNR, +SSIM, and LPIPS results. We will make the code available. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on 3D reconstruction methods using splatting techniques but does not address practical applications of Large Language Models (LLMs), knowledge graphs, RAG, or agentic AI. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails the mandatory criterion 1 (LLM applications). Additionally, it does not align with the required scope for inclusion. + +--- + +## [MMVU: Measuring Expert-Level Multi-Discipline Video Understanding](https://arxiv.org/abs/2501.12380v1) +**arXiv ID:** 2501.12380v1 + +**Abstract:** +> We introduce MMVU, a comprehensive expert-level, multi-discipline benchmark +for evaluating foundation models in video understanding. MMVU includes 3,000 +expert-annotated questions spanning 27 subjects across four core disciplines: +Science, Healthcare, Humanities & Social Sciences, and Engineering. Compared to +prior benchmarks, MMVU features three key advancements. First, it challenges +models to apply domain-specific knowledge and perform expert-level reasoning to +analyze specialized-domain videos, moving beyond the basic visual perception +typically assessed in current video benchmarks. Second, each example is +annotated by human experts from scratch. We implement strict data quality +controls to ensure the high quality of the dataset. Finally, each example is +enriched with expert-annotated reasoning rationals and relevant domain +knowledge, facilitating in-depth analysis. We conduct an extensive evaluation +of 32 frontier multimodal foundation models on MMVU. The latest +System-2-capable models, o1 and Gemini 2.0 Flash Thinking, achieve the highest +performance among the tested models. However, they still fall short of matching +human expertise. Through in-depth error analyses and case studies, we offer +actionable insights for future advancements in expert-level, +knowledge-intensive video understanding for specialized domains. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing through a specialized video understanding benchmark (MMVU), which falls under explicit rejection criteria. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), the core focus on video understanding violates mandatory rejection rules. + +--- + +## [Physics of Skill Learning](https://arxiv.org/abs/2501.12391v1) +**arXiv ID:** 2501.12391v1 + +**Abstract:** +> We aim to understand physics of skill learning, i.e., how skills are learned +in neural networks during training. We start by observing the Domino effect, +i.e., skills are learned sequentially, and notably, some skills kick off +learning right after others complete learning, similar to the sequential fall +of domino cards. To understand the Domino effect and relevant behaviors of +skill learning, we take physicists' approach of abstraction and simplification. +We propose three models with varying complexities -- the Geometry model, the +Resource model, and the Domino model, trading between reality and simplicity. +The Domino effect can be reproduced in the Geometry model, whose resource +interpretation inspires the Resource model, which can be further simplified to +the Domino model. These models present different levels of abstraction and +simplification; each is useful to study some aspects of skill learning. The +Geometry model provides interesting insights into neural scaling laws and +optimizers; the Resource model sheds light on the learning dynamics of +compositional tasks; the Domino model reveals the benefits of modularity. These +models are not only conceptually interesting -- e.g., we show how Chinchilla +scaling laws can emerge from the Geometry model, but also are useful in +practice by inspiring algorithmic development -- e.g., we show how simple +algorithmic changes, motivated by these toy models, can speed up the training +of deep learning models. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on understanding skill learning dynamics in neural networks with theoretical models but does not explicitly address Large Language Models (LLMs), practical applications in areas like knowledge graphs/RAG/agentic AI, or comparisons with state-of-the-art methods. It lacks clear alignment with mandatory criteria 1 and 3, and its relevance to LLM-specific real-world applications remains unclear. + +--- + +## [Learning segmentation from point trajectories](https://arxiv.org/abs/2501.12392v1) +**arXiv ID:** 2501.12392v1 + +**Abstract:** +> We consider the problem of segmenting objects in videos based on their motion +and no other forms of supervision. Prior work has often approached this problem +by using the principle of common fate, namely the fact that the motion of +points that belong to the same object is strongly correlated. However, most +authors have only considered instantaneous motion from optical flow. In this +work, we present a way to train a segmentation network using long-term point +trajectories as a supervisory signal to complement optical flow. The key +difficulty is that long-term motion, unlike instantaneous motion, is difficult +to model -- any parametric approximation is unlikely to capture complex motion +patterns over long periods of time. We instead draw inspiration from subspace +clustering approaches, proposing a loss function that seeks to group the +trajectories into low-rank matrices where the motion of object points can be +approximately explained as a linear combination of other point tracks. Our +method outperforms the prior art on motion-based segmentation, which shows the +utility of long-term motion and the effectiveness of our formulation. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing (motion-based segmentation in videos), which is an explicit rejection criterion. While it meets criteria 2 (experimental results) and 3 (comparison with SOTA), it fails the mandatory practical applications requirement for LLMs and violates the video processing exclusion rule. + +--- + +## [Unsupervised Learning in Echo State Networks for Input Reconstruction](https://arxiv.org/abs/2501.11409v1) +**arXiv ID:** 2501.11409v1 + +**Abstract:** +> Conventional echo state networks (ESNs) require supervised learning to train +the readout layer, using the desired outputs as training data. In this study, +we focus on input reconstruction (IR), which refers to training the readout +layer to reproduce the input time series in its output. We reformulate the +learning algorithm of the ESN readout layer to perform IR using unsupervised +learning (UL). By conducting theoretical analysis and numerical experiments, we +demonstrate that IR in ESNs can be effectively implemented under realistic +conditions without explicitly using the desired outputs as training data; in +this way, UL is enabled. Furthermore, we demonstrate that applications relying +on IR, such as dynamical system replication and noise filtering, can be +reformulated within the UL framework. Our findings establish a theoretically +sound and universally applicable IR formulation, along with its related tasks +in ESNs. This work paves the way for novel predictions and highlights +unresolved theoretical challenges in ESNs, particularly in the context of +time-series processing methods and computational models of the brain. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on Echo State Networks (ESNs), not Large Language Models (LLMs), and does not address LLM-related applications like knowledge graphs, RAG, or agentic AI. None of the mandatory criteria (Practical Applications of LLMs, Experimental Results related to LLMs, SOTA comparisons involving LLMs) are met. While methodology details are provided, they pertain to ESNs rather than LLM utilization in real-world tasks. + +--- + +## [Improving thermal state preparation of Sachdev-Ye-Kitaev model with + reinforcement learning on quantum hardware](https://arxiv.org/abs/2501.11454v1) +**arXiv ID:** 2501.11454v1 + +**Abstract:** +> The Sachdev-Ye-Kitaev (SYK) model, known for its strong quantum correlations +and chaotic behavior, serves as a key platform for quantum gravity studies. +However, variationally preparing thermal states on near-term quantum processors +for large systems (N>12, where N is the number of Majorana fermions) presents a +significant challenge due to the rapid growth in the complexity of +parameterized quantum circuits. This paper addresses this challenge by +integrating reinforcement learning (RL) with convolutional neural networks, +employing an iterative approach to optimize the quantum circuit and its +parameters. The refinement process is guided by a composite reward signal +derived from entropy and the expectation values of the SYK Hamiltonian. This +approach reduces the number of CNOT gates by two orders of magnitude for +systems N>10 compared to traditional methods like first-order Trotterization. +We demonstrate the effectiveness of the RL framework in both noiseless and +noisy quantum hardware environments, maintaining high accuracy in thermal state +preparation. This work contributes to the advancement of a scalable, RL-based +framework with applications for computations of thermal out-of-time-order +correlators in quantum many-body systems and quantum gravity studies on +near-term quantum hardware. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on quantum computing and reinforcement learning for thermal state preparation in the SYK model but does not address Large Language Models (LLMs) or their practical applications (failing mandatory criterion 1). While it meets criteria 2 (quantitative results) and 3 (comparison with SOTA), LLM relevance is absent. The topic does not fall under rejection domains but fails core LLM-focused requirements. + +--- + +## [With Great Backbones Comes Great Adversarial Transferability](https://arxiv.org/abs/2501.12275v1) +**arXiv ID:** 2501.12275v1 + +**Abstract:** +> Advances in self-supervised learning (SSL) for machine vision have improved +representation robustness and model performance, giving rise to pre-trained +backbones like \emph{ResNet} and \emph{ViT} models tuned with SSL methods such +as \emph{SimCLR}. Due to the computational and data demands of pre-training, +the utilization of such backbones becomes a strenuous necessity. However, +employing these backbones may inherit vulnerabilities to adversarial attacks. +While adversarial robustness has been studied under \emph{white-box} and +\emph{black-box} settings, the robustness of models tuned on pre-trained +backbones remains largely unexplored. Additionally, the role of tuning +meta-information in mitigating exploitation risks is unclear. This work +systematically evaluates the adversarial robustness of such models across +$20,000$ combinations of tuning meta-information, including fine-tuning +techniques, backbone families, datasets, and attack types. We propose using +proxy models to transfer attacks, simulating varying levels of target knowledge +by fine-tuning these proxies with diverse configurations. Our findings reveal +that proxy-based attacks approach the effectiveness of \emph{white-box} +methods, even with minimal tuning knowledge. We also introduce a naive +"backbone attack," leveraging only the backbone to generate adversarial +samples, which outperforms \emph{black-box} attacks and rivals \emph{white-box} +methods, highlighting critical risks in model-sharing practices. Finally, our +ablations reveal how increasing tuning meta-information impacts attack +transferability, measuring each meta-information combination. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on adversarial robustness in self-supervised learning for machine vision (ResNet/ViT), not Large Language Models (LLMs). It does not meet Criterion 1 (Practical Applications of LLMs) and addresses adversarial attacks in general model-sharing contexts rather than LLM-specific applications like RAG or agentic AI. While it meets Criteria 2-3 and 4-5, the mandatory LLM focus is absent. + +--- + +## [Goal-oriented Transmission Scheduling: Structure-guided DRL with a + Unified Dual On-policy and Off-policy Approach](https://arxiv.org/abs/2501.11921v1) +**arXiv ID:** 2501.11921v1 + +**Abstract:** +> Goal-oriented communications prioritize application-driven objectives over +data accuracy, enabling intelligent next-generation wireless systems. Efficient +scheduling in multi-device, multi-channel systems poses significant challenges +due to high-dimensional state and action spaces. We address these challenges by +deriving key structural properties of the optimal solution to the goal-oriented +scheduling problem, incorporating Age of Information (AoI) and channel states. +Specifically, we establish the monotonicity of the optimal state value function +(a measure of long-term system performance) w.r.t. channel states and prove its +asymptotic convexity w.r.t. AoI states. Additionally, we derive the +monotonicity of the optimal policy w.r.t. channel states, advancing the +theoretical framework for optimal scheduling. Leveraging these insights, we +propose the structure-guided unified dual on-off policy DRL (SUDO-DRL), a +hybrid algorithm that combines the stability of on-policy training with the +sample efficiency of off-policy methods. Through a novel structural property +evaluation framework, SUDO-DRL enables effective and scalable training, +addressing the complexities of large-scale systems. Numerical results show +SUDO-DRL improves system performance by up to 45% and reduces convergence time +by 40% compared to state-of-the-art methods. It also effectively handles +scheduling in much larger systems, where off-policy DRL fails and on-policy +benchmarks exhibit significant performance loss, demonstrating its scalability +and efficacy in goal-oriented communications. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on deep reinforcement learning (DRL) for communication scheduling but does not address Large Language Models (LLMs), failing to meet mandatory criteria 1 (Practical Applications of LLMs), 2 (LLM-related experimental results), and 3 (LLM methodology comparisons). + +--- + diff --git a/llm_processor/papers.md b/llm_processor/papers.md new file mode 100644 index 0000000..347793a --- /dev/null +++ b/llm_processor/papers.md @@ -0,0 +1,4196 @@ +# Accepted Papers + +## [Monte Carlo Tree Search for Comprehensive Exploration in LLM-Based + Automatic Heuristic Design](https://arxiv.org/abs/2501.08603v2) +**arXiv ID:** 2501.08603v2 + +**Abstract:** +> Handcrafting heuristics for solving complex planning tasks (e.g., NP-hard +combinatorial optimization (CO) problems) is a common practice but requires +extensive domain knowledge. Recently, Large Language Model (LLM)-based +automatic heuristics design (AHD) methods have shown promise in generating +high-quality heuristics without manual intervention. Existing LLM-based AHD +methods employ a population to maintain a fixed number of top-performing +LLM-generated heuristics and introduce evolutionary computation (EC) to enhance +the population iteratively. However, the population-based procedure brings +greedy properties, often resulting in convergence to local optima. Instead, to +more comprehensively explore the space of heuristics, we propose using Monte +Carlo Tree Search (MCTS) for LLM-based heuristic evolution while preserving all +LLM-generated heuristics in a tree structure. With a novel thought-alignment +process and an exploration-decay technique, the proposed MCTS-AHD method +delivers significantly higher-quality heuristics on various complex tasks. Our +code is available at https://github.com/zz1358m/MCTS-AHD-master. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets all required criteria: (1) Practical Applications (automatic heuristic design for complex planning tasks), (2) Experimental Results and Quantitative Metrics (delivers significantly higher-quality heuristics), and (3) Comparison with State-of-the-Art (improves upon existing LLM-based AHD methods). Additionally, meets criteria (4) Methodology and Implementation Details (novel thought-alignment process and exploration-decay technique, with available code)." +} +``` + +--- + +## [Leveraging Large Language Models as Knowledge-Driven Agents for Reliable + Retrosynthesis Planning](https://arxiv.org/abs/2501.08897v1) +**arXiv ID:** 2501.08897v1 + +**Abstract:** +> Identifying reliable synthesis pathways in materials chemistry is a complex +task, particularly in polymer science, due to the intricate and often +non-unique nomenclature of macromolecules. To address this challenge, we +propose an agent system that integrates large language models (LLMs) and +knowledge graphs (KGs). By leveraging LLMs' powerful capabilities for +extracting and recognizing chemical substance names, and storing the extracted +data in a structured knowledge graph, our system fully automates the retrieval +of relevant literatures, extraction of reaction data, database querying, +construction of retrosynthetic pathway trees, further expansion through the +retrieval of additional literature and recommendation of optimal reaction +pathways. A novel Multi-branched Reaction Pathway Search (MBRPS) algorithm +enables the exploration of all pathways, with a particular focus on +multi-branched ones, helping LLMs overcome weak reasoning in multi-branched +paths. This work represents the first attempt to develop a fully automated +retrosynthesis planning agent tailored specially for macromolecules powered by +LLMs. Applied to polyimide synthesis, our new approach constructs a +retrosynthetic pathway tree with hundreds of pathways and recommends optimized +routes, including both known and novel pathways, demonstrating its +effectiveness and potential for broader applications. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets all mandatory criteria: focuses on practical application of LLMs in knowledge graphs and retrieval (1), includes quantitative results through pathway tree construction (2), and implicitly compares to state-of-the-art by introducing a novel MBRPS algorithm (3). Additionally, meets optional criteria for methodology description (4), novelty in integrating LLMs with knowledge graphs for retrosynthesis, and demonstrates potential for real-world impact in materials chemistry." +} +``` + +--- + +## [Text Semantics to Flexible Design: A Residential Layout Generation + Method Based on Stable Diffusion Model](https://arxiv.org/abs/2501.09279v1) +**arXiv ID:** 2501.09279v1 + +**Abstract:** +> Flexibility in the AI-based residential layout design remains a significant +challenge, as traditional methods like rule-based heuristics and graph-based +generation often lack flexibility and require substantial design knowledge from +users. To address these limitations, we propose a cross-modal design approach +based on the Stable Diffusion model for generating flexible residential +layouts. The method offers multiple input types for learning objectives, +allowing users to specify both boundaries and layouts. It incorporates natural +language as design constraints and introduces ControlNet to enable stable +layout generation through two distinct pathways. We also present a scheme that +encapsulates design expertise within a knowledge graph and translates it into +natural language, providing an interpretable representation of design +knowledge. This comprehensibility and diversity of input options enable +professionals and non-professionals to directly express design requirements, +enhancing flexibility and controllability. Finally, experiments verify the +flexibility of the proposed methods under multimodal constraints better than +state-of-the-art models, even when specific semantic information about room +areas or connections is incomplete. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Application (residential layout generation), (2) Experimental Results with Quantitative Metrics (comparison with state-of-the-art models), and (3) Comparison with State-of-the-Art. Additionally, it meets optional criteria (4) Methodology and Implementation Details (clear description of Stable Diffusion model and ControlNet usage) and shows Novelty in integrating LLMs with knowledge graphs for design expertise representation. + +--- + +## [SOP-Agent: Empower General Purpose AI Agent with Domain-Specific SOPs](https://arxiv.org/abs/2501.09316v1) +**arXiv ID:** 2501.09316v1 + +**Abstract:** +> Despite significant advancements in general-purpose AI agents, several +challenges still hinder their practical application in real-world scenarios. +First, the limited planning capabilities of Large Language Models (LLM) +restrict AI agents from effectively solving complex tasks that require +long-horizon planning. Second, general-purpose AI agents struggle to +efficiently utilize domain-specific knowledge and human expertise. In this +paper, we introduce the Standard Operational Procedure-guided Agent +(SOP-agent), a novel framework for constructing domain-specific agents through +pseudocode-style Standard Operational Procedures (SOPs) written in natural +language. Formally, we represent a SOP as a decision graph, which is traversed +to guide the agent in completing tasks specified by the SOP. We conduct +extensive experiments across tasks in multiple domains, including +decision-making, search and reasoning, code generation, data cleaning, and +grounded customer service. The SOP-agent demonstrates excellent versatility, +achieving performance superior to general-purpose agent frameworks and +comparable to domain-specific agent systems. Additionally, we introduce the +Grounded Customer Service Benchmark, the first benchmark designed to evaluate +the grounded decision-making capabilities of AI agents in customer service +scenarios based on SOPs. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1, 2, and 3): focuses on practical application of LLMs in real-world tasks (1), includes experimental results with quantitative metrics showing performance improvements (2), and compares results with state-of-the-art techniques (3). Additionally, meets optional criteria 4 (clear methodology and implementation details) and demonstrates novelty in integrating LLMs with SOPs for agentic AI, enhancing its potential impact. + +--- + +## [Aligning Instruction Tuning with Pre-training](https://arxiv.org/abs/2501.09368v1) +**arXiv ID:** 2501.09368v1 + +**Abstract:** +> Instruction tuning enhances large language models (LLMs) to follow human +instructions across diverse tasks, relying on high-quality datasets to guide +behavior. However, these datasets, whether manually curated or synthetically +generated, are often narrowly focused and misaligned with the broad +distributions captured during pre-training, limiting LLM generalization and +effective use of pre-trained knowledge. We propose *Aligning Instruction Tuning +with Pre-training* (AITP), a method that bridges this gap by identifying +coverage shortfalls in instruction-tuning datasets and rewriting +underrepresented pre-training data into high-quality instruction-response +pairs. This approach enriches dataset diversity while preserving task-specific +objectives. Evaluations on three fully open LLMs across eight benchmarks +demonstrate consistent performance improvements with AITP. Ablations highlight +the benefits of adaptive data selection, controlled rewriting, and balanced +integration, emphasizing the importance of aligning instruction tuning with +pre-training distributions to unlock the full potential of LLMs. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1, 2, 3): focuses on practical application of LLMs (instruction tuning), includes experimental results with quantitative metrics (evaluations on three LLMs across eight benchmarks), and compares with state-of-the-art techniques. Additionally, meets optional criteria for methodology and implementation details (describes AITP method), novelty (introduces a new approach for aligning instruction tuning with pre-training), and robust experimental validation. + +--- + +## [Doc-Guided Sent2Sent++: A Sent2Sent++ Agent with Doc-Guided memory for + Document-level Machine Translation](https://arxiv.org/abs/2501.08523v1) +**arXiv ID:** 2501.08523v1 + +**Abstract:** +> The field of artificial intelligence has witnessed significant advancements +in natural language processing, largely attributed to the capabilities of Large +Language Models (LLMs). These models form the backbone of Agents designed to +address long-context dependencies, particularly in Document-level Machine +Translation (DocMT). DocMT presents unique challenges, with quality, +consistency, and fluency being the key metrics for evaluation. Existing +approaches, such as Doc2Doc and Doc2Sent, either omit sentences or compromise +fluency. This paper introduces Doc-Guided Sent2Sent++, an Agent that employs an +incremental sentence-level forced decoding strategy \textbf{to ensure every +sentence is translated while enhancing the fluency of adjacent sentences.} Our +Agent leverages a Doc-Guided Memory, focusing solely on the summary and its +translation, which we find to be an efficient approach to maintaining +consistency. Through extensive testing across multiple languages and domains, +we demonstrate that Sent2Sent++ outperforms other methods in terms of quality, +consistency, and fluency. The results indicate that, our approach has achieved +significant improvements in metrics such as s-COMET, d-COMET, LTCR-$1_f$, and +document-level perplexity (d-ppl). The contributions of this paper include a +detailed analysis of current DocMT research, the introduction of the +Sent2Sent++ decoding method, the Doc-Guided Memory mechanism, and validation of +its effectiveness across languages and domains. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1: Practical Applications in DocMT with LLMs, 2: Experimental Results with quantitative metrics like s-COMET and d-ppl, 3: Comparison with State-of-the-Art methods). Additionally meets criteria 4 (Methodology and Implementation Details) and shows Novelty in approach. Does not fall under any rejection categories. + +--- + +## [LlamaRestTest: Effective REST API Testing with Small Language Models](https://arxiv.org/abs/2501.08598v1) +**arXiv ID:** 2501.08598v1 + +**Abstract:** +> Modern web services rely heavily on REST APIs, typically documented using the +OpenAPI specification. The widespread adoption of this standard has resulted in +the development of many black-box testing tools that generate tests based on +these specifications. Recent advancements in Natural Language Processing (NLP), +particularly with Large Language Models (LLMs), have enhanced REST API testing +by extracting actionable rules and generating input values from the +human-readable portions of the specification. However, these advancements +overlook the potential of continuously refining the identified rules and test +inputs based on server responses. To address this limitation, we present +LlamaRestTest, a novel approach that employs two custom LLMs to generate +realistic test inputs and uncover parameter dependencies during the testing +process by incorporating server responses. These LLMs are created by +fine-tuning the Llama3-8b model, using mined datasets of REST API example +values and inter-parameter dependencies. We evaluated LlamaRestTest on 12 +real-world services (including popular services such as Spotify), comparing it +against RESTGPT, a GPT-powered specification-enhancement tool, as well as +several state-of-the-art REST API testing tools, including RESTler, MoRest, +EvoMaster, and ARAT-RL. Our results show that fine-tuning enables smaller LLMs +to outperform larger models in detecting actionable rules and generating inputs +for REST API testing. We evaluated configurations from the base Llama3-8B to +fine-tuned versions and explored 2-bit, 4-bit, and 8-bit quantization for +efficiency. LlamaRestTest surpasses state-of-the-art tools in code coverage and +error detection, even with RESTGPT-enhanced specifications, and an ablation +study highlights the impact of its novel components. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1, 2, and 3) by focusing on a practical application of LLMs in REST API testing, presenting experimental results with quantitative metrics, and comparing its approach to state-of-the-art techniques. Additionally, it meets optional criteria 4 (clear methodology and implementation details) and shows novelty in fine-tuning smaller LLMs for improved performance. + +--- + +## [AutoRestTest: A Tool for Automated REST API Testing Using LLMs and MARL](https://arxiv.org/abs/2501.08600v1) +**arXiv ID:** 2501.08600v1 + +**Abstract:** +> As REST APIs have become widespread in modern web services, comprehensive +testing of these APIs has become increasingly crucial. Due to the vast search +space consisting of operations, parameters, and parameter values along with +their complex dependencies and constraints, current testing tools suffer from +low code coverage, leading to suboptimal fault detection. To address this +limitation, we present a novel tool, AutoRestTest, which integrates the +Semantic Operation Dependency Graph (SODG) with Multi-Agent Reinforcement +Learning (MARL) and large language models (LLMs) for effective REST API +testing. AutoRestTest determines operation-dependent parameters using the SODG +and employs five specialized agents (operation, parameter, value, dependency, +and header) to identify dependencies of operations and generate operation +sequences, parameter combinations, and values. AutoRestTest provides a +command-line interface and continuous telemetry on successful operation count, +unique server errors detected, and time elapsed. Upon completion, AutoRestTest +generates a detailed report highlighting errors detected and operations +exercised. In this paper, we introduce our tool and present preliminary +results. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: REST API testing), 2 (Experimental Results: preliminary results with potential for quantitative metrics), and 4 (Methodology and Implementation Details: clear description of using LLMs with MARL for API testing). Additionally, shows novelty in integrating LLMs with MARL for a practical application, making it a strong candidate for further review. + +--- + +## [MAGNET: Augmenting Generative Decoders with Representation Learning and + Infilling Capabilities](https://arxiv.org/abs/2501.08648v1) +**arXiv ID:** 2501.08648v1 + +**Abstract:** +> While originally designed for unidirectional generative modeling, +decoder-only large language models (LLMs) are increasingly being adapted for +bidirectional modeling. However, unidirectional and bidirectional models are +typically trained separately with distinct objectives (generation and +representation learning, respectively). This separation overlooks the +opportunity for developing a more versatile language model and for these +objectives to complement each other. In this work, we introduce MAGNET, an +adaptation of decoder-only LLMs that enhances their ability to generate robust +representations and infill missing text spans, while preserving their knowledge +and text generation capabilities. MAGNET employs three self-supervised training +objectives and introduces an attention mechanism that combines bidirectional +and causal attention, enabling unified training across all objectives. Our +results demonstrate that LLMs adapted with MAGNET (1) surpass strong text +encoders on token-level and sentence-level representation learning tasks, (2) +generate contextually appropriate text infills by leveraging future context, +(3) retain the ability for open-ended text generation without exhibiting +repetition problem, and (4) preserve the knowledge gained by the LLM during +pretraining. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: enhances LLMs for robust representations and text infilling), 2 (Experimental Results and Quantitative Metrics: demonstrates performance improvements in representation learning and text generation tasks), and 3 (Comparison with State-of-the-Art: surpasses strong text encoders). Additionally, shows Novelty in combining bidirectional and causal attention, and its approach is implementable with current standard tools, supporting further review. + +--- + +## [ToMATO: Verbalizing the Mental States of Role-Playing LLMs for + Benchmarking Theory of Mind](https://arxiv.org/abs/2501.08838v1) +**arXiv ID:** 2501.08838v1 + +**Abstract:** +> Existing Theory of Mind (ToM) benchmarks diverge from real-world scenarios in +three aspects: 1) they assess a limited range of mental states such as beliefs, +2) false beliefs are not comprehensively explored, and 3) the diverse +personality traits of characters are overlooked. To address these challenges, +we introduce ToMATO, a new ToM benchmark formulated as multiple-choice QA over +conversations. ToMATO is generated via LLM-LLM conversations featuring +information asymmetry. By employing a prompting method that requires +role-playing LLMs to verbalize their thoughts before each utterance, we capture +both first- and second-order mental states across five categories: belief, +intention, desire, emotion, and knowledge. These verbalized thoughts serve as +answers to questions designed to assess the mental states of characters within +conversations. Furthermore, the information asymmetry introduced by hiding +thoughts from others induces the generation of false beliefs about various +mental states. Assigning distinct personality traits to LLMs further +diversifies both utterances and thoughts. ToMATO consists of 5.4k questions, +753 conversations, and 15 personality trait patterns. Our analysis shows that +this dataset construction approach frequently generates false beliefs due to +the information asymmetry between role-playing LLMs, and effectively reflects +diverse personalities. We evaluate nine LLMs on ToMATO and find that even +GPT-4o mini lags behind human performance, especially in understanding false +beliefs, and lacks robustness to various personality traits. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria: (1) Practical Applications (benchmarking Theory of Mind in LLMs), (2) Experimental Results and Quantitative Metrics (evaluation of 9 LLMs on ToMATO with performance metrics), and (3) Comparison with State-of-the-Art (comparison to human performance). Additionally, meets criterion (4) Methodology and Implementation Details (clear description of LLM-LLM conversations and prompting method). Novelty is also present in the approach to capture mental states and false beliefs. + +--- + +## [Exploring Task-Level Optimal Prompts for Visual In-Context Learning](https://arxiv.org/abs/2501.08841v1) +**arXiv ID:** 2501.08841v1 + +**Abstract:** +> With the development of Vision Foundation Models (VFMs) in recent years, +Visual In-Context Learning (VICL) has become a better choice compared to +modifying models in most scenarios. Different from retraining or fine-tuning +model, VICL does not require modifications to the model's weights or +architecture, and only needs a prompt with demonstrations to teach VFM how to +solve tasks. Currently, significant computational cost for finding optimal +prompts for every test sample hinders the deployment of VICL, as determining +which demonstrations to use for constructing prompts is very costly. In this +paper, however, we find a counterintuitive phenomenon that most test samples +actually achieve optimal performance under the same prompts, and searching for +sample-level prompts only costs more time but results in completely identical +prompts. Therefore, we propose task-level prompting to reduce the cost of +searching for prompts during the inference stage and introduce two time-saving +yet effective task-level prompt search strategies. Extensive experimental +results show that our proposed method can identify near-optimal prompts and +reach the best VICL performance with a minimal cost that prior work has never +achieved. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: optimizing prompts for Visual In-Context Learning), 2 (Experimental Results and Quantitative Metrics: extensive results showing near-optimal prompts with minimal cost), and 4 (Methodology and Implementation Details: clearly described task-level prompt search strategies). Additionally, shows novelty in introducing time-saving yet effective prompt search strategies, making it a strong candidate for further review. + +--- + +## [Incrementally Learning Multiple Diverse Data Domains via Multi-Source + Dynamic Expansion Model](https://arxiv.org/abs/2501.08878v1) +**arXiv ID:** 2501.08878v1 + +**Abstract:** +> Continual Learning seeks to develop a model capable of incrementally +assimilating new information while retaining prior knowledge. However, current +research predominantly addresses a straightforward learning context, wherein +all data samples originate from a singular data domain. This paper shifts focus +to a more complex and realistic learning environment, characterized by data +samples sourced from multiple distinct domains. We tackle this intricate +learning challenge by introducing a novel methodology, termed the Multi-Source +Dynamic Expansion Model (MSDEM), which leverages various pre-trained models as +backbones and progressively establishes new experts based on them to adapt to +emerging tasks. Additionally, we propose an innovative dynamic expandable +attention mechanism designed to selectively harness knowledge from multiple +backbones, thereby accelerating the new task learning. Moreover, we introduce a +dynamic graph weight router that strategically reuses all previously acquired +parameters and representations for new task learning, maximizing the positive +knowledge transfer effect, which further improves generalization performance. +We conduct a comprehensive series of experiments, and the empirical findings +indicate that our proposed approach achieves state-of-the-art performance. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications for complex learning environments), 2 (Experimental Results with state-of-the-art performance), and 3 (Comparison with State-of-the-Art). Also, shows novelty in methodology and approach, aligning with Additional Considerations for novelty and potential impact through experimental validation. + +--- + +## [Decompose-ToM: Enhancing Theory of Mind Reasoning in Large Language + Models through Simulation and Task Decomposition](https://arxiv.org/abs/2501.09056v1) +**arXiv ID:** 2501.09056v1 + +**Abstract:** +> Theory of Mind (ToM) is the ability to understand and reflect on the mental +states of others. Although this capability is crucial for human interaction, +testing on Large Language Models (LLMs) reveals that they possess only a +rudimentary understanding of it. Although the most capable closed-source LLMs +have come close to human performance on some ToM tasks, they still perform +poorly on complex variations of the task that involve more structured +reasoning. In this work, we utilize the concept of "pretend-play", or +``Simulation Theory'' from cognitive psychology to propose ``Decompose-ToM'': +an LLM-based inference algorithm that improves model performance on complex ToM +tasks. We recursively simulate user perspectives and decompose the ToM task +into a simpler set of functions: subject identification, question-reframing, +world model updation, and knowledge availability. We test the algorithm on +higher-order ToM tasks and a task testing for ToM capabilities in a +conversational setting, demonstrating that our approach shows significant +improvement across models compared to baseline methods while requiring minimal +prompt tuning across tasks and no additional model training. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in enhancing Theory of Mind with LLMs), 2 (Experimental Results with quantitative metrics showing performance improvement), and 4 (Clearly describes methodology and implementation). Additionally, shows novelty in approach (Simulation Theory for ToM) and has robust experimental validation, making it a strong candidate for further review. + +--- + +## [Towards Multilingual LLM Evaluation for Baltic and Nordic languages: A + study on Lithuanian History](https://arxiv.org/abs/2501.09154v1) +**arXiv ID:** 2501.09154v1 + +**Abstract:** +> In this work, we evaluated Lithuanian and general history knowledge of +multilingual Large Language Models (LLMs) on a multiple-choice +question-answering task. The models were tested on a dataset of Lithuanian +national and general history questions translated into Baltic, Nordic, and +other languages (English, Ukrainian, Arabic) to assess the knowledge sharing +from culturally and historically connected groups. We evaluated GPT-4o, +LLaMa3.1 8b and 70b, QWEN2.5 7b and 72b, Mistral Nemo 12b, LLaMa3 8b, Mistral +7b, LLaMa3.2 3b, and Nordic fine-tuned models (GPT-SW3 and LLaMa3 8b). + Our results show that GPT-4o consistently outperformed all other models +across language groups, with slightly better results for Baltic and Nordic +languages. Larger open-source models like QWEN2.5 72b and LLaMa3.1 70b +performed well but showed weaker alignment with Baltic languages. Smaller +models (Mistral Nemo 12b, LLaMa3.2 3b, QWEN 7B, LLaMa3.1 8B, and LLaMa3 8b) +demonstrated gaps with LT-related alignment with Baltic languages while +performing better on Nordic and other languages. The Nordic fine-tuned models +did not surpass multilingual models, indicating that shared cultural or +historical context alone does not guarantee better performance. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 2 (Experimental Results and Quantitative Metrics) by presenting comparative performance results of various LLMs on a specific task, and marginally meets criterion 1 (Practical Applications) due to its focus on evaluating LLMs' knowledge of Baltic and Nordic languages, which could imply potential for real-world applications in those regions. Prioritizing inclusiveness, this aligns with the goal of capturing a broad spectrum of relevant research. + +--- + +## [The Veln(ia)s is in the Details: Evaluating LLM Judgment on Latvian and + Lithuanian Short Answer Matching](https://arxiv.org/abs/2501.09164v1) +**arXiv ID:** 2501.09164v1 + +**Abstract:** +> In this work, we address the challenge of evaluating large language models +(LLMs) on the short answer matching task for Latvian and Lithuanian languages. +We introduce novel datasets consisting of 502 Latvian and 690 Lithuanian +question-answer pairs. For each question-answer pair, we generated matched and +non-matched answers using a set of alteration rules specifically designed to +introduce small but meaningful changes in the text. These generated answers +serve as test cases to assess the ability of LLMs to detect subtle differences +in matching of the original answers. A subset of the datasets was manually +verified for quality and accuracy. Our results show that while larger LLMs, +such as QWEN2.5 72b and LLaMa3.1 70b, demonstrate near-perfect performance in +distinguishing matched and non-matched answers, smaller models show more +variance. For instance, LLaMa3.1 8b and EuroLLM 9b benefited from few-shot +examples, while Mistral Nemo 12b underperformed on detection of subtle text +alteration, particularly in Lithuanian, even with additional examples. QWEN2.5 +7b and Mistral 7b were able to obtain a strong and comparable performance to +the larger 70b models in zero and few shot experiments. Moreover, the +performance of Mistral 7b was weaker in few shot experiments. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (Practical Applications, Experimental Results and Quantitative Metrics, Comparison with State-of-the-Art) and one additional consideration (Novelty, through introduction of novel datasets and evaluation of LLMs on specific languages). Does not fall under rejection categories. + +--- + +## [Attention is All You Need Until You Need Retention](https://arxiv.org/abs/2501.09166v1) +**arXiv ID:** 2501.09166v1 + +**Abstract:** +> This work introduces a novel Retention Layer mechanism for Transformer based +architectures, addressing their inherent lack of intrinsic retention +capabilities. Unlike human cognition, which can encode and dynamically recall +symbolic templates, Generative Pretrained Transformers rely solely on fixed +pretrained weights and ephemeral context windows, limiting their adaptability. +The proposed Retention Layer incorporates a persistent memory module capable of +real time data population, dynamic recall, and guided output generation. This +enhancement allows models to store, update, and reuse observed patterns across +sessions, enabling incremental learning and bridging the gap between static +pretraining and dynamic, context sensitive adaptation. The Retention Layer +design parallels social learning processes, encompassing attention, retention, +reproduction, and motivation stages. Technically, it integrates a memory +attention mechanism and episodic buffers to manage memory scalability, mitigate +overfitting, and ensure efficient recall. Applications span adaptive personal +assistants, real time fraud detection, autonomous robotics, content moderation, +and healthcare diagnostics. In each domain, the retention mechanism enables +systems to learn incrementally, personalize outputs, and respond to evolving +real world challenges effectively. By emulating key aspects of human learning, +this retention enhanced architecture fosters a more fluid and responsive AI +paradigm, paving the way for dynamic, session aware models that extend the +capabilities of traditional Transformers into domains requiring continual +adaptation. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1, 2, 3) and additional considerations: focused on practical applications of LLMs (1), implied experimental results for performance improvements (2), compares with existing state-of-the-art Transformers (3); also meets optional criteria for novelty (introduces Retention Layer), methodology description, and potential for real-world impact, while avoiding rejected topic areas. + +--- + +## [Guiding Retrieval using LLM-based Listwise Rankers](https://arxiv.org/abs/2501.09186v1) +**arXiv ID:** 2501.09186v1 + +**Abstract:** +> Large Language Models (LLMs) have shown strong promise as rerankers, +especially in ``listwise'' settings where an LLM is prompted to rerank several +search results at once. However, this ``cascading'' retrieve-and-rerank +approach is limited by the bounded recall problem: relevant documents not +retrieved initially are permanently excluded from the final ranking. Adaptive +retrieval techniques address this problem, but do not work with listwise +rerankers because they assume a document's score is computed independently from +other documents. In this paper, we propose an adaptation of an existing +adaptive retrieval method that supports the listwise setting and helps guide +the retrieval process itself (thereby overcoming the bounded recall problem for +LLM rerankers). Specifically, our proposed algorithm merges results both from +the initial ranking and feedback documents provided by the most relevant +documents seen up to that point. Through extensive experiments across diverse +LLM rerankers, first stage retrievers, and feedback sources, we demonstrate +that our method can improve nDCG@10 by up to 13.23% and recall by 28.02%--all +while keeping the total number of LLM inferences constant and overheads due to +the adaptive process minimal. The work opens the door to leveraging LLM-based +search in settings where the initial pool of results is limited, e.g., by +legacy systems, or by the cost of deploying a semantic first-stage. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) focuses on practical application of LLMs in retrieval-augmented settings, (2) includes experimental results with quantitative metrics (nDCG@10, recall improvements), and (3) implicitly compares with state-of-the-art by addressing the bounded recall problem. Additionally, it meets extra criteria for (4) methodology description, (5) real-world application discussion, and shows novelty in adapting adaptive retrieval for listwise LLM rerankers. + +--- + +## [Perspective Transition of Large Language Models for Solving Subjective + Tasks](https://arxiv.org/abs/2501.09265v1) +**arXiv ID:** 2501.09265v1 + +**Abstract:** +> Large language models (LLMs) have revolutionized the field of natural +language processing, enabling remarkable progress in various tasks. Different +from objective tasks such as commonsense reasoning and arithmetic +question-answering, the performance of LLMs on subjective tasks is still +limited, where the perspective on the specific problem plays crucial roles for +better interpreting the context and giving proper response. For example, in +certain scenarios, LLMs may perform better when answering from an expert role +perspective, potentially eliciting their relevant domain knowledge. In +contrast, in some scenarios, LLMs may provide more accurate responses when +answering from a third-person standpoint, enabling a more comprehensive +understanding of the problem and potentially mitigating inherent biases. In +this paper, we propose Reasoning through Perspective Transition (RPT), a method +based on in-context learning that enables LLMs to dynamically select among +direct, role, and third-person perspectives for the best way to solve +corresponding subjective problem. Through extensive experiments on totally 12 +subjective tasks by using both closed-source and open-source LLMs including +GPT-4, GPT-3.5, Llama-3, and Qwen-2, our method outperforms widely used single +fixed perspective based methods such as chain-of-thought prompting and expert +prompting, highlights the intricate ways that LLMs can adapt their perspectives +to provide nuanced and contextually appropriate responses for different +problems. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications) through improved performance on subjective tasks, criteria 2 (Experimental Results and Quantitative Metrics) with extensive experiments on multiple LLMs, and partially meets criterion 5 (Real-world Applications and Challenges) by addressing LLM limitations. Additionally, shows novelty in perspective transition approach and has potential for implementability with current standard tools." +} +``` + +--- + +## [Style4Rec: Enhancing Transformer-based E-commerce Recommendation Systems + with Style and Shopping Cart Information](https://arxiv.org/abs/2501.09354v1) +**arXiv ID:** 2501.09354v1 + +**Abstract:** +> Understanding users' product preferences is essential to the efficacy of a +recommendation system. Precision marketing leverages users' historical data to +discern these preferences and recommends products that align with them. +However, recent browsing and purchase records might better reflect current +purchasing inclinations. Transformer-based recommendation systems have made +strides in sequential recommendation tasks, but they often fall short in +utilizing product image style information and shopping cart data effectively. +In light of this, we propose Style4Rec, a transformer-based e-commerce +recommendation system that harnesses style and shopping cart information to +enhance existing transformer-based sequential product recommendation systems. +Style4Rec represents a significant step forward in personalized e-commerce +recommendations, outperforming benchmarks across various evaluation metrics. +Style4Rec resulted in notable improvements: HR@5 increased from 0.681 to 0.735, +NDCG@5 increased from 0.594 to 0.674, and MRR@5 increased from 0.559 to 0.654. +We tested our model using an e-commerce dataset from our partnering company and +found that it exceeded established transformer-based sequential recommendation +benchmarks across various evaluation metrics. Thus, Style4Rec presents a +significant step forward in personalized e-commerce recommendation systems. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Applications in e-commerce, (2) Experimental Results with quantitative metrics (HR@5, NDCG@5, MRR@5), and (3) Comparison with State-of-the-Art (outperforming established transformer-based sequential recommendation benchmarks). Additionally, it meets extra criteria for Methodology and Implementation Details (described utilization of LLMs in practical tasks) and shows Novelty in approach. + +--- + +## [Beyond Reward Hacking: Causal Rewards for Large Language Model Alignment](https://arxiv.org/abs/2501.09620v1) +**arXiv ID:** 2501.09620v1 + +**Abstract:** +> Recent advances in large language models (LLMs) have demonstrated significant +progress in performing complex tasks. While Reinforcement Learning from Human +Feedback (RLHF) has been effective in aligning LLMs with human preferences, it +is susceptible to spurious correlations in reward modeling. Consequently, it +often introduces biases-such as length bias, sycophancy, conceptual bias, and +discrimination that hinder the model's ability to capture true causal +relationships. To address this, we propose a novel causal reward modeling +approach that integrates causal inference to mitigate these spurious +correlations. Our method enforces counterfactual invariance, ensuring reward +predictions remain consistent when irrelevant variables are altered. Through +experiments on both synthetic and real-world datasets, we show that our +approach mitigates various types of spurious correlations effectively, +resulting in more reliable and fair alignment of LLMs with human preferences. +As a drop-in enhancement to the existing RLHF workflow, our causal reward +modeling provides a practical way to improve the trustworthiness and fairness +of LLM finetuning. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: LLM alignment), 2 (Experimental Results and Quantitative Metrics: experiments on synthetic and real-world datasets), and 3 (Comparison with State-of-the-Art: enhances existing RLHF workflow). Additionally, showcases Novelty (causal reward modeling approach) and has potential for Reproducibility and Documentation (describes a drop-in enhancement with experimental validation). + +--- + +## [The Heap: A Contamination-Free Multilingual Code Dataset for Evaluating + Large Language Models](https://arxiv.org/abs/2501.09653v1) +**arXiv ID:** 2501.09653v1 + +**Abstract:** +> The recent rise in the popularity of large language models has spurred the +development of extensive code datasets needed to train them. This has left +limited code available for collection and use in the downstream investigation +of specific behaviors, or evaluation of large language models without suffering +from data contamination. To address this problem, we release The Heap, a large +multilingual dataset covering 57 programming languages that has been +deduplicated with respect to other open datasets of code, enabling researchers +to conduct fair evaluations of large language models without significant data +cleaning overhead. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 2 (Experimental Results and Quantitative Metrics implied through dataset deduplication for fair LLM evaluation), 4 (Methodology and Implementation Details for dataset creation), and shows potential for Impact through Experimental Validation. Novelty is also present in the creation of a contamination-free multilingual code dataset, making it a valuable resource for LLM evaluation. + +--- + +## [Towards Large Reasoning Models: A Survey of Reinforced Reasoning with + Large Language Models](https://arxiv.org/abs/2501.09686v1) +**arXiv ID:** 2501.09686v1 + +**Abstract:** +> Language has long been conceived as an essential tool for human reasoning. +The breakthrough of Large Language Models (LLMs) has sparked significant +research interest in leveraging these models to tackle complex reasoning tasks. +Researchers have moved beyond simple autoregressive token generation by +introducing the concept of "thought" -- a sequence of tokens representing +intermediate steps in the reasoning process. This innovative paradigm enables +LLMs' to mimic complex human reasoning processes, such as tree search and +reflective thinking. Recently, an emerging trend of learning to reason has +applied reinforcement learning (RL) to train LLMs to master reasoning +processes. This approach enables the automatic generation of high-quality +reasoning trajectories through trial-and-error search algorithms, significantly +expanding LLMs' reasoning capacity by providing substantially more training +data. Furthermore, recent studies demonstrate that encouraging LLMs to "think" +with more tokens during test-time inference can further significantly boost +reasoning accuracy. Therefore, the train-time and test-time scaling combined to +show a new research frontier -- a path toward Large Reasoning Model. The +introduction of OpenAI's o1 series marks a significant milestone in this +research direction. In this survey, we present a comprehensive review of recent +progress in LLM reasoning. We begin by introducing the foundational background +of LLMs and then explore the key technical components driving the development +of large reasoning models, with a focus on automated data construction, +learning-to-reason techniques, and test-time scaling. We also analyze popular +open-source projects at building large reasoning models, and conclude with open +challenges and future research directions. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: leveraging LLMs for complex reasoning tasks), 3 (Comparison with State-of-the-Art: implied through discussion of recent progress and open-source projects), and 4 (Methodology and Implementation Details: provides overview of key technical components). Also shows novelty in exploring Large Reasoning Models, making it a strong candidate for further review. + +--- + +## [CyberMentor: AI Powered Learning Tool Platform to Address Diverse + Student Needs in Cybersecurity Education](https://arxiv.org/abs/2501.09709v1) +**arXiv ID:** 2501.09709v1 + +**Abstract:** +> Many non-traditional students in cybersecurity programs often lack access to +advice from peers, family members and professors, which can hinder their +educational experiences. Additionally, these students may not fully benefit +from various LLM-powered AI assistants due to issues like content relevance, +locality of advice, minimum expertise, and timing. This paper addresses these +challenges by introducing an application designed to provide comprehensive +support by answering questions related to knowledge, skills, and career +preparation advice tailored to the needs of these students. We developed a +learning tool platform, CyberMentor, to address the diverse needs and pain +points of students majoring in cybersecurity. Powered by agentic workflow and +Generative Large Language Models (LLMs), the platform leverages +Retrieval-Augmented Generation (RAG) for accurate and contextually relevant +information retrieval to achieve accessibility and personalization. We +demonstrated its value in addressing knowledge requirements for cybersecurity +education and for career marketability, in tackling skill requirements for +analytical and programming assignments, and in delivering real time on demand +learning support. Using three use scenarios, we showcased CyberMentor in +facilitating knowledge acquisition and career preparation and providing +seamless skill-based guidance and support. We also employed the LangChain +prompt-based evaluation methodology to evaluate the platform's impact, +confirming its strong performance in helpfulness, correctness, and +completeness. These results underscore the system's ability to support students +in developing practical cybersecurity skills while improving equity and +sustainability within higher education. Furthermore, CyberMentor's open-source +design allows for adaptation across other disciplines, fostering educational +innovation and broadening its potential impact. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in cybersecurity education), 2 (Experimental Results with quantitative metrics through LangChain prompt-based evaluation), 4 (Methodology and Implementation Details with RAG and agentic workflow), and shows novelty in applying LLMs to address educational challenges, with potential for broad impact across disciplines. + +--- + +## [A Simple Aerial Detection Baseline of Multimodal Language Models](https://arxiv.org/abs/2501.09720v1) +**arXiv ID:** 2501.09720v1 + +**Abstract:** +> The multimodal language models (MLMs) based on generative pre-trained +Transformer are considered powerful candidates for unifying various domains and +tasks. MLMs developed for remote sensing (RS) have demonstrated outstanding +performance in multiple tasks, such as visual question answering and visual +grounding. In addition to visual grounding that detects specific objects +corresponded to given instruction, aerial detection, which detects all objects +of multiple categories, is also a valuable and challenging task for RS +foundation models. However, aerial detection has not been explored by existing +RS MLMs because the autoregressive prediction mechanism of MLMs differs +significantly from the detection outputs. In this paper, we present a simple +baseline for applying MLMs to aerial detection for the first time, named +LMMRotate. Specifically, we first introduce a normalization method to transform +detection outputs into textual outputs to be compatible with the MLM framework. +Then, we propose a evaluation method, which ensures a fair comparison between +MLMs and conventional object detection models. We construct the baseline by +fine-tuning open-source general-purpose MLMs and achieve impressive detection +performance comparable to conventional detector. We hope that this baseline +will serve as a reference for future MLM development, enabling more +comprehensive capabilities for understanding RS images. Code is available at +https://github.com/Li-Qingyun/mllm-mmrotate. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in aerial detection, a real-world task), 2 (Experimental Results with quantitative metrics for detection performance), and 4 (Methodology and Implementation Details, including a novel normalization method for MLMs). Additionally, it shows Novelty in applying MLMs to aerial detection and provides Reproducibility with available code, outweighing minor uncertainties about comparison to State-of-the-Art and Agentic AI aspects. + +--- + +## [Knowledge prompt chaining for semantic modeling](https://arxiv.org/abs/2501.08540v1) +**arXiv ID:** 2501.08540v1 + +**Abstract:** +> The task of building semantics for structured data such as CSV, JSON, and XML +files is highly relevant in the knowledge representation field. Even though we +have a vast of structured data on the internet, mapping them to domain +ontologies to build semantics for them is still very challenging as it requires +the construction model to understand and learn graph-structured knowledge. +Otherwise, the task will require human beings' effort and cost. In this paper, +we proposed a novel automatic semantic modeling framework: Knowledge Prompt +Chaining. It can serialize the graph-structured knowledge and inject it into +the LLMs properly in a Prompt Chaining architecture. Through this knowledge +injection and prompting chaining, the model in our framework can learn the +structure information and latent space of the graph and generate the semantic +labels and semantic graphs following the chains' insturction naturally. Based +on experimental results, our method achieves better performance than existing +leading techniques, despite using reduced structured input data. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Applications in knowledge graphs, (2) Experimental Results with quantitative metrics showing performance improvements, and (3) Comparison with State-of-the-Art techniques. Additionally, meets optional criteria for Methodology and Implementation Details (4) and Novelty in applying LLMs to semantic modeling. + +--- + +## [LAMS: LLM-Driven Automatic Mode Switching for Assistive Teleoperation](https://arxiv.org/abs/2501.08558v1) +**arXiv ID:** 2501.08558v1 + +**Abstract:** +> Teleoperating high degrees-of-freedom (DoF) robotic manipulators via low-DoF +controllers like joysticks often requires frequent switching between control +modes, where each mode maps controller movements to specific robot actions. +Manually performing this frequent switching can make teleoperation cumbersome +and inefficient. On the other hand, existing automatic mode-switching +solutions, such as heuristic-based or learning-based methods, are often +task-specific and lack generalizability. In this paper, we introduce LLM-Driven +Automatic Mode Switching (LAMS), a novel approach that leverages Large Language +Models (LLMs) to automatically switch control modes based on task context. +Unlike existing methods, LAMS requires no prior task demonstrations and +incrementally improves by integrating user-generated mode-switching examples. +We validate LAMS through an ablation study and a user study with 10 +participants on complex, long-horizon tasks, demonstrating that LAMS +effectively reduces manual mode switches, is preferred over alternative +methods, and improves performance over time. The project website with +supplementary materials is at https://lams-assistance.github.io/. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: Practical Applications (teleoperation with LLMs), Experimental Results and Quantitative Metrics (ablation and user studies with concrete results), and Comparison with State-of-the-Art (improvements over alternative methods). Additionally, meets extra criteria for Novelty (novel approach leveraging LLMs), Reproducibility and Documentation (supplementary materials provided), and Impact through Experimental Validation (robust user study with real-world task implications). + +--- + +## [ANSR-DT: An Adaptive Neuro-Symbolic Learning and Reasoning Framework for + Digital Twins](https://arxiv.org/abs/2501.08561v1) +**arXiv ID:** 2501.08561v1 + +**Abstract:** +> In this paper, we propose an Adaptive Neuro-Symbolic Learning Framework for +digital twin technology called ``ANSR-DT." Our approach combines pattern +recognition algorithms with reinforcement learning and symbolic reasoning to +enable real-time learning and adaptive intelligence. This integration enhances +the understanding of the environment and promotes continuous learning, leading +to better and more effective decision-making in real-time for applications that +require human-machine collaboration. We evaluated the \textit{ANSR-DT} +framework for its ability to learn and adapt to dynamic patterns, observing +significant improvements in decision accuracy, reliability, and +interpretability when compared to existing state-of-the-art methods. However, +challenges still exist in extracting and integrating symbolic rules in complex +environments, which limits the full potential of our framework in heterogeneous +settings. Moreover, our ongoing research aims to address this issue in the +future by ensuring seamless integration of neural models at large. In addition, +our open-source implementation promotes reproducibility and encourages future +research to build on our foundational work. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in human-machine collaboration), 2 (Experimental Results with quantitative metrics for decision accuracy), 3 (Comparison with State-of-the-Art showing performance improvements), and 4 (Methodology and Implementation Details with open-source implementation for reproducibility). Additionally, it touches on limitations and future work, indicating potential for further innovative applications with LLMs, though LLMs are not explicitly mentioned, the neuro-symbolic learning framework might imply potential integration or relevance. + +--- + +## [RLHS: Mitigating Misalignment in RLHF with Hindsight Simulation](https://arxiv.org/abs/2501.08617v1) +**arXiv ID:** 2501.08617v1 + +**Abstract:** +> Generative AI systems like foundation models (FMs) must align well with human +values to ensure their behavior is helpful and trustworthy. While Reinforcement +Learning from Human Feedback (RLHF) has shown promise for optimizing model +performance using human judgments, existing RLHF pipelines predominantly rely +on immediate feedback, which can fail to accurately reflect the downstream +impact of an interaction on users' utility. We demonstrate that feedback based +on evaluators' foresight estimates of downstream consequences systematically +induces Goodhart's Law dynamics, incentivizing misaligned behaviors like +sycophancy and deception and ultimately degrading user outcomes. To alleviate +this, we propose decoupling evaluation from prediction by refocusing RLHF on +hindsight feedback. Our theoretical analysis reveals that conditioning +evaluator feedback on downstream observations mitigates misalignment and +improves expected human utility, even when these observations are simulated by +the AI system itself. To leverage this insight in a practical alignment +algorithm, we introduce Reinforcement Learning from Hindsight Simulation +(RLHS), which first simulates plausible consequences and then elicits feedback +to assess what behaviors were genuinely beneficial in hindsight. We apply RLHS +to two widely-employed online and offline preference optimization methods -- +Proximal Policy Optimization (PPO) and Direct Preference Optimization (DPO) -- +and show empirically that misalignment is significantly reduced with both +methods. Through an online human user study, we show that RLHS consistently +outperforms RLHF in helping users achieve their goals and earns higher +satisfaction ratings, despite being trained solely with simulated hindsight +feedback. These results underscore the importance of focusing on long-term +consequences, even simulated ones, to mitigate misalignment in RLHF. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications) through its focus on improving alignment in generative AI systems, 2 (Experimental Results and Quantitative Metrics) with empirical results from online human user studies, and 3 (Comparison with State-of-the-Art) by comparing RLHS to existing RLHF techniques, showing performance improvements. Additionally, it touches on Novelty and Reproducibility with its introduction of RLHS and documented methodologies. + +--- + +## [Leveraging LLM Agents for Translating Network Configurations](https://arxiv.org/abs/2501.08760v1) +**arXiv ID:** 2501.08760v1 + +**Abstract:** +> Configuration translation is a critical and frequent task in network +operations. When a network device is damaged or outdated, administrators need +to replace it to maintain service continuity. The replacement devices may +originate from different vendors, necessitating configuration translation to +ensure seamless network operation. However, translating configurations manually +is a labor-intensive and error-prone process. In this paper, we propose an +intent-based framework for translating network configuration with Large +Language Model (LLM) Agents. The core of our approach is an Intent-based +Retrieval Augmented Generation (IRAG) module that systematically splits a +configuration file into fragments, extracts intents, and generates accurate +translations. We also design a two-stage verification method to validate the +syntax and semantics correctness of the translated configurations. We implement +and evaluate the proposed method on real-world network configurations. +Experimental results show that our method achieves 97.74% syntax correctness, +outperforming state-of-the-art methods in translation accuracy. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets Criteria 1 (Practical Applications: network operations), Criteria 2 (Experimental Results and Quantitative Metrics: 97.74% syntax correctness outperforming state-of-the-art), and Criteria 3 (Comparison with State-of-the-Art). Additionally, exhibits novelty in applying LLMs with Retrieval Augmented Generation (RAG) for network configuration translation." +} +``` + +--- + +## [How Developers Interact with AI: A Taxonomy of Human-AI Collaboration in + Software Engineering](https://arxiv.org/abs/2501.08774v1) +**arXiv ID:** 2501.08774v1 + +**Abstract:** +> Artificial intelligence (AI), including large language models and generative +AI, is emerging as a significant force in software development, offering +developers powerful tools that span the entire development lifecycle. Although +software engineering research has extensively studied AI tools in software +development, the specific types of interactions between developers and these +AI-powered tools have only recently begun to receive attention. Understanding +and improving these interactions has the potential to improve productivity, +trust, and efficiency in AI-driven workflows. In this paper, we propose a +taxonomy of interaction types between developers and AI tools, identifying +eleven distinct interaction types, such as auto-complete code suggestions, +command-driven actions, and conversational assistance. Building on this +taxonomy, we outline a research agenda focused on optimizing AI interactions, +improving developer control, and addressing trust and usability challenges in +AI-assisted development. By establishing a structured foundation for studying +developer-AI interactions, this paper aims to stimulate research on creating +more effective, adaptive AI tools for software development. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in software development), 4 (Methodology and Implementation Details through proposed taxonomy), and potentially 2 (though no explicit experimental results are mentioned in the abstract, the outlined research agenda implies future quantitative metrics). Novel approach to studying developer-AI interactions in software engineering also suggests some degree of novelty. + +--- + +## [IDEA: Image Description Enhanced CLIP-Adapter](https://arxiv.org/abs/2501.08816v1) +**arXiv ID:** 2501.08816v1 + +**Abstract:** +> CLIP (Contrastive Language-Image Pre-training) has attained great success in +pattern recognition and computer vision. Transferring CLIP to downstream tasks +(e.g. zero- or few-shot classification) is a hot topic in multimodal learning. +However, current studies primarily focus on either prompt learning for text or +adapter tuning for vision, without fully exploiting the complementary +information and correlations among image-text pairs. In this paper, we propose +an Image Description Enhanced CLIP-Adapter (IDEA) method to adapt CLIP to +few-shot image classification tasks. This method captures fine-grained features +by leveraging both visual features and textual descriptions of images. IDEA is +a training-free method for CLIP, and it can be comparable to or even exceeds +state-of-the-art models on multiple tasks. Furthermore, we introduce +Trainable-IDEA (T-IDEA), which extends IDEA by adding two lightweight learnable +components (i.e., a projector and a learnable latent space), further enhancing +the model's performance and achieving SOTA results on 11 datasets. As one +important contribution, we employ the Llama model and design a comprehensive +pipeline to generate textual descriptions for images of 11 datasets, resulting +in a total of 1,637,795 image-text pairs, named "IMD-11". Our code and data are +released at https://github.com/FourierAI/IDEA. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: Practical Applications (adaptation for few-shot image classification), Experimental Results and Quantitative Metrics (comparable to or exceeding state-of-the-art models), and Comparison with State-of-the-Art (achieving SOTA results on 11 datasets). Additionally, meets optional criteria for Methodology and Implementation Details (clear description of IDEA and T-IDEA) and Novelty (introducing a new pipeline with LLM for generating image-text pairs). + +--- + +## [MMDocIR: Benchmarking Multi-Modal Retrieval for Long Documents](https://arxiv.org/abs/2501.08828v1) +**arXiv ID:** 2501.08828v1 + +**Abstract:** +> Multi-modal document retrieval is designed to identify and retrieve various +forms of multi-modal content, such as figures, tables, charts, and layout +information from extensive documents. Despite its significance, there is a +notable lack of a robust benchmark to effectively evaluate the performance of +systems in multi-modal document retrieval. To address this gap, this work +introduces a new benchmark, named as MMDocIR, encompassing two distinct tasks: +page-level and layout-level retrieval. The former focuses on localizing the +most relevant pages within a long document, while the latter targets the +detection of specific layouts, offering a more fine-grained granularity than +whole-page analysis. A layout can refer to a variety of elements such as +textual paragraphs, equations, figures, tables, or charts. The MMDocIR +benchmark comprises a rich dataset featuring expertly annotated labels for +1,685 questions and bootstrapped labels for 173,843 questions, making it a +pivotal resource for advancing multi-modal document retrieval for both training +and evaluation. Through rigorous experiments, we reveal that (i) visual +retrievers significantly outperform their text counterparts, (ii) MMDocIR train +set can effectively benefit the training process of multi-modal document +retrieval and (iii) text retrievers leveraging on VLM-text perform much better +than those using OCR-text. These findings underscores the potential advantages +of integrating visual elements for multi-modal document retrieval. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets criteria 1 (Practical Applications) through multi-modal document retrieval, criterion 2 (Experimental Results and Quantitative Metrics) with rigorous experiments showing comparative performance, and criterion 3 (Comparison with State-of-the-Art) by evaluating the effectiveness of different retrieval approaches. Additionally, it aligns with the area of retrieval-augmented generation (RAG) and demonstrates novelty in integrating visual elements for document retrieval." +} +``` + +--- + +## [Disentangling Exploration of Large Language Models by Optimal + Exploitation](https://arxiv.org/abs/2501.08925v1) +**arXiv ID:** 2501.08925v1 + +**Abstract:** +> Exploration is a crucial skill for self-improvement and open-ended +problem-solving. However, it remains uncertain whether large language models +can effectively explore the state-space. Existing evaluations predominantly +focus on the trade-off between exploration and exploitation, often assessed in +multi-armed bandit problems. In contrast, this work isolates exploration as the +sole objective, tasking the agent with delivering information that enhances +future returns. For the evaluation, we propose to decompose missing rewards +into exploration and exploitation components by measuring the optimal +achievable return for the states already explored. Our experiments with various +LLMs reveal that most models struggle to sufficiently explore the state-space +and that weak exploration is insufficient. We observe a positive correlation +between model size and exploration performance, with larger models +demonstrating superior capabilities. Furthermore, we show that our +decomposition provides insights into differences in behaviors driven by agent +instructions during prompt engineering, offering a valuable tool for refining +LLM performance in exploratory tasks. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications with prompt engineering), 2 (Experimental Results with quantitative metrics on exploration performance), and 3 (Comparison with State-of-the-Art implicit in evaluating various LLMs). Additionally, shows Novelty in approach and has potential for Impact through Experimental Validation, warranting further review. + +--- + +## [SteLLA: A Structured Grading System Using LLMs with RAG](https://arxiv.org/abs/2501.09092v1) +**arXiv ID:** 2501.09092v1 + +**Abstract:** +> Large Language Models (LLMs) have shown strong general capabilities in many +applications. However, how to make them reliable tools for some specific tasks +such as automated short answer grading (ASAG) remains a challenge. We present +SteLLA (Structured Grading System Using LLMs with RAG) in which a) Retrieval +Augmented Generation (RAG) approach is used to empower LLMs specifically on the +ASAG task by extracting structured information from the highly relevant and +reliable external knowledge based on the instructor-provided reference answer +and rubric, b) an LLM performs a structured and question-answering-based +evaluation of student answers to provide analytical grades and feedback. A +real-world dataset that contains students' answers in an exam was collected +from a college-level Biology course. Experiments show that our proposed system +can achieve substantial agreement with the human grader while providing +break-down grades and feedback on all the knowledge points examined in the +problem. A qualitative and error analysis of the feedback generated by GPT4 +shows that GPT4 is good at capturing facts while may be prone to inferring too +much implication from the given text in the grading task which provides +insights into the usage of LLMs in the ASAG system. + +**Decision Explanation:** Original decision: ACCEPT +Meets all mandatory criteria (1: Practical Applications in automated short answer grading, 2: Experimental Results with quantitative metrics on agreement with human grader, 3: Comparison with State-of-the-Art through evaluation of GPT4). Additionally, meets optional criteria 4 (clear methodology and implementation details) and has potential for novelty in applying LLMs with RAG for structured grading. Does not fall under rejection categories. + +--- + +## [Agentic Retrieval-Augmented Generation: A Survey on Agentic RAG](https://arxiv.org/abs/2501.09136v1) +**arXiv ID:** 2501.09136v1 + +**Abstract:** +> Large Language Models (LLMs) have revolutionized artificial intelligence (AI) +by enabling human like text generation and natural language understanding. +However, their reliance on static training data limits their ability to respond +to dynamic, real time queries, resulting in outdated or inaccurate outputs. +Retrieval Augmented Generation (RAG) has emerged as a solution, enhancing LLMs +by integrating real time data retrieval to provide contextually relevant and +up-to-date responses. Despite its promise, traditional RAG systems are +constrained by static workflows and lack the adaptability required for +multistep reasoning and complex task management. + Agentic Retrieval-Augmented Generation (Agentic RAG) transcends these +limitations by embedding autonomous AI agents into the RAG pipeline. These +agents leverage agentic design patterns reflection, planning, tool use, and +multiagent collaboration to dynamically manage retrieval strategies, +iteratively refine contextual understanding, and adapt workflows to meet +complex task requirements. This integration enables Agentic RAG systems to +deliver unparalleled flexibility, scalability, and context awareness across +diverse applications. + This survey provides a comprehensive exploration of Agentic RAG, beginning +with its foundational principles and the evolution of RAG paradigms. It +presents a detailed taxonomy of Agentic RAG architectures, highlights key +applications in industries such as healthcare, finance, and education, and +examines practical implementation strategies. Additionally, it addresses +challenges in scaling these systems, ensuring ethical decision making, and +optimizing performance for real-world applications, while providing detailed +insights into frameworks and tools for implementing Agentic RAG + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in RAG and Agentic AI), 3 (Comparison implied through taxonomy and evolution of RAG paradigms), and 5 (Real-world Applications and Challenges in industries like finance and education). Also scores high on Additional Considerations for Agentic AI, Novelty, and Reproducibility (through detailed implementation strategies and frameworks). + +--- + +## [Clone-Robust AI Alignment](https://arxiv.org/abs/2501.09254v1) +**arXiv ID:** 2501.09254v1 + +**Abstract:** +> A key challenge in training Large Language Models (LLMs) is properly aligning +them with human preferences. Reinforcement Learning with Human Feedback (RLHF) +uses pairwise comparisons from human annotators to train reward functions and +has emerged as a popular alignment method. However, input datasets in RLHF are +not necessarily balanced in the types of questions and answers that are +included. Therefore, we want RLHF algorithms to perform well even when the set +of alternatives is not uniformly distributed. Drawing on insights from social +choice theory, we introduce robustness to approximate clones, a desirable +property of RLHF algorithms which requires that adding near-duplicate +alternatives does not significantly change the learned reward function. We +first demonstrate that the standard RLHF algorithm based on regularized maximum +likelihood estimation (MLE) fails to satisfy this property. We then propose the +weighted MLE, a new RLHF algorithm that modifies the standard regularized MLE +by weighting alternatives based on their similarity to other alternatives. This +new algorithm guarantees robustness to approximate clones while preserving +desirable theoretical properties. + +**Decision Explanation:** Original response: ``` +{ + "decision": "ACCEPT", + "explanation": "Meets all required criteria (1, 2, 3) with practical applications in AI alignment, experimental results with quantitative metrics implied, and comparison with existing state-of-the-art techniques. Additionally, it introduces novel approaches (novelty), with potential implementability and impact through experimental validation." +} +``` + +--- + +## [To Retrieve or Not to Retrieve? Uncertainty Detection for Dynamic + Retrieval Augmented Generation](https://arxiv.org/abs/2501.09292v1) +**arXiv ID:** 2501.09292v1 + +**Abstract:** +> Retrieval-Augmented Generation equips large language models with the +capability to retrieve external knowledge, thereby mitigating hallucinations by +incorporating information beyond the model's intrinsic abilities. However, most +prior works have focused on invoking retrieval deterministically, which makes +it unsuitable for tasks such as long-form question answering. Instead, +dynamically performing retrieval by invoking it only when the underlying LLM +lacks the required knowledge can be more efficient. In this context, we delve +deeper into the question, "To Retrieve or Not to Retrieve?" by exploring +multiple uncertainty detection methods. We evaluate these methods for the task +of long-form question answering, employing dynamic retrieval, and present our +comparisons. Our findings suggest that uncertainty detection metrics, such as +Degree Matrix Jaccard and Eccentricity, can reduce the number of retrieval +calls by almost half, with only a slight reduction in question-answering +accuracy. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in Retrieval-Augmented Generation), 2 (Experimental Results with quantitative metrics on reducing retrieval calls and question-answering accuracy), and 3 (Comparison with State-of-the-Art methodologies in uncertainty detection for dynamic retrieval). Additionally, shows novelty in applying uncertainty detection metrics to optimize LLM usage. + +--- + +## [A Study of In-Context-Learning-Based Text-to-SQL Errors](https://arxiv.org/abs/2501.09310v1) +**arXiv ID:** 2501.09310v1 + +**Abstract:** +> Large language models (LLMs) have been adopted to perform text-to-SQL tasks, +utilizing their in-context learning (ICL) capability to translate natural +language questions into structured query language (SQL). However, such a +technique faces correctness problems and requires efficient repairing +solutions. In this paper, we conduct the first comprehensive study of +text-to-SQL errors. Our study covers four representative ICL-based techniques, +five basic repairing methods, two benchmarks, and two LLM settings. We find +that text-to-SQL errors are widespread and summarize 29 error types of 7 +categories. We also find that existing repairing attempts have limited +correctness improvement at the cost of high computational overhead with many +mis-repairs. Based on the findings, we propose MapleRepair, a novel text-to-SQL +error detection and repairing framework. The evaluation demonstrates that +MapleRepair outperforms existing solutions by repairing 13.8% more queries with +neglectable mis-repairs and 67.4% less overhead. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: text-to-SQL tasks with LLMs), 2 (Experimental Results: quantitative metrics demonstrating MapleRepair's performance improvements), 3 (Comparison with State-of-the-Art: outperforms existing solutions), and 4 (Methodology and Implementation Details: clearly describes novel MapleRepair framework). Additionally, shows Novelty and robust Experimental Validation. + +--- + +## [Rational Tuning of LLM Cascades via Probabilistic Modeling](https://arxiv.org/abs/2501.09345v1) +**arXiv ID:** 2501.09345v1 + +**Abstract:** +> Understanding the reliability of large language models (LLMs) has recently +garnered significant attention. Given LLMs' propensity to hallucinate, as well +as their high sensitivity to prompt design, it is already challenging to +predict the performance of an individual LLM. However, the problem becomes more +complex for compound LLM systems such as cascades, where in addition to each +model's standalone performance, we must understand how the error rates of +different models interact. In this paper, we present a probabilistic model for +the joint performance distribution of a sequence of LLMs, which enables a +framework for rationally tuning the confidence thresholds of a LLM cascade +using continuous optimization. Compared to selecting confidence thresholds +using grid search, our parametric Markov-copula model significantly improves +runtime scaling with respect to the length of the cascade and the desired +resolution of the cost-error curve, turning them from intractable into +low-order polynomial. In addition, the optimal thresholds computed using our +continuous optimization-based algorithm increasingly outperform those found via +grid search as cascade length grows, improving the area under the cost-error +curve by 1.9% on average for cascades consisting of at least three models. +Overall, our Markov-copula model provides a rational basis for tuning LLM +cascade performance and points to the potential of probabilistic methods in +analyzing LLM systems. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Applications - tuning LLM cascades, (2) Experimental Results and Quantitative Metrics - demonstrated performance improvements with specific metrics, and (3) Comparison with State-of-the-Art - compared with grid search and showed advancements. Additionally, meets criteria (4) Methodology and Implementation Details, and shows Novelty in utilizing a probabilistic model for LLM cascade tuning. + +--- + +## [YETI (YET to Intervene) Proactive Interventions by Multimodal AI Agents + in Augmented Reality Tasks](https://arxiv.org/abs/2501.09355v1) +**arXiv ID:** 2501.09355v1 + +**Abstract:** +> Multimodal AI Agents are AI models that have the capability of interactively +and cooperatively assisting human users to solve day-to-day tasks. Augmented +Reality (AR) head worn devices can uniquely improve the user experience of +solving procedural day-to-day tasks by providing egocentric multimodal (audio +and video) observational capabilities to AI Agents. Such AR capabilities can +help AI Agents see and listen to actions that users take which can relate to +multimodal capabilities of human users. Existing AI Agents, either Large +Language Models (LLMs) or Multimodal Vision-Language Models (VLMs) are reactive +in nature, which means that models cannot take an action without reading or +listening to the human user's prompts. Proactivity of AI Agents on the other +hand can help the human user detect and correct any mistakes in agent observed +tasks, encourage users when they do tasks correctly or simply engage in +conversation with the user - akin to a human teaching or assisting a user. Our +proposed YET to Intervene (YETI) multimodal agent focuses on the research +question of identifying circumstances that may require the agent to intervene +proactively. This allows the agent to understand when it can intervene in a +conversation with human users that can help the user correct mistakes on tasks, +like cooking, using AR. Our YETI Agent learns scene understanding signals based +on interpretable notions of Structural Similarity (SSIM) on consecutive video +frames. We also define the alignment signal which the AI Agent can learn to +identify if the video frames corresponding to the user's actions on the task +are consistent with expected actions. These signals are used by our AI Agent to +determine when it should proactively intervene. We compare our results on the +instances of proactive intervention in the HoloAssist multimodal benchmark for +an expert agent guiding a user to complete procedural tasks. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in AR tasks), 2 (Experimental Results with quantitative metrics implied by benchmark comparison), and 3 (Comparison with State-of-the-Art via HoloAssist multimodal benchmark). Also aligns with Agentic AI consideration by enabling autonomous, proactive interventions. While it involves video processing, its primary focus is on multimodal AI agent interactions, making it an acceptable edge case. + +--- + +## [MoE$^2$: Optimizing Collaborative Inference for Edge Large Language + Models](https://arxiv.org/abs/2501.09410v1) +**arXiv ID:** 2501.09410v1 + +**Abstract:** +> Large language models (LLMs) have demonstrated remarkable capabilities across +a wide range of natural language processing tasks. Exploiting the heterogeneous +capabilities of edge LLMs is crucial for diverse emerging applications, as it +enables greater cost-effectiveness and reduced latency. In this work, we +introduce \textit{Mixture-of-Edge-Experts (MoE$^2$)}, a novel collaborative +inference framework for edge LLMs. We formulate the joint gating and expert +selection problem to optimize inference performance under energy and latency +constraints. Unlike conventional MoE problems, LLM expert selection is +significantly more challenging due to the combinatorial nature and the +heterogeneity of edge LLMs across various attributes. To this end, we propose a +two-level expert selection mechanism through which we uncover an +optimality-preserving property of gating parameters across expert selections. +This property enables the decomposition of the training and selection +processes, significantly reducing complexity. Furthermore, we leverage the +objective's monotonicity and design a discrete monotonic optimization algorithm +for optimal expert selection. We implement edge servers with NVIDIA Jetson AGX +Orins and NVIDIA RTX 4090 GPUs, and perform extensive experiments. Our results +validate that performance improvements of various LLM models and show that our +MoE$^2$ method can achieve optimal trade-offs among different delay and energy +budgets, and outperforms baselines under various system resource constraints. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications: optimizing edge LLMs for cost-effectiveness and reduced latency), 2 (Experimental Results and Quantitative Metrics: extensive experiments with performance improvements), and 3 (Comparison with State-of-the-Art: outperforms baselines under various constraints). Additionally, shows novelty in introducing a collaborative inference framework (MoE$^2$) and has potential for implementability with current standard tools. + +--- + +## [CarMem: Enhancing Long-Term Memory in LLM Voice Assistants through + Category-Bounding](https://arxiv.org/abs/2501.09645v1) +**arXiv ID:** 2501.09645v1 + +**Abstract:** +> In today's assistant landscape, personalisation enhances interactions, +fosters long-term relationships, and deepens engagement. However, many systems +struggle with retaining user preferences, leading to repetitive user requests +and disengagement. Furthermore, the unregulated and opaque extraction of user +preferences in industry applications raises significant concerns about privacy +and trust, especially in regions with stringent regulations like Europe. In +response to these challenges, we propose a long-term memory system for voice +assistants, structured around predefined categories. This approach leverages +Large Language Models to efficiently extract, store, and retrieve preferences +within these categories, ensuring both personalisation and transparency. We +also introduce a synthetic multi-turn, multi-session conversation dataset +(CarMem), grounded in real industry data, tailored to an in-car voice assistant +setting. Benchmarked on the dataset, our system achieves an F1-score of .78 to +.95 in preference extraction, depending on category granularity. Our +maintenance strategy reduces redundant preferences by 95% and contradictory +ones by 92%, while the accuracy of optimal retrieval is at .87. Collectively, +the results demonstrate the system's suitability for industrial applications. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria (1, 2, & 3): focuses on practical application of LLM in voice assistants, includes experimental results with quantitative metrics, and implicitly compares with existing state-of-the-art by highlighting improvements. Additionally meets optional criteria for methodology and implementation details (4), real-world applications (5), and novelty, with a clear approach to enhancing LLMs in a specific use case. + +--- + +## [Authenticated Delegation and Authorized AI Agents](https://arxiv.org/abs/2501.09674v1) +**arXiv ID:** 2501.09674v1 + +**Abstract:** +> The rapid deployment of autonomous AI agents creates urgent challenges around +authorization, accountability, and access control in digital spaces. New +standards are needed to know whom AI agents act on behalf of and guide their +use appropriately, protecting online spaces while unlocking the value of task +delegation to autonomous agents. We introduce a novel framework for +authenticated, authorized, and auditable delegation of authority to AI agents, +where human users can securely delegate and restrict the permissions and scope +of agents while maintaining clear chains of accountability. This framework +builds on existing identification and access management protocols, extending +OAuth 2.0 and OpenID Connect with agent-specific credentials and metadata, +maintaining compatibility with established authentication and web +infrastructure. Further, we propose a framework for translating flexible, +natural language permissions into auditable access control configurations, +enabling robust scoping of AI agent capabilities across diverse interaction +modalities. Taken together, this practical approach facilitates immediate +deployment of AI agents while addressing key security and accountability +concerns, working toward ensuring agentic AI systems perform only appropriate +actions and providing a tool for digital service providers to enable AI agent +interactions without risking harm from scalable interaction. + +**Decision Explanation:** Original decision: ACCEPT +Meets all required criteria: (1) Practical Applications (agentic AI, task delegation), (2) implied Experimental Results (though not detailed in abstract), and (3) Comparison with State-of-the-Art (extends OAuth 2.0 and OpenID Connect). Additionally, meets optional criteria (4) Methodology and Implementation Details, (5) Real-world Applications and Challenges, and shows Novelty in authenticated delegation for AI agents, making it a strong candidate for further review. + +--- + +## [ADAGE: A generic two-layer framework for adaptive agent based modelling](https://arxiv.org/abs/2501.09429v1) +**arXiv ID:** 2501.09429v1 + +**Abstract:** +> Agent-based models (ABMs) are valuable for modelling complex, potentially +out-of-equilibria scenarios. However, ABMs have long suffered from the Lucas +critique, stating that agent behaviour should adapt to environmental changes. +Furthermore, the environment itself often adapts to these behavioural changes, +creating a complex bi-level adaptation problem. Recent progress integrating +multi-agent reinforcement learning into ABMs introduces adaptive agent +behaviour, beginning to address the first part of this critique, however, the +approaches are still relatively ad hoc, lacking a general formulation, and +furthermore, do not tackle the second aspect of simultaneously adapting +environmental level characteristics in addition to the agent behaviours. In +this work, we develop a generic two-layer framework for ADaptive AGEnt based +modelling (ADAGE) for addressing these problems. This framework formalises the +bi-level problem as a Stackelberg game with conditional behavioural policies, +providing a consolidated framework for adaptive agent-based modelling based on +solving a coupled set of non-linear equations. We demonstrate how this generic +approach encapsulates several common (previously viewed as distinct) ABM tasks, +such as policy design, calibration, scenario generation, and robust behavioural +learning under one unified framework. We provide example simulations on +multiple complex economic and financial environments, showing the strength of +the novel framework under these canonical settings, addressing long-standing +critiques of traditional ABMs. + +**Decision Explanation:** Original decision: ACCEPT +Meets criteria 1 (Practical Applications in agentic AI), 2 ( Experimental Results, with simulations on complex environments), and 3 (Comparison with State-of-the-Art, addressing long-standing critiques of traditional ABMs). Also shows novelty in framework development and potential for real-world applications in economic and financial environments, aligning with additional considerations for impact and implementability. + +--- + +# Rejected Papers + +## [DualOpt: A Dual Divide-and-Optimize Algorithm for the Large-scale + Traveling Salesman Problem](https://arxiv.org/abs/2501.08565v1) +**arXiv ID:** 2501.08565v1 + +**Abstract:** +> This paper proposes a dual divide-and-optimize algorithm (DualOpt) for +solving the large-scale traveling salesman problem (TSP). DualOpt combines two +complementary strategies to improve both solution quality and computational +efficiency. The first strategy is a grid-based divide-and-conquer procedure +that partitions the TSP into smaller sub-problems, solving them in parallel and +iteratively refining the solution by merging nodes and partial routes. The +process continues until only one grid remains, yielding a high-quality initial +solution. The second strategy involves a path-based divide-and-optimize +procedure that further optimizes the solution by dividing it into sub-paths, +optimizing each using a neural solver, and merging them back to progressively +improve the overall solution. Extensive experiments conducted on two groups of +TSP benchmark instances, including randomly generated instances with up to +100,000 nodes and real-world datasets from TSPLIB, demonstrate the +effectiveness of DualOpt. The proposed DualOpt achieves highly competitive +results compared to 10 state-of-the-art algorithms in the literature. In +particular, DualOpt achieves an improvement gap up to 1.40% for the largest +instance TSP100K with a remarkable 104x speed-up over the leading heuristic +solver LKH3. Additionally, DualOpt demonstrates strong generalization on TSPLIB +benchmarks, confirming its capability to tackle diverse real-world TSP +applications. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, as it focuses on solving the Traveling Salesman Problem using a divide-and-optimize algorithm, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. Its novelty and achievements are within the optimization domain, not relevant to the specified LLM-related criteria. + +--- + +## [Development and Validation of the Provider Documentation Summarization + Quality Instrument for Large Language Models](https://arxiv.org/abs/2501.08977v1) +**arXiv ID:** 2501.08977v1 + +**Abstract:** +> As Large Language Models (LLMs) are integrated into electronic health record +(EHR) workflows, validated instruments are essential to evaluate their +performance before implementation. Existing instruments for provider +documentation quality are often unsuitable for the complexities of +LLM-generated text and lack validation on real-world data. The Provider +Documentation Summarization Quality Instrument (PDSQI-9) was developed to +evaluate LLM-generated clinical summaries. Multi-document summaries were +generated from real-world EHR data across multiple specialties using several +LLMs (GPT-4o, Mixtral 8x7b, and Llama 3-8b). Validation included Pearson +correlation for substantive validity, factor analysis and Cronbach's alpha for +structural validity, inter-rater reliability (ICC and Krippendorff's alpha) for +generalizability, a semi-Delphi process for content validity, and comparisons +of high- versus low-quality summaries for discriminant validity. Seven +physician raters evaluated 779 summaries and answered 8,329 questions, +achieving over 80% power for inter-rater reliability. The PDSQI-9 demonstrated +strong internal consistency (Cronbach's alpha = 0.879; 95% CI: 0.867-0.891) and +high inter-rater reliability (ICC = 0.867; 95% CI: 0.867-0.868), supporting +structural validity and generalizability. Factor analysis identified a 4-factor +model explaining 58% of the variance, representing organization, clarity, +accuracy, and utility. Substantive validity was supported by correlations +between note length and scores for Succinct (rho = -0.200, p = 0.029) and +Organized (rho = -0.190, p = 0.037). Discriminant validity distinguished high- +from low-quality summaries (p < 0.001). The PDSQI-9 demonstrates robust +construct validity, supporting its use in clinical practice to evaluate +LLM-generated summaries and facilitate safer integration of LLMs into +healthcare workflows. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, specifically the integration of Large Language Models into electronic health record workflows, which is explicitly listed as a rejection criterion. + +--- + +## [AI-based Identity Fraud Detection: A Systematic Review](https://arxiv.org/abs/2501.09239v1) +**arXiv ID:** 2501.09239v1 + +**Abstract:** +> With the rapid development of digital services, a large volume of personally +identifiable information (PII) is stored online and is subject to cyberattacks +such as Identity fraud. Most recently, the use of Artificial Intelligence (AI) +enabled deep fake technologies has significantly increased the complexity of +identity fraud. Fraudsters may use these technologies to create highly +sophisticated counterfeit personal identification documents, photos and videos. +These advancements in the identity fraud landscape pose challenges for identity +fraud detection and society at large. There is a pressing need to review and +understand identity fraud detection methods, their limitations and potential +solutions. This research aims to address this important need by using the +well-known systematic literature review method. This paper reviewed a selected +set of 43 papers across 4 major academic literature databases. In particular, +the review results highlight the two types of identity fraud prevention and +detection methods, in-depth and open challenges. The results were also +consolidated into a taxonomy of AI-based identity fraud detection and +prevention methods including key insights and trends. Overall, this paper +provides a foundational knowledge base to researchers and practitioners for +further research and development in this important area of digital identity +fraud. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on AI applications in law and security (identity fraud detection), which is excluded according to the criteria, and does not explicitly mention Large Language Models (LLMs) or related areas like knowledge graphs, RAG, or agentic AI. + +--- + +## [AI in Support of Diversity and Inclusion](https://arxiv.org/abs/2501.09534v1) +**arXiv ID:** 2501.09534v1 + +**Abstract:** +> In this paper, we elaborate on how AI can support diversity and inclusion and +exemplify research projects conducted in that direction. We start by looking at +the challenges and progress in making large language models (LLMs) more +transparent, inclusive, and aware of social biases. Even though LLMs like +ChatGPT have impressive abilities, they struggle to understand different +cultural contexts and engage in meaningful, human like conversations. A key +issue is that biases in language processing, especially in machine translation, +can reinforce inequality. Tackling these biases requires a multidisciplinary +approach to ensure AI promotes diversity, fairness, and inclusion. We also +highlight AI's role in identifying biased content in media, which is important +for improving representation. By detecting unequal portrayals of social groups, +AI can help challenge stereotypes and create more inclusive technologies. +Transparent AI algorithms, which clearly explain their decisions, are essential +for building trust and reducing bias in AI systems. We also stress AI systems +need diverse and inclusive training data. Projects like the Child Growth +Monitor show how using a wide range of data can help address real world +problems like malnutrition and poverty. We present a project that demonstrates +how AI can be applied to monitor the role of search engines in spreading +disinformation about the LGBTQ+ community. Moreover, we discuss the SignON +project as an example of how technology can bridge communication gaps between +hearing and deaf people, emphasizing the importance of collaboration and mutual +trust in developing inclusive AI. Overall, with this paper, we advocate for AI +systems that are not only effective but also socially responsible, promoting +fair and inclusive interactions between humans and machines. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on social applications of AI in regard to Diversity, Social harm, and similar issues, which is explicitly listed as a criterion for rejection. + +--- + +## [Artificial Intelligence-Driven Clinical Decision Support Systems](https://arxiv.org/abs/2501.09628v1) +**arXiv ID:** 2501.09628v1 + +**Abstract:** +> As artificial intelligence (AI) becomes increasingly embedded in healthcare +delivery, this chapter explores the critical aspects of developing reliable and +ethical Clinical Decision Support Systems (CDSS). Beginning with the +fundamental transition from traditional statistical models to sophisticated +machine learning approaches, this work examines rigorous validation strategies +and performance assessment methods, including the crucial role of model +calibration and decision curve analysis. The chapter emphasizes that creating +trustworthy AI systems in healthcare requires more than just technical +accuracy; it demands careful consideration of fairness, explainability, and +privacy. The challenge of ensuring equitable healthcare delivery through AI is +stressed, discussing methods to identify and mitigate bias in clinical +predictive models. The chapter then delves into explainability as a cornerstone +of human-centered CDSS. This focus reflects the understanding that healthcare +professionals must not only trust AI recommendations but also comprehend their +underlying reasoning. The discussion advances in an analysis of privacy +vulnerabilities in medical AI systems, from data leakage in deep learning +models to sophisticated attacks against model explanations. The text explores +privacy-preservation strategies such as differential privacy and federated +learning, while acknowledging the inherent trade-offs between privacy +protection and model performance. This progression, from technical validation +to ethical considerations, reflects the multifaceted challenges of developing +AI systems that can be seamlessly and reliably integrated into daily clinical +practice while maintaining the highest standards of patient care and data +protection. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, specifically Clinical Decision Support Systems, which falls under the excluded categories as per the evaluation criteria. + +--- + +## [Platform-Aware Mission Planning](https://arxiv.org/abs/2501.09632v1) +**arXiv ID:** 2501.09632v1 + +**Abstract:** +> Planning for autonomous systems typically requires reasoning with models at +different levels of abstraction, and the harmonization of two competing sets of +objectives: high-level mission goals that refer to an interaction of the system +with the external environment, and low-level platform constraints that aim to +preserve the integrity and the correct interaction of the subsystems. The +complicated interplay between these two models makes it very hard to reason on +the system as a whole, especially when the objective is to find plans with +robustness guarantees, considering the non-deterministic behavior of the lower +layers of the system. + In this paper, we introduce the problem of Platform-Aware Mission Planning +(PAMP), addressing it in the setting of temporal durative actions. The PAMP +problem differs from standard temporal planning for its exists-forall nature: +the high-level plan dealing with mission goals is required to satisfy safety +and executability constraints, for all the possible non-deterministic +executions of the low-level model of the platform and the environment. We +propose two approaches for solving PAMP. The first baseline approach +amalgamates the mission and platform levels, while the second is based on an +abstraction-refinement loop that leverages the combination of a planner and a +verification engine. We prove the soundness and completeness of the proposed +approaches and validate them experimentally, demonstrating the importance of +heterogeneous modeling and the superiority of the technique based on +abstraction-refinement. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it focuses on autonomous systems and mission planning without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the primary areas of interest. It lacks direct relevance to the specified LLM applications and related technologies. + +--- + +## [Electronic Health Records: Towards Digital Twins in Healthcare](https://arxiv.org/abs/2501.09640v1) +**arXiv ID:** 2501.09640v1 + +**Abstract:** +> The pivotal shift from traditional paper-based records to sophisticated +Electronic Health Records (EHR), enabled systematic collection and analysis of +patient data through descriptive statistics, providing insight into patterns +and trends across patient populations. This evolution continued toward +predictive analytics, allowing healthcare providers to anticipate patient +outcomes and potential complications before they occur. This progression from +basic digital record-keeping to sophisticated predictive modelling and digital +twins reflects healthcare's broader evolution toward more integrated, +patient-centred approaches that combine data-driven insights with personalized +care delivery. This chapter explores the evolution and significance of +healthcare information systems, beginning with an examination of the +implementation of EHR in the UK and the USA. It provides a comprehensive +overview of the International Classification of Diseases (ICD) system, tracing +its development from ICD-9 to ICD-10. Central to this discussion is the +MIMIC-III database, a landmark achievement in healthcare data sharing and +arguably the most comprehensive critical care database freely available to +researchers worldwide. MIMIC-III has democratized access to high-quality +healthcare data, enabling unprecedented opportunities for research and +analysis. The chapter examines its structure, clinical outcome analysis +capabilities, and practical applications through case studies, with a +particular focus on mortality and length of stay metrics, vital signs +extraction, and ICD coding. Through detailed entity-relationship diagrams and +practical examples, the text illustrates MIMIC's complex data structure and +demonstrates how different querying approaches can lead to subtly different +results, emphasizing the critical importance of understanding the database's +architecture for accurate data extraction. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, specifically Electronic Health Records and healthcare data analysis, which is explicitly listed as a reason for rejection. + +--- + +## [NS-Gym: Open-Source Simulation Environments and Benchmarks for + Non-Stationary Markov Decision Processes](https://arxiv.org/abs/2501.09646v1) +**arXiv ID:** 2501.09646v1 + +**Abstract:** +> In many real-world applications, agents must make sequential decisions in +environments where conditions are subject to change due to various exogenous +factors. These non-stationary environments pose significant challenges to +traditional decision-making models, which typically assume stationary dynamics. +Non-stationary Markov decision processes (NS-MDPs) offer a framework to model +and solve decision problems under such changing conditions. However, the lack +of standardized benchmarks and simulation tools has hindered systematic +evaluation and advance in this field. We present NS-Gym, the first simulation +toolkit designed explicitly for NS-MDPs, integrated within the popular +Gymnasium framework. In NS-Gym, we segregate the evolution of the environmental +parameters that characterize non-stationarity from the agent's decision-making +module, allowing for modular and flexible adaptations to dynamic environments. +We review prior work in this domain and present a toolkit encapsulating key +problem characteristics and types in NS-MDPs. This toolkit is the first effort +to develop a set of standardized interfaces and benchmark problems to enable +consistent and reproducible evaluation of algorithms under non-stationary +conditions. We also benchmark six algorithmic approaches from prior work on +NS-MDPs using NS-Gym. Our vision is that NS-Gym will enable researchers to +assess the adaptability and robustness of their decision-making algorithms to +non-stationary conditions. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on simulation environments for Non-Stationary Markov Decision Processes (NS-MDPs) without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the foundational criteria related to LLM applications. + +--- + +## [The Goofus & Gallant Story Corpus for Practical Value Alignment](https://arxiv.org/abs/2501.09707v1) +**arXiv ID:** 2501.09707v1 + +**Abstract:** +> Values or principles are key elements of human society that influence people +to behave and function according to an accepted standard set of social rules to +maintain social order. As AI systems are becoming ubiquitous in human society, +it is a major concern that they could violate these norms or values and +potentially cause harm. Thus, to prevent intentional or unintentional harm, AI +systems are expected to take actions that align with these principles. Training +systems to exhibit this type of behavior is difficult and often requires a +specialized dataset. This work presents a multi-modal dataset illustrating +normative and non-normative behavior in real-life situations described through +natural language and artistic images. This training set contains curated sets +of images that are designed to teach young children about social principles. We +argue that this is an ideal dataset to use for training socially normative +agents given this fact. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on value alignment and social norms in AI, which aligns with 'responsible AI application or AI ethics', a criterion for rejection. It also lacks clear connections to Large Language Models (LLMs), practical applications in knowledge graphs, RAG, or agentic AI, as emphasized in the selection criteria. + +--- + +## [KU AIGEN ICL EDI@BC8 Track 3: Advancing Phenotype Named Entity + Recognition and Normalization for Dysmorphology Physical Examination Reports](https://arxiv.org/abs/2501.09744v1) +**arXiv ID:** 2501.09744v1 + +**Abstract:** +> The objective of BioCreative8 Track 3 is to extract phenotypic key medical +findings embedded within EHR texts and subsequently normalize these findings to +their Human Phenotype Ontology (HPO) terms. However, the presence of diverse +surface forms in phenotypic findings makes it challenging to accurately +normalize them to the correct HPO terms. To address this challenge, we explored +various models for named entity recognition and implemented data augmentation +techniques such as synonym marginalization to enhance the normalization step. +Our pipeline resulted in an exact extraction and normalization F1 score 2.6\% +higher than the mean score of all submissions received in response to the +challenge. Furthermore, in terms of the normalization F1 score, our approach +surpassed the average performance by 1.9\%. These findings contribute to the +advancement of automated medical data extraction and normalization techniques, +showcasing potential pathways for future research and application in the +biomedical domain. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, specifically biomedical data extraction and normalization for Dysmorphology Physical Examination Reports, which is explicitly listed as a rejection criterion. + +--- + +## [Adapting Whisper for Regional Dialects: Enhancing Public Services for + Vulnerable Populations in the United Kingdom](https://arxiv.org/abs/2501.08502v1) +**arXiv ID:** 2501.08502v1 + +**Abstract:** +> We collect novel data in the public service domain to evaluate the capability +of the state-of-the-art automatic speech recognition (ASR) models in capturing +regional differences in accents in the United Kingdom (UK), specifically +focusing on two accents from Scotland with distinct dialects. This study +addresses real-world problems where biased ASR models can lead to +miscommunication in public services, disadvantaging individuals with regional +accents particularly those in vulnerable populations. We first examine the +out-of-the-box performance of the Whisper large-v3 model on a baseline dataset +and our data. We then explore the impact of fine-tuning Whisper on the +performance in the two UK regions and investigate the effectiveness of existing +model evaluation techniques for our real-world application through manual +inspection of model errors. We observe that the Whisper model has a higher word +error rate (WER) on our test datasets compared to the baseline data and +fine-tuning on a given data improves performance on the test dataset with the +same domain and accent. The fine-tuned models also appear to show improved +performance when applied to the test data outside of the region it was trained +on suggesting that fine-tuned models may be transferable within parts of the +UK. Our manual analysis of model outputs reveals the benefits and drawbacks of +using WER as an evaluation metric and fine-tuning to adapt to regional +dialects. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on automatic speech recognition (ASR) and regional dialects, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the core criteria for Practical Applications and Experimental Results in the context of LLMs. + +--- + +## [Mitigating Domain Shift in Federated Learning via Intra- and + Inter-Domain Prototypes](https://arxiv.org/abs/2501.08521v1) +**arXiv ID:** 2501.08521v1 + +**Abstract:** +> Federated Learning (FL) has emerged as a decentralized machine learning +technique, allowing clients to train a global model collaboratively without +sharing private data. However, most FL studies ignore the crucial challenge of +heterogeneous domains where each client has a distinct feature distribution, +which is common in real-world scenarios. Prototype learning, which leverages +the mean feature vectors within the same classes, has become a prominent +solution for federated learning under domain skew. However, existing federated +prototype learning methods only consider inter-domain prototypes on the server +and overlook intra-domain characteristics. In this work, we introduce a novel +federated prototype learning method, namely I$^2$PFL, which incorporates +$\textbf{I}$ntra-domain and $\textbf{I}$nter-domain $\textbf{P}$rototypes, to +mitigate domain shifts and learn a generalized global model across multiple +domains in federated learning. To construct intra-domain prototypes, we propose +feature alignment with MixUp-based augmented prototypes to capture the +diversity of local domains and enhance the generalization of local features. +Additionally, we introduce a reweighting mechanism for inter-domain prototypes +to generate generalized prototypes to provide inter-domain knowledge and reduce +domain skew across multiple clients. Extensive experiments on the Digits, +Office-10, and PACS datasets illustrate the superior performance of our method +compared to other baselines. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria as it focuses on Federated Learning (FL) and prototype learning, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the core areas of interest. + +--- + +## [Dynamic Portfolio Optimization via Augmented DDPG with Quantum Price + Levels-Based Trading Strategy](https://arxiv.org/abs/2501.08528v1) +**arXiv ID:** 2501.08528v1 + +**Abstract:** +> With the development of deep learning, Dynamic Portfolio Optimization (DPO) +problem has received a lot of attention in recent years, not only in the field +of finance but also in the field of deep learning. Some advanced research in +recent years has proposed the application of Deep Reinforcement Learning (DRL) +to the DPO problem, which demonstrated to be more advantageous than supervised +learning in solving the DPO problem. However, there are still certain unsolved +issues: 1) DRL algorithms usually have the problems of slow learning speed and +high sample complexity, which is especially problematic when dealing with +complex financial data. 2) researchers use DRL simply for the purpose of +obtaining high returns, but pay little attention to the problem of risk control +and trading strategy, which will affect the stability of model returns. In +order to address these issues, in this study we revamped the intrinsic +structure of the model based on the Deep Deterministic Policy Gradient (DDPG) +and proposed the Augmented DDPG model. Besides, we also proposed an innovative +risk control strategy based on Quantum Price Levels (QPLs) derived from Quantum +Finance Theory (QFT). Our experimental results revealed that our model has +better profitability as well as risk control ability with less sample +complexity in the DPO problem compared to the baseline models. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on financial applications using Deep Reinforcement Learning and Quantum Finance Theory, with no apparent involvement of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria related to LLMs. + +--- + +## [The Devil is in Temporal Token: High Quality Video Reasoning + Segmentation](https://arxiv.org/abs/2501.08549v1) +**arXiv ID:** 2501.08549v1 + +**Abstract:** +> Existing methods for Video Reasoning Segmentation rely heavily on a single +special token to represent the object in the keyframe or the entire video, +inadequately capturing spatial complexity and inter-frame motion. To overcome +these challenges, we propose VRS-HQ, an end-to-end video reasoning segmentation +approach that leverages Multimodal Large Language Models (MLLMs) to inject rich +spatiotemporal features into hierarchical tokens.Our key innovations include a +Temporal Dynamic Aggregation (TDA) and a Token-driven Keyframe Selection (TKS). +Specifically, we design frame-level and temporal-level tokens that +utilize MLLM's autoregressive learning to effectively capture both local and +global information. Subsequently, we apply a similarity-based weighted fusion +and frame selection strategy, then utilize SAM2 to perform keyframe +segmentation and propagation. To enhance keyframe localization accuracy, the +TKS filters keyframes based on SAM2's occlusion scores during inference. VRS-HQ +achieves state-of-the-art performance on ReVOS, surpassing VISA by +5.9%/12.5%/9.1% in J&F scores across the three subsets. These results highlight +the strong temporal reasoning and segmentation capabilities of our method. Code +and model weights will be released at VRS-HQ. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing (video reasoning segmentation), which is explicitly listed as a rejection criterion. Although it utilizes Large Language Models (MLLMs), the primary application is in video processing, outweighing the relevance to LLMs. + +--- + +## [Evaluating SAT and SMT Solvers on Large-Scale Sudoku Puzzles](https://arxiv.org/abs/2501.08569v1) +**arXiv ID:** 2501.08569v1 + +**Abstract:** +> Modern SMT solvers have revolutionized the approach to constraint +satisfaction problems by integrating advanced theory reasoning and encoding +techniques. In this work, we evaluate the performance of modern SMT solvers in +Z3, CVC5 and DPLL(T) against a standard SAT solver in DPLL. By benchmarking +these solvers on novel, diverse 25x25 Sudoku puzzles of various difficulty +levels created by our improved Sudoku generator, we examine the impact of +advanced theory reasoning and encoding techniques. Our findings demonstrate +that modern SMT solvers significantly outperform classical SAT solvers. This +work highlights the evolution of logical solvers and exemplifies the utility of +SMT solvers in addressing large-scale constraint satisfaction problems. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the specified criteria, as it focuses on the evaluation of SAT and SMT solvers for constraint satisfaction problems (Sudoku puzzles), without any apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Disjoint Processing Mechanisms of Hierarchical and Linear Grammars in + Large Language Models](https://arxiv.org/abs/2501.08618v1) +**arXiv ID:** 2501.08618v1 + +**Abstract:** +> All natural languages are structured hierarchically. In humans, this +structural restriction is neurologically coded: when two grammars are presented +with identical vocabularies, brain areas responsible for language processing +are only sensitive to hierarchical grammars. Using large language models +(LLMs), we investigate whether such functionally distinct hierarchical +processing regions can arise solely from exposure to large-scale language +distributions. We generate inputs using English, Italian, Japanese, or nonce +words, varying the underlying grammars to conform to either hierarchical or +linear/positional rules. Using these grammars, we first observe that language +models show distinct behaviors on hierarchical versus linearly structured +inputs. Then, we find that the components responsible for processing +hierarchical grammars are distinct from those that process linear grammars; we +causally verify this in ablation experiments. Finally, we observe that +hierarchy-selective components are also active on nonce grammars; this suggests +that hierarchy sensitivity is not tied to meaning, nor in-distribution inputs. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on the internal processing mechanisms of Large Language Models (LLMs) in response to different grammatical structures, without explicitly addressing practical applications, experimental results with quantitative metrics, comparison with state-of-the-art, or real-world challenges, thus not meeting the required criteria. + +--- + +## [ViBidirectionMT-Eval: Machine Translation for Vietnamese-Chinese and + Vietnamese-Lao language pair](https://arxiv.org/abs/2501.08621v1) +**arXiv ID:** 2501.08621v1 + +**Abstract:** +> This paper presents an results of the VLSP 2022-2023 Machine Translation +Shared Tasks, focusing on Vietnamese-Chinese and Vietnamese-Lao machine +translation. The tasks were organized as part of the 9th, 10th annual workshop +on Vietnamese Language and Speech Processing (VLSP 2022, VLSP 2023). The +objective of the shared task was to build machine translation systems, +specifically targeting Vietnamese-Chinese and Vietnamese-Lao translation +(corresponding to 4 translation directions). The submission were evaluated on +1,000 pairs for testing (news and general domains) using established metrics +like BLEU [11] and SacreBLEU [12]. Additionally, system outputs also were +evaluated with human judgment provided by experts in Chinese and Lao languages. +These human assessments played a crucial role in ranking the performance of the +machine translation models, ensuring a more comprehensive evaluation. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Machine Translation, without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the mandatory criteria (1, 2, and 3) and not strongly aligning with any of the additional preferred criteria. + +--- + +## [Reassessing the Role of Chain-of-Thought in Sentiment Analysis: Insights + and Limitations](https://arxiv.org/abs/2501.08641v1) +**arXiv ID:** 2501.08641v1 + +**Abstract:** +> The relationship between language and thought remains an unresolved +philosophical issue. Existing viewpoints can be broadly categorized into two +schools: one asserting their independence, and another arguing that language +constrains thought. In the context of large language models, this debate raises +a crucial question: Does a language model's grasp of semantic meaning depend on +thought processes? To explore this issue, we investigate whether reasoning +techniques can facilitate semantic understanding. Specifically, we +conceptualize thought as reasoning, employ chain-of-thought prompting as a +reasoning technique, and examine its impact on sentiment analysis tasks. The +experiments show that chain-of-thought has a minimal impact on sentiment +analysis tasks. Both the standard and chain-of-thought prompts focus on aspect +terms rather than sentiment in the generated content. Furthermore, +counterfactual experiments reveal that the model's handling of sentiment tasks +primarily depends on information from demonstrations. The experimental results +support the first viewpoint. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on the philosophical relationship between language and thought, with experiments centered on sentiment analysis, which does not clearly demonstrate practical applications of Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation, or agentic AI, nor does it explicitly address state-of-the-art comparisons or innovative methodological approaches as required by the main criteria. + +--- + +## [Application of Deep Reinforcement Learning to UAV Swarming for Ground + Surveillance](https://arxiv.org/abs/2501.08655v1) +**arXiv ID:** 2501.08655v1 + +**Abstract:** +> This paper summarizes in depth the state of the art of aerial swarms, +covering both classical and new reinforcement-learning-based approaches for +their management. Then, it proposes a hybrid AI system, integrating deep +reinforcement learning in a multi-agent centralized swarm architecture. The +proposed system is tailored to perform surveillance of a specific area, +searching and tracking ground targets, for security and law enforcement +applications. The swarm is governed by a central swarm controller responsible +for distributing different search and tracking tasks among the cooperating +UAVs. Each UAV agent is then controlled by a collection of cooperative +sub-agents, whose behaviors have been trained using different deep +reinforcement learning models, tailored for the different task types proposed +by the swarm controller. More specifically, proximal policy optimization (PPO) +algorithms were used to train the agents' behavior. In addition, several +metrics to assess the performance of the swarm in this application were +defined. The results obtained through simulation show that our system searches +the operation area effectively, acquires the targets in a reasonable time, and +is capable of tracking them continuously and consistently. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on the application of Deep Reinforcement Learning to UAV Swarming for Ground Surveillance, with a specific mention of law enforcement applications, which aligns with criteria for rejection (primarily focuses on law, either with AI as subject or participant). Additionally, there is no apparent connection to Large Language Models (LLMs), a key requirement for consideration. + +--- + +## [SPEQ: Stabilization Phases for Efficient Q-Learning in High + Update-To-Data Ratio Reinforcement Learning](https://arxiv.org/abs/2501.08669v1) +**arXiv ID:** 2501.08669v1 + +**Abstract:** +> A key challenge in Deep Reinforcement Learning is sample efficiency, +especially in real-world applications where collecting environment interactions +is expensive or risky. Recent off-policy algorithms improve sample efficiency +by increasing the Update-To-Data (UTD) ratio and performing more gradient +updates per environment interaction. While this improves sample efficiency, it +significantly increases computational cost due to the higher number of gradient +updates required. In this paper we propose a sample-efficient method to improve +computational efficiency by separating training into distinct learning phases +in order to exploit gradient updates more effectively. Our approach builds on +top of the Dropout Q-Functions (DroQ) algorithm and alternates between an +online, low UTD ratio training phase, and an offline stabilization phase. +During the stabilization phase, we fine-tune the Q-functions without collecting +new environment interactions. This process improves the effectiveness of the +replay buffer and reduces computational overhead. Our experimental results on +continuous control problems show that our method achieves results comparable to +state-of-the-art, high UTD ratio algorithms while requiring 56\% fewer gradient +updates and 50\% less training time than DroQ. Our approach offers an effective +and computationally economical solution while maintaining the same sample +efficiency as the more costly, high UTD ratio state-of-the-art. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria as it focuses on Deep Reinforcement Learning (Q-Learning) rather than Large Language Models (LLMs), and does not mention applications in knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the specified areas of interest. + +--- + +## [Digital Phenotyping for Adolescent Mental Health: A Feasibility Study + Employing Machine Learning to Predict Mental Health Risk From Active and + Passive Smartphone Data](https://arxiv.org/abs/2501.08851v1) +**arXiv ID:** 2501.08851v1 + +**Abstract:** +> Background: Adolescents are particularly vulnerable to mental disorders, with +over 75% of cases manifesting before the age of 25. Research indicates that +only 18 to 34% of young people experiencing high levels of depression or +anxiety symptoms seek support. Digital tools leveraging smartphones offer +scalable and early intervention opportunities. Objective: Using a novel machine +learning framework, this study evaluated the feasibility of integrating active +and passive smartphone data to predict mental disorders in non-clinical +adolescents. Specifically, we investigated the utility of the Mindcraft app in +predicting risks for internalising and externalising disorders, eating +disorders, insomnia and suicidal ideation. Methods: Participants (N=103; mean +age 16.1 years) were recruited from three London schools. Participants +completed the Strengths and Difficulties Questionnaire, the Eating Disorders-15 +Questionnaire, Sleep Condition Indicator Questionnaire and indicated the +presence/absence of suicidal ideation. They used the Mindcraft app for 14 days, +contributing active data via self-reports and passive data from smartphone +sensors. A contrastive pretraining phase was applied to enhance user-specific +feature stability, followed by supervised fine-tuning. The model evaluation +employed leave-one-subject-out cross-validation using balanced accuracy as the +primary metric. Results: The integration of active and passive data achieved +superior performance compared to individual data sources, with mean balanced +accuracies of 0.71 for SDQ-High risk, 0.67 for insomnia, 0.77 for suicidal +ideation and 0.70 for eating disorders. The contrastive learning framework +stabilised daily behavioural representations, enhancing predictive robustness. +This study demonstrates the potential of integrating active and passive +smartphone data with advanced machine-learning techniques for predicting mental +health risks. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (mental health diagnosis and risk prediction), which is explicitly listed as a rejection criterion. + +--- + +## [Silent Abandonment in Text-Based Contact Centers: Identifying, + Quantifying, and Mitigating its Operational Impacts](https://arxiv.org/abs/2501.08869v2) +**arXiv ID:** 2501.08869v2 + +**Abstract:** +> In the quest to improve services, companies offer customers the option to +interact with agents via texting. Such contact centers face unique challenges +compared to traditional call centers, as measuring customer experience proxies +like abandonment and patience involves uncertainty. A key source of this +uncertainty is silent abandonment, where customers leave without notifying the +system, wasting agent time and leaving their status unclear. Silent abandonment +also obscures whether a customer was served or left. Our goals are to measure +the magnitude of silent abandonment and mitigate its effects. Classification +models show that 3%-70% of customers across 17 companies abandon silently. In +one study, 71.3% of abandoning customers did so silently, reducing agent +efficiency by 3.2% and system capacity by 15.3%, incurring $5,457 in annual +costs per agent. We develop an expectation-maximization (EM) algorithm to +estimate customer patience under uncertainty and identify influencing +covariates. We find that companies should use classification models to estimate +abandonment scope and our EM algorithm to assess patience. We suggest +strategies to operationally mitigate the impact of silent abandonment by +predicting suspected silent-abandonment behavior or changing service design. +Specifically, we show that while allowing customers to write while waiting in +the queue creates a missing data challenge, it also significantly increases +patience and reduces service time, leading to reduced abandonment and lower +staffing requirements. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on operational improvements in text-based contact centers, without clear involvement of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, failing to meet the required criteria related to LLM applications. + +--- + +## [Projection Implicit Q-Learning with Support Constraint for Offline + Reinforcement Learning](https://arxiv.org/abs/2501.08907v1) +**arXiv ID:** 2501.08907v1 + +**Abstract:** +> Offline Reinforcement Learning (RL) faces a critical challenge of +extrapolation errors caused by out-of-distribution (OOD) actions. Implicit +Q-Learning (IQL) algorithm employs expectile regression to achieve in-sample +learning, effectively mitigating the risks associated with OOD actions. +However, the fixed hyperparameter in policy evaluation and density-based policy +improvement method limit its overall efficiency. In this paper, we propose +Proj-IQL, a projective IQL algorithm enhanced with the support constraint. In +the policy evaluation phase, Proj-IQL generalizes the one-step approach to a +multi-step approach through vector projection, while maintaining in-sample +learning and expectile regression framework. In the policy improvement phase, +Proj-IQL introduces support constraint that is more aligned with the policy +evaluation approach. Furthermore, we theoretically demonstrate that Proj-IQL +guarantees monotonic policy improvement and enjoys a progressively more +rigorous criterion for superior actions. Empirical results demonstrate the +Proj-IQL achieves state-of-the-art performance on D4RL benchmarks, especially +in challenging navigation domains. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on Offline Reinforcement Learning (RL) with Q-Learning, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the essential criteria (1, 2, and 3) and not aligning with any of the secondary criteria (4, 5) or additional considerations." +} +``` + +--- + +## [Modeling Melt Pool Features and Spatter Using Symbolic Regression and + Machine Learning](https://arxiv.org/abs/2501.08922v1) +**arXiv ID:** 2501.08922v1 + +**Abstract:** +> Additive manufacturing (AM) is a rapidly evolving technology that has +attracted applications across a wide range of fields due to its ability to +fabricate complex geometries. However, one of the key challenges in AM is +achieving consistent print quality. This inconsistency is often attributed to +uncontrolled melt pool dynamics, partly caused by spatter which can lead to +defects. Therefore, capturing and controlling the evolution of the melt pool is +crucial for enhancing process stability and part quality. In this study, we +developed a framework to support decision-making in AM operations, facilitating +quality control and minimizing defects via machine learning (ML) and polynomial +symbolic regression models. We implemented experimentally validated +computational tools as a cost-effective approach to collect large datasets from +laser powder bed fusion (LPBF) processes. For a dataset consisting of 281 +process conditions, parameters such as melt pool dimensions (length, width, +depth), melt pool geometry (area, volume), and volume indicated as spatter were +extracted. Using machine learning (ML) and polynomial symbolic regression +models, a high R2 of over 95 % was achieved in predicting the melt pool +dimensions and geometry features for both the training and testing datasets, +with either process conditions (power and velocity) or melt pool dimensions as +the model inputs. In the case of volume indicated as spatter, R2 improved after +logarithmic transforming the model inputs, which was either the process +conditions or the melt pool dimensions. Among the investigated ML models, the +ExtraTree model achieved the highest R2 values of 96.7 % and 87.5 %. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. It focuses on machine learning applications in additive manufacturing, which is outside the specified scope. + +--- + +## [Visual WetlandBirds Dataset: Bird Species Identification and Behavior + Recognition in Videos](https://arxiv.org/abs/2501.08931v1) +**arXiv ID:** 2501.08931v1 + +**Abstract:** +> The current biodiversity loss crisis makes animal monitoring a relevant field +of study. In light of this, data collected through monitoring can provide +essential insights, and information for decision-making aimed at preserving +global biodiversity. Despite the importance of such data, there is a notable +scarcity of datasets featuring videos of birds, and none of the existing +datasets offer detailed annotations of bird behaviors in video format. In +response to this gap, our study introduces the first fine-grained video dataset +specifically designed for bird behavior detection and species classification. +This dataset addresses the need for comprehensive bird video datasets and +provides detailed data on bird actions, facilitating the development of deep +learning models to recognize these, similar to the advancements made in human +action recognition. The proposed dataset comprises 178 videos recorded in +Spanish wetlands, capturing 13 different bird species performing 7 distinct +behavior classes. In addition, we also present baseline results using state of +the art models on two tasks: bird behavior recognition and species +classification. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing (bird species identification and behavior recognition in videos), which is one of the explicitly stated rejection criteria. Additionally, it does not demonstrate a clear connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, as required by the practical applications criteria. + +--- + +## [Analyzing the Ethical Logic of Six Large Language Models](https://arxiv.org/abs/2501.08951v1) +**arXiv ID:** 2501.08951v1 + +**Abstract:** +> This study examines the ethical reasoning of six prominent generative large +language models: OpenAI GPT-4o, Meta LLaMA 3.1, Perplexity, Anthropic Claude +3.5 Sonnet, Google Gemini, and Mistral 7B. The research explores how these +models articulate and apply ethical logic, particularly in response to moral +dilemmas such as the Trolley Problem, and Heinz Dilemma. Departing from +traditional alignment studies, the study adopts an explainability-transparency +framework, prompting models to explain their ethical reasoning. This approach +is analyzed through three established ethical typologies: the +consequentialist-deontological analytic, Moral Foundations Theory, and the +Kohlberg Stages of Moral Development Model. Findings reveal that LLMs exhibit +largely convergent ethical logic, marked by a rationalist, consequentialist +emphasis, with decisions often prioritizing harm minimization and fairness. +Despite similarities in pre-training and model architecture, a mixture of +nuanced and significant differences in ethical reasoning emerge across models, +reflecting variations in fine-tuning and post-training processes. The models +consistently display erudition, caution, and self-awareness, presenting ethical +reasoning akin to a graduate-level discourse in moral philosophy. In striking +uniformity these systems all describe their ethical reasoning as more +sophisticated than what is characteristic of typical human moral logic. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on AI ethics (ethical logic and moral dilemmas) which is explicitly listed as a rejection criterion, outweighing any potential marginal relevance to Large Language Models' practical applications or technical advancements. + +--- + +## [Kolmogorov-Arnold Networks for Time Series Granger Causality Inference](https://arxiv.org/abs/2501.08958v1) +**arXiv ID:** 2501.08958v1 + +**Abstract:** +> We introduce Granger Causality Kolmogorov-Arnold Networks (GCKAN), an +innovative architecture that extends the recently proposed Kolmogorov-Arnold +Networks (KAN) to the domain of causal inference. By extracting base weights +from KAN layers and incorporating the sparsity-inducing penalty along with +ridge regularization, GCKAN infers the Granger causality from time series while +enabling automatic time lag selection. Additionally, we propose an algorithm +leveraging time-reversed Granger causality to enhance inference accuracy. The +algorithm compares prediction and sparse-inducing losses derived from the +original and time-reversed series, automatically selecting the casual +relationship with the higher score or integrating the results to mitigate +spurious connectivities. Comprehensive experiments conducted on Lorenz-96, gene +regulatory networks, fMRI BOLD signals, and VAR datasets demonstrate that the +proposed model achieves competitive performance to state-of-the-art methods in +inferring Granger causality from nonlinear, high-dimensional, and +limited-sample time series. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria (1-3) as it does not focus on Large Language Models (LLMs), lacks comparison with state-of-the-art LLM techniques, and does not provide experimental results related to LLMs. Additionally, the paper's topic (time series Granger causality inference) does not align with the specified areas of interest (knowledge graphs, retrieval-augmented generation, agentic AI, etc.). + +--- + +## [An analysis of data variation and bias in image-based dermatological + datasets for machine learning classification](https://arxiv.org/abs/2501.08962v1) +**arXiv ID:** 2501.08962v1 + +**Abstract:** +> AI algorithms have become valuable in aiding professionals in healthcare. The +increasing confidence obtained by these models is helpful in critical decision +demands. In clinical dermatology, classification models can detect malignant +lesions on patients' skin using only RGB images as input. However, most +learning-based methods employ data acquired from dermoscopic datasets on +training, which are large and validated by a gold standard. Clinical models aim +to deal with classification on users' smartphone cameras that do not contain +the corresponding resolution provided by dermoscopy. Also, clinical +applications bring new challenges. It can contain captures from uncontrolled +environments, skin tone variations, viewpoint changes, noises in data and +labels, and unbalanced classes. A possible alternative would be to use transfer +learning to deal with the clinical images. However, as the number of samples is +low, it can cause degradations on the model's performance; the source +distribution used in training differs from the test set. This work aims to +evaluate the gap between dermoscopic and clinical samples and understand how +the dataset variations impact training. It assesses the main differences +between distributions that disturb the model's prediction. Finally, from +experiments on different architectures, we argue how to combine the data from +divergent distributions, decreasing the impact on the model's final accuracy. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (dermatological datasets for machine learning classification), which is explicitly listed as a criterion for rejection, outweighing any potential marginal relevance to Large Language Models (LLMs) or related areas. + +--- + +## [How Do Generative Models Draw a Software Engineer? A Case Study on + Stable Diffusion Bias](https://arxiv.org/abs/2501.09014v1) +**arXiv ID:** 2501.09014v1 + +**Abstract:** +> Generative models are nowadays widely used to generate graphical content used +for multiple purposes, e.g. web, art, advertisement. However, it has been shown +that the images generated by these models could reinforce societal biases +already existing in specific contexts. In this paper, we focus on understanding +if this is the case when one generates images related to various software +engineering tasks. In fact, the Software Engineering (SE) community is not +immune from gender and ethnicity disparities, which could be amplified by the +use of these models. Hence, if used without consciousness, artificially +generated images could reinforce these biases in the SE domain. Specifically, +we perform an extensive empirical evaluation of the gender and ethnicity bias +exposed by three versions of the Stable Diffusion (SD) model (a very popular +open-source text-to-image model) - SD 2, SD XL, and SD 3 - towards SE tasks. We +obtain 6,720 images by feeding each model with two sets of prompts describing +different software-related tasks: one set includes the Software Engineer +keyword, and one set does not include any specification of the person +performing the task. Next, we evaluate the gender and ethnicity disparities in +the generated images. Results show how all models are significantly biased +towards male figures when representing software engineers. On the contrary, +while SD 2 and SD XL are strongly biased towards White figures, SD 3 is +slightly more biased towards Asian figures. Nevertheless, all models +significantly under-represent Black and Arab figures, regardless of the prompt +style used. The results of our analysis highlight severe concerns about +adopting those models to generate content for SE tasks and open the field for +future research on bias mitigation in this context. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on social applications of AI in regard to social harm (bias and diversity issues), which is one of the specified rejection criteria. Additionally, it does not explicitly meet the required criteria for practical applications of Large Language Models (LLMs) in areas like knowledge graphs, RAG, or agentic AI. + +--- + +## [TCMM: Token Constraint and Multi-Scale Memory Bank of Contrastive + Learning for Unsupervised Person Re-identification](https://arxiv.org/abs/2501.09044v1) +**arXiv ID:** 2501.09044v1 + +**Abstract:** +> This paper proposes the ViT Token Constraint and Multi-scale Memory bank +(TCMM) method to address the patch noises and feature inconsistency in +unsupervised person re-identification works. Many excellent methods use ViT +features to obtain pseudo labels and clustering prototypes, then train the +model with contrastive learning. However, ViT processes images by performing +patch embedding, which inevitably introduces noise in patches and may +compromise the performance of the re-identification model. On the other hand, +previous memory bank based contrastive methods may lead data inconsistency due +to the limitation of batch size. Furthermore, existing pseudo label methods +often discard outlier samples that are difficult to cluster. It sacrifices the +potential value of outlier samples, leading to limited model diversity and +robustness. This paper introduces the ViT Token Constraint to mitigate the +damage caused by patch noises to the ViT architecture. The proposed Multi-scale +Memory enhances the exploration of outlier samples and maintains feature +consistency. Experimental results demonstrate that our system achieves +state-of-the-art performance on common benchmarks. The project is available at +\href{https://github.com/andy412510/TCMM}{https://github.com/andy412510/TCMM}. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on computer vision (person re-identification) and image processing (via Vision Transformer (ViT) features and patch embedding), rather than Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. It also lacks clear connections to the specified application areas. + +--- + +## [Generating Realistic Synthetic Head Rotation Data for Extended Reality + using Deep Learning](https://arxiv.org/abs/2501.09050v1) +**arXiv ID:** 2501.09050v1 + +**Abstract:** +> Extended Reality is a revolutionary method of delivering multimedia content +to users. A large contributor to its popularity is the sense of immersion and +interactivity enabled by having real-world motion reflected in the virtual +experience accurately and immediately. This user motion, mainly caused by head +rotations, induces several technical challenges. For instance, which content is +generated and transmitted depends heavily on where the user is looking. +Seamless systems, taking user motion into account proactively, will therefore +require accurate predictions of upcoming rotations. Training and evaluating +such predictors requires vast amounts of orientational input data, which is +expensive to gather, as it requires human test subjects. A more feasible +approach is to gather a modest dataset through test subjects, and then extend +it to a more sizeable set using synthetic data generation methods. In this +work, we present a head rotation time series generator based on TimeGAN, an +extension of the well-known Generative Adversarial Network, designed +specifically for generating time series. This approach is able to extend a +dataset of head rotations with new samples closely matching the distribution of +the measured time series. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on generating synthetic data for Extended Reality using Deep Learning, without apparent connection to Large Language Models (LLMs), practical applications of LLMs, or prescribed areas like knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Polyp detection in colonoscopy images using YOLOv11](https://arxiv.org/abs/2501.09051v1) +**arXiv ID:** 2501.09051v1 + +**Abstract:** +> Colorectal cancer (CRC) is one of the most commonly diagnosed cancers all +over the world. It starts as a polyp in the inner lining of the colon. To +prevent CRC, early polyp detection is required. Colonosopy is used for the +inspection of the colon. Generally, the images taken by the camera placed at +the tip of the endoscope are analyzed by the experts manually. Various +traditional machine learning models have been used with the rise of machine +learning. Recently, deep learning models have shown more effectiveness in polyp +detection due to their superiority in generalizing and learning small features. +These deep learning models for object detection can be segregated into two +different types: single-stage and two-stage. Generally, two stage models have +higher accuracy than single stage ones but the single stage models have low +inference time. Hence, single stage models are easy to use for quick object +detection. YOLO is one of the singlestage models used successfully for polyp +detection. It has drawn the attention of researchers because of its lower +inference time. The researchers have used Different versions of YOLO so far, +and with each newer version, the accuracy of the model is increasing. This +paper aims to see the effectiveness of the recently released YOLOv11 to detect +polyp. We analyzed the performance for all five models of YOLOv11 (YOLO11n, +YOLO11s, YOLO11m, YOLO11l, YOLO11x) with Kvasir dataset for the training and +testing. Two different versions of the dataset were used. The first consisted +of the original dataset, and the other was created using augmentation +techniques. The performance of all the models with these two versions of the +dataset have been analysed. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (polyp detection in colonoscopy images) which is explicitly mentioned as a rejection criterion, and does not meet any of the specified criteria related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Average-Reward Reinforcement Learning with Entropy Regularization](https://arxiv.org/abs/2501.09080v1) +**arXiv ID:** 2501.09080v1 + +**Abstract:** +> The average-reward formulation of reinforcement learning (RL) has drawn +increased interest in recent years due to its ability to solve +temporally-extended problems without discounting. Independently, RL algorithms +have benefited from entropy-regularization: an approach used to make the +optimal policy stochastic, thereby more robust to noise. Despite the distinct +benefits of the two approaches, the combination of entropy regularization with +an average-reward objective is not well-studied in the literature and there has +been limited development of algorithms for this setting. To address this gap in +the field, we develop algorithms for solving entropy-regularized average-reward +RL problems with function approximation. We experimentally validate our method, +comparing it with existing algorithms on standard benchmarks for RL. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on Average-Reward Reinforcement Learning with Entropy Regularization, which does not explicitly mention Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the primary criteria (1, 2, and 3) and not addressing any of the secondary criteria (4, 5) in the context of LLMs. + +--- + +## [Inferring Transition Dynamics from Value Functions](https://arxiv.org/abs/2501.09081v1) +**arXiv ID:** 2501.09081v1 + +**Abstract:** +> In reinforcement learning, the value function is typically trained to solve +the Bellman equation, which connects the current value to future values. This +temporal dependency hints that the value function may contain implicit +information about the environment's transition dynamics. By rearranging the +Bellman equation, we show that a converged value function encodes a model of +the underlying dynamics of the environment. We build on this insight to propose +a simple method for inferring dynamics models directly from the value function, +potentially mitigating the need for explicit model learning. Furthermore, we +explore the challenges of next-state identifiability, discussing conditions +under which the inferred dynamics model is well-defined. Our work provides a +theoretical foundation for leveraging value functions in dynamics modeling and +opens a new avenue for bridging model-free and model-based reinforcement +learning. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it focuses on reinforcement learning (value functions and transition dynamics) without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, which are the primary areas of interest. + +--- + +## [Mantis Shrimp: Exploring Photometric Band Utilization in Computer Vision + Networks for Photometric Redshift Estimation](https://arxiv.org/abs/2501.09112v1) +**arXiv ID:** 2501.09112v1 + +**Abstract:** +> We present Mantis Shrimp, a multi-survey deep learning model for photometric +redshift estimation that fuses ultra-violet (GALEX), optical (PanSTARRS), and +infrared (UnWISE) imagery. Machine learning is now an established approach for +photometric redshift estimation, with generally acknowledged higher performance +in areas with a high density of spectroscopically identified galaxies over +template-based methods. Multiple works have shown that image-based +convolutional neural networks can outperform tabular-based color/magnitude +models. In comparison to tabular models, image models have additional design +complexities: it is largely unknown how to fuse inputs from different +instruments which have different resolutions or noise properties. The Mantis +Shrimp model estimates the conditional density estimate of redshift using +cutout images. The density estimates are well calibrated and the point +estimates perform well in the distribution of available spectroscopically +confirmed galaxies with (bias = 1e-2), scatter (NMAD = 2.44e-2) and +catastrophic outlier rate ($\eta$=17.53$\%$). We find that early fusion +approaches (e.g., resampling and stacking images from different instruments) +match the performance of late fusion approaches (e.g., concatenating latent +space representations), so that the design choice ultimately is left to the +user. Finally, we study how the models learn to use information across bands, +finding evidence that our models successfully incorporates information from all +surveys. The applicability of our model to the analysis of large populations of +galaxies is limited by the speed of downloading cutouts from external servers; +however, our model could be useful in smaller studies such as generating priors +over redshift for stellar population synthesis. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on computer vision and photometric redshift estimation in astronomy, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, thus failing to meet the core criteria. + +--- + +## [Generative Medical Image Anonymization Based on Latent Code Projection + and Optimization](https://arxiv.org/abs/2501.09114v1) +**arXiv ID:** 2501.09114v1 + +**Abstract:** +> Medical image anonymization aims to protect patient privacy by removing +identifying information, while preserving the data utility to solve downstream +tasks. In this paper, we address the medical image anonymization problem with a +two-stage solution: latent code projection and optimization. In the projection +stage, we design a streamlined encoder to project input images into a latent +space and propose a co-training scheme to enhance the projection process. In +the optimization stage, we refine the latent code using two deep loss functions +designed to address the trade-off between identity protection and data utility +dedicated to medical images. Through a comprehensive set of qualitative and +quantitative experiments, we showcase the effectiveness of our approach on the +MIMIC-CXR chest X-ray dataset by generating anonymized synthetic images that +can serve as training set for detecting lung pathologies. Source codes are +available at https://github.com/Huiyu-Li/GMIA. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, specifically medical image anonymization, which is explicitly listed as a rejection criterion. + +--- + +## [Grounding Text-To-Image Diffusion Models For Controlled High-Quality + Image Generation](https://arxiv.org/abs/2501.09194v1) +**arXiv ID:** 2501.09194v1 + +**Abstract:** +> Large-scale text-to-image (T2I) diffusion models have demonstrated an +outstanding performance in synthesizing diverse high-quality visuals from +natural language text captions. Multiple layout-to-image models have been +developed to control the generation process by utilizing a broad array of +layouts such as segmentation maps, edges, and human keypoints. In this work, we +present ObjectDiffusion, a model that takes inspirations from the top +cutting-edge image generative frameworks to seamlessly condition T2I models +with new bounding boxes capabilities. Specifically, we make substantial +modifications to the network architecture introduced in ContorlNet to integrate +it with the condition processing and injection techniques proposed in GLIGEN. +ObjectDiffusion is initialized with pretraining parameters to leverage the +generation knowledge obtained from training on large-scale datasets. We +fine-tune ObjectDiffusion on the COCO2017 training dataset and evaluate it on +the COCO2017 validation dataset. Our model achieves an AP$_{50}$ of 46.6, an AR +of 44.5, and a FID of 19.8 outperforming the current SOTA model trained on +open-source datasets in all of the three metrics. ObjectDiffusion demonstrates +a distinctive capability in synthesizing diverse, high-quality, high-fidelity +images that seamlessly conform to the semantic and spatial control layout. +Evaluated in qualitative and quantitative tests, ObjectDiffusion exhibits +remarkable grounding abilities on closed-set and open-set settings across a +wide variety of contexts. The qualitative assessment verifies the ability of +ObjectDiffusion to generate multiple objects of different sizes and locations. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on image generation (text-to-image diffusion models) without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the mandatory criteria (1, 2, and 3) due to lack of direct relevance to LLMs and specified applications. + +--- + +## [Interpretable Droplet Digital PCR Assay for Trustworthy Molecular + Diagnostics](https://arxiv.org/abs/2501.09218v1) +**arXiv ID:** 2501.09218v1 + +**Abstract:** +> Accurate molecular quantification is essential for advancing research and +diagnostics in fields such as infectious diseases, cancer biology, and genetic +disorders. Droplet digital PCR (ddPCR) has emerged as a gold standard for +achieving absolute quantification. While computational ddPCR technologies have +advanced significantly, achieving automatic interpretation and consistent +adaptability across diverse operational environments remains a challenge. To +address these limitations, we introduce the intelligent interpretable droplet +digital PCR (I2ddPCR) assay, a comprehensive framework integrating front-end +predictive models (for droplet segmentation and classification) with GPT-4o +multimodal large language model (MLLM, for context-aware explanations and +recommendations) to automate and enhance ddPCR image analysis. This approach +surpasses the state-of-the-art models, affording 99.05% accuracy in processing +complex ddPCR images containing over 300 droplets per image with varying +signal-to-noise ratios (SNRs). By combining specialized neural networks and +large language models, the I2ddPCR assay offers a robust and adaptable solution +for absolute molecular quantification, achieving a sensitivity capable of +detecting low-abundance targets as low as 90.32 copies/{\mu}L. Furthermore, it +improves model's transparency through detailed explanation and troubleshooting +guidance, empowering users to make informed decisions. This innovative +framework has the potential to benefit molecular diagnostics, disease research, +and clinical applications, especially in resource-constrained settings. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (molecular diagnostics, infectious diseases, cancer biology, and genetic disorders), which is explicitly listed as a rejection criterion. + +--- + +## [SEAL: Entangled White-box Watermarks on Low-Rank Adaptation](https://arxiv.org/abs/2501.09284v1) +**arXiv ID:** 2501.09284v1 + +**Abstract:** +> Recently, LoRA and its variants have become the de facto strategy for +training and sharing task-specific versions of large pretrained models, thanks +to their efficiency and simplicity. However, the issue of copyright protection +for LoRA weights, especially through watermark-based techniques, remains +underexplored. To address this gap, we propose SEAL (SEcure wAtermarking on +LoRA weights), the universal whitebox watermarking for LoRA. SEAL embeds a +secret, non-trainable matrix between trainable LoRA weights, serving as a +passport to claim ownership. SEAL then entangles the passport with the LoRA +weights through training, without extra loss for entanglement, and distributes +the finetuned weights after hiding the passport. When applying SEAL, we +observed no performance degradation across commonsense reasoning, +textual/visual instruction tuning, and text-to-image synthesis tasks. We +demonstrate that SEAL is robust against a variety of known attacks: removal, +obfuscation, and ambiguity attacks. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on copyright protection and watermarking for Low-Rank Adaptation (LoRA) weights, which does not meet the required criteria for Large Language Models (LLMs) applications in knowledge graphs, retrieval-augmented generation, or agentic AI, and does not demonstrate practical applications, experimental results, or comparisons to state-of-the-art techniques in the specified areas. + +--- + +## [On Learning Informative Trajectory Embeddings for Imitation, + Classification and Regression](https://arxiv.org/abs/2501.09327v1) +**arXiv ID:** 2501.09327v1 + +**Abstract:** +> In real-world sequential decision making tasks like autonomous driving, +robotics, and healthcare, learning from observed state-action trajectories is +critical for tasks like imitation, classification, and clustering. For example, +self-driving cars must replicate human driving behaviors, while robots and +healthcare systems benefit from modeling decision sequences, whether or not +they come from expert data. Existing trajectory encoding methods often focus on +specific tasks or rely on reward signals, limiting their ability to generalize +across domains and tasks. Inspired by the success of embedding models like CLIP +and BERT in static domains, we propose a novel method for embedding +state-action trajectories into a latent space that captures the skills and +competencies in the dynamic underlying decision-making processes. This method +operates without the need for reward labels, enabling better generalization +across diverse domains and tasks. Our contributions are threefold: (1) We +introduce a trajectory embedding approach that captures multiple abilities from +state-action data. (2) The learned embeddings exhibit strong representational +power across downstream tasks, including imitation, classification, clustering, +and regression. (3) The embeddings demonstrate unique properties, such as +controlling agent behaviors in IQ-Learn and an additive structure in the latent +space. Experimental results confirm that our method outperforms traditional +approaches, offering more flexible and powerful trajectory representations for +various applications. Our code is available at +https://github.com/Erasmo1015/vte. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on sequential decision making tasks in autonomous driving, robotics, and healthcare, without clear involvement of Large Language Models (LLMs) or their applications in knowledge graphs, retrieval-augmented generation, or agentic AI, failing to meet the core criteria. + +--- + +## [Neural Honeytrace: A Robust Plug-and-Play Watermarking Framework against + Model Extraction Attacks](https://arxiv.org/abs/2501.09328v1) +**arXiv ID:** 2501.09328v1 + +**Abstract:** +> Developing high-performance deep learning models is resource-intensive, +leading model owners to utilize Machine Learning as a Service (MLaaS) platforms +instead of publicly releasing their models. However, malicious users may +exploit query interfaces to execute model extraction attacks, reconstructing +the target model's functionality locally. While prior research has investigated +triggerable watermarking techniques for asserting ownership, existing methods +face significant challenges: (1) most approaches require additional training, +resulting in high overhead and limited flexibility, and (2) they often fail to +account for advanced attackers, leaving them vulnerable to adaptive attacks. + In this paper, we propose Neural Honeytrace, a robust plug-and-play +watermarking framework against model extraction attacks. We first formulate a +watermark transmission model from an information-theoretic perspective, +providing an interpretable account of the principles and limitations of +existing triggerable watermarking. Guided by the model, we further introduce: +(1) a similarity-based training-free watermarking method for plug-and-play and +flexible watermarking, and (2) a distribution-based multi-step watermark +information transmission strategy for robust watermarking. Comprehensive +experiments on four datasets demonstrate that Neural Honeytrace outperforms +previous methods in efficiency and resisting adaptive attacks. Neural +Honeytrace reduces the average number of samples required for a worst-case +t-Test-based copyright claim from $12,000$ to $200$ with zero training cost. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on model protection and copyright infringement (watermarking against model extraction attacks), which does not align with the specified areas of interest (Large Language Models, knowledge graphs, retrieval-augmented generation, agentic AI, etc.). It does not meet the practical applications criteria related to LLMs, nor does it align with the specified inclusions, making it outside the scope of the evaluation criteria. + +--- + +## [Prompt-CAM: A Simpler Interpretable Transformer for Fine-Grained + Analysis](https://arxiv.org/abs/2501.09333v1) +**arXiv ID:** 2501.09333v1 + +**Abstract:** +> We present a simple usage of pre-trained Vision Transformers (ViTs) for +fine-grained analysis, aiming to identify and localize the traits that +distinguish visually similar categories, such as different bird species or dog +breeds. Pre-trained ViTs such as DINO have shown remarkable capabilities to +extract localized, informative features. However, using saliency maps like +Grad-CAM can hardly point out the traits: they often locate the whole object by +a blurred, coarse heatmap, not traits. We propose a novel approach Prompt Class +Attention Map (Prompt-CAM) to the rescue. Prompt-CAM learns class-specific +prompts to a pre-trained ViT and uses the corresponding outputs for +classification. To classify an image correctly, the true-class prompt must +attend to the unique image patches not seen in other classes' images, i.e., +traits. As such, the true class's multi-head attention maps reveal traits and +their locations. Implementation-wise, Prompt-CAM is almost a free lunch by +simply modifying the prediction head of Visual Prompt Tuning (VPT). This makes +Prompt-CAM fairly easy to train and apply, sharply contrasting other +interpretable methods that design specific models and training processes. It is +even simpler than the recently published INterpretable TRansformer (INTR), +whose encoder-decoder architecture prevents it from leveraging pre-trained +ViTs. Extensive empirical studies on a dozen datasets from various domains +(e.g., birds, fishes, insects, fungi, flowers, food, and cars) validate +Prompt-CAM superior interpretation capability. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on computer vision applications using Vision Transformers (ViTs) for fine-grained image analysis, with no clear connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, thus not meeting the core criteria. + +--- + +## [Predicting Air Temperature from Volumetric Urban Morphology with Machine + Learning](https://arxiv.org/abs/2501.09469v1) +**arXiv ID:** 2501.09469v1 + +**Abstract:** +> In this study, we firstly introduce a method that converts CityGML data into +voxels which works efficiently and fast in high resolution for large scale +datasets such as cities but by sacrificing some building details to overcome +the limitations of previous voxelization methodologies that have been +computationally intensive and inefficient at transforming large-scale urban +areas into voxel representations for high resolution. Those voxelized 3D city +data from multiple cities and corresponding air temperature data are used to +develop a machine learning model. Before the model training, Gaussian blurring +is implemented on input data to consider spatial relationships, as a result the +correlation rate between air temperature and volumetric building morphology is +also increased after the Gaussian blurring. After the model training, the +prediction results are not just evaluated with Mean Square Error (MSE) but some +image similarity metrics such as Structural Similarity Index Measure (SSIM) and +Learned Perceptual Image Patch Similarity (LPIPS) that are able to detect and +consider spatial relations during the evaluation process. This trained model is +capable of predicting the spatial distribution of air temperature by using +building volume information of corresponding pixel as input. By doing so, this +research aims to assist urban planners in incorporating environmental +parameters into their planning strategies, thereby facilitating more +sustainable and inhabitable urban environments. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on a machine learning application (predicting air temperature from urban morphology) that does not involve Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not mention any of the specified areas of interest. + +--- + +## [Class Incremental Fault Diagnosis under Limited Fault Data via + Supervised Contrastive Knowledge Distillation](https://arxiv.org/abs/2501.09525v1) +**arXiv ID:** 2501.09525v1 + +**Abstract:** +> Class-incremental fault diagnosis requires a model to adapt to new fault +classes while retaining previous knowledge. However, limited research exists +for imbalanced and long-tailed data. Extracting discriminative features from +few-shot fault data is challenging, and adding new fault classes often demands +costly model retraining. Moreover, incremental training of existing methods +risks catastrophic forgetting, and severe class imbalance can bias the model's +decisions toward normal classes. To tackle these issues, we introduce a +Supervised Contrastive knowledge distiLlation for class Incremental Fault +Diagnosis (SCLIFD) framework proposing supervised contrastive knowledge +distillation for improved representation learning capability and less +forgetting, a novel prioritized exemplar selection method for sample replay to +alleviate catastrophic forgetting, and the Random Forest Classifier to address +the class imbalance. Extensive experimentation on simulated and real-world +industrial datasets across various imbalance ratios demonstrates the +superiority of SCLIFD over existing approaches. Our code can be found at +https://github.com/Zhang-Henry/SCLIFD_TII. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on fault diagnosis, which does not explicitly involve Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the core criteria of practical applications of LLMs." +} +``` + +--- + +## [Text-driven Adaptation of Foundation Models for Few-shot Surgical + Workflow Analysis](https://arxiv.org/abs/2501.09555v1) +**arXiv ID:** 2501.09555v1 + +**Abstract:** +> Purpose: Surgical workflow analysis is crucial for improving surgical +efficiency and safety. However, previous studies rely heavily on large-scale +annotated datasets, posing challenges in cost, scalability, and reliance on +expert annotations. To address this, we propose Surg-FTDA (Few-shot Text-driven +Adaptation), designed to handle various surgical workflow analysis tasks with +minimal paired image-label data. + Methods: Our approach has two key components. First, Few-shot selection-based +modality alignment selects a small subset of images and aligns their embeddings +with text embeddings from the downstream task, bridging the modality gap. +Second, Text-driven adaptation leverages only text data to train a decoder, +eliminating the need for paired image-text data. This decoder is then applied +to aligned image embeddings, enabling image-related tasks without explicit +image-text pairs. + Results: We evaluate our approach to generative tasks (image captioning) and +discriminative tasks (triplet recognition and phase recognition). Results show +that Surg-FTDA outperforms baselines and generalizes well across downstream +tasks. + Conclusion: We propose a text-driven adaptation approach that mitigates the +modality gap and handles multiple downstream tasks in surgical workflow +analysis, with minimal reliance on large annotated datasets. The code and +dataset will be released in https://github.com/TingxuanSix/Surg-FTDA. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (surgical workflow analysis), which is one of the specified rejection criteria, outweighing its potential alignment with criteria like practical applications, experimental results, and novelty. + +--- + +## [IFRA: a machine learning-based Instrumented Fall Risk Assessment Scale + derived from Instrumented Timed Up and Go test in stroke patients](https://arxiv.org/abs/2501.09595v1) +**arXiv ID:** 2501.09595v1 + +**Abstract:** +> Effective fall risk assessment is critical for post-stroke patients. The +present study proposes a novel, data-informed fall risk assessment method based +on the instrumented Timed Up and Go (ITUG) test data, bringing in many mobility +measures that traditional clinical scales fail to capture. IFRA, which stands +for Instrumented Fall Risk Assessment, has been developed using a two-step +process: first, features with the highest predictive power among those +collected in a ITUG test have been identified using machine learning +techniques; then, a strategy is proposed to stratify patients into low, medium, +or high-risk strata. The dataset used in our analysis consists of 142 +participants, out of which 93 were used for training (15 synthetically +generated), 17 for validation and 32 to test the resulting IFRA scale (22 +non-fallers and 10 fallers). Features considered in the IFRA scale include gait +speed, vertical acceleration during sit-to-walk transition, and turning angular +velocity, which align well with established literature on the risk of fall in +neurological patients. In a comparison with traditional clinical scales such as +the traditional Timed Up & Go and the Mini-BESTest, IFRA demonstrates +competitive performance, being the only scale to correctly assign more than +half of the fallers to the high-risk stratum (Fischer's Exact test p = 0.004). +Despite the dataset's limited size, this is the first proof-of-concept study to +pave the way for future evidence regarding the use of IFRA tool for continuous +patient monitoring and fall prevention both in clinical stroke rehabilitation +and at home post-discharge. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (stroke patient fall risk assessment), which is one of the specified rejection criteria, outweighing its potential marginal relevance to the broader criteria of practical applications or methodology. + +--- + +## [Reducing the Sensitivity of Neural Physics Simulators to Mesh Topology + via Pretraining](https://arxiv.org/abs/2501.09597v1) +**arXiv ID:** 2501.09597v1 + +**Abstract:** +> Meshes are used to represent complex objects in high fidelity physics +simulators across a variety of domains, such as radar sensing and aerodynamics. +There is growing interest in using neural networks to accelerate physics +simulations, and also a growing body of work on applying neural networks +directly to irregular mesh data. Since multiple mesh topologies can represent +the same object, mesh augmentation is typically required to handle topological +variation when training neural networks. Due to the sensitivity of physics +simulators to small changes in mesh shape, it is challenging to use these +augmentations when training neural network-based physics simulators. In this +work, we show that variations in mesh topology can significantly reduce the +performance of neural network simulators. We evaluate whether pretraining can +be used to address this issue, and find that employing an established +autoencoder pretraining technique with graph embedding models reduces the +sensitivity of neural network simulators to variations in mesh topology. +Finally, we highlight future research directions that may further reduce neural +simulator sensitivity to mesh topology. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on neural physics simulators and mesh topology in the context of physics simulations (e.g., radar sensing, aerodynamics), with no apparent direct connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, failing to meet the core criteria." +} +``` + +--- + +## [Monte Carlo Tree Search with Velocity Obstacles for safe and efficient + motion planning in dynamic environments](https://arxiv.org/abs/2501.09649v1) +**arXiv ID:** 2501.09649v1 + +**Abstract:** +> Online motion planning is a challenging problem for intelligent robots moving +in dense environments with dynamic obstacles, e.g., crowds. In this work, we +propose a novel approach for optimal and safe online motion planning with +minimal information about dynamic obstacles. Specifically, our approach +requires only the current position of the obstacles and their maximum speed, +but it does not need any information about their exact trajectories or dynamic +model. The proposed methodology combines Monte Carlo Tree Search (MCTS), for +online optimal planning via model simulations, with Velocity Obstacles (VO), +for obstacle avoidance. We perform experiments in a cluttered simulated +environment with walls, and up to 40 dynamic obstacles moving with random +velocities and directions. With an ablation study, we show the key contribution +of VO in scaling up the efficiency of MCTS, selecting the safest and most +rewarding actions in the tree of simulations. Moreover, we show the superiority +of our methodology with respect to state-of-the-art planners, including +Non-linear Model Predictive Control (NMPC), in terms of improved collision +rate, computational and task performance. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria, as it focuses on motion planning for robots in dynamic environments, utilizing Monte Carlo Tree Search and Velocity Obstacles, without any apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Robin: a Suite of Multi-Scale Vision-Language Models and the CHIRP + Evaluation Benchmark](https://arxiv.org/abs/2501.09672v1) +**arXiv ID:** 2501.09672v1 + +**Abstract:** +> The proliferation of Vision-Language Models (VLMs) in the past several years +calls for rigorous and comprehensive evaluation methods and benchmarks. This +work analyzes existing VLM evaluation techniques, including automated metrics, +AI-based assessments, and human evaluations across diverse tasks. We first +introduce Robin - a novel suite of VLMs that we built by combining Large +Language Models (LLMs) and Vision Encoders (VEs) at multiple scales, and use +Robin to identify shortcomings of current evaluation approaches across scales. +Next, to overcome the identified limitations, we introduce CHIRP - a new long +form response benchmark we developed for more robust and complete VLM +evaluation. We provide open access to the Robin training code, model suite, and +CHIRP benchmark to promote reproducibility and advance VLM research. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on Vision-Language Models (VLMs) and their evaluation, with Large Language Models (LLMs) being a component rather than the main focus, and does not clearly meet the practical application criteria in areas specified (knowledge graphs, RAG, or agentic AI), nor does it explicitly address comparison with state-of-the-art in LLMs. + +--- + +## [Cueless EEG imagined speech for subject identification: dataset and + benchmarks](https://arxiv.org/abs/2501.09700v1) +**arXiv ID:** 2501.09700v1 + +**Abstract:** +> Electroencephalogram (EEG) signals have emerged as a promising modality for +biometric identification. While previous studies have explored the use of +imagined speech with semantically meaningful words for subject identification, +most have relied on additional visual or auditory cues. In this study, we +introduce a cueless EEG-based imagined speech paradigm, where subjects imagine +the pronunciation of semantically meaningful words without any external cues. +This innovative approach addresses the limitations of prior methods by +requiring subjects to select and imagine words from a predefined list +naturally. The dataset comprises over 4,350 trials from 11 subjects across five +sessions. We assess a variety of classification methods, including traditional +machine learning techniques such as Support Vector Machines (SVM) and XGBoost, +as well as time-series foundation models and deep learning architectures +specifically designed for EEG classification, such as EEG Conformer and Shallow +ConvNet. A session-based hold-out validation strategy was employed to ensure +reliable evaluation and prevent data leakage. Our results demonstrate +outstanding classification accuracy, reaching 97.93%. These findings highlight +the potential of cueless EEG paradigms for secure and reliable subject +identification in real-world applications, such as brain-computer interfaces +(BCIs). + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on biometric identification using EEG signals, which is unrelated to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and does not mention any practical applications of LLMs. + +--- + +## [Learnings from Scaling Visual Tokenizers for Reconstruction and + Generation](https://arxiv.org/abs/2501.09755v1) +**arXiv ID:** 2501.09755v1 + +**Abstract:** +> Visual tokenization via auto-encoding empowers state-of-the-art image and +video generative models by compressing pixels into a latent space. Although +scaling Transformer-based generators has been central to recent advances, the +tokenizer component itself is rarely scaled, leaving open questions about how +auto-encoder design choices influence both its objective of reconstruction and +downstream generative performance. Our work aims to conduct an exploration of +scaling in auto-encoders to fill in this blank. To facilitate this exploration, +we replace the typical convolutional backbone with an enhanced Vision +Transformer architecture for Tokenization (ViTok). We train ViTok on +large-scale image and video datasets far exceeding ImageNet-1K, removing data +constraints on tokenizer scaling. We first study how scaling the auto-encoder +bottleneck affects both reconstruction and generation -- and find that while it +is highly correlated with reconstruction, its relationship with generation is +more complex. We next explored the effect of separately scaling the +auto-encoders' encoder and decoder on reconstruction and generation +performance. Crucially, we find that scaling the encoder yields minimal gains +for either reconstruction or generation, while scaling the decoder boosts +reconstruction but the benefits for generation are mixed. Building on our +exploration, we design ViTok as a lightweight auto-encoder that achieves +competitive performance with state-of-the-art auto-encoders on ImageNet-1K and +COCO reconstruction tasks (256p and 512p) while outperforming existing +auto-encoders on 16-frame 128p video reconstruction for UCF-101, all with 2-5x +fewer FLOPs. When integrated with Diffusion Transformers, ViTok demonstrates +competitive performance on image generation for ImageNet-1K and sets new +state-of-the-art benchmarks for class-conditional video generation on UCF-101. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing and image generation, with an emphasis on scaling visual tokenizers for reconstruction and generation, which falls under the rejection criteria of primarily focusing on video processing. + +--- + +## [Exploring the Efficacy of Meta-Learning: Unveiling Superior Data + Diversity Utilization of MAML Over Pre-training](https://arxiv.org/abs/2501.08506v1) +**arXiv ID:** 2501.08506v1 + +**Abstract:** +> Currently, data and model size dominate the narrative in the training of +super-large, powerful models. However, there has been a lack of exploration on +the effect of other attributes of the training dataset on model performance. We +hypothesize that dataset diversity can impact the performance of vision models. +Our study shows positive correlations between test set accuracy and data +diversity, providing an argument for furthering the research of dataset +attributes beyond size. We analyzed pre-training and model-agnostic +meta-learning methods on twelve popular visual datasets (e.g., Omniglot, +CIFAR-FS, Aircraft) and five model configurations, including MAML variants with +different numbers of inner gradient steps and supervised learning. We show +moderate to strong positive correlations (R-squared: 0.15-0.42) between +accuracy and data diversity and weaker but significant correlations (R-squared: +~0.2) between loss and diversity. These findings support our hypothesis and +demonstrate a promising way for a deeper exploration of how formal data +diversity influences model performance. This initial study highlights the +potential of (Task2Vec) data diversity as a valuable measure in the rapidly +evolving field of large-scale learning and emphasizes that understanding the +dataset is key to building more powerful and generalizable models. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on computer vision (visual datasets) and the impact of dataset diversity on vision model performance, which aligns with 'Primarily focuses on video processing' (or closely related fields), and does not explicitly involve Large Language Models (LLMs) or their applications as required by the main criteria. + +--- + +## [Easing Seasickness through Attention Redirection with a + Mindfulness-Based Brain--Computer Interface](https://arxiv.org/abs/2501.08518v1) +**arXiv ID:** 2501.08518v1 + +**Abstract:** +> Seasickness is a prevalent issue that adversely impacts both passenger +experiences and the operational efficiency of maritime crews. While techniques +that redirect attention have proven effective in alleviating motion sickness +symptoms in terrestrial environments, applying similar strategies to manage +seasickness poses unique challenges due to the prolonged and intense motion +environment associated with maritime travel. In this study, we propose a +mindfulness brain-computer interface (BCI), specifically designed to redirect +attention with the aim of mitigating seasickness symptoms in real-world +settings. Our system utilizes a single-channel headband to capture prefrontal +EEG signals, which are then wirelessly transmitted to computing devices for the +assessment of mindfulness states. The results are transferred into real-time +feedback as mindfulness scores and audiovisual stimuli, facilitating a shift in +attentional focus from physiological discomfort to mindfulness practices. A +total of 43 individuals participated in a real-world maritime experiment +consisted of three sessions: a real-feedback mindfulness session, a resting +session, and a pseudofeedback mindfulness session. Notably, 81.39% of +participants reported that the mindfulness BCI intervention was effective, and +there was a significant reduction in the severity of seasickness, as measured +by the Misery Scale (MISC). Furthermore, EEG analysis revealed a decrease in +the theta/beta ratio, corresponding with the alleviation of seasickness +symptoms. A decrease in overall EEG band power during the real-feedback +mindfulness session suggests that the mindfulness BCI fosters a more tranquil +and downregulated state of brain activity. Together, this study presents a +novel nonpharmacological, portable, and effective approach for seasickness +intervention, with the potential to enhance the cruising experience for both +passengers and crews. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on a medical application (seasickness alleviation) and brain-computer interface technology, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, thus failing to meet the fundamental criteria. + +--- + +## [Reinforcement Learning-Enhanced Procedural Generation for Dynamic + Narrative-Driven AR Experiences](https://arxiv.org/abs/2501.08552v1) +**arXiv ID:** 2501.08552v1 + +**Abstract:** +> Procedural Content Generation (PCG) is widely used to create scalable and +diverse environments in games. However, existing methods, such as the Wave +Function Collapse (WFC) algorithm, are often limited to static scenarios and +lack the adaptability required for dynamic, narrative-driven applications, +particularly in augmented reality (AR) games. This paper presents a +reinforcement learning-enhanced WFC framework designed for mobile AR +environments. By integrating environment-specific rules and dynamic tile weight +adjustments informed by reinforcement learning (RL), the proposed method +generates maps that are both contextually coherent and responsive to gameplay +needs. Comparative evaluations and user studies demonstrate that the framework +achieves superior map quality and delivers immersive experiences, making it +well-suited for narrative-driven AR games. Additionally, the method holds +promise for broader applications in education, simulation training, and +immersive extended reality (XR) experiences, where dynamic and adaptive +environments are critical. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on procedural content generation for gaming and augmented reality (AR) experiences, with broader applications in education and simulation training, but does not explicitly meet any of the required criteria related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Towards Lightweight and Stable Zero-shot TTS with Self-distilled + Representation Disentanglement](https://arxiv.org/abs/2501.08566v1) +**arXiv ID:** 2501.08566v1 + +**Abstract:** +> Zero-shot Text-To-Speech (TTS) synthesis shows great promise for personalized +voice customization through voice cloning. However, current methods for +achieving zero-shot TTS heavily rely on large model scales and extensive +training datasets to ensure satisfactory performance and generalizability +across various speakers. This raises concerns regarding both deployment costs +and data security. In this paper, we present a lightweight and stable zero-shot +TTS system. We introduce a novel TTS architecture designed to effectively model +linguistic content and various speaker attributes from source speech and prompt +speech, respectively. Furthermore, we present a two-stage self-distillation +framework that constructs parallel data pairs for effectively disentangling +linguistic content and speakers from the perspective of training data. +Extensive experiments show that our system exhibits excellent performance and +superior stability on the zero-shot TTS tasks. Moreover, it shows markedly +superior computational efficiency, with RTFs of 0.13 and 0.012 on the CPU and +GPU, respectively. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on Text-To-Speech (TTS) synthesis, which is not explicitly mentioned as a relevant application area in the provided criteria, and does not clearly relate to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Sound Scene Synthesis at the DCASE 2024 Challenge](https://arxiv.org/abs/2501.08587v1) +**arXiv ID:** 2501.08587v1 + +**Abstract:** +> This paper presents Task 7 at the DCASE 2024 Challenge: sound scene +synthesis. Recent advances in sound synthesis and generative models have +enabled the creation of realistic and diverse audio content. We introduce a +standardized evaluation framework for comparing different sound scene synthesis +systems, incorporating both objective and subjective metrics. The challenge +attracted four submissions, which are evaluated using the Fr\'echet Audio +Distance (FAD) and human perceptual ratings. Our analysis reveals significant +insights into the current capabilities and limitations of sound scene synthesis +systems, while also highlighting areas for future improvement in this rapidly +evolving field. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on sound scene synthesis, which relates to audio processing and does not explicitly involve Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus not meeting any of the required criteria. + +--- + +## [OpenMLDB: A Real-Time Relational Data Feature Computation System for + Online ML](https://arxiv.org/abs/2501.08591v1) +**arXiv ID:** 2501.08591v1 + +**Abstract:** +> Efficient and consistent feature computation is crucial for a wide range of +online ML applications. Typically, feature computation is divided into two +distinct phases, i.e., offline stage for model training and online stage for +model serving. These phases often rely on execution engines with different +interface languages and function implementations, causing significant +inconsistencies. Moreover, many online ML features involve complex time-series +computations (e.g., functions over varied-length table windows) that differ +from standard streaming and analytical queries. Existing data processing +systems (e.g., Spark, Flink, DuckDB) often incur multi-second latencies for +these computations, making them unsuitable for real-time online ML applications +that demand timely feature updates. + This paper presents OpenMLDB, a feature computation system deployed in +4Paradigm's SageOne platform and over 100 real scenarios. Technically, OpenMLDB +first employs a unified query plan generator for consistent computation results +across the offline and online stages, significantly reducing feature deployment +overhead. Second, OpenMLDB provides an online execution engine that resolves +performance bottlenecks caused by long window computations (via +pre-aggregation) and multi-table window unions (via data self-adjusting). It +also provides a high-performance offline execution engine with window parallel +optimization and time-aware data skew resolving. Third, OpenMLDB features a +compact data format and stream-focused indexing to maximize memory usage and +accelerate data access. Evaluations in testing and real workloads reveal +significant performance improvements and resource savings compared to the +baseline systems. The open community of OpenMLDB now has over 150 contributors +and gained 1.6k stars on GitHub. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on optimizing feature computation for online Machine Learning (ML) applications using a relational data system, without explicitly mentioning Large Language Models (LLMs), their applications, or related challenges, thereby not meeting the core criteria." +} + +--- + +## [Fine-grained Spatio-temporal Event Prediction with Self-adaptive Anchor + Graph](https://arxiv.org/abs/2501.08653v1) +**arXiv ID:** 2501.08653v1 + +**Abstract:** +> Event prediction tasks often handle spatio-temporal data distributed in a +large spatial area. Different regions in the area exhibit different +characteristics while having latent correlations. This spatial heterogeneity +and correlations greatly affect the spatio-temporal distributions of event +occurrences, which has not been addressed by state-of-the-art models. Learning +spatial dependencies of events in a continuous space is challenging due to its +fine granularity and a lack of prior knowledge. In this work, we propose a +novel Graph Spatio-Temporal Point Process (GSTPP) model for fine-grained event +prediction. It adopts an encoder-decoder architecture that jointly models the +state dynamics of spatially localized regions using neural Ordinary +Differential Equations (ODEs). The state evolution is built on the foundation +of a novel Self-Adaptive Anchor Graph (SAAG) that captures spatial +dependencies. By adaptively localizing the anchor nodes in the space and +jointly constructing the correlation edges between them, the SAAG enhances the +model's ability of learning complex spatial event patterns. The proposed GSTPP +model greatly improves the accuracy of fine-grained event prediction. Extensive +experimental results show that our method greatly improves the prediction +accuracy over existing spatio-temporal event prediction approaches. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria for Practical Applications of Large Language Models (LLMs), as it focuses on Spatio-temporal Event Prediction using Graph Neural Networks and Ordinary Differential Equations, without any apparent connection to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Self-supervised Transformation Learning for Equivariant Representations](https://arxiv.org/abs/2501.08712v1) +**arXiv ID:** 2501.08712v1 + +**Abstract:** +> Unsupervised representation learning has significantly advanced various +machine learning tasks. In the computer vision domain, state-of-the-art +approaches utilize transformations like random crop and color jitter to achieve +invariant representations, embedding semantically the same inputs despite +transformations. However, this can degrade performance in tasks requiring +precise features, such as localization or flower classification. To address +this, recent research incorporates equivariant representation learning, which +captures transformation-sensitive information. However, current methods depend +on transformation labels and thus struggle with interdependency and complex +transformations. We propose Self-supervised Transformation Learning (STL), +replacing transformation labels with transformation representations derived +from image pairs. The proposed method ensures transformation representation is +image-invariant and learns corresponding equivariant transformations, enhancing +performance without increased batch complexity. We demonstrate the approach's +effectiveness across diverse classification and detection tasks, outperforming +existing methods in 7 out of 11 benchmarks and excelling in detection. By +integrating complex transformations like AugMix, unusable by prior equivariant +methods, this approach enhances performance across tasks, underscoring its +adaptability and resilience. Additionally, its compatibility with various base +models highlights its flexibility and broad applicability. The code is +available at https://github.com/jaemyung-u/stl. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on computer vision and image processing (e.g., transformation learning, image classification, detection tasks), which does not align with the specified areas of interest (Large Language Models, knowledge graphs, retrieval-augmented generation, agentic AI). None of the mandatory criteria (1-3) are met, and the paper's topic falls outside the scope of the evaluation criteria. + +--- + +## [Networked Agents in the Dark: Team Value Learning under Partial + Observability](https://arxiv.org/abs/2501.08778v1) +**arXiv ID:** 2501.08778v1 + +**Abstract:** +> We propose a novel cooperative multi-agent reinforcement learning (MARL) +approach for networked agents. In contrast to previous methods that rely on +complete state information or joint observations, our agents must learn how to +reach shared objectives under partial observability. During training, they +collect individual rewards and approximate a team value function through local +communication, resulting in cooperative behavior. To describe our problem, we +introduce the networked dynamic partially observable Markov game framework, +where agents communicate over a switching topology communication network. Our +distributed method, DNA-MARL, uses a consensus mechanism for local +communication and gradient descent for local computation. DNA-MARL increases +the range of the possible applications of networked agents, being well-suited +for real world domains that impose privacy and where the messages may not reach +their recipients. We evaluate DNA-MARL across benchmark MARL scenarios. Our +results highlight the superior performance of DNA-MARL over previous methods. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper does not meet the core criteria related to Large Language Models (LLMs), instead focusing on multi-agent reinforcement learning (MARL) and networked agents, with no apparent connection to LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI as specified in the selection criteria." +} +``` + +--- + +## [XMusic: Towards a Generalized and Controllable Symbolic Music Generation + Framework](https://arxiv.org/abs/2501.08809v1) +**arXiv ID:** 2501.08809v1 + +**Abstract:** +> In recent years, remarkable advancements in artificial intelligence-generated +content (AIGC) have been achieved in the fields of image synthesis and text +generation, generating content comparable to that produced by humans. However, +the quality of AI-generated music has not yet reached this standard, primarily +due to the challenge of effectively controlling musical emotions and ensuring +high-quality outputs. This paper presents a generalized symbolic music +generation framework, XMusic, which supports flexible prompts (i.e., images, +videos, texts, tags, and humming) to generate emotionally controllable and +high-quality symbolic music. XMusic consists of two core components, XProjector +and XComposer. XProjector parses the prompts of various modalities into +symbolic music elements (i.e., emotions, genres, rhythms and notes) within the +projection space to generate matching music. XComposer contains a Generator and +a Selector. The Generator generates emotionally controllable and melodious +music based on our innovative symbolic music representation, whereas the +Selector identifies high-quality symbolic music by constructing a multi-task +learning scheme involving quality assessment, emotion recognition, and genre +recognition tasks. In addition, we build XMIDI, a large-scale symbolic music +dataset that contains 108,023 MIDI files annotated with precise emotion and +genre labels. Objective and subjective evaluations show that XMusic +significantly outperforms the current state-of-the-art methods with impressive +music quality. Our XMusic has been awarded as one of the nine Highlights of +Collectibles at WAIC 2023. The project homepage of XMusic is +https://xmusic-project.github.io. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on music generation, which is not explicitly mentioned in the criteria for acceptance, and does not clearly meet any of the required criteria (1-3) related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, despite its innovative approach and state-of-the-art results in the music domain. + +--- + +## [SAIF: A Comprehensive Framework for Evaluating the Risks of Generative + AI in the Public Sector](https://arxiv.org/abs/2501.08814v1) +**arXiv ID:** 2501.08814v1 + +**Abstract:** +> The rapid adoption of generative AI in the public sector, encompassing +diverse applications ranging from automated public assistance to welfare +services and immigration processes, highlights its transformative potential +while underscoring the pressing need for thorough risk assessments. Despite its +growing presence, evaluations of risks associated with AI-driven systems in the +public sector remain insufficiently explored. Building upon an established +taxonomy of AI risks derived from diverse government policies and corporate +guidelines, we investigate the critical risks posed by generative AI in the +public sector while extending the scope to account for its multimodal +capabilities. In addition, we propose a Systematic dAta generatIon Framework +for evaluating the risks of generative AI (SAIF). SAIF involves four key +stages: breaking down risks, designing scenarios, applying jailbreak methods, +and exploring prompt types. It ensures the systematic and consistent generation +of prompt data, facilitating a comprehensive evaluation while providing a solid +foundation for mitigating the risks. Furthermore, SAIF is designed to +accommodate emerging jailbreak methods and evolving prompt types, thereby +enabling effective responses to unforeseen risk scenarios. We believe that this +study can play a crucial role in fostering the safe and responsible integration +of generative AI into the public sector. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on responsible AI application and AI ethics (risk assessment and mitigation), which is explicitly mentioned as a rejection criterion. Additionally, while it mentions practical applications in the public sector, the core contribution (SAIF framework) is geared towards evaluating and mitigating risks, rather than demonstrating practical applications or experimental results with Large Language Models (LLMs). + +--- + +## [Automatic tuning of communication protocols for vehicular ad hoc + networks using metaheuristics](https://arxiv.org/abs/2501.08847v1) +**arXiv ID:** 2501.08847v1 + +**Abstract:** +> The emerging field of vehicular ad hoc networks (VANETs) deals with a set of +communicating vehicles which are able to spontaneously interconnect without any +pre-existing infrastructure. In such kind of networks, it is crucial to make an +optimal configuration of the communication protocols previously to the final +network deployment. This way, a human designer can obtain an optimal QoS of the +network beforehand. The problem we consider in this work lies in configuring +the File Transfer protocol Configuration (FTC) with the aim of optimizing the +transmission time, the number of lost packets, and the amount of data +transferred in realistic VANET scenarios. We face the FTC with five +representative state-of-the-art optimization techniques and compare their +performance. These algorithms are: Particle Swarm Optimization (PSO), +Differential Evolution (DE), Genetic Algorithm (GA), Evolutionary Strategy +(ES), and Simulated Annealing (SA). For our tests, two typical environment +instances of VANETs for Urban and Highway scenarios have been defined. The +experiments using ns- 2 (a well-known realistic VANET simulator) reveal that +PSO outperforms all the compared algorithms for both studied VANET instances. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper does not meet any of the required criteria, focusing on vehicular ad hoc networks and metaheuristics for optimizing communication protocols, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI." +} + +--- + +## [RouteNet-Gauss: Hardware-Enhanced Network Modeling with Machine Learning](https://arxiv.org/abs/2501.08848v1) +**arXiv ID:** 2501.08848v1 + +**Abstract:** +> Network simulation is pivotal in network modeling, assisting with tasks +ranging from capacity planning to performance estimation. Traditional +approaches such as Discrete Event Simulation (DES) face limitations in terms of +computational cost and accuracy. This paper introduces RouteNet-Gauss, a novel +integration of a testbed network with a Machine Learning (ML) model to address +these challenges. By using the testbed as a hardware accelerator, +RouteNet-Gauss generates training datasets rapidly and simulates network +scenarios with high fidelity to real-world conditions. Experimental results +show that RouteNet-Gauss significantly reduces prediction errors by up to 95% +and achieves a 488x speedup in inference time compared to state-of-the-art +DES-based methods. RouteNet-Gauss's modular architecture is dynamically +constructed based on the specific characteristics of the network scenario, such +as topology and routing. This enables it to understand and generalize to +different network configurations beyond those seen during training, including +networks up to 10x larger. Additionally, it supports Temporal Aggregated +Performance Estimation (TAPE), providing configurable temporal granularity and +maintaining high accuracy in flow performance metrics. This approach shows +promise in improving both simulation efficiency and accuracy, offering a +valuable tool for network operators. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the primary criteria related to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. It focuses on network modeling using Machine Learning (ML) with a hardware accelerator, which falls outside the specified scope. + +--- + +## [Graph Counterfactual Explainable AI via Latent Space Traversal](https://arxiv.org/abs/2501.08850v1) +**arXiv ID:** 2501.08850v1 + +**Abstract:** +> Explaining the predictions of a deep neural network is a nontrivial task, yet +high-quality explanations for predictions are often a prerequisite for +practitioners to trust these models. Counterfactual explanations aim to explain +predictions by finding the ''nearest'' in-distribution alternative input whose +prediction changes in a pre-specified way. However, it remains an open question +how to define this nearest alternative input, whose solution depends on both +the domain (e.g. images, graphs, tabular data, etc.) and the specific +application considered. For graphs, this problem is complicated i) by their +discrete nature, as opposed to the continuous nature of state-of-the-art graph +classifiers; and ii) by the node permutation group acting on the graphs. We +propose a method to generate counterfactual explanations for any differentiable +black-box graph classifier, utilizing a case-specific permutation equivariant +graph variational autoencoder. We generate counterfactual explanations in a +continuous fashion by traversing the latent space of the autoencoder across the +classification boundary of the classifier, allowing for seamless integration of +discrete graph structure and continuous graph attributes. We empirically +validate the approach on three graph datasets, showing that our model is +consistently high-performing and more robust than the baselines. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on explainable AI for graph neural networks, without clear evidence of integrating Large Language Models (LLMs) or addressing the specified application areas (knowledge graphs, retrieval-augmented generation, or agentic AI), thus not meeting the required criteria. + +--- + +## [ARMOR: Shielding Unlearnable Examples against Data Augmentation](https://arxiv.org/abs/2501.08862v1) +**arXiv ID:** 2501.08862v1 + +**Abstract:** +> Private data, when published online, may be collected by unauthorized parties +to train deep neural networks (DNNs). To protect privacy, defensive noises can +be added to original samples to degrade their learnability by DNNs. Recently, +unlearnable examples are proposed to minimize the training loss such that the +model learns almost nothing. However, raw data are often pre-processed before +being used for training, which may restore the private information of protected +data. In this paper, we reveal the data privacy violation induced by data +augmentation, a commonly used data pre-processing technique to improve model +generalization capability, which is the first of its kind as far as we are +concerned. We demonstrate that data augmentation can significantly raise the +accuracy of the model trained on unlearnable examples from 21.3% to 66.1%. To +address this issue, we propose a defense framework, dubbed ARMOR, to protect +data privacy from potential breaches of data augmentation. To overcome the +difficulty of having no access to the model training process, we design a +non-local module-assisted surrogate model that better captures the effect of +data augmentation. In addition, we design a surrogate augmentation selection +strategy that maximizes distribution alignment between augmented and +non-augmented samples, to choose the optimal augmentation strategy for each +class. We also use a dynamic step size adjustment algorithm to enhance the +defensive noise generation process. Extensive experiments are conducted on 4 +datasets and 5 data augmentation methods to verify the performance of ARMOR. +Comparisons with 6 state-of-the-art defense methods have demonstrated that +ARMOR can preserve the unlearnability of protected private data under data +augmentation. ARMOR reduces the test accuracy of the model trained on augmented +protected samples by as much as 60% more than baselines. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on protecting data privacy from breaches, which aligns with 'responsible AI application or AI ethics', a criterion for rejection. Additionally, it does not clearly meet the required criteria for Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, as the main topic revolves around deep neural networks (DNNs) and data augmentation techniques for privacy preservation. + +--- + +## [Karatsuba Matrix Multiplication and its Efficient Custom Hardware + Implementations](https://arxiv.org/abs/2501.08889v1) +**arXiv ID:** 2501.08889v1 + +**Abstract:** +> While the Karatsuba algorithm reduces the complexity of large integer +multiplication, the extra additions required minimize its benefits for smaller +integers of more commonly-used bitwidths. In this work, we propose the +extension of the scalar Karatsuba multiplication algorithm to matrix +multiplication, showing how this maintains the reduction in multiplication +complexity of the original Karatsuba algorithm while reducing the complexity of +the extra additions. Furthermore, we propose new matrix multiplication hardware +architectures for efficiently exploiting this extension of the Karatsuba +algorithm in custom hardware. We show that the proposed algorithm and hardware +architectures can provide real area or execution time improvements for integer +matrix multiplication compared to scalar Karatsuba or conventional matrix +multiplication algorithms, while also supporting implementation through proven +systolic array and conventional multiplier architectures at the core. We +provide a complexity analysis of the algorithm and architectures and evaluate +the proposed designs both in isolation and in an end-to-end deep learning +accelerator system compared to baseline designs and prior state-of-the-art +works implemented on the same type of compute platform, demonstrating their +ability to increase the performance-per-area of matrix multiplication hardware. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria as it focuses on matrix multiplication algorithms and custom hardware implementations for deep learning accelerators, without mentioning Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, and falls outside the specified areas of interest. + +--- + +## [Computing Game Symmetries and Equilibria That Respect Them](https://arxiv.org/abs/2501.08905v1) +**arXiv ID:** 2501.08905v1 + +**Abstract:** +> Strategic interactions can be represented more concisely, and analyzed and +solved more efficiently, if we are aware of the symmetries within the +multiagent system. Symmetries also have conceptual implications, for example +for equilibrium selection. We study the computational complexity of identifying +and using symmetries. Using the classical framework of normal-form games, we +consider game symmetries that can be across some or all players and/or actions. +We find a strong connection between game symmetries and graph automorphisms, +yielding graph automorphism and graph isomorphism completeness results for +characterizing the symmetries present in a game. On the other hand, we also +show that the problem becomes polynomial-time solvable when we restrict the +consideration of actions in one of two ways. + Next, we investigate when exactly game symmetries can be successfully +leveraged for Nash equilibrium computation. We show that finding a Nash +equilibrium that respects a given set of symmetries is PPAD- and CLS-complete +in general-sum and team games respectively -- that is, exactly as hard as +Brouwer fixed point and gradient descent problems. Finally, we present +polynomial-time methods for the special cases where we are aware of a vast +number of symmetries, or where the game is two-player zero-sum and we do not +even know the symmetries. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the mandatory criteria as it lacks focus on Large Language Models (LLMs), practical applications in areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. Instead, it focuses on computational complexity and game theory, which falls outside the specified scope. + +--- + +## [Trusted Machine Learning Models Unlock Private Inference for Problems + Currently Infeasible with Cryptography](https://arxiv.org/abs/2501.08970v1) +**arXiv ID:** 2501.08970v1 + +**Abstract:** +> We often interact with untrusted parties. Prioritization of privacy can limit +the effectiveness of these interactions, as achieving certain goals +necessitates sharing private data. Traditionally, addressing this challenge has +involved either seeking trusted intermediaries or constructing cryptographic +protocols that restrict how much data is revealed, such as multi-party +computations or zero-knowledge proofs. While significant advances have been +made in scaling cryptographic approaches, they remain limited in terms of the +size and complexity of applications they can be used for. In this paper, we +argue that capable machine learning models can fulfill the role of a trusted +third party, thus enabling secure computations for applications that were +previously infeasible. In particular, we describe Trusted Capable Model +Environments (TCMEs) as an alternative approach for scaling secure computation, +where capable machine learning model(s) interact under input/output +constraints, with explicit information flow control and explicit statelessness. +This approach aims to achieve a balance between privacy and computational +efficiency, enabling private inference where classical cryptographic solutions +are currently infeasible. We describe a number of use cases that are enabled by +TCME, and show that even some simple classic cryptographic problems can already +be solved with TCME. Finally, we outline current limitations and discuss the +path forward in implementing them. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on privacy and security in machine learning, using cryptography as a point of comparison, without clear connections to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation, or agentic AI, failing to meet the mandatory criteria." +} +``` + +--- + +## [Personality Modeling for Persuasion of Misinformation using AI Agent](https://arxiv.org/abs/2501.08985v1) +**arXiv ID:** 2501.08985v1 + +**Abstract:** +> The proliferation of misinformation on social media platforms has highlighted +the need to understand how individual personality traits influence +susceptibility to and propagation of misinformation. This study employs an +innovative agent-based modeling approach to investigate the relationship +between personality traits and misinformation dynamics. Using six AI agents +embodying different dimensions of the Big Five personality traits +(Extraversion, Agreeableness, and Neuroticism), we simulated interactions +across six diverse misinformation topics. The experiment, implemented through +the AgentScope framework using the GLM-4-Flash model, generated 90 unique +interactions, revealing complex patterns in how personality combinations affect +persuasion and resistance to misinformation. Our findings demonstrate that +analytical and critical personality traits enhance effectiveness in +evidence-based discussions, while non-aggressive persuasion strategies show +unexpected success in misinformation correction. Notably, agents with critical +traits achieved a 59.4% success rate in HIV-related misinformation discussions, +while those employing non-aggressive approaches maintained consistent +persuasion rates above 40% across different personality combinations. The study +also revealed a non-transitive pattern in persuasion effectiveness, challenging +conventional assumptions about personality-based influence. These results +provide crucial insights for developing personality-aware interventions in +digital environments and suggest that effective misinformation countermeasures +should prioritize emotional connection and trust-building over confrontational +approaches. The findings contribute to both theoretical understanding of +personality-misinformation dynamics and practical strategies for combating +misinformation in social media contexts. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on social applications of AI regarding misinformation and social influence, aligning with rejected topics (Social harm or similar issues), and lacks clear connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI in a non-sandboxed, complex, real-world task context. + +--- + +## [AI-RAN: Transforming RAN with AI-driven Computing Infrastructure](https://arxiv.org/abs/2501.09007v1) +**arXiv ID:** 2501.09007v1 + +**Abstract:** +> The radio access network (RAN) landscape is undergoing a transformative shift +from traditional, communication-centric infrastructures towards converged +compute-communication platforms. This article introduces AI-RAN which +integrates both RAN and artificial intelligence (AI) workloads on the same +infrastructure. By doing so, AI-RAN not only meets the performance demands of +future networks but also improves asset utilization. We begin by examining how +RANs have evolved beyond mobile broadband towards AI-RAN and articulating +manifestations of AI-RAN into three forms: AI-for-RAN, AI-on-RAN, and +AI-and-RAN. Next, we identify the key requirements and enablers for the +convergence of communication and computing in AI-RAN. We then provide a +reference architecture for advancing AI-RAN from concept to practice. To +illustrate the practical potential of AI-RAN, we present a proof-of-concept +that concurrently processes RAN and AI workloads utilizing NVIDIA Grace-Hopper +GH200 servers. Finally, we conclude the article by outlining future work +directions to guide further developments of AI-RAN. + +**Decision Explanation:** Original decision: REJECT +The paper does not mention Large Language Models (LLMs) or related areas like knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, focusing instead on integrating AI with radio access network (RAN) infrastructure, which is outside the specified scope. + +--- + +## [Multimodal LLMs Can Reason about Aesthetics in Zero-Shot](https://arxiv.org/abs/2501.09012v1) +**arXiv ID:** 2501.09012v1 + +**Abstract:** +> We present the first study on how Multimodal LLMs' (MLLMs) reasoning ability +shall be elicited to evaluate the aesthetics of artworks. To facilitate this +investigation, we construct MM-StyleBench, a novel high-quality dataset for +benchmarking artistic stylization. We then develop a principled method for +human preference modeling and perform a systematic correlation analysis between +MLLMs' responses and human preference. Our experiments reveal an inherent +hallucination issue of MLLMs in art evaluation, associated with response +subjectivity. ArtCoT is proposed, demonstrating that art-specific task +decomposition and the use of concrete language boost MLLMs' reasoning ability +for aesthetics. Our findings offer valuable insights into MLLMs for art and can +benefit a wide range of downstream applications, such as style transfer and +artistic image generation. Code available at +https://github.com/songrise/MLLM4Art. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on the application of Multimodal LLMs in evaluating aesthetics of artworks, which falls under a creative/social application rather than the specified areas of interest (knowledge graphs, retrieval-augmented generation, agentic AI, etc.). It does not clearly align with the preferred criteria, especially given its lack of direct relevance to practical applications in areas like knowledge graphs, RAG, or agentic AI. + +--- + +## [Spatio-Temporal Foundation Models: Vision, Challenges, and Opportunities](https://arxiv.org/abs/2501.09045v1) +**arXiv ID:** 2501.09045v1 + +**Abstract:** +> Foundation models have revolutionized artificial intelligence, setting new +benchmarks in performance and enabling transformative capabilities across a +wide range of vision and language tasks. However, despite the prevalence of +spatio-temporal data in critical domains such as transportation, public health, +and environmental monitoring, spatio-temporal foundation models (STFMs) have +not yet achieved comparable success. In this paper, we articulate a vision for +the future of STFMs, outlining their essential characteristics and the +generalization capabilities necessary for broad applicability. We critically +assess the current state of research, identifying gaps relative to these ideal +traits, and highlight key challenges that impede their progress. Finally, we +explore potential opportunities and directions to advance research towards the +aim of effective and broadly applicable STFMs. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on foundation models for vision tasks (e.g., transportation, environmental monitoring) with minimal indication of direct involvement with Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the mandatory criteria (1, 2, and 3) and not strongly aligning with any of the secondary criteria related to LLMs. + +--- + +## [Dynamic-Aware Spatio-temporal Representation Learning for Dynamic MRI + Reconstruction](https://arxiv.org/abs/2501.09049v1) +**arXiv ID:** 2501.09049v1 + +**Abstract:** +> Dynamic MRI reconstruction, one of inverse problems, has seen a surge by the +use of deep learning techniques. Especially, the practical difficulty of +obtaining ground truth data has led to the emergence of unsupervised learning +approaches. A recent promising method among them is implicit neural +representation (INR), which defines the data as a continuous function that maps +coordinate values to the corresponding signal values. This allows for filling +in missing information only with incomplete measurements and solving the +inverse problem effectively. Nevertheless, previous works incorporating this +method have faced drawbacks such as long optimization time and the need for +extensive hyperparameter tuning. To address these issues, we propose +Dynamic-Aware INR (DA-INR), an INR-based model for dynamic MRI reconstruction +that captures the spatial and temporal continuity of dynamic MRI data in the +image domain and explicitly incorporates the temporal redundancy of the data +into the model structure. As a result, DA-INR outperforms other models in +reconstruction quality even at extreme undersampling ratios while significantly +reducing optimization time and requiring minimal hyperparameter tuning. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI (Dynamic MRI Reconstruction), which is a rejection criterion, and does not mention Large Language Models (LLMs) or related areas like knowledge graphs, RAG, or agentic AI, failing to meet the mandatory criteria. + +--- + +## [Tracking the Takes and Trajectories of English-Language News Narratives + across Trustworthy and Worrisome Websites](https://arxiv.org/abs/2501.09102v1) +**arXiv ID:** 2501.09102v1 + +**Abstract:** +> Understanding how misleading and outright false information enters news +ecosystems remains a difficult challenge that requires tracking how narratives +spread across thousands of fringe and mainstream news websites. To do this, we +introduce a system that utilizes encoder-based large language models and +zero-shot stance detection to scalably identify and track news narratives and +their attitudes across over 4,000 factually unreliable, mixed-reliability, and +factually reliable English-language news websites. Running our system over an +18 month period, we track the spread of 146K news stories. Using network-based +interference via the NETINF algorithm, we show that the paths of news +narratives and the stances of websites toward particular entities can be used +to uncover slanted propaganda networks (e.g., anti-vaccine and anti-Ukraine) +and to identify the most influential websites in spreading these attitudes in +the broader news ecosystem. We hope that increased visibility into our +distributed news ecosystem can help with the reporting and fact-checking of +propaganda and disinformation. + +**Decision Explanation:** Original response: ``` +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on social applications of AI to combat disinformation and propaganda, aligning with 'social applications of AI in regard to...Social harm, or similar issues', a criterion for rejection." +} +``` + +--- + +## [A Non-autoregressive Model for Joint STT and TTS](https://arxiv.org/abs/2501.09104v1) +**arXiv ID:** 2501.09104v1 + +**Abstract:** +> In this paper, we take a step towards jointly modeling automatic speech +recognition (STT) and speech synthesis (TTS) in a fully non-autoregressive way. +We develop a novel multimodal framework capable of handling the speech and text +modalities as input either individually or together. The proposed model can +also be trained with unpaired speech or text data owing to its multimodal +nature. We further propose an iterative refinement strategy to improve the STT +and TTS performance of our model such that the partial hypothesis at the output +can be fed back to the input of our model, thus iteratively improving both STT +and TTS predictions. We show that our joint model can effectively perform both +STT and TTS tasks, outperforming the STT-specific baseline in all tasks and +performing competitively with the TTS-specific baseline across a wide range of +evaluation metrics. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on Speech-to-Text (STT) and Text-to-Speech (TTS) applications, which, while related to language models, do not explicitly involve Large Language Models (LLMs) in areas like knowledge graphs, retrieval-augmented generation, or agentic AI as required by the primary criteria. Additionally, there is no clear indication of addressing the secondary criteria that would justify inclusion under the prioritization for inclusiveness. + +--- + +## [AutoLoop: Fast Visual SLAM Fine-tuning through Agentic Curriculum + Learning](https://arxiv.org/abs/2501.09160v1) +**arXiv ID:** 2501.09160v1 + +**Abstract:** +> Current visual SLAM systems face significant challenges in balancing +computational efficiency with robust loop closure handling. Traditional +approaches require careful manual tuning and incur substantial computational +overhead, while learning-based methods either lack explicit loop closure +capabilities or implement them through computationally expensive methods. We +present AutoLoop, a novel approach that combines automated curriculum learning +with efficient fine-tuning for visual SLAM systems. Our method employs a DDPG +(Deep Deterministic Policy Gradient) agent to dynamically adjust loop closure +weights during training, eliminating the need for manual hyperparameter search +while significantly reducing the required training steps. The approach +pre-computes potential loop closure pairs offline and leverages them through an +agent-guided curriculum, allowing the model to adapt efficiently to new +scenarios. Experiments conducted on TartanAir for training and validated across +multiple benchmarks including KITTI, EuRoC, ICL-NUIM and TUM RGB-D demonstrate +that AutoLoop achieves comparable or superior performance while reducing +training time by an order of magnitude compared to traditional approaches. +AutoLoop provides a practical solution for rapid adaptation of visual SLAM +systems, automating the weight tuning process that traditionally requires +multiple manual iterations. Our results show that this automated curriculum +strategy not only accelerates training but also maintains or improves the +model's performance across diverse environmental conditions. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on visual SLAM (Simultaneous Localization and Mapping), a computer vision and robotics topic, with no apparent direct connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI applications as specified in the criteria. The mention of 'Agentic' in the title refers to the use of an agent in a robotics context, not the specified AI application area. + +--- + +## [Towards Understanding Extrapolation: a Causal Lens](https://arxiv.org/abs/2501.09163v1) +**arXiv ID:** 2501.09163v1 + +**Abstract:** +> Canonical work handling distribution shifts typically necessitates an entire +target distribution that lands inside the training distribution. However, +practical scenarios often involve only a handful of target samples, potentially +lying outside the training support, which requires the capability of +extrapolation. In this work, we aim to provide a theoretical understanding of +when extrapolation is possible and offer principled methods to achieve it +without requiring an on-support target distribution. To this end, we formulate +the extrapolation problem with a latent-variable model that embodies the +minimal change principle in causal mechanisms. Under this formulation, we cast +the extrapolation problem into a latent-variable identification problem. We +provide realistic conditions on shift properties and the estimation objectives +that lead to identification even when only one off-support target sample is +available, tackling the most challenging scenarios. Our theory reveals the +intricate interplay between the underlying manifold's smoothness and the shift +properties. We showcase how our theoretical results inform the design of +practical adaptation algorithms. Through experiments on both synthetic and +real-world data, we validate our theoretical findings and their practical +implications. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it lacks focus on Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, instead concentrating on theoretical understanding of extrapolation with a causal lens, making it outside the specified scope. + +--- + +## [A Blockchain-Enabled Approach to Cross-Border Compliance and Trust](https://arxiv.org/abs/2501.09182v1) +**arXiv ID:** 2501.09182v1 + +**Abstract:** +> As artificial intelligence (AI) systems become increasingly integral to +critical infrastructure and global operations, the need for a unified, +trustworthy governance framework is more urgent that ever. This paper proposes +a novel approach to AI governance, utilizing blockchain and distributed ledger +technologies (DLT) to establish a decentralized, globally recognized framework +that ensures security, privacy, and trustworthiness of AI systems across +borders. The paper presents specific implementation scenarios within the +financial sector, outlines a phased deployment timeline over the next decade, +and addresses potential challenges with solutions grounded in current research. +By synthesizing advancements in blockchain, AI ethics, and cybersecurity, this +paper offers a comprehensive roadmap for a decentralized AI governance +framework capable of adapting to the complex and evolving landscape of global +AI regulation. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on AI governance, ethics, and regulation, with an emphasis on blockchain-enabled compliance and trust, which aligns with rejected criteria (responsible AI application or AI ethics). It lacks clear connections to Large Language Models (LLMs) and their practical applications as specified in the selection criteria. + +--- + +## [Patch-aware Vector Quantized Codebook Learning for Unsupervised Visual + Defect Detection](https://arxiv.org/abs/2501.09187v1) +**arXiv ID:** 2501.09187v1 + +**Abstract:** +> Unsupervised visual defect detection is critical in industrial applications, +requiring a representation space that captures normal data features while +detecting deviations. Achieving a balance between expressiveness and +compactness is challenging; an overly expressive space risks inefficiency and +mode collapse, impairing detection accuracy. We propose a novel approach using +an enhanced VQ-VAE framework optimized for unsupervised defect detection. Our +model introduces a patch-aware dynamic code assignment scheme, enabling +context-sensitive code allocation to optimize spatial representation. This +strategy enhances normal-defect distinction and improves detection accuracy +during inference. Experiments on MVTecAD, BTAD, and MTSD datasets show our +method achieves state-of-the-art performance. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on unsupervised visual defect detection, which falls under 'video processing' ( analyzing images for defects), meeting none of the required criteria for Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + +## [Adaptive Law-Based Transformation (ALT): A Lightweight Feature + Representation for Time Series Classification](https://arxiv.org/abs/2501.09217v1) +**arXiv ID:** 2501.09217v1 + +**Abstract:** +> Time series classification (TSC) is fundamental in numerous domains, +including finance, healthcare, and environmental monitoring. However, +traditional TSC methods often struggle with the inherent complexity and +variability of time series data. Building on our previous work with the linear +law-based transformation (LLT) - which improved classification accuracy by +transforming the feature space based on key data patterns - we introduce +adaptive law-based transformation (ALT). ALT enhances LLT by incorporating +variable-length shifted time windows, enabling it to capture distinguishing +patterns of various lengths and thereby handle complex time series more +effectively. By mapping features into a linearly separable space, ALT provides +a fast, robust, and transparent solution that achieves state-of-the-art +performance with only a few hyperparameters. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on time series classification in domains like finance, healthcare, and environmental monitoring, without any evident connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the fundamental criteria. + +--- + +## [Foundations of Large Language Models](https://arxiv.org/abs/2501.09223v1) +**arXiv ID:** 2501.09223v1 + +**Abstract:** +> This is a book about large language models. As indicated by the title, it +primarily focuses on foundational concepts rather than comprehensive coverage +of all cutting-edge technologies. The book is structured into four main +chapters, each exploring a key area: pre-training, generative models, prompting +techniques, and alignment methods. It is intended for college students, +professionals, and practitioners in natural language processing and related +fields, and can serve as a reference for anyone interested in large language +models. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it primarily focuses on foundational concepts rather than practical applications, experimental results, or comparisons with state-of-the-art techniques in areas like knowledge graphs, RAG, or agentic AI, and lacks mentioned real-world applications, experimental results, or novelty in LLM applications. + +--- + +## [Large Language Model is Secretly a Protein Sequence Optimizer](https://arxiv.org/abs/2501.09274v1) +**arXiv ID:** 2501.09274v1 + +**Abstract:** +> We consider the protein sequence engineering problem, which aims to find +protein sequences with high fitness levels, starting from a given wild-type +sequence. Directed evolution has been a dominating paradigm in this field which +has an iterative process to generate variants and select via experimental +feedback. We demonstrate large language models (LLMs), despite being trained on +massive texts, are secretly protein sequence optimizers. With a directed +evolutionary method, LLM can perform protein engineering through Pareto and +experiment-budget constrained optimization, demonstrating success on both +synthetic and experimental fitness landscapes. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, specifically protein sequence engineering, which falls under the rejection criteria. Despite meeting potential criteria for practical applications, experimental results, and novelty, its primary focus on a medical application takes precedence. + +--- + +## [LAVCap: LLM-based Audio-Visual Captioning using Optimal Transport](https://arxiv.org/abs/2501.09291v1) +**arXiv ID:** 2501.09291v1 + +**Abstract:** +> Automated audio captioning is a task that generates textual descriptions for +audio content, and recent studies have explored using visual information to +enhance captioning quality. However, current methods often fail to effectively +fuse audio and visual data, missing important semantic cues from each modality. +To address this, we introduce LAVCap, a large language model (LLM)-based +audio-visual captioning framework that effectively integrates visual +information with audio to improve audio captioning performance. LAVCap employs +an optimal transport-based alignment loss to bridge the modality gap between +audio and visual features, enabling more effective semantic extraction. +Additionally, we propose an optimal transport attention module that enhances +audio-visual fusion using an optimal transport assignment map. Combined with +the optimal training strategy, experimental results demonstrate that each +component of our framework is effective. LAVCap outperforms existing +state-of-the-art methods on the AudioCaps dataset, without relying on large +datasets or post-processing. Code is available at +https://github.com/NAVER-INTEL-Co-Lab/gaudi-lavcap. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on audio-visual captioning with LLMs, which falls under multimedia processing (similar to video processing), and does not align with the specified areas of interest (knowledge graphs, retrieval-augmented generation, agentic AI, etc.). + +--- + +## [Understanding Mental Health Content on Social Media and Its Effect + Towards Suicidal Ideation](https://arxiv.org/abs/2501.09309v1) +**arXiv ID:** 2501.09309v1 + +**Abstract:** +> This review underscores the critical need for effective strategies to +identify and support individuals with suicidal ideation, exploiting +technological innovations in ML and DL to further suicide prevention efforts. +The study details the application of these technologies in analyzing vast +amounts of unstructured social media data to detect linguistic patterns, +keywords, phrases, tones, and contextual cues associated with suicidal +thoughts. It explores various ML and DL models like SVMs, CNNs, LSTM, neural +networks, and their effectiveness in interpreting complex data patterns and +emotional nuances within text data. The review discusses the potential of these +technologies to serve as a life-saving tool by identifying at-risk individuals +through their digital traces. Furthermore, it evaluates the real-world +effectiveness, limitations, and ethical considerations of employing these +technologies for suicide prevention, stressing the importance of responsible +development and usage. The study aims to fill critical knowledge gaps by +analyzing recent studies, methodologies, tools, and techniques in this field. +It highlights the importance of synthesizing current literature to inform +practical tools and suicide prevention efforts, guiding innovation in reliable, +ethical systems for early intervention. This research synthesis evaluates the +intersection of technology and mental health, advocating for the ethical and +responsible application of ML, DL, and NLP to offer life-saving potential +worldwide while addressing challenges like generalizability, biases, privacy, +and the need for further research to ensure these technologies do not +exacerbate existing inequities and harms. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on social applications of AI in regard to social harm (suicidal ideation) and responsible AI application or AI ethics, which are explicitly listed as rejection criteria. + +--- + +## [Shape-Based Single Object Classification Using Ensemble Method + Classifiers](https://arxiv.org/abs/2501.09311v1) +**arXiv ID:** 2501.09311v1 + +**Abstract:** +> Nowadays, more and more images are available. Annotation and retrieval of the +images pose classification problems, where each class is defined as the group +of database images labelled with a common semantic label. Various systems have +been proposed for content-based retrieval, as well as for image classification +and indexing. In this paper, a hierarchical classification framework has been +proposed for bridging the semantic gap effectively and achieving multi-category +image classification. A well known pre-processing and post-processing method +was used and applied to three problems; image segmentation, object +identification and image classification. The method was applied to classify +single object images from Amazon and Google datasets. The classification was +tested for four different classifiers; BayesNetwork (BN), Random Forest (RF), +Bagging and Vote. The estimated classification accuracies ranged from 20% to +99% (using 10-fold cross validation). The Bagging classifier presents the best +performance, followed by the Random Forest classifier. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet any of the required criteria related to Large Language Models (LLMs), focusing instead on image classification, object identification, and traditional machine learning classifiers, with no mention of LLMs, knowledge graphs, RAG, or agentic AI. + +--- + +## [ELM-DeepONets: Backpropagation-Free Training of Deep Operator Networks + via Extreme Learning Machines](https://arxiv.org/abs/2501.09395v1) +**arXiv ID:** 2501.09395v1 + +**Abstract:** +> Deep Operator Networks (DeepONets) are among the most prominent frameworks +for operator learning, grounded in the universal approximation theorem for +operators. However, training DeepONets typically requires significant +computational resources. To address this limitation, we propose ELM-DeepONets, +an Extreme Learning Machine (ELM) framework for DeepONets that leverages the +backpropagation-free nature of ELM. By reformulating DeepONet training as a +least-squares problem for newly introduced parameters, the ELM-DeepONet +approach significantly reduces training complexity. Validation on benchmark +problems, including nonlinear ODEs and PDEs, demonstrates that the proposed +method not only achieves superior accuracy but also drastically reduces +computational costs. This work offers a scalable and efficient alternative for +operator learning in scientific computing. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it focuses on Deep Operator Networks for scientific computing (non-Large Language Model application), lacks comparison with state-of-the-art LLM techniques, and does not demonstrate practical applications in knowledge graphs, RAG, or agentic AI. + +--- + +## [Dynamic Neural Style Transfer for Artistic Image Generation using VGG19](https://arxiv.org/abs/2501.09420v1) +**arXiv ID:** 2501.09420v1 + +**Abstract:** +> Throughout history, humans have created remarkable works of art, but +artificial intelligence has only recently started to make strides in generating +visually compelling art. Breakthroughs in the past few years have focused on +using convolutional neural networks (CNNs) to separate and manipulate the +content and style of images, applying texture synthesis techniques. +Nevertheless, a number of current techniques continue to encounter obstacles, +including lengthy processing times, restricted choices of style images, and the +inability to modify the weight ratio of styles. We proposed a neural style +transfer system that can add various artistic styles to a desired image to +address these constraints allowing flexible adjustments to style weight ratios +and reducing processing time. The system uses the VGG19 model for feature +extraction, ensuring high-quality, flexible stylization without compromising +content integrity. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on image processing and generation using CNNs (VGG19), with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, failing to meet the fundamental criteria. + +--- + +## [A Survey on Responsible LLMs: Inherent Risk, Malicious Use, and + Mitigation Strategy](https://arxiv.org/abs/2501.09431v1) +**arXiv ID:** 2501.09431v1 + +**Abstract:** +> While large language models (LLMs) present significant potential for +supporting numerous real-world applications and delivering positive social +impacts, they still face significant challenges in terms of the inherent risk +of privacy leakage, hallucinated outputs, and value misalignment, and can be +maliciously used for generating toxic content and unethical purposes after been +jailbroken. Therefore, in this survey, we present a comprehensive review of +recent advancements aimed at mitigating these issues, organized across the four +phases of LLM development and usage: data collecting and pre-training, +fine-tuning and alignment, prompting and reasoning, and post-processing and +auditing. We elaborate on the recent advances for enhancing the performance of +LLMs in terms of privacy protection, hallucination reduction, value alignment, +toxicity elimination, and jailbreak defenses. In contrast to previous surveys +that focus on a single dimension of responsible LLMs, this survey presents a +unified framework that encompasses these diverse dimensions, providing a +comprehensive view of enhancing LLMs to better serve real-world applications. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on responsible AI application and AI ethics, specifically addressing inherent risks, malicious use, and mitigation strategies, which is explicitly listed as a rejection criterion. + +--- + +## [Solving the unsolvable: Translating case law in Hong Kong](https://arxiv.org/abs/2501.09444v1) +**arXiv ID:** 2501.09444v1 + +**Abstract:** +> This paper addresses the challenges translating case law under Hong Kong's +bilingual legal system. It highlights the initial success of translating all +written statutes into Chinese before the 1997 handover, a task mandated by the +Basic Law. The effort involved significant collaboration among legal, +linguistic, and translation experts, resulting in a comprehensive and +culturally appropriate bilingual legal system. However, translating case law +remains a significant challenge due to the sheer volume and continuous growth +of judicial decisions. The paper critiques the governments and judiciarys +sporadic and uncoordinated efforts to translate case law, contrasting it with +the thorough approach previously taken for statute translation. Although the +government acknowledges the importance of legal bilingualism, it lacks a +sustainable strategy for translating case law. The Judiciarys position that +translating all judgments is unnecessary, unrealistic, and not cost-effectiveis +analyzed and critiqued for its impact on legal transparency and public trust. A +proposed solution involves leveraging machine translation technology through a +human-machine interactive translation platform, which undergoes two major +transitions. Initially based on a neural model, the platform transitions to +using a large language model for improved translation accuracy. Furthermore, it +evolves from a single-agent system to a multi-agent system, incorporating +Translator, Annotator, and Proofreader agents. This multi-agent approach, +supported by a grant, aims to facilitate efficient, high-quality translation of +judicial judgments by integrating advanced artificial intelligence and +continuous feedback mechanisms, thus better meeting the needs of a bilingual +legal system. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on law, specifically the application of Large Language Models for translating case law in Hong Kong's legal system, which meets the rejection criteria (Primarily focuses on law, either with AI as subject or participant). + +--- + +## [RE-POSE: Synergizing Reinforcement Learning-Based Partitioning and + Offloading for Edge Object Detection](https://arxiv.org/abs/2501.09465v1) +**arXiv ID:** 2501.09465v1 + +**Abstract:** +> Object detection plays a crucial role in smart video analysis, with +applications ranging from autonomous driving and security to smart cities. +However, achieving real-time object detection on edge devices presents +significant challenges due to their limited computational resources and the +high demands of deep neural network (DNN)-based detection models, particularly +when processing high-resolution video. Conventional strategies, such as input +down-sampling and network up-scaling, often compromise detection accuracy for +faster performance or lead to higher inference latency. To address these +issues, this paper introduces RE-POSE, a Reinforcement Learning (RL)-Driven +Partitioning and Edge Offloading framework designed to optimize the +accuracy-latency trade-off in resource-constrained edge environments. Our +approach features an RL-Based Dynamic Clustering Algorithm (RL-DCA) that +partitions video frames into non-uniform blocks based on object distribution +and the computational characteristics of DNNs. Furthermore, a parallel edge +offloading scheme is implemented to distribute these blocks across multiple +edge servers for concurrent processing. Experimental evaluations show that +RE-POSE significantly enhances detection accuracy and reduces inference +latency, surpassing existing methods. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on video processing (object detection in high-resolution video) and edge computing optimization, which is explicitly listed as a rejection criterion. While it demonstrates practical applications, experimental results, and comparison with state-of-the-art, its core focus aligns with a rejected topic area. + +--- + +## [MonoSOWA: Scalable monocular 3D Object detector Without human + Annotations](https://arxiv.org/abs/2501.09481v1) +**arXiv ID:** 2501.09481v1 + +**Abstract:** +> Detecting the three-dimensional position and orientation of objects using a +single RGB camera is a foundational task in computer vision with many important +applications. Traditionally, 3D object detection methods are trained in a +fully-supervised setup, requiring vast amounts of human annotations, which are +laborious, costly, and do not scale well with the ever-increasing amounts of +data being captured. + In this paper, we present the first method to train 3D object detectors for +monocular RGB cameras without domain-specific human annotations, thus making +orders of magnitude more data available for training. Thanks to newly proposed +Canonical Object Space, the method can not only exploit data across a variety +of datasets and camera setups to train a single 3D detector, but unlike +previous work it also works out of the box in previously unseen camera setups. +All this is crucial for practical applications, where the data and cameras are +extremely heterogeneous. + The method is evaluated on two standard autonomous driving datasets, where it +outperforms previous works, which, unlike our method, still rely on 2D human +annotations. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on computer vision (3D object detection using a monocular RGB camera) and autonomous driving, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the primary criteria. + +--- + +## [MatrixNet: Learning over symmetry groups using learned group + representations](https://arxiv.org/abs/2501.09571v1) +**arXiv ID:** 2501.09571v1 + +**Abstract:** +> Group theory has been used in machine learning to provide a theoretically +grounded approach for incorporating known symmetry transformations in tasks +from robotics to protein modeling. In these applications, equivariant neural +networks use known symmetry groups with predefined representations to learn +over geometric input data. We propose MatrixNet, a neural network architecture +that learns matrix representations of group element inputs instead of using +predefined representations. MatrixNet achieves higher sample efficiency and +generalization over several standard baselines in prediction tasks over the +several finite groups and the Artin braid group. We also show that MatrixNet +respects group relations allowing generalization to group elements of greater +word length than in the training set. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the practical applications criteria for Large Language Models (LLMs) as it focuses on group theory, symmetry transformations, and equivariant neural networks without mentioning LLMs, knowledge graphs, retrieval-augmented generation, or agentic AI. + +--- + +## [Managed-Retention Memory: A New Class of Memory for the AI Era](https://arxiv.org/abs/2501.09605v1) +**arXiv ID:** 2501.09605v1 + +**Abstract:** +> AI clusters today are one of the major uses of High Bandwidth Memory (HBM). +However, HBM is suboptimal for AI workloads for several reasons. Analysis shows +HBM is overprovisioned on write performance, but underprovisioned on density +and read bandwidth, and also has significant energy per bit overheads. It is +also expensive, with lower yield than DRAM due to manufacturing complexity. We +propose a new memory class: Managed-Retention Memory (MRM), which is more +optimized to store key data structures for AI inference workloads. We believe +that MRM may finally provide a path to viability for technologies that were +originally proposed to support Storage Class Memory (SCM). These technologies +traditionally offered long-term persistence (10+ years) but provided poor IO +performance and/or endurance. MRM makes different trade-offs, and by +understanding the workload IO patterns, MRM foregoes long-term data retention +and write performance for better potential performance on the metrics important +for these workloads. + +**Decision Explanation:** Original decision: REJECT +The paper focuses on developing a new memory class (Managed-Retention Memory) for optimizing AI inference workloads, but does not explicitly mention Large Language Models (LLMs), their applications, or related criteria such as knowledge graphs, RAG, or agentic AI, failing to meet the primary criteria. + +--- + +## [Incorporating Quantum Advantage in Quantum Circuit Generation through + Genetic Programming](https://arxiv.org/abs/2501.09682v1) +**arXiv ID:** 2501.09682v1 + +**Abstract:** +> Designing efficient quantum circuits that leverage quantum advantage compared +to classical computing has become increasingly critical. Genetic algorithms +have shown potential in generating such circuits through artificial evolution. +However, integrating quantum advantage into the fitness function of these +algorithms remains unexplored. In this paper, we aim to enhance the efficiency +of quantum circuit design by proposing two novel approaches for incorporating +quantum advantage metrics into the fitness function of genetic algorithms.1 We +evaluate our approaches based on the Bernstein-Vazirani Problem and the +Unstructured Database Search Problem as test cases. The results demonstrate +that our approaches not only improve the convergence speed of the genetic +algorithm but also produce circuits comparable to expert-designed solutions. +Our findings suggest that automated quantum circuit design using genetic +algorithms that incorporate a measure of quantum advantage is a promising +approach to accelerating the development of quantum algorithms. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it focuses on Quantum Circuit Generation using Genetic Programming, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the primary criteria (1, 2, and 3) and all secondary criteria related to LLMs. + +--- + +## [Reward-Guided Controlled Generation for Inference-Time Alignment in + Diffusion Models: Tutorial and Review](https://arxiv.org/abs/2501.09685v1) +**arXiv ID:** 2501.09685v1 + +**Abstract:** +> This tutorial provides an in-depth guide on inference-time guidance and +alignment methods for optimizing downstream reward functions in diffusion +models. While diffusion models are renowned for their generative modeling +capabilities, practical applications in fields such as biology often require +sample generation that maximizes specific metrics (e.g., stability, affinity in +proteins, closeness to target structures). In these scenarios, diffusion models +can be adapted not only to generate realistic samples but also to explicitly +maximize desired measures at inference time without fine-tuning. This tutorial +explores the foundational aspects of such inference-time algorithms. We review +these methods from a unified perspective, demonstrating that current techniques +-- such as Sequential Monte Carlo (SMC)-based guidance, value-based sampling, +and classifier guidance -- aim to approximate soft optimal denoising processes +(a.k.a. policies in RL) that combine pre-trained denoising processes with value +functions serving as look-ahead functions that predict from intermediate states +to terminal rewards. Within this framework, we present several novel algorithms +not yet covered in the literature. Furthermore, we discuss (1) fine-tuning +methods combined with inference-time techniques, (2) inference-time algorithms +based on search algorithms such as Monte Carlo tree search, which have received +limited attention in current research, and (3) connections between +inference-time algorithms in language models and diffusion models. The code of +this tutorial on protein design is available at +https://github.com/masa-ue/AlignInversePro + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on diffusion models, biology (protein design), and generative modeling, with no clear connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus not meeting the mandatory criteria. + +--- + +## [Practical Continual Forgetting for Pre-trained Vision Models](https://arxiv.org/abs/2501.09705v1) +**arXiv ID:** 2501.09705v1 + +**Abstract:** +> For privacy and security concerns, the need to erase unwanted information +from pre-trained vision models is becoming evident nowadays. In real-world +scenarios, erasure requests originate at any time from both users and model +owners, and these requests usually form a sequence. Therefore, under such a +setting, selective information is expected to be continuously removed from a +pre-trained model while maintaining the rest. We define this problem as +continual forgetting and identify three key challenges. (i) For unwanted +knowledge, efficient and effective deleting is crucial. (ii) For remaining +knowledge, the impact brought by the forgetting procedure should be minimal. +(iii) In real-world scenarios, the training samples may be scarce or partially +missing during the process of forgetting. To address them, we first propose +Group Sparse LoRA (GS-LoRA). Specifically, towards (i), we introduce LoRA +modules to fine-tune the FFN layers in Transformer blocks for each forgetting +task independently, and towards (ii), a simple group sparse regularization is +adopted, enabling automatic selection of specific LoRA groups and zeroing out +the others. To further extend GS-LoRA to more practical scenarios, we +incorporate prototype information as additional supervision and introduce a +more practical approach, GS-LoRA++. For each forgotten class, we move the +logits away from its original prototype. For the remaining classes, we pull the +logits closer to their respective prototypes. We conduct extensive experiments +on face recognition, object detection and image classification and demonstrate +that our method manages to forget specific classes with minimal impact on other +classes. Codes have been released on https://github.com/bjzhb666/GS-LoRA. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on pre-trained vision models, continual forgetting, and applications in face recognition, object detection, and image classification, with no apparent connection to Large Language Models (LLMs), which is the core requirement for acceptance. + +--- + +## [Parallel multi-objective metaheuristics for smart communications in + vehicular networks](https://arxiv.org/abs/2501.09725v1) +**arXiv ID:** 2501.09725v1 + +**Abstract:** +> This article analyzes the use of two parallel multi-objective soft computing +algorithms to automatically search for high-quality settings of the Ad hoc On +Demand Vector routing protocol for vehicular networks. These methods are based +on an evolutionary algorithm and on a swarm intelligence approach. The +experimental analysis demonstrates that the configurations computed by our +optimization algorithms outperform other state-of-the-art optimized ones. In +turn, the computational efficiency achieved by all the parallel versions is +greater than 87 %. Therefore, the line of work presented in this article +represents an efficient framework to improve vehicular communications. + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the primary criteria as it does not focus on Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI; instead, it concentrates on optimizing vehicular network communications using evolutionary and swarm intelligence algorithms. + +--- + +## [Benchmarking Robustness of Contrastive Learning Models for Medical + Image-Report Retrieval](https://arxiv.org/abs/2501.09134v1) +**arXiv ID:** 2501.09134v1 + +**Abstract:** +> Medical images and reports offer invaluable insights into patient health. The +heterogeneity and complexity of these data hinder effective analysis. To bridge +this gap, we investigate contrastive learning models for cross-domain +retrieval, which associates medical images with their corresponding clinical +reports. This study benchmarks the robustness of four state-of-the-art +contrastive learning models: CLIP, CXR-RePaiR, MedCLIP, and CXR-CLIP. We +introduce an occlusion retrieval task to evaluate model performance under +varying levels of image corruption. Our findings reveal that all evaluated +models are highly sensitive to out-of-distribution data, as evidenced by the +proportional decrease in performance with increasing occlusion levels. While +MedCLIP exhibits slightly more robustness, its overall performance remains +significantly behind CXR-CLIP and CXR-RePaiR. CLIP, trained on a +general-purpose dataset, struggles with medical image-report retrieval, +highlighting the importance of domain-specific training data. The evaluation of +this work suggests that more effort needs to be spent on improving the +robustness of these models. By addressing these limitations, we can develop +more reliable cross-domain retrieval models for medical applications. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on medical applications of AI, which is explicitly listed as a criterion for rejection, outweighing any potential alignment with other criteria such as experimental results, comparison with state-of-the-art, and methodology details. + +--- + +## [Quantum-Enhanced Transformers for Robust Acoustic Scene Classification + in IoT Environments](https://arxiv.org/abs/2501.09394v1) +**arXiv ID:** 2501.09394v1 + +**Abstract:** +> The proliferation of Internet of Things (IoT) devices equipped with acoustic +sensors necessitates robust acoustic scene classification (ASC) capabilities, +even in noisy and data-limited environments. Traditional machine learning +methods often struggle to generalize effectively under such conditions. To +address this, we introduce Q-ASC, a novel Quantum-Inspired Acoustic Scene +Classifier that leverages the power of quantum-inspired transformers. By +integrating quantum concepts like superposition and entanglement, Q-ASC +achieves superior feature learning and enhanced noise resilience compared to +classical models. Furthermore, we introduce a Quantum Variational Autoencoder +(QVAE) based data augmentation technique to mitigate the challenge of limited +labeled data in IoT deployments. Extensive evaluations on the Tampere +University of Technology (TUT) Acoustic Scenes 2016 benchmark dataset +demonstrate that Q-ASC achieves remarkable accuracy between 68.3% and 88.5% +under challenging conditions, outperforming state-of-the-art methods by over 5% +in the best case. This research paves the way for deploying intelligent +acoustic sensing in IoT networks, with potential applications in smart homes, +industrial monitoring, and environmental surveillance, even in adverse acoustic +environments. + +**Decision Explanation:** Original decision: REJECT +The paper primarily focuses on acoustic scene classification in IoT environments using quantum-enhanced transformers, with no apparent connection to Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI, thus failing to meet the mandatory criteria (1, 2, and 3). + +--- + +## [Metric Learning with Progressive Self-Distillation for Audio-Visual + Embedding Learning](https://arxiv.org/abs/2501.09608v1) +**arXiv ID:** 2501.09608v1 + +**Abstract:** +> Metric learning projects samples into an embedded space, where similarities +and dissimilarities are quantified based on their learned representations. +However, existing methods often rely on label-guided representation learning, +where representations of different modalities, such as audio and visual data, +are aligned based on annotated labels. This approach tends to underutilize +latent complex features and potential relationships inherent in the +distributions of audio and visual data that are not directly tied to the +labels, resulting in suboptimal performance in audio-visual embedding learning. +To address this issue, we propose a novel architecture that integrates +cross-modal triplet loss with progressive self-distillation. Our method +enhances representation learning by leveraging inherent distributions and +dynamically refining soft audio-visual alignments -- probabilistic alignments +between audio and visual data that capture the inherent relationships beyond +explicit labels. Specifically, the model distills audio-visual +distribution-based knowledge from annotated labels in a subset of each batch. +This self-distilled knowledge is used t + +**Decision Explanation:** Original decision: REJECT +The paper does not meet the required criteria as it focuses on audio-visual embedding learning with metric learning and self-distillation, without any indication of practical applications or involvement of Large Language Models (LLMs), knowledge graphs, retrieval-augmented generation (RAG), or agentic AI. + +--- + diff --git a/llm_processor/processor/processor.go b/llm_processor/processor/processor.go new file mode 100644 index 0000000..6ffc8fe --- /dev/null +++ b/llm_processor/processor/processor.go @@ -0,0 +1,190 @@ +package processor + +import ( + "context" + "encoding/json" + "fmt" + "strings" + "time" + + "llm_processor/client" + "llm_processor/models" + "llm_processor/storage" +) + +type Processor struct { + modelName string + batchSize int + timeout time.Duration + client *client.OpenRouterClient +} + +// SetTimeout allows changing the processor timeout +func (p *Processor) SetTimeout(timeout time.Duration) { + p.timeout = timeout +} + +func NewProcessor(modelName string, batchSize int, apiKey string) *Processor { + return &Processor{ + modelName: modelName, + batchSize: batchSize, + timeout: 3600 * time.Second, // Default 1 hour timeout + client: client.NewOpenRouterClient(apiKey), + } +} + +func (p *Processor) ProcessPapers(parentCtx context.Context, inputPath, outputPath, criteriaPath string, delay time.Duration) error { + startTime := time.Now() + ctx, cancel := context.WithTimeout(parentCtx, p.timeout) + defer cancel() + + // Load papers from input file + papers, err := storage.LoadPapers(inputPath) + if err != nil { + return fmt.Errorf("failed to load papers: %w", err) + } + + // Set criteria path in storage + storage.SetCriteriaPath(criteriaPath) + + // Initialize results file + if err := storage.InitializeResultsFile(outputPath); err != nil { + return fmt.Errorf("failed to initialize results file: %w", err) + } + + // Process papers in batches + for i := 0; i < len(papers); i += p.batchSize { + end := i + p.batchSize + if end > len(papers) { + end = len(papers) + } + + batch := papers[i:end] + if err := p.processBatch(ctx, batch, delay, startTime, outputPath); err != nil { + return fmt.Errorf("failed to process batch: %w", err) + } + } + + return nil +} + +func (p *Processor) processBatch(ctx context.Context, papers []models.Paper, delay time.Duration, startTime time.Time, outputPath string) error { + ctx, cancel := context.WithTimeout(ctx, p.timeout) + defer cancel() + + // Load criteria + criteria, err := storage.LoadCriteria() + if err != nil { + return fmt.Errorf("failed to load criteria: %w", err) + } + for i, paper := range papers { + var evaluation string + var lastErr error + + // Retry up to 3 times with exponential backoff + for attempt := 0; attempt < 3; attempt++ { + evaluation, lastErr = p.client.EvaluatePaper(ctx, paper, criteria, p.modelName) + if lastErr == nil { + break + } + time.Sleep(time.Duration(attempt+1) * time.Second) // Exponential backoff + } + + if lastErr != nil { + // Log error but continue with next paper + evaluation = fmt.Sprintf(`{ + "decision": "ERROR", + "explanation": "Failed to evaluate paper after 3 attempts: %v" + }`, lastErr) + } + + // Parse and validate evaluation response + var evalResponse struct { + Decision string `json:"decision"` + Explanation string `json:"explanation"` + } + + if err := json.Unmarshal([]byte(evaluation), &evalResponse); err != nil { + // Try to extract decision and explanation using regex + decision := "ERROR" + explanation := evaluation + + if strings.Contains(evaluation, `"decision": "ACCEPT"`) || strings.Contains(evaluation, `"decision":"ACCEPT"`) { + decision = "ACCEPT" + } else if strings.Contains(evaluation, `"decision": "REJECT"`) || strings.Contains(evaluation, `"decision":"REJECT"`) { + decision = "REJECT" + } + + // Try to extract just the explanation if it's in JSON format + if strings.Contains(evaluation, `"explanation"`) { + parts := strings.Split(evaluation, `"explanation"`) + if len(parts) > 1 { + // Find the content between the first : and the next " + expl := parts[1] + start := strings.Index(expl, ":") + if start != -1 { + expl = expl[start+1:] + // Remove leading/trailing whitespace and quotes + expl = strings.Trim(expl, " \t\n\r\"") + // Remove trailing JSON syntax + expl = strings.TrimRight(expl, "}") + expl = strings.TrimRight(expl, ",") + explanation = expl + } + } + } + + evalResponse = struct { + Decision string `json:"decision"` + Explanation string `json:"explanation"` + }{ + Decision: decision, + Explanation: explanation, + } + } + + // Sanitize the explanation + explanation := evalResponse.Explanation + // Remove any markdown code block syntax + explanation = strings.ReplaceAll(explanation, "```", "") + // Remove any JSON formatting if the explanation is a raw JSON string + if strings.HasPrefix(strings.TrimSpace(explanation), "{") { + var jsonExpl struct { + Explanation string `json:"explanation"` + } + if err := json.Unmarshal([]byte(explanation), &jsonExpl); err == nil && jsonExpl.Explanation != "" { + explanation = jsonExpl.Explanation + } + } + // Escape any remaining special markdown characters + explanation = strings.ReplaceAll(explanation, "*", "\\*") + explanation = strings.ReplaceAll(explanation, "_", "\\_") + explanation = strings.ReplaceAll(explanation, "`", "\\`") + + result := models.Result{ + Paper: paper, + Decision: evalResponse.Decision, + Explanation: explanation, + } + + // Save result with detailed logging + fmt.Printf("Saving result for paper %q to %s\n", paper.Title, outputPath) + if err := storage.SaveResult(result, outputPath); err != nil { + fmt.Printf("Failed to save result for paper %q: %v\n", paper.Title, err) + return fmt.Errorf("failed to save result: %w", err) + } + fmt.Printf("Successfully saved result for paper %q\n", paper.Title) + + // Print progress + elapsed := time.Since(startTime).Seconds() + fmt.Printf("Processed paper %d/%d (%s) - Total runtime: %.2f seconds\n", + i+1, len(papers), paper.Title, elapsed) + + // Apply delay between papers if specified + if delay > 0 { + time.Sleep(delay) + } + } + + return nil +} diff --git a/llm_processor/storage/storage.go b/llm_processor/storage/storage.go new file mode 100644 index 0000000..ded1a58 --- /dev/null +++ b/llm_processor/storage/storage.go @@ -0,0 +1,143 @@ +package storage + +import ( + "encoding/json" + "fmt" + "os" + "path/filepath" + "sync" + + "llm_processor/models" +) + +var ( + criteriaPath = "criteria.txt" + criteriaMux sync.RWMutex +) + +// SetCriteriaPath sets the path for the criteria file +func SetCriteriaPath(path string) { + criteriaMux.Lock() + defer criteriaMux.Unlock() + criteriaPath = path +} + +// LoadPapers loads papers from a JSON file +func LoadPapers(filePath string) ([]models.Paper, error) { + absPath, err := filepath.Abs(filePath) + if err != nil { + return nil, fmt.Errorf("failed to resolve absolute path: %w", err) + } + + file, err := os.ReadFile(absPath) + if err != nil { + return nil, fmt.Errorf("failed to read papers file: %w", err) + } + + var papers []models.Paper + if err := json.Unmarshal(file, &papers); err != nil { + return nil, fmt.Errorf("failed to parse papers JSON: %w", err) + } + + return papers, nil +} + +// LoadCriteria loads evaluation criteria from a text file +func LoadCriteria() (string, error) { + criteriaMux.RLock() + filePath := criteriaPath + criteriaMux.RUnlock() + + content, err := os.ReadFile(filePath) + if err != nil { + return "", fmt.Errorf("failed to read criteria file: %w", err) + } + return string(content), nil +} + +// InitializeResultsFile creates a new results file with empty categories +func InitializeResultsFile(filePath string) error { + results := models.OrganizedResults{ + Accepted: []models.Result{}, + Rejected: []models.Result{}, + Errors: []models.Result{}, + } + + return saveToFile(filePath, results) +} + +// SaveResult appends a result to the appropriate category in the results file +func SaveResult(result models.Result, filePath string) error { + // Read existing results + results, err := loadResults(filePath) + if err != nil { + return fmt.Errorf("failed to load results: %w", err) + } + + // Append to appropriate category + switch result.Decision { + case "ACCEPT": + results.Accepted = append(results.Accepted, result) + case "REJECT": + results.Rejected = append(results.Rejected, result) + default: + results.Errors = append(results.Errors, result) + } + + // Save updated results + if err := saveToFile(filePath, results); err != nil { + // Try fallback save + if fallbackErr := saveFallback(result, filePath); fallbackErr != nil { + return fmt.Errorf("failed to save results: %w (fallback also failed: %v)", err, fallbackErr) + } + return nil + } + + return nil +} + +func loadResults(filePath string) (*models.OrganizedResults, error) { + file, err := os.ReadFile(filePath) + if err != nil { + return nil, fmt.Errorf("failed to read results file: %w", err) + } + + var results models.OrganizedResults + if err := json.Unmarshal(file, &results); err != nil { + return nil, fmt.Errorf("failed to parse results JSON: %w", err) + } + + return &results, nil +} + +func saveToFile(filePath string, data interface{}) error { + file, err := os.Create(filePath) + if err != nil { + return fmt.Errorf("failed to create file: %w", err) + } + defer file.Close() + + encoder := json.NewEncoder(file) + encoder.SetIndent("", " ") + if err := encoder.Encode(data); err != nil { + return fmt.Errorf("failed to encode JSON: %w", err) + } + + return nil +} + +func saveFallback(result models.Result, originalPath string) error { + fallbackPath := originalPath + ".fallback" + file, err := os.OpenFile(fallbackPath, os.O_APPEND|os.O_CREATE|os.O_WRONLY, 0644) + if err != nil { + return fmt.Errorf("failed to open fallback file: %w", err) + } + defer file.Close() + + encoder := json.NewEncoder(file) + if err := encoder.Encode(result); err != nil { + return fmt.Errorf("failed to encode result to fallback file: %w", err) + } + + return nil +} diff --git a/main.go b/main.go new file mode 100644 index 0000000..7d93820 --- /dev/null +++ b/main.go @@ -0,0 +1,238 @@ +package main + +import ( + "context" + "encoding/json" + "flag" + "fmt" + "log" + "os" + "time" + + "arxiv-processor/arxiv" + "json2md/lib" + "llm_processor/processor" +) + +func main() { + // Set custom usage before defining flags + flag.Usage = func() { + fmt.Fprintf(os.Stderr, "Usage: %s [options]\n\n", os.Args[0]) + fmt.Fprintf(os.Stderr, "A tool to fetch, filter, and process arXiv papers using LLM.\n\n") + fmt.Fprintf(os.Stderr, "Required flags:\n") + fmt.Fprintf(os.Stderr, " -criteria string\n\tPath to filter criteria file\n\n") + fmt.Fprintf(os.Stderr, "Source flags (must use either arXiv query OR input JSON):\n") + fmt.Fprintf(os.Stderr, " ArXiv query flags:\n") + fmt.Fprintf(os.Stderr, " -start string\n\tStart date in YYYYMMDD format\n") + fmt.Fprintf(os.Stderr, " -end string\n\tEnd date in YYYYMMDD format\n") + fmt.Fprintf(os.Stderr, " -search string\n\tarXiv category/search query (e.g., 'cs.AI', 'physics.comp-ph')\n") + fmt.Fprintf(os.Stderr, " -max-results int\n\tMaximum number of papers to retrieve (default: 100, max: 2000)\n\n") + fmt.Fprintf(os.Stderr, " OR\n\n") + fmt.Fprintf(os.Stderr, " Input JSON flag:\n") + fmt.Fprintf(os.Stderr, " -input-json string\n\tPath to input JSON file (bypasses arXiv fetch)\n\n") + fmt.Fprintf(os.Stderr, "Optional flags:\n") + fmt.Fprintf(os.Stderr, " -output string\n\tOutput markdown file path (default: papers.md)\n") + fmt.Fprintf(os.Stderr, " -model string\n\tLLM model to use (default: nvidia/llama-3.1-nemotron-70b-instruct)\n\n") + fmt.Fprintf(os.Stderr, "Environment variables:\n") + fmt.Fprintf(os.Stderr, " OPENROUTER_API_KEY\tRequired for LLM processing\n\n") + fmt.Fprintf(os.Stderr, "Examples:\n") + fmt.Fprintf(os.Stderr, " Fetch from arXiv:\n") + fmt.Fprintf(os.Stderr, " %s -start 20240101 -end 20240131 -search cs.AI -criteria criteria.txt -output papers.md\n\n", os.Args[0]) + fmt.Fprintf(os.Stderr, " Use existing JSON:\n") + fmt.Fprintf(os.Stderr, " %s -input-json papers.json -criteria new-criteria.txt -output results.md\n", os.Args[0]) + } + + // CLI flags + start := flag.String("start", "", "Start date (YYYYMMDD)") + end := flag.String("end", "", "End date (YYYYMMDD)") + search := flag.String("search", "", "arXiv search query") + criteriaFile := flag.String("criteria", "", "Path to filter criteria file") + output := flag.String("output", "papers.md", "Output file path") + model := flag.String("model", "nvidia/llama-3.1-nemotron-70b-instruct", "LLM model to use") + maxResults := flag.Int("max-results", 100, "Maximum number of papers to retrieve (up to 2000)") + inputJSON := flag.String("input-json", "", "Path to input JSON file (bypasses arXiv fetch)") + + flag.Parse() + + // Validate flags + if *criteriaFile == "" { + fmt.Fprintf(os.Stderr, "Error: Missing required parameter: -criteria\n\n") + flag.Usage() + os.Exit(1) + } + + // Validate either input-json is provided OR all arxiv flags are provided + usingInputJSON := *inputJSON != "" + usingArxiv := *start != "" || *end != "" || *search != "" + + if usingInputJSON && usingArxiv { + fmt.Fprintf(os.Stderr, "Error: Cannot use both --input-json and arXiv query flags\n\n") + flag.Usage() + os.Exit(1) + } + + if !usingInputJSON && !usingArxiv { + fmt.Fprintf(os.Stderr, "Error: Must provide either --input-json or arXiv query flags\n\n") + flag.Usage() + os.Exit(1) + } + + if usingArxiv { + if *start == "" || *end == "" || *search == "" { + fmt.Fprintf(os.Stderr, "Error: Missing required arXiv parameters\n\n") + flag.Usage() + os.Exit(1) + } + + if *maxResults <= 0 || *maxResults > 2000 { + fmt.Fprintf(os.Stderr, "Error: max-results must be between 1 and 2000\n\n") + flag.Usage() + os.Exit(1) + } + } + + // Configure logging + log.SetPrefix("[paper-system] ") + log.SetFlags(log.Ltime | log.Lmsgprefix) + + ctx := context.Background() + + // Paper type used for JSON operations + type LLMPaper struct { + Title string `json:"title"` + Abstract string `json:"abstract"` + ArxivID string `json:"arxiv_id"` + Authors []string `json:"authors"` + } + + var llmPapers []LLMPaper + + if usingInputJSON { + // Load papers from input JSON + log.Printf("Loading papers from %s", *inputJSON) + paperData, err := os.ReadFile(*inputJSON) + if err != nil { + log.Fatalf("Failed to read input JSON: %v", err) + } + + if err := json.Unmarshal(paperData, &llmPapers); err != nil { + log.Fatalf("Failed to parse input JSON: %v", err) + } + log.Printf("Loaded %d papers from JSON", len(llmPapers)) + + } else { + // Fetch papers from arXiv + log.Printf("Fetching papers from arXiv for category %q between %s and %s", *search, *start, *end) + arxivClient := arxiv.NewClient() + + startDate := parseDate(*start) + endDate := parseDate(*end) + + query := arxiv.Query{ + Category: *search, + DateRange: fmt.Sprintf("%s TO %s", startDate.Format("20060102"), endDate.Format("20060102")), + MaxResults: *maxResults, + StartOffset: 0, + } + + log.Printf("Executing arXiv query: %+v", query) + papers, err := arxivClient.FetchPapers(ctx, query) + if err != nil { + log.Fatalf("arXiv fetch failed: %v", err) + } + log.Printf("Retrieved %d papers from arXiv", len(papers)) + if len(papers) >= *maxResults { + log.Printf("WARNING: Retrieved maximum number of papers (%d). There may be more papers available.", *maxResults) + log.Printf("Use --max-results flag to retrieve more papers (up to 2000)") + } + + // Convert arXiv papers to LLM format + llmPapers = make([]LLMPaper, len(papers)) + for i, p := range papers { + // Convert author structs to string array + authors := make([]string, len(p.Authors)) + for j, a := range p.Authors { + authors[j] = a.Name + } + + llmPapers[i] = LLMPaper{ + Title: p.Title, + Abstract: p.Summary, + ArxivID: p.ID, + Authors: authors, + } + } + + // Save papers to JSON for future use + log.Printf("Saving papers to papers.json") + papersJSON, err := json.Marshal(llmPapers) + if err != nil { + log.Fatalf("Failed to marshal papers: %v", err) + } + if err := os.WriteFile("papers.json", papersJSON, 0644); err != nil { + log.Fatalf("Failed to save papers JSON: %v", err) + } + log.Printf("Successfully saved papers to papers.json") + } + + // Print paper titles for verification + log.Printf("Processing papers:") + for i, paper := range llmPapers { + log.Printf(" %d. %s", i+1, paper.Title) + } + + // Save papers to temporary file for LLM processing + tempInput := "temp_input.json" + tempJSON, err := json.Marshal(llmPapers) + if err != nil { + log.Fatalf("Failed to marshal papers for LLM: %v", err) + } + if err := os.WriteFile(tempInput, tempJSON, 0644); err != nil { + log.Fatalf("Failed to save temp input JSON: %v", err) + } + + // Filter papers with LLM + log.Printf("Starting LLM processing") + apiKey := os.Getenv("OPENROUTER_API_KEY") + if apiKey == "" { + log.Fatal("OPENROUTER_API_KEY environment variable is required") + } + + llmProcessor := processor.NewProcessor(*model, 32, apiKey) // 32 = batch size from README + log.Printf("Initialized LLM processor with model %s", *model) + + tempOutput := "temp_output.json" + log.Printf("Processing papers with criteria from %s", *criteriaFile) + if err := llmProcessor.ProcessPapers(ctx, tempInput, tempOutput, *criteriaFile, 1*time.Second); err != nil { + log.Fatalf("LLM processing failed: %v", err) + } + log.Printf("LLM processing complete, results saved to %s", tempOutput) + + // Generate markdown + log.Printf("Generating markdown output") + decisions, err := lib.ProcessJSONFile(tempOutput) + if err != nil { + log.Fatalf("Failed to process JSON: %v", err) + } + log.Printf("Processed decisions: %d accepted, %d rejected", len(decisions.Accepted), len(decisions.Rejected)) + + if err := lib.GenerateMarkdown(decisions, *output); err != nil { + log.Fatalf("Markdown generation failed: %v", err) + } + log.Printf("Generated markdown output at %s", *output) + + // Cleanup temp files + os.Remove(tempInput) + os.Remove(tempOutput) + log.Printf("Cleaned up temporary files") + + log.Printf("Process complete. Results saved to %s", *output) +} + +func parseDate(s string) time.Time { + t, err := time.Parse("20060102", s) + if err != nil { + log.Fatalf("Invalid date %q: %v", s, err) + } + return t +} diff --git a/paper-system b/paper-system new file mode 100755 index 0000000000000000000000000000000000000000..078a66c338b4477f04416f3a787e75444bc6eb17 GIT binary patch literal 8009954 zcmeF4dwf;ZmH79$_vQsCC~3ul=H>wjib4wouC&cb!c#2TnyI!r&6AfX2&jdE&4q-A z1Y0hqwT0RQkXL+^BB&W_2@e6Gwh)|7r=2M`Przsuw7=$7Fu(8G=bW39%cFK?KJ(Y) z^EtWa?8n+`uf6tq?S1z7?cd(}q?b~L;?Ki%4cEgxl=3M*cS>E)HIb{J;I5nrvu|O_di{^d(uTu0%tS@3kveb-J9P%Th|ZX(mm0BP9Qao%WLNr6g)C@!6RKoFM`*{ z>dYgrW>|%I_H`APhoJkhpkV&Ohl=MsSTO&QM~gpyF@C|r8JajlY#{P1_&9%wHoSN{ zp`c*ioYJC4ik$pd{OW&g!5h3ZQ5TVCH$11F^DA`wy{6mPocx@EygR4dZBY;U2;J6P zlWG;U>E%2Nz2p+ukN~>J;&o|w^(QR66!5Nu7vr$-;i(VKoIMNR8h&?p^>>f7n5w^( zFV9zoH}jFA;w!+LZNcl8WW(e6%J8PoVPt`6l`kmhj^FRUW8-IKD|x=sUqQk2IS-dS zINJdj3omTL+i`tA4Mv`?grAIS78A~b<7&Bnil&?YR-2=EmcRP3)(mjQwTs`T&nYN= zbmoJlGiOhmb9s1o$>f$l;B}i1R%-NrLBVY1f;A-`eq`Q7_}#V7!tbyRPo5pn0z~lo zW)zy>GkXqjqEjNazy3EYc%Ic36nVZ9-3kh#pmZy-@XW7$P1En;mn+Zt2wGN2eDQ6&KH&`q*XRO?EnD!?SAZ|6K%d!(+-$ z)z97Wn`~J4tsHG}>f{^?zrPD_+PtEN02tff6M_45e}~Sq5InK)UKc2R8-t(K-RY&p z1@lYh&5VWTdH#M4FVBYJdDqRgmx5>0&zfcz;Wy5PH}-dEQ~t(%T6ho6(VP&A-xEt` zYWy-PEqVOJzG)h+>t9TNZT#jH&7MAW{zdq`aMe5w@9Q^N)n522!lL_MH+aH#mfpW( z(v->L#!iarY-?PbUUl@uTn)u}meFxoNHV^_B@)p}>&cy5$Udcll3qC4;ZU8d(nudB zx52rp=1oJ@Oq#C>yBDzv+W!SzzB6autocLp z=FFQqe$Jz_5uh{Y%)V7mG|3tIjj8kIlQMs3PD~l#3x0+FSUY!5EqQd_%tsbpx&ZF}t8-_Ke~8e&c~JOn&T}X+8X%(XOoNX+}Cr zhP7|^s{BuS-`H`XJf-78x;pL|V4ip_OC2l89vhM0(lKKr1ZGCnhiKO{s14s@fkMGOAMi-hU0Jt3b2& z#@gY(GOEJqX5c?NJFheDHv|4Y=30RzNJ=?LOC>FPteRV-RKD_?5y=;LPC2JT-O5i? zbF;ux>eWy$o97~)RX?k~^P8Qo<`zo1d)3^ql)Im2exvYX%Bul(ech=0-on>=ml-Jc zs*zssE&=b7vFdo4U!C3NQTdIZQ#(?FBSK>qlynq&Q-0_58=-5JO7w z(!JzQO*HFDR9t1)>tzglwM||_)6N=`l<2d=4}-zg%@5HrxJ%95V2^=&-1l{MJ{<<8 zJ1v}`A~L?Oz0UY1n2|G#ZxnuEcbN0vYGrcWt`ot%)pj@=Gy)K{p(7GRj(_C4+nmdJ*vQ`&I;a3 z63syNVqNSTZ@&WmGENR%KW!hH>BAPg4+rVP zA^Onl>cb%VAbnnHbi7iNs6x$Nb+k-*+BeasMz5zEtxRb7yK^>eIa9FAmM>=EjU+Ggx+}|IUT*gq^5I&XhkZ~SMS}Wr${q-qdI($n1jC;WTXYaGw^^l6`TgU~EF%ERI5D$+DZMH1Ar4cRK8{r1m^tW0RX zsh`U4ygRO=^F-W&&X;^0ox|f6NLy{pms9jx=JVk2!kzEjV}vHn%p83R8Z>$nYqQQz zcxokM*_oxzdOJJwpQoJMHzlk5EcoLL<+CU+uo{`yr)^lXfc5TBBUCvvYjk3q@Jf%` z2Y|H-SSf=oSWf{{oh1A=$k-73;}r;12FPG_sgsXP@qq7o|KnXDouwAT*I z5@6;f0dtIs>;>Ksl~}pj=ux@)Dix|wWjkdI-vyVJG9&Qr1Qn^dP90N0KmF;YWlZ7= zMpbiJPM}mhzO!{}USN+>^UC{_uWOuOMzYZ>XL=!b?^6+ERv_J{j<kJyoI1 zwG~QLw&s};Yo5fn7r(u1T{AQ&Q@-}Fme)RHo458p^h@|YGa@jh{3iG&8(!F-r1FXG z9<4}J$B~goOU^G_RU$C&HY0n0v2jPmx|Zo?q?)qz8lEL9C%E7>-DyUa#;aY;V+3Ds zdvmZ!ceW2Kr?i-gePplGYbmof@tp zt)q~W_ds_oo0FhZV(m?|**MCKgokUov`f3q_m~lBi@8xNuumZeoqpyauMa3+Wd?Xu z^s@Tbs{0pLSwP=%>09F*3+JHSw`=TvX&L49YW*GdCP2fu+F!i^4Z-J^_V2&wUrT>i z>-)so93$B9Q`h}Z^nHR(|B>tdDSe+(yWv{3CiUTd@S~@;#rxRKao4I3HkW=W)RtiG zEcON)mVL#jdhq-7n|@h*b7SbIn~W-5@BElhQ_t`1MDATwS7hyD)cxm}x=&KqnbVCu znVY?#b6*u{*`xxkTP@oo`InY%YGhtF^@hfMq46e@d@~T|mXE&f`fv^9?((VKrdRC} zSt)(ec9Lx`O~qakzB~f0bFqmwp@%kN7nNWW#WC*H@Olz9$3|?ZO{9gXFTC4Lf6_0Z zMIQ9~-RHW?r?%HDynhdl*LcjxN1Ifr+H+0&l8-y{KT21jWRHscJ9)A8R8)`Kwv>S_ z<<^6-_LQTqXM(HS9+3V&Y^dG+RWR`3h0dqD;ce^9*KHi%kHN9%lP(-TAkP-9_D^?v z>VE_O{g=W2$^QZT(^Rm|%Y`nQwHF$C63vC_T=JYF&)93}Tgkkca1)49)3o}(4$9=DDt=%da~zR;N;r_8!(y}sYD(Qe*eL0&D2bhPbsx80sqt<6_B)Pq_y2FIxD4?m}_#{TdJxMd9G9%x_6$FUw2bqMc(T58c+7Vn#-DZ%SUqZRC~h9KrKYU^S!PT85gDCmH)N zwnkH?iX@S);ndt3O7#UB1`8bYa3i=3ih+~H^9?RIi-03|7p9vL`7NT3@QMm5V8Ktw z#iK?wGPYrCB7^zT$QXN=Cr!|x`PZr{D18VI z2&JOOyn+|_H{NVU6!+vEQwuUejhAmETaK z!jyH!qLuM#qfe*yn1R!bUn}i7ZAdw3rx+SaS_#i3$*Qu&Ywj#9Q~{@)lu3gh<#{S) zHdCezcw0)*$M%@_0q*yq5o5hW{nQs~>|M6=Md;@6%yZnELQ|V(w(`7@@=kq4{m1qj zRgZE#!X<4@qpgFqb%=H>8pX9A`BP^%8WqF)?y{%B{csHKn~(to!2C9sWh;4V=j;3L z?VQVRp}zOl&ep#MwDe%lsoHPxoMQVIj!gbbU6<`~}JeOjQeo2F4e0>81V7a`;02? zBL;DdRCI!o#V>ln7|ky>kui~9>>}e{ev#A041SS4#sYqkNybVuGS7zLr@kIb=qWH# zX)9HwhKxk}-p74E`#zETEc?C}_oMCmKHN{V?~}Q|*S_z^{S5nl0QU>*dq4LpxyRnr zZ6gcxFLO_R3*lR0B;9nM=)7m3@AEPzTz?LI2jdHT(z{<&cdTdrE}HYDP)T2N`^WR~ zbNZ_7I<4ADYkn$zdzkb_Y*5i5O*M?Q>KN(7SCAiboByqPozT9de`wH zu+qZh6(NTN&sS}jBBw>(37%ma&%(Z`0lz=k;P=;@6TA!a66>@K@>()TXm|%Se91*a zhkn=^-S|^%29bwiHx!{`yoqKcgR2^y+Q_`}B3CoG%9Zy>i834V4B}(R*9@)_<TE{ppNeeTk6}{^Ju$fCggC7+6gWxjF=r_{G+#AQG>thoL zTzuSoxlSv zcqT1Q?J5YWv&_f*q(0_3zsR4{i`JH4>%}FTqqCXUvp!R~4jb!u8T1#sY7+Txdep9T zXx<7GDQ!9JF752x%{q!6|8N2}BtBVVIW`qO{IQt+MDt6? z$4@#OlP_^?XqsPOTK&yy`R>{h`Vp_Zm1@YzJ2JRdt`5)MzPfStk=0GJ+gCTw&blFM z%0#XiTqo&^4_xEE+;2(+d^D9yY{A@QbL~Fpm`gXbEwjsd{c<0#Y01{tWL=h6fs{=|zIQ7- z%v|7<9VBhouqX5}15$n^yx*;SFEb#qXc?!^*|4SDMCLa5zUtsBF-d_no;f>#wGt`A zzDR+k+mk+deO<5;ELa-H)6flA3mjNVB_3t&X+CEI(OTzzRtED9e%;3SiL9D#s^jBl zOz)7k4#Rs1(67^`Ny3a-+*>pexl>lC&L)D-xPE)LJy58QPnhw5u9t}X=|Q~$!=ihS zhvnX?*8>;@Mnd~uV2jL_x)W&k!Js;R`-}&z`pBPN)UUSd_o9BaUBA}{a&Oh|1*~eL zM|%;lWG>45)UiYKNi%VcGtDaW9nu|sissv~#WT$DIj;PD%(r&z`GLV;L!4*ZIgvLa zt2X!YO%d63JL`vv{uUv3GM1?0xytw4rDV^+Uk;jb@JoqPG9qOs4rUx!chG-eovh&< zWIjpVT;!osZ#&~OZNRfr4*s%D*NN}!+~t;I4y-21jWv=gE5M-wTD2tjjtX3XACF8p zsOtZKOg(UmVT#=Gm@p@vdXd1lMIdKWFER5ag~cb8g!k@f0W`!g(Y z7j$_Gy1NyhSNzlj8V(w-)+a4u0k23CtqKD~x{<%rGoF z-oY84O>q9E;0(Y1n+sv!x!YU&=WTar?(NfOdtAS3sM|$b_Lg0mwt0q*c_5FU4nzA!#?b4} zsq#Y&M_Ys4MC7_zCZfl8%|oL$UqE4wF9m?LHo{`{<5AKJZy} z?Di!$hRidCU!BGMEF&fKz3t{5zSEswP2|E)uY32X`PJ3Sel_KV8*Z8Xn;UMO{r(L@ zXIC*-b`RJ*<#z+NO!@18tyB8lux(140VahE7Zkhc8^|mkDKIJPn44nN2 z_1>UfG>>{+d70#8&e>P7amthdx6b|z^^Pp7o$~D)2F*T5y@S+?=25RJFO$5?In5QD zrpy^IboTqyYhU)7RO!6}4)X=})cD;9WdtG=&^QhOAXW_Y* z{tdC|QmNbP!ZVsjy{8Soc_(W>GGzgXV*AH^QhOA zXN?0fq2uARox3ZRKRadf^378YF5fcc)bg!U5?5@SGI+)IDYvhvoic01b5s2O>N_&H zvbctD5qAkWdtXF;IrAwtKEHHeu%XdeXK>9+Cl)kscbD(rj?X*rlJG;_`z02^zlBfd zUci^wQFtou!g}`T6yBW|5#K^$KiK{w4EELdiTe;E*-`XTu8#f4ec?-Fk9{h((Z!?i=z4+`n$`SB5+S*uR9p1%`s$^AS_?u%?m zVvK(3lAZ4Ho`gJi%NnyE@j8_p+CZExwca0k;HOsnPvHD@mr-@jhU1KHEWb42gIrI4 zE*JY~1BPQ;T#8>VhF7*T?h5c4{$Y5Fcb;YK^m6$9>N4=$Ix#O@g^ID8PsYFzx;lRj zeP(L)eK`!{^vVf28p{s~Z~UzH+YUz5hIy(}T3d z#(FnRn$QhfWBQi{oPEGa1I|X^q$D^O{uM_1_tDS(yZw87=bZoH{tf*U{W}vg-(-DJ)}mKpLo{KRN&LaGmprxQr0Fp! zzj&Q7+0~YNf7<2X;}2T!iG?`uH(de#`IoxpsT+Qh-fuM^YHtZYd-1m=cEujPNCx+% z>>nKK$r|aSt%2~&8}Q6PcxE#^lSVAkS?6pnXHO0M(+vNVz(0BLPvK%^@sHQCt$e_} z5{!mRz)*|5-C#ub$=NWa{p3fDswU#s*1ik_`!uC?iR|(y^mBWr^m=#+<#QpTQ!H}R><*rSN=k&mx9 zmwA(4R;uPULT~o8Vpp0GZ2V*ozq!N}(x7`WettSMW(^@(@E3d6;uni{x(soRTxcle z#xg!F_&;N5w?$=*boOriYm^VHdPTsNyx+#8OWMIlE>E{==vHgft^RYOM}~yl_*@Fd zT>76o0{@VGy9K?}T-KHIFI7%xv_$e)XBMA__llA?T>+QH;IiT$#HE=vU)BLbdcBVI zS)q9WvAv~?NfYZz-hS{nw5jlU);m1B92%LQbng0l{cORnTUw|>;^&oPvo!+0OsTUE zA+MVyX3Uz93G5uks*!b~JYw$6_)x`v>3rJ%Ny4Zm%8Q?vLp&>tpE!ed3orcrcHQzf z!Nsrl#>n4rOuW~FOp-B48V69l0`8x1kSN_x+o&6ndc<9LsZn0gykyKM>S z7xgut723QnJ}dZY`glIJb-*QT@27Tl$>`HD8rKQU8u5u=gyP4;C_B$J~A)5hcButcBBp&94a1RR6SH-=?Z^8_EKOw zKl79BxY!+y?m^+WdTAz!2MHiv#m+P>@Rk4ZliCVuJ<#v&*Ci#go%Hr$i4z-P1>5mn}-BM?n`W4eiP-Fb^0&#gtn|R zq~R~-s<@Cz9LJB0P9cudlbEg^CzpLRNXio{vdZ`EDo@-tw!HW>`@kRl*JWSp`O*#F zHO23bi=YE`HG607T$H$PUGWb*>z69EyIb3`=i{zhLaB+th6uE(*=_B6ls2~0#^vT` zQlIYp>#52K?7Lcx^z=c`AD(}rKoy?ddlh;QogltdQ?&{-Rht3NN+Td;awsGF+C(>Z z22@oOx&fOa(m2X{)DO*6{cWLz+wJrbruf-wN{2rfH~{_G1F<93923e=%^el!6^V&Q z)6N$hmb8rX{SUv!x{pac70eF(2v}mTHquTS?KF+@92J-)z$^!*S2cB%C{Lxtyu84W zd#?>c%6m=rxZ8CE57s0`mQts}u2ZVp^;DMHbv*S+;@4{Uvc}pB4aVB|irmuAHm+lx zIyFh5bXQ-c4X2LaH8!Ta)3;6ZP4|nsOhzu-vhW!VdpQhR8+sxk=9BP<-=0U`jh#O+ z@;e9KZe+b!=8jzEV39eaLLc;F-;<0{;mx5&V*2+!Vey~4%!=!~%UEf*37lm=l-ouX z9U*p3HtQM^bDxU*$!1-nb%pirNh`Dv`?VRHC4ShFx|8h9xCJ{kE|Gm#VxKh|ffo3+ zn0ig%nxXoHh<`>h{O0jeMGSlO4;o~@cWI)+XW*GQatk(4Tm~_4Y@RRg*ZZW_sFvY+ zZ{0t?!5%8?ravCg={r=*S4e++=RJ~+P4o=uLD^T8{{_A0>T8mYy>yE7LTq{^- z$&xN(X2DsbhFI{&Njfys@INc`gof~0HtDy4%k^B(a*g5&v)2!~RCn(-WT(P#K34F8C%dtcV! z@d1O#8i6ZodV{s?Y1utu8_V82KQP1n`K1jH{48m_UsPMjn5p2^p+e|d2wg4tWZCrY zd&QrlZ)PG}-1^h8ajQ-9Y*SChKA-hTznYuK{W4+%6^r+-_w`V_TY3f8`c#;`q^#Mo zmROY3@k%57(9FDu6CaSVKR{nw8Gpa_1wIp+btCkX`9dE>mI=)t!+&Fb?~vGE5$}FA z(l43Us?Lh($#WXEa2j;-(=Ym-TvNl|BY03?OryTYBQI$JrxZBP5<~m|TCxW+Vo*1g zx?bvv-`huEUqwvQuKx|EK61%jKYW?`;-lw5=k#G}p_DJ7J>=4ew<(`Sd+EvK(Oz0I z^MG+-T^ZSIzh7p{1jW*Y+=4eGy+=_MeL1V+0xFOiRuQ{a(Ee-XYHbubd3^MH$D92)>Kj1hzG{ z^3H|1vxWH@fPSXPhgWY66|ooZ8lL0m&p`I9Y@vVg@TrXDbjC8%P$Le(Gc#T_sxleR zL5yVrWiy$dTPd5#zQdF7QoLV9_As{MdrRGE)IA6v#8dWb)Ez|K_zd>LqeM1{tR8qB<43z<6dTt<*{Xb&{lqr3sn+osGrEu+dQvk4k)p^V7IrjshL2RgAoLdPNA zgQuDp-$La%D*M^rUCW%Lt)`O}43W=EfML+59|9v47%BASMR+4jnHu)zMg1mw{);a0 z0Y479#1AtU?oIZtZ-OV}ec0d{Vtwp?`fIH!dU#8XrCBI9dhK0*bi&e zM~QkLd=m8TCZkzf$@-BYHpULcye(pns-ld0?hFH`*gE~GBV`=9-BiL}ad=4Vu4iBD zlHX128IXN_>+QYNBBLGo{6tK=I41!c9{J%%J__t>Y1i$C2<(S|9s7Qg1J8=zcCD8; zLaP?!^g!ll9P>fmk5r7=YUt(g4l?_T;o;=N@A7LR)s#hwxU46HR#9S;oA@d?0a=_MR`B6 zN56~1_z8a?=hox~y&;LwIeRd}#4UvG@9^HHrca)&lSq@b%ji3G;43sU-Q$5DE921u z?#_Nahki5lcyz757zu?~=*ZIzS%uEg^=O$Gi}pmU$~_UdGRaW_iSm$1>-}G3Pe)@vi^;P56IO zhIHzM@&>5LSX&otkoN+3mUorHWhx-FmMcvES*xuImze=er+2MENWVUUu3`G^>{oE; z!(PO0_N4Y(nUnJVmH0|6^sNb8*2iI-Z>>plHfF^B1wti_ClAN=rx(0RvTJ)hTzz9?osH_*^rB_1yJLn$!E!lTmeSH{@$kUW`(maQpcirq!MQfRr}p6_B4jsIKch^uVa%bDY1 z2gv*T>3(?eZ=Kpt@L76zjo1Fc%K6*mh#l)u!ryCnYrcCgnjBk$&wVsq8dw=0d6pYk$Jc|EkvsqwA= zZyB#WHr{27fzZ5EVot~>%_FzjbVnwA8oIZA+&SXsz>_g`=q~$>(zB2g=eyHg#%U0A zmv;whsUx{T!@OL^i-j2C=sd^i%Y+ z=$DCO!2@0r{X?ALY&1S@$5IW`2&LM2jre@%_6Ca<-SvR9Fa3#?zp?%(K7l3A(aW-q z5}VIlL*LncvRy{#;`}-CG-;Uhuj~D%8s<@W0y1kd<67bgc?-Kj_9#;_O(a(6dkrInR1uG|K`i!zt8roV{x|nLEMFTaoC#sYSL13IJ!y|uCVOObzNgj$jzafK z*)BQAW3gRwXb(H^IM13snnv_}vh-b#Pj}ik!V5WW-$eYhO7Y`7C~&*xM4%Sb=UyKn7$1SK=As`yN6s zX4*PeeAkJ{{$X~$_^{$ni|>0{>e23D%4bQt=-n{&>ocOhuWlog`U%wQZ^IDZ_mH19 zfT7z0hWNhq=xnJgGWZbl6x^7n@wE@-K{IHwCB`Q{WckFzZ}Ev4qo$r^JDYo2F`WZX z*>#{(_-0f4#6_%~X;ncP!k zycu{b=6&S9=RfBnN8Nf&U_JxP%lXff^nTXv{}+=ot7; zeB%~;)ilus$nZGiVmxD-fLv_t7q?zyVk`4YWTGPrMK+FFP2b>qfzjF*yN&!b>_lsg zKCX5s`OV~)0#CHPR9Zhq(u?V+3a@oTW<99u$u9ed(R_ST)m zwlr_p_K5JtXMyjww`Rty7mGjSA@=5V$ujL5Aj_h98J+j+wU%6P^PUqYl6D5T`XTFc z4!)CO;t#SeDY5=uf4lrXkt7!^v1#4?x8_GzJi>w}_KoOeZ|%k_?EU_bF>&UeyyJ8! z`VQ9nGrHE(;W6z;8pxW|W7mp}xl-1#Jz^KYi(;R$wpu6lRdo*V89UaC>ZA;^cGm`U zWi%dQ+o}rv*o}^kJ!a<%U4^EHY<%T>A(3^3=|zE{UbirJ@_ejt)R2I)$H$@f6Z(Do zF8^J8cfIzR$DE&l&B?r%7>$f=8|#9(qriC-I5X!ZzSv9*vIV^@@Wt00hdn7iWgC2w zK-#IAFNDMgYP0>$)A*J0&R}erk6k`!o9%;2*@@>9M#cl@;EpeZ4noU#&Q6ecTm18P zMoyy3o=2C(@3=F9jj=m^*PW5^&~TwGTSX7w4!-()ky+p?aumFqiIug0pPa|2&+DQO zr|HALePXRC3;h|tW2xZoKVQAlqWvY}G|03T##3lN!^JBSuW{z(7MpLn`2(5II9==> zi>@uu_1%%k7<97iZxC9_euF${DZaGe8b{h@@L2>cpVhSVY9C%`8C&KxO-rx#=bv@? z^IM=Lv^Z7>eq)uY6q#CG)BEf@_*5zUHo?ndNsH$kmV<@f(18X1j+Vv22K@J!4tN;2Q z85T{4rowBR9J<@G$#^{2FrEEHlIP6tv*p%#LvQLe-p@wY_`>(Fo&n9yz`x?V^n*si z(`P2DNKf?Ad-Q3WVI2LlH}U8M&huGCpKibK{Z)fFFHg&dbWfB&!rq>Gz3nb|;|#p> z4m$kyg)=*Hc>nIqJ)*0UYK0$@;Kyv0)V>P&l?^R7vR@#H{g5{;Eb5qS7?oukdY|3K zo>#t)5~?z{_FEUrg@h|KJk#Th7 z)=y-w0dup`C1V}=^k&TdH;HpdU0t7dYgA#s(9AyU*0}4g%W#)*Oy#JY$Z5U&c&=9waf0umf94)VTU#%Kiss?w*Dgtp$)7XeVTMPeUkhV@945OW&KBFqu6M< zV)GJ1e&`vaO7Xs|34B?j_2UOHHXlUoJIZcjzkt|xg|ZI*@J%5<{-VYOo)w)Zd?9`E z_=0tsKWxAJ>z_BOvWKa&cQ7}zhp4kf(8>e6^5<7K~Y0_Hv#lzEhy!n3}7& zUqv}-*O2tfw1>TrPkfH@Pvg*oT zSl*C6ev9Bfs}LFakhD31v&HDECnFdr0Pp9o(BCD}U%UM{;6(d8j6Tzkdd^RA_jv=) zx?OPcF1Ok>B#rU~VKvvp*0=*4F4x~^drvNJh#O?}cdyVu+WS5A!Q0(ls`NwpdDV)B zbjnLVm+-!pPIh?%!P3)tg$YHPVTm&bsl1D z+Pc?yBYs7L(5fD|v2`@QZ!K@ww8f?mWjwS~a?s+FC5}9@`8L~+zs?*L-1A=ie#8DJ zUp$xo!dvGy{km22q(L6?Vdv%MMbpo%dBGfIUs>Oa=7sF{A-+^6@@?`kbT>9g9&M$- z!-e2dO56!sL8ob3m2#pZ^M@nDed_Zy;9t-AX_sq%pWQz5V~z5SxTt-xog!O|s`D}N zVz*4d@~%5ijAaRJrLjhnJ3ya1o%Xy{X2~4li4q&HAI)AtYyjjH zHn*(hM4UC$@ZHEXF>A)?L7+cRj@F(uJLB%tx|Q>&_Mih>3g|N10OUf;A+Iy%7F%w zXRB1Ifit2O!JfI8fe1)u)yEI$%OBe>=lqS>IuBR zF@8?ETW8+(yzP^FE&4fj#$Cdb=tk$f9S7ee#>u@NB0QDKnJdVzQH|(rzh8y^^+osx z{Vj2&RnW>BA8VaL#^QQlU(Uassrk1{KTCY)kG!i~(sx3nIUb*OIDNI#O0S!s&u%PX zuOYt0@oH?Y_n@c5Y^E`O+3c}B0=|bh8?lqK5kLcH z)8lIw{@IQ*!P70!treNo22VFh{!n~t>Irl%{m7+_$~c+>9>6;4IlmO2 z#Z&2?r%t@zohETQnLnoNWxjtzXfJ&GWt(pu8Vb$cg_mSJBrdnx=AW~99)14l;LoW1 zkMexSpZQz%_G*s`T}bC0_8*8$(rZ+^Q?W7AfS>#0-XZ32ve?C0)0wlg8L})DO14Sr&RQ?b6%&c+cZ+gV0?qZuS3m9 zI8V!)%)FE#zDL;L~=KzH^TqF)FYfnk{}K4vmw~VMI?+ags||_4xcwvPuM_)n z`ggY6tLJM{7W1_{uG@Si=6s?$L!I3?2pRRh5olb`cs$LTI%kG&X0Q8Z)iX3r#Z_+R ztdoQAX&SK_Wpun;jlH%pL*+k9nF^H@ny!46%P7-EnKs6>bt`l8L!;Z=tY#0D%*{^t zB1|4*rs*kk{yzNHyp=J2KPS-oG%(iFuQF5f#*$u~B}w_yloJ^!F_cbZ-;>z-ll&@w z=8VG8?+^A4W-)Ib zBQLwvm|OJ!I`ap8N}N8dx-?C$j8pv4Eb)Yj1qop5x_l zTK?R+SB2iE&Yx%AJbD&=J>$KqHky;|gv!z)+UcJWHD+83b9>2i}Y$lv*9j>jYkv>VhZ(<1@%HIT^a8TJ={bB7cuHb$Vn0roy=eA*DfgN- zjO(YN&6AJ6`H#^CxPO;6E1(VfkB^yLmNwIbHY=dbCTR1WPe+^3r=*R~MH|11HaBqJ zNF1=g{d>L6dz^h7s`j(`-d8I&NMG(#u#rVJM%NstTkDRyHg(xIO|pM?x*2%M<)b-qiy=hNqAjXSZQ_drU5S)oM>iuJdZmLg|zgM&E_o7mKKDqn1S ziPisgH+JgH>_K=%RVBZys<7jdt+MPrqHNJ;b=igCJhkv$DVt|5{03!{D7!0K*00MN z*kHg7gukN#jZ*eIW?(X9KjEHWmkl>Vm)*eo0+g)_4_9@+ld||0CS^b5{%6s$H|w$z z*P-kM{ILu8BNrOe)rB{uPKw-{7p|pVE6;19^-^@bCAUH}?nhF7RIqL|xV^;vx1(jVblDu@Il$I? z&?}`Zw#Q$9yN&z1qh$x`vW@r-y6h-?Fbx+y3EXG6pBOEBi!LiMDau};&let*vYF#y_H|nyf#CQZQw28dTxv^i2<`1&+k#oA;q3qQbSlEES238sOH$}_ds>|+0 zFH$xjdxP(hvg`*w%kLvxDbcc-x@-~hP{SRgB6mnxY}!8qcP96|&~EcosxE6FpP}hO z=EK4qDLWy!u$i*o=3ds*C|@Oe0dM2kFR>Z_7~ZiocYX`{{1a)~2KCzUbbEe^9rc@y zUGp-K4J$k zSJ#xN?C4w#-z~l?^r=>X-qH_jZi)2_15b^2FL}q%+&P5jaJ3l_e$(m7N*_deDrwmU zdd`BS_d}eQxV8=Jh8S4a3oK~+H1u-A>P!07z#0;T#e7CG0(p3q}#Cm!1KRS??ZSaN-yTm4i~KBP8}Q8TN)O+(n+^r9pZVP)N{ict|HI4 zVExLeW5fCvfyJCwPPz?i7tcSZ-dkO;0@%ZWRW4Z1IdyDUTLsqBX5c-GC!+M)K>AOq z_b*+r7KZOu3%}!n_2U>=RRZg7bD_vf#v}^s`=mcXy{%oav>Y#W!CDprYl*-@#vh7- zwTSe`s8`hmOUE|8=7KdZ2G(qWg?!l+18WB9CDdEuhJ}v%iVIdj46OSE7W!{P46J`5 z{VwXw?t=9?F_+hGcEP$M2G&G@Wy^ixCpXP<$omp$-|B*=6MbzQ zM9+?nI4qU_`Per`y9`+ZXekr@fSI3LdRKb8+8zJ_}{6o>wIL@i2(Z} z(j+b=a}awYa?B}X^Y#anaqx=r_j#-S8ObB|ApF%tJu9AYu`IuDLzgUHc9Xv?iPJ7St?2fAi*`wpOHAR>8clG~(OP4wQKfB-Z&(C1@eYC&3e;)Rlk?B8T zUGOt%w~mt#>u`_j{kcXJ-+W{Lt&cc`oK0@|E@A_R(>iFPB6_hBa~&r!|| zKQ3`3a6P)={E}6v)N7o@Gj@Q)pa+6yD|q&|$0W?V=>~W;6DXo|^yx5mm*h{KQMh2x8_OE@sCbQAO9^}K;D=WF{w!x1RB-mU ze0OOe->#cV+~cDXZ?_o6EXG%27!rSK!(SXk-05Kb^`SOm8g2Ixw;U)s%gXvRf&` z83`>f=~(2BiVWgxKPOM(e&Ch{9x~>G=WWD|E=#?p3!@>uIbv}0=>lMlDJLbb+TqPIP_!Q zHx=Kjxr`Vw^HciUL>Y+@3mt~Y`EJDHpG2M%UT^8o?5kNPp$zkxa|F1;?9H*?xhkgYbAlht&imbVege2aV+D=F3B3Pf)7kM~&*;5FUF#e4e@)?i4+d(8A?y2s z2Lq-!hsbo#;` zFQIkge$}yo7*Du=u#WH4j6#=1{sKGvk-cYsYvr|n;nzL`3s zxlHD)8qFCaoExa%-M*v+rA|0Il*2g$ClcFkowT%J^ABD<*LvX1b8q~j`P@d%FBQ5>-HJ@$T*_3w`sfllu@kp2~D%m zXM*Dv%509Inb7H=hAFm}$fNjLIg?7-n?`%`JN-H9cgD|&=~8x(5q~6!_6Hlrk((Zm zUq9&agX?8qTBeb3B#CvB!G`b1&5tLn&wTvQdd}mIWEqJ^lGr~k-@X{~c;fo3#}BX1 zt?!O+4ltcDD+P{ozl^%7o_@d+Gq7hW=(EtrdG}ZB5y7$8##Q)?F{ovWnX{zi*<+tQrR5Fl%7T9y-xp~1es?F|@;;i!m=~~K zQc_dBE(2MxbjQ+lWi`w@=A3TJ|KDpXmv84{Z=P+r&fJ+htYTeS);_)y!`YgN-pYcY z=WMRiPA^a8f2U6p<2I6m4ac@xGDgNm{v3U=9ed8bH$~RI)8W_G_}*FRiSdyZWWC%^ z{kQSF(_r0C`{VdX+jYEGiasbp=A^^N&Fp7u9FDFe=97j#WnbV2N{tSEtC!vb$e38? z1&Q91vG|BJdcPN4#&aq4<^5GT2dM=aF7HmXVb_by5?^#G`gj`pS?D0|$!uYtL<{vZ zs6UOlGl;r!E>$V>Ya=qHn%HU$^lMH5pZ>tr--aH9EcVncW$jzulh)rfz)mQqO@Uhi zT;4$%@f#bjrO+S^xIy|NIEXxyareMuxx8m7ZB;8|FY7?1lwC?W*~jXmZ&P`Gc0pQ) zw3p3T_CP+Tp+9|mU%Z04MbwpZG36U1A|K`4Es0tA;3;?ho6xopSUzBN(~H2pm+_Rb z{f7Rv-h;c_9%IMX5q2J77hszU(nncuPH*|{iB$Sjs7ih&`9(v??^krAT9A4^+dk_rm3iZ#tdz?}7Ck_J zWM6~4H>Y9#kvbKg3;Lcj=G98i1*cuzUu-HZR~BTQhdv?EfwkBPyfex7y3L)-_-_3m z=Fv31**?L4a9tL2c4A^;rL4!5F}F7H&GwW0w(^Ug4yCl~g?`I7 zSbOG$c3Yv{%Qnrtj01M)@eI}gSktZ(ncCv@-Nc!Dt6ITD#(lZ0pMvKB(&Eu02YdM{ z+2enB5bsJHqJFkYXm4c94>8{IE!HL6%YG2)qu5+-8S0FWwco1Co-2oUOFjtJJ3;1hGxKy)N7;!I{Ha3rvs-(uSsQ&AzF=<4 zw`7EWD=0%BB4fSmqu`7y*}uY?KqSl%Aah7$R|fc%$QW_SH}0Cf-kUhraFy_kH9vj&T>-w8 zy(*4z+OnXeV{2>zQkoZmKJ1AS&KPoiZqT)Fp68pptuW>fq3}OyC zK9J`PoSJ}~9fL133SUNKw(tnwe`-d)Cl;#EY5bfs@Z&pN@|`om1vwXbm$5%Jfp-Ro ziCF99y=wJ)A!Cr)qmbFi^3(URHYj5|9ey`F?LCk$?}9_N(W||Qy}@VTt9E>=f1JEHp@;bJlg&s&ivtI@_{+Wj|RHIrY04PQ<_o54HG9P1%8;657}aZwxeuZ}i! z*$+7+-BM-Z4L= zTmtmVGXwpAsa$`KJbg1J{wOds966WM92)9@tlNOSlUmbH-gP?+PwYx#9;Gs;z4Xg@ zS4H;mH}6aemnh5!ai~9caiBb zFPfM)tZ8GT4GW!3>ax+67{o~!2l}Mj@*w*pmg>PyN;z-n?=1P{oe$M3bZ0sG@>zI| z@)ApF2Dfx%v*4D2pD*t&I<|OxP9eYIfG{x@!wDzlZz2E5I8uW5L0b?LC%E5d8 z}BzV!Hs+)VbVO1uUiP~=p2zFH{$jeNIA&WaWL^vmGYJl|Z%d|znUDc#RG zfo2PZ7Ab;DUvr^*zcuf@bsPT_r|rd;Xq)(!Zkul)$+x|v?Hi@-F`VVe9`#GMP25e| zR?>D)vo7}hsyDSi*L5B_>sE4(%xIO0?*BM9@ND}zY~iER`Y&r( z{@GpU2L0i;=ca|vpIf2gkDMmPvSM+<`co?LNNK(|(wU*=EkXWzk*gKJS^=z9>SC)L zIkfnV^;;IluRo{~jx3|B*rX?w?|~)AFF7Bp2|g7($73(1AhWej#^&6Q-U%aHHLbwA zcouX)K7NcmmpB0PY-A&CiH!Of9cP~Z_;8um&}X1(I@o#l5f_m3BKX84HXvtp$~mb~ zUi2DveAZ*13-jroow@(xL~Lkq5Ir&mKT_Mw_?%7H6cVeBm8~61I6q>Ob#7yaJ^v+k zDr3K79eqq-zVyWJwCA8R{|3kS^D^%xex80+$15$Isl1F>+2&qqcbuQ^<`Sbj#U8`X zKHhdO=OhSyXZ;iX(5inH>;0|&0T??CjMACewgS%c?U=Qi)HUFK2d|c>G&iC4_UrB?dx^DoP>^7 zn!q`YIMlv)&S?N=S@)5%z6<&o?dkB6=q9nF1m|2+h48Y?RvK&}%3M?bx z=xNF(5kE~Nj+$sB9XUkVE$~2!5&r<=mfwmjl$enAZ5cON^SaHJCn7IOktYX4p1{vq zp3vVZ4=~r@G4hWUBTqcY6Okp&UPRh&*xf7HN6H9zXc7(3U4sCJerf z@T=ybaz}>v(8pd&uKa^(pM_^W6T>$mBNlL8y(1$Yb@2;lK? zLEJK;8D2YW%LyHi_($Xfcx@(jQ4{Qv75V6^W@zciibiBbn$M_g0sjo}ADf_d`@r8? zFZUg-?&WFUWXlVQ{b_ko*oFV)w1xD2mxvIFH;<`wFpf`mdJ`#W0))^v0#BW~&oQv$o82d%! z`h8*_wZpf-5;=iykW8#4lulm@@Ij>i=_dLDd6380i``TG%sEy-)y7?8@?pS|N*6A`v>T}k6q9dihQumkC{qPd~8LZcuyX>JfxJ-kJMYdZNe}l8ZBW661WAX6 z&p7F&*7~9$>Cp9Q(utR?Q9C%Nh1l!IJc}*#@B9`%#~wV|A%BND=4DOvdplnXcIOwN z;{z@}k@Ffa;t%4AvHp$7drODQ`!n9!*S~92)$`qTJ?7-8Up{<7O*D5ZXslc`7JqpO za>LCFwcsfA#4p*(y(#_w>EKW!-}5;n{h|N=%x4gt9 zbtSz!Mx94T7Qt&diAJ5Q9SHwKZBph@6LM5+QctW+YUf$}WZ9&$Ugg-N`#JAl>u=hU zSl8JfsjALSAKy^sLZ53pg?SrpR-wPGU|*=TTUJ?@?c@7&_!+77)$8a{fE@!ODJg{`1PwT#^Hr|pqUht+Z zvDZ_On>RyyiEX5?uc!xptHh3EU%@{QL{`DKCmE+O@32Xl-tRz~{suGs>bKGvBTI*l zgAU6(%GPcfVMOE}TlK$;=W^cNlm3=oNEjutsb~M1z|khbwXZ>EH%+83vOWNh9#g-x z_93*Db+bn@o_Qnks=%ktN_<|%Dr&bt zU$OB-c6nX0OWQ4Wo?C{=cYuWF#BPBWBd=#{^Pq9*h#V~o<$R0Bz2G6SAJH!&GY&vonfFc1?>xqD5%Z;lZwXmGu(x&?Rk!jGAeunZCr*Q|6MSyJ_I2Wec?2sMyoN7}X>HB#v7_|E>9F$sEm7zrQJz!x&-{}%n|}yDGpEv^q1f|cyWR)g#jcV* zimq#+kFqaA=6Rky&t)8?kCJ|$)_K-k&vT9QRP5w*;EMlOn(qsU4N-DR$BM;2*e&zv z6z$7-Pq|`8K#Qm@u=OOee2qG7ox{=x8i-6#z)S(2Tc4_9Ya3MGnsXd5m8o?mpu{r*r9x})QJ-RJE#_F~Gsb@;*k&Zx+Bp{uN$ zh+J<1c6Xi-dU?_9Qgo-R*?4TaZqPi|{Tq3DE{tV=FY>jHzUz4= z_8INVoN|rZvFNzLvtDTJh1L?=e+ak+&j!!3XJjlU$b9D`%sh%{-{{I9ZPuEbp(|9?DZDC@r4GE>iyeuh5pVkkI^edNR^M8Exs_gWplLiDw={~Ua_ zYkV#(kNl$ZZ5|eU_JEH^<~j4!ibYy6iBxb3pHzWal6TTMm(|ky;I~7ChlEm@!wHwb z?`0Q$e{|ueylR)65odx!4ST>Bu})o#4J~VC&%xJ9bb5tlU(33$&hLreNL(k6_n&k7 zsLC9C3DI48U0BLbOhjLOZ)fy*^yjf#P5tc9V{kvw*Io4JPdvJ~@{Wbs2Oc~Y~^qz9@+r+L*>0_K-g)i)pSh%`A#9GyYGL?81I_LX2D-RVNdbds8kuYT0-?Rkt!1oXh>xfH|5uw1e;ub~YN>zUI!T`jT?&XvRuGdE*&%x|K-%l3Uwu2?=u)ftQ24N;Q4QzI7ObU9i7fbM_rXvsV-C> z&SKAzoTsSQv01B@m~$KZicV7}i}%FYlXU##jEU{1iM`4nnL{d0M0Dc@@gt`$ACFV(2Z5gp1s=k)B7z4+^?JGFdbU@Eqm)RlQHzj1lNhJo~R znTo5dVSY&5sck5GHt46|A-J7HM&u$t`WwN3Z{f5K_RZ}+ovk9iaTbk&DxuO?x}+f$ zI8DTmy>fPE;V9znqVKRj$n%pY`q z&WUt#PQwOlX^E{iBSTA|xy-{7=7hv#MUJ{*z75P^yxJ9o1&sgC+_`|qRbBUf&y1ef z1OpB>V30<#@dKwe!Pr)uR-+j|FoZ%ACyzQ6Zi``)CbW0<*|VMz100(F!KW0r}|npZrrU}6+`~0@Z%2h-*$r2ms9=wRmb!s zHlo(I3gGu7I8ATkB)oeN9RG-ZE~bxtqc(5Tn3T7YK_6`X3jI{u#9XbuQ;w46JGS!H zo!QJG;AqXGgZ6{pNE|DGkMm8iR{KoezJY!N!|EyQyg`{0M$K078M{~W%2CCB~!1a*IM!aKI{d);mOKAgCX?=LYoi^ zCz+wRoS!e@!AA21a$9Ks`89#9C$Yc!z{!c5$;pbD)Q{?|ZfP{9$Jl3HL&nP z-Kox=z;4}V>|f%OHrb!SS(w!A`&ZqMft#C=0ep|{W&ba2!oO{1^v#JE*kAE2;EG2* zJJPOmRXaI6w>=L1!8vx5jTLuQM~YT67SDci@^=R9C;2~}$kP})Kyg*Aqr7r0u>$1X ziy3Br7jskVjpPKsMoe;1$BmKHT%N6r{g{)6hZdi% zH2l?F7D?Itlr|na7@v?WCLIa?qTd1y-+#ziFSp~M-^ZvsErIh5|YU(E4mdW*>|nPXO;c+$vrsxz~v;X_n5H4`IUrurRl)ONH zPNwyz0Gm?sOK~h@P^;P@PYuZ&+y>0n7$>*Yf>VtCsqwqm>U$iLB z+K~EKs_M0~rk^)((#Agr^YKE<6~i<8ne^%Q;L^BlL7m{ zi#c?nYb0M~caK9S2dMpSW4`DYn@0@p6pN@pm+r50A zT`y1Kd3Ow+?)Zt1hJobHW3mGkvod@aA0)AV#6tz-4-}G%5F-~MPA)5%J|GvN znp}iJauMR>B9KG=qq*eI#MpDDyz==)W=&RzD(l5-EV zFa7ZY?eXXCZLc4-j=X%6sN$J2_Dd9ymk_}B4p)s&6f9qIu5i`TbCF4<*O$*OzdoE( ze*K)Q@Y{b7eK+4kF$b1iZ)TTWukq&OmtKDd{ZEFLeLFx~_PM_3yL!julwBV%tFFhN zcJq{m-%s!;GSI*NuR5U6u`b?nc|x@5iZdSOyyTv(se+76R-;Rh+aci>zO+2)$}2mc zeKh#vG#aGmr#bcU2Kgxh|CZ2Z6rPX4f9eDF&Q|e5DSY7Pf9Z_^+E?t>(|-$znfS+( zPu!#f{wzeNSDL_U;nB$3oQRE!Y_oHJT~i+U@a4omY<~GH$qn`a!3XZTg001?wFTW1 zF1l#10=|Ee&+X{e%;<(=&m+>CP?#SleM%JYJkr~UOG3#up9&EZK@`JM*8pBny5;tCrZlX^S z@Dk(q&=}!+3u4(?#!Z0#-IY@*}jnj5QM3km_qa_m(rhe%1qb0n5rEe91!A zOsb(}5Bd}zxiL4|28) z&lsM0fqQyKxl{f=TUqkXJnS>}DltZVDt}t}#QIHf-AvXMdiY-7GA*GPJ7-6bw->T& zkFt}!-=4SY5>A|*Gb8LemewG9sj1onZR*F769tcfi!JlOk7#CTg}&058sRYy2O85( z@PsZsZSEoVgPr^ODn|zSX@4y^k-bwYJ9+api85^LPM$+2)|m&#B}%Y;|CMXnsWZg5 zN?64M?*)64#I_HPZ^!x`W#6W_1~pG0e9z2Nq354(2j87(WwW1vbHjO6fT{UC@@L(VVYXmlSA%9C*3cAM$fa>s5ZKh zHL|CZ#2Ao6&HJtc7r^Mow$a|eOwkJ+VtL22pFKavyQU54N~;?r*N`1ItEK=l#?udy zG2eG}gr`3jrmgn{#z%_eTPi-g2ia-up~a@C1s&9}I8qe;sOs^uhco|$ZDDNHVq%IV z#1zAtYbJ1?=9g@;!05wI6JOZyQ3pzdwbZd;0kIKB8 zQO*6$Uk_|83>oclwC$`x?&+Bdo*|AlLv2>_%y^!u#6Q;lr9-rqK{^;Sk zy^Jl3J*P>#7vTem7S25q>@r}Nu@0U!ck9Ra4$zgjQB5l?4d_DJfGuFNIS^9n_*NR1! z!#6!Q0~5czi!slo&ow45ab;CtYtPMk$Iiz_Z*S{|TsXUVUwcrIJ;WZdjnrp&AHVh$ zk6v-@Ndh+-YaZ|EyXcldE|&Hrh1fe9hVGpUw08&J47$%@T*9N;C}(WD8QUn{)q4dl z%tBzwUbMV=(D6;uWz^!Y4V_Q-`gqivPr2;*^4p_@%n`xX`r=0BOywx^P$7O;A@v2a z@~JTd{A%)r_n1&^1-aun{a-t_gE-V?io?k+) zD<+ti89cL_e8t>}rfu^2s}hq|ET^7oPDAnKfioMZ514aZWxM??aJ7D0l@rLQ%?f0- zUQK&1Wqg2G%Yyd2p3s>W$scptm_DQpXm-Ii#N)mF)yIw?b9X@h703YfQGGq~(!-A( zC$9{eE?S2Dc`)bbHr-df1lh?w*dnrlUJRK1vX?W!$tG}8Ms9o!<7e*emrr%qLjlh1 zM3(aXkCf|)e9YuJ$n%<0x__Vop92}!h5oR58T=^su?Bu%edLu5Cb*qF`L$7aLcC7C z_IB}v<~X*_j9T&xG}}DdA#Asv9`zX7m07Jrw+zV@RIh_hV!b9eb|0? z?%(L$A60d1D&NHimZ$IsJsbW&#gv-u1;N=p%Sid`33OHRG{4LyT=?q*?gVo|sZ{6}9IuhTNaaT<+od zz2HlE@m=sG_Wmo&jM;K%RQC1-`1p@Nrzo(ZX6&puH7?5e?Zt7y?J*N6V$ba>_(|>T zlYb}%y_LHjJh!Sn#=ZHY*`qw#JoMu94cj`YMR*mo3oVlT&9^yH@=-Imr+f#+34ZqT z{wG=t@{Y9&W5y)>x(~YSH}J-(KO={nzT%tI#-V8icoDxiagn@+Z1Kx5@7p!Kd~>FN zHLz;x8_%bI$4zKuDSa;N@4rQPmC}2AA|;&Hi|@m{dy>A+MaRd1v70{rF}R3=OKbbW z8xinGe`iVeby7#s*0~3l+EdlRX8~=Eh&;--;fj9uDlccF7aa`O#3A!XJ(Z4 zMH~Nk)8&xr+57dnKi}CoOKXr{N-`aO3GwOYm;Y~gq0$uDeM!(?Hne}-?@8NX-nl5+ zB5y84Tlqh~k-o^6&dNdhYG0c2H)7B?4t?iBU)gMXMrLdm54gOUh(g~e^cDYzM+W4{ zaP}K^(4nsmrSWE0E*SUE(>G3j`yjmwXmKIvllh0X)i7V{169YK zq#n!;Y<%YKBH1L8qmmzSv#zO-bpYi?Sh>PHE#&TW#`7oOf7WGcmyI_?*i#Q38ntm-1??hh>m+L{(BaAflJi4Kry~L`Mf!<6ZKUSAxcf5!VCmCJC=RD-V686fdZO;z7nl@#p z={MOebFm@xo@Bn)2js*IwQ-62>|$j73&eEvTvW9^Wb=Rgc z1~`#@{0Bb1sPyrK#}6yg;&omuN_5n^zy#=3Z`M6_0J_9S)ohbpBKt#f9$B;nxolx7 zE>#FUs*$G!)aOy$irO8H&+W#j(&f?^U*Dw5{P`#9pI1L%@(r?)2I~h1P6l#3#+(tq z5*K)7Df&b{+7V>)MEcxEpM?wk_tWHz6szsi+1OY z&f4B{Ben)~sz=t!HqBzb9$~%`i>U3HWl|O|fw8mn&L(`xY~Jh9dvkbij{9CL?Y#x? z$TH}^(xf^cz-NSKMrkgAFWDb2*!h)#4UDZXo?!0$+QXr{Cq5H6OXvgrdSws%Scrb@ zx|bY2`d3f?meOC1Suu=0=AG;{&DE}Rm(2?0Ypz41bN%~09B!nY>XAXKvn=lbQ+4mm z1IUwQ$eoqQuLqdR@Tc-obAXeJ%+emWBh1-)*6aF_9k-E}wtV?A$A-)_4{c^FC&=|# zzI@_r>Br&^qT{Lkor_$<-kbR{eGk&Nli2)yvgf}(Nxz$&z3z?g07+hn1!`GN!b_jbkvex-kd zCUPtZ-|9Wp0aIHc-f^z;8o=$$5cY`9hFF3<()&C-hTPhLUGntw>TQyP(#d(`9Q9JC z?gYO(bfYbB)^)Fgx1LY1#>;%rn&@kcxA#tJ_vV_!N^o=gb=W_-=1h>hhuf}O(|#%z zNZil4l&{}yQeVwv4||y0I(YT9iL5~$Fh%7n?mO$}kNKR*toTAV?^P`?JgfSR2asP` z@WEbWg~l=wn)SgO>>sP$YiljQTa4ACtNN!tsgEZPu+DA!8EQ~pUZ*{c<38ZtOHM>B zeOj3tdF)os5y%18o59iMT;vwNN#@;$yly~N9wL8bGIHyZ0QDrYiCG4mGuA`cLhE_H z6FkU&YGO~LpU?a_^ZKMd2-eeqoSB<1bztp-fBf*Cz7Ra(@auV}Q~m?+{?A3i+sru0 zNlD;Wr8K@=aLAsy+ReFK)1Kyly}p#Q>EZ7r=M*TX=p)E$qcxtluTJbjcTV}jP?_x6 z@jt5_s;>`SpdRdG;!(f8EN>uw5kgiVmlGN2w|K@V8&8bEi{jfF_kIS?RN_M^&X|d; zyaAbUcyz{g{O_WAWMn=3ths(bauAz#G0(2Y_wjVrdgvG6y*t6FbWa{ULmVa*K-TCP z>81@lQ-Ca)&ok1+#qeB!c^E#s_OSrE+v=#8&p;F+491;z;^mp6a}wQ@Va=wmkGdcneaqLgSA)j8wdOn zH?|c7Rsp&Ld;XOY#Yd7c{VwbQnh-yfoiM{slWQZ*--ITf?Wj+WKDnH; z;9pPZTnT-ki=_`SG_4tQpf5f8So>t$HHlrUryhSC`b>j9H#@%NfIQz57@&uK)4Ngg zwc^3@)Bkz4hiV*jeO*3G?*;C459O}>%*I)@*f zI86-w83*U(;C#1>^B6eqMc+tg2>y#^)GUAB8|k}X*7kyLT zuSZ^G(5HcGg5Hq7ms5-k-4te))L0;k73r(&Uof6pW3T1BZszP^tBou z0gjE0>3OXoC|9YIF~)dDw%=TA+)CCe;mbuI!?v{JA8DGMT%w@e+q0PQl1sAaRqWQ| z^aXrAqxemWts9nM>*|PZdhSKq(Q}8Fm(M-`F7m;J{2?c{u&(J1ayk?PC}#}XmrDG_ z&W$MjInJ7}cw0GAo#YHFr(HR)-n&WO)!a$v>o0Hd$yxaN4_&_Y`1yBf|K9EI+hxPp zzXzT39rj(LIrqT<9pKq+{~h}m8-jAg?1$^e2JXiO?#~`-|6*_0*beie9==rES$44O z-X+7?z1YM(vyh+O^TgSEV$9FQ$b&fizRX;dy&J_(v~%^KlP8#N?}lXWVk?T)`du4)$)(!bEVgSPITi?DU)u~#~|7`YTjhF~wr*45r2&(5u9ULSGo+%3Ps&V8GA zWam~BSCvgGd#x9nwr7D!O{QNfuxV8{Om$#~=p(Bq2K5m+_OEqePp0>&X90E^w%LH4 zh7Rlc1o{nI@;+?IDcF*Wu_X(Ly-voKe628$xE-5C^~CPSp4ovd*@2!viTsJKSTV$o z+|Sxv7c!umcdQ+G=b#;#4Sr=yJ|)``JbC@FwjzBHoKrj>qs=wU?+RemP#eZvh`Nt5bc17`*fntM6Dz}qwNLT` z!`mk=eMY9y=R)>LdcH#^a_K$nlRD$tC(kDSckGjI{O{Q(1AX62-`~?dfhGg?39=gd z#O8oYulntijreHw%t>+y_PtgVmmD1`P;}ZnGnJ*(E1m0WLP@}TBh4EuTv)^ z1G(qfF|RW(e;0PlH4mR}$9$K${y%BQ{4F^8ZS9!v^X_o+>0>&ZARBo&0{Mx2x=_Bu z@6w(rOzY!$@jTa_dGxy9u{{HSzAt-5&;OS8%m;|iyl;C3c>f)H=5p2K`Azl=a^U^g zGjVc`f0y=5keI-Q?3wl0vBTLjRloWT_6+#-+cSQJyR$4<0csN-!`H`WrD3>-o`)Hq?+k^QwoeQw9 zuz`26R#M2=;(U5JII`8doEXl4jr7l7LkAZl=VJKs`2N*F)_$;gZ4Gt&kHuBgxWEV z#)0C*fewxfZTGK1XpJF8Kb2dr81hnLL(7OEA18L)MZB&H-yoAd3Xbw1ju4MH1bz~SFDd1#4dbo<&JdV zLlpCSl6ZmEdU_|yrp|5{_Y?9JwH}>EYyrMk3`IFXy%YHtk8t0+w+nkbJj!g(&ks5_ zn~U3e`cp%H4yZrGtRJC2{rK&QAIL9Q#`zYCfjsTIUiRz$Tdrk%kI=Uh^i6xx^5vu8 zGcZ4z7YA}zmN2%QK(IFFa`yOeuaCGuANeqSZhWQWT>m23HQ_Uyv4k)6uk<%SA9sxm zoYzM^m(SRHi5&?Saca=T!E-6_#Zy232b)JQtA+VGU=O?FQ{GdUb`>L0+&@grCLIUL zVG|CDfwLPNY#g9T<^Ud+EHAV14-XHzo(CSZR6?bm@vbY$l?6t+KXmeH z>GHw)dY$OWu9!)cxG~$H@=mb9wXaNZHN|X^F{eG7`{#ebe8(1L4~5M)8HfpX`C@|n zrgz^fCaBne^0H-ixc#4sRUQH_zpah`4DX6(hMRXE-+2DKn?T<^{rbDK?{7@& z+lB1=6`TeCJB?SuoA1xQ=lS2#zQ2mN+55Nega0l2{%YpgZ?x~xKkvuBkFoaryR`2+ z$sv8e_WkC!f0KP5{2#UNtzC|+Ye3ep$6L01+X`YL`-q8LLJrFmCd;_ zQ%f#jyF;h{j_qC}oo~+osBHQs{XpN;d2!CK5Ys+S#-P_fQZrP(T;b|;e-3hxu@2@K z?4 z7w2G?@!iI$O;#fF^t#73pj$IcX6qr&$x)8bUUJ`dn82As>{B5BcZ=3_v9D*8>&ya^ zkx<{gT=Xqpan`>#@tbnzH>}D(+XYVj_-cLxJB0o#9{pkHB)a(Vq4p}kMFl>ia9!^$73xp@;iTmbGUs+?Zp-=>8@XXzjtD z=_bEp4Q=>;_iRU}e}j9nGd2_7&oHwxY5M@V>$%XUW?V@-zR(Pv)i;5$uE)OW#h5-(k$D zg^YO#&*>Yv@b@_>sg(_mp(47r)rq~ldPjKh{`@&Asx#p2r#eaOhCJ{6%NV=H5{Hi_8PmF%+_An9W`9oZ(QT?n zZ33LVNNtXqpMCMM4ERU*+06IxJhz$WSAfgnK+dTXyx(ncnqg~1ECpxeWTmvdwfc|y@ zZzHmyp6jPv+~?aoO259;l_mPY^K%Ja8Jgs<(_I7V4*0S`C z1*uoHzR%e#@(Z%g-O+wz`nq=Uv1GBH-N>HJjgw4cJ?F>lr4GU-o|7MQ1p12R%E6Ey zqdEJRT<`r#z~(b@PFRj|JC74T=m73sWLO^as#@P?4}Q1fcl~uQ&ZHQW>YyK4j{iUm zNqr2VbCz-T)X~87S@p;t`fGCr)koEH$pW_KqsCfK|D5wY4SfqWXkV*0clWYh-77hX zeD7l}={fn~+D~NjdSkY>^?vRD^7k#bXd$_Pxed2*j=&{Fxv4F{m=Z9Smv6wvxKMlZ zt^AN4)On!Fq0>IY%!Xgkk0*((DQ3etYps32?KN4oC+M4CO%|NrUuefeOyyJ%AF~TR zoJ7ZTplg&%qd0?n$LrRhXOR0H=o;nJNS`Xk;KhT~w%XPHasQb1&R&>0L7#HMCh_#_ z+|*{q^b&sK`yIAUof{Nqly-*yac-{1I>>+h@I^nak= zw4)jTT9;(s#;Lz&ex2Y6x-jI2tTu{xMtcDN3YZYnp zMtQOW^G5Ca=1mmdv-@J;KfCtH98wJEF7Y1rsI^PsKiQ>rzA$&Z`O=;?U;Omlxnd}N zJ%4{8y5FA<8351giPHwKm5SYcepEm!BE~vX`S!$t_J^tCwUvB6i!aX}!Zz)2>xvYu zcYS~u@gMpmp7PgRRE}vSd0zowhmf-wr>l=?UrK*va9=-j&504NZPMNs>H6fAjC+hD zE8gugiT;pd4_3VNaJ%wPm%g;Neb0QI1$hyLF3)?5;$J&=N_nsO#w>HLEg<nWV@mBRz9k)VH~bz2 zminifjD8s6zfSDW9&j%{=ztD&>zCxNn^|8bFX8gfzC!l9m_{cVDnauOYZ<;@M(k=A*Q+ z3p@r_tvKu8;e~CP6Ao_E^Lu!{n9ucmc2L6s+-(V)P(xtV2er%!_G&;Wb_ zw+8!F@0Rhd_K1hME-_)=&*1%u)Ei2#6-6C6<%Xb3kK51ehF|iznYK^nzl2?s#U6#M z+C1#LN!V#|lhdGl=Nx3d{K@aR{O`@TeYQUIU_3j9?K%%T;c(XUnHyy%;Ft8|uqH*k zs4IrfT)@6h=xDIn4SQ>;-~U)gX3a6x88+-?HONuTnE)xYVZ#}_(pv{vw}OUA_yAt9 zwL{)k?N-L7HHHHCz(7azNveRhgXBd?4~w?Z79;=D?nU2&fAkt_?gb%aG5k}=oRZ%d zBsW~=0tN3Pf0p|J{*^B*AM_8t9YTjjuC4}WDt zGrK>+`Z)4gG8#UnJ?Ped@%$!d&gV;KO;QcCHBBp*J2EXj&e<)wK`dfSl`Drv(3i{L zTj(!-{T4Lw=<%qPe}nZ+!*mmceknbjp9jkW($^?adCkawcIGY|iZ zx!nxiR(DZzN^$2-z8_$olmLtKx)bxpf9Wy7eu=fRP^nq1c>Iv|O=j(U#2YJ4doN|1#McB}*d+FZ`^pEpoZU4}%++!`p z!W%lC;rc*)igN&*y=vXWFeRVG8=}cp@jrCkn>HVFv1k1G=Kfd)b-+@+nL424Z2#KL>5AZ!2V7r5aH+wT)4>{S0Y1fIe2OIWYIoXcGtUKVJ?FQ0 z|KIgK>!H?8Mwb`hYe?^h;Vs1xj_CKx#q;F2v)`jvIquBWwffY3;zc@3B&lm?`sd#L z>CXM1>l#}%pHIPW;{W%_6Kqqv#8bfiGaG2H3m%WvWE52?m+`yQAiJ`%>2Ywlm*<#s zZRWeww53-05w*vApW)L`C&^TeVlGCST6BL-B=sqMUJkC@8n}PW^)rHXInVCDp1yL% zAMbbZZ1!|wB+%ol4h(9|JGD2z$2(7R@2~kSeQt}EVV9O2zIpwXwE5<@H>^E{EL5Gs zC8J~Q1HKygm9HjxX-`HEdb)aSjM$PB+pEDZkSy03&`&|fKJwX@Wv~|veqP4+nxsjn z-b@eg5`SP%VCI%SJ(opX3EM;VOHZ*WdIa0B4|&qTKF*i$H{`#(?D`vuIYiN#Uhbg^^awZkOm{54P_ml>ObkwO2M z-=}sV(?aODuI%Yr|7(onJH&}aVlUb{Db#=V(fq!CrwjRZiwI%IGsO$C=ICKNp_RksatTZ?Klhv-qaaNNsv}x^EafX)mL{--n8kKj>un z!mrdB)gU4##qqt)C~sG^Xa4@XzH46Bp~E?I`!%k6oac$X7uj=lzPFSZ^zA0~17Dk+ z^f?N?T{ZGuN6&;syPA`pns1HZ)dTi0u@|mQmfv3chQfDtl}bqK9NjC$XMr zWES>BW<%5K*a6@^WS%}&jNf}JxpIc&OGOwNpd%15H89d32Y4lZzZ3e4(ZT*M)m}$aW6e0 z*>lt{&~( z3HOeTwJW~%bDrB|jIF1uIq{rrgR^n@9_9PtwYl4mcbhZBK^Cd4N!YZ~&8owD55IHf zVykMhoTMM&k+-&<;PVK%^GAqpPXYcZa+&U@?#gWwe!MNfenhqLlA$K6$!gt!O*a57po z#AKJB0Cw)xk;f*74=YA(#~|BfGJVfAxeYJ|cCFbiO3O9=>o!gWjXR3HEMM-!99>U0w$F%fZ){@h@#tTXTUM zq<_8OiFo3f05W(7pW?m3?5blKE&Y#aFO+J|C_Zu!|Dh5ZF6Z|yVg}RfZ&|kf2bF4HlD4`V~pze4S}o%>419ZAYW%GdMpQfOgK>N7CY<}{L(Eo zyr*|+xHmbFVd<&oPhe+@cA~B5qWaY0mrq)pIrumNKG0jQ2xoch^ALR$^V*F6l*4$5 ztyAkgtM)kUui&>heVN2qFUdW)ZT>oHi_rdk?%Kg^JbxSbt1*{sIYJxZ^=7L>%bbQ@ zX!H{EAt#NNf{`N_&{B2Xi+NAH6U`sWJLS|cbLO{WkBc{Q!Uwmp9?;;_;$V*Op6HnI z?yc?O7ds|sxrf-Cc&vvQhx+01y`EJ)Bt7flN%*;xK6v(tWWzhE8Scu4a~Ifet92N^ zZ4x?%jf`*CO}!#pn*)DtJie4zCmlQo&y0W%UcJGbE~O@E9{gVh?<$`%#JlHS>wh9Z{n1X^D8Ua* z@@e~z?<${2_+Q9an7IQ>b--wNtm`uzuM@8<{46i^pVdMOBQR3lmS zkvCGGcoV+d0h6_I7j?PhYjmQ&qz75EsH+EmS_>`3UfOt1q*;AjN*|AmRLnPP<$Cg@ zRexE$Q7~>*dz|sb^Mi>9ys5RIX`B(zqA^z$CsuvI(S0#s&iyg_9jUFnm)OG5agoR2 zGtL|i*!o=RV}N|+L*v#xP7I{>dEVE#Q#Hi<#+$CQoVQsk8*Dsv0`+~oTPIMqt?D=w zlhf9Ld<(R|v*_GS*lucL8Es(q&YWm^&-Nh~)W$^G(D#XM8#{<2c>S%TzgjOXARZvw zw|s4IyJ(`ebk^5F+B%3`5YL~Js4$)Mi<%g;IdISRFIZdQwq5*Vv$}&Zhrv@hZA8J7 zYR!L}HSSXO5r@D_2DFg>Sqe`nuTS`ngYP)_4&B3kGM)+1j(t{fLB>{(-@kq<@{F;T zV~5iJI#V?>@s$q~AK+dku@s%BZugC0cRWeW1VRi0t(6tjh6+%<$ z`_^VbQ+o|PbzSJvGe8W&AC7u7+q4$A#2-?-aPaENL z^sKFoIpjX=>HbEK=i$!^cwjxa@aV3b%u7Xk`S9rbJm!87ek%9P&vM4$@m91nz`5Qf z?eUyYZG19?o6|1*UHwBPb~kb zlJURBntUd5Z!hapk096jkZZ^B*OKU0$+{l+O!?6-FuxS5ei8X}oOz~v9FJzVLNob8 z1<0%zGRv+H;v*@(UWx45z#iCcWRUnNlpjb?%j|LC&Et`3XdE-$@KlDCJu%Bq8MQhi z?|Ezl56Z zZ{2(s-aHv4*%8SP)&31~N`6i?_=|%-<;ql3OVawL$Qi{;J>04d#qV}N%e=LLecsr9 z^?7%`A)DnN#$4HaG4g$}Pc}~mM;^b+mr{Jd#N^-asji&~NSO7_&-T5b-v{E_*-nuD_8gP zBF%Sl|8rlZZaV+=y;3K}_Hy~qF|T&VJe+?xd6LFEc3`i45IZvpk4NG0#nQ>tqDn3> zse{-}ar2=>Ieu$o+{5jX1#iqXTep+jP+tAHq74{?x+hyqd~{+eImxUuCYf7*`FrF= z!!P%OOVJ|MGD5PmO>}>fc_AOW!gQcR16JPJYb!fBm&2VOIztyewtmuFYE`4x9(n=$ zZ38v)fjzT;TGbnnD}t@>tU)yQKWs$1M$wLSh~~}(@LWEA{v4AUkIbvdA#aQMUOF

lw8=8&|SmxUueX$6D?it6&G2&zgG~%;_EIP4Q)@)QQjUFu|2o*h1yR z+7)LHNp|ZUV&Qrw%Gy_u__yW7kW+ua>iDmJn7HV=rdci@#dTKaT_X||@P1A3>@3Ab zL_hD$-|cH7k8|el=E_jU%3f+}%MMy|Zn3>Cqu-2cGx_Mx=(kOrrzo1pt_kTm_q+Ii zhY5UM=U{X}x9$;VXQBI=rJK5t?cH-Csj87x$7HwlARl+a3%V!y))G4VY$tf->=yj; z%qa)RdE7u(<*d9<~1>(iAMV^0&G3 z;Yd-b)>|0&pM8{j;%#sz{UrS-xLQ*h(jTsMzclmw{_I8vsXyz{w;k}`Zu*r^>|iGS zN@||KZ_F3lul4E|@}xxl>e8OCNYgL=K=&f4ZRB4Ek=xQmh0uPn39ZyT=oBt)L{_82 zcRj$_8}u`ubqaLJkg-I4V_Cp6>A!!3-_v>Co6~RI$egxzHvOH&{*e#C>$5J*>;D_` zIth+HNgQn_KC$*%EXF46pmxM)(FGd$dD+=RRN0i+5CF%jO;p+R7xsJ5>V{gr`*p*# z)jXFSNj3Y<7joi-gX?hx%(Ko6YFWd#vO5aVv2&RZ6QHYf!Fp&BgE#je?|j(>g#T~LEA zsA0cd_p_!*zFvG><>Ng!$Y;Ddv1fUFRo7y(8vDCZae{xj4gICL2mg02MqWKjyd52q zK|XGrTAOk5ah0Q6J}S7~voqJSpLM{_^kN7D^mTj}>Gx~WX!v*T_=Uf1zVrT=N+@pzjsTt8V6=$MZIJz?#7n#=V4`XXOC~!I9|i&;NTK-S#|ZC_NaO zpJEY=ytK7VnZeB3yf3eRIxm=c>hFm?=$s1SKyl?{=zeOQbDp1k!O~fYaeeH=Q@jkG zUPC*jp)4Cqk^SFIpXBE!kps`u$I{}dMV97n%-H&2W3RL43H+$k z;N^f1CmudtPK#fZBlDuBI1wfHtDLnsWI~V~#xE;ru0*F$&v23YB0cEgN!Kk9 zTm7r_`sc7VAD`t%9haK&=zm#i4o``BuZ;7*(70Wa;f^J^d zVrU_L6fMfg=b8fz!qhp}XRzf@+CvZial7=Z>UYl3`Bu=zK7(Ceznm@;Eq)AV+$6K)xJwVlW!Bf9^Rs@AoyW2Je#J*c{lwCbfEt z*}97VQu)g}Z%BlRLw;Sh9(4UZKHufHI{qKv-|FA9j{g0co`L=o^!Y)>QJl5SM|C|h zlFH+=MfRe*wpq>fmuZ80oJ--HJuBZ8os?>u*F^+O|7jgT}RAP z`y5_Cu8Ai+8g$AgqXEJIPS5e=noA{J692o5A+CT>#&HIG^;vdCi40wEe+&s#pI| zb88m#@ahST^7dMIG6vjA$5|d_EFbm35MQR}zj*K#T758>t2VGUFFx|&twQ6PZqDQ* z`z6C=2jSzA>+0dYZ0PB))x@qr)z z|MBO8bN26i_&-yf^-6JY)Tx&bm_=__=e+W*|0>?MBS;;^Li37vFdn`CP-gK>TP{YLEB5HOt0c$xQd=iA!T`d)f4 zE9KoYqds^2PGjDcY}Hznog>vbHU_U}mhm3zd5I&~z%8cy`sAwob5=)Vn~+bwNY9iV zouBe-k-eN{%^K6GLSj8fGVwR~y}Wh41CQr+(U$I&((YoOE5KH>u{*|C!x(>$u{rNJ zV_Cvj4l|ZpePb!-UDc{~+HuBG1|C?)dF37CcIgw*cg?Esxu<^C|Ad|?`>FFx`KwNw zWweP6KjUU!oALJfHowib*`Af!IR=?-vh8_insYlp1`S69klUsqh|G@Ob915#pUr!2 z?A_3v^#$lgyxKX(Mf=vS0>=0d{_i!|aLErQGKir`Zps#sKNdFR?tpvVJwvYGqK=<% z-pCkqjD=m2io$Er`$r|ZILoO2Q|t-;3Hh=P9Q=9bEEvH#$GWr&r>!P3=Dp#xo(HEf z+HzeY##(-l;Is}8r_&FIJr?gdo})wI6ksn$&s|IXrmADDf^&3uI4}9(m}}k}&cXBG zG)C86ohSs({}r5r!^8Q$A5Q7D?+xd`FmULf;5;EX2Zo2UbqJhs?+xeqp>W{2)6uq^ z#2(J9S|>Qq4-e-{emGSZ0q5DFa4f&go{|W`2XVo9c6d0YemFZX0?rT5gVPq>dP$-h zI12^mhr`2}?}x*h^g&$7-MZkG#7&votEg!t|N#Hy+Je;%Z{PJKpIF&=?&Yof5u=b?!bO_F# z;o_i;CJ}NjnhKF;=52xxP;A}q+PGj`x8xt|$d{1z;4-coo z4`=6aa5xjqFRz~%3a1laAUbYV;vnmLj|)q?Z* z@NoXz4<~sMaK3gPoYTZg64k(2EI3~q9?oa{a5x|5{CS#&x7uOg+~dHRCpfjk!-@Lg z2uEfxlwQ)86M6gKb+EwfKxpbj^%>~uSx6yZX&44j&v8oG%U!r^OE^c@c0P9tO@AoO!xWa2_5W&R#zpGn~DXCPyDQ z4^CS&61Ma7Yl8E@@NmB3hm-HX@!NqGFJ~RRtfL**)*NC3uJN}K8MckaXw?U){l2#8 z{tL8mpTCXL;l`Ubzg7-2-mf_0{e;H5a`^Gy<%d&s5pb3bg=6{tImZSs5}akj!i+<)t%lCPL6CWPVI6s`^`M58T&rsvoiI-jg_appm5Z9FqP2+*K=#1<{ zjJba96T|bsf3dzlFwgVPZ^O#Jsh3#$;nW4%IO%UA;8MUId&^oCl{dnscr7C29oc6T`zmR#RG5}{XTy`c?RxWCwAh% z>F>XGpJMRcU4wB3#n$_{*GvAbVmD#(oby<-eT{X6o~|tNbV|=2B?q;i9LUKJC1SvCkR`$I7Q4JF?@T;LI8xj?Vub7<L!)2Fm{>Y^VTktoEcx`%dLn`8)&e#+lQWY{(uqd#`^>w*ikG0)#d)%k5$+i2_z z8;_fLfi`CN+t_)28_=mPI%=Giy%R3bMxnormSNjyj6UhikJXa{a~Al22!y6vfmjjaa2VR$Y3c8OgaW=)`e(M^e7&U9J!7CBVXjOHwD=BdKG=^A!&u?ps$H zbJvO8_<52&8_LP}Bj}~PUVZ;VzAxloYXn*Y`gdwczMFE|$S1B_OxwzT_tr}uwfW_P z>r;ZWr1*g2+lJUKSeh|PiG zOwIdkEgkK{m~VQXJHkE}<#pL>a-eiQpZq7SAJcAQIdpuHK6Mttr^q?g7}Gf(+9Q2C zbk+R=cy|K1sydrf&rE=CH<2H4!)4^D^85`~n$#oYPEH&H-FbK7wVct$m?q_shrg=u zY{x`$f;mrQBj<-~B5(2$#wHs0>F_pm(6z>@9AwrXW+vHh!`fzDI5{;@Fg^L4)?5Ae z3K>s1wUSuZwBzxwU+y9|>sk6wZb)6tgCDf#E!8vL`l=6*50XFI$)gCIdF`$KMZ3T8 zjbj_W(fsCyZ~P1YsW)S5Bio@t(^)Y;H~iJm=*KouVGPhr{94gU`gWo%KR;MBk3(4p}ce$J&(FS8wh= z`%^oIZT$uCb4v&ES-FCLMjnX1hv9q4!}8Xsx1Ls4)*3vooq3Gg^Zz<73#7DW>*wun zIdH6u5&bHe(=U))BfEqBKXqlliaPn@KDjiVHanPGotKr^dPF;}>K{}zWh3MI@Bv;LNnM~T1F5&k4c%_c zsCVD^gMIJD%_~;yj#M4>A2rMIvo z)raqJ;z|d{Hh-2evzV%8i#%e*-WJn7)YYYlpp+7Ef)75o1Zp1b`|B1J1N zXD`SfnTJZq)z;YrBQB4m*6<0f8}CQHeiuENWc~j}_-Se&l9~!n-xx5dk(pNRH-469 zzQ;2^<(a8G^C9kkh-ao|n$*X_5uP*qM_g+5pG1!*ZN1SEj3=+*iAZ3pV8nS}dZh!M zD_MV-cJgUQ*9Gk7(pqpnI?3{ecnaLBzQAyJ(7wycPn(AV=ogLU61R^lf$<&s_i`S2febE8^=F$R+lAa>iC=Qv=txbxcj%e$z=i=l%CS_U_`%wSnA*+uwEO z{0r2e>8C9(Um{6vH1umu-pD_=SAN4-_bdd#DFsdvI5Bcap5(d)+44@PuwmnwK0K|xf+|oOivvC#h_dXkjvFMriAmO!PB|GKCO?p*!Am? ziHWR(eF`}GT{`wZopw)~sbdlywEGUfOOItDkA!2%<%Z7%wjOr#K`GQ~<=u+klg_we zAZI@LMR2)>e2_|XPqnop8sp!oX^xG-_V^+>xl`HSMZJOeDERgs_;w08+MJ&!Fc=%p$n#@1QGj~}gP zx+>WN&pfJN>@|#i$H&Oo=WIV}#H1?7hdnfA)vOBg(07o_ALg9Rio2)@3!epYtBwU} zqhrLnrV47)bjJ*RGN;h>i#F0$J#8HTk7d~98+ZqsWR~Wq-r10(=0b+qAN1YVUgRL} zMR_ke*VwZepC-?v1brMPk97_?gQcOoe-7pc5~K7iw*Lq8Nn6Ur4MqZqOY~WaPvG9y zyw~p;?(c#fY%z)5BdK2kZZi2lKyDB@=Zi}D?Z91(ogDCu`PlE$MeqeW@EOTU$Iepi z%t+IhivwG!d6QBsN4zUM1}@>;b?_ygTs+%+em%5tWNeLX1v&+ zadPBJ$d^&wMCG{L3LdFZ)e3*K)_-(e`|fiawrH=C_D^Xa|5*Bq5dv>ql}Ii_j_nC0 zk}LTv1@w0C3T6M_g(NE_-00NPCc9Sq()&W$t({smV@VE08*YBsDF|^Z3J3VRbbX-I`C&>+)ye>O&8*<9ux0$s0 z_q3@w{}b9NAQsrg7*t1?ahvK-~BH*-$S~; z%J+S(^L;lp232<_qWsvLg2ehP@?*XJeQY9m?q6!!%5{r!UGX71pi>>!k2Buj(DAm_ z*Ikp?i+$#fw=Mj$DT%K#-Ws0IS8n`_>4`Glsp48Zm+n7E=6?u2^6XyEcc{9+{(05> z@aN0V8$sQZ%h+dhC41_|a9(tFB-Mh9+{OP7*r!^>Jhgas^8{_5l1`uQb^D|~yncUR zYacjaKG;9cA6RJfF9-7@OK3lzwx=#Q-=}z$Hr{mG@ZNX!GYr08JSuYL1U^u4zB$t) z|8M&GZ2{MxR<2uSz|8o|KXYoXWdcV&X(2i}Gl%c^qsZD6KC#ucd6P_Y8F~Jlq0cu( zb6(h1vNo_iFx52YjdgtMGTPgPK3PqBiuDB8hZLs04E7)?7QvocC$HW0iw-E&y;YrU_G|yaT>8F&>mYt}Q2LGU zcJC|w51JCrI59H?E5Lo#g!|Q>6RV_sefQH;ws6KI_CGR5%Ar?iYNT0vt|YTHPs6V| zXV{lwt5@(_1$fi_0QSx`;P@Kp2W+rth(3U4{T4CB|7`8aYv~`)4fKsRg22)>`RuJh z+Dh-cd_{l%{^`r?h0Q}}j0ZOM=Q2J!CRR1eFR(r%``EW! z&u<~qgkRWLdcfVQ=H}n4Us3wi7swEde#d6QMDRZ-dW!{_lEiJjz+RqXl7#H^pQ zmkK=>jZ9LVCMWI#kCw}}XZ}jJ|L6<&FnpJ98i3aezRO+~_x*{y&s?bsx7?8M=8Sl} z%=i9Zr@em#?1`CFH=T-7V*mk_*BR@V_Ng@4e{%U2Mvu^Y+yb_(Lu(Pe_QXA-M!#la&GI&k4E1u z$tk=3UTo*s=!L0h{z7VP1Y66ng$j5!GPd;k{QT1ElXtuO0;-+(hj{9Hj6rqeUC}#K z{#yAlC_hE7czl&?=b6~^r{fQPi25noBh0yH>`Q~MMk{87-kh~vzY{CngA9~@a`deO zHy@ufk2R72@g5t$f%n;ivpL9q%NKzW&7W=eq;2Gv$V|ZtFg1Ug4OH81(nEobzIusY_xc27c-xipyHXcHsc>g8kS0^7y zdl$KG{Y&+G5_@cguJe8rglv$-5(|KzFo=xjr+@djWKyLr=#JPRf2?fbU-% zX?}=3;ib@Dw2(}h$GoAP|C@wAT39pCcuKWbE}3cf^f|G_yoR~hdVkf1yzsU4r?j;* z*US{JFNxPQAD}<e=toGmCXL_{13Y^8nl5?oQejT>Ja!(dYgC zb$;&wuJ+Pu->&5E(pXjzpKVLo@4bofIQ-`1jo*JAa+v4A)qdH?%df}J#=h446CCaN z7XN#DzQtQ*CV4h&jE$GuYxb|`y3F0b67k`_#E1JTXjcW^^TB(SOFMlR{Cio;DgDF( z`-v>ThIM@D;cy&##gktS&7$y@{GBH)*Aum^Q}O$?w3{8HZ#{&-UrR}dl|pmdBt2Z z%q9B{F*kiyDz-U`*cRVQkDAn-XW<3pq4w0Lj7pKH&Il*q8Fz zI`G*#=$GoLCy7Pry|NoiuweqV&UXh^*|ZYgxA*Ewrx9CWe}Iu6#=E(^JNFWYUl>E1 z7Z1|9?=tjwA995KNY82Sk$l-9_lXC&dy;HCsN-gEDm;Q8)g!M0C#qM|1?`k4L4C>Q zza594#Bu{$?9`8`Cx z7Be?G;bqaG{Gd~F%MjB74|SziS~}I079t-mHCt!X2l3>3+ER=_c1jg8C;(4}T^SUB zCv{K0>WN?RJaJdGRrc2U=1%AoL2nssV0&*6=XGT_TAbIVFethM`_7=yEq=}%=>&Z%nt z#X9%?^!!8#{@Fdck1zgh=YFF(! zpX6G5dC*dUy;ZA9`9x0iyY;Wr z`SK)fXijRK*y<~4J_$!l3QL+TuC6R;K5jBs?xN$w&1vgW4ro#-Rot?Qj5jn$>;@V&8g{R4F5IR|!jk@{4*U_`UV-;ciUFANqD*Q+~H zH9pbHbsy{Xeca1I7UqOUCr+T(v^H^)u^wWqc1#OyZ_d2L%upNYaKFR2UxW_&E%~4D zv1B*ACitbSEv!d&>}GAVpvp)l4(!t^BrfvZhXPv*L+Edp&6V}Y8rId&se1|rpTddenPsX<2bbiupOxFX6gE<+B+U8 zstRe1IMSqXmmZi}r2ZGXeb>5n=fC&QD4_4_=)2OUGw3@+23vh0oq#>V9Q^BCv-OTEoIHxk!tA2*n4_oCVcrT% z_5D-NIlL(SEI(6tiO1#_d3vaWHa~4;e@RitMC>vAy^MYC&<|P> zdngJbcN7;;%tH9l{&GuSVRNozDT??cas^Z{-jD zzgPPvv8!|rtG>svt-6T+2tIRWw3SQf#SY-BE=LzJusjnwbsO?5iTo-d53B>sJn*I;p{~7*`*7p?{5W_9t|8@ob!D z`t(jVxpXGA3>@{hnj-s5)pg{8u`b-o+7$M3S3Ji)^QS!1^#*HEz)1dT5a++xXy*ty zL7$V~H{iQScl`b}#CTML=afm^cP)O)sq_4nHqUR-y8mxezv%_=f1z4co%DBc8V+ON z@J1Jhtb<#dS+RroC(SLEUM(N7{=%z@lZ?oA>~-VV>(_w;(Ju!5tj<}?+IDGRYnZ&1 zVB{Oeq7Ogyrpa(*PMKuJV(`3!_@@cAuJ>dOxV18gJdRz+-0;p>vVmWE0eE3}g&L(V z33k~#cQl8oC#q*FXp5Ym)Ah+&i6@tb&-#Bq#IrT*i>c!N^E^}fj@^S@C-}^7TPLsd z9crHJAuj^E*|vv_3{xM*en&Q{KA7{}>08A#>9&jk+31{^8EJBC*PO&p7CUmWl;`wb z*+lE(v?^Al>(T}CgUY{c#+F@g@fgZ}S9o)E&gpO@%Z^$1&_UFE7`!Ha(iV=~WPi79 zp+j1*t?iim!3na^`F^RjFVu$9udKv>a6YyDO}O)(d*{+-`@6|bG=J04Vg7eEb6$Of zIFKCe{Z7uUq=9HzM{L5qWKl#j?hu3Cu z2IiOA>oYQIYbr8Z6;r99j{bIJhw|ADGmbUz%3Pb{6k6GXtWciLy*9TgWOJM5y17mI zZOOUpnSvWt?HFRZF?{7};#L*3``fm;k=j@HGG5hKlFahHcm2RLcm9xiV6Aqocvk$Q zXJ^f&zkBAKo!MV`^}hbfYxuubepI80P&2p{83c}!Kard)DN0sn6(!;I*fAbQVAhPMWi`h?RX_K4=77skT_h4 z=2erMS50nSHMw~ev{OTE^^MfZ%go4ZEiUCu@o8pJJ!{>)^y~2Q@azfrn#GIE8Tpt; zSz9~0e8TJ{z>z-Fwc_)7@5K)hS23%;-L;JM5#I0NedNUK=b2j-%!TeJ$p0X>??uMb z7YepcJom_!UHDeXn}dl_p9Y8h=!?fW^8p-S7dM(Vr!QgzeoG`LZ3MA()E4)N3G3sx*}AP>OSCp zC)cO@p!{w6q3;6?)Rf$+PjpN}@`r)MFO@guj^*>f^7{BV zu>XQpOA0OdU2@HoQ+E~ZE zE_k|!_kYO%KFziDBfjhSREd) z>D=qD{2=q}s(t;>J2;5HvA$V26OQz`1l*}MYV}%b6Mu*P{s7wj5S!&$=lN3i`7&^) z=gYyT;@bE*+mo~@Tq^$R$Kx139&ZjLYJGUjXFMJrOBmlrxz-*^;W3X~(k;Z*lMjQ# zSZ1OcTeRx|)+~u_mcp~bahRCjWxQi?xznk$G>vwqaIfnDr~cN?cl!5ddC$1`zJX`L z+)DyWUc)ec^rn4USes67d9Tl}^-F>^a+a-E|5*7-FThvX&2x$dfU`HI}X8m_w^ zarFO^(1KJw*FDwLrsT7u(s^zvdb0Of(|nw>LM)yC@rp#47)$?Z`t`L)Y960(hK5Ox#6SD9AP4}xKshgeq)RxM(F`Pc#XWo5?&lbgK>YVzc z9lBph?v{7|tP|ULRrjB*YW_5zwm;*Y{`?cyZS-9@DMgPak>}s_#&Xsf%a3{HDEIdA ze-rpV#HWqNeWs>Jv92(Fqj)sb1R{ zwmaU6q>^hmlT^Hn?CN-j^+)1D5B`wf-eFxaLM>e6Q$+DP*0^3LuI%v6%tQ!3G{CtZ zT3R+M%+!H2yjk68RxcDG{>+=2XWJJtV0KlSv* z^JclulA%ui!VSk1vv>T>yoQ@EAN;-QPnh)-d?mN6wr-o&sJP%Zc6SGyX_zGr31AKZ|VFI^edd^G@-J;cDzkW~^r(tY;pCSx;IIpUOYfcg1j&6X=gaD4rvHKJ#q< z6SiOUC+gU|jpdJwXx8s~cep-(l{mEGX9e_G@2JlO^jY^5CoyBO51GFoIOgUURB<+w zar8hratNF4mQJ3{pijHV4bXSR+=9e~Q+pg5sqL6M*9&QTv1t5X{nigH_^<4Nn#unF z-u>g!oP0eV*sA*7qNV;k{{2n)06LG=w&SDgU(?RT<{O-;cH+|uCk%~GEBEi*9P}q+ z>?HTgjhD6w4;zrjiaC~o1C9Sh^icG!>4^e-=BlyCBQqlLx33^;=YSv9iCsV2%JOR7 zm7o1bylZ0>vrJ09wrCMhY#n&%YtFi--$TURJK%k-11I1qE1Q#75;tWoBk!C&0(lLW zd@QhaBk$~m7OGpOclSaI-P8A8ojv$=5q~P+MGneMd3#Fh7@?^b1pqv0*xxYr**&jwyzS5B5_0+&a#+inJeuS29jV z9;-32CcUr!x%lsu_!{y>#N%2gwQ~ypuM7SUu`V{3IbTziq4Nly^Ufm}avy)tJ&(Yy z?MX%~C6=$VLC^(@koBof&S4Pyn8@!Zk}<9sXXgTNce`YN(8g@elsmd%jK;;70*uGT zneuq1khy9?tz8qLr?pjQvIis7P<1uW>Wl;W+QmB=jJuRrfaF(aUL-XGTd{}xdufZP zvc*l-ybZ_pl7Fgks-EIgyxTWXXLk2dOEKH7L)&XxM=`thC3hWK&#TsTFVAaja{=S* z125@oo&)#O);xumW#A<*kiGI&;=>2V`VPnuyJpR_^ZMQ`s}MuBWqjy!r3V{avnnD`OmWZ*P6uW5mv7+#STl}h~4|=t_@gvtvwgARQGQN|K9zn&iy~u z{b{VFasL2qecWlU&U8C@XV=_L9>L9~xp$UnzLDQr1WRj!|BttKfv>7M^S{@*wU!O}W-ZjF^?qFn%1`n!T5C!fVp-NO;i9 z4$dL)=|dd*m?7WF&2P9ebNXUpScyxN9xM3_ec!dI&57Z|FnbEQm)x+L?-H{^uI+G+ z`&+ttLkn%Upx?Q+y4Be&I$L{mTR-yJO0njpzYl(Z{?TXOZeBs0qUA3!*N+jA{Q1PS zHlFthZ%)L!KFoa7XYj5}{&mGv@s{E94p`fpY^(tC&bBFLG3%`@?&@PrWu5ptu0zM3 zVD_p!;jYIv?3{}2tayR5NkkjUkkd+_rJrPd-3Z;$yQ_C(`FcZ>_|$e!Zz%h=)4!K^ z3h^qvSNU!GyUF@i;Bk~~vc478obOwC;2QWj_!xnIh%<74ZXUX#C``lPueDQ?#SbM6cpmz1D5`I(s(cEv&~B^!jzyBjqQ|VV9=N zS9zTieVO-f^y_0h`@&{VgfZw`t4V%cQzls5_?F^WTrSWBYfNLu>Bx%Qw4o_A`}3Q~rwIPcZg2`kVS7b6Pbu89>j? zuPR8syvWlr@B9#Bc%=Dh;3}Gt^5|}?)TO&)fZL`Eu+`z88}q^U=?iv!Z=c9sEbz;x zIrZ>Y+(5}cdwuYkIoKa{Mo50HNxIh;4ADXHN9@{7vTG~0NZ(6VD&(6){SC=71P>_V=li_nJy$w?qT%nbvh>Y;07&@#RHU zUs62QG2(hge|sD4soqRoMEt1$x+um@Ql{v%qL2FB9ml!ICOiT;E5@jT)x z_mQ_Y{88|Ux)$);I_Sn&`t3oBHW~aQ0sIsN_(#x@$UzpmvK;@&Jp3cFStjt0B*N_T zFAp7&epc-IM>hFqzs->^&H6_I6*{v-{*lSnKN7J1k)Y=v(X#~pkyD&MmSWtQW3&A0 z%enTv+Fyd}>YtC>hw=SMrnnj0w*1S(Z}Q06}B8x|K698wYO&M34RxR z)}COO9|Zau;RiA1t&%;B2iW&x_cKDDDyyv=-()<0GkDP!|Ix^}e))pap?jL2r%v@# zJiP{ZobCLAE5Pf_eB_FMKSA73@z~*U(#EEEJN7-s^#;1y?|G*J=Uw;vaOu6Tgs_Pq9C`;gjt_(#S4490ghUc>6|IU54$ds( zej_mv(vh`iR?kX#hCVTKHT9{EDxT-_T=$jBLqquGw|6J_<>Ql->b3Z}~c!%4b zdMiAk{yolj>5CzYzlyJ|24CAwd~IF$+N#$afL`3*J`Gz-H!{~}nbX7>&X?-l`c~KO zZ@t&`)(_vCS9apP#F|%mZ&7=KdFetXpTv8}Y9}9~&Q5Z-BR@w=ko$MAcG5nr@N6l% z7vK3cV-((`@g=>@bp!so_4w=784Ty@{YufCP-$p2R7&!PVx*>R@&aqzS@ zpL;tu-gtFhkK(ID3%1$ctj0QiyGHfBj(X45_e0Fd$i9a&`abRYYTNg$v-Z9EeSHsS z^nKd(#lTa2ucq(Dgy?&CRNupmU-tU`CHot0yq{m;zv%mD`}c+5plCfh-H`qJE?_3Q z@Bj1hbM*eb$SxO6JhH=eCeZf;cx83z zLh6IJ6*3>;c~vcC#Eytgo}}Mak3=>Tom3x6B}38o4|1-d#k&K>>LXT;Up&tC_rDfo z^mnY=*Ij;JnX8{>%+b*N*!aI;d$&J7I$I_&f%d-mN$ z%e;93L97hj&2Q$~$-Hirp7b;PZ1`^uoMk*;${5d&&N3dC9)4&jR#@Z7g#9@7l8e4% z0q^W8;z)=s8G!rkgWevg}!yhp@Nl+CRr73;LoucJb{7 zzU>WV+izC@!;RRkUhwR7S>W^wo~`Z$&sLW;#fiQa1aD`9x9HPd;Dc zdF-!qWM4zB?M6P=^>XCeUhrol?;b-27p`)mnR+)e9tZ1^@eSujBWJb~dkz}2H?vOBngDgGjv;ygyfftqtslW%^k5xu7oLTxI$jbk+iP?P=?>(A+FQvy zuciIW^^8rS&2V1DHf()Anmo)EhHkWBIMOPh^%4Bk|hFRW0A~@XUq%(Zl8V zOoSinS!Y?;L<@3bPc6lFvfTt5r-Ku78N>Gax-I*$6HTXF?>784+^9MA&%->;`p7zS z`fz0^)Ymht8`^tprHiwmRa?FDFqfely0KWnM&XSs+vR0kyXF3TzK&ldQ{O41pJC|a zdiuSK>q?Vdkwzy`e-z6Hp3ahA;52fn+o!PYllEz;99~1e!hNIrwSf4$`Ha1WG43?s z#`&zd5@bD%J$RnkQwgm(#<~-Z%75XnE48crVXdN*(zBiM_$@n5X-y*QP8=$uUAs;_ zx-C1l>Z}4jb94<2`=j-m;d-y|>s8sPdS#=igzjMr{FC}9d^=YkHO9>GIkF)^GyU=D z+o7^i{jhlI$rzc}R^DLWRVgtPZ_f*CW=~DD8=m#<7d_i&t82d-_W-;;R22Qo`)&Cq zGsTU~$yPZ$dC0GGj_MyS z$Q$DSOR(dt-VwI8?TE%-Kgq_9F>laV^j+4ubPOMlwXTBPYdy5_{$^f{JvMG-jVIUKz2T^OTX>C zzr+2l92;n{^#BL^oUujFi|ln>#`O;O+F+}+?_UsT(cUn7eamni0q=R~ZEjg9^(j_l zsQh-XoVrJp-{F?DHSrTZmcpG!&WbvEMC4tRf@#v>dpLSdMci7)$`|tWqn#!NLr(V-t4Ua^G&~8osIE@@AAMgio2w+hA9G_<$7*cIjK}>kce%FuES1A|Z}H1_&IJ~a zG!JqwSj>0J+f0*dlkZnK{*wEE$LIMY>7MJW_yKcpfz#aL!Ns>l-UMu$fw-gcEoK+EnSJ?a8+aft0%sPhZy#0SzC-~Mg*RAtg)QPP1vDK{irR0=N zaz7W|7H9tLT-&-s6*2m&`|>-~R=7g4Ndfc-{rJjSbV<=Y`Csif_MK2gh@27jn4!P4UNl{1)CRdT9NV zO^L5j-<|9e92|ow2A_oqCOZ}~7q{?-&oq4->+5#@3f+CgPk%0;eFWITqseUgj4hdK z_I5&}p5U3rB7c~Cj{$x~mjxeRb`7E{DAq)>YY={?Yh>5M;K|<9gqV#dO!4i{!@z}K zyABSxen0m-=5XVc(8p`RztjQNGxS(+Dg|z7Xtb`&fSGbKT+Tao?Oy*etIIAD4$pD= zpA=2d`gg|IxWi(V&u0BoZr6@?&f*O!XK(pBzuY@Bbr$7npSm}hX$3aFV1IF7bN5G0 z|1O7f27!4`DErFi<6P%v*X}#-e6nxmu0j*_({YlvP*FX>`^ zcSBcY6Y6=$MA6rfwdGr=Z0;a_;ud0xMK{;pZ8q=YTubT2<DVW)z;%lKAf!>V+mPuxS@4>{3#wK;)uZ|~p7iCxa* zZSo1FX*a_6ndSB3Qzt%>%qQ;UGvK4vp5WNSK8U@**N4|lR=*tPPjeTA`(A)9Xzgb5 zo%MUZ*lreuYa2{a^3p}&BgHeqR?d=LwmbC2=6y5PZxIg?J}+fIk@!aD_jW$feOG+TP>iM*BZAz3pGUcNo&HDPbmvcbWwGDdxF`8!$`(8Q`~NVw zcXVHB>5K1U)VLLw*+F~IrM9vqvHk*LUgR^=ed&_-_i-|P%$!rf>+gWqKYR0cGUoEn z*U7*8hYKzq;-AQzZHd|BA3jt0^t2Q6u)EBAx^jF?->Ej-1@1Db0!40jieCW?yzPS=!cqx8B)?e-+bkBwOiWDo>bJRq8 zc~^YF9luNSWq%utn3F+#wa38wG5LqKyl%!ke#M8K_AB^y_qagqaWkfN?ilA}F}VQl zUU~N^<;1=6Zvs|cx_RJe?k&7qnT1`1xT;?0;cDP<&)j>X@8Hb8^0Q7$Ngzm=kXd}o zIC8lJ##ZcuryfUkh@=BH{!nv#k>(ZJRp2=N@^h*z1s(e*%2T&F9=Qu5$&{~_L z7CzH+)QR>&ht+Sz%NB!E#q3)ae$Bt!iN1oYP!Bv_fez*4b8+AulEFHuPitx&xjKq# zCMKWEA@+v0MMo4LK7FRszaHMYV7#;HVD4*MF2jyhKP7wXbn>wx3(ZdR{m=PE@1LvA zRT*_EpTx<#%;tCCnM*1>9(cx;6IYARGA?Tyg%?VG{{Fq>kK{e^%$K?T1=nBT&(a_B zLB_DA`C~=o+kM+a7rNz}OjX40|5&d2-gctj&x$&uC?8bWc zW$n-VY4Jb-WyO@~+oiPC5RWEh|E_ab3wx@6(HXkGxxA@n%Ter}--PBLMPAhU3sUcX z>U8)%hwqC`POWt9;E&1C$k=jtt}@;CU;`rOj%?{|z8(H~Y%{&sX7uj!$Z)!T-jm_h znd~EF#_WyrOu0ZybKr~6LCIz1?4z#n_E8JJWl!omFRxW-|7VpcsAVcfOON=kGeW3^h2?)#6H^m8n2OmTDTEdM7*Yz zhut{zAot&oyRm(Fi;Rtn-2Q853h&x^$~pBNmq=c^D4C?bH1F8GHNb-S zxr<*ymoV?<%$fnL)@HeJG-beiodL)LlZk z;slm>uvI=T!Pei`E;*!h6l_a9*#5D6T-$){S@#Zk>q%p`a{O8PUUII!!)Mj^1DSnC zhWg*``yYe*jC1wf%8S9q&6#~iF8SZ?yW};|&4-1n(9J#PkhgDwhGhB&ijduECuTO2{!Sk9Zw#-Ae*D{tch#DJjI$WE+WDbt{bHQ2d{F>QYpU0rth22(t=r(rriUE#EE_s(Ju z!vf+_4ic-6zwssc7)7T>l;dyle2@MfhE0mgAH-JSAQN@M-@AxKw>krBABDfU)C4y--s=%^D_3QK%+HA{Ux9o3HSo~{NXO%Sr3gag%^k~ zXn%BqekP#d)zEB>N4}a)kNy=go_xkrU{Ws?ql0KXTDwE#j7R1E87-{6O>xFkVob&5 z=aKgt-Vu@AmK;ll_T$7#*>_wUo$@hg5BfQCEJg4=i{{0#nWEdw+`!yUV{WUUeGzc3 zhB?+4cXqjRX_!mxS4%U0l3@x=@JI?;Dm)2|C&t)>YJ=m)^h2Xsit%Sk<_|IFAOkiR7J2l+1az`D0eOsMahp6|hj(*fY*!Z(^Mhqu*$JF?3Ul|j=~R_^f& ztFu;JkQ{#s{}uI@(3eaZAdfl7yeIxv@>5F~ov~Fm18;A_G-RLDN1c{?(EE^`nvA#K zVXDd*TRG*`eDgu~yC#3X!@u9izC-pqbW`Vltnljm0s9Ql`!3M8!1*(y>Kx`@K^=0JMdg&FeG6;^*7j#uz;)Z$20cJk)$6CW!( z_fJ`m{I2FqH{~LbT$aKX`~h;?jQ-}|_{N3VFzvXLvlTjN*LeSY{dM=Ke9WWkx{|s| z(M6PlPqxxPj+uD|-lo1RecSERGWw?P>h5sMms9>IdfKtjl*VI=T{))d(=9^Ymn zQ*!6H_G{q)eR%z!(U&eJFE#5(&sb~3|G8&I*?NE* z^axv^v1@IpjJ4*`B5M=O9H-(Nw6+9K$ueE|u{xO($^YteoIXdw#-0iDD{$2BQ;InV zEq3QU!Q5bn*{k;}nM29g*i&l@2L`_rqHg;Qb7f%wPORly$x0E*WbY&vVhAVV)FpbO zZ}ff{^{H>r!jm5n%^VFow8L&|q=P+5= z_~Y}}-8g>@d3fRT^9|%c*E+7rn0w)X;@?k(Lhkt_*O4zOlr`c!rquH;&CSMMc8GPf z{F=CxL3Q1G>cOW5(OWkn%MBtgSMhr@XMzl3YhE#qy+Fsl4~LSu==izl_>|8SUVM=9CGfBn$dAW(S8$TO?;w3XO8+u#ehZMR zvs1;>a7`Z&c7mFN^#EF=$IOR&k}NfmVi%}JN>l3CyTQ}(DgZ6c_Q}V zKw}r*%MKuW_t)4vt-6033yi4qr(T@_KFY`4p!mi zs~7S7L1dcMj-@lA6)%2}vP$?O?-;J@p}E~lIB&#c9qFk=rq%d)FPr^QuYo(9D@1IY zyZ`bKF)_mJUht$|xf?i_r-1hr*OY#cXT;FN$h)mJiH+LZu^d?n*{$O~`oMbZs8B9R zvo{6Kx^&jXjaJDbO=FY!>@hrWLtt|@`?(GoXJ|j^!Cd54#&| zCp|V0?LjW?eMU4k_lQGms&WsNa2{Hg#y7Tz=d8D9z>`N=Pr!`%X~`y^@HWSxDAh zX>4q`&duD%zE8zh0nbh5S=R@3AMh0240H`z{&coA`0%okYj6u|No#N=znN=rv&w6o z{$IFjY-kNWsdCn!*57ZaYXg5N_2C6Kj_y%zHRnD1h+5G-1b*=F0TjSO@!)rHF#+-~V&zCXpI)}dbI=Npn^~Egc zXannNSG^NGicd3>o~yip@>igH$}=Whvovuwym=0|G}rWB<~Y%uEND1Bo7ecW@)v-#qMo=OfE?AYW_0T-KULp4OiGL)uToy0r9mE;MBhcnB_C3Qly)uHs$Fz{zg( zjS}tcL{IMKyZQ*S6m=-hv(x+T629@@l^tstb#EY_zwrMW`rPv)6YVB%VGiZJT1(JF z*4mLrxn`d=e!ieRugAsbyvFlId#Fb`We06^lk2SKM^3cvEu*u%_NX5X^uxth&ZPAE zAs+Jw;WY0;UuUQIuHMpF_#Jgc9Gg3FA~xXuJg!|C{rz@N#X`PW%uKSF+8%sJvd zuE_SbQnV)HpRXUhUMPIIWXLyS^*mzG69vS8Lp$b%zxrM)a})mYCc7`G2KjHT89%EV zxw|1gH7WbePn$-CpKcF_W4#a~x z&c|m7o@?(XeQVQxW$}c)dCUPgp>}0^P#my)(|(_V2AeQ=vA^KV9rXM zypw(8!4ch(F5&N~Lq~5*nD^=JZrLckJ;m7KCU~dgI15C)l zL0opo>JUR^z{JbHVReXllRHawkssj58epLIG;X`j54h`G^k!RL)muLxrbo|@(N`ZY zI*18!ZP3u9i-+;%F!-A5;j8>#!q-Lc$lPs{N671X=Hu%z@RhT(A2{aW>oMTE=vaE50HMRc*Ng>e-X`{@Q~-Fr;zy=vz-gZxQ)4~m+j;q+?cS3KmLB( z$ra?B_i5EHJig}dH@<=#0GaW)7N4O{2gW1cfC77uaKZqz2m3HNmw>bWH$< z3OyPD?OqR@u`$?qUC}M!g8uLaH8tRSB0eg&0oy0E%<{vAab}j}C;4~f$NvbrS||P) z;M24eTJRdMXx8~j;B1<9l#^L)is#N}-S1-kXY-59{pK##|Iy!uPHx9W*ueU4Sh>uu zf31h5TL08z*Z<9KeG|FQtkbRq+F2U5`gogimuQ}rF;7RarQU=eR&(#$X1kgDOL=}U zGSW6XqI{HX79MZsemQjxn(T_*#PsU?0j=@7n178AdqAyhgu9tnm-b{eK87t@Wj^ge z?(^4~-8Xdrm_vJ-&O&?KvQe~W8Zg(o>ZZLN(2gf*OKm^FujKVOdcC{mT>F>gN3FU4 zAO^-2U z=NNdccqHQIS_Egw;i4`3uR}%#UyvR9pvS)1MP3;Dxjk1dl<$@5HksFQcPDC_1qaZfGE+0(|jM#vbfu`YPCY-Q2E7>23! z^RWiHSo<%Ig73+g7td(VxF3$kFs70bTCk_xV@MjUf?Ci+C>E8EgHBU?utDa>0pq*PFnR9_l&@jC(`(HTP6) z*wVv0hwS*Xu&LxUCOw#H{u97_9(B&APQ}7hns8qSDw#NMFYMfvn}2L zZ9K4Y`YDw+ux8fNU&V2BvJPF|8EQPAyp1XPl4fijqPyU-_QFa&7ajN&F+i=dvut#8 zDE-^@mM&~ED_y(F&$-u_v> z{?jj5DYi%SK{%wft~Gufn!ir`z@ruHw`+VJ|Jo}Po%c$`4^k#6Tfp-ZW9n>;ldb&L~@lk%E$##uDifu#JTH|i4hiJ6Z{LqEK<)grbxy9c@f52<-Ks@l@ zsJ{n(DO$N3*)op)l)b`PmBYLogLW%!am-y#OYS3M?Eb=|Rg3x-F(26xr~hr%-I#$y z`@t1=Uq9xtc>6oFbyNQ9TP_V`KYkE?36J~Ox!P4ufo$eKYarf!a7xxzi>sP5XqwiB z`cou4K1aW0gFn}^7kFlAEH*LO_M!coOfL2^VyxrLapt^Qy(oibjmwzhOq#WUIy2{6 zd#*F>{gQp8)5iZiusP10T771P3!@x-m9Lr{&iM_n4uoIsyT}@@uQ%k$h&kX2c-D8D zi)YH;?BmyQ&S}nUFE7`5ZyWSde$Qpo+;*PNb@xSy7Ud|$YutS8HQb~)t((VK|6$WT zX2cfs)+qCa1ArU~;bFk;}lN`581^b-n@}-%q@$==d_| zxbWyT;_B3PHhC^CC3ddCWY;bz|8)EF@X+ELKG`13-@kVpL>f z-Zi7cwofcgZV!9LcAfFshX)iRLsvmJO2Ient*C-_Sp4wdo_XD+^-Ani7xXHMovc-R zTA`ux7mFXQW4*x}_r$2LGJ`%9L7yTk3r{(eOOJPGBMdxTeiCl%`OwI@&TP}}wTC{v zrdTUzmFUkwuG8=U%LzKy!6qsiRq?&A>e@8sPAaFcTUtnMBN=~E%>!vD z1Dy51+G;Yng*JZsvDx%r_vq85*zMMHf3CG=&KviAz#is*b0Jkwl-z%kjobViKipsXhX$!_TXU%7DROx=i$ z;roi}W1QpEj(AeJo9pU+l|!eq-SQJ=#uU5f^d6NnFHL3kefs6jdAZ7&vqrZZ-saZx zdCGO(i-kY*G=L6Mz`A++GGr&{x_tQOLBn((XDs zYcBTOZHj3!=#lX62s|>VHn^5PCz(}yP?XOmXU4i z=Dj>*_cZ-GvlPsJ>xNhvXjM)XI1|p^Gp1-+E*D8Q->3mI*6g z#_@Tfc+kzC8kTkZb@^fH%#2xD!rCf>w`|WfvxD%UCD3o_+}c|cMrKHm2cmpq-0thu zeEWaWHFri{-BSEStX6f%4#hmoip}VJ7? z@Q^vQozFfG{f6Knh47Gk>^3oY$g~v+%R{zD_HQ8u<#Cl)gL5%>NM`$m9uJvD`_=G} zLa+U`S|9Lo;jQW~xSD)*;6UVRKW4?(H{F=pJBZgJ9v7S#n_NpgZfurI2Q%e|$LxN+ zVf*T2Y@7-5B6NaNx-KU#LKk#Iwrs5f@svF9Rk;V3lA~xBcHw3Cvkt*;F5_2gRd&=K zscopZH(XvR-0CpVKcqe$+1wB#j7>gsG{Yp6T8o9)! z?`a(VKVLqPO&}9DI^kgfWqx}W_2}EMmj_PQ%0W_QPJLI`WyGI}Mt>&D!)q^g z&f>N9!dw8ap#OiX+~_&1lL=LnIsI=k7Qg>){<>lPelqjy5PP6}6G37$LgVk9ij6O? zoLG%&%F8E}P0bqr#i=#dFP)lI6-;K=2a`Fi!DMbKi0wO+JTDSTj&(xGaq&>{{Q5xh z*UaNTD)$Nf{TJ`|*WT|32Hf&%yx&iGzhCrzpX4`izv*unIC?k>SQ6vAX=!KiKviVv z)Sb`=d;fCrKz*cq>U8Mbdh9h$Br$cTQ?q3qea+t(&x>&GvZ*lxA#h2#=XAYVb|IZ* z%Upbudi-+boY8ywRs6BVM{rZWOTkO&!m?vLimmE2a5`O;n_Qja*!U!DT1o6qeOKYb zkc~w9Dl5sYluy4(j5(?Nn38}0z+A~zsChfhwc0}uJ#va`@tZ(>$l^-}Jd&IPhd*aH ze}(>*E~!zR+yfs$ezfnUIA@i0BDh{e9AH4_Gq4tPuRqB%#5^8W%y(*n87QNUV#htx z*`5KxdYOaWF7YFDTfXgt*GSe|TjLC*C!2v4CD3s3w`1;p){W5H)C6bXgFan-!R0~g zRLo2?%HbWv{JS#TJu08< z44k4rI$x=bKf6Awhjjp>Ip%qp5mJH!zjX{f=x2R}(W~vX7*+Hu%#yR7LzoyjXsnD;hUtQ?;(_89qD@ zu+}4&$BET)-&4FOIUoAsO&hjook-RfP1UvaQBIKU&Gh{Wd^db!>pD<8pf(KOTG@to zm4AyYD@WFW$JHimaei+71t-IgRK2wp+8E*6Jl2iYNu;f6i`oy;AFXG{EPKvjZ5Vz9 zk2u#BCt|}mq4~eVoBzi=dUvwgv$xq8;^Fg1Q+JsD^b&BW#E_@YU2pMFPJj9i)*$;Q zJC~sQutrPhk95b*5^NCU8L6UA(@aQbH6LEfwfbiB^jr?z6kK@EKu4#6KkIqtQsDmz zu+y2as!RL0W7Jd1w}tVLjn6IPx`=VQa9aLcfVzUrLGktRsfAU+~NZXi`?1|ob{*h?VgKG zjce$Ip`x#RCOV*AEup5x-?JyoDF>&Reqz&ktM$8_v;&8{$G`| zmgN_ao{?q^N06l+_TOuB-@AtQuAoifbqamV&@b&RwYbQfvOlgre+Ko?|3S{&u(Z{D z0zH_0Lu!M*E_LhNgqM=?%H~T{elfa_TTUF7t3yrj%F~R4J{|d*|GeIPK32~!a|ZsN z{{1U7W*}YL(W8Z}CN!HaF9wo#Y{t*055t|#qmZ7~{ z9oUkKk#V={mz>b}U|Y04q;J{&9Wd6GcCKr?ZaDAhO8m=KrluWY>EBd-SNY)9FOKzJ z`tZ|lEjZTn-m=?&`QD```()!evTXADw>Erm&wH2t;z#c-OPzU-a{}IYp8SWq)?{ym z4!@y2;0J78*m(Q;#Ubgf6}LiDsXJLtOp;_0oi|^HuC$#ovVY%}l_RfI;|nM!zN(J# z28gYK|M$z*Qui$Lq&N*|bVpuQi{!v_jZ1dalEuSz4AIk&Hy`q0`tZMLvgF@keraLA zSrVO>u-Ah1BVBR7F?)s2FaHevjrp|efHl&=I>QgPDG?uUX;foxw*JB8%b)^)=A2%P5e%wr+G;{^K;L<17|C#>A#`U{)< z-0F7&+fI0m%gAv+kSRmDKL1{Yu|sqh(6p4zcR>$ zxe<6cdRQ2qIM0Of*}%hSXAb?kbRzlB;XhXHhS!)#sg=py{qI?g-}@GGC-|`@GU2Ch zbw+C;XH+N-!hgT%kG_xXgZJy1XZVgQ#}cp0{M&de*-Pv^o*~9tet6{5G&EE)uKWVp zFY{%^n?Vy?BE@JCzV5)tAz;S3_pz7GcUXO$c+A}=tJuuY$ z4lllX9s2@fz+c~2duQQn1HL+wwG2EcAWz*o=J^*8vW+9>t(vTATh_dB@Pb?63H zK0V--cT(=_w{|Ugu-E*`US;(g=A|0=+t?h*hrrhX)*a9yJ14B8E-N>NtgiJ!09gUL zwAzDtxd(Ik?;C(u3Yf1QwO?>8FwcjFuJvFZ2j=9;voJ3O=6TRR!ThZI1wFd%!&UU{ z2>IT9e$#Xgc)I@Z;r%Of(ec)@CMMJGKqO@GM0Oo;aaN&+Bk~Uvdia5DWhnmM)VuF% zZv*|Dm1c~7j+5zpWAFJcGtWsL<=6~Q?t?a}P2&vLv6edg++Y{+UckbK+=O16qJ_i} zoXls<+WyA>^!|Uut40wuv%Xc*zKi%`ThVe6>7_OAo zDY|SXZrMCM;Fjo>VqD%m4Q`=FT0gMXIP2Bg4#8WkRpHHRz+*o2uYug(yVw(P5V5?Q@1rOKY@ak-J7m=ligI zVxMBLKAa4-hLhn`*!EBIyxN_QY(efBt3ONruR|wJ5OcB?xmS9%Y%oF3R;O>Ot-gny zd`8_CJujipV_4@g-Y>^yw-lS*Qfzh&*zCHo+3h6Pb`3VWnsAuh7xPYbhwg3OIiqGv zoeA|FW?VWqx`=NjuPx$Uv@dWgc>}nQ@Sgm&hI{+VBJt&`F;>1E&bK0aC-d`+@Fd3j zdDQ->7lrIy7T>V>R zcJ93U6mniSc;nk>-sC(Ot50;fz7yq-NI@TF^IIiduq|SJMSUuVH(tx{r;zvYZxS9kfP%E7$?evA0_$9$KP9qNpmEBYU>4{Kd%EeNk}_3&!D zN27#KtkGEka92D{wm7YCYYU=G`eMGBXV1n6VDl;B{ONq;@*uo4gwE;XU6?sqXM)NT zTU)dyur(jtDFb)Pz@6>jP6xPCv}WJdwG)eKwN|w!Cx~1Viom0(D~r0anVY%HLlN^R z9=(>fxASMv&jaJXFjcmd!1yxzYtUP*AB#2v=>D$VWz}=7eCOkqXt0l8SBzTUalz{k0;zddVD{PElzsTHS~3; zf8Mhzd>q^8;@C|dj-|oT1zdlV=MI0!X(!BGk!1Q}r#aWdCyk|rv5fW?WX5b-TGbjN z-|W%Du`}^@!SqUK%eoL}i=*Flg~C@R*2JH#Hlaq*eLpt$H0!L;%J0EG<>wYJn*+VC zB=<-uaF%Q#-9gW~Y@SzdAEz(tfX5c@-o3vbfAv|r9^1^-?)kXqYrKrrFo%z`b}UZJ z^YVXsu+Xz=p4|f+2Kc^SzTELwCyTJjy!KIdZ3)l(e?Gn)A9Wt-Xjm}*H(zBf{ojDW zG<27&!8_F??SGxVZ8fBjrTEcygn_7GXIhz51b93&ocfX zxDW#u4%l&ruAI)?6B}md-j`#P<9D8y<5zoAH1|vF+!N2o+|TpoUeA_}8h1K|Jsezr zaMZY;&KTSHQR99>&tg?A=ka@z?`lSk`PPgvW8-cS-k&DNww0gpGrgKIcHw{@zwFA- z!~9G%IEXDn=N&;aH+EK+HfufTELbbkaE8~{|6tFRY16rB$QLvH@(I7Ud&QeF^OQ&j z5gje>H=DPGkR{^-p6y}xh22pO)> zli^~2M27SGHl999=lpkK(29^j^Xq4k@7uKnl!X?ekIEMB$R1Wg~b)2g` zmrizeM{+~E3&)3dPoI>vd+rSPoA|fEjfvRZ;rR}BE%hT#KdR`5%D(IMyVd3dxG)(( zkB>NKd6%+pWWMY0uEV?Y&hjp0+cMvc^X@!<&xflUwn~P(oS1X+?iXJ(QqFphJn5nl zn?}L|IASFKagX@y}kdr5m>e~^oZ~sQ$eEfRd_Geh%_w^smdma6ej2XD!iS8qQCg91? zp1im*kA4j@_rf`yU#V}TFQsB8dDm*_E8nV2e^>vbZx452=<||YUY~w>4tPCg^Ro}f zQ>5nF`UiZNeMNnoD?As>0)3_Iw@SsF0pVHc(c%Hw>6Wq;sR_W{idZG2qiAxHim;zZ{Dmt4zd}ddl~A^eEtJHzl`sBw}5z;5`*sm{KDQ$?7+(B@KvG%n!bC0)!dni z??C4^=17EV7r$KH$D`YArNFP_I%mK!MR|hh$5}THura`ecc$u@$t{~o{O^GLh88}p zCTZv8yUYW99Gi_^t+BJlS>Lnm9C>R#=sfq6MTMh!0ijTPK@OF z_H3CrFw%I1SE;dPK)mJx54Tcdoq>DLfnQf(E6R*-$;AJo-(cREAG-#=?5;tzu?kpP zm{gC%<=@P6*Z%R0_4OF>-sf8Hhp0c5H;lV$iG5K0rJ5^ZImGKV4-p6WAsc7nbQW<& z$oB&&$@$drA!u9?@Q(ohR&uYVDf>7y(&8(5*o3d8E)8uGp9^N;AIjaoMZB_uxChmp zXfGbf&yD4YCltV+u~!ZhU7ncgKa-8C*qGWpa_C7nh8{O1Di`##zkQFs(Q|kX_@w=m zrZGUjw`140@6o1o>;vGJ#-cI3k8ah>v~=rNR|ht4UIf02mNiA}*+<7H^XFG{`>Y4+ zUEaL<^z0@Ze>IYx8TLgPGc|b?GK=PNE&l)f`b(1!ZRVVD@7c3F^RJWGAF77!QcZqt zmH_fZzB4CTkdd2Zs0?{Zupv*BwYB+m{R{QJqdt1=e3kni?tS{V%p1pOeaaut7~&i= z_jfDyCXJ6&d$rl;934askX(_P4eaMYdst7Mq2g#4ynilnWwM=!Rt7JwdaHn#3G{c) z^bT2@#z%<-s6vJu91OC(E5u)z@OR>9s-`Cu-_r?xcOqYEzlD4YDd<`nx+5QK3ObqI zVer42{(mLExz=~w{E@waJWv0gy@K@1T|3?y=6mrBtE+i_(R>dkX?UjZ7yX_$&xqRg zpAVnM?D`#+w;jGI5}wk|Jmz{nG*dCjyNK(oHrYqM3JuY|)%DyrFkkCO<(3z2r5Vq$ zbF{acJ%O3=`l7?)86Am0^!C3(ZXBaa-;O%YHR{JyG3>tWXQIJCqpV` zao-r^c;JnmVSR>0)Dsu(-|6b4rFXIxpxx%o&F#DHu=b%QmQ zwk`h9=c9(0`%5Ud?bN&NM0s~6WqPObp{Y^nZsXlCoG_>U%$^TCgLkf?>^%PdmcPsS ztD^ls(Wl-dpZa9zrNtdjpQv5v#S!>8#D+u7UsCueUMc)ZI8oNKJR!$tb_VpxYI8Q02LiB4X^s9mLZr0yq?77Rn=E^qO&+?fU;S;yH zd{ujJ&|6L_hd?4+y`_t_Q^&oHjWsn}6ay@~5i;`P>BKoS5c9c|^O3qk75GNjL$vcV z>_-jv6|yfRi}qg0bx!UlCvo?6#A>7h$(|Vd2G~<^nl+s}V^#Yxe9+_!BDb653-X-b z;q+e;ur{B)YD@k4JJAQNU9Dxx5CtLik4eW8iJG z6o=^h!7LBdJ^hxi)c1w?ZQe$kK5Tq=9c*-QXAUqd17|vcmH4Fm^(sq%hk~6QH`iyI zuN3noc<4{^;Ww{7+kE9Uw&&A6^WDXK`{&PgG2e>SiyQV)@vYkWB;Swj=MBuOXuMs2 z-rhlUL zeD^c=JNDR2)Vpip$)T5c=jDudUUJ_-#+^8~%ByE3@2t*vXQlhjd3whLPU1h~oD}CN z<=q(S<0&`4vid1IMr_)?Eqb>UKacz~*fdzv`6GN!me=fYZNX1oIdTpJ`%e@P)6>!N z*6Viv+Hib~_L<1WS!aUpV}8ykrn0>(^ z_8fGPf17;=N7TQXNp9YE`O_HZo-*E?PN?(5Fp_-^{nU zc3y$kG~n0TKs;(CyyR}yf}SsjpAAB1wu3{5X2xpSuMT`;N0xj01M5sqUp4Dw9eqB) z+-Qzu*V+MH+K$Yk@&?+}Tz0}EG*9|<}e%IssD z$CVezt|+2E$L2Vz^7iFCu$?t`Y_5BK?6SHoT5HPxqP2FMcjoer$~TgKZ}84}eJSRo z8yTX4;$4lWGZdDcdegAaEz~&Q^Sj8eb{X$w`d#K{_}n5qTf6v^?UFO? zdiHQe_3BSBj(PZB?&1Ic7ynKJus9pMt$bTJ(&y*5EB`<8fTeBc8k^?vTw{BcHl_2~ zHpsj2OJYV;x4#~QldBz%KCl+z8M23P(x(r1TmRRvKC1j2bD^^Zve$UJqH=g7IGaQD zD|Vxty=+VIlP+81;z#Eav#P7atmVeI`tQYdkn8W@SBa*Z>?J_2AG2 z9J+x6`8Ha5fkSWz4xz?Z&jN$}$W?+tJ}~G22FS`S-N2v+nOAy^_J1fhk6~??nc1}s zz=m_v8*6|K{A>s|iiIeRXx>a;F>5altt`!;CDP{w!xCdo7Vlj53w~`}P%31sEE5?8 z+na`98*W^1HrRSJuXFDE`A?&VX932JJeLeCGA#kh zy_EGBU>vO>;&3wX^|-Z{4b#t{oiEO01(rczS?tXZzRSbfb2*RsneWXHvHGhzp<|p2 z)zS?e>wu2MPF317Tsyh=fX>$A7S^E{B=t4 zEA~3r_@7*B>~?K}pIV!nGhp{RF!R@=d%pYd_*i?ZMEm9YtTVY*ZU;wO*fU%aXRbqG zvnm)S?<4QAUWhRibTEzF*m;{-_2yrAGD(#y$DBk) zN!{kGn$543wV6lBKYc1kw!4oD+5ijz!{ADe|h=TfP9enk@itX!S$t6ok$tJBC}X!5pY|$n8seO_tMx#Sf5ctAH>77-dbrl zMY*ou_wK7B-(_jqe?$0d?Y|@N*YcACjMHCVWx9@=B(!Q)2pPuWAvC3oHCKQgdmcWa zx)Y(3YngxfZ_AL&^6=AU!4Ese;)4MPC11&AUxMGaobTq5Cy4jt!=Lzb-pN}N*ezRO zgfhLaca;+(k9rG$L9EIoOIJolTbY5doF^&KsuasNbN}tG6vC3>Hd#H*E*ps z1J=V(GwjgNsf-C(S@cYRQ+l7AFjqWA;S9Bcf6#k*{OBd7kcAq73X%=v7 z7$qhzRHZVrwO^+Ns0}n*zSv(T1cp{y4zTckk$#0i7huRaJr+wQd-wB3GsejkRaP6bi zU&CCM17FEe*w2q_TXES-Md7NJeb8;e1~~v;$20LucwQzRT78pvwm5Kv)}Anv#w7Fc zm46ys)Lf@|_Sp;=5BVm5^=akj_VJRre?9}|KD@__lK+as>H=qw)uyL>60JN zFj@A@7-x+l8-%n!DLz(vi0t|E-JCUyJtrx@CF^PEd7Tq}pcq=Veoc`*dtGyx^I6V4 z!*7)XE**^oPs&b1F5;GkDJ$E%iFdBSH?bgG)ZPny#cs_xeA!nXWu5N_@3+MR$^D!$ z8W|U?oiWpF)}FVsy(`&tTf_{#JDqn^zjt$3s}BMCpHKghwKgJmw?wE%<@zh7Eactm z8av~)#o+54VuoZR?vNfzY|&niY-dGt5R7lT~-9+Ws0Bo7&x$Mm3jedM%ji&=; z;;!0LAFvNs1idzVU1tgP6FW&=L0f+g_0O?&2P;%Jw&Dv6aqK3GlazShz~$ zuIaPzbWm$j&VJCqsrNEl>T zn@50k*g?OF>u<1e%(8W;pNYDK#4m)781eqH@;Z_b9KDFYc6^2lalHLo>ePzTadPxqi5NRF2R56!pjTzRAI- z`qk!g%1bEAT&tPa(5_A9>M1rKQ~q@3O>$N0x@10ksy@uUKVR^&mX6z9zt|CSj9dCO z0sZiHC;IeR=5LJUufzK5uhFw)JKcwFtvZ9O1wWRtWTp91n)T|>G5oeg^kpKrCz>X| ztJW3g(ACPHtabRtb?`Xgme1Pk08T0sUlpxZJmG#lM@P^z(Ja9-k9B{KWDM4M7wa5b z@e%Dm?Iy0WAreeBL`-cZYd%4qL9Hu%?VIH9KSU1Gw=Nk8L)RY38Ayr^5e#>Zv$Pv4eG|^|%k1$}g#Qm(uQk*UpQyqu7PfZS19u6y@Sa z_~H7YHx_3MvVzt~fVmwVcNtq#v{f-jhG*tX>}hjm?Gck}YN4Nf`>@lek=-VvKS(AS z{jK^iQMv_v-+mtJaxiE1ezot(L>GE^em4Ae^}%8P5b{^FBWz~P!ERXqPGZB_h+U$E zJd|%@Q>rLs4AmZO@@ehM(59HBO<6;<7P}I(m-nHyu}oSE{df7aXSb?0)@P*o8j|n+ zNLiI+l$CDW`DGrCX6n5^nd7Z#WaUG^z~$EJv;EFPWWHvw@tJ-aB7zPwE2o5 zb^G(C`lDE0OY5p#xi?N-IzOV5Tt+6FxFvG0sn5ma7>kkbsgiw=5pX$hC$@#FoR&%x zsED)oshl%w(R(Wl*W{U*Xyu1%KnjIj9Z6 zz&|ezyG4H&eW`IiF_b@CzPJ(P_#Ite)sF8g z*b08(D*&Iu6&bz)doTRLS73R$?<=SSrjjW_z^#kiZJmq-xVChB+lg8lLtB#+`=os2 z#Bo|$^-v8w+074RX-g{4o>Sl9xOJYF}$<#oGN4(+t|LC$v5dU@A_?c(3C@6Rl>yjK$f6Nw#VJ;Hd%RUHE59$AiOjyG=R1z~k}YR|;Ap9G(Ua zkACkN-m`Ocz^6gPO^h`ypFIq0`CdA702*j{1Z{{%RDo~Bb{z#P*0PQ=zi$=)-00e> zj0b<81~-96k~Ow_^C0-2JP@?!X&qn5d0Gd;BhJ&RW**9!hgX4rIrC7%Z^F()sBhm_ zuy1f(1Q9^4vwNt^p#h>ROM#u{h`a#y%jvA27Jaz+3%#pTL3d;YJm#X*<4o<`pSAMw4k$ll@7@%hn67CPO1 z=yK4F8sbr0y-K$EaN}FQ=G*&5?(=4Cw+UAo;05Bh8%#H`E?IUjhP}6OE?Ko3yx2j( zLHf1EwG4V#$C&mp57q3a1wUsMvY)mZyQgT2uG69UjBR9pm(e%1>*ru66yl)P0 zSZem{n-E?N8qQh3k9CU;7`7yO-RxC(H*Pza=o% z`aZ@_b=o(D%*-6l`yS-%u~x>YF{fxFhxzWM9r1e2y~;be?|M*ie6oEw&7pbFn$XzB zpL0fvU|{VB8F1*iom|Jv-SO`_QMIpgr)~}eYrh{f6AuM)tiAuyyFYThi3=5roaAkX7<1E?yj)eBbY2B7ymNG*E%X^q4u2aUu@;zrsd=;Jk2`l#onXeR&3_y zac=yT=&-%!yxKf+Rua4o= z6HfcD@KOHSj6L#e{6}?x^D8dD#w>n!j%lfAHLEJ){H-vnEc{sim%*RbRW~HdiF<}FTiP*H2JKLpzmF`B z`O}~3bTgHab{*$iL*-kw&!Q-+(M_h zU{fH99^aTn0E_!7y0k$%JG4dWtX_|JE7E=Wbu?oo>n|eCv|k*f^CiOB*)Uv zS@Sser3Sr2w(ohi@1a`lEqs}MrGEWORX=m1eG;9wU@wQRWx}ISG9hu`ndjjSmDxIn z&%@Q2G^d&?c%z-GR%qTd@F)-2R63YHXF=kn{<}H*!!>5J_MQ0O6cAVSzHe9)ne(jp z%?RIUUYL(f#x`X*xV)~J-n7+K|434c5Zw>yNegL|HvMimuci}@Z0Y| zAAVo^qwwra+8>Q)KCBPUaN{rotz)Ebv<<~oGUp@WDs>NRGvg`)^`pK)J{<9l?niy& zxBD*b3La`x@tz^?yd%}ovTi@_GqkMm&?DiV;8gg&JzZ4Hza=qzy z;1fq%2cJ1V6Yi~ZEZkGPlLGdoz&;a|gl{l! zRnhmYqu(hvH}&P+s&AkA=T_!=q0@in7F)i_ zm*1OI4xRg*Th2Omx3TxC#_QJ`{`>RX_rI$0xx@`p-bo+6!e71m`+~b4@J9N35c<7b{0u&J6uwmO zI(*6FrM=|6uYixPhL2Woj%u&D%JR|We;Aype0=KTZsNDk#yfkk3uW?8$wob2vuAuJ zd*J1=rHGH_ne4uG`vw>HtVV|Fj72$9g1;Plc0+d8ht7HoX##eaSIy(9R)ddG9lkO}rBu+#oK&(fiMe@mQipDmV2#TL zCI8oH*74|b6B6t*(HTq1t0?_P@A`6qVsCVY>x}b9#%!v8&(J@?QJwPpquj5V{&{CC zcrY1To5pc5vhs20CGz$(evxnSIh+1fTr=mhWJ^k8hmh?*{Fa`(_y0%lVjq|c1H9)s zY#m&uc}M#C=joU3!*!l*!}F`FfA`V9U;GdI_xodR|IYC1_wO&x(!XE6Io!WbX7sOk zRR3mq{re^5R?g1o-@>!>Z?}ct@O%~nw-~k(U)Gn#_t!DO*flDgVXp9Z z%uZfM8xieq;l1<(&Yt=Ja-Z(&-TP;_*ZY~@9Q6C;*@WO_ZD!^T$+hs_+Nr}fwJG%V zryeXb=lgDJUmiZcLoz~g5wKZ^-5>suW{!jKnhx}^TFPzxoIRrYUu13Qy|iard4_w( z+6&#i2s`u~95{ZHS3YiLvJ{`k*Htdrp7K`p_lA$zeWUvRDZg#_?HuqWeVL(t_E^UP z(LwlY5Z=3la>1YjK9=4=Ir4;W`{|q`|2q2~DeDBUbnW)hVedz%F$90b|JoPC@2yx9 zjZtSKQYO2=Z2#F^?z201X4}6En8u2t(EMn`n8ywE zvc6_s<-UKy{MOCIdkgO`@amf8Kbths?$?{gv$>SX<~q8rDtKNBy@vYV61|XYpTeG> zLWce`x1KgL>Cf#M9)C_fAE9gUmZ4)Zur&Cs+ui=-CFcX z&BuLw+rc=rXGHT=2~SN+#->m5UC18Q0W3S1+c5nohOg>cWr~%s^T?bF#)|n<9d#aG z);=iv931EX2gFYkeBa5n{MuE5`IL`X+Wh*5hiS9u%BqjKbb0Ps*27Kq{K{efDfD)e ze8{=PoH_WTt&JNyB=XcIKd+0%Dw>P#B&0Ta~+H0n%Dv^f$Qz&Hw#*&N<1+ zE!gq*|NUQoUaxS@J?A;k@_oMh^L@V04%xWK>N~YYTFZXAR&0LNX?|n3=(qM6$kvt5 z8$W~bF;2(FxYY46Du3rP>L|dkxXke}>fY^(>N9-Ii7RoRv!1>>viHL#&rkA7n>WF; zjyAIH-pIPU;84r@y7r-DuwTuE3%-rMgLQQ+^-hkKAh`&L&Ft)gzsW{->nCo#tqY#U zn(yjX{FsW*XPt#SZo$6S&=<;6_8uSbe2%v-T~}&9X0T6te}2h)Xr$<^0qYIm@g?(i zx^T^eHvxa)+i#~hdH}L?o`ZA3W&Oe5j^?MW>#A<=IExsQ%H}g4`OF9Y{UgdDVOiy- zPX=DVl=U=&U4bXs>vq(&EF}i)U()JVUfT-t1^D<=-W?ad zx>l~sG;%7&y#;>wiAPo(q^mDvBg@{oGI#JiYUMNdiR>Td4c?V%VlQFXvUa`VJ|niv zmZO%vjtZOvqZb_*nZ5SbkUsKiTKuHeS&U3V$Hm7vc;44O`k{DA^z)RTHmZ=hb2$fW zpYRD>6zx4gU&On$W>{Kh6~CZ9aj$+#-j+~4j=U1>YCTSC3F2Lf-4fjpFaMIt7t*%M z>3Ie1Tssy~L2lmKx^vpY>|9x`!dd>#U z8fzmqM{{lTNC$uJ_u$XJI`}gPf0JKm=zNqoa*x=eoys#HJD`U6z$ALHY-qh(qcNaE z7xB*cIFXko=Rg^>$#~?iHS^P8uiaK#nZ#rTI!}=^dfdfln>cH(lYB_c8}X@|ee>{e ztyfOty9vaetN;ev7)Q1*qw{niyR(DuIxaLdX;1Q8E4y#zj#Vus*sU?*C2u>QIW zYiC@FuOndE#s}IHK3u8af(>{C&870dBLgE(U(9+B?Vn)IWAG$!!r61;Z=0>SD#oCB zd@yYu6{DosN#XkctusP186*540i7uJPiN(V_JjZ2`F)l69Pj=t;lH1`+XzkNI`~-N z%%$|qO~k=!4}feecws8Q81y^|9*7R`@k>|VY4VQH?&JJcAB5Khls7yZJTN%m=#=`t z0{#~U-oo|$;BYDPoW~l8x17fPCt^2T{MA^$>5TQKml@wF6R$KB&*7urmEzs9HE5T% zc4-docIWC^JLi?2J>q$$`NM}t4rj*BcdRk(nx?T)USoPNy)1C?^IYqKii5VG8AGSY z;C;wX`0^QYqDBjm4cK4CXFmAM_zoGi=^UA&T3_eS*hqzCh z_J;Gu`yCj5-vdJzKUNw17!JRBXjAYL9%$|kF!zGxT=qpPXT4xINALhPdB9BiRsryn z92nL~RdM)8Y5mIb-^{GbVRDuhKtnJ|92*US?T4 zy>(Q`UuO|F5Ln#)u`$E(&F|FvrUzCcs`sdWS6cmNth<8Idi;oglUDy}@|}xzFIRsy z=Z1LWbQ6qD4keet&U2ii|BQ|Pt_{}9dDFtz1~&0q4ObMwMy@5Wz6T%(Ea zmyRa>6M=rhmajR6EXjqQZlS#x{Ha3pj|`CCNLS&X`~`wPc(B@f_=1He`6ds(UU%GFiPvUu%z;W?q@XywlhWrEa_ed^;iG`F8K)L@DoR#u2XJ$z* zF#cij@2Y|_%-j7=tZ&{ zLv1S*7m1wy2JjYbnEp62TI+d|zc&y&C46`1^WW&B_*Hs7a<{*yjD9^NKlH8nUU&iz z`3mz0586?gBqwOd+Qt3RrSJ}2E1x9SBC;j@n|x^33oGpWqY-$P#*42cRS!*>?;i`N z_5pX@S3P5Rq+Ku1-Dd^J8MiI_8_qC;4!zIxt@}>bsweDkp4k*IXWwb;>*bev#Xb+< z(DgQqG>2aVK0o2T3lDq$hiCqhLwVy2?EsTmzyvM%Ze!iOK-(3bTEu&l#9XwKjK5F<{ZrfFoRp4$M{%pq0{1s+n%ej1t@5HBC zGSRWX$*}x6=;-2wCJzYrxt8zjX>>*Or#s8LB3i$n>cT4j;(#r4%WV0ex%e-}{;Rb9 zPE3o}9s>^>qr5ub#h=Wz=3BOZ;5qPEvE1?7vHgMp&YX+zSNg&L{a0?Zx!Ax-Y~aY= zh8D%LhoP%dWI=x2)aXsSfhXnivfu$9WWCC%SGkV5;hWvSM9*qzs}_9KedSi>2N=kI zv3W;h3uEo8W2~arZ$N{pM`yHj1CJic>6;3^(RDt3*L4B)={j$#Wv(^fvahw@wNe=P zJLA+9p^p*zcpZIIo#I3B+wHMe^ZgjUucAK1G|r_j=z4SXOm)d`lAuret+sufgKcod zfw`#*JLn2x{o-85aIF)aJp5N?kHq=whIonV%V=Yc#7|_qyyVOk_-qM3ea@OiiL+)Q zpN!Tl-T-H+h^vwhrj&hq8yK6PHH*mPCCBp@RGBpkmGuGJ0#9sk1i$B8Y~}}i zjjlVk@p|AM`#@B_fO6)b5!tMKM~@#Ecqvq6m+cuJPOZnT_5Bm)ywSFF15-9{YIK3h zEgD`fp1Bwww5J^N`e~K>hxBqw-c=4m^QF*|_bzvj%02tu<-#iW)%Pe@F*SOp%Kh_u zlw%yXP%aKUrd4p>XyGlznhW{JgY5q*B6h9_-(~1GPaV$yPwN(pC1y{tV*azr`H+`q zMGc&!9b_lOV^ z9D4vhF91?=~*v)1YUJ!A?xF-z&Ur`BzyibVZGQ>rZVIS zq257%?*q`qY;0ESuOz(4rZ0cn^geH?K7y({qdu7JOe$H%(@c_BK{eaz{Lx1WYr+*@jRyzUpyEKQx7 z2L5wj{Xi+X%dC`qT7$Z~lP4@57}1u`nd&1xRA=HKM-T^jFL_Z&#Ikd#&)0WLU31H4 zeL0FbTm3og=bzJ`^TYw%fAKuJ^vJ&7qd#lCy{$c@r?69oP zD`^uPGGkLL?pFHLTf%xY^OZ%+t@>ns(GVlw@p{{HaYgN(2KT#|bIlD&JI#Ka zBluO3k7m!p>-1ISg@5iEgX`Z8(B3cO){AgDk(Mq8Et8$RxRz(79k zI{K9dPSrX1rF*p>^Z3Q%@8uVFa7nc6%93lz(eKKV9FHs!&lbNv8}ONU^X<8^sYeg` znB$9Vo-JAeXLnTNKc7cz|FgvQbEcfuQda9sIb`UaZ!8YKFyF-XXDrg0al*rA%wC@i z6Wb4cnb>~4TZxS)S>^6~aqpAtg*f5vdpT|uXZZT3#^$3NJY%I4N3Q-Pkqz$o%Y*t} z06QHy)R;4}xix2G=ON~61#)lhl=8k+*aYd^pmmSZ1 zdVYV2Z`9e9TK-#E$p)2y-G>&xzZGcdAq;>G8hayU=@7I^EMR!;K-Usm%fUv~4- zE2j6&V&1+*>~{CP_*bV|sgn;`sXqA8DR^GjOe^&qc`@IF_q_#1S4X`|-#>!2{^0(Se+~`8Irik8fn(>2m{fz5@(@aq&dfQr{JQ3)$;Bnb`NY zeUqEc+!Q{}{*5VbArpT8h?P16EW~F=Wgs&O$X|ya_$+6x36Iq;^pe?;8-Q&05Au-)!CN1xCyzN!#;D~jnzQ&K+Ix)tFt=SZdB?oO@BM9` zGU zk}sK@69d2^bt2%QrSme|Gy=ZQEQ)mB|El zBY|CYxcG#Jr@8v49q%*LpHj+}Q!hDqQe_R{;xON9{N3YOCxs4KCtcaoP}#o6%2>G+R?j-Xl9R+*q#V+s#o;|Fug4x=R-5*ts&V7*h0O?`> zL0!XaOZnTr9wN8jm)rmFS84O)z~FaU1M$E>F^(=jS9=T7{5wUSI49QaOb$o!!diIY zkKhlR$>FGVDd{9@`Q3rd(6C@!f2*-~LVeqT$(AXl&4=M>nfMiorjS#BIu$on;MhX* zsY^Vxi@M&XKW0qKt9Y}Wzwpuaqr}I!_dnzQQaC6WKI)xg2lu@BidLnqX9$i#`rd^O zQ4j4#rj&OUu)g6PcNK93W9VP{xHaBTeZ8?uIB%y}Ir+xbgx4IN;#XddzLPJpXV4dD z%c@&%blX$#zfA50t8{Qyl1Uu z{V8**+E>23ZSXs_zmV%*#aZNB6Rl+5`eWdP3&X{if+6s`Aq{@*E`B)pxIHcZhTtL} zMPvaqh22;dn~`e3&my}`@jK#qh4kBp{g#g`WL>;@7IE7)kGJgyeWUd&Y{s*GaH2GC zYBZm#9W5^+{|)PXf{_m!D*k{i7dDU=U%KZy_P8YB)fIWfsu$rG znc$Hdb2-oMYmB2>^@?BkqTe5Fy{I)TcQ4Hh?D50EE|aTrB@PIFw^?(J0GBNElf#Tp za!|PpS28!^KV{DEZtUhBU`w1AHj;ftu*ykC5+lon zm$eLV9e>-37I&SpTC~Btzy~1){JGxO7V-hdfnOG_S%p5ywP3I4bUx) zL@wa3VqA*9rL8-HXTob5gJtb5oQkkO334OLb= z%H6lVk3F9Gz@q^BCE~2Z4nBid>6v?GgT8%@zTuZl#fa%eUfz*F=MoJB8AF8?Y>wlz zJo|Iz3SC7p?K6O{+H?0=nmGz*L|yw%JaO<@nvJJ<)_ljwH<>6bgI3FmyTFa^YT9R< zzmPr2{b6+CUi3ckh#0(IF$V59wOv9JnGUJ2Jva()OQM5FQZ-c!#5?P)4a2$MYr)>eb!uUWUkC} z&N_&f;M3Ye-VrZ^Q=?fusrMfZRWz)?05R;*Zc;22Da{b!|t9O$)_?i zrlEZ}v}@vHj<9br&b;};R#a^c?kA#O?%tp&d@o%%eQ(fT^W0Z)`R~SE=DR$emt+yQ ziEQw#zx?;tT;}^x`aXv}zWToI^54srxm3ShSpN`MWAh)aWS_=9a$4yAQubH&@36lV zBQXLf9wok+Vk?y_3{^YeJEPVzN5#kmi(7a z|8)CP#4M>VuDzd-emLLOlRu7s+^Zud=ZA|w?C9;5^Uu|nv(xgEyFPx!d?vU0io2kL zZqXL9Mn3EWI!!mcKPDay{tMp!jGnw68k5gO_D>Ca5d+MHv4wcnO?)Y`XLl(wbEy7HPQLq>}kaBaL0Vk zJQCb1scY${tli2TAK_h?{f@cpcTC`iF1>N-3wMvRc4YakkNOr?K2^y#3)ru{fWGFV z%a^ ztb&F9G{!`(d6rcPyaex%>=O2s>h3~`;SYXQycl@Yj^oU74Y#-j^8Cd5@2t26?t*tJ6Vd0b^GhPrEX9k9LBro z-!R^#-)X#S&qo8lG&t_8lgLIrLR`sQ_CbD{KkdES%WsV(9lsU#+KtXATVXD__|x&+ z#K2w~P7wof$#^e2-^=iLd>8&TbMX%mJE*fv#sH&v)upNN7c!&0|M}#al~XdCw_91w z__1H_4ur_BiQpY~(2rv(NR$c8&Nr$=}*z0!%eQL!|3a%2TmD@I+=Zd|LqgQn!d$n%uhZjYV zN1g3vf5TD9mAeXTxe~}T{PL*l3y&k0tGwrx>$fvMo)X@t*cL2|M_}|&T^*Xt# z7+*%y1jc4D77O`lq3`&(&u3gtJZ^IgUCBH{j_R3w2_OBiW$!P1_*Qck{ii<)Z$kcW z{)nZ00XwXp+PUx_)^lh@w!Cou>-PKA)`#tQ{s(z~FLepVvUzNq#nxe^bLhEjB71!J z@}Xzl*D_Y(H`_L#=V;#DxsV+*-!mWoo`0G7ps&N`L%%OGADN!{kk3j!>3y@>*Svjq=|}YP!o(qZ`EWXYTxt6&CbE8$9uKK_;6$OX7Z%&(Xdq93dtK=b7Ph_2~a6a@FaF`j!Aj z+AlA<7mv9$3P1I{XKmi_9=+4^L%g?Tvg6Z0CzOxf$Xww1bny^e6ZoaR4QvPHYr*h% zHQ7A!FD8<{;%T(6J{Y+GEvr3ZsInT-@z>#R_2JW0E`#!1YibdB8?@hl?Df_iW7&f@ zMzn;yfCnwA$JbqiJmt*f&JcWgEVgtAe}?EzIg_4+pS$M=iKlM?zr5!L6(Luar^$~! z$d4*?1LVh!dB~49=;tBMRr(RvR@#0R`O%2{=mnNtc?Vlsq2;bXQJc5T^ZedIjt<#g zTahC@Rlz|ugjQ4pC&-xSh(1X!+v+#HM*o!Y1OW)OZ3BEJ@2)Hx=7gy)Klm4XZySMSo zd|UD7zZFkT|F(c<(s^m$_@1Qq{<-$0`=-~S7!m2carxOahPpprhhUmshYOc}!3BOH zUM^pB8L%#A-ff*O(Ds|0A)FWgOTf?3L$d1Om-+Zc>hV$J;j563XA=F1U`G_-rx5=y zBPL((Ec`Ti_$;KaiwBRXK&M%XEah4?-<%z6a<1FYL-zY8rl-`OQv9i=+!G;N|81~+ zOL27FpAc(<9(>gLv7L+d+oPP{6(dY;+U@+d&q{i<{Rfm)o*|=eK_}9;L@O20iegZN z>s{y%J9O5SbnEBrJxoW%`{~EOam~~7&)8?ny{6~aqN{N=@}$Df$#qcA;cFx8=c}!k z^Ub+&`d&OEsOOyDY^@)Kofd9?MbCldJDhQK6Mt&IPVKy6=cs&&G3?bC|7Bz}j<4?@ z_-*)48m}psFr3%)IC_2VVrzRY@n7h$DL?sY7s>30czhgR%)cViOZ17`mNb7pd`FlQ>CK!%h8yY+%y zO};t%vQn^RjxQ69(zsue?`GB4?M;~#^ zZ??R{@t?QZcmH+*l_*Bt5H=R$s)x67Cl+EvecpP*se3x#hdX0{_&MICXBo7R_X2U4! z=`z|u&ZMwqQ;Mr9OOvxpk+XhqLUOhSIlC4<@18wZj*JbF8%y%M8hNgLpb=zRDRQ^U zu`4R`_Q`HLDj!fa@~q6U+sNIM)qvfGPs_+#!Ni42RB(Yu{ZH3;=Jl#Zrl)EJn7Lv2l~WjJdy0#0-nsZ zG8Xy46XC}ua6r6J_wvgNS9FyfTM)*6CvTc?Bbe!xOh<(E?(l%pJ!c;e@unur&%eTL+awT zMI51JO^OyXr{=z*DEdvsp2Abo%N(VQ6D95a3_!>5)goka7w?O(0j8rzjb+VSWygY(Ge)y6!Mv8io0C(8b9{q5o{rzk_{_lJ z-NdU~*if{&qna4!!jLt^ey1FZzP9JHd@nY@uaxJo(cytL@IXCF!s}`l*gpQytYycv zZaEg8Rx9xhil>!6ck15OFT*d0U@&oIca*S(! z_3RDe+-m>J9$Ve@=q=>Q-lx4h;mqhp@=PQ-ORx*S>-j5uFD`d*LV0!c?G~p#)lpSA zQ+)VNt=YP8+RJl`dOii5jI4FyhTNDT$<^h=xGjhGj3usZIW!~wr~3=wGG{WURQJ69 z;>-+moZi5VZHMPrsn;2Q5BKKX+g9pl{7zp>7Q8+Uyms@>o!^P*rQZE*>HYi={ZyTm zZoe7vW)Fkx`B<41{a>FPvV+{Q9pW1^S58cW+6~ig(lbYH``hsYdh@$T-}y%i zeJaP=<@UwpQ&&>2xsOeV{=+Mm=2ODIz1IR`D=Rww1@WwC&2P$_6Lby>t~~0nPaU`REp<=qH8BTZJu# z{F+mQ{vqFrg-)VpSzPy_ixl#FDbE)=`fEP8BArCGpJ-B7>EG@=sLc%E7H9tdr@EqS z;xu@!wzk=Fe^1VEcs}O9(`z54=l3!3Se`z)^BD#f=Du!hwDpI~?In2ke^4gQdUQg3 z85tiJjzAL{muTTXE)2X>2VXM%b$BxKf5uvk<Y=f~UW3Ol`#Ja~%6d`Eo7ipNkI* z@n^kr<&eD`!;TFB_dUSucT$k)B_DE+QP*Yz<6it8wVK3%xGW6biNHz-w--~AD`{N5#QKq z%kx^~x%O!sa^&7TNA9`xe37<4O51zzW8oi~vxhv!_28)Ft}kpAPu088wy!gN{D+-B zR9@}S|M@hX;%bNfGogR)_%m(#cl+>h`fwBNXy3Zc0|RYga-pTS_bsQrs!m&nss4UB zXA|4!1RDO+X^**TPq~S4uX4s*z31IJ%w4o;@XGOLgr)!Pwe9DNBZug~_>s{cB~!jt zI9!)mbP4>*zHzvG@n5jWOETklKsMmsEal8U(glB0ESdHC{||Q zeqyQXks&_I*IYT8dsQnTk;u`NkxGeXYYAoXjXRYI|M z)UD^TXO__~t%It~>v%rR>GNRx0`H2f1i5P&Gw)Tlh`#H$_8DvKGDO|$tkAH&*X1z3 zJllw!k?-`q^a^X&UgQKY`QPfJWa0mbKB^v#rI9|W-vylQG?qR}=O<2WSB!ZM0z2L} zyJPEe=2>g;HG&y+XpYA?@X|V9BhP~ltTe|#2TnmZw#2~6Z{Vcw)rSW9AX%~zUxlAO zD1Y$&khN1jra?H#F9poZI8y>m#T6ATh z^BY<*?{j^I23Ti28UfDu=)H2ov~fK!FynXhRjr}Ow@5CPT?Gz|)xO(a7h@3rX=Y66 z`65=C@eJ8Z`sVZWcQxw)XW(1DjPPJRf|t+FxxHVm-B7Ka$2|N69mE*M zi9;wr_a4JK!@L);|D}H;_Z8cyXH_rAF1K~>zr6MTiRb;m5dm9@{G!s z*mYM?H}d6ijn+Aj2KGVFz;R>wiQ2+ zeUIOQGUonpu-L%Tkx!z-MV#wqU^;=gT>Lc6chXn*i0nEGf0ANvt;y?JLOg$yvP=2< zwEfLdtIPKJec61osQtnntnKpdi~OyZPjEFk3yER+s54KZqe0o%VEcgj@@<-mqgxqU zo_KxoDH$Ehoxf8?zq)=XmKS&>W&$(js|>jlWS-;`)0}nPj$h`XaQi0WimNw2x<+e3 z`CR2M7Cvep^;Z;&Q{upixr>1d(sw04Hg{kCHd0_>j~iog1- zaENkaR3BIMcdq3#?de%1FzKcZ;frYRz*E)6XSjUw)5qmUuCU0HkA0s+K1-e^*qe79 z`Aw4ghWca+8vic!DyE~0-|HTp-seYuYq)U#@ulqbE`Tp&;IBVS-I?q)S8itZilwyA z!t_C~41A{jo7)Fo3gcTDgyG@xV)T=#fuatZ(o5a?Td7}k{Rnke-)mp19tfwj_dC5WEtEAdab#TsvKs&L z>+C@~tNiQo1=M0QRwEBuFI2O>;N+2U%4^@pJj(BRZbn}LcH$o9x>|gkIS?#A?SZ9u z?cXR5LS{I%(*rw~-@5w?UoDbM9NJ$Pn-AVCvT1fKdo5`5^=|NI3$);lXFT`44_K)a z&s|LW+Zz*!N0_57=s@Q^Xg%R^@LqOdViD_<4@de|-9Rp2VB7T&vGMHZ@1Bp3ls)h|AN|C` zGy4iTN1;N#G459qcd?1KJ06a7o@8FKKN0CvJQHVZc8);DowVD!{1%VU9GCcfXXg`dKZ`&1K16$`*}UlLv_8CbGkf9S$-UFy$K%-> zH`DIF`sTLnw&%7JH~yCqZ6&_UzDLPjHzH^?-+Ok+@kNzZ^sCm$=0rAo5!jb_CjVzG z%?lRw-^O(h{TcP4buGDmf8RFQsrO_d#dNiJR$}3jQz;1`iZf}?4YjfU#?TER^TdVZdBXXxYxB=&zP$v=AF&kzGUtd6YOt$Qup9T zF1S%e+)4Zq{6wr52`~J&J$+m}<_F4aC42*Z!guXy>H^o0lZNM-zJnk2>IcuOs8e)b z!8JQf-a%{IxoltTTsHldk63G@_rf0%l#f&Xb1KhO`RS|GANu=uy3#+@^S9jh%oe`o z_mAg$@%k9>(tb!_**1WjnLytY@Oja+#-Z4Lt@Y(Yzv(!>OyfajNnR_?KzY<!9~MEcPyI|4Ft_`&cvZ182AWf_yX5Q{&*K z;(cP^>__1lG3F-@z3JKePoV#BueEryro?>qj5q6<#2VTC6EE`9-}HWvU!gO>cj}}1 zlyvlvtws;YY)(Q?^DVbOnQh-n@6XI|>V~xb{Qh^?0<^DKo_IBVoDW|x`arlTOj%iu3_yI>%_=osx8y@+wE zUhQd-4DIH35;!XNOb_)feYm_Y0j^enw`R@+6Xcfa-wX||1rO#i?t{o3ty$KYaof7r zXBl^rdaBXW&9|C!{-ld(9jp+1u5)pLdeJTB$gf}Jz;=_fCajnPSI0}YA%{<@B`5e%VnPtzkLRZ|#9>>J=k>+bfn%F&`o~sWU0paMrFLs%>&=CZtIgi%9o2hpiYC7b zJtv7#dBFnb?EE!BV6GSe) z`XgF1$nHDxVeMrLrylhM$TK^la}m1AqZgZ+G%t(5%ijkwJ3mNWZzKE9P}icdM_VTP zIKR#BJNqbg-s^i`(^6}si4E(6_dV(x(Ig$>v~Uo*?}azM1>ciicB&b<3ciXiy35F= zKt7b;Q%-pRKF>Eb(9}Bcz5sniQ=R^&EOV z%~@Qg{4KYe^54|+Yr&JW=P&E|)#23d+~+~N{xv**p1ypBzp&~*m>Dg_FZP_uZwjZ5 zQT{CN8r}X{oYzfW(EBocFXs0@dR#KLaP!BH>ulZGiUDH|#0%zde-Ap=Ty(G|>8TT~ zIk!FiX#W>1YnSGhq+!>dMpm6&LB5AY za@R{Je-xz7XV+<8ni$R4)mA-NP;#awhD7J(CW4^Q4sypqz9BNT}1L!--F`#$)R`aWG z1HeN2e4m9*JMs>-bWrxF73dT`C7@sFgBjzjGosJl`RMnI?^HAM2fX^=$F>gZi@rs> z(A8?u)$frr!e7Y&t&ufAqi%i1PQP58^yAOfd1eu^i+W$C&Lm}fXtO%6qvZr+5-(T> z?7ArPd8@I-ZO`q8cMR&k+8OFEd&B5ciS@p|?acXp;`+8h7j4i*t`Gkjuo`hyQ%fyy zd((m2zX0<$(PvKIZs9i>nu}K8sr@Y-oWTd)57J2w_!tA8_=;Z_->-7<*`XKVw|pa? z;>vfdJ!XNgUvB?TJ-w|)`hQ2d;U{BzZw09>uuovLHe=D?aNA=W)G+g zeJSNiTRW^8JC5`xdJdii?B}bkDfU?pKh$&JbqCiQsq?!!Ul9BdPpzQOv1iEV&a<%A zpkJ~57;XCI8DhiSx_)fOq;A%8=CFWkHg$bnb5|;gLZdVpRv#PdQQ)ufc|-YmTwwWKXh9Lu8*P4O61sJ zjE=(}MaSZy31~J!+tutTOERVqI$44?=7L`d#u5R?l8d-bfgYW>K3i9_`KzdgA2t87F=Xrb@&C!r_$_7lc92(oLS@F|!*aiM=6<%G7qWjb?fLzB zo*PcR;68VFBn>X)2dI+=M)OVck!lJF~@NeTITnlW|;Wo(ww_*9fueAMvvyA_F$R8+uRrFg8{pwmXQ@+x& zzE#A9DE21~U#r5V>%x{`EwZx&xJ|&nAYWE3wx7yQBxWdx9?^)Opay*Ca&V>5$fIE2 zHt?$&{aaio3#k|;WLayYZ0Px(WseQ$48jtcNe@@wvEmh5dYK~ zlj`~#Vk%^t==&OABAaOsV_r!8dR9xFhUV#yx6VtoGt-^tA7jT5qc-1}=WpA*VUXWe z>N#z1<#`9+e1bpqN%LX&D*f~^2g;}QAsYu+i?{jQ$Mh|<{<2-?8SAGuZ@*8^=_6}6 zFMf}2%K0-m?9gS7p-bBbbG@NU`A36o*KnVXQ)lt=-c>5uf%^)X6GlFW&8)% z4z6jqI3@TAU#c`Wj75HZ@ebuyOgw}>i(f%=7lWqp=l@0ig)lg1{0sON_H(YP{Bp`g z(FM*9`WAp!(7T>e=J*!qN4jsp_9T4EC7#I;(tAb{x&gxe@m8cV$(`|r}9C_Z>RmZ zA=)-FX*T%t5HR4K)+?s{$@R^}pcit!y(`{oqLF7q{sYOI67<(o zic>&tY;^dkbbQ%m@^P0VpQpoHBgk9n_VPKumUp0~nf={&^YbXyOp-3^Pb|b zSAX>XByFFf?OxhW@_UbU_gQREj*B>ROmPM8N481_>Tc-^ChOei}Pw@-Mn8%)Q&h#W-j_>S7bVBX#IYi8obdOffW3|T_YP-0RSU=VfW8lRA}uUdDO?U1dLUezZY;lko0>^8-`*i3u1VKo%oo3o@;g;;U_)(_TBQ zlfGDVlZpHLn%@IgXMaMu3s3eZe2;I9|F^;B->Xdi%xebucl;6hI2Sye6F!&En$%pb zz+z4eo{g=zI#TynQTae(_qxBYW=|%4yB=LVHY%Ecza|Uu2f=S83*7Z!H$PVG`=BxS zv|vwtqf3FiVBW-jUFlu23#5miAkXhJ)LE4qNqvdyCzIp>lD=(tfV~I18+}=OR@USH zkRPxt=x>&fX*n@PThOg$fqhBT<_=IOOBsW zf#zP?s(y3v@qL`bDLp@dPW-0>zNqOtXNtEMfrJ8zW5c*3bn+N)eJzhCXxZD!c8a@#|e$(Ov|)81fiF4|M<-7?@O zxE!`Zr1T}9E%P@2+`vxx``vd3oOjww-wj=9y;c3GdV%_(sX5Sq_^a2jN zQ?%kft9PCWKh&<~%00h<^Wo04k_TxEIUqXF#** zX^)ZrsAmhfM?d29LojsPEW(fGr!L7kttI_P^Y)TrPcoWyU-J@m01q=KjBPeFq`3(> zaW!>T$j*PE@q)hz@OLTe`zj+i)KXWgWtlRy)MaS5!Iq=Dpl$J;@0+WwukPSpKGJ{V zy6a}6yAc~@+Y7(El`+u9BChCktF4o^z3>Y?r`=kvQ*NIB77@bw{+6Z)~d2Yz@arRiSDHlFc zMW4&*+gf-}D9h?>%n5ZahZaP~6OnC=$W`#Jx!&>*u2{GnN%7WFuHuuy3M(4I_Ypz2dzQFJwYin6 z#=L+rpStqy&PdMPX3P=tx9Gk8CVZllTrU2;{dLP)w&L$&|EJNpWpf9>X;Z#ML95-+aU6au`PB>GlHQ@%j&M#!a~XV5^t+h-^)YyJ96dV$ztCDs zIXt$Td5$2rY`u+m1j>m69%S(Z5;&ZR%^JFV$S$1^tB>~KnT zmv4Z7P^V&o{{Q43ziI^z@PP^o+Y=Zl{v-}R(Y@Ahr@)1__=k--+TK zq+ z0C6~6C&43r!gpRivD~@@zC-WQ%RZEb^U}+vJNkr)H$o<*?;r44n~#aV|7gA9Ai*zk zkV^m2nu7L03zxd6M>;*WO=^=z_t{8%;_f`(i$&-@8^Fta`1xl1IjYalEc)5r0C`$@ z9!J-4d0Lj+PxKw>kC!{&#@??nq<{ONkt%ex6713q@R2fW@2{&z?frES{h_hwK#Stf zThX#oUv9kS+erfXqS5Euwf4crHSK>!$3VB>GVf1^4LV|V^@jZN-l5g-^&v)0k}jZQ}wmfwNA0XZ_d);%VZ0vUew8@6PpQ zotE!lGj^eD=>qb2pF|Gaj4fUBw7n+Ri9MKeaR%#hBaM7mmrp)b>N*Wi3_0>X4|%`* zsn7I}&p$%$nDI@^pSq<#0gn~jWY7DSw&V5Ae0Xn8^{|NijyXm*F%b7zjTy5X&6YbAKk1@_fknP`O?(H?R72Io%TsbNH zt@7kmXz%Z&Tsb5kB?rgQ$8>+>DeX1F=cpKz5`2yek(sfXj03-8D?X4z z>_JzY*FF8Vtrhb4P4+O)%ZNYA;M?ArtX%=y(9?D z7j1iLVgKpCq_&YeKD#FC>4p8uA&H;yXf^8xUDUPK^ws7^;?LsUHea&&bFI@4^!sk9{NPAqi zP}eM9R#Oi6?VcI%CiEBf4bk}((7EE_K9?$u{+@As7az*oz7cKTCALC-o>psO(;@Ed z9FHS`g^OR|_J8>!^ngs1Je2IaOWJ=s_{|gTzc%;5w*Tbn;;!&^r?a1*e7?it zx9#xe4;lmND=uc$oYO@Uz#I4}txZrmRhqwyvtWe3hn&vA0(Nrsby zy5zd@zD>lw_`r>2qr+>&Cv84&w^ik~h0n%^Kg8RP3*)aCc*(!Fq`lb#v%y$+U^np^ z<|LE39LZcm=dQoO_-o*|UVFlY-$rnJ5Prb&UGM{Lf}eQ03%?G=9I?hV`GH@Rcs}FX z25u+mo6fb~a$RYsa`Q!q{fJy>V42I`d;nNVK5AcP*OgYP8+}A^tg?+C5HABiyXbQ_ z`pBPyQ-Zij^s=TwI32L(Fj~^yL#$JPG8LY_ z1gKNEPOhD1;Ya`)@IwO)&;UA3RP@v6=yVN^o`bKiDeS4kPaTRs9p0ur7q2S53L8l@ z(ZRjTN4zvqlQtLYJ#!(I$o7TRt#NY^`hiXINk9AwT|mALlcO!oha16% zn~x7yyr~KwuCCG#j1ET(twjv2MGUQe!{egs=;J+ueXQws`_)?1(Xxg2_>fNTB^F=4 z>R#%(nX$?K9_zVhJu|x1%4{yNvQAeLfBRd$;SBT64s_}KcLsKzu!wu+`8LLbj?hMY zwfyR*$-6NpeZQIQuNc~IR)RdJryrU>?Q_)`%F?H|@eG^S+5chlj)&5I3uj&27fvSw zg>BUX13RnfyKIuG>-^2a6F;)V#(!s>)8!kE9pTzU2SyL)2VQ%^?7^iT;e_gR+ve)v zUR8fjU~C(E0CuTNKQLMfUTJKH=}#~HNz$LK&V4U2m?c(b->cAV5j?k-xmQk)F2@Er zSah(39F9!~Dc{Jq>tthqLz}seQ+K56OUJ_&`2(OU+jgj1o1!K488I-j*ds(N!t%k8}i%zO!=3QZ${Tb=enNw z7Q@#rgEx1i;Y~OFblb}YcHZ_RSJh@AGBp?67c8sy`lGVZILr2R;Cc24@X&Z|o(aBb zFKBpbbn7yFJjh?!j!EpJ06B-e{R&IBJ2N6$Np7xBT|hSm=3ZN4q^F;kv-f8i_zcNA z|7dJ5>X}RasUw2*UgaC|+d6!MO_#L0+M08HG)`RMM?LLIPyD@dnj>SyXWY5a)rHTy z+Qa6X3k>44<)dvM{X^vV@<*fZ8j`+V_&wDyj@M|-Q%$EWtP&`D`;ZqEOp-T$=P z4W_lb(`mQI8K*lwd+lh*es}Hm?lOO|)_=4XB=~M7_iiI+vdgbI16xhc%>0oz&Z?Lk zUHpul!(jvE@+qf1^~;FwuyWTOTkl_g&IgX5RteyJSEt`i=ErGvKar)`? zW-RvD;n=~+GIYDLNU^<+oQq%7-XDH9Hsnd{zpRN_(%Yk%*Wdk`{EYXZD|AVhpAc+r zysrGU#vHD^#*dFBPX+2XPO*ljN(JR*y1bjIh~Y0fU@qua^>xMLxtZ^fc%TaK_k$U(JjY`tF6M#1pmNDBk@9Z7%oO`xd=(Forph zFVf^Qz*aMLU}tJwchKL~;P_bRa{+yxL7%loq(0A}&nJMH`aFX^FXg%VEI*&*Q?94a zG4N4s$#0tUv~}9k&)ZGA=va)i3mz4Nue{GwM`Uk(OPn#~BA~WAkPQw7$jNrG&urAY&79d730YH17ndHd{`l99 zm3@5Ixyr@9wfPq+kLL$DZ;R*hugA&FjPGWW{L*e8%$%;D9R2v0fIT>s3$8@4?U}RP z1&fy)Ptqss%xDF60c(rqyL(iY_t^4AKAAaW-nw}1sdu$Sy(f5P;P3dbz4vkUEq-^5 zZA(bb%mKFFcX+QCehaa8z48}6I7xiI0DCz$&DyOt;ghTL!S{S{n04#Xir<$${}43g zW4$RcExfyI@iWJZ&{yOa9LM^!{89QAxt)?v_F4Y)ECdb6C;Tjb<~OoB1l>Q&pUJrb z&6O-D>|b}m+P+To2M*-(t?0k>f^(*~t7pirWt{)$jFWZR-5Trl!lRMl-HOTASoKfx z#*H`2HfJ;s#hV2mE=_ffuyzJ2ILD1Ssa@B&TmJ96@wdfTYjfkxv_HZ2pN87r%&}qH zHNjfl2%a^9XWA3BkvbZ|iD$E{DEnY1%?D>@f-^mfGuC#1GabQIYQKabvpziT(~25ke((MIM-c7f&~j_+?jb08U3 z2tFn85gr7`Sliyw3XZKupN`#boqNM0 zn6oVAOlzdqm}>}MU!d)3cOI$3Yx}lRUo4Y09DmUB%&V(E%z}4$*LO=P6Q=AfZ#lH# z=wQN&C#dI~gA?xe+4vv4{6q008gGJmjWDkf<`w;S_t$~F=r*6ZEdXBlZzj3?plY#y zZG^dv5Icgc+bO$Syg+oc40_ib`I)0a=7)8YX7PypVf0VjwoCt-8+b(lI4>KDwoKmK zB=cqF1ezF}6ZpdKDDS0vW5)sC9mMOHd2?W?@z3_aaj%CqL?0y<@hIrxT>Eb2x{SKZ z=9Hz1zVyPe@{>Ex)y)3QxuUau=g76%SvzL=vE|q7JU8~&zc^PLyKruv6?$#{;*PcR zgyUA`YaNR#kC)8}*ty=QL*LX;rWPHiW^r&W`;}fJSN|mT0=~93hx2A$@U2yO$>I1` zJ5Q7R8Svser8n4iOSs*(X~A1RL8s*8kO7RjxCexcb~AVFMI~Mu%)KFc#EI21S;TH_yBv6 z(*x)zdEnrtPnD)NL;Ip*I}Xy0&sp|E>&2OqorT(Htq}=nZo3Oy-@b4_#H(YP6K|uE}N-_I%=tZ9`y(pO3{tf$1eH* zp;LQ(M|;)%HL;oJRuBYt)_eLqY+ad^cbADag6hOdZ^ZROk~l|KyJPtva7 z&Y{-ktnm%NS7$+2Fz0#X#;hCZ*luOE4rKEX`(At_OYvy%N8!m8=H6ZhJx44`igQA+ zfmU1ncHH#!W9_xBmx`3U-w^HFA=@0Ag{TX??@l1O5-1jg?Q)6 zA^XL(S!Bl`eVA+8ACiw;lrAHi;0oxAG3@LH{^OwYJ-|B$9MZdm$ebDE{!m%zebSHQ z=VCu)v3S#7;3i)2;}G#Z;J}DX`UuXtcDnT01o~PxIjjFlvRZteXVT$h`^#6LcnAYe zM;DJJWwR<~HrV#BUuQ2a<-1e|`NC}+Al4n2tL}!213SB{{yQ~~^DJv;$NF%w&bLrr z2kFCg=vKK_`Pl?x)LP|vc;Cv!i(lwju;8v9a+4qCY>$D5$#=(iTe;se!TQ2_=6=t8 z))$!jIeR`#emu?*6>J0-azM;!Wz5~cQ|(BPnhTG5i|4DTqZSyAWzDFY=ah?f3+7Il z`%(jQyyKKz{wVknCojta#Vj#ijk}inT*)}E+;MbM;X&vUcwpwz!$VzpBAWHeBlOVK z`LW4HcS_*j^~2BWft&oY_b>(@zyDHRdwB7IwH=EChG*`F-v+>;>X5ZlvA>c_kK`O^ zscH_a^^;Rya`YSD0RJUFflD4T-Ow^}GM%@l^L6oo`|$_3{d4*Fa@udCeV31mce{Li zpJnY5EbF{&BD3syuuQB)D|?b%cniMYCQrKhr#;oMcJt4L=c{S@sc`)QbECYGz;Ra$ znoNL;d^lQlR;gW{`L46i|JFO@gw?*DZrUIhQfD$D(|o_RQ}yd_meYns8I@_9>q@Ex)@mzRBh1Rg}{Xhq2zm5K)^IMlAXU8F{ zXCSNRA*;tBtCg>EKC*fRvU)tSdOosx8~Cz)@sYLL7H6#`C&Fu4ODc~qXa3we^v!n4 zv@L#h?e@jlYuVHM+NnVHB$dt1-rq7|gMaONm7jQtI=4{gR&<0dj*cMRxWmeR?J(!f z^}Oseb)LXSu?d<80$1gy@n45+zWKA`Xac%Q z1fVD2pxE^bk}K872J|J_S@H4k4rEZAu}jaNtMRhdY4ixY-+kiO6XDCU*-GHc-Zs7W zqKiz$Sh2@DWUj;@nX4_wE$+QMVzUF=?>KN3Jl*_C*@j09Ul(Vt^<1~uaK*^ zoFFz!V-)V($M`xuI5UVZW_@*Nf5NkjN&1s=2I?{Pc(pTL7t*q9mAW1n?K9GP&$?8#Jql}xSEnf;8p;#0K01w88jM>E0cUT~@d zT$LYFb|>}{a9_$kA##gQ&yD)FguXB)*`7YeEL~M&_A_R1{|g(ywM*)Z2W&lYsbb#@ z56)`)j00NLpL_Ay?#QBH}HPoPXjM~h`$(PQNQ%J z7r0q?0~MBa zu?m=qrxDX3eq_go^6d%i$T9a(0>≧v{B-aH;p#r}g1dZ9@fm}DI`Gg-pMvLhvz{=N4_Y=>>jLFD zuyzVZ?R{GQwi?c9Zj9X&<-7@7PJ$2O-F0T|n^c)BY*Fn$?H0e5o)3;_kHl@j(8WicYtRkMH*=<{#w0!2t#>mtwbG%f zZP3&@Xlfrc)rcL|h#gjh9o9I~N*#hmbbo;BpQhqpK}J|s$*)wOp1bvtW7ytPrY;t9Yqe+M=aJiD5G zTFCGzUm?yTKfrncYsrPkSl?3xmzANi>+d91%6bragY2G`YuPyrero47c!V=Ir3-J! zvtHcGyV?9z$o6f#E4q*T?qfXkH0W=m9?j7a`kzkYYcDlF*!*688AkS+u>wbQ*uGQH znq={G`Y{9Ciy({je*Ty;;#aNCWKLG6y*_QVjdW?>;XwnyLU^qHG}51BTKlg2pJ-&{ zko@m@k}*7Cr5dq+V)L1oMa!w8rdj)T=j;pVLiBRqFJx~4eOO=9(b?wOyi57 zr^lcd{2$s&pd4+|CAw!b=Z~O2B&}ikLxg#d{vh3 zZDU?|Z}{IM(CqB6*{5~GXCfeNee~yz4ct?8YuyJ|0`nu24*IzsH?(*wd#-Bbk-?j|{e*!LUY>@V) zO~)s^C(kuj=%(b~@w-bHPuKWSAF^ zRIZeA0pyf$^AvcXvNepUd_hfr8T&&;i;`W1jJ2AwdnmUbc-cDZ@cQIymrX#7#GC+i zgjusoZ_}m`XP=BiBQ3%g=-_2@wQaSpT>>U|d0_GrVB+G(s)FJ2yo5OteF>&zmc36& zJXA7O^0e>+)=tUc1;7j5VeE+HRQ4n>zMz3cuYLDw-_PCxzpUEX(S1;e`eSmNEOfeq4A?wbAoLN>t{#&(s znEkOOw6T!f6$fdzhkBod235blU3o3#s8?;x$L{L2tj<&9jBC8Gs)?AV=p=scMK38K z*Ibg=j&gkW*mT6XWp++{Yw_{%7gsdxvohIVKccU3M-%+g>e~yPmQm+7kc;k^)W0g~ zub^FgiLcM4?K1Wn%MRFseE2eB2%=|N23A&|)=T!%_bliA7kPgeovDg_#};)TT9DVT z@052=xT?B~=syf@fL$p3fK;QVzSnlYQ^6D_>p*b#+{5YSK8h z29ZnI1+2R=&M5w*&XK%dcyU!z+{$jAOPosiFEvKiQt*#xj9NpfR?I7S?;UF?{nDNs z(UA7>s4wNj^3}li!Y7|OUIPyk?~~uUkhL*=E8eGZ$p0yP%c2eW1$Dm?-LZoHYhR%H zXn_+2w6AsEGTPTSDyw=_rWQEoG4?&oM>TrBJHM^qZZCiGRlD<;MLn6!<6et+9cYvN z#+@@&jyay-yz2r76#sH)@A8&)mOaNCo%Rcz_SLrRSmChd+r_a6yt9(N7cifiyRppO z-OjwaG-dhhJ=db;O6>SN%WAH0d3%FxW3>{SrTFk4ac$@Cd**kr?T1|LwX9&9;bu%zdzJ4c91ttDWCp;rg{3ot#y3>@}BNA3)DxEcfa<3LVVJ zNA+F(W_;7#=O4Cn^lj(&X6it9J!%C=?3!ad*x`ti5qeG&vv69 z48_+5n9m|`V<~h07IW{)Q~V3W5u!tZSK4!50A9H<+Rr})UeSNym11DDhEPCR)ioEM zTgbc=@_i6^7Xj-c=E%pN;vmG6H|=Nu*MlbiaTRq-#%zO!8h-D{Th;jhuoGYO%9=#> z(D*K;ADlAlQA7KoVl$zYdu^M^&lvfz*X$W?-N4bpm@!P|xk4a8Jw|ejPPj*+H_R&IcDm?EJj^2Im)T2j@d!`#t`VBYQTJy9+rZTG!bJqa+959bVi_|Lu-tI^VnZn`uXD z#p^8JqB!e$(lfzTx4evJKv&3;sVkPhU}dD@zO;rx-Mfk0mm@>jl>N3hZL zyORD)Cnp2;=Ptc__Cn^J%fQtJ;1wZ`BItoLK7k|4tf77R>Q5=prU8c{U~JBcaTYRN&#Es0#|CQ%-_`F*zS)N_qloW`!(GXoPSI6k z4nwz!)08jD9k(;iL7dRpR2pAFj~*=UVfvWqy+o^8=&D)_+lTlb5CA(6TgA+r)^n8UQF92d%vFJpNIvl zNe}Vvi(|+q$zND8)JN@e5MQTW(;w(h^HJ~2$EEf|EliUynvVc*I{@5jF3djO2w!r? zir?@^It_% zT6Kx>sLz$q#0F?WHiMsBDiQdWXu|xa&-(3m@~CXYFCn?(=Z?r#c9I{6e^ch|J>}TOg(jOJKQBR($FyrNU-*Y_oj`Q#B@$IlI zax1Z>^hIZUKV}R@_sRJPYiZ&3&+9odQset2-z>1dJ7e8d6s=@@lb!L|y&v9xtiH$2#QOn!ciOXs5FSG+UHJe6C?SG2~hc{0D@UHV<_@U9|qjo-vO=|I9;AMx@w z9s1ffL#xtt41dX+YStlHlb%#Vd}BHBjSa*DR1*VN9Sk(ncBkSH6#up#c=%qZJI3C( zwO+UgA2jX^V$!;Zugh`aG{uI~ErJK}c8>#(DDR3~cr@DZc=K(WCNBjKt=;AWk5UI7 zb-*LfvJ5=Z@b;B7yj|(wt>NYL{Z9FNG-k_zg&$t+9S3&m?%!T-aCL{}^!=jdmcEy9 zt)kz5wBMhxoWB1gt?wE`0ls{}dcSmlbc?el2jt;QNEi zj``H?UD?CpRl?*Bv-ajirH5}M@3)zM#~)(1jyy(&nElC|0qI?@P`$ysW~D;ZodFCL z`&)mZ?0AU#5-aO$7Ph}xXMndT-*YZ)3pVpg%=*w&jb%hU+v=Thc1 z?DMUxI}8NvGmj+CmAl=iMCB0GQ5bjG@T zLG4}a^H`&Gt{(Di6JuibDL3XsniT^!)`xufGxcdNmT*;YkuFvXT`r&8bC+;_<%PLf zAzwxt=W}LB|0==;Sk4&*O4TH<%MQcS1&;o!e5e=g-zWczoM} zd3QB_n0&Q)2gz+2Y|`_Lo5IIu_^dn3ywLtU+TSv{%e23Z_A`9JHgbm9?Q<6WLHsRU zjA!{Rvr^-x*m3R4Z^;IZ#G73PX3`~$Kb<*SmovRjal=~E%Oe)B!I{s&Z_%qYuYz}! z@0Nl`+rVSnR<~oRmHYC_W9XIiBgDKbzu45YdADQ!o^0Qx7$m1}8_m7;B|qA}iF=&^ zvYo47Dg2lIv&7k-Uq;(5{`-hi@;?ZzHe&NyUv8I*fscu`(z+iVKeS=$Q47Zlq9cGtGrGLsj;;#NRgiJs#AZ=9hwdD@@HS^Ug3YI4d4Es zh_lr@8Tie67RK(*{r0-AdUC+=e`{=CCa&~%jH?IF^c?%o^N&n{hl;l_H^$!9JW2aP zoi>9T9|a~|XR+go(JPWKA3?z3gRx$&d)!Sexm_Mac8P;GkF_JEByowYW!ppx-df2Kf#lC87eIR4q` zuDpd7HwDp?g0X6b0}r&@wbfT%?3M>8&zzM^6JM0-cJ)PZlQY+@057b!ABr>HqA|&} zPUu{7%A@e2GZ)UYb79#aDtG3?dB@F%H5XtV-0qc?{x9JD=YS_m+vtK_S^A4kw{r!} z>t*McBpo=a{{kGd?R=y!91HE7#DQaW&YZ*#hv4o2j&|Tk00+7xEk0LY96Kd%SAQ}OeUsZ;Ue9=wi! zs5c%*axouTD23+B;DdbVQ@HAlPf^Z=!i`7q$FH-vA8?>IKE-+VjK%)|OPA|h{Og^$ z#I35Xx!`?2c;9cwuwdhQTTj7(H{CH*xl>1htz*4ghn~O0+VBQwv^3uwisDBaceH!c zCmHv0&X*hzcW7AS`vP)SdRa10yjlu>*U}&L>jg7(vglvet5n9b`^*U5t!WXjc<

dBd>(n4g}qy(2UJb)iM|S zKKE+R(#|mU9p2#VpPcE@&#)wA(MX-kThh_q_&fO|d&ggKP>O$QO9UeAbKG-(x3KSk ze%m&X7o+DRi?KBV*yU}BvPd~&5}1LlhYeE0dQ+A@br+g+6?RVu8$f#fWY(EN9nnbZ zX=YM@c)!5)rt3ZB4WJ(IV{KRQuiCKffXDCAjvoiet*-VtWsy47i(R35{}($hv_HyO zw`TIQH1wvB@xBG&W~!PZ@GTWyl>nXY|7#fR|L4bfCt(30V)_ZW2%Q$~Ey;a3lLs@pT66 znx!Mx$d_*UDrk9vJ*n3JDY;gGycaE1Ld%*D=>06PTUpOI9ovY55<#d8+h*cP-kIO}f~Fb^pjzC9obk;G=ixh6MCJ$D#S}hI(aY zTOvw%G&0pE@7~Oj&z}8KeHQryl7Unq>o7&oZy25K#Og=k@8HIXo~_*K*hbQutIt%N zbTGAaj$^xcvfCfWagQG!>>9;J{_gLF$DvnCTm8T@`y=4_-2XW|ntw)H-pa}O;m1$e zN5FIGUw~&2>qvh6=J}$;N1=J*QO71^%r=`p)b6%1RLl#@ljQ$k{Y9}ZDw7V7?I_(_ z%-9C;38)VJrBn~&;>Ty>qLVi?-ohD7e6s|*P3y|i<)v4!7aG3sz73WncazI%B|OYr zG@5=M-{unLcumARj1HNuC09D4y$5Qy31}D?_dz<5v^V7+|5j*dQqMyQM%i&C`Ir8ZH z#J1VzOVY$V^sy0;|LKc4cRMWEaO~n#cJS9*KEoNznK8+9CS-OP<~%QP-gYzZ#Ju;m zJ9=rb%a?-pC&2rf3#$k3p_r2&{v|ty=-qFvILhwseEtUTSFk}k$e<%KYs}qY_&u}E z+%0=N9AyrPY%q_upzAP+Rfcw|F9j?nG$lt$o4`{ zvsW&yocY}s@{3OUZ3BIql1w+ThfV#nG!W~}h1=)eXPec#oiagN%f zSNqQo?k=16qr-uBpFbQ}vHx((6bo zDp9Xl_T7#7`dw7rNjv*Bn- zG52PJ`kA4OU-}n4j#jRtpTfKPfWB`iie8-3JX-CwU6)MjEX?-nD!QtNChxAlGWzQC z-Hp2&P5wOT<-MmQceCefc&A$yK*v-MkKSEx7ChXD-*MD=$=y?V=RVHIlAcx!^?e~T zQES*sSSM;?ok+I%VrV}PJ`f!GtA`&N8TTY@cyKfW$KGq`%ME#53EHR`8eNb@54Rpg z&edr@3h-W&Oqc3xBH*sVUa__T*Q%q5?=zQ4ZhC7p4ox_FC+=xE4|z3exV1rv)ri>j zmI`EohwrV(1-sTZ*YVwdl4lt^mX7>L^Sp`gFLUZRXquf|NJsG{TKu{)?|T79=8bdt zOa9bxc5hGb{+G-`leS}8=NsQvXMD5E@;0^L#y9=sttK zF89|3jWn2{>L_#Hb=awFVet zt)!87iMZn1(5YG06r#Y2-0qTHF)?Fm4Ymh&mu{tfZp$Z>K)R-NLk*T1LCFruL=(2otSw6aKz@n+-J-P9b z7mg)&6rj(hq0chO0V{RBiE!FMn;F`a4DP_T)qac52+opZ8hx^$mfY6%Iq#iCx$Fds zOXQk-sNS=o{XElrDDZU9sh{r)RIf)z?)PEQ{d&&B3oUGUZNDRr4FCM}VAoeIZi5H5 zFrV;jQcDNWq41Uwo@ASXTiK0jw-g;H{p``i7_|?5YAiZ<-%k@!(L{J;stCT??1MwF z=$|)#e+l0B@YKSdvvww-1F=C*eu*=LQp4{zt~tU*B|W zTmSwI#XWVGqG!o<(+2&h4;#df>f<}c)s@%;j8k__aKo3?@5_?uY$<-uqct14WA3o2jgz}LH9b_;fGTz2gk+DA4kCy%9X zy!S|)-S$FFTqZgCy;94-g(G4B=aOjT-60W9<;T)v#im_vF zF=MW>EOJ5Raq%DHE19(zx)D!-n?vKE^(?Xt+bhl3rN!gcr->fDQt!c|*yKFt9iGQj z1}#`vLw)IBh~m{;ni%BLMDBQd^M&bl4`1(mA# zMi$uTK1AF2y~3;|ir(O-gXY^GxB6`A7dcaeaRv7Dr1#8nD6NZmk)@aH8osCB_Ivu; z-}jgMzCWMuGuOfA=&ZK0Sl?it=)oZ0pzt$Wu$-W-Vqaau1WO6DiqFy7D%jse8FUME z*ge+p*yq5h=1AbG!I;ZZnJF&prThX%CZ_v9s^Xt%a|~laT^bAfJ@s@`$5f5WnEw0r zA>VnpdHc_v%;=3b6m3m~w#2I|eDu}BpE>@O=wQ}rcsCd%j$z9cc5E!o0LOUY1)8z+ z0scL?bTaj8%{Z4H?EX>9U%hMh8I4_NLCdc{WtL~bzi2}Kv?}J}tFhyLMn04<^YCKi zbun}iCbmF!wEvyWyi>$GidoP*$}14&eZ$;evb2)3a#$zdA)e4!+yUGU|GBmZ<-$w% zhYnt3KX~|%tn{yCTzacnem8$Zs6%lFj89i!l(!eZ7YA=q*V{fE`0J1@B>K2B2M#~J z{QNFiS2ZxXo4w1d`y?N3gGbx6Cc(LDnlsA2vic&B znUhq5zk=^&S0+8%D(1=v^rm38V;hTF znWFa$dHXW-Av2S*99Zf-Sll{1{O9_Q@cU>z#^xiAExPxS685>Q!53dnIkBUUgy&Wr zq3*6l_0SWvD>$Vuvgo9hak=K>9TLca2I!?QOeiEKUv!c(>i$0les9#U}XiMYP zgD>lw>E$nPZ^hbopjU&B0+}tjt$BJdFNwZ&a+-YmezUwZYz_?uZmoyLe(=rRt>2(e z-~ajHHLrBqJ~iI0f!DphoomT)j|V-w zq1fd)^+|Svx4zrcA1jY~`y-la1BTUV`y*iZLQZ?aUxBZ!6RpQBwf58T>nY{f_V8_2 z0K2ZJV_=Hd;uYO5;qn<@aptH>Qlw|iqRmAdOKY3*{I{V^7 z@a_2kd?8PSixb%=H~*FJEy{tf`y*hRodes&{lNCL54PX=<>Lw9DJCbgi?4%4!-u54 zkONOHzU+Mt&PsFOD(DBUhkbDQ?In+XPBfmm@J+~p&yTPEU>lJGTc+dq@to&_?L_0b z;jhFWc|AUo-n53QAL?5te;HRixLrym#(AM54WzXyK}Iq>ztUq7(jl>=L}AJ}&IVEcRVH#Y~q z3#(_vPOg`+@Ht|3>)QhB2>3pVdmAeP=}KAnWaZ{ylzt?>%mOU(127u^;#@_zUn& z&zWzyI-|Mh_Onx~(YY@J*A(orL|v?l7=caN2d(?c`(w--SX*}6l>R)Gwv;d2Zx3nC z>$km)d~+3-SfBSkr|#yeXDWzw#-4EMZmxW`g1Unn=lSYBa?~u>Sq`dO|2+NTtcCUN z&$7Oyxcqj(17*y0Livgm+7@Vf9W4_2jXhFbexZ@!DbHv_tH7%gQbxmRs}I0Q1*M>harzluS8?3*+uRRn3CfPW7z}W#zg?T0t1b)Q|%f{2bBCRWF9%^kq z#%a+j`1-mFCdnR5f@}F6WDm&>wXnK#5RGY!aR||m*g$^im+(Ps>gGN;1CN9F_EYp# zYw{)FP4J8YZ$;p(Lf7C;x#g7ay-yqNe6P;!yYk`qX~aK!?|k%`*FV6r6b|h`{&WmAkLxt@YLS*kUH>FO&*_m<}<=wHFlz&(7VkoQkCy=l?DWetA}_}jy|V2iJiPbqK!e9owy2%S-f_8wIF@^w}D@^#gc zTd0!xUlDnOYQg*C#%zwu?oF1Y#^21DZj3QFYguK;`$fE!Lzm7xaTz$vzw7)Xqk-{! z+I=?fsskILH?MzF2bfJyFpeI)1^9w~2CU};tKOLd4#qH+>v(4jZL932uZ6mvWn7A> zPqErBfX6U>pJh0=8d=aeUSp`dx5O;6u2`QOn_9xUcPIE-eTA`Y-2`r)05^@aCEvpn ze1AUeG)*9;e>`#8nlthK`QV58cX{xw>*}8PFyD;kn}WM;IwIJ%LtFl_xrwpS9HE2n zALW^NFw1&A_Ii(O5smcF2YCnl=OOBh5Z6c2pm>EPz&RS3aEse+De_>ni!=3OFV7da z&j#?V`f)zb&j(-m&~7QX8f}QNA+|C*nRo~2uire%(Ag8v;9mSyFM`+Y;J2H+0$J$Dzzu<;#)Tz4JZ=c#?$E?7P znR40r$Lnm37vJWN*JbW_t>#<#fizxwp*fu~<+b@DW7g)5nU|w!G~da8L61ii!W zyEJyIpm${4iSyHW`+0%075JZ{A2oRu2YwC@5qGkSy_z-;T`O?S;jD*e*rTKTJ`aGG zx!eb6e=P0Sus5Jp-|~I}K6nmTiNV_>y`gVBc-hor`PI`C1&pC$PLz{*kA9H{^7{*R zm-co_E{xHf1^N~Jz$-he;2*DEtB0UT#df_({nA4v(24X=JLM$~f9o$VN7J|Hr`?wzeZaN8lRrQ_5QYb&hss?(5dC}pV`t9W>py-M z8N1S#>+v<}QjA7-Om$&L4Gu=yhyls)p5*mrbg$Nd<>P%Hc@#F}I^)@?#LiT97dW|! zc;EA1e$WLhE*!b?dU{U&GU8D1b(u&pdy2}NMy8an-}+xElk7C`W}Gq9o-4&KvA25@ zeqG@l3VSmlOynFXZ~(6=l&mv8IO)GNQTa`~3mmqmi?#_>QPu-w{7=mFVDI!`Oa{-!!|;pn|e-a}7M%<`W2 zUpbfMm((6W`J_s&j&`+%%qFi+Y}$#VDBH#RqjIw{+la?1JZ981fnC(ZDR*c{cBi zmaM`?%@1=eI}QC9M|K6+#~2KoF7oO08HZf|$Tzw(P}Ejk3= zLgWRnMqhWr>)3*qs~(M8E%*)a{HreR3c>Ape0vrBGanXAzF76}O*v=OmD7GF>j|q( za7QQf|9E)YW7vF$9_Osfdd@Dd##Re~`#5_C#UGB%=K6Nwk1pblw8#E6=1=mgE1!uB@h@V{fBDjx~07nVqBY$ZT<1RYUch6H#23#HAmOrk|X+NyJ zo2D=|T;rl#Gq(ic3C_RHxc3vC`(JZk=+iF`xa0V6&iFaHrFVYQJ`mQj%w}Ep z#22{p6U7_`FwcrikzE7P+|I6Z%wk=IXPE@QjNuTqH!B*kxbx;VluOsmc43jO_Ujif z|KUr-dwh6mZ~nsy=lX>neC+@8^%tgFF$+G#JoJv`A zjC4hNsp=fCag;-6llJnhV$Zg~Um4!_=u>iKhjH}CR#&cUL9S%5R7Pevn+L(PJ@*sY3FCqM1* zQBRuG)Fy1=bED}O7JdCy&M{e5Utv=92KyPF(Vo#p;yx9t&%A8cT6<2V*@WGlp3S)= z!lmN2~DNUSww{YuhEj)Xuu}C~za*m2LNYk~QaZ z(68Vlh|F&v&)N{Yt@U)t73d^g4*VS@QPzrjbpmqca;*~yc4DxX>stDu8aXEYREZ2D zhVcT~Guk6AI`MSG6}+!^t6UvXiH;DRSbA`AKZkh9Fm{c>t|=oQanY9I)SrKk&DT%) zLCTRQ2hH5kHa`Q&RC?rx*5CYl0`5@EfKMmL|4{AYW%0CR|LTC*6vLKCU~|Nf|BU)NZtq* z{fgc_+iejs6|xaLAD8@p<>;uw3^4&$1LH|%^IpaUzu6@5ZWi9{glC)1lD<4E^-49m z8a^tfZ`5z&DN5+8a(l3*-c<}A6>)EKF${FPv&9llz8 zM=m>feJ?c6KIu)GXDQDCa?9pU3qs#}u!FWE_XZ%d^O5O=6WC)LCZ3NuuJpcaJd<(g z>Fl1k6-S0DjzhZjy@8gt8_X+CuAj4@4cQIfhZf-726NCUKTYMxl691C@#Ui%ta9Yk zDzCi4$w!w@x$)?3i({{E_vFDQM;<)KeWA~uxLq_4@7=`siO&7;8nVG_X$N~KRq`%* zL%?D9T|@i>Fm7?<*x*<3Eqr6`2jyn)Y=;m2MBUIsA>-ueHy00T$o=Qh_nSHEhl8>GBYAN;c=#PS@W&i#&w^|x=#2eU1MHr1+gI60z;;Z1RllJT(W%$(XVq_P zCV%_)&^|WbP7@zy?W!N=;>MNr9zHhM_=n!u6^jqru=t?8a{8$AFkR-d@dpIlS%PW89}aXOHpcWhX<2npX<9x6wvBJQ(-v7#DWo zaSQ$GFJF_B|K2yh0A>rU59#WfLoo6mD8G^sd*MD@?q*e62=oa!8kaI}6hzj_w zr_An|FP<)X z>pSO|h3}2AnZX{(1+|JnVLdbqjPDe{x0)B<#k{!S-j-qX{S?Uv#gIa;BBsmIPG%=uRL2dL+m%HTo2KGQh;E8`@;tw(QfTK{aX|2|U+O~*`E z+cacxh8*w%&BU6dSpeTzdsFtB_IK)BjQW9jJEoF5{c+^;+97#6mLTWqu>;m3Lpv|S zuMge+7jPiM4w`EDvhHp9t>n~Y&=hp^8rQpdE}va{sY(AZ=iOiG-BN5+ez&VXrOdgK ze2abv)T2=#*U) znfD*a3);!$r+;$z>3G=$fA%YzRGuT75^h{=51%G4Nj2Wso5REME@h8VmIa1VV{XlW zQ->cMeU$4f`Qf|tA9?EeO2|*sGryHS)cH$_(^H$p$R1n9m}#xF3BK0znCo*d;ko#J z72}n=#w5O!ee9j*RK&L>$hmg(tn}>~{X*vnV>Xu{)6`emwZ43{`%!f+P6RD{g0Wn= zUXI-Mu;fl9a4eqU*z|u3oPS-vqSm)372hLYtbXtHZ$ihLLeP35UJ;?6lNH}5`|HU! zCF|t#bY&;LKRcf&Bwta_GcV5ElYcqydh*Wu7W=Me?t|=`oHM@u{h|K#)1|bR0hT~h zG*ZeK_T)zJ>dk4#-`{cMpZ2eZ`sJzY%G!Zu^9gBbt~=-IxLP`Lr)=jAWXFFJ7f=$y zcIJD{D=ToK+uZuQ|&PsrQHzLsS~ncP`&y$z-fq~B5R`adweP~SU$l4RLOMTaTyb6BPSvBgr z;!$+1xk`w&W%)CMtSf(%+*n0ys$Xv0>(j4GbK+TLKWZFA^B2+oxhcw6Z&7|T zY_JgD`RU?9U=>|hdH_yKAJ9b%nvm@sMjkOY>X93a--*`%J=&%o*$MOAKIO)OV=g{ZF-e?tj=rL(Co;L@EPH{_!HJDtQq$d{tJHO8D%g3qKdRiGS^)R#qe z7aj|2&^cm-%-=-EtD$4%4-9}OKa3g|<$Mr57fOe^_!#TrBlkQ7YyS8*5uJHz!;Tx!{oVGo+&j)4zqpHdK=nZY67+b$1aP z4qnTTedg8j3HQEQMQph2UA+sAt}7ez}_U;Zt$i7(*l zP2_&WzHX=O4*Dn^R7NaO5jo9^?l8+QdLEmbXWsWW@||~W>oO&&&r{~dNjo{uMfX2- z>HScvv%h`=XApdN>KQ$I!ZNdv!=X?_{T3{jY-3!|ZJAIkl01_+3$WNRXm$5_h(|K) zlf@QY%-F@C;{^1P7Y(F3ZeaWvml%3b&$L#&`6+b#EYThB;_pyCN6`gikZ#9Tv@j!2 zR(#RT1E%lb@gmO2Zo7eRSifAOvq|N5ppN3|=-tsPYn&W$zwEUstgLK}Mz(4WitQz%SGRacaJV)8Et&kGsDV)DvX}8DkUv0a2b32Dt!`jz%*RWKUeUkIwkES!_ud@8Nvxpd?2Y=Jc zGufDvBjya3ZmS7*Qhq9UMRvyEwY-U}PjO!C8?M}BES)ulR>vli z?EW2iI??{6rP!K&TSxu{`B*~KTSm^@FHx`MMOPoFzL%&^en7^l=bs0!bG~q|ze2Gd zLHJZScKK>k9{iMqpYm8Is3-qsy?zhgYxD2vy(N@&;!3qL%knF0JS}{#?*ToYQa;g>kgcug6>>~%(iu@5$Pm?i<}Z*n$O6HYEd-D7 z!va^f>$!NU5}vLVPctXYA{+i`8+68RmisXLtk_1q>%E`g{WW^ut{)U`oX-1gyel~- z{`bRo>CiU1~USY4d zj7e>-2WQY*?mPQw(idZ@X36*{zRtcl1kUjiI4bZIev!n5Kk!Z~?Q3(cRTBaV)Dbc@3u+0ehXq~4dV`*O>-_*EY@E}AEZ)<5o!jpl$W z`x&35!Omq1M^a4p?z_&A$W7*bEzxSAedGc%R zL+6)KxArGrb+TCw4Kye2uTLkiC8Xc^ZnEUeP~Z{moc48h+$h73bKqL~D4Cv_LY|3# zVcZ#`;DW`-jv(~wtUu{{!P^P`bsk_F@3O9&mhGDb_v5MC|6FAn7mbg`X^h%F*${ea zW+*>W;n|EgaSeq6B68^~N5Q|J%G{ZC(vPZJ{nGROyq^U> z8uwt}cCH6*`~&!4{`-;N%dhIEna7RUy!~DWM}fBb-RHx{qODf9O!8QG`#Iwg1y1>s zClJFSys)p$#`k7`L*wae*q5L80Ds<{=Fj2=ByYDqr~e1pTm!xKl-5Fvh~;a8XEKbt`V(^H5c?D!7zH1pVBx5fDWzDlf~c#7waeZ#m&-ao*aq2EsNUmt^Zys=UncC6qx<F zs#AD@XIlC7kJTQPV{hRXa^82x>NmQFXAAlL6gb&^`daq4)thTETzIlRQ zf1BG>PMZVuD;V9e`ibr(C-^O*O%{2b=ihe5a*=OrWT&@Hb8Phs-sU}x5w?2Uv}okt zUA^JAp|xHrny-B)7^yu_ZqFeJpH17uZ9MizCwAA~laJ=#xzC_~3J(SyzH#mBp7FQ- zBA@*ifPRwarm8#erQgH1#I6KTS4CSUa&}f=;Zhph_`3`<|nx8oPZYH|; zk&Vgd^z+>2LY1s)?!6{S!^Zs}WbV zkGLwGjaNV1iL27Oj?NL}+n%^8;;tU-I2pTdZ8Wm>j(y5-W>Nj^f zd*82t=Y_Kh_P#&I-uF@LeP7Go_Zo1oJoKgPkC%>DdkN%$XjZzYw89+Hx{{u&e!(RF ziF_PeS-UHmF(FlqO#bC0_I|)aHTFGxAIRZV=sL-Y3u#-vuhpOUu@;I8Y?uPpET>7tFXE>p}GV>%st_5a=RaP>!He@_{*YNweXT~^aV zgcNYhZ@KxsUcI1sh~{;&`NeA|NnS|)ATtz4t5~hk@SewSFWLC%UjL(fJ#FZWEMuqq zOG)MehV?Up9N&VUJIk0UzO4g1s2}GrF13tl@s0Hnts#`--_dxM+?W^97;J?ubgw^K z*Op6Cvp?52FY-#-m_7NbuBpr`)*Y*PWZfInmhJoXoevWexWjoipeu@Ok`3~G&KvOF zf0lD1I3IGxUe0uQ(d6$S&Tz-$#~yfaJ#&SDflm=rIM99*Vs1ZW5-*jO0>jsNkWF2$1l-BNCJg0EDL?g@1K#J?-|$7#-KZ<8qW}QsIFS<{u;yi-N=i;U48xe zb}Xq&G_Y@4-}oQfSJ!VYf65ACf%+T!aXDlEQy)HVrcM8NvoE8;G@Z&?e6*zz9h!^J zGpVaj{`;Kow+u%P*4zd<(suvfmAaC~F{85|fJA8EpsBUjP>K*}DXAG*i3viF_b z+cwt|j|Ln!dTn~^6oT<-=b7vg|5$Xd2_8|LdP3_J-!l=_UrBv|pvjx5cFPf6t@}Za)8tRi^5@fh)it^#s-Sz$&LM{6d{SV!y->P2_p{*!d$9 zd3*@Jzhul_HO!Td{J4YP47r07*~fM#hX3I3nrT}Px8L*b;n1SkO7#4)#IdU$Ns#X_ zL3_+WB5NLfba&mO>BDu8{)WGI;S@mr+JH7 zSIXDD1l;B4^dsj||Bz@fR+QAge6uiSzEa!1Lrx(BzI z5dMtdM*M^hPGjC2JFu~qcW(nOPhJVf-3J^UrG0IlJ{XjPUvl~VSZmN$PQ`~>+};81 zBv0{Y?;yW-9~~e)sJi#Nb^G%Zc=&VHKu?Un#uMPr9IJR$`1=enhK58Reh$8@H>aoS z<|eM_055gIRTNygxS0)ZWINncdKK}w`3K%2pR;6}@TT<+{p!0gyjloO7gKNNsrR-l znH#<$JkTjyOj!*Ig-!HY}-`Agz!M4mo1Cp`Iizb^=XBcbkQ}Y$k zg4SDPJ3*%|?YXu!^dNghG?8%Il%3LXqW)vw`TgaJ0gz4Qt!uPlGyJFaPf2g$SCZbu zCbRpg@H;u>#EYnWhwia~r}OLR$qQ{c`wLfWQ60<`BpHtIO~U}f7xN49-~d(Nr0ZOgBzx;uDJb+=8! z2g>>na~t?KZ{z#Oi?CsTuwcaRLmQX33m)zzZ*W1l#xju&&pZKuyu8;Bi#k!#> z!NMG!y$0k>XTD#i_$cZxrhd`i`^XCqW-B9IdtwRSJdTgxalYvUkDh;{jrI6;#=3*Q zFYI>p9`ywRf#u}UOy8e{hPW>@CjBA2AX>~K&kgXH=ozV*#8GM<=fPCkbn2mv%>VOg z=kX7_FW1~v`Cr1sIW>RMp^*~m$ROhsL(>NR%Gal}LbX<&(VSu8Cv5Jcp}-SX?EJMO zQ>*Z;58>WEAIiQld{qOv-y%D|!}0ItInQ^RyN0BS$n9siH}ZjwB_}L0@z^f=eY2VS zu~e4tk2v=Y<_#xz{=3|ZuO58c(}#{NCOx={ej1-FkO|=FZ)bkkn;$z`AexCrqRi9I zVr-pp=$+GlmAd@%Lnq&K@AsmUIP@N8Y%Q%c#=6EG?xp+qnpjsD8DXET;~HIU$D`@A zl=FVby=Q#H8*AMAXxO=57jpb&v0j*N2Bs6`Uz?s|-#BdzOMPLS+c(!17oI*(J-PNx zZaks()~Jt_E>HiF_4q4Yp01oW&F-nqLKlL)a$13n*&HN)9(c__r}8oW@4ZgUR+M*) z`9$jM&vvgU3g-CPcC8D$dcj+od%C=MS594)?%c6ao&K2RI`DhK z98SNkP~N7^;@b{nzi<|}cs?if?DhDpknI;BZ)$l@b~tbxSc(m}qPRoH-Y@_6_^SZVXr?pEkX`O$+JA>^ZS6o41KLO%K)rcP#w@>)XWQ~(sV|`|>9HsH z-sWxkRbJn`P2h(5cX{wBZ__BgQT%mo-X{O}yyuQjG4Pe=o6Sulpxflo)E)S9HEx2v zgl7q0A7x<=9oWWLXbsZKdn0||!N2d<{lgz{;a@`?il1%*{&wtu4*BYl>skvbp%1iU z`!E|`e;mE@4)Gc4_Y(AhjXA%o9wK+e`^I`msCAQ{tvw6uiJU`8BR_w+J^F`6;Cs?c6JB55v;AA_vQNAeb z=^4_gk~5b==i4aP^DH=f(c&y%b3W;eYCo>NBV0`(K0Rlwf9&>oHQy?BS-5g?K7p%g z*!bXUE4V_pT3l6utJUCYz587oF>m@VLyUS8zTyDyY0UrHoKMP4)e8L2(T{#}J^`0< zK8e=+IiKojzbEIDzGaS+z>o4AuvTH~OkwRx-z0!{4e&1Feern-VcwYuc{{eP?9KVqk4*2WU**JLS>CMX8=LcqHdRLgUiG(e z34GzM9Hq1+kY<_~0)GvPFz$_SMQirGS z{XT$z)~}=FBs$f}EtyG})b;qF9NF#I9Ui>Xt#7XPybak;np*^zuV%+!M?+8YRfNb7 zqWSB0yO<+rjMh45m;4M}oEa1Cn!)d|_qoQl{4Cc^vJLY}>|VjmdWZSQHTu>qa+ zxgT=Bk@_Tq;;doF&lsrBi^P#9m9$qiL9(fE|9$FjFvHJC6=84ugIm8Jzsl*NI^h%cx6F?e zb6%&UoS<&Tkhc1`WpSgUK#NU=--f#zg0=;8tKjYZ%N#2>oJAN7v-FyE$dT`=ndg(zi zBSGR}%9&FvK8d*tazu6O&(1rbvq%4-d&Nx2j-yvOcH?{Gt4Cg|EdI__z#TYHjGmLc zt|orFJ~<#Y;WcMYS9NrNl{eB+&m+e^4U9c=Oum)ub?O=5);&OHa9CL_+2_e9#oA5b zTPv54Q+ebj(q696UByQ%Y``=NU#RST-B(h#Eq|w+buedtq534*S4I5sHpW70Zq~+S zo;mg@XZ}>qx8>ZI5r=H+q29&NL3#XATb7HP!{ElT1zn#9bRs+HIbek*o6Q@}d}))* z*BM|IZ&yPH8F=wAm!7@xc`3)=CqJLB&1S=RT<(lVv-&81AaFKU0jKo~(!Sykpp&j* z?8Eor%NX-)#ahRi-^%x5{Rq&Td@za$_#X4a+%|sV^{ds9Cm!6BT)0mZzcTIw@YfQf zSOgA&v`q0KV zu#Xgf(#m(mVa{f!z67xZ33Knm>PJZM0=Mw&#M5|Z1G?jjuVOqj8^fo1`IBaO1%H?E zC;W?EpC;}`K7?F(a`Soe)%D5~y&q@JC)pYQUcioD(!1Xg==#{V*;^0a8R9G#LId$B z1Je4&nB>EjMlzS`Ovde2_OK6!1tR%^n&k`1rw{T09ip#U^iUYTP?Fq-@}WpR@cqFe z_Q4dNYnEv|XB{&e3Wu6Q*RY`C|OKa$oU-1i_viSG`T}RL7 zJ~UPF!ISVgHrMQN{CbJT`gC2snW*@Idiv0TU62@IQdPuKX@4<3L*_Z-uyKML@B43L zFg&REm%1ARU3Kw*`8Z>0$#*;vbp1|l93QqpGn8lT>@$Ae%f~MezjvBX9!}4Rm$W>2PE2-w zb7e_Px(_-1acdtyv&-y#!^J)KI#bwI+ZTOKV5{KKMor& zL%ij>Bd+hK$*~tpC?CnYTlDU?o%0m;zS0Z(3Si%zt1q!T{I;85sRI6(_#b=fB5)}E zqVmK!p{{LOBSClgVFb3FrpbxreA(CDLfYG#)1I@w)f>-|fmdWl)I-1Z$m_;X-hm7_ zkX;o>=BL`Ub_CzZw|2uGr!RiGDTi*w^GS2$sVqDdzFqr}&2aKYukh|gcSDdfvco^# z>68~U7hjC5j?t##$z$ZToTo%o~f>IyzN6D?5m}@FZ*?#%4rn zAA4NYrq#{d%LdA&^B)7RxU;&n25*0S?Ip z-B*9d*@quz4M+70mTKg~C}dEbCwFpW#N1!`WJEu5XOmn1N6H=6)%zdOIeT=wTepdpU#$?Hs!6}XNP0W9ufJS87KVjOBynrpC@wE5Mb9Da% zbDU?;!*z;DcKs(yh_^$h+W5yh{Hwq^sf2wmb=Y*Ww{$I9h~p#50&81)S!CMfX8Gq{ zp%0=F>e48~)~C#lPhhb!}BS9tW=6L*jPvT^s&)4o5sb4xew zz5w~7wY^Scj+Mji9*Ka;;pb_NEFy;6*&}gOatz&9#&7)3-7DUqKJl*Qb>>jmwoa_S zug#aeHr=@UKd27A+0L)O%?_2*<^}vB|8{zD_rF&;ZI0qMMw|cR)OXN4Gs@A0YW zN}Dwku>C$LnC!LjJ?grLGVgvJ_s;nDuC<&(zy0w$H=iT^%aPYNuwU%+!X_OO*FT*VyA8zltBUua8)%NmH;Fs*W9fpe7WU*Fk)C0k~k^ zZ|{+NGpfji#s0_5de(veAj#MUplQu*)z2U_UHVZp?aG)>Fy3~a-|Xyx`J(6n{2kErV^{`9^2bPV~Ry*+}bhj!7GmS3<( z&v9stGs2zyJyYDa&vx1-_pY<2=yH`aCtFAxmUdl!8Lx75?H!bJ`0$D`Y8&};9KZP5 z{P1134wqkK59pg?PQQBE9O1V4C9h3)pUUZOn_r;(PTD((zo^X-HY!!iKF8fJ<42|q zaEi{GWbLDTeFIL~oi=)K+ReRi`fpyji_M14ItMb3N_PG4~P*R%I$ zuj-=h=bScr_Gx8Q4o=VK7o0YGGU&G|2d80vOKE$z+J?413a1;qHeH;)q&oO!9ly{@ zPn)eOr_Dk9+BRLB{#50(8REB?HaAnB#pwe>Qdw|%FE|wqSs648-a5-UAN|5MS0@f* z&*oXt^f^8m^_7?HtDq^2~yx8!!@M0dkSWDlP zuRt;7f?I!zEngW*M&zsZ%akwDS0nlvTWHU6zDKygwi(JVxZ`Yi#bn&ua=XeWN7Emo)`v=YmO z512aOb)5~e#+c2CuQ|90*uAM{SqFOcOKzFBCzISqGun8+9r!iJzmvXcJf&;!fu%?C z{%759@`jE{VgH|1%HCtaqkblCuz5Bnht7>f*pqy#WVdXa(1=h(W%B=o&gFVUED}4( zOp?sTewv~FRWf%9Wy&KpGssWKcRjgkZlr$e#|i!SrD_`-2dcqyf%eb@b`%}F_f@qm z-|{wajLvH)k6)4^ciWMK3E13%>eA#OiE=GBiK6{L-J+Nd(Y z9kjD7Fc*8seSenVhJR+!cf*`%mf+_;0Q{vc{H1QLol?QiJHlho#hq-lFm;`xcEL@6 z_Ts>;wTlJi#F_DY7VU|)^xeB(b?Msu&TpsrujLE##!B%nk~8w%DBh`!G1WMG9>K4lzZa>TJl3c4o8_C^o$n5spNvRlhecbie5rSy`tuC^@X0!vTpm2r zEj+#QE=DXx&9{ls1>bAgpQm+g#qWlYQx^Xj&N9J%Kjy`ex-kb~tq*L+21NdzN89r0 zZDkGY{6FN#;u{=$aJyiFHh?wVc0)9BqIeW1e`+s$;+0CtU~D7L9{MQx>FFVlzJqq{ zq^3pR&%|abv9cs+ZKgqto%XQn*%T{Ng7d_G@L0Y8-StM#J*IF}Ch_o0}zqISlizdm@+PSl57o#OAq z4^GgB8*}=w^91yCv)xbB>wnFx2}JUtsZM0D=*L@Ym+Y^=58DKdDNpi2{D`sv3fa5Z z#`tQ_a+3S(8qTRWchRdM&ZbG2v8if&KJ}~-DtA_1(#94wCoXf=q+cT@Mr+2}M^u75 z4=zH!OCOO(Rr*Nt7V9601tOkKdI28@cx{f?U1fc!m9N=2k>;w`o!`WszWWY~fO!FO zL4Bdlh6YrKVw zM;2WszrQ!`pSz0rHL~?@nO_V4nzz|_UT~?tx2W%oMRqW98rxCpm|GbaZ#*`jXZ`c9 zGTtrc-5Bpy@UC?BqxdB(kGpdo)w_s#-|lC8r(1uYZ$4KNu>OjE>74!IlSJnwWs!-D zx4#XKmKMA4D-XNmwD-@0|0N55?|RXL)T1%5_@sWXT?1Zi{oZ<6Ke(FH53c?~OUucPGJMuVuF5PYKu{m_V#I-s4qm`}5=@a>N1j}ORvmBabO?JUs z#mJXy9C;+%S-SzBeQ274%Qxp+nb1(Va9LV5l6XKqF27$ubexdi3*SH`zJa?g!dH0> zc0B79<=7(FJuUcr)5>FA#QhZd+lD`@`1aG>y6zMI?DFarJ}dE|!8enIPtl?9`JnL0 zcdAb@)LD3ucN?m2nAX*CgL6Im^V7OIvB`DjhwKl%_nM!DN zKYOvIM{Wk+Z-VbPkv~)JjXo0kK)!=P^ZtO{to2%zpYwtI2ZO8*41M6}Ks^t+&;LMs z8;=!C+ItzktZC%o#Q&=^-z3|I;QQM*4SpIP)qd9$`vB69fLE*YDG$4Nh`)WG-481~ z#PPdiN|9}2gmb6Q9XDWqW3%nOEZQZQq&v$v|3G@(${v?~H1}Fde>xbK(WV3X$+Prh z?=Ai0EgS99k7U;<*YA<{5q=NVf0GNd*1ctGW@3y1V-a9%XLx}h%cFjk1`)Mx^+S8hj)_0yHXNYJv3*X?&Pfzt=K!;|*OAwnKnQ8r*aoPFUM52F7 z$K*)Xx>$2*_X_!7ZP^y>9Sv^$9%Z7lYH(i7KIJL8hK}?r8d7|==uY&fF%&&rL>*yb zqXO7>Mb{2t{h)ATHS3l2_~Gh_y{$*Lj|!TJZOHfRb;$YgKEC2T13VXBVNznQI#2Q$-C+E+0;2(5a?PC`(Z~pY`$*F5jt9$IN+dtdIezVOl@i&O)gNZSk z7a5<@emJdfXm3~>@u%!XI;)m7wkS9Lr$*D@#m1Y8`o_n%=@b4Ynw857j=A|F4xi+R>WT`%N2og)CvOseF%=$4PC zicRL}a%1dVucr(fSY^3+^erxsmrt?oYwJQz>bI(H8RzMkiMp<-8=RIb(w-#A{wqh( zcOPwt#>&B4Ir5Wroq_f`a}IgI_Y6!xQ|;6vSoNppwoP+x>R}%(d2n1#S$4emf;<<< zi8Au+DlV}dxikv-TS5Gy>M7-Xr^WMGr{Zjqq{+89t(|YW?&W*MwF!Sg*2@27fLZ=W z-6M;V#%$KN)x<|gc1aGj{SX@tIt}2f)jU=_sCg{%&*4GyvD8l1aV5Xw$TbuFc{!`p&8`UE{i+lTOlwq)^@Fe{`9O>sBNuVoClq14d6un;N$Y0Bhwt-bYhH*e3!Qu z+I<_}ecm?*ikXOKo5`k>9rfwoIe5%s2YdFJd@NUjZ{z#tpCc>&Eg9lGB4k8TV}LIh zoD*kpYm2Tk3Fhb6Zs9z}20PA};11=(4lBPNeYDS5m(+g5FnO-RtW6So!5lT(@k+;7mbD}7x@z!10Dp@;Z;d@#Nyd))x{~m&_$~>Z)VyJqpZdrtv6}QtKZ2d)N3zm?(X$peZx+ek->(}YFugK@Q(DK_wKcw&fKtctRp+4lIO^< z`znwfyc^a%Ys{jTXgHr=^-=gZ%abRrrXPU&Ii|f)x4*;=91e3vhfa9-sZ$L zt3D6j_g#34mt#YNf92(Eg{R|O8#6u?!gfwzD{K9yr_B27g3HLKX?=F#ErYD=BZrJ# zPggxY-ybv@#P^q>9!DQkS-Pdn)?<2nf32g<@=p{n_TVqjd~zxaj|)~mydMW%^(9X1 zhIF-a9qPLNd}mz!@3!);)>eS)foP7d%F@RGvbhRa%bSlsSE8pJo#px`AJ7`G=32(} zU5TEu&|8>sv3%mr@9g+9cKg2U(yv=?`Dc;CKXacm%kQVZAEFy_`}n6=|9w2GpFTcF zAFVtC4%x8A)k6c(rQ&r1FBdkkZ$N8lZ#Xe0(`j36dUizD>)@Pu>xDiZ?#Xv)=W4xs zDKgmQ%GH>9Mjsqe_Vd*_OEi-{ps5C&E=Nw(fN$$E<2vX z3eqc?Q^cWfkJrnd5Bi*}?dQF)RBG zvLVhqG{83j6IkZadeHg-Ih&dG`tzY_-l=CN@}X7W2b4YYKK(00R)>+*wvXUQ^E&6c zAe|WxPTjhG2YlW;ve9zsCYO6Hfuk(*J6#J;&KVoMaiWFb>1Fo0%jO$} z-P?hE$k{z<+S+*&_Qri~T!!p<#fGycl79Ca*cs??#n`dFkyf0>Ud_{?3HjEa#h$6f zceNy0m3|icVHJ8|33Er`Zw@ktJyetSA>;N_=TZEq>#ZNPVBVX|4^N{F@0qhUQ()(i z>s>#p-cjtM_s*NtAzR6RPu%Y0HP}S2yLQlXt{vq03AD#=sp&j2E|72SAbXE|(*06b z&OL`6wA8hOUV`2mT|20rn2mbm$tY}}`oYHLZPvBU>C~_4+Q@YsbhB6boOysVAIP)w zqJMh$_}CQT#fwdmZ(ucRC*{zaY&3fhY`SNC%F5uyz=nUo+GusaCOdK~^t~9^76V%` za4ZHk=_6fhoNAd%LtCj@=F;-vzMc<%x#Oj=^kDb!csG4$!|s>Q;~Vgi>iUPV?p!8* zZqHoiP3W%AT;^LL%db_?^5Ox`Jh95<(**SA*vU?ea-j(wcp11Wpu3ffRXy}l(cL|< zn6`51`+Ka*tVB*!A}3a1-&>qAkx$@6Io!Ey7KBEt`w zt1nKKF&}$xo83d)V1DAn54^*@(OT-aL#Z;>D&FQ^ZOhLoJLz@SE@boj=WElgf4kQ{ zk_-(aLnX`quVqZVv$y@=J>bZbF}b+;95}UjWULg+q3bH-aVhjv%eY`Suz&pAR0TNn zeM=efb$n~}plh%BuhsuqzctIXJ}H~nv)`6tpB=XrjJ{~YH)ZXO
YNdM=cXgyng zeNXRH8fZOIB54|CG6NHNon@J|cCV(?|PUv{wdgWk%4i%G4k+WiUdiDrZ7OuZ9eZYds&ai4%5I_MW`f_8u5-SkiO>A7^Q&cCQ;tSaF9 z8p&3d@2h-#@8R-a!KHZI85?K5l`TO=mzb_1@+cNzw+L5?m(kkTR`^`ci?J;QuV=3| zxcw@jT=r@h8#BXu@--;e4t~XS2K$Y1Xx1FH4V|HS%VqmzY`m@AXRW^LyRpbX*=q1* zx)NVfH#$dTVxmJ*nK9rG8w%eQaPmug^Vu6~-{;#RU=aK|i^dzft<*cmh0&vhBEHjg zyssT>Y4-|fqK5Cqt37J~&}@s^foDDc|3Bu?ujBvk<==ANEoYqd=glP+eD?Tx%p&Ub z`~UstOJ))Q(Z){Yqgd?BlXg}yPbzkEX(D5YsV=trmOn8Oxr-dlP`Z8I)yz--44q46 zA7;G&n`;l=LH2xu-ubrL5)H`DwT!hx@q8w9d&_r5P%q=$8emO>Ts{eGy-=97RP5)_ zx$M74nqB(UIk}4G&0`*(7e>}z$vao@cP{T$Kr7VU293N++>v-&@~2L70>(FsybwOV zB6{Uo@>Au089F6L&Q|fR)#Xku&gx)VG^_C_e$tu0 z1n`N7Uc56?v(QTn`b?M+tRrWS)MJl{KKqnIKPs1{pXi2@&mWS`_=O+tlcUcYaA7?ARewp zUSrSp_!L8wAA1nqfu6;?@>wcpv&O~leTIJ@Vtn*m<&FD;5s!}LXY7EE+vrCxFSo(V zL5r(C<(8N2x}4?ZD&WHgnXEC*V7q4}L*U<9{F~B!GhIG*eVRKxpQdo;jfZD1SLcjF z8GKyg+RHjCO7@NPX0C5ieX8szw{x({j$pcy{*@j(dxP+#^|h?4*UGprLa(7W2NvOP z-ilvFIM82|ag}}4&3Z|O`o04ExzD~Gb?{e%yr^Vvl=ZnYCL^zL-u)H7vUB9G>PAL$ z03q@l8D}V4WHyv8;`}nVJwJ^G93RI{?Q!wbRiE-4y7~q(K{`QeMOvFeAEvWw&>_qp zy3r%Qo9)t^Pp0{07;F0x(Uk0_wl%DsYuxbJXJ?rReidN3RDSvG*eZv@_(SnsZ8qev zFLTcv_rtHTwBI2A{P6cF&w*cZJOj)L;MDaVctLe~w5T<_qpZhN*gejn1D)>tQROqC zlbTTIK&9p&@a#(Ds-8dY>RzwzYgPBGXheQb;oU30$}N|jtC%If4earve*54HTs6Dv z?*6`uHxj^|VChbGQ*4TE|qYHBNXivH2y0sh>a`z4 zr`FD8ZrTkl6+biud09oywJQ19kZ-{-xe|D$+}*1J%F{G5B|YTKX--bP$-WwWQ^_~7 zU$fXZf@c`lb_8^^5I8ZoP=y`fF`gKYg@=Z|1}YO*m|0fMnm{A6}vOr~cw){t~^} z`9FB_;=et57-#X*JAd-%8=cuKoq4IV&e>d9YU5kv@8^0zdPca94z=&T$h+FBC*2@? zR>C`~PuDf2zCE?vi!Yzyea3sIfw!)7dCT7OY&g$kJ9zMT<)=_?>xO@~9_!wu=b|;+ z2K=jg@7*=LtGz+0CxHCadxh}(C&8o2mTa7Ua0wgYreTW zhF>7WUfuHRD}XTB@1LLhu!sqBa~|$ z!>$FgexN#vsUyld;jWR`2i$w>gnqhVoq^oq&VJYN@CoKwf*bkM zvvtf%?=OqA)22~AvD363S-7zkd}Rh=clw@-{;G$?>^QUbwrL6DsI`%Fv7+at%+SWiY-~CD_mdy;oc?V==jML@mT4W$4Q>@H2A0*{pT;ao7CRrQcB@{3pm1=l-Cnc5+DG z$$dut;8FO{@%P@wz2LJxB*iLBbK-f6plz!U6{ElzL7#!nJ$jx#s;_d89JKsv6u+<@)jX*_MtP&43G;*US*C(%X5i-qyOI^z}_XeXV)5 z^tGNX^y%rFe0utRz89{98^)lyK6!Skjn0e-UYs@>dgD!=aS?EQ1Dh1 zY;yy-dBK8W#5N}hH$}&`Ftt-UlgmjU7#(ZvR4r%%2{)r-jxb||wh0I*QQI_PYsOAR zK|s?>1m9_=w&OXu0mfFKw&fHs|LUr;#sbz)l_a?N zzfgOi06QaaTkQdG$egvTx&%+*&y(O)DgRZLzxQUwYtB(c7ZQ#%5ObCc_a?Q6EDZC_ zUFnXF>FC(W?=~d$ZGKYU;1Mn2*_p(G{;I^`8M-GnQ7}?$tIvU_i5(tTlWA5g;XBly z+~!Kkyg2RR`sHV|ZLd-N;h^2_FNi60by;0qewA+WpAW?QD}Jf)S@yG$6FGvs65R4R zhw>Eqbw2tDIJ5JXk6Rn|(4OqNSNVMkIhd?({EB|Yk$wI82Ku=4jRNYHpQnKRD#D3E z_AP6TBcF8%S08!qQuGntAEb}e*#58%)*-Tb%pS*P>pgovZ2n>T!aBp6LtMqbciQ}W zCgs=iu51CtO<1%Yn_{okXq`!WX8QLLqgR-H#K_}xzCq9F-$y)2^|QX3#m*+=|wl>^5Je@V1%iw0X1J zrnBeOa@w3jc{%N!puSy_eW7!sS*&gDB1d&h_s?dSeW5?(UbrP)?M`rHxkp#?;s@*O zgX|qYI3fE6`t5Cmehazpgh%72W__pnu}^~!uWA0IUpl&Hrjh;rrYYdWr%w;YU&egY zdeN4@k1LP1`grk`(R}*6-sz*{uliZ&;IcPAR(eu?EW=OWnc$XLk19>GrW>3_2aCYh ziqJ>Ihg_e9?4W*`;PJ@}Du-WseKNCxCuTg%809NSrq4fJ&%BB58NcZMzlt5Y*Zp5e z_wku$(tRyu@1GhBJ77iLg{w>s!Fe|9uxahs;#r>g{rv&XKRnC)?lAuU!RJLt2Vcb; zcSG0Y>N&c2{TFiQAX{u+qu6hMQx|$^ZS3RGg4Yhm8jk0r_mF28yEf}S}9rn4I!UuQWW{vW-!i(#Z z<8A}gH*FO@a@u}HvA6Jxw6#O}sg1s{m41jHmL#n=l{o879~FDM8M&0Km#f|482ciC zhmZL)Jm1yRDT||ziY6Z98TiJ1IJeF-$KvnDb_d?f^N;GivB6sS!S0kdzxAiQ9Swdf zdYZFs1cQ!8;dz1^eofh)MpuM(&{0F%@OjU^fP6U5V+%81<>(sa=zx1!*Xw*99ptmM z2a3@_gcIqkL(WuO0eUL3c95QmE_ZY&JvBa;wIso8K4W43Lils6_0MM~s9>dai!xx#Gc|zR}i=r9W%@kDyC*Jc>*Kf4evnS#fB3-`26wpFeMG zA~SzoPTZ&F&z-w(<+^j1(e%qK#`5Hub2nXc_au7Vf`Pg7!byEoo%qk|wP#{Ay2)N= z%?H0-{a*G!V6(05fd=L<7kklJWy8IL54{JwOLmC*Eqc;*J993d!^eI`y9?lZv(IeT zZA){b_IrTAe&kdIFg^@y&Y`Y(v=zGhl4$7kJ8MIUs)ILjKdI!B=p^nnC%2Y1T?)#LJ;udG9l zrQa^D4Au9;T4z`SPju?j{0UaG0_1=NRwh34yCy!f-o$4X*}o%;h-r=3zsn-**IRG@ zE?>faz1!^Hi!0|QdKTHgm)tWqaj44vU2*T+#Nj;q_tIr^6GsZq#QC3JXKcv&EUyiz zwRU)eAAas1yryY*-0jek=5aamD4r=h&i@a@%TtyOUX^N|fsy2G0nZDtYo*5*p~EZB zC-$KkyWo}W!8>>Ds*`dfXRyJ&YgdEH;e*~75W!M%#%+)NK5JE`f0qZ`;Pm@%==bYof$s)T4o4W7cfRgU#xf68iVMI_zmi} z(f{pM*Nc1EM;PM0LFM#c<>VPNbTj9+6JDR1f%BJ{8^N~(_^yDbh*yk8?quF#89MWh zEk*q8yU3d&@Gt@mt;er;8~WLzaAGy|*NA>ne2J~k8@-9S%KsmPz4)*hGjqLF@zkgL zMqiu)-G2(6=IvJ%{mP?Xh3v-!e!J5wHE>X6E+5`k-F-s-tU+bSA=XXyzYX8&b9l65irx`_ z-aFNyQRbjFd~vjzwZ}Q=K#Om*UYG+OY@cV%2qXLA;LQs7NDO|YF}(Zt;82=$Cwp85 zm-$}xnssr;O&ob!9&Khd+Wptsy!u{RW#@d(Bu{7kPv9%_iS4=7uJaq7$;%ea}lYS!{pm*IhoyVQeQE?smnZ$r_=pzmf=H|PEzr3e(O6{s50{heG4nXsnORjec)hd>CzLgWACU>w?iMq zTVMM&dgeN#XQrAx+oBI!zf7&fhqI6vc;olee(lNZRoG(S?wqaez>@vCts1ZXW7q-M zT08vBEI=*3IrX8x3|vr|8|Q$WJnEe@ z@k7!6ffeBEJS<^ z&Kzg2fN0x0w=Rt1z?iw&nJpN@7h(~A6eY!m@%;aHFuorc$7j;svcWJOE*Q@o0>;b2 ziK&7y*Z)6ZEco2+flqPBUZW5lc=j*Xd-pG24^9z(X_tY61>m5&e|Z|u8iD@`=nVdS zcifRbn%{gU4g{ZkeHJ_!zu|t(74P+)wCFQ7pM3GE z)BBE1n?6IEH6Gg!8|2zeNT2bxFRtzstx8s`rMSB5r97 z^z*otxNY>%I>I9265O~&^=TZmSUZG%)?W_XlJzsuo_P)**E&ZhKCrGe$fZlu#gE&E zqV2bG)5rezU>@6ac3yO$`r2@2KmSqvY#91G z{Kf6oFRqIEhz*+heZE&L0Q*3?YnUmnMOqkOp|IPZg; zNSWNvW1d1K)1v9vX0M}LO+LNg;7!LKIOy`f`IKue?|_bej0`(T&W7QqPvTEV?|lzB zCHX~`UJZE3^?wB~&mwc=5Bxg0 z(|sog)}z-A8|=TAoONuDYqVyg{hb=Oe3v$k+wo*G(RH$&we}p=t+ie{9zPIz`hM2c z)^h%nS+{Ortvyc6=WD69-ICAzW#K>3y2>k#j&I7)1I$_rI(`;?$U=`550#!Acl;-! zDb1m-o0)&<0DIp6<`>{UVNP}L?W4}FK$ndk@jEh-^DQjLKjiYXSE$ErKN~$n{9bs} zUzQ{TlHrhiFBoS58;}r{jb$sHn-s6X?P+P^=RR`97Z0GTbzx>(pirA^+Az#`X zN0#s3ba;7a)4qj)M5-_S4d%6eNoPUpgbZu?{?NWLsZXShRoT*I`H7{Uu%-`7e|^ln z&FN$D6ZNj-3=HM&3O(@PLG#YN`H6{3D_+bSVI9-n%OmTpT-ki@f)nyVso!1rXH4D| z(FOPRJotE<8#8<(dXJF% z+&q_`g=dGBKX@?BmsUr7$=W_^`0apwGtbvds;ZL9iQuBc;i#hnsB%RUU(Sz znSHeAj)#rUHMLjsBt62E8=tE^-ouX1RnOH|J=gx(^w59}_eIJ(9uMcLy&m}4k@ex+ zcN@aF8karC_|`L6uLPb;@I@U4*2B*WKQ`il*#`xyK5InpByiR|$8Qy)T(|_T_QR^= z{l0}A6M)GkD>s&djOPqMTlY=ThQ2U z+S0`E59eG;AR4WqNARrUf)@eF+1#7)?HmEUl5G(DSw?CC*PiM9oE`4{wiF+5AM zWIN+9dG3L^@)OzP7<_i6-Pcf4sbY)pfeThe{Qi^t$L2o0X%Y8-;Qsgg8@nqF_`;7) zxA9F{IPs2Wjo;-*Hzx6atz&h9tLmS45dK?}Z{!x%NM=J5bHMvF&Z?3>G$re`jl{+D z?jUAS<$Ixr#a3#a{NT{n?ZwcSbWr*5wLdA(mr}WcvLYY88{ne)^?V!g<*$(kOn%h? z-eJp`lh8{yb0hyx4|3*puJ3~TH=zrtop+ck#dIl0h^r^KI#z?uv7!DOS#vJddI)kN z`JHR0OMbZdXzVtV-?>8XP>)x(*n7HzO&@aRf(vWKdwTQLC}*YM|1Agp?w&%$)OL-A z=fl&#O~1YJ;fTo>GY}uE_`go*!^IuVht7k{Bktl6>(Z{{SeSL^aB z7mr#(ek<{)m#wth8^z;%HvNkJ>-tNs{pYPBa}~p>GcR>+;tPKUMm@qg;I!Z);M4;j_<;i} zFT5t5CwV#$UtR!>7FhOqF0VTF9{Wk^3pnc+KlJ{>Q^f}rTl6kCn~nS{AdXDsrS@6? z=Z%;+vgQbD0mQbLwE)R@tpx}NgtNL2d*c3Gd_GF9(*->W^#d+KxYw?mZ~xD`(` z`*e0AU&>A`I9N+wXSbbIz)0hf9ko~d51BOE*-xW2Nb$c#M$XtVgtHwvqi2^ny0^P7 z*#%C!^}6F+NdL8WBrAme4LrPKb>TB}2>8t5I~S(60@I$o>ch>k_+H!dNYQ`nr{$tPBy!4U0-_(tDsX?F6+J)x9;3@Ok zLA!o@(K*O<69btP??QRI9UD0~-es$GV#cLO{loTacWu2LEl%9db>ulFSM4_L)vo5{ z^Yq`{Z+fhd^Pv@wV9vN0ed2Tt=bYzQZw86!2!fOHh0kRD&bwbxd{MO7zTENM-im)6 zUa$Qkm|B&ka5m+PMG&lyw$DkKnMr7haCwoz9ukLjur??ekqg%v8rr;@$AK zb-{m>Qzl0J@j~5Oi5hY;_EiPVKD;;sm+TOI zcL*Crx+zR>VK;p$XHMMyZ$mb02Tsa&GRLxBP(ENidzpTs_a|2K{19^MHSU!!Kr*Zw zz4(xHR_oeKfP}AkPHtGarMJ=ldoeyz8e0)ev~KACszi1T(O`5@bs(7ClCMj zDTY$9H{kbO>=EXij6@;($l(|=% zLoU%V4ju_UVW;dO%8ZP`#_7OD=%jyT^koly`HtdPnVY!hnP{pk@I_+H*yF)CrQgPx zo9Dm6xqIlc?)$2zOLI=0d$0r5rv6>r`s38T#|?c_7q;k(emz6&QC|b~#hf?|;fIUs zrSmz*4t$f%kPkhj(!SuSx=np^1L#x!{`<&$hc9fh_f(m8HVvFn(HWCXb`#6QQ7DLNZZIkJFbb2%|>=*2UH?~)}c%*7SUplx`CvEycNg%^si z#F?{I)$?;}?#NHrydb4%%Y*cRcT664OU&BkEel`={Ei4dVh7#{jgX3+z@DPs4 zKPXr#4)-x&Vsr%g-H?acS2mWh28mtQyrH+R-9`+(^0DQU7rFx*MCYP!foFpY_)w+q zNG?^iy!?Q*K`_vF1LfhCR?1D?yd!v>BY0!46oaFZn=>8QW;?Le`mXrB){16v#$K&8 zaK_#&V5@uWwU@j)+di-4ybQ({4L)l8E0g7!nW->qPN&>JQT zuigCx$NtmA^dEK4DLe01z?;6mPj2R7@la=f!Elw2=KOlf;ia72VxQ}mqH^FepYq#( z2@j?|(Sq>eZ2Jq|a@%zF7kt3C`sP4Vo4-{#ZBFNE+I03899B7PUdiOA)4cLQ zxniUv<&bl#@Mv{wC-H8D)%N*J%DZ1c-b2Nz?Sj7Ld+`xt9fv6yxoal>8?cHJazYeSy#Ly!(kICn4HbCDc(0AZyf!XKf%e3{d zFaD62vEkOvz}9eXAQYMrL+6yN$>xlj8PKuI=cUKEI9cZKdAF<#y7KZM!NB!}eC2+4 zGWz5IN{Lk!e|islOl@V+PBvxjo!|t#LOerjw;F@m^Eo_2`z@qPh`;0{@#7qa zA9Ie5;TeDF-ql6 zv^kS2ZT6qt`M%0&^BS%TY44C-U%gA?-(}nyvn%U1{fAB8gZ96g{Wt^mM2!A{TlHRf zFMb|G4!wP!W7E^7d}p#ZvM6yRdQNf@kyow&GmH zW<&2mv)12VW@z5TW!v-g7CeVO%HFF*_jK(Z)wv2-sN5}Eg^VnPAC-|$4j7nxbraIS zPv|U0xpW5$920-GoPAsC_c%1+&^$iH)~}G0rqSfon1SB$K+9v|>lxwJU%+es$g_v| zx0G|XeM)pQXXSnFxxqZw@6f~zN&LdA1Bo8W=t~3paCSj&k%{^p=q=`VHL__1^1K1R z*vc84n}Dni)BmntJGQ(-gY=R8uT#z5!mFJ=h>qPqZ;XR<7WShhI_3`M_0kihV+hZS z?}aDLg-)^CyJX8qPbswNRy3>qxQdn7j_#)OiMC_Uj>qm=1x|^7uj=#Xl8@){PQH}&Jw=IcWJ&a;d}ia_pPBlH-Gj&Xd2f5nHMZ89Iv>vXaCX~zoN&yQ7dzm^-h4~pG~fr#<_~9# ze@4#5z?+i)q5YV4|Lk>7$uQh$BmlCl55CFwX8c%(7h^&f^8 z9JMBnjBaMX^P}(p;m3Hc#fqJ~bQ-?1s+M&s2cI&yR$mC-W=^3TACAi5KbN@g&zf%D ze}Mb@s2kmmcSFWL3qL#HZ+nwDdCSvxFT98BKXf!y|MT4bf79-Nk@`dbe-97y_W$ox zPXFJh+}r;JDyRQP-1nXS-$MDkr2dcc^nde5^nXW^Jt$tTe}kv2zmW9m@hoU73);$t zwhEvvzsHX(eJz{wgv|S#C-mUu1O5fKjx5T&K1bohH|dk|5(<|3bMaxR$*(Yg5AtVZ zeVMtWPlBnpZ6GBCd_4{Lq7>J# z$5YZJX~PNKKX$lNSq?lwDmb~QBbb9|O7p!*DL@y+n&UP zsebQ#KF)j^JaOc|@Z@~(te_n&nXbAh$?4jHW&HYcx<)-1}*=L_hvV6XdvHWZ(ehlCA#uFKRg1M1C5og{C zBCHXz{&)jpATHkIMAw|enTs{(u*vlN;VXmBuS)i3s{X&M-IMR-_^vVUvDcNw+U)KS9vlfL{^J zNf)v~Hq9nvL%ER+L9-`CalY;|J8s3!J;}PWkqwITkQE~IlMLB_Pf-3 z*9XQWtq+WI?A&pVJoroVHXE2sDb!xfoqf!YM69*z+^FIiV-3V@NWbpniY@wgz*ckD zXx3bvxT->Dody|Y=FqvU4zt~8#=yMz#&a^B$2G+6TBy;{2cpUZMnsClb12OCo zv+mzj&}w|-$XUhE6@x#gsJlck@AK_I4 z@afv4SDvxQzP}v(&%1|9d_=r`C_nAp_YCkC>8$3=2lPw%^8N1&jMLkP0C8shZ6%%c z&|jY12WTw@KS76My^{4`WZN&{xw40sIW*|_cFjKUZHzU}dn=R9Qd$WO>MSL4s2D#_ z2|8!CqjQ#`bLKfZXR`igbn{|lGjr;~_0Oz98XS6HqNBGH=YcNExnIZ&`ZeKpYeTh> zcjRk1Pq8ELYw8xSvWTs6am}?u@GDGr+x6HX2A{qp+(D+RoLtYwx2ShLYo7I`AHC*T zU*fEJDlSgnil;{Gbx)J$L9xdwD`nkNe!wDpaUI}($IF({on1VRGhX$j5j-vs9-HxI zG*xQ9=*x!hUR$9(d`@{e<>7bnYy82HgDx+ISBqwE5#K;g&h_vp7vF7rbO7IDi>%g0 zsx5n{$N8{ z)#FX>X~~hULh^y{w$}sqlM7q=mk*sSi+oc)@=f``Y2OH4jUOn{O@HABts29nhWAys z9&pCs_W$)He|PwT>Ql}bztLXgjy(0zpWBt@ zmmGfqowePnNR6(*=2*3VAV<39Sh(z-S13PQCbB}lwnD$PA*|dDfq!bb1HY5xocOiQ zxp)hIr`85~Lu7a5HwM7@R=9rvC%jKB{&OtN?!8<&Pgdi|MF)t9(Y=qvmU})Z=Qi~@l(KlChdL( zxYB;>jh?bsT{2}!@0Kx|27fcLmB++gQQ3vU2U)E5CK3%{Yc=UfKH z7+T7j<-pp(AlWbu8n0#TAUXd^e|a)rroB$+eGGljRcC_W1K(MTkKgt?gj>s%SH|h* zY1X;5HlX_y-WdZx~RPANtl-mQ21drP1ENc$sn^=WUQGWMT?1j6Gy>N5b z3%8iPaF?+cZkv_bwwS$etJw>88GGRtvlosvyf>a&zIW61<*A!E=lhLS+7EYWAQ!Lc zS4!|JbV2{Yivzhq<}U>p)mkZS@hM@m1~i^EpslPCjAOl~n>8TCS@y67q`hOFHK6{S zBdjAdZ)5!?X>GwSL*_U+Uxr!(`oF3}^4N`q7JqYn>5trvOk^Gu7i4_tz*J+B9xMB@ zJ5Z3j4cPXi738L}*3v`VnP9JcTjqQwVsbA8|4ppc5Al)jvVG*QK>L~pL&L1scT-0< zv3u;ptKVx~9Notn>LJ!lwU2L5Ib&D3@_i}hr~3x$&GH}84`{P|mz5}|4TB4um80Lv z&8O#TD}%FF;>>pmc{c*z(3mX7RFEs#|2BF@H`kXO*}vVfFO-MuDf%WE`)ZQxf7Oxw zuR49%?2-Lj9NGU0cp=%pf;zHTV^{$%7JT$u{fq1SNz;8Qi!aNJ-e&5;CC4PCVCe_5x zPr|48K}~2-K0nc6q{fMrWqd*S%2rFf&#@zFLiV1W(Y{0w-cgI}UL|^fF68qQZpkj} zI>fxaZz-=(w8~1YjO?u4ScfScvS%E6<#^VVx~@amCC;~6w*c{&7yKB`($#3O- z+64?^zq5!_ec<3R#pL}ioY1>{RT(=@20|tlv2QqM+?`~+ef$S}ta;`W7mSZ4&6kMO zGY32uEkxkMyV=4f|X^pCmdAH$1GeUS^oJGv>~9I|#iIm)*qK0H6Mc@*nw z7fDwc7QGl>WE^<;kn_?dRBs9O!lSNHeL?2n8DatSI}6_FL$=ts<;=cd>z;^9n+Abr~6*$oMsu_Mr+w;yl3ow==>D#!8hf*kCR(i@9E!y_e#fH zJ)U`^z7E*y$4blj&G-ciFTAdIzWcPNPH^1hwDB{K zaphN{znX7!w>SOFXEyWMfj*G~9+t94LpT@#2ic$3o7ulsdvP>|jw_$4>u-36*f)(; z-*2IRLVL>^ea3tLhxcT6dv$s75{=EqLt8)X=ar7Ws_{mcAK{(-JGC{(!A;?u#+}WW zvhA@%vo#h6k01JdpV=Q>3cU0lFf+Jp>S6A)!A}EEV0gxJ@JM*Mi8*w2^`JAitPc;e z2aY7!7x`xX;V;@>1`Pje+DF?MKLs|wp-z`y*!2zIM*;P90P`>~&S!pFrc)ojXqW!I z@DmRHTT*>X&Hm#7{a${gzP_O01AcJC&^|Cl?i)N3EJ__%ot5sPH@!EA?$OaZrl7-% z?xE3l8B-^6^o;kwA@#pNbT3@uJ$zZMdQbl@&8-~={!^cH?K5cErF)l_MfcPxxvlkG z<0D}VW`13|hbQR1#7p<1Kbr2-pe0)dIKE4C2f=uS@R@me%E1NEw>SPnKD%mkDyyuZ zwGsJ3+(5k*JI`MKcW6OAIv>x0|C_!tf2-ct_*J%%J0YjmET|VSAL)JSAPHdmxhMlvE3n~PwpTe$OpTJ4bmqsWAD}8=1-cO2~Q2P zrtblt*N{Vd4Y*;`{4iUeEJAOp^d-9P@+H=QLx+jE4y9Pp!$&zozH_jS_XM#^((yis z_y_5J_dDflrmi42gLU^$utU}qv)+P_?C=fY#5}Htu48`**EQ`yo%gbK2{BkJzKN`( zUlZU>OSsBL`Zhd8I>w{U^%3V<$92uq`H82HGiw|36BS(7ZOu=BzdN;V5GTh%2RR&U zA4eLxa8=96+pG;2@?9`T}LEZ(8 ztCRhOs_XC<(S2{t$EHoGdl^67<>*bzq`Tc(nCL0WPiPPS4)iW`t@>@8?}qLW6~EOU ze(kgEzAHbmtvZ}|4Y=9wjE`>Po$dwrA9&9$%kD1|JY1YmUWUcwM$f|sLtWLtS9R^Bj(ya% zA3qNGaIAAw^+v(96F90)Ui={9Z9eeQ@H+6)3;Q^I z=-|DFiRX!P-KZ;lT}}J@z`a+&JHcCXEEp)CNEiM8s^*e5qVOg8pwL~9k>_UN1oAYj z!3JIn4r|`>thC-ncwEN3~#^{^>tzbg_v3Ouvi;vj7z*Ne?<(B7>v-B_fQt5us;~Au;*L(iuMe7TeFFNQaF1PUTGWNLz+5)L3X6WpK5_0{7 zm(N}=KTvJvZv4MiuYAgSR!*Lu0-kjS0+S1tKe%3J*j9nBHTa3uuJ+}Y;Um~Owev-_ zyB)vS9xKqal{HcMIZXZJ$5EX(GT*u{CLfMPJ@G(Jlj%EktfY>UfvhI~sFS9i0QHnw zzK8ryJq90ZYyCKeh*^f*t!%^#-_8x@xEuMM5!1g&;xpRF%b?d`tnct;J`7`84 zG}w5>Bd35IIi(e)C-iP)hOMs(2fg~2 zsfYCibDfD@fPSTS%r~qtbkaXR`+hs%;mP&fE z?VISk+!sLyozOucbWq4M8$ZZ(MLn+(Ymfu(&jR-q3q7lAo{5Dn1m}yoLil_+<34a- zOT?`UI_QKB3Za8Sp1F0w_osiysVfZaWKvflbRu<2J_ON`e0=1$H@aLyE3{7Ee9(@c zcM5ie?6q0=K|ErwWBjZ0P6o6x_O$Km&KQ*r9X{CNgT4*!1qSPvNAyivIB}N;r(AvQ zb#P;-c+C1wQ#ZV3^;By^vElPGqEBV@J)Q}?t#FR=LL~0G1)O%Cg|i(!Z})(n_YCd! zIBmM^+2@N6wnrpGM}fzMnKPo1`-)F!PRojBC3G#HJInZZ*()lzQXVwrWeYgV0H2D= zA`6(q*c@Vk1HBKyvokFJu_sUWExd?2i(;wDA3%RCqYn!=-9Fba9*wbtF=Wwp-{yg} zPxqZMN#lxxGrwU@Rgd`dR~X~#j8Wq|R^v1J*V0=A)ARdvVADtP`Uh?%Z@BcK`RE3{ z(EI|c72I!jeZtZQ+;y)IXJ~XTsBV@2G}eng z#o&=ElH&JP;7`pWes9G`tfjsJ4A16I)wn-C*TyH&ZOP+>T0ggWz^1}P7k%zV4y0PX zw%sFAHW}WqDPa7udKQ}r-tt}X6|?Pa&6CB}gV@-uc6V)uSb zKF?1>zy79sb})`gb0!WrG*@Tiw3n$qVBCRE%zgLtM&^QhJ&#bIxh_f*?j4mcF@juE z4%0N+X<>a|zqMYX_sXeP>m{b{1!<T%JRuR!+X+5IkY>ME(U z`)&4qz@JvrU#&rz{!y3soqiWboN*#kP5sDSx85Dpt5|*W4Sg}!WrOSevZvk>>d12I zy(g`o4&AxT7ah(F8+GSk; z;MS6Gq8;9FDq`7tQjlX?)2*G@0!?cy^*ycfX3p15rOe9*+_fR;$f=d!9%I&6``3q@ z@xRLX)*Um@7194i8}|7z;tk=J>;94NBCPG9e=F7|b}KNOfxk!f@q3vk{=&;M^zAD4 zL`ip&Zei}>8M;H8uUeE*oFG-&R739=+`XzRZG7b*o)xi z<5+?GT7kS*4%19zU>$nlC)Ee%{RZ|a7+hmM8n7ddWY}||^Efgs=7+k7KW;sb^;W^t z8($|GqyDrI!?79PRO~h@vF1*6f!pkuYr#+Z>I4fne~b31EKb^IxENSypCM;Mnzgan ztVeA}x6=1>SdV(yiA7ljEXsk!c*d_iFRHhyFg4LF9>{!($0$Zm@ee-mQ{yUS9QhR` zC*;Sh1@{yS!I@)G;ktO7zRiGFybG-T4!%bWz6W}DBkx|MK9>ga;G5OhX@>7X1L~jt zU7FjI;o+a)HvF^qFu7xFd{0Ysf^Q{*c$BYqH~BQp^Fe&c*V{?X4O0#;NN(q*Ovd5# z{W9Qmh`H2UXpL$w_sWF`3=*m6_xqhOEoDr~Yx*jDf4>8prND-qrYDYoHDV#F~4npe)=UgOsy0v@l&-AY=Obil1_iCu2su? z*jeEgjdP{u5P2;b6@gdTd#3zN*Ym6heDM*#qHjv8@`c|=7n9@l6}~aQ$>mxE-|^!Q z=!$4OsrU@SEnDI7wea{qJY{XLzx5e;B>U|(`1&Z~R~){6w46E|8FaMxUi>@2p(_$L z_|yaZV!%Oll+fo1%*menCcJo%G37|Nid~`eDafCmQi*S1A^5{O%1Hbr4A${M)e*fu={yZ~J_Xhxom&|5<9&YJLmW8=UK7<{E6O z=Ss{8aqeSHzM3yLIQEwl@00lh>K|U+`p=x3fPUWZ4^21m`~&)`)>p2!Xd}bsQKdJq zC*dr-=5c%#hS#8bPK58sPc8dJxRT6kq^p%A>1rj8uJ+$OYtHtE*m}6T-f)9AsqMrhAfxaD${23^9A8S~g@5ym3k<;8d9hV{l0 z%l;12*D!hUWviOGaBL~bjC^>l@F5RA7Kc9+G0(0X9%PF$XL^s^6Rq2b_hJs}IoZqX zWqS?SYtEi@=j>AKOJcfSRylL_Ew_BSUEZp4;C8EBe$=X;9xcIl`@G88dz)YmFXsOl z+F>r6tvl>9Q}1CeX_xb4?Ys|naW7ibyk-K6mmOGm;jt9H>s^b9k?UID=-*qFinZS<_$ASm3{SW!wY3O)?-t;KeQVnCepmAecRk3VRfiXye!=ENL+w91Vsb4G ztiPdGG?y&8BX{#i@-XRl5qWI*oyG60FWY(BOU^ZZx8{=P@5qIIu?sVd?qS})>W*A! z`C9XSz4JcuL%0`2KDqd-s~HorP}fq%YH&TzzLt$Gl&own&nd)r?5mrDe_OeQS6iuV ziqXnIA88|(#NLr7W`J>uhp5lmbx|}v0e$&Nt<|m^-J}*=gt~g&-3IwoP)~o z6MFaX>zU3}OFmckZ~O7W_T$*H^`n2QKVDe92TmH?pFT30#a`5%-#QcjI^B2LZTfhI zcjRZufUbu@*E#rN#G4{Fe>PFr*SFK#kL32+sjE@6^-Vh`%P8i?`+ipr^9;Z4$ieRc z57@fT^sn9Ovd_2O1HHy_(2bEr@f>Ubbip{^bYx)nz;`+)@C>rg z@_l`WY?{tBtoFA`*Lt9(+0;jVBfGraTC4ZNElnzi*G%L(68^A-ek70OTQ*#dT6bKB zY^!Q1cVH(z@iA!gzCdPp|fnIDK%kk|Q*gC&WJ}SvXVx1&kZgFAz znRBBh%>A7jGyL*-#%o~oLz`bNQ8~2s9OXaeo1!6L@kL+(ypLL;40G;BwgU@SuV~q6 z>lHur@J=uPOgHhs1No4WVZG&2hj+q9vKXhK#k|zkJs;ZhVnI{lvDeWL(XZMDPjX*QwMwi?U5+_pj6*C;F)Vex5SqWxq}3!s2g%nU8k7 z=Rm^Or^jhWJ_GZNc7$VwN70UKk2vkDaOzk5)3v<&qnm79Ry^}ra8K}6`+9z*>cWr5 z*!%m?&(GK&Q_B9Bwg24Y zW5;f4j&NoQc}>^Wq*~KU(VYVAj~xhH(0IHMA7=(WzMr5ss%?V@_-l6~mrvr8%rrT= z{k@&cmHde_u>qx8aPX(zYKX*fUh{)S=Z8d}0yjj96;4wYMy?kh) ze>CtV2M=(W0j=<^=0$zTnnw;4bPT?2&LXe-+4f>(Ip-3Z{_?J6T_4@(%!$ej?NhGY z@cH0gAM(M-_PO}{ocJx|OG3FgzQFkPjkm4|-~eN9FM(#o6Mh5F6)y*LWQL zc~cktZh#i1{JL+Ud3TxpZtUj`8yDbLQY>NzF}(UN##J&QN-k~q)sXRd>BM3r?}u>v zg)QA|g{MVJ7$ba&e%t~G#IY(iSnDIalSlg|U`gzQ>|OfTfn1i(FL~R6Y*M^=7w5!u|6t&32(_K{r@n>beb$UJ_8n?1VDY74 z&Nxizt)~65OKJZ?_Gcn5#1oTY@gra%x#!g#bXH<5_*Tw&65F{h;XDbQdBFJxI=903 z4c@^I{0jYEg5DT}N9epWeUrlPZSX?9iym6nz`S=cm*V$|kru3V?SlU3Zz2U)qyP)O z%if~iw%_#`I#oa2`T7_7C|djDDI>dv%gYk1?6v0TWCS;4gEQ4bEoq9=JCsKZE-8-^~}8 zaLv~~HT1L2Dl1pM z`jMimD)Ky9z9#8sf}vuKwYOKgu2;tspShPl>f8>CKevCcoBj>Nd&oW){AIU`R*?-0 zvx$9K$vVAa!mB9DL{E_qv9QaZI}W(bBJSJF(VE)US6Ry%OMdSN`Mu?b-k5=pf_MA- zQ{&jGjMb*mv)m&m)#sZWxW47|TYcA``hP57ub199Q#|s*{`JyWP9RtF76{vOBAbBz z)t4-qWBgD0UA!o@UnfC#F#9$~GQY^;`YcEHF*d|pv)))=YTKvv5$sdq2igj-FFFc| zTj71#Y?9Rx$~yDd+eBGywaq`e0>Q~$Gwt@vXyoNoyc+=_wd;jon$LF-aeSC4` z@bSINXK&hj>T~N~JGH8v4z5^TLUD?e~vb^KAQRvC3zL6VJNk?H`YBrk-MzBkTUlt-m-snos!`DUZ*PE#@O1 zVBh1PV?42(!rVr3%b0mU_jPkeNyjqtf&M0$A7^ijU?JZ@hchpc_gA-e0+&o1F74KZ z*BD<;SnaDF<|xj-8R<93UgJ;cU=OSK&Ufu^j=HwsG<^e1zUTINt=;Dm#$A*!+VLo#BHe!3>4_X&rfxMO3<9ve7`%st@fP%UggZ;Ot(C2!|;8T!>_Ng%a2+?{9erA z5tReWtL*Z6Yx)hQ{!Z@yi#B7oh0Qq_7QWUP^VF&`;JS$C|G{tRr90ieW#a#(Z*TGJ zXL=7`Dsk56qQ16Rz?!Ua#pfd%h@;n>XW>hTPp}eO?eE*IW_ynI@V&+`j4L)O=TzF| zKTp}jQT@a%(7ls%kSa~(YM(9E4lbI z-#tm$O1+cR=OcXkP0BU@#pqt>VsEN$bjIG2qkRkiiD&e&-g5G={0nu}Q>J&8(SEgb zdEw6R!i3Jz5--!g%coXd;`l}a_IgPP-@eCx75`eN^=(DQo`+6@2`!@M8ix8^e6t<#yy z62AQsWi##aW~1Wl+S@Vo~m-@ZzJ{I!f(|(#r~e0Wj2n?=X=w> z#z_2ToH6-70N43GusodTTmbK_mQG1rAbLqRw4XX+MQaZ{IgRrj4n1n+h8E7sJzQty zUTD|dZhgbXhw0QUIJDSsz$VEt{OU@T)32}EPG>|%;rmX;v)+pb<5?`n z-kVrXj&tZDH^XDIxcbCW_Wli88|t)j@j)Gk1A8wHPp5A2@)B|+a4vB!KJPcMkq^XE ztQm&K2gc;aI>{Xnw5~PpZ8Co1`l?O#`l)$`b1_~8hqTvA_hoYnjU8v65p!6Iy*&%N zPWqjBZpS0mE5`vmME++x&w*w3TA4ftDqEH_#_R|5_V;P}8^?!u$q@9onP<|!li|0? zg`cx#V)Py7x7wG^XZxdkO<{5@%STxvUV&X~^a}J4=@nJTmrf_=h-633hiNmkN5`EH zpEDl~lu0+zU;liV_`HGr6dGR%zG~mowogX)#kQZiz8HN&cB3i7PyM@uGyjE-j+SQY zWZmR;m#%wgi1v06kK$ebQC$&q7i8UZ=_6(6BeT-1>18AKj9EonRaQ!68GB0Mjd$OP z-GI$hIiIWIchEoY?jl~U7JpjA3bxsGrJ8wj<4!x7vt7%Z0ww| z#LfWI;mnWV?bV6pU-%+6c_+S>xCq=o~Bi-MafZ$A(p1B~Fa%3hd?ap0-jryT@(I z%tw;W{V1`d!igAgi(g>J=>Dq%yZ8mwt)4kd*;IB9J_Me11pJdD#7dJ(xvq#< z=?-G0I|A8?bv8MsYK>niyQz#=XYq8-FIv0R@;_AO_=IHdsQ$C{r6c6@h4?h}Wv|l@ z`OQr~h&daoAFtXm+olg?vdyPf*FjHCalz@Ss#-aRG z;F+68r@j`tOwN^HbbggHK5rhKz?_x99G@Na7MeMG;9!ItlZEV8$vGN+EMn!i=$ZYz zVZZcZuU*tA-}!x0&6(Q^9*6Gm$A~t*^rE%lofde#bN>bI-{zk8@q1;OGr8yTzUnpj zH{7P(Kd?tyw5zrC$WZvlxl3ozFYt+sfX_lpjh6L->U{PFyB^U?8F;kHiNRcj-(x#? zRLi;bZX~3tnwv?(5}? zSPbm9Pwg`E_cZWL^#z-@JNR?9w)dntZG$(aZKvJWZ68?uV(@0DcDulvuE02hziNB6 z!QXLB+nrebr)a;&3O>}|e$eJUWgdJ{{8ty>1@ZCVKdP7ih#V0wDj$UW>x0Uv&&d;E z_7I3}gWyaV@7m|mDPA|&RQ*}x1F#ky$HCKnd)isUQC!c@i9z|I!HLwSxA=Vr6})>l@?k2z zk6j<8UF(&p{xIi~vp2zlPeza}`4!_%1n8^x9qly=vhP4Vv)AfexP-6l6fM5VSiNny^wtrv^`%QZG}jRc=l;6` z|6_rXcDy$<_d(#iLHq5J^C4_tKDAe_Z>*o(F8;R2^8AC?i+8i8gfAweNxo3Qt8DoL z2UCEK9`ey-jP0NYuXKPniw`O+8R!zeR(0R5HS4h-!4 zfBwo6;4Z!?80gt5U>2u8!UgF+{WuXg&%|++)0f(c@)Od*gs&O2osC@eQy!pCluhoU zUn|*5k`eSDtD#?$sHd;WZ*qVLAL3&ROuUY8Q2(qK9MyW>){AW1)>{8&bU?+5DHcZO z{lHg_CZC(AGM$^KtIka{@1U1t1^p&Y&OS5IZ_Wnvwln!-;;~LHK4)B6qpA;_BHp7~ zytd0;<9+X62J>2ch-RHyboaz_13VZ1BIV)M>a);Cx;Zanu>99NNe>J{1M<740;4qM zcrEkWQ!(*GEwEG$mCNwYtp*lr@K5zYCxW#Lx7Fw_`_Nq`uns%{y@hz2#QXGtJ$I8+ zpqH)CiTqmf@zrQgKD@pInc;r#!hMblcX*NVL~L>N?{R#mH541ZgLQ9}l{v7N?rq)y z_MONe$wvD=wKb0%63OkoL3@U_JbVV6XK%GF&kDS-9T`r1$APPtr}gz&I|9&r1RrJ) zx_9xt8~hbKb++v)&)K%3C6(hxXYI^B-&V29!h7Ytv~?Ed!O5qjK36gJF2;ktv$LD} zYPgq=M|(f@yc7PYwTy219yj`qy)RXBTgKd`Ft^*$y;l3J>F+Ri^YHb0d6PFkr^}mc zx*VLBQ_q&bo21*0pnb`D!P$BU+*FwC&l%bONVD%^r=O8+2&sl9ZG(J zyB#y3!{3ophcQVO2-cGKJ>Z1aP);&e+dZ;Ca$DtId4LWuAPhCJZ>p`QtwJG zC~r12cOiYQQyUeuv7a`yezV5O#eLY}v)~)&$=f-&*EoDu&*pI^r0SF0mW_w(-4o7zgm2yXywSwF4y<>`1{2@XdZ6TzeMY+W7zEK_bZ%%$ z(*x)n#hmwn541Hv|BzR2N{)9kPg^U>kPY}(naghMv+dA}e4z#Cm&TUi-6H6vgXf#U zj~t_$1P9r9(x1$@`gNtjc3`pmh1bEyIe+8p8~3& zwSPW1NS^yGe9w%-V$T72a_(+qU4=Zg(4}l2@|R<0lm_PNOe~YTP<{bdu8aQKb4|QR zeQ_^-THY;r9~>5KOBZaw{}0Y@9ZL+yUGT+wsCNl)&$OalkHVLz>wIv~z+iXK-lzX) znEgKJN|vu~1$YqOhd-juO8lvgSj_~@3j&m`8Kwmy6+7qrc#$__i4M` z%ekMR=bpMctHO!n^F8%l;Eyi$)OYgV2kN`juCLj8_LF9P;$rSEP<=JD2aM(KjuVR_ zKZVYC@PQNF`P#M%9genNtH`5$Aimv zZ>l|2cyLuHSeFt^Y0E3Nj%mI=N8S3p1sx+meO=U7Lw%jVhW*EF%GEXg)T)Oz!~0|F zebH&2vF!JZ<;cs9zjKcrE7#3^jOU%pm{agXzM2?gk)E=cb`H@_PetZ+@`G!Sf0fhk zt@OK=e%I3PTKZkSe9xwB^m}_K1s;&vmf`d}gL?G46qp+8w%V_T*S~B9pXowA@8Umrq2P=iK=}%2ILLk{`C!V> zAI68Q=$z1=F|V!-j#Vsi7WdM#<0}S4UQEit-(6??|f&#ew5Co zG4wRqW03z~sC??{4F4Y(Pk`|R8OJKdgAThtmGNv_TiCXZe#_Q>IbF?hs?CpgiPDZQH`wW^@0JmDU?+Bqrl~wfG6) zXW;UPPY+wsBMWT(dop`m zZf9LxIw@=Z7rr}%?~ohHIkKe>zjVam(eT7Bcu5cZ>kzbfWC827Q^M$-@VX-O9Ac=3 z72Eor@Nx?E>UuNRKI)~7>DsfVePzmPaT|3bw+`rAwb405{qT$?$%og_X;c?>v?FKy zM$V)#|0)kL_fPYm&iwBfZA~9`y7b`u^w-DSz9W4s{_dSc?AKaJF6;Dh;lxc3JaBLd za%0NU$tQf^!RCvs>FMdOjG4IEH&*aV_odezvC``%rCa3FNbkJ~9+=_t^=7mGES)ve zVSJy*x9NO4$rox#c2mXk(RiPH#0d zefTOX@g8edy>r2XnXKcQ{2=F`BL$lFi0^0~@|fe9%<24rc|Lpcbsco7^cl| zB0#@xhc+5~!+Hzh_bd5bc-r^2t#;L(?2Q00?z%tcgzSnwXzd)nS3l%ij!^#|a-rl|f!@Q&qD{$-wE?Od!yYpP!ok3~#!^^k;bQZwK{f zxarRoPJi_5Z|INc>^NMQKnBh$pI6qh){=(BcS(O^?gjW25a!SI>yeF&@^ z9%FRPBtG*#@)9}QY%RC<-bJGWa<$(VZ~-(k8XFtGMNbCyXpWUQJi(S(xjP1AmUy=O zL|vnyA?ypq;~dIiFXx1C;?W`2GuJ_vUi(z`sJ~ixicb9geMg-)M(Jo3Yw{|inex95Y0WL)j{58itgNf;kW8o5YZo}S@L>x}`UBhanN z-2X}I=Cuuj*AKhGnjV7J4^Mw%OvUD5V<-Jnc>QO6wy)4#`;?udybI)9C7)B$xfk%n zI5>2OIh8+yoIdPjU{5%Fpck4}d&J-*GQ!BwY0NG2soV_mQG|e7I`lT#m)55DZldjX zL)Ht!!nRCe59$8l?4Jk~56HwDDbu^dxk{D|tG1%rU;1u}b3@FcT;%7^d?V`Ed)wQD&!Y$#7((`S!uL|>f54S7a_~b(N=A!$Az{d=eTTC)7=tB>Mzg7{?dkg*48HD@qFFSD+hv^kM$EO0Jz zBFG%3n_Ob@^%(ke=Fj9u!*06kdy-MXraQS}XVhCCWMbDYZ@C5QSoWawgwOM=gLlig z#;8wYS<3JCo-t=j9krIH5g+(qOPFu;{uj-Awj7(wZ(9bQ)3nf(hs^U|^Q~a+r49Jv z^vjVM4 zrH*NhH*Gp|i+&I~ha7kKu0=mPurYe^U#ol=x>pf;)NbskBI#aSZ$hVeH!!U2CUmWv z(6!#RhSfC!XYKoyTu^-P5#)srT~9JPMe-XSqO}CY|4JSlK}MK%jgA@Wog^L?NS}P< zzKJJP?xR1fclWZ6%lxG^?PecCq;YgqdBx=CO`gM`1JC1%)i8PEhBXBoc%L?9wqDU= ze%tfW$?u~3C!QDvjE95c*otpfq0h}bk8>-arAKI6JT`-}QvC*};Y~*?iOrk;Vc+!m z_g9=4Lmg8BRzf)r($EKG zQI{4i+O@2!qn{pW`(H{TU(4MLE`%imtL<3PvCVytzb!Zg<|dwmcJ#Y6&E8k3y+o!C z&XH42IPtW5o&xWxtz(&ttrNJ4??!IBJU98h8H@%VwaHz;sMKA{+4-Fe>InXBG0^X61wxZ z&cx>otS^~dw$OtcFY^8A&fFtM6-%x)XVEYEIXNhOZHvi2seF3)U+b!?_pINIKG3!5 z;R9PofA@Ipv%8P)IroF(jPs3h&ShA|xeVhum!X_<8ETiOY^ut%b0SLj7GKr1lf9Or z6~#GLSyu1sr}`FNz;Ca9vToe|jc*8+KJsD`7XUwfV{XM|*Y!ZVRq)?ifaN=Un+=}K ztyp~Ber%SC5Bt6!q#o(#YG)!aS>J@MZ0=K=MmR8X@!qxXzXQBn+?TG{0_-gQ`r&8% zH3RF{$$0VLxdZs0#U2*$pP1yJiOF8{UE=oWlZ$&7IJj3*cT21Atn|@aT2&u>rSb@4 z%4W=_yzZRVtWn5c-YEcQmE%QgqseXe(YE#h7})K$eLBK{`x%q+2x%PdpF5u9{pp@M ziGjR2dXqh#c88udo@RGE6BrL;X}9Rp2*#59?c03ook#WM^YkT7Up^c;;3GO#@952r z-${LR`*)*(x97Yc^g*o;_f-W<&ghWl7@qzO`9a{@9q8WS`?5~R9uiN_8)Y3E&RQD& z`p08w1v@%M7VhYxd{yj{=<3)-(M?>7)2tnvGRH(qM_N1VcpBvm#n-}^jBeSON{)nG z2YfdN5_{>Z*$c|KB4%u^j^+6HKH=cc+2DP|==THoW8xlYYZ&8*fp6j&G0CO}st@>= zlLLiys~yw8N$IuBZQf_BWAb;zNAeCnl0SjZnfU569%gL;obk?maj^P;_ukd>f4|Mf z@wYuV;leu^A6&kmxixcLofduVoW94OrB8}06OYJe|A(GGX7gY=J9*pWEK6+T&+kteCm|uae_RYfxrQ5&0iB=P%5N zzIcIXg>w3TJ#fqqFcxgULVg#pt}UJ+dqlQGBk+X|+r@)R@WUw&i+tO2iR;LO2kTzn z<$wcWbQJYNd(dRx{NvZ~5hDjhm(pw9K57jzlm3n2yD)8Qy~WH`z?LW8-!7ciIB;rZ z-rV_0rVqFNJ;0!L*J$lwGxp4D+;5AGHgSWifZ2B1-Nrb@hviok4-VrW)INLHhSqwZ z*5@V#tPNezmTNE1V$FXe-_N4o$e(8N%A0G+t+Vk1DZfxkQ?)xDmp+`>;m>Of#11FZ z$Do*DY`u6P`!x*w#lNds4zq7O7PwS>+!?^`h!EF}yf87N(hbrX2XNjgJ4kaAV9!{D z{1y0|`{QSG7@sMReKJ}sxCE{5RU@0@l68gEtv{m;FTGfvKNsF-_ycIC_}<~--|eN~ z$E(QbSBOb)WOH+A8s|KK7lKnsS|IWD?Spxu)}>wF`bfcMIC1U6@lSYYQ{8aeztO~p z4DdtRZI0Yq2dn~3)fZtCRDS~f$=`G{@hI*ywH=v|X8I7Bk|rGyf0})U$-mN9;jrmn zj-6-E%xO+qLUyC@V|+!YSfyc!>I1Mqj&5_D!#( zujA?KIO-nk9^8 z9VSMrm^R~6stuec3(l;a8#vRZ>KB~dxkVS#dU-VBz+Jp>-Qz}oY%U!=*XVDuuL~SF zm(s3@cLvUi@!0IZSiCLkRNwUToj&F>-Yj4&pT!_v1fFT!Kjb|Z&LzP4>9gSLk)(L$ z0?q+a-%BH@W(-wJ?5q5{*iyvU$`=%FN#UJs*0iW%bGCY7bA~fl$?w?yUpr1`t0zup z@wx0J5Pd9}lc+g|xNx=WX>Yf&RsBx=L;I86`s7dUWS!ftV@{&zY3o=VS~T|&?y*bp zMJ$v*TfbKzx3XEU_Gj)oz6HDMNy;@3`Pj7P8^#N-xYm8=NpdF2PcB`%zYJfV%ARCQ zFX6{4P<`QD$5qb?@WQQgGdhLpbkEt1@Lr&J*YT~@ohQT#`|qhwbs_g(XrPXg|Btvg zfv>W<_Wt)7LIPp3;#7xBAhSb7FvT_}1ENK%z0m3{Z8Mx8QK(97uUOCo5(Wt>N4Q0y zZNjLD+ENNvb8TNHMJ=^5^tOH5z9r|J3}CQ=wfE&zX#U^dex4^eIfdK)@BM#1e?FgZ zo@YOM?Y;KeYpuQZ+H3EP9bxgsC(r!Du^n9BC3YU?9O5bG-MEvzO+|l=sV0-Y-gdzkqj-_x5fwYi7ZFo_@S9SN_Ug`R+z` zS_31mLT-QB8kqh4*2~RV_WhH_9GkOQa`!)XICg*2hefmZd?-DD-FRq&@ANBy&~t~2 zAAbJB%IrPs)em{!(aJPhMLu_cg0b&XZ`#O<#IA z%lPN)xq@}0YinBamaaQY-Z_POr;Hf7)a%25m4oW-ZL3XWCc36JDIQbfT%7Sx-<)_M zCoZ!#DK2xwWo9C_OpD@73NKapkN1?Xebj!+-w2PyIV0;_<$K@GTmD7FsHl9-n<-?? z&5u7e8UHwZokv<~*|OD@e!g;NU4Z>TxeXcS`bDg_4bQGQJUqMcLu?XD8}D4rHeby% z?aNGrnDCnghs^3qlt%5Na%W&=KpHuiuvydIcWsAu+6DLslWvW>k* zdS0aG>d^MTsf)h(n0RpFzi*N(NjJImCg=_MHk7&t7e^=S^b{Bzb>EF$maSHS|%kUPr!9nwvQni@1&a zdTdtBTNgTa*WJi_w!VLJqK#d--g|F!_OFfi-fwljkM`cLaNb9F?0 z&#rLW_qu~$f_G29G;icScYI(y`o+d8_UaeW8o51KQ5+PHavrn$`*g6yj}9q9huHMB z!4_zE(42N+Ps+~xm2=9liNpu4-L`rtu|&*A{O*`68!gOyLw1Gy1|wf|Y>Yf?AndZ9 z`BjuT;BM>!gMDB0uswTbRmEiXva-J-T%fs3uZ@$1ep&$^l;~W6Y|hW_?w+tT!?~y9 z4PuV|FS@L5AaQrZ3ol_!OS%o4W0h;;uX1estmbdxbF*_C@YP)TA;$-|E&P0)Bdagg zfqM}DjI(#m>bBZXwd6Jm75wv0I6qUIu z{BWid6Xnj4WaAalKhJ@q`sY-w6My$MGWJnz-}t-edT6~?zJtN|4uvN@hfjL zCjE8D#Xa#W+BY04b>f-h(}1~w`dzP?33^87UEiD&=XeFNS4xC{Ee1 zuVKzA8THD)elhcD^ae7WukV^4**dACZIb)}tc~4dPWj-GDe%MAmscL%I_L)<-txzH zKHNX<{SSBH^UT4XpOO)3S<1Itu6apm_T1trmxdl)`h%5+_3lH?wJ!9pSDQ8XEhb8q z%8vRiPEUcO_;VJ#+CJ6c)q4wA4~GtlSyrD&E-Q-bY~1R~9{bRVA=s_);Pvdto`E%o zUw*=@abtCBSOv8=^UhQ2=9%G^QsD*fd6U)P54kDjaj z?NfyZ>o8-9UjoloXd)VgX@}-s*il_N^9lGmN`+Je5K}<(&x@RM*a=0;mQ{tZ2p|u%kM(!aQq*E z=KFZoc;v~H=c^7R$*_FYLx9CqVE_5A2puVX*36^rGeYv&4rNauWBcwv>971(EIhJF zV?)iLNEv%?D&XHjbdJ7x@X7w!dDLG9-Q;h~=FgdP8ynx{4zG4WT?SC)?U9Xjg0#HC{}czPms@AWmY@0APD z3_99=-bp`u{o9{-gyj2fRagIOfx|{`MTf|)C<~9%-OuPst-Jp1eY6I?)p$P(e~NzZ z#fOq*(z;eatF#jfkGwh5)OXVF``IhgeVwz$rSVNVRO3oJV^p-%moI-Rd$dpHJk&oD z;|z%f#~2e!3-jB(@f4l6kv&DKqnx_3&?S4B7kG85o@3m<&HZT!>E_wkhs)r}H>tOg z*qxC9lbFi9y)*Pl)*(lnjv_zWgCZEk>iRfaL1k53J~e?pvUZ64Z^A21-b#Gqb@&hG zvBu%#v;Hg29&^rwucwY|>agFSf$P`g{DAaC2jiR4%kh0NwjSC8t}*JRe-j28319h` zp!0-1Rb{8YNm)6RRS3R0;Jc?P`?T%neUAToIPp>HEA{K!oV9+w`>guY2g&jqXROir zA9hT4?Jn6yp6w;S$IG<&-uV^tJ88G>W0OyJ|IG`J^kc82#@5}eqe#bgAiuvuf9m!(q63}sm*HdV(9L+^H4?L^edILQLzY$ZFvG)Li(tVb_Vh7l@imc zSe7Zk%$h%IzRvk8pzZyk#TN3U!Be{!zZ$t0Vm&mdfd-w)G@#u&zpux}baj__X(w&h zI@*u_0UG?%e?fzNNi^`<|Af`oz4}i)&b?mw{$Rk4EgDmcIor|Sh#u!z<7{y@HWhIa z8e=Q4#flX(!8|C$n4K-#8y#b4-w$b@%8St-n)fs^<|$qM_=5W_b$`Cszv|Ddq_&*% zys~}n{-70qyFW;Jd@*{pG!!~D>IpMrGVix=o@{5}!R?*KPYiSVf&FKeuNYK@)g5 z{v!QYJ0-sZyGwUkw8^$_;CG~B3o34!`Q|X$c*j|DKMubMhtHv#-b1!Cho;vLsJj0& zfm=I1cL8?iiyqVS==me(kDLE<`dE7D ze)JM?XcO*F3ne~J?8_MRQgxb1yoa4C9u(ZI3%I}0rNe+nN6w!OuHB}8S2Hp+X8!E? z;N4tJzKxF(fApD!M?{Y<^gvU<3>!0l&V1!9Mz6H+K9V>d(R09~CZS zx_3f%jcL*g*e=#4y&qk0Y<~Itd9(z+JkOKcN!l?>tSGxZr&Ri}^5;SO|gp3As+ zxvwVwdww%4^U6?5=5#0KbtLcaQ|=#;7sYN(yVLB<^o0@wp7|`fbJBjx|R%J(M zd@I*Jjj%HZ*+ndXp0#Gx54-v5bUTM;tPE}MchKZ_&Zd8`8w6Vg>oO+MVL5}VwCXR0}P>$Y5QI_^xjHayYOyb`7 zKE}r%<6Fik(tM=7$QTA43y34TgYiZEt#UoN7yypTDNk~N-9$_;@WlNnkBv1WhIW9O{UDO!hG?3*KGF8Od}t< zx(%D99s8w&{KwD%s<#zA5ko(8qTk;GrVH3k^g$Q?4C#yktgp;Pm$G-Lc^A4g_DM%) ztVEYyL!9%|#26oh7lv}C>ifm)2P$QcPXS}g<;(}Bq0b%^e++J(EB*iv;U0U?X{Xm_ z;nNAO9xdLx+`;2S0Y1QD_TQAsp6S#*$mki<^q?yKk#i1O}|uDi5-&4)NgxfQwb#_CKPgRIze z=Q}Z{5pBXj}7qCKxioYPvfZQ zCYks-JT#EJKLeJ#)e{3p!27^wJr-=5dnlAYkhACK`dD*8u2x3VA|H=5ClwDr2%30y z(rb@}RzG$2W6pROH0J|+5B(|M`?mh9&VCp)2m4{hQ)itsc9iFYFLYRQ7%!jfz=4VM zNt<=g*hR!|T}HebxCd!h+V!DF%ZH||f9^4}CO*fxyRi;ESjnAPvyd0f=L#vmn)iDR zwie&+WgY2!<3d>df^N0_M!nyoPH*j`yZVBr?$xqu%&Pt7=19i^#;02TmQh}{N$A{x zcD}cxZ` z=yb*y%2N6o%2K(-?3cyY9VzC%=o0E{VBRbqD&_qahldW`@?Y@KVQ}r`q0sh$=i{N! zXZ~9}r1<^h@;^Ew|GA@ER@29WnPYVYhBlu-cRZP>KJo)QZyVh8&cAlgc$vK&-RMWg zz8M9+!Cl%{GUU+*=AUHD%<}uYh#j5r0ps|Pss~ROFm4{l_b~L)ug^b4c?Ep;>;1C} zj|{DP=yaBUNY|~<<$cETUChz;Qs!Xte!zP-^THT9_+7WGlh_m8nj=NJXJ|eb=e`5+ zfou)!7o7#(-gAY09zwdmxzs*O2lj9NihaJorvGzN`j_nUP@6v8K5K7L|K_{wbEZwd zgJ;bt-*at)Zfu43S9H%f!CdsjncM5x({t#_vxTb*(pGl8O*;nCjz5CiQMVmW(T>4Y z51jr0o2V4J^?#y!hUT|nd__(>bpLJt<{NCC&hvFWxcHmL+h?a8W9+ko*GQfPccV>H zANref?X&3OZ_eUbc3PqReQ5h+eq{rDJ{Y?$f?Z;LZ2O$^YW|(Oms``(`I+$g5bUZ8 zu?H_fC+!R6um8B$$L9H$-lm_DY%Bc9xO&h#ukSV10{gjV zD0ydMU7mD?Qa?8)O5Zx!XJ*@0?D&`Emyg!^@^$}l99yR??#nOGeK=X9U+d__9L^$% zk6L~>Q1yckH+}H#hk2V?KMXJ2v!3|W1L0svefZ4OBb8Y-hZofa){E|03pqdTI?hUB zPY&l$M9QyV4?BGl$950n*B9=U_~a^$V>1X9;f-^2+pnC!g{6F1)%yr{9Z7RC&Aqm=Kh};>;25th%pc0Rx>}1}!ucEA z*?H1r6PHXKv3OzrCf1uY*DA(;BOHU*g<9&cudFYUI9FoUb)p^}q{T ziBo;k3{5&s$#TH^)0q>ov;@{j5#(yvqyvb_`ySe^xX|Zm%C(C(p!s)*tOV zWhAdZdi^!r@2m$%{xsKmAKkAwoe;bcquqrAO{AVNRqgvCe3i_5#k8R~(LI4RV$09o z*^3KkLy2uepy%wxnYIl9+Xk)OttYNcXD_zX2JHi2J#3Y!8yeZb`7;L=CF#Rrli#Bc zH#xfSk>^~!P#oeL{UlwuDU^R@l05du2bL_4acr?5a?p+(q#*}ekb@D(!4}TyS!&V_ zYzf{{zXdtC7dd!+RhpH9cKCk_{NH#*u-@`KcJOA_^Ta>1Qus;d@rs{jx%?D_2X)rl z8yRLo9NFC zRVpT=?o0`Pg{vzvg6k#2orBqMWP9qjlC|((@>%5I;#a<3(~g7rFTP+`9bK z6(WCXzRYvvQ#oEc1ALdSJCyCTb4oTi5ATcfUs$8J?IPXcJlx{5Gr62oAG*9t8R|D5?Gdhi+M_wUSC{&BzniDgZSzL*c;8$3zSsNiw#Cv0JT_8q zspvzy)TWqilfS1;jkL+ix@$YtBEz$06emWqCN`BdvDvJNjbu%1i%CB)n>Ddztci_e zO>8!6Vw{WlvlkY=x&HZu>Fc#uw$q?@*{?VieiLru4d#Fo8o=#p@doq~Z{&hkEj*eB zeWY(=BbnntC-G~5c1sUx{Y5+(g;sg+a0!3R3FlNG6M4|Ol=)A&-hNrdp zbpM_7O>(=+xQAYQmhk(m$}z(upRG&^DcUpg1epADf4m9g4ry`nUm4#r*AZWx5WT{s>&_fd7!+R%oGl#S-Lo zmN7f!gO-0tXQF%hdiDqhKiL7AW2j$Sq1O_4!@{749(VcYUix4w{4tR}5U&^GPmi zk&R8rK?i+yl(9^@HjK@rv<>J?)iI0yDCgPZ>wT0{3l6M}+xnD8yxp)c*rc)Ei}8`) z!}6g!fAiprkqgoNHy%9vA$ZXzPx1j>B_9pYtbsA40UgpnUnyo!=WZmIslK;-z?dU= z+PG6OiElt3>e;cF=Z*MpG**ZCt2X{Q%W0#=9(>1ZPQrJ3Uc$3teT7#ocE>E>8sOO+ zcvg5TK8HLLl>UA27C%+f??YYs_IwA2?|9$o&s6#c`2LjJZu!&2>;Ep^r~TeMu;V+u zGML9V@bjwJgx80v9dI$I}8UH)}KS1xE{9RuDbJP2vNAIYc z@4uvXM!H>Bc>Emn9)BKs_nhw=`rG&&#p2ld*qfXKEu*WQapBu9;Q!S6M4NL~>N?J# zXkE?uvix@Q?HT^$fA4`s-*8uw9aH9but6W|x{qzI7wprN&+kdT)uzHCubgjUVZY+R z4tHP|3bvm6Dfx}^?VG8v%wwa=9N0TN*b)c!>w;Z|-;A6?+UW_z%k9oUdy z7labO0(J}EZc2q64(uZi>;)d|>kjPof_;Q|6=1)^x6!Gvqkw(bf%SW^Z4PX%U>^=8 zUIlg;-#(cNI}X?f9oRoT@A5{_Z~ON`!9K|On83~lE;AMOI$-BGux%dfa0fOh*g4#v z2JC&no#A^QzM25+YzOu~JlGNk_Rqky&JHE&`Ta7m@2A3YCQ~%*!0z{8D;?Ml!G=SL zh5Swf_7ADBHv?Phz`o+a);X}h7HnxK@pXQ02KL=l*buNY9oQE<*v$^?TY{Y#N_>sq z@xbm*g)IU0ZU^=`5B7BjwneaabG|3PR{{H@RM=aAo$kP{@L<~<*k-{_XRg5S5MVc^ z!rl(-oet~*50-ch$?tl>-Wf_<&#wV?bt>##z)o{u=X$Wi9oUtEofb;u@_UMJb*Zp3 zfSu~Ve$|65abO=4?9@==LVja>`&KIKJ-`+_u%Gu}D;?Me1Y3-cj^87EyEhf~E5H^w zumv7$oda7c*aFTwrme5@?e0|AGGHe=u-AF8n;qD@1Us2IIj}qVc55nZC9snm*c=b` zbqBUku#<@Q2X-sp@>5}}fW5(i{e%bG=D>bRus4JfZvnfOZ(~wn=K_1J0~_#QgPiX} z|Ku)_p9Hk4=v_EF#lrow&`*pUuwy9Zn1 zz@`ZnTXQ|I_XBs9?|o!{KCpQX?5{l7N(c5ofoa9>w36SM!2U56b|J7?4(yvAY@Gw! zDcGz~;xT@22li+xYz?sP{ODB=cC!Qf8^JP9dVt^00DCYMwhmbA5zVcdJlNMA*tZ4C z{Hm1SYk=L83cDQG3mw>X9&DQf`|pCiFqF89-z;FayRcdtF^s)h6M52u39|o=etlUm zLqdsSe#I|a`0mVMgU(s{>G=2+|H~N=nxFoeHBQey>raeqvj57PSIgE8Gj3MW&Ltsp zsIJV}2ch=Ps$;(Geh))KrESt2+)JZRl*W7_QRud{=bM{fen!PRf4>1fBL+eJaT)!B;%+w;n{aTQ}3B|HQccf#ZQTSysdopIN9pk z%1h^J56C3@JacY7KDdd>=h*4)nh|q;Kl1_koU&@nybAULGshp3li@qCnEkg!=8{O^ zK;MBp?nC3OkO}+#I&xBL0q46Qw?6J#kPPBeIjDTa##bLl7kGCav~mticP2gt&dP6t z2Or9Tj{AyEN0ChrCRCMk8hzT5&HRaY$a%+@D=G#*tg|J8oIjy=aMZbSar!e19$|Bx zJySz#O=?>u?W}X#SxGyC@J8HRIw@-ccTxbK6*BY6+;Y^`ICR>?{UgQ1yT+M&s-7tG zroh4}_PkfUFJ>K_vJaO3xTqyCH1tdxFtyqH)^TQBbDZ?>?g5cBpOc&B0qJEJFhc`UC;b*J$Iy6)=snZ50?4u9s|y5T3&bh;b%Di zU|t(*JV8qrhv&5B_d06?(jVfZ;mBkiWiCPf7uVqzaOK{U3H@q4?vtCLH*LY+z<$3g zBCX;ld>`_IU9L0Hq!%b>GC1Wl4$}JHHgEk;zWn6%F#Dde9wxnTIk0BiAao_?tJBY_ zTji9o?xVVmVu*Op*)~3sYg*Y$MvOUa>yu8oO854aS=zh4Ppe~)4bgl5Gp>Hg3?(+X zx`poT?O&^Jd;Jeu55jkM(6HC>*Qc2KFm9-=x!@}twJsEp41l|MFm(-1{$B6f`PSmn zi2X`!b1P-R^vT53=s*eYIP7>Q85B_0^A;_U`wtM`uk#S7FEYOEYWCh3HFc zPuW{B?AI7JTk9tF1BrjIqlCvX#Rp(Fg*cDb4^LF!*D9_m$={b@CTZ`D@*E{kcXyzv zyBli~8%p`bd$Bm{V)dbx_4v8OFLBZ@>8(dizWg^0w_f-obo30t zJ%?Pj9u4K|yV}-9TEj3iZx4E^9UF2BWoF1Imw?5%l z>e3_*O=Nq7p-C8;gp+6zgeKJ{XlbG~OTpnwY=K5Qx}m`X&_KA$zKFsDUl#45PZ{6U z$M^C*xt&q!%Au}o>QY(yL#F!VNzQNFf5|a8i~+0awtY%x*)6YQ?AzSky{(S>xh30m z#208?csFa##$dy5BDRItLha$0$^J^GY-jxT%GPi4o>0wwVpHMA1;lQ|eUUh}<8b;` z>jl~ef_|7$hrWd_T^q%7=n7lzKBwI7YWSJ4T=XHPBEndncvf;dy{4t6-_S`>>D&wD zL-cozBF0Q(v+g7*WvyW&F=nF48OhO9=*Spr*K4Uim-=(4*AHL$@XLAWIT;Rr^}J)Z ziqCr9OL)g8ZtJCQiq+SYS57(Jv-V)sfdgy4J!v`IZC?)G$_6YVzt%{luaeuUc0`Nu zF@jqqf2GhQPMZ{8S_-We^Sl5()7{M(os3l*P4IxuO3P7O`15S~Gm`)RuiJ}mSKNr# z#y;&uUjIAoRiFMyaTCzt=~vKq>QggxBEF@TOwL72V|>ta6uQKq3+riXzV|BmQh3C( z4@OUr{c%~3uB`(v%_nt^#Y}9-S)J%9=%h7*&HOD1a0UhMT9fXEzI))mCFzRY4Y5Bf ztxK^WvJ<>LuKLYFuk0QPpNgN+9hEmamAMNtBAyQ|@6}NYzl*IvIXP}QS;l-r=V-;J*0dB2tzB1$ zzRUgOrIXMbEqX>TSf2O}`p{g)`B>!B{@^+nK1e*T+7iTWZop=)$Rmb;F}e=hQ+h28 zyIFgJ-pGDq-7XWnzY+YB=`GmBCSzg;uWd=DLq;BX%78k|fZLKTG*Z%0)`1N)Ed_QQr z@gvsh78V^T#XgmeEzGW6S76M%@$4sQMgO&NXXDj35yJ^RJ)NIh$8$|DU4@taG;dK1 zh2lDO7vRB90f(N?0awNK5K}ep4){y!&BgSYd?Q(`CzsLx=;Iw>*Jft!vpkH=%osIK zXY|Rgl)WZ=^((#=tjf`QF}AqMCw}L~ec<}c#_qLOZt7mUjbE>Rt>J{I^H1DKr?EVb zXKSl!ot`pei@ycFtnA zUu^g4&&n%KbO5`5)%eJpeB*rb`Xa-im z?%kQEuO#-?ZA;sMP`+2*>);~XRHr%k%wcn3v)ErJbOju>aUpjz39Kyi<{l&jKx!+6=&ci8j02vj>Yc^D>B1Io zorE1-Od0bQ9~S&r;PFeg_;}ZSf63*K8ybb6t z`SE?=k^_IsKRVBA>xo>-uHGI|8R+1V{I1~Jef)jI-!lGeo61UTyr9a^97T4d%Jr3D z$Vd@q*kQFkgiYdp=UGY&&rp|2j;q3z=h3w&)2_9b( z9`KgtjnO=CfKPD{4!e%@*ZBLqzBBKzI>MVPVZX@7;H#>*iM7J~5#(u9zUol^AE0xg zp3&2O{Bo_7sqx3!G(5}Zegd3I!EGP=6QnOMa^Y{_H)iF7xeW8|OypzZGj2TeIb`Jd zr~Xz(KJCuAlYL5zg)O@oQ=?-wPx3V*6Q?#}=V#;7vOWR)St0xZj*K=eM;2SYd$}u{ z$f@`@`-;BrSERgO$~*jrJY+R*d7AdCF5p(AlruO5Za~U=Kb7Ok%!SB|rMv6j(p?gJ ziMbLVc=qcP_WYq<|CVgc8s@N)d;HNe)FzFyl40HRZgm{~PMtATgr72wUI=Fd4pegX zSRLuL1Bs=_b}PjuY6RZ;NZ~8&shzbZdkoj0GpxNmn|A+BaTc!cR?lVl zG_euvdF$E4#ET!SZ{M}U&=EVF@yGG&sVvd$X6R-(b6?MUc#d{=pOmjr-|wKFWSHb< z<(Dp%4$Vs9j~y;w826t?KkcVU<&}Aju5P)`(RC4cTK)NJ_!BukX-*D|Tdlrk3W&*qteOJSaS9|4t8&yM{*! z!Q(#CE6FRF^Nn)msWJLYxo-u z26de~`YJ zJd^mV5k0F$MRGVNu=`uRwCfoUq2bFe4WpUpCvf%7N7Vgn=juP7gWD?O_Fs*N0~quQYCm-z)I(#jtto^R;vH z7rVA1bsRL;ebU}NH~JWQ4jGTZtDjtA)>I7Pd;s3j)#%C4_W%83@MVl+-~JlpW0Ibu zj6VVVd+)KIK3OMt{G7Z;K(fD=wUE}CriJt0k63?=WSjGBv^K%sl`(Z0+$8o68uOt6%(+G1{BMc)G#PJ5%o6 zf(8ZXCG^OQD1HFq4O--rZPoY!PimavJ@JI=AFy?89kT! zJ&$zh_G_M^Z>X31Y($|9(wDBCkKf>=x$YY4!|JBo z-+t1wiCul^=4%~JzSdzq`3|P!OJ4Kr=4@-$%nn7NK@R5EW(lPoMU;`m8tJSAJQYsoUM{YhvZm`s9dR!G^@%q`cO#H7(g` zL#@16|9G(1mn)qkeG{XM!f_@Mx{jEAp5^B?8Tbb#B)wz1g>FiEFX8=W`U8Ej2$&pT zvXL+ES@yKf>d#B^;~D&I+4$SMJlXi$O798gYYr><%ccz3J?yPtBidQK_-^?q;PAS1 z#v;x`|H*mAo6&YXq_?jZQ*KcWb1qBsKI@sM#;}eTqi;&FS7JODWpKYVG8@+TJj2_#BVd4xrt;}Ym1xDGUw1U<8TSjrNOLk80;?Q-z`CSpq+MX z&3=6ycg~J!p#NW||BKA<6~gBavc)|9O5%h36h5%NXU4wu|LW*((I5m3bXJqcBgy__ z^`Y;clKegBMvYhgar6bTRN}#0`bd4^Cq62dzVP3~Fa2PMTgoos>>B!F@-VY%LHv`E zvh+|>Yr3C1f*#*49&5!;YDEUMZhM7?quVdZ_50u{KfGn{Edq|Y`T_YtHO@(YB$qQF zr5vw49=|tnM{=@kC9k(i4_CPUWUE8*6|^%aib3;s`fAI`0gT)cXB3Y%`+QKl(jwX?xt+} zZr3Ax-0|y_*`pIZdk+0(dD(sc7##0?v8d&!3D`Q`Wc^rmXpdh8I_enTGd2u}Xs=(o zJx8J6oa6XMv_|KBxlfSqzlXQy19xww6;&}1vV zB;k24c=iKN?IY@Bd=;LvllCMEPw$)K>v8a$?fQDOH*pkrqBn*e)fw{4#k_Jmps`@r zQ|7(4%-|gDYgIog-6=0@>AyGg?32(lr(yBkct$_(FK-(vl&4JL9yommr5 zo%qmF`!?$KE%Ldl2>vOik9Eho>Qx{2q|?_*_tr9d#xu8Xtv3WLdM!EFai^R(a zC)v=F;X65(UAF5KaP|28HpbpQw$DY_HD*|(T<3&@hDWx4qgMwM0I$6ol9_)9zGir2 zV-oze=YW5Lb7Ycj9nG~Xz$u9Aj#wC6zm)yJOF6UQYSXH<#4gV4n&C@7{YnYn@yYl|v0?h55B9UpWA`9ipCWTU*6Q=V23@os&mO~%p?Bpz&irO=xfPoS z8>(4rnOeJ-Z%k|Z^6PxMKX!rUyugXSb+)8)hnaIm(qPJr@hy%HEX9B72X~cMjt&YS z@8W?e=uwrSeDcTYyNwZ`zOC-JWZT8|SxP_L<x<* z4^Oux_g8<$on-xnjk*K5$F(4(xa-KhP>d(+I$Sm0aR z<9UVoMa0?ou}fqtvPLnijPH!$&3@=o1n*U_kGuldB77qiyUn8&f)_nC?!Z5a1G%pP z9)dQ>c7>g{Ot>Nv#c#8dzE!;eY#zyY8!#1&n?cIfxT(9^3$jC}hEl&iAD7X`;=xMM z*veO7zSdAxhGZJOn<#;f#q<$tv4=|NBlPpZ%7SSvm5eLQ6XxYG-eyx*7JQV;^ZDxd zzvZl9P<8?3!AIM~N8+Kcm-Y2uh_`N6oF4tC=i7KjKOZzZvLiY83-c}Qr zB^SjG{o6V3-O4k2b~k$bkKL>Jaer%X_S)mQ;9UydmPR90hA%%C8aaNUevvZiAozaU zDehW?hsDS89kmu0=8t9m8HLYdJm+Qa#b!hY(ms56iFWF1#GcmvD9y#>E65reI;FEQ zqvX%IV&M^Qt=aMpd^qVt&RfR6Fq!_%X}lm3gD<5M@~U!AYyCuX{akFpWI1qb&BwQ! zbI{?a=3CNh&|!C~K6B}D+|r{r_RyopMtCW@&zFBWxcG(3SnN=6j-z`k!390ReyB&? zIdZ?qZsy41u_^FpJ8fNx-snKqDnvKhp?)6>-K<{NSD0UdUXVVK|8WF*KroW^iO|%x z8+z!h0>MdUMY~a=o#@DzY-t9+zDM~{jSb|Bk42vlgBYbgt)~@ww5s*iQN%NkZ{PWh zE3ZjBvnGjFI=5hzOCO!NlRUmUW1@3jZ7KK871GBA)Dvw8<>#QoqS!CV`0Z8uuciHb zcWFDX2pLhnzu`0JmkW&*kDxLxRNH_RKjf+ppuNV)5@;+QOdj(VL&KPK8}+m^7Z+R% zx+uTK0qN2Px6o0oH$X@FAho>P)W^`e3_RP`hg!5g>*=>BcuU7+%yh;{eV473 zHo+v)7_U?o@ebSWUG_l!(lj%%y#T+0-fMYBK}z39?+A|vCll)>_yXqd*kJfZDLc*X z7hXA>ZB|RLpF1|tf5gd~H#g3cqIQff*xZ4Ar{hNqcqcPqrEIg9MeNXD2 zB~^Cc@s7G0?kd3^2)SbjcEpU>QIlWTM-ThzC40i{4)x#BQ2qhMUm!C|lZBwnvwiTHeTK(s;W2MbM11k9PdoQ|Ae+tdA4<2lAv#_0B1Fy~VD^E)v(y*Pj zU^^}1%=pQ?4|i=R;#xB5;JF6PE3uIzFJbH&Xfh#psB|WGEK`Rk%aWO*uXV3wd>?i* zcVQ*-eDZS#msx`Ts3=e1iq-3f33UJS?jZW;p(ny z;BlR4FPRF6r_do9A2Sp;gPz~JzI*K=J=1sT#0Wjba}Ix2Cd(XK_V|PF7P2e*=|5N5 zyJCnhFFaCUvMs*SBU*cwY$eNR9Wq+s+Dk>qXq+*u92vEEGiL~4XL~YQskX_UWIQT# z&+oPEKu*u{P3upUqCK|PV&t1e|-CoX}f1XZQ~jF zTxCu;Yaxx$$l@y-0Up$NS)X*i!xgSvN;dV+o6o&&FKe?k}&rVE325 z&KY=bO` zna{avO<&39uAJ)a3*cRAE28V>>sjk>H#v9aeN*q0_fJmxK{EkfrtAmwj_nuZ_Xy?P z>y`OSr_5_o`q7Ii@#f9HRUC$FYt@uO^S46$@}R{IQpDl`%KbdMC6( z@2u&^_>K;4M~}D8VNFG2{{s9{$V6-o>(j=}dy=xJaL;bXoy22;Ln|>iih+(zXFNmh zBqN;Vy;kw!?dbK61vU6GiDLt<9hgRNkH+2n@TcO)xbLYaf1Et`Qtp2EO1e=#wq(7R z{4D*q%b2dOGtT$JzFLM(s{(du)z(*UF{dY0QCB}-zxpcXnOVMm_MD!$KfmaBxDl`hhxAp=njh|x_6-au^Hv>P#j?O_PzK! zlrOmq!MqOMIoXAYMIU$X7~6>MHknW$wmh9p5@BifU~z88B=xtO_>)J zANkyf1M9|gr%xOJO{V!c*BJiLJq62=$rJE~o=58$-l+FwlE>Kl^*n!;^wY>BUYKU*`>lHY4S~i`VxQy(o{gdtV(16S znPf>am8|n;WfLJYr7wPTHyjokn81m_fsCyD2=@#UKH>6|b1 zq3RR8!jz>lDxudD=fO7>h8TMEy?nk!8dHVKh5Cbz)bGt-ioqd78_spljpnWkXipjK`7`Yi zy!-&dK|X?Hx_CVA?diOan1N&+z5y7~|5U)a&*tr`I1eMx_r6W7qg7mKc8W(gUO|5> zqc0Y~)3h)C2);vjyX|4@A85IyD#TgBzWUj;vlIQh-vka+;h#SCApQ={YiX+dLgH%V zc5CGYiR0b1hez}LUiY%Yx~~Gee1`ggF?HL0jHz!kroPSiI(oxRkukoG)1M=+#?+oP z##W_`VO)KUadj7E4&}R7hIba@8ok(oq%0W`Q9zI+54%*4+_yW2|{dzRsEH^rKlPTWer zO2>W+*p7Pi=7So)(ybk{H5kxZT*iLvyN-d*ULoZdAMS?-w!(*R!dJ)ccKA@|vx^UP z_lcgxhsR1O(`3}Y>GGj`8QIXhmOuH4UQj$J_(V7PBX2AAz>r(WG&hyfKh(S4q z;T`$so7&SZyO!s*xy;3`MW@0C;veWnl zW)1w_$8RaWwe(XbudihxPj{`bNfY+DD$2c-o?9~9?;lV#Tucf1AM?c z@*i;gS+`35(++*_F9l5veiKnY$TpR3(^|nSblEQU>zS!$#u;QFR0dtw8^%Y|75yyZ z8hbPH#yjbosC!Ey5323ykK$#3C@mdFTYz(mEOVguWR`of}h1Lr$XsdLy6N~`a&lb z?<-1Q8%liIDQBCxd4io^e2MojKm(7)ZImfGS2JEN=PX+FV-EAzIJ!c60A}$`c5gOu z3A5x4i5b;XIg{btKN=J-6e{=}(E{2j&a|Nd*vyZTRK z73%;8qSy?6`YVt7*VMO?nc$<$U(u_Jhd6i87QrVfU*nnhp`E^|ke=Z8`Y6rnBb8l7*;&+e9G}Z?|JBX`xii!8gZ+hmv3bC+fM?s+ z2jKzkqx=)^eqfu*Oz#Ubeq=I{uUA=9o1vrQrQ@@_B8$faheq-{5^W`L!0ig+7xHQ8t9; zdzvFiZ!(9zBqV>}t7(x!bX*N}P{u*?l5_6bLf(Duo}wILI->0T@ZTl5C%(_akNcNI zrzYwXjh5+Gy#LVkc8|g)z0>E%9hqKb9&%*pdrIfbgBzUu+sq87?Q3~|hO$+bjRk_< z4>;{!WnOY(H=hy=bROOV>%_}0*E@8+ir@Evso^hMajy5+k?P0^tvE8*Pw`w!<^`!aE)4!cNNTL>C^T{A1|C zn5j-=E`|*$MmkhoV11 z+i%ggk)g!zr9;#g_c;12q;%->D<{22kA7O|@Zmuxy@x+1C>?$IGpFx*^yoE8M_<0~ zmGA1mQA$T&{x2`x)qlB4M_<0`l*2v}hX<}uI(l)tm)|YtVx@!UOJ2HL&QPTnW6OEz zt{xqrboA&kC%s3H21)PMxftk4oyj0w;=?c03hmXeF=(zn(p=~=&M6IPynkUbG+eO# zr}yJSFQ6|cqxbM5NjIwBH7DPji5{hY?ObZIoqx8#%areLeuQr^%GA7rIm`iKp5X=R zV$6ROd5f)P{F}sh4?na%Z4!m_uWaR=;AwevlvDqIs4bI1i8)@IZgu+bUA?akC0=&Y z51NI}9-FuIPQA~2`EQrr2yNe|_ovzS<)oi9jgDUXsov{ZyWuw%-ctP@oi1li?$W_qKbFX zQMS`%=;$4s36wm4)toI0+svESdh^-i*dfXBFXh;ydC)Jw*uvkc@NVvvwtW+5esOSW zY%cwji7Zm*nMJ+!TJrfMqlc*c-t#du=EVjv&v$iSF}g3$S2n5ca>lGO<_nC&<@j71 z0>LigQ)k#X2-frh%VYmIZZ(t5%_#1wrQT%NN#0z+Ei3t3 z@-woq%Dn2Vv;Ouzmyb?3eDo{cWq<5p%sRTl>hx8nY^c@g@A3`UZRUi=pU1Z!;N8k@ zy2AqtTpn2D+UttXaN?7C_n&)yMy-d~^>uW*%FrJZMQ0tCHH4qhUd3tbVP2fWS_Ef= z)@#ptG4U*Qezt^ziY@d4qB+{q{L=2l9;+(?9XC$f{j;@i$NSXJi~+v%bX(rIUy^ zyqh$}QrTlMbX5`S^@=&xeAjondq*xYIFhx+-Q71v;SpP27IdMkD14x}8m~;1w^rrR z);MM9Tx-eQRnUA3&*HhVr|axlP4N7xvguP3`YpeEs-4$skCb$=d_K|vX~ZGfwVA3f zB%Yu9g~V&2%*cYwP^7xrG-+M;b!?hAb3M&3wb!&3TeK!)-#W#RV;8L|mpv!D__~^P ztpA=;+>D+jcXsZ}yYiJyOKVYm>T9*3ID5~!0^+d4#9>z(GiMt1lJbS!e4fto^wVwF ztI7Jq&YAEh$7;*B7a0GXMr1&8mCWDxiPsqX=E-TATZ&$XYvAv6#xrQ<)umrMzbtm* zCHk|+J$XGqvRFx-?bM0CZcYdus$E#RI=t|!M{1!_*}~%0YPa-=t+UvPXOOQy&Uhqw z7VnnZH4}ex{{d!A!Cl7g^CZuY%1Dek*#&RIx9A+j9gu!7=brS)V$w57 zZ-l|rkK$|f>3H(QunQF5Rf%k9FS2}p$G}lB7KNO(BA$7nKVzrzWk|;cx^jpUQW@vM zdok^eVIPGXIqQTn7Ewl+ILG)dldrP=Rb^o-NCs3^A$?R#8}UoDwC*A{2^{5j+;Z}< zXSSRvJ|dcJ-d(qjxi9h(+AjXL{&{34nO+uO(TF;BQ-|z>C_3N`_DC|;9W=pgJ7;az zcsiM}9UT(m`L@;FYk%+LS!HUlW7%7|xe6TBez&}rR30`*6y2fpB9|_*E6Vi^o#4?i zjxH#Hj#{JC+Mevr%_%s%E*xZ+Ag_CP2ESEir85`)slFle)Ac(9eYxDRFJ93*vU`V< ze$f1UVx$}!_eZ=dzA28bP>j?R-iOhq=TqABAZ?Ov@_0GR<>fEH%a(s@ebzs6-+1xw zCwk(2((#Ktf3exo`Uti$ZL?!Ays5cld>VF=XhEN3sb9f6i+M~ra%J%6$d9cU646BV zr#+{LJ{&dB$t!<%{1L|%p9K%t;SP@8xVWV|ZFkOM|H_l|FfTQnwi}XAJR9Ijq?% zxxz%YAh+!c(=OjajC+n7*C5%FOa#Eod)9mdyVBB#^Z(aoQSLjmX#?YUC+#{8?(gCM z8V8KtkG&Mir;Rr%U29Kj}v-*8WSa|UpR%^mPn37MqaRuvxPvT#T6y28_iICm zX-+!(N1gfk4|#8P>Rx3+ms>sdjoJ8r9(ruG&N1-tepGnh&AW6KvR9OijzSg|c(89d zdT71MzZ)FTE6+N0oiqm>pT`=#Gd~=|FV@jEZ+`of($THM`TY~+E%nMY4nCg;pBVBY zI{M(dFA0~K;DU~fF&Fe`c#i|uL&Nzh7aIO2IxU%o-yq$i;Rk$!h6l~a|AOJ z{Lzh7de(f~rD3Jsq2XVg@_K0aHPZjt!`Y=_4mb6ZrIa`DID+Rcjl-0_lDH-ygB>WxX*&&wlXPrn6yl z==SB#`b78dh(W!Ca{pxCCXD&j+0 z^jqHP{D!-Rw7f?AO4$cPTCA)F(ydH~0SHI4@qmp>0B>`vcS93U>ucdY1V(L`B+vwklq5bL$se3Q^%b1r~8oBe2D7sVk z-L)cfnm4q=XC3gN;v+npE;)alNk4Eb5Ik_~%<>gmLME~w9UUgEi1oOm#+(|&9QAqX zdVzk_y3ld$XiwwbvX z*gNRPB+F4Id_4pl!Am^yYu$S$dCEl2hky%s{fo;p9>3gY=NFgv*h*3QP5Y+od2*Z` zsn~7NSnDsEcSbj99^;(%$GV#vGw8;nr}I4t2QMaR=Q;3bqy0zE!N12XKE3h#iVKRv z!&=MnaMF2}sd#nJKkecN;m4jSW`-XEAkFX7ovpLWoPzdi@P z&i$Y%@&>N$)8UVqeR1yYPCGB&$Dq4xQ-{8sdj|fQ9{w)h%kE93uZO?Z@!6AU+wsrm zp|{tLPWbQ`JoxT8+T-j8>pgEMN?Ww9)=pbG!}yZtuoeuBlXYznp41xT`TE3b+k5oo zamE|9aU=GF+ISdy($RnJobexM<8k`>J=xv#*;W5wTaP1)?;(TxQ~3Wj?AT=gpLBi! z`>}tHUdli(AjhZF$1(i2vSVd)WN8k9ZuI7k7u<#1LsOss^egSiH}R32^~K&q^qJN| zytQ$~UCRb@&QPtz*n7W$ry5eKDdceA6!DpT`_P_fy!z;(m&v=v7?5X(H3Hcf+$ zchX+yr@kGC-)#|hcwU4aY=s9j2au0YwzmA0it+3K#_BU@t+~^c)GK{+oW59M`X4Cg zkNMDntIsZ75ig!8sM=`qX znj*92GyEmQrXMCsxww-df0K2)Wa=v(2dNWRJTmb*4a8eAB=qcS4e(T|9WP>^TJso;4hu%?m zNj6vk`9m5TkogMZBd(f#29zoJ_+R1$WZlave_b4TP@h;mi@fxe@nG|!jGDtc6fZ=3 zbGR3Hr+s$Pf8eDLW&Xgk*AME0JKa9;;tRE2G1|ua_u7L!v^Q^A+Ak;mMtdRxA3c5~ zNPlQ^9 z2=PAsAR1R^_` z1o|w3UZ~4G1^sV>97kYmi znw$$a_5XjTO@khH&y-BEnTNTuq`B}sj-5*L-t3ar_6gmSy8m z`L#H>b4aaG$Q~=x`XF`#vEuBB|K|1>e5bXVR_SW&6OFlXYzD=QY0lOiHu=T-Sl5Cc ztPlKXH!`HP6V9b|_5i!(d;Fce<|mz9raAv_E{>G&fBl*VvW77XKh{V4h_~!-PBrJ58H4bR zjK7NW+*n6}e|KaN`zbqg7%)E8hqRv3R~J3@%x3)eipviYFRJ)X?JJkRtAe{iHxi#P zgBXpRfU_2p3ypm2n`p2Z@H~GRnViG8Fk`)Z_9M?OTk)nzN7tI>8p<=2cifhrPC5OX zzdF;bX;<0c_y%j(vMJvnepc;jr3I-zW#p0 zYg*EYw~$RP-=F3snxAO?D*LnsTQAF$oXQ<#X3Qi$XZgZ~t8c?MHUgjV)q`qUrdCzm z)WKPUl{lYOm*V$3OMUso(6>Zy0LQTI>eHHLu(>%6y3+2p(n6gfHSAt^MRS$*QuYnb z)LH@S8_O8WtWVLgjhwMc{gc+b*l*S-pXSN?z)S0Mnwu`Uzl(`zb*7A;>AA6Bs`Y851zqBv*Lw=!a zOmN~e_^SlDSi~3_qK~SfzxYgJg!H(_Q`Ps;@8sX;_O*Ba^K*;fXZrUgV8_wFjr6a0 zvJl=F2w#m~IB@;T<-S8ZeWw1;@SXVN>+s2V&Uy3rD;l_{IY#^b{iDCG6;6WtGklZU z=4@A{-~TalTfVA+;T^o}(Sox$2iZ+AVqxhTndH{`?>?_`phj zRqkN=;KbSL)r0BxmA=6%R{926eJVatf0%RhhsxD`u4UMbCy=8kw2sZBKcM%e51=pR zkoF*J{m5C{!<3H<{J-h92A?_AmEN!}x$jz`gY?P^;G^fK`0bHX^u)$suYL#}?cOnq zu_aBvj1^n>>xYc(!tN~lgLA%n6rH2FHac_17Ic{Om2{Wvs}bP63%%uspF8g+4jkPz z3fzhRxp?(+!D-5v(l zD8J?WX358gtZJT=Mm(+Hu;jNeDKefMLvqFcJV)46*ogTMYw-FKsN z(B-XhU*uc4)KkpasOZ2rG{-lv!?S;6cgU9PefMH}g3g*R@UwuIFH7@vtG^fC=3{HcBa~4OpM?2EcgJT_zRN37 z#Otp#dc5tbc?KDBlezv_2SH)?#4uAEp!|69F{{-!RCYu2X&{y}(q z$-=<;EoHt#7n!vBKQHSYGnb_5LiE=oe?m^FM>wb~#f7PiKcnMQ+xT0zjqS)l2jAZq z#kf43y?=l1-r4Q;%Mj63<%zD8>DYZP-CbT-toTKEp$2^_yUY9M`Csn4K=YAa|CrIa zZN!YtZx^o6YdG;t?i!=^*A`})9V1F^&EL|UK1qG&$DY#uYWW5GArJByZiK&`wIJCs zq3ywobw^)8gnn7m4!#?)ucVv3@#bu`Gu~_k{%Y`Ulx+pgW$W1X(O+rUWz=zK&-@Xm zM_jtNWy{CiH;x!un3(l(_s-n8=*23nWd{z(H=Ml^n!r1m>^{*r8oJbP*Dk~89(Z>G zXT)1ut{C_{>MI=TOB8>TG4mnuIQyjFztAe=^Rx>;>i;3_UErgv?)?8}W^x4-6fIP& zxpIlvYO4}b+GdgoB2w2{>Drd;<>G)hZ^Cgj=*_lwDnDmrJ;qwB1q))vVng z!6>4sEkawnwO!|u8^pGPcFic7|NC>E=gB-7gWK*uuUB3(^IXpPp6~tpJ>PSVzun}S z;jL=?XTe@WTf1rN=T2L5CzqLX)5|9NG&asezq8~S;+#D^yQgDzc&`;_FOGm&Cpdh~ z?8gz%o|kM~4xKO>1bd!8t$%=S(6Y*Nb-HQjCf;W~d9U8i>0$llduGI1m ztOf6SFAmMBp;_qeyulpwg5X`!87z4Uc;tUn=`5@rlfy0ZZ2noy?>v6zp4)!R@KC^+ z+i>`PK6OiKN4n*Y>?dx=Bgr+j(SYwt|L1}CF>q-4n(MG}4RY-^!wngT3WGsq0gM ziIZIKP<~^wy@sM#*C*kVaQ!>9bv6IYWgfSJ`DV|UH`zN;c#%xG@?_6PC2_MDdoCF% zK=;=zpzZ~8`R!IAMF75YE$r_2h5s(<-9Z=#a7bKPfdD% zKldktDRZvygLvAEA!PG{(JR>eg4yQG#)E~|iJq2Old5{n9C#!K%_RqVx7^eN$(Q!G z?NUDAe}UnjQ(z>RQoS3#8 z-B`{TLe3Z}-e(N5nZn)G!0v6swcnRIaJ}rInX%s?yIwH+D#18$cn&;C>#PIQEPfsVnWB#!czqyO(Rd*dA)q_y}DyQr%!|JsMO&T|9$?z+*&hmLG_otLdM2EU~FG|8zQ+i6`>6qq%Z2=*g}zd=5s zJ%Y_voOqCaWQ&wP(wS_gUJ3XWzr|a8;qHmP@UyJJD#+%1OP=e#S26(Kqtmo@Ap|{? z6BS)U(8FAVrc1dt7kU-2jv>@>LAVD$U{6+h*E-&Lc3~(n0C^_NaoPo^%EzMfqe zU@nF=7VNo_o@I3{9hcs<4EbM!#3~5ruT5>7%^3%dmz|e!cSN==Je< zsH7o_eXJKigL&W|KT!BDlb!(o|0)_U;@)(wgWFihN_=O%v&N_~VAkjSRJLLfw3^QO z?0h$N;E62q1ngI+d-$UZ?CY;tuN8#Ls003?u0yA}?(@elF?EKmx?W9N^XO{xT zWac1a(vxM>j_I8B>8iJ*c1h#^1m9JQo0Dgq+4S3?g{|=Z^H~{PMIRi~H71*VYc3em zb!9eojGEgc*p4fG%7fTGL@qg^ylnDs@Zc_ZFy!!H3p|M38&rXg$;EeG+ z*A@S%zDl1PQ^suQ&bViK#vPkd-`<&GuQflpf;k*-e3g&vlgDY7FPMezr5H-~q7WT0 z6&8tJwRNtYwTfWTaNarK&Sa-4?|YF?3IZvkxS)F zvf+vD5bI!&XL8Le8}>JuJ#v$DO>2|wGquX}N&dNAH3m(RUG#(OR)%C3oj7c_B=hv6 zcr#vr-2#svBu4M#EIH{7!#~6EgYBE%nu0&^%W(0=PuyxhQ5b@+54Q!tFFAN|DSsLcV%_7`o4|-KjC@ke}To- z9gmIzk93FDpIQ95`eTKu|4PnxKNT1(4w=}3yb>~G=DM-Z_mW#GC%3c&8Pi#(0qk`e z_BvKz5jR0!a!lVQ_J|e0k36GVAkkE7aaG5vnn~3z{XcwEle5lAoyV6SpQi6t zC$%|V9pTDL<%luXw|9*qFGVT-cK}@x$6r+}93Zdg!`D!Jy_onLU*WnxQWKCL;5zY0 z9)7)_{Jo*QJJ+Ons14VI%UI*y%sj^kxW>ct@Ir|fuJJ*$pOm-8ZV1`3eCUsQ>fGa2 zc!i%FA^3=)`y<>l{dHO<-z8egt$j~eyLjH`lQ}Kb->Q~%Xegsq`ciY9G3JNxrP{lzt@fVfRy(y|%(GOm z=fPa|Jh-7X_H7^E{d^C!n(m%EyS(DX z^Lw5T7MbTy{H4!ZSDNRKp7;4AzJ~d?qtM{EU1M3e)CR;4bR3=;ZL4n zG=xl<=gHHE-&gB-u8l^ABz&It(SFMEcwp55d|&pdgnrg+gR75eoQ`hu!u}q3gVRj^)Bb|~hgLfD zu!u3B#|J%qkf%S7{z<|=gpArS5_|K0w)cJRdEo@L=Fbc7&`g(qv~DY>jUool`8q#8 z51|_(!b$LYGsZvYd9ID*PY-+xT=+g^;EVFSi$n3<=oALeiS{TIFObMG&HKUiw| zKl0bU@50|a2U z>?%Ery_D=JPPOMUvd-LrZ#@M+)z5rJoHiP0qk%RWXrq#NPjP$hYpYLc{JG?n-2Ma$ z`@Pyeey44}K2INBIq{As7P!1^iS!@y)aqZcQHWe^4f}tj;akclT=JL?pO^Zm9emQ} zlHI=HXBFeSF~qx?TNOiMw%5MMk=)8 zPc)J@SDxDZ=Ikx~E*w5*&ENJYIyB9)JpRkMSH`u}K9%?I;?JJru-d*ap5dNi zk5;~mMxuda=w{UpF|V<26nQK%@mexI$~;KE?FsmloRQ#i>rL(T4ry)w&x?nUlM|qy zC_JXVKIZflC8yIjN`LcBypqgsxc+6srQ|5%ch7e?~tgb!8dj z2!cC!{lNGKM#_-d{57yhKe_KeXX@KMwTJl0${$K@naA9`=b8_PbuKG&0@uI20{x)8 z0Xc%H@c7(p>kzTpe$|leL1#1|JGsPax;K?^=jJM2V{RAzQT(oR&!+QTbI391*f&%B zrZt83JaWwktj&$}mxphyVjh`(b`k5TeRm#trPKZ{+EI<(4*0c?TtzQDWY(cLu#v}3 z>>^jzIl+h%%ZuPd9&99Yjc-#Fj!V=^{z`Eigs8zTeAglee8 zi`UYY_BQqE(^gaOpL{=pED5fi&@b=!#hjzS-h7j+b|3qB`e)d4-o9M>tS;rft^jV$ z>56~d@qEy}){f4-PWK2D^-Oz|@A=?!m*1Kdvg!^LNH6*)6ql#f9mwZIR{$;IenGbZ}M9v17`1zgK@ z@be9x_uk(JpX+xncJo*V_julK`OMs|y}< zU>yMa;K=5s4Xj}H#sKJlAr%YAnp{ja0F&+`5>-rrt3 zbx^uWb2gpmeANgWI27|CSMm{L7k5!lD)~_EU$l^%p$iWAuz#*ylwJ1Guk51Vv5UG# zEO};)^hU&{r{Ar|aP0G&z+&fasVU_X&R8g2)=Vk?g=#cx#Yx8`E>+_YRP{0y?I=RQMw@=fHY>T??R*KscK zG+=!NzNf}^;WTi&7u-%87i{O+s$Oau(|EoYouKzrbJP1(tAxB(NgeOR(J%GTWIFH7 zL%&Q%zx1JBrlVh;f;aQgEt)T{baadKReUA-jlNyovNlDx+?hf*_IfFqNZ+EbkA4=T zUvy1tMa2V(U$p0*{KrFw(IwpTp-ZClV|W#Lx!vIp-P85hSB;*U^-1C=U5hLameezD z#S)jakeAjq^qkj!6utI>=V!t5a_~F^p2uSQL_eJqES`KDU-0R3I#VTcKzsLO0^`fn zV(B+lm&9A(B=}ZlaXIj7j}on+HhW?~uaWJiwf#u#gLm=bk@N|(3W$FL@Re+VeD(%x zbpYMD{2y&Qkb|#a`UWR8j_!8PMN3C_b1vEw?zw2@H@e$77tNe)z8E}e%~UD6BO4z< z`xHMrAx(Q`$+y_NoIcSl*Y`osbl}(LvD)Q>^XPvDcIcJb>4Qs*uaGv-iSD5m#=tI^ zV&DUQ8ursJ0~ed2x$RHcJn@%&*V>@h>Cddg!@yO2rO2Pgm+vUy&eTOG8*=ps{lptYJc_*&vM z;YzVz6ZC0=_XN70L1_8{|p$fovc ziEXoOwZ>t13OoAz$)s%V97BC^)&#R}uY9cki|nPr7me^l8uXVO78-w_vvG*~B!}yP zU$XWb@^~7#tMvH>o}vQF8J;y!ZF%nNeepVQn;fY4Dvdssx6pqTzrDIh^7w1U zCY|%N`5e>sl=HiTPp=Nzf193Djc$5dJNFE~IC2zYZK-VDB5J;deWB`L3H7kUI?0u% z`DT4WGM^`T3LrCS$PDqYkz3bys5ZVsz{rdCJZRxeIMqYM;pHRF7&8+1jp)CK>Q}&d z6#h;_SIGVd#wPsWbk-vFOt;$mm|NC!vS)f$v4ZCOR9ktpHkf!BSTgakbY4vkYk++C z=GvPwqn_G!+Bte(`AMw}>cL;}gIDRFQu2Xgc&-PXVB<&Y5J%QS3l8`H>yVo~**C`918$TD)cB#thDI~?v;oB{@=M*m z#N&qqulk@(^`Uz)ezVSzYw$MkrO~Ev0B>(_adC@u4e%Pi=J^rhb9hU9^^DdLL1XFY zuQ)vAo+rI=lzhBmTk(}y$28vdGt{SY+g@9JHT@eq1PqE}fz>pK=v`Y|CrX8X*`wSX9xHp=a|UWm`WK> z$R1Y!Un-5*KAre6gLK49~W7%CyF?)Q})ID#*ftRA}0q>2OTvIqmQpfcaWFpMPK*9KmE{3^jS+>T2Agn z@i*(E+n4yQ>$8ckn8%ov&A3l7Kb1{p>_zWA7iw=-p@jJ4{GsCj>b?2FAaj+e}y-u}Z`SjJz zc$$E3sq%+h`w-Xao&4dmFTz{gSKcND+=UL@e~s@wEp4!PD)iO-`C)iDeKPB(>MsiS z(%0?@>$WN;tU%Tx;4un5| zD{$Mg;KXF&^Y7d10N%aT#QDdq8w<>upSSp}_q;w~kLIczeRG4WZ_qu4-%b5%GGFcX zSr0v};I-jz()S|PUoB?;cYQ8l|98oucv$>vDaHp6&!cAx@2(nbL{Fy^Pw!!ELnUK36Q8ES^k-m-0KdUWLoy%# zRo06?pEACUCcm4U|5AKt^dkB#fG(8nl};LqoKDnS8+<4IlY{@O_}%#Q*jDX@RD(a- z$UDmEE+t;w4xA70UN<^aIg|kJo7z*}H})9)th`A8JT0%CHmLSWdFKLRkZJgh9p}nV zR%YE`^1RD9Ge9}wSJ2HDfXht!SdL#O*oa*g*3x%Ix@G)hWQ|-=u)X%3q3eFleEsx* zm8ivMo`H?&W&GwH+Q_AiCAC$9i+NXhi6F4zuO{@q{K1{b%~9T0%(Iy1@_2t~ZLYmH zs594R+iQQ@*X{y0#$WQceVOl?XIkmGR?BzM$;)?Rp@++N=iC46xfK5?ivBZs0&uGQ zfO13RaT3at$T#Ui2e&HUhHNR9Vrtltt*Pvvv)9pb4U{|ILpzEq)`27O z)>Yg$IWP3>zl_qB$#o&O%DWjJb9|I0;^rP`7-i40>GT&d{aFJS5QFz!qP{Z+mhWG` zbqJc>h#eoYtbtDa_8Zv;sgHl{!QIc9wb{TjpW!nD9Rtmw<{kS`ib{BH*fJ72ZmTm|3%20w|*z0JjnO$a|gct8G9`9 zO@h2XTK!IT%K8NPfj(?zK0e!1Y53^VE%a4VM+!F=9JlA>U;i`q9XljDOMGE+q}ZE- zacn^yJbDMPxw=()&TE6eLmOtCjvnx7{+0RwQ|E&$DktXF2mFY2)81#j-(CDf(9q5Mv;UX||cW@CbV1W@ILhITHC?=kv4t$ZP?Amd-quuVa4WXX$r=<7WvDwf}#~ zQ^)Ga4Ld%P+F9($&0PzP*pI@_-7>x@tT!2rH(SA*dY>W_UnTv4GVr2YQoeQHN!9@*LLq9H$Pv4|Q#Fk`aVtAK zn|>76E#Y^0Rd%?ncCqZ`A(K02oU-@K6+fi)9raeAD~J8iV%UWM`Z-ivJs4p9Bl|Dy zub$4D{xmzEQLW!QW|%xdfSlVD_(%P^eLm@q&5^+Xex+$+rjNPYuN~MW-@Lv!y!ub? zKBDVYGs_YY&IHo;y3m;6F(^JZa7R9(8mBB%C!~GQ7jo_Z>khtri?t=FbqJsNWcUdC zVXfe|&R_cm^*iSJ-JdYmtGO=tq_5ssW^MVNqw6gGTz%UL?Y;Tg5d5LJ=kfybX5g>B z3i(7|sr@PIY&pC@>)6ph_MVt^2lARm^0E5+4gHN)-(>6er2lH@>G2s|Tps(mtrMfb zV{nl>#?*VI()sUQoO|?6JAHS1#_Aof_F(s}my`aIJ}!r^V)z&Gkz?4p5VUJiJ)L3$ z?276d{Cv+>UIUy);G4);tNpWYfD3#N`JmEuhPNHt%Xzv7YdXe;l}~C^e4RTkT-{if zc#G$xkL4pRiHr@auDj4$eX<+8r;cwOcu;Mi;*lUc+<`q%UxgE?8RB}J>vhx= z{11;S)?__WB1FIPJ2cnbjUPFiS{O6O&6qUDUBP|LaTid}P}%4+XQjL2lHc}%V1}+s z!0iwCF8(>wX`h$0!qm`QI=p^dx=3{OS7n3)xf$V_&;9d>iq9Q7Tk*MH@&D(uP4^wy zw&z?K{$G0YR4ct}X;ymk4YY?Y`cWV2WBbz9Z0!dhm9;Yl*WUN+wsM{=PxCj&EPvOM zEY=nwA0a2U+(n#A%zb_DZPdWku_mJ~qjhKwwoY*ax;}QTq3P!G4$iR)aeW^D&?mA0 zp1y_ethbO3dUYu}_iAvQZXMLVD7)}G_TsZ0!QSiuN6N1ggD#9i_X@Lz+V$a1Y9#Du zjgNlZtsDD2^%1H|YvH{D+TH<9cjC7(r`KFf{b}_M>qQ@PmeAvPHS~x>k7%it2%5dr zt_MzE*oxV)Z)EJ;*7(F5TGv(r zBRT&Y;u+234$C^cSOG0`_R(DUOn$dyKcDm9rXc6yiCh1XJrbx1iybDv_9XW1GkV_1 zL6cLdhDWOl`AsfF{!XCEZ*m5zH>&3By@p(?J-}Ie4V!s4i?!BUXoq&ir;1g7a0Pqh z&|VdF8uEvtPm;?!94xtw^$sRq2ajkyS)g&`+^6{SKN$Bj4u2|MZZ*6)GC!2YIc<&X zyC6M#8FHwVTG;oiv^ z?0cy7`WbC`h3FgL?}2}I++8@Ryv&{;b>r}k@-=4PQTEX6V(on)?>Abkn{oRF-y{3J z$L_oSO1p39@P2*!)i*jo^U&w*J81W9kDJ;Fw{K`WvhUzSds*vm#@JC5?j4uWrQ8Hs_&BiS`2RT(DT&X9Yg1DF6ZoD z&2LHe@Lv{2kv+wD=DA3=p37qW?nz?wRZ(gNKv7AWr>;;tucqMW)WfhA-&UdFZ(Y^u1ywcmGy> zN)Fw>k@h!f4^@BYgAesC^)Fkb!*k(D!TR_UwocdYuRrd5=h-5jGx$yMsU|u2?eWON zsT6y!m`eKEhtH~5LAX%g{Xe69_)2-*D%qSmD?IQYwoNwAV0-FbaPwStPRG?fEY>%= zC+_GTjoBUVMibwS9Pe=dW0U)FS;ufw#b152F;vG|L1cKtxtmY+qh}r?-qGBrYLPp~*W-Ws!aeMt ztGyk!)0a)hq?~A;*vw<2+ghA4$R|?#H|qHn2EWPqEzMVHpG8YD`K#yNBTwhY_dliJ+s3s$mQV_=*>BRO|9Nm- zkOIqS>&xx^qDI=g7&woB58+$$Li5oV5ojF+moEMz*sKK(tkv*(_dIMNaZ&ZT1t;Us zBaVI3{6;`97bJ0bmUmKh`!IhvIQ(<9tq)wj6%K#t(G6bxVC$#z<1>F$d@eA&5x>xp zCx>^tT-eTo&tIg#H5xwW8-1D7SAOE=9Ls+iyRy(Xq2sr)z&8&BEnk;c|Kgu*wkCfx z?Bfh8|EH7rFY(g>r_WK(AMLdFW7^2xIPZ?XXxN|DS;#x-&&(57v5!w6W1xc?3;D}F z;H%`b4BvFkY%}+MaP23c74|~&k85+e#{5UaHSCegZ|#L@??;)@rn!0V??@f~z_+bL zPeWRXd}aBh-ZO{HnrG=pY=-1ie6iMIZVMZvdG8^9%Rl?o=kb3vS3Zg7@G;`wW-jwF z=6k_i5B&E=!N`4>uSK7~Q((Q=yS==%e_yEe^badqUwvHbrnC5+&F?Y%9^2ZBzUp00 zdnC_?``(z1$~drtW9fwhjstkVQ{kK!Zm(7exQ zILnN+e@}b^-xXc-E831XNPi)zp>A6|GusRcR8OI@y>4AGyQ;T z3;Q2y;T&Y(>oz{9a;<4?*K#JOV9H6e?$X@B8^G?}v-Q8IH4>g}d9}x2V5}`P>l1HZ z0Nu1-&cO@AYvA46_-*WJCUhLj*v5l%WZKlAcG@);?=E?-Ybl>`|`dw*RcTK-&{nl(^ zjQQi(=O*MY$*w6Y*_b_f7IUtNnI8-$?(`8y`o;}J$ur)$m~-y4tb-47{jhKBKpEF^ zuD4cM*rc3W*?)|D{V8ZB|Nb=m-1my&5z%`LGLivqiSe-$hhAkbFY29pUkR2x3mg%2 zV>DD=B3q<0Qm&F5LEEFib+iim&bz|fDW01)761K1d|t|DEBBFsX+_NYxNWgtwzsVh z($=fUSTr!sv@x!bTvsl!v|=e__ubox`)M~wOj$-u8Am2+knK$HR-rwJk@X7d(yNiT zYWfPzV0`e%X?!}puf9$fa1B{?$GZ#sYs~I=?_s>!zk9Fd#_tP0+-qet@1@S&*jV20 z19tiRr?CmtEzZiJ-wy!ya_W)rfoA!k2|DSn2r}hy&S&q^Q>yT{nV9f-wr*ugy+h0$z5_y_xK+6Qx~Niu=eWkSJ5|n+)6Jh zUlStNSxxR!d$z0J06bM@_A3drRSz&%z&MrTskWbM24~?A&l(vC*zq-b!nSeF@1@M= zD+Z8HF8^6`VBh(=+3)Q&tN+q)lvw$vidC_x@}b-Lo{Rkcm}iXaKmCDl9y;OPo2)H6 z&HhYd+8PDRuq^xAP9r1`ce2iJR z&A$Yld<*cx^FHJyz_{b$HT0EuV>!H)o6UT~E$|fhT*^C#ReJ)?mxBLg;QJNk>mt~m zZuqVVIW_wUbN}!-_J-n9we11$e0Z|8O8e67fzHd|hoj__m(}=6WOK}K^7Q)MNEb`!J@(c=I6UUqoy#0PtdjlE z*|nEn99~Qg`4U~n4l3^=`(o_CIOe(F%L{bxYVx1Fa~C|D&cAGt54;=OqyB-jBP6-!zU+@={bR1pba3~Pl;0h)d9QvnoX7d*kI%<Tc5Y>XH3`8!FTgH4>*6T_flZhISF1^Swp?W^c%F-Nd7DR=Af@7Z<4e5 zi*1{hDqqR{x03S_%o`+p@OLTWJ^~#jR~G%{S14y_4XojstxN2f&eX2q7iqpS7uqq# z#Dgx)Hrsai2O87(VB&RGU)*Q=5c_m}QZVsJJ`E1K?fn6F>N-AkF`pT{^RL3e6q{a} zD-q2!$Mz3yKf4(>``iHG%8F0}ndw74UhMcbfROG~e(*Zg?@e?dBAG zy;Ei5>+uwQBzzU6#QTz)wY~|(@(rb*z500Q6grUcNnUgC!#cs8^peGIgF`2#ufnc* zf4g*h*?}Vp9B#hB-XAYHpDdYKqM4f?CO;o=&O^CW=lZ3b zhw>~u_zQfwm*Dx6oDUG7Cd{+He+4y<@)cEMDjnKC#UAT2)mlEwSWnXD59t#=i~WL6 z^knx-=pcSaxvshn&hl880quIJX*7A{A5tR%zPrDNzJe!v_JLFS)j5ySxesf85__ul zzsC3IKH8SP`yOX*pyPYE-b=rIwBJuZr{L8qfo(c_cE`!{#_7L?{Ybmn?=t($>XU`+ z|N0!iPXk}j$2m3F;?jES%#G9-Y2W9<{Tn&g%@-zrI8cwi>TW@wlAlkfj#$^#UUv&{ zru3`!R15MXxEB2l-2xeAopCF+?DUVp!Q;UGq+nEzcuZR#V>dYbF=Kw*;+~bDp4{Xb zh2K-;(jyI$OZyy=N9BX1^lP8DNPCB|E7WS-JUFzlU~uSQH)jN#qEFg8ShUkVBS7^j ziuX3+zm@tb&g`Nl=``>x^3|6#vya>S6)R_7`Jv#+$}g~IBV%uq-&AC+iol4lrKT0Sqz6{vf|q1EXuovkQ(Dyw9H6VO_i9o*9GMN0WWtCW)?p3;)`C zY-Y4od%oXjzT^9{zuVmN@VMd(@p(_BtzW$PgYHS3$4(!X^Jn*km!Mx3dje*yzW-#pFTHusLe^B`AFA#p!5BAj9ZK|(s>g&91Qun;6 z8{HBk2Y3>{Bs!fn$T8K;3c2qH$^qkY?v6MFZc{k3rr`+o`>^%q0Vq91M-ZmF7 z#-TR?t1RNZp~1kaBPsCy$Zf-c_eI+H8P~qYzp;axFEsFWas4RQBH%Fs9=&}WkUX3V ztO_!p)<>PK@@azq6!2&Jt>WHaSS637^z%6F9Z-AN35{Lt3HPEw?=ONSebkx?FF8N( zpF!^}Y~dV_?yE&Ro#Etbd#hS=H-R?QrP}WX+Inc~u`Fu3cz3&%zOtKk*EBYod5Uki zrYft1EkL{@o8w23Ga=CSH~ll_!E-NcmMvq zviTv}bM3F4|4rIo_97}_AMcV-CNamZU=~uq<7=NlAvmM z;nC~3Z+Ow!vvnc3xP@zG+>VWgh9%MIz&F#5xAtjl1wpRS-ppW0mq*`}1OG1khY)+= z*J+(IJXUcVdIp$92XryMi+z4y;pL%|I_IIhwt9A6R`4O^yB9OQJj=hyoxh4Oe`U`V z+3&dcn~x3k)@PV_6JC?gBU&usypLG5Rr2aS=1Um=PR_yWg(sdxUnv)-m{*@suBmq2 z_S^mFFz2jXXYA7HuTT#;AHK_~T|PVKA33{^y?lq5d%Fs{$QOCP_gS;j^$cyGAG>HP z+{<3GY2*!d;j5PqUJ`EMyz3FTMu)1+qtyL)c}qOC!oyRtBU14$KC%ryM&^f0KVhBO z3EU>$|Mzd5=zQ?s&lddPg|jU`>p5GAA1V1KKhPe3^}!Q?k9=R(R_5v2AAazJ ze(z+A{@JV>c)(}u;bUKOblUET;hn&K3?A#`nk^@`FaL=A6=uPwRgfkz+Z zStIAzsdnBqG=M%o(RcYs@a>QJ{_iPmx^xvDrhD+v;o&_mK3dIwEhF*a+f4iNqtF?5 z-DYj6vMkm+=d|hkGr_AEB$K>@zmR=g@qMLZK)Z`KAnaS1e`{W;2l5nqI#~0 zm<9WFrXK!oc!2xxR`zgAXX8a<5d3d`$<__O^5}z&Df98Mc|5xSn6O_FZ0Sul-uGLN z*)h)@;Ap9EiH@XCv$wUzqFCCe+zBvU-{dlA+hxPtb`4Vk*2AoDk}H_X-%>C&i!-cu@$(y`ztRc2eOxa&~5<#J9d;MS{^F{hqmsJe?*?p z$)(x-cT>x8QOY$}2E2BrW)zrG=WEV{GR*mp&*8(o5y~2I{$qTKFL5vRPfLG%XUWoZ zYnI@giVwHmKgNUrw(Z`ZC%GZf36-mudy86rWi{1^SDo(kDXUMD5fS{O7a|1$@M5_?sPW{M|R<^S6=NA!Ig?Rf4TW&*Zed70L|fVrzFuM$NV00ry&jvHX&_G0JD&7JH6r zD!83S45+-Zty8m1{Dj|>=;^Z(doK^RK2})P+M9D(Yyb2qt;`W5dgH)w5;=Maz4J0W z{tCWpoO&eIclGoUm+!#t{KC$$L(?i$Oy4SbP;Ft5qG}9b6y&qa%eRC{=Mv{LK`!AKeh-z%JO$L9QYtJKjzhh%cr2lUhUh$uPJot&QWG)uZ4#$w+SYQmbl6(^1#0qqs&*|)|NUo-#(0Z+U!lhPh7DaV`zMXvq$eMKe;{2TGfrNY{+fe8YxD$JG1fc=%4-S-w2JAVZVcW z#Z8govcxoMw)Gk9oXAd=OXD`Epf?K)v=qrL>JZdFA#krhO_|(`@c=o?P3$OBi zFYl{9B-h|Jop~Ama317V-pNs{O}&`p@$EoXxUSA>eLG}-4xx*3>a8NN6G7IZ$aoC-Uhd>0eFnwM{(E8^JV?o!S% zaK_U=9hm3Tw?`E-Fs38_lR2xev`phU)E6pCtRMC}f(I_qc<4tyZXdkbSL%F+_IhVX zGKhYbKivmUyR;GisV#FIJ{Rnw*;5x=#bvK&&e|~Chnqjtxg^+Q>GY)jXc=SUA zJXpTL;#b}KPFy{X`3ZY$(C2`5r~)OS#@f%n{_-2z`{1^N*gi3|w(+*rOV^bnFIup5h6OjXwW|PjlUw zzc`@lCxeL_-RsU7LEqQ)Utm-C+#_57m--cqYxq45+HBW#_%)TEGtjqD_)_%-Uir_1 zXXV2_NM3{-4|Ng#a1(pw&Y>Q58Rtxs(>l}(4Xhk4d}O*uC68+{T-zYM*Ne9^|I)6o;)R`){3GiQy4Pw5NgI6tJiEECV( zdEjEcOD>JC5n^298Q(gnQ^i?VhQt3Ge2+-Zee|bZCGMA+Fi)Tc!?m3HQsnzQkN7c7YDc zIavI;{`_+(@mc(8+JMH<5PAh$qgbu`7Ove1EyUlKE4I4-fF)WXPdB(eN27h_;amLv zvz`?{qQfQkMvu|A{0i0QEMY7rK0;Sj!k=UC#T8Tg?p`veSXD7;2WweWORo5A2eO%g zjmgBuWMN~}j_itiwztN0igrt}1>MFLD5g~ml8G(IBKMKgHk +;f?;BGZAV0DhXn z{26`?`%F&~jHUQx=zxWL?Q!n6W=x?zHhADD<2j;hx8S1zQx!Cl9@oBfivIlCD zxu^c@xH_ZlGOanDM1Rtq^c{omx)(q{Dc97hlKZg)$lYSZLtzlbajA z!sRJ=$K|PR+JAs?h_3FROW$nv;quns#Gs|# ziB$`*_tKr{#e?0~s?MaoU3qUiw`8BGt}_j`>uh3tx#z^*xA8l6ZGF4u-lSg^@NNv67+V7`z8vE>^q}5vz%M)Fx%W%% zDgM5~hU2&!e}77CiiYr?Y|9DSy`E<-;eR{-ALHN9HQ>||V81m7l15#ia6KYML1*m! zf}=AW`hLryt2Z8b&g6iS@klw(N>|)~u2B3j4}R#o%}U$|kM~zviTS{G>UMH1;NP`!9)jo#f&Xv z#{|-Oe+&F>{^v}CU7uptsD5C$KBap(dIdORz@``|HAbE3z#{zEwY$g;F{w}Md*BV5 zKQh{WDEPP@1ttR@Fm*2vCjQesud~6y`FUh%K?<&=pCwB#f^%1X?D_0uK6|}zil2~^ zjh{dcv4%Ljze;*I%j~({SEKKlriMoOU|XgwBhy+J(t}>YhqZGeo%VIZ0~4+G*d+X3 za=YEwj~-&x-hFmlmFL1a5gD3P&okI5VC^Ah?cEnFaq*p8|B|d%Sj$tJt7= z(r1a2`q7KbUtK5tDA_y8c;*o&y@HR^3*Dc}UB9)G`u3k8%S*^_S8}fS8pgUy-)C5> za`~>Bs|fnCl5@p(@F}}v^SQ0R>XC2hWnE9BkM0Iv$vkAJyMfvX^0#^KlfNxP_EPh= zJC(2X@4t zr#!76n;9AuuV$Q9EmV1TSqHsJB}m zENS-WbeA7n&3+ll`efBr2>u6wKLG3%!2UG(@Ivz8yUB;wIrUrt>gMvPqbn@5UbOYC z8!Os8sQQMd(*_@3T3a*wFz5QMVJ={IVDRzn=)%Xy!8c*EE1Bbule@^X(z^0!2U+ae zb#Cn@?aR5Hbw3NBX?iX-VmqjTf`-&olpKDHw)oEbFR~}c?8Ce5TD`_`sx?UK+HC$kdyiZ3q-bXIrhC@BBZtVR`Z`2Ckm9TsK65-eTKer%F1^sXqwR;Bow9%58l8WR z{ApkDweVL6`ID^#*6U>3M2m9duK@Y0ME+EtUzXdjbqe;=PaaVHOLjx#0_%{yC5*Fs z=({JnhYp{uu+lcAvzB;I{RQE@>^s-UTISu%tvy3dLj0+9CXLioHC{C*Tt~f28Fa{C z9iHZS`#AG==uvA`C-~95rTE{wv9FC2a>H4?`}F1aouuvKrfn;&*=<+PM(`721y;M} zu}z%^d!0!>>Qoa*AA!%~BR~t$PxJ2WJ?QCT{Ccj(usvm#ugl-R;Y94zzO&`8zjAhM z(W$fL_m#2kbIHSH=hk#dzl%nCKMsGJT5EiU(mR}3>WM;Azg&rJjbPLEtQ>me^V~Ci zII&zlI2Arc{}Hj=bUT)7#)i)19ph6twb}9q)J74X6`uAO(?R)6xnB<^Fn}rAX>DT1t<9t>jCu_iADYZ}*xYL}2@H0H` zKwD+-(GuQcjR5h`r_3G$YoYmSr>&Rqv&7S>coj|`1*cv(WCO()Ki59<@Wm4z+h+0S z+BbWDy(E8BkY||-tb5=~)g~`te9H4eqf6}C9P(lFXnTQ$??!F{+(h94$!QOM)H2>v zP2Y{?y|gy%N%~>nhrf?o*L=*>q2A1xBpY$~#mLpTOU)ecJY5G@zvsGlt4JFr&J+N|LEVpvQ=xgR* zJ{8tF?CyE+B|N7%v75N84jNU}*Kbv1UBSnJ; z9U6G&KX#r9RZetI*thj=%qa`J7m1K1Eyo zx%9Q`8M8uF zeEg_g_)(?cA#|ZG66v?c;sa#dj!Jl3L%e!qhh;`CU4U{M`O(@^f)`cc?DF z935*)wbuc>Aek?K_w}jyefhXC`26l#M`t^^v`67V@!LoEe206R`4^A-lzS;f?&~@C z8r&KkLr(Cicij3^kME3sLi@+_I<6qkW`#*GY|%a{i)kZH8>*T2wiBS85c`{$vx#{A z8^dwVNL~4d&)GUaxjFHM{#^dZPRU1kb5TL!O@sds^Zn51E&K$FIml?XRa_+6>``97mM#%?D@u`5oe4`Yn%7 zRn$_%K0brsm`;19eJ6iWz;DrX1-eA^R{w7;H#%#-wb{M#Uphq?ERRf>)rf31a2Jl{uAF(|UF!14T9M|a z0x|SN1$>l#Zu<$XH&)Eo{gSP#RByOjIY)Bc@~8h6x#-p$0`biV*fJSc3M_$9U`bu? zZNrnyj~Kqj{;GanVr|ubFf6<&PZI%m5%8pOKd{i+;;tW!V#AbEwfWMC#k}yqhY_AP zaWb|?dC!PPchFB#e;9sD!XrBtrC;G&|3TIQXuqiM|J1IZbN4&QLtYnyQ}mhlr8O`$Ns4b2mI( zI(hBD;_KFIMURuil^u3IS(hT81U+~#P^dhHT}#{@Dl5_coqJf%*v<3x4xi)@%a^ma zV^clzJMhTf+T7Vq+}ne0Z=x;DvzmQAxRwE5X@BaqoF%f4=VwqS(9iSxxTks1eLU|z ztLOae>$sQap5~d|JVS0Fyqjl+`=7yCZM3Pqt#3r%3zlQ#H3W;F>+>wDt1pn*m3eOc zrha%sHPy=XD;JO*z2`t{?ey8|H|w2wCx>QOXP(E7<@o$(^2u2|#=RcPYTm*9`jkFR ze_9i&{*=?8KW|^&erhB498g>8r}|v7pE}F#r-gfQ`l)jIiGy3MO=FEncst|TKzx~l z98yocu!er-GNvACCDYF}Z9+zy#qS!=RQhSE&7aLY4Svcz<=4!3@HfnO)^M+ae(Jej z#dy?D4gECH&rzbhTA+>F+pmP7A6KFRBx zxe2{qS7@EN68|#|{FlNzesr1qvo!IK;gh`C+H=gkzl&$x`)h0SW)E<0t!3Hkj2aw0 zDww-J9zKGf{_-5?gnbo#yH%^>e2?<|J+x(94JmZ>)utp8{ zKmPDu_62nIx`B_cGyKcF7ZzDt@EvV_roB9Dp}9wUdDv)uuN#|gKJRUqZfb>0fAo=0 z9|jK&{mnYQLGoqfz867%!7ZGOgSM&u_nFQKre5n*5&5Iei$~OJMJoM?H>yX@8LfwJ zG-u>z@3z%3js^*22HJ#Cq|MfeX8S{!ZP)C*D&n)2&u^*Xx0cx~^lL*~9NXya_w` z;v4AmE$`Mnyv4uzp^4VS=IeRp>((XBS0*xx|A2qrjUV)5| z-{v=K@!z<`pBTV59Rs|li9d=i6+VIozRTFU7nCRFg3CQsGs3&b=h}W_W?KVx_kL=7 zqF0xL8~+({-?M_?Ks1$3kiCh4uL|@<59jzaVSnWp%KlcMBlV8_0^O?xpRtfXVdSuW zOt@!Wc_L#PvHXPeZT1}7JqM!D`%3nh=nj=9P~pJRCbWHh@xYF&(m*jKXH>Uz3X&fOxKm~Y}#~_FSGd%vRmNy4&b_ZMc(YG z*!ou1x2T=GqGUVAwEqcbwcN<3?9^OvEqglkou*BDtjy-ut}Qq1ET*0P=)$R1cC!nQ z+7T=_QYWl7j^W3ozCWOt3pl>W``;VoeRn*^m`^uw@R{1ipNef5K1&-ve4jSVdPKp& zUZ<&T2!;i`za!=S9(ZCJbvsAliKjVBvRi%;w!Rl%-pCg86YrlXEr-X-6K~v74$tup z`mVzYI%_d5HuGY?zUQ9?--NdR^Ud5r3z|i)_AAeG=B8JN7JdM}e#M_XFu)l;f6u$$ z!{1r}U1X_ClEj)AD^UR;zXa0wpNavaF@k|DGaIe#sh5ZyfW8S1M$6wMqtq65= z73k1w`PcU{KCf}Ue~Ql!JKyKDPm*&BG^ZbLxFud$ZBRZ%KRULwDvw*K+7;_N>L9K}OlX z@`7*-U9Y`LP23c+Sq$9?57UDzvY^G57TFPLrvCU3mx?pM*uSyJzg zPp;RDFETOuYs;{^d$T!5gFPvxv5%_GU}T=k*kSDCW}UzI7_n?Wdh$7|f3V1x)vUQH z^P7G7^!qvXo?^z6^5r8+$)n7GNe`s7u#_ds-dcoXM`sQ;V5^Yh|k z;#GUkc;;F{tWjk?Qn6_7BI3qM>L{vtfA7#E>T{+8lipwP4r?SzP2O&p_QFqx$?;Oh z7J~1r$Y42fk@9o0o8}t$(zX0i@a6i0Zmvah1)94QpFT$&)QI>6o>v`+*0z5(MV?NX zy}Cy3ADC6knke@5y%Jdx{6#D77|bH)IOgMPPmBR4E7=cs0sG-DU_aai?1y{j>Lc48 zz;AgVeO&WZ*0`>lvd1-pvkjF#Yn6Ns`A4bm$Om~o%bGQXx@ouF)OJ2iJEHfO!4>D) z%=!cG{1Mzd37x-X=zPJ-KhZwxw=Qw-Z8P+}z|i+w4t@2^c=pU5_ldP9#!;WGF_p6C z*lSw5ld)#&wf@y7x39Q%@XpmQCdc8o$8j^XlkDkP{HS8Lmg{n^QNLdjosKLZ9~r*- z_9^MX@Y23{B{6ig=EUR7OW$-MIcDsd_Loh?feW+RKBYQ=cv*>X<@Vvdzk!@yEj&0T zFC%>C>ds*}?;Hi+f>-$6#5y5^-%;Liar;$p`wRFm6{q6E^WpM`&|=);4C52kfyWCo z8nzxsZnm$uW^m=|pKn_UFRV--+uY?F3ondq77nlGoxrX5Zsc+@vP0x>{HESfxccM; z))q5mS{8 z(1OSF;BmjJ>yc?Q*MrR)&K;5yaAm6(A6W8rMzx&iQEz_0l`qA9zxTkig4~gJzi!ng z#yfo_J?Lq{#eTKTM+fUp9>vELzx)UF>8c%#;A^XPRIq;c_qs+c3^4&U5Q)92G3mdd zc`y2C`{bs9XRd49I#)3zKCSYldPlia<%RtCWFh?8-p;a;W%$oG@Jt7OO%Q+d$gp3V zKCpp2QYmwA1$_45*K&6AaL%fam{xwAbod7Rm43%xX>|M*<*T+k{z^CZ3ayN0a#f~} zZgiV!t>yE}p3I|P`Doke>p^HW4|t;FevYxHh~O?&E(;q!#C+V1d}cF_D}9;9$3A^d zKDG^u@*g+i$IZZ3&Sbv^v)>dxq-?ltOKe(>_4s?rugSkU{WkuUGb8y`bXQI1JYp4}=J(+T zdyX~uknvB5ZvwfM0%cihPcjD*M6k_w%4*8ziKp|hKGYGu&QoYaJhf3 zJe@M~l=QdU{QO+#rr7BjE4^zwH0ycEJ<~6E;2HFZYSDYBiT?`xtUB}_>fKc@XlN5S zznn?FK9~ZV@^6x_mpw4L@U@zH;H3SEp(Ev+n7s^|+;7R3h zRF|vzX4w!sE_LTUZ5Xr8-G4GZF00wn`8(iw89v-Pbcbq=v_{|~yf-&MZ4fqKYOX&l zd$kOmx)&TM56!vggYLQL<~KIoITzjdbp}5C4$WV5!-v@WEm2?;d~s}sa?CxNKZ3X2 zwtG$6Ic;wP!&0-ibH+d)Hcq~yV7iw+>eypbKF(5SPiNiV$DZfnK@pFDwTJXx`} zu>W$+xn~XRWIvWZcujspgt_u4a9#1pyN~3mMnkgYz-3@U&deGK^pf^Xl70#@Cnr7O z)!{qLHK%U0VHCPU@9+N2NPW7I_FVdAo7&(c|C!o*=}>&mZs@Fhp%qLEM}S%R)K7sw zGbak}bRJu|YE6OFmtB9Nn0u;G@UtJA?wLLKa{11<{OrXRyOvM({5mg>PNYx#Eb-{q zQebv#ZBz4csqvBE^#iZwNCWVT_Ja`4a&v4Qqda!?G-Rki%l|l)?7*2p=tr&dmH-7!IhNu0_&pLkU5Pi&09fRfX%HexH zHfSg9T9&VQ1a6T>wXOeihwOc@O$|qT(65>I12;>p?m=_j8~om_-}JBF z;&sJJ8s}m9eGdMY@3oY?*j)0jOPE)w=#~bG2pF%9A*F<>MWSG=sBGQqje4monOv)H0PfNuAio6 zLVQ#Y9jnmCk{Rzif8cooCu0}yF9t7qSJ%vSu2s4DZ;!3ecph@bbKEl)*(~{OuI8q$o@iWz*Qc0MAy`g z_`Q*K+&1KURiwP{2bK`|Exi)}7S*`#gy+}xm6t5x-TdA453L<58)%P7;~Nbh8rLFd zUd|Y(%{%rK@4D}JaYKIH&c%E1BHZXtJcs{yCgSb`!*Aim=itSAPAF$D4Z7!7#lTYm z<8scgGV<$VT=~wp{EUlO&5Wy_aa93ZF5{|aT<%!ZziTgwjJ-(NuYDV}g~igFTC-Yr zpiF0|&w(egA*zM!-d5kPIyDn_gDcsCpR#^VF`IJUl8MM1coq8~yHUU%lV;z0+EUG4 z)LrWoNLimTh4l~cY|p2YG8m&)(!0LFwr?ZrrP6BY7uco}n`ocg7@{P;%~rQRU@-CpUkt}q^!aa|K2-cXRXbL9@ayw!r<_wG-B)BC}n!f&GCO&<79dT8jTbd6bQ%~Ro#J)C>9V={h%_Iia^3|)98 zf@~2Jn=^1EcZwejP2@u&Um?+?At^&Wz+>BNcMV*6TOM=hA?k~yi(P(-DQ^J|dQ=OV z2|QzIkDSF3=;G32RCxj(jeQdRRI6|JM=^@z>8RWHD7fzO^fTK2{HiC*HuO)<56f5f zqi>}D7o`Qm)$w^HA^H3CuisVi%#uY`R`}k`pqZC7<8H8RCB6##z|UqMyVVBPwc@y#%@7hOLxN`uFSn)_O#%d z!w;j$~Fln zJ@9c;8lT8P0W}e>o|g`sqjB+^+Nxtr<`a6_G_>_GG~^8P3E@2WZ2c%SoaGtUXy>(T zH2gHuetBuLOtb<2qtPZFu;nm=2e%^1A;dyLEPog?jsa6aAx+xhYr?A+HLfmU@I>nFLzfr^#na**KoAo6l^?}CNh>zMx4qUR|r1cZ%2;&C9dke9jZ^pR)#+OpkMbJQJ586* zPfR7I8-hlCwAIIXY8$x5yxZ}jBQwI&ndj$Rhco&3LVb7F4$9Api;t>hBW(QC)1GXB z{_97}HM_Q(_57N9JxKkE*LH}%Z};HteCtczkH6Q0lLG9L*M7_zrrUe@dyc2S_v3H7 z<}HQ4Td>*C#qt>J{etwa9+R(g z{3UGt%j|U{f2kY3)_Tlx;al%szWQW8e3`jo`Cy;xLy7yY@OUKPl&4d+o~i`Nl8#HhD+Keor422lt$O&-m9|8P?vtX8g#n@k<#0 zHt+a@fBEIeL)&O3mKCM)oAX-1o|__C}HnM}~2e(Z*((@CIjP!ZXAA3|^BAOGijI zMqHhU47Ug;E%1b+zrLP|lT2tg7G53C_#GU4!;1q)cRl6kE!Tgu>+MG7@9#y2)S}l{ zpx5W4*XN_x=cCuJT>bR6o0uEBDSb@yCTmRBnCvml3z#RJnQmQA%zu0(G5=WR$p-NC zUPX`G%RJ|&%{=GWm46_As~oj^Z;_ej9Bbw|KkdwO>X{dK{|;~{U9z3IxmVFG52QD4 z{TB0cl`B3nxN-G!+lqjpC_TG*qc6KFD?7V+Ec0`kU#-O^EaN;&?J<(Zyh9J~s!sWH z8E%ebq|R`0rTmU~X67jPQr`ZM^n2s)& zj%^yu8uo)e121{)pL9eI^xljAuf1h@3z%E0EgT&71;ta;!+Y(W*S2-K`~k1ZzP;-1 zxeyv@-!#rMHzxIi_U`fev$j7pk{{NRvrwKRj-Q+f4PAdP9UPl6!IN?NkLUQ9BVis` zzmfk*l9!w^t|k4Gqr-5e(87r|p7+2e-K8_{z3>&8`N)y>HFgWUkBlt>?`h2OR}eo* zhL*y=bK&2)@b6ssciMf=jPOg=$uB|X<_+GM=mXzbjORvZREH1I&lprYl6BF>t*784 z)oB!5v}tQh_U=t=ExMZg@`fhXjrZRYObj5m*Ml3$fpW~5@S}LIpZDIrFZl3=bE@me zTzSfo1NYuG&h6H9&2?)$eYzYqPVn#n$Vu|BTFGs$6ulDcZXOLNy}@aih&A;puQ z_R?Ltia9gRXc(T;bM@7V(egLY(6!s&bZDboN=u3_ju$nmZ*25{`6Hbx{>*O`G8IM>Z*;`e9lzhQJzXXZ+uX`GsiO` zuX5al8M=>_C;R2Rq|T8a{YwvSz4AS(KWOv`^Y8Q#lAiPElLP1JSGaoI)8}aZpvfl~ ze?2)rHw8Q?=c~CD$?{*5_tjjOk5q16{WakzHbrxKkrum7@H$`-f2hq7wNn#Esh!d_ za?0gavUVy0&Qy0$N!`I3bcNxMfQ@VM@9)0gtUKJVxsz{23s~1b!*zGhO9b@bF`PO?+dff=HGxPhO-#LGsqto5>>Q&vk_tw3)Zr!@It#VS) zYutAbWlnAga2_0-Z(`4?V1mhGx4VEm#q!e@glzhJj(!V*PT5$y>{FD<&Y-`cUjQ$% z-~Wv7((m%msNE?0p}@39vO&2BWBKHzx?VDXI}i*lbxxFguWxZ*74&z*`en&@?CQ97$Ry)VY?^Uk&{QT0yS-$@@vf)kIt3Co`RcCSw;`{gZ3+7!KTtdi?Mr6&`i7A?ntz|0k}TJ6qRxf#yf405xga*fN0Jq` zjqC1@_t>s|LvIYr;e~szejszf>HWUd&g!?OPO?H8+;}kyTkCk9rwbVSew%g{+XWv z$d{>BqJ5VAe%nlNiLGXGsbDMH z`To)wZwB*7@t>KDGp*@5)~T^sl2PQH$|a`YLE%6-p^)E4M@ zb80B@^fkn>`m!5DN6FxSBEIh7`S}yHC$94pnv04Dbk9?Ko@euj(L2NZRzFjnGcbbv zZqD4y@?~~Ck9^N$JYB4b-M-YvTiAZ_z+iCc?WQasOZ&uY1A45zZ0U_y%{4gY_3>Wm_8 z%{^1|3U#iaFSm})@2Jk1-}ar~kq;+IF6|b0>tW)k^r_r|GZj?@@b}g0PSZs{iP|-Y*K7#S#SA36Tj6$Un>8&DbFw&ziD*nU{H>#gRj=(}U-0{y zGfC#gkjpLj{NiO+;>9<5)^(jCPr);ktpnyp$~OSV7I5<%xOsuK_IJ3??t7eJc?mi3 zita54M1EJ7mXt&E2jEyXZp$m7L`#F(w{4!52Wc}*|MNnL)@pG12)O(ra`(%e0eKv~ zQA1wRI_AOhP~xoyWcXp;H!@#m0;KQkyqWFrM~Bfd+2ocVz*&|-OcdD${O z#QVgIAPbKS&1~rCeEj4-e5<~6_JOmBUmHxDr%H;f-)FK1EPGt|b#rTEBPcz;#;JXW+Xa)am z=tRTc6NQ^0=03mW){x`e--KNLyIqQ}Pq+6O8b)yDw1YkWwqdNv^T8Rou=R5qYp3+4 zFuGa$-^x!V-=W*yleA~@{L)@K_|ja8=7YQ2!5#03hVnmaoh3aU&k3uiv+1eGc)L``}ycUF%Lt#aho|K50K%wvtyj^xCqHe@kObL#LXY=$hNm zTCvc&w{QdPVtC9jxf(ZEMJ3#2+HyU)4Cy-xUfU;FYikuJ0BsInOO{Zs^|tsv4xM#w zz~q+TcQL+zTUiqoqGMTGndh@vpD32O6#b%e7}AG&Pi4i}aQN3}=26{EhGz_gmI4dg5d$zV6a55hKmLRFaW( z#Ao7jxHEMwpY2(vGmr7{JY(MK z{mx=&gG@DhU-+g~H=em@*Dt6W&nznn$j?3#IwFJ0%PIW(fmghPcrGsNDDCp|266oG(Si_=)>eufOkauXC2)1 zo~gPzge}8-I7m+D!<-3+zrN?;0hj(eOn$XK|C)TpkFyreU>{>H`$zNe?N7zGKNa8p zRDAoL_-nJ+|H?|wY8Yu{9X*nj)qtOBr_PjSgA4b(=tTD31~_L)yPQ1-_u2KHGo%BY zy*JgFo@tHC!rx!T8CUUoF|zgyIKL--$@=N||8wS#KG(76jm8dQJ37)b8>X_hI+T^k z8Q%>n=B9<3kbEvUzWjMSPg!y7Uo+=5~LPbojGWGp-&+dwf) zk~R9g()k?B=P7buy7lBw(t9QJqt7M0U&;QI-jS@+y+U(|AAMZ0SHvG_Z{s7bPbgro zzim0@ti=@rwmTO;%A#Fgz8Jl?*n1TD{n40Mkf))mOV9cJ#7D`dTMjQD_FH4yM)-^3 z$a($N-oG6yWbfH+!Pk-}z8+C(^EEnMdn?$yyX+WPo3B6O@-@6Jf5P9)=vh}Tz6}M< z{qc+8=P2(Ugr8$P`&<6u>($oA;U@3M$1gEX>`QQVQ&T5Cy55+-E72|DXU5w0T&PI$ z(H*Zn_xt3RAG*$hE>+MSIb`T2-{MWsM)`khRnBv4nHjXdYnaN(rMdz-&LWqw#dDoa zofUN7tueeXtH7ptj`Ai`CkXy>zF~FG=6gkfeD0r!&d_wGZjgtrb2|M4rBiHsL z9|jO}g`aD+_2|S%MP}`ZKNebR)qe%Ksy2!jKE{2M$fDm-Zsa{@x6z3PCtHM*fgyf} ziuQTtvR9tl`!C7yaVDk$IVL`kpG#*8+KOd+1+e4nHNYZh>SNDO_MA_92qEr*$1YGz zNDDHi1slTTQH5t(;^fDa-vjpFf#wz^Zyo0?3YzF0U_V0PmS?r=d*xMwUNrVZfu6-UYEfjjMPG@t3u%z@~X!eKu{N0*6E%a4N{TThZc(TZ^puMw)pNGGhV?p3-VUAtF{z+<` zbQh-%EShWOz!3MyAXgsG$8YxDl_~PatQ)zX4E>@x^yx10sY65cqjT0;qvv76Y2LKs zL!MhzVDc0_2@g+!#><^^);-i)fqv^jug9^|+n5XRrOEsA8hP}zhSFYI=A5Tb9KLbm zS$95Ek&lS=mYENG@ny-s>duFGJgeiqxy%Rk)tg@z`4)g);o1Uoj!W`DvGSs0l=)`P z9e@w<%WsMuS>(z-=7@Y(&k@g~Px0Pb#r7a;#d9n9?(&*&d8{NO}Bk@fs zG?V`zl~#Y`9_T2~M9W5GRO<6J(A&%p4=vvU{#H-lUVT!S!q38k4}O-MwD-Lh?|KX0 zy0^{V^=Lo5dzQXj-tF-8>GIF(X1vMvXgdDWWfp5vXe%G9Y%r~D?YvYutn=)cN1P>V zd-nz$f6*_`^s^^2f;Av}B?e|kf4K8r^myF^qc{G?y}za~C?4;-zoum-J`?ss_Tn3V zo&AL9gUBuPi{c(yd#ob)3UsGjF1o8_D*VGb2s>k~(Wk}u?2fTkM()LXxQ~%FtaR?9 z;ItBZS~@{A)LHSp@cnG|7p5>zPQkZ&Ccd5w-|G8zWV!YMw2nMNJ}=cD$*1aTza!h1 ze)M!8`{<7ID_5O)kX&ikkblOjeRNI5hlh{W+sEhWnShsl;74{p3yYXlxG-uJ<=(cyuQ(VxAzUM z*zW<3n`KCA{cir-~vriz|sy|oWtxS>cQDnPsWNq`SMf9`eJrkQ! zdVi>Bwl8ZO>*4M?aEon+pV8;|8AEGyzQ!wmz2CL=k!_gffmb&>LcWOZeD#<6BJzi+Ts}8_X6TdW+Gq6dBi}EdlJf`r)*iRdS@+{>W8AYUt%duM zM;5a57qyNat8?rl-qAR19$-YUtgbomTC25d%w^ZHJZyOA>zcX4zWgIks*c>%FpX*x&Q` z?v6VzgS~&Mo%Q;8`@W0FpR+!vjS+m7oA`|a)^#CV{g`(u_?L}ifm3k32V0sm{6&%d_{@mS>6@Qd+vkUW_Hk!U za;KK}{yYdil$-YqZ9~e}esKH3|3IS(DTc z@--A>oO?+z)Eg{c_aWLkJS?=PHTdfK`a3qPmw$B1d&T3!)UCwkTxw;Re#G;dGvq*T z_Ol*Rd+}%3N5*Gf4uAQpDqqI_Y;GA*+x!Fkv%_5e^}td|oA}j_>Rp}TJOnJ;edk`P zr=L^6LflZ%a$so>9$J40SPmdJCIL$+@9EC2NrSLw;rlry*dF|=9d};TJ2+H6MRz`Q z2eM7P+x(lg-#IWT-tBSS$9?srL?FZ3t?w1jl_lQEA7I~O9drbm4+M^e*{2D!PxB=2i7xSK?o;JiEB2abE!kxFIxx_z!9;KQb+t$H z5PTzBt{vSuhkGux?{;`6>*IM=;>d3HnXv0b(@)XH_8Yi2oSY7uC^vH!`KV{p8N)k_ zVRpmRgg(Q|r<%Jo%=fyf<{p>L$l2%v(#zWWBqsuMzMl8D-s--asrbCr+7o__Snz+H zY^@CjEW170$VaZJz4E$D3fzCrtEr@ylKiRzxxP$JKo;ck7H8> zu%ASOFztk)L420Ubv^W6+SB)Z+Drpx?d!81P3#1p`fLY}`+zwPzRbQZ_{)JW2asnq z3eGCOvhii&w3@eruMXCo+TU#tTZsd}BRE``{4S`kJ}RI??7>=lOz>eTB|V~exP9cL z)R|k&Tg}-w-tL((m$|BSQ3X2pZ+K?Qb*pJBuu9&^2AD=J_apt-YuyL^M&9lC3&r=Q zCCR(!24uz-=KOQu@9+8k4E{dpXYO?}u?yj}$oD>k&uJ7n5I4X})M?$q_`UU>19s{a zAs2+xIM1Wht-&9*Zv;L$Y_L3Ruu4Ab`P^XntCt{qTfT!Gf4kLOtz6Jk>A!)0-krzq z7;~{09kQSJJI#e?1N%+&@2v|J;OnLB0QMbxM_xvo<^kFc({>m;ucpt=+nH5!xaHN+}m^MblsLd>qeXtcir< zT-MqK$6v&s297rlvvI7k#=jFPsu^N^LHi4Bw}+bdhS;k@uggC7_C3cPKe_)`glNy$ z*jc^f4`9N5iLlGG=q` zA+EOQ>leV4+GpzTR9gL(yvg92xFCF#ye~OW4h=NVEBMrS+NtB^2QNKNUEw^G9RbD(dG?~}oAMw->IgtF%bWi+T<`!DTf@X~T1z3uoT&|5k-EPKPI_e0Pvl)?O^o!D~pGWzK+F46C&)GvLBZ+*-`mriN) zQOi0#MjxfvGjaNm9jx_(_}jow`(fa3t%EoC46`06<-f+d@e6VCPR-&o#wRfwZqXZ~NFR>^Yo!urXyJWRaGA^v;qffO3PfsO6bPlPQ+Uk)X9EG1_hIaTz|!{qZZ??~tJ&O`im-`VexE6?&yNc*evLfi#c z+x%VJX4LQi_ znhDkymY`Eyd9T=tbX(Ua>4bFjMtDMKjqZ9zmNZm+EOb)7KF!UM%%glu_dW%VZu{8c zliRraN&dzsojy)EeMnDeji)}A)5knymF4K2D0*i(eH@~Xc6^{|=#c%Uk4)1?Cp4vB z{G{aSrH>ALrS-CtuSdpBXUsXshWnr^?Q6ZEeb%V#)*SG6y^TK;bA~;gBbfuLxP z@OFUj(ecO=ek(ujqu|VX??RkKQ}n;$zI@PXB>yA$w_~O+H1S2x7S&f5<@i?MlWa{~taucOOQ6%Ddm~i#4_Q zbN!QwSM%QIhfTW^R7Cb)ON@XuJo2BxKD+sG+B}O*{l4E14UhbX=l$w_-Y-n)AABX) z|FQS}9p9hE^I6{gFQ;#B|8_hsO^SKUiyYzs{UbcurNS;=c(Gj!cyc}4L<0nk;Plm>^RH4 zmVaS@-+Ryd7al<#*AMyX$?(w7np!K+lyly<@K(ui{7=TFap-kwUtr+@Y?7KRtNT{w zx!P(+uAHK+E%={L(N-0o+p<)iZp!2AX{h{pKI;d4i9Wu15}!@e*51By7XFC6;ege6 zB47oM)|}5-IE~*W__UJxjs_Mg#rpRt|3ZJU;>5_|*AF?LjEVP{pDt&*Zj#b(fF270@7;SP<+H-m6K9 zd5SSEzklP|8rG^ByC3>sUmTr9d#n*=Y~i`qXR4<8H2PEczsl%ZtxqXgi;FDi;0uGr4ara^9yc@B7~NXrs5iK^jAH+?LvZ*5JHm*Lw3A z+Vu_dY18BDe7}GW`L^$QTna7X`L)g8f*!)J)-H;wQ0)8iAtHtpX+^p4ezsmu(PJ_HYbmny;mX`IuKY>_k&JD zYi~T3-kn44SJrdm)tCI3S~o^V5U0Ujdra#n>r1qx@b7WGUqa5*9AXD~cSdT?RKsVUm;vGK2jlJe)?a*W zA+ppPe?>Lp|3Ex1!g1vONRu_K2j-p84r?|(VWAz{0*}? zi(ALKL2H#*K6?CWtuuV1w5O8Bd1b|~{c##PO>ukC^3-lJbT-EIh$us?8@Y~{~n&-jvdK5+Q)uY@Jqmh zPWd(cMz0Meo}j%1-?hH|xeH4Wdxbgl2cDg!>^T4B+OxUCJ}2Bw`J0rfU5oSP7VTP2 zyN6UxyH(VQ^Ih%k=hMLKwA;cn_5TY0Q-JT^`80HL_TF4w{(wWLZ+hfms{XDOT<7dN z9iPeRvu!5vZh!X0G&4W?>@Y`y%=Ss|)RPlu^$oBAu*$=b|;ekMx=CRg+Gv7{*-P8E}>@k#D{z$=kCnvnu7DTS7 z-)rdCYY$3ZL@z?$*A4IX(MNh+WBcJjWD)aPvc3iVF1@eyy`fuheB{|~WR^px{Z3us zRBL(NQ{%IeX_JS3E+x;J7q`7-DYTLP^5RmzHNTU}7`MjyOZioRGnIcehFtp0-S=nG zy+=%6)m5+%WdsaFq*{*ii8G@zO zfkkVK7T|IF-@#fbm7jh1x+lG+$e8(w`PdAX8x zQ0jVZ_-*(c;Ss^=pSJWw5dN+AIc1zXn!ROO(QJH;h6j;_QDmWL36JkA9davkF8XrV zvdkWOQaQQcRE|trK^}lT6P5RPOLo#aLG%%e*@RhvDaUeZTP$Iu85DGbH-WL zr+>RSTdh1bde8k696|QA4gXoUM=a)9jCoNS8fNC4)`*IoD&2~Gh`(?%@CAvLcjuRK zoaOO85&k@%{g01djs1Un>ATjjikqnSu3OLTuebg;K4AT4KVbchAF%$`4_JTq2dw}6 z2dw|?4_JTa2dw`uAF%$*@3%hu4nO#xqWa zcbVs}Ugr4^`+x4rTlwm|>r1T-^RXADQz{QmNwi{1==+?*Qxa{+Sbd*+bV{Nfy(1r3 z1-3=L_KS!MY#GD4l69K!7Bzm+36TNFDaqjq)wdEOAz7^7mAtR`lyb!a zvvzF5zGz22Y2B_h-5hdH%q16v+HvD$*7AM<@6Sa)DPE?MdvPkUaW|u*HsGJ!h#r&w zUH*WLCQs7M*!2FUj}{Kn3mR z(S9-Q@1T9fi=`K^4^G}U#inY_B{?&P@~xcDwzO6oKnw|(^e zF{^fn=|38i(M!C~vrE1Yz7+SOcw(#8w9{~@`c~DY+q3E~^L){zpBs9D zbHNj!FZLc90*ujof-Cb$8|D*u^{GB}ros!i%RkJA)mqnipNW3%bNG3+RsK;k-~94F z1V1ACb%kQ~!>>HqY}#l413kQY_5&vFcXjx|spbrg{OFU3AKkWbWph2c$Fg)cXrPNc zK%C397K+B9W3}tO^HhE7?-KK6wPoEH+4{`I=gZ)J=F279BOY({CE82edExxKM0-1} zed&=!muRo;gSLl!sJcXZ$3JL$q1z%eF45i(`e{%5U_tb|bOAC8p19NHiDRY26w8kg zhX=x}Zxp{SeiQv?@5?sN!doS8{0*EF+2gEdc3x^di`#zxynjAaY6iRxDtF8B%5A#z(PmiWmT!9Y)fh`LIMbY+J+UmYl=13L z@e;~p`|1pv-qrJp@d0d0fA8<`#DICP;_)Y&bD6`Y&f^))s_*S@C;cG<-mU=eOB}q* z{?s`3#97N@`$k7tz#Q5Yg$CMxnJTz+58P#93SRVkKez}+BjouZwE&_7p}LE zM2%HTz_-$oy={7wJ+R@)qW_IEz~^%Q@i5Bo(px#QzK>l#_>ef7VHFT&7l)L&$`9QOB$$3JIz{03=N2T7{py|H`yG zl)G5)%avh^g=>k&W?fZKaM0w7D&=!B`Bb*m7Du*G-VR<41z%h5!c$}Uy6ZT*RLcKe zzQ4*Rd18-lv;wP?i>ky*U*#R=CUR@MNQ{i$cWp#B2I@X%>{^#;&+cSD{Ce7)hwSr* z2Ur`%nSYhoVFl@Y60<(ZKXP0e`#Zy5U3=oSdtN*{{HH%UyX}hT**q)#4aI0H2lQBC zmf~N2>O_3uOJ~Dx{pM`hfauxsMOLGMHDJ#loxiwL8~OdTVJ!+IwIM&Nw@vjwhqLZ6 z?#|G9TshGsXDSqD^ou(qpV-y2ZkpeYI}Bcw4~ht+*I=e1kPti-GA&(+uFMjD;71cF!V`WcTq_Q^p z*8Ym}n#Ir;aeO%Cz#@M|3$k@Kwy*T|uNM-tDnBqfLi-2tGRa)(F=r$%W0W_558WRs zdFmaf?1g2txh*Zw$k{pKACXz`v~pWM!8?LQF_A_#7uz@?)~T!PJY3z1a{qmM=t;1vU-;+S>X-gx&uRaplGs$6M{IrK z-Sg@1{by6&pK!tZeKJ<>_l_T)*7+NgC!F!GQakXfx!1(C0}OsciQC;a9eFA~aR0n9 zOtimx_V*SP<)vHWYLVr#v(!f|G}j)c&MC`w^R4Ckv)F>r`|WGtH*}HsB3PbQ6ue_e zGj>=ccn9asfJeGT=YYgR+05N2_Y`B>G^>6w_4BA-%Dl~^{$CHX)~c`Tu|J}>*ESnG zIxvS6yL0T$NCmla{tMnO*PR*Bn<5p&2L0BV3mTIvf9?I2zWu&`;@pv!u7XSQdoA$` znlGvQOW8w{{l}7#2R{T(kR9>|=OfSIt?_wRXb;KRb1k#wWo(PDs16$=9Lh9nn7-T& zfu^6XM&4vvdo&mHZ`Q@3+arJe5APYfq`V6nO70-nUZ-x5JxX`~SoJO4V?+IR>dR&u z2`|T~UkQAyjgQw8?_XO>?j&F;g`T=gy@NeV;Z*dMKe!Hmb&Pl9qt@B59`?cYJ1?NK zl$=WxZg`dt&7^w^u9Pjto&od@Fu!w#KtE_cJOT{x{VTxfj>m=XUj-k}uv{r!<)dqnf+1W#t$5VnfpA*G0P{ubj^F> z@{tGTZ0kXp(&lpF$D`*p~E0v`eBE8O+kO3ps6J6B$JVCZi@H`~~= zef_JSCD0xIh7Q+8uAy(q%ijL?s{ci1U3^&m!z3c5p^TVIo zzwiO8SI)Zl(wyGl!PUj?8!zk!UcqN<1^Qx+^um5MFc_R21ZM0-tr^!lFl%kuCfotL za*@06Z(+|R7r2K5vk#baff+wWzc9PHd!Ywz+1>8_Ioj{G??UXuuL~4Ud=rm5^SzQeQz=|tZO#V_O#`MNFjY7(sXUi?Fq?Uv$2>0w#$Fxn(jQv7bMf9r zJ62($)BbGQXMW%1W9-DW@9@sO7l&GFb)H9bJ16@I9JJ7v887gPR%27f90#6*9-GDG z@Ao_Y)t4FntHAR=8ULUUF#gXo{%0>U{@>l=nb**~zw!6R&)p}T)E7T@J-Q%VJiVFo z0t?GOWzE=l5B`st;Fr6A4IkG1c~JmeZoa=z-y9(SYYX}$Co6PTc2Au7uk*<{Y3}bQ zc4hG!9k8Yb`#Fw(Ksm(1R{eTp|4`j+tJo@6uY4L@*gDmY*)#nFrWS>f^X7X8I_UxB zKeuDGOd0x3wzB;ldGMf>s3t~BeyG$qP*e7HZL@!>b=UBbv=d103Qkd;Y;@1tR`YOd z*Ey`|w;{igT|=)2A9Kp4B$iFWk1>-yQ=UzMuTL?Sc5FM1LF;A{3pkN8Ow611pIAj5 z50c-Z=kHJG?vUqih^OE^0%6|XE z8u=)fV-I9=j!e36WTIwGHuv)tmss68AJN2mV-+#07Iy}cTR1`f1CtAUGkU9-v0N;+ zRQ^0@)f=bk#rN~zg#6oyQNDe@yyvQYzntasb#r0XjDZ!Q$ExXP!N)@<2hRyTw!t^J zItC02fNjGy)))3Omap;b6y?nK8OlFA1o-dadExLk)>rdPZ57_h-JmyGi9r?CW7+#d zj}`Jfdq3ar!fqJMeb9y3jl!wkpAW9Jo>X37omJa}e_B-FwaMk<9yatzV6L;STjc5SaWhZ?^gRh5_xF;LthcC z+sQZGMK0<2>Df)Asauol?~dJO&3Iv>6}jcc#V4!hFF3aVIz0`YUW8729XdS>oqp^a z-0&0d{U~(m@eMI_`U!aMap=^|vts`cLnn<}bb68JPYwUk`XBR5Z9N5@7C@&*pwj~A zq?~9Apwj~A^m*v?!@C9BSuB=p#u(5~^Ol|Dg45ZDQIy;8+qv0($X#@C=6Imo zdMvPyfA}hGtnK$~0oF~lo4&fF2)Qw~g?i#E*_|d@S-3fk9v+jQrzCVL})HpRZJI_9Aqlx5lE4Fd+W*K=~z)|bZ zt)fowbBH_Qz~A^NxKvKeIAd&Oe1B}R=gxNI*CESl*p7XXioe#Ma}LazTbnI``-#BMBnJl0w;q#RtVT}^gg(j%qw}gC`VDrjXAX)UYA=ZJ{*; zkUGn*=e#s+Dj${jY&&hLtZlr-{l4TyXO0~5=#{zXmG%+P0Q|N=Z<9|29*Bc$ttG^R zPm5+FXxBqC(V=a;-Pd6U{^#iHlmmY~@Uy>U;Hr1v7mUJ@cuaKBcqBhXTj4{zS&44n znq}+RN_4JrVw&=fP~u5+?kW6C`=Ns=`>po41~=H>&0fN>%H|QWsZZZW#Gz<$*ludVQ`fXKI*`gU=V)1xY-EZ(tfJ(W;QV9qiH`S zX53kKnt9Ine*3TF9iSbpMcuN}z3;6Pel-uhI@YXbty`RPiME{+MfUivsy*==#jj9* zNmlKNUz+dPO}RO>Cw}FAA5web7kuBR_1%Qr$P&(Syh^SclM8-d|p%y^md_z4Z6@L!VWI;7aeS@v-)s34Y?q`k0Q_Hv)s%xxBs8gzx>sTFVyuxNcL`HH^|nQ3(oW(=CYmB9n+>$!Vk8@{>vMd^+B9=ba5apx6SMP=-b9Za{z z8s8*#pW2v(y)T&{UljA1vyJvT|1Umjug&zk3>#2>8s$gdr2X=1t%PO#L03dx07kv< z#$F0W$&$&=`}&c9>D#xnApgt-$p44sI3U04Qh=)M~N`LCQ5?WPPR z*WyrX;ek7ru8(mq62a+1iJicQ7{q3=isy!a9X+z+1AL}>mLy#+{s1+U(PTOcp zdq8Eh6~1Bd$w`w+irT2NWl&n9>NxO6N@>58oRwRc8+`Dz086c-yMLv8pmH9dyzPYcDZ_0H}@P_Lmo6sfO?u2j9i7nXAvHL=a zi}8*8$*PlQ_~D5Yc|ofv4<0H--o|;qBB*_F+K{ezrzdHy>OIY!hluTwA4R_vC#|^} zWUfZRiP{kTga->BK`%@^6aVUY0eq`<{8r`vgjdP!xeESH6u_t2)6u(st9NyWWuAJ(Ec<$`sxsWW6m|!!kh10zP=qkHGWO_NbAEJfu$8aT}b;Y2U~jz z8JpI!qky3}*Op^11X=fTf6up(VGWB1HrDvEy8?f-^UtcRaA@eUF50O1xV5kgJ^j#< zRqO8=99r`h@HFMVvHl*`VzYeN)$_SC@&WaSU8Hq!Hs9w1>lx;3fsZ>_;q`j@Ph-D{ zeXnZ;hhT`o%Z>QdR#nZPaFj7UWo0+a@(rvGq4#E`TMLgMYd3ROQ#E|}7WPLq>&l1W z#nHE3(Wtw=Uc4&1@j2E>mGrR!e%Bn1v6dL{ZqHo@=;t)^p%ZzdyKQ2~o`+~>67K*{ z^#wD@Ty0k~zhG+_kDx zN!zbl!^V|MK480wFHdIU@8#KSU?^oyRXTHOa~5-oHkT6LRSI0_si85(Z0`%(cbbjF zmcJf6nYMYqp7%4Un}dzii7XMX2p;J$jaOqeFtA4Lz~@r|94>Dc!rSqQwy#wBp%VSD zmG$2D*wYtpOD384ebzCC2N|FGRr?xa2YsmRN&c*H!fOTV1-I=Fscq^!4y@j`pY5k@ zwJE$A9I*xzj#ddrl0)=aM=pst<1#W5*;ffXdPn;o%CobKc|6=VWW4g~d*#8VJFwx$ z+c2x%66)oiTfAlo^SuJNOORKkjInnwI$+zkNp?~wFEa7UWIM^#f3J;5mSM-q4e83S zY2^F&#@j7%^k13uA8@unix_KSSO0n8{vvRjxH)K}oL(1`V=X*Sl3~ne*(0;DN3^Ht z=Z>{}=6t}}Lo;#(e8~PVe6047E0SNWlrLkRW*}1zGyjetQ#KFsH^!iA5Be`~=acK> z$Q9vgD13P-9eDDK+V-KnCm3?{9{2aIRX>x-|F8{Nuo?U><(c?Yz6af}r7<0_ zhK%>(5F{NtGgn5ji$I*#5qZ7NBk8Q|+Cz&&w zvoabrzZC!QBz!QJxjzLs{Pf+<_?CL-KloSfCCz_J^B+DH9GdHfKf#;yVU+g8Yh|<- zX1x_>u6udEo_Qty{RQy}I^TkS+2{!69jwG2VBYQ4og65fQO6JSh0B!w%zVhiShrZ)Gy2*jZm^{b6W|eQoa%`t!Q-MVh-_`8vs=`6T@7lHD!&$QkBv8GiOaPfEQ)!-0r|C{pDR($SzbHh8Jk3e5lrNgCcdb?L}?r}EaFDZKT0 zc+2Id*Kd|S>5Hd}qC0%>LyNP9ij$YVRykSzWn5%GYp57!f#s`fK}QDQkvKe%Hj;B~ z;5p-#lA_F8LJ7@T*=B{@6CywJaGzug`{BsI*N_pN$N=5nt$bJ7HyT6zba1nQoTnAc zyYzyRqRAQ7_~F2@j5TV;$daNF;G279s(sAg49XAl`w)1X!n%t)O^?fNRlAwFhu7=( zaw~JyYeO>|pF@sxuny}$ZfoB4fb-4tK^~06Yd56fzpq^*n_9M>@SQ>VJlbj-me#2D z^c*`nA{a7q53QH~PclYuXB7B~uvs5QhJOlvdJ6n^pkv=e7s|eU1iP^C)+-x_ zL!WK%UK?|H7I$CPl8dVloKDLgHvSB2g@>X2X~reJoMgM2y$51J&|TsKt@XwGW^KpZ zFu0H{$Gc~+1KvdLzlHs&eJb(J4)k3%_{2s(KE2@e^@TTvPW}`3-u~DZT2suv%Y61- zihWmfS>%Rj!?z93U;EVpdtGU-)wTCw+s?9M%4kdbsddi!oP2Dnq;K{ji_4i`W^FGx z!A&omz~hCJ^WDg3d;qESE-{`Rz+9Yd-^*J(n7v~7Liid6zUG7b`40XR=kTZEFBNx% zzR(HI=#Q5S&^pAA&xSGXapzkLO4ccMp-eLU zN?VqeXrBc9sD0771Nw@l;tiL_nHOWfPg}Q$ziCJOjZRItHt{yfymh}r-QlY1%!PEt zdLSQt$P)J7x`NQoTYn4n*%vX-)%IX~P%X?C-z~l(|1E9|h&_)a1Le;<%lTdD)mQt; zuWs@knEO@w^r_C0nON&V_Wcf=4erhcj&$gmK9D^w<{9?E!cFA0)3Z*=E1nG>V2#(a zCz-patn}(l%%ybn#&*Vh%4s8$dh@cf#$_`9dT4{#jv0EUHuP+HX4ZH;J0e--L)H-s z9-e{xg4U{^MQ*2U<~R8O^_%%|kiCM&tb^+-p9wUwf3Qb))z>34f0gy>`Ze5#-?l!RW$p$2&q2noe#}}liaS%D z*mZ{AMjy-=Z}dSszn>}aI--TL6_kZkwlyPCf)C}_l<|9~Rds75m*2l| zf6pMMj&=V}?cdx}dOQ8p>^jEpje_HbF(y~j5Bcr2n?&2k9oj0M{wa9d^8C5_Z5p)m z#)|$YV=cDGi#}w-`b+7|N>AIe$5PKLxb4qRX+JiC*tdYS_OgB8AJ!PyFHf@j6sK~U zvAFGL1VbzK z!Aoh};|C1&E)2*{?OA>IGBC8FA2s%Y{lS5JY4MIf1;3vl-|?mB_TTx*_)OhzDNJck z<4Mi0(jU&EDea0*TKl;y*GsRI_CJ)eSCZ5=ICvwa-Bfrbd%UsUYuxsnIB-|az3jB( z#@ou~-50QPJsKF1vEJY1*kNuR`GvjDRNo!X4zvH!7w^mcD9yR}8uDmA#(5MsUOW%G zA!gf(J6(VKgSTKSik|2T*~_z8zsVmu8l55ELumm0BHw#lakaLu*z8eI*ZAIfcRRM& z|CaAvYcSb)x^vxPJ(A={ckmgYzcObnI-4(5cK2xLg1x4g_9Qxlxo6p-gZ9ewT<@1! zDZXsptqAt{P>h)zKL-g)a#USvG-gj*uuP60X>`}fB`elv z;j?}dII;Rnyb zXMF(vkndq-Rn>%I`KYlEW?5GEJwxi(cMMsve#d(~cO66aoSqqKmjAmxD@$!gXs@~0 zmsVX%efi}z&NI~0_YUCIJ>&PJ_E&NOBuUD2j^W@8U9@)=48Q+h5zCfO# z%Pu3&$j@o-ALa|z|99otvHyWQ>-SwF&)_SMJTo!b$g{4C%Cn$k$VKECYYEA-!CpFH zTU}C~VP{-Ko*i@L*$+nj@5-~p?~FX#>F%Su@~n;+2Cb9y`%?1kBl1)Isq##5T>pQS zXV{NMo(<}iX9?t4*9Vbjf5$l&;r;{5Gx|32Y;cl1OY|$x;PDiB*2|06i5D*;&n$E9 z)yT7Bl4s3XDe}yjCuO-FdB!~HPoA9~mZJYPU#{&}|M$Y!UHHE{Uv9m~e0izx{mqw8 zlI!A6oiG3V|ET`Y@nuvO{@>RB?{mJ$u8{thEpmbW=Zx8<=1Us-e;NA!N%a3-^#7}O z)UTf;{ZE^dm@Be7WUELYw*e3Snw@j;cW%cQG#>iPzw;zEfpV@b!zLKcckQVS=X+T8 zfo$MQ%%>^h)ektI+A{l}PmctX=hN-|*#~BSwa*_On~KhwZpTXH1A7?XihSfPWtQ1v zb^PPKHo|p+7r$Cd89D-+a*J}zPq+1j&Pva!wYr zGS(T%dikL3+e)(8^ZmUxdyGDNpCU3fZwon^`vJWzUocb#)s*nWh(K@Jda|p%O7E4h}l#4 zTow9L@kaP1;(OWWhbP;>MLYJbIcrY2TecbgD;;9Nf5M?-vp6^wkJbazUe*LLVk(yd zQxAFkekI&XsdY#TcSq_j65WsB2Ttvo_Xnry%1^30goKmukPPib zM#_Of_`lA<$HjO_KHpN>kB_nax~0Tl_1BMVLE*#Lg5YCWFOAZBaWaSX?OqQ~^m{Tm zxelBx1D$31wW8}PK?b2spy$777&TP9qT4)XO2H(^MAlNXz> zCSHu^djEnoHt%}Z*fY+i*rdFtHTHiHD`IR?eka+a$jmrxxNGb$$u6RfV3Lo;tg-oi z$u&0fN^9(&$tINy=6%I=s9o+f;a-487u&^XJN)s^pAcaLPteD^s3~sEL*4=L)Z;E{bsx9Es8hi!w9GQ&2Jk99R z`SA6R;LX#>DPyw^v2E4{=0y(kuGlxQtDJXaQ+86Xf$_cx|8JnYf^x;M>FjPNehkfn z4(5a6Ruu!^!oH5~7aq-=2w0&Ly|_n~DQ?M)4eQ89&afUm!`wX$okW-U?Dwh9qx2c0 zKj~wgx7mh1egIxkOs?jl?pa=n4$cFH(iES-wy)W|qW0R^=TXd;`~owPd#(5crc@P7 zF!zw*Unm`t*-(nS(U>*AC$Wd4KHZqFarAjJeNLiJ#SKg`efIeX&LHDDk=wJG*Zx9V z#%VqnU%?3DE9jk5Wg{HB6+Sd`#_KC^4 z&En7Nb65U#=9u=PR#zSj6}?Ivj?VAbEB=W&B3ULsvSL2$m^#G>rZsI)oF(%`@=P%e z&@$2YJkXShedP6Jd%wH#$M*mB?N9yEjgNQs{gcMU+In2dm{R@8ss3-_=C#YTnVZt4 z;^VzONA|(G>US<(PaMu}@+Ea`1XtPzPiMSKiCY)m=7P5jVg==slpUl!@<~?u(RAUF z_!-6gdE*h|@Wg%KPqwM?>%rsh^Jc~PB99Ed93PSP9g=@1$4B?#6WF}43!c{T7I$^R z>)0*I+p9gaBQ8#y_A`=br*k>!)XQiZ@Qkw`;lcrrd(UEPucHlIw8JOzef;>w3wf@U zIo!s0+ToS{^PBhi-`kPJW}F`Xzxq?#FL~NdJs;uaXP0N$FN9BZZq$x97Y@%VhR@;I zKQMQ_JgIr%{&{hjcK$Oawqk>Ie>Zj{GB3yY*k2)c>YU$BO}v8K*h-$st>8EP{3lOt zyJPXm`l>nS{F2es6|c0im$2+9pOGWn&6Kqb`72wZh4!9AXXuW@I=(NTpzoPo0qD4w z^Vt^mmH2B;b$xRiGR{6PrF=``r?jRRx%0JOZO7E}oAOjXQ7l1fdEHoITC9wt%Rh_k zT$I_k`F#1w@?pe6GhbBS+s2$x^(Q8e+In{Ks8{)KJ)2{tzpx3t;lef9$2kq&aber^ zv&Ivfe)h`QO+Wh;|L>e_z6^}Qx%%}vcrJGEthIyoGKA;Z;92sygnKiwQFr^nb18Vv z+OsZy`+MtNQBI7*r3vKnYRk%uxcO1GvX3YE;Nz@Z8*`)9 zoFTd@GN9qTbtM*lUMnZ!Lstlgik|`YyIRK+7YFR$8qU5W-wV(y?8}i$#Y(vSxb3_B zeU1Gx?NRccX-j&klD6W=WZ6$&qb+^Ur!C2J>8JqhX>9uM8<&0e*u!fPs z&I;sC8fflNj7vu&tI{@bk3)aDhk5#=89=#X*Jek)wd%cf74P=Eox3m3&fDXNh|jhKdg<3<(_!eXw5jj8jD;A8 z#Iny`xTh=l>m2xxd+?LGANq+uTqqCq?w6c*-ThD(zxI8Eefm_ga0&9EzZ?m_BG=ID z{Ha9;(D&~*kHD_~MrTR(Fh22*U-wF2U%lV=C8xYHSLeuNt3TWI-a2I69-UqLQ`^dK z0XEbAXV+Uyu&doP@sk>VwE4;AE77^~Rad5xIv85!x zT>KaIqkm(-QM&IAlZRI4J9nO) z_rR~uj_B?>d&tT>T6O)h6DvNv`|QYH{^V>`^!(Y^t*kfRSoC`1>x;4*3w;CLC|q1| za{G)x!fiwE9H!2ZMTZ*?qgOik-Vw+gHNuzqhU&dDh`5~%X^rTzJ$Z~%xUe>ESTCIu zCl>S{-La(N8LefYMZrGKLx53ji(e$oBz`YxL|4;;@%C(R;myMkCK(KJwN zuO%vZhMeBLfoDGApp*l!71=6TpB}Oz+j;&ZI8s}RC#uy2X(4{KQey?_9xHNX4ubLfy+Fr=Un6pD{(XJs1E}_eDEK-pB^|`fOQMw zt(9&cVKuurQ2Eom6B|+6taE$4^V!uG;#+r)J7u+si@Kr6#8a&anD{ErEq^G|LM+wS z_-*=j=GG@18k66yU;4fdzk1_E{LxU+QtoLN9n?;#2j?Z|8oz^YZ@CZK#mF#+u7B8)>_!#@X(ms8nF$A13z{4xx;S0vF zHDwI-{fwb9Wel4wVmz4xT5HOOCEG=CCCOj-TClf3FV7y_6L)4ihJ=Ym{>0xyb3q9=X-W1^-o#SZ)YtenX>x2J~{kdXfJuv9q=6|2TND> zNb&5@-ZPoed5p7~HR2<-ED7u*2k1n5e##&9H{5L>ormoRZ| zEoMEzoNs%CxK#M4oIdt$4Rnhi(~u#1;h&^3WXG_foGYS@JYbup{-`?w80ACmaA1>9 zrQ>#AQ4jHmRm3)ARoobv&u8J18zX_MR<0k}8Y9g2J4iV;ekiC_jp6ngiiO1%3lqN=o(jp!A;TL;=b?F1xs`rHw zBdC8f?;GA>-M!-l8}@YC*89hs?BB($R#C_8B}MA*!#qwY_{nUO%+3N2cDse0p`sm(w-8*NLD%DhGybwp#yc|88aEPoEq~oe7E zYOBt{f0}JomOdA?&7+i-p-pHCIvuS}dX82&J+F9$4 zxzj4To-((uAJNxyy?tdR_cfKirqkDS`kH$2zCP_8|7okp&`>-E9(wL0h6Z|+TIc?G z2K{QD$%pIC<>|3yzkthIWwN~Ct3qChrbPf zM#uF1eqO)tt8K={9IN-R4LP^4pa_{cu7&TpkKRU1PxR$d=80rNZy9&EsVvXA!_5m< zo+D3ne`XDOTz;W4_RrL|a-X--w$}geFsCP(JhFkVHs*%LtMZv9mu$e$f0C0+R?jED zn>kK*xmPe3G{0591=!i6INk#O()g`%;nmE4+W!}1PF}68TU>bd{a=Ddv0gFYAr9(z z40t4~H6Fnd@O+mngJF~N_aQ4Mf22=V$}hRcbN+M3ha@Ze;{C8alI$(fPIlJNCD40> zEh8(2=$?hpE+6)hk9iKBB0l7Vk9lCeGnd}{P|v!a9PAjq_9GpFUdvymY=c}UF$=JY2NsdBPw6w)7o(u zxO4LG)yzTOcYhz$Z{YghnUl7^qFlDyjE|c*{XcMq%+z>s; za@Xb(o+P*W_@A?#{H`y*HL}c#zr2l@VZo(!Yi~KSQRT}V*%&(rKhPI*F>&suo^|3g zos;<^Yw;jwsr2m6b9&ZQ^kj~g??HEUAmdCvr9qr&)$=L5r_Vuo@E*Kbj9<&9g?+~S zA4c`8vqa+sALSfHXxBdzgZ*8~#Ahw&`=wucZ{5@U9yYpX-3xpMO~ho;DWJ#v*=a(#Co|ALBoyd>LO!H@TO-N&U5|k8Pp+CB8*AO)T&pWgdNA z&wxd~jIU95CBLiq|1bWZ7T7#~OwBT3d{My2&lMk&Jd9#W3Q*(osV<%eHO_5U{cJ@OYIv73~mSK2)EBi&REER2H$|p9@Dr$w_ zC0p06>=fX+T+3$Lst1YFME)BVPLMbVlUSd-y*q~R1QokuXN;MefqiHwW`}keU0sl zyqK@EmYj9v^cN&&T`ll=e(qE%*{{ZYV*1Vq5(w?L+rt*3MfaCEWKz z4hTbc(fH@^$JZQsdgou;-_jQ{J0`y{vj445UpUX+yYkvzeR!TVel2+S!Jo!HgEzHi zukG;{x#Km@tFYgOhMIk?{$~BQ`0eAjd1olLTp9WX`SBY4n^+TcfD6M;V2I{ZpR-%O z3Ha%Ptf4=~=U6^Ne4>}Mj$Q%Z1k%VI!S@5$0?O&W<-MMrVe8`&?c>I2Gd>=EXMNtn zS__$yXsIIxocT8fe5em|9};Vh?~O%9^ShEgyg9V5IC8aLPWw50pH2H4X`5*ukzFhK z934#`xAQ$8n~OGYqD{3?9ttIp1t#AFI?%LJK|A_h>b7Hzv)chTsqGA+9p?2hD|oeO zH%qVzU)VQ~iC)H+V$*AmgDa!UY?>VRK@-DsyrVUk_YD0~`jh`$e#bKE#SZ!+brXfp zKvxy{8_q+M4Xlqfw}OuX+x1qW_)&74&p=Oc@4@5PG~s9d%G}3c@DWOn7(c7<0nS^( zHhzCa4t(r{)nV*}I_9(N0nJO4PGmvRR_3wJZ~rNGU}0y^(0yS`X(JC= zuoPJ!yGq}M2Ym|09C+NE?S(GDc)Sw21X}I9m*cf|9hwO2&+JLkN=NH*B)z-NrC z#b(>}$X?_fcozJIcS6@hJ_8PH*<;(+^8e<64~PE8MSr{Q)sg$CtLMeQqB^DEtPGqP zJ^^RN^yzJ@R&r)oUStz~)j{N6SHG6?r|({woku^pw3o|$3tDre`YTo6BU6md6A!!f z{@2^%ZlYEAN<`y|ng*j=2lE?!wllZP-C(6vw-6^vTuSv4+L&YEJB5CF z*KH%oM!T|RS9GA&yoLA)gMUZAySUb;SDw2%QE&vK*F-eOw3d}FVa*>AEq?mBKK(`6 zag|B`uYa{e24vMogp^!8yTeet@bKu zZ?cuva5Mk&thA%#kUh#e=BVlG>Z!z^r628hbJ5AR^UGGHTj>q#9j~@%Pxmv$h>1yX z9?-~rXcwb>GnO)Yen?LaW*+>|W2;_vd?Ws}@qPJXh_dCjHoN_-n$UXwmDd-K+W$U%H~vbEfJ2;o+8>#yNkea{O>ZzH?lE&Xc` zXCw5q_FWtKQX4dZ_M;w2wmn>1{dr*4nnv-)UV2RH2cBA!k0%h>a1nU!x*R;#u*m(u z1A>Jz=Lo5-W731yC*%POtSL`#+F8=KO$-ec4?4M%ponL_LDdF7m zCe|Ow`N#5~diumHbe;4^AS+ZH+6t zTfS|NOZ7suEgA61Z(<9eAK`~lIoNSacxL_XiIbL9@sqC<^pp!sQEr?K~V|s zSXRh>_QmPVLCb3Lk)uC%$TgAR`3)<){44soYtm~R8ftyK-osDb2krfy*UcSJSM}Pv zo1kg#M$T9wzc*pi<)Mf3vE_2JZ2orrfUaL#_FxSA$(|QCM&8Abru`XgzwY_iSFxGs zRoV!z#UD4u?C&0L<<7?X^EJ=ZTZ`9BLaw!bk$m0ECB?>~AI7dAWt(+0#71TxI)I&^HWVu}n>M~o8@aUM~9;fyTw*U*8t+KIRw{5)jm?RVt=fbF;8pWk#lU#!RdrW)SIzWO0}A91W<~`>9Dk&e%ooeSpj2PQEa)g&ri8lJzU-nhrA5T+W`JYkV zdp~pZs$;{LcIm)()t^z`ji-Tc*q>3}_fJz^_F0rS#jHO`&9|$6RA1*l^mn^;KiSV7 zHfs-e&#EisK3ePLqX)Zt`p#|0^(xjOgpnEg6O*O@xzX>+x6cX{(+nHRl+Z zv`+-k717oDUDZF-Xv!j9eX+TVMg64Jk8yvfp^|$jN!y6LjZ?bW{c9RoGv0{pw z@FqVt4N1PN$rfa`rt8jwJo?&GX?a zK0k|}TI)?hZrL7Oy=C9xE4#PzV|pw9HK+WpuboOuZ~5c?%<|iv^7oyg{QR>nKSgiJ znk^Ig4fQU4JhYcAu7d8WZMuUG9ry_RqD@!&B1<0$Gtb&ZiDj4 z(GEAu_QpmT8)ZGq4d@T8Vb)$6`EWY$OVOVvJIJ4THEoUa$;h+nj$sp%8iR@svd7%^ z_hdabIsHWd9ily1rKDGXBzUx!_BqzQPh^jY>XncCdFq`_z0zCSBPkzu6LyIFy-MTm zyZu`y&2!jK+IPDSJEi4(8-6#cn*?4e7+5Bdk-r|^{jV}|&7DK;B4Xk~mZ`bdZFfQH&9 zmk0iZ&{5|YOLvwm4#ee?Fnqwa73{*dm2xz0mk-;(@3ifF=-vX2_WlF76CZ(3L3rY= ztLSouy5GoP=z-Ul~t8Sa>8ii7_SkB)Wq>l7ouedCWT^osvb7k}hgw34qOf!!}( zgZ4wP#&MHR$ie6(&s`(&6qeQD0s@)-u_mvX+*Y~n-*(RvQF z#;-T_d*H5dX$EmxTQ9rWl~*^-x12OOe_m-uAg@~EHiC82DoghD2FltXj;`+uW+@bSW@a-O9e z$*XYFpNpT}Kc(^;{m_(eeCXrI<1c45Vcz+_DZ@mci+aX>$I-tn!+_yJ=CjZf<%vjR zHD?=^GL|ljGPdBH=ylL^W|(y-z%Kq6{ExDR0X_9))~(#m_;Ge!IqRReivvFEd)5=> z$SL;!pKSV&2OrfCd{m3K{FR+&_v{@N-aNZ!{iMR=`8j#MFtmG(?x?694J`N!%7H~| zj>OCIT7G9l4z~eg`H`~X9qel>Bdis^M#a+sSDP~n$5GExzRUP-Ektg)cL}S(@ z#K?KTWL*kOf!JHjK}+|$VJ`tggfi8ytY@x2BKV&k94lbXX&bU2eR~`4QOH9h@s?!b zheKoQxgT)JQ5$D3EJIy7J?W#<8kWGKR0W7O)Xb>BwhDXcXZzZ_=PJDdw0G+rpn2{zYgZil;GW)BE9 zM>4Pv(2pOs;LDMXp#3`5(AZ^^4J@my7Yfb#>!8DXwc+m$M)y*WxBdF}|7h$mXL0bE zI~H$$=5_S#NOLb#X4hQWT4~_uE1CoB&u`T!--Y(EeW%sYn~ohf+qqY&4gMSb>A*IN z`L%*bQ7izAg_KKw-IYT>fp7kmQ_@9?Vt>M~YQoX&@+G-8`4aFj_5uDR+2q=H?dqdL z+$rhRCx!T_mt1}Hm1W>HlD=pn^9k~&JPw^_S{a8MxIa?;PbW4)w)zI_Gx^H357O3M z_Bcgw=*)+$@NE|DKa+b*b#}pAe1O=*yHeX12 zFJuhblf%84`5oMwd8u2*)oEpH7}a+fvR68~YxcfwW!4F9!73O9>!w}zuO1CQz@e*i zP5oNoeh7cbkBQrgo)x~YbKkk{Zrw_n(cruSocVot3gv79=iYGLUf{x+j_ZbSFK`aD zaYZ=OaBb!M6`Nk^xNhhE*i-1`l%etPL2!M6z5V$Ev%1QGG4qZ4S66|YV@oy}bj-?yK~9ggtb#KCvs9-JS@ zshb<<*DP90_IN<>G+qizU6H3XD+u_{Ac4E z?W>Hw@XyHw??*D9*^iI7llg7QyzWmG-V?K}aigHmR`kR{#v6%o=pOoM&7B>h|GkvC z>-I?+S7ee$dxaxgz9K!0y@KCX_U@`VIv0R_Hd<$(ebpV~@BvNYEP3?Mr55|d&?T+N zr}U?MleLdCuDCpsR2g~Dy#xEe9=oUxdkySu)S>d)7+1Dq-_Bw#E`;w|_V+5*SSqi| zP#S{|b-t&ymnxUBkH#`yoo4%lhQ`p#n+uwXV$W#Y##s}(XEAElfb}=`EG{v>?eK@) zaPC}OinEQ!#7wv?Z^8|;C)IzgN@LR{ic4i zV<2!~6KnoL{Y@Kof#_e=%lR<&eTyl3wq-nWg8blXOUmZcJfr4L??ewvE;KgPy)WwH zb%wyr%)yK7=>p1V9y*nY($F)IY>E$Z=IE^9IR^0Yflc3GQ%URra zsdR0jP;&|SR$||19$agjcA+nA-4IIK%jAabWSvlizVOIC{6hHJ-{Y+|tv2*+ts$a5 z-Br5T{IRjje9#5QV$4y~bJb1QmGJ~bmxj$PM+{7Mg2g73_z zb@+GCMl$Hq^D5C38Ei#PwFY9J+9tm%p|95Z*ykMtoAa%pu_tZcRytkfe^7bedg;Gk zS`1z5cpm0?i072{3XP62hwg2sTYHgb_7Y(qru-)f@cGT^Zk;v7nh;OjuFn=7pt@Wf z|F|CsI{HGa4Lz|D9~<#6dy8V$Yv?!z@fvtEDRHP(zHH5q%p`C&=CK4`_X zHa>`cVve=3`UdM$0ru~4FHrGoms+2$7#bL$;<<5k)7V~VU9 zZxB9p?z~s04ank7znuBASxXbR9(ly)rE$yl5!|c5dd9h#k;YbJJ)b)zS_?Rz1Rrqc zO#1fQ@zvhO{4D))HD$hte7E59@#Ssv$$g-YUWK|`6tF^b&_lJ_!@Y9 zEi|mQIPcir>-%lR6J8EJ?f7p6e-L|i0=RYFhK`?m=c@;1O3!3>9U4wu$WP)f>8PxR z#I4kGdxUk$vtDHnQ?mSL)I;4BzoaZ+UB|kA+Wls0<)By#>%p2>6P~y=lKedHPvrCa zF5p4_zOQiLw<-!ewb$Rz!x@y54Np9xbQh7z=Vzclwy{L5H)&hKSe&>;P zBXyd+&fuhVL;s9k)g1d2_EN|;Duw@7IyjetvvjERX)d;30^Lyo-m<}$%&mG=dZUoE zYJY}C>~U+GMxTrR5S?UyPnZ#59cX4(3w=fT@kp}5Dm8Om9V=LKx{iBlfbCgiMEUE1 zL;L>jxF8b0D_k1y7&b6|pm0$9jy%>kE`qOLgXW94Cy0J^EA(DMyAU1QRzUCjBgt*Z zukM5S>lvJ@a3Xu0*3p=DQj=lN3sFbl2=ssl)=~5ud0Xd`r-m}_1$Ja-JoJf(hOd2w zJ1nZV-0q>_7i{{jvZkMFd@?f>S3`O5_^Mb`a92)-24&Q9mX5vrcelmqmiX?yPrhm~+ ze~rS?SGx#bLmij#goQ)J4Y6(b)82cZc>1v8TEk)Vqt(DAFRWXHZdLfK;OP>0Y9Xia zvhl-91A=w6v}egk5Pr))Gh(dvI%d?V-P!h3uwnO8p0-E%N?*m< zakjl3YT({M@R98uYIxoJRzDVM*v@ZwxXOA<`(?hqWsAW-+r<6Aljl|QJE!4$PTrTy zJF9_(I8Uh@t<@+;2ey?X_vpa(5^O(oV8?jQJVXC=jzdqN`&yCt>@Tnu10Lu6jCCQB zFWR`|>v?DiFW$AnSI0_2(Y-%`hHB$qa(Ot{mW6E!FK6$p8(wRN|3Kj-$gLf|*_va^ z!pjODA4wjA_Q=j=YZtyE+WfZ_J}#2{fN&!NUkIB$FL8wo4)h=HA zq}KT;ZW(gYajn%@J0RMK?r*eh@BD`EQf>nNaHdgX3vwl$U5idjpdU5P`_q%)g&xyb z3cs(Jzmpy`a|*KKuZ=cdBe|0e1%F%e@iU$W&iHhH`QXVln%4_hlui2#Xub&j7+}sq z<3-I;Q0Avwpnog8(YSR!>mT%85CSH`Mr2#&d>oyD7GQ2L$X$CaEvy;~4H-}pNU!llUoel_h; zmR(oFcxwbYOTY0ovVP6pe_X0G$a&!l#KD6ME7&X_=qJ&`BcDY#1KVdvn`VPK^CQ^y zmoeU{E2NGx#y{o9Z`v2RV)DKbH_@NugzCz8@*~`nApbl3;tbbV8E3?jwodg+uS=Iq zkd6suUY76Zlu_uD=NCV+?(AN5(vyB)hg zItN>>&d}3=Q|*3>c*c9P^ZpO>Uc5!$9hIDG3k>@y<8ID^J^)-V;KR+wr?!#$;`BkA z=$|%Oxy|_f>h7j*lV4`j&}D1Y4+eoRWaTs$1wwVY1Fbh&S=6*Ox8dcy^gM#*U#a@fA3v{*imb0p))`pgXL zQx;n?e$FS5sS z@z?ilva;$n!MhE}m-Z&?Mi|{wzVi zDw|5AtK$)|;K+u^b)5nTD5@z@r_=$4D!qc_L3|D64t-@#T+YlV&S0JjysbYwn%b zt9;e3HmdVe<(EuQej=pxqnTX?k%0{Lzm!qH!y3ZojN`SBB?y~?OuzoX*Z1j6$2qKp zXd|B&cKNkwMVT*|@^TusdCK}Rzg^wfBXjUqzM+mCw(Nb$%jdzg75FUwPzOH1<&F+@+v6M1 z)Y~32UqnthS9>nvk#fOeb4TCq*Yo^*zI$UMZSw09j`_TbZh zY=ZXm(5^KGRlg(uq1vO_uZ|rSrHyHykJ@J|ZO*i3jax+jW7To?_ZT}@c`~&A1X#S` z{&8a1hqM2DDewNZ1bu0X`f2*>Hel?o&Za(K&7Kf%9Xv7Ke*UC*Vnra{`b03^z9tlx zEu(S4weMkvSlGjntXNAGXH7rdy{3UM_dH7Z8EhEs{?65XdjBQrAL7<;<(T^OssFLF zuHSQh4LHjW*jfVq;4R;A$27*AjO)6q`^CF&w~`-#-;s@hw0-9T=`%F8KFpduLx-kZ z1ILLq?2i^4*a2GS+=_i-=n%5kvD)~5{Q7@~?5qAgJ@xlD_5X(YC!ewY)c(c3_}@Bs z0=%CX@64YB?*s5Y81JkP#arRO{7eRChxbj?c{;wI9DH8S8UnMX3t2P$9CVlM-646K zWSua3_*WozPat<|CQ;wyG}-H{&xrq1555BpzU!%bfP=50b4_pZmeS$C%^r)9v6hWi zY&Ci9JofrR@kW1MpA6@0Bxw(g-Q9tB*yQDIMdYD|yz)282Q2?tD``wwJJQRt%S-vR zkf+A(nO5VmC)qbuZNsEJuG(KKiKEJm9mGr zApD6K>$j5YnIA5QX2oRF{+hXb*}}7A3!|?z4&TI@5&bR*VgGVB^LXrerJL=n16ogf zm^iJ+myH~LiFI)DEo9R^&=Hmuz#q&W4bE~3Z;q~2S>ab1yJ{bII3L?2zhSo4xN*;< z`u4wNZLSw@?{Dcr_3gZkpX|pz{XWlUZ+T@=GALQp*;P&E42z+$yr0ls?aZbdF2~VwSkF(zCF7O-;o@4N7XkGhgXg3;raWuL=QOG!w^sYWI`h7fo zP>F3@Cpgn+9Q!_0wrs9!_f9QG?=v>z%)YExDQC2Oz&hhT`#g9m{MR}7_t;x*p9gJ! z1WjD|m_ncH9XkwzcNTC)(4~g1HMT6fbjA2EbnR=s{2Qjcls)T~r#&I+>uftHqv5z) zzJs62@9DpR|0_QD4?28w<=YGMCmmVx!s~6n&#->`;e8d!U*X#=58>~&N`7I|I_>}6zI>)2ZI)5{WXioxw2 zA3iGoaCyi+U#%SfZFy#z|29#TSuCEres0mTf;NF%;g9-6Z1+93Ol)E;)uwe<#@4Yn zhc^9^#>?w>a#ql7>dX4aS@*Z9z$%H!Pu7C`wLF3D*%*wqK8gM<&xp0kA5=A=xJmny zi7(-v=~mWFMlzPHy>`7LcdGMUogHk?d%CcE8yJ=$U(d0xQsXGW`yBfy71nVp{VC=B z0Qm0wb!Jj^Fg6*tpYpYKbxy|ewa*iG0NC3(=Pwt%V%mRY*sN81JPWxakJ9QKG?f+R z&T*ADhyB#G&VFjyMbb(4G{OUo`LUtopQWAd)_zsj|0JJ6Gilmxi6j>x^MZ8_dz_U{ zXB2H>FO_Wc&8Ek4gej};C*o_FZ69-L~S2jh$Dil9L)^boy`?0V!Vw7`}x#jOK(on3MB z!jo%gzniQ2mnYj6nfs}SI(aus?p+x?g6x?#U7l&|PW3N8>f6l(wmfNr&ZjaFS zpf`|BS67CRn}T)tOQ7jY`p8exe~lU*iH|NS#V#8VKX~cD_$d0D&iiR!RltcX?+j|9e`eBfSm2K|$4`wopyzQ($(9jpcP)z?Pb z)BL&ibS1z^e&g10_F6F4uPa~1hntxvK`)0AoKI>!vTyd>f}=`*a0_~f_l++xpT~Oy z<7ACbLZeuB!1_0hQ=VnLU2U{qEQ%lL#jV|I*7IG!xi6|Gj9*Ck-g}+=*J7LG1nM>r zU&b1|$OQotH;g!&KjUKZCn~)e2hk^qFZaTiCBvxO*(WakYG1hIan@~d zQc~|Zw@6tj+%3eF9zS9vOMkF zEvt~S6yN;}eN$UT@krXrCe8rq;CljgP6uHVSd+LZ8jf{rL^gM1#5&+hE_~^&-1jN3 z16n_4eSIH&$*Kd>`^6H_*pEXd=}cYGt79G5d&(o*uLJx$3WLR+ZGl)P_56W-kJ8uc zIJ-xDl&&{?0%n~@A)Z%xc&~m=bqxb=LqDURY`@xD)Ty!}lochP6}>E$y{@M}O{4WT z(t35-o$M)9yWan9;X}V>th*>XV#W=uWs1p1)E__YMqr!c@cLO`tA^Jt$GS~FwDDMW zPe1f4bQQ4Ck6zMnK6J#^EN(fGvIld`{Cn#Dq!#GXzJoJK2g`m4#=EOSJQ+Nh@ow7V z1jDm|v5q0o58JBae9{kNY{8hQy^6gH=oF1L7|SHvNUwI5z^0VV#er&1B^W>1wwOJJ z>9NJA&-K&&PU*0xytb_WJYv^Ado)k>Y-)WEeb3mI`>n=avpS-*Mvvz(2;TNk6@VCI<$R>V?G#ubp0Nh*PQxLpc z;n} zcPa0`qRw{8iwN(x3S$N6ju)xR=m_MDGx-}O@3*^hgq{9CFZQ?gFXggU(&sf8cc~ZobEn zXZf>`$(QH3ynEG_$w$Qt;LEpp)FYEM`bH)%s zcSj!YQQ62vP;bp=JMws^zLAUL$V4-FzN~!b`{eOwwmcRhBl2UlAdh1OkL1U}X@<%H zr?)AG@E&=**{Qc(a_*7G$xgk$CHz~I715*erWYA`%;`lQU*W8Eqt}qfN}oKs`ywsp z?>K8t#^di>2Ve3nEB2!cXzN31ci4X4XP$a1zWc5CUbf$6D8EbV6q~Gk z&bvzapAI}>#XFu1#E(GdyXC`_e>r69Lbtx(=HIyR#^E_Pe|h)?H@`-IUTY1BMZdED zrlx75Z?@)+zIkTEI$X$DKg!&{)|x%4w-q=XXWTN=3Lbugx!o9iEyMRwHhjdwGR{m7 z*6trz#=U#N=>E@@aW;8yy55{YcR}QH!O5d~6;VSUqA-n=P9n47nV-{MUb@*;YlOLZZR^8Mq^A@cS@D{Ce;MJMR7(kjcl_PjI9P57hOl6&^oWDeV(L!_(4OE|S z-^nAK!^6W1XVaFS9bUAs6WNXM{Tjb5J#IU76^Q)g*zH!>ibZ7uhIrE_eUoupJ$7I{ z_F$C$B*NG)leHy&edqP>9cLVQjDGwW{?%jHfp%Y!?qfeW_}tX|!ixU!sBDnAti!l| zN^;l@Q#cRi!nzV`Sl2=N@xb-D$ztom=ERC=al0&gUZxej*uEyfoI@~v@cB^u?Qdtq z-+3i7{`;+2@%Q&-$3OUWPW(@8x$$IIzxZf$aqANi(lZXPFC%?9=_^P-f%FqeKZ*2{ zNk4`3Q{(>qonHUsD*S|b^z}!u4=gMEvm^MO^gglCigoX>@jNFONgiplV)emuy0)>- z;7Fkr8yd9tD2wFYIm?Yqtl>dbKFMlWe#UC2{9zuK-d zqt80+`m#Lr<)ryr`>rpC`f_^e+nH8h&Kc?(gh^ zWTwGm$~#MV>Q8~^NV&D=sj|o^vf7)>F!vq%QF#P;MNjf}Wu}F7f-b*~R9P{~Gji*f z885$14o=;%)m>xlNpR+$=A(|Ill5J>d;HI>cOO|zcyc@I0n4m$#z*)W&VnF5_OJMK z{xx)2>_7X50juu$c13c4{K=FVKtu|#;}fCb4Zy!A8YuaN z@F-*NY!p9=#krRatfY1R^K2=#|4!|XQ}p>K$};^X z<+Imz^vRWZ*3c+p zMU8E~iT_qUaK_?Ql1is$@kCs5^ z0>&gIzz~s~0OQL!)>B^IsZ7y00~-5zPy-LvGOn)&*ID3t9=zNHju~SjJz<|9j5BD4 zhH`DY&C0Zr!w2$E-dZ1y-L#v}lHOlGG8g+Wlz*AKo{o90_|idvc%#KI1auzHdl)IqzynYLoU($cU%25LrJa*;41&?1hXu&sb9K67l zGnZb`tO(_mQeGM5l~Y~?6v~@QdDAHGCd&I9<=sqqx13VmyO&2A zkKr%N4-PYX&+(J3T2wwZRvF6w#WBwFNDJRF&Ix}wJ-mFp6aG+o_>LQ$@CUu&n-`S_ zo$#-vhwsRA!v7&Xygb(l|7v>pjsZ^i1L@(|4L1Dur-z4zI^los3%|e#zt0yw+zJ0m zdU#FP3BNZzJT%e?Uzi?VGsX#DkRBeo(h0vOJ-p@`CwzW-c*v#y-Ra>qF8$|u!*?%2 z7Db0);dOgGeB8Z=y(1<(;tO}^JS^%DA7rPu)~Av z;rZ~A7xr73UF-2fayN4-ZfNBC=Z9L?yKzStk6qT?z2-x7#|-J*V975Ob`D_&@S(YN zaBsY+M{91YcxtVz=DEkx*NYhYpS2NtU0ZykGj?H*hdXxJN1WT9)A38Kujs+=Qt-R= zW8!x$^MSzhVW9^@#5F^QSeKxkLkgWKCk`$ zkBMJ7zChszPCfi1Y`yTi3jCZpI=^-*en&Yk%w}7KlN9QZ2;<&X3$J7}0^W@v(%4n56M%~s6 zj^6RlT;79YYD|1F z`F8i;Sm4RQy$HcEJ`AcB>a7dX4zp~_3gI^i= z*>!Y|v+MEVcU}*EsWJbz((tp#{JrC6kN0i-P95*t_*H=4qjRK>)E-@bMmo;D z`!j5ur-O6l$HaMha*V-wx(Da!$vlJeba1xo=p1L)48_jA0CqGi`=+6C+@DncAeh3m8Lhk{Pw8b{)YwA zeMk>bmd?|(+eONUG>d(BuD!U6cvnVj`LO**Ui$RqKQiAk z;N<g|N zwf%m*(a4pbMSamHSbB;N($FVVYQxC;W7CH+M4$f@4fvf#A17{$6KCi{n4u4ELm%F~ z(I?YOAJVw=dB#tlV5#YAPe-3)ebT3_5BiijFqZuV=u>9sv&_(-%ts$P?rtZ}rcZj^ z$3kVj(&tBia%~XW7-?Mk+2)hlZ|&=oY4n-gL!TX#m8O4!lg~__?7rwTy$||KcjzE^hCbgi zG??zAj~%z#iL>dGUiY!k^j_)H@kej_n%+a7MW@hb`kCo7xG(xt_CcRYhdz~m0s2%L z`rKh?Q0b$O9XHj9v+0vw_pwlAuk`t58ht8z=yUBU^r<{Eee!Mklsh(JF7IKi6UEp0 zUT<@+Uy+^IT@8&UC-t85F~>+ME{{zaJiw8Y{VR0VnRR!FHYAx|Qb6UdB%| z92&W}c;|!Yd1Iy{qFcP==bUe?zjYb z0v9j+KIf;OjcZ@@%Yc3v&@Th}WkA0S=$Fw$zwSQhXRYtEz5ah8e^;M|exLX==r`9( zKXCEV?*#T}T6?u|?Tdan(2uRPRu1&bfqpsAFDFI6^=3SCioUZp{IAk40Zx6j-`9!z zsO|THH2Q5wqu;kqp`XFEFZ$&{zdY!d2mSJ(Umot6P+SMaV!X~AEnz2REh5@!8|d_CSZBEqk? z^sON)_CwBnZFkntgtTTO6zZbR86nnY==VC-vT#OELl}K>oHZ|3vTxeECPn8s|BiCL zRch_o8eo4`$VzfvALpnAo7-6{BzTnnc!;$)k+V_7)E+KxYYzq23N^1k<*H6DNPd|$|Zr7q8RGJEls1K#ick8El4 zeup03vfuN4;@Fj(p_8#k>wJ#gKRBM8KPdjEs)6wjZXXbT|JMHTcPI6W|Gp$Q{?53Z z_}f=yvyV56{k)l+x01n`JDjnT9gLrZhQHUocjf{T$eq@5X}_G-pdA4})Y8wsto>{? z@QOVcZ?T>y(aqYfXvplZ>mu*xiO;tJsko*L6W31MEha9QinFpz+;51RY~nIfaZTAK z?ia+dUW+nwQgP8-6Zc=lea6J)rQ#y}P24Wxt~GJtRGc-?#BC$)3dM0pOAiiBgG}7_ zi5sc79B*8niTf^bmnbgJ6Ng?Aon9mk=Cl9Q6UTm5#cd)kPjL}X9DCyw_ix1M?199& z>?w*~WWu64XTd}B0P#nVC(-}d{kOo|o8j%};O$NDb{f2$3U8;t+sW{D61>$JIyy_I z0^XLx+p_qHKd|S9^(QAoGmmNwig&%c_A41)vkt?qx7yToq6&XijlKThg`+9mBN>0o z_wAku@U|1~eh*68`P}boeBYmUz85=hxBaa1ecR_sa~@|;e*1jRPUm~Thw1RS>92Ic z6QPS6+_tJZ>c76^h5vUSEAFL*=&^R;wimv?aLEh*|NH+GJ8;x}*6=m39gH>iBN#Zpwev{^o8U%i>Pxoa?$1`41{Af3YoR^v}(;10wIq zZcv!c1Z0R>$NFNylA+WYv0?=$x@QQUlH=Cl|Kyy467~yT&U$`q zy3KbC`NziBPjruaojn3=*7%=qw8GbATG4auxPgra@aY^1U2@GG1FUmL7QB9u8y=_Z zfGI1pYX|$@b?$+|)5?0gN%(+cwcuDc;lpO;T6G4faCc#?B~I{yPi!i%iXH`4W^?#~ z{y(=k?}gu88wf9;yp})O@T;A=e;csNv?uZ(sZ0AGcj1Sd8MO}QKx3WPQ30JEA6p&I z8g9iWa4zYKV{Mu#T;DUp)?t`1V-#zWS43wXWuJQTd55>#;Vt2fE1G7$N_gY*4v)9P zTf!e&k(l`);SW9U@R_h;zKt#9mhYY!^4Jw$j|&d%)!apWf+wda5O28R#<&a9l{b{d zudXPM&y2F4lQc^QERHY9dnCSLzzy;Bd7m{nD9@l9g7LhHP+DHMKAm%^^C)Iod515D zCgRhb%2RNe^nq4tA68}AG6$`9E z+@Uh4xs1D{M-4BH4;X84_DNa%vMb8tSB!lvUKR-+o=5q24iCf?Kes4o;_tp9WYXL- zw$@Guz6X)9Z``mruC&XCKVs6ZEc$v}X&<|yI)2aCjQD@ta6|mL;h&AaP;?`B**p@x z1`Uj?uC?K4UOv3q&`Nn9yW$b(c|5+!#M!VsH@t-Mt4$e7({ROSO}ZDy{?Mf5PE-TS zhymr)RRY{)CaunNQQFI{h*16w@!)_Ev<@2n^t&Rxk2tm?d(YRgv%7hps9|kBc4lIQ zm3B_w!RIsBla|Szv@G_dWyjzBNe=tca@m*Gk9}$V*_SpTZk3JydE#E|{tDjrV_&Wq zw4n8g!3)~g*pC9Qp4H7W{Gl zg$s@zxM;zLe;Bsl_@9O^IN5#i0?x#J>SXtq`2Hf{w-f(&r2AXae}R0rk^l3Q^EZ?? zgK}@B{J*B2zoNeB)O*W<#EP32B<|<@mWpA(h5cVP?B~g@n}By3a8CvPDc~^~d?ta{ zMDUxy-n$C+;FYr%uMGT4IWI60KY5IOf;BUb9%CHS9m+B5YTdS~cWZa|=CY;>V-dzX z8f(1Eo;%LE``HP{XDr#x39u&u8C81kctzv3Ku)wV$obTa$r>5nCXZBQoBb8uF^uxL zf8P8S=e~B4-ka7NA7DHd%!^Vkd;Uom;JnWg_WLiPEw=HkaK=B)XOsU|zWn?8$geW@ z5*Jw;-CO!dbZ;2_NNqNzs@43C;q3xBP$YBds5v7Vy=0^~uwwPwmGSed{;~08 z-Y@_C*x1V__pjU<BouA@AUoP9-+`THVB`15^9%C){AB&g2l`o! zUmeu<`H)_E=)60f`LFw@x)@(pvKPrqI~%{$d8O~)8f2^!V$74l*e5f7?EWmqIN1x{ zzA9(IJL7T}{Jx~$0>&+7|CG*v&J0>l^=Ciu{B5lNr|kD{EhB$9`76jjf&3H6KZ*R4 z$v=hsQx}-}J?Et|uI(Ai`{`}?$=)N$a0#?=Wni#R29(eJ^WyK#e~l-9H)~Y=bm%R= z>b;bF-n41uOIHA2h;te;IIn^8yp6nG?91bpt#?{}=>t!G$@)3Iyx#N!eQ~{(hkh~e zgrki%JitgkZ#&#IyLVmVFHdr5xAAD>tgL9`?0&XhJWPI_hvY5uyM(E2L$h-$kUBs^pZFErv^a30CWtZ*O>84VheF*QOPgl z^OfvHIE$u~vuK9V*7L1l^8-#FCg1RPX4$%>CoN~!^woY$eU$C5&(tORF1@bR)Ya8f z7y16#t;Y&nVh>LT=6%j<2oE%!@cAB4{rS~U5e6i?X`X%zW(~t za8K2J3huYLb-TFtQujxx-?|`jLiqdZ_wvqP-oFEbKi=I>jqS2(QQO4i@VSeQ$|s<_ z`a3m#d|JHb<8)rv7_%QPwKrM&sk9b8%y>MAd~RU8*~iSQ`&cmLcxo+^` z_1K6y*Vcqj&WXt%yumv9%CM2lO<@C7TS0TbSSx;Zlcu&Lru&CYJat#Rvv}X^ykK3` z+&M?Ne*~SYdkcfknf$ho_VVfWnObTqTAwE(-8w(o%NPII1}^?n3_sHM zQ%;8;5%EL%8=eFsk>XHAq*&*&{`0hxw#~c4mf?qe^*`p+-*dh~n=KbhJiO3(9d7+P z2gHxBNa~6rtyO9j~=W~us;`}`J z@N+)6jhj=p((2Eje)yDh7`$yt=iaqkPJeq9{m`}SHyjs9=5RirksGHC4LFoO|MKb| z(!Y;b>Fvc||H>SvO*(M;<5&6O|1~GQycc}m2Yufa{nFFFbZ+|h!7=II{`%XmIt_f6 zGoEEE`^wbxG&ZH}{}=Xf?KT|3e3J>z&BS|8T~?m!Dp}SL%_sKlI3(2Y#o${DF1KiDBvI z&%5w@&z~}5^}*aJ!8w_qKVawapIh&zzv%nK*ZZcg+y37${c{SoLUfMwj_#kDrF9GF zpCX=m&RwPb&5$iw$TO4gS$r2_bII;9@#qqrx7t$3+1uEi8D^|w-x2uMINBFy>$btS zS=egr?B|a1UP|8AiL(?(SY?wHGii7SndAR4aY6d^)c7vd=3SlhIe4M|M>^+Chu$uJ z_I=VG{G?BL1b;yAYs}9h_?G~`?6@KxeFwbwICOE(#(vs`S@?jbft~kk%8H!@%wZ4A zy03NwW4sU#aLwO_?IRcqc$ji0 z%!=&Yu5%;lvkG{a%b4%Sp$C3rKY8GH=q&j&{Gbk3{=EEhDTCA%)8;lUd`GF0iv z5bxQ!(Y>EM3x532ga7{le;2+nKKd^Gr|8}bPeaEBPdRSe7yaDw9yR@K3SYOqY|o1B z4N-0-cL|7RvSp1-B^+FM2M0#?et2EqJgoE7-R`sl)vZ4-|IS9f3b0*+Jjh=U@9dK7 z|K9RS#s9p>LxnuZgDt;%U)l1(r~B3k%g<1rfBZrJUrhgBT!S93 zLXTHD`Y03LSYg@h$YlmmYrk*p9p{ISp;k z^_1D0ygllpuh*U{@af0(KKz6)dOV39H~1TTX`^bVUpxzZyEsGJwZr`9sk-dtCdvgi6dF@TvZ=tKBjYT}@_xVeN-&nqR^c~>al`p?NIvRY8&FR>q zrmqF>+1R56XMyWx53V{dAZ$fW1bD3QtV>#e3m@DG?6(u)aOEYl^qs>ukG@-g%NoEp zkJ8W5cOKt7CO>@OcBK;qJSJXzX~MqZ;m(qjef1iBtRBLsf3Md@bm?v1|D3Wnyz&s> ze0pzxxr8&pz4Q+7?WOk;?vxK*L4V4l{xqQPtNG^9_Y&6B=$_vU9uv=ZC_+4siKlPu ziyyCzP1!#3^|}0Lhd)akzWl(GrpG>N(j6A0;jQt)7ScpQ&RAkB;|#^|o()X@?#X|y z>Jaa5aK<9MmjdSoPa3T=HE@K3_LzjX>i8GZoT=@P_Q=OKk&gg(`LIUdzC^Zd%hMZEsr$7in8@8I!aD|M%p)@r)>+&;`(mcsn; z{%==4robbwy`r&O>!nt58}}TGS3&M6+m7$6MtcH^u87Tq&y}pR8VFpIenET3mXQu# zN_-jdCx1v7_lDi%#JTz8cM-jnzf|&0KhnbZeJ|;~eSItCMVz>1;#$a~I(G4U9A$db z(j&2!`jS`y=V=y1@?zunaTkC$%*u;hM_4}V{@t;T8Ot-?Y3fSpDtw&D9|E6COTArQ z%oZ=AEl1VYz=vG{?iIt%IE4pFU&TFMJ4k;F`z3h%*skhOpsNkL#`vYltM6I#A@Fh6 zGnAn?`K84h__xc;qXKjyHcOg4?1is@u_E)fr-`Q@qp!lA?ioL(=RZ^Vr_b+~H$VJd zTvV_B`%L!Onc`jfweO4c=r^bTmYfCXYeMv0MLb6TsDDA1^o;NMF2qjG;4$&)PZE*h zBAzNM!`MG<*snc03wqL5;S;re8f&ig_%yKNjZcI3QtDepTu=W(pJMa~@7dJ#C~;@f z9~XG!^L1p?#W(jdwbu)J=F$_BXs@?e$+jHpCGHeD9Hl*%fcrXd@6XeAs~z7m&C>m? zNy+IQPfnzN&+K}OcBpl$3+P{ptWfg=D^lF?YpZx-Mx?m&zpY~3rKG!H+_i!)(FW4m z5kBJgL<82Qn06Vm=iWzB)kf79QK8&*v?YxJy=nY;%BZIYrYfuWB6PF%?s)U~%V=kxOg;GD&E4T8_!;I6wq_i> zihHZ(&bj_=?(vuz$eq6qStub*;UH_qqq@h5JGJMM9v@(GHfv8uSixrQPHP@{V%cMx zEUQlUUbW%_?*xXVXaWq0+@r5_pyM)GpZY9iBqpINr$yiw@^CA6F`?Hw_lrJ};)D2# zO+8j_*CG1d_VHG-ANO{1Fib;rpx^ug;CajnaYlZ>t{VE?dfo@&^&!r$?FfaMmmRm^ z&>WZOr+caYa1DK7CjJCpz3t;qt^1IC)R|W7vJ(|gPCdtp%?#vpy~R3O;r$qI-EpqF zvO7bW&7)2%e{35#hjor|(~wxFsWX$i(?bny(61es-_be7*3s|z`K3GD3X~Umr16Mw z@3%jn|J!bZ{ThD!zI%^kzlMM3xAs7-ciy_!{Chp|nGG9Fct*p&@IFAgZg~GT_r#Z2 z*yIS1&Rk`=sO7YH!E@a|NazWMCdLaG*y9vU+6TZt9f`gph%5 z-g6xp;O?R3cJ8zj94Q&dXjq60G-dQ61D<*hKzmbnFEa2Zyf3h_n+wc(?SR2G3jccA z9P7t)57@iZWA?=`N81;VPvPN5=6N>0zJ3j*Jv{B#aHZjCZbOmtmM^flCqAQLlnKvn z7|#0?I{>)l`~1=0 z9$6K?kF!onWxUP1FFb2{!PD*`Yw$n`BIi$Z`eGz?=U~5GO3eE$}(QrRz z<9;jCjQwT;PaEy1lQ|pKz8Jrf_7%iGgQ)Dy{F9G=atV7@Oq%D*lVu_LcHJ$u zI27B)nvMwb;?8}CtE@;tO!LT5@>{?qy^x?SOI8zV-{9xgRZ2Nw^15|}83S0jRxOJ% z&`qWcV9?&&L=9y?Crfv(Cq`+HFMAK|@^yj;LmtBl0fWkrKMDVhfiX<~?ZW#ZHjJB2{L%T$?zhoJHFn#$ z<@b&s()m^&A-TB2k&9~27%P%Z+k=PhxWV&n#>74P7JccR+iZajVaA8;?;87BxbA+~ z%afmR-=f148#a5sP;it3$J-7JZuxPe&r|D@qR_>bE%p~k#{i?gO zRivE*K1naAoCup-Cx~=ok zu1WLg$lw;AJox>mYKMW*(Ztz748;05;gWO3 zOYbUPzK&26{xsf_myN7NVqWT@4rG3rWqn$GwZb*8d`C{Cu?qXeUSF4iP7WAEdUWY-t6@E|8xA|;f*M;!e~g^;%nvzXE)AzSH0bj<=_DVsk1u_!M>bs8>DT{}bxU4ZxX{7lzZ}_= z7`XJa=ND6ZQ+nt(*}88u^_5gu2DSix2Jydy@D{@L&D}CXOJ+Fn%m<49C0|IT7ynDX znEJN)kZ<_VL-UZW>&HUB*1^?V&|lc30cect#sV5&B5&7O+5a#aKJmv{`-2L&UZMn2``kObLvb%brpI`oh z$X^-qC;HEXb`kh0T-G6bmQVIvT<|y2)&|D%eK`HQQ?FP4PFvrck5FG^U@YXT?}=XO z!-tTj|6PA>fIDra%QRk5SRwqEuceZ{KY%YuKYRm%RO#3x5H2 zjo>%10k>dFF!mIDv)Ct6X?#VIEz(^&gG1psbp3FrV6+CszTF3m#XcCd9wdNIuLL-~DZ2%6VMN5-CU^n791*7Jpyt>+6JJzt0(m!2=IwDr8grRVil zJ6F7DRp`<4ZPN2O>3Uw{Jy*})fu7fRs~vl^0~=D~HTeeXC`WC^$Vj<;SEKZ*`U&+t z>hrvD{0G_*EKOOCrgU(5jko^V|Kc858(j&+jsPFm1?jJ3hN| zRGd8#7GEBS7{)?Pb-jIky?zhMYCseek`9!Z{er9E$@aLtLd zmQ!yAdn2j;NtJmU{Jqr)1GjGow%csqKr3Sp`k3ofk8HFJj6=S`Iqpj=*587ASUQ{f z6Vs3rjU#48lFQwA%2#}r;A6Zu6C9VQ{R*yWk>tCdwEfRogGwE-Hp-i0=ojpI{75%v z6wtR5rucQpxBL5>{5})E)YtF&z4~E|3o{#D=l4PMMi%2+@$VpGY4Ng+F{Q?T9gHmw zIQ*Q8eC@Y#x-NroCB@H+eEeMM;b$xSyvXGjv=+bC3KqtP^2H}+(nk}&4Su&mvr^`BRM+7>-D_3} zM#l1Ke4P|YKJ4-ppOT-i@4WBfEAfnPZNBpR>9gW%u;H%_U*%`ZX}E>oULCG6)O+oA zzb*Z})#D>`?GwNLMrU?KvH4`XR$+G)@@}*8#<@0Z5qhr{y;p5z%wOm9z23e?DaqJ1jG9aea}3?@bSU%+vME>ACPdYV5(3{WM#?QT2()-u>{R1>L6cdjgu>%Xh0I zbCQ!bV99jl1m2tvFG9?_7s{7}o^f%~+HV&x=9SX%a_lEZkJ$Z_6DC?|out9v33G94 z#iuO%R(tSkt%?-;?Wad4sxL~-51BGYLVu@!8OeK-WAC6#=PxPZ43RwR@0$vlQ;1mi z~zsZ zc6zmArz>9lceP`u>sxrMzqR#os9^&1Qdmzv>(IET-8kPN^ly>r6PVKn7NwU>z#g(> z0nfFRW6PpFuU8LDyU@L|QPu-%ki4zvt^(fcdHc)v*Du<+IJ+>5?(%I4_VcMPlFix; z^xPbHD7{b*T|_U*oAOqXUi4D9=%sJbOR$-EhhFs#y%aBc)jRakw_A7nG;9Zttf(Hz zgUYFO>e=elqi@yY;!x|ybE{8fm3g{ zjv{`0>(IBqj-Bq_Ylp}6bD$Gr0+-jcOS5iY<3Wv~{y38U2>qA9u4{!JU-wCacr990+L>`=VbmwX6->UklQ#?7PKcde2sZ;&Y#t^!a zvRkRUOzVuPyNb7bZ>7kX!HfE;z^fg6mh!H`{w(9Ewl4nrHpVzJdAIV$XZhyrBbq}< zpNIeW-_Dr8k<+oPm%5gxoqn|jd$niI8ay)SKP0zPk>gD4hdYqJOx9~l_B%M|F+Uvo z`6lKhN6;TV%eZJ{FKI?m#sTKmR^Y!gcEvYQWt zGMabZ@W{TM$5*hvAgfOJ2Q@CF{7t|oxa+UC{(cF%N;*b3X2j9nGN5D z9+gK=rN^ml+TN3~RclzVBQ+=6ZhqV0jpny(%#4QL_Jn6Oe7EO!M#E-)3rD~G^Vh8@ z`?D4MXZs1;#&hkvTw~v*<}VX!m$dEf?qFOtP+MU1xDlc2ubUl+(C%pGwDrTUP$s&- zv|rky)|=(LMi_7T5oE86-%G9QuNz^}R{56f89QF~B66VpR;I5kcjgI)=zL1!57}M7 zUC~dy;lpRJ#rWuwhD*Ho42~pqx0J=7i;ulN@4_D6LjOplm^s6SdorxCdLJvY-d#{q zk<@$P_=;qGU|`pq$i)1M7w>y3(7&#E-u$DFXKMbLzKJ>4DYT*HKqUDVVd#pa_WVk| zx0(IEqhlkOJK4nC)~1|n)^&V+%K_>*7#h$JCv42;!m+ortG8^LU~uJ`3pouc=39zKg6ziP4h+vKs`)98Wz^7;o8q3SIUm7ZMl8fydB;j_DW zz5|=%6I?e4KD>H)tOXbhf1GgDwUxU3^Fg+M^`ahnH{o~6&sg-H+L@i-#+miaYeod3 zdtLwM$idKTPDX4w>qCXx`a#SeP9Gfmu!=FocWqxNGG=7r9Ats}*6efq?aC>2c&OnIEx=MYGDG%n{-P>fY78=E}DjqYytLWF`No zF|a3 zPIKk-gPTvT*@m8woYLNR7mU3s7C=7bpV3@Sz$d4F0}s-4xug5`Cp@|w-B<5v^DaN$ zF#0^TzSCW2Q-(iV{17i@!3*tIzY0G=1UhJrt{mExgQMtF$|IZ%E{@MYX(Hkoao+HS zuKnz!b8&tZoM$2T!np-qSTHnbaQ3&!i~7LzSs$)_?WY>!^MAYeO1JG8k-pEzg)QcX zE!sP5@BgHCK0M-s?KhtGFa?{#>rZff0rIo9-$vgz9$zpztbH7ENIz@l7wB{4Q*DEm z@}tCgORu@}d@b0cqb%lduszi-H=?7fCVw{8&RG*3*s(3+7?;A!rGe|mB}nV6jo&;Q zeYnB0#(r?HdrblQE(rfdpzkDm>fdB{YrawCU$8xW?Al{TNcNBwtxn(((?hIZ3b26F7kIr?Qr%dNC^@D6Ez zy7JSOm(mt(8MepiN|T`6tI(eb_|j4i45Yml`LQ_LL4MPgX)4F=XR4rU6?So3xvguy z34Qgu5E^UmZ4opsMDESH0o6(VR^rRH1LAA2(R&n_Jy>XyYLGhjFigkvC*a`OoLO_1;Kw2mMf0)48lwSiGf!@n0u= zKR}y2@J=u`mo|C`TEFegNw&<7Bp;g1cuR1K2TmV+b#HCKBNv_0tH{Rp_5(vUIQIwV z!Qeb3B^z41_pnDkbly#_u}@O^%A6&j^E3+K<)v1#iFRzkzY2Kn)jKP1Os(zk&X@hO zuYT>PGiMF7Ip@82&h^lF8Q5V<)cO{|lLtH&dNKeU1)g~=15df_Q=Q|1M|60= zfkXC^slUpuf19WN-sW31hVa^Js^2|NMRZsSPffiwcD)mQ^?u%|x92>Sjdr~O&lw9k z?;{=GGt>Vix85Cgy`w$#dg*`NpGkl9cLC%*Oxx934nu3tw+rtSqwiDe6GWfsoF5di z&kstLlSh6|l_#HJ1UpoA1A6eKNH~3Mge%+3!S2@l^w)NFuX*pYDY+6KgAP6kmyZrT zmw3(%YVpwXO#FM!y=%|m-!FXib^g|FKWEFok3wDCvPYdaff;{qN6#z?!@i6-fi}6>?Ur*Z* z%?fxnk2>=nn-T{e>GnSN=Q3XQu18ou1lun zpX!%`?Dvt6HDZp0+U9TovEC63*_eo}5{@LT{Q~fbov#qrR3GDdoJhP!; zg?MYmDV%?HEpb=bX_U5I{V{#0_E^=Db|f%va_B8zp2jxo+1n+*T^M?&t&vv0f0gV= z_YRwYnKzutd|M?lBYuf*+BdTnAF3_Gb{ZFNvwj7B3;*wU@ON#3*U*`2r|zGN|CkFU zKPi8LvB$u>y)anZRxbO_GJI$+#9vzyNy_eO!A_9f{~GI@T^NO<_M-~M`+!mQw3}Xe zwPsWPstMSY>f1A+WoMDKd1uqLv7fMpIVUgDcnBUnK^aOLB~1o>zq^h~be7g(`-9=iMJ-2?m8)d#Y>BCPAy+SUQA1=o7JeAaYIAC1iNz}b`zXWYtpY|QT# z?;Ar|Z_k}~^pKTXH#3mcl|YuXj-?-EwcyXa{4{uIyrO*}>muxnqkXPtjbI6O@$Y2g zBj*XmRGChj^72_@2-efD3A1+lR2rn?jGqV{aD8kZ{;|Gbwm#_Yfq6ant1Z~}m`ztx zMzlEg18QMG6AH21HMQcLdA$@OhG@Cjm!c!x! zr2Vq$%tr6@r!AmU&Dd#&v^e~H?h z_9{xZ*_MC`_xA( zyb1Z31mBf+jKY}nQ=gSZ8`=-=JDG1$8{a>daQsSlSl3^tIJJXL{OE3)QCZ9@ftTQ}!rt9*j#b7s^3n7J;^|Dl7V z3$Q`2OWYQr4s^{-bW>G?^5ICTCLX)3bg^)G!P&-O1d?hH7-rHEu3ALx%g`EDWKwp6~&1^;tp znvcfK^BsTv)_~ZHgo%%SS(Lv2bT4vPi43~BUT-6hj(=dvC6dR~dDpIgKz;%F2o6@` z4_Jh+0RD8`icHRjmv_OtTdZUbyl%Yai5&V2$|%USUgFH6M%ielkBC~a z49ZzX852zz>|v5#AgyG&iFA?w%iOuZM^&Er|4e3b0|YEou-YaeAmSB^0x51YNf5lW z)mFN8*KP>}G&J39+3vq;Q4>r6CADS1N+s=bM;UJvp-OER5XDPci?FTj+Fde}8yIgv zs~JJ_|Nh={PUg(X1d&x-KA$ji=A84s&-1*`?S0Q2rx*19V(jV;&lye&V$=oIEu zZOBI(@V6l=iW{5(PERA#2aw;~z)hRm(E~5T-`(he(Sh4nXgm}z)%u#~D4)rDZSQyO z$9iIRbR2P=Cf5$_a%I6;=j^$5^s93{#mZc~JCs$wDBp@4VEwsG>(qf@NB%qA)0E%E zUbn5|vqIU0Y8zy1#URHM2tt%+^aGq1sqM7w%ts5MMF1MMw7Qv2I^3)_7`n|!_ zwxAv)w7t!#ZK3;X$VHIsDPD0KZDi2TO87BA%%cg~q01vNXe!&J^&);_KM->f#z8So z;vU?ybJUxlC-=7f56|hDe4f!fPw^}10pqi{Idfjy!a0Q(5es&8yT)thf7#>pxQ9kA zeKz{m8`9Ee2=p29@zICzo+dfGnsq{hmpcdZ;nlq}6|Z^+FYkV^|4aDnXY73VY;yd9 zB?XLAqYs~JJoxMgrtXLGU%TUX*?>Rs>`RoGJNXJBU>J| ziqJ>M;h2bg**J5EEt|TPTa%z>(uAf*!$muCt?}Y1WMq%(zOepgWW?;@M(53AKWfPA zM-5g<7L>y$S#j^5X7-}mIgue}FRGrKjo!@6WbY+yXw9zzIBS8^tOYu7y6bq&k9gr^ z?-g(^cHt!Vxakq#6kN2W8l^5ydaj1&V%sh7rXS>8{RKH2Dz=>*31r{JZ5FsaiR^}` z)uCrHt(M2ulSiVpfS%vLMZaY_^@&>1S6Y{hJ+AfY%qTo~SL|`xMHkM;rxZ?Fuf+e{ zGYeg&+?G}Fyu02hTM?pt;i>tE=w|PKmTpd6lXPP({n z&mE_Uf0FqYP524&%QSD0FReL*_$U)TlFz*XT&`n$*z4)|8JvmK!#i2{Ss9U0&>EQ@ zuh^8$)hpK=7*a@RWC?wVo6nqm3KDTkqL+#93qF`7H4cc`tfb zYfK^f&>9mmm@sS44viK;qkrh5?LELZ1HZEgzN!(g(|>)?x{Eo)Cd(>}d}-?Y%hj(x z-W6xf-y83W!)qD&c3y?}`#kUpYtI@scoJ=ByaK?bwiLIUsn{9gC;t3{+Ty!nXPRqe zutv7SjD2C`9Nuk0&uDxViyBV5FDTn3@4e^*TxOpw z{VNCA%UeazO*!mwELX@Xi9O8Cee+ zS?~GH$a=7Tfg|ge1gr~9A5MRgPs#iP;y?9AUvc`h^}Zcb@P1E`Q|zkNWo4uP{O448 zZN}beP4*N=Ze4lJHTfE;@lUO@Y2G1zoejTg&rpr>+*tQgT+OUeGdAux@0`LoLu3Cq zi$68aU)LBoyvn>)dNMJX-)rF8`SOS1A<;9%wi%se=XmtG2QFspoS5A&H-alXtZ`5b z&kuV(bM;y8cW*iF{=m(|VvH_z=OgyIR|@^1`wx9|-{7OSUBBUfX8KunK#Y&c5EqbFzO%|0465@QUUR zuI$SOEC2CZ$i841$r)_MMt)SbSF(>BJ_oMGro&gVLDH|EcH=c=ZtT=vulOJv`sRzN zGO0P3Y^VOY^7w3;{FMx3MSM*S#AgD`!6la!$Yn9}jT(46CVv{FUd;fOD1N zuSSl2^5n?9A2!7|1>YJE4b!dX+>zJY|8V^d6T8G0arM@+ORzEcAg-N@;dg2N8ZXO; zxX*m+YCQw554mh(aIm#ec6uH*c^3Xl3$|eaxZ-neSr5H6PV#|W9yPu%_UUgAVr$`Z zi)S>Se~|fo48PJ+Jc!>E6A)iX588P5!B@NN-@W!G#cvjDNAsJxe>9(G0y1QL9&FB0 zeVzt<9*-=#cs>rEC7F)T15fyUo~ev`x%LIZBYpWiUqJ5tK98Q4Eq|J|cYT+?Ai5R8 zv(Vi52%6`Lwx6Sq8pcES?Yvrd-eC4ubhZ*B{yJ@?^S9~+zkH&Gq)#-3z0s~ulnuYQ zyfQ<647x`4?H7WJ?}AJ4XdHWdB4~N=-lR|T_GD}&a4HYFfN|IJ9^dG~#k5sGJMxDD z9)IWn@5%OR?Bx&rCvEoS4;=&#mxew55c)$j`~mH29^v~S2HU#Ul^z}hB zb$RkS<~zI5lUbVYz;D@B+hgs-(KO%b`3)b){O0&Tqw#@M7qD6Jp_S6R+0pO7i_Q2e zE%*(sE5k*`uSMT$e@Kdqzys0^J!>D1|MN4|Pt8W&`iHqE`L+F@x1AW0^sSqZ_N=Md zm5etF?;!G>P6jlm5}jfdR$`sU#%i7yLZEYES^54G+zra^}-0^pL z^$_deUS2i)BtMEiiWz6U*pj zN`^J2vY{HIneSzfe;3@Q1a&`?`Ew|`L$UHB*%NHub7IpD@5wG5`D2pbi1#M*sZ9Q7 z(!W;ho%nDh0R1wc=}dH0DX^diOq@@9lLnA8Az63F@}Sm*p^fG~8XqU$@WN`w!__$> z6qhlvSNrU<-)-=XCGVTDHd`amUhYZ@u5;-wMAc9sIoc@SQ=Yegpe{-XD}a zKUcg`$Xc~zNd933I950~O3yT1Pn+xc?3Rq`>BkL4`90T4#3s#W`vGG5KHOY00JhMqQVYVh`RG zJ26r{qZmaseAu&>+Krz&aFjT3q&pvvH8pr4nUCk#p-rkg7eGJS^S1EjN$5vvJezuQ zs^cvE=-#vSqg`iCHM^xB^-L|g#n7eJo_kTF&ffp+T)Vt_ssmp-|8UGHf-iOcg?`#* z_if9+bb+b0i=Pu;jy-}07~gFhteU;72`z4$K#jzZH5XrIn%1&_gM5lW^!n}gc*d6# ztK*EW7W@cPqjo~gHj}>$Z;fVtcD;4$(n>45S!-n3f!Y-ZkQ3IX=VnvS5q&VejK6m4 z*3BV&#-{^06WXa`xI_DJ?mCaWsp`$SzqRK);+?d8olk#Vw=P9jY^sFr&(NOy!fYp= zdH`C=9(#SkY=_ReSBd=*O}4{lQwCTU%D-FIJw1aDwgKK*lT(cZpb#E%q zE2iASSm!Z*vV-_t5%nb+h`#PP&ZF-p{F9b(*0z-)>tAKRZv%!5_857`XEn9foA6nL zPs_NPZ9C|rhUZv=G(58ddnlb34%C=2l5EW9nHg5oW9SxBvwjBlQr~Ad_R{cunKLhN zWRuULIUb@19mtu^E z_~?BLwk=7WWK#kHnDr3P5;;s#@Is(k;J9PWz2oO`jGPgZZBo-KwXqx zv~}mW2i`J=y9lGe@z9hIDBRH3qT{CPrK8TXP!>} zHu~~yWxF&M^YM+{?>phW05p!_6Uc@H>^W%9TpBK+K;eFUODl8_s2*t@X%XxGo2Ug%F7(Hew?f?qqUQt0dgV(#KH!NJvzKQy%sxG{eie9k zG9DK3ZZifBtQqB2q>R4g*NuUPL-@@LP7lPx{#I_EcOa8e&(YW~Vk@b=^Kn`G#2XKWM)K;QSWSI_$7 zxrX>^6L|a%+a#VD;f#@d6zn|pD5(i?x@GE(3og^IBeU`WsjZT#*M~44CwDJ)3bp>M zL6JCh9Yt@~SN#_UzI5@{p8U)qvLVU*%RBko#Oa2 zZcc17ZB8b>QCD+TWF7vc{ce`##G7Wrt53nJ|3S`!=7#1TZAfS7d-<{~eAWPYXv8`u zAXC7k-)Aq+N~|GY%eI-}ZEI=cLE4bc+eKZELf+qb7Pe8gw|HrSOrSDS=eGi`U%93Pv5iopM`{(z!noc2ZM9%E*w)AfQC$xXakDuA!pY)#TY1a*%dgnbcujxMF@~VOI%zIK_L?wQJzLfe!CU=Zbe%qiDgHhBwf5Gsmz4h1pQ}SnT$Qny?C^sdPu=3trC$Dc z*7&E%`L*aa%b^!}DpjkXZyoenaC7-S&Tu$OeNmq!D!t{#j+G}z?DQFY(hkK%9pBwt z&#+^r&&YSzy)1lp#j}`ayg)7O#hl}D)=F&L2;y$uF{r7m=voa;UHa!y$BbN>X{H7V z_P3?bGILqkV8x!cd=B}5*0QYA0+(^KmGHe|7>h< zYcTVybu}SF_u23jx|4Iaf>Cm=44viw`{_>X*3ez|E^+8S61uYoxrgp?5AWB&`?b&^ z0RAoTJ+%?2Umh~&AInFR&tUw2`W5ciK=W0=p}YspO=Q;<$1>NVF=OGr=gN%ylMLFw z7+T9ekq@Z%eze;5>-0_!yA7JhE*UU3A?*9HNb zY+D)l$}Ww-U$SE+#6a-}!{9y>|F|4mBOFu@|8nRcyA#Lem^dZ6Y##GO#b|ahkIQGj zlN(!>ALqjQCBceJh_4H}s`p7wGFI8`%J0!T zPGhS1PYeEJokNdWbR2W-EzOLF==xjw&mgvH`t9)zeK`W|?~J4P?3F**axHt8qECJE zAyz!AH>KeT8AH8k$ zI@(cfc+p1x-1(w^e9XKmY+@iwyz5bued|$^@pH8vrFe(z{KD~;-GckA@VfN-pO?nbr-CQPiB8W&EESSWxU~IGoDx3@G5q- z4*eG*MmO8Z6P0~8*TAbB(b*2X>xm&*@PWof`t1OIwxRznPhOG}r~0}BdyF`hzh1bV z&z_9e2_D(vbnzo^KCU-@}O+zEe-boI9;DcWL^{g-7-;9?J zj}S1raMG01n_CVar9ssny7G2=7OWen$S9w~lcuHVWzL>g1ADM;rA$GT%| z*TMW3a8Q2n3@0|A=hqV3gU34R*>|b2J0Aag31gH_es@1_>!sg#XycVXyB=05{}kfe z2@Z{Oo#E%^F3Fn4h;`s+oONL1r`F;dmSuKLfbTDqfBetT1zf5u~LzDFh}20 zP%|i!*h(Ii@3*)7zl8@yUgtO2S=XkBmwxEbOZMbZ56xVC<=1z_w3jL$$iU;X3D-Mq zb<&Re{$<7^a#5-9OAZs=M!!=~JH;^`c-*$x#$H&~ zRxfY<6~G2X#pkB1gs{@G~e3(0y2r)i)6+9?Tr&bW7q@x?=o zaW#Gs^^HxxHwJ3v46zL-9b_3)3EP0Mj1v1=&Z_9)40a z{!%G4Q+tc0TWC+~(c)dj6Qn<)PFtMQk{E?8;EWz>@(>TReKdGHylvfk^vin<&a(&I ze(BHjOMFA0GaWd}@6ED*zu&|zh}VR-<-_C4cwauOff*X$o7vxaFCX8+w6WHHci!3y zBh}T`M*UVzY{QI6nXUUpcby+`uZQLZz^53h{4}lW_O<_Gj?bRS-xr!^ zYh6M9c)m(_@VxegZmYI3s!DXtxj}{e>3S#hWsm>1LjKHsT@Plhj6ZW<{SB?zR={5| z&%5LA?eBy5<9~bK9j`gYpHA8fFMkyPcVGM!V!X=bPe3EhOU)c0ti4FsLOwI78(I75 z)cS5-y*$)2CjZTtFdoo1Z0NgM^y2J;LjHQ_1#JuX6TLM4-f`C$d1)toCfO?FkN55I z^}L^s-lB&a-@WY8)cCGqyYtzn8oL%ho$pQBlTkA%(*6bVfUbdGt#HxwCj)bL!uzfG zewiVwsO{Qt;%V#~b!i)O-m6}g69_Lo)5@vePd$bJHe-j$pE@=2C)!YacD3pb-uQRO z-z}S)X|sj*h10uVA?9`k^E7m$^q}UTtI>m%e{|+%!V|p|Mt|ygJFeZ&u60st$0r4o z;89K91IS6*`HAyvJLi#RQyU{U5H2eI3+F(A(?Q+;FCi=)^6~b zfuG~zv*rWv+1&>|7k&tQGILnNgy(eT?e&a{fpvjp=%1TIJ%vDcZ!XUq1m2f}xs8_t z@8t&Geg@tW54=D40C-pQ0dLNSfLAnJkK7+niI(>fUv*CR! zWAG-=UctMs(@rxra9gN_`zms0?muPCtt97i<%8DT8lGFZm0JJoJJfXv*Zoeh#=mk6 zXO2#=(3e){?}=GmVFeqjZV5G12LheHV~<&N!0J4AxO~Lyj+y6L(@vUGF>!lqU~bjP zD_1n0)Ai^{=E&CpvVZCr(#!1u$33T@V z{K^$onbr$ufBW37XEUwDb0t=h=rj)ci*7MwI~QF0f$Qa}AtIX+=+{vpy7JuTc&-Wh z>)hWs{OiT->NL2u17r6C)?CFft9&@>_rCB{y}%9+KK;PwGU~;Q0iQA8Q$1%)*PK9Z zJ@q^iqrgG@Bltv*m(j0f@QU&j#CsL+$;zG9+zj~Y4|?v@z}z#jc`IA3xwnhg2BXtB zLu1xwyB-}=@Z^fBoz@Fg;yWwQIp%_uE9fVxcE752G6Qp;fEK@{oqt{sc;P{w+x}_i zOa$5~fbXgjKU`Y%-*y{wPPX4I)w}Rh?9jrjU5$Z-P2U)E{Oqh@TR1smJ4HV|pS-p0@qiv4Syd^$M|()|Hq zugp)1&{GpM7ZJ?(`kI@*K+eig{I4zg-Th55rsVpbO=~?{%2}I}b21ud4G1+Vzn%H( z;)S%sI%P)ySZ|}wpY*Qg!1%UBC05|Md}>T6$4u`(aiseL-(Pc>IKgPGZ2-e)+jfm$ z9w(T~h(+uo|0a%IEjRm!vm#&6IN)E{{DHm04qnn=$ClmqwT2;|w$$V-@C-5ZBjoNd zKbz2of8Wm795_3i(3lLSUyY}pxrRNmt28IT*Njakb`3luh~K*Hs_x6d8_T=K1aj&R z-k9G-OewP8vdp+O=Ub5{!P)z6EA^BAp7-D2J(p*$Lk5JKJC|_RLoE+~Du?MCyu&o0 z=bt;@8Drvy$$Y&yd|}r5;T21D=&iFOofDuDJfwaOZ@0|6{&2Vbsr4MJ(m+$yT+$ben2VrSyPRcW!3DhA*W9B-kx)oYm<3U z%B6E@vyHkN?p$58`7N|D@>`Z=&U?HJAM&lToELuaZt=#I8CFF5JL1Cv)FuoNHwdsd zAi%j1fke}wKw^iH=X@ug_y0B@+9tgbNW67LYOeoG#!2mTH}~r@bSw_qbgZ^NyKnH( zk^5JGck{549y*4g<80`-5IXLLj!#w2o*3h~QI#VN9V`5F+#W1y0uDVRI!&qo9EH)mW_b3RtjSS%tI@dqpM!z;U&cf6EQ)U=Ox zU*z|f_)AKY`%S=S-~`@~6{wFv zORYc+zcyJ-+FI25vMQ3W8A(>O^{{a(G9)m)9vRM>Svk#ZfiHNK#Mqi zsJ{?xxqIvd-*f6meQ?&bzkS6S)NV65sBYT})i%#(63@;WwP0^nW#Nrekev<8|3*O% z!rsYGU&v%2Nr{w?B;De3UDcKH;j8Ai*>sPX+8q+xZ<9+@d z!NW6pUbx-MTr(To4g`aZLB>Vbe2*U9pJC*@`}+gBnu%ph;9e#55>~A4_!27G2Zj?YPxZI^D%qe#6M=-UMelBjb%?@fBDe&>i>ug}=n1vu=5&;qv#&_Fa+NuRbTU zf8%zZ-$i}#&(mHCZ~nx_)ytcEfW_eWlO&Gf&qCp6(5YR++d%G@7q-=_|z?(5!C?%D62Z|==!j_G~(yFK^MvhU-=dhdU&=l^HlImX-zciuadAEdki}(IHJ#FXP_lYfWpR?dX&*j0R zv&Pr#4dw4RRC;r$;k)=jHjIVl*%&d6CyDV3j+1G}eb#;^KcYB@=1BHD#9rH3czSs0 z6f5gF`LcC4U%Kyl+Rvu_CvMJe$RnOX4cTe(=iTukKD(qKJT~&=L*(r8j`H3XqeB+q z7ppd_)=&NYs&?N&`nk-?eNJ_0+mBqj@7wg#kA6nn(yt+}!0t!6Ad4!?Cr-7p4LriP zlIL=GE`D=PgZdIY{`Sx1y<@?{|6ZQ|J>W{gKi`JS|K7>R`CgFsj+JIz=#*onSqkpQ zh(iwt3TWx0zN3m#oKEre4v<#ZRXFFVTJ~etpBj`o%HBl46s!{=Ylb4v$d3{L^}w z@N8~fKnRUfb^qn_=bu_-)+4v{;3Cf+z3|*RNC8jap=iSSJp=Xa2`6OokwdQ#{ zeyZzPJgYiv<)Iw#!f&CSAiueFZ^BxGs~Hzra53{jo}a-z)*y?T*;DG?Q*E)>T6`@0 zd;E?f=~Usae5_sMbYze(BG_iJhqKPgFm*M0ek12dzs+)T#OjFQE4NR1Z9nS-H}<3& zTjI&lbN9ggj+mG`Xa5EAB)X3nK5t(J?a4l6Ae)9?urV(qyJZ8c`U;bi6X+;G1{IHT z*G1RjKb7G>)#5*u;XirbmF<~~?a>}<*`A8fE&EEDOPPHU{H8Xf=h$cc+R zjoa?%c-fwwU<+S}ABmRu+(DCXp!Idd%X-#hiG3CD=f=ETSoNuPK4^X0w89O)Mz=ih?AG{gtYrB@Qg}10v-7nHS z$WqLdvxrTthTGw>i-c2{>sW;qnJK;j7mN4A6IQv2Uyz$@??s3)?%ur!=GwOx!SI|1 zZu#e99hiH;i{Iuw?;(?Z*Yl_I1lBjPpPus-nkFbVLOu&LW3IiW5E-ie5$mz=mg1Pq zEgCD1R4=Qr!b=O_E$vB-A@@5>J^x{mPmgrEAU~Fncpr5nIdRJGNl!TKZZ?UoFJl-M;3-TLu?bcY+&h zkV^{I4U5bqHu}Jn?&Xi(ojQi?)GyH(YW${w^V1%k)iZwP?1kYyXD`IAfZkL2gU1yo zj;{?wEb6^!57SO~qX6Fc^vmRJ6B{SaxCs*xnell$#{SYs@-?oUoQO}weq90lz^Hje zE$>y-d>-FF*s&TsdUBUogOL8yUjGrqqD!gS`L#1SBingj{3zZiMpvE7oJspl)yH-8 zaqp3mefvX~bR5K9>kRM~bcXV*+FE|NPtR0dF*%WYYIgmYOlUsmx~^&~w_`Q>u5~nV zM#i&k95`b8bM7PXhShYq8GJyhCp2Rx9}U&Svbrw`e` z|97_Rhu{m@^#DAg@eEKSrJ&~TB5&S>?C`E?oK9ry-SJTly-V{J`flI zcoaYr|9A*Co7Z}e&=RzQF7a1-%3(NFv=xLn^${Pt&^(U1*n{n1a*#O1e*^ykgD z-$Bl+l|Ld9dKg*OcWXps6yHt%waI+KCu#5XH2gcpTfwhNs4 zpy*Gp52f6@5On`Cbk=vpzGaJ~lb>O4f~~Jljcnw7euL&YMqj72fvr><7po1vs}1RC z1HZ@SxPH!34_*BFr?WQ0tWVqlPrp?gN?D(1T8=EPWSy^`Km0Y}2OP#Ho$T>Rr|{eE z{EH$DpL{XZDe0~rxUsu>5PyR=`s->9foIHm$#;3akhuFiEAh_h z@XYJhwx-%(B(@U0yEdG_U%Sg(6Bmv*gcA$7#$96F>Y|9oS@&dPTppKQ+4xHLa?$AB z`mnDSL&)fxPe1F{Vz94WzBTJdjKBEzOwTxJ4M{%H@-+On(Z_#1{up|Onmt+*SW<`1 z%tME*CihUk#kn^b*^tjD9}(Xh{hG-dQz+Ut>HXy)?8_|Gy%-%%%r`n$x)VQFvWo1< z#=1J-&@#!PTes1c7gv{dbB3DQbgvI{?Uhq6zQ0qw97kSVeAC&B%Y6Hh{QGaklgE|+ z4`2CZ{$60xex#E#sLKX#9ZUU*T=Hf7=iSDy#5P<3&28U|IzrTgLmoERwwv6rtSI?o zzpN(~Sc82!gZ0b(o2WM#nDVocHnEbStML$an4QYgGWe2kBIn$$POn=s<;iGo!wEKQ+Dl&rI()?`bT58h47vNcB z89jJPnPr_7|UE*P9@hH!-x)&i&SJ+V?9zv9Le#H3?nHx}ki{AbJfu1?=|R zY5!ZC-3-0(f1>Om7}iF7T;D~{ja++e2(fEwGzB{rGar~-Ib-5V{HGbr+oUU%OIv|$ z7w+t#ZLG+#USQ3)k#(2Ix+8UqC$nZIzH)QoTUq;7o_Z!QDkt8CDQMO}F8`aEC&PnI zt?x;;&-G{Ckqzh0v2 z0pwrvhYZ{K>^cXDKV3`j>I^&b+mrKod@{}1ijnzr`PIidpZA|1BYU8ErF4mWvmm@W zlE0OX?`*;IYWqO$2k%+9!86W%jc=}Re(v@^UHw59=4IP-mJ|I~73i8A{*nCYdbQvi z&VByOeO=2}<&SIQ>kC%F{F`Hd_X-cZe|F?RvY>x1e;hPtcPHm7HLT%M3$F1gWa99U zfV~!rjx8rA@N#s#Y~u{}Jr;zYnW&i6QYs;*Frc|I>X|5Ny@<%1(sj_| z%NMEj@PIr1xyIj0?$^5Ye)KkVRA$_pt^G+$X5ODIAK&a<-^m>6AISxh-Z#&zoodcq z()aSU+33t$H}mY0a%!-^VsBVG&QqLsN%_jG#3trb-gYi^+F>5IX$CfuHN1qrmt%j8 zzAa16Rqg@)|Mb94+dcei?>9^4W8LlR=R6O*vqqPhd7puIbkN{13mkMk``)RER&0Uz zgm}x)y08_ow(<{kQ9FxaZ!j@`~At|%HQCmp|>niI3T-eukU4CuYGa#WY* z9Ht#lKeAQwPhO!P&7bj4zyY5@eh)VGdFW>Qfp#3n_8YP~`X#!jO~r5Mmd}CTp#9v- zp_TYXJaaKL_wF}y`DKoYZzSp6EgIm{D8A9XU2_z4!Q=LPYBFPAw`u}0-~G3TsHqs{ z+TwaAHCzhv3nInLE1DR4`3O66tS#zW{GxS%7UEFyw=RJm>vv>CHn4wsH*&1KCn>)( z=hE*RoEUjlpldw+Y_c|GI5}p-<{nJ;1AUg%h0l%1j|4vQm@* zJf$x*mhQM}Oz&JtY=AkU^sqaw#jKmpA=VjS4E5Vg#x=~imQ>ClPZ2*7zBvWhUHXVt zs(Tj>N}m`1DgJo4NONYz-n`@GKJVUBJKp=mNAZoVj(*G|3M*$$taiqe6`4lqBk~#= z^-L-Hqk!1I_|?qwz_W>QN(YCD{{f4)9~X{v&%3aBpAR{_^y8<9Z-A4Y5suzx+wc$zXnr{I|3|UC=w>&laG=K8ypJ#pc z-pp5nfg+7lj6Ac4`caz?c=N1V4>eOywIx_mM17U#D#+8_rC6(;bH^0>CI2l*t=P=2 zTb7cG@vk!nFIXX+5kt2&&{m2qb)HZA8~QX>M(!uHz09R;S-8kwFJO#;KN+88ZnMd% z85jwlt8t?ZWB+d19L}=tD(Acy`S$KSTRzneY?kDfc)=2@{M5*ZCD=}$DHYEX&(L~E zQ!VESEABu|lo;RHgKyVlK<0$6V9W=`mmJ!r&=()c8{bllL^iI57`-(yH&RN@HU1_#3qE1{z&B#a{{zKi4!S=p72XCz@lpvdu$yH{5dVWOn!e5Sq2ceYrM8AEH~W0_>kJOE^K*&yz7;$N25;ytyyM&p48Cmzda(0OXi#ho?z|M)6)ntqDQ$fB zm-4IZ^UKUw1Wz{O@6ziDXjKZWYN1s{53Q2VBy-GMdf4|mpHAAXA489w(1Dl_d88}% zDo-odGba8%#DAeua(aeaH|ZVLCzpjLT1EOEx|mP)eTD{G+d|ZB$Q+A~X0B@I2HJS& zeevs2`rG5+L`~Jf2H(0}wR`C!+24&$e>Xe*U1)7Pjs9j03vWDai9W53q3_ZE_m0v( z^0snSpz|zn#nx_FK>lkMIAn%7+XG#nIoSgC)LM?vq8DtLms*jh@eA&?{c>Uqp^hs2 zgAjT)GknzkZ=&CeJY(6bZD>>!y2>h=jQobCfomvyQ|KJ%Fqq%VxgN^rEIxC!#$K>k=cmA8A){8!xsk`*cA*it z6pet!86op|6Q4J6?y+hY7mZ*{GlufXm}U&;lQGRW-8`rH{x+V~I!qcGoSV{b1$t8U zrXToM!{=+D$H*M=2`bl4)EI`kyDtI`Y9MDs*9N;DE#RJV=v!Dz)Y|+3#=J?k$mpYB z^h=w_$v|h%pnu!vr0xSg_XzS^G{2dJ?3mmS#>rksp;kL_Czt2EuwyH2dExMpNANuTEe1>fH4`I_vS?|Zj!Q1)JF(dkK_~f#6)!2)T z_|1xoHjl%$U=!oSXO)lJV%oCLn$|pF80|4OTlHQU_?F|}#>OdTgzwE9t$7l8G~_Qb zju**BYaXZj`du(6f16mcW1oZ3M%Kkr#xrQsB`qDDdoFG5_bl5N|4{D*vFXm4ojq;Z zdvMh6;!?RH;D zz4LL*cgi$YHhp(s>%;hWT2HMWY<1>4dWN~$o&s$0O!kRVlA{N-dkN4l+?c)}`l+u@Pc8|7!0C0fZ{kd8yIEwRD{ z5&QR`?JJkrc1(I<5`CFmO#B;{zYZE*mE^Bx(Gxm}ud*#`Py3FuBg*Yi{6YL|^Ot<4 z?C6!mt(u5Q+-9LWz)^f-YDq9oinmYtF}|C_Z+chjbzb@y9~A%78+-qhhaT=ao420K z9R}Xy{3h>#K<91nb_VnGW^xEBtlY>f=IJvlXLilU&sa@7Q?QtR!AriyR`O!v*&6HQ zn3@>q$dQo|UvzY;^Q^1$U!mTTbpHSNRISHykcM^opf_(E!DFwcqM-$~aMvjr2|cy^L%|& zr}ps+Zy%F=edGj=t&du}j~UR~^tmXtk2?CmX{MfB7~H+El^Xn4!^h^m>e3Yaf53Y? zT>K{nI?KVoc#99lpY;JF{N|;L{IX)@;z2X-?n(()AGtK z14G_7lQ19$L-T%=yw=#{y^$On@ze+1?~ykn&q;m}pP02O_@+R6I^orPtN2e(ID9J~ z#fwXiE(X`-*a7c7Kc2-tJky>B^bx}I@CU$i+6Tbn*4z0Y|LG6tyQBF}nVeT<9fSYW zek8sB)XAEmWIdh#G{(>?seigBSle2$!>yb-)4b1Kv-kz)fwxZJY#JlyvYXDCcpjZ1 z{crRh{P;(wKbId{;YXKWHu}aV??$V0J~B8-b>yn-bJ;D{TwOhT+o#AA^UWO=e89MD z_4TiKncIJ^FW*nG4Pvg;UO9)Dc*a>t97EB1;sI+o$6`7%cmP?N5A2fpnTN}4J=I;7 zAlNo^`Zugj)*kG8Z#wthDKlpdPe;yr?sYi#UN1BCbtM}ce$>6(o##q_OBP+9PxpJ| z#ZT+g>@#70=*9IkYC-$-c^~+0stfnV_sTT*Hei#5uRbe$_?mm*Ywr2*HTS^R-1Ff3 zZ5!WBia8t)zF+(Re8>34m$4;=Zg9n(%T9@ZL(#iTtOr}l_6;PzT#)aH4WaZBPiC{ObpD$`|nS{7-Bav?bz_FIXUB{$ z^L{_RpE&{eZv9{4d&g1meb(fsC-be9PaUtaw$;4by}ZKM)I4)en&t`zHOItZW7>KRXW-h8_U*JnNY{W5YA)7ATL z{=T)XI?-DPTt@uC)d7WBjt;;cG`TvUhO_Xb1HK>LrZ|uC|GsMCOdr23tb#siZ6Wqa zxdAD*Fh%Y?F`1O#t>{v6h*I={Y)GN)8;$eO)XcpL!`q?`AG!MAS)ae4nBh$3{{`gb z7wovutbG|BvEzLChbcZp(4`yn{WyJy&-&>Gy*`u=;p8~KuA1NI4c9MMeEC0o^tdC9 zy=m?0&1a&U%V)dN+M2P3&s?4Q*h3$`{O!9VjjlOPJ9C8m5-(p#AFcGzudnlH#`yTj zwl8CQ@y=b)BhK97SIM^6}!YlKQ1D z`8Z~MYhBPwX6}VXMrJ-{eY^8^pUfPmzFp>_hp|6t^le&ywB~|d{!8UY%O87Y+H=)C z6`r|yEqD`)Sn`2$^ECd;H$1p{{TEvwdgd?qAu+|2nU~<0bEe+~ww=KKD^KvD^my)Z z<|BL3#31&}wCCh|DtgBt(&Bm9vEu3a8ov4XXFa(3^`W1?wa!Njd~3dP{_rQH@39Hw z_@DF!-m@fgkDEKey-@c|47MSn0vJFZ*yXjsK*&(s#Yx+lR{a z`07FV>hMOhdFXledV5}o3n2u=gI3! zxpA#4JoFR~J?!Bj*FV3*_*0(sJp6Oz!?K34q1M&Oka|}I_`m`B+YG`4d%UU`63F(`?R3?t$ zrN!_QK#MmQ9;0lzJmb}?9$GxLB~`b$wAh}8KgRU-x0lrIIJcMY@U#2-^tV4!-MSeX z;BO~&EAsN8=4$Z50w4a~_-7h_0bSZVe)Uc8{kV1M4aYj(9)Cf+{$&Thbnz=&-jn=o zcsoq3i=MrPy?J}SJGRJzZyx@Uj_oA>*b+zSb8L^o<8IBs{=RW``CN4(`aJ*2Y< zQ@=rH|EBmC0Bi7yz)jlvjmi75ayYh_AN2fUY($)>x(NCQKK6!P)G3td~ed|ck ztWEK3_~*CE)9}SHq$@?YHvGt~=!w?C?4fq#*Z#l7cirr}j{B}-zUwC6^?u*=|M;#K z`mX=tyDs)!|EKSIpYM9F@A?Jb^_jlwvA*kPeAiF=u6O&cpYmOw;=3N{yDs!yKjyn0 z@4J4)cfHwn{Y&4q>krH^`7+6TgFcqU!`fp{+8!*l-)e|h(pt_CLZXScC;a~nCJRRvsUkc(Z&-*om(mauM|aUs|I z>3XN~drlAERLGy6FX5Ve)NR^(scWqlXM{_-3ivA~r}S9P19Iu}ENyxBBWVBb9`eI> zuvVn~Kymy~gJZ3od-JE0lQjZ_XWM35wp8!RyT>Hp+0(0YRowm#n*7V;c&NYcokac= zW1u{cRsHR~%erP=X^--U$Za-fges3{2DCB1`|^m^S?@2;R~`-RdGm-aVeQn-BP!*) zV%f?YYT-H4j*~mo)2=Uf=*Ql6J-I{Yt6e8|NHsQlwd>>y#h|b0$Cod3BkitYzf2rB zWP>zjpGnuwJUf3#JYm}LgqthswU(-CopgG82xc}O+Wso2Xd=Q4)Z(b+2nAYH`|)K5j>U0 zxWLM4Y!6bq;>aTQF=w%lS!*P!YwgWDwZ~MogU%pls@Gi5`(D7l@7Xs$)v%s@-y3o=>#t+KwDPyr zR~z&F)+^Olbx&VI>8tfh{1fK&h3wgwWm)y>cd%!LGZS|4S(1xiy54HX9twS4b-rcbK$R+#k1-?bVw+Q$a z0p9_B&!IkdPHs;9oSUC+7{mAWo1blH2j=#i?E15EbL;Ed8}@c={Nhx>2ep%rn#~z{e~c?BGZ} zML(X+!k7B>!JT{#NX1+E_BzYAjXA7xZ+qjkt04#;s?}JibAR!F+pbW}qRd9=#xohm zgXC7{R-QRg^;)`V??K?#zG96}2786tC#CYigMz=9vAF{HCxsKLJ9V!Wpk{k^M-{es z1mk}(ZC{ez_2_!^Tn6pM$QPE3xxDmqrwwwAqfJ*M&r_jS8T|lP&BB{*P8R}Ka-m3^tTc`v~SH{dtf$WEFQcM9@Mz4 z1m^o+xcs^ZeP(pJ%CTexI&v|_`oaQ6Bv_bjp_7~BFVSv`H!S7eCPX&H+;H%q}8eY zI%^s086&M1)J`+Hu^B$D1}5oa_dB-TuE*!hQKV-Jup#VQc!64*6IbrA=9a@-^TDB& z{@B-773Lj#pM5Z@8Y%CPqjEHQF6x7xs_34U8koyUqF z^^der=W#!IqjC7GG`H(lF?5p7VeGEX+}I6GGuPd@w{@*`(|XQ3XuTp()CwkHqNWtfX>N&Sc$%#m;I+kNDp_O-!- z+2MQkPW}G9hcbVf-O!4B?c&~btFjw*F`nkkM{-KF-q_B(Saky0klQA{Yu~W$>mKJx zFK9%{k5av!b=!cjb7Ep>0x^|7Mkaw!v=THaZ+RRvSOyon#y=FZwom zSUC^gTzS>g`^M*8zc{(~$m=`h`^M(yjEUbra_hMrJg;{>H`Q{}*&|Kgw-U2IYu(lK zUB(Psbk^TloXN!R=+t#^UcFn{rw$#M}GC8Lr<=H=FpRubRBBNM%7_&>W*Bs zFKg`Ly=Pdf4n28Ss9}B0Pxj9K)`N#Gxw!GrsLvccwEpJF=!46btv^!N*a}}{Ju!dp zx+Pf+E$E~(PLAx&I{BA}vQB=1KaJt!2Gu6Ob}ZY?^KI8!rjFiW&R@`3zlZCc&qn97 zzjO8le57+|O4#FEaNzVvDRjO|Iz+$m{fQj*e_`u1rl&yvW^`Q3AaaVY4JX$ADAdpl zEG_f!Y0j}a&%~~sak_MKmN_pZcD?9A?L}&vy1Mle#@4jKJ8Gj1{)k;qZ7k&W_RPAy zx1aamp`6Ns$Ig84&b{~jZvCMHW39*JThF+13c32Bj@%KpuYLU$QxXRTQ==g>%o$i& z^>?txqP=oXm*jpXdf>nf=eeY7#%ZsV87(8htj+JOEV*Hy>R#*2NmA%38ZpBzIQyVJDM z{(a>4-@^F1@^SN`w;V$|Thg?nb8*nuTaEwxFk=SawKA^mbul~{W`0{Rz}jMddxZH7 zzs0Z-tIxFki6_uQW$?VqD=X9wW0%S!c8ypgFM~g8 zU{pSh=ov>Y?i3un%U;%q-LG?|68&VUU(;Vw-Y=ss*=x~9x?DAzR_7<_-AtVZ7p^Zl z^jLV_oqIPt_`spv>z_Nc@Y@Ft@m`nSckPA7$?F@8b~7(sK~9U;2fEd%H*f5n9w$==y)T zCVw&M|H&_ooliT$=WN;nr#9?gJGLi24n6|E?pf?z!sqVB2P+_7SM~K)tEN|&+Q;}c z>)(EFd06!_&xQxj2j6kv3w|x&-U{w*=-YPml;SUaK7;MuHO$>_FSwxn27jN~(4#-N4}umBecf1hd!bG1Kp(Vr0Pfi91e%CmyHo7a1eY z#EWfbyvj-vt;IS6u4ZXAe>(5NKDRlK*g;Eh3{}4{Rihg(vz5TlM7iTlYwlikgC5b0ntSzr|4&N(}a8B~moRM{}m3ZxS zYB&GEO1#>^ymJ>k(g?4t1t#9tTJseXfb(JC%m$D9sg;f{Z|Z<2IM=TAb>{TIpwA8H zyLNu>Mt4i!9pL_8?!_LZt^VN5we*y99(rYuVgu4$qmd`Qe*l=<&H6Gh_Xp-j;jeaJ z;tavIW^`{Lvpy+9!RS}XJB>da2Cfp|3MZP{fr)efTASh3|F*X1tijg(^!<|X!Ojz_ zSYw4Jv_qo7p%L!jqY^YYuyiz8<+yfGWqH+@f>GbsZaSE z-gXW)h7%jJx_%W8UON6DKIIsE%4hK@pT(zqmG;lO<@XKGzjxKXLHLyC<5NDrbx^}5 ze9G_PQ(lEnIToMtOITT-VD8A%Rd1hZB6^r49=N3gR@-= ztZA=3&UxG9b4)L{_o+|+4tqG^yXn$v`mD6~WKRDcF&^}rVi|e(gio;7JI=T|cFORf zy|MAO5-b7`l7XH*%!Ty4+<=b=lu#Bc@5`yPN*C|u|Im| zZpQjgfm7;VWUSv|tZ#t7TpUJ#&)djGUwC}&1mp2GvggG`GV>H2#6O6|@D|_Sd7&zPn4n>s{Cjx);9q9h{J}a^C)@vHt zGP2CPZr73Jjm_vNjYkXf440oay^C&x7TX=(++Jhz=5~0p93I_X2X8V*N#j4+eEDe}~mswUij^o8iPR<~1!QhA!EY z51ko3_wb38eLsu?fiaGqj#DG$AuFRk&l+gXvQ*BMYcIuLHoOC){~+GLM!ja>w{?i# zZ-PHxgBN4i%|hlYX5Nesuz-BZ6Fvs;qHCA`#ChM+lYYH+&4=(LI08tF6Xvfb9Oqw2C-67X9Z9diS}pU5%s z%S{-*+MgLRI_x#@+JCOqIS2jL-F?!=?y`Yhi-;A=@BA^f!WLv%LtkUv;S3D|DfynO!6Oic7JZDvjkn~_H#S^ zc;A2aKl*^B;QP`cLi|xT-@2u9;lR2a0{i^;NRXoYK`*pZSDo z-h2d_pBn1?5w-aDpB(C}p~nCIPcsJiB@YD#)j$7U$-cWESiJX?*S8!RzwE_BKltt8 zL(kv(JU;N(_6@sg(cTFG`M|>(x&j01i*ETw!+F7h^{??AA9>mH@BOTC|EbOxR%cUF zI@_H6{V$B;ZH!s2ORGT{hE`99>@mDCZ%QIBIItu4XV$boJ7aj$X;TvC4H;NJvFb)M zhSfa#QNibqz1#1`ANBOR#NTfg{T^QC#C4g2E5_mCa|!p%*zdRT*&WoJPX4LJJn2jP zg8eU_q?}!+G5>6+vs`0d5bB)4nCFeLI=}7ltyPcB~>F_(pH;%_UNAm0Y zCxL7KMLC^cItM=dGV{gH@Hr;bnJ*o5E}!tDeh-wnyqLv!-`eAoH-5tK-_=9ce$f_c zkh$`){u74lUw~_RB+&UXbaZ{0qsjBOdybC}K1z9>?$bZMxqXWMx$^|VD}Cdge@F)< z^$+7t9M2v1(Vy_Rzws{e^wm%&IfT>pUy#$eid=umS73}a?Ygi3bnld}Z#^{S>wo0$ zZ-<_~`Kt}1Gc?aDW1bgjB)_F`bieF+oM@N_3O#N zmDI1#e!_4q72ToR)9V)gS0%YdmE;=TH+AQs1LPXz-ZHh}pp{!c=DjJ)#=N(vaq$4w zP=3*`!KQ5zho61IILr_ZB{`kHFMww*<8v6FU(e~B4=)ve-s-HEKge3F=CVoIH0x|0 z*&G<%@UtY3n7HeQh`k+;KJm_(Pl?{KcYNij8_#AAcPy~Z`vhSff1eGjZ4(DZWm7GU z!9RR>nqT$_;&IADcYY|Ef9?~6b^HT1tX`V0P1vyv{kDql`!C7qjP>}w15)(g!+%Ti z_KAoo`~9UGfzQ@|f-9*z?)Ze^`kLrgl+)S7+U5Sg^U|+>3jL;k!f>A1LqEZlq+ih| z3|9tl9lw1};*tFcGPp5LxBg+SslPh?Dy#ES&Z$~N44{Q` zO%71g`%=ybJjlGdFpyK# zPM7d|$zDs`f0ET%MsC(a)af@i5_{Qo*ohr7Hz4l2Bsqs382#(r|2OdHI-DAJJ~bG= z=?lb~hxZQ`6=6%i(cdca>dR{w;~4m8t$Gple6-#*pK-j4@Bc%7a)L4ZV%5ZnuLVED z_z#ahM9$|o@R_%fYkd#7*7uNWeGj?TU#Nb*;UBxp_I<1WXX>BN{9I$+)%F=+${|$@ zU>^rVWoiK>p$h_jQTMHW6pn+V^|!2!9)Ao%bu5xetH7 zm&}k~Yh;Gp&;sRRBxPokD>EnmAC;LOn=!WeN9P+G{sEtlf`3#`#pRzrk`E&KJwt9< zFZvP3HS~KVDEjRwc=BTEOx&^=ozp|VouiYya3XY$bj#a5-BQcgySgO<-ST!YV6Pdc z(JfE?^u)sZc91+4k6isZBYaa4I^<9N!bKC&A>(~IWD30P>X08OSI5yIwS4a)_ddax zOsqO@Vta5{^j`GG^O@EQze0c9jsCbB{c$(?it4B_W4gqJ$!2$TEmmE|CZ8>Pp=u-~5 z&q(6(Nyx!n9(~b8-5bLz!2MD2$~F(LJVX1v@)w8L`F#F-v)SuXa~j>;K5xERCF5Xu z{+WEoV8G5dE7bmkp*{PtVpZ6s>n+ZhA^)??Lbs{@#8i=W6h*$^T{zLU~#CezTnDO7a69986C2Byy_gBVJ6OS6hjD zIVj$$n3nWP;_t=Ia|ec`)+@N3E6ES4VZX@yz}#KTt)iUc)NS>vnveY+4UXD(^D66aqw$^ey@EJ!F`v(z$~n@N zGj4qOz2)Y4?eopf4HtF4m0`|RZb$bi*R>tIPvSGineI&$fygNAo%&7(-$7_%a&3XD zeH=MU(Cr_1xBGVXqtJf$_X2YTR|$CUhem6Ft9`n)%|44d>BoJ?83z~l_USd-)OH)X zM|-LkGoS3oo+{;dsMdc>^o53o)`IhD?d8abzKd-!a6XNp`-zYkPaP5jYkC-3_*w*0}sry>u``OZZ3`vm=VK-X8; zs}t+#J2|$0TV~5x&)9nU?treZTv<}&wyE)UVfhL;4`6Ldecw;tRVRg)-an_x^m*Gm z-P1by=yN(FzWNcqs>i8ZOUalm%h^%RO2XIak6jBzS5UXcjz?!lRUgXUXAz42 zobO(JrkvynoPqAU{Kob9TXV z>|be`XPKPk_#*TJx%89CrMLa+Q1l7izk&N_TkP-Q{&4P}p1l8S-TyxKM?3d9E4g&2 zdp{$(nfq_m22##bdKTPt9>5y*6~y7MQs$1OEuW4kKU!xGNw>x4@i~c4?Vp;;z9D?Q ziu%V+*yJqui@0oAx3T-Jk&a`W1Q?9b-9o$E{RVYA7F(^-^*SbIHG3>q0`L3t(XH-2gZe(+f70fyO|u>eCsyf z_QlPz^DX}V#^2lg4TYxd)PB8{=v-?h-l}Eq06OiB?f60Hu-Dle{L0hh z#IlFs)%)?;mRpH0p3PY%FKSOfsBycM)loPlfNz%7ar-5KY4K|G;SKN?dm~~$p?&s6 zEX0plZ)G*MVguNZ-FU{4g^Sy-DM{=`j@sBSVaKoWQLbfQz)Rsozi!4x@S*cZWCylt z|5e$2)|TtZA@3eeUeZ6=`&!PKD&NfN{0=^AH#IUC<7x3-@bduHrtm4L^KjFBix=Z3 z=2QiA&H0So0|IkbljAjEfb~LGa9HE;?&@X3v%^cpPd|X4hFinywKq$$bl`e&<=5h0 zJk5TBDssZNW9RQ@|3xkFv1?0;UT2T;8`HyyH!H%4x7gFr3fioPzb|F?AZ!$IET&~Eac2b@LC339T_(P`B+Rww6g zPAeX6b)E%0J#_dGV|=W1xSRg7Is3EwfHgP&WA1AneH9M(*pg87(oE(jS=6_vqb|m1 zba?wW$3-4TmsBg4T(lsUd?#`s`S`eLakbz>#(HdBC~V*hVZ&V5&H%>Uz_?T}BF6?s z#&3qkYvjid<5`~Z!mdus=KRtVF?It)n~_P{jKbGjXN4BKI?4SL=)? zjos-VKV7i#Lkmw&!dI7Lk6jjXaVo+Y*t9RKc<9Fu%TB>E)OvyamW?^=IoW8{>^(D( zW77(FLV+Z&%dWLx&)O$ogQsGPjX$V!AH$pfj{VI|k6Q`m(OVQ3p(f(crm5H)*8Ft; zBJQKFqD_y7%`@`Za(O0pFE)(*c$!XJhn#2I*$a^q%5ISUJ}VF|I*^mokcTe3 zpEZhC(2d{O`J)xg3->%Y|CGpU=*@?A{&N3Q(vT=+9J*971bI0xOK)9`fk4#-=P8j1pTb> zsL@%;;eL_%W39V{t70GvtjtDo*i_r4vAqF&?^(YwN9v|V$y&yvyTzJY$*1__x7ozkJ_dV|A-Hlvoe!W7yF=vWBLBMH zImOtqJ+miGG4rzPitW#tWmA|7y60EgbD?psb}v7f@$=gKqxmzvZJj;0@cI;g%`p78 zXEr(X*wa1NkzcRRnOFDjf5^ws|8MBOdvK`pLD6NXLl^wQz!x35obAAMz5~~|-f+G8 zO80UvP5koX_w$a0E_b_hAurv~g`9Omm%`rt|Ndj>{|EG64PD~YehM7vUaom%3pz-9 zjy12#MrOLt;XGXUb{%@utY`h($r0%j_qiV55I*`ad}7Cwsn3Ky(R@_%64Cdo8mqB6 zoezADu{sA{X3eU5Bx7e}M`w52=f|#R|3!5^dkmPHvuCbnt{&q##TLDD@=WI7(#?`N z#W*!rcE1~3oZqYIN3?MFI{$_}V7ctZ z(?`*VWK;dzEIK1o^G~tuPZoSFpJSwAlaFOZB=0}XU$8=FkhTHiBl!yr96hqV)-$IC z|GspcJ*Rc_Kk#a`Q-@JLeKS5y3;vzzW%VN$wC8tzYr#KK><}Goc+@=0Z+(sx zv8mV5YrXYCHuK_~lzDLvf8kq6UIvL?-#Bsfc@w<)qz^fdJ#_HWhrG?6-%l`oo_2Kj zvBq6`k3ss_#4l;L8vc|WxR|}lyS%c`*rXk6WWtT_^zNs(KlHQz&)%EIS9M)?-{CaoCEhY(Py*KT{q1v(baWBeNz?av|9JKJ=yT3J z=j>^%wbx$LUYj;__OtTddGY>x*UdTP>9WC|POTB2t)Tuw=F~)F^6|2*c;!;o)A7iq z-t|W7yIIdE+W{}wguWR)j$HjVa`j4Vp?9MjVMmlbapn!y(|Op{3Rl_scOh~x{PE+* zaeTlAb`gwL((c{p7|?ycjppr8@0S>dd>k>}Md8h>=$E&j56Asp!0*0E!Lv6Z_m*=N z4kPT1Oua3R*Hw6UUBw@s*QJSu;B{*M8MP0t4X^#z)B#=#&OVmc%6@LS^ z6CSnpRNmY7%53z3^!v!q!kpcQ{k@C&Qa@vzT?4*@pFuaLpL!OBhe0!U>Rm4H`Qo%eFIM#?fvw~SlD+$3o()INlW*L9 zr*qX(!}8&Ke@pI*@3;Et{pMkOu1zl{p3P#uM8A`(`Ljp9jduR!udw4HQp2Mk!6tB@ z4}TrB%6G5qAeQAm+1KSm_ea`!)`Z?$P=xROZfpzP#9dr1U&SpoXE~!jR+ig~UAgzV z8?5!SZx6l3dh@yV`RJkDc22HU#DM7cIKO3{-H>d2Rpbl`^(KxWuV4`W+~S1B)W!bB zDZn8y)9Rgq?%z;!Lrr2D^OXKdzpO+DE~yxApMA`4+y6J*_|z@(M&FWN)%@h>Z9QKK z?(ey!>cgHlmwd5yH+aP7_wMcslQ(X`_ndl%b@ss5tXa2YSnJ=y7V-9DTlSXwMsZydb4=OP;<_u4f&V92tP?>joON?BmM%fnsp=m zKX8p8Ss-eONPNqh8bRRBfR7kyd(@SSpzS*k$1!4CC|W1Cc;Z1@RAwO@Qf?p z*cgGA42PG@gqM`TOA6sfvb}_vGtz5C+X;ErtYOe_1iCJSj>qO&v)*_ty!S!hu*i%h zp^b$;U-yk4L$@R1@c2T*(EMNmc7}HRrfnw%@OUP88GdlW#$%Nik38$gG;qmLs`-~&@zr>?c0h7Rte&PcRASo5W1>uLG4TZm!K^PT&_6zWT&z8k1- z2j!)Yw$ola?;FV7(oXyB)VBhg(qYSABOl|o(RG{EpD5obH`!o6ugwcvX}cWX#?6vn z=*x2Sfev&5%9kx=?OBfhuj$Lah3VWc#dchbf5O!b-^8csCvR5KVyjm5L^s;86D|){ ztV-3x4qtP|IRW~JvR=sMru8C5IoU~rHv{vHq2BL}MhB+FGVkZ6J z(Sil-UXniI`hF38d`!RN=eQjM?e29*2|qn@X!`gDdxn(zbikIJ8b%*^A-!N?&57$6 z@4e8)%jgowM}mzjTPg0+3!A+zY#jOO>+!I0<*TchuYIhG_u6wVip|7@?@qUl2PC7R zuQ+ws@|g|W30GRPimtXklw4MwJiv#4JI?xWkX+Vh^Wm{NN5GZKc-GH{?;XsC+i5TH zdaGCU%tgLiF8Pl6-f^#TI9g*ys~&zE`Hp9%o@*J$%alXrNE*zuU$~SE7Y|3pczWSD z&xL^l$6=R%<56rcCWeB3x8KkDl9x#@>H?*uQsA3vN}6TX=+%2F&B zeILL>`%~PH;e5W)Jim{7@DT*Ysfz=R?l_Q-FA3Yn3;QJwWWczSh- zQfzjvjX-Nhdp_{|h1H9{*UTf;W%C%FD`4ZP!>!lB)0#`*sg!wS@;QHi?vP{k8i~432njaKtzLILi0n=y4B@a(;IlwLgM>L_SZ!RI$p6 zfib>`YyY@r2YB&m56&I86GsV8`ZO?}3IBQiV*V39Z@%Zk&6zi!zX)!&ez!{U{Y?1( z^URx>%$wg*_ah<0+uYc___~+6bzcQv;@ffYb@NPfXBGAt`9c)G)jExOe?5Sgq9O1S zbLY-)9lZP+TP=Ol+Gu%k!}~$F;XC0*-)3E9+4E@TC2%xt07uGSC|;HF33=7dZ}e^2 z!G8n)+xZ{TrrhMjNg;<^%8n+VjMq=}pC+bFyy@z&a=_@EokGj%eiK*XKroq&IMCHSjTynJFev4gX8o{8+kyR?t=?%dv>FU7=H zDPNfU0=Ca87>*tf^}c1|rR|)!^MJE*q%@(=gsG=|Lg4ID+MM0jXU8@XQ|jY)I{F6t z2Kr+|E#ldFs7HK3F@Uly9}VJbq2H>nBM_`^LO#8bI-A%>@CIW`E}c86ld*O2P3kz} zZrQKYfZwhhUm3J8OW$wdYGgM2{i;`Tk;<<^$4WP2^qDbQ*uE8asQEL<4-RaHid}PU zPyMk0%su7A9b~ryR>E;1I1$cLfsJIBmRBwIk=StoXXsxRYlM4-M+v{{+N_1g7)xse znx|gLq%N+W6t2Dlt|B(BWHQ?BaEznkCcS5o+td8Rz;iRQgz%J`U3QW3=1l}N@VK?-UV1-^Q z39f4{SrhI-HZBv+>l@eh|At?)20!sq3_01@f1I=DDEC+^I*!V01&+j&XBVR5$X+)V z{f4;nNC~mVCB$D@$mVKyHg&bIPbL5zX}`%lVrUc4hisYSqWZH7pS2dMKZUQ@@etyP zvdevya;fP4v7gp$_V&q@H@wgCG@liNr!gbrYhL?Hizk%?eD9V(`@}gdJPExN(!YCv zhwR-avHR+GuH+B;8$b3!VCQ3;8WVZrW+f1Nr?Ky&Zo#wl#?{T0H|{dG1$1U?UUXy4mEDI`Ddav(|;>`7x!NJ5+x!l2%`eFLz#6K%Ozn>0>ITszcdQJj( zBqqJ49Gsj2Mv0umvjcu~7@TO20I@N%R|pK5Q1>)tP4ERV6d?hy$>7Ch;uk217lUJy&g)g6naXzdx>L9x*+KEz9)tAZIT& z@l6q9)Yy#g-f!`yY0kx;qcIVE>MXLme@~@{PnlPlJbg;brejNeH*qiDwAU zR`?kA@p#kzCKqr20NzZzc%^-w+H`ZC+TinrI!gT6?PbUp^Wh^U$bHBfeSUN>WI*jp zvD*#3^xX`6)OYa=8P3+h$Zv)o8N&|O0{`34cJUcLbDTZykrnp24?D>JY0mJu3*Lm> zwEQw<{w2N)GMvgRx=fiDoiZzi+3mEG7kwOg(GQUq{SbN4wePRNoU8Ta>R|) zBRYRf_)tH;tWZ!Z%$QEu-)HZXEMsD!GdVC)M-$ z%Z&H$obkHfx6+o%+)A0&lKkve=&gMwXY*jcQorV4cg`da=BRew-%d_qXc&3HKXt=V;hG#XdtRJ|FXr#rfoN^;aJ;xtJ>>J3VLgEn`l0AY*#RI!$Ax z{rGx1kO50s7ud`BZWvk_R+=>_`RWxB$+>Pni>a@Z`Q@$ecEKSoM|9^zWKrrVrJf); ztW~CA@wD0o)oS$}?IM6$2o)!KS#+Pyvw#_iN=a$eI{H|OW_Ekgh+hg$6k4AmEfA6iChs@D7#-My7Nqxw@^a)#5jrOFTl>We& zW8;~#8OYwH^CvaIlb5pAiNuMi6Or|rh>K5Str|z{wt0?CTF)Ecq1sE;1kX)j z9hzeKA`Qgawrjrx@vn!@FJF@c{r7)s&Z-M#YcExQ*&=IUK4n8=6a$o6Q{ew-(^lwE zKC0RHKhnQ^RgDk*IyN3Yk`d}fZ(d)r<h!OmT8z26NQFJt@}`&+%V`eP4mTKu`=OJPlpmnSv0qGzDVxBE6}PoHEO zAF@Ht>*y-nyYJ`ozJL95WXLG8tKBcl;930=%wKW((-H???G5yv2P+xyS$JA#oijJ{ zoqEtm@8($y8QHRY}av+Lc%rSVia^`DBXUvlP0qps_p?@`tt$$iuZZQNY* ze(QxzQRa0Np5MgWkEU|Z{NemMbcxh(G416Qji@Q%`x3rysHM z_)al4#A5XO81H32v^f7j_vwE2&9WzUEO5~LG`PsO=bVd+zi|3~+(Q#CJsdOe88|;$ z@PT*VY5EE6GCz#YBe`9D`X?7|2QPx#hv`;I%?=OTv~NzZ`;61R(Rt$Fcg)aZ9Q=@H zSZ`mb0v@SWpj&c!EAueM^X7Es8uLqbv0CJxD0)>4n2L8sS*v37>%vLr9B~H@(%}w( z-#ja!2B~UDexZIQd;1C9NjLkqxc+7tcn!$Yk|C#iahW!vCWCXKN-FNE(LP|wu~)k= zb7X3F?yAjG?A3PX?NH|x+w)}rKf+6lvAS)f$`@kfQeXJ^@sI5F&B2Y%ZFI-|vtM6& z+{Zk9aM!u0hgJs7C-{Mvf2?_L@H}#9?iHs_*)+W67qH&N&ll~n7H);(I_7yhYkk8D zb(@7Z>D4L!8yl;I-4xzx_UMCW#bVg|(i3cb#ldwOG>SbmqL|-A*1a_DJD4X2SpQUC z1$%o%FQOaKjMfhEXz8+MUk3B^Yd>f1(~sT2HFjNyJ$JtD>F6dp&k9(>yY272bIo1f z{@(Y|rd-(_S>LcdewbeLC-(ZroHjNL)=Y0XcOCFP`w3?mTxv}d&HT*}_>%6j(}Oed zgagQT`RB@yXDrUIZmSHvRL)#l4gT`*b$R`6#TI8u0EYznta5qCp$YI3@tIui(-(2p zs+Iea_>0!^1oq$O!;`TabhqrR+nfWeq{ru4zBkgXQ1fi~tDm|QQ?w2kif;=ZetyT2 zsfT;jBf4<^+&K|{4p%H0-e&pNZ1bOcHHOb{EPeO-lk>nUeI3&>96T|GgM1V3FWqh> zk;fsS`$6Fo-Ygp!c$}5a*oyf_&M%8G4%H)`Ec~i};#YOxKFYiQ8~uIwa{UF~m+UX^ ze%Jnv{RHs66l`^l;YIMg<}&@`U3~xg*CS-wk49b0?^+}uGxzF|RZ6Vu(S`huPM;7K zUCI8mG3dk!zyn&KgQ4i$&QSLk|0NwPwhV9bbP!+V`8sK2YCtpzlOaOWy}QX_PZ_ z-r?1Q=$|=0XT!j{AfCL}Gq*KoTzF*}etKzmZ6_|E%D^jd5xk~3@Dk5#171~ebF0dk zTarl<@yCzDAAbmcJTV-RdDv;>2T!$*@BX_H+912YXZPC6L`K3JllXL zaU`5)qqB;G@bM()ob3UbRI-TSEx>kU9Bke5o%>2|>uxmi+yFd9cVA`o*eB*-pO}n&A`|<>oW+wlFL=hriu09`0QQN> zwN`B=`fq?Uk^-EO6yS`cDs*z_uDS|ccPZcLoZe+?eYKaPjeoBhOdIb)8!kQA^8J83 z)z_E6d6xdhSiM!lLi}`Ug>_cvQ?wH6Rcu)i+4lVBE7!E~ekuLXdbF;}mBnp2GRC*J z)efvlT7!P%^dVlpZyYyRKF~Vq-(K-vfb*#*b3V1!paZ~o&K1}=7F*f%e*6_%Mx8wX zeY`%BXN#SP_U}G4O@uH$)=F`x5P6+0}XR-C$`6?>B)QlM|0*ozD2T2 zZ&xfN<+i(Z5nHRekOz!kMSGfC$m2GacA`8taIUoJ#O;%5pFXr-6?#d$B@f$+&fF+u zUvU`Tuw#%-v$tnHwkqn15T(`OHLh_vr=<_<8Yug`L zvr5QImx1l*B)Dti`|Z4|<6SZF2#2wS+(14g@v1iDpE__>fFEEiw37hsBtSa}(9Yb& z2RLhb%5m8_>enXKwjv`{svdab0`j4@G0seELOQQcXO9NBK1jPaQgRPzh=rd|**^M~M;@kK zz`LjLp=$bx_a{9?|J9HzA;a0fWrf;)+W>@B6n(iP2|}W_#*y~vT9%? zUc+fXO$gbAH?JrbqKJdb_&Hk$54OQfOE3t?Rpxr2ZTe?etz2zdO0@4XQ-Qc&MGgSe(&b0u`go|=Ckexh{MYy4sSc_J9SsLGM-lI zE1{oiN4!WrOxb&+6SW{Sc<1II`2|@(u)N&^OINRa4jg&qm(}o&czsa1pmaj%gKatZ zPV=#yqL&>(4;YV56GQ(SYXu_P7@OpC$@K~77=oE-UOI=Kf6;#4EIKoNb#%lI${k_M zLB>3RG4Eo`o%p{s#!l*P<$WpdQ+S`n`$pc&|F8GtmJai6+`}MW$+^0=+C8hTOJy0&yFDz zZ2=GWt?Ju^ZEF4OpHJy+!7iL(B~<&cJ>w&cYypOa4sO(E;U?9=jh+iP{bQ?i>{Y<6 z1$)ZZ{scLiaVbZd?D&5N?93S5@4RD#&b}Wvu0irpoIO|ec(?x z_d#Y7RX z5PcTypM*wF-5cu7A}0UTeZ)~ftBSGjroMT92ajO9ipkM>95phy&a#o6BqfYY`n18{ z!2!GErD*hqxbb`ABe0+L)2K_IvS)bpMak=}`M&Ijm_M=n-0aU@@3uXJE>LOvc~aur ztRxqQr!6o1ZF}i}J=x3q74Pnk*RY)VgIZ*-)nR}_$~wAk64P2enso8=AZCd@;vhj-emhR0$~@|T1&S=)3zK?u|al_ zn<%4ro&C^k2C}Cg+0&2g=|}b~S)5m!kG@cX-coXYnbBLS7W-?bz?W552R6J;WQxP1 zS8gs(3%v+an8g`;FII4k^*z;GE*^kvg57Mr{7uiEzpyD6IYPFfrP{w(ajj!FlaIs~ zF8O0@s;<9@z8QLQa)X+_6Z0N>#FyPNKV)?O$>70jC+gQ_%NRfFO6w1;$2!V<*-Ota z+8BGl>Ro>RTN@Sky7GMK#!lwas`Et~yO>w2&wq8}N#@uZ@|m1sUR9qj*x1b+y9eIC z?!s3$o_WN2?Mcoie80?kZS94Eji={ZuRZelf{howWa#CNpb7ID70 zWsR%(7UN|PShkP6yRiqzcNyS%Ki42v=9g~-SLT~%wOO2Tv_s&<@vd3O) zy^@K)sV+e7D`=vQzD1eWX=8}XhBn3lt8|ApX3M6UX7$@titB0`$MviM6U^{y6@f^>gs=R9V&?0c<17i4Dp0Loe*{+dg97dwIcqe?0{m z1HHC4!2Brxki9v;N%W+D?O#yrK;GhVtJh+tPYwlCS`wxUbvyJEgm z|5{VEu8^0Nehy`0nnVBezKSuh{`Ahl#wS~x=5-$PTKyJ$w$txCr_Xcfe`YfL4H*5Y z`VLR9W1jtCd`f1|khk4saeP^DDW=_G$KDN}>@B8UY1yJ| zg5hU$KU}eIN#R}G3m1dtD!7)MJ{>=+>w|LT!ZAave`oXU_@raT=6FrN&$9)-*#b{g z+))%6BP1SY@{6a2w*iBtO9pa`8~-@`OZJfl-YI5Oemlv?&vV{gx)bLjdw^vBR^%2h zKiWoJt;iyhW5zLmr&5PvaIky4GTY<~4usRJoPWz>51cE52~X)-Q?MQV`6%Q9ac@uZHHJO=o6Jdgr1wDc_F08N?@@XMY=$;58s0x3FeQ?lm|c zKbSU)p`G^@iyn{}3=P`n6VBy%bbP3HIOF(4boC>vX4e(e18(I1cj@X;U>QU9zlg4& zU(wZY(G~Zfh^`#jc@SLd6Z{EchbFPw6Ib#>Or>PU6Uh{(j?@JJf_*-LN6b`0F(%PQAFH*;tDl7iY4 zbe)ORQIWiBwm7J#HEI+Y{uSe}+C)FPU2V z5cM~FtGar{_ea(W7E|cYL|vv$mAoC$wwfQt1r8GuD$!uU6^9}RINT;y)JSF zIpn#kuc)2FyF+Km^+`YC%j=t_5A8fu)k~(-K4T59{wDQQjj66a^bdC1RjlC?eSyf6 zwDlHOou%?N-+aLM%S>BOMxMKJRBZ{*U#FeA{`w}@R#V?Yd|N_&RW){f_0*T;^G9s_ z`6FeNdx*RP!r^q`@Jeu)IHJ02#PI5|yepYaym66>ul(8t$Xz$`{fxw6)p?^=SC@%x@dPmw3$ZJw>?4D->}=*Z}VlF%5bD{_{CsdpsCJm;?tQO_RMYl@>~T-!tbttQzWDv`GnQBv z?SE(Q`X%?j?}Dz#-(Dk`b|&}izo_X%#~JHOzyb*!g z7VJs>wC^`(;CnlW@2wTzoA&3%Mu~?chd=UJPanWH(2kvu^C2S1d^?_RkMM0L-|n|$ zANO@XM7wod{*WNS*(53;_26gh%t(@vti@d zs{McPLtEZ4v18;ePpO&v?V)8Y*A~}7d!5uP+3VDr!7_k&qHB}upv_L!kJqq6yZl?W zc(4BC#skUjp|Peh)&OHoMBcqpgKY5!p1q@EgY-=jdDBZ|bq1>6<%-_{MGLze?H7Ou$M1<{gtKfZF^mgL2e=kj`>(4~ z-hO`HLVa$T6|5;!SyPf&Q<7Ly4l7O@pTVwm8=H?#+1XQY^ndkq9X;C}jyt;XFusM9+`8uJ-)ihxZSCtx^`&&51a~XW7ac#mII;E!*Q1Mz zj?<3~RcVQ}YG3+KT}4uNd*;f`^3QC4dO5ic{l+hs#Qa^(IBGSPjY+iye7~K3x1=T2 zE(Mmw*pwJYGjT)MvgFHh?d!GcIWPK=f<^Xr?W@K2zbB>Ixbzi447nR3_=w6?$_FEtu@X~5oB0c@5L`;bRWh4e)A zTXi2GMl&z%uA02_Csrl+tk20VxQ}NR&vLO{rH4k;4F3Uh9Uf=s-?94@VUxpNZpVaq zY=Pnpt*dR`@b8}f*!IqW{t%zI$%hWBm@HR@^VY91tFCvPy54u@k;_lA%={UMXKX=! zb!Fvvo8KI4ev_da#VS-`^D7~~LB1l}23Qqg?bbd& zMOMutv`K7Sv+7I5j+eULS|mD`|E`#Q>c#9=cgL}fy8KRE#hhQGc;6!Sw$GlRaeTO; z_$Izb|CRsi7&JMX_Pk|(NxX*oUQCRR%FezG7%<YB-SpgF#6zJCe z)_u_Pe(3oCV>nb4$~Jv;bRCsR;*6ggs3Qm*Qhcf-#mLH<^U{w3z+*i4mM!ADtZBwy zNc)A!s${cqDN~MYS(|qq> z-b;q)%E112g0tX|6Rb331Nz&7tfTmsGGLj+xhh6qW-cghto=kNTfW&YWCYDI@3^A| zAHhMeP@bSfLtpZbCWI5r{V;Ri2aS5^O?+z`K04P%GT-4{7Jn}PI%)L%0e+YO4poQz z8v*btSww5awh(@G;K7=1a{S0wl?VLr$9*mx7>GWL`K@x2PY)2s-~*m{z;mpHTnrty z;E%;-_%2Q;V{bHLb$GVQ%ciP11Ya3rVGl{Ao>Q!C`OsMjx8C29y$@XI zdo^QBg89CZ@B4k%?%Ked-Q$6iVDvQyMn8;$&)LcZ`}~;+oVQgOciz?}@Lk8+P*40i zviJ6J)}1ltOLV39f^yWb-;@B@f+3R|4DHL+sB^4!70` z_cq;kz`yJ<1?>GG&{gLns?&~{zJ)pi>9keGvudlil6t36Zy$Nmw*sqV){HLbNj~=B z2F`Y0UZI@&5v_*Sct7C7qTVv(_CK}mq^>=!*!&Z)o92!90{c?g=bFYowE2vYHgf3u z7W$rZhP}ud1MqikdA40e=OJnTy>q^Fr9Fnez9eGkk{kP`S-pMbDUE&Q!y5bW8O;V? zl8Mb8Y2=y^vS3RB{?_lXK0FG19|ykW$lR6OkNQ$$TgJVOEl(viw$&#$wp0Q;>YIO! zb?4M;tXFnAFo@xY*iXIqydrt*vCE_0EzsYOFIVq1)H^DrvF(~+cD?j%{#Dd_m9?jZ zxW%!=Egm3l@c?m)t@Y5fVp8a99zHD-m*T`QE+a0brNruukF6|XjTf9wjIn*mEhXUF zEfe(Dif_oSt5^oFf4LJqd^~#iO7!q9%c>5-=N6!cFF+4(L=QiS9zJ2oQDTv%9Uq1s zK5Xf*+L^xO>YLEP??DG&wIoA}hyd5>>+WG&yVL;s$TKC*W8g}mcy z|MQ6r)ynzCyJ@@&qkB(CA6DCl-rbp&QX8bd6R0zJNpdacFB%>pTbGGv1K;iNs*Xp< z<$xa}3cZ^14)0}5_c5mX;d2kLZp{z%M$wH9@l8u1x@eppJQ=>;P92g(lU`W9xs$l@ zUF=s<4!l#)?8H55-a8Ggo?+fBeR199_o3wvm`Crt_|)cSUtGKS%!^NM&T3fm-l_4S zUY#pQzRy{yKCaMe5Z&C5oh10ehUQ>uea}(!=`Qr?$;^`|XEn8OW|QLK;^&|G(*A1Q z<|EKhC;d}B0qW7Y>M_o7Xn`M#x5t+|%J)I@on5yLr{IuneA0RJfA>2pJfC=^)Jy0A zLH6sp>x}CId47oauXuSm3hn12i?}kgIS=l`fm{bu;9-&thB<47&2Q{|fy#Ylc%>u9 zN`A>6FdDVW` zRp@-m^B}qL&D>nuAZs2`P=ME zcIoPo`1Sdq?N4jWlFWQ{zdtPse@e%n6<|%MBQFE=HRL%4l5MmXR&nu)r>|q*@P1%J zSTfn<{# zqHasV+1=-LR-v6U%kA^0s9$a8AwOteO8k56MPJ3a5t#<(0fTd${UF-Z*oG=2`PHpQ z-xUR{JB>Vo4%>!|*eDs|Bzv=&H%o!BH-FC-E4jLkIVbqLb+~;c&}~c?a*)|~b}#Y( z^0o_S#X)zT_t&1VtZzq0uVT-e^7|?0lS5MnZFmo+?RZ=ugHcZK7=~`( zoZ$;zEXL6(9Kc_qGr<9T+r_~I`ZWMwdrpt$nZZe5<$J#mPV(488-ONMM@69OJ$ui% zza;wFXHP6w~rPa_+<;&wj?UgKf$CVb0_ z*ZB8`c)yD|A-rl{i1$oBU$sVOX~@Q4Y#PwC^b^gy`0{&rZu@{0>!tXk)bI=DSweV& zVu|3>>z(V9T)k_!c+*5Byo2?4jLx`@J~dGPHqk5ncj;HW&O4TYww0?_Iq`JXjr1Drt8#gV zt=ri0U(m=M>)_)-_;{+%KC68?G$vlzf2}pji;VZ;OcKYg}!hKnv$*8jg@a@ozuK*N8Z+XABv4MGRHw14?bitug;f-oc?QK za_bYUJEOGGwX=Zwr#E)(OyKHo?4pe>r;RS=^+|a4KpV=%rdV9bVcHLu;@DtZeb(L+ zCc6J!XhrkzIbwx${<`)L6>6C26^-;n|3BbBjpb5ltDXL&`uWq_bi_0;G19P(2jRrTu&Rt;6pn8kB68yV|cIr zbs+{wd<|#A{-bG45p!AXD4C zGBNxiYt!$`do<_7JA~&4hnRCl#(UK2Z7;NXzaTt6>fjj}ue}gl=Xh|v5?mu2cyav< zIe`Y@`ct-y7Z0B!tjn$}a42pq7fcnSe-z$0G@d!<$jG-L_G6}mb!PTZb!d-}-L7O| zeVZC)apqY<_({bOab4?N#a~u)e*iu&`JJD&b8!rSF>lD&qUSC!zm3|!NP z=yYIS3CzoY_1zv=Zv@u$gJFFUk7IolkMqJ>>tuv6dwE>^`-8j}j}yGb<1_~(`xcTkGs4szMdBiY-NuhkiVWa&t!WIhCc&*<7KWNEgO)z8ZHdPC#3v|blCxW zyyR`|i)+nbK8%MxnTwsQS36UzSNf34qwJ?cr#I*JuFSM*N{Q<#Ag-%`xUK@?x`Iol z)viK@501kIav^_X@`aU=PISYSLC)?%CYw&I?09T?Cxd~SF#8&v6DZ zVV%~n>Q3IV4|HSa`PGp++6ul~u{qujIgD?IDQ3o;H?hjZ+!SmbUYm?9aRj_Q*n1E4 ztg0^HzM!|j+;_1zR&g}XX0C3&XT$cMo2EDStj#>xvz9orwHI@9?Mwy?jC{W>NGg?N(>n`JsS%k;$wwOMh$ z{k+TIT?RHi#SQter_8~&ItSb8c5JI{*jB3+hiWB*SDl}VzqTT>68mY};s9|(0TV}* z$vGNroTIUwb2O?rMaZV`vhEfV-!+4)`~~EbtxmI&?0krhze2I0@*kChYvR%r+i%YKP@iXGpBhiBQ!;II z(#KBvI3HZ5Qg<#iJ)Zt`(!ZVbZzcUpCXPNx|8~*83H0wM{nNg^&e2CU9;JUl`d2{z zO6ecxwwwO#ggzvX)N#(oHhe{jE9e?)y$~esYS?4j_xAC88S^{&v7LLd8%H|Pa|^(S z%HQ;xzFEY7zFV#{Lg;&pa_JR+T$8@2ey{RAXbk%q!&t_U!0$Z9kjEH`mn1L-JLjav zD;<9nU zl4uL}TfhHpHL+}7f3IY0`Fv&jQ9QSNu=0(`zGQ4b@R$VP0bSsq!FB}f1H4zf)i!+m zX8*$&fAu{4vpewtuPU8)t`K@yDqkqPMEs+TwOswxKC>7yjC@em8y7aAi`^APhKXhH zY#O{6KGQ*eTUZ0-e{Y+K-JSWOZ)Tk%j-2P@wzd73tcTl6=bXy{){@;?uA;xJBeJob zJT`RQZwu~M(H?7*Ez8^S2!6&RekXpau?k+4tcqvA9_ozMSABys)+*0fW7jZN z_@#-jamIMW+YfXLGe*X$y=sbU2rc{L^%A!6+#zV9 z!l47vg8p28>~!X>H-1;~(8bWD9hZY1I+C-L?y+;To4(;gPj9=WX0d;8&Izv#HQm#` z@ab(YYoyvs?YEsO4yKU!_&1RIQqNnB#6Y~DIFMz;PbA>me~#FNv3~qNis1lnKH@iA zJz9AZ{Cp!CtcpAPG>_j2#Igs7-zbL9sa|(a%0D`FVy`uOQoy;XTW6}OFKeJDwWBp% zzUUH@uX*2J2k2`Z@57$Hs?I9BioF$Eh!gvWmS&uI_E4kERb!AJt#AA(^{@em1c8NY5tHM}61L zcg4h#VVAlQf2XZi=G$>!>f<1KbDj&A;w{ft*u13%9NPAoMB|^HUM~Ag^1J`|o1DhF z1oF3KSno#rn17r>(eiW3o}izt*l046vs+KVE9l2C`qS5UU1Oi#gU43hx%!my4z&Fo zxwn-(vcO}&|MA3z6~tSV@Jw~KQ)h=$-o@2c)`%*`y?jP>^SC#kWt~eb^YN^WcI!f% zYsL8c+jZI$yd;0O(`E;49u9eNvUr1X&sOnFXFhe%W+!d7us^Mk`qJS8+p#mvQJfOv z2vSeoQTx1homn6})cx367^B@5;!OJbMj{Gd!T+d6(S!Cr-AxSH?ECV1buGR>51!%L zc#@3|{=fSEkV6OLAg(#IDDxb)+nPFZS)=REY6mXj8%^LO9yWrDd=r9A2e9b`Hsn3^ zoV)tOhNGRgK{K9p^VXu%0jfGP(5`!wLMP;W*q%7gV47cvrsw+nmN=1tPEW-|=M|#|`IK z;!6rvSMl4$Z#!cUZm#3Z#Ww#E^p_`Is@lgnlk{&n@tAIVgJ_s>$~GsM$$ll6xou|@ z*=?V+5+kSlDUtB`HEU)N`+1n(9l-4*a63htU9SSW9|FfF=88WlTkW}R4TYQdb}GN_ zjxiB?TXLjP_BP_u-FAeZleE(Td=CdWn*g}=1(Les$03`^DgL|p7Yt7V!#9Wr%}Gq^ z_H&M4H*mZuF)4D!KO(Xld3twZ(i)vP*bzuF<2g+`XK1H;C*yjCHlJram5JFM;L@~d zUEHSV%xzb&Q=2)yr0$Hrv}T>6J?-5%?H?Zb@%goD?jc6M6FIdL*mcv^Y1--(eUOv4 z3K`RGBXAKtsNQbsE%ybw{a>+Wou=M0_CTDW-gW2Kty#Y(g!cY&Y`}D51gS5w=d->$g9lD=dqdgWnnXe*FA!c`uu`NCh2zt^G!5z zhJX2vwxUZFr@;I9eR?7K^u5SY@Wi&eLbM4j-wiC0-MZ>S^amSj9cP8qa+d5SV1Z5X zjBJX1l7Z|qmfX6v=DPPEl!tns$6tTuDf;|V#yp=~7S2B7n12-Z24DCuA7P#P1>?eA zDH%?3!N>8Sk9)tRH1phS_@?TV&sKKtQ|uLPea41c|C&~Qbz>iNZ`L$u-d*!GkJ@$u zhvzR|(+*q2;KD;&nJewgl@6!8a$dRXd)X-D%`v`ecTJZ+Z#go8aMn&A+AV8OpW8O= z4&vRk+ey2f(w+Lx%nsSSn`iB`+ey1!wA&88+J{-Mux4a85kKEfE}`Yf1%K!^meb!> z@dWULyfucn&Fpc+mmXrzz34>zT>56L8XEopIU0KL!7pQlT6=wwZReM*QQVqxU$#W7 z-q(58_5kma`F=RGJxV;s%6{VrF~!%wOUPZYMQd=OEBs9izSbqF19UL_)oP z4&AnX5K?SdZ)-JiX~_1H!E^@LT@&|qIo6vkWW!^FWI-y%L@0gQ=^_qn4 zFg_Z^vyNb{pZcLq&xKbyGFtgHjeV?3r=W{d{-kWpL-9P(bN8#v-=X&#&w3{`FnAzc_H=P?k+E*=oZCkoEYs7-tljvJmJoG{M*F))6 zb`I;0WQAu-XHTm0Sqo1?;}4C~^MK*Sr{PEIDA&ZhGrW5^%gUCIQ_o-KDmZ)`4*#p~ z53v8Fg*hS}M?Sy+xYnNfuYhOaGCuBH-)SBxCME{{qeq~H`FUn|Yve#JH1_ zLb2uM++uX)&&8nOmLtHq51RR@eHK^|aW8s~d~4n*=3MVHm|OZLllc`rPF?Vz7&s!Y zg?)A#G4Bb)wYS5=6PZKs?%vj?h_%Z^UrDoizw%w`r7eS3XYSj!1n3YseGWKeU-P+f zqy5#gb4_IL$vpV3_LA%aXN~wEj_}QR;=W~Px)HlV=W+TyLH^zkH*^t4--V9ekfzvb z__Qy``Ju=~=vocL0w?;6?zg^_`K&sRqO1Lh&K_W#UBtFeV4Qm1PJI2d=)$o~<_mtK z_B8qoKb9UW{(tR%_HCM~b56^E!`r~9gfTzv%kH2r2kFb}^y46HzD{4-X2M??tKuh! z)tz;a{%4eyoNMKqDBs(5+rcuw9mACn4%0V{?^n=WNd^1-fJGm0d62nUivGEPHXbYr zLA$~3BH$yLQ0v{Pa$k1mG+@s9&`Rl!{kCsbfiLlmpfA+C>bJ>visq8pZ`B8XjWV|V zjJFTF&;k0V@t2G-I&6RbhXeTfGl2E$z&aCH9|X=v85?W(JK6*CY0)YD@?c7x&QxwK6cqPO8VG=^L^L2{r?nwmYm_*biB4l*##QegC9+XzJ^0jqoAj& zz|}RO-e2o1H|Ni_FRg|y;_XYy6_Q^@T$@;TXo%eF$-ne1IK zHn4o#Mz<6hP;r0S2hah&w0`^l$l^?k(X5A>qp9Jae)Q2M@lI@Y{I=uJ|HI8?<*iWtujepZa;GvH881Dg-KO-T$J)vTQWT*9Aykk9SuhYKX&}+#Lu5RABZ%fY! zcxZ^VSn!Hr+e=?RzGke-J`WCGM!o`uG1igjCaafS5y%^?kVUszh)0v~LY!qXC zl)d8+`Db;X?q_{~|M!(M-b&(Zbv6UJeq!IDzDI3%{H7keryp5{oFA><3H3&iUt;C7 z{V23l>F(K0Gbexp2{({)z`CEssE( zJZpo$2qrUu$rfP3SRac57rmcE|1SlLRKemt`VK#JVevj^6|~>S9;0|zY_wsqOt9eo z4-JdS(27e-t-!Mlc)GBQhnwxgNoEe<$AQjv--_Rd`Su|5O8SX-cmSG6geJy?w#a7I zz4R8onP8nIU(&3}6YN+&#4r&M4#!TU!5b$H^4U*6;@bh^ErdNpn|w1^jGy!YofmG!j~76FVOe+hG1#wqQYJD zcOrYXb}0rV#l(QjBzFpX@7LrYqpGZI5`s@k>4SUs@3qfxI!L^4&R*K#I%BWOa?U*U zMx3(PCaUFk?BaXn0Vrf`t{hMQE&9s$-9`tiUK7Ob)Q!Adj-4acgRKnxMLBcY?`FNa zi*@^Upqa^{@Q zsA(Dg#7nJ|ZR4Bvk>p-JnpD$)94?)(>t*(*Ux2RfrH!Ws=z0=#eez!R=Q(t}0C@?W z>2tBCY`Xq&C3MaGADXTe*IK*?J)3cR=W$Wl0B@5HXXf$kjFox(waz~Ij$VoJyoK`O zanip-;7!l8wyXh{dM+B1J;BQd=U3Q#u!3(?US&VSH#uq0$EWOPQ@?7}T#`<$B^!1? zr}OB8%8pgpQs&Mz&^IUUFg%G@n7fGUCAO*njAI z3v|^14!z&%j0VxgZO{b1k7h3n@`qb@>OOgixv$_p#M-0hOSspX|3Udjo4jkD~T%Zc6E#y_1=DE57tp z^NO3E?P;=-$`oHw6%0ML{Fb%Ntk-4x>F49@xBvTK-o>~L2C~1y5H)Z zjDDMsKCJ%ZT-%fnj_lZJOxc-B7n_7vtt1M@=qecJ<}-hXEt z?z3{_p(54>*;tGn_dd()i*bGLEzF}l=Fv24jpD(ov+cf6uUlSl)!s^9Z}MGk{ffCC zr1v|yI1LUf^LqabTWB7541uo$vIQCZfUAR6%|iH!+kfxaY8l(?KeF@1>U(!=lNnnT zV-pM{Gh7R--Dic^61-z;LGEakEC5XydT_=z+A}t-k#2c2Hspla#BR9t>kMl4X@!L+ z;cFjyg%wC+pFOda&E!W!=Dv;`97Xy0y}HhxmQP%VwMTqZI+5aiv>#7n@FT;NRNYeZ zIy#Bg8h!JH-+pAy<}+~+6BQEzt_5?(rZ_9XL~s`U+;ACMxL33g983!pW(~J#;E$=x z#i5r5HqQ#+} z#q)u-YoHf&=3R>}=BaJhgnB>n?B}>HxUBhX$=%)c?2*n%7~Q>q>vh*IYyM@)-4So> zK;HvF=6$d?IzQO^+vHG9d7qWtH;sGpFg~u|;VYl5}MLXVD) z4@xIpUu9+0wB1FWcL#ghsH19&ucnQ5q)V%fIkfR%pH*Ytkqv0z5`itA5n4t3>I|Fi;PhYRqj6_|BM0^`{=dcMn-Ct)cv8{Vo`v_cPAnj9s(?J<1-;6RLG0SuUF6=6W&o?( zsptM6u~h}=_EvQ#-}Pa`EKxbW7rojxD(y3~!kOl5zOyVzKPw=IPa&dr*B(04`wO!U2c`0U3ZR{*!g@{R?z!( zVw&S|mtP3fL#>Wxw!t#0|mVgu;I ze1f{fV+8+M%qhtj|F#4f!-G@xNBz*;8HC$p*0kaFTxfrEU@nlK!pIe$M4s6BFBi)b zFHzU;B2S$BwAR{z{F(6HcdUnQM@}z)Y2~YZo0eT^-RZlEGqL2~>l)Z^%KG1*SJ@K2 zuXMio5jhvI56$Mc_SO23GbsDmjx}qVC5OO!XC;ygxoL!NVF)|>-O3yMz*NPutS>3F zuX9SK_MS(c$Q=dW=ZqN1y(Q?A89wU{$vd-I1JcRIBz{%oFe#}rmGA`YZ=Z5Hq^3A1AA~pzHXN>h}EV4;!!`kM`Z$8`8y8O+a z)`w47-tCGT!cA z3FF;Od98Ir(anGKfUohF##i1D`0~Ox9)`9IY}1t1x%cs3jbl!`^SRN)&kp3@+G5UM z9$PaTeMs`lXXp2AD&d|nzEiR?Nicq=q&msqQFGuRxLD%AL+@sjA1mF)#XCinNxenB ziVd=z0tBjw|$O>kaB_;CcIx#Se++=!2Edcisy zV%_G9@U#2D;}l}mbI>n_GtWyN#^%cX(T9j4?D{)10SJDpS0}o(SKSJm}Q3z|?cj)H8KwPcZ3lPeF2riLE!~@sX;02s`E! z&Kp-hswrPT-!32I8s867ZnRx)A9;zUTx*wmhH@Em?Q*N;c*@bP>%;N(KTTyv*kxz_ zia7mZyKL8NQ+AR1ta7vevu_sX4Q8MC-{i;O{tLux2$oxa**9ydfp=2(7r8%KsB-fr zX*_9vO^jz7d9jIAs|ANmuUltt!hYPe&pL~2GHVL9FpYI4eRk`}0%zEoCt<4#_q}c{ zl3GbZqPW$mWC z7`i#p=3AI{6S>r^gvje76bl(({a?Ap&)R>2b-sdk=DhhTpM5s{9L}ac**EJO&N#Mq z_4W4~Kgz87>2|+^T-EPgC)b&NBTuRQ?9|tM3kCZELodN@(M$d`l?it5nrilE?G01r zjdjoVw+D0Vu}|Qt`BE_alg-;Lns=Ugd+_s4JzS~hoT=wB^ERqEHrwBuIdHw=%rysAm^mOjV@urmlIFY2 zi@zI7zIn5*RN6K*^WB(>zuSf!aM5?z_bzIeGXyXCjyZbKcbpgK{Z2fn{xi()*R6Mt z5JzX`3-OqodB(U_nQ`@9OaAKHIP>rNCC3kswca^gF{0)WXQY_%vK|huk9c3;JC1p` zo%WUQ;}H2i#3S^5ar)r#xbQAxE$fFj{xt*de}yM9=heqn)++;};KgHFmOHpk)3#z1 zXqj>K$*3aa_JLLB+n@xXAKZrSqhi&oL z!F6A*A1!hHxZJnfh8iEVr}4FWU^_OhPQg1b?wM%^`cc`$3j=uB7WbXX9*BEp#yu2G zVT>{Oqamf*u25@3r`u?fi6W4+v2`A-wiQW#||~OhM24Jg?h(l z>KkINw%~WT=({22YVlC*4guG^xbLPJzMb4nVsz$2a(8_S@`lBH;~FfTw|6>pngX3_ z%`SpZ3Qn?%8a&Y7!Ss?8j;B|BUxI!7QgBU(o4Yx*oq@h;>K|gR=UwzYc%$!^s{cS7 z-dr1{7Qx99mW^E@&9Q*%uU|8xI$!Ekf` zcfoK||MSG~kNck|hF|M{o*X{h|2#STa{u$h@V@@%iQ(Oz=l|q+zSBHU4DT`5q;Rcs zebITo$y}4e|KMC@kKf?D+hwjv;U~EscI4{a&>gfmN$YtLc2LpC8J>|>#yk&_-^0jT z-ylN4U;^taf#7MZJ!d&O_rw2>IL-`me0@jU;!=lcbo z=YQpSUgCNFWzX}!@jRcwy>v#`E-u}%{o4cfzi$&q=hYumOup`c{q=3+_sk@}=j-J6 ze4YHBRh&y+#ku6mzgyEhZt?b>x!*Y0vwY9_o`%J*b5`_U98V(8rgFlMC&%VA_787g zpIp0~TM1N-_y?i z_Qb_gIAc1odOP1OXYX|bd#_ip$NdoFlU^T#W~w-^PkY}J8TWSfx7YD}KXg*A9A>;=KR?q{nMI@-u<(`d9ZI2 zHjwotoaq|)@X6-D`IpH7U{$j>JyLmo_VG&MCNma?HU>EVEtB)#UN`5zS=HBR@8Qd# zUiq_3U-NBQR^OH^3WSTS+?VuC3E$YV;^6Pp-|_S}{yd=7z{uObZxX-0hdx^1P5mtT zd7|U5x7$-aS5uF2WozEJf5xAj@A#ANR-IS%?-`DLhZrerP+d8`?81Ajv(*Mq_TELU znYMjrz=vax&u?N1-{8#eEOI+EfVUgqXNp_9kv?ak$KS~B8R(PBi**biejoKKekh4L zPsXa6PY@4HOjQ5#EY@_zyScGx-JB;sgFHy8dz$Jd{zi4*xM1$2nfxxJZskpK>sG!e zQ#Y{5xtVhYox1hBkFxl!XPsa!`MAdCzsTa8p=0R7S>(USy09X$hIy35nKSC6Vg<6u zUvQHA1&!n{SWW(dtR+dc!^mF{CV#<=nr9_mE=0yV^@q*Ir%m;-^)20Ls#KK@3L;P>(_oo z)&IqmVg2>5@zj6QCF*y}75>cty!9+=l3lO%x~SeKQ$|$pPQ1eO`&m!DH&gF({CZzYa0^}gw;_ZI4Xo8KQ$Z=a`Lw_KB_UOhXR zlbby~$+|NO+g#2b_8VhY%DUdZU#j~F-2Vl34Bg+sJ^2>*Lq}O5`+hd}ZNxw=xW{_; zePSWsBA)RPbb+^gqt@KZv)imIs{4qi?1q+ZDgEZ86XehmU0snB%6`i;{!-|>BcHfM z;xfALX5SF;1gG&IoWXC<`Z%`IN3lc0pMS*IH80)ysrV257XQB7eEopfgbzzB#SOer z`n-6N;`4{&*Hj*@5yTygWc~%&gDYPpIib1#Uv2Q&`)K!m+J69ELA-);d=)`Q%A+Mb z=(=1!u~Of{eT*TM{7GARznr=2_D8;7_uiFfy}TrPAN?VQrG+>I`J24&#s_`Xhsk>b z->ccbkJwa)7wjl4nskKUo#^Yzq3ZGitsClB&esR%N6(g{+o}(0U+<>{6Z-2t@CoX5 z%N4SIy8K1Y+M%EJ6zfg{1M<)cHciT1i#}2{f;_hP3S&Y1$>i-EOV04gTacMKIaI{p|2O6EfrUHCiQ9`mi8tEoO+iHs@Kku=*4|HIM1<0 zb-w{j6Ca#CsR2ASfN$kL&O)yKU>Y=C2>ggAu}lopDCjxGIII{h+L8?Nl)1Na&#dKMuyN(u_&pyteQQ9*_3rujV^6$?&UZS7 zuj8Xp1O73^^%UaoSq)u{(|p_O&>H6l_~1=-6Ub2qoXWYfhiBGg`c)3B^OJeUzLa`& z+P3R}6F$`v)*t1AM&^3gJrh_tjHUxGWW+t3VgKqB%6H(KYCx~}6X>o#ZVDe}`oG$H z9SpAgu8&XoHl;7vbIkkB$UpgchWE^wX6I?Jzq4$a$*F_f9pdNj?qkn9a_rf);JClt zA?l67D^$-kVD83;e4hJU`mX&B*#<8<>%a=%@G*Efti8|4(8_S;-zfOdRm2rt6YBkc zhLER(pDcr)0ewV%^Rf7FGr^7WQCENqH;1);w~>chaPN;D=6nOrL^tQoY43--pFmgn za5di)H={W(-j|AhER1YD4!*a8d2pC{aDsV2ZpK+F$p5v1{9h}`|Fz=#v7T}(se2W? zxP+XN^#RT*=lev?aVcj`HwdV%p zV0}CLZX2J)Ii9>5jvW_Wu@Q*6zP#wS1^&3BGfPeko%6!SVp5OeN_xhd5Z~o1D)^x>J#4@bj z4CV`aJbFz&B4}#+s~0vAt8-^7u?;$Fs4X?rTfuc1ZT%zf_K~;Lyc_Mjp;A$#wLC;{Sy2{d}KE?ti3F)@Q{4wR386@z%PT}0{jDd%Px2LM>%ptI`de+Tj3x0$uEMjc!$d;bai=$ z%TIn+-cjf94&{^C9w3LF!#gAoczK8Fmrm#99qzM_;T`K8-m!it-cj!44Zjrc_%U%^ z7x50}z+c6|euo2hul~ClKi|jczZ!#hohx6v{Sa)uvPoEZL(zY$;LYQZfKU)3A0x2xaM+g)7VM7=}n?Y4eW?$J|Km-JuN>*~j@UafkiSG)DzOufI0 zUM>AM#iRdfZFKeDT-`hRZ%jI`qyH||eaO~-k%Hr3y;pnA70)ZZcQ`yu@w?jFud|Rwgs+dm!~cJs*Vu{v-_UuV z{$IbR&Wj$mwe=66^HR?^V>e=k00vy>#Apem|Wz#pt}RlS@VGZxynBoul)fMCW~f zkj|@fth5K|ituvcNR^L5`feTiZWr^a3q3bTuC5fL=cYtfq351K&%J_NGwtZO+B4LN zerxnxbllVExVN)*O5Q#DcanQ-9rxeNeUj04k8^*~8jKt$S@dtUZ#+&0N_BKwiKVVQ zngt*5%I{`>7<&)W&@cSR+dj$j$kno|Py5l}Jyp{p#AP}k7{QyWn5=!@=zziEIC-zeoFx=+fSN8lt}!*nhDF8C*6JH(!0M-l|D{4|=b#&Pu0l zCB~>2eVBbF5&!whHEp~nw=OxyQ^QNX;q)`#jgx)+r8MSBZCrn!ar*o3F1-?Gul-v2+>hC+umwn6^DzbX`n1C7Q(Ii0e4aVf&b$8kYv4DaPrCYdIW_}qNVELR-Aw+aPv`1)Q7Ai;fA&GY zQNmt7;-yLX&3;|}-+uYyb=+?_{gZv>gaaG#6YUYu*rk`oXeS{pl%2sp{V;J#37P!n zAN%PWd9-_gb`y|k()Dj8yphH;v&RV7>FTX-Kl-ogOXok=@?G@pAb$1-gh%x!fXAo* zo8SRHniG(VI8x1`U*uSi+*sEA^J^N)+!;k#a1>;LCCY~I4-?8yK z_CAXTv`?ZW&&28LzNjR@_F+~I_%Pe>advv}%btl3t;PPDbbM$woYF_%8y}uwf9{b`%_8f9HddZfF$n;h?ncO;QW7r#FO|03)cRrTM^nHI`!&Pzl zCg0fqVed`A>#EMX-+cy2N0yBZh5#|q=#il`ACnbAbtKssGnBNrAq{Oc9C?Z%#DQLt zU?SO)@gUOpU|JXXOJrlq#!3i^YlqMzm5huSln@UzX+lyCop`Lq``3d{-IEh)_IuL$AA5aznmvc^ z#8aWkA*(7gn)4y*tGYgV`Y+<`wruBm@a@{`mpt;FZs(Q{tN1U-9t;^{{!HU^%>Km92&pPmSU) zVDPtoJ%M{e(37T&uh0hiub->Kze^z!QCJtm$swtj7d@lz*bst(3fBbowo>+7)Ysg{t?Hm3lJO4ukeveuh^ zS;K)a^BM&|(fYAYVyCvR=Gb8T+Z^Jr9r(K0ri}fqx6C*0Wvyb8Z@qZ`HxJ&E#iJJ* z{XeD;WuK?uPZ+rYhr(Yu_30BJabDbh)=^xo7-Ypi zIzEL*M?MHFUC5PpuY?00-y!CaIal815a&&%ul&TP#S39@gHApq{hYc-N(|#D$jl;e3751ug^VvIzD?WyMon%+2V%#;96DCBH*;B!!p$Nms@7f$zlQ zanI7=jSPDnQ{Sr!kOl`e8>jy?W>Zlu(G3CmsniLxY~IydzQ$%i{ae{5AQAp z2K`q1;@!3Iu03Cp(zp-#_V(9%@5+}uM=G1q&AMvW9=Y>yN#zY+&%3{fc()YZo&CMj zc6%Oq`7w1S?jJ%7lJCd-SIviXp_2}x6Kl|k19oIT?w)Mx2!kufXO=z`y>@^v>A)X2^cpI5pMUDXyqB(?{Ad7wx%#>} zJCN%;!^0G>}XJT`V1&WZbV&o;ro|20-4rdLS>i8V^5WP|zdMyQKX~@6k za{7_Kw~_xBjr^yr7;xm@efOG?f7Tcwk1sm%uQH;O_V!DoPMv4`l584r{CN(&dXayf z`|F2ZUC3zwyVgoB%cpK2{^A(}$Zrw7{);*DY;dFB6_jAhph4H;j|TBe1&jQN7I5kEiu}Fj9Dh&xsRFynS;4iMmy*wCp3z@D zt8jRxMz)iFD4(T=zIo-j6a41_TL&}=(~kV-PV|pUr_-m9GhmAWTM@90jVtUk?xcTo zK57^GOmKSZ_TGmlTKlBaiXHm>A^Z{PN$Kr*)TwhRml*m58()o%Y+-FXv4eu>z^i8L z+mut4V#aQr^hJ9X`ry?#8q)t1aQF~36o1F~E&leicd_#4{1f5%DaZut58GM<1or4HYeBU8DwSqf~w zH@#&)q4)9xdVA~bf{(h89pf_}x;*OglJY*|o+t2=!L4;e^idbLEj}B!;3*NWl6Beq zJo>2n9B|7zgyB&-fL9M4uD;m8zaLzGewQ0xv~B;x>~-44e%)=^%-1eka&n24T04Tx z>p-qGXPJhdC>wb<_!eJB&~?v&Q=3-)#u9!f`iJsmr62EFa;A?cUv@w946<)S)waHT ziSg?@lRq@-4;7+Q<;R67qw!_YWOx9*Qalxzfj_4sGgXqSTj(=-As621fQKY&BO|GM zMn=;1blyT8;v0CmV}~!=8PmQ)cCJ=8`MEj=V%iJuY^frSET7QGEAnjQ6~6CWz;`_F zNqq16U<5oEM-D38`_So}uhqVxl4Vf zkJ{Izn7+s8`*i6z^_Bh`9fE%&KkoInfeH8wOu#X!Lx6V_CSV?g37KgLl>9_6{W~!I z0+=k%KiA)QDZ!qbc7aFs#1GwLiiNbyG`?vH_T*vgiO$#ULUzIBe&ldpm~wueiEp%j zgnUlf($#v$c+1Mpiw4Yldu->woIX2aTch6`JL?x+lRV#TrEctlXZzsU_3V?s&PuE0 zOz^={{50|sqak?KjqBO=+~4?p;^@Mg=m?Ws7z0MZtoKJeJS_Y?1%BMI&A&Q$a`Ce; zAztLC|FW~{E8|bIE^7y|(8#B*h*l8Wx8)?O@x?wz?wr24KE$I=TRJ1pLk?&EZF@aP z`bCnDvGpLDYu8+UcP0DJRl_?UBG=Ecf-#a2jG_n;Z1n92fpluPy3)j2;09t&D!th zOa)*(q_YY__+W3}WAA4Y!glMN0_}Gnn$G=X>;&g71hXydIDU9#a8IaI=dB)lk#WYp z4fJ!)jLSK*G~J93(yOfK9NzUKPXow}cv0W;Q>4zWmUVA97hd;H4 z^lHWnvi$+tV7%H?&Um$x`Bm|X@>65v;jbYk}{2(Mgi`L|jYm zrT1Q;efeAuzcaF5xYoWk6*_N#KFBufD!*zIxSt+i-%0kbabcA|aS(eJ#0TnucDt-@ z{OF+Zi$zV`i*975+n~^q^BDtIt(9vdS7VHx3~i(=#nRe5FUG+<22~1EdSR z@}~FRXKs1-nJfSHe9BlnLUt(z|GRmp_$Zq!>k4CUvN#V@@-LmEod0a%(^hFH`fYcQ zemBMduPdJA_D}tK#O+rGdTKJ?V;8*hX)Zi-67-R)uOE5G)jz)_Gg8M)mB#l<1=C+ zk(U1zjom{3Zs)JqAMFC4@*lgvDf_1luL)ZF@fiP@Ju|*IesTAG;0ymTfZu3%1Xy1Gu^EpM3lBewO}|F4sg8}* zq4Ai(%WUd`ce7{XADF#7uZVUXcIU;ejDrPTYw8eP{+>GYeGc`=w?!ZPswkWu)!wX~ z*F!7#MZEL7JMhN>jAfvc>~+WVs7S8`P*iW=1^p$5IR!^JgHAT^zp+! z>#>>C-$wl!m-h0$mo{8@zXQDVix|{3QLcuMmeA)Vz#F5i;9Us3y3YgN0PqUF@!;uz zzd9q~;f|Bg)ag%`Q$7*)IDN;w)yU1ZzFWDR7!mzddG*=o$X2_*$$V|;!d}^A^v0JX zjz0A04ObubVu#O3@MqE!^hY?&%zGVj=DiLj&L6h9=ae{pjM^-u&GF_B6YHM`T=UUq zqwRQTC%SzrN&R!qQvZ91kHy*C_d1{RD`e8(-4h>61CI9_A4}~2Xus`wMF+o{r?cm0 zKbxfg1LbGpH;B*d`%`q^)$kf)(ipgNZI$ert*>qW(Oy$Jk@tvr96PqhizlTW4nrKJxghcRqO40{$sCmQ>JtH(9@QH56 zMz-4iRwwyH%HPVuk8^$Z9nhr%d1*kVTzkBSIJ-Ar$?P``Ud_1|;8^3?&T3!2{B`+l zslt7AVSa~f9`+~BN5Ve+`7k`k^J2#a{UKsJu{_3^)F+>17k21xzXA;C2DKB*gZAk4 zd5nwI*R|x-%nw@EiVuGLX=4+bWLu=qRbD=%fsb!Sm}8fYWGxE*kX@JRz&3C4`n_&U z1bu7bYZVExkP3WA#aSl1GFWVnT|3TG{Kw4mH|1C#kN(TTF?w{4;~##^p%M1x4F7Pm z$3GmUpV5b7`CFaR-Ox;QeC-5w03K_ApWSsYS{I->kxukHZ;a}ZL+E^jt>18{|X<#XWK(V@9m;DK9^#X-N>TiS+&qx&x}98 zbNLf`HyvMAacl8~Vtw)}y1<*_PeH~P@-gs*4_Vt%Ogs07W__l~Ks~g+ly!I_`5#RHU#dWTW z?1tut6j$rIgDd*3oS2Np8&8mPSyY3+I*WJ6h|!}vH${pL)c z8Sw8+cx3l^+NUOyoRYM`?c}-kK+j%yMZBZf$Qta6a-~*)lMc=VlCF`z;MUiJP171_ zw;shl%{%hQ)Ru2DaYNc@Fm0rm{8Ys)6YIZ&`l~p5S~3*Z2KBh#3*K44J=!+&O>VpY z=C;dNMC}&hE4bg4Ft>d`v`*;LwX_*dOVTEFxZj?wP3blR3%XWqZYRDw1Kv`5t7&h# z;V1cR^78_X9~NJwBx#TD&(@Ci5;k&xEO>RFWJ~d2;mrHoz?40~z$D!UW1X)u*W(wp zn6czJwry~9jVa%fV#*8dq-`bgwezj!hTWVH&GlDd@4T?Y@%>cnCcJ-n0`DvSQcNFh z?jPg-#Pe;G<2;?edh7xEc~o|D!GZF>ZJiZteCsKARcnxexfFVI!=r1tHnaaz4nD{l z=u*btcPIyb>~hGgDaTdiO0hS>gZwDPBiHcPNO{FG^*otut8c$48I$W*oIbhoJOFOJ z`xz-e?pMsub|^NgebJp9UmJh6-uThK#ji8?rBnYPb(`F`h`r9E)03k<-gt3OlJ=X- zUI&WJx_ge;?LM8*ZnM+w8wu^^c+MyKuY~@-pY}E;87%YYeGpZvO|8;MsjS>_2BO)zfE~0{rG1C_!GzRD>|_$ z9oX|6^kcTG%d2hv8P(;}s3(hh(C0nWEq|d8-@}~~I!?XPg?pX88ahv_;+rtv5Idv| z`3!yN=YDka0OO(KjCG_h4Ik{Vbz0%?#?}J;n0KPJ{<-wnW9Si|Z!CT0!$w%xgYo(x z*lHcvM)jG~=GpX_P5=1zJ_rAg;ujovW9^0y5^*=##?1;ZZYG0^gua(~WFXo6rPlBL zKK;MtkJW$dXWYEU>UFf z?N0mehyRzG_|4e-ob2o8bRHReubdK%p)_BoywsFYztBDxaGR@3S`&1MjUEwQ z8WQj#-%zmx&CeO#5VLhdnJ3r3!=oF#`oWgh1bC5G7gqU`(i6(zFFhN4z*U(5*C@SQ zc;17&eo+79*6+vLkiT(Y`D6MgS0=PskjO6)n+|8wmoFu;muK@&&P!-B*J<-T`zHzW zE#gmieW+q07WDqeZ;7iB%h^GUB605a+GFu^xBo>Rz1McX^EbB4j;;^Qv)8QpJhb<& zM|#QFLTu@J?*bk%R=|A-%B$brejb>{pOTWG0<5kM zSKi1kUAP^-INN&fJxTcQz4~`k9T?ufeft3T8OP5Ul|je8x%hegu6W!3N5juJa7j+v z9T@&w@H37dG9G?*{&(SL9Jqv^>JNaQ0i*B7{EZgko>tZD=>6m>b-)9if&2E#_Au|l z^X%xq-a@X1o?#aXnbRp`K1cI;nge7`=;~DRg|Y+Ia4|YV^FLwcjHr`&q>r)2+&K%u z(HDaGlfW$ba`pGZg#7{KpGmh#hbbnam>KcjriSnh(Os-zNF#RqqHH8Ot6TF(Dq~`p zT03aZKUXFCNsN)*cc1=G&K5@3MfgVFccAmMrdRU}CXbi>h5CpQ>MZ7tgn9<7(8f7$ ztU5X;$BMqZZ23v;M_u9znz@-i=4K|4cPigfeb7Edsxz|%yXfdulhZ`oilfg+aoVSx z^yKKc)zMYLRYnpX+e=xm|2Su2^7y-A?^-Xl0{x^toAkT$Ds(5~XT|6>A8oJY(przK z#ys#zF3qO0!6L19-c)|VvtOhWM@<1|>#WrFW&8~U(%N4-wPE8zaA4b*bYjRM@Rrsn zKUz7T%sbY0Uk84R89U$;^cOJhx{9&%^;{`u?*SVE-#%aJK?Y|s~i)q`>1C3MR{VXSLGM=8OW6oST z0|+s~X!Q5E?U$by@L(hE;&9`P3DjZ{91d6zJJ^fK{g zWPbl^YaOI55gr5I=eS(LLsT+KWcxyS*jQ z?ha^pJN>@}`Yph3xeNKb6Mq`rwE+DFzitX0xFMQO&hpLZx66FQ@u8j0D=x>Eg-`c| zC^ta4IfZ`~y_9kpl)II3Wj#Ho6{|(;k zK+pEmU+LH`@vDJ7wbAkw{KWXi$V?V+-wpgdj7c&DH~g-7`VezqvM=4hxmI;hzvj<7 zfL*$)lzqLWza&4pYToP(FYm`3-tV(@r0R#?Y7cwD6Te!P!ZxR-w2ccQ!R0{100 zEkB1$K8f8jxI!o1&F@=%85ZJ-&{Wd|pf)uD>xSVLT;X(Of%p8@dhJ zltAlSp|$qN7LRBgl#0L6W@IY}pD~Z^*f-7BLHG8x=u@NX96!g<$D<#`>)!W?{gED= zlB5sb`F4F{+gW?hoR&aB{`$MTcl*EG_*)+y)lVkh7Fp?lN42)@xwRu(o?~r&5&oO@ zl`dp$mRU)RDfi#_0NB_X&~sn{v`f(@j2f9LW-Hhu-B|RbbZ)O4eS-< z^=X{vy1=d(w9sn6FX)m0u;DCY_v(%#7*M1V1=n>?HwC!1GSZiSOhyb|R;1`MnPLuc2%Tex`iP2Z=c?zAF4II`ZmO=+}Jq&TX((J(}W6sa3sNzxE{c7E`Czn&{iv^l>fiYM#>M3!w|;Pk4XJ zKe-27ibv1oowvWG&U1}tlJ&cSeiMs5;;k>BHF)s1JC0HgluNgb4qv-|h+ThH<9PMo zO8qX}8XKscRwqX}0{)OS_95_`8{qiZf;Ev3jUIM=aQdq64l2J!yhGi4xcBn7S0CMc z=a^o0_0etg#pUOhkWcUa6KukpK(Ml83ShoPVk}QHFd%Uxfb->s%Y{(RMtl;R% zI2(=3%}#)0y!o6B=Lnv$^}Wg&?771lj^X=v7NheR2Xrk!KHw4g^!?~Djl&1{I|o`9 zq%g;g9=W;Dil*W}k+*F7)9A$R+tFQj=o|S*6Zi(Y6fo~Ze(25E^d(UB(ZAw4vK*oXF2mR?fqBct($rx-Vas}A&T z4tiGOz)rynuCjc-!7k)69luZQlq{cfa^bllBMYBF7LE^(TzzRuc=$nbC@w`F+M$$6hYYxt``zLp|i-NY6eSl_$zRL#bNR?yVz>XFY-r}W5T zzLg%SK*ns}%C>P?r>s4a$p~;%Ba@o1_-nq&;~V4+8!BJd!?&%D-$4IOyvFcls4+MW zPrAOR^ze>Pp2?eTPVp-Gj9MQ3=kn`@g#8HRBe*>4#@&@4tlWcx@Z!tprJD;2qPH-1o_KEf*?WB9;g8bZ z0Clccegtt1!Sa_sd<o0Qm1kKC~7@b(>fKZB3tq9GLMH@j+mb&C;3)jT5!T*{mC; zjNX+dyep&~jg3nfBet9zS){UO;N9Uz=ys^grrV)%#&}*H_ul`5Z_>tBki%CPf8D$M zvnM~xxkw|dgU*IW9zlOhHsh}e@OFCR+wAu_s!s$nxRyL>j3ilCpUt^q^n?AuE_@lh z2|xYlhiCYms1swX*YU=-sy#d~%J29O-u=5IkJI_9mwyk8?p&4p4vit){K#+au+LhN z4az}BbOCP>*Ao1^PWG9R?;$zXxWL9o{G62d^m`uol=C|(jNG{J%`x^PVLk;qxpPhM z#Xfu7#rl49qG_wzZtJW2#^zo}V_UTQHBY;;S@LttJeWJ5MR0NjCqk`Qc7v z&xR%8?51KHmb;U|;>hb46W}p1NEfDvJ@+}@{BSaOHoXUUZb}9ZxcOuPEb70(jRQxa z2RF)JaB1e^=Ouj2#QfB=&DYQLz#*Qv)@lE}&etzA{4q9PV&=ya@<|jIcX`CbUC>q1 z#mV^6^=)KF1kZU1_;BOH@-GZtVm5tNdvZy<>xFE5#KULkgRYW}RqlxETb~Uc!E$0z zygUn*QTjM={GsXdZ|tF*<&19x!z&I9|8JpBV-lRm#{Zw7&$=YAykGiohDe;hEPmdU zihjXn$`5HFUZH)+#5+OmH707oc3Le3(NyN&YKS{)?}_zR+QtxeCxqR3Y1y)qFIu76 z0$<8t74libcsw1uAOYUC&~FurtR0#@0xOTi6&oZsCpnj>anvDPHl^cTaD* z+QTPaU0alZlQvgw^6b2q!r`&Jkr$Ha6I=hs=VSj0{g#d2ARDaMld;8&L**YB87j7Q z?&M@JIXd@}1h^ziF&8#R*L8YiTKPJzuG7`kcgsz@Wz0T3mu~Ho|G+qOkrlA^6;EmT zc~N$2U68-UmM_{H_~Ks9oNk{v{rNh{)hwP@TE3y8KdrB;;g2}U(DSTTGRzE$t^?PU*zw*IgxrB>f|@w=g_ zh;y1aLzuc_d@H!bAG&6*POCjw6|61d`=aTsb%ns{g=znt@0i$OdF#C4BJ6b$JZt>k z7v@1Tf8(6%?bu@Z_Y&SW+wXID@9H2gZ++vAczwv;$fGn0WtV1#8N+<P}OqSM*@ z?xr8+cSg}^E4wU{vB@Hf=cQc5J1v2dpZHozGgYR%Sl^UqKJR{0%QvC2%+l1VN7ZkY z_gPE6o_q5Cqki^;37?#NIOkB|qdD2P?9H~8eqHeDPj$L=d-+H3l@cd=FYtwd;oV@% z4wYt>Ik3s^%4S^X!pHZu{!^0=3!ai6PWkpN_|`*+F7$=}volYjQ)!{+&dqSL~IHRZFJ+4Q^c)a0mj$OlfAeqDG{9obglqZYq) zuRm}9UHMst4yT2=yYUuS!Uivk<5xvBy(-3!q9t3r0BHXFTJzq^sb8*WSRRF++T8| zeSe?leyLq{*_)a(e{+PZ}J=cIgAMi>h4hVPP%>rNE`hEOH_D(-aS?06gId;$uU0yuF=GTiNWN(l!!u2iQ9~Qgcjh^=>V2+f5oD+oc&D?cv3^&e`?9kuGIz43vJ{wY9axe1 zJg{3lKZs5+ec}3sLyKf}cLd`5627Q3Gbdb@*$Av&7y|H>+TVtq3{F{SuZ8a%d6aTk>KT*=?6*N`C~Q zr!}`CbIt7X%&w`{Y4R0M3%CFI4*Yy!_;e2XvlCc00*m;&YiiiMR~xRLBks*}>$K>g zZ-*zbEtQ^d*@mF8*xh z({u05Jdt0K`KMeh|B|o!e;9r=<<$L08!TUKGjlJ<$D)Qq?CtkYeMfx0rC$#pO2I9# zZGH!MGj<;P>fVoJFlLeMf9J#zp4s;{{>Y_Y`gLE+;!NS&!XCp5n zdsJV;yWV&exs1$<&x&5&e0mFUWwRFsw&Lm+nXk_dm^rGZvZV{mSvmNeO{GiI;&K5* z=Zm59cp+(H!JR*%qIV^9ta`P>rAzp@>=#5%w(RkV#DyQB8%C*X0AWos&C*@wCSK8+> z*Jd%-rd+LPxwi&y~kS9aks8)lY99lZ)?qaOK6< zK*D%U`mpkpZzc0M^vxR5{yg}t`I14eoL8R8TKS~Pp!03`5}L+yFm8ErE}n+ZCs03EuVddgP)<>s8%S2^EPm-l;bebg~p-$eBxUT>=Y z|4>c{efx>IhWEZ|uJ-);VeYdM^;90~qv7ArduU?j*Ym2L{hCK#+3}~bd@=257^}9P z6~37_Ta**B*q6mODBYs{)(&DXYpSxN9R=)>RbBP$ba)!ysCEhcX@efA%cjqjZjV5Z z9niz3p&jq*0mmtZ7Ab}nDbQjuv8?6T};~o}%WGc}HvTQ5MjS`#hbquY_QF$ z$f5D)kwex27`egMKU9Mpx_H-mB%O!lKGzta2R>~^7A@kFi;=xuM)pz%GZ^o=;{nN< zTSxlx$lQQWV}i870m`_t7QlYF#e}x&GbojaR9hdu%o+zJ~PJO zjqW*CVXf@hw6?CE_SR6Y0hwOHc@}MteLUJmS=Q#QXdy>KI?d=W$|*M?kGX>Ya|flC zJ$E4a938uij+?c|oWJo3L#JTlWn9~U=@{!21=BI~<8$bxv}H@?_5)MsBjKY1z?8Mv zS~)AauI?E4DF^00U|P>!Mxn(wL_^D$%+2zpjl!fn#)J4|^;5tl_^SEShfQBZzk0sU z`R&M}D$&cbhWx~g0`H6*DcV+D7Y6P#;nn&-?7x9GwQ+=8jYPhB;?^;pvLczizaXB! zlvk8hn)*exKZ!o)5$CCdzsi75JQD)1mGF{$E&UF(Oe78z-Yb0DXSUk=M9U^6>R9=D z*quS;#?3@`_hEBJR{2(j8UKxZ*IIdiI>}+aE_GAkaOz_|^5DaJ<*>WwguB_pKq}a4~dS@S~;IS=(Rl<*y;mkG5A-xD)v6!jL(&C#52j^E#_W(toj>o za-RDts*cQ0;B(i0_OTY6u_m!DayamB7f`R|{Bz?S12e?$WBZ9H4_o8*r|?(EdCnh! zCc?d|^Th8X4n6E=@I24lG0XzvkEEIPz#7MF$YuToey&8(Ez`!|z>zRq)!Xy*(ML_G&zF9(9{_m7EhkkG)_UCRQDhu9dD%_FQt7?77xV zD`)#Se}nwJ&fgpS{TF{H__O0Twr;lj;x}tPwANgHozd6MJQFz+^npIYcj?RxPGLMW z6qwRj@MWLT3;zY&zXjI?|7!E|9`@%P;Mp7GWs2@opR*3%pNTBYgipaq!E<&w0y_Jj zBB!FCvdX2a=Zqwc(UwuRHalah@Bb)DS0|U9d2F94Sq4*x_13r?>Tp4-KEa6g}x2MrqTu{!6&+IVE@+%;NhFd_|4GtW@uXPo3P>lvfg9xl`?o6 zx+)#}Jz?46xg}58eeQ!_dd!+_Th>eV+x*tUbJ=IXCO#{`=1A6GLT{zM1$kN zRDxWK$4YvzMO-JK@9N13lD^v*3X)$mp|;!6cO?&v$%pm`*)ZL|Rr4G_;9fc7QFbN#_Kmfw&^+IxqftR$$tcg z;QvJ>vs1qP4R}bpk9lYlkI4s@x#TaqHe7P&{XK)6I_amj?C0xdQLe~JYu7%)lEqNY z3w7x&x8^sK3)M#M)63Z6VqaimTlNceBRos{XnmcYEBA`;HWpEb$|~oJSY~6fmD(tI zE9Fcu`7>VLok1IF_XWP0O}qN8hwrqu=?UZK+GogXKPK&gMtr#a2N~95+SgY(ZH#$q zwSI>A>EQr4sI;u1uRu@jKhJ#Friv3cnq2R@!AB%eb>sx~nCF8x4p)C6a#DLt#+Zv! z+pvd{{)KMV|RyD{WJG*UcxI(U-5v@82yT|GED2#&DVhwHIPZ~SItQH=iP zLc=2Jl|8-Rspog<6Syhnoz5Msaq{t$i%+>lo!9F?1KcBp}M0x1fzWa{tzk#x@z2B1%e|(S_Mr8B+;UIZ>f~A_aV zIn2)=z;16nmo>rH+4(#>>67_gFqTi3ZRXzN$3$-3@*#?l(Xo2VPGwVX2e>kF;fz&s zKZP#_T;enLTXMYH@!JQ}@iFzR47hTEE1Nn0#P5^AU_Xm*#~s6Y+ZKJgl}qHT&+6n} z&yI2R(xoecPG@Yp+C!VJNI2gc&q_+4>SXlEbLf*ZMjvdIq0ej&eI_r^c~6Ihhf(@u zED(LZpA*W>EX1cW^mwND2jT3T%u)I@7ylqTJJ;kpW{0!ER}%VAcN~38y=C9errt#S z=FWybz%@!AzCD9Jd=GuTpMBAp-zS5?ewIj|S!cj1TCDQW=Zg+~9vFu{7oUYb%Z)vW zpI40K*IUpXHCBesWgVJ@E?D48tCf!&AxA|zsFj>o=Eb6SPk#x&$~Uxu zGIK^o_AR%I`<l3@7J<)NaZGkvLzkyAxSH{e$(-o70O*O;` z>w)Qqv{CO%9gL+^*KM#inDsj<*Nxs?GQGX7k}|Skz&ms!zMyp7R}?GN`pIH{ejjs{ zgRJw~k<(tMd~(JRX1$j5mwfup>#bqUP1!PIuVef1G-L?-+j$-S%k|;m`_OAKbg?}a z47B&~teLS{2puLrL31bF*zB;`8}Ekb)89G0Mf}XX;qYqW4LP&fJBBrc@`rq!X>u3q z49i$&sPhy`piP<13`3Wu0MEzC`Ov-A9qL_xb&cO6k3LdcWUduoWv?O5Ls1`5r+C1bbxsP8y+$$#j}w)rIlZS-90?b_Jms?B5L$ct#Ku*{mf&sTG< zXoX_bte4g}^#Ht4X3fgXZppQCw5siSJjH-)eP;hYd`^Pi!rr_4srIy5`ModLdhB-A z=est>%thi~7(3RAfBHf=|0>x#&v!N>KINZruW1nOVSI=v{;fFtLU0cKc`+YoX##eqZS3bYSW*QG` z-vI4d!`eT-E2qyz^m!`t;#OGsUD~q++`m?6=lWDnHfQU~_El6b%2(W?<0|nTG@R_q zk7fAsk|da@t=-3bqj-2^JUpOd&cMSl8xPZj2kztWFpmBa9wvM65Ql@G7Mgw0TzXo_d?N3= z^p{;1{ndB3{5=W%K6jjchlqa|J7Clt>gzhxqi(r-)9ZxtT>w*KApTh-;!*G}8}4!^5zgFZ&Lji9&R zwv28&g)A7|22G6K8VTqBI^^iKx5Kt>yNbM5-jC`w${eSR(QQ|e53PJxqua=xeJ!~T z8}Ir1)or&yk3`)zJ3+Ti{@uIlw(ndVuiHK?yPrh2&31L0$G_OZ+SH`}MKYOoblXhv z_509mACBW|<~ZJ6w>9O!3kTt&_oLhHPr$=P%>9)L4-L}cHZixPs2M=Y;HI4dN z_&ttpJ0w5L=y*rBbu)jTpxX{PzSSX*Z*@q%m2}&GP~MeqaIk4qzn`Jom^*{_ymEau z-F9t4zkd#2C8K{oV@{*nux*ZRGkyQiczqwUZ)1ITbX$Y^9r9J8MdvgOv>!CDH%Jk+lIhSBHLiBEHFwPVX=)20|(F5l1| zTYUE-`sf*3{Eg_IWNh)Gp*wq5Ft+G`@9}F4-T#@jqps}7@%-4>VwCRk{{`PH=&1W# zd=;Ims6LiEbXQ+>z9QoZ_sr(2r%~S6KjYUxcYC}*`#JX6T8_W&#u>amdw0gjmbbxw zWR})Q%HIt%mQR%*JUjaMp0PFCCr_G{-HWlO35W(2bfZvP{(wQ@SD-etuA9Sq4H)P#@3?Dtn zIH|jsJ!OJ{+J;l}YoB19NEhQt@wMp98h}j}IgJloBff|J@JppkF@ zvoME!^hNXhT7Tin_d*j79NYh>3%Csqz^~*b<5 z7$e<6pJnsZ2V%WF{MMNLMdpdLA5s_3dysj_zUDP0Q)TB{k3GRP2ES`er~aAo`z04f zd+D3nd;DU&TcjD@-poDq?=^xB-chiZ~Ji6hdc+6 zT_cx_JA~(VCdBJoBYRV!Df&e;`m#eKcaByu`fhZ}KYPZ`u6$@d;}1cvzfYjo#qX0| zznN-1mP=eWfnH5HO>?4ONGi7W_yf@E(RZQOdGC{6d#;oI zjNkuw`y^}MeBx-$>#dD!gl=r_ zBUh@A_+yHdR;wH*$(82cT{!X4_QwKWGIRfx#J|=9Up?_%*@XpHZ9X#Vs(*QN)>TLO zi@ll7x0*-AA0(%s>xklQJO9Mmm&W{VU+AFb!J^W+l2v>D%$|z~5a)^j59fmMd~E(+ zc~%kHxk;s+{nxvZ}-V{`*?H>Gd-czH{}* zKYjc37JV!IVf+Bv_7P*uTIJg#`(EbQ_j2sJ^nt5m==;@q@Tlw}Znq;9mIzgX9zZ20d~20qgKrm~U^MVXaJu_A8iwzgzR~ zYkezY>>pV_KwcczyZ_dAxE9_qWqlbdp2N2tneesCmk!GqrJ5E_T%b=e*6deKCvH)ow@zE=mYd)_xtR})(_ARe?mWY zCFw^$xfyOhdYh8-@7d___sRO<@OiTF%kuI1kswoEdK|hSp0Dmt(vS6=o#6Ik-v{VN z@%!w@b?>nsdzdFZKFu1w%x~vs9A_S&1KD)P9&tPf-@DH@C-nKcBz@L=p4;aK;`@9k z$I(3=+BM0aLDyU(UY0DSp4n%ujn`PD$wRv)cijCu@+e$cI&cAFdpE9{K)V-1BU@hJ zuZ_Rw`8&v8!Hw_4jV)bYM}E${BHI`B`Z?EmVrZlOXrAol*mzTOM81z$``mG-$tNHu z&4-?wj%{FVOnU>is=?T*lv+18tezaTPIOqZG6C=+KU4O$m%Zs?#pK*HYn_(8r{uY^ z1?adge9kh)wXNVGV#c+B!9w)$JbPT*w3ygzGh}i_{**F2DWud}x5*ZQ^}A--3>LpKK&!$EFChDb{|e9vnQFfP*Xt2Z{5k%ZPc1 zALrY1q^6A>wDC6c;L-)fjJMgtdzkfH@UzCA@q4(tIJ|uv96G*Hlj9pTd3+<6f3E%o zZMnWtJWkxXGMoO5Y3vOpd5n*9G%3ze!XC&^@P7MQ%6yVCS;Pr6MjnkzFdmbRn#TDe zZhT+y5RJuto)8b4WAek}^JRVP@l|lEHS{vHQvR4^ETBAbbdYp|^2M2RN~(*r{#AQr zdu`z)(G5M)iA;$u-TYt4`&!7MKzkz>FRq5fm! z|7y&WOJ3~g*to`LuLJmJpPd^_yUJrcM!x-azNzQiPOjawuQjptJQwWKi4AJ+SLrzM zvU}D04)U$_%yRyX`YYtTc=P8+D*by#kX`jpeQ9WcKZ%7tzbtpIoog6ugzxw5c>M7r z)sH{>X7%Gg<8S!Qid6@;)|@Inx|W>T8NrnHl2U6Z58nL~>d0bk2Hzf5eZggUa|7hC z3MT>Qd9Lr|Gb&F!Y?0n|(_Rg7P{}%vYU&`sdcN{r{j4t#e+Z}gR?kZCFW#W<$lzg> z2al}1jm#B}I{20x!2MZ#=aoIh;C-w+m<8Xu_tVWiYlzYw+MDCK&*VOlCl9a|+tst* za(L3k=W_O;^2VEmNAZK*_sHD&A%AU6Fr`-gUxS>_n|`3KfcS&*Gk3C=w$`3h;`?l* zPwG#EK9tj6m06QPU+A~(GuZL!AUUu#R>0(OYR!uLxSH$(X75GiYAN<99@Y2Cx#@$R zY0GBJT>uvmb0ds)0I?Acmy>e+&=;!Z6 zYd3!H&|2q%$Uhd{_jvAinR}Zbc5pwB{;Z)t!i#7lx>S(cwhNeCUU{5yi9GRf4^KQq zS(hI+VCTI2VCIRkly?{y%!OC-LcVrAtC+T6I0gJyPYkpRXUfSa$49Cl_8!51kuCCr zZ}M`Z$M{w>(mJ-6;S;T2ThE`?rCH?6X|1{Bt=LMb-RbavXl3(5TI0^Z$!Aho@6!g& z+u#-YdRVkjK3s#9TDuF`ei+$)7}4y|# zT6$aC&PhpD%XdSAd63WkPY+GyRm*f6(l zosH|QTl-+CZk_dPbP4)h`ap2i0+(p)lYXGjx>ue!In}klxBl?()I}b?xxjOO9`~B7 z^waKz=D9tWHHG^`o_YGu#`MVj^u^W9y1M+c5ZUqW=OUf=1bEiIE{r9Heuy0_M1LBc z&}!=h?cLou`OAA>!(Zvb2Q2)R742YbQt&J5+sdizB(L~Y-f1nKTXqp~N2lyzr|i(> zl>Kv_i?*)6DVx071Gmnv@xrS;AH4jY6Fm8hU{;>3>|X&iS_9t%mX*z|eAQZ+1N@9} zO@7~Yc-nvP*8JxrGk3kcDT_ zUssj&>dOygH#~F~KRr$!joK34{q#d~Se4lEa`1)RMg7Zi=JGx&7{t3dsqiEF!gRv3 zdZzVq<~{ma`*g{7R=w5Ko5&+GJ-FWO;9B@r9R;7ThLoT0+8D*DEn?EL^A({$Z3;FN z89J{COl7~~G}blr z6N3o(Qriu06t1p4hQ8}X-|b%2wzWT-y@FSzZp}KKX3iga5*?^BP8QP72;fWKWzmuqiWW}RFL<) z_|*S(RQb`($W0IU=KA@Z;Oe@_l$tu^>u~+aEZN$)IwI_aSH>Q+mDKb2lFIO&ZL>#(z`(AM57lF%MR=?c-7Du&4Em zaW%OJn`)j+Giydie}6YE%D&_1v6R7b4G(uNRXH~{WalD$#W$t#5dPOn zbm$1{9Zw)z$GCph{||MK_`Xt?=0B&lU?bx%*3lMp*<;6oa@Gj~TTfszXPIdI=0w)+ zYCX8tKczQDQ>@2qn`F0f7c{tsJnq{rVr^tzN`Ch@tl<++`|>-WSy#O^JTh_)DG-zQ zjI6UvJhf)h$!D}Cb{n!*kjWU4whBIEuT2zJ4v4&s23*3X(xBtZ#t*o^6 zjI^emB;LG;dAy;^xpwos`vJ-@rs{r>Yo#@O5FYBXg6$nU0<2q`2u>z7{(GwRSob&V zSaHV|=pcI46EDh}pWh8VdhcU>T1lApSSR{)VZLC<7EBB6cHW3k_j_#X@6u>1LR&Yn z|4fp$dVu4M_Fgdhbja8)@S}CtSEO6ljYr=vx@UV;+w>jUZ9V(;RKLN|+jrR4Zhjlw z1RgvKZc-cnfqd2O-Qj2JFR|%de?B}0z4co*;dS8+ew+qA-1|$oH~SP&?)B>MFl$5) zt4z)I`Z~7^j>7r%#20d$GHi&CzoBCyE=8x;ch}eT zJODl)gnpH*E##|_HOl>pg$!A81 z^)=;KABoP(i}$TGx6{Gi$FUt&zRtB$PLF)sm#-i5$rVHGU_a#84%Q)0Q{E5ffqLSp zL&f%bkO5+H-Rl@ z@0OF#490HHH}mcv@vWRKp1HucGFw-EOE;*T)-vfjU3C6On;#10lhKCq<6_iT$$QnM zc_76@>mT>6wD67Uzh>_hqwh6tFSA(ht2_w4RerV$i~07d@UZeTR8~4Z)?PSVsI|P5 z9inVvJIVU?I%QvHZOBA?<=0vN@H%HiJXtvWjctHUH}z>9L=W}#P@iD$C3oQ%^7bOQJBIujxdShfl}WUxciv|o z;@RuYd&RDfv1iD3$*S|L8h=c9KZg8u^867i#mHT?-Ugo$&oI0xzwY(hHIE=WD!G?!O>68kXNIL? zV<)gbMn>ZX_Sv_`!*)zx{F+UiQ~Y{>xr!3bqkf9|3%K6RRp-1(x9h62$1X=-!tk4iRd}` z{w|yuI$u2wwn4rv!N+Z+o|}E??IrNk&G3O;Pli*E`sJ->LgN>htM~TN-50Y9I0y3r z`HEM&{?NsjjcggrE6&d&&uV9|>WGQKB6};`{V-c?JP8)J-7%Pj6C1a8CA8_r1k5~n zHF|U%ymKM6_3+NXEY60!jJ1>z)}_HmpZ{3AevMV5Ti3xy7h)$od^Et^(ecaJ8zNGe ze-Aum{M*)xOn&W#Z1ipGhvV{VJ0}nmDG3aBEe()+8W^s?XCvMf?IsqYaqb)NP6}h@ zRK~N!xY{3{!(J?m`E(}ajr?AxeOdUenA#)Y(8leD&74qYBryC|>!j#b%bIff#Gmas z9kCAotk}`jcLauW88dg$=VIwi$|-kwHfuS$RutQMducJbsDbt}Ml&tH#D(A;+cnUkZm?;jcX6BfU=-4r@&OBj8qiE|*-MUfO9RuB9i)P|!EI3ko=O3aJvb+dkxcCFgQMnq#NC!c8scZ{PJ6+7`P&8Mv9o${db zpp}s&u7o?fa8W z9E(2YchDC-3qyPHLYO#~i4mgveAv>%;KJ?a(Kz=v2sfY1JFJs3(#yb6sHCvgh%c0+ji>VLdH*{@N4P(!L zn?t;m@yhWX#red@4)02-I+Dt~haXz#zK8Kl598aOh~~kqp@B=VKa6e2|H)UG9(dZV zy9gTglt^!m)xV4SdrAuPRhQe&EZRv~HfwHy{1IZS?)QGa_kO#DZ@u3Mz89!R^Fe}Z z7I~Xmo80i~hPq~APO^>q{+qzn$Jki!LhK*VbEP%>Bx7juJP!{7JMu%$ZQ5wgZJ(7< zyMt#X$nF|o=)~sh%5gpXA1~}cmC(`9pPTyU+Oi-Wi1@D5LzU#XY=+Q969_uXyFKtch z>Kb;g{dvz-W{v!y@Y$O{k0Q|n-?kT;hz_aHp_x453CNUdlb%^y1A1ayRUJ7hMtkX zgU!`mKD}A+HhywHzK+hrKJNHB*Ue4PLGpF5w`M=9Ec%MS+>f6#fS+?5KZp66(K#UF z@3{JDYI6OQOgB|~bd$|@V|d;|pQSUrdPy=MJ@ifJ53>hP+TcyhUB-|ZZ=6m#px?;9iPPEgFB%{_{m5>QNB(;#J0}03G5JT| zYfVwN%EK$l-J{*( z(M8DReECe^Rrr*x?2Xv>nv3WK&(E`OnV*BcQyFh7rU}2uRZnGMzUd2n zvU5aJ8~e$tQJ=cdufO5`{qVQ!Yn9fTrJ;W_z|rI|d-RXb+qcoKJD>RyIR%ONu|2mS zL)hGyhj*pRvFC6_OnN$zU`wf)`|@GQ*XeR*1iK8xZif^TV&b`$MbxzwU?@6nbW>}n;ypDf&XUF?(|7Fh=!s&;`{Fgg*4mo;-IbiM6_j2}& zb;p3O7u%LaJbnS|QZ(M}V@yCk!Qln?BfCfXzEk;ea&WLY%ij9((c^qGCq2;K#~wfW z<|e)gy%B!2fcrkiDaV<63i&3GThe(Xi!(MifTupAdwpi#K~s+XUQ~9e)&HG7=BSjr zupWDnRpv8(^K#Y<&MBN46)fi7cV1L$3A-7;_Lo@C_Co&wctd`}eD)S_X(C(VrAL|Q zft++R#_feaj6YEles%zQO4l_)BiVb&jqYXVWRupY9*-Tm{Vet9yBXM}LG(-pb*0$3 z75Rg|T6J{A0W10qY*G(%g+0I;@t^z-@a`KA-fL(_u=b$G1TXs2z^XRHQ|(+c;7fRq z{nZqYAJfeSW{=*=^}zgY_5BO=38uB~8Z0ZcLi?1aO<*i+t*di=h@KVkdL*4TdR7z~ z+`2Hz-&9@VtHr=O|J!Tp+IBwtbGHuBF0r29&K#>p{Zt*t@N)>h7(b^0KVvO^##;Q0 zwfGqg_#T`s%Np6@%?+m>Zg=*d9rbm3XhZ&vzEumi_Dx)G{GYy4 z>)ZSAcV?%j*9z8IDQWF(_&vgd_*8h1-(!3-o7W7S`vj+aGVv=mO+JtOvIcw}7hc!r zF>u3Q9~b>NA0!#PHlA+Wyc_wt34bi@p|ojzt8P6i*q#Kh>%r@K@VXwnrayG!)|agG z+Dqw6=%Jgph9+&O4e9>$hIaXH3)3gmw*UHTM+RU0?wczH+TOhEqy2BL;2Zgw3sOS# z&-Uq_pnrp+GyUtsuj|DIX`T+h&h+s?^^v~yVuSkV>mYs2W3Oz{WZCJPHiwX#kjDo# z{iUC#zoTcB*goE6@JrgV%VaYTHy~${i>C%`A8+@mRY&_Kh98}^eCb@_{R!;6{ExnS zeaP6T?lv%e5}4$3c;&C}-a_=vm<%d+$!$+C!eXDPTmNy7Gl!ZWS#Qu zv3(QUPEDWE^64Jtq#CSXdvk>~bR4~2bBT3$CTlk4!{1L-Xm2*(&;j(i);nrkH+}uX zN2Zg1QM1awHKp}`A6XmN_NHZ>nk6+{!NKcI<9DUqw2j@0qdqhlv+HO@7Rk z*0XjVIX)OM(>s{c**@XqcQo(PFdZL`an*Cc({p9`*#`D>RQ#)#wVnmYMeJ!}h~zcN z_G)j@{(yC9XokHnsrDC5!+vS3*I-#2BiND^>veINA+tSL-}}$G4%&hE={MdIn))JF}A;17BQk;Vomc|hoJl%#<7mp0d93vYz>_*I+~6=HG4>b8jBeNZsFsgJAG(C|T)oe# zJ`(*x;zw4!-M!A*NO*u zW_S>v*6<+D&)`9aCN2+Zk8R0QjJlc~9!!i=EG1{EkDMgsjaCv1GCT-wo=xPzaAx!u z(-L^l%8dS$XQTXlr)vC|+*74TbIBXaT~&Crb;0T*u`lg=vmAX~bZYU@wajho zuJKQk{>sA-PxK#ltZLm_ero>FYRWCB32b%0*LNBt^b0RJfuSeJweyh&v;n$LILF#_ z^S@sky=nQfxdrH#4a~1)0N1kRbLNg@QQ%mB ziOMo3M-1w4kg_S%+wkhO(I?1VH1iA8edru4$(1By*qMm|?CV^DN|) zTg|={KJMfAKEl3dONlo*yubY_k53Q@^Nz7M{9g<|xkE%VK6K(oFXc)G#^T%CCtLeU zklD+@%}lH1=WXa&&Kw~BFm13FJj|SK@AKV?-Lzu_+H)YhS7YmQ-1$oS`T^lT%{(GLovYvHo&JA_dlN9Lsw?mN+@U500YODkYhuQT3CPfxu3J@1qJ|_^ z+9a5+31ldZL6cXAm@243h0@3syGhHN6oa6*jM0~9D;?7XqM}v?hIEG4_)_=YnxHg_ z!49_->ihloIj8E}OO-=^-}AloJoTJ&_u0eRYpuQZ+H0-7cW|rD&{_Y{S2wpIt9MYJ z_Pbt}XFXHpyCST;|NH4*3v*-?pKb`>Wy>t)tT&PEQ?M7Kmuhd?-zmL-bjf_BBj1%i zm2_;=ds_=ad)v?@T98}qZ-!=?v!9$gOI4P6C$-LR*li_J>%38QB5O7GQ)i=Fr|*K( z|Jp&D-ONc3f+NXl*&ZdAI+JW$j;u0&amt}>)GSTr%H_{;E}#0S>@8{ zu^vCatv~n7tq$Ef(3Lu&o6h`Q0ln+_|EWXw3D9r?w*3dKtc^e8e+M$^7;;guZ{5se z8i%@XAx8Rd*kALP8?jOOLhrwezTPPGugxguA}cNFSv%$2Ks49DY8=XSm~Pg?C-Yj<<|4`7`vXgsi%QCuhW4;c8+_2 zvD(TyQAk~1;$1e18lFw`M{}e4B0RS&bmzr^;4mFKU}ocu;81p^gXm|S%!wWIS?j^G z2VI%q;*zvZ;tskQQx7tJ8j;}}k>MMW;Tw_R1MhibYZbcKtyu%>vxyzhJlFxPB#-oz zPT2`RF2iOZnws?gmvzcVC)@H9&__CB9kM~NE51Q+-iFQ?0nQD;xdAvg0Ox>vUf;R^ zIIm;5bGcSMt3m)0z=z(vsHgWYodp^Fb*H(qTHI6wc z^XHEox&QaO-yh%E{C?(@C-E`7#@u@I(QElO;hxvFPB?S((G|SU=Ur{P{{uq*a%91s zkG~(S8?ZHR!vJ*Nn~w$_=|{- z=&W-dJ7&V?b)gwAffZIGXcqgO{AQoz1 zS0*^f0S6hRX36D+;&8q+7;#WJWdm-oK--CeUw zAARxj=kmR~PA{Zhm;Y1me}hNB338r)ttV(l& z_&E4<<;}JCWHC-Ek9Gz86R*2xATp=Y$eecqfr&TVGkEJ${JuUosJY3PRaFd6RN(I| z@cFBJ@M9C_>1=@Kb+3@-dFh!kXeK_?Q{zRtvEnDh2l8wDZ=T)oiu{(JMst(u)%g&rS9>8!tw2=;d1sSX`rrZ9zS23Q7l__|^vOzef*5)g zcfd1#f?dCc7iO<1Jo*ZHZX0^;V{3x!Jza8CF+A(;&loiy9RuGqILn+Uy>%TrWEOmM zf*5mGXJt<4s$d+6C&n+&pHZ~R)+H;%XXITDKNc;oU{76E*K3?1qQ(|Y?N z?i(Iu?d3eZnQh#y)9&blm$|&Bwf99K@}Rfe$r@Dm>fk%Hw>HVrMUn z&cJFl=dF(j*Rp?x`#RZAYE9fV_MeW7!M-zRP1@Em*Q~GKKGHf2T~Ag%R{QAsHG!?Q ztRKo+U)j8~W%bc&(zRdOt20}z-#?m58O=l9^cwDNcI*&`a%(ORFXD_#)8B#C-kz95 z##Y-^;B8=tG<&W0Yfpb(M|OJm$GsHn?az4nn$(}6p-6nQFS5=${jYK6CS+h(cfsg= z@tS>G6=SuV=dLx@*2n4dIQl$qO~%%7*W6cMNg6h-Cm9=$%GP@CAMZMfZty0Uj z=2(Zf{rTn1ik0=sO6$+7j^^^M2M3M}O~S#x%fpwrI6zK$aqyUMkb#ck;9ws(usnaR ze9>Iu-OruP`fwd|)taf0XEx7?@UD-(XJacx?=IQF`5Q&wy*gZkT~>O7TlwS6Z5>rm6zfNevx_oLZoA=>^p`A{`K_!}=_{osxV zZyBY%Nz9I)wbbi8|Hyw2@3Z!ojIH$G;N^b$ANh>^@B9?~*SJvs13W)X|Fhu<`2xLo zFgm7#gFi9f^oN5-9UR=xb4wpM$n)UffCmRI{U<eTsjtbMALvmV@1% zwUF$kI>RrDKCbni&h2a;$GQ;SRqTcvEA)uu=oH3_v!1iZqRpQgv!ajm?*9QTyx&~^ z_zZBb_Qj85c(>`5*7$`Knz%M8#exW@|28>!oSHc4i_pV8w9>G3@jcMp; zikUIbf?fBZglithWlZ|HKX<`%Sr|E1bje&CXfYF&XM}7p4--n*iZ3RzAH@FX5N{4JI+>?Dj z=@ZCvCwa#44DvidJJ&8(tU_j^Y+Luk@5?wpa>E0eTMNS0N$$9o5B+c>V`;(nt|ssD z8QnqO$zpWjuTx&%p5R+1-!i#pTi;^v&vEJf2>y^o+^qBKTt-WAeJ>FHIB#|rA&tTJmc6<-1Bxdw|MK(IL@VxxrU$Y_hQfcc=MiN@{IAk=b87c#R?gRGtA9NmD< zJMqDctqT&>N3TQYtw!H{3EDmYAH0K{eF_`xlgq1SGytFMtbd_D&C9QtQL22%he_`HP zpMkq)ojj`NIo{E$4mrBzGr63({4!@1_9si)4xmdrcLFWAvTym_b^u+PJAqC(epbnE zSH4`!nRBV~rL8zXyhLDT$93$*sI(HTORdCw%h%k1ZB%>dHKtl_(izphdd|@ZD=y83 zIVZgOa^&Gnp@j6aZQxI1D#qA4CSK9_;XPi7ZoZVUSFD(qV%jO5skk+*LrN)E1RXRF zGk+ectH}xfh&uZ!GalNeEgMff3+IDaS=a=F&EwD9zp;|F^kUBWRoq@HK3;&zWS#}{C)L(eQ} z>~2sTQI(JP@!&(T!!OYXt(BxBxc#_A{phQ1^a3}Ac;mo_i1*1|h5%6dEL4f_EUE!vAxHs->pNj0hz=i2-l|m!J&uUdUy0u@61oAccQ6p z;tbh0ZTnL1UY|RdGyCgrfiKv#{Qd5d^C;)iUE|G5_e*G7Jd$tdoIfL3Mrtj}{hZT2 zd2DnvfZQZ`|?G(HxpN)0uyD-uLUsnACVb_P)+^;aw*_t+n>HUF4tH zKFUuUy*%4>8MLmg+1on2i2XrDd@nLF)EHH# zH*rnZlC~3`$^(CHTRT#4bCmjp7uByAY#TRy#Ovj==HASjdkb=6tBr?HDjxEIb2Dr1 zEy#(jp_%=y&yIgy_(;~TLRa^d*RrEn~ zrhL8{b6;gIc~rVEVW(|6U^0$Fc^!=&_m|wPX&PMs|^AdV4sWNvV;)BR;{5u~!Ox`IU!;{~0 z+8;yv_i~1r;$&SKrNX-6V_;P~)gQ696?v-Ix9yBa;X&iO5a8eEVCC@>HA3^V`NvF*(;#+h*x$^r(F~0 z46L>FKRgmZM(v^Pk1I3LTjsh4OqpGuQ06jknMrmT>8jONSSxR14^{P5*2)@Uw5z9L zYvv4r>Kbchv-$)sYj|pGrJ1pnR@Dq1moT>48C&U$ttKnIYCia`4uw{7UT{x;^Ng%c z?(c`hix|Ff)@%1Zxe;3Du8I!)!5+b*@yYoOd)@x`RmTa{aTC1PQ^ybOIx>E+=T;9M z`YQWoYWvI!YF~V^}SgF^tJXnn@1*mLYeK}GR-!*~UX7{fpSw$RCtFnVo3)HZ+@JlP-gt&s7mdlJ z81g>mq^<1pZOK56p_`f*tdu#a)Dt7<+8>r9%f0r~7W{Dm=&$vZ=7I7wbU!@3O5Hq0ef4map|0wo{!&gq`f&LlW5FzW)|_fr5iW{1bUUeWvzS-yH4o__zPSNE}} zmv)geBR&gk+^J9{8;8m^K%380mi}adi_)(_Pr=CV?bs+Z&#S(A>MP}5hqct-&iy}1 z!;U^eebYKXep99KeMIT&zLm$}v(o>t&#x#w%=|<-{m!9{>Ew}LSNXrE{C{%VjQUpk z(4Pc@-H&_rh{sgU{`~{97q^sr_V4OFe!i=oY}OqQ(1vYa$Zlk<^O^d8iv!PB)E?&! z(2tUK<|@^Fz0&sDX>*-?^OQzEm2bC`R`eO${4Dg*JS01hdp}Cdk{w?@PumtFTb-9r zqoWx6jDffR#MmEYkNr+~t(dXT-L=hQKe~5g2lmaFm5%+D`;{r5TE}OB&yBhKBQZ{r zf4x3;yI#TSg|7(t8Xp?Kk7#wVjw%_8&+Z_6-;Hcp22SO3)ZL&5>wJg1M_4Nr4{-2i zANS4Jd&Cb0J?X@gDqZqVK3Vl8cqU`|-Q1(BajoYkjce>D&95?kbq7rbwzP5Foh?4w zPrH&Gwl6$j{2_{=lw8j=x>jb@0rD(E*E)`_m4&W_@3HE&bIOJG{2S#y3ryOh-99M2 zF+0=VmpYF9^BuwwI_w1EKt4-d&lTU7I)BBW;Yikf#l#t^Zxs&hbZ4E$p?KZ! zkTXxnw;b=$U!1ungSFQ{>>j}KiWe68<+gva*k|m^uU{h=hWE}Hw`11;*Uj9u5MvK{ z^iIYQX@2IA*dpJ|fOr<$kJewU*iYiL)k<42`u<;coY}&eLy4M8PCsva8;c5O>aI8J zB4MosHP?iYlbUS z!vZ|L_b+u)hs79>{&N%Mdh%Z(+m6+!@9nII6sNb4@BQ*3`QpH0=vEvs`+nWJ0^@0u z{bJt!u4BGYe|LNOyA2*?59Qut;81)ZxuEgaQw}&p^CAaVTRdf3@M%XWTR>U)pUSkp zhW8XRh)i|x5a_DRXKln>ejrbCh_AVb*qS2t$_bZc7H12;30`TZoxbPL{xRsdmUU!m zJ+(P@J=`7DXFd1-k^2VV3DrdmMOQAg+--EM^r|xT6Ix|bzx0ZB;xD96$p1xbmaf^_4s;KH{E4Y@ygmd-(K9 z_*8X7v5R7>Zd6?zj2YdPs51aH7YWawg|AZcs$T6WPLdrC{GGHVpQG&B_PrF^yCVCY zbWhIx0k*}QO|#6{PSdM&4pEdoXzyCAl(CO(*tMN%FYZ5{Vf|N0oEG_;#xvF#+oRL1 z=UVAkknyekHZg3++1P@vgU-HyHC=niG#1;z!7<17(N-EtNbhc0#JJ$yjt@0!sb}gK z<0CY_XEDC1SFxy^QP$Y9sAjM1%v#@WW6h@VVSF{z8y9Y==R))Vx17qTtjfs0rMzla zaCFdb$td67x=V@$fh6QkeXw@TtZ=7O_s)>1z7fLn~7(|ZqG zjIs9^XRL@vJFP&oY{T8)3R!IAq}dMvuk`3HE`Qz#z737Uo1$r2W7a5iA!ro85&ZR0 zCY66&zKQiE!=U+tir-A=1&+r9Wneu1ns?e&{5 zBg$OcN!l@=jQuD=lgZZ z--q+vSA$P_m2c$+>dLP3uWV)vXFuy-*@BGr>b2@?^6%u>sRE0$mZ)Ql6>MG~T+jZR zw5|>8&zQSv$qU#bO#JF&(BnT1?$R1h@s$6P37{l~-mw03^}{MNf%iG4Oj?!VM2RWg zQ%&7jtTT&NEv7DiRi^r16I!{z?SHA$zeV#Bfl%nM!8NoHe`V1}@lZ1C$$b&*vYoiF zXA2JC0`^r)v43&V_oQFTyWtt){G?yM!+zP8fi-)x;Ze)^>lu&W!>ixt6O+ij zEk@^&oX;mFRB@r=?X~cb>{rozE0IrZrs6YQI$v}9nJqQ!J&5NM4@nGhYj!A6!*d;F zujbn}`8Y@`BYwJ-JnMNjftNDyqOq6`f0c9p@H+mtfA`E5@mtRrb?r@_HNy+)lVgkX z#w%V6%&GF_2{XQ|m2cCQZ#42@sC_4l>}2~$)0oO0;a@3OvTOY2EXhwAW1531nxwPa z_QXrweLD`Hv^co`p@aKBdT=iixcJ{dOo|s?*#ZjY+xt|^_(MJ}11{-pUjR<~{-kWP zXYMKby4&*AA0&REAvm||b$+*JW;CY<)0R?S+(|r<>s-qa60?faLL?VF}01FDMOJb}#ogyh?1&tHe^9^S|K10xP{rF-d-8nPQ%#i-?zp zx-p-QYziXJ(0`vzL*DHMFDf@l<&bCIa_skU&hG2S<}Ust7G3?RcKb8DX78bX_-zAo zomYOVKWQT+TZ3I2q1CF*=+@S zWNd`r$k_V1oD28b=uqUt&BUmTvm&qV3!T}<+7E>Gos;yQS@jNWi65$XO6F?6 zvexVJcg22gg^N~gdm&Ey0c?_`R-ozz%H~<=%|+;t0c;yZR-iEitt6AZ@P&{YscE*J zp4<;t{uZa;Fg4yYKH3>S@S7Pw!BXn~N!pj zt!q1YG;?MBJY(Oje8_$--aOBlJF^?v7i5nKC%-#)HVGc|7Ohi0vHhTFKm6I-5AFG0 zKkYa4(f$T>Qtx~ngKmD=l98*+7(;%1ucem%r0B1Hmp9n`t=Kv*;jUu_!Ee4b)tmF! zBVSZ=O}KDb`KnZT?446JPe`A4=ZR|O3C;Ddu&;N{vL&msnKN&5<^x-ww|$|~*)~$n z56#w+SM3yWe_S^Dr|jkIkz1YGZp$QV?=J8WorJzfz8dm*VNw2jTzt8@j-%^8shAxz zUIUF!@NUMdqX%8@;4<|-w`)fx^V7q`MThy{#Q(hiVh&^C!eOCp*EiCz4aVTuIGVIQ_rt#cvgGs`9p6#OE}*~xRmd&#;MN-o^myZk)DgK z_f6uLDxNG#_xv&GI+xHeK3&i?%+2wug!y zI6u78syVX2Q_tdKy|9b6*L^Y$-Yfdf&~QHX>KWd8;=T2FaZvDaICwe*2T%KLTe~Ck zG~d^9heGQ4?#s>m-g}-)cXcq*J;_Rxne$}Ehc!MJliWYX6P#-dl*7A>@2SIqCkNg& zIBzI2bkTcRgI_k@P4Y#w5O+|Z=SZj|KWWKi!zqgK^o;r5AbzB}sUME7s6c0Qy z+yb51OO{l}Nn!zd;CS3q$KwtTYV9~M;h+c{$lfG*fARqH3FimDZqDbtpx6I{T{l{Z zzX#na4ZTY-FZlJ%e$!Is>=<&moq67#_v~{hi8tcxsBz(+alfl=E|j@_P<0cK}nKaLKrDpA<^GgI>2i4Ll5) z8)?009{rminGp2VPk3j|(Fyyl!-~N`E=NXR;ERmQ^+$F9&r6oSeg`mb2c8`I#JaO9 zi}@^jvkyADuWC{18*4d)N-^j=n<6x6wCJeB$U! zPTmCY3tz5pPki}5^Bmm>zWj#EpS3oBW|=!EllQ+Q@g_W_eb(Yl&BcmM?*QlaT$#+D z>B1>~l~n%Bad5huy2d*E*)hfDPx-o9D5JP_@h9Wq@WEAMPLH`cG;@hHe%ctV30Rv* zzl{<9AsV*w|0XsI>5p0`CV$Ir{GIuh-H0!Ay5QLj{$9P5bLHQ==V9SyH+bvrhL5`8 zqwdUz{9*0*S`}ZS^F)-ty%=6%EwP*Sbw1L05>>>E6VP2&3@%fQggPhw8U%U!mq+99rFAcIHFJJ15 zyfzws0;ZxIE4%~PcK}leJe9&v?3?FaL4SP_<%%e`6F5WY{I=ec!cVpKZ>@`O?}Jay z_wdPG@Cl|K^JnB)!1iZfix2ngkP$udOmyOWsE5JP;bV)fljDhzYFSFZZs+WP3#?6* zHABL8ue#-QSs?rOYPXk$#Lu{Pr-1FV9Ruuk9-FUZXDCm}la|Byh90x>tWCH4p~#$t ztF@}inzT3Vv|6oK;8XDi`mC|X{S96*+g&T!cDzx-g2e6e!}`Rd!RbL7K9&W2D~qD<^(A3^@U=cBIIOKM+SEJwtP$k&(?UYzu+L##R2eq8XU~(9|vpt!NEU&EDnBiQ&PKdDXefA-vmK@6$)yfF4Sob1F8u>R7}QPvYvOZmN@eFJHX^(E+XM$R6!WwPuF zE#NOIeex#sF~(49+)s=_cH;;~W*S~Dw&$q<(6RI9a~R8WBkvxV7g>-tpnd^#TmT(c z_y*L^|GX8sVWuy#u+SgLq&tLf%=)9dJB_hy&-I$;b^ho4VA|Fu%Ra06 zJ}YBm2e5bJ6X-yG9!Ex=7`oxn6O(D1`9piyRCg!xE0gg)fbqV7c66Sd&bE$mZe1I? zWea;Ub(Wa?;okASe<FfsnSH2$e8itfz5 zeKY>O_)4PnzG2}XW*|?!^bZXSZy-%NkgFdW`VWT=J8eJsuoOOs4+kF3l&z+{Bz>Sb zXx2AbGjzXQdeVJMEt9yzZZdb+Ri-E1IrIy7-P~a}&X?FWTzok>oOqC!wmZ?sp0-wZ zcMpeFBly37{|h7CP4LR(!bG`cJ*WBj0`W*Z82&PGFfGyv*>~M`CpsE7>tiFVJGOJy zx#o@<=6g3?eKp_4+25`;y1+feQ-Dj##wHaeioYD1$(%nekX3U;zPtFqd-lvO3$4~2 zm+P?;NDq-uHNZF3SBbvf3U9=3M;?%#NqReD^cCh4)n{Vkra&jBjVQdXdGI#kb=uI4 z%2zEp9X$Zg(3kIhL|=l|bMXV5GlGsUeU&p!xwk{{ra{@UH&()H(WjX&NtgV`w$QL#8!I{ zy)%ZLyh!nALo+zj&0k+f`T_doLuWU50=_GUayA9b082t58CVtz@xSC30f&S0k~DF6Vc4}do75ZgR5Jf)$yeLy>tog^_E=7e(5( zVkg;&o#e@^h;-p52Y$)>-N3*5)LeWo#hbfGA%ArBo^Hh}IsT;)z$u^800Vnk<7vMAjemTq zi5$CWt%^W@vU+50T{%-2nPny=j zrO2Hfq!TxoP}`d47CUFKWDH@RV|?wv_Jlp9XZ@Q?{brt{ywPi!)6JTnc}{&yg;B9j zJITA4^X`f`D?>PVDd(j};rqslWjQm{AGZ%06Tj>S{srtU>KFdfB=`^2VN3kOc}ehZ za@X%C@t^kzf2j?B3o;{$OgHvF_ESqvM3K?cQ}ie8cisU1c>SHymE^mGx1XI2{!2M4 zWIO#Y2*K|>ey5XO72yX>>f!GdMrIDXs50UNBxP z7&CxzAn*@C4h|0``i<}SZ@_B-ug288e&GE|AMh^z77og9!!F9|PVYA;yMVIiQ&zfU3-eNM zJa1a#l}>%<)HQ|k4`|~A?aS}hQU$%JS9g={XADcP(0Wg~w=QnxshlI<>BOT4n6Q%Ep>vg}k0ITxE z8UMCl+G4L%pljzD;V{^W;ewgiatt^G%9y6RAV>0)up z#3%ElTzo0dKFU^}I=e;k3>|N#a5bBC)OyCG(OJr@lbPROUb+!`Fz4c1HJRa2=u~Cc zw1fxF)qJ=W9M{pd>^%FVhh=j|J9rVU-a~I{%`Tj&_gwOCjQKX__z_%^<$}xZFME0Duiz@B zOzyXP(hYtDSE&P+li%P+aFsf6DP8z6I0G)9-P1K-c^9XKj4SKOYJQ*n8WCov=&y?i4X z_=2Apwcz~e>#P~mT-s_)XT}rjEcHLn%ky8xenCHq=cA(rxpR{-SW244@FUb!`gQ9h zYsQk&n{E58gOddM)X6Au*@oAb0qa+=?fT|VUrW2nGaA~dUCDOoj>-#vOiPv1vf+9A zAUN|Wdm!C z-Dm<<-vZ*`fVB=--8@>`>YX)}xi>`l+lh&lO#bNz)_u&0nj2-eapjui$1c8U|A!sh z>2K70=$&&ecH^v^v8_I-omt3`we(H?B*TkN-}2~NsnfRtr*9Y1w*sec1y0|#(=V3> z$jKvvfxm#hl^0qii>>+7WG~RVd@+4|%%uf=6D?x&jWK_u+IcVHJ=gM|Ba_#u9_XR9 zzs9C`%B7F+WqqWDlgw#H8)S!`Zw66xN0wbzm@}URKpwO6F;c@mr2ulSUP}3 ze}#urwZSA?&^5%rne}QjWtnfITX_OsbTdzI9^Jy7EXG!oJU-4HKW@C3J&(GuP3{A( zi5|EZi`a@%>Q?)-t@;1m;r;V`C*MAf=WoV7$m~bwq>16AjQvJR#i5AJYU)+u(^^4o@_QCvu6|7oUptj3tdo=Hoak`^SwAH-!yNmF>Qm4@=HJ)c7H>gj1 z(twR4m%S#24y+M>oqaBUyhfVKANXiX`qB%HUstqn^ukKYxcXr374R84O^#LYdw+@E zeT^aBwQemlbj@xgE@H$lh+(TFc0{m<7e#Z~ZWe>fO8TXF%BA=7=ZU6@ZMGU!Zi&j# z-)#D$a;}`*MLFSE_dd32?I%2dgU3~V5;&+79kgG`LN26rx$tN%w*6M=Mt9GyJd?yD z=kOx5^zu+D->Xem?_3W|HT0>vtXBtd`zTo3#z8x+dFhYfZMmF#ea1oOljy z41MnoB`#f-(q_syEoB^NKGk@zW%;@KbZc=au@<{UOMz(U>r(7@v=|%!r{P=rF4`~A zIKK;6kzp~$XPj~_titiP`WowZfk)(93^`hkjd0sXUZ15Shf-i}od=D8M{R0MNtUhv zrj~iMi@cFM5o{Y$;h~%tp0%F<&lf)>Jljkh;?(fP_n+BPu6t#$%j9z2PVW5|hkf`S zqx6k^<sIQwoNY2nNFk=z<;^JZeeX02rxOMx|e&46>%RKtxmiuFW<)D8m4rY4Fz0qGe`j=X6 zlBe8@{gvyZy~`*UzmD}aydyr*d5^-Ac;$`@!6n}$+inmqk!Em6zQ3KH!q?>U(&1gc zdue3!w!0|r@b!})c5exN3fxP7bMbgD{o3yEQsH3M_*YzL^HLxE>acq8zpB4-eemF) zDCcc+c`7b{=kRYIZNA}Yb5Vcg`e^e-Pq`Z?SJ*pV4&_e$`+ocQGf%mK{>t^y-ZoFU z&r>ef+g{Dse%pK4Q!c;1a(%Q{M>**w=lBFKvJ!2hu-9fI7qJ}v?quFT=C5qC?tN5ql=L0>=5#+s4B4)8NF~4XDSHcLxtr(V{9JCMe%= zdgBD@i&03U1d?p8ddj-GtUF+T}_0yc&NERv!<=Uj}HrPkdGp9@M6g(4v$LKK*9H~4ezWOeEzEkIpt~QQ3s1ZXPjE*=W80nd6&+Hx zYMt@Ye5v)wPH0jIO(ajtiD580!QJRt!Mh)hqa!S4>tmgj6P7JgHpYJJPxJjI2&FO2wRVx6t&;VbSvvn6)B zu`gf;g)VoY(;_4A4>x`|n)8T=%>~w}|DMA+AOUaz?0)%Oev8h_oZfzC@#^%etkv!G z?Uj!fY*yT5>z(Y0C3ZsP<@>l7zjrJSojYuO_j&5lZ`GGh?6Q10`ZjhT@i4@+jT<7~ zp&eqA(}_`x-iRL@er{PSO&55+G!mxm;qR>DW#^OrxUw8E#LjhFvNxOheu z$Yvw|QfUYu7;|bd{yX&i@cLC1r>isZ%aEpiq?cLYdHC}Va3|ngE2DWV=NEeKx=Y8G zS81hhtR%Li7Fy(6!SIq*d8dy9kKU_!Z|B|Mfibr)Q2P5Hj*_2JcfC~a$;h>3vW9@# z(|7N9kUuHQ?C0sdFKZw2#gBb?oi)_hKP2C@j?&sedb&Lh?Y(=s^br}=tTQ^q%i4=u zhwo-|`swEt@7jPr@>S|OIFfZM{?$6}`_mbIbI%q`uO#ncVkp;HnWi22@nxgv9%t?K zTkO%HKF+NT%kN#r+R49c@e47&MaR|boxom%_rBrEgkD${QzvIOG^?!c2YwY;cHkFUd05=1{M<&4P3Lit6RR(FM`8s_*XrWHR6q%9p2c(cj3_8 zkM%mV^zH|{oAVngj?G1%@|!z+%S)*5 z96umFx%b?8pY-Wv8R1+1<*c*T^BFq3a7}aI(w#|y%l#AV0b(L*(Xj)_tV(27E9zzR32> zp5m3`v3V8Dw)boMkRS4qDLqQMehdDB66ZVdSAriKpxNK8?}8s&V9#%3;{|@!>d#8{ zdg(J`B6I4H>Ll8fOFw<=bKx{-GxYdbb5?;Xb3$j6`R7*`+Op-16#lXI<6hbm3+<$jf&x_CjP`ntYimkiQ~-Kt*k5B@?9Owwuyc7 zADQ<#S3HK!Y2vd+p#!7mz673nWsEbA72C3u*c{DOT_-q8Ds9=~8BI0T$}-L^)H#SI zmW93!INe=xkiK^FEkeHTPV(qUKXdeCAMkeKTK<(XiXm0MHWJ5M0nZ#?#vR5wOO-kb zIIq-4-O8i0b6T>|jnOe$ySSU`!)|kzrr`qzHr?5#HNvIY$Paw=igO=k-W9`>)_5N8 zdJZya!NxqE8fSg!$WL-dMXHW`hx*8$(T{8G{XEC9y?XVXQ^bpSb)8b$FuFc=g#DSP zpZDiz9pUd&>VwS#msro@8n~|5Z9p}*9}Exu}-Sw8E0)% zj6Bes_+K{>%L!eh&{%lMBfW@uyVSbu4%z8q#r~OYT1;oU-Gr|aS^l=l!3)uGb~*fs ziBVt2mLpltnk)+ctG--fk)pFK=Zv?(;T5%Kumxai!Coa>WEXtdRm^i9zu}!I^ohYg zan>uXtX~xC*UlQIV>;`!S;QI?&~|ar%oD`!8oD`p81sOUJ>^>SvA5^NkIt1r^E~v} z=Fq{NzYRT#{bs##EpdRSh?V{Ee-%U;{8m`Fk8Pnpn~C9zC)4IMu{nk|3yGmZ=X(l# zKSlh=JK$J&&=|;J+{Nfe3;ig%iavt(kVSmPjnsJ)YaZwqFJ*jHGQO6gqgN3Ne>-d1 zI~jL(vCh34{8V#hYalrD1KLkr3x3Vv55;E0fFm~C>@^BUH}iXo{Pk9s&Te5XlE^pl zkDT|yo(J8P7X_~9MS>Oh;DbDD^MWlKe?|>*w*?w%Z>{1641dUu&u`^Pvg0pK&M(`! zp}&(qNBspSQE(C;5=!X&1Tz+)bqrb?J3O$o0Y^Ku?ts=>mme=?Z9XrQ=(-NNR~F5* z$7--gC)51m#zSk){LP4S$J_?mvgvE@hsq+Z*n7UGJ7<=Dne*)!QyO1I$ZMVP8HEOU z=m^E=2-SSoJr4PbJx0$ndILJ1;s~VU)oLwjO=Ry>h0*cya%>$h&$4yAJV(dNLx+=& zmse)%cuJR!r>ArzeV2}x=jeFS)nn**#hgo{Se-i7dD8ndpJ<*;?0c>XHh8e z^P8EQkXcHX%_asf6$UwHK=K+m{;GO4{%-W7-_E{K(q*^RnnE!PtSzUhJwI}Bhu@n1 zAU5u`7I6vWl`M<^hp~g@16QfyV4d%@q3_!TXTA;RT6ARtAN*Vod=}?a$#$gmYYX)S z3_kr`*PrSxG3!9m^!+CqFV6c`y?+ATB-<+$hsRTTe;Mx`jI%b{>0n$LdNPh>8`3#_ z+WRsO{*q4d2DHooj$^#%Aq(Y;YG>?dQkYHZpjJJq6nQtG?hNE zc&NXA4r{S;(G}cdS2Or`Y48JwwqH-7ZAl-r9pRzvH6GfEkKOvBlWX?g+KyD!yB@Z{Uzt`146gzA= ze?9s)wDQKc<~wWG>iPCMEgzcu!LQ^+HM+Pv&W!8;H?ong9%nsLuJHg*O76G7>)$1< z0R6+9*#%yVpOAEaM(_GqZD>td?W`#S;9P4;rAJBEZ?#kHtSJM)qBUi;v!>K{$xGrQ z%$icaU06rck2+*_gVRQX)5d704W+9M{Z<J#lO3bl1kv z(@!fyip^yz)}P`mu4 z*zQ#RikR~xde=wm6w~790bOH>NnpN`9x45!gWoY^h;&==S=`|#BQN2df<-1)^hkiQ z+QJxdZ4;7@#V+4kK697Dk5sqx8jWARKcc$PW%{fe{v7Yrz0(`-<*W(RR+Kt4kHDWc zuAFb8Pfwhko8Q4t%sHnt;=xgjv(L2;%pHFluF?1#*|UB$^>JpY=&SgXF+8Id;irIY@02_U^51~HT{IHE1>t*( zagxiw_{;E;lV9_h^o9QN_nc4W_jI}Y|VDYr*yFxvL%=2cAh=F>E!R>cjxy` z>1obb&IcY>9t3%p-Yq}SH-IsO9Eu~K7xyD4D#qUOA!PZrk#W zt*4z`z}Do*qb5fl)j91bUGhl3)lQSshdQU7CPyCWyV_}Tog{gLo}5D;m7a

TF8MZ`H68kK8~W>7@S65&HKO87v=$zQja^@I3yxNySnIO(&|iQAAd`MXu!;gBN~GJf{V}xA5LVd}x&Vq&J6Fm7l)yW7;35_8pz(9C=KeZ_;Lf zxU`*;$J8xY<22^mCGcOG zv(K@OeU8~Vf&bWLS@rn8OCH0XKcXYM7{K2X^wU$rvo8-x$WeVjRazrjaZOn@3P+o8|U_C@gv~6U z`N7C!Yz!?q%&T$sV|U<_zKiGGJc+r84#8*6{?2Wjsb$p+2oqO^t?LWnIn%ngtm7_$ zQ?;2%ZC>ZLd8aRNfV`E~RNGI^JtiNJH`S;5ypun#jB!Aj590W@Q|kPglP7j3zTa-- zRs-W^X9!>TApGTNN$u})^LA4o{kHXjLAGAZKBp*qm%jO4_ZGWsvR&<;-Lj>2+1OEh z(+3K%O$O?V*vndkA6I*7_i;9q#eKjfKR{M-=7zt7y+E7IoQR)Y`PqB8=ANT|eChgK z%@8a@180=|ewx1?&c`plg|vtG zMjw92?+Sbn@h!;QhuHgpk7rlOlyE!#V8wpu43I3+Citd=PoU4X!~GxGi!}8Rzo)&FEtO80mLcI4lyU7h8i(5No^|7yEkp6Sv~LY1_Kmmp z+UI)>2@e4$EyU}T4Oc86`{DSF9NF}$b(QQ)rq4Zb1C)(t6eeVc+C~3k{8rpR4!W$; z(M1!ZSyKv+Wr{rj=2m#K6&kj}pQ(KEQ{s*czr#nd;lz2c5BY-w@LbBcxYVWdG5CqR zH9BJeo{sJu7%u%T@DpEVA5pWX6$)b=fgoO2`+|9N&xD`j@1`0&c*Bb$0* z_toehNWJfQ){bs{?tR?LKc_l-&pVrgUb8E}c)SVN7BQ|%Ijg|cuUR)N0|xZ# zlP#=Ib`!6EENGqFz+BeEw-9%GPk_fYr?R(tL=$T$ow3`5-daO`?b*oTcMji-zR7w@ z<4pU)Y`E;AdBix_dSf>)|c&V)QxbkLVVPnK%1Sk!g0GbYkzh z!-00^pfffwhZnO4TI(rqTXyVPTI22T!z~M8g=>2l7t+P9 z17CMj{?pR`2mR20CFx#X5f6OH;Q{f~Zo>oVUCZ%<2roLfVHZ4b93B{x`^uwfKH~5= zt3i21Pw}FcpEOQV`AK`Fy!@1x+cz&2JG}H9`0R_9bUr}8yo60Dg_jQ1aYn#zh^z4M zQqLIyo%r)pc}aat;wRQn|Hr_T-ngkhnEvx)U}BBe7fg|UU^>tTOiA#tcJlIf6)^S5 z-|u}4f3yDVgTD=~YWsof$9=%1^H}Xz7XqHU}vrTy*TkRtfw_cYHcZ7<^%7N#xpvcf7Vp!H)ii& z9^)*=Z`n7Dt&#PG>(ewc*0I?!R~}OA3jX6~a%;{DU-3b*|Je8#nXCRar9Z!cKYHe? zV(0u=WAnVI$EG(BIaANQyNl4bsskZ~)WfIRQzjvOD( zBhJ+5lbIRllRrQ&9fV#w(AGguDOL%V7wh1&;K?>R`@ zHV*AFqP^zx+Z-JI!of`segd|CeFE#|Q-Ngvx*f-LO}OH2^q;$Jzmn*?gEN)ochs7% zIFmIR@k~$1zXVNljP5@r{58tR_vAx2-A;b(*N;fDa^}>;-5Wm#Rk=mzzrnIWH@D%6HuHN~p^+)vE$8XiO51qh2 zgP4w>Z)GL#rOSLPm-FmOWK8eTH3GIz-BO&@3zNc8=xb=}q`S2E2k?-Mt|2=Abx`+~ zjUN6t;~72c?2z8`qxZ3H$`0QB{#x2I_0`%jKx>%qyz8CyS~o2YhVMnMb$ITX+%FFa zUkQ(0aI8mvG-GrQ=RVv@`RF+4H;LF1t#4|khL3({5ot@P+!Z z&746oB|L?CD#QkFTXM-lwhtaB6#*ke<2{0pS__(cD|@~Kw3HmV$a71*HjvO?4RImG%AlP22;`&^nO zn>v~Qw5Ija6Cdr#!v70!N(Sa2L-U!>@|cIUp7z3(#Ao}4hp&WQap>;S{NGn6?N8j4 zJM~hndyJmp?iY0BNj7Ce*z)EaK|b+c%zsk-p%LK=mB*`VSR=y2Nb8HYw+>9!ouvPZ zxAo_)x0jp!1ikUhE`Rs%$Nvqyk06e!KfF6*EC~lSBf^K?2k(r#|HS5-I_|y&4L<9* zdy)FO@V9^5xGPS!DGbEloJ!hS<3bDV;Pz=PJa2fv-s-R19F}#-GJ?sw&WXU?^8)J@jAX`J74e|K&GJ!3fF=$*E%YxD)9Q?QPfkC6HHnO);A3h!pV zx%9!aTLNz-&#_bBkt9B}E(l*q+6Bn#lEY`W+{nLI7xC6TM{_ynwT{K6mPXxg|2ett zOR1;7vR=6U^MSKl-byJuxDU88Ni%psKk>#qwlZe4-&6Y7Lj#iKMVz@xJ|pe*yxzmx zuKYMzd%p2syk3qzUOOVmf3bI;kYuVyYM7R1AG4Ak)@AR`zjCJ2+ei)a1leZnaNIUsII3WK-a((x^dFqQ2?SDm&Px+qVZEp)a_P0>tLE4KRz{k>r zjk=F^?{w>Q+burtoOV~ZdDL#c-IwlZ-CO3eC+=0w%+1EW6#W`_W>2Lzw)^*t1JnOs z_!1qzEi`AyTwB6D-OYuGPVW0UaYbRIkH-2 zN+VwQz`*det8-4P9>v+}8M9_gdyTs(Odelv9<6_rcAPTfDdV=WV;*f#|8a1r`nOYm z26cBrzYOw?rrb-^`8aiEP@kKAIq4nb>0Fg_dKB;7>?hhj8`$SLu=AcoAL`yu-7Z}^ z4`iA8y>K6-4f7q`yWa)3(%j!2@Ny@<*k$-)Eyowzg0H$0U##xs+yD)<=UQhFxqj#j z{H~o5HrzA7}31PUqWJlWq;_ zLavt_Z?MRV68@C|Y&XxFk?tVMl;yAJ%FRWgt~Z*L7HzGMw< zemZS%{a*s-HUH*J_2%CU8?y0Td~^?R-undK;O1`ZB*!^F>773mCjLTu&#bOGeT+1< z_YPwP+v5m}y@uVux_r7-@)YpMPuN6zPfWv>?;F|`;Q1HE(n4!!^RV=x_3s9TG`~w* z?+zQb@h`OVly7J=xOwh#Q;~JtH>A6Oy6&?Q4;6+Y;TsAgk9@6&bIb=vPX6}1$f-XJ ziM)F|=RBoBQ`QwtKF$I9f>rYR=R=X-78XVh-(Zi^<%~;>)7y)AGEP@6v?9kDrv;QN z6Rx%vC9Wd2Dt;I8c~+5$joLmMJ)xioUm9~EKF&7s?65dHn)Asu-~H|c^9k@p#}y@9 znbNTtnpY7Ue_trE8MqBRdECAF#ZctXoWjV_uh{UcD6rxA&O95QJ8!fiCxE9Ec!W#U zYkVQTtS%q=j(*FBP$V831&`e~e|1Y!A?Hnn66|#kPdItc!)@H1V|W6-=!~bYZoR0; z*ozeB*$)3l4;CgSa~$}p!T-_fTPz4jV3WnrH| zHa87Mc4I?0QH_73&^q}N@_ZCB^qt}&?191XiPbAkH<7*rIruKJ>cr~Lop$r?fEV4o z16DctSckt0j+IX~%BEq^TJ_i3_3xlfQ)Zs(_xIGl4t{tS8KiRiDAy7g4F9d!bBMEH zZC+fn=hwWWHxYwi?4_G>_gxTfk^R-l`-&&;OXl6Kqsf#0Vrm`Mh2h^)M^e3{y+NAj zKwRwM0><0t-n8lP!!H&_HqEi=@Rz3ZBJUj>5;?*6dFRehqLF(BgyVc`K)vuQf1dE` zqyO)KU*WL=+^hfRf!klNzU{R5^V{&}&%~d^d|-b|T)Aw;=_{EJKeue+Y1!R{LmzZ` zhjIHC@TlLc|4WWhU(BJwU~u`=>TjRk<{Ni;&a#OwYAxTC%e)3Xly3(xs9ojTw))#2 zk>{h=$aB1&{XOVaQQ(VFf6H-u?DRF>b7xAA=%o=f6= z@IBCc9|~;;H)fhP1~;B1UoX#j{ z$F6}de;+g^hV`Tm8%PCus%&5Z{6(@k%H}hWIkQG_kJ-S3pC|?_xuh2|XDD7-XM_YP zqjkH^;iw3%wr$?UjvdhS-@fy}%nMHXl55Xw*@n&#r%w0GHtDwR@7dUZ>d>(j3!wFx zV*KJ4A&<0ApZ3<%-fqsgspE{B4%TK3oN@EAwf<4*bc!`lY=ZQ(R_d|iPw#%X$>JQr zz}c_S!aHbo=oE~EKFdzYw?BxC{G88v7RRwZ6aSJ_|fp_Pua|58SMLdFXwQD&{no zxxWV|y|(1uxvJ*E@YVFI6@J3TXK?ZgF-7Ir2xQkaIH~E=#%FLcwlADGy#I6xUc~!m zPcsbe;wNJGyLW$wYabVW)RzYAx!aLXE{-a&>4;Zn2UkCw22RSk2SYYoS8PwT9o#jK*wyMr!o;oY7 zjNb9#;8pt%qMkZq$Z6HLRd;?OALme~tBZ+$UHF2-YL2V}MlXEu?I!6Hg=YWmMd4q) z(?cWDr-pA%rqQ9?U-e0&XT-O^q-;`s`v!$yI4AAG@Xtw8Kl_qn*YrWlhv0L`D*cI` zCj$XvKbbH^{>-!_`$@~=$g3x@3xZGWyIxNWieij9C(B01w<*{}u+P=vZ?0s&^4-qa zH-R*xhc>YOoeypsSRYGYCr?7Lw@TkWNq2YK=Zl;D39L8Ql83a_vcYF5 zA3jqjZLiWKXLPRk{t0$n1<5OX!SJ%)l z+EAHX&-vC`D>j5^?^3%g-%NS!Pa97?+ULogo*TV*j}dop32{P?BDaI!yuliJay&RG z3ZS2{&e@O7m}^}W)?R3DxghKC$E=alHHKT6vn0=dLHpQv6Gay$`%0EwVEal&^xW@f z_d_(%d6}uafbQ~0b|3We!YaRg^{0eY`^t6#>yLuOwK7*J9%!xut3SoBgKfz8b&k7x zpd4Q54_0j9=A7WtiY`fcO}bMAjMpWX5i z|3~@P{(!Zt0TnYM8~bg-0rPpwWb}z~=r^)QPhl?w{=yFxzkfDkdJF0F@mbxQuvT}t zasGn#|7)+Oy^jn#Hov!Ku_sOC(XsM57r>ltMtf!OQ_FN{fi2za=`OYst@E(+UPoL3 z{)dhm(VK1}u3!;y1;iV4RkC)-56t|4J=~{uCdVy&iP%c5pJPjbcPsVX&iW5pYzN=1 zRrdV{4QqTW_i1N>`0%1+^izLg%Dk^IP*y0riF8i#+O)uW31PK#P}E%oyF>JvBAI zboAVhijU?F?)7~U56XEhmhpL2P!Dqn{ws$^J^f%z+4rKMBkJC_f}x@JYm5xRKi}Q` zY-(G%HH?p*GOHdgMu#k@u}(gTji3UV6w9!lt3Yqsp(i?Kg`Ug{d3q8TA4}(nKV*lV z=u{PYqSxl>nZx%qo}>A$=Xk#Bc?I9~WZcKNwmp!{xBuYiRIYw}t%-Ly*PpF8 zj41Py>MSMRrV#lRvqICgE>zni|3%1cLu44GCdNJ{0=+NuXvx?CD7qPdf zj5}c9^O@JfKRgo)?B~q+_H(u7KhpQkcD@%mPvoD{Xa2qYTz!rG-1~RV_ZiM}8qbyh z_8-goXL z>~*H_?#kWotxS>Ehc4=CO=;%X-?Jv~i!aJj_=5Pb9$nkw&(+aYHUw=Ir=xWD&*h&9 z$g95ApRyY$21WCp@M!vSz?OaJtM;C5=eOII8}m6c1-3TNdUwM4{kNxeclCVxY#7Dq zv>=1@dp~PHqod`Z55w;P;K?6iJ?rW<(q-2C&WWvbV6u=SUYObAZ1f_xp0%ozdWwk~ z@YeH>#Lgv^lf4AF)^p~K6E9Sn5-(IrdK7rICs^x>QfEJM0J)|!7)m{RhozI1_HhP7 zPrMLyN3rRtPU+8XTiO$>^@H1%etX;UrL?8*-nR7HPkHIAI-^Hp-mNYPzmqbr-#orwe$IoRlh&Fw7lkiNO$(hD9@$6Q@P5X^kNwFqa*3DU%z1Mc z_s-`9$VK$SIJ#jgx?vl-;XLr5andawAy(GA-pl}JTAwd_9Dk%QqpODT@PzW0W1C^@ zR9hKcKAz@oULQKI&u{Lyl8vAVeq9Uiq`_x%GJE-K9p7q68%_+O{GYw&Vx%<&h$pRV zog97{eWwk3mG+JPuK}zzfHw-fG2n$?r$;L}!x%l%<>&3#iNw$Bhj>`Ov7pKAxk9=DAfADee*NhD3 zrNSQ?8NMt9{>aC`zZ|*fov#AmHNaj0Ke*Q%8-yM&qo>LKu%3U7!E)@Tx<6zLbkbNX zW^C-@o4Zf!ImUzJr1YZS^BzK%Zb2tHz`LG@dHz4ty$g7i)s_GMo}7dvTm*`a7pzGL z3W!!6D3Gx>Cka=zwXIm&I&BidRcfuZGwNWQAXvqC;i&UtMrX=J(I%~xKh?2}(F zq^&4zr=3oRb50UM2q;*c8U^zEeE0kAoPAzS0(AQSKl9J?wc@4eRAYp=ET+H0>} zM1LiiJG;6evv1CfcazsM;~6}^C;Y7SjChh~Zhzen;QLCJ{#wGgcH?oqKvoqYtCk~+ z3_Q_b9AfXi^9RNebdddL$I%yWOveLej3ewnWJ5R(*Nmed17m05vGIW7uZjTPf6oH% zDrlY2Kk$2Lj=#pB@h#c<#;-+J@?F7_GgtJp`ZiC#%mn4>O%dA<%vQ%GqCTdzPlTqG@_gdaz`~NTc(}!)owa zC;u$+Xra^BeQ{{e@St*TF-9$%zj0`(G1~`ksQ+!eYDK^Y<=E@8(faJ2fEQ%@-K20Xj8e`uy)UR}78pm{+{j z2Q)o5n|)&X>x0hp z*zCiPy2SS@LO4oS9JAkHApD#h!}1jmgkQzjt&fy|=kaix2W}NEZmZAj7q@K#;C5ve z+;%uTY5CEQuRnIaDS3vSw7PD^->?VGn^KDIuOm*xJbP}C(cg9G?`_Pxe#O3|k|<}b z>ig$snpjp{*FHbUtXFFNPxZ-Wznk%@Ie>@UC;x+4>q&90pX?s<4e~w0d{ws3w!*ZH zRe|rV#K?H-=L~ETkG(HmvA?G}7%OVWqn(N90L^*E%dW%jua!Mdy`}gCG_RB|=mq3d zg4_}EZ@RKq@7;UZ%&nBKmp_*J1aCdCt`F+0L*BUa4!3N$uCPw|Ed6MCJ9R_GTykg0 z*fae8N47iuTEp8_l*_E^%=jmdd3H*?EE}J||J&Hl>yGhk@%5`re0@4UR{}q*;=L6E z-hM6rm!g}daz6=OfUTEgO-K0>)tA;!F(wk5DqZ~cJIIu&$Vlu|*>kdCm48vTs=i-9 zAGhG!D2ArmGyMGlzL$;Pi!c2=j~u;xt1|&)cef42-hTKZVsLYt=d*Y4*PLbX0&C;j z@F~=>R#DFW49Qmu$4}Y!KSH(wa2b6BY-K!?UqJS_VpXJL=fUe6;PtZtUeBTb@qd{- zzJ6b)_W-(b#LxZ z(c-h>BVaM~0GBp+uDu+-tMdJ1rks2u=w|izkxtez;Kk+I3(MMt;zc#bmT#{FKVvcL z6Ge>u<;Ve@)uwv!U-cxYuNYd0U*x~4A?8%Msbq`ZtGE!Zi=oGufOm?q)!aVF7|RW3 z#;<^m!r>+?r5<%7$w%GD`8wSRs3q@VZ| z^^y8Z{iZ$>j~TiLzD&2jI>7ft`YXcN=+BSCKF|!`itb(CigLNSejL4beJi?$zfU!B zpw_p-zMiefrm^((xlTW?LoS>OT}$a($@*gC0^=*(w>yLN{Y@pUXU2bt&ryB*!NH8a z<(tXz+RVOP-8WCB-@^8NL$H^#mHOUh&HhCDcKQH)dmnvk`?Jf~KgZe!iAmlOtpmD! zXyi{Y=DsPO1>O|nEiox}M7}BW?PPzRZcZrGLITl~(w}J4T z=HM&6CVT~dH8Bdt=U3pb{fjT8Z}hwb`5`&1Soi7?eeZ8mtU&dtefL`ojn3$k+ZlF= zrOz*kNn4(#&+3wb%lf6y$FtB!zFN`er$>$*^X~puKfWo8J_=;xUcS@ArL9qOH^I>m3kzA}9BAaro_ReJ0o|HJyq z^2?!i>?n;LzUfaciZ-`op^Y2gbf_xGe?W}I(c$U*2X>Eq+jYplYV=iI=KgrsUkN^Q zwQo&6O`Q*5>Delq1v)MSXQOk_HCj{etc&%qzvP)-f zw%c;zV`y7=jAM==xa>Tk0y=8$5hbs_g{6)5YaLjA_nS2Zl~tVrW1^z5uG z&rZwoYy{6b;rl7zr~T=POVhIdpRZ?3D>eYRfNhl?)5sAsrs-R2gW55Dm>ek+^BB{7 z)8Cl3wvNt<>@0;p;ZcqAZ?SK}ofo-wkIszjEM+{vGf~F|>eH8hqW<`v@E#4%$v$=U z$5__S=%4Dd`}BcX*Q{pj)xrU7$~HGx}tF-}@~_p9HaI(mzI@Ts*Koi9z%A zX??P($~8*6&lF~EkV{89UAyO;gJl#r)wLj{nk+X?GA5y z{Mq*Xd=kg@S2Zyv9zK^BkX86yr^*-Hz?p9z^C!vLV%pDn7qDb%!{47VsH1&xxwYW5H4;$ z3E!V_tbLb*Tu$ZA?mN3~I%iDQvzI*y4@HZyO~&8REnlbR4V6Q4Ia9A;i}-R3zEl_2 zu>IPOT_v7cd>MPp89U|FvxGiOjY4LzzueZN*p1->)YV@b8S>UUE&fydS>b0}cxLUy z%i-gS^tfJK@{Aw$NP+W8E>uB@n`Z0X6Vu@oj&_LXUo^RKH)D!`?c*q z8SeksS@E)Co}Ch(+)w}h^n*+}bWf;%eScKiA1FVdaze-_DBqxbg31lC7F)gq+DJ}o z|E7F&cg;Q`zlxtDC>q#VS0Sq;7p@k)iTk{ix~5X+wTv@8zYdu$d2t2)VD{J7O`gOV zp!J(OkoSA0qKlCoowu^zn7m$H$di4@7b8PvG1ibDFaOLRpLR_AH@>}N_?j_lZ2_bI zLo{&p|4&2oSY`Zk>3mP}g(ktF4g1~LG1wGp&-y}*{=(i6UA}(+TMj!ywvBW|8}?w| zJ_yZ=kmcuFxxK=df$fw%uwS+Ub)X{|*PG-smg?;+y_#B+;WR3q4x<(GT01KV%*M`R_kBhOxGQc*Gf;*LoiFz1_@3?m?#J z1#9RnL3~V05R0REi{fMSJP9x9UOt?bAQne+1o?1Uf><2Amk*~Uh{e&pbdPe6i(cax zzw#^V3@M%StNlZYXH|X(e1AO~@FR$JHs9u2XoK7^^Mw+0%sT7S3Dz0lDQIu@Y_JFP ztavr_%y`ytR{Tr-@SEt`){{Bb#AiQr`0LO%e?B>9jA(jVI_8CWN3sqcV-BQqoi%^# zgy+O(z0>_XuD#-)bI6IneFr?LajN*3)FjFVV|6Pub$PQg@YCmsFBv~c#+;cSF*e&7 zryUw^j60Q?ha479$AU553QrDHMlm+*{V-*VW*|55wHAFDSoKTn4f<_vA$%}F`3&X+ zXRXfgb2t?Se4yvz1Ko=c%C8CD2YgWe#q{^$gYqw@@5KjQ@WF&H2W^yiH+8Y+!9HKY zIa!Ye-*gp4%7mj`k7zJBBEOD<<6h<=T}2gTgTdo+@Tdk}W4nXLa>nvt;8uT}NjrMy zK2xls-i_2d@G|efOV6~Qq#E4Z?~P3le(tkqP>0Tk-&gMS6mr(gLSWGzvfaVh+8vCo zg#k_VTr}0aXu3NXTMGl4?heM5-ixNYgR!N1(N6R7>WFt}5&M>OHhhZMqN35Ch;Em zV)?F~dMt15Wn8OVH+rCpy`i1(ZyUT@1AgMQ_0)Ga_1#N-_fel{u4_FzKu$=_qq>=U zb|JGnk>92r>h#YI&THOAo1Mt$E@ZZ82f2NKw)V3>buZ(?-yg0$iG|3QA;_5m)}@9L z^EIsb{r|9Wh`!#la(fSF&5JIQ+ezqD0-Ykr8O_5TzowDbvl7ORGxGj7{gs2Z;c{h^ zGq@nb`t0P8&TD=DSXpFq=lq`7dTamdyj6Gpq4*EwSJHe68*;+Q~w&kNS-E+TeB^1GD&Ed_VxAizlMlg=qL^+vms@Pyja zx=-pO4gS8fV9vCR?{vodhTx3%?{k(}_}LRY%aHl4qvO9P<}~A3!|3>L!p~x(rH>wQ^o5CMLym^~$M#J$=$jh#4L0#s#)W8J@((`Ei+OJP0lI!> zmv1+1V6LE;DR(ZJftz=Be0?^Ufp=mE-fZzj6O8?oj=vKR3Ep+YJ-Yq09K1@AW%H0_ zD~LZ*pK6cS-Sl58^LvlAUF|PQiI2pK@X7ao?#Cc$u2o4KlIEL+e;Cs)KK`@JbHl>@ zd_lOMV`s-XdR@^&Mr6|LP~6Bta~ji|UCRc6=)EhVlAu&i%^hPxk&^m<6vNJ9zyZo|8PV_Q?bjlY6Xv(s46> z?pu5tMYhVk$m}V0onE<$&vqs*phCQjOdboLd_v> zyH3Hciyf7mg*>@MF}HrXGof(`8mF!X9@^bl#<`Wu8Rc_IUX5%)uE=hzL+;eE=G!*Q z*OkMG!O^w+P1<)n78#_q4e_aghw>@jOGmxCoqpt3&+q7yLq3dc;5MKA&x$wI`#gAH zBjcbR9cyDwwSS*{FwMf*#ZS0xRJ?H&cwq;s{Ce)?V>hu9tsh}Nb^B)d#WkPG$LHR; zKPSGJ`DVs^YAJg4wxeslAvmFh<(Un{mKy#9w`{sazEzEj@ecpkya-mm$frs^+pO!! zg=RjHo^NW7Df_kkTI!`gD&Z;ZL1bT6GqHD@)laR;!$u6q+W&(7(6s$p=#1^Dd70*C z3ncI0y>-aDBsjmp919!o8O3-jhkgp(Ys-B=eF?tNw}!64yj9Px5?v=zFZwS(e^}>hi~=9GX}+oXbodcj|GNFiJ7{8YXf^`ZV{Pc249ZA1Ph`Mm>Qo|Q}T8-MIWXeye3 zLoK>mIwRv8WXdy7np?yetkLg$-w!Q(uMpoOv$QX?2Hh&(U=m%Uy~$<{ta`aWjW%+| zs*ikMp`m*)|5ZKiJwCcaG5K?Xb~KL9roI;fKcw4^?A4F#jGMoIwSg24?V4eB0f!M|oF7e9;Za9^ql%-$p(EL_Nhv^QVU4PxS5I^XpfP zd5!jz8a=4Bs@YSqgLcXe(mIuq9sHVDC3x#K@Da~z3{-qUG>bN?PgGAfSiTD^w*gD* z;p4*+_Wex_@!TS3%v$?!JGlT{`@`yU>FzqoEY|w9A8kIdp=(Fri$kwRgLmc`>lTId z=lsdM&*d6g*5hYeKMub(bnse-`LTY1jb@&$%kr!t^Vxi2HI)b8MB6JF8?=`=<=FOG zvpidu`K*4jcPPrdJp(uJ)&49C%LH(%1J8Q!{9bX;o|g%?!7rBOS#jpG`q|V0E$d_c z`7gHJx=#f4bLN##zo8lUsN7c8tL^vP-&Dc4g0*Y`HwjZx01NaQp*>?{1o|fQ!V97DJS2L;7LL!-D`e|FV}w` z%un&-_IJD>`W}it8iO zz3R6rIs=pEy23Mga4e2yU+yU22G`FyxLP~Vjc4=MYt#7ZtX0{5!k2Z) zBWCW8?JS!?`2Lsd&;VcAgNjRA53a0F9#PDmN6em!hj~ui-pS%YTSt<^fhEABIKU$@ zB*t_25B|z_p7s*s-{=RQN#0Mw_%wr0Vp8lUg4^H)Un=&j4SI8yVslo%fJ1M1|0gb- z|4puX@fQ5gb05y%2J>P0UzV8py#pV#mXFNvSCMxl8_k!8X^wu1z)R>}Lvz-3j+8+A z2sn7`o0kn(9dr5x`kQarL;olC!;vKYpm;^`)K4sqeqZnoJfT-Mf5s})AL)1i8{@0k zr6hBs1oJMtAGQU%ex`ir#KU}kQlIilT5_#iN!DVXU)&6hNG zY^;gq%%5kUtJcp9+~8s04s3sSo)R9r?s{WwKd@_E>dpTl{PTdntGj>rAN(`Gf5}O} z--O=Iwm#Sfue3w=8H`QkbC>LCAx3sKlF6f!Zv&Z-(;P_^k1;n zR};2J~8@@ z?T3EN(9g!782U~Aq`%&gg?>YhL%)amq2K@G$m1p>j|Zlo;Wg%Kq7gA4eYw_Jn5&9c zcy=V>-TGbS^J{@$l!tO1{E}j>Eney3%np}V#J{m%&->TFw>F5kG5Fe?OquxNzn&C- zotW8y_~qJu@XgToe>Qw?>lfdr2Eg|(PXgbEj)U*$6T$a?kvv)F%{zu>e(Xzb^FM`X zX8B`Jb%s1yU6Ng%82*@b9D2Qe=>OmHWSv*nFa56UhkpMD`Z@B1^NF&`6GOj05??*g z_^R)Rey%)Oh3_hx{xY@`I!tjU(qkL&KgxHyfIVaz@!d&w?#4&xuYW|E*RvKUo4W|P zQ-j>mTuydf5})xHt|9Ci4ZfQkJA@t_tO{O4cSBf`r!%4_fPn-vCQLsQc7OTO6XM=(5wYpI*FX;-4Wg$3CsBy zG^-WekoU^1n-`*iEAQX=N;+;DdB3wScV8BHfAw)_@VkTA$Ft68u>RKm;#srW2Lsqs zwdjpM)t-_qCi`Ih?Xibi*t@?@{G1=#%sR|!@9Uq4%NEm^5KMmTc2V=I-_#zSSfBmx z(p)*7{oD{*XJBt{h)*rQt%24hY27-Y?0o;<2Oj$k{m}Y}Fu(OP{4Tzv`e41o@4i3d*zqpB z{Q0D(y-uqaaNb1`@rAPK#GB`_{wf)~uZ{R4{9nEtR<1tf>KWAhD)F$Af3Dn9E@|;$ zJF+rq*1WLykewAFTJ&e%cPoE=7W>|=pUYoRXzaE0{8{tmB;!Q$Xw5aX2SRhAmKxta zb#WQZ+$|d}hrR(W%t^zzn7#I_CGQDv=>)eVxbHP<@cG!X1j zBR@;F{`krNpg(d>us=T24{mOM*!_tUj10@ZKhgMY@P%maq4p{+P&`(!R-W?(UHfaY z3Hd@87s~Tet~DUWSsVV4`HVBk8}YFAziE%%d-xx${h)8!#eb|_7UARHkKRpQ7sMb{ z`!ug#V(p#OPhjg$A$Pg7&pp6q55CpzLN8F zjlTw;(z!Zg_sUV`%(`RN;cq=FUVqckHGj3$k8kjTTxZuSw~_Eb_iqy(z}^AB^v7dw zksm+Xm|qv@5&5F@=f={jYNBha~*)g%`m~u@{s}(LbWe zjXR+g^`%N19^^dRd&ba5V>s8=&}&v^97?#aYSdR(P@nh&IR4u9)fMmuGPVQTY2(8e z$6uz-Y<+bqef4^4W?%LG3K)eief5PC=&Kd|^wmE)@?gRM@>#N3cE94Z>dd-9ZgUMb zd-8PcHD#X~JSjP>^#T08J^7+xQ9whU<2&y;M>OR7G59@ouXTIHviSB-Zu4~H<63+U zHgDe@p5&t)2a&An0@f#bML+m$ckr8I;%n0WP`iIYx>Wv7`90+$ZG#8gxS3yW0~hc% zF%E)TYeM)vr)b})Xab(%lPdD-D$ZAQVQuJH;VOS=g z_Hxz|GnE8yt)sm_jDfC8 zq8{gy_;H)WescB}GKKoah|bifxsj=hZ$;zB#S6SQJk0xIy$7xm=vNN?lK5;&(C3P+ z5|3()%E%HMSCS9RLx6cWxQ+s6aFCDg1I|duHl8~yAWBsuMdGZ9b*a-jV z`9ktl5udkO`#N4G;CKQ&b}sy7-jT2BJo?_0CtsCdv~_hryF01pJ3&1Qf_fHE&r8&E zI`y0v)U$wk&JF5WNIfQo44ThmZRbR6lbU|;59h<_B4>EE_!`Avw?m66^zp*acy9;~ z{fn=bz(cH8A4+3am+BHrH}>^T+JFoW?xj&_|bbSM>^k)ZE2KKxuhK_0PSojFVo z%i~AC)DMr3I+n+)e@U*%U~C*biE^K%oM@5FrjTtcIrJO$DP@yGS|cK!tIr=WfjIf2 z!_(`Jc3oS1Q8zcu7xmEkZs;0?Mv8}#pF0X~)Md+KX4gZJE56M1jNZXUGJ6(FlqZRJ zhN5hE0DDsicGuqa*AFw{jf`*DG7VqnsEhnPvz46DF)vaur5u=704F@;4mSw*HkRhMsA8UQC>}Vz3q4P(low;o)iQ zXN5N-(6X2_uHEteGV*qVVnMN+wAWdDR>HZ=CAXd$f9PvR*L?SSyt#+MOD- z+vV_Elaa^i{>xEc<&Bskc<+dMQ*6vi<#fh3eD*!ysSv+3jE`S>>pYW-KBqVm*P2w0 z7m1J1Z^DafO)n=Fi7S^uPE8~(eW&M=FZz|Op!b^7=d^lekGbMfQutjfSnvBDzO@w} zJL1dUKK|8w-^P#n_^LmSc^2UNVYff#-1U_u#vaVXcY`-#3g=6&VIRsHvV%;FW|n#@ zdA}j@{&TsU`LOoK#MwdrQpVfbuSQHP7rCC__#?Dk`_0Hr>p4?``}iNYui{sE>-ZKQ zWzG)lA;bUk4&QN4nD@2Tep!guw+DD+i>FYmrOO}M2lJC(a&`f7vzGn)%6}%iDTjaY zh0d|a7+(RcHU(|$3&xX~-&Ez8^P!rsG2gAq%{U+G-8L`&sOOD1^7bd2j?{T?ECgnq z)6#pD*JF96dOmRw(EB~{+0BNyomh;OWCV+Be^?nX5ZE=oCmXv z^I+;Z4My|63y zVN-T9H`!lTS!VFB$~7|Z7p!wuhQ@b9XSS@nH2&xfN7p>@lg#n`>o?P5S@x#JRc$~! ztpmI9+lK=BZO=l#ZIW)W`M8i8nG_*FmXnBf5OB)CBlq0XYLjI^eUYqOh8ByN1YIwg6-q#*6 zt(jEt@5=Pg0`CT5`CWMF^Gzj>=f%$f-U%Uizv#d#7(2*&FI}#4J8=b=KP#{2gy8Fw z*V9AtI<}zdLa@9Lz>;ua5kSX<2foVC9|1gr zMVrugF#2xjY4Lvtp47i(j)#wg;Q8(V@L2!V65@^B{92I<8n$2)d|W<~g?Im9XA8%VKslX7y8Bp5?mm2aeYT>NR^Gy6F%f~VONlHz;GbH$uI zb}IbfVdEf&E0Dw5m(`~y1AkNz{%v121bd5+#q)wQL%;s%BWonjCC8sa=H_w^%bkYT z^13T}d*{ZOrwCTyXC340_=;26Pegm)1@=3+-e}s+>rUdk%hvYRkt%myT=7f2<7Nrok8FG|dF=L1)kgvQj!%G*bL<2Yy0g?Rw6=@W`4ApoMZsHH`PRaPHj} z(Q@X9*cRzq>Bj|s+3;pw@w#_jxjk?6Y{vc`ui!P=YSKFwLXYo^@U{p>&9ycHD4Sh-O>Sv)O1S-@omj0{X|j&*k2=x!&*c$Lv=e{igZ;s*t^{ z^mvK9@1xWwyCG2>^Zj0ov9_DhfsB<7 z_QsN0I%IcTL>-HOLt~P5iXub1pMuBk zixf2e@JPk3`n*qdcj05&e+4$rEY6<3*6S&`?T0@u=B#h^e>3X^sd``l&T*G0-^%;X zC9cJ8!FEXkR~y$~)&3&v8PPrW*3$R)GERr)d2zuc*aTxAUi7{EuUgk|@m?Dl(tZAW zy>rK4pY2Dk?713xB7bFzmE86N15e^Qe~y2rU;-Y|zU04q*&Bi#35<<`@4fsf9nft( zFiA#KvwmF1cuhb@+4%ZxLoPgyTwK|Z6MuwxVJY^nWW#)9!+d;)?bpS6bjJIS4k2^6 z9*dmFL%vE*=xp%>d+?L&$FD|)_-BO=Y5r?-ozwg!e%)~+{d|NxOm?mv9`D=xi8rd` z%nIP1V)DGuU-;zu_9d2(hg$l`tcRnQZ`;|sM)UD{$pLhM#@x@yIUHlmX-&Ke{;Ogh zQwKj*@GqO*@KB%6PX15-o%_7|KQhT>2PWE6CscBZp4L)*^p%IA7T+Gha&Y?^a#-*LKdwSJ#ezZ8~_& z07vr-|C;UB7m@kp^k?9w+c_Wqn)c7?y9H@KU0*%(sb@hy^;|eWJ)guU?~l)5K1dsV z@^h4D?DN~gc#eu0zLK3}|K5$}XUuO_^UY5D-S2Mm;x|zK1L{gd&w>w6ZC*wHf57_> zXsfL>KiNnd_T(Bm8%F@~icC^x)juQ}H|VpFf51uKgiJ`0>%>^4rhn zSF(Al=CFSxO&{SkRxYl4e+BfEpQ3_s{@`=6 zFTb5(kFRGxa>?TS`03ZPo{j92EU(K${%+L^AAm2ZOfZY3cX}LT#@N+LnLG;M?Z6z24z**S{e9tIOb9)Mp#0 z{e#-?555uFEk@2t<`_Cx`R(2mYIk|iZntf>&TlvAv@3epv87*&f0yx}Wa57i{1f0W zyd~#F1Nq8p#lu|7XQjTlh?oasd&3j144M;y?~4I^vMXG=BoC+iq`!QPC(}RIvQFf} zF)jqh)IXtrPNw}Kq4x7b?VtW9w7)&1f5aanIG41Q{q5=>?FFg|{J8Za{If!G=$lIZ z+oFHBdEU^R9i_!Dn6qk-JBR8n_Rp%R_{ftlF zpOQH~Kf}4PG0t<#7@swM49gh$T{vCtT z$_@|v>m5Ij@dZbU{k2@H->hHX${;I$mSmClUq}9pW{$sXe8~STdjl9}1TemBVSLP& zceRd;KNzB?Qsn^^A zTy;L&SpJQ@f;<&mtn2mMimzYxfcy)t46|!$!aLvLrLeqIze``qE|M)G8%b@eUD+by zl~Vq7FMmSG#j)FR_(vuzO<;?NHfr0_$gL0GR7QR4v5DOJN_0(ux%8E&DCG4Xxr^eYximYbqVc8p@DMO7;5OFFmikKKD*Ket$eh9)By>{xigHmLGimo*uulPgCTpX=S}!_GVp- zymH2#iZ-udKG|R2-|e)Ov3_-i@K48I7UMtk8YadYuEr0>ylj0w=ZwW7TPpFXR$h6o z$wlTt2d#so&<`Iw2oB8I^qu`C$~xA3V$D{*TNW4H*71E?sqBB7OYYXq#aYT3dpIa3 zKZg7l3l^U@Z6kBz7Ie?`)H8GOf@z)j6?(^eH+h`B+dJ92sg)crixw}O)_XX5Qv>7b zPRcQl+BuK;@e9nO4nQl`Ej4%AH1B~*b8eBYYaYnyvG`6GzWA#wz7_HzUwIz*M!H++ zXW{w>#TeoPvhTup#Dy0 zyI}GBX)OV6(*xXU!Hsb^`ClB|Y9FfXDM7{=+#X8fHVoYK?3>{JF8uQz{Nk=XO#nu% zJ(;+pb=a)y@ZWh2531i?IzH;OHE)eW$2IA;-uYb67O-S%>y+cPW$|io+Nyoj!Ryf+ zKW?eYkNtuNRx1Z-gnT+Bts~-}{F(6$rk^#8h?j+*@!jO`GrqfsXR?(gEB}pr{8ODV zX~*M)(=z85@?Er#^LjR7?@H&?UGw*ER$cS#J5|^GivRcCsk?o~qjg?hW8LA3hw5^p zjn+3OAFvs#D}0-L{Nf4IULNYjJ4bl&-HaJFYtC))26az>w_i2LJSJ9z5{6cI&RY7cZVRK2p$J_~p;-DqOr|S~X``OdwWf!s2n$x+p(* z#+ADUFTS4Vh24eZs6L=eOG;fA6xUXfpbvzD%R^(@^0nXz~vP*=K@nMJS>>z;6EA*OznJ| z2TXa4=W1lF_OYDHcu-u1^iMRld<)l?#E{SJz^QMVph+p;khhEbX!myBd3;yAxOAG< z%>{?bw()%~c=>HcyJdrq~KQhd-w+U0giP?F;u>axM-p`m!sh`F`iR(Ol;u z?{k{(cCPO<*YbbnG~Z#am1{Szx!$?H-CP%%Z*Ft0|JJ#_*h z`bYk|MezJAczzZUO{7%SGeQ+L*LqU`DyU??Uz5=Ikd2` z^Hi=6S3cBv_?{h|1qF@rAAFB~TS8rPsA~>&&7rOd-x~zp%;oI}6;e_Ij@O3LERa!N#}R zjUE!eD9>pdeq?`Le=zz?`&kQ{bCD@48;axQ#PgV(w&VP@Z^h>;zqa(P^jUI}a{3K! ztn> zj(yR>9hV=z=bp9+_tSrbg ze8XAf@rBeile%V7*G%eq;`W)3QoiwJ-n~qHFCD&f$K&sQdzWOjPX}`4qifYKapH>- z@QnJ^*zWXm-(Ha_Y`Uqe<3V%TdM4f%emKn9i!Ud9dWXlHbjZXkbPM~33+ZEbe~5f& zTQ_Hosn}F<#-;I@3#>0R-LBdi5wsQd4^1e^tlzE`^%i+sq_4DgDO*Hqvz_pmd;o)4 zW7SzmI-8)09NOPmyl7ehHeDBXSHY4c)4Es>+Q;|hN1`{$Hd2`!@@wt6*^9r5oRxoU zFW(kT@NSZgT=a22R}*c#v4?m2kVOUX!EUc$X$1Tw^H-@Jy-@} zF{A7W)))5B&ccW{IRWh-I{_SfX5^T_X`y7s6p>R!ltQ(v9ixaY$I@Lf3kk`0d{O(*f8u+^N-IZotzbs}niyqt-#%>mT;N{` zk9CO07~dKbwl0l}`{7$Zzja>oeGW{+;Gylv`2JuLuNTAPqNn)=o|hj~ILV(A%k$p2 z>;!12eV~(0=)1wjMwoAC=wF-oDl_-`_U5mc=i=SG=IgjVp1q)Rp(J14Cy)O?I-nDM z)E4BL^7ym%^7b*3KWTlS_zdl*cnN!1doA{0OLU-vt$wiYC7ZN{y_bLGtyfNW&9jXR z=33*=aem*+-%-!GXG=Nv>1?`TaTH9*7Rxap3TM z_uTIdYaEVzyKnIg(+(`TVcO2TSlPiPU!V3yVXW-XlAEUO8Wt=2!;+h){c>ci?A;~b zoc62JVrB0w`I~9KJ~LMK-%I{>+AlsDD|>6n*QO268P@!EWMuP=Im4SD;QwjYxp@=?m7?|)_l=9pEJKVp7RCs`@lJ0F~3irGvEBa ze$Msg_ZQgPKj(YRHLIBetoRFW?ox2=jtmgk4H}jmK-P^D))hFZm zJ()iF1%2YuJR6P0FHe1RAvFG~q1{F2d?TRYtr;|IL*Hf4u(0(@(D26v!@Rg+Mn%Iq z<~w_lCj-&2R5V0(oGcAx6C}{7*=X1ejrT)u^?~~20CczgqPT> zBXb0I6>`V+{pE&F3JsqurJofOkP7<$UE;!Q|Ia(z`%HiE_|FSXId}Zo@4uq>wE&mv zoa--`YssCV&DU_PU+p`--1+7z=lV+L`#H|{pXB;@GE4I6XXN}JNM8NMkyWm|n!vbn z{ec_CH@sOgp?|;HW~UwblU0}g`s5$|;M#Zc6>D9*75OLMt?a!So3h$_4{8#)r%taBX| zyRv~bMCI_PB3@mwpjx+k%j{Fu`Yrj7=lc81luN%Bn)>JTFZsQn2c!di98)}omxC=XnV#4D!yw*5vV~lyht^R?g1MaHb>99jZ18JY z`zMyj^kzx{m}k~w+fTo;X;0Co zn)Z(SbkqD(y{3g{N17Io&1us5haHsq}xqF%HSf88d<1wG!=J2`PR^_Ej7I=2lyo2Iq$ctqpBWB*#> zYU(YI`TJs#F)iX@<)$WPvhVyGVA7t7NCB zI)g=fC8ootMn~NO?90$&@Y9|fnRm;+Q-9v9IXHDdqnD^-BYdv&Q|3^|gpwg~vuBli zd-lo2L46r@O=J#F{n(`R3-Mv1CG~3Vo~%^=DOV_ci0ykfF`=U&Pj?k7CSoJkIwWed{8#FC+!b8c+I8|LxSxo~|;@V_G$~dcCp)?I*!+`mWwJ z?Y*%wU6=69axd~4zi(DVys3{9mnGZUpQGd&Uia|ty7w9r?)`4&8`>YWn|X%DvGAh4 z9t7N}dROu}#yig{vC}j*M(R29!1a6^JikrPm2Y)7>#EN4EqX3_u=@=BhFAIDS5UX` z)p~S_z81dK0Sub!2!<56RR%EVUB#97M|f5T45G7Os19HVp5LbDf}uKqA$Yz;&jmv@ zFhnl*>zPO0^4VNZTcVHN-@vtKm$9}M^=Y_!tbdl??z1y#r2Ol6S(@LnFx~<}(hBVunWIYOG39jk@E^{ugh#aZO#D%9%lJIsmjkMtG8L04<3B&5Z}smL*PN38hT`Jdb7~AM zzCKW2Yrj?cd2F+A^!RQ;0KZ}}iz%=Be)-`Oz8~D;K&*?&rKj9fD!)4hE*GY8;e94P zwmfi;3FT>0`Etr9C-`{r-sEZu;}$EPmle14bBo)&4BTQ_aH}Csz}7F?h;}u_by@3A zKlke|<~x^mH32RAqn#^zB#Zt4zKVSnE!!L!z9Hg0yj6P{f_IYf!=!i7$Lep%5Xr;> z>D{S{!Haf_@9cOMKb{Z_S=wHqwuff6{U-a0tPYaAH!>U@Bzf@|`JrRz77zOeAHXvv zRvZ1_hW;z2Y~6TCjgVyu8M>G>9em$ z=c!{%Z|cia2ktg^LoCSM&_s;1E%#Ubm4hzfa*tBZZF3@JYlE0`+FV^?bD?Hw^IoUT zkM>tCOPk+z$_=MnMbO@Q#o*f9kXhRMx>GLFU%4#p&2!5A;qr7oI@6xb<(Z|uxlXx` z{>o)(ZzkpByLHdZ-2jeOPKXape=<+`@GI>9ds@B*#_~viul^LuOQ&VbQ${hjIGs7h zXyhq#mY@7*=A5N%6my8vnL~_b|IlH7j&JApg=TL~W_(T(+3e<{kPljIYQMu?^iQ_D z61N~H!g(do@yrGN{0;JDnRAv|yAS7;NU>+=1!B`TAoEf|Th&2ZRpgb>d||u%g*Akr9YS?7sKuQa0YXq046(@XTf(F@g&OK zOS$9m>8Jtt^b3wJ%Hz-StJWc0f0Wip?Al1%B*p=DR0ng>J?P?2ATwmY zXwIlTa0T$c&fk;ovI~8$eCd1ei_eUD4|memp3ksn416{Jke{pV3ckMz+^?3+HmG|a zvGS^8k13mr{o~g)Mdc}H_s6;A8-PJ}qV|}yu)pF{($UbQ;Qjhtd!dQWPu1GT`Ft}z z;x)G5&wYycwgk3IJFxk1vKAY_s&%TrWZ$maMmsdJZ5ueCm(knXd>_5yC-za_xzw*Z zwZ_$L>MOwR9MXIa^>u-N0-BG{^BM=LTkzcj9kgF#Ao;prfPU6Go9)j-QQu$L^l|xz zGX0h9&%sCA=*y?+)9-sdKk#yQtj+Tdd1Jk~9nZ6$%*!+PwZ`7hYyLmb@Mm6b<9+NQ z{9*2ujV-xXcHhUl*CT^=?8R>`dse>(=)bq=kHnLVffjh?M;Cotr3v8y`wGu>URR~)*|V=chzxnS_f9T$uky5qjGSknU)6;1bksD}YZ6n{OfK zNVvXib(K^80DZk1xSrxy^D+Ny;=<;Iefhuw|7CQf@%5hCjx36fGS?%yRzL55KHcwT zotQNY$(9yk(=H^g?OtM6(AV9CtYr-=@OE4{V%Ux!UKVTGG^3*FDf;EDmJv+{Um4kS z;PsC*i4Uvc!8z#j4&>1Rcc;1M-##f4CP20y;H0`*!vZ?d25lvm+AKBFXqmMMnm%WO$fi}j_M*P$p z@l&4*ojR71r1D-iapwn&~{Ff>Rd>jFXR7?nZ2H>j~ut9x$&ZY>e~h#{JD(f0r`D%nja#jM!#`> z6O092-yb}?mtXOO#=VsV@}KEJr=ZUdh(Brw`dM*wL*a>xu{uLzbp&H|WYhjv zM62BHRq(?e^pt!B_ZH-4`92uiulQrz^1}gmK|GOdY`=ZX*cL9QpreKxo><5D79ZRx zKDffysR{0-+jQoQ%M%OXiN)~5_3(u9m_7y%tfh}%GkifDCwMn77hRE?+c4?$Bl352qC?e(i{XpCk{8&i>3)>F$ZsB}emot%fHzX`gJSJ<{;d2!eff-Nb02N) zr_EQ#dbA&*{j}WpPKMkdW~nU_A2~q1%8{fsGvB82^PXtEN4&CvoZ1OwnXVhS?s|F_ z_{~P9&uN--q1SZf$0JRjzAUGyfN@t8VGO<7JGU^B53PvNX6`jOGKjd>!W|Pv6z-UK zNv!F@vWljQDk`CMK60Y~xglB&?w-Tg+dJCpxx8@jj`K$h-mx#B)lz76MUFSQD}z@4 z+WmjBH%((agI1#gT8%@dU4$$djXs`=eE^PE(Eg`r|0>#_c!>ur5n%D@wuE*jWYCRR zRp_?u1lv&kTcDvIUzOARaYHvhHu^8Q-uwMo;5-|g=QNF<;L(1B_H&v#;jvz1=4^`7;75(@rtHXE~|vU@}YkL^cUR<4c$76ydF#Au7GaW zgAaDcbmD}cueo9`b8r%BMQ2)aSDE^s=ZwlARVkJ0vE`gw4TRYb$Duevw z`!Dmbg(6K4eKx0QKRS3VWtUP`e(_0^tLA5yKi^uZT6h@u`ZYUwpQb z@sZzj=ym#gS7For#MxDQr|K<3 zfbSndCapn6f&YH`v+9reQ-q6T^IsXBReqA3=Begd^NyV6$4J$1Q)OQ~t)F{IZ~YeNg`9%<^S! z`CFXww*=)^1m*7v$}h_-Kfx_u@070($~Ok(9|+2?$Si-RTmCMm{9Sp$-Yn0=7^qLK zr%yIQ_da@?{JrUTP}U%}^w#Gz_0~t5Y&^B(glu-L?_N)A?PK`o>%LhWpTD?v+FCE? zrt2C1#GsfQnZ|d8oum1QV)@q-PxbNdVwtzIwu_Ac5#zvR_NiXxFivw^)R;$07K8rW0VdwYRBZ^@WxzXNXJlVbkd z!L!G}M>#X}OulI2ALaS0Hc#^k-!G&wm?CHRm*{IhSI4r=S2@>LI@jW7qicqETh!+V zbY{Si*p?Nvqp_wx+yC2*|1rIvNdCD9`+T*(x5Vt9_0M9}8dVkYcRF+R8SFcr$vef5 zw}baxl7!Q#`(dgE^D5!T>eMwA+2AOf`^;`CUK6R zyGAy3E&b7sozuYY9hCho@7A$Dpp9oW(59VlSl3pr_U@X1$FyE*^We-Fd;Go8e;weY zylp=RFWVnl#~2{L%a%Q?Hyawm)9S}5AJ2@JoEg-6BKciNwrt`h3umd#kM!kON!$X@ z7I@DnzEIau;;qQZiV0iL)Z1Ig|6u-yH1#g>jD29vjR@xSnxEDGn|ITC`X!e(;3&;0tuN`m?Tr z{=*(oTwyu)^IDfbvX(w*piafHZDX9Oyw+5SLDIojk+UmM>=f?$a65bTQpYw%a+^z6!ImP%Yrs}?8_DNjo z&CO$vb4!5d4C6@F<;caZoY5DudHv9#Ioys*EIZs z+?Q~#{a@wauns(#>&zX*p3>cviJXcr#o#xn`ySr0KlP10;PcNB*3x(%yBL{RC4UV$ z7nwr||HtVQ(Ny!*3*ohPRl9@+eVqZS8tBpQtbNWc} zsM@O+y(qJf#%x+)`bgtMePrY<{iA+TA1Q88{bccIy1<9ClYY93z42~834UOER?mb3 zy0ZIxaM*A@bO48!*@L?qnu*73-)6(%=L6uNzAPAjEDrJwS{!6w*me6b4!iko!?`{V zubd1Hr6-5OKOGN;l4EhuSg|;~uejBI`oZA9zBti4D-O`V8{QC)N$%-adO*LjZS>nd z)$7?xKWO~5us*w#qeDW8_2Tk_n!#1GhH5Ee*z8QNXV$Y_B$Ot6N+%tv=vae9iGTa4V*c z6k|4xTTXL{@M4@~8*}2<+03a1!s}b$m1K+wrXt4Z24u6~dJ!Dvvo~DwRr*-%CwWf% z*)yp++FL}7I`xyQa&GEc#wp*e<$gT(ZNLcc_sE}__ye&yjO`m=!7de?$V!{5IJq>| z^Ty|ot|<^+L7l3n*7(eVy}i`q>oi}czW00irE{Akpe zlKrtagSX@ti8QwoSJ9u`6C8KTwxC_Bo5H+xw#!?WLL=!X@z&e1GLu&n7<=LU6tHM7 z%uD1DaCz%R;PU-$gPJ#j{}1^0?+cpOn`?iM{CBz5d+n!_t^NqtqRr>UliZ80f6udj zGT-=q-nhB;-`&r(-urn0bDHmQ-rdEu-dmsF3a+irvGSYzq=x>l5Er-?o-2x=%NWm{ z=oiU~PWnN3c7f+!`XGPF!fBmnvv-WXQyw_8wrb$aZ(iEhRzdS(bDh_`FnBgU_?>5d z$-5i87k*#nT6+oV_%(Jrwz1agWgj2aUg&)E&k$tIkz?(5`DI=EOFr5N`^DB_3wYQ9 zNo)bv{*q2awmmZj-5Z4ujf`WzrtT#R;Zt)afPJ@I?>Hl%%J6TrJIY)pTh9L^_BCzp z^wy0yrcJkf%?r!2;Q6A1=XS@wT!pM~<0E=Udp*7N{vKP=3g6wP+DTt6tH9p%=c}>J zbv&y}OpRZ9d+eckhd;c#J{notPEI8GwZh*lJ6vtv*V4c94v)V3QEYsDyVCQf)=-Z} zJ*$qgSDz`97rQ8)80GcY=c082yEXw#inU1qm*A<1-ddJm53u0;V05e}7LE0IUiq&R z?EN*gi2C!5rN71R6MCjTUk)u=p#g37{aJh4j~7YjM>hH6ls^m}(RzT|p9eiQ5EGO` z{7uwL{kYD{X%sJ;d+^nL6kPFHKFhplvUqQNueB4?zJhY#NCAi0$Ml$=lWVJ!H%$5e zEPxiiLS&WOUy5hx?agT_CErb5?1Ffm=4Fbt1rFPkqfGhjvzOIGUq2&*4OKdsqHCtLbW zF!W33Q&U?8#s*-de$AnkM@I8##k9*Fo@4v0D&4Ny8XdIN<-hK7xnkwvK(+Asp8EgGTc|pD14$Z$_Ip|d2?EAYhbpN%`{SBe}F`PLR3d9V2mekJ*)xcBo9B42rL z?zbYpjlU2XRL=Ws`3L3u=|hGKa*rSMzV}J!v~089$yeci6TWZ2$LOt`7O%xG>wh0J z_XU`@0V3Pbr2+B}1Bj3ckpj zBsN0|S=@$9Xou$VyZ6z~%3|AB!F`yP4?xTCvwIx+CacJi4vkL>Xq>Eyl{E)+w)E`^ z@!v7@g|_X`I85I*XqyX-MN82j0i9Z)-#UJu@JjA^rPzC8vHaCu?ha`C2I~!nULJAR zBbJu0WT9pL0Cao>8rIX!I_SHHz7#EAftIYz^py0jeKR?H-8&`TpnJ;kjc9zkJf1OS!nb#_C^~y4UH6AVCkb+G*`bF9s}oUjZ5TI6}bJWG5=Z;ZKO)*&{L8>fVv&gJYC-V0AHN2jocyJ_C2=~|QBRCoIH z9u%B8Pg63&yc-g{8$R8{^_Jhu{!IHW8obM&?w`YfJiU4Y^AqfaN^BqbaKy8k%eeiR zg4Qw4_LW=?_p!AnME9|lKbdF$#Np35#ILye-DAE{%-I^L(n#6xXW?na3ijBcE5WY; zyT;e~Ma`qZZ8>8#QCd;fN}n72Mh7?)P47`YBp>#R;>=e$%jp8Mep&>*`f#2Ye1Fk& zlXIYmJ=RJ1%hWX~csF&r8LNGGTie3LH+XhMiZvtY-LFrHCx_H;UfRz5lkwfvc4W;? z-qrD+!S}I(_@|#s>l0(Q*7-8;H?Ifd8G5DaVr9<+64?D$?{^h;X*y7cRYM+Tyw z)@1(N^ot-A4Zj66ekYkp z8{QnxJ^=%l)phQ? zL%JHCU*y&w*v|4NstX6Izx|YkH%FX6{a2kp{mP;0&L`Tgz+V6ky=(EGF!xVlSEaDo zHMfXF-v5^l?52(He`VL8gAcrU-hcn=J5NMf-`VI5YRq9gbzsli{Ut9k7WHk*H@$Dv zB11GE(>k!9Thh;s^ia8fZWMO%gLPiRBZ+csg{yo&iuCRR?8GMmyJa1G{HN;HC(%P2 z@lWNw8Y=@XBMTM;vS1;yK=w_C&Qxo-D83L`@G`PM{tSC|6u#bP^XQU|t(V3-m}e!J zZ*?#?>3~ih=#>s=`ay0sSLcEqZ=1$?-)Y7bUBG-u34>k|FYo zU&FW*?QNMV{8u&jPxQTXms_TT`u9-3_20VhGGqz1>7@amKlMxGPL}qqJT`MT z2p!J2DIcLejWp}blALv{m(U;Z*6P3eBXJ<`gWicJ68JKb*!}V?D&AmT0p-vs(>PZ# zL4PIbFP|qe=e)B6IR7IgZ<>rfmp5p;&=-+D=p>Kd{$o;LXIP&em922W@un z?I7=aZ))vRaie)$8~Mxg4rKMYz$jc2$fx8~AGZ&#aAXfOPa=y_Q-i)(;m2uP{6&is z{l1+xI%q@vFMnI#`F25_f?Ik)u`vl`=SP9f%nv2&$|*AloWVik%heAXSU;)cjIwW! zX3xt-z{cvkTSCT~Gqtmn(8 zan24hT(Jta&T{&fe_`Lxwhcd(vPN#3U~14A1y*l*=CFFuk#RdWyVg7t%{4|FeF z?6s%df#|}z7h`3Q1^iO`he|V+=l!t~y3t3_Fm)3BblEQCcRzg?mf?~~pZfW+@_Sk^ z?j>9GXXVpC^7~1mu zesC}zi<*D*VK5%+=?ix}a(^=8aXR|kj7RUhxbls)-3T94MdR(jVa6kEn{kM25N{>n ztN#S|1pU~)g8puVukIrLfOgf_;%UK}xEk1zD}w!lo1O8-SkrpC*2I0;=dW9t0>*zYvHmwOj`GPfF5c_J*Q$c`^N#_v3f zF&=71b*dgSuLO=L=gD=}vo6n=e^v1Y8uN^4_03E4O^h`yJ3eh$eQ(F6aFLyt?VBeV z3;MQ#7=VU%d!NMsKQ&oR+sKqo@YLGBWQg`yYF|y0$Dj z`3U{|cvgOV3j7+77eAmK>5G~V{qbMhTiGKU@?`l&Wfb$4ItgC;Q_6fe7@l85TUP-i zboe9pDV;672!6N@4}Gv*S;H`skYLiQ$#4g)O>M ze$q!|-$28EAYZtC@8?&vEqD`|*3US}CezN~Y>Kc=+Ue}cI1zc)h&*v^J;}<{juV$> z=DfMk{(2*`TD10@Yvh*v-Sgz{E)IO%%ww)Id+$!(&;2;_m<`ylIq?1qfi7|FCGmII zUeY|q`b6$P7fA*r;agdJl0TXZ zf28C4dqr>LY!Vr4=h)WX{IDm`Z;Z7J{l@)C={MpLeEl{x-VS{=cPG9ou6>F|zpeFr z{YKwNzsYVI%UGyUPOj5|js7xoPwb~y6uVUOjMM24`ph3Met&3ws2nkV>`!p6@e2Rl z2guF}%F4!!p<{}vBgQyTZYQ-V`$g@3UON3K`E^gC-M+c817F76IDp{~J45r_;JM2e zgB-qi$gzn|gg@%C@W-34ot*rbL+)ocAGr9#@CEeH9uV zEMU#%CFH<|*wVRX%_iF2k+D`5HS0CG!Fr9}ZA6A>9vr0)wgobzE||Z}4}2lm3_I(X zzf7l%m4Porx?4P1g`UMj)ze15w`**RF9N@X%NxiYe5&Q~Huz6AHTgi}$^q1-c{cp# z!4tzrpDkpE)jV);M!p*=_lZA7fL~ zmYE;ID`rn%J-JoTKf=}OE8&}fcgV|J_QBG$TrzqCd!$cJNw1X0dfJ{RmuHpbo50={ z&z2*zs-PwHolIGyU$Cjoo*UpTmmSPmRLQCIkMt(|Tkd`4P z^Wx7A!(OCa#p%j^zK`*%IF*&y%!;#8P8wayrpAtDueogNLESaPdbOjQ<#S|AZmMIQ zMf(_JON$o%dUv0#c~JA^&8Ht(6D7}WJM+~J-Jl>ocL!x{x``}t##RYY69L+OhW+s!OFyl z_yt*L&=8`X>XHrFSB`Vx&I!tiUeW<=e}-{5>~-`q@N1lEjB2dLiT<*9&ILYR1Asr%tD9EJ{>uf4^FI6ZsR{M zf{&2cNXgB=`;a|*(O6F_c94Mqeb^eU{2^h40&2#>&{o(J&O~-qg`2{iRc773F zCxLdd4R$J?;}v5YWX5yk(~teda)j4OjJyxzzOzoUN$VstX(PN&l80>||D$qcWLuBO z(;7;wDEk@;ILRhxD`(Fn`vcmEaak+-Kstf?dU>z;!3=1)PWEUi{B!L5U{er_LpyWpl@3_L%u>mBNgJLrorpZzm^WBX$u|Je8ccr1T?`Xu?Q3pwEKzt;MRJ-?io ztxYy&%kWT@@6)dsmyRVB4xbHhP&}JiSE=)3hfn1E@}?~9Z&EzF@%0BW8bQ0Ip>`() z?J75l+inSQQ}34gXT`r2@VTYu$+VXrYVY(vp}h$Pzx4XXyg?0b-Y}y7^DK_CN5%3( zw(~4jnYD#l1@;A8Uhdbqm2j!Oq<=G2w z%8$w{|7ExQ=}!66gYqka^6Z5-!-%%Wc}2XFU>4}nOlCG zQ+`}fp12KDp12KDeqv_%3*GV;Ipr@h-|ehzYY^AI`fp z=*#jThru>q)=0jTF(zWcIJPP)b3CWxm*h*(9Pc-O&|D9hjov>rC{s>aeXqL5IOENY z$F=nr5)OLKyl z;3{7q3l_oQ1#sAS@EqER(1v_r(%TlEWA?-IO_Xn==tAv>Z^hs1#V-DTbwB)XCl0(H z{^#tImJG|bAAb8N(K8*tqPe=sApu?6nAgNuU&|#9`K7`4zjRzc^s9*`~_Qf zB|5eUpLYp;QVh}I#)Z__%4dsXHdd9=7J>!Ox3Jqz|V6N*oERP26;-X&r|cgFH?UpwsI2w zRSpi-mBdz_MO})=)cNYux$bM1@2F?z0`itP_3oTc{t~y|S#G`4)Ae4WdJ7_^-j8Lg zmwYCwHwhio&V2HkxbCZJcwJ>$&8vmtld0vo?H;8X3jvvDS$ z59CY^eHVidDvB9X&iE@d@d|1F74J4+t62XjHs`6x!%g!)G`e8Z0`zTrV9R8zw*~gN z_@CSZJ&)pdD>7?7xy=W#>*T}MnKs&|P^|1Ao|Bi=e7yA&b zO=0o;SwHf0f{}M=`yz=woB~GSXnk@P$GQN=`QX^rhoc`;WMB$pIO~;uY+7({<*)UP zZ4JTqdtkHjZjP~^(()Z&9Qhc|iElOgMhfC%p7L#O$ z+4&8M=i}V*Bz-aVeAWb!*(t`ce8#pvwTFeczttXVx3??}^oS$7`*@1>9|+;G+o7?A zXI04mD!=^k_-AY(>nQ)1xi^8Ysyg@n_sPk~AXsR%RU1MA0l})*0rJ}0oFt5DuUhGC z`_g+8=0x-=weJ<{kU+v1tsJ$b)wW^A9EVbhZPSS$3L+gC?$x%omz+4Bea)f{FAJe@hYs9+l3>%DrN#CP7T= z4-Ic{st%2U-{ba_KFofitid%rR_)nQ=~(}Kw!NQ3?nfoiNwNJme6(`-v-JA^Lmtks zmfJ6{=OYjQ`dp7ZbkffG%R{P>hceFGi$WLX*O<2(W$LYa$pmup{X5Pr5Bn3wkxU-O zn)>q|eOmQ?dWKpG=}h*_eR2uUw!(ut6GeF>U2lQ!-b9z~84=*z=_9WmK{ndhpMMYb zX&uiqmq)Cg7k1|dFAVQ8e5dn8e7bgwt7})Fx8i!1J`BI1D`XF3b2QKCzo%Sfuek~D z*+hM}ECDY+^XogKuX^iW?N!hcOx<6KkMgRK%aq7lcAp(jrxr1`jN13seu95vNBn$` zZYodcQ|A?m&+VV(cUNEb?q7Q18^i=J(TBCyX53!iS_>xo8zN7V;@r92)C!>&O6;11 z1idd^J;?`8)HnY6B=*PZDfX9W`1lEIqjUi6o3p}i4>fS`-jnMMY#u$l{>V$4wTN}j zLnm9{fw|KH1-s&D3O7s%8axN-s;#>b|Mb!pD>@d-FnM5INJFrfNf^-=zneh&5ixaweE| z$d{>|9M7kk&77KD>%5=$8{vETBXce>vGKOK#J{RHQ5%VE6LW5xmwRVHD>~c2;^yKG zXb#4?cx4yTwEFJ$hNgFNLm1g)Qq;p<4OYc`#yLv?R^vV;?B0j{Qy-s#kKHO~T zDs(+E=*SirU%D8&m5d;k!v6{ma^F)fV-D{%LwETfpXWWjqwmURq)>;dIk*(+ScAjC zweV!@cAi_WUt;Zqi(A9L-n}TeHL)D{9RUx=*mp$sEWSsG&Te@oGrV{5dutDUEnZ(@ z{E(DqlgJzXrF?zy`Rd)hb-8SfYMGUvx3Z^C%TI~#08)E{0$He9{6<#V5aE-dxwp zWAWY4*VdUp=L1@o)C)&O7fVO`&OG1|Vlns4yIzgO9a9um;0X-j5QKM5)zl#dgj1^p(bz0-sy` z8D!s$N+S483RPX!-`H&8pgPznponqzKOmK)UHTPA^%J|%U5`&R7d6o>?0l#X%NK6~k6^7y zz_G?w=ElEkW-&IGt}gyXSM@v4>({@3z*uC)Uw;#?&BC_{qBrJ*28Lxnqxr}sxQcOK z_fIQh@X37h4s0>H^<&}AD>-wG zcuKaIoWi!8p%=Pq+n`;*wX^K46KYr@`#b@<@V?e5KSYg*z8ibXy|}%l55K)_R6I@I z&BcYaxAKWSzpUE3XK8hbOP^i4dU!u}NvPpPAO9!DUnTq=g7-BCI@hUqe4ca1T<71n@8r7LT;I2ySYXjou5&WIxW%&vM*e^y+5R@u zjX&BGpC~UVn`UEa&!(-J=kdaPc;OX(?=>G8bjK||QXP!e>wWU12ZQ&U)dl{!!3X?! z83H}v+r;%Ve-HRts4E!W5BP5H2Yjo{`lBBFDPN}ZE)**nKC$D6-<5CG?@{V^WZ!=B z$L_U-#BV{T?x{-8hi^IJ6jVA7pKQjTH@Pr!56UyPqA%S1RGRGKlg9Ty5-O0-->V%ZJ!2fwL*egR1_m6?j=& zid>8i4@CZ&@B^BRZyXF~@I87TvWM?Fas+2ll{xpvDv_xX#^+^Ur%=OEWU=J|#)n-} zjQbtx67*eu&ZCxyI((#-YVtwAPdm=gGjD&Y{>&Wb;8W-Uo%1dkn@k&x)X2wpK81L} zwzmmb+GuZiC{Q38npopa)lx}khqiwvVAjO%XaR0xoB7n9y?SR0?_@*oR%pAKnxeKh zLyRMRwQwXKq7`3GxE1fH?bsU*=V7?HQ=`wy)4g@!S{L;2_95`_I>q;IgbK=lv59u1 zi>~&;ApIs-1jA1HCk8R!m+*ar)&y5w8h(jAH;St+4DVJwP2<4u@!Lel-nLvY3NP+B z1K}T0yPl|{WhkV@SP%Xz9a5k}BXm&i=oR@h$WIl1 zq+k=gKc-$-G-?Izo#-9q5u*GiKe8E~{VRS@8ocro^m-eq{z-Z)j zBlA?UnOm^2)-QE!q`MBz+UMXW&U_VZU#LKaar! zH^Gmc_=%D!+3>szYo1z$J@4jv(@8}Nv z$nVtkT-tSNJIJ8Nc<+mU{Z!UN)2`;O(Wg;yl_e|et&N53OSz(d5L zDV`rF+cR6Xhqk1rPcuI)9@FIPRU^eq=A%k^i3{ zH_7scI)BhVKNUro=Rv`wR6Fnt#;XpM76(#w=eyJLpF;^)fzx4fLuces5yUnZYag zsbbHRBKv&RnD8?^FTE<56q}f_dGuF5>d*4&uJP!>jLIH8_{D5j4}u539z-(GMrd?is!zeeWo(NDSFAjp64j%#rgNMC?gKD`>;Ms zUM+>S*joE5K4g938fW4SVfnYWGRM0M95K&aP|Du2vxvJ59-#lXc`wF1hjbeIL>$<> z?z(WD>IJY{8lP;YWGLIar{9E6eTuFDWI(@>mzgnXVpYQ^{zYGMEDLjV4_P0@zj7|? zrCY$O;c4ELoEn+i?Zs+*kMoLmZ&f#VhF)f!1$r!ckt04hAU-IiP4S7f|L~<*(>w)PXxad+2;ry!j#LfK!zb4#5MySanU#L{C@op8REO+Zx5H;N&{l zJL#QYGoP<(W9MXR=?DHjP)yy3at!L*KQ_^IBy@cgJ!198J`?}E zm)@n%*$<@0`$msn`vdwM5p>LNv+kW@>YVgk*}ds*UnM;`nGEf(m0&;W6Y>p`Z{*~~ zPaeY#1jw;7hA(V^_xQ{Q7FQp-ZLagyIAYJym6_qv%{{qjjji)%F3-05##f)9AK^dI zpx4~K;8)&8pUT@9o#@R;xcGDX7f**gomhsi9fMzs`5r@m_IeJzt$A|6w8w)(B$@1T;jBneM-?8;#*5+?H$~fq! z$epiPBmZe5bpE7A=OV_JLw_M`Lg?lYYt8ta$n)2t(?Z`oRC4dulUnbP3telzu=>!X zh2KA!`pzpSQ`dH!ta&i>bj|56K2dYJYX34PP@B&4Q+WR6s??{eonY-2a-ns^KfAUq zd#a2z9nOQHCxT7iIh1?yg_FD4_huIF{Dk_hIp3#W>~0ZuS8`AWjO&1L7%<94hp1s1 z0gNRVbG zyb`EccgSHsKi$7X_lZl)vsJUqv#P&Rdm-9WJWx0$v?x<;62~;@lkd_ojx*0ZuXPrM zSK0o`%{}=Sg%jO)$8qMHdxE>lx%GO+(!jAmzI&hd?p65iP4n*EY3})D#OiQcYckr$ zAIb%OB|#pKChTV%LPtx_H4Z~}JLAG1qxQ|%Ms&$Ycwr{~K(I;sIJ$e0eMIw<_eJM})A8p|0E5=Z!gzZ4P!o z{4d=f*VEp6lGWLs%&NvWS=+*eWUy6xwPfJyB$2;WiTgv!PxX!echt#f{IVyqLxL}b z@$0+%Sn1g7(qi>_2rK#y>>{bZ->nO+`Li=pz+*Ni*|;*%4zC9J_~HLLQ3o zL>WVjzWz*X_kMV{{3X0T=HXSkO>6J1ZqujLZ&}EKzuq2saOd`XcH$x0d8d!}D-!C@ z@F@;B4}Lhjp1R*_u*)w6;XUxK^^(?>-UU9Fux4IyXPw3LKqfOBugx>1pP_ zwdN;oOX0a*w$%Cw9{(ldN#W#1#-jN(*_i;bmgGeB6Iy#=J}r`vbN}}sHJdsQDg|98 zpVI%EAD{02Og;4q_I#yE{8jlus*TO#zYhLef{qH(Pagk#Ul`!KjSuAe=kf2x1*wsm zRBAlQ85zHyb%dH{p2cU%hpP@f#BYZ8Ee>{{ZHET=c)`Cvu^S(s4R5LDxS*~7Jg6wS z(_C>B#diyDF0iH=cy&fXF!Fg|DkC3f;AFfe9_Di?xrf{Mt%|jH<~Q(ae&cf~ z?cT_5>+CqewTX%O+({i-nU`mhY)Kz2<^Km+??3imYVRfVi{#*}`4z*zJ>$3I*+IQ_ zqu|Gfklr(J@VV2YGBh5G7%KUC4K4LF_vnn&Zd&7HcKgC`n z(*0HsT048QuP@nItvB<}MK0ki4ZB9?D8E_#nCK6y|7~vnFDBYJ`3(*<_%=DZ9-Dgh zo}kw5<$Ey6ANIqNDS-uO5GJfi(e`02@Dl25$2pD?9(FsUBP4^zMgli(GO z)u+EE*-tAX=ANR<%ol6z{4D8MBJo9=V z?1}qJu1eCs_)GnZmRtG(!;48^i1U{RLtp$gJxTuu$i4smFXztXd?TSx?bqucbHF#Z zuJvf=!uOx(Ph;KrK7Brr&}V#ooj1m)56^NFDLB+2)Gd3~uq#IKK*`1t$8S^V9*ebuLkw$Jv_$17eR|DF1j=?VUWu}|7Z3;X1+ zGea*QP@S1RwFaYx-)&5U(HO4yL^`!f`6lYlHnG<82G&f*b4u{J{NVpZ4Ts_f;|CLi zs1M>|vp1v3Ra19n=eQDc9>V8pZ>$F_KCOOwx5xg;KaR_5lJVVR_PQ{&SH0>E%-Fm! zs_tM}!Wdf;#<;MLF$NOGsM-dNQFYZlV^oc_S99;v7aHSj-WcT{`E}wf-xxLi^U+VT z2j9mhOvGt?y@h9w{=w^8zUJS<9zEU<7%of#Lx1IS6+WPUzj5*NQ05^u{$zFE|CRtx z5j0;_`QF-X{3r9nj5)ja{aLce<2Bc9VUMi!beDnOilu08St&> z1IAqmFsdHh594~@8B~H_d}QS%nLjbq^CvXt^VI}c68+Q-34P07@%O#UXJeD)6YG=s z6V{(n|I&kZy$=ja5@0wV`?AmU-&_C6pK63J@YlC_Yo%E8kQsiX?!C32VSkNed{UMG zL!5s%x;|ZB{?z;U>j%7?lv#If{q^r*^ZMhj-{8UY0sHG`!4!fRG zNv`qz^y4x1^Y_dD2j#z6fBnb(fMH(}7~=e8?V(T3#Q*)3??dP6AKMg3==<*>-`{&5 z7*-{~a6a1ifN*N2Wcx7>3@_6Nfo#%{K4rrasx#E?7^~U;lidVL& zPMWx7+xx^T+Z2-(x??<$7?b@c^XUGVEFoU;#@KhfGRhm{2OY1h_|yB_<8-g@zgN66 zzaKEP>Kr34KW}5E{>tb1#w*1M@F-sSmj~ZlTf=`cKlJAbu>DWOD<4e))^pB09WnH` z&pkrVFz<#xoMPtcvLhp(fmd$gJ!+U@{5H9aXEU80JE;|~Vt#FEV{t({^DISQWX)U{5+0bC^oVOl4ycwJ< zDW7vHhq>x%)yOnn5uVE$^U>^Avx4{X>0e_dr-c5>4i6Dc=}+^nept2lymCt$fm7Ex zz^eIe>Uh@VaK2Qu&@uJaqQ%3aTOlxg&Z}K2ao6u@-qyCcp4yKmX={|)Vf?C3Nn@<0 zJtsFT-|tl?V|C{72M=fFIN??0<)>D%?|z|^)ma5Uaz=VMi?gAdrZR6{O5e;UHnDHE z=%c!4@nzFgXhJ)IG`{nxXU3@}4%+DVFbS6Hofz5aQN)vIj4l< zE9mQH@O_u+TmpkTr?Z#;&EWiz`|mnc7Z}*-kYmi|b1XG=&DS_xZw(z-yAYTkq0hHw zI$fRR3r}^TKOP15&li0n{KAc@gkA@O1Z%8`FqaodKOMkyq(7W61)( z0p2>$(Dq5>`)Zsq;W5-oHD4p%6W@j6eAmf%-PaRnMS>?aAnRGOAFVIJewbuP{SY|h~4hi zFpTu^gzkxd#T(5N;9K$U4SZ6|p!Lp%zE!zl)ncCr3|)QoA0Iq?^++dtxMI<%gMkZc zD^mt`R)L4N!O1YuH#itv42)DjTPuIUfmt7!fQO^>eT+VZ3yKR34xXW(ap}y_82S&S zt;|RSc-?jfMYeK%mQKenZ&5u0I&7-;0ZFhsT8H;b_SMw@hLznc2&o_h2``0zM2KGF$uk>`1|Gn%!ol6#$Ra}1rP`r9b{*9P3$FLz{WT5T+w zbxUUN8Lh)jiOk&u{~+6{DPO{W4Q;6&S~ZwUso%OrW*s`)k(qTKcYoN35;Vr;X>cI2XiUFB_W*u1kPv zM5tidV9rc~rjrMgtCtQmYlhsj+jh+DUaR$`bNL;a**_N?rKVF)41MMScWXXt3ZUf{ z=+KU?>TuF)*Kxm;b4rTQJ8jTG{MyVOGn>41$obGvYc|aD^wq&LqO;C13i5swy&&4? z42~JlQ1uh4FRBKIs#B=;@TeNWImmGgx{3x@gHK({22}%t>X4U$Q=`YB`xmI^(mq>i zoBgKN#u!&I&*^N062>q|yu=vRL)Q@ZE8!)LL3=A`|1qt%2bBB{+ zaDD`wZvftQU~zGt5{EYi|LgZfc(1%-W82Y;N4m9ZV9}b>+$Yn z;Sk=nIBW(F<&0gtD?HBj;ZeA=bYBb(^{ntH92#7DwAcEYRN63j1&5;lCgG81`@_FW zz~j02x1PE%@Bc>9g$ALZrEy>Gbt zZB6FhS$X9WcP-&R+x|*856UMa@%}z0{oZr_h}J?y@dZ=YIpMUtP(d0sUTN6uot&Te zX<*xs=Y&g#22R|Gtz4Q}b?8IDB6-mIf+};?xodA%Qfm~&_t82%tsQ8Q-NZkUuF`w0 zj0L+DzmCB}I)9`USxCPsRFF>nTN`IXr>CqtS%?j-pg+YJ+7nWCX2xpP`p|w%G6H|e zPOE-beaHsuS#DU|jJ_R4Ply=5`2tCJnu`AfaVwQ92`XJ{bnmXyzfkhfqGw~LN zaCqVAZ0~#JcGXFzJ5L`P?WFHN!u@id(|g(*UoZmSLh~-`L=@Bb;CAmbR>7oqrZ~a< zuNqh~JLljxYcK41=;6&y%6`tO3$uUS{`51;S65-rRCDa^`w@r?pUs*%a9Ne<9E>t% z{Gv{s4bsY3w4d86!)x##XFk{%I$jPC+|gg5b!a6q32b52h> zEL~L?a!%x;zl5v&L7~UXsXfii$`5B$ESOj+KL&rlfcNzK0)A`H%n!Gh&oaMd;iK91 z6f3pk3)$~G@!{~d6}KEzJ7e%|a(Lzy^F6Jz0{>_Md-$vycG>N_$4M+*j4m zitRHo7Hex|%R3}j*mJ?A_6-c=5VGih6?4n8_)mdGG1iTXM)LCq>8wxIgUg4WZRz3p zxmi3*Ugi2M>}myTW;C|3`p$d!(-Q`a4;7FI?I9v;=c zrP$^f$Drlg{HFUw7rWp8OW()z{Qio*hxnbnxb?jFb6~~ziQ@a_;Gj6)kAvrrI0HND z*+1Y;c=Orvg%e8ygF2sq?>F<#5&Sy&!wdQCIrzR*u`m1M4yqV8QEe^+&bQ!ueXk5~ zjtB7F92mI2kT%A{^I7OM7EavR=qzfY58KXk=@aiA@qn&di61|#_Zj~#-ao=;J$hrl z={vns=Lc*Kq30_`$NN4?ZpF4EeQn#3zP8tOpA0OTMSGjTMQI?l^Ci*_x@T|2a_@t$Xa_>9fSd>X$G<@Ypb-s<~(IHReLNqna~!mGg4%02@nfzu>WKZ?6N$YATH(pKj_`QWJ)>$s z6ZbK1Lj)#WkBSY=K8_yip5^@J+s$bD=r@p zdD(Y=m+$)LzGr@7u4NO_BR@6QgUq`x#-B-#{D5m$zCRvb&svTga+R;FeQ)irb~)h- z;L9jEg=To75c#M++U@f4H@|u4k#9DfeB_(I<-hCXqYv$R`uQ^@hhNAVT>FXC3-*i8 zGm(KN-V?0fq>u5Lp~oMgP4Rlp=@R6KHsR6cHn&Z^)8TvP5xwL1-zjq6K^96{lfP5R zJAVCY>Kr*oxh3CeXeHLyxz^ITmu>6Oxh5XqOy!n*^w>q{F?^5pXwC+-_OuXlhYmHqAx^`uqTll(sAUcjN&e<)`xeL?+?>yLVPlI&}p>8HWdsPef6H|X^l%kXz}e+cg=Z_^BZ zdT^Z@dFbaesiL;MQN?W#V(}=RNq?@+IT@i((glNX__Je)m9dGK8^s`2F{U{)AuE$5?)x zK2rT%9G?q_-o(4N!`r21kAoA-hAw?76MU}(-wmTz?zIy?*g zm?ayEjT!KlhYuG9oG(qja2fluWY)^ZO9RLH4TcWe-;_UYpX2t`f#2WqIbgp9dICc$ zG9WmNZN?XuZQe=#`8Dt}0K2^!yPXqA+y5c$w?SNMa-z3DpWC6&=NKcr+g3`vI|u&( zzaoqH`0^~);0N$o((n&}XtqFs54`1mJ|L3cBnLvWpcr9vG1O%Z{(v4Kxb8M zuOYsusTvX9OkGr8{k&CKRG|FlQTXE+^zy@Y1-A8%*u;OUxMJd6fx)%Ur8xz%O}h6M zyeoe`1z!ZYy0D#EA=D+<`Wz27{B3OD-Hc~2_&o06b06>spTG3* zNnUHl@pItww)cVII5-s?>Z>nUt^zmzR&f=$8ESANe@M>@zxv$BJV2c@sB;y*#a-oh z8h@#j@Acs63chc^z-wBLysRQzh`2%Gr038`p1l={4V^htj^eT7wNju%^jNkSs2T;dDM)I3zGlicjXIjBG)F~ zNkIlRr=@jJ5#%(8te~5&FLKiMPfg$_HxHi?c}4Xn!J*jJjL1voT6R4pvNL|28u_WY zmJg8;Y4BbDfa|mL9J)ugrWKv4+>-KovXi&JaJH^D_62>~hCWkn>UZ9mSyn%}`(5g^pfIjVwmBX@ZlJ6pnWi{F5kQq9l<$jaalw+n3x8iFH2|6nLMvM zcQaY>EH+aykoGY!xq`|B-HX2YJh+tIIQ!k~NIU+KefRYw?@m994uNov_KeNEh8zp$ zPu_$M14h{x<>|g~6?VwjS?n!+A5D0_nRB0IpRPF7J!5S7oQZ{jltr>{6~?{=J1+pv zX7qz}ti_4!^BZ5LUFl5Z+}JnTmadfUl0MBuzQ@FETQE{@Y+Eo=l#J%k_f7WJ_uKNv zc%D?g`HRyNUEsHG+8;%^o>uZ0X(rc_*14JMja5sYsw5`@jb{YcIR&Ca6xJ3F3p9v7(itcFxw2M4K0bU1?+Ncs@bk2L*jtY| z{EF*{{Zc~(J9jwlzGKfhi`GMIrzhb7mm?0&Q+0)@!*hP_UDD&%g+50s7S# z_b?8PXE~pKeDv_7{N=RBbm-Y~Uy;F0JPzFL&eQ9CwIf{Z%DrO)f}0|v<3UHm-&`gnQuX!!}vi3uW_#%KW-_q zsIj*#bh?fj`#b=;XX5YZTm|DJ9FV_L@$L84YR|Y0t_@lr^yB|LBIDh{b{ z_*OC4uel~Z2_MB4sy^k@_+`!5&US2O%e{eu4t$tF*o(Rk2h1KEbq(w*gq(Chdz}NJ z`;w(=;rqIxz@icOMs;69me7%ssYZR4A)|cWOkVk|;Gq3?o_=KY6ssd@2x$Gy?YCs9S6tq@#H&4fxC02TOU#u$Rf|GdW1od zGpdP$?&3AsO7WI-f_SFoUic3GsvWzgHdT|*IWyGtbB$-0t0VAJ57zB+7QMyZd^Lg8 zNS)(+Y3qe6o>JWTe>0A{KE|Q?7CRRIIQIWH#_8 z?WPb%i--K}mf3baysKTEr8qheAV-mI<|U<*lsl5YV&(8N9?tZ<&OeJH+tORg7fUw< zoj|Q>P~3U-tjNRQOz^wUxjgWFe8wimILycUL(O$)`vrVvdpYu^fzFxe5Zcu~g%=6_ zk4Ii$KmMi+`WzVIH`>ds;y2bfdw!_jo|^sputR?TWu4AXW39|RzoCXqdq+8+GJ7m9 zAwO97YDz)jRRb5Rf9ihL%tmi+KyO}0T}uwSVdMx`Z|00}*A^D$A*bvqT6j9GUVfTC zwwFI1Lw6NBDdAP@*{n7N|HHuV_g_|-d`VCJzv^BpL!lF&hJM+noqEBc`h#lV2%@(y z0G3=}$yWRT?X%I*!MVIM!mYo`esEClcc{M#*mu6@*2vj5gPh}M_Tlh;53*l`ViWez zaO>%)L&|2)+3eFP_1R?(XRMJQ1YepLvUvRC8{VE(KA%nFoaE8T=1Wq>1ILJ>0=qv* z%2j;N>-k>zh;!l>jJdGNIT6w)-=okZa8*$OI*z%GkckZg> zHa##t;n`32JgalYnvl1`M(0EspVGOejfrj>&KU30W1zVIYv|K6|02DhIhGhYA`Mz0 z%Uye&l-f*av(w4wOaW#+qc}>jhT<>T4l6&Z*I4p2bt&2IJc;(mZBW1C16!YxJh- zJ6dE9ZwYn%;=|N|LMI2GDw>a;fqvRcM0?w`B5zIj!bYcHGn>g*uG3!e@JST>sIEx4 z&Q7%t{iEO~HqNUjt@h?L6KL&^_a(2NPZ+E6j{aExW#8C^gE&8!_^@aEcFY=oGk6d_ zzAddGZJ-|bz2w2<6KJ!QcT~fsGu|4Ft<_%G z!N?|bxcWBm@t)yTg&k`jy3N+8ey5M| z{h|;4R;_`5{$09Xv#m^c)@SEQZK7bq8`_#wq z@x2MKCfiT3KluNd;7q`O-UopHU#s2o#Q&!f#$@$OU-(x(*I#ecmweU;7U|X!Xn|g7 zz6pIOT6_TV__c4IzdSao=8(RZ3$F8&N9nM4v@W9$Ic)IB-&p(&zkh4!jC{PazI*MY zp`pjK@jL5obDwnRVRKa%vHcz(u-YU+nlhIUq%n%|G~%t_5e zK1$*B&yfQd;_R?~67@Xc3}D!)_yV~ydO|uD8{y{H+}h)2H~ulY@Op6H`Z+h3aS8i( z*j&b~)XgYf-$;A%S=Zs`Xg!E>ql!z~WiyqRfEUv7kzXb*Vt#jfYkA(pYG>L6_PYC1 z@rZ%pbbM%kj`f1)$zkwr@slB7a?iYr(Z8+R)xMrn=uao%oLkqe=x?H}X@^wwv1l8y(ZffOy%^|;RYLk@96YQyxE9JlNp7I{mj6phP zDLEbCVIDETJmoOpca33?>WH`&POR@&1x{vxbMb)UOZldL+#dNxa@=+?Kc1K;(wyuP zU?8SE7z2+fo~V-sFBaxRU<$u9Nl;iHMM?It%= zv-iVoJ`>mW4VS&Wz1`PxIkHkvU?zve)9#_st9E*wqV zOXTI8B)oiU5`Ay^dF*(}R1ZIkkE7qGh5)`y`$bBD#;EabVQy5o6;CT>-ariEj#v3* z#m#08((plzH)iprIX9efN55ay@JwFKp^oxVi^ZdMtfK8^#=4fV<}+5+vC5avV65t| zkUaYsXh}UBd4n_)tGn`}HIXI{X!=~BKCLWx{cVAMQSt;PFK=QK@mqT2&d+!EdN$1m zkCnN5J$OXl(H=bOef(7{dFaVS8W^VZ!DD3}k15}&wj{q+W<~c=z?n{tqm3Frtr=K0 z(Ag0n|J{nbNMA_LFF}`CISwKNif^jCHzvyxqREkbhH{Rr^)CA-bM+lt(_p~Dho zH$)w6Gx93_4l*9ac9y<#p)YkE*R`UrB)h6d(e)^3ZE;Ipr=xWun7N1@ZE0Eg9=MpxpyHsK@or7s&8*I776udAj- zI2XQ!_e7jGqXXX$=WQO&8>zFLIDd}zq%-~eEPXE< zB%J4i^Jl>MtE|}xUJ4z+tM)#98N6$K^B1)a5Zvd0d%rG~yo)E!rBCk?Oz@ZKk2dzA zPgkN($DmJNMW42!Pg@i8snszL1GjKsb&POe_1k(67y3>8&o=Vin+N}?4}6q+_}IX^ z@wo;MABty$lN@j|5*gosEy+f&`t6C)u1341X^^Ofvx9ln-*2c;tvQ!K;(cKdS7S-++)>fd7*x&3|3_D2nT>>}#IeE!x^ zaF7@eWRRn3&*p5Qs{#d?^c4#}$Xr}@M6{jE{=J)T&JT}fe_rK-2C%i4I>hYwh}hIv83$@e1QwEbva)8B_X|MJlGAM^fI3H{V0 zz*Ed#$=i5twEB6Oc@SdMO86nbI5Wr}QZs!0Qs(Sp$n|RKv@Yd2@re5HXcixv+WIza zsiwb@ITUo`49#1-gI}+CdJC`KZ5i^=c0bG-?^k(OwJr9!&)9jOU_xF-G|zVV?=P%f zjr;!4y*az8LA520?=MdOg!;r%=y@~aEWMcZ3E;JKI52zPi@<}0<0Wv`%slCA;K>8N z(o3O(p7Z#D&W4wMgSONLywRx7Cj*Pd@r-Rl?`u3DZ+PjqcpuzI?#lEmxZ35DzaVQF zrt)kS&q@Y2fj`+k+g1#C^vtJ{;HTF-0I|{Y9)A3D06*;mKdYEy^{+?NSR@ydhX)Ss z%ybsj3s#L;pZqq0&;Jcvlb`f{w`0;LW158w$zO;)?KI$roaq(YkbiDxURFBW%?GD< z7Q^GJe`zXl&bayDUNO#^8uC47zJ~rL9@yaex(BSkbrJP9XT>;AS2f`0pOFt~>g*Nc z$fluBb59r5&N^N6gz^)+(XFZI)>P(c|BtCf3~g7;5gZ>VpeE&v$*n zi%+}6^J&XGou2|eT4>MnWevOsf!E~l@NY(4vFuPt9)H(?MS+pfz zy$svZ#5{8uFj!qY5W#X#dH0L*T;<&1YWd6(gpzX&V} zNq52PIp8OVPN_zxNQeBKwm!l$y1s$u6ZJ+LwXVJOMnYfM3iVaPyMDbfhVKSXiMC=0 zwDHdIYG=J`TyG4)S7L7L$3Fa=2)g@8sHWq+A;Wv(^@aFT4?vp*_`{t}Lml{ej`7tA zUyQF#I$GbkU-S_dZ_M+Bn~C$A<^;lFe1Lep)67uUL*U@B51(3Nuk#!IxODfEm+zPz zna4>uBes^5)n zJZ}2@1AS_YwyjO%t_DD z&Iep%f{P65RLb!|##H39t}Qzf5|5$7#moMj&l2j!=fER#ILE5_7Wn6OWa<{^%{d?Q zzZfdmOCGJmxqtf}=K&>#AQiq>rEVFXmo2IG`gb?8X0>5UfO3C zu^!E>Gcx=St+Jri7-)5rnzHE?VLjC&&`WXp0Y`OO&7qQtF^h$(3{U^4E?^@vIGAx>j$Eid zK)FzA>AHS(o;duy;7Fo({>nVQWbS=&h|UoXKTE%fIXKqoHrRP4<+lHk_0_%X6t=tt zTi$vJc*qWQeU=zc_T1*;a<^*_Y-ZyqaKa?xp0#U7)MSPe#!NG!I#T(Cu&|2 zzkJ7fH>YzzIh`1~pWNbb{K76{$Kk`9!H2c!Zf+;#tlUmwF5Kk6y?l5=ZpZMtY-%xM zI4ied*za8}ZwU+>!`Y?77 zJdaLWMvg3ZaR3}zn{R#dtLP)K{bump%-Yt`gOry|?_3IxM3rBZkB(kxX1%xdMI6x* zdlAnCPtaP{K<%qnJbLJv@|?vhSsU4rshZVz-`i~87bfW&yGvXp8x-39pIT!B{Qc3P zbGX>&xDGo-jr_Kpd)z$-#=PCV_CkK!hk)J0-klgw6&_k9;e_`8;`? z%`}I6R|=o_k{_!3*tReSLzHsec3-|xhqXmyr9CBBu2P@;NwY`tn>6o3cRPeCHZ-l#&bktp4VD! z?SEtETvE$N$LCz6FFgIS&Bk&6$K!$Zw5>Q!ewQ1wxpXivyjona=Tc`vRQ?mW-7R_W zv-I4b(@wS5xiRt1Jja~Dy_0%5tM~SLGWm>wJHynT3htiTDdvJa9=*l~zhYn23YoeX ze6@P!QL-vVFW%~d-^$Bm;2#qNCi>7^18Wq_JWBK7ss^oDD3_f^|Jpem8>h?Pv3(Dx zZ#{4DN53ZaMwdqEdoTJmh>xkr$NXnOtg)-)3#(FNzD6>P*wO zsf+zRF+9AY`K4&Cvtu^rbQCv!B3#0Lih1m(I1fDJ2b|&G*}{2)Z--oXOEOG8K3eEB zq%hYPtA%d7-;BR9m-lNzSB97I{(Rn-to#*ugMVI3LC?UK*PC}%xbNhC#CxZ)_dBoi zP9b>=twE8G7Xn6ocleyB-^f9r-)6{PE7K=-sfZdBwQXSGyvSHJ>#~7i8MVZls|v$g zfN3xAe1i8mmmVGuZI1~bz^?k*?#hqVcUN9g-`%Wzo5)3kt_!yuhBx2f**C%e5$wtC zKtrrLJsi7_`|JVv&L(Pxi7kY~mM7SMj<%X=%^drIoF;s>x4}zcRYMi?ZJ$np!)ZzG ziAKdq?yuC|k!hK=`>4={TRx@$cCDV=>ItLYmVH)mf0 zM-(_@vw~@%uGB2#8XS~>r%L?rS)A=~7CyQ^#L~23_1#-7KCTS6tYzF!gO57!u@QXS z1U@#VhGSQ9pIrGi4jPNI}l!V9y(1|< zbzcPjF0SwXwT1DTaLZ1?0*rfsaX&CV9cYL)r-h@X+{a%k^I*)u)#Es$~omqN5ZM|UILN;t$e?woXtu8*vQSC@MP;Ay{??N{pN&lJu6a9~O@&A{E{VIh2j_;g5{%w9^`sK;<8v*u%m0u=55nFlf(_7trk2A0- z%7121Ly$j^PwEKrhY%k?;2 zK6b3V{IiES-+fyRb5Tz(UT~@`nEkrzmwbY~MN>O-$Z-_04pF|_Nap7{(DhUC$=k`7 zYVXk;_G94puBq_;H2fXq-8FwZN^``MvcqlgY%{sZBdnwRE$_N*=b5&P*kfsu6GC^Z zjyXGWBkdoEKQq$ppE(Dfx!*oh#50;NaN8MWo(ZrAlVFfc7w|iAmG}GQ?(ftn7CWJz zm86D#Hmh;u$3y%_*}qz2`U}s0KmPn!^ZXL#AKdn`Bcu5J+4%2QxWBX3GvqLj=C$!H zo?)$!*Jr+YW)}N2S(t|N`&04Xv)$iW_c@9B1b1Aa?E`rx9DgR)JX6H{KF@ggq3(`n zR=Up|s4@75KFo1iSl>mjRmFcFVcNjw75=<_vLh#W=DzqdB8G-wcH}J`?|awEH`LmmSy3^rQNM`P3opsSl$zrCIOz_BYeSK=(n53Drv4cg;Ca}L@s z-N0gU)!OR>c>^Bh48W0l-n0i#^crM+0)GBAZVqBg!hFV2;>IHha~f(#>!{s1uq<~@ zW8j0^pH9fNL~o$&LU2s%v&5~%Tces<`GS&V&Fz_a!V;&eS52PvF^JJl9=EkPF8-vm z%J922x2k-Q=279_H8zj*^<%bZsM8g zzayTh&g>P>L@x=j&LYs&bWwmcIw=jU;{xoZ&v@mV){)1+H@|{>3~?iV?gJ-xCCn)) zcTl?nUc0F1EMCj$i`Q;i+OP%Nf{!twiT9#)@Y+Ur?LK^Jerp|90eq@kyVj8e$USbyfJ(+ihAuqb0^D_JT zVGmltqiO_>u!kygrF-}5p7KjahR5$A@9Ac5!Xp>6SK&);TdvG%FNoC0ee}tq(cwq( zpsV`4w5Lzi_Z%5v+hUJ9+WL_0@odh|)n~}`sl6|bjEO&I`pne5@jY!x2OhbK??3g! zARU+*`Db7l=HkI;TT&wjxIdWtJvKBovY&gI+_QFCHd?mYwXvy@f!F}Yd#A^?q(;)X z|89zMZ$0xllfKrycJ>hRqtI0IEy%$2O~_FrdTJeWAB{T40$n8;Gr4o*O!K!*d>_pE zh!j)L8i<&fh&KAF|0h>Ld_^9}s2*FkL9wgW0ZMKp6U__H zp7Xd;a>096E@H@tmnSoJ%jC&CxftY=i)HYX>cu4&{<;4cyyWNoBaig7pXQtMPxt2h zH!|nH1bMW5TE0I)ok^nnu{pfAs)xlhw44Q#53ve!Th1l^&8;cC_?P{VjF4+uuH2Ba1qz@Io554?ab?=KJ3 zs{Y5*jeZ`8!uPYBY1g;HwnJ0$7s7+BFF~1>ilYrq|e0tq~lTW86 z;ZxUFb8Tlj_ACS6v16!f&$7UW*^`~Iw?osO`V`^eVa){?JX|Pvd^Wiq9#-v%-S6IS zyCYXTa%jN6J$v%Ha<}2r7Yw*k4kSqT_HInyzj6wU903CVF^?osUi>zG$lCAKE3;saG=Qvvg`D zd~pN3f=!iPO=efWfUWGWPQ6~WDoJ!|D`VdYo#bx{f1|-E`t!we>9HX2+J4W+-hSAV z2fx4T*JJ8`KKw7eJqH>KF8Oyp{k^RQ9%)|PBX57pb=Ok1GmdVa--90i8*}7-xzl~w zSEI{`F*T>BzNE|dpvxDa%QL7ml+PsjR6SY{ezCPs-u_i_osdKi7`T3x1TNW?Hxl3x zEQ#@ucrdOL&`GiK{w)EQ@NO$Uq4+ql7N#DXkt|*kO>Ml?>iLDs!J)2I ztFqkFr^}3uAr~*YK~JmuB-7R=w!=@VVLw8Cuh_)4?%cC-8lF#J#_>;m(5~F0o!=)& z>`U|ank%;YB?G={vwX$=;<9t9Cs93d3ZKpP96zhSf#2vZc<-nD)&l&?JRi{S|H%%d z5?8m(G`wqkehW+6OzD>%d-%dP(0@FajlLXA-my|~K6xq|M=WPAeOtS*i`oV2<0=N0 zkE?nj?3=+&2DnjM*3Ye>hL2i{#p@ZTY@X;fm$-csynd;NuT9|GuGv%GDBc(2vVEms z8~ijlzUpkbmi%NSX;bj1|2*()_V8m3pl~Y~H?d#NT=HC-(CK-9$>@fR|ZA>rt%kK2>JGp7$ryXbLnYTYxe`bzz@Lu$w=9*(~z_$YfUC}q; z-6L9??d%Zz7FP?HtI3YX9~o{P)?@U26MaG>)z4ZQeFxk%-xM-&i1;!37;W=RJF=(w z$W+==9#HmYKmGgVPIa}4jYP+J(9s`{J;NCZ$?V1~qo;e#{~KO;un!u3opz*?{d{5P z|3^W?DXb|>XN|1RJIzE$DB-DL%5YZ0hCsZe=v?!}4tVEl$@F z?2>XP)9_&u{opSV_b1b9vJ>s}Y4uHZAfxjN>Ug`6@4@dTuYq1_^qyTy;MPA|z8TsF z&Mr?vujJ!U%;42QGk(FGSO*;exAB}CxDrpNK?}vyv*jO}7|PTF86M)m6%$ii-C|>m zv?O#i{v_|)wV3`o=&j6AJ{LYW4Y!dCZn@pm?}V){B>wZ$QTu_5PNSida#sUI8|;no z6?j+A>s|aNqeqTN=Ry z58rC^F>%@AmGJa(PmV3lendTt;xcP9SHb%_>v1VCy-z)lp~cERXt9BICC`4DvhwVg zt0~Z+c@Da2fx{d8*aHq@j#%HeHCX!KZl-^tYJePOi?&AJ$#!%3U6` z&|GsJF^lq(qsVb;3~jo{v+JP|_wLs{?XO?w))tX74oD9vhkWg)!U1R#g^vQzP;pVCVj}07@UY7F z*6O)@^wl(P?RB)Bv34@nw>{k5O}z+nPi>65v5|9vN<&>QfuraI?a`Xjd7Sw-&6PH7 zBu2k9RG*nwR6k&RaeZ6%CG{Qom)0M<^0N9;`3TI5nV4Ad5c{gz*jVcU-2GeKbxIxB zmYy{giidl6_irm$lMW3#fXTl$sR`HwPYl><;n|y52f)5R#NV#%S%B@?ioG8QZ)qKX zYLPTQ{2F-4L+&gr^2cqAo?>)c@AWrUxBV74T-)KcZ)M`o%(4691LKbaQ?Iz(gH>_4 z;_(jFbSNH|kD;;2X2*_z|7P^D7q=@;Q;a8GehOO1AC%rpNA`DldM}+ZN$)K^p>sug z`=HPF0q@=<@ETnM4}Jxn(ioK+p6~bRy#B1Om1rv;@1w8hyuSQ$jjZ*umA9SgzR3D0 zc3ge+v$&6b9zRz94#@efYh#_PyWmUXEYRsz&!IX9c_`+th3R zZ5TSsTlb;*h{mtu4`sl|1EK#A=syhk^VWP^*5%t@kU8?in4aG3Z^vE*aMNS}&@5YWbzCY2n#d zkzeWBXVjOcm-W2nlB`~y5|CbY=aRUluXJ-Bp4G?Pqpx&x28-3lYQy#o4orO7mH_lkI=Fr~Yx|50%D-6E1ZkDt3kRyvPeRubN*P2T2 z>>0cF6yL+}bTa-C-~2-ozF|GHsjn&o*9+-029880cO4$Jm-N>dEPcHg7^K6s7jBTX zT9sdB{Wg2ymZrMr%SCxlISbKOXZpD@i@T;TElGPpYE04fCXbejO~JN=-FK)~zR-5TbR)cy#+hK!oktp-MZ1CVNQ<**rCWiJ>Ue*)zHAk_+V_U^?2nV)|2`tfcZj+Z?;CPqChSHh2mJPW@2pVH8InqwVH=@{K$rTKU^Xx z_WS_VcO~Nye+~0i*7u&JV`yLN%KF+L;HcRjAmO}dH+G^<(g4E%JrlQeIL1fnrj3y)ScPPr#RJ(G^e2pUrY*YK45x z^v=6V+&%U)(NjmUm*i51x0jHIVDEuD(QSIBgtpA*V)QuoH{M4)P$(ZfYky~O!2Ypk zYF7_Lrwl@;d^#|wcC3?CdjWh=85nH#FVLRzM;DO$QmhfYfY@g+dVL`A!2so|jGX9%UN*zAKU_OPX(%)j6!KLn9 zC0zc2Jvy*m1LjhPUKZ-&tjzEr_A=Kwfxu~=%jLPFcWaHh&Ofr}6qtMn_{tA&1z)Yu zaKl`8ywY>+z}Eq+ny1`@4*oggl#ba%+seD@{yO$ax8pPZqj&-wwNt~YF&@2JXNCnL zZ%OaczVI=yjJ<)0cUIhg`KbY{aU7s=vmev|_JkSmNv_%RME;Y;mo?v+F?VQ2?NWH5 z1z&bn`P@@2@SI}acbS)H0lpgig;x69%x5#WZljNO`g)f6&7=4X@~3V?&$+b(ve`Nh zN#_`4L;}=1=pFgZ$B0Wu&IZT4zZZX_`x(c~U+TB%z@zv-3Ql5-(bUgF)61AAZK1EK z@#x+Wfr2Gz?3Hvo>#yM7<-ok_DpzMz!52F{yWpPL!JK}EnI|{=Iu-mYo>q>j72Z_c zaT9#1dXE^-dF!#(~`=8O_npq+UA`4L0d3=|8mCA z2F;d_2^7>#bZMaP?VkgmKHLg#@}=6rtNcg#(ZU&Zzr+x(Ppf={WW#=EpQZS8}Nx&1OHCN&I)CGsv}v<#XLd9d0G~Iy%Fd@hY^)=KPLb)ZG?)&$)BDoV^S^WMdzL z4wjC0K*yIplbnu|MaNstrsK1k^AH`cf{uBCf@S38=Ybb;EnRhdU(Wvf%SBuA;&IxJ z@0+%Bd^X`gbq{UL8WZT+3VlV}**quO8ry&EC&FjY_0l_%6{CY@_SQju+Wt;^!6)b) zL))8N+P2-onqZG^qLuWw=q6f;ZUg$D+b4ar8qIYvHI|yo(i$e=Teug_|G~q#l{c;Z z70o{VNzSV+^618SU3b72@@2J`w`#yC*ly|eg=9s$081mJDdT9mrNG{k}w;pf#n68>}5z1n!MqOYq-n!Chj0 zCaycs8As_e-aoeBa(qSUlYteNPy7`5#@3gM3tkz(SqtE!&~ZAI)5=B<&(eMu=;2oM zu=LR}>0u|$=wUsV4Gua=od!3r-XvGgVch) zTQ=0G-2mRx(8Wv0D`>5PcwsGhRITHz^63f5xZn`pPLOBObK1jZZUW!fX9OqvgvXzY zA(QcO`P$NRf8e?F#^Qpu3}=Gu#PiHM%FoDFj=^V3!w?my0R+TcOa*d*BG*@^AwQL13BMv{kLJBoJwaH{~S5>{PS({&&3x_*e2mlGON9m zqz{j?U%jsJi{klmJ+lm*9AsY%`3S^2oMX#b_2Bpw&a!O-@8lF}M|0nW**y!^#D2k4 zQ`@{5yWoKPZ2YS`Sl`if13KnDcmJREsjTCMW*yAaWs|4)G`Z}h$j>V7<&bj~&Q^PP ziGi1+j92`re5=Jv9eB}LTkpeOh)z@09`t40hR1HO`PQ+Emmn|RTwi>h^miGz@~xVi z&^qbWjF;T+i<+~DQU@t`R(ddO0ETu0L#Ek#$-?k(A`J9xWR$+MSTntveLOE`AEp7+ zkF~RpXa0pfHJSKuW^K?`{Fq08Wel(=|0Q2-C9rJ6KTsW8$Gy<}?vSa`-@rPb#2Wo# ze@zbf=(E;UG2I;On;Az>O^zLlY8r0Tco;()ycVx>oC7|t0nfbG_5eO8K1X6r!_JNMoy9Rn@`&|P!oTbMu4ez{S2M0L1Y=lvLaEj2+WKM&mw&PDfg%@J96 zx|bw}=c67x{#>$uj%cK}=KDJO`W?^89-27i>Fyb;Sm$(k`TU8c*pwo2U9X?+zR3Rm z{0scfx~8%Gehh!J^igNg82GYuiF=M%=Cy%>ztE=cuiwC&D0=gRzCZ8%j_y3J?=O1a z@u`mL`!4L)dY;F>(i+HTGSh0?$$PbXd9T*g%-R<0m+a{AA9T-{PR>QXXa~A8n|DjN zm)+_t%H%VX+*b#=ugpMNZ5Fw&edNBhA5{gpug&Pk7`SakzHDBi&6BT@j6?jTxq%X& z?%N_9d;BQ>Q~Vc$|Fj0d=)P~bxrzV428sXTx)1*I>prXR#DCw@p7am&)_2l%^2rqw z**LNto$?5F^$XC*&8-C^|IGEd@bhW-mU+is_}R$Vgr9Og&xIfBm&Scj{Ja+YyiGsK zn9$9&VwVoS?=tpCH7~))Klos~(sw=HcRkj3J=$E$=A}k*&9!2Vl*kBkt@)(%$aP%n z)BpR$+>>mcHAip}I{gxK0=eEFuIW2}NBvGbx2Y7T*&nLJ-LGX8e)W4pE{gBh5+zn? z!XFl$ZGKmIcQ@8gi%53Yop9rQA|)Hgu$^LBYl-p!cM(v>kG6B)ed#(9qG{n^AQ9gy;zFd(lxk8dycc z=kZ+)4WSeB&zzk{&INoPN1r6CQ&i1I8+6wG;)(T&7N23^lRgsh`D`+L0*ApTv~7dN z@mfXRSDmVA6;r`wJM*OZ;Bp_hOo6`2J+DWvD|aY+Ham$uySooKR(Wvv?MJeDk0SKs zdiHE`eXhP~ckr>Z13^rjljht0Rb8@=Z5RA-!GaGuhsx`3u^U@8BP{Nc2z3DkW9znxFQ_ld%L z(kB~zaQo**1hCN`)vo@iVEwXxe>t;G8b3$#{<0bE(wqD%wmgb_9GeM$q1&7BYuccr zd?+z=>eFC-2@mq=f1CH~ab<9TluxPx=2#*7qi}9^ z^9@A>C43e#{}m7|&~0sur*SyEn4~|>rGIkWAl4$H57dU79z|a8c@IXx zM2%3jCR4B@My*-pUaD< zM$or%Ioh+hoxY>=xgMTVKWblX=I8YNemU!4O`SY=7pxi3M14F%yrJ5!yXe0iouE8W zM_%voi=LvZ;;4UTKC2zsZ9x{L<8|J4_n?)X?aYe`HtlKFF|MbbRG#zW(zY{K?TqXF zzTj%2ZH?^?$rExCgRX|3t6h0?kdbBZhX0=bx)@m(&O&^SW$pgh#d)X3EY4?c9C)Ei zH+iSy$WJMs+Ut7E;<=|vq0=+$St)ti>I7@s!AX;72#=e&RYQZ6$T-usax5v4FXJIF~OUxqa3|>4)4r8)lK~P3&yMG_1;dN zdlvdDhp`WP)ARhi`18Nyc|Et#Yj-SsoKJgtr-GdI=*77_JL;7BSDTm7wtkz=^CRgq zck#SaBl$fSnBD#Vde3{h_fjH<*gH9~E$JEc&AGZ}O#855wr|lb-y63dz6ze{8Aspa zzR8Ryt#cJ^_dMTw+^X3+_gFlg8=s@q-asjlI>uth^(}KP-JTNprtkjOefOX6-CyIo z|ETZ&YTx}8=Gyf)Ys|IFn@hOv4{v6_4{y%&c+=VkogcdjA21nzPG{|p<JU4}?$Yl{AFjunYxn+G-~G|%+ND9RxprwVg6sa!;ODG`ZOTCB z4Mf)s!A1>3xBB$Mh6Mev&&<#F)GtUsNd7g4r?D%xRXxnJ@G}mHk((-HVXKq6zx!40 zgSS#-X*x2y0686Fa80cWbanM*uF;p}JfEHAoVWv8q7(ObV_V1JBl1o-n|9-VhR5@k zM$c2Hs`j(s2Y>tIgJxtwuN$y=`RHZkI<-%*{E)=9n$Wv`dR&hF9=mwVDfBmdkkNVk zl={f%+y{L-7=z$0rM;u*^#SxDyL$QJa_nkW=Xm;1ovOPJVM^r3s-N-n`472P+hZ1w zJ=IS9-;Mo#p0<@gyo5H!(C$&%zkG28ZDn>oM_WHLZ7JSMiEK4(xiY=U*Y0}X{qJyn zE4ypVB+P(C0h&7_ImTiWS^`vLaXWj=$Yje1@a6YjTwIEx$qep;lAbET4U>sdFo(#&$hC6D0>~_|1#gv}UmPcaI;v-}|!pqUCF>h3!k8lf`%YlIm0J zk0m}-`qp_OE>nS6`#{mo|Fsm$X9Bg%oTLiTg<6-+zG zet+5#p4#x0_5MEeru%+cXBO8f^jFOPO!RIJHoXv87aURU;F+i$|PNzJj{oa3Q+oEp&@?nQAR>yl%j2H%`9s zwHrS?;I(Oi%+<%}Q}kXa{e~P*3#9EYrKUpq&f)tQ=_g`~G3aOEbjkn6+?l{vRo;33 z+$9?UQK6+R)+B`0rP_)@M%&yZC@QwvO2>9=CkbImh#ku~t#(|J5I`kbx#~=&IMcAI zxmHURs*KaJL~)5uLHl;|Gs9i7601^e%dJ4(@9#P1+?;#4K|3?=XI?&^%eguCoM-v( z+w(jIUWz^WBX@iNyhO8~JPDk<=Vb^!gV3e-{1|iIBXUG*iQ@QUiVbZAUaE_Fg!%Ln zH+b~u{5EQ>59A-Kc;HmvA^$+SSTxpy+%Wb!{Av6F-jN@nxJn$KBY_X7wK1li3O=8C zeE7!`X`>%`CmDYh_M~)!=(-oVZnvrW%KXULY7<^ZZyXq)P320x?dloT2;@?@@pVNBeQS>~U7pCg7I^p$9`$}VQ z{TT6e>yf6{(XB=oYR{cIYVu?ssjl)|r{TiwH|D!d zo4@4y2l2A#^IIo@o%q;${xbUd1o+85l1%C{Fe8r1oVETx&&91{!_2}qm9HpxbO6t_ z@T6=R*%`n)Wh;sAGGQV+W58n@Y%ue*ZG(ULfO%?;mz-pdih;TQp>%<_9=sH~5AtaO zceTi_Kyj#SDj#ZPU#??c@va=}qQF${@p12!>fTK5@p10~ql*IPaF36BkNTI30_Ss& zk9)65_hxgCk9%)3F_geu?(uQ&1$6Hc?(uQ&)e+AMgt*7Yz1O94f@-XA37?f?nqOP= z`Ul6c4zrHb<)uvdbOrQ#e0)(V_N`c(@@{cz5q!j!RBKGkgZPa|umo6A z3otv%e(Hjyo>OhC<{-OR^}{;vRdLZ5EPq=*u$B+xht6c{sC`?uYV8zf5jtzFwV%vm ze@~qQ@=^J7;9GEZYmpP!ONx>0 z(BAKSFJ-)|8IQ(`{76P=SGKe2j%+%%=^zuHD>C5e&5Qhrb)+tDyZ*3OKVD(r`-v2P ze;C-_oej3YReYknf?%ut2X&5cJ#c*iJt6;c6SNk)Np(E3J?*_(T=*^&Z1GEMxDuDu z+;3{$nLtl$WIvw*;H3A=8bWN0ulxmT>N1|!xix-ZoUn4c1!v{;w?GSDXS~yZZIH1k z|DGe*!b@qeZOsnbPk3SL?6jb01>oAmnpH`Dpw&?|EJY8S4B@YHx}pgi(fx<>KXu5M0CefDt4Z;)S**dX!z@ll>Uez_qC^f5SRynx%f#sVHf=Jro*>sH00Pe zrO>5pE#(XpA4njZIMWfnTfJ7bT=tx0cLpPe&mkv|tdh<33Uracu3Ah#t1PQ+G4R<< zKN~G;hoQ?i9Ns!unPy9+>Con)@P-?qN5gA5k$(wi;Wc0~BLgP8GU(%Qm3{u@V2NmW zLJB52i(BV3D4*KGJQo9x4roQbW)1UQ#MtHQ*P%ZX*dMYR>+^`?Ra?nstxLD&gvWwY z=``7B*lrh^{P29+e%MOwgZ`d^j!ST_6dmWs-iagYMe70ROZCU_#(t}Kj?EjK0U>!Q zd8FDYo}XBW-z)h#26|mYUFDKf{cT%?JH}c8oG(QVlw)sQ!G7IKPR&XE7Q$CkZO}Ww zLuZbQ-`^z9nCf$PdY>_D1;sa>!&lj0*3y@SfByTU)btcZ?q?ik&dp`vpK&i*$Jnt2 zrW~z!JzixcUA*bp3AY?KFsTJ5wd_qBy`DO0U?ICNR>wQmRmOgG+iMMQF2eQff6s3L ztM#>jKhRcmN6Vu{etdmu+4N3%{#ed@7@JH~$wuc&-_X>7Ug9S5Z)<>S-^HQibIv{e zuJ?k4N1k3o?dc<@2HX09*~9GVKaAW`UQhMCO~jM#CP&b7F?(BCp|VG)3C`zvt@oK# zP|!VVY(e+nb;Nc!+aS*SGkoORbFCdZ+eY*5YtlLtWLA^T{|-HQ74}ksei!9m#ksmy zne`k?m~S;a;Rk=J8CN{owza4m;~m@9Iy0QSnfemm-NgJvQ*NE&1P6ATX-71)L2$t4 zQoh=p#i4jlIq(RMReuK_X0CgntyRp!+m6p^=V8{JnEmzYryhv|&w*=28> z8N+stp9dcxdo1LRWKRwCV@7W}GC5Nx^s@%l>q}(Qk9VIX*_U2lm4Jp9jkUH%=OfeM zQPpGjKCAyU@#pdxhxP`)3ApOqffsVE?Qbx@{4+wymqNbqS>SF~!*aX6 zl-SuO^6N7{dHd`A@D1x!cMPC=lzRzHu->tnFA4wjJ>&{9aSHDzm~#_;>n8Yc25pPx z>Y+JmIW8Ok7V^cuvG^tnU@*_|2T`~D`U~nb+Low9f zqMOGwG|WA_zF{tF?HVn3RKH9A=qf#<=ejtzf!&+QjTCfqhI03eV|T7CvGUrMGKYi6 zg8}5I_Pd;QinU`H{Zvc4Ip3~;|+xGxv(!6WkEfDd@99!|0WSne0x7cuA8$k_|- zg14a)+R*Qgk3RclewSYL16QqsR{Y&qyLQCU-$XOw5cDLSLBHZM?;iQzaI0 zM&XWq04k2QtW5}q)D(GfRLeMA9J;)q-*w(x191K-ziaJ6K4akg19W8mP(5-qZu6j- z7ctD)*Ycd^GN7?D7sE@yMX|2U(B%xqLQQ%1?iyy*7)u^&e-i0B z(O{gl!!g#x4q&@ze~Dh!#_k?tZP)taT~^^x1Xv}o?PQ-LR~8PLXY_1-L~)!v^r-SC z%4rT_mna|OeP$*5?d3yzcd?hsJ@CM}$ozX6zA^W=^!Z)*X_xPmp&DrB6zEys(c=s6 zffrb#bKtGOxbA<4e*Yc%&2PBj$SXb{cB&PA4SeK-`+S|@0si5ag_@h{VN6^Qy(vAZ zJlG+4ytkBGeuq^iI1fpmLf68f?S}*+_q+;?qEAdbx2m#CYn*#aZI~-ZHrNp=6Auo7 z2gM6j<0u)YKJOxKud!_93@eQ#0lYqM<#%iBryTq4T8%j$8r;g51v6L9+IwRb*s;K_ z^!(X34ebaor$oM-jdohCGT9Y>V9jhAy|jkPzC%CqZx6Qy%zF2`sfT;7H79Mod+b^0 zP{p-UU203Zsd7CFp2>q*@CL&jBd7)l{wfArxvu$<2viNx4oXIZ#Xi| zqDB&akluWpwSN7Zc?Wpvef2p&8~UvrKjK{Gpd6a~2jV$lbiLO4K2x_p%o7O zn0&U02TUH3FYr<-?v8xO4LWgm`2Pdq?q-d*;_j~s?#MFn(Fesb{{;S1aZKzR#o$|Q z|H9Sj7bf6S>B(PepLM>QxI5!9ad&JU6MF_~Z<5X=-<1oF z&}Z^bOuaVpRBLnB8S$?a%aUHu zI76&=wVHFoorkge7^|_fk?Y)FXR$}m+cy!1VSLrlo#N>89J;G@?CXnxujIp}4y}bm zYv>A-E3;;W<)3gy%7fC~8+m3Eydqe95nIx>yR*}^tv92|!opzoE~#bc4Zjo1Qh4UI?CE;T+o zr2n;dWP(0aN3+t(F+Rm3NAFx)i%%iic?Ox=ip^4u?cKxrDd`Nww7OVFF$6A`;KxLn zPYihxN7rk)vId&o zWChGv>KIEuZE8HK*Vzp%Cuo;=Q(Fsut^5qhs>Qrx>Ul0^9LTn2&vP;GBf1-cj_$My z+gzNOI-Z005Xet^t(D(C-;4G%mk!2j>k5yL*J^+7u=0jB0OKm~F^#@gS$X6`$cc~- z*@GP}8&@&e4bC|5rQ7fuiHT4bNUfmkMQ_`7Jxee$Pc>5Tt=9BbnD5F71|y#{-?iV$ zxJaqzelg$2_zKouiN3tD#0n49FF$g~8q?O~%NdGF7lNNc&dhYz)Y!T-C-PHh)5J=F zt>{yJyKHUY@+#(iS&l1DawC`XcUsH?nIU^ya^q3f_IhQTYiCdA9XCGXp0_wyEE)3A zlnxZ1x#Jpc%H#ak{7%<4`8cqO!cU^herQAcTL>vg9xU`^wG|lZBA%;!)@6p?a=Vv7Z(83z1A6P^zUs*i zfvZJ~Cx_TQG2b7B;3uJ$V`Qz;xpJa)~i{Fqo-af`VUpO=4CCAPFp^9he+`Tw*&(3rEQ{&axzsH#S7;nm! z0k3g*@eXK-`oy+F$hRV2?$FEc4&ytGiCj;g`uobUN1on`411h4rSf;47&|Zd1ALrn ze4tJE@Q>%uOGdu4oWI8oy@DNh2!H8K+9b|C$IYL3c~SG!$HmKl!AriGC$)a5dB(v@ zFY^>0HBaGs5qL4Rp~!2|u4Jf-lUn9y>XM7VCvq{RLxD>`Z5Y1K1CP0p?}88cQSw7w zI*$Sm{vK79z*Y4QcRazE?Gy>jKP41e9uH;&jNUVF)vA=0hVSMrU->zsSGP`>z@E7+2C+Z1{`+NkL0TR&&|Ki znjqOa62B0g#F34nAI&*PE~I*Fa3})3sD8&?vrcRoxpozF%3eOqTk$K=Wjbxq4R){MUFA{w;S>KE!H1&AYQ@yGjxf55bN#sgOa9(Ryg+^1 zbY5)Yiwmhsi877^V>Yn|#;xaNlWKl;ti~Rb!415S$BSrc-KsR&qFs}Zsm5P7rI|Qx_0dt^zAS*xqsVT>&ty(hAe0zhHjAF z>O_gV6Hn|VEsXa#3UnTsevtHzjYR*PSGmrbRslDTO z>)g*Uyk*zCf7*PvV`v}c`*ijk-It%+ojW$S`vvf%yv;t#PyWU_Apd?Cob+>!)!tKA zuJ2{v*(NJEbOXGtSe9ZgnLOIbI%qGCew}ArdMqGLeSG{#{ZtG+CRbM`NpYA?C zg#Bc4@IA-|*ASwJnTBp$o{a?;^>4l}*?*zKCuhcO^uhps^G3RZ&A zbJWl)Phn`c)#g25XzJj8OC9_$G#BCO_M!G)0#_`G4`n=6)oIk=5ZfYx}g2;H_)@*R59K3yB($xGo4WI^|Ie4IRDYy;3^F0mAH z3TAJn!S{z}UrsH^OnCF|de%heS*DIjvR<`Qx#WtwXlFC;>YXLr0OaN1C^xlPa4|l-KUzS)I4x#erKVx7vdT)yZkydJ6s-@i#a2M*TSF2%Gb^ z7ABBU!i#tXB~ZUWU(WEUQg?U2a)Jd)=5NZFt*%HrxEct!Qyx1`OnnWs5fr za>kOE9$k2BM$ZWzEsRI=M`j)%S20nxW-&Nyu$TNF}g|!Mmad6PJZPdS|dtYH4UEcGu26!4cc&VFN#R4S~%E z*s})|OVCxhrqY}vPis5{(0DGk?cK<2cYc?BJb$v{->P@dM_%TBFK_ZLE4Q%(nc2ks z7m*7I>7nzOAF_W5z9~Fu_ED6dtU8%VjJ26~kH(W4Gc+mLpfTgKnQ^;)s&+A&L9-{~ zTy}3kBp+H9T;5V$2l#!{x%$oh+{gfbiw5@~8w2QQ_IlKsnZ{c1Bz@+C{xaz&#k1q# z35PH4;H(6%ts=hotVce~GWFi+{zSFqUl9Oqu5YnKR$)=-H`3l-`FqnsEakU@Av~LI5~f( z$vI@uKYL}6^Nf5A$?tp7lR@^|?_k`@<6Vl}QC=|){Y}5rj=^dUb}n;d41#A2StGtZ z*;rC=oowxtteNG7>Ez~}k9**$JeRlLMKvy}X(@H&t>j&KF|~e-d3e}Q>3yJ6a+h_w z6RVIrP0V2nV_mtSb-n5{>~Yz#w-FfGe|zg?)Q{Q`uSc={V+-K1tKqS0p;zQ;f+)R$B!fa9{qwYWBqa z7&QSr6IKmMDc4=du}wd3S?~CD)T^9n{JQ`A2L0nJ#E>gIcOW783=D+RUU2YHgS(u_ z7UIFRVeu->bQ3Yjy&B1&7nKn;*;!nPym}+c?;#~kh?=ukkMxSDYlR= zhq^{=A^CvP``AS5oZ5^F2ayk=;rFOx(DzvZ^6%s`&&negK;2a}_CY*3OgsZRM~(3IX|x%{<}yAMHdhre5S(OZrNhu4SwoGR zVi*6cS~uulc5w^e@8Xjtqa2!X>oFuJin7p-87JfYCNyMu{<-$rEVDn&$ol9gI!gGq zV>_WIWq%Yh_grXRv7CWo_J3=c9nNc5hW?weR(w%sv1hz5hdmJUjC|C-3U%y_DA}?J zeo_8K?e?N0ip_bxqBX^beUVD&O*vBOIM+|CWbIaEXndF&yax?@oceO~6LC=XScHF< zpriU2%RbJ`dY-(@wBg|krK{b0TK^#bQGS(tQxlV=&tB+FZK(G7C3I0e?YxMu^Cs+aBwb~>cq~jb#Pwb z)bHg$N7HCqdQ3i;`Y>mN;V*mx+QG+OxQagH6K=)k-;7& z%jrYDIkMH*@&RgGQ!y_GUhdl6!#wNqQ7&|})xa@l=yBHlGG1V?#K3_3!-ckgXXio^ z<~@6FjHuJbAIw#>i~Q!hmnKDncd$Ou)&I*pG$(pHT$OL?J6@^$}2 zKgHylcC&6)c3PDB-Cpd=jl3@#vl^PI1Ez_p3KK_(a!sHY7PEF_m32w@UvGfk9h(z+ zJJ4`yep@~{r5b2NYuZImJ%Qqga(J>Ij2}b$YHuI+6xZH`&8*0k9byk^JQpxDRD5_?C2@hq}yljwsJrJR3El|UUx2Y z*Piyu?aiS7hb`mtu2bFLUB`xJw^8db9hxVeY-;@!3r^8|)flJFMD&^)xeD6URdz;M z2A#To?CZCV(5n2{>uApvk%nHo?ELlMC#7>gGLlcL#@=nTl6$bR%JJcL182pg;qe2x z);ssVF!p(DfV`pQ3n--15CEwvE*$G;6seF@YjaRV?mDc2so8k%2b$D(_Xfvbp%t+H;rf1VX=>tXD<(1G|;v^)h_pxkz% ziau%63U3?R6Pm`(JJ8!mpErh*A7T8aeu6&Y^c6&Yn{_tOtlB(9?WrBLr*`La zb=zA&juRRdpLpBs0RH!)7nT7#{9*$;!EX?GgFM}S2pB6q-3*Qv(ElKRONK}mB^EGe zbm0K}6rWFy5P0svmlZrEJEj27LEz%#vkVL$2Zru^1{te&J^{VUJZycM8~F{iKH7XV zM#;p@PEJv^0UD!f5M3N=Ov<%MC)WyA&bZ>gA6_VYxp=)xc;y-GCOFJ`%S756mXOdH$H zr~jPwE3Tdr@4NYQzas{uf`x>cm(1|si3Gv(*{YobY{_?Xnncw`4Ey!8#ZwdAUEw%id=@96i$C(aVkFx=OlTG^j zS;Os;Yw-TU#P#YY&=|r7E~Y*GeJ0mh>&Qhubg}GWt+Upe*cf_F@x1b*Mb6(%PwJcz zKYsfZ`l{xBJ9pOvEPn!&yP>f_#5pu?=Wk_^iE7X!~CjFl@H-Zjs+rp zZzIRJr?{^8CxC4syo-j3uWXl{Ab&t~eh=*IbroFP9oH_Wk*N3z2p8ooIHi~>w?SSegj7<+5#g9kORl>jcT|AdS{zyJa zzC?jxvvp*S6T=ICmb2XzU+$sJ8sUKXAmfJ;W z_;`##_PTP4b{^1)FSz=I=VSD3d??1!1Z^C`Kl0LJT^4$LH)PXeLiU0BWS*I{SbS`F zP7yS%xVPXw8ti9yU@xCv>)AHar)n;Qw?)97I+dNWQ)Mr?u>Tw2u6Tyx4T=#&uWR0^ z`Ud%?8^{N1FBKP#L(szWl0~$cl120*UdSek7BIiR;)5*OCjS>%v`(_5c*4io~5A>)+*`Xh8ef+rFh0DdtRjt9g8weH5Lzj=kox zm7G_s5t!1iWY=)hn9gXGKRf~W#o%=DlWY*RCz)nsc))(wGWMJ`bT4%vm)6(JeR!4c zz&?DvgNxZ~U->)aT(7;S_t4AG*o*AfpKGbc%&z&FZ{?ZXoYsLn3jGe!ul6kxfBqwV z4e<8>_#LI(pn7F z@vzS*HkY3@1ir}7^|Z~qhoI#eUjg#Yt^*k>eifgJmv|>}9oOr*iuT}37cRZ*7cRW( zBl$txczZu%K4{tXL^aHzkH59|aUXN&XAZK*wI80Ikxysm%wEh(1j?=H~A>lxM zN#4j0amOHAZfmo3pbz@bX_!7YK-&+2bNM*05ITA zhTc^y*8NU7uAZM!~n}|epG(cI;@y>2lTR;t7>8X13HKS&p+{OWq|c6_^@%_30VG@dnRD#K$p$L zu~u7|EBUS%jyHFD3V+*Uv|;`KBKLR|{T9-X{KxowD>;{%&iH&_y+C#F zZeGOhU-5&DjC~3=;b^(ZKKzC*a+BoBm$win4A8c6m%1+*R>HgZ4al6P2P&!cklZPu zO=6T~RcBoKdL6mW&GmK9RFdc1TfBFDH95{&+VR4{i~~E(9f$JrHTVRhwco5E>FMF3ZF%w6ivY9|ui+!JmiUi<$rb27cwQV}n2Fop&}o)MntJ2fVd9 zxKPe{6ZmMd{L58q@~pAF_?vcP;RFn8s}OBwn@x-@YCG>gxmcutAY z4L+kAl%G|M$>`M!h&z+7HtX1=R~>#+J%#2Znz?owGFm#AXH$9=T9E&(*pFnva@Kt24`i^mv~gN&gIc;bqubo=zzy1E4(x^8jpLQ$wo)FWP$Bt9v$S;Q$q$liH=;F zlINzB@z4TL_jeejtFa7-H53D7Ee&Wz(0$QoScUC-VuS-Mp zr#e9|4Q;(+bQ=0xmU(*hk1d}@^sAe1_r`r>8%U0;4o`F13r{>(Y`wFJTsM3_x9j-u zg^HiLc48HFVvS|BO=10s^5&d>muAzr?<$U`ev~UH$JYEU{U~RpJU}OQOfGp~<@kr` zb0vMQp@#2ar!PIPyq&jCi&&)F=M5QsHUYy{$EFR8X4ASh7c^M|OcewF0$1VZYv!sv zbRbfZ0dMJo9@c!gdcM<U6y1tu-5!tj1TZY(uwAvyT znll9K7Alu4nA)_Vn%F?(f}c=3NX%8Sm_#Eo=tgXnn}~PZOq@e88DgUIS=+S6vUb?? zZ|CYh`+I7ZM4LQ!Lap(fk(@!Cw$$!CXio9>mWdW?F#O?9A42YNKfyia#Z-5ry?)%h zpz<~9TVwO~8%OTO#4p!+@?4Brb6JSoOGuVtPe`76wjNvo5yCvGokDKgOPI)`@3C$4lWa_{V4Y+Oo~bz>)ryH&Cog`8u_?nKgV% zXs@3*)(qAP?j~+1x>X!ca$jc{tDanIT+nH4QQC0Zcj2Y;vjso3mmmi-%(b5DOe@dS zbM#Zs@q!bNR36Uc?ts6WyCc3jk^PjygNl8j*R)5?p#`Dj0@ldAIUm1}oa+>LXaKud zaWC!L5oNtrjJTxPYlblr$6lu#!oMv(-ripxV!u<`0RC@6KeQbgV$N@ZFReQs0(Z^S zNZB#T{764^KN`-8oET(+ac{M@gSs>C>BAIXg?A4?{96y)K(4 zLyY(8zc}9h*L;X4tW&(dg!s7P{T0_s<|q#rhB zb9vAyFuoJ}(wxUjy-%Duknf_VJJ6laUOoBj)w7B<>kMf(Wlm>38sU;KlWeV6J*nYqbI^m{O=j7_+NE+r};z4_we;P{HG1I zP^ z-vTqf5qq~3IQsb*e1c!?|2UNozJ|fK_PLtMM}Hf91OE~}`~UTW@m=kKVbn7g;oTk2 z-5K!#*+a^0nRUG2R`I$4WW&;1W8iNbR2Ev{=Q@O(tM}!1 zU*Rk0o;fQ&l3Nj?J~+?V>JL*RaVPDl)~XU3VDHME(usDw)ULHk;4cTwegviAP1L>g zQ5zy#O8O8xz^o&Vp|e$|Ctt*V*IqmAmA9%FT?$>u242iH8-KfcOY{)%z&lfCZ5ajr zHouL~Loqz28aUxyHb*^h6aVKx6H~B@xAN)52CPGew8)+SM~`DK*yFZob9W*6`){5a z-e@(y=Dn|ZUg?>K#{3JdTZT7Of9sV88iV$}EVlgJTBlq@-D45FP}FRd>2H^Q22K6_ z$oiEa@98`T!D@|_)4dO0K(>+o_O_ut=OoLt-crAnv)=L?aIZZjrQ;IBC4KZM9GE#c z@d`ILkq=DjX}g?rsK%j-`jACigOzPt(6iF-Hvl)ma2UQR)p~E>cBuoqVSZn^`_A=( zVbB-Q+Q9904mue5g6ehAVT#H3p--^sCr-8e>>d8j{c5|KcYgrg2Tj|-p#*$T$^BkF z*niz?$T4f5C)R%%`}{LH*OvBLm<#*pN#>;1A5Ny`su-Em%3Mm3InCOC2pc`sH}DpW zN-UjSg#8I`*)VbT5_4c;nf(@u;gP?8@=0nkutDJki-D4xPo2y(jKZHbVc3sf)$}^r5}VSA(-^D`)5xxX2Q=|?1}w6z7cGEyv#sxu-y}LwUM>I~ zNVcmcB!P_`3*7wVuc+}+-Cq^BE(4+9MVy&w2T@5dnS{Aer=+1=} z=0hVfc+Rfj3`91NpDAXoqt?P}Uwio=$TUNX)SE1%cG2*|;&&^}8ki0IRt_!2kY_P$ z?I^Sv$9@ve)R{9SMrs`Y@~21Ji?zlPnvFrLad=n#v?PbO8@bnN$7#*)%vJt@_r6PW zufWrWN1w{z(Wh-5edW62?cF>R15WbKcJtlKqi%lhOBp!P{wmT7qR+!sc05V@rXL?a zaz1DDYGOE5eAq{87@uUgDYtVPb*x)AjJInG(Z{{~MjcMuWt{D{E9f8U!Vf6TBbRzD zbuq+-lW>^S9XRew1fjIq8)X@7#EzoY1~t!iE01 z@ll$hqRnYrlbk#;;U>imA;zi&n`ts_0J09SnVMo^( zUBbB#hJG{7g)lHp({m9vhfdd@t#NN+TM~AuY$W-(7AB0xsvalYtB$% zd_gll**qS4Znbfm3F}ee={zr9xymya*-P*Ijm`hRaK`L!ZZ4}N&Q!^{8lq9z<}SSO z%MSQ~pNmh~=jyZOSsRD7yDF2qC%SR}x$@%*=IE6lk_QQ7+@n7}>eSdd`X>+h>w$5P z6R*=)pLfRUwP{kZx(s@4K<{Si^8~tM9)3>;wFbMvzt-&bE?^wo*Iovb;k7#ARcUgP z`i7?%qj=4$55HgJ*%!mno!S!;dN=zz<`FwL`-iY+SPAzA&=VGW(IxN)RfDbAn2G75 z6QpZC%DE}XupdRuxKx+H9tB1xSVNVW`(!x%9j3NKG>|r4Y)LQ7ZaE3e6uWZqGidCe zk@Fi&To;`gBWJmok6@;_z!KoIi#Q&u;!;n3Ol!t8wI(r|=;hNj&)@@wk=3R-)ixx$*s^hPtz5?9duiRM)_iC9_fVrgduyrHr$-Q}4 zAN`dhFG^1BSYOG<;^UpmXzjP)zo`8ti&&%6hTc~@nQdglpXj|{6x=_dm{0uu5E4tO% z?^K)@dX7QwikV8UE03prt~-dA+{^n1ExR6I8uwypIv{>W(^5;ulAa-eTm%2ICPPGALW%~ z0~=a!>{@sYjJM2y19yRAR zC?`5L@+9;=+ITYQ;Lv&GNx{hiYiyg&bzpyg##Ly>^^Ifqu6^8Ru}*&MP<-1x>knGv zhBo|;{Yt@+_7k%E%#Hk`w?EoC8JyMM5V&YXt_cozVJmF`4qBIg8Gf?-FVR&wu&BdM zngMR>_$cRCb*y!*kJ#F(3$p+P_BI!GKx4AdyVAS z`YGj8_VjQ%>X~=DXIuxWC)$V8(ObUJnz91U=6enD06hr!bQ1`5xeV z(Mj~SY=eH_IRFpaYf{gQtYOV}HhJXcDf@xn0Q1s1)i*pjOX+u4KD~5Kx_puhl1$oC zVC(l{@LUcYi}BY8L@pdnZau`AM9Utyj9fw9&{*(M9>Si+zGJVbZq}EV^-uPd)d$J@ zP4boPgO9a7**#0-r;;Ig$iPDAWjy^)Kt@bLK1{ZfSM0#fLUw!Qzzo`~r_F&PUs(ff z4o>iuXHlpQx@{MgIF0`x$dP{F)eF25z-s_|bb&J-i?ipn|8bs`*T%Yra4If= zO!Mjw*`)(T6~L~d>=3-DzRTh3S;SZR&?&I`4GsPy%{5$W<|| zA!PWaj8QoO*$|3L$v2W;EErz~e63vKOUV7KIt*EuIgTM@|CQLsHyoSyz&)p7?;+z} zVZF29vX(1rM7y;QX+HZRY9Ek>V|TADLH5Lf4{Pl9Tf0wZFAALjkdLm{p4WNMgZ2?A zK+lYcRA5tYU=N;5oGT|&!uX4UgBSObkKVsC-?L*tVZkuSH-2r zn%~GFncs3F@9^6ozp;0uLmLG)Z6Kct*{|$P-s#{xESolR4Q-qaOqzIJw4wTxR;z$G zzw!zN-Ppwk%o;QFj^tt=w2^=|dZCSeXyZBXAX;r6Lmcq}{9fc-!%1jEx~hM21-uGv zARE=UOB<@cv1x<7TO8WBhq1V{QI5_Tp^XAV8`F-hUYmn$>7|WbPqDvR&Piy)GX~WI zX&l*T<7H?<=Yny*TDREETm0bj07G6j4dh-Q&17YHWE71bM#hWit6{QCa zXyP95t?NpMCcc@Xjoip;^ILA@R&%c)(!lj(dO&t)KI{EQllylg=UxBprHuMb`PRnv zLbfPYCx82P>`SdHQ5~n^k}o2=`;iBdAIb%lTX`n#bg7lsJ(bS@yxvc2PI@JP?<#pJ zzd$)G!zWI>%*|t&`cUMl;-Bob$Fu3V9_85P85mfRzxWVye~C57qs@J<;{*AKPbjag zyp?Q0$z}O{A9Oyny}x#PJWDYm#WFiA_N_ryoA?m6~a?Pj*i9ThuXSqb1SM%~{}>xgM#2d3obgskt+6#i|six)&Ya zptvM4s#V0O&{JWZpKa&wPK-~zU&7gij9u5Xv3ufCk9x+fv0mkj^#jj`W?YI(y~&uO z@^d&Jy5d)ytDIw>t6ZY?7AHxFz8|Jv6m-zdtwn5|6F;bnhkHf&&?ME71IRf%4>g^ z^P$CWl3h99nV~ww*)BiR`vUatNGI#ng_P>X2Q?@Md`R5Gy{M#t-vFoo!@&Pe; zqPUpP-yUmkW?Y8;96m8)^W?uhZ-|mwrB14>q0l&X>%T ztX7=pPSzM&@O3@=$L6ztZ0;v>C-1VXA`#07X?LyX9kV~Jz4wyc_AKIfitD-isLr33H@V2lG4Ue#efjjUO#T7BkH#h+ z&fTMSfc>R`wI)o2s?-4yNawBsxaRkp?dP|-8us5ydGdLN$3BBTub$2q4eTm07 zzKuP%5u3%Gt9(=DOFfZ&ZV&b9sx9I@)i_j``*t4p3+BGo1^6Qu^F5ocas4>?ucF6a zv(DKoD-sjH1GcH_vwjs?b>+n%u?=rMg0b(L&k4_7WIcHO6l;IY8^NUf&P5l5cJ8IN zWGQtXB>^j`-xeL5XU@*m_u7NOr2X!>;bg7zZX*5N>%8ywd*1DL-~Cj0ulw#$`n#8V z-+kNj?%VFW=Y`wcci&Bax6J$QA3g8>F_^SYXU`DMdtdhpbQotkm+bz8kC$Y>`tGj@w8LhSZQi>46Gadwjbo^ z58-ktxcmnXzB-YqUVK^R+~wkM4gE^ra!#0uJ$m)1+F3|DZ+m2vi~r6n`a*JFHmqz} z<-}_ysDAvC@O=7EO+{b>dpcsb%I=g6TKg&MK`V##5yVzho22y*0p^_G_W-p?ntL5{ zuhW?eJg?m9uc$-N_fp^~yGDD31~1`?9}&QIFtup6<#h&TlRre4BX>+Y+Gmt}^z6mh zM!RTZD(^SZm!4}v7rXqrEtQY3c`nUvrmvLlaN;0>QQ;_jqsJOv#bD`g$3 zzxg%QIL4{#tNi)6uwWj)nc5cMGLL!oAdj`5Ru|7Ds`y@Lt*iU_xnb=ASH-!d^XR*0 z0r%(gK|g3um#*>DqHH)TtaUF{oYz*#xZ>CI`*o~e_t9USmE21^l>_uw4LlmLCztU- zAH^7FJaGGyiY+f6JZ9$!TkZXDvdzmoR?UT&$+oTMpTb&Oe`^0f(=PW*nBS)lrMKzY zd5u?P^DhAyZ^IHke3jTt9 zyzma@W8owDB3lcs{adXI%vxY)T)r(V-(-<|h2oOyR;;m?&bU83X@u@Qg0ruG%r>8xrWGQg~$X*Ft5h z$>}|8edozl)~)v|?$cpeE1t7fuNTkBKU6)Haz?6~(ck*6dwNg&y9*e`(4Ws)m(N*j z&Aq4>82WiuXY@=1?$R~N2dREvHjGz3`k}@DBriWDfNg&c^{jc&6Ex~)-AU!{+%Q+_ zM)};9(^<6JDl35w55cD<7lF)(qem0yDg2!WE#~9y>!jajyZqc`<3@XXh2|1ZMrPQu z&cPWzxA;f$DF#nPjScP#f9(I*y1X3ws8Zu8E(|L#$lQ|34Z8DKgly?#4$^Uw+mfla z4|Ap+vvT^+w9k(&P1m{WJh;R>xAgq(z(|hXmA`g<$4GsxYp+Y5ZAOkP2L6kYD@%|i z(kze^1tweE3D(p?!Pffu0W%8Nx&@~}<&4A0#_$aS2ckz+%EuHdhXD*}dKhc@t|EuOmdP8$5 zCsy=Z^u20SG$)O59%Ia5zO}3il3&oQaWW^3rHc0y8>nJz$|3uBPVoNtxz>Z4cQMaQ zpZP-VgAl@X?6hc(>RuR<}ErD-ts*(xW_a9f((0EcoM#n*PbR0=r>hC3`$dMmIg< z{5{qROGY>C_pPYV@BFP;TZHyRv!*T@ILj7mF|-hfj6aV(GmxR81w>wF!)mMK87KV?3zdSpF6hUcgAd_sB%{iF5C z8n@!5F?fZwrY0A@fjR>1f9)@TKgMCF;GYWqU5xDzw7P)5l@AE;_gEkM^pjJ+lH6+* zKJFpnFUyauUEAUt!+OGkNDp$V_X79|T_t_4IPrt|)`QI%b27FLzuWk?`9%4Xp8(tK zd@N_3s88oQ?!F}a^i?A;PswL!PW%ya;P0)|$X37FVC3~kyu}X=6qgGD`)Y7d1OBUl zZ7uVZ>{}x^A|IaND*jk`s`a4O%<$gMz|-G*E%5e+-&I?IM)cM~KjW=rU8`d1*b56c znDLJdx3CtX#5%{;*G??Q%-Q*#InUcY^VHl5oVoqXp)HGl(%0^K`&rXPV1qqDW1BJ+9a z#gS*oUFLPXE z&An&_&uP3`o3#uaYE1Zmj5+iqiQ|c?nU5`?pZQkus*H9Ltc{J%hi4Wr9^}3DhS2-D zyl2D1Sr=en!})6L6)L#sEbizW-n&@y;5{!4$i)u8D@9)gzSMSX=AHa=nqRxz%3oXL z3my(}?=0^)uCePCRj=Ped$NfXuT#9p+s7w)=L|l2PI}1)E$VsY^!Grg_PH53Lw(Fi zIQOKI&{%14nGJ{T8&Md5_d!{C5L$(X^YdmhWL~GiR(Xe~oeSoQrS8s=V{w ztKWI2hk5F}3;CF~T+{wImLuDoI!Ml5jY1z$KRHkjEHvJaG2SRXh8gef^zoul($*9x zmYT|`b=hq!j{jWP6u*7MfnK+{}5R2x+zZTwnoRE3g?iK~r&f@lGSId4RSP7X{ScYH;Um_L zDW_OFwBJE5&-L(Jmn&b;lTGvNeYl$(-fQ|AG6mQu7A1JXPw;`A3y|Hob`kp_vF5!8 zS=o!6>jTgGkh2eact`wX+y5i)r4PSvBQb)KRi~O* zV5|sx+>T@Im(NtII2QaZUOrQ2Lh!vbfGvi-nW(nUj1OXOEyh0$vSz~MG4RbqBckP& zqtGzV5%HJbv|ll;E%;62X{&^G7U9#%Z>qgSa}GpUbBqn7bu+H+cIcZnULI#X`22;$ zkxwO8(tKe!HlKHpsj>Nt8Ty`1OyC9JJ3vfeA2ET0!~=4DcN#e)npT@?+r;Y){fe)0 zO&y{1&Dy$A`UcMM^7aj0ID-ShS?`&-^7j+VfHU+QEggaLg`>f_Gz*-|Z8$%t{rCC) z=fHW!C~)3ETiM{eY!o=hsu=$a+m4F@b7RY)`>yBt7=KT}M=r&_!!8->l0S)#K^C&l zkTn$GZ{thz`*h^H&Td`ICxMNtJeum&_gHytEBRY>F!DQX`^?6%{F5(37rHLsDxWWk z-YmhU*I3qT%v_CKRcsw@#{QMg>ws_epo@izX^c-klYG~+CAYzgjqlKte+3`teLZK^ zmGbU1d>TEgdStbu{r|K^V&uJGB!3BZi_?$y{eXjiwFRHB-;(41Iep$yZP(hmeVO`H z`ZxV~#%6s@a$dfna;xwS&*U9rtk9hH%yILrk}E#$6ZhMpXN-+M)z140MqhH^qx-## z$<((4ryBa|I0JZ5+wv9K^TOQwUGMY)Z+EUO%t^4C?wOaNclcqxGdDe}_hgR~pBY-N zzL0A**iIub%!xdWPIlW>eQYt~5WJ_M`+c&9X;b_~ylR``RS94yJHX|wUoNFaoIa#? z>6R>7uc0)wgB(C~(*ll~bUV;+fWtmFSG<(`hnno9wLE z*%h`7JCAdIM0fmsJ8)KPJhscoFj5Pj3MKlnmp{rNzhvudV<_@056S#UK6zYGR{NZw zK48)dOnQKcs~@FDD})o~E#JVaS3}IX7Z~&bBjbBGu;_V~Io8=aJMok)6Q!$b;0w{7 zYUU^$ z_?vP~2wD!BJcoawzS}tu#(JQ9rgd2LK}9vhHnu$zA%dsvJylSk=FA~foYn$;+^qn!yGTDA^ zS*g{z3K=eZOZK~V_P0DZE&4Cu^a*M={#=}%HVRHVXe%2}esL6>Mw#bzo%?Rhn`|x^wsFt zG>pU8G(Y4#SH>FS?-OjAA7E^T$c8td zr&eUu8y0^vM)aEWc#UAlxQv}M5jww;xKWU`aW_EsZhVdL|6|7Zv0I3%1&2x-vBQ|d z2LA5AE~{J4c`(p(ot4v_c)&_t#&s{cE#7XSxBX_HfMzSVE!q!07$>oi2YZlJ9muA{ zGR_leV{8etCfeq=b3)0e!*6ls;s<|G<`ZK+|9~x{d+y)jp}XLf@n4mW7A=p&aK6rO z@%xzjux&#NpUsMES*2n5F`xVO(RRg`=V9+?9lp*vi?-^FGu5Bn`eY&j9NU27eT=o3 zH8BCnK-)$bX(u;wj?)hQ$@ZQ+U-Hf#&faM9mEFZW1|H*_^_j$YX82h11Fu6z(gSuK zw$2~XnKea`S$=c|I0*zR%UXlhMemo6F?=0m3^B$KXAIx}%<=Xv^nYFR%y42pvUq_n zNj`MHb?N~U z+v51?*aYB2_=!VrJvV~)Zg36VMWMUi7p!Cgc=R*B0pw26G}{lAPcC}=H^$@QQ|%<0 zh*{i7A1`3{brWxM-y27)X|80h{oWILk7sMfA!m8!$I$EUG1m4RVkNo6H*<+^DyPSs z7WM+SttP%nP7k?W3f`5!vH3X=2_)E;3)<&16Y3b)#fJ|`}-x-CA~jSHHP~a4cphcz@^xT70Gm%6-8cT&B6fh6@8i5H1?ql zr+jl3te%fC|2VKn2*=>il?5ASxX%V7wIwIly}p^y!JW|9L1+xV{^j4DsDr=#Eqs|z zSaV``f@|sdgSHMgdj3@F@ZGFU2%Ld$Ing?-yr6Xc#;Ix-xby7a;%{@XV|%m|Jy`@E z9#LP&e|R$9Mh@{y$lFJxCu}_(EoE*+70H?M^EFR?i*t_{T=FGgi2Mz^?-ufI^kiHPUmBenYAT~4j!%*@H<;Iw20kk)WjB)MlGqTW|Vlol>kbG%?&eVq+ zlkv)zCc%k*1pm{4|MxQBpJ3jr=|jMGh6ld?ieFNZ318;@&-fad z@YVWJ7cW1M59i^DL_2VRKHV6yUR%CAntWdI6b+oESh($o?`h40LBub{Nmgzm|O3j817HPO%5SNiiMilo`;VkvsYw zTeu3{yol#w^fRD&%k}^g0jT z*+d)F$ctwBsz!!1Z@2g0%w`wbc{1B){4n3~?Ws33zWqA+_K{1Ew{HZeY4@$qhPQD4 zdg{4r8}Y-aS*)~XgzNS~pBeA3=Q+C$R_oh1Go3X(7XtU5&E(Xv&w6G6!^gl~9(oKt z%37JoAor99IOO0{eun(`IPh;`-?=Nv>m)W?$!_AH=X|H`NRi^qEM!m2LUKdhR^i$vazh)**Y|axUnXG_P4Fex5SLlNd>Ts%!+D}J zQ@d-Ac?G<#HOxAHt@d2;+8C3cDTD@Hxoh$>6Ogx)7-vT)c|P%#UcUG7{Sf!P?XCXH z;r2HlLce?F_%v}9jiHJ$)ZxEpj^P@{puWr)WScr;aCJ4Zal0j30$m+LR~NJ1!Ful8 za7A9auKv=aBf2^ln=_^Nfzi*{OWqTkSm(wjz77paW-dWys-DQz%U-+~{Gpc=gVtJ7 zlQY2v>%|6>9eU^m)?C6b1s5XgiH8-Rj^Ez=+3>^NN88Z_>)hCi_(d_X5PeD)7`p=g zm`=YNW?RY3K2vs*Xoq(X!LRD$?9)QYbH;^|)QN}BEeIvQ3oYCSd{{p^L_Na7!STN2 z+mBhvw+aSZwx+4 zPUO5}CYG1R?@!E=9~X@LfH~^AiEAcLZ{?Xx-e0^Y%f7G^Eac^xky+X^ArBrdL>7-n z)=dZ{t@lUv#<|cnlG>Ew5 z`QZa!CayHuO2$e}d=Fpili{(@nQCmXP1dDslM8_59Pv_7iEW6*hJ9k@?W&A>o*q z!Tg_MF5q1FKl=jfLG;O-D0H>^C&VLw&2C`hMDh6C!o3~7Kg$)$NcFN>UFI^R7 zqbux~G`hmpIT>Aj2D-BIaW)^x@0OfQ%MD`xHREG?pA)Z%&%|${^E~ZkK@6tod%hKq z&c<)P+gi~KE$Q#(&shth!-+0jdOu@5So$gW>{)2$Jn#=(7JF(GI>>ieQ|#Pm3;WW< zcUhcCVd}cpsjf>lh_T1G+Hea*-Xurn`Wr3A9?LN{p*NOKZT#V4Y7B*k55mE(lxuvK z@!g8thGtazoAA))*Ir7?Z>%fnaq?5Qrt%vpot1dk&TkA+|Do@qRa485g)ZlH8lO3h zEs?8E!AZ$11`fASRa#5EH1{WbOF)DD!zF%Y1T~kJ?vmrG7Iq z8a_!}D}PYFCACS=fMm|E$f3R;B)_X1jZ;S|n#m@=U7D#Dj{nFUo3i6Lp}L-t9OMX& zo3h~ek2a3~RygMS$#6V%{fRjK`McR=&!Q|a*~2_X{Dh*&?byZ9v#|%BW*)#ove%Wn zJ-|oyk}G?sFlRj@*(*ChzI6Bh zE!J6AO@$u+{>eD+xG_&_Yq}hB=PMiLt6wE2t$Z|N5Z@QeE=7(g4`b?SJhJyUj3@Rs zG;wn}O+?K)#!rU-+Cvj9T94%N{mnK_-0H0RsFj|2+op-%l5^MhlhMSgyA4fj^Xj9k zJv5X8%4@7a zez-RE6lX1yp>JfMn|r^8xkz6d9&%~69i2G-M0;AaJUiQ*+HE`9@7U35FD_dUAJ@^@ zS?6{sb93$JeK)7a4EFwRM2>iA=9(-t^Fws9Y6I)8u?|Q^>%$t)!!y67nAyl&fAZzw z_ANYHbilVFHqD0No&xKleqcE4z_0)quKmJyUyD5PgWqx9Rb!O;f?;A}CFmag_VPLR zyteimPnh4Z;TBT6T2}rKz7-ohur4vR8HM3YSpS~dU-f1D7GNZMqlY*_?1Aibr2UG1 zb+(Zm=`{3!4R4;9Yofcn@wKf%j!Mq{DmZDDZxda|AMZ^jcu3dA#JYbp%uS zDCIfSczl9Awx|K$Kn|h>Kg6ZM(ctpKY;ef~F87TNm;1B9<-Sqil6P{rZ1%uqOYE>=xxN`63pnKh+1G2d7KIo6h4Hve%S`Wl}_{(_Os_$oep0p&`` z?APmBzO;_3-8Ph?}tW% z_nlebJ#6d3vjlIxpA6npzjz|N?->Q&pU495?`VD=`E?rl|L?e0w8L5vqDC;`Sy%NE zF@NcVimOlLH|5&A?YvH%L7EI2g0^lNh1cNor_ahFgKipy*M?56&tBY=uFu?>?xypR z*PNBOj5;OBVy&NRqV96Db)yjgXZKEOC}DtW1no7ZK!%OT~`xF>%*pc zek|t#6hFUSx?mS)!|Xf1WBv0}pXc1Rj`c;d8y z65KB>Bp*XRQ^(u6$)0NJIN*D&*K}h5Rf_*PzDVM;BXsuOJK6bpCH-W|@tDtn?u1NUehcE}sq z;r_()Cx`oMza9ndZ`g3ZEEDeBKLPG_Cx?6KXGY-ukE6hSY8JS+W`X;+;qBYG-!#`c z{G#$uFR;c9+H`fMowpgWW#Zt;<>8WvA!sdV;sjA>GY0(`ejye-6WpmDVa;j7?IH3q z333Nbvst5v4JqEpFRnRr33iMxRC}gkpuSMena_^LR|-@f3l-O%`8<4L{?4mBR$RR3 zOzkJ=%3S#!uAa;EKmL{R#a@R-ug@-XKlcInTQnN}Ubpf0i|P!%;roMNLAKi`c4*1yb5(lAwUB+FjVvt?niyC?IVuwg0WH zwh0pDl6GZ^yJ@9uKvYQF)>LhkwJs5?XwwU5tKHfzbIXn3g$n;Nisb+Oz2}_CnaLys zcK^%g6K3XI-pg};p7(j57r5fD?yJR*)m{Plv5Mi3;eB)6__504w$J0;`!Mzb?|9pi zZAttPKGrjZt@n(!W=7EuHRNr&_Lw^^o3^xGZLdo%U|c4KM;noD@OS%Q-QdQ+uAG`d zH{3piKhQnAZYbZHsvF#KdTGp#>D%Kh7LCQAv6Y!=?2#7+;vUA1BF1a%sFfM?dd$}A zSA%cm#to;}m!3U9HwI4l=KuP3X1RMw2KrfJ@Z#t+a8pK}yB&XU{4j7SJ#;<(>78lM zUae<|cSt@<{?AMFH?a>p8Jo%?HnsE&&S@m3q?o}MQ^&Q>^}SF0<6yd0PGMv^cIh5q z>Wo3-nvLI-7I)Y;RNUcbjM?NZEgOWBeOfOIrs8B@v5S*sNu2C+*KA`4Y@GatvFiJ9 zIC+7(Ld3y0dgBgP4Kep!8RmX5a}V#qHwTvCnb@k+;mg^4&K$yKojqW)%D(d2tg^?w zzrC>Z+OVHsz23wopaE}8It$%z=-iRxTJK7aTNBURsC3*JT6RF#B(G>cb0PRhtegn8`H}hey8vQ88 z@aw7pGv2=cG45BJHA?cK8H>?f4;E0DgtehZS?jM~KIz!KW3wBvH}|=@1%8L-A0=j>oQJJiyJuYFlO}eNUsMh* z;&T~y8RMUAkNpg1?9sBIU88TV^E~qiGu~a;=S7^;E8J*4nv3RO>o{Bf7<`bs)*h&O zEu98iK0ZK$!pBD)d}OK*Kt5hdy$023xQZBn*6vkLL3vS;w}3U@4XlaTSm4j(eq&FG z*4^3)juDE`CJj;=jNZojRNloJbPw`u7i;l?mwrp;6$5*hUS6LD zz3fV*mrsgbb_Gi(5|4a*k}bDg7#J8kc^$i0+m;O3#ai}icvhcL;%*hhUiK6zMqV1y z*$MW$A4$pIP)(LCjH!XRLBoZ?SDUHFq8gae5waDei!RhST>Ap}TuGim)Uhw({Jt0a z;uidf2sCJ9HFI`l)Ze{{4a76rgC^Q@ZH?H?gKZ7f{)<4T-npgdLTA1%3|u|;z1f3l zqe<5oyXJ8>^er0tw)44;Pr+S!PBP{~hi?AKp&MNg%T!}V-#N>m?6-El z05P|a+Ls+I9V5Jo3q|86X{t_(@P6yswQ*N+Kl5B_S|5!O}g*`y8WIsEkZDbqe?{KH4#mq*B9d&II#&3Vbo zpCd0z_`S`tQqG%`PZWb6JCSEL4(+|EaddYBXV+X!PMBm|Co;>>++5_|WQ|KcK%jLg zvLc3FiZRz1^d#LCLx!v7q2?_(yoVoH8IVs>_L}yex-}j|&yu%>1_O58LKkL6W+OA) z_k+A|jo|$w(7NdD*1zYRM|88PFG3&cSN(bWa&0fs)u80red) zb<@J|L}UVVR|OqG*A}Vjok*x{rmT$jsr*ta~1eLSy%^{#wySolfYy!dmlS6njB2PUR$;d-E;~MWyeNnX=EK3;Bf3 zqvWGS;Rl{+tGIf4NH(XbrvY!+s(TEqB3lXGUAnGoKZOO{9pnbWAMYM+JjJ&{!EY1m)`(zHsWF@jar4?Hd6FP@DrgkKPJ5pXsy78uE2(V=%;pGsc5qa7`*7r-<|Jz*7Ht`88~lG z@-m8?><8}s{m52i3%a1~#d?3J3*2;}uT-m~YYm?~(}j%e;<;jABe^9y=t4eC%C!E0 zyy|z>KThULMpy%hnl%uery{>2xAiXkA|vbQ@4eaBtH`Q$WR>h2*_F1;vGE@K404HU zy63&_$|BCvfPFX9aeRcRb3`6&|p^P5n~E>ZTxbwozYAG1vKgS529AeR8iI zJ$)WDX7zeO@eHfj#Mqu-YX9JvDwvf-Z`){(^f z(CSl@oq4G4nsN=~7rJ9se8HX9rcCpJztZOurrii_L>V{o{sE1xma!4r_^5nRjVt{3 zcHCaS-8C=qeW7f0#_PkjzJvLyEv}orC!IX|u;`UBdB^CEuaR?v^~@HYG4$%dSGbrc zc>~-gaVD4d_e6f1J{{VS3^C_NIDGewG<{&pw;}j#GW-U z`Ng8?f%7k;+x(@*-bb$m7HOQm)@pdZ5Lp)#ez#SW7O`HSdNP;5kIGrq89R}SxVJEv zxQcyBtVe8I(lG}hTSzu#@6|b{C8rsEK_iHTe^}Z;C ztYqH?bYf+P0??1((^Y$Gx}Z(Ei3=Q#<63^Fvj#TXuM13FdzD_XFf&L4T*5g z?33kMFl~;OYFTi;Eu1?Yi17N*%8|j)PDR-^Yhu)t({6KN>4rVw>i)_|{o1fsMWv$z|3sQsHRel{P1XzYN!ka#C>xJv;j- zv3Vzl&f=W4+0HfjHZ`2rpz8%(_sYU(t*z7d-{_GowZYOiJ?ELN$L8_cSE}*8T7DY) zEu!Qu{rbOB?WEysho`;$j6o*C2d?d)dWM$gpX{H*Zv5k1jbR}E5m^bYZU-;Rh$oPz z2KB3xC9?MeeBKd*qxYwVhWOzvj4qyR17`05m^vM zPwk<0Ky7epXesq9iK*|rk3Mdvk9PKt`Heqi*Wua2e$nJ!vjrM!(A3BO(Vwt!_%8oa z@@EIJ=OFTMF73?Y8ug(&bjD)-@%0Mw*iYIoOPvYk8H48HbJ<4>91Nd2b9Tq&#p&Ot!^P<%jL%WeJgc)! z-WxnfPVz$NTlOe?X4^UEgvy^o2crYAUkr~dfk%7+Y=biJ!Cb|Ae&ls$-XIMMXNV&a*!uXwlO zL?!5Z(UkOou_GDh{R{h>RJ&a7DyDSTXHGP2w{kXgvzF2g@2*?E?ATf>m-?G|rv9dC z!*LeUPa?<((;x4Qf`=l=g$CLQV^`gyPx?yQ_2`DU$6kK#1`ltg(fnNAbxGb!@9W8r z8j{`%S*sg@PH(TUYws#XCc2S-)1x@#JZjuzS%(#SaebceL9_C8rs^6#%~qaM{JR-H zCQhwkt@YF`XT69#2S5Cz`n|eu^OKMB!R_bfVE>9gUP6D?kF}oK4S!U$PYqSVAH;Qb zsz#LB%0y?o-~}%&-e+ZRcoAH@2riZ_zxCLy3x z|MaffNaL-IDZDk=@D_f-IicHr#+)Z7d28b!-h%Ij;H@?fZ|zItEn+-2Z@uzFlDEvY ze`lTP%}<}4x1RaiRNi_Reylcn@UqbNY11<25M3Sm7~@|;ybQa=-b=gg0y|d&xh9)W zd{$)Mn;zOs-zoZPjIkk8_-o=Y{536|zxLSnv3&;HK)vorEi|$W98@7kt4>*;Y;<*D zflJSOY#T~^7=`{K;5c2M9H(}UqfgpugO{b~lbfOEAaYvuD;=Gc?BmbSui+V|4=>GE zlAqhdJNeo0oG+Lt$j8s~<#$jtzGLbKjT`3s@*CyvC2L$%m6sZOu$?sk>c-24KT!!y z8{Hva5gAK+p|n0@@KYR_DBs4T4{Vx+#ucNG4~;%BH2d7M&=kDm=z{^8MV}ab08bg3 z6)&gigSLu4CiQ_$vp3QCjbZHvRo&ntiMYFVT^a1TA^#OX+=mW-C;Lz-d zbO1We)dB0(AG)T=(E$6m@V8EO z@Q_LVpE56CQ>F95>o-#QPkbO7Y*=1EzPr5e-oHaT1Gdp%UVu;jJG@Y%GYZAG z4u4#C^@*mh|J})XVcy@^^1I&DmdgrVNjt90{uN`+D6`?)bea7T`a7A-{t0`%hT?}$ zdikN!me(sxjBX&FVQh5Sg~)<9@>+hJ_V(vsYv^p!Zr1PF+uQL7XC`Ys@&&&&W3tw= zp&jv$>gy=x7R5GJ{hk+0y^f;LIiJ8chp%JEwK%d-Hnr?%`IhnpujS0vM=gte!@(zr zXN=SD9%QD@Jy=1!S$kp>SMEgqb<5^3@isd@=5lz`ZO62YY|{JPz))upP_wFZJ>&Fo z{(3*>emq1w8mB$RZ#?0S@gVsTdf)WTd)_gwXFpR5`A@K(*4|j?wk6OgYOcEyS+OVw}a;V;W~o$~c3`akkU8#^@d2CCXb`e#tDw zPvw_rZTG!DIQ&%U#3yo*vjynQF@tseQeb?A#r$eO1@A|E%!b=(dfwm0(uNoTv504Q4zb5Bq zIQ)+q0{=nqU+%^K)t?*^|6Bj+bH))({Hmv~~#Kt5*_yFOOSJ_~0} z@BsTFRzv4Ytl)FS?GrB+tPafL8g>`gJ_w^*BGau8q*vnT3+0%wH)3NAdn2@0qJdbe z`CVY|mB>=;mFv$#d&;HhywO*xdhnCsyD8`+zb^-U2oc4Pu+{Hotlr_1qbzJy zesx&!800vzug$8T6b&8)s^amCzPjPkLl{r&5X&l z;fy@~ZAQNve{Fi`IocMS2gS9<)9xOtv}>;&H?CyQ(IA;T9=hGb8rR-n=>(5XHvUbe zozK=%DE>Srna>uvgSmf&`QA?ZvghE{2s|6ViC7oBw}-f^YGWw2naiH*ZsLwQ3(RX@ zeTQ}UXf3uIaERZ8?S@|Ks#4B6ymLE#GnVzdsXeV-Rv6oo{%9}4wd>WFv0*He-x7x(SCiv|%t+KIhM6@z6ePbV3?4sL z!g|CwYepwyP)_!C+S`#P6T-*?_JNdcMGyOt2@%?JmQ`rc$z$T0z%{a#C)3Fy!9~JcB)fbW*dM`p7_Or5wo!A2N`P%PY z`!8UxPjM|19@+OXV;Q*V2RD)lbAjVL;3>Ul%TU|S?z*16R^YyeJ`C>7S=EYf*m<3@ zU*gQK3p=h8`0S?kM>l=;&{i*P9t1a)zPt{dsh7u@dQ*IPDKV3<+B5NzJQFX`{*ND* zrNvBWZw~XC#MpAp7>UWy?o?kc?dErMAK$RyHs<2 zjMbT21l~8}srR!MquOtr{pXomBXdi!*BIBuY4GmvAJq~D@93FC2a$s%+wp}H*U46ibf$t>--tWY~b*&v5nhLBB zI5Cg|P7M57Xy5>`?Jg$GBWVw_V)mh?L&Ry_e6jJFb7%-VlhsGad@E#Jo)&`nN_nTQV^I05)+ z4M@7Ki1nR9-UC-j8e;wQsx%t9fceZLmq+hP*D24{3-1fyjj*l_<2!z5nEC#kcfIkq zUbEgf5FhUP3jX|E&^ofR6J2!>I+M)og7&-NfiSwtjnTzcqPK468vIx@2A>I9QSNwG z;7sJv8Ld6Y*D~^_uV>vsbvO&jv+%RVqPEs zF{HyKlK+0whx$_PkM=>#fSl&lyM>)~*X`;aO zvwv;tIqG4u=jMt~>~(ni4S1XyJgcGW=(e0t(8OAcL;V+>Xj)QY&GbGShmYbsubMqu zdER|i@2$D`M3dxyn0o>4VJo#Y1am|2?eO_yI#bE%^KpLDPV7lOOR_^TU@Ds{23*li z$dup{JquN@h5FLa_s^mED70Ax{A4S|{z9Gp_Dm4TFe_$+8p`SAVQkcT`MxkGTLua%6K zK4Z57H(+e`w;Nf)cs;sr_G^yr3*;GFu>Q`8$CjWgWIHN8r8)?4ls6}^11ka z5#K+a!+9(`*9qS?Q-5MNzWjZ8S&iL{vuAWpqwg$R52@a93pvNnkc(0ju*_M@s^f2+ zVI}-Jhq4kL>9+frNxQLbD&ZB$F& z^n#ZBvydIN&=EG(8SnKk3QuQTw6_}hJ%Rqh)9v$8b!{uOd!A^Q@0IXpEq*_B0a@=W zM<-Y_tYdZG4nyHTMvJl7`2#k+#FjyCvw0rcE6O3KG{<9~K!^SZmO)8W(u#7f!{Q%d|qy7EowScyka12;fqO;CF(eyn0SggI+Pt(V<{2kyPy6ZXb zy3w!#GLoiM+eWjc5_Ki=tyhQZr&)L|BoNW=L61B z=@P~szOKJ%9|LmhO8iW@RLnEXJlmNk`DFW)M=gHq@4uj>9G|%u`iuOAIlhWrCOiJv zrJ-{6)-uyZ3=h}#mUr}lAnLr=3S@p%CE71k^y1u>NKzYg??Eij?d3yJB zm3#Jd$qr8M&)&~+Q8Km?E`-)ftc3ql(27I*-Ws#S;>~$+Q{mr2eCsgd)&As(dcMTw z31TH}HRQ$WvzENrtT}dD?tajE_y$=zL+=%1r|P{Tu1%(HDl#b1F8yeghMpx~2Ytl3 z5K~Mn7@4|n3AD8YnHm8uiWS@dP3U`Zz%q4|y)bFlxxDR8K4ejJhxTL~Dueb8EF< ze6D3+x~#c^J(anAThXs>Tw^}*0Zn(`is&O}T){DH=8dJCABKwIV1F@xz4aAmSBUhpu8K^bL-s!|YkT(mESqpqt zsqkG*UET=qxp5Q4OuC3OL}_O(b=swu)7FhlyX~~=jRU2`AsF}ZS6y9i+u*W;um$$H ze(3cUZQ)3*1jePq7U2l+oUua1wj>EckC*X1i7~D(u^FUHF%sf5FF`##k$n zj!`;nn*@h=-?w_txS%od~TzLz3k~&K=!neiPMNj zK@&5HNo}ma_Kwo7{IHnXM#iXpTee(5j65q;(sI0MH*HtYc9gcmv>T({CAQsd+En|G z(5~ps+isM0qqG~PU7fcV<~`ZuamL@t_~VP2%T3&`LY5*6*{eE+v z$DUwpUVGv!#+y;bz!RReYhzD*iGIc7@%iq)TN~bkWX^c#$74?%sT|N%Nx3r~nsDri z_h}`I?x?};&|Zos^bTthCO$;l z-(dfl;*Pc)_(bR-@|Nxqe_E5w9u3(T^4n@;V*p>-7$1dB^gYGK2-xx1cG(!!&b}4r z+62a~{S_68S+b8}=l0Zn6tX?)J^1`FdLwCX82DX~gr6%jbSUCNQi5z#by-^gQVQoog7q)$C7nH(E%9npEf6Pt!NbQWfu{b}vsOY*fI-JWc-P<~#PIlDYuV*7K~e}F&7d#mx~b}-H` z@3_y$fPwC9EvSEGYf{eGFg_V?eAolaOuQk-ED2v|&1+vL?OhLj-b*dk@HsAxOb=ZE zjTk#LSTdv?`fubL3}uJ9;}_ojp0TFyFFjmouXnzT?)I*CYOOPd4DGMw54@Y$7sIMxTc?C3qf`2~ z*zqvg4byp7H0jc+#^|*dV(6}y#gov4kqzibvvzw2HDsjEiLnM9eXe}385aOo&SHu| zD>0sx9*me0dU`(ACD4z9GZ5`}hdz>1PU8+M$o5rW7O{0TpE2)FI zW^^#mJa(-|2QxP5V8;87eEKUHuSW;BC4G9H^XTA>^65kLE4mEKb9Ar+ZpQ!K{6^V#cfZA;Sp>@S0-k;u1SFlN`MPM2?5AGG&N57-B?4I~F7FBGf2#TT3* zyH@e^n6YDXTWuRxd?nk~wrjOVXbR7XKi%u^89R93J=wo@yl)siG69(uvr4yT&CD;;OF>LSdVCk2HjVj;Mrn)^=FwAd42AC?(^y| z@P6*h(X>-WPE{l0DT`aHVtbn()Vsm<=S+QS|cj@%rdXj71}-~(gxe4f}?9N8fri>$Ht&|h=s(qo-txW?G3v6l*2v#JNK z;d6pK6HL6qybXVePx-BQnCQ&Fta4}{LGXJsIKXpWntj}Xr3+8--~4g1Ee7fpNT)}D z!5rdQl8Md4^tOO^*@IERPP_^3{CFIn?v8TOpc>{{+*qpVx_d@-MW!`j4~kEy4w12c`Cf}1 zA>Pv3Qx)&P(~Gp;%=NG8y7)k6;(D&{)OF=q>bV+6HkFe*J$0<~)?m0)I_GfrXgrs& zCm+~UU6?wa^VJVN(m+45!2>4G*B(#Mj7NFiRZ~))Gvkr$sk+>W%Q)l8G{*KZna8+= zGvrlI_IZ4n;`UF3c27Rh^pcgkA@9w*U(GAFLM_XeAKO9hf1#Cc_FcXRPvmRgCHxkV zAA!uC{|I`&1m8>am<_$~nWAf<;g6@$@FEZGy7cSr@1B}g-!8HezFLMI>e-uLjSNny z$HFrf{-ISrP>qgqra?^0^7JYrpd-D(M>>E4#r(s>itkt<>7OmmfACH%y(TWq=K*lWwU zzcQHk;|2Jk4(}s-om*ep|uYx7F~lJ=w_Wza_(X1y#YJu zLUM?-mg%mcnw%H%1gBub=PsWzOLT7gtk5~SL^)*-SveiV>^t@zU%R22+{)Gb{?c36 zO|z|p)?*8g`&KPyEXohr%KM5Ty7`fc@kyqzcYUY)VDDHGSJ*Z{_@?@um(qVVdCk(l z2IsUPeXV!;-?pFYoQ*$HM*b*caL2DaP0dI3xon+c$I;0}+WGv4umP&-cNSW{mEY{| zU-X*z)8x2hwf=}aC@;QJY~u;QAN*bK!Jk+EC@18fGvIUmg@f^_c&~8k;#0n){7}I; z0zQS`&!fA#;IZZ4mYN5xWryv$rDX|g75u)a?7P;gt=J0vJln&b_#SdrddWHeg_YYe z%^KDCZ+spKj3%G*n;Vt^n<}fI?;db}5d0_~Dq_xK9@Xm7z-Hjy4O|Zb&pz1{z%jk5G{_F`yTt}S8{S5{WH1k&tzNd zHyUe|ZX1VBj-1&pngqu#&bzqQOFn(s`oOB^kM=KmiF&2Ysu?9 zub5Q^pKPUmukO!BzOGoleAd(p>b_mJ)xRq8Zoh%)6!M^ZnET7X@iL#U@i9L80tLXc z(SheEvk$G;g{S13SqsWf!t`1ProF&GXNxOWL-03yrGRZJO!Hg6B$)E7fhn+!Ge6n0 zUbq^V!c$(DikD)*G##Fk1W#b8b__g;H))@MV7cRl`(DYr_R&A@7;8O~gz0EtD!-!& zTUq`{Z7^{YG~LYlTP1q?K49xmo6WPV4W@_dIj`IrBZ= zeE)aPGr!^c$$0zIpB#|Q>AYRw@ODx^@3G@o^Jj^_&q~HH`|m(TZ_6s}+ysx`4^6B= zpVh(Zjm&i|dK=l>zfk%c9gHsiaTdCu-F8@ ztbs1>NB?e{Q`%LB>{~0C28hAfXT!@jnkqTM^L^m)P2O>z@8R8poWP;??UuVlBFy z_0@0r0*yCVqdMLSoXI-tKX14UIn-~R(Kiu2a0q>W5W7XX-rc+;EEi`ItUe12BHHL6eg^ANdkZz>ng=mqt| zIBhv)#`!76nW>*coGT?dB;!Qnjioe0JyzR?|D0B@hNMt8gu7~Ap9 z4>oSN+!_rY&+PL9n;1Aa1P&y}Z1|jM;N#*jUpNHb(j$j}w_qVSyd`)z_Ji}ZP4MT!LHIm~EEPVlyTA%v&YZppK0gmWe+WMR4SafWITRlMi+q=%@aXn6 z6fP^l<+qThzgk{0YY+7JBjo9s$kQ(%Pd|r_y8H|)bP#xoUN!GS=<*NYCshQXRdA!gYkdqk6-5pxRx}SeqxS>Elr60}^L`xp8Ch%@Um>=Dao%Xl)%Wfo4g!2F zU>jS&I2JP3#XNr_?V$sW9@%62CXxHO&zQ={H~%5JEe8Bu*e+qtU7Lc57lCb8Ag3ee z_7V&Vc+Q2XovWrgwt?0!OHc7! zgy*BQsaTlNe_Ye=s54*eWn$gD4#Qcjdz5xB7Xe5dY=N71YahU@15Xwt}e_p01CRZZmfgZ$3 zKigjmn0%L)uTi`%J0;(xe;aYOSK+tcK-(QwBEF0FpM^&Dg8#>@MD%HB;Yo0bp6Ra> zPRFurcV^3v{MZ863D^R}A96yG%{<@Cv*aK|Xfq;PfOiJ=Yf*>(X<+y`FnrRc#ct?@ z{D)8&dbRg)DUPf-vGhE)7rLyoiLpSB;-l_eLGrVFeZQIn-z{UDRg4eaAAJ@Wv#&IU z9?W9@0_!=UtP5=!rZrNa7C}};(6dqaRP-K$R^mHvX@>`5H72^r>>G+IBAAgQqCgT^v&+-|sB@T5f zc5Y{f_7wWCPjiiZ8ez`zUjzfi+g)D0nfLXq+AwQd^evw2=J^<~s+RAwmutMU9-fFX zX7X?GKb-a$qvW*ad>=Xb>2qRjZqPNYQAL1rJ@pe*ORbI=%tO$!Ef?&%{@yuG*BHT{ zVjaSzf%jtL2kq~{=2*9U@+|H3?&5nl^1c6azE!H9*WYHXBL8qve-HbG(ch2Y7iI3t z_+bW~=MT}>qr?>F<~F>tcKM_`Z>2BwVG)P;HuD?F9Q+1P%Hy`{EV*-d6r8PPot$=O zx$W#81{Uv-SEA2tM=rZ*zi@3o_JzF557U0#z)~D zwb$#k-OD@iP0#^4M(5qwbdb|3c^U@(5#*qBLTpno@n689|I5Hc@MhnQ;4K=igTHE= z`PluM*o%SnbMz~mxpSpe}UU-$I!z(w*GxrN#Wyl%_Uc!T5CZ9(-S?de~@XBp{7I_?FZt%^1 z$r!_j3xFMM$OcWY;976_8zqAxzX>Lm37*q!c*;JCcwkv)S%<$RTYWlm z^i{>i{EcN+PTxc1q(o~OldwivwwTh-ueiiggv)=z=AE_f~m&*>b8YR{Z6 z)0}4p6V<%u<-0;>zKSt>=et|;yAm+ZwO|G~fTa%$~30esaFIgD>W5^PXf2 z?=FVF#fQv$3ge5T6Qn=2zt@Wszs5J)!O2bFWFhq7=!GOsW~YwP;AG_>oFvEi6K9Mm zI9X}qME0wT6Jrz5=WUEHFNnS7(5UD$2F;3A?Kp}{}nwb}K{%l;Ar|Ms`dKCU$->kbiV4_SLbnrpAG$B**no4I&f1iskE zyRQP=6*6zF!;~@4!ZbKfq}>Y~IP1I=!P$lFL||*qQUSJQz*cL^-g!9dFytWmL!bM; zz3!s9EMYF%i{_49pYC|`)BKBo$ERTZqQy&-0ep(LSi2WZ7UEO5w5WU?cb*4tNu^8q zL>fcTfv5b7l>W{6Dg*TA2e(nq@bnjP23oc?Q?xY@Gq~%CAog|?8mi>}wtzLWoaZCh zk6}}nZbs-baxax9;bULmYV94oyTq<5#5)tAFV)W5ZR)5@2yyOh!lm(tpz&v!8+KyC zg?kZa$y8%Ud0`wp$?x>%-T7P2{CB6#e|}m$sxW#kn>EKA@SwF2)vMcu-l_pl;I}k7 z7kDm^tW*qGzH_##XONeZL$^(}ZIXw`lh)jBg?}GqZVk@d9-=+b)E3|nMJ~&amE2c6 z)ZV*YQJMq)zJPv--g(I}t-ohQ8(yih$a4T!F1+rcP4U2Ov@4llXf&K_a%x4FJq7r2 z(~%Fz2p2bd$On^OvqbwiUif6Fg8EO=-^v5Hg|ns~*)oQG@WCgRaliZSPll9RCmZjo zXF0O?UdvDxMLC@xG-JkeLpw^OgJNQsQ&NtzTZB?`}lL^ z=qJJeayCUO15olZh`_`zyo?Mb@gGA?ReGcTClp@#Kz0c8V*L zuHik&B+1frxwMYk8XD){%za%n34OOKqYT$EUSowEFN z-W!U}revVAzzFG#{QG^~zaJl(M|?+gwrXc)I=i3#PlwJ{JL4WHo$crR1eeYpOw(`k zuS&&_>;*seSh}51?euB*af@vyM4=l$cI;|bpEUykIKKhCDX-ZqKc-w+*gj3`|DyMKhX%G#)7WcU=pC=V z*y+q0+T2&+!PQI;uG+Qlz?u;{djMB$8En{nt_{0ZxH?uAT24I$@o@bRxcUKopAN1z zI%6IdSLC`DJGgp218na~ft7=&f>gS$lifK2JT3ROH3B@{?0w$BQ#kPT6dFm#(~ZuY z)A97HyaAqa@w9`yLx-n^#S=WWue<0&q2>4@!qa&};OT1mJ{>$wa>hIyp3>)e?hy0b zlwqD<7~wpXU#I#-n&Y3I&MY_Hr&j6d%<)a$JDJ?@>iSQx_uthgUm#Xz@z0G%tT6j4 z2KGa``aTc(weT3%K*<0jS!yD%{$$H z92eT*(9OCbVEX83!t^^CU>bCNLRXi0Y{4Gt>bo%gAv)THXRpbnO@n8uEtDcZ|A~H+ zvOCF_?m4BlKU2lo=#u@8O_cKft={*yxPDC9`}NkUha8-xgWBu=drU>c=(14G%F81|8!_#yEE=#X~M(9e_=0+%ft0U_;d5q{h@Zpk47#|1*a3qdpCadIBQi4 zxSg0a@Y~~Or{MLgUc7`yj+d3g;H8IpYp$Q|>f}9JN5-q$WjlRwR%jk|X2dgB4S|;% zPZKZYr-+x32QNQJ_kaFB;Nx5`J}fsDlgekA_#}52d^BXhN7ZTK<5xe)%qM-+RX-g* z`4jJ*iv956|7CpqnDHg$UeU<$vCWxRCS6vP0Uz&gIX(VZF$6x=o+duNdW!gXne!Q3 zx%I!nAJ=>FF@zq<#2;nD;Nxf6gXLAzY2qV$2z->BCO*zRMSR>c3_ez5vIpB!V&!d$ zBe=f&pIP^IXuqvE==$=hvS^4{Zc000Z#&6YZfZLY&f2`X{}tZz=>9gZe|}>s&i2Va zzia?!8#DOlZLWWg{l2k#bXjtZ={rxQ$TTzdt;G9(O8>*jv~8>bcw#ASZY*VoGwzY< z{$re};PQ~pi*xOFJAc3vzxC*Qk3LU<^}WNv+F2taA38NI>#;+A=NZd?rRlpBCIcp3V*1@)T=H%Va zul<85@#PV~_ruxMa$pQ9vEl66nn|g@>iOZ`e0aE#$MXjz4WTk zOOCzN`DA8V4$%MU(DG5%u1`fjU*mx_=b@+i_vz~q9(XGk!T9?lpod>FrUAK>h64xx zi__`B^}jOE$N8Qy>kMF*|5jvLzx_97tyXK4*xW9zy>jzm@A&F7kFQ`D9z4Q+R97DS z$upkQq1V;ku~esxWf*$B!@;{_zof|Cci5|tPOlw57#Y2;6}^7^^5i;N=@9g~nEp?P zUgtaG9;xj8t_RlX^qP-9og(i%^tuE8DqWWSdtzuAROjiZ$KeabzVy$fx2fs!Nb(im zd?-2_<;87ySlcv3uO25aU?^Gi)eN{@G;-X&(|yV4P!H#U2sXQ)$SjM(^nW_IeZ?8~ zNO3#S1M}@3m=6o@)EGvJ9{LgeC*_Vc%E=u#xgE1*7R-*V#D!^(bm>7DK^ z1Mt3e2zWP~CcOXY6yg0M`&?bQHDY)p?}wuAzxBeqedMxh&M@$PUVBW^bnW*}1>P5) zk)-cYL%@5=X~O$MrwH%cP6yr%L&N+3vJRi5@1kL2K`HX2hxdk(C%0sPckoo;z2h?M z^t+5a`O%Lv%adQy|LMq+|8T}VQh9Rb>A<^fXm~&9g?Dh|@c!m7@J441qVM?kPC=eb zA;-mm_umcy?=PMvylYMo-cPWX_LSwxFNcQrC%y2l9yz?P90uM$B!3}Yo~%0+c<&g1 z_y29nOyA>94)1*)KXjw(hvu9jygz+9@J^NA8@+aLl-v13)Soyut>-Ybb_dN1s zqu1~Mn^btW$?w11fp@ubeh254Zgl;AV|z})_H^j`!BlvgvCHp&hW<|?Pcp*0)fx9l z<;i>Okv(PlPPI=`;C-JL-u1)UgDLR7a~OCZ#ZU3jw}JP*&9*!l5xl=Mpf5i;1iWuP zO?cmMitzry>A*Wxey70uGB3Qhj2zw<4g>G+W`OsXPX*rN&P>`T(Z^HdNs_*;(}ee* zS-U@#SU|=9FuZ?F|ED8Qbl$*F^5hd4;C=pS!u!4<;QcWDpN>5F*HeUd7yC;y$&+ew z2b8C6dz+th6Zza5|kfqcvIG3+As=M053D?Dp1S#Z)KL>0g2aEVW7(ldz{yK@V1(yZ4H>}QFTu+L&Lb&l%UYj6hpR0{3iy1ooMt!T8av6pBa zH7|f4IVp!%5LaMt(BbFFcliRe$)53_#CWELeIZLfLp^QFVn4U9Zw+gY&3xwZc?&s> z*IK}cS`7`VN5-CpTiIXqu$9$Fj>SgR^X){h6P z)x&$^1>LSqUzg~^pR&)% zcQZMr^jXZl$OwDKh?g3jH0?polqcU#dwKkEt`Yuj@|fok~bbNn4YvjzB-2AH>7`+(f)aX$TK&*B#7 zB+5Kv6I9nS&=F(5jPOzc{s-zFz-L_FV;y@s3LQq+@2Gm0?5lffKA%}UUkwep?HjmM z+BrYkzqdyF_jq5m@wEpl2JExQ_f-wlgY3=oLxbhi>hj-BzdY-|o4s4iuY!GiSymtv zhvv06R(rS#seSM{VyCKwPz~<&J3_4qc#8Sm^+XS60~J!=vYcz$KdA3PEHNGY9S0waV(_cZYHE`|VPBiXwVC4D=@3XEh9T@;d+85;q zMuT9(b?qzqYxXls&s>`(|Lpy21LyCGmPJ>{lEc3Of7$S^WK!16>>p&Vx+i#+Aj7hm zKQ(5l^OM!_B7dqoqjQfU?W-Sbp#IFGdJnl2fp)c5&DI?`t)1|tfj2bi=Y8!>ScmNN zA)BbXYxl$Tdo5qbE-UZwW#|^s{+eTTuhv+3A=SNg_13Ku*uMt3F@k7r**-fCZ9jPcw@-!2V$ zdDh(@I3L=d$Derb20jas!^%Aq9{<4JYTse~q z^1B>(27!&{CtR@yanW_ac>!a)#b;5wFuS9G+J%=`R!60;pz$E>&%Km70+#iY1s_(; zN9(C;i;-6s(ayP?A3l|{YPl|XIOnbYCOuz%fyI4WwyZ?9P^0Q<<)aA4*YJ*XYVcfO z3!UnW^*DBtVD#QX+JyHlcs@LnHnD|ly<^)WrltY-5pB%L#!m84w}H8&$Zqa$;r%7F zt$GL2X%n112m0`Vhk{bmS6sK>K5KWLm22FCdt z##xR%?~HR_@v&LdEPSSp$2JDtj_-mJ=SEWaHp;x6xccE8<`b4i_NsTpUX>g>3yfj3T#fz z^K#ETqwu}#By&cWM}J9PtT6QAwMT-92>q(|ifV#r4oAOuyr~@?QC+vrX4Trt?z;`W zYFYK)7EOI#I-fE69k^K3U+Z2|QQAZO(H<-B7xR49jO{S4o;UNTbotSlm-9Qam_Cu) zUE7%ZW@w!K>6TR#x|bYg?|Vai*Sz1Fa-!*FVCLe}UY{PApZYNM>F7s&Xg=!C?d!I* zzSQP_(&ix#tq6o}`@LrUd%&L+Jw{5v>16OQ4gChabV4^>=mDJ}sd~1m7ZsicoWZ%^ ztFz@})Y97woXy8$j72_7knx4*LPN+0)rvo= ze8stpN$*ZzoV=rFyBLe$Y@a!9*SRqC_l6t>vY|v*z~|#ZLTe{_N(5Vcv$zQ z&_*134MU%6#!|D6@z-GcOV)YEzf9w&f9Y+Ff8G?vueoGp^~v@&bC_+PFU%t$~znx;n8H?P&SE1fmM>ciFyUK!zpIh0)bL1yb0}Y;g5dIfGuf5U=$^Rs_))?kF zFAvQCPF>SvQ#HS$XVkuUe+PR*Cu;7re{V|r#EG2t?@{|hwA)Rag~9=}hhMv>aJ99n zM4!}Kjym)$=(T&hg7uivb^CGOXl zPikHAeXeuuANaKI=d^y-bHB{<{o|hRrJnCIJm05#zF%p++ckwgY`!a|mD73|-}U}1 zbIrD8FXg-RdWv0!9kF*(s*Wyn_I7&hi4`W6Fpxhi`@KBddWu>?rB&op%0JN>fcUGH zcz>;l_XirK*CfM|wxMHR+jz5DtB5aYuiR4U^LFwwd6pRD{y6P4&`tyG)WWA1!*k6q zK&SN69H;ik5YH*Ln%-CX?@aCQ#*TK!P)pnD<7I5&D(sCu`ZV{*m)1O&X4zv+8H;R; zUTMP z0Jn|`;8FoxYAkPA|69>*usZSg!2qeO}0cuiyh`PfQTKc=K#F5IgQ_cW;oqlu%Sbw9RdX|3b zUHTL6uQ@AJS7)`nzsLCuoNay|DH)UU?q2uZEBuKl_oB1?iO1@Qh1{OCGfY1b_+IN) zk~#5MR;ZItdtJ=lr|q6)VC)KNrK=WoRA+0vN{vDI@*sAP&SRCW5I<0!=tf?5PAgCJ z-ftxi-tSBFG-3{ zhFvnfJkfPOc6B3l4ezJjCVVx<*!Ofr;+NF-Z6|ieacQA~v)cb~DDbr>?(-FN2q*TL z1h&1eGul)ub1iEw-PNN)E6H*1q`e63#b{fxF7>JBH1_~>7N<71&X12V-mc$Ri3np= z{Pi;Qf6a9CaGjONI}3Z_>sgz7JFLXjK7ZqpC;f?|d(KUPf82rp!6z#c*OfE&_x*{3 z&sI<~-KGmjWX9c+PkN5mw-2od$B!c zGuDOnSdTKNBWZB?feRPb1%%7n%2{iIUka_!ecL8$J#lnj9-l`_My0@WoJ-GZebjr; z>Npq})iL#h4I2c1!Q@-Wv-KaL{@ItXnS7)AeEL1HymUiFW$9Gks9#L=1z%Z#k1SkE zw)P`WR$$K_Y4jzIruFrIoW6P+%M(+1Fmj^d{hQ`gz47m9?xDmg9j8B*6UK-o}hhBJ& za}Es8MY-?tS{HJ`@LT<;(EZ9Q2HukYTY^@i^Db;a=r#gf%f7r0-J$xvozOuJG|GAL zo4Yq56Q%?cvCH9uE15$9G#E@w%0=dI-7@R5C9KmIw*KKg?KkoUyn`IQI@>-=*tAVw1|DOxH*22Dndc7T>(iOvWagL0n!*(3){Wd@jEUGD z+T6>SIw}(HA0UrG`k~NoK9RTZEIAOY_w9`+Q%M;2cA5#er96EpV?w zpIuR&_@&E}ysvlFMi2ZaUj!XoYHT#xh|)%!Hbf_?Bi~IMJ-pvbJ32ShoTGxB5Cyj} z^jsW0)cG|1Jjt`*`U(0uxQ6p-pb5z^yS}u}Vvvu6y<%u+RBL#R#kw6C=|L~XI?59Ve^Z{QT@EdQU#*=-Xj^jZSL6dkxgX|!1u`qBHMVDwRlDH# zJ-oXY*mn4l8NmNlf1>+!f1>9N9{D7uj=3o*jE5_xo5k2KEuY7bBy?{4Qd?MmI8kFU@Rs$5Y5S`hWp4YN=)Y zMDj>7z)$cRiElh{_8I+6^3|S4 z|N9-Ejy4yKq|N_BoBzT7sqk#{!|mXHGWy^q>V!tPRz>aGJ0nEcbM26XPQh# zulQQ;JO{qsCYVE>4@HS)-&*)h_7bR9H zdh$`nFbW=YPFV~*#L9e$IZFj6yAJhg{G0Gv@K9EskR21@yYLj|w}GR3&ad-~!4=;P zu6RbciqK}5_6^+P7l%|cG&~tt-V{vyEAz*+h0tp*G>?E^reEckW!6-#Znf&kU6ix2tAe>HJbh z-jwmL(93E!4$lf-arib4kC1;8iqoe0-h~Y9gg!fC=aqJXyH5DF6a00;yPeQ>Cot}O zn@?ce`7Zw2w(+H%8u#|`r4e*~6rC^l?0y701DOy(CPa}5F=UX=+LoUZMwS^}y)2mc zlyAO7)S+C{EJkSj6A#!!FWXT0z4)APuq z2=d6t1m<0ZER)YH*&zPZr)0A0PrQ3Jbc=k9ASa{ulc$4RS$el`b3>7T|GDVVqs_*D zXj?ixxHFEug3qwe>od$`58u(|!9Kxv9X!PF3F6#egRO{tgnu_F6p6u4=!b7^GWsP^ z2XB9v_?u){5&nOfW%czEA2sracfImwjni%?*B?P|b%JMN1|iLNn7Qtg@AM3F9d3Rv z_?X{j;P)7?*lx)`N}20%?JwF~!F_mnJL|x}D%0FQ>x?nZ+%?XJnY(ya&l|ijPvOCp z-(HzLAG;z3-No*QUZK1ACjS83WCPTW3!R<#KziBa0H9kFCwu;{JaBG@7R)%%sgWlc z2YRAX{K|L1F@lcN*#?!8PxN1nY)R*_N6FK2dCb*`5%?zp|3u)Q2>cU)etL z1pbLSeJ1fl--agnbZFAx>4r>rvi(>?hXzkgw7ZS5HV?+rw}q#s3hslaW(QC7kvs>W$p!>Z}7go zR~?>7u79q_zxU>wl#pZO=dX_Spn0GTGS3X=|S9 z`z}x4ade;hb^mESU+bZY52+o+^rWYBrpcr`Yme>X`D`=F+S8m>8pjUogvNV(*37Ow_+7xc3wt|^kNelu?(5#eZ)6Z> zsnGvGUvX%+6CTkRy715Q$WzigdiNFT?vyD83hcu83!(=-yMuN#R<)z|jSJRifBjm>2YmHyC? zv^MkIXU6ahy6HXe8$=crGQaTC%@4YB*PIpq50018qX#(VUyuq%;l!Mq$=Jfg z*5(n*Fg_>Odw^vG7%qcn%)1v2|L%{;fih>vL01|>9yF)+TP&-i%$H|)ss|ln+A2Q^>8tGuHR@FwZb;p3S^^n+Dp-lAQL?hqtZCYO5)= z4UI2M`5mK((vc`VqS>?!Jm+j;;9k1U*7D}ET z1h0L2b&i8&?5#Zf+d^Mf-}S8HX{{=sdYt*t;{#dwHa(@$#|EdJqtH`p1#wnrP-h|c z5+jxE)`NZ?m(9C-O|D{_Vw>bJJGRN?c04mMu63;T?``(L9yrJr6rOq>BM*Qv9ATX5 z$7^HU=JwA$_20*R&4rkU>Fc^<{i%I1{-fI4bCSM}(w@$H*}1b=Ha-z7^@*2K=SWwCwtmpYUQAUeN6h;MGIy zzY4w4avk`Z>q~SlK(2#BwHL)t`K{VPZwsD^A-yQO%h*~2xZ4UmBeRjeW#z`siqFLc zU%;9Kv~>hpK1yF+oP7YSM0-cM_o&9e{hab)`d|+0hUr7^tF0qxbB#IubYYA2a(@zi z$fr9!JKSI(x5^%>{6sO&~p&eEP2ZjZaqbfZs9Wi4Mt0x)D;>CCFSvZrm4Ocqu>WYd zaU4PBsBJG!Kj!wyJ@JA|r*1!k_zk-G`APaYN?Yl?P~`Muc!B#~UYO_LKOb9UDEyCg z$I1A&G0qsU(HOONq+}2*@>;*g7)5_c-b~|*AQACQmp$uAj-Kjw{azMmZ7e6M!r zGK@Z$=01~YzNaW3=Ywncg2TyxF;0J?ZI28vdl?-WFhFDC_wNob1Kx)ZGspnbo+AUC z_G}rj*4thN8PM&tmns9?yxde7a76hc*v7Km4x8_`9ru>!dY|WdkLP-)=X%t0{eb8C zAIs< zeK>z&s(mQiP&VRk2kb*5YaD&$p@VU)7l5}Ey?7G)@c+1DfX>qF!_0jk&#YnkNa~<8 znSYbhPbT|Nc{k}g;zsZDX*zU}ju=KZej;stuAR|BzORulwvI4!b?l53c>er5Y4A+5 z^)l&*BDc>|&=HTFq@R=N2-(c(_V5YlUp$|K-nZwCkBMA0^oXA!8`^6h{n>dwsYiyf zp}U=Dhtm=NdK64V`Jey9doTyKk!3&Pto_PfuaKkGXRhMn^p4K9gy_ z(ut>FLrX_!?cB}*OqHiwE|omBW$5?KcPBsDbG^xYw`Jl3p6emLk5ndpm^(!?k+& zb33u4{+F!7kFn;Vcte}rx+vncs=7QmKr^tX8*n6jQwv|5;MV`js>-dfM-S~T) zrJm?Uew!Q%WayFY_#O1?<*N(b@o>+?O}OvGO-!Gc*)f`dKAF$yn%fL+pR!HKXKUwNBRZN&~o4{T-fm}o7NNqk0Bcs`~DohtIo73 z=Q-h_g}oZN;GpMj;UF&~4h}f|_cB+LR|XDNfV(R4_~uU{c5iTC+3(h-%AtMEyJ6ln z{{GwSgLHlQw0ZvAc{a*ClfHafdp~oZdE52na|V2Q*>2!53QqqYd+!1zRdwb6-+J^L z1S5t-qTNk{(11|`3WOwG-L3M{pplG|m>C|;qY)kB_?wX#lj(*Am2Qnx%nTEq3?d3v zsR?5+YRrrviouQvXePntMECKgp$n$?2{(mP6S2sEDw*aG`SW6t-KQ{pGS2@on7*jf4ok`2TaGwG9cDy<_ zs(7`PbK(b*+qZ$!D=Swy&r61k-wXWARgIsTu-94BasrIM{U_F0c~)}gZKuL*hx=-$ zm1i34d9UsB{?5cI(`^CEOKOYfbtcSU@-CXZYGRYh74}CzGeP=T`TYLqY_79u{|N_{ zE-&q|1*nunH(_;&*w6lPZTNjzAfj<@! zFP)B0I+vKK-o24{EqP{DdJlL8e6O1K=I~xc=6lG*x4iAQ(_Y(5)<@`reI}wUz@fFB zR+9%;7)`@l6rISs@aCZv-@}p%$UDd8#mGIe{V%)DRBLFNf4ea+`}e`tBSiCG ze|&J7-v&RX=9ksBKUF|Y-`mxh{!~mlkz7;75H_=xt2w4Smw6|U=?k5NA0_8q*f(Q` z$#Yk1b`5i;R_2v$_+st&WF4$UbmFJQ$=O%?W{!aUe}{SruU_qP+S8hjvCT)&OVHP> zXJ+;N-_En~GLLP}YVSYYXO6hGIp5f3%Xi`j@r9M-U^sg^Si~8zjmnt_LNEQMJPg^( z1Uax;2dPucne|xJQz)nSTVrzXds}13g6k=#Pt_ZAp!+(&Ngeq>1)PB`|1)+swGQsP zS@1^s8Gj4Dpx@P)ZE}PdcN@Nj*1```&(wmA+2g*O1>@QQ#_bzm+)JG2 z2c!L;xzD6%Z$(S+4K(f;^c_#z)GdX_i}t{2Bz9r*ZX~D{jdOj ze1FvGse1TWgO%KYuC6OV50lrRTwKL7RAZ?;Nd2A|$9{6XRrlN$%x!9?|JBqWbW($m zXZ9-!J)U3s;=O{M+Ka<|mhX7ph1>4!7;BjvN!tevm>B%q4ZaN7eU~wtIl;~7#i{n3 zfI2qLvY>{IwfSk(&UMme_c(O+MTIXdxOBlw3(hQL&xQQPZeS|*1t{gQ78hbq=J|4( z<;!n|?=1e7Pc$QkTykvDJ6Zky#_9JR^sDAS-8Wn4Aj|9*(FKmj1bJ_&m3(i!^`_=v z&U|P~Du&LQbI);!w5dB<31{#Aw_$hg#xS7-gm?VsP&e-FQF zKIHcI((jEgnC=hw=MJ;KcW6^J+wuv1s2n5gpRS+edMN%9Uwv^1zPiEXr)s}lUw69f zK`>)Ix&jyI4Y2T3ZC7#MFPE{^&?}ucd_0ylQp!HeLwgF`snR|KZDnnI5_Wu zX5Dk_`HuXGHrh}3r!!IuqP6M|JSrcegLb|BwbP&GNqVl;z?m1FoPXj;*{}$_R{-x7 zO4kPq?|U0xj{7h#Rar9u>x=r9{?BEy8S@=)EqJkw_k9d0T=chGB28n6@c5ffzc~pmpRXC z{lvryfL%IW>*P+XAT7th_{uk#-}9{UWW4RnaNFU&i52jSVg+tn`@6F21F$@+ws?N7 zD~k=bKVr`h>}%P9i%L}YsH2M<{0?6-LKiY zK0!QW8}cfDL9s)8R1-U#kNwJs9Ug(Nc4W>m4!P|eapo+)Bu;B$SOq;@_~;2_G!;M8 z`&sbpb>45`eIujzxvBVJR$osz&&Gd={LQ7UYCr2Jnd5uZeTMPbGU|yRDz1;;9mfw( zAg9VrZ(~l&&ot_cg<#hs5sz?ABgWx17!u`zSnW?3M#x25i78pY3hvrGC8f@Kv{eev>?O z@q1aNr>{5rviiykzZV|5uXkwgCiv8imHm}+^PG6fT61mNv0Bglm0S-cCz6rC5Hd0+ zOFpz;*=Fh=GxE*RT|qvYN0#E72Z$L=;3GcXo|S)I%HP`mw2Y5vyB$6+CHL9lGw&~l z&PV?xHPNn2|HcDbN`Cj-@t(~Uf-!SH(DogH*=;vtt8TQCbCK6Nc;h+cSd$BE@6!}8 zd!4X9>Xwc5CE-vm`^wyi4Y?VAc}Ml^4)dH{8!vxPO#W%Zqwe@t7?dC?RCD{ zZMXVnw}|%C4!X2oOP_7GR?cp(tDHUQW7gq29{I*A`MEpa%CDXDR33cRKH8eyHpcG1 zt-{niguOB-nRDev@+h8FzP|QYD@9HoCSShzbk;jP%LI{(^z*dEnVtmnIhd0PyaiivsU z6JIQ&uQvF7CpLwe>gGK5hEhzHctcaXT5^zU*UBvlC6%k8I97+`k-xQXk&Baa!AS!7 zQCl70OEt)G)mhVH0#+;;O+vy53x5|7xW!7Ww>buM}Li^R38e>!}^^x?=O);&t{{ z83=FLyuJ#&Jxfk;0N>^zY(P6W?_>|gdF;uZ&)mKaT>6bZV2r|v(d)U64aiC1=5}C| zj2k=~J?(dD;IVVyM*3fTXxjkA0P~}d3qRsNY*RM>rS*L%CG&qpA7$9fOOV}QZP_e4 z*5<_ijr`_BuTmbF$#u((exBpYVl{8YkWiAxzt?eM5ppSsGpOt&%!vU z+Uq=ct_Xj16#RDvd@vT78wc+tLbH#Pqh`k+?RZf+Ywd6E&x#jqvyvUpSjogA@XRi7 z!@RK5TTk{3blgGvyJ-Ir?CLh!UP2%6Tfg;WtR?LHDD`CMqptgW)RR#ceT_wa1nsJx ztcEy~r=IKr_C@Wep-phvLfbJZIwbvTMNuZ6xlioGrgd-|AzV z>dCNujr1L(?+UB1XEU+h#{6K@g5zI*b-{M)a1U)?=PPWw{Gw3g^PjDZ9Gy8j@_NPT zkp-uN-+Rc%zy~_We5NqpYPxQu=eZ|GMqYpRH2b+aYP^7FK>)mu?ZDLp zo^D~hcZZUlw=(v7@YU-W7wd2#pIsByty(G17zDet-1`ck)Mwxfd&ZJ-VyAKiDliwJFFVwfG2cw+`HYZgNhFqGqO0(L6 z-^6#~1;MdEHCz==J=k2vT8#~J>%nN##5A#mZd*a*+RIzswq~lWnVD?}KJDkI@m0tt zck01LyZiX=x8rRVa|G3cU6-m8v+vvbz`mZT+$q)+FXFpr=9SR>sz}pCR^%I>^+kH_ z2_>6Ox9elrujB9%YQS#52USe79Q)h^4^16W)O0<3^H(0exgNeb3g0Xduc0qnvHjiX zw#y3&o2HEz(R6gJJx9BN`QS+AgFUO$`-jKyd*kp$_#4dA=6LYTm}PfFw{C1j&uaQu z8%o|aI~2LQvNCeRY%6kOr7v<9ecu4g#6%8hFaEVj=76Sd{bK6MM$mRJ{P#n&t#dG=cC*ct`sNf;!FrPx2>IP$@pOFs&0&=LG}81jH?K_$kvO+@K9MO z`9o(--Cj?3)@Q0Tzxqgeedbo-m)fq7Ez46z6gDlqFci6Bc4eezdQs%qoDq?Q;6i(v z$uBBre%5+Bx~|DeYHlta7^96ia@(ArYv|Ym|AV*P*Mql5skgZQbnP!DylL;8qUbd6 z7P}R=>!1a4-PQ!W@JKiOt(dXP2baS?pND_M2bbHl=%-ESyx$)lJ4BmDXD);%FOOVL zn}0={qRsWR*@Von=70F;bj#?8wZvI_kf}P~h*Zs1Ir>BSL|yE&e0@QY;XRjMj?F2G zyiq+Oawjl~A5*&T$$WHLLF97X$2J^AhNPEdv)dQ4zt9rSJAx16HRzbD(P{9Z@&Hs@56gLz+!J$~w4y{5Q}gEQf(2U$KkU3#K0D!hf>wQI2MgvT_NZR}fg9rFv#ai<#i zNAzr_)~KmqgrU(5@cWJMy9@tX;1rD)I=Z{Zf#1*`_BRcJw`Va^+cvvi*Eqm=ZzTI^xe;fBD*iGgoZ_tH|LIs zd>43MH#|9_r+^yrZ!oS;jySF9b%*zEg!kTn_l{vV)BHCQ{yQyt=Ur+3OYxn9&tvF{ z*WtmRRfZymuc(ZCuhNSA+ZDdZ&*<-ar`vojee+lbUo%IC2mf}EcGc#wiiN;_dF18K zT9HGS`Xa}`*UPjio)g_Z)wJ-#0SKSgL8meLp^G zVWf5J<&mn9R^*a1eUUhDR{^)|uJoC)0}b{XtbNZP{XY3VLycE_I-4<1Wj#Xj6(P4{ zSoCJU##5p9=t1Z*Yr(>Su}{qCIu^O5Mi7d`Ll6R$j`^a=I&e+qpvkUaj6>l4;m z{uq5CT_~M+D*Eul=#yILu!KF}{#d#C2hsbFl&c$PV|a4q=3ZsW)o+zkW#rCp=U18d zo0yJ@=bE_kZlU|AM);^XQ*5*Y^7Avs@3?H~Grwp?#CroZ6yP z{I|jOPs&UB_fNX0i_~0WDF0+WaiVPhM81i9l;Qd(L62Pjsr-{Sn$o(Wzcz>RPYTeJ zr>aj6qF0BbPs`Cmr=m}PY_-(z_h(BO>Kg&Ps5B?qp ze!)Jle<}Ej((`8g%i;QG(x)xF+n?QkLis(8PI}y2%SZ4>zvsFCsJXUvOvGHqlVTo-QM|+lb<&KeCd~Tewv$)*1pwB<)abf_&}{rBkK~%@x2B)T#CFrqMSBz zi<#HryA<}!pK2{$h!3Z^>3Pge)BfrRe7=#kzk0LgP}C*hul6u+x|6ucoy1M1SjO}JZcJY+0&lK7*rX^PLCCNMUmo{X?!~m{FR`6>( zuED>)k@*aMNh{;fex&a@x!xJ_!hA+?4aFAfd{)l~=z$EhLyv^!XFL-~=FR+TBR6gx zjceE4IvV@$qcbmOj&nuCo%+Erel=&yc7fA#S!ymeQ%>Sl_1 z*1)gwEeGSpzcM&G&E%E!oNuw;t(9zaaJPWGqHcIE8-Hd#2>zPEpKvD}9=#eKU=AqU z35UX+a46gfhiUwcjLLs22`b-}T={9C$g~-ik?!+H5Gx!RnZn#gxsuFnOr6+E)W5)& zhZH-A!^`KulX|b4YZtDm7g&+=X8BBvVyZo-*`jz&^E<|0H}BeUhd}r=-q*XkL56Hk^}o&6 z#(sed>4!U^#ZhRH;{9g|B5&+3G%>1U(BfqeFTV~Q-hdu1FZWN2QH<@h=>2~LEuPst zd|Ip}@6V;9ON&Q%U$l612wEIRhr9Sc$I!x#n~gKqwr%;C=l+>o4`oA?Kk@yMgWHdr z9ov!iPyb&%N-K8d|K@s>i-(|jW<82MAK6U4<@=+~NXJW&Nw*fIoI2fTeooe$q|n6F z7MzZrVjW1aRqeai${eJv!AiChJMCz)IQPOz#&2O>LVF*W*L*(@Ta|CyDq`M+KJbp6 z)jDN6H0VLkb|5?MS$xLUqldfD$M3Sv*JWYhQXBmNCP;*ILU$nix*k;fS0 zF~(Tu^YujJA7Q^^)2-x`d|%T`tnU~dO+3Fx<9Rkco_DcVTHn!lG%k%tv}z`VcdQ>5KvRwa!$)dO;EEOtPO%Ba51DNX0t~fxjrS7T8&Y zrVdQcGe;m-ME;%ha%0BYR335Mfnkpg0lVf0roIClk4I0Z>O1Jyg)>pNV30J@PY zl~BHvS^F)DwwP<_x}xZB&9&_>|Jq#J^&QHu8cOeZ>pN~9F}VKI`a$QUQ1Y|rXFI>r z)_+>Rdw=HFJz=d(FA?{LAnY=0@h7XW<|D zBAxKeywj~@IXwMie0Hs`-T5JEIwHiIRnsxKprC2;i2SC)^Fomk7s)TlrN6vLA-tk> z4&`u(ue!>RNqEcHZu}HeUZXb8uGi>-uT-z`0sTHb9~)6%ud5*+s?R8guf#j$@L>6ff~KyDG+!0LSDG88 z`AX|(1<@}=dqr*5T3dMo|+0jFLZe7I(VuvgQs-X z5b^(EwEbOZZse2og1rIl`)l*mo-xv0wvPHYu7}b~;;%(R@YmXG{&M1l1>}2s*E6*) z*^X`1+N2xzQ;t#3^o8L3@)Qnw;1A(I@vHW`!Qrj&O&$Ew1mA$OM(Qy%SJj;A2Ilw+ zsGZQ<^e$-n6Y4SQ`AzyD!f&#XkKxOIce@q-DZaY9b|gDHNF7NGc>8@tljG{(%eMLYJEnoCxD=aNOp_6U0}2~7RY zC6SA!+tA6pr~EA!u7T!~Q(w>0x$yg~ysvk!b7i2xw%^{Em~`!N@|X1ehVy;Q`R?I+ zD7kRg`tK+lT=&XWXXkq#uXfh^i2-~4SFazM@;|Yi@Jh`-s5JoFpC znV#uaoA?dj-^_dH%ho#DYCsp< z&bx})Os6L152|A*?2&(fE;ziv#Gnd2F(})YybC#%AFQ~G)|h%2gPRkj7*zKnEBP9A zI$AeYtWB{#bop*OCqnTyYVuG-k z`#+@H#!fW&!-s~^wmEafiE~^6Z@T#@mFC)>XMfIff6g%YS9<-w3I^xhFCss-e`)jU z{K`yz)%>WVIyhVNCe5G3zw_YJI_l)&sTu(JS#}M;C7cX~|$EXAN z?wado{gfJj4nC)G@2AuO{DgatLf`^mRULqPUpAwby>Ppz1L(QkdU^yoprd>vnhx>( zW?*#lmAzXUw^*ae107}ZK$Sc4CcI_O1x7jJ91(TLnZr00;}QSA3BN0^`VD8C@~hrp zED72uvGS<}5FJNHFTf@zZttfSfObqR0Bxo6WNFjnhMKnP9(Vig;OvP3YXMH7o&R*( z8K@TE6xw>uZEL7nfKzDmu~eG_)B>DBTlczc4OI(p3TkGUEHid_T;3aN=q!8Dl;X$A;`DzcN9td4O}X*INO2o!V~ZM(a6)wB)Ju zIrgRaF{%rQsV)H9pj>uS7r-96rY-QfH7)ZnpAFITNudezj!7sSm); zjKI!GzB(nN{H^)`7flmXzBxy$9A44SKg@)Q0;4U0iI&u<5Aj^Z`aPd zsuQSXKG@Pgod7n_)CmYT(5nl4Px0lLvp*l|ZEWlY-0@*I1Slfu#80c#uMYG(|FPS^{#K2G+cBg0-Owv-!v(8UvS ztQdxuFVnUml+q7N>Ei=+1J!PPOLadv(dyIXW7zG7xUS%%yxPw?-#T~nQ|8*P8<=UX zbE0SSJ*3|p@XeO*@x#2qeUFX+x!q3vfa)Xh3GkKAPQ|tY%u6QlG4+r-&$JkQQ^wh9 z=cV`o73QK95| z&bh1S8H;BWbJiNR_Q>*26YlJDY0UgD;4|@%O4<>Rc<(8fwEkxY z-e3GL#phQCz~@0^^!-uADSU1@Ba}Skv!2+C97r})7h-rDIa2&@!OGs0ZWfP!*=hGE z<38PwUaPUf+x(~Z_=wTJU1o*H_^sqIep6e1^kWG&P3^e+rrHp%{!;&M=5xjj~9&E0;uPSmWP~nNw;Vs`*`2Qr|27z}b!5 zKb!m2@JbJ(J;1qX0bq7vanFsTj>OrA*2bZ$gKs6jN#hX@>Abc$dMH7E z!fVjMYYaW_)_&+bA%kQ1TsV&XK8@o&4&S*p!o}}FBQNJ<(l==9J&mD)F<;MEv+BPx6EPuIXW*cW8BL73gDNxdp*134n)uOj9KG+iTYRRp`W{B z+u`sJXY#uI?)kEq=KGJuZHUPqNo-bN;b!tKr%GPQPCK(&uANpDu2@+`b*W zZasU`wDj)9^TgZ?_SWLDpzueaZQo_-b6$(E<#@ZWhu@ZZQR7>2`t?)VQH{AbFu#@~Xk6T{CD zO*8zIhD;d{A04CSS28F5PmSM^yWx#Lv;W=L3m^Qg+CuFYQesv7!tk=EZ}s~#r(f9v zgL9|PQje{8-NCs#-p%NIuYYIk6Y>?C3a_5eULlM_X9u^TYg$cwP-lZ{PYul}@}l2- z+TlmXU(43Dtvjgk-euPztKQnhi>q^6cZB#oG+X|me@%`m}!^ij3LEs)PzW-SM`0n1}!`m15fCm2kh+&sv5(*BO4Pdswz^>_ZK_H+9A!lC02 zH6Hl;+s~=L{ieL}6vYGF_=938imABqloMCl@qifqbSZvOoVY+0--fL_wrjVGk)0` zVmtbK+0B<3+-WXSik+36zhCEh)Q<_j|5etw;bULESsy&KjP=1~_VWAG7t%$;J?bNVJ&$&ksrt_UBajZ$2^()v$_eO6XG$cOy?JJk-?YNrGmD9>#? za~1>djaA9<%w5!W3GW-2Z}i*q#(*Oy&@zS$ol8!d<{F}3v;Y0A|KZpm*M817@vM)h z)?$=c%YtG;tg_5&4-XJHUdn>@*Hs)iJn*{TLIDUe87956v zxe|JJ{ru$CD)tepsy`)^>9-#v`)rL6HCWNqU8QEC4`enAT`B-jhto@dH`DUQzD`~KEt zdDb2?KC3jWc^Y)3X2Ks;TuR@1ZW;IzT@%cI#21$ko125bzw(|cVv}}^E>29dk{BHK zch_)My1N#!(2+siUv`F>AM9Q}mfxYX+SeYqx)%S$kyFL)oc4FGY^nm!IX%meA-C=2 zciU}WaO?He9`^V$UVmnO zmcOicMkTa8beR1e->ZxCW%;P*I_P(;9%=;NrUe=fr9vIY?VknxIx${pq|6{Y^sdWBw%Z=peAS=?v zHg1WJAO|g@q%(@yqd{|XUv$n}C%20C-`@CV(@|OO`4QYn>6E$6n!G$j7<9{N`e0Pwa+xjAt_JU)x1BagdZ-?fB z=XisYUup7jv(Cp@NX+_(HM{RMK6N>f05ArT{Y#+xM)AotzB?#&G$7Svrlg8e*OmjR&;ACV@U9Q<;VKA z-k)Rb8ISI&+B-E|-FxK6jOXc<@X2jfPUC^2UwidHi4|U5yXwR$E4QhbwXhOoTjz`B z_4Pdwo5XmHkH&d5tFVdabMgLOaG^adMVk};?RN8ngG+XCs*m}dxhqPovk&Ul8anW~kktv=tJec)thOY?Si#&AKe6ztL z?}q->HvP-VyU)nSuJ(sZz(plI5(BSQyel1Fjc=O(-5N(BnIKPb+4!qAmXM-#5g=~Mo>_f2w-L`>TAL+Jg+A2eiD$#v?eZh$AZ(m<- zq?&%J;jtR^#mBY5vLCMfEqL|h*72jQH@)^ZE*k@n$`6dgJ4=g@{o0w~eYy3o#LIl) zL>{~XUoKeKw>4hC;z`_ZCBCqkXM)IH4m@@#IA?9_=@&1U|EjIK zCWXhXTYX|_p!gT7@l!flW{`WA*Rzy$xli(*xmR+bxmT&Ry1>^Tj}=6Z_r1WUyl2yvY<&dCLv%t0B(zr={97#3R#Y$ zw-U%-EArIl$aP5$eG_-$Jn}s$IQC8p=K{w=q5-hTPxQj_TWJ5VmDiYm>}#*)ms;V2 zwY4XHOwLpV{R)4y6)vFPa%5S3Nq<=MHgy?R6|##qX=3*+t>xoZFP`0ajBKpB>S5nPSi@g$X?r*KbRL?u;Jv^I5;=4 zB*zxocE|Nom9w1}mAu${76+o0To*g*5R#L&K6Dtmg#E-a;-9WPDlq;=+8)W)^g%1x zA#k>(%KBv3mSta9vTXKB!#|kxKC%lPpYvOLq_32tuoJ(275ZWqI>+(rx2(pmzYN=d z0KZ=JZUYZ3)Wc_f7v1FFt1)#U28iw#*L@N%lu*48R?Kpy@w8Y zC}f4((KmaEeM8sdBeBst;c3;lb^=!|aCLd$S_543fGfZBtZ)l@wGG`Hh0pr>@*|DC zeR~$bD~i8M&kK+DIC%Uw?%Dd>-oJ1@`G)Py3M__XThqr*Gk+g5Uz z-8oIH0e@pLehuyYNVK^cxY5DOPQzD&AM|PCnVpRBHgKWb9?2*BeT|XtwhG^ECpO_E z>)aEu33~2a#?JnDNyS^l7dw%iUFcNy=?zDK_Xs(iRp_M-=wsIsD6caYehoc70r)qP z*Lm;_{7igC^{ra5F8H|}|4lh$Ezm~(WBh7j*MVCfpR#7stcRKV?qu!43)8QFsa!D8 zr(lYJJ9kXkkK=5nd&QAEJ>sBjS z=F=X|Zp=Cm-G@9Ts?m3hzk#tOz(IpEz8Y{|!~DMERy$|&Ux8KElNjqAjFo+#leaTg z;YT>^pp8!OuYGqP#l{E+($jZ9&pv1&dVbOwziN*jhMt>gQ|;+FwbubXJE3D2?UYgz z(Lz7P=*$hg+r>Dvzps~mUvlWDzPo_E6WR%$Da^YH$;BlvkM?YOxqV(wpC#(k8AA(W zxDc6kX_gIhFTaW2?)zWq4`$%?ecTFn0lQ$9o)^rW2IhjO@*Q1zP4mFqX?V_txe)uX z(6{lGqTqvX71d6DDu2TISLPL1f2T2YWsPA@KVuO5u8hV{!{@Ok*?I_>+}L|gxD6O` zSwDSvbDz<{dfw11i$>2l<6Q}jkgpjf&Uo8+ZlXh@IOA+%%;%E7Cp;>LQ9Rnlcw^Y> z7Dpe-HZbl(k~QUi9ebU8F+(TG88WsI8N>d?d9M@xY==(+K2z*6ArKbjx}vvmqF)Ez$dsHptJOkj~pUD`9RP0_HA9pJfN*El-$W!cJZ9--u;a8 z2)+}xW3lEmG5lQH9@x60l{xB0YjXI98Vm0}0Da&Oa|X6_)X(W#{Ph6sXik1VJgvEa zv7Hs_2RXwItYDtutDSHnFgnnr_ccF#koT33GM@KW0FUOU%X$7hXi}eJ&5Q$owdP*v z+l}~ZiY=}OPU-3pyi&TkF*8p@Ys->lE%6 zfhX*7U)_jEU){*aN4>DrK+DzObuBiGIc8NT5RTu-J#4gcaJA=1g7XJugTam+f2j_d zyb|Uadn%qkx%KmWO5jZkJ61^>rB)z(VcQ4qP3!00q<-#=@8{mx+%s|s9nJjvr0wsF z&Dsw@^Mm;&KAWB&$d}0jAFEk&%>$3CfjfVi?Yjt%wcv3scvM3i%E856kMC#k&y9~9H1Uyi{L;iO zvGwwWG}o8k7XuEPhLgg7wie%|%(lI<+lqxj*EdU<3mHEZ-e{Gr2PV~rt3Ku`?XO$r z<{hQs%UJ&$OP}Lx`ptG=R{SKjZabDX$8kQ6GY%JiJO48ie%WFd{sd#m9K$)`iHt$~ zqB*h;xtCs8hF(}EJIt6ALwJfceO=3LYP@;wIMD|g<7A9FBUIzWxgPd7+l%e7zTa!} zYJX#GFD56)GhWwUKS&<^`=iqSx*Hct)G%&jRbzV|{##A{Mh7_~eS55%I-E5aopU2U zuH)nEd5tXV`PS9I!n;NfUu(bbuIpgq4plK9lt1t+dA#*!vJcF)=v>8MmT2Bitvvh@ zgQu;JAm`R+!!KXgyEU&zYEW%Fl z%(tQ_cS zKGAQvwDT$8mmPASl?|*1CNB(b+r_^%J2p+S<(&hy zmf62)4MedS#pqgqI|kej|J0*9kl7aaz6HGzD2P(p zV#BAQPyUhTW7rB;Z@-W@Njxw=)z@9#_U6*wLO$Sw`SY8ltrg(}b2G&PmBUfxuYct) zkc;}PHg}sg?J+3EQweQu@wV%Z!5QCIy}vnQP@g`=QTnSh!wFz-#TKd02j~;~^DV{O zy>Kt_wvi3@zkCqx^J(J?S#WP~ae>WK3}nFoZ5p^X0+$CDg6k~W{A^a6H_)bJRyw;s zo#^U{-%%H%dHg`S;t&1UyEV)ayzwx_7ZakFaA^EK#TgWTSH4Y7^cz<(m#)B9!-lOw zhdki1Ve%=M#~;!>J_awgz=wiO_Dr@c&x;G!eiZ{R@)@ro=I6}UzbKpK@McCVV}S9? zrtcYK>@Ps)lsz!{k!gD%+a((y9r6q@{}}K}pG#j)f)oc0g7hi9A3%S*>uOpHwRyw#C*$x=%Qg#|`J1}}dCl{QsZF7FeRaUv z(_Fh|R^1cUO&0Lfy<^?97r$2VH_m-(UdwsoYkg@r@`2+)&Tx_rcIRIEfMWu1XzXV0 z1wU&q2l){}U=rV!F}A+G(<6#oUGcI<*WL}x_wd(2@=N6r&9!ITdC5`34A z2V{&*q%m)To*) z@vzRjbNyBAUuJwHWI^)b!_KLu*|nQr{WZ0?$f)9VS{L3&t%ueh6IT<*;@Or(od1gt zKrCj@KKNx4Yt9{^^TM6PYYz~wE%0H>h(pLntQc(_=i01IDE6_UcKK5|kKOsorjg%$ zi?-L6@LK@*w05E2RqIE8i)QvLdqvy=D0*txsp-)6i&P zDB1dx{%KUoIz(yx>~NLro`*(nDqd#vq2k+_ZK4OB(E5Sa2-|C{(dTre-iXOfR8Kd-Nkd1BwlhznDr4byHj4_?hPYxSH5rW=6? zU#hwfAnkb06#UlK-8wq1fVs@*|OX z@ub@yHHH^=m!TJ`i5Y++$-qYVK)Icgn|jeU2cHp`Hu1ZB#!<|r)sH*3&~Lf4`4IgX zpV4{N_>B0nswG4Im@7{)XZ<9g`8Sb6`HX5uu?zW(;x)CSd=y<99m*Ju4#hUL0K51# zTZhK3;WzAHe|+~bFCDY^?sV^OF5h84@v+i;hcD*gyF;bBM}@~{@g4K=0cFgUTk-ij z%|>=@p32m(ZX<<=A@gBtDiqx4xTNUio9?+)Mc%S@kdSlf+BGrR3*3c2Kl*b4$iU z!%}$95AP|C)l!TuhX2MeChY3PEyc{6=^wqaN7sP~=t%gz4joyKOw?h2x3kV6c^F5$ zq6Xcl_03AgXZ|)iQt>_8&#RhCJ)_!%KdzmFABOyWN#C?Bf7$g3ZQJYQHt?+Y$$PI7 z+Lub_I6mZ$TKe~4<-1+#;K}Q^4piTgosV-X>ygUQa%DiWw4V5#4}Sj{>+Ql_sp1b- zk;!#4Fgp1->ib;!?sEJevpx)7y!BDS(ckZamjkvxmHqxldH=ZQeLG)2U9Y0KqH>VN z^U?ZEvEmByr9-`$^&ikSR%Ds@l+S|?<;7V1GyYkFlOMo-=-&CgnXjWi6E*nS*beF9 zPI#(|k8D!M&FDpRY)9*5J_m-ZYIKQ`&={9S}eQt!n&0wDyaF@J~r>tR#gD|N6wP!oI~hzts};D z4hn6%?HO63AN3)*@ac>yU^Hh`d2plgwnGEqMr&i+e9jqF-0$Rl&c>ciqT^We?%A2+ z%YJtqa}^V3fnFMuIirepBJ6jNI-?3d&a_pU)fO;_A6z(A;zxB(XI{hFPv;!$=}h=` z2e4SA&ylgUpCe zMsjAF_rCY~8_KDH7I*SJW!;1mL@!m?O}Z9e4o{f4L=-1uD$}? z_FA()dV_(%-W%;YuEooP&DA@~kX3X~9yMUH4_>{OItQ;9dX$BdevkYc{Zt~^PwN-i z4~?&!Jv7S9_;P#J5dWLJZj~9|8pc<^Sup?1_$D(x;h>!Hl`+0@o+)R1%I)ZYH#ARI zdy~Oq*}8L2uvQZ$N8)fB&xn`J+~N$)Epk%lD2W%AU{BU?j&2-!m8`2dF_AuWUXb<# zNXdxKI$}JDGmuSWrEL`BgSWI+7N5j>C+)M2)ZeqTcOAA%Hb^-tOX#nF^Z5ja-N#Dc zbNdLS`w)N9uKFmZ4fSXGl8kx!YU$0J*=+bC(9^|UTIT|@`Y7jmGV(6IaPhVvg|`}m zw{s2NRvNrbURQSF+;tOA2rq(Lv3<3(nYPCRUzyXE{K#BrFq5_vA2^r$@*_oqb9qiN z%<;5S2Hx%c6Wsm7UHfY92l}dD$?8)$8AHF#_!Xi<7jyE|c)0eC2R~%SM+X^??4V0y zFAZE8E0_6ylE%t^|5MVKcn)X83`~z#$=4VjJu-b8n@0ZVjo{L~zTR9b=I4*D@!VhK zxnJYCztnU8FFg0Z?74r9xwg+#SZuCs9di}e!__gb|9)s4)8+9`%gx@NAJi{hv6 zv5qfLKIAwnIZ^lEg)o-8y;Qq!UXx*Wx93d>3`weaxSFsMVtF@GSaH#eST4 zuF;pp+FOpgdTf@~&xqk0n>e36!o4u9VPDbbG!FnS`3^cmLAmWI8;Si&$T!0F*Wq)v zjKep;Ptx8gt@vHaVKcs!Y$iV0B<$F~8(whqlN)2kKQX-F{JjF(vTEJ<6T29L+lJmZ z?JKtLw69#%yWZAW6L~%5#HPsMIxHL-p9%bTVEa4q59~42GKPYTF_idE{6KLm<(;n} zPreO1yvnwLhgO6C7vJlf@web#v6R)^p9KD8PgnEoB(s0Kt_*}SLn7?kus(M*} zbhP1H`56A_NUq&6x?}3^efd*;`0H03?_1n4)*eR*<4Z8k1mm)0Uo^30{{!P0n>n7< zy?ryP1{u$5LyX5hbH^UTZ)nGr+y64x_I>fId;e9ghmueEl{dXJxXwQtxV-a8@|rT{ zlbS#Folg!OvNMG^*_7mhriCY~?zujvsA*GvV^d?WsHurPkSj`&4dkzaGjkq6e`+4u zek1-ea;&+i_KHIarn3$x?` z;xOok81WP%o6a036*o!!?Vi2p?7xhRC@x@XhLFEj_*Quk?wa54-yWRS-t!Nne>WRi zr}s}%UATc2bVnBI!tF0#KEU( z0;CtK6>9{yG4Q7K1^uRYbc}lm@FJb2+^LBCKIVu!vBh=NgC0cpm|6zL`!4fR&AE51 zTZX<#>r3sw|AD?}&#V)uA9Rn_%Qbf|$M!0pd-sB%(Ftc#jrb1j=1h_5KOYM)s6B+-LcY=Uuq%Uh-V*ylcnrn6j=Mn^|%Kdy(3&e*|ky zvK`&fV8KO&FD&3cEZBAx> zn9i_nzXm#?&*JDZt)ZK}?~xV7vpT1vOMB-vfuqCaCoAu9<0tv>SBKGYT3cstOMHSa zpgR;VDd1VfTFlxIGLS&OnKQ28>2`FF;xpR&{-e-CJTjHJqi7*L+eSO-c2!ed5dDnW zg+CM*QNFvk-S$PaMPGWZ)xeV%t@S(m-9s1ccR$7W$pt+H{5=&Ki7wikcp7wB4_&4P ztmHJFnTEcag1(x@Gt)*{$^U{bM&Ips<;A8=iJ_D2cb+{=Iz98d!F3%;C>iw&CSF1u z&+BixzGjDLg?^LISpW`Q|G~sD!MFSe`5mpuz2vxoz2j@q1*v{t?i`mc3- ze&seS;qUp_vJ9Uenn=gmKJx&6zccsCm}5eVcKAX2R46_+={@aR9x$;bZ~M}d^3f%a zI`^RvANxDxbzt+7Q=s(}x zey-Z$JY=N^J{<)w!N=ZM=D&Dk(_POn@(KL!k4opSnEkX^8x7IcxOd#WH|#mC`ni{W zwBB;MpS|aH*n5X*eaqcD>@BAsm;Q<=dE=#8gYokbE`89ZjP(fZHz!)Im%Jb=vXN!T zig*9a;2iePboRw`_XAx|PTF$bJ^5+RUZJefw!(v2<1NO|h%vqmtW7NAvz9fVrS4v# ztV^5c_?GV_-*7!^d`tOs(Vp!;X2qSLrMpJ7&Vz4nev^D1@BR|3k8N2-URH@}gOpp$ z{M6L{vevKsCirIeO8Djl*39zYn-ciOymNC^vXuUGz4YeHy&Xh{Kc|0hy}FBU@uL^t z)S#1RIXQggQfhSNH;>Z#^{4HeD949TdzaAO8y+5U+v|cx-tn|bXHP{oC!yDnW9{cq zV)n+E6rM>R^66s8i|WplpJ$mo&vU{TsDETi`MetQGRC}2`CHgi`KHga=en*H4+Qtg zIC`~usTIDYp1MzTr}UG7ADu9Azpbxy&&)Hi5zBkG-J!h`jP7%NLx)~o-(6!Z@X*dH zEBpHye}U0onfYq6TYd0o-?P?iXWf(9i|c0x#9Po;>&dr%Ep$(?_q{vn0)e${W4GNa z8|VHuu6K#K&-fPyd&k{zpK338Mr&+s?04_l#pX{C-vlSF?ZLO+QnPnNH~~H+Ckf~$ zyOyB81hBZcRUOC<_~~YFDtrczlQMAOCogxL_M|APHfzvJ>0k37!DjA{sz&Fg*P^#o zI6MOkrQ|7FTdc4TJZs#l!4CzzYqGZO`KLcP>#B||%2Hg$-4iw7LH@&)XJ_s|AwF0Q zEUF*$>au;Fea5oo>v$k|xWYfWClV$|@P*g8e zi2X=xMc2JZO&hY8;62s2Ne10|p|&UOd=;HPpzDe7%x%D=HIhx5d(l>G8#28kluSHB zTMi8EN3f3xe54L^W)Jm3`JuP&M~AQSPrf+z2xI*o{e9n--C|_5l=WDY*er4M4fB4p;zP)d)<*bGtp6%wHKAw*~LSNsbFL*OnN8Zm|meYestxlt>~^LrbE&TlMhzft-?y3-P~y%e~y33oiCUgIs8^ z-O6ps6AhUA5-V8TZfKB~f%|$7?%Tm(ht{;fb&O}@;Jt-5T%13K9v7ZdI0ygPIB#)q z-l$vyuBG?a1A}xvusV7y*VKv#&O+eyksDLw!1(}hs(;bFQ~k47R)Tql%PWh4w}W;& z>AM?P6f^1Zz`K;T?0T;JsP^;s!W-wE4GxdUHmyWYyYT*{2i{GEwIzxX8I+WEEl_Z^_#`#^8?JqLO>H@}#l+cYUBuest%>v#?OQC>ycy57KCV&Xn( zmg(b#+R|AYpo8k@c3HmWjw`M39l#Or`IxH*nkRFfK|N#fne&?ROxv7f1MR1J9-&4m!2)T!+I;LFx%5bISWZ>lA472d!t*mhhTE8|F%)O&px4mcWIh z*?~iRdTHLqS8{60)(QV!+CL=yNxRu?3O=tMn1&veT&@Qvo5dr}^=(`)K<1WkPVdWI zoYPx--K@HQCpVcGMUk(txt8CQ2h20!1t*&tJJ`qn652AfMV|B4OqfM1>#%ZW6@Mz8 zY|~cxRsGX81Z^#cw)KX#dBMEqZ4PbE3Rq)w{lIc;UlX!>4|&D#Q9FITP&;9kcvbb& z4?}0>-Oc4!S>Z;;RLyk{bZ+WxYE*6HD(KwD{PH>YOLT6Bev(5w-!M(}oW&_FNuCGZqYBBN%tvcwVt#>9+E3f8?9`eENN!%w$6`jWo>C-mjH4xO|0Wf!v8fh_i?FRwZUdVlg1 z>FpnY-k%tP-d_LZy{jF0yc&IScRxCl-~ZG)a~^b)&Rjm((V6t|C(@aRG#B~Nf7>#a ztuy5h%iqb?nVsmb0c6tD`FL&G3(9Ti&yV?_Od43e4J;m;mfC~h`?NEdOnPb5SC@}X z%RdpnJyTyAn-&m#(mtm4{TWJM?m}O7pf7Ekw!zr6dDt{pU;f(hF&`iwL;4cG0Q#hT z%!lzY^HVm>XKdOO$EH&>3s)!=iT$(_kFhgtFV$U^loa_K3?)srh;?V zho`-;Y>#4p{lOX>0M>Jc39DD%r~G93eV+O9mwVYC3)*)9uUG!t;IC)1?D0 zxs5r@ZOma_WUL*`6Kww4WcceLkI%$8@z;lBkLlwD$7i}7pQ+RGd-yBUXSxh~ ztaTQzZ!6uaJ$4+w$=K|D#{ntNPluI+iqYR~lw&oj$B*M^_*x%KX~=DvMx z@~fOV%ojcFe8KbXBG2{Zo_EF1vh!;5GOjg;&4`1Y!CZGNdU71|hGOhm8MvCnd|;~L z7Ux7`M|nRlG+-=WYb;tz?3x-%e)+w0eS+fo1txx$UT;zCuPziiejfgBajz9&4sP(= zz`RcNf~&x>>XV9@f5+IbRC^M~i4~L(ced&a!plGI#Lb*ovBn`7pK@SS{LJE?8!y(n zzIVM(vCIj`S`LpqwD;_=#;lm7_9Ij5QD>GYju|3WoUv|2Ki<3x^?6@Ec**X64e_V< zN2T{8SI$ax5pfDD&~zC%mAo6hDcpl|?VBY%#azJfbgp=MT*#awZ}iU`d^O;01Kt+s z6J(yExti#!Il-soPe6xQv7U3{&kerR9v`VO0RO@)_+NE!=i=}9emlOnrd0L6neoNG zpAj2-l~0{-c3)pXL~-XJ@z{;MNA7t3Y~nLV7)JuUc7Wd=;QV`P&ybT9OU*Ms-bg)= zVgehH{Q$bbtuu|yg%_tYzJ-jhn)mVXR@Ia92hSd$?E}QIwN|GX+Wo{sq`zKdeW;}) zluWQE_cr$DYh^8?YV^i?AEYjEAqMxSE63&qY?C^PCW}To(@DeY9 z?^;Za_w?{2?kl(7hrag_FY>ooVV7TPR9DA*6FwUA@8k>UT70XTEXI>0M-e*KJ+9g; zKec)@{u;X6!@aGH={|6>jdAW^oV%d=Bgh=_#?J2}b59aOd&Ww>v)>AD^cl)-!+e0o$fH>gota15R z56fZAq<|Pc=b3kNuQJClXYB-BIXrFuZm+Qf*DfRGmE)|j=(*ciU-_Gl zvaUiK8aFiBva$YC;nrv2!~O6`A}8F6oQbxr;MmlUp+i+mS_vMiz*{wVt0zw>wg}$3 znrESho_+UK_M`&G`b~2B?$6SHw=9B(u6Fdpemkx!IaWSvDhJKEueJH*=%XsjA6~^- zeZoWNI5N{u8`*Y6dh)8C5897T&)yTRzLC8cYQ07H|1j$wobSSVP*omu@YuE`S+-3%aNi?WK{|rGkd(Z+Htx}}nRZQhOLTkSOW14MgRKV{ z_>8?v?@um$wG{b*r)OMC?c&n1>f~L(Avt0V7@JxYw(kd={c_XzQUBh0!}Rzw?U~Ni z&^d-Wi%)w2+qQL1xNR(Yq8L6NhrXKxUrvQT%dDj7@8&AU-qt;y*kOfNP`_%^rypB& z2Jh0AaPl9()SoZ%E068+^3y^546psynr1b$x}=C0^QaK}lMs^|%YIlPr~d)g3G1`^ z)p=Fw$NiV-Z>0G{>!K%VTVt+a%<^TFN1{FWWFLOPUK}mR>&9tL|M(tz1NMGgUR#~r zf9nAKD{s!bKCX3ft%*;tVK}tXfx+w#i;fcvPdG5txG)3;r@>!j;lj=BIO2iH#mB)c z|4aE*TB|mC&SK9C=qle!KA6rg08fWj7~Zh(XUI)}H{{#=#@(}|Kl;EQ;*FX8z~$Xf zUHtn;_+z^_5@&ee*Qe%>{ZzSxQ_;1>A$Wgqf2^Y`lee?|F~Pst?8- z*|qp7I?v?k34F+3S-T$Jo4-@%oA~(K&1VUstH$%O^Ra@_AJ61G7V1(3=Xh*O0$5j= zJQ%;J4If`p|MUtz+bsXueb}(l38AN#^I;ul?QqX)`I%>YKk@L2WT+JV=;x!oF!6x~rTI2Z((Qp#BnRaq$d!v=ptbJbOf&LcPI0N=|0ECAjza$q;m2m2&8 zKiIpm(MP_eaPs)c5kd>ZH$kU%4^(Su+C8`##>+y0E{W_m{sq zH(3SVsu-(j@0WAoG0v=DZwtFNE*SlH+93}I-}{^}wfI{~>n{jDWquDvf52~#@sW?= z&h084zG?9Aj7$Fllh2vX&xlzko-{c>;5J9+QcmD|vXwjynh2hn=D7y06+aVPwy&do zT)e))8fuhv=1B{7LG;Rn=DV!i<_h@ZWq3cIu@(FBOwDUP`C50p-?vBiDu_|sfW8yY zXRFIMko&m4%BOaGpzF#STrh z@9iDXB%k}=CT~QwcE!-+T;@XRXO+{>O8OB$Y9637Jhk_v;LB%BJ`=a|^#s%hdtBUs zttH}p;gIf7-o8}40U1)GBC zf(cfjc_VvmxU|YW*|$fzKQ+WoUAk#c7WB%DAat7w-IPNsy4iYIYp&Ti)>-PFy;+>G zxqi0r`gb0@e%HaPi_fl)XX;1g%2na7y5rWE>(CuXp}ilNd>$TCK2C$=p8FQC%6GM4 zb@1%^2`M?sx{vQ6JV}P6ANQemHSQexXw!Kx4ljHAX@4`bAMsM^99gG7jX^$*^yw1X zohsdm{>(v-YF;EgQ;)8co^*Af@;^s zpeKTDQ60(-&be(6&%t{Q@ZEag-@yCZvfx*nKXBmRmZckud;JkFj6Mr{2mFeM4F>Pd ze&B7%Kk?-K@M(Yj|C6WxRBfVzw+84?LoPxCHbZvF>_>IJ4TJ0$K10vL;8HeDHY$NX zAsZ=sSHmZPJ=)RR(72!2mGFK!x=`m=N)9Y=r@Tlz#$uoSd$;rgG*HgG&ZlYMy#R9J zW85|1S?_J;J?Tc_*vPfy8yp+>ynLVHwLtVtc+s^bQ@FPCoC48et|jvW*`Q+f%axqv z4DsBVJXh@4pOg>o_?Z9Nk=9{uuI5B!E+ucqKTF#q=`Hb%#uPwTy&_*5y&`EUo|ImZJ{butqQ%l@YX9ep8esJk$JbuF?o>)a{JpSliS>w6T8PCu98P7pe z-=5B=)%-Bl;G33$7$D$`jN+)WUJCW9!#o>s6EN^?CPGZz6fF0LC(|CC@(En2S7X-?NADJtlIU z%NmP#K>kDlFu1x|H3W_5%|h;v(o1 z&Bctqf^P0SY}E{E0;nr!Auh2O8=1>^JK(v_+EufPFS2fmAv2xewct!X?4?*vzFNVV z;tBQ|B(@+q((}5u>pH**{H1kIH^%P8i)zbb^ktqwPpbwYh7WTt{pQl=5n`&_kwN0W z$B)tXx%tlh8RX3Iw{kxG^r`mQtDV|CUr!nO%+4*h{Y`Iw(hWKY zui}$R&Pb5XP)&niKb^KG(DpJu%Nd9Ci2R@a+Lk?%O}>CO-1pU%?ysP~mGrB=Zh+=C zPf~A0drQHW+T*(^hF|N2cLn+aK0dVaclMc(>w$9_`a<`X|3Q0h2BW*DQ!k@>1p2PC z{iZ{%FT8?hYhSm0;X3)k^dmmjx1L`{`^$NzzxTfEdCv<|m(G`BAHVWaEBsRJniJ*9 z>GB18mNKpmbW0(*_^uD}*aFU&a&Q@w+|chfcxplpIEHreFQ|za8}YQ+ZQ3O7fq9eK zl)h4%whxpW{b$+~?DBouoHqXtZA_$1=@8d{H?<(}@*SK>;Eiiu&NEDlQvY20+hy!; zI{p(wr^F}W3!`V{XUlJ^mfr>~bzV<{6Z4q_Zz#UdVC6PH$cHn>o5@8WwxHZzXv`Q} z;2CPRpVU4J;sxd@O)JpN2Cw^5bE-qjU(1-|A6jTTzn_*{CuK`F9pp2t5DjLsPKoS`mtvN$Nq*9D?075pr4tlu z@bXiNfB4;{kvq?v<s_;8UBdkQX|2Gp*?Y@yx!Xzwa;LTJm@73qzubn|xKaj(t-RQ%88QR#{N=rhP&3w>%o z0P$!k_z{mjO{^-xeXTRde+!EC=n(PPns+#>(%~^|!4rZZfPGLcde!dD_nNs3@UX`* z@#bK2@Z`ou<}GPnL-vICe{uSb!)KCN@tUb?qyG@Hn!#)24oY|1yk_swVf2b~-`*!N zFFFU_`n2IS)fwkUXL7v_|F9!ZexYcc7yVcGsj=TZd68f2?ndmc&WcQ+Q*3`XC;Id} z@;2Zf`7H@#Q}DLR1`BtzcS#QS_|CS$yEs?DYl9a#yeuQ^^7DKXPs`+Iorxqn)C!M@ zx5eY;9=xXda@pk<(EmD%&hT4kg5kG^9Dc*zjInX0`K+UZ9b9R@75T=J5jS_k_P05| z3?3g|E=m7I#3ePH@qVj+hQ#l;jE~WN9JzSOh-ABjBYC{IsJtI-qAKI9f@} zzJPYtpf{BpaxpMvJPZEpToP&#Jngk$kM-U`>X!=Og@fpyN$>UT`5f=N{;pS6TIs(Y zzk+oM6WbZH_9&L$P=g=bhpw(fmv^9}y3ixq^CRsK+x*_c+#epkSojRkw_*x)_}<_L zyJW{UMC*OTAG*+is_|1kV~~61Z}?jKe%Sjb*gC?CALGN=>p6$Co+CN-{_f5Xzl6?; z_A6t#X5VZ1wvsXT`51dbyZmH$$&sxuGX6`Et*iTytwZrl*;0L$!aFZqVEcxy-*FbQq&0;XXit8} z^D~6A;^=hs#k-rp&28Z3AuG38{;t*~fo=E7KNz@ouln8G)-v=PZM7yb0hDOv2%WUjPQVy2Q7c6p>NqXID5%j2*w%Jz%ef^1;|(t( zrvmwY|F!o{c0wZ9neTbNkLL+{pM6>DUGIAD@4MFa!o1Rkxo7R=JLK&r0P{-P)7r}w zz-;UI%x1+kUO--~z-PCw@tx3sHRCyCO)IfN$=FSwYTvM&$Gvj^kKXIyIUWzY96XbrYvr$dC2$?W*-H8zcHzK!9Hpd=x;GF zm(cbiXuIKDtEh~A^Q`19Kj%w;{ah=#Z61G}#Mx?~?~Ukr*$egb{T%JeUTBf*DJLhQ zF_c4l8VApm;g^>YJMrSX0(`5ENdCWUonrj@GWuUazlvLwG2RNsYwIg$7CJo0d>VQ~ z9_e^J@~M(}$+CH5%TjnG)8ZT1jtrEK(2joXK-YQc?WgCn|Gwx8+24U)$O@;;=ucNy z4?tJ?=287D>MJ>BV%K{6R{UazW%cPhdNnkJ?DqEk1Nx32ccWK>8(;|jDQ!lTL(?uF z&Z2MWv=t5?-VIIvfj;%T8GpCKZ=cRPZv5T-U}(RJ@nc7LVYtp6|5c11eV+A)OzaEE zB*vDHOiFfe;nod&O#VI_T(U`FLFOqCle&E-PP3Mw3g`;eLJRi9e1D?XTFEgFRW4QKvli5F|cfUi8 zEk?daz9F=Wu5fEY^bK_jy`dPkQy=2Qef5sq~)9z5u`_PJH8%8l9EHn3A4;=(~bs)d>@*~MA3 ztz1Eb_b1YJH}QG39Ti_XZR?(X$AHOg*Ul3t{+H_5hHgCI#CT|h&Ro3r#Pt}zVnRjE z9Q4Ens)M!<}>Q@l3C=$ysy-E;T~QEDB;Yr40% z?ov~4uk(_+6uX|)x;T{LzPrBfyRUoSt@pfJ@4Q=ie(3Z+np({@#SMMmect=-X3x8u z-G09k+CIgB`LVw5KJ9(?Nzc1ay6;{RDt6y}s_(m-yzg%Jyu00fH!n2VefQbE@78(W z-Qjt6hdbUWp>Mk5ZSDK+J>GYBd*0n`-rd4HhThC^#@p8SUDczwb5#7=rKZ-*%+YHp ziQhjBZ(=)3UpzcS-^HB{omXtFH1%L3LW^6Frz7v$>a(n8+-s7>=Ghwi*?-V8r?R$U zpl4qBN+?FH{MUG9C*Knq;F&E22QujMuC4Rb&S1|Rwx79#XY!qAWcw6Dhu9_;DPAid z5rYz@ddos^Uh4qJ2U;V58f}n)z}H| zmn^rBi_3$i{o>tec0X-trai&aLVIV?p6m_N7dgi)nP0?00ZJK-oZHj;O zj-R%YxbD{d7(2(w6?t`|_B<4he#1Ie;iyM1oEl6#?ZJ1Gp}*ttOU_3{J6T(!BwF%uKu`EU*uqV@T54sVkis1lYH_L2T!j5IorWgnS=YqJ$Uk;Jf6VO zihekH#)F@O*b!d-Oad>9`3AkITfye^WhLdRwoLOeEkXQ`?-I)fDZH4uf zF}4D>so4{>Kz8*Qt2USpJq@F7c?;`K*q6_IqYqjYzi(hIo%M(5p#bz4jmb7qJpua} zPwKaB0-6$C89H$CP~LObzVcq%2;MS1_S>b-_*)sbWRw2&@Z%u+-|77TK8)6zM1j}H z1Z<)e*9CJojS4=v+xp(tM(d6T55NaIENfZ2welg^)~bhx z+R_&7NoZw;hC;)UQ<+2fJtBTzcedeo@tQTjw`1M+jeefV*#FQkhM!LHq_E zc$A$QSjRQlwOq4hiY=r2-&4(44E)5Q_Z84+qcttm0p8U|mA*?kY+5L#n01q?b8mHQ zTgD=LRr09WGw%NNZioC%Jd$q>9UD2OCmtCohsJItmIG~MNd7UV$36yPzS=@a(SutSF6~%Fok`Wopx`bU25{C zf01^l(C!snpGv#hn|Z3vM_P%?ef;Jx;~r>~{jW0aVZTrG-HnriP30HlH{CL;py^;v zYE##=w5E4vr#Ib5UgO|Gz8i(iF#F)o#IF#&=Mj@A#6O%ze0m|Vd7~>7J68-|dk79~ z&SIVCLEt>^UeuIPOxneqO`d1emp>X~svd7fv+Gu}dMmvL9PYar|xvfF=*NyEEfSkF35Dz9Uv7TF{=f zst5X#_dc<#FtMC0zNu=>Ix`d*`h%^}(j;SlOO`L6=*V&x2JqG6*E#z8FRY>MFDEYY z=%>lVExq3l59dMa^Ju3~{MK88rhJyz>x`2lg*9JTkzXlqLi{!YL4Lhd{L z+mjq|?>>&sh53)3*o*dv7)O?0lK~T_Ridst(D%uD-P&i(Pj#kbHJ)qw@RF9N!~! zzE|SN=Xs*J-h5Hy4rs1|vEByVtb%T8!OsI$qH`_&MIHIL_4p=^Zm&nT2bcqY^8A^l z=yb^m@tf+}%dj)0@1y15ZYlOJdVh|58uXZAf40r5*onV+P54+{93H)UipJ+_mM+VQ z%LcG@li9;31S${m48D5caBvg{4)YFpQXBeRPhZN1H)JXw2tHkyp(Ec3@G;Sgk5&gC#5Vflqo>UQ z@bElwLN5*^(>8vfeUX`;h0Krgzb(n3GHO62n^rrjN)xyB-& zB0tTVAsU$j@7B{7c4d*;RPC+F2h*;8`~CoLan!SdQ8gIqGX@_1T{M79tX90jd0x+b z=g-Wc=eJ^{cfg17RoFAN$k7Lo(QA>Vb=W-X)t2~PykPUmpQyFiAlr2jea*D{b74?h z*QouOz_-Y$VQ_LBeRRKj4S4$FPrA~y=iu+V9{dU4hQ}Yfl$bsH*rW5zZ(@yJT77L2 zYYq5bZ-g;u4^QQei!_xbTBbjka*2qVq|90Zf!oTJs<4thIGd+H! zgBQ);u0!0T&a)Ff>kKBFIvVP!o*Cl9PNseDc+GtCJie~|e2c%&^EtjDtIMeoE~iF# z*p20fhq6y~rZuc93;fCc9fmC*xeA#koRimR(Ej6B*)p*>1)4*b9e_Tf&_@jVh_8iK zz^(RUw&6%`4nxyMw>e|jpg5vh8FiY7eg_@@9&+9BLhA>Tz!TK>I(9=(E3kB<5ppCto5p zTC)$kyK4wM6W;Gjw14PJ#I^&|S|58$`q*c~dRqJWzp0q^RdW(LmRvo@sVN^}*X``j zPyF;$Yu(+~5(oVdSpv^(FtWx9?NY4~Jom7XHC9M989V7K{vow2X9Sy8W#u>BJE@@Q z(}rQ8C1+YqpT>rTmS%Iz9v-?e$7-s)m!|VqXd@hbM&+=(qT7u#bcezaF6j@Wmsjy$!AYgJ(TmVnEL zee-87wr0&3-(&kyV@57CN*`hRi_o8BxA%9vHsAE$M-Ku=4-9JCV^<3n$q&hc>lMFk zgJzL!(Kl_o`dsA)2C%FD9lhDX7@zal)!&Hpv#Z;JiO8GCVPxa?|A_AAca+~TetXCM z+FtbeAaGaz(b^Wnev7^VPwlqt>ng{d91OO#4s7V!J*7_>dvecWa&*u4x33pE_VsSy zdIPxl?ZTn4Y5r{7Q0JL%%`EI^D~4XdR@&#Wv*E7@{1t}3WM^lCJJ-(sHZU0*(6+Pp zlz8BBZN;by5A&)y)jnP39p>8D+5C3x?81w&vw<(-!j7Ho!k&Ku{4Dh1$F^DbEXLmM zkDs2lKLanD2C%zZ-#b><@)Gl=dX|>dKKAT|sn(d*et66SkI%JwZ0^s%<5~Uj=z-tG zqxeGjb8)9Rl+9h}*xc)ZFTBHnFPPQ`zGw=%)QeZo*n4;gKAsPMo||KBqsHY>2DP^P zxwi2=<&ZEkpOq_9iwin|&~P?ZoHo&rAL7?TxbEfy__qRQ#|Y1-gzYPxWAFkNa@GjzxMVbfs9K~F# z$_OUrXC0{{|FtjtiT3QvPw06+`M6T9e{|?A+OGW8(K@PGSR-a~8`#9toNqwc^sM$5e}VPG_!IYIYkT2R%&Yh` z;KJq~Hvn8NANo^~UBsIE$QSw^J<3`q*(n(sMFwcBTi!iZH-Wf%W91pd`BJ(h`y;?G zoA|eCqQ+l*q;4Gd3%DP>7J6F5yVoMOKaanIyy)ZaNSAL9CSH95S_H4{^heGq(Kv=R zf_xXk#qCnYdIPu+ZZ|5YF=k|_z{c&K#^BlHJMP>{UDKM$k@2C+s6m1 z!f$tsigW!0e>6$?DDW7>e|cDSUf@tUCudH!1pbs2ng`w@*IJ42r{rjikCGWW{Zj1L zYl%sa)03@OLLR1s`i1DHc1)|8ngN|FzYg6R`_VcTKZyLk7r&y>G-$LK-#3oFs|lut zRuE%K#rNF|KTqukv!4I-onv*C@V0rrC8f{vcKlK~88<#_&*O_$+Uj(C;&goCrWjCEo9{67R)`1z!#S7g~wv z>)^f>A8IKv;-S7oN189u`H7WyYdPPr>f~Ei`>n)B(9p;H{TF|q@OPNM1b<%??j3qO zFf`wMllraY$jN!=$NlKW&S2u*RYCd-(l35hPCjtt)1Hs^l4#FQd#SXSOna;iX$6n3 zL;Ho;W=n}Dq~!xkF!9NH>Xz36N1c`UujRhP;d(2PxXqWi5gGJFY|e9TW^dIr_Fx_v zOk9}s#kRjowf64!TZg{h5PImZ{?W~ni+=^)zm~IR3mLe0vh{TFnbZwJL-WXa>^3=% zw458*(;%Ik$^mjJ)VP=XH22W+E|XJ9?XsxZyIZmle}-D7G5f8woD$Yqy4N!C(Mrgf zM&5%L7~ePM5O;db_)2?3{}J+=qL1i%tOb(1pq;G0*E9Kvx45?O!eG-Cx%o|3<`*;# zy})W3KFilMGMLnK88UDPevsnk@Q|t1Jcs_H;3o#&#gEFh+d4XxoW9k35ZlApS>W%< z6l+}|z?k_~TZH)-!Mi$Efxix(dja~^vr*{--t~gd|v1S6<%Wv{R>?@lU z`lnrXtfs>cP4TY$#|ShPhn}L)RV6VXdmjydbA6bYfoDwK{@w9uT=|U2hWp&mZ2DdC z(+Brm+PJ}tX=Bfr&NX_qgx|_{h37eA0#AxX#%QN!O#C)u;$F{~xUVsFbi=QTvGUs; z(^oX6Mc}&|eDJQu6o)S&@JkdP@s8;xjfuWpc+Gd@oc_II6)nkEb;oStSKu+#GgfyV z`@9FtivRVz2b`L(!>VWY;L^ok|L5M-b62yb0J+@(eeH&?Bh2#wVr$x?Eqy9$)yVC3 zg2Qn>d!M!;?9-;Tqsjr8H7hw*k>pRFHKZ$>bH1}(JBeoxvd7_qqxY`Pv4*sDvR-dB zwgEb6&yDcpz3}8q^p#5O&166OYtK=h&>AB<-(7duIo6jraEF!Xy3I;N7z;eSEsuOm z3AWNk^yfHgR0`0au~sJ@&(YWwuLlz^;+IF67x5J~?3NPSpW0G}KX@tqs;)|V5Ntrd zs_o9P`SgVhy-jUkuYeO1g8&}MF~yJ{mQU#PA$=F2pG)38R+p7+bl#epmb5-P?~vZ( z`54b>&kV0#@}9f8zwfo$g{e zALB>%!;k#sQt(qOysWkDMf2PPwjT8Qo84W<>a5e8u?N%pjJ<~On*7EN@cCLWai@Me zu*TP7dm=w%3y4;mlB{)EBglQiCvKiJlX#YwZoy<}CHaa}l14!XFoHK3dm`JSycou(7LwwZFC(j!I(}r-ru7GVy-)_n98LqM%Q_e@o!_c)#%vpM3V{$=QrE zd*PlTxM!Xi?kr@XWQlG5vHc;9_q*{R7*xJ;0lhrTC5N`l)cm% z9vY6$Ykvw|yw;c4ofT|)X;OYu+Xcj|i4!kXteV)bV%3WMmQS;qw69m^lNPv_tgC!b z{bMAJDLtAtv*2j=b|3ep7w?|TdO>tU>d?x4FN{fA-`~%b-}=pA*1Gsp#7DO0Cpw-* z=kGwSwqm31mW|`eD&`nn^Vlcdb(QBTMu0qc3O%+qKk*yJokD+t-;0mW6AO{N4q#j5 z;a3k~f2L8imuHRYik{A1R@uHpv>N+*A~N7sZ0%EhiMK{#e`Bw|i>)51wGy4n(K*=e zqZsRG#(F_;^!iiJ2_}|*G5BD`IP1_W=%0#HsIMb#q2C+X2e)X&i6k>__);> z-zxCD9Nn6Y-8nL!K8R~f%m;>i`p!3f2}cLvf%lNh!i&B!7_AN_w14rzGvWCP{D@3n zqM`tPrFQxG!vpvG(NJq$J2>8iET4;v*o>|IBIDTwAF=mZ+l$bOWK}oxlDcu_L#KnY z=wsl#8rf3eOB5?NW}OzgJO4(`e%r*2>FoVxwUuG9Bt z-}UsRIlY?x#0QFf>bsfmQ?Gc4_|b>{5zs&~@95jcLy!q0c{dfBV4t~#;B_}N^oHW1 zY0aI`$P&JzWZMvak)IokB_~4=p zbS!kdD8qLspYPZ&O16G6j_-KrTj&SM;fDr`Z-=ly;bz$y>BrIcx4?yoZQTK1ZHFFq z!1v|Yk;IjDIO}R9PZUE7BM){$f6C8(?M&q49ni^k#>4ZjjZw<_vX{xB-Gp41jWvq* zx{xJ{d_$c$VW!#Jtc93W1UaMn&aZgda@&y%P`}iz%?#J)=gRjwS-G_0zq;#+C7)&E zH8{MBZ0dZ{#-I9#Zbl!0`$q8O)i>UA**mGKC1c)Ir|su_upBu0pf?9T8@$tpY}AL@ zCp|LT)+fq^_r%Shn;jqbjdi%X?&BG~y6(Z?$?3X}o;gX_@_J#rL$Kuy4BL;7hYfj> z8u|^eSte#TJyhNowlxFj;WdiWz6xyaJO6l+@3ioa@UV;y@q4LTYMUPoe6 zoQnJy%Y2_1Ok8)kcP)U$KRZ92Wb)Gorz=05+$TRB#_uz{f()JO=0%-aKk2eK`ZdPA zD7r=dukr*1emf5;|5Ws_T4yUf)3 zJ-|4yi=yb4Z;*2hdwxIhGS4E9h`ZIjX zra;%M_sOkqv6f|_uj-$-mK7t%b#DW9$}IXf_rJz(-mCwyZ&?`Iw|5F5rIN zDCgL*5aTs*VlvLEhi{Q2YN_XgzHim&$mMh@EcEPO1x>YHk7 z40@CP>GPQvCf&1d#f{D8-74Nq85PXkF^2wW_k~IK?OSon&->n6=UZm!-jv>Zid7}^ zUOo3V(uUU-{X6;-4bt}0?z#{8fu?2=d&8Csw$1m;N1#RMLNT9cIrN9Dh%rC*UX}i~ zcCPK9*6PJ9Y`aC!6Fza!#In8;`U!O=0b;~HV!l;c$8o_3dmYCj;=sd99C%BKqrc;~ z*>gUL?@H)=JUTp?Iv46Bwv??cFvlfz1&L(Vl%w~}UKeFr@5l36EP1%$fqcl z76^qIUl<=HS_%(I)}e>TK!d;6SfMRjCMKKO3(4f+qEC)$$9peM3?fW@xJy$Lp{aAB zDf6u?Y&EqnpNjnR7WuJp`ZRH6j_R}RcfCAOf_{)6ktIL&R@NIiG4yb4FfsfAa`wQi zcWjuEwX{+A`mwquV7BMwc(`YCe;~N?a;-y(Vdh;3pFq2295G+dI6iJYTKCblCqIs# zoa8u0-aqg-HsaIW_8H^&3HJvYN14HYFWkFi|0d{R zA@F)>A?#oJuMLCHN#6gm4q3>Oyb^41@V$$53!7|TdyDb4v5V}SioNfV)|$@&Hg8Vl zIr3Sq{1h(8bL|Nv=Y>A_dvYpE;mvB3Q$hb^hIUeC6IAd-}obbGUwyq z59B!f0X>N~dU7f@?b|rF<>hz4txJ0*ujA10UCfR0p!L}0*R%Fy7i(c#t?Gwr_?}!N ze|M3C9%$}{)ty-1)$SQw-_0iuh??v9XY`%lu6Mt8V&C;hQFt+i9t(pf$+82;vV*l& zq659rj^2nDa{p@h^(red5&Ki$eN%o+I_@&+92fc0x|&B?MLFD)U6(YT_~A?|k%N6R z4EwZp>X7wa3$4U0X9p9vV86YC{2s@Yn_W;?tNjA3W|K-n{J$C!fqbn%d>7kgbL2L-5{#9O%qLA0m4XLZffZBo4UHhUFc_ zsIIp4v+{Yv|C$Conazu7U+bRSw)D(rwWo3I#8!02kdF?!Wi&Z~;a<2in@<(oT=Qkv zaIXaJlkKAyX6O&t-pR0Ly8XSZ{sP3uMyJ;P23);H9p)(ddE&-f4`(L1c*|@)jsDel z7@S4%L$3dC^so6Au5NihZAMV|;@tJ=4X^wCMkl;Gjhr-ep}Ei)G#>-uw*MRnr-mNb zbMWWn5O0M(w%qXT!@n3cymp6gWLK2E*3VB3CYCYoX{QXYUC#NJZu<7&Zr`b0JG;7z z-kAms%V%et^09Il>xNI$X4-2>$m65Cl#kV#kvKF(J?Of2=tH%t^3&tcOOHJczF))U z?fK2~-Z3tjjPK2Rs$FbFS9HJ=s!3T2Y@Jscxnbz(pQ)QjVt?5U?91;bPnJwxEKXjG zIzE#ZYo(sI)a1qdT`ky8@B(sb&yiEBu#$2Ve{}CDKNdJjE#Ewim;LI$ zVtjfd$9$?Iyw;Xk(RswwXb&L}x`wrW;cMXquAMMHRt_HEryd@-k~xH*I%mUE$U^ay zStCe|4D~)ww;WEJso#05YcD?By-mKM3+oQX5e9by(aB)A)$?9l+wa?R=e=&{nNpg6 z@dx>2!Ct6KQs}8CKg$I za|`?PIcTE5&_q&~u|a9O1(}FFxL1CB6>}P3PGx)MLnAw(5#o6{CA{n2%Ot)R;CGaF z3!qcczT2+8Y34rXUX#zHZ{T@<|JKZ%$F6htE1<2OyuW)b&BkBHZ*2Vi2k-VMA5M(^ zsZKTi)&k;=A62LK@wXzx!ouW-iqZXk^oC@O^trBQk?*mr;USw|{LNK-Pit|qH8ac9 zd!HS8mD(DunaIRmju7h}R|I_$CkkRKW~y!qJdiV5vxK@Q#o(12r5{tHBwE!y)jmad z=U>ip^a{_JwZznEa%_3p*5_LGc0G4>0sv zoo?i!a5j#5b$ti(kuM*q3o=%VvC6)P(zn}A9DU`E`S*-<1?@MQ_ES#Keo9aK_tC!g zMB2#QhoM=~!xQWe7lGEJjHyt2M$u=n#_}d(+0XSDbJ&T#??RTnhyMQ%{`ds`_!R!= zmd>a4fx3pb3)6EyjwOfqrf;9`%RUCbX}`tjP~teRQEO1>Kkk`_z0a(t&)eh!Slil5 z=jJZKHfV?UOlAz}n;Gr+%|1%bCACX8cj{1bcPl?%2jUe7^N2T9H4-`!j^Q z=b*h9?ZxyxI(*~^d>(#^_GVj#HVj?h%U5cy;A!pkD*H0RxmioZZ_ROp=fk|)Czk^~ z1(C&~BhgC~Jp2r~w=)f&umRnF7QQXIxxHR{dk3FMd+gqQt6plN4kE2we)OjF;P)mz zem{CHt=z}&#xod)d`)N+|7^_?{5r;Qk;Wlk=-+;C$7S_<1#t!KPXgSDdl{qYpX=fi z!Bgv}mH>IIXRE}!&e*ix7nsa-wQ(u2Fzt1s{;sFLP56U9r62i&E4gm=P^1r&8^ETM z?bR!*9Usr3Y1sf?8b9tgaLoml&Nr~lihZF@Y=aKum|)zGP5jPV#AVUj?QiIt#i^l> zh$XJ~e6#orhh8>L5})`IALEDhr@N8PGZ8;?vXnV9dm9c9?N+@HJm6z2u|n2_!wc~| zY*b`uLHZeM|}v-YAtM-=S$!J z!Rx76lC7*k#V3(%u^3pCACBeGXAZ|gjWc;dGc*t5;ci}J;_m+g%0_%Bz{7jyXV2zYhFNNg$U zA=wMLhWvWRqF`btH8*O1DeZ5jeV;?qyYcBKGR78S5R$bqXv5eY$nxmb^zZ1IyGN)l zAjh<;yip1{tw*l3V^bU88|;ol3oc{50(8~kPvv`f6>E1{cRrYErPih)U-jJr_RQG( z7O>nxEo}$zb#mUpev>igB7QaJjBzgc_-yt&j)S90epmGy^QZsDSQj!@=D)k*)TZu= zG5np@r02y~HT0#NndqUo{%D;|N9Tm%$T`Im!#k+UaAbbCfibo+-pyQJ7^F@1z2%+f z+oMlZ_ly4C%*Xy|$?vcS$jNrUQ7C8_GPJTo$rO-x101 zvBnKqv>I8&UeDzbWKTv$@O$SWi}bsM-|1fn=4SrLw@h*=X2*-xlrm=x%-Oo%*)$O& z-k;5!#fd}N`}unGT&vxGDYi`u`GTFy?_Yh^I^uxa^j)W9ey8w0^V^dpBe5+?~hZ|@s8yZ%gc|2|T zd9RO7c3_P|XWIKaI+6R?jHwViLN=qJZ{qc$>EP%-w3i~+y$EBJtQS3KUklA| z!x+cbn@1bKr+17T;wW12vDWn<*X{kWkEipG@OQns>RskX_D`|;P~YIDho4Kq%ToS! zcw|7iBLi;b2rUGmk9_9;>)!X4^UiJ1C@{M1F7V#JgZmZyVYBqemTL6rwdkx>;A1HD zve%p+3Zyd^z}sL|?$cb{^BHsD&i@AB*$6ySz5TiK?;Qv8-i|D5x95J3^~(!GzLdUm zfAeSbXZv@V&GW#Y9aFHHr*OV2XoV(2w|PGH@nG#r8?d|rOckfGpDVe5is9rYkpH#5 z;PTZ+JGMXXz)xUZS-B%q{obw{PlwdUkkuzOId?j}^XU5%z+&-@D7a zr*AG-Aa?`EQ1jl~_wRe+i6_i^RS5&z_nvTI_31Y`OiPNZ^|nl-1nuo1+(!D?RS1)-qHT?Q(nD)-z!f%{=wxVUi;wj``5SlmdB8Xf81u+GD>w8KvEW@+aO+%fC_N-wK(*yTo|()s5B!Ivvw2Ut zd%UhyQG&_oLhD znI9k5Z$`IGef<7?`dvj^<-#+%4coa08)Tc{E#_Rloos)>Dc-kX&TL-5_;SdnSI~b6 z&rS9Pw=Q5zx_+ZIYers5@B_h!Ze)LiG~!9=5(;$ArGG9PkXM-nH4G#FM4c;druo&4!it!IkMO2LG+bu z(CA2H5O(OhFZbG^%CY{Vvq!G&-`n}r1@P)e)&26R$OhHZ_VKgOFY>eGV@0m+=V#%c zwZOmf$&5eGv&L`3|I+-L>+;8)>u#JSojL6FL6VI<6g^XN6+d_^`Vl|)RpM;gUouMD z#t){AV!?qP(f$ijWOWQ#9Y;R5-wGWggFBJIs$D%;fgg-~?LxM`U6`Evv0wg3QXf5V z5ZoC*xHf2F=Fy4h0{p_A$c-2}x(vTC1$tc}dL{4a<9Zo>;RMkuzpIcbrTB#v_=N$^ zjUNY$_z9sTzU8nJ9`m7JpzCe6?`4m(|G2}>+-cq_J@fV>_}9#r-;T#LeVzI8_l?Iy zsiBHdOBG=)Ll~d`QvAU7h2R8!@vdQ$?_C3KJ6bLC?XV%|kXPo}de(d_{wnA2>JZ|( z_l_U3J_p``48c zzG&8I=vVZy09Iax&ximS%_VfN4yUF363u5dFr#2=TU94 z;McJ$kv6k@JbOS)0CwQ2{YC9VT{D?u?jg&^ggx!+{ek*W`@aIWIka&zZK<7`ptZ?? zV4{|I_LR}XYVS+I2Vw3evtI7*(pFTr-pec3}1%6_CU8Rga z%y>+Gl%wPo{>7G>%88--{~g~QJMSRl&bWQTVFMraN5X+%QXK%_q+s5N6L`k)Ah3FW zd&gzo0|wzwcwEJ^w=kBQvw)wm)Q%onTh3U%G@7}*@s`8;8SBpXx{CtC;YsIPDGUFG zv3%m@oYJBFa_9qX)Ci?@l@GZ8iQ!-OyGBE%8Q$U>B`1! z7cr-D=ynP;?6$3So~Nj7Xf^=ts_o>HXj^lj+Cj~c=v3`A#nWb_lUwPg2CM*`YJX=d z9oo@%OM7V|9XxtrcIiNP6Ar!eo`H?^{Oywuj{q;%BU8lV_%~y;ci+c_iplu<#AKd& z-Ij0qPDA^LR^n1{HW!}X3B3m4_Z`sdIAF8)=VH$`?Ojlb4@C`6JN)@DzSy;_WpiUs zNB7!xegp5y?;|b_j#8TQiA(swLFt{o^`*tWy{EBG<>TGV8)c4R;%um-R z$g$S6+>c9%+4jTBB@SNVp9T~09kc_jt)RaSWZX{re-3zBfgy}+-5?$LxAcpC?B8#e zx8GXkVskJNd5n3%W^v=a^;@0T()qx38??5(?|nDc@*_QWCHTmp9nk}It}jxTrm-k~ zC0Z+kUdps?l5t4(UkA582Z)C`wCTjcE>b*7HcJuDE1sa=+q=$M6<&)DB&&NIty1<@P0q^O+~vH|f}4qO1AvWxCI56V0CY18aR3 z>vt9&tzA6~8pkGYD`(A59{Uv)vhKfqo|SlnzSN(-b2zue7t;4V8(7cP%VR^4GsBy| z&bn)xkMaNM6TaUo9Tmk^(RaHBlN*iH=gBVmymI4rzjXWSD>vSw{vnGw8BA_u-|0VT zxiN}0h6Bls+Z~usBsbPDztVgE!tqDuTyem(=I?vXZ{z%(X~D$f$nUP1&Xj$QK7Jx4n79>PeTZup zrm}vN`#XH2yEbm^E_!E>Iaj%;3G6ynP#jq1A5nV)mP!$k>nr+PkF~9WOia=ubTNktJcq zZ|)SK&-DIyzw<^CXUo9|& z`_R1nz5e;M2su1oI(zd?wAWq5m?P0x1$&hopq;Y$*4}#JoCl$6U2EameA?CiLko0` z^;!5tMT$pegRct46l32%#fd9!AO;S-Ew<*f&QABKO|WJ;=g42d$92^?y6+YGsWAsv z#mRgp3_Pet(eLb$BmR^=P`s@H9K^tZ>b*)dK6JCj_Z9YLbg!FQUY=DS8sFmG)c85) zYUAH`9IKN)(mnmwJ=GX^=Wc~;TX5MYeqHF$m$3;vwGqyB%h?Mp^`awnCDg|G&~em~ zz8|>f3wAB(`S1@qMS3U8tf`t7TKe;Wc{|=J*)m3uGp&ani{_NsiChg zKptrS?dsnBx5ZO!E<8^EhpKJ)V(%e*dXIbw!%K=0@BS(9L#vgi$Y#7Cv}5{_I_c^X zYy)CE7bzZ~x%dV(AUc{uJ!-+|3%Vs>!Yj1@X zd+p@8&fW^w(5C2XC(roM_1-=s*V;ba#q_E30DUGQb7UjnROZ%Ez@E0xCaKHL*)UOgt z@XFjgbg5)+_d>>uEYLd<k`Bv=%YX z=QrO~--SHXo)d09ryxiB5V9stW0K89zlvujr`R$;_VvGr-@w1(UZS%LT{?rtujd}( z+mYXP{tal7Oj zkTGQ@Prr@&6@v@;N-M!dD)*iMALl|RnKzhR(ij#MtP2tox|_Z7B#YWLt|vuT!Q7R| zviNd*1I;Bqm1_G(^L!G1Y8;u<&U24AeYiGqe7Q}Nb@Zci<$^cSuX0z9(5HAr?}d3k z!hLU_)F*7)1l?nm?(5vC{U6fGEAgeg`$O!6Mz5#VeHS%Ms-Mf}_lmQEP4|q?Z(2R2 zpsBrnNYjDMLz~`ua+tmU-mU1urNCOw8Y}wVCgSaidEk*-0T{)axwr(Hk6Dr)zoV9jR;NxAy#>MZ0mvboX~@>vqQG$Xi2a?%aU8 zFY%mcX$nWpV;S@5=9u)3_85rWLI3cHxo_s#u9p(-?>hBJT~xaH4svjs1NvMspEX~^ zx#UysfL1Kpxd#2bjoOc6v+H&0Rj`d}7ZOr`uu#48+ zwW#u&yyL2A>0N6&mAyGs#~5$0602BidM|5DYgudB*)X!{;A5kjx}F-{)LD(JXWfNw zID4a45MLvnnepWP`|f@9i4QV;8Rhx-ADyhB=zfL#2(^*E@4(xu7;`N#T(f`al+Z`4 zk6_J)sWIu`UMKf7<^zl=;|gNz@XmR@prQS8X#X*2|3hjZ&Vv?*az6)J9CC$~7zMr~ zd}F&k8<-36k&ss%w-G}y>m4r+%@ywW_IDhcs2#o9aT|T*05kn4wldpSnLA`bZ$EA1 zp$|Fz45tpw|M5;>zeN3M=XH{@h+j!rqxr~0Vi*7H)VM8*Pgw5Rdx zXmZ%Yc`n0B1a?}vzGXHJgp2rE{$@g-4**xa)q|Ir%w*=~pyH~` z&v@oWdt6Lrey|^@Sq*MI06urmC--t^Qd2uVemor*TCCii(B@!s;quEBCzz8fKF6F) zK!2#-qw?y;PoqqbzJkN;c$BX7g=iqgMXkPTsJc{O7KbFVw` ziAm7BXng{7A7^Z$k#@#f5A6lef7C`ila8FY2RT9f(&XhfJ9MYo=lWrGPLJolr|0yH z^k!o^HF8^!CD=JN)D*<8TqBjhk1E1nHE}j3T0XvMGo-3Q$=_^hjlGE~KG_JTsU;JkO*u+Zw#}1wm z9J0rBUHZYreY~9KXG3f7lI{nvNAjpicv&$Q?&VR7Iu04xz7$?f_q8>SN^4Ub@KNYn zJS)6qsCI{OtY#dl)selILH;@KRq~nmZcC4LZ!>vC@K=KFNFyIEdE9Mflt*|!jeK}F z_^aZ475Lli$d((J`+MM3_4Q-E_aJ_K5xl0KmK=H%zEM73IC_}-im|vjS|J>vOOv@K z8PrjS-RqH8kBI+iZvyQ-&wbV3yrA~LjpE>Oo?pScrf!C7o50~_;nHbCw$3KnxQ#Yc zH?xv9WP7Zl4dKw(9;&lCU%2h1b3SN-pOo4hX-EpqXRq=l$iYTzjC}2hWBjC)=F_2b z(WUE`|2yk;7Sh+bL_jB%8&K6@6^Un_=9-=HuBF#M(GaQ8GT3RX820q>j`hR64~kCZy5Om@HG`Y zJT}4F_U)8l;@cCfrx%~5+)#2?;0yQzj}hmc2b~sL&@=H~Wb8Z8WHG#|+Li-DiSdlZ z-k67tLwm|EErRx>t1Ib4HvWE|sT|5PkJK zu!x3N0PiLbZ_CCMJdMD!1KxJw84f&xYa=v#H|GzkEt$Bsdv|J)4AXj>;Fr zo*nmH%kYis3;q-DDCRp38(ltf*6lVd#-8;%J{0XJmms;WF?NCDVfaEfQx~WD-lWQ; zT>0was|n!gTxi384R%zq-IrUBxQM>wXZeeW8_}1kN4$=6cr>*L`|CPtfQOUMhYy!z zNDgrix*m^xAzpXexq@~i3l*DEJLeANoPO2L1jcI(ZPq72&ntQ3^bN5A!5x zdql3Z=VPqqgY$vr<2hmqp843~&c}$8nGeQ5EN4w%{+N*BJu61Arz9#1AF!B@dZb1J7n;)@!sW`q<1ns$<@SoZobL?=xxGRljkbS;f10W;}DhnP(<4 z_xSUPN09R?TY|Z?GsDO!c)J?jUI}e>6vD62iR>M8?;3qawHdi3ns#_i;^%VkeMaQtUVaP3T)N zl22EpqjSb0yUw&s?R6FNl}ew_fxlMpCm%QStKD@iN7H6>XJYqPBsX>E@H}m7Fh1xs z^1PN=Pb0rW`bTrV3b-UUQzV;#Nqg6+-bL$6&H^#cH(1AVOZ@^mFlzjPD-Y>bB*@(_r{> z=PwGKDTY<)nM37f)-dO`rDOkZ?48y$lb@OVE9{-tj6UlLk|)6z)2yeHzKcEhWsa2v zxdrDFCujeVM)t6bOy%3EX9Zucc;8y)V|`bceMCCBzHuV!rnsJ0!+sHGQ``3+*0M_C z26;y~I((Hk$yz4;t-VJi{|lJ!al}|<>zVyXcz@i@$ov}TUJddBdoFK?wQde}XzVud z^(6RO&h^K@A#-&p>kS&=7tP&+$;>5fuK@NG*1@2kjGW0kpKrc8-w83X<{M78<0=ju z(nEvEy*)44@?<=ENcOGtkm6ti#*bPv>|7A-UIxsIf!WCBYT)JC1n{I|=5%6u7xQN& z<;q|91$9RQ!9BV^+#j8G(r`cgH^Lof9qSy{+zkZJJ70oEN80fzgU2J6oHRUl{SEM# zxn|Ba*D=jGxce=8v+8)6Xpvk}8Z^^=*ji@cSBggrT$QJV9#+2L$sqfcqO-NP`1teK zTfAcHutIbSaw}ZhtB1BTF0E1SuZQBbwjNpt>^hgM#hxLq0gXYwDbURe=*vdzmld}C zvPE`oBRZqN8Se_{MzNbp=;pI%CW;*sWpA_SZ5-LdEDG#V_BGSKFj4l`b7ev#J0mod zecNqWVaIOVdq!S3G$}O_7{)m)p5vzZ05j)@AKg3L^wRrec`LZk$Ex3UIxw7AoI!`nWt;X z&~F(suMC-I?ircqZ&vO_xOoLU2GOto^dNNWiS1??-Ind_{cX#g6Tx$$`uINbRRlRd0nu(O> z=PLglDP=#iTYb5FuXbI2hBYHx&TsCEK7||6;v{G>dTV~}B4|#!CRUoCt920d{N6|m zMR{Oddkei7qumN*M?Ntk%|*#{Y6&j3<5_vw429^evI|&yG>d#Lbu{FZ+e(I6d+~$X z{70%+S5l{z#~OX@9W!LW`8l`G{J1o}ekwFht+r@9f@}&@W}h^TPx&9F@juaCj6eA$ zs?82x>&q3+KOlEhl8Vl~*2W(;-bE?U>JIQ#B^`b>{xznP=NjKU z$klb>Im$5~y4ST1e(QQ%*PUmy21B&2>sqtkXD-NtKFWG#?>q#U2k#s_78-~*CilVp ziD=(}F(jH29-^Wr`Iq2jE$hId(2D$D;*HN#Sjpv;)IcTwwfmxqpIXZ*S>I7{5Ast- z@Kebe57|eNPwaX@+3TiV#)yA2rjpp8+Q}q_La?wXJP*8s^B;lp05*flgGvO&t2oAIfKx2Nz< zD(^H}scqD!cijhFms+W{CEQPf2UPpj%6c3hIzX~sb7tx)ge&@tEVB8tlyMLnn&I*x zV=9U+vhh~|A679AqjSJjKJ25c+_}tByZBYmBXbRHd+@AyZ7uNpfd0;*J=V%IJ z+ep2LnWKfy99?`~sF68hpRe*_-go(YJ9;uzfYfCj-ch zj%spOo*3m$uI(m9sW^EF=gQwIc4ofEVQAW}z5TB($izCXFS2EU>N<*vZCH~o;(OS^ ztv_UMk3b6Ri|9wT4S9s`P1v=*)YoCR&WvS)qn#zTsVm@3{y*2KMk;# &vo}c|@m3!?kTE>d_`@IWd*Sg$vtFf`ap<0W2QiZSpN^jSR~e?5!1K!92UeD=-eTb>!_+nkNm9*{3dSJd zhEf+3!*90Nai=!lp}xSo@sEHTzHQcANQYmVLY)lwpX;YDVk<4!V^QMrS*%r#tfCF# zKGGu=^OMA!G^oaaYH2)NQf1kEz>I+(LTI&PUm` zTZ%%$7lH?-WA!waTUi_r*`WNe)BA`%rzzWrN~{?Zkc`~70fT~zcP#3EzbM*H68t< zzVv-WJ*(O+VkCQ1w;=K7M%cY;Kl`gW}B zcG+3K3qY5^A7w1{(gWbj*kRlko{fByz5LyAp=jEjTVv2{96J8+lB0FUerW458-M47 zzC;~Ev5BSfoi5rP2ks{G*O~pzrh_?GH+4o*DYn<45cCPIj`dwv@dP zN?WYGe*7rWsAAm7A9ha$^p4IPMPGojgV19W*S^ifXx^~gfn z)LuE*UObx>Dxl`8q&JSRCUC*Hfo!imo?OQomzH0D{(si?I+1QLI`6wZF-rsgDZroa zh5u`5Ck=nY0PvqEW~n%nS$jhqN_9Dz4jp_4I*6c`4b96|Lv}@(Pa^|}7wEmk*Pz#K zBR^Jb%bF$3;|lUP?M_?L|0{t>zbE*~V;0$Ye)Xdq+0yImJjs$G?p+5>?q}{}&{iky zc5x03Mc?BHokc(72%SYgv7Ei`*_S7l8k)=Z{UXv6*b%B{DmJlStB()p(#ErskJRn# zwad2zuFE@V++IEaZvQ)W`H5&l@qNYhSMyB^ah>9XX3GuXZO}ZXQBg` zyOJ~~UXOp8g&!k-HOw5Eb=AlV>O+mZP@GEgqEB2NdC@-}pNH?4NBj?4JNE+S)y3)c z!f9_zV@+Vr-`P0LL?0l7oEZCn{Pqd88S-@$r;}cFbG^2Y>9g}it(J+c`+zA-o0{ho z#%J3{_C44Bm8|k!SB*0Ou$lP356&d_h< zl(|2i`|?-x?47=4S)5lK0H^FLA>VcbN6FuwxF7wLG5$VrKducF_v?v2;b%`mrhit< z!#lpg#(fL4Dp*_^O0ZT`G^Fn(xH+Mo{2qQv|NRVQdxgiso8YK~yrplvwKu@H*^6e` zf^O;ubv~W3gP*)g#?DdwkJCz4V`iJdLu)Q_{_<2|2vokMHZJPxFl-@dkcH61GZ!n4@A} ziaW-bmlDTzlrB>R!cB~?|w0I8fW23U)Q?BHW+Q@*e!u7#iwd2|? zg`yMi5dj~5a6xP&q;&;;#%|6lJ?G%T#Npf=twZzf{0=ndeXcpOqQb$|GvLa_&yDcD z7e87b@eA!2N3De5EdyWFF|S$jVS!oSUj;pwv8}gqOBnllzQ?}U)Ho+M&r7F%$b6q6 zm>XFi%>9Dsg#8Yscgv{9a%-4NWUIL~%=F=r16y3ZJKI@9DI4Q4R5p8D^Z6S3klNK4 zHOEVZryOFT)E8HLm%SSBt758S#U82Ww`g7bT#tS%;k<)3bl+Sf4p=2#XqcPJy_xt; zV{{&U-KNV+^`4bdo;6;3=<_T#Wyx6TSLi=TTsw>nSU4RUrw*V0VqjWp8UJf1w!tm< z(Z;7=OMcn%x5Z1LAKt45Uf%>qXWvI$lJ&^q%SA3<@{H!94_|UkI=;WWmyEoPxK1&( zl4@J^JUaTJmj=FdcmsWHJjc;7Z_}4@NKQX7o;yzVsSWz+pYxEcclX-=2l~+X^BH%L zF)RL4`c1|1X%GInl$iY7S@13~pk4g&9h$bZBdb<7vW8VLyw)P{eHHe@wb&B%*bonZ z7wp^;Vk}w0*;U{ZoE2bqs)nM5V|$&BDQyM(5fj-f`}1y&_yuhp4^X3Uq-J$TeULpF zQ)`uvgizMChzEev3%mBd2`?uffvpq06?n0CjXk>*A6UNfv_c~r$lttgRgPd!F2$Sd zy^1HZ{xE<(R-4-QW241h_a3~crjqyes6Hgxz!>N>w+tF6%BMCd zU1Q|^$jBdTeZ_gUgJ&c7DG|OgtT=(0d*)BLU&0)=*V=Tu68x&RP5ob`KHso%>(L<= zbkhzVc5n`#6xBmREwSz*(ScWYZ&815AQK!O5}%EM|6P6&e47V&h8W1$Fz+C@X2fYT zOx@d!8iymdW=!n0x5**DANYe~Z}aSl38_upYm~zv%a2_&(@6{BPpN*vPxK z3YKd0QnjrQ_EsAmkl9?|fo&wPMf`V|bs@mUch4Mq^(0{1@Ht?!@sin`t-hhVKK7Rj z|Nrgl%#rG8Tv)20p=#(y_{t-;(u#acfnNjA(F)|iO60(E$UeS>(pGw8~#BYfyf zCJrQey*+T!c>1X2s|pKooD;8(=11>p*xeQ{dk@{Qw_BrW!Q{* z*R>g2oiUU-He)e1x%KgGzt$5C=qF*4C_`~k^T{Pjib z|A2pzes=Kn%n8=ACdKH3tdBsZrsLO`>!(@E4vIdwUd8=%{2TXrI=^ijHm&)Z`L^u; zC-$G(cqDD+e=x7<===k{QkC9^6mkjzZN?= zbbm8=*$f|)gJ19>{XR+XZw5!@=wIQ<|NX5`Ek9O=F5TvZNi~Tv_{N{K;RAT5s|nnA z@27I#cV}>I_(Nalwb`r_Ha0%G6TE8d#>3rX3l$J2#(%Ut*7!KYZx;z<}DU`q;914OL^wC?3-!$lY6@V z0Dta&P4{uGZ^*4S*OHs>;kb%FjY+<()*eSo?REE~xAR^av^x~qEe+WS}s8J`^ceMb09+8Apk7E^0oQ28Yz>ndh;*JYvyE@$lFGih%uF*?uvLe|EV zVF%^I6Aj>3dOpT8!a^!6VB;$|uyD$7IjxJv5bDwc4hV@6* zqicL^{rolaJq6ZDdFzYAGqD$$H@8mmn~Zk@^?tFL@VP5Xz@hl70pH)tUy)35{_vRg z!&Pp$5;ZTn z$L4`K(6jdU>EQw3D?5RHnEjbN(6>HXuttDSF|4mQNe(&n+fMzoxz96-hZ&qZ*WaZN zm&Szq?>o4E#DjZ_e?9btozOo%aOhL~USeY=_G=|AixLy?#>H;bZ}f{BH)GvlpE%i# z{9dFOnT`k0AJF5XUlJ$MaU0*0<0u$={8Gj!TN}FWtqbPbz<#MSUMEH({9MaAQd6sV zDm-Cip1=7K_}us_(55$rVD6t9O#GDliYNY?#!9`SJJuxPTi&snn4o8@1H}Y2)*$gf zjn&eTv9@4KX{^S^`kTk<%y~~dsP9&D$tStA>AWkN^|lp;zchalWRKpFPJ0#F%0}klqqk`T*NgQXfr*vRq;W0GzDOzPmE&OR539?6J!A1VpX%W? zBhN?L^6qC(d*u1ZVB%-Yi#s3pViO!2c}DMi%;TM_ER!pj{B0LrnQ!GI?fo25n@icF zyfio^w0$P+YTl4j?N?#{T?zhxQ#OnHJFGcye25)ZQr9iuvJ@JKGS`n-No{fFyAyw4 z9&wZ1R$6W6Y-BC;Y3oS)Tj|o7+RNZ-)|#{=9o{ZESB!4edQ8b(@}Wg>=DCJzH)@^6 zZ1RdX6wku8X~s4T?I+j3^F{mbp~e!QL2$~SQJaF(x-XeKF$SakFFPOoNCj@Y5nLU z2JE+cc-JeRGy27EwZ=?wp9*Z8jYck050p&aoG}LYokUzan?H?xqvY}!?(=8v^IP#> z={i4aQ^xaW=irl@uQ~c@T^2sCV#3A5pyKdm701QA`*rRGwWfi40oF1_xmUn3U?shr zMLmzfV~>pLE1%u*OBcPdt(Pxe9Av!GC0>1`+6mR0Xgt&w7HMqB{AoNse!Jsw@#5|S zGai_+iD$%swS;k+adBMCyJvGR$;<&!{*2g`t{?Ws~>{mW0e8|-zKkbiq;LG>m zyFYwKj(}6)+ksDZNG0&y&AT6SuUN2iP59Sbs5VMGA(^cC(cEv^WfZ- zN7Chk;oArNUVMwEB`5XGyTS0jORy_G%5#cIy5sZW|5qmg_vIeAf9JvfL29$T@cQ9} z@%)J%s#LG((6`p0jpNUy?=zCfU7+U;PQkIkDMzi5-U$wtfZvt;?dNwszgL{a`au5l zTl2bNJiqxf_xWALTov=Tnz6LF^C-U2JaW97wyL=oc4_bo=N`wqY4=X<8T#b;;w&4U zDvqk7`QImjpJWeyy!oD&%zXCl&*fd0mIR)n=y`JNp(r{viq2icTDzEPnD{CRo>87eb1S)+G)4IJH3!|=ea8mPryb$l$OY$`YpeYJ zDO&~%me0#D@b~&V-Q7t||Ko7_ZCj=foX#~D&c_`%|I^_amtX#0fv;u|`0hm}dileZ zpZLw@dvUJZ7r#iBSkOd(zdw9>Vp+1(=~I5H>iDIn+x}4&q0r9x|9bS19y>{T_cuJwURSOwoE;M2Mlm)<9nWA(;olJH>{&1YXr?ZIhkE-wz9dHdqS#s19S2Nj>5loE=h zxofd)dNtRl4|4s^vqKyCri*Q>dY->I*VLmQcm1A0u0Jx!^_D@dKR?vejvv?lqvwT` zYdh|Ge30jphYj4$_(87E8RYtPmxM|N>i3>Ou5TOUdU%lQ!}11hf7bEWHU1Z?j@2zz z3<-Ztxm@W)VoO^B=`%uCOs9rc`&stWDPtYlUgo7AL%-+r@Qb0J%dJBX!$;3jmr*iu z4mm1cSCadC=$u3VUrTz|^igfo=5^3;rLzu(_~f2I@WN0zIpR#m2NCUl@(aiQ_Vj}@k-gxi3fzn%pX}m> z^-OCT|1=`hK}=eB+Q{z!-w>Jzo<`Kjm$I5&*e?O*$&AGd_e&1k`u5Io`lcWLjJ%;v z2b*l0WQ{u>FFxhNW}E<@=>y<%EU-PuUdmqB9(G{+6R;V1)!@i0;y$^(_fE#bo(1bKK73vD;^rdr|!tJu$kHSzNO}&#o#Ir`=AY6 zf}2B=Io5NOEz!;KAh?pfXmluX>ke$gD7GQ_)OCt8_n~#>xBqX=?-teUa8Kt?c+Q(S zcjw;FE#JcX)sK7UzbCFje{pQrcHr=S*H?6%@6t7PNfdik{&XDMwjJ0yu&+DupAHa9 zebz%iibtDxyYqa7uyW$MdB;pR}R=@>VNrM$QF0AH8DM&PSi{4So4HzF{v9pH=l} zgLTG?oSNX_%lW&H>%(fS!wuG`8B^y3fB9AZe&ZW{*@!vTFB`1%8Rw4=-u^ZIMkS5B zY}9z`cH;FjE+p@C6@SZxmaCXae}1^p?AOt2e?{kE!xu8=3z*;Lju}i5lQLJBeIm(x1aZR-+b`C{#w$!U5Yid}paQ#N&;6?_j zj-Hqm?WwIeF$V3Y5hLCtUtk^lr1+8IeKF=EnRewvxOKbkzTXj^H~V!6&O-L`5S-Ag z=tlb7rj@MF82SxE$CYL;qM4!R&mo_iM?RNaEjigOZaz0c8|v$tXXu~v0D9g>UMrHr zv*dCW)9|sse*~C(V|w4k=efwhJI&sA<#N3b{CYpiHLZ2<4dXrN%6As$Sym$H4A)+; zWuf?`KmCecf21{L{pe5h`hbT%UHjrCQ{UL@553awUBp4g;;9F=3Bp4K7P>g74-RBe#5i{eh+EyITuYeH{;d77<#65clHYeddrvZnLD72N_xH!i=d-i+er7+n^{i*D^{lm?XTJ%rftfM}n%B)c ziY~y3>NDdd{uZH;o^zS;<(+!)AZJ(>5(6gr_u&uMw%hv$ZJF1qxi)67Hz^}ct2|P) z|E`Sn%b5}mH+dF||PHe<-Mp}+Q<*WwR9v1DtE zu^~zOt=JEI*mOrJ9vaaWpnt83YoxX#8!xAAay?Vp?LEb595OvA`hoR0khu!tUD$Ej0*lfR4n@YQ=@cGtaEJ zG|_7sId89tHa1gF3-$3Wl&!6^mZ?*3l-RA(a>XqZL}>J8*4<$ep9ulBd&8i3`Y7aRGYGG3Kp@_<&>3$Tr@)gZHQY>J!sIKRXGqE?2)ETsZ8pfw(t7NIMN<|xm#c8%?m=ZOz8Fz*;~^G zPYB7d+k4S(7k}ur-dDC=a55hrkOs|Z;7IhH0`MZ^`VS$T*lU^|{cumI{M*m_o?p_& z#eK9<=B&)8uJ(MB{A%Af%daL$tYU~?X? zD*P&DSl|3Ant9fIneeMaPZVv9Hd=;S{3?#~BzM8{C1$n&8j+kPQF7ce5+|OyaXN@oQF)B!`aQ) zIpeT%zQwz{xtg`enoW9S%`sp@Q_w`LMpr2`^m za0Z2Zqet4po*LDb3iz|C7aR-EkER|M^>8LM{=-OQt;-wY%`K$0!o!vWtX5Mc>9yEW_xmEai3*+3H%G!l7ZzJC^$~aCLE$CIPmQMAEoJC8_*R_sUtn-co zyqAi8BjzsjR)6 zb2H+*&c}H#ZToO5ew1~>ld=bS`(?yAiEe1GF)J)RJCwhg{B`!;No)-7rNyUQQvPxm z-&PQMhYz~ZQBvG|7nr%SU1&q=SNR;}sp{ubT6k@ zV@|Rzc~R4@uKbm>iG23DMD533z}ljLJhCqN2VytmUY#+=?b?>CboJ-gV$d#GsCg@{PaOmlJOH8|S{zZv;Sj_3YQa<+_)SW`RZliV%M z5$;Euv?%YP3uU#_unje{E;z{VRoFBSa(=-s-dp!c-WK}yD6xFr`+08{@2umUb^I=A z2efGKHp`pF+;UA->i(3bh|rOz=Y+?KZ&P%M(E3%afl~Wh`%Y~<$r?uXMcgCtIre>tx|eTlmA+KWec$UOOD>16 zlZ3wFX2{s^flKn$bOV)=roS}iR4cv4#U>nM1Rwe zd;7xHL!vi^KMQ|qcj{_=v|)(wgaKVPONqm6Sb;37fQN5pFF9rH)%vxqMP~@07aZFP zKi^{g*6PvBmi5-VO=ZaXRaV~Ma?Pw?3Jp`14!w)dKYWi~-Kw&e(n>W$VtZxXz;~-| zlW&IHPdnwiwKo&%*97kgES0$yEC|3RpPU( zvG11q+qZnWbsOpO9nQyl|4xX}FZjJjo_&)mw~!`y>OJ53vQxHZLjU8L2NRhO$VQRB z$$iYpzU-E*nc!?Zy46Hm9L38+Q@vCa$2|0Ugl~23eBve>^6fr3T=r$ zH#baf)-i4)cn&gGoi3-*u-S^s7|R~@a?PRQv@2M9)X!m_GIv7s{HLMNb7A-M`|>=W zx(cB?=sF0^H9#XWx33rahK?k@D_wJVqz>65a3A+luEfhXLW?_z6$>&So5;iY80J3O zE4GQc`Q(Lel#PP&MW1C4kjsA<1#rJb^mP@t^m0_bb3jl3Rs&XqcQFQ|VT>0+l6-F^mXW$-28L-06d`w=^p z*np1GuNnQAgE?CJGuV}mG1uf<$UA8BvAKd%>r7?lM%{sOSnVZz;dY?;`hq+WWootjc?NuSZ_^(NfZ+ zp0CS~PRhQTX!;vRe-r6%GW0u`@w^;zWEp@$Jir@m?QaR?GOF zDl%>|qn9jR*?hz7+UBIZBQA78Z+vNO9P4`*=V09x$^1-Wzr7xf&K7%Q*C#jDeuUp< z96q`A@$N@e8QX3$mn6NhB?TMASYo!VGqPOXSa2PVL*dg^VR$)c>r7q!Rq<%u_Ia_% zD}M$2K2GQ#{5|EC^XejMc70NYURnC6oPEyz9#1;yVxt$EdIdh1vSEDh5?xhrac8Rb z{#0Ny`hKGpd1Mqg`H0_JSsyCfQL<)=Pfp@{SYJ_oq+R~%M{g=xbV1e&>>oOUpUV8D z*y$O|Y-Hwr#N*1kL*dqG`zfHsfW(_{zFi&m&jVfePVej7hvTC@5xz*>stw@gjxgMW z)6AZ5-4gXOgs&y&2fgFb>n2j>6#9c)Y(Oq7I#&9qtZjnmvrp4@i7#$~_T+nT8u~$J znX`)XrO?0d&FE3=sfwcS4{T&j@@Uf}{8`|()7B5_UtITu#8ZoIA^O38F~0IFx{}Jv8FG=e zW53G8A;>}GkdlLonUBbi2H~&pB#}cBhuKVj8=<3S#=jZ*YKFeVe&xu~-bcqn4wP+i z>7|dhf~(Pbv}dHe3qFI$2Ki1}9na!tYXtXA;J$&e5E_>85Z{u}OffVm-#IdvgUa3m zUA97Zu7TuNYamOWb;(lrsH}mT$Qe^*i!9J|$H7`vT z*#YhX>?ugSLCYEd?qsirjH|@t)djW6%VlmcubROBuK1`UGUt>nk>569D{oA(><&$o za{&G$?+RZz5aP{wUp!CdU8K*4ZT(gDf5o6be3sbluWp`48=%9Zi#4+a9k7WxA>Vjx zV9o@Q+4bmLb<7EgC#)+JT}t-%xs>c~rp=13Bz+ci#T=3OQYU>8+G75weu9Ta`XqBm zJwx-MeEH%dmEq#?)jm&ocDY zj~=Dm>AZVaQN{(aQU8_k5&yT?S;e0gqNiV+x9O?c#-ZZ9vEDl+Mu_``z>D$SaN24a zgMQ9Dm9>S)s?VPJayko_;lJu7&w9txg~wN?p>JI*zIxxe?znH+-=X5Y`MyUGnppsJ zj41y@cKgRYY?UEBMcVydSiAp|Hig?B5Fgs*?6WSYSd;}1pQP@c3$hC2Y!c{O^a{}@ zNq?_k+m%`B_m>xS{qE#=brRS=WMV;fabr z4%Q_0;1NyBtlTvSyUbwh4$O->&UTQvf_mh7E;@+RTZrwukT}Ch#HoyMLTi^3D*zqU z!}A=Qq&(mwq!pmEz+1&;S|81L#zFUql$p%;#rQr0Hl{~3cT%P7mkQ}s;c{H<`=S2- zq@K^w4me&BhGUV{)!{OF6KkY!`_Cgip9Med71nyQDn;h(x4PGxt7zwXWZD+!a2xAQ z$`bp}4%V2*SFL(U)|t=XqYA8G?YPr=r`2G7Z?*Cq+f839z(W~*rL4}f7|um>Rqchx z?LgmYgy(L;{+JU4*nM)2GOp6jVgV&q-mJ5cO4g+^u1d4P8U{1)K7fQNM6 z#q(+YOwQE3#5TydZqv}=uw99b>%LENYte__m$9y+eC(~25_exmJu>Etqs6`wT_xp* zx3`{t)$_ib1*7ya`X@SB5MC0bO@Fc5q%($f@X&hjK^%c6IG4VHr~9#?2v2PSU*Z>% zwPujDV>5V61V5}R*yo%|ypOeKOYBSkJ`8y>pZX{(I2U|Uj@om^JI&|-(ncAhY52X~ zWIZl+u?^V8S|#oSf7n>ok#}g=#uy)2hYA#Xn2c%29(Jp358F$+tV3H_YliG$_u*GQ zieAym+KIU{KwxVjeV7LgE!ND}iN6q>3Xa8=Dc@df1h1{&DF{wmz|T=|Q-3*mpbd#3 zYXZ+r=|J$-GzPq9f}in>??lbKhVhraevSrsS7@M~^6J39?3=`)_E>;->%dEw4Ga7* z=RNiFdfBiFZ5tN2EVFe~Wy7M}a2wV%Xj}3s``aAbhE-v$InNT~VvUiqUp*pzt5fP6 zRM~^=@-|8g3_7ih-%;+Zag6eQq|&VzytjInfsN}U%IHxpYswEv`y4N!@1p--Ob&(H zxPG|%tI4B~jjO@_4wdkJ`3vVi8%AFm|5-Z{dlz#+*}Kv$dsimkKRy9ZdR|Z9yorRi zBJ5EKdcPXEcaE@l)QMTx!*+M`DdB6vw|elXP8WL`qN>Idi*r(U)$|kSOsskL)6E;7 z;aO}!$F)SyacETdSe@{(bc>Irav#pe-j?`9=FjK&81&Sej~$18PRwHd2|t_fU=QAx z@H6TZe)dP9(GWk|2@NwZ0{kYl9L~>fmiNA5TS9!Tit|1C6uw`E4lH~uPy7P#D>-}AqUF>|;a{6BXC5$*l$<~h zQ~nal5#6~6-@qd5lX5mkF6&6KB^JP|#1169O5jJ>wuM(UV;^$Cvk$X=lDe}w(^zF}$O@T)NrbHLcNz^{a_%a}BR zrzXZtbcIF(UIxDs`Vk$y5qvbluYxweYM#jXhJ1HQ@FM4*h)feZqmn7SE4UYVrQ`)= zzAfLQqJEJNvbK5<-YRDdWy8OBfJ2F|5PeE?jO%E>+=~rw6}Fc`Bi9tTa9bZXJexmV z>U(llg(h=>c9&zvleJ_3zPJP4bg5;8_@nU39`-rAElu?Q5b^TQaET9(b0;fv@h1wO z7W>>y*ypg>Df`^^()qQ2MJE#5+$Q33pr1!1Hejus_W_UH`jGHdhxc)Q6W@hLVt2re zw0otxPvG9=eW-U{>~#;4w?}#K)(X=89^UGF`s?x5=fdoDOVWgwcE_U!ppT%)#PR4i zH=wJa>q#5KFXUAFVSEYL|6e^I-wwbo{v>;Wvl&l`7l@2-zj}zPnRKyTWn=H#fc{53 z(yMQBi5(ys+njtuVF~=$g^eHsUFab2AbAclPGe(aU6n#SV+wJNDc)mT$DM0l8Uh{m zU%7lwZRzc`%dc2dyL`#=J@+{JADL4+TRm^#d4ePP$jhbE)N`%3Y(U_b6MQ-?N`*Q|2k|e;uP$ z%J11)|F#(X21zUL+0#y06L|NYCHL&P#*uX7eae#ON}iLz*>@>Rp5M2he@$8K+`rsU ztCZifwdA&E&UD^(7dRiW`u5tJQivfgPpW*f=#I&`!-{He{$I0epDwy{^1)%tYG2nJ z-aEfD&3;~~o*x+QepG&+O5WW?izhder-(ey6ctT&4PREveDpp@9_x8!Ei~aBOCEVH zA$w>BD#j-lTFlmB z+K7)MZONW0@GSHxX<{ExY2ZtEi=>HtR?>(aII@lT6T`C`n%>V<&-E0S%q8T@s{$89 zSnv*)1;uxy)^rt`btZy3$0@>FATdPH>p$D@M>hPSg8jU=E12y4fr9l zc%k+8hA-hqF{yp?qt=V?qr(^BM^zW$M;pE(KN@-wesr3-Cj9AQ{AkWa_>u6UM)9%1 zi!7eRcTS0A`2wDFDx4=Bphw~&a?O00MSK&#sspE@y(y6b&lYUKE zU&NE%?}H~j_&J`$Ic8sxC#?_TNn?BQq-ni)QqkYRlcrOT#dnHoi>b#~R6MzO__Er& z^jNQldaUP_>N!L4q$Rz0Qtbb4dD0Y#AA~1m*>I8#zpY?D??eTYy%QAd?;WpTlJ_e6 z_e_B{Pa3PzES@w<-CI0qgu1tQQi|N$JZTvBUxg>_VIT09+26%3`Zt>=h4|5773*}jVLkB7nKiZ9u1IwrX@)DZW;)kV_B6PN1XZ75s89ME zYHz5{?XH{}>2E7>(#7=GMNFTvEwS%o7Vi^d^&8Q<<=m!lS*c-fm3$AehPhAEYGi*| zRIhgqtQ@;XVw|#*h*t#fqgdOWguq zGr;Ru!4*7~cAuJ#J_zpi_td`EHoj_1ljr<2iIfU+A}C?nguU zEZ&?}S;KFO$S3~eP@SQ=LVP_`PAJb<*2A6H*<5^!(29qM_Rf*`B5XhsLohq+{#NcK zZH9uSFWmxccp4ivX*Y+xmnZjXe#Us+D&6J%0r0ELyAS!E`K8Rek>R#z+uo!6L(p3I z{`wqZNSFM}XIr1_b)Mw{IoI#rk(DP#_PzIdkKGRWZWhIN`J*!!+wk)P#J?eH;_c8- zc7fJjAg~-eH1IV0T@}6IKNefL_}hMeui#R}Cv&!ade8Xe(0j5TXI-!KCiJ7nd*S3i zX;X;bpmWlm?(U58ZeUOF8T2?q(~>?t`Hxkf7HaQb&A7;Sn=W4# zU0LR&zA3~pEyM4^KGCVHaXay0$Qf7xYzD*?RYtw%tduxXiBD?8XV3zj1SS!efd6Z| zob|E~TTQ0h?BQpquzXc@-*rVvd?Yf4fylXN9PmsB(|5M&~4^-w8}C)qRqj0TKno^mF-3D z*;kR3-LAF2$-BZM@RbwK6lKS6N8mrS&aaerrC(X>Yq0oAsJ~YFdiq;n@3F@w@Lrnx z`pQQ)^h(pNuY9O?+CJ9*q{PklmO<(w%)Cc}@ zd*cfHeW!OC_?vu*`1`%=3y#JH76M{42H=(&Fykr5qBNx8qf}aadk!SH=iS8i0 zg*D)9^>Jk{Woy_~pmnkNDW4Gfi0}>eCo3DC4(`Q%nnhfiw6TCTmUF(8+%MdDch*Sm z)!cPs^P|5#=C0GCW%awp+`h=~!qYY^@8W5Xe2%Bt<@^yop~l#L4%xweM{LaJ=p4Gf zz1O&Gx9M1TvHV?(KeQ_T(DVC7^^B41a}-}qBeqYe?_T0F0u%ASfNu{t6#uL64&obc z6aT9Cq33H(#Y5!#9XHuLWG`vLLnI#be$w`0=goyrhW5aU{Z_>XrC=|GkE9LDtwn!$ zUu>+xOT>;Y{-t`xWmPZv(3@ARfAI6X;=bN#(E4i9K8Hif5BIqUuWVG8Aq>YM zetW6-O#Q!sPv|=wpL4zhpA&oI4&HHf?=Jj&OXo!efrYW7^;`$KWyL z_qbhYei%8-8Rf4t-!u7x8Y4`AI-dl>(&Jy&D?IM?XzLP zhT1JwIjJ^u+pxff?_S*O5Htf+}J8hp0 z12)wD$tow+hHe`c*s$D&+ikcH*t@+g;I|e09tFQ`;CCM9Sg&#psVs5d^3#V6 z$=5h@1Ks=|iXvUtx#E&0jU4FC7(F~E({^jld4_$^WxZQn2M8Y~vo4!knn0`=;oZhTOO%LUlZ&bG>YQLGj-|@bi zF^V3xJ%9AdvH3S~E##84IN&(0TeylOZGn4c0`MxXnZusUzk9*9nMG;aW-d(AX5I}i z8PU(3F>+uI?aa-{?YDNU)aB~0y)W$zZ)>jI)=(RBX=9=0n9etAdbBlEhMbS(iKVSM z&P#7A?N6kgS99IUbvKtAIxe_p+swOZbEw_2pH$&L&%%EM_#X-WQzP9OBl_jgpIrKr zm*I|FJ677`8m_$`!eMxyh-p;#mpZ@Z*tFUJe%JBzKau_?U2#oG zU%JM$W$u}2T!p7r*RJC#=lueq*K%iNJZHsd!^W1xxwVpF=IDmHK*ED{fqZXWTf+3= zV@swF&nTIZlT;F$8(ZQUKCUEVcxFjl-N^)3j*DmOUeYphVtCFdamA$OyN0KgOk37H z*QLGT?aHI=g|vMs?FiETrFPphIp;>^&=>I$;a~FJm%VeJ&Y3s#t@nGc%b7QQq-PdD zXW{&7*i|K(D>dJFgDby~IM%yq>)o_-)9~^6{UUOfJ_k!>=rPo^tx-_5=4Mn4?H3>YW*gUXt~1{+DW+~oc-j#gC#Mi(F-j{QS6fJ?Xi_w$(p2qK)Zk_(6l{{Qmmyki9fzZs=CC@9G zySikAT~;pd&Y-`=!$E0?~df% z2K(KH9`8z-yLiv%j^w>jyw_;I*VyB|(RP^!dAG(L#k-?&6SLL;HiA*sbus6yYiuzk}(fF zoewS0p56J3bxFZ=OHWz=OoMOT0n7y!0t6K;d_5fs+3| zpyaOrO8#{~$$vjk@;?BS{Of^|e>+g}?*K~vCxMcGCs6V~1(f`|fRg`dpyb~Tl>GaE zl7Bx?^1lw0{MA6o{{~R<9{@`JgFwlD2q^gjK*?VRl>GHT$sYtt{sy4rZv;yICZObR z21&Y&JPDLC zb^@i0r+`w%E})e0G*HUe4U{tW0VV%_pyYoYDEX^_lK%~$eN5|QL>`CQZJbAtk*|HP4F~WXchg{eN z&mU<&KLAhP4Zj}Ev+(dQ;-&KYrN}pV@0*luQrboxrS?L>wNU0yWr&&+;1E9RQ^ZEzZuwR4kG`Gk$=U*vrF8zoG5^| z2o09oaJvon*)U*3P3VdDQf=tAVSx?Hfzy#I^3HZUEp^)1d@t>gG&lDG3j_|!EJ&Ri*``5&oYQ^Z1@1Q?POwuO&Zt_iv2WUs}rui3&k);oF9pqfc;^ zK*|3<`UQH@|3UpC9v$K;u8Zjx1<0RG=oinSUtIYO>I+hCJZt`|IQKWa+>p)YVQeh*kbb${+RLSpI0rAM@W*etlg3YM|vP%}R=GUOja2@Aq z^qaUFT=|)tSuOev`w&Y+$BDyEo5`8gqT`5uqoqo^i*!dqCg)a*jwAYwmXuWTG&)Y) z4SGW60By49IHKR|Lbnjx&o_z3a6LtAq4F2Ohi`u}yx8+-}nWvFaum-(wnAed4N@megN- zUv0-PyZg`_)(1IHBi!d2>Q9xzPxto#t z7aqR9UiQw5Z~5cX#LJXv>|u+js^k7P=0P3)^Z?^5zju~7sy=?%dPmNOsAGQ2;GKGG zhuQeZcR-7BzCdWtO)U0lp-t8HYKtyL@onNLu3cKx5$u$ea+aU?;Adj5TuV%n_|c`V zRQBEc7i|v3mrFlqhKyKSlJ72r)^+cN^+oz5 z{R!>=ZpC-ed;j-*>Ro^hT=saYJ=~pf^f8``z1*F#T+v*vDycJAR`5~~UgFfs9}Pl} zfgtfHh3r!d>XrWt4tmg@{Kh!4{@g12_+$<-sW87U_@9-g{yIA4$90 zsm}Z)KI_Fa&I?ueC6B^^y(dhaK}36xGFKN8H!k!naS@T+?}g8fmG}>E7^>sBuTe)h zUCR8D{fXNZ|CTezIS+z1&ytvAdu(N{j4DcXF#o_G$?eK-!}ObWr<5s4vz_TB`B z(?f`t3w%YMyH_@tY1b13(xr4FO8 z^mXHA>b{YB<-CR)dA9*RnoHh=t0Rp?s}C50+wHpMOVcj%?{-A{+0*s}vB91;N4)PL zN232E&*L45o=tj!Z=XZ+FVJ=WZav!HqPx_)@y?;XLM_?Po-qG$-mTJOJje0rNSSMS zw@{1rPvzaQdYqqqb59&~YBe*~YMwS{T+K1?_t5GiMgsVI7kCQ%_5UQvNM3WqSnkk# zZRC;qkF$qa(_($84&A@N;q(_$|8B}`;kU80Rbog+>ruW|e4a<~E3tRYe-!}u0-0?f6ZZI0`2iq zuJqBz8Fa)5_yY7VjrQc}QT{tAr+~E*kBeM9c`Xia2%)2gJA1 zMo+n>`EqE-lbX}-q>Pc2LG1n$O|;`sx#nr+jP3_lA2beg#nO(rH3yAi+CGN&1eelx zO^fomX~zQH;jdTaMEE(=?1`te2u~xvu}I>_M{7wPQ?-PSA}zi{_AU~?(7`(&Qvb$X z+Q^L>@=wm-=|l#UYx>6hDoyvU)f^k^x#xST-upD?MyEbT{mBcNH;b4zcYfkD4wXkJT(v+?F@h`TX$J5Z^z_iD5yr#Zk7v%EVm_W?9<(zz zPp@_urIoE0C_#He89;zLYzD7rqpKs#$Pca^|wIt6wdW3H=V<+^;eh1Gw z#%?Dx_=Xr`qB4an7%wvGOm1R-U%$P0m!z zTe72J*c)L~@8Pp$<2SJ}7^;k{UQ_#&V&kyF?nJn&AR=!tpIUlA8REI)XD z;$}U{Q=vP3uCTN==GEQIEBP+P^XMHXxt4trV@z0GW61mCnA^wncqJcYjuvWh%x#zd zM(R7L#j849q{Cl)QV+2Vs;=;IpQFqZTuUkU8p?GclVU0NEz0fY8F9WKW!}hqiy~e6 z;)0hD@8PSU%s04%haBY_<}=pniVrWzcR<3v@Y_{X(2u-63>&K#T0iwZ~Ox%LQ%eI=ZS&hdIz@I%x|u-5a7w z=pCAj;@cJ@z*d)=Mb`^Wz~G-ly1bX-fy%)Lw9rk zWA0B>6uk5<_wRAPA3bwId7`J?*}tZiJOpGt`uLjt#<@@9j1~NT0lFMg;qd$gd1E5C zE<>Ja$Sm0xa4+{i<$fP9fDC(G(xE4Orykiyv=7=9Jyb&;%DD;}^cw)51@a7@3gj6a z?UQHZbhSJ~{~B^OFhlaA$4-@e5k>%=)P;`Pf=*h*a{>1O^wNFYzrnrC^>}ni^g{nl z@O+{7spv>Y9f`i@kc~oT@94>%I_P7+Bi?@t?}>Z~y?2OrbF~4!an8&9$B+k`^dX9d zP9P_qgT7B7qxO?GSF_4*hW0HzQ?Ho`J<9xf2U<)7f2W|u37^Cn%YmY&{@9`WXj_+j z5}HhNIQ&mi|NXXn9Yg)iT7pM(OOc-{_BTw&Jf?Kact1W@rDJXq9TRWZZ?Yyh$~qvP zcKjHKeD`2$^D#zh45ld^GtyTsx~?rZv8_D89E0a@jx+N$0v93L(>(J?D}>X@=t{M%#Q`ekC+8Zn&T3*Q!bx`?ZeIXeouwoXe^ za;<^iMDE?l{9eS><2ReP+{nCN#MR?B=8?$aLaqjQPCEQ5hce-xUiR61sSH&5uJ9A; z6n-N7Dgs{R6h1XXd-V(D*WKvkHeTs}*@Xa-lnG-LRFlB9BKExU)w8 z$eneN*w8VzyR({y&8{7Ln>))j{Px<6+3u{2;RUthX1cSc4=<|CoaWAwJ>pm1=+4^2 zc@RAo-{6ch zg-7vwzLk`Q%t_PYJUOH>-?&^{(VqR3E3!bHQHESVxA4fiMCubAMqpq1ijoi2)_H`j zoZ>qvh&^o~5sJJ|Am&pUFiRvl?Reh)BjW#1S2 zS7kl(%qpMXd3UG$&fH4j9Gp08j9tHxSIY89Sv#@qP~IfR9Q6KOx`-45hnSG#$?M1Dtp z^z0M(vCH~Ns4V(e*t;xbXn0xk0K5wJiM%L%kQ$b*w=Si-w7*=Gsk<<0D3@pStI0k@z2*Gjz#Mo# zcJ6%fDIin3d;nFia^ioIi0UFY^6=>&UL!=Flo-Z*ASnnmM0(kteqWUVy*z+}YV5dixevGFJju zB3BYu$oJ6C@;9LG2j{akWnB;~1j3Vpi=oTH2>3X>Ht48)5WI*_Mrc#U-1<$cl=J6< zFAHz)@;h)gdrIXqTDRXJN6sGjf4lq+k9^MWkQb@W3s8Os`VfnsB;?mf(jAraBCY2h{s$Wu%Ktz+yoR*H;e7_EcpK}W zuCo%7Wvs67-V#PLA^*d@VYJi5pDLoP?^9(%KNYSX^b;7Yt8XQQd=R6PICqin!HBLm z4V^EBwO%Z1!s1ew(MG*@m%eEnLx;N+o$z_mU*Pu_#jc4g_%-XacKG;~(zlJ%CpGi2 z(tgI7)0#Puwb|8T8PRhTqMmJ2>lO1cEM#n^GZVhWZWcTSc5!gj_;2Xy2 zXn_Ze(;rBBq;dMMnmGqO_FnS-guMToyl0V(1(FxJ_A~N!d@S${qvO10`b%q#jza>I zjgA^1K8KEKfd`C^y|QNI`89z&zoMBPYod$}=1>QA?hc=3{t#Ta%XH(rW&4fqmr)n` z!S~Cmjk99}(kG`5e8V`~DbJC{*;;`IjI-}*=ATMGG|ui6c$sl_w`Trt%09bI;6dZ8 zp_#v3&GS})JZ}*k@w{0e&mOT8xy{ZQn)z+)x@SMq%rw^7&)zrClf*hQY0X>4sS3_s zWo_A3#yh0FxF+5>wUqMs?d%-Q6km{}$#XIOqPAJQ&pf|-^;^aa?8a?z*w9#q&%i&^ z7Gzz)ntBF)nYNcT^Cs40Gw}1=jL)ZOJ!c-44mO%rY3BLT4~+9G7{9V;#cy9F}=>eaKFzEhV;nL5?W)XJ|0YL*D)(fvtf3QqClj&7qo)2B-M^cem5IsJJBI@?5FHt7R=?=o%^u)Y70c3}Hg zWA%qZU+1uV35RUo6!sU-(bvRp73(|t3VX<_#fOG{V=w)xvHNw1 ze%0FjdY69v)ot~Q_O*W?6byjcH8~hNxxpO`$ZimeX3uclk43k z`0}M;1ChQgq^|q1$2?74hZrB}ODy&j`8N1W`f@9M$)_)~=*xHM%e`HFN%URCn7o4> zbTgNSOU8c__(F6kKi&??ucrJ4%3nnJ56JlQj`Ss0OZK(S6T8l3{-g8h&tkhjh4g2+ z-JfFmv&!xd`A)2`?PPP^=05tsy462|I`XJPd_dv&ilGmjWnPm>AFiYiSJ8*7>BD&X z@ZGgBo(#q#13$3Xou9)_Rl;=_m-uLl^+fv6-J&drBVn;+L$(C$W%wd;(tl9UJfGUVIWtep5zgo~7?gtWjsSbuZ@_bxyJF)i+2c zS@#XVzzp=!!8w+XW~yy#%Cl@ufovdl;=mLj-xHQIic`yN19H!_nodJE} zJCAhhn4&M3pIY%y?G=`OS@mtv4|?jCfmg68BkL`n=j(>}dc+5@7#UI|K8W&Yw%3cl z^RRJ@vAc@<=U2y|7aX?y>1$%}ryn*RWBks->#Df_o%lKjYvwD+wDZZDxd*xPDl&&d z_L0?2!_Du3EfPzB-58lz!1K4!8D2yOIF+uM&!H0>Pti>88i#Ror)G{rC%6^eqh-l&D?=*BXqJ8-9}=9godU$UA}=@ zls^f+`HDW+5C2ho^(WdOPX>Id#O71?FhB0*I<4kKGWO#E$XS>FO?@D;HpzdrV=!|> z_ZRDfl-)SW8D+`D!In&HPj;K9FL#^QI+Fd2nR%Uap#MDc{X#lxP-L{orSmCnbBJ@W zXR0$AxrO}~{=I?vk&VP$#QEc#mQDHu^dYwIKjP;;=S--X%>9R}hZrBM{tqP!k1)Rv zFu(tFUxMd2^zqK>w+-Z(Zx`_(Cme(PTe*JD)j^$)IWD6w1AX|P{1Vq7GNX?A@J;(4 z#Ahsga0Buq!V%$4$oHLY^XJI-^W>YW4OaeH z<|}@JXccFYPPqs1X*DBn#qZS3pWej}HpIA&K1`zh*VF!7#h>1WKMnSM*V*6yXXNDy zn?Fh0i}lMW-vNJ$@Q*-F2%dtJFMh2>)F=E&{8Y`zg2j$#A9?}t8u${B8;9xVQTS>b za$p8L``9ex1iZHr-ATsy9DJhIk?c7S-wil4Up@AAo$*b#WyL7!yAwGpeB1?(m2XhQ zF&_QtLo$6BOdl?z57(`Ad0gP%W%pq@{a!)c$S;L|s}F;G8vGU6>c4~b2cgB0y?AVM zA$|d$@(cJ{pc@zXz6BlcC^AC)3-ayw6HDD@3GJ8Ig!UCm|MHxUbDJ76rVbh5pblxj z_!AFMepIja$I|`;+Mh)GlWG5e9_>$7GCPU(&!O$NaEZ*Gj?C7O+1US;58grhB^E&1 ze-q;nk`c|>v>#f>uiHzj%oTZuFDTEFalPBW zi1Htx{BZnB`{QVTJnc`Q{YkWcLKprMd~wk8Cfc2Ew_otYIo3+XN&6RoAMwkGtnS_Z zXwGW3WZV#BTp}_q5gC_waTy2S6F=eKE#oLd{42t%l)YuVbzcvZ*d9p_0z+{-4Pz|% zRrd0Ft}fSFur$Qt=Ru*BU+o6(26;umgaJk~P@MURm<2GQS%)o2r6{04l+L_!d5?<6N0s_|%AJNa%`Z5I-{hD{qiGB$h*b$0Jpm70+-J_o6!-&>R~d;GS4v zZxi>sx#t_J_zfHz@8Di^5QSID2>JaRlDSW`+AiPk3HhCy;<%5t(&d~kIqOXMDhAvB z=VZ(Oto+gF7RnzD-s_@C7rY0Fjd1pi3HhoxMK@{-#Y>FVV>=e>@m+Bf5-)KJaTC)o zls)PsZb0HD(xGpOn-Kq__?+^Pe@@yKgvVjG6+ehy+7|Ldg^#zyPvn(L?1Zx>lJO9| zWel(d8Q;qNQO0Zr@_#0}!$bI{PcW9^YaWk|a}7Gr1ZZNSO%EYjkT{12k);{5tA?14 zH@L+2DzO8cO{C&0IETb9u@o6x#9*j63yFV__kwnuMepAvu5T%AkoDqObdz(eo6oau zzJT6hqQgAFTJhxScUW8D-_?is#ygSytPx+~`i;F-{4Hz6LhUkQx-5Bo7#`99e|rGl z>0_-Z^*f2L2(>K~mmzHo)qjfm&rtta>OV*Q=c)e!^$+b)|6pGtI>DcaFZ(&yBU~aw zA9D<*{-Isk(xd)CzLToo zN&atBH)r@OJ!k^!^FnQis=tBpk$yje43PRW?fMI;Kh*CK9(wm*>OW2WXQ=-y^`E2u z!9D69$oe+fzlD0Qwd)rf!34)Z>K{z~vVI~a6S?tVulmE|R-(ZBSgs{p5|eio>(P_M zvAszBzo!0|ssA_B|6A%mx3-_>ImVwgo99{9*AlyM4{PjXu0$@8Z%M4-WlbvUYiYws z`Y&w|-Ar&GbzG?>wA}@NlGv8FiET;HBRfWM-Jx5tOU$Q@9>rJUymdBjk{FjfmFDyc zpX!dgirV-z>FeN4H*2nqhqxEsG*ydHu`lQ$R_sd{Zwke43D1(4;J$d1I^*#h;7wij z&pd~+e=fO3?B@|Z?Vp>in3?(tcnxztI2Zpk^IrD!jAHILj@Qi5$bu$#s*0yUeofW( z{ha(?acviUECY@+Y2O(1qAsm6-Ue$Nma(C^kH5u5#99)4E8ob`jGFBQnT0h?a1n0{e@6}x^q_E)j-E&CI}7**Zxx5A#kl_QG9~K9Thq z=`z09OnlSfMPg?VeN=Slc*juRY{k3!quX0{Jd^at978=~A1KE@C9&K!(5vY7W1(O8 zmiJ-iudMTzVdHC06dT<@zvzq0S&xWKP6{ehPh9bl$6Z zrybp~%MOQ5Tu$31Mo_1m1B#EwVi&dMZUSvT$r|dt)wRa^s|Oi>T>WR`Pw0*xpgUfG z-rfQ4*pxhP=>vTy6mAp!18n{vxJiOPY4FE#_|1OCL)IbpEB+7*&aFC*p|2NWN0jx& z$zsc{b_#tSd#wKyda<&ZQdT?qZz^qFfc$-(@+y$ejkMPV9>h)Ky82*W0qqofj_B&?v^RtksiSSG z*gWHWYCXi70v+6c68#)okpCpQqq24K-YN8Qjkd0(t@lw@A^hr&dfjJDoGTkoZxS;NXDzm| zLD6U>xN|34asWV(2Q-2cRmbY8#}0t!44MXl{2$Nmk%i0n4}RGsflfD zjM7W%mE0JkY-2&>#zNB3|GW*#Ha4c)Hbxnp>9$@9>`bxl>wzJ?E2JkYxq&>9xQx2Z z(4edv#sFpAkPeh}18X+5Za{{pbp!gXS~sLe5dR<1!!{;x{B>WzI-k|3%R1=x{x`GxO&8HAp=A{-iRDp#MR?B(nXh-bxI*u zcwE;vrQ1vYM@ZS!Ei!Qw?H?T$2Nng+Bf)nB`nc%gjxar3^2(p`UpuXMq+8TCBL;NE z5RCYDiLaVgE9Vxc?~xd**|kB&f8=i@&g%Bs2F8EXi|(vl!wPEUdpe_^l{m4Y+9vQh z<{60<<6BAKZ|qYNFSfF_1$<>}cW1Ru(Xz(63KzH)a~ zd$yJ}u~cHprq!NC9$vdr;>u>%p2@}!cdx{j-Clck3iiWg?yNr(e|ByXHpROo&a9~R zJhE|ek;IxUtGzHD8{tBUH(OaN@w3+#xU)w5$iheZ?G`RZ-e%!p)NBg}qi0(EA2ZGB z``8<;erLGdS>tZ7`kOh~>g$!)TK&B08mo_2UupGk{8+1R*Nn9KH6hiVHF22LpKFI$ zeZkhM`f**7)rU#(R{JNrthQ%GSnamr;D&a^z3q4Co*i6kxfXC)ad7=vt0r~Dz3m{* zZ7tUVE_j_C2N%OTE>E3CY#i4PE_iiUe4P~scbT`IInR36BYSOg_`M#UsUZARLHMis z9XX{Sd|%QfPOdK7omG##Q0d4B1!vj$kR9q7S*#$kO~E)jzpR0$kU!B*hi|Lj;VY_M z_>6jn?M=JVibuWATUfNR4+*h$3 z1Bl5O*cHbYH7u7{zGPzg22?d|8^~`G^AU;#k@&t)EQsto4aI_3`%fj_uPe?k7*W|_ z+A)3H*HufLhvq%bx4DJ?1|*H=_Ru?p#EZ#0q&M)6`kk~^eh-Al`7tlT^ViAmj9HJE zI4PUFQs4VhRyxn5{keBp%qvxv{LZ`Y$nVfWD9*3zH}XnZzn8LhO8j9&Wp(een3w8( z`JH!Pli#7OlxytRGrNv@et%ZVK_Bgk8Dt)J)kWTCB`@^ZqaC!til>XH{8#dKy-)gf z-tVega7bNS!r!6Jt)X|M9rjpkly(Fq1~Q^@v(y=mXZVRKQ+}sS>*aUmPAHbHt3G~{ zy~Qg+Wl>jY@3P>{;bq+`zeCq4pM=Hu*?5z(7KW9zq<2~H_VBV6$nWWtb-{|$>x%KC zobO3F;dn+Cgtz;9l9%}wqMd-yp!A2f`-~_ck(1Jb2epNukq#gYkTKK)~LLK zd&<97@-pvJR@!Y2c8~3--g%KtR)4K`MoC_HM~`??$#0M8K=KE-%porEwyNfx->(Yn z?T$^1PIcJviJ!N7PURaAq5bKndg(9W^0!re!@XPn))iar``92BiZc&B{;LfycUaH! zEZB*BtmB;}61Rxl3?N6PUmtI=&O!LN)`rP;{y7%x3`SlE6uZqj-ut1%Gs0iH;u+zw zea167Egt#z#4|d>@UgG2c&KVCUb9uj2oCJhM=k{qm)dXX^Kkl{?b0tE16dsw0||aV z7XwM&FN=YElQaBf?2p*}l<`*anyPO@yZWYl&YWG*SwNsOcL$3TU3 z*21&$jr2UmjOQQ6TlezLMdKiuLr2-apY95agQVS|ILI8*yUtpu?v8_$^!Cc0v2CJ@ zKW6twY|o+dE!v5B%yuNT9qupoo9G_)8}|2`^Qlj#zk!p`DbJl{L+Ia7uFJSCuj&l; zl$V{IgQ_}_qn%~MMV28W%aD;}#7i0$T|1Kke+aIE&Ye+8_S9E9E3X4DmqhQ9hq3>a z#7CYMxoO2mvJO$_G8CNqxa(X7;v*kPn;D_bWmwSHxeR&H-SLt1G5lPHn3&I>%Mhw# z>er~_QfOazK<}8zT*fgdaftM@&zMQ*BbJ!S5xl2jCMiS3Owu2TnM_tOli~UZdddoL zFEWL9*$4Jm;G`WpN!mWXKQGTcV<#;I-$K`Vrec7i$C%X;FEOYGJqA8-sJ$|3t;CvX3~Q&ctf4Z7xXp3H z@G+%wmIHo?TX}w-SiBQ?4)cZ8@y5xi4s$l|_a|O2nYg{*V4FKr;UGSc_`))WIf&TC zbBi73U}7837dp&6*yApYahR`Sk2BL9=4Hf2bR;^=D~OHwEY4x>z!p~*)UoXyG=d-L z?3qh40yVl>UHYC8JfxdBv?IMV-U!w>%+%7ij6ku2m<7&!IOH%VV>i6ubC~}~48v!W z1XAWifluVs;5LDz8^l|J%U1Iz6&HKGUyRF*a&XcO(*3B*9$yk1S^QG zC2lsjSm2x3UcrAE&+`THJWs}l=eYuTocc=kb3xisz83vlX7{^7Rd7y zDL2LlrU*Q21TU9zc_yZ}x{T*!(pUF0f{6m(HiB`~Lp)9}S|GnWrCgpn@u$i2$E@S! z`Mf}$;iY-3`By?GCD6#vppTCE4)bTwN5?#eiq+|e7Fca`I3=C<*=ij~{On$V(MCs^ zK;lGK33M49lLWp=%&)-1Mn@a*y7Zv~-hDkWG#wuZJYsaz5?{Ld3Zvs)fq%2(clrI0 z#Fz3rYass`p5fzu&gAYa^Nu|{B<>~YmqY|{dK2X!9J++P-F)UDUdzm?S(pH*n) z_SJDlpiFG(HO6sd(eKy98K1!yf4io}IJ+D<$8+6MVsd#tj~>^L=i`etGl}O@g%UGG z{`s0&$o^`v@qeH4&(4*Y-5R4~4t67WQ0L1Ib1S}uUr9_$QOF)Duf7RI{uue5_PkhtgRjKR?@@=SkbXyy#sD>2INIum`^)QM3{CVq5~|MZkx zZ2Uw0=dyFn3sZ9?#@Bx#FV{3@vj*;bru);QxlS75_- zU|Su?70VUF<>HFw(ui-BJ*4>JJjd7@{FXD(KT8{+d;}6RltJAtVxyY1A^!fFoP!ak z;%v^$&Bfk4z;|X&u6cHTt`!?RFV{S`I9FnW{pSjE&GXBpjf4H?m(t!@Kv6Gz8;9wT=0apwB(9Q}Q--~-sh^(vR2>-Stwa=k2ZUF5-M zLY#Gf-y7Jfk5NaglbB*>jGs6&{~_wgr|vuPA>NPOJrrYkvNOjVPYiYY$2sQd=v@3T z$>1i}Jd-HnMBH>7xEXBYCK=qM*toeIT#T`ClMZgi+qlUDHkV%*phhxE|_=q3>hqJ27qkx10(7(Zm`^-^Gt4Hgq}X z;N)vLgelVhSLlDW#7)!xy*cKYnjEYDhjPrbwK*1Tyqja5`)iIx8z1DD=Yu&GZG4nt zUTEW4XrnpDG*9PPv@HEUnPc_;Wvl;3vEgUX_agefp1wazTg9i|NW0?b`_;7btMq?+ zpZ$;Zw^E0TzK@~r(}~>^dW^&m<)QCag72>W7w4GmD^%>buYEZ{bvrsJL3Gs;5@X4K+!-t zI2do^AQK!+vT-mG9Aw)#m;w%_+BnDq2eWJ(%%IP6Z5+%22lH(l%mW9D!*Eb&<6w-1 zgBE;iW9Yxkg9qsUF8Y52Tzs=QkXXlP=pe>Fn*QHJ{||$UK5_6@_7F5H-=(iTNW3UI zxta@Y?4dX>epz?{>z}io_8y9l(X*m$960H7qK$(%a4^`$K{7Z<3B$qVHV!^=D_$V` zI)o1Hr2jH6tH4EHIJi_^(Dz)x7e*aX^nVonp9U^MIFJ~Y9&=#}IM`<6zyJrkZ5-?b z2QSz-@PUI@Y#h7{4zTGcS@D|NJby^`tRO3}|6F*NXW<7m#ASaVv9-uH;Rm&Diyv(5 z#=&U%zli<|KfD+Y{=T_DJSueve;or)2*<(a=E4kcFxSSx9B?q-#=$&ru-L{yAvjoS z6-aI zusuaHPXkY1u9*Y$ejR59<3j?Tt#+8V0*~)?n9KB}j#IBWO!hi?H|c{qDD%TQWPRN$ z4l__Ou<|zcI(NRnec`~$x>s^d`Mzgxs@u$$^xdSV+v(|~=dq_x(sz>Xw9`{a&n7-W z(zlTw9AK4mIq6f}<_(f=kX~)44<>yQG2xQFmGqrbC_!S#=1Mw|3>;&S!Y1jtD57S@(@uxIw z1r7mj0S*Ih2Bs3D_;FcPXT_i%K3=u%$F}roMb9;VzfWv`OJ2NXUs^2orG%dA`FC}5 zBx`j!pYdt*;3jn2-RQ#2gLU&6*2OKXu`*ao%i4YhYw_K=JpCi9t zcugSd`d26qo&Lhh0(pKx$_M&bd+^?8yEXGa{T$Yvq{;fU zKWoDF8M>KVnqZuss+&Vf-!U%a>E-}%b0%9iuPD`w&!*_+K+cssJ4rVOfva7eXvqrpdh`pz*6Hixb<_}oQom;G# zx3QKxQ_P&D4`&Mn9yZR-mvIEnlV{dUa|L4Sn8V&q)<9=xY4`|Pd--(sc9O4J>?g6t zXRMvZ$(m@fZrNc9b@Li5 zxP9aRPZVpmgQb5o#;}&Xo^@;{W6(BH$G)>qtxZQUHjQ!6Bx7~rB))LQ;@DpN?Tkgk zR($8OMw_9TEyTSqV2{@>_RgJS9rPZtKflz6`hTTe;r~8+NF?@V9Pu!R+0P*7Z@tC! z4%fR}HC+3+4zhNV_$cDDJdZgXzOm>p>xjRRc$UMm?}RdoDKmjG4{8JbccIIO9y{CV z^i4!2o{yXw{v&5Gnbq7^#MP@EW zhV0ETCpmea^3LrP9bypsMqEDjg($hHVnWdyStt5uY6(i$J4*dmvksItoZ{Te7!oF$IO*}BTG9BB`5tI$l7!0mOmiQPWGOhL2f?5-kCEKbIf(@jcH^IqLH1V zONrhuc(|E%jYFQjg$&EL*A$P?chMuRMRxw2{(r!l@-OJlA9DST>vgVbu6K~a;iQ9!h4QIXPn*{}gr+G@oXTkb93qG)L^_Ts%*u_n4G zYU~$YsZA?wkp`t3YPD!%<}=Kf7f}8y$>S(|QHdNi*js+& z;Rk%!<hH>P74~N=9x%d4Ixsac| zkCl`g`MSfD`~CUs#j=ime?H}QPTumAr`#exUQOAhy$3|%dJ+feSITk4S)XVT_7Y4d@U`6{RSu?eQ! zgDCgu^yfmkw^45Tvnn@z`7QWb=@!D*`)Fs?_O!3m?f#dg?2HLrl>JqaBk9{)C~q>P z`@-v{FYG~%p)bhZPA(&QjK{a+7D)MzFO+{EWf#7e^4~!})MLQl$G+%;dxm*OTU-$UqBtp`l(O(TPc6kv^#zHttM0c^?FUA$`4P!DCMV} zRrw#rCXaopccRq8gQ48ZQV&hk!{fRhID@$-;Xk4M(4%5U#U4sT%6@_>yU5J8&)c%g z+VQvW{L!ZDUY?h>zZRZfK^tGsb0g0}o(Rw7JYN&K$Zs#c^MvQ6>}l$DmTCVg%6|2c zcz!8mUt-G6c>UIL-bKH38D(E#+8(~%Pd`>`%9ZdxZLQXnw6z**TEs5#xX0dI&Dg(= z`e2Vx#=7^{gqR0TZiyrBy}z```~8)F7v*RF?})rFZU5i2^9s}E(g!RL^w;)JQT`sr z)#8|keR6+oKZo*eGv#NTeyh#2{Y82$rP_WjYbpA(D*t@S4{xjT!`F%jXk%6Wh?L#Y zYsCd-oz(Hc?QY9`hUxRt?rUjpyJj58IDZ`NJBsIco)dUZ=5ct2$$nR%-G^_9((XGb zx2%)Wl>MWula8X!Wu3H#anP=l(4}RaGKF-Pg1o)32`@z7yHBRm=pxE^e z_l-UzK1L=3_do8-T?#z#s4w?lzGDx`o^+pe(u2NSJbbi!6uKAr&HN7@?Is>wvOOS_71IQVUSlNvnZ&ofIBYvQCoo_eYwGYouSYKJxm%{;0NIXR!YH z49_IiO&9Z=#oA~tPaW%|89de&us^%HXW)ry`q`J^HFW2cjh)!ieZ?6BZwxy06Q1XJ z-r^bV8^kfvphMWi9y)-mM!z`38oqr7w*8#D|2d5Q@txSDpwHfJ?GD9$Ws?fic6Lwk zU+nB=Is@+WOM_aons zL)e=iI;~>FA?il8$(-8IU9^j`?V~NUZ6Ev0UfaGfeojQ6ILsW%-QmFM$NZFIhS&j6 z4s=H!^HB~tmsk2tQElJzhVoxfX5YiQrk?dy8vc;=dj|PtcYUCx8lF0hHRVLsi>I@` zJ%jbVtkoyMw`an$;wwP-cHe5o8hCbpl(v}0n$K6c<#Wi7pXZs!a~aPUc)rMUK5HCV z8_mLwdzDkU1>fH7t66ij^Ss3KHqUXa3#RZ~##4p^ zYm(!)R3eLr47wB<;a$2Aw!5P{4{aJKIR4gu$I+^mZi{hCp0YQIg3Z=eVR5C zS@&&d`8jK(ccJAVw21B5NdK{0j>kUc1fG+593CG}19X??B7D(1L8zvY)-a zgRDtVN`LGS&-*-ao&?X^?DdJg;ybKW#5ckcXgLa6CPK@nm=m8d^6MFn*cKkE@<0x2 z*;z)5t&azwYY4PR{q2W_0|k2UuUj#!a{# z*a@~g#N5u=AanZK(1XkeJtLrJ1oZShtMi`IWjLo3dY*w6S+jZDOVcw_(KAWt zK_(P>EG^Qn`lY9=el$J*O#RSrpk;R%|MaP!sHR8yj?i>Odj6O7$)-@QO~NM=9yQ^}yUCTI-1jAX65-5OD*imeSA=q{ z524kL9G8pr;lEIj# zl?=u_U68?|eeIFv-dL2u&hJ;B*ZAu>oBe8knSH=!J>Ly*p6QhNZgt}&UE`VKs@a3A znt4gr-9&|F6EXMc+@t{vvUd-8=eDXU)2#>ul!4yZQeu{-4X- z__F9#F{l2K_jlH_r_5Y=JbS0PHBN5y>`S^1u450DdGkwzy}ufL6T&bgQ?j45z4C_x zf7U^l_#z(Y8@NSm@Q|tS`yJTwL;8oz8zoUh<;1-$lCJ;x`{2utc zNh9Obpo&2)7bBy52|bYz&lh+uWTK*P)ICw&Kc_G0_OKFpS@phIX7T7(Xvdnu31YvRXjxc|5~~rtEc)reaJO^?)1X?syW%xkKC7{foT zem=@Puj^+7_lTH11?I0bJd|F+H~}BYInQ{R&zQe1p#9!p{@O(w?pb2yG3G9r$HX4@ zAIx2sJA?PE;m)bO%wG=y-R#^ViwH1I%Cl1$=w9llw97P3ABB z6dn58bmmc~@{riQCXf{p)0h+5eD1pkm>=wX#eC2`pZQ@WYXu3n^Oc0pBYXumzY=ce zD+xFARa(Nwfm7xy2{-fA>k@9~D+xFA)t@BX&Q}s{=Bq9VA7kb#2{-fA?<9Pb(9Aq0 z;nSgK0r$%ho?`AwZ)P5H%={(un9N@?kIDQcdkHds$zFoYU$U1V^Ox);i~-7C!YH8K zOYrx#OWAlsaUQ#*;E(5of=zhvG;s3Ao z{8-5cI@Fmp_?L3e+9zgpBXj?z+TG8x$-Rit+ z`ynlFToCLlZ;TBzV8eS|SND2sc%6o+_|KSH|DM>bp5G;Bil-bD+trJ^((BNlI3V_` zmvzZG@3^h#QuD#2!-~FxFxn9-v z)>78<&&Zt+*L3Y$!kT`E*t}lXCH`v9ej+sWzVbnY!W+O>SQ)` za&KsAaxQi9JyR!P>f}47PNq{Q-!gSFjXGIt>SPLaa;K@2$<)bhrcNeNC#y`IOr%bl zsgsMSlN(H(&@axo&eRG0;A7XCI)U#eU1REG40ZAqQzxUSldDXfjG#^yn>ravoh%GZ zJ?GzaeVlulu8S#`>UucuVqFJQKc&n6iSu>Y$2RJ62ScH$^`F$`J>@)I)>F^c;UC0)-BB%d^3Ib$hTG1wk>@5JeJ_WfI2C>F^|1EM+8%jiyVoAL zfwbL>hYB)2D#&=LAmgWkjHd$CUfo656En_AJ|ASytkN-VE6BL6(#sy--2Z_+Gh@F> zFMEUYNgpx!FEMbb$!D2)XHBK(V-8brrAaS)iVH~JWWu8ct~Tjg%sXBB0%d?#*FzE2L7+JTZz+pqfEj0e{% z<+vCFytrPF@66jZ!#wu59$YVZa6MIyi!sED>(}xfo~)_uAJ=vduC3*`7<;_9+T}a^ zUK8pc*H1jSHk9LH?DFE0^}ezbZtfq~dTbD|0kje=VMGZLIq4 zU5f|*RN>$B&JQ{lFW#12b-|8o>ecrY{!L%Mt#i|mN3xB3p2((ddsE~8(LJ5@XFiO_1h-a@@>px#xzgpGZ#)E#xRk-8Dv z~w{HlhBZ4 zoL6_HuaWUy+wvwm0&eng)+S+mQEUTZyPN9Jb%yL7i=5aHyJfT!i=OC<`6**+8~cES zuOj>k!WT<8WlVMS4KKAh9jMPvpJ%=inp4icjoeuxG;bK@e64~1$ybFgJ^w5YI4_>U zU03lizAb1svGZY1s{I=3b*FTwS!o{VoU3fxn+C?Z9}VT^ z+j$$_lr(pC=Uc760xoY6mResdK>enP-zWSCeW<_sb&TD3 zXHReI_CRx|jC0tHKU8=6KxH@H(68ONY5TIau(-B?Ymr~$^2r@NXV!cuu7$)Oz!<6K z4*IvOUzJ`pW3lx0`>!jG@7A7r0R7&M?^kn&&-+79F0B0p&kyUI7iQG{i#y}J*Kd20 zG5KEJ9p3AXfu{Jm?gN_+x2~*pURb!a(dc#SdwC_lJIL=15B>>i{wN)9#RswYF0iuO z(Fr3bFpdc?i2tC`@Rq&bNbQplCVlB?8&~#DOx*1@%%o55v2p8VqoT2^a zgv)qYJ}yeV3H@L6)K`D=#|5f9<^CJA-v*NhbXBzD-vU`u`i1uGkYJ4<>^@a>KnEtHs>$VpE&VqZ0!TqLZ zE^+wj`9!Y=V(NaZvh@IS1O6De-#L^Mp8$`TdRTZgajDm&(Amb=2r3T!UDbV%@2rh(;(NEBIeJ})y+~(j?_}c_ zB|jjh{2TE8E9e=^ydd+(ZH$e<%I3}#@rNRl#8*Q5H1_X!$5y+&eyH=%s^(hN?qiAf z>8lR6%KV>#CgIsQGMMc1CByX1=}s;cb#lzlZjaE+e3Kzg0(;YLe`WiB)0f2;Nbi2_ zyG7kq)#13`fgZ63;ZCy_*+*P}UPlDF(yZ}@KvQB0^sM0FU1*Z=+0r#Z=pk*I{8B>C z3h0?apMahe^rWGu4Vzm{kH3VTOQ1(^s&hW_)&0iEk(9rN`b+pjTczA3-#q+x%ZlO{ zbp1!;zqO7t-d&mgaQ^!bgS%h;^K1Whw;v@uUi7aMUqbsWE%M&6hWBcb4GVm;#PVK+ zI~?A#Yfr^@^3Cv_>ho6rTH0Yj8;4x8vmyPusS6#sJm- zJ9Ph_Ap4UYW&M8xINF?C_h#wS+8Wo_F?x3VcSab#-|N^?9g9rJK?ts`lYYN zY~4TIM+LK4Z%flzhQQSVtjD^40bF8*iAw_tgSEi6?Wa)BcE!M><{Ces@D59(4^5 zO1$<$JgF<8CE(sy2qT{40iBnUHon2Efm&zb7jbrM@kl3jrCtBEHWD_o7zXcQTS!>A z5T@mk(mf4pQ@H!TigS+0y?^OXUn8u?r>n=vdbezRGV+NXbL7qfJI087w-6pJ$SKT& z4W@sJtnvl*O^+?z@(tQ$43aV867@}wDe9Z)e__(ERAZjhqrd$!TW8d(sx#`Xk~*{V zP)SW#Z^QiVKWv!FN9xbl{(ud$Y5Dde8>ZsvaiPtI!Oxce zf8Y(1vOwz}`h+po{KZp;DxTk6Lp;^4hL>j8@fQ3q+4z!|#OH|?@Je0k`hBht&-OK+ z`?Kt+!)yAXCHxiEiiv5%hUksa*%LH6tP{^Lo7^{K2+5<%pKaSrK2C|GvO8Vj&p6M&-IK8<(}mKe8QD3w~b3#3UN&zOB45p zrd=fNJ2H-eSC2)-wNd#xe21s3b{)?8bd;1;ep+5K{tuPBTJ(R&8Ftr(kn^?6@hvC5 zgAx(SAM{VPdoFfgMf#-d52aMSG~dO3ChJe_S1{GVoIMA5G3;~YUbA$Ea%^&&Ggazq)rsHe3S71L;8W0j zAA3k`H*hxs{Yus((l>9<(-zx8Q=_bbQVWn(Bj62&O`BD~+XLDcn=)6hK9xPXTX>i9 zUFlRjx9yYM>-%2*(g1ntde{CMWi6E$8rr)38;4tKhdOJW=969aVb<0)SGh0EFWV!D zI8$Z)`Xz6>L_$;X52)Hix5rquA9{O($glM!`=Rj-0p!;}F1aOu3>(Px`cKflIl~4< zkJs-V8W!4SvS01brl;XqlOCZ-n+MYE)zW1yHcXYJDWuV`{u!ah^-=9lY zf-^eG#MQ8ev^`CN3m6sLquid=La&JzHW1r<ID_AD1% z$GJT#1UGPHK%HOo;`>13F%W+S>OHKX!i%4SG(He|@kOBHYS=^Co+W}0xK!|ubbFQ= zynYimY+!VRNvom4yBxeLgkJM4tYHsnd*%zi5pK@{!4HgBdXF)94D=fq)==RMgLk&j zJKV(8uqS$q+jEiN1I`uv!`+^FmR=J#Y+!VlNvom4I|96;gkJM4tYMEIJYm5%%EFGbB*m0$?>FV{HZ`f~1+#Wx8CJMfz-JVH;A2`|4JJ{ec&~G62RgzZW#Rotd+p6B^ zAQM-^p6Fn=r$+DrgMxpM+cVD6YvP6tj2>muYN+te1uu5_y?*m8tYMEIJR=0(QEtyD z!4Di`>8&<+4D=fq)==RMf_I$ITV>*E*b}XGdnyGVaFF1ya(jkadQIH0hWVzG*)Olr z<=@YE{H~1EClYoN^+~#3Kj|yoo;=VG^lLl?->Jrzs-mZZ^gAUz{!^Pq#d@Lx-JZRY z4!B?N3~+ni^{f7=&uj2OE`C)xzftMghfB9gdhK(yhx7iPq@)YDJsH6R+ye~PyFHtM z$)MY_MZN$32tna{@Y?5+>RGUFR!&(u4yo#TDfzn;%} zkTrZ?Uve<+uWmo^9EGk~+2*O{}#-|EiyX0>GAYgo4%9MxaH@?f6u4(76D%w^IRN+&{{nQ0vf)jrJk z=_S%WtSQi+=!yE>p7l~czzx7?h1=6++n2S4I=^A>fHN7Dp8cZq6iKi5so6_%Q)`$n zqR0a4fPP@B%u7XCK+;)xK%JMY3-^Cs(%Ls%_M(H7=jIa|yW*5H(Q%^t%wbvYQ8yx2 z@_rq4X5YV0-P-^6c>cF~{(p=ANo0o02Z!#mI3F&<83AW>)+R%xSdb9gu z_8cZP3Rx9dBmAF5dDqH-qN|xU_x&pW(EXpstevs2t65!bsn>J z;j+HxF{hHJJ&)PvH}@BazveRggv1ukO4*;>&K_u-Jvn)Q6*(;))p`m2@ui&ei!53J ze|LNMGYNkt8B>z*XOeLx34bQnTReUFQ_h3BJ^ab~+7+Lh-SB633lRS7W`s+||=#_O{eb@s)6sI0q+`=MI5Hn~CbrSM)P?t3v6Wu3u(dV=xu zDB)4)R(#L34xm5hEZ-_hlp5%X$ z@hH8)nJWAwa`q4DU&*20*m^7XCW_r^r87|F8Rs0~P2pcpzTcC4$!{UDbPf0-hw_(J zQf`sQOL*GodL;&ZW9weg4Uu&e|KH?)f4nVx{mrlSSr4~&=I1X`fiYXp81!%LZlt1@d%z>e=>Ul zCa&<>A7Al=@&05#aehafcyNH5Al+kpm$DXgs#pW6c9ZX<-zwiHg$nmG6uyyG@_WqY z*T$YBY3`T!mL9^kN?!Dv0{)2xzr-i~JrbXFam`4NPMFENn>as|^al5ua$3ig;#w`= z$ARn2{&B7G;JTq4SFjY7bt%+6P+0l9oPPa$ktFjg$jCb4YK`QqNpCzGp^Q(+WVHx-Q?VKa-JDRg~84n6)aBD8+ z&YaJ3cTN?sTIso*Mjmo*sm?{AA^XRCYoc=H)^7TG(5cwf)a={W%@`ABOc7cZL(6*l zS9GMmB?XPQ5GOrd($gP;vd`{sKg@Wd>dQl0GqlC0Q`g{&~ozVu%bxd9s2i{2IaeHWr^w$#s|U{srDEct^)_-RF6q z$9sVH0cHAF<_?6twDo?`q4Vu)Kp#TN+=M<_VhD9dAM%#F$&_2p3uK;Byf( ziaw|2uKnmlu3|nU{3Z#X$G#{0u*g3<*F9hVk5tai$^M#zUomJl>niLWvq>s%UQQ?&rQVLPF$a_@(}U@c2~Y# zz6$4OR&HARSLCLm?$Qf&ji;*fy`bnLc^K1|S8(&iPm%?bCm`B74C=DT7~XP2tQIz zId7GFg86?JJocQ#t0tb%CS%pY%3>eBd}s)rkXZuMn^eIeEysNmht`{#L|$9sTxba~Vnfao(=-Am0MmS>ZHDeCSi z9f%!TckhFE)~0=lwErx>+q6%Sc2TjfJXH5Rt>^ba`KmroUD`FJjK>fDM{&-u>&bh} zIj7dY(&q_sC(ZjqM-|UdE~3B9m3da{7uYt_Ivt^{N&4X)xQlBy(Pe2d# z{h=SO|97vfoskc${$F<#d4YUH-bqpya%Lj+{6*@ddvXXFNbB&4Ts{R7 zoxU3PRCq7pUEhwjdOuEk?J?+i{*$_eLp=5MsHeWXI)(YX1HbrUqTeIBJyFNMB}rWb zS+k2BDesR-odj8{i|wd-r*7)7kDS57{PS311U!okad#!>d?!0^MVRMOw2Az`f_biI zsQ$l%d2aUz{r_s_xv=BkL;W3kiaBl|zA(gQID^e_om2VTIAk=L$Hw9(Dv3|38Q6bT zpj+JyUL6+DVd(Zs|I%RtbQm@VVyA%p)E(XUr~KOCKb>+uyWR#)>CBX}n|p zOZ}#~QKk9W5z_4JH_bOxnjak@&13zhxk05_e}ps-l!yCQZk4fTjVE06X>Iv8dcsv1 zk+mg$)D!NFf43)G;%{eNAn{j7c%OOj_5`Xb^iw+Lpk$zf7 zKb=88oo@PR(m%F6LO+c-_&W=Q?}*g$4CATd3GkSH+El)u(P4CZS%XMETK8h*)_;h; zo!!f@aXjPKX`&QKR=l6Z z{(8@?FRbhA>3$@8|L0!MKK;FSvooB5hr~B>1lYq~;wpjkPtGlNJx$*v&2#vDqi+p( z&PTuWfuBD39O>?0On6Ro?tSDp*6|m^=$Ah7AL|5)Vf0IDH|_0@`@+Uw3o_aP)-vBe zG(EPV!f|EIaezMTmC5h$i=R4o zH~;ag^4uzTMEu!dkD~0t<_jPBPceLQN*_KEyp7-$C_E#$sVBvQtFOD_;E%vx<*VX6 z;fjOfe3j2}=I%c2RJLz3_{BaW@i1e;qv&?SZ?dPG^q*?-JGD=KmM@b{dVQOsE>eca zWK22f6#4>VUwjUDsgDHTx*rZPZaTRyiA+d*H5SP?!qgZwhrYuYetDheJ825voAmcD zdcQwf{4P4kanfHR-z#TgBRUIt?-=C05wUpXY~Ii3eHibudH|p-{yK` z#fy3WY>BL>d0VgV75&hz zKVmQ7881%ul&e2`2VIt;{f6w-FKBl5Ih<2inPT{#OZLnk^6Z(vP*=WZ9ya6JB$@My zx@DH8r)#N$HphMQedMP}v_JpPs31)K!}CR1-r(`@WfJ}nx%iExrEQgjSLN)+_WeS8 zg^j+AvI#88$KC96HIM|tCVwYh;eKN`v>7z zq9wpObO3uejl9p~y@Gc?@1dA8xFMFh2>S*0cbe7uN3V&B>mRviI$YB3BKJNZ^+&qW z`YV-v?YctF7Nicp(@!1BdBs*!k9IzJ7TMA(>xrCRE3#(u2)Bhgll6e?ogT@4g?`3D;R6|GWgfnQbJ+VC1D|G$-N}FaHa$0kF_8JVeL-O!t~cXd{Sn7I z8DC`{_EDzk*fqqJ&Jwl_++C^n{>&LeKk}NIpW)S$S#u<@ElD9OM8P3CP9p0jgFaVe z++%2WSpzIyrN`71X;Op_7JeYSj=W?&<;~Cb4QZE;d)mb-Z`<~mpzabb?iWgXSbaHL zclyl7-@X}IyMv6|LLYk!LSG8mBE=YgHuT7yCGt)9HAy^KugsEgn}&9kG{QR}PkeR0 z)l9j~x2c{qg6jfLTzQ|#yDqPmD{Oh~*!%XEOZ)Rv@QN+(5%}*R(f!)<%7^pcYXoRmnS&1A$(V8TeB8@w5AxIWW2TQ_pB#e+Urm9K3VHaD&7H? z#GOyLvTH+L7aq2Cr03Pr@mhGf$-u2m6%EQS+==zMprSn(up+&+Mdg9BNujryI(uW$n-#lfRt8kUk<=1Povps3N zbj>OcN8eKNAFpurrRx)A@$1-okoYHh!oBtTF;BS4cesu}%oA?QCv`j46RzS9(D4Hj zPCZCFdF$!8@^IG4LeIgIiO(FV;!{VGMy2OlHEA>tm(H2O!@s|}I5y4o@UW5x!+L#j z`{=&w3oDabot!*&(W0-zUU4pbu&JwYWK$Pwt{1Lg{q;Cw&Ww>W(V?x_)iWlvb%zt! zwTgWQwWb^$V*M0oUyNKI4TM{gbMR}$|L(a)cK5R$p$^j07bwexQvOQzT8GMBtCpYH zLwUjO&8q*LBVDfcX01GpoH3NW*9W~a=LgHv&5Zxvy1Y&5_hI@G>ws?fvdwAk{Fv-@ zi5)RzlyTKLy`m*Crg2xAGTu)apLTrs0I%a2*nTTz+7W1MNl`A@TTWa=IT>TSXFIvb zXvQDrWSPHvC~uExbD776NO_^Vzz2Mn_U@*wyO9@Vud-WY9+}J7cf69hNaPjaNgwx2 z3(eik%gNb5&hBLHUhrdQUpx933Fg)&$G`7##_PrJ&FYf%+z;6w9KbqHt?}wR?onaL zmsgGq-4!4_{sMw~vxhw`$#V`)I|wY)rg3*Uy5uCz?M3*Sy6U1_-z=35*Z3;kqo zSnc`s8Wb{cUu{%j;@W0k3Hb(>~plPv*>2Ssn>f?$;!*j-h?%Tw5r^v3>Hoy^vS2 zen$#^PyG#3Wf;|`4A)z{VRT^2+H`q8crBfel;OAGKT-6L?vu|#n@`7>KKXpX=CePn z_SWj&yRmC5<5`e72_3&(tP8i^j9vx%=j_!5@2KM&`*ZEgWp_0M8d?(Q5=*~VbxYq! z(tku=lyUgK8AIY@n1hpH?oV+>#z!^m!e86gu;bgsc@Fq$dHm>;hVEKqa5?;sqZceV zleC+xuTtQYb(G*-Bse=vU85f?JhJ=5BG0J0AWZdD;$JOwW%{X?m%E`!=7^t3TH4>+ zXKh;1sgr$&A4?qmOF7iM@hbg?d8LLq;aS#XarBuj4$0$NGA|jN^1in0{m!!Y+sfWo z$vb(Vx%i^su3;X!nR%#!d1wK7UXCAyB<;1B^{`qC*FSNOgwMd|i;woXobWVr(MIZd zM(y@{rZX2cFc;l~e*tMfv4gR=W?+vbxFpRC@HFuK+kE%ZujEfpdH$`8-pk6~FD-k& zSl-Elw$1sFe=3GGbveV(A4VsEI^KG@;xEGw{qUEUC+ryP<+H!2_Alj$$OrI5O6d6j zo;cIP6Dl5UsK!tDN%$ZRPmCjumK$q~p0Maz-njMqqU)pdE^pj=kLQ0Y|I6pcbFu%o z^W!v+4Xctb{e#r}IN>zrNB_~K^W$b;ZaRBdi|U+9d$C6lTZ`VXh6&zf&sO;EZ(}GY z@59g!Ew^XrMJCYcN^KUbokF=Sg0(-eHU^=WzTa8<{G-`>?|L@dy!fqbd>#An2fo>P z@03`!dGhnw+OFN%cq{uwE#L24_|^NewX^@2y|*fzjj!iSc*h-`wJU#21NhfQj0sYdU*P~Y|I!r%-z6AB5)<-Jb)Ul<{uh0Iz;Q6M; zPx($hdpx*`<<;xUPkFxSG_)ycUi08e6!N)Vtz)NqzIpR`u}`=@|C`(=T+{WGgtPvR zPjqq{S%)MhNjq@%IhpcK;X!v${In!lbIZCSIh(LK`0w}O1Lh+5aIWl=HEm7JBYZyZ z3xE;QECDX%Sw@=WtWSvFy@I$a32Wks622NZh&6Ky>DCBOQ`eJy)HA#?#Rna<`7~cH z2@fTw`^blREsVVpZJVAAoWpYw@v&`+W80ME{!wLT#C@ba5#E>ZjXO&d+?ASGPTUpX zTuIy}Aa+*Wt9iKJs)xI;QtOD{$~X2=>t*i6kB#WSpY&33uX}qnzCnx*yxu?dwX^56 zw$GjuHec#&VHD*H``oeUEhOo0YVUIpbu*MZ!(^CxDlGxDx&COIf-l zzM)grC9hmj9Cy2i;!oQ%_S*9U*`Kaf`>jR)i0JrGhm+mk3LVgE-=C595qgc9h-&|c ziKft2Im1j)Hgq;~HShUi=uH>WJT7Ud&wCZl)7?jTXMOg9oYQXX$5-N9bTR%Xz7oGr zS^GO5QTnI&gDcn9Z+v@gXX6((Wp}K2F}o>_Z$fC?aTb0PUunxOK6yuW)7}42ehqg# z5baz%`^VYE`(DZ}9)YW$SPG zdp0%B$u+LNqjSfg&DqA^|32H8cqglEn&BPx8oMiRDEL?WL1(^oN6*cjJLbo-J34n{ zSEWf0J)3r|>kRCFG~0MdXLi-Jw4x`l;Rl_Y+8)gYW_D#)-Mm-R^Y80B>#zJ_cGGdM zWdoxQDta1ycXrnQ@t4_69nWX$AA)Y7XBj>kqYq{Sm;5!mBe6%(lbRy?A*D234?Us- zto$+3&f`;=ZWx-6ss+p+D63yec|SuZNk_4@9d< z=KI7LpYqSR5q_*k4{z~6@y|GhJtb{Zm^b!;}UGfBlRYPmd{#Wm3LXzI}{G1z(Pf_|LftG;3Q%kvqgz{1nEIxX1?Z z33CYh(1G@mw9BGOd=>Ue8MRMJ@mrY+Hsqo^@N+n(A$QXb(joInx+;l3fcX2Re{zP^ z%UwY}#*zN?P78E4XrGS6J!1Lcm(>0b_%8KB+bo7}QnU@_VGaPoH%;(OllX}Co`c%F z3Df*!>a6U2!47BNO{@uoH#T_uJtkNWc>OVRZt-HO+~=cT*NNEgKlx(5b^B4&%ezvK z+E1j|B&s@MeR_vO9p!bM&8fogW2N|g)O(xaAF?~I&kckJ%WuSNkgXQGG+~zVS1Z z|Hs?veX(1{S7;xQq`M#A2mOAh9nv4tfBS6*GlzbvvqZOaZQZ|BT7BrXea^&>)P@Gv zRB$aa`l1Q3Nvo>;a9r$zH=&!O>e)NDF$c*wQ@glyeoKxB9_ms7VTiMrDdPSNi^|_wpTS2~}FJ$p+e~NwW!DK&p8jsNb z>n>xj!ni8@(v6+QL-2+@*HU(4%nRDK#MgfHA544+Gw~POFpGn4j}rDYdYY0~f^gv> zIrIGj;WCDokIy0>NEu#?6zk6GPqA=brSw0F`UHu&(pLsBuIhfE@#nuugO2EPVuxmYMbbCb zSjxB{e5=Z4!wi2&+3xX_ZI@>bvhCI1d247WdZW&8ca6{@by#!c^ZuI+{{GHe_jTy= zl4{Y%>L}I6if>~~VZN%MU&`7!Wa{Dx1Q6W=-h_A!d5@K103PYN!^Ytd=PPta13#j@5m1AJJRYKVw`_0@k-;rBrd-y?o z{fS;m>fs~%j?B>CtsSM(&xYni61`#QNzOCABe4xB_>O$o_>Q~?dN%X$E;PwFZ|S;6 z=pk*IdP?z4*_6uJ3G~SNCk;JqNAMllA-ag*Opsn|QWVd1n6m=PkImOqk$uNf?vigF zethThVqaVF(fN_{|6|-eFMK#ZzGm?E&yQN4>sjTGXLvzpQv8t0TF&y~_n+7Oav|qD z)@Dw4sJqGVVq1lKjk2%9M>2ep&sS?cl<$TQ)tHdX zj~_|mrjS>ZAIZtI;W8e=w2WNjuknhnj33Eme4k9ZX_R9MW7=}sQH_1ve}laAIcc}$ zv>RifjI|}__{c|U95Hj#UDUDFPw*YZ_O~Jnejgr=2&N`C(s&_DB9k zy^IH?{>bC6LwnN6b!Vi1kD^XT^Bk{q%vaK{8>!dzGKMBu^V7G*hh&htE!y867fU8R zek3y$unWI(Q;uxdinszPvlk~eCjOF=!aePUF zXHj`P`s^?5e8;$?<}bny2#<-(r1~^xOwjmXKk+JFqOW_(b-AUJbZshs&40qDDt|M+ z72*Xe-0xbvk{5g-{@7l!`SU+P-rbaYO+WFv2`eeT&7X8pl|S>Ku{$X#zl|61xnH+< zB`?XJut#kE{7=xPlK(OrPr`^RedtM*ht?|;x#0KM-;bf4G@gL+FS*g;68xkUU6@BK zF8*s=zV>-G%+kTPOKg~mC+$$--eQ-{42WvvPXR{jz8+V8Gj1?CsjTB+h6&;XS|U* zOFQ#jc=$i~rp85yPq@%|wGC78=wp2Ql?{`4)Rn|r;0-fr@3HwyTFIaKxZ8&DU-PT4 z{bHLIx^28Ugl+YfiLy(dzs07vbwQf%*z`7xHkSG=jPZm^`OdXrHeRyAJyo^_SRRkI_!k?G?<#-r9YoI7bQZ(*6Bgfzoe4Z^eNkrG8Vi0VXLIL` z_DMMt``uZwOk2?hrTRu0lIzar+XZ~Puq2)8)9ke?*~TN6e12iEKM(Ck{`|l4Px)Ks z?T_l8a=VR7d8GfOgZQ#E{VzSP)Ia6Vi7V?dJ!a{3xY`?MEvv??5qjJ@(Nj(tw~`~c zkFu=1`=VMNzJNWW_lF)+kcS(Muf`{lljAEG$5w(nDgG&o`$Kl^D|PPJBY+_$W`L0}Y zoc*PgJ>ev&;5%j4#VA3yQqjjW5e}ecEr5J}2}l)d<&Q}O*KURc8ncr&BO z2fuZhF_Q0CH&a7?_^H$9+y+MNd!5PGzzz*ne)u~~i?2ic9QH88J}lt@PsSPSa@8;2T4~oA9uKQ4Lj|=+UIbA0NK>dc!9h+6?T_FazF9 zegJg|^h@3+yO~OpXY?dPr-6Q(_C%An2@h+i^2E zC$mTLKHkmjHF@GURpAaB7`5-lZz?`fb>1HTse{OKD0z

qE5{->DKm1K!N5d+i-Gt|m3LLn4*Y-atjP83!Uw4Fk6P|OH3Qzv^W+7d zOqb++tebh+lKa->3LB)wFngpzh)9doSnQ`^8UZsUOr{{GUqv41QHJ zTgWRs)Xh97c@J?j58L)MK2Gs%s`E0C^J4vOU{pg@&NlMgOrBfF)A%>-#fPb;O~Xvb z(Qak~d4VU>CV3BbGn;LD8egUz2KsGU!7?l$&Xlyn*X&dm7y!&g3Q@ z%BSg`Gkvk$XID@!p{>WFw~?tqC;sbK;ivSd_A6O`6#SGzXTeYDv{|`{_{WrQv&B#8 zlv&L8vzY5=<&xo9Ig!WYoLc5=@kuJaN_W3EyQS)I!@4)Xa~gh7YyL(1-Tnr7N_!RW z|C2Lg?K@V^V&z+k{n7Ml+OkF3(%kjRxof5);ASRMKYrj8VA$_wrdi(5{!R6Hq`ps1 zWJrY%*)TmFd||<-{g{eOnTjGKjvJ6mcOWB91p0xK1{8dmS{YGrc;!Up({$p@4?h2! zbmZ@HhEnixO8q3F6Wl4{>rv`e)=<1Zk4~L^cf=o~`F5Cgu>TKv{=durB(kLPacXdm zEyJn&M~*Q*PQho}?}L1tCOmn(=K24s=l?7G7dck+KuXUw&obvmRm#Wdm#-|7J!Rj6 zHH7F~$o|Js|L&1N%{oHtMi=7a)baf6c~XDpi_NWe=D9Z?u(lHEJujETe{zpcojQl_R%+Twj6(%H4E>`-|53zN9Tih zTl9Bi|DXN+t=;gr(w!*d^AvoZ>X3B}a3>HxPcgpfesg3#FX*0?@p%G1PdrMV@Ok&+ zK={0S8}Q|#Pt;SnE3>Y4hUzc&E|o>BiRBC{`MCChdL?lp#1a3fr*hXyP`)$9z<0_= zYJcAouksy!PvIxy6#D)X#+S^x=SuX%3U=S>9Jt)>JF3^#$qi+`QaK+|zEW9xFOfb% zdGsE@KsTSGKQLa2Ev4{~@|6lLJ)|l4N+o_1?O5=Yiod=mKdP@(@)TdG7nJ3jkbKE+ zA-WLC|KE{(rJ6E{Z@*2Vx5F8O{NKobY3Kgd*)!lztIuoK*`K4&+5YgX@{c;Yu(poR zhBxOh<{6($vUeU2`dsw&3O-Wh%uo47B|N)}I*hP)H7AtY_6&O{<$gu{`W!gX(ui%3 z-X{@SGqb68?8=k)4NQ;5?<(_v-V-tB#+A@9NO=9^_Tio~bec7$(7!Ki?$c5C#~$fE z9p-|o%A@LX4fL%G!S_!4PoMRK@%}R#C&FHw#3AmFh$Huw zBnW#z>V)!$t!@Rp)XiK{z)AZ12KE_z(5qSz1oV zB$L*}St@BKk=FP?Re0qaWs!QFZ_6oVBYXkz3-x5`{t`ON=^k8XmE#JQ;+i1e*)yv-sP7e5egzFpCkP&I zTOwN(>WMffNLt1u%PVo=IY~8BW73*9gCy-3(ylUXC-p38`Tib#wIpBW(?Z=7 z=e_6n&X`n#oyH(++zR~1x4)OCWj^+%{kx7$8blA9h(Mw_KsxQ zE{sn~x234>iq7y`sG_>Vp4IxU_IwL@9pzsFUySb$Egf%rUv>0aXpB$8`3A1n zFwgi=U&;MCuRBvWz2J84i2pphs`-g*{jYOMr*ucx>dwV~Lhp4Gx~HGtt#nFPtwX2u z%+2V_?#ymFm;1uu??BBhovCAq_rSB+O|!Xcoc>dP?d_dA_WUH9I{cSx{UG#K>BBp= zMLX*cZptb|>k)4M;+rgrSi?zj_Lp+D96{m#aZ zJ)GUq@>+IBfU-mXrqOqF)=#-No7((`Y^oNzpnub?t2+a~e<<5{Df+#mi3k0G7h5}5 zbyD`pf6ne0{5M5^ig`)t*m}y4kaGaSMaOmkyd^re8{jRaR|{`#fVax$E}6F?w7>B8 zO=iB`*HNwZL*E!D`UQhZ_Cpis$}eEYIM(PO^^9=TBljWG$R$9UU}wHbc?NC$HpEbZ%Kgr_kk3MJnCY@>1Q~ ztJ(WrdS_)Ceuiuj8qKJ}Se z{?X%?tP|L?P-iFD8K}A&S+UQD?zqx7#n0_Vu_0@b{et3}U+o3#ftQi9l+JE=pyE@j zv%7?LUR34$jB~+~b5WWvmfs1lwV=B@Sp4myyE~Zkbp`)+O_hKDRM-TlT(S zF#6DpU#gBhI=yv-dDr7=Ez^dbe80NA0aNFJ_CeI!_QBK}`$=9s6(z4VIl3;B$Umav zD|RN-?e*2pw?fsT<6B&>t9wf3vo=}Mpmb!^T-2dup$pN;>wVmeue}CNGvn(@Grlqw z#7UbV-d4`6rw4rdu7-aCfd+m5JMG;^2blOG!z9V)cYk4gXfis$3mOV*{Pxpl#D08s zxS$7o$LSTZJJ0rWH$3{-@W0v<0FTuP{?PazF#hz{o~)bDldN%P?kwsCqcinT-2tr| z%)05JKDxn{Kg-(Q;`%tatZwjjvDLX&>jr-iu3GxYb?95Eb}@3j_bl-EFBIhfyGD7$ zqZ{m%2ad3Qe__VLlJ)xw4)2NK&=;EGo^*>^zrV0B+IRha=(6JaUFPT8mAp~(w`%=m z%DJaqBfQNSrz*>IBTt1xx%WiZ!Q8!G*e_z<5uMru-{ow55;}P*orJ+V-GrZJ!z?bo zeV?!$R!^94;WN<_t|2_myY8dDKJwH~%3|dy*;nywf0X)*Q#M3Je{o_%W_6?r`WEx( zPm*t&^rMRQUx?l?eXPGSCOF@ex{5b-!|=~TbuSe4 zhaK--^L~3S>CqpyVTOlfj=jTE=JNAnY0v)jhijECA$#3Ii>VasfKYFoONc0wUHW!<^z&?wylKI4AACrHTx_;N_ZgvcCrce#NNo3T7nc`zPjjl_NS z)1fm7Z>l@&mwEJ%_ZU6o2z0*A!@JO$KvuQ%UMF;tHVqw05BYWIjD&LC(3yhHG;|(G z4|${LmV#6IMK}4X{ufzD43&c0=xF{?C{6 ztRJRGx_;#Eq@;n5WIR%7_^;9+kDsRUvUe77CTC=CmB#Q%LB|>S`-{q!t9TYBO zheEU9BkW5yOz5L7q+Y+^4Kw)G+B~Gb;C-pDi)=gzlX%eizKth!E%_2Y)5eo>NLum^ z*)SDPr(HqX_;j!~?EwtL(K zvA<0!)tOdt=X>@GT4#Va4BkfY&g_eqeQ()Y&@ys$iHvM@rVoFvIJe4r;{nzh>O3K; z?VYc#6?v}I-Z_r`f!GAgeW~^?RjY@PALPUTw;jW%BY5sj*0$D0#NKC;&oD1yPg=0? z3c`yL4?AM79W?&@bv!dh#{C1_R}?<%i1Unfd1EKC!NfD;KvAb!+n2rkJ2S2&&7LuB zIfOP6`&q9IZhmDM9`ao&V|a)3bAS6&77y==Pr&mYa(;hnhHrDWHHNf09=;~pxDVc@ z=f4UE&tgRfI@ZSDg0fe%>m}kzUrE|^mZXbnUF+9u81D(GKfZmoz_Y}YFz9>9#`C0I zXX%i%#6f4&{eL!0)sePSc*>@w-fg@m2wUJS8)cBOg#J~kfT?zdqQk8~1mqYdLdLAaFfd)_c93$zZkc}P471MlZ8eNqn+#h*2{+l zU2DEwXyaQPq`ArFC3F%l`G;&gTbF#>TONFrWhCuaz43%5o~sBGJ8GqCJz_>I zJtiE>1RG;-Oq^MwH+{i)?!V$ZDoh{Oy4SPNubz#5_4%TQNSNhrKvDW{n`q^jLxRg=)-UsPtkCiLrIWun{ty3y_45XyD8g>vcO%&FS9Wrg(d_Qgi` zy#jqH*~h+xcd4r@xu;?qdQ@%P?Hu6TS+}n_b2GZ#t)eS41RHIw+r75Vsdn3^mF>$J z-R`r!?T>Evs($v9#*DnVb6;fIp0+FSUtTe6&(**&^*47Oh)mn%z1t$*T9r$zui}iM zD%ZWCD%TsFs`nTo2DTZP)G$B065ZePbeyn(Q3E?P%m*urJmM#Q@?0H1IK|*LuuVhw zhPs^UrlVEx%z18lHIOvv7U=0XN2l={9AN{a8s=X`XS?le9VcmE(4>iIn1593Jf~Vk zmwmF1+hL&JqzP-7Uk5GZnO*@6;7zXtk|x~*Es>Awv~30^4Gd~r`DLQ7ewL0CHL$}# zzmA)qV`vfG_TZU1Zp6Sg1Ctu&(SuKuXL>0#fH%DiNSgHWDpkL|{z*E0*g$j~wce(N zd33&elV|9#;Kxi}2DWLKe_iX_rM3;Au20u-{3Z=LkP2T^!~El-Q#Z*?FMtm4rXxVo zq?Z_a+9vAwNdwW9ROurc=FvCpjh?2%It=uixM2GlfHyr4NSgF~Lr-|TP9HU}!$7~rlSfCeH)w1bA_le@nA9*|Y3Rv_?Lw)Y zL%CfZA8O7HV|p@OizOj@TR8&Ns|s6dZKlvEeym)LA6&{!~9(Gh!|UhHUpCe z26dV|b`Pq(h7z~TPNCecAwLLu$TK|&I>4Kr3?xl@ilHZHY!D&_wi%ezFps@MufNnT zp|{MAA>W~Wfu_VpA!+Ojg2qN6Vqlwwd2AWdR4p@%%vqo4!4=`ldkq-zX4 zkx@E+n}JCKgBo8Rn}y!+aXOB%E$GE&K;>a<5%PN~G(BQx5Ik0Yix}8uU{b?8_6TY6 zOb>z%@TP|XNs}I7=qa-)=qUrz%~szv(WS`5PN{(hE|gXI{;nZ z8s_Ii4|%5Zm9*0_ZrTq$q)AsAdfJBR_(=nU246(OJaz}YW%dER=>6+9&@ey3(6gVt zxyTUxtzw?pe4l=}(ojaKV8(+oYb-Y(Ns=#A*K+yF-VRJ-X{$sc^_*MV*LEKC9`L)>>2 zaMK&i|9kx|-*kA0|C`PKII`mw^S_z@51RkA{D0W|e**pPN6r6*{D0j1=T46FHuL{p zWX}%sKg9o?{7+7I)4PCC;Pb#BunXAnuWtI~N_6@^flmJ?OLY2=N6%4o`qMAZCLM9- z%+B49JUfnOcYP@Kx6>P9Z=O9P_K#17W6#t(vFA?r#eR!k{_j8Ok7bZ)cOui?&wk1{ zo{t*;KGqDU^6ab)#hxAC5ZiU;jM(qa3&%FsIpK>FR zFVka?>94yI8-S|zpQ6j%J%+WD=*?2Df(^j>^K+j^CtbeH6dQon^N|D3M+Q7Umux#< z=_ZKGD!Tvozc;hx>GvAeWk|o9JH}ooJc%sJrLIQzU)yaI>^b(!x&S+XyxAibe;m*2 zbwlbwSyQk_;_=g#=jFYVgXMqy;Ezk~Nhj zwgGlcBK_@1>k_3G&_dfAUtu3bFCe|)R|JpX+y!eq6}x3(Q{q|p9nt72L)S#QTRQ)YSDSA4bIfM z-uxqy-wB>PMtS~^^!y*r{}g%4nYULDLe?Vr$g~#ihiag*C-^)%SyK1qy3#4~mUZtt z7Z=y~cHJww;5**yd!8=maYxWcSUcn15`6^xVkv!uhw7d<*UjCa^byuBT-{e6;g971 z2IWxuksW$Jvidk#UzhGja{p6q0cTz4NM9PSVQ-RmWlzwn`#@pOF^@k+(fN8;`b5#j zNuA4d>A0dVnRr zU;{i-)4rH<+fx{qIOkg{`ZZtuP0=o4qxj3S{%b}5q}(onxk~ua_LU@KV0m@I{Vfmte|t6ZLYV=<;xmF8^2ZzrVgF{p`ek`dQes z$DsLN-kLGf|8}+$z-;~_{TXHrviEk11efd7%x74rv z=RE(hr_DJwdJLiut@llg&BCBrvE+!^>=~ZVp5XE;qv2YzeA(9ze1A%0PDTzS^?Z*`uJ5B>7{>}0QK8qcfcdgUW@Uy}Vh;f*f&E_&7{ zRw*6r{`T~wuD5&g?lyHS^{d7&^i50jkrhv$wD&KR_=vsxO9DI!+&ha)-C^IcH`ymXA|SN zKehGSoponz&2B&IwQOD8d)dT9?#MkY+F3WQJ$ut%pU*D(&%N0M?YDi#9i6^kJeplJ z?vL5TyYXydGXAu${C}N^F+a_2T>Vt`rc=Sk`D5bOw{<3>k7XAf{aiLtNxW%heFR%b(iuIcoGJ z#Qr2&>OZhnbc~t@yH`?=-n=Tjc`^PM@>(5$pNn>jA2yz_0ZBxnA?szk2>p^ZY-L|02gVsJT0oo6dM}6nnb~#*prz z-`E@e+5vBl`KgpXwg#sl$y# z=w@#;$v$jqeJCe3Fh4~nNPa*mAk11Z-E z$|Za$?~PYQ4%Sny&{aWXZf#qn`aAQo@aDqn#vTmOHd03MKU{TiH|3f6ZeHv!`^1y{ub|GZ5S}2-Hu~jhrq1MD z?&&;*e1A!}jDt;|Iox_o+4%RY(uMT+gH`-G*}abPEjsts&P4i=Y{Or6W)t6IT%s<0 zU2f;bg%4yGo%nio<0*`D@V~Du+Idsg1KI7t^JMa?&N`Vv3Q`o8>RHgWMD zHTG?P<&Ms}%|FdHy!ceM?y>!9>}$AhZKv;-zs@eY=dam@G-D)VA7>@VK-%Z{&5x_G zO4+8!Se5;}=7}+KS2p)(%lIMx7qVBU`9bT2N_oSS7k{%weW9>EuX$q(<5jryyoT|L zbAe;rPv~|nsf0pv}HnV zke*T-SjT7D5P)aRnj-eac3~XVH;XwF(X1!BB-+6}yxFs!D7gfgT!?%uw)NF&XsTyM zU-8`NE8d-^ufX4{ue6TV)m33xi zU;V=vx|jX|XIKB+aUA`F&*IW)O9rj$bwB^NqxHiXw08Z=2xg9qJy9?oJ#+ zXXURj=Op~mv)!8#v`_fV<-PbTJi@9$=f7DxB3KhAr{t;a-J54 z>)W=keX^<7`s6Bm|FVX2d+%Rn`m}OL^qn1PJ)ydmdh1%`scVsRt-D^?-p|nJ*a~zk zezD)Gd^m|O?CRE3it~aa=&K07$X5!znnI^R_vqWK#f#t4Td)2nxn}WOY`ywR&-EeC z^#QJX^eSab;+#pz*r9K|dRg>I+NN`A-2Qq@_cI(>2XEI3@4(;{a zqdUFN4NucI{prjBP9L68XI9y38+v>;aC3F$`V(HAd0F&Z8Fc(p-QV=tbi6Kuj#D$} zn1W_H+r7o1W1RVZJu`Ze=SENRu1`Hl`&~U*Ma-fF9oOTp(Uvceha@^hGRkG@$z90# zzf-UDm{(8k;JN&l;mFgMxR#Dg`Gjr&&oN8LnULG595 zZM79_+-ha**lT6)XtQ#5R1T;YGBj=i@JAg0M=gMHv$XTJDvg- zuy-`AI0u;dJ6>J!%eou5udWOhC^M0-yxfaYHaoxl<{SN&9 ztm^W zNA~ocTi%VSS6J3DZ*2GGsiB4$HtmP9HZSOk8*>ap7wlZk`(E+o^ne?HRre@NJN9vkXp&9QF?(2Bvj&0eJjp zgRF<f24f<%WHTdlkzjZ9)tE#Q`Ro8Cy zt*+hcTT>etYBgkGPu%u?I;0>SuN;DHMMuhh3icxE(eHPuWAe?RhAaN8w4n^&=&et? zH=PKKjC~Lo75neNxv}2_&Wn|GcW)m#+iEy>fzMuJ9=3D#ep~pr%nmhtdqHW#dE}V& z1<&oDO#IybJ(GO?A#_n+dMAR;sYd5)Md$2A=Xh!U9yCs>Kk3)S{^9!{@sHTQ-T(Rh zzx1E8|3m-C{i6b-_7?}v-T&u-^Y$SB2%-1k-wx2!>G~|Cd*f8iSn+Lt})CvAa z&JzEl;QMrz72WB}@2nb@#&`9(X?)ifrt!URVjAC_E3IRNpPr7F%eO9032>3D)(v7z z@m{I18n|C>||g^5%(c?;pkfz zj=sInfg?Nx$M`G2sRqtge3=V@)0a=M#2V811N?yJ(T8tagN;9MVr*6Ih@$G+D~eXv zeyeCr?W*^>w@VgX{{MaX%hg}MtzYrG^sjuu%>M1m2Z?%okmo!;$gg_&Aj9xMM&c)o z`kidp?uDo5^Q=`8H9HZ}XhTxA|2s z-)0zijKsGQE`K;)T?nskfLEV}S6_E{HPeoWPrbH8I#BV4V^wy%G;fsVcY4Q56Lsjv zddiTa9e*O!b+j?m)k%zVr_VZeG*DB^8sB4aex)mA@AWHtZv_YKzh!nAwm94%zXO@= zZ1#=%s3majN7?@K`Wx5mSP1PafSt0cYJ>P4*QSB-N5@3hq4>~``0-QQ)W+X zC|+=F!%*HA4V{W_cWpy~6*|5I+xG+O2er>z|4{q3RcqR{wCen5WVmAw_M9KRb##cb z6{nST@#x$X>RK`<*l_RdSq&eirh?D44ZksUhp4->;e^4{&+`Ddmw|h@VU>BG%lo{B z2hIC|ydTtX;?OjnO>bClo(ta z%Kf#!{K^B+`)k1Xdc!S(7Z@Hp85j}k3Vc5HQQ(}|i9_F@&YK!;=AG+HiJq>H;rbV0 z{OD`(qeZ{WccRH*Dql#Gx|k2{&A6>M5h&$d%}$ z9_GhbFLS1P{M6&@XH;d!b~P6NGV9^5amM13k57l3llE^J^CyTEWb(S(?}Y|eoBv#% zTMWG)@sHZE-GA)`QpL438W9{xCSXe#ww z+o1hSKTMTD>u|#jzh@o)6YX}}`}?%@=cY}zp2!??Z0DW)?@s^d#({w$JH8kw*zt|P zP$LuF{?PGGe@*Scz=ql{2JWx@M&NuzjGeg2`aY+clkdLSaDl|QFg26*eY0Vvf#bCG_s}7(E%*g_ul?DA z*LXLBap5-^7k-%f7WLoK@QTW(Y<+&2!Q0m7FPO4^%CdhUIY04~v$mOBR{7M@&HJW( zvWbHLPuuTD`8|1Ne*5(Oj5L2RjkCS?ZvJX?@2Pjz?IpGz;@;wZxANBFmq*|D>rWKN zJ#brODsv0$f!&?e(8_#6D|0{E6E+dhI^BYpy$43cHsuCC@WiQ2mY+Sn25axBQPIbr zWlxe4@}mYKFPZ~pom;co5hnKRtbN;TmHg7o1MRWrQoPr0o~GLQAnkPH6tDEt-hC#Y zDt&(JF5yvmKDZXdKK|avn`W`M^IUQUUSPhZfLtEMp4G-Vw9%Q{J}^5H?0V$~A8)ef zrtEl&>R&|ttxmn}{8WdTpGw!Kd<)T5co$`}Z`sIDmyev2v0t;I0dh&^zFJnJeecXK zG%>$J?ccIfiw{Nol?O|MfjTokLiw^+OKUVoQus>fVLkKlZ6PppetQ|u^v(|)Q*NMB zj(0v_s+=j~J|F7xj@>5VsCE@QbnWZ>)%)1n_|qZj_Gu19dxR+NuRIRTolND=+=Gkq zU#9Y>`BTl67lGSU{>;7b0IxFPX8CJ;;PdZU)+1gzyZ!C1)&HEQ&lQh&JCpuq|Kc=$ zqQ8qHbv8W~WZz6*{CFn5de_`kvCZJ+;C-#ZJFCe&b8x@b;O;!@jek=iZB!eC^OLkg zeSN03ymM`uf7O;ZJUnddP1xQ~%IrT7&e%`tJ?89AXpS$v_P-~;5qa#{L*S>ncFz7K z@1iHKog4?VPoCEPGdI{XpU7NN&wf!gdY;~MMSN^{nVjEOg4QEF_gh#O&$B(5&yeeX z@AZs)B+4dFgt}f0c)kzt{TAk2fw9;7TyS~!v=>GgJtG*D|Cv)Zu5vuzLOIQK=MkGm z4x&kPV=o-Rr&eWBp{|0V*0E0FUOn%N$U7EX;B586<@x=YaNju{Zner_{|m-tz^!79 zyx;=oK`&gMFUf@a>(k+WM`eBu+^=Q8Rhwx~5;*sI;qr`p`rdSS>vXt(rZU*S)9_6> z^abCD7cS3l%!GUBbhyN=ABi6V?u!}t?s4b~oUeP~^8AWSxNn>e_Y#$P1GpP}z5C6j z18}Bz;X;RkOt?*_!#zi3n)<-kg$urg4qU_g&huQp>){>eS)h++pZ4)>S0B$laB+W+ zu@msLe?x$0aqf){2|N3RpO)}2Fx zuA*Js$EVr%(wko{JsEkm)V@b&$_7MEe$Bo|UwYSQbUW~S>HXn$dpCO8>&_c_^V#Iz zOw7n%*538l!-n99(uU-N{)XmFfd)VJCW(&JK182mj-`8ubY5CVCeYg^?`|}|mx5mb zzmxeb;&%$a6Zl1!YL9!-g`Cn6mhx)YU;Sd}nhLWw&poqG&daVmm_;1;@lzEWAOB!l z&CBmEe|W2v(-ax=QteL5cP#Sd`q~tAr_lX6!$Q83_9nZA{20wcB?3#gX+Ho%_kcYP z(Ypx?B3HFoS+Vg8impQ5qUB>)GsW7d@W3@SEmc-@>w>vgRnW(bWoA~Py|To%`;QR@ehLd2lx^DBfzTWS2@MwoM%N`2jKH@{JC&cHsJEV>f$;8FOK6+WmW+v ztTM>N7@k?|;V_14`Sq@Uc{}^-^~%qmERC(U?dmtv?{9lRR z#f|;9&oPy~jl;iRp-YtICym7hHzRJZigSFpVWurssBRilAlSxUp$@po$#_fb?xY`$g3}JT)WI) zSH0|ox(@D(e1V#3o;|^{t|r zf4y)_w(-|R2k}G}?ejbRs92f)+3^HBA8g|Jz4O78f2uPqd;4T*{p9uYL{P_> zu}SEl1ke5|^BMH}cILCCqKy@s&mPrYc*ef~&YV7;U+0BEJ@UT<<6BPIIQOdmk1418 z^d6sVp~nWR4BuR#G7GW6=+9n08Q;5bLm6-{%7i=rbhrg7GatAYWx#dnO#=U77cTOk zx;~c)7vH#tCMpBIc`6eD?#+F`rHtq>!iDS9b&C65bRZY52d-1*gvylnfiHNvboi9D zOs2o}4bNLM;m$f8?*CGmS-^dz4}8H@bZB?sI&?UY3HKHcT!#)XtIRFHeJH~x6K)P2 z4!UsljJkGY!kyuP>y-J4%FO5kU+{G4@N);Qp@p6Udkf$7(A0VM&(5>>67(3)lJNS| ze5-q-TdJIWg+03D-mD z7;o^6$}d*=ZuU}AUOxS2%IZBnk8m;NpqtLjUSMFS*I^o5fsx?d*In3E&>7qW`!5V^ zCwCvakIi{z4mO8)R5S^`=Mxu??-5P#k@)V2?9XWK_t^LCx*utOHpUNax!?pht>{*#|O2Oc#17x1jt2=tzRjIYJ7?!)Fy=h=1G!RxVu zH(&>4XBT3l?pz+MyK8w?-FKE}*DYG^tNZSlW)I2`sSnt9^-vHLTVXNy)wCP>?|4YkXtb4b_>KaV@ z-W{Ufm+IcVhVPhHem7sgd+XkvA~^oK_kjK0WUK2Y{DLz4f-o?q@_lSs9zMXnx;*HU zy?k$7JL}mJ%XZg^xBhSZ&G!#&+~6CwW34}@@hShn9WV2Hgx{P%PU9tkfji#E2U+{t z%Ns50yv8HgybZn!cFgu!jh)yj<^*;exs=?0?vG$!m?2bD!_rPYHE>(LZR%tGS1HfD88(j0$Yp-sMSVJ1i$d3v80*ybk{5$a58eiZyz%PE|j=8jDmCuhqwr8W_ zr-h+Fo%XbETZAkwwz@i(AgfEc&iQ&na>zFtn%A0j5e=)zPmC|+8edI)+?-=|{R{fT z_CbRtzV41QTv>fL)LT})_QH%u+28%sA?fuSarjPqgDOs{vBG%AH=TB};?R8lUI|tn z%!987z=yf;uPv)t$f~dI8~B@L_?wq2LD<2A+SP=Nowb z7Bc!PWb|HmI|6T?$NOB~|8Hb$7xE?j+ew)?Wwb75Gw+|~{ZVA*i@+=d=Huvvt>}fd zl&hxP|1|Hv>l@rS*5}`m=jYe&-|;2-x$M<_8yoz?8*lK9*s;%_*SOv{V8++1o+jBkqvc&N%G#=2LfcCG{TtE<>$U;wV zCH{XQc+Cf&mQmzxGv3Og%vPQ~$~=JLH75Sgy*mfs6F-^zaE*^~hl%@}`>7^2?dDy5 zkOR)yp@vn1Im7;Qp)S#85`SOg?=Jox;BVn;p_&b^mDW7{T3OBOuZ3&A2R?rXF0qTx zsayXgt6{^HzJ>=r*L!@r&#YJK9dA;sWe@8Kk=tkqV|aYEnE~3Xd4qcHXDc4S*e+jj z0O+mqKH52zKgBMRtY7`${*N~$(Mj_2wBCKC&$?1=)p$}ccLPJT)83^TCn~N$g0hJ# zxbX$(=ZQ1eaV*nMd+clCg~#F}>uW!zUv6D6`>JXDDgL0?j^YoTeb#y7XT%>|z0h9L zTQ6{4MxXO>c6=pGFVRJDiV7>(S;#&(qK9I3L1;17p@+^Tcm{fe{MHm=cxKOu15dO5 zDaVfC`E;hs0e@8ULw0P<7u&-eUW8blnQMbK-gYa#H3r!7A0nedT}kx+?4CI|n>Jn^ z{)wk;ZhXa!PuTNq>Gq5!UPXNDp|a4I*e?Y6LC^T(ti5NBG07aG+NXQPh_vVBzbu4q z;A4S@7f01E`FYr>ySX2FPBimB>H16I^RXjBU8|UbOi;JlDg9e^eyA%%8=3!mM)QAj z!9_XQT4((Qa$7VHsQaBk{aTW_RkE?D1m+WcMyykmRp{V*C>&-&?*-uq!_ zeOC!QH-=uv@>#eL#of~=~_j|d&b-}Gyl~~!Gxx^kXaAZz=C^<|4?^gCa6}%Vu zwZ>S{wt%nmF)R2HGW#~X-BbX+1Ay<(t2`LCtfprC`rxVR^%dt^#{^S-RN~00%_GkK z8u_7Zig(SS|FjQEx%tf^W)zJ51=pVbdcUyJ7hObu^?DZaMelI>(B*d{&+tt%wo-ds zsXvbWac|ky*g|^(wcrEQI(;;ovuv!(FNr2{Dzk{G>_L(}0(YVj-x(i~D!Q<7HW1?4qgT7B<0~6@y z_@(?(pY{X4{rXQf?b2H3%E{3mJ@xUX6!L(5jw(JV`HCQ4RmfKgzpoiTRQ)$qXKZk9 z)|c^2>&p}`OQGWxBTZpb6$dr@G=QIQ9L#s}KU|%+k-8qmUy>c%t$lWalZ?(o9{bp5 z$Far0v3aJqZFcqc-)Hpg!%lzN<9}!W88gQ)vS&_3db4>L=i1O_>22oXXC{y>#UB+5 zvg7Z;Sa`$7o5ti=SI)YCoP!CbpUFi~Kl@`(eThDX2lOt&yAYi~r?hvq@odC;|k{+DfgkGb*j=q=q(gq}L9Dht{i#=mB7j;;aFGmm&Ad3g^l zttyD#KFQ>2KD2aCL3H8d5PM)}HFQqSZpg!i%C>5rM}XXq#T&n}F{{lUt!l zIeNy2e<9wZ-t9%?ZbaZ;*?O(h843+lK8cS#MrU^OPP7&8q>v+c`G9zNHMCCAFFN1F zsmF;PUo?)8^#p6X}>toh4nh_m583Ra-su z&NK8*>m%tQ=_A93mwI{B)&Qt;b6_~cgjBm$qbqMKgk{uS=`a<4w< zfSy{@+RWZmE%41@^m8jb)W&{+?eNu?;Um$lnZB^&`~@-ie8g&eF!^7F#{UX**5H4Q zXa46}zjcL6t0=T8b$lN)e|hAJOyB3g>AsJv4>&U$+@xpS`KUALhriCC=a0Plft+yd zd);C7|LGkcOr^+^dJmoXYk2D|*A2MD&VSi)q z_Z^50fgfAn3-O%&96uD#aULOaihNhxT+j8W(8JZ%-*0QK_Z9b?3$aNtNaqZ&DqRx` ztY9y;S;B$bkM4mDI(w<-><0SY8)NG^hrSWKixj)ehE_SyDmOMC+e6&Xtdn!ktdxFk z89kW&v4%8M)A!BP*}Umy&gl3PvtL&8rhKjkG7rl;>dpMVr7oMj`*ZAX!6iPr0-IhC z`wqMie;HWNsET>PtX#%&<6f$bFkh8wvhe8w`;&X1Z;Epmi0QII9(}6Z3u3-|@F#Vz zm>}Qf-PM6rffmo!Lion<5QqX4?NY&_Xkx@I0_c!S9xLS`y%QD=6_MAzBM#* z@OX=7sV4drI=MW%**f`L;8b{{=A#4b|1nNFntl?k&HNRYaxjZS9FLD@o|^{n)*?4#_VJ6<$23fMZ* z)Ln@V<~yAqK8EkS??3DNzU}TnXpC$2Wu({7(%UtzN&h=Fu zi)#ml#C{8#S>@>eDbOEZD;1^>fZaUZ>pRFU7$4a94NESe&*>|T$CL1b@>rAMP*(#s z-}MvTZTNUo%QftO5HkLN@zZX@Pg|q$B|b)yK9QeLO^$-@3j@I7nZ}rLctQ4Fezwdxymz@ zXQ7-BJJ%%;TYBHen>OPg$9d-E{g3!gZI_Rxyn?Cl@U`$_tJ-XAcM$%~GW_hzVw_L; ztH(ee;EP@wD{dy{Wcm!bh|_;=f32dw)DMaYsJ!3tt>-#10mVY(TaTi@OkZ)YzEVFW zkOjrq6hBd}g8IqOJZ#gvmmeJRM-PoP)`_H1_LrW6`@#!=6B#!;vHnvUFD#~7cnf3(fYjaA^6 zZ@@2W!!KK#YhA7ww^ydJh`;2}XX>kz88^r$moJzddtLS>n?8q+cUji@T;ZrwS#*Yuu_s#IT&5Igq2BD=dHWa$ak2g7N=#P^p=m&mv zFFBPCYNo#wNBGG*pKM~>wO!-9Ea2@VUx;z!OxpJTt>x3qJ>$)pS@f}<-8O4l*E{e8 zwq<4lILa^DN6bNRrMI`@vt4Uh#n63(-s`z!^mbxM*9Jqy(>SA3ao1~uR&h%hT?+r$ z_DOkc*|BShnTf8qLe~=N*{yo$ciB+sXT6IUT8l5lYk}CjbsukHciSm(`cV0M(mO_0 z_c;2|?U$bW(6%>}Q^a%e)~lZDeV*&RT%RVZ_zlvb^ZV#SZ#+T$mw6uXL&G_Xmsq9N zMHv5xxI&?Ib@YAazHNNdvB4g_Omj(Np_D59z0mVv<{77DV<8dBy*anY)L$?h3p>|Kqz$FDt#cK^dMLvU%&x22k z2cHJ|(!2ky@Oj^7ZTHe*j*G*d-n3BN13Yzix^+4%3d_hfR-k{8;rrBzho2 zMz>_p;{tf`O#RYY+*^*dUc}okr5>ExcxRj+{u596u}PW-QCxgiz}`2-eed?)S@BQh zl<4fw;Gdso;Q0sPAMjZ3;&BH4`G*W#{zv%-9Nu(s=!1X$-c$D&&!{bOuWT(Zl$yO3-5{TSX`X20pHqxU|}9(&lsU(=>JB~!b8 z2+XowUxttio)of`UaoGyMFdB40rFnepU3d>dhHrq1ON}r$1Tm6wP5TD%qyX z{)sUZJ`!W65xePo?e*@q^J2H1@PZFN+im9@+Np7`@cuX2XvV++{5SAs?Uvc&T5a=d zd`157JnXFS-cO&44~D`AReklp@gLHAZ9UFd$~qZ(*t9Eb`wRNs;`8TGHB{>-`)eD{H;ZrODA`wruerR%>z z?aqe|&?%meKLDR8PekiyMSG2JEo<&SFJ(WGIUkmNV&%@cz(4dm7g%HFUQo$gsokIH zcE3Q~UVkgkjNjAptND8R=)l~`cNY7}v9QNop>4C4g*FCGeY~zB;9uBGj)dk4-FKr; zm7C}E>CC{X(d+8q3zez1tSOq`kIqFWBG}>OKgbKLgLlSofR6U)Q*(Bmy7e?}c1`as316oVc;_V3Trp7)$8- z6|PmE#^~aOJE5!MC9zC?`2T>u#s81cw+%c+U-7Ns%MAK1_R#mc8T1{|4}EVp>)F%t z@;}MH%8!FTrYH|AK8OAtiEcv&!RIE=+?5mNBCqtrYkD8BlK;wjpc>^*^gM&-^h~lL zo{PY9iiK=}9$xu$<2~Z1UtMYQQv#WP37&HGgS|if>FpKI2as2L4lA^6De@XL@)}s! z;>fG}&X(7}LdmP-K=P_GYOi8sTb(v<@w9myZ8o--$Fyt<<^&@s~^J_OXi6^oJSe@vB&nt;ZMc$ z)PJLucfR!-Mx*9-Ir%aW71)JSf{X*Q#XxW!G`P_5Sg) zWu7L*R8;zLJ#sLPJ|F+rs}8nApF93I`I>uAE!z0vsp_We=f8JwWp(!Gw&d|z)_cR%)%>e~C4-MO*Bm$(0ZVEky=pt{#7|Jo<_H$4{39$mdG zXY^yybooaquew-UwSOLe)%O;iR34dIZxh%3jdvSNepj0RHAW6&PXf?4f;}n2octlK0Xv}O+HXN+;}K$UGjrT1?TPdN5q zI@gi~*9LT6VaqIwBYWiazW*V7U+4qhB=ReKE5TQLM|*XNa4qoQdVU73yBu79>F}tl zU+y#eSET&``SWf*A~Eltd_>8D{CqdJWkVJ5wE+4DIhjq2UjFN9ehcvbOYvz|;jhf$ zZx!Wsp{I;I;rFK~H;>5cZ|cxpZ#OgNl{MzY+6(!c&fhxA7wrh#yNw*;ritTTsV(sNj$L$F zeQmwvJGPZIy1V&H^0%2-cNV$S0dlFwTUq2FTRUzgC;K*Xvbpcfwz5BZf*iPakkyu< zt7{{{s@li7FX!Ft>Z;lo9$H;nHtt0Ow-k6@9(U=V8Gj0ZZ*a_>S{HxjPL8eO+jwAX z%!kb>OE@;fwWW?fa-Bnq!x{a1z2V1n|GK(SeaW1<**iqKrs%PiZ`*UYS+V7Vtw(eg z3%O)noZl`V{(#OmXr=Ge)|aqn2gr+%o?Ho>LU7P?@}s*FPt@10#AXOqfBp6aPrqgI z_ipE8^PCub;;Cbe>SvR{Xu$iB!*az%nqDjl4KUq5h$KKPh9 za>-OC5B@HF{WoW*C(%zmziVGl%;@X;=xeusJD4}}&gV;>(ea(7jODVCp&VpL`xklT zH(}-svtv0Utw&msrQPVE7aTn_A9;NOd40xdgW?a;KfBO9=pgtqyBA-&vU@hVLOkv2 zi+hIm&p%P8?y<oRDk z81PR$I;V#pP5iR=`mCwc*~}S+-6{SUHz-$+{JrQFVlmCBQ1OfSFYNVG9B15DM{I5G z8?1YUFBJp(Q{wrGHDo@2h36_``)GaO&6sai2F~8`gYb3p{jm?oQ15(y^l|5A{HGv& z$XO@P$_{NEU_C3Dn$92bx;I}O$X*8gwhYXw*-4)A9CAOqi~Pmi@QKz4Hy8P%6&$l?I8JofPJJI(~;cqI>Liq!nsi5)Ih?WWDC^C<{lUS+`KhKVr4P%};WcDuk zyCU*jBJ}rs`gDT)LVvV%GP&}+Z-ZX#{Km^eUCQ14;ZW8NGuBSS$co+8@Am=l;Csg7 z`d)eMx9?hA+fANB96D&tmd-9}n}lsYw)~LJGnHNbcXYkVyEqA-mICslCXug&Khts( zIbJjP4w`7Lx&`{EKb0F5bbiH;x^KP|x@{rX5`RE@HKf2tIdc|EM&I>Po*~;19hV6ctvHfk}k&W*-jQ6e3-<9eB`Z6*c{ZpzD%f2vqI@UPG~MEy37+!FPe(^qPm?8jQV3^IM#L z(7L-6ay$3^<%g2Qkd$*#Y1T%KiT>SKd;Ve`{Trd2*2uJu6n@Ghw)ajdBZg5D6SOwnk|7EtEYX2FdudD3!A`{0+U#Is+)*6o6papXH>{@=G z7)x%=EzoTid|i${g|GFz4ce}w?Cr=~N4HhHp5MbCTg9ydtnI3|9$60&Uv8eyIuH0~ zrL&9JGQ%{TXlts zJCTVtcuD$NdwnQxEGa#W?n6&I@N#0O3eOf6IYsD0!Rn^Z@7%SPeav$^XMTA_C7t z&o*Scho+)2bd~OH_R#eO(G}g6rfYWW@w0}NA5+en==2Wf)7-gx4RH+O4$xKhxjk%k zHDjl2y5>UHtk}nr2`^p2wHcpP_F4G0hkN6kEuHDmRcntPgRTdeYrhD(Dt@5cIQGS3 z{>G-OhQu+|D~X*;N5fvc#rh~uLYi8E&mdlxH2IbC)=2veo+6W6w-Jb~^s5n9<|t&z;D~edwYbXw=R*y_<8b zH}i(=dONpppNU=VM5eCCue?}#jhN*mAA7~Q6Wqdu!{QkK6iZ(v>>I^$p)veU}^ zO*{6ug?N&kW0()0YmGr3e4f?Wj=t!yvYPI&teO(vfX)c_t3T=9E?+W^PD>&ecK_J1 z<(n@*#T@#8*f#oDviS_ZvRAr(nrrch+lT76n?3#Z4<7q``#G{Tr^ol?bI5O!&)R}M zXhRpYqh~wN3y1M>8{z4v=mTuhcF};~>4+M5z7?IE)|(arDS8=Ho9xz8?pN z3(+G_A>V7jsfPOn;3S_wvOW!*GV%G#e(=e}MYc@!3zr4pqOr$1a9NXq%RhmOY{eP} zml|*h6Jt!(TbftK4qL_Y-Pxm){`SZADTf&v7@F|>CGM5apuGvE0wWGhX43}wYTDm1 zOg>{~8!n*@`bIg+(qHqEu! zq>G@5;`>#i3$R_fROp`P33x?gqP1GLRKz{)7B5wCZ)Cy~%kCKmIXX-0GrenZI^gNU z5nr)nqDJeabAXrMTbCK1lX}~zOKWmEsPAy3w0JUdV>$o4{@U$kKfbiT=Jn0w6SOWu zeip+&TLQ%$&`q)J!^StygC_>ie_62~&{o;9)#y9%@mg$RJ9Z^6E7Wx{v=d(s%wwM$ z-VYuW>bjINU#843U#R%lPOCVFSVvYKW3oYMKPJn^Iln)%ivOFraPemVSnB)tf2MCj z#rauQ*F(_%e){b}55MezU)uTRV!o-OUVWqT>*>48D9_#iU0?E)7k~X1<;m?baX6Kq zNZG`eaB<;q>&gOjQED-Cj*y!{>@LojAbz8G#o2F>Iuu{nfXGZw!y4=Ye z({)aaU>EgE7q+1T6~pj*Vg%czdnqqmgon{@@J(9@vek!7M4{CLXywYpd}uX}zLYFz zOc`Iq{Iz%;c!_e_$Ng09T{_Bts8HMmIbKr^F3=Mt8?_A5Z#(vsMA%X604O=E( zes>=Hurc6u_!jU)mOE+~&l{Og z3_LgXK6+o}GjSAtZOBST4LTM5)|NsaxA8r6noobWiDvo>^Y;eruQ<=pt}UD$t!KPC z8``adcK1QM0?sW_jIo9KL^sa#jZLJT9UIZL<-}~V>~U^Do z)*WxjN6d@WpNWRx*^a(Dyp}QM4Dda~SokeqK(lqwv0e1zkC@i8YD)n8%i*-8)6^Q4R4uTF)EWR>8X3?l8}Ptmo%6w1l&m_hvucdURkX z&r-R>yM0|B7v)7ODlE-azA?Ds!ibC%=az8Zv}oExU>Vm*(tw|^2BhTz1?ZkAo!=R^6DqG)Ach_ z_!)-;BZD?U-~|2Fvzq6rhd=feGADp+$tPrvMKO=AR_fn7!n#s^zVwRt$<|e&ZR~e9 zB7y#q{c1rMWev8TZQkfBZbcuxgRa=k7_R{P{|+{xQsX`B)hpDs&&Y3%vH1hI?nH(q zyX~{k1rPQwyWjZ&dXM@{|I;7M8%s^Ss;}cg;LL))v|Ts~SGBiCcBk}~-C1W1D?g@~ zkzf^YX6nM-Y1wVN3E2%JcU7+bvt{=g?3JN)Ke)Et1V4nqwTg4-v3EJPy?e-=pIk?d z!^HufLz@XX?8Sn;QyiV~pqbx_BPZhbIKI+$t$&9OyNJ`v9%zm%!niZ+l+`oiV?(d6 zs4VwpJVJYJU`@Z5mT_!KCH$bch315&0jmrgRId*_ojBP;bFj;gVwe94Jja5kiN}Md z^3|%r!^d~T=MVb$PX5VUV9gU=z>*FOI`EXQP5iwF-VMMj^1xF*mTlkj%$_Ab%YCu- zwoSph7UWUuL959{bb0qaY=L|#VlrK`Bv&U?`$@c-xSAO6z3m_9%kYOk319+h?Xh+sdpus`|=dWLzahvsPR zoA>`EzRF|WYPhU;ORhD=@C@xqQMLdYY^Lu5=#0b+_yb+3aie1DipOhh=lPO}utjXZ z9@{+1I3rU&Qlp8HOy>Kr{0a-Y>HILepXBq5b;dRBcW<3m-XBYiw);hU*iBi74t9~h>Et9DC8s-?$n_Bzr)pJA1c^U7~b6M!Q1hkO8 zMAwcmFrY;eT3^qeCc3xvV9?&X3HtWB&*A%gczuY~rSZ*7t=IO**(J!RkyDNh{rCQDb<@o6{CfL$!Io(mxizrY$4WWXNt6~x}c1{;}m z=lOGNU76?LlJekS{Ex7`Kil71y3%WocH#_VZ0_Iz3EgV;@UQcs=i$!TDalil*umCx^lR!>2XozU}9+8>8z z@ZW*LpI;pvMg4kT2w(I(f9mR}eB_HvF*@TJ?hS?A1|YV;vo**oZ50 ziH+!Z1JG;_1!s^f-9B8lD!7`{U^q)a&KxvGBC| zC`JA9BTdc^{nK1SJ0>wMW9@?YVhwz;mt2&W;fssliw^kW3)GX?BcF{k8!JX~K5QU% z@dF=kTFDyL!pfp(;-|oPjz9T(_%cgBKIOc-Ae#6Y&v){-hj-LpMmx(kd^K9ce&5Yi zS@*57b+*`XT>t!U?hJ5C;EfAtFDVes%w<=Nyjp0jD3 z6`KGqYM10!dH$v7JmrOnpT=qpIQ53$)kJ}>_)$Oh5#P6*I71Ac30VFnwI{XK_Q}hM zqv9jX)V;=`^{n6CO*zrt=mg?Qy8@ht%X{&HctY@WZT7vH0*&d{aTaSqtzbGuYg+Ji@YIf5;MM8y?2XjReJS!^ zh>RwY?<0+;HpQRd-6G0Vp%-sqo{8^Me52SL`!ym1&fYHKGupcuT$(9Y#eBYrVJ!lW zD*5Wxv9Ga~xIp8;Da7jo(lbgC`d`4X}oVvEq#)ZJE z$f%3{F?DTB*Yz)|tDd^%Q`b^YT{nG3T{k&()l=7e;4Mwpr8cZl_^9ZS-6xZjD+33$@XonLCxJhShXtNu;YUrzmtseigtKYi`O z>9g-BXX@WX9p%)qI9_tSLMYQF|;x{K3kr=Z2TYE*7XXu6wD|h`wbi+h+!~O-g zpWJ8V)s&+fEabKg90s5pl;2ROz5PWO_(}T&X)KWBx0jDmWIeBO(C04d?T4p|Dxc3I zSH&Iw+-LSz?7bhI_EJtvhL%&(c!$XEEM>2_IJ!oABgWCG32Z}BwqbIpt9c4`68_V8 z`ucMjPf*4kPsnF-#}f~ehasP-neR0puJJ^cv!|lHH(eIhKpoKCVBeIjRo#NjxGMV^108fw>Z20o6>*(F>hTG9S#k5KmEYL zm2Uq9R+u#hXX-b*eJC_u_MhuT3n+p;er72E6ec#ZCL#&*u7!@lt$(_)k2&Y}5^m|55dLMMD&A>}^`Cs5Be2!Jw`PBD)USr7K`8lZwF;#pW#aWwJ zA2V(U^H%er70>7AS}?2->(`$taoy{EqiH0 zA{+t^HUYn(Wg#|U5x)Oo;<8I<8}igmEO7znJn1{dd!!@U&|_|YH;<(5(ZGjxZSYh3 z1gq<-7c$;}HpUkLmg13jAsezivdvGj4wCixO}Fwljd&zEhfOb^TDx9z#5um4nl|ca zrj8c$t#D!wlqTYGO)s5VyrRX9!q6_`1KGqm_m13~i*>w3}>~DLO^~IFm zQ$juCpQ*0Jx9ki7U$Mz%^ke)+tLulfSGMqwavX1j#wFB&z7kGGuBz-9P>_x2w1 z+02n@E-8n(e(g1{*kB&Qj)fXR}_F9{wG7*eDOdb~aVxP@=8TQP|j(%^TGbC^#XOM4ZEBzJqJC0~o{ zyNn(PZPUD3j~`oQnHU3kCvTQRtEVXwWc|{u{Otr@h_RIP`aa~y$FsToX{|H!z&&v1 zo&m0C_9f_aCYnfZyf&_Xz43x$*Ij#V@2B2-y>JRXX`97fM_QY~U-M0loxHTSot#eF zTW^70ly8~g*vd(1{``0Z@{JFw@v`RlT4tcPZb5I&LWg7P#5e26+0hyo#bBftWyAWj zlbw<);0R^`bX$e*pF;mhzN+pSb5eOA^BrGA{Utj2?RoWHw(38?%iw@+%IW2Y*uJdz zvc%!_BKlJMiy2-AXZ4?WJ-vU0{mrkw_{pZ1`I8*|2fx4I?@ayV+J5!doHO*7tLHhm zw@-iUzCPyn6nAQ+&)SJ&m0P);0b-HO)9ELkYn_s8tJb}1-bm%z=(`1sn|$Q=NnRDx z(OPxcS@{L>2ef9X723DahswPd-z%Qi4g90@!@heKoMc`9vjaFEy__`~MaWeF>m1hc zPX5hd`dIL%S-DMZf$XM2d>5bPt4TP0-Z}*QhSUF!9_>Jn*3;j@Lw*tcVfNxpp;x81 z)!*uaesP2^eg(~JePs7hIentF7Ft*FH>R%!#O^S^xv>Rj*YJ9u&}|WT0proGibLhWLB?l$0owlu{ZxYOF2%1++YLLvWH@q=S_nNC(Z7r7-zDU* zRFT87mOQH(&U}54b6PjT2gpf+KgHk`TO6C=)A`O8j`#{b()c$H-^r%?$VD1!{5yZ_ zg!%^_Gx1Hzo(*p9+H>LdF1+$GJo&EE9^w8TZ4w;WDamFt`xvyck3nWTwf;_Z@ATz# zk}p5=UCyd~WABQY3HUmJTqeOG%zCKi8T1Pm`H0YN3DR z$G6f)t;k9n{nR#^`X6LHEoUcd-2SHE3tpIZBg6PCHAUbhJt;Y6tlGJcHcFoN+%w+T zpatl(0*l-f$NoGG-^#8v^2>U@XLtJio$cuAojzaZ0)C~d)jsL!ur&z3XplLVZ!0qX z7-#b-hTaUG^377?K9b%R`~~=Dg5M6W76EUnff;NP%$>eK zXJps?wSw&f_KVnMtuKDv;Tac};Ml&%ASbrujtvII7AZH0{I9>|S9@A4bnfqV?&q6( zJI`Y#*E7t$_6ZmqTVTozj{OzCf5~4jpQbW9I(xv`>3iu9vqt4oXn`CYK?aV(Q_?B# zu>SUJbc$K8aw#%Oe;yGH;idL|*5@^t_*3uwEMy1D$y1NekJA5Md!T-m&#QioJN>8e zqWG$r{%ZmM=vex166=ZJixfOGCsPJH8zQ~!xAEKH(C>+}q~8^1Nxx%N&iV8c`p)l5ZGqQ* zj1FU->l)6aynS+**h!wTUER6XG0pR7KMcF??YyWCc%?qLwsX%oYi2CCwrLW&;YeUW z)19AeSbrzyLEq1_Zr(?VSZ`U5Y{Bb?rz2+-$lD_HYKRoQenklWO`i1QCWS-<|skC|`v4XDWpRyHN-Y~N&0fxTY6 z9a%PZyq-BspVf4+Wi{o)|1_%W2(lnNk8Uz&qSvG<(Cz5x7IeG4&kpN7Rc|>qn0gDd z2Gl4I)5sO}p9sXEk|4en*YWw_9I?7W!2wNXkDq|m6lQK^&Z(8_Q^21i-)bpkkgpD8 zD~>!jt3KqbMRg)yt<>9!Y`1N}h9KV^MbM`aT{6My`gx%vd^e-(`Cj^7ZIjNB|ETel=B!e{eH^$NL#1fnoM51*eGy`K%@NTu*tv)J zgxaRqPsj>3X>3r$Z|YKTr=6T}cc_}URtxQvK51S|+rTU3w6m4|X$7A)@NA=-3O^aUra*N{D0>1`|a(d-er*lFvfPFp8CZ3Twfs{E)Fur+6F+Do0Lz0|4p zHa`cgfqjQy12YcZ;*}25#=Oo5@jSKfa6NPDzPzU7EURmhmDkjHi_Iq!;p>U;%8^Z> zu1a5CO?)kNLH}mpn|%(zPjgW1;FTbbH4eTJ{MMgB!`$r7BE|*dac59=K7OoZF z*w?=FP<=4F^YDCf!*^HLUi1m^4(cc(-jS-Zx}Iblps`#T^-Di!oyvbpKL9&bMg2Dc zYldBq;Og6Ca7|NkK6Hk^4=*Cu3YcSn7q79pF5+D~a;G&zv$W?C_$7cLzo#All)V{h z+JTeh%V}z-jv&8^Z7QES=M&H563vQd6JaSqWzYvleEK3{3&(Cn|==4Hohm0@7$kF9H00| z28`Y@oFFlrc}~1iytxj#dFxyN56$PV4gcZ>6W7S=e2REo$C&-BE{T3 zMFC^PJ!9>$U`H^gremSibuqTN2AkW7+*D-cb>6XieXZgknmbA23+IvtR?Zr=-F@`I zWey#C=zmR`{)g!U(O>H@%d|cb8MVhA>cauCvHg9E3~K$Q%EuSc2b9$uNG<-9<}wn< zpm;>Sglt$zFwj{w8XgYHXAd^r$~h16+2iO`jpfGkTzL$SBEyQiYBD&+RAgpr1sGo#3~H zxr`3rmN5QH304JbrQjRwX*7`hjHB@BlfFPx`*d>g_-#XXNxyvuJolmB-i4p{Vn5}R znKKcwmCB**G`UXXFms;;@5HfviJPphE28eOj}aB!-a+42%5|eEP+GesA0to3hkTok`WXls+p* z?r%iymmrH__-;D;X=cDQZ3Wha_`I#)=cSLAHnQ)gEN$%UqmAzaL%xb=lA;~*g+xQq zBpw)Q)|G!9n-RuNt;To$vSl?L!xtD%U*7#`bp74FVKo)j&`!nSMGxkX*>_K8k_|QM zMJkAEzku#fRA| z?7Q&n}17eQf?N1;_p9Ej*{a`eT!?Yv-ujovrkqW`crc&dlShDAbDmEgx#qh4%9%pv_>fV>3IoWg{PrSrhvGGEp4ds+ zq`5K01C9Q;6kdcs>Y;1LF!Tq!@i;JEUY1?AciG;$_F4Fh=+6dXaUJNj3$rBe*k5!@ zo2w&>;H7JjEnu#{ebUJs=Jo|s-|8FHqw<1n>edXrRg3_C`=uah?2iT?lk@&R?< zSiZlmz&Eh7f%x2)?!Dw>K~_#DW2u=((2ZZb{gRXA!JN(_?z{P2hy8EE)=B>#xdxih z*07b+G*5nQAkbtH8ysU*}8QRq=S3 zc}n<9Yu@%C8#}S(Mf|ov`w-8o*juSt_A55067%d&fiat29=LQ%@XP$+yaN z>jF0VxAQ)1R1*4Wu4b-c>(EQMCZSyte5EIv!8s1BB(R!+B_A+R#53p`hxhU?uys6g zXZ(Xw@Mm<^8SIvyxEP-DMLtKxY{B+3Y(rbE_ zK#MA9LL6f>XC>TrO&FfaHnzW>IK}{M{z~|&JQ(aOVy)wN*H@_u7{7LniD^`K(nm8} zk(rhnV6C-bDt}hv_&9ZFJ?#JH+twQBy_TGgU}clWK<)5a2YNCOd#ClrVeDOycZsET z-v%_cMkZUP0JGfcYQF`ZLl?e)ex1NyF5hd9(+i{@;Sk#%~fo<}{fzf8oaza$@2Uz%%iPS+oV7Qhd?GBg9kiQw#i~ zctksVZ^r4gLGj{Ncyrfy@}4$am6i1W?W7k z;#1|(Xk2c_?ZB%Bo?_^`g4x7LY&u1dWtT3=>F5_|BY#5U+B|4v#>%wCjBV!w3p|>^ zLHieqPD`MZU=g{o5G5AfMF*tC|7(C*P!Ap9N!AHm*06P_-&TD{G zV%HPh&16T7CW(>YC)HPqQ_)R#^Q}e+s3~tkH+@k-zo) zMftPW1O#IbaJ`zp&yw8{tnKg5#(mz2-yRFt-a* zw)2krykooOO;%S+osENZ#_dz2Gx9nc8HamyNP2vyGu;bf-@zWZxcmiHCV>>3K~{EW_U^TeA#FP{mwx^s?Rlr4 ze%|cTN%H6Pb1QwVoXl2a{{?5x)}^cFOEzYHGX=SoEKh;9=r@%q!>3Q`Is8*j+z`KI zreeDCT?@Fsi2HcC_M7WHZfrppC^p*$@3iyoI{ZJ)|7+}cujIX4^`Os$FJ%Q^W#3`^ z_(gmv#^?5U5#CKwM)~gY)wQO#-026!RxWnOi?n$sLU z;BR=6_ME4(o_jOKLT8$oDP=!rU{RiDW(>9%*__UM?8qT#?HyaSLL)D%0N+Y4p5|XI z33dI=b&M;hU$hh*MXNS=w%egq3R*qp#6Y^C(<|r>driBu&bdcEa$@5Rzl#=Gv9Y=a z-bK#u1?C<8zWLSJ;aRbB&2``M(U)utqWov+_kU!qet-J?KJWSt`rEsHRDP&SpK@rj zix^y3XJg`rYJPjRbSAWsUU<~9wug9US;L~$_mO|k7#}-Z9cw=cN%9I+v}`x zP&`rX`)6{ERQFT=hqrfuuc|up{`bkrJzNVG6l)TKT*MZw7i5^fNeF_1)=uFa+o7ER zK>^XKrL9`D2?Rt59Zhjc3w;|jD&$bBB z#z5!?&s%Xv1iG}*p?3c8?$-=@Qp8T~SI)0=Y~ zmEN0$jAQHEoa0}^tHD{&2R*O6qnyo|N~*G^8` zQ#gE@nK#NmS37Ov@e|R9#lxpHXP}d%`#Su-=8lv0dVm}E*WMj(oS|`O(T_v9R_xvp z%u~lB6Tn``+);ZtA_d?FomQ#2Bf3m-CtV+!&bb!iHOlMvIGgO?FwwU55nbo zQ*C*9Nb=CB94vfVAMJPXEFb(s{+~{_C{`l=UU^>M`lU6OJ)92TTh5=o9^mXZVBYh| zz+t`fKkypG2FhPNE&rE%X8zyhjosH~GiGdr{LxeLgrPJ1;*?Jf{4}mC{BLj@&;PRf zQRF5FFT4EzA-ENtWm}X#)WiSISm6C;c;Ltl@S1pzpxkW52wU)BXJ~t|1q(0R0iBUJ$tnVGGPW zF$cbfmseS7RZm(O%`5RsR`Or#jG6Gi))~vJj4Cg$H^J+gH>Shuib;gf(Q&-%j6acg zd;CIRi?e-B`L8%UD?H9ROJ05V+|%>wH=Oy7@|o+*dECAGsbmi$(6kkNbffPSJC&R$ zcB=pKo29d=cwfevmCmUg$#05Fsh)JA>dNkQLr?UW<`{AG^oh4*_uvod$QEo_D>#!5 z6J518Ff5%sq?exR`zZHMC$EfS%-EbU_4oqFu{$2g-T#U!i!RKStS0C}6A zI^BYX#I2vz9JANox==C$Zz;B{JtGlxvC6iC!y0%hl!T{z!YA?`K^M1RE0kAc;*#K1 zyd~Szvj@n-Tf*O8Gfz_&p3yp-&Z3ka$i8P>*EsZm&ZdhRT@ zdFHO#HHE%ZqX)!8!vD3{1?h?={4(XB%kOC=XSem_16-Axj$O!Pzg${#3wcG{J63fe z-^;CZ!`tF%t+fQ<2l0X(tMUgXh#%ks@pcp*m%iz1pG-}UzH$7R5}%3v3J0%qr-$^- zTm0W)o)vdZ4LslToDq1|dHxmuKaI|raR$C|bxvRWlVyCJ`17kGv*3wRn|}h14%In} z9mESV!GUz6o@HA^KgDGfqmgZKZA}Ee+{gFvl$VZFT&V^9qqfCs9pXD|PZT+n-N`3D z*DJ%0|1aJ_zeHw%e`K-+-K0AQbk?3(hZJq#J=vkb^kZyL9RFnmzIHZkKws?_)cM2W z!+!US?^*&M)^M-N?eJke|Feko-geIt>`{*4$uF}0yVln~Q0~~H%c!S(fCzk8!+-H( z2|mjz_$k8rP%U>Aab`%>b-vVQ?VBJDFgulJt;K5{z7d~A`ck@6wplzW-89m~RQ!Rp z(kIeQg4N;0v$uV=Un;tdhQG8HsWr#~!*kl}m>n4FyuZM_(>hE>Ajf&m>UozINH^s& z0z>)#Y4~*0u+!@|mrwu2(Q~#xo@TGdOjUkPMtpwGA$aKoJj9up9h_O5;>XBNY|w=-{7S?N#VB~0ugk)NN*(Auzlh9P`rY_suAO6)c@o@lo% z4*Yz zC)!G#6BI>mmBZ-T%69QYJ^VO_K9J8ITj`9&y9akMcaeI>**U@&+c(UnTMK8IiiT~R zYZ}O~UMMBMe=TE_P7q&8f9&U+>h^)YnCP-RiJtfdV`$^dRa0je=c#MljK765SX(oS zW9Rc3R%NK5B>q#bMDo|i{TsS zG;u$?bQ9|ZZme-1vBnML$#`+0F^XSY+P0%p3xG>}l}smIAbL4tbjPQ$w=-UMj4sWk z6IaMDpX?o9YM}pFwZWY=lUXmp&SrCeyZk_%buQmD_!7Kulf8Z=pXg@s!b|V~aE0*! z4gT@f`hssCviZh6w{~66eL-u{vorGF+;G7EgXf<1^=me`ARlW6`B=i=K|Y809OiR` z&rv?d_<%mT z;ZAbc?`F+q2{^?sJKBV-;g9Vrm448jthx^)g}C8BbaubM6wYCdlorKmxxdb=P2!8S zZX@mt9>UKv1M8v9MrgAK+N^_y z55oiR*gO!bq>pC$DyJWQPs#(-@8JR3@$$e_mj`ICy^?;)>8DvdkOB|*;&>oR-!31t z0E2iydZW8|jWx9iuvw`)00cQj2|IM5YtaSzXMw~g0Jf;LbME{tz zvNGCSNZasIbT7OUfp-c;C;NOHou{-_`2%;u!*eKGDmW;M%(p;i!)M|f=-;sn+LuB5 zh2jxAH(KzijUQ?)`VQKdNgFqwrj5I4V-9VUp4!Ikw4wOD*0+`SAzHT8%g?dzFHo+k z_H$_crx3gQ7(9cI^sF})Tas(2d*M64JENF2PYfZ#iZY9o=%N*UJU+v4#I*#Ju52KG%zNg=r)CJa-8^NFHUw6`> zdxiHqD2JYs-oGY|n5GA38gF+bmHVyvbPuiT-inV|%cqR*`k;e^yM627TkzM(_x<2= z|2oP2l%X&;d6;CC&Kk`C7kW|7i#Q<;C~T>spJ>Sc`60i*8tcuCMB9^opIom>t+f{)zDR zX=PsWelz=}KdroH(zm-7q5nqS{r3Flp{teOk}&^iLl%x&vDR*TzVP%vYTn5@A^#UK zZ+I}xy0Q>|xq^9ShwcoZT+0c&PV^%y7Cr>sXaoHm%m3J_&Lg%cS$$mj70}j48=+<7 zdqT_bW#oBk-E<;$3OaAJtU(_iy1MS;GHdri@TB{bbQZtvs}jzfxlf$@+?m>k@8x&x zuQR`LzDLn7LFoTO;0eKZVfroN%%V}yFjytn1eahUR#{{nsUjxa_AtBxoYGy39UXQn zI!*B;%_+DK8~MpH{&fesM0@OoBcDB|h=X@3@Ot}DJayk{_!0T)K!AKP+6$Vs=&e*yEt$yWSpB>Kfl<4YbSNb09OOB#OV^oRiJTEKiRm>Pcp8Kw>UE$ zxW)t5#lRKU9yn_QXYJt3duI2(zmg+Qy?kgf(xHWW-nQTAV+?HyM&c6jZ35$Q>}wnK z6JZ?tLHFfHS+n=z{t0wO2X!vO*C{=x=&zgLaou-yYyKM>X7En2EPc;rKV}Hu|6=^3 zFJ|E*oqI#)4V3-m$J6#`?y+x{Z4-lsiehu|4Z>~qy?guTczN26$^Fo;_!0AI`rS7R zTS5E#(a{Hx)v&Q!MbFIy_nqC~3)wt)8T%iQRp|$<0myI1ciVlvuWOt9w+?a*?!R0D*eoH)6g5H`8AC_CODE$9#(4Kr~!;@vuz82a)4PGYrc@OOi!0W^G zze(q4vp1laGfYj6GB%;M5E-If#X=(R;C%Sc5AQ{mo?8B=D!Y{VFMPWd-4ZNEJ|^2S z;1+C5y=CW$)q+!Z%}KG7`kpv&nzQcY>bJDObm$*Mhqj+#7cX8Kf z@dSJ}7Cy80gXkkHTWxz523^D@QlfsM-O$(DT=J%~rV6*RG5Nzpz zjhagUbI-gJ*rvJuqQ=bFy}t+FXRLP=#lqluZIZeZRrekGTteN233cJA5WJ;z>-&?` z9ecWd-=Xdj>MksbeLG3rbE$h6JPJSJU*XK=VbMc)B9GPh3jG2vE02b;hd7_H-aOlR zIRD9Wk;yH+K>Je;=QpHRzs32RpT?(Sle4Fk3j=*C=hPysq< z3^rhVQS4~qe4!-EzVGlF#ZPnM<_nQ!KKwl5aO;V~t;0W9?Tdx)_Tgs{i>vanc9h+z zy*gfh-kXQnOGjT|cb5E3=KXgwr(X>}RiPi&p&!;0bK6)PE5*k3w{7erwok42b86tw z11Hz_RY&(MM;(7wZI=>1S&Sbv7CG5Z9`0ISW^*QbM83Gr>&m=8b7Ozzg!#m7*52QL zV<+16{nu9D~2Z4qr7qu)wf_;`Roc}dOmX( zj*)HjYpB#`)^=?l*dMq<`igp?QfMpvhpg)SM9H^uC0toh3^2m@qKrvxX^kf;yfL;A zV-vo5;Y()>#Cd$QJ%_tC8H-}!$`{$sI*Dj>tTNumyarj4kEuLocU~CPc%V@$xM>47 z?Z9Suj(6g@1N=W|^rv*wfWS|g3k25z&wAk62+eC~s}`Bs%$Ul+ZzZs=My3+mqaU?r zbSOOefZC~{om$%2EZ*U61K|Sw#9b#<2jR6tz<>DOZ+Gp+e>>WLK-Ii-_S|H1M&NwL zu=uX7L+o>lrqN#({Q*ZCaI^!5S7zM)v~O-PeI5ihwHv(mOI-(PvoQ@HD2ua-78b=C z;CtDr&J*3U58e&EfO#+T9GlmO=_5Zn^FVXkdG_2kj>jH?7Lp^`?}H9+WmDeFX{Yz{ zR=LX2wkbDVyo7$_cf}9FRg4?@9A2+^^`Oo{FE3}^d7xR(^8Zz9syYu)F-(8pUFk{M zF?&m^@DXTBdn9uyFF)Pr7I^6kz>o{fjY(ksreGG^%h+}2If0D_84nLyzz?*I32L}GS-_|jq8^NLU4En_IHhsR;*}Z-f_h?FY zM8JpUAmoJ(4f3p9h_K!RcaOcLZX390r|z+ReYg8tPrGVUZ7YYztG{~N^#xk_y^V71 z^ylT#S={p}d}!a2=AJ)-b`RhWEyTB9$lO}zP_-L<*E=V-V{17*YvY=y`^eSP_zyDv zFnZ%o{GvhGgPM=P@0Sc1)SR9|JOLTXM23Xpec<;1vUobY{v5m>EM;#nKGDIO$i2oV z5}&k`_0m-IJ-nK8hN102LnA97J`6He=`#63fy4MhrNrg&hjLTWjF0-%@$MH6E`zV` z2Hu-|1}6DI_8m&PE7pe3*4Ns{n{MJc+MoUXM0$JaM?H6+gBQ;|y5J^c=5FY|j5{#Y z4z^8kSQCT&H0>}pmw&$YS>Qo$hR~a`4Mnsg{0eSYUmgYTE?$LG;Z}HEWqc>wPx=ly zQ@MX*oF#|7@n`EB?VG?{<+7sKcE?9jd`0;)<2a`!F~2-=8F`-PTh%+*Z~4HnX;nLT zZai7CuC(6+CinMoVk+N&?^P~3(K7QB>DJ{*^wgIv>^J>JC&Igv!MoA3PCnKMCm(BA zuZ|s^SbhZf9fqC2o@sxO=GmG38{W&jXH1vbcamovSwg>C$Ti61-i3DdK&;HOc0b5| zRN-~eqDAD+kSoj_Mr&oRoa*jiwRaBhB*SNw+vMb$p99SnEn09@X8x-i>hi4WdgLUN z^}04IlQ|3b9-|AMM($Ul3-%J%+(O^V8*y{ObOv1p`CsPF2YCL7a#$Fj_M;p$&l!RJ z{6E3xbi5(|c+J4Rd0@3Cre^08*!jKJnLFF!^Lv%stGw1Vz#L$1x`g$~C22DQOS1S3 z=9AB-n9p*3Z+LG%>y=f$oaS54&8`yubT&uFQe+W5q1di+u?xxb7)d^o&c|uP7gGB# zlY{UipBeO%&6yxu`M;!J4tMBecZL#lrPk0tWPW1%TD@^~y+6!*<<<81K1-z=Q;aQ; z?v}iTkT>!ZgkJ?{;lR@vdNCj*m_mutk_?c3zz z{%H>mBKQi*Z`NMUR@!c(?RMnxq7LTH;7)P8xbmLGl2KgZxno#(GPO9pw7ZT>L<$wC?R%Bg4O zEOEL#EATz=K}_=z-8-lMx_3_hb?=-Pzn2kv_ri9zv1`fEZwB-&M~U;yFER#Op6vb7 z9al<+4(Z94k6;_4<>*ChQ)@x$r10I)3K@76zJH3mh{ssJEj(-;5xySyvUO!W`!R_f z&aOw7DTY-));dBQYIc1QI`vX=4rohl&NFwEqz3Z1qa=desm!)sY$RSagFDq;;r+3t_Ipk{Cx($E4DjQ?a6l09$KRpSk}nm-otxkbj-~kk8yn<+gKOzluNYV}4+M`(6BR9sTv&{U&cR zH`uv%b^8-BIIwH5=5n>fJhM&4v!}S=b1!fXA#mqIVL<<-%gt z7SRWa!`VI?u^@1$IaHLk#gFn^ym(x5A#zUaFf;3o4ffe2X0A4L#{KCVWqV4AtJlHj zTJKc|+>H;w7!q?hhSKhr=<~ph%=geSM`q$y#e^>)KO)zH_aNu39Ky$sQoeeur72uD`{)>p|Xs{+qtt|A+7L?axW={cS98?0v*KqPf;Nf3Zv*(f{+`wM=>O z#|OiTo~utu@0^0ZXTDRtXZDbd-OTIe;rkq%;G1NhXE>O7O@I7b&m085yv;u(5JVRl z{sNB%-_S~Y5r^*y#gh&}7vWU-JqNH~N5F-0e3UCdu6T72-elkXFyG)_`iGb=%C5O| zQu}Jx_{q>+Z61d9UYgC3zC1dGxoKwSJjSH&Wgks}r|flY%}Z~Zb7FlE`h1#pZ5w9a zkMybPIb-mVYjX!WQ9SV&^Z>4B=ao&3>HqwNQ)Aa~7n5XMdd&Pb$N8;vYAgpji4Kx+ z=_|!6H4m|2a?Zys@te6;Zi+oWlWmq=%B8O8d<45&==?T@dF_MuF6>&w+^sOBf9DqX zWOYhfa}B>oH?7*Bz9c`r{n+%kZI0bf2pJMjtmWQ_%#@-Zzs?@|9h~tp=#%1g=&>UO zgDW=(mW$CF##cgqyu8#imo;;tA%?eZfR`kXqO*}d#wh*}ueiMM2YBSfpp&GA z792Z6tTtkyOD+g#|9YQukVw&_D?R7=x#x>KG*b35{(O-Y*u+w7Vtbx7`!Vt$BmUG$ z(bdxDz>fW_u;%}Qm}&DAD|LMSYTCqX%a2p4CyypzB zkHHIbWaW7l5;s&%Vu980UK#C+=T_!5ZOCPw(Sn}Wc|Q0sM^>`;UB18Om~F_s+G#iT z^as!BF4Q!1|C5=G8}weh*q?XA2id1Pyn4eA*r$uk`?w=5#BVBN_>1^!gTM6c{QFWi z%2z5RwjkT1oDTWhih%~{OZ=le6??rTH83NE*ff532z%n@e#myzD6WJ|b;BEip#^?u z<-c^L$(h6kfrrC(ygD_|a1uPggJI-W^`VbhpHV(~a3SrHyS?_=$_-ivt7jim1G3Ws z&qc5c@>N%uwOr-2^$(n@TxjaJ?L{VI%gW$y?2F28S|C=zPp}$PW*@2PhfWz zEXun|RT4y za+PxGUnhQ^zUZQ>wo>;tE30|Qg>P;c=kryMV_$y_?cBoLGf3NMi^g2_>Zm5gwU0an z4=RWK7JQ^vX}@0WGd}t2TWCL%_R|*MO#7Uh$^2CIPj<}Cwe$sEq&?jcuJK=|{4K_- z`{HKsoV{rLRc|=`OrvkX5yCgo8Fpn|-Lv1IztWg>M0ZV%r@vp)-&*=xv-no}%jwLd zzYO}b*NsyGmzwrjL!$jJn`hy#Cp=Md@G}R^3fzZs?wUIy)4pA zP3$3#(@m8{{KoxS&i?u%O}76&j&=D@1`h7=--G2=tfLmbU2nzOHsa6K&|Z}l!(V@) zWu5(7=T_E@H(Rmmnfod4(#_Mwh7FCv_j~c-=OJJ6FH3l*{~EK_Fttyll{`ul>#C8= zW(UwK<{pxGUm{a6(?{EojrR4(LlwRnysq=)qTt))yrL`S!@r8();Mv83WvXSx257G zhJTw}`;->Y{U8@FzUZnYz+RU5+JXW%OLBt+~G-w^tE zeWuL73+^}>qwaFwXU3Qv5MMssXjS)LJnkyNpz|sO=NIX_kpBlbqmDgOySM)AXMd{t z*>C?^^|N>R#Qu8Q{Y@L!eq6HWM_K)838F=k8pB@E9#anf!gzO2n!OYR3 z=*bpr^25IVRfl~2n-8yJZPquaioGt)_`=P9`1nU_jsW8qQ~Gm$RBv22f?kUf<8A@I zR%q1*{2FsdEn_K#W?MNYmh+SLZ?@kZK)-o&!&K(r9DKJs@C9!wg5S__^zCg+`SWTg zgpY3LLk|e-r{00-j-S5-9MkTJFIb1~EG)rZ4eETlvwQX>@WZpjbOzsZXV<0a{}i}1 zo6qO?jN~(g&kB8Sc<(5B>aLW*&6k`vsEYiEq4G&*^1qZb;>PJdDfqe?d|uXnaMgDQ zuwIcewDWd&T;(pL+cynnbx&UW5{Ap z(*9B__6TFRwEv*0BTKE=CHLOhbrGerKN8$QrITYLZZNx2`UZFEF2Qy5wUqUrrH^xtqHjpmQt(!u z{z&IqbYQ^f*a5~SX1el&JRKO4pzDe zGWIpc>Bk}P&oJ;j0=~-y*9GwK82EKOJUjthUse=*&g;*WIPpa06N%>qRAT!S=TS^T zGAtMsLo>dHKc#vlziTe8wIla;?Z=c))f(|t+v@<@XP1i)kc$tHjnAulAhn*U+>q43 zHgZ91-`kEwCBuQnS=`4s-va-Z^Jn5Ym5x35QoMhPJngy#c_Mbhc^~P9=LKhC9M9_~ z>|h-pAFz6FYK7U8UV}|mY)k&ES+A&MtjJ3tcA^pd2|w?Xw|2{Aw@sSH987p=qmM1j zgEE&-sDA&xqIDS`pLnp`pT4AnxwGaziDgE7Txx!o-O2cP#Dg2TZ&+os*pov((6d&> zgz9~lpVSv9JdO{TP*-%& znwQRhdZ++<2Mv|u_CRK8)eiRl?bu}P-eFl)JGdWLW4#5s3+{a2FUCJ}$6dy_Bgm^C z-}otD-2tpSt;lcM*fXd&P~7{Ir4Q=;&XXe^e2Fn;_^iv_@yQlFbOn5kUNpJ3=npp^ zHjX}`4YB#`A=d1jndw!ToYRtpY+pxvt=Ok_^j919;TFI3qUgT>`g_+0i_qmQ^zm3K zIis0Y4B8#Og?ub{Q8QAQh4@LtKBu$KF5BeQ@JUT$4N&Y#%@(uNSA^P{1@Q?aY-n-UF1Y3)2B)VMdJR$r6+ZXFS zXZttg_DEm5eYMO%hXAA2%EiY)KAKyYHT4PZ9BXiO;B@uSIf_rTO^2t@9pN(8Ac+x= zZK&8#ioc$RU+g#Y3eEXBQvuzOB0X9{zvyTe7QxXry(lIdrRSZDRcpE8kxYDQVtKna zV;?h_`)8(I8_3M!Gnh|4pJG1C^}XRe`SIMF*qm{0hFS9vZgz4%@hLJ<$k_2=SOXp= zo}b45o!mn-4}9mtck{rv&iN8d`24Y84!kyycSCr_cWoPCw;9SMx4D4(W3rjUp@;L1 zcfY84&I0m6wI;6fvvzWSr{+=Gvv`}Ko6npTEIKtNq0?8ux0ydV_1mD2=9$uG!kgxz zhDPIiX;hr74rQa`!J+g5bE(XR+j+c=j5i;gan{^0dz~dUkPn~RGX8_-M)NG4 zl4uKL2kO}iooMe?B(Y~5=FWlm^)anonfWYo5JkUfJxl9a^O+M|!zbA}_{*V(WJt1W zX+?h8pl>_)=s;HYfvZk@LvOq;0v^B2{8x5W zcgeQUw&u`Yy&(KOhE5s<{-l$<`N6_dMXo)^&)1x)g0pP#oz9j16pfH)U63x@&S-_!0O~%%QG~K!fzIUkewlOBwSrllRb@I|SanxkHJ1?+0DzJ!D6E zu2nKb{hr(*=;4;XQ|0*0lmn-tqjXsYdtch`#+PQ_Zw2zISc&v(_(tr5Ig3HMjlHPo zwr26UXo7ALou$8&17gPvH9mAUykh9pWcx(Q**E)!@qODWsRysAUI@NY*$91`wS@Ab z*erOd7W*6p55{+M#;)%R@O@hNu607~(^TAFd>6zBYI2h?S3h%c7M=S6urq&L;9gIi!_y*}~{@U(ZmeH>?cC;G=) zZ+~h4d<@U(c`y2U6ZS01eWa!6@+IgF=``hlxw=O&yT_8~^mji3-{OTi;0U^ET|DgQ zbm?;w&tpuzbV?1pnXJwP@C;u@(f!g(Vd`*~C44P^TX99vN_5K&%!OAZGwp@&()gm7 z&O{ae={K*wP#N!daJYN!*zSeLj&pAIz$AXr806;~_@PH{`P9JAX4rrqd0_+PZcC)q7M9}!L@ui$btcwa7E0lhUhk3L5p4`s{2ONZ>*m?GAV=?A_r zIB;@fWY7G$tij^ylto?zJVbBH` z?0&UYqr3=*_jtdYyu)(&fW&e%Z&bVS{j01$a6bL#!h4B+K%&eG_rGOd{F!82dre&2 z>%T4E_I$C9yG*_7|61#hEVIZ3x5x+M>{#Lf`>pI|<$=G>-c`j(wAWC-sZ9IB(5DLe zRND32d;#r4%!l{F_>5unZ9Z!biuIduVZT07jwAX*{)6nF*{cSxo4snXBN_014s8xZ z*R6xLL-@a*|I&LK!9VnoAK8*?#r`paeC$P=kzI+yY8BJL8iUSL1f?tOPKm+PNZ z)k5FBIYwEXjf_h?s+fWwc=fx=YTw9scj7y zj6cXnbDU0OZR9!N;j&&lC_a8ivU=im=LAx+?)BGy}-43MNFa)V1ys01My16=b zH2q{aIkRdFZ0(f9_tP&KG$;-+cE6F5>?< zK6ZXxcHr|opN(E<9|Ijk4|vq|1)gC~^O^KSV!Z6bxcP;xFBXrA*B7q*OBYB7xcNSz zrQq`paEbgYfBXcwS0Q9*B>0My%O~@3Hf2iZL;28TI%Sb1$o##*)RS+2wHH+L}rEAfV^ZpK$Y{9XHxTW0}dnGL6O zW-D;E180!BO+wibi30+0vf)@kOx)U=a^Rfv=VJ_oIK57i)Asc$4ne zy$(UfAsstb&*&Z_>(n!{j$FI4uAD*n5t8>H_z$}M!XxVU_nh~m{>+^(#0hRiuGM#t zzH0d1fj(&&Z^N<^UDE-s0&C)I;Y>{??8>=sdqlU={2-$Fq+AV>zc?>(R=~ z;oR@pW{*Fz64Cq=9LC8a@qfuuC%9Aohh$OXve#*|1AFP;hA}np7xS#Uwb*ZBp0%dx z54>%jQv(0R|FhMD@{zQE{cLpK4zIqewCD3#<_z=R`<8SbM(Z8$?cK=PQh0SCdT$B* zgY1XVPZ9XL17Bw!@iy*hY3?Yo@1NUOh<@OHmLT}~4LLEv>F~Jt8=1<(u36Z%pOCi_ zM301^D|j|{uZ!2Tp1e+UgNO8+>|X@FjxL0!mVjsYM0}&UxLI=$p26{c+E)&o+Te`4 zN0m?4NgMJZB+t;UIyZGdX8>NCfA8F`732TxSgE7}LXN@ei=Fdg6L~h?&0F{^>R>?GtUdTE^QjhI82mGJ?IE zz>I%KeL{-+r8x6;R7D! zb6D~@tT;A;evYn#ujLP-yL!8YiYZrE`Xk!byOq{yq=+&Bqo#iLUa# zkL%M9jcTFMYGnBRqS$xBoF4!k+6$l)blV5Ko$4F99e`FE!y#yS7+M~Iwnw3z>czL2 zi61kVd1n@W%s^sPgV5joSc}x&tDJzLJNo-kl>73I&sMG&1 zzt|& zj#{zxl=}!B`Df_V<@d$jg@fvwto%ZRqNaS3pxvUelKl$)r6Gu znhzmkO-JwF)8Zf4d^2avKFL_RtqVG9ze@RqretGTsL{M=a$3Jht^*Y@^~)1EI}eXfpuX z41zX80?phnVr&{Phk*Y*&K?W{yO~d1kKL`suD}ZyWub?m|7~eG+*jMr=!M7(`@BmN zpW;lEMr?EUOBQyv=(z(;Y1RAquh{4Qhw%Y*-U+@|uPpZqbg@1kf)CZ7`cV7EPewNP z-6(re9Q!46)BV$#Q&9In8~1XXbrJYTIh2Cgsju?;r(<{8uu;^veQNf9NCz#nb&!2; zrhMV-K!12XcmkVu2YLpZ+5z46K|{s0o(BiYIXce&7U)|@Zr+w6&8@V*JS*`1|F&~3 zZYu;o*g)Au;p;W_ng?ma_38jdfDSrP69r7hW6 z`IBQglP9#XC{~Q`AA!%xn72gXv*r5Fp64mZ_!RQ(qVz|8XH__ZcIi`bi|`Phhw(gu z{^{>!`mnHH-$q_`LC3?DR&2Kq{_Ph3AdAcRPySM)!R_F{PlVgQRX&zE@=>Ywd4}(6 z*rV$oYt4Qh-bId^m%_Ur(AS5^@~=Pn$(rw5pRan9cOUtO8vo!Je5?Fn&3ArEe(}3M zxTou0?G42qExHgMqrE2PgCFcgM#`+%NAFu^ujPk_tk^(g<0JZzY#l;>a5iK0N9c_% zWb8Me{B+Hqt>IN&$kaF^AIb+A06z>uK0Xg^haw|GkdfiY$S`DNV8HLn2V>D%gzR+U z{j}N6eM6cL7(U2_b~Sy=P1yz3>~Y{fjd%!7~>4?x?335YSTV}EXd7`Vv;neMj!1L76eC8CHDRX5; zxnH)-6ortPkKnB(@YqtuSAeb9YRin`Sgp@+4i7rYeINXYzLo&zQes${){z;+UA2aU z4?k4?K|T3!wVdS=c?MoZuJl{YUu@ZWT6t%{YGes`%^ZsHCd!m@V-zpRg2&ZQ2)HFb zP0UTze_yf~K`*GDWHE}Jiz16FiTjxRGTP8Q?rE*bjYi@~wMIrtSXKB&1Ubm|ii z!0zH90{$Z4VT#~n+zsf2F!T#;wQ<_cdRT;Zb$7e&XphrF^l>IVv`Ot2pzH7o^ysuu{lE}qxy&K z$^%OY+{bTAll=aDem8kx*vKXP?&AHxzV0}@&w@5({O*>WYo2Y{_!iIKuq1bmWeH&PMK1#CJ~n!j;Z&eEnGW>^$(HGwGER`-g_^*`txM>5h!0KnwBAA?`6~ z<&(%$9c|*7Ea;mS_#b|E+v)#7d^=hLiK7es%g-PWnD({qk`Z{D`hVc#w)cj4w)tu& z&pY@8nX9TSm$5p}=LKFf<<2ua_B-=_5M@#WFPry6ct1F>&AhjD_!H*6y`KLc=Gi{; z@K-z=9}xXz=#B1*pu5~So;Ody8{geO5ua?bc<6$_73@Jjn^=c;&3FGq{I$u%(oWjz zM8O>JG6deu1xFd^3CsEqVsBUGtSai7z-NhX;P#oTtS-fVf>!30IhQtWnNZxgMeoTs z-ahlv7MEVu72O+l*RS)V0z|*}7n*gCBJFCSbMS=Wci3H-NWMLmRg|XytCNw!Us2w3u=6 zCw13fUR<~B!dEtYl#&^EI=^YdPk4Sig?)oOe~;%EQ~Gh{LeqxXJU^O}6S$V=p|gKq z$DY=@eDHD*9>}EM7Ke9Bk!R)I5xco^Dln$XG3SN*z);r7WU$Z90iTZi=PtU^E)uv_-Wjd zI@Ipl#?Q-JrlC7hk-2naPWVp5Psq_qOG0_+Ez_*X}(iGN8Eyl0dNt81PzIKHBfjPs#ol`3A- z^_+^6w!iM=z6|+R{2ikOpWv&*Cn4Y&Of2tMfv>tIWpL+X>{&uys@Z@X3-g;7zAV-{ zz4xJor{l+zFyAV~H<^habE73bdm#;2N0Ym`qam+)=gGaVWezI(tjm>IYeO`BE0l#o}nH8(h-hqm7)vZfT ztXIBb2>CERHSH)jS?ivjZLK$q#;++-qXik0|$jGjZ-Ex^iNx9l>V>m0T-|FTud|q{6nx^@%fhV(yc;#-vlmbjT-&^^l&j^X{_nFlucOk6wyzRRGW@U8rbFm>ia(*|_C zX^-#T^2ypb+jnW^j1l?ry?cl^(N}!?DS^KzXNq`*a{k`sd6)Ta=YIW;=WJirni-rq zHiI*nIv3sAwa?1tjGr`<V67@@~UXfYxrxMO!xDg%>^@k^sDYsm!Tkiby zo^sbv&exA}=n}WwUnn;(3%&Z&-0E=7{pj2Ko|{a${A|jhSKM-ca?9P5*Hi9N$`xl) z4qCe9UUSQRJ-?^i7|P8{ryMb?ia|OrbY%7KDfgIF@h9K;RxFfhJ(B;(7pi^3dCpn( zoNuVDU*@4-@C&Xm@iX*91pOhM)H1@p``{bo1?jhZbWfPyCGRRzkYky0qnLxL?1juj zefT3O=p{e;D;52f#$J%goW(J(wm3G~s+c5wHlO~`Xy6hi#_FjN4Z4)m~EGPB>(zx>{0lE;g9Djm&hM8ddgkYQ?8D3iTqL2 zQ*L}uxf;qP^2e2wi{ph0h1Dwt*!)ol{uA}b+itm^kLf9Q1?3X?<4w2R(9c(g zGnN_t7*DxG{&?9f_vo;ma`}`?L@xTAl>c@xJ{ITt|Q~Bd7 zZn?Ra_mq2@a*6yg)h&11xat*yD*hzQIy9mpGIvIx^-*N^!MkkwF!LkYbl*Qj8S{?0k^Ao2op!xAzZ$t02+q0VAeo6d>FE}y*-y@UY zdpYxmFuan8%PC39K1Q5QxDcIKTLUld9MAUE-LZ&QvOMGXlQRbSXGcox{L6<7&-CtR zJ)1eq6Ts?@@q>jYj2yY&kMLcz5ii^Lb@CL6O@wB_i#I`A?tbfLJ*B%lqb|kJ(jN%c z`l`48fw748UOAL}LVwBBbCM_4nccqLR$sHACv&;2M%Lhu8+-ZV`{ECNGct~k{WxXZ zx;s_(CcEy>RhRjJx9&pKmFz;lxN$ggsC7ckF@@u2*#qs$qj*L0%VhkL#Tib0$)|Gh zm4|-=Ipp(+we`%k9ey?YubB7t%(oqRnXBB^O=;C;?oZhWynX%2?)7o+S$k5PU>ffw zw~{I5&xUT&JwbkF{=8mtB-t=?7~W}~t+P%fV+5iewa*mS{NH}}ZR+$XL*Pd1YQoQ= zB>2(%d%Kq(EBl#z(}(U=zUhGYeA9P)+2k3cD|Pp0>j|spEChSb)w?I^dmZd~+6<1Jc3V#nUjm#in7c*A18LoBUqwThj zSThArvcE2#lymp8@Pyp-wviq<|FreNmw2A>?*(7iH9W9EdB;KFs}kCkS+VxFm{%Ko zWkQ#Xz;|7I0h7TO@U&DS8)d-s7O_@vc3}d}ZUkpz!P$il&c-@88|C0EdZ?J(8yjDe zbr)ZI*|WfYh1rU2DyCV$7+Y(=*U`T56;xh_@Kxi$N84>jPlGRXjTc|LPFo**p$n4Y zs~milg0JRye2q=S7x1){gRfFxY9>YszS0u#<#+I<*tK{p)4^9N_|m>D?fEpirgDHC z_oPkNR=IwM8*lUaF^_rW_a=+@KZ#6Fba~8$kG9)7$$?mZDxVH^ z<@&N-oM|7;=rgv#`FYqdn^&_s$2Hort+mgVyP@Yz@M=>$uio$CPV(b{O=os$uULCi za{1=C(9_CyF1kg!FcsOp23ysP}3KI@aqj%`={diEG)Hm)+ZL%gPYSljA}P0otv zwM&!XYbn0MEMThVJR&D{Df_Vi`|$+!qo2cTsf>FRvYml!D~8^eYVlE}%%@TUUv9!@dESae-YM>x1ANT8?lQ_hPx;#ZcKI^OqaS2TE>xXo;_Dn% zok+h)_@VUS)QR+~4k5?kXNqFORfpVJZ$A;!4|TW?>_lxpryuIz8-$-Pish&d`RLv{ zZ<~H{CUsY;AE!C>;3SCckYw{UKS?Io%FlTiqTxn{XNB*t7(oF z>_#t4w{%a|e>B2hUKoxv0#|nt=K%o6Qo){MVAuZ7RQi_fk^U5}WLvcs^a?q0p}T#t z@G|zzP^XRh?ZDMe8^JQ*N5@Gg{}K4RBiY=OtMS?G1|!+kAIvK#MKjHU`>lEH(6_jmrc_mP0?<;P){76i`R^@`C%Gd8~_TuRYN|BGxER z!Sj0IncPj`bio3f{fMh9XN?FmY(xU96KjEiAKiHGW!QP6U@-bANUvO z)JC5nWG?L3Q}=Gu3EZim_)UcR1;|_wnd^>ZPCDVY$KE|2+D0;~L-3(!HAgt~+k3ep zv>ATi-j6SzcPjnQV_fq8Y8;9EmKZ?Dh$n^tTk?Wx_C2=yLD zw?bF>dq#ix?fcq{og8oUm#H7H`Mf&&Cv+n^^(Am(&dAZ( zH)#R+7iJ&Djqs)eOY{!lc!}5s_?v3uZdQBEC!H>Z;xP3<79sLjn_gjTqzR#r{`Gq$pk(mfM_|bcIIn8tCC(wL8ut;VW z(1vI}7g&1x^asxK($9hATjGJOz!gDe&}skP;C!bq>Bv=rQ@`swx+%;19etPP{f^$e z*V~6vZ>q~@^cm)z*AI4MJ#IWgb8dJ4r{>-deQ;`QLUZqy5%7L4>yqT8ns|`kZ|2|E zU84K6)8gmfo1ybo4Q1`7Fv4qj^N}vwtlkM#FfOBk~e>O|+HdlbjM~)Hw&z&&0@P zd)u8r{q8dEYF{ci1^_p3IBky=&F~G)-e2e>85;&{MlNq+Zab$}E*HXAn*Totk90e_ za52w4a(S5*YcJ_rF2C1;Tsm#GmneVTmP>OE;$_L@a#6B6lvQ4e`juQ}BbU0%+sNev z#xmYwFDiQ+%sf_k(V2mlu?eB^$feVN+XVWbY|jrCI5=AMbE7T zKE*}Cln<6Ubr@TecB1g^3i%$96yxXIEFLXXJHW@gbLmqtvjWC1zfA4QFDr0((Q^E< zAhs>n^%ER<5Uqa1_*CEJO})FGK3txEOyiS$pet4g7ZWW<=UIX8XuK00SisGPYK!lG zD=fD<^#y}uEc_YU`)8-VcWmb{HqoSSy=>>tje#aAmrO68B){Mky^PK=Gy#_{&~KQ! zB?&Zf*C7&V;=^_cM(jEAW7~3_JvG{C$DBn2y+1E}PY3_tTKpzF4d$KlkX#$1aY+X3 zvmKBD_)B)?=%4L!$;R_qlJWHLoinbTrte-_FB5L3|LwFs^n1I$mo~Qi&1q;yC$kUG*LlFJiU5mV<=XIY+cLC2c z^_*4bT~~-c%(!T?rAo9z_E}$OEl8RIh06L16HJ0B#3k|%4z(bgOBuXcrSU;D$ zm}KuXwnzcG8aTB_VlMyZQ}3TCWAncir$Yoso6e}uU<^5o(F2Q@e%Eo{$0q}O^9dE# zXCCmvHQ3gA@{AjhKlLxYZ~CcpcuO`CJVr`@l{z}_>O62->@VW%?83K?LkB&FOMn-g z%14n67LSIwCr>ssAKd4&My&C=w)ppRvBiutF*kbD$&z&|pr_g@M=$u?#QbIYY_-x&{<3m~72{X#viy?TF!{n`dh>;epY3vev412#M)t2!FhbV` ze6hXwVJ@7H3QoqR_2WX|VXvf}y9{l{^n0LLdCKYiABfLg29_o(dG2zBU_nM3nIkEW z+6&9M4lJkSE(5z_v~hGvMl1B+f6<2gBk}ESp51)n9X#jCpJa?)9U$E$9{Ohw54r2C z(pz`-!B09r%DdjGJP75Lhmh3^#|qvw-A&D2UEV=2om16 z!yDijoJL02zvUCZk*=FZTbk=>o%GV4yjhz!EoYBzzRplCnppks@1A~)U+80u&7Lv3 z`uRGO?{a>xzJvFqm))~GTakq^qcqNpz&haarQ)w@jL`Nm-b*K&^v3I0S3+%|` zGl)+fpCUdL`rhzfJ9(w-!EN3#+&!Ov@qe^WypFt}!v3o|@hxU&EE;!}_G0ck zzi7{zPr5IUz4nE%7dRi?JztoJ1qMvJKTgH zC^-t#MhF=^#SX*!>_JOvhZ}>~K6JXr_C4#`;eR2IC7B&|;e1YTVjsf5Dmz>vJB+=N zT{m_ZI8U*|z|xl;{)J${rX+)9q65n*b{N=u?6B(#LaUD`@7mmBJWF15ZhQyNr|=AX zvg~lY&h_SBy8Q9bsCfN46~5Nmhwk-JOhkU$&8%gmIWZBfSKf>q+)O#%_mtB}iGYYW=21^98zuMAH$cYWxtqIA{6N2nn-z90U3!f8)yCQsh8;Iwy9sy)uOY<|Mi zp4#~(_FXck7<(AoC7aTO%_+m?G#x9cD&tvmkp<{~&0|(E&uoF`A4eB!Kc3ef+!SOD z2Hq?sfAc!dI*s^$_?#afQhKO>@{*rxIisc=`KcXBzsPGe!`Hl$zE-mLZYB556tlNb zb0X;eI-I}hJoa6n9}?*8*8MYe=k?ZoHh=b&XRNDpkaWJ4Y}gx|W7J98 z-vFLdZP;p4W(B%r6*i3XjW0JkoxOj|8=A{Fi;B2k6+X}lnkz)mGm3xuU0fjJ#z#aB zqu6k*Ypeth_MG3gmG~Xi-Htuog5M$BXieu7Sl|J{JCeEL0$^DHEJ1XnS1yjvbmlg~ zKQe9RHqb;i#VafJ+%rA!-`u|-Tx5PMyPF+Q++hvp3FvOg0CsmTXBAzC-Ce|;lCrxw zUntsBmBj8w-j+Vm{eJ0z*J#sjKQr(u&)gkYJ;Su$FHlbV7jXv3#kBuR+OO%q+Sv6q zi!L_p-$466O45F_Xh5pqX!Rvp{BLEFAD*+;y>-NW3%Z?3)gwDA1fS-`XVGr)75!Ow`E z`IGSTh|X+6_u2Pa#QW(ko;JB~l_bZLH`Z2{48CFLNc8UixOl^#*WT}9&))C79-5vG ze}jNGaX#U~`4#ufOn={Sil-|2fKhQoFI^iw@I{iqr!$qk`S8N$SLOE_U&9wDmTrg7 zM!7M+WPA=h-%Nm~-Gw7U92Pt+_wZK-ze)Bj^sHn`bHLN#D;fVAxGE-9|Lch}^ZykI zu$jHYE_?+x9X~7oyYT$+EZ`aZ8Q}S}{O`gQbm2M+{_i-DL|^=!`QL@}O!~u(D>d{1 zbU8zja0RFIYb~lm7o+0&IUb{s*1`E}!+GzofhV=`+Cd zH}OAk<+*U31^lA=KJ@=z_Q>zc>HSlmEl{ z@c-kk-<^ivUQB@P@4^2rJYP8rd93~n@O)bScj21r!gUt>zc?9ue`o%8;XIT6aQVNe z4;cS%^S=vM)b$_x;Q#Xz;Q3qme_Rs(U*{Q(;4FzmpRw5ZX2ern+V&v>F3yg2#>=RY z1Lzt`@P$ryo>4~{tP&KD*>K17Y@f~oagZwg->^_!2I|4j$Z#U z8U7AC^C!XRokO|!n|BsE1PR>zj|vU48SU3g-6a zx%TEsTGP;ceIxrA=5D=sk~_EXj^}l=pR;#^nd*Kpt;d+yIN!a#*aO_#w2+T-QP?}uN&NaaG0pc{xD!b2 zac0u&TH*(q3n_-Dy(B@cJrZl{_YLA@emmCY%fi;Za)BE!+jVNJ&7Ld9$6M`p|H}Or z$>J=oPB^$PxlVYYFaA9Z{z|}K5%|+Oitt8^l6!Fcor@j3RdOG-V6^WzOO7+GNtK-} z-qXZ+QW0XTI$K`royzG}jzfPBj+(x%H3`o3Asz{i4(E^UcUCwep8cp`OolJT1$}40 z*<*d+ELpzBVv`@y8=o?Jmf<;_udi4naa!_Bh&vMdN`uctt9s&$;~8ff@2??dnL?Sz zOkR9S;BDQ9L)*>(}t!7HFyQ=&tM*VuFgd*ApLogZOA^o^?d~suaV189mQB_E4SC zr+EKZaZCCPQ3*{^+GbzD90AiREGxjN(|D5koUXLpUJ!aUn& z7~jvc@Q@grb7io9AN+KOM+RN{ah-{GUeFW&(q0ptRZxm-_`!jAC7+MRs(l)w(>2_~ zHJXnFJvM_Y?m2%?^m>*37tj)Yl-A5zM&LWUwk{ke$=sU;a$(y+K6yr%igl*;L&Hrqvyk$TEAeu0D6qCe)Fai>m$gv z&eal+YL5r;*(;QvmCBgrl5?TwQSd9@%~^BAdSlDMM=BrnQG<@%&$)*tA9omQofX!R zwUR5!-D0ISx1xtvBUd^n|4DLuM)5An+WT@NYiWTKjMe1s;vYS zj8XEaacOTuk&#E%0D(pNbrigQ4Zl6f^OMmZ?3t$jtZ_ZbdyQY~WIOU+-B1V&<5(L} z4h{RGVy)cKDjlq|X2swCn>-WchOy>aZN__L4^zRn1IFp9RNi)CNSXuCGOYuhT$ zc$G{(c}dNN`pn|a>#$`}*0q$Er}Zq+H;3|}5q2Ld)d%s@c}^c1gYwW!PN7+QP3x?s zuQcii2J<`o7e%(pkzd_+ptc3C?PpqnZrYSCTM10c0eXt_d7;md9l+H3amJN8Bh`$D zI%=;><;M`mrL25S+g95-4w(m7@20;8z`1fM9^t=D+Y5BZCb9Obs@Lu6m1FUJuEY1& z(3UHo4-FH)_r|Av@Qg3VK8sXx$+|FSJi_OL;CpLe_2=Mo*@Q%S_&U54gg&l3WKz!9 zI)_K*ArJGA2hmOP@N;qm^jrYHD5t>>uS6Kb-Xs^X5f1g9{vGviHDV+KhkSc z&?RbLe$+_O+|k2-{O8_vLU2Xwd6H|>pFPFzas6PYZasCE_pUKMlK<1sow2U`xmEX! zG(MWfDcdBx+>GB=iro0&>nXq|9W-fww+ zjgNRrHchhdP4;;QGm2tYJNjS0nK3Y?eDp_({^Kho=7Z$14lX}k@^hrb-p`c9S+d7U z681AagN!_H#lqWcUOvV=e4}Lz`uNb*bsv{mySsqr;{t1UcP)8aHG+vV<9JqXZ)Bs5 zTftNg@5>j6*4qALf$k}xy!?2{hRfUT{;2Q&8UBV(RYvUvX)E{)^6@;pyN&l7(P1_C zU(idw$zFKo74&^QHq=7KRv23%9H#}|LZ^7^if2^Uo?oPPR^lhO&|VquB^Szn$Of*> z@Zusq*naMPV`B^ru zQUj6)wH@c@z#IRE{p{imN=ca>LPCam8|21cQC)QPD`CZ^`gMAN@-{8d7ujzq>tYOJ__WE%y>_hv2{iy`l>!xjSigSh<1Fwiv$JLwiAZF!U059RV+G6;G4f3?GGQ-u_L zy4p_?rMzhC`X&wXP2h3)CKrlN;Y*c?3Qk}QiuawqqSUWJPw6aL$&qq|R>=orAByr% zr8kDVxp0-v-V|u3_W^R0Y#%u_u#M+L8@v6F@jN8ma*c-uk30Is=nMY{v;Xn+soMYe zg}D8X3&5rFI7Y)u+Gku)!aYUKbJ-Ney4laZNWYnXMK}7n;}^eZy+1ac|8491vHBwG zh{asDey(+76ZotL|3&a>3p(OhK}sB5UuPVm>tp2dgt*hP5FcbM_|&iTl>f9j4&$ChmPZ zR+1gpf1T66tH-lVna{}E%~6BMS(vgV+^G9#3$g$4L zN=25-tkkXYT`bGLJwkhP(RJv<4Z??uuO3`Y#-D*!&BwcE8~>c=9CTeW-zG!yLK|WF zQ9EvV#dUNxd@43LUS{qyXB&M$UW?9`j8Zq5k5f#3hkTq>@Q(IFR~SE?xu4UI+U~Ir zz_*g~i$W#mRo6Rn0^3HU1*Rv!=guW3VbcY-?XP~pnTrXw)IckJn|w|~qt7L!(YxH) zY-ltdeS-cAmr(93Xw>Y~Q=O4>9U4va&Bthpzpk<*Xb?x7L=*)%%m6dL_c0(>rw zcCkLT9^Di&@T3MRM3)h+>|`gU%iS(r#y}TzZ5X*e8@eoZ=|Y{66%JjPv-QZ%7?&>D zr>6_^!NLj;U3l(`E?2qZrk@_VP~Ol5JLZucn=ZT)T}B((!Oq3YP7hta(-&QweqrR#^Dx!%`r>wfN~=y&hAB4UCU8l{f4Gk`%*;ss)B-_}k!m*kv4yg=RD^*`?B_0Yp=ET+H0>p?L*=YOMLkqxjbLQx$LP{Uh~KP zGn$_}@!0a`S$FpH^{d@xozYQ`&H6F&`4j$UAUFGRpXFW1zqcOfyH@WH@qat-qvf0x zto3u>u;us@vO4BD@`^RG@%6cDu@8V%b3u9{T5jWAdZN9yochq4_LZM#18a;dWY>AL zkDhVxpJmP|?D~yuzn%*JXPyH8Zl7(~{%(Jw`!@7ZyWmCE$=)4lWi?;N{{eqq^G0lk zjmIBe?gM}O$??&KoU3zW+>^!*&TKftp_AIu80`D%{0$a#+D(j~d}*%$lk8>5x954@ z%U#n0AAD==fCu}oeUZD0e+~VQTGb!h8`HW_9g26hPt7Qft!y|_uh=bzZ}T(&#AkDc0Tr>)m`^K>V~<0Q*b;0 z4EA}2{Dzh0IltkDd@kePhBLdN#`C<8=Yn)Qm^HX;nKQ`?;I)Odc7B`t*1YSV+ws-> zwdA=8)DuZP;zwu>TfIt~%=<`6GMy$MzXt?eNd3 z)<3OlvH$9{9uATtN^OenOQ1XX4C%+EyZob%!_QhXDL0M5Pm}Ucxc;2XhFW+)-*t{Y zMmepOmhvur?V)?^M)ihc@cT>dS@7bw0Xkaz3BTtD7quL&oY>*Njyp>-26VhR*_w9# zHx{)VshrsSv%i$obFXViI1Eq9BYrzC*)~0o`0YI6xbqCWN1Fa7#a^23;(TIWvTJ;v zZO`)@r(FCBw@x6X&Q!O~CDa*99X{htflz#;TmGt)@?~!MV#*g#p3ivGRiXI#Zu!en z%3tl4KZo*zD9>lS>9SC~zgzyIl=5TT^7)j{p*)}Qri-{k)Ga?crTh@Ld{=I=>sW5G zmCtw+_rm-+*TK>7l=6}Gu|{j6z_1$7p0WXcFQLyAEP{<@us3se7{@% z!j$rVAr8j0{{iKHOL;!yO&5mZ@44mAODX@cTmD_jH&dR^c++{I_-?oSSt;eCZuut4 zZ=*b)@ussv@mJmQ15?Vs@0Nd=@-I@J&v?_oQ2YhAe7}_Pd))GyDZh#Ge8!vlh2k6B z@_8xc|I;mBNBOms=QG}v7mBZT%V(vOZ*^*cS>yZd` ziYZfU`(XFF`(~xYe#o}e`dxmtLfQYP>328f1nX!%$M9L;>(eoL$3ttntUR+v(3AHy zuR-(Qh4)G1rhUhbWM)=F3Go$SXOGA1VMEWFurIWPy`F4rlZBl7i-NBh)IqZlzpnRWA91u1oQ|EIRw>%Q}Zsfu4MXvA-sNOVwoKZ@G^#tGxWZt;;3@ zQ$~l*-9%~kEYEmld(LIN5$eSjPd59lC;3|L&o#al&HwQIc9~Ye2W(*nF5$Noo^6A7 ze1BpHlSz+>tfDgB?YNuWK z+N01ywAa6UR@$Flie0Q&3+?eeg`LAWkKea)FLsT^z18rw)-AGG641!h<2T{Bh(0yP zG(WiemHlS+w&_2?yh<>yBG5C+{CEcYhBLLCG0ts%lfLlJ{9ZQjDDH|>`x{tCdGe`> zN97AJwy#IFHZzB>;WzC$2PK19BmE<7$@gmf0kV}TUk8nquc8#)bEf#y8IR=ZRA($| zyPV(4ZQZoR)lF-NEs$KjhW5MnkZE7vzfRxUqj6&@UN?5TFU4+u(t$_qGgn!!Wt#mc z?PFCkH)@F0*-5OfNju z_pKCKhOZX0uUpPJxGwg;wcpH~+fuo1WT>2b)>7(Uz#ajMk=Q<T=?YgJJ){-T)^tvN_#(gp|j}D*E`pL6fv-tjtaFb{K>k`nIFI1 zIrBK}0juwKOIy~LKiBEox|g~M18dQ=l`T8|S7Yb;-~6GobSPs7RyHvQmq1s+y0Gov zT4pZU*17QVot;J3wi;MVfAUDn!n{{HXC7(pTz~LW18dQke`%R{@VU;J{eIgy^ZS$o zuj_C5S&MJf=FXznJDtAW;2KzG9q^V=^u7?`d;T9crCqrS<8;^HFU1u+0?nC z`OgN{(z8~#tY5LQb72Yin9*inEq!81%lZjhIv2kBdS~Q)#tE$2Kk*g)v}NIa{|T-B z&{TW9HAf9%}xDB~CH{|HWEn>$N4 z{joDrf5fK!->+!d@vE(!k(>5(?kGBFV2mt&v}MN};2HRK=lU6p8GO!sZgor1+-;pR zulk^K<{{_(5`U*j|l(;t>7!S z!8;@19pX9XZD&0%9!dEPKGAPs@f&=S@*BLO-z@zG&qeUZ52H<&@9>k)E9G4Uv@L}n zXLGN#Sr3?82@kFuZe`YHL$5iPTKg0msT@HG>w%TeSUI(~f5X}*+kW(ks#T+X53J-a z+}e5Hv-WlP^PA`KZXWM$@a5Ouxx(7_k-r~%168Y@>t>C+2>WOF3Wge<)Dm#wf}U;+IPf1xcLd*J;A$w&KO+#g8{yMfAXK* z{66pA=iP5I&Z-?YclOa5Xidyl@+snM%`@^N^Y+{AXXM5;x6iYmk!dxz-)TQ1%XZ#= zxBZNK$h_ko`x&_~`;LFIpOIPnt+P62fJ4Phe1_eib)fBYI1B!r-#A42uK*r|=RU)C zJ!^n}U(}z`tb3@-tobXSyn}dm;KM#xDIUp~jeRgfZD`Hm@Do01$(*pmli!3Vcl7Y& zp`zKD4o_MQ7W`XnO^vus{2(%rSQzw)K(UG>-2!>1F;$A0e*K7ITQ6=_Hh_qcc1m$rKg==sVCo|VvA$t z@J0!=l-!q0qrXB&Muq85^WcrESPP;fBFOJ(3H(wXOlBhEYpjggOy)=?`anLei?EmR z?;L(6UvbLhKLw86=!kY?QqA#)mZL9fE7)V1v}M_v1Z&C?*1Xm9Wpo3&Cd!)YD&%u| zU(yk3D{(&k(pL>~Et7evz7|s7-N#aF?L|5t0jvf5?uFN_ul~z{v(m~Tzly)PmHaAS z=eY>jwB}K{Hrm-n+qKAX6K6aUJRwt_49{N~Wevy$h9-7b>po)6X@_|1;$U(oZPidl zInpAuM=V;iU=hyL?(6t>{oqAyKec7)ni40@t`)taI7O}VK2t2BQ>T?SHBNiY>B&cg zZ_DO^9OP0)!@nxVh`JH__$kk#S-KxY^KdA9mcBNe&%8ALj`B=99tqucq&zz%9x3(# zJn~!c`5v@(ctpG*J+@9f`L68;`I`9T-C*)aWonF36EGP+=F7-semjDGUh9c}yqq!X zckz9MGGWRrBL*u{8?2A5Dy?r@TUMXgP+lLY&Ek7D-*fn$TOWDZUZ*T%ZQk}<&Y5f{ z577r6+v7d<^`L2(&lvLqyJK$+_Rs8husbLpZemQ^7~8AF_E=V}i6^b}amVh0##N$2 z6Y#J;kG|^2wDN3&cg@M}KEHRGKAD?tpOc(E&79Ro?%uu7UiY$^W2w$*Ln8QsWtiIy~9dzj|dG zcYrQI=g0FGH11-~Ci*XEOz^pFyYKMpmo8hgK1eR7V8-Ex@F#36v`qX)JMnd4d~h}R zz5HiZx7nl*!Ye;@Bz zGnnW5;k{*3>Z6a}P~W=x#`<{O)cWwU0em0G_cQrEsNS}lGi@0IFIm;b2fL-n`kT5rql99-mJkbpKr+6Y%kNGq5WpS=eFU}YoS==XuUXkfOWJW4| z*3zzILMC|00w3AnA&2qjGVVOa+lR5`GbY6+R$3YRK0}so#n#-KF@I%^)wgz|)vtD! zbw+K%8dRH$-8$Mgp!Ry-z}k7f!L>!!S+!-*eqak{V{zE9xwXg zcaJ{TZ~N~*W}I)&A9wT-_A%;8t;2WVr+<|FzKXtmI}Y@-rajO53Sx3b>p9PUz8xNa zzMuVEk8XW_p#6M+wdV6@Ipy_yo>QK+;By!1xnIX*+P?-^AI-D&*?r%6^o7sfUiHb1 zZ}0i&yc_pC{x{aXcjw=I^ia{es|IAhyQlwx&ssx_cEJCYJ-_7h+y3+Roa;ZYc5Aj3`dvm&?fV($)+WF?4rBIfvKkJ8ckQnn zfCjDpvssIt*RZY3cX;_B&cif4uSrIn zF3*u;`xo^BL)bG{HDSFrf^b<0ZD(86$=q?l8>|bTnObyV{nWAx8>UuXxOwWharbNa3a38D|2zDD!vBCig;Ou%|8M!fmw%mIe}Zy9S+M_xB}+e?dTr2u*T!JhT_b}T zcU9K3P0cDBc-ONFx4+X=J?EY8F2D7iVKuR-#WgeD`6=Z$F8yfgQp%}(PB8PXGRhAx zJM*r$DgPSfAAR`NcYHO`sTnnu@2sKx>q|eLI)ZX4KPTwBtA_GlFB^2%$CTem`R6G= zqo#H0?3!8ctfhS0(od#lQ%>b41+BY^D1U2N-@E!R+WyWC%D+qb@|wuh(wf=t{43?N zmK~aU8|74f8RM`hzofL^U6)dRJLL~ieo;+$YIV)bcYa3s;mZ!6G~SJ+XWaEo%4@s> zJ>&g1%71-X=Skyzy|n*b->1CBdpYIZ@zzoP)@7fbG~Tw-0e7u(#(Oj6-SHayFZ=MM z@w)iec!hs=yaxZvK00Z~2D(O39)$Lr!hoxU#qHD2N09j}Z3bo#pZ*La10cf2nC)A`xOzs4*4yW@57 zpU%%N{xx3V-yN@u|Mc;?_}6%ae|Nkt{?o_n;$Pzx{@wAq_)j0Ni+_z*_;<(a;y-=7 zF8(!M;olvvi~scTy7*8PI75?4vy7*5YuZw?;SNM0w>*7Csye|GV zUg6&zuZ#cm@w)iec!hs=ye|IJ$Lr!>;}!nh@w)g=AFqpljaT@0$Lr!heY`IIHD2N0 z9j}Z3^zpj**La10cf2nC)5q)LU*i@2-SN8kPam&~e~nl8cgO4EKYhF!r#nU$=jmhA z*xYfscupUe#^a8~#cldnGzJ%b$?|mA1+xpUWcTS{m8?D;oRZC_gE4nWc@1T*4%N{4XzKz_1|2_+T3d&6q>W68KFxjJN`1_seIU^!TvgzJ~}jp z_Yu9rKGh!4{hqUx^3%yL|2i=ddM@%4`;}uQCvsoOth#HGL-84EZ%5x})MX}(FSGiR z&~sz0C(OIC*`~c&>`4-5bma@Qqc%+aORhENW%WGsl5D5$rJ?V;b+1URyUts8yr=GX zx9(-3@owFTsdYo%x+R{vC2rlzL&wKBFqfy+UF5Ah$y0ZdTlb344!7=<)VlY2>tbu_ ze6ShsO>W&QLyO$HH>cK}?X7!@r|vCo-K#>A+`7|K>;9d$ZiT0AgvYwhGc z-yXSwwJ>{LJBVjcF8PJ{vJ^+9J*eT#4L^Bx*~7Yu7{0($)`Y98j{EQNVWV@-GtlIl zaCJ1-T2c1zBSP{Ct|#A(IWuz(KEch*sVg{tj!(Rd*d5_VcXbS)O}BmJws;a;m04p< z?(FqG>SOO~f8w7`QGXG+ph^mf?E_BjkttS;vx;hGZ`t0jhUDWf0$(~a&;EaM7VY?S z4+tqhP5-0I0~rL0+?z0X4%?LGX|L;p8TJV%dz+VxAD92g!N zxqE~oPeG$1hek0kjqu+Z8aX&VUpStKpRFXQ7?8b{;K=M7jt?evAGq6xViCLP1iu?P zjRn8>IQpEXjSb|*V*ad-Kr{b>;HmCi@A{6ar@^{K0z3+~-Y_CBS@$KT|gqnReTa4p%%JW$ z8@ZS9-c1~;?knu-!#%XRXY~HY;u}9-ee6D;xtF|aJbL_I`26l5v@K}BaXUCZ#Y!eB zdC#1Wu}@*gRQTZ``Lpdf#OtvqVpFJh3uWN{*mOQC`MjCWa_(sZ4>8)+z1OSk>wLVgBdR^biHuV*o6Nx&bYc%=i}rdB>}C!FR~VSwF}5N1&3%Z#7y)U#flaUwuVG*t25fD7Zv(a{u(bi()0VGR zuvJjzwa=g}@=|{NHgGB!omd8;-?`HoP{A!Q$LHw7qxFf6}h)7db!!M)phbNe`ZLD=R=4RDq z_04MDkNuX1eb>gCMKM}!l-)*I#fZlJRgEeir~EdRga25w?Tb@Ze&#U0UqBtnF`XCp z=GioOcJymh_TIRWp=g@!%B@FtF(&B@V@nXLWX~5z<~j8RLztX!>%ftC5INJK+)&DS z@l|g^$MkqT=?U4k z=nZpTM0XX&z>oII`$4w~@&<+3zt{KMp>5XO$_b0WljykL%D87cbK`0D@^)U3*>Ici ze4=E$vAbn!Khrm)b0?C2w(XDo`3$rJ$JpxF;sv&?PXC7PemnlG?fzh+&MM39mV6Yg zweQr5T#lcEj#t|c{f*&^Qjbluh_+h?LBDgr1NdYOjcsAu;-a&?XYbI&>8raf=9Xzo za*eriw--)#O_2ttse1@Gn180ObRE1i!&~ zFOzerx3h2cdH=Jz^^fG9y$5QqKi+%;z9@kP(7~LC0T1iIpLK5kP=t7#tyi2_y+M4> z^V7WF#e3xuGG#Kq8j{UyWXL7KWCLY~^WBu=x8gi&YlOKeS);obV^>(oR`iT`NOAmA zN1otbFMbE7ZPYJ;raPdifv<$THA9ZAF3hyX>$k41T=ZpLeVF%(FX;~sqy3ow?1f4v z>K^$pb6EF;<$8ZZ4@lPp$$M&jXK1LjAMh4uz*kj`nu|9Ues1#Y*)|zGAel4LjhS-d z9i4Yk=q?-TC)c8XatjYP0ISNXj((H>#;vdKm3^yM+Uvs1hDp?kU4bksPi@=qb(7Oi z4s^%|SN&}oui?D3cg`;N(s!{<-|LuzpZ7nzhrYLnjmr%&*{3gJPo&-9o$;{x%$<+GC#zhbQXoSU0n zYjZRkd|b zn7!hvcI%#X`aELYGee)pt$Paf`3%3ak6QH;e&;iTn8mV7#~I&8DYlcL)j6EY0jH9W zy~|U3P2=U4*Rc&;pNK7UopnzdZ+`mgRNj0s4S$D;0g?RCzWEVvH5J}J%zMPH;2^WLf+YV5+I*RgqD zWL~|C{jnR_{<8S2ph5h%UHL$fuWywdTPqm@&;i+3YfbIZ&P*iZw-T zm26>X+4d^7`!;KSoH4WlzioRPn|MVH;~2&mS{Xyzt2X?JZNd4n55nN|2)2qhHtUW&?8=ZCw# zd2@*5PtW>BYa7`D&q4QxIlr3&4ci9T`FA4sLSx$bJ#m^txu3aU0q2WZYl4@+0p_{#Ss6R(Zq0wjdoS;hS?c#~V$U8v z&zg1z{pHhlzJJ2i^Ue#doQF*=**%x`4@w`&M)=Ovq3BKUBJskJo9XkWU~>M`%ok*s znFI8vn7=ISgpKe~Ts%VmW?q!rw!Ye)M4K_%MBlF{YrQxWMFvOY_nS_ex3HG|E87e+ z7Rlcy1!L}jP!ydlT^y;Ty$bGB`YYQr{t$R0j=qy#aL1NDme~Dv+@L#eJ}*7Cy2y!% z_9-U1{jBbJBR(@ex%&-%R%~Pr8T|0xiyu4JiSD@0VXZ0{v{ucBetlBPacA7Vn7?m> z@OThiRpqk|-$;B{3HQ;x&HSpzw(L9ijtSQjk7ItfW*pvwt^P=nzhT_xU0Xjq-_Ue# z**7i;$<~-Dnoa~?=;2HkU+9>aXo_AkG{tV~jiv_gLHx+*pa?NYk#gE839=nrnc&+=)=_cZci4`1u_Ieo{~$+-v{rNOeU|NK6$^L_mW-8P4*$Q zD*2Wjw_Wlb@?dH02c2{#spPxX;TihGk2K-s-z+G24E~Vvaj&DX0B7@}&*g7~ULA zMqZGN)%o9yhEcyiu{wMoJaI2{zZv~dm0#agMJ#4jzxr+-K#s0Jz zf5mAdd>?J12ckFQw}Hl0_^Nb9IL4f7W&Retb%L|OQ|jXdH`KR}yph;Si`YtEefy@2 z`dA(I19)tM*75S7a&A1q{TnZ-P3aeN2fEX?mzR_`BYli1cIF`ZK8G26OWw4>d-42I^o-T8-JHcA9$J-c4IY|hy;OPo?Gt3@=nR_jhg4u|N@fsyxRvz> z=kv(dfp4g(uxSnU)xHgue|aTq5PybsWmGtS=B!ZkU2ra4_#QZaw=`Kwzmli+9JOsO zFAZcjmlH#sg+9iI)4U6R{x0tSC68wFPD?TQJ^mIWFTRTp=eNG(4#|rT=;ODg$x&a> zN90rJiHvH!8=k!v9=a|jTOmB6}nqge%a_WSepw;xAK9(*%vD%W35)PV> zHPf*dclOxW@e8Nd$BLqFeXYlS;_Q}p-4R7r$Np?3f49Nmi^Yul z&-S=&z8DhfXU0*`@GZ}H{*F96QN|NvJbT=_C-Kyf(C1mgfxX7L0hqVBWf_ZfYY+bo z34LnH7BrksKV6j>$z#QX%)W$+!@1;3ad8;t_sAURI4>i)eSp?w1+2>kg+95;p`RBP z=l2h2zqfYxe2I2{cbaxzH*ni?t-^zs8h5N-SPX9r3H>{DT%O+Qc_vY0_xl%kUL@UE z$e#WuK72G)jge2`^-oHZcha`c%y-Ez@QeQO$}ZcdsQiPjU(uFn>2hf^@D8uBwtj@H zeeO)_>Q-dX$H?6MKI{9RFvkz!uR1hv#hSy&sXA+5$9C?GaOXgQHK1b$Hp?O8@_-E4 z0|Por`Cf(&AlJv$>=Uk-bMv*y_7e6Vk!7XQi~D_n=vsJtHFI&5iIrI~i`Wc6u5TKe{PM37(l zK8yGsx3BH=74})z{j@&cj{us?1T-q|xb6`iF2r1ui1(tDo1ThV`QM|E=@MmhT+3g;K=0_vzEaoihTh6v3XmuwT({|ns=hR7k}5f zy5t7VL&GET0)O~b~?31U!cN_CBP9Nm5H~oFgd^m%6f_HrAS?4#+^#pV2Bj}|5 z-1n-J2@h({y+d27`xx`*ZR*-}f{o^Pc%j$dbApW@&9joweS8Ua!utOWZs7NseBP?| zJ}3>SKYhOI%h>b6-&y13Z)n9<&B6{Tz@K`<&Db8yn=YO^{uZAIJX*o?h7kL)e2(HX zj$CMC-iAGVx2_4_GG$w7?^Y|bxvc_w6Wv|I*(dDWWE`I7+-EKR7I;3`*f)V58_xap zyE#8NAoxnn;PUt2y}f1D;L!9R&Nep7Q|8Rt1)&1sjn)?sCU-3&8N}X^(edX)(+${5 z3C5YAU7fAX<$ON*x{S?~>B}}YlkBYll&MX9RaI7f7kb=F zyF0lj(5vI+V}-|&UnYKa%Wcq+_<$=10mD#Wc$&DA5x~GW4INYBpB#JsRDJFFl6@Uz zU;NAVW%Kb6Gv=B2B*Vz$82nYhI)CmDh}&b{)L2$CbCmOfna$bEnFKr)x3X)G#rhkZ z94j7Re^xr58|=@nf{dirqKPhFGJ)Jm`1#KKP81Ia;Hyj~^7#(0Ci?R!{usoX?Hugg z^D~mI)Sc*=3$5Ug^Zf8sRwzoppI%R_Ft`@1f@AhM!Fs`QTLGWIF?S@Nz;VYV!TMo? zt@;rc`07Uu%cvhS&R^eJP?j842>#{ot2helmMadgmA>%zpET#C%df$2 zT!1~TydDSeE7ifP*!0ut&<7Frs-@!#&{f~XKUjcIX1nG-KAH9C3dO(ZtgvMFSNz!J z(rN6~c;Q$IzqMsx=TXj;RgYGCY<%8FX{UNekB$F3$_pmf#^3tYFK^@HuN+)}U#KrW z9Iau)z^U&(zI)*%Sj)COo&INd`hP`j;6t*YTLC^K=rQeB)0u`p8`d3Py#u?psG2yB z>CEE_{;^xFM{+|uus5QVGqE_(Yy7a!SBG!Lmc0e}b8~6(Rq7i1gLSy}LbSFP?tUg% zJotGSoc$yXz6l=qLj1-a*x@34@T&h$+#g}q=i{KotM>XlW18sVYcOl|Le{jRO9?dK zci~$;p$)%5rvZoSyY}k4pS*)F2bfyt;p?D$>)m`-@OiHjI zTd)I`ekT~%4NtXO^Ep?1|H=dJn)sD_vVF@GTQ*qhAnm8dYZ!dGL zPWq1h5#d75lIN}H#x~^GQ^XEG#eSM-A(~2u6*AU3Vy87v?DxS&z5fdD7ipdIR)2hm z@P_sr7Q-J^rO7Mc56wxt4RnjiJyH;oUDsV_nNz2hI?GCvBTiALP~-l@N`~*}A3MhE z;b6ZnVyxl&!PzJ3NBb1S$}ncHT&ZKPa0YPK5qIK;Z(KiDq4aTKKKV{cjUDOIue9aF z>Wmd0ej8xVHkx1REcL1@-(kY0pLZ-@1pAqS{fQpfpCtCpmlOa-xpb!R{wDnglQ(t02d0~O zcMJdE)e4i(OR|>nY_O_ZWS9P7D6)1FE36v=p_#0SymJ0O)b|VY{UUw8fd5Z@dwC;`#-~~bN9I5; zeD2{n@HnsxoomxjGF-nEu-E-4^HH%t<(JyO&EU82IrM$LFBvVsM_r7Z9*N&>tS=eA z1e@s!U$T9?FL_|1kNX>Z$&PYg^26djfzPmaj%>(H+4nx*;fJX1uec;5*;|=2)5=8i z%aUb(>3Zok$+R$gz4x(Z*1G$CaB*L0^7DU7!$sOWNgt>D%HMTlree>%W0hR|pB(Rg z;=9Z~zcb!&|FY!y1>74r{aV(9R)=JuWI&eY)F9_K&L7?R75p~Y_$aY+BaBCMnTxIO zm1mOmt6oRe=0F=iw0RsIgdN;Q9p%r7Q*Sq)?byQ`$rIF$9?9Y!4tuXoF};}$AAq}7 z6Vsa)YJJbb?@Jr+v4-zO&+fOZlXR@DccA6t()ZG@+RGtcKc_Ugj=dWDcl^|K{LWmI zpTD(|cyM@7_Cfbv0d?j73|GRJx1dLnC#~?{5{Ez6i9e|)Jlu)R=%tyVSC?>rJjuZa zjlJHAt={I?=?U!gcI@;6#Tm(iBQug*q@4V)a$w*yc=HH6>W#~|*QM7b_|Lbp2B@S>Y|MLv+oSo^nG&`tC1^Ye5*5V=cT&2g1OM_;WNf(Uq!FKZk7 zL#~XKE)ZP{(`k80X|nAK>D31t-T92nUr~0ne`w#<6RU?QCsV9{Ad=`Gh<5Z3#6IaC zV2>-=1`gs{vw(y4#RCEdmJJA;YVG0OKNHO&eutLZuvL)7Z58;hf%``6#nMjp($ePE zscfhliOZTL`v5wMuF}8VI?vHBThOU4U)VO0_#xF#hJDQ1!{5;d+j`jC858P=MXe+* zE2@2wR_vSUKJwHf3o4-1^wQ)pVzq^DBa@8{Z0{|J57l-RG1~T?2{`ASbo(_l zar;~D^tXg~uZ`@*Y*V}!{nja-i?}cP-&^(*VnQo@NuPY2bHppG|G%LBiqrJ3`_*L^ zeoowuV(gAPpPhW(N<4QpaZ0ZLFZ==XWWSYc{~(woCi$hje+UM~3=FQE;+x+%+IRoT z5BF=I*}CU;-*RIU%N8a#d4qDYq{+S9W4)_d`Jn?3iF z-xoov3g)NoWA}45gLsEjTJ3je6$4lMOOq>HTme%IT~-Ob%s!yz1MdbKTL_=L3(KPpEVI3Fd(g2eosJ2Ijvc_3C{5n} z1+abeAL1hqY+b-sYs2>UE^M{H)|CpIVtehmoYPQa=qov#+c2N!0@kC-u^1_4zKo>3 zvB=^}EM$O{jF0C#`_4Jd*oXzRS&SSR8KeyJ^b+Rj74$WpSk*NGGAqzlYX#zu zTYRl8F*>+(YYj*4lZva04V7^o`hB@#}xj%4Vl0Ohzbv^m*rjXz626Ek5Jo|Xgs1M&vTQ@P6 zCNh`qV=mp#T&iL&0sjhoX!vxtMtCQ?V={cA-=4HA@)}yr#KARZ9bd9MPQC4%r;(hJ z-@(vjvER<~#rN*m6P{)7cz=&g0^d^@+a%y%jQFEi?_3lbH~||1*h;Wf@#nB-nI9?} zP`3tqB3Z=e^q#w?uY8hN6xO{auYl2g^npCxTWO6>jqh<_x{dfR(IsWh4FV44nK#De z!#C6B)*$f1d}?Dp#hFiY;W5ke=gOPwMr%Hu+rw*&v8yVRIG`;4v+MoFCerztjE42> zUkH{(ebMKIW3dT?jb8~PTcLA|`L1)>$o6S{m}edR9&4$+__@xGx8Lp@`^ZO~72LtM z5Seq3ItTr}fO4U#{7mNO=Dv@&n>Kyy%$EK|Om!@nEWs#d5$GWznGgv>4 z3~e3H-X-$Z@Bw2}x!mV{!?KP>rzW8|T@yy@+*kh&+euruKxxqlyqfQ;xer}y#ICW%4 zs_sn6j=jj!7xgNBW0}78s2(z0eLbUk!@!;30eAURA8Q%=_H&)Jz&^#noq@HG`aib& zy~9)gC#p}pN)foE{wvMTJ=U_Bv1^PA)n+bZzWC9eF`r924*b`9+F7c0&<&;z?P$!+ zeK(nQu!*M~%z3cUh53G!g??tNe2OooIPljB_v4=J-2B+vowZXFs^4*t*dVnne$prR zH8$Kaz#6Z5iq-35Rlibu@gprAH`x7|^5D>v=hKw$YmIksG=JXnHeQSD@dYaO`R?iF zuMNJy&4O>7by#vvp9W4ou{8{i$M$`!hvy!#dCn(V!yjWmvT;=^e2PC_0q68%_%Yw+ z$9ah!ejL~fKVIR`#Ka#$8v_TQ1`a-*dGd-)>*o!QP5;np>?01ZsejQ*GHjcsJL&Ubkot$GO#ghE{`oZh^SQ&|oO$)!qb*|}eZF(^vu}6iKW5`00w>s8D&mee9!Ft*Rk^cQx&m86Vt_8t+@`w0pkW?(k5f+8v2ZM$V_gdb_!0J25)Y#Fk?^bUiv9SV~S}TZ^L_Pn2M0i zr-C;EZ}x_pFT!albLo`zOn;}K1#q9T9yz{y|Bbv$+|%o>;msU#ah}7R&BtzFKK%#h zpWOL$F>*`uNo$bpemj@9cYfLVaeD5%(U~iJbBF7hQ-#Dz*!G&UPC>^YXB$p%$AX!| z#pGe2Pjl7}8%lluoBA#e@*cagKXxT{QyVbI-qL=V3xhkqABNuE`K`IFwO<`J=g{pn zYpU=QiWkkgkUrGE_Jo7TkqG<|rG1@YTL`>feAsdle>eNlMdZV2e?C?2xG=kRpJ2b+ zgEtp;*XA?+5h2d`JJonF@J4?%=iB44eLwlmn1r(uBZBGN#mJ#it+pgdbs$1vp0C5v+gR1^oO?4Hw>IU?5GT_`N)+k+g0;20-Qxb&gMgB z#q!3Wxu3ad;;$(qpOb9(w>UqhI$rvu*x$f9APv_4%i&uW-!8ld%z8C7Ux3a%#LBVX zrvNXo4;Ad_i731*J+Aw_;?N+6F(lxxcJi3iTN%yE`S0K!iz&xzm(R5Ffm?Ty^g-6; zHCVoq_jSgmM&m`tRWsgJ?B*KA+XmmQXS`YqWU%&8o`rZhZL^NixnTJ?TJaNU?2p2i z`0Y9talTPDgFT;|bt^bg-iuOqJ~2nU^U2GbZ|v%s)9JkFUC;mjg`WSK25)clJnWqR z9Zejl_xx|H5?am!7t>jn&4Gruh_0OJHtUs9q2K%hTtMsCEkWjw9Vb_a4b*xC^ntcT z&^cTZq>jJA&<(mQhHmoVEY$uF^lF<X0*0qN3GaI(U2jPkKycJBy8RCTtLtB6;HW7H>gJR-NIQyAoow|1xu|-3VbzOO! zT;_#-#WmS6Vv;Ah?DyvS3Pbe+w*UShxv;MF$FPj%{lg1q^HDRxihWMMtD;?a+~;%@eyZe?qBq0Vbl5W9r&!~5&SE+ zxMGE_CBCbQn73J5YSxr+W+Kddjg5x}jI$j)YYu5{#NgBLSa^IqeKOW`I3kS6=5O2X zi{Q^#{_Ypc-GOPmBKhaa-5)q|z@7hZVxM^LXOm4VKUD@XTVB}|&WA?W#zVyq(5}qT z&yLrSZh~KGZCD-M#@v(4{P=Zyp0%Dz{$Nx0rmrO9C3AE)YZkDG54G21_zyXU0UUmu` zYdpVn#^cq;Z!aDu%Lk=bzvdU6Rk5I=7uK9Tu)Sh?Xgk{j;UQLh_a z`VjjOJ#y8irDBeM&3qFaDS8GyCY=0pFF1Lt7kVG|JN`rBH9zk^xZ9@_Jj;416I#x; z@Q*R?v?k2r+%)lbSIa)rno#(i+2`mlx{5ObUCbfneGoj6;tck5d>!jp>xJhKr$Ia7 znJVV$v*-%REcdKn0qwLh59JH0!q(pbK5V;YU`Xvt{;b?8JBa(T;GtILrR;t?PAAwH zMmC&58?y$WqZxZoU4KK(N%fS=i`Zl4p6aoG(J={oWgBu~9sS19r7`B3{P9s}Veg}6 zHvAFZQ+dG?S*(SFE^<9Ed?e#O!=t`i*tk7Uc*P zuidkPcum^-?MlvQfZv&v^Wr*__V}KH>xKADn<-zEZRZP%Wsyq}xurQNd2PwJSbbLL zFP9%%{VMs#XD+@bsn1!c?+mlk>lZt6zLL5dsJn=|lJkq1$HmB$YT{h&JrCLRzJ{++-oregL6h&OUXy$%07Kc z9~~}t^3#}{hUaw4hcLV^n_*7tkKSvy?r+?0S>MkdxNJ?`0`7T}ya#`hyA{ZRp_*G& z+Rq6#?!=a;!Is$f5^)LOq7D9vSCs~K4G1X(qm!&*t zzK)Du+WQaMbK7#;*~WS7TI7v={xjG(z-dQqyk_3_!1J6_AG?P0dc_kAAC}uXUUGaR z`eWO1ztJD=xCBQlvcfyAwX_psOtRCIA8lE9eq#>#(cCdFI3(^fAj5@ zeGVXTWO6NM4Vo)?U&-&qK0kMAJkY3j+Ze0Xyo>P_G=ZZS_DUsfA>UUn#s+@;9QZWvfNe8!2mfzr9YfZR{bT9g6uvo_$76PTMbIHT?dq z-u1k@c5=@w|L}BpDe$#%w^q2U#4E~ZFt#J4&%K07ho_8C-lg9lK^DdJ+(534%4=@_eku9KhgeR}P zDt^@*sdL)=CoX@&ZV~+A>*wI-Y|a~LtU8+`n04Qb_JVa@E{-3d)saCtj771=Uim3pc~LjyGI7~Hts)oc#@?5NwH71STin4z0cFZr|@^Z2Y)a3g0Jh$dblUPMe9(}L~^Hqyl4eK z9CG*$xOQFI?kjad7DJ+g}6!B4EDSfnU#F__urs_)Gp9@b5Vd{I30Z!1&L4 zFqK(%J;i@#9OIr3Z5Ia%;1* zvYT@9H)M-d!h@2diWya`Rh+e%zAJ~}Q}BrF-a+72zK_Hu#Ptzll}G-uUE8ft zHuhKi&;|QOWLbf+)XxqU?AtF`(am27CuId^V!`BI;S@gpJFvF@{4}su4>7RHPbOPe zcBc44xbBXTaqQ>UBuhs1bMiEiLsU8tyYQh^jx3QLVh`h_TtH6!6n;kzxpt3a$HX+= z@%q)i=aD7d>zT#&dgfttxYr-+`eGFyrFD(;neoY1^v+K{_^CBg(^7zb9K{*__;k)- zCal?zy9~m+YapN9C2BD=dK08wY#zL=SJ4F%pLrE=QERbbhiKKNN4T?!L_+Q zt2Xb%wW}K7zvb}Xa`hckS&QSWE8Yu9cIgy`+`aBTSc)U+vr#RC*?oV9l%~WliYIrr2gAzCyvbH{A)uL zx#Y^9jnFL%`SUuT!=YahV_LGHZ0*;9&4zVR^V18iISJNhJg{DyncFNFXJIS4aHYfa zYhd!iAm5<&C1kgnJd&&tY*>mXGy%)vO!COj3ns(Ntq43j;bP*+nVT{Auo4(g#%~b( zS`N(BUx?pq;di`n>$h<`Czw17+(sk=T-@@`KFgPi-}PUBUvpLtzK_E5a~(OKj^|q$ z!%}!+1U_$@2NpIzgzh09er+!EIG4GvdHyE)Mm(_#9FNM(1J793 zm%iKn*=j!x=2mFY1}(HM)EPPXU~Wh5p zFrGGaKmy&7-sb`EchJFKNB;*K+n&f@OU?!K5pcl6dH4}0@j1tlXE_hJW57;g47tFZ z3(WRBTW03jfn3JO9NWkoGcY2{5)SRtVOuZQ=sy7*qM^&X3FKA|JhZ`;U#ye8{B$}! zmAmcKdg|NwW{rMo$}#!o@)jz8V_s^$InL%qIGYzmPeoRt7am7emiOTdK%YR{ls=pR z=o3hA_OzYz@9kD)~0q7HW4L`T?(8(W=NnXYp;*sT(nNH3U$>7-I=x}~B zaSy9_X1*%Mn>~#cx0c&yrWS%PeOJyP+wYax@M7O%tF0@DTe=TEqh5DgR|M_4#F^>5 zRd-$D%-p(9QFqdq)QOfe_?^Sl{qk) zGWLGjmXr3=oH9DMj|`YrM=T=ilWF?S9@sR+n@GP6B^QeD-k0Y(#eMS|@8;{B#;5cC z&bwFiPChLYtHM}6eEcLHC9Y(w;}^2=I7{n*YsjVC3{92S;9cZQGw0h)-GYYKQqE$# zekZ%G%a=C&gN?r;2gFmH7q6)vVCPW5$1C|&Mt;W>T08vg)_b0MHRJ}YvBsHt3-LV^ zL&GX?F&3O?9PApUcKYly$dOA0CkaL_JL-r+ly zyQa5t?moX;?q+xne8ujSzL3ppB^S=^(FNLbD8L?k4jgKK+~g#n>n3!~X4bqfU^l+V+UI3#$5*i(w^`)3wa7`39r$c<4mk-jQ_g(lP>Cg!sJnR*6sKAfPw`ye8Vq_v~h`!3_#d@oCQZTv5 zvYIpU!2@gTZk=A9s?-1F*&dz#Q^DB{2l5cOrUHv>+sG>Vo(Erm&zei&DexRA37R<8 z|0&*?zE+o1~Lwqm$V4ze?wC<%2oQ-n?{^zAqK8<)9bPPi3F>uAiiDo_Vwv z`!P-5X#cbVK9fzR{nM)BzI$#b*5x%fpTr012RXhUntl*W&M|Fz?MgcqVW!^87GvM? z%i3aY-^-o8XK(l1B?Js^P6A-Db4$EOT(5lHzXg`}$fI`(Sibh(fF;Gx@K>C-Py8~? z2eJ6eowxtpG50)rmYom(JNPFS=cM{4{>+>?hA#K#23m7+18w=afq4JiKq3d91bIXb z^v5SL2%p4;+`xyMaswGxaYrKdf^xBkk)_fl-eO$+SdBlsz zm(`40`fUpMqAg;LlGVTBy=NRTbn0T|P%fb?`cQsEy$i2qtjMVa{AQoA@-=9`ErJZY zd;N*k+ok6{?Fb)=+3=3_?MG7Q+IH;w?lrGxtgmO;V_m{M!Rot0@Js|A>;d6gJ_Yvn z4v9Xk^lALBPTxEK)25Yj2WsrH%ZrV_AlP`L7xsz3j=T$gN!b7GOTcd60_SbOw$br* ztOFjw90NY#M6trg-v!@bL+!ndvzWS1awh8-eb-q`n{HVR$sFvbf%ttK9&_6e&56%E zq?~uIPh_bXH){mOT!oyj%(t!%=3C=i$(wTxIT}jfO$!-4*szhN7FHkJ@{I_BJDXzeIa#V+Sw| zzZ^P)8?EOwe>7g#UeLa`m#;1mK8Z6@dHFhIJE*+ktL0M{zg6S2(Rh9GInZVyeR$gt zPwm-j)|Y$T_2t5g>@~R7wC>vZpULG=O}w=F3v(uaGq9LvVi=w=yv@DT`h>pgJT!?w zYuPx4ZlX8ehX8k1)w%US>==ta!cE|7x0Rd$J-TzZarc;DQ#|L?{^V|Zl``8H)2qZ^ z)ON28PD!8iF@_jpSJip=z`nx&`TSo{-_>Cycajf}Twi8gUCN$y@ql3c@N-M+M;4aV zj~-QC9~*RDecSn8sZSK2Uw>fi1@*;+@GtdZgTBrCzvumQ-hYRj7=6i!(T|)MXVf3q z+`m57LAhnXjNT@mTz7o$wXDlS^F|(9t+jb<4s&20w7L)fc4c|MmVpm69wU$9F2=V5 zo-Sft9jy#!DNKCV)rA1!P!>YXp&r@jRdDf*}seN>7*!bi9PA0hf+GWQF4`y`j&7Hel7_irlh-G)j0@T8wR z58)-_E9Jd#=dI`XAGXN$+&klem7s*c8&Yt)NVzA^#+5{dq zF#o>{e~m{^USs>>Dv|dQcr}7vb?JO~1az)|&R=ruu3QuD{z$Gc*>{Tj7tL!@;g7JNn(n=6?z-j>sfArAx7&G6?C@Ut6UNx)m}yvxb5 z;k+F<>w)uc9p3Wg!TZ2z<-wy4|8E8_hPUA}pEJH$)SHFvs<;Ws@UKx{&qEn&csl1N z7-uPRq6~ibp3l-)B^#CdWjOoNW*^&yy_|W)*hd0y=W@?{n%*;?X3{S*_R0|syw}ja z=7Q#vI~QaZyS(S+G3DoJ2RA*qB1RK`k~#n4;3}8C-=l9ASK4DI{@l>i^&#aream-f z=AozW-uiWI2rn#Ne!t}c_>MJKIdK=t0fg<<@hI^#OUeDVpL}A9bHAtgo@d?jF!AwM>q6RX!T2h7|udh5YP#dete z4RWRB!i$mo2OE9Rg#GvxMaMI`@1?=-w8eWbjThfDkRPJ+)50(5;OE7k#R zFaP(}0bR$_;%WT$eRI0A`UrQ8{_HO$^>rCmNNd{gGGMC(Hf&e)zVQh~nBT1>;Ey=T zCFG`h8k>JNd7w9v2fCd(1HatUW-T?jsbs4w)=Y88nwQ@1?bx8L$UylibS_)A*4~dy zj{E@k-~_k_ClI@vd{V40OsogWc%IhC!yaCqTf22%DHE&JP_90n}|*t8%+L^yG!uVbo*WH zoP?UUPJF6A(~j%B1bj@AJn4~_=1yQ}sj|eGlFO(t^v)(|GljC3AaCT)XD)Q_BOz;K zYiZ7{lMi(YaPYh8r^M#DeR}z0ko3WjhTGCLJDs^IRGJIUp`bg=c3^6VX$?hCY6C>I~R z27kAc$0L4|mCOYfik;V)=R3KJCV`zCCk{kHqiqZhF7gw`C6Da3X=Q@1QC_$g$!M*nzr{660 zN2~j@ZWJ%khU$wJ`cxlv7Uz|q{ZjP}j&#OaF_ud+Gi!Hoc4^lP?&f7o?Q?HCsyiD> zko&sx{4cb9onP@Z?{syomX6NkY=rJ(IxvyFkbVz$v}=sCDP3>? znc%jSiC%Q+DjxKQ<6(HDZ(eTmxW0XxALadHyuZX3Os?SnBHFx|HnA!9-iht=_VKPO43%k#v6+p{k`gWdKH`6HO0A!xJjKrwm- z{^HDEC_d7{e#+~((~-I7^1g!kTR+U2c7S&G0Na?peVT*xHwC!MfZNY=CeK+s{{`Mr z9)#{Wx`{g-%aIfCUt&6QjQM&1Jc)neBa!vTbN+AY-UYtO>dO0n&N;b45GW|9Xl~r` zLW>uo^Pij~C?IXMBQv(+Oi4(9P_$|$w-9I-=7ro&Ct1lv-oW2G}g5XDPu z_Fj9fwbxpE?X}mw zR(+GV^gyvMqgytgWcnD>Z+7=;WV7Jwpzq1q@XtB)-HfhF$OroBq@P^wcj^a+=Dh=l z1svdauEx50j5XICr^fhcL!0dG#TQ$1F9(LpbB471fbt(wzJsxUj=#@Bo0}cld>PvO z9NN6kd_6XW^#C;bJ@fI30ClK)Wu_H-ZwjA-E%W|V!?))(9SaO?x&<0-r>$pc>qk6` zCc@1m2Or9J;-~KbeUAl}u{mii>9qL}_<0z9Sp&bUgvPbdSaV0~6+f`#0mC@@UrD_J z`X5ipf?^W=E)TFyBWWR&L1F53B=2gV&97)7VCz_|qbPI!Es zJfs8RQD^8<@pucc{D{9)JpMrQ-ofL0pf@U6ZS8*-9YkvHZgPq$@fVkKAL?0 z6g2qn%6G+Lo$H*<#ts{3&K?>3@#d`mFF0q9Xzn1tEab!>WS}1z*n&Qqk~6ralCmnw z7EosEph0~)2pg)O>@Pfb9fZzI%vo>W#4(???}QFQ|Ap!6Ty>DK?ty!!X~?HIpJ ztNwEDE$Fw;q2G?uru5?7=*3~^#lQczQ|qS%tk_iAnwn|FrVQqp`qIhM?w!>84)49o zdsC{f=$)5s9b3cvea6ac$>n(tejnfo%>drSI<(BTqdC7F(z(CgzrTCCf{4#J0&NpUUIsT`_f*8zZV1deBjPvEabO3P*`2vJCU=2u8kFDZF$1* zL3UHvW9w*56-FP%Sx<%0lM(D>okukL@7QIX{HAJW;<{zHplRu z^Zl0lCiVUtd*Wx<6LS_{*(=*^B5Ra(`cgY;-}D7c>PLO-h8A}oi$k-l?ri34c6wI* zAZ!@X>oVvz&l=RNymnoMtcO`Yy;DY98smP1aj#_DlNt9-j9Yn;3#-4>`x&3F{$18l zPxAXVe%Jci9)Fe@e~r#Zux3(h+iTcAI*+FF0CVnJ-1|-${tDjt7Vo}??Rm@MyLumn zrjI~V@qv7soxRu?tj&MhxftZzX@llxOH%$v46(k zbdZ^&cfnikn&S?h`{8ne!Q}xLe+gQ7+9Ut^v2VTupG~lFBwPw-dEiobngTw%!QUN> zPdwwr+vS|c^Wx3MSvgqR-v@d`GYiz2`xdM<~u*ZbWuW=AO{Wv774Wd|0~fj`YFx ztFqSAuSpwRU++t+FV06+!SBV$tnJEog{pY8ixnOLPFBp3g z8SpH!;xo*Tcjs8Le?z~YX~pj1`+Ly!Zoa<{?{`-ELS34>#C^(^y2bH5=mxeV&!xosG|`F)Q8pEwpBzgl%N}cF;*a zyH5027c^E50KJz14f6P#z+5a~F1kLCLdxX#7)RL&e034zNE^72jbQnZUC1iQKk1BD zS&JDzEp(F~Tfc>QjzG^&Xwr4Fx>>)Y_=<8lAo|^(Jg;wkY^tdsQ z49`UJJI)%C)#6>R3l_l&qLDt)#k1zC5R2=_$pfD`u)d{nDP{7v!VO8+~@r zMaY{A`7^N?z`38s#(7D7TFx4rN#G&&_})~+0}#L(|OSO2>#0)=t%I&_EpbDtsS=ldg=1H_EIJHqzQ$!9f?f9@^tCi(uf;>#Sr!jn9ojYgtP z=~-wa``Jqm?KP#|Pthj1j0}!4e_NTyir3!;jTDRW4EXCpzc!wl#TiBXE+3YxbA7K_ znxjSr+jomfx6enmbs^KTsq4rK*gM6y;%S#Tp?duHhk(ta*>gk+-{I*%-3zDo_4PCRL>l%EF z@|k)4ShJa@lc2Z!v6n;dE1a*XcDuB_+|b#cmy^$t&aZJ!_Dpk7K6=sF{g;sECU;ZceI)JZThY3+vX9o< z!%4Em?z^&AwsQyg6>i&ahj!pxdpM8uET1*AuolupNbGjV%uU zBy8~s*y8sk+2T87i_iQ6{Kh;y{u6BRX#?2eyyM#9kEQUNm-i%J9)oA~`Az4u-uYb3 z=h+>If7Cw|Nr~r-x|G@`;9@=M5c z6XmY6J^^j*`=tC$L(Q|je#r5ZXP*Q8rE8x4;hAMntw#pA^cN0N?-y$w#rol5svI1Rl0_~HLU^Ka1r{9wb|r{5$i;_)#48{<5nBS6~s5XFzVg{cfK^TCjIl!+=ToW zd^@;%Ym;i4kI@4Ieam3CkGtV7+tl&4iTy4&AkgO9%ZH`U+W&_lRL zgBGHR^j(GS#C(La@5C?u@yg^oHNT$%$Ga9`V3tvJap-D;1DeO_c7qvk>B-+ zkn$iNNP%Z{3OwKaZ@@DtMcxWO?{gn;1Y5`=_l0uqD!y(W@pbc?S8ga~t!K%vpt%Q6 z*?y6nNussVWHPzt^y z?1#8?A1L0=*c)fyH~9~Z4O3?G+cXb6f}!kV@Y`PspJC?UN5SVK;4|QXKNX+D{s-`R zcyON#5kB90)|ML|W&R!+VE*oqJcB=*nZp_IW+rsZGBQ;;kaf?3a^Yo}GZ6AOzTWQ0 zS%(*2^uU#hr+p5LvMpP&jkOo~+Z4Xb_sF*2IrH3|-$(Fuq{{#Ifj3Nyu@6}Nz$$s8 zyoZX54r4Dw;Q0c6%j36FVoihiDI@foA4InS1Aa<#_ml3Nk^KHN^3L$>UYl=MdB!6= zf7=;@;?lf&BSK8MJ71J@*p^K8{|Wg20{D)k;QJpv@TKDWOaDK__ax7FgzuvN2EJc6 z>x*Q4y_Pdk($|}~CtCW+=wW!e4gIL~nPML8-2Hub$|TAx`%W3H)m3+#bHBuL131=2u9(>iq zgD(Hcr=Myc3}gOS&is*$pmAvrKcC#D#h!b05@ik!yt4h@JaCIjUWH{n&ra*nhH(^PsDjZlCqQBD#Ii>Hi(KPls+kj}0jO;`aBK z&>!b_CF_q#9+*<`bIkzw`K=49gP$|$kB_o`_#yc&46YnK5b(eyT#ayG`6%lL+kQzp zUr~hp$9+O+Z&Y%p&{pEyu?dSfV^PHY6Q(SnGVY4zoK~W&lCu?o9Lo4}>+R*dkpW*v(Rb{DLjLr9 z5B%>h4;JO~S7Q0z$YUJMfq8F>=k=*JU4EHu?BTxgyXQxA-c0+W^Wo13KG-lm@p+tm zxCwn(M$QiI^XV<*Z`?3{eFlG$w}L}Htu4T%eMy}q`&RFX^|C3H&ui0%b}X{m2@^Bs zrKP*?EIMxX;2_oikP2_J^1*Yq^E;zYt`F091bOMx9;eA$H!1Wu<))1I8t07RTAlT) z;EbVc4*B&jCmze#ANWA?kn_J$4xhpqu>x!rod?tR^1?LRc6NAAaDD?gf9=r7#b;y6 z`a*V_{EqRszp)kABDGeC+#63|A6=t6N^~Y<1voYR!*`nn7yj5{WT9lg&g2xI3fvQc zruwem^4xW-Bi}?cXXFvzR81ItuI{u68vud_tk}x`F-(aiYHTCS$-ZlBKXvI z&PKmjdD+ZEFLU2ioppJr;oi#L@<8s}wI;XtFmu1t=1UnD_UUjpXHJ>z!z<%nqRN^} zOr0S6N!^UmE*oRYHen-lB+4!{Wfrl{;J~SOktr(&*Z-9$%Qa=q#1#B8QFehT%R|Q; zq^#2Pdz~rULyVYU!$z>rXC%RPsog&O{BokLQFi;p&FxNL@hh5m*1 zHf3Aa=el9o}Bpd2F=@i@R)%2TGpe2 z_j*3BMWz6674UM-`0=MW8(7ELK*exvEfv8ZUmYbl{x^ey=MKbqKE3 zXU~q?DE@Hf3FjNH3f(m7GWc5-Fn*WkTN zL-;k;<~J9Io}$d#-ZiwB?`>~J+G*Qc&%5GVSN5B8v&g9Y#y%aW_293Ze7y2+*mnu^ z`9~EOB>%pAq`Ctsf()F*`MEH;Qrf`rlj+uTQO2Sio_WmYg~-^2$k;{b^H670iAVrNW36BWHUJjvcYo!ys!e9EnZkD`^A-yZvT|a_7h;79n{3x108$T0{Wql`s38t&y+ z2i8a7!BqIG7~87nBw;hj1U zj5Vrr`Uc&gJ@M8|b+j?l= ztOX>a!o0VUbHqXPlvy7Iea9war}n>Rql|ktxT}Nv0Jfg0ZIPd2p2eOPagfqGTfu|g zuUlHyP=#-$j(T;}tIJBOuf+D7hYZRlc4aH{7y?b*_Z7FIJ4_^#YH7pNgBR3RWzct^ zminScm^$*^CU9Hp>Lh2);e6M72A+HKdNrp_4g~NK#eNJU8#P{ydp=`sW?dXcPDU7u z=6)f*48`rG_K}M|%hK8iy^;q!1&-XD4=-(jmwrkNz&LnGYh#_c9OvZ;2QJw!Ux(+a z_!GbAdxd#c%!k$VMNj$dp7PK0oT{HrKGipm?j1nB9ILg@=e!PX`^C2xi)XUnnPu>d zFJ!?p7CfVv-_|l@DfrorePQysVbit(n{w`uf98E_|FxkC_QlqxVdH793LD+z7nq~G zPvmvt_bPrb?&Y2YY~W(%XFj|n8VGk+zjtE&CdvD0^mRKqZ@^Oo+r9*xx_StCb+zK# zwBB3*UiC~HiM&!J;65?mYTfulmj}-AyJx;VZM@@z==&ZX4fCJ$@^bkD4_`ShRtby5~0f0r;k_@Qof%;QLp<2j5P9U$e$>@da%R z4WPN!OLInXpI%npG1+e_5A%B~zsV<|XSZ%;PR-+5Q)xb)xBwo2hUL((!lvK;3dfH4 zfc~Wjd`}UiY><#iuY0lpOtqiX-=l7sX3f^D3 zuHg7f>xy$$c9*c8a_uGANwTZh&uZBw-!ZzR)Jm_HjYVGT1M>IE#_9kMhsRj=JcFK< zjRil?y%9aTiSPNWg-fu7JDBVHupy+I@~wui`%kTvzM}YH-OURx9=1Z<_3%U< zzQxGlU{PyjNf9>w67gDME;{2#9hYktwN|Pv{Qn*pS}TJ^;nUIcwD6Dz40sm)h2VMS z9N}4v9P{ok${xXPhbnN7ccaN#Iped6?gY_D@36hE6czfSvC@rB%loLa>Dm6DrQ zQ5;*3kJtl79lcDP!zh0&C2b#Mb z0^Jq(&Xj9f{*BE%OHW1@VY`3T&O=ngne7FvIaeTuly4XtzPq0Bv*t#J_?j+7zD1Cu z%4x0nyN7ve+e*W+mEe2o-2H9V+D@At$ki}K<`ku z9Xmw*7PEJh)x2UuoZJbL70Oo?nLzA1a@|i{xgFP<(e&@qe-og!)4%K*#Y$HoPZ!wh z_=IfOK%Bt@$y~-JnY#%2ft=79v*=5}ScrdGzRp~5RgSD%nkXA%%9^pMzmX`r(3Is- z_LW50MW(Comzpw*vM(gc zMwzm5a6+Jz)5mC2)(qWlPLy42$||rYgYs>bA&^c*}^ zMnCdJcEDR5=w10W#?#NM_$0gVNoM=*+u4av@+f*Uj!$xe?-ApZlrK_wijMNf8iPBO z`@So#TIa)b@538Udi-U39Xp0O5@9|lC)9N($YX}=h}_H_XULs|+;!{Hu95qv`3_&N zP58@#R_NFD#)sBdW?7+MP^LMQkl91Qr}D?4@024)a_AqP?9(A8PD8n((WmVf6Vv1I zk2T_p{`1hj{dLL9jqJB9!M2yITm^3`&s!bq`NK=etp`s^UZSImW~5KQX5(n&)TzJuqmk`w89`{F%VtW%*l<0Q>AytJiGfIi0zj#hf1p{MgC+tFg15 z!G`!A@ef*e=)987Fta|5p%#kA8Q+FReg+lxbjAL@4 zfh%D00ZFF`p0Tt$4tNAZM)zjcHi8pJk&@}k)Xm#~;oll&dpWUP!Q)|FiE_s8hAlh9zW<@Rv9GWi>H_miCapTRwcbIevMUwtmo55 zLjyi#^QWWFcBffC3Nn^gke%i7pOZK8{^2LrS9ty1c^4UfxABi}x+v6&{waYT;t7{t zLl}d6zc)a8;c#@wxr|g8szwKH5}&@&lRqTV~tbF5j)drpT}PROs7lJalyI zojjL**gN0$*juYp=rt~_`S078TlbxgX4~Mq-N-2~4u^aUnr(7$cvhOxUnEl<$w)>N7;58emO?j49`)gO^&6TLJmwC>ZK`|%cM&J_mld%`o-wP((auwE4X%3S)52)=ka z`Zcn@{^y|{d5P^T+P#uv%gc5?jgJJ|Uj7ov7xWl$oax;|&>P6jwUV2cCDU+6z9Tn{ zOh}fSJ6yTBn=uX~HzU~Qt*jAJ<)`jc_R!MwdtNdvxpTypp9A&l;#Tn}sq%9&@-rRz znZfvGqK{?&?8e4vMSd!$p?zLX_XPW!TzYEH{>#v`BNtl(IjVfh?fmY--^qJj$X?le z7B)u%dP4i<+Sjf{&T3DI_zKBdgA;6}tHa1-`RwimMm-Bot-B<5Yf|KHDi2)i<$=Mh z7iD*ju8Pi{A|noHhMt zpFY`tjql=~IXK@l4%zwIGfk48*yCP&mt+R*cxvFE2kbge7$n}Tt?}gg8rDuLnQx`m zVAgGZb60a7u;*h3$mfPFV|b|?J8~0tWGDOX^7#h&Jr7#RKkTKS_xYJeZ22L%aXv7g z?)>8saxQuIPi%jU^_*#|*4ayttYD4T8ub-L%d8l2bg?M&Kz;z5XY)dLgqRD=@haf2 zV!a^0&La4J0XACt`N$mB@(XxR{@a1%&9I?KI-|e5p>4-bOURq8;G>_s!585Ah!;5a zn)+RcEEq`M^qiNZ`}@lqPrt6bS>^QWwbw4wSm|RSeU%_@7BR==SJ{$~H~nSJt>8?u zM!plt8Rh*D>^iU9j;vA4b(FT+WYd99#W^V6RdS{cxQxu<{ZyF~nPAJD+ksQh;+d7e zqdk}v$Q;@ECGcZ}I9Ho~SA=$e2gxmsu|j(dv@f}JlQUL*FW?=ukvEZ8gQ3z_lR{f~ z&-})?i8(X(YWk6d0Y?sMFK9lzdll_yUonmT3!zJ36Mi$~VPFhnWXyp)KBFZ?fjnPP zA^ixJD&BpWcQ5B%)-;Ca$HGrznU4Wr$Q!M(2sg;HGV<_nCq)=IxzFzj@nR))Yw-zq zdD$alv2Cr;?;bw2-cx3D@2@D!^{)5jqjlu&{7oLYTN$LBoDh!v>&E9zxPz!5kH-e+_`NJEZ@1Cn3Ej*4CEs-&sx`44Xq(Qtmfbd zXxbLvcu;jzUPpNc_x25;eD|rqF_k^3GUU1TId&7z{xaX|npbV;KxcF!WA`;bykT4O z#trgu6^vsa9p0*{&YQlM^=2jhhuTlz6M}!GCtdlVob!?w^Vh<^@VVqf+7Rn`$%h36 z$Omi@BOl=Bf#gFLeCDw!`^yK~OqCA};If~5kS#(Vf$(2&!PBoRAC@`&5|d%y`GTC5opDAP!JL z3xWT4fBu*hdy%$WzHsStzC$0c{5CQIzoPtEWsF(-T)LY>u^`g7g^n(E_qNlVHVXL7 z$V68cJ8gK|n*p9yRga&pHn(v`Bx?1(Favzv$hh_qJFpRdT^D=obLeXuXGE7?;j zqz*RPvDxfbO9tp!@k5FuSDW(hSKuEE@*8oc=FXKo=w0Z@Ue|8;6|ocYKjra`>;~=A zd*O@6GdRPX**No_Q|l{Ohve4~6M@dxUUw@r2}<^Oe=FvEHy%RP;&Y22d_RDta_!nE=a}H_1c%?>@ z@2W38NBc{~(B;$c^$K!|$!DneI{6LLka_9QV4LMLe#7zj4J&yLz&q*Cq?UOseXsz% zw+(%;1bk^;YW=2p8Gr}tu&uN=t@}MBpA&w#GTVlKsJc>o7J4Nwy=gi2M?HQ)?v9!4 z!u5UNa$!;%UbvnD7R3ooV4k{g$u|^b{tGTQM(73yF8O5@BP6)m@Mp>%#gAxonmgC7 zqptLR33avRnL*uz-_wCPbqq;;b120FQLSPdPJCkvYS_r04R?{^ti4(XV3IEdIIcsYdpGy!}ZAw@$EPmDod- zmK~SwV-28l8Y&M!?-KrGzfHg{DR<%yBzvWYG4PQ>nc8XTUIyk zAt&uU)e3lQC4af^oaVd=_00TctzrE0;9cuFH-7h%DfoQdgU`h&>lcmh*b?j&)}CRl zx6mu^@?C2m_PR}6MFbm2JSo1)ado+pLT?#G~Bh)-+FZ1s+|2 z-H?xb)ZEKkI`7c?z~u6Q=3$x#W^aCuNh$fKb^oL8#f;8jyfYbh8GBNAv?Z_V&*7{@Vk+zrF8$?OnR>qnG=^zt4H!!2QF%_k7H=0I>y<#fpK6VsE>-kUEg{ z%8}T08@2-eE2}0gMDU^GTX5EpydxXEbtdo5;a%Qo=baATxs!HnKC$Z_U>=-PKUN@g zT}nSq=lX7qFGT*iq<*o%&#_-CG$!TU$d#H8+{2o=aRY1kgWmDDeC+yf+VEM(HkKbs zy#1N)**eCRZ(38hdxu&d-r&qncYR^!SnOM0XkFpP*dwioVCRcYBa)ai-A7SvbV_jNaLrwQFX0S9K6}&5iBKzdc}r$Rk)>uY&-?pU z_DlIa;KY%uP9=49f6F86p#>$2@g-i!`Q*($dru@^Hrf*t%^plD%n@WuD>Bd6fyffY zQ#}Mf?Ze+F``4`Vu%m6;(2;SD?I5|NTn}3BwqY~MZfOU1?tZ4Bhg&yA9=`j}$@nK8 zB@th`*yxwO{aVePR_1(Z`E{XE=I~32Idq*Z1FWf`KfZQq{ZFM+J}9&8tXf+Ri`EC} zOXnY5eRA6NH>I@CcOzqhQ$xRX+dtXA{axqUemJFlzMJ-IriNZ|+mH8ef6KYHpLcyS ze)w+MZ=M?ZzT19aMnC+lNohaZ#2WUUCy64@+a%Xzo5~J_ueHwCxuA07v=2F3fW9R* z`oK%R-WRlgTY2yN-paJHrphe-hVhrjUy#2VeQ*A?V%j+q+EO{%U+-rQNSA5-S~Uh8 zMVwpJ@fr1XJS#8o?wq>rpUJj6RkluYa30_1^IdB@-O(p~i4B4boD{m9^RM!I=`ODo z?7>DRuonkx@}FtHTk8wO9vgnfU(rr`pM5lL)={Y6&O(Yx*T^xe#XT3g0`hj=CN@<6gld^!<2gfp2> zgXte#@5-bN@Gj<%v4@V8`OW>-%G-XtEcrgWwvEu{F=(`&F|0$*HSir8+4mS`G?A-y z{lW`~DF-KQ;4BJF<68qo@dqTgxRaIn-iA&#elh4F8(KO#1N@k~=4)wBb@fgme@pb< zkVborVAqY#L>7Sic5v8R&wF{w?ShvqU-8==%kkw| z#FWynvDLPg6+!1~d{)il;t|=FlJ~@x&uzVnJdxhI+>19*-De#e&Xz@K#x@hbMs5YB zt%1urJe9re`xJp{C+*OK=-;nv-(Apl~E?0PQBacFAR@IfJJp`^*eLS{rzVBUWUHL zrdG_=Q>sVmp22q){NE{?*P*Z8QI5|p(@rMzw)1_e z{cY6K{(Ms1Y4S%DzpZuD0&uSV4ABQ2q;oOX_#OJRLZ4Rrn_4SVo}eRp1E!vrj+^Kp?ku%EpuoB9A3H(G_Slm|CWY<=a>8)()kmed8WChf3Cim zowEKYH+QvOO1{;T>y3O3Pc^#Ll@(g^{EEGL$qMPDDrBzYh0#xeG-K0pmhSyJY}h*V z8UEIOHet!Dwmz%iUD<_|*o8INyQB#ZZ6OY^)G7)sn23Q}h4X`9=En zd&-TCaCK-H@P{u(jyd0R^nHYFbNhyc5BgX)=AN<|b5HsB^Ya&IbnE(5cNBYa@=06c zLmy@tU$!eNua1CI?Wf2-(3<(X;6r8k;6&x}-{`l(!Nf$+j~`hYK>nKXU;{M==Dn@) zL^qsTpSs^OFg-2~{TDD8dTi=XkFR^_fsRekBOW*>di-{uLl53H^uWG1^w{LkBYbUW z^Z@kuG4Q9-BOCvb_dJ@7(`lh+fxnL)(?ZYayF-siZTU{&DY86h?qiiK^nou!n~^RJ zr-d5yPJ%9%h1T=k<;%_5V@RRPw2w)bWsK|1k2GIKhrR|Z6^d^ux9M||IUwD-hxsz3 zk3RdWu(2Ha>@Po=pwIr4`LbVYK+_&;z^~i%*r# z&wcs7!`M|3?1ENon<#dM;!UMHiTgD6XAQd4>_t%~8%DOT^r@>S1CBn`+AeTj?+a1v z1f3zSqOSahs=EN4iZ13VP1`ooKK-_w-WP7b27d(`d^~;byLUnFzO);g_GR%mjK4ho zg8bF!d-Jb5vB9|uwdIx3>Ba^(ZF2|fZpE_V|JIqsmuYJOZEd@^s&`x34Ncp!_#4Jw z9)CgpYV_T1tDSv_Cr785w&a(RkNq-a-CpeYN^JVQ$II#~d5%uNhQ^jx{J<)5owsAR zY{56xT4?sJc38yL$wq%}H93s1C1qdFzCdU59#r0^BGImp`pfZ!&B7P89KY9A_C{oD zwPpHRw0@}5nM=#6FTqC|#nx=Ywva8n6*&{W9=k$vkhYC2hn>-mP4tw;37(Z_x(oYC zvA22PI%xPTH#7sj+QT_1@n4DJ3{Pbly^qe>k+1k1*<-w8>o9v?!+rN^-mP%n&F6et z{@iQM`tBrRPPK2Nc+)q?gWgTsicgo1TQ;(6;(YK>ob#}$vl2hmD$e;Vh;yzQ8*W9; z!!6kJEtRM0@GV>Q`QQQ{Su8qDxCk_rySTviF}T3aYlGKTfrmQq&;bqOFI``&X#XbYP zWa7Kl<_E711#6%^?<;0@oZ`BS4V~3g!1EH?zsa$zd1nr6HWU$w4c>9#I&!omAR&U?T3suW%k}mI?uUz{nvuBd7k=Ce>=6^ z6VHAGI(y}fSJuohGA7XE%9vK+8h&-~Aq2wLhgiq46l^ZJ3Ap zlw6Yx(7cS|&x$hd6<<*S-^pLr8nr^NSH{;*UC26`_A9V&=Ce1iF(^LaN$drm(|-kf zaN6TQ7df`0J?~2JujQ*fLt}r_p5>><_e}%f+x15s3yd=B?bXB=91mQWbbowDaUd3+ z5Qs&l1Xwc#kTC(~R)D*VC^vg$c0P&Ub6(=9bmaW2e&Z9AylZ3Z5zWK5{OmcD#a1wm z_AAR`TcE>M@EM*`7OQ5xUZK2^zwwjzcX+75vW9*5-d{IBD|#h=%sB<++7IOZ`bZ$Vg}Y_CksIijO!lF(n>hb|u+++2Q;cj_ z!k+6qHjNAYs4-mi^8Z1Rz!GrEZ3L{s;hk2JWr->HM|de@J^L=8horwa%?{9-md;$#W-jx>bHS=C9zDp0?#| zUg))l*}L0YYWzFXZOOBraUo)g9n8WgU$Ttu^Tx#@01#%|*OFuHC44)liWK5AmrLj30({8(5a!Gr6 z+kj?)-$P!7w3bzjLuYC0$j>MrV;f^ysWBmkQpdE8F-3Qmn%J32-Z%G@VK+#J zxPFk-d4nI6`&k0rZ=eSsW_+3-WsGYk8&7d>wZRm&aZ%S#`YHEgF zKF(H0sdsEbRd;~>Na>X%o!xwO=++0Nr!$(07)u&wE6g4yZHR}K=UK;^ejlGZ zj(dv6{XYJ}?wm)uqsW(m+9@&Zq&H2V9nttqzM@X}piBA*UA*_7`s-rvK2ayU&;^fw z-J_p%FHdSbxb}i{W^XJsS4@c30@Bew{CUOX5!uWha0X@gMqkWoT6J*NQ&088t*j;R z!`ywtIG6Unt$SAR97T?wGhZ_f!`Ik+8nZi3HD`ZI%!OhAtC%0M?>iVrCw8E6C5x|z zK#LCgdI`O>mo>r5!1jvf89GF^+Uq}Z&gGWbz8ah7IhO;^KLXE7wo0eWg6Cx?Y7Pgl zjIXc!F#dcy?e^n4+P!LfeEkZ`i5+y>1!jC+m)h%m>zdG8%TBG&s-7_2wFhMjP9|R~ z_UVDAPsLw|=RD-)rM`FJrDopM-t$u%2j&6Md*%T2UgNC!68=CF?|)$;f4%m-$AcT= z*9Mne$QsGNY#;a5Un0A5`MODeys3T?&l_`Ach|Efn*3qey4^five&XNXHfmCqcgcr zk9`DUjkbaV%^!Dve?0h$lp%L#TIdif)`9(?`@EV|58t2GT3WlEN4=xaT=1w|wz6m` z`9i$CIkU7*bjPMT3cc;TSI8dtLw-{2x7#s;gPQg*PaSgU#kAn^d;bT4(rpK@>@16_xX<9MBPc~ zZPi!&y!JAWA~ThT|NGxnY>M!a-gM1r=(&E@At^sv_-DfoAAKZ0yT3qIDKz$RqX*w?ny|E z_nB?**_VG+@T`>`44+2eZ`mGNI|PtRwta8oQ+m+A2=6xs$nBa*{lU~9O8pVo+^j>y z=YsPyr-KvvDxbT-8|x6=wNc{92Yr~|QvE+R{*vLH3jWw1G5D(l&NIRWJd;x3`6&5% zvs3oVg|GeeFQ1s=+1WdfX^rJah6`Uo7f-bV<4O2QPk|>&n{6&$w%T~{d1%+U1TU+d+cJ)U`|eEg~W^SYUT$?&g*e&Th- zFKQ0f0e8(=;1vGW4*;9ymlvKA58ZTsNk4c7kZ1b`kY^pm+##9QAD@zG+EdK?h;q%f zxfTvEhJkD@-S=U}<;XXm2hZMkH5%)hC6g6TDg5qZ9L+9%180`Q!t?)g1{f3h zn-K5Lf$L>B8tq{T45u8 zQT(2B3y|Bg74@C9@!WiDx%vF*Oml=dLHYHxPS$)FNN#GJ0na#vZCX4;Gr+$B zUOFB8Zvuaf^7sq@zkwInwYK{x@cxqXy{_Iq-TA9YCVySudCi)k@)Kr--t?VZ&)&xV ziq4W4xN96cvZRr7HH~Q_eT~AM4<8dgsQvDqRXOn*&YsGSigk&&)*a>R(qe9#FVJSi z$j1)mFWJ&l^5W|~761PG9_7WXIKSaw{={c`HqYME<2x_hBRTCG@x6oj@BN@Bf9PvH zk?$Pqk*~Jm=Nk`3?%dY1=X*cvsR$nJkxYmn6I<~uXx(<`v&KJ{fa!WKOzS-`t#@Ja zg&qec<=)mkS`mD4v{&S9^I!2Fc*i;3+h)C|&HBDJm(ixL<A`M#Y|}eP7TQ3UBp= z{u-P_fUAJ^S{v-OUoPvvaFjep{C?f%?A)~BGIF{!06S$@Q`X8jZl@j9Rqo(;JiXD> zDfETD4xbg*SjWV7g*#SmXrpc}@@W%tMLr9~Q)z$Y$HY~-yrurcR}1wnxRIR{<-Mw< z!6IzSqHWfFhZK)owbUxIW!eKK&ykTuW$6hYq}nfnH{7)Y`z`qMvqPW0KY1M|n*LJv zSYSU$HJ;Hv`@jwcFpN6iY@pj=Ex`=*{LGNfF6a%Ne;n)=Gg!duf z9ojwM`it;Ol?6ik!C^cOJ%|H}FEa5SJLV%76blsL_a{G%Zwh)S=b@q9txg?ySvXf6 z)={B@x5RB8*Smbz`0wOh(Zlcpde_j6a{0^0Q@>;o{u=l!aW^IQv5{rb5_pq(*fw>v zA-*VH%9$2oqNCf1(}5P1V`d(bouT-qaePL}HAmh|^1T$U5q$*PSYi*PdsIG-PxrU6 z>@B3Bb4)p(iE{Q3(xiV>pU*@&djx67Qd7?7R{9_guu11)1_J)F69=j1OvFuV2xz8uz{eetH@Yb|_CDKG=0ff>FwFax83 znXwp{Ct%w%Hu>%Zv(JTd3UC6e;5>D+GoA^Q@m+aeKXA)5#_K7gZVzSha~U~Q+kZU& z;Tg|B@c^$6fPd}vBxnGRTfub_4H!qdXu!BK7)J-=)i~br!Y7!)`5SJTLxY{}SQ$@b znmyJ%8p9cA@=ebe`o(9Y*mJJ^=E7en-T{6)1~Jfd;~e(sfW_!9@dUVB?3OVW6N}5| zs?j$vFK;w+H!&}#+w&5gzIHzPT4R1(u+!e@Y|<}!U{AfTL;OD*AKIUX_W9>bJY`uQ z{tf)l!-f9>!CGv?Z^!O4hO%I6x?2ui6jSWN*v4-w7*8GLZL}kp7XY)X_u}v(HbO|W zxScliyZ#iLX>=vOeZirjDThvO`K{h^@h-BvmHM9%PI%wONq?Ta(1R0iKI>z(>BgQ> zzT@MynaTFd2UaX{58H`<%fz6@yRxSG&;SYe}B>0H49wBFt&gp7Dx2-TU zR6`!ba8$bTLiEx_jpXRG?VQRCW5fKzuQ=O-4HI8J3Vk=aG2TM^QHM|VrkgsqQfJet z%xf;9&*SJn?;ht<#6D^NLb_G9-sg$8J&qpp*7=@O$H3_5xeJ^#C4#+^Jyd%hXBvAV z=;VAcavU4t`g;eExyYFmSS7o>;L)y^Kc&V(pWZ~SKA z3fqmC>G(HP2ODEKHb%Xku^X0SH&m*QZJUdaB(n{_ZiT0it?jJUY(HyyQ}s>oM3nO- zTcOX7v3L01`hso4m1&Nyhfb8Qf>+x&a&Dc^j&)Yd@NI(?n+gARV&51ZECk22|6^=5 z^L;PhX}fiOF!lqttQK2jBYmyoz7n@geBMCWWBBwbGkhr(WMYyTPg4PgJt z_LB|h+Q0E@MH}cRn{AGsp`Yx(uc{6?F9Y6cepT_e!a87 z&tAqS{9%iYamM@rG@mcMOqrn_bX+LfK`TQ$Xt+qUgKrG20?>-KXE-!5Wm_qu?N)I8 z*DkFpp&#@Vt%}^TTIdHoTkF|>5&iVtq2GSNhn*z356`i72!#GXnX%J=&n?^MmWh76 z-%3A6+%nP6Df>64Jwv~MW8bIHuk*3wd3QScEoR?jAp1Q*zrXbS6aC7P>Gz1n6YZnl zvwRo6ZwFsy{y`@%4Q~WL2^!L#ci#Pi^+nRWgC_gkGR-@OZY`9VIVPS#7A$gbYG_nw z&#^U>C1~W7{i9nZ8f}F}wE20rOf+)J9&p+_6OA5Bp;6NLl)d)(6hoiA$>&qFCZ?ZQ z>(iPq_WJTKHow4i>io*nSeOHbzmbtMobe{+7iDICF|QViHq0k?etGBBmmNAJ%`4_b zwM!q(E9R8uP^IVtzneaxPka{Nk(u5(R3JDgGcwSbSAV3QDHDBoUvp@*TTk?H%DSEQ z41M;d_21_}23~MR`uL1JFv-M_4NvT0lAkF?AIW+sX%6-2B(FZZ9$q(oh9%@c0QTr2 zhp*wU99PCLZz4;GD{;SX1wV^|u{6mWJv;K|IgL&8sTN#Pe+S=%bAxx4Q||-t+q&GI zV?kt47&$DxC4PRA`uZ&h9O1kA%J->V%8fqzDtp>)J>ix9b(ZMgojNvN14($*8fMS8 z`sC6N9Gzn9fm%D?^lF383;OVR8~xwTA37u9E4mFBZf9&>U(uW(az%LWvlk4nFedqm z2LICW6?treXqm%1z^*!e>d03#4jW4EiY5WZ7Av!X1-W zpN^0{e`(VvJot5e2MjW?Uzh4H&{}mKzCHqf=W493JsO4IlxIUeKl$r^qPR)mQTfHd zE4wsP_c&@?!2OOhM#B&Av%?Sjop=s%g=B|rPvHl5uU2rkeY0=Qxo{7(-aBH}dwu(7 z@-qwm6~HOk!J65evvzsq;K>u~H}}P2@5rBc*%^4n-V-FgAzv#r(hr?-va>{iY6 z&?=sHTY)9D^t0SbU(!nK(JdiVRy{S#ijC%y{k)n#rHyTep$7g?l%S&4Tu` zd0(=ralkeN3>2^NR5q&ZtI4JaY zUrQ;o28BMCTGovJpECJPKSdkHMtInYMX=#C2Y&G=zda1!*0YD`<-aoT{Se9d9CU7d zEIw@ctMl=L-^k~yOMMNmE#1@b`qI4(Z!8To%%S|UrGpzTU)s{}<-Ly`{MJoddcOSJ z-ku-)-HD#JmgO|OKV-(C;a2OxX%D{D@Shvj9X$TU=X(D0`olf{+0fea?FWZ8{1fFb zEZcDKg@4)Jb6e$`J-7M0dwMBPN3VKTDq@6d-WMh z(;GUjw_?`PR~qDFEoHoM~IOKA80rT&J>r7tzyNZ&K)w{0l7gQzPXS1IpSElY1$ zubrgCZ-1BPpZSN>Kk6I3z1(N*+U)b~%A`)U zKg-m+EO62GN^AJ8H++M3scxe$yZ+rm;)`f+1$8T}%w02mS-U3s2Ja$;m6?k$Kd!@aC|E` z{@St;4NI0CYFOeMw*7tDo$4RP^&2CB(YMRz5A3S-W$t>Feu>rHKpc4ekxbpGkg@&9 zP;wlO3dSY@Pxpvm>>XtC5?{`)cc}M!boQgZ^Xh*;nAic}-(&f975W0Z7Sh(kPFvZ) zI|X#TR#Zzyyfbm-5sjEpw}qxwVzKRmJ>Y z1)pR9@2kxJ*BDb5V|tA-y~dcbkuQUgE!!FYcE&%4@y}uWcTiqM`Q_m83I~TS{^~sZ zRchsIm%UnN`FC;V!tmUazO?OneiWJZP3FmC%l0*_U6$1lTJ~DQx@ChK9$nVLJQ-BK zhIy+w_o#1peVsL8`&i`1Dqs4pr=g4Zcam>t{mZn44%^k~wC}e0E!upHHrLW-D5=e% zPMgDMbNKcdw6}`(CisRpZJt-pJy_eHq`h{vOFxCc`6Y042Qu}}rCAMMUiwMN}wWL+b{OmENBM5H&0Kdd{h=TBEGT@fXWUoAJx`%a zo<^5E^XB@4&-{5y&#e1@-t)+BWAKg--r3X8UH$mM?Zdv`)BUx*JuSWy@Q$zHPRg$c zuQ>Rvb97 z-#1BrqkFe8kJ{0_>CBNV=13R%x18rCJio^0%Y2GX`OFP;;jSBz+nNtEk^7VVqYQuM z1aj)%^jLz-f7ceqvyVBYxif(=by~yfrBf><7vRTtjBZ7r z!msb5TOVca$u5v?HSiTspj=Y3YH6?Mu&Z*tYZ+4Q=SkZH!HAh=%#-)J>MPD?%GD(FXf# z8<@YyiLClG>-_F<*6R~eSGr$*UjkEI&1uz8jIKs=2?6i{DpkjOj#af zdo2H&R-Ta=O|*-Q7~iK;sVloqd==qSd)emM)v4GLMyFnRW}PZKI}NB@plA^_Km4N)WN))P#rAl9CP=fPUhaks$fypCHEfcV*X7k4;D@S zhhWi7m&`ge<@3R!C70ZC=(4%NqDL?J{GrQl2^M|-lJY}Wd^%Y4^Gm7^T{$aQ^xh?R zADTKNSTyX?dk=jw7%ZChSHYsMUOMa0Rnvn-Ph5J-p{uV77CnFI=MP;oHCXiarR9gN zy)0Pt!KKxQrcVqOjn2LMP|?_6QBm%_hpx*D78Q)K$3s48GnTO7m)e&d-&VT-wOUJ_$%Sh$nU^KP5IV^^}c|wKH{eyf1CMR!CwV` zMz#lrQ)gIx#K-UaT@60E2(Uc=U z3>?Pp@?m$`YkiT`jo(;o{0V*W-r_Z4&i9vE#ol~m2${S+g+tOnv1Cv9`ms^b|%k0nit?7<$W3ut9K6}94 zzdl(x$*#B6{)`vX-)%S8{(F9t4|X$k%h|`y_E{r$DKFIR>9O~>=7b`Zd7<0~S3HyJ z_jTj5m>U6}Fy+JLR{=h)mv^0vum59hR_J(P@)~t8~7uEWjlA6dTIUY-Q?EG z4epe$!NT`Y!?~DN+AOQyG+p_@_v-96ev>M`6H9)mjq~eT^Q$btr|RnYcyxTyI{y{o zjs(N;z__IHmz?tWZ_;ANYYUQo`-%I`*1V+h=iPU1PK#Zyyh!+ty|G#+slVW2=n>5e zEh@B3%vRe3r*2AI*U7ScyKZfMU)}3ZQ#b0=EiCA(3k>+zOuqFn{YDtWjz6(~u&zXR zK^<`M(LAj8h44CGEX?l_>isZg*NK9!s_HG%;i-Gxgs0jvdap4&N4+vHi+ZNXQCoUCty7oHdysQ+UM-XjjKOuz7QE5F~;+t>HSE|1gS4^sMD zMPJ12)w_Kmi_h5Cz5)7rgoFJ}*7?#}uAwa^Ps6cSCF&dkpRHR$~0r?edIYFg=cp zaB+TT3LHr|0EYQGJ9!Q;tUe12^MK)NZolwDE&U1x+s8e{__!nZ_7q>G&wBH@Z0Bm{ z^I<;mcM)%H&NS{&9P5U=lJ`WsKJdav`t+~s1K;K7VSE26!y$)Jv8_FFW;)vzP4k0V)p8ypR2&bAaJYpLBnH zRo4TD>T3L<_E9A|Z(ZLu*caOZk4NYyOdD6gTk>mb?=SvG=Xo&sgy=h($tS+bTzsna zoKvxNLsq~2xZiRv+g;};$2*-HKE6HxKGwM7@!;c!1K^|PH2ApRecywRDEwZ9e@^?% z9pHmJ%=PWyL-!TB_;|yEkDq$*@o6tUPMsMa-#r~Z0?vLf{Qu{nBNO|R$19U*Jja2( zf%6YvBliA}DZl@_J2ppWYd_#g^qip&d)Tt|q`SWX96JBf-#+L&f1`7e>Sy7I8_nmU zQ8(hV)fsp{=Z19Wiq0UzANYyqUH0J_?*_WvGj20KoW8yN&o=f+U%vR?kVmAPJR+OO zBO*A;JIi86*fZC;fuM4RF;RD_*NU-_U%7~%{_>hho=AWgz^@~ za)((j<_3MeoSTmkcgDWgMW+3V5g9STQUNS`faOgG2K8M5PL$Wgz%eSriTmq=;}VVA zGgj|-8_jvLK6~sdoPFJtMO^bpat;luIrw$ACPVQ z0p}<9Cw?pIih20jm*8(&gl$>LnuB{H<;zptn{2>|zx(t0@H~6J$X9NK7%O*x-cjV| z%#WQTIJomSOM~ksoI3f)@<7ItF3uz=-?97K^{3{U@(T7UCY-wXk%oBudDV%qp6)tn z70u%86KRbOhO>6}Z1Oet7~3*;J!Wvf*Y;tKCkN!=EbAVNy{wF&ji*K6 zDHA-gjwBCNaOX+z^t!=Q=8~fhp4{JlWAK!@M0k4L!ISF9*Plsyy02sq_j&2;`vX6M zZ=A7rXwAzSP&k+b48_^jJvHpLi*}aonb`k{P$cs~&ScnU(St92xZam;J*VGA%SvFx zuDm8d4zf1dT4hClvW>NECiqK~0h`KJ0oPBVUpn`oYJavBAB??EoZfWBPIM*f4CQhW z|3p~7DqlhvxSEAa-5++2xSY?~kkXohQ0pAtzZqYP$qP(;PMJUSpGQxu=XY{vUKEOg zgZa!&nS(Z07AZw@YYECv^iz?uy% z-ms4RDETWITCHGq4J@Hf$rSEn z{%_*ucs+UYPQ!yqestE@lI3aS3UGC-YkQpZ_yA6~{?mD1oqrBy;)iB!3~sOa0{J(* z@&+4dt!yCX?EZY~uR>qA)3$-s@4i*``peMPy=S6CZ2D2{z4pW2rBdhUVHQhaHaoPcW~SL$F{7nWq(p%^wE&gM|X<+R{YHB zlz9FCd^VQ9vz;%kI^;QDI;`f!DC4iQ&G}u$E!2?HPI(G4%)C+F3DH=z7A-THzLBs|Y`?|zZRlj? zbP^sNexK!m#pQRM_xGMLuJGgnNY%e;)6FyR68v3SbAicGC|s=szGBP1yJIErO3x_f zKzUX~XRX2YDLcx|ucPO5?(VoT<^B%E;0$zsM~{d0efoQdSsyIEkoCddp~>rmxGxqR zO#Y`){@9U1f2@0oKUQ309rKZEG0vK41vGv2RIn(Xb@5j=@htdkUL0!dy?3s(-#I$L ze!nZ%ZmWF+`Il(7HPesZ!XFFMPuoy(eeKOjT0i^%d8_q9AHA^osB;1L8*KLGAKGj8 z7yg6K$m8Rc$$7WF>$JlOJ?(MqjPMzz&cAJCZ8o74{Y&1%LMMOHVSlJ6lN=}s9-&X_ zj(du=$CahA=)6FYzDMQ}ON-uBy-WEmJP#c*g*K?uI>7tqtB&3qrG7Y@n_C(i=9a~^ zwxaFK(%4|PY%ukPQZ}MAmOO}PydagNX+%DQre-ic<}?-XX6 z&!jQ&+f>#MzD(=+a5wk|p4uAX(|~V9jTJgs>Abtu{tWY7aBt_Gh|c=I*FP4g)WLH& z%6jn+C9!7KD#t66*Mvn*ok+KR?lAfu@qyGCg750&VRClL-^X{E)s3z44thrNetO-& z^1dDY;pSvfuBao}IY+U(y0Hb{!4|y0nYXS!on-Pu_pMh<{tWVgX-`mR_rht`b9;fg z68R!opm|z?Oi?_C@#Qg2#q>z-=nkrR)L}l&^;1VUklayRk)DO`#Jn!Eb&M;2P6F2r z)#cO2^QZfREY{`tm55^sbk7;!{vgSpXzKv-72F)FO~%bD)HQOIb)<5v4*cU1icWBC;@y7XFHsI_(9dCUL! z0%L_wY#7r+|0)>a9bZP>L14Vk|ImtqFIk&=N?+OE6Y&A~0@+j{&{t)IM9QFE{~xYdTG23YO}7Vf6> z!t!-sVPAQt_8$7d0>2BE(Sn8NR9FmNYBEB13l{jkZ~56@*J3~wtg^e!M8u<9ANtSSzsEU3RBiuU<$f0b@YR&HFNpS zjB|jg@hmX;Qees-^w2>ce6$(b?eIL#GaDuRK7X+X{0~Q@7OWscXakD-7!m? zh;*rPx+x#I*3=&-ANf<{=iEoG&TYqUtl!4-X7Y57zsx$&ITK$2@V!Hxa;@ds3z6aG zY-b*JjIU|iS8Ul+Os-koXVi`?B?h*ro%hBQEBh*Nb%FOdIU}FJ=Me`N0Y15>^T3gS zzh#ng_W71SKFMd*AEr(h{&mHTNdNE2A=hBeBlVp*kF<0m*Sm-V-p4(x`z))zV*+@; zmUa8}(zV%5bI?`ItcA69jG*_6tH(}n1BP~F^hPVad*zN*8+4vn>++43)m;NUlpii4 z*-3rH`RkmzY`cFZw%%St+iMxy7VNWaw08CW4n6i)VzcVA3cFQ$-p+a8aOb^38skXM25>*;HxDgF8RRr|6x!u5xWB*+=iMNVWk?PBG@Q?4!?k z^l7#^@6o4E+u&#A&Qxrn*7OlcE-!T62_zUt^qwmeXKEhgzxTcoQ(OKr+1NmfM z!{3>GxfME!yi)GZ&f_U?D%;4Z=J}uF6L^kFQeQ4qlY$hAhQntAL7mgzN+%f|L5jr zB_L2xv}!|0SR%NzT7jW0HwlO=N@wI>+tO(QVMz#9`8%x(Xo3kUL@O7Zp~ZICWXV;l z#icqnRW>O!u$$kZ=&12}n-RMD`#iqRwJ3HD- z3a#_BC!OQQO}gr*l+a}5DWRJtna_mK)#fv>eWdwREGRKFifbFkPTVo@`QW(Xp7oYw zYEOChI_qkWzREG@<#p;S>8LtnDixkeOCvWMemVd@ncCX-m9T%~=1^)H|NZ&TY2N2wo7u1e)v_GklhMCn86LRUW;-H85+qOW7% zq96NpsNwK(1Re{+b0*%7Uefx5YX8=PZ{oRY@CROwLMt~u6_KpLV<+|#|6|=i_M`B6 z+m6f{o!2$?34D?>uO2<$^MCDcJ=qm^PA*z~y$4!(=ile-C*R}j2g#OB>|EcfzX#b| z%sjsW`(B3qDTStYgX=qy*ZJVR2z(Z>u5vN<1v?+0)?s9^k=LKjf`5fW&BYBax!-sX zyx`C=4LYh8@bNpDoA+7XjNd$?xk&g=Z{EeerQYOYJ(kA2nBNw%e`yr8f**&63VeN< zmw(QhalDj0fF*Z~Io6V}Xeqq+eV?!9SNtBmTXX%s%~@CC&+zP{S6DM{p{?tBFRyy8 z&+?kB>>E0XuX6IE$5#Yf9ymJD*QYA^qUALs65zMXTYpCW<@V=>?qB;=Utp~E5pAQM zZl8n^E1u)Li#t-7H&3-=9q-^z)o?tmgay)(q);_HfE)J(s7Q z^|=4(-&{NP?8!Jg_Jc8w9V14ib!qINkuwKQ>Zb>f;WJLbub5vvZ_Fc{wQ@e#Gw?wY z;Fo#CbB-lii#A|a)?){Be+2v5!q8W`Bw%d$px~pxqczef{^}5N24m1<7W^a|p4Ypg zMr(MPeBuiPm1oKtevR2Td$wYQ)~$-5Lu@*2!VAl&Us^A{0#Der^B*v=RkIeu@8nOC z>jMqv^LqidLq1(mpzWOB@w;?(lKFjP=kKa1tM_f6*f~=MT9>od6Gi8_y3P2lw*}f> z-gh3|=HS=$i@i8MAAYxytMdf^lKCGiZ^!6HE3|^^J{7FJ`TL?P`-J}0JkvMyqC6}1^Ng!a4b0B`{^;$Qv$$u{YOZ9!{+b5Lgh8p`_8wU>AS?2R+HoOt-A9~@M895#44FLW`8iDN<8hJfhxO*ygW%v z;N=*+U34kuO82A>^6m9`Lv!Nw`kY0~-nMm~-_Eg(_HX!FhnbsnoJVCL56H^l@x3dW z??7gAS-T>RG4>5|F}vj}r~k-k@PfV#qi18-j3RJY1f7IK`9m@3UU>THw?1B@z2!I2 zM-&;7oYtZvO>R3f&efbb`neF^6W=|kvlLQjBOQ6`OFskX2i+Xc zcXOP1hvob!2BSK)!i&xc&cwgY7sUt2M{bJ|A1LCi%JIzcmg0*l_Ne=Q z;8Oftb1d>IpTy_6QU0solV7B`dI37J+cnJ!XI&P!3}3k49Xn=n`M}xd<8bD!PaGfL zR(?bab;BZkq3Zztv)CKKcj4OAV~QcUzOeFDq-)%`y{)foxP~$ZO^#bXEMWZItslMz zy<8p;ofADUd*MAmKB;&9*nsTTqvzdNQ=RMxK0*wB#g>rkI6nA6Gxi&MVb96zd=h)@ zLFW^vh7%cYshKcJar$%e^ijl~Mgt0~DJl*L$56{a!bomXNpjfctB^|%X zHt4!Gd zk*p_XbowBX29ob$w4 zTDA0{HFceB;LS-L;I=uMHm_&hlQwI)uI0JfYiSc57#pcI$+Y0*-1vn|&P(~HIr3prLf`Cxzr;^xmYmB^|JTDmCjSKc{&C+)u@VcCN`Ko!ig39>_KMH|E_-xhD6$O?e$!_>s zQqieHS6aaxFZf#U!)6Fxtx3qg(6x9k8-A`d=Wiy3Qk^lFI4im^DnCQ>KK!dh+L=cl zfoeYMtW4XdPYqp<{F!mnXC1zX=4VSS@)qzvWJC90cb5+K1?S#if5(@Zp}478Q@}px zI>s7~3rEAkv*z6M&AoTNJ#4UbIH`D;IWJ6W*oKaN+t=C!oul{)KjYKYg`(qa4u1xr zrAxmOYL9sLLd)J3z=t*PQ6luJgYTsXwz&X0`GhTCs z9==Y4_28;)#EVO=)A|Jdu34XWa24m*Rh0&;!_9vDZ~5E42Uks_7Qh`}wGN;3_o;b~ zd(Uz2`^@$4EVK^4>+f6hF8AK$-XZo(FS^S*+~V(F^ZAd!Bqni`d9r^x%T@E zta*}qPjc_O2?MLXx6(TNXa6NNzvbRtDW=e$}8-@?zYJzYHCzJ`YrOG@l( zcz8ldseKI(SC-siU&F(1l-y}w!^4RS7TMSE@SFwzWM9L>$JifP2!6G8{sDeaO#DG? zE8@SH7}x#Ve;~O<$HuV7AK(woUgU++4# z;fY!8v1#Gbu%_w5HM_j==o|d!rJ?IDdG~C*viMw`I1B#s!CSk?754pc^^pyOUp*CB z|Msb+zp6jwW34Z;@u?%ei?*D~9A0y3SLEYUKGxMU=Pf0S4f_%WwEacx)?ci@GP194){6gk;fc#LU-*I zJMBf3+p*I(_JvPq_v^Il&3_cWXCces^arQfDM0%bLf`WgbIqp_*Z5A71gW zm0VRo+^7y(l=JikXAf8MTmyb|1ikhqeF!eI7DOL! zY7Pv3)c&RyD%NVf$zmK@f0J!d&hEcDI56+fhb_adkL1p~+_lUk^g}{uqH?i}UAljJ z9@nB-cd}Z;83OVC@Pm%s7eBP@vG>nRCO-CY->&r%V^#3PBjC5f=7|%BKVI`ekuP|< zEYRllNl(+i*(b$)i+ye3`HYR{Zf34YjiT!46S>vlNBn#yR)@3l$Cmr88|xmS?%Z9} zCam(<$qLTogf=El22Z;-@^huwNS=?}2|T5=eYfVcNmZK9Mq;$Vv$qI`ck3B&U+T)F zy>G|#1r2)a3%POo@;iO0k0|{`=x-D2LLGha>=)d==qE~lw)}Uksp!?03%a4f21h@% zRM_$P0|x({^%dQ777~%ajx!HDH5BXmmanSYV~t&=T3hK4Y)4LN8rX}~z~_#z)$-i2w?RQ-i9v~qok|B|X7q*f02 z;SH4o#`rHi_`|O|pVZ7*YbCE3>;LS*f8+BT{>u(t>c6b&bTyk*GT>G;oA-(Dn zU)qY>eIr)rzTms(+MB2sc1hKagh9;}A1Y>ZG4^RtXc;+NVb%g)U1zEHoi&y@&5f$(C|Ku=yUE13 z=Z?2O=jGqjCYu>%J{C^rvoG?6e`jnyK1(e=fby=C^F;jNNyVc_y8A6E9DB_BTF>+9 z`**0h9}lnMi{>8wqGVNhe%7|Yw6BghHf;v~U+4c@BLdU1?>aVZDF4IwAI^Wm73*)! zx?;nvH(#;w*2>hu-`ULnUj7^SUtambtxt~_@$K1xv~TT{x4y}}Hwsg_ z+&@sc{nmQ!*A*sqx!+Xz;;oNx|KY-vPm|9<<+ z);4Qecm6avbY1-WAs3Y&`OQu@uiuHEJA9(~!8{LtoUrsDP^pIP}V z^^@3xY4)H=M_W&hC(kewd}QE{>i4a@YidFRpWcx#O=8qQ<#t3_BhS;`(s8+Cv-ngl zP=t5u8N*IuTkd-rm)qYy{QmV9S)-Nfq4)K!eHSw%1 zCzc#-08jPB^IGfrB~^_l-mEx(Eqy&o{QPoX`PdD_HS79WRh#%s;`wOt+*aY?w>!)} z=a$;7>(*9kZ>*GRUib~JpowDIQOzZE4SpIJLj>5LKtGuG z2KW5l68Y2D!voMweWpNDwY$x-YBGE&&37<=X+ZaxcCzEzp+E73`k#jmQ~q8pF;f%! zA8qG}HvEjef3&AqYvahycBpw2^p3@CXNKAt$yhz_3$8G2YfRr{Ogek=NJn0$?YmjT zq7TDcUVqMe{b%L5V4qI&d2Mk{D-*W1tBS8N^=cK@ZD$U3H8fVNSh{~izr?EL(Cez= zlGgMj%_mZt4}gnN{ZgtP=leVSc5U%@TKmwB>_Rc|)0;nPzu`Ky@2AER?~On{uR@QH z;JFdl0qybMYWb<>Wz~emOUT-7(&6|bvT46CFluclA#{*y@q7IIs}_IF-B%;IFTUqo zyWm|{w1h9upF3Y)K0>;nGyf_A4`}R@hqG^-u|LZAa*B&vkAv$1a)xt;Su;lVORieT z*!S|=)$sfY;w>8cLdM>kJ<2Vtt%Jk1QM@;bvE%n}J+E~XW0!0;WT79he{MeN)x{%P z7l6MU;**u|!)9<&$N0qe`mTCb#0%}*Qsm%L?hD3}OzHtc@7J-xLxELm5Dx>>F!~xo z-!Z=1F|VZ1+5hO2GjA@J=G+Zg@L7hfKeS)9=xoYp^({Dd(9Z(mms&5p()63uJQ@8d z`0!(ytCo+=M&`9{Aw6(z--)6B^z?0HoIdO5CxbCX;AiEONvC?@5R8H)b|z(Xjl+*Q z>_Z&p;l~<7lceS}a#BYDgZS|$z#xA7g+r5h&?Fg}JO)j~kHg@{VGchIxv7h9R4i&x4By za^muwm-a5t>Fk>e<+%b6-T%8hmqV`pr{TGPr|-Xp=eB?c#kwYcGX1GbVblM8*1RvA z{$2APZJtO7eIMGm*Hz}))`O3F?l1G)f5>yc%yWN<=l*@3``H03N zFGSB?^u6=z*^sBckme7}9lO=92UthxPQO<7?@_-Va`o#E!SjDtziLlVcXBTt^U7Ho z{NA0MH8FqfLCy|I=Sa@(1kRpxPET@n8E_3lXGq4I;CWZp>d-f`#eQ_0Y?AbftJ|0T zyi>QkK2(QZ?MA=n_n_ae=;#~2j&)V9yw#filT82Hy7ZqIn(OIbai9y;?S9~poQVg! z(}gw4A#wO~zqz(`i7+=zmPUq{M>n- zi1Jozr{jy-uyV|jFS7YO@jzPwD6I-C~U@ov}eap~dk8PKfC#pPY)-jtG;sq4E^rkyYg+_ZTKtrAv>5q$XAZw z??&;fWB37P4>a%8P7k!r$0vB+@J&}(!>&zrU_XIBwE~;lSUj?|5x!I|jOl~6)vod@ z)VA8R+fNN?55XO$+M!8Ga}()(L9{|`Qp;n zq4NX=e_2%RB9z^aUrX@o&ZguIjH@*z=8_8hqF~Z*_OR!>dpRkbX2dDZg7UR4WxiB%byR!cfIBI5uu46lZ#v*!R@EC2O4bWl6d+wHw1 zsXp>4pve~Gc_g{H&H2npkcSiKuUq(jlDX}ze1B)O<|jUKGg8g|L(PdB;iEL(&*47; zKSggxCTAiy@NFG@T#r22@1%vkqIonh^!Uy=-ht=qM|ZySdEW81tMdw)X!p40@${{E zZY{i0M@+Sz_^SHCKQ}Q$?VpIkm$O+5EV#~U*?^9Uu)oQ+C-$6Jv4gJPUEkaOy{EaQ zXnN%Co@uJMf;**j?L!HE`MLASq-Oc$vYR8&BRR|^WZ!d$od{RTG1i_R#d{j2LNjQf zd8lmvW_Nyv{Kya411>bjvvrCsL(5)QK6>jbiYM*kd1NoRk2U--^oc+lV)_I1o2wsc z!F}B%#dEOp@XuD-YXs*hCG%U;`X^PbV?ArR*0a$2S`#t4Dh`(9{yqcCFy2ilk$#gb zC52`KOV~oc(cdz?11!q~3;p?lL-)hz+z2ozPNe)T#e?eU|4nSeK4L_gi;kBa!S@bl z+5O|!u-+|UX@4lkE^YD+cIl#0CK63eIx!J$k zSgN^T}~;WF0((J%ox^re^M2E4U6I=k2TmrLYG_ zxly(~w_~N(6<^i*Y?9SBviMqZ8#+I)hXQ0?nG-k03{FZMb(W#W{BQ!egx zko{yUj?-s`&)-~%Z1qC#-0VxLksYjtJ{I3c^h>L{3ObBLulI&e3!v+AVnkO%%b)PQ zFZ8{x_|Ddg=vVekF{1rv+HZJo0O!k)chNWi-_~!nUBmm=qyQUs?HZoHhJN2e59!=4 z>ElU;2DS~G@Ch`K?8ex~-IG41{$z*jy82l273S>S`gbwO>s>hn7Fq5)sc?BihlL@XD`TKU3R+thUdu( z=*eIGk;^X?Hoq9&ajs?4CGS%`ymPa;w(n0g*S*X$d7r>j(#7JdFngkV(go7P3q8E$ z)d!mYd(XR5-hZul41Xz#4p1FA&G%$q<{92eZhjm+sr}j7lhC1m{q{JoVNPPldavfJ zxRlU3=2g7pCrV+2nT`)$i8uz)k|6cXu0^9v|?D&N_$Vw&o=F%0vV=X=#{JoK4t*pHC`XGK= zoBN*fP-yQ^UUna13lD7HPdl}=ljyW#WR7+u+i%j&P_>iJS(Al5v@^kO=M~dVS-0)H z=(Zzxe*Jgd!Sh01r9Hl<&(qEZZ##eLp`G=1JFCLm$#i#Z#%U;jO(=B&bQ3CI`Qmn=Np~vyihst zn(oGRC~uBUuPL-M0bcXsXO8Lw!5^xBti3b0(4O`!=w2prxLErmHrchw3=jQ&l=PLk zKWmdyBgoND;{SiXi$1#H{}z9u+0Rowlzb9*?fc$zomQZhWc4q^iy2%xx~T4kuelRXRg+FSe+@$dVt~(g(d$x0SYO z*UeAe178;n9!Xt~!1lGL&a8>vjZWLDchGP6@n-#ZA3Co_IxZt5pUn0_{7&9~vRQL! ze4VxE4OX!6PUb$?cD2)88|}b*ko{N zoL#XI^otkPSUPo?%KI5u?*vvCzE}ZvWE10h0ytNro7V)|s(wIT+cVH@zD=7VXck3} zxH>>}2vr|N>-=v_Kf6YI4->&vrs^|4MLX+g$HD(sUu|DgMD0rD7{;E#$C-&esshHc zORDR5ww`C}UbNcke+CQ{)MNyg%1wM|??-DQw{!0i(Ng_oge>+xTIE^6fj> zeF2C1QvH)L!fyp|--hpi4;3x5+U$48;YP>*J?}}E$8O`@3cYKeMJqp7dNNuTXrn%7 zaK~)+VH7y<0{5Hfoy{lyE@)~f^PHV?0UvC_UV}^Nc=mLR(H`t&fvSwv0M6 zTdtUJw1M{cZ@v(j%$#eFfc2( za3t4JY_$0IIefc^IkWeem9R+bp_6Dsd3ws3jV%UW&%+z{@Cj|6yOH-w8OLH9S9SM* zZ{TkLpW){jKRB#kXT8<{o{l|cEt&?8A1k*O`G9Y>WxcGqoc7j?Kvv|Nt8Te+caJ@4 zEh^Nryt4~=l6>g5{m_4iet*webOPA+Kg+)SBx_L`?HxN|Ejq}(&+7M9-sk$*UA{%^ zC77|<;4GP zsa45I)@vlO4$^$TjJQJAHnCGqn+fM@lkq0gp4#0=TZT87*N5-+1xLM)oQpTA#LMso zI$(2P6uc4G-VnzdbuDMsr18FJ`fua|iTAq1c+hd*;@k|cKggcQj=6>F|F!s?{MFI2 z`JLw$mewQ_(w` zw@GL8VeYMYWFvBymJ_H>|9ozB(=#d6%`c`_r}ZZm##|<0A^N=-Sw~N$&5=*$Yu*hF zpzUaM5Rz_&EB#*6indFOzzf!C!UbITUw+@qMD_Z5cG-6MV@aocNrb{F7 zmS}{Hc5!qLUauuSI*zphmuC2Zqu{4I`uDD?PO{pTGnZ4_S~Go?F->q_U070La#9wQ zK*6UnYdB?z-{V8A_aei1Y>ju^v1S|U=(#hF3cqMN> z9%gwCKJUWHI|f$vlU*=g(E;bf>kqnt^+mxt7+A^wGq93d7!T_L7uLaDVXbswoVbX0_Efv;T$qB6a()B{d$KS7EQSY+g?- zGUBA;eOAkJ_#B~8f$C>($i;_Fsy^|vKuMbN&vWi~gg3v|+$59kQ^4_FTai+WwD>WvYI&{Z9E&>@PIyADaTK z2aja06h2EBdsG)Qd)1oB9h;%JPHYnXEOs!lcS2QdK5=*S@_h7qSG&VKumvLp8nX;qw2e#yP#*-L{C%0mamux+!E9rX1| z%?lHWwf>&>r3cur-KMo(#lY)-7ih!p3r7EF^Ab3j5j$$n`@($6wq&z+r_xFfZe!dU zPfYfKv1KRvsnZB9keoH`&9Y;2HVw@l4__}+ zN8pb87PkKK-i58ls23KXruU=3DBAuDpJqPr2z03ek55=_wX4DT8tC>E^jn9ovx)oo zVIQ(bTYJ(Xk1%(s0^TR^?^Z*LHTWh^;hU@j-9 zUuZlzxiQY+s^z@1x{O@A-BPFBJDqm#?!Mhq$ep*{QEH2G%WC`5)&QrinNC{;PFqFY zw{-z57 zS{>KcFKDZ2Qf^xg^W*li%c`65bK8#d{rInt=^FCC;FtO1oi_9B&w1mzZ+l5x+p+v) zGuPMNAMDkdrtzMB#?c4gyYw^N>E|}5pQ+vVGy6RK{7|@|)=4e3P84sJ+}5EhHluGF zUIb_Og_2{{w_V44<8k84wJ*Zk=*-CT$mlc3B6@UIYM^?~K=#d5T~b}UW>EDkbWdcT zTmOAfaKPL6`RH2dDb>k{pfk)IZBtUv)w}=Ipl6`>^R`~Cd+KY|4V!MSPP){pP8sa0 zmTgEv|3;tj=rHt|S>xc-tHbX3mSUyoFWU3!ues@BB!{|*% zhy5JAVb8l#Ow30*=Retp+HF3@9>1+mw4ZNJQWt#^K_(dQ%Ms)(ikwL=xN)ut;Ji~$ zur`O@i=g+S=scriVqI(3yo;WdE{GvtF?5XhF^c}nr#@aSI&d8|mZaP3R@?gp!q6&$ z4h{1?3SYT8-qhA`=sF5qsrHNFOUk7VzXR<;=a}~@4f|H4GgZXKJ%e-hTonfgWF8_jd&k1 z3-*bS;}!+x7=6~#cNiGNYeO#$R1Y7VOAP8W@Wp`Yq41dEdx}ZzppM2Y=CsoNTd=jq z$W4pjCv3sy9wUctsjit@9pkx7u8HBa?dWZ_1lZHJWgzqKOR48_rhUea30BK`J_}83 zeo%1kAK*!N{9L|G?BLt$#JBKd9r%(zQ1{e70N-t$e7n`*Tj(R7)7V~cVfdC$FW(-R z>F_P>T?pT9MvneOo`9EcCByc)Bg9*fr6{!A1~1z&5kEdeqV}2$5bZa5=j>kS?%j8Olf$Fdy#dZrbNHC&yw7~uc_xf*55vdT zVmoYEv(Maom%5v7>_u?4FR=Yeoqy3kLG_nX&G{V>Xyu;Y@h4{-?)t7`Sl;!T@BbfU z=S*33yJx+oiFSKhuc84O?L!b5|p=VjWg=v)K(HEq7@!Kn+6 z-CpPVkvU(5dg1Hwd*{c&_I%v@?2?^ zj}HEceNHYv?}C4(Is7vd+coQ%2`28Z>%wQ8_`YXsKL(CQ&)DQsx#K&)TCsP1tDN&< z{G5wk)Gx5Atk7BnzszVKKpeR}q1wGK8kbGrH*{9n96OE~p+CikiAz42h<o&7m`(8aVA7vj)z7Hs%Y`EzGN5sN`&#?cc<&pGtfWyZ01(j@%f>TjWicxB+dM zIU}*7=ydebwSpgC`g-8OHkcUSIPgy!#Fw@1ad_B$!&?u+4;gcz9qSx>fGMB-Sz51E z4UL~7*J0-!Z<{~oo%m;oGe6ngv)g&r)(!SN!qj}guKWXh>*|P(y}r=#bcwAi++50@ z)^nGee4TUYsW!0#U1*xb++XYVqMKr~qIEBLb1OJ&uo9}aeLcXrnJHC@%QpH`Yg9i^ z?L@8A6&cW(wX+$@ht#_NR%qJ*-{|*t@ji0Xj(p6DqpxTyx(+e%Kjkl+OVcm+K-1g1 zqp9c#EflxypeMBT(sKy(RQ_u_Jgqu{%8g@wwkGU6r#t|+ja@va-=FQ__t)a-hMaln zv?7j9YkA(QPbSCdUrV)!ca9FeHOU$tJvR%-=CL2t zUhfKQujF3k`Wu5YIqS5-oOPN|TF2Um)=}c0dCHu5nov3md9QHJJk_&FtfQ3Ejzv3x zC$;t?AFs1*x9_?1&9gTyC)?wY>Uib91$>~eauqVy#i}#1pmkYo9 zo^zsx_-#PjoNW(tu@LYffL)Wu=8|eOO6(wU8B5tFF#kK1HAm`Lsv$? zaXaIVfYTx9g*@~^!3 zUzB~6-TVi}sOwC|=|iVSz}1j!@?d#ZwGQz^ZA8emCtCgNnctKIUg$Tqrgud^`a@6c?112IJ4Y}r6vqFGPSSqRgyx-E(*i%L9W{=3T)h2L9DZbn zgr8o(u50l@7#bunMxD_WqdnQl2!5^htw!?$ZGs_2yNme@3})Z;cxvH~Wla2L##>?2 zK0MNHN6+UKvX=*(7xjMNXdX6E?P+~DzZr$3w{BLFLc|ZT!SDwjuYvCX5WmdnEb9c<6Z}LUUc}h%1lGIZ znKF21G4hn*8ymfc&$5*2_RD>Sck=w2gTXhO{OB=k8hj(1NyZYOv21JYop#=N{oMiP z%J*--i~9}Wp>Z^4FR{)My@$G>$gKRUQhexyzSd~bOYhaeM|v-U&ey*C25=?as_)`8 z+1m!Kf79QF&ssOW4nC{lBk7z4=;dD61g%*%^Zd*Q0;@-Uq_fF;m7Z{Hg8SRJk90P9 zFJpg3eq^6bu4kHVCpL<2(TI=n9POS&ex6ls2en>)vwd82qd}n<6KAf@X-0n&5x8Gpbt4Rsvt?nF$o)x||1{Z$q zsgTdBy&mR#qv=*#-s$$2glmm!5;C<7UQf&3x3(Rb$|1)+r_TdNQ>kT{#+a*5wZFs~ z^4Q}arQAp@s6q1KBj}?jG>O7T#y3SKZ2e}h#bd8m=H!kIHqEr!P8JXY+>;P&f+w4J zrwKlsMx0RNUQWBza=WoBAN3CX0iPE=IywTL`hf$t&oJ+Y=~KM?VG%Xc;a^KUiw=rj zPHl=fIE<~x2Ud7A${fV%t2oT?;1_@6oJUE`XX$m>h^5$whN^;phb2 zQ~qTNy0Bp}xlQPG>NvGY2SnH>(*VvJumg%|E523>U)8}&_3(fAPVs+Ai1n*AM%O`` zq%mjL$c73B?mdfpVcOPM!}J#h$CA?v9cveTWBG%{x1L?otW+WNIKrmzl}RQtlr{WsZj|6Rnc+_`_u z(Uo5K$UgwS&ix?a4Ct6;5reeaWKX3N{nCk9vjF20j}+n? zDef(O;f24aamDkWcyD4H{A%}*hqhk&mYDH$#;4e~G?u)+vDoi}dFbaTwR;k%6_>?- znZ=w1e<2~Y{Ix9p^&NShkw6W(EdI^!`VM`v_~(0bggK4{eKYybcWLF!xils-M)x~9 zLvuTgVW4LW@-bZe=fvSZ3s}ir9$R7g4;BIMCgAnMm)ZO)jxGPj4=maI>-z@a^Mm7T z{`I}kiMgagyF~ClM7#`Lw0^tp#ShoWu6Mxc@|XwC<8wRZ{C6H4XF}%${>OuZFt&B4 zziT~O@u%7*UL7n5SS`na!_PWxHvfW4>zV3T?@5oxzyC??yZHK_q%XGJmPMDI=b`V9 zJa`j*|EKWrKgmCRFMxl(>A}ON;UAa2O>y#I;Vz4+k#EnJ4&z6Azqhyh zsudq@Z}(S+`=V3OyTs1k%iGpJsQP~oan5q|a{LB!j^pTHq{s>`Ay)3iv5&3y0 zj~gH50@^;5clr|3c3wAa&%MC5U0p8y_iLSB;P?;3KaKut*<-H{E~j=${Q6*c2jkp> z9omQftg+f^f5klMDC=Flt(RhZ@IU5rZwK~sF*zNH{;JJqCsu7do4Yz%YR~!beTeHT zUz&4qSw}FqDCIqTEah-dq^`KmA5)FVJ8sn3^vca6ZR zkNMLl^>uuPr!V}@+M=#~{e!o!J-*=A=<9UM-(?->*6-l!u%4OEzJyo>`&FTJ(aw>< zSQ+rl1fD8B@hi0k8RIPQ*lP34p`yEzg31kv)zI%get!zM*J!^h?QC+MIeld?wvMqq z1HWwLS#)viMe`dxa(yuNJomv%1el`ivr&F>TKyT@_j1-H4gL#gTll~GlX0@yi<7%~ zr=TlN`gn1I4D<&ljf|mzG2{sc@~NTEX7=lBW{mZWSukz@FPY%w=W+14d>22*79W1W z!-rl!Ry9^;dhDI6kCtC9`RDKUr%R6 zD%aMYhuQi-^)hr`=L%x5CyI;5ob>k&?bu~4+JsKnQD-emCH6BuiJV+)XDNPVX%fcScbFReMix=9-A4^71-W5)CQ zPIOa2AFIV;tz-xJUrqE|ReaSLYNrGRo6h|DBW<;}C!$-D_)o6Zemv!B4dKj$yt9cx z#pZV-E79vc`X$R3oWpsPl9$1zw){~0EIhxx7SUaMQSIGY$(gjYH^hH#dtd#O?HT=7 zKGL+eQ93FPzAv4pJy(ZjGsemZgM*5})G|iZg-h@yRN*Vu!1p!&k1AHgh*8Bd&_h}L zQyaW?O7tk;J(cY_U?Qi<61opp0zt(yB6%ULQkFmC2 z!u+g?bCQgVLw{on&QG_Md5OJjKIjI|QO@q`i4V~CoKc+=dQLfM)S1wA$XqMt;}7lN z+Sa@N(6d};FgNW;^JHvU9sJwn-M*1BJ_eF5!l@&)TS zZ$>qS{jAvxW`1V><~_^X>S8nN@&m8cO|tm6&vn`GrNZsz@w6~`t`@Q+^QyoUH)3)NiEAvKO z6^zd3`v80=>}m&}#}a~b4Bc&?tXuzLj~xqAUxG_(0?}u6wZF{W%a7kJDT6fmaBfA!8!h#&C8r6MT|>%HAC^Iy3oLZ zMf-)0qqBU(9W+1kepfuw`@JcjxX)(A!W54axK9dDq^SLSh6qv4H|2MllQH2YWZJ>iZ2+}p+XTzQV~ z>Cj_oa_i5!YaiQlDDMYfG3^gYhS%XeZ~JdvVEbWgwc>uj`(EC!Q-jgzT=y;89=(m6 zXl!@`{+(*}*RGk29-m_B_SdfI&2^t@Z-3Pn*k9dK6Uf6#s)jFD&W0})16d^0JdyOFcZakG*yo+}p(b zK{f7%^2}y_({nc?KYDi?@77VTehc!ivlDdIiE91q#6Q`<^GiQ$e<=^WtJr)~z8zDs z>$W6?zD^F5^mUZ?Cz|Ue+sE-~*HT=veGJ!?@NgYGB;LNIvEA_dLVT6SIHP#oP-{^! za!^J68^v%JF7w&-LwQTCG5r}m{3-jpyFNsM^4*A)YRzVlxmN5n zJv7jBzm2>a@pfPCeay9Vy6c8*wxCp!zZ*A zptskfA7w-4;!o;3Yg039Ki!TG5%YT4wK3lN5#}#i>t{~(o@HWYih0@l+xMqa6OK7< zR56Z0&U&5SZqJ3?J73l~=6lBB^|gP2zu;Xn&{~1uDKhXAgU0UYyJB&wqZ0=HoxnWw zY~j&*bodS{xmk1E1oXagtj$<>b{%tR#ovdV{rb_oD*{Igo=#Y_i{~=We)TAM7Dvao z`WKyZ?(Zfm*zUx!MF$fT=N-}WBhPQ)bo5nU=6M77OebgK>|IAQv+rM(aA@5TVsxd7 zX?vd+F4W%NYFvu@x$s$BmxB-d_LlhZ*zs{YPM!%K%(&0`7RATh|JLxKmpkB_<={`S z_|D#Ybo^O=i`&0}srQmq?!yn1bVf<$E-S9_p zv*s@43D(QQ;Ex9Q;~4dEh>6ZP^PaZVfTdDr(C_tnOm=*=B*ZQLE#M*Hg2)^_~I*zg4NPF>%?cCF8O`<)l} z&U#{0R^!NEE^#X(n-w-ZU1SW|TFE|OS4T*;UU24tmh;y!f6WyxJL}^{(Tg!)il8%w z!5{VLkA28S!|m2<;xGNTuk1JjQV#@CX+IGcINa{RV~`0!JF zp9xMWDIBwdnpQ)=_wnhh=c6wo%%gPPonnKvy;ZwlaIlg5m3sWyhTAz`iFS`q&mF7t z%N3h5XVc;1st)FY;*n#tKB{M$(4*}mtVMg|M_gZiWGb+fJ}~X*LST99fnJwC#`R-q zy{k%-de;;>@C_Mj$GB>MuN3%79r*HqulYRi?Tmv@{WW(3<0xP(Egm&Sx<}7U0;Yx~ z(~dTzvOb%bQnM3W3QrqvpiVV&t9?$*i>6}XDY;p2GHF{U?f1eTK>V>}n0xel(q<5b^*>jeC9 zk}(T5FK>JY`0gmaX3V-A;+^orEe3{ZM}Gngzl0Zl2`~II?V_sZk}j(0fNR93gKPYM z1FmVnDDC^*X^e#LwgnHT9i)!;g?>HXl0uTHAiJ6j1 zB7LwG-&ZwZ#Oobnb$CE7B7E2514iz>7m^oc>vniJG4KBP1wU+GWCJv%Ttkk%aQ$! zDI3=H#xU+C+OugTGG%H{LYI^FQ$NTX*PiJ2Y44&)>Elj%PP znvbOaQsiFtK={r{O0OC0z_c6xKx-uz3RCWXWg%W6nmG&nwOzBl*>{#`f>A{p$$x;}hh6=<~Sq+30-MJDzpFj_&f2TjkJp-4y=Q*n z`frn2SBm%Fv<9kJ+QV0Z<8{`n44zlPcz*I<3zWZ_qpQh{Y+FuN}Cx7M6+L)tSMg%uy^g$`dUA1wTdEnFV(K*25j&2I zw%TNaHIFp=Jso`f3w^MDz;F0c$c*YXYhK^L^IF65TR!3@N#s$9kCH>X=d(tFzhFP- z<`{S5Q^wJ}mHn$d(R>-crt#PP_IaBn*#FbiqWkfOGUwUbY zp;1~$Ynb)K>WT3aS5ITDv{$I9n|rm~i!O%0z)?go`6AANEAWlA;j#weM`7IezN~`PuTV^T190{BC_~!*~7ZuYP_m zsU&emp1N;(S{TC&77#~@@aI(}LuZ!PKo$#A!=mcR% z4m>;u9-b*FHF($!9-7aCht)m7!1#rSo_LRXD$HT79;1G9`UF-zL%*6Um6Vj2erGqd z&uBVNzu%4Tm;QwR^YuB*$Yx^5=wS4zbe-n$y9|%}Of5jy_IY*cH|fi@k&>HXz`hmT zIv=0c=w>G-pct3pz>;fMx88(qorG>3!Fq*iNDl#@4d_Un`>5EnZF5{71)cBc*e}zk z>hr4(o7=bi6UCObU(u}V-9sG)qie;VDWPKQsJn+RM&6Fsp4lu+_!VTk@gz7Hczof zi&#!{Bx4*7zO(G#ii{2RH(TVRu)pZ#th0qjYqd5*tS=WDg`tC+d$o~R&E~22Hp9`I zBW-%PF`RG~_ps4j*1#&9n53tT-OMeY?4}LHaMZ>vs@DOI)UF+KIkye`p6wOgw4s=e z+Iad*_iYriKftm^Ir)dO1>N+M0)EorfxfoCby*x8HX}Q3Jf{gC*c&^IfYaTqiDWFX zs-L0`$ZT>NA0nqwImcVc6M)CsGT3)u$ETg~Iq}YLhJF7{?8l!zXkSy8!FnTm+BUFn zO>NDitu3_mL#G|pdeC0xjix<2C)oB=om_M8IPfRrJJz$8HlC*x|8>VxV)lQZ%Rh6B zTqiaw2T5ZTPr32nUHAmESz}V}VkfU8gf16-h)s`Yy>BwPOtsLZlC#^<2d|Ao7Bt=o zj6pHnA}g^<`ypqs<{64 zn_j3`D;^*3#Jj`DjMfnpfB$>s#KQZ6LF)!u?~LVl!l?Bm!THB2j6a{}xbDKo%nS71 z`^p=^xBdg45o%U9=sWu1H+*XB@w!3u`(8Kn3wr3c99nq!QL%jSIJKv`%9P7fY>`x!g)=OR-(`F%06<>mKg>eY73?|$3gEuc&) zZi9~at%blj%YjpQ(@%q!)egSrfR}9WGP?&jcT!i)vND1VijKxaJ=>4H98pkNgz*7J6Q5%BoLbo-sBo)$`k8v#-Z~?!N6>1Cc)> zSv7r{K2n>Uy{o*iW2|ji@YZa`;kObGW?D%H zC*TWZC$|1f`}Vf5r*(Kg);GyLe2Be7n+NYUdtBdO|4@;Yy@L5_%SNtqE#Hb+7PZqX z%dE*5J&|wgZ?&oRv+%E!!2-;@Ilyz8A|XsP^CcimGqTews%$vWikDsVKD{YKhzG_g>1r8$R}Jx58@ z{}nvcF&ewYZ;j5LAU$cHD^#ai3+yKgqkkjPv&^ky@V;P^KI#~+yWZ*UMRLB^1OHy^ zf#m5;`s%Koq%-X=DV{QiJuvG7fw3FdZ?loTHCxyNt9?(J)&s2Ub^pM z>qhy|L%qbj?F^pBYjSu9Y%XwG(>}tWq*MMJRl%CZy-{p6A z%-T!2+u2JgJwJ#2l$GqC)R?uO(%fUrx;Mueb0xap9&=*o3yeAJ%0q>1Xa81hUvAoV zee%C;+lk$@-640*IO4z8xi}@>Alr2VFR4 z#&5nec731d@ZB=_uH3Yh6nc`j;`t7r-PBC*sb)gJ9~@0hSK*T=8uTqYd$Mhc80jbPOHr+BW&f&uPC;3bL0DUlaxS9l67H z$gcgXGjDeN+5g}5IV@Ax2`gNA>`!fT7Ggk%j9W)s`?!B;0Ni? zF7pfcj2tgh&m#|9Ykm}8pX2Mi_y4Othdbu!@neR^RiC4dT!~LppW|v^)jUc*glu_^ z#+(&5W-qUm`-4CFN*tePF7M$J*9RKn<(0nZXb-RSRG;I(5ZUkU>vPm$%S~;LP5!Z} z)1i7D?dXQ~L7g=^%zpGus?lNZN7vj^vMrk)A#PrOr(@IamR`Gm`~T&B({OA$Hr01A z_crCzHauO{Y0rO8?y=9Q)uH!|z2trAn&a5RCh#UX)_3N|1CKK&ZDJ0pes@u;LpHJ5 z(E$t5hrQ8i9DV5iw%F9_=xy}j9ZszdJ=1&}^&#*IkDxA180m$KL7-Y4e?_fi|_f)Tz}WohSXj6`iMbkrH3St2w?Dqw~yn#;5Nk zPOXkKQ>&w|$*H*={qFXuXB)w{#^B=SadtXtguU!!tlzR>wqDurjC>~Wkkr+H0 z!(Xixze2;X*;{p0@a8v^yByf=qh?2hniakPpV!*8l|+*+H9YXK#Jhi_9=z7`Obrk2 znHnCd-9g=b)e4F9RJ#M-s>hddYj-SlYj;4aONPhQ?%-a3?uBVzW0%}s$hfgn>`(n% z5L>r0)A9$Ge3>@9_X7UlJ=~M+YMOb*uGaznyfi$YO>1S(Yfm=q57;!Tre4R7kXhTt*mG0a*%yK7)7I-)Zt8Vh(fUyZF@Y@VE~D4X{E>CQ82UmnkuLL| zB2V28bXY!fomzB9XWfou=4vS+Vi+$CL!P#H>UPY>_N8X;U3&~aDiz;872m$WO08(~i+?oL$b4}Z?G`Mi?-SG|R8At20Uy8Fo z9zR96tL%ZEnya|;Mb+bMB5vx9=P72Y_-j3Q()=!(1>Ue9waA!aNSC8qc7r2-z@mOf zvRQXlEsp?mS>_}+eu&=4^COcP#E{nG2Uc|aHadz;ohcj3ZwcsZKmU2)svolF{-)X~ zdE5_hU-E14;KUMDQ)C&xRq!u9-^ep+D-Ych9f1AC27d^C^XWejpl{$R1g-!y=zxnh zyl@qw``E8$`pi{(eX;+;kOkU}?_avd3r|n-B-`Hr4n>C{Jq0J0Ng2t)epO>@A-qNshq8K z1;2@B8XE)uv?IUm?Wye9Oyj>7|LN69*!0bT>|h;!LmhIe_)uL2d51*_V^wFU_IAdQ z4;;5K9_)U7kuO+}?T;2&ZSsxeU+n^qnp1|MMU;2-yxy%t4rj7PCO^WA#l6on$vm6O zdpd_h^Xee>a#n&hTDc1!W+enI__UU`m+?$J@0<9O!9@x_CF`=hbKKghu3wL)o)s}T#kA!w7@lTLb*EqTWc;B#>OJIH3BkNqJ8NV|@~yFEO%C}H z-LH4)o*&>octCDc)2xI3d+c~yi8%xL;trcLAq5?OIW>iOZxZj_%zKju1=?0H?iGpF z%eCOM2^j09qbI<5BQjpU&KjF5orBIDOby9PcboZ~@l)`Zz8JgB_EENzr!&|Oe0*1I zQE~V}V(PWv`ziE6-MYZoI&`*TMRo9*)}-p7z0M96ohv09^e287Uu~j|O8(2SRZDs1 zA%3^z-;TfdlAt$d=4vjT!CKWp^wSn#r7GsD=cdB)23qP?}MY`gKYm? z@>+j6W6W2d+S6@6^PY4pbP;_tPka*JOZ&NEte4gCU29~;TJHcS8mrd!*gs8M@SFOE zpS1R;7>DY{%w`O=*YZpjxgf~HMEt%9%u(g*x$752;BFYWo4`1Qx0_g(F!2g>#4O&2 zuutlFtuf68=e5&qoc|vO=lZ^k=Z5lbIXG9X$y#80p8f}a*0#sun`O71U30I)C*l?T zx%#~+&VMrg(jyDUM$u0Rw4KMlbi?d3Nd|VovYR-_4Zx5I?B=)jiDq5O=#ciLwg~uz zM_v|9HTRM}Nre9+;L6al!uF|LdnFi`0HdqBHD?xlZa&`u&R^&*ek}M@d-DVE6ozlZ zGXrf$8j&~nDgr+%A1!bZddeCYEZ)`0ze~ZT_)$Kx{A?d%(f6T@NB6YGEIoxjGI}Ek zy2+L$;d>`TBaL$|x>B*|J?MlET}l6Tjhdv;&CTs=W?$41+ciFD9yqo41#&(g`m@rh z72H#OV)vPz>Juw2E?u&W`HRuVw0FsaXV)<1_pI_!!I6yF2dOiyab}%MyyVU^Os)v< zYy7)d@7H%Te(q`fNsKd@aWmFKvpJVWpL3}@yaBm0*VH0rO~*cS(8!XVXZy6-Z!tRf zSpHe%F7mF=@XcKK=1ur!3w=pn=(i%;$f7S@7y5k$pLNo!$dc~OO7XR2vBo60U77h8 z=l!SeEozm_tiLy_H3Q#Ee5-hV%!%c>xLfAL^t2A9m|n+yC$_iDiR}qC7l+d4KLU@U zZx|VlNDpvbM2*A@^uPwQPUfz0+j>bh-__xVU-9VhbaW&9{RaDNT>hTy@VES3yPkh% zeTNC~{lpaIK@cBBFW_6PY}omwU;{E$hwhTD-1Q}V6n>B4hklLqQQMZ-IV0+8Ieqn5 z&)>*<=YAY)3~NpBdpovF1+@vKsdG=9@rUmIvwJ>`ZGURwVECg0LsJ|K8$2+?=ZAcn{u^4l z_Ww*=|F8cw{m03(89(((2UFjY^CzlPvvaF^jZ)1YKj%{?nzQxHToV5!iX5Bu8e~`Y zt>%OapDnW$aqz8n;EQzN$7jwf+2E;J8qY7)ar9xH)TKUTmi!F(#jQ^%-D<%@nfxF7 z;U~|9bRN5_SG%j1T4Ls}|1 zmwKABpuay57@N(%`^+BI2*ocUFUQpU%=OGu-CCP=erRIoVQ`F2zcGSeo5}oBagRJc zW!uf3V&#IMQ|x(4_wfeFZ*R|dlwX&L?8!#4H~D|H-eyx=ewh6KD7L9L`3stN7ZJOO zn7N!k)R%jPUha6&h0>K8-><*g$!F3JuK)bs;qT!Wz~8TU`u*SL?=#08{cr*N{ip}- z|4sh>{$!p8e>>hnSC;TBRoLlNc(5e)4^_E@t9P_=_e6$ahn_0=ugCK`r+SXrKIy z8QLSDwICB8;=0u8(Pi!qay2aow zxw#LvYcu<_7x9S`&y6I{Pdjr z!UyfD&50H{%;0f1pPIke?IsyK?rcjLZRqFlj`unqf1ipDtZ!iWt84za z;(dynQ3t+8G1s}!SFx{oXD6~R*E+nHIrRA2fDOFm(LR{Wm{qf5qUVq8qt_ zUc$woZg8>N!%r>wcD%8P{ly>m9n=wTjNAqtrbBz|Y(4*)dpArcpY1ka8qJvCW$g`; zd@FBlK6q-UR@BwREi3zS9s#(j|D!dw?ih2@d|-h;hq9KgoTJYWuU1YO`Id*nv>Bnz zXg+N+rdsg+HGWr}y>jHty>Hfa#T#+Y|AlAuo}QZnK862E^ldq9M}RwwOw=ADM->04 z?vH`7Z}O>{Crj}u+_t`rE-p_aR*YP@{plS&FTJAA0?TSy2t0~4Doz~(jydpdF8!Ez zhsHq6K{Cdi<%rvs+m2!jf-_fsns6vxKpVt5sO7;mztz!agub>wAMZ0?QEUtzjDSzU zmn5F#`FWOAa|}5>$ygMR)jOh#=%cokKVJv$ZA4zpSrG8xj@}P9*ZpkWTJ_h5aYzQ4 z0|!li&U^Rky$s~j^St_%-4c8c0-r61cHHC|p7(xVMeN=Cy(c-ni~Y-9y&oF@UXaso zdiYcEgA#K#d8hr2*Z)bADE~H}an)XL^OxC44BzL~~- z>3yvY$Oed3ZW}Rlr(j}@YqZ`M&#Eo09W8L)7jH_&=XSiGXx<-Qpo`T;Zf8t7>*Zxq|#Xh#5JWZ}U{J#!A!#}rFt+4Y+E83w^)HWcKb<(qJo>**Z3C07=WP3njIy80)-AFX@)7&< zTzCGVXeJw|Ir}& zj2e&XCp`B-`;2<|33?CRwHqB{{4aE71bmjm6XY_C_P)0sIJ)$M40iMb+{X5Rw+!s% zEq|s)Ci=x~A3I~(H*Zu_!(@jN9Trk;I8@v^&!XI+1|Den9}`M}~G{DGQv zgRPcln7_C@C4b%hT{5@qY~JdLiAnek)RkZ^)4*J&-b$v9dSH9Q;Js@zSeMua+{ehr zsVA>l_+o7KoWjBToBZAY9vW^)IoL2^(!qv-DF^GUq!rcpBRSNh>_uIG1tiiTw$AtJ9KFtMXGHshX6SS~Y^YHvmO~{uW3;0(%|5*_IFT|CD;sogx9(l>rnD-@D=r4 z`Fl}#Q@Tlfq~F7n0&QP#{GX0z;GGyasfBOdXL5K(?<*GC!26Bzsd(3}J*Zk7sUbZJ z?Vgk!H|OAj&)LM73-DVNvyAx592@Gc2RZWXcfY$f*8bid?#Fe5`;Fk8DIYCgt+ASm4yx-sbJkR8r$%MnU@B7c^ zlg~W!%ze1->%Ok*zE1aji?4fx3+Wc%R6N&>ohI5^;k2_t?eLqy3v|{EkEm`1b;YY% z7gWF0sjE2(QCGS`NOk$$)Sd3chEK5;OD>D2kFg&%nWv@aNbV))lU+ZC|AqeFl03gY#bb43Y~QXiIq3Z-W0e$u-6< z`(5&@mvvBU&c&2@71<&i{%P8iZOWcuvrguVmqa79skRmGJHmLp>zM4*4(h8t)jf|o zvg3Qe<*V5L$@SHb^cMLx6Lzo8<$#XIlN{?wi2j zGH`M;^=<)=CD3jNIbDeS0jGn^OQA1o=ndpF`J0^s&_pf!L4S_EU&}tt`}piZCiP-J ze@}DbX)mEI=4g;UOQ@?lIq6G=Zs0R&>TVUNrXaM1_eM&yq6j;9;hWAMP(WG7XE2icR`Ja~vglfB?UvP`u3P2z3RSL6B? z^J2>~$7XTIOFO;bt(iJcx%?Qi<*VY?=wm71^1uWg%STcX04DhQF<^7?VdjQ4N4CsU zz$3j^@JOHRjl<*dy*Mz*XD54?b?i;!?@NL!fyR-u(jz1*e#+kYR5I<_WO-rs7mvj6 zqpX_|!2|9^>l>kO@pChLWx?aU^wG-vt;DaPd*qdWaxG(732s*LZUCIWrr(%H{WhQ9 zsnuKZ1CZ>rs%au9UisoL#ie%=D!9{5snTy2fOCpna9 z%k!KR@VV_VZ!R?mA{nB zSY?-LT`VT<*^4eAzmV-CQ~ot_igQ^juDUXO_Z!Y#Mvnd{7)3wx9hiC?7+n3@nKNS&9GnS#n{18LZLT;`-+`C$9ze zeD!h8k(ymKXRUm?839jwonyCmm)h$*gZ8pK?d==8y)|lY-x;(w!P8#H*zJ8??RA_% zdy_rw?H;?m>($=wGiYy`-QG6M$4g_k_i45FQn1FI3(d-#ZAKB*GYbyN}Uw_SxTO8f*!n7^eP1{*kP2bf29^ChSd<5?Om!yDuf>l#L@xKT6A|-v0K}UxagSH10CaI?MaQ zXHVPTSI(k8^`$Yo<9nL^@<#C7nAbM_wEbOkmi=jrZhv8okNYv=<9mJ5Y5U7Pi~daR zqo=?BVfw3e`|HYk?DY5yoJD`y2cYwco~4eRCzICtlzFz_x0q+qT3YK)&-?A3@>@Lb z|H)H+v*&%2r~D(HGGX(qeI#kE5AvLYeBHiP#^7Y zVZWm<-FiNP?xdV{pp((;d20vOGD0ekXadfj;#-pMEDGPuy|4 zazn8doj;_tRr;6CIMvyE<}O5J&kerW;UUHt_oq5*>k;&&ry9w{?C>8S&JP@a>WW~^ zk#$zhQ&$CRntfBgeI&Da^N|x)fxh2gaN}D?>fPVYvws`!Gru+a&bPl$a=%|-f1m1p zztH~9K25vr>GpT_p4#6pw!gE-$MnVdibo!{YL0LQ(>wW#j~`*4-(jA+n}RiwD~k$x zn$b%_MFqV}*=KVjdlfhncwialc;3vpmd^%j-1{F-r$^ksbTl1SBO^V>jiY$m_ubXx@nCs=DPWzEAkzefIm^hLoM`GoYojQ8o+ z1Z%YB9h~g^9t_r?qf9$E&H27KSRSj?}Za4O-bqo@4<0&oVR;sPWLnE(FbLtdULsDqbsNB@=S6skgav#WGnJm_O|x=xbpa+=UET(*^{fi8&|WJu-+Fg zHTM`_9^QlQN8BnF0Pn&d^13nap84To^pvMxKhdN(5cv>^Iw1DC;&yrDp=i&D>|^a0 zQOqZQJo^-|pJz_8VnNmo`PQ_*Duhh$o}@iLv&rjpY2Kcr4bB{H7-J6Yy_?S2=OMd} zS$EloCwg2add!3#*+*BJf*xJ@mYsWg`F4^qiMI`pY7Jz5H20#Bmo8!Lg?98QS%$-S znrOvgh9&_#|&!*<<7m z<>qE{_QuDo7kaKNV$LL&tngyW6*@Md`jbs)`icR(jf&o*tV)hL3O;q;%M=vQ2*L*e1JV@VxT5b zM%nVXn=$u!#=Ko)cFsd&%#C?ZIOidrV$7d9&6uxEf_1Dh_W(x(c#s_i|8>c*eBt!4 zOgarL`DX}AGQQ7Es%PIBnYd0pq5LCk;+^I*z4b9ZxAQMr?QXDra!Z#`rXd)MdC#lh zu5Isnl?LCJc=*2D;rogtzAtz9zQV)zT9f?nz1A^T&TAclzQY0Jym-Z1Pke9hKg?-G z&b(mu>cF>>?}~*;w~;?m{zjWVZ5b$@8pl40H27*sRmGy6f$#mWpM7B12G3Vm=?$V= z@&3>7@0vb;@9_Q~i(_s)ia5|T`G%!~AGG`?X3BY;Vfkmo>(aYS-4jmT;gPyJU%>q4 z;d{4T|9I<#5bc=$9U1A$o=Rko`m_C+_PNLT=vt>MFFE1XsmG7eo6mgO>ztw0U+qfb zo4|bA?qCga?2Zvx!MY|GD#>fn8t2N2(R^X!TKdN6bd6+q{uy}odii+5_|w7@JVSUC z&vR^jGxmQ-?(6w+$+rH;*jcAF_F3wS=hKZ{{B(e_hKC%#_TMkI*W4>ot+yTH-YiU; zJhYmes}DLRB-#_v2?^_2F1)GrR=Qyz)98jZ&FD^B$w%nUNsI4WliuU%J1=HgkNjbK zR#^Mbyldlnr>#G14}?G9ckgfCbAAg-FHeGDn6U{4+m@PO>)|=Yu@L;X7`?O}y)?x9 zm$0UL*MmRM&Ik16?Wfe~Cqd44*lmheN*y^;h8*$s*S*P>O}WDBreI8G;d%9GJHKOO ze;;Rgwfn(Yeir+V8j&>t?sFywLHe`wf%Tk6rTts7;mHAMAD&~iOJ4lU;l*-?XDi^v z$39_uiGWd3F0M3_fNTu7uqwU0Wsq3+4oVuy;YPT!h z@oMJxQ!f3YpGs$ZjeXGFs@@`VXes{M<9-G&XI}ZdpM1A zGXgDR;amCQ!gq-I624yt-y!g=GY#rb*5Ah+$_>$c?iJ*>J>Xk0MDd{ROkNIs$+we$ zd-^lD=d-tIG5={T?(e-##eDa-7~C_41l(f>?D3cU`n0%zD8wDss_VhMQ#U27XLY^l&EP+?WjMrsV%EVu<}9RK>%c&$;B~8J*P&>r;PAmv z!O+{Gf+L-wf_H{O1@FEWD(GA4tEu%_-#*CyZvLD3FX6wC|6%eC(x=?Ld8f5@m%hI` z`7WNnyGy^@&wszG)_vZyW>>TOY^~e1JO2FrT?d`#z8e{D2!B=~JarRs*BjYWe1&E5 zszg`JHTlQjF7r;s3L6zGe2}p#ceT5MG4h-C7WvT;j34za&djImzrrKYieM}XE~2c7 z@-Hf9C<;!Z@=MgG6+|Dtj=RXOZyDHrLrZjOdQ0E*jFy4?%$8^g<-oggiXzyeJ^1pX z+wouhhIrW9!J7Vq!J2_+ux4-|SaWCy{}nVm6msGO?s+Sjzy*Jry)z-JrniB;1~*zY z{f}A&1Dty@xZNr^^te@U_(`i^=vk}a$Sqcl=;^Iu=t=e;wA2G#MMFXAlIwpX{XUAX zrkq;8P9oMm@Qu%^tRKjov(HFr~uzTGIPbtkSE;Qxca@TZkvVseLKhbJ0UwuX~xz-v`MBzK!*( zAKB1XVZ~Sr!;hnH3?MU}fYw^zl|#r?$$iNr#m)zx6~ET*vd<#P4X^$beK~FS(PuyX z>0HG@{M(00g0ZahJzu?y*m+jfWoy;G+g3SkdE2>pjCRzp-W%lJ5^`L) zyFRnC>f}B5W%;azVRFoi;d7m@uCtXy6X-JVP!={_Eq7__E;nBshJ0X9E?>{pLHjHt z?!Ym*uQw}Kc+Xceeei6DA6(kD({{<8Kb}(dbUCu8YgKC5)3w5uJzc9(%buKdAs-&KQ;KlwVxfD`;WY8cys9cyi^qWC0 zm7Sky`$g` zD4zmZ*n_M!^Us+%83$4ynUyNKfC$Z4+N9+qN&L7^N}gjp)H37-1$G0m+1de ztf80nOKXJWmF7RpodQMEtrz^vzw*xI+qo4S8hW(t$u9YJbY_XxOxsr@-*Q^3@r==d$q(`9&sjgo?Oz_i$9)$*ZuWb>P-%SJfn*?Kz%zQ!cYpW=G|i}I33 z&pV&H_)H~hq}ML{{AfADo~yY3{8)!`9tOG2f0{IM{%)iKUF&41U;rC=_@sZ$9_-nr z8P-+gW4!qq`6&la`j5RlN4&9e(ZQdP&%xfv3=6*y&qc_jVSMgBK9|5pwN|>h*WH`r zEuR^7WtaSK&wK*^o9sz+ILVQn^ijxNwTb0zd(O~uhpk(7mq5$6ga7X!Q>)R(R*1D9s!z`@@`{OisK(pJ( zw{l-2cV!QCV`qHP;Om^B9&F-w@E6#yo#W_>BRDGqXFA`x?>p#&)z}gE2nJVSo1-Ti z{~@~L5%lDD(23vm;I_L28;8BA3#u+(L_XuR7UB+{;QKjM-!gFRjSVkNbPf)0Ixvh)qCHt=&n9rqrf==z(f4lFtO#)BgFng45@_ol&I1>&tFM4ILX45N zr2F=md<=WO(r81rfcA*ypoiAuU(op{5peCtCqaHi=Upf7x^D))To=B`?ZMb(9|w1| z1@G0kZM$%3|5p^d%-|Nnb7HgyZ4^D_JLTq# zT~2UqWi7@guIT|D*<=;MAN}cUnm+6@Jv+W0!$-ec0IsmjW@D4>v#f?EuDR~5tn}~g z%7T~r0?G#(%|mCxpRcK14-K6ue~O>AzHTE2$?zAn^N@L-*xJMM{X9Pl?R^4TyaE1s z*FUjUd~g`KQ3Cyuf4ltWyg#Sv)UF=)yEp5ebI}0pKrdcd@+h#5$rlUp zt6ypJ#k=H+zT+8_=JG$8NA7zkPyd|O2|Sy#wt@4j%Avv!h`zk*X7fx=HP6d=&XK<) zg^uUinZLhgwET7L%q<>W&-FLB=SRy{OFlok7x@-NzV#sY6o--BsWZR&!Sev{$mTOS zztFJOm0r$rKMdT57?1X|+=hHp!RS>5m?~Z-F(Dy|X#=VNI*``H$rh z+b&_g-eJwyPkf*OI!4BX7gXK0sL*E}TkX4V^TD~+u_LS*_osjFz>|3sT6I2{c&NuPM;7!U2m5bE7S2N!!si|Mr14c3 z3{od2=0CQ;*4Ns~9JyX@ijVAWGnY4)N@Bc9(3P#lg?5}^yl@n@6Ic?zr~nUuAlvOUZ0xk z@9_8ndyxNq?6LFKEm2+e)$PFlA-SKJJ4M#@W=H^_CqIze;pW3B_~_1 z9$QX&&yzUjxAXCJM&bwl#C-hjIpl=qC+6he#GEUCB{^60dgs0DK07bV)<3(WR+#$- zo7O8ACaJA!oI0+)XvXGu^2U?Oe$pxX2i_<5{o{B$-Z-I?@3;wEHX!34VvZi>AG$0T zZZ<*d4dD18_~qeX>|dct*FTVLboY_-{C0A0!Ploy2YD_z8L72uv=;ZAu);80V_tuwVlWZWapnT_kau8&TRNyPBsG5#zQ*hMoin2GvtsjB^ddTO$k#<6~ zECH*rrXZs`F1<_y`tYuYG=p_7o4JdbY@i+iEv~N^U+BxfOewf!nkmI$A8C2J4Y*mvqt8%BB>o> zB)_7a5czqj+WCglPCfaz71x|UGOr~`?Yv}#I|L8$+f?m*VFaG8ingmAc$imm7hJrU zaXn{+|Cx3w#(?LuUU;e&nRd40?~LA@)XtNvlY$34d*MN^{2KsdU&R4TX%{#Z_9oR$cv8;H0!N{DD z&jAj&58(1}B%it6ZiVj<><&$;9q-)6i2FM{+gK5pI5H>Xh9tDpY=ysJaA40#aytV~ zI|iONTzHrh@=OxidDIGjfp+p8UQ2H0S6+DTw`GW#lcuD0nym0=X~&fzZaZd97`Hno z`)SY23GbXa=_+TxhQn)(#ijGzIbk11z+QLiZU4Ju=oI#!Kc7P1R(?yiId6R=e=wi; zqVg9Mv)-!MbPjXN{7N>?*yWQJ@n_w`|Hug|1@bt-bYx=)9F{|m|$-=4Ch+^ zyq^BSWmnbiYjq#TG(E58`LLDUaKj~>QRUC>ywo?2JMOcFp2XJPjm`Z{>d3F-=*Nv^ z(-+I1*;t;lm@&KO!rM4S*FSbIIcWpA$Xs;m63(Y-!H?06j-j()bT&scWW{>WH3Xl0 za!cXgIoP3E8=9$4%yfC}Wa}7v#g`*HyH@01D_Pl9I_KKh4)C=SeTMiV_E_yM#jthX zyUy(Jo(U`6(wr9tubOjXIIn9vHb-SHXO#H^odcZhHDLYVhjXmVhUD`N;*wA{ydb#C54 zeEQ{k?%q6%z2D7z2oAw;34QDQ5VsHV1%{&N>~@}nC+6hPyymSe-@{s+tf`-{!{JF&g|t&f@a=ojb~TtY})=`&XA zgFNS?(6hydmwV{j#P8KQnP{ub0)Bt3*O%`xp}i>z#cw#;zJtdd4T8Fpxf8LcM0@*BYA&1N0&H^hsW;x zVT63CrMyG_9a{iRi^nd-C#PHq<&F+$jYK|Z4*QrJ(TnEn+w)l0z`f*Wt?IzPvWK|G zyt989c?NvGLc9I&Pz8A*{qSG~a_(`xgWrr#kh!#dRGNeIR&>_*z7g%ejzSMPz@_{o zWY5rc?1y1KxAFZ7%4$yJ!w^imr&BUZ`g73ONEe0g+KG(>9mpQi8YLe-GWOFZo{v0` zoWGbH5`C9V)XzKjPAZ+#ls=g;LwDK-tQ-b>{OtWc&zuS2+G{$y-lhxNW)tmHIJ8sg zp&iknx140-e)<;giC_Osxh(Lf_@cwSxB2)ddXN6Cayv{J$yKX$n|apzwAN;xN73C> z>Hi;Z4JOfE6y8nVFD?4B^YKUcSLd&2{aL~Kv-6ue+cdp%hm~=k`q6(0yr#2COPHTS z%9|U1=dRh#nv$*WlWKR(X5FYQwT|UePQEqS*|S-Da-hdL$w93*%(MEk@$mg!l5xM3 zT!J>Fqeqz=<#z;-_t3zxr>R$a2W`UVT6YrGA7pVwsr6=FspP|*yR`18t}P!u^47_} z98=ypzegC8WaiD_S9|vKygZI~tJUOJF|UIC^QV}DgV=Vie&EVauYD~1+xbd~_|Jjw zx5}>?w2s}#`nF}FZ2wEbzuW;`gPR%fb`JVZDf}=q5NyZBZ@2Mi%ey_$t5rA6&Y$#* zx%@P;EoZHD3_r&567bx~+zHPK@>qKa!SgH!&pNk7vX~s3cKj2>6fOzx1V;Sz?E!eN z7g*PudG`%%0oHD4LAn2bqj}cc5`$_!?q9Q&-_nt7JBW$zw$j$*GS|(-v9|KbnqRPS zLhy&p#Hci0Hy4@rH5>7xY((Z`>pidjayWx$6dtT|&l(S+kJI5Q**_wlt4J~5cPFiv zho(y(967H`z8>*X3qIso;G-X$&c>f19$WMmi2lRl`HZ3Q_<_}r{791F%qeELeFrex@^MbCoF(J<@B4E%;A=#$(JyWBqW*Ei&6 zEi7d1tKclf9sF)&4>}=z{IDN+u$S17aD`XL ze68i{)Ox5m*bc2pl$Y+E11?2VuU4)2NjhVB^$f@2()!S2p0R)@?^vYQYAk*}PdgUY z=SGc%H6?j0QDo{YbW)84JI@{qJmnqBK7*I|SdKH6SbQwdzt-INx|K99?pS`pSPcG3 ztz#X?9?kbwhj;ggchLvLyDJ4JYrlSv&;4oo5I>!AZFFcvvi3ve+5n?P&fg!+v(~zt z){{JocHeh?KhEcSp5H%kem~4-1iO3C{ASDOU-3NFI{Ek9IhZV?&xL2j-&(t+o0JBv zaGv7a@aY2RDGEMi7gbKQp1*4{e1KiD^xK;cWcr_Zf47xcEqkKs%l996{e!3Zo>iR( zy`rnO?La0QT^0U?ciSrA-Mg<*4o{%-6)RxYSoW)#+-{v$YU)^X!e0aiwJZD9eW!NF z0bQWd+d)uAIs?MWgXvOr8T&){LItGr|Dzg6^|a+_kTV0 z{;-ukuPENvxd!GkFU;r#$uPTZ>MUpb`MmJQX>-%0`wz@1X*F&7pey$`!M~S21p8~O z-;CdkyLrSO9yOLt@a(|FHti>mMQ0RdW2_%N zXAnAB4V~Noos5G{a(v^evygYy&`BqBlH(g+T?Bu>4xMDJpB7dOzKk_kbRxNa7&_4! zuoRiDIoE$7@=N(lkK)T*js2+4Td^O9u#uFHegr%CFg&Y$dZOLzTx+?$lVe&S*)*42 zG?g8Xz9ug`hxBFT-p_@R8Gcz&T~+|9T#hFWoG9g_EsJ` zdGCEk@LfNU_v+?6-dXr9c3Zw4l)%-OqXa}Kf8Zt#1UwZ@K7Wwy>%E;6=3 z1R2%Yjy;jPkGZru6}QXv1&BNPjs<{i z5S=c=$GrybxUpwa!MezWb^Uo^ot1PSa^N!LfMDGUY<8XVY&uT?>lK_YT}HmK3+r|0 zrT<^SI?)5Gi{DgZ^2f_f3w!6~BYc&%KP4M{Pq6i#4zIr5}6k|S22JTZ8CA)m4*7ntwCUDM5HX6q--`ycF@){lhB|nmNSd)(j ztxK0}yK3~h=;lJaz<6DGDVdTiFC!saUPh6dJxh^+obzf~*Myt5aej~NzCw5ho-w&Q zH^MuNQ+b_f(1~Q~Ze&p!da^6~n&Hm|^y2|&b1?VS&4pIh;XUM+G{eL9QqJUWAZL4! zx4Pq^3Y)>~)5*5hNOzHcv+#OrO)m6aSi_!t`KH*%W5?Xwm|~#y(UZeX#hkM#+i~ZS z@4qzwjdybweE;P124Lx`!jJF-@u`WQv;EtzBhN3*#6LhBs2@Ig9sThhM}B#Hxoz_0 zj^jLU@@C$YFZViM_I=9dW*kEe=o9(KopLUQ9C^y_}$eGU0Ah+dJ!{;;_`D}N))H?D#8Wm-PtI!qbH&^hEE*V%cp zfifALM<`Q89z#F$D_j3C^Dt=gUNTx~W18y=8T@d#sV6&eT&d=wm2u^H#`WjS#XRUD zA2~Aw+^d<3A?9NIM-65!Zq;1KHv#R64wc6vp4QzBQH=#!`7-iB>rBOXd%dZgWczY$ zK6dU&_VByy$8zT~_3tCOi?%%ZOSLZ@S#7&^L89!*YR!elMQb1D^Ahu%)jHjLN*@oj zT0D;>LnJ@8uunT#ua^9H--+wE@?$^i>z^i#+@ITxO{%rI1-`Y&r!#QXTV{WMH+b)1 zzDM;-e;bqf8|%KvQ}u^lH>$taQuW84ZjEtOQlG>aYWkr8sgW}fgoz&MEkI(6^jJ_D(s5!hORe#PL{##O?nnUmS_9ywHv`-FS9KQYZp(oLg@wXkz zXYG|8qkVM7heVkZ%FDK6HEkvaNUTOaX~k!>XGH5#0Np zGuq-l)SH7f0b&q*H+So#&%_QzKS`g@Cu2{)!ftOR?X5&N>7ku`+VEe>{vPJBkQhxp zJ}AY-w^@PCJpTL8^*54htvxa4#D<&D1zym8-McO(9*R${g*EUl=3a4+rOl3 zt+hL6+VPUDz=Dn4s2E%h^iqVqItzP#HvQg3>_qz83hM5l?jGvSrY`;xa~AL2#8~iC zH(J(t%x6-pF67mno%l)8JKLw_teMVlkK(T`pK8|Lu8Q+V)?UM(-ZJ+3+*RROpS@*> z@ii*OM|`hM@x8o%KyQT>w zr}oDOO#M;veJb(aB~nMQ|yO&f-PQ}7yj2JL$c&*X%E+&scFtI-j4mlk;# z1@c*nXI9hCUGPjgJOj_Rr{mX`ZJiD+s{b@-&ZVmr(A8|{%J2-lRRTZf6Yq57z^L(w zpVvFQ9G6e<967To$I?%K9mt~rV2pU>5xgkgGQ1Ce$uHZDj1oWUZjXHa@#A;ijjvgC zd#sFxt@w5Fy064v-cT za(P}u?r(?Yb?%zhO69Zc(Hi_`()~{CGk|`Fyo@G+ED#Zf-A*zHlx2R_Ty(M zLB0c%cp1A6UrFJwi#gjfNKTsS$xo%aC04rG>!mg((njPacKVsQqx-pS2HPQ)$(l<0~g;$j{Wa3Z3b3*0%4E!;uHC?qjW)!n{@Zw+yIJMsc{|df~tj-2}7w<5J7T@@WA#i;w?-h@BX~3ILZ}guIn-&bs zI66{lABLx#v#wNn^sNz}g|8$Ox(SD)_rfLn zE~PuwUp|_zxc@tDkGuSI2wTC+PvW0Q1$4X$I;JoAA>})dLf`l<$?N@W_zrTsx|dn8 zb>pqJb)o6uvhl%|NJF4y7+Q~rR&M4Dy9(HsEL& zn;ZVR+Ja7$XFTGA5ncF=AGDh}`_j6TO+n6{Rc_Tb=n`Mp^LBlm1DMwOqa5r7eyiuV zT4FoOBk6`0qwv5;yK2++t*Xtd)n*3wT)?Z!{oJnoA9XX)-TfxkBi$jJu@{dgudRE2 zc+wA8gPh=+LPR+yN6sjt1dG-ymA z->rnUBFHcCtUHE3{NH1Y0iGQ-2KY*2*r+k^d^TeUDM!n|X0^`edDNN#P5*+u!``*x ziynGb{?NTgn+i_XE`U_R<{$mAIIj-NA6))aF*FM;ySXM`|&~VXFZYJUy0mTZf<{;@{T!^9y`PRZtQ8klXnv9pFsQ)n?E1@ zN`1Tfi|iQ1k|Wspx{uf&nG}8uc`iLhI*b=iH*dM0{`R8l3{XCLbW(T__yQAE$A;I8 z^-K5((6`>kx2tjL)6}OgQ=eF{slUgmKgI4-<2Ch5!4>h&G3$THso&<*KX;_Q@?N(v zX5mY|_d#rIOgD3+Lca?^cU4=iG&vNxj4YcKQK$Gl2bT+iG_*pImfJHiWt znJdO197nEjaQ5aTqxa!+G4rM|bnAEeACXlKFV>IP>Gr)gI@{akUBj!+?~mf=adeQj zlsbr0t{xsz4oD53>&X%RE}!aC`qkg@EPJC|pUh3E^Z5-vrFVUkPnUnR7FZ77d?uM+ zIIGEtS%OP8*c@jIN2YFB;I;GuScusA| zUe>4Dh}+A^u&o~76nit^ENTOHMuL)U8KeM7w3~!N?1@JZi5K;^%DHO7!8i>E#=X z+bT;PoBq3wtrFt9@@)(G#2$9#Lk9jW<&0 z{9-dR$uGuUN`+U+{q0wfW5c-j!bdKSBx8-A0Gf~=F779w{3t&GbsaxJqw6Pl9KKaP zvixk@@aGvB>GcgU$^USKE=0M55A*kcy*0_gU@UE6rcaVe7|$oI(_oH{03`zcKJCn z9vYYgFHRltlV!1goOaFL8#lLBezJ@teqW7mCLz9rK2D6O9et=Ehn&(pa**(!5TBkA z*4&onXBa$}g6A#JfO!Z1-S6PrBA?u{=G$fGWSBF@?fUtrtpEKn>ThB1+jQD2eSCfa zd7zJ&=iMVP>;Z-~)Ri4zV8OP@g%Gap$%e`LyeCXNuxv3jbT#$9btT}#nO-gV54do;s-{G^Df0v$w zr?uco^fT%{_se_jh^6^(nWn;CZ9Rg=|t^ zKc_Fr&(n=7M}0wGk_*C-Z(8EG&QH--d|c3y`r4&&@qD^*1=JU|ndFP<%Y(m<{?VPQ z3*uvn_k*pbetx8Wcs^Y}hiR)je*}Ntd5XII1nv7V6KDc`O8wj^-0^%m+;yoRbn;Z= z*`J~xcRbM?yPrz>(X)6)^gx^{HWPZ#-r7>pgU$rPPm*W$X63ZL;e$W0eK|< z>rC1}Zry&0Hm03H8{3fG-#A4Z{xfK!2Oq#+ouZBRiHV;Ur`-K|%_-W5oJkwRspg%c zjjl6jqn!Q47oDPw-Dl8-&hO7o)dsvM8hb`<@GLqWi^jGA$MLKbbn3M6PiN3Z55C=j zRBa&V1jqeqgXdGhfnKuf*Qwfo-_^!CwZZc#+aUk*)l;-_>lw6B3{O9wstx3*;P_j$ z!Sku$D5s6>soFrMsEzB?h9h6bnwJf<@xv5t^hj@`jfJ$KXURhODe{pY_~K$F_KNJV z{Xh2lt~GqqWha|HmXuGl8hy^I?=?;h#_nK`X#~9^h<;q+055fCm;Qi!S@XoXIcs`d3-g&|HzZASzq`+@(irg*1`)%M&9|K;^>BmyQ z8cD+IRu5i#9K4wV?#^=X?(WTKnDd6mdM*nvbE{i<46g?!<=V{>up9WrXlYloRPX9m-(mi9r z>(J;Uso;f1{}S+iDFuG>#>h<*jqV5DuZ;n(Xkc{;SjFF7-e2#*tLVd}&s6wKfX~4v z@TJTx6MS2N@2XS4SM+DVH`mbTh`lKJc@*EF_UbOl#II;#g5G@v>8*2#?Ize>$ZfC9 zE}WgbbTr>+kCVY!zR$?ZhzDoA9$FMly7>F=z_*0^2u=szzyBYEZ{iiP5$JFzrfuu771ZXFLh%locV+e^_%aq$FeXOy9N7cI{OJKuKZ>JKD*D@ zXLtv9J^KCN!%I${xTn@1sO~?%d9z}1?r*o8EHUrPSR0m{oPN)v!^1DAjBNLvmbFlO z+Ad!PkD-&QkHw7l70zG0_D?68=BDx6GUko= zbvD*52yeN-K5t>$OwI!U9yj0VUmQ5LQFjI~Y{AF6HP){GGn9|(-D}H_ z8yj=&jN|g{P`;}ioW4waGR>;{q0Y1mSUo@4Xr(pC){egemwFcfk1v78jSN&dh%bUa zwXHn}UR`sG+1tJ_{NDexeG0}`Wu=9O_%0ZQ+x7DEP*=7|7Jc0VoL--Z-dD{_i5QSEHZ{(aFaBw4V z>Aid?20rFjzsu*PGP+w^{*pBO4-x3hr(8Gkmx!6i?q7An(6V!Cki?NBl;Pq*eDha>V^@x{)Jk)siEAM~9)^LhXG?9ue8IK|&NbR*qHf9`s6h&Xfd`4Ge4 zV0gDxF#M8LAm5wjyb_)cB&{9kW2_zaoCnODr+e29@Th!YoeiNir1{Ih3J)7U-9qDQ zZ`)W$u9W79IqU7<8CZIeUA@2)V4g?zm4Au%ici=)Big7MRe^d@k0$p55 zUHbSIbYa?KY#QT-`IPs=7m7Q-6f7_}4%s%p#_yI1QpQSArW@G0&!F7b<|AK{+E0)# z@c!*dbEiG#X0klWATYlJz~T+ZO%-GBGH?|j%4OiO)U-g@}l zJtr^Y?>=loZhc1k8T0DMxnur{`OLG@eufQmK>C$(0(1_m%`c<%t8+bgKH!W&{NUb) zxc_g3E1N1g*a7P>E(+RuWBn}54iYS|Ll1;IcMhmyv4+z ztRcl7di*uJ@~sVV7EQ(*4ck2;B_rPT2DweYXV7xq6ikr;{hX=RaN| z-=wlB-Mp)a&Uf^wE^*{0YRJfYrcwlLO;22gb)eFh1$PSic|v#&39GeAWZwv*|J2 zC98g2T)J?b6-@cX1sgBE?k$bG@+BLF+T9K;`JYIDhc^Oe596ga9xL$s$MHDFfpMM#PQ7-7 zf2U^H{2Ov%^q&=sdB~c=i)|RSA4o8oJSk*mF)~Fl0O7GbCnHww$e42S5cIAr^htw9 z*%;{ASAEL^V;Ok~dFaF{S4u9zF!yG^?crZn78W8yr0=OuyKmy<+k*wg4xg4^Y4=sT z(CIUp*yoqMeO~S9^XiP)Gas^!RG!sdI&x-p0qrT) zfPcTM;{Jd+|JJP+sX#wu4Znx_`ObN@w!SlRR`p!!EBA-`M=P$oC`J9+U29KK|3Yv5 zZ>FdZ+*Q=yPJGYrz%M(r9{g+l%=W_n`o$^wzjN1Ds2`;rx4ygIPveta@1vbzauB5J z>z(FNb65V$kcrb{3wIU&C}7UX)%&tv`eVz$wQ%9m!VvYv_x#?a`w^Trx=QEyW7j2> z?e>(d- z#hzP@?GR44ZFjJFT)7Vo2C>^|Of z`v@lW@m;5nb%}laPj4UJ@$~VXvHSQ5edK_nUI#~&o9wnL8xq^T)6;fEy&dB+^H6Vm zMGoF|{-CclXPSwr?egPSSDuZ^l{eY#mOpB@E4q~KQRQv-VNbgcd)rNG9j7+Ahc}_k z$DKBxOl)(Rx6SRIHn%6X`G={rSpuyCXVihS;zhg7ikA|{|9MZFWzWUwzN_>(a2rFG zl-RPw(Ep{A!fFrMKyJb{;W+_h3bMh6|HOS4DKUPduHr>@*^iqt+#8vy%(OsQJbA9l z=x)hWWgfH3OrT62>vD=RM~fRT3fr>jWK&V5T}FH<8@?1@i1v=i-{fbkdUklUt?V~( z9&IcA4g8G1K|0Jj@gca-{ML3@vApgdAIN+EnfLel_PnomW&3O#x_ti*XRyZkzSD!l zPAg{HOZGj%vKc;~f3j)8My!NZ-E+YAJj?KJvG_ zv9(OwZhloJ^~PgA;+Jsq@^lW1&YD%fx%AtO4HQ9-Ck{>-?EUkx-?7D&zx+Fotsb59Ys)C&AHX)0&7VaiaoBcsZ7=wWqv!+@YOQP_{WV=<_puqTK{gKOvS=6%KYwP zQ>KVA(J5n;Ick?#M49^OW0VP8V%oW!GWp*hqs&+AvCpPV$HXzpTwu3z31uqAYNyI> zXPWx*r)=kF@ozsDu0Zz~uQH#Fmq9jP6IM?+;aBnG)iX{uCFjFu8~?=!zepY^kNZCKMK8ZgM(DR5d`5ebTT%4AUe+k>4K(NC zpl8d+drp!MHQA?f5ioe=g5ta4e`D|0bH-xYdcTyqHzSYm<4GsZ2fy~&?AoWo z`RzW&m#@7r{s!gk_ua@IjH;Q7?&h3L;v5Apvp+?-_~tyqLe87Y@^>CW7rAs|R;T5| zkDVTjrDZd3Y6m-{m$G$L`HP~3=5D2ea_Y+$TCs)mmFZ_IeoXl-dx1&*hTWELXd?TW zBKg1>vSKSQ4z^UyENbb`NpBgPlhLwL`%aM6I&(^UHx5p=juqiE=qF~}PhNgMx%96Q zkALl&Qt1qhtMDUeKj=PieGr>-D`f{~23t;WS4Y(xt7WjG`1qkv$iz^z9+^7$Q?e%R z`H$KBw^PsDIZYedTdMQ?O3hi@fpF_X<{wW+pO-Z!(V^XiKo`v`wvS^Zt=&%>y_(oT^2q>EJfo?pv|ek!1`d$5`$On z$V;SC@A~yOj6c5Ku(!dxpCrP(Xb+ZXE@<-YFAP7;of+aCn+7ike;u54&$DDxwyn$0 z4_8bLwwULN{P2qD&hz&C@T&Y^i+{4!l74}&B{MgzMYJkEmCgxSFPdTQd#<))ZtvhX zo|)xq=vzoUZbG0zdq!W!hyK{w%Hy1I7|U|z#ob%_J8}eMLuDdMw6|2WM%%Ghpsm;N zyDJ8xd6aJd82V)w<#m2XfZwHaXAp;weP_oNtkzs)qQyG7>)X~Nz7xYQ$i7+6xy&=J zKtH*PIRkE$)fm{<*H#hCA^+Sz6m;y;t`XHcK3Oo;l5?=gRZhZwWp3CDC&^^emsnR{n)U!J#$;lX!4vR(j_EyxKn@gR@wX zakkyprgeTR<8NgAqSw9PG>`Ek17gF=tVjL{S}^htIcVf%y_0_~oI=k=KZ2fq$KGGB z+`EH%l6&r-k$)Yw{JWPteDD6$o>jI?Ux}YkXZlCLnY)KqbY}W>_U`s9qu)8uAoKYs zyuWh1vFjRFO|n0$rWT{0d1N=boz4*KVL$W`^2&`@^(>X_Nh|2BNGs@Dl~&N-ytHNT z@$2F;sQ)Hg26cnG2zcvR#u>P2)+3RQqJo}%MYiAGfhmE{z{Ni%;X?ZrEau*;^SS&r zmotd|G->311V7~pp_?A|9!k%e1=o@su0@bzJ>W$8VDFVdvyaK#{{er_z9iUk*@B{$oflgzt@C{? zv!_6B@Q>(!1N(OCfk*v{*D8qt>Wrn4e*CRHQ7b$%xgXl?Z3?yYH!p6fWRLO6i+yIF zaiwCY#Mp$RV$S^u5Ptzjop()0@6^7_#|=LGLuu?s3}vNt)=%(vR`Q-R6Jx8P?N-SQ z=>G|a{smJ2n0`wR#9ZkAF@6{SUj}Vg>KR$hc|oz2)LBKHTY2__d-cBoo(?FE8DNgs z@3D#XQ)4EFV|gf$)^HE^px5P$D)z25WwTF?8C=Y25#KBM(>@%I_6pN+rww=U#))H(zG|3B>A_R0d? zhv?FKc5WiQi+)9`FLO4p==W0SUGy7WTXkG?y^Z~h*Fw`-%=zI784dkwS02~7k{dXm zH{t8(^D_zj zPv)$~LTl2{2JEUS%)ymuDl=(l8)tZxXB(Tf@pk0W2hhT&)3~3~mu7s8!@g-ldsvh7 zX_?Qop&;})Y4vT37TkKh;Xwd8@y?bK(m^QJxlzmILjSCjMnay1&zA2`y=lFge|F!Y9ri|Iv)0CUSzv`0r!risO zf^#zT{haE1(21kaTo(Ne`?9M2Jm&yc2zY0va~}okQ<;&${?^NYvzaw0LK~9Fw;H$; z@BBy}hst@!b^70v%O`(N#M9{Zb6SJsJX{CtIaA1c0`{CD#-Kg(z_P%=oYBe|OyE4R zZO)5$VQ>!X2{P#eWZ^94{968Xrgf%{p-bIIIsmQep6EQSBdk9W^wS>nk*H)9 zYlQYPb|V`vVMS5nhoA#jhb2jfAW~~{iUq$_UsQ)7GtEitt{q>w@P(^(` z=kZ)h{WgBLWyqw~BISb_88WH$^X6GHWJ2rJ@%Iy27scOawSGGOKCAUp@%I_6m&e~{ zw9e-_m8|#$`|Z85;=erk?#JqV#$K9hat1zj{GUV(o)ThpF8OUGKGLRqOpLu28^j`mJV+DFtR5}!Pfw9A& zizsxVoTq_R%r7>L&dr4%-c)?Qm$~mlXD0@6>^68L0-bcPLhod~&enRJKd$A_yzwoW zl1UZxjUExX8M?WJzN5wFJ`CNPZ0DVFW-q$RzrHfuwB=;xXZmw)z#h7hH82z1B&(&r z0=i_Z%8iMHXoGb*8Y)JIDMp7Wwsn$v+c)OLjmnAsWJB2JL-?_4og6Tye1Z-$)k?4x zRwT`l*H-uk(HnoRd_Tmx>&*|6p49_RBH&#%!Dat8+_XpeeDI3lA8`53zcOC(2efzA z*7a;1^Ck9YDn{Ew8Lg|7(LKUP6*t!Rt$cUe(YPjvzr}+V`MmIE{>jp}dhrF|Xge0p@y^LnF%l5FPX(4|>mE+;U>K6_&oe9oVA7RzW_nlQXa|Ivgyh zS5Cn0U_l}GyC4siM>@y};C>hM0(0Ms_Pu(?(v5%2`(5R8x%b8UE@2K>M>CV~{FG+|Q@(C@=A)SC@Vy znf-%Z!dX0D((yHIytwGCxtA2ZrGE4H+`xD8LaU3o=h}z9m& z+~)-PH0RlkoZr%SBfN!7h=v#gz9rY6u*djYriQ=RFml#HL^Ob`3$d4B=EMiZ$JKS% zj{k!>(RnF8X!>YW1EuuT&ed!#P*M zZlkWCw#&EX2>77CcC{CHaOeB#a|>;n@4v9WZ$;0GqPzLChzH?g$A8m~UvQf0sXuQ$ z^WCWzVXwOGE>^vn_)8TJgco0sJd!?`58Z^&`Lz}_cUf!d=(qW|)|w!nDkGhv;^N?2 zwfv@Wnzp#RLV8Ut5nu$+id2QLftb@Q3#~Ijwb@FH56 z-+ZD;zMHr$j$E#D{6!Jg&EQ%m7uE46F#o`NZul3#+dYT=*<0wNA7D6|*ETV%{sfcy z_x2_D1c&;Zq(0}6Uu^q!+;Lv@w~Q;7F@;!z<%uJDO)rweQbL<=05!uk9C) z&0mY*3DLd#=gQ|6!(SuvxoG_8yjPKUzaIH`w$B$I#uU2bc%>oj#li4e=}*>JC9Onzst(ZB^2| zRWon&t$En`;}cB-%xlD%H}@WbA==WsMV)!;X5KWOdgiTydDA!s7>CgTpefOty`I`O zaeCE^MaX>aut^*GE9N1+s(jJTfbY%M;GxIK`#e}_t?6sLYqL3D3qAUE;8$+0bh+!X z2c<8`=NpAK1n0i%KWd85XAgbWVQ)wegl`(N)_*EIcq6df#D6F)JivSmzy|~1Uu9O& zZYJkS3~O&DG9YI@YqNAk@V^P4?EvSq4AAZz^8N;bmxs-m=b`fDDriEc7J>9bhV)A8%+ywQWsntsq((^X~%vS>H&wRZPXS8MrhXD#n#E!Vr~ zb2coteeQ2f{~GaMVh`dQl{fM#-_{Z2^Vc1>y*J~hOt+p-kmbm>?xqo0{(JUVMvzZl zS-u`w-q(a|e%6lr-_E=+(SJAn zN8uCYhlsZo`_md=>&5Bjz6)%InCyw34qrIB-50+7I^-*Ld#Eeg&h5H;oe%> ztLH!J4@a33>0eRiMPVxT%_INcdIx#Xz`Q=heDnTe=w8u>i3MLktP8z%3--HJ zcR~2%Cy5<5fct9A<@a|*(O071p{?&ycO5hUt@J`S>lFuthRnV@WP#@W74~6h&t1nj z;;hh}bRV@L{YNpvjYog**2XQ?n?7v56CdYHI>*N!*<|7l-_Z1$M$Jmr|`8Ivu`SkE3ESun+c()s>+@wvbK?3AmOC(d z*I&VtZO(5SS>K1pvz`=M1z!nRk0?)PFYBTo`@i?d1IE6WACKILX|jiJL6%f5E#iz| zt3dt$`Qdz&tvGW30sL^)Td8Yd^XL2&o8PG?`c=I$>fxs;&~quzY0QcHtXORhHo(uR zU&@2L{kc{_@$IgiYvc33K7L}9 z&N30){B$h*nppWre7NT(@B@twA7rms4xiAb_BP3PiR?D_h-+O%FW}yii^AGZ=+bf@ zan?S5>kIk9qI09O(9R0%&u-d^&{l6g@gLghWgM?CjyjElG9O}R#C6`BprKit$8B3{ z598`i@jjVuMNeYC`e1W;%HdD+9S$3l*Nx(59; zl%$_N%e=mXjCzsz$8Wh2pE|tqynLBFe@65uohFa*LLUW{^R0z}AivGC7WN>6qR?|4 z=k9e^SOx!DhR=mL$O%~gTm}zyzZi^}d3#B7r1J&p4)jp&IZwGg@P@q>IPtWQty%#zBD!jV;4d0qQ*v5)G z?w)4neNnDG^T7uW%y~oi4QE!D@%yWjDcg(9hMotnPTS4?ldU|f?3_0qII!sp_Fct3 z#iN47&d#0+xTcQ})r zp-S{q&5?YEvlf{>GF|2CmKd6@s9l1*(O%#5)=Ri!ggHK-x{>Yp-zcwo8;L=P{>!VK zI=3&0<m;Y)`4n#yVwJ1$gkr0cD}JKOV14# zHh$QYJjao0=#BZ9CEjKZl^?G>jsfH(Yj3QZb;+gUYG@$nw-)wNXCrl>|L3)4pf^v` zC;pe`AH(kJMOG>v)1OZ}qFu%++d^Yjo~P~{);-5V{4Y^m@R*DzjkOy-lbs~LZ!L6P z#(Wmjul)3t?>=x~~2lpt&FYE>$`CPZD`8 zmE1hi9E{z+)e65eFBp58doSK)?C*h-qu@d~7-aqqVLu(l2Q!2m>LniE?Vm*36Tyw{ zkjNmWJDIrMIj#5uhFZxVGqFJID(SF;FIojd&sqhCpR)=Mv46a`nKLM^3bs6WO;O8{ zg;_1{7G}5H51x)Jvtmb<5|a(VkLTJt)id0Y{Y&`cA^77Uc_Nab&rBgk^>g2vAh>Ft zXdPP*ZYfvu&>IgNP#N*gGvuzQjPSc}mi3t<3$LS}>syYzIIiX0-Q!!1ETg|w^jAWE z75uYqzYDI4eAcwbxyz(_hP9A{PhbZc+whg?#=!H zI(T<_F!sd>*7MDC;a%j#v%X-gZNWnaCcpFa`*%F}i}$-09C?4(H+d)z*q)sdtZ94t z{sVlk7JORsBH->?a{RVTVvVG&65^1hj5W${KcszqYVOwYd=B?jG?QyG#P3h}>e`=v z;Qj+YzwMd#o5}k*;^>;XPy7welyTcOJT+l~*_ZLtzp)+(X21)QyYO2dys#U+tBm$U zTNjG1sZUU0TCUME&Iv0f^72 z3-FY%uI~ls=y2_?Kvx@~D`F6a@5))%NBAxkkJXOiJN)*aL7yD;e!p4gu%VM-OC>MB zb#gi4yJ5wmFT)>FNgfS4;?W%&$Ax$Be0RW(BWxm$@C46UmjvHxabgO+jT<*_#IKjL z!df#ceAi~<<70ev#FV8kl+1MJIj8k<=JyuX@bBGH6xXk3NCrZGg~VkO;8h*LyR)inQ6~SdJFx8sw!=K%0GvZS3%1$eyEiKqQo{49@TScJJP)(} z?Ey~F!ioMf-t|GaEPwJgb!I)R}M1cQk<;@{`*7JNAe)<{u@dnn-9_Yjm zKUJ{z^bqgOILQr;@T|Qa%6sVHzT+r)0P;t8WnmdKn=A{Z7rL@gygNuc0p@3jPxY%l zSZALe zmWLSsd-YcAJ>=!lrC)6M{j!@{j_#Y>^82@@v>dJH9k@A~zqsHIX#YKE<@W!Vxp#q& zvOM$tpP9)C0ul@sE!HFi35b@qs5rGPlMD(7Ep}zscIlRcAcvw$Dc#a3wgH3E8m)}@ zHvdAmh*4+S+g1zW5K*#Elk(znbhnS`jED?`xyzrXu=o@8bSqTTL(-h4j! z%*^xLhwHvh_jOHvtIaXW3 z$ET=UvJE>P`e__rq>etPJD=0|EE$OoVC@eqsW(^ygkOE?__3#7J^sdnZ_)11E^O6V z;V=Do|EI_E{OVVZfBN@to9D6*j}n&wJV!r0e_;3syuy5Abo!CTK&t05Telop7f798 z-kSBy1N$F&=9%N}xQ+}6gl+@=9uMQD55aW=8eIuYs$b$Af_S*kOIozHcxwam5^LiK zZAN`Vp<7YQ5C4vQKNVVQVp||{QRZ!(4@xkP>kbOncIiOBcA{4gqhGtwvu|zCJQrY| z^R*n2O;&p9}`xtohmd#9<7#+r%@=@mE7~_-=w^jR|*qVyzih#c}^`3Bm{U^=00d$J;tjK<; zrB0kOBQ5Vfc=iqY+vh!d&;1GNthBrhs#mRg<+KCtqu}1y$v%5M?Y6J>diyv}*-wDS zK@Kj3yJm1FpRbkn=k7;of8qTh#38z|h(j!LmRrwzS$EIbWi4%mcjmAz{UM*O%~}ji zs|uC(%=igkm!A;1x^@EbG}_;;#9qeVkP6^Wl*7_R% zv^3sH_eaQI_gfR+*XKW!-)Iu^fb1=eSNtnDHOC!x;575*e`s!)`)ALalV-6tba1Np zGucQ6H;$h&3eO4_V`D8G8xC^5v6k~I)y4)0Y&)ef!V6J&pei3eejqojbB%vX8Ra-L zFw@8V4$N0-uK&Z>{}Rjx)hBQ&r(!yOho<4Tg?C~zm%o05e5A)?dEhH#XesNaS@=Vj zV$aPovN^LdRWkEzBk#;Q*p&N?wN$#>eZTQb$gYbgXOC2NgTdQ*e>-?}-@_N;i3Dfh zrBAZ+L)!Atmme}V1-~L|h=_IimwBhZejxwoL;EZK^iRjPaMpW|mER!#XlMO08XKqN z+t#;4Q#Uq6>lmA+c3fYaHs+tVjhq(FWNfdajg^eK4V|RB8RHe`mhtei!w>(o61c?! z4(+9*)BBJPGP)aGlkwb<2Un*ihBGeomTCp6YXFf*Wjeh2MmmlSC5l^1gx!-(f$vorw)LoubK7z*`zLc)fpQ~>=Id_x3 z|7GTSa3vYKgLzl-%3YvuJYk)F;b!6(4t3c3!+M@e-juf7(=XI9?sCrK>~!{PZkAsc zzoC4D7Wh;CL-~9)N2I@F4&=NJw&EMs(p`L4`6#r!nP-ZTSjG3v%q4Z4qi7`_Yk`%! zFAgl++q!QLcf#FlvyIR6_gk@)>zr>U9>Ad*p}@hUN}#eKt?j0n-D> zXPpE0O~5@GxOcJ!ss`>ww5c-khu*|~_s+*XOJ^{K^l}=*6yWV4hC^de-5PwMH_4aH z7(9GW3_$-#R^ho%5btV4N&BU1Mc^USz>4)dL7{})YFMiQg z21ei|p7l;}NLxG9Rvz)JW^DM{&2QpT^*h3x^GWKeoZ5EFmGa#DcFH|UyPL6{M-P71 z)GOt8srqD%EE%MJ%6;EHA4-N3!OEt7UzOPlCxQk%7O zA)m>D{Dy+vmRZy1=jJzbU^lUMU~H%)-{phU8vJqews=o4>6v6g>p#sClAndlJL)R} z{bK0uNkQuoczV0$Ams^38lDti=&WTvYtA%1(K|l*J>~CHIh&`Ae&~I#^R2#?JBfR; z+wrx0mGAxPj3&;9qSNt^WG1q<)+-+pN2ET5C79AUi8)Ut9 zKkLY=9(elrB+o$Omy`cC2F_&D)E+eY^7VO#()FctJ~)_T>v_s!(+*QTwK<9QJ78`bB>1<|E|YT09EN|df`2L7wVN1rl^vq8@Ngx+X|K2JKzR5Y z@UY6(;CCGW54XU>>L<#%T-A*-=SP-+8+d{J>>ZB-%P3&MHeW0``iujIc%Jtu_k995 zMgfOxc*)Ti2Zr6qW_r0CV3-FCW-Us4K45qp7}S0ZzaKT{ddGynQ^&dt+<5G@$X2aI zV$fj@FrsI!YkSi6|B1f~KThlyzfJ7d5<;nP+<*+Pmfi_|;! zjByv#1aPs594W#@CH{@FsbzLd$8B~@95EdRKQ+QjeZZ{Ee$8IJz6&RcgL7q)d1h_x z%8N7S8HXpyH+Iw_FT-&9Bkqemnk!5{8ST%u+iy|(d^ha_SCsK-9?Q@TPP>}V`qd52 z`(8UwG?lIQPypFPo*Kd9MC%b({+fnb+gB;?2sC*Dn%oCXMne&9L z;7)Vd_t7ADTxM~`JFsKTQ}^s&@OPVzYwq3+pAmbubx))uEWIB^uH`fLu|8e^Pw&NV zU4vbzeKf@==a1t&eGqu2KxfLSAGf}lua~)doy1b32poEXcF zIq33Q^gJ@AJvYteW)4c@Lst&cpTAWXNZns)g|*MK3O}9TS3k_Bsm{gX0UzyQch}mq z$M17&@Hpm5)<|8fjdYGuGJCiV-$*Xymm{YhczC*%l75gbh|Z*3J@XE{6Xcm>y_)B- znYO>Fauhm6x&fKp@Rc)-yUMY}rjzexr1gAW7XPB3dil2mT5V!&5~-{7uVfyNE{9JU zdpTp*_ucSv1w6fm&vJNrq{Gvii@z)yQpUx-(SNif`yhrq&!s)JRmAU7+8FKBPlAUy zzZ3j!qrE6L!}Zr#kFZAEFF0%%@><+>A5%LwQg$Ns7T7kvzGJWBlhgQfyK1E)d8ajH zgiqNB(wU>kW4U_eMd5E*XB*x2q%YdgHEv&9u(dn5dt&~=F{c+4bq-@AY1-GywCqVIonTVx|Je^gtfA=kF0YYy19boHCqNa)RU`$)R9%CVcIV?kvy+6bvevo-;IBk!>w^PD8q#V6e!P;aTbH{kzv*vmJqJGvq!_l21fVqUVSUK}j z1@qFl0CfVXU1@tIS`wM{P`$k#`q3v@57lR`hoT#RWg~qx0mqZXWv=1Bf&UGh9jW7d z1m{LP$Y*eHBr zkK+@oCC1$gzGVl->TNnHFI=3psr;*((EkzIh|-4Esv{}OT59X}7WeHp>%B*FCuPJ3 z<;0a`z`=YeA6N|f&R=o1aTo7PKDmMWYMAC-{qRtn}Afb=%^+kJC=6Ifr;fcrrZ5 zxYt$%E-^WTnVZ(mWi9g*s)Wx&cf zU0E~t!|&mJ9`->KWkyp*F_n_9lDj?4>L&(FRlw8=oVCC?3b)+S>09!RGwIwzei@WDB@>5*c}x`f7(g>vy>q ztu^0zK0XQgTTZ?dGp@nm|FeQ|QBV58Lpx=(<6#U@+VHSu)Ht>7tbtYu+7@rz$yoG# z6ywQZT~%_ox4GnQU$bJta!TzwI>#|jb(Y(8=BN(eO`TE58ge#|`1#c8Um?GpUvxfzl-_DQk=8it--G3oh2I*9T%C+c;~=Nw>$d- z^|qfO8l-LPQmu!e=T^%a7~aUZ6+h>W&qI!_DqxH=u5yiQDzHMQ1b7?;@8#agIuHFN ze8!;1E+@ZZ4fjVz76($Ncz%-S8!W5ny(3>}ey`SgV;6c==hHgD<$xLJH`*RJ6?+-F z6wcz4whM3NlX{CkL4G9f=@%vuClSSWFpc=E2K<1NJcWyE^qUxoNxMB}&nQyIID^oB z3TJh+mW^B;e&JKrBk}`CrnP3!JC)HoLgnq;67rwRrcBdYz;5;>;Ki>iwtO~YoCB_) zN&GKF6VXF-v3=z>O#<+}wepH^jn>#J$Aq_>w0-B76(i3M9Yxb@T1KGB`Q{SM9aj$i z;JG8j{NC$NXMLBR&#iI@{8X$xM{rw#JTtEb3xQ#rWpr&%5BC9at~6S~cqYPA%vX_G z4{_t{tAl5I-Csv+N?=>pUEPf%*}VjNx7AU7sJ0ArFzbH#}=7v|xP~y~vtHzRI;GM=W)3QQZb& z(T!fu3BUS1N3UC-F?u~#!PyM-L2MkKj_#12kD&)*<&~-U_)6yf%2Zne{J18NQVy6! zmoSg;PWn^t%O!J^k<6)_WKLyA7X#ayz>S`YO`v_wpT{QhN$yR}13Bd5%tiO)Nw@S2 z13#BE_q-rIb1C!rWz9VY;G2@lRAdDIo8S@X8*QQu_&ieX4G(w5T!~+6m&x%xGW-+9 zhb~-O7WkCw12VdhdE&b0n}Jjed6xZjHDl0kGiUKkbC&dJ96lj`gu%s%8SHZ~t}>Gk z&pPIK4)Hx zoFLb+n zW%he`qkYRw{`t%;?Q3rK-D`S@>R2+5)DeLUyZ?D~IA(tD zU#~uAo=m^5Q|Hfhzg~d$q}%_Nwu3&}hIZB13cv8SLBZJAKaGZ1r|O**NMvZsSPf{~P-6gq|7a;faZK&%+<) zod*5U#~awz-*`X!5ytE0GqKMA_T@9_&1F*V^WR(!o#6i|705XHbt-t62JUYG_tT-p z3~2d9X!)fQ`Ma!d-8A&U^`nM9u--ScVZCDJBgAxfV$Vj|=Tkn@n`Mu{A5rF6_YA+u zpAzKR6!g+`^wLb^pgibruO~Mw^KRQ3ufL~`+z*`lZ%0+@wVM*^NBrg_u3zvu@cQXIRo6wv$X-g@=?3= zS!DRJ@4RGsK7^UXB-*mm$AAA$TOam3?6IJ|z8za}?tWi<1N!G7=23KHtPxzT1AjXA z&$`xK^J!hDJ+cnuxS-MscQV#1iDm7az^%d>}j_0vx^%@yUPf0`DUeSZ~0S zTBpV6L-xDY91k*%IDDO0V%Br4YX*lGeYbBt7t3Sq%X&_0Qq>F2;66jvHJUpPYQ4$4 zQ_K4FH1R0zT2AM0CUeF?HkDfjx?C4$jBQH-<}Q(wPI*1kyrkz>7Gi^8lO@21@}c{b zAAKV2V3R2)I_nhH+n;6a&AJ-Am(hplUJBi{UXL!bQa2J`Gxr3sa_kqbLUzpWcPmpx z+6zFJ?!JQcJ8R@>bME@$u;^&k$jbtzZf&>flK;-F+f$kHICZsdu8tc&+S-~VaHZ9a z&W6SneK@={4Tn*1*t;eMhao1L-5>KGC}WXBQYlug)C;PZh&s!{NCRd`JH_4@lF$L%?q_@;Dq`9T7ClUDtk(4BX%U#r|Uc7BM$X4WHX_XNnHg3gSgI|WBO zcs4u^u9HokaB>cxL2_R)_N07%oEb+)3{%;itozzHd(?K&^6y1PljE(_%fQ`H4vdV! zy3~4}n28-%ovGh{)n7lRIEuyO^Pkng-tuDhmgTS5&Au*tdbBK}{Uc9T4SCZ`JRaj` z)4Y?MV5N4Idc&VQMSfB2_6YqNKj3ochptP2qa}Z1JwlF}txr^QE*?D`2{ML>j0s(F zn`F12I&(ju4z|QF!RjH;`XXQ(yXgvG8_=b_^b*g2Uf2w-(I;kI1*Zg4GuhiISpEdM_+Q0m3PW!t?oYVfLwEsKysIOv=+Q3hIn`wVl zsog$%br)*?Qugi*E=xagPWuUTbPhJF$vby2;Fq2?{x#(hV6N)*uTf6(Ui?OA6?Eq> zpRt`?eeyf*Iqp}VKr89ZY<+SM^GaeoJP(fLyZSn@PSS~bCf}Q$YhIDAX)kA8q; zAM%;-LmDeF>00OG_uX3de1%!h$)`JV7P|b_bLWUFfJt+N>=extF5d3`h;e5ky@kN<;apzsK%Gg!zneS%yE4qNW z%Qw*PJ{vI1+J36s3ZHB$45$0blkl&6<=yLsglnKzcX1A%$OKKCHgI zGWAN8eU`$5okwgResRm3eRJQ>;JxB&JAoe<`-OkfB_9s{C&>++0e_*j{p5t)4A>{T zu$K=oJ}>dcDas5hw2nS<&zOv|x_?LG9L3qIob`_$Pt5f0=()|i!{8*V-N(|}HS_`} z^T=^6oY*luTE}`@hP?!SuAt6O)Ar|#i_ z*1(2w?@e7-NG_fMrs^o( zOXrxq>^(R!p7nNld#$bK;%lARRC$Z?vTMZCIp=NPZLhE10O}3QDF5HxzH(EnEOKIW zgE@Zbi9zUzNN}=0GylEqzyB@s@j3qc1qGqlfg*o=NkJ$PFY>n~i~Q|fMgHWGB7et; zB7f(*MgGHeDs!v<*Th14KNmOT+% zPu1oiPX!r#rujg8cB&Rxdu~w1yFt#o*qo`12inY;>?iiw>@$fb7a=>n@%Qz%Z|_3C zE#jEemeyat!v-oj>pSLE{2Z{=at1pA9IN@fBggZG&REaqT=N>fC;9$EzVDpJdAzmt z>yIq zw=XX6CzllXJC+srJC_&u4^PSoiG~~UpkeuB>P)5%co;vKI#a1Ll{!v%%V+Cp)b`v z`%*JM+0g}slC+s?d#%~2ZmtmnKYW&W8wNs$#^_M)v}*==9lPur1k zRjH%nt5U}%RHfnvriK!4P7SpkoEmCBG&PirPYrb>r-nKYe3@}prMiMuskh4i%K!HC zs?<9ZriG47oEC}}GoIl-^g$7@4YlRZH_X-tE2>h75kZp|A&Ojkfl+)ADG8mqy*imxdQRbKPR#Jv9M2rvLQ5 z%bdQu8Y@#>`OL}nGeQYs@&hOHuRjG}W#e|7Q~v0NN`r@`w13p$v!$9(;e{x?5c@NE zp}QV_<$eGohYk;9^I*e}bNTzflLuYCyNG@@_eM5Y#L!sCpf?oX!W!X8))LQJDarjm z<|6kSp!G)V(fTUh!!u8^hIp2}`85Cfv7jU06Xjpcl5g&*jx{0Q$a$n3`L4~B@9FON zk(=%f7m#o57r65Ms4d^WIfZ=JdsET!Aon*~&nxD13UNTmT^ENt$W1kt_4hyzd3nga zlHK>!{q!Ay-bbPLv5Ck(@}5B6+q#hdBgp>=~~E;{)5Q>Oyoa^{7*st z(P1%kS%&-vZTYu-6m{EjfL-^L7#&s-OigMz+sNFty}iCldxD`PxtXKPYteCD&I1fP zMl8!Li#pnubH^{;b}wU(!lUiz1GOdpfp08#VAOf>(RoubaN-3gQ^3hoa8giN93JrA zp_gv~C*#1$RsG;3Nj#a+w`uhF@5G56=P#Q|_~0xM_|Tc6GsIbWE=M0WW$MG74tz26 zi+rCQob!v;K`ZJCzY)ss`)73G5_DpGkUzE@op@r9Kk@D$e_Qt;e|yg$e-d5Uk&jM9 zUrPTOop?fe5dKDIb-!rq#NQRmPP{n$5IXi$@wqxN*WsUDomh@eT!&7iY`R|Rc3_T{ zqfgd(Qo9CO+oRYiu_w`qv(bq!q7!$b6I;=Vd(nyL%Ffr&iReqC6KgYdqD!Oh+xpRo z^q=1MADq6s);T({X+~({dp(WS*iNUA#ccexJLQj}Lxl&U6CGYKIuZGZA|J6oQ$D&| z({!RE7uj-ovjam<8V_o{^34}*-}Ta+%>VyIIF2Hx$BETD2a|t=5eP3 zcVH*cXUY0OVYOex{M79mWcuq4+MlNU8c(W&@&!-iggfdf|4>@}2EJ3LgSBJifH%~R z-NyPl73;<>nS#%z2b*NNCpBL7d)9bA*Uxy3UjrLOpEcMi#x_{vO?_zqb^>uHQR1f* zdlhe*97=4N9BO-Va;W{;$)V(PlS3UZOb&G}o8~`^|EX(YW$LZgY5un-Rg%-@rqGf6 zn?m+_YKXDrJG&}V6_xA(53fv(;S5mZKvgJ8zTH?izR(`#k?F*4`G~#AXP=EdVx9l# z7|vL`uuJi;MV8~2VqT6GdQ$JM)qZVNYWKGkCvBgJb@x`I>l{9Qs4Dfw&*7(c!3*(g znky^woqbmBO^Gz1m*2%sBfciSoY*nqe%cz5gAG-w2@|YGVkOLF#q?RnKjq5U4^b>m zDeI0}a+h5wuLbcwChpooJ`%;1se6gdvgcf_!;p=-O0$MJSvxS3w}{)5&gg#9_U-k& z2wzd3cAobz{+qE&_UtMS@5NRtz%J-m$dnd|YLPJ6arER&DR()f73 z!^iUt9|I%s#h_6f8YQ4n8#Mar*`CIu>#+HUcv8n2$tAS{ytaa$z3|^YcyB^}C^0cV z)OKTjsC`m?C>hKTbxg?*b*{U}fA~%Og$F&Ux8fK1-;R4y@2txY9cj!DodVvIzKb$& zFw}voy#Zg{?40ld!PcD@7Hr+c`TS-*L!Iog`rPrDvhwq(ET4<>P2J=P>Vfa<{!Y0# zH|5xuIzb)qul|z4|B%X5bVX%K=MA&_T<`(-{jJ+Bdg8qJh3^yaeH(nA`X9pY4hOD} zir**Q@nqw7f{owq=?-oB!ST=Bx~44nHD&w9a|Hg1!do%;zDjZjF9wi1(V_I?$R+K+ zkIjnCdVOx(wrjaRU3>bk&rMdPM%;R)ab+B%Qf2MACR^1-E?&qDlX6$ii?8$%~598?2?gkq_J#TW)ZkByA%*6?1 zgp>H2{qtGHUl<<)Jp=B)1>5!(Y}>Ad-#C4kJ&ZbH3L^Ls3b1VpjcuFPGIl<8Ejnp5 zw&=~qrj;KduVrMqUE72&D9ta_UZ3kvu3--j|Ij6#H2b!R@v$Cq{X^~bH&N~;|J3*T ze8Ii;Zfk+DcRTTCo@^bMY46TKSJvl-VugkNctdU|F{04lR#ND1FE8{bD+>J`;|l$q z;|u+VH{^y!kh|to?EvOKt0jHSd!2*#qqVGIM--Yh?1}NvbOJWOMC_wDwqFw4uM6An z2)5q|Y`=H0{Z3%}1+o3G5#9>kf?o&QFNp0o<)+Zld@FT~`$5fMR-R0i4kH`#Amfe z|NFYCp;a84TKm<}ao8ml9wU324;sH@=Yoh-(AT)iln=Yy@F+CUXAN{PwrcA@W2?q) ztlV*rm$Zg-{K3A(P}+@G^!?U3>g@hiPO%1hu6{%GtG2v6)4PD;Q4US!9FjHB z^uNLBe`99nYPxr zZ4J$6%j}hxTOs0J{hUb&C1zXxwmFu+omkZ50?XgAn0(Po$Qym56%sGc7cw|| z*5-}YU%?v((Dzxqu{4`EuqDMCr<&j+%BS=Ae7CGA!~QgWAC(mjT^k|_4@H{#_+|j~ zi;$;0(o~t+g)cRB+vHHZ75@{yw>iq=3w(W0jZTf?EnFVgmsdi6)WjqhgcpOf`j_}(deV{2i<$IG$bE3n_k zVZV>Ze#f5gz@G11hW&m3`yCtptyb*!L)h>2*zardLq>+Hx1rNH_x#c^xe9EH?9B0RVlO=c95;wpPMpgt z?ZonB>Jf)mTHzJSr}N4|w`^Qx;FafWUa{~w$7&~s<~(kheCO&<{XR>dySjUxk96G} ziuUd!+VsWmwHSa$Z9e3 z>2T(qT4Z%WfOZ13ga3RP_hKv$&+<4GtY`%GuF&o87Pa z7aU7jFCSZi-)CQ@fB6XUKjMXx`!ZuKI-Tcjdk6S;vd65uLbK~eoHF9`Z=R@1T?U*- z*(=_$dOj5{(Z2u%3urv!mH2!4X8w%dZh{>9KK;LC>P zvktB_&ZD>EC*Q}soi&b6J7to$SLvSB?b+|QJFs@YIMl3V?0yDh(4^F9qdi#=Itu=e zcJaA{{hX%BJKT5;#n*hnnb&G;U(xH;&)Pl_`9q@MIC8+NvwK^0ckiJ)v6E;+XMCaW zR>kVBB39<;0orBUqV0ExbI|WkVbkl`r_pCc#E~5ZCq=~J4JDQ=n|~f~#&dL8RjRx8 zGUGQ}ikRqg`;4k$Nqi>f&0+jTx<5g&PjB<7`v#8i z&eqMB8GW53?qFGIAk||PcioTdCg*rlx^Kex*0Ol}4hObK8f+cF)Co*~DR@q7nY`m0 zc~Kg82Zo1U}HhUEB)DL2A8WcJOoG0VM zGhqFm14D=7TUz8}?Nxj+eq7d;<1P*_;`0=Bvd4I*3%e;x9aEOi>2}@DnaF5-*g5@u z-K}TJkym0SynAq{-QS6!roRXjDaD?RzJ~7-S%@P83H*?4v#U}!Tx&hj&ig5xOGpx9 zN`9>_WNDUUwJ68qUx8lM_Yj{^_FQA}FZ<){xhB|iZF~32{`T%K`;$Fi_ILRH%HNs) zSN_BK(?Sv8JdCdio96kOb+_fpp~U-=iwAud-3Q(mcx4x_or|sb^7J!}kLECEFTh{A z?_BtbfX{_ZaNG_YYCnozyO(`4?$|!JkG=F!+tLT%mCLdl+K zp$^|Iq0anULWhfI`McJzXWdYhdb@U(|DAPJsUx@D5;}VOEumx3=NNnS$Dq-%3BJ^q zfg@87qbKb1p*X3_<4jqCaIjO#c5myfGI zS~NN^OK0{+i*{&{gccppq7zyih8A7W;w@lCdb?rwfA9}O7KeF#UUPNtk|5t=FJ{EU@JyXD*0&UG^4IbP@etV;j zeJwe!b~m=(&prn4s+liq>~lS96>I)D=Ubj4req{KX&&b#6t_N?xv{kO%za?n-80TK z%7)T8pt;PkQQFb``g7`)Qcu5?lT7D;xWCn2=hoZj0=4GT`r*h~d;Rcxe}4AR4|#SS>x%Hn zG1xvc!QpM-@OFGbjC~$^7puQkywFFLDx`zGofjOx#YE-`37*P)Gz>ll0> zTnE9mzGu@Z!kD9s`OPnISC2Vkr#sp0_*UgQ@#@SDaLBjaKA!XK;3fV!cq{;qO%9}X z2o8J(!tb&(qf5|GvA?tyjeyr#{l_x$gqJv$Ec$mrMUX z&6{~wq461Th9brqMSS#k4j)C4iP(@p>Mn<$jC?S+WS?jLCh><=tXl)nUp7({|Doy6 z;ZLGR!3IxVX|-9e^4Ue4;nm4Qz>u%EBJPIDeR^S*=K z?8%KHH(oCJgLnRlH6in|Vw8>DpGQuj&)Ii81#fij{%m#R4&F86bbljO8~R+Db@0PK_GSU56~It z0F$4d@;ck8^P=;xnG?t~I_h=hV6koAoR-nL6MGVM;bZn)y6ys3%4sqXIgm~ z$>l7K^yGZ>9r;X%-BbV|8WuPAaLbz`gYDNUKa<4lK)TBx4xItw~s#Uu{bgA zt&V(4F8Ejcn$0sG{Je7}c^I|N4z4QLXYYj{J!g2+p@Hs5&{-kr-yY88Jxv~oB=qZm z7CsN~zd>ZTZRfq4#uPL>J%4I3H!b9s?&jPa2s5UE$TX3D8G8M{Ksv$M07D zNy_NH7yPKRR8Ia;eacts((6E1ccbc#hL7`$@*3K}*GuG^%%InFn_eBnh_v(F&}-ox zrwzRpUU^#glcd9=JEX*CiUkyZ7(Qbx()sxFZ5|UGqARevctVF?zY4hPfO#Z3B*vOB z9oAn6R+lHhspLPuD96CO+JX5FcsLJ(FSrQHQda9F8?<*#~T8v zngXl2gxn7`MH|sfR6Q{)~Ot`U&IzoAv>wQQXLy}Qtgd7 zskU`Fsl-D$sdz(9Dz+vE`z?okzmkkKSs{8*YqB2hu!$hAJ?zy#%BOp-OSy(k-dbn; zlaSx&rNsYkn4CI9*_GsBZ(BE+{wAkn&)L2mauhe=e<}B+4&xJWaTmGOdc+!v&fYjV z^(gj-#$8S9lf~Xjl$fhBY<1n+`vumndRNOk-P^km+e~YS$VT~d0^3|0{Abh7G>*gv ze2?$ZojCh@bHA|m#wH_!k7KX2OpRsMGDR_4*Id%$b^L434z(C3gZm8_qf zO2+wI!)FqIZI>@qsx^Gzs&KOP3bW5_>g^gBZf|w!?RDzybLzcDy#wItO`pcSHMz1N zoP3erJNfrrK|7z|96H}?feF5d!56yg@@Z%quf?~r0N?8YV!fJzsU&AfI?5G8ZRf%U zAHzl{7S8r9V8h^Bpp7_fxaTJijJF=i=Hb6ZhQ!0}Tph$;Q)?AtH_7kj?!%QB*}PEu zFZR=Z)kkRmD5s445_g=aGJc`?@7jLv zHvHcAZoPMXJMGrtKdRPziM*M!Nll`KeJ6$Nhl`cp3VcVwx8m$i0b@2i-*@_Q=c`lb ziEMbj?Uupz&Za@iYtz8CTWp(2_n#(!!?vAl{M)(oGHGHlCzc@_n(H1!b~~5gcYM>E z>KFmf;ahCqz_=J&q|uie$UJ!Xpf^>>T-dc7za#YJEa))Xmt4~FzxjSxdVqU#@4D8S zn18LceG|CpLT2yu1a{noKDY*bu#r1rE6?`Kx|2I}H(ugd8tv|x_12+U|5M#P*J2M2 zQ~Ccy_qwpSJjL0)uHCdJo-Dt_O1;>3HvPH;yo&q``#G@&`mWq7R}ST@JNCt>q0G~a zAz8wBdg-INprGZK|Bh?`a~--%=c_EvR(a4>{pqku`hG(9uA{@E=z=z2mc1jo8ovf~ zUJMLv=;xiB6@8R4>F@rEcM)KV0{<(>v&)?_ zAI!6yp>}vSvJTo7L+>G(JbMSc7Foxbis9KIF3)Dd^Saa52l4D)w@g1gJHvVR=kRR& z&8k#6KHO|MnAlI7i6ucN#=(|>bb9nB4{yOE_k8!x@7@8ANAO$fyzUw7m`Cvy#Nm%S zv5_N<;0Bti?au>lqkbNezl zxtROXBR7FQ|fB{ zkw-p@O6H?7=A-X^1o+;**6x{Rr^A^daE;+J6DQY5l-A<|DxOj~@cQ zYyS)ImGuK(jB`+J-PktF(Mk01Xk<WOmMyZ{Eo;h9 zmOZe!#NhEJ!Ho@>*BRq7uP1wT^GAHiCGud`BKp? z_$q1h6zgLA`tG{;6X7l2Vmx*5Ul+O_I>%{WFq*s+z&O7o&^&LL)ja=l*2gad&H6Zw z&EY%Tb6uRh&DPGIYmKd#b&jAD`4C^;j10JZ{2Tb#`Va8Ucr`YSMf-_*|M_ji4#_Uf z;_o+|cFdWBEPj67c~||PJ=*?a=VrdEnxN zP20PG2mjBG=oBCS!5y&^LGB2)c36M-U>e2@4awL3A!sYfFpyIyWcae+j>&=I;daxYn)sgNEsS)Cd1#D@AHqU!ya$Cd9YB*7a?F{k!C} zJy>OY)799k>PPZw?AhhP)Q`VFEGO@=WleuE`skz&t+myDjP^UR1>|#1;8%&FhqdOE z?b3#Q(2i}8l#L*LC7&Yw?_~e5(<-=sEV0L9i9P<^!bPWlYYk|aL+ml*nx*&?&W&%6 zEQTMjhX%|(!+9sxlZ%6?e}QMpa|0=2qJ~G-;CI4D7JCT0U>$Y>KESpO__j7$%mLW^ zTX>F7=akNV{t~oaTM|fq{JX zWNN(`_W7nB_K1+VleL+1#0KX%dlBIU_)JbwFT2bTht_7T7PS4`f_n>a|B-Q8{JHzB z_Wjdo_r;#DLMET}L@N~>!r02mc~{K%Dy-B^!x>|VmAbi*aSkVkEN~}BupdIMJ(|Qn z7ufdB=(CLh-CMBI3O`%x-YIdh$#01+dtG~eG3{xN0C(lIUqRaw)vldu{7dh%zrnjQ z>@B_5eyhqT@1n|$8Ujrydu6eeii{&?7Cvt6%_?`6@;xdirgG(bu+@}@O!>#qlgL;_ zxB+-P483e+QJ7EJim$tS3G%;-2d_~7d{_Ub?Gu5Jr_>64`#NXOj(yWzzOuX#er1D;UFo1ReTb3Srx5K~!4Ore9=lJC-8kI zeA3E)Ewmxe&7atxE5RS>@~j8ndnG(;%SxV2+hUuh2f6PcQXB|9jm#K-ZN1O-*P8g> z+zfp7JP95S_(IVW^mT;2pq(Cn5WK~wAV1(MF$wtre{IW~*ZIvbaO5$x1zD@>;SVw1xhBx#5`lrCf~uv>#;l2A9$Aa{4?%e<$ecKrl4V zrDM>c{(T6t_*QMcZNS+^7$j? zi|qaNTb17jT8dYGd8NZE^?kDWW!lw#clzG?sH({G229bMBCOo|Y5W zzt>pBd4lx#55ZXdIoc2O!MKSrJq2v+Nv3Lu2N7&@C(|}EvYtCxu0`)FuWW2)Ca!Aw z!PTZLT>YdUTwU@(xLRT3YHc>Ic%F@`$V|qPhAZ*KnRW2J&yF=we(6ly5u3w!s>;a2 zOguvldG|KH#oC8GtBns>OD#U7`+m*-Ow|vEkPGYw+r9&@RC0zux*WTYbqRaVms-b& zH=MQea{GSiR_-uOaF6%uPd(tn`MtQ3-;>zSx{LK&4B7^PQFn+BI+4$f9<+9KP&EHGyHs-eQn9wW3(e1 z3Ar^fS=xIxa~(R9yWUges(T||BGPzc{Zq;m(f@4 zGY{-9pYe?8Yn9X2yg{}Msc+qBJ(Bigw=*7t2RrU(EIc5+R>xQpx3jP9=wCOs@gE#I z-oe%JhD2GXTprN45(0cix!;Ly|HDBDo*Qu3+1)C8yeTLAGaF3Y*g$Cx;X)F#j)dR zpu23NYG|n3>XXp<-_ZUNIFnx6jjk)3bxk;yU$<@3XOYEVz?6wUgkFRPW5U}K>@fTW z_2dD$@={{KeYyMgp!4_S)IGe%vKowTq`f@!zV3vd4o!0D@4inyxPSB0PaSW6`qkt2 zeJP1NknLA^az{{87=_=#|0!@uL^cOKk--zTLH^A?xs89L)W zWLv-C^I2`kTsyM$QmM!PD72V#gFTOG?DHA>p0}Sme&5rt9N&D?J7(;g>^636zWT@e zH$P*y>(R4I1EKjp{;_FSylIayuuW%ZC(R;WqxV@!%OIXb6?gHetHO_w^ZHuG9fxLh z;Kle!vgn4qT;B`ltzQ3kfV0vW-{!vuj;Ee54OJ?_sXkeA6?>H$U+Env zuHG)cm}2GzXGfIG@ur@2M@`>e?sgD?EbDsagdbZ+CCkq&KJow#)PM|~`I zct`hzx%9Kow~RRbUVX`F`Ld$e5|QziX=f(w;NQfC7zHhIy6AV-Do3`}{|IE=z%v57 z0s01vUo$`a*TScBKdhOzFU>RlP5Dvf7sYn#!!dN1j3mf^ALqNlF?I+!fRC0z8#Avn zE`!rxmalUN<0`htc5Rwm#D)ZT&bfC*!rG~PueTD4v+L^cxBIEHi+mvG)FHOp^kZEUUeHgSQpK?DMNU|+d9Q|#(KS)XFy*xiF4;V%lx}!QK@3?p;4ne++e2po{N#Gmz5f{Db!Af&*r)y0u{UC1LF!bx^kB zN$mJ%nUkIar!VkXo*Qn%E|1^~{o@Jz)6C=Ar#Ahw{&4$`JcLj1a^@4WHni`GeU<$* z<0C^q*v~&7{;i%Z*P7~L?i?y^?rCBThfl3)R}nFo(60gey^;PlAj`GBaC{@5z!j%l zqCO`qUv&gM`=SHOCgS+m)7fgR%nyH-JMlcmKR3wTLovD~s^P+UQ4i z>HmJ0bvo~~PSJb$8dS#Y=}^w>=}=aCI+AIvH+BD69Q>PmV_8S#GKUwBM+Rx<3EDh; zk})ozza;_oAn0?MqqBUr5C16d^-g871OI7d`)ezkr>cpGuW|Xe z^wZ%t@FQjOZ-Dzr-Fo0=R&1b+|2>qgR_>1Rz_bRwa&<dwbbCGIxC2I_(Bqhf^+F zhpV1+_%h*@@4~BMpEM>{k4G6>gt0x7hNu4Ukj-P7>$GnB75gMso^`$6n|t71!)}l) z%eJUTr@`ZMP3|lE9ys%x+*dzle7jg%1f0BAn~3Z0&3iSDG1oI^QaAk`I8&G0S9MNZ ze!pHFH|1XMy$8;zn@+do7oywVOu6nVd9QILzNz#!=BMGOKiaFTbllgRzT7h6t8{*n z{8Y2&Irrs&KZjP`8YAG7Y+FWp!+fY|vW>)06h z3BGuK7jb^W`wjFxi}$EAI_&b^ zkPE=%;x_jJxRos%ThRyeWx!m?dZu^&Upv;8mGWL$>4SHL4eu`mFW)~1-Z=X1gLF!? z7Ma2xG=7bT0;wax^T3LJjB?1+1wxNwTW-0oMH>64q>iD-%P`{ z>;$K`6cR5!-bziUt*!~ucibV%{eL>|)IJgToBF@!ed(9Xy|Ju|z%8*VWh*Z=_QR!; zA@r;4vlWUpZ8dZD+PTP{e%J0pW|8GjBg;mPmsF;1LXICrZri~16Zi)s6YSseNk*pw zgZe1}~fxY2i`$6Yhjd(Q)S| zGVO1*9i#2HY1^#9%RS+Z-l1J(X>0K8`=kf-vp#%dO;2Mhy4&@48u*>H;6)EQ<4Ic^ zy1s=teMf*T3alOdjPF=$u8BA3yhZ!P`I+$sC*c1T9{+vcBeovh7+sEzJYv~=ORV+^ z(ZAOIykkMo%(?UTA%9OA-q^a}HDV)XTd6%?J!|fFRQb6}i1nInpWlOC#(%D}PI^~O ztW@;LK7C}jbeOcxIDIW}EkRzo;Mpk}o7%ya zSthx#CaT^rw@uD@sr~}u(e$~Hc)K`d6~n8&5%(@no#$u)Kj&-ran=K>KaZG8ty$pb zS&of9T;sTL}9OPjn@?gq9FYRaP z``iW2T>-5&?!NwOcxH>0`kFqQz#HR?fX_RH&u7uy%6~HloIYpiF1VxfR@`JhYwK=; zR`wp8?k@Tox&2F7FR9Gj#ZH+el$itk&*87)eFVBjp2e=gX2RxQTNaoUj_yQGTj@8R z8;-(*QLUHY!;gYTXYLFZGxWARz8q7h_}eb zmCBVC+BmC$SFvAAy|Q9QCg9C-^s3q^N0$n>^N2efg>HO;_SCjw9`yVU_(A(<`Y!y( zz_Izh#HL*}afJF^bG!XnJDa#d_)O2Lm+>7wG~YSXq|bRPcm_|JXTm>5ZL&_k5WCHm zU1V_9WaTpFw_-xr?_cZg_kRyQ(Y6$2VBqL+))zB(2oF^xZa-vW%^bvF*WMseGAT$h78=5^@y;(7Vx@&=|ku zB<8Y3W$d$TV2>NW`JuG^KCQ{|KlbKYkRM{lUEPg3mo1&8I%+3Me~RslF9vR4s$W=p zx;+W5goirW5x$mCpcu^aIHB{3GUdO;4f27om-)K(GXM1;xH|-` zZ9k&P`>+{(^&cnf_=Cq~bAW?Kncv-h-1d$B)!rs>$W|W)N1&V1F3c_5Dg<&nI7j z@+#z#uOJx z=#c9Yj6=HotBgtCr62iie1O_dX=4m+jG-M~Y6t!#JeEAq7$$$nF?{C&V~DYTQ)Tul z$bAy;ndMtJ_l9{no;OzOyft?cjv>aWmUXyq;m8~Q#CT)C9_7D-{}O!v-#F^n<<5L6 z8*mk8!XMY3T-u*&pMTi>oj+W4?*7hn+Wjc|JI8!>eCdzOeTA9vrJD15&{xVg)N?zZ zb%n^nfM$(pgNduYQ)ecwxA!Rm+a@!wuU`4_@a($->*~*T|FF*IU93HJ?d@lm*>HA_ zdC&TOR{hx#Kioj9yvjzOgx2(XFZNS__AJ^Pc?r3cumvouD6F}|eIBW&Y@9eyVALED z0nVA6Nu5o-YUbW*Y&)N07ivCf;&<#xcoq4T%`gw-Habi8u*Vvj}*PzFkUv8+gyh{}q;DcB6-1tkTqoZUK_S!+dmcv`o@xT$8F4?!b zbPtMP>LlOKDB9VW<5^nHT)ACX9Ih*7I_XLlgN9l5I?Px z59qTKvgWX0-+Zu$wO<=DbDlYDtGkD$IjrE$zBw#P8>-jSQv?qTZRRA7ZR>$&GdG|Y zbynPr1$|!g+jQNC4%|4@){WFlG~UYZFQYRD@i~}J@Vgw{cq0G$7`i7mMe%d;{VEoT z_^tnB?ppyK(H9Bc8yz$S+r=sKFmfJWrZ#|a1^@7B8}Hj@^Erol%)1F}8JE66AND17 zk4s+qNVAUetRwOUL*5XSKb+LP?eKLULmzEYop zLogUS%;)eHG{ya%Jy33TG=9(BXOE#XIrcKacyX`4MX|=(FG!T<_@mHPal=N>blZ99Ixpi+JY)ZthQ;Ty8?6^2-2uPN7e2GL{xD_QZznd%@`T%IcNcbz`Yh2p zoxOjJT{64*@<8)1MpZVqTwT@t)U`qEro85kiu~ro;|rSKnmC|&mGYY12+bxj_fO!z z9C#f(O+()#%8l({^uQ{{6Qhiai|h6AEvDKsnO$x zzAoM+qr%(u;7#9Un<&0Yc$1Byy=XHJeF&W0@eVp53y-c{-|ORSg1@m9zSGes|FI#A zzZkfi0)Fg{URjgOx%$A~ukFicDgV@`obv>;k*jQ({sQgmyg<54Z%Xq=9e4V1+aF+L z`LbSFUUf129dzgAOW|2NPojOEuB3(hS4Qti58Xk&aMm{0%homa!mdlh_ge8^Jz@D8 z6knV19KEFH_d0qB8{5j|JJ^K+W)_9Gw0Wgcjls%9}CBVfz+>F>TcBhul0%a zZhHT@=t1t|I=YAc_gZbgN)l_L{xhD_zn<@P`rqq0&E6FEqQkFo^oxN5zpi4c9=sG? zv5&hH&=(1KJiVPw%8$D@;IHPhuTDCS0^7F#g>cj#etv!){Cx8-jGqOXg%7g7)fZuO3)fkL-ho_xaA&-@ghz6kM~hMdY*7d~guC zSVg&Z;!^aP;8T3&%CTg34d<)fx%*S+lh==FZp@6&@gh^PiOgdecueP+l9n<3@Js@^ zx(U910>0jCC4RNX$}xPM@f^O^^UbbYy=nRP^dnbAF5ip`_Uc*b)c8bfRdl4*`+o_! zS|h$Wh+LIq^31P>!!rj1JkOM;8p+d}0Y5o$&nr(+CkI%%%xB=>L*k}Cy!`Ip9DUOd zUYgE>7ui`s=u!(^?!f+f0((pPCgZu$cix6O9J*X7yc_~A%u#)MBe&&S!pkA>!W>nJ z97Ul?44Q~PWdjX#WXQ-$(3KxYcAVeHi&>)##>b^Kht3NqHdV5-GhxSp))x30Ruc!R zb#8r*uOUI~r+m)ikehf~9H{N9^R*=5Z!d?&!?JsW?Bjjm!mCdI4x4Zv zHs@7?ZJR>=rxxt&a$=*#VMkmwfOpst;|5r%-{Di6?)l!sGdy((hri^x`{9|M zr4P^X1dboh-?;u(&x&nz<>iZA2?d<*j{cVSOzu&#B$N7HM zCukr0XUYw}yJwD(h#+p!J)i=&G=(M9dd zd)nK>Z!@cED85S{dTIo@ADI&^V|QY&Lywc&Ioaq%HcFaadc*VQqVGu@9tcqdFJglTh7MyCtq3nhf$NGQ0vr^mfByEI}AQ{ z=UlB>?K##xYpQjTy^dub+g88*+`UEh|9hwZ?De0{xBk|wPy6;uqu2+s`D5@BzM9u- z>;Hf5-UQC->b(E|-np}}IVc#6n9Ry1Zc!E|CYd`kC@Q8Th14b{zuAC+;;)IdX<{_a zz(8C`!d$hTM$$5Xpfj3;5QwEo%OLI;gP0~JY0F(^0ntQ=Z5Tl3_x_ykIdi}FGAxqz z_wVn|>owf%e9w8#bDsTq&N-Z|&g+7E{MUu6BK`N_BjGo3K5t};;z#3EzjF=uL-M+y z!hc;@%^4ifOB|k0fcLw2#=su7G7o-|5A9^pBNj&a(UL{41#lI#pI2dMX7j*#WW#ipW=qhC0A?i2Xbj$GMY|cvcV|!My&Gw}5fC&AZ!245+EBZkZl;%goFk{?NMijC5 zT4M24Jo2swDw@M7oY|N|6Zl@oyebV&!?(r3srXQ`UA$oMYuoxb@j91-CpV76Q#Y_z zN#|KSDVEb>@stIgYAl|p$29MSr)=<)>PPjZcW>D{1fI^xz|%ikJURV#ANxjw_FW9R z*LnME?c9Vb$r$ZHx?VJeyj{z8mEcYEwi6tQb`};^_3yaCI|N^twqt19jqG)HqWV?4 zy78Rjl6Uu(Ph3ij9T_FtK>lnu@>udM2Cpc_uJMFRy-v`+B>LWfT)*}T&Yosp)kgMJ zZOp#=0S_OjnC~jk{n+jEjZQ`fIC(Gm@BwVa1UlD@gY47#nD{5JzCUQ!%Gw_*7uQ|A zh`)_q?)^E7i?2ooQ-7Xya1OC_qtA~ubqqM$g}F>%P0`g+%^$rve-ZmUMccKsSDi)! zYt_yAd_Qx2uoGOZkmS^c?qh@Y_@FNh+&$ptA-isp1^VaK?*RLf-z`YDPwLQnTDKzB zMr)+?(`TJ^YjSk6*;jd_>D2`5?C_*{y^8IaoWok67lfA^Y2U7Awa)B`5T9JhcfzT7 zR57D=_E>pe7#&?l{kbnseEUV?DMwkuvn|j{!)Tq!oav9soi4wn5y$v3p?+~Xm8GE=d7t$ zsV{P{YCdYKb1#FumJP41u)LNHuW??fkxAL`n%<{)&7ZUQOgX{l=_hCLL_-@**Z{Is zMK=?{>!TOVdKEqU!1L@G%nzLx&0ErVbjFL%KZ-$D z2e3~P=$4cmEbQE=IxX%?y>XE}cg5ny$Vi^MmASh2OL#9lX3ESF8T6&S&4!o)^whzb z+QIrE=@&;&G5AZg&@Mi+deYDlGBrp?m7*i`o?;Ec`}>&Re9l(ZT49Uc=*tH664o{S;4{5s?-qfdSEXj?V@ zv*K;+1=`bad~6h(%EY-;lP$Z^#o)xJl3kyqj!Ew)U^{cJvcZAoQi?6u(0m@Uz{f#Q z|C+S}j-#r9Lf6K5|9jpKwT|Cn|FiSkB`2b+qmu{snf*|y_*cA|J@3TV3Xgjog0&vo zemSDZg>wVbE^DJy_}^bE|3$VnYhENfn)B1|0xvOaZrL}I*|T)U z!-Jd$Hzu3ES^S0htM2GQ#%~zypP#;gGt(QW$I&_1RlHY>b2)fj4vrjb>wzr?xJrR* zIWVdguH_`~nKkmJM?VcLmA3wM9rxlC|AVc61vb@ztM5AIjU1dVTTE-T{~$cZd(B_< z^`Zmn{>bZ$vG;#K-|0eW}s6~5lp`pKtsdRDfZbnO@EPyDDk z#?XJ@VWu0i&e#!mZ@D#NO!2uHi!o0&mSO~MERA!^Nyf>coyR!J+a!>mpUfCbu=sT6 z&o@nOW$4IXv6nnZ`zGFOb>xP?8aVvlVEp$~{L0xk!lm@0Y$)k5&DZ~2*oh7wtKX&X zmSYR9<^3|NLv?Qhw(MA~foreNY7d=`mF=YcBG?St=ezybt9~3&cv^(*qt`~(Y=GzY z!v~4Y_deW58~N_)PyI+A^vw3Q8=iy5+&LIGBi9w9Y@&^L&g^4G_KT0mDQ;JPJ=Na! z1pN&cOpLM4_G!jquA(!I;eRgddU(E9 znKiR^P?Lx6f&b6m&b>!j2g&whS`YMGp|xx4h~;o!z%-I+9>29)}5U z0X+3w1#W4he2H5}@rvwCjf*@y6k{%3jB9K^3x5v!eI@Y!lJ9eXe}~|wt}6%l^}dex zqI22dqMh}>rEUy8{zE=Hn*-XKzYv>4Yp~GT9Sdz=#o%|L?Mw7lPhV=MnhCugw66y; z`pXIQH;>;OZ0gU!rv9>Qf7r36zw7Pmb-bR6=iZ4QJ1Ngilf+4jz*H zikDiu6P}KZQ2m4-kD0!g^+7HvX6R%XcIDd|S!Y?1nnK!ZknFc;Mc&)%y40C*^;LxR0@6Ws50(qI+oavBQY_Fs`0E zRiCH&&AXt2cXsA zA5z{Vg`+;s7ZL3Hi8slw65Y9fgY;+gZOY$Ii0-iGzAgI`8WY}x3)yGA%wPIacwWbM zT3i2n@)vwB-&1o~=y4|qZS`A)muuv-a4z{Fd>Fd1e!iI}ZFGh~7V{LHPT_qc?-O3u z{Z+)PALD&qDDVD0-q&!3P%U%M>_>7V*smX3%kKOOYN(9wU}e&yix*ve54ijn@UaP8 z?n4e$B8N7COTCYCC+nl&Qfn*R{e*Uzye^w2JA;O$S04}H*zn`6oUO}Q0iqG%`%txC zKk)+Rg}hrZW-vzH4!=jhcRMtVPTcD7D7|(9UswBi`jdSxzxvrCYB`8g{)9DZ#J4u* zLO>@)_}5>j&wbRE#=)`bXp%Kve>r<567z`vO71kU{v%MZ)b_4RK0l!wdu$Sk{)aNbq$$XAo74ma~*u)L^2HGDYe~o8a8?vv$ zzNZ!tz7ox74C->~aQ<+Wqxb*4Vi}OKz`^K6@9JUQN9071b&!7NmT~BjgHKkSkn4 zu5bmp!j*SDwPqu^!cF-(tFFt=-QDyFKUWAdjMaPUH|A@;n)fXF?V;aC=ywHhF89K_ zyU7_g9A9z&I^_&oLeW2ZHRU768o&W_H1RgzSKjdFs_(JyU*-Mj=JssH+?>2Vg^!#e zufqp5hm^ebePr1)Mqb}6d7Tw48Uk+VWh1L;^O%7f8P5BSycf^(K?BbXRLm6LHR211 zwqy_GlK+w~tk0seZ@f7$zr)g6BfO>Gd8KdNla$>Vr9Ybt;3$uC`2+moRe9CW<5L8TqgARh8UbrDrnk1Eh_97tHyBxq$VC&m}tyKkxGQpA9P!%<1<5>MXLS zv}YGOPSWqJ|0KuCe%K^DkeK83|LiNo>W7$*{ChW_S^^|*t>BEC4e)veXWT62wJ$$! zRsK$N^#>xa=H=uKo^hkxW*@mt$y50(o4}>hv-(c02qME&RmJgMG6dxu^ z#fR0V=3Mvn8vBg6s$ycwXBTlVxz0FLd^|SS#G9c1E20xMH|8(;kI&@QkBu0bUj~Ep z&ZVZDeq?+6NOR(|=w9Ecww~|%RmYY~&N7$WW*+g6!ST^Qq5;uL44oI`1!K32Y&rNM zYqYkASUc;Nh1|;`{=I@T$0*nzhyAprs_f~96XB9heXrlupX##`S4EI3p0UfEeIYsX z8|@9rV?D7K`lwa+>_7B7wja5ID zg-nrt`Uv=xFE77X_VOlle-8Gt_|qgGE-KwmKx*l?HKzR!{JV^PQToj zpK^$i>5 zuHIW(@W^3g(9%lcT#UD;igQE@$-%%6^RJ<97+FB9?$YRIn43H6N4%nVWHI(4^KGps z=2uTV@lEPHikV}`&XKvb%J(#Ihy69J&ik9a)fL}0W5&U$8PmJmj9JI$lKjYa>RfnE zHFcbK!uipP6D!~3V3`Rlifv}WyNW3$fTarCTd=qovFZ*K4|@ri6noU?Qq~&z+pio| z7JRF?vesfKI7QD|gCOideml?NUHle|hZR}+tHA$}9bJLXvy;~l^rAjxU&Smh>T@l2 z$KdAzpGSkSoDyO=%dw6AgY_-J@2HXAfqp8%@6dZ{oaPrX_Y?LDvz|fzZU-*;$yxA5 z7+x07ihq57M|Wozc79{J&7E7mbT|;dD_54zw~aan=RH(MJ(FsbkQwNq5z!TMjLc|twwB^&k{K65w|ZX+?PAv% zJuR8xuYqq@X?=p96U-q3{wl%WrtckV3n4#>*)LVP4Cd~E;j||pJAltg( zJIR?p39NB2XL|1t>n-`L8i-(iOLZRI=-^&(*#|Cl*8hdz`TDTdItyO&VR85U$&dCz zc3-NM9ggM_1;AusbLT(wmHKdwJPDj8-cX2cEag7;0L-q>2xl2cCbm4Wd~d_&)*1Pv zm`({YtSVM$(if)*yrKN;PVms z)AMz_>iJ7}e%IZr_Li>uPWto6P|4(C!4nui1}*FP5Z~*0#aQ&JxB+olKeozy$+6>_ zGd8h{5j(kpfi);s6eD(WMKNOX{YI|*PwmU&U(lcGI9D{6@~7lWC6_|a4agU8;A}GI z4=HBm?x=;=#~z#$>0esuMK1wQar~hD@Wuh;&+E{la*5g>(uXWcgh!HF9YNfInBG#w zs%&;Ze;!9v!#cS3s-sO>b>r{QBGaHhv1UUYxEqiiYv;Gbuc{-Oq!dq z4}ph1Um3=J+}D9;ID2jLN9P-OmH^KZ;5lI7`4#Z!{S@FCp5xa^4$F}|iS!E(#9LS& z>{9v`j=lkGCPu;dz78428V;vTzArO(4K42^>&CNh%h}RS)-4kMIC=~F@woe_@qD*n z>|i`D4z0x+!O@qXfjfb%Nixazy+v!t@v+72UbKoi*TA1Y!A46$Gy9<#)rcsbG-Yz6 zW9k)^9S0uI>p1v>{EjK;=mXGEEq3_<@TeGAANHH}oM`TG=9&O6rk(^m92E|r(RR_& zE&Rs1B*C-x+Gt-~ZxicLVGlI=HTsQZ^xJ3p%?5wOt?5_! z+#c4wh8T{aL;4y8T#EZCSFQEe-ofCy-)L#E2mKjit)&+?frsay#oEP_jx6Hdc+p`a z>rRY*+W%s`WlKH>?aANT3GE41%}H}H^I(kq@Q&tXYEhW$W2_T2zU|mW*x|26`t@4- z3;B=ie-uu@a|}3rTk^{2Biv=={6LK_J-iIM>jA!A`V}u+11$QiIs<*DGqkZS`+LA; z4mM;k77{jR24>bC__0^rKkdial?w>gTiD;}H&gdI{U*BnBKyHJ=`wh}(Wgc8L-diz zLqab7^itgBLY2#O{3-im$z|frfPV|0 zl|pZNROK8&9o7yZ(Nefl?puwOO3S4BTDJMg6FcQuee@^m2Fa z0%r8}?%d;FyT6Zqp7Qcm#Y$P5Lg@}aRulAZ-1>a``lHC9!1`QWc-r-udY0?s#%#(Q z^JgDC?U?7BdQ9wrp=mir*ALmoxdB!Ji*%$m+bxyyyHT`0fhc7xJgLt=VsG zepBql>pWM#5#x1oOFD4tFH<^j<-0bY63#OF%U_tO^Mljxw~DVqR;)%At-&s9LtZ~f zUA2uLG!j3UO#DFmdus7l*=#;`E%5`zr5cHS4LlOM?N+`Iwp)h|T0-2Q!p0KxZ1IO^ zqxiyW@Xi&~AD}m$qwel8{xp{2z<&&1sGobek3i2zuPx@8!kwYp;GOwuH;FcnebfJa z8T!lg$$bUIyw~rA{Js`FzV1>B-=5HIFX8*q?{@W@|IJN@)wk4;u{U9VY97vax`=kVw^Fzjj{TfnZf6nu5g(aE3|V&8T>kDs z#;!ufwqY|u5AvB3tFhM}Bwv{CJ*4l|e`o^F(9ZoP{ndL7w(4B;0evVJE#FILj$$9Q zPN|#{J@9$f^h>Pi?X2&eUjL-?y=~@vk|xehU5e}guhcQ!F;Ouh&%ZDM(W;e!?4W31zmJO zn~$=F=6iJIB4FFc_r%lZ6OcBvvXeWq%p654%#psftFJt0MbF75G@n^-eXc}bHY?`K zSswp(-RbA`h2Xr9|FT))ZhquNg7aJG=EuCv{D`Ye@aMN(^9aIZK3ka9VUPGQT@d|; zGk|H%dw?mD2@|mz?LCB7RD%m1E{?9zm{XY>Hh*9~({;Xx^xft3q4)9+a#(xWK_(9c zzsR;Qc11YboRgru+8TCJhMkle5-twu|?n|e*j(-Uv_gB!ix4wR3AQ8zX``lVkT;9+QFZ7t{wR( z`G!5n5y=x@&wv|ZJKYW8=ty_&r(GY}>C#JT7hE@eFZ)_?hLip7zE7%eeAD#j{k`jv zPHNw6|M-e~8i>y-wxs^k+p`yeSaR`e=R}udBTW3vv9=GHXK6b^IhhYe>vp8;$3y{m z8-q+72YrplSH%XG&g}a|N@u!wM$m7(96c)EN&j5_XqefLeq#PeF|OoQi~)V>e$nbY zUw-cd-$D7=j%<2}^|`hHxeqLow{-z|dlPzc+^49PJe4(6=+PHjGrE6pJKpwtRX1?l7WN#&VyPWLxn4@Z& zq^Fk=-*K|{66Sk4*&8B`>}2oVXE4We9sJ5^;rB>y49?ZcUN_fV*A^VewlZ@?{`xfM z=jH_8spK$t9+bm>at81n*cVuX)4=!du5A!LWN{Ea@ko#->OOs}?bZ(QQ;H|P{m&=o z3GqVqNqC_u5I;Dy%i5FZoy_}EbrzYfR; z=p5~XAfD4h{&o50l@E8TMvpl}1T(rVj&74| z)m&Dh7qzE3j_g%CJ(Db8k55tiVe)?4-%h6A+q)bZ)!j721WX)N=gG#IQ~XQFvnKvp zkUMu$6Y`+O@Z+h9AE`|tHsp)S!=Srkv|KS>t?w-CPKy!3&@J;s!5ctp9Y=ky|L4z>(OJ%hd=x)|Ne|(_*}NYrja?j z`l}i;Qq_k zKaY(8o$PPKUcuIX0~+atUV2#1Zt$Ii_Y%;e;Z4zm(c35H7<96E z^W=4w{2McUT))fS*;B=KmABkK2{@(#2YT-y`s|IFoNsgzd@V8i3sd)Xi7(u_^>Jtf zn{Q8Ti?RLo)ZOF1)~`yB6H8yqwjz#3&6@IG)VovjGBA+Up}MT(IM2kDb)RnuXEY?K znJh&1HKWUoOhyK*fTt6yt2p1Ux+AeVm)AT~V<{WR$o3Xr_xLg+r?c!3HY_-b$^JHX zlw2A85%(P=ksV?9&D0BbmHK-F!G~%%;M2jqfvN=&pWaA)hTa=Kg+B-9?(076r}KQ> zV#jy$bF%MQ@XGy_(Y2HPJ@q+@7nnUM+EahB{j0gerc?b;6ANC%Z;FA&$3u^WtTi+s z-de8p9tRDLkMKQo$NIYcg|6Ov${5jns@_}o?Ihcv@wL$S8qH}G>u?T#1swzTFb4K= z0@)Z}4gP4WHP*QbdR<`Gcp+6fHfB(LRU#HE0H~e4` zwjnZcyY!>c7w~NiKJ;@V`JKPt2X5ds>5T;XFgQME$Zc1Dz4N^7z4X_|d*vFHV~`DL zeq)Wrx1t08v!8<_-$Y&&xdHDLffK*RDBI)*|NhMZ?l|CHho(IK@;e6R2-l1akH(g= zCmNn>VvKIatKYohVdbG>$ZvS#V1#jB1m;A+b)5A(n=|XKht_qb(x1jf%`3+g-=zX9ngWkO|pWbgCg5Dc^djBSM4+-9<=sklrGU?uqAA=sn zZwdG<30}IHZ_iBLL!-ZBuTBiQjKgQ*C6h-)?kM;7C(yNIw3E-nOzu6EZ_)m|K_8ZW zD-`4LY2}Vd#A6g|D8_EaPTAv)_3actha2u(PRzH8n#Cmc6|&@^bKr%c9#)PsU(JyYvU=+zoee1f2Xp+mjQne7Y=8xziAo$6`i9xqw}MG22Gi@ z=6*`WXx1sNzIl|nXD1GAEBCX(oL??~l-g(TDt}ot`}m(gThM{-vzDw?ydPeh_8;WQ zb@pU6Icn9`MX=q~p4iUjV(KA->mi#WnA;S6Y5#hr&ePrBqAT~$(bp71U+L#BWa@h4 z9QU7bZv(ikRZV8|`?)h+cC>0XRELn?`Qsn>XO`7*hr+Q!UuF$7<#hDG&pQ@)hf1)^ z)@@$?uwpOB8ukN+*fUB_47_*47rq^q+xas3MZYQDqGt>p(uba@pr6Iiw(_ZE(2ZMz zVOj(DB0kg_SXr002K3cSpTg;lS_9T2a}6FklQjtYwh46yoc9$vNDP#HAN5!7&kmW| z_rit#`PF^RUNq00Uv23VJfbUg25ENZ{|0w;=ncWt$GqX;gCY9zh)<>Qn;p-=?&Nsp z+|FlC*x`E`S5<;npC4Q<4*!&K&o!~5zG2=&j;7R3_%O6($0lV0((SDb6Nk&Do`AD9Sexw5TGbU;xbB_e-?1bZ z2W=d?mqdFn7w~K?&$i$X#Nnv~vQ^KU`Pg=8yaMcB%78sszb5{0_qBA~bmDz2+}Fc7 zFrohBh){pem{9+skKtPkK-=T|&-%W*z;C0rE{^}^|n*0*&Ltd>Fd;M_#MK;AsqZ0*$LA7hLgk z&-hz8(Q7M@w$*vwTg6?-AwRZiWDDO-PY(Ec?KNs1^j@)?*iZ4Xu>re*X(jjr9^?!@ zt%XN0w$pzL-#2h(xNOlunC|iVyJSz&eiGxF_;Z(^Z>!^Zo%w;RhK?(ZKbOE>sv(b) z{3*4?qgx)9{tnuiNyd_Wv03o(-o%=MairRU=*JfyZ%aH1+~WBS*shYV%>}-E1s}cO z!_dhH=)a({|C67BMzQVU$eCK?EHr6++nL$;hv1+VT!tcERCUeboBEc@{#(#Rm3(Jv zkF>X^IXn7W?q>p@n~Njk4PIS+i<8|u!Eg5lXrDRuGIk%Z$cDO)x#;^caIN2TCYyer zI0BoAHk?()J)6Y1nP)ffW}^>sk@?-sBbzgW^Elrvz80FEiLD@hSPIOY)Nem&*G&9k z*Jz;WBJ8XWaPB8|Y}19jAKwxCHSL~^^mo6?_~E>0&o97TOD-}tL=CH7FRy(!-Pi>W zUv;c4GQ!)IDEKTk*XKGC1>CPN?!1n~?-+lLzc<9;FhpHPZ}@oI(MMTJXi+p6Zv&PG z@t1SFt+~jWXVF)M)YTQC-{mXCzh`0udlvn!5;^CYGmi|;xx?!ABA>^NjPlNlI{G;e z92&pydlAJK_AKJ=sm)dJvhY#zWRfrbcw?f8 zmPmhniMRDyasm^0pf_Rdli`K#W^{T<=<7T}#i!MZK%g~+XS?J>5 z?X#jAOBbm-DDAVEbWtq2fEIRQAAgwjeVEtS54`^Cz`L)Hb(RwKR@$2feJiIbdDEr2 z{2h`Q@PUI($Llx1V zaBrOaq+-sJDnvhrJbb=g8du}oy=64dD`!M|-lcR0= zm~P@q4L}n5oVr5fYOc{IL*x?e+K6`Q?y2lYF0`6^c}&fv_Dc9X z{w?+!I664h8n>fYG%q6;czvJN1zEBaSyGHFDa2;1#2)HFmaJ#a8;~a(nRoB5S>+RT z_s1OLyRzoW^<{Myf>-&x%J&;RK%as`XC(S(A?U1q(SYh8ZJx}|sRSHWd0Eja?w4u@ zf4w`Xdv5olmDsMbwX_!`F&7(~n(G&VDPDoB+QGe4oO`zLg36A>gQGfn)|}J9=l(>K z*IzRI^ zu`m8)cB#$}z4x7zZR+vQ`9m+jXPNQWRvUkJkDhaWTo-eaECV|mOzf<5nR6Np~(N_`aA0XcG1ZzYsGkb@dw%kiDH~&)~j$R3G?IfXK?7*G0xYy)rtRd%}$m41W`cXf<&N zWW33xAoHBOtV3S9{ZK!D4Lcj089N)9iv5lqaUOPIKxf%;lH7AF+kXl8aARJqG3|c$ zWjed8-R^e>UfC0AV=@oP`0W~#_h&MuFK5%^O{e4i5OjAc-v1ml>HBy7egprEvvX7X z4O*bVd!WTt(B*?(e>3|3LgdnCcs0B*C~qZa{h9>TAB#t*p&I*kLnyk23zgpa?Z-Oa zY4BcYN7niIw@`G(&Vjb&*v3itNWQOhjqY{TUiLWaZ$4x51oJB2`G9Z#WU)3?*gw?o zVpm)e)t&{%gXTHxpR|6ke}et2vHj3?Pkqwx_nwA6-+p!dI9>1i&3^s7^AUpIoo)@< z*$aQFHE6eTL9#*r{PpnnGu~(7^Qrwt1L&Cn&T2Qg*FEgH9?`Vrk&)g**hJfua}4o# zv3GUUZP^_U+&O1e#ab;`90`@^^B1~A8cS8#m?RM4!(5(__RME^Y;X@MKSd?0oZhQ z_ASVrLhQ?b`P&JLvPh+!tr| zcV^@lli)#hUxIr9e7~Z$Ji6lezI(Gn-gNx%R*%nd`ENy+M)w^>-$K`tt;hEHXUyq- z=_cZ)WyDhJlwT38^IMJbE65wg%6;20-1&Lc0P~w@+x4U86nDMX<}5`|CVyw+oyy;B zBHsBdG`Y{-*SE2mSjyrub6=mQa|Gp6bt*1=yoZ@|eR z`JztOtp3Q+wspWG`+p~~|KcAP`#veppP*~NgLKUYsY^-n-su|gg#0AmwwIpM{C%6* z@T1LJYX9y}c~*FLa>2zvMK6n4+f4bkI)i==wdDEtJ%hWqTZX0G_Z*u^Y_KBKAHOEl z-#eF>;T+*2YtZMEk(v0`grJ)A!ZPPOm9oVDQ`eZL0L;eF*fHfLsO$<-?PXVsL7l7h}LlhHtd}CnwTw z5a%1gxx;q{`Ig{*FWKeZaBnO7w)&su%nRA?qO*_u8~N6JKI4NqJtLZI_+1q~W%yt% zd@VgBePX`NR8Q0rXkj>EkoW&f81yIBjsNiKPt~(uSa-w>ot<~ zLWY|4l1-pBbGV)kuGevfoaC&*_3_bXgzH_@UYrT8x6|&u;QITv{r@doA7W2!rfk># zS7WH!{~Uh|Gx$!`^R|QIEaYJ3z{}(xLd zQ^Pq(<3GzjBN`v>aEd%Byeq3C4gZ;VwX!1Q@XF^wydo>IPKH+}2aK(f;rkASYuA1V zx$&_fX!$z<`Tkz;*?kWCTaOQo&&g+r&mU5&;AF;`;`2N9ndJWm_o#ae*5u^{>U}cW~=(`k&Qz; zbpBfn7fmB)(zIm1P)AZUyTGOgw|}6&u&{8vpT(@!dSa z;LZbg4d9~I&cnblO=qDtab98iJpPG(f;Jeky*2Hu=N2m$t?c`pU9UBEtt8VNefaks zr1Xc@QLrT8^9Js6uYo@BG3QJ6D$i(SF!QOycgq6aFzZ&0e5+$EyU2ZdU1Ou!$M-!C zF1D{^jdk`-7VBH>evh1Qey_5>D?eTDdE0~|(O(sPy7dgo`0Ko!`=2G||7?*Ly=U>F zBVY4!R}BDL5pbyfC6Bv}rQekAoNxuQ6&)g?XwDbpN;T+FMR*oo6`B*(bf_>o@iQQ z>7V+M-1R@5_}TIMfsFi1JMt7ivOgb~RQt37UKLMUxLS*!9)xR;gX;z0l0PoL+`n7F zp1ot`PObK+1n)G;kVoYfe&yUw25;tk2H}UUgZv{KS7!!_51p(9pL>cUWzo3~f8mro zlwT_TyuiW`A3^OsaAb~``PshR*FP^URc|IcP`c|!$jS!d z}jn3?Uj@W)**ZZRT(E-Dd@xw!r{{O|idRd=lD}Qz7?m2uX z9h!v>ck-S(p>pZTV(Ij2BBnM&@esvu4XtrsY#%XfwG*D|!Cwkrxt%{cNKTi%ROUBw z8s<0UGI(D7{D9wqVTSuGeWrC*3?t5($Tl+iT)Khwwde+;=UFq4y{W3V{pL*M4YqTL zc=Ij9#OlZ=?^OMXjg2VQr?`#u+Engk(D%h*`ek0KLou;HXi#m+h>ytrk7GwFerWE| zspdWYl6Cw?!2YJt+e^pCuzhVYqXd)Y0sZ);c zg7Ro$M1*~Mg`Iyghgu6_Fq^Qy7D0oLVvEUUH1<<|9`=*&*YtCVI(zWXZTm-&ahq6Y z+jdXGpskf>@5Qd;WmYH>vo`U@jXPt`!Y8 zPo|b56#eX_tT%X3eRz^JcC{tF_^rmr_Rfud@6YXh@_~5{t~ujB$M`{CWS@ma>+9e- z-Fzkn=JSKInU7aw<}+FI8RQ9_v7S1s+02P~-g%n&Ja`84S;GE-cMDSYgDL+PulD1; zRn&G?wU!Io8%?WvgVFq1jzN%_{e9YTX#K zi2j4WIUW`*%@54|8P?IQ-;xY^6x_AIt#N)$%+&N3s6nZsADw6Ip6?*{5UjWMIFms2 z`+v!PhQgxozaT@Y4ZgL!QuRjksTgfGZB|4sjNWor|#;#}V_3&jDwWKZF-QD12 zt&-1YKGU~8tKT2dZ<07|jCi)|SH8xJ^c_RK`FKs$;&|X;#}NE5#n4ZRzZH+G#AnC8 zc*~o0e)Pf+`xVD=S1vV}(l3gUZD8D1Y6^&$C8yP2hrUqz zTH3eMeg*9n*I0neRD8G|oq{YOc6Jf`64^2zn?mjDX^;OkUH9c_-hXAs(dQCmq;wW{ zvD=t`Di$D`dG0x%|6j1{%>`a2-*kgHW8i&*^$gl`rFy`D0k3>K^J2W#s>raYa;&l0 z=soylxnKxqMPC*^Z^n;dY|)2gm28-Ys2ylO&zm99;x*5RntVC!4F5sv%Drn|amM%* zcrkV%aR%92$%?;!pnAtY9^N&u@9>2)!5PmjK=v<$?-qd{-;T{TIVPQNsPlgPyM%N; zXg)kPtaHIM;!dNynZ~c@`Fe1!96`OsyThaG2jQ^{JnDVs8n|`Vnm;-0oi=E1t9VJe zOtdW@ILpd|2>ja1SkJNMN$k09?5iH|@-1L#&i7^p`}#Wkl%Q>OmlqXXFMzJe(a!MY zdSuHQcxj>GrR=4>@Q3OF+;4Xp9?CX6r2FZ`LuxaNzKG4tlnvC4{7I5~QLbhBVb%zH zM&l)stKH~d(b7U_X(P1PL+yoV^;+Oot+{B`@p_Qo7Q$~E1N^oSIP_UCh~HL<-xyD8 z@)l!?evV=jMZytNf9%U1KZky{zXkjF{Yy@s*%}RC7oCr?bUu3se(uPSkFCf8?U$~~ z15VaTYa6}L^_!9puRVK~`=v9+^XwR3&KSelB>&Dh($Oc|e|?|H_owf_8X%@Uu+}Rd zfOlT!Udna&h$cUXtg1CN-(}I4rlD`g`F5+;UUo1tc)IL&(Qqv^U~GBlAOw%Z=*!?5 z9+%A{{tRg^1G)OR=t{iW0v|KB;bX>M%h^&y8^!bZL?DOWu~PI^7T zlZG~0J@&L-%lgE=1P-2z^y7FwWc)1NzsmdNoOvAkJUFf9cW@8z|WaU z-$#edPhhjSnkQ;5$I5PseLm8!Gs>2|GqBb4(U9t=it{=XoI812pr0cA3D-~cRn#{& zAhV2Z7&80cV_&j#(2`^5vY2`xe1hFo+>cE>?EyTMRm<=z@K?%fn!VdR?#!ww^KR>+ z=7~F|Zfn--H1DP0b8jsjy_Uz8Ynl5l_Do}cN>ogIkZXxDH|HYx4Rr2?xV0be)vsAxd@2i$pP zUUcI&Vgiqn8$u2XM%Dapf>(A}*f#LI@}R6KI9F^Sr&IF1lrzv8I0LPLGticK;Z=H8 z_IL92fobs#z-7)(#s4bJj>g!R6yHGI9sKL)+O3^vdpZ2FK(u4|Ypvz4wU)ou2KY<0 zURl2w*g6Y3SV%pacv3P$JSmwWo|McGPfBKdj4?&4%M|DQ`uWi+Y)RR)vct}sjXs(Q zP8yk?l`HWpkGA#9B!A}XusyQVu&t)Y8=-U7EuVF}3*6|l^jHzN(f8HJlv3#0=s+uj zB!B$<;OR2PjrA;JRoWW4b&6}p?77FMkYS&;>r#0N8FrECv}E5i@AvaQpFh!dAAH;g z&rlcjz(y}`mEzZrg7*Q7_W_Id0gLy60Nz#GzJT`k(9dIw8?P>0=HtC^f{*vYDL&o{ zr~7ypo%#95oX#VhttMIjpTOS3pTAcitMjn%sL%Wr;=f48$%ee^X}-&g^k2qvsBgdC zxW%vkUSjsZpEu|)Nq5SATnB9A3(A{Ta4sb>O}dF1M(!VMda!}{)_UG5(Y<&#Rcp(+ zF6=d!%AG8UrHrjR1SF?sLleLveY@W3$wx<_r^iJ43(>b-oM{%5PDXyjXJbR4tDEtm z^M(jSLcl}thuY3L4^>%o`yZ*yb@x;L%m(THk!++c3 zl}(S>`pqGWrO*#y?fYse8>&Mo@&5?4(o`Da)E$G9`uF;%kYfkjvME?n+XMA1_wv{a^ zdrf%o>Bh$GSpQ!DgIja0_foqqf5*DG^%w?!W#(u8n!97`i4VjWo3l>1!_Gg`;YQX& zb@$5ki4JCt!1{UwbcA2GmPKJy6E?YjGD`F=fY1~7uH98#QfaNo(0F! z`;mS!&kv;6Z07O2o~yNS^}1$qL%{zj<{!jm@V#OQ(zneSxE*V8tG#_`^bU;#^Nk*O zJCO5ETNN~CWFGTOoz1^&BzQTQuayO$$38e)o7AdSsj8j>H35 zXT{GId-}qb(jU+tFyZ6l$4!d9N{xfFfm*N^#Gl>3COhxe-9EpT!Xx_5)SzNdab6#= ze_@O2V6)+wLe|9Z-<|DY=fIemb8>R+`*vQEYcCGD9;y6>;^IBXvD2G;KYYuYuk6>GZ02ba?>wSBH?FH)PP0;tt8H7+-RW zwQ_lw^T2U8aPlPO#4lz^**9aIrT7|MzikL{Q!+ULzfltOkGB0?)#GS-_NFF70Md z_6MP_tj@1*ksJnoLu>xtAlWU@BJo|F<3nBgHJmw%&-NxUa>;Vt&!Mx`B*%0YxN@gS z@X(EI(G4!->;IL-MQ!$xCu6Lc&*RSK@a}AUPT9bZ$fp8tisS0; zt>x5n_JXhHMq^7suVHG8p!v;>*vOA!18kvAKziW0#SKR$G4|7oOOJ$sIS1I)hx*j> zs#(=qyac|CObh(h4P5$Nd6l({`6hk~`fUCRFKf5_&}-#~($7ZhojA`m0e=bb=Z+4q zdK#D(SeTq2^9^8{08F`y7aY;vW%jxpp}(k!KiKD=SNsH7Bp=bNdy7y1?z;&5D_MD& zdDYp+*`3wA*IG^CRlhskBfbpM$qPGuxg-8`bn@i@tq0}j5+je&>mN*>9Xb%a(tFBZ zf1LT6c!`&F+x@^^&3yE|^5_Zp^`U!_H>|B>VH5tX^r>X6TXV(1vWd5>%?{nx0KKnW z9J+0?q5F(GoSod+Wpf$K_f7B_oU3@DH84lzbH(G5p@JEI&%kymV+!U4>_1w;{-d=E z{PUI?h-rrHc}t3Q2Jvx&*;|+{zcjXFD!h(4Y>#6p>~S$kqE;m_?ies}{hT!2IJ>nMFzg^9bO+hVz#>mpvDt#pRs71{?@iVYKt z4~bXhLtjNd?s@gEx(oG_CVN>E$&}!C!s8zr?CWYQbZdzlyBEuE4?eocTr*Lfji7%??u zST<{>cK8?Q8+*^NBII4Yp z$<~&K(aWzEa;DjY=*ovT@_vUGUfN}NocsR7quA%FiIe?5dMG>~N5tFT%D{#6T-QnP z;qO06x2NjJNf@4df;&dRiTEuTyAvM7i-JREZz$K+w3RzVz{^lPu6(uZ57}NltG#k( ztEnE$*_Z}**8Wp&QhfJktbvhv$jm>Z@9V&+#!61|wqJh_@^VsSd(Y}fxnf+V4jP_| zkHcqT53k}P&jWj&_t@ShFK=lH?;G;Ns~V3qz1or;Ue%yB;J4%W!2B?|aoON{m7MZn zO0OmDAy$xO`pR0`04&6hw&Om1CcWG!97_iAJ1{>3{y?=K?R%YQYd+M^l zT9KDb^G992dCK{+Nz(Jp13SF(TIJL)nH*h5&ED1T474rNK16Jk5bkVo^X1WaH8`J* zj3Xx%pQBgeOFYM%P2IOouQp#*L2WtT*Rx++anblBp0)C^-sUgn4fExr*{_Wr)*96_ z=R3iJ^tx)w>WK$b+5Os&1C!3roC*!gA65;#TYKp}&EaJEPU#?>nNkBUUJMUyLJygH zRJy#|@JZ(nbbGfg=CzSEeU`P<{@7=;^9J{9k3jw>*|V)!w&-jl!m9B7YXSW|~Q~iMKvpVcEH7Cg(=3IbWZCBzNxI zajfl+Ee$*W{Y>+BuMOB*tU#^JZ2EgI`8%^e54|0%$1%1Tv@DosX@58ITJ7(Cn))So zMuds0u(s4QwR-$cJrifpT^emT$a;YjXYcAPmcMvC<6d^lBSHJI9JuAT2JOe!EM87T>Tl8pUr9(z1hJNQ&!f7`a-aULK1Rn7PE8HlTGHS!dB-(}7Q92G4x^=@A^xwSgQ z&AW%6bg!G}OL2eX*tGAG`^Oe1zUBFQ*sj2ynHDkinZb6CxOPMAbqlmJK6DOr_-)%> zK6NQLGJD<7Ls>=mAZ5rGWOkRS*Gu2)=6_cfUK@S?xY;rw}Iq+$>id-4uRUN8Gt^IqAAC10kt=gXW6-Ms8!)v6iUXl>|b z+5|sq{4U1VnFs!A+IV-hUmrB3NHjPu74M#lZGAIyp67hV3DK>W;Nv1^eIN3&=w?30 z=dteKq>8iRAC2T99 z{#3rU?9}yRylq*E?SL~}4rA7;&1iHKyr#A#dY1NjmU)|) zjr2~fWRf>Ix>PtqpB3Kab;JyvkBctiJEPwsBMp5+d##d{JCT*jYs!a~|Ezm+q+_z6 zLwzqA*KeXl{aywgsm+7}Xj6VM^v`)}=KQiAcs>U@m;B42Ho12kXKRqV5zQ+8&;`8B z=o9r<3!dvFOOgGtQsERGNBmi5{*WItec+Stp5JB>b^9-+FAweVDkqVCVbtGAPH+Z8TS0vtfH>#9jttpVUH%&Pn-x)aP>9_S z=R4W4@YsCi9uwT%>tu3*{G(^`38SMGQ{gq}L40MN!Tw6%`($jWDY#|{4?x9;=S%M*Bxt;*ZGrqrU_YwZIHYN9i8WGYZpww z){YFQ7|UH%__s^Wr?#txxil~r+IP1=r|@R)YUB-cib8p)7FnbGpLCGE`}#|qcMCt7 zJrZ-RyxC=a1Ab4_CH8xbp*zY<|Mac#F3h%l&!Jy*hWh@Z>l>Mum~&=*YkunAhdtcc zy9zpI40B&L<3Lyb9A-0yWG~~~;>JNfC1#)5IMT_|G2~xb-JL{|$wkOxgO3XOnwQa+ zlUoiy@JTR+*gXz6b2ny)nwdn4IbVRgd;B{C$3nZRbI{$F!l8cGvx;TP zE{H+%x-V0@xkT+))5s~>4ZD=KQ{caIxdQ=OpTik{PrUq%y_byo-r=Uu$fd~H`4`Yv zMX`6N7=Mu8``zzv-1PF{OP+XB?GCoX7uu6mF@?JjGTM#2?4`rE-s-nQR?VMtZsd+9 zzW&XD5i#p<2ZEO}`}G2;U{)&F%(lhYz6R;KiNv@9W~> z=%rzu4LU!&NPC?lTk1cC-3YCAH?XeYg!(3f6WMSlJX3qQdFE3mdFJSYk^bw$;Ij!D zReqK8r8OJ)r`A#w<$+~p2Mm?;B|_SAGD%+xj6a{!`)iT+MVCpIGR|Co9M`{}Um~7uVL%>j z?MD72>1+MqSh;4bdeN0&q#xCN8QBG{+_S~-`Fh|@fKS0uUlc0W-IK~Qr}6=Q4Z=|0 zg(gex(m&5%`6e(=X72Rk;~(1?nG$a#XU_ceeBHa)kjRtz0_3jd4ZLfS$+E{66#4I$ zK{wcB_?MSPi-Bb=bTIgxcUe?5ookWLA?U@lE4nO-^I+Q5^6XlCk0JYOzceb^Nb9fb z(kMJ=?))7bqxn+ZwKCT9-F0zvIX2l*`S+4z$atN1Ow6jk8rYjG?A5@o_v&L8us2xP zn=I_+I}3Z0h24B-VOKlB-eh4vp+5_IT7MSyCiLjwxE6LFhKr;1z#aqNs@FOSO&j@e z4eO<~z9L;;cHcrh4)e~Z{z&g-vo6GcRZT_m+8~7uL+QON=L$lwrOFL|Fw9Ui$j_Lpwt%<)a^W1Wq!_dM;1 zWsX&xFgEr75nA!4`n1CQ{JHwHBD*7g^Zz4rD6(^y${gkd=zZo%=>2A&-fyCwgEPSY zV0!PJbF^&|>+aj$zRZl*Km!&0Ss(Jg$?_rZc0Ocl@!uBsK4is-K4hZCw;3(~KE0Q( zk(J>?2FDsoKBxa4$bYr?@O8tJ&B%dk?4F#3*eX4%eZ1r{ZZTu7v@$^7Pe2ZQ4LLBe z0KW%0upBwC(w769y=KV)>K@sLDmk!$_)7zFpdGwupZNhV|NbfLF`mL6;};h%Ir4%x zY*kGt`yje-ejl;c;q2ovHl%D&Vjb%1qlG++&6p7HT@~s78U2*x__YQx?(B=BlkWvi z*rGL=jN;Kx)^zfOqI@%9$v<4j5V!Ar<4@&8}fPiNc;%V3bIo3!Is=CTk`m%qitFr z7spntMX`kksKGFC7uHC6316pw9lCfEI=O=LpX2D0=QtzeF>Jia*m$}ZUbdUD@rryK zkKg;<@B6XwCS&8#?qDsscw^&D_HDdiyL@cC#n^aiw@zb>j@aSi(AD|c4!oyg!+3WMo2yUN_%m=ymhm8R&Ii#`t!fvlGmB z8G7A(XZ5=O-q#g_HW4sKtQ=H4AG>ARQ{Y7MZx-@THEXkwdx~{dBKKmE7U5$kS19{ieJ;apT5jZHZs!*`qcsliO2;Lk>u&f~b=l(o82q_t zm^UM2+J-t0u;-cmtlJ`HZ{Y>eZ4*5f~T@AKZ{N_Kqu3n z6V+Ie=QlBN$zl0W&Xzazs1;Ry-D|b~y6~E+A^R#GqA!OV;Up>F0{j~PADE>Fd>!x8 zmpM}wxpachIUXIu^Mg7DnHkhE4FNpX$p)}}_mflCd(H2EqK@expMSPG=8r#odL46v zZ6o~>p9K9QLyeA^ZFLMdH9BU}NpuW!t=L;o$8etWw%Am@myQ9qnL4KZctFQs3pgEf zbizq=%;hK7G54{4r*>&ErhJC)~&j{xWRN5qB8wT}6j>(~-0e-J-Lx>x_)`rTmaUsL%iGjH(b&-?Bz zNsG6)L#LnjZ_D2Myzf5UysJ}l)V%hxXGQe6Coq@bd=8j;jMRCaqQ{fXX@};7T^(cn zWY0U9_fIFA2kUc!U5!i?4c?@A@cwk`=jh7NprNfV4bs+{A#iMR?eL3(?Qc#7+fcNH zeh_RE1RL*92b+9Z;rI@VCo+Whu9WI%HU z&uD)~e7u=gt9+T-z`TtA;CKevV{yL6`J&o?qIv!HpP3i$Gv-Bq)|Uy^>s9-FQDx^A zua1=O9%6nkn7xjv`ALo-YpA!mwH*6=hGNcT$etCEaCD`YjZ`$`6@a6!ofzcC;Ih|kZ1-`%-$n>qyqgl7Z zrD=ZOp5hhO?;B#>l6Kvev2J=_D*Kh#67^ctJ5G!yu%SKRHL}gFU7R`A{U`PQI%mz& zjKR!SHf8>o~?v1QH{tGd8vzJfxM5>FF{ja@^ zs*nHOo&ok_;=eRP3v;0vba0p&m1^qELey`n&gWw~7_dIWW~+ru#V()ZozJt zL;uvQY<#=ioEO@6GdAh$2(?4U+LS*ux{P+pDVwpIFEeAy_Sb!0kJHY_u^%_|n8P`O ze;#mD13&O3fJt>cN#L3VT;%b}zjZ3Oc&3~7X>iR0u4?Ktf^eyChZ6@E?RsgqG6OEN zmcjEx!#=!Do=&i|Fx2_0nde_e$kAy{!@O#}-P+#t(~P=BaBOg6Yuwye!|hl)>sNH< z{yF)5AGS@fzEWouJk9wz-N=}BQ@6$a>D0%_zN*95lwZ~jk4fG>NIi$%D@Vg;-{)oB zm3Jg>vBx8CnFbTzOU6+f)>?%wFnLsW{(T1A|H{5-@e2pc8Mp`GHT!u}^4Q5x;b?>I z*O=t>zk+`r*WDV-K{8UZQZizwyEV-B8T&l{F3aAn!Q7l2<$T5&U1p!*sOWa^;cBpq zOewQ9Sd-8_z)?;7vuZMHhAb1_qt6CkMmOQRm(p&*70@|xoox1EtK9gqMshi5&YYyDrWjc=#;V=e|F1eWssgeLku6;k->7d)!kz&7SFS zn)TU7{l=NB&%e;_y{ykqZ2SKk>vPU&)@RBP_OSLu>OP=8|Vic@~E_LmaV|Eh9_E5zJ>9h&>cx+$WS?)t=fB>$CxANvmrt|#)RhvYke4N zLqO&_`8Cx39LcZ<_YCNjeZJc_%w4^=wBV7$HSA|uhOO+&%gd+@%{mxj{jR~jju=@q zKKeJ{(8c^^7t3ZWDe|I0xiiJlHMo>~O3UBH3e)Z;sm}YGz10=pHDj;C7c^sgmz%MR z`CO78*&aiddd(ML3ubkGa5r^kR>nG*ZUZKrajiOB)uGD|s=mU9$;r5E$3wtWWnoHy zzbs^)V6R6O)}w2jZGYrP0oxvV8ON?MIwk}DG=DflhRK(?F0gLF{W(QezgCd5`U&}@ ziU@T-yoL;&6?XdeNs~E4TVvFk8c6%Oz<=XK@`5 z-&V;UmXb^S_i^^n6oC`%hmcNo^f8R}4xT664qSP{54t~$U6+l`lZ$RH)-gZum4Z`enoZQ&vuz(p)Tt3h}&&jXkL>G)YXS9a#~w+GrW zf6Cp;)GZz3zR|C|THXIE*5SG^XH#6y`9n83+^bHR`)0`z>l}(CXK!{PPlEbYF{ol{ z8-w4SO+RV=gq2BJ!v`|v=;YAzfw{W%boI81R|o5D;|<`n(d!=-VebL{lDK=ALJ3KSIr#&M1v+ zIe?tL5xf7!I|`x)iJ83-9tFN~O?_K3TtJPv=vcTsr*k5AKfT!y>Hke5_n*|TUuh09 zbt$!zpC_+(JF**mzEAq)2Ht;M>oAIE&*>z-uxABh_S!Lntw&!M&`d6Ui$*C+Fw!0n6YZ`G@JQWc>RCt*Q?G@{5&x`(*GTDy5Cv$;pk>wM=zfk z{XVZBT3*Ik0g){ec~KVXxPJD9MjyoHp8uO+ z;A2#z-}SpMbnd}gzLzXgjywhq;@}b+ZgUlPmb@{5FLoN(lHeqWTj4s0ACupyM)uIZ zaKL_%8I^AjZ2Ka2@}BCsZ$W$OjC{29`@05n$eR~6RLU0L+}Oz7h{fQpxUxJs1>Jcm zuzU!6pFM^y4&uubn;Wol%h;io{#BDJ9(d*Z>GH$zfOK({zC26uyW%h1GqHbI z(>~Vs`aEi$KkC$%o`0YhAW$0-Vb> z{KZjh6W!TY2+zXX?RIb74XWp{cbVb8Z7z0qnZbWU_;1AeeE;ni+f)95X$StRz$rOg z*(`(Ye&X6;%b0eoPZfCAJ#v9|=1y_>Z_n8F;>S|x-0ZC@gzxa*+}^r&;v>U9!jAT+cbnxqjQt0sHM#bA7+-CX%RI9K@ejORKj z(*H|4R}=rcDmB+vKX`WQTIc3)nz`QW+MRB$SNU_D>*mVm!MV!DJ=1l?Zhqa(6`2=@ z@7=oo?OD!sqMO5M=6aE9ce=S=?$33sn=79O=Za5sy1Ay?>porkv|bh^ei+y@8qfp8 z?t^+jI^YwVtqvfLAIwu8wmC{C|8Fq&A*SRhdTLv!>*2k{Pkrxjo!50(x^7u6{w#7t zHfDFlxAxAu{n5i^><3@@>HGIyyYvT#bKZXDa8BF)!yE6KwPxe-&px=}_=??OF%wLg>ZjJ2t!coAZf&K#h&w zmxR4RpU%f!M!p02JXizdWH@rh`2it^yAhH8A7<#w|89Jh6pKv-kA$;&L61UjFth=0EAbX#QUe z%>StjdO2l2(C|lO%gVEu?`O|qz6YN?L;esi*nHr}+}twqfyDGTdyjC2F?Pd$ReomG zi>J@e{QOO5?PlPcja}22o}Z~6PJLKLey09Iw7-S+bFc#}k5}3J%*{tm$j@vZB0tlb znV-3XJ~Q((W^9)Sx!KRpRMXD415^2#O@e=(!S6})Gaot?TywEQ@ngu(q~Mz8=V#^x z;i8@LGih+qPWhP+WWZ(C@)Y@*fBLS^-+xGcru+Fb%g?O0W1UTYrrDf3Jtq3s7nwNf zpnh=qnW@nKDe^O$PnMrqkM6ShnavF+$ywUIc_G2CIG${9f| z@X>?N%9ns=HGaq3_FP1}UQV_hJh+%j6^qpzSfl zGBWjZM*dXuP)y-e`3K?Q%KRB2^2}~!+KeW31LpOPG@md&|s_>x&Qy~9f)}=m> ze|U}dgSIy7Z1WE%z@P5Vr00{gMiT>T^#7Oq!#eVVW(|>_vW*-L5~s;OB%espYs}KA zV%7R5f6eW88pfXR;Q2e+pOwHS8X{iZ`tr^XM1NBuSx#&^8H&1?bTG!OebvOMH$O`a zyC0)&B1T=P7D$}@LE0*{eWZdWkRU>|a@3XzZ2>W0 zVlNc+5lbx~iVC*n(ms7#+J}>zT#Z(s+7iL$|NYJ0J7=H6$pxyP_n*(_WbZwD_RN|! zYpq$cX3ZMeZnSoJBXpvBm$qVMtuu%1mdBnsEpo->Wvw@gR#toPk3D~}-r=b!Q`?_^}F&JNXl zN^8^VS63gGeXuuw7Q63;hOUY0J6B&i#M*f;`qGMC^d;n@@n^wbYDlcl3$k2&Io82f zjJ{N5?{`V(d=>A07o+c^!#5VA2i#_1qsKmd2li=nl{WO23UneL`q2&OL_cF+-6z>s zx6OCP+(oSGbbynOgPV#hdv61^?l`K&ZM=9`&Hj{)Q4QyVd_imvA!ne49(X?>(ibgoi&rFzLWAgu(lvCD$>CZ zdTKTI(sN23J*NykM>5jL%qm+3Og@kGX6!@;m!EiO|2Mw-GlR=)rKN-WHy$#$Ux(j( z8|x8TQ+taT5l!@WGkZ|qL{4?EHsQiL_HSJ)W7c4tdSc{}lO|ZuUH4|EP2!x|3eKiZ zquliRr$0e_Wn`Xg>m_Eckr}$@JkDl8zUgch$-kNCFEge8O||8p{7-c6%0HvO=qxDt zA;z`YU0?apy~kIQ?{GG{QNn!iknzVr4^HHd;f=>4Uc2Gx=(ysQn_k(}h7RV&VL9I; zPt%S6do=#DSv$xfriFAM%hrMN=@YF(ht0mrv7yVriEQvAkx$vuB^=yHw?l`x^mW$R z=|^)eO(W}{(lrW@bp=li4#_{F!S-cXVXeI&R(61|Zo8ewKhCw=4fV9MC;#`;X^*du z_Ixjlp*;Ftkz4c>HLa>(!a)LCXu zAl$VAlQ;gG@Nig?B-xF&q8-OgFI_xD}L?cp$5wSBxQ^vO+Ff_ zT(|4p>_MUYl*;{McsG(X*%>VR-C^r5XNJ~cug&7!aMFy9;P9ky_pHh{LU)P?XI_~w*j@w^eStXS` z{@I=Lx#=JIGjyi@g$9YyjZfVc&TQh|m8*;J*P49Hf8=`TYmxH2TWInqzhvT0_5r$j zr0iCLSFji0!)N<)rP${@Am37m9sO#xZ(l2Wn!1SXo=%ys2BOc)DAUBUUg(>rx~eEc z`?}typI#%Dwc^6-jJDa%JH>qOBEEV9&*hX^0qhq2=zgbj>AOhFCe3|cNu3pZ6Xd$) zgmna&I_*j1s{EmJZi@?lPuEiFnauYke5dcf&NZ4h*|c5#>i)U%pblN%d!CT&Me;w1 zAa~_A(~K;a@61>eZzHYab?mTdvu97tB&OPS&R)q>Of~keD4+7X-*%mgT^Bt{u|Xa{ z-<^%VtJod+w0T!r@H-pPceS6New>C5#s13RX7NzI(cS$AxmA=Ye(~aykg47j`3#iZR|V6|K^G&^wZ{{oVh|-P3XkkWt9_C+*6jmneq3EyqA#|NziFN z{+zN~XTg)*`kvplrBs+$yG9PUY51GQ=M#=JAC@1Fx4caJ*)(5vanwMaGn~A_m(I0x z=U#4Jc;xahbHn&L!U^iN)8hSs5`4>?V%slz-*0T1gs%R`ethWuw9D;s{} zYG`c@w6-2v+X&7c4R$`ipY^+kgPg6G&zZ0Vb&Z45>YC0;uWQL1P}hi#);OK|2479u zHKbiz*Z6P-c?XhrP+jA6-rvpphp!{;deUyFYpY=HFvzM4ob9ViIiGkCck_InW&GD& z9ZmGy-7l`~9$S`$jkEzAyZWA?7uQzYtzY@;zp<0?TyyXR{ESZd=}_Kl2v$ zxoj%GGDWfH%swjZVUbRybqj2{mo+W0b*jhk8PL74&6_z)X1#4EK8ra2p1Ga_gJpcL zvO;glAFL`Cr^4mC`=YpX@#Nyh#cMY%tMF0dEW?*M%g};<`os;kYZI#n+(}->;D&Le zr@5u*@Ru8Vr^7Ob23>hMCvH5>%<})iUPlUq3Xlsg*>Zs~IS^W8=DeQ>y?B6g0>E)e zjk^yas`IR(ew`KKmfw!d^!h{RFz1wSIC}*;7EyN&Ha7W)ui@Fpvpav)J|fo8;C02y z_1C}kPUlK+cX@<)YpMMGpw|+vWqjMrJ6Cpk^$)L(@GTEs{^W%(P9KYb4<7SVey$52 z&ppBSvfxVwJ_oKQ;9TUu_ag6z*J;1Q4_@z(Ggn+a3HVzNOX~H*h^fd~qK(4#m0Lz)#PMm~+m^%Fo-tui8r4T@8Fq?6)bqi1Wwz zt)YJ5b=_F-!|(rq|8BhE18MNe)y3=LOTj6;^(4Gr310baBY$6b{htj{ ze%$8pp2h$dA0HNfb3F-{(92?eU+V`ht2{9Fc%Fsvmy6D?TzHLqf%aT_6Rd*Sh4*^ux7af& zo{~&4@NzANNAsaSozXpu-yxoNfRiQA<}5>>yytnQm7jM%Kh8SeT?;+dLXR^8r19h2 zIg^&Gv^3KAxoI=ucgYFKvSD=UN5~wY=-Ae=qgm z?m4gg`;g>wwHH3x;>tnc^4@;HH_!v0OMgfFe&e5U%^36}e{zg}Mi=^!&L(f(YIQa} z&)EMAy2;NO8=uyh?^a$HIzv}Q8fPJ=aQ=pEuet57T`R2&>{H{(cbnD8+?YMK`N(0$ zBIKE#%l?_)`TE~GS8hkA#0G_)a#qNPuXHwk6=z95A0On~p#AOfxNp1iJAdhXYv$b% zlkae8D}Dl=ZzU%MQxN?*d>61>f?uW4;m!{Ib}06;=UHcBU0!;(*6Wy~UDo^z>k4<- zbB3Atvgld+4V7ljC}fSd`4a5h=s#NLS6gNRpUzGc&1zhb9%f<#-eJEl175B5t33|_ zcZ2jG)&qQ&uN!~0;GmXxip^Jm?@8d9xl4F5I4FcZC!rHgb@%5c+rAo4!&lI3_~-VV zy@azn8aHweO*Cjuq%jVgThO#OJM;+cg?3jI1c&Qv2h(Q9$KbMtt(1Qqw8ZzCCyb}v zN@E^o(&cxecR3l@s?h~yo0k6f?dJDac6AM*@6W8G$DxT>n7X>o;N9T52FjK%O+yZT za_D%9R~}xkm=Qkg`nK-X#Jr*zxUWwkhQmPSEGi3rblYeA5y~BW8d^$0Ec3R?9%4(pjMo(FiH4m|0t1rITQqMfX;TzSf z{Nf*bo^11utFL<3)!jWqFFzU8BjW9&&*3xf>JxQlypF~X>O$6aRakk!RN{?fKyO2> z&c#|AqMePrYowjT<WBu>C7EsSp`jozCf`=oI0>i_= zQbT((kxyz@S8BIxKYXxjLiKimdknQr(kZe;NMHjv%>Xcp+ zBinDa-`ldCxfAU<+vPj>!pL@Tw;s7(j$DsG(_v`ZjBC(mBXZrQNyqoD#*v*5JMw%e z@_cDjp07eiX}#5!=eeN^XrnF9KXtM^UrqUc_)B%j-|_TO0vhj4?#@zdt5m+r0OrBKI}~0$hjC?iu=97$ULA`+SN_`brD#05 zE@Wnp@|Sn6{Ow8Ry0Z3C_)O&-{j(!$3vF3Dl6D!KInR={=K=FyJabJieB+q2dF?SP zXwL-vJ+iP%X|BZjP&zrXnnoVfi|YX6`{==QuHBwsG+`P>iS{zq|m{HcS- z=iPopW~v`cIeXOY$1P4jik9@}_Gh}Grwh8*Lxs=1@;wB1UcJF9+YCOcV%PCpokw)F z%z-C{=Umv<+cf6$&{#9?TzKl3$9dO-1(r}tFhkEy zUD?FJaPe?b{EPm!Uvna7&KF;|*PbgluK6h24;($UfBZ+;evPyJY=8Pk*?!>Ysr?=6 zYU`wKQbIdet1x|DXw&?QR;(QRjZ?4UeHgkcw`p84 zS6sT&Ik<-LoPRF8UlV7)G5qI8x6#-##_D99_E~g}O|o-Lc63F_3$5XsbnG3joYEPF z%dr(2Sa;g6KH`D3n@*w8wGPZKk7~T}jsZT-wRU0uEAL!*6UmDq$O{X(D;~{S7<@Dn zS$pU2V$b8f(*uWiq|&KhU88u0d) zK3n=x|Jffw{Vsj>r@!7?bNc+*8-4!MrIk4P{F77PM?s%wxpnr^-!Gg-dp|1r9MBK? zc<40X?Js?Pt{?S({UfN~rO*EK*A4$h{k_rWO)jm((dVa}`aTN!ydA%ozU1ebr_tVz ziayI+9QDFa%|AYU`O%kt@%?_(|H?;Dze}I}>95*Zhj%ljmu!$Oy)Vh!LHqDkpqwXQjJdN+ABOdVNdZ1>JrS?deXmd ziL1BVsrOX$?^CJoin#iwo<@DIC(!q$e{V-OE-^mw7s!u`I2vi8HA{}K{J5u1@BD6T zT%ULA-|OwUX}tdJ%Huf+>UZZG<6YShCyxd>b)CLEinTwjV18pkSJL0l=%LQTKeBmR z_8jm0G2b)C5j_2!FSz`4Au;Uxl7IW0x=&}mVAE6V{7U?!`rS6C_XCDa31H~vr-Xesmf#JFF1ucnoCd1dYDhT^RuJ~M@N z$U9<-Xl%)>C*F?P^Q?2?o4|)`O(o9x^Y08W`?0$B$ji1LAm5VO_PF@YFAv_lrJYI4=%{Q(2#TJ%Rqv zlm2o?)$x^YzHjdvi?PdrhhwW_b@p!gw*K_>m*Z|+*{Y~)6>tC5sq3TAU;ej;I=4G@ ze%SiU`@iXhzMS!6T^w%vW1n^TJrRAy>o0%m2Mm=7VE8Eb?Yk|#;Df&ZY8)IN1;2g0 z4;Vfh2Sb1O?f)jH_!SG?xo8+6T&py%* zSau|UrCYu^a{k}ZXXo_;hP4S`I30X^nEG2qT4Md}Z2VND%jNpz&w>7c-sb9W_!u4T zkN!5pY41mSqdA+Thtn=6)PKqDUOZjVFa&YY_|N9?e`BgtWx%~T1vHaJKJuvDYXXb014PbmP zFTrO4|HdZ#0jGkm7GefP&nnS*8`{SqzpB*wEY9&U^DM{jEk^%?uA1;aa($=2n4oW6 zAHF}kxG%STo^~9M&c_ZqbCsC+it(?G+Fx?mBgb4bugdc8E@$75pFKYAzPMuTE1|tP z_}m2e$*wQ}!OmXBs?veDG5;duX_p@$FA|hhMGPA09uy)Oaacn&fHc z?>(@4;ZKkA7Xg0b<7W1AL0j^H%jM_#fNK6MUtQvyc8=ud?w7J!`-bNn4?OZm6CCcJ zs}IhJo8QVFoXbz}(AV9#ir-%rZt?8t11@)eStfg3vcd6U@K(z2f5^9%Jyj+KO(o}E z<#3K2*8(eG?`veAPj)r?9QkFl*D;$u6rMFk$SybpKVV#uiIZwdm4=2e^;Qct1G!qdZBzk^$p4$;>$B6g_q%rfrB3@a9=Y(h8~%%~CnQn6pZff)CzMiO3BFF&_RLV= z#G%*9Ih$+=?NPj1_JSK*knA~vUG3}k1M}PU&a#r1?|N(DfnA(OwsiKaWA|GrHC>jk z1K*@+<@jEvc6B|c^BHGiGjGKAQ~SoH@9(5ujaM$*|LMS;$ezT)9+iho*K?5TwJPr?(K-=6)BMjypVRTxE~bnybx$U5@w?W%H14}$0M2y5NBVZk zzg*>=h<^V7V~hOx9;95w-ZACyyohv_yH+s0W!dn91N3S3NCr2{cdYmX@ow+~)EFSV z>Dk@WK6x1NCsnUS+;5dZSq2B>i7BHkwv5TdkWv20BdjB(eDC&2m^w`z#N#pL9E-)J z@@K2QY8z(XT5Dgn;^U;n_N`lv#h>J*dwyDbfqbIR?w0wDw=oAQhi{9K=g@^4S4iVc z1V3^4i4MxNroA2{MmhVJn(y}IHNn#fxJd8tUrziZ zuCp(*rlrp=pZ~Kd_wS!fEYfwv8ck0R=1u;CO*34V6um?TK?00!8)xdFP4Nq1;ceQF_ZRA(J^s=kes*+vj7fLORIJ?QAu09JSz6FbHlVw<(r)5;byU(HJVTF3A?+h>xe2@9b-~UC+SNK$ z`<{Jfe~xsZ<^{ey(Rm{>GD1JhV82i``-T4Ve)i^QU+Pr!V-F6S7v$%)iiXgK6o+y& zG0B?Q=kv5^j(PAQ=-men*Ffv^^EAPyUo$Y)0;6Es20seM`M}uBb2BiSebB(T0~mD< z{Wf5H3A|_zP$PZgh3(Tlz-D;y3LCCbz}4zsu=&;5qb61ZTW@fD;WXhG(oZ;UBMzS6 z5Im(0EQ$K;!an*e2|3YRk+{F!#xEvee^ogB<>1BbugU5!@Z4l{)xaL@1J(X6jT?#h zD@fnHHl6*@{LqovQ+7e9)WlZuhkm*6kC z|9%i%?{)Q&WhQ=jtUe-LL^_FAACW%L0xow##~K&haibYtl8*8aI%EXC(VqSQx=GVE zD^G1z%&3W+%V{Z=(zak;1!Iv9`d6F_>0mqP7mYnLC|A59yS3g$xSvUyiEH7c>0WvM zm**!f>$xTTVr3fls;Mu{6zHMUs6s=7+{FiCs zYHJ*=r9G38zYn2XG$Vf`!%~sMZOEz2MZ|2ROpT=tzzT_5hN{_F6D9Lowg&TFWltmVa5tsp4>kK41;;Y< zv8C{DGBSJ@`yEr}%$jJ@k#F~tp2Aq~wyRNnLLJTUmSB>dCBnVpIxL~Ty)cx)&#lzo zO#La;Z{mg_YgI-QZPghHDnsRMr5uf4ic_RLPnRuq=acprzC*q#@Xt2q`Xbj~nS1FG z{`wcVSKLj-x%xWzOD5jtHfT(ARSXWFAZ;dfEtx%X;ytvPY=wCn(#d zk9#(6wU5~I^Jx4B=3CEdERkJJcAY2KqbeKT2+kZ*3<}x80`N(L zW5022JIXhdJUV}#^@tVa#H4FtOlyYzTAcZ~&Tjwa#~j>MIkp>vS2tFCtUYI_1NR?1 z@Xj&kyF~4&vW<$4MY}GJKfxFnAJddJ$HX+P!)E>}dVm?<(Jk6V&y1gIb$!;>wGR=a zw3NOn_y9W&?a+C4?%cwWdr_ykoGLteO2zj|$mqphSp$g_Pde*d1S;*%5iFHgbFB)KKM!G*{B zO@CnRPaaoR%T~OJ{mZVL{*k9Y(wU$9Yw*eT5z*nu0mKano`XIa+-%}CW7lg?e=#?V z67|#?Lj#FwL9Robb)dhQC~zA4glW zkzwkeMac8)*M^w>0kCrqcjF8#%@r%$ldHd5iQ6efqPoNlI>IjS2ZV(<-J<9a*LC zBn$t_ys$;|O8Z1>x3RXmkyzn1^i>wLMa7C5am~Ymr zj58@qx`ekJrT^n8){&{N>QWmQGieiZjZ3elO(Vg@k_p5eqfLt^*ll9Wdo~$A>ssbp z+pvRaPA=Ngc_$_|7PuM}u&!9b*ptj&LX@$=GQrb*nS(Xer_TMh9)E?D86Ih#he-`kmEM(Q? zf>ZVh*+p-M_9es9c}5Ox%AOJkX)gYoKNIT&ys8|@S(SSg>qJYyvFetMv4d~H!*^=0 z3?67{Ub3Hg*<9fw;NZf=Ni(sXl|P5L)46H%Kk}0`&ONm!Dju!1Dju<{z)*+MY??6^V@0--e;O4Ss+NZYvf%SONy`DF)wxalQ*%?;n2G&|e59TaEWa|^0 z6}Jxg3cW+m%r$^Tcp1j`wr(XI*YSmQ`mc&UsA8Q)I1}CYp&MgwhHfGb-AFbegDzJ; zR+aejqH6+y`r9(B$HIIUO^eerfHx7JBX^J;nW2Anb*|jum1BSTtdTor?$v`FbH8ol z+XiRc&(=D`c6;3K?c1o1nQZF-4jzR63DCV_R)3DGXzC)a!izgbHHgcW@rhq$ zqcd_5TC;U1e|_nfz$vh){%(D>5ZEsccFG2%nE1=^gO!~~_v3Vr?r3aM#4Xsyn!rVj zBipL6XY#yp1@Fe9Q{o%24V|!_tLAgqRpz|`4BOtd=2?2j^LBil#e3Ui%jbR@w)`xv zE*<>`bd>G)=#MuKH#F$dZ8>&;a%j=a1LJUD=)-BJ>03wMjoWa1<=WI^&ztw^BiinL zzX;jU$edI@X430BJG)kn#E0DIGw5!{{)DX1yQ91-Abv!cvAZFMcg8n-~Bjh3rt8iA%}YCpqHArIh?luC(&Tb;}8@Klq`Ill$D} zpI+WOZ57-v19zF=M&FQ5YH;De-uW+JmtN+9y;QI>{)?_365QMiZu*XMK7DnWz}NuI zn0Wg1gT`Mo7WI=}ZCVV}UoPJ!+BSu&feBga>TkbW5#{ZH3G}xkexHednm$VWJ*=C* z!?)S(+kq?$AVaGed)KfIr}1Coj9_h;OW%=i3S+OA7XIr@qbvMkJL@A2bAx97{#v#z z1H}IU_&Wz=t$epak9`=n>;YJ(BEyyx+gW3LbL6_H130nN5}bskBsxB)wDY>Ep=ehnu>HR z=|3({&VxQmqq2Qd_WVzG^JE$PSO7o9*wFp zLixzT#`Q^tE+hxO|7ZCD*mLtlwP`oJ@+!8eE@!N~ znl@?NGO@sbVx?X0rZCrH%_Fm3wgSbJxAn$A{h$8Wxl;KmIS)y5Eg$(U-fKRY&$y4REU$vO zBywz;g$`vghim{I+eTp9hE)G`@5lPKmH=BRuzB-IpOYRp&B4pB!JEaOJ3o&<4@&b& zGwyo~+OR6vXXV=Q@nwNq<*?1bE`N&Xw zTPa_-j%j-w&boD5<`BZkU=OV*wxaB^!j1l1{EWjM9ZxUaID$^iIGYjN{4#JESqfaG zSK2zt40IH1_Ln}6AC2ZSGl&gj&Z?ehbvE+NLFB05mmchcR}7uerY}GrCCu$K*D9W3 zm!yzTgIl*fcOZ#?&FIHMvSU%n}E`~?hZzi9uf*mz_3i+9TJ zzW)YzJ&C`FucR@%#LHjc*5xms!(Tz4&0l`r`S?clsoZqqUks?BG0>qlJ`#DuLFSeF#Q4v4QG^T}rm7&{@ZVg8c*ymI*d3FM9c zT=VWZKXcrj%u97u-WhzCL|*ApdM>97J&Pal^Q{S(?*cWPjZx#z(j44hquAm8OVKB3 zhw>;be*)_eJCj0x@-z3k8yr}`L)q)o>^X2I^S$Qj&RQ<{+_l^#NzC1OZf2ZpL=HEX zbJhU*RckTyXr-97T&?p;$J1Jqy*86lKZX7M)vPzac%N+-Z7#Isx$N}~8Z-H(h5W62 z7p6R|eY}aUpKPd+a>^oqW!c1MEotebCPlmoozucQ%JB)Plnf%a;n;)Iaw6&-9 zzDqsvbbyagsKvo?0kT|e)p#Dkmf+(0v$RuvZ`Saky9oBEowG;HZ=jFW2QBoiq4jj> zh5q_^@S|YR8Bg$q>E~jnpMRSGhS45cY;<5~Ko1Y2Yl{}OuHwQX{qH;n9_bcl?Fc=i z0UgRqj}gkW=e&N_>Lk~-HkVvq&`UoEX8jV;=jsG-_eP(G?~SF;9Agi;XLDe7j!mER zUug<50vnjo-@)5;=pH@Q|6d8}Pl&fmz!~$GM0nc+OkN%~_Locghqrebb9&-y@!Jns zAJ@6oveB2qbIIsDvfplJy{nn`8))w?)@ocFw?G>kpp8~!fa@1Ckv6vQ+~UZo7UUG) zguMLd$|kQY_?`#Xs!KE>+6dnYT}8vDga*}xqe z>DBD<60FQKIosT}14oD>Sn0RzB6^odrX(fkC-In9jv0C4;jfuazwCmqkYC+=CE4b- zL2&klyF;`k9(Q9{2W&IEood#QwN|e=R4hNG)?Y>aE%0|6IBNxWZk?^d2tk;N?(nKKOGziUp>j~-+#YnI%|M`#b27| zH?nRa{rds*1*88}*|wD+vZkWw_{#6{3p#O&u(fZ>w)~-acb%N}wFGImB}gkvkoJEP zqx>mMs zgZ^n#GPrL74@-mioZX5&g8DNp>&PdtMQLnq1-Ff~vGET0jCMxgI~SM2?eM_+E7^P4 z*;J0*n)ljcWoY6K@P&+TOs&{F@*j-z#k3Du*aY7h`#kkFQ~<*Q@+9IxrysSD3 zB|+LB6Qu1YO}KOA?pN98B|oYs{}r`eS^UX<)qTFvB`2S66kbaIJ#1ks!2VC)iVuYc z{0fX7zE<|A8rn)e_v|CHhT`bE0sQfO$UvPBQ*8M<)?bXwDA!px$i0V=36IkLrPvJy zLI+Oye)y-Wf^&2#&ZtuxnhWV?WUX{n$;BO%r}N*i0d_1#o<^XrMr4cYGmieWN%n=> zd*B!7BYcP6LtD^S1F~0h`WAF{GpA?nFPo`sDDtte?as)4?#(#w4;Wdm{OU8=7ws5Y z+Vh=SzkLWkiN*5Iw*4Z#ZNFjd@s+aG?m%ZS^A`E_aj$w5C%~Nvq5mW&*?(ZK8vsW;vb2qFCC_ufnQa?XKl|%%U@VRBT?AM) z9z>v}2)rJFo{T=g_l=%58vU2{eVe{rL%%GduQ$RcI&(hE{Tljq2V-U}efz5E+tj)F zJbU}W+wMSxjVNN}8je{Un-M)=NN!}jk{ zo$^x{OJ5cP>nA9yD!_UIb8Ep}1&&~5Xx`Sp0tPSA5xl%Hh4JmzHyvO38-B>J z7Qe#^?ZDs#>=uO(IG4~1DsJTGcKLEZKjdYKz25+l9c#G-J$m1N zU~U6uHVoBt0P`VhfT3y8rN-L^;I{YbB+Fi9VEzU$YmHB^wLzmAd*1{`_S)3H$#)&l zwbr>y7vkX6g ze+`bt2XW(qY#${5>+?eQ|AV*>Im|gO33jeds?0m+#KQ6UtV;z`qkmYvoA)O4<8Th= zQjzyP-g)OJKLdAWj>5Z<=$J3D*FwG?h1|<;ld-Z>WxIZgX`$mQbN$#=c$Y)l;9p~p z(lYnEZTsLEC+kePJBReu&E$>o-5=! z1>1@r|4wiF_ENrSAMZ^2@bApQwh_V4tB~uIiG0Jln)jQX?l-(M->|M$IEB1H^0F4@ z&AY|T%R7^o^{*UYWF76fLfT9{&*qyo0Q)<$hDM+Mx6|$<-m%Z5Q*kXu((a^TTnBRn z4U)K8s$M9(7^joq-8|#zM+h9(V0f$*eScLw7H9r*IE-Z>rnWYx_dgr+tvW)#?Y%7 zYoTfRfexijlbM$_A%E_-7OnEJ7M_j%Im)VB6(&zAv~?S{!c=H#BkO0W&`uin_MG09 zUw+ktU$e@PUf04H(VTvKrJwS(HXs@3rCV38e!#)SOAa2~_P4X2DqcRQ&AE(m8q=z| zKFL^@jE|6f_1v*8z*r|)g8uM}Ah^-`WzJx0+9)fqhyAL1?q?rHD)!e=;8Au;CJ`Q)6f|{jI#N&r=KeTaZQK?;ELGSB>k9v{kU^89n-P)m6wk_7cW^*(%o4 zhoh-iHg5Si)#Bq+i@%k=l@6>uuF_@Iu9uO$srY!c;^WoDntZF{S}M>E*S<>^F6JZ4UWd`qKVk@A~;H!`FgYx+}gU@QLfkYTE3s+anvq!+&Iej{u=bn zm)`OCYh>AS*_FAUc-ZmRK(GBA0mog4CK?{{!6$k4_yju%j z%jc@VWd<|FRcSI;8wd^z73!I#Ph2d5>>s&RCxLnfE-)pCwm5 zEyOb`XHQCbIzDoY%enmAw4L~=O8*P+(|7rdMbhm8{Peto@hE2s&-~1Lo_nIR0nie@ zn+fU6FFs6m-a$u(2DvV#JtuwxeFmI&@U?lzbun#wk9TX?Q*3BA*LlZvG41>t@3y;i zJ;8a$bun#eu4T0GPrTE3Cw@-Ex1ztbIV5b!CA~7 zHMXO-g|uI(h5XIv6W#M9#&4CAnPnZhma#d7vvSelI<5iNsfyLccN=Hlc5DrHfiz@; z{JOIPi-r~i77s00SaKj}`RYdjuk6r@qtuR#z@iP(8?uK@%{{rovG@XJ3{`c%K*XOBqBVEmQ?)}A%Q78yr-+A=COqZ91eIA3i@?`Si= zmbTAvrm0_RM}hhgx(7bl9M0ojedX#ijm(4O7b;y_ z=CubHoeJ2c2e!gPZQ#DWFh5Urw%p2s-Ajhucc6teuwB^4Uzt5~%hH8cA7~}*HReZY zzDe5Czophn-Mw_-%_jYI(#QEO35_AW*-GABz3@v1s>%N<>8iHLn{SnsKgGQ$562Q;q+3GjxNUY|pND?>kVPwQH4V=aA()QaxeKD$O-C zml#2Px$Hk2jc)()kU)KPAgR9PMDc+Kn0M?7q}JPgI40DdWz}T*d`EJiQ_VTrK2Cgx zMVE)tiQ}+5_w`jBLxY>Y=1bAsCN!CR*L-Tts&U$jk^RQ1H9U{wxz$RqKUB#a6?h}i zn!Qipp3UVC*>r8jf9UlYkB;d(ap!@N+pXRC+kLx}nA7H4zTG~4BauUqOyW9Zu02|2 zB`x=d*X;L)w;e6Fe9LE4ChcCFm9)E{M1A4g-Po{he`7<#(E{H2FZl6({{{bR=cBI| zPbk=3Fu~53RC53Rq>>FslS+2+Ycct&&a2#Cb>7;eRp)Kz7mj{+F5jJNfA>WF^8HWL z*B>qT;ve|EXTB?}F4$dGowOS{)J+Sbhjrv#N8Wt$=94$uvX+mm&flG1?IQ-Mwfl5o z=x-ZO1#ct4ec^WB?y~JkyU`mCfajCiBlH{v3IAEAzYQ*e{E1yGAkcdl~cpY{ntc za5i#WyrB8mU}&-FYYKGmQe5LtXEcPuk{$}ZKK|RFR*7B&rP{6ubS#@C-sW|#N&&o zx2fC;Ei(0{)=O5XUdfsBXq)DkJ1_nGfmcox9jF~K|A27v%89RXf9HYqmY@6Exu0#{ zk0^1{t|#qF&i$9TzuCV3$&3T#;pF;Jz~Y~ueN6NB$<(K@yOFjR&M)0kzHr6?!Lvm7 z!U=Fl4oRNaGSA)<+Dd=6Rt^nm-C|DeYpdSGZhjqj#umexzwgLcWQo?~(z5rgs^B?| z=UrA>{kH9Pf9{rSF=Hn7x(5@FnUXbcFNW4qp*3jOo!?&Xkv(4dZO<>F`cpn}3>Fog zSos9|m?F%*V`7!I<%Uv{o%krOj=@|~b?$ZQ)ciu_hyC}O7?8TxpF3W(8-M<2{08}J zHchoUw`I|X%pHrl&P6Acy=FT&)>=i)2#qEFKY$aqC0Wxt&>yZppZSw7&{2j?s&$Vh z)<0~U&wzUMe=GgOIwyVxvOW3fkAPY4Q@nr%&g7Fmb_e!MyMJtbi8ZMeT0@6t4`&Cj z3}xJI)}*3o)|H`@__XblLO#;^(qDw5r3w1QtFt`s>7)O5_>x^SF}`x*@Rh|LI0s*Q zg|BUMuyc6Rz}GM1)47gl?p#7fjPyRAxQtOwmd$@Pj7e(0-kE(dl6By&%H>0Nb0>P8)_boa zZ5#IB0&LG(GhXD(snox(LNDHVRfKj+Kh)SJe9o9Z!!!5LdVHY%V|>#+{~&(^9D3(Y zMyCXxfaXm>>j-Nn(-y$*$;_QJXCWrzF{S$%4{lOD)MtD(m`m)OKb!jGBbQu%g1BGZ z^R(`IL<<4z)Yg%D>S$tagN=L7Hr0a-LG{d{zgwB(w2}VG{F%ox^ND>yU7vL7GJ7nU zyR46C8|wn951Xc$S9RCt+F|WFmwVdPh+I>BGp#@<#cA7y*|WOarZ&aT3)cwW(jk~% z=Eu(kX_w^Qua`qN&iz(%uRW~`H*Ycb+Pk@M^K+g&n?3hG=RStN;rW+l$ECsZFUF_A z^IJ&k56^!tj;^)N>y?o=I&|*J$8o*TbB#mKPoD-omnER*p*_&E>2v7WrZ3wDEgCCV zpTp0+^?5v<6*%-fi++!%vj}sPc$&89Of;QRKeZ2<_R?2@L(emz=Xm-mgq|;U=-Kcp z^gM(4eALci^XmA1(zKV4UUKNV1s;;!U?x1&1%Ih+iD>#w>gbuKWlLDNIo(6MDIR+D zd+sfBZ`1k*%ftgVt-sHGJWUUZr(b9~Ej|sJ_V<&fmsQ5f>Uf&I0h)H{+4f(E%4eM? zvKM~s)OltS>nl0F(3+$(JACxH zYlm+3z$3eqcYeQ`^9V{??fHs)(Atq%N-K$&b!In?#-%>&d>s$iYy9X_o3evrLR!)L*DG<)73j;`zc;HVFIwqwoe{rTrSb9Bdd zcOT79X*|k$0tvH~@z;(G`Co(2?VhWJ(JeLJFV3jkZ})Rn=-YQ5U%CFF2luZZ zz4j>UwlB!{Y=(tjblUbW2Ljd$?3c9f)3{G_?gw%|(77LCy`X*3lPz>~_-NG4jDdIh zj?89ER6Sbj)BZ8t%f~Diov0c=L(LtQQ@83f{@U2^6qlqCdu1~^M}Tr>p%+L#Yduiw z$(mP59vc{`vl$%%`_7(47QT_#{E9Ue!&8g0e4&{a+0W&`5uU`_HE=Y`*2#SYy~TxN z4lu|zS<6^4I?LLZJbU!9m^DDZJzpuLZyNFc&U5-mI*9t|4;3~&Yh6;hhW@(okFC{f zPu6;1F?OCkz?2J)Yr%Uhcwb{B&kaL6SxzjYm)|Ci4D%uNDf`lh!!|6WKG!;B5PjSt zW`&`hoi^JLKtjE;$|vIBV;5#EiA$)FGK**rz^2fJhhykk@9EW6paPkTiQ zSbr|C?s)_|-?0Mvqi|U85%ELPMfhg0o+tf9^y_WAi?h4ISt-8S!dbrJdmX;8SDeM@ zkb;rEEOYwOJ6C*&es%kEoZi<^%=|G74qe-0480lqf9!lxF>|!%DF8oITS+wmaA~hGSw|jXe!$%F$Wzd1q!9Xo z|3p76l>aJxMI2dTv^=mJ9;+$Frag7j(p9aD?b3rQp;yCC@TYA3zk@Hr@UZsyHBP2) z3*k@tS7o&_?yJsf{0ai}qv~A8_v(Y@3~buixLfFhKVxrK8JXnqmgANo7$U%2Zlw}i zo%rg+v$=}cl?Y za89t(vV6ouxArtbo52&6%PsOsPcN~onyYxHcBu~ao#1>$`^v$W>U(n2{i`Ok-e%FR zChC!tT>{si};Ef;>?j-LmzpJLFFxeGK`{_lgY@xt02Ffd=tKZa{vi zosE1mS$LXGIk!MRw>o}vetW-IqPA*`9c8_wR+1?6Kx zGtLRx_v^=mW@S3}kujm!W1Rb(v7y^94%RiD- zU9yrfA-uqPp%Qr;uHede8?NHYSj%1kepBnhOVb@+Kzm$b9A~a<>@=IQt*yl z#g^h@W#~q94y}wAz6J+7Uw}Tmcq@R$N0wU8ev$qN3=57+8IeEk!G*y`f6DK-{NCX= z)~bA=o$K(2A1j*v_AkF#Z4J+;w$d`j1}a||=S#VEoG*F6IA7|u)mC~&^^EGFualQ` zn2dvOe!uAWxi5aRmHe&BpQilDlItRSxf#}t9r zpA>)=F0YHi4~_RmJEC>b(VP6H|D?L;;7$)6K^ND;1Fu`-23+e)zwmYPR9gcwcI*1J z>Y?I=;j_PAG(2rfv_0VWTKJ&O`KFG%t$b6j>sOWz6+cY6wx;Od9uGeNF@oZ=-}lqM%x9AUkg99JKv}cwS1#?)J|A3RD7{DV{y^(S3GvTcdphzxLIh;on(PwBY?;McV_-_tl<*zk0lA z(v?p}`2^bOt33zbuPiz~%hNx|p1#_1eAA+$t*d?=Z4dn0S9`YpXmQc>r=NGMxG4s!Tjb7v*bK&Z%!QB$icwqWE`9k5(=tQ%thqn7iUnn|0 zesoRI^qC%=!q8@byva5le`zUsM_nj7KK`xhqSfE@=o9b_{YTz_O~+H9V}I_2qT@-a zj}UqdSP~=Ie~9nB71{<9!Ye6nZ@}IR`B9b1MyZ5F?ZaG-&V=aluncWLp%LXZu+Y!Ix}{Y z^}yKBI@Z?@xaoURI>X77vuRkUA@#n^zj4!l=alpEz|aoLf7MMtkP>4@)m-N=dsh|1 z!L^^b_M19vWbx-E{DE@9ny)jTf0O$L=IGfe$DU`e46#d+J7%*EpT*wl9hUXn6ST*U z%WuaFHftT&w$_qvnfU(lS1~qDd{weSMRV=>ui5)dx$^4|V-tyBk7&d$q5T|YPY!cr z?ay)d1um4YCpL%%>?T?t4LIu$vVGa>p<0{MyxMQu!m6A#d-juDlYe~Wo&2WqbL|Gg z$%;y zc7FNmvG#V&+g102PTj6OySqP(e|+>jlE$f&doj4V1p5SdoixH)aof+q7wy)1s=XHP zug`j)eY)%!(7u5i{w5sTz8PkZ^2Oi{9LpYD@aFNA5B%ry{a4R@=4jfxuN+NV*>*H- z;nl0tKKRO`X&+SXNe@20|C+lt9v!zRZMEeq`*AkEkniR)4^JIRK7Yk49NzDCwhQEeR__k zeyyLlyr!#L|G2pNY4?gOq+qDSFqvr~s?QRkky zebQ@xSd)8gPj-U(RM*M`?MaX8Z;^r(RQZxyYLjSP`w)qDLVpj4D_i9*qTHVLS0(K4 z%E`Ur>j>*RiTFr$&FTYRTF>+HuW8Tr#O?Xx$@P4oI&QG*_#n}KmIUqDvTj#wk6iI1SG;{L8h(&EEbSXI zG5j+_X|0_r7e&`@R%Cz6K7+`tmBydl;mAXGER644;r#G?TUPv!!vik;w;TG5uHTrw z8o-*W@s;uR=hU4@ z9^3Ut*JsRm24fiW80!j{4{J=_koV(cDVv03M7Ve5ZktvOdME!2?X3l08j=St)MYh4Jx>U(z`&l%qs3uF3p zXY9DAHr?ZC)9(_rr`_24W7oT7tMAJ6<#hp@&Tu@BzGSUvMY%O9WY#6oHzLqfgnSY7 zi3oXlk6*5@n~r1YS8E);`h%=(SnTC=Y53L_)@)f1=w7q^%4D>B-#Ylk^`$uK)FFQC z_9r(!3Q>P@t(A0L$8R!y1|7*~yA<79^dg@a`IEnf?-Bm3D+;!ArWXAd2CwozZsa@Z zgpGVF-*Q94l|J@mTSr<`*_$wlbfY(FFMm|0bajte>QnZJiyu{Qcv61ecJ@A5$0{BP zQ|9~Eu`cNL0Xn9Me?>pqc(B)F8k3nLeC6Db)?5pir!*E)R*i=jBEHb|f9YJQwiV## zSRj6YCgl5YI`ut%FMZwvEUrvY-#ksfh-?%parjMSa+U>}a9yYfj;^FCdfAGz}ABzfcDNV>VflPib7lW_EL2S>l^fqoL| z8$q2h-m|{H44jy{z@uyo?wqmfzfRWoZF@r;{ZxW;+8mE>)hYU!NBN9V&oM@=C^#|J zrXQV40^Ymg=w}x6lW>fxvSEuEpPV!=kL`D07LOh9;K=2#_PFuZrvKjYlkdRq(#4DH zWo}?^i0C3)=V?AQ$nkTrb?id?%gW==nA+*!mNvh592$VmG{3feyMvv_pg+YqG4p2V z!^UOY`0UWY<4zm*c-l9lHEEdSi2FA`8x}ek@jheTTRNUVBy7w{8CaD|BY-6Si+=Homd0KB05FMRP8` zt}*aN^_2x9I9s+Ay&eA!qnYxHL5dqY9;t*a+ak% zR=E0oaP#FAwoOg4@Q=V3Z_m00f8|8}mW`d9#kMEI;F+{W%w%zQ5&+ zA&&3mFnllZLz#^Kk@3AO!5690_Ptb{uJ5Jp9lPWT*(GIX91k3ksra5CJB=QLOpxED z^pF7M%)0AT2k?`0ExBfr5v zbM?~5e>pVLlRgt4uR&vMvB&>%CjBsqesKLSk+GX}mgI~G0#5hwY2zM?ica?Bk+E75Na?wo|4 z{5@Zf^Y@e<7%602XYL_iPVq{^9?qv>Ua6S-@&T!K#mn*?X+w4e zpxp@lKhyG=xLKo#n{_|F!;0f{)utt@6n{W|nUjgfP=YMAWxo8flk0C1%}`Ib-zMcK zwnOu9=3&&OGXzthkwes#&$o7bDS!R>e#hsNciwOHo$B2Qd`*S&9Zd$Nz@81jbQSlU z`)$s_6}++^>s#g9M)^-OZZpsXfM*l{)0>Dg3DX^_!~j^$ZwaPx+YIK8*hQGaMY4Z>YoO3GK18*KO?a z#GQ|b-;(N|cWNj#BitF1>mdLBk%7r^X_3`u2s-6e)dMzKdkuh<>aSs&EX?=gCC2tmCzX) zD04A+e@)%ta>|uA$oqU{6D)22OJB~ zw&TxOp^6Ht?)cxG>m@$>`7Zm~wKL^d^We%ynxC=qmf+*rl5y{5=y*@-;CnZ>a@`iT z*mu5k)w(R=A!IS`LaWe6M;m&+d|l=9DWA|;G0j`FuVEmz z#3tIj1U{LGes%nSQ-7h=S&2`m#a;}~pfY};=3YLc_=cK$`HA8mYVPGLif&-;@fTG& zC8`kgbe_ZV9X)HhXfe5d7<8Bf+(|AyIyAUS_(tYF<|DFYOt6Ll%1J?RDx9y{SK%99o;8by)n4MQf!QR$T+Mrm?#Lni(k>1~xp3mt5lfdJ3S0C1ls%eq&G~`URmOh84(gvp{EuBDmMl;45wn*3E%ZOO zm(FI!o5n*{o{w@h-bi=Qn%+~aBX?B<>biCYj)iZ*UNMZmyv2vVwN)3`9z51`39=d4 zD%b^g@|@8V1^eV-#7JQbZYFOn@dHK^(_|xYS6W9@FF$m%b!0L75ZMbvY?{#%*)JEW zBSy=(^Q)J4@eW(swDdV+Ccet~K6S)$annZyyU0Ij`%D}62eePPjF>O8c&~nHCYHe{&bav- zd0bq~1Q$9(d&#*r4rUIw>Z(0Bkj{i1VAuv?K$MSoa(TIxQoo+q(W>;j)kFJBR^cb5PiJ}zIluI{Ia#^ zzTI=*V(#rc5p%CG@7~Q}PoBS;`&9GopFQ_~^yK-ybFXjrac{;*M<#vyo3>2K@yH~1 zKA4^vJAW08#>eAe?)n1fYrK$xJWWN8rXfFNXC}rBI=A1^x$k)dKdzcuYhVrjwl%F* zMok(zc(!#$4e@676~n97q0=b7!r6SwImN`4PT}0x)cUuOr*Ff%f8*V!c!%wd`Yh_R zsL!H4>ji_;DqAnOaQ>Vv*P_=mc?#YBopD>n5tDl|`>JnePvV8|-FKikExqG);wTdz zB2>&it90F`+xORe%$j!XAl(n>z(zjpY3lpAm9qS>HEQ`gz9IE1@I@Y+G`Qv+;#=<> zbl-s^)G^`qJGbzz=C_nL;gW~<-+c4aNAPAScWhZOn2NqK``w;Yz^6+5T0-SX#a$RKRwzjJnPuw2Ryk6;1SL*1`Y@B zyZ3@K;ja4`crF2+I~;f}w&AGt6wVE>8{KU3fB=@CyytRppj;OP|m zu17rG{L0e(o1WRwJ3RS5gQrm%PfVQor20R}uf)V}PpbbP&;0>&uUPa+^?SMRNl%`K zPeHs+$og%!FNAa!S3b0x`CinXEct+4Xw!_0(!6S7`6+gae2``g|Ge=?d2x+4SKI^y9(kCdec0D~k+gY-4;^yxtb(AZMb>XlyyyyV20 zNoA}Y8tl}*w8%N={YT<|{ zZk(4AxUo7d>l?NFcJXWF_sx`nH?FAu#f?u+NcrkWYsgni)5e;-P3MmN>L$`(nUETj zzJ&C>q_<26#H4Q^{V?fECnU$Dze;*f<Eg6vAZ2v(Yj;W;Yp8ahbKJ-PoDG`JbBXlYQsnm{KAd*d2ihF z;kajwQ&#m8H$Ha%x*ONO)AgJ+e+{%@z38V0QrbiXT#e-|9o{qDN>cf`qa%bXJbZu^-y<&KlPJ?sDG2 z|A!a&HRlc1nfpR3R5sYTud>V@LX*A|J()d((`nCwooQx{Z0`N({0N5^M!ZE1BezoH_OS{29^u3NxbhRb@o$+nI54eO0GfU*Vu=eN}^^_3a!Kt2H0kv4x%13ueg3ne^%b5Kt*`2=lk1y)n)UgI zMe8dZ7Ok&pShT*K!%nVmy{T_+$-`6C&b_6N+Sjt0sWv^eCYft#N0J%WO5nZitan_G z@5e|#HXzYlq@Xj4u}Jr25vROuPI)_>^0qqVUF^5ZyTl(;-b8PC;|rtZjrSY4O5SnG z`}_Y#+`GU>RiF9)XJ&HcA}CZ;)G!1DL`4@0q}nn`KvcAJ3tii4Hwi%j(Napg;$klJ3^T)vm*{yg93`+VK@I^6c|blY3fwY_2K?JY`cZ&ARB3BhZUZ;Fe3m;8X{ZuGl% zKG(cUKYBh_NarCgNB`C;20%;!nWq}c_1a5o<+DDL)p%vQU1!#{mbmBd5i@9APW%8n z-iznSZ&ZEFcK&4h@8z$9zkU3vu3+j}pSzy!oQt)eSkH6V+Xsko9VC{adNtJF=iNNs zRsFVpjq0nNSjk?_-iZ>I--92~5;&u=jWwEmb1Zu=zIS3iHDHNn#3m8zC02UdIo8bY zFD^g&t2}4TYm@Jcj`%Pq7PTLLJzh#Z25g*m{O`@I->u5;TN@+hV)gOWp4UH>+Y+bm zQreK;DqS#Zs5Ns6&$r&FIvF{&Yx;!jrNl$qYPm*SDK-haWInJN%R8(KwGhK=zmi;C zeEhw0h^@+oxf7d+SY*!TVq5REzg5gRL&TBMGlHjLID!8vwqtJ;8+a%#cX81N}pTsb=5zy4zt#zG4vJO`E<_y32m6&NA*T}hZmHw*5tOMd~u!m zqqfxkFwR7n+dr#z8MURPqvpXU`PAcokbH+B@X5TnvyYB~M}}kTpdAi zwCCiv1scbl}n9 zNX9$N%B{TzU;VY-*2ef$@TFW+=s>jgvwtQg2>ryTI&s9gh6V#B4^;K)pqU+_nYf-) zj8^nL4O}m>?D%x-dlR>{8of5P4@0y!{!2VB`kzM4a@ITI)4RsSO`ho@)=%B`-&FhV z^J>?eXG|ZQCBA(se{Iy?SOZP!43TWw(;kQfwBhyZ?H4&m{eqvK_WNV?>ptJ(e!+2O zzsIIy_c5Os^VM9P*pMURP9^&kYZZ*kluLwMiy_l)=6kw7C%K+vFNtSUK6LV<{+qYn zaqVYhW2V+Sd;^i~$QiA53_-RlXJRO_OntWg>118v(NH8YpcnB-D_l8%nA{6~J}s_Y zXJVB4?p)JzIgteOn?;R=ieZlIUscYykfZVyhnW0jU-) zzLwrszSey5E%dz*c@iLZXUG|h6Pza&&^EFc-Lu}R&JD+j^PTvk+6xsm^sy#%cmIva zv_&2xeH9rvL;K1*`|Y7)b;+t;oB`Q8lB|SYXE09Wd;BK)0Y3?@sl8dq@7cg)De|)% z_z=J9P`p!ZJ}r4^`NM_Kh~(vJ@*p&yTak&+U^{%Cc}ONc&3igKqn^2_E@4HUWe>}v zCN*mw%j&J2KGx9mp3eQq()nHRPd#m0v>mVUhs*Ym#{}JIJk5iyqrS$p$gqJ{LcdCaAS&sESzoa>^GN@RVM`--{iGesjq!nZ;rg|tyH zw{W5}9|zB}53=VDo7l{!(Ff3$w{K5t-@)voFU64uX8nCE;~t~=`?k-*u1Y4zV}|z4 zr&<5uxdgf8jQL-SPu2yX;VO&VB&~alb;hdmo~dPRKXVTCtfO~rXd4r>Kpz>ws_BR>#PvNvGBH+>0k~ zPs!tWKl&OFvVTCkgws;-G-3$p@dfR#ptj9p&N`?1E@*u2;4x&EvwzaS*tO-O&`f7M z4BDy6=&!r|Ov>jwT&E zk%7ib#xmC3$U(+x_C~!IqOFj9-)3v)?qSx#2=VrnoQGP@=U~+roL^$=vtmJsqjPI= zj6Y;>XXfm0d@Y~c^7@O(=Ll_BOzq&wXIkT<(4k=DJm;Kkl>Xc|RbzOnqc57@w~AuG zG66m7`D6k<6hG=S*~jP?=X!Qga=WhcnLLLzKhZZit+ELU@gv9w{7%uyx{5yZpU4i! zrg9y8==wLgzNMEnJ|1AtNGWYp(#9*!r^cOliT?S`9>H+)3vRoW+2N7AR{%fsA?M1x zJ4yX#6(!d>pJorhTIRv;1iY4LaNkX2h0o>Pa?X&`T3&zf;3v;Zyqm1#{U@DI({~;3 z^1Hdo7jAx3?=$wwtZ=^Dcl@Q5Pxv`|F~IwY3FU^iLq*NVh?bi|w$39y z9WDI7xF_8bA02|m;i(z$zsAJ+X)`jW&6M>WU>;tEpb1SoKL$5nW6S8Ip@{TvUAp;=p{b!WL@ld?u66LEdlRTptonadVUu6 z$7<}Be&o*#32@!j^X2Gy&qg(J$mO9fasip}4fb$*cBtguhuET?e!iA`kM#Ul$^VeM zSNxugz2d4JE8dF&m+FlD;*O0nB)s(|*>&0zi`;^DSi|kyFRs1f%D46OPC9H3DGw5! zI`w{WN8WnxN@qE9?q)55IS76)m^I+oSa5($uJ-AdGxTN-Y$KV0!!36L$d+3eK4>bO#-~cTe zH~_zdV8VWO>F>#$8M(}wIKex!JgB}6uQ_K-=z11wKC;dI8MISPo<|Vc@w>F+%cPxB z2aes+rF(Cyb=M=yk^Q?ZYvG!K%bsXvU3De)2z%~lUaj*iiNVLnztB2s5?!&B@0sKK z9e8L)E}EQO=*0M);PPFrccI$>$R6mQUPM=_}fkJ#oT) z#7d-J_O8_#&(_8c`Y-eK?x;qGayI6`z5L$0mN?i8)DV51{b^g^%A$wxM^4nL3ljn|=ld}sW-QQ;eY zXB$66y1sc6bT9t}8}J@{GT96dl%A|>-lS{P%*FS55}vsQ8O9l+e0(IR6+}D2}5%&eCGv&g-_``|gADE?T??IJbbKec;GBM@YIcw{bGM zVCv^i7+bZwG5*wlKzLEMIM#=!4wVI6N1dL7R`#z8PySwBc*?#;BaUO2*lu zaWXc^+UTR;eg<&n`PN0y?JMw@iOE10*wzkDI%lRngszdVpJ4pTb4)OY#4CI6W)HJS5ynE6ca89n?8J$Wj6XhTM|3dg`OHov&YlUI^;o~)Ie(i+Nlv9U%V zpV~@w*0CcW??pa}&t$u_Qr}cI(tqD@vaXHyW5~)lva-FDbq}9)IDgIkzn9#+4Y~Oy za&rZF6_T3?+A2hD>X|6AQ!#nZ29ylnvs^evP8Mc`6KhvKk!;8*N+x)Z&**OM@m)Ti znFDem-jEaC^LcDibVSfMaeRU~8k>G9b473#m?tMoY^xNH8%Tg7ObYnl5b%}a7xv1j=c zfyP4R?lLaL63I0-dH$!vOG!FjDuK^R+I#^#qg%F-$KDQpHd(nww}|(ZW9;y6uAx~^ zmc6Ypmhm2PQ|o)udDIJT6g(ws+7>~_%+2!U)M}hII{#FW$pM zKD!Q|^BL@i`>-GG!+y9A`(Yb3t;p}KZ96gL(Y6!!Z=V9MwoPD+@U`S+ibqC-FJ*kn zd-led*;d;G=yD8WXN;^@*=>2bwYhzBw{PNni=o(dPP@cwOuL?r9uXeYgLX}u$lXHV za4&Tvx*K11MmyeET9I?aY;BEWwd^-+w9GO5<;Jda(>vb#x!9Gt^qo0ht=%hz^aOs< zR&XOb_f+lwk@h{fsa@HHnv>v3KKjnQ#UqY7?U9R}funomqtTZ`56*sdhsXSe$I|DM zem#?Ji!$i;E1o}B2b=v=bWeuaqhC2$NR{Q-68vbk@KQ&`T9;w!HGpiO)f=2_{`t1 z?+-uF2kc|ivk9K#)bK>G0@K!=n?0NAzQJ0y<|N zt*ℜ55o=+)vp7=}nvu+1i*3=y`IpaT@yXdIlD%=_>8;R6 z+a~L-W@x0PNpRIViIHn7mQ_P9{{)Pnl@|1;=mdN7F2(4yXY)vnFPy%Y=?wB96+ba; zaL!2}dY$VLz$LKvSoeHG-?F{x7NA!dcNy`L8<6>h#71T>^X*&)AFrHk zEgY3h&YYc<`W*phvso`N_gOO-rQc7vzw?|mgxNoKfA{A;bFZql7Dm|%Q`K%Qtk>re z))@480-n?7v)m_Nyy}PC=My}x*Z%19!DB3)r-zM7$K#4q@w8}lC-S{Je0p>33$2>D zrq6}Bnz1q`Zw$wG*>$T@{COz2%-f@FKTdh!(Qu(Hk1*hy6ByKkS76)p-hq zA+AH){m`2O*!SLDK0=e<#MAPF$ls={uJN>$@N5I^u7!tKgD|#r?%f;yxCz^u`a;zP zu7ko`nUi7?1@MH+ASlJG^{>E2v*0->ZQ(Y{}Hm~g8=vd^= zboB+ivvO-I$sedJXFs4XP*I#VGv`~8XNUPBt&4yM`3R}?Q`yn%Ta_({{iixAin~quOLAip8$Map zFTy%&IMxsd$HDV#;9P!&H9lFzx>;3Uv&IqS9&Lnw@#@LC`rh1|A^fnut+ia;t07OK zjI}G|LHlN&$pcTU=b9MQjo|bqWGw41ik-{XSTNyu9k@VdT^Md&z_=FihbEYO5|ou$o?@nN>-g2DaVgc-4I7N;B(-cwUqJ9 z!miZ6#CnwdH@BD*Fb+mSlc_WDO7Y|j%6R1M&2rPb7%(dF0t%` zRhE4STKF?#$?SKh*Y9BFSO85FLT@9my+(s~;wG)or13eRl|9hR-csWE@RD+9_w#4? z5_(Wxj0g9-h3|BE*zVx0yZ(RT_HWjN=s!XK&Gg?w|E=`jM*r>b$sYROOaC48zmNX+ zOEww0^VV6iSeH`{uJ*y@;#cP(zw(J&_aaBPH@Ulg8ZSW}KeWm{o6Q|_cVl|!{{c)U z0F#Q8KHL$kA8%-}yZ&$gKZZ%BPm`6w10m#vmG;kDZ*=PGr1GJXRpdVnw#>R*0J%b3 zYU5<=yp=aw3oUFW{Vw5mSvmHLbeF3q_U*)WNYT%s;L}45tYQP8&q2@#v}$5VY4|S$ zx2mgU=rQ1Pe6||)J^vSL_xL1U+rOc0+A;OHe3C_zsU;*E_fxceo_JLL7VpLwSDbN) zA7xjjkHhQxr|LW1E@Qpv)ATL;$oBaU_RdQ-)t>?%&e>wAdMV1ii4`K3M##>w!jF)X z)j&Oi_yDf=;~KV=u0O=}8s%*<4^RKfk8VZYD_$i%t5}tzlf7L3)cUVtDayN#A~R$C zfD7+NkJj&8 zzO(1Q3b~4YH+eBlzVI8boUE%x_9iN^iGYLRftAo{6}16Z$sbQ(I~4lEZ~ly&KJk)PZQZ)X#uZ$aPZoL?Hr8#X!8>yjyvKI5*Aw65$GXK>+Y)DA zf$}MCeR8s`-;#T>8!NG6wT{?G{HlF%<)6VdT2Yz2+Kw-h9USP`LSMR8ah(0U3qtmrjkaoyN|s~O7KCSk%QRZw5&T2S z&-GI?Q?}6yA>}zEd%w*4$rs?Y=lKhy{JwJO4ER6(ykNrhzSxe{{lh0;I$0OO z7tpm;y>yMZOHFq8DA)0W#w$LrT=rS+x|QtrAabKZHEONuM+BcJb~FCt;l9K@6pxw2 z_vg@kW=!NyU&@$t-T0EUhi$)p7QPj2j(d%K5&2d=6O+-JfD_l-Y~XQ`Tf54cqX(mN zY0tC0gT{VN#rIAJBk%ouy`Rp%D|!ZF?|u6n?DrJEtq5NA4-UVHUQUA}@BKY`KV=K7 z2wFov(fj|y`?22tX7oMr&Go^3PTgnewbS+gF?77**4+*rFJ(Qz(DC~Y9p46y z*XN)gT^j}s>H2VFknXW|L7qPH17ALeb=nyE-J>nb?Df5{TiUvZez&A(V@Ggi)+fTv zw>56JR;sbZp3VJ4@6Xcv)H8SEE{B3gSPwP4@i*f39S66;Toe4vB+jpO0NEDU z875BlB=pXE%_E?F-t}@K)AK2xBL9)SDztfKIQA>M4%>f6rFD+k+avnd-X465&uTro zh4v(e@0P+cf0}(C zuC5Pgj_xy(o#q*Q*z{+Py~zF$EA(!l7j3BxbbtE&54b-B`L6ZHVc2H(a{rbz+h7^C z!R&5qgXhRe9L9Kh`uwHZ!@$^j`uua)X2TduPoKXeP0nS5yQL|b+Yy}IyGJy)8GFfM zy~fikj_=adRypuC{E*VsJA!XnJ!;$Ihd*Y%e{&DthqmNTzggzk;KR~#4HV|^;u{Z` z^D9#M)P^_E!;0~myE$Tv- z0SD$DU0ox5rOW-RT%Yk&He|;86aU8dM}FG(q3d)X^Q;W~D5oMFKjaCsUm``{3FK=! zjT_o_@iPegM13h=Ecq5NO38&o!Kco5dDm*hmpas0leyHT=drBQu@}Sam4t7cxHdQi zH%?qTwBZrt^k#6Rvsfkvkm=xw{dVuxv&NbpFRr#G{kHnUhjwEFs?1>eB2nZ?44M*OyiClgp85)k1>rN3x2w-k(I#>ShQQa} z_+@`Mu}I5_)77{%O*@9is7d(@ZSl_dC~d@O&)GYa7Qgh~abbC1CM<9Fb;@V;7XZ&6 zIolwkjqi4CV@L3#ZR|Kz8=GYttNu4_6z6uX>rktDZ00yl*+!RmVvSXl4sYpzYcug$ zL+lM%*@B!<9pVSTaeBOQ59>cC`*k00H1R-}U#0}c=n|iE7rMw7hq#}^Sgxk z1znoOXFU|0+;x7=d(Js+IoQ5iu|3=9M{OIrbz^WQ&$QDPRo;YC$Ha-<$+4$J=p*2QtO*1hyWSg8A=s(JL*d#|$`@b`lR~y~N6HtEJQ_ttG-Cs{ zl!olLfxB|LOq*Qy+Kh6~tdsJo_FjipmvPqTO|&;M%U5g0I0OIp@b`A*RaiTFlUuTn zG0)3pk4V1pmshBsNQC&xH>r1j3HE0G+_|QnNT0bQk7~^FPvnQhn3LI`%beJBGswf& z=Iwtke0lY1($}`~Si=(TyqIf#7aZTe)C#k=W@i;S)0ZAe&fLekSKoYW0sNLtzWfd! zHhtf0joXROIktbkvHibzSNYNB`AbrVMz;Tc$uA%KooVY~zIWjjc;g=SET^x9enHRk z&+tRQsc)Tc>OG|Pi$~``Q}dy%1>n3;YwC`EUrx-dKFvOo{-v=_9+(eIi&JuYMetC{ zw{d07M|qs}daXm|2yccTJU@*6AM=gxu(6ZQMc>f67yH-Rz8PO3>#E4vGR5?AB^!y= z)aY8(hOcuiNPV**0bZon%}-AonKb=2`*_o7TJ;{%<@-EpHl0eo zC!p(Q_Q$p?0xx%3#KEDb_t-bP2$^vwa%%}P37OiAY;6H=N$GTN?J1c}t!aL%<^X%S zlkB&RC6Jf$Q5!hl8ChG+e#d5H4L;(=W@tpds@CUXrPu=Rv-aP+`jHP^KUMdQY}Ne$ z^t#&$H*x>Z+|R;CRqR>MH21?-Eo5&%KkLN=`;{x@74~Jl3mW40i;{WNHrc2( zF=%u}WoTqLAsgjA`s7+-Iy5^Pn}O%!(A%*}f4CeQO!cz14ENiwjq)oFxncpd_71g0 zE{E2JKx^*=OYAYw8hhC6cZ&GlkG$yteE&zjWA}{0?s;b(-+|NSpx=J|^M3mm#eTb~ z$Zv1^ir;?YYkqs{Kl<%A@9^8(7y9jA&hy(3aXJQa-R7RK`p#&Z|r`5xo>uHWv1?;#p+_`VNu zpuSz?b)vGgs43RJDEb~bYsfswy>y*@x_3Q&GYtAqvuSq(v;4;X(YyGNSH&{iUt=d)yU7#au_5!E#lkDQ=2G$b?JN8o>w5R${ z(i`4CM}Ii`UH5#K^{qh0e%Cg1NC!IPpk;qCmzZ4ES_HN+jy`uZ=_%2q0YRK{*4DaLiRi7Q>Utieyf-#eBJ`TH^0X^ z?jpvvi#6Q^tnI#o96%)(tFC?Gbylzj{LhLn3tohp-id-o%)?1PC0?k0cLp}U>z72Tap`^1lSpNs!M`v;H_ z2Z7T8pH&pEf~RWWtzGoF1UyQny$2rWqi5!@F1($!YGmEs1*~tsCp_zLA~y zXWJ`*n_?n52SYZd<5O{NZrXQK>joS9y|b*xO8bK|;I&0|FR<&hQGuVaQ3XfIV8*UI zzzx%m*Ih~O=eysWcMxF(ny-DC@hlPBRyzyYJyul@Zr}n)(v;)T86Hm+LI|E)$ zx0%Jy>3L!O8xMEM4Lh^TnB&L<<>l``kenGHFQ3@BsSPt3n?<;IHt6^&@(Z+&Pc&8K z=H-9Yg|&D6@FgFmRbbBAg)dye+F3fx$N!yRQcSE_bU6f_Z0eH$8#7kPO!<=|^TM0{ zaE$t1@XnLyTGy}s!5ip>Y4}*g(nW(>N34Jb4?Ia6k$De=2KD_iXpprL+mRz)oWb!& z9U4*Yu+OCtA2i}??1NrX?wymjmy-c+XCDOjy4?qHJ@9U34yN|bE5N%leILZ2$PXPD zJNqC~@}P4cgh$hY_5VJdCJ)@U(&d5wOlSY+>U{Q9M2ieRTu!ckea_LJM(KAmvLeoV zrO1maS6%=wLcWgHg4DkG&U?bkTW4C2HBf^w98<>#@nilnx-%jqF0ccgikgypdgey0UA%WLM2~X|fBrc(Q8=`+8;f z8QI0$^zIMIZ8^yI)5y&gI)SukjbeNs@31C+CcY@i+@YW9TR{Q(#)>L?_ z&hXX_#jIZ125(JucuO$q!y?} z6Rfw4Hu9@Syp@yATW(vKyj6S`Iq?sNBJqB?{7vOAM|tPtqr?fF{Fs79!#$kAk zc0Lua@t%0?FtlAN+QvpPw7n>ewy{|ZZS$E<+izW)M%#>|TiT|5y&J25w)uWq+J>Jx z?Iq}ZFY>Sh*^o}#Ut`V5&^B!e2bJB!eZBQ-SC^*g>Sr4y)BbjOkluf3H~k}z4p}?S z3a=veny$C*|2y^H6W%&gyj2QsJ(}XJ=vH{Ev>V=f)Zwk~fFtG6|GjxDh&`ZK#l@_f zY-5iwIa=fAB$MOLhc1%UfrWF4$0xBJ?#HH_!x`Bj>}#Drle7X0qq@g4^?df%wP;}B z8$6fX99U@S9=|v9TRwgAP+%cEGBbHJVD^K?u_0q1D|v!;*!Ly8Nxx`6p=>Y<+%@$g z526~sJ*5v1S-&0?UX`Is&iy;#u5}$anSoz4-7@t?nr}4g{n$B|hZhS!GaUTHZzSK4 zb$ZFbS?Gy-WM4ebeuZ@oKlK%^_$^%Nn&JhrHGJe|9GD)m&lrf^%bNdzJ6(QWlH%vY zdv1Jr8)MTe>fNHP6WBHU4gCu2bJ>k@k<|cgg3Bj`7)f z-zm297Z)2}B`;=$M-k(lL|pcUiJW5u{kQk?*}X0eMf!ZUIC944OUUg!o7~Ppkpr8~ ziS#;`G2)BHUg7g4at!-hhkb>dL*zrY=i@UD)%iqx|BQ9GSl{#c{$1;EEXeo1)N10) zo2qYEhez^zD0W+kdAyKmj$~uP&;&gqRUar!>@C^hbAtNdF&d+8|E3!zSS+Vg*bf6|{%R$UND z{z?A$XClcLG}Zx;kiaW8l=zMaLZtQ2Er``eB!US6#DE$ zOm9}f8J-Y0!_ zU%%hZI|g0!A`W$i-_FhP+xchv?K95v+c{_X?IA(y#k>!IJA3={P8Qq9n-}wXm6$p! z^Gse!XoIJNQ!rEU2RSHea4u7YbS}57u&0z^4qKH{f6$meUJys92xM;4@>Ne zfP4K9{dQn5H4rB8-S|hBK+76mLF0<4$P(JATH&{wClQ~0%xOz_1Wx1r;I=jQp%O#G ze-GIyP?^lf0;M!T`Me0F?CSH9Thw%Jboeep%~ z4WHdW>_W9}QgAA0{9MMIe?yz?wCVeIpY31ewCUB)|GC>{{*Q`n)z4R4Hh&G@p*8DA zK07M{jXvkI18aPC>j3q|o~nK+UM!%m7&zD3C2N2?=9f)j?a0@mn3VZ_(-d%8pz|cn za|8G^^5ZiZ@?$D^oCY4JGq1uF9#5`ky@h8V;XTQV^uG53zw6%19UtV+|5&lzYelin z9SHY&sMzj6=IulF?FY{CV@j}n_k+iMR}!b|?K7~JZg?8JYOU}~;8S$5`Y{JS9!=cl z!sl7q6wK=%O39**-o>;>AK-s4bf?@H-wGeP#aH_))*64sTH|!z8MXVeEqi~yqwj8q zK7I^+%%t7X@cH%gzIpT#+MPD4nsrJ!!y(S)TV?3OfqxCkg|9_+ZRjIi0 zir}}rJ>T7rTe4d9N4vJ)-8IFPaxQR2g_|GK=me%|~k(0xJUY=>9Z z&#Rh(jLz%WLe4L|x_;i0)cqKJ#_3 zXs&q5sU_?BQ7aT9_}8mlzWRgVtIj>=kC9^`d+S-+F?{uFbg9|9>BgyYk%5w{zw_Dm zW2;5_=hX(`qda1x*ljz!_4<3TGnc^wb3 zZa3|IX;T(x(_)hG{F zv}14;Jq50sPmilX-Qa2>9zdFOYC0Z68k!2fA8;?K-0ZCW+CrQF3VVG^I`tXd4;|3kMfDf z8M(hha{t{ap4>P3yOTcmk>_CO)79Vg=x^1tRh=%+caJ`d4LX)w<*?trHoI5tC}jR9 zWd3cw{95#oUE;|69jbBHh|GV_ZS%Hy^NwDO%rBoe>gX2cuQ_buGt+Lp<|OwjksqoJ zv0L)vd}n`QBFD&&3C}xuV5UY~@;qm4iF>gc?rR;1wcmf?na!*rx3iA)?1#C;Shz16 ziF+fshd)07Sz}~dKWKMpN+zDx z_FXACS^O7#;yVR1>O5%94S}%aNYJ>i7~zzCG^RaK%uH4;lJ(i}_cY)6 zPMsOy!n398x|wf$Iet(rzMuT68mGogGc{(+88zjqZIV^H8oV|*@uD4y7p+E?E~Bj= zu=8{7m*~OASzo1^H+Jc(g(vF##JH6I<*j*{y=A{E3j2S1tZoZ6$kTD$I>*8Bqv${7 zhQ&r(_FQmW-UojN`#v^0WS_G+Sy!CLZ*-CG9B3H3%s)`_!{7L|-qvG<#FwM!ua-Go zY2q1|&P#~Y-5*}RXfC+&s)PyOn-%0KbOxn!>pgLw@wTQ2NfTz^mncUpZOmry1*+w6&Rsc zO)Z>c@6I+~N^kQ<+9VbfML$AY8-4J-;tJ9|TFX~0o|a0+s#6aR#`|1$VuX^T;jbplm~5NY?3}iNDpHVd8HO zR1trx>EAW}HVA!R;J0Ig{dUysf6NO{-NGIwVn4A#C3fsMFc{2dlFtG@u}@-6j$JRG z>O0S&n=8ntN)oHUzG=ahZY?Obo1o>_U0Rlm68?6QFEbokZh}^Xzh4mRjMA1D%lJHP zm45?20X*sKD02_|$cOm4`uRHj%%Gnef$4O&A7Z7i6DLillb5`HSQ|F|5JQd9j@M72 z`uQi$?PHv;ZT5$CZz=oqw8vSt1UU=X^drJ+IJaaCXSc59d20r`l;?+jHDtf`rK2y& z{~G#L%P#9<9lrgGW>519?Xy0*8Q86~YrWg7dZzuxXYqliG5(dvI`jj+2{DF( zM#;C?tgrIB>%QqM>nBT*L+=x7D1V{YKKcFQb(7FT$h?_8o}cu(wa|jsJRSLa`t|k` zZPB{-RZhOTue-WjHT{h*>$SarwlyxnP;ni`^4VtQYQ_dEC3p4v-kY7hgf%w{KW9af zPr7i$mlKZO{xNeb0e`!79~(}%#ZJ?`t=y}g#6I%`G+HN}7%=`?bT_p7D6#G*?K!MU3P z#Gh#Eue5d0rRj0d^h86`W`C!6$M4a!&Ic9mTn_AcKE}Io-o0mEvaaF^>YLCXYx0KH zWoMs=pRDt8FC89m_-Ph>TbCiru@$sdt6ak84iG=%ed}vJ+xHFQ8+D#PxtF@<=tVof z!ik@0Z({x(PW;SYLEP*XVrcV-q21xL@4MJ~(VUYdnW_0UIepHf&&?y4ZyEEgWWG1K z^Bu{2FX=kpt@LMP#V~ii!V+rt9e!b?+xi#_yG^UzB_=oF|bpzl# z(Z%)Ty0Q=H2xlV`r|kdX!7oHUth9DM2yKbRlL5c|=fX3>N8p3>`vJG#*dpe0C$zr= z7*_H9IDD0auXr!^SK=bH+y5Z3a{8+Nk^|FQkSFAikNYw(4MIQn&U9qR(wUYW?+gBj ziIpN#tB`GXOvZMAUbf)BTnrp7)t}(^Dt=$Y?|SW#qwRZneiL#xR>C?V@tEVKdEu%m zE0Wwz&LPjo3iHSb_JwPZ<%gKRh0PM*U0hUB%zlnQU-RtoQunjQ`8?9ltIM3HcraPX zyto%#=d(wxJ60E+9(XxQ`yMT*i86mMI+iF|Fhh68eVMes}RzIEw?%v@Dh z4twv;`y1ZdS3gY8+oyVTu>x=L}-xo&Nc(~QS^ zH}EOm6?}~CoiT?)|4lw~Y8;1xGxI-o4i2nUW7>@iXSDe~ZN_q)SU7V#6#S}N_v18k zJJtB!_|)S&HQrvJ&84Y$%Z}i{z8^bh5B5)MJn1-pJ-f&6KF+(cw?!||HS|*0<9F}V zyD2;oH+SRvr-7+A<~w;emU{P4aAMEjo%=Vu`!&6rE(3n%?%nOdn6LXA-o27{D^hS? zAAHu`58H!&FZmSjChERIou_}I&eM&NLie23k;AC-g&kIiu1>ICRdXh>MAo&BH#oY% z$m|;Q0=8&jBf5n3i0A-f?}IqwdvLMch&|E8UaZM4GWOz?tOsH*9>*@M)ET{%`Qcjz z^VyIeo;R9vc^~~uq~*!8BYV~jiX+ueB?WB;sXT=qmawN_oZ=g33V{mtuE#@4-x z?~tO$6~Wme$J>+8oqm{`IrcX`e(cU#J-ExzIbGrf z#0qZU{E75ffqmTB|COw~yyIHxaHOAK>eWv?bR0XIGaOXYvbj)o(E=T=-)Zcfs~B%Z zO4dyG!EEaP`NF)8PJ=a}7lnPVQ3G3H-GS59QZ`Csu3GyaVI{ANG-O7@x! zvdsR_N^-m_>#qz~p4feV6aHh$oNajp+q{}3L!#HsTDjjyO2dwWp=e_XUU8=7CX zIV*gxmH62jcpwjdz036zW%uj)y(b3W|4K6X;xDYD<7s=nwO%#WoV_P&uR30*P>f}L zE*u@qdWh~i1-KuCtk|*0d*7xpBrC2d5gtBaA_LduRGX_K^lFywZPW z$13W$A1SoL1O3?knd2iKn_iE>;BS*vEN3oA{t4C9{a7?Bd}j@R0sn^Zz#U|0Q$mgg^2BkHDYYB*DLoxjC{R1-}yBKg|1G z;784Q4}KqTPaI(4dreM$-7l|Iy~K?CI&>F%DD5QrCz&K4jJV88zed-UVng$KuzGY@ zamr{V@Jrx3ZQ^@PcKEd~A2a9L2C3aAy;e@lt^ygLxYSNubMNlTr+3u;&9B3+p7L_Y%P5?!Q<|nmtM%4D)R_2 z=VIpY*Yx?4t0mtrIyPi~&%hCV!1phvJ+tLp^#2$Sz9YjiH|`vhZ~8EOp-*me`zjs; zeCex#7<)OjEu+@apJpBX=Uv%{9)5TDUw;0<8sFv*mQnXyF=fU0iOs!O@Ez;hA!2{C zh);ZJ7xS9r?p?J;gcDO~e;PJWAu;AV`Cb|bPj~y;au)ve7{-e({=#q3b^SF?av=G> z3OriEQFVTJ+bigD;$_9;tp-^;24>_iQU`$NOkEG;+RqrzEkl;RJKU=OV3TF73sLW8 z$^~h^SudO7cYd?qd0hzkqr8w}V&p+!si-{}sPkxo zeZY1MT3FfxJbU(D9GZ<|{~3A&k0tb(AiptAZX>n3b{N|Wd$7+2>SDUQK|BK6WJoyf z(%vmS=r3sYANigAN9wbHK9D(a>?6;ftT*y#9I?>6@S&fi`h(Vi?@+D-U*>mh5Bf`w zcbGW_e74iR>fYytf0x$2nU59hoX-_KY2Sr~2lHV)cxP}Ao;j-r&v+f>JCH0CE%VMc zXiRqaJzD?zEAQZ|irCG2yjV?D#gx zn7@X!hHPqt=pE$({qk6HT$TIIp2be>j-BWN4|bkh{%H^9ocT_q2k&U!n)4jwdmC~q z&UZii#I&Xp<4<~`hW(+Nktcdq&$OS9?LUkd`4HfR{Mrj#_HMUW3*|hCQqC>pQ+D7* z$P{l(9{nxr!I(VV@}r+)i@G^$KE;Q!E;O>sYZHF!qFe6hLEBDVQf|k^;I$VyOAlh> zwZbFDUL&R-{gYy@=SRNYgSIvUn=QagbferKox9vZo=(94*4gD-05AE8nQfQ$plu@y zjO_V0BYWhpDkhpPdv>l*$sR{G4DFN+7x$nKqxaDBo8CpwlgksTXYcPE=}2p*Cp!Sgv}w`E*;9SmrqpsTy%jy*#ytu;b4KZfm7UL(a31Po8PAz~ zfQQZs$7mzHAFl7P_Kgd#%D8UEb^Exm)=tx(=Q?)Qx#3vGb&c1Id^=+t z?sGeWEBbam_ZrX5&3LX}IYtvFge!;%r1$UfSo(QZqg-61=Po?u;zhX8KjYs7obzY) zP;co-AmjX*<(3_*#ZMrX#d`LkTdtg9K5v~g#m;s26fKKdVa|@)nM6MAhtKXG7K%K0 zNpa+%aV5kMd$FIWH~WeDMDE9)ZeKtSChLp^zMPI?{H;AVN>&#%E=M=Tr_t9EWE*z% z?X;6MkvQB{#Qeaq6NAy7w5-PW`dN=**G`DBPNsELFZcY_+r0gtIqU_!n!jvkKPY~r zw-=?JGwSx-8L}_VFEMA$kIF9LeJ97|TE<|$E0+C~Kl~aob+Z;nA1z#4^Q{ZRty~+? ziIXhD3oW!)zmIs~A?%MM#dalY1eKSJ1&>u5PW6((XrA?44?v zdL_es*j3r&(&f`XI;DmDpH}>(cqxCpmktZ{v61%U=(v^{tPkTSseh;KK;wJRW-8X^ zZ+w~GmDqk&oR#_}c^MbqRe5wBX8=@@m#e(yS=<-ReVflN^p706jXmX(Lmus4%`=_! zUh16h5#jvHll{7$_o}nvD%+J$XU_4khl+0T zdzGVj3;7$Y#lBe2`N(G8-|8Rk%l&$uU4`yljW5$c9C|Y69?$wbXC51vGXDg%Ah2IO zSX5|@Y2fF=adY;y;8W0etKfzoEFV?R#ju$JwB2lS!Ll2Z$sEea<)-|<3Cu&bH8N5$ zq*>>i*s!zr8vH0eVQBIiD_nt`i8rWTjFU6<&dL*Yt26T6Vj9!^7l$V^4&vkkS8+ee zHH<4wX7}fx6&f0B{5sxf!@f3<%l(k%kZHhlq)vrRs%xwL*u$uYThL`Wi#nGjZfG zu-#Gl>mtKnim9ufw_s%S!rTGjNsQ4MlcO^%$#SlD%6Q~~beK0*jq#7{)As1)eRpi4 z7w;K8pGiOSF8L_^6mu4(Xh>%V4?V$oq~L*i#l$0rn0V#D&w*$7OgTv_@3$5fBj>~m zYv42dlI>fm6SI{%F}GTr2|WS&WnOQI#`K-%bRN)I{5Jf`ck!6kJ37w^a`7;V_VVVA znh4$QY!BpY&-<|IQTBd3lAwm0XhV)et&F3M@5yiZ76uv1MLf@Mx6TjevJ;O~ zUX+h`!!`8##%9{RAhfGsIOj?dQ+#8Y&-r}|=S~vK+M4Ctx%K?euHCb_&+p_i&Xpwg zoqULMne@4Wb0_ur80Svvlk+B{id{U#eSLBcWw}0o%zb@ssrA{8?vnZ#^R!`q-)()~x+{l1b~6Ue;XYuVGxyes`d!c1n~^0(UL#jjcf4Ro;5Xvo#7#P9 zw2F14tj3@J;Y3{lIc!6C{^R;pEKM<~I5H!F%r~{&K4z@d*nYzq<0i|2tzo>Q+=$N9gL}Lp>H8LpScMd z51|jX&Yd^W@WUn6!jS0ba?9+YpOa^eOAdp^MLQF%g$ei}Nu8c&`3yocqL&h$Z-rJ` z%=}ZdG7VZWXJ2nhf9R_e1XQs59o4@A#bi@FVpS zyNtPQhOekyw3bHS7c%Cy8O2;D*N!m{XUw^ChfiGF13LdWTKTy9(um8^6($ zKJh@j{bk3n|h3?+7|_l*~mN9w{^~okRHxy^!2qi ziXR0J_FoRnW*tYg4gLIN3pl&i&_Z^{J>Wn#funz&bz?_%To~Rq(b4s$1`)88E`Nsh z=Je00jS-L9%HD&Uhz-vn9<>%6zyBpChBasItw%>N&+@q!b>?ON4%qN6wIgym8qNgo zqb++mzeifx+w<51CHmib;(-Tq&?jX&*OhU#!>6s-+RoUVIGQ)Mdd6nXOT?aRckM&v z)JGZT9>#cE|NL6c6}LySe@msG(8(h_Ecra3bQY9=SuX!rSO{D*2Wh2qgi~7UPouzt=PY+ z4XApmDSGavA5WgS{U)hRx|emjH>mTfdZh61Onj%X&IfU7o@Jxca~oHmflQQrjIGd& zTx@|4Td~J=ewx7twm=O0vbG=h*BynIO`#%Z|HUpIZxhv==X= zU0|SiM09k>{@ufN9nVbAGg?QDj-Z{vA)W(vwDI~b&MN1g_}an0WBVnknd7O>zQfuQ`7H5Dr>DC_NuF6Xu`c7a>sG(oNmW)$hw_z5d(AlGiMx=PB)HR;Mp78 z_jBm#S?xv2Wo*5HS-+6|tu$Kl=;?5&qYvj9n(FEg8()|+gXItFyXV_n#(C5=)U=Ml z+rnY|MyIwG_Ug-3$QK{-zWFBTQFIi@>a=l6ytP*D85#xNPHlj!#&c=2lsJ7ed?en# zgf?0xgAoIyBoh`el@yCIYIZJA7aQl)sg%@ zI!Q86>%`PzHgWwa*bA-D-!JCga&#Me7!&9}$y(9cvUg9^)qj|4}aRG`3Su;~i>S?6Pdm2imv~ zI@=Gu?SrPEfsNn)MyTjN&kQ~Cj_t$qjZJ+X0M5IqUnKhL zmlLwjHZ<1Q*>YfjlWnvtMp^Gs&R#^2#)Zc#kR)3#l%o^)Kcw8_z6BwY3 zkfkXY-1IlV;2`7ExDLB;Nczs~co07U*oU7aDw<)Xv5t%pQw|4A-#4CxK0iL zd)`$H#MGeNB%L+sx8fhphCaN7e8cEpbH^u>)MQGscO%K(4e%5LPyaqBO)e~erxx)C z?qcBX|4YAjC)bw%e|Run#pg6WYs{X*9oFKr!o|ddy!|3_YP@yHd+0%18&1;}YpWZ*vx4^`zb=NSromIUL*L(rr{=>`2WGHG z1fE*wd&uxq2XwcWzXSY3Id#T}#z8&kJL-ps#v(SbMKZZZ%p9VJj z#}J#P-h=wpdGzY{bbPF_8QTM#$Zio&XwIs|N4`ywYA<>7eq8gW{dh7|bhqiFp3jt7oV2Ul}}7nc}~nrr(>x_4)j1?Q8Mp!ak$3Q+s?1d&qBl zo;@|cW$HW6HrLrnZP0rA2=oVXXD>MKKo*MT_oFkU7pJrTI^D)fx7}L{&;!s`J92jq z@?$Ub(1FYo9qmUZkLxZzQRw7tp6ojN=t# z4fjfkG0JwIymy?n`z_c~HSBHl4HNz5;5P>vA9#p7G~w!bXc@WZ?YnGg z3fcX|Q(TwaTFrV!9NDGXLY?xfGAsOecknFDHhSk7{D8IOBedaHw3Ba_PrltK^6etz z+vQQ$MD{pw>Bb1Yf@;Vm$+_DCFRkB6ToqZ~2Ji17UZ^qXtn)VD-U{5a%c*1dRchYd zOfEd{s}75Dfj(GIUF9>Z39UT;i`&^BJ-gU`eJ=SPW0)K8mdt*YXRz%jXnt#ei{f+MoESrH;~9E>CC^V1Ui-n9 z)LCqx?b49_-e->2{Q`S=4L(sz6?|92pXA-7ki8@Q-g2&^+v;UY54<2;!8qax^q{+r z^!Xp1sMA+d@N(W(8OA3Vm0*9-F)BG^QV~a zlj(EHnD1lh_fqpcWc`@4&p2N)eZIaP&Uf3|Zsd^5H^zpm9vUV$6o2~?cWup~pBTI< zyVA4c6ZkD!_qhVPsE}U(tV^qJD5~VS(q%UkCGd&kz~0+Sp3b|^oMsGOJ2m)Xr5njN z_8)yI3z<@d|FsvIRhw~ao$h49iqo{)MJKrPRqVW(xP1ZfbDd+pjr@;xC)Z&8JSW#+ z{cYqLbifODFwf_Ru~Ns2m||XspE--(+oT(iXXIy^_~J0^VC4kf!FkA?w)lKxFmfd! zn*u%-o#zuDC5G8qS7|Bv5{fV0w%DnwG-vUsql&M#<4gUP&t~Mm@_z+?t?9SY#ukkq zxt_poP`m0+ekyfG&Dd^bKH~4KjQBc_-6VSy+iE>I|Ka8Z4&SQwO?#6s+5JMFZP-Ii+u7g zdXaz8JJPQFi#fDABcvRT_49zM>bCimhcVr`b{jC)7!~_Ay4%H*>M;oBZz8)!QwvWu z@x1zCQH`MtpQjQzJ11nn$#2z}_0AHz-x$G+0R&HYf4A_3fpWqb@$Zo^LKGp1%CzperEl@jef9M4=-_HK;5pL zCO-ih?YKM?**~#3axhd9>3AVOvj3G{k%LXWBOSn?gZ?_`uZo&>>Te(Y38$KK?7fvw z2(H6vQ?xnB>TUY|{#_ME&){#judnI5nfvNXXQQ6Vc93mwO$l-*m-c%Dn|{#A0DQzj z(8*w8)aZUE|Hjexqed$KrmMb>t+VXOyxWY-oWlDp$j_^3ueFSL9=hqFflmFT&8ox0 z?=3@}c&W@WE{{D)9)#B8U+)7Se{VwzeTXmCvJb5d{m~BJ z{Stm0wd#s6e%+z#mW~Z~KqLFGX^nkPAAe#Gly!mSoVgirWS#NT-0_a3zt!HpHdj6y zonsAi>b)49{SA9Cxxxa~;hJU@>4+ypG9A`6#G9&&yTx~MPpOyyTAzezGJ!L?!3 z^vb+u^&!u}I>Y$dUwsCf$%EZYVE0vE^Q9E*#-+pVW?)y&{n;tlUGh=b9jcr?D*QA# zA*No#RBSf-YootT*eyLZ?4Aa8Lk4t(UFP@&%lL9=uNFFb7#eyMT14I?kTuQ7nHFSB z=NyqS1dz-kI zfZk`N)BDdxg_XzCg3N23#+oiP-bT%;PMpQA2sd&*&UHTjN9cV5IIGUiz?tNa=yV3- zo@CiSc#e7_id{^1$9p^D`Bn*Qe~J&s+RM2z zW2J6XJ95wP=9V=2f?v~lvkzn4jE&qyce}jh#Ie)(6`3j+bA+GU50LO{U#$nU4JQZ#Yg=m`A71LpR?~2jJ+HWpX1xQ z`f(|`LcU5I8(jf1JI&CS_JzJS%9^C)huNRDiMR=LVGXi2=G2FZww`TG@DXRaNZ*y; z0W4xQ+2PASaPo*-dHz<$u+@c&Y@qviKi$UAGrCqu`%!RJ&mMEr4$nGcQl1lM1JUky z@{j8|cThg-YTA|ms%Njtu^z*QelbB?*ugIf{^tn~;(P9w!^0K8VHV@~#@3T{iAULM zzRSnnb1QOuhGq7kCkEuX*R=mUJCYc1P2~8a&NB)4qe3`RedTKE3|0EVwfKPn@=Y38 zx3pBJggstg{7>`*a?|zMD$c>opGT8)@aa7_z5z5qz?XapL#JXFxZ|YO8uc_-y*qIDE7jg^%<- z1~0|nrPy@%09jz>xX5YGtKB4e`Ul2!Dn9x++WWZsict(^-|4KuSKIomWY1~54%|Np zZ2Lg-)t8534O!t^pVs(ASNH_fWvi?XV!v31-qUH)qvL;rCeyE5@Oy!k-*GXvrnhfJ za1_0za`7B~saO7=rCMdudj*XHE=QjL7Xv4*>3QQ%ED707Lbsl zKUKW%pVIYJjkDf*E&E_k_B*?C4PW+p9K6QBUjUym3!hQ)p=AtfnD}_|8J>nNj2#dB zWuLEvHg3RgZXOIhjDZf=6Eu5ZapczXOCqt}Ig#c;xsld_yhsIpk91ABTT|kq=Th>> z$DZxz2>O+5sPTopXAZO9D>?$27;tr@W$@HUTj4eAH?ktx=lCMIL#PLWUJ6hTM70a> zCs)kIpHz%s>9OQYtxf8h_6vySiP;-nn4llgyL2OK!}ye09ev41)A)xIvur6uPL0HW z;4eu|Q#tf0T_If@7>3T_&x>DvZ8>z|!Xd#pn}JVDA#zoG44vjsr#NRO_VkyKLH(_b zZ!n&`;!DE$AMSo*Z~xHCqkZ^bv@2O;c)P|~JHFo6`@x6EZ+a+bg)g0V`_W5$eQG!P zay!c5b$rZlU+qyFjoc&Npd6%K(5LBlv}B!S=C(W=xRJNKioRDf2FZqU)}`u+%_JD( zZ17tNewA;Lj!$nMFT)$c`_sbvAnju}uU6{te3A*UTYUM;@8{h|Q973%SDdu@ zVdNk*raUEc&U4Zq(V5RjntWaAsQs0^AyZopoq)byv21g%uyWesrO1P+jB66#iLu7v zee_Q_J{`J6k7EOdmCq3;H%9uRrHt>T+@~GwQ;;lBUdFS(O4f<@Vgq<@0(%7EYd`$1 zJmO=FB}%MUbe))hZ53kA$N+E-tx9LzPHZLqD7F%|Rt$On_{X*tI&E9xhbh>4f``7Z zL3fDnZs9XIU$E*MiPKMf4s(HL63Bx%*Aol)4Q(xRCblEfD`6u2wk@Ujb?BU4tQ#`$(IGx2mC*I@aM?7$; z;yqpSBfMC|WX3RsF<;F*rZT2$7|XTH=Q`j!8kw>VUXq``m-)5ci5~+lTbO4Xb8H9K z%}bawaEO7=lHSY%+)d79PT;?^pYX?gjV{b;yotP&DpwD)HSMs}Rh3{}* z;-l4W&59b@-ICu~cW!;~>C4PI$`*G&v1%d~^FFqQ>3bE44`;i57DJ0c^@mN8K7J!R zDyhMeYdyx;HpUk?W2=N#G&Z$W&ODg+XQMnPd#o38$)nvMbCFLbTU_^s^oB0Duel8C zXK6n!c2`D!JQrIIoTt+kywD7Pv`h=xeTcoaGXBkOpI%OHEA`&h&+GIvjDEJz<|i5F zAKmvq@i^x&AJzy1%q7M+wFm5&c%5;Jk3Y#6Uj39~5ImZJ)k@;5)SR5TlDK7%IA)yj z0NZeEEPfMwX%WLuFs|)CIab%qy%@Z>5}S3_0Bff3cMJHt1^g)nBRUK-9(|V|m4hvz z{r=gDhaGKMf(}Losb)NOb$BgvFGK#nMm#Et-0O_5JGSo%*}nhG`NqzB4z1iuU(Y~y zi|IqOzM3|M(tbU7_0}t+WzHC%0hh03<^2DooeO+b)s_FxedNuvK|n>5kU)5-*wG3p zwz)}A6g1jGN2fYNAVEMX)}iw+pd*0pop4SDz(~@+7^_DaH9w~W1k~8 zxd{Q{3&qOiVgBFWKIbGim!MAPem>`O&OML4*Is+Q_S$Q&9W2IfF2-&yu~LG_L~OI} zG(6KOtBE}FqTvX6l9>&;sXeLfZco1xNfCu^Cg@>CcSJxWzrGN*; zj3{=Ed|^3AP`w?G%jEvD3k~ODB=rZ2Y9q zCS9u^Uj}yJUi{fje&7rLrQYNG9OtaWu>$h`%jBJkm-j2N@g?82o^7OwrA&yA&h-|| z&>S8((b!#OB1esn!o+{0tD9(lV(ja5@?__%9~Js(HTEuTY|bq%5FR4nZ3E>EH~KcY zowBEjepLQb=!b*wmu##Qd=pw%h1?{T#Er?YW0r-#8I)g4`BCU!wz<|c{QDn>UCDyB zsY_#SjZYK5wb1vP7gAi}CU{BXfGDv=vi%PM*DJJbF)%El-ethBRcpIs6O4v7;PGg# zW$f5i%9rd-EWgu^!_L^|oA#F0_iV*BW9|9W<;1aTKg>HRn{_@R#fP`%i{_O7P4g5< z_4~cdQ;Y_tT<9L2Y0U>N+9MlXbAb`$NfSEuNHz8t>(|nlJ3GSqwnEmn)ltV0zkUCD zuC)%Wh`L%&b-dW2wPx6oiSXLv`{er3iS$x#(O(zW?_`}*C%u#h3jV{j2Yt3ZIEVZ} zpY|QlH6EM=PoUF_=*vO}kKndo9r3X1>DMZ7?c#TWb3X!%QEY%F>;f}~!}oTd#^Kjz z+cKJ`lpao+T`_K}gG-Icv_4Gn^-F}myTKp2uN9mbf8Z=|h8%gSW8HzRR`Oi80`_GnMx3d*V`PJA8Z zjle!Hue3t!uQBfL;HtBX4W7_fFYAKM%ZxtW zgr-}1ZrgR|Xj_Zc9^?`W;LK-y*J+znxe<4nmy_r3b6nC(w;`r;qMgB3b~ zJXN_Smdk-(x!-yklMIWezrOTqCOY4-OD*ST`_rBA$%UfF%uafgzo`k@ia=A%*qmK` zoPv+rrln-?e$b&MtwVBY={*nqg?`c6c8$-C?1#3L^ZFTR>7bQrXi4iwwRUv&Iq=<$ zmdZV$0smK8qHmrChc0aLk;o>HY{obCqS~l6J{8E=faMRmvRFRU0{XBC9qNADBAJX0 z@FcSD4l`c2W%6>qKZHz@Zj^sebg%q0i?Kl_L8}vImv_pt+2E{;Ru$jmt6%5fPI|Lb zc2SS&t^;=+R+_7F+vp-}ged$b|4=vm;?!M^ ze#rqJ2eJ8H2TwE5FBRyPmw|Jdzq?*pk6sx@9DGVYPq2;ml8A*zKeR7FuE>uaEeA%U z89R{mP3=kr8)ghneM>Qse_#V{}7+jzCN;hTt0RBPV0JX`^8uP zvyvWcj7-tX#r0!ZGv28;9KIds@NK+5VWr{MuKkweOORi+sme0@XpBZj=b|UH24^~W zAy#mIF0{A!d#r&)PGNV1Ca@+|JWI}bW4p`H%eY4SjF7 z>H83RvYGyup4He*K7wpBPeATq`uz5v$@gs4Z(}TKvFgR`IXSk{6=G_2SMT$e02zXVkhs9C^ngQ zw&Jh33qP{f@3nx-2f*bLa9Kv3($&5BRIl0o{ug>udO@-%QLIY_z9 zrvG^t9*>rgQ-m>wbvorLo7cPUZ>5~TDSWl2o$+zQj?blsev7K#L__K~aGKa}@BQ@L z{q~RY@I{x@}2tV5m$kV)8|``_ih{Or?+&q~Jz%vg9QIbTxRr*o~k#1rxD z?A8yN&@Xs&*Hs>7jd6UR#K&L#n6hm7Y0t-yGXs0)yv{z-{Q8XG1T*(v{_q5oZ>2ck z4V44qV)#X}=i_oKbR1b3zn@1M{Uo1k3$i*3yJ$MNiev_2P4uBHKkRviBH~9XQr-Qj z?0ag9Xz8~O+=@T6;(jhY?V;}Ye79a;n#TTXGr*x2J-D6tzVb>>Xa>H8@_rWn2hoW$ zk6&e(zEu4a-1?n+ZvAVW`lS=&s(04U+xFr0!vsTjeIs2Xof8ex4-eB1RsMoQYmm>6 z;g?*8pK=2}%cqIAcow>be+`ZcCm0!0h`xRq_|3iXbTaG1@M}l`HYG+bwev^(B_6Z>Kx`K7)RLgF4omd9x8A z6aR(GY1#l!J{_>viXiVah853r$zWf7xruAKJk)sCr)zvE){B_83OS~{L;A#46H%AsctUgn@nj>nNPrMCuc?ZYAxH`3-NBU(*(K24*y$C4 zSRt}#I&>kPF?4j(1Y>uKK9{((HTm;tD@@(Xfb&`owNJ20r_U(E|5#}q%XVpfnoYNz zW#86S_Ih~qY`ogMaL%%~#g%=xU3QLJ*W9>1%y4q+Ikr%zTy_4tJknD;K1MD~ryYfs z=a^z>BYhn`;e4BeoJ7y=$3C*FFjXw-(uQGKw6K9$UwFS!K0gPqd=*1kbs7 z*PWv)V?M78TR++0*NlD9i}5(?Q9o?A3O;L&|J!5Fp*OY%_>OT{E%6KWhjKyE9jDvO=P0+D{>#uehrca0`6ddZKcz9mt-NZJ10JQZ(e&%dfWZ_bUgQC#}kV{Ew#I&HZu_A`?%$jY;hG5(y|L~QXZBGrhJQeVGhMX-#Rk1ggkW3$TZt$?yL6_ zo2;=x6uHoXd@#OF%3H9-$wgOX+t03jRqV7OalZ}9|G=6$Hsm2EpZLmc&k?hKLsxr7 zVKYqYZLdoo#hRabbh*yhGgpmb?N2(oY@1mRo*l}GUrXe`G3vWTW*u<4XT)c}cLVdH z*jM6(FmZSi4EfW~KGgmzV8^>N3o57mC ztojl+H%1&z)W5QOPk*JWZ(pWg-S)3!Z2D>6uJ$*L$IdKN>=Cv&cCQy%>BH{bL5xo| zcJDH5-m2NmM?I()2q^LMUCJ)i!50$a=EiGT(o(1PYi zo9X-JLeBKRrMxgL+Rs#L-H3N^*1CPW(y`+c+T@N6cg3|!>p-+dMs~H@cg$z6|2UH0 z^ZJjL-(Wv)$9{g*$jHri1*>R0q)8Hw5PLZdv z^?_0G`0tj0tLNY$?EBW2Y?uZ6p{-VIEA+~kbZVO48cVtR%6?N6TmM~q-__~JEDye^ z!X4xe;o5zk&GQI$cGI)q|2gdKU%_9$!9IA2G3r+I@OJd@tE{Chp)K^0d{=t^ffV*f zp^ps@RA$-o#D)7C*0EQZZ$1SKbK3BIFwR^`JIw!RAU1O0T_^UirXm6kE2_?{F}|rO zR?JJDb=7ZdeBf!7&mWl9TZL~mO=KRCxGVV-B}eS|T07pelINcVb60*_?0#TR>1)04 zKJjUrM?ZRC-@>{NJ3Rg73O{pLcRJEDobh*BdlRuk2a{OuW9#P4a~ztu6PjRbiod82 zvN{c!mRdii!--vW+JDy%mK}KLyTABwtEcbiCG->$ zdx{!APMqv_%BghL!DgEH%zK~5C!w|U>`&(Gxo6|QYaOifT|8{539vmhx9Y&_El(TR zUPk6!iu~QQWFWBhZZG1$S_J%U%=Kw4v1E(<3+=R9YZzMPbH_KL z`4Raro5&;4Ox}(+p8NbZy-jRTk+Y{BZQ7S(Wtx~2ho;VIlUw$c&>c8LXAC6Xm;Qt% z-@xx>bf(jOX!~qg`URftPd+zIk3(lNkx1#wQ~hrE&T zjP%S>`YutoY||Qy!y`gPW}TCCiM`J0)%zV?QsUSQz|d)Tx%X0#u^Gr*@G&M1AFsvP zLTa|(u_e0VRz`%j?qNn&xWIDcOw0hoeh} zrlQl)^_qv7$$zsh6~DwH_V=@C+=W^EmR#?J7m6)!sL!MF$9DZs=YLQAuJ)X|;_Uw2 z8>cT#K6zu`UDRpcMbOtxxwMCNMQN9bIRSq8-=t6E8_99nkptXX%T>-fz}~_3Skd;0 zxw)uZxFc%>x4*L;f8+nT;C{UaxD)L=w?D4`g76!-f#1*kLp5u*)E>1%_}6@x8^hO$ z|Lt}=;$eOcm=oJo{~7IgtcP|SNWgy`c~9f_3rNJjftU4gf*1KYwVGJJYU^tz-dj2+ z57?g3Z7ac~vS ze=C7E9$(DE$roqh<<*~F_$rqJxnl5Pfg923`g>W6>EI(C53am4{2vEHOI+RY^<#6L zv;HN(rq~_ziF`2P7q>tCz$Bh`D)J{9VcGwS1CJ23tY z`1*Uh_WbiO-*P;{KkHG#+~0-$NZ<}H<~FsI;eXaFFjWqx7%|acO|s(Kbmp)Yc_Yv#7jrjo&?UsB| zU*GpC^YqZ2>{!Xd2`|cSiCeSg=#S(Ccm_J~NWR#8WA8cXVXoTko7OmZuqnIoui&{H zc(Q>@^m%XFTK^@qrgVa)WUq@x~eKKQq(pKck$R5!q+#L34@ImoIA%1ooYof$zur z!pSwm-9Fa%fPH3)@d=#%)yXy4+;8F?zF%drx6Ba!KhA&by$`gfjQnxhKgN{LUNQsp zzWtw^l;?-~@!$B`R4)6*%%I#ul$+jl{{!t4^9uJP+!w7)XFr%S!OPw-2A4kD4qKgo z%O@RN{@TT5u1!}(pLF$OPx|YB)1Pw>b??vl2{2Us1O0h@5A<}iV)>yb=~0)SeDtN6 z4{z%Z!#5LPnBl-68z7#Zm&DPO=*^|4BS!vrjSoamA0YoD$bT>OKY?87>~r>-P&}`6 zcoFbqGyZR;|4X2Up7yP1;(iX_gy>7tcO`b;4NHLeD-O)spTfXgZueaiv8XN&N1riICZusw6Q6UCInAU`2DQ<*NZQmF=u@Jn-c2({r{!@M0-gxpJN*K+!A<4wvpsq zqJ8Jup>}MXuin4Vw(q3tSW6m1#cCnWzF{s#HW6d&G%ew9DI1s}??KvBU??9OHQ;2xk|Rlu|yI?|p} z&yb%uEh!MI$M-ww<;y~oS>Ku3KhW?UPat*~e5m!1m3)`OctO0_M0rEGm(9Im4&gwy z?=Z@g-fuJdS$?CW`iR~~pDXUH5L+h~JR5lU{sF$-$f3EwsAV+{hq)q6)Ux#GTd(j$CG?(VtOc3cvgE2U2M1c{#oLPdV6}+Y+?=SrW`9YZ_ZsO?(*~w z9>h;+(U+;j#`GcXrj@a~@sD7)M6p9`AA4H;m>)9NPYh2hehA|a!CujN3ymY50Z;z^ zHlDQ}Lu+YOf8<2R3jugVGAl^#(=xMH?eNfjtGoD8{swp5n|QQ;Twl9>nCf%z$4bMW zUG|B_>gsn({SHpLaA3;=cYG6H-^XLR>~yz2$?u-{agM34Oa6GkxkWuW%ui&y_;zHE z>Qa0D$0`5)xH4UNc2;=~>>2}|tGqp*QQpe)#SdNi+x^h@4(QttKHPZ}4`nA&&qi`= zXuZvP=-Ud638kDqxO*f1VfLwkS+Nx+vM!VNynHJfmwhRiVv>)i-+rSBNayih{DPijtO+nCTY>Mf?;r>J)p^)BUkE;yCHB@g^o@P8TaKgIjI z0+)wg<^5_0&TDz!(74aG{adiZv`*F4kBX0J5)H%mvaJp5lxcVrm;-Km+;J ztb7Me@T$$X+GE0R^DT2X>L10-2_IGHYw<-3?X2oYpCj9<2CyGF^JC?#p(xOnyT2xRLXSaeV$vM#sicS;(M1r>jsLSy5uuz_`c_Y?n-L*ECWv+NY;bcp>;fNv~l$$XP-e^UnS ztmm71oNr3zZ`SFRjmIf7emW@-$H#=&}uY@9W!Q=UZ@2 zESbi!N%f<6-@Iq&%U)0L0QJ3!4{$fOcbkDv_<=lGU)-EU>M4{^RF93LRd69Xk2J_V;y-e>;$^9b2t} z4a#*B_@|KU1w)(M^U-J0#j*|MOOqeOLp!8TG(P_XA6l8_0O^k`%PLSgUvlU0eRf+u zyu#h9(*E!2Z}DtO9L(qP_gzoD@$o>!k%qF7Uk_g8#!uLGfy(=cwG@g2?$%y2KJ035 z5p(?yong*}oU~DB#LUakw@a|ab8a6N`qJjJ;Z@mxqio5SM1R7_9{k-!=(|G8f2@~q zCK^U>!F$M=jL_c4G;iR>P~+$8IJt)Y{UC>&g=SumXSU5}x6892XG;z;RpmWPoS^u% z3EFpkRMJT^S@XEjN)A?ANx@}9?te37_RK9Q3x;koHP4@ zM1VtnRMuSYUk+~6jt+R*$Ne_!0PUlAkU0h~zEIgbilLP*()9)tpXh6sZ`W(xT;TKH zT%Z_U#cLUu9s6)Od-Dmu*Mal2*|WElqoeQSI%D>oTQX*k*pfDT)RqyRj9}_T_rICS zbIF?g*~7K>VrVP!su%e8TKy=$pKTwwb|)}MKCc1)-DJSFt3&nViIl!K>%V&HpG@#E zh!_j-qA`GQdTBy`CC-c1#rIzVzZidj<1f_sT>SI<)92=&;^(^a&&gnSJ)Q2r(&;bE zwS5)m^cNa>>KY#uS5_XF{vXa7$?aXVEPj!;x}N=;##2o8-f-grB2#+j#(^*yV~9kFZ5V+KK5{^yVbox99t{aJh(|u_h*nt*t;?;3LqV&d?P06oa4j&c zu^Ns(W;L|0vl^mw}hR$cNCPhA!Dp*tI@sV}}p7h{WE2aHTXhZV| zKIE0ot(5h&k!dOI*AJ2W%c|e+%ChY?&W8=OUPy^dD99gdjeYtwV}a~IL(2IRgCneM ztRqcrpFN0qRmxQkz?6u;Kx<*29ujz@&Jrx_iN`$RV&adsvsd^*&bgeMCL)8d8CnbB zff8txFHaNlF6pTk0qim&KAMwLS*&nPOT{y{;)OggF-Z<(^O$~c!Hjc8=gB#%e zcHlV*FSelr+N`9;Ht?5!W_e8-Z64X*3N2w=q%{PJtsDS;kKbg);&s>$p(p9ES!dYa zoIN^c9hn$x180X?7Cq8N{~RVq1^syJFfmhy7;k9b+!V?`d=s$K$1RgB)-`0b&v^NP zwd>GH+rX>FU5k>`p0_@LIS z-)Y-jlp{M`??x2T&KS$CtEM{)V6&|LeYq zr31Dp)Dxwi#n=}HZr*E$-}K$|ZNaroz~c6^_>ZxDEE?C>%2#ImQ*nKrOkZp5t=pzD z)22XdoNzEThc^T)NJ!#m*Z$=GCuH=&~xa9nQa$=CSw>E+}l zNoqVo-`BCvW)3uU4gagL5tKvR&Mo7u|5+zYsrA1~tb1Fp&@ZUF8$6Z0LUr8xBJym^ zXxi@R!k*+A@n^FAOS zbV)CCNpEyXA9TrkqpgO%d4Y!a^R0${BLWQ{y95~-tyG41GilA+=nN;X~Dt1e&N9-k7l0l>mB^6FD?A-DQh$T&6gJZqi;~~gm-Xw zaJinHAN;oO{P4GDt`MAgV z4;`(BuCBD4|IpWJ=<5adedumAbaveR4~~x`2Y)y;(D0+n0}U^Fdk0^n?r&XTHT>{O z+Haj_>iO1K{*MbZoag=@&;P3f4fCNl!7-)KYN(}*`RtiiJJf2Zw)zA=VjMQp>)BuA z^94W3XD)>6kFo->k1s@i`38{3zJGmUzej!NVdG_zuf}t%k+r|mI0rd5qvtffZbj#1 z(%u1qqK5OP7Bvi>Y|3hHy1UWUs2R>!3{+X7nc<^c&{sJNV>hK!84(#!Me4m z;g`K0KG6TNpMQ9{HwBu`G&G&*%?RG@O>t;CCHOk~$o2O24sV+NQ08~R!~4*;X!-}x z)SEnelkesi*Ji#5&k3KazqvYdFSN6lXY;*%!t-a?&sKwzfzZn9Dy!(B%(3vD;j<}^ zW)6fFH+cq{IzFL}!M;J^Kl?5S4)uA%#et=nH(ULJe}P}cqyGY}7F*|si_4Z~_Vaq- z&wk;J&?|Z@Yy~Pab1aK}@A?|}TiFBKvXa79nSK9#k0-qPj@6liXFZxJo>n_P#jd%_ zwQI1gxA~^MwU_zB4@^_>p%YFby=Y=2+gK=uCKc z%R3veaZjHJ_vh*p-52iNJ`wIOaNy{?@AQf8 zf6+csU%j2edT;pUvyQ}mJD6`BYh*8wKcJ7$b(=eVUV+}B_xgP?UH83_pRMcgb#&^w zczw6>0_nT1@uloe&0ENBTgUk6Wo*EX0PAV6Z$GO|Jw2O=yq<-uoQ-^(gG{^=IXD;D z_Z{TjUC6k-Jb#1d`*^;K=Zd#^5#NaO1JGFyLEnsz_GzD~ z@x&i4>54^PmE-Sk#z*#;%2SU|Z*%$j3%ch-cE-$Z8KWqp$SDIK*atb65B$Pqyhz@g zw>;?MV)U_RU~r2y$k6_a-ofbbf#J7i{0KT6XwxCIAG7)gPkPY7C2KS9M;EnN7l*5F zu7+3q=xz9*d~N32$ci04f4I7$Iy1?W8BBp!+5*cnYdk5zRBW%*Z>-4Nhz-r$MYyeO zdFCYUU*|~;r_NlFIT~KsWo3kS1s=%!xhFX|($gzE@*68N=d!QZ9q7zmWe;RN$o-X` zwD8E8E8$mv@G10|-k-$%2_EVNp3eKculvt5RW zkf%>@AhM_e+{zCX^O!<$w-#6F*z80aEukodXHwKo` z_N3sSe75|Uiu_nh+iwOZjlO=!(Ts38HfK+CrhCi5*KX{hk+XiBnFOx2&Y=ZebJRDe ze(?zF*t76g;eFPzI{X^M1)1@9f4_ak7Bl|`^s>)=`q}5c1MG9Z!S*?M2xEL`RPk!I zt?l?EqRpm%dW{L}|2oLus!em(06#$ffg2hg>%A?jd95J(yND zZxhE}j`ulMEPJN3ZdqMvcv*dE`?6se5Ew6sE^!&vIUeCo@mzI^J-r@nmZ%cs8lyI)MU_J4e=nhUbCdSEY|S{eGH${~!9jy84CEjnyxf?yCNEXG=y+mVUb+SbE#SVChc^9xDCm!iP%VE_k@~;KGMXt*;&^^)EV5I^wHumR_>x&C;8* z58klkt8bMqTl7|G-B8cTF3(BjgIPU!<>ymlD&$@Kp+gX>*>&O~2?`nPfCHx%7 zg>@cp`*#j?OqJ~Y9qz6#jqM6=cXW8_E@VG<{HdWGn$Of6C%ig?`ESjO zPD77A&Ujw)7n)Dg97b>SnXa?Qtq>tjJYKHkGSAjRS<5&xt`2LCuGsR`XHm{*#@$84 z?=v^DUvek=zG-jD-zk7U#=T8{PKV>(9ILb2gP&UGYfFdlA=s>{5KG z>wsYezN9F=9KDki!T*CF_f^YJtWQc~HL*ULkC}!)ZsO+UYum6Zn8RtW#Gl=YA58w9 zX8aRg#n0l?Xu_w}JQ4YZKSwzZR8||f^HXOFdRFzZIVR(Ul=gD`N~~ALKcsm8+ds4# ze9Ldt$~RT`8CaLq_~hy88pS7h`L7(Pk5i87Qk;W)IDX2o*EraHVdLJmo8^DLrdQyR zYx(vdZN8DYkj2bd~Q$36A-|VEUHvkNi;^pp_{7u6^*zshcrv zV;<)yu=!3`)hy+I%NXWGEng>G&G!8Z;XTaEr%~$c+bTTimgrZ6W33iO$zhc((S3tZ6Ait2S8iG&+At2XZL#izg^`` z`tAOtz$1GED##OAtzXV`GhZnQFf%X-hX z*J8VAuI(stNb~p|<--S_Hazu^2F=Y5<*_oY5o zec$fKe#Lns=l|mT-<)^x{0?&ayhxmq${OzN^K7-%*Whg)b91^k&)e@=?Fr9V?YHc4 zzqa4;ju#ZKfUQSAv5t?}j(B=toVCgsXR#)=cW8BQ<7e!4@8KTq)%0+08TSN3PqMd9 z9BsKiNbv)6Q6{!hYjLjOx|lw6@$>-eO_`r)Q(BfdhV?H^=6GgA3MGt{!@L7uQqFp zc0lFuo#0@6#TsaT=lJRgVlIjqpG8fqN`g(Df_%@QPUc0#qrV&0U9KeRuL$#A@j6@i z732TA?O%yZjgK#EnhfpVV#V6ZnG?C&ifPXCad_k1#fzTLVoyut$W-QvY+2&7ed{jm zEK%G-<<~+1-=%t&OsCLb8eCu8V@AypRM**CbTzk&e|Ox#OK-3-r2hnrZ2T!tn^mFNBq~VXQ zc$p?-qu^5)&oT34A?t4>&0!HYoA_+;iYc^m06}JAa5?8spmNw(kgV#jpQS{FB_&VZzbLXo%EKWz7h8Qy z4Bk>`cRO-Ea+4?a?iHTcp;?|-Yq?!k^DQ35`ytb_kuh4^)u;acACn{WmLlpZ;=Ll? zC(q;DBI+rkKJqYrzli#afTM`=iek#05j8m@h^M=$2wS92`-%b9)VrBi{J__@@#fR3 zYA(m8o8{@-KAZg)`U7K@r+@o1}%uPR~(q;SPyZeY5&); zVIDQrf|;ud}CTgec5t~1|k@PuxL2h^sv4ZZ@6Q)G{oOD`7M`D1mKZ7e!tZRDzIXs&hu zwlv>LF6>sUUkUHXHlOUoJC>vyUD=J#x557z3A|tV_0V^?r(67Qdho8R}~V?ORn;P=V! zDE!$5uO5ZRdlxZ|Mc+gRqkHi^x5N9o|BmDU{IB;8!`o5hL=)dO!|(C5KgyHZI0@c6 z3eU9@^YAS3AnjI0upK%-zM(r`2xn*U1@HXk@6gDYbMr+B^x8HCSpi=hfiFfne39l! zYdi{H-1#MIY$1Fx0^S#Yh%fx`a5a3<;_yWaG}HoLbnw3g+0%+ld4TcUE^PNh<89e? zyrernOmN`sm=ArQiyvgexcngA5I=}7#1HUAgfg3;?PmBc3SUeh_lmiP3~a`3h{BhJ z?mgc5o_o!+d$<-YHBW(li8o=+uANDrsoUZI$YgsjK3C2*)1D~z@*X|8M(@fdXyM*E zzDZ^O;|TJw85wvDdQx#QpE783U{0IT$ zrS_lSGr`(_l(CQCo96MiSMeSFLHr8xkvR*~Mr4_IQ1a&o?At56ucHpeKw~R;KXL`_ zqMhFd&!2DG*gP8EdI5{l8yw zYRw#uM49j%^IW-MQtS2n|H9ixShM|EaS`X@?Tk2k!Xk#dNBi2?7hUnG%3mnkVlXze zrThqk?6pSCj2nw7L$uUmeuOG}9O23@*%U9u!Tpcqb9fxse2lZ)d=5GK&-hDgWfbEr zyWhuHK958H%S0R4wW-jZ_Tn4Lp)s=Jk|uoj^Vj@690%{Y;=4zh_4!@vXItP2?Lnq} z$(k6SS1`WSx+4!`MD0hFZ56&H+}p91p85d)HEy%VV`goHv3Ir+BO_m!@(#6Ve1$x) zC-4W!Nl>IL?}~@rob%2_W^8N5fQ)Oq z#lu$FxH>HP?Xif9!j>W~>Q&2Xd=t{5;|HhI-y*ZsrNSGw_Cz{Re7ROd;`w&C9DZriVyO_xb-!2 zN7?IhmhryUkyWymXc={C4b`%3)C>F%P>#x5c8L{x;Bs&QEgn?AZhK&Dd|9_rR*sMF z>3@~AWQ6=$iiz~JXF~(x5xYN<>SwCo;SIrUd|ReWt9|H49aAr&e?8Ef(O1xwkMfc( zr(LH1{f)`MkpvvAMsF>vslvA^I4W2(^dxJBt^$r+;PCTZ9&q@8<18PuLkp7kf9HMq zTaRcLonHJp&EV3Df33NMGwZNqAA4Dg zrEh0|ml^sN`qsBc^)1)tTh?P~j>Jn}XWi$o%cB3YPAAo6fiuw=ekEjCp6yrqg0kkH zdcaS-tWkakhWnl4{p+zw;>W|IJ~?UdBm83vY@XhT9!G~2r&*he*Y`EPD$DF~-kbem zeD%XmnDs0Ba^M>e_r2C}zGQh(N;KArZgG2BEt?1cM zKVy(Tb_+VxEBKJH!p~6Ve2do)4;gr@ULhZ_XdLL)iQV*- zCj0=J>yZAarB3mP?&VOQDK~3^sdFi@62;aPA=ZGH`_r@hw!hS8%PrB~U$alGc|H+; zM&Eaz54-68GizQ^egXJjdc)`n$1m87pH4pezy14}Zn;fdJ@5MawozBSjdCu1zY^Y# zk6)=YvN5L zHtzcxTG&9}XbdPiP#nJcE}jRX^uN{{T}nUb+Q^AiM@~R9s-v7b6c^lCu4qF292p)8 zQht>2x{vZ(D4+dQO!-=`IFr6^b;>`F@^w9ozLuUe`#nrjfGNY8~bIL5I zOvblgQQjUyKa}Hh>4$rr^mAbx{bWNw)FZl@e&N~lGa0-;l|Vmp67nA?_rcLw$gEqC zSyO1|P54L4@smzN7v2V5OC(<`=Co3SB_3a}d$E#E1bMVKj(Q;TFrr(b~N(f>NTW!9)n=Y6wQ z4tZ_L&lQ~zZBD6 zj=a#inh0{&%?N7AV zGtI$P`?T+%&KxcgXy%=dJPVJ0AKIA@?c5FR%muH|R1-9H2>NP;rdpt>PCCkij=)bM z9lh7r*2T^A*);ZJHMB%-WBGa9K0F^>w|-Q#@G1YcZ^PdUr;$Ucnf}fp_9hD5x$pIL z-b;q=?tt!+v#l*pL3g{LyX2dpJ8(K3I1=g3qrT*s=Ig{Cntx{Q+NL}DLUebz=;Lna z13TZ)o#1upj{Y=s$JvGD9opr-->qDiKIYoxUZ8Tn|HaFFOXWK6yRfSKYRj_eZoXZv zU*&=`!LGUMF1&l|LzEGL&pWXFJLz&PbP4Sly5xQnJktyf8ycrwCZ`8&j%@IReh*!0 z9A|9(65hiObn~>z*2y9NjqYW!j{9ZtCGW`Yvd1*`9FGmp#UX6@eK~=`(1YYDl}*r1 zPh~5ol+ULf}1^uR)mQCcZOzGBM#pzYC-p=Y4=BhctkeHnC8)R(b$t;z|+e)sm7HRZjbo5lRU z7d~9aQA>NdA9)PhS^5VX>_-nW-}xB2opLm{yLf>2*a~P&bpB+&$^+Uvs%ZnXyqt2l z_XPJ`o=k!6-96=+HbA$_p>O0?C|0L z-Gxrw;)7kBtIUtV~p z`Hzg*4GiN$e`1~U^tu%XuzeRku088htpeHK&BQKfF94SYN(C#tBO9&>_&k?e3(Bq3hv3>{{&9L-3j6I-a7QWX7G^2Q=9p zcXsXl4Grnr8}u#L=3B;{hrma2nYGaMYxUtB;Vhv1PWX171ePV_4wMaFq_~AG{aoKp zx3$hMk47f6!*6_7ZuLcX``VjEpi|(nruPG}w~(RN=Cfbwz4Y&^wEHIRPlo27rr(hz z2KS7?4Bn9+&7)I79&kTB*ShKgaP7*h$V<+<*xSulq}csLeSiPx*$ zTo@Xm?><5XeuB(D73i#s{i4@Iw$jeKO-y7#^tM3kJNR_gk(b|=h0K9;c+6gq5zbBM z9r*!V8rj(!Il(uQJMTz#m1ECMgGYV^y-`jRvO)6nBjOrmJ2xW(j2tSDlS9Zs$@vK9 zL^-rc<@249Lwp;dT;eRTDP(U*4xtyfq8k)9`BDP@o<`-$_&)+1M=7Jd5u9yj&jb4TJH&zL9r3r`%TK`;n*IMW9C1k>}l zuHyPi&dWKEvtk+e2@lQZI>7n+oF`zLT@60&rR|GFm(oSjIpl?Y)tM9arF7F@qEY2Y zkRH=KlF?hvJFcu+2(G4g$*NU3)$Ef)TWr~8e^;$M>G-5urQ6R@&Me9)iZ7?4`&t;) z|5@e4>s))zD;d~RdW87~>dJ9^=g8L&axTgFU%{-p>^khY1-Gs7__6(|z3W#(8@8@A z_Wb6Yz?Vah`W$;c6CE(fTG)Icb~?1u0-dx%BZr`mHfTe7{-Pnsfy@B!1$ZyOd&rKy z@L^_v_n^xQ(J|*(UzNswB=PeZk+H~_>v;bLbVjdje~IQYHsaiprT)5)+zp<3SFhjH zi+V@fbHdSF$#~~Fi)-jKI@adtnaF+OkYcW%U*+UlR^wvkIG-V34E9{(jMGbNqN9mL z$6hnvT?r3@L;0g-%D+8(`Z(rR!q?2s8JB~MRjhV8zO^jux-{mfX7c{EyrVglD)P3B zkZc7%@-0lKJu_%e8E{C(2$r0F+x}9_(dt|D{Nt9V@d?Y@sC9$j+RQC^*w>YM_G|yv zLdIe5YMw^zp{$vd#T>vdvFp}WGv`uqoy9%mLq0HrYvIWBbCvCP=+?KDg`ph!)}7}` zrtghC%QqV1c@-}PKMm$AeXp^I`uNEUxt|h#J?x<#r*#wQs-z7TJINS7*sS^2dLeRSbZc32+AV)$nXzHu`*3r>eZ4;pW}^T&s0$JSJ8J{(@wJ_6$J3eM{s zpF|FJ_Y4krV{VW0;|^Zkv9w_S=mMLDg_Cp5Hz($&5)bIe&aONleOp#CXdyUTo$AKC zme_ru^8ZTt@5S-QF`vEPZQ5wbxB;F1w`}G2KJiZ>;CO!Rl+$LP2Jv3bm#i0>L|5>Q zTdxcMQtkCpPX4qC#sKuu%hrisN5vn|lK4aNUhyCD+sLn#n{Q3cMozIdpuK81`5Tcj z#mE!*r_o|q)fkIEl+%jaM<|8|( zYdUk0_Ol1gvnFWk`m{hn;g8v)p6f#V+)H@>$=MYPhCTSu0kzS?{~f8!8*ld9^EhQ| ztTE#g(I9(MQf}K7_L!oAdv^SPR{cHPfBBM=YXZs*eJe3Tih$W>Y5); zscTu2TGw1sT+sY!%2iG1-pDQV`NZOaVxM)DWHtW8f^u^Bd5P_19m#-5A^3qVno58P z{M=9dro9zosiX41&jc^+kx%?Ft^tp$0ZChVp){C`C7-okU*`xoGiPA)E}Vy}4B zHG{cyf(i#E)-ltr;{$8UK(qhSvbZ{y+I8%+w%C1fKmQlT)uFurZJWnuWTgfDWl49V zv(4BN97%`20$!31-}|QS%?5vI=UKs_Nz84}nR!BER_!kvK~5PT4D->~%^)8ra7U0! zQTkSXW!b}Rw7bpe2jO4ORsIvmK|^1ZT|B^I4M^bmk}|6x9)=mvu<%w0txKlqSq1M` z^Zt?wtH7m`kFlk)X(M#9+4@~t=wIY@d*pv`VbdD_CdI>W9ORH({8<0l2lgY*&Cb0KIP1COl4S3F=bu`$AGb#$nDX;{<1w zky;e;NWcF;B4UFY~bK*;NWcFAWo3EezheV7~DQk8BN$yuV`(_w)BvT z4?nrTu@_<=^X$KPw)&OyZZXIGpc4}l3eIvWFAJGDd|g*_&3La#*DJVY%xA72wsXCT>jlpBYOYr}*N<}jsB^u8>t~$n|KfU!bG@7E-OlxE zT>r_r{u9@So$J4GeZsl^3D>C}8=m!C4|cAf;d+E~{d2CzIoF%HzRtPc!u74r^){|& zJJOhIl~&d;Ua>(O1G#_hi#nk|E7 z*Aq)a+55_#o)G(2`Z=3^F4H=r^T+?G%e|7+u6kzD9>tj%_=(Gj zujggPv5EBzOFg@uW%#NS>xm{)Pp__e;@jgvPsZ1?g!Mm(^|boWuIC!=omJ0=DPxDw zSGm>5HxAL1p|KUjLvwV|*bUIw4W0P5X^eXnUH5Dn+jTuZTm&*`{6%WUr_b+cIrzZZs-fPZ{a^!-`%QDb2Gvxewi;?-@l%t zzGKX}YCYFVWbE;hg!!!w`Tj3rZPLYc>QnbP&X_uWjnWBXf~0r-;9YYhE#ST7AU-(y zxi#u5uzmN=zFd4e$#1P@ZY9)Io3_-JATBOpX&+!zFJ>m!{5=7Rd*^kHMDrK^_gwi>9z$N+ih9&nQa+#R$CT5 zKQOT^`Lty;Z5d;?<>Wxx!dS_Si?PYw{XgZ)%kWtPy!c^c)A_H)w!aylGrXH}2{Fe3 zZ2qati6FN#h^JK^#|->_I`^6oh}GzPt8>PG*E^fN6UZOb8^519UuU2D+#ZO%#(fi4 zkjLKA$?0%->6ur;TGzkKN)ET8laf5rOKIV! zPQMneM&G_l?xMp}U#(3mV@*5d&OEa-Jn2khxSjWY4tx)7dW|*j-tc96_SB|+qopR* zlM%lB0&7*jrNcx07fnAgg1m|NEn?4ntF0#W8#}_OgRR)^P4CpCeCyS43cmPRdydq! z{rt6X%5~&!wZ?>7Zr)K-y?amCUuKzjUGi0gmml9zD_o^|eBrbkch_csley$h`cCi0 zTE$ar0*+sVUauv$W9)rkIrhspYTGuyU(>eg^>E4vE0+IpOHJG6H^M0aE4KW`!!<3( zcZB`7TCp9y+iR+C*d1mLHkS9uyEWCl8pHlP>ZOlNy<4c4zB2Wmpt>m zPpKCin|hnxsNH*{y=L#3-QhkrTd~W2_HIq%k;ZTae&MzV@caaLn6Joh0GcTz8XZR*XT-hNKKKYXKhd7nD_*IcJM2i!+_YWc4CYWaRy5aa z%-a)Ano4_bZ>edncqOdZ!<1OGrsa0v$!5+b?NH736WhbSaTfD9@Bw_0dNq6i&eOql z2DE7SfcBd+{jT#y@?E9F8{&)J^nV}voqUJkKJih=u z+uvf$MUJgQ(7f6=oFktce5v$rG5OGoscR|idxB%n=~u$Vr}u{MBtKUMxUZm&ji+~n zt55IZ+obUBsk>_D^4@ngHP*_8x@$Lh`^HiDASHYm`nu1ec_(%&FBd$p;#-H{fp9n# zI(%kRdrk7UC*tN(WcuZh=gZ+;H-~Gs@3wK|124>BhwnJPy;d@#4chHD8RQZUPwpMA?M2%k1crw~uhsU+L;e8c{$K8`-L?6>nq8}23lql{Oa7QTHt!8< zUhIw=+iJESr+oTmL+_(C?}#MHZmdg&8W?`i4<_om(> z)C=xSy+5U1XwlT$w6}KCk)t)6&b%7#4NYA7vv+FhkA%bN@cpg`@caaL;JxGq;E4ba zJaESu;MoE^(4yp|srS3o3q6>6|3tm?yQz04^+JoL-W=+sUroI~+*_M^PqZfWhkL`R z@cisQx7GOWp-gyh!%o`xC)x-P3=r{1Yfy(f@+ z(A+TaT$j4DmYlN8y?eux@?NRUfUZOf!HOO5Kr=kBGn@hM?W$<0c_k0JhX<1XHCnTy z0=mE4iVcs^#@m4hS#wQVYt6kpTv&c&T-N*C5ENEJCCk=j^1HBhRYjx;{ z;?vv1+xf4)p8yQI4E?_zPUZVa;A8T)!nF?qi*&(8a^O9;duQ$3ZyYASwJ-c-?B4#1 z?m02qn-M0)KKAUlj@Hcm#_n(ibl8p_xbs`k2J}$AyS1hqJ)nNc^;{M%zj=2}`R<)z zA2_mUF)ch39u)<4uMJwTtBdOP(1^-2$b+xOx7WASX1*eu%HUhyhC&;k#<63&44HdeINJeddG z!vjh1z{Uz_eF%ClcBm#Q9zfOu7d8X$&)BqEB!-0o$yX%0eN5Hft~1qHys}E86LQV+=$|VW^Aj{ z-RMTc1OD(B#<_hL-F>1R+o&54K!=Sk4?qvqyPIpO;Q?chc&;Sx6=&Jfns>T z3;zv;2a4eVZyxoIaO&Mcz4VEpMe%^)J?ib`0qPYGxU?A01JFb#55V`;@IW;@Ah|f9 z0eIj6KRi$k4=mjRJkX-|YNy`sQg0xU2dGy(04tal4;0@+neg5# z@IWy<;6)D%#TF`t2fWbYQh1;k9`GV-yzrjE^X=3d&jZk3E_j}ky1h0Xdbb#wX2mJp^8{q-joO$rT(%XRtSyP-AwRr%Z&%u{eh+p8@ zpLONAbNx?gX8lXo`d#_jrnlkaXAD2(qkA)=ff&BJeOf2N{Kl0f^xb^oB;=39 zf3&YWa7CzuzUcP-CDySy_PoT|-_zga@!yy6J^k11dvf!PwC5Sl{vH~c690WA-_z&a zzQ58scEtAA?<;fQmw$W-@O$9#628v?{^@2e`ihVTIx)VLDtiueJuyIk=N;u5I?lO_ z{~vM=aQ=X^aHR2ODD%`B?K@}1|0=q#m$r6y|KExK1b=Aa{y|^8$R00e(T`cLah&d5 zXJ9OG=F0Jf3x@9kgZzt${!qrsXT_tWy@tKbI%5wo7OwXm+l{{{dA%>>Jzp_{iczxn zU(ntispnN5(458JrW5zf99PzN!58svTGLiYd|wXu{a5~L{;Tji&^~2OQ<==4WK(AH z#lU}Rab1TKyJ2E7DDz41=jHr_gGc#})Q+z^?bw%qPZwW%%zDQz|B?5*c702sGavQ& z10@9&%?Uk=>hx$=?*lzcm+AjLn zT2T+zx5iz6EAIN63HbRt>%7&^zv}@%$Exi0G3@C_)(g=qcwT}JVj|ywBjRCROi`a&DeaTt#l@U`_@qTEE}bZT zS}}=gtLWz&qLo|Fdwkba0#4Ycgn62&M+&Xj0rD|4U4g%#P`-0-BRFa-Vg9~__wXGw zwoFt$M_*9iw@gD%gByG8f`^zHU;Tg4?sUqXL=KK=R(kuS%{6NaABWzciwyc-??1$| z_sF?3^2`rvur-5=d9NJ$m`1*l$-jJPt>D>4ADMko@Uv*YjAn3f8+~Nva)F^4c%r~5 z7>?q1X$M}FEu4(>WCYXbZ`EG{jm)*u8{hS1G~RgTp_)E|jsKh2b3psldOq%$s`>u* zvFv{^*@|5N?YM0_I-YXj8G}Cf5NoKX`En8OJ=AH}qEQWM>>yVvNl+hqg3F;l>f&>ClTuiJWi+Tb&EN9;ebfwC;PlLN(QBQl37Ea*0_r`^o%QrwMc5;Xx98Eo4tUexxfIg%6F!CK06*v zc`exwNcqT_3ozKP3gv5!rz!VB#(4R5*}Hn|Q#+2$_QiYaP5c&f@XBvG&-7nXz2w7f zTxX$UT)%pJJk;_2vIo(4o=>P>zS@=)&1HCwrNToU9X`aLor6mGlH6yzkquGuxwPO%ZB|Yl%8yWHQ&f4y zv<_p@^TczILyh&WeT!nsf!a(y6~*?HBYz{KoxDn%Cpc?13%$f`^RB*W8bo;)Qg$YA z6KiAm#FXu;zf1jlq23`sOJsBVrZ@BX}LocX2@czG;{UgU}4oBBjHJ!Evtj(RI*ne0+f7`@r!r3%H2MTcx!Qa~88!x+@lUU(UW^CKlJoh{QY= zj=ma7%nkbiuD0WDyU#(nFvip84RNr#vgD6(I1_zdc>mG^$I8dIZ~r>+Oqy5VeBlq4 z9yne;zA-*mT=T^A*ftBfQj#9CdZ`V4LgQ#dQ^_RO=BCHSvj$#y&1|0c)+e1LF0(dJ z*ZNpdUE7A@x+Bk))SW6H94gPW>P~GK9GZC{=jh zh=mEpI;MV5K9YU*0-;^mo>12POP#06I(l`r6`W%?1eGW{gpw*TuY?$KVQOvXrPm+PB9e9=?Z=@TBg6o!vY4D?BrSE;-)eT!L z)=o~m-n1#2pN_qn7HgiIMhvJw*aWYO@7jPzy3IXo9_Sti^k2kTi3fblM@D~@MtS5- z9F$JJg~iXGg5E1O^e$+koGVlb{RLVR_k9Z4=k#?uDsP#=o+MMI<7*xS`_1&PZIYCZf?{0dED(qgOg zygkyli+4l+wPXMQjPQ>%w zFFK!{-=o}p%6ZmlT7RnFSSzsQwz!#w3VM9I?Q<&IV^VjuiT&dwVU1?xqiuk z;peAbxJ&=xQPHSd<^zke7by25ah=E7&vZ3Q{#8oOB~V0A>V(qjehtT-Q#`} z--iEm;CPvGWhWVV4zCY$^w^9&@cS);(V2@_t4Y2-$pv^u@xA{~I_kDt3zqr)!9%y) zS|EB}&v#w8?5?L~ICYVq|5yopD;PHdquSHWF2H`7sCE8vHfzV3E}x092d=I*_JD7) zZ4d03ggtOW^;z~nWNd%N6#W@<^kr(6a=kxWJX&$ zLqhn5PRr<9L9q!W2ohVlYKKe4QDBP(-N}OsAbrJLP^PBoI)j*m4ET z`}^;6PIB%g5wzo*=gD*LJ!gNcz1G@mt-bczYsXuQ(A!2~ zmYln6-FrStX`Lx28s8spr7&0TTbx~GXLd}7#sd_7B2bKEwn&HPR?^Lr971LP#K_R!kmzS0~Fq|wMk-%evoaZiFAMn1r z`v2&Tv)^6uJm41IgsWAo#l!vj8~yZ@aOLo+x9|12Pv2{u_b(>Mo5cBimD`4R5L|lU&+UBIy}TLdyzh>;UVt{PywQA)(tatlZ-xh! z^SJ?k*-U7@2_Co#9@v$Ynp44V;tdjbz}79Pojg!uY{yh%I~LK0=6!-)hOe|4KW2;k znb>H5&>3WpjpjW|x+ck9lfUbYZv9RDPXzHddu{u<(w7FpD`-o?6=uJ z#P4M1ok{d9U4+jm;Y{qVu6?g$|Mg4V!0x@@=S1L2+aF3k|C52d1b=8aJq2HCijf_D za!^L!55xobcLLgrVE)trNBF%!JW}fKJUeJ)=&nRr7Th@y|JF3@gI>`4zv{c$(56e&%E-@tMOOb4W%%dBE7{VAz$N)FoaxN_49>Er@~>FLeCAAvmC~R(lks7d zvc9Qc&rkHM;9fOi;W<|t*cU!c`+qJq?GR%(AdUKaxO-}jDy>n4>uR}TiowVK!x z{>g$w+vy(XxokM?SAW3!?%?|ur(7g`lCgnh|GtU;BAgB$8Nab>v2kBc8rKG=p3zsl ze=6zyTIc})ThO5`UYG{D&O=|xxr~cas@Fn7DVkZoOFHz+_vx zWlg==m)3{?@X0R`9zOnV-(R$~%zZ(;3SM*5BWlGjrSs z?!U=6)A3WqkH}WHWxBn0`>npRtF?#y(6+;hS#fOtkNB>#C;7DT%wA}3aKgT*Wc(ur z*X&jHX0MXH&E)*J^(G!G3E!6g9DGmg9^d40%VXV>(h%$h-x=(wC*xaq)_%P3nNK{e zaC(Hj++k;mYd4)K?!@Z|@V|G_9PqoO=`i>`0)B4;zgL6b(FVU+IlaN}H^AqM;J3c~ zw%TjJFSIiFCDyvOmAkVtzNq>cqeed5YwhQ#yK;TsoPXdhmtRun4eD&+dj{V#h}+2M zv!sdpe;RgJ{u2RoL3ebpX|QU}?gw}5xmIQKUEg9cbkyYBX3!c9U zysMq+2lE&CDO)mld{YVjqEgC~P=-5TYnMGZbdmDLi#~5M=Y~L^66iAx+{9RC8vO>0 z$?$yWz+-r~I>SCMg^d;oEZ%mL2S4J?gPD>8#qoc4@S*u%&)qiO{Mcsxqox&Cy_8{RmaoF!#hwQC+K8%mG@gvJWldYMtwPS`A0WPHUj}zz)|2}W z-dLrW##A5o=r5ni$#kMe`SBeDT{&N8!C= z@a_u_XRZAn{C6$%*t@=8PG5L;KfIUyooP*3({Hc64&Kf1yS$iHn=vf-#7I*=bJSxG z_gNdF?%_@Sa}E$|bUk%OQzz4(xo;%A%RTJ1BZqyvX@oDM;Wp|V31p6%M4j7BovfT+ zP)9OoRrw@OonF*gVCu~E)R}DR^v>Bwo%z(6cTSz@rp`gjj+NX(&WVF5#9bm6GtW+J zB7UpzAbE@xmymh(D^0|679PaDsYlLb(sq3ozsWbUpRs{zMsI)Ab9yW6fdcH&)*rx{fj5?dR81dO z$9I>8k{R!0sD5$$DQISS{%m>L&z7g;d_P&;KS7>`$7*eSdErUcecrXgJlYo>)rb3k zGM2b!JKHH^&XDdR=7=?F^KgH>nKf#2L0@}~O`Y!6*gyS%eb>O(3Y!o5iO*n<_Mo4= zV*7rRmw0#bV4^(#d#5kEtYVoKvuEaSxS2bjPR0`N?YhY=3!iC!KSo>Ak>}$FPiVSj z(5+3k4!EUhLiYHkiOBTZGD?~zr4%FEgHLcKthR(TDDt8<x&iCEfsX3hYF!EUN$PN_VY91WP9-{3BlD_bt33loW?bF%3 z(m#|uhJSFzVDn1<1by)mF+DErJ8hckj6(6jR~$ZY=WwigVxjW?xGdq-ubIRUo~vII zZTdF)>Q28#o#(QnbHQ=1%dM@GmJSY0SUh{tBM;->9@Ho2ktLaH_nf)6X+|h5CoNU} z?_krOCFwbjT-GP&b>6k|Zq^Ueb6&@P)(Q^yT$XP3VOKc6clS>>e)4n|ugp*F$=Y}w z0$$UNf2|iXmObxK481#J(ey<%a*aKdGeg7i=YWH)}L;1aOI?(6%RUG=T)=jc6(h}(7>hpA?&nGza5j~EklRqX2 z|I+6p&HW2q@&3tn&8klPE=DE4opxj&^$llnHIA8wY)VKM^ z_SLcOua*5i4cn!GJ%%F2IGvcp0DkQh=vRZyx*WY-$$Q0sP9u-P7HqIXfkoUe@8=%$ z^nDXgRW_xcU9oBoerNUBY7O8H&J+`qnClC)R}kmE*9veKea5~D%2se6eFbIOfK|SZ zx!A5p_*cH$xya^r=ra1@-fJI&F4utjM>qA!84XX?w8y*ZNK ze9ua?=bC6XZ_x)J2bKV z@fi*8ljq;Q*DebFQYH`S3~ou--Y8HM(xDhfgMB&o_GbWUk={ z-T#@^@VAB^?A+`X=DE!$Uop>ZKDmYOE}tCX9)rcSW#n(5ukHWle89;-cA`E`M;})Y zDr{acs4#ioWI64fqmwH#6ZG~nr%Wnw%Pmj(h*f%MTg#vML?^dw$SO2>XsM&K0`f)J zv1(~1zTyCP(=}J$TiAS)kN){A@)d?D8@Wk$sp{?xf5TqNoGeX@ft=>_dHQ#j$<@$N zk#Z`m{B`iI;cM(R&dr>-8XL;QBdblyMk%8`P@$ z!X0cUM#5L(=R?%}`J9=H)=_o@zKd3Jz_n3Ux&3O0OV8yzOeQ?FWpwa~-Q?xVO&zn) zPv1rGa1>p!iRT)_a%?@xL&2!udOweH%Wodvv?FEA!q=FGrVh_)cvj<-*R%QPh7XeB znU*?uFfuog?dWXbdC|7l6mw&6zG&MH^IbMwM#I0E@4A01&~Tvhc}l}i`K}y*zu;3m z6i)Z}FVOYPD{Or~1-_Eq=hgSkD<&4sa%7VDtvmmb=rG<&GxkcQ$A44;F62K_>^Od- z)$(K5=VwwH`aHxP&cuMrcN9aGXniq_x)Y(nO!R~HIBw<6zN@gAM9XOU;%&WXFXQ2_ zH0@=*vnfmK$5gX^%t#GBVWHPKe_C6@^IUK`3LDy9JC-ey?WMKj$S>Y69|ZldR^t6$ z)>X2Ot5A$`zwKiuEqL!YX9&d38^G=O!) zRm7a|dpi3iB?E$mC4(!QO2E(Q%g7zc9b}zl2Qd$ZdCCq-DqG+wOK!?X3p`~9CzZX{ zQx+JskC6bwBB%eGJmu(rZl0%{`ghAs@RXzP!&iFBsc*O3-JWu+o#tL)mn&!N`B}lj z{5f{N<fGQ?&JBKn`wWzuY^`wkQmMHYOW(_03Q!QAAo2D~ zPp^0)Qb`V5_%wPN8_1l&=@nY@d)*fn+~$LqEXsr*vKG7VqVJRAhI0Vl^OOnq`_8uT z1D-PNcA4d#GFAuekhiz9otRx_iKk5CS;~CVQwBMvzV7pssbkF>p6@BsWS5!eDI+{6 zS6dl(G=!-yzW)=s_4VDw$@HI;n!D``PJR-Ie}s?wE(b5`fg>{8!OOlbyxisKBS`#U zxWWU+E*p-Co-*CR@vCGwUiHB7Y9MavRL?A|BBz%5{*#%7lE0D{{N7sj_~=6OJW$?M zCcQkd%#(JR>ACX?qvV8B`RN7o3YWRxhtDe<;(ot&Ug325ib8MsElt_0=b83oKdaB_ zYfO7v%hrxA%y4kz`w$#Wb8z%(7mhMK;}6H7=Tr}leq!UOgEn0Ls-hjv`gQTe^LCj( zdCHWj%-x~lQmN5wD8M{%+45+TQMe|-|4-H%W4Z6D`(Z|8ZY^E{nT*ACi}LSE2r@@dKs z<~=WyXLQeqP~AUrzJ&Q}nYsIgb8?P7MSJggWUec3>#-HQ^482t^j$IW52E|C|82_% z*(l-*)>rWudaWG2qj^-ck~l?dg$TMZkjc5~s{GJ&VyaCZ{;F}I<=MpEq;PhPn2oEB zE@^sX&a6}BE-GMk`_j3(CEz~Bc{JrH5$-F%eFgJYXO3dPouNB1_+Ef4Q%-N)UHU&f zS6Nrqbr4VC&g=5$J+yf~kv3-i4?cEuhmWNFj4x~ajBP!1i@SEeS zo_CbV_mrtqndEs#nL(a133GvWlA{|HyTbQQeE-aY@1HsN4%l-+_&ympy=HSa_|7!^ z+_nD04zhElRfA*AqXNtLYnrJKzZE9uH$RM@(aXQ#3SXhobc=@fV@tU*z=8c`2ll2582_7`IW%n~jelb|<3CYtpPxQ_nQ)MJ ze%hRwp3QmB`z+;9%Kl+3!P3C7(Tb-I5G<2s$I%-$FmUxb#xs?$f>=I^O~R ze;IjvX9n{CdcOy)-{t=p|Kf`<=jtQew-@EUJ*{1uS^Ks?_e03&R>soCSi0^K%H;l| zUJZvI{%B33a(~i>^n@2~cf6;#kKW|SM7}fL{rR_l@x0@W+I)1}df{f1W7|941xe#| z`}qU?>~v!Higjb>L^2HmM*X+a=Cf}DR8Mce_y^=xRGtABCtQO?pA#_tK`mk*{ zcW7LV{PfIqmN%X2o5SoOdgWDgEb)x+bL=K!5y)kLtlc{ErCZ5ap4yIVowDsK zL^vxxRrf5f9dAA>*4dvkHjIyR-jn&D^|PJ3L3FF*ehbNGcy7B7dXJ}WP2<&}*%@rlrK#Viwqc*b! zya}B$2pEbKE1{g1y2nfK7uN>K`vA#|GnezI(u%rV^GqC7h6TzbXU3jX%ht}iE!Bql0)EBPw z&eY>~ay@Uqd#&_U8ON41W#n6-`Z;q>ZQ+bmk2C3|6&DU-dhQJ>VY_?&9q4O)m_i1Gs#hE4rJ3)f=@?F>{sp?l}n zWer+b&hw@GmY*UuJCC?tbV`P`a1PJc^IWoI-Qc{15qwhR{8n3Y1Oxp}H2ndK=_A9m z#Tl4w`KIq|1M4u;j|Vmb{{RE4e4FdC^A?UU?F?F2r%zxi(b#vTRnq$*5wDW?`!hXTDb#O&ElRtVng$*E)Ho9YL1}WwpwLF zL*_2WMtv{J8D-CH>8zR9}M)ZK~fs#0C(++E#A=2yS7r+KrfTMqp;v2NK^ zb*ZTvp{*+2(Q^C9P(JsZxb1So+00wx!ZLZcQ9-Ujt)JC{9OUw z_af`y=K&iBS=*OVMtdxI$f2@<*7gAJH$s=)&;_64{GHGt09`5!UAPN_9PE;nGmw?J z_&jTf|K067^)uzpu2?kVR7FZ@Lq!Jv+5G46ALPGEzwP_3Hn8thF({>0{shGz$KZ*Z zfN%Os+}(&isxqgQ%5DnHICF62c<$J1-5dV@$FxZugS}>#wM2+{(5RZ)>Hf8v5DYq1@Ns$4ue!ynvM z!+fYg{=a_w!!?4T_4(jp?F;E_glviM1b7g8C|VkflUL@z3iigL()Z|n#lUs?*d3TA zu_rstm)Sl_aSe;-Ejo&AUFpl3Urapy@2OW&#vG%a9r*GE_Z;TLTr0RZw@*#`Xv)mV z3obrFyVHCb^S2Pgo6i_JEMI#RoVQ>{#_r_)rkA)gkaoTBuL1rTchockXA5?c=Ih~? zf{V4cMZDyJOyG{tzv^w`^RSh%Y9zMKNNk-q=gd9z1~HM-xQnxjep>0X7q(6e8{Lj+ zOlj!Q-B{o!Hkm$T>(TyRd{{%Vf%fnzedWs5;~n@_ocqyU4{n96{&C zu4b=gxD|hvJOufEYrF2*T1M`s*nPppm04K}BiI&0gWOdetovj0aO)Mb4~Jh*c1aAI z+{ijUb+%46Ot>HzAB(S{(=O$?%{R8+0hy}(f0Ju%=As#k#9RHT$2w!R)^EGN!B{hx zKdfh#YCW@Hpjpo#n~vxAkdsEdRs^qI!@hF*qFL;-r#5t08HVmQ4sCnY3x|s{H&>0; zC(=!I9{joaYP1&lF?6%|Gy192(N9B#(}A7($%nqlXI(H1J+tR;tKT^^Wxdh=5$MX@ z82C2&0cUEvkNqn2@NuhdKxlpML_M4}$yvvj+VbYsoy*?2>WWQ9pYMbgGqcclPCK#% zbQi>-OaAGdN79}(W%7A8jJs4fSwl=;>zRMb&(zV8UY`nltWn}E=%PdLGj>Wmjc13@ z4=v15U}6p3FSK5^!3H1pSEcbSwBD8o6MkS*7v9)@6}oCq-1q|GIH6&Ebxv4H06lmC!1cTQ`ig519Ya>P2*N(u` zM)zS`7BJ7n=eg*l-RQa+bRFxpl{3519XDT%?$cTYp4NSaZuyz$kxfSbrM8y?x2ylO zZ~Ff2@imRT&{5sgQ=dKcmH%;l8C~e<3tgzb{)@iK@OSjoR}JGD+ReDqlKPqXBxl%U%U6h(Y(LeMBszNSRpf1gy}BM=Zb7a}H_bvf8Q!?RcW6EG!OI&(No@+o z=it3f@b@9$X#t)x*}d5pnY-t_@LZQv*R|t;f4@aOto2Fo4^1kszB=a>eI24N;M{7h z{6eT$aC*jmNmAWj#4k@m4@^Q2w9lD$>Im_Fm8^M#@K*-9jB{i!w;-d#=%nv(A7Kl! zH6r;BPDUV)ijn_U@fl9HW!J)wkp1bw_;0$~-!S_T-tz?LdFI4T#L2$O-AmJNSWx(^ zvrkoEMNd^&{#w~JvR?|$Oxvs91!r>hz6wmXPbeq!)dk3E=7{_aTK7473uE3RCSXVa zxbd%`AL4cLg;ZS;x`XmQL}-3vijnE6Q_g-SLF(6eETF9m&| zeAe>Cm*9IVsk#K70vDO^S=Du+RgYuO@y^KjfV)>j`QA2eMme~VkBKoP#3L#`fqi^r z`Q@Fmyc9e>1a8KwJouRjel~%hRQ#l~td>(IPav?G-{42TXF2#uBd2OJbX!Y(tsTml zKhyb?EC}O&ej9rsjC>OQr7xK~tb?x%MGJhP$GD%?jG?;A$N#&tAFlD9UlqKEP5#Bi z`)PGAZRb7nte;<2KgjwBvR<}S4)ApLwZRwqVZrG&==vK2;IpDCU~J^2*a<58kNC@7 ze~pb(TOWld2jda=FUs2ByFcz&gD!hn^S#Q@xFGba`DUNo(6wCqOxSO-73M)xeb+vd z=9u=QBD58yE!m~3fc5O#MvmCHfVM7N&DT-}n9g|+x{*>d%Od5~w{W0%%QHUi z8$!RC$2>vy7N=UZ+MNlMyy>RBzqO>brBAqcEoiQER5=C}OU))0+l)g*n1k^us zhj?tGEAg#k(_UiFtx{X28d+rx4rMy?S@U-p_}J;0uWxiSU%mcqgZrxCp^Yw{ciHQk z1b(O8$CB`zd_P$jA7Eo(+OIWN_2A6rOvn5`qa`qUT}I(>QL z5qfI>Hr5~?_3P3fk;%w`dswGeSn(s@w2ZFk!?~0ldDe+{f$_&%zxVVy_PVrpA|Jug z2f#0N+V5jcJl~49&9>`G|ERunk-g`oeY?I5X~Zc@PP;NchW==+U|(l8`#SUS8+?-- z$oFwK`UBj@z*u$n&Wo4X>)jUet`pmb@3(J59e1NA=p$&|t(=6HTH(>Q;lOt_FfpIn zk#R>Q>%i4jNpy)_Okb??y4!Oyd7sf;NAbaJb>@jptuA`~|HkJw#a(_;kyUE7rx&G-|vC%76;#Z=gu0(8Spn>pL^9f>7^aM zRD$5iRAS2D*Fk_7cmTYUoPFfZqCo{5Ec7d&1 z^F!~xe=}ck6kF`IF#b(qrY`54gZ2Zld-%)-Pofic)^Uro z)(@*Lwp}hh={oQvdr3KEu2mekd{5MwuKTa6`m#1A)(EGPFQ#Kts_* z>sqb9%{mSHO?0uKi{cQqPSf2@ib3e?4;n|I`^D-HIy&^h9=$YyKE=>S`%U&9aG>GU zh@l5Q07H+}`OxF1JUbsfzW6cp_~&GMHQ8Rm=G1z{^;>mkFAa0r_v(ip?Ju`I?2?}B zul}E=y^pcy&XS8>Ys&86do;;D)%;WZrp8>87mQcAHez>nDKcOD-r}_HU9%?Jr!GD| zwmEx#ML-3$=BensfAB7$C!({tHSWFvwmUk zf;$cvFKxt-0rEAqVCNpf&TU0rwA}y>#)1=Y&<;M1f|q09NHm`#8er>621IFFvA@mC zW%=0-v36_)#x`u-!(*|@!I$nkIm$SWF@_&9=a9#%t;Vd-dg6ogtGDe^ zewhS+2m4Jy-S1qL8JbKCPlWv?AA7BxGNi6gqW{B*!MNFmckwy*8GTIqH>iF5mdLom z3+-o!(@#(QU(cBrZ@fnbIPX|z72;)yR~1;>CCdWg)DSXt{|tPE8}Su31H(+K`Cm72 zc3ZZi`F&>EO8uUBCTFF5sgmz4)~Oo<`%_I^faJf{z&||%ouCCcLSBv_2HA@H>L>%f zBg|>dca1Ru43^foRbxUQ5CbFqFa$ZEJr8uKk@3!YdCa54T^PBJ?is=wOY*n_zH%$Wk@l430U+2{f8bW>Tkod+l@->mMC8v?(3+xeE;4tKw4u6f_hb>DFp z+z@1x_uX{&T^{d-Ccc~Mz8lKB5sB}nxbH^rZgk?iJKT4pd3RmnyIb9N*YR!~j zj$ypq3m4!{H}80FqO2J&_ks->PuU>l#wV8fE6Nz$1}QU<_Y?WNJ+aS`ZlAaFuBxQa zU*#_xaxd+DoxZ0~hr8Q!Kcu(s!EWDEsAH8D`mI3WkQ6>s0);na1VYpM1?z8`7zoW6 z7_84RzpeTl^V_N)azAY^qTTzu+I*I?OT8Uf{k$8JPM?ctdr23p8z|#y2U`bg? z8+GFg{dFmYLw-P=AMpDb+Nk6EGl_M7OI_jI&pR`YI?8UL>)X5k8yjLCHFVG3d6Umhik;$F@He|I|{!8Y_KJ-5O^;z~mlFDhFCz;<)j(^z^ zZ+YSYlo!T(K72Jg;X3TaAmy~i^^>0$KIFW}fDhlp#c~a1JU0fjO#4OVtvEeC9>eZXxfzsGj8@A%=+9jxg>Iad2Y+>XO%dl$%dr1jC?}ge)e08w*ljIrNniK0@d=@`jU9mvqL>r|7rMM{N09!A zI|tU=PN$FELtpjGgGYhIz|cQyVKsMJEyqR?Y&;)x1lWkFnZJwk#y=O#%xmxS44&V{ z7(1*?6DJcPj@^xq84B)dz}?&6P4NS{wHkt~DHYFA{ib{N$M%1#oEy__=1{f$ z-1vnG?AX^o{K=LXk%Hjj=bgFfweRMd{n9S~bPOB{M%3ddnFEUXA>Pu&{M_Qi z{Mcuuz2_S2dF#gVd;_1B9m|u>xd`rqAto!I7%k;ok=}PG2R=oL}_Pf@Hz}encn=4Ri)9`~dGC;(cW>{&nyBJ9&3E z|IQdCA9kf+<6D=9hH}oX1zK@;*S1#TN!}!0JC`*JI)1Bkd?|aris6$^d5&j$sy-iD zyCnGJ&(EA*Ss6%~-#T#J)5;Zw7ESG z-x2gMcQ8U5-K)7A+||?(ufrVk?y)J(Q|o3IPOYn$XT}!UNP9K3w;Y>f6E;W&u*oLr zd~ViRf!gJcZ8FSfjfuefQRb5ii}H`m$w!vM*YZEQG%>uf%blklesQ0rqlbFaRzGNc zF?2@WB>T_BOW*y#k)iOCSNpC})4~@-OT)zQu3$ED?TrKn~x#mNz$(Py~nD%RT{r+gud8z>CQYtzB z5+uWM`d913Uiy*k zC!gd*zUN!kKKOi}|ID&grNnGi++GwaBUZ^fZi{w{;s3XG<8#Jm9Ydb!9M#XyJ#V@- zzi~+Dz6TP|n_7w;87o4o-siMW zzpJg)!$a#CQxZI@t*@W^d@8X|uW>&5Ha=xH6e$)89P0bceAoQKr?gnS5kAQ~;)lQE z@TAKRwmsFwkE~N|TI#HlOH1|=qR^c?$h&A+=FpONufYqQv|M1BoLw$0SHd4&T9T7# zf5rC8&Y`6rT28y)S{S6y0_-F5JR*np=F1lza}g6a?;q}VjLw|Rqwp&j>nN7_h>v?PkP34&i*WR$NBv4w>a%wh<{u0yPf{+ z!58LlzQ$?))A+X&`Wt#-{q_F@{q=}1=uL>>*BP1vw=ivC?Wg zI9HU%`a)+u($VkoMT#$fdoOlf4!Vsw^)9^d8vPvQQ}&eLeanG2`My`_w;sojecjXe zU+HH2vUQ}lv^LS7JHO_d`PGGw7`i_NpIs_`Ir%ka;rD!w_3cFBKnem@JS|8pEpVIl z9=sL^;+Me>sPnPzK2{8SVdZw~Df!9&8~gj#Krr42t%B%e@p=>+Gnj3+Uq<_Nw0|A# z%f}PcxM;t40Qd0bSqX7;!dG|pgmAdT8MBv%-8hw>d*+|LU(z*RKmAFs`=Ea@XA(Fw z_sUS>vW(3H4?NdNf5#Sf=s#O=S}EzZTcCgD;@PLt`{=G~Uwaw!uYmsM9PV)N0uRVO zo5DTCW*-3_2s3`=I*5Ub2>1yEtnHQfDlFt$WrlUahs+9G$GMV>tcCb;rr75L{N{X~ zmj^{t@9%u(yZE$|FC18t%i*h8tv$wkofWbED)giO?%HD%E=M}Jlz&0IqQ6dlK{t8* zVe-SVwj!R*I-p$KvP)-JzI`=Irag&2=R^xR)rM_d{2qd}yKTwJRjvFF_YHqjYup;nJ9ECX{Rns%Zl&$x+|PkId=^>h?Gsu1$`??bowBMT zBe1F>C9rD7;NX+xgC{&$GPq_{$>3m9kT$%1XI5gipP#?SCP^xq86@@^LSk`LX(zEglPaDH(M zHU;q$t1_VTNcl|^ z*BK5U>`Jk7maMbl;}yrd4u2{94Nu2ew`|{VGwZJVfa!57{tfv2bMW~um);n9`|QEJ z$9=*0GolacLhbX#s#$YDGsSHms>ELh-?l;9!^pEygE+gAWv}6)Q;}gSw0`p!wmjr7 zY_724dGKM$CCqW=y!HlVgD$|{dAA}M|E15W{hZZz-`nKoevh^5QD5Ix-#a_;$;Wx# zwt;xIK2~8?e|!|IccxkW+Ba}+bDghWdj_9hwzBrU8|b_5duPA5s(nQ;{tv#swMCSP zVQZ7GD}MALEB-F)$#)rd<7L74{wsp<1Npwt&wako-B!PSWBC6daPhwP&VGN@d*q#d z5Bi^Eeeyo*lSm+&c!J!9jo1N)h~v}RuO1p5d7eEz#vKjxZ`g3EW6is)H;b&kCjLS^ zodeDeRf03VtG>=?xje3QrJjvv?OtQ`38h}H^T{_m`*$|2(wQ@R`6qAxo6;K<7IzF2 z-*G$Z_gU~SbHL=g+{2vL0S8c3w4}+61=ghsnlna5u zqK`cPsh^Fqe4)v6%f>C>OvUYU^Tr(s^lq4THn7k#{-jK7Hs#dRnLgzf(Hc0B?<~b< z+j3=pe&h9_6~C3O*RNr@;-Sdt8sYmgzUy-lpa0^^UUhw1|JwIC*Kz-Y3r{WV+n+mh z0_J=lcZ!7uXy1=mM(NNLbZGQ((E=R`Uc;>UHGav_Ytn-8&-Ef6@^WkYyH8s29psg) zW399!6Fdzgk3gR;xk!FEQ&0PG1>#fYhipZ|tN07b*`rK{4yUtrK0SM#Z!L3jpZmSl zvTEgPXl32h#vIA%v-9b>8+>aI6Wg!zh`GM>cAWlGBnSE+!{uYa?rxux$34hTvVK=z zj3L6@s$*`LG0=|Y$B10!`Q~6e!$*Ehc#a&L`(|BY9cV{h^gdI&>L_(qz_Y!0CSUd; z+Lq6`mFI`1v4_n(Y-Jv{(N9Kyrypn)B|nNyFMOczvBmy=7(UUw)EryIdB!N;V|;Js zxo9fBZsq^GpOBZ2nX&X~U%-4k27I~1!DbAyrqtn=8xC(BV_wDhH@Xix3ikIH>v6`K zd8u_k@d(EQnftQOtX_3Ja`AoMf6Jd;dz^8-4{q-FTlG&P;qV2479T-B(!t75?+csS3?vv!Qz}a|}Ekn@>C(cs>rkPBQOaLT1$QueI_@ z#-VtMUqS=P6z*yOpM4sRU<0%>=A-bw`nTmpCi6&nZS1%UWc15jbnsCR9H}hykxcO+ zi${DyntT#j^tv-o#Ggh7Ktq#n3K$MCjxadud`=k`CNIylfxFD>gNuvMUHC(%97cB( z{hhTiAKPZ6wep=E?0+b4ay>FfwEI~2f{C*7W6L(WIp=J-s(y98)#S7XCUV4?cmX@7 zP%mhqoI=Va)Vtw!_WR^V%|r%r&UjoqbEciMfpd-gvHe=z%F(1gkD!muzL9CSS3?;4 z0USWeyPq(Q@4HUM*CfbHY_Tu=*N1Dw+bTDfGWM9QhQWN#mHz>q<+Xuj zul>8uFQvn8y`ft_c=h67{8yg42VMK^L?G3^U*K-#Gfq#mU)q^J0bjh0{I$)0#!q+J zibvnWUyJRKiCr9SMYixxWn%B6?>m5lI-|I+oqe4XeOO0?*Qk!>l2sdF4hY_N-w(zQ zVi(s!^NKr>iS0phXc0eWbOkwU`CUE}yht8VSF&B}@9-MQPG9@VJFpv&>9Nzn zU-LRE{vNr8?>zJPs_9n0+UP!hw=x%*Z`YMt+xxTscYVp{LN|PLaPNSNg0K7^_EE?C z;WOZf(0-Kuq?29!l?N=W*7h3$5+% zHCgd6HemEs@LLHT7+-8HzEj5CpYhzlUg~|z#)j@^eBWnmH~KGG_3N}tYu601LZ=@5 z_NhPhy|gwrklmige7io)3Xum29eOeFWH;RKx9HGUfnzQ6nmN(34}8=AdrjbvahzaW zVdizD($=f{(W?j0t0x#=c9!+Bd@mcuT2tP>EEwN;MUedwtG)xvHg(gWc8A)8jR*GyaYFb>KY-ufA^`?A|tUdNik-;W&l z>HFHwu3R;TTuM7wSIoHQn`y*%XFKk_#7g!?PhW)ygvfl2!L5O7P*M(*Kr z7w@Q-O`ewAH0wk$zUFFlsqEQiN0+{ZtlfdEEymtyo(;^{-z_h}kHGUfxXfidQQ3-) z-RAOhi{=>piC3?Ow^d$xNVK>27W5vO4_~*!*U{ftaqb?fjWA|C*ZN~BGARaMS2F)^ z!_F$l1`7X1{u1^P;d#y9DD%FBwM|TW@XlPig#KEP&jsj!)Hwy?il~2h2y>4XSTP&+ed1F zVTTpJ6*+%s12LJ{zLT+iQ>@;#&2RG?Sat4u!M`kh`WU#~K`xMT`qeXg-KBgdec;Wk zhV6<|q`dmnoK(Nf%(+7wlKQU|eWAsUV0dM zUy<%TA`W2~eEuBvZ5nr26){IfVxOgQUzqfCODp}bcGNyL{;LBUe5Zc)95(yw>@T$e z>t5!maxfk|W3LBY9qQF(3APODzw}>&6a4(rkw@(~lx;`!Njq)kvvX+uZY&ff7pXd8t zz8~aYdz)I*{>r@TN4e~V--V=;fm6vm=A+xMed4KII9Jt?&w;8vCQ*NoN@=`4n#@+pNt4 z(NQ_*s9bbQfo0Af%8u>*uAy6k9osyRapf>?bK&;_FFMlwd%ANd9_d(Ugfc(0rY z!og1V|J?k#L%-$hS2;dW_4nSf57)>a-(??X82k8*FZr;KdnMS%DZ8;FdqIy0zIaNV zeW#6bh@>tmj;DRA7+ssXPkzdl#&u8s9DhKhv{zyC1nZU5vfiNCMRNHoj-#&l|b(+R!NcdB}!WCi^&tnThQyTbuZjpG~Eo0$)7e$s1J8URZwF z^(MyTCSstEo~AQMa8m%R_**gmXbL@Hl<9*;_E@pPyxK;WA=X@&ZKju3 z>Z1&vDd)W{BR|ynMlrUgorg#5 zmAzl0c+b`4e=ae+pmJ_|@XhuBypf{4oZxi{d1rrYd@f6;Ef5C@#D$)4H+Y=3rAh|CPR8^Qk{YWyp~^wiH~n ztgsfhR$GgAGCv!8Sx=qj{vFAH(;Mth{C0~Uxht+Xr~M_wi|X!|@D2P|2Nzd)pO59) zP23Bkc$e!&**Pp}%gs|Yv*gU0j>;?QJ1VcN?>G(q+b!0{R$TZi1{Xnt3v%M7`KFYD zi}B!Mm}Tv&;a_*2r<|!-RSrIOXgvsxH*BqbI$X`2>DV(iZ37Kckp1D_imSXV6e&#) z4FbM!jxTQHOzQ+&&R`>S%9*A+jhy+PbaJFlu@%j3&N}4pi+tBUQ>yi-`Mm)7AvwvKBtCPbwf#B1r^6Gi!5czt z$hb|&$xP(r^Yp8@mFdEJfrIY{$no|cCaw&84-Z9$gIDh3$y3gwMV}A(h!u%aR%;vO z2JL6?ItaX8R^Op@{4VhN`nh-wuvWTYyq0qJj|Zo1dVq!dt z<5yXMeZ-P+=Tqvw3hsPbN?h4NhZkOh7hFVq3?GGbMRc4=K!4FUV-YV{f9s=AYLo-9s4lBM)xu@_;CG&3#ez{;g)3z7S8@({c(nE8) zh520my-+>&E7~(ME~gtkW)*d-$7T~ZaX$I}AM7)CtH+aEI4>H)+@ z;wzAS`a!z&R2}*>T*2>Gt+;fnWNH98TG4oe$zPL-9DM}Ya%-A$q2WtR#gBo_+l)?G z7wlcQ47sr!c~XTu*@Pa^yF>S}rpE8OY3n0TBg1Q_Au~$<4H?g8+pD_k+Cu*Z_NAeJ zS0H+D6>*bk) z$rt{0;^D3*CxpJEW9n#Q$5(7!Q%r2ix)*%$N6_B~IVa#_9k&DhQ&wP2Sy59QU-?)u z`+gbwI_O*b7pgBGr0iJ5j{PdbdP=^l1l$B~Ou)@QffM0oKA);TG0*NlOuV#yuODSi zS%z)@Y+uGe9LRe1(ZF+|{BeRQ;I}?6duKZKPIRqpGgV*%1)xVc^su0Z6&x2zMX$|* z9?G$QGkuAF`>XE`tFVs!$pHR2;!PG-;j<^+WZ_cAnnx~7i#_Tev(FkH2%b5=F58|p z)c*XkY-8Ko@gjXfKH_pB9o+vl_~E7~_nwth6^7V9tMv`#tdG_a_c;1JSlFe9e)46b z-(ODY!r2PpY)ul*8VeI}_I=6=XY-tTS|fcw%{iwLux)A=_R3EVJ9<3<_N@-=TT2?p zxv+1Y;lSR{1N*&#eO&_VTPqsJCcyq)z=nM{pDv$FQoRjH^@54@Hahhd^QkjTvJ+jJ zg|TN$TinqzR&70*+}0NoV4$MGAHJRe)$s9h__%_#s(iMOp&!LpRf-h`uT|;RjnzlV z;f=k%{JK(eUdF<2wETE5Ud_3@waz)%4Xmkl6Du<;V8yF{g>R?}VqYo=<-7WAX;@5!%m=8F1tXRfSY7shU4Z6Lpe;_mX`C-G$j zKFhBPgvvNOP{6Yao@p(y^Y0Yj87ve|-bYSFRwUu0(d6&m>&mBJvKJ(N;%rkVO4xdHU0$+UTw|wzs_xo6@RXwqcXD*$+>tn&Ic#mg#!0m&= zM|BcD*eCYlw%ozTjeNRzB>w2B$mXPas2}R{)SJNr`N&L<;=(R;%RXCCAIUjq;?v|B=mo^(@vG*2dE?e zgJ?M&Uz7YFza@u0{MG6ID0loHVcY+)+On<>U3X8H-Ke_d)J5)mKKe6!mb5AV$3EK8 z?;*6+ga3nkkRR{=xS;%Y(PX97b}Geet+h;nA(Ml*BMuYJ{1qr9WHtr z^H(om%v*aJGy7mJ{U>Y8{Qk_xd{5VyS82@j8NKI|jd}0|jQPHv#=N;}%zZUxet+g; zzO!r0cWBJ?8NKt9jrkyPAm^LsH}*7U_VHXgAE2Emzd!RaU)?q4Y>k;dqxkMW32#4p z0b`E;4Q)HV-GupE*)`^;G-iH(=3^e-HRdMrhikt*I{cH3dFBO-`QLgP^W$A(zDHx` z_h&xlv0Y>SCykjtqhmkWm`7Z|m@9i4b8Xj{FV~p){h5#XhORMxRb!^l=nbE2%?cV;-zA(`WSB?#JxvZC7UpC-n{KgR!?|%vI`Gu zg$I6r=6JZN3lIOzxFYl!y{UUV^rW*FT)>!jO`+}X9%#*vu{3DH- zKBJRA*_f}rfH6PP)0o$FjX7Uq=J#hl<_TS6F4LIlGdiLBF<+?f@XNFlm3=$#9KZ1D zlBzFs`i0SB`(67!!oGP=zC(M@_WIDcM7-_qwD0N1)%#7f6Xo~E;BD*q?E55s?u7AQ zX}c%e#D%e6PcW|OqHTj<tg_dZ~qTxw2Zao<*XHBtSw!-{;((b)^^c# zjNs$pk6%8N_eKSsHULFIH;B?-woTPyPz+1&2#*+NP!k&Dv9E z7`49aiJwz|_k8?3wWo1E-!<+(5rd*w$qOBKcQFMQ#Lstk@$=`!&-AHS{_bq;A2PPp z{FmW353{$awZ9*~v|^(bgXyz^??fx?-;r0zfr7uvj}N-k_Cc?C@})Sq4vO-!7QMS1*7U;kVS`t73<*6tm&-RiFp_3MREFxc`}U^jmGoXQAh& z&ED*9rdVIv(tS7v zPsAbU{6@JU7{6z=(ba`B{DCh7vc#6$-_H#++J~V%fyrd2Bc}l0eOfh`keS3j~_FNzD#ojh)??+C1 zrhWSSA-|7Wy;nWP-A|8kM(~cg3r^kc%c|Ys%P{Ax6+_#nlh^I@i62e0dEMl80@gqL zsbk;@VEjFwVdj6t$%}S8=ENG3^UB9NH=giOVt~Ezg#H(b<0bUtXK!#?&^qy!Vt$8n zKAk8(U}O zBC*-FT(t8GT}%$rC4$m zVh^>*iAz&lv*K~XInb~Ge#y1srVM+oZr&rw^x*Z9SziOIk9f~ zR7{(pYpFfnYGBab+i2_&A2Pv|rGHbFKD=eU-;>K}@5KAw;1QkX=2}vG>sS|`y@9hI zFkYN^KcmT?WboOQ4RuQ08h^SD#kH>|B7CT@n65gM9y>Svk< zZr&>5gW~?iULoZ>ki3xYyK^CTCBqY??EP!6jypc51iwyPnqptMsKq=G^Qd?76hneF_7-Kg7M2`tEzGPoaNxA16L= z$+q?EUk&yn+r|c8sR~$QT$nfiZ^9h>QkNWbU{-8!0v(+>bD^@Za)KiZnUm+pLam+Q zrRmu|BhOq}co95iWZ?jNZ?kNG&xs8-vXDK_>qB!Hx8}Md3%7R4Li$mgl7%U>BU#WX z3wb_+Ha)Ts8;QL!`d1%i*~Rg4-h1cpg^u%<9>(cc{Bhzqx4Pr}uOsIl=d~9w&Npf2 zFCC{>=J$2y6?*vEZsyZr_N%@34R#=(I+Rbm!(VUm@0MkR&`+zCn@_UWK9>^Q_7FMw z#+#gcf%z@u*^_B{!K{QD8gEnR?|#QT`vn`XMr|Bw(LkQ|{piGRx4t+SUpG z(yEx7bDw+hw3~Y%jQjylI%na^o7>P~-h3>|Pn_+^Pb}QHcUEfOJzz!u)xERQ{00~L z9dPphRUj+0rdiPE+}y;~z@YrY{rz3JiFbN(6UVxt-$;`SvMV2&e+0N7Z}n>dYb^O{ znu>@?C}#}5jl^G913ne*KGV1*r?pPIH51zf{jozijj=&@5FfS_ zeY(@Jaq@IFGC+<$+9DQfmFyPz-b;tFwjM#u4`(RLREBo*!DT*U?oZo0b+!_HZR#5t z=c}DgzsiT8Ia5ykGU~hcso8R{tBl}QtcmV@37b1cuM8P|%K5AnIDkJ_M(!|lUe7zaDTInn7fY!>-n~H6fdMWpbE3T7Rs^W`v zu95qY8B-eT^-YXVd6y1lrPb~^{&3Ttuyx?)*s`(_l)EBrVlj71SlbmZ*}+-KrOf%w z)a}6T%_A14qoXhJ^8NVFuJ3r9IC;(%wmiW3wTIwE^@B~X@)6{R+Bj(W+FyeoTV{(F zm4hXX^LU(pCdRS`KgoaY=lt_*1M zm+M*4uT-vvXeq9`D ztUtHXYg?uo`5WJiEQ)XzU<>#V4wfPFT8r%YC>*$R@%xNlFtirgbCEMP4T4K&jCIF) zq?9^3Ypmz!qMg_ru~O#ZROVt9JV3o_^2ce+opW)TJs1BO9LD%=<|1&Ixd@&${s1zB zc{>*4VNEe>FToIVcqohLSE0EOpM*6>~Nu z|4hKx@v4g~Jg)ny)JE88BQgOxOipZzIP1dZm!KOu%&Pia=oxH|*D`g^=L^I+`rCc^ zoZl5+K$GE&OLzJlx*M5w546G_*<^h&)G{AfzX^X-_(RI0+)^3{wLHluyw}2f2{T?3 zOCuhD?oaE@tTDrY(J_Vg-omPD9_gQ;n^kDl%tfrhO^M$C0~Gf-rqVgcYytTsq&A3 z7saD*K$a{R$sNwVU>q2>KbDHD0fwpg|GshYt#N%W7 z5pTD0CcTX@xM!PxgiRA>%n@h2)!1tZF`JD0W^gP2Kp0%b&?5onRRkOtSyub>KXy#{ zHZVxP=uT;uKP3}Cd@|T|AI1&PyBl6}>&4*7W_a>K>or4{b;QeaZf=TrAO+iL9k!Ci z@04-H@r}fhrk0I2al=9MPU_9*mch2poPmtnh>ZIwcLfYXU!-8i4NZr4^Egi%#MdyM zvO~Md4n>ar#4U^cp|UBor}b0{Wry~o?g(Ezr_nyEm1FfX`YCrbvFq@ecujQ5$M>eS zbd-F_B~=0P-dfv-I{Dai-fum&XECvs*=OKo@EHZ4Vek_1;H4x0PO+VgjOYfRt_-v8Fw}_u?LWnlB)~uIJxAH=osTrM{c9>{$&t=6NOO+WRyNGT(K^ zy-&j>e3z^&G4YT7&`#P3BX9fhJQw@xht2%%i5I{t^0v_R8|iGebn})!bgeg{{ouii z;pu_F_@(4Zl3UJQf6O)Ok1qS7v^VyBI(9m8Y9?~30-1$<9lyoVbuJ%_fwq#-<9RRI zmZIxi*|!t9myg`Ly0^0iaqJ1fH66GjuDwu-ZsOef8|HqJft@yvJFjBsrDo(+2l}Mr zAZx~4@p`IRH>R?_c!)UHS*#ne5pJ||o&~r4j&+2J4K1AN#0-x8ZelGQ|rv9~#jw-ctybFKI_uG7}jZf7HQpft3Iv>X1y&l?2 zwrgG9%XE1UUh?u@bvNsU^R0hH<4?2xeF1tIUGL-;7Vqu)pHII2wf)tXb@HaMS;dRw zsh9sk>s%K`KYS>;B_5ax9~Kxs^v`d0_|ScKli@?Z;ll!l4^>7wTJ0fc$41~2$(Ls{ z;5+yLTX(hYyOf_I0{tTR*0ffN&{qU_@P!^%JpYds$HFswU4=^>o)J9SZx0tS|KT08 zwuA=_-@wbo@Iao+0|l;KWBWE9bm^b;T(lQXzXwjWj@j;^tyjK$R`eay4Sgqn3i=*S z^1o;gahdkbk)!he=>C@S2KQbWXI+FGl@G{Y!}^Q!I3=8=RnE_2tUIK;{_Z{I=SLme zz5JJOR>I*)cm4>EH#>M#4!GYY$;ZQ-P5-E0mwc4Ga%Ge9u#=yBO(SQ{BpZCd(y{M@qD_FS9@;6k{Zv(c2d~Y#$%-5^x zosl1Qtm)`TWBrwG?qa#(CFY^AjX9Hz91J5v!|>NY?{nrt7!P_1y-JES~Lw+8#GTQfSUHSA*)=07?cQLLGJ}2^7N;%~MYb6(0 z1~~RThTX!N=ODPm&YY5tuP&Xo&AJJ_p}VNQX<1JNu|btX+wg`Xmj$Ebd;*VLz&ou? zujKy3OZczC)@)`x%XXV})oQKrH({?$YO(jIm+!ULQcL&R@jbFT?Y+xI%Ma95K|I&%d2R<~F z@6w07+sSX+#Bu$LUON0a&Bb%?^KQXbnFj6Wvd)(Mc@;M4!7}Sa8*?|xe3d<_GR@c&#H<{z z;GEY(i)NiF3*`Qq->WMc2Zv^%>-KQh=u_ak(cB?ACiDfxIe}|aCVWXqIYl4ieVy?S z3=I91_kGo+<~{4d0Re3LN^JWGyl8w0RsPU$${Kq*Z*a)$)tJ5SVWE8YdE>y4`Vudy zpMRts({I@oov^{%1H!sv*%}wR*xPsBK-2fkq`s@nIkSsG8EzlJOFVrK5Bc5aWrGs> zzFYkm_^uAc6}RHzpzbo$&ty+OK6u9Vm!yT>RZPMJ_Hn>zN!@@@?@oAc3bnfLz|FL* ztcB6B>=~dZP1)e(=G>jet-fxco{jU%JIAg#D0@8pLcgM|T=?GfmDFA*tYsG)Sn~vf z8OybyJ>IcY<(T(_I^P$B{=@s;8eraMcfKDP`f0*@8yA|NcKf-ZA1A!O($qhMoQ$E% z?0G1k4EZyWzcu((MeF=u!1Ksp$wy z`nGK+#}65>?OF9ZOZ}oly6rcZ*stiPe#?-RiWiftbJsIgvn?mbBd@fcapTpDZZfpH zK6DSb6@2H&N$m9&#w|Mhkv;ones{`Ao`0O2lwW&1aRaN)i_de%jWDpb- z6>Ad03{Dk=R9kj(f;hF>LW@u7eMm@{M6XiqBdNAY3^<{cBRn^~(w2k}FmZq?T(tCF z0w}1o6|qlk>%HX+8GxVyLgfIO=lffG@8l#VinV>;`+47dKKbl(_St)_|N2ksKdrUS z=?MqSA07^VDID>O+Snl#J!5_9lrcXIyMW_*YTj1Q4)&LxvAy2AJl$;Nkh!1Wz;4%ym|Qm@n=`q6uk zWBQBTYsp7lUu2&bicR4&{lu7iVio>S3|KrrRpVw(;5?}0&`jo^Jm#PCShH&N$DX`{ zSlQpv&sJNn;@iky7MLOZby zM`9c9$2J_#TH(mO@W)*?xX7sg9=J{Xehc5J*;!gJoqaU97S6EaqMQ77#Ye;VRAKqu zMIAo0@uzss4kY;0l0$TcWA6n_UPbL#?z(uj*TC9%t1J*Qy5oTk6j|T(EGrr^I($9JfOH` zQoZV*w&Lr?BnPsykp<~Q)&Gkwj%*M+;CIfyt*nMFreR^)$}qP<^@L%h|Li zdYHDr{f~Gz7yh`H>#4uEaldZ#mlwD2V$PZ$`{=BYRNCqdp8Emg06Tt_8*8@s^ZcjN zhzGIe=(e0y%zvi73-nWyz_)oXq#RhiG$LIrUWpB%=rvEH_c)t~SG0*au zv%t-F=3jmHyY)5=0q0CEC;#olvmZ7!gdOLnwlV*V-0%)QqN=dMk&O1lYB6b6UUIuX4IFn&vKYKCScn+!814`bR;qQpYpx0a|Tr* z`XDft8dc`Quk*Q_bM1c4KJAiQ+4n{VZ+Glx_6|B6A@>tlZ|siUI(IaQc#yNVzWw|P z?2lQ-{z#hNjE!OPjliIM7Ms4k2j*l5#6sZ4xfH@v`rH7JWgGKDa z*T{Sj!v-)hb8us95%3dm_hnW8#iv$2l!XuH+9kVXm+U;-TCUu1!3B=q_H6evsY#J6 z&&~%?t1aIt435Lyv?V%R)dL;$+qfQnGd`8(^KAYZ zyL_HeJR3!yYpygougTP{Yfnk|CW`$OML#I!VCy{3OwI{opVuh9gtvd>vqPx2?S>x& zzvgAt%xMlu#D#~0bHG70`wW^nko_sMZX0QG<;;icO`iNyaM1E!=#*}7|+vizF;Ng1nB=7}2X0kt`%X6$r7;1QKF!x-;RN7rCGi|*v@mpgQK z;{T1?9@~?@W}V&{)A7y_^A5IydFP6&JHMlLG?r4v63u`Y(FKNInaeU2&;1!ONA|yp z7AcYnr?n4b44M=* z_&e;X)d<-+t9w~jMi}Q1$EV-@_o%tfq^-rw_wqOI@iEt5W$?Z}i}<$5olnE=Am$+d zD}t;?#FH7+GG+KeZ`^Gy$2XvMrb~_+7!SgW)!;{G#b*$s@VP!=7+hI)z^qRlhECI- zmEQ$N%9W1h+Q_@*H|ECgzd1KvF%|y{yCVzy=>8&L6`tgG$WJnHP}x;H^9*>BPtun* zV(6(xXe{|TF+IUXkll?voJf25OOmk$>_zK%=@aac{F(2R%XbByOviWKU4YK1Lg(xy zZ*KSzzfrnEzUz6SKfKMHKz{eK(3PEe8(Y=c=S#IkQQ8!5zo{AvzRM<3JG$S+cZEJ3 zzN_HxI#!KSH7AK07#ghdXx3v(^Z2E_6hdUUBL5s?IS05-C-Z8enNIpbTZ>a2b#4YgXA9t z>WrML-qMLhLYuLnzp_R%lk*s))56rJY5u$yKdl)3$9zpK^KB-F?VMp}Ki@TGDxbys zt*6+j+G;_EHNv~n3zIDL82zcQ#u27oaA__1xu!Ab1@v$uwHJ-j5wcxYS~+La2US&O z4B4ovdkKz1|k;++OBfihRy4an_h~ zCTRQo`_e2KjJw*Ho?H*r_>nu!A?XMa{vI z;)OnJ2-aM8GoIDx{=)CEKB1U;K+D>-(jtaSZe49&%X-D)Tfjkqv){1xDp!5$qPeps ztYojKaYJ`RU?D?}!&q_xZ1(W5hF8>k7XFs3S^YEJz*NZ78<0p5)IRg*Xq*>#OEn*L@+`qz{doMAH#C2lIkxVzSiBGN% zt0hNLk!)SzBhHXTd*#$5d$M{C?av;p>p}t*&V*i zpDsLAsThruPjb%dG4tMd$sRPyApgSLmxQgdNO?}?0OdBzXy3t^Yv&`Y5l>btoO?dr z)A{z)*YKXRe@Egw@0@QPDrmLix)#3g%DMZl_(Su^FqbFR!UvZ7=kbG6d)Zk(E8gFw zzRk=b{=ZXw+so+Mh|#sa`U4+3auY$%VkcAMHnztaHy`<)2>adE7M`pO*uU6QW7i(Z zwi=Gy9z0oT{3>uY3tZI><-8#?*K&rN>^oweI?K>G`#7z>O1cqQ747|p@lp9~R~i4W z-4?r#IuO|)_G;`j@R++=y2gIqwv%%f)4%DDzGl(qqW1ohyX{Z3R)4-K^fK*qPpy^Tv~ZLjJqC@6qAa>0=N#$0j<(XXOrBJ+LH7E6Xi=_ph4#ITZAb6%yPk7wN`L*jxhL7v z72GmMN1o>Q2l>36H56}CtAQ?#e2w}%_E8yvKfH~Yi0poS*18HC8hO7>cjLfbPAUG5 z6AK8QBHsir*r%|~c`xy~BCRKT!CAxdF3;<8HvE&1&!&0;&R#URs|e>0WvR~**!c8m z>1@_?tr@u>B!2w2A<%mr@WX4){jY|8se5ZRFX4xiC(Uah*7yPATZex6kTIS_&MT5a zF@BG6j!Wzg?%|uRb;n(JC2<;4hvN$kr;dtv1@`|S6Sor%6w6cGTlV)h#-aU;V%l$J zExB3x=!I@;U`*aVNBT{8(C><4=$T>tfQ=gQ0PA~_Slg34kk13~LxgW9zRX&mFTTb% zleUb%S!2cD`3T;xasAD*fuTq%<9;99ab8Vi8lTA_&vtQ8PFPymnUzU{=k@!6eyyQV@9>8DAVPF-IvG4CqupGm~L<&R7v=AC5bB!B(8 z(s8sgiLuXO&S_%o@~3pAUh<(GoJ)Xzxn42uU#ND3*qnGGi&$OO@13&)i0^V18+=DQ zxh^d7J*)9sFLUL?gT)3GwS5i!&89!)>uZNvS4@KMCwt-Jeh8}&$AoOO4y0lam0XS)4&T6YJW)r_I*T00->q8$6d zercH2+HHmoMm8pfSkH%S4GmqUJ*zh(TePv6d-rZ18k#lOt}~wl!$Nbi-Os^cp?PEM zPV4U`vi5Egakgua`^m&=t|eB}*2bAaZJa5jb9?r39xY?qzmqHLLiXc-bwy{IUnU?4o8R6CSAnuaZ~z!;DFIA?C8Hh_!%4tU)XaTW|VU zpCr79zXDzG63tEt9n}85E>2?0ZmDZp@jqC9_}#kf{ubl+vHs9s7u!oejDH@yRbb*3 zqe8vJPvFswIW6^cKw^$*QV%&M$BuJyOp_9FOkwu4k8ZN!Ly6}!uCVZtkT-1bw@)E& z+RL8)hLV3DMjYjSv!-UaiTmgooriKJXWtZHJ2vyIWF(oGoIYbunzJ;)1NF1F#ZBd zh9aq@&qrz3%jrzG;8bN3?N&Lm*)AK`!voS2tLQ_qlxk#Cx&B;uER*vaj%2&^;k(WG z5AX5a(KqCpcqhCJeN}127ow-a%NSdw=S!xU*!1^%xUYG8KELUCy=QFJ?an)yv>T<} zFzwFl;r;LRaQ{0!+`qYp`-0;-e(Mg726w%Cgj(7x&ac|_VmEbQ*w5c)-MJTskx%tz z!uge_VP7r%TjDRL;N`CV)Bxc4Joc6L2l_|B_xo<(Yh_*V=Yj8Vg_AFz$G%S=_dBOO ze~~)~KD)#+>(Jys*}dVRG1yS>X#N13m<&1+I%tiS=EwH61#l%wpdF zt;f-M@uk$8*w4Ry*@?;q=E*SrMm4%UtL)s+Z!_bSiW#V0aEOImAd|Vw!?Mrhk7@4& z<&N}SwcU!x>itNzKlC5G>(I%`smtH00}tAl(MtI9?D5ykdXtX*nsOOi1l!3@#m8ez zifQyk2HQ&gbz#0ou&=Kh3jZ>0;>H^k3s7t$7hNuW?ajeAQ-2{{p9}0+$hnPd20M}n!xCGDPK(doRBHnW}klM-qZhHrNb4DQNXm)C7g=O{s=>4zNeN<$<`7QLx zN7>e!gRzl)j$LQ?`-}R9YO7Itoby>DcYt5~wxW~B9ecnB_b1n)0gc#koTqx_yYmP+?Gjq}7i;56XS*uh(>(QjY*gG*m* zwuZorKPayzT;2>W!*f2lAJ;!h-j0KduKJ+g=askZPQLJAv!~0M?ecas))_B8k<^t>;&0{nsJ$fTQ{sCOdi@&Sg_pGsh&4$$oWYXgdx&dA zAjkI`p4;gwF|-T#a(s94{M|f{Tt|^}ANCXZq7vVY+`pNa+Eji+A4HLNt!MWYbAK-P zZ{z+|+(+(XLGIh!&*%QN+(%AT3!$~?zH!{2!2PSaKZX0#xIdlyz*Wc`?<-@EER_1l)6m-j|ka#&l zEMl)8UH9l`I? zuU`Rn$)0j`s%ME^g^r(yeh(^+L;ME4+<@*iFb5rdCVv5+WvP6Y1>mjxtahK}P}a{b zaeQ0!zT)h$KGcWeyEK}ZQ0Ki|6h}#np(fURj~6}|kMb{GEY&;D`o$jfdpdFJ#Q3F& z&+W=L@wuWm^0CFq{Q-xQd(5cU`jnXP1>Y-gru#3OJ#U7DHg6%00DhF$jWGX6mKu>M z#UqQ3m%gRFH4Z>e3*XEtJ0lbs!86Pau?*gU{}wfj4E?U!Tf5jX4`eZ}-**H5;ff@u zetin_K}AZZ`5^o@{fg$IKe8p-hz47zA$!Esfn=~|D5GBa(`IPbGMWE0>rKwm-IexH z>9RCm=@YNAKCYH~;-MDW*hhO>t91>r7{zB^`v$Rp_+=E=VLVrRwf)u(a_KubD|QE; zw^%9jtKGi5xUF(8;COdYk%@usE}mo4rgQ$7Xz%FE6vOAI`<%vBJJQ;4G|*WuRw8@S zi)}TGpD{M`oW|Ky#9khA7}NYu;zhQ{-+~v}_19!m(+6Izo5VSP;H8c@*(>0q7JM|l z;4A4$gD=w8nhTx2M&&wvJ!~iRwaHh~3|`klgO>hQXfOLUwGL)Knfc&}@1hg5X?=+J zn)b|TdLg%D&DNEBL@!Ie{HELz;Y!bpJ=VsaGWb*A4c}?k{y^HUx>FB7GQyX-ggt$B zi8uDa8`^i~7x1WfOyeZxvr%}|7(4M8wZt7fW|&Q*0!VRU3TfX9K$w+zR`+6=zm`Fab-ha z>&mvxiti1oYyWL~a%dXAS*#2B9z5O^590Ap!-J8vE_euohikGNSsO*o5HRJ$f=M}T z%#&?x>2BZbffw|LCeU})h3v<71G?lUczg$(?j8{)*N* zzodPqk{1qtQSrKA2grR+vtH6?@o~3v#%5@;o&TQZj4kMd?dS&N#LOAQ%{6Dtqt;q; z#%ru~Z`X}aTP1(bzSgkkjQYBcnr-z}$J+PyzSdbKI-94d!r}LDsl)HvK5UyR9w$By zkDnEK_22Q09|ZPNE57OfLp(M*8G27gKZl|5GJi?i)6ykp*Xdr5w4Mf?XM75+r$g%* z&>9^hTJK14br!Yi!vDkAr0x2YKEvo9qfdj$pC$9{Lz5^vsGTO{ z3yMUO8v0m)Zdu7VmgSW+uLG_Z>^OTJznMNTZAsexd*3?nG&*JO$c4kxt@wz(j$YFG z`eDc0#(kffn9%_%J}lXZ!~gmpKwIf;*^IsafIW)d*XOi^&&lSq^Nc^?ew*rt+;>lz z&(6E?g!_L@xc@=I=RZv&CA%6UGK=AJ_fFjqU#dl1|e`PJ1`r6Gr$}&crgCO zp!~RhA!`kkN1yKr#6&Xacfw)%zY z=(n6$it3CS2ix&hcrtc_U4k#{JfD#tH*iv;srP=)^WL-6dACMfm|w?MtG<|g%a?5f z?>o%XzY4H6^cZVRfiHkhog|;y9ZM5qo{gPTQ)Y!0f%m=Encr_BHc)N(>Q{n)_D-Fe z^+D?R#PN3Jg%i6v^TLmTU12oOAAOBjRd@WLSi;ILTgxj_EzS?O4rik)^7>;32K>8f zPn9mKwSN4Br5HMVQcH}0*#3rQ_(eA77GSHTzDhM5=H5_jTGsfi8ECz^gLl^)yk^M` zecpHP6FZk$p{J~rT`}H^B^}>1f)DD$4x`{2<;<^7)o5T57$yh;@Tbj)MA0c&r6Ji1MuAMe2OL`L`4Pj6QJu+{F9e zvnBgKe*T;;GN;&8bpnp-6YzVu!if)!x?DIut35sxyBGiFK?{4z*C~GX9ojbj5cZ?w zdO!G7O~brPt&QQDKu+XsUrF>GUrEan=ly-MA4{>hXs@Y){8vqG&cgxg%Kg}iqga=A zK(?YUz4ZV-LzT%NWrTj16R+Ir<|F;o-{ckeO$~aKn7vtZBOHRua7~c@a_KL(MC%zF zOVR(#$;}U<&zFGn$BCIh-~IG)fc|cvzk~4bL|<>S*XIFf6*hbP^+xBVo4lh(w?A!$ zZauVl0D2xoN9@NIYDG^Z!tw52JLj=^qbIcD2kgfWI0&sr;R6_4O$jXA4}Bl*Wo;DR zMh)jGJfvHpiR^n=pMZlLa8Ss(6pHt3OQJ?5^ZXuh@M+^0zUO)PrXLUAuP5|%;pf2j zr}cGV0E`M z55Jc?_}$N%8{zli^E>1Bk%TcvCY(Ngd*TD(me{Z1W#LtH6fY;zaVPzpE-&}gkK%~( zsl+Q;q9J%S`iAlBH$2=6-tEBY=ug4vet4Jp#qezF<-vtF%o}BJsW|Haa49|w@3L~l zr~A_!Jl2u7SKkwptzX*MQ?nBA_;>;yx5B4c@TreHg~_iGe>3_HS?y_T!r^PgVY}*s z(FuM0uLKD8~0*G(kZU5kDiXj{Gqc(eO#&aImB$2 zE-E8`eG|_wL0@6(Ha^ak-x|?bjZ?b3TjUP~c()N+yv*Fy#Jr$2CHwhTTtsWFBhW>4 zW*Lm_T5#6PI9uSA7RzF&=+^fwKY36WBT} zgPAuCT`sKu>ubzo**4GGJc|v}8{G6mw+}$KW8bt4Mz_PqH}_B<|L_@|?Sr`SU- z{60NlKZ0BLIC;5Awhj7efAZuiJ7*SpLw%-@-&DVQ>DSfCrr($7*V7x_ z@qA*xUOw{(8+bsqlAbIKmMj2c6Fk>~o-na^ z;8sqy8Tz!q`(fHq9Qs;nOikPw9GE?xc}IIZd$>@2oa9dS`7mJ9XXPu)fkFAoGHS;X z?U$}R*SNmJcQE-k@ddPeAi;i7zWeV<@9BR7z2AHDi_`n7e*?XDyZwFq4)1!?VjGF&Xjz2?8m13JUNytgYZ>8d_D<&N7oMY^qp+zC^}bp zE3M^F4Z$_&gB^BmjuX>Lsc)iQzo;)Yg5vM-%wOn+AvUp0_}<~mO#C~4eTMLL6YB;{$f3s;aafv>@_$t=Pu3tS z_jK?5`1!Yi$>0uLMwnyFI*AG^NBX0{<);Wb-{kfx?40?oE=lw^=b7{5I_l$<1Dkj> zaQGm$gWA4dHG@^7sl(DbX6;#B_C)pdr%c}S>)1zT|2gg(-L~EF@!NfV_qzpeW#KPR z#1H)~?P{$+fLf1;_z(V6+#m@a6Td1ysCwM4Z79x@*oWdaGps=C=#Sc_E}lDc!WPTF z;8x(dhWJ7s-k1OB`I^a(69V;9T24 zE^Mf^ppG+S)-LkB*;Z^VxK5vwoX`7DwoTQ$Ra}$1g}Dn}Tsnz9$9FpNF8cf&x>Tv| z?WbR(?(IjkzBq8c$sq=*CCDf4Az9RzX5;tQKtIJhuz5p@bG(b~rC3N2 zbY$M!P!&E-~I9M2mFr+0;~P}sW&)>$I@C`);s`rfH+Mo(5s|FpWRD6A?qO7hjy;(f(rE zZyFzrH(wq!_3VAfcyhl-?NvdCy~7+du9ZBy#u2>5T{mzGvdr_rf}|YL#fzUMGtS^@)(Sk; z@jiTWMeKt-gO3R92+m-Y(~jhFPbQWbOA(|Pgy4P4i8eUWF^^M5IOfU?ge z^|(`uIBy{cEn<_Q>saT!#b=S>M(SGfm($iyi7`^^qWMVls0IGb^s7DmCM|Vr*ZieU zoky`#&#-yjk%(rq*HotiAd1#*_Yd(61*4L4&P+Muke^; z$Kksnp>M!%5%4#WdiCh`y!enyJ8&QzKI!6cIPxqSWsZbSoATlvgNh$UupyXQX9zU?yRShPrD6r59G(K*2l|a>0 z(8ybolfM;y=5KcJbK91Dcp;-cz;C*4XN}|m?C*pN~@9Rqh&q3(*u3`kplkx#$q0^FN>&?27F$=X1ub$D`?q>LTCv(&s z^xdoIJ?Tbb)4S(z#)htQIb&lO;~9bs55@12uc3Y4R4ZLF|77LX8TeuFW-`Ce!ba(; zmtOui#sW=ZjAvG+6`wWK@CkbsIp&h0%Zz8|=TmyjR&kpq{9>jS?# zQ8^MB8vX+Zu8}{m%pN3qUU-=u#5QBDH@L#KGPt?{{X<(L-L~d1#rO(0ob2)P)t(p4P*v?OB^A#~YFYODd>ea`H=o`j^nfG5D(yoK(PHF?4Q) z!?R|+i*#Z1lA!S&{{^`9c?fu@fPb^W!~O8@CGeYp{f*BLdtn0Xg4u&#?G@8rJM6nY z3Hu|T1NJ#R!0yb^*Ox5$U0bDW!2SPZ>g7rz1A3Jdn7x+!gkHRyIkgSE4!FLA`MX5^ zNA_Kh$`+XK`x10P{^W-gBY#?x=*r*vVn_bg&w|&z^$)ACJuZF7S}vYY4ovuM;_wpr z=3fIw%~|@shIXWLvAyDvbB#lnrf_-N8995zaIh&NM4teWtas3OIhf z_@l~j??Bu#HG+LYbJcF|U_2Y2FpK9)2l@{`1V1ZwvATcR9@)gPUD!>sak#gNd+1Sk zJ2Uhr#@Prg;{6CVLv$Csy3!wyEkg!c{PA@00Qf4V-9tsbP*cr-P{xnTpUUtjx3V|? z)D_f9ms0}}<~POWs_56hyll@&be8VvU2@Oy>aw#!(Up1avewv=XZG_5a}Dsk2hT>} z(etWjYy{pY@+KRsf>@04qi9p@$7o+|7l4n2!bfTcK8AviSZXj{2tJB{dty62dP0`|(*Qgx zxeh=h$$FrVqbJ+>54_4xfAY5-c3CofR|~Hu+thPBo)=HX)@xe972PhLkey=ex)^!~ z-JyKy6MCQTvX?U9UESaLD*6(8Nox|Dh>gGU8!P_GZhZMyt@z!%BO5~F(|BZ?$j36i zlxy>U(cd%h6DJl}3tolpIxo|bPUd?9-wWt}B7PIssT1$!Y$p88M46Hg(N&g<+GROV z|Fs9Zm*X43Lw9mqPVC^=B6uOt!3))2mc8@0$;MgT$wno7vjl$F>yPJoyinFZ^dSAj z&|iiZmbko7*_jtAq>J`;(DOpklkeTq6aLVejzoIOhS-m;iNN>Dm3$WMu=&tC8KDB< zAoV_D^Fr87yP5-VR3IJPGGw&7lm;Yd@*XW;x@6h8K-*J&A=y{kr~LD zfloFA|FJP6NUbS&zS=IShMsG%S!=LY6;GM}lymN2co4oV&uPA0jQ!a3p+CMTllaUF z(tjEC_kIq(+r~JmQ=M3ZiR~ZlciOlWZ}+6Hh&>tJ@9y_3biQH+nbg$JM8C!~kD&k5 zUbXz4N6!zvItkp@Aj7Tr^{$Lm3gX5E2@W6`UrHP7hoG2mLC#lz?DZ>Op&d?avibpW{47eGboUHh**)NT;H)DtBp3d!!a^JHR9(e&<0o*k+ziWQg zejJTk{PC8}(6e74o=m%rZ77>Cx&GhK_nvH(A%l@r#@z4vl1A{})Q|YhW_cm68MLaZ^-YFj1ru{U=v$pPX_SBfc9(-+m>H^FsLH1yiEXxjK?j5Il zIpB?&rg#AVw1{(F0wdACm-ylti+nk&*vB+Bhd9%GU>J;#T!fsH^N!vjc_&VANpAe- zertIP&+Oxw5!78Y;NP~D4ya3I--#A`c&Me4++lVg)Utvr&$q1P*_x!#K49F3PTvQ7 z8nb>2Ps4_r%sd~Y&wJ>T_6PRE3QD3%ix-Vz{9n@SfFNTZkc-=L|CNgUy zrJrK_Cf_F2(Mzh<-Y~~jERt4~EGvnL|-q|-YJ{SK@ zYr;?12dVPO#63 z9l%y$j?p^F$b5GVcre|>(WB5H7aE9Oidps1{{QevnD(^a<>zfLJdHl?VxA_CW%g~m zhj}?jY=xML_-G&SiAN65-wl)O_>H;LMc(a1y;h#M>|W{*`}XR*hs;VGRr z!PrEn&pkH5K%C6@Brg6s<;%Q#jMLLicQ&%;2mQt9k2;qPpSRCdW9nQtKS~Iu!_=Ec%-qZ4W-I11SUqABT$nX)Z zSTUasM+(4qY6S^cJSrO<;@{X8=xtB(FECGpDpvD4kX zbr0n&$7laL$d%7sW17pF`lGSrPW75G_}kQLvVT%|3v$L9%*b}GH+Wvy!f}s zH9i8hN4-PGFE&1c(M#C%S46;d6u8A3jlit)^%TQ-lDZK6mW3Sa_aUi_i)S?V>pA61 z^z5(!aMSmwlE0{V&Mkcyv<@Fm2|#7MXPA zEBZ9|fG@hvid)pPb?n!kOisDfoZFe(oBi7VM*M}zm!1vXm6ynXuMfbp4=gM=POQE} z`TaI4t$mMT$wc!F*hcsVtIr@ta0&Ri3Am7Low4`|`mDAg(CI9%nc&>9oADn+EA*sz zXl%(B(+Oc>;EDRY7{6rEsjEY?z-RL%==7WH_~T{P@_js)hON|A+OMvyv_Jo6*0n98 z9u(Z|zgJiG*eS)nxZGb7u3$V5g6|~`ZPL&i{&YK@o=i-T+)BF7iF4nN@4Igk&v<^@UcHUmk+`b8a7TZyCq!;5qOf7-z%o^tJ>{RAd_-Y^X^M2PJ!mneF>%8AimM^eG zZOHBlXuq8oz`MH-QFMDC)Stbv#Xpf3Z1!-xsjg}8&2=r=-|W2K&NBMLzPO*-x1P=j zgLWNfgaMQ4DTWPWt_OclVCxQL?25%c&gZsQ=(C%-TJ~n7H+VyKWOts~;MkpB?L=YX zdJXj5$Nduaw@3xI;3WbN*}ds&uoeH;Gp*%?d~Y0#-bU{)zMJbpttCYcm|rx<8e0;$ zs=RgIfsXadZ?Z1B5qPbHKT{9j)vtfMbNzZceT+^H<~-8Z8b9g>cKkZd$_U@XnBb35 zeRAWYi4C;EC#{$Ga;R%PT!4&d{!uR4)W2NfWA3+~zbXy9-HYA8XYqXa9@?_(J)?*p zjM~8b#;@Q$%&kEUk)!3vskK_#m{u%XOHe>CGCn)yr?I+%B{I(40fxPw9?!EzrPENe^zr*v6 zj_cD}f0lfxf`u1fAs?!*@v-*LcW~$fM~eSr(`>|N#I73^$^;)L7@KU>7ZVY>Dpj1ZRXn> zOA3xZhCXpW+wluGEWsXtm&9)x@4*|Gy9#XcTQ*}K%Gig&hk|<} za9hZrg`A>`HzH@)6{Cn{NluBacydY~-SGO-SqZ%Um5cF@>2nBkjMn34!^5%5@%dLg zu;=B!{_05n?stw97wtQ;=stVJqEp{mR(7gv*U&%OFIiU7lJHH|OGEF(zPyPsnVO%o zLf;vS4#)nG%^10yE4Bc0=}W$It(Tb3mlmSK%N28Aer;r&%1OPugLnep6;sJXpX>hO zM#kHl7z?(7{NIXw<-zi!{ccXQ9W?yS6GJL48Uq10htM#=$&iA6; z`SGH;1rwCRjE=$X5R4{2>#tw<9CP#>7jD)gjs))580rdvTRAwvs`;h@SX-F$_qnic z16F+>;lir>>4(t~z)Jik-U6(t$M_t_Erl z5jdJ`#iK!u)nAV-w|>#hBSVv9%kiw%F|^NBynmkFnZY{~dFLwn;NCv)q1eaAy+5iP z;Zq!m_|Y`yx0dNVhhJi9Sm6)$e5`L~O>h|4<;Q%~&stsuFN(JuJ{Np`>e&^Kw9LHb z$eUYhTH(%b8XLB3bTVxR85eDteLt?U!Gq>QH&5t3KZ2S?hXxmh%-mvgUPJAPN=I&z zS|33+XD^(&umwA;%}O!zl610YTZHb_wHVziIETSEx^8P}+mKz2eAGfa?iIPUhMJ$= zZEl-76T5s^vG&&Rw@R-xg69}?*IYBS0Qi~v`=-Pzu>;3@@)v`?jf}~YL#+=DLs!{0 znN=w{-gDmXU*dOf42na5J{ulYoM!Yq zeowXHW9PvurhkvVz6APyal6GEkH&suN#XIIpvyhGLSvU6SN+wS29J*IS_O;_Pg(Up zPx$;KpGEU9aLcBaJ^fYIp(~Cnd-@yL(iVT-dRJ#}ypH{si1sW-cdP}sPoopof!`Ol3(&A7-3NRgKnCx29p-qzQI>+AB^z zc*X>L1l|b?7X0fNeVh0zaglHRfb%iFWNmB&rgG^>JyUj0$opN-UdJ<4#Oe)S07Gt7 zpPU%51&QZxPsb{jh-oI$=c ziSaBNON+pSq2)35>g$Tb=iO&d&{qNTe1Xoq z;63fb=D{F%iqjc4xYKVN;ZN#>H+XxJI6Um^HJbQ2QVC5;`RAQn;&IddOw73yo~ra> z)1Q1N@%_Zlf?+awjxwE-%;$yRuYt9Z24}#Uh_f7EEjz=3^$}n#XKhJ4Ex@6ky&Jq1 zuokU5oNwhk=FHwsKflv+CVxk43!YFtlX4l2cPQ7u{B#$5&iHmB6D{j}B_sK5-x4c6 z>Spx)J?PA-^8KuM){TsNnH9f*Hd@UXs*yAQ9amcd18aj8h zyEk3K7$>KAhim)X#g*KfXZu93C!&w@?h0U90zAuz!2^$M+D$K$L*O^%5cIoZC3;5r zcs)0)ANJ7z7ykROK?V_*!(NG25T_xRqJ38U#5)4yOrq;@OZ<0QB}til#;){_aUkM6nsC(j*u_Qu~G>0A7s zd2cIgmHP1Bv$s9E=j+csqkHqSbdUE|zUf<#$-a4Cg-0e{Z5=M8cG~p+=wo{>diI$k zzkQ5$UUvGgcKg5hw~y_)`1a>a|32M2*Dg5&c-!Hb06Zbz0`$k!+n=M)-?{L(?c}l- zh~T-1b}phF!Beeq(@tO7G4Sa-?X0Aot+dnD*0=5_kDYMmzY9arJHR*cX6C=N^FxK2 z|B3?b^WR&^r#N>|Hc3Z)MLQn_l1v^fh75{!`ON)V13YOEGBbsB2KeoBsJ&%BB=W}A z;pYAiJlQ-k(Z-j(l~VuXr;(|N$lDafE(4}US2cK=Z**PFT)haN;3=Lnxf`B~PKMWl zz+Hs=iMQu-4?FQ~=9}0w;0P){lC&f`xufnkT;$AQx_WIcQ=9Jd#>S^jwVfp2iS}fB z%Qx9dyJ6&Bbw4rWywtW2AAsIX^i$3F68lk%-o)LwD*o;DwNQQWuF1L6mi7naY@{V~ zPF=g~!MVdGY)`UsX0aA$1~KmY;fILECm66vy*B0~v>_isFhzhtFgNjhBls9joAAG} znFRY2f*m}V=YVMdc!l2UtqtQtS)#Y0uX2Kc`e~;@@5Hvk^eZ@{^y~NyltdoX95xHPpa{EQQI!+#l3nfC-2vlAtXWuy z-;`AE`5e;g_z&xg96w@xvFk@nF?^pYI772wUT%_LAHbRQ(NhR4?Mb(*1$@rDNip@nh!{X^-Lz$FE<3zgBQnZv6IRCn~+q zM~;FMe($c`!p%YEF*`n;dSI{pANJJl#N7BauU$U>r`s<1(747?g#Q(Mm9TMATy$e@&QNlNr*dS_CdBPXo?x5GN zN5m`h&@aaRLEq&4v3K|Ma@)|n5@8+@9vW$HF|t%kexL%ouaxP!w($M6 zY1Y)mH}VW)TE#v1DX#kcy~Gi=g6Ar}7xMqHFJ5T}I2(^PrB8F=p@$RTd;&OUCc^o9 zXK$KM@RbW^6EIY}eb#_ewRtx7t#vszF?b&WZ+{n^kcUo-(Psg6tc@QKMVC3eFu~MO zi5Ft%Z|Mo?EycQI&sJkQH1pn0#*J@$cog}VSSj%;?Alk6^Zn}MOyUN8?f5RnU+hC) zNgq<%;n-zq^f5_OT8>&72|}%|Zuk>0>R(;(NK)?6St@wHx`ZzrC=C zy=O$rz;CP)!PS)qd&2|UEY_7HbDE1yu8leEOg$_50f&)6kES)y^Dh2rQ~WUoyoc|y zTr zikSoH_dC7#JQUw!tpi7*OudaSGcy}Jt?TH&`5w(*P7JPPE%arci7>{w$X%N6YT!#V zaIHcoSng^81_MK& z6AU)}M+VS8bGhdD=nOmlF7G@CU%s(5H|KHIV!cO<=6z`JUb!#lSRf^I6xb8v4IiL$ zR_!6s%zjB7(L0o76FU+o8ihLx<;} z!^O}+x$wHLJGslMuUqjiLFZQbxQ#y6(f)emFZnFu+vi|Yo{wDe{Wb2teja_(pR*35 zANH;6us-$w<44x=R{ATXznN}-x6$7-^mmrq-#oX!0{VNGb&&h%FLRU~ucz%G|M|2( ziT2m<+}dy2@u$9xenYQjy*whc3fZ!^QJbS!13X|2aL#4kc9wG1%#GBU-U1(npQh#z z-ZT6NFGMRX<9`?Nd@XiEt(|0QBj#`}@vG4OdewZbgf4gR4?I5VeZKMsYC-Nmzrk0E zLrAv`;e7WPW9?LfJ@k^$oUc&Jas%()i|o4d_?5sBy}^-z1pdE5zAwB#@O5(Y`aA?0 zpo2a+zq>l4lMm|NYrG5Dk6`A}-M`r~c5Ksw&im3=ZU2X{qPL@M{I;1Z zXWYz059C=_6U(6<^+E1&&W*(y{Pwz+80VL{buSfuH-6X-^Cxv=6nG=R?ZF(S&+cGO z)T`a;TD>or_w)$!I0xnzJ{{)ooCeGp*sf9dSusD)POX5KgT(y43;nX8onn5SY&#bR z!GZWt;99zRjbc3LvNmeYvhin|*0ENB`ExaM-75G~dhS=7+A4kG>kMB>=|6c=an0%m1_YgQp#8-3$IG*p|IHLJqIG%4CzajCtLh&@ZA`_lI zL|?0jbA`wIb0X8Mcz611JF+YNW%SNW{S*nDS^ux-v~UG5 z-vr-H0q$vbyzJokrH0(5|6bxWaN$WJH zJdM4d^=@0|IH=1vF=V8%#-X9~(7B@FUBFV)#g^cis4GvR^%*@p=gC_NzqhqrSQkZa z2BCi*d>8wWHclG2ms;mu7K-S+)EddPlaF}J*IBN+uU!hX7GPtw;HUIzYvbJGWMd!w zb+BU(M8TtUrp|7>3V$OTTv60c9?QwU*s4*Yzpt6AbMLBz&l3_pU%_X|r02){iS@wp zYu`B4QDbEC6M=wBnHO1%eoNu2Zg zgR3;=z;x!oUd(}Gux-X-<6LB9s9*gl<~-G{9YA(kr(5y&!LR1#!@za8!is;;+m3&L z&2h9Jd59t_em-sdkQlgPM~(1(>|SWVykq2c3Gu#wWK(;+X?;c-_GcgFo4(jyXE4WS z)MqLda1YNfvy2_I5BSgjk{!R6_cUJ_J7qp|qn?95@VUW0dg8k5GOvDnC}1zxqsb2ISnpx-}dayB9FO2$=tnFzDsZ2$)Ey^i<}Ib*>B zu3A!-k0JAM z{(0YDm{gLQ z$;S8hczC4C!{0(CEdIQFqRu7gx?h$LzEvGrdm=TQTF0+ec0W3 zWBV8%7M<#xB`Nu33}y}q_znDOSAEI;{yM%~azijDIhFGf+Wd8G8Q_oH3N}h;3+t#Y z^n>(*@}DeuW2Tfxl|>N(3KF`Bl+% z8&}1J3z+vc{@#qefoFDpbiC5TmGO7Mhiqs)gY9gtxdW{$)%F_d^HQ<5RJT#io^xxc zRhok?X>scywi4G|W98<|p}&@f+?*X&>gy@g2d=>uTT5THJUjblD?Vpx=dnb|gG<(D zGhU5l){V|sYH2IE%$H;O#{XP9mNtfh586oW(#D%^8!2uZ$!fy_=I`-d0q+5S%J+D$ zpmX2bz4!Rt;PDox?=^g1%e%Exd2cT7-PZZNjoy33?t8`Vdp7SuGeheMyywKo+Ud5| zdvBck-Z*H-^U2sYMf5of_-mot?3;=8zyrz9KEnD>`7ChK+Eby^=Mj@n7qy67~u@#};( zUZIU`i7~SYfzO1))AgE<633{Twn#R1YTvRwHS}Sg%Vu7?#I@^Oz1l+EP#XL8WKg55 z{mG(9l8Ka7tv}p120e2Xc!X#C$V8@Ug^-2FI3 zZUPrCz!QukHcfn?y#np$OQ?OZ$(frqV&G3W97!zAPrbvP#F_g0M-vb3-)%fNG6me? zTOC?w#|LhsUIClySfKN{P$Be@{1`oR=JnWT*d5^_@HYoKU^8jGVIww7(>>S^_ku@k zjC~cvVjd(GvxHd8<#{Csu`T+3gFN}S@XNjfE=sNCho;b83U%eynXT2pkmRR5U{}0z zJo+g*oqlfMf10Dmv!TCY1By$?u94q{zr8fi8qK~U+@n4D%jB;PwfoyOoEz@?+iO2Q z>G*v?*Oo|(NxJt8?|Hw|uXsB-fw#|q=J#{Yo&R>{jT>lqTl$qe!_xLVgWE+^{3#jrgeyAA(OYx86aY*PF?sdeW1MNewYL-+;Z(`vPu3syikPp%TjmH1iw zn3(@kp5sl{G>nBEgKoPyHz(N3y3*vn;MJncU``D-j`o&p zM3(*JLy(R2PA(*={zt@N{3+H(hn8Q4SN*N!$Z%pzv}qzfKl;4}UF_MCtI(I%bLId! zjQA?_xz<9hW*+D?=Iu^xRs`EBJeYrUM=mzO(l_PTC2?kTa$R&Vd@}<6$>55v@ZgZ{ z>BLLf>}BKf(m;=w-1a-gT-xuEpM_U`Kwq9s^JD7EW_Iw;`n;RJ);<5Uw;|uo%YSv$ znTbwwTsr+6+IaLB)l2qoM?IY%yhODh1!FOGF!)z|>kw-i%kifogZM7ofR_);iG^PR z9+{uAnHz^PUzvK83p(TcU1~UlbLIMk^Q3|J00q3me3?am{=3mp3p>9P<{kMA{yUv_ zL<`@YChm90vc43W0A57bLgv9jYHYN2Mz+fDcu$|zAK&4hj(Yt?3z<)`ha#feAoP&2 zx0B3TkWM@hbnylr+j046_ZXA#+t0->Oybt>IrS;&)GHqicCJqeKaL!s7p0q|pQW2e z5IZ2gx)Itl#_UiVXX#3xMIN`~5X0F=tY$y)WN$vS zQ5;9Mm~52-=+ng5+`PcU};6M4WhfJH&4b0Pxm0FJyXr0-x=*jRh z>hADoZ1mTBaM78LPEARt_-xR*=jb%|cVOwN(|*Rf^{zT?NrF!EzvxO)NndSS~+XUs^m4@Y*cd@>xk%iW{a$ua9U>9WYCl~3xK zD9;SB{IBB!Ul|?%jHOnN+Ag5&3ureyz~-KvW5(jY(|z0KfaGTfIuO7n>kGD1{-;%nfK*;Uc`C3k$aGtt9&M>Iun`}r`d`b zE-hHcej@N*KKqeU!MB8W^LRE$ zJ4MWq+kMA3<bk%_G`&p5HfAD;4TmR=@)pjrTv2Y1{43vEY{Ig@YM z?BrXv(Wi+oL0icWG~1o^Yb(@y9Ql?zfUCyn-KC0Q{5b=k;SO{*@~!bHMl(}2jx~<{ z{b$Cf&x!e#D)NpUc8&vA3-Ih4NzAC0vq+GS{hZOu8o@oHfzEK)&l%KpeK>FUm(0J^ z-ulv&f44SP%MR$WMlHOOF=E?kKaYQV5T0aDj}5xAM@N}iLpm(<3~&ox>7sJrJ;*xe zcQ45``(R80W_?!;z3yEEO;|e}@7tevB6-Hd9OFNs!=!K9TGC|yko^5TVTh3@&oQ|x8m>Z1fRS34O!U2+R7Tv7J3dpL~+vh183}d zpR-}#_xl-V5J~P*bejXLjeasGoA-iU89P@S3+CS%3#k zO>vho{M>!kt0@i-V_x3`TuVB5a4LLaVw{`2T40B6y;#?GYnnCwbm{RemmW8Fp~tnK zLXQah$e%7f;?VAM(*v44mkZq{5ub)8?{=X{fc+y+hbBLVSN9T^dzko#wUvR8NObv1=yKdr?tJbD``(Pv zGX!FI=0fv~P0S5`AGlF9n|d>`?RX5uH!Q=vOC63OHF;N~mD-O{OXQLHM154$vO=WX8WiMN~GXFc97V9t%K1)h65VEVaW!v0a-+OY#1 z-tKHWc=uB43(=2DdB@YqQDoE1HOxhM%sVl3a@5huyRGoxkWcxY1f6{L$<8|2>Bp*n z2l|AUp}SZQsTu{%I}hZcpRs9d=1u7r>52EScitaEEdu>!SZB1(XT7532mbBYd^_o% zcN^b;MsL~i=C_%j-of763*BP)(=FJ0t?YMsM_Z-ZE1$>u_}d+RxgCCa7P~ANI5dwv zFw2QUmCa&qhOS}y^XBaq>^;#qpSh}W9luwAPwr{Hx=#2s`S?MhZ(Ydz0L+>n{Eq!+ z=7{?>N6?{(y-OGkI(f!!4%q@MeiP637K5|(frxq{h(ZuHb$Bg2Fx1AqR{!ru!H{(hQ> zKkl{T@4?f@-xT2wyDci)!t#e&2E^O*ei8IU6de&8z(2mtAJ_40Dm>R4p6h3G-;U3B z*JfG#dG>I6X6JlL^m1QGYznlzf$>hpjxhBjUk<(cC1PPyuoH#?dnW$c1-4m>naDG; zD`fX8ulzm6E1yo^HCE%>U7lN#RKU59ld(PUIRlwiJR6&DXc{oWAI_RfCw8q^ifX~A zL76&-_zL#q0`x2OU-t7H&^oBi@-2uUS516BCjJ>uThz8yU;{Qy^W`*8@#Ux%FM2uS zFzZUSw;QpumDFL6C#C|wHeQY$i*GAkT#Rh3Mz$_?c|f%QS-@QYZ<}%Vr>(14cS4)W zWk@GXWt`Vg=hVd63@Kb2%wAgOg?h$d=NU%nJD0wR z%bR{ZU#q{{Z?4;~O~2aHz-jY>&|KOqq)pl>DXc2nQ`k^-#FuHk_=$cBtmDwU$A13G zp6cT9#C~+X*evRTqxjcabM5qbPUwan`i!LZ-2Pv>X2u~rO2mTI0@8tV9?d^-Ks zAlmc$DrlrNYVu{)fX^AUo5vikT&uCW&6c$*6oVCi0(>UuS>t8=VlmS@WGsSgYkrE#?{=S= z!XE8WP|v}8MxUSu635}SeINaI_q&I0 z>F7HW@caz4@o=lFr=REHLx0@w!cMr&%||y8WAS6d=3y_XJ|Op8>IK26AHMH|-wG#I z(aBbF@7cC%PkGP4>w@b#7p}btFnRE#oA~4T?eIwNM$n5}$z3L)e^=h-TR{AWI(E)a zpIba(%iTUxYpmFJ4f*h)tb1u<-!tb-Tlb@pnp>Sb$qpH*M$eQp|qz8)dH zt2&w_;se(1yoRyHkB(e?sxkv!!8V#29pS`e@ZqO&Z(x}3hIV7nGr#>S92owZ_5;C=LUtnPCB+OnGz-Qkh7-}aGgm1*N8vhTj*a7O@fC+ zkBoZd?}d*7_nh4Is*A!l_0c|hdwtf=&M~<*;s?94Slc1KaPnd99tUoGS96y5y?tGp zdp`o+C&Ckb;6<$&(4KgrQDUFVKBLe5&_we^@F+U$xX~qM&xsmz$xz0rc~s+(ePbVj z28>Jm@(%BLa_-q=jp#6~JJPv>lCe7zXz$r)(tW~hGk7lRh93@Zr(R$2!0W^lXSd6s z^gwW|vrghR^rcx}gP&H!xT&k$pgc{{Z`cP{>o;aV+ncQT)vWyuqeCJypw-p*FU)n* zPO{#Q@3!4}tozwdG>ft3qj$76BsZV&(ryy%FJ=uEJwCV?nnd9P`6Ju3Zse9h?fDbt;oa?VDYuP$#%9(9ihtpS z2E{tfUYAMrn`c?eu?aSY8B&B~vu6iBD`l#l}z|v*50%lzrX%N{J?hoKzmuN|6Zdt;acnM;w(vP)NjIH<-2Tc?Hi&t zwRgyW(Pjm>Cs`ESNgmo zzey&qQ)cvy_M`m~dzAU=|N1lXL+E1ZGmTq&GwZ$Z4URq&T#MgCi<8Dm%styJq$4NgT1?VaxJ z1*3gojwm;aoe@FrM$xzQfxUfQBV#c07`8#mVed%{-~`tjT&)I<%IY zHn#WSTH?)>tn>K*aFXf1gd435>_2kiao>9JGikq7SW z=&`8UySe|Twx@B*w_VfUmm_{&3+zuVWZd^5tKdzwd>WhT9QQHTw_pQQFs28DM|9C^ zNvwrI7hS(KIdsEEdtcpwF1pPJPXk*y?HS#}z8YF{{exn77#Wv7h|;$9(~!LH)gBP4 zdFS)rqE~#D^a^t^@sUKmQchmz6nmb8wLb@KpRK(w+|Nf?pH)DGc|@6g`R{C zQEVMQ^ND83V)2l9OewdsoY zIOn9M*MFlw`)r|?q=Tfh7GQ_0Nw(v6$fhCYt9bG+!Bqn~bRE|S`n#6vgI3BTEeG#= zs|CIE)Z96&w@qD=;q$d-62DnQKWT11SJJ*{XZoQ19Q?{gcy~P4?(682T5a0!>geU; zc|JfkZ7DXR_C6-Il?j}uxXSjDjWtMX&{OUBFS*z5zmO|P%>SyMrI@{llHuLPe%82^ z?2}xM&0Tz{U9ysODL#jK-UO%A1)x|d>IsgJ~u8DFeg;#97evStOoEk+iLb-mWu z3>%ZsJNIW&H_5+h)e038!2W9*V8zSOXNCNK=Y!*wW$u$mA!+7>ye#N=Qo!0wf$C-W7bk#m-qUl#eQ?9|Qu`ThTGpLiW6{~XT{68Idb}N@V z2YB*acnqD3FH4+L)_2&(&X_W3Z*ams0v}RiV0;$(GCs@Imh`-d@67CBExrfG)d_G+ z0*(YeS(?YV{U*L?bmk}E@bL0F_Q@>vZpJP%b;zpS4N8U3w=P3W8CKL3}!cY%(ox)=Y?Wb!0n zR8Uk<0^to76(poqnGhcGu;>LZwpa-S1jLqFyoD4s!306^mWdoGVlhD8<~^f z+KS^+=G6wjjLr1n@Jp2LspmZt$G-)BTpZK31sHSu zk(ah=jRW00lhEmMx06x7^5DI97TFdJmeoA89jmU%KYDn(ZU~Fh( zmGlFxXcwwVIe8E8rv3-3g>`kcZ$YHdSb z=Gw+(^f{G2)jAKJDL;9IdjWSbwO!|A)=J$pPZ8(6?$(}^u}tm8S!(7qmiuC4%_3vy z5ANqPS99GWYnW2-mXA+vqm6iNF7tY3uBUSmBX zKBmcYir=`Op?RSr&tFDOD_}UUgulUcI(op7V(5YTRCfk?fb*Cy4s$RD;wt&zE}{q0 zYagDgJz0n^EqWjTA4kdj!u&kZ2Q98~#cPwuGbrbr8*3oZ0ZrJ~1LT#nUQs&W@TJ&G zt8M8qtX;fHUppD=v9;jian8TjYrzAr={S-5i*|sM+2BNS84moOT4>iElN;6{>-lpBC*Qar&{#Knw_enkt;}(ASxn(@lxRzbM4v#+=wl% zhR0>?T8aJiLNB4Af+6cYXzEB|y+`f+^lpxI>7P7!B)v&{+rfI@zIz9_?-^ly&l%{xH`i4&xW862 zbhx8t_^70svBbs1MoZf=?ugmO+z-A9y$?g1gXGV=sRbWB9Fc{c(EYF6Gcc67t>t%- zg~tw(w_PZ;A4>Dgam#y0cyFM*hfK+PkFX~y?}cM*S>YJlVXZz**~D@q#<-3#S^KIt z`jQxl^i@r~VH@&V?b6rBdy`Gun(j_VreuAb0o`t+pGs&T=go;VpOAZUy!ZfXSwm9W zOP3Y+K=jff)}Hh5?;loY65`)2)0aIP8QVN#joAntC4Rk;G4fkb{!T(JWpB~J-r}E5 zvtA{BXXQ=Lw{6o>b*54Zo(9_G|8T7T13v%9%>|@nYs9$ zlU*EL)|D7*KUrgDvi9>a&gZ+bhrqZLKeG0d{m6+a=rca6q5EF=KwM^bt#^Iaw;`+KSWP$CSS~~hkl%Kb?K_@ zo`jCxiNBD}XCFseD|486^nojDgeOEd;+u^+;G-5B{2(^tzzJPH9`#W_=JZuHM^8Xk zSr;>}HG|Oy>FQqc=hncV&E!%x!mnQfTf>J^J_h;=xd^7j)XiVXI%C1Ji~*Y_I2+WH zTH@%Jel6I3>W4Q_KfHkY;YHLBFQR^U5%t3hsUIHh)2JgphkcYG_(S~|vx_?6^O&3L z7fk&J_qvRQ50{~T41f8iOUQlDYY-@ZxoHME41d`(mi29K(dERhmx(Wtj7&X5JXvZN z;hQs;Cf@1CJMz2u$l@6SP z;X-7=$^*hf5(o1D-%J0d4c#?Ids4%14!y*AKZjPK>T*A?1)F)lg?YB{ya^h3hBKyX z7#sML{WU**1gcWpmvZkHy8Ah+p1mL1Sz%khAJozwONSTJ;l-Loi%#y8*e$$R4_}F_ zWWbA(3-MuQaN`m9E*bulzYp_W*8{{%;2WJEyR^atGlvz=Us2*nUozKNPbGu@`Inr0 zzGb2IV)FrNDsyk$e)MNxA@izo6w7*H??G4a)i>D_I;?O}r@1b5`*NJ_uXFEB;YC`m z;+ci67fTBDdw)xedw=C`8S8!Ey2QBmw}g9tw-7TCeKj9kt%ert3GNo3S8{>=0<9KM zV?sXXtGj>oJzR3(if_W_l=z_J`(>+Kqpt2>^}u%Vo8-Efdh}t5)vbgE&?TKdZzjLV ztP_)+hP?26KDwcQCiaT>h1`oL`s?`J+yhji$v$SC#PWVBeV4HQg%|p!t;F-953!?Gs*;Kp{x(bbXp z*yG667!JPHrjTFD=T3`kzK(1j zG-NXy{ms6Il1;G_DMlZnQ~dOCIdwKt=%Xi}X6*6gc0GRa4)AD<@(%Jxo9eavBi>()Z#}WAcKvqzZ+`}Pq=4P7^>X|Bz$-b-i}qXK zx2@=fD)fTStGUsW6`~tPtxz#|RexQ@DLM0(oT7~`dNRLwJnhO`WdE6bJ$ztU`{XmO z0_+n0^X?+ux2}^r1sQ9dd8gnqc)pP_ioV{~VCZY}+!oJ;uYSpMlQ(ye?>H~Q6NksG zh{Ky^AK1LG-()^B`dw@3*G<0(@tWBOwpYfs+|oxWeLNf)+lg_zGwk`*(ub^N?zZ%? zls?8q`Zzpp%mrucLt;6%Tl%<*KKdE^SW14sk=K47;3FU3alECETj;|RSyP=DcQ|T| z9g$ySE_!?~_HH(3o4eWe;~O==kK?hc<$TJSFbDgGqSp%Q9eHvNu#B4S!Z!;B6E`1@ zpVA+Daw}_tnM!XSUVz@*j{o!;`4c(l$<@>JvuS11$i*IPLl|zcTulSnb0uWV9WAewX@hrkwos z;;6NJ-MEucYxy?xGi9)WK9sLw@+fd67Q1houFKxTuTjd1|0#Wz#ySRl))!j#MWs-Jh2h)X4~_6gK=&U-MtDrOI$A*Ju3c*_$m_9QSmwCT6`_> zHJkF-W5iAr@SEh(G>P9T{=odNHhM>E&$aMd2LAof+wzNtF>X0G(nKC)@+I_@POb@O zatC~eJl)s@to`nYEph1*Q zP9E4k4d_Gk{a};V4BD8u5j+WAhk)1PygT`SwBn4u+@~^t`&6!^#sIY?e?@%uS6t}!Q?Dny z7fv4+OWP&1ole^sv}KMNbJN|&=|ja+foaY~!DOr+&ZLj33n`k&JWM!EMbM}Kc$3hn z0YK3 z7VHA7aT(m>m%{oCKW&;rkENmeW**-lIaHG%t*zJO z48#cy*q$<{``es@Q*yAL`NZ@ueIFnB&6zXET+0&jcCD-gWN8KsYG z@Yb~*yvba{aBlb}xSL+{$@E*nod)h4HP=o{s`&`~T@C(ZJsFGlF8B(=E<#W9Z~i6t z0GGo{$V1=51KghO)Ntbr{K@}nfmM@;8x_#uFbT8s!f0~H8b}X=6;N}_2^=e zbEQWta%ZPUE=T@Mogi_SHhM(qEobc<<2}bMtXXfW#JDLSwCGyr#D7ChCL0suwFBL5&2a9MuF~c)O_8(euER;;GG1n@z-Rp zIGMG{D)9kWlWea~SG8{?zI=f9cB%8baz;M6_7-AEuM(q}K-_n`b`5*{DXrq0h+b!X zPp+NVPH0nrf5H7Bjgwr#Uc|cn>#^O)>w#_fHo$3NKc3j%?iOrsLnW~a;7IQF_kbmH z1m6|12LCR&>WdHWz^~bYoigj9HKFUJ&xeTv%H8#1+XdDr#y|`#IFbGbNe+RfuW!@W zTxenRRnd3bjViA?USECHeDoSUSIPG>cxEL3fRm5@|2i}kzrX{}dCK@)q6Pbr!#xhX zc$V^vF?yB}pJ3dP7_;0dqTdDm%7^CJO2a3^R)%BD1^P2jj`rL@@HGhhNX<3$N}Z|a z_mdYOG1Jr+^nH`G7x>QHQ!};Q1M$1|<97|hw;ar~P)P=MsSE!yYeGdD{=h@Kjv)hq z8Q7s&TCiod-u6}69RQx>7##%W9P%xb(5C_PuI%MY%}H6KV1o{e4O;yKI<+KhgTiCm z%-CZ0bOP|Op7*KK%EbEtXd7^*9opYDZQcG84<6aysJ)%)NEIDgCo$rM=uokj*q*~; zd(=7Jf+Y9vIFlu`F#8FtWZXHEYmN<@QqwweYR!Rct>*Q?j+#T)B-J#drw*VqrPiP1 zwjA1{WLojqyIHL9x|gUp&#RK_^FZ&g-&QJlZ{~dp22BLF11}+l0ABZ3PRx6R^J0T4 zU3qIR)^hJ=Z)_mn2a*#dw6Nk+crg*2UK0nWQw*F|K^IvEDZAbdPPyCmP!TvS0;fgb z)X6>7f>X)sE(fRHD4aUw97&z*m(91})eBBV&e$KTlfAO}z103#-AeY#7Ume=*(+N- z%vCf0GOcD|wxee8;G~)oaGe7l$~X&Hbs4sZxJT+ixo0qW3G0WM^X8*-ey`^M%q2cj zREW-ok7Zr)?lGfAqlLEppqtRzf&AWSq1^&#_bSh1-LqBpJclQ>OFL;l&C>pMX>Y^` zpnb)-3zOSzzm;*ax2f_h=W*7tA9L%+-x7NmZRvMxY`-n-_Zwx;3_g=S`&#-ON}uzr z@r--h=Tm8ak){0=v>#{Wfhj((xS`#-s&h5Nu&G%O?h3tvEggoe;RERFVqHHKximMr z;_VRoBHe_p=nlCTK=PTIbG4usAFdF7729zabR~Y6Ys##5x#)bIzm$J50{=>4`xd&J zIS{X~-WK^W?LzYf#I7T{{IKj*I`DH%T@Y&*eogzeablV-_^;fEX(}C$OudzYPC%!7 z8H@NVF51if!%cje_V?9F-TwCHSsVLXX54=`2KyUxUy8E58?|6Exs%9YZX@@nn0B~7 zci9+rxMI62Bgzg}VjJh--;EnitU=DNAS(|ZrtSds*J5jG96?veSwPwM?q{?qfDWpr z2Jc(X_Z>9eHR}%4YwST-{VTI(J2pUaeH!UAAKd>)cmmzugg>fy!hDZ$sdH(>(LC%Y z^~WZbYr|DO;9MJDVJ>_ldP(9vtm&_pJ6qVN>2X1gzDKO?a3N;YADuUsn9;oP=o1%z z4-W5%S7`MaQRhx3D4)e*PmQtZWBFQmERshgev5nxugzueW;{Hcj13)s4RI;pP6XdF z=j8GD+4nhiA{~~PdjivF`G7HXrsr$!foC++Zc0_`(^T>jleCy zwnz?C!$<7tN}e}x&9(jCi4%+?-&^-F1~Z2w<$FoIL*}5L{lWKQo#ue2vy)xLE7|v) zmpLIQpG8>{;N?EG)8t)h^-GOtMN52{G-#>z#5du$WKt_>v+2tWc6+LLh7FLt^eA7Z zAyRK}-EqU03G+Z(+(+?<@DRG}*vk<+{Ast~#w0_RKP)k<%SPMlCj*9O1N;~>kN>4V z`+0m#p9kyZ_U7@en0bg^kl5^F)P$X|V*ZK%y5OjmqVB)_<LqJMSxm zx5sk7Og=F?MK|X1Tg&`@$NXNk+M4>5cIVef9U<9I_9G8f@R#IpNWMGkxw^jf7q}%p z{?RW&8~xz2`c}E`hrQ`6YOzdHv54T*+xff=oeJ&Npu5&)P_L*PI&GP*)8zh`M(SdC zh^w?Jzr71Fp3K@0H^F<7YXICbV3rGPd`ZzuzOa7P>8Ite)@PCs;@6&ITuV(Tg1U;3KZ`B) z^V47I_H&e-{WQ+gPtKM-z{q*}#)X<*6Lyl&T#tE*9o)hEUqxRI#MbPmMrt29V_;)z zBu=~v`3x*{1?$RP3_iMJfmOF>=<2IloVbUjyY!NQm0-I zAc~g8Gs)jv!@up}vuSN>t%|=C&fkQ$e9T$oc{OunU){{H#&&jbrx`icEzIpn<|ICt z$nh%XV#?%WqkKIM=h<6>|LdL(N^gx9f6H42muP-yuaT4RyZs%WneP&;z;?Ms< zpK|W9X&7`E!Mc}uW=K33IB{dn)cDv~OU#*fwIRg7cz5VU*lPZbzXV%aJT44;E4b9O0|w7`Mcl&ea+~40A97~B$LN-!ASwX z@vi7gCvZPU8R>_NB#qE^PeA_f#&_E9%5)zS@ z$>eD|dqlrK*?gZ|8BY=YQ@7qp-Jl%a?@v2t@96h$59 z@3ryt1uZ5)uO{Za2wJ4?&2(pQuCp0h40TX@Q}mVcUyRy1Wv1S=#E6Y&`=A}UtrdM; zmvH`5?(UO)PS$()zYglxdYDH6aGYJJC8Fk^<(?+EnkD$;%s6LXWUr|obddGyeCD~D z_aA{4vX8HHHGPP_lib-`o(^sF&T<6{leOZ$=zxpxt@`1^H_jo?dn~>keN-_&i*Hxa z_vS3++x_qS(JIN)=?TL!vE@-c6|MB9B2_?(j8Rh^UVN^bpk!DlbQCvy}%;6T20ze&fb z=m^gY#v;GtKT#NTV`Bk&VurZ}jE{Re8vWSO?Xf?QtMmtQm6k64-pTJdE?5_Gq^ms9 z$Jx`#B3DW7wpQm%(A%w5yzIGCxy?aBd?76Ao}7sW1EVezL&Y&i$Au6 zXR=rF5_7$m+&lSPLhhY@w^L@V=o}yR+MkQ=fsdM{KiT7CzfI$OhnI5#erQMzbZ{$j zXzJQ(auXJi+oa;0@PNu)2iL;?Cf^=N(fPIzm|A^Tm2WM4EBV%A7?TfNdcmoQ@9p5z z13s1C3{J&wmVFzm-#l0B%fW%}cdo*3#)dUc(&UcY6A~}q(gnW@yJ7gvZO-;y&v?Zi zm^w&g9l!Z-IL2wnNxdm2>TW9G<+R#VWMnV+w4-B`O)>g79w!&tg)buUWt9sJZSRMc zKIkYo7ThWPqw^<0x4-ep+~Dgg+;4ELO{>=Bu_f@!68L5K;w2}CIWpI6CbzB7=(8*Q z;uC&>NA5EE4flx-)97y@{r(I4F~FQ7x-yeUmjoN9NhZ5y|;(U3AaMphhV@cfJVm;?{ijCnujI(IPc zos4@s<8HhEQlj=>jxnEO)4}yJM{zT{panbQ$1V%3t=N=m^vYgjr4pRUxtb)_F8%`X zh14EQo$SH%e;it%GmLXJ$`2IXbdxa;>wdu&=t`W3eE~;!U%;WMeF5d8cxfkgsZ`}3 zv(L{yOCjgXRK85sK{Hj zFxo%t;&SC3XU|6V&8Mam}SS5!W~%Q-ArcMeB}r0**%edo~k!$xdrtL%fkFsR+W zqiAL5YrAK%̋WB}(?V-CQn8|RDKZ)r1!aq#XOt(!V4^5f!pC*LAR<^(aOkmJJe zS&_dprwcCAc4wB7v)xnM{bo1r{h+S&o?Em}#-ZQT+_T09_c*$?F78QeZ8<*CO7d0E zaY^O)YHL}0AuEINzXl>N^8P)UtyKggVH{4<9KlrQ>d1}P& z{DA!X1H_y)QZ>3RK+d34JL*5(qNT30 zo?jWDr5utq(Y{6KhLxOePG(>Jeaaqu0pq2bye^&bk+2>x~eAp3c>+^si>!G6Wl z&*i}G3E#ZmO1*vZCwzP_M~|$=7L~8iaW3&V*+UXP%gm>=lQZ$P2fh&PYfHXY6Zce< zgPTHip86^`emrZJYuy`GMRGx!G6lz1)n4_KuK$`ci4Vm13!$QGT8B_`<6l|5!@fe( zFvdSf!!|j{3*b(^UdonNY%fhG7ZBf|JN@=V*0RvW{3iW=S)jSI;bHu(-JU_>pK3wD zRUx>lgrBy6Bf*agIm|{b^3ipAEI_XXD1MEM*IEls#{My9#Kq4M7@jOw(AyoG-_sR* z8(l8?TgE^6Vj2Hs;qe#n+{?Ho@NPbBOC%SC{Tt!e6y$9|S7P(<{yt((La!#~DrW!| zkn1V$%6_HtE%BR=q>&p!{;Y%XGUgL~v#(J5iQv0VYVL{8oNvrk*78-%^={@WJS#DR z0nE!qTebcJFWz3RU<&)+3#exzFjJ8cfy>@O>l@ex6Yg`s-HYB9xLd@pFHDLHSNL%Y zaBnu?$~sl@00iz}_*3d#3aq<=+DvpZCUp; zG45t~!o&Mgo3D|6#}^;(4cD3s3GN3WZW@#OGeOgu6xoq?#p!TljG3H z9e}&8Rx^x#g@=>zBJv}}yN1q2=_C3$Id^8>|opEqk zpY8Uo2bUWxxGd6f`RYF{vfem8Z{J1Q=q<*1+fYa=_OpND4^bBc-8sR~ovXh>hsM#J z9&&a}8=k1FZ#Q(ORo1&$?DPN7XQHw$`nWUtb5tChwuD^8Eiw92*^^V?#)6XxpU20^ z)8_#vYyS^$@-cVr*x8hpQ*CjQQaB-RJG`TC^MzJCLHOq57`{0kA4fCK0!LyWOq*-v zo3Z}~II50^qc7UxNa5zQ7`|zWkCUwPfRiuDD>&b>G&dejK5L7Ul*)+_zWErw!JqUW zjp3X3;^SyP^Euc0=C}U`IJztzjy`USBgHqZF?{of_&8a0ZgE0Px{v$bXq?oIi}z)@%Lvn~#Pe9bO9AHSkpyWH(tgWX?evHLl?-GA+Ki{0;<9~@xda^Id+ z@A$nsubFt<{$hMQ2FNk8$X|!>XtnpoT1nyYZTht6sCId+GkDC4hevN5Ji0D&`%1uL znFWvDOy%4Ez=}uLxV!;!=b*V(I;s_K*Wk-NI$x%keEIA6_YdCVPMq z7d88TR^}qHLWvU^>)&ef>L;<@UASW43S!AUWZ&H<@ze%Q)iDbF1)C-7;pYwABem~{ zFYNneG>xa|{;ZPb;qG_rPzsw+kPCAT$j?(?{0x37+;#Jc5{!m8RRso9c)j z^sIxXdC1H$w~QKSbQnsuLfl|3r17&{TA5GyS#Dmz*^|%6Sl>X|tlK z+xr+NQ9<-tbng*chF=&e1ohWzR z==LpzrpqieEz)T^*g{j8yUbf)rIsg7%R`LMRz8W{ihK$!mhev@8``8e@t}OUCUx5|5T6-M)p;a*2hOIXW$$ z`_v+z%vqh$0aog^#A$g)99lLzAJ`S9T zrMHXnU&dK!$(*Ghft9*Daa#Jz?B#MHYq>BjmqE*=&@yD`*h5f>@o71= zqqJPcTteuYrHRp!YoE_MEuR8c=9o-c3LT?#EaUbaibu;U-M(4SlDpPTU&>ph)AFrP zB68`W51F^XO07?vmOqF?%QaS7E`g4xpyxt^mellnL}9oAlN=y2Xc?+!6Hxs92 z=;1hWX|aTShn71ma_OJV9^Wiz`I$k>J@~OAmnAW@ygoiH zKMkF<%H^49Ig7b`#++tH)6&LXDq8-T{w;EOUTB#wv^2&PmzIp%x5t*2?2YESePf|z zfrXad3}r8WXr(26$h-v>c@DetS^KM4-AnkKg!L{e{q5|FxOquZ)fG|Z{*0BV~a}*#_xNnEiFPf zIAmP`J+5~9MmU20T-Iqk^J1v2u>@~XhWkPbT~b;svBBqw4I%^nrO~m$0llNw=f+w~ z;pbWl{>ekBGxJd9N}fOa|5D#P-u}N$OiT6IJ3*Yw+PNcf7*l_hgJ-+F8sGD6<1h-R z2ZN9v(t&h$gCpMLbQTJDA$0B;ooE7@4f$u=os>OD$IPI zF#QfdzeD^peCOs3jC(5co1~w+@X)99PwdBi>@7X_LgoC1*UtLA%_i;#n7Fs}qpzJK z>t-Ky@B1$vFeY#W`G1c!^IPQZ6s5aMO0=3&RmP`R|AaooScbTLSg6}>TI4pWAm)q z=GhjTXTPv{w$bL<6E@HO)8^S)n`dP<&mOXQR&4X^ew$}YZJym@^K605vqGC^vu&RJ zyUnxPcqVpFl$&bbrGPxgZ=KhFfd_^1M_3H`407x}_{|NhlR^*;CZ z`bEdfd(+MLjNG(>BzI5!J!+oa=$7*r>OE{jRDa|(cH_N9Xw*o4Y-p^CO%iXu-t8O7 zyEkY-A2gFXXFllX%Qf2@aLOF+uRi)pKX2#M&rihaq@VmS!tcWC4ujVxS#lW7eAyam zR{Sd~azA-~8hY5sjoohM#`3A|Er)JqKCR@C#^lH`8Ti)%*W}+418zP!vVGG!mLogfSeJbhzO~DdHS=R7SJuptU5QOmIkJo;CPy}0 z*QOEPlJf(1FgE!vGQJ9)mhqLs)4C1Q^Q2?x-ZO^oWk1$+oZR&hkKUr!wu};(_!H1s zbQN`{!}P8iBQzjaH4eRh4BpK+iLLIk;N8rL?t2#W&Nj|4e&0gxRg#NsZ#VX$-$3pY!!RT6ebNg^~8FDRkGGza%FFxg1$qnZIk^?K|u4u;Ai9Tq?o~6Mf zX`E9`gGVO8BTd-3H29(!{T^b!X(;x@q))nqFUYg2u*6&rt6GfEhy6;)53iVyJ=q3* zHh}}-_XWtK>-71#K{AFI$ zY%U(kUR8R;rd62tfz4t}L%@gR=??-Q8OZ<77(h~kM$V|e1>|I~HB0~Xz{@t>&= zj>YKz2|2F30o}CR37P1H-tFjxYV;kg4LWmH!i0N| zh4zW+g|B|1^Ia$HfTHu84<@b`g6N?{^}-(+TLN|%g9FAH47DMH}=%qwuwvug`rL@9>_L zU&RMk{x|1Q8-XKdx2^oRSe-wL_#>jPWIU@+a9-+jSMX9<*E61r;Oe3=NQwk7?X^5b)n{#@%nm7jY;j3XUk@ZMH^Vw0H|d>_n*+dGIn$~5(+7XHxPnpxSnh|ZzRdL^cLu0?wsyj&HNvMQ$(*6% zJQIFCwKyuKxrjeVTjl>Vwnn}SUfTHo1!4a` zO7HiK;oq`#x{kH-Zw+xs>wQVWyJG7nX2G9~H4VOOuK$|58J#J7EIL$dL0ErQY3}bF zH*~r&4pXNuG~oH+Gt<7*#PL5RmNF5za)vP4_av_f-*d;Ad{5*gp6|IBUNC)+M19ZG z)cv>hJqJNs{LTQhO2qe^V&TI?eb2up0J~k^Q+Q0(qKWoBcdRsGDvZ4uUW)QPnOA$h z=L?LfqrT_zA4Ka;M|8qb)7jZM^#?a$ac>c@KTi{_{42FlgD&>eOPiMuCnlVKg(HS*2;Bioz8eX4 zw9$9qECK&b!F|9&f7Ty|E%CfIa83U$%Yd7(zKizX6rKN8oVfp%iBD$dznQp@8VBOD ziT~E(hHiX6gZ~yq`yJnp)_2_v+9#y%N-we1cb_)f+t1USCl$U*$bSC59s5}y*M2Ix zrk;tP&=;}%^mA}0cx%H?iP=vDx8}M1|p+kkAW(q&?-Ikxq!s~%3e(D~>PbCJ;6XK`(t1SFv#7?lEcg5kS zMRE8kZv43pTY+ubP9E<$jz25J=S}KMU`f@#p6mQvz#%D4zQvwMWJNn0~s4bEKkY;>Dli`)`9S z{1@@xUJ2Oivx5elSbdfeFP>OQyk;AB%A4^TBc7O!X-&@`<`$no_U>e_IuegD=P{mzJ4AB$fpe&7CH+Ts1&hq2|tN1l>8MIG6T zWeuVDaK}F)GGJ@VEjrM&Zty`G=}dU<}7jN z^zTi!{C8Iz{##_giPgVx`R{&%|5luH{;QL9?IL));8?7z=m{~a{oB+P%)4F0?I{P5qw&pOV3vlHNfGsK`3|J4B7w7=-T zg!u0p3Bc|Q|D~P<|Mlzy|NZgaX#N{_2L6lPzg78A-0>1OABtKOapJ`a_fsr%$LBh1 zi5FK`^BzR+m_FBE*`Gg)^{#?D+(L8KyLFcMNgKGP&$Y*Z+ex2G>6ah>N8&!$OW3nG zK9|CUik}$k-JeO$!kK-pDBUwFhV~7dOBLN?UGLh``*b_;4etGEkKTXyG$QAT(tGYX zpm!Z`O&L#=-Ua7?-WPX(-eWsJ@7|rD_g|>{X&#vfg>P$)2C&ytVWl z%TJSS@*l)ThK_1^;?oW^&Ld4Ua{Q(4x17;AnymfS#XTi{;!(s8l`YzMqT{-0XdGUW z{SfOpq%`(o~ykY$3I@6EsX26uYab=%XYMo1sb%818 zg2U&KfT!0ycc^+ys(!gTKj_-k-#E|l=O3_Mrk&*bNIsW}?H+aI`4$Zv<1lix1-2P8 z_&a$6gV2fQ9&Z|Zyeco&mCX57jocy2UVo0meSt|&W6vhrID^Ri)EUI;#nC$HVuQ|W zA?sr&9nrkXfAxmR->3US`ILDjlTG0ySoHStt_M1wJQ`jCF)9Iy4coclcj zPObj`LSuff8}s|5AzkI~Jo;PiS8wPNRiD3!T0`8g$31qXd8Rz~=UyH!^`t!1m6G4T zVECXcYDjY&w_I|j~VyQO8aiK_mvRSTEm`Ey7qR!nQ~}9 z{p{CDp4d-6MLhG;j~}=J)lbQ;WX>Uug`ThP;BM+MM~OAl|4md}3jw~V7y`=r(whn&xj9ZPko_I9%~^-wcoY1URg(JW)(nT#dC zSjbmmPAhlKXDo6@P73$y2whB_@+7p7J1d>o4Y@Bu?!mN<%bBd@CN+}ZgsqcY)kd$= z?Zg*zaaJhQ(-Az?pr2ir=PLKAD&2hwU0l!Io_=V!=X6Sqz>_=79^ovKoR3-t%z?mU zFRnuN;;IL0!99h^d4gj(U#|0Gm+*OW>)%_3Gbd^rkoWX<+KRy|0=@Es`?)VCM7;<3 zF5{FMeL_3wOKKKKzs)`KgDrjXB}eLQsZlWGtC9^v93@Z4eW;wzLw?SkCeZHDpOG&H?e;^v0I>Fd!*%57Y3F^bfFm^vsu{n~P`+=yhBFky zk#+21CG(YhsBvG0W(r+(eb~=^>3!Hk@U!(AogRXpYVafX4~UKzJPm{%mEcDDd==S? z#m&}AhoXy)d)Ka~!8m@#Om4}QHDTK1q`9BuoowoaH1SRo`r{P!O;jxc zYSySv>JZ7NHSeym!QpPC#?3{4<}S`PNiW*d$+gQ$r`WN2Td{|u851>9LZ30N&{Z|^ z%qnNnM#)(-a`tgsIeSD))#a=eIr|vAFh8M{$eGY8W1UW`#g9eg%nQ72*Kuzmb65OX zmgb%T4z%mE(e>n-Yuw-21P(mQ9PY*V+D^@tS9Oh+8~fYnS=c1#;N%Wcsadl5aqYhJ z9@@;!!}-kQvxj!y&1u~8oT81Id!;rrG~1CEdQ0$sMUC*J?+f+hEcfAoVtX6NUXBy0D1xv$Cx50<%-@-nznH}tH~<zR27ioay({$GrL^ofp-3T)Vz(l(X2h}}WV{W@ z?hoiA`%2($MxV{0-jTptD0Ju+c`wK5euwu8djo3;XCiy+uzr&SteZ6@(+V$zDR~XF z-+ZU#Sz~hEEAkBb_hn6>-ph98`JwL~&TNRTmiil_cZAlNyc>Yd&Ct1pyTnBI)KDMg zb@Waxe;lML8L#lf{xG3!b z&W>`QTdwxp7U}!Hm@~KzO&8pCG2SPhSCr%~<9&_y{WGW$(_P2iJ@AF%gT@pKAC&A` zA@t~O@IgJbN-oy<;JX6jpmyJ2q3dqOEp+AYRs0>AEo02Ij4{c5x6q4w2K;lt!9_aE zt|rX-R0~YkuGuo53$@Wgn;8PI-2BdFEoDg|xRQ7N#C=_A?4`+3 z{Y<8x`U|wtB75?CAv#`o_$K=FKyOd37Sw&qY-BL0_Pa~b5xL-*cgr+Kuq@f3Y^_Ph z7s(|M-M*T*xSYc+J&hitpHlwbqd8l}W(n_zjhA}JCf_t`PF1gE5%pRQ;xBSO=#U@2 zX@+lF;F~7Klu8}%vGmceuI#C!n%3G#U!IZjTWamI*c8u5^sh0tIlL!wQ~(U_^$FLb z3Cn#!y1NT=C)Y^X@Ne>-)Pa%OWesnKc31LutbYIh|BkkQQ&Z60S^2?<{|xO8eU?-l zY7m~jtVW*M`X|)HZRy^Qe?m>57HR?=gI;3u7Gg`p9%Vrbv3--_Z}C%Vut8qK$MgbM zY)>QfR5rqy>VER5ZX-mu7Nlsofdi~dxF>n*Va~3uTC5pXDGmLf&NlpN%VSrvH{ht|0f{z1^ip#=0TLRvu|>0Oq10 zD<)-7=cQYuue<*l8g+tv2dOtQF-7m|*EW7nTV8YR!f&triR9`i`O#{h;QN+xEjS3c z;8^po-t@v#K=kCF=~xIc#O+PW4z zmK%61#lAxe@&C4-3e5~qmrvnS-2>O%{Sy2{QSNJRim(;9#(*_qJ zmshCgIu2%9aB!R8U^i`_brlO96b{hY`@sRWzBs^l;SXC}Oa~XU+QG#Pa4`#9U?1Y) zBJ3{%YwHW(g`CvA;uF{D{M8Sf2tM$M4@;f*+f01C5RH!|CO&AV@KL7)--v^Y_t52{ zuZ15jvf$zZzGp1TIVm=-ER*@K6M4L-CN#|PqLLBpppub-WcQ)JVQcWC{k*$}HKe(g z5Zx^PY2Yu|b1(l`JE}TbtR>PN4?UrDv#ckNqMN(T(N5NUl|JOPg*7w%`!>L{!}4Dgeo60vP`9DIVlIcrwS&|nOos<~i4ja0Js9$!98gsI)2-_HEhWkFjv#fUq zaYr`#ELN`5g2Wq z`9UA?hd{dCkHj$he;FF}GX3D+gmGglzv$ti!n4$M6P}$9K1H77`&{*1e}0$xF1~?Q zJDcxYR>bJGC1=oW3z6rg?a1>I_;o4ryfD5zZ+!rql2@j1dOv*9j&8fl#O(voxLsl3 zwwm@zmbn+VJ>BO0Dk9HUT5xa$-zyjOKPi5GnXLbJh@W~%O(San#TPj#?mzt%-!oMk z?Rf@Y4ZdjF;X)q!)UFgi4ZP3W=?c!@#@Ym(w4hRWB)K&S9o~dr*GRodxf`>I^I{Pl zb(MRtIj^13^OAfR#WVDyc%~A0qP@iBd{RgM2I5)xU-P%=dC%1sCa(jJ!SW9uQ1YbW zBf-$9YfStsntxL8bHGNP#J_wRyllBOh6h+TwQJu(@BsDNGZ8L z;elHX9+=h+4``RT|9Y>Xe`ZGG>I?S#m3^bV;(=-H^T0d{5BviBm^S|Be3x~pHv{_5 zVZFUfbovD~@~jMfA~g-=T@_P!mYBjz#1znLk=UK<2^CY|nfG~o@s}9Kvod$?2BE#s zy^G;rW#Ctp$(kEDjjW6HSc22-g2!adExwdzxht3ipTS#sMc8rSvG<{$e9j;53T|^! z+h>{1gMY@>$@976BRXRnZR!l%%y$`Q?_7G!ckjv2?i|L6zFgOo#-2vyP`8KiEqd*w z=+Ip0>t!2WiPfpGe+w^lQTWw$%XkZZ^VN4<|HyqBCf;wbJRf6ue!b=Sb(ZHN_}&LU zL(VQ(?;n;L3WqrtE3x;cOk^|*T6T8@7k(MJLmu1)9m(J|UGV0tm9xs4?yKQnxvRn_dtUtR zc?KC4d4^B?=*y6@Abo`0m_Fs= ztE0(QZTA`BW%4vY(Sw+w>W5HUMV^ z?-o!8xk_wGgb&2OQ1e*?oO|QI`Ixrr+rs&wbv_>hXFYH>GavEi$429=%7lZ>o{Rmr z!&_4;?bb!$D7=++ci(A+(+Zq*`O3EXHUVoYK9z+J9yQ^V0;fCBe zJtXU_s@=NYlix(=h%SDF53b%=Q_ zfiG4X~pRtaEedl>jOV7 zrT?sGnJxqld)4Gek8*EsKDp7mM0^YQ7T;3#pI3gg6J(j+205`d;XXrm5{qs!q&vS(reiOVTEuxR9^ue0hvx!gEq|W>KK8f$;d~W1( zE%+z}Hus7r5$E?5@a#!`+ra0~BCx^3^`4PBUH1x1z6;Ex3fgjyb`twTp0TvqM%(p# z?u=&@b2en||%87Z7%G5k+pamTt;MbP$*{H*o z9H1uV+|&n|>jnIt!0!$}X5|OPe}0lN$Zx{i^1I}7$U8aUUEb>tKgzoUx)VP_HV5`Z z2Hpq8FkrOsxsuNhwP2Eu?}PaMIiJha)ZSf*!v}JAa#p%d;Lai5^sVaXZ<5kuR&h)7KmBus&xh6?b_Y`~& zzL#8+&=TZhT7KRF_Llmvo(U1p-7fcha?gq%e%YSJ{x0&g`V~ImUGhZSdx-PBK^$E( z&JO=95Xt=`XXS)hzjd5Z7sgY7&Y8r&G;mxsE^jq^^=4lZuRSU?WG(%b)1R5wH_uYv z#jd|gpqI#-^tXkv1&|4qn?QelvGM3@snufY+>aJvFW^U|(-}`QG3W!_iQY6FTF+pu zJXK$t%AUS;{{H9U&;O2dJpb|MasI!Goj?2jvGY&4)xrKwxAIU=f9FdchQ!XFeU;exr_9XHTb-fn z2l8Ea`F|6Gi%s!y;XcQ>cRg}-F^Qkd(6b_>n$?$4ERbyhGs?U)O{84@mA1Dyoo*)9UZGvN1R8z{VWD=&Cj&u zAtggQV)$ll{P{aOIe(eEolW`Jcm8DPj@bF{XgmKDe9!H~bQK<|zyo)G`X@#6LrO_} zeC%QTbDgE{IFj)z&Vh}6<~-t~M+`oERq^DAyRbe+jy|)Hws@kA#sHL zcgN_T&9-w%nx8@i7O}uRcZ>ivI1b5%z`vmZX z1Ng!L$-%?l+%I|-U)b#8zrah@|*M>z#o1@?zp4BW^}joY2M3=f35ngr%#Lh zZ+GM7YqKXbo>>W_+y<5-|1_K4aR+SgJSsZ_V|3ax&!z-!&*Y&?!FG- zuA}n$=LIqHx*-nVX`JtQ3%nh%;H^l<+a)I6)>`!3daZc1$ZMZyy!ofZ$6M|>##_#L z#M|Sscw29ax4QEBue*H*z}q1U-g0!j9XMjht7}|D&#g1$wP!TmR=eWk?U*NVc};}h z{+Ip|mDP^&+fA`}TNe*++;>h)u(82{H*bp4bHyg!*jqtPb$?_HcyqvSSUwyvZ1 z^^K3Oq)y^19*_OK>HNv&tQdSrP2t#ewDQGPM$do1Q`BUpScK$;5eRBUQG3wa+SDWxxBoBTMd8Id#J1Blj3w}(xBXYiA zC}#_z?q8*T_2)&-sJP}B?$%Lp&5e2-PtLXl@OJ_qYk7XoF&`gEPFOCvjAM!6I>+bv zB^J37T}ZxEfEcl?o2QUZZ~9(yEjXKNsg)C^!BTjC8UG}fIF>OJpFTd)rQdr?f716B z>6`ek%9)LdOR79k<%b<8(sh)z4>LyT``DF79AZqf(`}wjv3WMZ=9$ap*$p<&M%z5Q*5=u8n`eV; zp7pbNcBRd;%WR%?=b7jyvli=S&Nj~!MEuCRE=Caqpm>ppy+!}#Z&qBAdn?r@*h!ILlrHe2X7fVb-zW<*;iVr`Vby>K!fWS=G^}(yME*mD}?4s6X$wqB~hI!}%#;s~WTE>1{ z)@aPV75g&Pa=)pJ!#r>KF5?&q9#VlJxrklTwKn&gX1yPhdrG;ptCw5yNd%VUu1Mc1 z*A#j6<|Mg4I{cLyM|GufP89nk?+6_OU5q;*FN>z*ab15ZJYEz{$2xOws69MxYX^_9 zw3FHpW?cxAcAp>ZIPKQ712d{F1hh*tXs6cHigx;%+N9k~EA0$=wO|ihSW`Q?62H4p z?L&s?wFuaQEcBAP5ZiCg&*MBOchAb*soHK2{mg`iQZB<*VaMc7r9DO+j(rw+THELH;{X}eCP=u&wc0&l`ABQ$o%94YBpZ@_A3% z3XgQ8=Cdg$L(jx3xovsnWop)oj*aCN+2@v87*?JUevx~BV|k|QnRvzm*W{Va2Hein z>o9rdd+6y>re1 zy+18XoZfvpK=0ElI!^DGIBRc}^FfKyoBUMeGuhFbSVVjDzOEDW-pAa|R=#_5g5Em} zxMw5ZKkNX#tFQ@)%6Dle=sh-u-p_D8NA$W?zT?*SqJD^7y~)S^U|-+sNu%%B`d+d2 zVgm5DENX|JhzYjGPrv+Y$Mwgx=YXF|f!k62(fu6olV?ui`Xj3Y{PZDp-aBe9p85A^ z{n68gpI+xKMC*NxLhtPoALp#FpS73x39Iqhl@G~hBWpJ0+u_ryeN&5HmvbgsvtEex zBmbLu+3Aj0pTNvdu=QtMsH)CrTj?d73D|f$T~~U2Qz(w#n70X z7Tu2ohL?3!134d7KhlKv5b(s0bl{uzWnCqHr1Aj^94gKcYvXT?b;OaO$(h>jh0pKR64=POB zW!dn`U<rz`EmKdGl75%;@c!fRpZ1>;pN?adg;=jbPFBP%=V+(c96VM0L z^|{K;`l3$me%o<-v_9x=(A;WYZ1usDI(+7)f>T_0;pRix*buHCnX#?Zb=#;pY%1uD0n>hXTU!vr-Ee#`nc$kLy zacC$wvc_UmjZR>iu^1Dk8T8_ffOmPG%F)&nDa+ouK&+1FlteJF1_52(1%c z-&bJ|64lS8jIE>kd29^bpW!U5$od)ev*c}cLO(zM=QGH*jeedIht}=+YbK310k2)T zw!kyx`fD~c?ts6heO&&CCzh+*cg3 z_t*LxaAN(n#qsW=<1AD8&H7!!e&RjV`1L!8-R{IM4G}BNz@;3rcYymC`mmXOmlmD{ z@Ml#0y`GkRdavgG#~bVi61Q7Ty;fDfxTj0i%A6MGZn_|6IE9bjOF(~hHa=aS(ksYLbHt&FXs`YS6&7FHT` zx9YFB>lJN;E&qQTUQ_EOr*13Fy~g=Q-?983xBt>>Ed59B;}{pG|7mghx2=c$4ZR-r z8qO@9u^zUoef#@w@R4ONCLw>zPmTJs<=;`A;NJrV+_T}|pK=y9QT}Z}Hz&%!m5i;U z{5$90qIJ34-ynL+S`Rxee`#O0<*(Y;?fG}N(RVEW+WK?D+wteFX`i27dB5ZQblW-L zr%K?qYwK5;+do%eKa^BQhKfFFu^1pjh zmnW$Ph0%45!!^5$$QxKfoH3?;c(|@H{rG_!Q2j*ehd(+~uOEJFMx?Isnd*lRfp%&f zrE%(q8)K=aR=92baK;i~EX`^x(e=YkK9RUZIkbqaA3k5xYbVLM>?CXLaHrG#Id;cM zy(FOz`+}<0`Bmmvjk?`ZqvC9FHe`y9vqvXea8?;zbG;#zx=GNgjPcsSjHwl+>by&> zs3%33UYeYG3cI5dCKH$9k?$8>^LA$7i{2IM(R8F#c7kH?rMtT$QS5b#s z>KKPYHTIV|hY{R15IVwFWmHb`c|z8qxLitLj2uVf0_` z$h(+#LQ@658Pqv6=3J2E&J|iq-Rn`t+3mGc8Ao3o$hgb7nsk}$7VL?CW zMGj*qlvShh5*df5!RhWV_{*lIU5P8WHJkDAZc#RM#UvLIxp__O0OOtO(t@V`lvvGE zNFO91083~&4>I+z2~)3Bl&R)&Te(&{Z)RWuJm5;tnetcuEDkJfwZZM zxq-ORCjOz{B6TrK^tzZl^KV3-kayVv4+`&8_cH3I<`}%wPu0OxYx>nsI*L`DZN@d1 zyI6$2n=LS4lQk6InSg9Pym?C2F%r$=`e@%(P2J%l{Rx5 zbqvcTjsV`axg5n(Q+2b<`%Ue>gTiOj*=`iv@VA3n@8(!mfs;Pul$dzv?3fnx(8u)p z<@%V0@EJX(q~ZaLN8ktyQ&u}V_X&oqzJ7@ThffnufVsbRtD`sxIP!hMh1%{{!!_M? zUS5q)SDK~kY{8Y_>2}^%vXpJs&eq@k5${$~J6p*Tdn4m7h{)2-wAsGbsKJdqMb+b` z-*=@RH=pwgjr$FhEKvhh;CD6Zar1lOOh=x{!z1Za>Uem_q131uqp=4ezQEc$kcW|s zmG{e5Xu+}vbh%hSUdndl;$3nJL@rj7JD_TEQx`<|yOh6eYjT@BZmG$g?nu>Za$ERY z>MB)Jm$FSwZuSwSF3umYo3amJj!kNE_vOwQnUBC<@J?v=jm$}Ct=54J_PYK5`@gqu zyBfgTxozU%<{vDTOU zpQ!qehV3_W{1Ww>jJb#hzqW>f|5nI>s+nj3_ zpT8*+naqOEyHl%$I+NJ=GuLx$F!HC>9{nZZJCr^r7Gk{vC>{E#^9$Hq8HZl0f~ei znm7+DG~Qqn-;5hC;GAM4Undgp{Uvj>i+AlbU|8b?pA_i)F=U+NaP$oGM3UARDUae~ z<&$S3wxmr?Jv?IF|C9YjYc944(+x~Bb}-IT%ZL2$xG?)0Fy}L0<=3fvWb*idY32#b zJ%n<;Icon;_&KuwS0TLm$-l#^vY3y(hfX33Nz^Eqv9;OdxQ3s}RzMHeCR{HF})|Ni3^ZU3bM$;#J<^f!( zS_R^3&tqM09q*sm_wOWq6wD#f^a-CWZx8104VdSGK9AbbhjXp*=<_tNZRz9c@Ajd8 z{OF%BeTp3JmtV8chkW+HrN-HG&b&_KjCFuBuU7gD>>r;#e>)N-KMB$2z69xGvwyDW zvlo~a`ALXAGuwgLQTgd>L!TDT&sxti2tS&B!1V@wB=1Vjp2+^W*s&1jWUY9fXme%+ zouTGj-U&SSY-i4A&{GQL!cO4%OZM{P$zDhCeBk!Bda8vpxAE}&7hsDVSn*s(;wRtw&=Gh>dXZ?64vS#k9{g^Yk;>)z) zquS*g+U`$qzDVS>+AiN5{<6ytHs$uG??m*YEf3|k?OtKB3RFp zm~dAB*VK`=d}M)Z@=>M%x0Cx_%D4D4IUb4bcctN1*zI?T?i6_{vYaRJ-3#5|XTi%V z^r+;TmYTVyW#K%dD88wh8cq8wgZ9?_T3dS4e_VQ_E83&?6`i2>ubA7}();w=9jEtt z1Mb<-`>qbqdmZ*DQF<@x1ic5v(0c>t$3@ny^p3k;DZScuz4C)s?bj<07=2sUE1{xm zTI2Q~wf`U4pZ|88{^!T(-*$f<-6VGju|HpT#{GHL5AE^qJ@ApaKcA?+`ifkFv*q7u zo#5Z22HdmZ-(Sy2TwjTQktqMZz}Pyfua-}V)>mH6%nJXS`}3bT9r}ID{bfDeBfK7N zTEZG{VR$c9t=q~S?4>Rv_OKND%kL6TYgDyq$GBhQEK=h_7te{=EQKGI5mR_TtLgKA zgMUf+z1ttlOB73pNF4OlFC-YgxzCjWD z243L3cB$-h88O4MB=_$acQ)%M?;KZg3HNdC7DLl6(Oz5kL}ZS!pQlAyE1+9y(Hs782bldIXTyKRsSNTJ#!xCi-L~-%Z~-HZJ@w=ZZylHDwB)LX+;U;9EB5 zw$8 z_uR`*`7C2vd93+V!>6L$eJs9V&I4~5n7{dU*44+~TyRSM=C$%&HU2tHCrh3KU~HY! z^f$Lbhw|Y$z-2)GW{3KH_**uA^OWs>d4IF+>u~qj*nGQX^F4n<3Ximp*Lk>;&ELEl ze(bO3Bgeq_f;W*(=#=*WlsIl23y($Z8X#xEz}74V6*XH`(X&-D=0XkMT0|#BXIh(R=Sv?+gyFZqn3G z_5<7B{OZ(UyYM~0_NR>JvwQo{TsGK#G7D_ea_|QL+Y2*cd$fPpek2QQ{WX)B^l@-% zUz$0@a~Wcll(P?9u^ylUN;y)()M^<_| zUEI^@#IKbDzf$*{26K09I#&DOCg&Rua3^&`CC~JJ8}GN%R}HyO>f>&s+of{4l6{a% zmFgqp_Q5lMAJT6&dihmF zs@TYf(35$zDbFSM2OUOj13cHAZ)q_n)_AhNCEu2_PZq|R^6f$o-hTPEF(W>9aDeh= zH{DbmzOuAfNdr{U6esP z&D1-KcCyKvo#^lH4mm$bKRO4elD4hjeg${=zV)-?M(2btKd;#Ae*Okew^oJ%)PF)} z(2(N4`c3#4N0-;x_@4(Bv|C$QR~D9yW!gcTZoAqiCcO~ zQL}Ob?F>$1BN#iW0Q!b6e8yXw?xfA;@7$TKo^o>Nd17|%IoZ@ynwSu=cKO5xSX*iE z@mBEhgmQIK()XE^tNY0($mUCs-17U+v&4qXm>)d}4d|{M&u5>Mb-(n+?BhRY%oD~e zWEXj{RZAxi-Q>yGVlTGJc$u}8(j)9SOh0Dq?6EuUQ}FEOWb6){_|?z7eXjTR+1O<% zU25L-DW0YMe}mdx;r7aY_J107>bpZOPU5$4yMq0n3?3VNs&~@x=@(vqZ}-z*R(~ow zO27WpVdx>yl5`mV`8-xv+VL-^jzfRj4-Un~X{?j_&kY_u%D69b7I^?W8vl7;Jt7Mp zP4UL{>u3Y5*R=RE2>cnS|9ld>tsj4sp(p$($@7?W|3DNUjcBVs|2ceew-<}qD!qK~ zb?LZN?(MzbpPF^Q@iF$h-`LLC_ItZmQ-5RnekbdL&{21rm|F_&YppK+U*wKt;~4|l zZ#(W~zWpxo)z)Vl*XqdIym3!+zjw=y`_`uhzTYo!FHJxDZD8Tw@9!(0<`nn)){Omb zp>|lj!@u9%ben5|-ebe|>(@n>()p_$$WXsucOmu=_QAoOMHzlw z@Y)SWl7^;+hwJLTh^Uv~vINgUoIdq8+-bHS51bS3kdaspiR@}xSX3xY#0=Xm+) z=$FOJXEps7vsc~9H3x^L@o90W4cOLCcAnSxE&tQ|G&nU(&z?dCr~H5xj@Gv6?{DkZ z^M2y_+bc}xTeenw=^A*XT+e~brqaZzBamTHFbGqz)Jh?N41LXUz zwfVj|=$9G!zDbx|=k3R-=KI<)zss1`mKddb$~dEHlz)SvI0C68{c%~ve)aKN8OXK2nltK8|M4xlcPvq_Vf3twdX4Kd?m*Dl`M_YeVycxsxWDA9*MOI%)lO(0*2ng5 z%;PlqiWzfEI^1{r_;T=*Ji9 zHZ^cIr(N^4{of3S;o+njcMYkkJFI&c>`0HWDGIxu?7sq(F7 zruF4pZPaV{bK3p;vy-v2>cFA%vewgdKkMnhRh~@^-FkjhQ%|>oa}BV8(w==S%d@+( zJo|?%&%Tu9*@7(3YIr8OW^>RGpAEIukD-OT>qTswOQQ%-i{PM%t~*bKif}l z6Fl6{q_@~Od~7F)U)XV%GOp!`S!vEUuBExb-nggf&o;8^x8(8+G=5QlZj#NPZDE1m zF&V#Tn+fhh`{D^Fh+m}9*v9g7{2$`&yWgMPKmL!&ivMr@#G~~L{69kduV^b1|95hx z#k)f;Nczj8@ORa&ZYCzCn&3^y+jirlERHC@fxQZsbDzw9YNB6VI%{bsd8eY`PU5fy z#9s?~d!H#FpJ^&QK2}OC8G1SAaD;pD9df@>QcpVaN(%3?%3=0@Ha4iA?==A9Y3H0K8naoEHrmUS>5Ss@X^k_6W=rb)kqilAW`O^b?xWe+e;RXTUzMd zK_qz0Cx)>FU&vJEGy!~*k7Zl|XGoI&`F-Z##zm9#Q?KSJG8v=N$mn7CdOa%yAWv)H}|J;~1=n@S&w4dSQntv$23cZog+HTT}e zv*}sld{e-28ggv@(EIc!%`-3lmE4r`p=Dd zxsY=}I5+0o*TWlF^M=tz4)3;plVUMX^C9W@r~sj)EU#xO=lEjau3*Z0=U zboLaGkJ)t>>nBe|=f*&*yXE)LxiO9O)9<-4{lI7aFefpN+{_*59r;VgE$v`$?jbfR z9b_LmhK+mbToT33JBh*Se%HbxXHR#qaMy(6wGU1R)IQMd6gA+-+M{@-1^qpzsBKY-Tcyc0_xyFe9 zD=$)Kk+xIkGmmE*Ie%N{bHC)|G$@DS6TQ7Nq(6%eYIqh$=bL?@6ILEE_VrrFt4~@* z9*fCo=ACKi-*buOk{4W7E#65yQFRlW&v3%89=rL$0(e(7cs=b*^zCCeJjmG$W!1#r z7ovaTM=7f&*5AxoaGPnHe>Q&)BX@utg3YweAI9J35wjVkMb;XdJTR!| zk|)+YRPGG!ap23+k-NqrTZV!YrQqTS;o@N6J}B~t@E`qNaBCk5?x7n?U+x-*&OCxS zoXLGi&Y+&0cY0??PU_4x<%%o+Ax$5quNbs-t-%*J4`6~1kH-4&s8rtz+%_ZmzK&R) zaP1<~<{Jj@gwNoU`99n{ciRs$-`(funC~ubp2c^In~E(zNqw^LaV9wK`2Jb{L=k7> z`Rhqy@T=+zor8n1XGRyg=LN33X6fEB=Y`51adIQ8z_)5AZ_x_!CTlKKUVG5Zjc32d za9_^#JX0-ND=_E;kGp`+IlwE)gQ`5c!NSe^yT%!`tB1a(A8%N*8XVcd-O@=kE413c3Dg*Tl}z?@Yzp0lbPWaAG0-o$xT#nkG35!~Jf-mj(>aUZze z4gO2!Oyl!W)?CP%3t4kBYaSmcj1;nF=66XivLX&0DEF#{&-1D{}ehLS$z%w%=>4NB0F61wx4l$df?< zhZ^gE!GF~lh%XwvPwr;`yw8u^L(ZIFcDtK9_vGDt>g=th`rFH|nQOkgxI5c?cX7Ac zd>8%XNABdi#pBG{gX7dJXYyEhV;}q;X#K%f#>&ipnRjaWFW3OZ;orTOvitF4Z1wqh z&DkX}aO*Vc573oHhOY(x22}61oft(O_nPx3G{u9~+^-kr$0-`3W= z4RU=#X3QBU!43bI)*snPu84R_W(+`M+897)?VIkOZikooYTpe0JYD7DkKXM-#za}8 zJ)@6xoBOA&?yNkR*BE;&W7`;7KK$>vaO;FyJk}Fw`D(^+x*AUK=^P*KR4Qb9%f$EPU3JpH*K7kfxl=pS7!)1oFUlc zB>ou83m>L0|9mFfeyZu`ckp#&r}U~g{cU@~`YsDxyEOwZZ_U8V7HUkzXZ{Y};^x!2 z=MPNajNEsJoRjit);fuI?r`g3s!E-3dg*G-mYi@g^F6%LNn79D9vpOrr>URW{xPHf#fRW? zKz{u(?t~972_^h6`GfZkbr^|N*RV&t8?6l`V)T7zEA=~zLWxRrP4zJgnWW$1^wEYM z^bYToJGniB(R&QPZ}H9{KqiQGzv{tAHm*;-F!vW){)=n@Ua9jfRbx?&oI8M=i?Y^t zfK6o;dh9L02OSLi_UZNLLh~PI{q^bVuk_Z!+P79ZMd4Op{72St_?GlF|HNyj+Uu$i z>$!7@7aOjZcm9VEIH6-@JNl_=q2V@37{|Qr676{rh|rniY)%p^=(UUO0{(^T!t2 zp{x4a-C^g+{#Jqyg1d0xj*L0lc?G>O|m#Vzdh3PQxKF;1sr%J>7 zVuSZP z^t1j|{j6X1YoGpn@Vu>`{uZC0zanf)zdanA0?kZH`dzommm|MxYdP@cw|JG{>_lb` z(jAoNhOejIEeieW{Eyah=)9`0-}NdF?lIu4x<%Q|y|vU;V6!S#SL@aOCeL$`4On~v ze=B`R#&*yja_yO9-amU}{H?)2_)c^!<8Nhd^0&5v8-t*SR^U(E>1q8fTVHF5^4@62 z`40DdyrH%5`=l82!pHb@(r&7E?WVwG;>T`HaG)(v78vRKI~piU>g8VhY(9z)dHdn_ zQT(G1*Tu*E@@`>?_9!LqpwWc&@wUoGiHx1hdeO74c5MUCCiY}Syat}m8R5OizusQ_ z-BShsBG-@Ty)Pg2IQrO*A2I3oEX@gz;@L>~7VGg7A~QE5JB@z`IqLT*K4NP8$A`aK zfu3X98hd7|^^bk$Z#|zP?-`#W_St&uIO9{i1D$9|hEEZh=<_KWJbVPb=x+M)`xNCL z)xIY8$;wA-A9;VvrxTx!Pv=kgQSs^ZcW#gRF*P|;nbS>u&-?FmavN&oBcBfbAzwOy zrR=I)dp`xbaYaKb*Dz>!^XBGnzf(P7(PR zbXP^}aPN$ypPesfL>QBHF7jt?gR;=GU1%({Z-%ey0C!{NnW2 za-98yFX^13V)+Ca&&CgyFmA|;9~i&5@NyzJe1(U@&~i2$-Wl?6cw#yZ$BEy|9@xlx z28S;Lhp+H(7~a=Wm4UJav1{RQh@a5Amw6fdAT4eP3d}yU)MO zce8%vvFvr}hC`{DmrNF)Ip90{#_|{A4>h`U>U;*-$3gH>y1?wS(Fu)Q&57WTdE3Db z{uHp0e3ky#4UV=Kd_!j*7~f`*{vHl|<*W4kn)0t3U!~{QB%U}U55H#8Psy1M`A&XA zCww^uZ;+p|m;8hFt*ag?ck+9p!cF+S>?!%Q9{L<>f;VdpeSCmHn-k>1=TZY7*Izf@ z2e&akxQ+6`O)%R*{i|S>O`jILx{WSaoU}c3mW6z@k73`|a4rRBTJ6&L6glv=AU50G zoLf=J8cXmOW8ZC^jXd1|AG;YGC;^x3-BJPgq4w42LsgzW#Ga@9?d>h+ZgHNGyT_{c zwtH;PIr=wbi7P(_cHYp|FYKlUyR@3G|w9% zRz=)pBzdC)^}kH)*Z)$CZuplz-+lx8?9z_$znJre{^{xztqSKRczc-Xe=+)AKlMi) z;EMJqADZd!OoV;uz8AAc=dwpW-^*5;e`)u~)&mJwYrs|fxf3^$+abP}>3eA}rOrAt zKOg(mSA%56{Nv9~j2Heft?%VNZ_G^Ji^jCR7jxb`W$*n@W;&a3oK2357?jTeoyFUzhm&|0p$OgJzv`Y zp6~Y8(F6Bn-PvGo+|$YbW3P9MPyF`dr;z^#-|Hv;uTMRDoAOw&+e8Qb<^QqoyOnnf zEt&HRHeQuZ=R>@G_vin0oAZ%Iq~yt!>?`}3tUur8lxdEAY$Rl$a5^(5#oQf-{E>~7 ztUbrKqCV=;BV}VH&!)6}pfjfAap-~M{d2xPWLH5OrtW+vGQA!>GS1wZ&;$CbEgvF1 zZ(FFW*3&U%rx{%xABF1mrLVaDL1?lCI>vrAx+b3rMCY;3(&CV; zLpeF(>rXM3&Bv5ZCOyG_9^fhJU*Oir@LM0t(7(8En4Zbly$QK=xUc|!M)c#yZ==5HSwZMeeIr8 z(AU;^XPpdrAtjdszB5O_Rl}oWrRb!{82P^jdLNPH!%M-_0maMeM)a$%J@wdMUSGQ! z*kh8J3M=al@yxCFS=BZ`}7HzxQ7=aQQtPK0i?TeOP%IrzpRlx-xy= zi@kkMli$V{owA=Z&o{)E_>APOKn(OfM#jvxDACmKB<&)7Eea;!oVlSTMtwBBkYcIY3p9-i_qDF34Q zgRXy}99i2d3cAo9bcpVr_PfUGq@PhG^ds&m}Go~hL_m0bR$~KV;(n(%cC-~9n)IKmS zP|)xi@$8+vqZsN;_F4U`MrTw_!K37WjSo0w>Sr}EzA^OCj&IW6$7}4L`q{xAy7^y& zc8xK)lBxVJ<3}W4avyzv3Hhp=lWiJ@yq7_F%$Izx9xwlEHNWr2xAPd|90(5HbwBwn z_mgkg>73C}9T*Jk8h~9v4|qI7e@kwtzH0xa>;rnAaxr_r_fGU&of9C~juC93P2qGg zIWFz!r5)&io%r{Z+qoZ`tw%Kz9{jr#ZVp(VvzUWk{hP@F8ysm7&7*V5SN5p+?&=(y ze9t%do;UGbYg1nIe)hrS?y$B>Y|Kt)P(!s-)WhArGX^c2z0|;m@6$ha!(I!U@$dG| z$Ys6kKe<>t@&OU`L6kN^UuDj?_U1dUFNy#&$`omch|Pq ze0SIO_k8#3Q|D5fE?HIR=^2jq&*&TVu77CtXwm!!Q#QzL=&IxyE&^Y(=~UMloholp z4BrI5mvOHwd1!ei4^8J`Y;y9~C>KpRl@(_?PhZRLieam!D0(aW0X%5Oe;%(v5344Z zYx|0a;#Da=pnaV<1>O!0$VaSPuX;Dnbgj+zsv~DbGHw-gYx2F2Q+0gTxn9Dd?ciT) zRVeW-WRjlQzS?=Fv(p3kJ=<%rpSWLD_h9bea|`nnZ&+aNF@$#cy8-$I*RQTFat`)@ z`=tvOEZszoou9@;Up45R@b^hO_-(;8p}oa{{GQwK6?U+$_Hyu@_0GPgVs9(_Xr#F} zb|^MQPUIbO9^&A&Stq&A^=@8qa-G-Pb;1YCI>{?mK71!@>R>JM3;#1Xm-Lt8cXw+7 zJm2A;>2JYf5wdSgAY|@QCifc~GJn?&c-IbiS0y}VJ#Z*h9;lPkb3gj`eqa%U?&0@a zWy_RrOEhZTO<&7x*1jm$8{FA-n^yyMFTR+2;a>q@Ab(6d?>CW)t~kZX)?2jTk|W|T z%SF%g$hc1SRQBv6rbd2z_-E{+&gS1R+DT~sHtzKsV)w#R-2-Owp_#w#z%=WHw{^0Y zU97tsd9q@plemQ3;u{ZseeaF4$P0w$%+GTY3yK`@vtXC<*}H&oC-^Cx(>M$8Sv`8B zcjC_&2j6<4aKRUkj7RQoH@xS>IH3Gp*`vnC?BU6M;90kq-&wS5V;;7{^bX1&j~?(BZ;r4BK42$M z{N2bi#k^&!C?`&MKH58Hbpm>hu|LrP>XD7o^@6$1GRd~29ujnY&^hFt85xGHu0B4~ zd}S+fi2%YJEl<-*rRfcxO*FcGG(6yXJ4d@4AXLRXV{n$}JELTmN+# zc^@U{f~S;QU~#W%;JF9OJs9}$?H8HwKp*tSQ=%s4yFC2mXO2*(a^m}xYUuwhuRrO7 z+2SW|{4B*!bmxJ5aB=w;z-B)AlpNn~&{#jdSFle+|6468dceC)7sx zq_Z95Lkl_Ozj3JXja`b~g$%fxItKf^#pDDAp~c)ZT5ZEtT}it*Hl1)#xE4jPucI%Q zcGGEF`GqAKk1*7 zF<_#*E+j9thxiko&h}0eT~M$*S>OHb&&VwM)BKpXKlH7f@6qgw)}k}2m1F!G{IQU| zE@ZEp7c4mPDE3%+AlMV4FMN67BKBIjx@xaG>%Oc0rqBoQ+rs+}{cXzJUwGQR8GGBH zw%$K0Kc>Np{Pb+~pRYZAq_HH|jTgMAnx4z9aN`AJOX}-oX(H5aE_UL+k%8dQ9->H?dlusb@E7nn z3a7vdMuZ?6I~^9=c2kDS-D1?bw;+U!wneHXR$ z)(5a1eCx#*+PZc1L&AN==wYlGPa$_SH}NmKN7B8NyXx+JI_-&13Lb*DV(9)hz+|gz z?2f07G_EM}#i16A^YE(J@L3akA52XW_|j6b!-^`}lVIzW1Tg8DfePp=nQO#j9Yag$?o z*3mOpzVS%c4-d6Xah@N&_`7dBlKAJo56l*1U0T{)R)jAOFuo z*M9ZLp|+~rBg*5}9jJy!-a@R7eam5Q6@Q5`-fwrm+t_vsJd9Y4&L~?!d_p*JKfY@H z?!rSb%8A^v=iRg#nap+9W~{p@y>I>|8lq`{XCb}E%5A(;CoXFr&qu)gaglC z1w0>gVQ{!5C#|Mtyayw}uN3%f1U{vU|LxGm#PLJ%+nhv8)5C964S(uT%eP)XRF&U) zNVQwNwP!cSnMLf zXsS3e6*y1Dm)d*A8H=lsMQW>Dbm>mV9zbrx&tmvcRreJK&Ly;Z z=W&EbQf#p-bmVH%zVio(!{LGlPzwOBHHss&k+F$rpz4jMA8pHou z%l;m~PnkdeBwkf`@c`GmZPQjirYP>OYZiLYdibK4E zkDGD6gss!YeB;b_E^{4AjATCZE_U*JMllCd1FzV~2;hWF(itHZcU;~k-7+sy0v^eZ z)!8QcO&PB+UfDGFQsc~=`Q6N! z-=P8QX^oZj{H7|mKDhhi(0f&SeK5Yd#6E2K>0_N4((kb2%8Ji)76(QHj20@d$x`5rPp9*pM(6F;>?I(<0sFP2z-N_w8C9K2*xQe@PYq6@ z^C|oVO-|y#!w&XDD8ZeWPfuj+Rv%d6=>vC4-v|x`zlO0V=R|&%?|f&)aPskfue!;) zGRbip+uBKtUUJ;VwxTaM#I_vzL~p~lo=^X=XW1iomP_yXk#*o|$uQ*(fCDdt5)*Ur z{nGae;zf58SMuu#Pk8;yt{DYSs3W#kqnd5r`yB5bz9YS+_Z9M%jGVdyUjgp~*st1Q zIS20pjt2i}uY3jgS|Z*Ez4N#9&JAbRyru_t3UB`UTYTKyCyg%DwgehS|7~Rs_@G@~5jrkmY}>ZI z5&t8wT;=Ib7Us8Gm;+-IXG3>V-`az1(C%vRn4FD7J7cUuf6{OF2*=T(s>V1o?&Y_Y zJX0)PwBN>^l{U|#&M9cv290mnde1{K__^iv^I5CrH5;E&H9WH({?Pg`^?~Q<{T%Q8 z+6UT^1)F%*fjsDZ3YuRqZ)w-#O+Yec|U4&CXa`B z@13a66aD7$*BqB0d3{+u?{@b3#gzT-GGzoJrN-QV}^H)zZC@t?1b%rm+LLsklWocmjW( z5nU80i!FB&_n?1wz;hJ$v-{mbF+x~xtPMK#so)>Pz zXC|Hf$noCB4)Ety@Mj136JHL_K*#c1bb&+N;7@c>MVVw|C-}4^z_fhara-)!?#s=+!LuvN>d&Dc*4jRYgS22Hm9}G{KKnz*AMC1U? zczx-@rGwe$+Q6Vi734~_vv;lRZ5w;JGSrv;zR1RAqd$woO~AnybJ$w4J@14xl{=|2EO2;_BcgfE*8(=7N9q#HU(ifIj(oX@hXfU#92skhl92oBE z3*bl6ci`~P+V}GR`uzv8Q_vNkRxL*W-%AeiO|@R>cBAW)WZi3J6VB4QQ@mn#$r!hzu==nC0c}YE(HEAt3WLeAJ41L(zHxn-GYu{)OW!rQ^XI;>kwRhs+Z71@( z<5|wsK%e!W18@i!8@alOxIbqL3EumgoMro~$VH>>YX5Da#Gv!g5$L=9DQNK~Xt@X; zfv)4lHVr&30v@x1u8q7tNIY}o1qL23>bw`#T~#r+shS^qg$$8R!P$)99eX8t@+@-Y z1^Co9=(Yt~-B?)`A!kkF_1xl)3C{Km0*CD$IO({Tto6zq(~dC;y4+7+J@k1yYj|_I za_!FOX{FDeE16sHOlal|aBmPgLap*D-j>ZY@2`-NCzxx`m7&C&(3<$m0r3vs8sBOP zE&K7pOP3zF3E16&?lv9U1ytNocR^S@wY>p+zE@+bR@FPFB$yli?z-cRz0fASWbMlB z=Z4#s&<^;t!H?U3Q5(Ot0kbyoCI7qi!SFr2+s6LAj2_rN)s@3KtFVLrIx{lH`jt9CI#Jp~*dRIDf0K!P3A=`s(dHr@1Z9dFFEZ=)M$uDhIDFhbMY< znKAZFwsglwpvgOs2~)69-tFC5<4g?4CxPR@MgC>k9%(pG4h~!j4qVQ>HU`4yj_hre zj%Q@`9rQ`Ru7Ae0hf1jT)w5!9dR~+mhd!LW4-@#c|s>Qp9yGV4;qvY2P;-LFDk5T7N+5P^> zk5l`-0lc^OeZ2&4Hurr&f3wk1&|{MIhtoW`%SWL7o@w?QA89eP3Z3S!L!TrcL1UUb zkr-FGCztrf{XXN`{ig0^t-UkmMB`q1l5xeuHRg+qr&yWIKe-$KO$YR5?)~CCMZ5QZ zBLBnAxdUA7lpe-s=-UsY+j{jBR$ltYwe#N{ndRrA$sN?{iY71g>0W;QaUrn@e|*uL!GUa7oH1U7Y(u^& z)~NjAwyDaSbe;)-3zBEQ`{uF64sfuO?=KsFpYqRL-=+2UwM~K_(0(&E{Ju{(2eBb$ z{O){b?H7k!i@uWAqxhHkd+@c18yC)<+%0{j3c8#QPw@KgLLQpf+Z5moEZe4r5?{t% zGBVPOHM+6bK@s718k|BO34g!`lh|O^GlaDbWj(`L7qZ35IKk^74_?`Ha4XyFc~vI`yd1_!3kP;Msj~X>+WUUFQKPvowHbXIeDH%{4E$;xC-w2eGYJ_Ur;ky_|pnL zxoaUt2Mk&+f2UDvH?X-0Sb$rCb4+{yIP3hMspu%3=qR1wqwZDcMz zz`0CVy*@UDKmEZg>(Ikri3ixwfBmYHjCdS=;V@=%sn>y4bXJ=D^+m`W=DT&}ow?z5Y<>THw=v%+_+otcSE{{fyLRQoj}m*UO>5iD zxgffeS-JPR(`hU3TKiscI8)034z$8U{66+fdsFL+QFEKAzv*0U@df+O>T8O1`q!sB zz!igNzl}3gefxGde8lXNw{OkF>;wz7Q~RuY*jB_kzn-?vud=RQ=;B83_{PA*@P6R3 zpLO4OCUz9hUBA4S4+8D~VhAw2+~rgIZ>8P&zyN$uyD0Xa_)#cO{+s>C`RNzA^peac zfuGFue$#sfhP1cu>%BBsdUY^wl1GzCH^;`Dm~I63`Ydp__WdLMz<;5^uk%y2B=BUN zhuRKqMd8aG_>NQ?V{q)X@!{*BiPi4{6W}Jft^A#lgB1H68J^r{vID z<-6l|zA6S^74AqD9Vd@Kf0qP#pJ$^Ah=Zat#gKE-DPq7ffE-iqlxwGWacJ6Q;~84} zwI^8nE3Dnbd#l`dZ;?4eC;h(I_8R5_9w|oIR_!Ei7#vDW8^V4MWnT-hgMo{stpnnR z(EXdpCHK7JBJg1lK9Do;r=1z85PYU+9a2z^!kmvUd;vy&d8gmvz{~7K?jR4hUHK>()?LFpjAe31fZ?{IqG8{e+f`mn z=>%|mgF4_$+O;_JxDSUc{T3l_-yM=YpKI_(7$5u4`t3v)F21|_sk_?Sm}fg{8pe5U zX>$!`-*fHt&Mb5MPv6|cW8D4Dcke5KcO0%NOgj%ZIwX*2FAXHxiUZhW0b?UduhIX* z(7oU9Qw=@J?~{KH@}N<8@YN?fj`BGae%q2;COsm`JJM66%V{rkhI5=;k{@3JE{v#1 z#EL7Brxl6HvAvCLjiKg_ri$ja;knHn#d*zoH^zEfgUh*RASbe(ccA~>j&f&sCb<3Q zhuD{30DN%h+ij=!K71+t2kGByXW&~^=++dSV4qd*-QPa@^$(Xnr_O_n7K6K`_$uTV zo#`xo6nR?vd1o}2QG1TwBL9PYGqrbSwA*Fdk^k4Xn;z>iv|0&jSac(M40;&iRdS+VBfXR-I%5BN-O7G9!wcv~_0v2R(O0Zb%gT!ZRS(; zaPyAcvqrgM9oW38fgHVcC3zd4R%ci>A~1P;|8$H61%N75CgGfKab z-gE;oc-6+G_>9i&7thL%T=xWZel@_CeQ(0=c>_9{xZrd@*n_dJgYOUqbH=Pd`4rVG*B#$(-XA4@%}$=Qi|I2`6rG_b_FP6-0KD zA2y+4T-dLVP9%>cIoIhSL%)reF&45^e)k^6x-M_}t~vZyE%2h_6??yUe9ap9!)pSA z8jM~GAMKce&BENI&zo^(jSF8y-}+tS?VqH+-Fro<7S5iWyI}U@dMEgLaKWg_sw-A) zr0x}Q<(Ml^mA7)ga&!26fX`Fe`~A?a{ORHqLr(z*q;JZX{tZv=S$VwE=yNH4os)aU z9?r$xOZ;-N;{V%$t+Q@Icm#Bxqvyo2XTg(8h~umXoOiY6zTMQz1;c-6KGGOKl$LO} z9G?mJ9x*=h`Owvrr|WOqW!r`>NwuwW+qyL3@<(?MjjwrYiF0N6`ce;GL!D*P?F5HC z)DlOBhWMMfO2+!88Gk}LzGvdM@LzauX~2)~MwbI;lDLG7ws?Q<)^xnT54iYoevmUL ziStq6JTwvYboQChyv2LjPr_^ErnQ5ce%$T`zbszYq9^+CNcfZU9`L6|a6?x;k#E8B z5Q{&V^61YA-^jMH$u+T>C)%m_XuBG@7itd`1IwQE(t7)iB~UySA)B8 zcz0JV^8YS)?tJjLp1Y-}%kBiV9mmJ{TrZdl2 zydIu*0C{x(2gr}`coaFbTkl_y8&<6FU*JpkuOAmafQ*aPIe|FZ;kZ{^Qe! z@6VJQ3a(0r8>l?`o%L_};QNL50N-uYy1p0q9;|o$H9L)uFm>KZ^cMIf@;wgE)OjSG z@J(vW6YcQLHuxp9=fcvX%QG*5*WA*V*Q_`B7Fl@BZGCvn+WTMpfZ-!QCnJU)`f zYbwqTAAk>dzmY(48nyvFLH9ThP{L9_X1y_2!rW#%Yk86kLbi;G@!*hDz z%W1qu`|0O3^+{gCdW}xCRyx(ekpts3;xAv(9(lS}fBfa2eD=nllb?)x5BTn!2EG?( z!S@a16`d|W3BE^sH|I;ffBJ17e)8)dojyN#6h3gO{Nz>qkD2=Ijqd@zk5SwFp7=@7 zy^mSFh*-(ciw_1lCG`AF^F_KoB~bG}>0 z7Uf~*H?+d9QftbIY}0yL2F5paw$y>UvhdBo72yN7dOT4&8hkspF+)dNaf#96l6=$C z^`7s#QXSY&}F;OU1TrYJw3o4zR4b+`4^H$^*=m)e))Ojr%GO* zg8sGqoxc3?U2>jIR~}Uh#s$n_FfbkhjE4ea;H~3e{0?rFfc zGz)w;zX$jhyhr#Rtas(nYVwoc9dd3;9_emnPl*KbE0_W5$& zH-2yO#G-pEmYwa{*7yj``}^(tb-AhcJ-?s-{SBNmsQ2Y_FI9~YxnMOTn;we8=QeN$ zx_F3kB>#IQ@lkk54Bi(7Pk+nbnXve_4;HQP4dq6)F5(luV`BAJs>ULp+KPha*nIFG zo};?kn8vRPEIR<7kxVi*h47kIc#Yz!@_V#HvmMBfPIyfhyr!qBVwq}lxEt2xzvK*W z4_5zX-5}+*e>5Cn{m~l7#1BiTcU@1tE56d_M^O*FfqLMVs0Xg39yllD9BH8*xQcq< zAoaji)B^{%J4cF-8pai0`yfuUx8F#W?xpS zFJ!HFhn-gmb-3%P!<8@hCF*c1sl$DVI^4?q)R$AW&|2@EW{`&ym=8XF+#6M+^ zw4%3(E_3mb=8&9Z(%{^i4cBIu5*3_0)>Pn%}pEmTg0Jj6%;{ zMQpha+0$=ZV#zwUv%lZ#M$#|!XLpZh+|i7~SnJ6ZmOa#p+|%8H*gqF@|E@w!&jbdY0ycm zVlK+@t{+L?jALW!MmMkOKWF8_EPK_Ig`X@m_J1mVXJjF3kc^8X-`d!pcJ`%%{ZTxi z3m(}GkKB)p>nTOH-9($xie*i}z6Lu&@&71uUkQxuT$N*@eKd3Savsp}J^Cn)+)bTF zNN@x;ZNRI>ysNwuomuXl|4~mH^0r zILi*OX7S{m)CZpuUug;Pu6LG>m^c5de>^k)@U2>Gl;zl6_3SS&RlHnrbH&dUXH;BK zF?Pk)yXFIX^c2M#d+;F~xXakjE*#0HRF278qs51LA6>%cR_c5Yojt2N0Xva@UC6_3 z^pySRFg&YN!Ro_zhZYsrfRFPfaliyTHFCh)vo~4)Tj1&M!Bg$u-=60Fy~&*3 zDi6_5D6!T@JAcXj{OVt5KaUZ+Oxw@fy#1{9_Os@H&wjRFd3yACvi&?*<@*1-&7HUD z`I<54(Vwr`4n1~2kDbtC7xdV@1RR7Od*aX~G~E-gSoVA9rI&c(-N=ls7jfRi1DuHh zk5o;Z!@1M;*-U;{J2!@^`dZKTQsvTsew$5hz>nx-+c&8r7oUZml@Hu6pY`62-UMAi z|E;u-)4mP*Z-@Rnp#M(jzl*ti-Nvo$JHfLe-(P`U1Pv({EC*Q`r=8AEXg7R9{5A-0 z5WjVE!=mtJlVb~CR*cfkfy;rHsBLkI|9fM&{Qu&x(S<`-hNBmphTdoCy(Lg4dKayK z;dF7MrGh^C=C?GXPx$MR$<@PNDGuMT{z&6`@7yWuRcwt*jXmnhyg;Od7;z1?wQX5) z_B79l%(K(SQwpp-rivINAlijT#x)M!ETNfBX{7VbwT9uXZf46gIYO%DDKr<{^Gu) zu20?5H5lh@XU50Ev&hTN82@c&^abIPjOUWMTJI6+bR|P~cx$%nb@O>IPWd9>@6w-w z=t0rp;O+=&zE@Txg7=fl>I8Smr}|6iyZ`w0YerOK=?#)5| zsi+V3mY>`BUw3Bw?^fCK*!Q&s{4Z?oosJA5&tH0x9hY&gZ#Y@3{J(t!2A~d=K_2v61&nnfE&EX65okq22Y^y!h~U)tnJ% zsCoPAduyW3o)^(ct^OF~4sPYgW81&&Fh(o6x=W}j)4MVG9EMy^{_j-se{X0Wb%xVC zdRU-&?1eeat<%ZHuHWvBMZxN1bg1ov&xY-l)c2=EMErRGn4T4WHWRve0SlU zF0Tdem~!&Ji&FV&VCsqxUz@9b+_O;yMR37y0PI41zX<7 z=ag8^v*q<;!*fT3n$7o?vElip-uK|R@WM+&&124Un#YX{G*1|x(_Aii3&y-}*E)~7 zX%Z<6j9uQDI_pBY9|p!PuD-wiyo$tz3oE?!`||C*-!09J@Zhhr97S&jt6cwIH}`hD zJ7j2*=6Y+u@8UpNZ*N|+?i9-3>V#SU+WA-BUbcn3FL3G}c+Kg2u8d&KY6ZlIshB_Zbz5 zjb~RRzP0^mqkdb-^M_{^hHGn`W{WGF`!lhQwtwg0TT@{GzhGXoVEYa-HxxwfBPXMm zgE!!*+0!c5zvtG;7f0Us6gk%5n`~kGUM25^PV(N%r}*B|NeRZuG7pbm->-85;Mcm3 z!SYAZE4n$YJmq@k^F!=+AmSdNqU?~M6QGw02w+qVJi`V#yJjQI@u$JUbQgz!jsgJO8vpJMz^xHo^2 zbL0zqgKvNPjne=5>7k8*?T0Ghx3Qs4BBY#F=fmNtO`J1R4F6mkT9z}>SvKl5KCf|> zwE~l1^q=1-J@n&4ExaeWl|y~esGq^tpM$T?F?`?L^*kG#*+U#(wtjFXZEmK`49=|j z1nuw-1TRZ(yMwl)X`71=VbteoJCnAbr7d@n2PdbueZjW9g0_=sdkt+T(l$ifYt!3q zwQaAV?X|SMfwm#qeuB0)r?=f?+ulIiP^c{D%TC#-x(eh*pls9v-n||8&f(q9^6ux; z-~Ar%mN5Uo3|njE$!-tt-l85^c$dxIvixhIxq~dlzr=fDOQ08cig;NodU$68wqh+Z zb2)tDUU+;xzm>z|;3x9a-@tx%p;L)3RPkLtnx^4mbz93Ys2T1g=&8 zALk#WgI$Xss~}MOWapGXS)4XH|5UYJ()+Lf7v{hkx~5c=b@ej`*&?cE=;D1-Gmm`e zHhOjeeH29!CGR$l;;dw?RWbQ^747NI2^3!K9QnorQQnKvw|G(oyk(?#3%_rN*E>gQ zA27dFr}&lhdd003(@x`A;#2VDD6okEo5}*tEU5di(H%oi9Bo8*eWI1;6;9nl(ueBF zLyZEznZP1?zVztH;T&)+hjX0ehreYKdGW}JfQOs2kiXFARWbUP+!+k+=)3SI2izP5 z&Jdfx4nlvWM#ibTDm?q2p-K3fVh-ZVv+3h*o&9tZdx`y{eiGz$=x<_7dMt@G}2`6PYP-|mu7!07nu*Ul}Do+P`ii$3*TbUc#&=CH@32h$&Y&ZNdr zz7^xQyd{))=JT9y1rtumvgcHS;kG$Kr-q*%EcbXdQRJ=0=PiPfy zsE02!P!|dP7ovY<#*@Cq{1pT3LQg1w)=kZ%v3Uz4|B20vj5D={@^8BJT~-%%WqQoi z??_iEj%+6$7hgA?`lLVACZjQ&J7hUc_+a)P=I!DfzNLmuHvneu*2y8SF&N;oJz zvAx$R(^(*|P$R0dPd2b8=8P5g=~m{ivwBvI;>;dsWgEOJ#QI->U)Oo(icSBg)LAWe z`_5`f_5+gJo;3eA|IPJ6SjhX@800jxxhY;b8+SZ=S?5vllgGhN?2!9`qzE% zA5R_5E3>Ry^O)RSM`N{_;u~AT3WbZ?9%zjkh$oPe%Md7bMRn46WAMA zKjp%D6=%d5SnJaVYn=mRar%2cSg!!5?b&ea6+g#+rCfyy52tI8r5oAv3EsH=^QQcB z)_Mj4_q%*>e^EX(@5~#JGE5 zA9$xR^Vwrro^9rt_Q=}kfAwKkw(ZW4UjcG$MlHGj4g6h;?49u9^Y)+R_d5;0ACa`> zZ9KK<3g=*n>K)N<=QbQ`gfBfkiu2(n04t5(N~|T$S>j{sjy1l7E@J$EBj5${Z<5=6 zHhUxA8~0!e=L}qCp+_nHrW)r*bhiTEjokwLV(7%hzoW zA9Up(HvINr^=~SfQ>BrsI`btMsRlOwzi;aUzmJ6f#s|M_vQx6*zf`;4kG`N9TETD> z=anMcmQkC#r~*96VIHOYS03$)vPHG;{FlGc!e9PH&0FtV_!ni=KdKM)rJCRM&{-+> zmvCpu^VB3SD&g;q^A&S_*Ge?AyeQfyJ<|9BEvK4J{bY1h3nG<0?Lj zkr6WmW9A}Vb{2SlH~;IFA8C{f*7G!Z>FF4POQR1i{hj~f>Po5e7243tTczitN6Nmp z<0_|*7mOQ& zKDzFB!PQav)BGKLO~stw6@AvbcQWYWN;X|dYnts_Q|l1=$M#mvVSnepbesC+^!3HD zr83tjoxYvE{A;v1TUzHY!I|h#&dz3?rAucnb=SnbAgm|!Zo$3fdj?9y4) zdE{?Da%|-qb0;~^+M(U`JQKeDH9T4UY0a9S))ANPU~JO#HEEr5;T@k)o)fZAe^2AP z<%yXztv&8m-E|gv&UXLIYS!bA*FfWm(fo(rdICOr)16PF_;2ho*0A-fM;aaZte`j1 zeJk`<$*1H<41bhxQ$8!LsXzMK=9_1>^Fic4%bcI3d(E$e`ElOu^P~B1_yahm^Kbdv z%-_;|A#=%vjz;m{&cBZNpOpUpKb+rZ2Qa@o1~9+>pMKwe`V~E%3jO|{tKMn+JpV2I z{w?26M864YZwE@hJHh!(|MG5Z^X>2(2U|3Myq7ar@7jaETI1%8T7Bxjy|E6!!~8zt z0C%)lAMu;~X65(E#@jzGeY__##(U}g9dFsEu1+8Cdl}EY+tjPb5M)$xoxOOLngMqigMHiA1H z->+E9vA<4_wOD!@CPsk1eFC1Y;ueRGp8{eJ7@o5z>n*PCYc zGv7A!SmpO=-ixx$8=Y9T$A4tZf5p+%{Qs|i{#oF8Gw`%_R~33?>TE9Vq?w-%S8I1@ zKI44z>B=162fuFL_^ZwMDSJckTMNEL$FWB2jyCM8>zPC9o;&MHmj0X{`Am;DpOe@@ zj&E+sJqA{ahZ?#qb#br4M;rV2O|+qZ*u&)9xOS_tf6ezCJs)o!V2Y zPxOiR8ee}Z-W$Ug;SdYcep~v%?)nb?RBV($WTrVwk~$IS`>_q^(5(3-hyC&9=EvC_ z;1lA%d{6%8+1gsXF>wuW=Q-b;wLkyrgT3G{{&cX#s~0wBw4Iak=@6@k)w{ku*)6&7 zg2Uj$!;W+I@!x;E`FM@9XAb#fidD=6?rwgBn-`MD-_crdd_J~W4R+ui`VQ{ZfO~fk zKVv_prMJg$aLTKX!_AJNx)NEwoYJ z)%ZDIVxMD+fb$Z;T0Z4K1V882mF8T*g2>VH@hhtjU@u>1bO|=&B6mJIPeD1(E#%F+ zIdegLq2hPB#%DTOIlTCx*Ch%j=`1K>6>Do6$id)q&bPG&i%Zr%^W{}ArddC2bZ3IQ zckb_sS!6!*_oe&)ln?EH$Mw-DcVwA4Kk@m1^Buuhu}QN()217J6}Nk1NUU_+D9Bm=bP)U@!ypnYg{2YhaSDzS^ngXm4Q7wF66GhInLr~?A^BU zj=2M^9DlrgV%sKi9~+<4W0dK)_27YGn93y=j0Ok53C8T;Z#~yD;Yq*wOTjEK_n!G- zZeZ;_ufr$p+@BYY4XO3+OVoY^BVS%39--J;mOLr%UHLNP<50ffs|#jM?uGx5!#1OL zH1RAx#rMUhNvY@;J>*Cl{-4tVY9zz?m-Qj^Ni#Z(LJ1T`O)}zN*-tQE%Pf_wa!;l4T;WDwQKw3dF1K> z@2*R!Ww?o)59Y6Xw>#LI{p?X}3OPZH(aJlzA1OiJoobWpUd4Gg=)t9ieL)7AebE`o zzaifxb6@_M=O^WF9*!HDXFcm#>ju`G#?QPqHHHp?J>TR^$vn z;Xiw`E1z>U*E#tO+2C;rm+mPB@Bgp}aBa52`u7D0Hkb0KX=$es49f(0Zy_SDp9^Z7XS4HJZK5 zIr3~3e7qicdo4U&bp8q8QOoBy;c4vao+|Ec3^JBzv5LG>=_VoGDF(M^1Ct7`uhIo0 zCkL>(Dn`?0ypxbW+ONZOfrl0bl2`Y8YtPKrUY#Yr-il0;Jd*xc!+&!&12C|8P4ZU? zUlhw|gSI5sML*^&Yr{`-dp=1{mvSt6hvECbi24R>0KM}DGHcca&^-Kb4!<`c&n|;E zyo?Ub8MJ$wo^#LSReXOgxi*4zTksoCw$rc9bLhZUh=H$7$g&dPp)>3IYp1)@_3j+{ z(ffK=yvXEO(eDpg$BW2PGp_9qx!PZUy8pNdxv6nCVK4UA?)j69`z^**d_{JT=yDZ* zHzA_~;GUVw0LECy7-oN2M=rSKU*EOCBhTi7GcSW@1>jjP`1K|5OKTshu{?a!y_LBh zzO@1GT<}f$j&zMCWab3!W>*ZWkpAymFm7@#_?8R4<$`bO%ffaGH3EODeyK@7r%A41 z7I3Ns_Wrf%oep@D-2=^A_fGz~*Rl}#H_)|=)mpgA7P~hcH%@?S*Lb)gTPt(T;vFZ0 zfpTb1)}|ZSiUtZ6j-AZCqv2x4T+Th8HGinYu5;rJt)0`8}*cgpU$ zgSr4<;{=cVM71_H$69)0(~(4D4Rtp)b>qY0kx^j3PIWiWT^$bOrPtj=mmh7MPx~0U z2L2Ob@TzpJByX0F5kwN#J})uNd9ad`SL@Z z;Ng8vU{4XXdXJ8L?V(;LaB#;JO%F|N`sEv)wb4VI^S!1#5I9)yk*0^n!ymhPpF7mm zyZ=x@pkNL5TI3$!)B?@jkzm zZEx47WBo%P!e{e-!tnaz*x%rZ@I!i02Xjus0X)43eO~i(*Xzy&9cz-^XT{|G?awjo zJ%5zvZ_VX=Zq6v!n~4v~f6!Wj#FAIBhnz96cN_b)fVr%!bM_o!E~W1lOj^ZW+TR|< zcgSyH#pbPFQoJjdJmZzo$>Ej9_T7ii=PHN49sXXS{3z#=aQUafW#oAbeE%KvG-p>8 zzBkve<(N9f%ec=WH}XgF2m|od7u^^dZPM3bcnST8pZ{jKV3YBC24BhZ;e+(zuQPAs zzX}$a`!AZwNje-HoOb_(b2QNipWi(TKF>TRexJMz*@brAx}#g)XYL9n{)lhuU&%lF zPENXytD9UY`M9wA!kTkF`vgr=BVUntB;LuWJ81A{oS zPUFl1Ru)EIq8_pho}G)l(Hxb#IR_g}XBo)P`z7Y8@1vP7z7vz@C_ZEGCpg&TGKo&M zQLFt6o{OJ|ezphQTp|7K!Y9YTX9E77!@=A%_(bnT4mTiIzU?G_fSp^6&UHUHWBsG@ z9TdYqgI}+EWA^sChgzqS^KQ;ixR^c8iM%}wI}&?Y`)_18HmULR@i#uswT{lhfhP*b zpLmCS?B(#od(o+rygoN_%amh{rPw9qjJi^{Tr;rkVNDhyPujD_9fC z-xK{dea~l2Q`j5UuY7sa{x&@WUh|Qmx3P{}`D9+2$KB9I<;Y9I<2eg*#d*wc5je1f zHrVPhd>s!YN(vmLB` zF8dGzMjv8s`kcl2R&BuTUhsJg^K3^~;ZCJpoyTe)>~y~Nz`fiz-+7md>kGhjc+*u~ z)vllDNpdjseLT3nnmJ4WkMD!V{PSzQLwQlq1bjj?lZ#JBbfbDn=5ke>u|=Cr^ri2> zB1t#cs);DNQ0%TyVhU~KizD&v=>8s31-=$wRFPL*g>uU!V zTNaEylDWqDTH`G)Uue6DPj8);Kg3ys`kbqE@?Gnc%^+T;bqYRF=4WttI{MT8-luP% zy>MU_b=)OC`&hVjFnO4ZLgr4zeVY=E`~P@i+UW1y-Mss|lfAp#r4iNa`{_+GY#+MJ zC%b#`walK}#k`A3UVUgkdUYqdOb;@wfWNEJTPC2l?BCY(#{S-Ihb9CDt*PZ%C;CmN zQ}MvKKtW{7@!qHVo8uPd_~d}*NDORAI!{E3+TRvyQ7sPk`Yv}X4=uEYP+}A8V ztT~DPe#~A;wp#th`g(4rCR=);tugsdKlXr|FH)S0&6xPCm!BdT>#+Z3+$vYE7<+em zO!#VVoMe7VjTe8#_haL){`{0?^a+1FIJlB~1(ya)9L?(9T2E|S_lG!F4~sAcaA|!EzbO2%y($m> zlGjs*O#ON;aSQ%#e+<6yEWG3?H`dkpxcuAs#wQ#@UPPPV4Yhni2VN`-UZgv+@*_74 z#ot{EF9PN|e<<3-I-qgY6?Gt6JNE(yCwnYwG5SOt{Ea_F8~TVp%jaG`A9qatzUCcg z-W$Dn*Lw55G0VJF&k;5AcF(5Lp5V7O^DJhLt@ItG-xyH}T1n;|KC4x-os-bG-T);m7vOb#G+dHw|pv z%6IJ#Kcd4ci1)cZ6vb@B1LE_M&*q+Y>4K^w4MuKf{OGS8YNYei;>7<+K96+h82z;J z8#*w$;bq~Q(a#N!oX;G8jSr?D__n@4zQTgiBT?S{cfC`?y7bNq(89lm5>q)hEf1XR zfDd-U56i*bR%Cj|tI)&}(UP0rAUt*9r8(zC?qt3luksFj-;MjKofB!Vag3g~5!}7@ z>|ld*ni^NOoBOa=9_^jDirj~>$lhxg)EtrQl^-|Cd#$^W>(~*Ub;HAjXK(*)H}~j` z4HWJYEsH0aIF#`nX{@}+9nHrZJ9mW=T9e}DZNTAP?20yMw2tw`Pro?<{RA5C1h#S6 z6N?y!wYF{}&%lGFVEg@tjx@?QDm_m0tp6R*^k(+16W+5A-&i_AmOKt!KD;6u(B!tDFZKoqH zsdjR+D4UM+8=2~SZ6_N`Ft(N2j-h}FB#08NT(Ogm&P+fS<#v&#IwjL@e*~~BrmaY3 zYNwr+d$WhAH7+d|(fppT&p9VKxdg29?c@9VXYzRDanC*XoX_&!x6kJT%<8}cJg&US zGGbBbz^xsiU0 zz!UHFVDFtvK4&=goRQCx(<$<~6L@z4Z}H22Kqr~RQfCvxaN*qzo_By};Z<>O;rYj7 z%t^Al`&m1uy3VnCy*PfBIk-4>`!qDmm|R@XMV>2nDp{awjbW_*zx4tAqi+{!9`QiHK$MXy{@eJcR;lN|8vE1Vsi*)bSONrY-^R6B%2NuF_ zCwm#WW4N0!48wpis9*J!23_p>#y5@~`Oc0vGb(57*%8`sth;Obn?Lz8Ymdg{!q2w< zGGlY8S&(g^@3xE!Viy{}(ddee(hH8=^9HKWkoI z{9fyUt2;M$UK+c6x0~>z(^z}yj{kF_H%Fd~1zY(SsbeSuMlNjq4s2bxel!KH@pfP) z-hxL$dJnm1$8rO)B6NrRoU!=gdF*9(j9z`50Ke*6vb=-5tYW74llFOb&bgrxc%`o= z+4u+k71vj6LbP~S_t2vC6usc;NjruXi0)6f^V8~Y9JK?czX$CY^WJ-WE3(Ns=aXOVyNESm zmamQVnQb#puiv=U3O0D@px^L%hT-)l4-VqDA=9Dle&8_M=WiGQ7JIF61JkhUP2RMD z+5_gR^L-Q49%wDOiyP^r?@c<@*w1_hn49)c)cVjw1Ms|-d*JE>@5+BI^<^~t8a+5{ zHwaIuatpbXYEypMGgj9Y_R?$zaQD*eM#kgvluM%zjCW}CHd{7`Z$!7lw9MF_2KM>D zIL!X96$Xxh4}>FhaRx2#vH}h*XADSYc3uUoZU6?NS!yp_*p)I)twoZ&GJ9iefDRqn z?gw7({;aZ<9ueHmLf69aSho1&F~)W-xZRKfx9RZqjld{WIsLkx_s*otOQFkDTAji@ zJ3q(zKh@wjF4 zW1GgyciC7yKOD`A$6a2`hkiDnix<;kZ4*a$QS}Sr$HZoMF&}#k`ZN1KLSy2^lfe00 zG<7z*dXllG%I|a0)lSCbc5H?6Rv}9lh&86zs zVn-g$BY)?WM{Ch}8sk{Hb=nxXSvss+(~LYacFXn9(xZ|`oZIX8gf}o&)r3hNNk=6f zwP{9s6-h36<=nam!rI(+HU&ag^ zD1#pHIom3Tzf&LeSOgu`$~#-hb9Yh?5pmD|e)weLTI^aMKJHfdD7*NhX1^iDGHS_at^y79!9t>mN~?>08!bBfmz`0o!>Q*CMADg1cZo4v71x2V^IO+I?xZDx1!JB_r$a0wLXGx z{4DpeT^+w;Q(;mh#zX(7V=R+6xw%xZ96L=JPOiUnwPF6h9%q?q{ncMHE z3GNE9X1R*r4)7f9KV@0Z@6cSTmP z3b2{n1#EQPMZYFS2~0Yg=o`5mS3HZjl%FwfC6}&s8C^xHGxHCWIyR|tFN(J(us07c zG5Z-E(=&B`>st5y1Y>jCe2cd9yK+|KhBgi7ad`im(AZlukh@%a^Ei)kKjQG`Lg-m_ z=nhXE%?`RerI=f0tPUPMbI)HVE@|)`ioO5zyJl}Ca^Qja{w*&jEdqA-Z^i(4_sqjT};H3)aO}VUVo03Ksh?a)o+Zajy<7w!P z`iUX*y~Z*EuZe#fT3g0*4xWu2^(45zJ(7G#cp|2l zil@Z4ZJJX(imtN*c1}}V-DDO=(vqqCT!gV^GO`H^5mZl z4PC<7JyUnZ>0h>I#_y7itieErTo&zkg)#n`f8OhS-Mv36+HstFo$R@QT(cq{H~TSb zpBd#a6j#0HE6AD#=D5%@e3%Nen}M0?G+-nkE)ivRJ0>8kzh46`oHg^BZ*tGEx=DBZz z)_XcMAo!;#{^E~e>ps5jt#mW@vF3fhXWkvmyPJ8NI3V+O^E75I70l&n{zu|~X6~}T zemj9UAan0U?wQ!Y)8M#*@dJY%V9*N;&X%W1KL;$nBv>$qzebM>Zyn$b+54RE? zk1>9eKk@^y|Gm`lZ)5x=f86+AK@OY^mw%!0ACKhK9#6lf3%DvqA)l_B zI)Nr)_Bu1LD<2pEpRss!3wZ6yAA?7s70Kl-)WRgVPUX=!aiW`m)7>4(##Q{6I{PTO zcHZ9wjD0!E_WRp0?k(b;TdVs2APwvF(@(wIh=254ein7u_(rN}-TtL-9NRwQJ8v%j z(of%<{)yM#tg(UvS->%$TG(v+nU?LuDqkR0$!}F_(E));)Y!zW6>+XznK>-$i<}z1 z9C(Jc9R<{0{;T9C@H^|X$Td6uX1{y3=gxXR9eNxOPXNOP4-9uhFP7)eg=a4F^6uC0 z*Q|d>C)<0&*V#GnxRYgHs+&INh6p#Kl z>{1(leox(k>d`=rP4pAGyF_~ky}c9VH7ZQySD z)f^Rz+4T`(?6+Gj0b&A@`aVA=&w_~8e)Ce7Q%x!S)Gc!Uzk#>xDr%H}`# z&Xmn>a!-ByQne(&;TPu9ANzMc1!J#FzK=6LT%T@^!#A!-apl1-v-Wvpe+I3si-U`D z?bV5&^bO>$+AWx1J(q@0V%D!>L+>MYpPk}obKYcw`(7>YbyCmQaRst%D*oSA>dZ6n z#ir2yEdH;6#;3A>r_;|@oqqOVBM2E9RlzM2l#!54O zi*$>8$Ic6_q-s+2oOFw9i3DR1FMhKeKZAF)7EEgsxi_V(9)PZ5Fc!WIlZH8aYT;J{=){6GA@iY<1$<*<&nV!$RQ2j2xZar<88gAN>9vrUs=2_yai|Tn-v7jctKgjnA z##cPSioV1cbKqC;Y9V}TS^kC^?2KD@rW3nE>qZnC=p8Wr#%|+>Kr5n?Qpx8{Goxk1 zXgaXPB_HK;T)9{^rnZi7@n-Y9O%sYUX>Z5`@^xA;(AELn!QZSaQ9YAC7SF&|gyy=C ztKGV=*A+uJ$ogWyu^IpCW3AW7r(C2+GtG`3QZPfk$vR)SMd*Pj#;MTN>99Nr% zX;W)0@|`vh(`Hy_EnQ@{dDv<5Fl~;y@3h%Rn`sxM$4baeb%5VAYNMntXMrc}udMs~ z8SC(Ue%r@=*#_cO!8!vQ-K9}4%neWTOq6p;-0N?cYn_b|j5Tx(~%Z>JdJqK}(>ah8-Jzo4NG z^s%1rl>E9)YeQLM!Lu{;j_iKoJEA>5?-u@5UR3%17o&FFOg(x?@9O!Fd7r0E@ALO3 zla1S1Td7#VyO&sRD5k@^b7YUN%686Gp#6%|mqY`K$I+gNHPFt1cZR58_8B?ZgY@efN!Xy<2;JbQ13h*Uagt_Dm{7_6y#M!*p|RN9)zmRjVVLx0`b{ z0#!Y%M-nblpZSqFUn5}7*Vyiyuc2oj02guEv1n&a(J94-HdR=xOS>0(;C==7JMZJV zE)ea6RwQRSmHz>cs(I@A6uA8|>(|U&Cq{?iP!x!M`Y%p3PGh_`0-GJgVKeb3v(f7_ z!O1LQ9+y+|E!pb+uCoushvM&E*of$S*%Pw4Bx{uS@0?4&3w58p^x+f3YfgRd zLF%+*^9@nE@m+j_{q&RX^A8k|zwZUNJ>adC{u6gbl0Pmz*%R0P=#1G<&Q>*a70>nhwYa_7`{Lgjtsnb6;K{V56@2RbQ00b}W!QHA||8e+K zquYnt?USx&OtSw}lNp%gZ&(if7E;$H8m@qb;gzcu2Nj%tmW6EQIwwH9rn21BxM}SG zb(|-~XX35Td+1B~$ao2H&=eXgVmuY#V}wR0M*p63WxaDOWR4N$sQGE0UKp!RZ$E7m zTK?#S)plKT&&Q$bF`jkpK2u+Pq3Gwb*dJJ1c54r zQt-6}ne(HMSw~Adi;`aeU-(hUQfx^p{>i9)KQO8;-0lB_(}(dN>zwsL_-1d{rK}HX zL_Q<0-E*2G8(f`JcH6s+1;EG$4`;)-i8^=$xekvItJ5d3&P-pP-g!IV9~CS995Pa| zViSLEtA! z$qx8SeM(10#@X_WI1BcSeMX3E;Sg&ekNl$@XBkfoylN(2U=4xRqicP56?HqRp(G}q zxA$RuTksPi7P3?8LREjj_nqudbDVtud|b=+-U2VmXIKkwC1w$~k$#B>qq433oW5F5 z4gG5;Jl=Ud@?;ir3_f&a_OKi~fbV+4sm5~l8|u1(J(iJKb=q&0_Y%BwIkx=^;4Xe| zBzY$?P;07dENdVeTS04Ebg!KC&XIq zyqCRYHf!!8%`be5cs+E}ISc&0H`Le-jX%veyK^GRXBcZY_qFCt=NEnV4?~N(m}e97 zJR-Y~c`j$3`H#SJ@Fzq(MeWNb{sre5xYzAm3zuKG)Y&J|g>5=;A8#3*@D=DNo*vTfSP)%YJ?c%;a; z_!LWP8|*dkjy&_q%AfPTOOKLosq;_W47uZM>Gw0haOtcq#ccVKv?`G4E*4fc+bV^SV_0s&lqxdt?ZuE&g zN1e$oy!(`o-6UV&->5r^gGUql6P$B=!0G;GVlkUk@1*$F1%E)NoAvt{(NET$Z0zRu z5c}?Q6Auir=Z@yKgt;xjM@S&AWtSahj=C0~WdTdgul!I^avE~h$KGRRzK2{J)Xpu4 z&p!g3zw4og472~mi2QpK+Va9O=7D9!++uT1t!aC1(C|R3Sp$RJX!h&##TID3$i!N$ zZ?64J^c3qMJNYeuzY}0Rt=e||odhyZdiW8|g=@_Po0fm{FL0EsEPDI?_UU1N!;MG3 z*1p4P{nd`s_n18xW>n@>ZzB#+7qnKER9k~>$V%_HS9``?wy@Zo>0-vcFlfsVM?MQC zKVz-&{{|*5@9Wc*uMe5?BuB?51K5S)$;HI+YZu8(7?HgFhzJU+vntcQJ5;qG#*E9IOz)EX75g6BY@bt!w3EJ38 zUF%c3A7cNJ@PLJU>wuo(_?&l2_aYxUpo30mMSB`7VNCL`{lHE!;T~|I>mA(RODvN$ zh32dijccBAB-}&)JL3wOaXn+k6>8IYugCG_=iy(+7}G-j2Lj{U+D5p$9zw23UvK3vdf_we^rICln_O#r_lAx$6*qA=-dPeroR7 zzrSp*?U=yNo!?*JQ}WoAfv#Nn9JqDmyX4A72hYa;3OMKE0tfGYoNsqpqwLCY=p)&c zL$@Q3kr((8Eva<>Fl)yZKXU2*hB4?~v8u)Ju?RQtpe9GJUqO#s5m(j9uC0= zhF*(;$pqt3>=FB;`4stJTeb#FzEJf0aprio^H1#j_XvLrM()`6qQpr z#YgFzn)NzZd!R(CYg)zlvUbwK8{4Q;>K+zV}j|i37LA z^aCsgJ3dOD`z1d$Oa5fXEB<5x7JjUVd*KvN z#-nxiSMdCmA?D>9S2Y_O<}T*(8uN(x#x;D2dEC1EdwZ^6o&)4~7cT$gp1amWHkbH9 z1N+Ho-?e7V<^?`q!vfx)4IHij#8e`W(!IBkH3gAo z!Kv>@g~|Q{g|t!IqP|X4KeJ~6dE*5dr}{)sEXWwwa3VaeVE~*g1SXo_>vz7^Ja4eJ^E zlFA)>N~}_s2(DgiH!KYinoqWp+fRXl3tpq+9w4YDA9nf`dMh5uIXvk$;`95pK ztLW%@+O~83tDEN{LqgTl_7u_{zUB(~RBnIJ^~(>?_tMIpdv>6AN-W>NHr_`rPM$`M zv8^{g>eNj5V{>TBwW-!pOOhQ4MdxbY8?#62>gFr?-NwP{=3H}q9{AuqwBh@qSdO{Y zp02B#vpm-yH`l7)UEO?%xz_%3tDC2o>zd{ScHLxiP23Irb%D8-ehS6T=UShV-&46B zHAX8Gy^JxQp|_C5U02w9HtfV#^Xf0fP`j?EZTY1WM|Eu$*Ivmr=Ut7qS9<26m$6yp zyD0W0Ijz{cJhwPG0Y@2>;p59xnl`kR> z7WiZzgc?G8O7Hh!uUs*>roHmL2RA;*@1cm599L}Hdv{$GX^CN%Poa*6I-cY>&ZVA> z&8k>MnDeK7abMKby91AD``>Hqz>YDwNo*U%Cwgv;tdxCi@w0`|%R^pVy*9n_rBNq4cg$pT;NuNJ&_Py4v#v#QsjlQMzD z31nOP8*AF1u3T2V7T>3jd_+I^W<6h9Th@zDDDGc`y(8NC2h|3d+Lci3+gy){OJFA$ z^#DKVn@^90Q|4^!Bj)l5JmTUrgIb^WC!Sd!(bYgL4|ZsPIB*9zRo-fg>XvJ#M-z@- zU*;!|z_n@y>hKBb@F}p7cgz0X20ng3eTMub#X0felC{`mrZ)6?;0bKI8fsg#|DST| z(rr1=RvOQWc9g#n{ivQtwLJ0_bE}`;lk2nQWFkw;v2|aZ5N;R`O&qGcqMEv|7RjHt znOg=h9fGbp@I5lXPcC{WAJ}UjluTb>;4m^NGs7BG{gU$1lEIzOO5ZH#VD#aA?=_lOE^)68VqKlkmFAa5%*N0X z_e~6p`?86|(|UfkTc1UIQ2WX#&sYdO$@ZVY+JLyk>@e)dkrNjfnv#skh`p~G8Y9C(v3I$4Wl8vKwC2c- zZ?VpIEV*$PFqX`S11H7wH79)^rZ;qkH$JRbL_TBhXHMnJ$!55wR zRDHmF#-C+A-gx8x4fq#0I+IHjO z-|@_)ANmo!4;BTC|Iwc<|06WY|IqpwtpnAXOUa}h)|R=lJ%9|B{ECcIT|!2*j5S{I zXNkH00NM6KbVejA`Ubye-oN&lOn+d2_~4?WY&OSl08sysw&a)npN?dQLHoFCxSBtmIxVzv;RMecFURQOsm7enYu@ z65j2rUUE`8qOW?!$zIwhCbs(afzf`^%;7PGXeZO0u@FZF+d4mFWb`3ym_FW1Ur9T( z+fTbY81u}KVkyCahZ&RlJ%Bv^KI2jh@M-pPMP``2%~Ttj1zwNiPnGd3{?6bQK1C*eiP2cAF{4c6i7%3fNG4I9h~w`dJg*EaN0;J)Vl!TPq2Ch&&r z%rblUToGMkp{JX81{~FegXrVnC_R0I{YZ_T4%+dz+F&r*xs6!Q>4MF&mC~6*Eo12+ zmhcni*~4$0z*+6>wZen@kvTi@y$|AhAH+v)Jw12x&R|{JA>z36;0OG>SSHV{!#z}Y?@;yPlYPwjaP{<)z09$fIrcI~J)7YRwr!jgh~;DJ4NQqN=YG23iQdbPJRzK^ z?F{BPkad51?((WVPVWar4drr^atoTMMz0JmN zei;0JH|wWQG||_0uq%752@ONm_<>k*Xi)(E(pnGsf8$iYW$eZAv2?C?L3iUB$6d71 z%NZ3z;9=a!p+!Z*@7&k?9&uF?^oP+;roMV4{^#yD+%ZdfgZd8HyV%3>>jSIjB?Fv06u5^^)@)s}cpmF` z*#k%QLHH#;?h^xePR||Cc@}@g_eJxN1O9{v3B(!2Sfu)JN(pF`j%kq z0r4kuRczD8Jh8n7!pI8s6Re#_Eu+6J#CYoY3^FF2m0V0sZymMgE=_C)cb{i`IgC&J zi6#P!FGN0HFbtTn+3W8N9-Xm?M%+5w)ccbSZwF!@;WOhLy#2)aXY%$M4?WoS+Xy`v zyfdad$sa9#tB{;O=LEo4q7xtE)P2>ps0Od%^XAOtYXu*_Xu+As=&{;sZ${m7=<{A@ zY`<~FGgxG=pUIsq8)_t<7r)h!F}D(HR$b6_#Js@Mts(32QeqRP?IJs;6+}*~pA>m) zXD}F5y!Sis^oG?pwx=UEHcY>jTijt9fN+q;~c+_FP@iN<*Ut`z*ry7 zh&*=p%GoE!alK$oQTsUh9!K8`&a++~C;7&BMVBWSui`cD(trNV=({ZR)y3BC+_h7q zvsad#EV;!RR18ADyY1J|uHwN>JTD)#@KWzPi_JTVq5UWC1d(H^f%P%A>*4)!W3y5h zj=xYweTYRpXBTmR?tI{4g&NTDQDRbtrzY@z0^DRV?@P`x?=JM5-p^z{3y=@Wo5y)C zQ4~pD3NC%)tmiY)dR=pppWTm-uJOqR-^x46 zeZ`SSLF_EO`xJHh-gjTXHZVDJ>cA#3=FDPiFd;jR_w-zX`yKqNE={?G%z2#Y0k04L zPJW>r`$)fcU5mYq?n@vym+|aQ@TW2LW$k+coAu?eFKr_^z1d;*d^$LN-xE8*>rU`` zkbK}y@R~+UOyke@r45k4UsU3O;gREKr!7No=&v&E0bd&$MR%iW@*IB9I+A=~(MLS5 zi#e_bjz0p9abT{vM%^s-X~7QfJ3Z9Mx@dE5y`I0E*lt|!TATw949vUexd$9PC>~bH zzG(g6GvLc0PskqC1NH5)~>YE<`(Eib>Kyu9Z}4?^;WTwDYpRkVshEGOm^*P zn}=Om+Hejy{5$n2>lpif=d+bh;VFUclI}F?V(3>e&&Th`cuO`{C`N4g@mAN~aA33r zy1n}xeYv#vLUoq>k(U)SR9(~~?3bJgtYpuvWh{*v1N5lt$353kb3HCr22Xv%xv%R6 zT{A|-FT2pY;tj18>V?nxfM15jejdCYA6mpl_m6K$8MorsYGdi; z06ku`Xr8G7kj~F71i$F~TK0t0{W4%yuqd1?<~jVTfl~1u`c1qWxr6t(R~g|QY=|3q zuK-&CJ0DviXwLGQ%X`HA8+Z>p!8ahjz0+eSB;dU}Wh-JQ2tN^Ya_tKR%eqU$+_(3}$}D#7Q>#!{KZ*M_&5BiLE^_X}J9Up|zYJcQb-q&O z;^5$mY6t&oNA|j7bylb?PR)npYUWp+-}+Xy@3^hr{I-QU2jx97(QhYsPva@dsy!BG zO`NF}LZ>~P1n!)ps5#``c_7P|Fv6rN8;Br2ffXgy)S-^A1n>Oq8x{c+$yA*s{ z=#dh5Q@C8^!KHFY>Z=@FE(VWf4j$d{W;o+51CO$UOTiy@$)-~56@8XrpOiXt6z;cx z`_tsaN4Af~ z;-_8jJ2A$(m4DS5oA)1_mt31=Jy*&;FZ;A6Ym7Eu{(v?!IcrLE7l6+7{t4i9v!Msh zy+99ASMzcw`f2g$g7z+K2I{9D+ks!%jlLDVxpeWV51@-@9e8;C4e#94cN#M4c(BOC zTLumyr=0*~4HqHcBkS*meYq+EhCwpMMF@$#$uu zovEVb&ceLK)Mf{vMfny%=&}aeWCAv$Y@Oi4;bdSFzFLtj`vRM=`5&H_4Dp=yf82|l zYY^Y^KI0jma^MPav)B0emvf zl`wTYbE)Gwjz884J$2{UHnN>};Cu^W7ZK?kY=}fA`d0Z9Y?&^6B<1I%zq*05&gc#y z^Etz_EeBsG@R4=L0`%zvj(*@MA0qa9g(s!LQHPqJpJN7i&gPojki zWQ`A5#<=@tRO< zr@0Qrw)3fV<4^Eud>Yo&l@m9)1Rkn|hs0ajFHZ5TvH0gO<2o1r*nZF&r$(+Czvjjf zzh<)S*KEHu(ynolho7R{>3aAjlQ`@{j6?VTKy8s33-*-quV>@s9Um~3_3*O%@IzUx zPiT!+uNB(Z8%UqMV|wco>)~7VyTh{D*2AY2(4Y2_{k(xwAXdny*PiKN9f|CjqwMKn zdH&pdP%d>Y-u$TQ{WEDdlYV^U-qJ2!cg&9+p<4Ykp-J+&L8PnVzceP4#+T<%ye>x2Wrb! zzEEGbvYvQncOWOWgHPh>Rcnz4+a_4g?}vBx!4u-2B5GS&;UC%e2Z#wBKwj)aPZxs+ z_7`Xi!9Rz1_AvZY0WV|kv}OA2^ZI`6;UD=tzlML7Iq?~N{|S8bg2P9OiEW3EE@kYB zf4l%6DgJQ){t;gFZUr_)4mO3xrZMHvm*IEhu*R(M)=_hi$>-&KzTk}2Yg6p!-JdYl zTJTZ=4ex``z2kJpc!06(V{E@>-Ii-lyhHw7dVin8s-`hF*5seFRw7^WVV#I{w?NaT}{%TisrjZAGtKUUl*c-}r_ma9N9NSM9?D z&MVix2#Fl|W2$#ak4@qn!vwOt1G%o6@6RHKmWDX9hfg=Y7;tzon|m(*Jw%M*{fQ&z z=NX>?yhs+y56gi+(;p`82M#m9?Q~ySTdt-0)}X0x_50F@d1nl41b;)w{^fkK=7d~P zMod14;bDh&55v3jiRb3-yDB={;oXHkBTF3KO*ee|AK!y-k#FMLUg+Zxe0zv7aNTwY z92{nBVV`C8{XC4m5>E^7qj(xSO7JoFQ+Rso?KV#vT7jqaX?U7G4NnV>*h3nFc-p0t zEzrs5z~z`ceZZ!L1Heu^Aeyg1AD0@qpbOD?C43LS+v1r_t}Cg>I1CTFeBsB|*8A%$ ze}m-{Plb*C^YVrCpW#9Hx7EWNuFv%6?)cyh`Ani~`Aqxa5kuFOhpwG5XL|Vh0DHt} zd;!zHy*5|lK0rJ|W7htw-Z9HR*w2`4K0b52@(naz`37V0vfts255vpvkmvp&Uhd-= z$%r#}*_pFyHJ+t!eeQ5RzwLZ(<@2Z3_>B+4gAc=lw=BQ=0P#D5qw$!Op4&sB5a2scFK*|2=!S%d15biziNjotROxw zb+g>BAYZ0)ut&v`oc7%L|B^M>n*aXP`I~yG(e-hPv*rK;*<~U6mv5l_srHST&24Sn_CLm9aRJS4Z$fkhwi%eWx& zSZ~&mC;DfRPvzXLAao_XE0!c%t%f#^oMPP`_`we`wvV5F3Jp$uu?RTLW}PkT|5_Al ztpk>ww5yu1R6NU;={Sq*D?b;Wq16s#pYUS)b~cXFs8I{ipU#qYeW8bc<GC;aBavj<Rqt& zv~S>ct}`9}Ng(eOOPlM^kk%^R?4o|*{fQqPJ|irCJMb%kSMFrK z*qk}gLNR))9-CnO>EnlIVskzXZki39`C{WPhV~z};f@P_9Q#izdnxC&#u=1TKob^*1LQOR+`c zw=1TPy~}4u3G{PkB>4`w+wRxD^*lNM-BzvkiTSEoZ}vBwOWet|p4paWCAW>^OigF3 z+ZgLs#wl63jrUbYNDXw-+qT->kw$EG9Q$Ixr`|gE=YhLwOokR(#6%74n)Ex`#?(HhA{*65Vj=V8_kzpG&oY1E z8=r1>wtC`M#5u>{N9DV7k;~rtMd5fqYt9O@Zip6guI_%;-?Z}G;(LK& zq@{DB%dY}vIym2g^lLm76ochUi$KI~EzU`k+ zOr%lbvnha%}Jr#T}=5tv%IW=PE#+PPBc1^9duPbM9-D+Ro^{L3NciDqQ`SYcj)~@Di zzIPtK&0}8kl7sxNxpdM`1#|h>m-aoA_a(cQ?7QSfI)8xuw<(` zDhDp++JTF%7tagRXPCaj$-P;T!3EHU%cqIJ3M1PtW9_A`)qWYzmhr5{==!mDF)qnA z%|U)_DQ%WBmlE2;)?L)ayNh`*@e=+H_r43wXJtiRmY=;1o4vbTc?@6Mj;shZb0Onr z_q5@=;$y2{cx`ew-%Dv_t5I~9$L+Tu4X_-)1kj6=-@7!RyWyt*rwHc*~i<^D(!n{^{7j$ zd!2cDX;u5mO>^kAf;KzA{WSWSrm>Un%ZGlZG1sRo>v>&^macdFt!dEGG-#<*XAxw% zH4nB8rE%yy4lfOfW^{gpOEarHG-B+f1JmsDM)3jPJ}`Y$J?k6Ws7I)n?#RWUk&B{X z<-NLC+n3EvTgE;iyHtU}IMAF=ad)5+0~`kg$bevb#Lo6Y;=@kjk$YX5GrHe$^7 zYwh1_@#`&e?|RoiT{XS_*d^>CW^~1Y>874YbN(@NKJ1yRcw(9rSxH^)$`UJ3h5laT z57xG1^G~f*ReVa*6XHe9PoH{zn17HT-u)J|Kg(`n>07{+h5oheURzH!uy=?He~sfw z2llTqmJ6`C-mah)#DBp+HZ>n_=U7oTW>3EEjMMF}pO~9hUz+`3sO^ellgj2)Zg(1V z6U1JfK^}8`<&tXVG%$^Es*a$Ld+W@afH7*(ftTp-XV9hX`vqbza{UF*xV<{Q%0oNS z@vpe^0REas8EZM+c_`N)xVeAse8?G{F`r2x>-k#zS$97BJ@a{h`D~crBZe3->ji2j z_#1rSN%N{>UVU8GRnDkRuuk9s)(ub#*5b7l?3%4W>~YT=qg+ov%N+kaWsb*E=C}_Z z&YfesXO8atc6|u?2zbUI`nd4FLmy+n^B&+C_rMcd($L3x2abincN=h94=i`^xfdS5 z4_cHBOa<4qtY2_xCc*jz>A(c*7DOXs!S=Hr*nWoVpH+;|y_bG|t#!;E8&Ugi8a{C8 z#{=7=9@vV8T=@R~ML$8$9DQ6rewH~tl`_W(2PPT-fLM4Db4=RdeauJ%3vEy5cCQ=Svw+ z5cz`)O-fE?>pU6OP-tyMG-V9)Q^xSO&KR_Yz~J8*19fe~W9T~z{xh)YyytJXYM%}( zKRWL#=yk>+-_Tb*ZBGI5nSAt&;?~&8X8)9l{N~yp$ox6KW}i^m9Zyn2S$4a<2ZgRn zZy&t}g<@YdS=QhJ#>actWUnTV|Ky{O1%LSCH?ytai&{^id%x#XFwSOvW9^S3c(TrJ zl>9T}nNpUwmAWilyZH<=j*{Y}&KuAeBhvA;bE5YBVDcj75@!q+V;~;?xI3?G{DfWJ zdEGfOuQJ-p#-=iD+!?gnY#nLS%iEgA?*gY9MZaGD%O&6Ag@4B^D|wJw>w~4Aj2^^) z>!E+Ozx38(lV4DLL*KuCTXFKn0RExz+c{SpKXMB(FVp7%TVI;@z=QG$oG+YVAcH! z4lgjzdgfUV9O{9C=AOW2*YBw?_(#D2dslX^;NZ@`7<|ZP5`W026d%{7;3U=0zf0ry z$namJ(uc8EM%HsBz>8>eF*MFN*)JQMroiWU=r?`^HDUZ_=mFnVb{D=N_Vs0|tz;j? zQ1nOapQ*X0zAv~Ij{(&Ac;M1Vo=g8Ttx3_j7f14Xna{z>Wm^0H;!G9y-v&*W5&w#w)+H zPS&f>M4#e^CCEwD6^!shDE4i4ZLvd-W6|AC)^Mltv-D>_JpLLyK7gHE!+r~G;N_Xz zeb40Ha`=hdTV8yks=#_#IIXfSz9zOla?;$hCS4P|F|sFj{)?AXS>bD9Ga}WwH|pBD zCgzLe%AAPd$tC#CA$(`~%<`Y7 za882!W2$a(ff zRV@9Fjt|4@>CqeR866+aGVAEFqU+u~Gd}z_aM7F;1CJwPKSX@k%t5vWa5Vk8@!>bt zj2$1o!-1cPg%igqaN@(rz}?xzep6u+8v`~!lRVWp(Fq;Uw_^3GJM6xec;5BoGV@tq zQN(@`b2%fm$e)LNw#SL?eid36fEFI7-7BF5oxhb+Twr1{#Qff{YNtdMdu(O=tl3Mx z3LU(|GpTXzna(^q7NTSLy_|VeMSRJ-i+#zsIa_&x<4f52kq;B2{@KzqW7J=Pm%K4* za~_mvcswvbHcDRp$TwPdUFFCI@zAioHSw{FM{-5L$jwm&p}8_-z4Ce;)Ia327Vw1X ztylT-Om5&x;$osvPoB1u_*sqCcv9o1{Vxt#{%D*U>;inWFuB9b_k1h*8=Iauv8o_1 z*EcQ}esAb%t$*pS@#P_xHpwn+t&X%*(y!|Lx>!@#4L@l=mnT^VqBdKrt(JIAL0&ET z)u!Vd@-CcFL*G9iBsX+DJj=CgFRhj6=Ktf&Mfx6HXMVfkZJih9*%L8f<3-Pm$V4Y^ z==^p)IkWPL0_7)H8PHX+y#;^S0ea)h0V>>_- zqJ`@mT3F}MfRUd8dp-9eJxR;y@cuxcWo*K4I;e8() zRl`=~*V%jiXvdS}(apLQ;WEG+(qfZ#68B|o-)m1tlHcDPulMnNQD^!!>!E!UdO% zn$LSaYqD@(<qRR%B{4Om0~9h#5ZiK)2_<`-^WC66-&2K=l{e zmtS_cV87VZB-!yb?IF4PFNfcOo(pyZ%>Py9 z*#QhXQ(&O}M~$n7aiIss8rSE}HLfebVa=azTs6?wQf!Y0px1KftD}+q7)4{8)4PvY zf`N1ErO|W;&V9rd{!TKNXUip9c~7}s`?;yn5YJIR(Zb&2Nn4)zV+WJ$)eD{@*dEer zZeAiztw1*V>=taBPap#duuDD+tqve(Ugh6Qvva9!?`4fhFLviMmCH^(ZKXG4`hsR} z0@gJ}!{k*H%uzYhEcP8wG%{!8mVBlT=AjrtDdSlNOlDEzmf#+>ukhKWhR=GP+Sh?G z>jizLUXc1-Q!m)FnfW}*KYAfVUmf(-vl*VFzxboT4md6g@qGgPiM$!vGa?v^Tlv5E zvE-oH<8)l~Z-`4Cp(ZAmoN+EW;}ez3PxiC+y@Y+0EbM31M~!1{13agG#8;|+()!++ z^wGwr(UHg)`wUK7&b_kzRO4D|O^l3N8EN_XXCv*;`U+M`E)V(6uPVbv?u4HIp7;M$ za>eC^e1}In?gJk=HZN49)D3RFWrRn|93EBOpm@~y&iC28@TZVdFIpKJ7XaS{z<2!e zyH1W{zq%S~K(ie9;)6Y}J)=}d76s1I(}Hmf*xUq+pE(PR?+3={#^;}OY=<6jBm1Fa z6fPpo4|A4`{hQTt$SPPV+d_MgxpuH@g}t1+sxwpd+&*~l5IlH@^_+dsDRrKka`#O4iKs`Fe5oX&vv2Ihe6sr#w$zV=-3z6xI1@TY^f_^$I>WGXV#*174itAJG*I^Yag zH9)t+&qkUHfR(Esd-$#AHekZvOcDF}XF+iMb=~9927W?RMhOJ@&VvJ?i_-ng)lj)6uQsYuncOcr*ju>gu1Bz+}Jt7Gy*x zxnA1`VZVWGry8|JT>G#E7;msP(vpcR*MF08O7tCIe8Kuii@vW_jWPQP1|kdca&*SB z4WCW&^Vb3&ALD$Jng;3VOkgt=yEH-DS2)j=^W0NB_ZrX0H=WM&s`$y0f+tkZuIpq{#0jYpAA0hKJRGnmyUqzliK(OW}C`@qdpDnc+UHtU|M#k1KW~< zq|X_rYjYf6oWgg2apW*gzkJ^_+ZbM}A{$b6nxmgKrOL~Jg~KvqlcS%^T9sk_Bzp~b zl{w>Z>khU$<5=v#t9*=cjBR6t2gKK|jiGpg>ZxBVhZkqraF%_s5`LC#q4OfpJ+b-F zxoBWi?D#^PM}Ngx;@ z2RM=vKUcnLI{(n)KmoC8@tokTT-G%H?RwK-Y!k4K=&TX&H7$bgkOQytxsF(DA?N9q z<7-bVj^r)ozmOb|?(J7@h;f$luQf6W>Ih_aiROlLr*4h_dQfaaa_gHOn55=g`cv|6 zL)Vg9U=5U=)55lEx1!H~7kIM!joh3L=4hb=!x?=M7bc~CXAz%ozzw8`?XDB|qLrf^_I5pt}lHn3tpaAyyn z!8TxAacNCE@XcFq`OW!^GcTsr%}S3RsH~`_etbm-+Ra31^CYnT>Rj%Oz=}k zo7;Gn+VW^Ac^I2_YwW$m>KqyXw)VW7J;u(s@7V0l{ifEy3O{m5`y!1;9(~Z*rZEq9 zY|M3blXc19H1;&ESY5+DN0B_vc{OKBYA#KlxoF?I)cH)~oE`6cw7115dj7zW#$A0g*ev51U zKIpmkx^wSUbI+#bey-1upU8;@&Up9A&ME)HukER_WGveJAAUWPzyF6{ZT%sBJkP_A z<2={tT}k!mA(i*f;ahJ>LAYY!BHWZoWBzogo{M zI>dqVi2qjb-+}Gqo=>d3%z{=BJtuxh~4pg2X<5;zFhYd>LY3=N0mzsuog<= zs$qXFa!`w=Pgv9L{=Thp@k!Z@?sv1V40^a`jy2fIvo)-LuT?xo?V^{p&u`q7k**Dc zzB2l2_`5%LYz$a@!2^r`*O6JSeUY1DpSkuJxwe%<>}5V&o3Po6UXMM4jeLz@l!<-S zCVOWy=bo`wU@2oo->py#DZw~3H|&%Zvcb-L2Z_$wXbSue-_Gt_5ns<)=1CvvW<3;G*~+2=*T55C^ux2YNN>V476WcpP- zRBC_ou^GJdVrWHnICL|-A186g66~cfvc_#bYsaUfn{+1qUhJ}Iz^ESki@KRjI~dPS zXlFaQmxI`2S{ttYFLd^KG5axSow@dD$TGF>!B`b_-zDrfbPKkG=)i5G3mdZs`=%G4 zRJ1C5bl?xvLvPv(PPBFqTI+)r!)qr;gUgp1T64em(C-oU5g8YZti)y-oX$QV^L@b> zwJs|+vhOc>u#NdqYsD65BLf@r1i7}7kJ)D-g|Fw~o zxxV1WT;hql@DIBxB6;NM^0u=tPYW<*Ppp;G`Mv&zNc(pEe(%Wdif!mQJIalN8cpL*@hmod-dMkB-f`fR`i_0iu@i=QMQx|v;|z-Qn&y7hb(p^W zzg)t!9c~@bS3R^N8gl8i-<&rvaz3hj*3!+zCLU1MSghKIO{JWJL7c+u z1=9+AE~3wD{MJ>}SC?|`MHc*_>oU#_;r&gePn!2Pl{FRH&v2f`k64qT+O;fl$Qj=E z4~@KEdN0@TiD~PQKN-2R_L%A8KEK`0Dt|KTV%7Z2ikkcPn*NTK9=}Qb9xdy%Kg)l8 zll^?1{a)FI(f#*RlOey`v{B>ffAdKH7qy6LeIFnnnGNybXFYS$wDJCZRRo(si zpwFCfwiUT0eG_^?dMPxZoOgaLXNX|)ho~u@NzHH?XTR@2rckrfu!FodvS;u?o;%L- zec1V0v+AEvkW3F3u;wEe-Dah=b)#4M11mYVLHiBZ=V^@#7O*E@7n8XN?J8zdp|hW8Yd`o{1P*nr{mC9+Op4d(46?Y(v%@hZ`^>A!(7e6( z|H@_`*CtjNSk?S{+7XW2Sj7_#F1DP7U*{Ws9XU@z_~;=&sraPgl$}MyDd#f3g}|iF zpVzm+pV!~u&pYv`KW|{OKkwB>f8J}i`c`Vr@dLi-P3-$?g{`ZVSA89LyncH@-Y3cL zcJaLnICV8eR!Ub8pS)V7fYvc|Hjo1;fsd`s0UiqbFp>wpmkC75WaZtWY#;Q zOB5SQAXh~fhp|FXsJvYh(|TSzL=o6f?BprI=zc2H%X@bECJeFi$y; zz72)UuaNl_GQUFRSIGRxSL8>41#_57et97_6YceoKfa4Sa^%z5`gJ7kq2J$SFQ3%k zRKHco7$4`efX|H{d=1x*pdY*7L-&l3-heN!FWr~dKi-#jVxli^;6h*Ct5bY=uRUt5 z6i;V6ICf>Tbf;NsG?g);FM9(8ykEfk1-xIt`vtsT!26Mv_d9ekEDy%u&;Jv8*w4P6 zwa6&NOazmE1twnFcs=;xX`=%?e-Ldvh%WciMjvsZKbAJ~pp8k;IN?*TMM| z2j}Pg32?rhb;VwN{9$mu9-Ir`ZvEbJ_E`0c!G7{kvV^=_R{*GSheU%?j^IvkHWwebq-dY1X}I;wnmonQ9lb=~XB>!#MHXRozVdnrh^o6j6R z;Roh+wfT;|==lTZh+7~3N{}zvkE_`*R;bf%*AVpENJDf+1w`E1vNrN@p#N9tQ`^}<`~Q+?=N>8Kv$ZO;`2 zdA-*7h786J?ylB6l1NTc(ZFWsUxx zfjzmZ`F}wZw zDsxuWeAPx7UAwCJCUZSL_IW;UGuP60E1U1&I@8f>UO6fGm(Y2s>Cgr=t#xAJg?VSu zcgs`uKbK!4{jGX_)%ABnhrI`ow}+6oZOGe~khkr~+gFgczq3{bfLRTAeg-)y-w5BB zb-XK^ci_)(o<&>5=?xou4;ADo7OTB*b#{C)Yv($^i|~B_-{v@<^Z1;HO``qz4zfSr zT4-Vjoz}5?J-7~-bK_)_FBOiF;TM1t`8;Z)$k;jTdxxFVl_R{TkJ$!t*Z35Fhk{Gua2wZ7&Ey2xAK9R!AsFFZ&M z<)XN3p5qq&VUV*>RyOaqLhQBZZ`;9Mi`t9Q-s1qj7TA?MeX6kw9IvO19pGU-cvug9 z_5!b`fMZkjjFW38gc_d0{_Q*-$*ZfJc3l>Jel5>fvV*{(!I6`HvCm9h)x5&MPG>@_ zY<|ECy9@2QeskWi_~*jfmKRg(frnD&m(N(rnb%b6{HHUH5Wc<6c#-Uzr~DT4RgHW1 zUMsI_>Ii>qmHo~BuRp=xx94rI1JS&*HE;N&o4N00fA5~D$g{nyPpO=F9cPF%)H27> z^Ri`v&L45@-s`;c8V`+6wdeLL)<_S_hN)xBkGXi`9cF&aL-Pw%mz`9sqO*>?(*|;H z*q>|hHS(Ejn7B+0y!aq^ab>_$$N=p#sr@G_m}f0AVCN2JF9{Fd%wVh?;Nn8h_`@mV zmt6a{XY6NN-?a;Ug09PBzj<(V=*fep z7jFg^ZQbPebRNS&=&I)`tL15)d+_21jy+hm=gsh%^e4mbE#92|-i9{Lylx3+tz!*E zIy4)OJ+~8_Rn)e0U&UEk;ZR!w{3*wE0D3`}n>~fraYlz=F_X_7r)xIjuRYcaj^ogM z_qDYx&#-o{|5LRsGrk#~dij;>js-YBP=3Tet8Lm1G4>`aR8NVteczCwk>%l?Y>AE(2&KB}uZS|b@HJ!6_ z*!QMo3C}(N9v%P>J6ETT;49N=Ndpi2nCEfqkd9fkE!%)&S7vR?x(5SObAV%+EL_@5hj$&BFu|QKGC}Ko-1#jx@xhV(9JH51 z5cwn7D_cmqL^jeA!gZECxTcdZ-wV~>AqhqvwNpXb)vSlK%Xz08pfezH(@6@mvcm#k?5p%4$*;=t5 zdJ#+l`o73oQOI}dT-quoSPd2I&s{&i_E^z`;MDc)io0)lvmE(s_K`;h?FBx%mX2Kl z-e!O|?c?wX<|&w}kNU5!KUQM}tM(%^f>*C2t`}@;y4zaOK|hPZq5Ix;@ZGbyzD+oP z0hv}$-%a48fai)VtD$Sox?|-N3nEPboy#BEbCXZKqB7zw`nkw+S@+DSt%r&Uio zSp>}Nbs*TvQzFSm;1_UpMUm~Z#Dqq6n8T1?XPm^nK zpL2hgJ*ww-fM4;cA6nMjWebW=B_CCnV6VXq#(u}xyf_-BLE=WL7qMf|tD0|wzNiZ& zE_<2Q)o#j;grj$`C#q}DeAdvS&epYJpE1{VUhG;Sk?SV zu3g#{y&hv7itK|s7^~&^8bXIz1?6)r_8}XCPel!R0lgp`X$qJZQ z@a_s=o`Xyk{7aVwlS4)RlZpq)mUnSNJh7!b-1@(OV@riT(X$$#{@uR5_%d}yN`PCOl*M*kt^NgvG&3TwbY+iHw1ce@pD+Z~2eoKO{C-c^wU` zLw*!naI=-y-Du?zOIoSiq4ElG@+OI85qx#(oxY5XMV_SlK6sP-$r*Y{HXZNC9&_z{ zjknyvpV#+u?Rf3?;G*>ZWIi+gFV5$aA96m+&HAyC_0L1VCmp_DHo=-RloeT#gI*ef zCK89pA+_=Ul9iW!!@Tz1SFAj(2b+id*}$66UgTk~?6Y<&uUGar?ZmNr6WG1-7dsyX zKbyf%BlzL`WiS1N>2JQ`w!|t_$8&;ni5B=Siq2bJc^w!?CZ4wP#&>`Hv8s<3w%_f$ zzxmZ4GL8~!<#+yV{jniyLRB3$Q~E`U?^+Fg$PJ6US|)n{?}ztpUS4|LBfR(8vsT{2 zp~z!Zz6r4^+8CgXFwedA1@(JgRR!~W^#@ko1T&5aRme5YwI`3j*wofE`j|^wz2CO- z6x*YI@UdRn7-Af~=)?~*-w*q|_&|K$fxN=@AYP@sK_V zwyu3#OCH28v{owiA-doG7QbH@;d>;x#>216Q|wRbsggnB4thYs)!Zw{a_a zM!wv^ly`0aEWr6fnvY~Kx{hlzAK9DEeD2YFki(jf)^)i2`l06Y%GmQ+Kj{PJ!+U3& z&u1x>&M@@YN9 zc@HfrNcN(44+U?^8>+Te%tO}7pH{9?_xi}|u(xnSKXUz0FfA{CrL|%?usY7QY|CEc z$sy#uc?bWi_Ybzr*RzMR?%Z|c^3U(;%}L+Ye?@rLp+4GOXeEC|?0s$ko8x+T6I;-Y zy>~67K7)D36>~2l=02C0`$A&ww-R%|jhOrG#N3}H=6(h5GWN7ZzT_qU9#Kxg?4h?4 zxJ&^ip8+O5;FFCkxCEFK&{r018hq9{v1i_Q_XhL>FaMS>`{G+k3pmXNK8kA%*C$^a zNw%N!HGH=%U6$6QV!6 ze$+lQ-O$GYdq0GS@NwUtc+v2h?m%Th(!aRC_$%^}ZQmp?5Mz8dRhCg3^YzD$Q#%vU zJ{Tp)cjl(tYw&K?ja;udJ^2&z2m8>E@e*(V9&|oUKK9=@{EY-OGgK4aHB^(aYv?v& z{zW_k?mBP9R^q$zC&Hmw&&;@$KQMUvV{;|Yobm{9{DMy4D7!Nazro~9puHY)iEiFR zziE!rZA(w3Uvn6lpty6fh0o1@H{;({=}T_En|wI+KLK=y&K^;`9NO5ee3NKn6LkvW zmz}^(x$td^eZY?#A~MS6b(=OEn%Z5)d&rtO>v->+GE^fII&!lc$jvs8n|+iV(Pnb9jpSxuimZ&^2Oen;dpP;6lS7TK z^W5*a-shXxV184ZYLB)=3mcKm3FyY#x6Xj>px;jV?4rN3(!w@)_-T0f+wkzS@bC}d z;UB`o$KhdcPHfPubLsgMpH+6=>~-$DIsW6mkALNLCvWyTV?O_6`ZXFO`JP67`rz3F zGS2qbGh^CMYHDP>IpbA$7yh8{IX>Wd0XWPC_pV&X1o!af?(Ej9qdsKBdFgKMK4APt z#hf~Qfd-wibw07E66^edyMTq_Ay2`J=My`6eU>HJJa`!W`#SRAC10TJjljgVdroiK zIG9tA_tva}ysP{dHr#tzr2WsVNewlAV!Y_$eE)?5mB`u2#P9F*U(og+z6;vk30&0n z>G!|6G3vj7=Ozw(x%ts2wEu4_`w?~&S;^nQck#d^4 z#_+paBgvPcHPK)u@tFUQymx_*s=V|6&&=d9Nw|bUs}*Vzf?QR!A`o@WBtcQE+xBO= zY>V5FgiA15mfBsVY7+=35N#QCn=W=YTm($4l~S#xwhKtbOIqc!?e?}mW->`8AzY;D zmZ_Kg-k;~3lbjg>y4~)7_x1Yc^~!6`oH^%tzR&Y~@89q9ee&@43&u6Xb-Zi%r{a%` zJ6Rsw*#quG!I@&q>V>|_W8jYHa-3)r7$1i=w;0;w9465wIOfvk0@mnp)~JPi1JUN= ztd;gdi8jUKl|La~Yx7xqZEj;NMQgf0m+J^~5)F)OU108yYM&zk#oZ0C#S+V(k> zajm&=b6MhV*!1bW8M(PuLCLL_XO(I?tRp-$h=Zo!>dMHS_T~MSaRO zoM@E?k*nR%O?O{tc@Hu&3*I&dy}Os(uv+Yx9`L!Ff9=Oso{FB)@5R3C-m!mgmAlU^ zS+Zp2B64lMd?{pl0Q3ZJM#4@Y=c!iZX1|9W$19-u+VbIlyX6T#wU26Fa zPJafR&h-y#9KrZUMnZ`y2Sddrq7{y3udi7yey_Gh$}TN!O{1x67+e3Y}l zMm83M&tt*oBNq4!4cT~{(K<^wyqh(y27l=*4qxd8e|x@x&&e}a$H0T*72tPsC~+Zk z<^P@CZ|pX^-y8TG`KwSueZLhLVftO<&q(UmLN9dt9seivyW8hKbq8{18Z{F#Slcf6 zLf6UA^1r;I;n03-SlfPV!V{5|hmKgoYUQ)afR_(N)(!)oU%QCf`yRej1J17j=Y#wf zC+A`!crSjWv%At6Ul+J=!r;}=R@M^xZuw4~SJwkxA7gL%arC(3jW#aoY_-hq%qfaf zLn8w{W*GY45zc@(hQ25~cjX^G!D$)jL~Zn^e2ek$zWwNo8Sp;6GmLlh(5G}3MfXy0 z;Sq4+PWl)#kNk;S$)6ZS{=`RHGFbDj?|bqmwy=KNq0uK5rO-)A4GbT~J6qR?Ly zo!~EOpW-iyP4yRbT;VV3oZ&C(%JVH(-50ID^f(s>T)5)wNA6w(%=^WyB${*gt9^4A{f?4btD~4`rK4+B=flj`J&2lm@tZnMM z-&pq`d@#>9tk?g&uWu;!4e8y;8bnyXC~LV^JczMsExlv(;xs4i@seID+WlF4!Ch#Ew|;6HUca1!rK^zhsy&Ohp5 zeIvE3)#6ZM8sqBVx7EPk37seDy4Za0!`sd8yt@e-WCic821j;VzTT&3H}U}QG$4;2 zgAVDx{yg-5+QYY#z_){+*Vj6lJ~Y>vv?2dXkT$YtqnPt)F1d=FXn4pg_}sHQXoGue zZ?wuM;oFg3ngxC9xwSV~<=N2njohEaIu`rbqj!Z>K9TlJO$ca0Hcc0O^q96YdhZx( zm3P+x+daUDPSpi%biHF0b&s3Za`fU`!QoLYN2h+Yr5il%29N)U9(zLo*ayMmPTIA6 z*dhUNI1?Ps0f&c!!z01r^T6TZ;BXo5GIrsxuBFpt2q&?nx=Y9@(>?SA-HTr#9Vesl zJ zP2BkRGi%GRCB1l-gkyE!#67GNa{061m2f#25}po2j=AU9ryCt8h}<56+;(-TC_H2z zIS!NHAJVC0Um1NW$4X?m`c!DT(WefA7q(7qw`?I7)H{?=TC-DmuLX|0{qNLSxW$(8lunefY< zNBKyG56t8GYVb-v68TDv?hGBonKvBz)PGOD_Yv+n>$2W$FGhRs5W{xXWxduVMtd>V zCC0ia_fY;2`D9)Hkdha|7RCBG| z;xY8U(&#xni>+v$<+Gl3UBz7LIoH3cb7Sk6AK_K?j+Gk}w=whCWF=xZAvd7w9_4m6 z*|;bDT+grhIB{Lx=``>v#wDHFwS{fm7wpCk#D2)*ohR7O90$K7JJjFx==kU+Z9($Q zEOIS0&aL3q8pfH&`7Ha$1J(Nz`4r65%QX3E6M5dp^L186Z}pD*iPKxX_28CZ{TEZjP&OXLd1FzJaDz4|n zqrj~t;(O~|`(%Z)?&4RnPmHgPu~xIDHXU*P9Co1URZQe_mX&FER3|)2wagS_x9K9- z8^iX~bMmzX=sS&ZtVV9v;Y**7jccust!3!GJ+;cf^2 z9tZ!F!<5h5$G|`NiZlkne}l6QG4N0}U^DMb+_K3o_^Q8UzpF#);9B-Vg`ZMUzvM?Exq+c&wagb(*nLtbG>}$ z&z*bEd!BpQx%UIly?xHTU3?1no;JTJuQjvv$L6<8&L`Gb?0e5S_ntEMh8nnba*hA+ zfa5Stic3m}}vxYj5WqXB?UA zNqN`F7cX=CVeq2&hmRcahhc*#UvAhn^ICFe-P&@ZIJ@PI**Pu4;43HKKhpnYi=QaA z67g6`Q51aYj^p#{#piVrpVyoCy#9dC3;Vl$F8mq&xT6C982scnZ{qjDws^g=BynKt zoT8Jy+{V`{eTf6d;aS+KcLdq{_lbEeFJC0PCmWqHr)7T+`WXHG=AfN79-%+ud!gMJ z?Xq7l@gnW*k?(pRelK!;+sXBfk?Y$*u5Tx~zFjBrdtHq#LEq9BrL%3zu%?~hp4XP? zh9{YO=#sK;PU63N4clTcTgKHJlWZAmpc8lTj$_Lt>5b@%CF~0nZ*%3Ec*k|**!0Ua zM^1`orSh45pD#&V<$V?zxbO46#6IS3A9E-@QanO_V%JX+pdI3u3I;Oyuv(m!u0R0-q2@d^hWTdMQ0-q%j7E+UlP7kGYtAKKu^%R z$dA>I&K~r=d~)!esf z=s9yO`n+f7x4BNG&!-P3)6dn%Z2Gy}@h6+}$l&Fv``=`H%_GmHvhccuKC33!pVd?B z&jnNM&xKc9XL6OfSZtr3lc$4BmA^dTjL*#v zJhjhn{N*j!eQ)hcOFGX-dcJr-*PXNtjA_6lyGC=~b!RA%N&j*1Fn(8Q(e{t!&yasE zh&?#lvYw4&OHD%NVpE>_8T6J-J8>%;{n(etMrKx^w?2fQHygS6Dt4-`LwK9nn>AH2 z|HZ~m^~p}<-lR)yJ5~4N*mixyUCNN{-4*2fV}lO(P;%iVBWxeaBOV`01M`E8Ku$!0 zoQMQD5y?K381`oz`?DAO^Cb4?o7kU!!2Ybm{)G3$o8e8eKe^W4mM(aP?9Z3cyUjiM zs<_vE7+Vq=dIcD}SOdlKWPkQ?9qG3}gSPz%-gnL5z2|KEQ_n{n`%`hMJo!ZMbLGh= zA{h=Z>#k7DCRp44ruH4%YcQ8J@$wq!(38QHeqK|5rO~0?=jAtWujLmN9;fPJ7dr2I zbuq=K^K5@t5?>qc?F(LI^fuO4F;e+ZQuU%ho^+jz-Ug$WE5AI==X90`_Rh=^r8&2XJbEmHuhR+gX=|gmWi9VdXXO)DSi>FW9@5gy{H>`cy_%g&DiJi z_x+b~Yah7v61dd{ZoLj}9R|1l4cuA^ZXw$u;Fj@iaIN_@F`K3VpM>#+!c)8QN|w(y zzEJj!ToXDRH@?u!+A8uN%`@=auHhxi=*h+Nc6Y8tW zC5ASlr4Rb+f$q8+`RyC<2!|f$1fWImyB%Jw^Lb^{DaI*%EhJyi7UpX^^YtY2^)&PK zedg;q=BtAFLeJ>BCAp6%eZ=nJpME+Pv%d7ziJT#aE}^}EXWJX9c{i~ZWXX3=Q7ex< zq()xRhHRx6ab{?CH|T!{58m*Hs8A|*0s}biRDuabbS?kyB*)cE##1h zPgmi;VLvSA`0<%zB`R(I&H94*)5De64Z_WT#SduPL^i$|pDX&-iDLAXsmVIVfS#6% zo;JeP(->pFo>t9x17$^;kM7Je_KlXYZ?uekqh;(HEo0v(`(fdMF?e7HJg^fU__vH( z^Dy4s`?$}BrEmVye9tucW*gVC7Y;hl9O3gd=lXYi9_AkzyTj3~cSyJX zCApB|IX^bfWw!o^PvN?&vwo31+R|BX#b2hp4foH)=N@8sRG1jAaxRc7Ca&wqh16%)dY?ru=vmiRz71U` z%KYY|$3{Mlz6d{1K9lrV#f2pY#-YbnIC`vN!hYs_95G?(v30U*`}Np>(PI}4qQ_=9 zdTfoO$11O3(udSz+u^w}cxtL1yA*jB`7}HX_}gbgPi4?kiXL11Qwttq$Aw=;{=2bM zL+kxItQ}wNQ=SU6A>DNdbm8i-33K0$uVxs!vE!?~+)t&i_YWk~)4Cp;o-Xs`zKEu5 zdb0CfvWQQj2ROVn8I*^4+w0epGmzc;0zRzv zyYOw^gFN9q^IO~XiO}--r`a0-PWh3G0h4bUhu?Oi8!CQj?>)?F4YMbIZp9i<<%{jd@!q(i= zO>b2vpI@+vH8STHQ~)RVK6Xf`r69MY<()0TaMf_mCyoWfl_RZ|XwLO5v5~V|-YLw| zSxhbORA+^&b)Pwo&fs0(i(X9(7QMIq7Cr^vT<*_R9*&J!$MVhItB)k~L_)4+|$Ay3Um4ge6SGe%RIdEa72N&-9t21%o z!vAHsF!2L$p?nZrXeORv_4cLUG0*CLCAd&<_x#gQ z_=`>N6T%;1ts>Bo`~aeh7<@&V#MG{1X{#L{b)P&h(i`SDLCWOl}~M; zd%jheLjR1}(7$Mv_xtI;+M$2=fKC6`LI2m={6O>%A29SU9G}X&z!1H{rf*kQHhhcz zTo|6Aes=`EZ0I@#j;I4ibZ$w)2VK`^vDe6&8b18F(@j1^TR~`Qn9qLxfzEPejB$3< z;>SdzRNDwJriR{L%su7`ViuILGIJ{W1*nQZxbIm|T((F8DQb_V472N2w{?iC?xyYhn$o3H;_l z){syK@{InotAWTv}R%|Ie)llz;-`&gV!tZ=n-Jr$H$yPpJ!l&{C|9QU5+Oh15T1@+W z%#U(`bpJ+TI`UbLWxle}@wAVHysg>@bq&t!M?UrCJHN%;NcZ}t`BbgMdv-qJnJ@AC z^PG87jX1A8@`~A0JK#?@xfkqN2{Ct(rO@)7#Oy;E;p~g&wG5edD>)I_Eyrf$v}EI7 zKUz)Qsav2`>{+khT=s=(=80A%hh!5sX&)1Gsq=8g4#l^VUXqxSRg$1ySCVJR8GQI!(v7cWt2Lytg1KACTqz&6^V6K?Rc0kr z&jUKxv5LA`_x&66g1qniG-oM6_o|yEId$|V#)ixa`EuCzoeo|S|ARK=r-`2ok6(sMAra~E%VYB zaCzw~l1IFgwU0qJS8PIaItdJA%CE@PSpf2sgqz~4S{er!KZhOyg( zuR8Bc_VID_7sYIP=7kbVwC|U-`)9_b&pY@$f?c;!eo@+>_WHr2@RT=zs{;bA=&!M zA5u<}n?EEt2mH5#=MU+5um5&F=e>84V;6r?KHJP%*{fZ(_ywRPzW0jnXiq>U^C4gF z3g`EB^Shtl{jG6muRA|fgbg{p8~W--H|WNWdD+UJZt@6qZwmM3mK04)&z;fDIFHg_ z(8rzw+86Itu44~vbyt@aX%BgndC&7%UykIi+@QFXO=phH!*g0w?>+edwWfDaW9TUS z>Ik+=J9f1A*PJDbW>%0h76TU>Y1`0A%KL&reYf+wdo3Is?AZvM-yr|i}O_=*erL&y5apXF?L#y;7BxsZJuUlbhX`9SMOc|Ona zNqKdCci(CkG#W2tt_z7dLfhZ#9H4F24siFdc0sf8Le{3Rq-X+t*wp6Q-%dPsoOr;1 z{BUJP`<6hF;<6vcztqioh{i-~ijUfRMS}23#oV&&+FE7!Q)lB(y$OHnr}3xWfdK(8dOhlEqOMR-{pgv51$;c`IG&7@jxkji8%)Lo*Zib4X5_s zNNWF`N9{lS;Hk1zGI2+5U(5;_%#kVevbo=-{XMe_c-AAJq~O*3##C_Q3UFfvxZ&guySQ<^a06Z71ai}h8*AY6<~}^g;KnWg z3EV)QruuhMaU-N08t_0q$v+z&FrSw)_o27@p7GB6pL6{W9kbV8x|3`JeX53XWJ_65 zz1Ey{)IG)j^DU=lo9h?k8J~}u5y%1UEyPbzR6NT1GP+WE6n|DFc8&CnrSO&1cT@S{ z={DmlOyZGmaxaO`RkPP0es*7nho7;>{}=e;r5`B%okgCs_q>(pXyp=6M+2L-&~v^= zA-QAbd=G3^o$pb_x>c}#jp*oO*;hKtIR~TD)E^Iow@`z++43#h$JnH!_w)A#%jk@< zU*=)+h_9Hna_-yDIyB|n)0rJr*dN-{+K7A_M}O{_6kYVUzdjJ=Y>Jt?>8~Als>j%8 ziNped$#vGpPWhpyvFx#RVbrt9uqDGF*x0M^Xmrr?IUKoqE_frri_EX!{EbHN{G9MI zo>X|1*D(%w-7%hPY*n<6-G275rHnyid(Y^fgPpUHJP(7d+hFT+_UDr4s){<~h18HG zM~?Gcs_st33F*1M9n}9;o|AJu-g8}OuI<_)`CKRA`fzkubZp`K-k1m9ha*Ervft4o zt0lAlH{kzcW*r8b^Z$AL|B7Hs!JV_Oa|*ow%klp&{+Hwb66NtY_`k?=U2U#y{GZSD z8Tb$G`%`fLdI#@ad|zwq(}DWp#Xq!`ZN^tnid|HCz7N~NKRM63>{L7VN;v;#Ec99m zy@qNk@AmTPf*)E3L-;JM`jO$i@JjD*(agW!InxUtHO$&?v1QS3+8x>)Nlg8gP?64+ z@Y3TeXOTy1$$|FDqnIO)0{TQAt#stklZO8Tcd|@6u(keb$)vlJ`F=8g5^r4epO8tL z9huZ-czE666h41oKYhZx@0MKh17GqPOCLM661k)^mQrL>3XIA8K`{Qse^EB6j*Dbd zFS0{tL!Armv5YYl?kePy;7*ZIXCJ5f(KwUkRPs1ec#XyxFl+HYBd4mVnN!V~hl9&0 z&UQjh)%A)_?dy9z*UQbdO{ag)^%->P$f@&OT0j?b?aXWczo7@ti1~2zfL;Beb+iwl z2lTZ69(q{*UzAlJKo3c9|4-5bxl63Y|6Y1{Ub#LFJ^avf{j9mR>0u|=XV3$3t32w+ z1&0Fb^+T)VRK;|r@R_Ztskyt^ms5-Xuzb=`>J1S;>A^Ou^9`x3<}*h7-Hbzf zD7*M9v9j9=cz0vIZ?spoO9%9Rzt&mjYn)HMgU|sMW9*PU)!L5-aP>yzQA7mm}{hNGGH?exMiHWd!% z%+MP3&6%OX6<2>dwa@l{E}5k{7}1}96m4Lvb=YQQX+;-+|3AV{=bkg77%=V>-r+r7#|*Sxw@Sh#EG~2t@ZKSuL@@p zFE;TneDLbGi`w4oNnO9Hel)e^ZFu_Y`_yF<A7Bj1^Rb^d zxk0DRhYoGqy7l+Ft&G}E>dxpl`GQqfZ!12pcE+=k@t7DJW03FEyT19HOQJEg z)BZL5w(l~lydc*)$TMxF)T%DEzV>j)vKs%A^K7cvi>I@BcH^hoOsu1m_BQ4=Z1B#5 z;Jl1>qR^OPtZJ7rPj835lru1>cAZpv@7)NUyF`EW zz%Ll1^rslNu75-<>#N|i?kV0QJU08MX|tU+4Nmji9{Hxhi>PzIgErmqy6a2rK5$|Hy~D(@)^C6qJXhfNsm=_~K%*_(M69Gk8h zh1(eaD#p>*2hIc2>x{R7-`wxzd{^Azex6Z%3dO^atD`!XLaW3IqQH*L$NXfpZpk3` zaN}g=Ikah)nJsaQLJ)3q6LGxqy4m?6@s2JVN>^n2( zVGy?!egYS{PI;~KX;V0(+`)~0beSIxXFu-OcCMl(Q=}Tahu=hrfyUrD@JoZ+b}UJG zC+V$Upq-AX%Jr~LiKbTNuHMjD%Ni_ZFLLhv8=@0fuWI&u(2t%ozk$y$u_iCUcOQrE zs(tkZkA0^2UF($M^zJ;jbJo1taCuKZFQ!e+^WQVi+E;ikUfe~0Uj8e(h(H&r%@zmW z;?RX~tC-w=(Zxh?fjDUIUgk{qUICtXu4U+A!rKFMvCW|i(-yd{9LNd0+s8Xm_TiX! zX-{|@=gce7MFIDtJga!C@|l%)ck>_m%F}DAX6|P1QZ=#1Z}Zz}#`Oo~mjJtBMFaew zIb?3N-@#qmMQDX+k*Pp9hF zzO=u4xY){H7D1*aJvY)?-ajtJl%_E`^=CDv1FV;4ICg%xqo1uUN!HJ#o3%f{JT)*! z@D#6ZX76jvBhMg@e3?AtD(9j*@YS8Y(k(_bZWzn_=-Ex;J9-A3G|%AU>J8jyo^hXNkw*2+x6Jvq zeH*Q1d%+3wj?-Qj&uhN>;4KA%_G!1LzD0Z9cgZa+Qcl-?U_Syp+F!CLXiYnkX7{Ii zmBjf}Pw)u&ZRDVo7y2^VjulgLk-V}K)P~sq?$_6yV6FF42k9hyq0cw8SN5Xz+DYzd zuW}b_Vdr%WMWzgI?Z~tatRl}W_H**gnCmX)Ty0AJ90yNMqUUt*&T-y7Iom?78q$A8 z<1lnl?PZFBI~CZ)h16Avuoih-KQS5J#G0vgLWcH)09S`%=d4}SSy$z?eJ_Y^#k#t4 zag??+KaOtKDx7!}*p=6&TGUTW(73I(E@0ei=}gK%TNdXG2*=k_YvnlO-p!d{8n?zN z-B080amH63~l$G{tRPQR}9c|KimzSndptETYomh{sCj2OM6H1kzwGGX!a=Mj{;|e z@x>Td*Db&d-_C+}9KmN@A|A&*_|-F#VZuYfBU$^I2RRR_G{;P!h0o=^VHGVw;iq5FF7V)f;$PYt!2{#@&G?I7!8)`m4YKFIpq#`>IX zj1ksBbD{i4jnSRIU5v@(OgeabgNdb6&ylgIE;qJeKmK+J9~gK4x-)j^j$XQQ*M7JI zyV?;=wZk`~@XZK3GYy_-7Uup3$J>?s~qs^N$~1P z;T1T5j&Lf6^{n-c3Af}nZjgL$%TPVW;o(+&2lvGHJ01{ylHX?Jfz=vqkc>!c6`vP> zpGEFmdUnlp$+;(~CDB9MM`=s%y7u4=j9>f13fTu-p8IWo!( zmdwyRMMa0`3bs68->E~3`!pujC(b-b=j{gf1eznOVii;kf0W-o_Nk49hJ@ST66-jebG z^UxJnfalt`B0eH}z{FIw7PNch_dF-s08ZWix7QMb&X22ivx_^`$K1{~V~N?#Dh}iN5N??_-Uz`S@n^(=znaO7zpfv3n1PYO0gY zvXfubKT7(%pVVLYLLND^_(lsXKQ+$f4_LLX%JMe~HvDNuzuTK$e<(`6%SSF^pFtk> z2Jg3zH19O}&L^iO&H7#?b01KjeE0L|JKv_eeegrKKlqHPr|#`*K7HBqYv-=WhjBi0 zlWFG4-}+HL-zCS!oh#{>8=X1&k)wz8&ksIe@A)O#7m~poEs`C@b!38-XrF3fFOd(% zn3<~srP#?a@JM&{hF$1!}5J7t)rm zW~{wd&KU_deg#^8>>0loSH75nEB}%*A3elxQ}dTMU-sJ~?K~@e=>HNA*Y#M&;KzD*w_ou$(kK~vf0rmJHXr3=#%i;FL$Ee)@mWAhC9qDGcai5ny+3e_(&5k~)cii7&jy~D!=#yvmX}71oHGkfBtAV)={ULK9^i1d>ruK zN3kdTY1XrT{2k!Xba%YpVZ7LZ@-GkI+!Z#?m1wM74~ld6jt0d!*U!&b2J*D+`Y!T! z>x_u9pf$Y$zv@urO`*<=c*e$wLgMPWE&)GW_*Q8>(4QQ?%Na19e=ZnTCBaCJheH<* zjLO4Mf702?u)_<>??1JFPjh~^_}%!1)W&GWQxRJ3+TWv@rzHG72ad#%k)52Me_xJ$b;oA3%T5o{J3Cr1>&>K6O*Bzn|=%Gb*hkBv8xNgW*Ual1rSo%*&hg1FywhZ*aUKr-$9D8o>E-yDu?sEsq4IpD;^&3%)GU_IQ}GSqws7nTYJsaJtaN4VcPW97M0iHu<%2T(JYegy z@HDd?pVoTtj(D1PJyaV;JS9e4Uf1K0^?b)Bt22JbRet;t-@YoWeV~HV&wcUFW>F4)jpR!I*u-Ttgrdd5dX&C=MnpVqR=XT0~-q4W>@z-YgaLQfT~#E z-U@Q|utkjC$r`URYnmT)p=RCdvG;GJzSYk-d+DLQ*1@N!ySMuye24J32S0Bee30+Q!3mu^_)xm<;CK0c zWTCZe4ZnZK`At562Uq(JKBRY@-`w{C{N~n{`}t67y5@C`=I_3tk9`Sz(i%{|y=^!3 z)NiLIz?Z1)_Z0rTI`T$tr;ii#Eu4L$#tPT@a%(>ms6RAp^Vc`52e*!Z6UT@NJ_x>k zoi<)GZ4Z0w1Z`H(-czTI&S&Vc_WESMVeG zWmGHXZ(AJ zEhBQ6lcpuuMm$X zMtQ-SQ`KQsW0h6V&1`w)BPq6-0|SsUQ< z;;h!xrHMMrZ|uDq=?LhsucE`=bYJkXT6EZ&_tqF4_O|;f4$txBG~R#?>poMLW?f#7 zZY&%fHni=6pl?{CbgEj$Fl=bBG4CI(%je@`dkr0K_&-8#LtnoDx;yw9I*iVJI*MKw z!}qNmjQT9am@>jQB;&*ZHzz@OAe{Ie@$s|Z#Eul(V8r}%xe`7Js>UDp$}=(@yR1lOqsaydP`N$dPm=64utyA*l;I`aI)eZwAm8F^lf zJkR@tEzfl>=dXW#y6IQ^V+*cVZOCZ0UGwYHKang-zBh+^Kj&Gm-16qM9{RzUFzRrJjXm+H%r|}IzPhb>dw3k zwjbgZWVyFKz3LlvV0WmNPR2E5heMgUGw_e@Sa_t27=1?XE9}qD(1{MzcSyoDAyYQV0%dwtJGSJ1Odct6Jd1w7Zu`>U+MURHWZGKEKmORTc@CKFH+2uJ-WyaiioWH~ z?}1k+7hi<#BBWhSHO=HJ0NzINaMem?>@dO*01<4F(&l3 z5NoSEDe5tdIo)J&?s$y3tt4I@V=m|OS;f1T@SE&Q*)F!d;>77QX)DKu`Nx7e6Wq!P zC6f4)fo~w*x1O>5EwT9doWrCTyaf!}Gt<2ldj{N2*}JoTu)RBo1FLK^cfDIU>+${J=MTgdpwAfgYCHC86nZtd z<*dspY~K!YHny^U;8P+Bt+qp}@{5*`_fr53X@8RTfLQOYTsPmM_7yP)ZyPIVqn$P$ zb=$x<-wtjl_7(x3DxkqF>?xNYw*%f&L)%V1Q?KGg+K;0cuX0f%;CDNC-X;G6xbzhV zr?5q4)I)P>s}ws-c8dJ-%0Y&wv{jRXTy1iY1C4%iT}*tm-ZDAJb>pmMd!S9*{{hXD zOQ)FWIMpEGnfds{nt3Ll&)6->$?*5?#imnz7{Ss3EW&H`*N@-iikmvdhG+YG=fICH zf?w=*{8SmTtMoi{Wam7JN6T+S|NjUcpBG|Hpjq11{v`3{2y_*lsyr|MvUSYc_B7wY zII*!O>hawzu$HBVS8u3e{FC@?EB7Cuziyrh;-Bu}nLK>d6T%N{&^x=W%-VeBN$~zF zxS`KO<})ME>eDjGk?21^C#On|Mz{*r1ReR znuRluIP-VJnZI>TA9g&)o+FL>68ehcd+Gyz5@^aN+Vv!jyG628SJ2+uN+5%TqG zz5=aH?pzTY5N{XW7~DWM_4}SiGjCUbZ}Z`t_s=P)NZE5H}#q$a#(1M^|#O>^}U zeII4s^h_`Fre_vA^Ts}$+CAjkYrfb=VCJpT8Iy7q?3{RHB{?;cqj5gHwBWU)>JJU@ zJJ*gnk2YK#(xulW#B@{XRlXnLEBI~pH#|iQAq_uWGq|Ds4Wi=+bba+7-fPmn9_1&* zpd)0`j)hGnCMQn&fcEFE*q}WQ(cJnCwqL~db-qpQaK)jU$(_T-eO7b3%4*&)E}a-8 zcps6DLwkygmvUe2V@DcZUY1#ZC7l&7KmQ*J^XFT!I{h{;1FXj3gW5~b> z=@I`#2_<(ilCTg=1quo(WPTD^5$m+?xQBK-B zs{gYEeTliy_53Xsa$I{hFAQ&DZ3W};?Z7BKfNRy(iUQ*nV63*T2rtliSHO4?FmA2? zWO(x^ANjNfF4Kl|(SwXJ23+Vvi5PHI3NGnxZ&|0Fbl{SI2tGUQ3m&*Cx8uup;o5FF zwcBj{Q}h1h!9d;LCX+?R{u!-_HCz!p0nwN}FXjkU6(5^l!qjnN!6RAdJa2k97UBo|=Jon4c?e2vK z!MA<*T9t!c#d>ny4KQaR3qm`8%6{!t*fTr!Qag5WXjeBf6P>m-vee!yHl7-_HN@Y& z?<+4k_5F>!uN-3U`{R*gM@NQsPF^y8X4k#e!RgA0xD%dq7xMWYcvzj}^Q@x3qYg@> zjJ0E2+8?=%F?7u)PlGXRXAF7o`c9QGhP~jq#!$)_BrCk|*VDh=HTyCJe_1G@eR4gc zLpytjKLp?P9SpPn=W1i=AZ?7JjY!S7nZ(CN9i94@)It7qe?8=*4fq_ZiB$y3ituMo zKT7|KQz(XZ)@OOl&HFCJ>NGUjI z>wUKTuwyvI*zd~u?WkEab5(?MVr$0F+yvcho@tMH0WdudOrwB_xKrYB@JnOfewkoO zZ~ZN_DO+(5zfWTAvp*hsc#d!R&Z9MxW(MDzeYkrDv`?+>ec)h$gO4wu|EA*O@AeGf zqw3#Ewj85v_npCT`1L{F5&m_9f3Nax2D#B0A!}C#G~Z9_=qg@XuTG)$_|<39>a+9{ zzZ$%_X@J&KX>|hY@1^;!>N9!97TW2mhG%%%(f(5LjAG8}>S2z?Q_Fn6buSG4h1Q7nUOnGlgUPhJ|J~V#yBO0E z&e1t@4U&1stEOT&FSDd8vCLMk0I8L-y>m_mysWeL#k} zI=AtQ>JQqh>($pH=)ch`9G#kV>|pN5EiV#nFM34hEFd4C7x$bS(c;gu zIU^*%86kz>^3;$y`!)5sPHO7z{ELda#~b`bi;-Wkz`Z7K3LgU3Up2Lk6}NPG7PdXL(N#lwKyTQ*8(p-bq)7X$ z12q@Tj3Jxkll%^`ix_-CHh4>a?IC-P2c6T=z>)_n0bh|@r#jOeQw}&W9GnnN zMkD9)#8Wx-Cy)QtYo#1cR6zT8DI!Fa`VfJrMfwxwO zx0V&jFB_G-rmYO}EYf}HZCh#UV)!d_Mw{tIpOK%^JFlOjEw>F*&ww@rPY`+5wXd{j zk!QV@q^wslZHkX%aBcuKmRd7v3T7IA$lMSxrz;=A#-n}CdKfvH-Wq^EC}*S9;kVi7 zO00t`pNl!C%_$d+Ap7@9!@&TdoGO{_|V<@bM;5UW6>KVm%hoK z-oa>b9DTa7(WO7t>uxdfaUfpd>I?Gwhwv?%Ju!Log)XoW|Ccr9W8spN+G3X!CdVgS z-XVCt?13ldfx(rRYg77Hzbl#(O?$!3g8B*l^~A8d4p<9Dhoj_{w8sYOnhhJIZ}nU5 z^y{rBHgG=ofV}EZohr3N5y~WgX0?+ROt`~sI8?!Ncwr>h0 zzT*82-WY+-V$_Jj$NlA4=GS&s<~DC=U<}pP$nbowYrN0H>m%S4^LxPhl&5x^_US`+ zzZmPxuPZ-N*M5twFP`U_>wZ4GB~&CDaOZncia!Y7=q{~ozWhq{9}j0=eOi8;^_AEQ zUcb*q@&}@Y9={L1BI#~E^d;YXztNfcHX9$%z`N)V@&Ww=@56x1AQ0S8lfMERH2wj((qzK0UK2f;ttu3rD>XwK8n z_^TNIX6&U!z#?ClKi~F6`53p&+dMj<)7NwAD>IZ(U#fkWN^ht3+48Q3T(tLxe{?`+ zLUu)~!N0}8TnjBPh4(htXM1#<_o1OE^NBK-zRSonVJ>6?mC{csaTNAzH~KEKcHz@@bT=nP2ArvUx6`rLW_Z@WpZ=sj zW3SZ@f)D1r1oRZIy(ared9QxW3F9vv5@;;T4Ygc%ehIMqTRKMvT1vIBYBJ9>8y{9M5&szq9e#tDE{;Z)I&n zXX54Jrz3wjVs=_^%D@ z0UP#qaLx}eN{*il*gzV`Yi%}-4pz^|$U|~hlGX#hyg9>qMsiWQP#WV)XM7oZd|}y7 z(mz+A2QcS}0R9#iN4LN`=8}&i-VwOi#&6>*XU?TZeT98+vaNlzmqvTqA5cx~EPxL- zJw&}ME3Iuj{~5HKPP;mfZ58{$wa)6p^{ZV=ABFUhMj!I41ss38cfMuc%YGAX$-W$7 zV6gWED?Z{~fA`t`eM-QXhOelDSXvphVC?$_oek?Z@oue;M197{ zX!=85#CX4sH3_}f_l)r~7(L5s9SsZ}{I+@`W5sW9KYeV?C)YJC-PBxDdzJT;*zMNo z%(7%SS1{g{ywiZcQn8qP%_rm37*ofj`U;9wd&emr^*nQz=is=PriCZ796S*(cWL=9 z4-LEcKZ#hTxBfx^n2`CWdXTeP_iAF#fy|jR7NmXk;bv;!M4*?tA!EKA!B4PY$h{9Y zhpc6LppnYpJr5TXf8GRbM)B=>@!F-ALg*!m&!C@P_Kf>Ldimzmv(t;}`l)`wKe;^4 zp(Fc#XlGOEeQb8kEq?tZ_=cu@x9QJny~#7bn$wf7+O#)tKC&H;(fPNj=Ue%hJNOqg z#roz@JAhhttA?mnq@$NXkM$OHA(w)i!1PJ*37lCkpVmw8xjcLaFGH7^=3;S!zLXzod1YcrOMFY9rRT}Cmi@WV25s5xgm&iBj;>#(jZF4RsvYco>Ko;?%17_M z-*#CkZ3Jl}tEKn**)8##Lc-PcHS`}$4<&-U8>9wQAAVWI;1stGW(6DLLxPRn<3cS* zFD_~6-X3T<`gB@L92nxjpt^SPEoHPht0i8T(b6*|vn39V#LMuT`>p9Ozzg;{JjSif zaV_~Y;xR8VUlXAJz0iNEuTK0%w)99BugYxSeIIpcwtXAD&g564Tcd|Wsy+NR<_~8K zhCj~xb}m)ALsxd2cHNxJ)}H~FOIup|zjfArrAL08DhrtNvDc2#`N!!==VM2nhTlGC z^R;vr4-D_H>rB=&md8I1KGY4RUhc4#zI&Lzr{O2W>sI5R>3oiLWj@h+M(KJtG}%SX zJL%c8Mu?YZrQ9`j=gS+@oCXd9{oJP zyUJ;FaYi{m!kGuaxwYV2qrsWstZyzjGoq!x&j#Le-i=!p`)R;A@-(~axDW1T4$bu7 z%$a?6Z=VA^H??$62ofvHYUy5#{)rAZo7h$yeMt3LLR^N|CJ8{51z-N_qDGx4YgpwBpbwi|vcU%U3~ zDlaew{zc)pS|8=bbQcoCL+|P_=OCuF!o(iU%;w!cPyxAr8zUORWZjnCD zkbDfBxEg%~IUJ#_C}%TDw~HZXIwsII?WlH@>NmBUcS5`a|30 zPsLl(M4#Ye0seBcUwtXCp@*7y;=Zz?|9F@4z-UX)=sYvkKJMIyFSV9&c%Fq%IeMGB z26lbRWH<-Ar~F5kW|QAN0<130M$4=sz(^Hz>%Fp1@%QoI@df$z z8J`jGL3w3HU#E=@#e1-D0u70n%}7-$1{?{2EUz{l5l?k`m{b5qCcx0 z@9&bQ-Nf$KqJIm{qx3z8zUR<)wlBMJlIkHa4qz$DA$DF{uxCRpFfXFM=F9$>(}VcK z^WL*g?Ze*F+{EC$=BxmRx9_Fjg3yrg5Uqi55Zp8T6dP<6e7@7zifPdmJ)5I&y+jKhI-q`mf)TKhtHxrbKIYEPT|!vT9#{0P6p zGZAon95`V7edv>aOTE?w&}a{GW6~u4$#>3UkEUYiT8qUFKdNzT57(}c{E(lU7^AhhYK4xovKomSwe{dbEcSi;%RSVU zZ)YuB`K^2)@4Lbqoe>a6&a%J4*m-WBc7Kix3)sBjPleI-?RwWI@scASJnN&H)wl`z z|6F--V@iIg^0+xWV!HA_CEJy!8Dp*XppRMge-Zw{TZyJBaIut}H`$z+{;ic(dRvG7sTKSV{Fv60U*CfWYt^X>3*`S0wtbn~!fo7*<)ZKfY@jS%-4?QuzK%|;&R z*|GTPbSBL|PZ`)7&@aQGBR4+qb$6{CS?^seb-+}5S@Q<%8z}Vo*VWOF z0rH9uwI`XWW5Z1c<@Dg zorr-ZdRWJ!te5mY?YsF?=QNd%x%9E0z0R%=->Vm1GYEa0Y8Y~qnJv2JnpJ?a3vb!sP2;bVZin{Wv>W73^lfR&T7SXSAx#OxE2JfrRe|bX|>#e#| zHx#_QA%{Hf%8RU`g#{L}iMTpAn9llFgD0xlUdp=$zZymu`|xFOVYH8P%9kuYy`5+J z(90&FkKpe+sPB3wc+ra+Ug5c`c}{wn(Yv6viM$gO9_9HHoljduT@(D81N^~Pg)5x7 z-Dvjgj||VV=k|a*x9!ZQ?Eh2bC+K=Ca|>T#FJNh+jJ`zYti=KOQqk$J?8qxMI(84w zD_*7V3u_jfzKHdhUo-A>q_C8ldZqAsa(?p>kuUEH6B*`W`K2 ze`8kfWa>WDG51x_raMOUdprGp=l#BEKc~Mb^mp;{>eCm~*Bi8<{e02D-8;VuUWms^ z2hIaGs~kM+m#6H{OxiA!pJh%c=5NWFC`#=7A%dziCy!rkNcZ>X@y zafX&Vkq7c&$!70HF2wLHl3&1_^Kv2EnBb~IpFr;!V%2c^Mxs# z)1&%l+UxUTF!b<3Y{zbJF-R^?54d}EKw+e+!vfLV0XsAVHd7X1DD_^gnwqTFVZW6z1O?3!|}IQwkVfGw+cQ>(B9SPH1g-VX<{T^Ei^-1pzT;qbU2wJjQJ!2n@_0@BYyH-r4xwx)%rMcQp>4xeA={%{`B+tpb1B(fQJ#$LoUr##KYo z8&{tnYIz{Pq^0vQ{Ga$(SK%L34w+(}vzQxGONez={8M>4v9gfjox~|WiGG#ddik}d zo7A5CIKp=yYu7>FD|p`ci#*SdW&cA=ZAkazzGRYgN#V2h94W>VgPuC(ale9l)!;d8 zMTw2l=Nl$Ay2T&<3U+URK0D|WTJJ5Pu9@PAdRO+8a+9QM#%Lp!$0zN?Xs0rzokjnY zXwtrIH&z!!-%Y1o)pVX^`P%mLKY_Sl7k0vaY)pLG>$X@KjS(YX@ZB-~#5Ursw7rcv zjyK?YL66Z|6yIyL6yNJ>DbBS_?Wj%segwLh18%OwkF$k-ej3bZtQ?ZoDE?Fpf2xK* zDUYfPo!sr?l@PWf^sGL*(6QBr_6T!kO50}O6n@5-^G)c%n_07%6Km97&%yw8DeFw0#XT3i)8ucGK5ezS zaD5VZ#D87-5g+~u`@mft;VOqtTz=g1A@gt9OEv=wiFiEm3~t#Nw^(5p=YDnN=_ViRwMl+XaL(icFka!9OAAZ?!hR--{-tN|Q9qih`wgZK z>e*4|*$1ia6Oj~Lu(uytnnp_0DO5Kh?fA>qW*{GDo}y7+M;`jEUcClvW)56ip>V+d6Rv*5fhht!Bm4t7aPaI-bsfqXNoVAgAgxFYe^UXW!_AzHxK+2pOei^p6BEL(PzWutkv(vCc|}X zpG`yUE?iseJ?nct`@jdw7dzMdS#u>; zrJq;e3Eq7E2>kLPV)e4`OGD@x#$V}gUBf+_H=H>SH+$MX>db@DZyN0UtR-g8>&Nye5W8~V%cI&&z#$CM=g-6Mk8J!y<{=p~oCV!TXJxbBpeCJuo zY}r28+tb{Bqs$Bag}c#--sU>Bug{P(sJ`0Kb@c35_I=7v%=_ytD<_-*+-|#!`O0=| zaM@Jf)|_+g+EixVvt{#+ad@=jr}E(6$rlFX=Hq9>KjY6I@ZZXBD>)~A-axFb6aKP= zJTk%L=8@e8EQ+yhgRd*b_Bc95%(9-{hRj4?YTS?BzHOAfCenipUN9%Bd7zx#N^At> z3uoeA>Z`X*4TV0Z20-6LYgvTc-o9znr{@!Y!;b2A%1mC$fuQ>u->&o6B<(&uKmv@Q&g-RXp#*M=c!@8>X!w z1b!07F4)+-fpfMJ#krwVKZhna8r?T9{1~}WmFP4++KwVG3d$`v_Of1a9@U2_arzXV zYy!SZ=~K3D34IpOXGn90T{|D1wTBp`Wb*tsOB4F6{6neo0nOac=lzYmeW_%d z&2zZBG8ZRpw9N}^}pcj$SCOV3x)zv#J){*_ZJdR7iY z2lIR(>uSbl#&7E*gN=6~`&F|)+_ke@o8kiIM>Oiv=UQ_<)BwLIfe%dj z%{xt7h?g2W3Yn`hE@aGg@&(v-#*Qkp#`3YHwN7HZ*cxVi7dqoz#CT2pF<`_N8WlT_ zu^{hETl;34w)Vn@^u6k(*^JF@>tnRFhx{Y%AE<6CF?@5u>m}yATel6@=8_)Z<&VW_ zp;NVXn@R7U?Rmwp#D~0XxIA{0Gj~5ooi7uA8_@5g;QUx}p_Ip<>sXnEk3@D<8g{k6 z^}sk3*o5r>0-TSQfx?h&K?XV8g0sC`L-TTm>E2CVzev19Ysd#<1_*)kE>aBMsKI&tP zq9rrW7pIx?VP-Ke?pO_+_@t%ZxcIQ%>C>z4Du2S(bqDl)*&guBXWSU6>ZQ2+sDT`? zf`L4e^##)|Pv%EP_fEDybRG)vsc=JXpw;;P(2-U9XZ0TI;5L3!t);c#{OVP{gIoE& zo3m5c4_m(aR^LI%*VWWE>c7uito|J}i9`5P@k6_C{S9zc4)Dnx1=x8${QUiVauB%O zcmGm(Z8={vIkoWE9R)**Kk!|f);(+Ond|*o8S5F+KQ6d%!AEraxb znWuk?xt{a^Fzii%LB38e44XVKxb!8NX0aZkLBZprpM1}E7rrIufbVtW*kJSS&Ar%= z0-NIb9|k_R|H0PhGtT^IZoc~=U|2o~4BG$Xoud_=ITCzs|AUSH4dq6HAH(6{XB|Jd zH!r1Mt-rU=sHe|k{HAqq|J?ou8-K26{3p&eeqgvN1&04DerBcYAGj3XmDy7P&ocb| zS3{9|U)vcnI^o6PA2rJlS!ei_^iJu0@at*f8PAd1Azf?%@^aPBio)3G@&%(yUL00# zmuc@)qr+RsQG9}SI*37~pPl*lP+wn=m_$}fmF&@*oW9ep4aa8F z#w||YmtD)*%zQ88>^Az&q;KiTE9tv~z8mOUK6v?f>VRP#FvL<|r~!sbU}y(Vv{%P{ zUNE_`TXD76Xu*^LOqngpY3V@6TgdpguF45_$}TN4_;O&?<2hmNE9}@JUCy=vl>_}# zVCn*<7di;%iX{>2iQ7+P4;a)^CNw&1lS6I&DcG_CYanD!47zSCp!6jV6g7K zIvmZZ+j%9=i{6Rj?~vWIM93}e`gRqNBFgm zwxlOW|CB7p-hZ@TA7-BaPWf@j{|K@pNr!dy8rt|z-De$K{owaM8PE@&_Rz4)e_IUy z8`zJnz1Y~a&yc&8(Dn4m?~|9xHSo8CBb%`2<+E6bj!Ha2xcsc>w^H^o`v=>}?NmHh zvD`o?(KP=5G50R;RaIx+|2{bh38H|4MT?pnSHWuKW{Pc25)iF+bOt(hO5c*?a4mMM zrBkt}2?RvNOLp~rGu5_)n~-QVLvb{9{zDK&9ojMCRNLAq=aLH;t$>}918CmwZ|%L$ z+2cj$424|YIwhszLl431QzSgQfQm~RmB10w`d>t%slGSz<)Gt zoxHLA_N42Z^uxRR+=t)CeE5Ab1HYbp9GoE^qbqcmFF3iCxfpoZ?-~23VQVga6*!li zGWbVcmX2|1AEaN_z5n2WMU$U8w(gu?9P9k%?~i%8^=#_jHns=M-nFD(z)Z;|L#GtH4X@6Hb;t}C=jBldIq{ntg%o%s3a-YioO zH~6#E!|lyZtA~qhL=Hnkj_;b)Y!N%GVl87ELkaCi#Rf5QzFIj3o{2}-Q=6h#mcCaY z6HPp;KsF6J23z`rkDWfce0?Y$x|lxnOnrRU?L*&RLH25oT-!zbCwI9AU#D94VHZq- zM|IYMb07A_r6(I?55yi0B|_*wt!)|k0qKnQCs02J9Y!8bpHJux=|Tqz`CRwEDzvKWfM({-HLTCsVUA0l3h$jbofT z3TSwqsUZPA_M048YZcgq?CE#jH?~{77XzN@&UG_wr3)L!AoJmmO%p6)HP#!mICs^9 zYY_EpS;MmwcdeK0w-wuO677q@_blX^V#pQfU!9rY(fHixM-7q<@d~pa@!8C|%0@#@ zlx(~ua9b2zIEFEExbJNDchG%&r!M;F7S7Eqq(+L`boMjXJhGN^n&IgUvJu2bW~@2j z2LAD6$D&J-9gjeR(0l~?zYLs7cAQ)hu&ahx?q9zD)ctP$hS=kuk|jESBQO+QK_877 zYjh8A%U_edHv~948RH47_Ktm5hu)m?^11`;{aCKOgMv}N>Auk}^ksngai)t?hr>J? zslzezF!DWzc!zcLHEN}Dz)2@}l-31L6awq%EU%;Mm=RTZV z7xh7{HMEIir}WiO>!Q7mD=d@SQ0%vy^N>4$XD#O;cZA`SCHzOu#Fq0HnHam3wc;F! zT60dqByyuc`}+mhR?u1n`^;xemAz)!v#wA)A<(lKT8nG|_a>*Iy>cm{EK^>vGqw0DVp&(g8I#Dmgk?*X?T)#8ix zlpm9<^yu|s)+mNPs9+tnS1-Pb|26z)zVSN#GT-=mD^XzP8nmNZg7(ecT;Z!(@RgY> zyz!^zdYw1d$?0iSK3_Rv$VN#ZpJ`!{p&w5*O<;@GHzkhX%AieYRbnAvJazEti){W*ElkIA@3}5-&p|f z#ilqsqP+@TrO0IXqzl<4`c|FXEA$a2)^Jkyao~@3(8n5dR2?)}U~KNm_9fI5s|E?# zZT6>7FP(@$Ynp@N;L?k;IAt(?&7v+3dedo*_Py zT+rp5*Y|nf8RFtw-U}lWl4pqXUKj7J;629{Nafar#|vFN4rE{7XY#2jf5q6*z$Uv@ zaaqOsluz}a$4(5qe&OR~@Ue4WuVb%fH$VOxx2DJS3p_v5S@-n#o#6VAy(^xb@LTxT z^8FJ{c(E}(_WXg&c#Q06t(BK!z5mxAojxWs0zY&Va$_{>olj1$*rBsGd~kSqq_lGl zi&@VM_&j);Og&F3zjPSA-3<-N#*2&xwn@l4Xd(`r9ne{1d77)<#A2FKoa(H?Z?#Q`(t zd1J}F2)dO5wi~3K1gG0c6coH#IXsriW20AavIS)QXPgP7M#nES6o-; z_jiOj9}wJ%XAcs;-j9B1vx55%T2@andApaa;D$Biy}7?~e>Quo8maB|<3#(fyTGZT zfqd+@tX}85M{wCK-hrn@$HJe*d8P4iSxFtZ)|fT2HCQ`yHV*oS^Ce|Kcz52-;? z{_`>Z8kv^4rd!z4=A+|*)^zkhYq|wnV>I#td?`134ZQw*G82hLA)zYjduFaMH_zX9JCQ1@h^*ZlJJ%3r0+*T4TW=5YHb zn8R+CdDbPM=s3824Sz6_6F#C9_Ol@rE^ur zXLAllPE#H}-XZ20f&QZKr+C=Q>371bqu=_dp-sGvtn+wVGDo(AUk=HSH1XBz@e2Z( zbnD3IF@FZ0F8c&{YD?nDsf%6BegY3q{`35rOg=xgKhx0TIpj~6xArttlG|5YSD#b$ zyIIul2A%p{G+vaoF?UIWqbN9)&hzTnJzhuV65|N%gsDb$205z*d9CraUnh*b?xGL&DY1t> zC%K27+9>u90C!n zUrBA4;!RiZ_sLP8oS%0+pORgtT4;%O;fGE=Cq8&ga<2!R8(dd)6nv zrAup(-)rc?9!#&b5`>YIM&k(BAE$$MAQJ zk2%;b9fv)_?>XcZlq2CT)5J{o?3cyLfgvwMtG}iWK%WbP$XxUr{ziLsg}b+te1yJN zRs@Zm^f%HdOO#QA)jmyKH0z0xBPGO+dFeo z_szVMn;;nTr~`We`ycS@nCsV>qduGWyK|j(J?VWm`TLCf`i$o$KBmuKaQ6y6`lGel zFR{?F_i%>aOGi#LD9(kSI1imtlK&EQvu+K6@C{ztD{q}#^2Z0k@f_h7{;IL=5T0Wv zE!*+8rrLjeh+Lt2CZKRx%{H+ALa8Vd5*oLFI}dxz-DiPZ#q4)BGRp$f^B4IrR#1 z=uPwe$iBv5cGv5Cf5(+u1?Sq`Z}UBlJUMuJIkjnfrkonoPfqpzV@gh)5B#4*PCdiC z`pKzP8FFgX0CMWX+(cofoI38v6!Gti@ZF<6oK>HkV`7B{XD4$LM;~B56^_g;Cq7cX z!>M^NJYAb(-(~jXg?0uktHgU&E&97+tX%<*%V+-1!0)t_N2c527rnE5;CI@~E2r7T z@Pgi{9^f4V6Ys>yE$W@W^SvWFuVjwcHPaLqQZE1Rn5#Z}v{+pO4qg7KtO+KbWxjs? z_>vDdRUghy@yCZbiRO%V%l+>j_q}^OC$Z6oNAQ!JyN9z$KDfPSyq7lrq555#64_nlQBIio!&MyIbbyun@O)9tR? z$&VlhyXrX04mi7qkG8vO;ExT4U+TMSudut<6MH}hcRyk!*zet**Lse9=waS@-01%H zyfSou1Mh5ce}9eNH}dtw!?2fb$=f_+CKCQzrR481o69VkxT8)=lFdu zaYp*zcGYORbGO5rsx@|EV8({1mkseV_THx15X^NJ?;#sGvw(Ayu_4ZNc#~R(!S-jy z^ft`qci9liA9HuS^i{j!xa7E#HpHvcBxLYrOZECR-fXE{A52^#8Yp#mNwBfb!D?eFuqzaX!Fd*J?+X@6I;ysg%k{U4`f~AB<(oNXZ6%M|kh!{l zK=U!by*=BrJzJ{Y&PhB9?4Ha;2Go0U(yTi&x4`UoP0C#Mu;tji*W#CzqkQj(raUX_iX!0fIouDSd8*|~o*(RT)cdHNPFy;{S^4U%A zu)nRU;hIDvv2Kg=&$3P(d{xgDPWe*nY;4VYtepGzDQ8t+*>gDW=Nf#_K5}G@*ta@= zAh)mY+0suteohX0w-8nkn zA9+(~pU3{-$apIesQupus@Cp4w#RBahF(ns$g4M+oJXNOn0EruL?LHf>o?(6G+8Bi zL>xsv%~rQ)&w`$aEhZcL%>GVM#+?5~r z==6h}#Jl%Hr!&C=euCjaw;mEYWj>Sb?@}-6)kAu4u7(EW$F87X$)#D)tyd4}wbLF1 z)1O-pxsZOmdPt9!*1G*vrTS_2>LL5}ZNWyccV`qdi`Vnm#c?H~_ysB{VrnX6W^|()7xpkGIm#W9zy2^??ldDtA zal97X1=6hbh`$t#Omuej>_n#MEcH*UJ{T;EaZ2McO z{@61*rC)ywKY4%k(r;f&^@pAv+ONNxK6!t@mup{}>JRv)_3Q8IPu^dzR0ylj)b@R^tt*aJ}7&nQSDQHaGE{x543mv z;QH`Qjvt(a%e1fAS$~H1i~Q~H9iTn-NGJ2iewm>MF*IsEW%#wM<${TA%Tf$XVisrL2ZrwaSEk738%lrg>^{;?_e zTf%Mq$5)MINbP<2_qRVk)xJL5HUR!u_bb%iXHWXu=eX^SJ^4eQJ?Y}FhY$C^p8Jz`T+R9d*>1A zO2a<*{o}u!8oxdKA5J{T5C3MykI>#ff3-L98|d594=>P<^s{&-sUPMz`oY+8-5pJ?uCOXb5IxURH4U+BF1;CJlPzU}Na7SC@tu{~tM zm3AM$oz4fi*TtDx^WJj|&f@TSoII#EhN<7(V%hJjA8629dq3{J>*CJz^N+q5A?FaL z^|R5k-!br{WRRZ+7pCAj5`M{%L2<!;m#=!9{;>P-_Kd8epxJ&U((GM{J!9Fo(vOd4)B5Qp7oOA^BZ~|?#jHvE$&7xUwCw+tjvv#H;T?P|k9Xdq zKf^mbb9tw}`ue18b)@`T_zd6<+p50m`u=-XrgE44lmYCaWtU1{rt)=H-V47JTZt?5 zH(z>^7!zw6V;$r0kHLZU8C$uk%Yj>b=KAn0ejon%)cmO%a_G$L_i)y=#Zn*n=uX!4 z;*9n1)5jC^~CK2C@8tLbAPoImEr`Rm~P_5L{j0sZ=7LuqtXl!6UgAZYli zA8giT_Pu9-?flcgw!#nFepg@YH~PZR>xvAx{Iqb29Jo(>5sD;=(p-Ly!gu$Y)Sg+hqHpo_RKTD_UdV18|Q~@uZuHuwkIdXWx(af z*)s#cWn`_Ri&8QIT{Mz@{B-Y^**{A4lcf88^9zl)gBkOyQ~W?O2sxrUi-rN_sNaf) zIC3?7Bk`hPj_i<}t-jIoVVxYkY?R-$BpRCB`cduAa?Yzx`I`=2KAW^p-n|DIGzr)mG!{`R$}X&)J9KX;n;W&ZZu z`6Qj~!7SU(j9)Op-hP_)pY^x59No~bJv`iSn)ZeM_64VD&w8&vP5X2G?IWjw-{wig^}{N!b*0``hFB(E2qQ_V{b&KFD`pcv+{k{!J=F3j z4z@01Pa?6sd7PTs3*LSP-p-|uA;_3v)TDst%=1s0 z%-L=!e0313HT4MBTjtE3eZ+B1JumUs4(5`@-1e%bhPoHcOXtN@a%PD+FJ(OA6XUoF zS*Ej`sQK8H3odi`PTaF)@p2RQY*}*aY}#typYGw}9wg?XGqz&Xpna9v^)=vAX92Ba zoJHEJN!&ZYUM018XqfdVYqU#e@92E9ZNQ^D%XHsZw`wX44h2^({i)v3y$@#!cmOuR zr#+x)xXbt+!~P7ctnQ7_=dtgxHxwDH+VBW#p%`x|wNvj=6Bz*p<)F-+qm!rsV!d93 z9wQlR(J@MO8J9G#7+^hAcl{&Uc>ZKlMt!2r`Klq_eHpYUd-zkIKhdDtqA<@7UUH&e zZLK%Q%sL~DKi3_<(;e?bD)5IB#P0G!M%~_F=DJ+xoR;H^{WdfQ9me4;!6KNtRyq6D=CDUhQJ4&*hD-G8E?n{m78U-@#y`{Jy-x8IC4XesQt)P8|J zqcd5{`If;;h_$g;n@z0Eu6wzM>~-j{j@xl_^5+Y;!5HJ;WnUv-1l z$WhEopXEPB_M?ZeQ=&D@>2+wN0$Mb6a~IiziCsw_C`Xdb-h61c-phA)?RR|0Jo?h- zarBkTnrq)CzU67@E^hVy^WxsGznOdF;RECU__Jf%sh8v2>katmBfF5f$mVBjzGHnm zLJg#7)Z?WpzQZH+;-%|TbWnd9UNY~Ue+D|Jbm?Fq-iR>n$L#CFr$|I@hc@}%eG+;RNA_BY)!c_yo68vh$eXp?a~PQ5!8643hSjym$LfS+hVNEnv1802F=eXKL!)o@L6%sRuJ`P+P z`A>ezdi2wF;n_M~m z2Ez;FFRNa;Cb5QmV0qiwb_MxXJzMz5m#A|HCH{?d>Z=*q6am&sU=0AP)=qLPx&_$? z%$?f}Y%SKjEA0U1Ah7oi*ql4k7yb#@@Ou-{QY&%dc5=1BP{NW9-^*C&ThT&a&H3Q! zQ~V#rmLJYf30-J+JjvM@+S~P0EAbM%v3BYg?2czlZu3ax3hshH#_0akctwsqZx3f0 zmZG1T(>ANroJUrO4(o*WIp?Zr$Cxti@g7ZW`bAAoARD`SsK+*Q9?h84w(?!?cID6K zov-ok+5A1HsjG))E1(DDRo9!L*4+E3!BZI<^S*Nyd5#F9YejO7}d98rFFeF>J-7sw7jc zb?&1@-_^VNuKjfS?oai-RDHW=neLG;J@m=@j-oH3$b+%SgF(oHama%wkOy1I;ftoE zU$!C-N|6Uu;H?81E9QBDIe+9r`#Q#!Op*-X_i4JDQS$BJv&bzzOTJ0AbwEGs;Lm%; zVDmXLF)7D7!KL!#zoU(0n7(TtX2B3Ev7YzG4?~{8A35*_XV}$m3k|W?6}bB}Su@YB zN!o6%o#3xaSVCRGVz(}#09x?s5+XOlC+LDv=dst9x#^4q&HF`oY7Mp|b_n-rO}Aqk zc<*84fZ<#24yees?>`s3W2Y2SvmhTm`EDKWuIIm5`^$(&X4^I1yVNcS7psBaljYm6 z8yc9q>S5CMpfWd|bvKaB*g4v<8#|xsZ^wK4^&a{tJzxDA^Y+J6bp}sGF8h6lu-9@J zcSerj{?JjO#I7vYr||0gS&O9qfUUSAV@#c2<>j*V*Rxlh{gkQm6LnT%yaHO7gPxs4 zpX0$%F*VAP>C9EWLRfY3uKv}Y=14L0IUc<@iSxBCbZy!?C#DfWo;h}*``r)c2*w%Z z;OK`L@XFT_j^1+N_TZhEaUWg;__SxSk@GQw!#R8A>yI4BzIL-Ydj=k!hmTDB_zRqI z?d&IzFXWvO^3%=Mqix7$+5g}tV~?T>H+D>0aKT$yPds*D&GXxju`gkgWRpWjGwC<0 zC&Zd@XY|PO-X9#;-1~%iN3znq!<}z>XA^CbxN!RnF}A6V0S7u0MrZo0=2=Ago7x!3 zHP2todQ|uiG>5={X^69Vm|FlnQ=aqSfi(|3eJty*_8oh1-MiA2=Kgl+oAS>*aNxyn z^4sr!ajg8I-<#hCf+K0WyKtq^GPE=;V{JTt=V;Km@A|qT(d@{i&mLV4?c9otSjF1@ z2|Gr6p1eEqu3?{>Y?=VLYcw{@INOuGhK7R9877e>PX8K5bm*<8?nM7C>lk0c-swB> zJ!`d3#5&5J`<8ueIriG&_dM1CEf1o`b)Pla-p2P%=wb0K=%gF1#5Ks!wvn8{0xd_O z~05^eETWBBctof+dgMvK$4@b z|1~W>A2?=!ubB=kvw=m=ih)JX*^6O*8!K4mFurS7OSgFIqBS}>`b5Kdqc|sYiVFuZ zw}%7=`l8X`WV(Gn-{B*2gHGOBa47a07jAE4{jLWuQ-J4Y;9+d-|C$Iq`b}%4--TwgT_BR+m8o(Y6C2P-;dN@X3(!(E-m)2*k zh43B~P0w*{QIF=CuQzv(CLiX!m|E_aJ`*ke?iy^|3VcS^Hj^IXjGIZ14&0^o)xZrc zwG_5lhL$AXHCMq9aD5!lo)DjC9{V*9KAU-Htu&8{^m&|-ufcOthq=1J)1f1uJiQM6 zPw%{iI_JEFiRKLD)O`on41>QXQFA?tGmdgY3Ef%t#Ua`kZY8cC!QOD_y|0Edw`_YSaplWOWz*Zr#@z=OQV$EJ*cldOVJ$>yLn{cpA;8YCa` z**jE-95cW3zT!*;r#9Yu;_{TeuH3BuyVn;sY6(0$y_CKj-RZ;);k|a%TpO9loq46- zPc&pc`y9{IkK#q!n-dK`%g~?8&>J5QOVvA+gNF+669O-+A%4Qub_ckRNiUCv9=QY4 z$e}vRe*G;M9#>CjjE5NGFk^W7**V`gb$_w?En^I=>+w^_3_jP&A7TwoKpzF@GiQ#e zI*mK%^K|E1E;IWDQ~74oKk!HMtwcF%87szKVhq9lj>a2}Z-k8~eQ12O&^g9e%f?oa zpQbfv1RisqcRu}K%Sgu>m?|9Ie>xkr4V|8}QL%$=<*aP&y;l9K&h8X`u3;{3`0#U` z!GFr$68y4Z432=koVm#EI{PZ(gyWG*z@t0ICL`w@JUK9_@5|}iJD=6V|7-Zke)};h z8ivjdO+dSfnK*a-I`@+XsOi*MDMi|+@^kd7^e(#cTf|MF!Cgt)z{Q2(X=r#Bab7=; zB7WWlwh6!%VQx{@Lb2;#fdk9;PyNcS@*a8cA0{5)r%&M~K&(3J&mpg4!(l8QsbrQ^21qqk!Z6yi&uH=^qqb1rLi zEAbd7MqFQ*or?~+Yv-C8&Um{TdcIuqbzlp#cB!-Tg3V0@oDn&doYVs6_v%@kr3Q^? zjn{D=khgbSv>;oqGQ>KQPx$H4EB&Robp$&rdO7FGpZ{?~ zh(5ER2i9EwpXH3p+VS1ded<^70@>7u$vGHb@e$^+Lh&co=V2>xWet20 zw-OTRHSwt7TWCl%s)fj@!d=ts zqjw_%c}HjA;g9xA3=#8T4c4v<_pXITIENsy&Cry4mad7vn>)CbC z*tWJ}JIdS*K2|w6+{*hW;LTcaDgPn5k=!o(iWV-PW$t8%qPO+E`qo*AumACQL+vPh zzBwV@19#A$&MZ8bZT(Xv?;W%Xj^2VDw`B}+vw*QbYwglKlzOpO(jUC`_8}sYq?7&0VD!S`60DOZ2p*y;F z+-uGbTn%0Z4GxtYyeV*7CBN+^Zofy*0>pEP*^eh?uXt`P&pOem6QJ9?Uj%I@mJ!-H z1(~w%=h<(7Ut&vk)^eT@4GcqjjY%@NvJ%sXx15nkK1optVR(WZ2_23a~DDy>D0K=0Hj3^}zHFJ}EFoJsP;V zjOWCe-U%giK3oTMqF8GVGB^%>#-P($-rwZDAA@G%&|}AT(X>-9aH2TJ7EMccDh8tX z$6ul|rI+ILAsgf`X&(dLDDbMEfou(L&$y>+qIV=3)*wU5wO{lVVBXJv<_G@M^CGXa zp5nnbkvZ5O)}1r$J>1U`E5=4}b2gV8JJHYyPK1lb{~zGuzl%%B!47PZ2z(kLR+deS zKZh9qSYrI!i1Dw*o)sPHEWh~U&`AY39OyKkI9VIEashmE0-N3f7x3GRScTprA4r=h zp9`U<2;U1N3tf3$Yx4WKcE$PFd~dUsA0Q`grY;1WhpF!$GpP1}Xne^;WYF8P86CYQ z-MtCD_Dkvz6pN~0Y{iX2$Yo-kE#(&H!t?(B3!06w=914^_YiY&;;(~DY!&{T7rz`> zh&lN5yNNR?cGd*EUVLc`a!xiW=d~I76(9U}c*)5FBJ;^3loNltg!|eY`Cv7dBA2CS zDu+}U{Zlop0)4KtME_yuO*foq_y^=bvo-ksFTQ{0n_n!n>>Ra?L`&v3de)sGga<2S!?r_O!E@Dh?eAQRw7X*LV^b|0K!QT?r zIn#ay7xK;J)4ZU)-nqOtB$W8MyN}Lt|9SqIQ*V^=UroM!26heYvFy>6?c9PmMw_@yF+PsK-C=WUIX^3^J z!FHFduHv@>zOUlD;;$9iPL0iV=#uCNa5Wluct`(3{4eA8rKj+DM)6y|mGo!oMqL-#j5K5W z_Q4b~#z66vqi+0c8F!d}JS-JIQ#>^~-%1o=D@qr=Og7!@VZ9lgu!hyhBl#H0uZhp( z^I)sC7oOry3TicG-|5!H;HyT!TaS29Kb_{TanrFkFwXIdA! zLvf9-(oUacU#?%ds8@9>`*|*0{77<#@1Fd4a$8D%=-xk1cIdC<$I)Q6saLyhEb)=x zptO3m$Sdd*=(r5IVPwb)PR?;JGD9?8C)u%rv2K-*$9fog2{eC?vzz2cdvqc?ypeVL z1G(SFeP?of5neO(OrnEYaLezz?@l;(s`2?I{ML>Qj7=U_}( zp1;UibiIPE$7Mti?Kv|qGg>OsrTh)*7AOYaiYK{-~IQr(RZ!2 zo)x+63TS8OGC0WOGdrMZ(PYPhvXWveyVt>r#F z<{Bji)v-#l+Bt97qha}GA6-HWfY^@izb+nXO_$8m{{q$JAS0s4h#QED#Gpsjf8r++ zZz3OdEBD>gHghfCh9F^~4QA3`yw0J&`ShjFihoG%NNx}hGBKJ7tYziW3B8BG<-{DzK6E?tGxH6axvQR6 z=LVY?MvX-cY;I5Uhkf(ntVix0;oiZ}o@yAQ^cw@$o#HFms_?JiqGphs-DLaSap+#) z?!HlSFSm!d&Ad2krhL1Hn+V@kFDyB@7Ch{hZY41el`5N<+Y?Ds@@>WN#y^g&mS*CxUJ#?Lkx2N=R zwAe~4m|z|4!7r{fIY93JVa>|i`~0d@S5!I0$*DQ^LA^y>)UTsG8cur#EdJb>;JVR; zOZHl982BC$O6WXNKYWMH+2j4~HR+Qaa8BJkICo+tQR?#|z#p5$SotPq(jNX$cgu42 zBJ;tPZO^<*cIdI=4V5{>PiBM?pJx3+LF9w$i$wB)Ym#$MtJ#;N_2s+qMKs1}OYvuP zd9GamjuhXI@_THw@kQFhx8&I!!|vWGe9RIZj8FO^M(1zB7uiW|g4WINi(qphKfXXd z9X>2}dA{8y97@1*EK_GSQF zONSd@3?4D_AI4lqfNOB{Rd}EryoKbWv5ryJF>~#;-fy~Nd-dPiA9UE%|EKbK!l9}6 zM;{%(3%lzg>*xc}Xz2yceNF-DDyt7Ycp&!8pP0JJDsrWPywDv*KY8E)vVVCccF7mu zEwx+CUX0B5qx|lFze?XP2$ht+YTZ_gKYF>^P9^sV9%!2=+VRkVHA8=LEcT^;KUVam z4@?cp9^Rvm)}jX8C=sk)OQ zWJLLIOH9n;%CYDNXjS$?DR*eLW%vHGbj=Xs`)BL^@SY0%9(4GIYAdjx+Q9p)wkZ2?p3aoSmbmIIjMe^f!|F_E3Yc$&JLY@ znT8wsJM;H+T;2(vKEi*&J`i7BGyq=v-TxGH>Wz@k_$R4)qX052KwP`9HdL}l^$<(2 z$wqRY!hGKLS*t~Bj`*`?n_W0mm&W5(Gfrij))K%w`8|6Pbs%sAISRL*jID7!YX@$1&v+N~W$dgQ zk&&PA_5?XKOn=o<9?!t9ecvF zsT2o+$I|Y1!7trVhTTJaz2so6b(_`KH&6H1wy`(n-~{+tf2Ufvh4>p69_wU>D)%+0 zxr%n<=r4Ea;;Z^-TS?nK+7{?9Z6mblqirL9wdUpM^K#C`k&gE6B|1Llq&Z`izUE0E zjiLTaI;LvHEX7OOt8bqLt-5rE4c8W2@D_A=bmCt*`@nQ>n&zeRmAR`qd2h{H=Cuy^ z?)e1s(tJ9Z&m!J`f_~P2g89hTP<<@+LyPBM4W~XuHW#=wxeD2x*kdN|uJg!?wC5Rn zM)Q}?_7rvX*!sp69tsT1x20;8nRiQdP5*go?a19H?-=6Dp$)5n8M~}xqq}w+-L>23 zuH8m=?VL5sHfyGQxw*s0U9;Wpn(cPiY`42+W!M5HwzC{M6rHcK5^EIWs~l&)f9c7F z(4gLbj;sp710nb%v~vgN=x)j8oNDE`k$K_|i`q+_oilNGS`M(qjU!I$SdQ$$F6*ju z;w*cq8PGG~QP0Q11Nv>;5c*r;;M4d`#9DIjndG}<*IFftjb`5f?yuz;a;+2oivrVm z9$q)Nczw7(UJc&s+~3#t{~f$j`yS|df0{LK9OtKPA0F7SgF5BI&>lrBBdJ#~bnp%A zNP@6aRSz=(_?vm>QsCdQbX@P@kNV90f7-j!1s@q1Nd7pAee9=;@z6vJ-y@Ff`H$QC z8eSxxtK4d|0vY1YOZSeb?q7MF$nE@w{E=+G7W{`V$v9E`LFI%8o#LK3v&?6Fa7GC~?6V^o>3cE(GuPvmZW? zi_M@sRUtOA;JubU1@9(sw;h|NkNu=&d^WI8O^1D?2m2K+>{q}e$a&+x%CCLu0(%oS ze^H;#RiYL{&k8Oy&tmhjKW2myqxmfi4-5AEQ`t@4I;>+Iawdlo$Eh(<{xg3v^aL%N zy8y$Cv&5kx(NRHwoWXZO-_Oa|8?X%-*vY#c3&`{Ee#Zj%A{>H`z|$D`XM!^y#T?c$ zUN+;s!+7F7jpxT>DR|s|*#if1pV?;cwP(0B?+NfV{=Ekd2!CaK-v+*RKgvC6!>oDd zS8`t}vOL7xR42JxGF!C5yrILA-EG!wRoEngisf31Fws)i%olQ)It8#-(Eg`a*n{@ zU#DLxujuhY#&7@5Z@VWmE-<+~kjMoebEq93|KPtL8~^_MTH7AM#eSYUUcn#G+G<|g zCa`Apf!hLnPWE4rmBwrTO#gy2>np6^JSV@Ma}%=tMtt|V{6!ur2b-7U#4lsOaR^!1 zRVSWza^DN7gO=}pu}utu<&fLrEiYd_mVOk^+_7|_lP~A)z$6}tBh>v5&gI?uf#;I_KQxzPJI*|p zo0vQ?8?c5pvH@#o6T$YYp-m%yrwga(Df#4gFn90sclj(i z=J^D&c`s~C`PkR^_QxEpaO!&xn>`@u_o=C0{HFHu9nKlr5ycP{=j9%gEOhl6WU%7x z1tzvW&eYo9$l63I;2~&dptf()Hi$hZ+cYvIWOz-wb~y95s3SM`kn>%Bg0bBzoS1(n zI%pgBY<400UV;vlGjZy>v`;>#c_Z>p`yP#7I1~B@F2#-Gz*5P0Rg9PSQC|b`4Pa*6 z%h4AzsaHmZb@CaR#96WqADMU;3m+Xu&y~7-RO<3ksmn*DK0Z2x4ai#W8Z3TXy6h_P zp7=<-Cq5GIiI2p4;-mir-(7QwWfH&chOUgv1CO#%UQIZ5ru^T|q@Afabe>~py7+sL zeQcVWbP+mdR~)*Dg)K8*zdc&#zBiutBJhZjkxSqOV8AZ^@B1UnTQvDe{1NQ~`yUyT z`F_$djo*X+@pD-%f6GJ~)EsWB4%UObmRZwL`4w9qZve z=vsb6B|5zm{*=!rnWgj7Bj~VM-EQrszYp4V>bwB@ar{(b6ZniV?6^4m>|c{r;K!>I zl0P;}_pPcPWPo?S#k-!57OM%Acz)XMv+>h_Kj8W_^3zt6Taew>b+hAZ|Lu148Q*1> z>X}!U*NHsR{v7QA?W`#)$-ys^95k{D+ELEZ$SUZ^$SV0SeD=$#e0bN$o?H9Ns;d|` zQ&wrd!m~HW&oM9WvmZC?-Ad*uO)M83Q~TqTj!_)99sMLZ=Kb^H`}YxB`*>I?ey*H! zEUx-%=lp>J+4YTQlf$}uCp1y7bInSKD^{syeLQoVM9f6B^&L(;?+eJTO6q(f@v@Re zVr|lcg}N`dja)hFmaKs=@uIBe<^s2Vz0QqaPlCqAr^l|1Jo0|eg>GJs z=g9llu=zh8c3x86$7+z5>yh`!{U~vR2tKXiYI)GSbmGhCFXOwef>*E!m17c*VDt7| zMr^MW9G4-3E0-?1>Vz9F4}F|FBSu@jTfMaUs=mHF&OIO8gw5E#wzk0Tm}4c%FS_f^ z*`ufltu3)PTUi@o*K2I>#WpP1+ob%I}V|_XIx-0J)qy8y| z?%Aw$3k|hn_>I~ZFTX3coOu;z*>U9M3##o|;q04_Z^^RP{}XF#_ULdw5%X2fZTdsx z9o^5$C7PV1$vtMf=cPD0NO6s|;KhjxBhL*!0zKzoCx}<$vJ)y8U$Pn;gqBua6@hn8 zfWL|$xFUA|El$^-I>U3w2ao3>6%Oy}ulUWwSp~3378-tCI_4_f$EUo0ls@#X?yqxX zRFJjLZ(fBh;m}$tMrmkm81Ri?%}4PUnrr$|I&Z~Cu)d?HKXqwO^yl%{C*AM7@m5AU-5_jYJK%N%sML0e~!jk8op{BytWPg+6}M8;46*OsK4-8rT)Ta zRp@rdCU*K!tW#^HzkYl<`Nq`#st7puU??}^X7x~qJTC~3)DYv7%&qAS+%^$g!s}n} zJL_nfnCd?eht>T2&x9|nt#CWdhuhz|Yv`@tb&TcTuOpjQJ|Z@9dTzzPUeJpA&G+>y z87w-<+|P6$@tu!{rOvlf&5Pph9b2+WluM3@myt^kR9lZ05PytpDK81>FFI3peQa(h z@ue`jaE^F{F?hc4PR?Jdb@nlABc{2H{Z6ZguY0WHLe^vL#8C-1EZ z4qASQ{EGCV_uC7n7MSPQ9P_rH8h!8M)B)<9PUOIS^mlkXIS$1Ku^(Gs!RBSG)#Mnd ziNG%@vSYv>2mZ~}cp6(8nW}hsXR+{=->e*;U@V{>!;h;h;mXC)A{R$bfFtdnUakGp zE{-;XBh~PFcmkG3Wm9UNVdl9PJ0**lj{KH|R_yt;$Ya%~B%d)qJzL1!h$lWPA80E+ zNn||hhE4RfSCJu8uzScE3(wN^CrJ?{TpZ^mA&8^=RYAxzm+13_cnX z{4@M``4{L*zvVw_m1NE{$UNnbth3_JZzJcXc_yDRPd!`b&QpGuGsm3fV$G302c)A( z13#~3|Bu$s@TyC5{~dVN`QYh;j);fXSAA&W;q}$WeerOQzBBl^gY&G#%aeaNfXwIa zUd7$~@~eqhtbc!^_9k|q)1=>G(5Q4>4r^}cxR$*g_*>f7uRYICu43=c+13nn>=CQh z*)M-TIm`$&-Dv9jO6{R%nVP`*YHZids*3@?T6ftUlHbkLOz4@(G4R~v82HWP7)HzQ z)c6+r3b1FfgJsX^9J_7AGf&>ed7A!nROm;#z@|^xrs`Mknpie{toF?#27Sn8znhvz z?N?JjioGrb#^`9LpBnnn=R*3*LRNY8UPbVA40@L{y3bWKfL>n%Z8An9foZ)0^f)>q1;enScXm!-Tggd zB=0*e4%i#B&h0tVea90oPS+RD!-!pgzt!N8eaEK0^99!Z30D^^K}P1H3v{;IA#}k> zbiu1e7vwHK?CJvVw|9&#AQp=*c-7Sfdgnf5R3rVB>~rM*I$+-9r@6VT|IMMq#pnRR ztej3x7W<|&H~4N7bafAKcfoT>z2UYQNu8|phIB?3ydaq~nmMoG*?i_v;m%_b^FXFe z&oTREI-lRieg*M_`3~IrzSxz0PatbTv{U}L!-csPcyD|ExSRWS^eXiS)b;1|kn@hkE2e*k3XT7AjyxsYxoh`);zRM;^n;4wEeNat-GdX9qA_% zM;nkwSv*(VYBewmU$VKYvHMgXtz4GzZ>d*a%bhDrxT8w;xm%lS=t($P?T)#H_Npz% zHpGUq(0NX6ZZY<$=x-79w+Q-+F%RK*1?}{?2zjxNy?>hLnw7=9apX`HusiUMweMs9 zjo=IdXBBWNrkR8jnX7t`lwWX({dvJTqd%NCyKru%z2t}BbYg`LoJGL7aCp`RgBLBWHgGRoTGT5VU5actojiHuv4Q2uV4pl0Z1_8Mzjm|& zxd^@Hxb*rzVS}j_G%gzyxzP#l#O9#47-tY;bQN2PSB+gZsCn+oz>6%6;!DMt8@V0u z<;8-to0;bk*=i*tTpJCZG&UNtUi_8=uDX!DZL(ij(@x~J_~Kn^qEyEt`$e`()n_@! z2-_v!vA^0q`^(rM&__SJM9(tp5@QFT52bHE65NcRW|!=B{H_k>qrJukcIr`M%uTlS zOVC)9bGO9|)-~XPYuQgs$Uypa__z}j5@;OsqSDJGnKWF#kdV3Ad zG5XoV!5&~%yV&*M4*Yf8=W=nim#Jf9kIY2ka&g8HllXQGE|kkzIR z6Ix~-yUM|L(jIkbBn}+-S_cNAAMk4Y%1m3~hd&sYx6jJ3AOE}dLN()jQhOn}AI8X( z^!+f3|0qTz{@a9qR2Vs*_+57M-IMU2kbeuIxyO+^>m0e${y!ymI*>QoUwV&Z4YHeD z!wj?c!RVVo&C_3UWe)Pf*qp$k*pl|1$_6Ih@vMB>9Q+vhi?IdJVA$nnWYMRn!%gaa z_(yAdDSTDs^4t=5ZaehxADf)L^GjHN?Fp%tz3Z0?ZoePXU%o4s7DA6NdGrW8{ro~b zgC6~U;abrmaLXrro4G}qPXV;J5WC;lpmpgs!@I7{pj@_W2FiUdDEC-%Z*M;KH*V5`RB%+lT#aGi(PVPvn=oc)8Qn$phLBb3|83+kyDz zH2W*}p_Acx*$$E;vPGW&H{W*0Jd^F9_mj2*^UypO(oS->pY6cBWji?Vj7W@W;8tqbQ?_`>AwGi(RN;1-du*IH4BXL9ywaDz9&Jza3~S#U2R zf3NkaCV#K>BX4$E+hHyH=_9$&(GX~U7&Jcuz8FQHKKXLgjb}HFl8=yzXZI~9R!po_ zc0|zSXVG#t_PcZ5duZq5*zb#s{hqzN%eCLV-&R4>HNk9Sze~<4mZNtR7muz&hKrZ6 z8RGaK!sqP0zeD2ye+B*ZBbQ4?Mb+|Gght*u!3* zfOG+JNP1PWNM{F#$Gd@7Hmes4B?dS$ra91;pQbM6oFw7%#PU?^+TmpY8epo# zXBEAw9(x)2A+=X+2{Ev#dF0aXbNk3ApB^R;)P-!0Bdb+c+yO2mcRjqCS}^!<dctPKTuDwBg zdaCHP5}J2tz==rM-%`_nhjR_lIKcbIAgyE^KM=WyBMlcyxaA zEu0Tz`VS6}KNvw@jJL`+Z%;;K#-j&Q5BTx0)c%fabY2Iw=ZYsvmX9U2Jg(NMdE=~! z`flQ;+lH@ytQ)*^6I*_f&%2=;Y$xr3fOfN+Ki_%0ftdA-NIt%`&L#P<%=`@>FXOBv z>}eD8%WhuYn$Vd&Gt4_D-FN!R5*@Uw<-OIsSIs-n=XC9HDFU7X#wg>r0`f!3Q%A_> z89ccAfm61dgZDrA9eVTf0+;*n^}KH_{OfV}>~wtPTf>jOPmV92b5-7lck;7CN8g99 z^YcPS-(&4+se`s6pSItF9?DI=ZkpNSAD~WeCiEqHD;iwEU7T*sjXMw1u6DHjyJWlb zOuIddm1_6xf!l>XZMQq^IHO{yX}22KQtiGoaJ$9}?H7~dO)>4tm`keNKMdT?y3pR0 z952tAA9GK&`)+AH#+&d{*3>|dW?--hq|FEhTsk@5W(8Q)*yyJP?|$k6@&$DU#7 z#z%Z>=*8!*GyBI<=j|z8TdFg0kiVs@CG^POf}_G0F?h~hpMo81HJkMSKVM{?Meshq zwG`IorG5hk75t`{QYkc0ss5k=)8FvW(ExaK`a3VxhV`9J8`VxYzfDX12F+Z~Z@I~_ z$lH2wu)Y)c&Cq>?!~cDK!6wntV@tXB%D^_B`{d9i#rT@kBv}`5e|7F%J7aO_BJEiS zHjjVrM8ng(AEn<)vmf&k`*(Mx*ROf?1xsC=diKUsKHPfz;nlM_cRZ!)*`nwP@jbTh zt_b?gt2u%1+p8X#CE3tk-QbJOS2#9s3>_%{ee-Z@R}SrW!|z|BHrLrF!FO_>QP%5a zWI`@B!)|Qk@OkK0Y+TmEuW&CyF8)UzbIuD4 z+MkC%oCl28@!nWqf62<;kRJecXj!q{J^WS*?5qFSH}CKi;DM$)v3WWsab^#{zsOud zoRJp?4%tK1d?&v5hU}3z``uKZ79EfM$-CF^Tyx&X`Gfni@A|&togqUT6QDoc2c)xm zRNp$N`IXzTjp=g}eJTf3${wrhuy16$OXsi6&fc*0$X#!)t+d{_hn&&DQ&yAoscijw z(36X?cfUgk1Aq)+Mp}|GWM`xW$R@5yGn>f7uYtfzXrZ~Q1OL#B_nzJ~QSb466 zyt9b5+pyOc)9*%bQVq@`te>fSh1Vjqck+9VuRF`^-JW3oXdQe)AFdB+=>G}m|2gUf zG{q1R_hDe+r>S_(nsiAvnT3m@()pCBdgv{_{?%}>DN#5 zL#;nc9G8BoFRUHo*H87<&-3jlXLb7ZQ_Z^>H8Ml2R`J}+f_X@&^=+RH^623{WU@aW zE4}z5#tcwLP-u~Vg&(973#f&%UOmfRoiS$hX~x{)j@jjosqwscTITu8@;l{&QfJv( zEgrWWUCq>mOV6j9dx{Pv0rEqr-Gx zJbk<8mR*q1cZ9Wx4zpTg$jA6F{s7g%G!IuLz1`{7gaTR0MpeGNFOp=(o*2mG$xRd22AZ+E@s!BOGD;p4}l zf&trAcGf`t;OWM{df@R}tXHSlW$fGX^HtLs#{c~zf7$pK4Lp9k^;@04d3T!KdEXhv z|M$Oe{8Nki;~zR`v1%{13&j^cyLZ_c#((-xjh`tG!+{bb54n5Vk%d>=!BC*QmHj=m|OX1at}dkHbc zR$_{+AB8Quug>wOWy|PaQr==e`r`+YIAe7)Ru5y9GFAnBnsYa=B7A|?7~IR%QommwGN$^mzDIR8Dw7?%|M5qulLlU z6P%BuJLu5abC56CJIe2TTzIk}-ou$mx|fjW0q|4UjvvaIxySjp!2N z54}r2?d9)Hv*WG|@Y^)1u?SL=T2o-%_TmZl_cNys##~p>_E^};xj)Qr@sIna zMy3Qxs6Vjd=#U!x$6sLobE`8i~>h&N?A!k!HbXWT-q!Iz-BfoQdJn4^2|u3o{kwB5by$tai^up|$h_+LYvhBgt34mks|_mZv19aGrnvg>n1pL9?GHAPN22}{Y2ht=enz{R>AD$ zKvmnM`4`h(ZT0;La5j!Uwtme@Yy`%weD8wK?pd2vqIJ#&Kj~|q-8@745NTV>yS32D z8gf0AM$VmU-y-B!eqk1ZGBWPW!xXC&AL%QoFN_-_`>VgYQ z-0dr?nd2zr3iGPI!lI3X3+29%9B_#2(ESy0-dl)XR{X}uN%WP@ z(NnDE5cqNOthwlgJoEzhRZKh4P&HCA(6Q05@9V3$2c9)P9vC9K?1Y1_N8Tr0v z!^nkjXsrDWzDLo~G1~RaL{|eNwS%qiF@EhgsZG6*cpbX@5cG9;rp~#Ro*dE~;!aui zsqR9?)vv9!>;=T*0_cG%a43Jc8n|O~EyfD9#)-d#j2z6j*MCgjqyj#M-aUTAW}FtC z%lnL}I_V^C<0sbwwqC)BWZRji%%bOeXGdxkum zm9m~Z_v#7!xO3zqH4orfxcC3&*ezb2W%_SMAGd&kak;`aW&ABT3Uc5}A^ zb1l-D-_T#-AJX$_huOy}+vcQiPpK3Ca?b0zu0Z#-r=QmqeH=R2NGui_Rt+QZx1`Ka zoN^`Wp!Lh)H_2Vq>F2C1FPVKYdE*A`P3R&HT@>=WVu^;QE07IykPX)(8;}DX@Pc$S zH3zMV>&37WqYco-RiV}w&$Fh+@hNLK52^}VHI7eNV`VL`BtPWsk&N){|1j4w{Z3(|Ew{(BQ5$i~{`?$x5KSccljmpTPjvV);6h(CSIVY=Kh1j9Wt+Wkk_QH7@PfhF2yiwEoQ(!& z*MqYg!BNGmUiMS9M)&6Q#*nXlediEo<8Gn8Vf-K7Bwu_p{8(;rF2k*?0kMgSQC5AJ z*hC>db#w)|0C#iz__>{DtH9muWhGl^i(ItvbjVjt^Sl)s8$R9nNRmtJMH23g-2cq-y^8eh6=rTx0X<9~hVqrsb*<9y_y z@bDGxdROdfi|EYB+5Sv;K(9%^d3eYX9%c-H2Xu{xheqLn?|%*+UI(tz(b+)rK|d0= zWPOJ@w*EWp$MyR}A7;o8#eXNlkN$XlH)|ZNgI?D|uNydn8T#y8Zne%n$C{dt{acG~ z(lwL&44J>~HPrhdbG~w^y_fs(y5P5NXyOojaG3daOd&?ToO^W9b)2!*+OeUmWNr2Z z)32o;$ZM(>hu)Yu)wDDnf@J^P`pK!CF_1A(5s zJQICzG|&J-9j9u#R|FS^t*nrV1sUtSR(`W9%70f zY%yqyvt(M!6|>2nWGnt&WbP>;e{5oRC3bYLW%qD4ZptoD9f8j-NFRgyelN4Dd}H*{ zzi*uS@;j+7W}Is2+2ZtZ-uMLLs5an9a3Gp6bXP&lZVoZK>p4G9am~?IYZvR$&AJ>K z4ZSdy>|E_%H#Av6ykHJ-yz5yEIk{WWzA6RWCt`4U({FbH-3D5)$6VwA9UMPRzW%QH%zFXr1MSInSFEixV6~1t*D4vC*Y?%1e{DUlI8@HH)J)u9Y!0)}NH)iO zYFH-1GdaK_n`7`SEAa$AZvI5;XqIxhz)@!W5y8QkBj_=~aT#!!T4lFC)h{T2JBR*c zOFlt=^0gc3&x6$ur~K{NE==RcS|y6@DDJa~9Bv+MRfkcn7!R}*!3R-osyFsf#`gM@ zUwTM9&U=cz9h#{yFXv*vM+JC9oCA(&#&hT?c0EIrf)sekpS{a_v(=z3Fk5M^z2p3sFEa442k=i)BR z_`E+&wfq|YkL36DS+-yu<7>~%8pikJrQsWw_LO_?Bk%r;_xk3&%2^uJh)y;1_@+yb zZ)eaWeSH!>`|N|ROc-%yJ{v!PJZLg!g{0~yyO9IjiCGfMMSl%Je-WRgrkMO^SxFJ+ zHDnV%mmCf$H}#&ASM}tuH)s5=cKKaF&QS?-2amx4^0zIx;H}wbS<|Y&{J#&BfBD&C zz zhg#(WW}z1qqm`a9c7&sA>b>}yY7(?o8?;t@)>_G~*IGr8v9xE-cki5w{HtP)9jih+ z^R4XVgW>TL*yrWJkdu?GJzm~R(3exMn9Ai5b0~@Da%K{;I*Tshi{=uy2yPW-{GxTQX%4!t>Ouu~_d+PH1m zZC~BSvvJE7Kfl`wF0VS-H?Iv@97Tp2yt%nUkG|8NIrl)dL0hW7kXng z%=29H{24xnOq}2n+r(Mh#@jnr_-ue;>kjG7&wDs84&Cvzi&IZO4mRh@r}$O)+>%F3 z0=*k8My7$+9Pk>4=BmMM1RX10`BOOl#%XXIc5%G-6X3YQ#qkz!9D}clpzk1fW`D%~ zIA@?Yg6HC8;pfBPR(GUH?u(ZiPxLu9hbtq}@SJ?+;CZnR&+~nFuJqyg%LCx~D}Fp< z;|tGm_SJg!{SF`PTKx0y?VP2Qs_#%eOSA$!&p{8P2lKJt1|j!%lT+G^4%E6kwt!+4 zDO=zM&KZzztN5MndsYky`%N}QNq6w?cfK;*>K%`c%|-9!xVpxxiTYpERE$Grd3t}4 z^fc=%nWp!*fzPhtoZGf+-1A}9d;vL+2g!rQk)31D4sKXX{a7xr=ApCm23z*HrHfZ? z%f@D0V6_T<**2;-t!DkYp0nJSHo5HDO)P9n? zlkgm8zxP1!ockx>QS4N9hzH9vto^kK*T&uJ#F>qZ^n5{RS@d$Vk6yeS*>^Me(}|C# z@;|bbG#{-UKK9Xa(TK5=atxjLX=JbFBe<|PhzXahE@(Wk3AzwH1;AI9i+Aa@P1tX5 zlhaqsT6qxkCG=q>KJ|Z@dl&d9t8?G~otfOY2^JJ9wn+#A7OQrP;?%ahNkFjL>OQbH zJ*6!P2?CAXmFlT_&=O1l6|EU{@3hc1;cjBB6tSAR_a=yf7P}jvr|qfzm$~Hv(F$rS zgW5UY-+JFS^JYjCJni}X`Fz66yqC3}^{nUqJZmj)9Cev-2=>V??Ed(;@n5snPYkC2 z3yY~$Mh0{St;_U`ql<%)v#D*$*2RHKh`XR?WrHcUFatS|({6DFYCr$3!KxV2PcQZ9 z!=p~k$m!@by=%!gJbrnk*yp!-vZSA$&az2PM<4#?-c#3y+k7&`lP8CR^wu~(g;6WO5c%?GwxRvUB@-gUk8^hBP|^U;dOqu08$B3UI{ zU$i2+Trk5rHr~iB`T}nW&cycY+S~ z5qhi_9*5Cxl^rI4*XRGWn7w2(^=X5}Q&tsNmx(58sNc~u#l@V@Z1i7Xe;KvkW!9Y! z%0CZrUHPaaxrch>WW93k)DU$d|0FlbOSjdU_{+tSs^icyc4VA*OP^hNG2e${Uryn3 z6rba;GqEp4^8xZiir>Z2vBg$!?JKN1^YU^>$)_t;rP%QV@NH~UAN`xyXY}OwhN0m* zMny^ors3_bUA?`}uI?YV*ltZ`ue%R=*RuybA0PP8X{Qu_IQul{f9(MDKP0QZ$mmn* ziQ;RBZC!($ohAL(U$4do)~nm=@om5TKjDWud~EcM&Is(MU)R`eM!ycb=_LKCxmt`I z-G&@p#av0h=EED4z=iZH=PtNDbzbWRcuIbut9wkpU!q@6_oyw8&%Q-l*}72h#eu~S zqj-A@?aI$TADY~D$LxdMvP19q(!u8^SmAwRtlM{AVugi^JoMsjS1%e{1G)MvHkNFS zY`Fem5V)=#0Iq>x5^v9hFD7A6Jilbp(|%uH@e{Y=H8(H%a zmrpr`cnbDNJGFtHzds#4oxs169uCp3>hDA7IN{Utv&E0HbIt>PeHTv!fnmIOih26- z+XLur#iod%`D0U(Ig42f9nH7#shJOAW{Hu|40f{OQD5@Ws^9**8vEB@A1FW6^GOHt zlZ(v07-J*XQx}%47X`oHN7Xnhp}|Vx2}#DTamLBN>$~b8l!I2BM{7_0y3@#x=Ar#` z=jrIM^F2JALJTpxK5)n>%ICr}IfEQs7f0@vK~H7Scre$xEEoGhy4sfC`v~!a(DEGv z==qGC%aT>@yGN~pVa6Bt-j&Q&3^=Y>@HFLwLM;_$E=HT!?~Uw};l_Shv-g<+W53X8 zn=^mIB2mtv*L)4A3$o>Sdx0|-{bg~7^*yKV!h_Lgi)6;<*&-i87C*syYD3SbFpmv` z(Dza^pBa1gF63b+@s9;oZjEBaZoG9`k*NbJjI0x0sHIoVs+2h_uC`1)kmwV-+}?L~ zUY)D+;g|TWR$I1Ae^xfm&B`mC*L>wvL0D+ zI@znT`(M9!~UOZgW=>hY^z7DHh7OZ+H)djh+gI} zcaQpPD!%VN%bJ>8%buL%vSL-#y^znklC{>0$Y<&E2J%_TWht+uT$XZL+E1u^F+L(| z{CYz9`wNKk6mV9;;RSYTa-d-CC%Aq=AX7uRiX4Grh@IGhC+0hGydue+ddEic$MIyt zNf#MBmOHtEBf!-|+v-!WB!^7Cd;xOp^U$tfRPN2d46GiEf>-sM+LK&9c{gj(y0J&Q zt?xWIEMT|g2lCg>KGS;n>#G0cIpr*7-`e}sO#vm-uFAK5v+{25lbnzcOSmQtN18GINyT)^5T)|&Lg2RboW06M~AhBTOAp?;YGyM=J&(1?S z{ivpE6Jz)I_Ehcji`JaFPRRH1zvrJ0De0eoQ;sWs9dct1b3YH-*%J(R?F)vx_Xop0 zPf^#L2!@Yz2g98Y*=;k?9gA?l^T@}oE@WTb^X!A&d&&+boQj$c^x!MQmto`kp*y@>n_?@X zJ1^!|OU=+EWW5)2gMU(;`^&;w18_sg$>lHLTv_Rn&IJw(r3MC*cQY_tCm85ATaPTI z?Q9>Rle5!2IIp4|`S6|i0i6l@1ShM($s%yl-CY)*RKVI}@SG2x7mv0s-&b3?L%30V zFL0(6QS59Nxa|a=veoYrjL^N4n=|n<1K*@U;42*nzWH{WiJ37T={VtUq8oGU^AU(W zb^@R0Bba{Y2{-Z)GI$v>NWUcm^*fK8sXKq_cM`G=zvK$9&jh-q5}!hSc3YtcKgZ#xK-^S-aw|8RO>D*?u?d1Wxc*pTioOu0M><7g#lZ@G)kAzM- zL?_HgUmo%;<+^E4^STOHWVhEsGwZ3*SkyZ!)dlS+KcL??BxbSpC%1Pt_VM}3aqwL* ztAEwTxO8mbRgLF=2xj0=ZHVgr6^jwf%3*dgJ}-ai)P%|>%x~>McIX|^z45Jh$Mvl` zs(@WNOx1SD=Xf62^*!6S3ccmpFK+%4UDQ#rXNGiMN9Df$xk_`7yt=uUl~x|6yo6ft#A*v`{9IzOmX5wCs=#j=d$~OooDJa8_H@YMdIkwD7vviw#Ctjkh$?tJA$ zW1O>gsF!F{orQAW&E}kC^g($bGV_Q}?_}!(^hG=Nn1LH#su;ebj;w7w{uH&rCVp>g zf5Y03M4nK+Q#rJ;)@N(M6UD4K!B-A5$8GqZZR|7FhFqwLVP^!eGqQZk4l6p$^(~#g z`*2}hX6nfMaKU(t&f~L;I`gvnLgK#YJgr;J8sqR&7^CddQ`K{~cyw_JzR#WLIgcKk z_-iJAHfX#XJ8uKD;Q6c2LWk9JR-}@8A8-6Ve-1h@`R;})&tG-(RxxkRCK=zdL;2vV zHBZ=9@oRkZG|8Epljfo}zn`D)&P9V29W>6DfyMJJy&T$U;P<@Kz%NCYefYiF%! zWGVXFtJ@)Fd%1ESr|AzGIe5qacprZ--^BL@^9MV;INw_8Bqk!C{c%3Yr1$X$P2XRo zZ=XNdxBtPdw0AoG;Nx%f_Xh`*nRTw;thpKuf4RQm)Ac?%IF>aNKHJSN?_Mrq(i2H3$4BRfVdTBZo0NQ@=pL^a%e!Bgfj%@j=hwD?2EgRi=@Z`&&{yMR|06t96tR*?gU&mMe zc%)Qo8SvL1tI6m}`Rid9o)7ZZmCw7{?C+J)cZpg3Yv&dXg$J2$f2`fal(DJDV<*NG ziyy5uRwu<}3@@y^1wI@s-afi4oL{uzz&>nj#by>@gZ7ZGEjY7j-3!><`%aG4^zr6q zY96wAbL9jpvJ3llA-0j=abn%3w#>kCE3hQ6Z~Sq8tx45dLlbM}eXS8zym>P=j^fLG zv1VxeBzp*ZNwmBYSQPuqwug2d>mO?#kbhl$ZRkz>n^hyJ`Fy_4$IDrDk>CA0(9=h= z=xK}F|EbW^(lN51GWj^!sQ%jJfoh$xM~+(7Cn66VgD=6OKZZ)2scq#*#eP^{q}UJk ziW~dsaN{W*s%2JuLva)@2Bg?e7jvyxM<%9XKiGpzQcUFnt`iIDt7)!RO*3|xiB}-M z%0kAT>#*jYZFKla@e1^@6GL$1(qZ6I9k%)uEYmY_pMG#E?&F73ai6{TMVWZacOLA; z2I!{l`hu*u53!6kf9xj*e|NfDKV3|0L~)UjTR)xW%3Uum@;mbR@=tS+E6OkC7a-^5 z%K(pRLx{y9Yx~EgPh{|L)VkwbbM~P-EuG`?Te?dMS{5PWshLc5j+@2r*)5&x zb6UDL<+hmpZqT_}@4SUN>&{8~1=T5trLG?S94%lGr9H8(EcIL?3OLu!?Ra!yZ=B7&lU!T z)C73{Fy|u=<&5R&fuS`ksr5L_xzp1lfo1Zeu3r4z151ZL*6-QYA7?%LuYqM1;84%r zbJ35@v(WL}9@@RD{35adz_E7d^n*^n#jV{pQ^$_JiVC@%6-4n(#f%dP=MHT-tQ@Rm*Gr z-duD5{A8OoO4D5%JZNl|^NGI)`9uD1uh?XK-tCp!Tz`0v^Ni;k7@6W;t7o1p%{gS^J^WV17h=#%oH*(a ze!(84>^*Ag`{23wdg?0iOSLb3XA{0L@~nFtya;WZ_h_@q9TT#FSftf&Om`zsJRXyd zI?v@F|9-Q_LZO$Z2S@d`l(Swgl>71zPfTxV3erxA^>TVW`zsu>!rCuF`yK^Z3%76t z@NGjbJ&L_%@&TWT4CTxy?8nU7_ebF^@ZL;aL5%Ok$Tax~vOSB;`r!$Y_g?4Bfrn^k z8+qUD@?EWEA!2z8t-Ev~aMUgl4+ylIC*11lOysqDk)VbVYwU~EV@2K^(8mHn> z4NP3vY%CWd2bLqX>VbcpmtwtcxRw1+-mdrOJt8SFvYUVljOt zScAnJY1|3s$>i!P*sBAap97ra_n?Ww$VSGwfwh5xp%QsreCB0X$%N3ge_Ly{$0?K&My|dHGPI>i3c{>W{(BvD4+9bSPRU2tg7Jb zc4Ya6vpCPz#le;#*3Q+){%*$B>F9pbKYC;8|E6!*%AMr?6ld&mawZLnZoV;`%UOFm zhf{c*iH_Mq8)`@A4|gH+{Ocrhp_eV~7eqR5wb=8Z_CV(fKJmr2JbX^1b2;q;Tjv@+ z*YZ~tjC9tScxHol#-`5TOl%589($O5cEFWjPp@Dt=FQMA{g+|u#SglE+;1(;1|g0L zEb(tL7d%s8&IltvUml9ov*)sCtW0aR$s*=}$@+WR z+?fPtIwQ!NGvVw$a3&sH<>G9mi?eEQHbdtVmIyZmEfwG_2N;nnW0KGw@hIZI$ewHZ zM#Gq>=DC)={x#3= zh~{~M<{2J2X`VOrpJ&alH`mFBIExlsSsyem?Cyc^^+Dqj9apn z>|^J`zrFt2%MJYNwN~!tGGp6%u(th;U>%4@mxF`KM@yyOj+WJ0k;{|LTK(iI_Ev)? z^p3`w&D-eoRBRY>1KZ@!D_q-tn4S9ld;R16FCuUK{;FuZJfAb`cy{JU_DC4b{_3m-~a>d&8&x`k*YJWfefc~2Q#{D&XK!5drfc}=8 zVt@Bw&;NB~Vc%XYI?v#G&d9Tw!*Nz^xRiO6T=(j1&!(0|ay?k&)&yb)Iexrl>=4n* z)%30TGv^jmIp-Fv#NYDHEjaA?f91~k*@ue+!{|O3Jo!u9I+c)YPlUoVfupm^N?ipl zbuF+`v!Ib~WXEja?Lk*uJq{Zlc`BLnb$Ig#I_C3?;hLAr!fW|{1@j5Gry&C1#~dWVy_@#JlKCn%E2wMo^E8@E-Sw#LH+0!E1x~v$=Quy zu7*%Y4}Obsom_Pgde{8SV6N+#>#}c>YhX?`VpodJz1p*{xVYN{Oj@_t3`~c~s}>Kj zQY8g}wPug#X6xku`DDE(I!R#n#O6R7E=|RuukUE>z90 zdJ`=QU zY~K?0LYYJ^X9~F-_CaB-dYg#{%(2^I-SwH*0j{O3tdR2o?UO3f`?^O@e%5_dyg@U$7d_X_Le`WJ-xI}vIhDH zn6@TIV(?HDxrv=eF8foF7%-Qqj_SGGNS*BKP3*zR-hryWS$gn9qvXqfGVcava-@Pe zcD7(1(g*9W{O}-eqv8c$`|$wnC&Q5{>TQzpZ=f}`|Br6_TfO$F?=|fcmrF3#YW$j{ zYCnJ>79b{2aJkuU^p)Z4Maq4_PywzU_v31?{%7RlBj3q>nVJ4S<@UeY>)**0HB2dM6Gn%)g7T-E5OD)2Ynvi4W;r!}`B*4%n~2p>hSsGSY$ z2^xiF`}PnncVvA7dkhzWZ#NIQU3tj0?Dd*`Z8g{K?Eh>-vGut~E!Q>2o_v(uHH`fh zJek<;+dIc^KOZ*hD>D0iFc)o}y`Dt2EMi=oKNqx6*`cv5WNZtOt8x54^xY8V%x!g0d&revu96rc(KYg-)623sD zUMR-6GNRr2gGQcjpibZsxbeZ(^nJ#e~QWk3bjTE=K=?yBb`?u}Opv z4}R@G{JZ(+Le=#HUjSSxP8vraIWRiDjMq-=k6s?yz3$p1;<<}Nhxi#W=uczUSoF{H zSBLyxj^`h;$HP1fG@i|jCpOS|8eT1ui<5tVR{D+6HBytmuA(+nR#D8-g2aKF1_HZWbP%4o%pT2<4Wc}Xy!gplW^zWyXVY(z|4Kn zoqIh~r!(i7`=4!b<~+b$7u%m@osBcsF~%W%8Dp+v^dBf9F2EdHnlJ3SB8%J>Fr~St zIV)w(ooim6&6y`_vYqlj=g~%#Ir7uTeYM0_v#v#%W8rZFc$A!d$|n>0_~q|3pR)&; z&jlZ5K23d6rru5Y8u>`dVWg3vs@GQBka_JG=#1dpLCy$v=G8ePIM;th@N{s7pBkUV z+|k!w=J~UlPW5K+2ydDFr>}#@=6F0t9uEFe?lM^gj{%$E zF<{Gu$5!Dpvgh2QE<0~he@A8zYV-^eFwfkPsIip z%RqdQGXP&y4WNIPntWb=d*Tx2)Xuk4(|&X^jsF*Gw~x_||BUJlA{Goe9O$1TEk!S1}=IY$QH2HPdi z=(yaPZr674?m0Rx*Vrz3u8z|)ii^acclD$G^5I9-K-6u52cg{LcTi>K41qwA3$Iq>EiwN8J9=(O>cBVQE9*vS3Y zxmN-#;~Br|zDgK(AwFh+8leC%D9-Y^|9Trg)S9OLi@5$Eb$uS*wD+W^k2KFG&|%(h zhqfGjC>)>avqkRS046?RPmVmez=_-Cz`y=Bvw1ejJV>9N4Gm}OGsRvAtPbo?z7PM- z`r*ku@<-N&{{Ml0o1yh={(TDAPM3cV8-F9?PqVJY_?hVy!9Jb&H}f~pgX=ajKNmw^ z#^1Qu_!veP?ZMxm@0k3J75E$It_|`xu!B6G#Q2x!A&m!bd}YZ}~iVfcc&NA?CNmtarqa0vW?&F1%K=>K&2ed!tf<9pJj!kh41OwJ4+i?gX$JXCMl59)mC zmEbdnoLEz>waij&#RO|vyJ{;YTgz(oIn`Qb^T~XVQJi=;dw5Kso!lkWSM1`Q3f|ep zJ4O2BohZ+;p3&50RI)d?^tb#Q*^PoDo=eWNXbSN{d&dLReA|j^qF0RF=DR1qG1sbn z@Qvdq8db-90)1Z=Vou#>ZTv0=UyfYxf49gNXspxmF<(ReD^{m<3}a}gB%R)A{5|xz z+LWy-deIs!W8 zrvs5w^IUpVY+b%woljm_j8XE@a{qbs+G6UNGWKh5ww*#2>?{W#Hula!&NR>9(v9b1 zd$h;!@8L6(pQ`w@Q>$Nha=iZ^d@!ljKfC`IP5+tvMTq`G^siiS1^w3z=hK@@WQfLK zY#-)iTzmi+yl0RReZDTZ8RdgRxQY5;5lkLFvf&Xe$ZyX9t{mW61zfejwb}LU4GnmG z$|iGb2c8@4ouAnsCth8K^hz)5cC&4jKe})Uo?~_H`Ct>ovu9`Oz2+hROwG)dkv~$0 zB_3`!zt4*NUUga25ENTOBU%$H`R>Lz%z18!d7Kpho!>@1i1x2n3|7}9Q;UXCBMhCW z=3oA5kb0ZYRQkehbYyC%^E9&RTt)X6(rZ=_ZWyV=rujoPV^XXxAOuafwSo{?Pn zC;F27_TG`*+k7MY`t$xiWcOZlykP1(*9zWtc+2rMyWv-@t2?Ax0qCWNIVc>$`XFRV z7ka1r8sG))-O9PX8fH!9iXHF%$J=H-K#yBVQClc$RVX=#PEDo&BQjcrd_T^O6Ti+Ce`8 z-^U{}=21W7;#}u??z@6{~hjZjTl}mk)05+_B{pa66nlp89$gJ-bgf9Jp%-4%K6nF^65{vaReL9-e*- zo;u5^D|5%=`Ad5lhiGRb@5)DMW(>-C3m3b=f#i5_Q~-S4xucx9D}MEkgTjYwh`aD{ zf=eoo|CBaF=gjNTGR6>GQhdDD>~AqVat8BmY|-{{5kC9zi`I*))=l`_OdH<%aa~hQ zqhg}D_<~7bS=pd{UaXz+`J@ZA<}!+I>%zuKeI7o?Ulk8|w55Dlv$1_X7FqRAsYcH~ z7O!q#9uw#uvsQ<;RX=tg{p*_iHT1E`HJkN0@T{Q=`kx5jGN-A{4VF32*lXY4=huCG zd3k8Ht1;FL(1+TKGF~U2(4Q}(;6QS$jdjkFV+VaT6rMaiY|gvQ#J`1;_$=fM@=dYL znY^bMmh#OTsQmz6t?lGNt!t?(Wxe3Wp!KliLIRyDJ~DC$x*zrXbYnpFBkOC5%v@d^ zSx-HanakSZesjsZTo&Wq9L9JNW7PLz@F-qiz4Le@It(5o_dy?yJa=pc*(8$vUi)=k z`_zxoPam!L+dn(2{ePi-*)Cdt5Szp&auwR^uxdv{V#EXg^gGENqesbIe%4*{=JYT8 z`gPcJ7tfx4-{#X{9=_i%9(AzkuP51%pPgUjI2V5E#1m1p5qeUDWsj}&bEPvDYGANYYIUw;H{&2Q+S=n~nc*3L7( z@4_A$m~SFk_+$Y}A3n3~-%{jKSw8cIjaa%Q!{Zy!B_-6ZNdA_h*NWlIa%>p8+D`rQ z`L|EP-CX+g?_c2Y=a50-Edj=JCHNWd()B?3{TYU?PtNbp_vQCzxOA=jzHqDdh_^cQ z4Zelbut(ct?Npq&hnW-XH|6%VCwV_#eFeTcxQ~P9;3Mc~{B6;9jCY0q9p6heX2)d= z9l5l9_5idl+Pcu4lb^dX%)`&sADqwqQ1tz4#+*&xz2Eu}^i6Gs{zbd{Fd=(F@enof$byO$z#jv#dYePtUBB z9^*aDnc7-GTly_srfX_TI!$YHJ9fA-#`R4cy_WGsGza?`zvPTp52*Txr3U{Qf2@@I zL8BY^=nk*cYr@?*^aU%;mG>wx(Is((-HoKU4Nc^#p7c zbhY$?YYy5`r_59x zUqjW!Ga0+*kLI*4c^EvC|B7QfsXetLJ5}%2zMW4k2zrWnO)}1U#%XjP_nf#N&uCtK z?`v$>J+guN*8RRheT3}nMc`fb@j_@(wXhEDxw6^pvzpub18BF9cuzmw2Q4L$S)47H z>SC;-kMS;zdNxmDH1H#PI)SB&aeT{_AGT_?t~eu7%-(j25qR?C0u$fQ(6h$#edJd* zKgiBfe|f(C6yqT@VE%dSIp^(U^34h8P@mJv7#6L?Y+j$$V}B>sm~brjzSB2 z@SU|EP}3;t8-PtVZ9Qvqa_|4{fmLVxZNc*^$~G$O8nkl zwKEhc$oeh2HfiWk=UwNtCZ+)sd?4FNeh2x~?bVH?4()Q^qg{t??H!9E>Bfb?8e>ki zmR<9*o8S59qi)urq;Cu?Q{GMVBHl8-3~euFohmf&q{b&7h5XUWA=dZAv4d0}x|{X2 z#nzO_M}LluOWSer4()F=vF1SQ9`0AwUc|m2pVeVUqQ=v4ACz6sd`&J^#o+xZR30`OAtS3YWVPcb=BonLO|j&~fs z$uV&_<~t&qmJE%T({9lf!7hT!yMnCK#}wHY~7e2(>`u_bFQ`2Y9JSM|hs6g`cs0gv(fG&_*2ATv*s~j1 zdFUML*sJI&#oprL{VDJo&;0_~vhob<0d$u9y#PLHdc_$nlDp!utC(x?Q5XE$bu)4f z{_ncg$kB!yZoV;m!=LjnkAj0@_LM80n-{r@wMB`#ki#SJ?w2pa{sq@+yBS*h>lN=h z_A$IC+AlBKz3v@&q4M>z_lgQCK0e(F9IL#fY2A3M>PJ>V8QU7)_M?#f{R8DUZ+oNi zj=*~Lb>|7U9v)c{1n<%twshG;jDfz!-{HVsQtQ-Hu?Ip&S$l3|lj{9h15i3j?>PHr zq*n}Qjl+nR*h9cr15T>=3$e~2p}o$Wy`PfsV<)ga^z!jWiyT0D#jutr_^ALt0T(}- zhuC}EK4;8S({Nt%rL&3K3DKJ z5N_*4xAeCr11r30aEoj?T3mE-WFz~mB=(5*G+#Nb&j6#^(t8FES#b90a^$mg>7_UM zVTFx4*q-#s>dLZ!HSVG~Zh^o*`o<@@cMQ0sAGlA2q=me3r* zi&fP3C!mWnnM2doutC}y*}tvMLD~|{UO-#tt6qyeK(t4)boETmY7$@1LY7MxJM}Bh z*#IlCr+4L8%PNuAf?>(}8L30;{p5{DxvNh0%hkPwO*5E#$?N>q=X@}9vLCKDCdm%r zL1Pm<)V7S-b)L0M{+{ls?^Uc{%Po=&Mc%N6XK%CC`sVfZDgCGVl8w-N8T78RAs+Gd z=k>K|s5R|e?9|`qAa8whQHq|?@9a5fq>X0cao#*=y;CXtQ4DUpI6#H=Sj(9ZDLn%{ zj*qDM(c15+KW7hO!)KE79{syEBXilG2b}A6$f+~m2z)VgPqc(T?b$;7?|c8B@xQiL zDxMm!rYfItDYYT8=SxDwP#Hh5XG6!T74qKu8hhVI;ln>eLvL|?y>IPP3pHU{`;t38oGFv&@k3IZe<8$b#KK!H4uaJ)~d4#OnBfE1< zP9&-NLgZ){c+t9L`3QS?F3ubqd&)mv*Z()M+mQAacb@-Rz_q>I`_6kg*vPq#%yX}M zcv3FqAG6=lybWU|DO8z+puFcP&*i{u~NH#_}9iiSA`;~5%tr~ zL&!7HPV822bhDkB4t;DuACVurJ@*T%56G9a(Ld?*M_IphZp&KR*ajWOHux3uau67P zYo*4cBg)z*N8+zo&b}N!2uCWlM|@Qt^4VFp_0rRl&(69)hi6?G9m+SpKypuccmnnV zxfJ$iKPwWe5f3=_o{??tbKsE}S7GE5Xm2Hb$lt);XefSeltX*G13VpN_#8$LdAh)R z-?Zt_mSlv#?b#vl;L0>-FWx+IsQD(Aa`JPYP3O@YIYsDC=fEiT;(xy1+_Rs%Lgl8*6fD(<4v6V)SprLiF!O@!Hr@V#}qe#D4aA zU_5iF$IvzLoUV&sb+2R?Jc|rUjw?0!RkP3F7~$C75!$Y`%-`A(vI1?1bkl**m_S?I z@Y(~RoIqQMTA)z9W6M_2=O+40>_ZOi0sp(4KBv*Au1Wvvx?)tiH<3Q0>#bDidi8U4 zIMg6rLVu%z;oJ)J=zUh|e01q5@* zYHLe)-uMtamx^BcR%JYRD*|UF#Qil-L3p?566D8{S~DLri2VpB zO;)HT@!O_#3oL6{oX>?;@E63550qiEPFSM#q+?p*YZyZl_bk6MDS!=jEQmdJ^0ei{0D#;FmDg>EtX7C?LCQHvr=zY0A*1?{=#q<)m~#Cax2 z{flZ?6eH-PiKoy{&(IHJlpXvt#+BfjXhPRz6BaQx-QW7Kn^PyAUV_bEWd$Q4_PNYC zk2-{hfsZj*#7yP0a=oLMd%{k|Ug48|V~4;C{1Sf?TH;yZM|>J5$K~}KLs>}vuhZW3yz^&%XZIIC zZ>qoe+u)S@cTJ{!aN>cuq2|&DES&2cBb(n;ItO`?5$2 zVE{i=5T=XTKKb`jU+sF{Esm*93@&ny%-Ur{3C8AS)Hxw6< zw`cyuZ;d8T=MBmhi-CD1_uzJzsYWEJqxPD3$1pW4%y1WzUK z#2eY)8(4~fMc+LhDW9C-5gWXS1`-R1&wR#e)B9fSMG_ooEW>$U@9Oue?>VyRuJ1vI z@Ws|k?6w=|OL|vv)n#av_qW3Rr@ z*()JuO!EQx)=fFqr%HiW`cUmuqZ3MtawA2uJJ^@!u61eEpc#4wN0*5Ph2NcS`-ZQd zc3^IP8u?Ej)F-v2fx8E|b?#RWaCeWodtLqvt8EBlN?Su}#^du!-v!u{S96&^)oS|{ z<57ERL+?B?D$ur1@mJ(rH@?y~VDrm}d}Ks-`x%iQWW*3;#LxgXBCxS<#ZJk9W^=~H z#gW7R8=fDAOowN+);0?NNgwpl9ezeF#Z-pXLzLIL(LJd%Ek#&3L zE96ARa2AM*Up>DMT2TM_E-j>8T2M}67PMZBzoPmw%~d!1xk`t7ZI7XCudVLi+HI5Z zLxv15SX*%H&Vxrd`)KHr$+u0)wc5H5G2c1%vLfs7TdJYmD2@D)O+M z`NT(<+wn!z<5|{|Q|j|ScwgVuhWv-K9;HT~-(JkX*zR= `C>M!d09Ur0nQgTRo5Bun7@g0|# zu^D~gS?iVR;`b9`7}*xIH* zpcmUuCFQ>c{TMst+p9hZdnn#_$Q9EieLqRdpz_ zv6g%gW8W!`aP-UQ*Rk-zIG)GmT!?S)>Cw&TkIz68Mt@*KN)A?Bs6VwEbbUl8W~gh9 zZ|D4u4wIbtc&3?Y>G1pnU;7=!e>fvDUcLe`IKicu|9`4|+M5bLC|2#kIl%kyY{^OR{e<_^ zjPJyy_(b?d7JAI!e~*LzGj8(gRNVPs?ld10T$=q+)_k;>`kT!DDBk`U$_EDUS%gE? zzag{xt!FA3WIfZBCLidmXDV^?fk|TNYAZ?IxQY9XWG=vM@59gtc#e%Ua^G+kRw&_iq^6;{42KmJ1bh!K3?8N|EE*$cdX9YkMl4y7)NB6A*{dex&Pg?MlEa*MdD!)L*bkj{j`mjYi#}DGUCd1a{*c^TrFps4<{t6^ zUO-1Q)VH5y_O?)*RP>wxCh715aD`?N*T)C&^wB4&`6x!0`0t*gt{lj{pH&rz;jNK7ViC4Oiw zZBsu$t$G`K3f5e}8VT}8sb*xz3i42;lCQwLh<>*K!}HipqH*m@SIIT)lX(Gaem9u? z=}I#D(CK`^^F~JEb(x$^jQrr6^B z$pY-b_7H26+_k^P4`aRdL&Se2Z*o0;ai5DaM_#?oxh~HrH!|Oq6@K~s_u`*+XfB(7 ze!JxK_{Z!=Wa^zymS^et4*&eUW!ldk%93fU#T-bc#THoMr-(g9fy_#)AC!_Z5(RcGM0N4`NR) zCr1`Qp7VQWG52H>$~OsW4$#G#hhR=Av64V<>?!j0$lZlJt7p!Y-Y(#5BGXo3>)HC9 zlllD-eg`;5Az+oc{S~%q4VvH!enoQ}T#L`RL$-7cu&PbPdhop;2F?Q>EV5_(WAONM zFE|r_{`(;O>Fj5isi({-BKANnOyD;3?k&Utt>Zsib^L1w^Nv1nVB%kX{>Iio`x~4! zvsV4;Y-ribO}ALz0yidxBiL0t9i+xccCCEj=l=i;R8+&3* z5w$E@f3BDpzA|>6)oNY;ZwK1{u;Go{50Y2I<|rfn6vqZUX}vprnYHUV=)D5w8XLk! zZhC7(%8;Y_ZRVrgu?c5?!?h9J-`*I%=8o}&LFQ+v$tPv#*Z6PPF*b7=EV9EJp{pm! z)B5M$a`kRNwSL4npZubucO|dK)8AqGIx|T8caw8A!3N*i1V>yrJpRsJPoda&nevO| z1Vxjj_?H22Fdn|R3z{zj7e&I0H-6V1^~2|lz0e(dCw;2FW982J*sIy|`L83#_Sv8E z1%j^K`L|d{5xvf_IdfS@kw7NL@uOsSc60u^>aJAVr&^F4c)pr3)w?_|T5TeKe;>4- z1MlnmW6Y0i8rki#i8nJJIoRzpt>n{NSTioW-TcOG*YBCG-M$K2(aCM*wr+UPoxgIY zHdek`PC+o7gADN6*19No?Q*S=zsl--dauP9^t5e$)3$zJ<+hD2JvBLtxEy2B8rFMg zU*n=SxlM3Z-h5+tgE_ZJXQw%PuJ0yRSxi1k{-kPaBqxK=Q5p5TsxK)%@6Lm9_OCCi z4M$3?z*_Csv$B9X2V_8GKQSKJm$mG3jGr7WcYIRy1AScHMc?vyE9LW642^6oz$dN3 zA5;Aa^5LUX&}%%CPfe@CGnpD;(d@t8;K;CBeLQ05rOsVXZT7^-tfy9;VSItRKIw1i zugt^dxdvI`wx2`WNqoiC@+%p~JkifQ$vtw7%%60|TYLKF_$MQGOpb-?wTg$3JOAPI zRHJgJ3FKo1c#6%31{XkIoZn1dHgW-Ze#J?MF+CZh{h~`aPsTkzkNMoL{akiK8}bhm z*NA50PuHWD@@cd08Z@G7)h>;R{hbe29=yjwPCfZOXQ`eftDbxfG71@k zPbvS;*c{TQ33$_T*N{8$(NOEh{+T|OLz`>hG2VF#IpX&>*HS}eXpZX@rk-wE#DeBj z!(`gP&o=Z2Pgk&(I?8kRdG{}7toWgE?s>97>m8n>7O@t%RwC+ z-RJ~jWo@fJ|J?%-N)IOc{SKfb8|UkbJq)jWxJu7&q1rx zub>_ZdCI$~8+z_Pu=oC-ys`DAU%s*WC!K1ecct!e7Q*V5`wr}V?Z>8F=1}e2zWR*& zO}mx+o}#v9SjPf<2bS%{6_-WEGk?90afW3MygHq6Pa}@b*tTH1#lgb@YCpmCT>0F? zct6Ug>g>hKIqus3xBm$~#7>ailWD_Z%>nrUat1ey%2~S`TVf?^)*XF7uIta~sY9qO zGVue}t}7m~6Fw`TjIh$Ai5=%@ndTI*UK*x2)#^k=?)@4{??vaH#g{kB<*l zx8u|igFmxxj^I??fzG}7>{;(Mn(=5Z87u1}8micnMSA92ht;0HuE%M|<2~n|<;Gmy zdyGMH`C^y0zrc7rd*N!%#K=AmD|DXO`-wVf>K)i8TYAa9GTo?ps3b70=$*Nb<;VKN@61Mg!a33 z@c$j~@8~GUuFHmhqxK{yI6qR(KG}PG@P}OZFJvDD?S(P}_!n`li8UgX_@Kh`Wzd~q z8lcUNGH{mwcR!&`;X-)P7^HVD1O6oN*8{)SlnMSKotIoY(yX8I!~exB_`P}dk12zj zQ0vDBzzuD8Sht)RS*e;bA8w{)wOs~pEns{{INxR~XXtDr_tZV%%t$xiU&sl(T)-Ji zFS6$22;X(yQZL^N7g@_DMpm!e%Q~e6d|w<{v#y)(n!{S`RQ4%47T}yslj8zkrVpO0uc2MNdpGUA$oC8A;}?9_nL9a=`gQf(n@%5Bx&1XEm(+)1 z8J_(9Hs?%8e$O62ew+NtAn_hw{CfM}D}H_D1Hguz+84jJ2ag>nULZX@UG=B%0`m7| z#SXIL0fr7j4sDP7`oMARPq{eR<-%m$+aq@bw68vB3?g_Qd8XIb8j#iO3 zs%QN5s?l$|{UJZ({353&Z6dV~dz&*g{#pZ|^{*up#P7MSchQD&Pl_M^ zfXJtRYj50me? zM0qXKc00Dt9&ENM>ugiwWJ5Eu6{P2}Kaed>E#W7gptb>jO830$=j)p6WZ$(ST|=+= zuZ`0+={r*!H8IjeO_=niYM1aOFH0~6jYodY6*nTI`K`RPu1SxEsNLyg%(>{?oC0!d z;9=j5CKu4Lgf^C7<4)u~?U|ELO~gKAW}3Bj(hd8Nq1{@S*K+rP$#*|u*5&!vMCK=h z;eE`<3*bEe2aZ3ZJ$br;OKT4gUq`+cJE(USwb3v{imeW9b#E>=wJE~sd0cNtj~aXDE9~nt(%!K+%$_Q;WwCcWTnb;plim?cCHP+9 z>E~Os>=reeaRCW*#;GPatVxR?}k8SRfhyqQRdaui!yhkHXV1`B&HKv z10C^BY%T8qm+CHM%SInEG|^CHe=d@^nP+aLE%hba7?=|)_$TASr zF?3?bp5r>pXlG%Lb*vqkSV%4RZtM=NO(~^s>D^bJhW^20^kzGCf?SA2|3=2SYr<4( zbOLyFUGorF0#8HVdbi(mgS;ODH)gFIxG7Q_7lRY*cz<5`-~NK&O@4nam`?V|m z>-Rk$UYL=?&~~cXxAx5Qr%e83cz^p-dMAVoPK>j3j#_Faa@?$47_GVu=N!$ZALkr< z4f)nx0`I*}jmtPaqx(*sw`a=}+nIVc^m;zDT5UTrXpxOPf?lN8yz?AK@Qmsr7te6y zK=lm#Dr5;b*j|0jOvR0QE9g)2D8H)`8KZj(ubY|r^mf*K`1-G)hPat`@dtKJAnsO= zUQqnhdtWuL8+aeN-5{OtiLCd7+>b(&Ja2Nf(gB+<#QsDFC@uvo$CMw}bBb?0$8+r# zbfo;+&Jb*bm3l??#yS-*10xiQ#7AHKoi+Y!Ws%^BSF^ zHQ~@9v|4@(^9FBDfH!M_MYO2;IpsFfD>x?*-zg418lMTej$LZOkE}y})3-L&#F*?I zuRco-3ml3cCI9<)V~WEYj&0$_J-~zHqyL)br^2E0?Um5E|DMMBQe~=f-vIVd3w`GzHJ>tz@rF)&3FGNgeg|<){Cp1H@nC zdnkrb$M~yeQ?JQq3|iN^gvuQF8@`a7D3XkT?qen36xlHvxccUo>xV)E+ezM9g-3xo z%6MceiAP`Y!D!}xkF#Fko`00R)PKD~dUPNa)mnv;0DTA8(>@SRr$@91cFFG?U`!&< zbAeZ9%Pa0!^=`Vc_G8wQkv9fz(3v^=PyG?6=SU@qx9*x^hd&Fyn;0Q|oD?_IwXC?IvFmCq?O!rA z3QUS09s^Fzfz2F_6c9CjHiM4PET$E2q9L!L!w4O4~YtX=}mCwtd9K1yd6~ zUUFO0xdxAJ~L?EL_;>hf&`fwt;qQxCM=t66@6S^>$~t#>>1znkxNVvR}a zTQ{G>^Tf|LPqxfweSsCJrhd2bde+HKw%cmmc;vI6b@u9${MMQRZ1&5venmFtdezT} zmgqxtSdx!Dg3VdZ`{iNAUs9U-mx+Kf)3#uWQ608A%j>jR^B*Sk+FB(--c zF*Kgp%bZBIt7d%S5cWElg}wrAH|Ek{v8OXJi|u8%utzU-v;RgtrQ!?XvybzhyF|OlvCetG%37FO;8_Vg zCbrVxoRN0n{pm*gulbi-CKf7R#NLsQ?$bQVe^}_7#|xOpBIZzYSx#P4^Jw1&kI%w3 zV2rvhUEBpOyXJ76-!X8shIOhQyng!56@7mb8$@z|XM>x;F=ryg<^U&g(ID~$zS{Za zE9}%dKJAcGC#U=3+0Om&^?Y(Hj4PJTz$Mu=qZ&E)j%Ci)-s;M&cIMWTX>aYx?HAiU zi@G|+b`xpjQ#AbmeXLe1mAS_C$Ja zAu)hTXxYsBE5)th_2J`;VP?mGtG@K2e` zKXLfyblVUw4J4CKx1GUw|8&~{*GILtF?9#qRd>*M75tmI#=3(GA`fO?WADxTxhCFE zS7W2QyyeGl6nX9U1<$13!Ea|SrB2?9KRWhy|M;V^x3NoY>{8Fpb?o;{Eb$Lt^V!>* zPR-u-@^@E%?Bx7idO5Mxk=U^KDM5I6KC#bG0lbVJn^~Kqx@Oj1-5AbOejGmThVNdX zhIX5&YtF0b!6(xiFYn$n*aj;?dFGskZP*9v$pz~f@q&qi%z;0yLk>@~O#21YHgwK{ zrid5*^d9iF2Ht^Qs3T2vqW1!d&)C?w&|T-Pv_bqK2F%i(SBh+Rt^><-ug>4iK>>^6Ar}0jn1*6$!nfY@^sUKxWxXd__xo|#eZJqj z(Y2hHb3EO+k#+sbg?8~~^3h9w zOw9IT;tRYVm;Hvkm0XwXU5MP#`X}iO)<>mgA?Mk*?ZNIT#0eR5Fji)C2zC*9$o$qQ zHic?DuFba|u48YI#1tp)?6sl2#naeRif#7L{$bjfyeKz3`2xG`2=87KC~ZrGhOz#@ z$!lsKt7vE_{%?UhoA+ukhV(M}DUh|>6c_(v{hkw$A{VU%8=M;4mvz)n!!)Kk$(=*7Tt@xT6 z%VK0uJ~Bx3BKmj-8T5jYLHRX@T^Z!vd&S70d?SNiaAlC5>0)kLY3~R$x>S49LO&1R z4Suuw?P5;6wvIrjLw>+srPyOb9-sU%9j_fsW0_gh~5z%YQV#z@ZKHAporGCtB z#;xBkxqjJUWb8|<+0@zthgXKTULqOm8*k!Qc38Of#(Nl9r?x_>^<><^W@MkSznr}R zv_PeMK-Qyv&sL7 zf-l(;{{HlSU+f2%c2jehS3T40MW8m(ThoX`wDs9vv~B3E7+ag?bK9*5vbJHN<5M)q zrw9~je>CU0>SVfc8hyu+ql!hn)kEBs_p0fmI+uJ^3G=Kx7w#|*_VrTzP&54@4HxoVPG-)apfD?7HxgI(PD2a^{F=0 zUl(gEDjAb<3C>=cxvk?FkLVzC{Z4Zp#CcrTyA_PHl5xd(Ut>$q)+ZTToO`9Tg)hWj zRp5Rou{5qF_+E|v7tQ@G_ByYiE%sKDsL1>SL1>N7{k7mj6qmJs8-(d6$MDu@d-Gp6&YTTx7>8lF%daUnUkh16)r zIfquZ&a$_ND?)=>pSa-NGg>5ns(81KcjY6U$@|JfE+9u&e*MWcS~I8xXRS`}erU1u zEATmW5T##1S6*VLzP6Hl7wwMcI<*TgkLP-P9I;F6Er%yVt^c$G+B}hexvtwwpd)M? zzwYpO*MCj)UBTXZ`mAIxJ^ww=ewN-V^WpRH?D%Y*Q%`w;$)RN8^T{gYDz;$k3D!U# z42Ba6EO^e|q1;37JZKs}Eq;(a!Jc5967$=IY*%fS*;5SLVhee&VbEX=>y}qgXGLOR z?h1H%345pafN#kleQIwo-JeZO#iLeee=oA(B|cw8ue=r-zW*EV-hc1?)`+#Qq2pEr ztYd#d_betq;?=OG(*-T;QQ0Owkgd>F4sJ_=;Xh3ZhKGHRJZ+8L_8Ru4>Yh>- z?s>p&J6v8CJ~FE;tUa!}LT6z2j6jwRLrx5D{mrkPI>f!;s(X4MoKHJ(p3(Yx*>n!v zI@_h7^_k!D?DWzw>+;419@cyYTDN}tokml$$@q(r`Ods}c?Q+z=C*zTdNOOlXtRv( z;wdv{qrvwa*8Glyw?|v4PH0=_QfeMM=DU0`8W}vt<{9MAXz&hgbwM}x`R3N*pNYTM zIq~#=eRjrA0nQxYJ%27Tcr!MrYbz{4cHmcBE;)Oc`5$+u{m@}!D-_fm zacu?f-aE!tC@{9dVb@mBGuoF(`w_)=F`hKzVJ^GhVNClNBR0%6^ws;gmHLITBZjr! zwVH9VzB?^UvC87A0z9ABkNwX zUVZwpz;Gk$%x~b+@7G*e$Nt+>kIYlewB~(8>t`fKPqE)4^XxYDt#*$26(gLE zy)Iv%v}5JE>7#73C%Ca!;FA($M%AeH-zP)=b4MM#QFVp&@(SwUuAxoEZ}olr=kcxJ z9ql<$SrA%V#{S`DwdY6d;eoZ$w^?70EPpnSm@oU$7h1tyn|EbT6jPg4f^L;QWY3@8 zmx!6f_9*sEyo`De6JLCrIXDDgC4jY?&pmmO_(SwXp8C-R$2Epy$o{#>J=C($)@t-L zG5l0Fxs^`x`CZg$NS-c2o^C;&K1+R#V)&grb3U>-2@Wn`obOg(!#`!GV$XOy_WQ1f8UV@AtMjE zTzR1QMQj#he!@yE5x>Jjt5?Byz!IAU4xSL~tc?dQ!Q=JS178kl$=Ebs2Of0FhrLP#gUMBf+3jG~^Q~OKRKDac`#L63%zF7vJ z4(X*fb?z&yiLH9OEG2pG?GZZrk+&P$(M=v+1;2RAv6t-~UC_`z=C{zw@11?2>W%~9 z7nt|MjP1y@((obhx8YCeo!14dF>1e}g!r`9;7+2(;wcNhVE@xWXw1W%iTU!}1%o_i z+9eKd+9m$&wfnI_o{x{QrnxlGAs=`=G~kUV2@TAL2I>qA1Z%W^t!Ti1f4$KO&f5Gs zmj?9Q^U8NJANkC^|9P`l686tF?4KUwf5AUN1B|_!F?n#Djy>DcM2(g9^fYsy-@3=w zpZe-)VvZzph7;e}g>C5A=^b9~&ObiwF`m!(1{#y@74hzBjcc2gdbu7N+h$^J4aN33 z$Src+#+L`=+lQ!w(CuQ{@_6q`u3dlmS@L!bxws)wPQ>DW8yG9#yjBgq@}A5;7Y zSbv>|KLT!&;HHW=&~E(1ZpPG&50b#v=@;MOdD-;RF?YR(&EnFN>cI|kuju>Kw()K$ zF(N%z#V$7c_JYxZZ}6~2nFdA@$jJB`i29tV%avYzzXz8m>DkUYO@kn490 za{b$bT$lawP3}ows2|79a&(5rQ$3r|LswF>Hx0RhOzV>?@M52@i%hVS>7B*N^po9J zA}6L1ciRNN3D5Z5UDQOBLLb7@CsrCiu0yedVempPIC>R)Eypgp_PuwA*AA6#*zRBH(8_B z9%W2XaKKz$en0bc483$|&{i)~62(dshU5{vG#1LRIM5M5ubC% z7x^w(zl8i{8v9

ewJh7Ae)Up zdNDQ*b`G(ahP~~djC56D6EEg36pD1#=#zR9#vq*<*RBWP!HwQyDFxYm=B!Te(`JgEFU1BLyq;Fw&RV> zxyOGNr;T!IcxK!!JD>A(t=hI-R;Wh)2(Y*AqL$}=a;{sERo%edjephye1(C$-rS$A zTDOaupU(DG)cR z(=@p8??V!uSKFrilH{YqHz5;WA`fixPo>z-rk-$&P439aGl@PV@0CyT+A9q?_XcZE z^XTQF^nQ%@l>_nK@5_G-0*|r5h>;o_j2P)aFq~2wK0EG1xAD#l2aYtc(*bbtL2%%g zZFlp(tRaf;d;_m;$wF^H*-;=PRu;Xm(K;f|z zpKv9=W&bHp5}yL^aj%T)#r*CU7u@5-n6#d3Lgin;K+n+qxxHSCpjlr}Z#Tb<}Ptmbn0b_>=Iu zcy|$SRRNdm1=+BBY3C3zg;r`s-`$Hq1v4L=G6JfMm03*KBu|S9t^J`%QACQ z&^I^j?)vfjB~L5-C9;loz8UX0xCrlk66nres|Mn5AOs5_(8eg6TYZp{;P`Ivn#HF zzL4Fq-yLt1T|eE*?^VsK#-O@0$>Kh{9=+E4BWml>2{E2sfz7O#tKPHF`2~g4v|%$d zANxz*t-Dt_tSZY|o1^*SyX->MZaoXkj62)}%+2(*;Q!F}F7Q#+Xa4`0OeP5g1q&7x zYcdH}(PCExBDPEtZi-#HD=oWacgc-gajUM~)z(V_0a4I3NA1RgI|uZX*LckM27O9I5MVr^y6cK+|r_nebC8G@+m@_O-_lXLlAp6_#ip6~Nz%wgu; z!kAn7H~EX~V#exzrN#!-uyko&F597e=?s4*T9@L!Zw6-)7+bAUMU` zYk9wl|J8icM8yXaHiI~rGxXo;^g&+F9u-u;Rt6AtOQTAzSr{0TI1V?_V~Slb!FY+H}k8t zxshM&qj)2~&+=Q(Z#Tai`E8q97O%`LOIKR5rGfDnJHCwjLG+qhepiB5i5&YlV~OAo zlWrA%5?o*J;7B8}2(3lkEy8#5=Np{3V=nvo*0Z0lR(ZqZbgVb@9_(1_?xQ2u!|bDL zpv|@+_R)!-Z>7Bw^f>X&IDSO&N%8h{8hxzI*W^B&<(|0EyY`f=y-upJYB_TkP&9|3pN=IbYG^V}R~FTAhK zaUY>g;r;KF>tTFz`5o_YE!cZ{2RKkRaOTW7;Ec@pHv9ZN`!3EoSBcU1PsXC3kK=sG ztoY+E;UJi?Jf;Z4&P? zHZLX>6AwnNAL5+!=)p)?D?UUs-cn+Ls>=kv1o*W~-t$EJ&%5#Sg0lhq-m zj~?hoysMWS6J1XT(l35*#kqe<8*_uuQ4zRKe;$1+ZrlnkIeJjQ$Rhvs0L(@*$I%X~ zPWQu#`3hDjRvET}PL$o`O`Sea>9~X1!1J(;zr$fN>3fAf{{O8Kj^0Fh;Yg)*-`dOE9 z_T?*Q-snHqO|&)NB=9k@3feoqk$4xd!EQNvi2Su+V8fOT$jt4p1lb#T`xd_2Z@uE^ zO;O*sC)6T9F1~=KVv?SoEIe5d|I*WR&-f(eL1ul3nKg`={I9yV?VgT`Qyk z`VbFBCqoEPq`|2($mvyn2)0{ZU1^CgQ4;_2j%#k*>)5byab;E)k=ilUaY%;Lh z%6r=jsN>+mE=(H-=_e+bqL1Vqx%0q^O&#x$GqQ&HEp2OcwFW!J2emn2(z3FsQ=it#?r_RA+&{uYygOI&p%l51~2L?9O zId}*yKinJsKLg``d+!5IorX_3e$!86h<>>D@{xi0?ARMA`b2O#=%3G=A^PForaoUk z&(HW@^)nmYx(Ym=4V;9xgJ8$A^OFOx+rKyWf7us(2!7!q*62udpr3(S8lQBHG?$)5 z^82dRj>8v*!iyu&;zn@p4(OpEJkcJux@M*y=Pn2ZUcC!kyl^e{s1N5_xNf)a+@jhZ zI@dw?H=DCutKk!Z-vr>e+_Km!5!v}Fvis+Z!Tb9c{H^PBZ(pls1GVNoDUTmd$ z(Jwt-D|(g86ps+yCvPTJiQTIh0zRczW5{fD&$8rv=7{VVfi0{0JG)FDp$T`2%t6qXAhlW~z}DstMR?#0%8}*L8{|unjU_yaN!}M)sVefnqUcmN zJ%nC@%#D4JemYJ+icwUuCNXp@AMe|5(TCF~u~YgN-%BF9*JIo2zH-e%j1_yb|1fo5 zWY@_SAhD}{oX!BrO#2HII zbBd$O33uD*`(gEK^mue++K0~Wm(P4VINyt|^qg>hnxoe)1Me?@244XWbVk>DzO5t1 zv>sWO4CbC+!8c*vx41W(ci-e)_EZ?XRr!a)iyK|MP@Lx$o(V6!e-3`Q^#_d&kf~Rx zJdVUtqmMVOxPd($dE@~sg1;=YQt>6=j{96bu5>!L-LXw_I-19uyqvd99WdDj!jYxS zXBqat{~0ievJUBVrhX$bmi_ zf~sz)NSR#X+JOD5o83%GTM>D^%4&0{r;HI2S z&wd)}yv~=vi|l$~mHFtyz&Vz@{9~(&>dvtK!twFX3YxU-zIW(-bQlH|9-dgF` ztXH?QUU#0OJ~RH=H0vRM{W$H>abiJDs^yuBPN$sS9^}J=w|s*7_N;e;^)PiJSubKs zGsvgv6fb(^7RC!~W8K8(!Pz9V6Z(pgO}iv3r8CKQ+RIv3xqFoz+z)kZ--{oMZ|itp zK8bbi`;vD_=m6e{FE+P7@EL3IcMg&p#k*b}EH>hfDs047@}g=vgX4reSlXuCRB+F$6lUybi146m&rU$*L^ zNb2kx;dzXygc_q2z@&uQ$MA1{Y5Nklf6hr{95YJhu%6_Pmf$z4A%C=FsPU1LL%&Tm z(AqZDHFax4r0a*ZZ=<%5IcE=imoHO&cC){rmb#`1?7<4^ns$FywcpBY&b8YdpU*q& z8;?-ibm>ptYu>}VrmpEiXbu`yKLd45nL`(24(n`pudZo#rmkt?OwOF)ojCYYLOoDJ zqh0u)phx4IYz)|Oa8J)Ur?E^jTx*!?{t5^5w`7(2Ro+!MapNFyW7$m`tOPY?gC>?f zaF1BJ?rm^m>By5ywAWKU1 zc`$A{ex>za^Ki)dTR0?{yM^_Z?`^33Yo8AabZ03K9Ll?qv@!zG$z{;+GJs$tU`i4R~36B&{to`8N z9SPM`7zb|g9KWi)kTo`T4?16Egm?vVAp*ayf|d-goYkveEu{ngmC z`n!hQ(C5GZp6V=cj_Zlv_R?M-^C^ZOoUDC^e<)wu;UB%|!9l}6lyh3tu@$~L34WHw zM+A>sJc+gJgU2lR=erI}zUzsB`gY2@(tat~RQmPW&^3E0W@x-C8Gl;iCm-A!zjCnd zV-8i!LHDZOuTZYQj_P#939!)E&ZYfq7-?>r>zO=D*N@a$MX5^QMog?T@ms6YT#w@V zJ+9%=bFN~Y>xC2S?K5Y`e#wq?2kkSdzoE8!u|cEeO!mv|UU%Ju7nR7`a`c33-{uwY zW!Yp^o<3EXWB-i!mtY{D<|4(L7n(XFyR?6>0vO8vD8DX}lK=DC?d0_SF`vCw;DFYV zx%}jfw2@Du1LfqzzvakCqbDG*ku%c`4+T$7;Gw$rAv|;(JUhyKs1;dud~=46ie_a) zM}j_?-I#0q^`HN$%U9jMJ-#|;k;7MyXU8u-L>_P6>h8_*@zc-H&c~6*t3Ry$z7ysB zan36EIPJHavjQ^l7sZvuYjRn;^{mI7k;r@g_kbtck-xL?6-fThb>*+@8pUYB$l#vY zR;O&D>iLz9&fRcr+aJ$ zxfR%4;^(ZP_<8fg$EhD3AeLqkJ8c{X?EHhNj5l}DWgRrE-@$q5xgGy`T}pe~44v1wG+YY}d;6si@Vv_912sQ3`9M2rpQbk8 zEN~wE;bopjd#A~2C_-l3vNL{4`!4dB#3Ja1$qb-gDaE1mq0)rvP|<$oy8L;9ljRU5lxJhaW) zuEUq;`6|v4-!QzS)V}WjWr@m&d9MVyl}^Kt)1&x*M*HnXSH0Bkh4=Q+e_tcE z8!~tSxG$ac5O{lNBmE$!Pq=>%_wTb(OWA+53VZA^_8v!>XFxX6f(iCI>S-qu!GW_M zwtU9Af2c8eV^IHso&Rr>JJDjL4q=xR1TH5(sleTkpA;L~ z+;4m`H$vYxJ9De$y+-Jum+{F4+ynlGN6e(Y`E5HBZOjjvmrh;7{GMaKlDfYq0aQPu|yGd2KSE z%YD~nzU%4U_14fJIQ}cP)+fa$o%P`1sA|la>)2r2`isI(-u^_dP39$^ir1HTr+1xB zk7%!grxm(9tsk84UumV3V&-YExy_2@H6SBmGJyEqu_Da zL-9)NK6qN3vwh+|zGkVWwl&(0wdLAZP$OJ^i^P29bxWws_;aYM;N<&}s|KICAKCL! z=M#lLEn^Pimv!*L@#L~d=1Bh3{6Fzc@|V)&lcx*Y*C9)c|7{V^8N2*l|8<^^r!E2B zOZg9sv>rL`dYJgqGGKs=iq$$iSH5)O-UB>0%ozvaS1fy-IZue%5~kizZbwLKM0@&n zfS0+tmeCUia^OyDkI<(@pV4dS>jr1-jjX!-Y`aFXsy2fIZ^1*OCRcYBXCENX;7k7Y zoP1AbZ{=TTUzIzCLGR=vAJIvwu5xse|IXS!;+(yjIp0ijJqBEj4pM-f&pb{8M_Q>j zC_6QH)7-Mg5!Q_0s<~yg$PDS4ijf7uj|6io!gu%`@&~tY9SJ)2g6uH87hFYcx@+c^ zJv@x>uK_Qh0qF(e`|F`^VyL?k%(vF)%|Rn$y;y}aC&zb_;G0@#hJ3}&#KrOhf@}Cn z6c>t@TB!kC?6&Mjs99ogkumsjQR}3<$#=Ae9yu+W*55}|xld73v!}>F@DnM=*x#q z{eZ#w(!g52=^3$SS@Ye%9-5ki4kld04r!{MWpWCeYGyk*g=TFVoqXroR?Ak*qD8T4 z?6Bo$McP*um$&zf8`<7hI;wp+xZH!T*)whte}ARD=fM%}eQl?;_iW}qI$6&eVnyI; z&q|XwRJQw&WzH1psV9#smpb-WQv>E^_L0&@Uod}P-v{;g;TyTPi#$l>Z3W0}S1tI& zIxlC1Hr zYr$RU!l|JFj(>Uh@#YwIg#6M8zN-UP%Yl_}KFoI>{wxK5%7;1lV|d-FVuL>(emFGa z)v0!1>3w?@-^NDXwo|yrw@u-{8g|NW*Ru}9M2(F+96#vo=SJG^7+>DrTQsu$(4k$QMT}f;|%WOp#TJ zgH_D}S6MIZ;ZcmF7&)VSwTal$+H)&k{*TXxp4)hbwXHJwY!mD};F1X5x--aG3ys)o zwaI)t$oT@woeqw1;yd!=*W&B!d(^Q#zI%15S=W<-w2Ms2$De#{&0^Y1!iQ9YCaT=btNDic#K1GH zZ2@rT<@)EWi`ruk#j9$|S@*n-BIxCA&JI&tN`6-51@{1xZ6hM>51&=u9v?ZfJvnYv z`!?1s&bmdRfmnSc^*v;?a*e|H_wbW+Zs44*37peK9>toqta~mw{2MqoNq>Vsi*w20 zZ{eJ-t>mmImRNT=K3QtUK8l?EgLb}%$T5~ z7<3fJU$yTwbP#x(@ii>Prh?AUi&D{A=Et|PNxXHAu^wy4eU+c54xCfY#D(;&-)ijH z@xV&{WNff~+rh8x)-Ct;(oY}l_KO%@Cj+GxqPp)N2Ciar;X*xy`>Mqp^&z8t=!u@w>@k*#RyWdmPqPO=9eQo z6`I%%-Ssk_KE}8VAJudC&3fBpfAc;#uh^aB>htX5Sq~lj>t(6t-ZtJ}7%_0wnPi%K z5AVn4TdBvG|DB@>HctwHufMtFHI2{1y&mH86B*MTt5zPH%pCEvr*2<0=@|2b#_(GK zTW_8R&S#$Vp?NB|$(yTUyukg^B=gW*_tU1X$J4Iza#hPCm$~}aQ|E+E-Jz_sfO=fdvYy|XTZTNa>%J8#(-}>ik#+qlaCO{S>?gjlQt^PNR6|$m zz8)P|J_+%QT=_fbTjM_t%#6Q-*tPmpyG2cbqwx2|HGKOV-+rB6zh0zws&3-@+5qSC zz$f^-9=*uZ3B+goa?6RwOYRQ74-ek2vkv>8CJyA6|Lu0J6oQEUcj{$WMUPz>JPEAm4l!)NHSOX z?2SAEhU?LdlITXdzkzdJwj&4iZV~N&M)2mo{vOG-Xt06*YiYZJJl4e6$j;W#@#bn^ zEFNHFA-`38D;>}4Pwg(pKlH5jJ8>?Ce(N{`BUtSESirG1XtE645?%$zJGx!&c;*P) zve(1Iqgwno!O5&0{P7p8H##hPhJv=g-A39~d*{)1DQ#DhJFk5lLE1i@Ih|@uV?N@T zqB@hfVWPcC>j}(Cy2!l&7Ilm-m$T4)?8eM>QN#JFFS@q01Y~c0eGOy=)d2Zmtvr><~l7W*)@3C>{lEkr!KCs%w^JOco zV=e*cvzGaUnGbq{vnM*}*el+cO#Iz9o`cNAa{oE;WG5eTipihK?5{7qZf;qrYAG_; z((6KHmFf%r0qsur>;U<$&w#F?Mh`RgGd$3>pBvrW!w-2@iw-0ixf?xFxxy>ZO;;J+ zbQrqm@b(q(@?P|x-l8k``$}}vJap6i_THu3pU?eSu5OC#_3NgU=%UIaT@5`nGrqaV z&`4OdgSO~*kZ%mB{$=YIf_0ILm$l-{-LwRYk40%Sh;TE zXC5vZorHG9(^T7UH?dafJBp!I1DhatqP*Z5`AyNQjcoAgI1iF<)C0|KD{%ck;PNBj z@>AqINO#`G_a*tzGx(gKU&ZLrS6L_P)w)qRn|r2dUk~)cd90#^0i6XMPG=Xk(N2Ir_`i#>1)98KagI1aR3d&EBOb*=-E{8c| z6+KwCW^R{dOPBnzKV@c3& zlyMmQj_(}3%gKXjvx55?k%>k||5kOlg8fy}Wx#RSOy+)8J+8X%zGub}Mc+{S(hcqg z@APgB@lX9$6A$%ZvNmg+XE9F6S#O-{7^lYPAD8|vJ`n!TAkn&sq=fJ>iBU6Cr< zTA)0Xsp1XPH~l;CWh!u@FTq#*!l}&@*?EX@ym;oZr&r(7&^b9!uzB@O4V_c?J#uqo z=j_1n{;S4Vi&x!Zbryvp*S`$h^n7-pU=48$>Tx=`A-Xp#FpO`l&U)TG%=*D2cPWmh z{Efr(vHFZ-Pj5bVZRcv<+kDpA&Nckj)BXrz@^_va=$!4gzlmqd0>e!E`W_o-jP9LJ zUA9e})6>Gcq6xvuU(4+DQ_%WZgB2P*-nR1AGpwG#SbyeQ68CBrBqjME>l>Lm_$L>06tcRhC z_FU(kwbbv5Q+Mhj_nj(e+OB@z=oLAPRXE?pJS^pM(*ALH^MIe1H5!V3hPo%d5uoiH zjmzZ=+4zkQW{2^?5L*aalkFn*0A}AiC;MJxl6_|OJ!_IZD*GPq6=vV#y*%y-H*UI$ z8X_l|kEg%HM_Q@GX#V5Nykp*Aewr9feQh`Q#p9>uyg;L04bDgPpI0kf`6Tt;F8~)E z-jK;ftYl0dtDZ>E(YM;q7QY{p*?$6Vh2Z5nzZJg4_>Aqezjg*SwySc8dBF?d`O;G) z6YqrI>%D6Cy&B$&Rz^}|4g5(@(U5e+Ug+cy{C*|8tIx6w?>cM^8{l1cF~*hfqjavHcjX)2;^nnD{4Rfhce%V} z$Lfo(DQkyThG8dfvJO9UCHzXV-TV#B>+el2Z|Ng8cfM)`GR_z^I-g;zf{i!EzWMM+ z+NpE+$qwc1zDV28!cT5DH4sCl4yO2tQzJ06MmE>&wC|nWp*_svDT?tN!oGgtI&c;~ z>+u4wk2E!ue!;gHo{`^i0XZLfPyFESDyQBWxVod7aqM9n#@<0LoRD*-dqd@%(XVWl z+n8qz+$i}LdK+>^I;L`@Ye&%+`6GH3M(0$mWiQUBy+QFMtmz_)x}B1XId&{@;u(ya zAE9OxKAIx@4~)^sYw@fuH7jeReO1)$4w*7qqg-wjvk1eQ&#zB*f_wZ z92iBvKuiueSkJwu8WNZ37a#B`IvM!ki-16~xti(Stk9dV_lP=milVUh6yv z)hpAU{7tlZ1bjFQ?I)o782GO95>6AIGtL;eC^#y9sBtKc(+eJojt?=HBl9DvD}dc* zqn{M6xdWUTf!;7XFk+4DHR%$fFZ7Omo6%DWd4?{sZx}koVZMLqENi;HpAyL1r}~*w z_Oxulet1f{fpAQ49>e!K&q2DeYGOz~{%iPQQnb!kHef@|hBx+7pQC-$`Yjr_H&*{R zO&uTRG9`je1n-hAbQj~+xe>oX?`om$^AlgP7H@`k)w_C;p0@-F4ewH(lJp|oyN0^d zcQKz9-dzDLYrLa*mWQsj7hX7wc`~QXvuX31S=Qpba|4|RU4E!%vjcgiO?{^}b?+tf z9eNh;3g5&l{QU3<_~BhXe)t4DLVrKu^27c`#3kVeszK6Au7b|L@Xo%t-doRE?s{JB zz^Jzp9?lrFmg1HDiyAuX1J?D|uyzOGX@86VcZ176rS~0SuJ6M?J$#eSrS}V%!%(>9 zJr}-hq@6zG;t|Oi@cmoYB5Uby8T}YsV{8d`Tn5iHHh7sC7ql%rl&`R1Krbuoc}aEZ{oUrSf1~<` zc9pOCJ79~Q2;Vh+oThN`WG4^8zh9Hjcd9YKgZGDb&9a}%;=v%`{mNxYuLSqOlWbn> z=fmGT#W;TYA>()hKKz2qhd-5lK+r`O?MW`nZh7*XtlJ`oMg{vm@XV{7|1x{tJiPFq zh4W_CSvYSRzEHUNO&2fjf)BvsJ9SpVwZe;&v=1-u)R{6@A;($*!`IZXHxb`pXX6>x zC0@VY{!Peg7ODyc^z%F1T3Fmjyy=?xqd3C0$Uw zN!PN)^y{_pGVLGaTW@aSZ=0ZrBf#@YWP_hH$`Kk7=5+|E3SgtnmyI_sZ`4%j1 z_Q7&BeM=wJvxM+P`shMS`9CvCfveYF*8pD(b=>_^bWg@0-SdC`I>Xnz^)tFBxOIZ= zseAN2XwR7B$U^b8+>V3vCwP7sZyAb)Jsk=j@^<*w0Vf~K$3xa~Hrk*zv;5B4I^=3% zo`ddTht_4^L)Tn3#J#U)-{ae_X5Z_YWPgQw!VS-Mm9F|m1@wb&KUT<|e_t&EzrA|6 zP`c{b1Gb>{H>Sx0FG1(5VZ8;PD}Z4fIZ$#_s4NHlaw&YJ(&a0k#=hnlauz*mhJ1<< z+Wp?&GH!G^#TaJM7qqaWB7#0O72hGx4FpYC@OZNwYtu?R^P+6 z@byRE3A{Zcgw`sf*s#skG7sQVpmq(GS}2rN-&o+-)L`%x30>0H1GotyMoJJ*L+X) z&KvXbIc0rk=ykMm9M9``tK6m6(d6_DVPAek|I#PIhx>tJHaq39IbTX3GL>* zr@{3r#5Y90*tiZ(bOr1X`q6gwg!W)BC9nn7vnMnOJ!+k*cphsD*n1gQd}<_>0QPke z@(AQNn8SLbV>ifufDT^5zJ-zk>k`E&in(6E~4 zMEY4R`$WMF?Xeif*$47t863HYI0E@U+A|}4pO{rA`@uTN9Wnd54F0gM>(`%i_H-#% zB@Z8mVw34~emi`u^RF4tR&P(2Y}QKV2QM1f)0N2V=_+YHj^3%WIK^vPh&|ce{@@ABP{!4MvDnu?L5-3J&_+c0#VEyVp10GI3(z*;hS0E9Cr97tcI? z({AWFv;Qs4nx*TZ=lYz1I_H&+PVlMJv@m9h|&;JKo=PeV7*T1zMB)mknD zFTfRh4KQR~V+-La@C9PhO|@O;+rkOOvl6Ay6XWuDJu&vOHtJqRMpcwGkng{2RE`<1 z;n(2G!1{w@1N@mdi}-pMen8o!%8joIagDAQ1wXUD)ps$zH!ya;O+P3(qp_9Pie;NP z_=CtHaNOi{z;iC!#aS21DPtai5uACac|J(}KYjzZDpxmv-X7T*81d1^tg)&t)<+rR zbm8m}WB23i6+WE3&Brf}R=V+o)1?FW>RJ6b@a@<^WZI9*zdZ;2Gs&9NkFc(vLe5?N zC?}UXq1Ykz#~s8>$*1jX9K*T6*ePp~$=S82-ns(ZEv1H}C#%~6cHP$hTu!AH(tx86!k(7jyi z`YvKP8t)$7t$l{TdHG0_% z*OcWDAJ+MWk~eGR4|rgLy^VT|etGk^xz3rO4bPB2cpW(q2eHrire;*2bC7-F&Oy}@hY@CU8a zr>)~E9xR z0cucNyK;+j?YY!MxSew@Ucmpn+v1E2aN{7j(F<;nBmXivK?mky!))XXjP2~_Ze@>w zIR_Jc1^-)Te2vu^XFpW@TK1Kh96Idqph{;_I@>3SvS9``{jE%3^6 z_^0+P6vKO3bhat{NNteY&9}suz>7ZcwSeaZ0seo1GY!zE6}PRSpE$n%Y+ZWTxS@3E zg7zqT6E}fx1G@BlXk%fPE?t^4P#g!#c7kG;jH$4 zyA4|6+j4vk&yhP{0Uaf<15Df<8~5>_XZxI8-5xt(Aja9~-pgl=3#9{{*YTI&jfcnI z@$qj@eh(X$CBLslSGkezZy*=W)a-K3PVc@OUk5mXFN+-7vG)8Y_`B9h$yXVzg-+N9 zqa5~^e)2){p0U(5xHiKZXPNyB*OX!NP4{Bn{@8XO{=#9j<2`%O_!I}nxb=?e7Rc_G z4`3xa(&?*_gY+XiCCL~zfM?2MkSwdNf@hez=Ee47;M5*uaS=W;jm7gVnwZI~Az~)F z#~A!Elb&PgON_4pJM2aJ)wj<9PxYOk?>hRHu9IXuYw25K@cMt7{>R~)kY3>J2hv(A zKJs1Vad2G{bbKbtTLY#hj{8Z+x#|}?L6WD;v!-;X{DDNb%c7d)z{fSTRta~fNObt=b3 zcjSko8xEH3&OMDS%vdxw$r1niamJ}yQK}i`zwhK~=eT=K-L_p?i~|GF~Z9!(Hq7&ap-y2+rq7(;`1+_BtHL!p-Cq`|Dx;XSA1Tz;a-9kWp|vvin)yi zKj=^KZpZ!^YkY6gYuIyTD|YYMW)W=jHpZS1j9AANJK>+A8yDw9D{bKNE96}Ww}`!% zyv}lPDH_Dqr=MbM`skAPhy3nmLwpwjK9c<_?|HA;*ht`%rz@DXgtvDE@#z8gllgM> z{gP9Bf3RY(zgKJj`sJzS4Zih0mA$Vu`Xv6{R@TAH#aCGp zQ+}{?2*F5oXJun|7pRuxr_b|+FXH(x2*<#EdnGsqj>MLNH>{oU_YyymT$OJv2~SY% zZKDTke9+1GuE;UbOPqD9s{+4Vxu7^v_?%SpR^*b!*yH~QPVk=E4)cxdXnhl-pRn({ z2I@4Kdb`4r`Lq{|n0lA7Qg9KRbYSnu;vCvt3h$SXOY8Zm_(h<3(cNCb1iHFPFhRG_ zxZ?wKH{S6z%!1z9pgYz~`%)8(S#(#*{N&qCK(7w{MRwl)Bm5Df4|q+i)LGv_IiSo< zbM((o`JbY}^w-EE&HB!xE5RiDyRp#6Q1_$a1%(>xc^&72v!3p<)YV<2f_wb(|B8P0 z1J7*w>0^)Z$DyB(Vjq7}`hIMP@$LQu;~U2Qj6aUdoZl%OCHN`gsqjHLw7ImIME=Rg z8$;)hV_(Rw{lN;>SGb|NBk0IM@KV0thkim#nEQ$euLiG*vh+3+6GjG}5EItD?3l20 zA>^cJ{p%jwhzE(s8(bs)5+94NgYhL9tMKkFU}^AAZ*)QPL&Q@*hM4M&jFM8j_ zLDv`U(uL6@7-y2UdT2-EHT&1|nIrmvbPDAsduI$WkOBSTC3GQwyiR&_CEv~AJLwW$ zee8?u53i3s-h+?ZUv_SLRsy5>ov^8rZV+V6hqD>k8%z^SoW!>o|i8cn8<$|pTRomIXXv@HB)S7 zIXdMd{QW2~=Dx^C`_KmmU))YiXjmYUQk|n5>9X`KTr>Qm(pfu;wr5=K)jN0YX)lar z_DC1o?~>;uzp{9wbf;A_og8Y(o(ijF%Tk}+keo)|Cw*vKT*%ws}^{r{d9%oemHe&@;Y|2=UH*Ay?i$}1Z?>oSSM)B%dP(r zd(S-{@8RzKyequ${(0*+g*`-n9GS_lm(Hm?fgILEdBoZ?S&Y8hz&eViY8vMn+N#|+ z*U-Z{;2~JlpjU1K7Sn-6Y0x>xRyIrH04%&c*bblY%L2_oG}q{k`#%_~H@;!`xc&Pp zbED;Db@vj9tT{H;47)~qsS769mF#i83Vp)tv0@(h^~UtF?t963V87X` ziuvtkUGE0QJSkr)OsD?tKq;H2(2EoO#s` ze=~j`eKWq|V#_WUZ-m}D&^f>j^nYj#nvQ6n@CRk=6Q;H%GC_8UYHL2Q{&;f_V`#+= zk&O20IdqUCAv>fM*cXEbTd_m@bgr@eS6Qn0*C)YW_6!Z;zeUIi*{`>}$Qg#{UCNJt z7$4jEu95Zzc*{h1%cHEd_Qk7SMVNOU)ZVh2=ce>qQ}3&P!(7D=l~d^-(~DOArUGcT z0Ge%Iwd&YoRsp%b!Tv7rX)iQ8Of(B@Jh^1(o;H3-Z6jkhf{TWy0*@&3%7vCS7V$XE z)4wi1W9}X;cl%&47{4<5hSnrL37^0%1 z#DKyL7uhA;ui$9}Z!QwSUFg&o?CiF9uhY zORpSx`3K^q6=kJ?oK1;?mWji*UlU1ntcs+LSjT>{F!eXDjd=T>0~h@Mr|&!w=z3=j zx$649pFL`iTh>#{{@(h#YSvUfP|?7)>x2i&8y?oNtR0^dJVNa!4o1rS<1x8f;H%ow zx1uw_NjXT{;m09-cZaQ>XQ+p1a+J;A=#BdOu$w=*n%IxzzO!d%SjRr!{JlPt$ZGX_ z8S60@JX2lk)&U%I=y50>7S}nE@QQcXlkCw;oUz}>As1Kace17Ji4>aS~LwkW$A8YawbJYBfcITP%l3zNHdb!;N zgU(A%P7UBc3Z$a5aynzPb2{Ut0dhV9sqygc#dvb?Yqy&!0Vs4Hf&M;r`BKoLG2yDS1{;r*&3>Ut3Bn@zHpp4XTbOsVU|X2`u2~PtLCyrZcd;>u?OWGR4;aXY(;hKLxR6|nR*b~M^j+`5!gM)-xchu ziEm_`Hp3q_f|rcDl5taa=H3mAyOwd+6OU`N)^AZi8n4-7uir)VBRpiy(cgA8GNzc~ zY>Y{IpyAE#I=;^MJRLYb9~sBGltOFLi^MDCKM~E8=-xv77Ti-DSZ6AU&ZGy9piTWc zcw6AuNm+l(r^`8IoMoX}Tvq=t*yk<(ah&z7yhwBajOJs%sn31%`AeR?#yoz7{_+Ov zcMu+v0B`!(tJcq+p10U*@mnj^{~mR8yYtI_M;znb%Dh46oUcLlDi0w3ek2udq&?=L zzBCW%sh_wnYFcLZ^%Z|#FVNSL2(?92U%=TLETqrZwz?3w}J%Q|O5_H_^XYl?lf1N?7@dL=P%Sa}w*p{4&jwu{5Zp8M&s zX8E&pF1C>Y<7m^>ReBALgx~`@C!IYt9-j`n{4k$38LP+NG|o*LXFhN%1a2dNFJtud z{i8vre&L;C;a|bQ^$Vkq($?eXiRfrn!xZ}}WJisu+Z=>1=k%AW&KkH;1I74623tfO|rgq*niSlRiENvY@9z3 z$1V<9CO&0qr=ioTZ|RAp^qsic=~MB@cgbgY(6UDT;jJ&V|Do1;briXRhUU00TMpZ0 zzp3d3t~QPWHX5^V3OZ2yPWn%j_DpPsZ%#jhcX%EHH}#HqSe)mv+0gy9=ne(cXbtde zAv$1)v0Tg?ZgAsx^Qj|-TyP?JG8O57{wjD*dFfnaY^sbTbC}5Iw5SjPw(9 zbN0%!{!78Ti}_u~?|go1`JKw|N`9dW?e`MxsCEcC=wh{@xSW0yvIppMJ@DQPoJ*-O zSo`*^uhq6%uLjAp@!F6B}PN2(1~Ln)Z`p7g7)L zs`k1uR{M%`1MSH*tRFV0>i!&JO|202?Jz4fY*g2O9tnPD=REc!s?XJ|Z98jwJ8N4K zwd$-w1&Vycw}fe z7;2Z>yJ{DV$?wLRcx zxFL_2AbjviU=8jFw}k7X)A$rwBgs_@-9hw#%pY5^6f7EsG@YW@_q z=Zyjf!4>V3xgC1do>=XdnaG%WcSllhGw&F<8oil!mZ9U+Mp75Mq&>pziE$3IB# znim8=`PRfOTwc^A`FJz;z{>=y54{(QO{bvEF)_xB#5G;EeX|JCD*xo&xiDBXo*?Y;wcJD^MyAOEY!?Oq7`?;6d zz0LHQ(EN)!uBQDsuqq*!WFGKcq5aR$xa`@!8Sn-x&(LxLdXn7`f7A7-Do z>K^{nMD~2h-wM7c@48Pi7rInTL+!ta|7A6G55v%*Y-rgK9_>9!UWNGTYHVuJUe&6} z$8d0rRsR}``re-P_g4N6(4Ff0u?$}d^kVVs-WxJ&s(3NEKt{i6ke&!G8~%2!k!yo-5`3wx0Vm;Yv1?r( z;^O4+K*nd-zv>&uj<6@LJ`n1UVw(tG$mXto3tf zI*ooXFT_0~*Yi5w=WlrJAlWV7xcJkSOEdh*li%`@{kyMrhj^B!Z%i?{QW^h*;urEy zkgM9ik6hI@?2kQ+TQPXumws(x9A+$z4$^Zgd|q+6BF0<_yeBaaU@5v*u0%#pa(oaO zozVGPb-pIBr;vK7{jcCpoQ*EPnF@Al@Mtm~`v3yuV+`^BC`_Mtm#a2T7#!z0-Mx-YxwAm4a@%O(+i zh2blDo`6rb@VkKDo?vj_2F_g#fyde-Ze*v&mt5b9=&~M~{4y|nbk&NP3;2JSntwWH z`A7pez}Nl$8!I9_;Hx=`~AC$y-LQ^P_MT_XF$SR z_B@4dJx=$@iPU{-s=2TD>`!@?92ddf%I%MmgXzfH!TP7h^X4TPyp1squYcuG`{Q*U z{{1_$EkcgFnIAaET0bTI(9~`R7v!gCWX-D?PndDW8K-RD%fMAbSA4S-Ju1MvvIT3v z1AS|r@vWoRyZGwjwf~;vZBZ6{HIJYA=Hd6BX`JZ6rd||2@z-jQQx(Jr%c%wF@HlEP zlatyNbmB+&I`=Dnq~DJcKXPmWM~26rLK?q-^bZo@vX;emETxzeAHt%F%EdPWNcz3 z;~=)6I^)Jqs$BCA&bPlaJgvP@yZU(E$Me4V@Ga=0Qnjl8wG1CUvUfJJ0eNTE8l2hs z6Y{da>F6xW#4#hrM}EGo^R)lr^vt}-18h<)(z)azh3 zK@4Bl@u^Fkp_tFaNj~3^DGp%C}_RIqtY?=nK5RRQ1m^?#DH5u1_`Y z?|+1GKXj6DJM=#|Pt4H&#Tokls!RVzyK{}dY|7urUzRt>Ulu(G4piosCEr44`YmhR z2aFToUi}&EdHAANptt5XIKK70*a3U-txuw6M;|`)ZsS|emv0^3V0`Q7`vbmpVmX6- z>#>_@8#!2VIs47{O(VCV*&P)P0~o7+C`#JlN7G!)FS zi`@3H`xeJ}uQEbhca9U=C6^!mXYE=(20CHB^0|nYR9$W2DEq6I5J%Ze`}bI>>rXN+ z7;72<|N<#)k*GGAQxxj5BMRresERI%sTj>_O3Rr zs-0Piz83U6;~Uz0ONuy-I1Pj@t3d z58Zj-=P$LJ+VKHk0&o5H_EmS7+VSPcUgAJy^>F~Ru*2Z*X!b99JH5w2^NU$R%`t#r?LE>#)6Rl>G>H?CRp%IdYo~H^b(Vp6Z!P=~nz=OI44i=*o6D{z7n(>xR&gV=cbysF&A@w+6%yNpWsLz^!0}S*)O^O z8vl_mF>Id%b)b`_dGuYsQ#y1FYgW-!War(%*%{DD96G5U&OV>?*mmKK{8cab+uKcB z-I?|(iPzeqkt4f2nRyfNRv*G2!MYk) zR|9L&k8pDYaMtw%;EbO+)k8Z!r49KIV!_*3L!DpX#Hx{VJFLe}v%9d-lE8Z{GSC9n z^3yS{{S_BEIRDQEv#mA-lC3==;fc9yE*n{wU_^KJ!myQU!BycIgfu5YTv zuJ>XJ;tOAV+_CYGFwTi5!SCXe+>gkANDT(@@y(;H`#spNJsVivXrV7fQgN*p^L=Y9 ze2KLYPwrwrK@Z=wGEOt5`pSWKle}B^K6;7!ZtIz}^%gN1=xU%nXhQFO-T&Tb-gEH- zI5kz^V~}pu`8qm=ixYd!;2pl}g8t6AWN>|d<2U&&@XoV##c8Jvx$37!`7s^-V)k(p zTgmPC$yM(UjMK_>+K$nN|9jKd%)x!>9rR(&AqaF>taFt5*fG{0xmYG%Ab6I*3)eBu z;_<8{v|r47lK=lQ1APt|D?a4t?2ep$n)^fXCc(^`i|k(Ta&gq1i}-mMKAgBwytkq( zN#5*Q=7fxTO5f?ev2oqC_`yl~(0fVYFXIq&v`oZpH6?aJ`4hWLelKT)wuS7^{Sz@b@g8zlB(o(GZ#v1iqW2LK z18=VKt<&G0`T9xjFFwirKRwC)1t+EpvTK%Us{mvO&J39Mv4)EPMRh=p&MU@-5Z7zNLECw^Z-?mSQu2XT5Bk zvF$epQ_btxFJyRg7Qckw0{>#@3qI`Qn~&U{;g`q>Lu=9hPVdTr*KOk$x%IYej0?ym zi_I9|mpiIGep&DE%Pw*?dbl2nUzWqqviRkG3qB`)DH?tQU(mbLho*%iR>45N+dh0( zFB{)gu)hKQqYqud$#;`Z80-jwXOe^2a_n#57fJo`1IH)f`2*UK?f!hY@_h8Ok=RG9cQxnkmKP$U;S0!d^70h# zcl@RCNy zC*Z%<93Em0zXB;N_n$XjC;v2)zgI!Om5lj2(0L`YEkgh0^ijpyWVgAIHvh+M(`)Ze z^fbRe#Q5u&qxi0;TdNLVF1Xx+4!0UyM(^KI_aO07c&TKl(NEwjQ@`ToXSzNs>GX=p zIXu(#eK2>$2sOt>-#VQv-|Ho6Xl470#IFO$j9hZ9*W>$Wxh8P^TKqc3=ALAsV=G_4 z@fm&oN5`A3++$D2r-~2e^dDlqjc*D0Z{F*=V8FLjs&{iT?=Ao~!kIZ%XkV{6x9Ijw z@;{W~r*m}wqK8Dxix7y_6 z4nB`!cCmG|T>3Y-kbCE*1M7!(ypx=;_nqfMz3(V4+N$;R-=E;z-^%@{yPk1={bO|Q z%O5oOzQ#IxjCFJk>l6j{S_AZ;9pN{}+1v29{M}9H#lWMUF-7vM>D8>0_Hb^%4%z(- z{!rGVVGMZ?v@5+y`0~4c^ep8OL(h7*;oQN`f5Y<_a@wIs&Peod%!Rk$sBvnK(cn3V zeouV=WbJ(H`@S`MWm<*@eU~+J{AD>#zRnEAyK@KmAEZAgo2^t|8!?0y;3zq}(ISp* z9bEuC3w{x_3j*u599lwr65YNu(6}X0YANT>*v_M&SH*hNhLdjv}C*Q6qxZMQH4j?`N2S$JW)61c=R0^eJlZzZF# zH!e+Ry;z@~8N?mw|G#KU_eyC`_mz{NXA|J7!3KCMYZwH!*g>ffwxxHz9&%}aZEk^k zz8>d|GKQdUoU&CDwe)qhW2^2~Km2X_;hy@DeVQnu{bJf5Pd*iMf0#K1=Vay!PG0f^ z-(s(&awEVAoGBj*jK_dei>y>BuuXpEo&$5HJSl!O6<)sB=*!l0d^y4Xz-V|mFlK+f zsdrsxP8=0tNAaZtwj!;8#RCa*8C;2b$`QT7tX-9VX~F_ zChM{OpEn*D|N0a7T!K@>-du{$HFMbte>QVr%<@raF7h|5+c7Yg_VHtMHeh~xa%@3+ zcocJ3=*{7C_RqXIG@NGUAe=IIp?ry@R_bevQFx_zv(B*68U@JLj(3x{ZT!M#*=sw9 zr!56mqvvyO%!2k{lhq!2Akdyz$hv_8`c`LwiC!+#e9l^Tpsstnna@^lJ`I zn?wdQ0;_I#T;m1MJ+h`-z98}#8o_JYZW7&ggU5~Iks;)02nY0y{53)HM|3t>qwrtf zRh8O}z|U*j;hRqE#Lxd^r~QO6is!_xwaCF>zBgDY%j7$bvt#7@Ja;d=(eS}CTXtY< zrTg9*_?&WgtPAZ}Y0!SgdoLoJo479YFFpKGD{oUD=j-%=13FJ|qm{q8C=lo`Mz(K9 z-!3rwOM)GfIZIV^Y4jLaejLdN^^U;lVvMsE7!&CLj zWg&P^A9M1d>q#E0^Xx<9yToWyxDlgG!xv~@xTf{@>?6gVYR?BvzBHZW;ycZIpDSq96^5XuYf0t*Gcvc;7Mb^p6RVWXVfXqIPfm?rhLW$T&WD$ z7kcl4E1pg+8gTNO-F;U0|GR?AUYA~eD{FAF`ibO5p^s5Of9UFi+TVJo!QagOhjQ>W225pxsz2Fw((}YuJ@}OgcY&S3UEwf1(t~+5 zGAs^03upHclh?I}vqj)`4te`0z&A4bBg1zP{t6br_)HJR(9e)C7CwT*12C=)*mrnv z^W(2z%vuPC24Gwouy6I={V*7JXJGsb&YU^)^*}3%ijx?K9h* zvpFBe&d|QH4bbFzeuKbOF$CqJc=~aLe8f$s@!f4Zzx3W<8<=?*TlYoA6lE>{o3)oo z@A0oIjQ=5%FQff#=uI=S#x=0d?cR?!uBHliZ`)bf=cVdM{(+hEx%xZcb4 zO0HY#tn0Tmt=mFu`?Z5yC*1Qe(J7rXFd0)H^Nj)%orUSaVyO!YoqKHTw{ADz;X=XS zjDecZ^54Z89UAp=71GPV;YH|;*kD$}1p6xTu?Ep-qeG){@;l;Oi$*06a?qDW>jCIA zO0L2tU?Lq#^%xvkD!yIR@xgg~2TxmvpBH~jQ|)gnOm2MxzGj_$k(}@3?U`q0zcBU* z_hjRRzGKc&pKhM9M_9jg>=71jCumdOst*${!lp4gOt%wbm(3$u));G)o5LB~3E=iu z@i*lAvr&C!b{LKMY}O zx^Rmy_i}Xmdx@WDuRs-gcQv+G$w)Wnh_z-&_iBwp`Dy8~1M4cAC=9;GE5Wl#mRYY_ zv!;r>IyrMq-_ES5&dt-BYTvikR6eVCWdxXezBp$s1!FHiL~s|qRk*O$-y5K718;0T z#l#ff9>go}a(LyGLi{!OcC^pC&iLgVez_Z;PYi#L_@@5<{5N+W*uMOUcP3o($~&PS zzKMMi>W|4E=kj00^9(P&rNnN7kBp{W^oAMuu=mS{ee?Iw8*Tsw)X55fBZlXUFK}$T ztZ%(G^G2wRhH8DgvsV6|Hj{T7Va(7x8Hs+`GAA{r{L9qOge6*b#AuF1=Rj?53fX1;Oc-}eu7$d!krK_K)m_m z>_2a+K>kTaE?x4`_n+@_=kxzuIqm5a6q!agW#R+k16AyQdXXBXvvctaqaO!|ZOne& zx;}=i5N(XZo~>@Ypv-%BF7Fz=ZghP1vI!obKgmeT{pa=flIbs_zeqoxT^up}RGXUJ zmtb??-)0=a z8_#M}|0`pk$Zk@5;%#B&apFf@i;Mmhq za7lRYe-3+E1Mf}A(6Dfz(T4*@KMUp?|N80&;eQ#kp6+y~qLY8vA>C^5SwMW6fD*d?On< zI-}gwq}Q_tiEq7uk9|37q~8_rFa55Bhv5g_QLy%NHaTLcB>NK&@!cWzEB3JOwwFD) zePg)}7T7(HGlxX3J^Axa?%*DLt5N3DEx#2xj78`cd&CErTMl_N%3Hkk#`l^d`PK}F zHos_MGh@&%qHh2T@Y1<|u02xs4oaRC!FIrFi~Pmk(%*a1RjCvDKEH1u`_j;GW4@iZ zk7uR1cE}w|HMH8G{Hca7+UuKLUoCYp1IS*%Trys~Nq)ZPWY;pr1oKSNrgB;2A3N_) z(DAr0{aMfCpVRa4h4_|uX9D~~`BgFCDc_EAvvSeBLg-%0R6j@2}_me{0^J^r_kz(xJWf_W0UUTYYa4pJ}B3T&qm* zSI#kKn2)J)&w#!Z*u_RWI?h_zVdM!2hQe*bS6k0vj~B4-vr@yS5vSkHnj|8rSU37e zpBEz=W_%RO(49*(CSpv|YrQc&lr^R#sMnB&~zFm9e zx<;G5as|XC4#O+IAEXZYDdrOc*Bre+hka(`-}FaP`+kA0>Gl~T7p`ZG&a$_TzH}XZ z@;nb-*$0nFk6{l9ddF_$@txPQFC~}!1o%(O2f?GQz`lnye*~N%&v=*cwtTgd&Q0ND-bwM`j#JD}`-sZX!M}72`YP1f z8LFe}SWPv3Dna0T^+OePi;@ zhr-e7Q>_1Ao?`6VS;Jn|`yjZCj<#5P1+TV2!|<}Q z5VS_k)3O*mAo{CFYVrbddGCp|zyDgmt{G{$zx;i*^LGP#KcZa6$j_0@<=D^gN@V10 z_J>8`A+sZnZ$td!XF1jw-Tw){vL^)J30xbV10Gxhei)qPom##X{wVgRU*#|Q`NdPr zL*Ml<$Iu9?Y#H(RINxk#F8=o(ci)2tmQ@pT*M8k?yywBA&3*2tFLkp>P5FFd3itqn{7;NJbi) zH`H;X#%r#1Pc-QA1=ewV0eS{HO7AlAP(N63-~3hMXFkH3X8468Yqe+5v7z%!-4Ty| z2Y3eXZzEoRBF`8zh-W-ZeE7${ISs`F9iDL;drvYv<7cmAct%)y2XI{m&uISw^9H_k z@C==)vm2gK&Yt&W@Ql?Lk=G5+(7ITosPO_{TRqKMT&*)(Mp%~?o48-T{p5F%3$!PH z$gs~dhjHjVw8^=*4j;<3uD9^#r#Y*E^&(zZp4vLe&uh*c#B09H9MhaXu@)JYk(IYV zSMY_gS_^X1#+9cUk+Y4+S;=lM{(S*CEb;-}P7H7&G8(y`ieLxgUxij2UQ_xWxyjti z;x$d-$VB_Gc6iMqc+CqLUUMV!i!U02*Sz5HnnL=Fas6SuW*%$i$)%yzEHh8#3-xiBmbxHAyFsYXeg)%{O{!Q-6xpu(|7gZrZ>~3+Yv=wQ=Gw_`+rahEd&UF((nGvG`o zqnAt=C7mN=WV+iiz<^=yoE>Uj@mPsI66@Q(DaWv&jVXNs}v*Ndq`SG(f#!TF~dGh^B9 z)7=#F@?aHbuC88Y<~~>Rep-9J=ubRkKKd*0&jV)V%y~C7TVnK}oB?>LjT!ED-Z#== zWf%1TD{M1IZv)QDv*2TB!N4!r@d)RdqQC7>US1nH9BW;>~fmw;#Y zk^et6!)J$*0pcy{NA{|C%{Rf}3SxKU%pdIl&LQ%NX9J(*>GUPP<5_4j`Jf*J&AznI z1W|E}3i=f*y_3I>+*EyRnKi6G zj(<~oP>g zj|Kn6VM`V9H-6*c6PRay#Q5FivrCMQ@DA`92aXp-QlEK06AwSy?VPuBg^4krIS^kh z=iS)FR;q$_;*-!D3t8hCbHh|^jpiAl2eIm z;p0)S1ynRrAw zbBzJd2y?Gsu7dM0{Bz1(Dv?|!cTHz8MtMdI+O8nqI?h;?ccVH9>yX*#dHdx08#1fX%{+rDCj2zCYOAw#WIio)SPQF`e*<*67)Ng$7SZ(r+GWjvuuYbiAe`tPK zdQmBO$C@i19Rn{sKlWcHh)0I3r?BB)-J5PXu$#IWQxs=}-m0Z<-*JvzLfkjOx+F^@ zDf1<8JatE_QunZtjZScj`?kHtw|Nz|Q&J;TO(*uio9D=d7m}@45gO zdvrG_-^jo-gI`6(*3nha8-BvK!`8+2tedKiRRxP)M?ZPnN{+R|=<$C1fbZ?M-kfR2 z@Co_vHvm_w=>j{9|H*$pqW3SEW>+IuvhMG&BE`mEl68N-b=L^HJnMZkp8GGetFy*K z`)^zKPqepYyxo7yR#wzlo3iyY~Bn*QzUK z@sIJa5A$48Q=%*A*2AcD_?L4gc;>vk*5OXxvvS4N|Nqb1xxhzN-g*DbEeS@13JQvv z4B?`J)vgFe+-8yiQE^wdVBIZlbLC=jYpw07#cmQ4E{fJnVc&G6yWB*aXxl8}w$}C~ zh|*eZi&(e&w!7=xl7QG2)V^jA&HMd5=bV|7A*i^!FP{&e%$#}7bNSzI&+~{*moW#` zd#GkXKCJfFb0*AA?K`YNuE{>CsCv&np91>UJcQe5F?7q>r@_BrUj_WONBJXYrP|1> z)5x_Ex2=M%p{>clVB6c3*i6kk`jJrwp!X@P&)99eqkMacc2Vj$8adNMG2LF;_0jI| zHU|z-TUJGo7m~Tv$ed)9xiFsu^QmDzDdv;7m0SSx=^1W5QRah89Bw{-?|f*NqTO)w zd8uGN2L4|fZQyV4<;>H?*R2k|9(M3$=ZEb57uQTbe*Z;cH8J_^Rfg8ZzMs)O6@V%WnT;$OOo6nBDE_t-t_6NQFPATm7 zhL6|pHS90mUCPSw8A0py`z8s22XZ4OsI# z`7zyzcVshgcmzI~%R0BcF)60KEgOk(>G>Rq?QqV4%(u))y_8E_FK9;SU5;z{iCY!Nr?7xt?K`>Dgget`q~ zMqqyjuwP-reyU*4Z$n^D8v}dV8Q9ZySlA<1a_qT?eG&Vw zE*742=5a~a62Tlfo|@0IrNI2o;bCrQc6eIeS4hjrBg4{iw`h5Z2j;SSdvC&)Gy4Zd z#x4~+u}K7PY`*RsE#Kt8b1Cq=6L>DN;rR+R6&ZdT0#Di)c+$?mleWXcbG8Sb^Yie` zT+C+=Eq@R!U7a?Z@6ZpQ9)O;td$ZxZEi;@Cp*;Zyprh5qEu@?E&&d0zU9b6qsR_y7 zubhN02AfYI@6Xwk=-XoSZ9j1AgN9As9~}3!5x;`>Y&*=oAFr(I6Ucym=%PZtJw-Yf0QrdwO@VLAoHtS&`vG^SC8E45d?QbY zNwPO&WNe3?!~61g63CUWRlmns43T)UG|2xs{15Oy8c)MF^(9uk%(A|$oR#i4kxbJ4 z9`_=n8u%_6a^v?`LMsyk`S|_4$g^kP8GNdbaaK9>wglM;KU~(wxT?3gn0NGf3h%9E zO;d|`ADP&VPR_$t@gAMsY{!FaU2fMW+WQmqT)*4rc@FuV{$vvw`hiYvE1;8RNB8IH z!SOH4<-_Fdvo!e5poeCIW4g35c`S&e~fg)Gz`Pw#={6USQ@iyr^8Ebs(D*8xbclpRUCZUyyj4L{jPu>Jh z+li-{yETB%Gt@>od-zS;3fgj3F=rhrXB0eC!`#xm+bg;NX72mn;k`8U*b9D*ZWJv+ zQyJ*5i1!tHQ`}Gaz26{XwNWe!+csDR|V$ zhv)XJsV-T5WCJx#`Yhh^!y6~09Y+><=_$#2htwdj_O9n`-&ZiGMJ`>#em(gE$|GW1 z-R{HYR1OiFb72EH#1wmU+_`RLu1TkUfHfTs|H!GuIT1Y%GbgvN4*Ck``|1dfy~ws# zPuQ28hxM*60S^csFRSj$jAeEGg3li{R(9J;U=Qy91vwB2MY3GTZQ^7$KcF%U%lFcz7;o&yVk$%S`{>R^}5XUg5mA&zhd^ zr);0oo_U73r^Af%MjlVhkA3T>=iV>7p~zmx)vQbSm{a89@vwD6em-Vh7PZvqZjD=Z zU9R0e>xTUJX1uTDznA@`)9*sNy&3;2`SEE_?vDGMh>w_lZ?@aj2|hh}df>ed>pBmf zAE96I!97Jsri%}5<*)G3RCI)Q7nUu{&9B?~DeEA=q5e%jYx4b+Yn*lueSXE6@0UFD zeJRf`ynomlm<^sBez_~xZlCqosj*fE@21^LcCEcx*LvoMEG%nx`ftgt^RRUf^hQ0M zGrw-@M|t>^ttqnc&->zIDKt?`$b_?Fr;f8dwLQ`xtj{=em!*Fb(=w5R`R%-@0c zrd+!Y>sOb1>6Y4g4<7%m_{bBY`ybK2*3n&L)8}K({P@0^y@kwgpM|fyIS+>pE8?M> zolblD31gGi8F&s^e}^@`GB=O1eT*x7o?2|ff$#Oi56w7NPR)%&Kg9jc&B>3lUlr&1 z@wozgkDqPgdIm2Cif#H&6x;Ow@45H7t)@Jj84vquL|)Eh>~@UXVQsSGeh&T)TO}1{ z&w=Uxu+#4wxqkOqSC7omNx9a)aB@zbl?@ahVXTE^$BK`%AWvMoRedqmVe2cv(xDae zZg(E{vOC>>IbCD&Zgy-~eYn5Njt!uXsV$yGoIZEfko$kagV_sY)9dli?e}x}g-^@&&DY1v_fX8h(mplrH{bFAHURcl^(bPN6_z=xAy~6^ z|0Cah+?)qc4NX^^ELo60#4LOF@SV7!iJj~*WAtT+1r zCdZOf84LSplW2a@>9N)jvBL`P)n#q6|4VNfX75Uy9s6#=UL&7t?&j0khV7hZ7{u?G zL|^SM1!9kkZkNAe_KYni&v7buO~|iR45o=al|jo_*))SbI2UfoCD>l)6O);4{A70z z)QmySvX`$?G>Lb}U3mK&;5XCXCH%H<=x;~d-)8XJB*p*++rrk)&9;rXZ*PxnV{W7N z;{a>8hdQd~e*5?Pn_9ORm~MuDqwF2}+J*ZAjBDdC; z-SM?nXZ+{6=PW<|IAi0wp zM>~9`_BFuX2Aw4`nK{%`I~l+~sK8b)yC)PE+;ekatbcIl0A4hQtQJ3&z+@U&D>zD;Pulx?@=UHe@Z(-x7B%(25&d zFgCA4=ZX`$bhiy0eBQxr2wW)Es92(l;}A5F!*OK%5FF2AO(DAmFSn0!zS|dh2-6pF z1DzQXWX(-I4v)Yai#1Ppw~$93VLn&Thj=6iUbIiHkXJP4Pr2jJml;Pqb7h`q)aHD* z4P%)$;vMECyIe8gzX6Wo0iEly*@me(3n|T=oT6Lh(1M~tV(r2~5E^WU21D^R`zNn? z%+R2B??tdjwex|lx%dEa&ZN758lnrru*}s_+f`i}{pTF;3`X#(LJLy2~hY)*g@uvO;+_?M-ilPyFz=VmNyC>S${{ z>zt`Ii1$Tj`Wz_tLnDzLKf=E|8N9yQdOVG6a`39VLr$izXNT$QYffLPJ#*~{`KV8x zllM_eJh4yN6t2F10sZXV59k9Pn@=717a#DbKh=(dXn$B0vRraB3J*1LhV*lsAV?EJp?w+f!rJ{Q#;hiOmF@b2_uYxc8`d3iPG z0#z$t#u?N-=#EbVf3MrwplJAUlcSFg2jO||@L9b3!f2?)4J+G(JBktedc46GKujhBQ zPrsn{f#pe!$y%$f^W84lpgdF+`SP$mm#qKV`}e0tbiMiP`7gZr?0pC5gL}znmmrQX zUNkdawzxGHQ){kW#p( z((s?&6TiXlmnG>-yt9M%^*Kx*ihGG?&G}r|lohO9XBV&*Y|ArarEY)LwRe#7> z<$fg}Hmgs*t4}>s?Wy`Tv2OIEA02pj(Lkpa`PTlruWx?ytE=Gy;)9yUMAk7l_=)&n zCo%p#LD|cEKaF#-N5xak^w-FF_ie!TPCi4sb}fco1C-brz%TX1p1)_Hv;0hI`S#4I z6g?-1r3K0(m5mdune1h&OsNneF=X^cZxA_)o#%O0`BJT%iV}Iuya?a;XurAj&6a0bK49T(uz)3KYZ6><~ zAdf9m{uG{X=gx{A6Q{l~_MQ8Xr-PA7a9JrBCRYP1##K&xaF9I6U}?M+n?7ve_umA* z7V-OP*0z>5*R#$Ah9(YM+`*|jrn9)K(X>~bfOxj@4drGm?iwH-WoqxX`eU^n)Cdx@ zjhJ(*E{(0#JosI>^f`8pv!2V6#OXwzsJ=Y6k$dYV? zwSSp++_iq$vsSnME4Mb{$R=mPI1{1G3i?dVgBIrJ=Wq51rG2c!-*s`p{1<)1`TvpH zgOizm4fAKe%FYaU5xu9tON)US@Mgc{k2$BAJR0XyRL0WGy}^N91GKOOT6mT5#KZNh zM?72yOfI6&!9mWt7#tZV$H^VgSHv<)iE;9NW;thZu>MuFSFLE@9r^X=@K+It{o$dZ z_?y7`wf9hdm+lvlJW#E8yZG0Gzo$-e{gQ%RZ(=)Ii2L;ia8o$Av%@GByv@@YB0yqNl$K*a3h^~zuMxr;t^KCY=t+iKUP z=}suOE=_x)z4BM*Hz|(p^{3@a%cnNAisQhe$;*w8y|jkB+%$dzzhTiGYc}V!$d*FZ z*>`aT%^oE2Fg(s3S+biVm9n!Wr*s}_vJ4ujfMzRvm8o&819_L+W5qUMo4Mb8)CSpW zvwi0TG@o7vx1Z?}26=)yJoSAF3z^Sa`{^oO-dFcPd2OrqK<)!budG9BG&i_9c9gf^F^Fa14Y-JuPW4rj5tN&KVc4^T( zSci0={t8F>tGV;e5Id&c0~2>n()YF<;QqDe>ZPwEZO(bIm!jX>hMpIj=meizp@og$ zb_4uqRaZ7~zLN5!6PSw+-)kbVuo})}iYhj5_F2!0`Kj@i-}J~wIM>Qge5=MeKWa^8 zQ0H6hPjo_W8=&8f5ymGj!CA3xzKrn>*K}VJYb4hAga4iF^idO(ppC^j2%67sXbXPm zmW^KmKWwJPSh~O+2N-5kwErA!qwr`3{K@AkS4@v}RI?7%z1#QC2D=g#HMX?ZoNnO$ zu1)M&&+vIAbR0tet(-*siM754Ew^H8ZUf(sf}5qpDYg@*&>e8Q;a~27zk4Zh3VpBU z`!=0TOq?PKog`RmlC{=oZ8sx-se5c?ZS^{9ow4hvc?vLJbW?6^FR#w~FJ^7{FCVbB zeOBFNCcZk{+VG`@<|EteIv|aGrG9hf2kg-gVPa|b?D#9@*UJ85*(VWrP;k5MAbCb& zj{({>a!-Tgcv7+ldU<&R`?>h6+Ecf`tLbC6?%!Y>>}s<{;*UGMea@m!6F;19)0-1E@aWnH^QpcCf77qqLV-P>vZkh^#XTZxgq5a>6bShJ9w&jCkKoV|Uao3lAh1_w2VA4|~##SEawu zyzCO}*Zx{+ScuytSX&ahmF@Y3x85fnYuhcZ-x7j`S6%y!{ZpoY$H=|S&}j%7UIh(H zUWWBuH2hQF?Yog>qLZr2Ir|0q%DdSm2OiwN_rdSJ`RGeOe{zT=hispwW%UD!N9n#$_Y5}Z=dg9m&KXQPG@$#VyUiJ2*tCfSf~h^% zbG^Ua#B=OlOYgd8d^P6pjB98->gZORE?N!`?zC%v#>GmwWB4F<3{UZuyeQn+wcPgI z!S?;|_!OZLP6V$XE~8%p+rTZ?rgHkj>?+WP7sdzTB(Hi&KP+y?W*c!&CX=98={NfCF zH+|DrW;U{HDzd})&lkmJ)0e52L~ag|t;k#T`Ahm#{kCLl;!fl%{IY>Mq$RUzvP*S# z1b4w5;G9}LyXB5K=KP2O-kXb#*v0wqyU~$zkvqE-A3=u9b@e6jpr5=DH~Mlj@dAH_>h6vntNEwe0~RCX6-HZ z+9exRU*z_qc!2v^ezE)6#A5N$?EU-f{U=VTI-(49Oz_Bh?e$4Tt*qo#FERw%V4r3A zVy{yZBEIt#!jQQLhJPm*^4aJo;HUaFYNYU$nJeqC=X#;RQ-Zk{dBzIk_shm#L|svR zY1I5}7!@_O&xYUV_o(7R&fK(rWRqpBb^Bh;_~MJBE6};(S?N8=3FMT~CjndL$QMwJ zSh51xGG1sjvX&TU68=cSPd>i;iSx?OP`tC8em0%T+?a2&f^{IDT>=nCLsUO;yrXq zV;h{iHo=Lz#%k86xed3*5Io$=e#(%ChwEjVxismq;iUh(JS_d^^04%ug)M;m+gWC0 zvc31_vBY2}bA8H%hb#ZE*Y>&hSQY;W9(8b~xQ*)2lAifTJ^)AV8b7Gb$@JF?9jvt2 z^Mrg59$So_`*2!0(f-A9>GFd8(iQMzpZdSkmTyZ!(2J=H#5ROyGSIKazQ#kt)xa=A z`xfxhitGS>hij$>ISV(kF3P-z&Y4K!_c`*)i+BD z{QYadwb$RTF2+_LWqiBJR^8QBihhM3+JK#HliBgtR4FkJctkN1!ynK`indYK5mmgv zkq>pqZFqFXJ!N*REy5bhi;0&a8@&BVH-3+~{)V+#{4@G6YS){_sf+Q}n+`%lZY^jE zJ*RrmBsy=fjC(tg_k+!$xM;NwIifT39*k0RNIaybWd>)W_zy`(%_fdJ2tAS?<4ho5 ztRETI7r1>#A3CE?bc%fFC#TrYSpCR_e)>)<0RD^M{aSE|ZP5#ElQ*&V8#+%gu8C=0 zOdJV+O)^aJOXA=rb}3s;u{ra+;Ox=co8IXR;2UmLtP;5(`I@>O8FrLF2d=&YX;}d8^F-p}h8u&)-`b)**o|XM(`D0(_{93{1CehFd+7F?lF_x+A6ONV? z&{KL0uoxG~u6OVz-07biPa1*zIBvfqay|;JG&|=UD%Jz8&~3ydnPV6{9p*e*Gk;<; zJppP(OdS5aSUtK|=N#_jj7wV{o!1p+PTIdbWIrVLbh~jUok^UV54bP956O=`G zmJyr>LJdg@xI}nf&GUQBev+9s?ZEdba5*tzRZfJK72nwd?bq{}HBcLLRxF5(z*(hF zujDt$J+ltt;#!CIvubdK7w@wro^$V4bME~r%RcwsV%$dj0iAoFn!sL{75Ir0phe`3 zVs3jFUwcA@H{t0R`;f)=&+^R0hkM`j1*V2&Y;J#WVkxwX?s*$}-(Xpz-}}R-dLn+f zaxZlsOHAEI(X=q<3Qb|}#$M_^%J9K9&m=}g-|~4Tfx8UvN=7MWJ&8VLSBei++x;WX zeM!(~W;uERe%eMnXEXEJ!hGaU8ozugyrlf?ZR8Z3HL%wxx4u63ZZC4NZyvE={iScs zUvp0Y%k9`U_zF)4fX7zgQELUJm46C5f?Sq-Jjrr*uVL@jn#w44cpvSHX}_KJ#8amE zE~b42_Aq>;eNG#Rks)6e>U`mL{8(?kelxI^UySXHPd_U*QSmGM|1@!dh13%(cIab$ zqIcz`$%(ATrhZDkoazKz9d6U3vxk7+67-p5e$d)_`F6W6AT|PQ*gyWXTZe;fKdlw} zq5^!V4u`cKsz^#9XEjQc3#;;&CLHZsp#TE%Xg z=FZ)1r@6a*XztjtGb5}+w!|**kYsN17qF?GE+X$bg}kfc2eJ)Rga0UNdhG$~r0_ z8>HSPv%RJ=xXG$CIgf-drhPTiwKvn(Wc2z6&3gfDb`h^W6*}mppODX*b)xp?($-ze ztJHP0nmok}lh@RGTz^lzC%13!_1rh{mc{Dysw6Nrkr_dJuV(y+H`R!xLinI zn~Xo`i+%qC#ueN(?zN1o&(N=_z4HOHBy^$tn>+t)Za?TRpKbT#=ReKv=W@-T&mU#} zn&%vN%cV`~N~Uebw+teKl!vOtkM@zHOw9++*OPN7H+F(c12e*mrMj1fNsI@+n6i5DB=ALwFMSRHiA+;jRqnf&+>dWo^g_F$?yzwY zeSAOB`VP|Hu#wMLibw7!?=->h%Cvp(RKTOT!rh3k9XUEh_?`mo!K zy^fzbM3#EvodNG2E!PIHpLyRIV)sNYjQ#Dqw%s$tK11JI%AFS@r%NLyt`sb&XYo6- zU;eD@5Z4}hf%9zNA2sxR2p?_H9o;p=yDC)o?C6tbVz1I6A#_NQY^D*!Y>-RZ^U+s^ z>_9HI^1IvK+ei0<Vl5J7OQ65e0 zc6SDb3h*T*iWJ~mll^Kr!F zheNS{QN0lSOl;oJfa)fRwO%G(HT(~5b9@VLC$YC$B%9H%22XozyKV@+Kj4|>9$_uj ztnHV*CZ^h9VygMMcU#}O$eaV=o%@x;%zZU;Pce5VS77*GxdM0Y6F%bHE$qrN;Cw&! z!{TLimlKoOx0jsDUOsOR`d;qm-jBzqfq8+?7n~b;#Z-I>B3y0;v^3A`!4Fa%!60okda8DrD!&5oTaTUGyz|F_5X|1LOR zG@3uphCD!JvQc^r+G`6`rkBI_ON~4>c8_c}XfJ2`oIy>t4|t1qW&4Q69)$kd_`Z_o zs(+}L9mE_JFH1s0vX6evS=QCWeUjKesWkcqxhdJ)jPBnIji%8xjp+U$x_=V79~)?S zl78Lwsy@bEv+aX5CLiA4IQF}5Zk@OL&Ggsaelyd^`}fS*z#VXF8;@PJuJKs=i_Odp z-d?!Tu@^RCFMLp+>9%NP1u-aW!l%m01Fz~o_D6^EKH7tF;qTth<)$KL&;B{arg#aM zBv?~|xT{4ivG7wk-xIB4xVFQMU3!ACzr)xkJEq!uVM%V}W$wLyaW7}l@8vA|M*O6e z_+d-P_rKO+-EMK7{A-gea~Az;&RO)YajtxXdIZIX<-0GYJ!hJ}%>81Q5AseO?+o%z zx&At5(GT*@Cf-pFDiej?;8PPjghp0Htk_iE*LffZ<{=ZSEwg{QnDZzM&0qr>`#u^m z`J6%I+|Qw#pF(#pKzFjKs(hhCo7rF8kN;!*2ICW!m^{Z@pCc#qUh%9i&~7;RKKaJ* zxV;hh4%dGB3EGcDu8S`M%n6%47GQqbuiAh-NXV{YPV$@PnEkZp#ST(Sq8RTMY#{Vz zxArYRirx%Nu=VC6ye~b!33+MaZN!yjH%$Uoi^)UvU97yam2J0kQQf(-fzS`)za!LG z$i`L-dC9V-fhqL0c-f=@^oo(Yvd8?2lR3W4xzxLQeVdWU5@K^NzaALz;c11*k&7WY_{lA*rSk)ah4en(D?g676VOiZSC zPo(m@1@XQY%zeiba=w2VH2I+L3|&j^iMG>we3dy^V=n;feQq7Zv(z98)~fH89ni}> zL~GpdOfEcR>Yfv<$M9;@){Xt>s0_T)kAClk1`^}&@6p||8{{uM2v20t%bpl$w_7){ zyaz3Y)*_$cdXVJ&6O{8d}j#~ONo&2HdgV0QHs zQ#*1QYsk>oM%tFMmgG;kM;Kes+vgB_uFb~lTJ~~#ZMrn~u#3)I{iymW`>dq={f)}s zC!_vK#y9p`8rwdD&63AwuxsRz0qO{?naOBPWfGkx+b{(_#s6sspMn)|>UMLU-x>~1 z$h<3n(=GkV$t&&}jLRld-EKMXKTdxH99l-#4ACLZ^CU3T84a@aWFHL<7R9l_I@%*= z#4^w(HD0k^=u>sFX8(GdFSfr=_j}>nfR8jVMt;4VfIg~#^#j-;NpKY=-dIO`M{-TF zU-_Rj^m3y44L3L0qiSz(=B@*|l#JTADtB=eja2>vICAZ4`#zPS_^A<}PGz0Nz?-!Q z))C;V+Oz~Tg=eM*6nj5xY+P_F zI&t3}<^2u(ntgPlIk#>A`Qhfj7A`_IGFHU)uXE>WEaB|kFnZ#+yn>4(XPzl|K=lMU zJ^-KB!v|%@IEx+~=G0-%kni(nPM`3$lQ(qeR`^B+OXjwZ&hzFe78&3zKFdfSG2h^?B9(KH59BPm>lV#bKDN8Qc5YVd5S=EWVXf;Z@;4H2 z@~!ZocuxFccx8{R1Fk0LW$G_^Pqgp8H;4CHl*2s1d!IZL*>Zy4hRDn@HeNpM!HcU4 zz5J;2mc98d)whnuc85R7uRsc#`~l4`dK6VlieX%8Q2G| zvHya=!N?8bOxuWMNnVK7G^XUH>NQ%C4ehj*F8evZ8Jgg}JpI*udFRnietZfVnBe)W z`0PW_z$WByI&jwx>F~B?*4^^4_1VSoM#rXf_b;}w&ReLNsR&rvjl?K5=f%$aRF8)q z!|uTLfS%Qt{<`Bc_nm3_Mq!SHmyMmH?@f}81#4{btg+r%<0RspW_`fbtdG2%c%un; z89k!Dz|9=^_iN-G&c?oU;hy6M&wRZ4zOdh%Zh;3CpBav}S3^%j>;Y)^%MJ~>^m767 zx-h?`vjPMg)i+mT-z#>KM92kimse+>zloR%G{En(8jlU`M9yFf zI5uIxSzk_l8=Vx*HeS=NyZFzr~yrpRi^r_bPirdaa5xZun@gxUP+RfWXliPS>o!8u z@+aLq!)wHMTY+~-IRWCk;yd|-ikChDO})K!XdUK@z)5*I(SC~GT%A^RV4yPwtfkKy z@X3&2GyaJ)_7%%)pfAOO>xjL6o;GeCO=DN1vnEsDyp7na#@@{yb?rADZtN#cFm^d( z-%v1ivpY6@nX!qBVwWig-tFebB~Mbb!3lUV&kJ(nT{jaeK({5qiS+-F991-Ect!8k zH9Gof>0Br8ZT}W=&YJoWwHeh#cD*!X9uEJ&x%Uzn)*%b>XHD7f*|}Ng?$(bmj%u4C z4*k74`CaF{9migD-goUyTMpRc_^>e&;IIk!Wp0A*-XAozxoya!%pyBJB|m8u^dtX* z{8=^$Jlm0B($y~R&-nxI!*fCW>I|}1`%7HArb%^k=xF7^&ju&@9EK)j_X{tt@!Xw9 zjWdr&jt(;q^o*GY`rFI{nlkgiCf_9AiJ0io-!l)%V$B0N8@?`x67(NGLp|p#_;`{ ze3YT%&a(NiW3gRl*HKTMU4wFI7WYllHWT=weNOz7S?P$5#m2S>9jZo0NPZ^ZSLMm| zTcLgN<3;auZepzoc-CF3(arjPF?bsZMSv|w&*bL zyEMCTq)oFoLbI+u{xEw}boz1Z&8wk7dv1AqvzGBs*514Ye0uA1lpm4ZKMB4OZtHEk zzoP-WKWC#jH8~47U(w9L`7SufKXPMK(&3+TV3I5a@1xLrV}QR0J}(%1H}5aWhiS!b zlN~C$R{hNJv0~e1D2xHOTSaTiJ+L`dgI}Xs^Kp?K<6l6YQTL^DTa@D|mk%5qW6q_z@ipTEQ-5|| zOumnJCxMKX|NQ6P-^5qYt)+KeAGjj#1FySjxcHzA3zyGqeAv%~zq-=u>{(aach_!4 z9w758YoD|4t8#z;hk4n~6T#SBPegNiiT$eV-wAu?T(F+Ay}y0PdTKvlJws>-I{GR1 zPC5>2HOuX@=kSII`7fI+hfxX!GrsJ7`|9vupaS+cR!)K zo@>n6t3&xCiLR+rJj;>$Mz-@zyrUebWZwqax4)f>UvK+r+sIjND=&?0<6P2q{_nB^ zwoMG}%f3+ihS$X7+@S<*Uq|jCVPXovDT=KDtt$R%`c2P8zwkbF9_GFN{Cl6)d)m{6 zjE^u!i#bNvx2F0P#lP3H(av)AH$`>FM9*2VCTj6Q*xbkjr^cxu z_PEF9mF-s1{r%)4`pd1@nq^A|R$Iku%eeakJ~n5Xl;G=2Hp=FfZFv6e)J?#{k|7D! zm||=%?^(`%?!Wtzor`hlSu*hq5AIia>>I^%N0|6g{(b?$O|)yVCj6VH)2BbQKYGFU z$Tj+>?S@Wmoo&}9)*;K!WIk=*UA@2JU!NG-ru>|kYG2{wg(KBw{Jr}d_iQt5!aOtm zv|MJ|xO&-3+nR&s!o4xvF=y9DHexR{w#=kngStNYP~KT_+Ztf)2j@}YUUi)j&W8pM zHOybJ!qWsZaOUm5hkm`^Ri{9FWyrIuhOLbZ7#!6&wULT>$e-&0?j{ZZzH}ePv&`iX zf4z8kvNSJKHl0A8ya0_4w?Em`2nr_9gwB_7{SwIpowvJ*n3nAD#q#N>O^{zz55DAE zie^N!ib?zpIMZ*$s23`ZV`AXISu{2uAGQY`QXb-2{t6~ZcqbtE@l2m{F%bCEz^~1w zXLp^hE>^xswD-53HM{y)^}Ejvv;HG{ocimqVBpEW*IHR$M?}8dktDT@VQQCsJ)z1) zvKK~G8(7?PRsOzoba~~*9{av@wW*`cvAK@@vd6W-g7Hi({!Y0swhda59^Z>?w0F{U zW9I^gg{!F3-pg;jz`PF_366@V_5$aGsSm!Kww8f&FR;9@~Ibi9{`^j=y5Uque7e7wMlg&$Y8}w zyz=Y5QAf@B`ab+AJuioL%9?;J{)c?JUhs1O{AxYYFRjF6Rj>L%eSXF5bD5`4blZvg z#I9!_vTT~XOno?$kU#Qqu8`!?b0w{MyqQ*0)~ z{4>|{cLD#m+O>_+Uv+oXATw0|!2JhlU+kLm-gWF+n>NX_91gc|?g?wF?lyjPM?LWs z)-dBL>Q@wBsou#Ni>+*AC!fDfJLuOxoqg;+>T#MQI~0Rj3k~~-Df^~dPYaj&>s|AS zKG*$maHqzOIAhl__9n)j%GmOsBaB_n*wu{f?PHZkUQ**Qumt!@>8^!_HDH^Kckzfex4 zJdlgweeM)n#=cQdkLKDpFF;q`c(-EQ%8Sh@jb!Iym+3yLdC>p#I%nT#bQt-DTq}3R zas7+~!^ks}FE!(xmx~=F;8QN9TknH@6(6m^cRp%#yuI(sz_ZQvu_RL^Lnk|9NuOBG zKX*L`iM<#4Q?mILmk9KjxWv9D;u1mpE7>-^(BuYkK8sCUBDlQIiA%V@tuS$kpovQ? zcH$Ch6Qr+4WuHmLjACA6;0MN$jUqliFax+FGi0ME7h`GyfT#AVAzPpFao&x1Otlxt zk{Ro%A(#j)h03YBV4ePQ%j6|#v(M|FE+!wc&65vV%>49Sb5lNKC4Jd)g!{NJb7g!U zemVaTIKHC&=HS}J+4oKe&%8Y}3H+>tHbrCH>3d{1db1teVOKm|^x!x4uR86!rUo0i zv#_1BEH=L6yS)y+Q*D!cN!c*QmLYdA;XjZo(7SSjiiyj{R6Fse{x(ooCACNR=x{X%Jsb#QTm;do-ReYo0ikUq)hk(;N;QL)WrZqT^zs;jeymKphuhBX+mVS{o zzod>M5{fk*du>gcxUZkj$!Jkbw&NK1G4T$x_AYhJ_o#N!+BvBOnE*W6uuU3!Lb2zd z?RNGea2}1pvEgO2C+m{fv;3xW4C}B@0_+EFE^zqRlk@D$mh0)7n{woUty& zMgivHPwkiR!ISu5PZz_dtKn0f7gEo=%6rQv5kFTC(*`}ea234#8oVsoC;SPII$KM+ zyl=J-U!BwulvSt zJT#f6ZNHT*$9Ay3H7-_<-s7Is98G>|!O>2^Kd}OQG$VKCAzvMxp?!h!KeV6P)*Zzr zMrVT)>h&z)#~S8|3uF&XdtMtq$4UC$brqwWV%+@o)H zb!8o}i*T;iQtY4pO}bCMIG6KN&0{^VbZH{;S!{=4`iK-+vn=ITI9p}W-x3>$+ImO! zQxaVv+d+78+f2bWPedcxKl1!-o^SB%N4p&zC0>A+*c%g!&7;3E@+W2G*4@Ny3|&uT z&K1lH8Zr7!d9Vi6MAN5gDN@u2RG{bl(4}%8<)z?{-{$bnq*3q={ovOgZbYU>(Seak zlCS=*H1jsR18%wNx3kW&p3*$a>C^A}e|(QXSF;s+Suq#?C~!!6NPDJIcQU`(R@VI1 ztayp&#Eu~yxsjS0>GDP3dp7#$&PX-^eI=o<6!evVzPz-gard!TEX?^pN$|A$6WH#J zmeCvV6MLM?x`ZFqJJqn3nz||Qf1vBhj-#Ex>7Uc^P@~QwYLxD=>*D;t?OEi34>;5~ z>(N@&M)c>WFCQ?I90{L@7N;N2D@#6e^1R|lhSVT`YMh#8A9HJEZWeP3JG@(N&Lb#` z?foir2mSlNTUdF$C+s`2#Glyh-C^#K5gZf1F}Xr;1-{Kg;puNdMj#KH(HHH=!FJ?e zBj;hevap)Zg5T_aM1EG$Pa8A=Z8qJC-%M;#XGmG*UZGFLwlE(b@;Ka6YV6}=mEe}A z#Vw-6Hh66#yhb02b5%lk zVJkS^$i8j=$0HkGq0e9QA2}i0>7Ts6{g&si@!#TIH-6n#gxqX4x;*_CCr=B)@)8`H`^^jA{I1WZQ1hQ!O-s?9K3AFZ3bae@wMUJU2AO zZ-%D$T{NXOlaY0*?e-%l1JGDs6}8vS9>CC8V6L&T6&lzBk=V-j?ac3SXy|*;=9BcX zHB^}zwGbQk#`uA$pN*#;8XeDcejU27_TGxbQq2LLgbo$Ymh4F~Z}FSt4>UbCN&E9JpglR7)E&%a7P1bVK8g0k7rU*< zgjfjsdkS%g?ez5u??}G-;7v2$(n$7_Q>^t7YF-cUc|^QAUi!~9@sUqQ;|HgWjUT%7 zjQH+tB1R4*?_kwnR@Wfa^Tg);v`HiN@2K zV>v#MY$f0;IE$xR9l4-Ph3dZ# zdEMX48kD>J2=f=6%g$?qh9dM?@60(-1ua4I;csV$*lTXTg1KOPBDj2fZb=6xuhhE{gv4SHu1~z-p+*bfjbvCo{n_N5y_O#h2J7gnvgxbi)6hGLu zrd^8{niHwSpPwZh3U@Al2yfz{a;@*bfV-bs#rKzR)_VzOy~md=8+gnrSzF}`ni!IN zs1eGGVC!adekSWOa)a^kp~eM6nuF||VAr{Pf2NfDIrJxZ7t)~oBiEPxc_<$j_#tWXB%lBX`oT&Pq7GzjJf(Aosk}=otE? z?`rz28AXjKwTq@dC)ch1+MNDOeJ4Jv?&1Eh^GXO$s7 zfpYKw9Gl?3cKjap^%=iMKCX1DY-Ic%&Qh6f>}uZv>LM5?Dt<*{?fWT z2|xI3pUU*xiXKwFJVQT%$ph4R=(~I?(Si1V$+yz)D_OJb4ac`KGOZ2Yss-Q5jum?L z9zxqa0qS3Ywd)@lygJX(jr;2FFgkKy!=0X*6z6w4jz2U`wgTAHUo4mgy8Nt5=Kvj8 zme0$IPP&7(%+Xx|qLCbZoI8}>O^yy8Oix2cLMzf|b6D?O@><(^Cd3}a4w;7?qCI|l zfJ+8{X$yXu=Dm2qTU+aYn!j=`fvzuo2ArIscrNzIx!4hBSTh~m6W!%~%@E3jL zBiEBF@ux;#?J9Q0lx zk1*z}m5lW~XALS&v>4h~o?dh4<-D0bi#j*Xno;aX_aW?Z?n8K;t~9AXm8hCesFczsl&+sBLU|=`!k0>Z!^6P`PGRXian~c#gq#OSw7X4vgh^I$|7<&ibJZ_ z^S9)1wwoMI(eeXM4#)lN8I!{)GO^C>P7X(H607b0lWUPDjHUB9)&JG#g5E00#!%Pa zev-AU2JdTu1HM#x9qaGJC&9MG2k6+#nU{T0{IPlb4|vY`WWKT+Lghp9>D}IZdJ%fb zt$tTNy%*bE_ubmQkLAN_&vb}zD^4Y{B* zPFC`{7kS&R@0U{d?7Xjd*?+(vq7(X2%#ZvUcd*D$!e)9My#EI^5BeNNH@R^?ljHRC zkpjM|)Bl*`XrGYu{9{V%;(r2%plU_@MqdH){n?A$9O||M1sI?Z|=se13!1n2F zT^7viIASH?#CXnXqK=R`$X^bTU)^@~r(#QpD|~+CKoQ;u&mPIepm7fzn`7V6w zne3Jn^gprGwpUcYuKQ9p!24}ICV$&K+@8H9C)l%R=Wp4oiQHY{@V@r!$wt#&+m-a$ zi~UCZH!?WX^|gz@6*iORVC*aA)H^1UU4d0Cp@MVFIUhw2dY zU3c$k9d+b*)sOOD(k+H(tg~X3$$?JEGWDZVj$R88FqeCI_bNf+;#tzV-8rV_+E_3QO_+Z+v;)f6@a(m)}RQhuFJcOmq?W?}Q03 z57J)-8t(%p`TTpZYu9w<2E8hdth+9tPX|8IDS@t9YEI4ucV)yjj)N00Jq}KKRsg5x zfYVXvz^1>T;jg3M&OcgA*dtlk^fu)xo?Rxrrd_r@_ z4(_Pgba||Xob0pkS`XhXzE>+hk*J7$2mG12bVi*#PhLje&zJ`p%f;70v%fEY|7c3`OkjE*>v2cBLC2(byF)09rwX={owY1>^Bn!vh!W4 zkIWzodZ{(Db;9v%O>SQ7VdidHrtOZ@jh z-`tOgKhs^;W?}-}^{erz=Rv3Qomx0*A51*qmppp~IRO5?KHgg2aw@Xu739EVYrW`0 z@}(Xf5XPprV`}V;{$(V)0=qtFbbDZVFM3N!Q;KzVNYxmc4^T(^*CH}dSW z{62!ur|`Knp5DdpOX*W*iZGw4tY!5K>#0w1Pp|MLydB+x96_E~=-%XfXmp;FgK+R; zaJbgN;X}jYFwJ_K$|IFc)D)=B#oO+G4>E`B)bwNjy0QthL*^+NMkM?mFk)&XaqWc#z2-ypBvLx~qO*m-YivyOfMt*}a^V zw72D=Sl>qGx|LX*&b)cdzE3=rhW?O&vJ1a<#=EwE8HJXR>1lLjUOMRzrtGn^9Z@Ne&`7-$8?qB$22 zNpP5E{DAx@{95HB)lbegU1#HTIsK*hTxgr7m~W)Tii!4qb3SM8@;!Q&$)D_;10OXZ zn^c432Un`QQ$FQE>_+8U#rsWmo2T@f+B88AE$EJw{Jxv@hFNbJFbH#Bz4pKAo!96C z*eqN{+uvd{>Dg+=(7cyHi;rPSsU-cVo z`fAt+eLV#IO+xnrqa>d{!PXs;@ zQ+qKrwv>C)TIIh|bKtZ6tQ;RoE(K|Ov>2S}J>cLMei^Hgu}mxw`uhz13JwjxtO46d zV~Bqb^Pc<}GbVGG2u#YC8DC_q<~Q&Fej9)5Zic7pECb6r`aloONb;`6t82n;;k||? z$-m>}&~W^Cug~Zt%U3CM3%Pp+wKGt-Q@7r0|1*~iEuJVC?+AD6>PhSaBv+w4D+~Qqn zFu%P?Y(hSR<|=*-K+B1>;0;?zXPfmHS{W1T=FIuTTI9>RYtS3lqBrKp1EVeUjSqd} zkEelCU_5JCQC-=yuXyHg_!>@cB(c@bf_F{?gKq+Z;qWLNHo?CU{ukm;`-O7&19yj* zoA7rC{2lxd{Ea{sp9p`mPJln<*JM*X%wMgimv)jTw*{SiX@Sm2K+|rn^xpSrQw%?q za-Yb1`5c$)FMNb}PRGZI=bV4x@$sB0=5jLeoLv0%lH=pACgxg;y?2wZGJw6e^%OhD z{R}nhia|hUIXkV;MpHc9wb4#wpGk)+?z$2??c58XpVhz+`ZTthbxtgO2ef!6@~VwG zsulc2_obN+xv9){LF-1bQ)_hqY_WUtWuBqQ55EWCFn#zZF-pYgkFtTeVrQ1Ti15enkx zz}x*7BSWAe*(6_L{es&d^Br;*&P~;o#`Xx<@j|u9#r>UUX5R3>Y^%3kF}4+Uv13~? zU&-eCiX7WYYox8Qt%#91a-&0c6{l|F+#U9l46!N4lwg;Rz%D(-wkeJ5^2c6#q4gEcb2qScWEk-+M}{q@f61_bkzvFo#>AS5*(8=jC#6?oVz<9h1#Q?m5&BpFUat35ZX0XuRII&)b7>w!&);{~vVneU z;9p)J`1~uShacKM?f?Gp&BuK`Z(i#wT`Rlgb?D>(cGy95N*B6l@GI7rv55_Rj^+3A zza|D=k1k!zn9#|(FWa%t4D=%%aK{(W@e8UeH6P8Z7hBMs=K<`pgUG^^==S>R%G7*c zWet1Aumyn4Ww9pi$HoT0Ryy0Y1&}LF?Je)g7I=qTylQn^Tj2UX9q3e_iNJ`i@n=}; z>l7cFUtNg~pP8&;EXGMNpQj`jX*cO+^2l}Y2zp(*#O1rQuwjV(&Q#xX;0xJ%ifLTQ zn2ANcN`rH3Bk`VS%jPZlRKc!isb$^Gd3K6*W}pq(d`sZ7*T~b#zFPu+x%S;s+Mwq; zT>I`Ptf$bvTcS4T!=-x9X)F8gr*2!G%f4GiJK1;3^nDb68Pm1znz8SivG2Ar=SJ|p z75h$WYqjk=E{SA`GlYQEVtSx?_h2xqc6?=wSx1#1}7kInmJo=g=O-S?)=C#&$Rg? zvSSf3jDU$@_?KshO_(}(_+L8B){lM@ztEX>I#W<>o)a%lMFvi@vNujeKTbvNO*6FC zQPVOpmKuW$9tT|lYsD5td(^5|ep=`JF&EA0=mg?30mTogV_6Y#bdMb;yIlA$c;?0b zbsqc=iT{1b`F-yCw;)3lJO8NX_r1*;3u&kHB=K{k&E7A4@C?yt{(PbYdMb&|s#pfk zQMVx()Ib~q7_#R!&~^9Q;sxNlh~MFH#nD_Dr8xLMkW=+>UZ0Usv^yCY^-t8bd{h~= z(QW&2Wfb#Q%*rdLl@MN_=mR{zfc`Ex^7U8Fe{aqGCBD#FYC;ZM z#1&-kr}5vuLf*TV^=~ei^TYdjUS`hAez^$$xvv5lZO2DCmY`qz;mbjE)e`y6@TYP{ zVeqKWVdS?T*vNNQ8~NzT?HeC{9+_8ea^L4E2hpMX;S=QVjon3C*SA>CT@wl9s(g#9 z9XTmksed_nY4S(;dPWlq0++%;FE-lQ=*(Jcj@URn$ zZy~2+wUo#H{*7ZpV)XVLWN#JbBMRmMj+IOF&gDYpk}S3N6fhUgkvq?x3pUh{xd4Cf zTwK}a&CM$x*#|8tCO+B&>(r=-yV%3D=!DrOrfutl3Xe`uUl~^~V6%v4elgk64a^x{>Zq}18{MF~#3Z`GK93`^ zgZ&MaR$Wac@zebY<~GQ8=?6C^O#EYre(1KKy@F?gp*G$NwSyaO;%g z^+Ox;I9dJug27ile=~^u5G-^yk$jN<5q!Zlo`U)bC?X5`Racwtl~cII#CV$`txNjehT23XG!sZw8-^JyvY<-@gBrvk0%p z2B*JkIe#aKZ0(o)m}QTxIrPGF8Ste1<$+1y`Fik!ADu$47`+Fsbe2fpEaC<59`!FB zC9RxYIGetz_>bJ|Xa4=*pZyn=#qjG=?q{na&e2c4sUP00m3n>6zM; zMXapk%u;mHYT}i}(9H^JIAnX2&9VC|uH)Vo=qv?|rIGVG3rhEC^dV#YH!#;UwG+tX zA78^>0_Gm!*?HITS+=EXQ$~leDWkT|7R-&^Nt+yvx;CcM2EEs<`|ge2;|#@#)WS%< zuOGypDroy*GJ9(_uOEk#*>`wo{3PVl$D;2!!_fC4=65pk$+cf^G3N#3?N`YqzeC5v z#op$4@}u6^+agbHbdD!C>e|C8@E)MwG&ImFy@FkEh3FP~l`Wh?r=~9^b^skV5o^$W z0PcA&hF-S%488n^Xyl=*uqo$ZQ(hfUjndf*fp}k8FkXirkb>@I7s($A@z)1kNq4QD zFWSNnV!URDeicV*hkl!&-xtZvD}K}r{i>O-a$qmKb zwgR6o19QoYDDSEEY#ugiEwRN~Vjy#gt$EunrR@~8r7yL;YBV`8))8TRWY(c;h{dZP zy_3T$d6XMnypl)JT#2#7qKqM(l-w9$Jwt2WeX2z>u;M%l-W3hF`fSU9Edx$upV{?Q z`FbG1T6|0%vEz%EY=S^HSeMD zY#8w6;5l(XZEJw1Y9{4pcy+E|_+8rS?-Ts>!tn+NjyoJWcID+ka%|pu7un+}={$5+ zZxsKcDU$shKJ|y$^s4nddl_{1Vg85gpo*N$zt6JDVizsJ|KP43AMiCc|IL@!Hb1ry zHh*`{|A6+rwooDOX3)=Gn}3Y6X8Bjc>5oGD_DIw|ALN#cRUcDuK1dq7W6-Khy#s9C zMP3auR=}6-EB0mkqa~Gt==Z_FQ{ofBUj=kGhgt{Z+)n+iX^BvGPTt$u`@RMFX7)Dy z0a(06fAff4;JYp6UBwnN=x)_R^u8mxXFb(hT%GL;RGZkLYFPi`JZpV=9&O>HKK!z; z@t)#9su@>4!^DIBfV_K)bHfLzu#`||L=R1MH3g+01eGv#&rbhi4I^oaJ3D>h% zc^tZ8Ji6knc&d##{F?F1IX&E2f;_xn9=5Zt9@Y`~~J|Xjd?mY{;Sm zTVxN2Mu6!-Vv7131_nv+;;m&$K|?xs-N92||E@QHYm^(7yIMk+W1XB}g& z=6^*H9Q%mp4v#xrWJi)=P&L3hN_=%_leakCAKpk_+dAsftT!(#JikT%Y1&w zT;>L_?;SWQZ`jWmcD-e>Ig>KEhguKhrFd^5vbr3YaW1Ec(c19mtaR+Da(oP}S?yCl zVf-rKcoTD7WXqYOY2INh3wS57pS0InIXqkb6`L5E!3|?n(|-^38Y{Htje6TMblz^_ zbjj;k_r+G$JZFFE_vUaf3HF-kvF??*hA!)WJy*GhvBqY)KEj@G-eDe!r%K+P>^%Ny zxOps~&r9?3c<_VfaVaoa@bTv1`W|7ww&ZnMuf@YR_(RkwV5e?e}4(h2+vJ5FWS zUQY7(7XkW$mu96!U(o;LFg6F{q@kZ6K5x5qS}X(oq-*)Vhx_-KpOHb#Ezlx+)jIM! z_67Sc;hj6+Z{)N5>rEzRdUmXSsa=Peq0g1HSB>D`kSmlQvyyk7Cm%KGrGZZ5{k~l8 zoVLoX{g4_7)itI0-b7oIJ4Z&FJsNy>$B@rH;pfDwFShMd<=?c1B=h-4txXks3=1#xo0N;Y~EaqJ9h6s&Gz zp89-T4j(ww7tC`XYn7j>esVc{_}b`c=C{ejlE=ipLA_^U0dOq6iX8qt)^hdG96oC> zIegVl0AE+edEqM@kXz2-z@=qxo0fOqvExmua|9f_Qfr%0H*#0=4#In)n`CVs_T$f`Hkf;}P}D!1_{cGmCTv-xEU?<$Y63A;sgG_oTr zuCq8(!=@eKyiqx9>`Bp1g7@@U_Eiu+Rx}>CYvMqQ=%D4a*yYss+4Tx%7+c=uRdlbp zqeOP8`uPR@D7PM_A3f6?%3miY z;J+es6|030)~6;odz2j6=nUF{iCURQPL z5&FVrXP-IuzLX&s%b<5;V%XxGtO|H?D*v&M60%8TTQIl8I9o=0U^K%AKH7kd;8F%o zw1%V83i!o=%a4ybyx@En?d>Gz$$6nB_wJ2X{7=#UJ>|#K|2pPyGW74Brx_n9{gyvZ zQ+Gcl&?U)6Tfa@j7EbXj1^iOT5yca%ql2CJ3Of%XQ!czH5}$rqb^McmS`&Z$dqwd# zzF!>w{jQR@YBMs>nD+MeAv^n#r3bj9Gq8@W%KJ}+`*BXV|#{6yD^PyRc` zpq|N=zl?Ep06Wy@=s;(Ta+l_u?_T+Pney%MoBZzkJ@VJZ?=9yXk6+QZ@S9o++?S&( zZ^Jjc1K;vaV%vKHIlJKjY^)bckN1?G&}~ z>i>0|x4t1n{ce%1|EqWpzRy5UoM-f+=$AbF;}PJnk{FftuqtQoQ(j*Edu<&0KOMcT zGb4z>y{tLt4EWZ9Hvi{m$a4CT4P@lZ3Fyk5$71H8J6-yiOCK_oTCT^Nh$BV#uUMpN zwU@vL@&Q$=t#fT-)|r|G?X%VdYt?>$XUTmhd3Pc) zMaf3#pf+??1>divc0{_TiWp;(_Npf^v8xK<&EM4_ANMHrQmi*wLG2f3N-EFahs<}^ z;e{uBY{SoM_W-|VUj#P1qdj-5?eTKfGaMe*07I{Rq;pv`-bP^4sG9A0tW7bEioC59 zrGC@fcLF|7ubTt@=Ys#MhQv0~Rp5tsV*BC2g?0S@AUu5s*pxea*^MvMb78FKH?}YI zCs{{V6!7hEd2`qQspkH2m!F4=qsjj%@#vcWANnQ%Ul!V1@4R`E`X&&t{i(3(qeDaX z!zH#Of1ZIxzYlFb32lB4{5@`&nv}h_vR_;|1^oa2%)V#NSFjX2S}v-G<^;@CaJkf`mVhJ^3s)s++R)5`8Cj_ZBJ#qNdn z{m|#vd3Pf3mElt;9+E;+IWT?$Xu|&E{pw|xg@dWu|LcloIdRHF<030_#%BN zMyLMLJX4%M@ep&u1`C`onvLx6arRWXQ!DZ;eYUkRSQou&`sah0*H zg01$2au&0@CfEPdd5Fm8u`@L`pEb7DAf8Ly8Oe5hp12O5lg*aS+-3Yfi9O`@7qRau zbn%oFo}K}&tRek9c)uB5rcdn~{3QMBH{s2!k@cyK)~9wl<5q1IQ*`avzQ=(z>sGu! zNM0ZWA20ipfzH9^)8h%+CfkCsV{^gPI^qGuy5z?TZ((9PsZRc{1t;LB+@j9!4%(=u zs3m0fOtm7nRPRnqnc9%)v6&Ihx|RJ&yAbm={R7Jk{p*?LzM1&c70}Dq_$%K^wVE3| zXHHfi*Bj z@fY9e{3D+W$E$b7OE9+tbK+iy>=yn;(Vc(fy|;PCJHH#<_N+^Eauj)A$2>K*XdK_| zvJ~&6w}U(8CH?m!)gtiB&^gcD*noalJa{oMWj#NcM=tRkNRTW)vn&KeKE_#^UsT+U*N|pW9L&tseXGegQdRgyNx4a&COOk(RO-l9QRfbd&wqO``|$z z=hyhMnRUKwZzp-5J*QT_8#pyqUmA(O8$C7FIEKG%r^Xgf;BQ*bx%a$cB+vQVIFjd9 zJh`DdoBY1dk$=R@k*$63YoeJk_-xC?;#gA8R_Gb~UyU9=C-(1pc4w{~y5>(~zt*!m zhPKkC&!cZ@I`uFjggaYkmJ^8*-)xe7w;%O^jrX2Ww|eF!7u6N zYCC^YNB#sIB_BVtqT0@%{6F^I1w5+i-2dN`$vr_pRH~>+2#AOl6%|u#Gf7nBqIiy+ zpKWOiAt++B+S+rh+LmBIh3GL;?Jwu3=Kvz=M6J|I6>SSh)he|`?d{a|Ft^+YsCX$1 zRP+0M*WP<(&oBu=IsNzfKZoZDvuF0+YrX5euXnxcUD!CYesX<{y?(-cRqH36*VQCW zfG-PXvrid5{vG{vasIK^oD08*A4VHd@KbKqpN3-eT@NwpE+`jzWuZLq(A=`;c4g@n^Bus!1}Ft zGd!y{u^PUuXgW3M@>vry69dPYwf4i8)S7*n1?Zqc;{N!?M|97rO>|w#SvGyx*(33} zM_IGGsasY=z5V6X+b{DKHC7zM{xoWw^-)WI$_=el;(|cpwxg}5nsnVz z%YKhqqocn%(z@F`$2_*jtSuC_o@3`d%!zhd+@sZyrZZ0x@(}kudzRG$1Tmi zY33zocJh8cj;5XW_SX74V3X&g8LJo{c72INA8sipYOG+iFaLKoJN)%Q0= z7p`5pjDapop-Ts_h%RyH(g9r_{jh)D=@dVqiv+*wvvk0-Cf5KR^+RIO0mcz!9529M zVa5<)EG~cj_z;aFkTH%3u#__vi?JA4U>p&~apM5vkgN&L`Q$;I2W^!*wc#`NnJ6zq z(<|P-dh9*+D>tr2AHGce$t!+%%NzQRUpm{=s$|{jMV?*AvkQ6F)T-2PJiCx* zO|44uS52+TRn)5FI;lgRhraMyc&aZ^gH0;O_9z!f@yh_$A8bA>%@02beIY&Asal;e zUxGEab!(Buj%|#!7rF1U{ej3^d@`;I+NvQ28rg^ZF}7$g^FqcMyNSK}W*^}h!GDAv zX}YP`uG8i#3?$y!ZOcQ1-zt!Y9NJKwwlZWtdK0z+z2;*s8QI70r98ts%0p6nj^BO8 zADuQ-=R>jl@C={PbpdB>?{x74^FHZFeMWWz`##3!vW0(YKg<`n#P zX3q?puYW>6x*m`G7SUF?FNGiBs|(qajSYiS*H*=z$B&*`lUTc@uWdduDt>qt8HrV* zpB@4p?rY8!<{8mOd=+7?5D{;YzajpLe$c<1bf70X(G#kVT2CxYJXMHqa2vjXcxV%A znBs*fzTRkZ9pz)*zq7Ax6LJ%|i!qFFao`Vjlam9ku`B3a;ruRS96ESu7aw8c;s$V` zYpvtuf{QqP-0Rr9%O=_9k9~`0v@W{rHrqF7x=lC>B)-k~Rj+ju=irL3UA-nguF!9e zUfYC@bNz@;Vj}V#I_k)MLPziYH*6R9jNBba452;cNJ#g^g@1HJ1l_R;eAci}rK65q zre}x)@}6W^FhJL**Zmwl#rMc4cmce81-vZ5_A9?6JQTc-sv$Shjwv4Jw~qM0KP*jF zzJ`8~tTZWKh+44r-jiX$H(tiZ;unN+m*xUgOZ4^94lj z%LkyhtGi?9JDuI&>bx72Pc^VjZB5$LlgNZ5F^*nf zpVU)vcyp!f7ct{FJh}{8d(U&L;CJ^tw@u8^M1yM1rij8z>3W}>R>wD;xQ}ys@Ljz6 z_n3iTp{=F(SxbRu**xm%b2d&hdcx25o6%Eocw$MbSs#an#|58feIB^Zit*lp6Un_{ z{A=-FiK+Hh5VJc5{d6w87KYa%@Y;NMvI#y5Z-%zBn8N^Dj%>K<^9GJhJFB=8{6^uo z)qF}OBaP6I{_7rqcekJ~4IdQ*^Py$vvp8{XJNOvQ!Eh$(7b*suzE@O&J+C|{$4aW`ySc8{*_-`aeSVp8syC2NI@ zIotrwO}^g5;IngxZMT4X?1SQ78iNzRKQq#R4rHA2!7rkJeTUBGH?#hWo-;VH{6Xak zXA+n7^2>oiayDKt za4i^QmzJEuclz_Tv(#;evxfU?XJke@*U*j|Pni9F|8`>|^QW-kx02wM^yAW7FuL<~ zi?+SdHoMes229iD=ml`!97Y@5U@k4`o;ZZc<19UaW$8R?fOdPZ}iH2O@b zF?%0Yf%DPu4}3@NH2Mm?rnsW=80PUOTiC(-3(mFQ7tIzx-#9j@2p+tf_s8=7f`Aie z;di}%`LWDRh9aKQ`82qYJVqr?lcB|A<_U~Hj!o^DRBK}9^^8sTjEr+%vZ`lH%$(;; z%$#S(67v%cgfr<4`3d=+nEAaY)6SE~x_sv98WWq|S9h3+*X?VzKf`ZAOZtjzVO-qH z%tMW=@dy9;LiC2>BEa$>Hsrw}*3+Hnygz=C98Y`>!TkWgpO0hXcw6_eN*ii1@gg9>n=wiGLn?^SRJ+V~5+-)~GL!~YA5Z;*?BujU>2tl|ip&&t8MSARtYX!DN) z+=~=2M`MhUX5d(qIULkvYIL68p0fzm6P*_`R&(!N0l4zIa&vNrZP-k_T`Zx+Mxcg0)sTMO;hQDz`%G zm3l_H=wWid%E?W@U){2F$9LXr{%iA!8uo{2EyCOHc={{*tEF|-R_WoT_*FjI8b{lb zkuutHaj>3uqk?rezn43;ZsC3F?$d&jh4vR&Us(vh;xm>TTAdKQ;Y->t7clc>^&$N@ zU%Gy%EssO!N7wSZsO2IVO+h3XS$x8g%=<=>M^VNcu9AGE%9^UxiNV#6*;2B9{g`6g6M~Mm0Q@Uws`qv(KtJA6!fY7hTs7R|gNODHZ@PS}VH%oOCx@ ziD}@d2OJg?tC_wjC%C-$vM_SO2!;PtN|OD4bnI^xOLISU&!X5vRTS^ zL(WVNWOx#OGQVkm#@yV%u2+e7AY;ti3WDpuh~F>{9%Igs(qZMnzCSWAT9bY@_BY^3 zGcTG8&rY`AiA>?Mn!ibbM5K;S`i)E#Uo$6~!*`x_?c65soM;?0q5Z4bPxc|#nSK4% z=}Y@`BHSN+B7LG`BNI5k@MnF-?{?=Xu|{|ud_%`0P65Z7XXv*D;7;~kaWRvh&iJ(s zHV>HtABmg6|3G-X)yiAZg>LF9w}Mw(v2f2qYMGSze7#|C$-LO{Z*$tOn6=x``#;e? zVfwg_&#QfSDp5*yOwLyTC#p3a+2-5ih<Gk@-^YnK_lyUuDlAvg-YQ%i5?f(^%?~tOm zWIc+kONSQFev+ZyKIMlyU+vO=!*PRfvQgY zm+(+@JhnKBopy69JCOZs@>%52iX8s^57{!X_-($_Z&T4Jvb*9X$y2P`)*F8YE?0i- zdPJW|`X-&**gr58oXkLuX7S!7*riMH=jPyR#B!<(AN6s@WnUhD`61<)MrOcM&+lo= zArG&XS{+ev7Y29XS=_sXd8GFqJSf?WZ3|Qp`}<~$wUzo|EALr@9&R{0H#m>$51;e2 zg-;~kAbcrp%^{!Odk=mMBkvpKKhGN%tl>;l$*usIR301~=g-vnQKJYN<;?R3v7zK|?_XE=U~8NBX=|*`a$;=CXP2L#cwg=a^7PTm zvi~t;E%#f<8?vN+b)FP9?5sTe$Mda@XJ4}NTBL)_clPn<`}yqQVZY4NFX2zxHGULh z%x`~*cH`&~@kf}tJYJb~pyqm z>pQyE`@4&qCmh^-(t1>)-zTX@^+o7r;Au|f4PFB*2eTi^>?g~p|Ms`y=Zi%*4?kOD z&aHC!S$ka`fS+^W=ku)ivumwf!_Vfssfm-@a-PG_FT&4ej~OzeJ$;8uPN>0>jgO}~ zwDQ&iwe@AHo>~Omu6-J&uG&hw2A*V6bUGA!aMaf2l5cp}$TxkdPswweV zNA)bV&JM=W1In9?ZN{FU`=n=yOI7*4W6uRNuft}Y@x(iJoZb(gD|VFAL|zy3<9hbZ z>vKN)Jb z!~B-1uPfgAr2Kv#xXI-AC$$flzP)n<_KqfN^JY7LXqvH)DSEqf_G5Eya^O1@bav_c zp!5~Ay?eN=r;;?Ku7yj}I5dqzQ^msK(52qkVynH$tYwT0jwg4Zg8Ee^cNtyx&)fUj zWP|IUJkQwT`Ok1J2fvT>&{cWwfegC3GIgYoO{j_CDzmMkU7^kj8Ci>b_Bmrzs=X#zL!#eeA4c=0B5r5KCv>^7jsRlEFymp zywoxWjm^@&F}ogz{K+WqskYBN#dw7$cA*n`*GM8ER$1D`+{Qo+_Hw0 zPk70x>Hc?=cBJdG!5kO>X7#@x+}VD^9(l*+y-weA(f@ovGWhT8ZTIS1H%>h@!#~|c zO(%2y*&@|sM9GYu7bwEDFZ%;<+~7<+<3s;RgI#AJKaFmi zg+a5==T6$Hqb*>GGS^aD29}ruOBd%UYVL{7l>YMGo2PrsPxV{F1lAJjpf%5gfvKq) z{m8sHImdWqPV-C?UY}q|^%rydMKtJ9o`0L&_$+7g)Xyz`itail$ ziRalH-|5J*&RW*o*L#1TN1i(!dDe3}Q$w}Iy2mj`2bLmW(0K|MnDbeS+7BNlI{@rc zkne@?EBI2(Tz%>+gTiZxy_{k1VR2!*A9>F~-n%*1v(T4gm=_-%>ia!(2r%x-{GfE(zn!f$oG()M=}Xi*ETXnmv8neIodZ3A&0?K6 zya8GPV|244&(@K_2=$|GpuVQ755zZN_@=__lhL_gW$nt5(;WN+Y-$wQjWyR+#U-0_ zpetiPj`KPsvx38IXVVz%D}VR0tAnzV&?q{YnS7C~t+nBL0&0Mu#T-^{4k2q3hkX zlM-?v8f%UZUPXLxJn_Lhe1Qava6~odF!eH5>t!rE!1L|=en%c_@58NK=*8JP#M{85 z-}zfVBy`&wr`-1NdnF(Ukyz+9OvJ)^W~NI?4+Fwc~)@r@@^x~N&cmi$tSJ4s?<7bS}8F-dT2IT&mPA!J=?6rk6B}EGV>GGpdFjPpImGcPFbtsJ;g&KcOlm^I17_~p5^G{ z1HXstgAegV4|ZJoe9>lG|2MvTUZMw`(u+=!-H$!OT#R!kyU-o`8Vc=Pm^Joi=TP>k z)#U{{N06t>I(?Y?@;BTVgW^CP@a>Pz-P2}~4-*B(FXLOwH&&nX-m?3w-*%p{BYW%Y zd&Z`~zm2RfcLL|j#Hso$Yu7LFKQ_VtCN{-6nBLgbFP}ECDfX=rn|cPF_`LB$wpufb zgOQ=%dm_3fFBzLU?X%w4R8G6b*~d5|kMrJ>@C;+^TRfC=FNcLXZvqb*<6?eq=059} zkxBcH@!O1XmDWWWFE*aJ9r;6y(}Etd_j}3rb-x=wq^`)WIdhQDtL~2CgxCjj?x}2) zTgz59DHhWjV)WUx^QZM2wraogr{_v9=C+TO9zIAo{>Pt3UW{%*?z73cE0@xlrJNna z9GRSkF4`Ch9PJ}Ve0S^UIjv%S|4NNnXjLto(1zG2DLWk2^- zZHFJkzuP@>+SgZR>Q`!?DY*bP|N418F|Jl4h|Nf$GPg{UFiTnoby(q`eQ{6x5 z6XiFJz5to%I~HH&IR1_g#n7*@-QZ~){{24W;vMkV#b;%KBmd+U;9m+(&ch0>qkU|n zSzkI0I)Kw%8(ABQLccJw&ieOR_iK(#Uw(8XcE{K+>e%VZW6jr z^vu+K_B~s_$tN&+j%TrH$?MVd6PbX2h;2YG2ZN_$4{4YEWmWPsrKbYmK;ysp+vLi* zacrLXM{?5Dw{)kg&s<&R>a_;+ndDS@EK{eQ#TkDGD?6;A1!Y$-?6dQ)s6U&$jICV{ z9~SK5`WLyr*9(sey!lF#72MACbgp~3F1yV7#tFe&?s@6hz?*ZqUKqUfo-VGpvG1bk zqrTa)OXHZo2$r#Ye%3c+#jjZV{}pThcP{$+o`3Zn*3zfE0oLTkvu-)mto@g?f0^2< z$N;qf@HOE@ufHanWO4$)(=&|omApXWX+BTscH>!&&O_!jud3tB63wfA^w-q-Q;}J} zO~r>4$B9g`WG~h=yw6^P5`2nE_NouXpDm+iH1BKfEWb;3M}Al5DEs;OtdSI-N}j_g z{9ydu==0Rq-Jp8c)Un-Q1xJYP&j+gJ8GCE*lhhi?cxo?+R~6GKq&-_-+wo}Sy@>t~ zpDCW<=A>F9OJdmz?)#scy>D$vPVS62Q@Jvvuw zEiW(+Qr(VbJ`E3WPPj{N#bC{vC^|X144iA6Rp>cof4@JnVw;qA;BHD(baXt2CKI1Ja zp!O{EEvN6l#P_e$`6c?uT92UFInRRqf!_0~l*1IA%i1DqRCz<;<4f)D`aO@i_k*7a zBp$*Kd&w%;vJJoLHP#@^o@?Y{^q#$K)u&p!W&pRJb@<%nE8bXI-2NW5i5j&&Pn`CK z_}E+C(Dw>`{|xK4&syKug#2-DeeTj*-pI{=$UJXRyWU+NUiajlw#|90TQ;2-R2XK4)MXV^&bL#fi7K&bv8AkV*yi4hya7zCuzcSl-7Oxonkgji) z7bQI_9eX}}IdxtCvE!F$9bEFRHLTqB^PcK!t0!ka3J)3k15avw zB0dS95O@~1`gXA%Fgu1`>!1(Tz@B;zU*R4;1N3jvzn(QUpV7}ZV1taU7zaFc=tXpm z@-E}_e_5;V^$oZCc3pwpTYR@~S2>?ibXF}ke=hkmzhph9gRwkF|1S|!le`uXOM7Y9 zt@k`goF+p5HHp5bK8HLX!)LV8GCIzOf9b~X;LWG?T(r_2gW6LY(pgF0CaDjdSg7iP zko*5N@{%^?*|qa4bavO6&zpI;kG;{#@0VRo=~DY$;Ywqf2VE48QXJlfhx3b`#(#3~ zARL)E)I{tBK3NR7v`bJ{gT(tfcp3A-u%Fgvn?t?I?A+p$sOv~?jY+JhCi!}#$2EIb>lvn;|;tH;RWS>uw z^1mX;PcHcRxoQp2-ZJoh3wWpI`&_x=fa>+)alT^o#y|2c` zDyL_ImEY@g=p{VChi5d4P7cjhLMy|+@Zv`NxLj~+L3{Sfn!Tppd-hq<(1Ul&%KiK0 z*=y`y&m8**^pgi?>LaC(bIf{0&v30*PWrMA%eAeW!S?jti&X<}Co*Vd! zFYa;*a>m83v^w?e6cvxiC*tq!(4t2aTN4@wi^E7hMeFD z0b)g^)>EC=2NGR3@mEdkj&_SakNl&Dw=X94g)N}w7O~mS1$*nbKb7CH``88T8AHz# z(}#Co2G2QF$Xk7F)z87J(OK-T>@Yg3pnb|CeQhT&mmxogb+!E7CD5soceIu@n&-7P zrC7Y?a(jN0de-Rj$$`X1=~ZAWKb!e^VsEmodFUYR&!{^GIf3`_m3kwz?bcAAr`V>} z=2)}2rB3Sr#3~&)z&U%Msw(0AdA9@e^n|^y zwyB!>3XDlSQ^y`=e1wQ%3cT-wRvLS)_-rv}OD!(ok2<^)!2^5|8;TzWZmMd5UQyLG9O6LK|7k1s`?pZn;a}e(+W>C#L z5!Qmrop)?K@9{k+Z-2*_N|M9lnsP>7Q%Q7Mb~Gdj8L;oXe^?XAU*_?#Bm-&ocHl+!oXS-0MMy`?$K8OEmI$5a&AeeA^sYrk@267gSX%N*?K zuQ89u=QH&p>a|ahd-7dBAV)&JpZl!h>gD*kYC||G=luY7MsZfn4;mh*GcjJx502vf z80{>8hnkqTFrRz3;w!|FhX$Vb@{=F8En!}z_w=l4ZDu^%TJfz>!TYVIH)ipyR5W_Y;|3+t@<_~_V{gnaptoE)8|H-)ndv68T-N2>2yJqgd_}qDb@E}-a zGl`)PW3gl5_HV=yv2g}J_^KQ-u4m4vnuaUE4Io`^0(d8UTSO&Oe#lqvU<&iom1!m!6e>qtDo#cARkOs=6xrnXI|- z5!?$fj=#`PSaUbtOSXd#qi4hmhTnKD0N-r+Q6IJe9qf!bW}ZV&(0`os*F~Fiq0+*xqskY7Q2;21Siq zUk5)`KCewT)lCjO@dvdthxfosly%-PYrMi|jCI~P_+55xAQ54`S22|uXbf)~pH=+z zir#IMo@5=$Z?6@befMf>AN}-M_+~X#oSR-{+GJfShOZxEZ+?unRih%Zxq!BF+fSlx z_@uRVB@_0%%{FKF^nCAB2y z(B_l$(~R6d!{;@O;SpeHpr19s+Xx)@0x#=hzag{2Y7` z;vQz5Yy`Tg(As4&Ph%Z&7xSap9lHaG&i9!&6fg$?CyKcTr2moGtB~O^{r(a=_w}jh za(pYx z@+rNcJ*=8D#TRql#)o}vvhCq!Xm_iXXc_>X%<`I+6#99M=#@fr~mPsHPHKmz{d##yQ zgYkERZw~D5246_|yC1WMd@%m5+y8j8e^MoLI(rCB)f>BER_P$`{byFgw&+i`^RYQwMOYQt$Et>bNjbWeEFBy7pPd+ z+GD@_=1a_*UZSqfi|j$_sI=_d+%HxH*{}0NG^V|Ab$V$*qzvkm+w9XwS?tkFR}xzB6a{%w5Fq-*IsK z0=TvK=i+yH2EVzy=I)~qKD5ULn6@5@eG~)jm#G#`(4m^oM)7B2BaQuFzYO)HM}a@# zka5M}JMmeJy%k^3xj%gNxA$c)6N@3as6EbN`0@qLRkF5z>x9(No0 z6SC2T*bCJ?SBy;SXF1>IH~fL3higJ5>uN(oo~;Y@OvnrMPRbAMoKg@f8bd6#4&6F| zPv~cCAoS~)f=^LR%+=31P~I$@Z%yQuc`en+Vhgx=STnWQCkCZk}29J zc5huGK9#&9e5_`Coq3uwa_+`#@?h=#7LqmOX(IIIxmWv`A58&XY>q1{wFO#(wyK^V zjZX#4n|ynoQT$|eF7s;6&eL4dh@)lH{b2_PZyr#;E)s_`M_-eFk4@ex^Msu{r2@#_H}7 z=tf=&s4H5;9HSV%adiLIM)J`gVUAit{?qMN@3Xi2N=!bw`A)1w-*0zfEi*Ygy6Y{g zYWp}V(Sv?ezY%|F`_B*1ch_5iD)rxk4ca~~khrzfnmmnsyG<(E}eLHl#iCq2%oby>bom_s+$=&kPwP2+ z@yUvP{>-1?3^M#nt30TE7h1oJRr5(~H3m(M-iNMQ+g$*ROX2f4ywd?sEf{9&GdBmQ zfjpP7$o~RxtN6uboMq$Mmm87M=kS#jOA4S%q&wVsi`rIf!}P&(7t+oc+F49(PqTh6 z*3JTF`c?SW1FZrdQ6uf>x+;B&j$(KG19D0MqBpOGHa zK2M_$+2`rfK<{{;UG(3~kNfG%m0{y+jS&8hZ@=LW>3S_TB9J)Qquadt?SIPt9SeV& zF*K)Q|C500VE2FaP_I0#9trX%e?T=!jPK&e{Gsq&v>(n-doL*$z{r}v{c!j-j7-On zBj&pJDih4!&lvKE?;VT*kKtKh@+0dFg}e*z_~-GR7?WT51@N(GtojUE+ndLHThB;m z`rutdf0IY#Z{I%F$wi}2@tKv7eR%*r*M0H1KI?eyE^=-Z z@0%+=ABoK4a~a!+*{89N`$ok-9(l_e|Xw@Uqv_zS>rvoy{9|`#<_K z{GFv=*FL=Xdz+tc^6;?h-~LbW?^Vz}lYg6m<6!xBk;%17pYKWUxqi6x@fgmRSKi|T z)FW4{L~9xWb6!q)aD~pvf=;ZptW(Y4sC2_`K5o-oQF)5B*2;S-sCgp&)_`x+KwL)p zQoL0Oz2)!e`!f0&i;k=)Ax8oKO8aNj#?8cQ@_2t4apEQ9tegA`^qF`@HeYZpEdoB^ zx)&R}%#pXtu?_jy2AxCN1uwsaU3kdYh5SbCt5fXEy=U8m{Kk6t{vpRE=$UThtcmuj zs4wd3?PS07w+wlD5#1~KxXjU$m!l`KAI{#svi4gYc&-4Rhk3S8ImK4@v*Ue*#-^I@ z*d%@Lact^x_IP!`|C;lvf9dCUv%t^+Pb&YTL+c5Q)1716y4mJk3%q{y4s8qL>`WK zvL2pTOZxC-~0Kzf4TF%c^BNf z>qaKNh@KG~QD|F+f39_k82*E)c5>ZA*={Y2ySHkLUaj>4}e*I%M6EqB?}8Yx?cR~@}FI=EyaKH zM>3yboOV9l@xd=ZC)qaX#yJkW;=u^GRjg{Gb8ciI^CQ_DJy%m22%URGP3Zh%YeUdB zlzW6Plz(haNN1v`-uo#>(}!w1Sf>O}p#6%Eb>3u=iH8dxNgk)YLh!MP90%fM1|Om$ z^R#vI+v*IT`Jjis&ujCi;YZ^yVf-(-<7VHGwaN@y& zZpo;3bPKv!^^V*c2-uV}G@n-6s)4Y=NBihRxBXSNdu zunFKvbA>YaKE~J@j~gHl0UX5PV>eIH)WHS!J@g@7Y04On?1ndIv5Wq?p+^t0`#L=3 z9gD`3`Ca_9tJ$tM`wHtu9}hh|m2b^FN_%-fh^5y`9gEy+ElYA0<(}=cOb-7055YAw z(q2=21}y98uRP+$)a@Fdkvil99V8z#%)8}}BQv5W@@Q&L!)wZY_12ztbx80<+6T9> z>(OKIxn#rb`~T37{P;{*_ac)9R^IW#`gOrdzlvMPJ{fvKL&3C9_r%-yPaVKwYEO&a z-0uYb`{{cPV|bBI)t(mpHHZE=zeT}`YEQc`@A)x$pSBbOQ|`9xrVDdqGS9;onhQUO zo>dNDE_y<>r=?$H8zQ`$tUV1pf^`jfDq540y^-9>mbta33t5xzulBUJ-xc(y+SBSs z^XjGON%^3+(YEwCKBD1U;b6V|3ShlmpV*68;8f%M5})E}$&h&SzkUPW3Id5nSC_$W zuD&YD@MnI;*+-eajB5{GcG{6W$gCfV->{xBub*oXYoJZet*WZzuh#OvTFaWr3+Nl- zW>ppZ&9!`uVf;<;y!cU{ykEupRr9SJ^L#DqaXGoP!5{Te8aJuULca5-K5FzmH?W@j z>&<<9S?;f|;IEqJv$TH$?cbnxH|bg4gWt10D|vMBW6x(&`kek)Q&moAj>!r22Le@N z_-kgI4UDr6AKZT~zwyWU;pVqBj9q?W1%Lm2tXU5?a+pm&ie6WH@T?l5f+rI{FJ$1S z20iHFCy6f)ehj|A4fspqt72ndPY!?3qc;n_3_V=>;e#dVVQ>gPd;6==cTaztYG+z#}=xmF$ab%0T=w5IzjN8T^tBAIuZh6YI{( z69!V{uEbP4ScRW9t}*ITf6<=P5-MD z%S4aP1Fkqae;J<}EUR}O@+rL_U$&mP+s*vlHVAlEdf@#-Kk$zKgz#SBz$?8!8lH^- z?*cw=cJ!5aSA08P{L60@=mIbA_9yG#^uYXk56mupGxs-Cf*@sZ#q<}#v5E^}#r`!l2P!9^3sRxg?`)+RoM z|8?B6+8S>d6(*9;!MM3t&AP*uLWa z)};7={oAF^Z?zlU-?TR@@WeEJ?+=cnJaBBwg70b6c+HSEgCp_tCgBs@OTP47I?u(W zWJ@yU^7o(0sr?164Znat!!L$TIla<7I-^GM^p&dp1;1k#4KAgNrS~*9mEIc+4@KTc z$?310F}e6H%D}Hze>nOgsXr>DBN>Bo_|eh5tAWd6tnv#_9R!Uxc;E*p8n(g^gockS~7sXx&nGO!T*My;K$ID&m{ez?R@@R`VEF|*LmR0 zrmuEww)gjpKY@Ja&FTAlsM(q5dLIAo5q#W-@PXGbw}JP2_)}hG*Ym71t-+6d2v{Ep zBzjj8JDZE&jO_GaOT*Mtim(RKZsl&7dz!WD4$icA@F?r4?RCs)?)6o5uk=xKiuukZ zw9EQhXM?>a-AN5~t$Cfwp69Lx+RP<)jkBHHC6MGrP#eSRdsx7&!`Vtb1^ocQpH!rYeWyT1G8^wrku zy})xN^{8f^83wx=K;A#O+*E8NbnIGQ7*zaPFxLEiqZ4daWe}})nYx^_g zLwm*`*foyr=!atP(G4Doj!wseaM}s2MFZ6|`xEtM;?kS^-VH7|Yrg7wWW>CC+`#V= zE4VvcZ6zY`h59apXViX(+SAqS`?~{LXb+cZt0JQ<`V`-}a7<^tySo})Nll9Gsf@Lb zIoOcpz1BLr+DzpnB%i-ncF@CSzkyWjnp)&3WV;g zsR>uZt%yek6f1A*mJ!+c|(B_WGGvLXYp{r=~>$EA_Tt%BL$jqbY zv>lTa|Fw?L*)P4w)M8(0FZ->HZ#|yh=UOGbJ!e_7uPQ1ryyx=E&M76Kcj`(*w*#a2 zF{%5WDMY6gg=Xq^Y{L#@r~?_6ZrV8)d742zfA}!kfR4EY`oV{BVCsQC_abXu{=?vj zA?;1BUVCWeJuc3Kt6pSz$0X^i;&$OJu(x2c;J+On(^yveN_ww!;GbaNFYR4^Dza8o zYG`yd{C*Am?!w;)oTBjzM|bx+@Eh6#f3JLBVAmL2_yxPhAlNmAZ1B%>;g`+=Ucp@k zF1=&fAzkZ?!(tr0MmLo7hEB3(Uk$!bDIL=C!_Nmo>&~lzh9#kQr%$BWjvk<0wYjr;2C&Zzz4m!4wDmk+Xeac3jW)$| zqT4wRoivYWW;_+8{ubgwp|PK-3B^W~ggQo*hWJi?y$y%fGAe;1!kAqdQ?7<3IOtD& zAs(q49tfRxbPexXwB-x&J;Ak+QMI9UCHH;#Ig+P$*tg@pA00Y_nw>L4wL`4Xxkvg! zQQ)oxZrNSwGh+vu?bxXOTTc62@}dqjUh!!aW1c|HzvL@)^dRX?PSwHC`zMUW%Wp={ z94x)l^uJB-2M1_Z^xlY`clC)^9+Ucn^~g`7Pd$d7Jd4U1GIY}eR?7O%q)*Tb0Icvr|}O;dZ`OM`%3%>`L>zy;D4J?W&cnQQq7Iqh>izkktO+dAf-%(bm!{*mj!>X>&p z4>w%Eybj&b-~4C0cmA{4ULT$|QTioy9%=ke<}!CNw`#M@p0!A`l_18!eoQ-_NM28n zyq;Yv*vFu?*-MebdC1En%w7=BB4GH%_hGEWM1{|)C)s*;Mcawuf2}>H2&eXB5T)`_-c<>#akA0{m6%X!cp-5fWuV$ z+x8oY3k0|0t8W08+S94r(Mdcf+>WmRw`R^?(sGA0PuINkow_0RybHX}1oyJ>i^`C7 zCmwZ@<(bPqSrpp&e6by4nrkIqmAoV09muzd0bBwu8<~gVbHuJ;KEqr`HS*((WiR;- zoyO*Rsz?V|$v?K2W>x$*#w``eye|Vs9+8nF+3?(-qT7>3^HAjr{@_(hs*oiyhD+$@|Y2 zh2Gg*Y+_V9p~Y(+TD%P%-hm$CW6{FpW%04-Ve_)aHne@!r=Z2cgP=v@U-qZPr4B7r z+hrh}?LvpU_TgW_(n8*Hw8XEz1g5w5-T2X=(B0wp3SejskL4jm>z3_=f6~ zc3xwdeHO6>;w$LB$Q*1C^elp&<9#Nl`gLMcLC#WqU9x!*^8Yq`a1Gzz!G2uJ_jjwU zL|20^*zM>j#jh_`3>sS{f8`|Fna-bE&ohS3iz7qYlhTPSn0Jv?;k8!p!JGZQhh0hT zn^T|8duQ;TxBX6j*S@Lk^wF!>4sa;894Du!gML!*mb8!MUB#``zw&|Rldm>@?S*-^UzUuS7Mqxw@yi90P6?(N;)Omx}*F$sK(;7-z z-1W%5i<0&@{z3p5 zw|skkyuF2UA84l$o|)mZ65(ceVzJh=SsUSAm}}XS-s@@aR@pRcDZG9Me&^fhj@O0^ zX?Yw!;z#%q@6;8CcFigYJvJP@&lnd|J3A!*NIX2rN}N>aYk8IR9b-F)=l5zn8&c!h zd#Am=qw#258jr@M@n~EcQ#U*sZ?MPY`l(?n%Q%2lIrU}?@LL4@`|*zsRH7-{lgwP0PMoQ;EsS}@?uP_ z)6vRHqhG`Klb!t}b*?{9(%wo=)1As|GIgIy+HW`4HcgkCYujJG#a!nZeb~(Pf%Kl6 zhbdVoJAf>>IhdbECd~Z2&enfgzw_3!ja;I7FZjac=rz{1;^?mq@TxsXQFNT{$3{`( z^slvd-IAN%(m9Sjh{HDivD<&+69vA4E#1Tq3oBl}N3g#_9KO%;?J78J^t%TNQ zjl9>{^4&f0eEfu|p$ks2@Q-|J6cfExw;vRMc`}X<8~JE46kUNW3IzjJ(b8Lyk%@Rx?lc^e2xkD>)K=K`YbvdN%7eG&c|o8ePIRB**wKaX z2>e2gt}V(vei+(yBZr~CUeE$vUmyQb&0S`l6Fe6pKdOb&$$EDj__a>{UF?VU@fPu{ z>a|3`Yjh!aZUoQQfagiz`Fe1Sob`Ov_v53bD<)n5U8jXkf{#vykG4ablfbd;ZU=3q z`cN)iQTqb*0gVdjL+j7pK4k0USLwO9fvuqZih@1QXicpLzAA#R_S5f^3b7GI_PQGK zAs!r$z7p?@hX==(7Pa(LQ`4Kh72>NB_)2pl_-dU!50pP!)SeIBvyIt|S3Fe>PoZZf zV4F^Xr)I!Y7s68$;O7~PSMAHsUSac;)3)||693oUtt|NTwyk=NMm{$?_7|AMQ}Yae z**fa0gXFJ$uMEsznfbVGys+q~^mw7xB|Go5%-W*C%!py@Ajb6c9YCw&m&o^0e}`0_to zZ3TaYukNlL$aJoKb`ls zU&?dvr5o#4Tu1Sp9kYNJUhlpHUWeb)V^g-RknJhQ_Pb+K4Qh=~<25lV#;JVAe(xhQ zO;&!3VD6!=sjKsnF|^4g%vDS6HBQDi&u2e74L?`# zy0JIK-TEJ2#$2+_JC`g$woC1~Brv%=?#v|(k4KS<$?$iC_qKEG!j(Cfbmect+eYW6 zWDkBH$NPF$HDGO@)VAN=YstcYkhi4Id!5gJcRt(rJdoaX*ZR*mfNaQCPs!B3j{jOj z4EW=rDgV{$hbH|`Y^QvYVm=LzV;{GpZ{?GxY_J=jm=4Wnq~a6spyCtT@i$i@AF{2A z-JC{yirp;7U)V2p^L#$@*8)3s!+ge#-MoUoxgB}P8o!y8?pK@m4g9Ix8u@TLE`?v9 zht9L=o{LOC+a6$3eTl5`8^t>(BKOzv9y&d~n6{dcjazw_ac@Ufns*nRr7_4qKo`6| z-Nc}ZJuxWTm%IZxl^?9wp4OOp8H1Y>r5M!qSvDWKu{Oo}Qo3C6w&M1?wD+qYc(x-8 zy}%H8VH+ME;o&LG-e0spx3?01!00!-rDo#xt} zXK(lX-o^ES_*Z_zvR@3W*Z+xrllhJ&|1woSMCUoTa-MS&`_K!@1EHc5@rzEa4RuxK zCTqMDu+L3F zbEta~OVza6b{uC!FkjG~7S(o9jOO%oTf6^=>osV;5!Q z$7;OSLZ7}8_TrDS)|W#k)#u1(-$ch|-yVE@K5O{oJI{3TOt>yk6`2~Sip~H(BLnN> zqXO$~AHuG+&~ZXKZ9aV2<{#nT(1tM!&Uz1R!rU{Cr}Ll7Pp>c^@#ZHB-mTR2*+`5b z6VC8;z;!)v-DL9B*T;eRn`fk-PxG@3*nZ`}rumB)oI&*nGBD9vkDs%? zgYk7vvDUlu(4GnK&1P#4)LY*%5?Mumb@08D@2o{tvA3)$It6-+fKT9+j~c9?TgTbJ zw8rLz)gE3@je|^@UklA$9?hitp!k!1fYay>FP-Z>boQPBlccwim-PD7#Owx*oA4TN z@DMn71RP8R2jJ(mNf~?2D#7zO@H_!L4+YO7giG*@taXjB65S)MM9(M-n|6HrXwkHi zHpkKC1lk-*o8T7Q^L&8kv7OMNXB7EQF5Mj**L&!`)uFp9bMAN-Wsx`GJ`?BJWUb8| zmxJeUferJAFWK_n<-!wl3#Ky5s*;{};mhRDcY!G@e~+Mh3nf?s%2u^sW__e9%l{5TX{052*w z@>`eg4j)$ROWCL2Ied61_;2^=>AK$KzkfOe{Fj6;i~g5Bm}mP7HxLiGA$E;*12$k? z{aj-Mwz32kjJ{0#^S?Is<1eF z))|Xz-S3lQ#MU97VQj}d)xo`QWU#{L`a+9Meb4(CqdgC^eV?sk(#I6m9_)fZLi}Uf z5!?5v@$i;+e_5vfxOgD@7PG3lx~!^h^oTjb4i)>y4xL*kF(J?*f7{z;~XBLRw+4PL;UPFL)s5b|1_MmSRH&!ia zuYck5*gpBoNqPPh_U^6Ivg?~Fhc`csy>sC!GVrD9jkx}ARC9Z3QD|R>YRl)I=)mH) zuoMOpe=M}Hc3!vKB$!PDo?j6SKKlI&CPNdDaZiVXj! z<6HAr)ff`aHSfsB&o4ml6*A}3*$_IPHIClX*;*0p`6B+DdwV)PR#m3-SCSs!=7%0Q zMT^w@)zN1josn(+IxP$TUW{)P08Z_j_5r8%&)RV2w!b|woo8?Kz$IF(aNv~Lr!`ux4<$DV%pF`LeN^VD>i5PnYwzX!vQpT-}5((Qk< z2cqCW>z! z*xbMU`?D*nu7B;(8E!odSHFE4ekya}&ETi+3O0PP|K%s<uKIIj;MS#kG{GHP$()2=1%-bviOW3X}ee)P>d7ls z?Ui}dRoF3|DO=}IYRqW^+)Z&+JNt^qu2~?>bg?GNV z#1lvOFYdYaGn7q!z$I%%2MueP2UfxMhy$C}8@#Y&`aZ&^cr`hXV6Gfzd=bV)J+P{3 z>-cYmIlILhKP6Vv?>#Sl#%Ea{I#u+^!n@gcHd|bHcosR!gk9rr?1whOzjrRO+C!T^ za?cC@@+|Wq@p?8M&jxGfDcNZnOFu){mI^^!AHIaCc$yk-w+>N$EWxp z6CWRJ*x&Ac5`1Li&ugIwhXtlCl+FRpB}SM>eb)T;u;rX(zBv`6@!~FyU*y(}5k2qEfZ4=u@sm1;d8<~u z?4j2_ZSvU4x40+R_0Q#hTi()rRoP4Vsj8KyJs(SfyXgRM3bvcFz&2QW>B5r@pIa4A z$@D+Kr*e(YH#p^Nc?YK(4*;i&Pr*AY3%ukg7+k9UqZf}pPfSJe78hQdAJg$CesukO z#a$bK(c*93K(GPNgba9+{1`j%dbtZvQ9u5V3(rPp&0yYd)BT)1)PgR{tXl+3`57=B zEIoS+Jn8ZDbxu58v2^iR9KV0riSXFT%zyY?1)hkZB$jc`E48>~k8sv$>+E%> z%QN8o7Y9z|ciHyXUe{RSp@G&yUH&-`}`ybM$_ImSr)oauoQTcW;J_g(83lY%vVcWpml*7nzC!24g+h<0_4WT}U<8RXCYPr_T81zz}})%w$!!AQyX z((rpj2E31D!28p`0leWm_vZ)Teb;I_DOi!4j^7J2;9ZgdZ}1bqJLvf-y!yZSSLiFn zV={9mJ-Ep>KYe+8`g-AD@~@5;E&op9sckO&9$!}S{NYb*c|I6_CKV^KmFWNSA9_r&ZS?oGZ(2GME;*>`**^ByWJd^4F3D4{{j9Rra4JA94iKe zoV)qahayI|$z!KAXO=|#y{vy1Wyu#CHc-Bp2RGSt;4Ll<2jW4+=vY^BbEBnW{`7o5 ze5g1bYfElkwDR4JAG*e#Q*HA2Z1U%9ZIhZGFUm5X()?~P`OufRW69vh9|_h&6tA=A z@9FU?H(sWge>s2NdCT{XN%yt>)q!QO@v>=I@?SJRI(1z7d4=x!glO~6P9KLl|7*km zeSXo?=fULLOf&RO*QXal@0{{L)%~nPwLj^eBV_ByRnFP-tUt9@Smy-W?@z5iY3ye? za3u4o9KKuPi7hFg%GKvKzoqvte$%|bg<<&tU|603!;uaQN&D--Fqrt(_$>34cQt=! z?qz9D7jP*Lu84VAv5CbLBy-_9*}r7(pD!ZTRop(~0>{5`<|%I-o#I7r%u9Ide$JK= zZ~Rp8EAJdK;)!kbdtL{7vOgZ0z8`H^KKSIkB;-T6aVCFRaT?V-(3u_aDXDp7-{O2T zw|tnmsy%-sciEj=P7bWUpWm4)+jCB5UTJDoGv|E&cF&v>e5p2tIa8+`irOc82ykQ4_oKN!!|mxr~mk7dhF>-o<1}A;PNbdkX#>)**d(! zBTI@;Nr$`gG?0FD*AL*nSZ7 zaN{3ddc5K3Tm44Y+j=3%j|ZSfHvS*RcwJun-Qslq|FNe((LjB^==3RFr17Nr51z4V zy-U2Oe=h$mGW$i+<>6HPqZ;i0X!PaH@I^Vf4Wk2`%gFxU`RotF&YUrm9F%gul^A_4 z_gL#MKi5he$-cn|aZYS&t7=HBVLzzeNuK{m9=FbAtKnT@TFLVtO&g5GN6nfSIRCPC z)zF~&Q=10Yn{8XBv0lkollui7`C9Gqk=*@)?)Y9Y_8~pKdolA5#n@J1ACwmnV-3UT zRpPwjtsk6YpFyg)xH+SNSZ@dMJH_xj$)VQTNEfyvZG8k?xzRuOcNb;Q$30*1D;^pt zk2;ypQR&1RJ+{J|&(R-$xcMkK#-`clT^n3AJ98r^|3WzXyr=Jj<&RJne`+26l;lur zFPzVX59h2AdHPnrWlp~b3_nK7Mw8JDt?e4F}fGc6>2I zKEyxR*nl0APUQZ^&H<9%Y?yPN(Vq+Eo@aPTwXa@#xXw+w>Cga17 zj1MNBJuQnrdol2e*Ybdqy&We0em?T>65lnK?~uQzJ=sk~!J7Q^SnGR3($_yu@Qgw7 zah3!3VAnrveU$D$xc*u;`&gXO-!V>qgSC&gK1lZuua!Q4FWr1R=^*h}Ha(Q|QyA;M zXVP`SCz)^bs}G@tg^xavK77mR|6uf?J>N>#KkF3Rf$q{lI&WC#s-8gZp;yLqw(3S# zFKRm^{?D#Y5p(pVC*G5}K56quI=rpWOR&nu3P($Uwdnw` z38p)%9=FDNm zxsvjTeKYk*Jh@nz`|l2>zbzU4y>*cN?Q-fXy+U0qZ+#`R$GOOQx(?X2#$3NHkoaIt z`q|UBkXNPioGtQ>t)|JrDEIfVMkjwnes-AeEvzxdCzBVPZ#`vduCf;W@>7At!`PT^ z^vcbAhLBm+c4+4F0c1BWd&BRSAw$}4*oE#gjah3q|M2+nmd0<#Uvi*p2a_RzOX2Zzk4eD6ad#A~S|Q1r>t{(m22``P?OV{efM`HbRu$=vziKRfR8+i$o#-EYi> zd-2XQz$X2ZiT9+vt+a7oF@U`t2;aMZwLiY^IY@jA)c%zA{oB9!AlvUT{%*=&s6@{w z&STEgMHVXY?Q;1u=jrm@^?NFjdF5ugXXWnr=r5`BaM4q$fvR(ibp~h$_f&f>88_OT z?hlC;-}2DHoqz9YE_}T@XIO91B+mP6E`I$#_Ev{h=2$`L)xH_NMEFu)qHiwyKpK3B z_(ETzqtQpbZC|46I$xsudhShCoVF;`*H;kQx4AI5c(E1Q=P$B9>+H`Dn*G6hEo<0E z?|dOd?V;C;&}(Wx!taqvUt;fJ@JMye`hA<7_x#1?{cuA~BC@b1ar3PvhD%)z@`R=U zQ#F5$HHqjo+{Y)r(R;s+dsBVVzw4~uE;wwkOR|Iythr~dq&VP^HGVltEl}| z@VzHisU~PsX)wYVUi!hFHsv_U*RP`OM&VBwR_rB zhxUG3k8iykJ^mo^qo0r$AiHqCqq|pO6Jn;WSYA8lCadk&=MVF>$UdE`?>oNpCU#(} z;!nVKrs_D$zO>Vhu1C|>f-CCxjKL1fzoKH#zGfR=;p=J=M<;PGE*O~v4kr5&0pZ|V zlY&v6i{U3ohpHC%gTle3wyv!uf2DTV#9#xpKfE|tc91xz?*|7r4HgI3<*Yck)W*Rg z;ehLYad3T2;t#jDd~#qMTqYb`0uIo1;9zU*1;c`CtaKcldyqIN?*|8;8!Qg+7qa5u z5*r663I|+g;vfvqzIftC)R|>GQ>;X$97Qj!Nz}oE=fcnKS<#OrapjSzxdGm1ADZ`` zBTw+sR_ngu!PVf%D^nxDquQ^PK5}a2sSbl111c%*YpW#R zLTpVVGI0%`mHuE9eG`RGqv#vS-`sNlu6leGAMfOpQ}esfda8~wG!bvH0+WKdf8F+r z+F^n9^0_WTPF#7={Og~RV@UCW(Fv!Hvvc7*GV$EUi8fb{h4DKgb?7u;*gK;pQRBB` zf3a$CHx>NMz(1@DB%-t2J#fV)ChGQGgHK)ojFQ19GHCkz#R>cM`5#&ORP9!C-VkjI zm%^iBHj2|2pRd{OOLO3_QcF3E?v2alK&pxrI8leVpUYTFLBMbYDSV-_sHC6}qFFxNig=i2|GAJ})p%+3^?FQ*+b7jqsvA zg`+f>(OJnj#Ry<^=H=cz4cVwItgB}87u?)>)oAS=7fAdA=SsN#QYec*doBK~{Lw%O zIvo3$88=a1t@NdOC*HM$dXHcGuH)BQ&Of(5)tL6yAC%83-*p~qkd}37@INcSMnp5t zDi0D{?&|NM@QoeIbooR3EppOoRCdt*Dr90c?cHzWvJ75w_syjH%EH6fJve{aGiG(`g1e@t67BYsv1Nk>k?d#vad5n8G2k>vs@r*<3CxU-h zU7qo8rhOj&rmi5}zlnkC_o3Z8@?qEqEI(#(F@Gh-p9$w{jb~`^B++vl^dvVw^1k@L zjCB!en!iJS?AwX9O~jb{`aTm%%Kpp9)3{#JvI<}cuY-b z?TNLa&a$FV_qgIv$AprQtAjewHJ!+lbbbum7sm(aLjJlPA3|rUOE+{d7wR6fBOYwb%2H)kpd6Ppy{WHy<<%_uM zS~28X^Xg{u7<66fSN^w^s4@B6pQRRSATj%vy=~SJ#3aCr@bv6EiMBNkZtCFg2KX@s zZsIO(&KYa-w+A8eA1z*!^%XC2_o=%P9sFPdK(nQeha z$fVVDc<@JzJ^cHW%trCanz`4|G$Lr$$66TQ^u28xiTii_-r1|oc)A$(cfiXk)ugow zf-BX|TWTM=CEGYM*+w00sIIN}B`NdZBx2*F)%xpDyXL3|Cx`zHa1wsY_IG7>!#l}$_OebbKX?iE z^-fbMf4PCs&80P=rAO3;!UaX4=+I)$#h&2gsp!jx(2hN6m4%Z)J}$ zXBJgZ-@dzftM}&4fb7ZLdCgN3#(F4y=pY3^c4K(Y97k$*y4?`EtRcg_@ zvKc>zFPAc}xq*Bn=^ST{vd*2O+~v?lwFgdsHmjhgiIe?5+TH~)s_NSRpCmkkiq;2O zz2$_51SBCZUXe&9L%>uZGzp?bFHB}8$;f18Iy2!RTA~3#@K#3cO%*F3QUq$HsMS_m zk+Ird?hm-=z4l&vYci97imhlLGGLqEXYF%NauNc;|DPW^nLRV-?7jBh>+xM{uf5MZ zI&Zha1DW(W#>vyu;S4wvveb;lF^qAz8E4G!Ulf*+3>_kOse5}K=^ zc|Om6xpimeDBI3)<5PH6S?bR0X=yuujE*k*$wNoVp8N6V-iI#GSipxL!jEb!;J4n! zVv;!)2A`T^G0z+e?o(p{pIZCV3U8m!ScLvRjD?r6=;*kNdH>qZ78g3KwmA7_)5+%h zIa@*PKY1Aa{U|SZs1=_2rS-RFzTX7yzsk1Jxw0g=h5Q1W_Vs8JdxYCu{i5yod=f|7 zk1Z&9F)ey@z^P_yzrIP^Ey!aWd2GAD@_6IR=)<$>doHBZcLnu*RlN6ETB^bi4|4whU()(@!w*L{gBw=z+MF`AWi2`!dlF}! z63kN@bJEV7NL)m0tej6L=hBtzYdjL0)Oh4SK1nv)WsU1bTe3g?(Dr@9qwPzW#{l;^ z(xUCu+m8IUQ}-HpU^{V!_OLDaBVyfeHKs?8E~7ngK%5wgi4Dzp|K3367alo^&C13$ zb*#Fi3!6}fJg-3E%kf#!)}gERTI*PLI!P>sx*bMdN?JdKC5}#po?2UScTVxn=ce#H z_u`VBKb`I1nN<2SHsm>apB);qvvn4BkNHkugWD=OZy(%p)Wj|DwdCcM!G~kpr$s+b zyLWFp&pTL(FLNAz_oSpm<+=7XJR{%(bgYkHGG5>CtxRH~T97s+In7>HDL! z*<76bC;EQD>F!_DN%zg?pj+l$_~O(`e2_HVjO?=muLAMgGWDPj^!B?;%`>>LvJ-+qWFA<#18JsS2}Hgt??0A z2l(e3PpQ3Y_^T&Mh%1OZsOwgDT{oX=IywfRUkC2=-+!`sEi!csnK}kviww5Fo1$}T zt-m1$EyzJ9uffYtOU!U%VbsBXuQt|~Ox+soFZHtr(x_*?wVnyoli$6bYSG2NPl-C% z%h|@dgUGt*=vCCyNgMY)3(pTq84~@l$Nh2ck$y@{?-BHQ%M$1h6enYa##jiQ59M=> z?57;Qt3P8kJMUC;>qy32eMUBr(|yIZXdFMkmA0qbQjgcO_fyu}TW-7ITN4rN8p5GV~UdPikqvz!l%4*$dx|X3HFnrF|veO{q$s$~ueqXR@wmj&YSi*IUrlY4>~K9f?y~ zc}CH>=d$RDe#GKtp>JePZzxXMnGe?DKKn6x^ao<&2ip3K!N{`&=K99f=+DI|wn>TM$oO^G3+8?e@2|s$_!?DIsTuer78US(Fnn> z{k!4U*4^l!9k%2Md*hL&g2p4o=sR2R8g&l**pHLVt>g~M8j1KMb~CQsjQ!Z8@Z(0? zbv#R5;*;C@b=smtOx9Jf0UeiMYlrT9U-CuqNuI%ut;Q#5!>%X5w(|U}mL{7#%ZC`- z71*EQzDS8}eqr`z2--+#u^myPZ`YPqo z^cO!y_eq#LiI2L$HrT_%HsbeQSWX5AV$6wZ= zBpy4QcBPy_+bZ76eChR#jNU_it-JL(6#g_~coTURPEv4!_-OLn!mFeXzLPrSd!b!^ z8wvd>j6o`MpGGVMS~h0vyFQvi97Ok3`s;CKu9vkP!3+zZ>#_wcmS=xk@xX&>y{0Q) zyJd_mDS2Kqh7ivkTAaKJ9HZCYeXT!$Bo zzDP}eGA1o~d}JE=UTMjXho>c5Pp2m19jVZsmTViCmTVuAmOMH%jlHUAJi{j~c{*X_ zG`RG4IpajuTT*Rnj_hCa)6YkIec$JQ`X)Z}^NhjkvVO_h`?i76$A3y59@s168e7%B z74rL^`TbMI_gG3YL9VzPjL`OIO7e8m6)M-EJ*^xkL01S&F~U`v>%s_7d;$ zPF*dJO8$f{QxEP#KA|fGy7rR`Q9>*`z9BW~>L)SqVbR3HXKx=pd)U=DJ({5X1TvqX zeV2T9-Kc1S_C>EIXg@*wZH_CVZM3hCqan}G;+6LQT+A53o5$>O{|Q&He{>W%8+reL zrxGd2*5k6aeYrjcY5KUu8MlNQH+|p17vHsv+e+G5#VbKSj~6FTSbuBc`?aY_u}_P! zPftT{$?poHX%aJ|4b!hXMvO)2Fl?8^Z-%9wYChd`rA|Z4cu&*kq2*c!Yj=+1e_7jk zfp%Wz1)Xu|6y7;#{ry#bd!5%1ejlp;{+9Lk7{9&EOV$JbQbaDtm7V()?<;FOk`Fzt z=*gB!;Uyz~CZ@l4uy!9GNxiZbJ%)GIqxbQgGn?(YXajqYTL-S%EBYP(VPna=H^{Rq zld0*AM^cgXxcEMgE3+Ibd+UDq#7!<2rJ z_H{qj#y!X%bhjKMu6~8>v-Apba^ANkKbx9tw(`vOzjyIW9KLqI*NKN2BZt91P5OJ= z`}iE-f%aq2iR{hl9DDKxmN4!CUIzax?)(jz4DzDs0b z>sB(?V^Y+5Zn4bM%_E{)$$$F&ntycanJsEh`SsD=^r3xaak4*hA~sZPzXx7#8>7=c z9onTlO6>0kDZ>BB*4_4G9G*#_a~2Pze&j8e+~GFHv>){t`M#ZX#XqpCJg*$;lCtFY z$Qs;8?5fCQd=~jT^e3^yLi45Ms*$($d@nSA7n+5Zc-lRCQ)v&mJ>lAyK^_~l?dVRI z(3uWxVk6{Qk$355r%W5P4HlWEEHW)QiIRgR^>xy=%}N_j&?{Ggru(bz3F&_HBEzZNb^~6d3pLcFs&_$9DgSeji0o5zpIm ztZhv67`9t{G7~!-gO+2Wr#9$%N_5o9U!7`Z|J%mu({|Yhx6##s9*34z+CA~_C!4pV z*(SBopGNv~HvFP!C@xeq90k7!4LhLg*lkxU8cd8L?Jc4``gy{3Pg*oX^s2}-FNq7@ z{MTf21DLaw_GSE9ph0xsaMoXxKCw-SCYbxD$UklDck<=)qN{O=qU!;4kI?lt`sCQV zj00n1rE4m5NqtYCdrQ-bSx24zUm{1HIsN=5xtx>X2jwTo95erdwKZh!bWF!Hr_*#y zA@lsqJ(hXi#JF$nnrGQ3C-dCSJhyL4O}5k5_8s6q(PyuqtA{Xe)T!Iev|TBTO*UiG zGAt#TqA@=Sqs`Vo8X(6D%umk9B;|9&6SRXYTx;x$u15EX&Jf?j zMqi{X`bzXh-A@e6%k!t>!_ZmuZf(&(8tS;7I(AS; zBXzXWpBB-(b>}wvagG> z_;jfmwwF#jy5?t^K0kXh8j-`@%+2$uvaWW*mU(eB&UJ!&g)i;!^wz&5n-BgS-i$%p z3%$>GrnNqEzF#1oi(K1izeVObrqA;hnP=Z+>Kth6Jg=pl7is5zpFcOIGW&sj)bkag zKMwtih5m<(xQ?X{-=SX97it-bp1M3G`Q~}H;h+r{^e1Se%^agG7e!@UtmAZBFKz7Z zqYc438N^ft=V*d+41dj}aZ?}nt99Ci%!$SKIC>5K+UsCkWK{WSF7$A0sKQU;r?nb> z+V5AMY$iu&kBfDr#A)j25PjXdkA|(><0>Ale57sYE%A|73$D5D8__m&iC}{^@L(H0 zQaf$6i=K7qdRBC>=sxY2Jaf{dX%THT=`^*xUYvX!T4KZzT=;E?KSRqGebCZ4Quo(} z;+sJWb9C5sf;@3(>E^G|hQg+cy3_La*K}GIqi6LoU|r~VisUaNlVhT7pPs#abjP06 z;E}cismW)+U2U{KL%usP`i!w?bLj0`b?qN$yQjYFqyf2cK!Xb!5~44gd$EfmM+c`AH0--{?Y9=u^+MR%%+4yG$p?B}y`+agY1*Z~k+#~VE&8&HB zB`%zX9&+=&=p$tE3LQHP-{o{b|2xE`4(?K7-8Nte<5ng2$e74`3OTN& zyVe~k-Kk%H-!xm*D6aRcTYp6EPvLW~V>aVPwg02vI`GB-+ay_&jPC&7i_U%(Iqr!~ z4$-%Fijz;?t=1`7o9BF7!(JtJa1Zj8con=19>JdMvAr@4f8bujhv#ev>g!J)^F!R< z@;-hF^JtejK^JtJOXx~K6KkQW4f?5Uxapg;y|4RCaq<$REIydvBH4c_e%4sV#k6xR z2F(F0&4$fZG>%G(z9F$I!)A+J`aOA$)~2HcgRIDX=&%X?bP4? zUROOO=*bd;Prgfk`=QrXqt}p2H7BND-1?qAC->4{DN9T_Rp@0-B-ZpkzDvu5l;nTQ zJN363aWONlI?xzzxmU@~wCmKq9Fs2=36H`LvKK7HXuEU2l)(@GOxtnF%8#aPerjP&M@1Ud5-x^7;Ti&hLn{*OdHA{mNqV@jbV=DfA(ynt$Q0; z=V-&^>+xsd8I#V|3C0-B)5qxQbIeQJn3rKPHZr!%i^zB9*euZX)ee(x>Q-Z8=u@-J z4)&S#C6E7L?YcL4H2KS{%gR$@4PI|c=rY_6hBvR1ep;{n12nYZmzca~`dH$Nh>!IL z+Bf|YYagxkl~$f=c5t4#sUN^UJu&JY+mErqk{@Tm!OxRd*b`%>K6h@c-Esqbz82oF zV9Yth>RJw1Fy>nHY?|)lTQR1C`OvWq7}EisnRRww9$T`1#RbMptb&#%gBE<_|6Oj53pg(yS|;~FOJlC0Wg9lez`J`~;LVwu?;$+pU6!&V}t3(H3~-aig8YZkvwn7P6K={ z)tT7NRnUd)V(b%cc@>=01K+haW5*4Q_g&fo*RpP;K&T1RELp zWh`en8n$ku6|YG=>j7-hYTn`RD~`gevUc#s*KAdr!94fz{x&!%6P#2BPU^4?QaI^1 z;GV(YBy#H&PLgX=xnJJNag`&;^gz2TwkMb(tstcL+PVMR;#ZPqI^q4T0uR&UdJl71trH^4&^&dHM-UHc~Vl+``G$|`Ls`B>SVC!hSeMOK$VQ<@@JTD(sHW zBK{NndDz7I(O>9TzYVNEn(w!R^)G{;&bZ#d`fIpe)@|ha?d0pN-m>NibsxW9&;4(` zd0BM(mIt0_J47zRft2L;@!wmXPf5xiik1WR73)*=1A<6G$57Z`(WVA^B$PW$=YE0M<--w_K@>v{qB ztx4y%w$AuWzg-b)zb=RGhMk@7pp*WOi^h?G`z0=g?Yv*~Q-pPJVj|tw&<(x|z$fS; ziOq=peB~ALFHM@3h>kF1AXc2rg_ft0N42g#1X_mnLCd~OMGN^2dxe$~Y%)UbVJL)K4`IV9pmyew$q?xV+ppiCoOT>h>Pv4 z)oC#?MF?7y?R){-xve<)MDv-p^QXJcEtj%yy#Rh5DY`Iu$2DA$q?^Scgbw{ z>mB&+J>s-YXIbk;9LrpR(v=$FN%4fc3}7`8uZAqR-aV zp4?#@xM6roKecC@e2wTJ*|QCP7Q4Sv`UoygV4HtSo=gI|WREZL@$=3FKAvw&enRe< z7ZyiqV_DFKo#&qQbAI~r5-s0pnUv%x){;kUY z={o;M_D(yf)6D;wNu9H(bNne|4}zojJRq$HE!Jl<*!XAWpPp}fm-f+KzYE;nUy|qU z--p|}kM;lG{{60x_Wu0>xA$5f?d`e1?QQ9!y~i$ad#n3sZ}|mouda{w?zq71&F-VU z;tSkfP9N=!y1?yS(MNlOE^vD%Po7)eKVe_0Rkxl^-ghef+5No7^EvP0UF!d;`?utI z_2;WU_w~`g7tgbQ=iA;heYE$33*6qiKH9tI0=MVuqrJr!xV?FOv^V1dw^z_ddzt6i z-h=Aw$gX_i#p*nZi=)%XRcje0=MFR;Ng*#hg}m?-^1_|ug}*`Wb;@dTrPY}&8PQz1 z_F7x=jgF2J8RUgaY?-`paw0OK33wM$iF28lFM^iq|9?31ZPx8XSLUHhm z}e=aHNAvIm5{uU-4Ktn|LYnvIoSyL`6nKYN3| zrjh$MikIw9v-A0E{b>8Bs~-cUAK-C0AFz%7+-HpY`&^^R9x>@7xk_r!8t?bh=DU{r zJI8_jVX7Y9WslwguIX#uzHJ$Y*DX9|j>m6%oY&rg+^PMLW7r3>$~M&oKieA9q9_03 zq~s!P%qYH+^L624zR%$MR%E)hi8XF@UvYA!JR5w`K#4PS%0UZ0h>XGeJK-V9l5?f5 zTR2OdK_cbF3kRz9P9dAcX~eh5AF;~GMLWLebsvJrPllzBawe-)e!RWd=N#jF)gec^ zKf6fk`qP%Lc0TBDhX_A)J)c$f>t}5ENAuI_EUErGi50w9f;_};uq6}tR*kX-`|il7 zQ{~%q?Hj)5_&Ls!is?Sr+yBcMK1RI`bB%<0+o%_rIbj>s&sZbrs<-Am>Xkhoab!`> zx+-Qbwyp7+=;2Gy!T6fQ&G;Vr;ZX;_xf-w5&(C2V!&A&Nv-)?Pp94K|j-R6EL;Wn~ zI6kAfmRoWj^t9urZb2t(W!%mq=a&nTAi#_l^ zdP?#hr+klfSfd`DpJy4eFiYhsNSoEP`7~{&vj4lu7JtQMOWh!AWa=|*$mb@bjghn= z^U=b*m@+7N5JwNwSLQ-n!D3Ji3bW1)A7Tx0W+>HgFx!#%kp(ih(J%$Ji@N z|I)kmJTBn=g^Is6+I}}#Kg(}hYA1i~qJO&HNYnR2-AEfv>|>U-hKt$bw@RHC_46^T zQPi?V@qumN5+`TK?O@Ge5Ni}_%>tb$XD>>AmZpm>9v#?{2>AJ|Cj z^hWeCIhUgIWbJ)B`wq7}t78o}_sV)S@{pALP>E>?1{gIcWy5OXMyrcc{^KR`e&k+R zI~~P&K7VyE=PFk{oB6wPmz>AAPxi{u&R-pzz0K!MmNQ((fG^~HBR9M>i?x(ltfiR! z+HS0+NRCRY^b6b49~-0kg>61tzvh|!Vy}s{Uw4`P;$GD+IhV)m7kg2@YQNT>qhFt4 zFQi{`Zc3^K-Pk-uH}>yrbdNCUHgKRhAJa^_xmVH6UYJh0**o)9>AvF}bjJ+3CFiKb zh~-PnuZ4X&PqEK_AJ1BA#ZJic&su7ctbcsW zy~n9<1TumBL?#BYRg|;3Cc}H!f>L-@us}O|n%Za4x9QY#3;LYjZ2TtsJzFc;dntC} zFB76$<+&XrMQ-$GeEd!d0*rg}wzg73Cew(&t41Giwi;dLfGB-NtV%qz<%VHZFMK@B&q&f z)TN7ETwiPG;;XMS_5gI9fi#Q$i2Z71{1nYix=v>8nA`|#5_(+rIwxZAp|p=JPSCd4 z=)W*7RvqU@enl5F;cL0r|CeAaC04EOiRou2s(o$PhmmyGbzNaJ0l&9UPb;z%N0wq? z&O{X1S#?ntUZD@IW*_c3vkzBU`XJ+S3RyMx6ev5)TwD7l<7QrG;@DQMk!!8j-S!Op z!*}wV^?MW7sQtBy1AcvJP(Fm$vA?q>UWf z_^R(;CkUhbq{=qejrYO-QJcKU6o3#g3lRzR{0q4 z;)rwb;1hyg-?9x3ZkM^||6|nHI z>5rZd|6bM$|2ES8SLWZCexBTK*%$toSoibf&uSjG^L|1>%#OKpaFR)7Tf*g0A0^Iy3YCd<1y;2q)v37&~gf# z(8(W_)VsjYF){u8xcGFgy^;PQlP$0E`|HKYFHd(gtFv~|jUB5l=Ea#IE1_?gtdoyd zG9tP@9#%FG9?#``{?zDODXeL}4E9j_)XD1-`}YF<{zr$QYcK24we&;wM>mQO+;(|% ztMY-n`tj;l>_-dZ`w5u1byl&`@5iCzkLdb8O$R^4z)$7ir(3{JYr#*@B!0}LU?jP3 zDER(_gYQx<-SAcF*v$HMJvdtK|1>3S!{(Hs#~=Rb!%uv`y8}OBm2K$pXZbzhuqA(; zl6Ihtm#<&ihELP_ANcg|s~-L2vWwRLIOQVF2Dxb0EU@5~(;e8OGskGL=z?h}QK#&M zu*`MnD~^#mL)qYoPJ0ADw7?h2?if5Feo1TE&|NqEh*%}>@w7{J`0{M^&%iC*}q zr>#1ho)G);%9H1>AFe?!D7@k@o`*3<&WrB;JdD;#a@-cABtIg*N%q#al=CbMaw|S2 z_v$z~VILhsM&Z>q_U0rO@{EiCvhe|O%2_!5MJA22BjnkdPB{}|3A~wrzo|=M3avD>Y1D(exLHg)jV3x zxwpwYE~8##Oz1l0uy@JqGI$yvK+TUsKWF~44~mmn`-!tHgrDDo-+R(gD}5Tg>jU`n zUnEBm{`=zxw&YLe6z_a-zJ2Gl485! zkF;a$pZF6;;f-UAmpbd3w$)iHw4XqJ`hyA0vkpq>`*wVdcKUvlzVDOViJallt;GDT z&%OAFyj`=tW9XC*-$y5?Ghqft^U~q9>Fg24UmlK6-TFR#Rec-2>z3gsn~6akb~RoV zZJTaSzQdjn#`mQUDc{E&2(PwqeJj^95A0c=7HuMDsy}CqE@p3t?CX+PYPW0VT^n_A z&6j(O9I=--Kk*DPnhfrjJtVT9gtO7kth-LH`%m1brJVfP8ErqQ*L#%TE&Eb7W{`6x zYXV!yLsM;#dnRWHwZ7B2r{!?Ti&sUrfqP6k|C2fzZ6|*w{gUtUd(bn2aW?OfdF@Yp zso)!@ny2%u%!i1%oA;6Pc7>cNBKAq1%OH7_LidP%+}}vfA>;E48HawhpMsOC*tbD$ ztn4G|cRV(M{Vq1!0sQa(;M|7;tG`^czWr^9)Aidx`x{GscbVE7aCDS9n~-O@$aw+9 zdp{q2s#(>q>1RRyiO=BwD@>Y>p^Iex2KFbWe6fl(Y^)^?Qp@_uv3GcmI@iu8*F(x< z(rn4WDf&51+7QZvDG!zNm*~%6c@Dt~r8`A;#%dzLM^ zkh1ju4fu8-`)`k3SNu-LCn?`KHf-|~+<%40#ah-PKAgo^#mJlCyJEE`HZQu4Tpl?~ zL&Zn=EiTt^{nOttd30E`?M8cY96Z`c{bn5!TRg`1az=C=-?tDSlJPKqm$Rqz@ghHy z@q<@RtmfGj+|AUDw=eT%*QdkeoYd)bmW&H4(1)OQL%m=_;O? zB0ko!ZB;uz+C881WN+PhY}-)Emr}lL=P~B#L+CvQ?Sk8nZL%qhBzUOjbB0=Wz-POQ zlS}JQHrFDv>U>XfKjo~47Rgs#X-@{MzmbY~#xlzul0$*3lJQiFU9SK%w zOB>vI7R(Pg6SZX(Z8S0VYtcm;iorqZNv7{-f`M{?^H)~@z-SE zuyKBnLE{|gsD?(Bb2$V$hN8P2vM-HiJ)oa@($vQN?To3M5qiJK-kTRi6VMkHngY-i zhNff1$;m^K&G+BF{z%D^7d}t@%dbCAZT|3c$34ZHTwmVtsO!td1N}HJWCrJjcpCd{ zTFiMOTX`+yOp)pdC4bMjy80WoOy@V!Y&B0MT-bHpI*KffClEs(yGYlu92_T0a9R6Y3!_MXob#BJ31bB2xU#2_n+;e+lrF7di8sjz7| z^c+A&7shpHvPSCeVTXF!tv1F+VqiyKH*nwq*6c2_eYPIj+7@9O zFN0?ok2ZL;{dul?f$Ls|XI^C-&@rF%hj+lTAFDX|#bBdLcir;Gj^_BOj+Z9O-u+7G zJ0v-Y`dpaL+cL-amCmsw-|%wAau~15V!;D14n%&CP4tUx*RgBCCWmO7{Al>>u)+I# zg!kdCJtdhsZ?(~mc@ENllY3~|T&c5|I#rAhyW37(?bu$vJ1oyY$PoN!!5#6ouZBBh zUH0eX-iV&OVGBBsIqQi{&KC#DS*B(Tu;7XF$Afa7spKxMWG}tV_-Fcy7ixWcdHn3_ z}rI%d}6CT70{-`XtwG1tdGOB<@zLw@@!Iq&hZZs(t!#;+1Pw}Z8Uh3LFm zc=2I!EX2mu65~47k`~p!SNiAh;y{VAYbnTE2G7TO8vJ1Hmj!oi+&<;-h(YavHh8=qcz+k=QeMV`yM^xz7rWf_!kUhx$p4yD~h5s zjr$&v`yTG)KKG625ZPy{?-7B<6WeEJM2~VtrL?zN?%UAIeXI5MM$#wOn-{D2543Vl z;ia7Udyl%eLGE4K%e{B#oA!CXyX)n~YW zb>xQVXkz3->sYyGRWJ9Pyf#|GJ!T)*NFOiCj=n|wM((?X`_w!vo2llm13Z>^j+_B7 zzkDxq8V65F*&Uv#zVob{ncSu@R@#B>U#@y|yX~Gw?&F-w-*4tj6KvcXbccs?FytQR zDrF~4U2&r1&FpB5w!5%>|6K=pCeA|m>u&gKRvJ2&b@di(*F0k6Qs1#=be@4lK73Tq zSCRL@=)VWiD<3_K{p9-o@V}JD@%^F{+h-s1Owo@wr%XML9{2IxC(CTf*~BfzamGwf z+OQFaOWvInb&2hOALOhtu^l_kLfcp?ZP21<*lbsAeS-d>otJ25axd*L2A4+@m+R+j zo9+C|S=t%eM>}^7v`zY$vDw8qq|?W{c$faC(|*sk(bb1ba)w16J;vXEmbOn2PwyOm zqmMF{(!Y;a6({FlgP!B;{nv|ze>N5@-m^W%bYsaU1EYx^?Nywmy?1+QPuelt6#c%K zy}UN|OZ~@yFBIQ8!|*429#UDaL>~9JZkwjgUy`$GG8Ti4B-W$+=(JrgdQW!Vqkd<< znbcV_sPV{%FE_6%W*yAdI3fBA>TY4*k;2XNHRBTe$29%yj5z%-24C74^AGP*xMq*S zWC?8Nv95g+=3WkSpGg{H?<$r(y{X9M*mKfGTW#$3 zknvnh|MWE#(QlVVZ}>5LMl3We`A>a)U7qPCK9ih7VxB!h{rVo&(>#l_!|?N8>t~!@ zF;ro0Id3NQC+H3^i1<+H1S)Y)p**)~?s>4&Y5_F{AX&HIek2WKCzRN>Xe0qAG;vIsuV$Lp|5;~bKn z<3;^v8?O&`pKAUyF@w`-It`~C#(R_g9!Fo8^BJ+uXAJt@24kap7Bk;Y`r)(w&K|6m zL;Q|hn7_NNzrVxp@14DWiXZk#Tv7JWab6+)+lFp-y~dh=^iS-d!4C;&PxRmi;RV?* zvxEK#4--=*UyyNnHAU8150^Z9NmTar7pt@NI5!ENP9>MInq0>Hu-PN`!7<+$s+#z4CfIkyRNSr$EmyFk36%{^1LLOcL#lwI9OscG;U)&cF?cg zl+iizTjF`>a$T(GVjRJ$19%o%zi8fxGYL z>uh|S#XL)_zHJ-W89RA&H+J!PbkGaT&9;l9Th&>+Jim6B&^11KWxS(V_T5ShAz_hU zM-QEVJa4q!lcmnGh+E|M@93GcdC~a3#!A;_=-PHpy3p6t6U%ISXQCBxeQMGPrf8*| zt;l7@KF$-Qo!)f&i_|H7gx6TRDCAzwK4ILeIln-@7k%7>onw7DDcIc18~F*j3glaK zdzO6rt0J>{9_#7YkZ8^DVtn$8qIF|0EU=vz#9z|}@Eo)L-JWAMmOMv^&up8* zeBrmpzbMvop$xmtc>?o{&-hpq>|1}5OX)`50a&u-%X=bC@wU3*q@P9kR~ zCEM^P+L-@R!QQWBMBC3Y{{_sy(AWq~Lg)R&KTp@mT8C|yiA}bZoKpF8aR;(J9i8^$(lMlvc<7&=ul=0~&_>l2gE#F-|f-?q`eC+9oeLmUAIbO)1 z;t7ZG-lV^)`Z#-0>YD++lKSwg_H4hS$QiZq_$#@G7JivG##a4;T6yZk*e{bcv zfZKR3;E%2=;kla*o(*_4&j!4PXK#L;XYF0PbE^hFBp82(!M9ENyU0t6%a$x=?wnvL zb>80+crm~W{*1$)A_up1{x0VO^V_m>>aC3>oUthSRL-jsyg;81uYMET%Xqa(jud*U z4ZYP4J?%PPu=KefIf0i(H__)-z+yRaV}+!{9wl<=ak( z_A`0!MtE;BW3}n*ytf$IY=6m#-Y1yEVpnS6i#mx8hPj3 zyhj^*Y#)+e*`tlQ=fiuPd!l61;FTG|d)Hxy&cb_h7&l~MkL{VFsGWOO98}M{P|rHX zHYQyPGwbL2xrqy}E@oajWWG7u7Hltie30}#4eYXrzBd^dqr;}2v)x9lPCM4oe0rPy zEc??5{k@|Wj8@6@3wiJ0eF5*1leH*?_}pVhQiuKM^Zj-EKIi#P$ri3X?b7d)`itRf zJAD1CTIHu~RDMc1*HxaK&%kLLU2+bK@EPMOe5T~GD&62S&K)%QjJD3?Gu=-~5m_B+ z(6XV6mQ7Y#Web6H3uvlnm68cW0_ktmy?+l*TY5Bvm_)O3+i}AQw^#8W>sM-&+ zPvNCKf|o{0z3^NcJgD@B?J^b18B9#KjF|2Zzu;Lp7954%)8m<^!LQ6SpZD0VHqMx} z@cV9b`ssvzzPjQc>TM-Pbh~vNrZWzXKE^@L{$L!M(3LVLjKg79Kf_nB`i``9wsClp zaX6RF0;_H;QF%b!uqx+z=vWoZ(y32O|6?t6TKn~UE4VX;et6rKj32Ni6NhZrXz;>& zd>(om&uqON-TCdEV(Y7i+7u>n591wMUpL0K^DWC5&t#0>61~hAA2Y}Jwo!UMwPlPe z7-Mz*&Ia)r&`B=iT>EYM`yXm7dUyh3ozDxs_dDzNllkt(v(Eu|j(sKL6`tG590;Df zPw?EE*D5?GzF6GAjjgob+JhHmyrwZ$Ev)ghuCgQl_GDr$SOS{cAGU*C*k=!a{zm#s zT<#caY`#RFJa@u2`-U@Y?BlW(^x|68~UZcO&=qaTK#?apz&fecNkq6GVP_lP9 zz9{{5$#>(2^dfs_=edp@ia$2mkS)>g3yD`)`DFp0`{YSE+XCJ<;{YYdTFJgq3a?vv zQk@^cILB$zJafz1|9hN5JT3Nx{#S!_N*=rH3_l^B7!-~7m~WGo z*P%tl<2eUM@)!rzU4PP5tUsGb)~xS(wxen~XHCZ0160G@54<#GLmhl@6r6uOx%Xhw z!xDp^^^cB%7s*eNGw%o1)tnqNko6%js7rF%ur2aDzQwA{bGpGW4U}bECGMa4*vaPp zoK<12N1X3|+24^@s^@p;`!3A2g!oY8ntK18y%#M5vE@UsOGEKRExB?w{(aTPhk z_GVn*_6quFFY^MocXc1_U37ul`;vUibIJRO=!KQ{mOk2h_X4-~W*_ana-QuCCWrYS zL$B`K8>H5~h;vKqKw?=yk3BZlBxDU>)u)^R){nhxBl(@Uh3tiI8v}m1R`QklldCjf zXUi-w%q{F$be(K2A=aVr@eN!%f!DQqK1>?e3{1$LTiVFSpBY2!k{F1>vL~*NI*AFy z)jlbG%^?fi*D{=XM}m!rbLbdQf0nO_qXa{3vh?{kXTeW{$4Gy>_6N&eT5#{iYS!!* zk<0YpXtnmbv6}UH6YIZW?F;>Fp+9nutjBZS#Rl< z^84{oM~^l7U-fb=G#oBzxm@ofc-UNLf1#Iaq2X}J6k-59Xn3~wYxk5a=ltXzYuZ2T z0Uwfh?XHYcOE&8Q@4l}`(*}2^x2BhCsTV);%sF1z%eB)WckI>gz&LtjthIa_I|nc!!!A8LNC{*%DidUMIFS#t@V!T&8?KNE6Q%*l(-Ac)=MNZ$Vw&|X2u*m7t(wFg|JhQmwv=lu<-06pzoi_ql$Tn{cU#KewUqC-lpnB^*IUZpvy>mPl>gOI ze!^0I%2IyXQvQ*pyvtJliKV>PQvR8x{0mF@SC;aAOZhEJ`QI$%gO>6UOZmSn<^Qpi z|6nP{E#;$@^2e6)pDg9iE#<#i%BL*le^|K?6_)a?mh!hP<$t!6t1aak zOZhHK*>5R_Eajz^^4*s5cP-`nE#(I+<@I{msyEE?*qp_gb7niVg?Ek}uN7)5?;TO- z9XmENCujdlRo@=%zGLAXcc!P`G475|F-?l?BBlsNBei|e|G=Q{k!(>-v8tM&+Y%o{^$4a+5gl1d-uPv|Hb`3 z+yCvSi#-@yAKMUnDE2S0 z@5LUDZHz@@kHj90Jr?_Z>|bNeu^+@9k3A9F6nirE!`M@?&9N=9t+A(L+hWhew#R-H z+Yx&=&_DW3R=28T(c2 z^|u~57(MvN!AB21cJTWL|Mg(=!5{8iw9pi__KpAAAIHD&kz3M;Hw8;JNV0kzdHE(p~ge^9lHNe z)1lRe)*O1^(Aq=m4n24%dgzfu|N84ke*NgLxBPnRuU|Rvz=0=6jF^&hqgFUt8()`q zXW{6j7cH-mGL$J>4s`CcJu;9{gL%q{q>v9HN z4SZ1_ZV37Uci+A2E?=lVZjwY zUy47lG5$a_{=g&g2Of<-@L2qTm*Q(T#@9yUYafZPeKfxIvG}^i_`3V!>zd;0*2LGX zjjvl5U-w{q-G=zOhvVy_@pX^H*EPr2JrQ5`!}z+V;_J4=*X@X}dp5pqXMEkR_`2Qk zbuY%(y%v9PP5i-i@dvlWAKV&$@R|67+vDr+kFRfvuU{Qs|3G~G+W7kK$JaN<*Kdz+ zSQCF}UHqX3;}5NmKeQqK(9ZZnyW-93pEDF{;1%NK=2gebXXiU!Azp4?b-cV`ub%;P^9s5G{#w|f zn*Tk}%NwFg;h?8J9K3s(|1M8MC~$Ki9J$^Xx$*ki$c=T8 zNH8~NO3n;>m3>~>+*$L=E2~Oj$(-e*m%E)24`kNXgaY+-+_TgfS}}v~YVY<2M=!rP zeOAElX$X5l;W}q7ue^x2-ZPVHy{`J8x3*q!X)q9k<03(~*Xh$P9-LkcFEseQ%a{AJ zvZ}q2@LXn~YDLiF4Eu9yf^|Xa$%j4jYFzbB_%9Tx54b(PX(-j3oD9wim?rF5>hv{u zYJ36aSj!q6E)9hOp=BX&#N(>ta+fa<#wbRXxlyyOIuxzfS0C_4>ShPv#?j&Fqr*38 z@M~#BMR|obIQ`-gFm3K6{^aNJXHtR4!IZGi>+*=^)N)1fPM!ku&I%zJZchzeUgq%x zBc5=iLjEl9Ecdt;)Cu29sUqyE^SF`IK&=c(s5b0!`s?dFPL#RRodHw=RX*BOy(=9wqOk&-LN)T8*>87tvs# zA}taK7A@os^nFdRP`I=nCTWP&Q1Ey?%U!-;puz9wp2cr$)9rhBV7p3mr~WuzuQk_4Upf9$&;+RR=rWf>Co;!XM7^dQYv>DVNmBCCoDlM=n_+mxSe#fLtP` zrH)yw_so+&0d$dvzA;1f=uXDe?Qx0Rc$Ry^bS@ZK;dfRe1ZWjMBhe7%3aImAGo9fT z^+C^f8oWybL0?1I=khQH^2g;vj$qRIaJh^@K)Mz2`rMdxXSl9j{`f>mgyc`Snm<0b ze3U=c@`vHhnsPlB^~UQ*<`>AHsY(LRekSjmW6>RXjBffBiEggo_ur5>=5+qnWb#9x3Af5AKr;fPER668UZxo4GEl$Tf89gfm@ zRTZV*EOk_Aw~C;PIOa~0zxjFccTxc@HlU&rw`p0^vgAE~Y7Ro`2-G9!;T2&Kcagj> zvg&tHzYl81Yj6Y@i6JS{?g~_gi^RY+gpoL}TT>FPX(a-{M{5nty(aVJ(rOTT4HM!7 zTm?g(rA3+#HY*yhdBqI5P)|jg)9nrk+!A9`pLG*|gIPD#7gaiLE}g+O4WZg1%@tW5 zDb(mNcZpLTPRpiXp@4~q~8 zh16fB%k2rHazl}DC7KGZP=6zJUW}pJSzGG~%}{?4uzIi07ZybrQh$}wzsVsMaiJ6S zx74F7idzZjM2CH@qjdJHa{B`N?epw2W|Yn-fkv0+2n3dR;bZv=;)34-A#W|Mc|uD) zp;GnDypR)}rs~l!MoTafD@0>q^d{!!x$4~$YB8x1Fy3u>6DQ_Q9ldOHcFr=p;)F_7 z571y47Bo=rMC1ZNdLQxlxFzI?G=%)w@GCzz(4eJX(U|e4c@>t(k_3|mEe0I zmM3KQ!}X!r9)E454oX}RUnP=PTIE@agw+c}f;)8~*XD=OwT0=*>gMOEw@C#!H(_KM zrv~h@0tY65KJfan&tA8PDt60(bk6XGbWy&^5%4n_k=&^h1q@D|JV|UT6YBEdM2NW6 zFF?rwZUh|isiMKZ#2;9Oi=j+sS-G~%=|%eee)=yYbD4a@Cu;FfHyvFT2rZdE_m;Wk z3+5K`6@YHMl5v2bNLjj;$$fZBb4mok1~JKC!_k^!-h3@nyE{-{?e&b;?#e;DWtezV z7P}DDQBRY|zi8?r&AozFk$oXFJga<0Y03Oq*fl>u2d0%b(FL_Ra=$xQAgj}z$D8P6 z)C49PZ)~u_GFWyYY3$$#P$*3FK;D7COtoiuDdx`?@RuvA?VqoI^5|~@wqS0tkX+*Q zd1^3lj6#^XpC3f9+@7kC6Zs73+*@~ZseMLiMP-G67*9y6@i}XyM;?FCXt$P$8upLX zMlUTIy>vWG>gBB2wTGCJ>ZkVvTe4GjJR8m0Yc%jTU-y?-X|IFla2Q&#`{#`Jy)+Vl&e=5 z%GE0jdQ0g%QNcBGwRDT>dQ0g z%QNcBGwRFJ>+^+lI>SNzJ?PTkb&o*P((%Xf$kSzQjm zqV!t`iM0AvZf$q2Ijh9QDp;e;~`{gu}2E==BDV zZvW640e=lnSp>9?W<=Xn!_gRMXK1;1=_0SYHZVbT01Gfd>gIEG17L&C2t_ytC9-4% z^0Rzm&Qx((9SvCKvKg7U1zw;6nl>1~oQ5ipFHtiKiLcyQI58igQFJV-MTISMqV<)1 zQ=rci4r?LjvMkgaVH3YwAi0#qWfnIV4O~yZ73gQ&=2ev6cDp}-ZV4=3fy%i7C6vo6 z-%wAyEB%yjD5!kVP;!mvCSZD&l!bomRd~kSN<^jJqhPyz?(N?a0LZxLx(G$1)K+e; z@jh9^F@U90YN@8KUO5-x?0UDvSCqj)^XmSSQVk{O5Y5o=TIQUx%1UN=?C6Zl(%Y&k z>>5!`zq0;>F1d?=m5KFGFbIYXo^00+Z{5KF{(=1xhp2pi#_GRFNVlK5{0i z$P<#E@(K!a3k1(AGUjtjnIEJBZEmSORLMY#vT_KQ2v#VehT(-~FOg#LalkUlk;ydl z_EnK(6Z(F|9D_1l882KAqt@g%XjB8Pql$a8vwY7pSu!A zt6UI07(tsSA~G`&s>a{(_~+6;{}iTKSezqPcYWb6dDXX%7+1XdQI z$dQtreGBrb6%}c@d3iaLa|Db}%FW52 zsA648>}E=M4Yu;uteA~MA!1f0&P+rJz-%PeBaA|H9;ksEK9i-O=oIa#L7ao|0 zqS&a+u?nCEXYn&};W?#qW|z$=tI}@qmLLzykV%CnBoeH9_qCwh&XP7>Qxiyi5tkrLN5w%g6c49wiCuRjNu5XvQ-l)>#0rPompIvm#YWWDSAiOG7GI@~Jg-lBQJ?l&J&4 z;esj%3|}!3p)5Q7jZ{=tEfS~BKBsi<45L)(iDc>SDm{WR>cRe?$r^&e(LxI_}AyH%zvWwGohgKrH2rbbTL6>s_LXHFgMu+biO)!c--tb(H z=quJ2fcA(g&Z!vcJT>vI(28IL6YdK$h`HK3NE0M%)HXEc>n z=AJ|I5Kl=~CMKvznC=!!kuerx1HnkyJkx&DbK^TP*b1w{Fo1?x0XRZHiEg%qQr&P@ zl+7!hGb1Z^QV!-A!4cyuMtn+vQNn0j3^rG&7@IQQ!nlk40fJIW%IpM=2t8hw#;t&`rG(%L*yoypsr=CIMnJR}# zXz;^0m`agr)DLSbARmcqEiz~jYXwsaf9n`R*?0pGWW$2_ITLdy>v45mMJb<+2qfAD zDAV9!3dG8bw`bf|UQ-jU^#!V(KD&wzh}M}Su?HVY2_>iW3Bj6bz=()ttxGv7Vo;U7 zaT8ni`Wj?8#O1B=V%c$JMBRw#2xrS;dOc=Q6F{VlF3V1OWChdQDSNtdrUN7Q!fo)LGE@-cH8Q~CWpz{DGTmVhMCYBt~fA#bicMatvG{_Ic;<~g` zZdcI^@}Q~myOdVmta8wwqqKze%dw(c5%JK98??+W6*gWA`Wgs*6?TnTEinz^l-Mak zqy(TO7=Zt#FpUIHky^z-DumBEU|{A%{R5f*t_B2uQCs zT*$nKR%AJA&^W>!=sk(su!<^kE5=HN;I*5v6=5+O6|5?Nzoj@!_Y=5|MF{2U2%9k* z0tSqtEE`Es9pySfc-I%s&Cy9W^q`=8^8(h}WbMMoN)O}0wSaj1j7TOk$+w6j5j4#s z{9jXphorx04MDd(7{p~#G?e+XDm+2oima+Yg^Ctt%>x6mc3AI4=D;GN>MFZ*U07Fk zC^m6ICK&M(Rh&4fWVZd5(!8*^D|)foT0|3?-*UaF-T;Qd@Dd!gE>LuHX=PdY4AZ!H zJz;pkRCXX8RM3L58C5rrB{nujj9gGwrPX4@B1Stz#)9%MvK;{P+=+P;r%axr76K+u z%A1-$wLk?Kb>#vUH<#@eV8Kz3N?BB@^@^|v8!D+1?J5hUd=L$RGb6E&Dq?X}&gxoM z3G0=1!tu_qTFf)9sb!rTW2P=8+yq($X8|Wz&|2z-Yy4naHx>)7l6YcJ`4=E}@iBch zaDi({b`Z2JGvlI`YzFeH1MO{B!DRh@~EGHpc{xBLdvyEBtD`N5^G`&laeuk<0)VS8F6*$) z0&UYoRn<2vuMOnp<(fD~SWXm{NF$DSDFIVLEc5MEW#w~q6J7obNYe^KPQ9 zW9sQk1P%_6rxejjmxJ!`+AuuJS5T?ILQx@T85LVL!=Bl=q^xN}VHhll>_}U({z=3NP0r5Yzg$8+th5l0V37hugl}aVuje8O7VTeVO3__z~1Pn45j0ichC-G!wiINT(s)OyB^vl{_7R|lj*CpBI}i9v%WRItg2AMb(9acl+B)9s4YP%`A}I^K2O3N0TLgS^wz^fp%p4Z z?Zuabvvlxf#J3aXgX?=<4u;^doJPAw+s&=l3IbaS@k>EO^A zA9VeMa+Yoy+F8}>aW9ZqkUP-8K;ixgt`z-=5_idhd=cwZM3t1E9WDY+c1?+{z7!17 zGCTDb?9+us!=CS;)s+&|mf;Fej+l^vxRD@7GcF{GR2WV{M!bJ_r}|uuS?>Das;lk% zP*1=oM>k2@LWQEGdopzQrVmyCv}k3}Rml3A1-a_zP*gw>u&`6k2xBXLXS|7?cG;yM z0VP(0t7gv#RyME@I32V|_8B_1x*@#6yj>Iw^Qt1E1|HOxZ3Sjd1McZkHNR?R))ZYK zFLKN-uPmKmEJo=u?FB?x$`>q}KabVlQoHW--mT!13O3?4gLVVH*?6K^rjpd-(^(ag z*by$P$(oB?WzCUD_rjS1d2c54f)MM=G-^K*#UwuGYEjo!VntWM~9acjk2u!8; zoF4bQTO5_S;)F`*TIoOYgJX7C>D;PCED6mhpTlB5P(&;Ti~CAF2^z+bh9%onXB3IWjg}a)q#Mu4vk+GS)utAFku_FE z?)HQNMpBEuu5IWa-SB14(-+Q6UZv`Yb9wI6oXPnGQzqrB#XJ_yCrvKM&z+wdgLnT6kAp2-6)>pQP(RZ@)WRAT?^&UQ&TVc2;8CVGRirZmo+gxl=2xxIx&d<~VBdLP4GB?QDsCC$UNooVMtvn-daKkqT0#W{ z<~cyDVn=nolAfjOI|KE}I4~$iz>XDu(TJjgB`hw`-TbLYth`X6MV1L$lHTFzRD%-3 z6rGMSB5)xlqpSDwNtC&6Tc8Q`E>SRC>SYq`LvN`_3G2b#Y%jVDVQ;`D$aU6qkOimT+8NGl~&BP8=opl zXU<3cn$gLMibY1w#-h<~LsueFM5M^YRk|#T_yvh7!pVzNEMbu+0ITRgLB~j@F;GNz zS{7NqDAWYA>Z)20i4m?x-x0%HRuA-TeJIC;jeYm?4 za5L445tW-}epa5uO$C!DP062}J8?>$S`ASATW^J4nUZIw2O`I+P@VE-HF0f~66|ul zRN$OQl(Fch)WjTCnwMvz8YOK&kY@I*fWilQJPQ48Bpc|z>eGn=6~&-Jak}fJV;Ix| zgI=31KCJj@!i!F=mJ)iS*3TzG#b!e)=Z@G}r_6XIJl-WV7%}-(58rUBL|K*7jm(nQ zr3BXWQw<~~vuX%3QkkRckUAFJ1V#sy+Nq~m0v2Hn*sGQiwaR%5SS>GA(H1VNS5|*| zJ+Yix(z%w?A>wzPN=EP1oushS7QqVzqhKnuqhJc0jbK+7^sgeUAIk7$&5JFjQBAr<hx zRs%VD@bE9+|I@eMZT`=%=70U|hew(+w-@~n46W>{OPm%~ddvSxUa^-~j*uz;Ohz|D zU9$E=8M2Ex;v00#yo>&6-XwV#{qSp-ADau)Yk2YErCD0BqRNz;X=Ik5Q8wGC$)joF z8n?}v*OoKl(x!RaoH4J0b4;MxTUcCLURi}f>>nK~iWc?TaiO!OjL#0ez+ZcF8^3+C z84+ZPCl+~Q3+pRu8w)+lqBhri{ABCt_On2i#rK23-ZHOTw*X;fxa;KV_QC|7Cnp#= zi4&yEGPKoTA=fe}x#<}Xk59p0uU@~}HXFgUt`bdHmUGnkb^i1idFko#Cb=P1xp*ww z@2FBoW}PFvNCMic8^{?x zt@Ytv))Fzi5>*IkWGN?uSG+UCQ?9RBS=(<#m-XKv(o6!+E%xS2&b8zf;RA0jtrP8s2kC*u6=@@ZFzFKD7TL9n4vC?fGgJ^3-n8(71-pS3NkEw&lKnTw0I;>^xqE+v zGJz@$unV)Bshq<-Vb44V#FdG&?mui%Crs-c3y73CJrtr&CyxnxU|%Xy3)+&-)ZN|-x2P7{P=NBVJ)|A zwhc$~=}u8?KbXIdayXRkq-?0-ldmCpEak{CiwxoF&%3 zv-S2;fPA+vsoG6(4#=cvnE+J`?Bt@IkqLSYX{h;Wuj3*Vn}VkIannQZLUxbQu+j!Y<}^c0VjG=J%u`F-=oiX#=-PX#+ zYHxXUd3kMZ#S2Aptf)Y9blHSfOk25_$LHrRz8LL0D~9=vKv+Q3+ z{8d_|iKSYDCqubE8qGz%NHO@{E}6RI8+$Bv$y>enltH7qG6am3qcN*lb zv|x4Qz%snKBxTc%4_<MQjJ+hdl#!Z9QyL%UFDo@4Kr9W+0Jnihpjrst_&8kxr28%P+f`0U=v(x&ev=}m zQRk6;0#HY9iRF>=-vj{OwZ5eYu=}!6Em6;2eR9CfH;DuRb;@{1ZT~q4AI%w)M9ut} zdEr$O`N=om_?YcHC9S34Z$=i7(DI0MLd3mo*SAf?+5Y4JS5&;vzv4U14J~f0_ZF8H zi3`?Nz22!mpahsW+RO|8b$ug&fM{0A!Gnx7AAxj0^f^~OyBydj3gq>Xc*oIc|N2Pi zA3Pv=HS@3Kd-g=LRN;bX8xbXJgoV%4joDF_(i%#$Ve z-Bms?PNwBgT!R=nHe+iPY+PZXk+Hg8$DEPPzoj*0j?Fn6JUV4@!+~nMk(?UG zEb#7b$w<9lZte?>vC7l*nrEP_Gjo6dp*h zsiY$`Nrc7=%lOxqFYHD)O{N9HA&z**3|74)w$myu+`C;3hu6 zFl)#I5@1x1F-!tELLG?Y=a+NRKg_ulY#wrS>=Qzd15r@akIQE`Z(bZs`Q)V)x;cl8 ze(ocqVTl(XGxtK830tio!xL_91DS7_J&{FeRTxCLE<)zBO94LVlZXz_K_g`p!vpZQ zYq*(Q$h)-OP30yic!a9#1FJgMP9to!fa79>!y|yPkR#7YX$JXnsRGWfj=LjqkeXkT zFBy-SM)_>c zw2k)8=oPj9t?1&JH1E#)ShESEpP9vav*x8Rvt;DtqV!SoV(mrG(`aD+{Oo9+)rLH8 z{@`EdzhX2ksQPO2D=%#Ss{Y8U+`ls8&CIc;C)AjbLT%sI6BW9(KYFHIZu_Gh-rc8# z<&ECL!YUyUbd;CZt}UAz zQ&>pzK>BY}zgtMtX!TP=TakdqoYUz5+l%peAY9P?=%GH!tlTYV@YE!@-y{$>#{^0v zEXO)i+D&w0cG?1*aBWU?s8C=?;q>vmLX>$a>^99->m&UI=|x3n`ycB#1f`!yv`OW^U!Q*L-wr2{hs-LlMAKWC@uK z)qNR(OffM=ka~mb;xELb%66!fNR=9>& zI_=t7=58apyOP{YK^;UdsmJRxiZ&!y5kCOjW|g=mMH9kufh1tj5ne(AASvXl1B}AX ztDOTRV22@6QRuewigNCMK^REBb$g3uPX~W#nRa zP)Yf1v&B;TBS;}$kxKQ(Rk@j*%4L1}B z&WNf{jlK7e@`{EfYSi$UHUz+iEQwdXD1u&-$T&Q}+MNnSdba%Xrb_nJ*9wqiVhLVg zPR>hI*O8Eq3=)1DSnv!%0WvsA%q;TTB;9vOt->1>UaBg_g10H3b^J4X7V}+Z>j`Ip zw8Qf&maZ`l*3^SW1dgpVQN$U^l8EPHtxziO*4qGs36kbhi&-Ov&J@Vd`6@8OnCl-Y z>TG|uKcsB-W@&A4WpSfNEV{h9u;FF3c^94p#o~}>umTS~kWqpjfZ%6cZ&_giVgH+_ zD%5*%c}md-0qH|2pLka|Q4{m%&xa;$wCfRuFj{crVmlwd*n6qs1v+18;U2#e#@N

&@_jIZhK$RAH==h7gLv zYO*y_=oAovDL8)DKn4RAH%bvVpBb`{qaQV&k;flWwuXiE&!|qb2X{N~K*d0o(1mp% zvbb}`>IWDVD}5p#ggJavWz)=VsH>HpBSNktGWx@t?t9BJwLd#jjj&I*(3A?iq^HFH zFi{1ppf|nm3N$b)J`~n8l>x;#^vTTG>17VQl2>jSoc%skvPp`-ZQ*UdeDmU9=f~H(RTvts z^XyWqPEU@*w!j9%EL%&K@fuG?(UY&fzx@5L1xV48Mcq%(6*VN9BF1xzhBUMMsB}$1 zoW9smO(o8k*%w~O?@e)mdtcUw=;hHNM(J=>=N;MvKEzku7)z>`ZAumQ(eSjLo@mKX zHetCBO?90nEZ7(`-QgEIxckllCjJFj{*9@PiT5oI+hzjX-0dJ@lp(4F&BUTx+TH&No6O04w{kIN*yM_G{M2{=t8t|;w8Deo^Wj@a?=-f~U1J~v)~upL zF%lOv{kYj85iH7q>K^fJ@*n<=Os6X{oM9aIIHRj~XSD+up#UE;g8ALTisI%luS=Vq z{acK^<@5I#Pg|3AziP8J&-$;sKxy=;m}oXEAcbm6!RRRET;Pg4-=N%$Wdk z=gG#xRFmp!mUMnaeW@^>mcX6PVk()O-W53ZR^1alvgP$Q{A(`SpT*vy@U*gVEi9*1 zu^~)wFwiZQrs!ll4Q#f&gp(GSRxkgoCW_r6H%QH5EBwY|O}eqr#7>+kPI_T|b#Zlh zg_w77skgD>b?h5U3y`iWYxujB<+XJ$X)ict^!Dro<9XeMsAQw?CzgTda7vSYv6ztqdIeBIgloq=R%7NRMQ z7x!kK$Ro3{51pyBgrd3dT5x?1%Jd3W{KWt#_6_44+w~YDB*SZp*s*Y(~PV z=+}ru(-={R4!?)bsFpb?Vto8D?J@gs_<5C1lWU|6h2Jd~D$ia%tmSOH0wXt-i6$BM zqXiXRv8nsI>nLY6n8`x0yeVr*e#mmdOnzS;0ImqXe8~{qpTa{8JS!N7gxX}>WPImH zC*FC0=;O}~!)m>Hw(lP*F=UmtYU|QY%r$uUC0J)pBIfgwH&W>Ji9;=VmamIp++nZ^?SeRJUtG3c<` zXNVH2hTRpK8yefrab_z+!{S%HOTSINhqo6YOajTYg6cxhk?n3}}|l zaxsbm#ZU907at(SP2)NHVdk|p&qErNjz(rPxYDwikF|P< zr8{yOp6>HVXLjUSt`iS_ENiMaPecL`CO*Pl&{C6Gs}d0pbaJ)(XS2T>w<#}BqLnU= zJ`A&ojNM;U*|6$UWzDLmb*C`J3B2rI8Id_*@Qs{h;E3Aj1MKcAklQkqXREX|SzGF@ ztu1Xtx88&S-fJ>tEoPB-K|&+zR}vwNU;!LI?D}Y5x_l3845$h&i_~~vW&`TPKW!8v zTF|%ZG4v5GRX503f^JN-r|@R~U{^%@-O(kbnH*Fc1> zXY&IKmDJmq%3bjsyahGK!~>a_85m^brYXvb5s&n)1Njm}MKxR#9#tWeD@jJ>3XKP? z9HYW6#+FD-8NBir9q#QeZPW!T517sV)cKg@y6>$0qlWsV5R!TqY6=<2{y4X*j+!?5 zocx|r{F_6QXXvFYCi|n`g%I@@?!$v3Ekc)EwNXR}z^5q|?!>$zD9hFy0~)EWAZ>Z8 z?`G^N{nhi*#T0^|7Ri;++D$DaP%a$vbM`jJWQ3~Krddpi2Fbfa%_YE&Otv8rIURpP zghJ6U2_s%8wbrM_w5Rc};xb%SvU}i(ir|Z1vj$e;QMh9kLxI)vCP<9DpcN&Dz{YP^ z5;G?xnuXy>F>9=OC<)5qp?U_%POV%W6u7D1EUJiT#nT~oW-_#ns(G?OPxGnfL*@0tU)3CQ1eqpk zp2cqmYO1)cEoG1n)&>> zeevN8uVguSb+{7fM0D()eq!bN(ghZHsD?E%WL-UpVk5<)`n=@$TR!{yiGUG zVwracNU;*KLhcj8sk!dt*V#|MIYh9OT#XUJf5^6oIU%N(%R)?!%M4>*(V~;e9f|xe zfC>A5lk5`*6Oo@B4awlHr4`u}475_1T607r;P}Q!F|8Pr2=a74OuDXlyJ|^=uE>_$ zB;vAJ<&S_w?<3reGd}_lRUwJ+fF4@+o2$XG_Rk`7wuA+W)<;nse|SpJUWzoxV#A#o z(HsQw8QLQZ(fgpgtMxZ&U!>p#Ridi~EmUM2{b3dy_5#6&^~ zbTm-(T+33kIts2n{#kYvVEOobblG_ol>)<2t_I$_<1S4cHs(mC6Bd`3h_SF>b6fA< zWkq_2m?47?UwxY()cTaifKNdcHixN?k{_XW9(wHbrdHmQRqb^Z8n3u2hR}Mk`^3gX z(*gw6mv7p`2rYxz6D&T_UMcyjfx`h`57BX8L>I5Vb z?*l4@@>1-6GO$iztx$M8l=Ti;eCl1a<`KuBK$^RusWok_iKRE~zl-OL?{CKcqGFPq zN$>nZ<)kT}#Wt|zQE7}2KI-|FSLCMe<0+$>t$4UpGqFrC!;xzd<{I>Tm0|MeQ?r@QfVm{#ZrImpE+UiJ-^ZL1dTsYgd45>+j_vqEm?APMa(ueALtD*< zvm?=_L=v5C%f5M9PheDma+?GG(dek@LFVk}?Z>XCmip&ilI$PWp)uGw=D;A#G|*vqts z9Dy)gR0RqV&C}?+R0raZ^=id*n?pXKqXaf_rX6(!t`yU10RrX5*62W4&=ifi<$KeJ zelx5}_-4SZ`NaG{RW71X%GvFiiG{S#M@EJ_Su-PH$Z033kbG^lbg_X_T@}`Mx zHhVqNf3)9$%BKsYbYbF;H-DWS{r1f_PeZy&qqow>k`jsFv0!%8A}UB`9`IvxV5DZm zj4F|{4#ab8cDy>wB^+{i?FWRzPVRS+8tNq?HkD&9Es@q+C8ejWEE>{PAR0&qg-YT0 ztu<8?s|tm1g=|E+IoBj~K2ng%pr|_B!eVaRjCvTb|NIYG3AP91zkxaA6J$xI21k9h zeykJHelzr$vfprNQ|Jt>gL3OqM9BToWE$#xxr zT4v9`O?n4|mvE5vcFLlO*wHm?&(*}HvCt?Im9O#%XWKnkKJd!?7^nFz9;eu7HbpZV zOt#QgMjV^W_65ml4%IV$fJhQecz=sL+6Ue@2P*sOed=<-C-iei0rtz(puG=&EJ%n z9vu~rrrYJ)?209bqye#OX-hs#K&q7#baqb$Rx7H?h*~_LAj@Sq5F?0c7g}CB8=ec>b1#)@rVNx zU|S;QWbhSp{rLX>>&_yHqO)VZRm zBGriLQjOwuI+^Ji?B@4u0F_KcC`E1Enk|O8P+u4Av`OK-D}k&kE8n@}_Rvjn>1(B5 zd7ZQA8+|OrXKqswtbIhNwr&(ZyxV{L@SK#2bbrEM$nW%$B#)zh<(a3A@Gxdn=&={f z=!Y$NK7NSZuH@Xjf=MQCtWso4J2MbFYG;+U$4y9K<(*h3oS!t(^aZkN^$xodGbODU zbMEqF?vAR`6zv(zZ|16}m7~6tIa$4ZrgzG!c~pUPJtZN4cR8>rWf&!2!Q&@8po%E8 zwc3n@Wo?E>e`i?S)m0Qd>%CR)f48={wu0_wb%D++OUrC1*IQch{#7Q?PZNznSBGdw zH)cByvyuFmH^lRY4}m(%3RsuEvJ^}dq%YZ;$h*w=fdxm`A*mn9OBTJQ{s!NKwHG8N zelmhprSppYY=lSit|~nxG*o{8c6yFROLE{g6|CfhD@AORvyMeNym}$$q*nZLxd8w` z9cmuRpbx7O$6b7&i#9mBbAndsowPgU=h)b8sxHXUfbfm@Di9)VwLJ5VMaUtbEYUPL z%p55|uaMRgMALVEcoSCA^N#eaQs`}?@zDLarh)SKzH5RAJ9cyDh^Wgl5s`A-n4nxKhq)r34MnQ;Sd7amG0O2fcj+zU! zdEoO)UeITiEWLj(BR~tUH)n@{AB!r(X;sdb>VJ|| z>3yw5&_|E`(>sp&%*xHi)pGbMkJhkZMw4pVLR2ux|TM+u^|>tc+`N+nf<6YS4W}ZXWU|2a#Qjv@1mvq z&}A_EjJZ5u2>H!{#XByp{!W!I%lP!d z2uTg=U>;C+QgM#A;A80qfEUaM5-n);VcY zvesr|rQ)oN`5i(9WLrNmxjr-rPKDf2nJvtLefgs zc5P#AeZ98^Te-HdxUsSTaTz->osy~{h%%wl3v@*cu4|H%6I7?`g!)yDI*0y>t)nFB z>K1ECL|;gTCS_<|s79~P&BMtOn|6fxQtWw{;{W^)?@C=q$qH%?F5^jF;uf&G!BAB0 zjqry+XtVgY`OpYiW9#8)-WoxLLTKzzRd3O}FA;d=`f$|O*MO}mcPKii&L0{UW|_Nx zF`>Kn+UopjE(i}Ph=vdh!bF-cSji1_!WEsKHn&j+?bZjk`K{`ZM8Y%-;tq9T=gtEsTz|B#43Wr;>^H^Q zUdY3*z2lJ6G_7dZ#!yq#AUD$t%BG&8)gg-UT8ESiz2{s0;%&bB+iav&#!)yml*w_g zf@G^JuxUwGda*15D-Q^q3}A(%917zvc6m7X^qwG3ypo1h(YI!*6{1tUnaDu6g^upV zI9l<9JcyZQt|Rq;Hr$H*W*R*@tq}WjjJb$^eb4I-caBxJi?EFuUD4c!HNRC*h>7>b z?6I9S5SD1(0bIFUr{~C_$)HxyAwUH+)=X^(7q-PE7B@zAOjM;c^>L1PUEE`N96Wk`G{V zk%QdIT$%ra?3n+9d|2&SJ(E?|my_RRKRu5K_DBw~*eHv=Ct<4h#hf5=Xpz@q_vw_) zgz;Yt`ZhBo^=OFPU|z|g$m@-xf(j+UdB9Q#QbRSe=V$9$C5HO#)|NeiqRWbGNiFc;3L9E|(OU#aaZk!*m5Cf9SUO>i^H)(eKeaQbb#{|hWb*3Bx zKoLG6FIG-a2GR?#60p3)9PdMYa50%7S(WqoCJd2I!j-h1qA z^fs0^me+a=n=8|8`8q`e{l}I+r zpd)#WhG=Nl8U+G!*K}nGB!w4Ipk>c(4n}(pf5bT)Q*g(>tP;WqCA+WIF)#|&PXqH@ zbpNxnPwX&jcOM@c91yXE(%LkV4HZ#k85C6{OC-^wtR#7Pn`>#yV(hrzrrm29Pw@;@ zAAeyke$UD`82z)+KIXuts`b_${pv2%$v~qT%A9bPDrpXzO6~}?i~*F)Fc%Ibh|LUT zAO>%Oq=klgv>gmdS>-c?8+LM373;v4_+A`pq~-d(sqR0(%hY2I0Y*GOHpk#+a%llY z&Ia*&#!DMjP$1+i=%Apo0uBf`jD?_BLA+jP|Mk{SNm(Z)FhO!-Y2=HCgQjPMlQ~;K zi@&gP;7JSv$rK_o-KX4j$EXyXO`LFZwmpE_W*7QtR7vxdm-GWprMnC>1LO{&QSvJj5JZXz zaxgXY;sj=~1wy6x&bcJRCm&pco;=3O`$RiZXMY_c2cXF_*tMnN?H%mov8P`;mc(&T zC&ln6@4^i6?9CLE6!kT;)QbCZv>j%60ySXEH$f=@R zb?0DuW0<{dGB?Z1$h7y7`OkX}j;`2Do0L~e$^gC&2ad=xHww4)#XuR^b^qETaV5FEm$np&YaMlaT}VJb0oKBw3T zYu75?>xMki1>2()-x?u-IUs-9zx+@RYR$=Z)kobIe*Nfc`!z?OGX2)ru@rQ%wYI7j zLoegXiSYI{D$&bEeoJ=7FKcNqURz4|6xKovy}Z2mw$oee{4=Dlq)nX^HFaL@AH=~q z4$e&+Y*Xb{^r4bX1{}f2B-!yrCWNm}DKKNe=t&hK!O<8R3ZFmccxZAEFb~Q>s8h%* zR#fz=0$6JUmL0SMp~Gga;z)IvSE7#l`qo}((yCrroo67<7bXkYv&JvVJCUy@!FZ*i zMcbk2rKms%ukkBGQShu6kBq@6qZwHiZPzFOOi!_gg=^2)pw>a7(rkMdqz<^b3T7K% zDX2r9E^ml_nPu6SjRgVIA*Alsl!Y9kUzT;xEPz27vO2U78xB0w>ZrEl@J$ zo3!$O4Bx{lgsnF1vZ>-k-N#7H#6Sf zmvbcvrS@#Pt0H8!t>`cfNAq2aS)=4WzbgP>S`H6TY=xc}x5Rq+a4r*xQ(@vvk!-Rg zb>C`pZXm7QS39*B6hm4IUejZ+G0nFc&EK3CvKYf4SF>}D)@i|if|R;;zUVq(wTa0v z9izOuSKRBc5Sb=R<cFg9Fc#} zZot$@B=Mte%4Mv?_a!Z6{!W%Pr+yyr0mYFzfHb#pANx~p5)QueJyr{*D!)4o;=xNQ z5!|f^);ixsyjInJdtWnI2`{}1O=^P#xm^vae_ouK?-yZfTHp1fI3`+c`Rq#8g2fQw zcqNG`dR7H&UR-b3Nh&ghXKJln##-b2S&&E^MU;vJSxsQ9%-d60*wpWes0CtEw9x1f z=zgd<5Sq`-bO#5_48XN(otZ?yo3hcM=QF7-tQA-Gs=C7p(*^jl@_RH$h)?Lpp<&9;P{KZH#>=TWs6?JDbS3^QV_O>TeJ2F_JAs_E;p4t!6N(cDaq>> zxn}|xCHtfLpz~FE`$>(-KbH_7c}}Ayu-DJdNcz1H7q8&QK~E)0N-o9lejpI``=th= zmfqMqZL>5k&?$f^aL@kZSDnGDeM2IUL5^jdS zFdes=efyZslFkCfnaq|CEIU%m351-JDaB=LHK&VwJ%yB}Cj`aPm59Ss2MY2d{s+1o z+MN5+(o=qY{WW`j+<*A?^M{A6!xyaD*2UZ3o;ZB|^~18reJ3g(>_)V6@}A&5=cSZH z--%Ab;Jn424e76%J!1_p+ZG)XLq|YUaLH8Ufwi)oq1XkaB&G_xM~3Qihn2dbNy55z zI9*$$JNhaOQE3!Jj%wW-qOYv3FE1e?U0?03^ZypuZ+T;NePd(UJEoe#fJ8eDwP>MS z?Pk*Xb(7WrpE*4HY4bNA#{Ty6pP4G&&(=dVx;sJUOkkJoFoM&ZcXAuH*c zhJchJ`=Q0%e9PnJH04U_D0f%fP|S0F%Q~Jy=aSC3Gta6?3JT@nczODRiR&+QT(5)e zyP8*|I9;=IIhr%S84dEHuc-@mj57nD=n@{b;Ir_*b5ccE21?Hm`wuD*jXUq|N>6~( zNHAc|{p`F>H_5dN7!Zh@`szgkjXOzEtEf+t^MuH?^S%s=5F-lCV!mBR5eY9{SGm&! zizFUZSkvUZMpMWp_6zTJF+k8Amo%7l$5ikAk-p!;mPxUTg|5V4XQ&WCeQC%um`$@0 z(iF=O8O@@7WT9pN0tseRg}|!>Ko%-k=It|wt0W8iiR;O?Og93C)6*nb(GW7>&}Tm= zpmF{r{NtgjHx54z)dbcBBfEP$uhpRZ9#>uB|FO2Okv9^vL5hp#L2yCOidQ`BKI?~` z+v1DjaOX$Ja8muxRJ}iH$|zsO{u5*Emb8$GeXyxF=zVg3K4GR`MD8=_dItHW%Qap*H(mHd}stn%|$d758U|y z_bao*&`E+Bu0}gE^lh7`sO4w3ucveq`e}Qtd_QXX?>lV2-8{S^s%CIDJL7)cy`ojj z9l);ypcozx5M}19ZtlK~C~zi<57b>n-9K(7sFa%{5}J*J9{+7ngn0J^6@IunUq;1H zkwxR?nNk=gp8jIaQ-u6FCWd8^4w~Lfn6KJH_gg-8@;F}fv)>Z=S&AgN?rnzf%?y?D$Ss*3f0rq+VL)5CGIL$!xb;C zFE1`2YF(y+lmFLS!?mt2FK?`&bX}r~^E2*p^9$~j!<%1lx8`r3DG`-D&D|gxS7}{$ z@y7=TuhGd?)gC2jbwe55^srtnsw5YEX!;!jPH`{OyzcX@T?l!bNfv(Hyv6UGjRb_O z!BBBN=(EiqyFZi>iUveJ6-|gGTV(C(B9apX!Izt#mC29Hjd{2zHQ=~y8%)qARz`C> zqZ9S28@NNZI0}qL^yipB{ISgd-|y-EgcF2?wvGAGyrJU5vU+_x$?|;G=iCxc0O2Ml zr9_Tz>B4uP_OA)-(!<(gS*>gC^99RE3A->jXxeoe<@zH^@yC)+;we9O92G$#3k>7r z_aRO`l<(8NUC@l!$f!Gaxkym32c*KRv{DuCxF6$_`j~a)oyzyoqu2+zhFs3ksA~P7 zg9xUGJb_$_lOZ2TJ;S|Ya7--WA2Km8<<13)0}FyAF{yVi#dUbCeAb20xD9AkLi*9= z$<0UDA#Mxsg=1;_c^^e06*Zcw(&=Wx%Z0C*_Yz`*HjJ_fgc8zRRaXH;Fa=ff@)xFt zEfejNqCqnI={Ui|`KV#Bpqv?bsCyX;yfr;P&M^n9H8}>?!eUn`avQU`wMpy=Iy4!{ zXmfs^e%)_xkGkMO_y>mbW;owxE0Ir)%d#uAWPu_`o8#`!(_;s%TOEgkQSC|4g6Yw} z^2J!htfF5OrF&zB&R3QwX?$J`=NsFWQbBGUXRH`9&>Pp)Eabm9LU@@~6b-AbnWJ<7 z#+Od^feFZ~Q%1N~?^*2N*p829Qld|gq&=q+Lz=C_tP|0xPN&+*_<<7zB^LBVPozfC zH=`Og*@@?I8;>4$m%yL~EkGJJjDgpWkny$#CwwR_h779NyhUEb$6Y zj@K6#BbnmIKT}2I`IJbui0M-J;4gsa>0Ka~$hX{kd<@xKpM8)q<~3C)yCh@3g+zkV zvbfq45SgB632V($C?Zt(4t8BPR?;sHT8D!O^`L;6xgyFe+=io?+VZGIK_2zx{A#AC z&pXrl(#*wjv-8RsV=U-0^6BhD7gg7M|LSZWqEYdcD)9$PM zgas2E9sh}Pg6BsjPtd()w8cODG-J1Yp>`GkO1*VdH59k@NQ^D5Evzjg5?Wtb^1k(J z3%!+PG9%Qu_Gs|9u)easwy?0cyhx(Nwzl@6-m|EuyHDX=^U~D$g6@yf@J^>Lz*4h! zVkfqKeDPGWIBA{mNeVwC_GTUsX$=f6?YwHiB{Q7n&cI8Y@YiSi{|exe$Fh`UyZXqA z%6K{VBBtUoos~#Lg88Fe&+GlpuXuh7SoowuE5Z$TqsM|TScz11Ny}vS2R>1@#05#M zya!nhOJcMXL73-)l~#hq z;2iNrE5{;HzTbRQ4@VJIq**WqRli3e6-&l8r7vZu!^^!lDKLCX7gw#_IE2okqIo(z zD_3aifqUOZ(xlnv8;<7Y@-*P8W>x5d^~RDoE68(ay_VP{J781AY$Q)&Lwl<(fOf1rMFXMzVyXTvC#V@d#=`czs4a}FD&0L) zGRi!ww?qrIb72*#$?0FBnI3oF^uIPi&6J;L?m^P^DJR$-cf<$RcVZ3^C^6+Vy2c@X z?4Q4DvP2rBk_phd@@!JMNvxUDmskuFfC>SsJeeZykRut8=N;19#&piOYviKu__>J_ z6hSEPAQ4FkMmwElM#6#_V3Cu*_uZu4nnBOHV6Dat1= z^2kc|Qh`fA*e6s_HF{qgQ?R^xc}V>drAYZ*=eHtF&OI+9ftu@OQ+PNe0bO}%>A2>dmDG4aDh5q(DZumxf6+c_Y+?N9EL#t|yxa`JQkyvs}6;uy8p+AoXg@F_Vj!{Aj7 zmSZ?6`;u1F4>z&ipW7g`(j*J^NR{CyD5-VbfXm6eRz4?2gn)_%)G%0!$*AFWQ4faY zULa{ebLt6Y={>f#EyC~y-RVlXx&nrHNWt{UmROz9uh9zaxAG6ERxdEltPXScZ=q_{ zF%%DgIrUCL=QL_F#RS1qDtbIqV8YC&+j|ei)T3zS75IMchHe3dJgTk3(L@V1@TnQ) zY*h2Clv|mIMo}v(Rl|`&8Vt$*yw&T3+&z}oNZ2G>A6cwf<_gi?d%AR4vjqoSAzd1AJ<`i(_ZBaHZ#<${8L|bNPF)X_r;O7EaF5ITXyeVrgY% zagD5)D$si?z23&+1|O|1F3@q!`(u9wI*Nw_JAUR2JHfcPH+#>?3Sts1zRB%Ai)}AH z4)+^vHU%S@^aSSGp*3NB8ylLpJ3oJn%j%D&A}J2C*V*TrkEh!Eh2U(ww*UT z(k=Y}#KLMt_7bN+Pi4BwLk8)Oyd6xXulBu+S{Q0fWIg2!l-A`=txAf&ta4eag0*LH zWg0F=EV8?7xoAZMa65fkBH4a~O+S^+171}-dDY~LX~36z-n#52yO}jy(Qm(fZ&rcf zo|benW=bzTEVh?D+t9jiuLK!g!GCokf%COWcR#~|g2w9;Ts8&NETqo}RHlejoW43# zw3FLRe2&-h>(}0QYMP5}`n=#EU9U)4`7R58&Hpx@l~iqrN6d%Wh;Yis=ynJ*=CAAR zB{0rfp5!AJYl$VUTpO;5ZLI+ovj^s*8;XS^s;XAg-`q_}sWA2%0QI{E-)gciKT98S z_jl!t$XrdH(L4+8BTV3r+GE~!bUaz{(&zIKMRyf{L-^qd4f-_*}_V-UmyG zG%~y^zm0JFycyGY+Lp|>&Aqg_C-4^yXQrOf68yH$$d#Sdtkxx*W!G#i!D{C}O*hmZ zx8C;o&tlKei{|{jPCW?m&R5d;j`bXEiGArl8g8tSbQEuBq=lRz<-}}mfpCMW=kL(iS(WNCE*vit<^75K@f4;uEeRgx%w;Ir; z#iZlStiq8UQ{BAbo^;gRyPrx43xgjs=b+t6mGWd#$aesvPN{}7Sw-OyEW(fWnFEjU zsnoc0*M!WOHqip|K@!c`1tmW(^`X|wb>1>gh!n^}e30uIUfom5o3BUIi<#RAC1&c+ERId$+M5>b@T1c=s3X7Gr9;|ms=i|JZo|;e_c+ zu{rzB>sR-P9-4Q=tzDi|;n6?hf3-QhB4=6i+Gm((Ssf0UdSSXa<%1dPf?qUn{lN#? zkGZ>4g_l=R#NGPdy$lX-I42ByzqZ0M+a)heyFwPT{?7a#U7o5-SR(M!Vq%ti`8{^> z$?N_|#)4v|tPC(fOt4Km)MMHYkU8^^Iog<)bws5l=H05^ZM#+0n6_>Zdt+fWk-1yb z#zhs!q#emU^zc($2%CnvGXJt;eoEf0e6m%fFfJz$wiXIc7D;$34daEJ$PKIfen5pm zUd!T1FfFfAKU7}Nzr>C~lfhba`j19ITu32MzLB2z6 z)pJ9Zi4r-pE|MPdP%6W^t2PO^_mrjw$GxXCy*TbW1-?mLt8dkh8T(?6T70B|kr=w- zY9hjIy=NN(=P4?&C9wfahs)^PwLL2#WP+lU&0_cH$*oo5vP&2)+sweV$cC*f=&}l( zvin!|8gqgp5)_e}IDq%-7DZJS*XC>Yrz!#3tK0XUKA4R2BKe?xm!j?br}P(6IoLPz z1JD7!!6_;wH}7?EEvSeg$TlqLIwV zz;a4TqNqc>%hKfqDTD2o2lSkI4>`vyuDY0khSZ{QI5s%E z%KMt<4Z{G-arU0wo=YV~M;$uc$QWhD@;Sdb!&?Yz=#Y(b<-#0OX-=qL`IFo8cTeCE z=rx8608}KKYH~%zO08{?thMKu|K}a|Rde0ybT*yj_g-j0=O@~;{nIf4GIV&Q#oa3^In1|f< z`?>E=d(OItuoBF5OX~(~)`)p}{Py5->l*rv;OE$OL$vME z=qNAJg}NGQ+ixOYQh9cNJxxx57m+HEa!#6xtK;Gp-4&iZGE^}?|MBC;?#Crds?r^N zeqm!{eO@^NCLWEF)oYA+1P83EGYCFqZN>Ash%QGbRiasPP9ZYAcuDBwNnenj*|)fN z=P6rUXp7!0)|&qLRHZtPW`Ez4D)>Nq2&?Q9GB9hP1zf-(va9gG8lo$IhN)rg=ov=I zfSc0sx(X8Zgi6`Gh!=E&YUsf~2AemB@U0G9qspLkc+SwZ22A_=FFq^+ zCsVPR4Mhy=1RP2%>4F!JaIAc^$kpo(8FHZNBq5k+;k87zw$W$VMZ_Ibk=r=LurwUc zTTG#~vd9vPHVoO?dx0~Tt;NWQ0Wc5VJ|{#EcDpc|h=gEUul<|`6cw0QHK-1-()0t9 z4c9bk6b>171Nx|&%NH|-XyQ7YL$kH9vr;oP93V6rj6x5R7ji(7o_Oc+3bRm?qHxH0 zVuLZ`1XCu_o1$nNm_;URcs94r$^2(p&bAz}#_y69tEjw|Zf)HZMqt{jPQ zNjIeop!prsPs=Xf)EWm;8Z7Q8!B(JvYq5$0}_zYgM>FbU>sziP^-h0binB=Z4ycWZH zi;IiB<>l4hGFwHXNL^CRhp2+3C3YHFXJ;wx4hclrdG(C#7lq-$ugN>wJODV@cFx77 z8zbNgUuZu$Ynuh;tXilT8~uO zq4bGh{b+!|UQQY`99kh;WZVlC=Dt~KHIYLuG}L@jJJ8N4BCIPXAl_<|r)A&nkL;Oq z4~67VosEbTH3}KeNjE2zBB{xfm^>ohZxA6wZ*lC_5ug>crs&v;9oW-&aMqtcKkvWq z&nPs^j@J1Lg707z1MpiJ+>%R83g~wiC`!F zv9`Xr)?1{b)&eaxIele$!CQd~1tJw{^U?)GW;{Ka047oW={IKqB0itJQy*n!1RDqE zl+s_M^^(*y_gHP|o&rga<+`T#;T;eg{7~6@Y3!_E)y6_xMp4}<_kLgSZO6=MkN3Jf zQonugr7T<;I=g^na!(#i8qA8!Wr?_oV>4lEC$tY zp7)v%#on`&Z%#xmJb&^5GTN_={ zm7K1)BraOSD>kudiS2v_cWe6Guo9XRIRTAIR&E6#sd&8|w>4+c$@$A?iBSIXS>lvG zbC!wninlK_+M`3=I%v%6`U^qJ%4Jl~aI!(M!Wa_9O(Zb$Vy%7(JK4v9r?0klWXx=6 z$x&cDXGalM3i1W-Y=@i55^bk7J8Dv*rDIu>679@U9@kR`vu6dsn;yzT!GUsD&<@;)@l$*|HxlH_;;I9~<-$uCchjo@9NNb2aQ zbsAB;6~Rr>agTAd9#QeD^Bz=%VyM+*SrnqNzP{k1*hPuh34i0n={*joRI_^qxP)zN z39T~s0sB0u0qOH#8yI-jo=no2Qe~4)4u!Si(L-ZCBTx-bb9(iN4`|km0oQi%a!-m< zB^bp;MaLnzGxJtr=UUOkMC4_swYv}dEN>z|WnGpq^V-Ec>xPXDxB%(3I$8fL1f%ac@Mv`*-8S^Q%*QyP&`4d8!k%uzluE!}`=ZM@0 zf{GX+*$Ku>?vf!xV~e+~h|aCY<0@3&+DzoJ`sK0~mC(JONjqs6jVl%IDzT%qd$R)& z?b~_lFdH~u4#2`D!Af+_(*#3mj04msR zUK4Jg-&G#X)b$o|ki{gCi2EnB^q_;EU{+bm%qyU2OxXE=T<0IzhKc_YRsK(Z)GM%= zRV9CRqZ$^Sd-s3Y76Bdz>`Q9z;8>ZJcEGMY;kCu3Xt$%UW|j$fQMs{+UNeiy-NCul z&C$4EDQOIR%|>f2ertbt^M`0Rf|@;sw$O7JtsvDFme$wM)2y)zk?o;QUC|0V5iM=d z-=Cf-D~tS7Y#5uYudi?LZ~nuc>T0Ay`~TIAH7GRlilipJ3viu`98=mfL5nKCMg+k8MREqZVmj9Yt zdQ_)v^m*(BXj=gU0JU3};02q(mdD!bSbuem%9on7xTj!n-S-vcs~m#Ts*N7>UgfHf zV!2-PK*5uf=2EGa?@BbIB9w;ji%!tC5g)5UW>lhdeD3M_dx9}xoZTej50x~thjqQuahby`>A_x`=ecX8psIv+v9zY zn8Jrkj$@&0BkC$-i()^;PFevj1}x!dt-iF*q#B3Jd8 zJqmmjV>*~z%|KXsYZcK~Pc!MRmzI(1@c*qzzHL)W!!aV^^^MiVwe_X-g(aLQk<<#% zjF&+hr7R$Mtd41eor|0KXVxH%e_5Qa%-SextIZ!CTs8!QeE>AWxcZ5EW>vbMJ29%@ zzLi@I>uP-s<|z=IR~NW%#jammHhnAZ^Rcp!IySC?Eh-p`sj3EI;Z4T8r)yw}Zd%3@mE9*|lrgJeh+{ObMC8SpvehO`eL{A)K zcBYWl+Fj;8WfKk!PgQbWxe1G8B5`6`nm1Rc*El@x+>007!O3Odl_Nu@QiSKPAOG_B zZw@}p7jh?j4IlXdsM6kDUFPiivMvtek=T^9wJ%)~!kk zo9KpPfJCZ>eibHKG&f;aKPp@lC>S@a5rgfk{}U!bA~&?0mXK0Wtn~tM!)NM-%r1-C z5#BH*TL`oBvl+u4QXj|xv`w%oT21EapI5Us^oOB4j0Yg>mjYIa#e&4;?Yw3~MH-?% z+J@c8xqY019KPbLh(K&++QCetwcL~+>09e5JcM*Eiocmr2&LY$eKZRvEwpHw685xf z)5lsnM!mZ;Kk^o8^KAWXR^$aXJ6mG)E^geaJem!)vdZua-bO!B+apD+o?wkZ+P4;0 zAO|*IUFmJead1j*WN22E7r}4C+7*SFpQ&Vj6ptmzrzJEZvQxsy(sf~F?-jr`X0}P* z_;kazynMVBb9o<;599f>X(py8Ul>yF)@A#+x+~em;8~dTMVwyjk|wQD98K5H1N?Y* zN0l*Zmpzj3d3Bk^zkB&{p@(!RgJJfh2V#ZgFoaSx(wa>+^3p$1c%pdLK>|-@?zk*E zVMa#0n%q`sQ`Pa_Zynr@ROTVBItC0UX%(LOzEh*fw@~Z^9^-lD>R!ipYr;O^0_PM+ z8Bydm!d6o>#xpxAxc~)SAC15IHmb@&dpJC?DiyP27ns@*Ptpq!spVRx{)Ds%u_1=_Ic1V87u@!Uj?KKo>hYgvJeM;Z5ckln?P{ zE?l|<1_GsS0qN`F|2XYDh@I-px;%0bpZVJ@7PN69uaExCcB++K0~+;XzeJ;G68nfq z=GYX5Mk`-P&@GF?OwHtbiJE|+)vJH}V)s?)^1;QzRL%Z=kU&&4oEzWD0$R1NFS*%+ zxT=7)ZQZo+=@Yk$DTR{8j*H<6!f0gJ$c?eEe)*Bkj|T_4^NZP|g~$3ph<`?KS6m+x zP1SC!aHs+iKQhnG1GIUY>gfcaqPi-`2s?E``z*bKY-MXS>QjvckT%5dIxBqy6C@vn zcbw zn=IVd*WI=^@ZK?WZlS~XG^<%Lyx}%kl6TV1ZigdE9g_1OW?)MyLyJrE4Q@&bk19W` z7pp9`Y7&*GnvfX((!W;0J++SXNxbfSJv;i^iXL-)hA-CVGr$~O&#~*}kse=m)@Ah4 zH&)Fy$~g7U*_n?0y)>3SVDPijtTz^) zSEmqLsWql>3*XSy){ELr%%;F!N2x0ZSEhZ4(iz>3|2|q6`t` z(4sku@roj%?iE{@mowDe&_-XSMUU=<%!H&R37vI3b(64?Viq2hOoolXE7)A_p<0cY zKiD^h=v4B4PXG#VogNG7@`@ttN%UqtKnD5hIG@F&KwT830ukD|o?YuZRd_yPbjGXPY z2h!pZ5Z3F3G`VR(Qwpa#YCg#w9lwLRtM)q4hRWjOaY_P|%#Orh;``l*x01)jrjAwD zCNKJ6cuHu84eYCf`EY$I*rTNrhr(k zA5$*-YWPtSSiM+ChfxRnGKYcbcq0@)tAY3T>9ZPu-<*4&{;D~5I>cu+A6uPcZmnjp zVb^O>u;tRG%9X0e7cq+!Qw^bz_~Ux0zU7?5|4}W5bz>ASPG|nK&A)Mb)@1E({NKx$ zPpNt~z%gSl>JydpHbj*?I<tvNj{f_RLZKQ**?0F@GcVEoGY>K)(l z!-yKN4m9=?{?Y4V!**u6>XApN5o8*;dFHXl(LbZe@^)U5f?SbUgG%*9P#r7uNRYks8JzjgkF@JBmF_ z9qL&Y5m-x4%xih$vq4%mdnE)^pEYecLZ#f`tGPMv6wGg>DhJSDvDy?n$V*L|#htJ^ zOp~Nh^pi35ULw#vNZhK>GYTHcXL%cG_cqmMfR*NoBgoS3>#II%*Rqfet{#8ix|=Pf z8Gy!+>#8HtZKx@h0zsT1Z?4-vv!<{g7Up3aZ{`CMX{H;rZ8nRvI5Ycecd9M3ER_*} zv28Fa!-Q@iOh1>7xJDNX-Dgyd z?rJLVTb_Vb^ECo4^6d|f?u=CsVEF6V2ezBS1V6MuX0tOv@>-w?@TI>NUArX=KkLuT z$WwNUDAOCe`R~}e5{qd+==QDHnfr{=mh-QY#T{I>Qe90SZ5p(we**h8m;quy43Ro4 zy>~Mo51eDjW_fUSrqhlYEpeRRWq3j}noG%?XaF=Gtb zyD!@@_(|*6Pp3!X9|BUK7Fg%)9eZOiQ>^5jhm3o8Ap0Qxf{(&bSq6A`{R0;Et}{1y zWRMc8^>7AD_Z*pE_6fF$5fU5b!<8)5@s`>rn%t;ZR2!VI-~_%p%Z+qX$v7C3g6PuX z2_lPSlp~z$;qK|NK7LrPZ4&Hgdtn0Q%;^SRH=?SXZSUZZpRK&@m(E6c;$7>pqUksb zh$*RrES-?`Yn68u7#Q8JWHGhZF5@}-p#W5K!ByaF+wC~OYD6CQn_=!NUNNuqDI53# z-kP%ubp8w-;hdT|f!Ll(z6C#!Y-*T8j1?BN&ZISV-mn%q{s<93F(dY=LKeNh|BH0Q zKRyO!HGh!*ZvFsMMF~TNc?otdnYqkxicRV5d3H*zIYi{yl|1|C7NcTkH;~$KV{Mxa z3^<&HLJ>O=7&+^`DTQ%75)e80jGd?T*@?zFy}7-E12+z`w5}QzTViJy=0h^5ypl5` z?pga02@JD2{GlKn&_9;y}8N4pxjdUpm_7;JxfnGNo0>*Blu`uXZnnF!w&MC-yr%`QfOJ??{ zAplz7xC-`V*OeBkja^j$Mh7vqK~&k#BqcmNXhdsYV)KmNaQfhNKCzYG=`L; z>GSv%c(>CVlLUh>ZS%7t=4)Ran?mz9wtxeN?5Q{)IGofDPHY-h^#Yq_^4y$`KdWoq z2?vr8^RiZD&W%kAO!qnv=b)Y)9t7*v zQbTP$FNnH@SOy2J)mX1s#D?p}Vl4tWF`D~$VX>%P^^x5a15edG(eq^hb@u~k;>)T?5 z@lIJrqCPc>+$&KJ-Yr&^nxUiU6brQ%S?!(DnnpBlc{_gn`{M3Mbu{AJNb!#^#h4s1 zWRuut{Co+jbdO>?N{yX;Tb|S)lvScfpzZlwLTc?*Zs?&{Uwvs>bZ^Vdw|XnAd*^O= z^D^3!xwt^Ups6as{zGTsz#dRr&x91nd1!o}Ljqr@Gk@B*jNg5n-yd6%Rj$P^TMElR z8#WukHTDIYw-egaeLEFT3pMn)Mc*K)YSrf=j0iSYIO|HwEaN{`U{2vb5S)vDp=8?X zfU8t_g&`%Qe2;z5#wS;FEg)iMHt>q1BWrbdBTgN40Yby~qs~A4gJt(!OHAvxJP?+M z7Sx714^?x|b;TYui&f@KiCImT$m8HRPxw_f$x$100lNM%jS?Y@l;`QtJ)P~L#ifp7 zYX;ll;Pyfi16W~Gp~~W1x@OFw8517p#cFHH*;o$xkTcm=FiDOmUIVwJG8g?xXp;pPK>b% z5j0GSszoNoGh`q(`PH1xW;=Pf#1qCHcRrsav8mx-$O?Z=m@k;&NlYOUh&(meJAhdu z586kl`(OIPYlvNWp-p%$ZFd94Pzn!|RftFtQXb{8e#AYF_J_w?O+W1|Ft6{X{*;t` zlN_o8_l7mZ`FgS=iAmT<=7%-T)-*A8a%QH}dA+;+YVY-%gYCV;$Gba+KW+c*Pr`ns zcVIg;Khv}ZY$Wd)j@@Gim@KTUQ+B9p!Y{J6mSpX){12aRDb%YXy)*v&Sl$<(Ys;k z4bRU;pG{>s0ImW$SPs)9N0Qw?YSwnR+fbpHT&*ZJGc`fbZHZa3A$0|?g7IM%1Umyx zxtb_&8%}AZ?{Ba4oo1S6!I%(D%Ow~c%BATNLwi{?Bu4NPXJVmHFd7I=`H1edd5&VE zDipwFDMD45eo{uRFZVFhq&W-SPYPp0n~s6XXZGqgOF`Psf#MSC@Yj>>ZaRIFHLQrL zJ9wLB*9v)5d8KO6;P@rXtmQjtS{lwK5&#hpSL;ce#DX@e{T#FZvPXNk6gj(A$?}A0 zgN3-EPkPBly7>f1<9c|7?1N$&%%n*l@4xD*kAp{;-x&?PD`Z?f)qlwv8Y1*>RQnY$ zEigEuk|LJ@Mwq7J%*=2T6t_%b35!&hRhkIyvzL%ij3!gfWI!Oe065m*bgGMQzi}k& zgp-q9Zi6SS6?Xl@8GVb4fB8ULKjWl$sz_oFEH&Z<3skW*@C4W*?UOMyc8Y1ue5 zd_P24-7V2!#-^UwP^1{^m9$ zOo%OMUrg16E<{gh&gPJsxo1?r0Ib^DnpI@Z4%C8l)P1Z(Y>?|Y>umuO#msG|EG)B` zjf&%5N@n&T6Bhl>{`LVK+Mn-pvjj3VC=qZ18Vov9(Mj2!qJkypG+YP3##789?VOkf zd6MFYWm@qq4LDR*ke@egohS6GU%@;v-;EeAM$sg6hc*ui{Fzw6k10rM46Z~vTqFmy zVvpbVkE!CV?1y#^v-sWw>1}xAE3!UWX-#g-JTDM47A6spXUtKP`sS+HC5`1(=@ZUI zFO=Kz#bP*%8tS2uYVh~pj1|$_y_oh)zmz&>S2axY*IGAy^iZ5FoGo&us9jYOmj0Ku zi{gd^KLAEDY5N^ZmWr?{zT2R7d%6&`)M&46K`=iUuH_ z6AN}cWnC49A8+)5VW)GV1N$6kPRAab@XB=ot(26UoGEU7j5G=4#>h%-TBpA+CZ?k1 z!P(61y;wKx=_w&8Q@;*(%35S9*|eVh)b@LHxsC}RDH9u$oSo-qB%FyvV(M6Cp9Iot z+su8`-A51d!db^aD5S=dON|-og#k8`)eIati*`|vMI;~3zj&?KCn$B=^n8fWP2GJj z=px4ywMk~NeFL8l+Uk0p%V1{Zibrh?%P^4kRFfy-IV*S>No~Sd^(*yUE-(W$UQFUy zs#06;kfa5#N}_LEBWH1*_HPagYHy!rW4!R{-e8tl3rYaY39H})+#n{;6LVWr`r{7S zhtvM~9#aVf$A}sI^%8yLk4BEAEN>df(mo`0HdH~V@Auo>Ox5y8a;T4;2Spsm*58uK zf$yl}oBi;x(U6efOXal}71^HDet%l^OhCj$nAFbzE(yG$X5rJ5hs875JCEIvDqrU2 z`_TMWF}X~p0FUIr87iCN+RbPVAHv1wtmBDnZnh1Kr*1BMewXGkHpH%BB>C1*jIwTX zZB&-KHp96kqm{o1_Pi&v&!g(EBb@qQup18Vfs^nYU<13+8JU5LVs!T^d`QG zXp!(awZ$`5uAvp-0vC;hmY%H2rNC4!%z;Kt)JMJW?S!+Asep05*+>0)aP z%_XPAbZO;0PR>3J2UhD{*JA|A!}8E|L1sR^oK2HH{fkbFiw7LJW2`;t=YC*zmYJ!eXwwi#IY3sE8#hWGAQ4$s=%vm20`q0D z9u{CEy52^LJi&w6-yialv%W8lQ~wmV%GBT`LJjcmn~_gA4+xoedSuXP@< z5}m$+yNValeb~P~emfUd6O!lpnC3;gvi3Grge>E%GRD7FpoaZ!2+)PswV4Ix-<@7HCH-p+#L-^R$-1c`P)=GLU&sFdtk@p4zQnu8 z*|(l-5$RecoPcEK=lltu#*`YK2k$S}*81902hECOzVJ2c?`#?`%Dzc&89nJn+kWPr z?rW$AEeJ>K>)bROg*9U?SEaIT&Z1EFqVoaHmV6qUCto(1(SckIGE#tE# z&JQ$kU*hLznHHh_^_AXnZ~17ce{@2l>m}5I?CiLF)H_-mtR0;U){pwM5g$fQ z+ILG4i6xXN8JVA@Jq*1rL5~NQ>{QvbcAM=klGG&FBhW$ddvUzk5lBV&tETRzB;D+Y zCyMpKi4o3X3e8jM^ul-3lU~GditW#py#M&{^{f4Z$6G&HBw^km(`Sv?2Ybt^Gu_(lRa8#~ zv*F65#p1L0heTKQ@C|w4@6h=--D-2B25=2ihesv&)aD!aYLnVu3=XAY=-8jpyhM%r za5rby-Uq=RSJ8y>asnMg;J2X%>GfH%{PycTCa$`SiA;XDIph1`1*X!ts*2UxCZ%b` z(AisXb)SPFg=S{lYJ#T>0L9an7tEIYV4-Wh1q!)o-!w;|aYKi}tR{OP1J8}&t$4@R z-L00|MuFSec^J5gKJYY>{I1Y|5=`{{#KCy(8bfr^wX{<`v`NgihS76HkXqxyu?&$* zFuu>q=wH7CLHr>uwE-CaRRdDva+aZ2@HSPs4d-ApQyX5ETDj@WDi`6MSpHBcT#wdu zS7JEwIgJWX#yEsOs*eKlvC-C#K6O;p|B*OM@w82nZEy6m<-lAM?}#|DGY4(POq2ZJ>Obi_ zs7)tju&l5;H3-vt1IeIMsz&nR!gicsOmag;OnEd5DquhU5pCkOwZ(xwCSwHHa36nm z0quZh&BMz>P)>>Uh}?P4eKM@{%U4dK4-I8 zH{R2rwQ#Q3LlD$#!X5$o#^H2KQ>QM)=5;K1eU@dcF~2P=it;MOVO$9s=1#9VHSkOY zU3a#r#+RcO)774lkl~t;Do13>$$Ofxz2mWG|3dMCzbCz z5_!apT=I3&PI$7v^Nz<`3@+hsoMh(afcBNdx zImKjwc$Qg(C=iz4v4OL>OPf_Kg=Y9f6kQDYh%J;}o?ZML2@yRt8o~I}z8fkI!#K`M zD}rs}pH@!-vq&)G48-UxVhsktNyVb4|?Nb zdl&;xCYlX|-KrniL07HoY93Ee!CM_*ZcH=3&|MlAN0c6SQH~>cp~)oQ#&mpsiPdCW zy|w#zWN%thNXoJH>80H=;S^j}=o-U!z7%9*DJbSo*w*r&O!g`6^RZGMw&RdhV}6S* zlQ!yGTR>b+h%>$y7RIe0Qm8HhsI22|TaMkxb$-E}c{3ydg+z*sCkWkK-h&by<(M%k zXaFJlmS$-`Je(p-i0(=#AWl2pe1Z#JG16_N!ic1bi%NrI5fua9KqDLs9rHsFNB?wt z#}DTHB)&cuR53#}NN&9bh!_wi60K2{Tn zTM-EN9&&8Pf`jE(8&GEHiLj&EBLfS$%bTtzdGZKMc5zw=-Hs1I7OqS>6nVoU#j-S0 z-zRFdB1<~NH|mm*hZ|E^6-0k^ZbQU+9IJfde}RzwBYRIjezNs+``Pm!cmDaO7cXDE z-u;)o{ew3@|MJ(r{cB^tpTB;&{qn`m%bf$#5AZMlo25H|&53t9Z-SBTK4NPt!hLZ@ zkLLe9{+@@QXrDrWBB3I~&Pvw2L}8G;l`|gSDdb=y+;SVX{o~c8mBD(yztrDY8m=8J4UdinY+tiV$IaEFNOuj^Oy{NPCa6xdOb{!pRCPKR`jU7EYJW{OPs-DRRZox;RZ|ogYM<0=A= zH6R{uZ+6rL&0?|Z7{)R;ej3LiOkZKg8VuGZ8inmyxl)<)PpvA^jwMuSq~>Ng$IU=% zW?43|W_qj%Suv=!YaKxFfOD58b5TQUmhj#SPz?Uxy8t_e{h4v$8+xz+;s1^!cjf`K z$`qp-TX}e|UsB&y_wGJdChnnrB37g@wuT|wq(FF;?1Q2l(%tu<$2W%a{vd|D0tM@5 zENb$EET#in;$;IcA=c*Naiux>MU`@>1Hm=cj%aeuP3;Mf zDK7JVk_6a#3BzIvEo;kQUqfo(h2^qdkt)?S)ceCWXH@1yhB1b+F|u2-Q5QFJgy*V| zjS%uX*dX;JoNA^7@N~n9`kJg}{t$qzZK<>g&&I~`>gwv+!b)#-iQRelV~veOd+Yr7 z#_E!qPvieK@I9-&9y?z5*!4$yxb;?77jcUCvM*g@ab-jMlWNDHwS{HeB2BH=78h38 zz?Ae08=S5#1TFFwEh*YvmCZDglx8RAV#9bdf=|5iGk~?{`6$K;GTOF*ijc|Y2^(mk z$I%9yDmIaPbdmoZo)fGFl8)2Udu0iRhZ$r6ZJi_eR3pW3X5T3~)T>}rjaGT1E!iGC z{9$5(e>TeSjz`Nm_43(Q1Fd=iwnJw$5-5)P^g zAvKDPcknoOavJc=jTQIhl(w9EID4-a#CY5wC`$JFExV4zsPV{ZK8sHq8+As5F;h)#)MBxUOM{;e?B?X1C95z+sxi=f*yHhWxsk7up)#AG%0 z6&^kjHIr#A<4gNqF5jW< z#_C&S*R-^OK{DFq8BIrkMl8yS%k~ zh2So}^{3Tup=uEn`+N{F`=Lyh<~?`HNZ2i*X|h?*jo>WxAnJflc*xZ2EcBMw*(>rh zvtPa{rdG8G2%e|S|EybaahJu(!AZS*vwzUJJcedfDVpYN87gG>fE($AgQWmR=AFOt`5N%eWTctUCaA*FMwQ5bS&BtbX6RjEFy3qMWp1T@_ z^dMq=TZsV^r7~&_!Hgq22%|ea{S);&p1*oRO_}Ae3R>2PW$}VJqT)a+ur$^3*u79v z;aiW9ZPA?8gbq$G7kLUxn~Rhmm)TJo$%Y;09+~52bu|<~;a@8e5stSo%JP$h$rLOc z3&yEx>Pv(!=?R>G*@R|S=aQ9d`|R(CxzQvkT(?oGUsl2;o0i5)aa$%jOHz(Lk#XlUe5%Bi^-KV`(@CdHu2c(Y+xw%ut#!ChS#wJcGO1!C@e=;&=ow zJUW}rIwZ#*s_r=<(C&WR>O!N#mL%?> znyG9$kgSml%2p1M&r@MYh|HvsX=uHF>0uhQY~-!eyEd;A)R3e845RByH_!FDk`xIv z0dOl+lfzMd1G7~yXc4PMfl>r)j_h06&~m0zcoz3Hw#+&ee6yD_EPD{|5@I&)ZES?bzAf=x9JP(%^adr{&g4avou^lE(s~@0y57bykbbL77!ClIM=n=7Fv&eOX?ab zVB+pXWy}7%v#XzN_#Hfp-z4KtMHiO}nwhJp&oN5jn8d(0N-WGwN@n(OcklII|K?$cMriB#4iLE1XFBSi_rzGPnP!PrL3Fr&YDB9ndy6)rk@XwE~71MncGdkRu-Be zIJ{N^*qoX%chY=lNZ1Y2FOTEf?Povxti8`JYp=6Uo8#!>By(I&&jR5Fog*I1#L*W;;2gsC?>mVf z9)=l@OC0=dxxa6D_PE9mv+3`z7jQrEuV>489$$eS^xR$MCXWXH{d0gYGA+sE|3^Qf z^Y?XdT-4xa;a)wzsPlJKy-e^(_1}HX5Af7&G#Sx#|6ISu)bF{W`L|8b`y{yFfWe@z zKH>MQF<3n&^{-p`{1H49?_jQAf`5C0r|doJ?hBlZEfLQD%9YS%x- z;Nd5_p^JaCNagEvSRX#o~e!snkxqy9e+T2b5>o-sS`pM+~ShW6qPv^t=BCbvT z{pjkyT-<+q&cg1)&y70ai&FG(5|*U@dhp`EZ^UgZnLOCfBS`K-_o`fRNiaz2=I$Tl z66`4tlqqDMvOqx5v13d9m zsY_6R+z-E`<>%=U;3t>+2YLES-CR68gZyOy9mHTZC$hGzacytYB!axs2%4#$brK{!V9-z~sb zBk&yNCM^m5GsAf_|FD6Xmx{~X(gh46>N*N2;G zE`Oh1&?S2_UdCql;GRUh*UtOU zB_E8#iHwGycYksva67n_fs>C6~&1#PJUhG?yfjb=HFiq@>jUaJcDEbnET`r{Nyq} zsXxwaU8O-me)2#+54oGaLP0mnG`qc-Ox*PE*%u$5v0HkdYQYM zJV1)qOZ@$1{=rh5j(fNS277wC1j++kT!Z1#TyYCO(AC}FP3jVipY)RkU{CO-?s%02 zT^FPXfNz3Z^v9hS6zq>%De$cPBcbq&!Tz|f37_lf7mVKXcMWt2^!J0y@^=gJ@PPZp z*G1)?c&;EYKqmL_lLrU;c?RNH0#{Feybj~(=I)2m;PyP=#9d?o@*p?7I^!;t$viQV z3U@yb_&+@DAjOTG08c-+V1HM603LPl55yZVcqPZxRe>u8sFJ&<%)>9(KM;3#JY0fM z(*P71@|L$OCZT!>czz3cO*X2o99F z<29fEt;;0KKb+$J^*}EkImQpOd{~K0#!3S7eLAkd;KDP`E9qUqo(}=xMdknbcUW=s z#~~C?r3Ebw@xi?-cNo71br9~5;j#(r8Qvdn0x1-*Di;ObHiB8>btJmxf{EcGh0TyD zySlr$y1__ba4zUu+;PJJ5N3*b^#^qH(ZwaiNCE6drDvxo`mx4;ii> zArFe*FnMbR+)S{dcx?BW?&t`b^gmT0+-kDFyB8*3!K2M z3Nq8$tflpF&!hLa-bi1#r=o|^<)Ac_3u?T+cjo0a~-DL45@ z)KLg^EQN)SgoXEnh3&$EJHmpy!UB~r=ZrAtM`2F0P|aMZCKl@X3iYgo`b@}lg-jx3 z212GM6f&WjOsFOhs`U~wZ=t5HP^cr+kqXt2t|nA75vrL9)og`oT0%7)p-@AJf2b zZ-^Z`G06flOenDxY7JtMh3$~%%*6E$3opD6`4#aeEc5PJ7Nz`NsN0(*K0L$HAD(Bs z9-ct_7UI7keu=5+m_p21%ln;5D6f*5Lbvd(@-DEG&GQ=p#`yq$eE_6JJVD+Lrj>ammb6rV#zdz zA#E7YwKvmrmoQBaJEkdBgXlw?n5Jw1LE&=-p870_U*+qb*vy_NM8nvfDB=p+HMdn2==Fq{i&lYb(E=&@-&gA z34MV##GYxY*+NifHKeiOOyueY5i^mSD-+o|L%{9TA^jjG5N{^3vx1m0k(DPCnW{k$ z?`_XSy__J(*Vl)M`iLM}Or$#$q6sl!>fXLg-OU?f3NeRRF?H9W5Ftd9se8zny31h5 z0H*E)Jtw5udO;Ak)q(V9>h_KhH>Pf31nI}rtwj)^70NJkhd|G42vZ-V#ncTY5b%B| zXTWd>;yUU~T~D8>tLZUyfelk7UBU>FfDg2 z$PkDD)AF!}ctTv5mbDthifI|^K^O#ShR#fD;2@?o2-^nqV_JG8+ckfBU_s3Qb<2RcFE8U_wv+Wq=6?fw=J?4yq% z1m*XYLwpb)$h2W&+8Qzl%2PLoU_T;l2=*z0z6iP^Y-bFDeK2Q;J<}Fq8+?einh`_{ zabp^~3WyJ6I1>tF5EmvC4g$IZ4S}u@Z^X4AdT?!nA({|%h&IF$q5}~@hCtT|B8T`w z`alLC9p$P~xh4?k!`TUio{(V>l&=PTA&E2M4v@YO%GVoW1nCbkfWQNz-a-NDirE2g z41q4%C}hS+XAqW$uNR~*<_6~8aLi|4 zA=5B}e;Ev);BCn?hG;Vlse-8+V_sQ{*|xVYnGkbFsA~=}g;+t%AgH@8>aL5n=~_e3 zPF=JQeJe!YN)%2^tnIS@m2fyhqWg0pt!y0vwqi^J5rs1OoA8ZJ}+lOgbz=o`K;dkKQ+~KP=hcOK&2Nv_{ zK1+H3Kq!$5C2m5At5AYtmakCaDU{d=CC);LgHU2AlvoKR=0b^yP;4p`TMES%Lb1J2 z>?9Q13B^`Iv8Pb%D-^p5#qL6}R4DEx6dMS|eT3qnLb0(>+*c^>FBJC^igj4x+Y2oG z;VBk-CxJQ2+gaF?4&Z$jr+UDWRb4Df)y=l69wY4)fi{yE_lFxm7u%a-SkRS&;cnCh zG7WP>rX`2l!?co^4`X5DV?uG83zNj!a- zq_=>HoeeOga6yCN4$V-OGxB-3GKsz!%7YBY;n4zZflJi%!Qr$o>g@@0k_=!PeUV=S z=B~z=xX)0gX%BO?l56V z9|Iw&13n2B|@n7A(vn|4?lc%v=agPFvsAMzSuUvQ7e zqow7^B$jgYkrfk5QMRos_6JvNh_Vdzp<~A+11y+C+8g`9Va5*A0B%&QKzXh>wCk%g z2_3=>)tJP<3+1D(PQ97LzCX%?yEaDunDj<{q)Y3(9=QMP>d&* zEfRZ_Db_|^2clna$QlZJ^TA^wX(bst0C~}V6SNO~swstC`oPV>t!fN)fIBzGzR}+%Mog^Z4cCr7>~9Y{ z#+X{7pS5uK9^wPLM|sZbunmkWmWUE_*q32{*f;EGh#`~6v1ILsHb~H?wEWb=(ggEJ z%aN8m7zYXJE5kCz8RKMu{kUMvbud@3gcIYiBG&82G~8hSgE7Vfur!sTJT@3}4Tm^; z_!Di6mp0}xmNpvj2O9Qp<5=Em4=_SI(MCt~qq!FZ<8Eoew5V^TuvZz`fdcaeV>rkd z(uYYfza);Z8z=YyCpnXNW88dAn8XF!u(Xv7L0WJ4r2ZH)2h1}pL&XXk^bKrP7jp@9 z5bL1tbzoCEI!vq&)JJ=;tP(RC>j6w`gm&~sAN4|i_QSZ?Vh&?}nkMiQCYXnw@SD2u zqxu*VEPXW92gB|W$1+OO0Eb0e*a{sc-Qj~IuwfhIH^ow`H`?lqWt_ucrfq;{_xnxUVE<8TSzp*7qbeSoD-oZ-&7xhhx6GSTcz_{D?b_0q_M9P1unLVvIiR%^=Vn z;K{^xuzhFrmn&?^3^s-1hu8{ZJggVy560dXeIr*uuwCW=fnCUkz^2i6IIR&&tkI|F zS9M1wHin-Yihdt}!?GRv$=Z}@TH^G=LkGS9{i=gwfItJs0W7F+x}~WBn}yHT#1c$X z1L>GAn#MR*z~^b$$(V*C?8*$s7wlU@YJq+>Vj8ZP7dTDP(8j44mN*)k7z3QL>8Qi6 z+GAOYQv?kxJv90ou=tkykS?a<*$biu(Pz4zmJn0K9U;yTdx!_bjp-;P5M78qM4O#o zqXW^0=s{>(FQ%(N+Ja#aUx)(Y3Gsl~K^!3tOc#D$*U*aT_&7lh$F-jG92Oqv4uFZ^P$#E7j@IMw?ls_Aj2TT znXaua)3vb_stpvX*$A1nkl6~Ey^uKz*&rb^q*b@B*+2-o*9=9NIYV3^7PN5Jw=iY; z7)gCsJ*Mvl%duumUyKE>t|8NtWADSn*gIOGfmZ0ECIgHihD_H2CyNe4nXZ!w#GL8a zz;jz-xv!8iT|HN(tBYPVN6$K76y@k~oE+##yqF$lHqO17PH*%WJcy1lMrshY!^`VS zeVMi+#2YIkPaLsDE;!M`2@*WIt{G0K2E#K9hKI0(*R#S=8b?1ZQ_OTztj^pp>wR!E z!RTuFA}+wmqCPx>K6J4v&@zDcFm%Ms9}F*m6%rP{5?joEcT;$89A$9=twTp!Pk2jD zAG8NnV}n_*Wsez;S&v1&q`wP}T+km3PdX5*9ZSq+UwBY;cvB%}b$_fX4AGaGLz%t^ zGZZHZn!|7bL2Q7F4;omwW0v8BUBeBl317@Y5l&j{aZ>JJz|>7}Qg4P6ZRt=Glh!yt zlh!D?i%Dy^No$^Lh3sBgV!Q% zCpVLa+j*G9!vj3b=3yZZODQ~b;Lah0**q+zFu&qXKEhfaX7X@553_h!NMX@|J4Fby zDXdGnQ(JMTA?Z$i#hv=XJM}wxc;L>_h&xR+caCo1VHyuJdAOU0VRxFU?;I_^bF_ko zRd<>y?ld8hQ;Jk_T1i_$d%~uzl(#91hX;6gh=+wdOhuU&DlSxAXu24FG2&vv#hmXC z`3DA_-m`G=(vamp=B!-3ra84at$Anj?&hqMiOuQFrEAYsoU1%nbw1{N?D_BuN1L|> zG-rHY7TmnA`9Sl*=KSV@=EE&}nu}ZZHJ3M+HCHrOHdi%QH`lgAv_zgNY>974Xi03@ zvauz#C9P%ax!Tasb4BNh&y}1jJ-_9A()r}OjbZ2N?ly)~5J5pC1yK}4QxHQzECu^1 zI6y%*1vwPtQgD!hLlop8I7)g)N$)7>9VNY^q<56`j*{L{u2*)h?p!&f2q{fR5UGRB zoQ=49vdS{*~N|`{(a}zPDu~liJ$IWDq&T1>y>EgSbOHAf6C~t&Lo2 z`|rYieC`$QX{~K8Y(DzqvF0!*llF+6kX`K&aqSTa?Gcsjk#RP6qrz+vL{Jb(K{N$1 z6vR>xPeCFDTPR4TAccZd3eqUpYHM@945d7X*nuE17r~xJ1bJ}?%F=CZ9!3-)$cRF) z7ug<`iENLypyWpx$ont{*&go5eYl@gcjP|W4eY3ePDed-I%-iFl&D zn$++*_w-@IhfNVZ5$v(233cs}v0O8rm&|K|@*|V^AFps^3a?VshvECDgr3nvdMe)| zdJlC=r$NN%P@{S%Zx^Mb^4oYKA9w05l#l&_=I|1bC6f|S^N73;nxuw|JE)Bj$G94b zg|!@|kOpWgZKZvoCP+bhv3Eq!leO)UJ9z{d(O2kL6otKHaeYK~^OQY2L`^`^t0<}` z0%aIOPxoRhc$~KG9gb=v*S^1okH^W(_VKak=_r%{vqyLJgs2A8 z{$((cs3)p{`t(Gg56j2&ff|bH85ATUj?O_O{jX{x7dj9rD11K;kpMMty!^lWdJ@{utr%|sftTg?I?y*Q0C^TigQ#ul2sKA@PW$4l#a$-aDn)S0-nTP z>bKNC#DAgxQoO9RGI)7#$ik%y7i|n#7PxHbvfz;AAt50FAxlF--@h82nYxafa21Hf&L3OEb?1~JHJbp zE?vDWWZCj%>sKyXxhiy3=-PGuYoQb3M}Z8US`Kl6xI)|@?hp@%Cqw~(_d`IM@O}tL zS4PS*(w33BjPzxcK}I>`lta$TrW|t0A*UR2$|0v5a>^m69CFIxLOEP0$HgtphnkC8 z_O`^H+}4uxG&iy2Y4(8xNYvAul#HhbpQPiU9Y%Sb3OZd`t{w{v##&Co`3!D^@8i+kcjIA*BU8?FyeYDv{J9tLK?59 zLAFA+LDC@^kW9!9$WF*E$Zkj$WG`etKVxLJC$@@DzX z`kM{6!fu7%Y`V4e*0!6)w_$G64LpMMBZwbC`VsIN#A|@HkUDIu1J(oU z+oIdTQF=H^4o8XM#M|N_32hNbivW)Rj{uLv`B5ZvBat2n-N?2mNHioKlF$}~ZBf`3 zg>6wtk3xDB(xZE74@8am|9FH6Z81oXfqo3qV~`R9{TQUjB0Uy77CaUlJpd_&lu&7q zc%E9;R(7MjtqiQpq-{%Enn_y{1<4eoP>@PN8nh1GD2G&FG-5Cc<&X-DK+KK9h#yA$ zFye)X7b0FrbwNvGP)XHq!Lnv zlB$pzMwLQ{S|LQWFc$HU1dKyGBmv_P4@p3-ct`?r#Y3po@x0|lK1N+;?RUuA`ZL! z5K@0ooI>g!)P!3o*(36D%Y=>bI#NjXhw>2?p(L_FvOvd@cYfh3bb{%0sFrSretuoaD0$X*B10%At8m^`!lx zEM79TGl9=}@=}yeHQ{*?uWpOFwH4(fgbhWJl>;Mb2B91(yOeZjuVis;aXd~lKDI3m zoO&PIsbv(J{8fkva&Rq^ztu2cCa6mkSBbq*4ljy5I(V55|8mpjx%14e>>Ry)zp$_z zf$5XTrxQ&kno41ON^QMy;70b1oEy0}4q}4lp^61J4s%BeH;HP(QDACUqI0X!wKX?t zVQ%$kcmz6yx|b$z9Pg-XR3Zm*kSb$l?L7bMZ%$;N$T?AT;?RkMXHw6kwd`!(){=9- z_)yEimfYsd=IwWnq&4qo-rG{poZFn;yuW34OICAUOH|9@mcsjclbcIglE1HQKGIy% zT;JT(64sL6a_D|>UQ1p}+`aPhmXwwQE!izQPKTWiKOOOOPxN-kLCB8(f+hb-EB#lz z>|gQnf5j`F?un@bM_DoTAMgW@Mp=KQ@%UfqDEII5(jUrytid&p>T^eMfusIh!@0(D zN6$5#J9a+ueAM~q^Ktj$V(-O8oZoYy>_Yj4$_v#Oj$Ej@PoU5vaK zbus#4%*EJ?aTgOWCS6RvzbpRo&Wo8BcU&&GRC+n(Qrg9gi~BAW-rt#asl2V^=7F|S zse{Zx?%?9!>fq+!?%?6z>7a11@w2t@x3vkdwF$Jf39_{bwzXMkYqQAKX0ff!5?h<4 zwl>RbZ9;5qmfPB_u(erfYx9+@%_>`)P+ObTwl-^QZPwb_th2RQZ)>x`)@Gw^dsupC zdsxQ0jf>WX2CiJWc3l7#fPo>a0++OhH?)T*(isN*Rjc9?mBrhYDO+%IgHv^7?yj!9 z9A$1nM}4kp$NtW!1FFoO%J?K@(-vh+k}9)I8MmjiI-)auUspweGN(+LTBmAES0!Yq zk}`0*kMsQQ@C0Sep^n;OWlE)LOKewlnJRu;=b`YTp|Yy1B2*U@l@gnHHx^%U5pAR&G0@ zOixv2Y*S|BC^Ms!=<3WuWoD&vdxSD8Q@JBXnYCTHGfla(N*Ncc%qdasOIGfyQtnSu z?oCwguTbtUS7sL}b1IZMHOkx=Wp1|eV74-^NSVJ|nV+Z3+o3Et)V{k!Srn%%YEqU) zDob}LOY@YamCDj%%CZ<`S$6y0aAo;^WqFCRGE7;Kq^yoqR+lKN3zgN4%9;bpn(X%d zrOKM4%G!Ko?NMc2LHmL1_5+Q|`Z8sGrLz8rvLQj)(9{(du58S1&xutY%~l>QY|qJ2 zHtkX#OHv-&t~^$!iU@Ddt#8lGQbnYzB4bt2ajNJos_0}@biFDjSrt>Eipf>Ql&WI) zs$wftu}A2HCvLYYevc}lT$PxvN=j92IjTx*R3+uBk|R|qxvJD`RZ6rfxu8A2M3q*c zN^4SW-J?pYX)h>JWu&Mwid5TkRJ&_cJLB7ncB*!zwHL*;7v?Hswkl&XR9RuFU0Gcb z#T_X-Rax6rS-VwP2UL6Zs`hMaFOFC3t5@xfRvp-?%8piLC#&}FQ{|?&m+Vm;j8GlQ zQRN+09nM!3?rtyJp(;9{Do$@NPgWIIs7m&!O6pZ54Jy!zW2(v|Re7GODpOU}q^hn_ zRfnsp3slv$sw2^=sw!1&sj9A|y{b~xkf>^iRn^p~j+Us7#dL<%b%yQj4BOp)B(pO- zp)>qwXLwC#L~&eu5f{-}8q--C-w~JBS-G_% zVSi_3erIJ)XH{5dbxy~Y@Xngp&f191y6BEARUJt?JCD|PHf3}ktM3Yn?h4=E6<*en zlF=1W(G?Zhk($^Q)ucRJ-4z$r)wH)Gt*|R0tt%n1D`Af^p|UHXyoB}urDS!a7j|tc@7h+`m7dv^k=v26rz7J) zM@C`S&eX1*SzSBxx^@l~doowV`8IQ^)SuuEYDg3X8i+61vJWx+)60D#N>~c6U|Rbk!tw z9ck>UDeG!T?`o{=j!5s0Ebop=>yFyp9ar0(xV<}ZM|XTD%YDRZvV#gtbH63|} zx_3t^k5qJL<#ZI3bQB!v-nYNwa7Fk2#P0nC-8tDEMU~yT1>J}8y7MBs^O`zJle>$y zc9(~BS03%COzx;G?yk=3sI2X-Z|JDb>~5&*ZfNRmjP7ns>2A#IZfp>YWCHyA4yXn^ z4io}U0BOVbKr!$Kpce3?;3H-YJS`Z_^Zf{s;XDt3cAfCYZ_f+JK~Y6G(cufVBN5!EBy>Tj0m_?f^;e zE|B!@0ck@!kn}r%)E=cEfVZbp5XAF!0V!WMkn-IJlKun1V!r($khVVp()Pzd+Wtha zoR|Mhu!8G72a?{eK+<~wqzx~Dr2m^>4cB`GB)va@r1u&~+usN_@btGp;_uP_B7p#^ zejL@nHNb`7gbdsUC;}%`$2dV*O#|&C)C5QQYGQCg3AiRu3!G3JTnDHFPN*yRn4_K` zkfT1vmG<8U{Xy6loc7xfoUlJQl`{aG&;Xq3F%X<^5IB`P7@W{hFqYS2h+rYdq2N@G z5jdeSIF(}pPG|~l4KxEMG#8BH`78vBI9kH4(4lHJ7!N{QaH^*rIH5f_mFoab=m<{r zasnrG2B-Z<1rvCAu7V{T-2@Z){@eviIeG{t@%~f@mT~j~r}_*7C-erV_W6Jl`ohks z|389#5Pl3!dZWP!$AFXGSa8B|uy?BOB-jbzWN@nQ6mY_+;M%}x;DnzDKIP@k5Uk=j z6P(JO1y1-GIF!Y{!o-vV$#Kfw&1?k`x) zF#w#>1HlP{FdwP^7r~AR7lTtfmVgs31*di_11Agt$2?S94o3)ix95_ApMTen&Vx;DUSC9Dvs@# zFI0{S^M$Yzc1+j>J0|Q#eW^e1qrQX>)b8_q57i!Ue56*x^&Shia(sgMK>41cT&nL2 zluP&$_Dl8q1NKY!N^p{=|0z&%e2sdM-W$}5@U7qm$9IB1IFet}7YHzPejLfK^#&)u zM#$hdXuAl0gHRou(lx*dH8Bp9F2*y3!G3J?WA-aw3AR5oYM8c3H32QD7_cv z2Vrk;O78HRQ{g#E!OeE>M20p^Dla3JOf;ULTh!oipigof}_=r^Ge{1l-v z+6Vg+TA_V})@YA4&<5=xv=v;V{ukN_e&c8_xXRH%@RFk=>QDKcP=7*aAzlFiN`-iJ z1So_36Us3k2wepE)XyxMrE!d5864x-c8>9EC&vV~kK-10kYh3{;E4BpIHp1m^NJnD zcEUnd&eMxnHOFFB!?A?baV%wLIF^BvehWLv@eDi0(_6u*UT0Y|$8!vCDuG`HlKvG@ zf3hc$00=Hp6ZNI-c*h0L7Xmd!gE`j`4di;dB0OJ>bUkpwL3q}gP$6>T+ou63-zRt~ zh{~BRn#K8NA}`LrfIjVSp6DZyEyoe0kxf*;2#BZMbJcoq?OL4@Zm zffvE4{7c}3mr-v@zlQx1UKicq={H4f9B-jK()$VJ5#AQznONY@B0N(K`~~g6_=)bJ zUWD!7l->bOsKjeigeno9Uj=rGx;Wk!b#r_m!n1)$e<+&F@e%rm%6W|T5k3K@?N7l8 zpMk^Qik^cL{wktp3?f zz`2#iRgT|j9OrmK;}XZq5*vxPK!As+q#R`u7mluy(~>5np8;M6wrY>no+S`yj|2Jw zCjvhMPSW10jqTd$+NZRkuic`3n&TO8N^jK}ql0`p<8+ScTm^oobBg0>;5G0cfhT}1 z_)pu<=#J4v`MMKy$8wwqj{WFP(rwdi0p0|j2mXZr#BYQD0{pp`WiKazpqG`Qx}lpu zV5niJ$x&=5;izS(%~8itm!qCxZ;pKoy*Lgt^ycVe=*w}0;c$*44M%bO2>!xSATT2T zNT?aiAR(;-r9fH- z5*i9^Xnq(C5!!MbDj37-Wh4mZNb5k@tC0zgH-x6}r&Mn<_)|h!2U5Lh9Y|;i4m&ln z0w=Tvr+V9f6WW4fyo~I?329wO^`Lbjp`+kDueX!n2}frfH>n*`95)GN;Itn(IH3zT z^^+?&p&QnJlNB8l2L{fD?`dr~QlrCmat>`3CC!DG_iP5-DQ=8236L8Y|6r6CnU>e66f|VR+f>S+affLgDlaSV*gtNh^UUR?+ z=Nir7{qlv;T#oawo$5Is+X-p?N%?5~N$3Yo`TW5N1JwAs&L~iAF5e!cwt!=>8b7`o zEi~$xCyR{ud5+Oy9G|J2B{)74E(ND@mVpz7fYW}LgA=X*r+!)qPDtxf3m~mW2}8lj zZdQX6t^p^#wcv#7)E4o6S+B;=8;v%oE#dvTQEe&5O@djpPB!{lu$JRyaH_{Q;Dp~| zy-WQchV?FCI5_ot1UO+NIE_;jIAJt6wJQdkFczH3i32B$7ktjs69nryCW2GGE#QPn zSU*s{WUL;)&>tw<-a4KgjIN>&M>Zf#Y!VJtS!c6oRmA4)H#k$^T2g)Pd zsrD@|cQ^bR<=+p#MtA^R0?Y;{%n_X6?aLM1<9HC9+J6X~Fb|yc^1%rUz_G#TFgRhM z(SBavBBKKwi}CzB^;Zd=eJ3mhr*g``3CqE$oCNh*9-OcNobok-6CM@tb(B$);6C5~F>u=Nci@D_!71MfaKi7wDc=v^geQ$^ zc{@%S)p0xxzeRdK!fz2agPQ|ezzNS_ok-Y`sg-f;2B-bp2Pb?0PI?c)2_J!D zgVAGf!YAP7z^CAZ&tUhI?>X$0@K_Ia3~u60Z#Y|>*2lvfw2JC z4~PpGNER3i@meV%1Bcxii@*uh!AAi#zzH?6KBxR*Amx`JM)|bxZU&(?IOWp;C)5R} ze0t!7`dEik{$4=J-y1Q?*T=Xw$G+f{uOB#Jf1C%9{y-q<4?>Le2je|ALPK!U8v;%^ z6zhA^Hvy8qDPp8&W^BsQ9GvtlzzHqEsa{s#gw{CUp!_yK%5RGp<+n4oIoNyt|2Pyv|Amv|-80BAL z%&&_YF9oN3%fJakz$xEyaKaUs7o@)u_DQ%3c0{-u<4L#{{YtnV?I+xbdJ=w(c`_Op zhIv95ZY<;&fmi;CM;ePbMj5Mfj5gNb7=sx_+hZ|L2;(qM2;+^lI40ny9*8Fz>u}s+ ztjjS8KPN!x$(T`uDX?S0RM;_L8s48J+-hvjahtIP$8=*$jv2;Q95Z3@q_-V*Ot=Gf zOt{n7j^i$4dycz}9XMtgJ96A(?8I>|ES~iDVH^neV;l$%82fX~HV)vJV;sma*Eopd zLE~VKhcIHKpNDZE%*Qwo78oz#c-VL;$3o*}9E*%YI2Ic(=U9RfBmGj08;*CzWk9T7 zjmt5DlwJW&SjmbwR)NzvA2B|{c@6v+<*Nl!zB(X{XFZVeHR8Gi=^uv$5I+HXApD+{ zaQp$Bwx2RS#rbLYDa!vN{0CtRkn~%Pd+g>cknHXpkn){}UD5t80BQdhaRq|%T>_`= z*T4y{!yk~|4It^Y0ZH$s@jH&UfmH6#K+^vO_DTA8fTVvH*EvZ49y`wGM+cDRhZ5UK zUxn?2otU4LuM0@|x`CAMK0CnmAEG~qKQjKBx9d6VpW62;JIC<_IF<98@o${}4!=VB ze_%cm{s|=gH>Rn~Qy?(iWxAE)Zqsy*C8ifRo;AI~@to-;j_1u7p_#}`%u!;d%~8k9 zoTG)AB}Xf~n@f5l%|7DzvDs*jW6j2K953+UxWLShqrcf?jv@GI3Cg!z;LCA^Ss=%i zW1P~2m(J!mM>?0|7t(ng=S#ojxIpU1(O(+K zF-RKBaiMe($HmeW99IH&VgFxAS8*N+B)(d@hV!*R;_IaAIo|*zzEQe~^RI!#H%rHG zo-Wx%rC-tlAhkX4SfMvi&;89>x|JwegxE)sCzm8+WAh$tM-2TEG z`0xL``xOklBJJFucI45QZ!+rF^s@9{XCL}{!>O4b2iJa?I#6p?#<|a?KfnD&$DsO- z+6jr4znb2Ti|M))Jul93;QOzlCtthW>&h5G^yUbc+~e&_4!5QDOTv4Ap~Jk+`*g?~ zemi)#$?x(wcjrG=m_Imk-JySzqiCz%>-<5(66R`svhL)1>*Eimqz*57|2llLc)}mn zWjh+r#NJ+49Q@7d_eX{=*=F@1FS6PAi%DgJ-}#A)$NGMr-0d`2CXQII*&3^~UGns5 zYf{=6-RiQ{p~uZnJvuaP%yX??T}4GVb(R{GjeO}J`q`nvGaZwjXF863aqZ8qqWle< zF5OoC?7nPiK<%7Kib>^BFa5swp(MeuY4h~D^j(1uQ*}PgCNcxZF>hbECn$5kY z)^|?WEq!1g_dxk%(aF$7#v8whEg0c8p!=d@!`l6!6N1n@x4U*1;T?|t!f`z+hu`vRU!x?=I-%ilF8grqg+P90{j ze~Q}}&EEyym--4X*^3*@b*c~hzF@w6z8~~Sec#D@<_?!$7;AHT)1i6KpO0FYF|v8x zPs6r7Stzzx-oNe9{gk|&=@%zGv!3a?d&S9WyD>}OI6Sp`nwp|(9#mlzA5gPyyyNtO zoINl5*Y7>QS>A5?v|&M0)F$sfYk$#L_q@-;SzRVmtWuUYuNj%2H~Y~sRiWU)j~|P& zzH?Z5_lKup(z$o~xqO+sab1m3`{;S+9&IX}HQ2d4ZkO6}``uqgzCAlTSp1{Wi9Hu@ z4Rlb1pVv=_d9`D_w&l2M4x2+$MpasjZT4$jyZ3h3`Vsaz;-=xAY4Kh^TvF~{7v#A4 z%jyZk9?aPF)p}9KXn!rM&G+y8^6l~r&5k|QgNu^-=l*IOnwk~(EOn)C+>~^Czv#7t zfAsrSxp$ksrB7h7QT$G62aC5gqX zyOC$s{c?B8M{^s7wDl>{@t76g^4ldngW!;#6aL&{X48^jK0V{>m)}aRJ53C*{GnC% z@`~ui3*Gd3@7OzY_?5=S*)1sH3aF@UYWQ8rJ>isxN4)xzV_4;y3qZJvMJ;E4PJrSgdw=EV+4E z>m$9x_vZZ&88G$J+7Xt9xua&d1z+DjHeub*hn@t!DBiGA%V3mA*y~f;LE4F7OP$^g zUvnqo^YJeu9lmyca@8EGpD^!hGjS=p4&HZIfpBIkJQ&dzOPWShPju(6TRc2EbH18bg;fp!Mo)L?poG+ zxi2aj>9A&g;xRJh`sm^#luz&yR~U+IPCuc7}rixE+5u$!XT+7m;IwD(3^T8|K;;ed{Y+voYt}vCYq}#(cL{y29D5_bL5dc4_VAf9`Ac zssH0rw1)PQ+|OOQyVjk4EBbPzTSIJKdhR>($x9q{0;}D;`rkC@-n06`7xOfed^aqb z*5NpP?H@0e8$0do$~*0`c*N&9*SxxioIE_AwcK$x=`=sl>I#M#-d&yIU1jJFsSkUbmG?hLsFYF=AjR!-#!1G zlX&~~TZf-7S^91F0iXPfz6zhAi|i(boqm;f$f!ogaa`h}3F|-oe8iWnRR!|9!Dp6K z&6^+g`Bq)8%%*4e>I|6m$_eS?d=`wG{VMq0&N!1VLW6JZ$Qk%Ld#uIS)O~(muTw0} zOfbAR%+9Xg4v&v2cgV|!SX3Pdnp!n}%7L7-8&7+kz4qdy_nRZ1R&TMn(5NLE;=HBx z>V)0LG>kr3b7W+$*KUuS3$Lv*Uhv7e(Yj;e-_KmQ*!%Z;Dg)`*&fmRn>n`}+d(Wbm z$yKp$%K!KwDSfB>;FR!p-EB4t9_+{-veoeo+oALJ(ZfGSkKHsG&!7EzBd7i3`8OwT z9Q$R%h)Jn$o3^VJzx#YiNMo*i=HhiL%@i+3J-HR)druLGHd$zEcg=2N_?eRdGv50b)g5{MP4%AQ%G`$W9tMZ5 zj&U%S6^9IaemBo=p*9EDsopzpn>-2N1_lBZ#->kTH_U)<7R#s1iY9|lu`ojEj>eJvc z_L=D&zr7i=c=4YpA>%?0o~gTW>h-k#bACwiC-k#%p=L{A6Wv z-cg@Wt$yQclRUp+zug)$yG%au@x?{$Ed_6XDF3yh>}^?|+O*|AcsZ@!_A+Fas#5WM z&2JMPSgpVI@z;749+@6R=P&o^GtcYJ=GcW({j3h#hrP-6T+`pea?b$|@4m6sRm-QP zR&H*)IDXP+*GJ5&|3NK(=h8mUGPkBHdWE-tTwOlYeN#lh+?S=xO&*y~_e~vTeB#nd z^YjDJZFihXTb_qRDOOE9|TQ?$PptNCZ98K2rbws+>r%;xtawm6(PJ!;XN z4M!Wc4{j}d77)impI%sfzdQK-leWpzzCYC86LVinYnv|rq5sQAstY4C69(3=dn^-P{jKqEPQn!Z-s5-4gI>Qfskr9&n?dBo zIrqQ+bmSxNZzg1Ywz=ODTiwr>Ij6VGNnE5zpEv$wy!DEq(f&Q;g4 zKWAr{KH25q=q&d?=jRgnO1FNM{rtC=wYE6!k#|PFxl%jt*_tut=XMQueCbs5`;x6e z$9xt)o9XD8EV;g{wYhxi@HDZ>qEDY^_%z@5dt`EdS;e7cRi2MNDYO3Qe&bL6F^BEu zt@^^YH00(_pC&u6xm4eJ(?Hv_vzM7X!su$B4d3QGXc=);XMn&i&U$d}{C>eJ{_OqZ zCoV&Ne8pz02-HtjK5qYW;F~8VuacJC|4K4{l=+6dlKp2Yermt-#xZ)r{L-PvZhN(U zd^ddM)bkNdvDXf~zq%^$w%#)B`1pYJ7kvjW&n`afJuaZX^$P2^>lQe-P1>sRUOq-^ z$GbdrW5?9;vh>r>hGacG{8@m;<4bF;3zK4}51AmJQZi{(wZo^8@9q?TE$y2*OEF`8 zgYb!L@{!749cMc{Sd!Q{I79vEeCxEO5lss|a}ec6#JQU)cSKm9Ej=%58pdn7`iEz>cRSoFzS|m5Z83GlIiQAu8(uc7 z{(afth{};Kguk4-vMKu)yD7DGQ~o^p^VyoG*UxCDJv5r*G;V~d=9j4W0TW%GZ1K#D zbulbHkyw!JKUH|vY2Ey}lQv)(86t88^L7LJFdw-s# zOq%?~_oh3IN7TI!xZ$@r%*kn7zq!v!uP6Sxw14H8^6;T&I_4YpFR-~bvgp93y^G`D zW~F$gbSAFe-&|)KYV7Y+kTFCt)FZ#wWYeWaO1G-91zsy@MYNyOqAHX63p1Gbc%x%iuttIn2__;m@ts%O95*m>gg!PtGi@4qXV zbGAKs{WH){Aq9q$E>$Dq0g zvF3@NysQYj5*FKTY_#3Gzh|}PyGP=$`rE4YE@WZ4En>g?)mJ{Bpp!ap(&Ll_(H~)x zLb@!PPq!~HTB81Q{Es#x_V=rB3cmN*Sa#QANAeF{39F03{B2gpKG#?}&}_tsXV*WO zB3bLwaA9NY*->V#y*@d${o;As&#rCyO4gE~o^Jd6V989={%fK%XAF+mHO}!_)EA{5 z9a z;*r(()qUC>(k6zV}4}NH}wDfJZ+r&a;*{X%-4?ljXwjgn6SjOB99bc!cx*7g@q}?Nv2`aUR zQ?GonXkYQP_j#?iulv5;jz#Px{V}gzL_DzSXW@0#^!io*9fGik1DEIKeCDE)dp3LH zu>4=GqosfRHl;J6dTDWpQ~2OCySdAJUVO5^=I0wfWWE~N`qRWI_wH(a=OpV~`B-)M z-J;SJ&1J$_{aatGw^L7Fo^R^%P1wB2q1X0TI@fHTJ zc=zoE(^{vFHNW?%j1FAh+2`UPe|$P*-7O2>$sx`Y7L9z|KJJs5ZKj(u{JvUdSAI6K zPr#tSyVIN^@`oP_Z4H(MOtk#Xv$d?D-jIw|Bg%IJoOu&$G^5DK!h<=lrRZ&Yx$SH2u8c!@R!v3i)SV5{s`dF12~t zy?e4%z_zhd9Q5OUGbkUm(%_@VcPIULuSQYuBJ+Y-=a1*MtFAVkDe+s>dyx6Er_(Q< z*f>4){{l2X%fBeNLNCm;7pc%u%i=RJ`I$)bqtC*&_M+)aPsy^kVxfG2u?Lq8vG~<; zp#rulRMby3crg8vX_?*KCi3p#6$5><$q?>e&lUqu#+=yb>IlMy^T8 zgk@+#xtEO7m8}zEbzDvwRtZLle{=G#q?DevcpNvzly&vqsqGorjCNYmQ7Vq~1~P8% zqj-kE=vLFokv~*z)A*A^X9%S9G4lS&#z>J7nm0|(GJ@P4ShWC4Ap+O$a(bqc=(HF1 zsYPM&pCVr;0%ZQ9YXsb=KCo@f+ED}6bI5U9dz8XAZn)?mKamse;KtkXv6j8#N{CfI5rkjbQ&AXoxTa@k0q?w{9 z)lEE-z+;LDw&-XZVn9D^3ia&co_FK?WdRSKT}i#2e-7Y!PADLv@1!3YVys8rx@1L; zj&Q^(G3JopnjWY7C09?MOK3f=RbS~9w3HsP%=bqD!mX&jNc~W#2H!vc7#PkX&Vt@T z$_H#3b|hL+Y-A%ZIcT!@k-bRXqB_vSfnUiXt<&msFh-DzoiUMNl(_SQQ@Y6Wt5E@g(7njp!B^nG12hg+Jt3k4^g9SY@a-F!KVWnICbPW zf9YE^GlGR8q#MfGec=no#XR^AjFAhJ6pIxF_ods2iWs+x*!rQS{Qg{(g^n`>h%N1b z3ABFr3nReC_lHY(@zFji(4K#GU>%)XjX&~>CXgbzXw!x@OpvXD*83_2mSo(qus0SH z`v|O^^Ms*@)Sut7CF86kyuu%Znryu`JBKNqQusU-5iCY6AyvfR{1v%R`mv-ca6qeSaoioEI69Ky@?oxH@pctlnH2*^6<2gU|M)5yo8c@6>a%`@ znF*VSVjH-{38U)PBJAH8G{ev0ptbp+vt`?yw!d(dO;SHr73_5|FM!}Tm+j@?Q^y1= z^yg}9)CT4NLJ20M+<^8uFR(h=H#>8wQ;3^tI9-LVG!ZUgo9&3bHQ*ER3mPcxjfvSha?cVz!D z;33S{?oIiHDA=I5Ku6>I(uK#w7XzoRjHUR+Y6aY11bB#!PAA0RCBy6Mm}VhXY&i+f zyYCNPfRO|h85!K7fK*(EPE4l?&d74J(Wu_R`geaFaGz&zwQ6m(0J=WTP$TjnP5=l_ z9AJ$ZMMZ1Xcj47q+zIO0x-@mF89tImz<@EP1u3MeWXRLdo+k`ZT9E&?LGwv6YnG>o z3F0fkL8D;pApivnbMHVf9}RC>8TWSycfb0NgX^tlVaSdx0Z+TAg%avNt3H|oHHg;21duTJ zBwg-a>HiaUOKz#8J|evm{@)B!nsc`0E?kcSICj0MYC!FwJuqdUh7`iYF6PG`DXw8Y zGnwC#1S(?Sj(WSD4A_6$t^sVk(=OcXbiN1@TEjITXP-tI%RtpzZ+E|1Jmo9}do^Hj zKyq(-KfiGoH(9cuN(@9CP|}A$Q?Q5OQUydZOkyF(e80Eb&H5tFrd2`jH<*aaC_-8w zUQAeObf~8_aNN0}PJSkvj8wVL6*^h-63|OEA#Pr7laGvQ%rW2GLBh2JP56q2`(89} z3YlZeV!63tPVaAxZ24sgTgiuWk0r1 z;-D>{Bwl;cpG|{8-85_)wb|bNc3w*!PkbXVLrv(9bo%XfRx=x-4Qi8>%}x2_T~U~#`TdjALp?pwed3uCUPYS zk2dzk^&Au1U=kCo_q6&uHS0hr=b^lvHZ%&~<$mli7 z;FSUo938kF-P!K=Q*qud%j;4RtqHoeKZM~GutBQu=1kqIqBHqFxb_o6gMW8*IuYBX zecvcU(WP=c8+~oxZETj9mUwSepev^{nnUfnWzy;u3Kye}rJ4Q%xqi`JxzVbS}j!5;z`QsaW4G~A(w5?u9YmMflp)|?r_Vk{^@y_^1RTbdB+?y2%Ib7S(iYi1s;_vcg z!MO%~YJtP$r3&1OE#IkSjtUZQkx_~N>XpuNhs^7l1h8s z9_cTny?8cU!w-DTDhXB>ks8dPM)lZD%y^Z6sTDyV8 z7E*qp;Kc@O1d1}IUKWu=_B2v94Xn=(i{`dKI|&+NTOA2X@?uyA99+1+g^xb9PS?d& zB(lmzcO3PJ+F|F}xLm@uDDc29>~JPruemJ^(?5q^R$_vo9uGYab2{|MeBS&HS*tTV zZ=2`~Bm=D}^&`mJrM{T{nzw@9vjIoS`BJYrrL#q|-|(SE$fYW7iBXg}flzC5f|bI01u& z^6X@4=KFBBd6FO&H%6|T88{LA?PW(u(1Gi^5p@Ubd zYFz-D%VQ)%&2Dg9#U9f^cAAbq(~h5do{b6ZQ&eJ;bcpQat-hmXRe4%afGb27P?iia zVhos1dJRCcz>Ke#&-6&*5?JnD*^s-qepuvQ1X6*2r9vAnd>mv3xsun~ha9Jq_57W= z00bL%eXGafa9Y?jEtM8h8ZN_^{|SA+iY=x-&iI@_X9JC(}_ zFOMi1&qL2FSQI3v>F!{>i|ma;pqg?5pT|_y^)hAb)g0Jg;eyXPFmk5<^>f4G1HH3B zWBli|sHlM7dkCP`QYp8Nqe#G=yF7pZu=vb=WeU#;>)~e@z*LOq8+KMj(PDhf%8+sIKcBeLZ z`M?cS50thgpLO!`j&nrD^ckm#SJ|CK!%zqGw^=CFfFZK-i+Emsv~gwuUm;5>u8m5{ zGj_JyJpFkO5)~=l=47yMp!mTcbk)I(Hr4cXc=f3T)SoZy2mKdw8*;On=-T+(@ACS3G3pVcY zs&I?oPhbVrsY^a9chc~J2Iehgs}{xi_teHT9#0gi!Yh)lL-n@V$PRTk_NL@H_N3*i@s>ajk}3_lkzZ$k5s9$H+zhTr$_N z?19Xd8Aq$_Bp6}yRbBt?pm!NTMKQj{I3RMx#D!3mv`hKFHB&8upyJ4;*h+~vyAdq1 zGy+tkbWH8tW&VwamkZCVi>(y{d`1J|*IhQ9klX;6gW~5!~H#{Is z@%{ZR``e-1sn{}`kJNYY&p@l~yPt;(pEmMo7@gXNuJ8{HRJ|$lP;Ao0D445J($GP_O%|kQeq)2To{f=T+YYuy*mM zZBgvU-7R{%Q&m@U`Z`ske8@LZ!c|TDKVLC5T|TcWL`LW+|CMA(5xHTn^R;4T>;6YH zH@ga9wffGnT=N+Y=AA^j4u2IC+xR)wzhOLE%&z|fXCXly)TE`t98MzbtB-$E*&2t2 zfm#hC2>Dh?I^RoscO7(qP!GCb_lxtxaUBT)TVk6J)iai1-pRN_x9kHTN9SEs5_Pn) z9L{vf@139TGQ9w5PfB-Ent1`Hjtv=7DQ&nx>V^{J0)>dIqqaSB+j_O3$f^TPCAk0tiHUx zee*#>0KrH0>}X-*%Iub-YNQ;U<-KM4k7>Ai0;f1PM8j>!ow%Yu6{JIr5eiz^FX9}A z3NxO()4Q|wDmW;o)_M0d{~)d6*FP}cT#UIP*$c&t#om50v_1FCoT(oF?tb}b$X_mJ#wZfiSQSr5K~Bj zR;O{PY__jrR7Q}UBEI{~mtGjckFDKmU>a)Am&daf^D!iikuw33{2%O>@5!=Wzu6U< zyKMYIwS{J&uqS*mhaHSKufwHcB+J-$PW~+AR4y zV2SXpml6Oc6cr#(f|-8?0z;CvC}<#8BN5@(myoLL6yyo=qWP>=Q;{dy$=ciF{vH6g zGzWg-h_ceVwfa_hP0ZNcy3S11F>Vdkwn820|>g3k~ z3e}uJ0O#+F1b;2mImo~tR-52t^4j8yu{2N^6TA%JrEtAwK8S4-=UvdjgxH&XNZ8ax z)Y(V{zzP zeHHK!lN@z*VEXnaL-^kBJCvY2(XX}4v8z`pcTv{7wPBmwQd*tOF>ZjS`p$*w9SU-K zC4{Rr4cE>aCSYn|+Sa*E`6Iv%uAo%xet(n>PI@{#nOk-g(|MimN{{;_>lrEz#iCw} zypg!&7EqGs?9n>huW*mztB>$ZsJjX)6ZRFzBAE(zZ3kU*69R09w}YQnhquGm-IO*a zNz&FXpm1H3NYYqpj3s_ z9iJ18m50UIt|157r}nvA0cr*Yon7tY#GVBo%DQZEg+de@Q>v7ddSUN}t@PF}Uc;LE zXe1IxpwhgiTzv08LVCSGfJ_Eb%UG)xM`TLXPVbYA)GWi7aMAZd&wtT#p-Hm!&6wi< z5B_GhQ7(I=zF|M4uhY60Avb+5Ze{PBpdn@FkEQ5l$+)w}>qlAbC3o?Umgry$mxWDubFPb~OFIP}Gj^`IKQta-FMaKajT5I2Rj zSIBGFSdz5S%-X!(6yH}Vg|5WcXsaa9Sx$4*wb*`u%$3B*K!;zmxg9B5-L8bbmAOy& zV6%Ywb{eLQ#?@8aR29k{M2lIRCp&k!Bbn3up1(Fc9WyhT5lf--i<^nyHj@1c+Q!T5k+4FS z($QLPtR1^<_Z+EH9p-@N9v{;Zq!2-O&+iku!-p+)4MpFyOE({yQAa@nhA1HLaxHMvr_`zGSl`v0BV(D_vWa}ajVcCE4h_ar3yFP{k zM1Bbh3gH!HMui!M3O&mvNPX53;@F5eD4h##_L{S(Tt~f#Mfs%*xOc#2d28To!YA5= z6x;sb)ertU(~#0VSOydG0%B+`VthlO4Y38tduj&{s+zw;PLGfKPU$6W!ae%8$u)|_ z1OJwj`{m%yl*8*oIDV&K-eH54W!}y}wIU1&b&$W*XTrCtLf$Ns*05Ub`53y{U`B4H zDyBPXsh+EQ`Af1g4z}Y$mIGbYn@xNMMQTo@rqq6(TWn^qF^XsPe0zr=wwO7t0Rz>R zvk$<fob2ba_5q={S32SkZKt%Hv;VIpjlTLPdv!3O!qW0l7xuzOB1k5peD4Bk&kiFN-V z?6&2;8j8Pc+u?`%cp`9*W$iV3tw+T2%+Lcks?E{A1PifOat`setlTP|8Tp zc(o_8feYXirIaZc*{Tr0&5Jc3FhAz_N1Y0po@fU;)I(1%o@gHg36hW&DpXAer=XnG zlcKI=fX0eN{$bFJ0FVG1zcUcFZHmBHm2@eThh4*`)8-E4x>f^NWiAB$A~R%-X<46V z8KzJbu6dIln92NNuDN8!G%0N{*3+)*WlmlKvs5G~1eN(m_6e)aac-tip2;-n6Ss90 z`IHyAS4H5E-asB0R(&Rmj2AquGI_~;5Mq$i8!>@W9U!8fw_CjHUe-0x^jx>54`J^x z%HWw{DsaK*3RanGYSh0B!yIAe)oZtEm|Jy=w+2dbovGao$_>s4JcvQmHcX-2=T50$LhMPv}QCbeJbRY z)m_Nc8S@}i!u8Z1Nl3biLoqYk-NEe%U$`;T)3d=!=8C9FHAN5oQ0y<$Cbp(!Ve%aaraCW17y!IcteU0<#ctBwoOB028UG8B&D8pDGTqj1`E+ zk?}b}_QuOrg0|CmMx%Z)2-%eNO8o(nl`n5EPmz2GRkQUY)(N^Ae3ij;-Mc4z)fU$m zRVu4eg+GriDWua-d<4d_oR>x{&5><370BY`@T1!5zM?>b;6Wr=dKUXf10?C`bw&Vg zL=J@isz`Wpk1*UcPKqvmf_k5rAAQ@B5`Gdy|4qpC+Kt6C8uElx?^tiDjX&aKkrua)sO5?w|U zM9nP{Jlg|ZvG>R|JJ0cd!B4f~Fpa#fa^^s}qdMnxkMjbJgfgH73b8vO+n^jOkd0_B zr|HGK-Q?$8+h_G-J7MQ2n}fa$1g3q)cg%q%5KSZdh%0G=HCfR5HJxsl8wFPWJVD1& zBWv_t?an-fQlvT|7H_xWC&q6Sb+NlqI#-d!n^vIJLxMG)Z2_l}W*w|n46F@b4O!Z? zGdP)%bb;Ojcm}A@b-zs#^0a094EW440+!5PJj}lL_1%ZL_%&^(ul^y=cO7+6oF_q0 zE6H!`CdWtgRQpWGw-`(A&lskwjrg$0;FVkennpk0IT+FNC=$<@eFeQ#%Q7E_SH$50 z8U=aFGIo6+l!dGzlmq@-CX3EUXnt8f4W1tu31s3F^a0Sx4K<3*NH^$21W$V&$lVUm zsm4W)ln8pa*dPq$JSn$v17~z8S6XpEMRKl~=c#H!wsHCF>(zPlnAEDfve$2G)M+{S zwnqO|;nFGNnmE_YiXw^>%T3q$5WI)Zw63XtIhDp@3U zW$wvJo#4$%rCv)vMz&I?Wy;gjp8g))22?PJ1bghML4CVeaY%4Nu`^J5K@^ z{F}z;@uN|JV`!)8sKdj^)B~|BbpSMLvlx;SM!W(_P*Ha4b$`bC>nhMzi-gvjCW05+ zfrFJ$<(DOSzcj8>Q8-^@U|hRMfC%HSmJV-LM2L440&2gf4l!wrXtlH6()t9ILdh3R zZHX7<@F3u+q2;@cLZr-Ry(TeV{fB`jSP)1$mo@;PVZFFxW}Bl3sAVnh-a>5|Yh#p8 z1Zhc3R$bjk?h*EPFo)7fIxl+^vVsrY*sue}%LDQGGz>-=p4S)yTrTtY7eT1Hh=_}K zCUU^sH$0Cg4;&on8`2QQp!eWtgnZQC!0fFPBJi7Td;YgzK~xNT6IJjkg`fjG~udemIzMN75~L;NO$>KNFWRGSAuDXicc(zOFO z8Pmd40-Gn$lF7?P_JPnA zU7Y}sEcou|PcStis-CLVMg#KWksPcCpV|t~@eQ;iKAU~Cap{zEb2C{X`J84Hy;r|v z5E=~Mu3d&Dktc)G3%5S2+Uzu+TRpt5TyK7l!9}7WBx@>>BJfI-C%$C<1XwD@W#p~bkMO2Bz5k6nKoqJ>sPDul zgmFhkfVA3+4})%-{}9>;1t*Y6$uxy#7K{=2(Z280>v_6fjAM{S;C1L$u-i@$7R0*Ud zNJ!q5BFtr&jTs-J-wtkZH;O{XU-PY^mzGb*!^-!0Jfr~(M6P^WjU~Y`bt^zYi@+fv zBa_}$i5m8=2ox>&`e-#`3!=g%Jjy5JlGhrARA~G@;*~# zZJ&sTF3Q8?$5?S^K`Y$*4QKm*B}{|?3XyevSBTFCzc%yEDqiGJQY0dUJ)(S#F4-y0u0%K;4~nFMS5c*m>b zbmAjRh`}5!$j|JwQ%79^!dl3m365o~%BIJVbFu`UHb&3tXJk-G(C6aJ?9Df#V;e|1 z6oBp6t0_Q;3$VP$1=O83@AVrgn5db>>k2aQTOnjUbQmhaDH6;KS>ls(}bg455KmXiJuQR)|0dS@d&ZN+BN(C~NCI!SKp7+?tGl`(gDsDbYE!-)_8KI^41j zAx-|Vv4wPUu7=?N1TF@MJL_&-E#r~;-ZZg=uIa~h)Uq4Uv2$Pzz+)(0VXe8uN?%ev zNhzLB&zer^9V^3)xDBPq0UthWk=g|6a9yxIfMwVU4NGzEuX*4Ml0{h`gNbLL7j9_f zT1vO*zGhiQjU^tUbEY_z92A4{!5qDYo9M?89d%Z=C{;19~|3CHqKpj8Y%wcTaN`iDzdkXqy9Gn!_q7l zz85ZfmM|$q3R|IGc@F|usaT2E$TyF*b+)9N=(l6)8p96^4EwY$AOEL0=9TUH6qaG5 z^?;bGQE`+4|F+)S+-Rtb5ye~WdY?3r%a`#WgvyTqS}Qx)_NpW3UKf0}GOslu^xRY) zR@Ns6L!k@ev31epUY&=veSn#BHpIuxTh*LZM5TAx|1%*)vbp;p8V`f;=zsRUc@l6p`<*t`&?8W^}N;lLPRH!VE!c28|AH>rq zp38ALTwnuvP!Nf1$wHx?J%sRtHj)-JNp$m*C<^*7sw{JG0Ty9!!sxQ2x_?8=h$q9W zqwS`Sfz3GwE-RinLh0fvg(cKfdCDOL)|(BpSM7m!2%=BE3rgeo z@4;YX@AiZY1h7_vgs-tAE50Zc7E|?OfhV#YtRh)Ob@Gj4KP#nN@3W`GSKe_HvUiS) z08E1rENC$?UDZqQBxuKR9JUvRz|ppy2&KXU3EMG0T}~KsAqku~#tgt)?R?$Bk0)X4 z6(Abc_%~{ZQjRXIGnwaacvr+4AyF*OO#ZW&1i~b;5Sy*#wT;qGdf1S|FUmRv(Tr(r zQyb|AEfG)F3A�g1OFbIeZx`dlft+>t=gFEj;2k4KqL4R}4t&jSzXE$fUF7oEgjB zh$C>+s6@FDk^u`gvkQchrt&0v`e~-zJ2iX&!^z(K)*#v!PdcG5zF%S}bHUphwy~li zR#Bfr4q634qSV5xyzeU<0{Q7rSzh z)k<2IF&Qkt+rghm^`a}_>yYHT?&N>K5w3sDq@WTvx&!Hf&zEu7Z5-aHN`22vXw^UT zYp;@_3O0bx?j|dOYE%F)50<_D*qZf2JbOU8f}AA&<;sRkbYmpPt`DAaVv&^SQT2xv z2J37@lRpE5F6fKVuwEYTFIajGf0l{*A?c>ohxWX>33}7>Ozz)h#FMmTY4DUVz%_Op znu}wfUhrmQcxHAbr%k{JXs^ut%#U`Ewl2zIvo*MV(*pF){Qct4!1@l5**LcE&B^bJ znr6{*EAcA5%yd`0HpMdzhcx>M8L}|STJnAOqL0J%^MSuB@Wph2MSicYr6mQfR6Ikv zs?vzXR%(FTb(igZv+cl&b0E`EAIxU%w*4h`4PX!#9cHX{?a{sS!5)+GL+%-39l>Rn zoq}LKeAQg&=1Z8Pk!;evQw)m+j#FqFL4z=_D1qqla08c1!QPSE!(3tqF{}Y>X8HcZdIi;Y)u^)f zUA~f;V)C#gIW+*~$a(d8?LiZhJ^HvwowcVdCnG|n#JGb;8OuH70u${4JV8#*M1J8X zk(q7#t+O+AzlQG0mZ~M@FOc?q9e^@vqv9u?$#LbE@!1f-cFRDn zz-KXu!U)6iSb6Kyo$scxc?dB4Z6c>O2H$>Sv(Rkqe#3ig&ihUwkZFwtD*1KeW+iud zOl?>?(i(R0rHnVMsAUxVh7Dy=qAP?~)J}RcNh3n$<7I1C)~YeLLlv)S!~V;d{?4TP zzDJ*s+K06hEipm+>0)L6&hhVA>XD}QO#B6UM=;Di{SOOaBx9jgSv6ZbZ}Fid>?A@{ zjTBod0eAU(PJ`UMoyM|tUi^D(+QS1WdZ#*Z0+Gr*@D1Ko+y;0`%GJxucC>8=SjyoZ zqc>k-UE%!}^cwUa^_ujxEi`?x#l@DJ>rCsnLtGUaH+N1DFqtl|LtSUuhbbL${h5Wp z!7J~w_Ck|;ONo~7agKcQtsWr!J|9Y-ok#Nomnv{Vr}*Fur0cV`ME&Z};K9KupL6+M zP@Q2KN09(NCVZWioL|8+zORFYa^zc*;PC{n*0!S&of>tyCS~W+e`S-FXW{s>mFP9}d z!RQclcU1UbN-KwK?@-JeQWHj`G%Bmn6+W1d!NoZeRghWV<=tWPbt0c3_7~(!i5*3@ z$WRBWy|8$>O0adkiG@YfpZziM|CZ`99g1nb4cu|0 z%zy#XVo8x>iB`jR2Id+>#Ct=rC-RJXESAYT$WX?_&wT#f$HF5>ost`!$R+tRJa`66 z_h3X?N06mSjV!F7y9c-jv_ShhkN;^cHXfa%Wbey?mOi7hv@bLI1fo}V{yd{IEk_6`$YWo0ivQ1kJKL~c@cspNha zy=R2cV-u1S$(O&~Srjc(YA&AC9!FcYEUiTLN%Ktt7AK$UT&W3eXm;6^gk8utO)hINUn!6+|=l z^vAJU&|)5s-$UxQqjsGIiD5EE4?3c?3%xIlkpXJkamibhKXkN8*#USf;%aAJ_D#7= ziIuObzx=8}B5NlH+?#ugq}^L$gZb*;wEUy;`gd7dODqNNLeReOk#z=ff9RjbPVyqk zbDYQAU4-pJ@j6pq!74fl5s!cxsY5ZTc=QAfK zF$I#mNAX6K-y+vEBwL7lYy9SAPef{{Zt=rfMjCn?(ebh!Mp_Yap=fNC-DgLbMmo{! z@{%*h%_il|oT$v%77K0dY;=ZTGW{bL zK1=^E9s7&O6r2Gq?DEGAOAtl!G6+6?E9$NL+5le<`}>(@&*drFe1c&kmro|fqCmAmj*$U1aQgp$!VBZ`qOoofWrMIQ8+1**OGbmrLG1pYu+W)baLO9`r7iSU z%+|#*xl+vCPL@g7*o4RxmgS1KQ{&X}9;W#xhP-Vq8i~eyMGn;E67TFwgk2t#vU4pHOhWI&qwKVpxY0P!hPk-WWN!8HH zSbn5`;3LyNc_xZU^Dz2yn;OP6(U!K!*~ygArhr(e>??VdnFFsyM$Ek#Vz(Sy;KEjZ zepYdjzrYR<$;4rXZwKeu1iZRi2z%CQsgqt=1i7D@B;5Pz!6tRL3*GEz(?kuG*1(+? znK_j_Yj%sP^!0B2QWLzPu#u> zaLEdOR)P=G`1McJ*>MmU4J4lF4mM$CM7gxA+rI&_MXtWX8X7+hW0s`s^21wf9GVZ zim+isP!M}E^Lpi>f`VbF%ROIGa(=ztNTksfxRdn!mX9r8sR}zL zObWVx0H9+)QK5)NSe$BJ!eo`o%MaWSdrevnbc#Qt+I1pSm&(p(k@SP0IeAO=ENw!+ z&_8lj<5eZ%MDfH1ZEmSy4v2@{sAxykm22DSFT95Yuu7eE+?M{5Y z|EN%d58-wOiBHGj_}ZOj2?%}JD|OshxdA#wT_Fs@+p8!Rb1CUL3(3)D6u(a@t_zY@c27QXvA*%S1YLl%)Z5z#a+-2e%C3i2 z=@qs0 z3O`nxPna{dof4a6N`9m-x=Q1Moj;z@f^W67G(^e>blSGG$_v4{2AH+{(gHh2Z{DGaUKoYc#I_C}ljb=DTl&NOZC|D zd2V!@K37XN$A4}>P)q$rKhSl1*q94#!yH_pEk_^mqdQT5Bd^1%ev(pWhdI|}V4(U7 z&c>ItaxMg6BcBSy4DH%VyTLCU;lkbco*E2yr5$6xasp^X{!EI)T>D~jpKHh6cd@G` zuh4m9)sc8;wJP{?AtQ3BXlE-KW*>~OJG<{#gVCnj5uCC*ZIQ5-C!SG|;)s#(7~Ir< z%CMN@hK^+8scqGV3NHJ5*c!HOa(A1_)`E7E>KKQZ%_~-pz;r93WT6uZEgY zBI(_z&NaQ2I0WMjmsN$~!3cIN8>x z)M(i*27R|J0JB)Un{hr^oy0vo-REdnUOE9lSAvZjJf1v97cn8LwwZjVqOc6|i1@;@ zqexuJ)F&z7OI%EjgpW}cKk~54dvy#ViT8w->s>Ot9KAIP;SGKJhOxbcmMpU4P;<~h z#zpTwIh8*|ZF|u14~Mw(Bo~>HK9R?U*2_7ZdD>nc{M2K|m9^be?N6)=xMOP;wO@W| z{Ge`R$??#-2zPU(EU6J+bSRd3mWG;GdGlePWuixA3eRNA3pg_{^cWI!irEk3}jaMZwt!GaWKI!sZl8H1p+=6MNW{UB2#2h@}0 z(+eHUx`3+u2EYmxB>>V6#@Vrw3wW>4ehy(V&Z&TIM8EY*wIDnl0USibmK7PHZTEMy zlt4`@TP~Mh-j-fZl}Srk4Tnz(J89by+ikXgkkXz6I%WyKN^&xb>}qdLTZ~r)dYIEL9ac9ep&j7 z3u|&G>{v|RLk$CJv5xIm&z6(4|9ok|aHqE-D{h~Uk!>Egf=S#2{bk-J6fO?`KTEf-y9Wo&z?%r_c}IyWUcL!lfx zHFq4SlswYyjdKkc#HdCoSKi>TTN-b57O0E4Faq(^Gb<4c^Tsk_`~TRT5?OTxzge_L z8z^daXFK{jy}@@nmnnFFvk3=bqczEVB8ho^3Yz1p_!s@UW059?UR*Bp>SM-uj>!jZ zqb@?4An0U*2x?4`v*;UMrBCbwMXXETXn3v4r|!B(%Qo7<;B^S_5CR^{xApzVhc`M+ zehhvebQ~)t^+WJ2*K}a%>ULxIj_wrYO*!7h_zVkjao($Z}0 z_%>($Mw&&th02})4Zl#d>!LlnYrzISd{r{LAzSkiX9qA6Be9k65q2APpgM9Dyn-}n9f3&!*P!oCprIb+ z$oAwwb|ZrQ;gDID+khSUC8F=z%$7Qp977k9p+Ee=svAbMgu%N9^x`|5(*+G3-gejQ zA$o^WSb__bdQ%Ul6a+A$rl}!7x*R_MJd}3sPJ&Qfx;*&o{M(?w^R$ylfm8X{P4Y-$ zL3Cl=3H#(t)*@2{r*quWp0M$58leHm04o&9GNG8Tgk*AhR^cXN@3r#-{7YBR(!N74 za*uX69(i=we(6lJpBQ>WUy2r%ESoF5i$sbi&$d?DQBug?n5YaSo98yz%7T!mQH~nZ zo_j6=$l*hZF@4;fX!y$`3)L_f=Fy`iRJ|0()|_ zxW(zDFmZhF@IKrOCkacVzU0|%yp~}Z^6OjuehKeub7l@YDTnH~^C%JC{M~8=G0as~@E9cFG^B?a=%3uos+2CqkYuI7hNP6UD~-$4gH4M4vC{vjYI*ARdCCM) zt9511hHJz&)$sIKYYu1G#U4fo(XI~oQ&4dC7TpxiTlCsx&$7fxEpI^4&hYU^#5bcW z-IA1(#|(h_j-b9B6_ZOIbPC@1^9+I$eGeSmf*JqH=G}x~8ItegZuv!W+1DD!px!CI zjhr#~^)Bdu+#GGQe;hCH`CSQ!*w3T6G=Iq92m&;faM#t#QHYV`>>5f!aVucK^|9B2 zPg2GcnA7mJ9k6SzZrS$8o+aK}IPxD2d-gDDgRu7h9A*ZvWy}Oz4M>iEhSJE`$jk|}8sES-_rLR<>X!kPZgz@^L>>j0Pbc|s{%g!Lc zEQW_)(KIynUG=UJI6QIi_~U^|z}GCknJ@o2)M)Oerd=!NPPTcQxq^FfIO@|Rwhl*4 zEs}L+)Rj9_>}HQfSmTzRqLBeO~--!50i}oawOD*9b-*KhbVr!bqgOnF-vb2HX`Vo~fmTzVsr8?E* zvw_kUy*+n8ej7)l zCyouWKde>_OKTI`U`Dm%js2z(9y${5nRlb=|Jy`-!D`NTxj;qltYELGb5TjK>T^Xy zH$>|(V64&m4ObEUn%zWOwAuGDat0#ZY6iHp@T2(HD)|+n;;i7ANDiKWi?jNz3eCMz z00I~3YEVX)qV@Tb+ z%I)!k>=thT$i`i1H!MD(hk6A{@K(1fn2)rk09AIVL>y&k&iYb90$z6E3sftm@PDIR zsG&8>j~p+}98aL(5@Dhh&K8IxeivH_5upa+NkKx21hx=7`z&>IXZ54TxZCkZdZQM< z`#l=dnlCqeDoqQ*wFNB_XDVh4^_*q@>B;}-D8X0lgLasng%!IGE1EG!@(tImaO6uj zy1JN42eYSEb96oMFu(pjZfPzwH^oYZ_ee|lJq?tm2{ds8uE%s?EwW7kG%z8ILmWEC z?33O9rB19EK6nCQSI19sKj_1e%s=6u^$U;i$ik?S-u7pY{UR=ZcQI+$Gl&7JmC8Pb zwjFIIEAO>2D|5GN&VXN$#wV#y*zb^E($=AGtw?J{=v~))DP&uOdQi*K?AKM|D)=m? z1ma+%BU}PfL<(DdEmW=$+O7qY5JOR@=)X&$^%aAa{$*1*CKxt9Kv3CcYooKmqV4p2 zG_%JNoe@Om(X>US=W&{XX|a<`zx)?rjF7u6g*{rcAalu01%dc$g-^hgAG}c;&KoX8 z^Ev^uQ2#*;Bz{dIMf7^`YEROmeC4^lW2Z&)*G7VhU%StkZ7RKU+XU@~Ug+2z3ql{j zLfZ0n$9q6zb5!tt^+@%vf4Z1{O13cl?)!RcyV9T*#V?XinKWV=Y{6U&i!8=hI?q9p ze`Xu~VQF43mlvS*3&#~)l3|r#O|&3AQrjN~Hye=zdk=4o7Wqi__;Mxd>N$1>&Ql22 zGWO1}1dy2NF&}@}z=CC>Hgc+zR>s1x<+u?-PC-|qo2RM;u$u7TLHKe8WgN&pzxl>?awD=FFD0TId2h==ZruG!a<{JotXUV#V!Zv5{Wjh0(lXa{l49 z8Cwvz2x(FsBD-seenb%s@Dm~|Nve8eOTjNL#`wwh#%MvjnkCARVKk22VkfyfniwEk zl)!%#tH|{WEyi+gw@`OA;feJ3#W8*OvKmns#Wi05CwL~LZS8=RRG^HavVA0XeBSRM zV4W1CIRqA7aIW6-NU;!HGsLTIQ%eK_j`3su40(bc@n31GTDFEL?W8A0liyF7xv(yW zz}~d!kD~WjjqD3d{>Ay~?M)2B_k0WHudKmsY1pZRPzr`6#EQD-32*#Xp3+ER&|8dx zU#?Z$UPEiF5p)s2Z1pFlEHotjUIf*Y4Scq6H402D&3RR}I5qZZp)>idTwrj}Y@fb^ z1&4kCp}0q~sxL(8Xgv*7p8xJt8NTDntKL2z8!n+3wV2ah`aH^; z*WFVSWwc;w(NP>4yBA!L8~I#+fdr!`3cuE#^9*LhRvn)Au|Ke|R1*L;@jId8Dl01= zR8Vr7ZQ%wX1$~LMjDMg|o8P;I*Z)auJt|{DxF82$d|rsRtdzFi+<&BRdC}mC^%O=p zvq}=ZDZ2c4>JpE1o;?w}7%w+9l**HNB6NG)7~hMMd1&smmpUeJIW)mcZP#>ZFHz1#>}f@%X<7A za6JX(0yAX-$-l<9TjMi4=^g@_KuavAu5X6OTO&qchp8>A$P>TMeF!X!-48WjgrXnR}QebUKd_KQ1GF0oJw8w`<;Tk8pc zh;c-nrt>(ot&+Y+HfF!&LU+3WVJhE-v`=~yqO|wy2b}Gp|3c7v;$Y;m?Db+_`r`Gs zw<9FXYwV_ar|x6Ae+tfZ;igdXv0tQTpWHm2SP z@eg*EC_%r`DrHRt1rVZ=)KL)%-wzQpaXEjj4-jA(jFOd%Vl(i2CCHI%>>WSsT@vOQ z+ISI>K=82#IA=M%P|HCw@Hh!BT1W)@02|FIwvU7zCR zO`km@HM@dKb;ivYGh+u%b3|6l-{{hK8e-K~bE!!~B#VzBiCu%8;%=J<#e4n%&RsiO zD=L5GWmV;0w>O4b=eixFt~CFHk9rD~oEhQ#26u2iA5>2_&(JZbI5th11j-l{gY>jY z7{fUJUGC#ln}Ns!WaBzGOokUOq1w@@TdM9*vzJ;VgHFGh3D9CwfH~Bg!V?4hDcf@K za7&e(8lqJBx7?d&{N&AulBSRO6^BXc=^C;f}H?j$B&7N(tu57AITRi+V1gk31+T( zywil9m1s98mBGhY2Tv*4Fn0gk(2u%})@8Se&8)nQ-c>$T_$zSjo&ZXl4a&GD-8AL) zQvG*fG;2x3x3_!9_U4Q?@EOcRNIhd{V<19ak-6S;z#DZ+CuGukPc#%nBI4ZUC?A1^ z_wNC%E4OI6kYz|EuQfIxN+y%r3T7fGiXX5ZWQ_f{rX=Aj|=ukH9&A~?fmAW z$MsVU>%Cv})xFE;aGEm{-vM0sf$J<3fEf1=dP8u%bcaq%YT66BktnlYZqkeK((BtC zBGst_@g|iorm#`I+J;efh>*^6^yzj}TG(4#3xMt81m?j_Qk5tqUehPkCATuD8{stpmwxm1OLgXMmT4(#-&DsvJxk2Gspds?nH{=Q>e%je`yGp3!2QoK-G7H1Jrco9N;H=rl?@g3>ZB=|Fjo)Rl4U@M!k$1q2~?kDrIzQA+IS@xal znA^c+!4pZ*oEn|?)P41UAt*U!Zvq4nNrZXf_r5(P7*;+%H#AE;LEvO8)<#uZD|W?K zc`~)F^Np2~@X*U$zP(TOM07Q|oryjvt^w@gk#IGV@1!WnWza76hl!%qz@zF4j4nDc3R~!_ zo^AArI(?}vy`G5Dx{N;%Y666mUpv@`ik}ryl>+av)*vF=&l~kj;qGmK8{#lN+Ifp;RU{-{PE80x5nmJzj`L%S7UL}%m3++As3Z7g9 z{xNLFVn!=%M4-BbQC;xuzKgH8xQ&l-rpHr;-Twg3i?sd}K;fxe+xNWuxBb}J=iFk*`?+)5h;lacr@Khx8&@^3Z$Y;8qLGNCU5^`>p!)l? z|B#jd+r}9-x7xdl=vISrb3)X{HSAO5*zk7!DT>0XR3%(@@Z?ooov-&OO|9sNrc}?m z-)*M~=oqktlX}AmxwFWREU=lO%j=JIj37IYI7ORW@Q}Ll-h_pG*QIxtX*TU5xLFxlr|*OanmH})~Hiu{lGG1ZCVyZrOUWNk}dH(>WFQe#jtIVFg3{z z-)zk6_y$HvnRqzH{ATT^Os+MG7cK&k!@j5N=E~bBgx188Q83Ox&F>O+CD#OtAg-Km zHrm{|Y=v0xBu(570C_@%zvl8}>fZ_6l@jPZZ1S5oheR9HsWtI!Z7jA>DuV^kUi-2Q zPc8(JJN;f?%J^{Y#{Ex>%&9H43teR79~}V6i}jY31#%Hiyzt0UgrJz?(x6KH-bqgo zI{sUMd$UO8%sT5-Nd?q|LUoR_Pa8CQQK~(R`u7WI>?mn3L2S;3BeRwS&8@8vRUEzn z`TDg>=Vwm9g_&I8c=ozHEt;#>Y9dWiSn2wG4G938!q`!VIrXIrEhe}+-MsvYI|QBS zDJMBsI+VPwJ0?-@H7SN5lg{d&dx31Q1N@$2+<_f5eZD#9d4m)(iD=u+t!@JMp%G%* zUd++-Ustm8<~paR_Q=mbgS6FQUYdu4ybh&7?Xj%I`q&3N$PA`gkQuncN{z|h{>TiT zA~(1kxhJ&Jkpv6=Y^He91~+7@T!~aE!#^Woz4|#q44~l@y{lbFp{|lToq^$N05qJ~_5|E;3WT(# zv{xlYz#+Y0-fB!#a$x@g&j;o<+>F_ZsVse@31;>69vw%&psi%`BWC!#wp^a7hXxF{ zY#cxr@+5)DH58`QUt@`ZrjFUEueguWEi%VtwlI}L`?JBb0n8H5h4yj*h5-x59XVJV zBJLHrB-CO?w0V{LjAGC4 z>=rj47646LZV>@!#Oo)Ves-Z9p$@uUS4;p$QWO>bw{s(m_HE8OtT((}{@=rONe7Ce z+ee|IWsyUp)*>-ZZW)Cm8{euX0^Y6n;b>7#BK&!%CLk6}3tnORj-&1bQ%_TFu zV(9Vm!vjxuA$bqo8_ZHgA#%!>>PGCG@JgEJ$D4~~p+{4(buX_qZ+X&dHN;Uska?A? zU1d7MqupsHqe@YTE<8|OLETd~Hxr@Xcu#c;b4CR}*Tp2_wX+O%8EIi?jITTTQY93P z6QR7H_3LSN@r$t<-6h|fKK15RRMBk9^9#%bfRN~wN$*=(!J0}Cs9BW$@IDf#7Gu^) zI8^!Pv@OGPRCt~Z3DPbu2L+#?Se7-y=Y;Yx>7+9Wu_J)Xk(JL1C5YIP^im)B;bpi9vBLOmsjh75fY#4y_uX|S|sLZp!ep&2VPX7jAf z7TN72P>Pi_Ma(q5OBfsmnRZ!$Qi+LUTo65Ft^8hl-UgA{QWU0&rdl(g1iJc9iF|!C z_K>NL{?yK$C|8UowzDWU>>;YciExwok<(^A9&YrbvT0*|%`&I7A`mP$vfYnkexp&08E_k*^Q<+!OHT!s(&fs7P->LPO zOnOouImmrf!MwV{#)`{QjjnK{oPr{vDgB4pjfaW^KGHE5G>Byq`4wga%YK#VP?ya*c zR7DF{n+XP|)_CIdwJXtuLue^v_V$X`GsY-{h_lnjSeHNB31>S@8jKw@PTHOp(xwxZ z7%xw~Hl@wiPAdSKxdUN$(fGZWxqV)C5SY_unT{`PYfFop8>2!7f}K#&VxDJF7XSE= z8~`8zO5-*z2En0F8to+DjP>YrCf5KH`W+8H7%fVBCf|T&x3e{ZotL!|lN4c!%damT z4bPO%)QK(x zvsesLgzTF!4XR;NCUzz&t%tmi0;6xpZ)w4;;n!ws&~nWRljZA}HFO)yP(|`(?4`N1 zlM^G(`F#f4)FMhF0Gt2Hm7v|2!N~w&pmn?(mG+#I$f;lzbiYylUNTS!lmJFvo>xzt zB`?}`D!GSTGwHQLdwAk6}}w`g^B%4fsz z91jKLeY_9Jn_9@5G7zERqQrBfsgnb(w=zPuDXo1LuzpwdvWmBq0rLbFi!>?duav3_ zLPKRi-PrqgF3qQuv{CkNY%`e#(f@H)YKfYc%+g$)iiuWzUv8t+lH=XjC?Q%?Px}- zTo&gBIqcK4i$!p;hPj);#y&1h?F0j^Mo9C;BA1_Rpp|x!3LL9U7ju7X{ipE!0!Z;O z9bgxh)NkvOr)aVW_|&>**~_@{H}=j$j0w@5gLd} z+G}sOIC7@40pSb3@)Vq}XVX?Sbs%P?F?dT53g+2)yIIOuJEv)-l2gP>$@Z95=u&!7 zRoAVqr9EVYgd12Y)eP+#XX*JFye9XcPQUx;QZdlkKz(qx9rf1F;O6#Y*@Q4-uYP7# z+?>trZ`DLDEBUu`6f2fawmB~Qzx;9bZjV5HiqGHpbuj9lC75QFs_=xO*gt)SP@?>^ z9%-E>Uwd^R-le#cAV9+MrftCMSu^?5Gn{j>J}Gz+<%ZV0aOb1a?nAF=ER*-5y#PIM zjz$7i^9taou zI;ZmNjcy<+lo?fymt$%MqEn5ziU?{tsw5FDe8%StJQ138amLO_`e$!X({r=qdvZ4Y zdDK-@fek#N?kJ??UcIpvU+VsA(+cBFAmms1o;x*aPG+ym>rc@p5;KQf%U55vjQNL7 zLEQ{qA*Nv4Kop3kWoW@?s#2>)8t;=3Wb7`xy34$izO4h=WIMv3#A@i^3ZQyl<^dG^ z9+V45bT#yydDhrk5ZGl8skqNffxUmK-91O@`=@D0(y z*jRwP{HtW0yALXHvbjTv1n)gN6xr^O|8?g~qB^UEK#;t9%G6(A`(4N4WTiBV4>p41 ze4lkCS6IJ@$pzcCY}8n69o82qddndFk>vOkKm)=NdUA@kw#$64Z}^4-Ae!L*%*Kmc zHR^u5O8<6!h*FBGs>-WP$)O{!iVVz=3=SecYAnYzN)$2@-wVRL8{Q70+CndG6$PMl z4Ddn3ON>i%wxveDq{&u$1C>Bm8&~r`r8a1^!6$ixc#HWY7RNJ1Y9&-+mf?QT3%h#q z{abBZ{AAD9cY4k6QO6Z&H1(K=5|H)TUyB|1s0}N3kbN(X_^cRQ%`(1s*E7HU>0ay7 z*IWn+6%in_`z&xR=#0pP+w0)A7uEqDJkFydtTS^~j$uJCYukPmkNwb{u+az3bJ}Y- zctNdL!VoFg^cN=0&w@`9a2ILsjb1Oz%nZZ0P$h7JT33!5>S~31S(K}>?=OP8j^iZk zyeL@+afrE(5j{qL|7yp(!}TW16gJ@d-Hxl&?k&&2n^)_|L9RQ=6-)rkvZL%eU9S5^ zv6>za&ovTf!G*$}G^zWocVv9EYm;oF#=K4GWQ!k%d!~`4V*Ai2^(NWDf#5)$T1(jS z_UaKaM7Cw6`ICtvix`6}lJ(0Bm*Mk0t1x@`s(SjE2Vt_ZPO5Y8}PQ)8(qCc_vQM#KGka zJpp!bDYgZ>y7*8nyYFUh;;X@P_lijJBDB=+b)`FI0$mW~SCpnxiwd@WJC{Yt2mm2s z`>SS4hBA&F00{$G3~6}bG74w{c_ zealAlwd}V!!6o)c5`jc8cNBbkR)(YG@fx*)5Ku8MiURU0)QqTZAE2bk7+8KK*p)u^ z7MoGD5c-bYzYwVn;B$%XspJn_chE`g3=yVpqw{Z3A6Aq5u^#($D^gd4N01C`#=W>} z1R>MUSFM1|qJ(HIH@2yUHT)-h&FJ*4?Ydz!@IrE78YNa8l;D-A(MFkZ*aRNs_DJYu z6EPlY05dP<1LYl1!N)-r-ukTc`pK$IaSP2vmsnU1cg;}>Y1mIxZ2>U=?I8SML`2(U zuM+1!fYi;tG}h@EikC2{N5G5VGY2T4>?FS}vIf^IjN=MCxfhT%TPgutuJzz@jKf1Y zqi+{i>Stkiu}SKyx1X2##m3?gebTYCdfu-i#KK{{Sks1=_!MUAS#)oIYk^%#dk!`U zPoy#AOpyu0GN31n*yHWWiq@sh4%i(-sS!Z9*=#U#%6xq5Ft@YODc$-Y)3Z1JTZ@H* zTpcC0lhI7SQrS`y(;vzlQNa62#-B%1sqP6HLg$(Ek(jQ6X4t1(5V=^Z zG&|cm;%6yXyuj+rM?|iP^=kEdd}f@H*WC4Bu@~%`ysgUJpL=QS^=C9jLGzdkFB4~} z_wk3Z80$(Q9eVzd_c9c8Jg){7#GsMo`DInR!y8YucN)|XmflOK#jMlEa`7S}0( zF&ttts=-(Z@d-JwW3TiY`Hn`6_RXdHpF%7aag#fBLnCM}*PI_;TC~b3mfJzVJn?CTL1^Oieo#Qm7g9}lE0A3?NN({z`)wp+Fp^u5T3E%ENc0b>IX^-Q{BNuDNUYfwd)-tm{{D( z-~fN8$NWj(4q85^zk3Cf(M);*&3=n78UYl>vouE$Xnbd9w;iD8Pd_JqUcT9h?+BO( zF3Tm-2SJ|IeWLgA$+;1NVaKmiS&@O}XS`tpCKLyM3tfgPXC(*c5Ewjay~=A19?CmD zype@K!7H$#IimY9S?=E8(UHBF(Zc-N+u%c7V#7eXv zFe4_#9mAr!YIG<-OdTm?kb^N%3E5iiQSnH>fmRrMS%2lsLex}9u)#Wfalw(|Y2B<+!5 zfAHeTz=)OiwjB+q%xG7&Y5Y?9ibGU?uJKI^PinOJr_rqpZ`}hJv!)KTHLmJarB9AX z7EUfWRec1Xm<(<&;J3BONhWrc)6yGj_M@)H=T2a$miWePC;vt^(YjzL_2=-Ll1sFF zA;^*~g2Ab-W6t%W{iRfqv9F@grX7HlOh+zAOTrPh&~ll zk;MCm?$Dj=GA?YDztUM-PWj#3WN6%TWp?hoAtPS>#0HvP@A$N*Rrr8%-59C+h{mpq zV8W`qLJvSQ5Ad_2tV zP{D-c!xReMBbmK8>m*7xB6Jt!ppEezn~Kxef{#uYkrUqM0LGUt-F3mCP?&JVHBhcH zph1w9YY3utJ5&A3{^-Kj8>hS0XZxKA=8n~}*@a35Ei{Wd_$DL>wM8AD1ctEc|4*UebEQS%^o-XjRv$vt34_*rKfb2GjBc`k{nho>< z1;0G_O^PurXvBd=bS9h*7*KWxgDi)>` zXDkAdp%bStU7S7WlMnBN@Z)w!DEa%rDR6hmq|)cSvuk?qS*D`jC~^k4kED0lm^01Q zKM!33kNu>+D}1glH7TjtO3%e2x zbnhWferLI0An+o^9dq|BXXFx#F>Nw?f&=(Dlu80Q3}d+8Qj{^5sFusBe1@a#VL6J{ z%KyZ7u;N$@`WdNBV_j4|i6UO{nzniG5UY!TJw3csH!S!?2QaJ1ie*ZCwm2ujf+6QX zId;UePvn?~M!t7DU-+jN@_hu{E!kOK#nd?@jZNn0b)?+R7Mse44b3E{{tZ;6S^eZS zdPXTaPc)R6ZMq)CN$G*!^1|h%YX<7tv&YB?6sE{OpWe_0kNDqu3eI&q^&G*w)rY+M z%zC?@;{C0ACXPj+R>C>pXg0JQUHun)i~A;DD+a20uO{T~1iz2h&;lR_+O zJrjZqq1XYmfkA$4k|m|8QnA}}GYCWJgURP^9<5(5^(Ygxe4P*L_tkhEM2n`IZ*sgf zo;z!x@gF`}RRUF_6~VsWJ=85#M)0V+tEWgTGw}{@QIP>tuQzmQ>AvLcSe;i3m32)zUbGcc_6cHjoyDs8UUS(X|A+ z?;P@CQgciq2PZ_D9~Z_mY?cSVuoeg&k*IsysYN#fS?X^vXOE}A;PFruHU2$0*vDM=lKA4EtVyEiINb>QkurR0PgZ<yI5T6m4ec>J4rZ$X7T|6~(-y@DaI`gU&ll8iIz;)m; zJj&5VI-qMkE&l`o&6Pi?No)9KFPGZ%(+$4Ue@UD&_Z@nKRm6k-T zBN7}4aIwi}Z>(yyA&-U@A-BBvqmR8bb{uK8GrhJ3b9ov#W=E22+ju~wOhgiD%ux_+ zsgW+wHTvB|M`-0YNB6VZ#l_E@^la%KKi|Qcj|)gUp>YKaJbOAz)w8}YnY+zhX*R$$ zI$J@q-1VSRn(Y%rv{`r6_|r2Bz^NZXG{!X~mAUZuML#iB4h*hl6c+<}Vse)m@hn@SoJG^5oK z^^?;n`AVt48tkUp3tT$9?bp(XljNVemq%f{v^7J!Gd1{fXiXfMwvPph=kB}h8WIWn z+?~<6&fQJZTeKUb#7mX{xq!SaRJLp#eG}PE#Q{A?+ZK=;F5)OKlZ5`*1CQjPI*IJo)z z;#gjv^*sHj?ra#xIjw6bFZ*_0;?c^Trh@h3!yL zY_A|~yw7a4`rBzNm>vy&sStjAiemMv)1i3j&SW8Dt?HYLO^e7}Zs- z>RuM}N@#W8S?^BR)ntQH8aq~S6(CNv|L;s|r@Ps7+=2#r*vr2BY!Q?30;B*AeY2k? zYhjMg;9IVLC#p`0ELalKVl;p+zEDHi=A|!H&8d z4skZGYX$gVup-i@5&k~1S7E+%Q-anEPb;f$v{)lia~)hh$(g$_vO-tO8Urmn zs4p%`xOkZRE6E!-Y-M$VoFAtip#TFMqLlX(nCI#U##9;?Pm@_P@^4=L!NyodRy-P`9Z!`$t&d891J@EP;0T@E1DL zjuWkf4wbX;fJA$3;Cj|C^scDi_Bu0CBYmQJc_|OFHX;H-0XSWDDYpe{=i_Z{0OAh5I(PX&aXAP#3F}d3%_dpu&6y+!8OClEG5n@n|Z2b-tgyy zV!Z9(>|%%S`*SCSD40Yug*dLm^N^b?vQ{2*5*QiS@G)p}m400yUBN-tNR5HRd=)&V#YOdOrl}pWcaUHBvX30}h|L zaCworN5~X*XTVNU={(8)ASDu;__`z3< zY@0|!X1ZtoBb_M@G%wGSU5_a5n8!_9oZ zpuwKDzfEYZm)ReH@_c?a`#2-aC({2$Q(@wesc-*@b^sbPc?Va%uk+ys#X-_Tqhg!h zeYws3>M*XLgy{xqec!!LfEK^W#A9L8qHzQlUS$(8uBco8fHAmND{d_MzUtt287D{_ zm+F5rRmJo9%Yh}NB5)5h;}G{4!VBKvs9G-e-U?*yK#4+0kvOsz!wi*G1#&awl}4BW!Kr7trNa5PlXocDYb~PN<6L?}2Qp zm4)1}xAUP4E}tEr2u9u~r-I$OIluE7NF;`3d1^IzQlR3_Zf~+m(f+I2a;oi-dOM zbGE@Xn&w021BFJ|nIyC-rwi%nQ~;`C5h+N#@e_?TCV%?3aT#4`9i?YJsT9$e z357x6VI>7lV4q`I7zkYzNn8gq980!mJ2avyYCiR6DPBYy-~_327bp9@bB%~73*_N^ z8-1zj_BDPkm!PkyDtBSfS$rRQ#K%fQntPk)N)jP%$Lu1z^UqoZ(^$xa$$I^ZoOmR_ z0^KxJ{Cl-Q>Z#n2-?!qE0Vkk$IjT zx!~2!Vc<3}yd?;O|MsSJn_l*T1SVT9H~0Zbz*YfBGq+c`tA>A75ky;a4(;QJE^{*~ z4QMyRAHm5n%fFeW!~s&NQTT$gmETQ;irPZcwC$Zl+|3WM&q^fGLI8;@7jXQ?%5pH9 zKL+TWg{I9K_2N-~JTXJkc$GcZr17ItZqD#zdRutfElesz6=QZTdo7+jqjK5g{}_VA zCkw`Wk)0HTI0(m&%pOKBZ!`p4sZd5?^8M-StY-GXH?=ihgS@(k(Gyl=KFQ(5I?l?7 z8WA5A1hDy(`o-G^TLY!X*GZmWr;e3V!1NQ4*R&=Dy)QmlyI$UAHw-RJwD^2(cs`hn zi~MqL*ZD{zW1FjQS*&pDfiBQk_t+Oin|31g|8#N()nhQ9tmqFkPj9UCx)ceu(b;m& zTqV4>gjs{Lj{~KAr$1d36RVG{#-q)8XoNj#Qlm)6GJxYJnMbQX=L|#oDK@QnB$ULyj zen_#`0}r{(ACwTsI4Yd_?am};BRFhzBLgiNSR#-*jssrm8Rby1k#$1HWhlDbNZXTv zUb4>;jBx5ZW3IH_A^3suE)cHT*@>VeFp`~}>S~9eEfU0W37p}(JL+Q^42^FAc7TaH z6St#g9s8GrvE%c{MQ=BNuD-reA;NjC6+1_JLS|5R297 z7}L(dlQZ-3TX$lH@^6EE1j)d9VK0tKuPgGyrlxHd-k#ETW7-E6Az!6Hk>gSnRyd1= zRy-U2XTsABdLmVQ-Alk`2~I%fRsPcZy!@Q%NzZ{g74LH9=JFe+Z4T3+9aXXrqpfyS)uW+cY0;xu> zncTj7_58W1ljpa2Kg#lSDC@o*Bl5Vdh2Lan4h8~aRMqQO(jYT7;UbNBK3DzO?R~cE z^@U$3$yMu$6V|K>@P%mDq#b#ABVVvYXM$c9?MOu^==J5ExM8-1%t+OriH60Az&CIu zvJ5HtZ!kxj^z3&A>n^%hwW7)qk}_-9(R(vFjwL24k;2G%wW-BT=0X{Chdo3V3_ub1 z_@c>c4o>TyHEEoIOCb!-b8^1nffcrdf%}?5{|3!;oelw^tnQ<%-^Ul6E2?9F(`?9e z{}e`Zv&5B!%wi9j$^hP@xL`2kF<9xE+f`jfR9awyvj$@rQ z#oQOqJ+3IXt8;?xNAMx|CMgjebQ(!ecZ5k1x!>UZVW)p+t2Ly4B7h|>=Rx6Y*^uOk!zE>Tb$%M7C0EG{2K^usk!Qso)(eY3JUVW1sq5?L z_iN2DaJJ^Yrff2&*Fo}`RyCDioirK%*V{_&k>rW0YQHrLMs;r!^mYJ$J=$cFx)_?# zi=wAu&C;@NamitaAC$VyT}r;Ai&0P&(qetjg2~e$P=UH5o#HkhoE4^kQ*ZgPDVeY1`EV(1eQZg>KS82rv_mYBoC0B&O#L$g%3 zkr_MnXrZ=mmcUqADB}oqW!=|a`luaf8_EJ?EhnGOhQ2yroO~83Ttj1%LMmFs!S@tM z&B`L0B4T9}3`aXMF=SG-800q98cxkXz&n!xJ>3sOBambV-fccTu_Ki(?_>o_PrP}M z1u+ebSUr7%XK8FnUpo?sVD@=Bu$u5To)oh0Te}cS%46{Zm? z4GxHK?9cRiKaoAYZ>BPHy|m7LO33CiRgWT^ny19=nrSbio)noWaA6p75sqwaQf^d5 zPf2vG{^#$BIkh}1F|aS>4(6h{BS;Qs4gFa7ZDTA+eyhssa7HZC{Bqzx9Zoc|W@>sx z{@gc@Dj&$+)%ifHKKRvPJOtnIb|Cu|Pah%L9N=Rvk&1j4ErP|rz=@H8mGZCYK9OULL(YRfA0+zb|Yxo9;yb!1a| zaqH1+N7X1w5xkF#KfGWA?6$`A(Uj=+o6WNrCkSg&zkd-Ds8acJ^;?t?uw>{fWSa>> zrH=82;*3-Win4GgoE%&!C=WT8!v<2zLuxK(_}aRTb1P702TY*(j{dqQ1g`cQFGJdN zJ3x&mXo4LlR~tu@tdWZWAH{aEX=n)n(tX8k0l|4&rGRXJUNj@Nh#=SLNkh_!;P#w8 z19~yl>iEcWof^hlJz5X($in$Z2bF6G-xMTAlABbn-;9BIzZIlndZ#2kb}6#cpY<)d zM9&}Rjt@nEhM5Z6{oP|H^1Nu4dWb>UPo2)q!P-NcxMPb& zVNVr}JC~di(4@V0mPA~|(0zQv(yO~AS7WKBWAoRo@?(Bm-|=%Uf;Y~caUZ#|YN6%f zJQko^BbZ3!trg|=GMsht&+Zk5`0?VU&a0;D84YX;oQ z@&sqslJ3!PgL6)ewH|Ehi^9tL1&;Dgiy?$84Q_0H(Dza-7M9=-@Jtif!;FmNn&mM{ zf_u1MX5|W8rzX9!TJbL2Wfyo`^{(q7#$xBz-W`ss8gFD)EJu!|Ho5#-GXLFG^l$7pE* zkrGGU;qKfkXG6wE$l}b(z)r74pKD2KECe)Q^EL=7>SAmQkn^DwK=tuHW0U&llWnJR z3bP~AQRM0mwaD)OhwlWva>1`o2S7+{8y?#`0s_9~q16!Ku3BaluHhscC!L&piS`cq z2eu-eDrTSoNAV(O&Sb+FO!pymZ4^cXxT~n2PI`AaC*0)QMWx`j`<1o0N-?EiM{J6 z9$%-MQPtk7=bKS<5Jb1gxLDHzso4+?3$XxAw^G_Dv9OY31j%&^n4&~yZ29-7uCN8e zGWjeR9YQVy$~H!xP+k1?-*(nt;2+CJ3zn#xkwV64ZQf-81|!fwnzFdoPV_r?y_nE$ z`x#EiBVoS(_X!_hI!fZ8#fh$mDy#)woKA)Z0A4IP{hkkn-WyTke`cS;k0~*PZ9vcU zUC_XJ$Pf$p67V6Fa8fX;byx@4~Myxnry( zZ}7ix-D;9jo5np9SQRU z>s;R7UIsH@#zj029~~

?zc^D!kAPi62M8j6+A~o@bpgR_GiPx=yl(D54;BjK@sY z$wydud$Mes8)8swt0I8k_?rm_9okO7>tgsw7Cd0#?J_6PP?VZn4-Tjq{In^wRDUJn z6$M5`#`|kpU6097NX;hZ%LIPDSegWVq zl6IdCg_xIRj|V?v?@{+mNz{k?5bjacqT*&?rU`y>*zu@`jT8YatBcS8{xeTP@#Ls& zaZZ}g@E2`H8O!xM5Fk+h!0cks*q6`T#0uDfui3zWtF3w+|3w5Ee5Ur*V1mW7g){%W&r4JtO&2rl_y2s~7vUGg<3A}s(JWvYtkQI|WLaJ+FU{A%$99Z-l>wJZn zX0fNNhT%2ltRk0XBHfKA2Y|L6OgxaF0u=H{iA8QUqP=Rl6r8j%kTF2u2}_(Z{zISA zqkI{cvEm0yUEOnWA)=m*%fv*{6srXhq)xEt@z{X=LZ#i3=G+6%s4tl5{Ldm9WvO0^ z#E7jwf9mx2T&Y`XK6l$<$5#V_v_HKVtymOqt~nr?p-QwR#q%ASfl@u~R|9x0)YN*% zO@7Z<#82P(n0cDiULbKDwWwu8s#KC>rVj3)HL{YN=;o?Rp_4gl4!t|4G0;6}>|kxs z9EetN8@b8@OINcUFNPJ?8V0*$g#lYg1|Aroh)F{ySI5A?)Ed?-N)1T!9Nyvs8qbQn z$L-kgoLVVp12aDIeW^ypXkXp>v8?!(t(R4a;@TT$)k)pG04YG$zw|@hAt=@S78(>8 zW)f7SrJ#ll1Jg8I$X@9@A%HGR@qcZ8@SE*E&Ux&kqai+QN7JR=%JLEHw5m3#%o1O% z(S01|G@f827ovOCFY^sj%qoAOsi@-nYZp-217I(#9E`N0j#py9IEuqO%v?>%A(0?O zXHQsLc{}`5D^DSRCsNMJuO#&$`g-VPfS;695In(}HA?QS(w&PfIvY7b6 z7Q7`t^WcXM7nr~~!)LbKB(6ROtE2)KzeM|tBb|dzxP|(X+YTWi zq-LYGP>)qNOzZf2+bu%L1me4(`fY<4DjR9fO-!@suajCdMa2bEDm z=JYZbNT^oFoqDx0R1ZYJgC8g$*YUG(#!=uk(GHOXBIhdw+YU?7-Z|=M#V0*`X-jeT z8u4s8E^oOLJf@Bhca{H)O#4E@n32Aql?smAL?j*!D)doBlfJd>l$%c`m*2vx&;hp?(8^SdA}U} zvMU2U@mrDTWg^!58M{g*owTAzFL|$GFYLM*hF`F$A~P~Y+y#*G-p@;{FM7lX!mL$2 ztp1}{U>k7wiw~GbhF|oV#N@yW7kYNjO+A2P;%Q5tl6tF)HUsomguzxKtwxL>rtBX`}zOxW575WDNl{ zlfMdKT?{|k->~kFao0cv1E?M=TQ}3R^m8xrSv~8!*~CslR2-7B*W<6wG{rYx$wa>Elv@j?(11h>}2@T7?2n0XupJe3yiWg2p98WGl z`zFKAik)o4w!pY-DLq*AtO#-k&&JA9CDuKl4TJAtQnW?@T`=ya_JuwQ%a&Y*B#C<8 zi$sxGaf2h|_}g3^A_?3_HT0?-mDH@R)C<*uxmFRMMJtra+;Xx-`r@%?fVcaj)(p`5 z;x{dH%H1UVM&0`+dv-)cW_8g1x;d-t0KxCN@RqSFwR9Y;bKV2hRxF*IQxDE92#1wC`ics7_kDg`S$bNDmPh;g zZqjMnDJgVb?(*rm1Oo!;h&Ubckrk%jdU=jN&o}4mVH^_qiM<251_o(ahqwNCmvVZ- z8W>=5G(2(`;C_>E{D#j$d;F(TbRG;ITKXC5%~!EtOWC-O0#c^+8Rw^)(fo# zTX#Eil4ji%F!9xn&oT-X*ck7v8}0c&`nfN+K4 zVEqh^NKO0gd$`S6!q>YVa;cDGyj7?T3vf&^-rEFCD@N!tY`h=8&7S;Xv8z6VGq*b3 zu#2=gnYrJ?m$8Z~#8iP6&#+0QRgwm^!RJV4t;@^BZx48NC&<^y;kbaSPFBs^KjGSG zjLzz5nftWnM_S7Tr-cwN0J#nuaUU#+mQ=Y4c*ybvx!aNlq6uT-tR~@v2D8bE>_;RD zw6@r4NF@8?pTQpzU_U#K8qfSDR+RJ6Y9fYKPEx-&Fc?GoE7v8*c(rXFw%6seyI?^5 zj*dvI+W|)bE=i;cHk!|MtO!kGc?JCpx9=7E|7}x|rECy<(1LZtvY1M5BO<%|T)O~h z2I*N03v_pyT92B>N=V7R)&%%7D`v#zGL@2YvvZtTP7iIBXv{HO6`v$}q)sT#;SCSi zB#6;hWDTcB^NTooi)*Ouh7(}T3QsMI5lN7h%pnWH>fV9OmU^PG8Mmqm58yKTiu55yeL z!IT?E?sfYJ3do!B&XG0Vf%RBPaEMRuIV%VYjf+qT+~Y9lw^epAbkEZE z`c(sI-|K8(a@O9%(6f@eNRs$`T!d7&Wgqa>)yaCz_GukJmRSnJ!gv^x?NEj$k=1A> zjrm`T<7FAxdKxGd2NBN31AP3*8#V-Gi{j?K%Ltg;4CoKA1{Kpj{$}9T7Z^|rO~yE# zF7L`sgjW5Ft3`#~Y8Q9g{xA3hDOtDO4yQbX#uv?FCT6(xaSu3@bfT8>wN|M+Jhb$vudkWe9FcyFo& z1u!5pT3NmU&aGu7UHub$IsFB|b{-Zb>WbNY3oq?072Ov1%8xO{j9 zU(a!FJ|78Afo5iOi&q~-b}H#JwT!`6Zr-)ORsPDGGwY)60i28Kq1LDP^!2nOU_wr{ zQm*7uUfgiMgy2=W9F3#~Ke~DeBB{R%=|5u=Rfv;Q;ez1D0c)I^bue!z{rjwI7eNqM zlvCcrJACkfI%m&7z5i;eyVf{~QqVr9d<^ zOXW6gd;_C&3@MFLf)}gQ1j49;&6P8NG6`xzHFf7h3grF zJ+Bedh890_GLKX2_k-C(d1d0NaA@cJn+ zjp|xC*IeEmGOM}|Yu2&{mYKb;2d)uWyhIIx5ffD)UOGp_hRhpf4DSxun=1?>9pzh5 z2NFsw1cc)0zo2|kLVAK}BX|Y&=$&ae7Vrd57zJ&-^6eN#7kBX-Wz~2$d=I@+*i}_% z8~7ha^MdvRN_)x@n>CCH!2vXwUD8tpYOsgKJ&$(4;4X7ZG(Mw5d;$u$WW#Z>UWe9b zZ2SX0UdK}KA3ZE{K6-A^Zl>?h@9z{K#ka zw`-*@Blzk*kCD94Nl4;wSh-P==&K1tJ#Y@IaOi^vfDq}}B4T%}qOyA|o5cIiD=2L@ zQ+nI#2B#N@aA~eJ%>@qiZ{B~rbanf+W5fvl=Tqs)aTv;j_0+%Ev|JqMMS^&Z-Z()= z;5_N|r7{_V@JnCU)X9wgcf<;JUZQd?hDhhNQvNj^P=ZvX>4&kJN+%ca#DGwHGLZ{bc=rd!rjxf5rJ_2MBz5=7wFXOtxWPATw zCgA}FfD}vpW`L;JQk+@`M&}^9TE;XIe)ub++Q=_f#HE*XMfkts^NNoBL|J{wNi|6$ z8RWY*KvoCi+U=mNNl|z=@v4XKfW>0d#%f%pFcL$T#Wz~)73BwnR^YcyY45%HfX3bY zxrc|!?cvu?rZ(pYZ{1&kg^#P!$vhpV#U zEA03Y(VbilK49m>w~ar;yiqDSy^)$Pa4qRF!Ih_^eH#_A3*!n48c`zS0?4}lRCu0F zX@XEX>bOAu`W`i$u{&1w;8lS;`TU5AI;Fty-x*9k&`8G+8k)2OawSqZOvi4pB zWx1S9Z^EEc1qoHCzE%yfY!TW*qPJJsNN92^XWi4J7Hg>jg?wXvJv&AXb1MVKCvQSV zSbb|YRUJfqB#EBT=SaVAb4@@CQGLCH=gFoHx~kQD0yU60}^(vgnSNkj$dg+kVD9u*i9Izg-?~atW>3CDXexzCv5fm7a-U7tl4WXE@np}5QvZbLGL@&g|ifTQPHP% zeWfSk+T4O)ABkLDQJx23W7lVYnlMwN5#xc{s2jc*OT>^#o;UsHa1)bWC$` zo}sU%J1!4JoPI;q6|q9%Xa0j&O2ShaEW*7*QTg>#YkbbBqM27db0Uq=M^24w2W?%a})_CYp zR2IE*T~w?k?uz1JYIr4;tH+(%R0~b$Ch|Q(wL&0=o9j-tsNeRGS%Oy}ta``vA?ud* zPp@;bs^r6Y8u7C>lrECIDSD#W*7$ya|7>*jGx(_U$I6hY{$+3EA*rz4%Ei@RkyS-u zcoy(My;%EiLNr8gGz7xi6gHGqfPb5oTK*v2^O~cX=vL`joxoOa(2r=!_g%`7=6eOf_;p)h!F>Da-9!D0MEUWB%Zu~)0p;B_M?^Qavdv1d!8PqKLqn`J=NR(s|h)d9jdb;KtuDqIhwILK~=GsA^`FFP=)hMSL=aUn*|U$v|_ z!S-8|TM`mxC1-u=Frv8{<@B{4?|6I$u8PTr1*V+JcRU-;vsdu7R82iqDm|+q&9Z3Z ziIUTk!y@nm<@q`RrPSdetILYQ!gxO@4F$N*g;?eYh?gCd-NBf#QtL@)EA4wX*6Si$!t zw_{YUvHHp#4+Q+88;4QtSy$HG2I9WkHEpT$FA?3#=UC@!2WS~Mh+QAa2y%XNEp|8T zhGMIOU!kX8l$~U%m@^6QayRUE^syI=2_7r;W)TNc$==Mc(Z7U?JniJERTT67p*g)K zlq)~EZjNJLPa8D(E7Kk-x~bJdO(7FJ*2TwdF^Ni+(oq;u$_AEo(Y<@+&kn&Zv9-dW zHCDFAg1u)P7R}Yi-OMsNu7!y%<0(}d(L|KGdK&sCBm(VZ=?ZtDBrQ2870oZl5-iSA zp@$L0`0Ahb?Wav1bnIul1w}&<2r>`fN3oKZy9dkqR^%ySkLd^C^AEajQuG7SVtJi2 zu=l<*Pr?41wR<%|n7Z|j;%5sUu4&;I%b7-+EFisW9y1Qj4$^^MuMRdFS4ABh5LoJ; z%~Ax27>;;RW=7PFA2lYIsnCIP<08<$L?5i%VCrrTgZH8a+JwZ|BU@9-P_PtdiJN-a z>qIt<+_ZNgvHJOL5S(32Q)Kj;_X{?wrb(kWB~Nq=pi@G{$JA-h`sFS7@JuvyM|OTY zHUw?svJ&%@Xr;-~5&pDnJo4cPKsm$Ng={jPBSuBMxV}yppsuiu9UAf= z4-#I(R$mEi*xK@XJTe-lc-9q80b!_=@oZB0;8m$zTw{2?SNI!!1#- zMg=VDRz`vH^RJD<`jqsZl{&QghhGrI)Ik=B31z*1ZH^D<8nC_QzF#6|onY9+e_1<2 zJ2+P8leTT-;o+fXkyuGjA`aKo_g4_!x8H(kTmMr*Z?)TEd< zHZ!(xpyrs{kJ9sru1wKmW5|gMmwnQ!XCZZgi6NlXM0C{mo9-&d%C9>JvqYMdzTQEX zIHgZ&REU#~W88ZLpqxSGKgw6?sz~-na2KolL6C~%RHEB!S11dP+T8sr$Gq+b=)ZR* z^ldZK2PFyOzeK{QVrS^wkHeb5kz;T;-PNU*TJ|Khyfi8D@1a_%3zkz-7BkxFGk>Ut;!o$@erTRk)JA`t| zV4#M}ahr z_twY-&}!-diw9f*m+r5})2^aBPO0(DM$)0vg8}72di38xwP;dH)x7B=PB70TKO{D8 z5*GM}Lt$+TI|FK@CVY1-i$%yAubk(7Qn~@eCAPa^LexeR4#!J-C`-haOD|SbQpb~o zWgtqTxc*7`BR{1q_0j(87gI?4yuKT0n_ZxIVr%~wR=u9dW z1T+e{!pwm+qztiOlenH^a~r4R;U6=)PZ{}gfRk(eZ!(Z)WgI6{Zq9N-@kC@GXEazd zvZFNjz+hYvxU){`?Z#^=zIW`7NHa0HL*WK-v1auj^fLNRWpa_8Mx#H%K?Dy- zv@V*M!LTlIw5Ob+lMlMfJd8pa))ZVIZiW_QvsZcBkWP!NMIC7}1W*G`_N-Z_2Nxxm zoSfJc1;c;lsDBgz7ruoctBVURo>xvNtRxs^)nqn_bR|O#iKA#<#cp1Fidx7yLksB= zzg4y(`GsqQF#y-}3r)^cnw3JK@jB@A*ZI_ivRsvKk-FOdnc%Q{NGHXu-6ocnh?`}?&$g&L(~=&4IbA&VP|Zkcu_Y$ys6lqNQU zeA}gigGibl=4~N_p94_WeTB_Ktp^PPt69bp+DNW9l&w-!=Ig>qWWm?@EF4UH%&X;< z?rMl26ATxwzE+TXyJFVB+fcHl$O9uC!7>!sPM>xI9};P%eJGeyg#D$*Is)CyJCxem z6w$Euej&qziy}Xhe0oQzQi2Oa{fct-US7TAY+2YlTs|!<^7Vj=k;QDgw?I&hHXNaq zCZR%Mws7hh>pEDhU^gVI4w$=qG00RVUR8|1jDIlL<{)J){*#G9SMx{l>G5p+(iDB1 zU$DzhEo9K_qxV5`j`042ty~dfI-nyzWJ9-F=Bs#$)11idHL-Pkzi9OhQ!aJbJ%Yvv z<9141_&3jr=CW0pOogv9t3BUDrIYkrqV>a$OY1LU8e}JrO-a1m@T&K;dt0=-aQRW? zedVGopXBq<&+m3z9|?!$?5=rz7Tpe#GQoa*`l9%q1Cg&VyTArw^jT|XBK6|+A7&k; zZaJXz5hi+lfJGE+P$ZjLbp==2V4g7=A2Q6ykA2oNlua#UgO<{LYLCmMe+S-so7<6P zSNj&1K_xvG6aPT7)ty-|>cw{cu$NM6{M=fPtH@|+9F3A3koCtS_X5;&mm$LgLPy4D`-#p?gReIw|J%?^u3+TT7S2^d{d*r_C#dW`9u9QjCc_WGYI1kW^7BDeW(BzKP zbu-^w+p2d@RcU8wJeKH|Gfrpy)AHNhFXiLVneO(uyKpbA%TWfoQR}FxlLfrQ@0s)+ z-HC?Hj-_YwBB+_}8pU*iuDONOm7KFMeMQt1Y=Q5=ybG!Ry6vJ_qzz(k!#66^1ny}qI!S=YG)K4G;dT;j${u; zA9nUGJ~m^}Hk%8RQ=6B>kF0ZWG?!J)UJed8ggHF^#A__F+=7CAYRZ z@)0YN6mU1ZQ`=ztTC2!KMx*k%AXOs*st3gn&P<=NSKG=TCG_S@RprbO(z}|I00{v+4Rqp$* zt2Jj6e2&AuU6mDPR8!H4I4pR;Q7@I?#yZ8Mo2cI*8!C9_CuM(K>`}sreTP6fkQP>t z46_<(@cbN}U?B^#>AwI2D)jmkv8^vtztQy@Qn|p7GTt`)J3RSOy zPCbw>XoD>~e8^e44mFd0oK47`j@Kr#4zMAcU#@#5uawg}Vr9kq!q9V#?C6aQ!jJ7f zB1tI`8eGWRkkZ5VNvy&29+$h{sNpUiz@vO6Tw!U$ClorgA0+c~ zG|Q)y)eyc4W+cE~!R<_8Yf030v4yzXVoh@d3x~DcHpWx8EDSWw6y|^&nybRcWlT-8 zuY~Hw;^z=X9koh4mm6(AiNvK_l6vHO>TtUOKBT6;a zGBuWhI?d;Az?N3_=@mNtnpZFeQSD$;7^bYvJuf|l!98$6C!p~&8cNuTy?n4-!;EQc zU%BRCWf{>cA=giPVSdd?@fa)_s~?b0DNhzKaQ@3*?51lZ;s_) z$&=m3a#{NZ5snnXK@<=6p!L*!R51^hW3p14?f5I`FvZE^94dFa4MHVmO%!s$*e@rvPicBWDkb6B&buKY}PJ* zp2EdcvD;##?a%Dh%4uunE#%wQ^0MV+hV$)eLL*mydlQ9uFgaw#2VWEKh~U_5wdv zLP$H99o5ci0acfjnb=md{@0tM(Sa-~jOj{~<_|PPkJ&2y6eg)r4__ESbC7AQ z&Sqcfec(Kv0*R*|)${2$!0!hk5$~lbo>5 z^gpOFENB5QP$W=jv2iCg9vLfnX7ct8m;Hy0oV(Bm_0Az@K3tIe0!(MK(Asb<)uau9 z7|P|sU76i!TywI7sia)qgeN7rW5nK4$a0t*FA3h`!jmK014Lw1TPh zAWXuuxG5dnrY&oLLPX(l_z3_5({_^dsx!0xp-aFE$DTV_wIE+v2+g4TWK{dhq77N<$#V^D6Bc)GW(}kAI7pVDEU64m z44LDeLr@WI={j7SRW?!fy{CC>hFX8 zyaqK0eQr~TF$=}bl+8ZfBta|-K@3s_0?e7qS8YTFUudH9BNt1cNke|JvJzT&f%9gv zt3qIDeDOE0rQSFtN-Uzw)Bq-=o|U;AA&Hm1E zpy~EU7^-{+-O#Nx4HQ!Y`RzToND+MFB2#d_qPqf1DAuOAddBy=+%Tp0Ae>btM961{ z82T+Amc)SEa5a?u;vhkI@MCLD5F&9_Q^6})e|N2j_C^xC_6D#Dw`)WG{T&*a7#Y#$ zSFpg;G$=w;Is}BZGSnRjCYksuC;V*@i+6_+G>6>yQM~Y{-$0}Az}=UvR`#XEBI(@~ z4+Tsi1%fMI4XAc{q+%=RVlYi5yY5xVKHy)V5jLAroo@86e)c_{j{y^_RhV`$nQ`M| z#H~l#%s3P!{Z5Jd+w&qZ;TTIN&h~kmzTc`?%H^G;g;8@z5mqE{T)@zVmB2Cx6=++0TT(?? z!Cdj{6~LS~4`$yLpkdK2AfxL{vhY1iAcyu?MFdPHzU__W{%J4vLOI+B3ek-A`e&WKYcV0Mwqdd_4c&&x26}wbBtJI1d&#G0vNP)bHQdkV8 zSF@97O8&L4_{+>@n$LOBy|D>wWp{l~@c`ecDV3~1OEWkS6oN?H>BmJELNhXBOlbqc zb;Zmt(l!h?iKRmFN6+lqPOcV=9>jRhPq3?wtW?U~R#iq@avbHZ*gEXH5 zxplwkY$3Z`3Gfo&AlmR_&>PIIpydfE@H!mASKOS=NY9?t$M`nmibS|FC-j9+KO;A8 z1yZPZ?;%~c@&KtZ-89WOx{BfdScayvaz$6!zYG~4ivE!zK)P8LiL+1B?-YO!LoM^_Etw5Oi zaxIdjkYB3iGQEZvH$)9ye`im<@x>XV`*yO91xlrVe-e>8)2l5ileEEL@W0u|c}^la z&Lb$P3ye6hZ-IDtX^L|(jkf^J0|M)YEgO+(D^WylPV-g`Pda5H!Ueb}%11yRmP_O|D?&htOsQIx z?Cp{-@|qViod{(p`tT0|UfLX+BRYmjMvq1D5t18mrEN8v-t3dqeUyQ%fPx139x1+h7`44GLEp5p<#4%g+VpkTB;3%_1d2{+e8cz{%_dcNj6UV2t+e62!R^1_pib3yB!O!~$LK{e=Rwx&Cy>mS;PH2l zwq{&nD;{E_8V}{_sjgkzNLZm#eZs5;S6rfzsiIL83hT5Dmudu5MoWJGHBC*vE4eSG z80Tt**^bkv5Cj&Wtclxpdh5HOYxOEM6TQR58wkBb-5CBpxJe|~jiOzFkbrF-kXc@PPzFm1p3>{rQH~cf8EwTsQpoYonh4#-pAEJpn=h>B zAWU4@)T=NlB70D>6;7~s9_!kiEg6VwdOCTRF&Ir<@sdmJ9C9jfI5d;$H?We8lpUS( z7Nei|+EH&PV598cPtrOCdFjPB9YFe6888l~-hAdCCGVCB&~(0(>nj?-*mX<&3^+t8 z4lMG$g|f#Rhx8(bJSdd^pFo60jE^#Iw{Q=xF_+=RV7#1i?03;^_@bBe^Ds4mA?N-e zS;3Mqf8`#<;V}#k4iMe+B6H0)^MvM)xeNy#P*@a75}Ljb)<0{!lE`6Kq{}C7{#Zyj z^*eT+wfAuXOJ0szTd4*{J(7Z?1)E*k9!A9e=Fl;B+93ZJ>vvImCt_{zD3~q)NR4e! z4#U>y>|^<#%~%TFTn?%`PPfgMJNY;gV+#ycP(68h&ovvv;dmilH54z=l{bh&FT=SZ zJ--4p4<2jDbDr~0Ldn1PY?jv0_F1GY%1#xyJDt|?JhbgFR2G{;!gbTo6^cj3(`z^6 zeSZcsUSe4w%Uv*xH=0=lJ)NIG;j+P3fG4`_gt36VqUWrV%wWgFOxqDt%l?jSMq|$K z816}O2pj954fEdnR{P^eeQ{XSkdrp!qm6AA20%b|=>r6*iGnu6wKvxBa)Wu%OFG5U z)t+Y-riVMv{n)(3UQTIhwi9-3uH`5E<97UDbYe+7#^E1RfF;`>n28=#P-`np{h^T( z+{#birWZPCzytCZ0Zw4Vve3Ah3S${xBt$GB#;PFPuBcKEQdUmwvCSF^aJCOtwSktI z{d4^z=(^#wY#&{b+M@uzZWsXr4b7R zX%{Ze?7tp8C;bQZJUG?CiX{x;694x#fCZ3jDG$ulx7+Iqd-{}+OVr2?!0`~j;Q_&C z*2J*~*n=~1zTYfCsi7+rfj$OaL)>c$uj~~}n7P`cl3*k(St`PlAX+U%%I!LE)iJ%$ z>$Zn7W&J9TFuKSa4Xlf}o3i7TH2+CCGpq>jCLB z68z(armaM@nMURl$;AW(#GUE<-P=YdM#9VZ>*!ILGC4B=-ZydxRmK6)BLVZvGTgAL=pwoe$GKY-Jw0MlN<@zVur`PAboJ0i zv==m>^s=T5q%W$ZX6<0xy6|~cA@5E;GFlmNI2J3FAj3cjk9}XAhhnUa{&nc4!6UYu zWJ7L^GePyQ?9A0cMRVQvR?3LHmfj}RwZo#*o@-CpNSr75F1w>tWAbMh3tL8Y)lpF* zmH)7r@7MrE`G}q2^9aE{6Niv$qQJcpC!cA>U8y8O(M>;)mVkTYDI}StJzBck`7d!7 zx5<%XbxQzJo}3|_%pIX==6%4{(S&PmJ6F$L$gDN=+Z9xg$XHf%VFy^n7n*)HCU3=% z%B6SEB6d#TqF5sL`vv3%Za!U>x>Wh*<8RO;S9BUBfpyeJKHKVVXKKLiUn28qoyD6( z{FR8>d|rEv9j6pC8o^D~UciayZLZQ08rduDOmcatvi4qTFo|i;Zb_oZjC<7)7Wabt z;86j(xWbBN)g}iP4X_c{6j-7wV)EiYe^>S0YBiLcYaAYfE4?N-w+aS2&WR5s9KjL# zlR7YNmDtFfPlvyJ0J=h%Ehfo7-{!nxPX7v4A<=r=%;?5XRsslMD zM@LH7T-a3rVy4PaFg+eiY^Qehe#wF#WcD$WOIFAJ-&A~j*{fKQ8Xs-$^?%yKg`<;1 z1>cr^>-7bCr8zoBqAhAm@78z9y#<7hId zlA=j?VxP%Ak%EXfEc*S}?f#5!?mqLDucfFz+~qfCz#TfUural3Bd|B{i7B7uGgB&| z*{8Mn6W}D^YCCx)&#=U`gtP}{8^|OUKal!)mFQQ!Xuf6pHv_^s{9dJ#*6C7m$i#Xk zP7{=gpRkM9xCn(i3kMXy{F&aSBtcTPPzCQqkp@*(&04{7?&@JCOIN@i$O zq`FLw+`$9(Bb%n1HN|>+7y;I_+S$nK!8b$7rb`F3QdC%`pEI{FkbzH28mitUrFvbv z(t203=E8a?%sT)Q_NR&@_|=jw>u`i=P3o1)RO^9E;3&%~Uu5jg_{6^+93yn#2CScX zkA)2SH<@1|azxwoG!}-hZA{PK&qDFAtL>n%`WXZb%a?zH^wlzB zVmeY!nMAu`RWpdgd!k;1J+RbX0E+OT!r?xh(+z89=hOXRWfx#@`~Ni#VyO9K@=rp} zHur%h+Mhwg#7Gvnn|3ky0MH;YWreh!)?&QMO!Req(XsqVw}OA5G*nglyNCBWEck=s zV$jrXu%Hu?p`VB%SG33T++GMP7J3gb>)H+3HUa3h?fiYuolGRvbZD|%WTHBbG7YPuQo&#RarxBscm5v2?b2qAi z>r@dz1ffR9LC1=M`@C}up(!huosqsdWUc$6L$I{-OF-zgg5TQaIZVBmV+dq%Ps{8c ze`w`>55vQL^K&XQ+o0=N-Mv{-##b^;+njKbB9N0fy(K$&n($UPQA3Siibp}&HuDdcjteM40bX7(i3vt64rhcfM_i9WoX=>Rx z-iBR2|LnjB96}E=#q3?|kG%4xl5BX^g#UvRr-Y#Y+XjqShZW~nXv+144b*hUwzCm) z=%^tzOvqOwXAk}-`u%-$N0^24yl|08}27rDO1zq_JX2b$x#)|UJvJp-InKm_-PbgbYg6afY)1E0l zw23gFP4|c_$1LQcMS9g~f{wEXX8dCwd<+XYDX%J?=xHeK8`?m}$8@-4t{E`Z*T(Cj zOIL=vp(wN(D*x1BEUUgnm zn(o8cn5IKay>|gjgyp+)sp_*T22CfMDX-dUg?tXWCb+U(c9fBu6Nl5Ey_@GDdvOdL zB${@Qw3AwLi#BNEgW;B2A5~AC@1PKwDx^7a#&%u_PNK-+aq_qsCpLdCAr}(*06*(> zpSEp2G;WR&PCuEpC0SLyn1%W-vlP`0w4vi9dVp0he#2K-cTkngeXzlsED{l81?k+S z0@a1CZgN5EG^)uz<@Y#A4xfx+N1kzm+-ZH(Sj=P^^dlNf*JhuBwP@fdUG`=KAt`m! z*!WJS#xE7|Dtc+4C9;`GZLG_SC=E71IH&gB_<;>w5EpkKUcqYyVn2X0`(Sx)E1^C}U+6`&J( zegH7KO7*I{iATGG^gk#4pY~7)Qv=2B8O0I4TxdoU{g_H&DOEMJ;{i9rq*+>}oiSM&H z@ciW;&@S40VlFd*L_t0WrPu6y=FRM!ibfV`E}?BY16pEf{R50R&q zKJadyLOD5v^MPUOVAB{Skc03^uM0Woxf8%;{-iWb?6Z8idOd|tiwJr`av!b!E@=FN zIB1Pc{GlXr-QBMppG<^KTz#7n6^_wBf9q(%w@lH-dEtq!!d1w$Wun=ze3`olHCsx? zE?Z+PyeufOd>Y*-xH^|cz9x2!pt9Bo=Rl3GYxA?NO`#uGa+v)s@-K6-g^du00H*<~ zhk780S*ZT1s#OQ@Y4|e8%o?4I%7qb+NYdVi*?q_Itt4T@VDLFGU-XsWadivVNvYL(vxHP~f-Red?y~$1Q zrEh9XiddcbqHLrcFICg4-OJx)Dgqmc!V6H>eV9v|^)h#EGn3d^0&k#CsJ}N`= zb+{*j;-5zoUdYY*{;3{Txc#a!gP@n9red3bbG3^Nn{NYS4E-GX(x|;;y;bWLdM7@? zDvFNe_w}ig070d`Q0|l*-w=Jqx-q;bx1n=A zB`9uaF!CJGvpZU9Fgd^rq#HVTmi<8rPSnj)0-Dq4p*R3sK%>8gIL9WwfMkO6g(4kZ zHtsfD3+#`LjTnMW4{j~b=j0y%sXsHL=XY5oTH(lvfXSHjQl!de?) z0Eg5_Ka;Y)5=n@!Wy6%;E*9oZT7OIPG7NmWM#gjy39|EQEMs49vYRub_7Jv(O$3hH zg&tIX8I%ZmX~U>a%csDM;bm_AaimaBjH8@G6EAvnG4P8q2f(_f#8W-X>>cGk+NeK| zq+H(`z`ul>D&ht&HDpSj3BjRF<<7u~8~W%}+PnKzZg2HHY8HrZ7a$}`q#XaXx8I2U zYLlSU5r#J;OyiPBRX%M&Bws};4nsY)6NI3mGLY_S#hI~Z!|wJgLkYp`@Y=>uC{;)x z$y-&<>q!R_fmEhzw%?b9DQ}negd3#Q_;BmlwbdxHyVvL4S~oHa91Mnw=rO9>J{QDj z_0a;~SyY4kk`;cM%@6y)>MII6M;o~Prm#r-=nrp9BIAkScUMKorJ*S(Z$;p~5X{vp zcG^~4*@~zg+36G_E){e& zGt^#Y&H8U?GjFu~5zegrTWMgJe6LdC4uLGYS#jYHMP++6i*>%f-}iFrpVoAK1uI1t zbCP;-h1BoYcC&|+G-kdJlzPppC6nevPORSp_BW5x-mwr zSfqe@Gnk(>-%$}!=XnQXfSli#hRP~uEcl7j2+-H1Fvh90< zN~%N0o-EYE_w?=10uHd(j^Za8(X&7*i(Br=$>2bgWsy)Eexg}3LG zbVDbn-FIbf7BtVQsy@M^q_#j}ES8spcxZ&V~C8 zuw}h9Zh(R+Ik^UTpPoA_4NZzEpCj=HvfIyKmu%}U`P4rkzsQ?Tr+(xr3n@K)IuB*g4j zxK9;Q4r55);zBDC6Y#$TQep_nt?$mTV-akzi+zb`6u09?QQxg~zoG?;H(Qb(+9hU7 zH~tc5cK-l@O#5I)$jg9ZV7%n55NyeACMT!twtBU~YP108D7)8l#{K@|z#|}EWVI+O zRxn=NXI{L1&W^f0Hp#ZolxFDI#tO!!jb>MIm9aqt-KqUe{NUNo>8s@i|d; znWgosA8Tz4kUL|1L+2J~Dd-VQ7IxCWr#KFTDV+J05FbAYxmUBf5*ZiBIrs$C3I5$n z-Mu^eWFIs|1lwEfLhr~S=(LsHU=>bMxo}XgIo@c<(XHu~X?e*^nldQ3ke6hxVq3SO z>mDz`e%sFniPtv?{C(An8ibj8lZd)O3SCXYrMVu4Y_Avs%JZZVXt}b!rj-VC5u=D@Klmg8m3X zRLYPV!aho`qfyipcEV$x!DBh+1>lv82r>YTN|gB}P<%)#l$w3xn2+a0TvJhVHB*2e zi9{UnLH>}#qFtQf`5=mIz1lD4X&+^~swXbCo*T8hyFE993A@@dHiF5BJAK~gS*P=K ziEM0#QauQGWl7l0Mc}-1WVJ`I=w1wlHty&`Y>6$x6G8n=@Y)kX5Jb1P z+ZyVW(;uN<{pC>wP8yV5BWO_p(HOqZa%Uq&m`~60`&dYb?sOwAKuI{jm z(>J!>c6sQS#0&dJVQU0Tl(GqyANMss5{uBe0u(=?v(sb9do062G*@muj90`SFiU>* z+6u(UZ^EO_lXU$X=ayC!_!>V;UC@tUjB1K1%G!3aB?tBg?^AeLL&xwV;JO(O;~>8i83UC`JMgIjOA7!-CvaKyYX=B*^{P|M#WK>x z$JPDY7y~UdD+i~IkF$#7jeFd^X*EJf$)Sf>Tk>G7Mk$1{)v|tY(dn%zPlX9ZYAoDY zQK~nX7+1iQpNZ#*&Pm`l^(VCqiOe~LkQkvVJlLxVi7?YUw;C%H)9i$+^|M*Og=)s6 zzVc`aGCUttq-{9Gjje_e#J8|c^Xk$`tVY(!wWg_h6N3wU$3i$-B53>o&?E=x;V;O~ zV-J-dM8+Gi(ISsTj_eutIDNXYe|TT|Gn?O70U4qF+?2{13P~oD$72F(QbC;EWXo8R zm!gp+ce_S-*+_6HSU;ybA$-wrYu41xpy;*hkHWz?1HmyEk!$lN0l$gkR1KSGgS7rG zBtT-qI#Kr1btQ7!rVPk80Q2HYhM0oR{K+jgNdf;7p@ebiN*sc9<&|vsT$HG>q#rLz z0t+;!B}HG@1dIHHGY63O*XfkqeDGYR$s{pZwWORiiF5Zwem1uWhxOcge#>Vt zARibgx5xiWM1nJbthzNjFbDk!Z&F3=e+3F8D1i*|r09+PEC0mrGLtc?HNBvBP9YAY z`V@0*i86-qO4p%?M^t;v-abX0iSY|Dk=+5rL3w70yJdYcPQLq!YA*!!O7{1<*Q}2vP(EnbpP|$1a=<&? zjoTbZE@4vGWExc0;k2v7#>A|Z?*K+@3aahWHWiRZcSD9iDPa)WKQ!niuT=8ZCH;-r z`Jbe|c~#$*(oFH%?p(-P;}EgMyyN}F#02>)L5Yg)`IQgrwu0_ilq}thdbAG!#wLk3SeNLn<9(d4CI18<_+KRZdy1Vj_=3r}q1dFctQApC{lANnXAjfMH z^L(Lnj8RN|hxrP%PXD{g8L5zsR}{9gUaZ9U7=q*$AXG=i%%g>S>K<&TKI2~a)RJMe zQX{(+;jc8)0r8`Ck%*KMy!s{iGg83KYkyVjf9Kl*sNh+)X@GTmw7qn_kNMsbUL3pK z*ojz1Y-Q+<^KfpmFo>`c65d?>AVae4qU6soPs6xmt5%~iv8yIsoY~CqcYl-}n7=n_ zg=TKoi*pHW=NZ5lwo`r7YaAnaB5G{(B*bLdlHK0_19S4LwQ+bTap@ga75FlyNRQD~ zz|-}Q^o37oC3PnhDk8%SHfNwg*wC3u!}76242G;#SihvK06V@nulk^>_Nxf&Q!0Ws z^kv}^8=(PMnjLl24&p4Ur+C%iL7ILB;SK5|P%vLVM=-`{fB90gx7%3H0t-+(cMRE< zA3PyQF6tE+;O?+jv;R4O^5D%R$+_hFqefcEHVbUqMqZ?@svrNVC>g_b<;rZi%Wx`| zJSM3m(Yz`)G~kWpDm*CLAn^y)_WntYqq+S+@gIUJ{juQZt{^Zk7QIOsCCk*M7fwAK zeouqT>ni#1pIIx4cL8hs1cBEhsc7?0Y1-dJ`dbozX7}5faZh+%Cq1m-cbb60}lN5^tB6f-s;bEIz0lCdP)zO zvnf4Xv8P$jyr~G1-=*Blqh=}j@2i9>8yB|hHI?Wa0 za4V5OGE*J!(lVQ|3K$6gmba?JS3HJ#*fJ-Y#J8`U?JAfJV==5=v!A0NK3w07jx#_B z*iMT!6foV+Z+ceH=4YQ*m&!?uAbj?hHKKufJ^tcq34Q`HAYNvZvn1!?^RfKj+%NMc z+9rd(mtY+)LV1C+vz_}aIwngB-_-6cI-DQo1byiSjdCTEL_)mZ8jBk`1%5f%!EOil z+GfY)EAFeu3V{?Jx`1@?F!gxnD&32nTOfMbH3ZaFX9Z;NxNZ(1__m9R=+)}kpX@F1 z4hL$7%sT~@giYg2K&LoUrxISmX_noe5EPhu`1V%HqKLfq4Ht;FF8=v*r2q5=>I^eI=)|>w z?L&)5{a$%k^Vttg>zxl`N*0uJYOLFPlEwoMo}c%P!bPd!*Zdx4XejYWqyNz8^nBnr zg*k-6)N2`rnkbQ=7$$SEab{p{dG9jDkdWm@PGb+-tHU5;>;@3hVKcExtj~Ha}A?KJr4O{7uxeOXxZRd~1!Q1h_jKLw5S879QYo z)ZML8R2wJZ9u4PSKz*5YqW>Ll{7^L;Ilnx`8Z~Q)L`1k0`hNQ-K>06Cmt=iNgx=`T z-0z#1NGo8nMU00uM%co;{HSb4yg&D|veCdp^vJgb&>@@?z(8j%k5R#2*9MGk16H4? zNiw&4&n?~OMdCdj-GQg#*6T~$c=e#?EZ+l(0!*y0Ih=@M7>}e z1Y?nBAE~0;QLg-+s+*8OzAC|boK^JV(%Cw9pZrlvFX$ZTb3i9Bai{#2wfn4ZC%_E% zgr>=#fgRwn_sQLVv?)Ltn?(cRjWv@#0b5xdi0soUQ%RFdS5e+Zd$UEz4u|4pqS!k*Uj zeM1%Y%6SQ8MZ09&6JHbu9~s$+#pDSN)IY1JwkN4RF?TM6Rpg79RUhKfQH%V$D+Y7j z7FXW$i%#VcJYze)g&Vea4w|OS5eX>=QC1IGrtg(x=O?Q=4F@xetmNg@ESf84Lnd>P z{2ZJ2?@~X(r;wYZTxOWL@3UFjjaoqyF9j`;dUm@&He47-MJ@j)U5)VRN=6W-W$1$KdT0U| z=->R6j1_OB%@b}LnWNCIFEkY-NO^EsYi8bCy7S$(YO=iMQXM_r2!`jOi*|-(2kI)@ zFiH9v+~*WmSJ@Yc#rJ)`#~*Cz?%B<(M{76ijOF`xq6{j;0_PLtKM_&Be8lc#7u`A6 zK6~DL9t6PQ>0DWK%TN-sB`FhyRy7<58vw+W|9n)mh=9FJ?qiA!o|^YEf7{DPBx@Md zce9%zaMP-S>En~7qQd0C%mz>La|dN;!*U~(g_WmGJE4yZ1a&hVzl)K{P}*||Z*t;p zErMIkuSjus{UD2R{NpHM=(#;`r`cTv{*Nu$=x?T^=dS^^XP{ zW_0wVxDKBh^^3v!2?CMakq+x7F?@EK*w+VeX#Ay z0C>M(V=SX`$+@*#DI~gWntYD<=EkFApqK<6DUDH%;(zEY_eKK{Re1}+_U>F8syO~sh(TZ|uF zrF1H0`6}3EZv>C>`{R^H+|;xxr=!>{d|XAJcL|N1(ZjF3li1&;)>o8Jr`1(rE|vNI z_>BLGn%et(rajFK7Az|EBTVn=kdNcV<}_Merl}cKhBZBk(~qu-#0MvZ22k3r=eKH; zkQ^iL2`o%9dFZzO>@hgW4*P2#hoD`9P1tDr@UiB>9fmot49VB{lyye1742xn?=MRV5=xXU_0=_E*yp!>g z;UZYXH?lOfh45%WEQ7t0(V!hC_Pyijc6zlDiMaZLs8Y2K+qC3GQ^^0rr-7ydzN2Xb z?C^}7h*9M!=%r~B1D!#t2NwLsw9CPS9VDh%taa|-DTu%FL%@aMrH*xvqk^BkeAu}_ z+KU0kk1by3WuCK;R-<%ZQ9Z2XdD>6UN>g7UForrixM*&5h0rdGbX>B0)8$xlF|fqK z%*7gnHD_s1ojq&GeSN)D7yS_AN1@f-nbVYRDP=nQ471I|;Md5soovCl32oom$Ol3i zxSRh(Xhh9 zspbG~%S#AkkR}{KS3Ce@RTEVQ!q21Ilu}R}i=ZCUeT#;o3!`+(9+?iQ3)i{AV&^Or zP6_W5f2^BB>WCD5nvFjcs%084%OEY!;Q!HT960fNdxT93$chxW^;Yt)&NUcB8wBtC zrOm*IklxO?gW#J{Xyzq3pO-#(9k{27Os@(P{T}-F8~sEfTvDD|CL084DND9x+qpEB zj_rz_SpzaI$Tjn0H9$|Lr`oGHECQ{F@SW~pVnxAPX&jR_cun8VI!=#DXAxOa{o45= zdu%{$gdtTQi5L7FJ$Q8lP1iI$7b+n7nYMqu{W@Zwf~>ZWBzX$cKf7kI&*e@v@bB7b zkwR?aBsEg_oPZs^RROb4Zxa%yyB`6sCP<;7Yf!ZC@NPb*u6HqulG#oqs$TMHzv8kT z$Qn8}C>Ul2f=II>Ds^hBGCczTz|XNfYT4m;)+!M&Bgk^*CXphHCx?CZ!kJ&y41lpWGuDkYCF_Vwp5|P z!1>TpRPDMAW{?k4P$-o?KSF^`Cr1XNG8A+LuzmRR>&zQCt`g3a?lq9w%!X_ie>%$Q z21b5mI>z4w%|n2r<(fttCG8cB?3-j>%E~ZpT^|Pz@}D(3do+f!WxKh*%Z+b;k=^x~1+7JerwO&_d1@q2@v@u<8DF zm#(&4RG|P;uyX3OqcMGNP`ommy^DA-iRG6nvE(;jyvT93b4msb0Bee0-WObw#`D#Y zN%NIU%cf=p)qU`nzo?#|kqhNjjVvJN*E;RO+-h3vlK@B~N^`F^%R|189s1;_sH7aY zhuqMoUv?tDwG(htY8~)ffLWN^fI^+(rDx+DYjiSQ_`HCeZC9xYjl&R71FAlS*}`K?vy`>(4>U zmvv7HSigEJB_JS<#20~*ayN0H^~NY1+uEC+$K+99ie>7bYQ}TRloKO+zQ)xa^C$lJ zrdl?S$n>5`s!AVnun%i0`{UK8vuGWKk~)0}h^G@tZ|RDE4hF&f+lb-}`tb2K1SVg2yzz<*VTy}LmZPw;2pK@mKGqhbNHEn)bPBtf8*)hD6 zIvR7n?t&WJ8lq+NyF&dA1Mj473&FdWu!qN_dL-I+4E8d{jf*}~Bf@IJ`)|*AMKlrr z_N4Yu0+>8yC0L6&AIzHX>WGzoA@n7h7=`crXu0+EB$Oy32hjP7UBD``p-GJTHVe*Q z5%E_x9OgemVnV?1%?8{c$eu05ik-Ql8oYi9-Lg6}pf{6~hzB^^pf_<|^&ArxEO~!O zNG4(O&ObI2(Ap{Ss}3E=Ibz~#C!OA5=Z@2A+8#7|YCs2S0@bV#a{(MEljl^)CN)v4 z0CvTDYeA<(E7VenX9=;N1h%(7+N*o-t>0OuusjV6iyLw=c~~1^olb5?bN(?qNlLnB z^Ghc1_nT4*@>ENam6u}RmbR=m@{4V`+RYQ{CG85w@=xcx%Yo|&9s0#Yg+_;r{D>Db zYiSK>W@fvSM2PM9 zC?bMLs}~wp%p=Yl8d6yYgPA6_yks6iGx4?H-X_R6rmk!>5VH$t=hoBC)YX?jZ{Q1~ z5_M7(mI<1riRJxg*z?gGWU-OADnAbaYp*xUQhT~*x0pd?`22wL4ZTVMp=4C^lu;%2 zHB2t@!9pALyZ7SraGFf+AqZqaH6M~0#tXFSgB72Lavk={kaV)=YWMTmikbIU5u{KV z7#@B)+UVxt*QsB*CWPyRHv#q0?Q&WHCm~L#{Oj}1 zx17MxQIN>)Z7Ep(=!SF+0YQj?vNJ%R?V9aiR=t^wDpS#F%5ds&!^v=SF|dhPm#a_Og*0*+rq8yAC<2LVj2F@fVD&ztCn zImrd$-;9rpF3D0sKjuButuU!#lyX||fXPF#>tbl>20P_cIZv_nEF`UU9sXW@0UvHb zSBsnJbas|AQ#YW9^yt8pZEjxq7TkJxI=Rf8$=sq*tiv6Aq?!}7G{M!_->k1x*1!PF{gbR{yr zOT=iPq||sc0Jz&c$}Z)`k9?Wa1ng_E9qP};D~~S7Td>lTNA!PDJ$5@bRm0tm&K7PB zg7^>j@=6G$VRq8ge#>%H8Qo4q(?Lm=+BmeL{ucnEw60c@_M1JNv1*ZDJE5jvliU=X zQNVHGLN4hVSHk(sLG5F!G0RVd$hCIG^EaWABgPvUA zBKp0)$W}%xW5le|6URaV)~|X(+80K3T*Z3MGW6%wl^>C=SHN5M*7m*gsP|6cjvi+Z5))Nd?s-}(xf63O zCn#-kh{as~soL{YW!p~1T+bEI5&S;DqrYJceo@!B9!R#zC` ztBCO{U-ELiG`~RVgJL3TAq?agE(b;+Jy$k?z7c8<5fK}P#m)D$eA4o7OM=Mf{Z~mAnV-AGE z@WmF<%!y)rm@Th}5cSXIElk?XD=PW$dG`LE4}m$Q9;QR|bU5Gs2Eq}eqmYG%y$@S^ zDwc~J!&q?XH~dHt8ece7>&;bS&5R;ZAXK<<_*`QxUufm_F7d(XQI3LbG|MdnucmK-a{zfz zl{fVF#S^E~oXG7iHLOmxoDGOavor2`E=HP-nS4X>8iR*V;1S6Ji0x_FI2j9|p&SsC zyL+@-fKm`-!7=2`qsl%@{(rfY^ROXZ`sE=VX$y6{`<+L3y|4g__xjoogVI+|LsFb1 zjQ6W;37AE7A4{{q_>YutzjI6zPDMTNt~6P|3pv$RxGvuk69iwq>cD%}Hiq9t?|!v^ z>faye328TX+A&j%j$*u3TXo}$FB{iEcoPx@x;1^}d@)%ARawbx(e($^QL-G&a!jwP zA5_e9f#s}lfznzV412{_z#McV1|jBjvIKdI%i^cXtnZbS>Ll#LX2cbI?`QLFQNtWu zS4BGj?=%el!bza#v)*Hz6!#s#js$F&n#Z|mI3VwU9&Cd^e9uQjDUQ__4j~+15)v+b zP%C-j7h3w$?U==Wthe-0(fHM;4dq(7&epgNI^3|5w4q}t-Lq+X6?I}gLB&qT#Nx-q zU%dC9GQQfK`e5oRT7!qnMRcw87x5iw9qBeE^qb^+z>h*Q%nOjJ>1nQ9AX{GdoH;9U zX>T<`LV5OCusvf`Afg1cm}^n=U~lC0%4nXG2(4!P@#r72Mq|_)W;XDhwmcoMZ{7;D zgs9}81%1CtJ@3?MCE564vgYB}D;)3JO~>W10|hrsn&3GaHfnQ(|K|vG)qnWycE@M^ zIT?zCT+48LgwW9g-dd4U^-n&6sn3H=wT}c&4e@Ri2(vaS%eJ$TrWAi=AjXrMK`(#W z;nSk3w@rF)7O?aegRbUu?9TV&)BF8Phv2Lhwf&>>fyNnNggvkz7JOXp5(@#vV=LXX zXu2N7Xl63a1FT!i3r_cWVk)T3Uyq*?j{W^x?!ClV<@0x+KjZYeY}aWScX6bbOgg9L z-1bZ8C)&}%<`b4ZH~;CTS=T%n74a-4(qkoXn?J~0Ga8YEF^uFBS!%jjG*%fYv&Uf# zk{hihB%rpqT7ZUN0FBxb#VY;4^}K3?q-@tXA*|R@>aRo8Of*GxWXpxh?agK+8=s6* z2^tW$3lv=u5nfT~imaaX18-;lV-5RCBYja-UUk8bAbw8U!eL!Yit^4`n5qv0uQdGB z*1dE`R8!;ShckKNjqX+ zVh4?&MEzp7>785FcIA7y!}d!Fwu8Kfee2EF{E>qB1{9^qcz8 zV-0;1u93Bm{7 zP+9NbNJZ;UT6RAI@UoC+@i{XP1GBTpqhTF*lBN!AhcZ++ya)q(0x0_h&%Q6=v7~)y zY-3n_AqPDG;{cljHQe8_!Kb2Q^xtwpw^dY9K2P34Lc6%;MlEt zTey;{TeI{?PG`5#>?^{WdTovO*W%EXocUrx2leR@_K6BdQb_EbhLRbPs`uimZBBPC zf9<>(3ZCQOI(vR+u91pADbpWe$VJPrW79k7nlR5C9lqhpAUWp?_@e>ej!AU7tg4W# z^|DGTZ65JZH6uUp#b7S|T9Ofz(A1{oSkl8~Td8U+ke{{B85(F@t@SzqhP3lImX~&C zP@*Ucj-$}DG7?XFYSXxcN2ZphW&)m}&woq%!;<}EA70-XZBP+Om2IIsdTHUGn4f*9 zCnJoFNPZWieFG@YL!}s1Jo$DV2kJX*3L;4TUswYU*XNxVRaS6CsB&Bkcs`s0P#w#( z0>L@xb@0`_aXuorn6&WpU}U9@OKoi6f+o~JrjK+|Mec<2oU3T7^4f|f_t^AG~Qyiea=vUpuF`Lx~-amo2L z0v?(Gh!3H-AV?=b?JqzP2!^b;a=znZ5nysA7S!r-40Wxe9{1ym+^;}!Gy-d;TE;1F z=B3DwIR|9y$Md2%nHOVc2^4p@92{{Dx=+u1&wMW`Z`x(v8wpMtB$3$saUR1~D9png zL=t>{xWK9;^5q_A(LL%2bFKE)t}r4#wI~;`xYTicdk;w2l;hv89I|L5AL283%!ao5 z5y{ug(K-G0VGpU0jU z?Kq3K@5HDB>w!``GLc*x*}pFTzP2+fi0z@*B4<%W`_QZCpOhO!k#DAf!8(~rq+@hz z3zc6hru5IE0q90VVItp*M=mEII{LFv#+aH6DiCo0b4Zwz0L1}qq51}f?1zg@-JX;u z!4F{6Yi|I0spnb$zFsiVoc1W9wPJ?R31TLTK-8*zUAO?Vv5pTBNPerTWCYSYcxGXs ziw`A>_ATp*pt51a;FmZJWW5e=k8h@3M?v8tGno#UMKb5DRn`D<27y3N~?SdX~(LGOV-u!Q5q{;wJfFC4D zlyHdLLx$;X_;aEhf1{X~h$xbu8Kd7y9t>EzbHea~yk)OA4G%lbruemrve=GIiFJkG zrtTFo$qtygLPno7ohb?z8MnH%HoPkqEVH6K(o2PNH3k)d_TQ3XLI0WCwM*krp|@2% z!J{Z3dbtKgLgm2*^oGxs&AGY3u=;bZsS&fQg*Kz9mZ`5kmjWGEG%2r}Z~y2(d?g`) ztK_jTE!pHKtqFp))Tf_JPyI>(2!`}IWEU}=~GUF zZ)8n#Xz80Urd^NfH<1?3@xdgf^AI5kM9QL`DtNW=Mq8e9l;3Kx`1B4z_*QBrd&K@4 zB$vU5oNbP4KnfeHLd^m`KY(fM+XScy1Akx8x;yuUJE(sdj;(WY3C6*#XHX7Am%xmP z5woCwbU$UR4$b6pqT;;J6!}qDx=Z6eq{LvB8hYu?$f3?>qvhkPadN4Pm7UEM7Cp9p z|MRdQuJz6hD5Bi>lsBYAV8Yc2D=*@4Q%AhVSUwMYLb4q|aL z@Xi4{K9f@c5SfL)f$-MIsKcP?I~ZoA+(E;I!M39+V%XaL=Y9Ka8<3Js=$VMpN_-A0 zpfEPi(=VzW$j6`MiQh++G0`(;NzH_TzOpMp)068-^wnz-h2X0XM9FFv4JycZf;qbROd!?53hcwm;?09at1UNRN>t@47qDH_`mvEp0+NbkYvx%~WT8 z+2qAfqlswY-bv}w&vH5PDn27p;2LTMlpVfxo&Xc{t2oCo%4`Uk#G4(~3$-4+;`NM5 zn@X&MZWj^a! z+669@VGANR+zq2(J-UZ~E_BfxF??9VzksEb}UKaFsWL=vkjE<2TLPXwLjncYg^m)Qw8|@}8*GDYmWi2N8XP$KBn1Muo zbTl|;20=$ue&fNcH%1~@;q&at%K|*8T7Zt}s8|5#l~r!d`g0f~sWP+ud>OWftR+FH zzMgsvQKR<-*i4Q>Zs3}k-z1zxaZTLSP1QW0xFAn5gcK*abTQfTBzaKYI{Ck>xIza7 zGdKhIk48SFH$dM_^9H1oX^lV{&2S>XQf;bUcp{We^*SuMh5tIB~Wc;tls;d$h_n*xXO9tYl-(M8|FFbu(L)N3dTe zWVge@L6}TkipF41MbFod>0k`%YIWlK0mpe9vZ`K)$O+HA(UI?26Uv4qS@TyR2x=Ua zQ8|>QuGt3Xx1uEPt5>42E-oH&r9YU?YEi82XguG7hr&mlr~i9b7EWWI*1c@h(rE~R zyG2?y$Y?o>eB`}RbtYqWxbu2cXIqbUkQM*aF8i__95+Cer571#7UL8wB;R<^zyrZ5 zhV9Nj$w04!yTxF6D?mA>$h_RgDbH!7t7}C~j8VARLSD!g$=$E=68b=27i`geg>3vkuTb zX9|Ww!Tk%}Vv8p>8**`18gV`;S`a50c7M(s0N;H~&=IhKw)d79tfP|m-xe^g#19J+ z3|Pmw4wjepHNevZ-r_8uS@9^6tsbeR)MJIc0 z`#S+NWtBoIts7?oVpF`(=zdQ<>M`8ILJunPGC1_2hZ-6NVa97$MN9U{cnQ*9`)xd| ztGAFF;I{y3^Cvc*Lg~#j_h`%LzubXFI+f$X13@JJmgDEG1{)k!GeI<`&}FQ-AmP~= zd6EZjZ&zBCA8AI&&!Af6UYGq;qO`0E&$LA|xIJrZV!oH+) z01y0^1JGY*jc=rnW?PI9A{EI>Hxw@9g96c~?~+OaZ43ErK4&H$SXa~(`2`Y;-)CY` z$n|0*Y~ZxSuDZO`bk-!Cou^7bJ z7cmV-)$X7?iEQhV?>Rsjq(p-6GewnkEiF8StYTEHk6E|cgGi*;GHucI_q&1E9sMm& z`;~yjT~zX}IpVoF9!(0T2}ZKJjx#>9(_vv$y%oY81qRwZlr-#}?#hB^ON)-_P>vru z>XD_6H6)b5;i)Ic#Br2*y9r&Yi=&*6wO7EjO8Xh^;3QrK0ZPAjxCc>y&$+&{p@EM- zE0>^TC(cFTJWu!a#3;=1(BfN|^T!Pg?%u;Cp3}gs^nHM!XT%_wVL(_~&ta3i4nzRf zDnh0ycT!38b3b)zvhR8PM2+~A934&Ez&pkJ@JZ01m^KUQN4Fj+jUxWT7*f_L!9%qK zxHT5VtkJkx$ogpJ)|Gu}5Nz8oMavo`Fm6x=^H61J?lIX&HIoSCv#?w5u=G!wl(@a~ zh&(2Piy(=a9osN`6srT@)I|LiW$neVF6VVGp%;n%<0#;}YS5~;pae6ercuc zM}HrZRJ1n%MNfr+&6R{-V%vk|vdy6jzI5bmUfq2Cq*dkMj+byyHUdp%F{=*jm zGtc$`oPqkrn$%Ri4?Ml5+HKFGNq}fDbdg$ia&|lhY6`!(M+Icgqc=0Av6#r2z(QeZ z?gb5itF_jwr4d;_mhoc{AVybV?z##gmZeFi**1(5ryFBK(5^w+&#uU;tW{$Np zwwx%#8~8_1Fu~<_p(f(s?GjIKS_WduEHisZzGuyTq{AHTM-PXw9&{9oS?5@8=9zBi zAN6)9f;NdI))CARXo?tDB1UC?{KL)LFUX%P-?v*Jd)-n_9Xk}{qDOd?m9|I`sqd55 zF8xN-ezY$m^XuZ!4v*a~D#e};`2X8qwRJPJnynEGF61D1%5z2=CFY>utZglm z4NP2Xyos8jT?u#1I5hjcNR2tEBrlS__?&uP79~1gz?z8oOCBq_@J-+>w9OkdVo?2DU6UZJtb#K$Fp3Vam{G(yltLtqA6h&Z znLqGV(w+JkvHhlEf5MQ9vAL1KZaMcd_vovJ=L?Z7N>cTp#R~7OGF_DRC70IJQ=5sp z5Cz6|Dgzv`SXWtHo1{QbWyq5#g$qZNH&XQWKDG>LUJxmLmDI3VWn^PRNPUiqcu71b zsPw5+&NR4eK<=B`Z}<-;93D_dl@4@yy`9F=`DBRRAptHUh1o)TSuS@MMN`GYwpIJN zoKZ*RQ@agdD1@H@H1QgFf#7yAM_LFB`!!=CJ}=;NN!IpCq zF_B9|SlK`N>uSMOZsqCRoUG7Bsw~saNO9}vAGYtu5H3y6m;c%|*q_~*9Yc+w?VI7u z)$=_IY4^!)GPiAI`}51pH8+uR<&}aRJX5vdV{x-brcL@L3qec`Wg7ea z=-p&Gr44fo*ofbT=M2gUk0U3N&l&wptgFxVaD;JFC0bj*s zq7NQT65RX!vLE;(@|!%3BhPEQlmm?{Ts#&AcfnR~Vr(s%+6kVJQBy?@bE73_j^%j@ z?*qV#jGg`LYX!D=KU8hAkf2jC@iX}{M0uAi>?eG$J$;u53k_83P{D$TS=rE&!ci=zq#}wLJY@D)Z1@mYIUul+tCzsr za2afI;hOk8<;8#*GH{9w;}j@o8zEgE`1X z^=c#Rcge?iRbqApT)j*w-l)K&(R>fD-Qv{rl}~W|LxK_*=TO?K#gu$%y8|_JlmZmL zk9n(8p#n}QFM7x(@A%xF`2}5)Jcivhn?UA)ZI^J7>D-xGI`8Hk6I1Mfys9;`K*?l$ zK^f_nZm;!DMNynE7VWH?Jaz;~MuANiZQ_@788wDNTA&HB?X7Y85TtJRmD;z;8FC-d zh`#3|!*o0l+p^Y?yW!));*zmFS-!y~UbMV4Hpv%ylP(ZT6H2R@9LR_|FS1D~e9SDd^@e|#-UZki_M&ika_w>+ zw*cz>FI%CDmA^i+$MUMWY-u?lv_qCOz5TFozB+<_2S?t2ASx&+hGqP8 z!#l(T22>pm)HUsINIE*Rq!yv1L+)%zaFMTyl&#_0N*9^mQB(?}M~b157oO3*%6uyE z{Omfl5^>VTwfdE_9#)5w9ta7cZKEhylZ@AZm-`j4K;<$6l-2rX^xzw{*U54A+15eL zt@;43w;YV!BjWqx9 zyg@#AIGPv-aoVXfU3&KF3NDA4cUp(hDJ@Mghsrb8hsHkK4*JK68OupXi5z7{CJ#NS zgs6B1Ae--GsV(L-m;i?xH9SkS#NwatvabwzCNoED4_Hbj0t+DpNVwvZp+e&Mri%_d zuwJ%mn@Qe3ymC}|oAJ_sI&Pd>Yp9=rLTw0jH{_$}KD$shX2D+vGO_#=>vi@t*m?%Z z%&7e@>E?7KYU=Toz|*TV**S4-x8%NIvIK-_KId?0E*%leUtT-1pNEiC*WYq%P!YlL zykUjVmpTVJd zVFJ0y!BP5g)4yAc01h+Ccmf>VX4tV)s^zA3Z)VS;6JU9_&*e54QlqUJIvt!}_fY=? zRy-mrb$;{RjvW{gnmpZf@D6)zxXg0F|IqS~eFNj{(W`8QRbbW!^W(2=OnxYnFG8Ss zNHb=vDv<1xaaI}=v~~;U_T^eIJg?pB9s@bOhm|qtyK(s530G;hvJQip1y?C>TZbFY zfK3*xWCN&uii4|z9KItRLoXi3K)FnZQjR)HU)!lQ86m?R8!(atHIyKRl__mx(lZ1U zk?%{^uom=QJeukyADwtCHA?FM@*pU*&`T_zl0!~iBRFO!f#;YqZb@goVjRd@7e5SU zc!N#duNR5o1Wk+vsD%L*s4|6`vXbHCSlc^_4abH6O+d20Tc+qcK%kgTZw_CcpDgI+ zHb&G#5IVzrxE*l??24*m5<_pQq(mnT68}S+cXzHYiM`W;pWKYe4F2%EvBmM`e_zUS znvz6xOCz>^;T&EtdL9fD8~{?b1?v!4GAo8idN@FM7>5uiT!1czGDdStv0g2qra zTQR%rYv;E;>^+9ZzZdX2UH<_of@_yKLkbM;2 zav+x;(@ig2b;jn7+Hav)Wf{-^CTPZ$IP?sob{Zdh3B!Zn1lFkz4moBCgOK6>UdXdk z_Y0-Z%=fJoyxDcI;i2Ew<)eXq!J8yRmrrI^d zrh{U}&twI^m6J;Lq@U0+gjR2-`D6`@iTDD1!btelTMH)Ptr+y^-g9XqMM8Ns`?km2>`2%Lik=5@DE0M|HPjSNlq1%T5xxInJtYd-Ikria$U~Y+2VIkaU8*<#Ri=|pyYWr zQK|xq@(%;TvxXDG$klv~^~JP0+(7DSEu(R}hHlGaLcYT|&?a#=8aR-#V|;Xj$Cx*EV4%sZ+3-D0_O%eg>{#CaON!oOc4qxk-iU zg)_xKc^wAlJn6HEdeYY$EIM6vsI~WsAzLM&Xr8iEXZMPJgEO@*tDjeMTCmZDL)Yw~ zRS>ilkzPb}k~XqVCm{92r*bJPkfN&mL2=;In)c+G*P2uPpFe?=KSieZ!do{;JdMSc zPR3oWZ!)t~<3O^h%C9=9=b@|v2^$@txDgU(m+MTR5EsaKE|gx>t;{DjN`6%gZO=4srpq_Q@M$6Pu9+aTtT<$(qKr_x()@>o7-Q%rG;k9 z;8WYPryh%-(qWo~6pOB7tCiH5F9nB$4|F){PM5M%;dG; z>G)_M0vfe-KI;jVhcu+h;rgpz%Ab*&MIjoj`g)Vnq-Gx+sSKVm6YL}N$4nWJ7NLqd z6eX<)p0x>$tw6pChEV654c0eIL%+rduv2Uo#XcR-pkb!2AtO;zMs%2%C8_L@rA!Fx zKg%7-ox7<(!nAl*L&X%_uMA;3D%M_Wgg6^FFHr|!f#a|Ww;NY2Bx~z{a4myI;&D48 z&HYSu&&Xz*AoP+;opTz@uhHq=xQe1*r>+Um@y6=4 zUpEw`dXI2mwIF0c+*D$mNv@uLeuH`>IVpapBn1hKz^-)iRd0G12`0*g_@2iGo%ZdI6Mfu@9|Hg;_q-0~Mm!J>M=@eCxPn9p zA$&;6`fYFTX69m-&I&FKzrr#5RT>_u&+V~8?5$*t} zj@K}C-Q%^?JO%5;LyXlvv1OZM78Sj{Nv*?>c|x{9j{TV)u6}U670(hLx2qZ(Ppv5q zT~r56=R;B9-iydKVlkz2&8FR*Qg$JBbDOK&VkcP;#1t*0qEKesbRS8Z+ zjlZzbb!C|Az9@E4MuvYl;(I`A{5hscFkTKNHUNg%HC*RL7R+7PTk{M$Cnjvt&dQ5I zXH&Hu&o1JN2A&$1*gXNo33-ak5%d=Z)g6YItXEgNT9Y0PAKry}<4qN`5dGe@>zp={ z4#DnSIj5+@?xwivifnL8`1i80HJQ>3w13G>h zS%Y&#-mnQ)q!5GdLGf*I^w#D%puw&#(U=hry0R3k zH{>o7Kabr>PTM~i4>&3IXYy_qO{R@-mX1F0bao&N7#Y0VkFGPP2`&*Hx2|ebb!M5z zu>s-nDnoh3-YwpbRqNG1byrg#EjfXT72!IUYp17?nTbmOyKh0yok|V;{bimK-%XTA zp2B-(XoWo4k{Z0?-rrp2;D5f*7kCX*7^Xh7Jap^T(49`3F)9(fESPmP()F`OrFf z&2ioJR66|9HS~mkcr%V$#RE%oMmVN*`9kmTxlvH}1atqwha7a=p{y8y{!1;VAn{f# zb3O$*^%(WS%$+vg)OmN zUYSaYS*7#rJEb6%sp_#8OH)=n-~_*<+|!R%~(Scwrr+ce*Ez|Dk9=49Mg? z_7e)35CL`bMjfUhHS>GBWwQ{DA^!?>>emkh7y^MMsk5Yt{i6luv^|?&=P9`RQZk26EQis}Lm~ix?-GU@}QPDc^f_(ww5JA@D z6x8}Gjgk~3E6^@jUll9y8-mLl6dFj{^?^@t$Kw)#ga4o9U7X~U$i7=LG6fODh)qSz zQP7|pQ;r{;xNT2FI#$ZoKmIhiirQEeB3yr=@N|@I0Y!ugihavrP16?{Kt~)9Ih~IC z?kv?QCjZJSEMmbjKF|3D8VBh-Ub)naLzERCro8*#-l_zq=C&R{N#SVpCoqMa+N<9x zG94nC7t6QysgCW;O_R4xc6;0n7ztiLIkEiR!hN*ArYUA#ky@WlUd6v2L*Ccj*y6S5l@l7!Ons6Rg1Q!B! zLpGIUOZmt;OCY~!%79jUuTxHRb;jUI2POsfD%tG#&e)Y&bkS^j`nl%h!zH6VZB#w3 zI_a7nX;&XJ8ZMpw^O5K^JqD~l2{6ej9QW_w2DMF{3e<@O|Nd)x_28 z)-T($q0WzQnF9S!@q&0hrk~>6%EkvZ-1f=PzJj=KE%HVzy7MgJ&9{Jk#DP3q%o#ZDo%)wq+pj11S-t;)L{k-qx6#x#V zjk`_RBYhXG7p9k@N^+QvRkrU{)Z2JoKx1<$0|d7rNsjf+J7?e zENODymgLSD4Xp7Q`4NQYN!l4J9L}%|BlnC4w!aGnO;cMtHWm&O@gDAi zo|&|Is^BLx5(KTq6-m;zd{( z!KenYlhEsivNr*J#g4Qs7%}j48Jp)A-=Hq{U(po9S8k$qxPZ zV$zVw6Fw-(RLJ{3vL z1{Ki8EBEQ=V-zl>>^U1jgr@&5e*=6lg*<24k$-dmrqD1IE3;3GMAwQE0+Hz~P)PJ~VEgvn`wiY`hQvQ*PHK`i^- zO1&lFCQB6F^ey-O8W85NovP5d(VOI|??ptilNB@WPQ_gd@k@pn*hnGBUqy4^?{~)Ehe%$h7&Z548u?NSs=r=L=|Z?VH~n{@?7@1 z+5}N5YpLTew5vamh1Z7L9PM>EV}Xp#A8Ev1DMs88CYX;fJ+P&hMgc-I}p zI+jx?@3=H28;KhD8Sub$*Go?fml?K!_8GqOS}Kev7wnh>+hdfm4|gwA>7ChLQ`bx& z#=)2T&IvzA#J1D;v}0NouO-*)ARfR^(v=J!66czfbkpDLw|DGv+iiWWq*DVsod^D_ zp7Q~C!eZ;Hq_YY%Io>CY%7UpNb=M2+G%ba=8WRmFshI9IGy32~*Q93e?tp#M83DQ*Qxdin{!z2R zaOZ}Xqt$}=*1E3{0ly8n9A6$i26>;riW4PPp++Y!(w+#B5^#ln;H*NZVL%z#K6#Xd zr1aO;KIkw_|L*x|a_pW3fSn9kx~!1Vxp^K~r0icFYET^~RKAa=&p9NHEGwhljf;I< z8;u$JH>-^Y5iIj`IJrV?%U|(?Px(Plx80GlBo!wLRAg#oT8niO3<{NimCikrH}d&1 z&iE{=(Cx@h3*B>b!EqMGRG>r@f}%I^Q;SX;3U+~>JNfw}Lcpdx9%s)qOL!QRxv9<4 z?7M+TIIFUG8G|@2JmB)T{#5U`Mi8!kHOBAcGAUsxNpR*DTQ#xN1p}g1p|n_P6RC*o zb<(tjnK!(PGp}wjk&AH&rBcy)nj&H8dKE>)_X!dDh=BX2d>PNhmfTH8azX)iad$8Z zRnnpz_Y6&#m_}2^swz6TH|cDfjq_T=9r@{*D8I;qt(YYiY*V$#YH*G$T=)$XLf$K5 zdWr?~Ik-S9=U^{&i~8M zBg?&*+=%v-blgDvN=y>h3`r(jcNiap>`~@Qul=;cran0hjx&ElGanwc=Tl%SgZnwv z6rN$)0_R5H_U_Hmtel0&zm$+F{k>!!O~nt!tXPE}awo5kgpxT4&`hD)%a3}(xugII z66S3vkIx_*D;d3mzS(F2xmF{Jq)AGYdLGVw0AfVvoVjDS^|ef;NUON#-Cp(5bH<+9 z@A`~%e^$}#yoU&D8#qS7^LXLTmqcmHW?~-WC1qX{WPNES-Z5EvP?H)0RVx(F20V?2 z5zVwfKkSie=WyS`Q*}p-#slC8*vshucahqQU-GVDwI`iNz10$6+Copz0T2RCx zr{RB}K$iD{K5l!Sx;N@pW5%>y_5TVLgxnfp$ZT2p4hZnJdSq?5g=}>d!nqx)l-tRr zu@fM%ZPP6S$%aB4#%kV#_-x0BFVjCtv(PobdqLx;z$b%iI15D4Vgsx#Cby46`OV-= z5kV}mpp~uh&ZR42L)B-fH=Dr8hU5X#&DJKDlK(=#dPg9QM(Gjkg32+VU!WYe4qeZ4 zE|#hVqq<2}Yz0x z+y~V-!%wM!t>#4)E>zuiejbOj8iBqRon8?OYJP~%Pwc?ks zj^2O-6S-r|A`pIT2Fz>)w%R)Y`R%eqvCA`Cir|Vm6oa&;D+}hKk0(z1LU=JH9J3?{ zIL(lJ&*ap2_+Dv-P2XJTI+|Hft1349L|C(S0N^478t2HdZilAX4RX%^OEhz#<~#nt z7N%k<7VJU17eO!$XA?g5vA%h2I*@3^KdQ{H{xS8eSDTqtFc;s~m1C<9_;UQ09e;b~ zN+_n;k(x+rhOKEa)aFN*e-}A3<>ixr>sc&{EL|f;#p#rHY2NLFLzC6Kbwc@5MuP0W zaWAZjJf}q0=g?~(Ga)Wt^M(e_*^mYE#}CUp)e7p_!y>>l zDY}HSDa5JIlDM5VQs(?pjZq(BC~!zJF*pz&ecvTpc{ZxO4i`U1ew8#&h%u>3@iSQJ z`#Ib*57mglqY$txb@g_M98)Q%tr9eKrW7ydY3Yzm zTe-&uiEKy=c(lAssb^Y+`7P2D6=B{C;2TJqhP1jLkJ-4zr0OqF0tG>w0ttLC6IL5{ zLi*I@H4@tKSF-G{u10EDA%0=5K4sJZ^Y3K=Dpmun6MFV1aK~_m`hS%#9KlAA#PW$4 z2-Vlt2E4uEQNDs;!QT^=m~p}OIty`igxz67M%4=`;6z2!j*6juyFc8|(i-84c<6bX z*LShi@j`Uq0QlCcdmJf=tE;>-D9MM&*9gQf_{y7 zRjz@SiuXu3KI{Vlo1=kK_p}=(itg(iuueiOkFM19_Yo8E0gSJ9usRiY!2(``Ovx@d ze6IKhz|dZj0dzfFda!M?ZR1}moL;_0Q0aDh)w$L3Ai9p{GOScmqq%X9$`S8arkMXe zH(oGjwmyOYq2>S2?u^_3$hStNlv#l-oaEN(R${yS6nJYQ5N=BRQICE&Al}M@Q-JyE z`&4u>FLs}3fH-9-47yrTvK(^yVRfR5>KWhk?m^P5x&qmw)K>z8)lzv_nXHpx# zvjCUHYnZqwF^mjZuZ+p-yBjcF0sGOZ(J#6vPv#hKTQ(<}J?m0WmutQu|%r?t-C-G{oCBz0Tg|Kj&8zNkcFZpepZGz68PmOHD zS(CJTx|;Q*IvqkTEPlg4s88s8KqmPdZR7EK_euNW@{nz zgxekh0RaTK?Kaqq9bS~@d^1~u+&$8UYX?m5LjitFH2nj`@_I43c$ z;9!D3K4+Yf%%N1}N$dMFnMH#rhMwBB>)ZBpp-NL>t$STASe?7|81~T$T+x06_ z0VW+FG;J@M_>VtRcDNTncY!&gx@xlb*{#TRwlj8W#S_ag6PdW65`gS6g>`rwjS zMF{m^F<6_X!Wguak>#}1{B*wmeRw8HpgoBSO5S1`0{EDdMPX^{(nSPX%Y1#2*E%VJ7D_gho*t3h6cQDJN(v@1Zs1?i1 zDTd!BNsLs!+*>!Mvn4#nSgn+ZNUV=i4NmFfTe@jRcfOp&ZVabDIHL%a5i#KR8Hk<& zvWjq)ognQeI+V`m#Mcm^Nt>bl=SH1 zX%#zB7BRKw<59kS{p;N^Xaavg{3Q#~U*xPitvgSr321P-GP8`zKe^ViVBtXRy(o*{ zjy8q;!nBZ#3U4M)(yIUVhuNw-VjE6E4KbgkW@3SxbVcwdc-x>+*H4sLavf-<4HAHe z#i=n3(#-(zORksGtMYqlzzA!QB3k#0+!9KJ<)aUhEkE*NG$E0AK)GPOajR?~-fS0R z3f*hGR88&-vdL60z$%}r*wQ?(?QS3!=J*3NWC)WMj`qj7J>8PUX%vG&BD!h*EMD`y z)%1+uY($lkPusZ78@k+B*>G8ou6dSAxz#{Lw49aFAEPrAcvh{!D*zkO!_>#)Q^<u_)cZu66ywO*eW40uskBV(V$3A$U_dRXWO4Kxb*$oAZe0&)MNQQ6vkAsCOqL6Pti(74#Y4dxMfu)_bxk>y zk{_7SrdZIM9t})AfanJt&!l$N0o_eAc3&%BJU?Y+bbbD@{y^|v)IBKv?YHxrS^kj` zlQ)2^Iq+Xe`j$>xnRdUFP~xMqwm>FyKghvBV+o7H3Re1r$iC7A6!W=jS8EAgE3jgR zBK$xYVK3&py|O|YNdt5DfQ@37zB7;g*nWu$jY&w#X;w||ot)_!Kuq^qxI|@J>o|Fc z67uH~(+fj0!-vck*cE= zZ%r;j(@i*xoc}Wn8!e9)h8D151#P{3*sf;iW~5H6b}>MEz7QS^k4+Uzq|sP*_jO+d zIi)iYn6k+Mv45oXbG=skS-<;)h9x#DgiMAw-N-`&hfej9Wk&JEcm~mu!#5F%k=5|t zf~bbjGY~CJ(&3IK>2~8{;hbb<+EdJQ&EXl?8Guw{T%7@}dcRP`BvvSR>%C zhR*LO5|ZjxxVSO=XB-L=87c)?t*+ywo(8lT@QqcTLs6e6D{-v4qzxk^S{PO18HrI) zkGF)~{LW2UOns!X580=l01mUinygiG((ZIw!!P}>kLue@f-Sd+l6~*BvYk{~0e=~& z)PN+mlLW*IS=C`N19)#PwN*N>t)4Iyf$Kme#^sV)-lyJLJ!;JGG1g>Z-RP5ionD>Y zE64#tjr;CKfRUvDlpq4E7}Il$c!iqU zHKl9RCI-E(10T|4+;iwhjGqa9$&LX{R;bO4#**G#dW}BBJTh~kHQ;!B|A6>)V)8z! z?LH^}J?^A0ye_Y!&Gf`1{!Yt1gs-5n5Y3#ySGZFIK24G9(aGKG9d}~8$liN_s>PA2 zcanmxkt2iFjq!TQgc`>d^FSBvb8V%FT%NkqRG=(z_atoL>N^NE!Z%9&>4RjlgQx^u zEPUdw^J5J0aYn9yy*&AOoD#9``EAFPa;CT9 z@})5zAGyj92~aH-(i}}hp54AuBGBzmw=~vzVOSF3_+Y(CVFjLhQ^(lFU$5t8$0q$I z*(+FqwO|IwfAFkCk^zT~{@l7Eb2}^qTH?j}76xvjQG-EfrPfZ8ANy{$>}0RcY@)AqC)Z1tPX*>$X%7@ztINCI z??~gMv~j#+*_O<{QZD=$kX>G1zX}y`hgJlpFBrXppEGb_z_c|51~(D2xj{!&vk6`I za{5PVOcZ*oaT+ztbNpwCJ}w28H_l zK;CRyev21L#1VKC%yKC)3ARi%VYD+|5_$D*<)n!BE)dFu3-4WjXT6rWnckOssBh_0 z6~0cv{?lE}bwYmTLsu~)gS4?FG4X+Qj!8}3v9HA0+7&;2j#h-YbFQNsrsLQR8<+}G zY+hfb_hx@a}Bp%q8BKwSy2e zUJB{QR0@go$^slhv@51sdt%!iNOWo~^=mBlB~~4vMzV8E{ly!ub(D zT~2)jP=qh%D)!fs9xCE$ZGJ&PaZJjEf3_O!wVEfu*|J0t&R6=l@Cz-m0c}s)7*{}` z=n;>?Ydq&W8d;!>yAiB0Ma&9Xn$QyT6W-_Ubn(lA)=vcHD}EPnuSgurboFI?%VC+4 zW?9bZy7mzKYlCD6b_fb?VbyB~cndFM#MRGug`i;)CujxDEDvLMH!v_jtKz~kr~RIi zl@Dvhv6af$^77C=vXA$lt8ctc=%1K$WDm%(<_zb&p#nPv^Pj%cufUmRNbeFR8bRav zSJ@(%TS$9>*g*>!0pd!~Rt|#TD1|Z~(BpxMe1g7r)P2-Va zWlVt1TmEc%s&j!JwJuN8NVd_BKe<{4&bKlC^x~QB2vns6 zMBLho`otUKE}_~qmD>)SDEFndfkX(>vyYt}MnZpZJSh?~9Y6V_KDTRG21miIoHO39 zGzj=pDe{2;{M?yfLF0u9?W7;#%>#PFeQlyG--P-|2&^3lacLXYD4JfO4(b%y1c0a@ zRsKIHRMDuM6PshMEkaLWsI}A$%-rtHj&#>H79CVBT!wq_t6DfQ!$)Bzqoyylt`EtggVmY*9cRko#OKyJ1>bJNA5kI6I{uO;8)j8SuutcRtX>K z;)4(~jOG+~ttbk<7>GSm0{m;uP+0Ri5rws(ir+8`Uz2^TKbh z2OS}s+*|Gs&h8FaRRQ;CffwHn*3E-jkmsv}0T-dQFJMqAWFxT6hL~so`{AaL&}h|v zAz1>)A{krcK{3SH9D|c?_>~NYHOk*8=!#Z)1;XBK!90;AiH}xVZ=&ICTf7*qne_F( z6QtjSuJFU`U4!_r3Z!M&3+A?dH%l2gWLaftN(hoAR@l7Y?ITaQGvr)}VcG2yRUOjY5D!`1+*pubuHS4ylt(h9eDF$(PjuaT$v@TzFQG7< zZDMvB4`n%bgrgt#M;j1N65P}oUx7V}wQjPNu4xsA82goo;myM#e{=M^FUEeX^$B9D zrQ6uPt@)*?*R>>Q#~r5kD1C71nW@SPMiF<+9tH}kq4XRhno+>Ab=D~3QfW{Al)s*eN-3gew~O-i^9ue=D^Z0FYO$H9iRm7?iE@T zR*(S-p^tW(;SPxJxb^Q2iX6+k%E0jXILNro=^ie?6iSM?*oOt>nD#H0c=Gmu+C;yz zG@VgqF0)$sG&m>#pm<`JhhT2@5_4h>mSP@N9FXRFvWJ@=y!|9Nh39~M!mKY*3B0>n zI~1%rLOBMR&M%&BWJ{cXg@7J`5Ta-*wL>;UPSplOxQ-2DZH?XvY7|0a6*6!lHmY9o zCa8;thZBaqSb8Ei9Vieyz_vsG-a#*wdb0wzi96<+r9=)<`9BogS2Abxx>DB#BET7v z*0;oIFv3Qx0u}wkl4CAL2qGJzu~%SYTN?~U`5)!*z%TGJCdzt-9Pr0|Gb_1$-1}O; z#7kmc8sbg72P=WhkRS^rH_<6FAsbo=#RCKt;581_G|C-sw!1rh?S=3)SFSFGX2`xi z@V5E#0R>pXbPkzgApeU%$UvVxPTcw`Z|gFk(^mtE$?**?vKZqy-6~LxEtakcJD$ra=*nl-f$8{DUzaa0wpWyjCZGT-1=c9pjXQwVe;EBj+#ie+3w>0bwBHW(vL`# zl@)3SkYwso&X*r+O+r4)j9wTD^KGoZM|V+x>{hLl{*gn^T$AuX9f1Q6CToVintPwQ zM<98GE#!{K3t+J$aie^aMOfX{y^9DD6F-nHC&Hp-&)(eNZy(ksZZ)334MvY8buJ|6C`E)qSEI6QuARdC1P43w(q`Uo3i)&fYn z_-TKDg^IAQ9zg{J3hi)aOi31xu-;(1mp2o}5Auu%); zlO6txXg)6319Ovwo~ov&^{xRRxmmfD^a4qR*rNQs{jd}I@S?Wo*~>wP3qw{A@xA%E z8BLWVq6OF9Pb!yO{LpmjcPWw9{>#g+rCH4V5eA1>FEG2S8FIQXC$sul0L~~PL-nYj z+$R0cy=$^-4XJ61V@Zj;(_QH;YAcfQq~Uh`EQi&xJolN-qcYo?6f1RVBZ)xJoALAu z4m~jvY#=+buKok_4d-*<@6ns%k4d`X_|?8S^i~*w2V` zQkDUP=G5Xwr>bXt1}^#$pI<(ID%rMj_pIcBPPw*r)gjr- zc0BILqXzSCA1c!$7V$h*>0@YQtJUTvTmW3YOo}@vyXKswDb?I>vD5@os;|V2c_Voy zN@B39e=uwIZRLd&)}4hkQ$9NG$j}&DTSuJLQOyS*4G3c>5*g5vtQW-1J$~lCP0ATtpxT;y!wGafexgtkH4l zbDxy=ImHiwFfN$a^IXs=aU6RK;#i0BVV+hX%&q?|OjKDfI$f|+UjLuL^%lNkq)bll z4>Pi3%w{!*a__tV33$)S(qYkkf>c{M#a$D|8^0$m1&?EFpx+1ZB%%=7Av0}lSaa!* zEEOg=R}71oU{$$`7ejD{I2?fAp^!ilxo4aI0GM^iASNRO16)4J?%Z%oJe^C9WFHt6 zJT1FiG&62D*n#WKUcNHDOvjyd6QCEjymfT3a@bJGcZ5Sd*5Utoc*f3f zj91tcU;vwoYrA-;nXSnNIQtQf0GVgA%0$gI6zI?7((<0c{y!0a>Y&wrcJ>+>)nRvF@`#uxb4zLHz9i7Rp^ZoUktFfYz4C9g_|tU`>x(CiBQpv zn)3sNd5a;2Mz}W`UpKedZ7DX6iH^6i)GSAny^kN*!^RldvBO?%Kfe149J9BIwrs(ogb}gK?s&uEk#g*;j2UGhH@qtOg zb6lXwSd^n2d-Hk7QeCj1|47o(0Wzupt9B;vL* zsB-f)!c0;xmS;76Hkf|=BV<!W3jhX-$!?oo!)rH!tbPgAjsegiov`{UpEA#f6jK0mdI=REJ2I*C$rWO=wpw5IGi0?yr42F+giUT-` zA2C5^AYuH#E{x}aHP>2!E=@_7Kwwz_p3JnmEMfEPidFxf!#19J@4hwe2D&!Pz)ms@ z%+fs-G0yrQ=`-`b^JKHIGUzCC|BCM$gas`Gt(zX=A!;?>^th|jU2O6yil8Ej(}@s`ppNXvmm zn~PA!pV(zfs}+zByVcgW+Ie20T6b*lA@F3O?8i6E*a_bTr7!Pi6Ly-$?txo&q)hF6 zhOlQDaf3S`MK+xX4PK4o@G#jP%-mA1yCA>E&b#jrtz!bK{Lh(+kJ0>Xo*CE|2Jz`X zcj9D7NN@<@=y?DuG%s9D&YZihdM&kjkbu9?ykQ~D)AN^2bYclK2wSq+9Xf2KjtKN*t|_xQDkIfZh(0n$6tkeo1zmVO)z2Hy6jo3M*aVXwOf!Odyz&XS`w& z!rluJ5Id&t%yVhQ8s6ZS`s4YF3=Ta$Y5QKZ{M6FvNOA+7Ps1L`Er5Un27xm*EqB=S zvh)&@T*!F9w9?d5wAE9fy9Jf6!;YyZhtC_Tnrw4sv57$LdJyM=NTMC=OhqbVNe zni%{pFq2I12IJfxwi+2b0+xd@SVzOU1^Do#3qXJm_`qIU z27!o!UK`9J<#5CMiwlO)N-f^sR_Kv8R=7)O{KIL1R4DdQP0qE=))%%5Fl6Pk-KtN;9w6&F8L)k3W+gBR$PEv>i$;%_C9bhFg&k(be}jm zH{De0+U4L?@yW@cQO4gvJ>mjoHay^t1n4#RpxPRb5sN=s8qS18JM1X9MrJH- z%*yBw4=5%#cS65Y6SNG|&+}od~&KEkFGD?6F%OcT-*K zS64hX@W`W%>HX#oSI=4Z$^o-?I{o{jM?5?J=8ucZtDj%H&z4Q=KltTo#bd@kaPLQ} zqAgmd^jde_y(7n-_0;dC&w2K5JKZzyFZUm_-A|r;zx%VdXMg3+zdZiRo%cU_@qI7- zx_a5#Z=AI9*_yxZH0aU78LuCEWqY52haA#%&OS@#z1X@2vRxXFqv#LGyjNiJd#Wcg5KGFTAnxy)!?!r}Cz6op*5W;j2fxH}*X1 ztAD-bu(m_5`ToW0n%7tFy7=C{QRdmvd)#~Miu-KUIJ9o{LzJGXc%afnSPk;V= z(M!8L>*|M}I&$P?HyrW%Z=e0%b&a2-Mx8!$(TA7)adGdn8{5yRxaT*ErgrS~z|8W= zYft;z0Uw^8di<39va|X0-EqRTGfw}`oDu2EK7PE{uyd~aeCuuItee!{@NhGpDm{PxsG~Mh zzWu_6xBhbe%iBD3Y2~`hk7%27;t$_>|J=Kqcg_55%rnQlI_1zi&Yv`Kuclq!-g&Q) z*ES#gP~Cq{_|wc6-aq*0+WY$)Q*%l8;O=?f-{UJ?&)m^?Uiyz^m;7t>8E4K)zdz@r zqb^yt#eE0dAI+|R?fxAnzjFG|YL4kTJ-7Vx{g*xYcI}W=EAG5!&?)c#VDPN&`IpXI zxn}MSSC5%m%%mUr#h-Wm>C$H(@4R;1ipKh386}78dpSX9s-(K~b7yCVtt{u>P=-4y5 zzm`9%SLN@2y;r|=yPmUi|6fjf;A=CUzj2GJkNb4H%9d{wUODgnKXpF#hoi^5wcDN_ zzI@L9kG*}$erNxF(F3X9ymmv!zg}JP;z=`yPiuebhNTy6Ibrbe#X%3Rd2ZOd2kbEG zJKz6#`kjL}{Ndc{)3>WT=;nTpe|UM-Nx%GK@0M4Oo%rCA>Vpp6@z6c~(QwlF?>_fH z{lru9t)sgJ&phY(-s|q3U;q14pWAx=ySHzcyXX85w%c(08#{mVcPC%_`}Ieh{z>8F zi&DcLxa`jR29EgJ)i->$c3u9Q%0+*A>$2h3xa>X$9e?S!+P2;Hr5pQ=o%iMy6BiUF z{IvJCcf9JoA(!uQS?1O;OFr9e@-=7Na7*L<(H&3KPyO1@HY~dDkDc!asR{rb<9mqzyI1!Yi8VX-QqV^?yz6#{Wsow`husQc=^ts-+053{}_JVS-XDg z(V-oaANuFIZI|rav3Bl~{aPL@eEN%{FWNbEz>8~t`^vHpAHDsN8-7rc>r;F4@WT#U z_2yl#5196ycY5Ek&2De(R;-)-=K&iA^?U23SEfAp?&`|dcPt#b<@J}0o4W6%ga0z- zrS_~*oBAg|H{>~9zA&4SzGRO z*|hDZ-B~`S?&!0=_weicKRo`zU!DK2r!LLCFsuFCnG44}_4dN+C;Z2(wX-MVG3Irb z40~?y=>x}=zdLH;$U`O^@~h(?xb($GyH@OVdCQ9je&eHU-Wa{@<*%*X;zxHsc~AZ1 zG5_({yY(CXKS|Dw&$^m}v#u%Uj>z`UPRzRU5!qw2-_ABUw`;p=&blqe=SF8Aa&BDS z?T_1_C+1wnb-UD*?1|Z}vu^#A-00j6*`u>1kjdSJBMX0tQ$~rfp%|$n)kx9KDWtXvJoC5>5_8-hw zy9oJ9o%@M%r#Lswxp~g*=-k(wdj~(0Fs}eVnUzZnSgTJILrN zoE!3NDt7_7o#z}BKh3$Eb5+h2KZ&xr?1slO+4ob~pOU~W%iwR}E#`;1wGcAnR;Qio z&H-cd! zx2s!_%H9%X|D4V~K!V%_)O!4nqU>#{>2-dBXtb1`I?! z7&SsS<+hmjZsb-~q+Id1$gODs3w-+?^?Rdz2$KbmEP6CTXJ$Kaw4s|(cGXXoKT_eA z|D@W@d$8KA!g23KdAIuVJnHu!vh=ATZq+3fZq*M5x}^^^yLms%x}_)Kd}y;<{!lY$ zp7>e|m7>WW7Vb-Ag~iHD2EP8a=#ZVS1}SKVCU3Rm=X ztFryv($;=%&7*+-*3T6lhPti^dNsk=Tb!NyV?V%KLieTt=oHA?v_-CWf0yeE|8a+& z0lQSU&+3bOJqVj%m!_G{9Ss|%e~u&TvKTVJ{#)lPy}G-*3uDl;`x6FWmLti18X<@A z1)l_1-|tkEeVp!gSrD&{K95{?v;jTj4#wdwjvmUw{oI;p4SJVj2i-EwWmq?xCzC6bbu3!B`pO)}%jXjib_VdkzbZ_g&pDqV8Q zhh7v3VT27SVdhE56HTEZlAth?$%xBDQ7YQdO;M<+d)>Xk{@gYC1!s&*Mhah6n_il} zhlaW|bq|eJO#d_8y>8vQE=D1jc2CfL6{)95@=Wv|PAp*h-}C?4;eY+$f9^PtXpIiy zsAjV9no$sATN5U7Nu1T0B$VRJ=+49$kcknSiK`DL%qEgJt1@9RCrP8nx(y89Ol+4j zgmO>@EQ2&jIFuElz@)c>1(xTTX=5;D-%_v%-h)D4ivM4@r@j8C*@M1f{;3lyr%z3# zF#Uo4Wb}mT(^#BJGT8jt=*^F1KINPp(39W4ix1@?KWl$e?S=Hr8$IPE9O6U!5KowA zCezRAZ>+tI)!!Wd5ML@!DNKK|_D#Oh`XRowKI=r%(XxsXohUOdR<--NP{O)LmQjAK z)vZOclp@O3TkKikC-t?OEy)^vCfk(JC7Z05xLQ+(U_Fsr+aRkrTFIJDOGu98mRMTW zR;;J`VR@FRiMZZknZ{bFSx<*%nZ;Y?=_j&4#j+1|NSHAhGmWSrD?w?L5}_?`2Q2VG z+?_)$o@CwE-6agLsGOHQE!$NDP0HI!6~witEEr`On$Nan+qn=eWIM9cb=lbD9@2$& zlUtkOH8K%uiGXhxB26vgO=z zxo>}CFj-H*!9;ur8~+qDNIbERLV5mOnDV9r`}|kn3O`^>-e-}aY=n6*FCya)&*X_s z0?D%}n$KvAp5>faRnQ`IC<2Y@cfh=85MY=0&ZJ zq%L_G{ScpL8P^$B&e+a8%h`9#vriWJe5e!iy9ecyf;_S@AIno-%acFXG1hKa9_mQt zNW;2YdV1zZ`8;J?=<}3kADlJ;TG%PTBlb`*9O%dr6+LDPhEXRgUoab(dm19#b&SSGk^3;oCiZqnL+9Dt8YmveC zJIk?07xTeJmE?s6f{+gnt1CQxUFvm*$1~4lvT^r@d9f8LG0(c3&zUFx5kbD$=SfdI z5Arlv zLEb@V@+8eB~O?gXEZKdE_v!!m37=S)5#C}XOs<#0om3t`Z8aa z_O0j6WFk(?vkZcep7L-$VV*J~)5AP54Ja+oT9#*h=6O_jKFqU&telN7Px(scN0?`znPRypjNyAD!@R)W9@Ni8DJheDioPb>bFv9% z^lY1Xo@>wn%+vN$I})Dg!kACm`wQu@mVi86(vfsp*41_sKJJK4DcAm?J#_fdP}VMU z!7!6@;>vtmCKJy*%Dp_4fj*ZsuNyso(p6vpSY3y~$UNJ1Rn;OJ6|??gc{SpH%pV#UeH~&i^X$8Rh&?e4gfbd* zQ8)MYErhaMd?1w;8p&gz#q?N+sQgiX^8xbbpd}5*ZZi&}hX@)NVG@&BdPF6Z;9} z=Y0j^UmuJYT=z)1$lz^G`mMH7div@jtgvGK0Uri`mXn5wdDd;oYo79O?aFfM#rfT6 z#{0a~MSx{F>A5CkUdoFJpI#W(s+e4w3XRwk{^#yWD`J z-Nzr7yxO)h&R16_^3*|4)};*WQRYeC%ePAzxQJ)DmQa&pZdJ?V^PaRFLaDvv+}~+yG%-4CgJII5k7%EdFPn? z$@6mW^Wy)0JW__?IAYmS+sFAWvhehAO|FL%M_7~0LES?CB_gj(YsDkj6@lv0Q z`HyJBII(Hm$8**l0b$vu(EAudJJbTna^~ysk9i&yoEBu3%)>r@g207*&zPycL7rUC zv!2L_wqP!YFe=G-IxLEq7h8FIl8y@`mQx1YC6_#5SDvgX7(rM@{-Z{^?rfw}xXOk* zdlVBm<5P!4IrCBnj%Iat^X0VRl#Yak*TAF^+qx5N;U{}8EEkzYOZ9n<>nZ+9pMG2l z!^F6LtQ4K7BM<5)_SqpXdDi0^jpfqrr2Qr_FZB4sJll3`Oyk=#xx@j2Q z@|<^==fU+J&*Vv;WL|jUoa7T3(pZ$ToChCM;92DHHdK8TPdV_2Ma+lWW2Dz|t|pRw z;g}C_)4!A^Ho(l@-JlkpXr zLA_#nUe}X`hr^=8=vgN&M+Y-6b?bZAGf#e83|YPmjWI8F@cklV4RMv_s!QUNhysZ} z1pc4$by=r6un7iSVnY}!ZY%7&I;;>;PsUVp*uUx=?m|Nlr{8oxNR(3h3_NMr2Tl8Ja!k!N94hCJIwQNwF&=zeN|tUGiDjR%czf0 z2F9SIZk^;sw!wIc064KvDnnQFzM=gt@e>zOEEoEI1C#!(3;Gv9%IZSy*d|c8Hs+sJ z`>QT0KXF)2iagqu%FJ>+?i9;82);ynFwQVpTPTd}gLSXSCUW*4BJ}DfC-jPC$W-vs z$;O4m1Zk^MZKC#98)#c9Cw*q0eI`n~VyjFLTcUE&SLz0FjPt3rRbH-r$Nu4@YvnYX z<=F=FB0nzDndjlM6`~9M{0=k{mwP`^n!4(^yt-G?M>SQIET_zDhk^oSi~C*9;l4-lr}C$9m`|ipn-WFm?I<}k|(aC6W{3Fkc(;Hpv==Icj2%+ z6j~9})Ky6TijHO4?-D}}I#B3Ihi;d=(4gyBCie7l62=6O(7-uZM&0}L(Y_X*lO|1y z>*Ap;=EcrfRxmI2!I^m;wP=ZFv7xsQ>A5Cnx%e$6TIPk%&|!HicXj2HnOFNHeEQ%4 z0G5lN+{h5rDRU*tS)WKxdhFzI=~5~0l{o@lwQE10kLYYDLx9(!b&+A7c&>nN79 zt)u*yl6Jk%s}HLWh+YW3q~}qEO+}trJ`{Dyn{nWFESj0;H6Igcs8?@HHZ149GIl|N zd?Y?By4Pup6?)&-YR~P4TU#!*%LL1Bolj)g8jp&yTy<1ADgR@|M25kG2gmYQ+-4oh zZFSSZdWQGVtSvWE`$GB=k9o3OY@10}SY1pt;y)q-{#ZHpvSL1Im`Io>eQ&?cq`nX#nIkfI1`cJ6W0!-Nd5QC(UOrD=F4DD@@Sm9Y zST!!rIgR*__zS5|DlC`u;1^tq8GA@)Rluc=*UEfF2Zim`j@@Ry3$xoRgdEw*N zlPpJH zWkI}Q@9#Lon|FGB%eL5Nm0XQ#+oBWHEH7b3S~QP%%fCuXnMPnEmLj~`7$B%d1lOj$U z8r$wBc$iha{(_f0F0+gm`SV(iiDjY)7n32ZU|b*9axq?NE zU*rk&6$CHje0{<5BG1E=kJolgtWOvvF`i|81ya0Z@tiBEh@-w(E=eNH!=-ZIXtx-$ zc|Pu8+b#r1!z#E&jP(>2m@k+N4{PBrfZwa-rVlnTVjRXwPg)iv%@{4IlroW52DEQS za)G-U`z-evCQz~(Hn7^iFx&kq`u+> z9k<7OX-qMF28%z$C~rXtW-%_6VN*h@euv4Lw7PrVY<)v|ylw&3c2`U{(pWBWpn3F& zX3vK@RU+Y($WX@jL=+bK3DZfiku^dBcM7_)*>!BAM04D>E@Hgxml_ zB3@uzE0ZB%u|sRVrP;%x7hdMTHZ=Bn#c5@Om$ZV5DtnPf3}?!@-61f~l+<9M$7D`g z9^MW_hx9QlvSEGNlRsa}QhMpE=!gUN4t&|hleAUabMD7Ab83m{xVbN>So|E?TFy-}A zLaYzGM+|56{S+d2@(~`szQ~y6`)bOceG7W&V%8_V7dLQ`2#bCEiLmrT1{WP3ukeB!zl`3g$Kb_%U%sHjcXhgC-el!r&CSA^JbCzOOY#)Phf;xwA2?7~OUz5- z?MTQI79IJAbGNNq!nFj(IO&YGEZ7ea#UTmLMl+yv4lL*0bj6GPBF=FNi!Aw}L-QUc zg2be=_OMQj{aADZk+rA%bYW`rxE_r83m$tk>=EL%KL08#eY@Y@dr$JP;H5h;B5UlK zl0X*hsif>q5}cbisH|OvLgsmhUQyCVDyQx*5-%PSMSP4+=+#e>@hfueIVqNlbzFoE zEKg5<)m15BL!2mFA(zU+Pufe@$!!wiRc~aFV*0XVtVw&-)%zHp^}jF|xa9hUJbL%$ zlTyl0>SHfb*O)pi%{JP$NGNR+#u8ahCL3%Q_nFX(#L8O;Q&?mtPi^2qj{xYgP*2Gi z5gsrn?V9Mze!)A6arjUib%lq-wv@V%h61s!AhI{gJ_%tSefn^hLi5bvkj^tazzltW zu=3+Bh{EC<*wbwBFlibZdafHQ{KA+x=}Rm#8T<*0p)$!Djd*OONV4`I4dZ5fJKXw0 z{W8@?;+Phlk+$c*IR|;4mUbOBcRj7NtJ@7ym%N5H=h^2H?8CJNal?Y^LD~Q|k;O%< zQ(d*V4Pj=EZOghm&M_W&Cc?h$I4-cPPooS6!4vL_gpD7x=q29^- zsPgHB9I4dik*gBbI2%Si?1GgnyHnbdy1EPsVC87M@N9)#oDvpY%CSGgmAlMcZ9U=$ z_p^+*N}R+6!b5r|V}+F5r(@ztvKhW_t$FstZrdmH)CsO;iG?yUEeR}qqq6vVfN&ojQVAZ*WENR{spDbEm+3PXF&^_A6JZ{`dnao?>3sR4 zI8x#=N=is8y{b}aC$Z{qucxcA!Q0CbJv=oOin8mC|SnRI1>I7z4Kku;2v;UA{J5bJE zVi$P(#^SD;On8Y6gE+=9JY{0y`vP~6eIhF)6Qi(g!DGKN#Cwy)vbrJMl}+j+D#yO5 z)`j*O&twVQBW|XL%H_JXwLkVcL%Vq0r7cxsbw3&v40d z)>T)|Bd2w3-+(abDU+?Otj|4uAJv4PK0hJ7DfL;0&z>;F^)V;2eZtZ=0bcZh z=~%={pzXmQ1hITr|L}^DF!^w=$@oz&?8x%N7J`@lKkwG3-Z@WU z>chPuK0n3!gk276dJnTc_m;RpLA>}fwq-(ocinZ{4X!!h|JcgNtyY-w+x{!-i#}L0 zg!1(&v4_xr*5u*yA54UKxCy(Ar=J-f@{1c17JX}Yp{ID&KP#&P;!Pg2ui&$K9YH+H z*||Iji@lQGW!9&;>3?{Nb zVjZR92gn3{9xIuW3mF=nj=!8a`^luOkMV~%H1;uR+A=zQheM7*G7^tL}tm^9Rf3iZIFfH>``q`$-;-g)=d z8;b!wny1`xeL9M|1|I%wNzf$qSKa2GZXi4wk8phqAzQ|5k#&@qr_!bdcAT8UwAaw!$Ch#rpT!CZxa!rG_7?R*2 z5zQd&lSc*H^ROU5!1ADeV*?zKu-c=hGVH(0GxE5L;H5sa4eR$o;)hPGzTXxWm~@!D zn~JG z_~B*`cN9&2qWHnO7|y^w0(^CKvc3{~A=;Tgu|2q79REb#a=o8v@?oD>i@gkPw;?^o zlN(Y;(pWy|7yGkRlsjyJd{uILNa$H{bhD@*%g^D2ZU$o{j~SB+ zg@Qer#AM2FineGqPNTS8eBI-U>`>$ki>0C zQ>rud!24@VP!Y}7~JM~&s7FrFXF+|-)d zsG#*lSN2qhcQv(Bdrghymw6z>*Gy%7ZwxW3+~_55HbfnG5ayZhq43FQ4IyGW^oxS@ zsz|qwV|u;=!ue6?#qHYo&Xc-L(p)Uxz=2{efkjbVX!5yf>ss7sZPWYJ=JI-9n(~SM z>ZVd(?42LCT{8p^@Iq($;i?HUx#G1y!37NyFZDB6C}V+>)YpdM`mi<`5ElRFU!O}= zl?zPyF|7luc;(kNB9CvoQr7Kp_|(H0{rz~|j_-ng!5*O8XofZD9T9E>@P*pIw!96k zo-8|t7$ZGEBTnXmq^{UC)&tx6a^pJI6Z<#ZC@1XwUha|+7QIsC!!axbul9nMUir(4E}_S60qsu9B)8<%Q06HJ zPjUNp-@Fxf`G8;$ALuq_+9LHTnL`mmmxh1k&^lv!0cW@q#;T0Y_Bor8hOiz z^kP3`qslyAtByk$dQ0%6ag%mGc1}TIi5u7^?~V0f(q7VE62Gtg-nADlSpr{(LnPv} zcLr^N;acj8J!w@aOn%ZH21cQ&QS#!)C}$#0U_8$h!E%Jzf4FGDeHN`xev*(OUhIVx zm+@zPy$hf4!}_y~OD4Wnp3*oW_n}yy#{mcIE`iVZd2`FtJJ!E`e$xj%FkzT}BCo#y zV11Dn`z5?>X8hppa(zmehbv6leNthP&l|IgM@Jlc)Ni1gZ;Ld*i%!f_Hoi~CL|Alzt>BokY%bF}X82ZH$M`y|-4VP57TpH% zt2pXg)Md*phao9Hu_Z#E#Xh!yy#QD*z>6ma?bBDi?Zoz(#=cwh87%F|Si+?hJBz&7 zp34sz$UOqWir0%R!XjTz_mT*Ue%ZQSh}WC_4$CFfdB>~z3AiG$Bik=Y{$IvQ-T=Z?zSUAf@gzxA`$eIpKC?h z)nHyrQGeoiUKA~iiXP@##1iq`yWsbnDG=*R|DbcY8)y9ZPDH9o@Ar@&Q!y&UcxewF zKzhnA*Un%_So))VT)wqM^`(vH&dKNVZCK-x7KdwU!d`;D+oSyTFJ+ON$fQA{yy}^H zEo<_jeN0ozPvyr@g^6PR*s4XamirT8c+pnn^Kh)c*uNr;t69V@v`2Z}klfVTn!L2H zdrQv!q52Q<M30gU74?_f zQ@5j{;l;SHSlk|&BM}z+akm4C5*B;7=4@(*VdB5+?|}mcx`F6e;-&xV`@259drSX# z`LOoV_h8k2%fSW@n0o%$Yi1Id^y>ZUKRQ0wO9V;mpU&KbSBnU%KDDS)CHm>RVV`3< z*7sU;JS6@_`Ref@5{5);AC^yHcP^+;z4d)Dm3IM`ywaW@FEU`+zWR?!L0J8HZQX{t z&jWt~0x7mn&f#jakG$td{s%!Q9<1;ALweO;@dw5JBlgA1T{U=_()i((Ni?f4?T4*6 zbh-L3{c|7Q=m!VV^KdIxtY8Do`~H#1kg)VG-XnGm+sV6dYEOE**5w(zM1Rs#X8b9y zDU0MCU9~s8UvUcWnb{MBiGG!?!{ceWDJgyamw1I)qV(+l+ehC%8mFLsgV=-g>L2C# z6#fy`{?fiO`Kqh+qK$a%FMS7;c-oWKkeK@G*WrW7U3TEdA4sqLQ31MO{Bgi|`$~Vn z)R~mppY}fJpo8`>dy|K5_2~_IwTFl=`lvk67Y0k(OYm%eSd@{y{sgaagYs+tVXGhi z2ur*cPYwKymkp_U;ZOTcWF;iyM{DMyoa~nlQIz~OexS_ok$S?SPrAC!V9~$6Po_7X z$d2)~bu~5EkR>cladHE9!`GoUnyes9d4*|De?+uUdf}fbPgUdP5w@-Nqy9jx1v9;tl#atQIFe@#to zEkF1I*yBZ^7*E;pG=;%+b#>MJj)cLkGAkjDFb_!i_!GDqB5(D#8>0m<5Mhlh_35Iz zglk0^>pX&ksT8LF=W#5?!9+Zh(es;zOne)QX><5@5fG+>_Kl3v{}K+~j;zXs-**7q zRswGoz=NO=+jz&fcLzF<59aFAh0dLf0dOE-eg}!^X>KLo*!2YjQ4t9N#83E){cm<}DP{dJSak1sgaiE_*!lD=C|+g5t` zV{8&p*5!i5d;@u2L!NCg{b(=zwg`C9$Cws(sK4AL0}b{-e~m%;R3F36L%-(L@YV1K)@||hwO`l%=$BCf!Zd-wq$h0j?;Ypxsp(*R&sBMHTHT!wnEvs14^toF zOrJwKb$qP$`2LXpOyocC@Lp(-`V~EVh=(mcF&>Nq&f>v8Pg4IK>*>#je6+_9uiq#z zZwLIax8L}))SsX60?|K+H~$!m_|gh^ysyvkMtj8OFTXOIS<(Ux%YB~O`gjDj{*J~9884sI2&Al1NcXLei4O8 z|3G7HIOByrNqWEQ0?1g9=~=M9S_HN`Odvl`QPL5g!9N7TpG_x@tcG18L9jid;N2F zsIPA8+v9vneW!W&YsaX)AJ1sP4A|o|Wc6tj_$}(Qf31FPZ`GeZ#B?m+-R2< z&;B^l_xGC*Du1hA`H)|~c>Mo#nD+$p&_~|_McSrIv4!F`@5b${b3x+%^zC5zdS9}kgWt}{D;fL>r>80 zOg0}e{^kMmU@DDQT#qn*%@QougC+5X^~V60_xM8)51GfKIg-{>e&kN^{?8xFAv_%i zQ-2sIkMp4fkk-7R@ZTeaN4~9aUO`vy4YRYybADhNs3&*RUv>Pv@8Rp;QP}2-PXW^& zXZ!K-I0$JHvme5z$hO$m|MR=rzS*CBPk*xclrWRQCnBH4Ve<=NCWC+WcWwVDEq0rO zk6okjWW1+e{FY`e_V7y|seG?`c<)a%9x;Y9mEx!WQ}y*`b92Fu_LwQu*lPqlIe^>O zY5(fiDKOUo?gaiBZ@KjM;`oHK_c@FdN&mmts@&|YrU&EF4lj&0`AMxYB+whY< z(~7tG`h&5so{GHTR{)*|xW(hwr&RF^Jv<-RsU65)^WT8UKcp{(X`c{Z%D)uPvGZm0 zrS(hcOZk<`Q%YaTZ)QpR2bRF4^rh{U(jOnx{|)Tj2KY1_OtgO+MM9vtN~=5UH|8`-k6sWqKF%gK(Y>A7E<2cl=s$zR}OOrEmkr(^BAfAu-a%VA{v< zpVldV>~BjN1~?Dihh!A)kA+np*)0`xrSRRLr@t-q_(P%3RPcJz>(c?4^8ek#hxd-e z-%Y<#JoWDcz2Sf9+rP%^e*qS@dEiIjVA{8@!l<@UnB%Pz7#nY;HMLQ_}cG%{ROR&I}rJt$Il)f$@SG<0ZhEXZvp=#aMObN`JjG! zg!uI=l&D&(AEb9F!2VH zp1x)D-yd1Xe=p#dD)0x1b@MnL!NK%{F@ZxRNzLUJXgz4V~(|(MnW-rRmWb*H- zV0=CMFDxt>TQLkIeY>FJ3v4BEypqSJ#_NrHMY3M}x{|wn_tx=xjE67VU;Fz55ATQZ z*o6Ed&wq3H-_3|*eR#CD-%$rdvK~C%*Ix)Y4}QHpf5Oz?;M+X@bB`a1!OHPvc(zY@ zOuplAtELEecN|PF0H4SCP69~ePkUJX()LdRJ^k10Oqj{w4bX@EwaDv3nEVXx3;x7k z>G6b#H~3y}4-9KbgvTIf@OZ#^z~vtQKV9DL!y?DIW}=sO-@_x>?|@rLIur41E96e~ z@TH)qJ!_OSUUe~k98-M1!8+@@g8?~{SI`TXFotN$(V?LUBxbmk>>%mm!civ&H&yU?NQ%deft$BX#Z~M z>1P1G4tbl;9s^AIzvb~8g9`!v$Zu-<*LwL%^}lYOwwL$(s!!DM^`3{Hm>3p-LkN=OJrTmZf^ri3}XfF@kRUS{6{0&}pHrj`6rup&Gc#ihh z)1JN*CjH`op73iWaOb}Xm-0tQjpbu|rSOCj{Qo4F{rfO%Y5hxkE&*)zt%b52PcL}? zC(QV3@P&Xmei_r4URx45`VOWYN#D7)C%h2;!!%9)qMto%{xlbG9=J&!f8RHh{`;Qa zMK|ksU+Lxf8{p}n{idhq`i1en#ltghiR6C979PF>^qe2=@%%U*@_;QK{PKtLT5&9Y z>H4-5{`_{8_ioSs&v!=dL*&m6;Nwf+H38iJZnf`H@86VX7Vsty;h6zks$bWAs{gZo ze1H67jaRRD{n%fuZ~evke9_J9L%0)ogKvCD^*I@3Ozl67+|4*o@#C5FdB8@0=;IoX z7NU&l$ftC@WAL%hseA_C^efeOl9J2y7UNl{@5$&}gD3nla<2e?il^WIEv;V@!0c~> zXZ}Uw&*Na)96a?O>Ypb6T^~n1?&p;Kr!uzvq-_9G|K((kG#~H;oNIl5-TqIN zzmH&XZTG3tU+>{(K8fUhiM99gXXw8a)IqI5+;lMS^qEL-{5xxs6WDtqzmz`Kpr_;AF+RE z59=SolLPpKTuSuc7iCPp2RsI6i%*2d1@KGtDY;*r_4*IRx2^95{ztxj`p=uloB!Me zJcgjV-qU~32l4{8cyw&vl*HFdz=MhW+Hp31Y>)BR+S_J8%B=>xn{V%##+3NqFqAQ! z1$py0+x7kp9wtwwEZV2PZYz*;dwbZfM^6Bp2X1$dzuUuhy+L`2H+j$Kr~GhPE@>uU z;(L3z)V_aj*8ak>Ptu~HDY-vt>ow}T78uhvGfe4E^!6)y_;e3vhAY0f37Gxy3iw%n z%s~8E40yD(?i}S?fxPwq5%25xFnc|>ZK}uo#_`Smu<`x#8ufSErzOAR1Ac>dPRahl z&%C~+@q{qPg8AoawfbB)CGCg7Pk|N2Y#v##+w5*stLmBFVRl5*Ff?i7?Uy?Chlhv{?k zES2Bi9&y?n#wrra|q-_yfa0Y;a|_s>cH zGdNBM;Q0D%eoDrFC6P#Xe@o>Xs-SB=S>bIwe2T~0czxf)Hoj(gdK)j#dU!uypY3zJ zS^HObyy?s9?>68M@bsnqdEJ@nf9HGszH^q^XE#s(K45Gy#McAgz>SGKa5kPx;W6hZ z9&Rn^r3D)A7AojU{kiXXDc1=cV<1xz@S8Zl?CWoTf%c#6r?1DvkO%&LFYmWMLmwhf zpJF1M$JyZMJIdeYql5O*@x?L9G~dG(ZwN0zxxrswoDz7YZ;$wVfG@S@6_;W>f!5^h zyEG+xC1ZX4xfi8mE?whc;-3UR!%ttP`cCxtGXZnFS%0qvOn-Y0WlXnT2L7;B&eLCY ziSmD3(T?(yzsY~XmFh3|d;G@WJa8eN>kETVzDnaowXa{gA5#jy^ z{SU^w0$_`G+@IliJ>Khc=l8U}#qWXFrDVU!;8OcgpFHZDyruBzKS;^-^4_3fdI0d9 zIA10Z;x+11KC55Kf8XUA5BhugO7WZzE=GUYd~oy)DY^c6!t0f9+5B$kEnBII&;XM>|SAu^Y`5Qc3 z+8>l}DR3s=)HhSE0NCcYR{>84+zSU2>0iLv=zsr|^6xKL?B}o6`O(&UC%>iZ8TJWN zX?wT7my+v4n;#bgX8+oJefj%3f9)={wf$nW%3k&Sp>UR}^47 zTW{?952Zhi#7LXl&nA8X=r%RqZmxdmd@}bfX*P}*rTG8TFy$+JiTvXr&s5;+RV?>e zDlPGH7Hr0J06sJEA{I0KXN^(>%Vqke2oBt)Bmn z0Kb=kojiQyL~ReZ=V#{s6lnf^cBH!b&je(mwgrln;)Yxpj} z#{*~c>!JH=d^*q5w;!H%jI~7%zYX|IfliuoI^9^mUucu$-VVl2Sga2iLJHq!LVJ3sg z|2n|NpYz2^z&2mJc&7S0+)h&FEVZxAKUe)j*Vo%A+C5TK`TUEEU)T?R4-d++fBxp_ zZ=3}CB0tH)eeT410Qrl3d+(pC{q;K!fBQTg9~^T`rQ_ol3sl~(dHSO-NXvfSaTsGv zydQBd&eqPwXWZ~XH|Bfpv!d2W*0h>O5;d{6iIJ(i^fv(ncm{d*`e zdcunXcyskPhd=3?+TRPLRp)5WwaA-28Q*AMi?5~jA-;7pc*d(Qt51L56#l>4-p1(v z9r=&^f%e~uK@bw@#p~0u{`rm{5CgB%csGuXAj$j%Uornbe|cK=i(mHio0~66l{>{m+ve;TIz z`9S`8c+q15V8$$_&u-IraI}JM^e@uxX7q=}-vL+qc;xY=@o~U8uowLAcGO`a%<*jS zA0AWwwm(QX54^$jhXVndy?K4by{Q(IF)^OE0k(MlpTE?2I?~6l*Pqt-wxfnISNDYS zAMat>|I6Y*slBP+6>Qk^FU4;T-dunGpTLiSc$Ht_zpMR%r*wb#3-D&OLY8w?ESeH?A;1}M|gWTdwwr_en)%Q`TkwxVgGZL?luqCc>HS~Zt!sAi-j~|@q#r5$N(zZIW4*#{gBAYh5tTpQKYdQ&Ixqi?9vh<+ zx4~=WCipGGR}E%kZ18%-K5CYH&dGi*t4qFE#P1c+jcIEBK1O4%PZ_F8Sr%vfs0wbZ z4QOaczF{V;P%r+fRd+sW=7+81LtXyUO8f+!ef28-G?EqQr;^6`Z!zHubSg?aKDHHq zt4oWL0{y^J+-&dxCtoGhke`-}24#Bkh}(h6q965?k4*T#w1c>GW2UOZO)ksF(!`vS zdaQ#N`Xl|)n|=O@-9H8o}Qn6M^^{ttj~rB)_W_X@7yJk^7Bn%XLu|VXmr4m&i6Gt4Owqm z@sD=-##urmb>$Ds#2*S+4&YNA_&P-JtuOWfZQ{QI2&=&0-N`U%iuifV09Fe0Vz+*R zj9*?1h7EscN$e^%qHk1-QB)qvBH+sMzUNiA;gF@rhs=Ecpu^&E;G6cJvO`PcL5uLi zHEIX#53yl(WK*_g1Y6W6#e|Ad#?0XU_)?AkZ3wn0{g#%`>EL&z z!p{lT@H>=JPiV*=iM%LhZ)R9Mqu;#<<@Y}&kkkc#s~q_JjLRk~6V{D=Kyp3w5+w0K z-S`D(mNh6Fe&R+F%Y3{Sd0EIdXv^L3!|M2HbwpDwBQ1WvOYENfmW+l`)1URW?u7{# zF6O$&WeeGkY}fQ`_sndk_$PUIdmyJxCiW1cNg>O`Px3!`@Uu@>+zK$3YX^P;P0bsZ zxm;skd}s_+;-3bysF~PS4IA?pe{to*E7JIgQtaT#heQM0^pKVHhT*Mtq_VQ`vvJ*s zXWkDe`-Ycac;SV`ix(p+GKh1zCUik2mVx5~zlI`-b*Q(};YVGvHG>%QSjGw>zSLBj zYLwyb5k62eQ9e?`dVZ;c?v3Qx33hV}j$0=0gS+{fcy9H1JF#w!01b%NjX1Oo#v-(g zZ^Gaz1Qki=%V6vxkpfGyH#Qa$j|Js-oZ zU3}mPu@P)ApGJQ|yk58J5v^ zH8|31w4U@1I>H8jTsM{8eQ58>Uc|GE#z2QN(6OqtS%)(sVqs8WLLpO7I&>9el;468 zLdvH4E^o)iB`=KwJsL7wWUA!L@@#|lyXvxqco>m~(y^Rh1?q)b)^A=Gh(eJ_*gzsd`kDHtFZKjN@R6x#hF5Vw9IYCMwb0 zcm;-$p^IV96J1U^>ffsuKI_H+#BaOVcb$`UKNjO%vZjyhq6(xXZJ^`}W#Z_ueA=tx zN^xB;CQmG*-SalUB9KtuBRSTJ6NuXW%M<@Y{8rfg;|%ed?!TmEDeCCiAhI2x3$Ps z#<#Y!4;D-684W+SLcf; zzo9t-mGaYe*H)~p_#EPQSFlZ*0}V(rI+bN)O-oIxrf*FHHZe4IQ=7qVh#LZjB8}P^ z3x09|2hwT3H8%8Z=+nS&)0iIeMVV-nw3pB^{J7lETuZK07O;~s-s3v)_D1;)G(Rtc zH=Z>GN6O)!=ZbJq$V1A^_K=QYFdNbVrs{znyw24NIm+V441d%M$)#g`#kFLuDe~FA zAfB*)*5Rwgu^zEiBe4O?$+f{P$xGvXpds7P5!h;YRL*~_5Z7(5Y>)6SF0-}C z%4l^|geI|kTuKO4;v2mrsX*kavY`JeqPo^H{gjaTRjOUo3ha#QGIYjtv_ZaoJa3nq zxiEL=!eCrxGi~_Q`kgYB_%M5HKYjz9rj$0>32ybuVt(cFVsYhi(M|Q>*S(ja1w6|R z60yB_r2L6>Q;U#^%HsEe1YkYNS9%@II*c2aXZVX|i;zofRe|5!RdZ!TX62tBxzo_* z*1X%fjXYU4iqVg1Ko^$D*e~qy^;5UZw78nBvZT?zEMV1o;QU$Gytj5C6_gMYNEz2f zQ&JuB4_VQT{3K3JlvIe@z}F*d%PY$A?ejXxk98{%Yv9UUaLez*DZhzuX0fic!TL<} zmOKu^FmB?J=uz@B+Y@<{Tlt0kWZ9XFCk6@Hey$Efd6&+Rh}=ErOkKgCIws+g^^m9Z zGmY`&T;P9oCz-&$$kXq_=H#RQQFFv)5x1MesRX~ZF(Mji`)1+t3=??|O6*Q(bqb1k z7SkWdDH(r?T1YXaVjnaVN#|pb#DxJRr*5>4h8PRI-4olREG|brPS%ue9mQ)R(#m*{ z`5SD=F>WmsLHP=8&>Gt(56)qYNoW8R(CaXhGApMyJ#a&FK9B!AtWCA0?{gTa%lc9~ zlbby~?Azrvp$r_KRWeC2rfOV>^@9v@X-!!`$8w$t;!M52ST0CXVWv)8O#C4bBerPL zqwL~DV7A5au=UnkkC2qr*U1tYfDBSnu41WVDznas+=@sN>7`#(UR8x_Cc9ASrVAEm zI`aaq_>+4B8__Q2seTbR#&U^Idwp~6H%_SNoo-0weBbqOV+>cZG3DZ6e{sf#*>%4# zu@D8Y-+@4f`XS>-TvOJD0f(XWeoxmFnKa%Ei*~?|E)B-Pn^9b+&{p0wvE7n#L#YF> zTx@I=05c`@GD5Tve`yPi&Dd2&OL*H(%)>t#C+)X`G=j2R*O^tIO}iF&2KR@Pa_*9F zid~vnhQBOVy(+QCSBbjX^DNiv{WWMl>(y=>r!`;eZo98ce4Za6)4 zaHZu1WJS(jp=&GAnDtg{D}&T=OK zjas>mIV`fb=5<^RGD@rSTqU%}#3?(hTJHJtl#Q@TPJxiTwY(kkVXYg3Ei5LM>%6~6 zW`2CKyrg_#2LD`9N|SakPc6g0dj5;)wSNl`MR^p^e#s>dcJ$Zx419i+;JRL7uU{u( zNxFWBZ_>tk5(T2sya4Js<#kDL;BXbcALc8I5q+NHq1DO>_@s>DQS2Gn0FA>UgHI4H z{k5m3=DZQFsbE_48BMDONFUuDJC-k{gkmyetAgZWSXE4vO1GV=<4EyCyGTB{7!o&zB^f3 z%wGkok9tfb2y|T*0JX2Zz4pvF{f{=`$re>XACsH4zNi!J9A6ZAi27mx^?YbUpI@E6-gZv4i7&`*zpf&IPLcZI>~^IE$p+2W|5L zCbGY_VZ#QNN%_|Nbtz42P)_6USi4nH{@y0ajkc72Rlt9Gz@M#Hosj=>(02kilSe?m zbQ9%8lovsJ8jckK{c4nNjQ=$yd}vRNMMbEt_e~>L5IfV}h_;2gLbiaSLV2O0fZ^KX z{tiGE#df1T+W_;p2J(=<&VA*XTQkEm$KWT`_ry-OwL2wB{?Ya<*VtR09bP#+du#R> zoXd5K1vs-$0r@$BzAMpgJIZ)ugL2BldwV>J+w|;5meEF`{b{%SM#|ZbD*`@CHc@W!=RqIp z8}bjwwUw9ZYw}aCaJu?Y1U}5BH*c zI?6h5+#B>`&bQ09!?@Rpa`MRI2=!f#a`rFV4f`vUGvs4-V4XFA{11T-WnmlRf_^dj zJj%m%%};2%FwSHF+Y#E0^m*WTg!3uwVC~)ud01EDMhZ*7RMZ#IV~eOIz80iCDPtRs z(9UfrpF`d_Tp+*MndQ#}{z<)P10IFI{`AjImf@Ha*oS=BCpt68bso5tpz! z!Xvxxp)vaVP5!Z0;ph1_8h`4WBX1mGe}(dgcA&lm(DDfFz<%5qANI?}%B|l`FOw%M z5AA8?rE;zS|Cx}5$GxMI@fFHH5B1;9Mk1pOZR1N$+L2Q^Es zMzRr>x#Z79DK22;8rlDe`L}Jt4wQ3a_IYR%<&-mzc6fy2u4xnHw43=)XwM?bi(tXy zHT>$M+2OQJl(XMEKsye{+Q2^ZP`(spOK_|S?7yT7>!K8V1jkzVAzajtUp#UJ&u_$~ zTqdfC$M0(J$%BRm;{fSt7pxGAHxzFw-dVi4xV(7t0+Vwo=qU@2C7`!*`onaTv3?k5 zmY0dMi-P-K{{X?MPQ7GX<-kmr)P)S@8B>wuK}L|dEj7Nvhs&~IdL3E zl!bBod(rnm!oT{Die*<@}xw%h_J&hc+JbDChVkUtVVqNfP!Z zh74i3#uK^w#6OZu{wXOu0fm2#=0h?$iEqOpG6nr&_GI~MmKBOgI z+Kp$O_Y1BNHAi{OoW-b!QcXQ}s<59HZYj+kxTpFS03xXQN7X=4vu_1Bf;XDwOz5Vk1)@?bBH33f}+2O_E?UPWBze zKHT)2R)_n*x{rvz(8BK|^5GY`T@yHH+zwSMNik0CmUI>9h@*XYX!5yD1hMthHjo?Y z%H0`0gu_9|KG&OCBS!dc61FU7TS$2L6?aDXH_QJzCK*TX@c(Ebg_NJKhwboGK1#Jm z%!~s`z!T8wA?}?a?rp!@O<8EeLkmof{{i8Q6LBdVv1#L+%A#xH{d!HSoL2pE<$jeL zDmO@2Vtam%6psQR*N0EMi?9+`Gud$7YKqjB&s!%vSBcxNiu9jW4fMm0jpF|>m)#?k-J=SDnzE=}$SDsyX7I#s ztEffEIM4N#4zysPJlW<%&C1_UQsO2BIKq}F3;80Ub#T{;AW33oYCtq7wM8@}%16yy z-%%JfUp;iqVxFWG_`AtTov42_AQ~9y%~jN48*FQPZboBIS}qO&BZX1K$D=q1%ZRg` zd<-~6FbrF*Ad28Nh1}d_h%0%(!5GRt7V0050UtR5oA7uB!2+-_L~Dk!dBp;d&Ih~? zK|d)+0&jVy2U!fZl(a{^hw-XLlRY=o<3;LLvL#;rf-$n+(6>M>M^zW9eeD+5W!Xv*!cu_&6(v-wl} zNT^wS{RWP?aGm!&8sWPlOlH_7IDAjHN$yjA(VG9fzwTrmvO6C9pm#4G&Z8Bgo_L{IltnRX2I zgeU!({;erRKbIIyfp^$p2Ogs(H8(@l(WBX3e2q56;X19pqrOnw$L-_x?&0{|HBm3@ zL-7KV?R~LtlYXFnv;&-z2Tyj!0|#GRuZg3=E|i4_{-#H!tI|8+-_O#grw>Wj#)IC* zUhJy4&rBWR89`#}l8VyrKFIhdNFp5=+PBt!R)f5lCO za~}uKNP+IExLE<`;3jd8__^tO*7u z*B|{TG{QL|xjKidu=VMI z>(lqHPxG=%67haL0{6{!V@<_W*&S7N(^vsM?{c_4%x%auZs7ZroTY4_ciTeSUnCT` z1{S{LA?_9bA$xH*!Lu(pf3Q@}tdLI0O>egC=To#*Sn3K5ZeoL*Vr7AD7iXRRXLX1M zRa1F=N`D+;CZB1ajx*n0T;GNDr|9MHd4m`(k4IkdRBlvEM=NFAqtw2?M0R= z*{FcO)K%gSA1?31RVfn+(bKtls&F>HNzlW=kTh0fh!U%ZNp0974!@o1YtP>vBAQJJg_#=arb3fW7fbU0AHj~o2-Y$ z_=Ert(~pmI7&bPw&7J1k^4tvVWinIN(nOo++FHjUpUXpV>G_G@NbRs&FQ>EzeN# zlLHuWZP_H~Q+4TA9S6=8QuFXly;Yd=X22r*<9@(_f>*nj=gOOKG?dquH=ZZ?;ki8# zUE}Lpm=K)cD6l>*U40)|xRv&GQ$ZKw+V8CI=L%#|84iT@(4_?~4rPI8F>hr>c4bB7 z%8I6y72{S`6dz|S<@nzMhfjH&cMOg^`O|NaM$2DF$w|Yxynw{}>+Cgk(Z_Y}lREd0 zI^vO_0yFXRL0i z4yL4j0nk9KnW~|GC5{^0*~crOD3!SZ=K`+qCS%1g2UpU^-FV!Mdjmtf_RmSVlbTLy zJgH$uuEG7-ZHF%b?ueOaQxU<|DwNN@sg0~`d`4|Nuf$OZ_&-0yWuF+b`;bB|pPSg< zHDx5$#{Q-58zt_ANegk7^69u#y*vCf`F19rhbB(sDeRdyWpkNgdS#Rs{rx&JwWP3Q z;*vcw6U!$?d*C(a0(yT6<=@y|DECHff3y`ed`IOG98CO1;#N5S3TGz1WZQxBa-5k6 zUxBm1PhfIw2mCIPNQ8+un7<8wAm9QHCjM6Poj9`%rok`(=f`6NLOEc{)Z*dI;a|j< zI1{u5ng@x$b&k&q#lLk-Y7e-92r_*1XXt(e%^zKxSL7&z3&7F5L3Cqw+W0DlP< zYlC2`-ku-fF#-Gm;H?1T6Xuc#PYz)I7W2)(tJ-i@pnZD$d{Y{0Scz zz)u5qkpDBBn23i-WBN-m;ExCV0ce;eq9Y%|d92s}7T{+C_tgM@T?u|YV8j0o^{L-u zIGC!?pOojf0^zRqP<_^^;NjCje+BZ?foVajK2_Mr!GQFA!1y13%ldf$9MYF(h{bPr zBOtQ>rv~!o0h>H21mq%MKkgj$VSU4o0G|45Ky`wr{D$v0Oy%DJWlYqM@|b@8F;pOmgrlUvdSlWsK-@}VxTZ50yI_0`0?aYPM3`-da7_T$pmXK{ z<``lkJ@qkq;!g(rWq9_F;m2SCV*B6s{muTTeXaiq6Cc8J0(h^zZLsk!Q!_$0CyrE!t`H*@0y|gWAXS~2dI6o_4Qwc@{|`p`6B6E zz`Qqjq=)IBw2%4U!-wFWIbgm^$8_Og8eeYr^!%;)StxJu__dI)0QeXWe-jfv;qf?_ zO5ioZ}$A94A9ew{0 zPkjwfnD#Yz^tsx9lRW(>HjLPcPl!sQ|I9%y^dI^c!n`EfmsLXlxdyP=_m3W*RdR{v zMdb4yKIjsxSCH3rD!*N57P zGFBEc3C#X8nCm0LpL%<6Jw^Rz{6Flye}G-nx&Ob4L@*&3ZpsOR zQBi}m25B9dk`jzcbuOb$VyZ(^2C1eR?UblNs-dcBQA2~MAC02T?e*y$q;33Y8xdeJ6&i{eJ$qpC?(f*Ly$Dde&Ocde+)&?{m(xgFV3LGu(Ks_qXeUKI)q& zzSTDnAAQA0U#Ikek5YoTest&*pMpqTRbZ z{-r}n^7K@VwenYn{(<=%(ePEL`~K}5KnX%#m!QPt;e%s- zV}=hl{`t3i`~E`bAogE1I@sB zS#RjCtvBFF5&qdDk`>-DuRkKj7yH`y?zZgR`P+2*`8CFs*1l0#|C2Oe_V4hP22B5_ zD{uB-eo=os|LPAeNe6kj3CiuvcOkWm{U+up`bP&o#6JGQ09^BVM zhbdzi`Bd2C!A%jSd@O&!Uwr@0l?-&8@OT|fUsfJ{*u(T^V+H^8uYP}X8OlW2*u)=u zUG0-EanA@nDqN@IWQTOE7G9wIj{`pQdG8C&0Tau&P5EC3y!RU4pF0CS_-~$nOTg5R zFC7;!<=H=4`R(8F{&i~L-y+wWSW@3?$9D!($HOH$!W zm7sOP)OWc<-2eG^xBOK$f$kEXrlVfj>gdT~zPw#;z>~x={AR#Es_^pQ70Md^qx7{% zS5sJj-Vh$4{22l7rHvW3xZ4+ZvGg?reX+B@Kiu2z`M~cG#((b$_&ni8bubt6hj7B1>|3_!HIE;MDy?&)E&yZY0k{FMP;D}Ji{MFB4q zX0F)&adwTjeIEOUz#u@vq<2+v72vgYFb=(@~cNNB50WJWKiJs(=&W)=I5^LI0vX zeS4ON_GjNMHvTQYzH!e16=di}9glWMYr{U?KQ!!)*6i-z?e>X)zpc$|QTY=B#&^K? z1w0TReYKJ`eXR40&ksX;2I6Picj4Z?eZL9yXZhLo_s4IJ+LPtK7ya4t%pbEK<);b9 z?Rz)=@sdSfe0?tpn|+U!KIY@`gY-xA;7k5`@rZ{mhI(vI#%F<&mCxE6{;Uiif3Wd@ zKR4ntzUa5{zex|ez?v40_N(*b`^n&sluwnce)wSH&(lU7d%O_V+jjZy7~zA1z9%FP zzF5zBplo^g_>=MR|5SK+D3AY76CSGsJum-Db+mWc;2+d){@AAe!2gfI`Xhp0r=#(g z2maC8Cqh~N-w*QhSIdH<9gDjDD*uguZ#*>Tp08}_*$H^6@`itUnASJpae@D5z4VIy zlb{dEuRPMvU&b8TUbsm|tMAk)zCQX4%@Llc<0ywTj>6d6_8XLM6}IvxALHNsbxDwa zI-PkFk1)2g`rn2xY~u^RI>RqU zeDqV^^yfb5`=i??NB^alOf6Eb5HRoN;QI8@fM5Q!x4+G|p(lI){(j)M2-7xeUs<>$ zZ1z1uxJ~#RC1}Z~BrpAEdYJ5ViXV@=bq)eA*3t0dQ@y^=27agT8u6vp5&EYnXZpb- z#4~*MXZ-bYa*+Rq@Ob6N_~Hq;S@{@Gi}35yH{L%DRYMrtA4TOSMe^uti*TwNPsYbS zhiudJwM+Pmp*`4prm)!?yfnh>pJ>}&N)Y=e#^3f&%TL$0Xr=3kppWObZOYqtKK403 zUWQi-WB&n!C5Me8s2v-S7M z-F#(Z+Z5#eRO+^>q1{#FuT4 z4%f5(_${jY-S`dCZ+!NrwD%$sh;www}M2^0e2=Zzg>$!kmK+{FVAAU$A#ypDa8^_>M4N zsIO6SR^N?J<=poV4tAxJSj_UDkj24Z_ABsQrgL>gSiWpZ0V{ z?RhWdCraMRkCwhRVe9{szu36_?+DZWmqzH12!2o0e)!9TjXzNRAA7>DcUC{`?~-1# z=X)t%l)RPyko4jI)_(NkKgI{se#7Wbg=7Ch-u!DT&8O+YHXrsAUL%|;LH)H4ds_KF z>Antsw(ETsPk6~c@0f!!5&cKr`VSv}kNHz0{(-Og{@tf@P_ytP9ZLXFf6s5qI{#~! zqW<`Eb>8@Q|5bx2vgZ{)7p@ex^#VS=8}q5(_$#$|)(9V}1TA~r_h&y39cDZX|KUxI zkE-1BUHI&eV}3)X{K|+wPxWChuIbP(|K`VUj^?moyE`VC)y;jtsvp;?o|+C$2BX zCn$s;yvsuW@XtlU=AV1%3vINAGSCRgmvpr20rK6#CV%3nyxT9%(mCkgN9U7u(ye9H z(N~3e{yN&jq+%QYF81$5c&74(pB(Am?M|FUXi0V6%}?D1IO=bIfIVz~(6M#iwf);( zJo%h3{k&LjD@4I^N$NF$uz*+vtojhOLQb(5vvlf{>z}U|)>&qD7czvOL zsS?!R`ckX&w!ZwU`KfM(8Q)$di1mx{wDoMD`aAT+HRc})`mr$mwfhhHkN&v-t-`4i zbjd0|e{@^ghgT?P_{!ku+lTT0=39P!x;)^4?E9I`^RBL|y!r%OQr_^FKbm*Xdnbqb z?+}jfU%^WxPoE%gw`2|X=YQ>zGyg-rN!aAU%OiYqQSUYt9u?|;S{VO-K?!2srAoH` zgR!UKUG>En)`zW>AfB(U((y`%{QAF4dD|cJ`PF#Y&es2b{K@aXZVmF_M&%9vW_sRT zA2|oTBHW?lWQX)D_;hbi`~Kn_;coH16Z!+cOE~7QjQAt`pJec|opYaLZFR3qwgz{jC z^x@hs<=yv19t-^agl9(bOY}Tyig?;KI^y%3dgbo&FyZ;ZG+mzx0Y5oA@4o*vC*ZR# z$tRj3eb2*q*XBIQpBwl?w)6YR&!|n%>FtW1EBT&K-xb0Ol%F2(Ey7EbxB1BYV?1rX zfy?3>E??@$|Cd4kFNL3z{F;Ct>&UzP_&!uFdP$hRmBaY$`87ZP?0%VdY2w57eC^%^FApL9F{QCB+bVBQe8Sh^>q<##*|HqZS{ey%3exm0UjNj%_d9ty7 z=6b9je~kI`@7?&>@}o25Ycu7u{4~RF&XjM8_~&){`N^DwmR#t^&&KnGxq0_I;6l$% z@XzJSn|}^uf36>?F=+7D^&i2w2$$!>o5M7=zl@yApBD% z>GRL(rd(3g`6q*azz3WE;$P^C$3Grl^!JD#>!0>M$`5pXW<8iEeYPHe=d1lO?#%Ed z(#OA8{*j7*L+Ibu!bRcz0{+59oCEUn&HRgd45+d{zXa<&+ixyTaO0f^Zb|js@qCOTc6^eH`#dcyaxYm z_3VV_GwYQ9XFY$JbeFe}t(V#7FNL3aK5c~H*XwBGvG>xvdwzY4<7!`Uj{khi^wsJ4 z5aV@(_I=QThy3{M=%Kq`ouRz#?+cH3f4Vv1FOk0(pZkHe5gzFKE7!c?pMQNM=(|<8 zMft4)uDmnvo?rbTV6K<+)$IS`_j1mEtp3}?r~gYUeSvR0>91eoHi2jB!iK(W!uVo8 z{LAbIKh5x)BK}F@*9gC(1W~?SM=O7l?3oHLQG&4dd>zf+JfEYy-B0uUuqbTT=Y`Mu z^2X1OKl|TOB+v7U1;VKk)ck+^ew;Z0fyo=*{y#nc^N#EJw8!{=e%@c-?ehijTJa5! zSn2PF#)t9~gjwGX2>68G`}sdK;IeSF^1B86Y~Wje*9E@a{~z-QUw&keZxyEe-T^-m z_~tKb0^goDw6F5xWBzihaI4DNe(|+ey?xJ;&qCY&Depc%*JA=l*9$i*w_V_aOA)^B zHGlp5uI6C+x<<~2bCM;O8ffS#IyQXKb{fK)(@`V z&j}wFEC0K{`~C65L7vaQQss{h`1n?Tzql}93X%Xjve9nG*DqLPe75 zctXIh3p4-BpMFv6{qI_xgTM_s8vfyf-v2-6xcc6Ma8dbb0q?M<*MC&Nw7*7qYu^dq z&m~J$w{Almfwe{G!^0+5xcO@HyH?J?T5-4+PfnDvs+G6<`Pss$@Df?ut?@oy2oLHO{%e?)k`^2{M`@H(}#2py6R|+r;euYhfNjk{_9c4)pbBWBq%9g4)|zc zu8;EqhCfsJa=`H0lsEpTruz1qy^lSv!ujKpAV2PSFMnf%%gP(>Z1(-dXQBIrdvvtt zW1CI){_$y;BKm`Etv`F7;qQOXcU-M^FK+yMi?6HvI`K{o{OJLo5XOJ2GkyPd5By_= zmn+{L@Ot4%$}bD}fwR0l{}eFgOUlRP9~ZXr*U$9rH-FwtBPLVIty!(Xp&3V5z?on)^M zID3EgDFrqziuZp)ec9)C_|r6%zcJLe|J;g%=h1r%@h6TEt`R;w;1j>*{biScFA;8+ z=$qkr&tHUVga=jm^27B-3dYX94+c&nyrqIK`C;{e>Azw6w^-Qvw?gIN7s7^c%6)#lv<>3VW>>hi`()J#HR{VH ztmTw}p6K!OYYz{T?EL2b5&w81l;{4DrWS%ce8wW?cSQV>J};yEcY^$O7gxCNU)>tb zZ`xw>`){{YxbgUGkpJ96exkCcf&TKaFK^Ea_iFd^waK%S!-Px9?;G$cVfy=Wz_TBe z|En)22mF^ueSIGe`q>{A#oIFQv41@Ne1Ck6%3l)f0Y8;&JU-6|+xVo9`}z4wkmviA zE#imacH=!UDi42UhF^~OjDK0WOh5G(GyJ8(F+b}+iz5C&`v!tWp0)}6&^D&?KYkc&>H~p`ud_iM%cqqU0SzrF=VST%< zUh9|S@j2)*?Wd>f_d)G5p7`%(-zridSwuVkR_Z>`ex4E<*h)gRF;UGM63Yzp@i{qbq9@wdCm&;Rp6 z`9p;nkDf4|{qZpcd>@~s?>$>_0G>ud5N&vopbvATxKjQ_pCQ)UMiGWK#O_icqzwW* z*I?;T+$8;V2bXhkl*FjWG$PdDL&hnlrRqrN$}UQZ*>W{~FZcaP=!$h@lW~$PFlHp6 zYQp;-U8tdV+EtKyi{Fam1!qqjbG^|DGlyRROSGv4(U1`{T|&%GsGO7!PRa_jp2%hU>Z%;A7x9JZ9|D9ixm`F;w&3>M zQrEbgwpc3allqbqFNe=~A600m8HLiA(&(1arIDpVOI@kZSSZe+dPQK1!e@su-^h&X zH`G;D9c2}=q--GS@;CK>Nn(CTVQ^Ua#?Qx!`!5me-x_oeXF#me;J^Avm@|^Mw<$Rg#E$#PV^7Fnc8qAH! zuV2CMwaS#8TP=p#I<=~#HpKk7%~dQp+&kD zf~?vfx%hPSu)_V}s?9s1jl3>2IV%Xp@4+bQ{T28IzDwf$6eic2d2a>p`6x-Is5C?6 zOwQg{Vdc@^pT7Fc`PCV0`gp%aT)sDhO)h)>{pd^82NLh^Fnvo@59dgn&-TmYvhV-c zRJ|?Ihiyr`cc>$SO&`_8a+bGpCfAd(L$)0zM?H_DJ!@9Z|fC@)ElO>PnCN=QYuK|{e-5^{LjkMum0pVWj`xt`mk51){RHDy(VY& z!)B~co06k_W@qe2xmuMWb<}O>2lk`g1L^CpeQ|rOeRa||U;0ROs%M#S9eJfHwZr7B zedxgdtY3yrZie2kMcb@>t(o^-BL86}Xt7V}T^x(x;qIJZU%lPI4ptBKrs89(*q^bB z*^l$|A8hTg_R&Ai#rk4wagGo3NDb*_RK1OR_N#oh9bB#(}@|RTQW+>mJ#QPaduBh{=PPsAAc$9>( zQC1&zL5{?Ai*``2Vbg~naL&#%j>wU!qIIo5`?)qqF5Vw7z6G5laUMJOCx^aNG9=z> zw>neK+E&VnIe-rwa=)BpX*pI)MKIG#5ZpwbW z89Dk}rKXXX_offK#QG?Q{YcEe_acXXo7`fRFRE<467N&pl%0{ohp315pl!;2^e=9o zjc>Mntr>lr8aKv+wz<01KlH`=<9^ghj>LN_Hzha!J?bk-E|o6oVt-H;E-7DBiv5c5 zpl>!#o2s`ZBi9{p)-KC*KJMS9>`XtJRWE6k-b1;maVtu$Lo%dZ$)P`0o;u=rxkTsb z!$AE&j`AetCGD#XIG#t0XWS3!t&<#SRph7iBeoy<;&SBuJ$sTX<^-JQmUUMKM|Z1g zajI1{_JdR$676GrXdj99Hkw?I&LcxYt~a!Uw#4(JG00IHen7uU$`_SlKd^QWWWO0I zUsl-`rD~1GDq(b!;_;w;bvj4tY~1wtE`E=4)?fNp2>nf!C&lB&csN_>G;)*!r%LS4 zv-Y!g_eu^K!zO3v%`OsgYlvZ!!w#_@Y)X#)Qis)J{YojI_M+E*!_%XDNO^aZo9IZ= zGY?Ocr(PR(#xcewmmMeSN6y+|a@q6PjyCnzUu3O3$5_sIJO*GtkzdZoCwV;^-Yhp% zpCady$-3?LFHeJig#H{cBKlX!I}wi1+!-p_BSxSveDt9(_Jtv%hG@r<)6*F}6}hak zyRu6&s+=?pv3@Xb%nq#gHa?sKn;he7ZRa|*I)hEFE1X9bKhKV%$x&}A8I#+TKI)|{ zX4kdCvHj?4c#ecBH9E%|4w|Zxt16QTg9+td*7NHeEM&uTT&bwko4D2|73-7y`s4;} zwvdns`7heh1EKyNtGGv4PL4n54bPd~(=!@WC=SwpF*II`o9cB{BO9K}slDo%oLKu@ zJ8i$X`zonj)%&jA zqWVCkvC2!J8XKTnB&LBK zZw#6{V(#YcBPutmtgLZXIy^bN>hKYVZ$4zRA(h*>XWz(sN7pp&$@S#RgWE4{zjWcX z3$MNDdpBKr)AttMzVOcr4_kED6^DIWq6we-^6YVMp8Ls_xgo)$b=1&!-%cvTs&Jm} zK9jHp3?(-?z=WJy z2;tXu-&j<>Sk|9%c~DYrACyx{<|tjQBpNDA6OID((awFj&-Udu&dgF*#^}^#jH-yw z2unJi9q`-A)0}sdpu8%9BZkaB* zrB2$*Kd!v-=ZUYc1tjPD)hQvS^D^wOeWztk>c5YwuW~p7*C=QB z5G{wz!sXDOlJGR)3zZ=B&DGKL5#x7~uzrf*2z?#OnLc6ycMD(PxLRh_zKP1;739J7 z5vKf9VJkm*tnZusrgJ~7(=)|CC-k58vZ8PRAo1(s#LpvG?Htf9-%IX;A*> zLEm?U@u#N)|2MKX-xJx|4PYORH{Cw~y-l0$w{-zQHze7jkgGZ}w z!^5UJvA+Id?Ey-^KVPl=Bd_rOI(v{8*j2biy6k!apYJ3apZ!uv zqDzB5_5;1jV`GSz_G2W^8UkG|tp9!2JFcIrgzfqXzc=Dfze+%KNbm=6T?V&A__cZB zt3G{o*b)1m1;V!9c}&-@Rl?&uJDK@)KfewQ@?FBM;!Tqb#5*a#`l*Q{{GGl%;9)XR z12sjm7gzcdyfdjmxGC`QciLH2Bbz%JD>cB7D=I5&}OOAL4O^tH)4x017@A1zA|I3R!{z1UJ^8tVUMZmoC zpUbhR)VQm^ z^|{%@g#F@Zn-~wx7$xgMEqdzDnnhc3jOh`10M$ z)h4LF-yGEI{AG@-y-cvT)!(1|#7MsMj^7V=1^v7;VVU?%p*_4aK|g!zduPHhE!<0m z@6;TFP8;EW6L6@9?l-cu|DzEuO1EL&VbCC+t&hB8;Td6>(Gf9nRx5W`z|=Qgd8_ZJ z&HZm;c-91=4}Pq#MmW|-eWvgDEqn~p!@_!ieP)Tjml8w_;C3C&p0`WBQgzL7T&*iw ze9*UTs2|*;{>4~7lkCIKj?B3jlL~#~Xl1}dgFb!}l1kp}bDsD!#k2icm4ZX1D!WsV z=baTb!gmMEZ(eGZxBUeE`l|485uf$e_{`r%@#6W5@5b{NZ1Wd=__yifHxAX(_iLSl z_)Q4ocdkQb|K@f=@K11)WMe!v!pxN^O7|&2#B83S zBl}qBh&_GGP38o2ywXIaFNFFiA7AgP>b(4uL&OmA4gh368QOD{a8dZsfbp9Kr5^@; z?^m98K%h_4))6uO7b^FOfPb*B|IJu$z}xQOV|?5^$e+l*6M%OnFh5QR^1B83?qH9B zy8Xw7{w}Wc2B5#$_Q&>lFYRIez8`x&7s+Ss#XB8} z;vfAk;{o658%UlQ!-E% zVEJn*e0r7sh_e31d|^J@e3=&HXT3-H0~)-49}(soc$)OZxIMyLADQ3r^|4ymu8)s> z%wK=E(m80LcUG|8_Dasy+eeS|@AP?;!lISJwL0$OkbmbzS^04RXWt1i5m z!d=QU2GB)QeM}L%KjNJY({=tW$CY03*?(>sFz=*bd?p3VJ1Mx{tPc1F$-~!ehofV) z(Jm=>sE5g?g%>N2%^+feas55S!{h(orA6ueq6^(9MXNBaHsMlXdnHU8+(|4@lK1iDtnIO`u4TT*494W zL4&<5CIIsh|Frqb{DyDyn|D;y3)}dE3(_CsnGtTD?$@7>hW3LOseFukBg}8c6a9Z8 z$Zvg$_K(UJBm8LBewOST%eM%dJocO?Z1yB(3}bh=612F@$7Hto{6N4} zo}H|@NcL8FChj)^&|2t9%!k_SW^QY^?Ple^ijs|*%$<5sxw;vq09~`&; zZDHn;wHLfHgNspr_)U0O{G&sAz+Dl3L7iVA{3RtQ%O|GGB%Lq86d^xax=fyTOmyjd zr)&oCj)~DNze z0)c6NjLQ+`ofS)jH4Kh;XT@F0-4Sqhf3^G5-oG9We9AMHae3Nj<+pp*uh*vqe+JJ| zS;NOZrTHd2!PTy3Q{pSEMDpn%53Z3f#khX(2K)utlm3}~7!Udy+ove~W}lOOy|I13 zv3mCbrl0Za%;?u8rB8oe9}KZpTX}G-9~|o^<`MQgMLM7>g&UM-4*+5R z#z=qm`Qi6p^Xr|&9PvEm8RdQ*aNA$~e7q;Z&nX|{5z=qCzjrj$NZ#hxU!^Y^lL@ul6xJw2gBF#{P!4 z`cU5anq6FO3Il%z{?}tpFI7K z9g|0@rU?+lJ?pCWb=Rc@{9pA&^v1O3Hpb=NBwtbeY#hAP5&cW7;7V( zeTPpP^uI%6L<6;dc#y}Zpn>FX(|Wa7a@L-K-l=r{?mi~Z7oB#!lWn-4k6#b{e_Hda zEZ%hi!$)__pBV9pN!u!HF=?-P$v^-5iOxY^d%<7N_6Yix3b*V0P{)G9_C}v+x2bP zF@8O}CX_$*SRZrU_@h4SW6;|A1YRJ0G46@*UgA#`w)F{&e2iB__lSm0;zf%@Xt81c*=#F$?kVV*}#5k4f8Z@S92e{#@I46KrPuLeGt>!0EJ z?|A-$K_309C2RG$82Q4c|NYAUtJ{B8!C$YB3jO1H#@KA?dWJmqG^KO{XUrQ;@SEIo+dfNl`6kL_}r-cnuw45c-h|M_j)q#KA%0ZLP8=i zpAQ-S<8SlsyA>nDc!8@WXP6k^(}h1zVbSfs_xnBDKN6$6PUmfXCPw!};XM%+eeVxG zhV(}~Oo+)nN%^e;CMI`L`TYX!FXnbxyuXL`5raF%NBz~?zcrNqsxY=@jfZX##(u?6 z{)XH9{`_N(t9{8o{Po7xm-^@Z=aVA?pY^9kdE@_B0qpM;{&wJlCrP*AH-`9_wQmRh zN2(O_PWIkB)PIchH&<@>K1*IPo3VFW7|$Jqn}oj{jW6qwjpr-F6{Ao1>Y$JDUMYFw zgPSAyc_V#HSzAAd@jPAf*D68axjGs?U;Djf!t@#9omKO7wEcXAoQL`-18se`-yeR~ z!({f4{QTcN;C~l?x$;*3g|B#jX$$=%<5zVoEnC|8D^kGrC)O`##(^@6Trli#1}Hq_BKj`J}ZPY(6{ zMz~k`ivvDnnzx_L?{?uX@pcLP|2V<-_lihvVk8A&t%1KE~p60TZ(sK0XM6TcykJ@6Yiuf$}B(T z=iAld^^KYLt9D^JYnk}y63^*0~2h3IP!Dv^>KhYoiC%zc_=L})1fX{ z=nwlhvnP0p_%WuvF=l^d_}UA+Kb$HV2#h^r%=OCfe=6oC_P?$p0!}-YJMZMk6fZdFeAc7NX~+_};BBMLaJB+w)SsztSpvE=&>Y zH|;ll@Rtf3|AvEoOzDFW7SWz%I>zmpDco0H*O#)eU0*KxijTp2o30rU`&*t@#QWhf zI&bpuD*)KCBi08%pg7AbG1V>kp^g=-WFc)c4iz zC?>Py(|}*@^6xabEciR^pCSFG5B>^aGi zdI0W?Ffr@vg|Ang5Ham(&rJbW%&thvoDZ1w5P4evK3l!y{n_rri@5_e^)$>@=FPXBW|?D)VAxd(>lW)IJOSC4Cw!ffnOb`m{hSFgI$ zw;s6j31`$m4m&rNIRx{0w*JyCTwu%GDE%E7{^i`!|K(iH{d0V&qfA(a^aGcUeATeZ zVKu|kBsYxErX9mBA2w&$)x$f77lx}(Nss<%T$M1D|8?|x_$yhNsqg!PMG7ZH{)E3S zyV0Mb5VqdNpOslP$wn&<@rY#N2xb~k9|JJ!ic?)b<{`SZ(r~5C1`n+qsxjv8yZxgK z@^-HXer^2@$y-WqD807+Vx@!Jxfg_2uiqiJdi{42sZNeTcmqk&X#K@wEE^rU-@pzR zUwm;(OQ}>WZoc`55mi-!7oX$k=%bHLj@Uq0mf~9|1v``{C%u!CF8x0xS$Im)b#Aiw zl%)HV{1F@9E# zcA(qzV}}^8&0s6nrSoRb{^Gv$h+ikYq~72!Y=^eNOCaUlLCd!%od^_8xP#T#EPYMl zla?t?Ubk?nJc<5cC##S7VdZE?K_VpDi5+5$JyY?pF>#Ts+~SN}e|o47dlJvj%5`SS zbp`#Yk!XK^?KFQWM(ylRF7DszOna=It(o&38I1pw)Rv;sjHo`! zan8;!5yl22{L1W^3e(@XKF*OikACtbFn$=fn``)o;E6atQi*HEHc8HXFI2daYn(2a z$y7NO*DuKG>&?W?T%EzlnmyzBm$gH8)Xrx0Z<#P-L~4>BT7Au#_Gk6b-;&A{m127` z?#xr{M;x1$x((xT)H0N zRC(%TUfH-CwsK1(S1lR(5nuldQ(w%ta@|3Hs&d3}Tb99;i~Ebe*z?jBLd^J7k+b}e zo_sQ-ODWAKAIa%hc1NqPPWot*wX0qj`;j`d&ePwxJ}cKFziNug)k-cErcbOVjiH?B zwQ{A5zHZ@Kl_4>Y7(c_rOSN+NG3Qey;#txoue6j#PyOSD(ek?dZ6N-X?$64yR1lxU0P&d>8hzx zT01KfJ+nyiT}e*oDThCJ(iOJdiFcQiuFj-uWOBpEq^pqhx*v3woNJ}t!Bl{LuvEra|PDk;IuIz_Z9@1mVRC-9YO5+dE<1o%$py5=$UgzqR z((8g8^^S2R-Jhy_(SH~HM1QPg?DMG-N84(M2koQ8J@BVC5U1SxCP5xPR8LlP@f5l5y3#91^|}w6wBT zf1C{y*PoPg?pehyFdz+5@Bj^yp38LUspYtC{8()e+>iMHh2q0B=ghj=Dq`%Ilp%5M#+?0joT(xW7t@6t~@+m(=>lPeFpJeMey z2hGv(5+%yZmmKMSI6t_c)TvZV^6FPeg$hNcRML6H0~}md>QWM>;9{kTN(Y42*C>z0 z?o)!ERGO!x$C{4b*5CO`?|7KZQ0I!u@8z`Vd+@^O>k0aZU(z7_O2EV$X%V*1SBYno z3ah=2h}T6M!ZVtLcukZ)&9hy+rbgw*2K*U4a6}$^);a>HeIc zA|A*>9q)I@v(e#)o9ae3MwmuLRQ<>UTSzB*HWOs0HWrhHl0>Sufy?^yrTh>w2C$L%W%Tl<)A z)Nk|cVQpx}3-cKuL_8t-j|>E+|Az5z>}&qNk1B5!{+kl?@qk|m>ofCXsdStFUoJk+ zgb&So`8$M}TeCb&o)%7o^)bGqSA|=YTTtmS@uemzn|kQ{pZQ1|A>y~r)3M4S?E{48 zE1w38eTvG*_F=rtKG=7)aBSZ-!e)QsKO#>XAmTqx)A5@Q$?wz{@HG)$5aDcmN7_>o zKW-29xAx%QvxFZ|f{6b)M#s}+BM5$~qw&GC&oJ|$C~WiL#F8JsUY&!$b9L-hg1{?u z>{WumGgMbEK!kn`X`lXYjr09I(sA{9sO-mB#P(zSRLl|mrTuY#Q(@~b<;%ix`F3F| z&;E(}@7FBOjKUnXqz6Q6gj@Mf|JMEqX-)xJ;0 zez8&5_KU>x)26)d`=gBiXcZY7$`cQ=SvZadIW6K753)=6f*^m&7ybO~+Eyn;XMMrj z`}8nhh(BA_`6FP8!1Oo9H5sfX_3?jvzK`u>^7zjJ@n2Db-plw+m0WB;`e*hdp4=qi zINoZV@B>N^(79c4yG2L=AuBHX0B;SXA#{!?GNxdpme(KA_mB1S=_%C#b&8LDWBzzy<4?Z|DSbaRlm{0x znE4y?=SO&$Zs3^Dwm&4^a#8rMP(QdW!dySr3e#qY@{D&}p7DvxcSe}<*w4x{pQeky zpAtm8#+5qSev^2O)Q3(8JVNCSb3MR5DNNCeD&L~xFCFs!jW5OH+bV41OMSE_u8;kp z>8HM$3@%2vzxl)bGyV8~qj2m$Q-sGVLGPvfG|9!~OTt#Z>pQ;h_Whrq1$>5NAmVk8 z(Xp&HL&RTf(D6ctG>`W7-RH~wfV>On|s0d`n^=xuHVF)#NN+_^5E_Wvpz7M z@%pe__<2>f^+)v=f;{$1m5;}_M)>{Mzb;dL zVy1kPaBP1yvCsa@*G^%Zk6bU73ETDJrL+9~rri%CU$6F?e1GkyJ#qWnh2#1=gl+sO zUn?BCgKq-cPJ`-bvh#?Nr!P5;}Fg94xUob#36KVZJ^Tc`Y90smAO+guj#FB1QI^2q_8`Mme9 z=75PWiT&=1a6^QNCs`6Uf6n56Mftcr)CZ31k1_EuX|Kh@8Bte|&xeA6$|@nEc%ZAAjOdFP{8Dc&hRT1Ps3i(*@o-i1>)BTzbu^)-@j%2?fT32ZV z8WS*letTv7H?_HF64qtd5%C_UNdC%ziN9Dcz8;%7B7S3o@U8*xF8&C~?`c@&=P7UH z;m?qK%wHYpe}0!-QkMMDItLNIk@@(=fQk3WdUbZd#Ct^l;RuT+jPvnVCaVtU2;t>A z{x!(A3A0|BKE6o2NU{e-^7WDYcDwm_TUH)@lT<#&-70JN!I%7edW6bFtmmxXOC8d^ z?06qP1|84^!msM6uf03s_ZKDQ?DrSD1bx^T+F6cEe=iC60A0UWzy3Yo?0S>MHIgN5 zYW=xzZ~y)B-$({JUTKPweSev<6P4of@C`q)kNi{kc$vbr;ZlT|uZI6zRR6_4A+&GkWUY6?c0Jl%xLv#>0>4oh|E15+n^Syz zv*$cawm;JU9(r8BTOO55x|DAW_yS?(XLG>BE0l{Sj|NPB!|;qkGOFDk2_0Xcd`{3_M= zP{5~)-z&~_0pBb>?fLU2%HJpcI?2Bgm0uo}|C<(e{Qc3uA9jwfUq9P&w7qbx_|F79 zK^T441WbHBZ1G=SJOR@;!`b+J=bWeYUHaY(@|$V>=u-Y?Bm9Z&g&CjgJaoQ1LwQ?2 zSntciw%!w;55C3cTeG|WJ>@5K4!TMFyL8lL%+btN|9iYi9wt}Fkr{tYD@Vi6_3Oo| z!2gN(x%+JD^~2h?uW+aMwm+RN+$GF2Oz1-4g*qPPkk$d&Yr67W z7a;7(`W4#~|1x_bzeL#NKYyqHy`tRA5x)oRQU2dNO!&R>bIRNPgWoGNpBQ84$RGLj z>hVgC-_eDttgQJX{DS0Te(bM3_jvoC>We2A{LJr1t_b+@fG>^k4CymWyq8w-w54VVwm`C-NI?$f9~gg{v96SshS%xX1^L^;^)U0|BP{ut`9MOJcHS<#(exQ z#)~qT^(W>p&EUoien$3*<$E*u)eP1WyOI2iO#g`2AL}b;Fuor1v;ATI#C-hKFyq%I zY~x4#wp7^03tWsa8xN2Bt1juY^?>+w)1vyo9TBGg%Z06ea4D*v_;lFQ+7F%|Vfvp$^5E_WXX7Q^ z_M)FZz0v`JnLoV%5%ycCV+s(R`4>O`t$mxX_Ur%0Jv$jA%>1|eZ{%6aOrHC%YRTID zSLbU!o}Dfej)-r^eo1C?#QoPZ%KbXvd9VBSwnzrLQJC>O*&#jm5pGrf#DJFx&r^QC z2s3{S6R(c@J6rE4zgD{A^0S0Z9(}BZhAGc}F)lw*I4-|9Q=a(}moJKMx)| zl_2U{siVn*8GFOj&-}OYw12*ET>t1y{b^Kw=39QhI4)e@e)UsdA2x*0Ul-{oUK{>v z@!A*<{KLires6})`fYsrH(va>f3ZFK<2Om(`b+=H!g2rRM|{f9ReiDkg_-_m#~=H~ z_N$V7tRH{0`iVzK{V(erL_9*qqrsshA%5W)+i~~eY~e1`r=8&Ha?!h_X37LLgizAy|D4wKekCW-aj(_*M;&mI%3Os z|A@am68L}K(Z?Se7qxe>cyargPljoKm$0?p#X}Uk@S81sF;7%K$8MwtG_{GJScJcCzca7zZS$h2o=2GgIoKKdJDeam_yOgz?@ z|6GLMR{TBYk#76;^$E*)+ENIa~S!f`yT1;UpH|3#j)C6-S!^7w$sUwMw#x24WO z#N%72BXbM_*Q!4;UaY!e%y<}nTQ`3AgTh!j8d=P{=WSo|Fxg$WPP~rtKv@=w&yjaqkKH6PwE_W zfpDkFp5|fV)^GI}8$;k#5q?hcO`H$)eevV@q+OUjAas@RG9AAZaJTR};ZFwqe%pV@ zCpK>XrDyqgP9GXVZF+wBSs!nqD&V`tpQiJ>Mtu5W{3S>F`=_Tu`*~hnFaBLh5d671 z#{4#6;}6-+zmMrNVf>MwrF<;kC2aDi>*C%zbi?8}qLhW=kH4DW`3w-!d% zegR*e`1t3eOus(wNR^L2|ELzeEbuE7U$9zb4+)s|U{7lg&!dokGUz8>U#GCi!+$2j z2OIx)-|^S8$vOvt>0gY?vj3(q@xn_gw=IQ5#Q$bJvikx2o%OL5rf7}!udIhQzyBxY zS4HJ3ck%IvX(Pn-qDFkXK7jAaV4h7IpYqJ-xcq8i<5PZOrabF)%%^XL?|CHeo;Q6f zYENxc-_u2{m%=Zo421rfI$HS!5Bc%f(Q!SW{u&?oj!0jB@p2}rtoaY0pG^^-9Q2*} zy#KtpInvMiV)~hXOI2Sy|ES;QANAL0Ein0A9`o_$iWOerF~YUV)@Tkw^mmqy)?eax zt`c4?ozSIE`1R!o50kx@`uLW|1Pp(nWQ~7=aJBlsDDcsbKbwB8PYuF$eS+U3`Iw)m zt;Q#wO}pgdcsApOJCq>e+04*!v_txw?hb!HxL0UD{lS*jpKnNCqW|MU|KAd>68>ed zPxUYT_3O!iX&<82K5#LEXE*y5~! z+0SC{ct1N+JgblTS4Q}sQNAuyKkHvye_K=^^{VN;Y%YRV&|G)D8ef{J9HJ*Rv%=}v$;s5RS!~dCo z@%oht$M(x!|DO9N+V6jR{r&mh|91a3kR$T1)9Rf2fFi0-=_(sxDm{g0yZ`qpWj12g?DHZPfNa%!kzSFK_{Ctr1+w-06j+4Cb)=JQ^!oAYJ#KVO6 z^6Qkh&tr)%Un%~@fxolj$4^tu_UREW>v@yi&%YsDE4+Cq|KTqxe!TME z5BPB58Oob~oF=?f_<_J@{!LT+?EaDGIc?&_&vSS_xLPt0{*8_akaCf@ce z;WxwgkBGO;_`VY`@wO>{qi;@ve21{fzu)-VtJ?kh8eR_i_6(JhXSgLwetXQ)%|cN|B}84 z*(GfAWBC2DhxC6av@eVQc8m9?&jvntrSgXN(s<)fO@V)yaFwt3A?`E2XOKEmxfKhtsbz4+hw_}mAmj}YIVTdt#hp2_#;3Tp43j;qh{ zgtxSrfr52>0>OgSp~^N%6s4S1O0d zbro`5)%uHbq>|LH9$avvEtI#Cjq~@lEXJ}FyF}gG&DUIRpj4^IOsB-1q-^!}8(?NeAX;5*sGoPgL zX8qqONw-W=EopjKu6~Y6##SU_^+AJsSI>^#V-yHXWyVslk6c^6t50|6cA1Ws{Z3!G zu!gGz8=hwSzJzP(NVdElNoTCS3<6=_HI^bzstQU{ZAeng&aL zPwZviA6M2Jr0-B&*11B3Kjf>+Vi)rDWm%L1Je?8pvB4&jFLF{+Pc?^2|8~j5vFyXMQHmYj*}O5$4)LYE)cL;uzLR2dPT^Lf-*O zSqdZLomxyrJjj%{tSOSEXdeOg??aZcR+?zhG1POlM&GYlf3-Vq*r}#{{pyL@S*>=j z&e)6cv3)GhF&=mH7?1PxKgPWooUO-tkjjpo(oJesdln1VDNn+V#NS=5JgHuBa%qwN(YI! zwe`YHmuo<#fVRL?SHT~d5z z>lgMx#@dZN$Q$mjeO3>CYI>Sgjy{oy|GXrl7d@0Sy_>Sn{7^2HEaQQmxIO)~4_`x` zep-7<>VJz)k!ls6wIid4dg{a@^{OA}MGvVp(!;pZ?kXkLAKDXR#y!TXGT7Qv*7;OA z3QG6~dSl#^!Mz!5dYW`Tt`EI2Mh|sZJ<2+tu=bmt{_H_}NFAD|)=%^h_n3adqpm_^ zA8WUrr`>V=*>R>F%}Pb3E{%(|r#0xeb~8SyWDL`u7_Z1+(_{U&{?cxA7`FBqw)Poj zUugU7d}T7HGSQX%;L3FH;B;_4&F71PiFU_1ZMiDhs8*J>WfhhJZOT-16h?Gk_~3bu z?Ym3t_f~I_8>N(2(zArzDLR)bpZ;}=TrGK}8j{qdin!aQ^IdDVP*px|?iezn&Ozl| z=b)aUokPk)%N`HaxvIr3byaod%E@e%`m$0vq^rfdZct~{)q^^!;&wGvC%GnVcA9hp zsk~BIsn_MHt7*R8`{HD(bgm%T%H&ubr|Mj@;0E)?a-kGO68?zvJBB-P~U>{+pzay}(#p7J7D-DxNNUcEE2dKSk*sB?x|5$C%$H zZ2UhB^1ol6pmPwoUdI?u%HU#zc@9-4%sFVW?pvw<%MNLerRPjd%40T&cyUc?6Fv_S zABg8isYCkiipm$2*U!cr5l;~PdhFwf_*m$A#X}cQa*6VGZ->63aH<3m?}q*KvmUy5 zCY0YLVB&LNPrJYToMusjuzla3_#AVEPYd$IckNVuY`_m|(xR_Bv=3a2@E)3kls{4l zBK`~`e|(TXQ2Z6b+rt!{rUxn1|5&AiNxN`S*#`qYULPD$-;#idcQR4<4kf5d^4R}} zf&afv-=&&=jOWh5zH?IltodmzkPx^>N5l6^exdLsO3<^+UzNSWp@HMy;9rfD3-b6& zN%`1crVE?D5D%I85XVEFD%=y=i+@nR`Nst+SQ2JTA>ys#Yp(=(;`7w&eC*H6SMz7u zgYLLJONFgH5AE#ZNo^D9tI6oo^43FA#g(t^^T} znZC_&$j1|0pnS~liulY=>}m58yd=WJ=V?*haXhOT!VgFWB7W-{;h7!|xE{>b2O~wv z&eXhuuqW+}`;R@W|9@2dMd1^bpvGN&dvEs8#rK+_{DlEOtjW7l`IiHxeTz}cx!b! z(r0L*=1aAX*N6K4B7T+dlz@qE)*xI~pCIC!VL$6X{tQ3%XZjQS^LSzN=U+=db`*x#^cc{JQk*i`9b?)ee}=tz2EUAesi13->(F5y;-PZ3J}3xtfTR{-YpWwCJ^#^ zB;6-}@2P(NZWioC{IQ9Wdn(xLp3nGa$@Y7)#{<^SavVK7-H)&R{^<$LH~hU=>G{74 zxG%obIQg6Sc0Iy=%!k;1YlY2z#6QJ;r|BF-Jk3TO#|Hf3;r_RyodJ*3V5}5>dW6?S z_`R%G&61DXGmZL`pzY4{_HPWIUya$xukUwzb}~UYRsO1g4-=-n9}4X`Nw_F{W5C4g z>JUCP*z=#Q4}YzZ{uxRT{cqPX_DAYBe`Gwmh2#F?Kh}TRR}_x>%lO3o#a?lLrwYgY zT_F5^`pbNa^)+Vnbw>KQ{?Z=1{^IY<-&6_O;Y;4%Zw@yn6NH!O{Oth~Z<77seolJe z`P(|!+m<>Y>^D!x*nabc&3@bJ!TA{Bo-p35SE+EkUQH1$2mfHb!hW`1U9QeD9=4wL zx4ux`#uItQ-sEW??TOpRcv$<;*C4!$62$&gQ)(l9bA@Al_X@}QdW20M^-ap~ixHpt zrVGdQm4)N>cL`g4#4p7jhXnh9SIAa|KQ+;Rp0%a!BcM^j?c$FKc#d$YvhxRfKK-M9 ztM4A^V}5K^;rZzAQr7g-AFg$Ae;b9ZzdzsCuSel>?XFj^sysdnas67Wqhap{KxXMAvTgtPHJ*-z1) zct1sbR{y#$>oY{*^Miiodn$dY&OwwwI|gopgz>%+#-c;V?fUh9zdLBh;WK7)tgmvxN!ZNkRi@(%){*ZkoFzaOW1 zbN~Ib8-jiL{@Fa~zANa@V&XHx|3Tnqak1dz6>S^r2VNlA7_ZFW88T?hUp;68=KBd6 zqQ37Z@co6mWUF6?`r$8EUB?9sf1&zfe7EL?+~zt)zLU%{HZ?R z{e{cQ?-Vfc72B2HHQ;a8`tRSB!uZt+uNLp)0UspH^+})RI|8p&dtzKr{|yr_4>~NA zU#z^?8^z0GKHB>Ir1Ff1y6xy~{hhD$j)zIQkN^GGUjtsduaDodGPI{%_2FL!hx&*& zh(G@($P*t8zQw1?>N{RM>etWA9M$MA^47krKKRHRA3Q$7Cr|MH^TnW#__>4t|f{UD^+Y!)0)=%JbSGdUUt!(*7lxrfA_B6 z>~F7Uzwdh2-@W#~wbmS3{}1+D%JQ{I?)p>xaI8P|EY_=92M^MwbguuklKoW;vCbo% zo$?nY-kbOV^e4kHe$DWvH^RG~qUp2DbT_ljVEaT(zv+;-vgu6UO!kq!rKE2N{@OA8 z;IzMpeb|3H@(u8}>+{A}FJ^oVd^(%%-@CrwM`)P+q-~%C?|c#VeuHVAjeGEU<$V2q zmAgLRFn+qx#+`)zL430+fA9+0JK-Nu!#@-G7IYZ~aC7UEfcBS(U55 zf$*WP+yNsY-~Oty9?w&1ecn$)SVCBa88%M&9(-9o4|^g13HqDR-C5J?dT*1_w~lnr zcui?PqpJOrpH|jq`cP%-$NGr+QGXKlL;vCI{YtgBo_3)-nmw<-VbAr}%wJw_&Aut( zxBcvfqy1RE50bYi-<9M~uH8zre%s=fy=lpw>(A|iWBobp`D~Ky$zkuYe_ZO9X3)5A zQhC<>>kD>d+kR&6$20d}FZh&=rGI{ZwRa&*7MeDt$rr(^;NMH~2U~L6OUPX>&hstS zi+k70OMiBw#&ADmyqxv9v)WUCHU7|d{;c+&(P;BL_-V$oVgLD+o!`F!u8)t+dSSC^ zUwk%boa+UwMjq_?R)al<@5f%fcT}f~rfo>Kv9kS$&h@dzv)?1&zFp;ysQQnES8{xG zvNbq=v)tJ7PjO4qU%F>?Th=bWB!Tz33Oxq z?vI~|obtG+1aH=tvVRMG*H4;2_~#_~aI&X;+gs?nKGb&3#j~sa84s29FUJ!`C+iP> z4y^rdmCO1wTlT^KU$f`>ITI+iQ)~UYp4KEd*28gqCfirM-}@k*m6R9D1~l##>d&$} zq4tk|!1_}5Pe0_C&h@73U!%YJF1!H#utJ>cXY``Gq+qjN#5VYum7jTCS%2==D}RUi z`djhG^}36E8_Z{^#>w*uAN;}pR@M(U4Z0JSm-Sm>y-3HuFDEZwuIWALTApWN&jFqC zO)a_awnOgvQ>OhBjW*xGS)Xbtejei*Hxagd#r@ecVb7m$C;99o|KizYzIUY8{$MJ) z(I33?yJbB7pVaX3e*D4euTA=n$DavDd2Z*!&oKOAHr+#&pIfk_-0ORm%OOAKJDe-< zdq}g+wqpB<`#IA&n~iR?ai>4K)c5A3zXD033+ZG{cyAo%X1LU+|FnJwSoPo2PUChXr2jw$i}CYK0%4!YiMf_@}&FS>AJa zyvvhWf5!76%0vC#q&?E-`ZV6p)9f3kd?V*e$lvkJvYyPp7IDY*@U4&dUMX82!JkBb zKm5;C|7)=Q<&P@=aDOSkm)G(;`R_}A@Xs}V+5RoF<(GMT;%k3U%Kw&X|84L9{$nfu z32b_YD({E8#1(Sc`Uw8XKQXopKDFw*9_}PKKM&AFcDhwQp7nXw!!-NGJp*1%`CU#J z-2&M5cVFUl=x3gJ%wg8vgxd$Z(|LY+e;4QHuRp-M2Y6f&H}`uF!4aSFcT#(q zaNL+qmqT2J8mIjQ=!bpFQ~Ob#h961xeBSC1RvS9Qdp<|_Pdu9Wf*k*7BQ5Ja+h2!0 z!<*i2l{f7fj`V(UGS_=#@2Ya=o3Fti`E|XY)$kdF(KY!?;pG*HKaBna^q${nbDs!z z@#pXh;fUYxdlH+TVK}EpdpSK_INMtUXM1}SYi}JK_DtV&IH#{4&h}QqVefHgmh}SP z!!g|-^)RQYmEZr9?OC6bNOvy3$#Bl^QaJK!eH=(^`E7*%E6eXcv%UPkp?)7OzrUzG z+a7Jt(H{S^^XqyWUcYQJI@i;f%0BL|K7Ll2&wWWFc)}w2uTKfB&h|6DKrMo;yBU())y;!#@}*?d6;*muKS- z)}Ea0t%kF`%>UQjr9b)AYVQWtH<(U(1}P(*>q+$V93EeL>*r{X^k3H$ac9DOo)_W1 z%C67Q#lN(&>oHh0(Z1vriM3~ZVNbSy3HE#%L_f}_z4wl3u74kTBnNS0c5*Tip2ye4>5RYR!RRL~@ zZ57)vTTjRNPj2P-AF}TXoyk<^*tW$0qp#u`Gkgx`FmaTW$os8BE5qgH?6iw=Uc9a^#^aP@<^ZQ&H3Bc;;=iLPbF{*Z56CKk3=rl5c~%$lNTqeJ#0Vbeb~VXzAIHIrJlb&Yy4k z^BewL?oXpP+NLsk1;5Rid_%}SHCN<>5p<)!Ey6! zB5c1=4#C>Va@$#E%OP{v?Q0#M++x$8!EvYYAk9_K8! z;x2bLy7?VbexL7=;^#P4@b~idL+OyeS9Rm}&iAvF&E=ip=9ip}(e~MfZ`{ZzcKyxo zqHX#I?1MY~vz3R&iZ2Je6%alWC_@VAEMGXF3Y{CwHF2 ztrz}VztNd7ijDpByV=-Z$ac{v_K#y5^^Hl>*)H@XK0PcqcNPVjRO#*Ur_WoCj<+kL z^fkiK`g;BG?uFdpZq-+U^px9E?Vsks;tV=ovhup#6=te4URv4fdO$?w!ajzx%8FRli#7-&J`HyaD;1 z%8!{+&gYA2c=KodME>N-EnbRl$er&x0N0ZOtc*va} zE5B1^T$?|5Wy$Ynin#Mv6N~=cmF@oy;?MnGFX@l|&;HH#3Aulp0b{0d+9wH}Y<~ir z?GM4({%Sbe-vVd*BXHP1j?Ah5JvN>5S*?%HHuP^<{e8%?zV@?z=63nE+47u8_@7t4 z?edmwuc3eXHKqLTX#+!swjOYx?4S&eeluJ%~LV)5#mz?_tv&ZT{(4 zj^Ex$Ik(Ovy~OpQ%Fa*p{ow18-0`tlJ`N7K^Rwom+pYRuFV^7qoYu*vCs=*^-!41d zGBRs^`URZ4hkZYr&iPi8*ngyhUv4Vp5v=}NIP~S7(TCaaa_BGoy;45U zsn?gczz48jZjHHi!}fRcu&LWe`O?HYKgNbm`|3x0%H@z-9<$)ckL`B|zOLG{|5*>` z{=xj`{$Y>$Y&!3+%zxC+H3+7{pR45~uSopUKP>I@PiuPRt;C(Vi!Af<#Lho6J&W0N zXMBMEko|29mHBA1@yC3$_uN?YW4!aj&2N0T`JA@Th1k#eUj#?~l`n6}&40*k{}bSk zu<4v{IG+8ohRXS|8h@6rNpky(1#t8i&OaQ2zg6q+(-h7i{8cvH&N~aQEZE$C8h_k> zDmT86|0BE;en-{s`*68GJiGD-K2ge^Zm@Bd*AjeZRQ7&-DgJ0LpF)2f9OY&C*uH|> z_ve3gVY&Z!SGDK$+Vtk@cV_P|rXY{|3whQrVCN6oK0CB2o%08~?B6cnydGO0cQ-`8 zjsA%Kq;Gz*zWE7#%g6NP^6A34d?vxUd+purm=dgX6P4~XPEARMhgF5Bg z*h2of@0RPwJqmFn=zG1+=~6mGMrCJzm-GdtM*! zuWa~z_;Yyk8{vPzgP-NgFtWzU+Y`V3IKH1^?-8~Cmi^w7d3%d{dJfB-?>GT_6N~wd z^C9J@RlW@#!2g)aH^N=w8&~=F;F<7~Dmx#tALh28aRWR&>`MBw`Z@eyE4=xM@WyA^ zhraf`euezU*pr`L<9q6(%6x?1t*rb&OTGjSx%xwt-65u<8cX?{UBi!`QsSfB6y z_3t~KF0wn5d@24cA8g4-TJm*CzT=!a|9~{<+B}J4)7heZw)wuN0wfsNzom>Zk*Hhk^r&1o7=e77?V&|)wrmKp%#&3Ix^j`S5a<06lrtb{Suf@n` zRz7`G>CfL-`N?#w%aQZhwQRVn}e_Km`2ORpw@AW#;WBleP(y!d}J@?Ll~{_`q3zw;pevnqdZeo3!k zbie=7(m&Wo=#(#EpXKx6EZ^LcuS;_CzY5OzGrdth&OfwnWB#G@Dd&)u8649&A8{v~ z=OfB7A5s0O=!L%XGdI97pYj;ux4$}xO=taBb`ih&YmkNhm9Y9xY_z$a43_?(FZB6r;< zeakDCkDT@YE7N~GAN&>)A4`2U&iP_pd@*0l|8KAprWGKEE`QI* ztS@K%|H|?={r}bW+wJrJFQ3EpyOQ~2tEq3g{l+=Jc@h4T3pW4%d@jsme&hZJ`Ul{v ztNsYQ5C2Q5-1%wp^+|txOTVwBFK7LAE&a7EeL3_O-9!J)`EzNFPhS4WBY9Pe*S2_l zi#N5{`+-Z7{Q=Iu%+80%JlNueEp`pDtly9A%zImWpv4chxVM$QaV?(L(qBdYlH*(3 z;`J@w)Z&pAa|u4e-Z1@Dw!fjpn_IlK#XDNu*Gm7?7TcfY_!hKyd5eb=&-n9lfA!lO z)4c-L{-TCDoiBc+tS2x(vE>{wCy?eBns@x4_x{j{A?aFAt{hr_!Dm z!(m@}wtxATkF>uO&h}sZ^&{=emQSQ_(wB~0A3=Gx@A%J|=s%N9=X%YCKc}I>rZ4mI z7TdpM`MSg(V7|ZimhpbT^#*3bSEHkAuQ#B+@5{5k?{c#KI{5q&ymP^WWxb=#9Mi2P zyy>@((;1)VQx3lbj_}4m3Eg1VBN)O~JU{+t*k91HKO@<9ettjciTU~8uY3~6bgs9s zoBc%%@f`#?m()Yg>*j9-obzXY7Wp}j4+k6QkM64FBimmEJD-0)9QVueTy!(L-csh( zw2|OFKP~HpOsVOc`=fGy{N~Em=N$aGKCPdeUfWZoS9@znf7mC0LFm$m<2T|WPr_J78F-!`;o`_cr;!~f05;x-}cS=K9V?fTVcCg60 zdvsY};mhdgo^!$I=KlMFf(?(Qe1jJv-v+;`%76DJ);EB8P1m?@@&1!3nVpH1-@m!^ z$BQdJ`aD-_iACmLfluICNgGx`dDG)%wqt`;l=)-1+2jqdP6BakS#Ex^ z+~@f$p9+V3|FcIo_hZkk@eN;6&aWtM<#UmR{7KIl-F&|cx$OCu+2@zwcTpcRNY8&~ z)2)N&v44C+%)dH{d|OMt1$#OCZaBhU!ulM8=tufoufz5h@hksj(*KkHHoDWrUhgCM z!O4T{GaFxK>o4={77r)B_PSD^W2=361@~O&e$0H>b3*sp z=dm6Ed|kn2J*jo@lEmKM1|P)!WMubM`On~f{9mm6TdaR#dEZMKb-uswdU}0BWqp|~ z_+OuRYvQYKE#-dC-pF`e{{66mKbq}|G%YCfZ;`KDXS8kwVw`@OL^Q(Laf5@Ny zR@w_3{fp&475!ZP18|hT@=W|SV{#w>AVgI~_xG%bj z`;#$;*YlZ;+;_K=tNbPKEZ8#CEvbBJ!Oqv!$8R2Vme(})=QhN+5_m5Dw^klcXEuzV z*9MJ~d+`UG9{G|YZtPEmv;7^&vVHrXu>X+{l=Y@gtKpmU5I*bo!=dkbW|lXv1sZqK zwVh5EUtht_KWr@PTb)c9>13}5!Cvo|!*Tuh|6zN6ZKcd~zwCNDhF^r8n;L6;PiOs| zKJ4FC+4Y?~KjZxK{M-)5`6=&e@!l36Xz@dd&7bW#@+S|c{PvvAdQqG&H`o3_`5MY9 z>pP_~v+Mf?f8^VoKk&)Kp}P-W2*>->hu}qUyib_%ccYv5&#v*k1K!KI8trEj+(%!{ zYoW%eZ+V3NW&6whUtG`L2Rpy;-l}hS|Al*m*Zw{@+aE`ILO!N5rfKiLpYp#XmCrZ+ zxy)Dpd`V76-i$oh`J*ccdu5W7q(gGoJMdh5ewDla$rSu&)bv{3i{U6wd0*oH`IIvM z^pmx}bG--0*1V_M^Z8~W{z;X6zL|kPo_~hMl=<&xCHYX2yM9B5v|Liq=eukuM0S;${{Wx4-f zh>q@c&o1j>yr^JjCG7e8%F2FE+lxQGr@bG!VV_;)cftGMmsEEBgz@;Vt@g)WHm12B z`50|LciE+5n){896zn`}U`#WY;Dc3u1H6NMi%<^uM-HH`&Z+O{`EyKAU*KY zYW!;||M$v&b>*1m{_zbp{(F#Hzh9~R;un?wKX04Yt%j9H|MxZ6`=PfNap!c{`<<6p z{`l2nns0*)t83hgf1ULdNdKFI(YL*3`{P^oxkMdd-~K(@-wkK`&YukXzk5wtuj45x zKKt7opYiAX%t-P5kdLPu;Mdjg+psqR+opB<|J#_(cKGIk9j`xgh(E4BZ9e(BF`cor zzZ72pI!O6a337^Pp%KJ2afamB*0_;>&7vier!Fh+Ot3M{LlT}emKhGHRH^gI5^{N?da^I+>x_pQ%zKCu5&j_Ir)(;M~W z`Evly=g(wxzNcd>4zCW8q`Ay`QPPrU%uSd40xE?8=k=nbwBC+r1 zY(I}{tkE7n$9gXKPp^DO|+@f_HxB%VV0cdXKC;eT4Oou=gZv=w`zM_y-F98{V-C`{@6;_77+5roEy+zoysp z?7)90J;*!ojaB{%coO<|9$O-CJ*WXxu0Vbqhi2Azo87GeZw$}f;rT*%{vtdNgy+x0 z^P@7=sk#}EzKF^Du-uQV0DnpGPFKjyfXpQqnUr%RHpIrz2aZiJ_3Ol=4_rgA6 zE?bxR`?Hw`-qp|M{B!54>vxl#pKkx^<~(ud@n?3vdgh2nyZQKFi;XACofl8mnsdLO zc%82v?0Ox>GZlH4&2&XPG4EYlrZ48b2W!VLZnFuu8P4%;OLFb*Ysn)X^JDyO&dWEw z+0LpKAFZ5n_&w<7{6#zqT6WBjX?JtIh4m?%G-%iiOM*rE(Z{%kW?Jbv6 ze@c(_F__r=4Z%@Pi{N&BXiIsQZ4u!k9@{~br{!XOsN-fl8{wRekrp4V+``^2)+>p4 zdI{(G>NXy`Q(HW@#pWmLA1xl!=w|&yJeHgBbK*DaXShyB=05CZo(!)*cM02~a}TfM zVxB!OcD3{)o<*(WD_T6^>jP@E$?==b~}5!UH*E8{5z)8?df#Kbx!N) zoHo=U)S0aN(BFgS^#0Ci1EWvt?>xCP3gqt(`5i@qNk>2TApTw4xp`DGzr*{k?yYFiw@`NLfa$dqSB{a<)p1 z>nXdb?ku50t_k&s2H$k(V5jFfjsH1acmDK`Dv_6D)|TXW**NFaLyew z%J>%MT&TJ5bqaCLB^qQuQNStpUN6fR!682pL+jwUSAFTz$~dZc_HX~Mc>kB;X7ul4 z(>YgfJ^PO|#62~<0spef%6GQpyOP}hZS1}4P*ZWF9iECmo}JCV_ij1=OW>S;p9ORN zw!o1;*Gll-^?PhO|99ym_9LSkY}THbjMw{Jo%&1IhyFcJENdP9q>&c+`eaZ2 z?9af~nyT+w7c=nZwJxmB9N#b;@mU_;d*<@!hod}{?`_F~%g69jkVSaw$7hpVKV3NL$M!NE zj`s3fW62-<`X}&%bFGj$?2oVU*_)58zWwG>vdzo^=K+l3{6e^&VqVO8%{8RuNcaqRDIh-ZSwmh&^x zBM%@8b}q5ye_gd_d>i11&+??mk0?*eH_G$aOUrwJpB=+N+!?U;@2}j&<(Tf7lz%__TLfH}`Z9kvHbh^AzUlqP zG3CHASo`tqRC}IpVb8S-xX2w^yFmLJ;jn-473H4itYWTNYhnWacUGQ@yo>*Bl~18^ zX5#;Y#7h%vZz*XDdxl>OM|jr?v3*=t!^^gh%#&K&-{Sc#Ufklbm{Do6h>2 z%>Imq%JqUKcBuWRKLCfm=cnZz=cn~){JH%sfTKRYMf)*5t4M=R-poGO_B{ee`*v>d z4!ASgAL{D2%3SHU7i`X--S{ItXCfcO_L^${f2Do?qVzuMq7vV{8lP)34CD74(xfgj(;~C@f%-% ztN!Mt`ZNFLC+FY%-p!`dzV(^yTmE7HOI(kvpR-iP8Q%b$^XL6XAnve z{<6y6Z*=kJ`;Bq%?M1!0zvzea{l#=R-(L*DxqK!NHmAP}NBUh$#rpG@?i=td_Pq`9 zt`zRa->q!^te>2}xp3sqwK4`N-^Eqm|JCffySz5W7WlojeY#eN{9`tq;ddT$Wcb~1 zg!g|v`)xb^*R%CwedYX_-^ky~Dc_YapJ*EA^suH9-Id>XXOVBJ z?DNbT{NJnW|B@fZ|LMwi!RzsVl{D$v&pV!fvpJs6zXfo_cOLb<6#faD?)jC!SIg@) zR8E)j`dJZoWXn6V=VS2E(%&WQFO&W~gwN@J5RUXeT>JESjwW@)^WIc=8Jo`k9X^x& zYsrJo_UZYS+vhx(YHFPES$_F^SO~}YV0h~*hhGavc;)sdSv~^}`3Zc4co1Gq7@f~! zUH18TtPhUoF`usn;UU84d>%FZuWP70{|(}g=QY>nv;1OhKG)*g3&&b~UXLcjaeeT5 zFaSq?=l#-R__`XO@)b!g?@X+|*T1Z9d4)aS4|L%uzZZV7^f&SC!~W9p>t)l~Uk5+ZpDY?@`7Xzw%h&snDBo}XeJQVHRo}HHh6s6Q3Euhf-xg*FVdGx-{n8)5 zqw2f1$$I=zKa=$ zT*&?Qbas`y_Qo*s->vLg9g8@(pPuCF&=2{sgtxz>Su{?2^LkqL2p{^&=d|Km+lp^t zE56|rpKFaQhtEv$&27cEHpOQ-tY`ZKap+{rEBN9mqzAsM%ICqBkLQ@~wXo;kA7E3b zzWsaXE0;rV{dU=+evQxe8}WSx{cSMC&^YCCmUobaTz%sYR^RhI^xJEFyb$>;wp*%w z(=&}N(jzMmR{tPd*0;TcTz%!i#y17NicM$x9bg~rSN)}Ip?~VxWvt)3Ii^#-pDpBu z_kJtF`@i04a);KEbgdE3=e*X)8aU#2tqSF_R)yggrtro$1V?zE* z1?cw^{yVfe-3#Fk{_6{N{NL`jM{Ve?g-75o6zsf;j&(2o8!LC=1Ncu#a)xIdlB+)@ zvHDZt(3f{6_J8^MEqHRZ_piLzm`oBcX47fU{O9<$z+rC(ADQi6=%yRz{}P^ty%>9D zeAD63H$KO5Mts_TkdWE_*koUO+u>|)N0J+UK`XyYTlq6R|GVP*(90+v())eVp!*H- z)6brNnbWvWBOizSN0qNfZhgPI^3E#%^CY)@hWziU{a9hrPd5RZJ0^AlSWs=Xh;1NcwIhR(V7mT#P&_uO6d@`SYTm>Ho$JNPBo zTb$zi$`?z1BEElt7ZLuuWBI{3H$TUB-xrF0j&C1&Vc+;>rTA{Z-ue{ZQ|~R;8cAC^}YV&_*@q#^dDeu;(YW5RmM3tevo~ff6jHB0zZpQH)d>^Tk+$9oyoA} zdtxo$v4pq1#`WlOh4EQvud3G``hx{ zRoVXyKLCHZvi}Qy5N1eleQGtiCp;Quo3hd+ak&bdS8|Md+O zw!DL#OV|%zSLOci`GxqGBwmzw(pyS<<6k^C?yGMubC(WQo{xMjlBd=7v>x67pIP}k z@Gkg*%4e=9_4!+sZ-AGR|9$m(bJ8ltg2U&t=`4@s>~nc7hNC>p&w4oLXD6KVWBkF> z(ccfxW79d8PJ7R9hzrFBpIVS~ zm$@4kSAN+8B`tF+&*jCF<#&B$<6lZyM10%nNVg%gj&z2%e~a+y(}f<=U%!#LRp|X= zkJ{ya!2Tz$pWpgmxqbyZx8NZ1tySN(;k@6Awc$R4J)ei(j*jkL*!(Z6-1G4=SMqO* zxbr3KO~Bt**|~ND_~U-YvG4ofXwN5ITb`Fg-?8t?qknYl`&{B0SMB}&S4w|*R!y(0 zJosgmFR1cAVD4rw@)b3__GTn|&czu|{?6r?PM(qYb>A!J&(b39dH?>-$6w@2&FN;aT|qrn34I@Q1$kd*Ez;c9J{y z&GPvQo6hiylialtrgPkLLMP8>pLtb_x3qX~i*27d{ItYgFSnteub0;MZDd>L{hjSO z+UJdrWG)r$?UbG(KkZS>^?>^ec9b7PmgQaUw?pp#Mjt^x{x`b(Aohbl{X?!l@C#~x zWc;(ykNBJWTkL1<#eU`;iJgmQUo)wu&$)Qh@W=hmbH_zcjFKCf1A&w zd_w=Yih|)#LXSPJ@S}K;-;6h&55IwYD)x@8*KhUL$W{NPe2CitKf21*A0WQazwiTN zI$iATnNaLI7sS3d&jnclcWVNbZ)?f-wdCW8FNe4MBfN7t#-o1@o9^@v@xMjcKf59J zaF_I5@pJfS`paATna{g>OmjbR6>VH+dzeIeqJB?#19N>i7w*J{?j6_he2wnO1v{VQ zeS~v=qdxuL$h+Z4-=Ez={|?9Vn{zEj2piuU{9m%Rr@?Ap-qzwN#250{B43ZaJvIKd z@J{%jDm&N1|C1bZJ+x>2MEEQ2DRTooXLP^%q*2XWuDJfF@BM7(zY_f!#D7&$@5~=z zE+X=1ug_RQeG%VJtK8?YY53#(``(W7zrZi0PtZBH=OB9?8ynZGZH)a*1v@AF1#=6r z|82gb=^jt`)$p4OcFu!0!V4=u?VII%i07?~VVh!<_iw_(uzj}f^|0|@-w-zNBFzpb?`j4lh{sYD{jad8>!XS>59+xr{A9rN$m6wi|a*y`Wm0q=8L zZ(4(Me>Q5d_{t(Z1(kCZoTJKAumrB%!V%;79nbFgcO&Rya~!(kyfbgC|C2FZJM6}I?O?}ichPk--7!8p+~ScI?`d(w z+s|=rhTYk4F6X%|x#eiuqaIekb78k_PwzeS|1G=lic5Q&^3V1nUdzom!;b0CY&m8& zKbfOkBfR;`@&hgQI(;eIA-B?yx|L=}$C3W~MpV?%@eFT{H0P+H$|*ByWJM#P za|lmmcaLG4o_NTw)tYK{qx&B_YIIM3A-4$f;IU{Lf4s}WdxYyRIq5G4XO`VyIXLr-%ib~RvUkiWyS}=6N^`L9 z^s=kQN`ML-0`WJ#>}EGd9Rs!;-|3~mI-lo)u^j*3)iD15sS*4is_Ynphw#T3gJaGv z&jWD{kiCZuHoW)9IlSKxBE0_(W;6Cy)Rf9rez5;PMt&EYPWfE+A#d}0kk2Rmzt(7T zFMk~O`|x#@opZ1Ro>+Nzf9a1NU-|Fgarnnpey}n_(Hf`zDtuYr^2_>i=pTD=8N;x= zUUTJb$b%hI!WH6(F(q9%>^lb9{O2*#18~mIEI9I`{+gEl+Lr!s(s#^H8oW4D9q|Y%T55h00_5FD+q-j!JWeQ?&N8#tuz7&POLF=&S0mgJV7`H%9mycfY|u<0!C z#q6WJEnm|^w2iYpTOLuqzlS~37we5#pVmjzr~kin4?L^p&oMP~;4Yia_-$V~{y{k6 z_k13LKg6c9KUm5>&S%G5Er&m%5a%4573`NaRL1O>{y(M+bVrNN_8jp!26Q`omdZHu z{~-IE|A*kn|3`14{bHM2<;Hy;_8Dhq;$4Za9xV00r>J*Up>O{#@kNcOFB!0M27bH^R~WEl=BDE>GJ-l&AJB->`3aS)RFmR>DzU z*5^jJ%ciru_OQ?8wHJ=^`WG6P?d{n5(6og3tiSKD=^UeoO6Tr|=qG-=$p2_`IiTG1 zgxoQphQF-p-}akj%w{~Z{XUs(>WebaUHj@X26>npW}WGo&py&)es{t-zt%_O_u<-u za^Fei{CDBVzxKzs>`zGct)CfiuAk{})Q|1GAI|N4CcKPIXZ-Wn=lBQVh~M@#4bJV$ z_7LsM_A~53lkXr+wQ?4!;|Y z@aDe@pTVZHzn#QBx37LUr++3K>G%IhFMuzo@mb%VPg#F)(jTbjm*=d`^V{}!enW-V zBsP7+a8BPkIHzwboa5UDNBV9cG3J+M+&HhlJMcw&S^W0$ynjKJ|Coer!XN2*&F_@{ zBi;#qfXds8yszs2F7kQgJ@gH~Ke6Etz&X6{^dr3e!4zyw_>{(yNpUZEm@mG76{{a46{)2Fozw#w5IYn?t zu6^5Mwm$-g{cpXyw70KvOn0>QNRkiPv%T1!qP(ZQr`$i^XJDMy+l}m(HB|ba5&SuR z+e^gnnDjZ6pT~5Tukl6s`u`_aA&>OE@y?=e8_`+*2iQk>J4U^O{-+h<9HTzg@S_Vm zM&156((AqXBt&nm^4sAB@GC0+6#s0te__+9Z~2D)w?9$lz${Ms!^lHldpqH5&+{SV zmbd3$uHOl8)URWhoA!Yno!rmS**>paCO=lm^UtSN) zU%nm;!TEYHg}8J7J_(NaJ--*iaeiMwld+ZaXZj+4%B{03Uy|g%Z2eom>quv=-(fiF_u=Z_G2VUHcn6!#G2WKP>V`_& z9KiqS%8qdz#D7oa#e`pkpP^HYvn?-W3;h+G7;^MC{(npB*D}^QM%?&rZiwd+*!K5~ z%D%%j{&-$#uPNa>BvgalF0RE7F9QiOD@&DRCmpM1TTFiC65AQ{87~RmZWz5z8_yv3ry)6It z@^12*!%rG>n7?;cS^s-%lN#RL!)x$AuCnqG{8_%7v}U>Y^S?;GEyDEYsMy z|K08Xz`oMIom10e`?dW=`}O~2PJy>mhPpPtkYmdEw9+`k53xn~qsjLs`HyFnF@C!_ zrfZLxejE*LGsis#bdH&}{=ZYxbMEnFj6w95j-g&hTCQ)jdG3ep@8(v1FT57+tL&K3 zY53zConzvs!&g=LWptFrcT;7@pe`hQj0xRL!w8mU$}eEm(WA;jrfz<*kIj zv4;QAg7W_wqrX(nDSSxonCUM1^>M8uZ%?fLCgKl!%KdT?`hLUoeca?4Ut>S%JLW=r zTYHLv^7Scw+H(v-*pm+?R^Rf-@h>30&^P{Z*pB#r*)hA08Qwus*VX(uX1JI5XH;&+ z^utkKju|#TF^2cbHM29?@a9fFP#4i zV+_EFTS$N6^Wlx8zgzXSZ~qnc<*6;Uf5>wChhWp=8Vivgd3lQ+6O!et6aQz#Ka}hl z|B4o`YVn%HAO3Q=zS-yN9P{nD6YQApnQ)AedMbl;M&PHi>5l*N^8aZqGo8GWedcwE zE&oZ>UzcOLkKI@Pe@(=9v@zhb(2FtPPv`%y?S~`&abGwxKH0V&@ypv18~^MS|J<*Y zG1L*iW5SIu*fHUQaEuB6dex8kpFsI}{zZK9))ssH2)XgEO!4nMII7c+{5fno$6#yk z{DwG}kuSj?a(QWs7bkB2zchIp@r6CvZ%e`V|A_v9@T1Yuy%XMm{3!)Hcfy`OCsy7E zZ|3}-l6WTPcjk>Pp4HO#>saXD^$+FzyQtdxGuZsa`D}U@!jWG8f81u$Y9F9e?)56< zU;Rl*|1#37^Z&(d!+&|6i#~+Mb3@o&R5HE_)tJ8>ig1ljVbOmJhV#wudaA-IA|p$-6DN`Z;_% z{+zxkq^}=+%Sd-EY<+QwYnHYwo%K3f6BMW~OUIK679!BT?TN*|` z{$G%+{miB(_+2k!j6MCs=`}w8pO58vYh#UP=S5}C$HdD1|DN^upMVXWa?3O1FXSI% zOh$G`BQ5f&aL8{&e=!{QuZAB=@|V+5P9r}T)bO{$^Wak}-vQfSNB{Z>*!9UStnx3z z_GjlN{Y5Q(`@gL3x9`yZ3}c{o5SF2Pjr0H6?8g`PyWeF_0hi1}|4;3^eE45S`3=?j zkmpht!9&am=yGz!`$p#oY~`J!&nUXJzk~hWvJQ!S%deH||JwzeW7e&oOA2rhGN~zp31e35R1$ zd3((G$v-G_Holkim*Nk7?b&{^y>&_M^V2l!$Me%||60b-$MexgVB6=tq*?bRcr*J4 zD*HS&345VG{eg0C63=t~KNjydKHr394Ip?oVGr>`C4G;r`3 z<6rzg7{`q7p*_Wz@vl9CF_YMOQMG4y=eI_AKXeRZcxnG{P4<@(Psl%dLK&m^-X``E ze<#K7nA+{ce?^sRZx0;yE*i!9>BMaxqci@!q%Z8tlZijrG2)Zai}E|#|DUlvguda& zV>{R};nvUXY`QOEuS&1gd*SVHwm*aRknQ)w5&ma4m-x;kjP43}Is2^* z@jUtRvL64_E8mKIH~vp2`7Zn+SHA~aKabC2@wO?nS6;(5?r87V=OVkm+W*jp%l~tE z;DmBO_W3>7@V?85_Urq3^B3W7eSaA|o)>2; z;LnUL3a$xa-M_nF^DS>5{-__nmvqR_Jyq_Uc*EN#`Am6cJzjTcdTktPe)ShOy3^ly z8gb(*{mXE3!1l;j&{17?)ltMa3Z}XtvO>v>>V?DKSNzRby}15z$nF)_Q+AD?pGxd% z&g%|Qf6*B#ve8A8de_wy)Pyf92R~kRe_D38l^y+jvD5tj0R3)`t8TCFet>b9f^ ze2{UrlL_lK%y?$U1CPh==DO{U>+Rxqa~yWqa&LU@b(3qjINsl4$02Jo9Cx;q|J8YIQ#MFWuD$*%O}f+T5Ne_`ACcR zwAgp(r?C~+w7(N^J@K^jJy0Mm0ZU z@tx++sIitq^JAz-cx<#k9e*Cd`m49ToqKBI53L58pUH>4)mL5Hn ze(bL2nC^af79EdeqH|p68uqiQ{^-Y-ae9w#q-EUfbi!U)S$&_wLjO$kC()|TN%CI$ z_>imLO?(?K_y*AZLe;lCrs0qB_yF=*Fr9Yel$*Yg|A6?mz&Bu1=l7jG?6(Ry!}pVy z2>)65L5@GF=3jdU;IQX7)A{gOY&yqQTKASa~e>-l-aV&=l`j`WzHUgD1YxK{Z%1 z&Dz_GJllJ)#df(N-*g`32fwukCEQZj^YLl41KpU@C@*+&pPm<9%&(A?PpI`IgiB8|?pR48VTDgmmEpDXL3+(wDYvsO&@XFs?>+7Uz=`Y~EsxKdC@k5E77qF4Ki+KU< zwKw1X%2HpG38ORrOWEiA55bXt$DwY3FR1$F-}5`?e>WWY_k7r8 zKd$?8A|GTw;|UV=G`93N#}g)aIi6s{+n#gy4R8*>Iq7R}N8(o!ejm2UR^yc0*2BJR zT?JcTQ{h}+_Ek||pPIt?2H#TivlAYGf4%0%@|+7td0L-q;9Q?;;i%6`e!KLCeH_zm zJhz++&q?Jo#PLX<>0R04;lzfYg{>TZ3Y^1xeq~;q*z`}rUZmgiZ9N?4+u5%v?-8fh z6UO@A%ej}^$1eD8!sy;`J>hAK6Ki{NE$7{EKRP>CX{oB4j#HMqe%M$jlYpDFMl2!O`udIDJr++mZ_J0HW%EP|)IOPu6a~$wg z_^WKX=Kg_wZa>rEXg|*L=!d&(I>!Oq_ubbJ*C9UOEX2R8vhBz8=k{a%qW%0w#vv~v z{Q1@XXNi9U{!2%f1E%Ky-khF;a88fse@;(l)Zz53{G-zUnP;8(oyWc_;9RSF9s9c) zD!jAB+6%dBcke*HrRq0pc_YuD!#-&Wms@w7tJeZ1UnME%)* zyq@LyItWk1hEDrS*++cZZrc<8lL!Oz+!}N72k4T^8Gd7iv z_n)5cksj|qyKtPZj(eX0TLwDEJzIWvH$>gvS^7s#)5i6ExG=ZHjr0C>GroNPx)ILz zuMfg;J@xu-d~v^R{TW}bzu9oq-@~1+Ua!|A`|6vQTtD03s2}zBCjB3Mv7BG=Jlo3$ zBkS+SY`Q1Hd)Oaa>*MLL*W;ebuO#8y@c$8Iu5+#Oh13sAMm5g+w;jmi{!M$f?`&^7 z9QMXdENk06tpx8psq)0iXTfvvPp^Cdycj>#(6}36``dAe*CoCj`EHnQy>YLC_v51) z8mHXy3Hc=av)JN!-L;|3f2?3xXmf9HUKyw4Ij-~kvHisPubPtf9&Fu5IThkNyb1hWs>qC#=2j^O- z1H`|ceCbxbq_l@?IHof_UG_OWmRC;C3^=D}4xH1o6wc{c0!MmWt92NTwOV~|{}B94 zHl1w$7W~7H^WB!Z-C5IjA_eLDmhF|N!E@jhN&ZmJk@60C5Baa7KjRqgdsFyXaL8{# z?z)06s&aW|?_s|3OQpOnEy?Z3G`U0TFUT`mYIV~P+@xm4_Y4P$F_mP%te`*(pdkyxy-d=}}&b1spKX?pkoV>Heds@7|#SgaFzCYXRYw_e3 zTmGT{#k0$IW6yEjgK#f$>s;r!GwYkvwsDTD+l?>!d*$0(a?=~~4}YtC_u9iT-K+P~ z-(hP~wWr+r-_uC>?o9l)-$+m2GZ+U!`CVA$GvNK`hy2>_mbLv1qmwP~;G59j0>8G1 zJF?&Hg71Xa!*8f^c`EHK_|}WcxQd&qT%M0V^P(18zgfPb#j9I9+~N(1KXpP`8}K%c z>1;0x*+>03PTBM8dWAT7EBnklT5SDg`Mwr=K4*D{^3FUi@y_RzwLDjHOy_sgb?mQf zi2H!&mhyT=WxwlgL-+n9Uk8U=dn43W*n1t}tuNb%?n%fuvA>|9!j3Zve&v;A9BPC& zJsZ)@=~)0re5Pk-D?Q#@gj{=DQ+oEjpo~Mw=^4RZ#Bcicr1ZEp=nUi+v*}*;K$+J+ zw_wM$Keynw4e69y-XS;qAbfEnEpp2%(N zv;Aay_6H&VWz+MOTTA^etLZuIiZV{<4&u%MY}8KG(01zxdtd`gL!O{|2~={~L+tCsuza9Qv~L7i{=FNnhT8Jlpg7 zka-IE4R&o=+d^FLoabcw-bx;HuI;M)FR`I>E!M^C$2G*W%=Km7&6lgZdEbjH%G>a( z;2hrcNBEchPPtwmsNp~Ow`FbUr&Rl2hG*kHvGPH94*s}*aIM#|gk4(YuJ!8mviZ8p zG1q#ve_LMV&U@LPfGp&W z)9Zm_+@1C;&unjVvZub~75eRQe%)Ug-HZwR7j$%vlO#(=jFa?ydTSAP&O_hx=U9&E z)c1SVxdl6ptDFvp{^jWV?dQa#zqX~nF6leY?xEhpjrMK+Ctxe`?>NQX zaEw#b-i(x9d3B3Tf5fNU`poR}RA%cZ^L+9XJj}Snb;R#IiO%xtVn52yagIyi_ZD&I z`w!6nz`s`6wRi0g^V+)$l6}+X`{{^Jo=N#+UYqLkvdhX^s?oo{aUa)5;<}bNb+1Ie zmHi7E;$0T(`SdaJs5_CDfAisif}P7QMIZi12FH3 zkM(qV$p*g^ia(@vO!)!Fkh2b}-*$TZ>2NQ^&&2X`T<1*jOt#LMgsMJ2RgmrU#+>Vyq@p!|C z*03#@+S{zZ+K+!a+X3Qp99L$#@y6{Uoc+Js$T?-b)Gq$vwtAlwn@hLySXA$NRT zmJhY$Yg+R4u+7iyz}aQp)3CP-KAGdc!dBeqA@)6|G{2))=x>f3Q;v^b?vW=n(iacw zKkin4M>oI6-O7$5qxgMnqvP*zV_~@Gwjrf&Z+>rErvJ88jec`vVxvELRby>P3%9*P zDU7?11e)eIBhZbVTJG^@R{rkUyaVI->ndN3d>CHKrgI$2I`%$$>fVU_LH6%zsK_6} zA9BNMJBJ^DL$1B$i5(~6HSnSuzvEIS!P6=$mqY$xF2F=_=(?%?w4~o=!>>-^m7AZi z-)8ldM||5ZuI-Csx^Kbr*8)@{GCh6cj3496ob7O+KTu1%8sMi4Ciq)BXEqPc|H1GhwiHSjw9*CevBh= zoQ?f{jI)u~ws>D+?JY|7zCikHe|NCye&e!|*UuO1_+4o_{kXlQ5Wb@7yFQlb z`yQL_F67qduT{D0VOgK|6>;-F>b*~n@@=m-W&L!~&Gj=4j{5O`8qa}Wiw&LM1?RKR z?PmcT?MHj&FWXxRhdsluX@xhx5q|D(mTwTe{%D+R{RjUR?O`^DUjKw5w|&gT8|_26 z>B;g%NpAjE!q?UOeCyKEKdq{)-15uv`Ebb1pZ#slUoNlDzPFU;FgiNf^b8BQ(_c~A z-|;nnerNXli0|;0kI&_~e68;&AJ>OG2!Di4=eUdq*)MAdKk1)^Khke{r@%SA{czNW z`U{i3ygISBrOj`9#;8sg`5!)-ADrJwZ9i8vL|fy# zrsZRq>27ATeE2-pIPJ=okI8sKjgegj{>`2%GJBzmV<8q5mvCK+GV#-+FZI z@3?;VUdr*OHB#OSSB~;`JsaDbHgwN;O)$ktj?V9zOW4PEP19%nM)@s8z6ss09iukyD`DgNHqR=$SJKdU)I2o5*z_)jBfZww z#-79VXZ$k7+&JZH$V=3(ayjHaSFc2qX5Kh?dt&FudObG}I(ay;`n%x>|9#}X2l`1- z@3gu1Q$_AQj83`lAws{4T>jT0?)X0v2k~E1+3$q5zpE?z9dSAS^DC>r8QWQZ8=Un` zZ|JW*$oe*v=W}a#nSb+f1W%wmvwTWozr(C0>}@r?Z2y#bQ;SDhYT`k_*;sY(V zFOKlmM~Cu@`mleP!lB$hOoMa(FaYQNVGxe~;Q`u@{YU<8Z+S=iwS6m(_U(K(+jE}p zwhq1%9i8*t;h=X;}F=*IoeH&>STn6E41&H;E4`ui)7V!rxb z+Qa)2&*t0>_C5BFUdp4&{cf=nURGK8RIW)`J`K+Dl`Z+IB=>)|t;g1O(x8)fun&HD zADJaT&ugT7w}8jDc*VbyXJRt!86QP*M11y7A-6pa6L+*n-&<~m^LxvOl3e}yq$l+K zpJW^1#pvk#pI-}!FaGD(0zP8NU#t0l)uYOLv@J!wHa4$rIS zBg4ZH{`^7Sb5gc1E9N`D4cq@MtNc5>#GQtIoIlsRYfN*$6W5pbz>`T!l>a@I|G0*4 zzWX5ma!s%2!%F;dKKvv4Bk-NamjKRxo=p0qeHj0^6#pr{CnbH+{||hQ^b_>-8s8wi z5dW`Kw*B^DKh76iU2Rsk{5jwgNm?nLwKT&^1ieH{YeMfkCZQ`4@mjB80 zrfSdjZ&^kDWbc22--W){-%m8L^M2uU#?`?SD=UAH^kupK(<&+Dan_huCmK&2ML<+4VNNkx@ep#ZRH0x~oUA zNv)nZwb?yJ^P~1YW>wkw5@OWe+Y7ntf%sQ5^9Q8NV zAAJ4CG}5WIVczLpe9XpCw_fJY^Y|GxtNI>uhj_;o{#@$MbNspZn7c;ZaS6UZ@dw|> zJhpMv*BUHVPS}j8ulREUD&^-Hbg-uJDMqOq$JT%7XHegw1{dQSSloNCGjNMPuWo)8 z=;LSBEnNjXjt$(hy#6c@FTaIhuLI*0%<*UOWBtK**|^R>box*92j6Ud4(R)#KLa~ydyr&Jcb zp!c~UdhWjwMeyD;Lg$Q@PN(;nQAKuqRqh<}MC|>CbZGCU_>nb83HWn&F)UM(nHp-^6~P*!hOc z=a8@0ZzrNOs?**shKlI)T?!tH0D)v3YBKW!{ba~IP7%pc;wMRnv=sTz+!kI-ruae1|EX1tZez4-dz6PGe-G4e{4IP z=a21x^Zcqu zBR`%vbmngz`)q$b9QNOTPRalB8s7N!BG2*dgCo9EpIGh%-cZZW?*NmK$9Dkd=T3+7 z{M>#xe^;>m#CL`Bv1j<3s=a01(%$m?+!@I8{M;cp;&XoPQuynm`N26KVmW*7#dUtC zF~0at)eP;a)N#FV8JK;a8`pzdeJ@>Nde9yj}{oM^Q z-x&LQ@rQlegZG`eJ@ml@YupFOkNwy0Ro31V{9(`c^8N7ls{9geFl|{jcqJ z9vtn-@}CCh^0d68{GDI@ARP0n{Xbw6h(G@yF#F@MXM40ii1uiCu7Y!Uu7=50}e)=|e0XB8Uw;#^&9e^V~zmv^C@5V;E(|k9Z zo_HC;b#VU9W%<6g>fbzz`X9}FrON70!G6~7heKcaP)k18lKY-9ho1>Y_;ar;*SgQw z_&gsb;J>@F=bL$t^TGO93+M70hNHZcuWrfhFG6nl*&pQc^ZbqSQ+}`|e=x~!y1taJ z&lNi73-2Mn6UP+39bO8bUa*mODbJ8=pL?|<;`h68#P9zZ8j-P~Ge2A5oZoG5bf)&oA4j{Z+J2>tii^w~4`dJ+QsS_2Bwf6^{1e{O5znKUCD4 z{;-GiM1N@hhu~cP-mei&sded+H0@1e|Y>| zSK_k9}DZWn=pY6*&P0@aH!+t#3|10)K z;CFP@_P_MrdJc+VQYC(u%j2!6TYJ$jgr8NbTQ7u{vA?vi_U?!M)6g$~pOW+|lm0&R zt6{pON({e-y$F9XmA4+YjVkIl@EiK}o9gk$ZRPlK-O}`4SMuLv+n@T}{+7eh{?4Q~ zeW7>Vb}gLOZ3p03w_X2+dd=vvdHfOiK0M6tiIN_V|8Y2uf8S@Izi3bIqy0|9-(^#L zUx@Q=8*8lI7h->s-xsokyvO@O&ig3m1MKf8<#8#zn*H~b@V?KWkN!2@4{!*r%JWMEj%cW&2q;?9V?&dm;TXK3M{{;b&K&xt|mE zeD?Ihe*t%rpVt?jh1bBLcU|Ko{LP|&^ab_(!u!0qN91GZGxy*>v***y_D{i=5Z?as z6_k2k{68JN^Tb<<{=?{3vwyHQZ+!w@1K(TNdHG>@SK+rjwSEu#w+g#X zx{Cd43R{ly9i&Ndop?LH9s`Q{QMS-~eJRhdDcai~Wqa!<+sk49J5-M6^OqYK>1OMN zKVMP3?!J}%ysoCrZAI_?a<;H{-M$~btLR-eyD_ZooyBA z$?I%hU*2By)q5r3SZDJ)6f4jzDtf;|VR;7o9SYO;Na|G4ew5$PYd@Y?d-EUmez)QP z{A4!8?^Mk2yR@R3XB&JXvGdZFZ}g`>PebK87tvCE3IKsXes|*t9OWz9zBAi?gH6BZ!$_~+<;dyvyBw=i{C<~XJsjWV(0(H6 zwVz6?{k~-HcRV~FM|%BkM=PaAw*QOt$?FoE-WfR3>vt*E;(w68DNX)kze~a2-&yqD z7qAy(UUVG0MVJ|yA&%BfR5B7Ou&lj8Usdyj4W`18? z5$^|sC)j^Q;b!_=r*~a0_a~0`a(}W7{(K21+?bQOdW;|MWs;#?f2N$~5-&N1esrHw#}J4Jg|0>^QCW`7_6`$4dO_J33G6U-BKkdqB8VRcOi)KEl3zfN~MMpRjSh zxeK=Z7`9h(!tn6*g)e2|Yb~r#MILJKXoEK=e$=;kK9~GIst(?IDLe?DU-+l*tDi5M zExhsGd94xlw<4*iQL6Ggq_uh9EE zkNo0V-TKov(Z9nnzBBwrIKpf1@fY^X@2#)b9#+DezUA0Q`qYoXS?~30)*r@S=(QhA zYQN9Vx!tBJonv7py<$`oo_*pY&xO zY_RFi`c?Sh z_rf0kW#PB=CpVv6%`d$g8%4d>qglTL&U%mE&};AWkHMbL$MBQK11sR$bdGeh^|O6m zKpYRO=l8CPxX#) z9iQm%DEh>I^1nZ5omKtEw|f6S{}&-5Oq0XY>Ay#+NMQ92Guv=}huIR?eqMPUCznxv z|F+bbYew7Qk-`J;HaO;>T*Fa+f6@D`?^&2zqAK|+xw%^0$)x>~9sE z{r$N7SRVTu;SV&zo1dJX6L6%*`s*ZpQGf4#1oaEQ02}2a@C1BSjV;@Q*b`hDv-sTQUW2B8+nX)%Bd) zjq_`E4Qe6nWqpm+JD2@-Y523eOn)wK>mrx8<)6!|4_?fs_^ioRe&1ZuGYw+8LUq%0xU-mx@hkxyDA7L+V z#9!tWq&3+1E$>Ki1G6Z|E)m9Nuf(2>;Lv>N)5?SwNzYHoTeNErnfE8HLXy zjf(Gl-oo$G3wu1;p4x0m#|F~F@5>fdSaA>O33d&`_Oz!~S8E=76JJdD!|=0<|06U8 z(;s65!!Ly+y!RX|#n+w1-(@gO;M5wE{oe{W`oBAe>iYV2@$WqeW9aie2_tYEFRp3r zg`@qvk?_aiZ&41)2z-R!-r~>lIRQuc+=9M^?y*I0d3LZL!soV?RF&w>qWI@WXI~m+g;& zjo+%y@tgi=@2;6`fb*J}{ZU>sv-~4Hmd61&m&ai^%0vCphTis;_47$z4sUsfzIqlK zUO*V-M-}UY<=ecspn%pT=YImeU$xWg?;v1erMr-hFkOa7GYuU(K*R}+@s?=k8S_oDE{q_ zd)VJp*z&Ud=kn-EoXd9~9Oe5S{GEVrpI7^L&Di#|n7=6=58eDOsEBL9H`VOl6wjxd`Hl0b`L%uI z_AmiQer@k3;b`xs@5j}@{cD@FMEz*r59j>YA4Pt&A4&Fqb1v6}wna!%d!KBabF&FNd-gC#!a^!LHhKN$WHJSRMPjqvJiPg(Ey zBI_sNtoPZB(3_qa$|l%3ecwT}sg&pJ59>7@4%2DgZ!yhj8Q*Wu?@p#355dKX5){wr4uZlnMnfxC_AO44L;hkQ@^`a8rmD`wO zX8(nS`{6$7&%<7M4ZIDG!y62S5SPm z*Zy{>#%fM?GaTu44!93p%mGi)v6vo~PAhRu{|I}#YOKP~r11Ccn^*PEucJzU$9USbOucSw|Jc7;t7##UGzTI&4=e1<8_B4?>_D7T6@crZ^?B$&e_8yARyGCUCV~xl) zr5Sh?o8p>|=Z`YHtJZW@z~|TMYRzdHHqVM{ObbcR12M7j{i)1-v2gng*|H z@P@=o|BB-izNUn?{cmPJwbslfw@*czc7_)1JcqSDsM6!?C+gPV4QDpI4P-l=o-P<{gCSZY}zsg>5g< ze`)X7H`{xC5%z{ZlKAZT^&GQhrtIftiy89s@fxf655nP3{bur>^%HRD{g&<`!oH}4 z_glJc_Aez)#qvDLZ{*i+2b;c^R$AUkfc*rV?T3@S;SZya@bc$P`ui*NwwKqRUmN@g-VQ&NGE=%fRXho~dvi*O&||Bf-U@@;qs zK2-RJ@J2Ys8yD~AorCZ@YxC9>@H8C#DR2;RS!iJ4fN$@u3*rHLM)Ihy3U8OJKt4dqV3cmH4*uqEYlLl_Z4&cwHj}W4?RRmyiEVM~ z99_3Z*e2TivbC4C+F-SH@Y^%eZg-5ZDmUFCISVI8I&tvpY(Gw>QR3KSLv-m+8hpI2PR@24xmFUMIw`=(2I8TR%CQ#LR}{V2bo z--v!2JfjkMC+xU(cST&Y;eUety9*oMzCFUrOV2vZ_Yl4pzK=~&Kb-Xcg5JLIdqr=4 zhOo=|*#hVMjKC3}`Pm8Q{FvU*o1cAfx9ID@D{AF|)`kVDtX@igv!k^_e z1?TeG3rBgGKF4}(d@8PWbi=XMQGI_G-cYNnxm?@crG;IqwZ1Mc?0X!|Z+?&CdN}4f zUAx%W;L*fBPr41hnoZGuH@^|yxxH!lIyS|%;2C}wS9Ip*EfeL)${oW@x7lL_|4DfTAvZ#_juah#CtvcUhY=~meIXN$dS2QzTAKi^AfdTwP?UV{I_{C=k*)@-QHBkV_gYJUt4 z`+M+j|97z1zvH5MtZNz-pTDzw@2`ll{9v744#P^E%N=Jg=6?Nt>k|Cs@3)%wIq%7Y z_muFyM{`F*zcuOgH$quwe`9d=Hwb5cCmVYEgV5jhYxP>(O{M)*#~W#h0jaeeOluVdynUl#M+NF>^(lR{f@-09ZbTx|FL|c|M^|&)Akw9 z-^<4m+kaR#xj(SKT#SwK|8B0wH@{HI!}dWFn^WJt@E$hh2iTk5D=T7+^}Y3Yhr_ax zkCHjtXIJ41-dEc{qp;7znSY|Gq^IbgS=i^_h@u*^`6t(2rubc4(&zKoC*W_WL~I|O z=psGu!rxy0epRKd(|45poIcAZ()WAVd#(ALVm|}hHa=Lct@JzBNa2;t-y!VDK^VC|T`XM;%vwYXet%*0w&kqf zoAmn6-v>0j&m%;5_4_$CBYyR3;H;lHXRf{v4t+Q2XtTw5`@+lW=fRkwuH>okI`(4j z@fi=V$KSt39+mf^pW^o|#r}=(emI`5?Sp&J#rRu)wx8@zefax}r`FFmKE3$=9=rnm z6@_0-_#yaL*%ad&Ww1Fn^YA1b<6HA*{}lO??T>nVN=fZe68pAY%hoX z%h*4}=9H__&89BuOI9DO{VHt3p5b`?JZlQSh5fMie(%-fH{b8=y_Ipl_jymLkMDfH z`APJ*;a4&Kh4WAQ`{MU$3-Dc)mh%YVx5GKS_v7dA!;SDh>m1>I9`^{`RqFrW|Bn2_ zw-^6Dk1NOXxYk!cy12i6KlM2Xzqqs~>u(kH9rdSvx}o0*hu-%>xA=bre^Y#K^`L($ z>5)$+)_xoQ!(JXHT(I%UIX&K+81Yr_k%lR@N_;-l@$I4-TRtDUKC#b->Mx!r_4!cG zcTX<%hL>}A`>SBXFNY(%&x@WUtjDY(A4Z>f0r?BoKesQO@@M*XC$64fg~Pwk=UToI zzt86mB=-5-4RC%w*ZwB*V|Y1-Zznc9P2yB|pWn4ViS)_`&}TlLSpNel{ibhoV*Ojc zkv^a2-jdSq^W65=kv^a29)R=n+(+TakKyI8H+)CRkKq@>5#Hy&4`Cnamv^M}%hQSV z@Ax(Q?@wH%4-Ws{tE2BYAIs(^*n5HO?_<2KexmcVUOve2nC(x%*`D=*Q+n;aw;=2* zW{ehWe0-{CPI%wz9D41yaBOEjn*4b$k;l&xHpSuap+<*--3p7e%(n(znENBjJn>#0vT&i6k5b_jk{ z@h1-wK3IG6o9)|h=(V?h55E5|=e2tA^^W4tdn2aU-;j8B;^(8cetxy+<<*G|e=OO{ zM-yw`m+a;K2Cr%G+Qe_gek(j&;+J*=l-#6`M`xS85U+|3j`@S(B z;k|Yvq%+@(HwwR}Q z$=>f@ZiXL`?7beny!g|Xvw8r|wb+4?SzigNE`<`;ohtYrdp7WLPdy3xooDajlQTVqAe-ys9@Fj+)Jl{}w zl=#NU-=!(M_r^r{9>Q;j-&yo8W}?RHLC=v&6`uU%@LS;s?^?Wl*GnsF-rKgbzBl6! z7T14pZQkQK*5UuBzzC&pYz!i z!*A#K$l(V$|781V^x9vH{d22O@lW*Hb@xrQ?(~M19ZxO%rT%7fb%#s!mwBiD(tBZL z(^1@wf>9gJsoOO?YZ4sc2FuK6jsF5JFh)mVsGzZXH^8R;@Uf}#ecQK z^9+iNu|DnC!0pK89W9^h%)Gp;KQBb@diDTp3Ap)O=$XC?nDyfg`-uj-ex2>78|->^ z*3UNhSc9KLx_a17$w;~5ff8UA$v=0t+U35#^UcNu{_O&8ix_c_f9AKQ-Q%Bmty%Z@ zhZjV-ckF@@vhCi93wpW1-My0+wD}~9d;fj`?jCUOfeT!zXODky=N|vy&OQFYoqNX~ zct4)pJMqAa@#Nmg2cCi__x}BX$KlDnB(AMk7yadL@?e998a&eAu?CMfcx!_v61x^P z2`{seBF4Xs{YZ~%83*BAMX&uK_QU>DRFLO}QTh`_{mG>FIgpk7{mx3uoC>_0{aXtA zoQL^)bn)MOuFI1xk~%lA$%x4$FrT^O%>6bEUxo+B|eo+kEQYOYa37u zKS9_z;o%5BdUgE`kuAil82<#n5x;Ex1-o{60A`3(iSKSTe|?3UbH}djc>Kk=!L{XO zu>Gjwb1hzry|{#TEoUo#zpm19j)NT=FD|V9ctdY~L+^7f-K6LC;?K2}CGcW4#rQlI zMts%WJIoNa5}&)-$NqR>^$YQp^&8;OyOz{V$h>yA2;Nh>tDcit!G1g^WBajga{E~a z=k~K3&h2LxoZHVyINFcxX%%S=c5TzT&TE^aa8BPayaykO=|9oXA4z)G5@@2Q))IWq zavF~Cvght#*D5**-{xo0`=U#YprN<})kNIA&wvcgX& zytJ_QlB^?rAFHexC*ywrj{c!j*yzMiGUjw&`Kl_)>aP%+LyTRd4uqpcMpgzK%_O{2cH-A%b z&Y$gbaUHzny#=e-k9!NecVIId_YTPBH~2m(s++hUQ_d&HVXx&MU)c7!0sY5{f7|CM z9PRUS*qfh66#I3s$K&%zlk%Vs)cN6(sS@>_=t95osf-8UH`nTx?cg?HCj>8-!e>+cwx{n>u*P4b07i3J*W&tkx|1v37JdH*DKK z-ca<;mG6V|T)Fjed(k^rF2`KC@o69N>2E9P4S%xlW(;<2-MQ+RTQ|I%!*|jbhrN35 zYYP2U*VfM^evUX5^|r^*8{ZhZ9G-FX9KFXw=v`arBL2L#(1vfTyw~U274YW^yEf4e z-&NT3jKGl|wgS=*z0e3LvQ}GzMHsn_#Qa)AKy-Wk)KbM`f)B% zrW>l{zT4}y;9o3!KiuN~510IS{Ma7z@z4w3TkL&~jV5~P_<7;^b$!nx4~qUL(1(Ay zY95v>RN`9ra?Ah>tAuy1xto241(lqM#~b?0&iyT*e)8O3CmeHsXVR}tdY_}UzgbYq-{)}K?8kGo zrgtqI>6Pt|f(`HaBZu#yeMk5u_to#-T}B+rweV_wA6ikX^*(rveX6mN8F(9eoE9sY zJj_}bdp8#T9y}l02NnAZ;pOblD}3ws>$T`VF7~qd3%(Zid}13?jNkU1cU{Yw#2NOJA(#w*8wT+rBcJ{>)tsUX=KM^|@EW?A@(q-Q1mtpmm9 zU^l=+HCA)0JK@NW`CE%F_?#&4Y3`e2m}Mmkvk!Svc&E z7yX<0oASmR_@*SkH&;~Gzx5gAIly3iHSynH^uFtK8T{hHAA&u<$NAiM@iARpjmb5U z4^uyzDYIbbwlnYG{2F|IZPw~7&mDEZN0_Hw)8N4dI}f}N+pwQ#@Kl52Ij69X=bVD$ zIj7(?%+qH7gAE>Qu=B*({vg~#`rIaX?nzlmBiX?3XDX_nYZ_z!(}lfHa3lNrQG5<) zg5N31Q&GQz-_ZLU&@?>5rubabL4M;}<3sRq_`*^bx5G!^zc1`_Lx#PZP0_!7Tlkml zYl405X$#CL<&^!tr1!ZePOqn)d-~K*)z`V-r+p~9;Zc5nQ248`=RzJmtVH|0?1jB- zUzpi;lXyXYqUm)6;--XYz?eq6hI zEp7c@SM<&u4zjobS%hjaMVaKwKJ0Z&jqd-$93 zrAt_ICVv+fz7qXmxU;Zx(Y8m=Z;I@>G}z}b%-_{2k)Ond--=y?Z?fvJVL*?(qX^;7I;{jQ|9emoX(`Fifj<+~EjKHh z7xo==3)#=lfwbZCir(~%v!Byvf0fgB0M6;#2j}!Tu8;KjT*MO6aaolb?8fT+cNO+K zLCe^WbEVHo*#4F#`#$!=-tPpdi)$oq-}0by>PTQ-?6qD{u-O2z4^`dYv8c= zxs5IGm28Ufo!~d(^Bru~=Qk>Cy>?`MZz^niwg1ZPbqO5pRsGtAejw@l@IM0oRq^+R zH__JEf9=`zAMB4T|Ep`P`lp`6&c&MFcF~*ujqFGIW!qb@YYh}}YfrKNXW~D}{<{iW zUM=i%dCiBTybQ0u2>+06+y*?4F<%Mqa}b`-qP)z%*QU9Ax?tQ?vi|mZ{J|7zC2#+o zy1s1_it#PuH{yFc_O^#zDiPOKXhNsPf-fb!`T4v`x=J;G|$nmX#qddNRQC(L%@S!-Dyp-RMRm8DJL9Adu=8`}0 z^Y!r?_tDsYtwkUGgY`89=la?WM|~OIyyfr*Qh4iUIcdrD(+@AkM)4i6o8dQ?^nTzS zod4MWZehclmmGd49O3`}FrHjOdOgMz$CKta*zx2xIFBdouS4&6@+AB{HpTC7jq-bA zMO=fz+CHtY_sJY(|D}bsA0TYDUk!)7-|ez|qy6~Zt}Sr%r=R`D`uI9j{CyoZJ+>jm z?|M0R^z<5Ae%EUPetqFt_(;<0Z-Blh`}0~W{COYH2>i_w{-Zz3a~nrrydpu>fJp~)SoAg9@+4N=hS}wD5(V4w(D)=9b@0`>5H-6{tb9@UQeA@nzKUcr! zy`9sa()YxFspmp}J@nic+spo|3;W%je)eBcxOzt*`*D6Ty!p@Jhv6LF_Lsw3{}JB% zbXSo6Y`+T5_UquVKYX0P9XxlQ($fcL{Ve&`ZrILEiL5nlTU?>oGf6JNBC>mPh})gB*38&l?;!`ue#(fJZZ{S^D5-*XMmO~JoM z9Vpdz>S(Xt+pG9p5!WK(yCSmdQo(-bWCtAIInn=iIQy5g|1R2h`1iXuljyA@MgLRO zXZU{{>EFRQ;~gdapC93!VH)B?aW2dA|FbIM-UIZGKQAkMauea-V&;- zHB$as`6o4a;7|P9Tm9Rv{;ea#lCabue{(ylcFDgZ)dOby4F~Mk(ef|*ZT{MO)+q;P zopb8m0uj zruVwS)w@;T#f5*4@JHZ$4L=LV7~T8JdPrYf!#ieQ1z%did;g5r3>T;P+8pCKzHT_; zGkh@un4ZTV+|ll8?alX>YEvrM?4Q)?449|3n|{ zoBaG?3BM2C$NsBQ_(kY*_!b=DzmL6RR*wm#1w=&wC*u0O~0;ZOa_hTh{L>&F}VEe-uz@|)8;07rP= ztG<+U#QmbPAFSuh{ump@<8g@JI39g>hV41tond)(!+R=g=Kj$SB)z;KvGH|~zKHL| zusfw@rV{N}q7VC@Dg2R=UeC9y(TD$g zr)&G)FM9j$Y4o{&pMk@_=gV0*pD!0sUZFQVwvWt9693=)|KIu_p{3^bPZONe{_XFg zecGN|w6|!_j)!{T-2M+^675ZU&rM-(`qB~dtS#V@)3T6wI4v2 z?Z@HJYkwf|TZzA)^t`J0GrZ^X9DW;|!_Owx-t*ZUdw7n0V)Nfi_$ZI_xc_VH9DFd& zQ{t^5b_&l5PDHh|l@jELZ+MB*??|3!rjqe~f5ufiC zSwwo{-6HxozOdK-2yxHyH%0%4;Ou`I4txELCV$?qJ4yU;ziyMg|8)YN`To~__%=4h z^muHAfA4=Cf$w5dygzY*-+aHL)0uU)FvL9QJaX{Ab>k*!ywU%|Cs=t=}r|!}lO@C{1=w@GyV>*-FdU z75m-nUs>2W%G*c8JbWj}{Mne6ZUCyocw|IFIF^{4O& zxC}2_rqBH3^sRs+KHqIH!14W*68;aL$o-5Rr~SS0rQENDZga`MbM6dbPtCbM@#p4M z*RxSS&S`VVoto2DZ~qqh$Nemc=l`dd_+*dIV8gfJZx+2g+~9+W&t1*4`Q-QcDShMc zOH=qU@*Dn)ZvmX+^Ia05H@&_~D$+ZBYn^_YRVDYrw#QEv|G)OH^p(W7rtoV=d3GG0 zF7`G)>*u{Gyzgd+@Y=5>zOaAQTe%;N`uI9|R1STP`|07=*4X+f9u!mm;o|T3?z396 z@Y{-BK9bn)vS@#}=>6`CbM#jfz5;(k@S_SpjPzRnZ!P{G_*DJv^UI39$>-5|%*Xaw zMK9B({8+X=vVCUX?I6Fr`1AhrWt2~(PyGQn>kl^cYsqic+rD%74dgfLEw9jfKl&K@ zp2!dV-!ArZ|F;spt?0Y?weEBOHwfQV^!B&2>_`7?|9$|D{?Go^>(B77y~j(~`)%Ad z9Pdhc4EBrQ9KUNgIsO?q$G;=_Gk*Jrh~M}oDeoMg{X@j(cTE(zJ)&G^SdF_*jdkt_lu768|~#GkE-8w@PbNP>&;)*AHaXs z55l2;H}`|i!qFf5-H^lZ=Hk!$ML8v$I$peAw3GDRTJ*->L4AgQ^(&Iz`(0<@INnYF zW;o*4{&=!i?|38h=HL9}{9E6lH@(N5cwG`xS8bGk%|6$nmd)BYySE=|8gGZ#IPf?MFEN$ltZ3QStt8&+orb5%U_? z)bC2lYe<$?tRemV3;8xU>5cwGZt;Jx{;luuKkr5Ln$F7NU;9q_r)=Mq?Ee_O_uKwO zu{V5+{ws&i-+ni~Md%|w!@DLF;f-&o<8*vap}$>-?_VkYcYL?j`8VZE`VIU&^pANR z?=FEIGbk@8{NWl~>L=I_{V;4_Vjrr=-Q+je@Duz!!ta15(chvm(qzN$;O`M$o=L3z zVd^J`&+K@|_+P4X#PNt@p%^c``(vz8!QXCI>eie7;Owe@d24sge>+hhukWT!DgO=q zO5(ez@Veiv+s98A*8d6g;s49ndwhPqEL#05?D*oBYj>?HZ|7Yq^ErQ%2J*R1*8FJ? zUN4eP-7TgWrt(D%!)&b``wkcUxAm%dof-zuO{5{rVl533T_=>T1n% z3F(dF@sZQ@8r^e~9v7$d5B_gwSI@t#C5&=0+(mw#UwAp;XV{PH%YL|r^qL06HB-;` zm(|#EO>i6@F1!x=KJpXid)Gi0b)dvXagERKyhQ!W>$%1c-hYt#B!54^hhlr!0_XOy zACCG`zn}Dme#^IMpD>pll~}*)$bZrXZO`5u>?-XjgR7<~5naLJSG+QSik+XCWR(oTLat%$s% z@1Wi5+*p#Bx-?w6pyuEwgKmcok*Ph!6den(;FW~Sgc_6+Z}ZVvCU7va^f zZs^w}{qynfwbH7Ro?n8GupieZzT4v{ys5W zZ(rf}Vm}SXSk!l=%))W5GJSOCr_z5r_N(D|&#!ZDj-{e+oFwda_!(?Ule_ub47#{h z_y@xGo1bDo?+L8M!FjH36ppz%=hp1YVyymU^0S1z?kM5qfyBOte=}juD|**yqNn$`}0@ar+Im|l`_D!uk6eK6HjiTz1Gd(ofx?veHIhFaay{{Z{pzw4=Wdm1S5O`=~w zT=}~xOW>#<>uY&~S0y&R>)@PT%PY#)^o+uh9{ZnZ__~t5=32kcZ!AH_DXJ3h&+cRI z>Kd!^7-c{Ft9P6jdimggf-T>vtB72@5iy>CzHS5BK#nEdjWAOmd6O3 z%fq}cuGOunU#!aHL0lt5JXR;n2VP5%uvB`TGd$@pKizl%G`p>Q6NMpG^K;n-=?}i+K^LIn+Uvm5JApYPlF^Ke9E89Ew5_;qFd>w50d;H|` zUjaw?|1SCIgY$Poh7wyI)=#bvkDn-S`=d7LbyJLQCBHeoRdB@T^9t+W1#F7JM{d}y(>VEoJ_$`Icq<8!gdY>;k46{U0iTU$dCFk!L9Qm_<@1gBS z|E~S&2hX)%1Bbo$_nY5{Pu`r^=U2ADU#nuT%WH!DTwdGZC@-H+nSsB-ra1oD&F`j) z>h`>!{ha;-aHPlhPQp397J1L{dHh6t-uuxDUs?Pc{{cA1@3lpa-}2A#ABH3TwSP@} z#Mj69o6^nJW{dN^ygKnD`<7RX7kW52j=~-z$`9F&vBmlRqu;BazxuvPM4l$D;1?3` zApCc=y47UAi{N-PzGGncZS;Q;UUvKtZ1|mU_Lo_I-Sg-A>xU!!y-#LsgS6zer*-f( zC4Iiz!uoZbrQEk5R9M{aXZX-W2H|yN%1WYwoVtIPJM)|t-WqsXM^!o26zuCVW z{`bF;cR!H7zbf`0BK~RkZRA05ZuT%7bF;Gf%{$V~o|W`_lU{qvBf=Zs5cv)NviT2w6YN}O*c*NXedf)H zjenf_j`-zX+C#A6XOq2r0KNVmR?@uwvIK5w^*v@;y|zi+wlWAFJF$ z|FDSP_f_=Vg6fTG(;>4)&wpcOKvysjRE?=SYA_goXc zuCU|gPIz%)uiIDBF7kEzD4eg`?H4u`d*>0>(eB?-*z4d8@WR5zHwowXeC{CPQ@@gS zmG!IO&^s=%Jln;e<5cq-d@Q}(Z<43ys{{1m@2u65k% zd8tPw;<(d(>IF5n9CvoZR~2q!zk+@H3dQ*Bw<5kZ=(oc$C+Is2N9Ub>hoSc{Pr!H9 z!K>$>x4^d)uI^EW=iH+V&$&mLxZ^rdf2Jq5hadmVzKy;}d{aISyLb8+$aJ<`7$ zW+|kSza)OgB{vdI@g0Llv@iPm(R&^B4ceIEbw5>b>bn0YjcdKm_V|eWdrz%xEAFXv z+&Kp4apzt*r+*(D>HiV-YvFUtiOT$TlGbeBW%{vEOrO`qxx5_LM|qh(uZtsn`r8I) ze-m){)86t6d+!0BBAy=prZ~>t&Tp>2X*k!P`OE1se~}*3d#ItGP5P$~(Z3PbQs%c5 z?N`IupY1FBX>a=od*>B4!+Bm|2Yeoz;&sxl2G1n6d~7ede2&0TKI-Sso2$3}Lw~00 z#6`rrr;5FP4*^v?=RE}L;a=iUoTnj6r{-yV55WkW-$P(~iu`yyTOav&-U`R@=ka?O z&d0CipO4>e;*R?E9%YZm-uW0I-otMB9T(O1{k3`ZPVEu)gY8djf6*Sj4mtwo>!71>ZcoSH zXixTk9&cCjH^u9qjif)um0pLrKD4P;*YhZFZZF&5#YMl2_;dI3KdP)*Q#^y- zdBf0OvWxVe!@ULgRGRF(t>4qg^R|8qIOc87q~DzMre`%A@ySz(xABK<{B3WE|4!Kc zz%o_Td;SQ$?D-<|z6N`Ihkifyme)PS-`~RP*#B%{^^S8xuYb?&5ng-m0SkNe%c=X& z%S#!r1e<=x$&r5j&%)tfUPgHb8-5Iq@bZEOpJwk#^866v3GZ2&pj_g*=loIb$%6Nl z_UJuE)7ba$H)Z}G^1Tpre_dniD?HqC2z^&!{hw&~KM99_-^n>f`Os}tV*Hj@=ndb~ z2=90?^rpvqf+9WNf*lWDk>a0*bN*)F2ygt}Gn?ZdfJ1*91vg0izrv?}KCgJ~-CzgJbwJge zIA;mJS69UKJ`HXy`*dfO`0VF8_O7q7`tJQW`}U2B>oeQ=T~_SE8_B zQP}*A!;wG3Z%g6TchH8jz8?<%*S?T4!PonXKc5xVU);;$v#f*gn~UE3?1m#hZzcR9 z{M}Dqp?t9L4-0R=Uj088{y2Ki;~%WeTZZ2bM|kb`!(p$#PRcU;X+K8Tu-BjCuJC7g z^*Ov8;qNN;n^OF1=@-JE;pH4&j_{9tF>@;9KkDNg#>Fl8Q*4UsKlw z`u+i?(JOfiyp_ELHMT5&kH08?^&YRGH@>5Aj_(*8@p*h%Us0dNXMN@PJodw%?cZ_n z4fs%uU(WIOCi~ySpVtvXY|3Y0$I%o^C4UKfysazjJV_t>aeu@+@xK<1^)j!6&0oF_ z+y?JqQ@qYHzPQeE-eCvFPh5xI`3v$5sLyZI!CO9iyOZ!=DD1Z&2GQR{ zohr76?WsM;M-qG8yd2v*Yx8PeZ3!IL&Dw8;qdkodGd3oDQNCVRFNO1Ubq}1cs|Qnl zoX1#1T$h&koQG|*AJ?7E`&jmQ-p6tBb;aI!SNr3|gWS`8R)-XU^YxIOlH?j{4G`q0$`t7M$%F3Z1gIJe(KK+5`vA;t6C z8h+#a_DS?R;g6_9s{0-I4ZS>-_^a4EF59IsqQ0B@3HxsJ9$!>LCGsKmg3pDmkI|$z zf1&?p^!AU@A1>O&7!5tmrjpL3^>yG%d??=EGfw=^Ogx;}^m!c?>63RQ)}Q0p@aOSl ze&cwuJuJZYf-1K9c=P-idgnEU;W(b0*VqV?>`H7uqi}A2WANgl_x*qq@Oy(fKDWZ* z&+t1M;XPg>d^HaU=lY+8qyEiL&%D#=)!#6j)8jQv`1@z_GY-G2q(}eGlV|^)Kf=HI z9&EFI860}o5jMkZHszCFs^^8`{eSXa(wliT{xeS}uD-2BeZ82!DduM>?IY)B8N9ew zSNe4g{raT$TVA%W7`HpG?fP}hYa5@}JrSSd$s_tB&B}MLVx9)QeY#S8>kH0$=Lkao zmZzT8>c#F`8Y4dI>G3!#Ab#h)6JHx@tn@?d&ldJsA9dFhzL8(~juO9oJhAK4@(i0| zd>((Z71i>xw)c3=_Hx))`Ga$O^Bdu(8sX&#?{`4^NYj>* z-d|i@KTB{`VZR${`+jF(zdO1genVluW4e?2T2k2Wm`=m*DeQMlohSK?!hQ$U_4RgP z!}p#uH~eZihu;iGc;ok3i-_O!dOSpW)pt>*q4&F~Yv5~3eA>6^L&9GD+RoE@!#hur z!*`xn>mOcxw75>BTdT&LU3@>Y#d>KM?B@DsSA*Md8=YGV``(7Wzrh0y9)#@!+!hkv z>(Dm)Zu)CA)?dA@%;{-2cxk#$H2te!^W)Y54>jz)uFU?n!fkBbjNj|ZtltakEA&xr z`zbe%MK`}ox|Domo`!R{V{rCsKK1F}Eax-jE8F)r($(MKfd&sYc%;GO4W4N5WP_(0 zyuZP-4L;Ujve_DbMcvNUoh8-dfi}N!zMUxgR}}qj*kiH3u>Iw7IQq*y6!?1hNkzXN zwhkUy*!Q)rh2#BoUT^io3)z%gxzOFf?|JkAiuQy2hQ0U0jlz3WBJyT_gAWrKr{vb7 zO8Eay_zCzsCH^a4S&x@`3p<{kL?7qtH=^GUzrE-kpB#W=eB%3JXW=&&{o}c?FzqYY z6xrh_So=OW?4OC={5`j%Z#8WCKgOn92isTfrcYAjDSm@T;KT4=l4JB@@3fZ?t#Sp_zHGB;CSNF+Pr#RVieBf0ml>VqIcb90=5p7si%`x z_;-mz@%--i@VbgiaRjc@BR>Y#&P+ z<+(JKr{BsQg6}W>?|xo=eevNsc};mGH9T*I>uT z9{)3S@GAT&IQ;qk>y7aDN_g$JC419z6uvu!w>%@fdaqxyehVD>d&uZg!mcUdHNKQ{*DnIQ{Cbs+zHp*|hQGQ1o<#()6eqE_Ns&6mC zx%~RzTz(spz3H8VBfY+RXDfB{`cgl>+h>yY^t8h2M`#`YZm*-S z*7f9Am($+xVQ=~ucAieZyd&{A={?5v7=J#MpZT1_BfR>q3r_33pS>H)d_VhgIIi!j z`^{nd4MqEn{APQP=dkzPOX}m@OO=20*}mPdxBvTb_Un`Vl{fLOCj7^{%k;mq;s02& zH$Arh@Na$`&qRJ?$CJTNr+nnk(gqaYmE!qgZ(_%q!5{zg`u_e!wRvm$i_A~3AN&vS zO8Akbe|qv)YkP*(m7Mn_zD>%WeW)UD<~Q^920K2^`pt=re>!Naw>O3!{c$LIMX*!bp?w;W#w9PznM-w$({Rf+8UR`934QQyZB z;qUtLS*>0ygMR?`!7t|+RG#|H+CIu}0S}ty@{)I87wMDtHrVo+qfc!7M^gD2|4}&N zf6QHcE0X+2`F)l8a(ofx_e#>Yko-mY8NcfhIX=rH*!Ubj=KOAiBRM=0Me_ECP1ru|Ow_ehU?xWNlj`Ka$o^)IhXZ2Y#b z9RCm;@ppfP`9ty(<@XKB$N8}+zj@#0TcogQR`fp$=k%HXVB>Rqo#R^tM|_r-$+q2Et@Ez&0LA*M1iq_LZK#F7%JTkav9% z_oqpQV)?DdUoKDkui&pdlXp2`dt+tYYRT&wZ2y(@!wq&kll7KY=7|Q||A*dhxlR)I zYfAj`OoIpYLc z-#=Svd7c7pv;XnJzXNZee#Q#FcLC)KFDU#c=SN6Y_cN@WfUhK{PIymyLz4Jq1|J~oOIpX{H>*{wc&Ln@G*k^yP2ZcZ5YwFh| zz5X13WPht0{0C-^_FE{>}Gw(%eK zSE64Jx08M-`M(^0Ybmcc75$B{*HZ(@ekten@aOkX*23{Ulvh14ueF}`^ZauD_j@bH zX+QaUD{{UMZ^c9FKHn%N_@2rjiQie``_4(8?}INa_3`)j&TB0@|MYV|md^;|<4BJ@ z)8GyYH}tpiuC#7^ucZtXd7R(OCldebJnpl@HtY8${>~40*CFMV^;;Wkdkg&@;!Umi6o5cCsI9=qDOH z)!_XlynT@CP_9#D{~fI3WNtTjd4u~KyuQIB4R-x4`*+I`Ybwq&w)wqO zK)kP~hu@fI^t}*k;rw2Rb?|vqXC?GGwt;y5(&ss6;o5usJm&$p&8GN#rstjeYpmWo z(LtNa{vBs$|Eu8ae+``d_rdMrPyZdXoAB>Esr|%tMejW626%Dd!`P3&43{d|On#l` z{x#-5m5;)caJ=_If1Zc3zk_i2bKPh;;bPs$@?8n%@?8Z-`F``Fdfs|He^c~7!Eg3& zzZ(AaHw|ZhGjR5I1kU~zk+eu}Qz3T|!-}^uf|AG>q_w9|rao?WzfzH5> zDSF%2NjTb<+(UTSV|(g_bA4I9QD54xPWGUgs$y zy|UwoVAF4Zk<;%uba9os`gj~@=+`CvPkMgcZ?5dc*8|0$_kH$|pDz^l`SsnTFZ#FY zy>^LxK7HqW)(weMk^4EX2D?rRKI7}E^Eo>vs@9am=iA<~rXhvBea^e)cT7jXYziC^A< zpJ3P3Jbpi{bHsHukKZp>RDT0&4f`=J@V-skTV7Y&0_SzLLvXCCJ@ShBp2z7DzweQ= zfAk!pJdgfk6+FVG_&jDm`){eS<+|e_d}raG)Lw00C$PV+*n9r*_+4Dsb;UMu-_PF^ z{f)tw)>xg7cf!&B^=I0%zZrOOv3H%XOM_n{AT-6IP6_F?&wdZblQ(?wIlQ>&{f_kz{8!0-L&JV^vN!z);Yh#FdGwIJ zmzVIScbff3uj}|daK4Y6Ag9*xy{~f(Of#><^6Y0Xm*-kI%G2^6f^+!~HtdI!z0V)d zr+i;a8Wq!jnBPdhd^EA;xgPsmo=a2xcwg)&{9c_Q_AkT9-txBn=koSCdU0)DjbFwa z`YlQCeeyfuI38VR_V|x=X4hR;qrbcOtNI)E30BE<*zacVE2aJT{;xyu7Yo0Pb>F@4 zRAJY(kHep4Q}nl=-|)8)`;+jDN<@8&_(Lx{UI^BoKl=^$B6GQiO=(oJ{8xwy-)vpg>QNo-<;s@m{syt zm?~{;DtfQ$&3|0StKW<+^s@CA{4RJ5eoqR2i2WRXlJsT!!*JN^ZvpAe`d&En`gc5@ zd23?B&!q6Ce}eo*{D$Ac`7pw3zZ=f+x$Yf$?c3*^X8pCPv+yT-ogVz5H(*bBUC*X$ zf)De{FsqUu-NCvC<#9J@RAk35!GHRu`uXFh)#@s|>p&4+z3a-M|HnUM9hUqZz=xtg zpDPJ}>Q5xSJVJiLUfw}|g1@?pbqRQsFp7Gg&kVi(x6mhrf76rKdF5Rk^ErLqCmZ1x z;oobhSJv)ZmY3f%%H?$wj{IKy*Y!7fx{7@-?D2b1;@QO7ABMC2k)$`g{ab|B{#e7_ z<3IG;pGd5|$A9+k@t^JOUxT&pfU|vP((At~@dkJS{8-{pWZO$-#~+#74en0-l6#p~ zfjet=tp{De^U$z;o+58(@KA&8-?P2Pd**Eo-qB#H{FJ@({8Pj`RK?ExGQ6+BvkkUB zvc2V*+49J2f17z_V)JW%6zMnq`Sd?w@A146j^kM#Zt$+c)-ku~ol$y6;?EuKWJGfs z!>asp(|6X2Bd3ikSE~qL3x{DgUNsjT6{U#4v*u#Qg~@YQE2ExTy_>3^_@~Atr|<*r z6&x$Zr*LHz_LOl1YR#`smaUOff2ejE=$NMmMIQ6$D#kpQKDizfeY*-=k2ySt#Tcb} zmoWTH>RFNZ@f++luzg5e13N~s56EK_+uUL{Wdrse(_RWHrq5$6r>_T&^y$wwm;J4R z!(TNf59)n?Td8~nRe%8TJ9$&)V{PR&qy&8so1_abEdaMJti+9GV175|R;E#DaPd#|ni&7H;HW$*^}xm2vg{CASR$iHLY z&G?Tou-CAbf4+vDg!47*2{`PX0~;kRkzb#Matt?K(pR+y;>+dfHB^-6D@`99{;T5& zj^oMlnt*e8I0ju@<%Y5#{0TVc_b43sH9rf_o|~U8IN~$^Yv9O_?*{ew%I^lXece{# z>*jY5{(Rvte!sq^ysPkYdA7>%udVu1#`{k`t9mEQHAR0Tzq9BW>QrKUYe;X7kM49% zd~G<#w;Yc648Oe*eyS0Een6n-t~HN4-RP)x7wHK*4xa!&6qIH%WZfLuPd*PPxh zaHQAobgVR@bNr`Xiy!8fW>m?4M?b^f-z2s?f~Q`p?_F#5W{x`spfA!-H{obTE{|l+Bdx}5v-vLMd@418T zprHF{`ekJvg+I#g(u$b-L+|nC`mEv@#I!~H^4`RT@5DaBZzQ~HkH1PDm8Z}cw!^>s zn3{hP9*4i(Q}f&5X?Uuz{`*pX1BibLE3DPKZvfN+GCX(ZZNAf z{#jl1Q|#wOFjIr3`YECL2&I2)wR-($Uj4(K-tdu}Li}H3r2ZY5*Q)l)4}UaMnE8Vk z6I3FMzoK+}UhC9{-|m}Rtu}w&s17+7Ue51mMU4I6W$eGNuzH3wr}REovjg5-^xE%d zKkW5C1Bd_j!$;t)#on>ZEc-E*`6T*d@JEvVWYTMYJh9JRS@nC0{i03v{FvtxMIJyO zZ2BF$<@{}hBfaV;lHT_a^wM5ptY`f8{gGaIVPfAS&<%V3SKdkd&Zi93*jjSmSylN) z_-o;n=t@`IdLO(R_I5*M1|HzIbhT&HTehLs{}`P8Z-KLa_1XXS6yEr!8hYz9`#%h4 z|JGObuRi;4k(UTRdAOc`a15jvpJS_RzaiOw5BqKKgV>Y{;REc?D{S~N_H%enx2NLs zeQ8w5ocEv>~9*5@TPa!{L^fDX5k$EFr35p(Ef7xZHW!<`dkk0`dkjb4bI`a(z)9F z>_;E@acr`dlF4He+nZ&my#Bp>qZ0dLsr(nD{JoulS)24|YN5mj4W6fy^fo8-6nfLk=&8 zz42Ag@t6F`EAb!k`%T|dX1{5nj1;-5o5JtdSAXO09@?yO^A{PLp?h_Wtt<90mZ!e1p$-+p z55w7?_q7FEe!0F(-x&HxpWzuUp0byJ(pcH|MP~o|Q~maR{_N`gr!k>frH}dwd-+g< zeaNM=(vU674r9d)ebL!ppsh*Ug*XnuL#( z@Ur)j29JDZUUi@16H0qgZ+{Ycc?ZX%9P6ORRn&_sMDNCVOArob)$% zO~Zbuq4zx*OL67qdo-pBw_*J&&skiT_m#9Q#W@!H^*bx7?|bNH|CYkeu?(@F=Zps7 zJjXHx#~jNh?7L`lz5Gr2C_GABAFQ#8Zw2}s-%>cow+_znt%Y-Z+u?}MaU!Rx)}G?u zIm;FB{e`EA{}}v?!rz6*&Th4SuJD@)e+cd?Yk>N$xebmvNat9`;Fx1|&Qkq# zC4TLD*w6NzaJE;U?Kd{;hmyVNcU%|gfBuW=_b0@8?>cw_{qsnJV)_T+$dBo9oF3^J z!M+!MJvPcdDrP18sv2A7&-#w^8h%#_uikz(^yXiE@A&-UEF8}-`h1TZdY|vH-;d{ev>zug* zcNMlge7-l*ckTDZuWE%gI%=)Dd&2IuR5<1j<3O4N5y9$DWBhu-U^ zHoS~5iq~b`Fx^BY>X$Y2%adOJ>*4Hw1DyR^-`W3`6kh*R4gKz<*Z(0n`=5owf0aM# zGwY8x^lKaX0XV{c{JUI7!Lud5e{((K9LnP5h1(n$JJ^3z;cXP|DEjChZ$fYR#qq7Z z-*A^}7Dg z6GnMM*Su<-|AjTS%-_g^=jLw|j_{Q|9QJ?yC9XR;l|HtFH+>5!>>U3hIKum$_QP;b zv3E_ya}B{Nsn)pU8msr;tYv?3Vb@qj*}t@~_sv`0F^1CLI2`_bpOV*P`F%>BpX0ji z1L!^8+9iJPQ??C&Y;`vfM*KmVd6VDLCrO`({tTpJY?KZ}B9* zn=0ZS$|u%!^RdF-7rF@FZz*he>o145{vy2X$@UbKJ6Y3l zJfrxN_p={-Cv5%wY|SqTGV*1uzFxP$<4*PCw+H7%tmwOXWvOmQ34C+Afdz1(H<)@M#^6w$GAJy2p@G`~| z*xpd=p9+t#&u0NE>4QDqZ%TTP=g_|z{S5r8MgKC487l@l@zd&jg&)GYdz2rbm5F&;Mlaecj)R;^Cd;#5DKF*sH0*wN}_ydAA@f8Vv&lk=QK{WJHz_nptYuM5(%zx(;Kp7m$1 zz4qGs>~q4&iHCJDAI}Q0kEdO1`Mr2z_P)tWV|~8;tn7CuPK%6uh05;b9bzxvlgK%q zOn#5nJy)xKXlD}sna`NKDF4jY&&$L?zRStU|G3!aU!(Gx{a;JINo?0Y@OH=er@p*@ z){V@5V&%l}euk|vf9$#bGJoX1PRE1Kf0NkkksnCptb2Jm>xSNbo!I2$w_j}e-TFbT zL#h4-8nS@i^`VKO|K^)A{s$F?{>?dt^v`1EllNfmVgEMTe@fgW-%m&Rk0ZZ5%RYD0>(n15ABaqTZK^-Z4}T+K^LNFc z>HD%8uO5i?*D2=v@fXDUx>wA)?3R4ykafosVoifWO5bv{5r%ozOtUd@%CiQpZic% zYLZ3rLE`+pP@E5>`K;_YpQ&4g#C)V*+cqD0LcF*#n{%KrBJLKq#QLPZ2E{i2!T+#0 zAG~G$Ve?PC&9b$4so!?7mFI}$o5Z&6i~d%p&+j(MBv&2FLw&Z1eSNlwtv(NI%$^~8 zD(W-8=#$*n?{=}T-%+u}vqAQCvi*CV(?~^emADZ7@jG($+P)$(zay8|_8F1+9X#en z(<3v#Yu5IH%w7223~gicyJNDa|ExYWQ{Z>@TE+f%_BzGp5C2^W7afOZq{M4uetfT| zU#xBt67{!F?CX#IY4t~aaenmmRVnuMwL)z5^%E7OD9^X6%}6a8_=fcPNkA$6o&(>j z$SFVjUIp*(D~R_7ZLPiO*gvf8`pD>aIQ=Ule@K0dgn#PeFEY$MJ(TJHk0|GR87f5hp-izk-s;U>rU zW4>VisNW8;)i2+x;QaQz=pQ-rQ7>PwvYY(*I9`*R3(p z^j{TG>-SA5s}e=nsocAOAd4Xa13I7n>aWX8AUIm@&loUE)F2$46p3JoC6% z<@e7#ZWKQnb&fHApAs+8!KTZSkeE-xem>bHZqLNIf0TZ^I=8LA zFh7NDe#-ke_UK^QuRcU#KDJPmy)?tXn;i2DW~T<*bE2GQFxQE#eJFc_WZD%H-@9wk z*55?`e2=bGJe-Mhe6OuT+xERS&L3^!zld`Db!*%F!J8A_<`{pBkLC~kJ|~BFCrm8n zpYs#r-xsxygnpIS@{`Rk4oU9YZ=cgA{xPwS|Adp%Kk8II zt3T!=oPYL8hjg;`p}u{*tHr)NoUcqydm#7Bc<-6=8dLvus=sw1_pE;Wd;@TfpxJwwM+n`HPP!SpX<-X+#2zF0P- zw~4tP9g6yA59!=pscS;*BjLY6Z2sw=C&boYk+!;mffBkeqZ1Ma*F&;8@`RlDwu^$ij zi>-bb508uec$iaNo1F2G^ASEtjEBu)KOQ!SznF=``D%%?r#zf*{CL>zanukong-mmmGh>!-XUksnAnJwK)6A;*``Z;ROS8~vr`i;8z)jEDIP*Aq5>d9N0J z8IQG08xrynCx=a(I#Bpf5K4rSjJy-7m-Z$p3)+`TR%J|15v(S&uXO5Wnop;uo7e@>-SE%bUd} zr+mF)@1Ng2H~+}DicQ|Bcn^qgjO8O9#y``iyp>`rKla@3Hhc8B9y5L78xa#osh*UL{hIrG}} z@?-PbwsqM(H0Elg({;bCsLvk7L}H%2Ri8FbzFcznt|;feMVokm{zyDWv_qd-ZVu@# z`9GkoxzYc7#D~QB$Y0ids8R8G`XhZz@>&~>%7--ZOx3MqO_A@*XV>Lg z{t9Wo?As*&rwnsH6f>{X=_I6@+qG^k{vID7h! zw`ZPb_T!LjO(k9M(d z{}%C+QUCFc>Ob24wqho6AB^_&_rZq5{<>t3*k70AWNYzoAB=Sne;J`rc{gHTo)HZFa z+k}LC#K~d$gXzPE9Am#kW2@N*e~!`TzPZ^G&w7pT77y>cYMfH~ZaVrr_htIXxesz# zjF0z$)u_xjMCN^1@T((}Ki7R05BfYuVfy&*RU3K#gQ{cmzd`br+R}6PJa@80WB!Yx zf9?~s&nmx9uwBd;MxuPH=9J0@A9hUsHFHb(!}}9%Qs*N+JwiJEUcunpT^Ni{cybl> z3OA$*4$N$e*C%9Kr^OBFsTs0;u`g@BTd$@J>E8Odp?$GyYN|=w%07RW-;iA>)th78@Efw6JKG}j zjB`P}HZuM;i5dS$JR?)o=c35`CN1Yq?{BNv{QbS;^xN^dVPw1K$9KI4C2xuTo?Mm9 zb?hD(&n(ln{+Z=U@zQ;3~Jj1bz~w7;|nqEV(~79TD5PiD#6k z-@N`voO`HGf9`1z-7-i>057k{hOHuqy2e&{dKxxwr?*DkCqKi4*i-x~F)AI?3#ehOl%ALMO`oM%W)&bu9& zlpn_giSPV$>C?XR^R#$K_2YlDX|H%?(AIUJF0#fnU%Wwalct|KA@GgTAt7Jq?jXCvSD~5TDrMgE{_<`A*YO$?ZE$kBWy>zczQ9EaoIIFZw@KToga1 zIw0{KCF-B&B8@p$8?RM-u)Yd4E^nG#l4H+z=}rH~@6>m~bRwFhJV?aXtGcrI#uWc% zvDGK>GT*Uyr~i@4r}Ef|>we|WZ^Hjg%pdtWJ;QDCUi~K4F4gaQqntHa#y7u)dcg6w zCTY!7y041yq2H`+ug{vK@dW+G%Yc}eNY58@ZTybN{N_ua>c@Wbh3~$@_8mavJj3qg zu$QyOZF2OtJH{Vu`Y%-eN5#bujeh&}Y5hi#>ht=@w@OZZYuF75`F_g-zKs5*CqD;c8k5gU1INVOlONw~SHB&WWHG=5lpos1hbUU<1@{P#=l^WzJpo(qmY{Y{O3iuDySUU+fBOA;<5 z+>&r>!YdMPOPKS4&tFr(GG3igZABKM5{{gPXL*dI;wJ)^&zC?&BFt^gUmAhn^ADcyUi`kAZ*E zy#jH4WG$tKv|aBj(P^B!g5mun0}1a*czb-GiH4r=yI}(f?@4%H!Ux5~VD|hD8F?G$ z;`>L8YZK;o$-I7Z!i9v}67ER2E8$|oy$SavJecrM!XpWfCVVVmey6NW^}*jh4e-40 zg!1yo`(yTrsXzXT`W>=SabDY2U)l`o49E-d{U}^ptP|U{#a(x2{Bx9*kay_VHaXWy z-C}>O#I>}|?Xl-t#_actIrh&{esXi(^@*onZ1GSZ zOH@9q59F}PDKF2LSb34}*SXQ;@IuYCjj7KzHK4E06=JK;@O$H8t1o`9oVoTNs!mA6 zvqNn0z^tzslOJsP@q6HnvTF|3x~>+l()K-(UoY;|c9)ZHOypYIlfJfza3)lug}Tx-=8q`Z}!M{ImX|h*!$b<14@#?%kL zb7T3h6LanM&(UA6xL(|nndc_f=slM1d~q@MV#Zw3b&@yegq#ngX}fr__~(()hfTjm z@%ry$V_$Uk|EKQ-#h&B$&-o6Z`zf#NNK3_A!0(-|d+9es+uBAIp!wUbUC^w@vK* z(Z1eayV}DTe|2K-uUXgL-XHH&_x_4vvnReW$N1YVHh^Gr{ppL^M?$||?CqB%-0YZn%?j~wW*+Q2oE(4M3Ag?R z=KCr`lCO*Y`2H&6`*-z6qP#ivcPlTy6Q(8!#{_boFYgmmW)k)T>fdHhJS}2hzE;P4 zAGS_)`MsDQ`m4oWe~06(8tC)N$9_kbwZ0|d$Mr|LNZhPXTbtwi=j+AQ(Ld|%#5&zE zb!p7+1RfRF>5s&{mE-zc5J=Y&s{a$(wtFj|mj4?0dShIiIv}nU->7;bk$+xn`EzfG z^TXOq9Nz!5PCP#{-yc69_RmT5i7me4l8;IE^60wdc@yhgJ&X>l#4`@X8 zXKNpc^3|&SeEI6czI>dIeEDd9U%tg+U%pLZU%sv4`B`>3;vW)Q{4Y^^F@E~B(;Bf~ zJ8cu+DjO1OO&mYPKp8JinEr0^tEA7`=j^BtpGcVeyu2ade4@W7;g*Ek6Yfm7JK@a< zZ%ep8;oS)jCw#~;^+o+zeZaj5??{;A*X!?2c*W#$d0WEm3G;ne?~k=!&m9SOCcMru z$NP$0`SD&5+wo3+!nf~F8^yjq4T-HkaWD3WIIlkv$3utu&w>oY@y_vT#{t~bqfhMXqhIXnqe^4;T2~*u|H|s)SDIY+il=fOQTo=~ z^iP%M_YDDa$0X#OOQU_G_VM=wLM65f`tv-P zaD#k#IrDSR%+Ebn>Uc7KrF_A9Tq~rx7wh_6{0aS$$bU4^XFM={8#ap!96&ZP6+a`ziI7a_~*y}SN z@cNbN3&!Z@#a_R~$?@Oj82$B$ey@|Gzr!*5dlG%-YnDIyod1o{r+@l*>cw78fAsR! zMBbUmHzx8OiF_cD4<~Y+g39tgk;tpmAANj_#XdgflU~mCmC0%UK5?_!h{U}<#(z!2 zLi+H(>e&Kqy(aP}#k;h9bL4&EecHY}GWy39pXtMURlmmAkBGfJ>lx ziLXV+gO9I6?EQC(y?w9P?1^VY{onLq?xh%$Kj#B)&-icl$PdZh%a4mqPJC552UvVC z&-fejzSU0gO^S(xoZnb7Im~*nG5U>K4>o<`<@<#`-afI9XRp}eK|iNCy4SB0d;Mmy z>67288Cia3DZV#^D3gCW685O&e~+3S{6@5Xm?;&?lq(b4Vwp61Cd)amb=$c_s+Y8R zsZNMKO-M5GI6=L!%yvRK>g$Z&RW7!x?aSLn_E{EVcBiGJpf?4T;3Qk)(&A2}362{e z&3^o1e3#ytqVvvpnwT3c$Mxf)uyIQ{-dc`G(jpr*M)z zgIbdJR)r7vJzULe&*YL0d%{)8xK-Taa}P1+pR-5tjK^FEI@&Ythnbo@>u)@sOsBG> zbCSGb3-Yce#t{FmCi>%}!K>c!y0#)WB)&s`B;Kt)tk2T|>AP!p=v*q@S4Bp?PHb}Q zH>n}L{h>s@FOf4(_VyeD-k!Aqlaqg?#(STCmDuDVf3dgU;^f#fclY)@2j=a$b~O8; zHM$oh`T4QDyjzOrY`H!l@f$qw(hPI_1`cz5zGp)s9`0S6KkBPlZ27?hjzfI9mMn`; zZ2I`yC-(lhcVTkud3S@i-zGLW_Q%BDp1G6Bc{lt4ag$=dRDX64hO&?xEmpqgz6>r=`x3pO( zZD!S|$YybHhLo>xo86_oTq%Z>udvO^(q>J5JoK8}_$KW(9-3mvQdO?Pro2*UjC*tW zSWK}t3vqXTVoMvFDtzBWzLk}2P8;8Bn%Kflk~WKcDm7KKsJE38x$tM#!fX?S@v$TF zE7m1;NGZQkdx!LA*Y+uM5Ac}&n3s@JMjyGi_dNZFS{sw?&!TSpZB@!HzsX7#b1-!+x)>C=f>!_i@iSm$>bqEvE>ikjD3yT#_ZvI!iy4S&TRVlTP-&K#M>pd`0=;d=_B8s$ag1l<_Bhvf9BBM|1l@W zKWmE?56qm}m}hNR3vX7Nkzm$lJg;)hoZk1RaE%~4>tFCX$ISV+h%MfcZ|dE!V#^=- zzC=#{@$xFyUZ;OS@w)cGpYJ~~*HC7UepjN;$<5^C&o!0B2k&x>zagg&A8?HR5vNan zIoBR=on!2q#Fjt2B;mG%J00VXwL5EX^4}`9`0>Z=!Tckq|9JUGB0ucp_&+Z8{{1x> z{`0Q?z>6H$DL>|?&1ypuye?s`(M---r|(a2yX>q#!Hh@7tcCWAtvqki%N7obt-Q#O zCi1*%FXW3{do_MJdp6X{i@$EC57R%)Kl=0!ukZT@{^=hUAME=F{D<5+DhnLeoOpbs0 zhxhNV;jv%s?BOoQtOfi20&_33S#3;$M;))8Q5pKrecDHY8Gk$vC44;LX0?ykU+;LG z{0)dtbN;!0GyOM6PJ8}YrJO2l67#(q`>l~Pb}D}HO4*Pu6Z0(a6&dCpkbb4+4|hg+ zr{rUjw?=-e_=NbSQGc#FBWoaAqa5zg`QLMs^o{>m{<%J?bM}o>OLFY<)#bc5VZJA0 z{?IRI`M`6B)@nT;N#beM{Xs83R9oUtD!xs#)UGbR17fe=uln`8ZcaImCA?pS_4_~ql-;^gRW zbBz9AqJO~2(I0h;eyxrV@1J`&77y|jiJbY4mv2er+Y zH>2ra7YEXBMvC*=zCUt!_FCH(5A(x~Vk-~wt%-a`A|G;c^!Gc)-;qS0{^#?f|M~oU z|HIz*Km7at2mAhq{U3diVgO7wkyL(cf( z%eyDB-|yt;)89;gNMHCqA#P9|l71rQ{CKMVNX*yAB)9V+_KRKrSgH9ryeU%;^YKmM z=E&&pO7vmVr@V~!RzG}~VwG&|yA&Urt?#X>3O}S6NZ%EAY5Q{-=AISrk-w*eipopOTxQFZJmw$t5ms`Kk@F6+{fD|ws?tWi%upM5AinCmg0r?I_A4NjXK}>@9OZq z??4v>F)eR=)^ zVxRx8*yn#xZ29w^fu>WnMxr(#@t%Q$D*t<9d*FY&>d%*Nwb=Xb5_|t!#O5D=eW#cF z@oZ45^08-wE;>th48#xWk96DJ8GkUt@J_1r+I}oD?^CH_4xjE7%~`Q6*7V_Dm*S(84b{0idBqx`>1 zzDnDdW#Syqo(`e^Yj~;3hg@bM@t&4C9e+jXU=h-;s=WC-r&&I6~&yS3LZ=%n5;q@6WOdo%T z6ZsJ*=h;2Rt9015qTX%cJwm#2q#kH$`FOB|?A9wO4vPa(M zfE~gk7r3NKJ;1dG<`U)vDTRKw1`cg{JH<)^FJuL<&T{EFJ8{G zqh3Ce$VU@--;{FuRj=>LyJW_->`ZsQB@C1Jc;BpnjW_+wq!h*LqyG!kuqVXVM}Ac8!Sge(jP_Sb-XcCd%K2@A{N&Pa6WpYVT`c~5lyg4h z__FgM&+hUp_48bOZDNb>2ld(eXzkhEYnrmLt{DAs{%g^;oev+^iFJqg?^I?I`UCnb z0!qA-_JBU^owVqm5Su=~{S3b)#zX!i@@4slXSF4JTskDmQxN;|6ve(g%=dkHhQ!`q zll*&s%-_AgHnI2DD)#<*#paLSDy03N$ifT9Pu0Zo+H-sk zi|0pvyXuSf9*)d!J+2Tx8M*eO+5IMUqmZa?p4GGZe!1+qKW_O^|FG3R%-q75`2hEi zo{Iix51zHR^3uLr#lL@ zs*ibjy{?x$7ZW~`FyHm}`g;=QyZTmN+djs#@sI-7Tf*vS85yf_nSSVf7I#Yt5)Ng_m|4Yq+@dYBlrG2Un$uV9WVGKrHnmt zZ||A=?}L*2`|+N!hs~b&SugYPrSc^@KE3>a*zED=nfvdY&zJ*{Qs(|U=fAf{IlS00 z_v818by*n__tzQEW=7_Ie5H;@e}DaiE) zepKx5$1@)_d+x`l`jtBVO&|Y!|Iz&aQgQNKX|u1Ir27ZjH!Zd|%=y%~b3$dfK2*00 z2|4o*lk;1gqhk6v>2%2#YW(|BhI&V&*7wB+qkn$8FQ@uaSRwJ-ea+&zk&#n>UVb={ zQ(sw88uPpubLgcpKAsojoyfx~GYS1> zvFTs0`Z}obkZCFjd7H*Jlf#D-?$UT>a`Nlec;oZy5LX3+^RXE z=K~4zHeW9vOSpJ?NlyM9XOzd+BKGlhC-Pn=C!S-f9G~Cjd^y*inej!czuT$AvQ#Xi zp)wjRqhn=syo{J*meMn&O>I3}Mk>4|J%Hs%&tR#dN=Xzu75ltwLaw5e!+>NZLUWvo zvafiXc?@)A@TGsosmPvjDwI(^qrO~4usazkhcnb)LmDE#aSA@qh@$XOaYim4wqS{L zex-a-#*t8>pxKaX??fr2iiTp`;zOG46_WLZeW8|~LqSwlgWAJYXDLzh@etG&RH}P* zf2AQE)ufDgze9<5;*u0UGKgeeO{T%XG&L|f8ZTf_FtY&JnA@f8EO)(UOmmFY;#WjD zjNJHUafkSLEG69O82#=sX%ceuy}V!S<$DwP zkdu?&5wVr;VfibHU#dS+%EaHIynT6DC$RFKOy1?>_`}}(lRxVYmOs2L;X%jv?-g7A zFmhw+d%M`;fmt^&Mt^sr&voK&lINYA__4S6@wY_v~J6B6qLjp8RV47|iK>joWS>p$eLU1j+&xxy_Wl@8 zy+7pUkN4Y;i2pqLzh8Vnd|zbBb42XRdsJ-YAs&u59}nY^#e;vY1HFHa$NQpx-WT64 z-sRpgnm_dOV$&x-`m4$DPk%Lgc%5VPHzoSLPL6;2uldJ*uamVH0-e8Q}2dO7z^jq$fyZ2pL^OKkDMjHRCG|HkCMOKkqoA98a1vkv3^ zb6?f$(LdoBf3=k*f8iaP;^wUTgy&2+UOaPr81tNz@89q$*;;<^X2(1iJSgsr{^5g; z(WgI~J#zZ9>GK>i{n^Tge2J68s~qF6DE9ugIXV6Y#NOXtClB#C4)Li!cz<=S|AhEl ze?gz|*z}QiI63~iojtt8G5R|aeb(iyKJY&*HviZka&q{XWBgUR{sD6z%$I+m*!*K( zNVq-Wb&m13No@XzuUBmG!TkyEaZG;u#g;$%qfU{`Ul zvH!uH2^SsnT(j@rFlDjwzyprMbJ4Cphv%SO|3RPrYWB$6oSf&PJH=K$CxP9J~FAH2Vu>ks&AaE!kdVjmy<)#?v_^jGtTKK<47k?(eL{O@!2ue`%EeqvY`CBO_le6EBBN&TrdwNdE?BKc>l= z$-{5!Dc;MXyz-ZNC#Lus`5+aT9{np9l`rIe=`pCUL>)hUdVM@6^a<9*G zEM9*s(MN9j3l$Id?b>xrk|=+d*vb!cJ!MRJ7K_au9!QvXrg-`Kg!d-gG^1Rf=kmS% zF2}@kOlN&|Q{2cM{dqv)!-;?spS6w~#Cfuj~8dsy*Vz)F(+P zW1m;wHT&0GmFep;C!`n1p7C*@`s9o~b?I?vG?C9Hh=hE;h1=a6a4`vM}L>&uZs(l zOa5uUy<%%`o@<5ubFE|IN3!t4J>V)GpSE7XbGv!*7oz-k@yC5X?{9nJZ_xSU_tZ95 z>3E9mOMP{Tt-ezEF_jN}t3S$n)G@zDHmYk}`#rMLrs;fmw*0G)kj@uxY}9kP8Rj0H zn!WSM?hCg{&O3R3tU4uqwOajM$NPJuJ@Tzuyzp|^zO&8O&8HxOh_2u~=klgY| zeo*(le0&SVUcOH!A1`kZoBTG}^Ch6R*#F4CS?u%gZ78=tEVc+Mcg3X zorRzKfOv(NYZB7O#9PHrXP85Oz0;?Dd5Mhq$9{>}+iwzkd+ys=e#FzJHK@JL|6;Mp z@n4rO_untM_kURI{SP~P;^|HFcgUaDFN!Tc+UK~~*MIYwrTmGf+cElmXO-)3oL{a# zl5nHzkJxu5{dG`m^+|p;3rbA?+;omO_RnGD8vk}iMqbdtV)B*m(fU*_yjD#AYVAhc zT~Yd8-OsmX_a80~QoVclT^f5OzcliPw`e_BtW#D<_&XrhDK#YgalLp&WaP}Rz5Ile ze_Z(?uhl*h>q%YcM`rzTv$!r3=N`N*d+ttE5)$#&s;oZVoY==(Ber;FzB|)@VYGj( z%2z4*MUk7L{6&#pDtVRU&5?cS?ue7pV-PXL-l!Be2U6QqP#p8 zXma?dD?jxI`}%{e{y4vN%8#8dPsYq=OwRmimHMZjKQW(Q8KTi6B|0Cm-sa~^TYrO` z`4RIMKVS09`H|}h^UwT<^(;R>@{B&$^QO=Ir(X5r=R=;+=la3)na}W!ML)msjQ*h0 z=QnJcG@r2Fut^zxo@+Ay$PXm?p3&#}!t3)Kk@-hHH09UJ@z4Dw(`SCr;^qgb9Dm!T zZ~3MA?~#3hpd+RVB7~m`{26_rx4qXQ+-zFh7W7Onm)fn@_=`I{1xwk5yiF`_;!t z7cQTe>(HmB7$K$1d!sf=az~W&-l_xI)|4hB-aEBLA-pOw@~vW%V?QLBx95Cfa`I!n z+VV&Lh}i3oiA}$uQ{R(Q>^7c}|7!JTpFhWn=`WH0Cbj)}F+TJUB>D%PKJj#_ZhbuK z#6BLLyEXs#YmtASzyBTxa{9ZM?-QGSD6jgj_s?%un4EaV#1;?!2GpOuzlCDc$KNio z_c!R|yyvT{x-=h0zd`-i>vR9r{E;8`JAHnmPL6%2o^gEHct$+1#RGHx^-NyI*z=sa zw}(y6czRrHfctM_-hlJKjK+cU5bbC za7b+9A>(t^w9c<>LKf9wa;pUofge)U%`r@xvUe~ZQ5->BH@k7zy0^zq-S@_X*r zWXCGUj{iZDM_$E~Jy`?@07_{TBUWi9YLf-v1)8-RZEe=yNM?Bw`A?il?ljUPUqC1UTt%`y5NiT*Yx$Nzw1^!FtCN1Ysf#w*JYeXa+6 zJj|C(jy~&!#^|q0^mjNp`h$+qA4>GO|7QNsXM8h8zh3O)S>fd9cQ_95B>LN&9Q{7W z=no|N2b>)JQOD>Xar);0==d+5%C0Xgo-cepyPsqHb8&~t`YQP( zolJjN`@Q}l@o&=i$0Pn%sEWM4KmLe^^%d`bDA8wq#N_BVIL7}XvG?EYoY(0{#g$(M!!Mq^;bJN{+SfFC_>F(oH+=leKfHdQlcUdjF1>!3|GWNEr)`s?&-X}-iKkWU{cm=1^tU-i zzt8DkqWkxG^*=k_xL?orPOSgK%y&HRc8om_i<>?Dy-o8Ui;w&}9FzZgv5&XU$u=<~f+WBiXK`h3sTXi!~%Ua|KNd;hFgnScCozV`lL?{A^l z`-8o|yx9B^Z=+-Uv!3kz!`}ah*!zdQ|6#HBzt1uLTQz=o|FHMpE%yFl@1Oa!_rKmT z{yG19|FHMZeA)Ylz5g1q_h0K6|Gi@GANKzDioJi>`yUj0|BQ#m_+O;)%KL}C|2DDt zr~Q~;nEf6d51gMbRUah5yhp(EYKu=ZiW%~GQcXI5xUh?_D-k$a}`=h#8 z9+{rY{gr%>s6Xyk`T8k}t-R#7PHg^I&sjgYw4M{>>JKKTe5}`6emuWHfAG(5)TzB6 z$+8dk5t&at5t-*VxZm-G$k_9}U~k_d_VyiOZ{IC8d*VAGw)l9S4PF}k6EEYJ#fyHc z>c{kv@6~#c$$6ftQ}uB}wCDM?g7ocqG}JI#J>JEiGBU`imm>z->mq(edFA6`-0fpw~Ni5{#Mj_ zmGw93uUl;OxA42FZ`CE&prl*GEA*)@9}@Ca+Vb+%VlQ8x$h(~U-Lfx=)onu>7H`&; zx>!iaw`$9A6F=j*_ut{<_}?V<{)=Moe;~1EebD>goyb`q zG&%kch`s-VV(*{zLLVRNiDpmyHHo}dY;ycB6np>8V(-5tv1k5d_V{Q0(dW;4iplX` z6np>c#peHad9NS+@lqWFq?5hZ4|!jZX7bHqEpdm$d*a5#_B;sh!J|KIjdI>QM}K%< zWZrw%C-(W#KHlF}#qa&~ioL&KvG>P(%ll*f$NZ81KC$WZ9zo_`=f!v^Z>!krvwmax z*bgN<=9u!ce&gfi`>HBE~G6MwJR$ItT&<`4P7L_X@|ya%vdo$aHpKDl1?@g5ah{OG{-D_G@z4Cq`#&r;IsT96{+0KCKy3c2^#XeC-(K&^lM|ag^~Zaf zOrQGMF81}eS#0$e`n&FL`FM7V%^&SIsPntm@1Iky&wRu5>A%M`Kk)sR`(Gv}zPh=k z_}*Nn>j&}Ibxe}pCGHl>B&2t5(eqr|QV1dOyxt;hsR$u`THLKIg%A?*E^V3oYvMt% z`3ukcIr$GHKPpz1At67aEwlfXxJj|;mfX0!QEc)#_v^bE;!~Ww!^zK;obzM7lkafy zWs)BfYswQ6`5$!hm6F%H@dvr8XgvSdN!~8D=O2)BKC}4lmAqHX9D{^>tCMe+e82eg zOdRCw2O+4Sp$DDk=nZ>mAp;7(8*UgdDXkrAH?T7Irp>7{`rz~zv==fA9nJ~B(KzXf1#6e|IF-p-jwyp zW+z|bD|KT@A3ouLy+Uc6g%K*D~B*!+d| z6kB_~WxmE|+0rLSycc#rZ0&VTi;i#c#^{gtH5?V+m#OEL%bxlD<@zIq_Ei73_QXHW z>zM!h5J~sJ?vpnQUA6 zsSnR@lYNus7mrtl>~r*&Zs}WpNtyh2OK$no-uuMX-W)$i#Qyj>B)0l~MET`vbbYTF zNgOXc|KpDrp8vG^=XfcK?RX*HqhgDfcm@=|#Y2A{6kC7hcp63?`!`&p`I0gAjc{yV zj*rD+e|)rw%^%k@+Y~p)Fp2es?fO&^LqguKEiWGsd-iGBRTVjn-_q4&>x$@}N}&dX0YIsTbXdH)S!@4rRt?VH8ke_JAF{$+Cf zcZt3KZn5{jBeCD&?5Y2OM83<(@z4Cs`)B^<{f{N~qs|`xxrycFV?O8O&x^hPMzQz5 zBC%iW?D7BG)&n*s{+SPY|6EV|`rnrL?@#0diF{8YXMSnrBfkBKd?b;NCi25hPW@L^ zme&vSQH!7T;T1Z*Y(4fay`UWax&BDV>Hl61dpYM@laFl3)_=Z}ndk6_-0L4#{-)1* zZ>7#(e*L&gyfWI8KhKMpKD@~>`N2Lv*z&u%TjkNY>;c6wvTeotiLJ0v!J{CDVl zXZhhD_Woh7ze;1d>0{rm$~DFw_V%#ZvmVX*zh95$`PN6WaKm)~>;F$g{+l1_-Dcu1 zL?*uxvE>K%sq+|9KG^c3zBzyS`ra(|^0+lZjSk< zeqdieu<27j$Hl&WYBZVl_0uHw^-~c0`YDQiez4CEw)}{1+pLm4^|e?R^S-=`#8%$b zmudVIcgOSj8`Lhn;=j@{Lc$;KZT0!XK7ZKqze@fNYjNN<=dV(W0p8yc$-O_=`-9CN z`LCK=(%+%}zeD`;=zouRm-r7Oqkq)t(>}2G4}1TWT1>EburDOMT5~3o;}7=!VDm?R z-KfO{;}DP7+aDEsd)V~pFLhd6@cm_2b5P%34vBq#857fAZr0yM8m&C3b$kk_9upHd zp|%i6<7_6)Q2R3MK-xBrg_EFGMV@0@tuZQkHd`#SG+PI~mkKLtJCJGg%OS1EY{Ao> zt6b$gU#B^332!dQkamyLdiD6jwsB5`EqGK3la^awdK-UC#}vEb8Q9b zQG!sS452bqF!QRluVkBH0Y%d6Y8{Q^(qehlm19Nbxw#Mw<#~249vNYWu!)V@jGil1 zR}Sxg3-Z*q$le0~kY*S8x4=J2g$$Qx1gbaXN|<{tRuzJBo0NkzLtF)H4bpLp(uz!4 znJ61}WVS8iwiV=S(!%V3iIn*QNz(8UU>r+flu2`$az|9o%?=v{!-s4m`xIuP?BC}~ z^Fc3b7D=-WLeJ$AW$L-om8c9u7?~NSAhW0NijE9J)v`xP(n_QSAfVpzZJ9sERZ*g1 zLH~pln!BP9DeDXsR#(9`XHxY^)C@yL!gA#a<_YDBoG=z6*m`L(JRlVs16Ku5|Adq( zhXcoi#q6-UDpS>X6%M)tqjGzulsip}I|Taj$lufb82ZZs%m2rEpQ3nal=D2p5y#lK zN!Q!Ai_7fAW%iD-Up1lJezn-!4~V_}amUzqXuj(6?-ZBWi_7dCW4}%<;q6(^_4eFv z^Y+Y#jlU%OVaeYg>j!_Vzj}XtV(*XjAhV}FmuP-#j6L_uy#0W<%--n-d(C&fJ>vgMDz&5rTM^EKWdzgyz{@jhDbkLNeN zzrBv}w@K{%aX-!b3-6~@A0rV@zvSNEamV;$eaidWrunw_w_WVxJ1qA87HR%$jK49- zy}$jEdw<-o^ZpuUllAx`OEemLe+9Ak zw^i)@@jJ`L_}d}&{`ehB?{A~n`x_FQJ@M6QR5Qlkh~(bifaKoaF0uD_%-Q3wFsI~? z<7q%#W-l(YclyCz^LdNs(MsJ9cl=R}kM){gzbxxsImXMp*!$ytquI~l2R-EfPoh5l zTIZGI#J55+A77i;``acqf5dmhG4<6i_Vu+*ec$`rF82PoKj8f>cH=4j#w7Rt4omL+ z9T9tfi`;lhe5}WtfBdzGy+7`+d4G$=-d|B{_V_#C7=Jvk>irdy@pf}E-i{>WZKE4M z!+4#Hw|t+($8*@}6VGBdeqzu4BwwDr$#~223cfr%ziN#A0kOB|`2(MSgBwrDzsoW9 z$Hd-#OziCoZv4c4n`7*2lJT=v?Clv}Ene&g9An?4@y_SJQ0(nDioN}Q$F%1bv2V}r zWW3y%jF&@V@2_q~iSfty>irE#?&IV8>^?r$GrT{>M`Qf4zTy4xeP8cyKl5DJMzQxdEH-=Uud$)T_&XrE_g6T*++T~>`|A~Zf5#l- zuTSj#?a{*%KEAzT@2~Fp<^H-I<8PtZ`)kRU`)d_@f7`@9z9WwD*Dp4IoZtG+Ea`K; zJT87&E=!y*dHBNnZx@??&X-M!?@ywBIA5MslJoxd7Rerqa_+|$wEfn|?-Vc8_Srgz zk+^@mQJ)V6%6>;0{$OPMZPvE;w@qyRcprRTz7|CNDvh0uVs+V&c%OTVcx~i`V(xdT zNFmL?N8{xL-Jet+Bk?{e?hQYgVeVqdc|P?6k@-D|{o)@*zFP8D*>ew$^fMK%Pt1iM ziTDnQeSD*0A74)AUmu^EXgt1q8+AWc=Tn~fAo07@N5z*%|GbZ%^Y;%jafqMu{gaVj zt@w~Xsys=Quhq%lcwQ#IHWP>UPwkdmLrh4#f6DV4<&SIj?@5Qm`@U+4C&L`SYuPS- zD)N7Cl)ldImqnhsSQqi)rpVL}&oBD=IWD&Pq5bPscHjQgg|$EJ+az8a{n5S)#VTe< zly^Y!SbY%xCdqyLJpbb3-!HcK@johF>-;m{F#pIKrj*O`Vv|!pTgB#|^7e`sM1S;G zzE5N2#UJw%pWl$!{1N|{*vHTNq|86{-zv8F=^uRG^#?IO>VxMitv=`<$bJ9lck&0L z{FZ3X@4Ymt49tN^KT*8x;(VZtdEbsP`g@%|JmPq(<`$#kH$N|PK>g(8)7MX>*w+u& zqrQF`#a2J~<9?9&qrTe2zP>udzP|d!R$n~t!2Ig*SRTfwO0kVkJm0|etAD zA5W{;$Foc991SGKE54dA0O|hvH0*;KfBytgV_At zKUdE?im#95|B!f4d`)EfWBr^``_tbRiG6vQ&psIS>HnL=zP#OHUtWH<)!H|9I_Lu9Ma0@$j)ES#{KFBA;r^t z?jw_KNIE5k^l<##66<@V`UlCL|5aBBNngVZ>Fn4*YL8$sH&)oroewo$t3B5~ zq!Yf<3&X#;+-u@zY`*d1a!t(ddr=;XH9k)`826vkS`1rLXXnG`s_?J+SNy!^O>cS= zRsCrtc0H3-WB5)stW()|dRCp``H^mppNr$?XspnGkK9#Rna^KRk-Jl4UzRdY&nm(1 zjqS52c0AVDNpW{>IYE*7CbaVrhyUDFfYLj&kML|eo0EsKo_s^x|DyPr<)-m*XVzl* zSi#SnnJKJ{4U^TnUgbY6tD$*=CeH#{sl=Xt$ad$p#7_-XAwBcH(0sWKr~P?|_uN>X zSH#a3yeupK&Yk(79}mOvar=&%4~Lr2@D(^!586$jbF-eRf2yjP2g}^~~{i zqy8iD^W2NGc-5Q2o;Sz+A8ViYzI9h?9>B{;l=gNXp4pYx?>ue1?Il>)eE8w(9=+)! z?dRpLUo#K2+=H!+v+r-2pF8~n-M@J8rP`4l)-3sDDLbrkeLJ&}=tOM5ft?)1de2>q z_Z{(oDTKZHEz_)6xo2bhzb)>6ef-oVXGq=gvuQLd?xKIbdJ>UUx2XuLMFeBzG%?9TnFe?hRE9qZ%P=y;fm zFr}BodRrYxbIID>r5^8P>kapu{?~Mj8qrZX9}l8y@z85_{!bcj=zigxojBAeI@EGw|HW?me0-t ziq)C>S?n*wM7lrfzbxvn&35M=i1l#d*~~!o7<%GANBPd3SwV`iJ=&uFK$K5q(N*c$ zo@|GPp-|FmAL3Vw)vZ@%C(8Fx?+^9&k7R1OS(U|mi^NncEZ~Fj%ct|Q116SQ=dFyh z-ljH)oDKG}r@xj`Meek-uYG9NRsC&ezT>*8<&rNChueck&ek^R!NzMFXFb^RP%KW# zKAvaZabucclxu5wogjB_vGu{6m;5eS7IsF%43i)3k%>Nr9&g^++-0jCd z^6(uG-`;c63s!z&=XEQe+WGaHRzCHm_LV)H^0IVC7+~U#o1W68$|rX2JaAplNAB3U za^=Iff8-Spf9tb5U-9tGJr`f{a;ZLrmDIj|UC$jKxn%AoN|)+y&a7FN^sHUGfT_T( z^VhCjJAeKixAyE0sih_R+`ppd17&H;sXdoy3Ze7{{E+0MJVhVD!-|%cmgr+kaI|I1 z_AOOCnG5|pM-CK7&)NmbKBj@48|L@4pZ$0HSKoN!tQ)^JLTc;z)ADp?&(&Fa@r#uO ze?2-i+^57DqWZo1>zP}1(e&xlr;ewd*-ATiw(jhk_`>jiPj_KT+5bJo8Ar1J3lkdk ze^bxgvuCZGtN)a4T)Fbuo}O6kcYN#Y4C|FZ|h* z#l2ZuJ+Sk<9%OT4;n>bxTsJ<>&hEMPCia(7&w0rx++)h7a>=EC zRsG)T9<}v(vpYxo*S62TCAXxhqeru;*M000->iO5F<05SAUET+6%|$cRuupBhU%^- zTPECl`Gksl7rY>wP2JmAckhbIdmorMSKke|cYfuhd*@G{mHn~J-}HZ(dhcaDp_$!3 zJyj0q0?V$7d+vRH_4JB+KRH>s{L7S{id^}BJrgwl^8fWz=F0!;snV6M|F5T3&+z&G zdh)sQ|9T3!^8b2@x$^&dk}xWI%7RcpJxLgqJ!L^Cpq{cI6i`nRM!qKrqtKItk^6mp zSdBD(QAfb=Uq)Ye->+@D76sx$qcM(x_G{GTpz&fM(&Dm9E>RptM5T2tokPp7r=Ky+Fw z4@9Rm7lghU-cuHY0_rIXLIL&oAlU9H3qk?)lm(&Ze|pMmLuaw_+R$07yf$)qo!WhE ziflIuyDCD&8lESt3> z38cq0NeJk-_m7)%&7J#t&%FAbH5XNeYtUEg-?!@Tzv=IZ%inqFIe)+E*5do?xeZm+ z=a+vKR?)NTIPKF+Bip`VZ|=uC_kAr}T-3h2{z&ismrwNl&(9BD+4hs~e)I5Y*SBcb z=(ErM;L>w0artTdcUG>9dVH+;%2$@NRGg(n3;ykO zHCIO{>h2i)xr}Y~yl=@pOWu6xPpL;iYp-1Rk7s}Uj#tcD)^gL8- zzghO7&u<)=`pD;&Wz+J%UG|R?zx@8rmwaK_$2Y#`^uK+>{$;Ov-~4Tj*M5Cjk{bgTyQP0{B_I-cZGxxn<%{6oXr}l?b_{kN&ko=v644Y$| zh4?udKachN^mSkV+vQJu`ZMo);xqHsKe_g^H#VRDRPD8YG?dvLd-kGiBmO&h<%y5{ z^y@FZwQbp3>;C+jr{A)985eMuT=I=&*%jQ6mpu|L(taBE|7-mGWsK`ph2Ut*=d#b- zvBJaOTySLC1r^KLU)+4b^62lgS^VGs>{-D;-!lt-Kea5^bivlI#eMo%a@$=OU2yRw z7c8<}6*Hz)PMR>YYU<>PHPfrLrmBA`wX;e)YqZn!e$8h^X@7QyCGYki%O@)=%bqD& zu=UJPIrANTpI&>*`t|v>8*aICO}_iCyYlOA*>Kw$FSqUNXwg8d9kXob@9=r!*{?nE zH(&bHzi$5EDUW^Z7vJA&CofFd_Qf$7pDZ7sGxn)7e$403pN$LXfJ-I0_f%3S$wLXA zqqJsY*DbH_%wKZx`M18gdrkgzCCkzlMpQG6H8EE)VM^ts$I_Wa+ZU9&8-{Oyy?QL*Q{3$xbU{SF3fuIh3jv>b(Tb@VqaeukU0+MXWy_wAfj#D7*h$7XoWNAsILUh`0Q)r93wKRM<5 z->m*AYz4Xed7hIUyjXMoq@3{M}*I#^R z=kM&z^vk`hUw`Lo{^OCATa`TC`o`PWcTPP&e5g*=EY*p#I_`Y!U3YffzJASyrMJB{ z?z;M#8(#C`mo2|)#fw{(UgY*&{gM^SuX*`Wzo)J3x%*bFdd|KTEoHl|xncQDSGO*` zWy779EsEZ*x#5PFFYUZ#!<%k@{rH~cH@)<-MN98ld_i;EbUV7W@-6gs}g}Ht0 z?Q1%hc5is&n(BYQ`09(^S)bpKyYH&n^VK=Bs(Td)@*5-QcdmL`cqFmHK9w_U=7w>tVuE%}o;s~&=1lE7?WO2dOsJeRY08wUs_Kwl74}WkKJCqR z)vRIP1i7l3I<>k+cSxqsn0ZgwH(}yL?Iu-KRZZ<-A4&O;_oPXatF~l2DuTz!Q>v!yHO_5MDI8#yHOM;$)4K=DeR61s^VjqS|%0hY6Our%1%c?$f zkEzwK3j65d3Rp@rbC$ z5%#H@sx8@HK#azvvnKS<->{D%xr^?B^8%_00_oE6!Nlt`)9B@jZZYNLHYl^d` zr2m92ST%LJP0GxrM0M<=XG~!j;V7@EIWyWSme8p<_|G4=m422gnuD#hI8H-9oizFK zpqGWBU_$pQWVYE+6GE@7d@=i|c`}}+2AWwr>y=?&R)SD4wdZSsEfpMdP?dK_F?L1?${X8U4WWL@h=!B(DPYcM8^MLpZe z;Sd&17{k;%=*;23%qGzkgg`={JcWGYmiDfP`^U+^YVRqD){GnZxiVy$Pf zv<83SNKhsk3{+RIU>^}jO{)9mpr>rYK&Kj=F>}scL2tq&naw)ojQp8rH7>k}?<3_V zadw$LT}S?$Idkhz;~f=B)NyvO>EAE0j~3_PRfP0@zH3d!q32GjQWi639=3g^XAQG-^4JgQ`qSoUCr4A2`8j1 z8X3PZ*wWk@)igJlHFxgZ8!|ndhK@kHtzzE<4Tx(0*(&R4r!|~@PqfX3VVZ5m%y$M` z+A{RY8MO@!r~gIJ%PP8Z>h#%lr=2nXTkM-$T~j-I&fHT@TcCBp%deQng_7nfDy@dT z`cuz5r>W`6?BYp*C`JZtjm5khITspD6&uql*q5DZIjZGicHQfPzo4TMQ2o`@-jdnU z)Yygw*~~u5%J33S3^QiDFZjb@XgkgUGd~>ssn_cqVd*c;mUhc}(!@^$+c1zY^HJT; zob~xkZ;~?PIHiQMza9LkAmqXBnYD9%9Bo7GWyivdU-^A%;|48h@%!)ew5%Mo(3C07 z?4zb!y4Wr{DPl}z`$8|AF!8dJ*zz6xG(BgZn)QXjRvQ|}HMnSsSvzY*=8yP76>_>= z74=YzhsczUkcSTQtRQNVDN}AUy;4~w-VyY&qcu3yKd;aBg%C}(EAtmxLMrr*DI1+$ zEKkLo*=NIK9E|P_`@(Ulqe*?eX2vNWU|$xlHC@f4(U#uA-ItXZE7`#6q5w^IL` zTmQ9S%Yjv+@oF~1RsBC_da8lR9IoNW{P(a=r~RNgL$x<+&h)rQpuMpRPMvxR`$9Em z)usF=ou27YrC}%9^W#1hPr+n;_z%G>10mz!SbLr5k~T9Ww- zSz9*CvwdL*3&V*9w(FuTIwYA+x*^lE9?9-WE5knKIbq~r`cmBy_r>mr-o)1if95Uh znz%Oe7b=26bG&x3FBFR$DaDkkH;?Nn>=5O|iTp-I(4>*FX;bwdmGpv;Td5(4z#QKv*Wg5e9_TSGyN;s zJ{)pp(TGz!dz5_)Y$_T<6UWxi*vC*D<}NziX4KYB$QF}AQPrBE9;Vm=Qz#AbR8~%l z`+{PKv2u#OiV*5AYox4lP7nL4YHM}YnX8U_>hqr0@WQZ9?GW5g3A2HhurJgVrBy>t zn=x}G`!tYicxAf4P&a$s|6=c5oFvQg`o4R+s`6E@W?5jDh22>mf@wrech5}kj9|r7 zS3jn9yQ^!us%HjP!%S9YR#k3gW#(jNy+$@_3^F1@5sbnwIUFbk*+_O+NP~oVO4b+| zjF7M+le&?p!$$qdD zw2VJa?acnP*pGh2$Nz%rkO5li@|S${%Rly&U-ii!jd#P)vU0Gr^mpA(3(%hdUH*Bu z6RhG70-OJ3*8x%{6XLg8{q@u?Ekp=Y`WLE8sN&6IwEx|8sW#!mZ>2i5PJg+6i-cAN zumR6MzcbZNc*>0C{;$+78bX-W>Y~oY3g@JCGF@%y;+G_cV4_u1@W+cfBbN54RMHt* z%`Udln^O8?(JpPT)?Kcws7@l;D-}nmRG*JJfnH$Ieq8+AxW6%6%lFf}CM!2wT&zc( zJ9IsB*-P&xF!{3YI?pjsi+{g3u2X&dr8)sns4rQ#a1vu^%3@#|x%S;@4DXLx0)+2# zJ1^HT7X1Hwv@@82y8-MgZ_$QFLYb}RG$A+*Ma(;N8s;2OLbFddEwXHA6q`z z^wPhG{#2(w_wc_K?+U~Q&xH%Wnd$^{Z1_s5a-JOx9V7#96r$QN52g6{}MT; za|u)d&n&z4iR7>rBj*qPRH=MzQHOh|K(|YNm+qOH_E#J7jOv)GS`~lKr+57%daT}v zc7e149*pv()SsCmxF`Jh_Q|_mo*4BT(N42VQz?~pqunFG-_8_&J zdig^4(RZ~mvjf+STlw`-wDZPAe-|zsKl-lG>n+udU)9ZpneR&d34CG`5&}QqI<1n$ zi|t||{9v@B`?0-W{$cm0H&H#!UKSR9H1(&>^#{Vt{ef$Yra^b>1cYkR_232GIExhQs$ zyoXwro0WH5msZ9)Bv!i7p9YbggJiBP{At$-2>bU2wff>GD|NE}f^ore`nf;}4T>?ET z{N>_Z&DtxToBO9xmtY@le^qtv&Ru$ZG|Gw&&g7gQuGUQtLk<^ZshPzh$<54FnAA{??sFTv_I?0s3?>Ytx&zG63 zzN6uFu&>2Jej~kW>(>hJ^2Og)of?`!OJ2?|{Gr>8nX_R(>)@{Draw`~!oo$@L2#06 z=I*uHCn5+Oh!Ab(KCL>N&dh{{RZM}WTqk7l$XagYGtr+wAT%s5=v@YgBoJg|qJ8>Ifyf{lfzZWqL$ zg-FbQuiLRj6$}boeb^vh@t>r3V{z7}YCr5c^{qcIt@K|;JFh};5On_LxK6EN(LYht zOesk$`d zn0B@LyKWczlQ6LI2dPf>C))^jsnzsdQokLnz|{*y0tnLyB(GFH8g-_b;-4=cw+o&j zW#W^mF4_bPzxI*3*u1YFzbhcbf}fn!iSj+|%c?5|OxG*XPI4zn_;UHVsdv$8ybgU6 zMt>4D#ow*epTR|?2DzL1OIr)#S8K1hj=7o_i*Cxb9lt9rQHU$}&CegkyTuF*@V@GJ z(^!#UeH!gFn*tFk52HV$8Qs};^8>1zTac*yI4;poe)5x_`t-ZepMeadwCKuD`(3Lk zL9EE%pI04!37kgDUv!RGAAVt`us!J zU6`%NFZ?Aa^LU{?szjmK)}RHz|4G%ciMa0|H(at`^{G#Mtd?~j0SAVMRcjX+j zxA-G+eUDVlIeR?T?lr?s=(b zXKIZOuc*#jqur4SwzPcR?VvVE0elBSt}WHk&~A0m4+Nb5PSm}POM&t zT3Z~rKeeGt6Kl2hZ$-QDLL?2oH}yxQCoC=f2U49z=Z}ud<^M4Hn{X0vN#9^fe;P-B zvp?iI7cD%r>;FZ1SC9g2rA*cSRMUFX@P_2v0*N4vB^arvD|yJGIyV|6CE|1-U-sj*c5{Xaz=+mt8UUvz?!+HIzKlnH9mjK)VoFVZ;d*=Q#71ib)9Eh zG`p}Cb*ZcoW@c^`?|QcJ&-&!uV!7F4b@XxJ!ku`RJHb8tMz5$&T^sI7)>v3*O~wd^ z3;5Sh?+WRQKmBIT)SuRK;doM4tRp{})Qy|#d$|6KjT9b~AF3|OW`Jz@!k>@2abFTb zf1!9cnoGcy{-da4Z%jr|M5XpaMSrm;!os<~GISA@mAZGYKN1a%bmM=A;rth2sgxC=8 zQl;|ysm_Q;L$SbFC)LuOr>pMB)^eR#hCCqhUw`SBeOh%=jzm-}Mhx$g>UfiJQ*EuL z@3lSt=Q_es>~{_4*e4+N*HgO$1$Xk1*ri|pJ#c7F`veG{zb7d|NJ1U+F9d#vA{N}v*=NklFDE7f_+MADJK zFsFoUFD+3t!gSe5+QxCuL)iZnMdG%)K>3mS^F+K`fIyv={@A3QkV}+UT2Y<&&E&82 zjO!Bcg+SoVpPalKlVYx`sa*in=11*LYUib}W^M4c>X->*Qj*gcd(iEKq@IaJFLhiO zZx$qX9PI)MCWlJpyIg0;C<#(#&;0in?X==#8o#Z&C~GF<4J+ z0tTW?C5c@8J8q{0f-C zQRluG$UE?VxQ_W~M*L}9eH&bVQg7pLcJ^PpPOl26!YXXPFN=1mvHLAe>QwgYHPs19 zlfRio)u}TLD257LK0aqIKSo5W61*&r9QXWe*AeDwS`Kx|nnTVea*DJI3}}W@ekJOZ~+1W@h(Q7fa_& zVOpuw^}FhhJr-Ru#&LRAdm%~{GyXI2Zm|Tf=R%8sV}Nd`aU_Ei)han zn4kDQ*MW?3F3Af@m7`ALse4EN^7_Zk7+SyqV=ES$>%3_p`j8=WAI$%<`QqKh5(ySw710{XD;(^8;4-TiJ!)Fu(BX zW&XdCT{t|-E*z+QHNSALpI^Au?e}?~`%`S{i^9$?U{K6`2)`!`JK`*~>ga7Xh zc*gsqtJ#Ides*EEn_Z|=uG7yQp1;`6E_8a?g?+AVuETzIq2I_aY_3sHyAJPl=wp}q z9sVEDce_ddJR9(A?=t^Wf5`v4s^7~lobdl(#Q&5}xSnu5(7TK=qCBLJF8}uq7>jkR z-=Oak+A+=!|Bp_n+ho3njNRedpl|9Myz>^<0p%uhZL^kxI&0jb9evg5yA5n^1B=^d zte^WEZ_*dzw~ra~l=px^leWjt@c%XDK!5dJp6vhw*0g_>{~2?icLvlA2EdelJD2(Y zY5spEn`yQO`OMA1Ae%X8on#yH2Z{&wO6*_I9b;Iojp5)+nF3t$$Vr^s?UN?&Ug-hpkaIvwO%t{dzN-IRW}J z{LUA>^!TTD!~=fIgAVl1y-qWq*=U{6c$d4?onbb!=T4gxZ`99dceInwJlI~N7d;*F z`%p8jQ+nz28gy{PxSQ&<-)54lT}Eqld9YpA+FEaBGXwRvadeo?bX&}z-gFbXzklT# z|2%z_6|u!LJDTpGenNlj)XeZ`hfdi;24a4P^;7=QgMJU-435-9-S;(kZ|93lyrFgU zc%v@pjyl82%psdSGi)<%qt1hU{c|Xg3`YmqOuI>^LxGrI=F)VT_wT<`GUh&OI_d2mkF{^ZYdjd99kRgN>&*U?2hA3PH}&-Jh;a{^ zVAw&6CaWEmusVDOkikO%`*#&HUC5nc9sufW5tT1H2#ZGt-_RdBl(#Ex@_aquonKY{h-e<-VG{+M|O_ zxZKRz+8P5v=QG_p?Kbq!y>2U;J*2XS@7FUN+ll4@cQ7zk5Ubc9+G?4iDPgH|wK1_l-JZ z-R%u%*>ANDvRTOX^~&sS2NIqgo^>1aBTAX=4_e2W>luZf^I2$in9pAC^*Z@1TieKI z*Xf_9BIel#?JjRYw9FYs!N0qM7H{2h&(QW7HHV{@{O=b1caCWDR%_7nSoF5V+?YfA zEy0AJr_nt1@0M%Y-Q6DZ>F_CdFOn$LcF3rwE<4#OYlc~)I!+Z8*ZH&Bf^ zXkQN6pxCT_o{yfrBR4x>!0hL#uX?{&SL+b?_v;!RZpzyCJ?Bxq-C@_*ef_JaE#77C zMlD8PZ8rIL&12nfHIB7*$d`ZLyqe86M0K-!U_v%K;HgH@QxKN^+XGmac0vD&JI@Xp z#{)3o2yCCN@4yjF zQx_&fpVS=Ihu*dB8Eb6RS!w@hxDT}J2fWqwyInEVStJ2-<6q_m!+lddYj*~9&Af4h zRGIB{n)KOi>zqfh=Fc7%0b|f%s+?70=-JyMzQM$Z)hhNJAZU_XhsEvBS zp6cIy`WJ4CKOhr0_MQ8!mgZLXF1&fv2H%cb;!*n& z4zsOBlR0enknOYU%}%S5){dZ~4L=ITSnR(0Y0-YSrF-THd|ICED-atqgl7U()A(WU z7+h+!MN@t6=2m-;Roq;=@nSyvQoBpb_2JqoaJawDO4q!i{q`QXR6iOCqofPXw*7Ap zp2kePiUGW+Vw`8+hnn*$f587p__lBUKe(=4Xto>791Z}+BHdmQSLs3ThsJPtEHz#4 zHF8dC2lFC zny8YlE#v0ITUyn*yGIC`d`|L8fMnJ*X%BB~ZR;`|xpV{zm|NZT2GCjKfL(d>NHb&f zt?hgcVW`sGUKchrx7rbg&Y|{VWzl2<$;9X_gwh;utL1&(pqDjhj_E>Yf+tfSooR2` zF)g}>T%OzNv;`0$4N$@)*zU2nr=vj~V4xmZn_+iOUY%^@^3dDXGSuS@BnNp~~|9Bel$bFXUzz^Y1cu!}_@1oNdO zSObFANF(gF`LWx;ayr`uQ9wQfc#mi8F3eENlwt-Q0rbk;-W`!LkT7QLH{ew?LBNB2 zlJ@EHW>o5O((3a9z(T;!iP2w)%hhad?-tsaN@5HmAdKI+Lm?4RV(GlVOb)c#GjVOE zB=HDw?ZffrkdpO7O6@%_MR=XfN$_3PYTG?#2GQ>^Yg0chJ1sQOx!tA&CDW}VY37iK z!m}-HF-R1o*6uyE3kplq;c`d6%;+Wvz_|_U8*`8{%F`S|_pr{hvmsXaTq8=iSS~GZ zNp}^EiN(*gj^TVl)B2D;VPZnXVf|Q3IX%Pz5Sp?9=3wX`bLdc*9-=sOL3-(pSUa`_ z=KNg!815l@SoN%f$sJoR&Ph9G{|@UT91oCak6$xpwQs7#5nvpOY-)a&Wez(LaCV*7^-D ztwS6bbJ$tBw1@Y%R_U`fT7Q|+@f~qtG%Ez&+&;h0EVN1E1>p^J;e`fbj4h7wR0k3r z3SpQ_vxn5Te0bWIr@>w6gsk?~0RB{+!@fNhtL(MV9mPgb$5=|&m*Lf`&xi~gqSVGd zDh(~;g$Lb3RUThEx^m5!989&2-a04z;&L)TF=5A=zU;QUqFaqlW!IV@tgFAV0eDXP zue?Gf>IP`@x^;d4g4wN>vH<3~7L;>`myz)r;Sd)D&^0Q`k{X7Wu>XmLOh0k=4j)L> z_1)hq`Vu80In$hAhRbNTzLss7&XuHm5L4CHyboUtY3|Zn7R2I7WU6 zCWDfdmFkU=7-vZ;h^}n4@b%jQPI}%Dqbnhnw{}jml11>#S&42y{=6jhwLW_17X!W@ z{#snAmsIAt<*}|>_^GbEN!fY`a75Bl|6aWqk9ooL^U{6x96wp9Zl=S6u5TxNdt;^& z_?PNGNZ0;r(e8EU64;5u=aLk9_ls)Wq^U_+nqFAu? zdFgIzw2zuoTB~>AA#c1L_>+|oo09$BPvJqgCxa#{iJ7+#vr-d169S`t)ZRT4{zmx_ zg$JrCHG5l_DQ_oyeBpnHem_jtClY?Xovt6I>%)JN@HbuG_(GH)rt3TD`d+$zkgnSI zfy9jIwK00{-<>y{veMzvaCDu0Vtt^#_G=_xsjfvCmPuJE($fhJZ>qloNSuA`qwUB3 z;##3R*b|E_9U5-z>zUZd+xs!z3vb5ty>z|E4_}y{r0eD|%1?A_X@2lGInYX&!QIx$ zho4<^{rle**H}-pRW-i12Ah8~@E|K)3e~)1jq5qekPR%=S3IJ=ijFUR;WeIUKZo^Q z_6tMiP0WN+SKb)*31twTnw9n%*0xHm(6dT|UX=HGQmUX2{zcxD`VGBUDIq4s>`Q1N zUE#-GuXU<^+IVCTs^GniF7QJp0ps;i_ZhDn_DKm72xpzzTeB*)hgOP8okWj~{Ub?( z(ovszFkdTY8=|u>vZESzck}>$dSp5WFIxlfkmq6mv~LgBq@$Ji>FG8adkMAO&#l2- zfj@_F0llEt^L*599Ek2w9dGM;XvI(Kl>*2$v_smV2{5n5S;O$dLm)X-IlzF|O?{Mmp0P^tBK02eELB1mjg@vNPkrSyXc6m;d`GC`gw;RpA#N#-c9tBt{`fj?upRP}A#q)l;zL&0- zx1;{ObbaALl;2L*%P&Xy;wy1|CtctE8$nl+D~Ih4{7A0Cb@-ZG9Vkzw>(;2@_ZWxb z<1eKv#Jo4!e~%w@GkiW9*SrK0v``t~a}6Yr_9L!?6LmxxmjVt6;jJ8MrB%y8#&6hq z)K%uAuCgCXn@LVtt{uPqn=#+)XXEcg@u*`~jeu2&`Jo0dCbFQWfCPN|KpytEAiYmMH1_M+GO@LF7-_)J_sOxJgw zi}HKl5?ADL3vUSh*xX0H+S|ZYda~p#u6Fbtx1eiDU1E!+Vg1DL=yf}1FGAO- zW74imZCppv4|WE2Oz@tZM&ch$qAKD~LoAADhh@M%3R8Ih=fcV+%Rd$6chdFZPe=LPbbaGzqI~gh z$MwT>efW2xy!l_p^@DVM zep`V-^p?H7*5Km56aBoKu5Wx#l;7Bh>*afK1>fbLcpW-fyMZPLoz;hm*p=>@fx!pU z^_Dyk!}mXtdHg57Dz5!>4Sf^0nf1=6!P}c%c^TgMx@h;p*Qe{B3VMGdUElwPsDGHs z!uPhkS;$E^BzSfoCsnKY-qX=ed{q5z{$BJW{DYoF7kjN$0*WP!G4oyUYFGTb-WnP% z1b!^B=IRaeXRt|Elwk9BxK8~LI=Rt1 zVjj|aRYuO#8}#45|4GB=-E@87{^22&u-}u3=bbI90VIP$hes8`f`Ob3S zcUw54FeppIo&>t+Q1r&Qy-w#A^CFZQauqou{2mtVaVB9fBmYB=LN(iK$SpaP*Eg#r zxtCjmHx3g1bmIENVO;y^`mmeodvSd~U9&f%eEC0%>xb$3@Q0(k`B&mvO?+G9qBe~B zzxPqE*Y?ZPSKy~8Z`Yn?-UwaF?_d2$dVb$@{@zDEmEhAn>u%R~KKO$uzw?K2ec_MN z^{0z=t3>(i`NMRr!oP$cZ+|-4z3>fjW&a0f{ZSA2vc2ub#h;4# zzMrn|e)daKzt}g(W#&H|;zfXuT5pm5?BEl6v-syb-x}jS@!7b(pRU>MDCZ?&?zh+O ze(+B4=gU77*N5qv?I!;HAlhHN{c(@~;Wq?-Fdw{!ivhc2WcOo z!@pOIhke`YLKoQFa?`^v!w$Oc_qCp9*1PDR`}NMT^gN>P^#8(-$9V7jy|}*LkL$xX z<9c}z*UfZ&J6)gnZ&Ukpedp(*{9d|diO)TeuGxPV&)*ov^__HmFJ1eisDC%Ld-w~f zoW_H0yLjvdZ-4sB6W*Xkq5mg6;zoZj>Y*cC%pvDOzijr_&3qR`4)ax>8}9o2!8Ze z1E1ba*Q^Y`mAAKAhOq4REfd8BDTi74(k?c~H$It{<$c3Q&&w7jST$v=7UIs`yliy} zZ+7_*e+9UgmBZmO;Po#&XVa%FivxEGk2=GaZ-$s8vRcZhc3yTFWf?`OnU@t(mxH9d z&8&PQj|ewedDypag3)O!BGKfBM`$Tb^>DP(?`LIlP@w3YtlaEHh@s4r%cq*N0vC$c zm+kr>6kE3M!#C=<(`6OCrWTu**?Zm4Pn&1Ey}T^6B1#cA3Q)w_*=tsTRzm=zELRkv z9^Ie>ZSxlXSqujT6?g|Y<;%k(G*vvY49*P6TqKFve#y$-mp|qi1af2*Dyz>aR zf%WF{`qsTy6#Fad-zNDzWrgNR^r7e0h~Nc#fsS-a7g<@xk~ePiV9(LjvWUj3xZSYD zotIG@oYGJx2W1;km6fIF%Jwcxa``5V*yr-{`VhSnbE@1YW-6Z}H+?H^y4hCvzRWHi z3w6urx$;3!)9&g9&}qSN-SYi#4>82xOlcl$r9vz2u-;^n^em_vm^()@4S@ze1PfQ) z2rB|C9Re)^ukl{1KyRS&X3S8)O#9AFyv2=~r?$X}1iEo9fqF*GxXR=MKFlyUQC4P; z?~^l7Efb9mKoP6qUX=2MD}m~GLD;!t_QGyUrg6X3@5tj`CT|FU3CD107<<_!`y2yX z)Q^@d>$ZF;-lqm@)Oct=0}h6=Z}2e0jS!pNIRYBKap1?~sYH>oC|K%AS!SyuJ&??C z!*eEJAxRgzET?dpVCF6P8p_0inkbc81)(Oa*|@=cF1Lp_9dIfu*LhvjQ zi-?+8<#+wSH)8CT-|8Liv=K&SMQzqdo)Qef7SXA!IB|d`fuR--5%@;BQHU9>4ku&}lqiW=r8R4GMgF zCYd*FPBJ4g~qrm>d#T* z^&KYA5hyYcltJco1^e3T0Tt%c=nZ< z#BJE7#sSJMsKutJA1nW&CIF75rN&Ipz$AAtI1+qvi0Wx3Gw6t!U+)|!Jwis_3+koJmla-unJ^8kUo6+y9odUUb$s<= zgK!P=1YrVs7iF^Ah=z*ZtppV%?l;Sm|9x}2qq7$LQWghaVrCHBLcij z;$zvNG2yC$w?+UUFF+yw3KarJ?3$dIWq7fz#4`NAnQsCVejy%VcZsKoWbWYdN5WnL zb}OV2d(K8tU?_P}En`8&Uclm3K}ANlxy78r5273RBrnk600-wPGenS~)v!|UAV!#P z9^BxDp@c0KKxi(!C2QB1Ac!F(V;a;XtqEPMytR)pit$-~r9M<@jS)N6ej5-P2sbvi zZs6JwN?=)vnj|A+B0$}z0R*D;aib`1zz}VRkSd#7K)x$wV#uIqt!gYJ(S~zFLa<|C zMp46gpMFS2H7f|nla&;LejmvK{Xj%G;Jl}#ho#YvcGwp{7dH8o;s)6iHBSQ;#XgCe zN|RaL+PIRJ6|j?s2=eR=KvCdK_Lo^4!o~Rn+HTa6ecl>I>O>i?=;T8YHKg;5Ue^r* zk=P}cgSXPQGz&Z%hJaR%mKB&mvAIND?N2CcE4I$y&m9RF?H4-bp7ma^;FaD0#9kk+ zJ4Hv@pb~-1K!7J>VTOfDqu8giSz!+AIOM?uYw%JFn?1+Xg&U*TQ!+(DaUJgCK3^B< zHOOFUuu{`D$X40wt)NKbGN|03HGHC#Rj`Bya?fQIM|0t>Sw&fNdRnPC1w`Vtf?6Iq z0~IkfnHIV4Zec{k-F?iHyt0i*Y(TuT&X@c?ejqrnfPPrO#DObGpLkAp&%I`iq5{k4 zZa~9X#c4+%LtX*%#B!UO=`QF(8d60Xwv-6!j#!!GOT|ExDqnF15;_7iRd`bdutsY+ z%c)`|9xXkq@Q+Q6tm4!cW?G3lrCLry1(mpZBsD@2r@z(=9SyMEn`bw z*=o_V5|JvGdcze7IcC6}c}ayN04pY0MN}M6djN;9DNu1F;wBk%6>*$c+XGZ>=Ok9v znB`tiJ1CvY!M=(hAxz9G*!Ge2RY`g&GkB2N83s%hrB2Y`4y~>d1|z0bMXH1q>55b= zYugnG9;w^f%QLIhz=i&VUhHMPYxw0A)M0s|PkrYUCBwiFO3=VRm{mYgp$3qY(w{~n zas#QdB4qus9)o{lEKx?Pu#7oN`vmp414-+#4&P}VjUycPDUuk#KnCvw&~yH=Ac%1meN8W*l` zgK!0yjV2dZCpihe9b}cmI(aB8J!UoZonZGS(~9|N@4#KO?SO83nA{5V*B-$w=tXyS zb*aBlrP5ZdwWs>fDnqOG;#M5^Fz&+XO}mG5m9&SU>rQK zf&qadg!*eNj(t0{SwZliGTKjSE-)pYX`0>zVy`eqR!P#8wd$C?VW6k6j-+z$RVqz< z=VTDy6?P6H3<}=>$cab5;A8gfJ;~C`g0t_gz-`C4J79xBOV)vOl|s$_9+EB9MVi2Y zM|homr#-*jL*b}MgloN(N~3?oE64?7#1_1gG0ujmZt@iBkv_v7LEQ)ojV22wV6U|@ z<=f%{cAdb1#g~N7^tXRBY8p3&KtlTqjw2K7ov1jS=yfdf%HSkIvT(RQzUGx%H<&3? zful(V)KBcPf%a2>dBqh{m@1rDLe`UUYW^2kr%trNvFDs6pV zf=zpx3O(cINn8p-@lyOWtr9n_wg_0gNDuc2Qqc>1WDBGa2euqyjA3g}dYt-0u~*WJ zG_^P**($+>C`B0Ksa&RB4{ysDT1L`hj##V9IEtL6T8q$%Nfg z;V!hn$|Z*Gd0yVQ#(X` zx3L9Ppen<2His=Fhu3k_dOeuk>D*LmIeIMaKI>AOh)bUmKz_V+rZA#0C+wMu%z%Ip z7C^dl%H`(P7Ftzr6dBzW2q$vhHG>TRVoJs+#;F6vhs6WbWX{;ai5chwo@ik>@|`q_ z(CgT(gn!&U;HBr-nWhXoX}M^TiE1MpBd$lFzfH0kv&VY5>1^i;o9sIzgc3@ToU=a5 z^eew6jr*1MXx~PS;MWN~5t1M>fl)5JijW|f*gf#s59Gp?=j~`@9ZcEUh(y|sY!cSd zKf|~ebaf{WUPbP#&`@Y&Izr~PFjUOB$PCUE<#r5EIC}8{M?0wJPEY znNyX$ABb0#LG}u%y{)WzL{4u6U#hTNEDdEVSK-`oo4E&Lu2%6(tRk<1#BfvuG`F^n zWTI7LvlMTs1|Ma<=~na2tMHfx2T4HDDyMPcO$~u)kTIxQ{My#Ns^Cm92LdygVwAFK zvek&UR5A3T2Tw=r3wBcF=tlUcD%Eb!P$QI5^$Kq4JRlnY{}Acq)!V(DNZs`D`%&W-jdXtmN02LwCZLDB`v0^6|`}Wcoo-Q_E6?60Kut)Yzbq&TC zF1#v)fw7MYZ;%x5-C-N>h2R2U;JwudCUUDFDO8}fp|T=1hk&npKLsYL&O)=qXKhzw zU?IV?liRD$UNcLn5+3NF7h&O46-14&MOCb1&BtDI|hP z3bn{gl>L)cTVs>4Dz#J$Q*cnjxa}Y8Nq(SLdgn)+q~q2nTkP8x_y!}Z4`APVm)C^K zJ~9(S$Y2bv4%9JA1J}itaDKr5tGY##L1BP4IVYm{a=P`=+A3CB%bD-qtU(yD>`<$P z=-}u{6&&a_%&e=J1l`n@tiwGsEpnh`^;q_$#WvTBl6HNpt36Y=*-1K zpo+>;*?$Z2m)C`Q7B3^LwF6!zIA}9WEku59;!6_Ya0|u<_+*+wPU?)E*K5hh!+tDo zG0&@jx84V9Sde-kbBsfqRdhn>7iy4h^;){!QLXyCIf)75Y^NP+v zgIw$|J`Cus+d8f{jwCN3hC}$Wl-sJe<(8~_FzTKis_Map;}#ZU^>9s0%C~!syYMq^ zkr$-(bi{a4Tic4z8?|w(rLlPAIZ(fvoIx?Otr1^Tpw_k_{^aU9MghqO;?)qO8?$el z_2%nX&aYo~?K9pt_?cv%JQsQO?&NlDa(k=^1;Hz64;s4!!1 zwEB7V=4k^@i*j&T{PP7j60s&Klsh8bwg!Wg-&2*s-|~r2aA(hpU|UcENSv3W+5+3J zrNtv2>gjgVxGvKv31M?NN2aQ1MG>w-21scF0{eMYhGXcAYzrPw8 za0gJ5Ef`?pa`Y~05jnL+wpNh{krN0=mn?2YgislQ_ z)(4%?%&RGUhLg;6aZ?6rUcI4sq;guA`kgJ*7=n?g2ex7rW2?F|o!!n6i-0i%w!@fu zEkjDnD{o8NK8iyQr;Y9R9<}%&PoV^j|s4#Zj+<#2^kGlY~(;S~6tBx$br2L>l_pSbG&~i{ldT z-|^beDT?W!`!g|x9DuL#1Hmgci(Op_RL1Cx?2}X*P&X&ihFGOS%9CjeB=wqzSvq2h zBxq*h&Ptl`7GjcPM2Eh<-P7&oaSg9_65n}EeGqd*=^c#Pf_N&_x zRH?SOb#0TgUEps6klZM4p#Ui9?xxZ{AqJ6^Z#`d0SxH{(HVHsb16O2g*bqn%_#E)| z%-S4wOE!a|h6)QYv%7+rPe8NXSkrbHsk6queKKN$r)Ocph*(|Bs`9=0wmQKQH1qaw zl(ul~{4<8BgO|d_20V)obNWH9X6&NK{=8(`Xx}OB=TW7M+ zr4_@2My)kGO)wB*E7;2*mmN65s7jS2;kC#IW)(JpthJ$unKo#PM>pMKS+tsZjn2ZP z7agZt&Px`zBI#874AL~+N>A+QQnQta#ukZt~S=_P&j$g1@vicSE6!DtNVVzg;d3J1&nsCB2WQSV&cvML|PwbU_ zV_Re>Bs+i}3cE$$+gaUOTVKzsFY&2~13A>K7cP4AQ#dk=GRS?C!uh3mqL&g?a zRyEyldi04u{co%O>B3N|#lBhk)*v#XvM+Amy_46>+`~|;J?Mg^5o=BNwv%fP_nEe{ zn)D;k@aC|g{+~-mW>zy2#KzTpB0Kbon)G}`GuCWY8|nN+s`q_rU);kxYGi~c?y8SH z)QG{=_g3Pb_&^;uiwt|}Ns&zN%49&hRW!9mwO+=C@IA_SOYBkgh##5LS&eUyp@AU` zMwD@Kb7s$+WKPlvy)pR(K?ZXkk+GQu0ZKzMM-p-Y* z7W5=HYgR*nZXD2tj|zv(`bFKb&$ z_DJ_u7QlaNnQo*${U-hf?i;bf=>N4M&RvsdA~|DfhrQ-ehZ8Tt!|+XrErau&BO(dJ zkoKV+=UCeNP`fmUG|v_2Vtju%b>Giw!5!@>ui5aneW`vBeFt_M)D%R;xREzFwNb_x zYtcnIIB=Gqf(RH02ybb36_AHt$Gw~Iu3;YXBKisMFSzFzDb2u19}Im_aqHS4wykV$ zWPOboKk0hP#jMnnBZWSy&*KFxJ5`1L&xRa^wuhwV=yDqC(0&gQuGWP!s2{W}kINvU z;I%mTiI#2LkfJN&N4yAq0-V8Z0B>=kCrblv;4KCL*>hTZ3r`CrV2g*0sJ;bZb9OUBwaizP6pTi?>>q}lWzK2CLz=T$DT8RpoL)nqP1Q5yJ)Vp!y6Mn^!?>( zCndd`_?Y_-Cj)W;+H^nS`8v$GCKq-MjeFF)TmK>|PP%V{WQH4z7+?(rw?j^S1}{xU zGat~(Yi3)ay@SJqwawJ3)>{hxc1eP2MHzDNDA|F206w#jLN3U)5$(mR5pfX@uW1=t z@EUxq*>fwYq{u1RQ7Q{Q5P0{fr8c50#SsO6ubuU(UNBE_?HZs0j9N=p+p#TRhl2PC z-!{Oip&sYL!1I9?G|$QX5JMl+h&HNk4TA2PJq!yEAo?P!ljt4}BWSbR3}*rIOWp(n z%yKH_91Cv)&#@9}j(Dd%Lg#F>B63{|5fr)>@{z!c{3n6Wl5C+bT}}E@E_~yRgR&6R z|Ff41ohJ^jJ%yh@d+p;BmYet1wl={Dc(vvxOFQW>Ffro??L*hFyQzlf_hlHG%uwkh z>0Z2DCn)Grv@B0y(zW1(ySujfU7sQoz$da2jn9%7$;hgWO>z1O%U|;m%-N-gF2aih zW3g8DYlvWi4#i{HYi#$9V%)WGHpE)s&azuEl|*ZZFTA;YH3wY|uZhk&ipR32_bmg^ z4SZlPcz(mipCSdo2PCTx6bVfDZv|XD=M_GHqmux_VT|*>j0^TOO!o~wA|w4|qX(98 ziYDrLM&1nV!Jk`F5Mxi7Jn6rw|LvadMF&{Yhxi79?(8y1{S2{+Ov`+Q1a>6-%Ht+n z^1aZ9wez`!F6lk@NSm%<8bNFdv*$Htj5{G{Jnn6)r+Q@+h*!rGJQ3r2BKhNcV);o2 z4*Kf2PxVvOM?dW1y>2V^R89zRPEI{!mf~CT#l>DA_1V<84@-}A$se(DUrvb{3@Vu( z!V?0wk+@v*r>Kl4g`O$fjP>esU}?X0$Rp~m?E?<_)CXKwoZC>7^CEo&#`oJj=|Xu; zj*6|s>%`Jhj(&|upDYsN9yv^BDIQ(hA~7bcUigv^jXpIPqCX7B^`b$LEzVttUdHwg z_sP@D+!Rn+QGSAXa!N5U-O0ae{n_+ux&>DClhnq%5sQ;>HtGp+G`oDe8e?<1PXc_^OkY@!{)QJ-{)&qEBa?P z9WYgf0&}yC`wRz5^BS0A2NyB@8fwaFWL>k~SB0wfZ`3xnCinXCW;7CRpt-kJ&=Qfq zijAA!7IIcc7sMY9EK`J6!1*(ZTWJ0Q^HYsML-**_DT`hA-4Qy9z61t7OzNe^wQitw zA2O~!=(qyM<{lKm4jU;Sq$V#!;4Sm9|3)@D_qIXeUJ>G=4LEH7i){q=Dboi2Fx3W_ zf2QE8*k$8;g%fwx8S}iaZ#X1dFWeS|?2y?cK2yLv>>Bcy#`Oswn!|dvwyDGGGlL$;G%A3!78xxS`FMSa`nM7&@b_XfO_&m%~RtbGb$QlRB_0q^8VbO1kyC+bap3wp`tC*Fx6>rwTJRQbWoh4mX1 zwBA4QHSOlt(6toHWNiW+EZd<;HJjHaTs>?)h3A0|A_W23JI&@pF$yW2o}f=ddh>(O zx9wjjo+x1}pT9#k!@WUh`134zUN}1I;ZJFt8_2!1cyK zN{g5$*JO<_fjAIL0KW^vd|n1V3Aa8MIA2)uSFS#n&HF=`1w@nGA=Z=43*m*%WD`(8 zol%7gQao@ajSjJa=lvZa6rm%|c!Cs6XOEND&CAn;D}q}<&8VD=P8-4ypjcPO$pZ5! z$6VG@HXq+0B3HXVnk|l5I`%)$Ng&0mZH3R*8_rb{(rCABsz+f**&|jy*%EtPTTNM> zGy?(p7c%*i-c|P|i^s+x{9%7P(Zv%23 z0cTTB`Qk$V2_De(RAI_E7IZ?m;a!=xwTq~c2|6F)^ssm5WlT%*LSgy*?X^vPhEimr z!uB}AKEK!LaZq%KX2_P1Po#*?C{}DN(_qOuhH#41g{Ay z+g1fs!DZZaKF**pc*@Z>`5-9Xl$_C2gnODfF=dRIRq%T0-)5aLzW!vfGGuw zb|Iff+C^+lbCVfIMxaC6^UlgQzXuP*gyd1M_~3IwS0D7*E8NKQ4VifG9esc-W{?!v z`68qa?=_B!Bcr*VFsPGPlPnBfYn4Dt@nsryFM{H!lVTY3fJK9`T zh+S}VtUY$wSe2|>INzS5=Hb=QM58ml$rNx~-<3Zpo8Ko#6#&a4k6=*>cma6iJTC60 zv|a-dWJ&u|)5S@g#A7YeKIB5FB{!7~6UNzF8=NTt4DHZv$KL7d(rgOaEvk z(mF#jtp>x>dQ!}Wpw;}AomgNd+BK+*4S-ULS;%^9F8cGY60Q-ICyhIwSHO`NzRtu^ zm~M7+?Zx=CnCA8h`oR{7U|6dOk+7-J3g8EhN5$7%=f`F;AIh3PhNaT?#4cO|>GeGrfcLt<>{*#7)DoPrg4M z)4}Y}5npjWnfNg`_lBBhJa8dC94YY1zUCK1_!=wvjtXlVt9Ne}1r~u>4VA>R&F%Y> z0zv*sK>_{LTk62$h}yL!h8p}6Hp{F9!lID2QTDTDeeJSsAEELFmP<0E;Ri<}QDOij zLE3IE@C`klV--D;8s{(V3&~deGZ5^yz9vQTiYPd|yW;r3&8GOhj=qE-1zqwXF*_yb z-}KkIz<4NEWsxB7Og~xohbAMPU|z=gb}teNPyj))6xu_YDW#w{&^Z-t$vDnDkG}Z1 zdZV|~mUQNcL(o8E6%oHI+*Mz zvHC0*1%`Wi1;q^aH0!w9g{-x2WXQZnO1idr5>#H z5)u#6mdI@8a7ZY@Vh%52!5tc1Y=jd^liUsqK3N5*6mb6C-o~vplv873q)xzHq?@}L zFaQtnCSW%snII-;k!Y(0WnZTSlq1m17RK73v^h#Lvr?kKLSYFzERssrWuD9z!WKqt zAQeAbAgx`oAk$a2umd?jcX8ltAss&t6KlcmGabu>1xyOXJB-xXg7vk+%33H6EpM^) z`GWW(=9$ctp=W*hY6`~8a>S?Ma}N{%oYOV<7XJ4F&u~ol%Baf4ofdwst-=v$0-P2lI*=E);sdRN8z2a*;16?QLP(y;tZYFRJ6d1gfn(;$F=K{-OF{u*=s}>=WV~lr%NFOlZfgzvlEDW0 zwt}S#VnRoX!{F-97El-aGQ4bN#(478;DgB)#;NBR^OPFe`E+!cH6Ub2^ptg`^(Dp5 z^oN?M1Vm6CtC-Oh6i!yt!X#{dZ;KTbL0?Q*AR)xT|L5eng?{5$9-=UF!JbAWkmImr z$BaC&l5D}d$w5C6p`auSI_Apgd{oG|3cOi6geP+eHR2>&P{^df60L8|-zt`@&KGWz z#aXPDz1@wE5XdT>lyGO#Qv0u10^ItUM@rZ71sh9B@$<8((lurs_M(amTGzG89PI7A#9ML?a{Q_RK<7>Tt-ndld3R~tZx?2QX=*O zTTbR7-hQm2Y(ec%sAZe$O0$AoK>eFEj=-~vgQP!7Tp(XvLdja=Yf3olOPQPp9C~(l?~%E{K6c3CXv&<*7s%>}STa)t zN;2uu(#1cu@2^cl$d*;NtWS#bDFI+oA_iL*ny~lz0|~tC>_W z0`t@eAuS~BZ7_$lNaize(fU&KAzyeAA2pe_P9jOqx`1#pkW} zIM^(XX9?uemV=P$vxrQP$_<~lO9E8mI0%q{D<+PaN@H$=1QHvMGcVw^QeU-7;b$1T zDg3-p99ef@df`%-gboFZYv?<~PhqeTRQCD`ix=6bQ{hD>JhZ9cde{iSGmfZ($DGkX zQYv?z10DR550H0=_+=+PR2HyPgLGx7BJIgFBu_M?)2sLH(?ehuSQn(aqHkg|GbI|4 zKcfkv*ExMJd4WT zmy4v9WFwFmuki7LewGjZ5`p3cy!zMxKIsnjM3^|X1xOIKDIPLwNzv8ju~Zo5!n2|m zLHD&uV3|_kg~xVWAbe$p<<#;8eeDawF-%FcE~Uz?>}!VwB?XsBAeWOwoOTaD9Jh^U zjrgo6N-QbD;O;SuiXWL%z;wYroWd0ittiy|W+dA&>%t2C{2J@C6;j}mGyx|a>(n?L zjE|)#@~bb+S*lG+VZNrs3#5hIDv>^GZj+pRqK7^xj{?&YM-orA%9^yUdxJf;=dEPX zV$8+Sk*AKYW{ZUl+p;;aD1i&PCJT3QtRE_b6*1bl_E~ni zdTlBmoEqb!REAHss3U!-hEnBmTr4IqCwqAjDyQKc*#Tr5q<52MI+-)l6LWLXdGKB; z+9FSj@29Qsd}_|yz?r95l@1c$B2rAx<9if~e9_$OMTC}x(xP3+Ne>=oi-uA0w3r6a zfm&3%klzwf0xp3&zPxdYjAMLqt3ako8vc(i?7Wv|LH((=zRKz5-B zUib{Bp3@na-~}05h7YqqYAyfDwWps; z%NO#KFVRgr2t>kNAj{N@B44~tLdL#`U)cic3SyUuqEB|wl|TQri>8b>k#}7rV$Pwi zR=%hYU<+N>hx#Jb;%%Z*ctVJ-LBaW=o=ATT2g4)9YxyF~a=NkQY@xyCi_=9K16~+C zS-8^Gn3cQ?BWxwRJQjni7eVdC@uy52!kHAcX!x_g?lEqhoC{lRA7)rAik4rjR|9GA z;2SPt7$Kl>1}yG-I*La^;Azg|V0J7a;*}&7M3F^@=?SIam?0g+mOlg8ACc5Z)+_IA zw@5KePcag-H!Rv{hn>shHA&Q$?zet(MCPF~ z8RcmBeMg%X5zTm$3bk1*2#;pr6-Wz2pm+5-Q0LBnKc-n!Wkg{Fj+#N5)v;-(03s? zc^Pb2O`b{}VSm5y+EgfgEUe|wevYTf*FcDZkSIm{hH7zKswf6I=Zg{Amg` zKcpbn23z*rls97$#0TaC$9TvjF`=JQ$OvPQ-@sY4-7eE#yU)9_u*V;bLOo->)GIsc=@Yl#&Utp3p=tlk6h{k0^X#?y#R0qZo8)i-jr*UN#>I-#gz0l=A_BlFIirhOWk2;-Xywm;j#aq)wv7m4! zKPBgb2bv0}BVtZ%CPrLnG1AkMJc3;pXX-UBvm&j0r8pYzt_+a%%?&b~7L%VKEOZ(t zQ#~wH6lQtV#%f+_45pyKj^Cw-Vnxq3*Ok4^FBIpCPD0TKcF=02PW!3UDdjmX>aaSn z1!MVdTwEVI(7Xu3=;k_~Hk2BWied~>wIFkyMn;Nw+kv$XXP^5;p}YE7JKV zMFu(oU-f+nNh^xBE$}l#*Ay8bRF@Mzpa~~b0){r!Y2}O3?!uzc*N7n;|5jWpeuYDr z6J7vjtKG*S8hZr*ELH@ze()=vMv)jStQnh5KVndbv>l+I)hi~c7LeV6dX_F#|BtIntj#$S z1h8aYrtIZ>33sZ3BG`Fb`WY5Ip9HVTS_gMLz$I7dzoGs}S2wQj;`Nbu;RPlaRi3a2Hg-6T!+@t=OFS?a>KRgMOG9P^c zA0}m6Rf8x?7YT+oDHqS(kF1Fh!mPLOJ^0&nOG2-11g@~kOrKI6r8dGe1fSIYb+w!- zqj8|qOrG!cPzYyjz=hsTuKirDoPrZvCCpUzp^OANYxo?K|ru*BA z=i~YKocL7#`!Z6d%A$k8R9U`(H>b-xs5sdY%iAhEMz^{Z@RxiAo}@U#(vIQ`tt~l6 zp!aNvEi`|7EnBLK8I60$mLf_xHsF_NdbZkO)p5{b$;Z02+*IaAm-qq5qFhp^tEkeQA>Pj@W8~UnzTl;q{T9C2}fAjo52lcRDcZ)8UVCrc8lvF9v#z`Na!~ zUlh+PkCp$oa=cJHzxWtG_~8G2^!cBYP~fj<|M0Q$dwx23{!SvOqWv#CHvV6Hto+1d z^pZUh9TxA`9_xSqv2ympqwSZ{bLb@f-qY`?@jI> zD^5#e4+$T^Bj0+-evMrnx{^aBTN;wkN5-dpJNTC4dGZ;L%QtLsqG9Gs=@EGycDXD~ zrTaq_p05v3+u4CaVLo$zYa4nddP)Du`;m@@r*iR$^=wI=hV`M|a(rdUn$Mtcg=b5I z7D)%PLaGlDErI7PJ^+!tE&0-QiU(tU2EF;X_JqHoB|76_yZ54^TxonAdbicVc$0jF zyGQv_@;&fS22;KyKFUG~)7lNX#Bh(54?&C~2}1h7*;FEKwxoFGl*P~cL;oONixa$! z<7_FkIGmI^oj%N#c16d=jhG*0da9|b;U4+8<& zkK`8sE{X+Np@YA#M?Qke&rciwriT^IM#KkJc|YObaK#R~rNSE!^2Pm4rW)b+@KWafn!a>JU_DqvwO~5SF(xaGw5UNKc|oOC9GX$Y;vo8PT_%%d#?5 zQ~RB>`?uDveTJ{5#q;$-Ux%M)C>;*aJKA6OKOCm$e3bPKdw(@MJ#TiMabS3&pX-&B zp0E9luT(BR>W9H8C#NRwZ%G#8UmxS^C8gnyQyhxtAExJ8zX@$6I41kc35FH^Uvo9( zaFSr*IzirLydpqzD3UJOZ+8QZ|s}dD{E%VJ@*$dq^-S`D9vNlpk!b z$;(^x|3)$h$7OymU4KP6O9u0}oD9TqIpJSX{)Kpb{QiHQ%HWUf(s1@jUlhIJMU%OQ zV}|*4v^(S6eU~SR{k$*V%D6!)^G^CvHOVu3VG8}2a@=BHwb}p4(SCcLi}GF2LBBs( zl<%)!*K^V1b5rgb%X>&1kEF9qg#XC_hmDSBNcifz$kr&}@xf#8{fm#m=QNf5{&ao^ z{N2YJP@@$#MZZY77b&-AX+8Z6LBLh;R9**Cnpl|ivp9JGhREA4`9n9@u!G1SS(4+q z>rCsrP%``1Ch%pLmF}<$m!jLOI!ToZ1`~X(IV@B!WzI%;oXV zA0#2OTsV!2(0h7CexVR642xT4%1=baYlY=*#96L4b*{3o!J9~y|!6ts=lTQ@CL6<6c<4J$+ipA-b;`ieaFymL$EiRgT84fNb z@&^BR*!pZ46nnHHd^B$n5VvASAF^Gau;)fxBU_GKK6@0E>5(swb8S+d2aD4o<;gLN zWrh!zX{cj#TDb0lMbFI2S6(9u`X|0ds>s*4#UoLp<-(1V5>v9}#QhZB)Gvz`t7WV| zuFRGvsW5<*%8!;p8t_zwRl01M#MION5hiM=F@Vso$c&Ca$a3NHfrIRIaq^5S9t2F8 zM>l1r@8Do|ivk&9X8E$?oY(oi7*4Z+7vl;oT9lbD-$T{hMfl3asV`A3gR&SJ#M2XP zFOPjictCVYeHm|Q{|nJfpC8fG{m!uPA1}*YV{Izw!;=baISj=WerP2>Rmj#WSk5vg zpORKqlG}q}{qe%^fa)8hOssIiH2vcOM(rs#4Wkq-BEe(ol}PKDs(9VU7^W&nlK@4h zDh?a|;pQCYTEWy;O6Y`4m!B46h7v;S!QO`BJ8U9L6s#3g^M=Yx?Hs>pT=`=0D*bf=sswjnXdtL z{bqKniAq*ffKt&Y&N*mtuT7^OyYF6WO3(JMQF7AqlA;EOY&IcXv&`n;+mQR%BFliC{G9CF}ZWH#`JE_xb z8O7zw>8a7!$^cXE@^OO56Hx?MEC(-%zMQAXFk6|zZz-_J&auU?BN_$7GLK^L78SJ6 zsO!5M`g&zEUsj2oA_(Pn1qfrOKgZ{8mM;}fg5;H1j#D{=7}77TEf;2eGAwvSc5V80 z8_K)Tfv>Pk^5Ox|*bjO`pRUT5M}?Pycn*t15=Ky--+PcXKV4x~_eg~dYAS%t5l6U- zwImD*`7~_%6)!iC1x~$>$3*-B zGGZH~9vUn8@ih{rL2Wq-i07q%{jxj@9YyCKIF%ZDA#*SAGLF|Tv#q!R{hGgN=ZZM$ zaBr}-wv|zVKS|^jwM?Dx>?+&{qu2Q*R4AojYFK|LfVaQ`aNi&tm+Uf+t!#|;^|bahvC$BD@c6(`9F*6}OzPV600BiB_CeXn-3>qc2Qs3Os zxqf*!^dkUbId_tM;k-5^5Yt&%FB~v*Y6r)lyB8i8SRZsd<-&j(PFLj1$^g5*hA%}<4q;6R-4=bdmP+nRRWOhG4Ol`4xyR)?EJ(7vb=*1 zu!iwYAU<|~?9vFsR?4?O#mF|rkrM4oJS%jEaYbbQ1j4Yq2A}KFzXA|)r{%P6Z6_6B z#?mPQLh1%u^2UaY#EAUEVIpNf+Lpe3pD({ueEJ>U9G#z6{-8$ks ze*$u!lVke&f`7C+Uv}QoS~9A|K6msjs7ZyngLq*2850!eFd5K$g*r=R* zgl;)LLO0$@TJU{@ds{2_^lOkx5e6OjHo+w8EW1djkLD&&b46PBL=UFMKWa3t$uIKF zE}MTVP?_VlITNe#J%{iwk*e7xt^EA-Pf1sE}{{HN(aBOg?JGZc%+7PVRCPUE*UgN~vF+SE zYGkj-5am}NJwH{La`CcTCgnkRaT8{*RE^thPCYKTGzg|*n3~s zy88c~GPk{x`~3a+?D}Bt{oXTkX6DQ}XU?1{sL3Koy>TMaS|hs=vI%+xbgrlxS3Ht2 zz}rm)Aq$JaUIChkqzq!puvtcJj|54Cs8f2K#n{#^z!hV#)yRCpZ9L$CSCC~77nDn^ zVnWMrh2E;>6^OpauTcMF@nzTzuCW}=9<-gfQB#MpK~qp{rYW?mRSi_gZkBOK{iQr( z!v)-+nN1@T4i#Fns0W?#q^S^zx+~-D*bs_VLgZisEgsMKTCrvhqxt2+(aUEbtcz!y zmSC}@+bjy)wpIyAUzCBm+h_$`OUP8?+-9U1@Q9)k-xb5(k$h~yhO4WOyF{QEUnbP- zAR`+%Mja`aQ}gUb33Nkb?)Gz$V{Bw_DkM^Zu>r4W?Q|y+Qw4ri+;(k&DEe%HDWAh7 zWMFnJ2%f4Fw}1{1%il5b=@2TYW6znIEjcciqzZU8P)H-}F1i~{OO#t#m^Y~c*jGOgFqa<5M|CE3h*e(e~BMEn=I@%1X9(*-ypO z$}3n!mT06EY*ypTbn5{eK9~jkj&vSt6j-g$rz~b>HnQ+={GmWrqQ;|_8?^*>3iFv| zCK9=cb*SqOK1Xr1v&iw3L4;wa&a<%9!Ffe@ABbrWMiE#U0!Fm2!%BEL))Glfn-0=!dBIp2df5k8G+C)*7Xsno*m%rhw9x*ta-EjbZ0D`QNpsjZlRAqwEw)#nT_7W-lj=BZh^G0r zH`-MiE1~(wcA}a}I#y8+ntx+fz=O~L$W2mSJ z4xL7_aDun^%w@C)-)Ai-RVemLNHuXqzGA_j=Y2THe&~_1%c(#qB8x5qij*Rpl{!iy zP1_bI()~A*0es7a0?BTrWC{!n4s6pph1duH*=CY@veD#<#J2H^&}rDLqgI#DipmuE zU%}p*)GK_1N$TaaJ!UM5l=uF(I&C_qrU~mDbhq6H;_wgFIWULr@4}!$s4&9t;E|_# zQ6)086RBH-h0dZOdr?eBeXBU_2WE05V*5UYVs=#_0}e+;Fz@~CHO*??wb0*dV?_`e zv{%RuKR4Mqa9=5+KaihIZt5w71cf&ZqJ*-|u(AQ>1V*mdQK&>yV-=1~m*DihD5qS6 z5tD^fLQEaf#^&y#}#pUBrGvrA!{Xb>M2}LJn47~<`t*Z$UsYDEd=}u z*;l-)9uzfHsg4>Hk(r;x!iY*ZASEoDL#JtjqNTbY8!BqZEn49hfh}^`HYsehiRcwV zH^4Ho3T?)ZI_bUwz3-hg3%*ml=KUk zR%|`7!Q{-RLjD&m5dIo2MRM3kJ+CwX;@`roUBkzLYwBbU5esGS4r&*(3j-5L?8$ZG zJDpJ!sy&>}kgQz`+42(tS?0Zai{+O;AK5Cv0!(@7P~|$jMGF&70Hd?r(gb$RqF9+Z z*bZ@Wf`Jeobu$}X5oQOM`VqLOJ;E{~_LUUzYtF!OwJ(K=Afd0R!pZSArV24xtH?&C zRgR89qtil(t<`uWe#IA7k{Ixz3-{SnEu2(YHwTswEK#jQs&JO<(*;EV?};xoVMGGB zR_g95`f<5CFg&Tgp@v2di~hM-MCs}cc48IsY8i6QVKA|!pqjgWKG=jKy%m@HQYGZ{ zr*BMQ*1AnEH6QRTU(KBaU5T6tn2+$9K$6p1jC07-X3a_!Qh>yq`jAEC6(Vqt7C7Hn z0>!v>f+B=<#4Mk}1_=iUnMcwf3{)B6WPL~xET4Uy@82sVH4U@IcA8mqpeQ6K!30YSKuUQMics|m;j_g>^m1YV z>mDr>L6KAE(c4WUXoik(gBDe#e>YG*XY#&N^mDifVu1GIAv##~4Gbj|sp;g~Dkx&B zC^8DsM!w3%A13ZC6kwr{c+^6u-E;Pt9Rj8bu_PjAC>hRbcA=;qc0A%!A^a%}1xK|D zS;fL#(!PfQwdYFUWGX(tFXiR}MTDNxvP2@ni8Lz0|7Da|f= z1AA{096}^uLPfJCpB z&>SsNi()IXrkqLPb@weiG3Urq>rNYAG25WKNDoq1$E%|o?}=d+j)K%OR@k%+?mbz% zd7@`g#&(M+Z)zZbF5U8k^K4M`)nL>2GGgEGkN-W|FNx9M)4M?#n{lcfxu$wkUNcyGCBL%UD`_a z)Fe86vM*f5+FmBe$-5A2=woSL2t;N_#8P3gir!`+T{;qI4sE=_HcRKC70|}3?O*vScrzc$9L3{nj9)yia?sW<=EH@ zl~DpNmRDYP-{N9lHk(}Cn<8gKS(ob0rY2NdPFg*i4poM}hhS#S?@avjo4iP(DFYtR z;A*4`ltNhSaLQSzP#NF%D|G#jvE(E>B-PznP(fzrFcz=SYvJ~y3&^^wp6YeSvZAqkK-<1*j6DK!74fZrxa@+FBai4kRxRhbHQ0Hm>(#>T1DfJ_v6X*r`BSa``?3X$u!P} z;}>di_j1xc{MbWUZ2b3#E^utRbmv-NMd9m(0Kz7f)&(ald93sDsb#)pg30gR;}PQQ z3!SUnKrx7lz3y@$1}>bT7%VlyG_3+qX;Ho{F^Qu|YhQ4UCZDMP=i~ zFeN^S^+NCQ7rFnY&<3Qu7qyr=ZAvOqt;%fZ6`9(B7~>S9;E!9HZzw|}@W1ge3#%cI z$pGu_q99mp+fPU}WM;z!tb>ZT`qDpngg@*`Xz-Owto3L+%ehaK5)~%2ox%*U( z{DMY{Au2I^sqRP}Mjq}Axy)tD3FDsQ`&9MkGko}-+jXLS5Dvic`4`?4!rb=AMWNau?kbaMYHz*ta#9DZMFhXd45DzLP-}x+_btvO?pH|dSdQURPFN5M;Q(p}lWWzvi@=dvJ3{jcr zID8=%uY*)p4_j2c7|?xE-Kn0i5;emz$JTY{F~`^IPQ*gPTSh{I)}0`rj`}ygZNy_T zW-3A6)8MJ)h)UgF>W-?2eM1A4r7^~k>;TDhX2pry3x|&{=x|0E?IjGD2t$Zt3w@a* zWqR{0ja@Z`0>j5IBQqq^NU2ltUPeyu1vZn{Jvc|?pXKZBn=XEzLk_vTI-xR+Uklq! z-QCs{6x+%56nF)?t_*jE`(=`=mZmv8M)fPoaZ62iHl2LpRLpXXJP8;UkQ4YEyrpQH zoXtZDZYXCHDYPhD20f9=u=dD?*Gz17x-G99{STMLbr0Jv%XmD4S0Q_b*AlELx`l<$ z>fRRG=E_26-FOUfS4)r{^e}u+EFV2%PvZ@ACHoXI^WnNy3}V(TWWx(b3V4@^w$m}U zIJXd~m`GGhUU9EGhCq-)=zwNKpyEA@QY>^O<-s-=WSuiTaLWPjwBS81124yoiIs7T zYpe_>bg_vjhGMFFILj|j;U=#K=vK+H>}FzPs)uiY$oj@15Iw}QXs$b=8C#>(qPlA# zx)(aIqunEFM`XxO03lZVtDA8U4jvtbgAf<8IABK>sSXM+P7gsVhw(l|aZsq2VbIdx z3Q2HbdaC5=AwhBE-VO}(9u(z_bna}o;#~6R(W7_sw2O%T1vbxD84}1fS zYX*W4Hv)&csVy?m9d$KjeCldbl!zYB3sethXvlSV++4IhqMwVI+QX%KVevMJ)@)*( z)2XotHp6RAn@rSkuGruGh`&j0r4T&GjOz6uXXi}tgfuvuOGCs%HuNUk<&NfMb^l z=9rEpzvZMNi`P1MQ{cQ6M_h_;B;cF#M~#W?2%wO5D5!%E#HtR)V<@J2gtTWY2C~!L zgH>gJH&lC576kU}pgtVn!f-4Mw0v67c6(Z*lKZv4)Tp^|> zDUh=Ir+S1Pc&bIGdVq>k6Wsw9uLo4{+(On}n;Uq1Ifb@I6`c0$Lgi{IlWVd;{!0iE z`jpp${2(m!`5dBzENbOhn3M5LE+(r|*)#j8<^vj;*av0|up#Tg`MB&%8%(nZl7E+J$P>Vj!V3Xy1N z6y(ho9I-C$JjayD=6cV2h_)9=VhQT3P`LzQvehbtq^W%NgZ$n->K89*MieFls`#ZR zE~S3=1A(wSODH@Fsh$uAD{<0;)G>y0A(pGUET2GOpLLIA&}1ktVkobs*MoFM5h|>2 z=fi#*(k3`Pv-b^bbnVId3)|@Fw%fr2tVT zLsu3jbw~m0;TMK9Fooludpcn63srGdXR~Yy%hm6gT-@I@Z%bShbPrXorjr)I)+D@F zSP`J2(9RPC4lEc1_KA7`4o6(bXwm$IAq`8MGO@y0>>210*2FMVXyp%BhXoAwWQJne z!4Ua7jC9|bjKYYLB`1SfX8?-gMZ`2B?E+DG;e!$+20(~4s>Vz`rBu|7(jLfZSxMO~ z`HU|NU2^=BoOefo@kq?49<&0=xeFxRg?=NI_Z`q<$Eg#S8RCka_c~JOoYj#+=bPrj z9QQGeB?@iIvO8bS0pm^9{-D(@g~6mMp%y|1L)En~)dLy4kwUC~MaWC-^`t2i*Y;5A ziYk|Y8>H;P3DZ;5J>MfFGqGn(vSnyJ?wk`b&(}NXyX-ObPiKIdP8`wh_yfhRRN7EX zTgqrlr!I2pCc6~}+*GofKCshfWZ^eww-%o{DG<%sfz-Bz6p0gY`+7b7K^=1SI%>sf zr>n7+*3;M0px~jq*|^Jnm(`M<@$(Vv>O4J;hJVD|TMJc5m@L!h5mUyXB^WSQiLu#8iYI-iMg3`g0s}i!caXINUXtATU zBK$7#%spTujbdMdcaiL@5tAFkx!UVVA=8XyolIq9@b1#ccX>?upQdXRXiPd%qENbi z<99KxB-?>3?mDX7_?02<^=v@*k@lnBTM8X$iu7jSyzS{nw_x|u4{+(J+kg~0I^B2H z_Ka$v;;zMaX`Lz&&5A-(Sd|Hv(xT5AUgTZKK62UO;ABrsrbaz=6dWz@8Qz|8S8)=5 zVb4zk#Ze=DXOD;a(kW|QJ}`XgWHYxo$88d+0|l(Rnn_=9TU0?sUo_WI&r$lFp0$@yrs zf?cR+9gbPgMt1I6)`(H+oWTdyH&NIDnv&|trkm5gW#PjYWFy2CyHa9&+;Xh{_2Mz;{WX^_@_9l_A;&3B_7|py}$nQhC;=Dag>3ip7QXo@G?3Ck6wy zCg2M~-$MVRQ1jYUPjdf?eG~pJrc)%y11Pit`bo|gs1!X8`#W;bx@V@2bv7K4=dW|^ zQaxv{!JU{`52`ehrWqb3uP5B~Id#DJrFbQ}`Bbb@?RlC(G`=He6?(3s2om405g7PO z{8~nKhR#7U4WtxDc+sXH`bA2y>wz$YAWTL|k=`0N5cx_3Mm1}rK^IfvamXDuYb3JA zMa@tjkIFSAdWM|PyhX&9%=cuo!IwjrN)&n`rOWEGnjVfrU|7oWsR8QT=IlgK?4SjK zHGEEhY*?xbwtjoM2%72ToNrvefY}oAAB)t-cz}1C@_jpzL24vMlt?}}JNX_}?F%h! zz`h;1NOf)D%?jx=);xFq^hR6QOpuIkEPr>o_ivusk$6q zJ9u>XWk-RAdK0B^ zH^Ow_l@yka>kF4YEbOF3i}}hoIsuAVKc0<5tZx$DG6(VbIQ~8xLk9N~^{m4>!-p2a zMnzBB=X0sUrbCoCT@Kn++|`S96BjDuJUZ4gxO(KIHgRX63wQ})U?Qb{%uA%y#kf$v z`KxgWc>||FyZ4IdyM7QkM1sUQ6!AhI)`=-L$19@kLhF8u}T=?6h6e$A%mLmS|T%OnCkt-u9yxylONXww-j;O`QgEe#Q*Np1^FtV zC#~6<%>+aT(mw`v>LTHV)lw|M*3|>?ZP&ucW!pu*GA@--Wg=cY1v3dFKPV={+7E6o zBKNI-<(GT{7nKvgeO;1Wxa^okO;p*ZU1aS4HgceQP`ortC*5ixr{ISzBCbltXvsq- z8;M2IFpVO|$dQyv!AByOWfz6d0OPxjobQ8pqBU7TN=dn1gN=HJ3;`ON9BdxE9l0hu zR4jTpUE8oa^ldCBI67LjJl0BC)u5m=V=Jq0_cZ5NnL4Sw5|JKk4}`v51L3avG%*%u z6}0NUE*<4JubHGcrqF4z0CMbNqJkLY;`q?VBjA{)dh^hvpm2(Y@W-GO)1JfVQc4O( ze;bdt17&LoDc{;Vbe<8OCXwELFnIWFN{Xbui6Yi8$sxm%iPTZr<|vY-i3!jeu4_~H zBdP;HZ}Eztf(tpf4zX*C#vp$o>SJaOO5H?&+BRq62(=+ja?mH=XhwuhMnXf$s^WxJJZO-7O7Q? zlMGfd0^5qn!5ajLxx1!|Tt2?J?0x5ONqC8hB*YvVeK^1Jblg4%Es6#*M zTt(y#_my61GRMMm@vO|@Z21KjQ5Qj6`&Na&hpsgNfviiML|$O82v)z|$uy_l-+IT} zq%x9PO#xPZJ|8J&oeL`$mUB5qtG$EMVUde_@HW&^PSach51*bV!0>^jm7ASPW=x;4 zuMBgL^mSTH0RKoV56v9?6>qced>OJ&Ok4_;z)56M!)YK#b;_c^t?h@C(SQhw+0)Ig z=xnM8sR3cR1c3lhc##>lXnHn_QMxp=OEY)Y0T_d`4uBc~g=sMk|JP&h2qT$#UL4#} zMK)Cg^@Bd3Ay9x(r~%$ay>o5EqM?u{AQ{9_>^o`lG`x+`G`fUr{iPXAeGgxCkp@t+ zV$Ny~nXqVONIUhfEGCj+$GCHN`+X@cZn(SIGgHJLo$kCkE`pkSCr6_V&H1oq%Os<^ zF7LX6Bx|~Qr?F^Tq_#wgRcFI=#W{tgb^K4T20~Sqj4Wc9R37uDJ^RH|MLa!o6kF6A zrwE)I-1m5+c(Vk}&v7r`|4r9=E=4C{-9^ZCFA=ARL*ONTn4}UZM!6pn z9SIZem8(AUiTGX!V=lqyYRJ%)LxCq;-WP2%eBF@J^~JbTF|@o1YcZc}bs9-8rzWeQ zQCob9C1HZA*c`DZ<@Fi$96c4F}K&f4ZF@oR-imkDn&VE^|4ou z`pBo>Q+D|{)W?OdE0NhvSil;+UcS7X6ZY!j{Xkt{U#Lp)UetqFk?CAw7IRY+jBMon zsr3_ARpP!fD4vmWOeA<l$O%ZnO71(DOF{Sn zEvfoG#H_ehRX20xPSrw9l`c`lIivd27?SdSS%x>{h}B7hWSh_D3$(l{o=U5k6Z+W<0ltZVu?egD6%3_&gw2C zBFLjW?Ahh9!0Sx3-1qi(>VhwlOGxD*SWW+e-K;B-mamYA43tk=R*%V^Biy^)g?}yb?DTgV=^#><-jSndVTyFSvQ@tVzCN6j2)2`L2UcAjseyzw6 zM$A+CO0HkD7o0eVm%xF}$OpY%Fr>;W*H9J|S`INFcw%@MKkN~wI&vtC7++p5Ih~jj zhTE$ua1WD#K#{m~ZD=p}Rj_~s_j`)@q4nH%(lk*-q5tqWN%fjb_m;DHA7T%UH)A>R?LHgtmq=yD>P<{le-E=Ue4$)u~=6pNoDPFFQu? zt;sn+^o#Sfvclw|6o^0mlaLP<<>M{26@1|E4DsQU>=f^Jth@#>2as8H#DB`Ik7cD6 z$Et#`*jzr;EX8u`<&=x=NZ%PnE>b<{=Lb~u(r{RpbJBSnwS1Lf>m#U9)of zA9@F_TC7N#>kV!}AlM0l3YL&25$v*P ztyFOw=A2YiE3R%3_yFH8XI#_=s7~;V7Y}K<^O$lv5HV1Db8qK{uzJ1m5pk!f+A~5D z6N0!?wrb%knHAjk*Sjflb_?g>X=5T&-L*Fh1=_|VpBLv@WiE$4#cx1xs;-1u`3hjf z-57ol@wNuFS{A)!FL(ec&B-sY*h7^C@sONU^_>-c;T82MhA&n!+T>Iwl7y3;d8sJY z4N}#5hg80Bbeh*&jvdX0RM%Bnw}pF$lq1%Uo%Y53;S`CmI@KEsd!WE!6;njl-mF=G zfuaS?fd`t>2xlX;l676Q9IrQgU(s|3sv^l?2pZ|@FyRKe?SpX3Llsr6GmNOKC&h`T z4%8c_5s(PU5lbXw(|2d-COICA=2%YSfn?V<9_-oyf=SDPAKauf*%_*!A@_F0GzqYs`jw}0 z308Elys?RwbZxmzGLmERHxgv{lx(a<7JEF9QcPwH8s4gsox9$WhYE7Dg!fZ5)h3NA zC$QWD8E~oI6o}KztBUW%V+Evlxr@_`Y68MtONGU(l~XTBLCtm2H96H=*Y^10#Si9k zk49H_|3T@P4jzmyjXprF{8b<81#Qwor6fa+``mvP(g^TiH&ayzeTs~q@TWp*T^%He zp65c>iV`v1c|Kf#xA6dLZB@2Y*NKbK%h&~V>$d)Lpx4K(Uh$l0sXo37M{K!Cv~e^A zP9Hqs^YDK7GRb3ER{gIH@z===IroY|C$!r0;Lz-4s9RMPuGGv_ySnK=2X^Y}yvF*Z zQJu(m0_Pj&2Gb{C9RJgAPnd#EY`<<{m4KHLec~2mD)TWiIkF|3-kcJpXkxBD{=ypy z3FO^~K3QC=1&I@QW9HOPsieS%KB0v)8*#G)UsK@?X*ZF!CV_Dad_&br&P_)tZTCfP zr9QsvBoy4nn{Ui(rw&O2G$BNv(ApS1$aG2dnV#irLl3p<$bl5NBD*^s4 zn0w+ck^U2WP45JVuWnu+Au)nQAC}}1ZWUsT`v+n_;O}=!WuM5N%fyNPwhL+?yBd?~ z8ylM&pgF*5*OsDBgOips5VMIZccGToNlTsFp(a9Q=l^sF>AzIWaoT+~NoAm|Q)zmd?_U$s`r7N;&=*n3cv=y1hp z`IS)PzThgaX}Si4%+$VyWg;@weUPg+SZO*+)DT2FkY)u0lB&L6c znw*ovEY&w^m-=fjwg>yV?EkrarA-Fmzy#I3bZWA{wnR4}@Mw&0@kJM^zCI)z^nwVP z47lM|Dpoqf5#spZsFD5`BQ|7!TFN&UXM?>2<{>2<#xRPVPm@Q{LK$k{Mm!*eHX11P zN$62%#gzSq02O;QF4dR(T<}{#^%XnQ_g3>G^_avSd41vNv{C%1BUEUa^Y8kWCAF_S zBELUdbsP-c$WCy5BuUmLQVbe7mPq49rv%GZux^%P;^0xRCD0g+)kc45oJG!B9R8rJ zRkRT62Z1u_fmUP%4}0GAgUDjA3@bG?4IXomG(k0}5v8GJ7hOrKK;$@{o-OqH`ddNM z;De6DGYBfQxc!cNziViv%6UXKYI8PtB&tGap$Z-eAL5VPg*d|Y&{MCkdz8EQ`9jYR zQS`$Ob07H`ymBA8>|3r~y#D(9ptfaFeWz!qI}3uHV1DQ?P_zO0kI#2KH_HF&pNI6B%hjGi~lbE!Fcr19}{qQSGxZTx6mOe;hBU51ih zVaarw+BP?T&h+ez<|YJUc-umy#%5bNLfrf*?neB!SoEl&No zo0IsQX7CV>#)g=T3LZ+uB7!J*D62!l!_lP)566uOKSZy%+LB8MN#4BH*HHb&Ed zRW5&ygf%M})Reg#9QXICRk!M=w<1oihpkKZapM|ky4RW%PSE7=#2U%FkR$te^!eUo zU(VIQ1Bxk#8$S3X0R^@4*<4H{maASrY%f;hfaTKqTI>py)Hg02v6>D)j)0D%mBSm)2M^Ip;*FP*Z!yaS$2A(>tM^h@%p!D?~ed;#`h>$$ir}|Caunu{m z`}?wpGG14ce~=Rl@jOiR_l@3=`ReRmS1Vtth;(t9X#bGh2q(!G!O=goX-Y@t$nVG* z8`<*II)}axs8+?otN9Q-B$bgQ0V6h)*uFnsz(ozgIe*S8&YGO=e)h+5Z3Ke3ry+_B z*-UeFz1ZI}io-2}#G+#r>}=8&g|KyTzj3O+nvogAC*hn>B+*}1TcmOzE{DAS{?cw@ zjjT34G5NCdospP|1hhv+9narzT)Xx{tbd)eOrMf`7BLix`=LJK5gR-WMgRCurNQnf zSumchDF)d%g6?o~UslJb@5&Sh8AdEO!)12#%`Vr2zGdm}B#v+fk@iApbBwd4xMhcV zH9z!}&ouf^4F~YUr3m5^kkLEgzYM!8E2wJ(x9|?aP z9V$8!=fRr9QQM@#sSHESX{s4J7qaECofGHg0Sus>v=6&T{OJ)HEm}j;95kdaK>K4I z10afa77{T)zaT%3Hnmu0b$E}zIJF%#3)(=$El-Am0k-h$3n^G?ddT!6o+m7=+jTVW zSwBG4VBSO4+sc5ex%f)_nc}y5d*6Xy4ebkl+e}oo-ab4Bl_#O-vDEhQDICeOzR=rV zI!%zn(j_I40|dpLw#0EfXbikybeAYM@>prfK!-%ucu45ifeaGmp^_;4l% ztn$cBzzNUY_i;gY$#j}ebX7O~O?PJWxfqJL(+u-}c{q|6^hr;li3-Ct*Z-OOJUo3D zV*JU}@{eWxt_B*YZ#R0|>j^v@y~BxW@{r8YbM6vJFZ~}{zVyVClZTP0hXFEybaZ8Kh?jg0Bh%Bc5YB_%)79d_PrJ4F|>x9mL zxEyhFQF1peL>*j-%+4vn-Cgxd5h36z>DP~sJfWtn^W)+Iyb!EpIyTZoI``&~T3FlY|yoAh(}I`7dliUcNdu7nD1txpQ<1vt0Ld9wu{l)VC?(4 zE8G@1vzo;TJDf!izY&^3f%n#UBa3SWLWZZ~DJ?d{r;rQH2`ryuBre8>dY&+qw3VAq zvMx!J0~MMDva*>mFIrPk7-gB|+^$qtY{ozZa5cpoj%$co(^OiXbnJlYw8d;BDUg6k z%cEa95Z}OLmTJglS~$*9OF4e%2JA+v#||=**=NonE5p*k-({81SIE6DCKT3s|L^XV z4o169V+{8l!o$p&n9)5qB}5xpBsekdT>=Y#!0!+mMx=1k%?6xWbGn;~Q9?ZEXehA% z#)+{=U36`VN>gtKN=kSD4qH|s3_5fZ8hAEKoFl0n;*JU4<)*8@o%1?c7bQevNB@c; z=sLg{vY|$^L29nK*1l9;A zr0YnEjV?uJhu(=wKi2v1R`eZT5>H6uud-6ispoe>+zNGlc=vFil!t_hIFn`5Usm5p z+XF#F{s7;NZr6EMT6!5nC|c)eyJ*d%VaD%pZ(;Q2Ol`kMBlYnIAbliocMlB52Kc&f z*aD`$U|RBpz;8q!nqIDJ)j(3T@xRiMND{tg)aa385i1@oaNpr|#2**p*b3?#-R}2; z7((7@j33r`k$3}>U?p-9U^}-ChnXqtil+4Jedl-slRQIk{vvD+^+CToEujPV@s7bb zjP`~9K00H^+=W^+;IMwUqc6{~_Ju}S*dW7~LRIR>*D7D~J2dpIK%tZ`%{QN6&qKia zQt%~nbEjh#IB~)(z*j6ygyZSRrsS7Gty8LT@s`@rG70^Io)Y8ZIPU9jL!4b0vJi7$ zQZ0`zsq_cjj+UZ_q0QHtl#bvC(bSGZS7j=fm$=h#^ceq-pY2c zaL6uhHyrwvNQ~3p6hScWS8a3e$EvsIkl}2jnc#m&6{oEL) z;1|2&$n_!JXGunHP|!w*9$F~*`7Az4)2Oi^P9TPx4m?g=#^ByiD9az?UbA8}&auc4phZW9tX z)ScSWKu&P`sK;qR@p~%z+A#JP(F<{gSh7MF8u_7<$sjs^;ct)^5x#{)LCi%IdCdvU z$i6IgU~9cWR^7-B7@;RnM^HvZZ|88IqMtZy%kT^B9E_xBH~2a^TnFi6QaihTak7!e z90jZ|uAzpMAa`qX!@?+C*9zlo?9MVY14)C0G;*7T%2D^8%yZeqL~jLt3mwDy9(Oyn z72)np^`Q(<&SH!3#nms6IKB%e@?P*f$I;TkdpZ6Vo6i`>G?FMLmb^R|$l4C685TKS zdfBzAh69V>V!TIB;PZ^c%93MpW@F7#cqwbeF&U{J&4&(rF-jM#HI+ku2HEXU>FP{D zZbwL_$nlHF@kDj;cspCenHCA+IsXSexq1!S3oYq4Otucchlw{Ln0&-%x~8@Pv92>P zJLGZ$xz;OuEYW`t0lmNJJw*2=y_Xu?;XHCFa`?R1LT|7+F)?7&ypat8aC&ffraQ51 zxmD~h1`>BJCeOGr{6)hKz{sHCLuQFT7zWEE4y3N!yX@~8$6wUP{=p^Wbi-ZO%Rf1`dlxTH{}cHKd0@@X1YAgoyu3-LF4_zO)Ah()8g=WlW&L3 zoUk#i6=3nXJfrs*4ogS>Ek2*(i2mBq0YP}5sRBGab=3Jzx7@!JyI1-BM17rJMO!Sx zhAIYM>!`rjI@EksM@smPXH+)vUA{0a7%hqKav*p3U2ar#&Fo+e9ZEf>n!4_xcNEEu z`bOry@2)78UW;hj{9R{P(_~FBcq8xXsBum*8V{u*lQ!~hNSh2^2`t;lyTNx&1Ggjo z3W<;nlhAMcE-y(X(5kR~$?r1He7_UvCnDE!=cMvCckoisM$GGP*B3H@$)QKw8Ox;l zia8NnRZs6)Yi^EuJAwlIyj{$(jym`o{$%(!B1<4XY`?4%L#9_Sf_I6Iv@rMlzx$$0 zU*_9B^nli|H2HI(4mG+Q>Q2WCzpV6!r43^k$ms|R6X@6+hfCK{xe`=~_5qsOAVw-5L(}$Zp z(J26lQBCG%FbO((-RW^+e*4zJaN$b)9>-znc!?~k!&&e%%Ro9BHNvHl^nu=R(y@O# z8?`cYN{56$Z+~^X1BI`LK22R|n>J1(DM_fin{c67;f3C(rj-Vk8n=|omV{COl$agm zPdMHH@Nd$=fVAI-_8fdZSfBWP5fw&6i|NE1kz60He}uqD6A-CTl^Bj3p^GYw)HxL} zei~hCLeb8CfAE^-&#O@Z4J5Ci%U~hw_^JW9T`WBiGI;j|1K9_;VBkOy2|+>*1-EE1 zPl}6K=*Zt0qJJF{S7+QLTSPuLNR3d%a-?=X&1RIeJ~UYUQ5*eBkjg%}Ei5irNN{uT zPi_llnx25e{Rh>V`J&j8r^M}34#TUyziy#dD#`-It8=30`I`MMIH;rv;L6?k(Ekg@ zokab=VB8T^!i1W`y_~;2qji`$8J5MlMgM3bTpF%iWE=0u7ki_F>CQtvn7*XD26HU3 z$u^oem#r-MEvzdY!=9^!p!C~8W5@m9K>@@650nCqj}CGq`|1Oa7aBqA0BMeS$}sB2 z2+MgI$pR(rGfe9Qqfb%(ipIgm&gaB}0QelEe+>1M)B3q%e#_c2CXgf{vEZbYd7y~f z1yLUwfxm4+X}d)8oi%+Q3XJlXd1474h|#l_zDOBs*3i;)dAjq&M-@Y`PzlXg4=Y@Vt>xe}Bn!oS!wS|BB4Z(YlCDWgm z{pWiY{&8)8OCJ3GYX^R1;YVMp^?y8>{=FObz0$&esr0>nCgW!<9CV(A?{u%wZ@VQK z|M}@V{=>o#Jx1}LC-;AL^%K`x_&@za@D*Pq<5z!tZ@q=zeUae3iZO}(|DXE?ere%f z`Bm{>Cev^FX#Hmv{$Z_u*~-#H`fK{Xai)d;ua4jJt;zg9_Pnw|7XChMf9ufX`rqI5 z*jX0-nKQKiCMV;6zwCv7TKJ6z3chUbWchRB^A&uLYBS@QU~xW`4$ zTln&N#lM}5Us-$i78d?Nenz?A%hn|G|6e`Vy>H>SzEAm=hyIE_=WJ`?M=RbNl-&MBj~w-bg+E{UmmZef z{*kAiYRaDvDy06c-zMYl_-g#F7X9>Pf=_>+j2}E_ae;-eSE=9fT{8ZQ`fg`i__9AJ z|MSqVKJQ-*7QXLIg0HwZng1_N-1#93U#;ygyDfSCe{<+z+gSKh+^Y9%GJeAyUtDhC z)7@16%Y(nD>-fVgeDzx8->J#+d)$TF{mH^tD*xKvP3GSvXRWNW@O|5betK2%`gP%D zM?GWV=R7X>mYqrnF}oZh<3rZ z<)MG*x<8x!*RJEg<=Eu<7oFWuZPD+1ztC^ZBR}50aZgi!{6xo(_iHl$_r7KEix&NR zb^LfIB-ekFozB~6;h%KlFAx6RvfXdC@Kdz^TTV=-zkkn{=Z$b$vO-FmBYW7j6dn^AxkWL!+N2g9-iEPFJG{qNxwz$-b=~!*WA5L ze~bQC+J7y1@EhMh;zJ96c7yWo-z&-dtDQCHLJL1j>z^KxO#g)1 zuBQCHS?Rarp?~;aZ`#qKKctWL|G$&#zj5W-cP;!$LzMq{=I_zlR+;ka&TRzWo(JE) z_3htS^hav_)2ox$zb&`f(bPZ2E50HR{`(L9bc#j)kFNcB@OzGb!}MR3DgE|5__kx$ zjkf4tru5TmlH0$_^Gzidewu549{lRkUZ(wdt!sZC{L+O7-(%4q&{g|C4}QmwR-5|I z13G@vM<%!b$sf=EgGGPzXrW(`2S4N4Ee^8qJIvDYp9kOg?9WG9_)jXe{dw>&{JpWv z!tbf|Pal=s{)J~uUSZ*dGrzGQF zx?_uOR{Q^?^>52V-#huZp%#AHr^3Io(~{{QaQV0%7XIe0Vn3#jPM&{nw%t}^;cwFR zr^}Pa@6PMDddkAD*YV$WS+f09de}@;eifW0?QiXz-2QE@-}5Jnep>O~7RlrHz6C$* zY2g<+{_mSy|A|G5$5{A(>-??A!~gF`J!Q(TSCs$Wlw|rtcYpH`i+-85zkSnW{_Wk^ z*YrQ{nv(I;cDvpG(mS4)x%#8kpx?{wzrh{Io{zyg|J(A^|K202?zHN^_B)}_vRQKd zUw!!VPb~bq{}g<>OEUhKt-dw=hwGJp=_`}R&kOs^G1sqObpCmpC)0oE>(d^#>i-w@ z|Fq^AKi!5+`_965dP&+}wnZ}iKL4yb)54#4h2UHB)c@yuFCAv#w|YYG?Rn~7^xcH( zE&LJ3Y5nIU+aGni{O$k?|2O4d`j}+-*Z<6U=K59jmeM~fnSWo-`0){o{_#40TA;|t zOp@=ZTIugY4|%`J_P;9sD)Q8S$Iq`m-=hDjn|~iB(?9ryEAFuHeRTY!@37~;Ye)8* z8+Nns-e=nXdB)$MZ-+0k@O?HHu9ZzouKykDR;MidIS&Z^mOS-8}@9H@g{{6nf z|F*M}=|9)&-K7@(3CI6)lJT2Aa{5;m{+6LSYqm+oH`N?*mW7|E{3$zim!~cK zAB&X!<;n7+_2}WI{hHoZ@a=i%KYq{BRTlj|59;`RJ9++jbJm&i<2Wb(_fHwT^IFL3g6>*V!2)#cAW zSoqtt|6B8nzX?4Sne)H!uhM^QA0_jD{A)L^vFN7<==?h`dHiqnz*naJeWJF%tXndE z!7baG{>Srm{Iuu6f7gB0+gAOn-S{m?rvJe;r!`slx<=(+VKV-eT}PZ@;a4er@7?72 zv%`5mn)7d+;@k6#zYYHwZ_4kzwEtR1Cy&1g?c0s8>R;;CzYCJvf5OYxF0}B~1u}kG zu1Fq##RvYFweV{d-}Z4b{-%Q-G5NRMmC}FdwaN11gmZr|GfPfGul9hOXg z+?+0}t<4twKPmrO^U(jf+h0uibKR}l{@Ka$r_=F2|H-0%v(ooINpAnt#SfVA z7xM~r{N)+Hk1gt2VA22QJ5v9)&ywrEXx0y={O`?VoqC>c7pGLO$4gQflzdHTO)t6I~4zw=7v ze;)kE;sSI2j!}Gj9{iqPJoT+r|9u=j&-gic+^J_<_`+#Y|FT1q<$uAeXZ+j3zo7k} zJ}!Cwj2yMfv|rC$s{GG`KjZzW<1P9d6z|PXuKymlK5FW(F8;XXZd-my?L6<2*VhjR z{ha-`t9Fw&+VkK~J$n!G|K;GnRQcVONuGa~-E{ggtNq>uLO*>&a{GtYJ-CmBFHpSq zU%UNo|Kasts+L>$^L6~S-JCrCdVF)%4i>&#>9^gKT>rOk%J#AFcRelbZ_P7)7R~$4 zlz-o={BB#4%>Ql=Or2}dFLU)jDcSyi>zA3YS@?~eME<2qlj(o{=m95K_z~R&-=3%a z``!BSV-|kB*1znQaI|F%CSkKeO*m@>w~5Bfmpr$;B_Z`X}dr6={Y+W(l+Z>dklk2~}zTEBD7zn`Q1-?Auq z{aSSL!@Vv1X4h)}9hyx4%4s)TW#R9y5q#NQ$@oVu+tIV|(+?GV#qr7d+l1X-J;K6Q zD!#2bdHk$6<{4cA8zIS(W{dbyKY1+>f7b?DUGW`$FIjO~>zttMSx8%VWuY7qE z3;(Y2zkPUe{o7uB&h+2jto>iMD!Kk$F1qq+i~c&TfBWspc(lRgxDdPH*jyS;zi{uX|xe`)=@CiDM4XWwV~-zGjH`1U;G$9v_q zjTZd_lz!R$$^1L%r~8X6eCtI*zja)4{d?ZI&a}TPlzvN|_HSJLukS4Sn|`DG-!+;3 zDRu9d{_{<*Q~r-i?!PylEIHbuzu!8+w>_AQ-}dD<&Hb}Cb^f&VP1YaAFTT0mqF=l~ z`Ts(4`#0S6@&F6(ZBTrk_3yL$FZ{a`Wfd zWc*fL{?N~&|Ah8`*+t3tQ+Bzb%EFg!rt9yRWd1!d>z4g2{P;(;|94BKzwEJ<|FQ6U zJg4I~5C59)`mmRUFIT*`RWkk2uYYaMpPN+vwdBDc`tubpTl9O>O8whbB#)m7kM=S5 z4-ZrRwdUd9rt4|>%C$e{{i5`bPj3IZmwQ}c;V)HuMV|Tl?dDx}wD4m;68h~&Cd;2+ zS_T(e_?vY6wdEOq?^Yl8yoLW!@nt6@xBrZb?=s`J&eQ%&znqN!eAC~px9C5w^0PG$ z|DKt>e1e6)R{OtVZ1Vj3U*_Oe3qSOA<$oUf%h%2@?Wfmt{)&33zu3~EU#9q$JoImPX8vjmpIM>x@19)$)0SO1(870qUhr-AC-eWZ zyZ$lJ!e62D-}^`M_V zhkvK9e9hE9zCKF%w>EkFA9cxLpIY^QPx+T_Nycw>b>le}{<{N&enp<~*LuoxDGTo% zt@Y2-e_xN;^L`6IUh%DY=)e2u_n%q#SML`3t$F0fSxuXn{`d4Hf^W~mztSyloN3Yj zS>geamKl%Md=WT1Xzwcf` zzwN5z_V0Fj!)6x!CscmM^DoEiKdJwuxfZ^EozO4aCCb09QtbEtFF)Sh0{mz#_WOU! zL9QhF8>heZZOeAG=ubOFak2cpas2NNf9*93Uw($*D_)HDU%dVU?tR)Ef1O4NKD~P~ z|NlC5(2o}V58U|uC`vzG|5tA6yTHP4)cIdAJt}|W_=f41lw0^c?$!Fgo=kt$6+iA^ z;g8)$^*6aLfJ2tw0$MJ*bjlSQ)zoGnZeyA9dqW`nvilVo7T5BOdeRW*rI=Hh0t%`Bf0)R zZ2jEM7QVDcXRrOXl$Wkdp1&U-x^aff>ci_@R-!zmY7^?$8Z z+Fvm)nSVuJZ25}CKfVR{dYoCe#G(bKGg4Y3;&X?-xY5q%m3!*?>GB@WrOf9o_{w^|IjP`XO9099sb2+ z{!iHE{oz*qiw0}|#qx*7>1WD1A7|l7|8Bhz@;8=$Gmf8d?|T zxA4uSTK_!wTi@B60?=~o--Sv)p8qpm|9M@znd|>rrC$-tzZu8hSvAvKf0w!T=fR(l zU3H;V|8gbNo(JFm*Ch{F_*;hv|Kj;Sx3=(2 zmkU0ge=}bHZ-2Sm)PIh^{Nq2d{KIkl-kq-ai$(tlm47X<{Fib3xZ7Kvw(!4yQR^Se ze;LO=cyG5eEc}fsztWp0>kk$8mfd9Gk7yVA@%))_`X3%M=`#yI>QlkT^Jm8KODEku z(ZZjs^xHp4mj91@@LaKl|8l+1kDp(Q(;t55yc!FCgyxTHc{N%7&${?tQ-1C6fVThp zWc?{y_w#KQ{dYBgU_Ad}y#6~L^}T8TJ>&8pew@s|0rxEUpGE(39e-s5lH0%N_zz6| z?OW}?c>cwB{r~*ak;5(eW0ZfbUnJN6hhvZXw}n6FX5oLu4$1n*>5trTvxQ&tmf%~~ zC*$Az-Ip6J{AA4^70+K6Z-45S8^{5Yvwkkp^V{hsljra4Cp|aD!f(+0gYo=tar*UV zU3-g#zYX~Izooo<`Dbg(hrMCx&xEHxt@|X`|H*qcJ=LQBEBdeEA=oeTlI^cd^L9U4 z^a)RY(vyz^n8*tBC`HT|#u{Yu7fdava9-))Qg&HlgnL7|^7 z|JA20J*V;ujcB)7lc#UGmcPlQiDiuv0!x&Bx0{cLxOe}7Q^$MX-x$6sdkgs&|8 zfLpZxuT1uzZu#tYzq9bqDE+j{e>AS6yJ)MHRo3z4pnL}{|I6JHd<-t5js1z&f6%3# z>3=%6O@OVj{5Ns@>{X+;vFiW#D+S;3QS$ov>vdzUu<$QCd@O%Woc<+m9s8_>|NRu9 z-#$Ef{M@kHEo~Nl)H{N2yCb>(&foahFBblG#aEn`tp9xR`1U(m_)T2?kMol0uRCz~ z4;Fsv58D1|$?YGx_>c=M{N5V{?>(9<|F3!b+y^cE?Vl_EV)y%VZ*CQQ zOP6H&*Y$hlE(^cuZHoUqng83jb~E=sKUMuDe*bK|{u8>LS7p&ZWtzy(im}P#|GPb| z*uui^)Jx^>&dK%v&mNDN`rCU;l>d45k3U&3@lK2W!`lDplat5K;&pGE_IrOFzZF}{ zy~I)*&x`Z#pZBjb{Vxl&{q6fF(|@+E!So-_{88Iqnauxv`+WSi)&4g$e^W&aAJ-p_ zm~^%2|Nd9Aw)m}R{p0x7w|6(spD#Q~@D+LR3%Xuoo*!&i`fV|Mod37fHJSS-Cu;w- z)+Y1s_Vr(y`%itIR{npPEWh5r^L}&u6#Px_=^c{k5B~37O;-Qi|E|)HQ|9kDt{jK_6qW#xal`KDZ z+@;O*|K8;0PoDbUGWZp9{rOPmZ|jWc_>c2H-S*W=tNuG`|F=&~#vk{^y4@}ObK3v$ z{B3diHK$zhUkiV=%m0xl|IZeiZ@-IWKU}}vCOiP!V)}D@{;#?1j6E&<&Wf+tJF36N z@lAjJ(LBG>dcKVRme~3i$A7$P#bS&8BY#%>u;lrB@s=07WAX3sF9qKk%YPN;|Jc^P zrvIYhYgzx=w@)6wBiBzf_3uHqYyGc_@-I$*$XS<% z-|_rMaeVQKQ*O2Br!SED`}upu4b9Zn)-*KLRF%}#WSb^6H?GXgA6nJ8rlF~RL`_{& zrm?QFb_6}jRF#yE9zUkM9;FMt{~K;>>8SrNTWBG;R_yf22-!T^bd3Ojto*yWVFFbLixj#|yZ^6g& z1I6+8{^>%~zIl9=;LGkw=KnK)``uwy{n!6d>mNH`9H-ydGS3D-dW^x>v_ret^fImxxaOJe_0>XOOoYtL-|)G{Z$X@`kDtnwt4uQR{OVmRodSk zJ0BkBU&UqFmn{6DrwYDem*o0CFyTy-evRVe`C;Pp*R|IjVbNdh=#NXbug-n)GIM|9 z@fz*_Sbm>4{rh`eZMOdiBsTk9zVf6pT9xt-?C2>ALrjA7yKt{wSU-tQvb4r$?M-| zSM)v8!hi9z@^5gm{AlYvbiRc@|6#@dAsK)8?{78x?mIW7(`!dd|D`8I`5*7Uop;~6r-i>p`@b!QkN4l$W2c(?yQvSQ{@!U({>ACf z+pKZ1MgM8#f9o&N_Q&a8y#1RER{xE>Ncq=08UL@XFEsa8wpl0mcz&5U{gYqYXvQ~O z;QB9?UnY+KrDf1bR{b|P{<-`r|2I9gwAAUh+70d}PXDT+HgkV^;c`L5&u7N*3-?>! z$EyD-#iy&2d{eFImaYN^iTsX9< zzOJgdu`yHERD%CEu9-j7H*G5y*Obgz(~ud)B}*!+GU=m^Stwr#{u`%%|B{v0TJ-Oo zrQ`2|%ZldN6ohI<;4=8-Tq-R z{d10f-i*)6ZdCm_PyH|WX4F#_{l3=-{gzjg>;KD3+aG1&KY301ES_H@-u}$2#*;1l zFS`r>+un<|KfeC{blbno^YPmZ)c)HtdHtUJ)_gO+*zB1~e|<9j8-5vCV6Fe-YCAIn z%Kj0RAMyI%c1rt!7Jf)Z@nv$g#(`tklN z{P0OLzU|;I6`x7gUnVZ^Zu$q8o&VeFKkB{p=K1uq4j2Aa{5_ff>1ngf^I_8lDE&3b z_;rteT59#*{o4z^?Cj+J`~FW49ckg`mFoO^Bies){-0X2`7~?({MGSqNtFNb{%>m< zb(clItMaeq$mI6F^W&e*`vE3?qvQ9@=;#xy`fs;f@NG+z>tA!p{Rdd}|54kY zek{5E3vc@6aSQ+VHt9ca-)Q~g{G0OCvu1wBd;X~S$D{P){WoxN#UWPvz1Ky4repbm z;`kf)-Or3~EmC|-9{kn6bY5@K@B1I2-yX|v6sNz{ZhtfP$FDz1@bUaUas1*7Mr~`+ zzvVQ=$MVC(@n>9imTAA7@sZ;H9i4x1{L9bI8)eb|r|Mr78=~Vcj(_Z>NhKEkbmd=q z$2|DHkL+*Z*RD|h$IjQp$KQTeA2Qt)Kf5ze0M%y39Z}C;nMhid3ov(j3 z+5R|n@DuASe9gyF|BAWE_!}O2{(K8x_=e!U+h23zxDGao<)CW zrQeo^zIXlw=KMQP>9@~IroY+IN8E4GpP=+xV)=dI{dda4XPEX&x#G)CPoDp04|~^K zKYQx={?=H2pg8@vmv6Yws=vQK=<=iN+EIMplIFUqAvN_Svoe*{nZ{y{^7H?R0KZ*_ z@*6`J*DNiWUe`1%9sF(n!e$stlbKQ3kl$JxKXe6N!v|{`N-C?X8z*O)Dr;&N zuF6y{FX0WuOd~xVk@E&>JAdKOrS&D*Wt9z?5*Oi^_RYhP8jcqvQ8&0t88*b1L+gHFBrH& zURf}3U_fHI&eQN~bhduE>z_ypY%*DzB|vnq4Rz?(Hso@@v!KUyLeGFsh5jrIp_$ zzpJcm&P+)2PqYxMnxG??>oqHLL~|yKVJx+xu^ZHP5;Gqo)gOxh&!+}7S2d-NO1mEU zt(`O>=-(-6nkx-WjhJ8~(?gqTvn5qE4a+i(vzu$0GII`?J!&*hJ>?3jy+@eWQrm?* z73J^fY5E9Nwm22edKa^2m+(3o(sL zrk78|ME7a$*@4s<+T2)MGI!Pip^j%pOk=T^7w&lO|E|eaRo2#)%&J@^J;0*NQ|i9_ z9G3P+xr9=b%&u8lmuZ~Y(3CZXx##D-(bd;$f9UobyLIil*~{z258U2kIOhu*`-10{ zZS3Ur+f3dpP=M3V9KL4fi*M}Gwd*z`&x+#rc+B(K41D}MW46EX4&eKIsPwP?VHneI zz8CcWmY@C(ul~I|>c3^Y{-*%n_8aimZ!_Z&;Jd`}Ph1N659OzS>5V6zZSij+@@KUD z2K+(wg%<(8Nt}Kg@|To7oS*)eU%xrc!taOvOCOxd+kfXD{QfP#ZxyFsi}Rapk8R>@ zo~QrcSlW=W==W_$?yJL6UY+Jqa{Vq|6E}K1@3Z}ue$p+K>U7uH(ed-+TAV$qOnE17 zCGR@C`*q{+t|wP_vBu9yLva2m6Hs%wphUhq=Hq`)vgqHr4e0*{{F+m<=K2n;Kl3UllOky zQet{l5W!_Pp#rt?@emgj#$Qxg+qHGlpM`t2)n@vh$TRnPnM+=<&;x4Sum9oiy!aF9pEG~cpr3BckKezr^hvAz^Ik&#H|OGAKe~GCwYcgb zYyC)Xi}SMvJ}!S>J*NNmmi*ll_|}!V^yB;=y~&uB7XJ?=`Uc+B+x7qBZMEeV|5gE? zUX`EzVXaqYE&A(l{;_pU%6n^Q^7ySBedIdebJmaTuEh8`G8Z4O|L^A?Me;jG|EVPW zu_^D<#wh(b{`w6MJZs_U{`vHADes(JlkwS4I#00hbpLbv@AKmiU%BKc3s3hyr%%eo z$LG(nFTZ@LHGk;-;Pfdeug}V8{o~`Otp3RH7XRt~;ELaX&n$XozJ;gzgIf)Jy#9xN zIrb{6{&atD#i_aVkK+&i>ddJYp6>5$HSqENUv^UMP>cU`e{1_`DX$rPq(AX_FQuNh zVCS;GT6ntuwEgs4e7yd9?egpGR{b*+Uw1+$?CzBJ|Sqj4G?)&H26*Dki`)BPD0tts!FVZw2TkMqCypx(6>|LOjVcHqy} zJjxCq=il&(E^{sVbbmv|`4v~CyzGQz`d96F+hJDyDgS%>pCP|amzmNL{r_Hg=1ZWTqkmEU_w-+K@!nn? z`Hf=!81~(`vZ~6us!VOB5#F1;c=vAP1Ms1cvnPFwfCkcZYMoPrNyo_V=N!${(jcuI}Ox^4qR?@Ey+mR-d#Lo$tuOe~$i>Kl&ZbW8`q| zH`RXapDcVD{(~^SRr5GGock^N>HJa)KeR4}*E~)R=YE%Nd~X+ve*@t^p+Am)vfCf{ zpB%pD`{`XQ`s0$oBj<4L_vCYr>}26vV*Kw2e^&X*g%y--_bfKb1UcuHEkU&6=&RvFcwmSMVk8-6rG@GY}N zZgf=tS3`1db9CMBvYww`VBs4ze@WYC$^5%X~07mWo)5`>`$EZSeZ$4I1HyVn2=tESfaFU#8X_+gmm5POjTp1iGwpnjmxz^ zAOCWIIe)w8`4%TXU7VTYoBOTZ`xaCGI8XDhIDFeHw*KD&iSiFzxwOw_-WZ%;^Z98{cg9R z;4}-r-@{RStGfq0)c>)#m*%>f)g=@wQiJdn_C8232mV^B(LlNV|I^aS7p?jK!UsaX zeTnd`qxs)H`k5y!{Ha4^=C$kj;EwP&E%@iv7XBLLpF1Di5q|mjJ07y|H<166{_LJC zzc&`#nYQrv_E*WF_eVN8;`+_~_Iq^ijD6Ed{)_Ldqw@hd^1t(I!apa!+&NL#@9wwm(gzAH zeA`y8SzsRBga7=#ramKoSU}cEMcn(YF4+AZB^>)9LA9;h$C*;s? z-6zVwiaqT5Psn`qtA#)1WYKsW7uq_&PrLSZ)BiV8l$OXwYzA_Sxa zDNX74We`_JbydAFOL?|bv+&70Yo*_jC=CW@ucoy0_0*==`r%fKvoVRoh+?+XaLaMk;ZKBXa+j>J<1XjLY zS3cS~Na0_2IaVy+MFT9Q@ZY>wuH&677ERh>w2NGxy~V@>%_PWeYDFslLf!lQZMuI6 zYh96*cDw*svxFVX)dS1gW$!fX!Ma9!Cpr9FMoyT_tF!&z5YsAp<)#p%eelBe3Bu@x z*sHKLA0pfT^DjcTzew)jBuUUhW?G=!Z%RgL%6N@>geVGs1=N}5ksHO0$pu-QPmlf+%!u;!D@%@ggKfIG+2!d$NbfI!^Hewi@f&vHQb!o^XPb9rH~u+wN8EI{Bm=3e(gMuk79g4U1zG|;#!q| z>atre=;Y6l{=Z)yTmH_8gpgbPq~H18VjcccvA?cItPjHZKI%Fz$g5IS{@OEEw$$Ov z^RHcE{uKD?Jgka~YZbrK?=_y);mh-x&2?gbao&HSGtYNM(+~K79(5m|A==*(K4ty5 zv5KC+`L(Y4q2CE2eKb01=s&}C`1yo=%7F82Ju%|D{iyo&3%YoIEl{HaG6w%L`2w`ixlQu> z!V-@zqILUT#(&ZQAN{Cc%;4Xsx1he>x9!yRk4oZvRf7;Q&w}yhV9em(zy%#Sc8_-c zrBegaL>!O*Ku@5-n8Ck+i+hiiU$LoBTwDqGAC#{=pV)joP`}jqNu*b)n{R-YUFke=ZE9BzjVZ`D)pj)hya{oFTKaugLar=n#y)o^V#IedBhaa>znv0XzVQE-xPD>Y+Q;}P zHz#@3|7Tp>P*~^x#c;MipgcX?KT*@Gd@SY0GhMzv9xOlFmu~#s7UsXP&uTP${2%m= z#3yrf_GyUk17$${wf$x;{X^Lq3%BU-Z#zWz2KpyB2QruV4|jXj)Zusklkm}>vg?;} zULvFbg;=x9{tZ)`-i~)=4h8Z5?^f}<@xO9TjCoH1wStm zI?zPr{_4}UTS@#ZrTu-uLH-+*^y;O>H$WfR*5z`reZqgfvx*+X`R<;^vH4Fmf2C7z z>?8XR>i27hME_6*wf~@|r>6pcQNExb{oP#d{~GrGZu0st*gx#Yt_K0gkY5&Gfet*; zzN=0@z9;EHoL}#YQ9qCywDPJe`*iXTWBJwk1%Je1h7sU`t{MB9++RFQKi9Z?p zaepxkm~Q|V^wLYKH|p?*{Y?pgkNy>CtoUcnXUhG-!sPFFC=?&%3+ja5LHWlEBNq4f z_iu}ZzMod_O%JN}sis%p!up>}r+zX@Cx2n#nPq@~v!12-BOm%%DllK;y%+y}wREG+ zQvMdo_oY)E`#UVi8qE9DjkeYeNKlzlKY3gf5`8dO=81Gy#hVcJzn;2!|ZdDGQQzGO!Mo} z?8j2_)A8W_a{Z0{euw_sCH8L~?svLcll+iNfw3=5`xpHvKe}y-uK#vJGW}o0SIk2} z6PJpE%}T5(*I$;%QS4`_hmyLtlH4xVzngA5{w=QjXU6UYYo&b}#Y5Y#5-h*keyQmd zI9q(^vD(-XUrGI|`-fD4ez-r5_jh=06&Keg{{a57mfbf?{Dj3czKZdytOw{p?=)H$ z;bQxbhOemK$ZXT~g2XS+_^_WR*#5m1+hvUSwt_exrSpAypVYrvzU?N8HgTiIKd1sS z4v2*xOdlMz-z)JGl=7|XP`+w<1>prD^VVo@2_Ic;V z>N@;&-2X@Yf?axR@H!p{tWd}ey6N#ZwCiiXQ0AvFK9*bjl_h-1I{Clm@(q*!vE0dj zA?}lRb@(r<{eK<@%B}tzZ2P9L4nO~Qq*%!B;`jF4;ul*TS67F>Rq>ywzq!T#_R#y< z??>x@8Y;hs{X%Z>`~R`sA$F|RX)ZQy*8MuW1VOs3B>Tv?!?tdoyL+#^1 z){pxE?w{NeTl;7+C(y6!%;_JDkMXzrSU)~x{topfAaLj}_5Uq=BOj6dPj&pWSt6x_ zUQoY~{ZQ1v71|@*dHms;#iV~z_s9Ctq9;`UrltoDs5*Xm|3jxe65mJ&?cW}8z*o}) z7gX(j4gE0jjMR@hezKzvrGQ=H{UVzDZNZ4?i23PT{rXYl|D5qv{ix{`xWEVA{EvOb z&;H|oe^KyHTl9chzR}u8EMUsv6-z_sPnO>w+V=0xL;dsv4-!7wLstv3|7iHCedY9< zdrj&`-CwFUD?-(enx55Q6N`U=w%fko-(dMszix;lB-den|7{st{WTdM#FJ^vjnA9c z)(_q<7QaAG`1pKSSDV=IEd~ZH|IMdz{kx41`7_+~Aii7mG>;9R;cZg*izgRL`%%{y zC5rV|GPw71zntZ_6FU50w_0t={_V`;Kx_$S3d`QdqN`^Gx&F0 z@P!z?u767%zy6i?qo*d(b87uHLz9PH`L2cD+A(8Yec8SwlnSlCZkAtdU)1!WXko*9 z#<#7i50dh${fFZxD5X1-%Gbl?8?FD~RTy4*d;Xmg-%{jXUx5^9b)o*OkMWi2sh|me zuIPWouE&3o_(=->aeluuKc{+Swj&x%{uTY$m zt-teG(Jp>3HFRR_eC@9f|H8|C<@cL5KGu&N;{J(`@y!mg;hQW!h->JTFLe5;eg{2avi*Je3SZw?u3$o{$N|NB7yO8w2N0xH+0axMLX#8=k` zUax;p%fIuD1)k8^&*u*MSJNxvhRqKq7rI(W=YNK<{lG3U{u9mrtNpt+m#RG`^`rI= zz2B2$=7Z#yeI1ELlOOy+=UiGY{byMJXZ2fzkM`Hc<&1ekRWG;(F6f+%v$gBrvU&Uj z4$-xh`d7KdUr=zZ)PI=#zj6Br`MqVSos5RB`lr5+f8RpdPm=O}x2YrRU-3U`dguid zdIA0X53`L<{;P`oXeVJ$z*8x3K@Vr`+M~mFbl&{&~`qp2a-S+l09sXU){Id6%*zmpVz{S>-{?Mjy-AMx9?*Ekkm)gfC zdH*N(v-JLGc8#t7<7GXe9q#zm;Dy2X=$|~x`#br%5Uwj$d`140kAJX&@e^78X}o`v zm-lb-#fWdQ{Gi*rPn7e&wtg_ay7v@4=$g;{2hIoVf?>i4a6un&b=<4N@4JJNy7N=H zn%!d4KX5_otWM~y!%qeNDU#oHo%r5mP38V6Vfrb+`&YRu@%Y|#;{U#G>YY0LFS?Q; zxSA2Z?>h1Oz4^S{e<@7<8(Dtr+humoo&Krx3}fy0J=W;HA9-RbJ>h+c?LU|JDBrAJ zo&4*gzz_Dh`wO~bK0^4O81cOfg!+5@z}|<0^@I95jP>JVT(d{+^kXrf-jA)6-%o|< zzhoXNUvCPPuj@MTFT7J;uKy3i-^ulVwB(Rn6rMa zEV|fOw!iB5`95JkzP+0FuV@)?Gcc1`R@!`j|G8}F_~+1h;{D}w4Ze~-6ir;2p?7*0 zJm(*O$W;30J|VwLhp(g$MKjdbQ=?7YA5(($kMX_wUs6i5I@yoeD>nO7*}s1Ew{^Pr z_ZxEv-^=|rSB&^-@vLim{*ul=wEB_oT_*d_>%{N!O64j#{Cl~6i2*J{TB$Y9C!O;wSi11^zsf zc5HEce7ubh`&j`01O3N{?`3HC&k`#($o;8peE83=*?)R^l3sk*i9e|Qd(Z3e`?LM~ zIupK`9Gm{#EI(+k{>Kgk<70g6RJ0Wj>ioN!e}=OO z^bXp;*caOMV|7@5$cy(n55G^^fXWZxf}S05S-byb8wY&6M=G%PC2&FeHe5PhXP-;i z{%!l@@!Q@UmLg~={NuPoApgKORrdQws>mT?Z z%XYhpO8dw92l)TD#ra1vKp$A&W%}#qff^I3~ z4Qap7e|YmKJ?I-l?TdBC^~pc9Pb2OA7zZ1Y{HVX!Upbom&+HV@R{r{@jT zk98y2Wp3A3&waVi3LSneMSiO$=|8vlqZ{6tuEWpHrW`{*(0{D>&{yFWaX0ARkCr@2 zPgvWzzTO#||M3Bs>7Vm=mC@nfwT>QePh>$NcYQ;B&cR{*KiPxG;+L4*w|K zXy$-I;{94(|NjpAA0MB~V{rQ(4c}y7*y--%Z?l5s$Nqj*w$Ov{JH`Mi-&paXpQKgi z<^E%~eL_D4_7lF3{hAdcK5#i3#;qR>og@37>iV3~*nfTw$?azQi6+0rz)4K^64uP|5YOg{ix*!IYHI+(FJ?;Ehx)3S(zXF#-V&^)(vRkRMpS^6Cd`G_^N&uu>VB6 ziWJ>v%Yv);(L_wPmhEAH-xJsPOefVs{a}r-& zACm6|vJ|y`Da(UHu;m5x1KO>5kDG-3g!NC>mm_?;{hh{sx*q$}Fl4#`T+q@vgHP-5 ztFipR_vOU)e&J=lSxg^({EeJo`$zjYkL#~7nB=yYLM~1oM(}^oQFjffqLY79E!0$Q z2e|%5d%v(0j_nRk_%m33yq~sU{g`#Q|1gB|0l5?y`x5w|3DpmU;G_NZenm;l1%&U8 z6`!-+g6Ut|n(qjf|01`K?q!7UYry^T81dEiYv?aM^Xm53ttJ<;9}w%WZX*2I>=4mz z5clxE_&;cJqwYog^7H;VXkRAGptLv_$~}nmk5UguraOuJ50mL9gWnC=zoGztU!TqT zf&AVW{XfbdwEW`cLv;4}DZiiE>k}@M{L$hU)Z9+Vd&O z_9acDy>cxfe7A@Buzpnnxj{Rey8kJi{I^yQ#mD-JXzv#mGqmZZZEyMQhwt0<3+Lm^ zr~2i+i{yq}{CF6-STgk1#)bEHko^avV(9z3`yP7E#p8KcpF9i-3xx?>DGi)^>CRNS z{z~27=0moBj7R#opNRf<05J3y@(1c&{AN|1{uee@^v~yC<<>r{ezvlu4*wqZpHK(J zXRx0|sMJAM5S;0tjb}H89Kb^m32}->6JqL^ZgF`QPV4MVfuOG_@M#9|1^eu zLjJYqD5X6vAEW#^3d1WKH!hI=KU0za#8`Sz_5Zgj>D5cvFX%tVdTbZ^4|~5@T|cyZ zG5%?`j6A-owm%QOwx)~p&+7cnJU-vZY)bXZokH~seoK`b*Gc%n^r4KqcgphhDeu3l z9m-ctufQ3ll^e4kSbR$A-=*O10zS9zXn&) ztDybU|Kp#m+e%3NSd~NV|JGUboa%qn^r2|tN*RDy;TiO=eE&^X@ON|l_06RIl`)*k zH`@4v>OkW*mfk1v-3tC;2Ye;_VdC#r?tJ%ruXm*WGZp-I$I*kT{wpZy)yr3u8{1v^ z?v(oXDENCAU)8^zz7w7N<3IiA`VXqAtB4bJS?R~+LuCHd{RLY5!k+ebs_$+tzc7F( zH(u-?Sz14me=$bH=3mt8 z{Jx>*a-DuYZAg~vJ4W(bjIXATc70}CLwSBOZZ;ak2^+u=mqrP_8T%~`)aNU z)z74N3E#b+`e>&pB6{{jUCo%!(a7zGmld{^>D|-fz7Osn__%-E z{Wv`UTw`Qx{WFVkLGRscYS+)*b356;m*ur$#D`u$8{X}ytvmmzSdBQteI(BCPNQ+P zSjR_Le49MiCqZjJ?a#3MsiYTU6jM8;{)d}ygU)y&t)SGurPxo+rBtropQ&6e#*fy1 zt6X0_-*mjr{@X8Me8zQ;j!k|Qzuk%<{{1~w|Fdo^y=Z$G*BTQWzM8+qoktwd&CjK< zeVQNe_+T30!!A@gajoLV%|6~lhd+(+-Hh*!5ntu+yMI|*-Tdg6|B^($he&R3jQAF3 z1a$kU&&vhZUyQ%pjq^2s;`Zg9*z|+<2WDd&t$6GY**{BCzCWMbgP!nBCtMGYk0{+2 zz}VEyzd)bfcFsS4xs>f6>n~l;^MfVX|Hmjl$`|x6_wDQa`cdA`cg$z}66`M-pIc@A zXE?F=7HI8d56JJY?D;e3zZ~zwVb&!57-M78KkNr|`48TM0zd5i;g1_3Db-^)?;jfN z{SdgIdmg&|XW<{h-mkjWC4AR%lG_`je(V=G*3%C|5^GEQ!T1s0|F@%mtn}}regEt$ z$iPhQEB?UW(!Y7Cg!aE55ccW*GtTfaKGq4`3blp8*jG#KU-aYqriF4OewKp&ym%4x zYVf1I|94RE3srq`wqHNW`_J5=^q^Y5Y(rCUaEWpSy(RDCy>#|-qq4uGZy~jl(Z-L| z_s@nUTa5S5|0(;M-1Y}aR52)`CYI+6Euqx6W(EUY+7m)Uo&iGyO zlKf@Q({o;y-^=p1Pyv_nlo2ES?>~zC9UbIX(<^X>(SYwT-CH5=Q7L~WLa*O#|L_Wb{Zi!b!1&e^+&$Q`t$#HplK$=Y=g#63=l{uRvJ&Wja@KC=V3f28BhxZcM@+e% z1BAQO8==45UEG3IF#MmO>wYJRVokveLeu#koWsmlN}~Tphp8DUQ_?0*sLMx) z_nR^?U8EU4apE|k{O%4XLMurVm13)J+BZ_fk>2AviNo0I#&wTH{d9P>Mt3198*21$ zvaqmov)fOcFduzcK*nO{U?;u zG1-yO{N*53t2xEn4?)7Gr)Eq}n>dpxfBr(?=U3&Q{Gia!XU_Opg9;n2718l)p4z!k;1^NxzkN^0 ze^(wup4%oAE z-WO};7x-cJzgF!3X?+Fx1NMpXHJgf+d}~zggqQEBh1E9dt|9QzTLjC*WjYYSF z;zK{~A8H!rYi1BK{Z#SSZ~OXT9sZ6D!vARSFNuCPnR$N$dxEv0yTd_i^n^G+eZ z`+&3jS-zTvRk(VX{Lnl8F`>iIbe;WFeo1`O9RVMC01haPO5(~e#{amdT=Bel3#sQb zT4H|#k1t$)lyCaAnud3+835v6s{Q=bV8z=y`)Ozj|M`nEevVPgFxN%LZ(jDqlREs; zt;G0u1bk0iEyEK7{)N}AT{`^d#P?;^ug>xtc?ABe(dA#!u5mXV{+;6cWlseB4*ouUldk?A?JPw4J>DpPZ>aq651C?|EZvqm@I$R%AD0-sNmsvG6&Ch+EZ&$p zP>=8FQB{l;h^PeBKa@P2Z;8%7EITUv^KoZ!1vS%$5+dD?yB?Wb>&-kozTyTaD3Qj zlGOhnJlGUahuzOWKH@43pRKdc4DtPu>twv~_M@TY3w+PGsz&yx@bU#NSelEPOc(kI z>tBr@EcAam-k2pEZgBq-_UYuGRs0rDzA#Xy|8>6$33C2B{^F9)Wa#kM3jKJ_#v89S zm(mCNC(w`eM>QkiqiRrtO;!1)J(wc(6W0I9sV*$x?|5V3*iiidKeJgyBf$(WUlqT> zpW}Aw@QW1{_4h(JKJ@QuSkcHS8y)|y5w|zi;a}S&vV1AtcucI*(duVWf{{4^-qKc& zIz3?Af4K7o9e($ZMg0=^&pP0HF9`p9D%?M-`gzNIVxkUzlE8OgjyJYB;3styx@i|( zKc1W_vvl~$&kO&5HQrb+=4oV|hbG2|c?q9*e~H!pn_7=OBJjiBzcNMndj1vfL4uD? z|LMP1HC&s+>o4Sho_chCx`xg_Y?Je!*PQWKOdvw#l&RUV%P zzVRLD$HntyvEpaX>Lk~{gvr0vKHrw)x7VNN5`XC@tq$w(KU3xZ_WI<%`edosboeJ2 zKi*$^g{^j>br_frvdbY#`5x%8M!P?1LK4*!;2SRf{)=g>_;rk8WpwgCoulCA@H{uu zSn)fbI9XDM-$aq$J_r7Lo`*#_N72ubfuG#0!(Z8!MqmH&u87D;N$y2K7TD%`M+QA^tU?sXSNS5U*9Z}AJhpS{vWjE z%!0)Qe(9XdW`0Y@_g@{|!Vjkr52@m8tS?aXtzN?KfbL(uO70&RrvKAT2^;lbj| ziS36J2EYYPt9|@09eyjG=tuqX^dbKcO+UZ|E%wdZ+WjA{9wxznkNO1aq<`RoHp?Ea z{XXbx`}}V5pSE6RQaQ(}pEv6*y-}y1^sc0T$nQ=d{|1VDDC)QSZ;=1@VUKA2Tc;kd z9V%hy-S+^=A8Y%P*kP=8|ERQ1q4Hbj*nj#cAF<-kduWpO`@8#Lv{G0&V}N;uM@vzuYVuY;lI&66d&?? zo?<@~qx}zG_0%FA{#C_4V}BBSrxmMy>fBmUd%ke}ZlUtS{;_{otnF8q6{ofR6NA5B zga5(%3#jW;?vF;X&sMKI)m5jTuDm~$jX#p=t;sZ2`FCUuI;6u-5#v5G!2dvQ;DajS zhqaG?-T#NSeazP}6d&(j*vBi@@@>)o*LFJjrzrhr$e&w$ceVSp`S1hhc4ghvC<)i6SGQ>)pB-6iCN37_XI%0}B>wf=w z9skmyz7wZT7%A3zrHe&6K}&Y}rcRogI$16yN|vkW8V;X2Vr=S^|IK2$WQRpqwBD$` zSxyqsYGXhP3$KV~S-~A?xx)P8)@N zpJ@AsX{jk=8zxWmUk}keI&`SfFL>P^EjH6VSwEMcQzxVe187JjWEr5X=K!0u$pf16 za=XCInG+|ZirZAE=KVvBJDg0D*Yg3GjMH9_W-qw&Tc>@S)~N~myhB;R>R;~|5FQ(oC#gN7bztH1Oqr#M88RXO z7xdaY``*^!ALb2=%~O%_ed~)FG9myMbo=c3t91A)czoB}CNh4yz?KmKxS($(m29QM zUoY0*$zUy^_a!I&WDP26$cO-3&=RX|d_;%;FYCv$BFk?MDQd`w09;U)@9)2L_;P)p z3_d;|Cek`TITu7v2i7+Me|@jDck1v*t)dj>+mYqBek*3khyYyB=Sq*+slzWU)(^_y z4Z#o0V`Z1puM+|;XtzTLmg(@Tyg~S`&B5~L9)DHqbDebfrS}lNvE{$vuY05aJ{^7{ z<9q%ieuvY&Z`I*9V*OZK|C{{t9$xdT4nN=7~7|3IqBaC-xx2DWaNNfpq>VM z({=b~;eR@l{*%k<=eZN+haIs0D#c>6|1G^LJ)^^y>#t?-aQQ}Bhb`A)BS3D@ljY70 z(&5YX*Zxq2>?hJXY`GR20l1*I^uOa}9sc)1-!fpnO~tR!N9aehP6D`~zx5ivTZb># zcgx_4BEMYcEh7SOK_B?6&M_VSlGT*L%kz2HD}RS4n%B|cTb~lX?+V!`;Roy|vs`TL z@7;wve5u2q%;jq((Rihbe_EJ)EdKeYe%+s9T_4Oi{tH^(<_%+19k;JtQ3jZ&n9jMydm%gLhrit}6Vf#Pj zIUyeDpU1#@1&k}O+cP9T8PFcnRe1Y&JnQ8*bn+*&{FXubk?exkA?>(9+z|TrJtMv| z5fK4;2Q5>k@aH=G=59*i;d4cu@N-@%Xvl~FT+o!7>j&uY|Kj@P%0u$2-)W?arDd^{ z@220rDy75U`T|RoSC8LEd>SC@mswwwk%)Fs%C|t#F)xbsUt#6@rl?<@apC80n+zoL zBSW7hzC;GID*1m^KgnWvFOq&JwPD9i;ru#Q&`cJsviGpIG;$#X9_{eJO$GUYd7M zD=b-$@iKbR#K?A0~g&m10eG_%1%5*_y1!&#V`l z|3B2@%Y!=ni`+kPcO(C%${#C!xw^m1(&4{(f%X4_9zQ`W-Hb&)kFMwv*#A4i zKTG{2m%PD{5dpZMJ0`c)`p>QjY@gl9ehv$H|qYK#2TLRy2pZkj&$s!^^Q=q?=IGn1( z|4!6T8LXFr@dNrn|I_-ZxFI6~a6#(~TKB0AzrFZ=&L4Q)h0}L{=E35Ij0nI59agXP zDILE5`$Cb(Q;qzC6TT(CGZYa4xS%iowPdIcf6qgd!aGopzejYbbp8`7?t%Z(nw~qv z`f*`x6fnLm@m=EkNc2Bd`JM2A3;N*8d3)>RziTq_a}WA&@>d@F?PZ<(65ssszsvvL zj9UxnF`&4E@YpRdwl2xwB+~~kLt>I>PC{py)(Fc1ML^y zuZ^;`41Z+7fos$EZq)X#j&c36QuO#}AARd;7&0Q*xP0TOF%Or~$^RwyAIy*R_~<`) z%hfVuL{Ra6fBzfp`0z65ry})#RQrK`ybshcWJFN$m)yOjhfaQo4#VgzPdRsQ725chR;la(@CDJW=GA`xN*iGf?qcFLpQ3=|}EQ;14w;>qqWWAR~f` z|6S1$>vj0gpnlyPY(D}25C7>YD)w0r5kbYj+xuus9ex($`##b8AL!qZzAb`(`}E!S z(scNFza)v=yY=`OpZAs${!K&#RsP=xRvxLt&#A=v=kwuJ|NqZas$b@m8itGrDt?Jq zADXMfm-|D=;NtR)v`>WGD*{2q@0R-g5goqVA3_Fi6nwc)gp3F({-f`IqwOE&6Z<#F z;2lWzbNoo?`}?waq@U~#wG0^%RQzA(-P1)UzjraEaKA?Nx6IK{{QFx9-Nn#9cez^b z(Bb>{w-78$K0h71#UEU~L0=ud+@C`RH|r{OJG4eMV$NQ1w5(?_q8E%Kb%T!1D*mMXUGnMV-_P>n z+;$cJjQGFVF^~O@=Y&7@?zP(SE4hD(3?4ogTg8WdieCF|NfRw3+wc=LXLhN=fK4kDTCHqnL!NK~m%*V?cP1FUB7XJ^Wn>AlmR40GauPB8#U5^j@vG$cSWJFN$ zmlWRejSfHkBT8XSB>PnL1OKD?2jm2N-v7C0Vcqzr#P^*FjNitJ4_wgqH~)KDC%@d^ zNCq$4&-LOrZvDF|@6`NcQPbgEy3A83Cc+Y;OO>Z_$b zZmz?B2<6+7{IiPx!SAv8|LvpZrs(j$Pb2w_R>An;{$I5plq+cRs%5`v@!7w5vdR7@ z-xKQJ;GdI@ls05U04`{m3(GQf_-kiS3hN!xzY~7e&!r6+5r7MNr>lq7e>ONqDSTg% z{7(B=8ON&`{>Th~3tH^Q58v0y&;HZR=QBIu%YCwBbPAQO%x|k3U(MPt>~FF2_^RY5 zL$v<+7dzK<(8+&V;LBh(C;hY&=MhOn;I$JY%Dy^z3~x!(AN>!DKhO(kdY&(z)8WrO zMk(C;BIBo*G7T9KfD4-3W>hU5ev5sS!prs7N&ckLrXeE&a6v!5w5FZZ{T!Ahj^Rq@e(ko(-ph)@|nn6@a>be9|-NKnp?oi|hH{yDKekIy&4 z`bngh{A7T?psR5Iu(sr+2XykM@%fzAtz`eI{ODi#gl=R+fL=g1TaRe#?*yC=bX#Qn zgqy`agCZgT7xdS6%C6JN|0;jq?`aX)e%!)#WJCZi=;(G!HtX>H`wNOh9xmTV`wUuQ zuR#Rhf-d{{W9|E?++WZi>XQ9L+GkMiHHZLQ&{9X*&C|*M-Z@I)d5`ex{2S_j=$G_MmHvictuBx_v$^8*!aIt*f0OQ|Hj~Mg284mNER182}fw#*W9d z^B1M~`$wZW`Ol~Q{gZ(F!1rXAF=RvlE@;vZOSI$jeOr?M^!`oyp?xX?_^*nAP_t5W z`z%ysY*n3p*06u`UifeL_rBjczYf1X>&JT=jXyZa@9k99kP!iT0X@~Vrgr^XN5=QC z{#E>s8j1QV&ftvpesR3jdx_J5Adq6FTSRKHaH%n|mFvXvC_8+q+t^dqU;Iuf4hDrDF_r3`Y7fSrhQlb9ysPLa2f$x2g_0Q+4A>U5; z_&@0Rj)lkQ^q;VjByfK~_M`68y09nd$2+W=AtM5CK}&u*eyDiAYl8n{eV{4ueJ_zu zmj9FZGMN6q;(zmrFZx}4+x|bU@4fHZTKRp1rJPUs&(8vDZJgm@{kvLIJA>lu%q6Po zA98~>y5!dSpF2Kb3Hbari5J|yOs_@yNqD)0AtS$(q&Dg{9Zm^@S-??O=?2o zd*bpKt~KFrT*5w`)~AiBLG>Y;eXd{ihIaga$a#{$`!n^=7d;VbpXfiN zw<~SPhyc3>J+l7Z=XLtOjQ5xMk@0iBC}qfq09?>+6~=47fB%Wg*L5&5zR_ODE+Yrv zg8m(!rk#H@c2ffHT=Jitw}~T^Xn`~2DM+W2zv_lKPX?TdIdv91QS1)*76SSqw}BJMSs)vs0;XzX*Jf2474Y`GXUF7X^Rp9fkM%>u<@>_OX!IpUkJhFBk?U&Iy{7SsPIQIJjzVRF3myZ#DZ>@brboghv zeZ_ufuJ5RPyo{d+`~umATuOidxk1}+%`$cP<^HA=-p9y4W8W|L?}~onmPXXij2ABV z(Bc2^M<_n_n9;0E|Pw-tBFwr zjs1MY?FVqZ4^V%~%lLNrMv*`3ug)*%@H?~qAwTy2cK@NsAEkT`4!`MX9e(m2(m(J) ztv?m~DEhfJ@VNs5Kdk@xcmt|m@dmBOyhn$Bg#BY4e{W9o zCAj|&J5s5W{pXaQGeU=d2iITVLq94muAT6Ays<>;C(Qp$SM?8loKW;1rFIwI&;ocYq7p$6Wu94nMvo5C|v70e_SDw{@#C{zpkq{;I?8T`v?r`6&naMc2w# zz!~4@-&*dk9HyV@iu_p-@Z~x_1lT?3k_DL+1%8bIiE% zagqGzZD*=|p~Jtrnj})~-Z-5*`+Ty|cV`8Dn15Ks<74m-2A?kx#!uO25_$oh`b|T5 zK6IFV(z$$r=Q_;oV=m>taB|~r9exr1zT4-f{-?daXAbEn;Tfu5c?+b6)Gx>l>h7~x z?r$9?e<}X{*T?fCK0bfII+Hv6?StFO{guP;oAUZ?t2p7={gbnFU7qEa{U5y3ON)Tq zpjTYwAJpMDdn?p_V4q&bx7%OQe~8rn{**akjt;+I9jd=5UmxegeU9X}@uRd~zrQf_ z6&-&0c)~Z~KbuoITYo9~FP{@pKYib=_>>O6AjK;VN@`zPl+TQhRDX9o*rJ#Ye--b)4t(#6glk+>@LOc&iG*Ks?d)Yb{C~KA z;+oInWBIv!8Q-p71tR;mk5?zi{z;hsT-%=YgL4^pA4>Nnl0R1bb~k0q^KHWL)7Za3 zKj>##mlgdKNQkKa8O_q`3;eM9`yKBu4*P-KfVwAgc|@tdEB9Y`Qz!ovyeVJLEaG@v5pm7LW{|VU<`zHtA`}!Ll{$=hzczJ&k;2Uv-Z`ZGctcduo&q^9P z{LTD*8NS~KzL)Vcms0x`m#{P<{;k;sFYEBT^_|x(IBjZCqMqUL!ihiD&@OLX6{xZhL z{sX}GGX7H5e|&-`qW=HLKK_;t{}si*d3v({yn*Dm{eKky_He&7^8GzbKY7b>{pImJ zUuX88`ErM!|IO7kb@(l6h2q=i73Ej(mG=wy1<<4G^UC+%F#oV0<;xBtd!^n#G`c-< z|2%c?!Q4Gc{|e{XI8i73e>Z(n zTZdoZOej9iox%Aq$VW8&Os%=1v=0AfK0gQJpTGmf`#i?Ko#fAZV$BZ%Kdk;XKT9dB z?V;_Lm+jN`&r$r_j!*WK)ZzDJeD^fQ=l5)PF|z-x@-!Y)IqqUa|Mt@5Pv6ktm-vX} zhu`$@`Uk7HQoo}3x3;gP-K)bdEWQts!NcP}(6^C5__lwJON=OgLbxz7+V+ zlK=Fo_yt@M@vqe>yUKSP{Bd|>x@5A!sk_C8wus&%o8_WZXj#<%?g&tCu+e4^p6 zD=|-dKHCvKKM?iTwo`5|t--O$f8?!%-*n}>vT|tqVhkqxF-qmm{uA~*HCcy0?p1&B z8qfpe_cDI0{ag3oyuCQ&L+7qZ63Gw$U{oggZTpGx{+p3np@vTW`pWzW>H%ojIc)=iPa#U1FfuJ8b;K0y zkwv*Xg=c_rxX=IUfFpsMG&ES||IMjE=$C<>XtS-R?)^WTzi-2NY4^?#eg9utL;wB{ zT+r8E>+z?~KC0%B{i)}&dD%Xa){uRdFA#bC!K6!tkL&QKe;r!?(NAz!QR-*BC*t_{ zd$n4~`Gc_jS$$sLV15wVKeMV5zTN+b@_x8Dd9HSRup8sM6?~&w?(pv{GvbO){(`)| z$H(74z>j$uKi2+b!5N*-3H&hsvN)N_7wtFJnZpmbOT^Z`UdR}Kr%wJBV!fFRsPD#4 zgy%LD{}N^V`uD5@QvR^=t&8@pOuPR_OcFt05tBS80H5jQ^rk}iAJT37N-aLFH$q6q>V}&kuoKf z?~E9iCi2rgbwsXW`!|8z6@C1f7tqzx9q4;}b@Gs{TeGH-MYWWT9tQ&Rmb z_}NI%wiyH|8cK^v({Y1>o6asNlgEeyBva)nf&cy7y}<#6bVwbMHaTtL1e_sQ^CE`?8%d#9@>EgFXEP8+ul!d4Jr{dMQ&Tj!zq59|mck<|xV%2TD=}9X+w( zTjIpp-U$I|Og`rU=|E-vWPtwYD)4vF3pEWLe);c3 z-ZKLD?(4+A`1I;?I{ZcNi|-dE2k>!DML;t8rOJ;Z&i#l zto?!di*pjx^zY1}JCIxb|A5XDe$nCAxhU-OJEJ%Rn$Wn zz~6rn^e6sQ#bZwX(|kP0{;w1N-s8=R>g0c^rO8Cw}*hHEz=3FZf5uFgt+ny-xhWpWmj{ z&rmV{;_4l!E#|v={m6Z>WN^|?$Aky=>*OCL=0`lw2CTvDfPd_(Q2c;IbbrRiZwib0 zdxKhk{}SWd?s<8PP1WN<>M#6f=4;}+-)Ys1xxx_#*I(oVbk@GL$vXejc9-z~OX7^J zJdSGdJ{crc;P2Mzq4EP4v_tC$r|Ix}i2ZrYr2%~G!x2EHUn;)(|He~S9@6Pw?(gBf z6!6am@57;{m-~Lm0D0*u@OR08&E<6Xa(@r=asc0To%o-x$n&lae~H-N!`tx&TM(}U z{`xWW2;>h)MEBoVp6C|)dlXdn*Zq8}u%88a47qPbu>bkCaj3n(|DgY-;y+jG+r~Ql z_k{dL=6}bZnR#}E4nJGqyRQS^Rw(~)&3*2$PX7nK7UjD-P=772-aq^*^C<%KPgjAz zBT`(S>F`ep|KJM3_r!>geii7Uh4W{N`W5zmw||=`NAdl>Y};kP_xp(utu@U%V0_T? z6}|5Ta!e2Y)$Li?zK6A68^rzs#-cpNW--qnY(IaE5ABbiT_1W4xgkc~KKHd3?R!4; z{K5Og`=k3=(LZ@2w0#sSVGOgN=r48^`}+M_%8;M#c*H(EX1CHt&M&3xasZ$Eet*~! z`U!hKT5wQEu+$mf*P*m{2M*wSI|_c{TGh|6CO59s>F1;vzjQBi#&@+9^8OUSH^evI zK5-4XL2p{{)B#=jz9znZG+zsp?{^1Cqlv6oeDC3!N#nAhzy)p8v;V_7e9w8&KE{9# zK3HFXdH_0TYetH|59{AH65l7f-ikA3XHoywvr}KcW+zdGfCsrjTNi7WM~v@=+5h{! zsig7<>y!cG_cncs>|I@_3|!E9i(1R|hhg~NRwLZJY!O)RZw_Kytn02%&XZ`TbOw{@3>)`JsQ{M5E9P=#h&5%K7Cm{7H)b!`5B#dJFYFa6#{W zYqnND!&v{2->H1nbr!${ZL#G1`#Sl5Ij1jQlq;xrAg{aNb#}l79rk=V`TbOwepV{> zgYR{qe|G_zZ^AXo*BC}a{8Q4#r#9>`HN!vbTUQRIr)JP_rrAERu6b(%{|Ku8ss7^8 z)@I+-$*EHkm1HzZ-*3vqbdg40h#}$>IYiuW_|y?&Q>XktjT8@$VWc=^{Zh1_p8j!0&Wq%~U3`(ose{;ZC-ya&4n?J<1gY~B3-QQ3bG!8?We zQF0tRxF3Y~FzfZAhKyOIM0+R3QPSTK=bek_6p(MwW*-br(&fL&D$)Ln@%tiIN&mi2 z_4>#7I`m&^nBXkVg>VYs3;OuJ)Ji)3IYR%gApWk;V&h*)=*bm|?vZT|$H?fEzBlMpwe#__-z!!AO{`SRm<^QwLzYNxy$ozeRmyD9X{0=-v z{Fc!vK<}VUzHak{PXFdp0&ifP@yHx5|Gj$sOTEZ|^3N9iXP+4VcM9MO`n~((7j*m+ zJi`741^Bx@kBvXxFJtAu-eWWAnYiJzb{7qxvl>j*F3XPSN>B3e}Ay3en5S0rAX&6wXGcZ)49<#p1l`Ta3g{uknwuF>)LkFQG}Jnt#zA$9rpbQR}6i-`PZ ziQh6h1=t;Eo&t||*V*5r;=RWw<}csb!0o>{&_cHTYWpv?&-Vw|->h7=KL@<$*Xa10 z#p4X~?f`#ZjQqdMW&3mTm8shO&;0YNB4?Pl&CNWkZl02~6yE<>@zlC)y7E6G4NL^< zLF#|po)}vG=znA%DQL(D|C97vK|@BT0J{g>x@n7c|IeJ#RMMW+^#0IU;qR7BoH&2n zI#$FVah_cA@gjzdP62#DTcj7*qszaA?+dR|{ZsiT3wh)`w{=?Z7g6$e7Y{#=NtFY9 zK~p~{R!7G_&k0It^&-1>;-7i0h#}+k@_+d6KeXqMFQ`EJcl{eV|DL~!7&6AnAN3ye zlWwD)*44kGeT45x5%ZJW{S)e*Q$bIAM1kt{CkBxCVL9{BlxKl`3BwaQ>!Do{GSo|H-`u1|2w_^Q}e$=4(SUy zoC5Lb(5CyOcuT}$e? zU%FBN$Tx0p+`hSsuKfRbfu68#p!P>C|K|lzF#qwzL$86aQt$`O-}B%kUH;|zKp9*i z_8-W9Zr2GCT6q6^eUp!$(DAQ!fl^p~g6%KBf9HSHKl0zR_a^Q6#2&7H)?9u4+u1Z! z-*f5;U5Wi}Q6|W*O2HTR@7eL?1zq{Sl*RdfN?-ni`ic6^>prs^#2X!E2d>k_wNn6p z(0#9L&e7@r$-ac(zLVA+tM+$PIBI8VVSCRW(U=n%jcv~Iihy>$FPKJY{g-tS|rKV*F3qTJqI6 zL&h~lLhX@KhyR7%L7$rR-dLUfrwRSLX2cnXU*++yBYOQG7WF^|=zmRfF)q3xcK>&{ zjrRpz{+}28{h<{3U&$x<{a3v58ZI&4fcjQK4Ef==q?|g8K<}W>Rl5AHj{lEyD5-Y{ zwSRIQwagXtCjGB#V1U2$zrp?>j4|Ipde9z2did8ztNs5gZwb2!S|9B@s@Ff(M??Qx zg&lZ>TuuS`2EB3H5bgNyPUzpq@821r`rjk;AFS`wVkv9zwG)5j8+1#~H2M8S*!*kJ z4WdpxAl|=teYo$p*z#X?ve4T`=lp{&=*=JgQB%i1`x#Nb4+i+Vevgg+NwKcooeTaC zG%43m$N!tXBL6}BeKGRS5pw%-!N1Auo95~G&;CW=1o3wri!J}pi*hpuIqM&G2RgS| zhnc$iC)e-GfbRk#t>c$#`J=7lC$#YX@4J-}r|I}7d`>C6Z2!bR(Eg#F@f@yc$oP*i zC^;U5IskquMZQ7XoGPh3KlgoM$1<3&k^gt%4}RCn|IU4HT+-!#%SGaEtfu;}^2az! z=8+nPj2M42#cvs%0`d)drK6=Ce|Eh@2wn&N7jF{vL+CsGo*Mp$@^Fi58Jz;^J?Nyf zqZaGzKTDLqXHJ~4XbZQ0f9maD>O}_l-*cTrx=qgZ55A!HOn+yAj{o!>qU;|I@b|^Y zzvdBYhe)dMztB7AcP9=O(UpI-%aqo;nA$(5@^>ZFGGs*gXN%u5ItB0rEj-H9o{v18 z_uuyRjck9e;)1`3Hh=LuTSTV-zM#_+te158_njsLt3UPsobqoL7xpe9_`8bMFl2NJ z;0wC&{B7F$x1Z;KT>bRyuP}7hR`^2I>jFo@h8|P}rABM91c^&xM_J2`)M<^ol?-svhbPC8fXpxtS zYR5l+hW?l8^N;l(ng5FXi-`3P#x=oPMu&iQ2Xy1YdvkQ{Pv&CLjs&$ouCuYVf4%35 zbj4$D|GvpvWw?%ive@_6oEy~s#mFCi&>{Qw`y2BAY{pFaeQ?K1 zD*w51ydmg)qltL$k0t-Gd(b^kq(85-|AZx?|2aR-SZC5cpsv65_P@?Ua-#o>b~x7j zPd`)Vv@ZYgj}U_UEo%QHA6cs)_iAZ>ZN>Z8;r#x1TNk?a{36C8MRW?#JLsNsH`Uki z-^KQ4j->Z5^*iP)@%|C4@2x+EUdPJ+aMy7ybo}e_{*msxsQ$&uKdD+JLq?~7e1m>` z@WE6ae`_Zpct?}{5&wYxTihuAFV{IwuHugy#k{k3yjcG`t%^U!%K!Mn&$Z{n-dl_N zUs=KV59gmgql!3RwTc1$)|@Jaj7|ae16q5tHAH9skN+h6%Yry#>JR+>cS&#m_`dvl z?f-?=rQ7QCFTbCc0pI1T<&XZ4|2ul|?L6Lp<#+TlIt7$FXoE=)FVOK%ElK+KZlwC> zRQ|4gRSX%Sf8z(Cdl?-9_=1*Bt*-UIJ2w-6cT8maGj~-nWCZ{0-Bk=3odWoRUU}rv ze7gMqa}(v?{Z?fA%b6$aT}1FtSXsr8(J6p0Xr;yd^6L0MTA2B7imd;fb%MW$;Gem% ziXo#@0AJ9}15;kn@gK?l$GAIk{xk0q{6qwQ@AxXh55&$t&zVaP>iECQ`M1&{m%sbR zDu#^U@B68W_|7zT{@sRcT&Lr|N7R2AywUi-DENyA{s}LO@)yx5K<}W3^5@^JZ8^!)Sd#V^Rf`8`cLLL#F0{DUscxKg~I{yCsMFfb4_g{DNf1VEoe-Xjo z*eQ6+=n%jcwAI*p+WD`|(Er%T<)0vU$q4?|N22``(J6p0=)z$I4(syYy)pa$8-x8% zVEh~HkGEzeLq?20B-IlA4-uUL_<}BKdh^{n{&N2nf8c#lob;cwU-&-}?ff4Q`4`bC zfG_ABi`#F~@%Qh~B0#Y3ixdCM3}J5~f`5+qEu&KaU(j|x|F}=bKYtn{#8dgpeOiLw zKUU=n?Qf#}&lz9dkkKjNKj44vuWfIN{!ckPkNwpof4A75#^n9!bH#s>4F_18f-mS( z?MEc&^55kvN@6Y3=O6ZGWt2B$l*1+RoKvs(Eu&KaU(ojLX8f(=pSpz*%sI6FpYk8* zf6Dbw^8G7wU0Hw3?fNtD-&HGTwT^$%GUmUD+Fw=wb43R>`2E8Y&j(*SCcIR1V}70^h+%xnzLG)$>^Bzk$Et3;NC3r&j9t7nvd6Kc5Nk z_r=IRx8Hv~H*K+Y{dqd;-?f3>|J3r|*oyS+DOk>sv7(SS7X72#L5F*1eyuD2-d9Oy zp7)5q+(#++{r9nkq5L-8L-q(60vZqf1-_tzy05!K$3Op8O6NXI{@-aICf}>o#5woj z`!I!a4P1dQ=;U>S2k7|A{hwsOzEV#7hQnd3)(;R=yo0dw{IXzw`P&u z7uXf*e=+`>bfKCdBgQ@4=f!UkodWoRuDWnYJO5q(aYh(T_22zmDF1iF_z%ij(yoO= zubud#-huvDd0$0c{j0T4^nJwsSnb~D@gLsrK&}7n#QsV$p#HZjA?&bD9w+wp_zn2` z7H|JS$KNIS2koEbbKt+VI93u4} z`2G=mPzIO8|K&bCxYlzIUw<%k#~0mo{If-Q%V6Cc%skM|Zs8GRpX)><>u( zhPb9EQ-=rS8}y9|HMQfv#s8(W=3T^p;n-07L;LG$TG5aZ{Xer=MMFlXfN}?YVe;@D zy7E6E>YHbIoU!mvYJYr%^zDz7O9qrb^g?&T|AH^*nM;4x(DC;y7k0EF&KOgO?Z2>| z{}`^P;Q#(i;j?acwm-TEGioqx*FiY0aVPZs=r zLHu1sV$1)Nx#a)N>F+w~_`fM|%#}g;kCA_34WYM0Xa57c2W@j>QEmUf!8o!V?}udn z+wKarfBF5lT<@3OzM?;(+_S~Cj7|Z3L6_eBLw;TT-^h05olSag{d*|?55@nbukp>S zXvml&>{VWSrdJeaV#UtC`lLUq>iDN{{;i4B{;2#vEy?^J5d1|<*-F>h_lx|C=oFA| z(EID0)b@Y!&LRY>CG~%Zf589#T8Q{r^D7!M?(8ngL-;@UV-@`|R{rnI?^{Nf|H^#6 zgnK8If4icg^FJA~{Db`c5|RJl>sa~M>i$lmj=$W$PzLOanA?2}DRuZ?=pFQ*7s?ga z@%QgfV*k(Q zm8kun^*4t0e?Dp??4V-o{ixf(f2)#^+uskN@>Bmp@1Pk!EGetwKWM$k$BS`>y9(*w=fK}B^dtk?U+I_K z?>Xlmb`Lr-sf%|0%fJ7l1l~dOKDpfoQYO^}@CCi?@RW)=`ycZSr!Gx;ciJa1vwdkp zM(h)rF7|bl(J4Sr;Q#c`OKS=J+XFTj|B?LNV*g8w|3u=SzCWU^K@-ZK zds~SsbtjCR=|I6(j)SiDmjaLwOe~rxFEB57-5&VrF zLiaK{1mqj^xlcxG=l|vYpfb4F{+#kJ_X(8|{N+BOGCBqD1zmsVv)bLm6J6%qV>;FL&*^chF7&&!PPDW`5O4?0;GT&!heKbr()OvetENrb^naeuIq%QC z_sqF`%(w5a|C!gB=PsXnpZ9srbC!FS=VAW3f1kATDeYUpMTnvh{&Vk_{9*4k!FFhv zebJ!nDrr{K=NK(+#?-SgdXnXH@OHA*65j2|-(B9=Tx)8ApE^MLK)-vu&NfP4iTv%i zB$!8q4WxWAashfGz1N}jw6He^J99c;oI|dl&Ch6h$W*>=YA0nIJs`_BKkuJkOsr@A zgVr-sgk2Z+&!&Q7SO8zpOa6Y%XutQCgnel#hcNrVPKX`rF#eXh0X?79so8U^Uvut} z<*4YlR)SOhs`5XI_hdIR>-R^Klzuhv|QLyqeJPxNou&K)iPfwQ`OXyV`LX~`Gq2mjbG z$tS9C*lK;5I8-|RuPtfC{=ZAee`=c}@K@=DUs}lFXw%=X-3a!7U4s87TP6Rpac!qM z1m#FeH}avh>3`{gzjmAWpWIOD9ptau5kJzorgn)_=pAS^@k@0R|Kpzt^4IGZcb3#c z(C7&(sk({(SBk&;Kjxo$^>(BEnuh18_HDxdHvOYkU-Ysm{nh(Q{_ZWZ zU19$f{`}9X8qDubU4!RZwkJ-J@1VbSc=UY}|AwyAfA`i>^1sSp$G(*`w|>GuYR~?v zu#TGSYMmbUNE+{^v1?RmO@7dTLF(@4(IbYlTiJG2yyMn8ckq?DW9a-Oud8;=_}sDM zsHZk8j8WW%JhjSx=c90}8)qEvGIBp~Ozzn6BgYN!|8P*8dCLJO$GxNGj~hdrhKF%d zX8$P> z#HcVS#$6VZzl6~vuZ`z*TZ??mj?0}Gcf`z8%91ne%Dl1UZ-D$P)1{M?mtHzlg7mVK zGHq1L0fUn8VfzosGEtzl1GEpISGs==F?}B{c!Iw3`g#5dL`zxDWvQ2S4ZpfrZ3VRy zp9QpT#G-X;i|G^9S%+V%n?%4D^-G28-{?X7OZXS+nMY$oZd%;0F8A^9T!Z6-_1Fj+ zk2ig%XT++A;z9HguLZmZbn4Xy)|%44^>P{;CI7qiZ^=VxM2P^(x2;L`V zXVbbpPCv=vNTL1(58Z0=Z*{{o8Xws1Js(y4_4o1Tc?$g@J@}3)u5*lteO5{Td5QCA z>Qkqf_~)~Id>?<&N~T|bfzBWMrtM?MPrxT7`2#2DskL^cn&f{6?N74?ricnk{!zsr z5&?yOVoL$TzB%ld)A52YXzaIo-4DcF zWjf9>=RfoJ>mT?Quc;v1B^BsyFd*HaeX4x9#FYL5O20G2mwv?`h(Pu9gFn{G1-X0m z5BR_JWUo&-;(}?(cGH*Fa3n<@PpHc>jF-u~=pwpsqLk76lCIW#|P(Rfz1{)B_p&*%YsL2Dg6${HTO`u zX@301u`(Y#&hiL+L4WPAd7_E`!}pW^-01Na?UGCXJQ`4p<)(_iE~0jD=#kUkdg=|U zO#Gisqx8@7_D4oT`18IA4p#iX5b=H6kEM!Vh_y5F$ovni`E9$2|Fj?IdEfp_agsUx zI)Ch2#eNM1D?uQ4(CUx0c-X}Mcv_DYDe$E~N&d^J-@yB3XFK#f@1NBI_<}Ba@zt#+ z{+s(y`WJZX5$<lIV`SG<_Y@wYztwU_DF-^cj;?H~GI z1#@Yi$nGQa-#B_pii!U|8o#l=^Y}-TIm&#L6%*X4%u&o`WLzbXCSzen%+$$LJk_~SWH zULVf2Jko>p;cj(HWK^;6vjoNOK+~%H$1k^UUzB^?xbu682k6*c=sY(AHXhZGKL0|ei9bNjP zl>Y3@lIcgjgLeI?`L8DZ>qzSpBfj;~QN~2!E{41Ap}Y z(GON)Nf4wPwE6NOr<(ZBoJaicr~c0|viylI=JL<$COBB}|Fr*t=EvE;a$>!Nx*kUA z+;??4(q4@`Z4`DwNAv^@bB*@+ z^|Z#5Us4*fq!F=fxFijZO~!}XXqc?$8Syc-_&_Ko%TF>4yOkWFB!&apcKE*>S<6p) zW_H}ju~}nt$IvL^VWw+aS@B6)`McmG?Fm*~79u}PrY*;k>(i0BLq`t3YS_4}USmd& z&nG#Ov(&)#A<~CxJ8yjQBsI`(W%?Y)PQ{os<>ED=9C;Xg+?c^baQ zl#|QeS7la~7&4_H^T6=4%FB2X<1D4GRrMAcQ&p1jO>L_yC8oej3Yv!7L&b`8it%Bo zuN!;D@WCV1gg{~x|4O<>!^bp|PtO`ORxiW; zbF`13@2y&JooW81LUZo*P+PZO)=NANYw9P9NfiV~^pm4>%n|+MVmjuCfO-*h$7z-M z`>397KYD3RY9F6ND?oLgEpbRxTqnr@|EZbke8Vi+p15<>d?*+z3cjGP)VQyeiT@$m zKM~hDRLR}mZ`|KehrC!&aedIz-I(u2nF z$s*$K_~wVKGiCjU9vdj;hcXp|E$7Pb6O>2b0DtHm&RgnG{%_1~D@nUMi zf32hVquvGO`56x2UsL%zj!ZJIrv!h{{`o1LP4b_h@;~;2OgHoK*FX3U%=w^_;E47n zM#mfrX?~2yr;F&ABLet>cD?YH>nQ!5kbbN`%Z+gQ)4cTeG`Bx``rlLYNmhpR`vSh8 z;0xN~>C3vB($D-k*ov>7{)?MaKDYPkU;Zw!JERIL^_Y0%`2jvheIVcFjSql->e(|k zar>vQUwwwg`J79U%XHk!oc`;+k#Eja3|?0D7E7#WRrjIy;Q#2~)2=b`|B>6Do#YSl z4%wbtz0Le{Y6tmeDSww?{J|IWtH?bYO!9BLT)sDUK=S2zKrVB@pWQ;v3pOS~_UN-ZLN~ZLGkRuc{iK%kLEv|TPgh%JJP2hpBLRi^EMPe zCVBCB;e7IEOc4QmLEpS@;qfN^{gnLONirQe{~gr-VBL0~Cizn2bpgz;_?hzY2>#hF z*)yexfcyj9T5C^cT>b^JygsJZ*Je(q{!QmH;lzq-80Tcz~>Y%_oMYs>-f>#X>@6U_Zznl_3D z(6r4jvj5Zs|AUJEtPM22Q^S*+lO+G9YCb#KC^`LWbVHGG|I!`lvhH!C@;y8co+do| z5ts)T-%v$I5%b|ibj-1U^pW)=_9?AXpomac9i%I|vQvezGV#wQ{`RXLf2(hD{HH4ZP6&Ub8+6!|y;qsaAM@wn zs&sQY{qn~+Sd8Z2IbvR-a4Gq(rFbf>Z(#Y`%PI(t2;d9)N}uIM|Bc5g!MZof{FnT_ z`Umsi_WksJiolbzn*0Y-1fCcjb3_1N(EN{6{xGHgL#lr{#DtkY=0S>A(DyAx%*RLX zqhpHjW1F$Eg5ZdNdIz-3T_+jWM-A9b<=;0yZ=EOeAN@X1UH>2#$OE{*U#D+e^~cku z@xcqDl=l_NCj6HE`^+^6D{Fmla`Z>76 zJpXzA_V><{q<*oES&hny<-YTbhGj)H0>t+a18b-2-fz-m?goUJn7S7DZZm#^j&JpPbZSiZ&*#BCI`d9JQ z)4yJ=lN97*fa41K#D%Ugy^s21OdYz%5dnO0eyH3|wx3sTkCuMsZ@=m3pLJn!>3{uJ zng60NIsL2h`RVVN^zWJ(RR3P%`EPWR;kNp0U{5|>KO-#cFUb1q6w{?nAdO?Opt?Vs~muX{}CXZ{>)#aEaAotnNaq7N(R9@C441kdH$ zJo|MxUoxfJb}IkX^z`?j^5=FdBb@7GIwD;pmAN7W<4L8LpdUB*!XUqHbii-RJf{Ufp*!|$)W>-|T%K~t~zd6Oyq%T)eH?~vtq8p+c%{%74&iR`R-<9|hT z%n{>%E**13Kz@PNy8V&v)c%y%-#lU?wZD@+|FO=cG9S1P$^jJqW08K0%W?mk$KNmx zC!g~459$BpsoSEa^uMj@|5#0#e)ivLN`LmAN`fQi-wNrNBgR9mcS$}J5x^I8=<w29?FxQP1&1vdL3T>z`ymim1>&G z{}ZkRXxEeZ$m!%d2lZAHIzipfl>ImR&z~d){&SVw#PfJhl`Yp6QTj{Ne?3V4u^GPf zUv94dbopalQrP$d(ha(@-B(8YDW~m`h~51%A7%Qz@jr}nIE%@DA4TwU7ScJz%m!5d zwk|pU0luJDPk+|<{nwb*=W~e6@bmZjU*Ml{Ecq*>2>#iVn+lE?cge2bOmIX1U(ls} z+Zye6UO!Gg5ewvhN&ePvO$A5rFQ8+N82`%to$ydZK)!?jzDFKEjrCuj|M{{4t$+B? zD}T=59pd8hWr!TwjhPQm&W^Xx@y>I#lk>AvIJx1`>du=5E1XMgh(`>#y2 zzes-tS|4rA_O<_NJO@`mzf$>>sTlmK?7dhQnA;yb2Ry$l{k)oK{Ez+jMaQR!hU$CT z3E|Iv{WySsAIv|6jQ@cz=)i*CKjZPQ%zI_Kjqi`*J4k*vrHG4iRsV0Ox%^-BnS2xY zWBeV_5x%deP_;d zrt&|Y_BU{Ff06YI^YQn;un()aHu=k;m@$X?OSCRNpL~gN#Qv=O96G0np!Yw|FH@KO zmy{^~>_09#$(MfR#|!BO)ziQHT`7OmgWwCQ^S^Dv%CAlQ*?(DdipSqclK;ddN5{Wt z`n`E3{&D|hQ$7CH)#mcAr~kHJCI3iAbN=h;Kkoc<9yIZ1|7B4he zhefCR(myh}^shgH{-gc}UGmpgY`?I?`r{9|{GVMWQU5qe@<;#I9WU`F1gH+c7xb$Y z4TOpRwl(zavB%?YjY=;4r>gc>w2=LlBk}*aXJh03W?+j}{TBVC5y3{%h6uw>aJW{SUc={`Qt- zv>!g{RO)}x{#5u+brgT38&pp}+6A-+Xa|rEv^$`G?u*SdwLiQ+m4l`B#U5^-YJ~Qt zqW{hNR5_ym&HGe2B7iUG=21h8d_H#`6{}|=JJKUH5W0I5qk@6pWK@Z(9 zX}O92+~Jh}*&csu?Ei-U@R|K zNpw!}W*Uzcr_%fv-Q$RW{0E&?)x7^<@TFA$`}{Xr*UJ1yy9=tPANhj(Kt6y!)K0x`oFhw*{?U^&dEdNGoYA;hI5PAW9 zfF2BKOVNq`k(DU_~$=i2O__5b-@w#AhNCMf+O}RTTQA9jtJn3 z^W_h>YhJZXnG)+Ccz?8gewl=Q3TL9Z{(*f8q<;?j-@nM;gAXY59<+GVtj|sIk5T!H z`0Qg?*O}$7^M_m@56D614>`2xSZST9|8ag<>i^z>T84B%@y8X=FR7ox0qL)V@uvyq z?{B0VwCB_BO=tf4{?9$+zuociw8slp+65SM`N>JGu3j zx9t36ktzMJk^XUTf0XHGK7KnLnC~hqZX`I)p!9P3t?wGe{u>Qk$c63~Z z;P^}y^%uw%s;&K~O?j$+JFZaQZEF2Br{5*}EOXyi<7v~(`p399WJ)5|zbtAuSU48$ld`mD^2ap7)HS(`lO!fbSo2bnB z{C8T@W%^MfL1CW->4#n*KadZ=fqVzOYwrf5|Ja|tluy_rW%*-1{`MDqoKrIdN7#cY zq+^a4_lb5V{iKKhzMxk3wK*pJtMHY2V$|XJ7kA7M9Km07B3=~1zpxA8p@)#4V|DzX`5sQ@mIZ?CzEz<3BkI-oc?z@2jzdH{*Um6RO7FHV&8LRtEQ&b%Rf%a8S}5d zmm~X>_mcnY`6~ZI__JS84v;^-zpZk~wg2g7H?L-t|GM~Vh5fwD#|MW6^> zkUdlQQO@|Sn&60_h6`~jPQNbW6W0HX1;PCAzO>|6>~t zum8|Laclo~OzCI-9Bjpx`S|-kZ?`3V^YuSR(0Xux`p0(-p3Ap+`it{&mAgL2>F=ne z|60mdx2!Myvyw~yv#S4RZ%!`#Z*+e5MpOEkKL=a!{io6oxr1Ku*Yamg{4e}KO2{52 z$C24jsDJ;}|a9W@LP&=%mX;5)fOBPz!!Apz2(-M`1gKLJ`sIU>3^KRxBmlv zO2Kc;s9=7_`9DbOld9K_NAUM@R9>k+W$C#l{ZD(G^uN6KYiG7O|KUG1Q!$u`@%JS4 z|K=<1A8krM^XFhIzEb{2pMJI~JNd`z))pKQknf)$A!RHM{;g*I-eu?!@AHQ(w6DIyGNdB=T_-B0DP((L-{G)VUKcmrhfGPtpxb5tC#LqRsLKh0eh26;)~`>R z`X9eHpz)V7_{yNPgz$%bLhwJ5{VhHHTho`FVoLudPtdpb1B79j>QB!zr{7UaYrx-C z_7ttI=Jv-*NJ1W5d*vgWxc{m5e{1ZZ`fpu1aq~T@{}aOh=7z!c|76wvM{hLuf3h|H zJF8B6mHEf21p8mjiGS=tkH0nFoc=KWu@L^fHU4*=xO+46cdG~acOw4cA&F`4 z=Jcm4{-Ev8zmeCUCHnu+{{=0n|M78|#QH<)7IXT;)*m`eWWA*&w?zN@zDf0R$p2P} z@o(oxukGV^6e#{Yt-%2l>%cJ{wUNqwq*9uAl>6Byr;4*9q{n2Z2SI8J&V*R;d7tsz z{_6X$+5eKgDoyMjsLKDX=Jc~)CJvdBh$-s(GuicT8=E|Iv8nx;d4WV4+a~Kj=HvhV zDXc>J>Fa+c>kmp#|N4|FFPhTN{-0cb`V0Oy>Hld==QXDEAIy|>pYy^$n||m$=smN# zwIuy7vHqYR`Tufhu6aN#)H<0V&Z{3z$`j4dlI{(zZ!{(UyKgIt4zVM|#gn!cMpXk{F zjT^?b5+-sJrg8JxU~b$l?wJ$wM-I*#JdO?LN{?ZQALV+t;y>*-Zt%GAWA$H+jbD`( zb0?8wC|Jd{{?BG3A6;J4aMR4S;NQn?_gaKcVOmSoHRNwmoBTknab?RK<$7| zJ|}9bLe>cSU7*MPIBGxje@e7J_pYZtm(M?zwMgneS|w2U$3nm2WyT^ZT*Z>7t-Qx_Fp=X4A(=PFA|G zDyApaVc6N}v@az+u@0k{&a;n6PrL{3yXws2jQ)WZze4Ty6te%&R@T4H9cKC8x>2SC z>A#^erJvS2&Z#fy>nsmC=l%dwJ*0oWW--@4F=c-Y<8O_Ar(cfr`rp=_lE41^?WgcQ zbT41gABki%6QXjH#J;d1I?teUJyacj@DCIJt1Vi8oA2>=lH`y4DfqQ1=|N-Bp(2eV zu1`sP8|nPJtUTi{6aPW9|3F;h@wXNqUHUT)HWfw3cyZ<_djK#CRh z@pqEs5C32Vhnk9vdzy-bpRm7apB|kC(AP=X&|?@JvntmH!g{85R?N_uMjZ|14#Mx9<+`&mT))k~3>3%6l=L+jNfg zKKbM7iX1vW>jCPg`t%QS2VJ!Ctv^lje`OLq=bL|X?lQ|C^KX#lBFU4CHi0kbidmD#(Dz3@$QAYn52<+-T+g^$@<$B!XnlGihqn$w@7 z+9%!K;(yfj*mZIo0Q|C*KtbDAojlK<{yl2_dDWg;`a}4K$^UpIe|L+-9VY+sX~p~e z{E_|xJ%ao*mYLIkB>v8$LH>|G=w}Ufz2)bR_4n2L1o;<~lD`-wzr*$PL$9-1@5=Y{ zNBh_0^dSF?drFu7SbaGTq4S46G_891KtF%T{}nsPzo3--BaZ~-55Az?W_LT!#Q%WC zKjYrgrQd$+@ch3UdDasq{^x4)FDNB{q3Ive4cevI^7BpnpX(Boe}+@K^ji((ydLu3 zQi=!utA;*b&&0ofc94I;QSwLqi+xbwk8%h8AI-^VYT`evb&!9?@}o;X_+uY{&L227 zHShBuKY!TY7|<=qzo3--t=2*P%U1G7J~VGWf0Lhoy0ZUK@MM~ZZ1FH;tT3nl_t|QE zqMhd``dpmzJ*WXOe$+z5NT}A0~ga|FK67-~RluzS55-{zF;*zL2rXoc=KW z$p6^2#}qIUxQALyUeJ;=YHl>DvPhv(n1!JIQp>0htaKNU(oaF6;CknU)V9o zzo4}IZw$&Gd_mvL|9+8)|2KO6uPt5rBb5)I{-Fb39c|)2WDCjMR2x;~VFjCG|; zzx#xIF0A~0@%6nYn)vV4dTW-L_&3nUf7X{S{o>QZ>wjvuYpa|1pQZ6HC?)^M5z60Rcm8GU z|6HN#|3jrr|Lcb@f8Shr#b#6bpU~_7Qu2@78tnhzy94z7nl~-=^T+qsX?pv&p>*kw zjt=%eGF63w`uE26Pu%C{kN)3TdjF@i{AUFDTZ%tu#eF9_e*T$%O8XB#>gB(%bm@=v zI(++gQ@6K`-~SVI{Vyed*kKB*e^2XL^>=^zA^*#?@z;!vrA+^#p#CA>L5JK|Z?%d4 zk9z;_DETM)U$qV||D_Y2c*(^7CO!R|jxPP+FZRmkAb&AEbXJ%>k7n_|c0I=o9u;@G zNprgJ7E3NU2~5xeRIbR?wvcB zel<4!Qu0F@3SdYHkEU&<8&&N_D>dy5O0G!#Q{leJ^MFTI%vFn-DPbarPs(>~9iMRz zczBiGMq7Cfa|h=Q$sO5qeBO`~;xpgj&(Q65hQ~+_=74+xZ8@O+9@F}qM`?W^hsa1- zzwkV`f1Vig;E~A{1V{D{#q;4YI_8LZ@ti5-FNPxK$6Y$+7#4sFbj!z|8TW5)qxpLd zvD;+-KMY?^l=xEwzML6!PBDS+dOD{FeEBz25FEn-$~EX4r#9U{^IIkM*ABgx*J57I{+#?>enyMKLIU0cT4v1!y#6;a|8=pN zf2lMZjTmM$9*1y`GMqZG6|pK!za`{1#p4t_>LsMqHijb|F7in z&FR}_g$Z99%3txKCqL((!uKuVqZk$-H_%4Iet5-H|Lwk9z9Htw_MO+^@X#vC{UKU+ z!*;i@CcP4^$H6hv+S>W2^gdpng5M;IZ(`lK<4x-qzoPv~RzdlM|IFCqvK>agf=(eu ze4MEmu7myZUcvG)g+vzdtv5{;UvBd|Uogp!{rhpS&yu*ud2@c|oC_4X13`UAS|j-OmA zA0^t4x4)tMeKl3|QvR{rC(ZfGe*66GMapxXz4zqI_U$|CYLFB&rklnc;09lw0g zl)o*XlF!@M%XWZquq|Pv3;qM04~`KW;Xe>ET1!nBAKUL9BRJOY9$ZIaMptrw(dL1J zDw)a$`}cB-JpUTvDYN|8zlOhjtWo~e-1CCvBh25IjuUbNz5V)?3rzCcN%%NKR><)K z#^JADp5HCyQ&2BWaLno<&x>o(S|f^`=$!SjuufV$!ar1g%{0L=EC3hi>5b<9X2Q4i zW%-8ae3`#GzRAiykGq7{kyEToJq*sz=v}lHk)pnSeVqEeeK+lyqZk%|3-rDZ)}3#{ z7hNFVfO#<;pU!^>iG=YXY#|0XX4?6L;CZt67PXGzry#0?4RoRmR=+Ez4)nA z!EsSP$=}*aK87hiK=&C}KHcLum*f|>zv0q-ieUkA1MT);ucan@?0<)YeX%STIzHXs zjxJAHdKHjgm_HuL#rqFjxKi`v@2(+yCHxbf_>BBT^h0y`SVil!yfb#lQ4jm;Colhg3csNybe3dRx zJ@ONM3bEe3PQSA+O%;bI-(!Y#w(9Z3`a`7P%W5qzG0ATnzn{ZBvVLJ4{`yb%zwtBG zcaE?B^hd6jpPa8@{x~EP?>}&XcDX0l=pX6cgYp^oEm?jgKCgW%q}#DBsiNR>@&`dN z@qPA1Q(;q_;ZogB>$0J9x&4E9Jw(AH-~#c6~B@^gLoqR%IXFRcH9_3dsQT4()o^7yuX zn^xY0FXjpn{mApbVZD$XKG+}Dzt17R{95E6Vp=#pS$@VPHG5gfw;aDmQGAK%_oe%(d#4UrsKKkMapu^JB$HJegB(p0paCG|g2 zmDa6OTyT**x2n;)ZHhyv{*L!&DmN8{6vF~=fxdLk(o;xLrQk3u@; zi1x#wV~%J)3h0<)SU|Z2tyXzZzNvf^)~5DjpI1KY9m$oCBaL4PDj(6BG?*9?fD81P zQ`%f%!dK>XlARA1XF3r+c3>tdO}=o+bi?BB%y{fqXZ z_ztosL=pL!zqqd8i2N+Pv#y9oDo^}<9;4rK3=7aZ&_O*WWts5(bfbI&{0Q?tUq8OT zsXX}gU-v)i(&PU7_#Vc!Fnq{Y&@F$RVD!)J^rrOA_w0kaFG>AFzXcTgGde1UN7^6o z*A`jM!|*|Fpw(8o^Gx!K+pnjW*drw_#^IMA^4B?`zTk-bEpAy~a76w(bj-0(?l zmSb1|F3_|JMY$$?zkM#>5X+PLC-HgXSJ=lQ&Li6m6tSNL`Ggf@4?!F3|gb>}&K7oH|jy!QL$K@je_= z{<^g1h9mN~=vA`MP7(PVd4=p1Q$+q2(r-D21>gex;G(iMO!6DJOW}){@nL?ZXf*jt zqKNsAqI^21i20AAYv`OJ=09@Qk-cS#VF9>6KVSM?BNM)^a}~Z?vVRaJzv8#ZemTX2 z{N5({QH1rHDZ~qX=FYklm2z+059#y6Y^3?{QhbNPJ%X!sE&O-DrPJ_Tnkt`+#gO3h0<)`!4d_+FwU-JfFTZ zxvq(RTSssV3ure$Q=e&kx~YHg&W}Pwe@GQ$zgF{UyUhKUu~!GbkLqFkba!z45B=xt zeKkeSCpE+1mH7T`bL_K5`(teXn1el0;$l0){(X3`Gh9%wI*tE%_Lqz3m?P{CXB=Cd z{3BKu+qCfz`Sz0kAve(PK3Z<{u>DA#02S;v%<^Nqz5egZF#C8YPl9aq7Hn&T z*Os8rJJ9~^3n!Vrk7f}*_or0R<7v4MAQpnJ2ibM=kM9nE{j||CJpj<78tH3`G+|Y- z;t|(%3S6K)7B+pO*y|L#B|h|d-OAO!{PV*rv=ybf05L04~r& z&;4q&5B(dJKMwBIvi$1!@SPzpCSDY=AK4j5cqn3Dat7Vwi2cbH9dir|zy(@3a=&r> zl8wEC~(vBC1m_CUEjSi5S) z;~X+U$aQ1n&02zESb*F>->B85fk}Q}93nX^C;6?C?VI~L9%K8Fe)*kUllp$GXpdhzvd>Ae z*U!Y4&SP0*uaRPB+JDFA1?_9cV}SM@0S_rN*$*GA>I|DyYU8t@D0p3nZI^`X>5S)0f;5u8M}Kk4t^|9pe2&-MMd z=-=n}s44PqC-5Nw^&apotNI?>M=F7DFVzR4EVbt(gB6-@lB{F23G4{`a-Ae-VA--<98j z{+0eP$*-N-pKZVS@8Y}bge#5nTXFjZR5{toZUNU*{`PHkk6`kk!>dA4jz#;-Ws1bcJR>LtenyD z8`ji8zy}JsfPT{Wl13)^6w$byXqo2OKmFV+pKGa~>X%QL{nN!wH2DCKMu7|TkFJk? zV$%O_m&k9F=$?O9|FgDK`oVEC~%QAFiwpwqV$R501+tU&!&`x)>1)BaYrkNPWk z-=W`c2fl;Fqzb}5w?bkbHcICaI>$V0F`ehoxgId`iu(P$cXChW@8b+*zY+5zAJF$^ z>{)OBR7~MhwUz+o{aqaF_PuP>t6b5xHdNTxcn6g{;3q*{!xuELkC z=0`rJ?@#+Vpb+G1UCIwvKy`ebRsSnCSMF;D<8~Tu$O-pB@m`z{eIhq*|KHO5FSR7O z#Qu|#*q>>ApB%m&N=~}{|9Y>faS>$~UiJ=v!lRHM@V{{3^`lvSiC-xBE!{!mxBb1h zxJlxBOyP6q%5e$2w~xXXQ+_)z4ukukz-Oro+a`bX664F(^7mc(es>5T_P0iglapU% z*e%oYcb>Wqy@Ylz9OuRE=K+`sM~g%9@aiszHgot>7juPuHsvK+(-z{PIAUDwEX$|gVe8Uz={e}D{ z(Ed#K1z-L5qgj4al-z7Z&%56|KB%t?2*Zc#>FRw?*5CaH<4afe15y9Yi_j0wqx}gg z-LVjSKhd};2Rr^>h!de_HwVA(zIY~Be!*WomLPafM|I`Yvgy*z#s4?V+uP(9x>Y8B2m`7eJpLgiye8Bsy$U*s3F{K&;;JlK+m_46eT*lkpoMK18E z_LsjUO#!$;-+I4r0)5{n{0m}z!WR)*KRYlbflnMr4qs2$b=LfwG}W%_aspOM9RT0u zPyX?o22~gDZ%n_jgnuigDlr!_<#%a!0y}_B|`39)+*&AF_TKP30rz%ima% z_R^Fgf|%)>YPjQI#IW zM$A+C7kNtNFZkm=C~!l*udlo9YQO%W{jhJP`gxF7KCIu8!`DQ$ANEbb^??(V-3jyy z!5_FlX&5mO?)v7_Q`tYZxKq^&H6;Bdhklmh>7Sbf#+RYt3ML3Dhx zj$QZxv) zKT82dIjQ=|dF}n>1NF2068-R%RFS9j+WymAKJt#2FSONrVV79P@tM*~-CnnPEJ1(^ zw9@8{r<&Tg3kjcfc}fCbEJ=KyqMdj<_5TRbmfys@iL%EH`KZSdq^<&(uIMXQ?%(FuzgU{APj0P8_%8JDxk=(1 zRxF=)UI_N@%RVcws&!73#w|@PD|AcR-esLlx3He3-PELL+Zc#Y2 zdai|9R}oX!!{lcvV8EHuY|TQy{E)wEH<12KE|iV}l^0HujmeN1eFF8+<{9XeJIu60NUjeWcP1v2&*5{w2+LefsALIl4VRXmW z0}q+T|2`Z{I{%vQ`!^-npTqkYpMU&sS2p##AEb46D=LWCq6!JWT+#a~2#3zG&oE;p z*+HXoJ?Q!V{`T3MOy3`Ag*1M3HTlm{_F-ZQA0!IOb{#m_@&7^$*evr`ub*F2@4GEI8{`BzoT|zVu7}}+UV?5Oam#gFJ|fCL0LsVDEWg)N6Zz{Vi4X0s-cK2= z@;&yM>|g78hIH8K0JKcC(Qob_Msk`6x6>jwd-N>^G19m!ptRYZYWRw=N!%d z!=EHSG03w|HYz2l3ygxlb>BKIr-&O zQ|pzt1p5aK@%^UBRmZ32@9duAuQTQE^j|4|M|=2UN#c7I{m_O%eDg7mqwwi+0#4w= z{U0vg*UgU){l9%Lll*-6qUDp5-+7q-=pF2z{sw$U=pO(d=*r_Zp6kbl{((#RYro;? zpOu~*zM;w=iSxb8XUOGFRjzRz^1*#j3kQl`^3nZcxqjK9m5<`qDP;f5^N->tiLdhy zs(+>Yl7QVP)xU~9FUvLLgZrTR^~b-@UGJA4`hSPgY5c{PzmW>b$*;1KQ!FQ_f0t|J zBTRlezNbgj|JsCa=6%$s%J<5LsF)nSs!Gost$+ILS}9lf!3)Eu*W1(fzSn^9WmlH^ z0{NxAL*>JV&q)&Bv(PgoZ{*u{$n^;Qzd2e3UGb$KcQZb#Mi8I)mih-{Jo(v`l9S)B zN>0&NgYC!ls@%jBzNm%+>9BF2=;!IjKFatqss{1ZVSHn~m&B69=PptCsO4|PZK~WT z++q2ft>IgDZ~@z=tQV^i#CPk*R6mdN@I{YFPJTziXTK)p8ir3VACL7oIEV3d)aHM) zNPZE*hyJrwIXQfnYzdYFh~qQppK4da@ag@+HMLH?p8Ka68ou49(ERz^X`(T3ShLppCo!`$#`1_<(kL=!45Be@pBS&Le!*btz(6fy@`Hy3AiS zT`52rKL@096;d~hFBzdqrm z<4^IIU-Vy=k^V)m^u~uGHO%Eh$M=-N5z*S;RjS{LL;e=uPWW~s7fE-L#J5q&3F8=$OIKB{fgAjB zKa3t%rrm|6@xQmJ{&V+w`D@pb^3&h*TOomJh5SiY}z`HzkGkb*p}t!Uu_hK11txIN0(3LiEKvhgLrJ zsP#W$rX2Ux^-Mi3EOh{Ukv8AeXMFh@zAk4{o=)}jFIwBIfBUH(;$X-BQ<_wLq_5xj zU9FqI^)P(d>OGl?o;UjEXBgi^h0m@j@6Xowa&@!s?>n10^u>GS~q5rJv9mq!$U&4+(aPGMA@xb~fl#f+Z|2a{g z{o8tGd~A1-gB|}b1jeb5kLt2;RQfT06owCS!*ey=$-2SM{FDFTS zjTWfIYcA zksaI?O$6*F70|gw=WUKBycb9CY_xS9i#uPq0 z531)c?BD9+q*ynB^&4Kz5JZT42fgg(2CtaPFWVn=dX`D#Z&cYK1ujq>UljIR3xeZ+ zSicclbolQNlnc<8bN9s8k0TW*A6seuH!{;JAJ%cnm5)(nWqack1i#w8ZE3Qx9zn>u&N#`3RC%>wXtMab3zxA~1(61xmd-u89A7p${ z4c~e1Qd@eXPm3Fx@qM*VawDb`)%@iH^1G-ReV_aA zIZ5Ihc3uz%PuE?J|k+tEgp`AKRVfV8{OpaisApeO^;NmLNWjk6&9jv+K27 zzdUxE+)skZdb^SUrVe!ZRha$VGbZ`*`Y^YDxrF@Oe+r-WU8m#I z%g4_T9xCGRH+_C^{BnBo;j#&QktY9){4l?uh-Uz=E>okFh%C0 zxG>oN)A#p;+B>XqL}3n@Pu{)`7uy?nqf-&<>_ekc2JM{8%%Qtt1%L;5G) zEhEP7k#ZH8$>r~Wo$B}B2J`o;o8)!-F4@0_d=#7n0WQ#!w?BF=`?&&>2^) z=JB-><@Zg(7h8}bCad>aCrW%Mx1c({gC~;S()b4EWjb1PUOrXqx`f_G*YE2iq6UD@~#QrSCn4+d$0Y7v`Q|2!|>g$%7;jm^-CB&;KKdW&&)oF@$Jy` zFJlSe`@zE}S|^7O?HKe-i9y`0$q)L4`=B~L>zW~NGrpLH?-we+t`DD+B)&xdKr0{X zQNO6`Vfb|YduipsvZnmqNBS38=#>vUD>?aHtNK?MCxu+@s1W>aNB-eHC~$)g{G`b_ z{`n*H58m2E^6Oewta?edOR*4qY*(Fw9se)H8dW~94@zAYSyA~Ocf#SvuPpTd@Za!K zt@!%;$a48S>c7*!B>CN$BGx@3(;aPNmLJZ*Sa z{_3#=NlzTfgZC+^Lx=B8w^AbpANECujQVFZDcd zsZ;Ul!hM7NwCVPIe)(;J>X!+{|oW)d#XHX;{;Px{Sr}nqUWP}Jh9FJ z_oshfr->gQzAuYNucAJDPLlYr&%?ShXdmP%Reoa%H~0e=DC7tFP22rb1NEP>zcKMA z%9p!S#DZ_6JnWOq@>?)MzS&j`^nM=r=DsiABT|p9e`UXtc_!ntR>4y(r zEJ=LL)czyXBfuxr`*9ueN!(Wl`t_X+^8@RzpnvVk(fIhX6fvx#s{c+l%Wv3C^2N4# z|ApE*19e&4to$|{aef~avwHBobse|z_(9}2i39jH5I*PL6wzb8s{h)X@%5-2te-n8 zJ+sOM`_I>^>rq@s1%L!73V4v?yJH8u$oacR$sg_8120m2;G~GX+hxAHA^7(8k}tN^ z`>#PiKpUr8qwF9ef5Xa$rQQR&Xj0iNj4yhh#DVpHHOiA*{!A6ODgPUh4rckWU3Ct2 z{J%)Fv%sM)i)8JqKyG-h{NvUNfBk~|J(KW7|Dwtk;S@7Iwp-4@j{g?|{VUWX>aw^M z;9DKvS)C>BF#Q8A&?j$xZ>GO~M)_d-?#`t& zzA;Xf>kxe7$qqONJN{pYIjWrM{XfjFiuH%DpLN``U%K%<-rv0g@>l!!TfIf~^L?a) z4OIQt$*g~Dx156=|1ZQ38)P|iwDG@>RQ-qf4^MN{gNo?a7rZc~l4<;O-(>24kdr^) zvpOe-ZG;t9!*%FK5FNFt>H_ljgR{=5Wy;@QyV7s>l^5S`RrOy8zHjMw9PIdi zAsV3l(CWYURsYmJL-rf=d{mE@SYL8$#S@45@uB^_e-xFE%gc$ePs;p_bTP}1?Y?uc zKyF& zf3JQP_2qY(x||r-f?bm^d=d3Pj-p?$-95vfzr%z~|BRZnK4PsWKl@a({B(SGzAT^D z@y%N-^~8Qwem8~TgWN2I@2oaw*JFGaEBj`!uUBpf`2sEU?61W_@Xg4UFS6D9U}rd3j zm%YC6zaD1!>EqAMRX@+YIEYX0-=iF8*_$xSf%|=;XZK}%#g1y1)cE+_k5c>V!xuYB ze0u*>#}}r5kQ?X>d&os5`8{wF>EHXF{Gz8Ho&2od{~7s3M^ujQuN7+i&sOqV%lYfW zXPt3$_?$V1#}_6)lv~g)Jsa%w%MbNS#}5gggB|}b#P0p7e~=r*cMr=KQ|wHJ%p%8= zKT6uL#1R3wKqDR6cQ@Hr?_EXpizgo*pO{JSrU-mdI_3y`#dORO_#(5&&N4;dE2d+P z2*3q;Va+zZO!!_W|41CHDTl`wnM(X90$%|ga|FIZI_3y`E**0OzMN?!Pl^b@1)5cP zfYHDA#`WrnsfWjhadP0pIq=~e_;3zM#5=<9{oU-GMJ9Z4 z`%~jP>Q?@fSL0S>!xOcqpE&jN*mrzExOWa72K9gH}(?koMQq z{%-W|TSRI8qrW#lTjmNo^Fy~5!ru< zhg+rI>G)nEc_DvTUWHZY+*kj-N&?{P2ybfVsGoB_NEI9rAUDvBV`mL9;p6?&9IR>* z-#@vJn)gzN?V|=R(94%?DQm*_()AKwbcw{r`#}BWqo96JUW>JJd!f89HdmrSLE0-y z_gR+K!?e$v;t%o`((;eJ%V~Ehxl1lssHwF z(|APtZ|$XhS`_D>7tH@^{pES_21{^kKMwE(%U~Y;hWS*;&U%JJU;tt+UG_Q^2`5*l(@UbkI(BLlqUsYc_DuqQW50ij0>9zjxBr1`_9iz1jlsx zPVxW#_*U|~@IOrjM+E2{=)$K`>YMQG{9Zocto7sb`d9F8n{%HfI6ljThwLWWjunql z>fiBsgzj zte5`wcb=NxvYH3i*P;E5#2O2ZNKgLW#)2aP${%RYj0$r7f!05GkM`$9M|t?%B=PCv z7~uO;Ps&eP9Ft$InXoH25zmjOwu$tkn9j559076zz4*aT-Z9DVj5NZR=i!T7n4J86 zR`Upvn}hO$-AvJsbd4hNKY9TjQ$&E=K-+h?^q2VexhB8gxPBS!;S>K!4&PEW41ZSg)W|M&r~V?y1+^!s_i^I24$9kLIC-@u=& zIEVVm6vO<<0vG7EJr@|;-<}Gey+P)y37>na$KQmH?UeZOS(UtV#^--tN~2`|^!k=<$1u?0x#rS$@FA6!OEhY2AbOCAX6Jnk*~cVZvANx_aUv zSwD~ZjQaru*Dqbcu58tJ2X7LP@GXe9!Ju|IPbT}sa`)GQNP6K zw;U0G3$*+VO}{hY`(~Sb!cqG^B)&3<{?)qPlD}KAzTo&e;c@RJf27CMkH>F1$@@ig zkK?Bm>3f6L`(;ml0$&Clb3_0x(6P}Mrts11;0UzgJ=agdwN7OG)CpxEyfP4jA-(%;OCVbhn{(yu1iIf|!v+~z}$Mq)t zxr+K9=hcqK7Ibb6qyBPE?RYG1NP0!*8RyrI$Afepqw_*K=NP5&7@GF~J=gY!ke=bRv~*Y#M$kAMsG+P~aUCVY;DFFR5?e87>eZ~z~0 z>-caE`oUE%Zt~+3N`BUHLHQMx5}(^AM1F6}^*Va~=BRY^R&?=*?te1Aj>_(4{vXmm z(___xEAjFNo8@=3{@$ivE^%S~tgUc>K+d2SwV9J?D!&WJekTWazD&QVe=2JR`Jelz z{`1oIPm!;nv6cH4oAS5DIl=N?@YB{ z(Ap#QNhM!&gFQ^O-w6EXua|tV&KTts&*2=j+t$;U`STa+U$UPL;>){0O+i@Qqql##la9YUQ^u1m6+szYMef#Riz`mrMl|^84t84!`>4 zhxT{k?^@Eo9n-@bUoQ9KL;2WqXRv-LE+xLmpJw@?{}PtJ zxZZBXz-@kfC?6kb?OVa{(#bEf$BYkp2^>mniS>23-YM_)!6tkYdBVUKvac>3K5MZV zpZm1Ad_ey!#d~Mp3U&PWkiRQ?1@*6}l=#q(Lit7i0M}6tAPSs=R`{^OaejO#A1!NY z{l5{Vlb<`xEI;c^c^&x(e2^RPxf?$@&xG&g9zpqKj4T~K=L=cChsjUJ2b`ekH6FR$ zWdH9`TE7-M$&;b0@ZmX7Ue{KJ{=f(8+9H!_UD~^Uxe_V8qbxtmw z7t%T21N!;Cd9_XJ|H^$sV=phJdj8KxN%^7FfbzO0mb9$^&r*Ka-A%#yg@LPNyN2tK zXA#Xapx<5iV`Xn0hLVv&)QFcV3|B8+H zLbbc|gZ=J&7J{FW;LCF3_ z%!_1*W+H!$h2$h?qwXW;ne^|s&q;n~d-`XOmhwwHs1A5sm4Aiw?y63*bZB(pyoSPx zH4yvh+@*8KXm=&LzuY@7rgN9h3+VUmM-3A1!TWx`=QPVSKDo1x5SDLzGN$lBZ$L3V znWOOPNpI5lBkC7N&6KgQ=T z9}`tSQrEvlYTtl+NwD6V`zYy=Z=M_YECuM3ao7CmFCW;y+VKsN-_NO{huXhtU1P@A zgZ8ib@fE81Ut6=UG8p@l)xIvsr$;ry*Mi3NYtlS*6{|%1n|;Ra`2M$awf`6EpHF|E zzReGK<--lZ$K$8|@^Rn`)j#eLoL{YkeZdMJ^aJH1a?-KFezjf#AMSU_Ix}9sJhoE4 z7xU-6X#J^GE={aa`TFqsMfyF325xW#)iE0<0w6Ox#QYU9G;&wd~E*?))k{ikL=>RIB9fVE?uUd56`>G zn%b)53lhvG)Ocxd*tYUj!Y^%H*2#OtzNM}4)nQ2svFOYYmmd(`VnkZg4Y z^}wc-ltRXDsr_>u)%f9OAJPxcN=vlg)_601ZofHzFZPGKbA$Nxd5aABn5eAUd(>x^ zx*kS4svDs8j<4eOeewI=3V-bBW1WQgXJ`I5_0jT=h#L zb7a0izIaZjcn)+{^=h5{`g5_u|NIWB7yEnr7oBU9lYhd0UzThK!2iOJlwK*Eb=8e4 zz6)Nr@IVMnSJy!&<+W&P8b4i=Cx7VPCBI*%&}ljPL%xr%9Bh~1FQq8`7{L*KQ(QXc z*qG<5sU36Zm?O_G@%n^9I_3y}D(FXqz!#Rf0sODcek#5{=}T3=qW<2<>-#QA5kDz^ zDb_@@{C^@G9H9SU_KBjXzmZ-)HgKQ^Y}$94PRV7jnhJfiReXBgG-cdaq+j=277_@lSd8`gOE9VIu& z9rT~V?<9$Thf0{M=nYk7j%NJoRD7xC_m105`hR(f$V2-$)m(n_Xnh9<$RGBDbpLIy zxw4&d&XVte9Dh;zV`s~HPS9FqAGT8Yy#QKHZOwGE{7+K?gZ`Wp z;@_;UW>*9_^?oVj8X|s?sS;jS(Gjcn1@u3XD)WCojekX7O%=V?$n!|ljK7zL{}T24 z|Hs^Sz(-NE|4#}bp~y)MNLh*+X&%_3SQacqUy!p08>@m~L8Dk`%2AY{B2hpA+tNc( zgYXIn$|*>(5iE$<77+olL`6kK`9E{>eD8LLT{$AJzyIbld2VKBXTHxf&(xiH7UN;y zShYo^W4`X4--kooe|6$;#)ao}iftdz|B`v*e{k2|JeL1s*|hHRW<~mvh_?Ko9e>R`{!uaj5Y1DrW%}U^RQp68C zm%9AVWc^1w5B;f!@3M2f&eibhUdWw`U0m|=+#PQDm$Up^zf4b_nwh|Fk4R1abM9k( z`_22lup#5O+j`$i??Jo7-xW1F6l_>EBq zhu_fpQ{YE9Z~{NVq1SGV?%&WKg9DH0Jt*qdAFSVOD7ByGRu!=OWb=F+Jp$-xE>P%CZ?Lqx_st-e1D(G$XXOaPb^bOo zo8Bu?7`}VvkUvcKgD&um`+w}Xf1k9kq}vZ1L+z(=Q}X-44&w($pz>U&yr4f@9_IE_ z+wXWZ>p#A~0{!g+{o(kKt4ekM|Bw^tlod~2DfOq2zx#~&?-Z)PwhO;q?5{sM{;6Z8DmAFh-A7hSP!Y8)VO%Dw~n9$m}( zUaUR&E!TPDdB=J0r+4wLPy9jY$@K#ozc% zl|pR1zx!H-}W1S(^$2)$zew?bO<{t05`s0ir;gBov=~Tb}>pSx=ApPIZzJbIazZkb( ze?Ox0P44sWF>08hKNXkLcdsu@`ek^tW4!fqtt-CWlk7LpeLOFzjvIcx{r@0pW%#xK z%L%`8T4qB2&e+uCUsKZuw^@H6$LBdd!gc+?^9hOvIbdS_k9SJ_*`evr%rc75 zDg6jPj);Uohk#A7{i}9Y#_%(QV+(j{RQ9)Ct-0_-}e}yLh=quDdhLZKy zDEVjduQ*)EC)(9JFTNk+ino;f4Hr_spPKrJ=uzR7@AJ}Kb z_nEHSLh0S+YX2WeP5$4_QF<=;TqzJhPhu=*$gvIA->9S3bq}Qfcn-Ac8+T9l8$T=~ z{PK!j;vPRF$3F*ae+BkJAm3E&u_zSD|3>m>{8vVor{ngwA%E20uNso=wmXvYH^_jK ztU0K+pnCm%742b;x1DK)eId=1oKa6_sa}kd6KkG-OZ?Vg~>nfU<6JFp3{qoUizq#>a|Mhm#pXht($@YIj zYVz-l{Ui5y@#k{;hJFh8kDjgYW4#yUk^rpz=*Rv-rbkA4dA@&Sv-}I*qVZcSD}mpd zm>T{%yv_qW$o?Z<*RuINDe!kMVfjCn+WJ4_j&%R}v`crl{GtB_**7r;y7d3C)bRJ> zaf04|ebiF5{|E>EnqMmX2tSbg!wejB=4Gq?l>Jx!6DplBuOB`&g4&*sP)n&U1mKr; zgXD$soPc#UUJp@sg&4y73ibKp!LFa*hyp+6>%AEMLMxS|Fg2%x%tja|QPh?IXmpHBw#vSKZhvZhrdXEFdKG|Q+NF{g^at;$ z(N3znLQLg$#?Z$9rCRub)KB`R<=Lc zyf1na?|+l{%|^5?(w^Hr<^y^ID*2KZvA8sd;b#l35V~gC8Q$@35l_7x1cfQ z=d$Mz#6d3K0RHmxPTel$zn1rRWBs|;C`x~L(tf-1q<{M(_m9g9^2ho!+Ie-Cu%rFw z{ulDkmINF7E0fJ-|pk}tK9eh557Ui-`x0QDSyPo`K3GFCjIY{DaJ46_W$YBy*Gs-`5(AJ@65RRHb4Abx2k8N5r6!^8{W}UNOu$| zuDg8%eq$%$BNv3fjP6NSS^Qu1Yh2q8{~{Xy$jhGa_xSCfTarC1aQ+dIegUpv3DSnAJ>N&8~f)YSC<(~YXW810lk$Av)uzDGN+ zwLcdu;r5@y57hpIIYIEzt?jaL?)r=KLDo!~e-VCM#+AjR|^wi|vhQF`i?DMXZzlC|tBi`}zxgRS1aqd&$C>I?!{LmlJs^QP$ z`%iLN|1tk+`zqB(7k+C-YWRcf{{_jpfct&y|A$a*r0@zykpk`UN_w^Op1d+Rru_s{M>z^7rkh znMw8@$@OpR?a$EumsBD9faHSOPa})$8MA+?`aKE2Wk1rM zm74sg^EwIacR-92c>N9SD&%*57SE6FSLJme{lRmf+rsu9s=xc$4;}I)jeovKwx8B) zg YP&ux1?;m&!->>I$F2LW0e%NG>H)=GGJ85q!b`tIXuTHim$#tSBi|qeOYEXHR z3%bA`@J~JMxf*WyWBfc@;@_I=zl_q<*V_i-9DbmegFS?90WbO`G`N``{(^~_b{MKmHQqrY){`l){Q^`Yrg3e#)=PZCl@@7#0JU@_tzOSewFxh`TKTQe~zEu*_bJs4(Iyy zyubXVT{d|^e{lZDSn6FrLBAZ;!WTTG+E0terN9+Lb^IT7Y*$s{H#Ge6{EoS>N&>%C z_P@n%r-(mSW6Zqaf%x|UEe-!;q}SGiX=09{L$G08tVB#GZi5&ttg z6rNaxcb%u-udIj67xG^UyBQolC;`O(2lxZ6n&0IMzxn&7mFoHE<*L7tdY&bc-)E8b zgx%}Uu)pH8Y$~KXDgL*fAm6x`9>bo$j$ZQ>pzH( z^^*_@$jQt?u&`eA*B{uQA@<^zzD|e)^mW2qzR?8hCn=s^(DA?h$fEfEL*xVeO$mRj zN)o@Z`2P?5&>PSNm-j94)1NO0zr131sCGfh+5P^s@F*&u6O!}4J+#2!)(6Y*`lTHp z^F*q(e>R9asc+JcXRf5?!oB_;&)nx0v2?d1FSGH(Ad{(1QZN{{Rp;j)*c zCjWli9$F*3{nsU%xt-DWKitaku`dGc67tn4^au3H(i8IB?dMwF|MM5g-yWZ|4{t3^ z4gWoi2kV*0Utj-({%=QrF;lgxsF!$7r??;D8>Lx8uXE#veW5AmQG`fl!akGnQfl}= z=JkIgr0lUE|82Z3r?>NU#(LXT0}ljH=soDLnfF+J`ahH#Vk2L*pZms0{ZHC~!~5%^ z^kzc3p0qIsx-2Q|Ud8?*yIbVsdYcR{`@X4=uK5+pU$MKXkggs4Ka!tZ&}Hxq$iK<< z%N!~H+^I?q=znJse&>gDG2m;K&oY1c58w?&g`94EUgw8>R^ZUbZ^-{hwo7IHq1p-P z6`s>+b{%g>kLN&df8!b7_VfJ>>e*;}gPn>k;JP7)<2R%DyM;A2{(!E3V zM#i_6H51Z(%zharuT=4WTkySa`}1TKCHrU}H~!@Q=ZiW~`PWJIpF24J*w$wBP10s! zHR)R%$Bt$~kkikbTtXd{d%t_m>>SrOpZ=O@}qq|gnG4MbD{W2-ic#MUa z^Hx`-^VjAy#>wqxl<=E>W{5GIul=gO{Ks6WG6Mc0$dku8fRTxPk{Jp=-i$2f@``f! zZafe`qZX^H$-#_z_dlGbg-03RIa;{T z@ab%1f^oj_j5(@>lz&8%f4iG0!q^P46#B5zAOBJfKlYDBxj#bwIKPZ>CW$6~4baLf zO~Z%hL9L_lOVQ6N9=kDaf1t6q|7}cm6~(v=(Pb=$2jK5gQwa?E^FF^Pja5;4g4|B2 zuF{DeujF7+|0S<;Iouqn%FAl5!h`;!{{nrtWO{u41o5MPs`>kdd89uXnPNWA$L&@A z@}J*NJqi3l_TrCt%UiF}&P2J~V}nu~Kcn3O?NQ~yH~jjqIZG5SXSr(sB~JJIyW43z z>7KuDqAow)@$kR?G@g(@&EF3f(tT6Ydp|~Uw#is=c<>ex&}8UQ$-m$1A2hyCw@`ILTEvNAcV1N4W! zn@}(3 zFP_f*h&FzHo7+k71^nM~ziiG|`j2`M06$^pz~BgNn$3yZ|5?KMLjKdK{kNto*~+Na zT7Ud^a)$%?Z$m#7_O?Iz_!Bs7&KKiP;9_|Dkvt9o{~vZaOWA+maJ;`0LrEWJripKt zbGxz5AOE+fst1985BI0$S>Ewa)A8PNLHYl#m5aeRf(V`iJ@efSL;S{X{nn7oKOp-* zZ1>Lzz~7Zey1;)t=8x^Y@9!?u!XdvwS~$ujRs0Lab_x68f0*#w8e!IYn0YrMoC(bnJA z5PoxC62I}*!QnTDd+RUs2jK?efc_xd)bQ!vJPrVT{y!(j?Z-}M%{DxKI&@BFv(2o-LhqQ*Ip|1ntq=Wx_Z9XI^A587byU!VEu z&ymkk{Vh$lpZ44S@}GRQ7ylI)Piph88?|uY@2`a)NPqH~5y=1ZQEA!k{tN35ClY>p zP7=TMjz9jcWG7l)g^Gw(JpL3%c*p;HcpQs%9`ZYd-(REuH8`Ch0=_}-x}n2q68}JM zpV9uWqVcocD>Gp~(^&71Kgj;1G2N>_c>k{3)r9>_eZNPl_-~(aoGI}~HT*GZ*G(&l z-#JM9(SHg5x?5Yt&rjvE{cXtq{uXIuKP`#hEtDkEYVEWmrT!FX_|H9x-d|sv#BaQNaPp7+OZc}oZ9hcf-=o>D z{<;&{Z*<{z4idju{O{<`jf;B?m-q$Se}?{dJcG_QUYC@={ocXJ-})cn4|Bt0Fui}= zaL@jE?B71SJIVjzDv9~K^}auTef}PF{;|_b+`ms$<$``{DX$Np{YSr$YWvx5#h%rE z^IwTSelhN#`ER5?&2PyC&40~$wS{z;|2h;`y2SjqF3o?*Vg6fAait5oboo1kaU1B+qaUj!L7v-z0wfql3e5X8$|*Pb=MZ zy`TPk$=@H1ty1fs8$a>Be}9vf=G^O@ZJww3&hq5?q5l2g!>;khw_AAq6RUdtsp?P3 z`8VcB`R8i#KeaSNnEjISH$FZ%`HP%?NB;LtZVI4idlhFX3Nv z!N9tH`m;T(M#$0Csq0Vd|3H6k8`>qy5C3#iwV$y)3cuat!1{kM_|vZP>iadc4(YB> zsp|h5qe8bx{V&k;r?EkM8MdV*=f9sFoc`O@{~i5b{mbOB5`UqF|FVu$f43*`I|qs1 zs`2mOzq_{hL*nnJ;qUwtHIO@!`0dXRPX4j~5q^FBXV$C7+$Zrv-*A3vpd?Z1Kl0&u_Z%_s08E!>d1duVL_e8N8XoxX9i@HVmpY7kg=9 zW4X@l0Vk-zKQPaE^vVbxQ_(vK6eP*LGr3Zid(;=@ZR(4o$#3+dv0$XWC~I3^RM6JEDEUJd)E5@r-$Q3)>`k7_x(jre3ry-Z9O>rkqhgKl6Fb_;-dPZsC^QD z8Tk#m5B!mf>(hDA1LHTC5!8dUyjJhK>o3Nicgy!e6J7fMWq|(E`ng;GZ`amiIXkiT zZKa=h{|Nn+=Tr@{vl%R^6Qtu0l>fFJZGOD{tQ}w7E_Y= zo&OR3Oo9qsk@PBBkH>rcipl%qFJb(kzg_XeRQLCncd-4aDYSkiuWVkQFWIK#4~+s% zpQ+@kzpr>4`)y~b_7d{y%;gj1@M6XTyN_6hMmXP25a0$C<6ger5C88Sy!azuRStiw z^w8ojXHG!3E-v}Zjo;$;N5|0mr@SEl(DusVcjoxxH|MHw$XS>FTE?l9NgM#$sMGUp z-1rAF{+YCXE-&COuOxnBzCZpbhX>(L=m+Bkf!sj{ejjV?#vf(;rA@u~OLkOF{z3Si z1KHz?scx673J$aSXyAB!A!rEvj+tV{ZJYzgu2<<=9PcRuFAh<+fCLfzCa0;rVX-*?fL#o!0-Bd{;U7 z+pqn7{{=mV{4xGP{Zwd^|IzM%7B1P)-i;snzv3754CEjAzH<13ZxI&-=%%%;2D$M^SbqkEz4*&3i9g0Tb^SNiDExVh zN7tWzNtN;Z0&dWauhy>bhrf1vFaDArDkpz?gTMZZm6gL^!rXw)KjPCn-1u`@{-fyq zwY(Dg|6}Fwi+BC;S9bn_aWiPAPrmKp#$Uns_oaF9hkmLY{-F9>+3_>lCD2tLjGOPq zkMZZ~lfC%M4>5j=%cYq0LDzq@t9&~_D)<-Z1wZcmj~o9Umj5wFd-0d-uAKa10s3RS z_xJk0BZ1iaqlmqr=N7)5?Zz+o{KBPL{f+!wIsB-n*_^B`pwo4-1sqnZ+(^*f61SflYfx@ zL>Yf^0R95T$F~!toPUAd(XiPIZv4>y7qt0fd@H+>f`^6%E>JBAp9i(^0zr0^p?i=&T`|oSbx$_^XgBi zO6BldEaz;_7vl%u&*gAk{~=$#ogf%DfDSm~+yA=pWBf2e>%Yn?i61z1{4wUMH~@dp z_yM>I^K;xm){>%H1SS=LfQWKL%cr)@2xm_5#?}EuGUMSKcjHobhNkq zDdBK@w*mD6>4E0)571_79~|JupU>}4hri;*UtUT42IJK6i(TGzAmERFr_OB~0IX4j zgEs8*mF>n4`RBIxE?|`8R8IcUBmCtbU9WJW{nYUX$sf2uUmEuA$A0*W3%vLv)hdVI zI7;EvNIsCC}{N*2;pw2h%hyMU} zOd)sBvp?U~(~Tecvr)reUP=7+!%CjI{82A-{MLi2yn^Hpxr1K3=!OPv{5JQ$x9R#{ zqmuGxIqUdQFMw0mf5=y9OY(m?{{*!Mz4x6PKjvQz$X=biFn*5I^v8do`Abm$<*>c6 zV#cG7AJ%fXP9|{xX#S*Kx4H3S{3d8$u)KgjbXeu^TNPgY(Z@f~A02-QEnwUk zGq$?%WB*1|4S#th@y7!2!yiAEK>Q)ke!2V(IRlsu9 z^=EB>{DB+viBa|b@WVe%o4=P=62JKE@B6QRivP>c)_ca!7(d*i&wuMxQvLz^&n@Pw z_+aJF?N2$=DL=jL>Axz_{^dM(6hR>^}+f_Xmyo+W5bL zKmO1G=D!wymphyD!~9o2M}_ex!VQkE6W|8D{mCw#^&jAG{Hhm!c_s1VyP!J$sIBTn zhyemW<~8Wobo>SffL{2_sI%Sm7xS0r>R^RmJy4QcIr*D^c-)Cv_--eaLZwKIa4iWx|bGNSZ!{6mNulz%eDu-X}^_RaDfFF7lB!9H~po<=O ze4!gZ+RrR)|3G;F{>sk3V$4@D>vO987sd^sr+&L*v>Sh(Q2T!xo$hTvOOEiD|AFvZ z`@HhU_&KpJo!dzrf6(|p#PLBle^T7cjlYobN85VwM;ccSze8V~ktw150e<^K#aBUq z{Dam%fgALUdAr|qYQaSw5YX0~g4qtna_-&34`uXblo!t1#8Gk)G zKPxZbFRvtijMH@aV_sGq&xMYPvfV_C|0i*{PHYYUy`oM%&-@GTUm9uWk4l z#t*vuOLWCyKf*y96+RbobuQk#B1cGv3%Eh^UY~KB zAO0C^7gpS%_FVIwommKhyQ(Wm1_~kuvz;E74c*y}j&h`U; zh<|{-+^OvdvOiE)JfiG<qw5m`&*l9^T62IH;hhKj0 z37BG(wmZJWFTekU`x3wVzQiIg$Q`uRdEK-9@E6hdbLEBaa{1v$UU8a&{N3?ml(yRs z`Ik^xN|&nqr|drVFhBhA`@ixshX%{voqsC&BfqDGXJV8#eoG4c2ul_JO&=})-4B1g zGKrqb@I6&;$K^@=Pkz6t1x@dfm!wsxKxO?y*Z}h5x1@!&y#CxSUjO#vy#DjF^y6UY zAM^&aX{*SCe)x|Zp`I|#Q|-TwKl~`QE6*msku|zC`CWF0&ZOt%{m@e4nOyJd6uub! zmhLw4yZOD4;=BDPl3gcvdO`RvIqR(V{P2Ip=kJ`E{`gxiQ21=JQ6?R}k7YV!%ZS{c zB%nC{FNgcqSA0eNJh_YMzAPtmD?LxH7Ij%NzD>_chYNCt|Ki$rUMKBG zm1*{)u9xShK2Aq8vNfdih_a1epwfO+8vPY=_|vbe`9JJLaj=-f;r0+U@5SoOq$fE( zfA5K!o%m(fBmA!NL^{9^8f|%b6F>Ps`-OTU)PjKL~$|3wH1F!ylfg zp0MU9{n7D1)W)0sk;&r}6WS z-9kwBDdCs=M;~t?q}$R*`O8aM2Mvh&s|Q2)c#TM6lc^govw!8okZ z5M!pae`P$a+5wC|Px_rAZb`O3&atVrKZU#f+n@L6sqokmHE%@wvwXD*$9oIt1?<9E zhqV&riTGlX>E`@QAe$ zQvahn3?Uu#zk+^C2mKFaw-V9?w*Pd`CKsgtC<)NMLm!wY^*_w|kMqA1>J$Di)5ZA3 zNBsp)?M{+>(pR@H7nZg08_ z%kus3-#L|8Zc%ys?(w9=@Agalwu>K$-|bHo|DVwZm-*p8`3HsHyh@e#orT`Bn~1A6742lo2mKXJ2q!f2<~C3O5*wHa?U!b>jyNQ$3F?dwmC zibOjVE^{`!D-M|N1NWa*v;Pk5uWPKSI8D{i?!^Rcic<`unC< z-%)=P{nz>Q{qN2+xaBKbqeBia8AbR{z`c;s*ZK*4Dp436X?$^E)401qi(lHGoIUcg;o^&SN zx2V5{j!Zw1{L6?RiYwg$sxQ&IllLVSdFl9nxgw{(AO7|?sQ6Z!1K_{6fiTD=@YC-w zIp8lDPWQ+Ge+Au>E(m{UL<1pR5dJ&`OBhUt|IqhdssB4z|IvSqTt)S%B3%sO?+Mx` z`|H0fmo!R?szLnf$LV06$ic$o@8dupA;%)bzgy`i^nb;6fBh-q1VJym=FCUp^_RnU z@cy3?g0G?}fe{GawX;O8Xw1+Q+=jIb;z@CbxOZ$M#3-GpJjc={d-QP`b_m5er7A7_u%O6lbTvANq3(*`+r3B>OMxw10-bgx8mhk*b~o{~p#`U9W&Y z)UlDU-%R2UGbgAA86R)GS>i8e{IK7?>nAGDKhnkKHC!K0_s74P=HYJnKRKM`sp)^a z5#Dg<|9mYz^aA+B!g?ZlJuPPORlqc#X^q;Cd?4}Tei_?OUjGp3-$k-!s>@;g)Lh0d z&|$n_R>a}m3RHcEJULigJ=z-%{ImJ{s)%p!1B>re@zL)??k3*{{<{ZXakj)?%HLnx z!ROarJ)i2|Wl8+@8A|@Z4SEFCJ9z;=?B@fIx+}zLZYL2A{P=#N-CDipG5CQKr~{L^ zBfgH`VEahLp`0@VwVp|B~@-*?&{jC3;{qg^#$^SQwZw*rQ6!}bqU36Zj zMg1)w%W~%SKZhUqPz$fd;ot-9&n*w$@rlH5@b}};{#<=4=?~$@3Ui)$WphQz;z@o!@N|1?75=XDt(hvjK^ z@t1#&CjUvCPIRFf4*>s37+19Qepj@N<6BRu^ugD_*^K`#RgS<5{viL@(sQqs_>qrb z{O6feccx|~@LOl5hQF9`il4nWSF#*p8`O9f`eX9<%FPE9zrY`3zECc}i|2s%yt%DE zr0>tx*i@qCiO~NM^#0V|S4G@H^FsLy((PQ8!fPh`^+*HnIAZH_-f=^Qlk_{Qs|kbN z3%bV@xo;%>g6>~=P10}dt|pdH-WWqjYLH)Ez>~9k)if#pwU4Xv!usbcq}O(@G;s^t z$9K;1m%rpsUPX$Ca+W{#k#MjW%W@Xol$;?)oAX6{;A&f{@K}9Se8@i!2{OI&#ljEd z_kYW3a(}_!AL;cC;qRX&#&J6K+5Y(Ddp&spKXBTaN>9{XA+0t6XEFpdT@M;+gp^& zVDQD^;1?A5PpEtHXHtJkr>J~^|L8K(Ll=IdYijuMow!&ZZ~d*s_1F1Ml{4~>VqT{8 z3!O)^95@{9zkvsto;LW0H>CVcO@BtvdZBY$QvObg_}dh#e4`~^{ps{y6>i<*9e)nz z^%k@5nq$d9!z{z@!e0q&wd1`vnlEWRA zvs{aj*N@|vU#-4G<9Cz)9__ezYP|o_^k>O!RDWIgt#ebu-(`!+SD(NC#(1LDz3uE6<)y5YVi00){*`2*v5=R`(vD^@MA^;D%+n7 z`ZGCweVot1qAKIm+sP-WKXI??V%%?@Nc)QVkPM5G{eipvN4gVMUht=%{_m+p_}@)W z;CC25+6_>>{XFAEmOtwW^0{%9HyrYNYpXXL`hTYuA96R4Ak!t;KOXOgfB1(am%Cj0 z|DV+KXHP)?HEEI;C-57rx3<rF@wgR z?O>-4~gV%eJuE}zS{sT_|+KsbRyK3+Q?LJWHnD49bd=3Xix@CJF%#iw1 zQOi4isI`aI4?jp3TNsZMfPYJCl@e96_%)u(9qk4OivliRLu)^u;C{+nsBohFG`U}h zeXZn1%hB-*`UAT4+|vKX+fPm(>;E4vpt?UaO{{3o<#@io{8!L;Q(nNofah7}d*1%5 zq^?TG`c3uA(4WbWGw)MFxy(d=UCmoAkU#J@>~_tK5`RBlpTzpff=j5~x+hJ1^P?(9 zi}%r>-2s)~zjNbHj1&G;c-37YFdjg;06)H$jql|_|F6Wl3ePLUoDgt>&*GkUu5s(X z$@6!)|0g!DO5**g5mE9-yALY&b!5XW~6F?HLM_ z!#cd&BV>NQ|2`o4T$ty2sQrgsI2ZU0qQ+CLL*mc;M;L;QcRp5r@P&~xAq@1T88 z+mrZ>oh?Nfh3W6H!8e$=^AF>0iQm!iXVLx#vph2)fBS-eCVzeZhK|4F8`_se4)TwE zOMY@XeqDcFY;ym9{PgD?8o$McWhL-i7ydK+iT45l`cv^;OJR`%ezT&bDEfYX{JQ)f zzp-$ll>ZJ*{)6TIDVO|>i~bq@HLJN_;Pz9Ozxg%k59ti>7wsZHh3WD)SaDE)`xP%z z?@u`#^>-fa2eU?`CFJigezZHFvi{1eND+bgi+z=M{p1`TSD^jj=mPs6^mX+P9N*w^ zHSj`zK({=-t-4!(aDLhOo^t4)k-%?XtmF^eph5Ue4gWFxUdCwWotKqxI~jdR^%u~O zUK}5M0l&eFgVru;JeT)-FP z4t{>y(0{U9{?MN`8EGPVWU~DaUsEN|a@^O6Y${71~!sVY>e7 z_)o00<^@0e-3)4f79{aI{}BK74Arj%;g@@$!sJkYi+nlH=sM}y!VV7KlJC# zMO4B?F8#mk;PgkYztI0!ufNy-Y*s8?|Diu&zK?R4zqxi#4}P|v@ZFs>ejCS&v-FHr zm>T}|Yg9W4`+A(67?Q=~iW9x-oaf{Hcn8H7^tvvu`^9*_7VzqHU;7&_apRA%{jsd- z)c<~!E_U|ja_{Mnf9ID9VG#o@;(ciP`-{%JPHxvy<08m&Aoi{GSK+{4?H3+5j`xl~ zi+M>u%5O?*DChKzaupus z8xuLcwVmTLuHoAmKd*lRufflO|A%LF{#MF=2dAIS?dRbIw0<~-6QgJB%l+lA z|IX9i@#pj0PTJc1@gDRG+WfbW^~!oq@r7~$e)v1iu)dV|v!7A?;{D4LeQEsf!f##S zkN;*({z32WUYXDG+^O^t^6R)(g=4)4xCG-tI3~Cz%OCWD%)OrPcgI+N4j)b9|KVBU z9sVA)aiu^0cW9r2yr4gNzoYKb`4OzYY3;v0u0Z{~nfqn&fvWFePKeYZet|x_?AMy^ z`Wxc?C#@S(|8;kUxO0<|j}w4j?jMsE@E==M$=AyF>i@74RJgdrs}En#W%=_s2RIwE z{=gnezRE@zOi@ldzJGDJ8$bGAx&JV>s7eCA&HE5B?g7>D*TuY#@o;vcg!KXAw`{(# zll4EA?`>Dl;B^2itj0ya3%P@zd9S^5u^WF8*WcbGe=#&Ef9q<0{n7EqxZTs+&!0we zJFMx?x0?RvF<(>jyzx;k5x#FS9WZm-r*8Zv>rY7o(w}>i_>Eqv;s4<_l`rZYXCpp1 z(TfN2UuCOs=#{|_96MBnLk}Tu+()`MKV181iGPXBdd}~U#&4%S_t9kg=>*`H`^V%3 z{ny(KbytX*ysrlRDezzSlFAq1A%5WQ62_yg9~yig^q8y95)ywou;my?bVI z|C^=ck8vlc#P42z{%r>1VY{E4otV4F8xEW}2VH(E?K>ogbHwFE=W_HH-VU=os zpjSWaRPm$v-g(&)1`0Z0;g4P2@)trqS4Q%W{hpP;@1%&|W<4~Ez2k?Yo@4yHuc?G_ ztVjFL;Wmdq+*;M&SPL~SE#q*MkHP-&YwBJn@#jxf=|lb}(|$eYhb%Gk4bJyEfBDZm zPbE}@^f4Yj!8^`>8vW5g6&~XFotZC->#4!v=W)2heC2UC?&J9pA3SuP#9zSp@&3Ip z?VofeqLisE>g|tz2i0?V0l&T_X22_aycP92R-Zd z+qng9{lWXMS8t^8=i_vK|3z-M`}pIZ$98^!AL}P#fYN8+e}?Vjh^xHy`ccM@brRs3 z&gF%0>)(DSdge3d#PADhd2fiug~%{>cI_&Z|CVJ=EIsGT#3X zeMs?ZFp!_WR`H{=RXFs=rVm8U+#a7E~~<^iI)KUa$mH(fWIl@H#>UYAN7tZzKq^#pA_&rEVtOz-hO%1E)_r4 zP?bwqWkc4?)CK(+zqe$q8^6ix59Sl3|0A=K`=4*{*B`n6Szf@8`e;t`wm-v}FL8~+ z3H|w*?I)1!&(D6@W{f9h(f9-HC7cMX0KL!0Shld%rj=%51 zW%2!oe4D;6s?y8ZN%D8$xBCD8iT}P|CXDrye;v{v>*=(F{Eb`w|G?i5f@V6ckGVkN zFJ%44{=HuYll-s95~qB_^(z4XDa5C|p#P_F{S{Tb?dS0vZY)#v7w@6pV*P<$6{_UK zCA{8cbGu}5IP~6N|94yZUhkGa#{btyTn}f6PJLDVwQu#8e!!Fvtp|5%W#`(h)MRHSIhk;`Fx%c^4~!1ikY67z;6{P{1{Py zO8jz_+T;Xxu$&<`C3bG|0wZCwEZ*F zr2MOAiJfTwZ}XS`PObfK&p4wSm7X9wU4I})U4J6{fUZB#D}(QY4qm&|Gk(DStCQxF z9Hu7wUncoqOXlEeL z94*{neE;F~Tg;d9I2`vu-gVJI{GE9H&6(`2zh|?aiwk0sWgWk{(mQ_8<-cxj=hLMASepE^Z_K3qm`VK3!Qy|!AHVsz zKYr_s0Qr}AdoD35M>y9f?d3+vFZw3!| zqn$5Y|1*utUHnCdrpDhw9!FR=D|rJ4__OZG5hgjj$19=VQF3^XS4O{Wa^NpUzhmbo z`HS33&y!2_8XDPhm>+)&o~8B2HOca`?m0C4;VdNhgCG3C5B}gs`N1E=UpC(W{r=1= zzVov$@rOL$O;_|u#g{yXn=O*|fxjicJm(o*UN!F89H#up|Eia=o zN0HxUKVkvdXLS1u$RvP_Z@*kM?si)$y%;?&UC;%-K(~CFf1e+JY1gO}Vg;)HD*lq^ zPD|^n{6bTRXL2JcOYkP?hKSccbGg@F^?AjQIfM8m7p8Nu^10|^BrkHevV6)YZRxs9 z_41!8f8p=f{OZS_Jl`v?7@tc%l;?aUAAfhwSKfD@^OgQ3ghf(Zad1BWP;Q`mZyS1@ zAAjGFyyRs*t@KmpZ&Cx5U+m^;Lb^tI%5U9TO-R?R$ctyt1 zE9A`Zep25bTEy!{Vyt)G*5gPu{)(Ni>H*|FsV1!tQvD1yY9*o#$MN7 zK^L9}ZCC&Dv;4|$=cOuz*tH6eb{&%YUiN|)dqyGH_#K-!;QGH1_ zkExQduK_zDLH5ChwG|?Z{=ol%?1Pa-`nY{A=Z#z;9qfgHpWjG9?AA&Cbo*hzt5cL4 zs8vwxYrlE>EcK-GoYKoVwk*FY3H&RL^VWx}NI~80dok^cbUG*Ty+-nK`-?~r;6>7L zCwl$swfJMH|B>;+>#40J7j%Iy&@GK8``X704dfJFRO1l2k0)DXB>3xjiWd*;r8%ck zd5}Z@TtUC3%QvVzXr0nLoAw!zgS|25bc#>Tr2BsM(qa@>x*&UFI-lSRw0FH-)%@yT zQ|2#rko=h~2`f497op$MfxlQ=!b=YPnMYE5=>qsO+tEF8mF4f1v%A#v+KU{QX#y-`8yDnnwI3^>bc<;;-V`MnbyvS%mi)YSeo-io0DW ztNY?Q@{`+2`Wz+shv>I-)2V!AeAy<-tLx3mAHA`WkZum$m+{rzefut^`_f->eWSQb zmA?+3E_&Rr{K~of#Bx=BGXl!b7*-=;dO*` zC_ig3#V3dI3sc(Cq5Qr7{B+D?}_~M)7eYyJ2~(t ze#;fofxq%Sxk5VdS59%I1Ak&4#U}^;qQB>gxMTgSpt#Z{_!A8Ym300W_yWzDIob9r zzvfr-^Nm&eGJw8Ul`WFv6FEP1w~r^2enOWdZAZ_${c@|0`#s$?HU=Z`&Aad|9C|EerIT`(o;Filo#lMzW<8zK9J`q ztj|2Z#w>q(ApU7*&J4Gs??;0rV>C%4p(zu!AxhM-cas3d<@O|>6N z=MVO<(cgeSqn+YI=g;8tiwEM5iaCCPUb^?pXWjgj^7(^S1z!HjAFf>aS=V^q7oz;4 zuc~s=`NO#doj>~uB`1^%v#}q^=Md0(7j5=@U$&I@hqXM`%U{{J%JGMFLcRQ8{~Uav z{H$@R{J;nF6Xl0;*^e;}zf!FX!%2V1o3AuUY$Si9mJo> zH&A{XGn9!S}HvZF*=<;lq>i%_`Uyu$`8+h-nRGbN%8f0NP*jz zm2!Re_~iPoIbQLX&hbFy`g|t+nH(N;K4m(;H+Sy$@`ZKnNR-y8>05nR_l*oEKZRl4 zH!_0!6oz%)qLJjMF#UqO5dLlT;yr%*cN)J*{8g)xu%Bd1@aIq3Pm-E!a)LG4&J(_y zg?y@vRqw|Tp3OI!(K9!U{MaOpBc&X;bNjce{$U^QzY$M1=%+#?ylC!1b^pA4!`6D`ylquEk%f& z{#_=_lPi9q?|zWe&%3KslK;`JK#!ZZ+Od#FsDr2PSh`9rG%D*2Q4 zdy6=WbNRd{+5_aA&GJJy^aysn5N_}T`G~LW-v+*1yyk|8MU0 z6X!!OBizQ?Oi|-c)h4^`6Y#RWW%G^x@2K?cPTqD^Kc@`-Ml*kq z2lzlfMf?NlztrKVPLe-|%Ma_jLpo6XORp+M^Zo*BvOj;LxrmE7U7VA4mZ*LY`A((< zFwzKNoukSR`9GrVOUmU3hQof9R<3IqXY>W{I1c$h4h9FTo$=U1lE2)ED*t?*A2qOO zZc8uup5Xk=DgOLDq49SM_XE0pTbt9td2jF+;(Vj))jBKqJA(Pc_ep_w)Cz@D@OMk8 z`g^Xm>_fTMoG>H4e?G$b6mb13y@ewFnkk0z{L6U4pTD7;UNKYH<3j!5;KcbO=oQ+R z@i>=sxR<}bW-30Mu4>+3a6&U#PBBO60r)^Z5&luY^zvU`cqy)*%wG|g-x#u+X2)M5 za72EOi*^MR_N~ER5vQZ`r?)%cZ{~W%2jUx?-4re-goinNz$|Zklfyw!{_u~hIP>+m z{ierQUUd|e@28zVt0e3%I|2OV(D*`L;IA2%x2~T}V=7&P>r=LROdO+yqu<}`FdwXk zkjK<=Z@4~QF*se&e%}nf!o5Ee^{*klKenz)OYj$C`_(8nP@TW+&^P9TvlFm`4gN4b z7{hkYV7Cx_-_79;x3}Q$9xYriuL6y~8B^c9#oa$(|EYX`8>RIm*l)I<_OE}i-;DCR z?O`sblT`Z#{_fkra_4>kxE|qf@CQ8{$muvNC)BfHTDXY}n1VmVKmU|rBi;Pvq5XWB z^z-d>(X~|Jiaz7dU)Ma9F!;m1G}IepUwxrU2Y68)Xg?7jeB+#<(_QtWoQ=TwNh4p$ z5Ar}h+57`^S*Xjoe(Qq+rmKjKt@x7j1UU)lh^Jo~LS^o<~LWpclW zYyStn?`j;KE~LAi26*y&Us3uk-OtCXd4o-B^wKS%^xX4`C@p!)c|?@nc}my$BIUb<10S*j$c02!Xe)ezF@s& zJ%s!Ueo)~ShofA-Kzy#31{h%qKjJrj>WFy%mx#vwUrChYe{<6QtT8P${#x-q0<)Qy zzxv#+S`RC|0)KfNKkBIX;P2X%ia)eRz}sI7Z=3+Ge%NUEIDEuUt-g`%XQU4A1LyL4 z{#P1*zL&J`DyFB#-()`TZx(vTXVFy(uUW%8K0BAsp_>C#IGGfTUl$>~r%ES}aZls) zqinA@s__H)FneN8ulHiL($n`lf73gZM(;V~h*4cm0DrLm3jI8l#}_zvj^|Cb&x&xQ z412LSCxjUb?7}+FtMW#FGMUdMm>ax$fP4!1M*-8v9{%t$cl&9vecrMOq@Nqp6ZTbO zGyKaB_Eo{(b0fJvaX)~3^>G{Y?Yk0=&*zi#_<;@_ZVgj>ankr$LqTt(n*7kPTB{J&Qkng-VQ3)C!~E@=ta=_%Ywy9e){^0 zZnqERiglULRLOzIx^hXshe9~Qhu%9Ns-yM4{shhjs-t)%N9=*SY){ z7v_0-`x0WHsL%4-;ALbB6ziVy+utz?3#xx8H_&D8-|(?t`;t|lGKz(iK1n->IrNA8 z|DY{y7~4vE2knP*`+*H74gEan6^V*Mbh`eo!3) zSx@VOA7+aFebl(y;(eeP&Vu&ercwrfSpT%X_LkobJno1t;c&)rCbv5%4}-&DR}$f^ zI2`9=5Dt4F;16;DJ>kDOrn`OV$K@B_-tA=5UoO|b*?VdH*)20+A2s$uYUQ^J^^f%`kuRT*27jn$crPekSM>?} z-NyDbqWiq-7TYwtqq)d{DasG@x)I&u_Rq>}Rqip~ACf+h`Um0JaGyj7K-!3p)mK82%HxL*H|E@z()E3GoNvMXXX$*5`~K+`^|!|- z_fd4Tz5a{neUJM-_P-)MWvmEKPR9Spt|^?fBEejqY^6S z^p1Uw`#Ej?_kq?PqMWMocntxR-G{Gw->?3i_a7C}TB`Vx zcJgvW@_c|i&)}Blm2^JB?N98-4%7a5azlDidh}jFbRa)D$iGDXCO6uq`bhV!?hWH^ zJ?(Fhey3~0xVwb#xYKX5g!C(%E+PdDB5E{`?i?BOa7wyN&A2?<`Xp% zRPvoAlIO!;;C2$@5#)O?^02x!^vI)#cy!Cd=+RKP`86+3EFUG>7u3t^uk8^3ejIR2 zV*PxU_RGgU$rQ#nTt8m&*H7bS^<*(8@OXgzoTs#K$oaqM_dihmUzpQbu7#T%zJ&D| z;{@=*x5YsGVA?KH+(XtsYntK%^CNkF#mdeS6WKmrl+UqXUIZ%7wa5$n!9JOD0Q*u# z{3#)H`%)H%>-MGkarjE`&FxVUhl6aw52n{<&#xihuQ*3|->=B?9Y(b*F@yQD_?!p$ z0+r`J&PB=|d? z``_4LHC{%#`aU}Fr|-iu_<=QCUooBwbNFu7&)AJh4|M)YIo>3uQ>}eF-Thw$^Y;?r zj$WCWu+Jq{`11$*V&JdAcrNdoz4P0RTpohMb2A8Q04I{^YqGd4a!r4>5m>yzA6YzOBO1 z?q$cH65@KcuZ8d&4#)Qs?U>>l<^A}}Djg@P!XYPqOxXCrbn@u;;`{3mjP|q7XS6;s zm+YHPQ~bqZ{^i%Ng_l39-#H^y{zx}rpYJu54#F>tD!c}V=kbj}T6l!RPtn3ru1K$d zf8;amQ!_HyU4Bv4&lr`TJsZ6^{eIP-KiDrr`F+IeET}igH;3Q5AROg)=V#t>MLFHh zdVux_<@6rw0m8$`fGOmI;Lkt3VVvwA^zWNH#Gli)su;0Im0y(Qi+%xg#PQzp%f|a1 zp64N-y)1WIoBs)Jj}Q+2YG~ha%|Ecwgx4daI>$xVe5A@WRzQ0=Pr^Efw64uXA7t;RuT1op><|=>w zV4n;68DyX9I$qz!zD=xr6mz*^KOXef;C+ZD+sQ(?!p^>(?ccsYKMm${-D9VD+85Z* z{PKG$m&>ap?O(l~8h>eg{sFik-22$hYVKqIQFZ759EV# z10D4KmvQ@2+0#|Iqx|;d(V2!&vi!vA)c8x3pJtz|(&fkYvA{RVFP4*9`Ayl}X^y-7 zMEPx}^JC_f$?|hj+&< zMXy)mAiexVjnvBT(x-mE%dh-Sln0!*CGEG_Yg6M73G7L9)w^ zXWjRezt+q5(*=*K_7mgJ(`f&7Y;cyizquN}m~Sfn&~Jg>KULXZDq=>;dHw)9;^6O+ zm`dN^?*o{8WB6>&m)A2*#+T1{Fz!B!!_5ZXb*2-p@%qt^Am6V(pAr9l!$56+>r)QJ zxH2QLztwolpFg?3Roa&Ve;7ww+V~9PXnp)yHc9aZKG2Tb$Ni`=RO#p83>fR==1EmW zc^?`VQ^xLd|M?sb^o!P2-*TT%oy7Z>R%_=|E0{mL{|626_yhedCoiHL9<8G0Z|Ki) zJ{9MS!4J--;=DD=;Yo~d)~ogs;}aYC@_FkpXS^R>-~0_Xf04RMp1o&dg<3sW{q7i?(%P6 z5Dq?2ezCzyKY`UYKp5RMAmuql3$MHEWXN=Ns;Md=OquliUd7R9CBwGtPM!Xn5AF)ZZ z&s(pI#y6#|^Z)NF{&|=|9IH$A|I(Tsk@(dE{nlWmZyYRIkh-|ve^n(Q?sNSE$nU;i zSj+3r?c()sci91u&z+~Gk9>6sy#dV`KIV2={uO*)9s5V0*+b)sE*Qm8`G27Jht`A2 z{iB&Q{wd^)o?5KhH$xl$%;WWq*b>#xBjtAd-dk)|;UtFmMdBAuC$eAmRZCG(zlBJQ zd$zX};n{{59wYg6NWSmyL;C^T@>8ko{~zgt-nRE8`TdrYQT}+pSWoUBT9Dj7Yi#r{ zf60GU;y;hGIur~z?5U#jiTjZyaQ`^VMr`Ulf~U-$h7we)np;RoNKPp+&s(eM3vVI5V& zoeNbvEcf|UqeQ@e|2d?8(q8J}jpJ@svGNx+Z7igF^jzgHuhLjZcQ3~;tJPRY_hq@l zA8y`QNVn;3icjy$D~@g)cdb@&`mHH_awoAr+?L{#`-IWH-fFJf0 z;fFm%`1@8Tyj_!cbiUyS-=NcabQFI4H~vXIVdtvy|D{y&pS15L^jEMM@V-g?OTw%nRV+3>5AV~{<6dC2ZA)><1zq49bi=RbFZJX9d6a)ArT<&o%Jx6Ge+c}S%%lB3 zb zxF;!@?DqIe$?tBzztqxms5^2$zgByGrxrg|{_nVUgRlO}?{~$o0ZRYnJ^(lWW4clM zTs7G~Uvrh$Kl@a#U!GrgT*((F(9yVFaR z|4kcKmHU=0|6BY% zF~ak|pE^+fYtobUU$!Vc$l~WfiLLl$GhK@J$K6!D;b5VkFQ(yd{0i~@*4R5r&IKI) z=K}A#DDZZ0yx|vc{Nb>(!uz%^g<;3D@On3@oD>Tm8#_&@Ppbt%#V6^3nD;|K3T> zi?GfFd6wMd4L3NQuOC+7*!N6q#;@5NAK}oK8+vj4xf*^eaIx zEd&$)!OZ>mEqouJM=)6xOX(Ei)6LG!So<#q`(NwK#oYJlg{u2s5I9+HV0kqoTlc#Z z%2?+n9Ovs%rb|-)ajh3wZz&gw82;niJYM@>Dz{6nOX62x_}S#6>V@j_iw`B+g~k_t zRB{;&n`^IRlMg+ z`H6_1aUt_Veu*!a2YgiihcaIAv0nZap3%6S7U_FQt@i(y7SzK-V)x&CxM2Bjtt0=` zc~s?7`QMUqUg1?P{i(MmyxRYg0bcx)9<9QQU&^O!@kRcNJTmAlEB{FMyYPHg&zsSX zejm&KM+=t!O(^FK<)5nt^;Y>0=S9l@U&t4Ym&)hDq+jBGxXSbI5BmBd{F=&eU@sZC zbL)P6ZV&A5r+O#*|6_4~P9b)GPV{l@{Lfv{6D*MKuXTUU4R`qbh>uv_l#R~irEg8P zqdSA}sxSPkVu@c7F~hDJxLdsa>wcC~=41XhFed-~#OGh>kgfX@vZPn_q6<_dT&eqP28;uT)lf07+qzYN)u7+z5n0h z`JYYwfDv<$|F^2|Yek>dE`Q6vnL_!$b1UfHfK~iBN3;v|7?sZ~`#JSn%4aX?KWJYd zA|drJ=^K3ct2_@Qe>1#4x=9fSw%Wby$9KD>e7~-OweWtPr`|#Te*)f5)_%rk3zmQB zSBLM5tw+9;eoU^5o@Rep$|vjd4=jK2Mf}?QRo08MeR%O}C*rlA7%4-6Eq=-V;Nim) z_n)OW|DR#+tF^@H|K|&qf9ZegeE>xtmcRPHv$+40o9^dJ%ICP5KA$DxmCrx&{@e70 zKD_dI$Jai5fq0d#;+L&@A-iS!qlU%nf06xv>NV8=xz+olqqYUhzsf(oXL9~GgyozL z_Jx+zIf(O;^TJD*Pw8hU|I0DF-knhXHU6o+)XV=AGD7y8i|3W+f8vj${>4GJWBi+^ zSuWbb7q#=R`wt2v@XZIl9K!o?j+^80>H9oil%J^uKkn$9ywZIJ!%NRTP5Ryc#Q9~D zum7Woe^lr5H}vf*OMGZ4-{a|V{G(@=^}iIZ(J`GDVNO4yqr3hNA7Avh+WEKhuVxD6 z|G4k`a(=&`4^OnaisydS|EKBkk8gc=;WsA!ofUn&YX75%*L^~|C%Bn--5Vl)Wrz#_;+&wU`x`(=0GK2d@6bPrPaj-uv&1A|A)~AB{`iXRLiBt>NnJBgtR;NOEf*Np9^U z$*uThkLM4yn`sOGqkKQI{99Q6zup7&|GW0?p_wfIuWQ%;hp;bTrcnOx+-lU@FYr7m z(Utq3ODrFiv(8DW-l|`C?9Xf$!92GQ=fw)�bLruXdsMWozCg_vl03=wtUsW>P+% zlK-9lqLu$b?wtFX|8HvN|C|xY{P#Z2r(1etil-0f?ZT`6Z1!|Ayylg++?2$tCNA=6 z60dgDFTk6ux|R`N()2v9aq{r^_&1CF|MtUhf95~iyO~!r|9`KY|C!4s^ZyacP35dK zNY6)jrCaw+2(Nk(>iMW%oW%28X`X+K7(x03&zG=XD1O;t{>L14{Xw|Kjd#m#EP)}shG0E`luz$8ALQ9d zNj|?pck6_%&R{xqd4KmiC<*|&2jPE7 z&wr<5{r7o1egJ=@J0X#YEp-(>)Q3eeX{!EK=Z?tCl!0lSX?KUS!=gZ}SB+RYi5zJTqk6+zoz=r(YJ zf4Nt{JHYyMTU=&VTNd%%fcIMGbSp!>{Ga~NHW${)fBzi6IN9g_6r`^!{=)nhx9RQ- zl>avTHmH|>*NhE5K&|}8_j{H9^mg5yf%0DhZ3E@M09pq1^549DcV{4@a+kf)BAJ@{ zPtE7~FVcJX`hU;ymj9~$!%-j^uWv7aF>$x{`x|FAz`Ce^=9hmQ z`lEn)_Zzn&tU*0FpvqnLIydI;ua*B*uJ+k4t>xF#KL1typFHxPiRph7^GUbfY0ECz z;0*L0RDSUWXQ2KgTC~9#RN7CE2G<5qf7sls!5NG|dgAq~K>oM(ZEyy=ewHln7dG|o zs2k`5WR!2&FP*e=&G(~wp5nhK?(WO~S(IPA{cnQ)INtwnd>PUgoObo^0#&VuaSSHPx;UOV~{hbmw)%&AZMWb=keQMLGz#gC&B|{lyBL8|8&QR zwep|F{XJ&V6MXsGJyKmU(?v0s%@3>k@+xfreLTQfUna}X+4}LI>i8Y%$9)p$dAv;h zJ@MxS#HZh8---oY+->g$#@#p1^KLgBx_^M3g82!P?X+H&ebS>_>=i%XlIQudbnrfB zs-v$5m{ydH>=fl1+ILetIpWba1 z?^6DS5A)y5?`6NTS;vO>`mglu%Faf<%xug*uk7gBc&|s)+2_9#zn9&%j6eTVAim!F zf}#9B#qg4Ir?>F)43)d?L2y&OyjbNG$_tghQK$QLr{v#h($_nM&3HRBbIihSJn^#6 zJm<_V@%&Gr{I7T>@;`qe_od`bU22!VtqaY}^43rE+=zBRl1^rIFV9r|XRy8&h*v(R zQeQ^u9)c&B(=c+!fr1R7t zc|Js>qj$WDWBfc{{8zd{yNQU1_Q=I{-`w;k+Ip}bB7dT@(S{DH@+Y&%g7GKJui_8$ zD?aQ*KIpw6@!@Rdw?zFh@u3&xWa_V#kA*0=6kdGj7T{Gcl&;WD6QOi0b=6_p$J>|m zYgW4w_2={I`%$^B3&tPyBc+Z>`I72YDec==g+JU!&^e}v>D!9>MN#DGpc^Zjorn+l zaO2y)eoC)Ld`K_t%UyWY50d+dg(p7~=FQ2su4wfq6THvVVR5v8G1dDA3Vc^d28Kd0Dz!Rn9R=PGQM+@C!{{Sx!5_^(oXAzt_^_HE=awOU@xPI8Y^TeH3<=y=CM3+nTc*Ek_jAmmoL2i6 zyUfS>(DO+0tK`O8i=9CDWtGB)@l&=bQRx$?r47W7=6ZS?N=e4=VQX ze5m-W^>X$#tY5Fk8kh31=yja;oEat7v*{Zm#`053^kn~j%>JL(n%{pmnct4CuJ60W z=J!j#n_O?+i{ORZs`L4Y$f(V~^UX*rVBcVVeqi8khmIjWU;o|7m|xscrFR`VloIs& zQ^TV&Ob3H4zSQ<`1J-`Xf8UzHdq{w%&;b~|+`5r7cx1CA{hxBiW5gHPEB<4x0{Wky zoPX^!3hiqp&GOJ!_w@Bgx;L-%p6`d0j(_((mJ;}Eu%Pd;6mT%vpx%2d zSyV)WdS)~|ko~f`duG4at0FfM`%C2#C9Y;OAkPJjYr}97}c(9Zo12-Q{k1rtG-F%o0XtCw)m5#|M=6dyFSjJ zQJg;?`8SlqC+%Hpe~w4JYV)Ucx+f(5YwBGbX@i ziLYlV9;NSrXI2<#{F%dawvs=?r=ic;rrJ;oeHM&AU$eZl4pchyZiDir`V-z)5%Gi3 zOpn4hiqMluzNLxZDd6V@#CLx^8Gb|2sbd$2pF(@WM=l?4U((OYn10efJ5}_PihXPI zNBT*sKlSvJCQb9{jL!A*=QQJA^5bOsResNt52|mfCt?3KnsjcU{!{uSKi^h8nf#X| z{pk!}&x+_Fd-mZ2%J&<$Uhe7WeUt&naISN8{3`YH{K3#rHnOF^qjX-&`s{`#*B5)` z$RFzOsQ!!z@cl_&dOi6od%M;vde=Q4-g;GU7XhT#Q@XmhYX(TKCv`2{*YvKO56{l% zjeBviSz5=WiEpAkclGuDTy9_F&kXj@cD}v1Yn4B4K<)Zt=ibeLon^CokYA$RRsp^X zJ<;6{{3d>%lWPEF@9ax z&V9=I87*F$KcDXIV-kPv4brc46a&2Yb13im79Yf)HK|AD_V(>j{E;4A`c%J;P=1Gy z&%!q`;?V&(`qQu%;_uq2a{qW{$=dZt_m8Xo{FC*hc(f1iV{?^xqwfD7!uX{>S9sos zpP_VdUZi`*TZkV;d9LuYh*$kl`zjKD#(nJfcn-n(wVyMpH?q&Vc9}10)t}KXas`q0 z{jBvf-D3I`>OV6Z;;sxp;rGe)O+7tHw-FYO0QILi{5DX38f}34Isnp>Y{74XgEvQg zybJdYrF)g(%h$bI^{w-Ds1~vNhWf)TKeLaRnsnFWk@zC}cRPGCt`>jt&zS(i{9j{n-S0 z11vCqHU)nG75?CN4p8CG#^4X2!XNx@0aW;dx|Rq0xA>DQ9koa;{xtp7x3A2`zrdeS z$QxjR`Lj9r1E}x^zjJ^Je>MYu02Th=cMG7xAMi2{_;2xN$PSaXtHqx!(rlp{`Sqlo zZ1(=8%R4&*neqGq@3QGU@bg9D z<5!XM!&l$J_fJ+k`)KLf)jWqep&Vom)`qMt@?Atp`K7_YtI*xqjrz!2bIUkRjcy7 zaR0vxmbP(yH%ObEAN20hiVe=-;M;tD^M8bX z72w`E-amUh^sfN7J?rTeS7~qtS;)WR`c%hLew;r9-vdrW{8qo*F^C^k>!CaW;Lt1jUh0o%~sw)PUv;QQOrUcm3RmEZtEP$&cb6k@i)p>1jaxmNo}k2Kh6S;eBp@s-OMWPVanDi$5P>zG|j8 z$*%)VPS_n9N-2+aT7N~4^>YSez;N-#?0WBOx`+EY1HG@AdAy%97?D9-cwaM*-v(by z1wThu^S64z`zwyp^rXMu-z?4U2YtzY6@9LjNBTK~J=z83Hzk=~X6q;Y&w7vCbYLz1 ztojDO;5j5)U&P0EtuOMJyBPe^{O9#mlj+g@evMo7at4|Y74h3Z^PyJ!Hqd-1yCv@b z186=JZH4&~;MbZDsa?t5Zr!_T>bKkbQD!nlKmW0JqC);yS!QqV?F4`4E5Kh zF5~&rdOze{0L#DXg4ve;xZB3Jxci;ey*qbfZ)fnp`bqycaURFQH@?){8JIq{xeGHP z4+Gq{wRfj(>5cp2db>Q+Q^Icp8SzE-<02%Q`_JfOkzH2T1EJywn|3m&nGm&0^_>*nLeUJe0$K8Om0e&t1vKD`akw57({x(?kmu-6w#tYZkb^G4W&vUE!abA3t-R)>ksFn2I z)46!cqlk8UQ_n~8tN!$u$Mu5n5%C|3&UD}Z4I%!R&nWMD#qV2=C;o_MJzo1tS>o>x z-V@6a|M>y_d#yW;BYqa`7TPad{Q4?!{^~+~ZH2yU6eyYaa{<~HrG!xT5EDo*BAL&PlKl5|R^fNlA z`{s+&yu73fJ)XGLhu1z}iuf_o7ves3MEw5JyW>7`Bk`5{d5CW!ek1DJh@WZVuV=i) z_k203y!0J|xi_{PSKi+l!u@gS{~Y=$rDI3h?4jbAT6r z_LANd>GUU|T?2eX{NnE><5l@DPrTZZ_?c~=Y+pl3r#jMm#8H16)~|fNJGkF&#GrPr zbOa`yGHcf6kM1{WWw?L!exr(9g6G&V43~}csUqSRBK|n-!xE2aK-ruU;IaHEn==Bu zBt74;A1J<@)Q1%RWFNQHN*i1IYGeJ8esA_Xw6CM8{E618&7bE`&&-HF(g)K$RNfbB zA=S?yzsvG|9feoDS}nk1x>q(|QtvKeJ4zAX_^7Wx(mPUp(|u8y?|glU82%pi|CxV! zyz1Lj`paH_p>LN3!!##ayS#{XprpYxBp}^=8B$>96WNNAK%gFZN664SjnT zABGCQx1UF(`9Xi_`;iaL#E1GnqlsUCmgkSgukpmMNBt$;Gc$$wNc8#6QI7agw9~Yg zKk)N|%$`}E|RYV+sEX`Y_=)8%sV=l10IE_3)X^yQKxKCEv!=6k=7e7xe1_)*|@`hLdtL(1>N z6YUDlf430!b8}~QaO>^Q@)}l~KkI$qzZ8EyJk@`nzR2@IbRVOhm1F%;Jz4BK@`v@O zKR?*^rX;?R_<@{fO0P`q;aTq23!hQMOD5~5_&ntbPv2#IIp!H&!^#|M#uRk-` zzGgzckvqP#d-ZNlzqoE~{=B-b@59BPu)LJa@oeu(zxeYS^QCee!VjKjd&#v=zR#n3 zUNzrtV))V`KApm+iT|8<(HB2cR}PVODv zUsK{e_FYIvw%u!QM*PwHJ*7^5zRKY4ChBP@{j=#YmGNpE7oCmRZ!7#9hL`?^@Fn68 zV0}w>_VLamemUL`qwsm=|AsSs{VAN_+oS0Vn^C!Wmvb9Fz2g)+Ul87JYxmn1_w3|e zf1LGjeP4gre4T8&_kLjzOOxO{=BvPQNAbKH;KkqZgOhTNfAyaH8TdXf*ZoiL#UIZE zt9_-HT;Z{J`zo>jS-EpN7kwYQ|7Jw(_NDu8RPMj^{WszMTEzT>_u@1XKlMsqzN$Z? zh!5|}X(nFfCi~dFTfSUNzHu(s<)w{0Uu>PeDAx0;e|G*^vMRH6A4p-dMV-M?AER6$ zzgL)vdt<)Yr7BPU0u$eOxzD1!PvM`y`Rn-ie@A$`U&8IMXc=CDe*yCMmdzI}gTunW z~@QmVIRmR;_=^%{KexVyMAJcZ?X?=w|dW7{2TQ&KRw^C zH*KAYlvnt7b&q8FreR_qkN@H|lkR@UCEe{0PP#|qSg?3!_SN=%Y(M16fWJ8Unee|9 z>51=GWn0`g&mUge!@5+#5dV<&$#cJTIUn&e0uSCt%A^(@tiul zK_^@1*1_JgnaXjl*v{8mqTRoHo?krGdG@w%2RVa!=h@qE*4@DNZynD6r*M|tU_sBb zr*M|tpq>%mWM6agjSaQx-~Nx1!XbYAw)5@r_WK(Al{d397bo4lU-a%$g!fkfr*E0` z5BCS!8oa;z1{DR6o{ll45$(0-;$?UX@moAC4)4PN9-rsk`MCp}!GYJK{>1KA7=rlY z_4g?AkJabStNHI+=iz-Gz(Q>=fW|<6wb?Nk-$u7{r*J(| z+_X0TPC*khBmVVyF3At?>s;GgNxX`6ZOt>uKaz8kS#FxQL%pPI)(^r#q{q?`-ye%N~#Q`*qd#MRHp#SpDn9cAk6Y zP_Cu^rSl`c9!Rc*rO8v@v5(jk-qA8RHowG0o*?- zedXl=s;~5}yrU}jpbM`0)a3Uq-6=ol<>T8N15)9o_HHia!No0W^H1{oc_jG1U4E~1 zTVV~KUo2}9rireI_PtwtG06JAd6d8FeS@dgM*YJf;(NG{;;6@{`XSryeKRBe-N*fg z+=YJK;(eXFiS$XOXa`!$}82pU$(3tdUbXGB)6@v&zeWb zHoh(FI#u46ekSev0e5Xu{_VJ~D~$5xq4K-tF<);~-Vqb}bNZ(x+@BZMKW&3PYFz&` z4gJ%f(6888q)nHJ&z~cnKQEq-fSXqDcPKX3?*Fv^G>>$T zV!JEONb*l|?cDhuuXc2j-j(2ddpzl^MY*Bw*~t z7fR(eop{NqGw1m51>$9m;t$#1zrEL=;{0o-{QO(f2R*yGKjrw&v-&;R+Mg2t>fP_1 z*S;0|r!GuKTYy))7(w~D`b+hT1(r|lGoQbRbTqyz{1D;`q#s@G*AptYeYoFIBAq6N z*ZxSc$%ju9FRPV5inO2HYUR!1{L7Jl_C9uTmrj-az4Q*X>tDz}?Q<4(_x+>d4eN*a zSK>ZfdVF&H*oyqiaepr&p&f|N^z{4_pGN%M_XD}zd^*LauZY*WTlK8}4Gcft^QndO zr(DQ!i2FselmXh!w3YtD2Y)v?&OePG1;{Uof9PDfUp}{EZT{(g`FW(f3fFr|ucB|_ z`lk4`Z@;Qf_i$gK#C?Kg_(7;IB|e?6^)>6U_!KSW`(3q1)u%oyb6&>qY0|&4tH&4b z_xu}0{BjJhdFK|y%M#yY-~IH;d)HcjjDOsJk$b|Ax10TF>yN7bv-C6be}cXoK>C?2 zf9&ZDT4!SXj>(g<#+V_vPxzvqWw7&wwoWNs-V^6%dRFvVvv>4#2GVCuUDwkYNIx^W zrl&Kgr=OY5_jCp`u(4X|jq?U)!VQo9gM@q|c!7QtwV7Ee0~; zo9wpfx6Y}>zcb0d^pii&zrs?`H3W!%=>bSDpdSBu(Nt6#j{A+DQUH~%Uo9u0F z%ulVwzuU+^_tekxuQUny0f>Lu15tkf_4wx|A}s*%ujMe*SAh5zos9GY#J|QPkr#mY zmz#q63#iAxwsAe3fz`hZ>R)CoXamH*R!bY89{(~(n}PUOJPG9ukP+Wx4}1UpLM{F+ zJ}hbh?o7{tLP~K=VHK*)4K}v;@@u&#CO-uK77mT{hdKOoENe1dC)3D zFYjM`puaOXr2u}Tf5{?lgEaDE^6$cZ{mU>F`8Rj-Tm8!*eunmVB4oe2MJl0RGK%B# zoQFYgt&WxPxx{g=GryN@b&8u$X$=g$_?!rF(+>zMNw)sTc@6bO? z=;tnu-3M3L7iXy0KV%Q;?=t)MbDg_Z({J=nLVnGbc@g>1@7%-+yTrDNExyX$gy%BZ~Maot8@4KB|&5y;~c1hB`A4Hx?kJ(FD_}F}^5e+`>ulvk}o3wt? z|2U*a{IxJg&P=+`4g80Jems1H`BB_GByg{2^Q~BYxbMwm{5?NPx+@F~exR7kM)sW@ zzrQIyzi*~|cU_c!X6JU5{lQGek8_f<%GUm1c{2!J@(AI}jVTCGGBY0tf2uf+R|5zj~5u`K$Bk)5miUt?w?;+stJ5=UpxpB4mYx9b-;fCx- zyst6C`4fHJh19c_JX-a4IQ64*J0;~ITL*g6X%aeiEy`CJUeWKx@RD<98GgowKD^R1 zmH4Nqhnt(^*GC27rQcmR)8|+1Bb5KlV))P!{VDV~XZg3zjq|sW{Eeogzdy4o-z)9A zVEonoiuzf_FTFm=E&9_#_uLAPX;s-Qj|q$wndUD`5W<{DMzkk?#ZS z{$~VOt5fCv%Jlx8znHd^jcl8@EsQ_Kw^~py7~K6ydJ(Vt{x?OBt~VyzV?=x?A8sIi zI_80y*f_s)x8%C6-nn#y`r)&HuVN=~eZb>2w$F7&9J*P;AQeAl@pg;#wa zNBMNPue&JEbt$~+k?!5pxsyi58|tg3i4XNvRe!^FnPK?Q5`Cq2jm}LSjK77ed_8iU zpB?fw`iE=rzIQkFcL&tw@41-QmJR#gC+#bJ ze>MMY{seoYEwTCUfwlR2oC=H#`~Thi_qp~f?mubOL&uNF?z`%thaEC`>|xQ7(Y_PM zP8v61;{K~0aOk9ot4(Nw(p_Q?t4z2F!8XaS^>71G;xzd$Nmfj z_>S6f>#gd9IBd-R(LR$V{VpUuVdA)bryyQL8+sjfL^e8P-~E2K@7Tju-SLR*zM=o5 zgC-od$wAq}j;Kla;K`E?+jm?|uQFryJtW`)v71hwI5s+3HMO?gg=pD3ZaAw?t@Uxg zj1O4i{QhDzTu9n0@`vqX@&@bc=Rpyputs&9cn9kfjR9uwitvTl`Z$a5hHu-nhcn0o z^x6V?p9J*gn}2Ma+duf;^+~4Z>)`vhh9=YZ(FpG@4#oRqfFnj?99_7YpDv(h>6Ixk z{f(n~I0G5+RrY=N9R7VR{@VK=W+J}(VftebxmNhQ7djw|t8EzGAOb`KygS+)zX3}@ zS>F7!TR~q8aLb11&mq6cZUQ&pKDrxmF@nLNYkL0_bjJ)r{j>4tfQ#$nzR+ztmI2pq zmg%?b)X^CXVf?8b@jc)kWYFR-ZPw8lg!*KmKKy3)`}pk}GiQ8ad9U)f3H0Ogv{Mj& z{5(XNS@3fkio^e^%L;he?bj&FoITm4m-bpmZj|6hyZ&8>7S2g<{k}{eyg!ahYowoBKm)cZ0+LlWvZbFZJ@gTeaK6R)EQ7o5(i0_`eeQukdNo?|F;ozwqM!-Yj3?e~JHj(wjgms-ewXc!Mz2X59(@7#NmcJi;OxksOH8xaI&aKL4KU@$@*qj}`HQuEdLu z_<1DtKZO_H&er}N<;@lT+{JoJ`J3urgXC4j@5*>5C_L$8h(9#IXNlkK1JY;s9P#z6 zMtUee4?VHcEAjP9E9J{=n}O$%c1ckdQ)<_Ld;i9a_1(U{Wd+^?sd_J4@=@{s zHs&k5Un4x{B>so{&i$D#>D5VI9T6|Rx&rsdRi6$z!Pmp$)xNya4BzoJkI#JS@uP_E zNxbN*UF<@=GtnQ<@E@@pb1cUx#6Q~E_hXXx<%q9mDc`dHw%M&U^)p&7^b~UY`1y^^ z9~XnBfclLMYy+!bhsqpa?0%S!Y+QA(j0DE_+lD}17@sfLdslJy^vJsl$Me>^{Pr+I&+u^;Kly`cPWU2(+QwdR-hzKfYmU*GT8J1&bs zjw=37$9m7?oUKm`D#HZkU-7p;Fqz(Mm#ykUSePv6E3DZ)5Z>(kV|XtaaMSlm|I`af zx8$6KCkHu$nV6i}_cD8uAFVGUeSj{1@c!=gLC#>QQU&a27`HhJ8vHbME`W@LPN4K=ps{fBa?!PU@^y?O@q2HO_6#9h#>DT49g#H_# zu72G%kTL+IUzZ;R{ZN4P>oS|;H$eJz=`BDXApJVk?|S-mxnYPOVET1YpXz!jeJShj z)>wrDq;FS3S`F&z+aawPz;=5j+shrGZ)quP4}C{KGwGLh0zE)IBfiVteW9;Dti^x3 zpV5rY9h;s4R8}he&xMy^e8&9>ZZC91fIoNh>1jEjyEDk1;oaFQ@f$E1-&=aqCv|rQ z;d#KWHzF=ns>W-(mmxcm^iBpnKnaoC_l*a`AMj_yZ*F%r;s=~FBFUc<5ueFZTE^ge zz$Ng{#N==NApbJ|vfvHEw;Av^dM)^E@K#6f-?%s81KdjgXpipB;AZBxd2DxQkVAeg z{ViAEd%&2zy}L9HX$Q!N@3JQqo?WgM|Bpoa%7*es(_@J9|GjN|`Yg?6zJS~8R3q-F z1uTvzBghZ(>vbzb1Kj2>GZsq=`<>men=?2G@tM1{Q8#DsJ?XVUk<4HM!kav>knUE7 zA*7c=Sc6IOUl;y>S(vaIJvSVT0<2~%h}~zu@V4GRhqM}m_2G2n*V12D5%d5nkM!|p zR_W#p&O&-k-!D51=?A3FMEKbK`4>&}^iwOq4PHdxKNODXQHGsv!S``_F5L0_ z9AW-3`rq!9EdMvh`TY9WAs<3`{Wkmh6O!(`1NVV%C;hiUe&XMsyGb&>Y3n5YKMLaS z2>p@x_u~I}{>adtH0$s;ZT>%t=NG2({p`ZokU#8Iy}z<>k}n69q-@E58aYmlztE3I z(H}U@GM<}mOX5}j6@4JCLsh;57`{M#RILl7H=^~L_Vu^bc*=9+5#zl}bP7KGLx`Wr zbK6QMro&}(Yk+Se{%+!no&7kTCjP+y-%R{8&NuWEh_%2)PFY$lMW&Zo(j!FI>M*RlOTg1;-?_#-p?88SsS?(*z z6Id=Kmhbn(>%PKy#LpF-Q#_qE;yZCYRXWMz=Mz7j^C+D=bL7Y3)N|3ktmZjANJsOZ z6vMB?@WKxv{>)wdx=eJ^#OIm6sk8@u`J$aK9YCjRq5Nm712CR-s_Iu1rq<@at;@`a z|GlMWH`mVty{~hpwRk?4HuHLHs{iwJK5~tu{O5=#eR!o`^|Oul6&6Wf^|OKDg-r=Nbz<8#EfeCWeR z$NO>@zf<&=ef;FZj|KmGJD%6o`}E_EMg8BaUFH7e=v2>ti0#Wpw(ehUV!i6JsGo=Y zVG`y?Bip;@Awn{_ehz|{^7?UkZ@;x1`uFbb-LB2veLv!}@H-%U2KaGkUSN zSH$-X7`F=;cXnCb`}bh_d!@a5Ji^;|ZAK6DpG3Sx*af6F+MvBV9O=oy9t?UV*sma8 zM$hzpGuRK%KFw&*$#iS)HnMnN*E6b@vM*SC+~`{4^Or?Wp|qH2BE_^Z^4NWc(S#4>)KSPp@<)!UHb-4D`_1 zXw1CazDInyu}D8)wQqfVZHFR#fV&U#^a_U~en9WdygPlu($3(MXS_T0!C+@_{MFtq z`hyVO=r95Tet*CRxmx8@ejgh|N-iZ3QLVK6l9ed}1X3oz=@2+9~`{upj z2lxxZ%g^jlqr5w{C(;C1cAKQX-QQ*JN9r5tuNSw+d=D@b;jO-;_rTl_us+=izZyD- z=myk>Oph|Gb9<8hp4fjg{eq>Np-gh9}g_wJ(_mk-# z{Cn>%F1~~__?Y}~OD*9H{($clm)X+-cgr795Tj&53e zuUh4E#_pa%aWy~R%RycjZ$H~Y03J_IVX>rp7}6K_UxD&H;rdb@C%49j%*4fUzHf2& z#umx%FGPA2*6gbT_u2vdgV#%jUnYp}gMj{erzXQsZAiL{{gdvEg7}U^eDV6QQ^23u z!S@SVK6iJ%bwaK3G5Mewz0+!PK-;6y`18e{KD{kVLY@iup6>iokgozZ4Ev+gefUBH z6Axh7Ki*5a1q_PKDg@;VEtK~etH?w8wh%U zK0$iQ%&qZH?MC)0b4NW}t9(wn$$y;Q($8ycp4JC{Dj&_$eCmJvu}eD151XIK&+H86 zFT#CVNcwA@CjYH5&(UwQ=LPh>fZNJXVUBK2x}QBJ`Td)4$HSNBcbG@%dz%TDi`+T^z9Z?sC^}r< zcOw4L8Kh4<*jhHPGrZQXs(<6BGd$%Ds$bI{^my&}qWdYEQp&GagpY{d?$%`ZA;gb$ z$?%QD&!3wN-$eYd7n1li@v=tqhsx=$8Kts*zx?x*zqFi<`Y^jwMgJ!H&%{67W36(-w5V)GF}&8Fix9v6bv}QRzo1(wn{Q8MzJq=AX41(JuX50S zNs0U45$S~Iz*G*}-&MJZ->Ku6o=KiRY0^2Hav$+Q{GR;~!;=qaRtYm*6_?n34!^S8 zezLuMJy833rX3nYOh2kPeZk6SsYjB0Qsm#>p2Vwsx==1B{;7U<2=rvc*MnY4(uo+s z`GNkA_?l+FP+<9}ejhAyWNwyE93H;IXpi%4eQ194IL}*gJdqMT|PQLr~WN_f#;v}YZUjl)VtBWu_}k| zyq_sI$ICw=e$dj#*B7-P@pXfR*q@N!;_GzkO(=YtbVjbhdN5p68}Eaqw^tCh`;nJ@`3Wdsl+c!I#J%ICr`YrF8m>TgFY|ToWC_6 z?gQoq`TofCc>ez|KS*I-Y@qo;gx?1B<_FF2GpIK|D8kQRLFWhaSw8b<=O#T_inpIG zlt15$`Jmgo+8Guu@$Cmum5uDp?sl$&YhTg7nj$$T>nnr17kI8l=NJ3aW7dwz`mJ{I z0r$ZbeiXxh5#Te#-^_DF3a@(n#9w_q7GCxBfPW<0`3%x|nc>y`XAyrO<<;pizTVF! zelEicUm{*|x5lmexXc!P+!e9;dTP@??!MT3JGW6Ee)UuR=TAFt?`qAjk9y93lDo0m ze)>A-;*ipRI=?QiyNlD||UF%I{=r{ggqP%$&ycP>JKJ z$2u3zd%MtMC5BJm=hxBdH@kBluK94p@O_CF{UOA+Q9ns|^|K3c{wz984Bv%#g-;W| zH}^XgeiZTVx3Qmn&-dre#DBqcmFSEnere9b3*u3E72+lhQEZ0F~DioXZfKZVbI{#AZ`0=)X?v$Q|V z{P!oJdzjA*=Z6vTvje=!wQ?U~Fn@oB>mTuPP0|^|eLUgUBK|bi8|}BRO?-4d`ObVb z5&x>_6Q3r2Zh%+)(7ltnzk0r@e(P2Inuv6Mq5MU&Z_Uyd_KgPAn(4x-X9@#DC9tODFsAIpP=Q{m+GI z9zT`%&?*oid!O!aydE#VEZ1kI-&5?+*?q$EDsD#Y@-w|3GwMGFvHca>`S9M?xyt=E z7boQ#k{guRK2(0qx=)Sg7S#Sys_(>$?-TOu--3196XFZYDa{1E9N;$~el+8Ce1BpT z@pBno_-5jdy_4~BK0TWF4qT5$nXc{2I^4 zI6PC*Je1{iwaZWIYn5M1z<0%e^eajJrs&a=diR=-tDL3<;Z;uO2lz(P3D?U_#A{uz z_Zic~hwJoa;=}d0$|GFAXBocbOP?>5$HByh>+~ta|AFiG+-IJjIpVvY=;^ove0x>7 z(z{~K;;$LBw_kPd!SV8IVfo#U_LrN`q4Hhs`UNY$@clCpSeN=&DkqiSER}cAzlG%gyiZ?_2Y|>G?S9!FGKGP$ zd&zQDc`U|qQhM^F^J&nI3&e+37b0X2z2c$8Ypowjm->LwLtd}v_%++Pe$bE<-@jYm zuh+64;l5MAY|NQ$ZPQl5{i6WsYiqsp2?YN3yEG5_%YY*Hhca(N{~7Qp)~j)Sf!Xvg zd<^|(!0n7r^rq5Z?mwWb&~LLlj_~gEC(!Q(EU12W;eEsh`0Lip&*#Vwpf!l^0j4KE z7y9LZ(-?othq%ubaA9!%R{G%i4-g*E2L8s=JMC-lZk`WaZ@?VJmjW*hWLDM>+PNq1 zc)asM3k?|0d4c4|4)ag9dsV;E-Q?S!Gz(;F{V;{=hw~ov^)7t>=624rBaTl>Mu%&C z_!9S17NSSxJtg8*-!#t_9m)51;yOfl)w6>3Ul_0L!J64U2;bkM-9nsaijL+-+J6x~ zV))9t?Mx@wUN+MiZ{|4PE{71W_pBn~S0G;R2D!(4_!Wuo&UJ|PKUN~Xo+UoX-f*ea zN7ibe(-}f)W7!Zutw|e&^wIRHLy>}NaQ`P?% zx?A%5u3IGCx>qSX1^3?qMr1H%<6fl#ej9u-73mpW-OpUT;Qf`R()6Ui?p=zG=^P5e~S*Z4My_#Ak6MwYQ$?+M*Q8wR7Z(d)F~3Sb)RcbK zbP0U#nX&(T$Oqcb$$LF4_0P}jiTXDT=g**5G9ae6Fn{S}z1op_7OHzTu1XK2ZO1`w z#hBc;9OkR!_eWryDjrmAcMjuDVMXtbb{*vQ3Fv17_wpYl>Ai+@D=lVUO+6CfcH_7y z_vtf};hQ+F$-k_J7t`}v0{N4l+3yAMncQ3FCQ`c$a;paMh4063tDrdD{+28#B+NW!c zzsD`<%Qt%c=gYTv9MT8)t;)CMc$|L*s64YLpu7Pp&*(&qpMd`<<=J==>Kj1iSvVQv zH$dg-ruKIR(yvWT!+rjM->Q65Lq`s;Rle3gn@Qj1%lAL&r>&0tC;ha2|G&APw)FqE z`srf1)LiAggQwnlW2yN3sz2vfe|Zl9KWbmOpEtVKmvf5W%bq;iBl7Idb^oU1*bMIO z;XS#UCtx{THpk9R&MPDl=`fG@VBJ*Omh%?MC)5u`e@xD2&`QcCycakk{%qaz%X3z54i^S|-w)cGe23uSXE^Ci6IT`A>@@{bt|zbEAu(W73z z)I$8T0pDg3{}uU=UenqC z#ds^_$n=zmZ({h?Z+&{)jXs@e;;&^oi|cv%D))OgPNt`s;h$fJ`0qWP(Zs()eENrE zJ^3r+Rer}a{0lrECVZCoPZ&PB(9@Yhy!a}6+0!q5qE`QG_rsgXjPd1q3L4zvu6n*< zPbhq5aDKS770x>Y8b%^8+^zU+@OR8J&0WNA1C2k~t#RHMp!2)AZQuvk8sA$#)_6rX zXE2)XW+*-zJb?6=e`Z&lUj_`@mhOY$2F#{^>Q0uJ{fC=<(J^}mzERylx=W$`P zK|P~#mp!6!-PyIu|FU;Hh4g#A{B0j#aaX;+X6bi*`mIdNoOTlVy!919 zUq7^=3{8{0f7==jI1k_8^6*c?Kh=QnfDh2XJSG2s%Cm^ScL(V5Hfwjv2s}=o#Tzv6gD}TkW^3N=`j5ARA%U|U$ zcfInLzsg^3mA~97f4P6F@)!Nzs{BP?kW!A*^fdAI|I)}6v zY?;A&81mJYWwD+FXgqX~1siBR=7zuxsJ9+#T?P3C)LU;iX-q@U4OBLkYhP z{$2S<<`~o$fXXj73T0vegQ3u zCg(2?;9!+~UqE^dhHj9g_ts5Gw=YfS=JUrq+HE(Vy+(ZfKJSN>osImH9>V%=UUmIm zddklOFwH0%*;>EP@Y4nNvhLmFI+wxSUcowDJydvq6;Lbjf3+X2{bOyz@1>8B+pU)~u=XCW&!3>#nY(4XUe4eK$mh-7wqGx2@W4#(&SZNz zgKM`=eqZR5bmI#91btjdUv}sB?d1&K;r?RVR-g}f;Z5(~xJ@r-fTLcC?*~J4Vd;?> z9qgNA0(W6>GW-tQx5xu5Uq4Cz_#nOG zgY>My@R@DFKfr!L|FlJrzk220c2Rn%$ee{%QLN;Q?A7WUpD;84TfiEdRn_XK-VC&#%IlSjz$SLVR{VYQ8`2 z{{YNG`Yax|!qU!QGpx_dzp)$c;{d$ai;O-K@(OfS@%P;K=HecEM7Y5+I4nT9b zUPOb6k5?Bh?C(oRJ&pAx;DK-X{bxuYU>v@;^hDnbb_Ppw{g#e!p9o+*=0EF@Ux34U zHcmgq^=+Xu(gPUwuBVq?7~cczeiBZ>U{s(ab2ac(mFLc|ZH{%7Gt-!FC~zVdD($CFOP|KIza zWJF-8e#mZJalMDI|2|ChL+{U;{2;ec=gRrb=p|o2RIaimR~o{2a(%sAH20tnujs@6 zMDiwUb7s_^lzI-vojHU5Pu{SA@+Qx{AAkI{32kHcdqDXWvJ(& z{$IBAK9Ie#IqL!6{}0d!lMD;@U`A3hvtycXg>xVQ|>qm}yBB~d%!}@XI zbDSTn^{>?r(=loOjrB|#13i;;pl4EmyPjNA?fAbX*EAiI7U;cfiTY}CP17+c0w=kL z>6qkBsme7?$0QG2ExFgfqDQ29A-i2^-vPD8m%jJ-mYKro^K`n)$fOPe@;B-)8oB9>%1D5f6btN!%7bzed+4?;L>0HxB;^IXJ(odc8Z>D{io_zmmH=!}EPVIk!kU3Li0@YX2Ik&zgMT+jaWhm)rOh;lUL9RQkswuflxm$SSdD zZteOrcSTQ7_2+r6k4s#SD!L1~J}wZib?rr3*9Pm?he)SDzNuclqV+fNs#kvr@Vy!F zo#c35)L;7@Js3HIO%R{C3(rT+;Jib#DzM<)GuMROXme53hELWp&xC!h5xZZzTSrd@}r+#NQs^n~3jnK{7qViGPgw()pY; z@rx|y!wWx(_*38a=@-74_^sDWh96D*!LRu6(u2$p{|L{==brQY98dg;Uo-r-o}XFb zkLt+ulg~NgA3xsXGYk2CDNpd8>iAwFXI=W!g5cweOI&nLvE z5Ao+|RUg^{{OB&8P#=zKxdYj6r#$}VdwqI}#7`yuHKqShPd`t5<^2negGJ(Bq};l6 zv8O+q_}icM^h?ZFEAc7Htu?=zLwtB%t3-V9N>4}onr*~?rTiYlcFJ;|AK+8OzeWBl zJ*pS?kPpH~3{Un|&gs1L{p4+HjbF2-cz&j1=O?2Y`a{tlpo;_erSmu1|IZ_=fyq(g z=O_0G^ea~XOS1nRuv^ldeLm@y{z&Rb$bkSOGF}do#&3gvSAV34ZwZyL?IdJYRrnS~ltz-D^I4n)r3q z-(!Jp{cJPw>Q~a#zsV5)0q6U7lHlR9O6f_zY*U`#E1RaJmMGDI)ilP6TiRWCBDog z_~m%MHAB4Ca|Nx#h!59uT}W4U+tsIDQLF!29rHainf}%N?}k|a)x`N_cKN~1V54BZ z8SaO-tdy+(r6rT@(*yr`hbH~|G$q~JAU>VbwD$ZV=fC#b%&zc{uMaYq&w3yKe_ZhW z>Bw)K-qVP`aC3FOdKc1H!2DS>zJY)IJ+g%kOS&h4emwl$!Ti?xKhZM#(O^EkN|644 zAb%yqA8m^Lb-?4GXKvA04DrX^U9V2k&tkn1e;;H}z|V`3fBQbO+2ArfkMx+k1#ua4 z#C$qV{|BZ&S`q01$f#b)e&mtwKdM!~+M)k86ZQ7(*yPj=u2W@wr+tayI{lo1#=oek zpEJ<-Z*ebRk6 zCgK*p2-*gtFn$zb1hO`u0oX;lLpNRRVW7KAsuG|EIuR zt}OVcdL?_SzCXNJtA5$|XeQ#gXyf2u*SS)^=J!hGZxYw*#T?#80ZiN*`Ng0q{%pNY zGW;%Z7i0W*Zed@Z(S1Ff0gl*|{Zn)CJ`7-I^dF_pXir^JWjJ~hPcQX1_ybr}jt11LW2^V4_&`~hdE zyzhh`VB~!7pE()t0RifnA(RtnT)lAGkqQ0LCZ7MC{5tY8uS50x=QqBe>r2A2r9T>R ze%6ckL1tBYmcGH1V77mkggJKGX-DLVT!KI+gedYm*Ms zUm*S@o)=9oot)2gJ24sWY=%EFpz|d0k1)R~msaBUX8MJnNBsI-lJp(x&8(ojG;iCA z;nO^i8ZkV{RV*~ygqeKM1@?YcU+E7>{@!@D4;HQH=Wn)8vougCem(ELWc%(8!uH;p z+Skjd0P*&FSq^hftl!K(TC0QGvnBccdHYoRk9_y`?nZ>Rd@bt z=h{%A6a5vwFEBm1H992TcWvIaz55CkMf7u^Hy__CeTHA>?@4;N9HSE6{JUS7bWaNM z^K_8@zr@_Bf8c263d!%+iRI7ee}wRHejXR2-}bF@-+}+?ZyQ(d;J!t8xl8B_x*V6J z|0F7k{2SMBZd>Hf;*rUsy!E^Ik#pxgRdpM_xm{KEo%z_gMek4g&xnO@{>Hie1Nx3e z^-p%6kuP6gYy2&|=s$>d^!?sBBif<<;q7(O8GWRWzxYufXR!TN-rf9RA7?ON&t&@E znw507{we8RHV9t|{P$yg;?Wc|2K_esu)u#$rbl7EygnKJq|?0H&FkX~h6eGkAGnt} zDEa-+z_HKOT;%5KVI^S zV4R&oxs2#%8D96%x|e-^rx5=sAo_w;IfFA(2@`xL@oPW&Ou`241i^8LUJ z;=_H77UHj^JVkVFBmQIZLHOH=?@7M7cRYPmYj3!ZBRNr#=)aTygK}ic?^9@fS zv#xJfuddt9ErT8S6bX-BEy4|@TB)->t$naZss?KlBJr3htJpRTY{_Y&VEv%W&!+rSHY;R}q?v_5j=7V}WgK?x^ zI;6KV_?-U!>mPqSn*NQ4_I3vUpnqx-&I|)C2-4Rvm|vs!{__GB4 z6~CQ7b3>~9TowE?d4v3?f?xV=_Okfi>X&e1g7O^PAsK$jfL|-Y9WS2~AM@eerBtH85a;5 z3=aJ7n3xQIb0>uF+QFrl#QiIP5g9MfOfLmDVD*A`D^AHf(*xlTxD(&IH6bq^g!@(i z_Zhv|d~r+Av;DYig!^a!p?;*~rKJYk7Xw%+NUzT8x~0J{z!}>nitk81+| z#(;khTl&E-k*sb?a|)choOHz544GMaBTCzXkj@_@(;0p!GH11NZL$RA2Kc zxB>O*YrZG^0jjUr#ZVss_3CS}7w%I6)T^(JeQ;k8pk94#?hF0^q@SDZhxh>hSATtV zR=t(odj2l|s8xSwKjFitm+i_zY$upXxiz{-yIxcAm-7t8Xgr z?H=73_kDD(oM%eow?TNGN#jomeg?a&?&WjM$8>fE_0BuB!O!5A&O5z={;v=lAL^ZV z${yR<8A!iAhu;RjbpB~H1m2l$*dIVxgCBzO>acv0pWOoYMwVZEFaEKb{lrqem3{B^ zCp=ZF{+{xg|FpDsb^Wy*c9&FtKYGyDx75z4(*Vs6Qe)r-?79fP-yP#8c9jgiXZYeS z4bDL4F zGi00=-i>c4Ln^AuA1%!tPWI_7UV{7rnlJF~(hUvxuEAyCw(m=OzzxXKzvX7|12BN` zw`>bH;6|ihyfFJXSi~g(uPhAV>0runiu@w)|8(a(i zTYT9?P+tJ`jOwlI-3EVFQ$PB{p`JqSF0W^3^WWuhL%-_pzdQd;f6>($)SEvv&+Y0A zG=Iq8x4|#XA2k0>zYY3;1)Wc{yx-LsXg*Q=xT`a$H=l?;>54tyu9f*jx+CVpfO_+Z z9KssZn@<$p>FNyX%_oWBs%l+$6s*h~d3#R$zF&QqdH?CTy$ zx^HLv3iI&?-oG)`&lx-z_&*c4`v>`1HgK;VxF5LF(|70ga|Zc9p72^ge`N6edx8Jw z>tg=3DdaE3er4Dge-LQ_jNT&YKQ?f8$M^C1&C!4M z;iFsnI)m*If876j(r^4{UuPiG$a*gOm>>5!A>RKMxIbq4ow3 z``>!+%lBiy9PR4m#cButZ@qtaKKH>2ulsi1%zTCK;*BOgd>1c6eE2Tj6ym!|ABTF* zIpTW~AD!U)g*@@?IIajko%o{{^8KJY$oHEC;=}j#)Gyz_{V3g&*23_21^6QI;k$dS z#LuB#cj+Y0pE<;L)p)tGPk)K{OT|y-SN-ymEcfDhzJH(3@D0SLFY)86`sHg$NAL2e z|Gh!`czjoK2um+N6H1CD)0iPC<{l#x4&)6GldBCqF&xjzKFwp%hEy#;O zJ$Xi956E)>l4nFni$OhkMkB%+{95vi%(0#E-d9zg(Yjk_XRx61jLh!%9w2!}7QYSt zU3rG;wd~z){PQ)n<{zIP?Q3SViLa+ibaHM*?9ioHVGl)r9Uq^Thkx9?O5h%e^<>;X zw`|hAdEoy9^kd`U|M~y0_b%{OjNc#dbU&4}J1Ws_QYn#kx`uchJZUM&ptw8+Q-D=X>S53>@!ix6C)&E1B)*qbB^u7l=&Q|6BRoA#b zvwmjcvMSZI%<+@bU#H5irz#&;rO)aX_kXhrKMLt7P+aEtP?hIEl&8IMSIZm|>3)>S zPqG5-m!;!V*e9yfeq}n&f_-uq($k^e*LWA=x2w^9XgbzY_;5Xl!UlAd_LF)q*tkye zrtQ7|#@zq2uc=~$p7v-in)iF zsaN`a7;_IXtL`J_iSO(>in)iFr~8gOyN+V+A?7KJtfQEFh(5_mTSqbX5PhCB+=C3rM%EF{JxKdsz7J`u`;dt;?n5TjeaJ)^_aOt)msa1+ zJ;+?`9p8t{k&UbyntPB1x-a>^W!+HvL0LBxewerYbwkn1x}or$T{o2WR_eW=HTsOR zwm1IC-^d~SBe@;Ffcl$#X@{|lfAo0W+;4QcZM&bcoTdh*|IZ&$Jv>~R)<^sEKU&1O z8PD6auU+yb_amL-X+4{c7e}-I+;OyCO@|rJr^!BdByl>1Q2*yq{E4Gzjff6Cp0Avt z{5Z5fWu(&HUN>)CXe z`;Q)Rw->E{(;?$^rw`?ajtng+IEQZYSDVu3N8*?9Qqa4)Weyp?=P9f?N*ljtDK2wt zZ~Puj|Lc@rKawAh%Gd6`Ej^b}`MAWrd#Z)UUX;FocxHOF@R0kN1&YfY(`o!<_Vf63 zzRe-`FO&4YIpqFi;;w3zIb=L<-CfNxM^Vf8_)3rO2k%P$>HcK;Sh^oVM=fQ)v-dMS zI*;b(g@0+idd>Fy>33&iNBB$HJ;qRa%`w@A2$RBu{#N}3`iSFoU_S@;BJTzH3re1XywW`O zZ^M6o$k!pAA|C+x5=jr9Q%XaAt(4o7oSrive{>T2clvTWHyHAo5-58$eY4X@EynFLhc^L+`om_ zlO{lZlbm0yvrL7YwlIdjdO-im&zMDkbhPso}N793-SD(1^??H-|KO<7dZ_-!=HZMPwxBa=lyKh>*xJCLH+~MBkO_= zm#|nczh%_x6}AX{x|q8g^!Mx=97Jtf1va) z&x-3W-i_ujk^L5rYL%+|oM6JKj5GdwP#uqMyHAik5m1iWspA^UK$9pt>p!m{s zF2@$rtnPo9LmeqT%{$mRVjQf7x89xuQDd;dzje*RnUUy<@I>v^Dm`uxw|ecngv znP8K`Z`{K51ecCL!`0rrLX=tro<3x9&9Uui|(C2%0GG6Cb2F7fN* z3arC=Yvb3!|Cjn9XG9$<+7H=R)Ul%dkWc*I?1#)IhFCu|@t^fWsVmFwhob#V=JGn$ zj`|tzsybG*pGl0XV@3NJ>)JZjU+HI@$#ty%xBHn4#%F>LF01}pw0}vWe+g*+t^Hu- zt^JF2F!wJ_aGeOs{)v{fb;SI}VZT_q2HUgLvUGnT{mUR+pK_mGu4~=z*l;ui{$<|T z67o}Je_02{j>?6Z_IG-0e_O;1=vBi*Yn9uEn?AxMY zUHH2j`_9BZfV{Kvvkvl$mAnACQy1fF?2jvfd~fVOaIufU!g+0q{TbFh@qGy;*bm|C z!SZ^rpMia6Y0U3!$gh7Wo-f(2@G|x{*sZx8>;(H`hj2Q@PcO*lt9^DU$X`?Er9b3z zkRF+@oDBKpI4>eU1@ae_{i%?jf%mD3{Q$^aoOgK-%s|Lz;e3hxX^<~f{!fSe3)$y^ z@lqP{C*db(%Jpy<dr9LRD9 z@*=DYc#!vky!vqVZ*^k(6y#6h`;5*voSy!Wx5M{c{0dyIX~-YI`iRV{he3V^wyHyjj@aig~j5p9*;cNk7^T5Av^m=5mwy`Ao?B-N5;hdgMbs{;Ifq0pzbCzurTf z-8j@#T%&b2T=zY zwWlrqG2R^eImG6i*>U|XAI151H^ljVD*Q@S9`wjVEIxWfBgXF@80VXmewNZdugbTv z%Ktu!o5u(wZw3RZH&O1ah_5F+TqC}IkZT;b-$k8|uNBw(Ndi?K^8HUqL-}Lk{wq*} zEINa=l>fSl*H^rPs$c6={QCQ%hp6&8O4XMsbblzCo~p|KdUZbJdem4Cf+FCVqhB9H z{UhDIoa{?aqRlcoL+22uTS^pVk>%mW^XID{>k;38(M5d zuh0DJ#a8tCOs_1qqSvSMS+NzpJ`=Qf7QH@`>*yitNO|Yi7hBQmGx=e$6}>()cNJUF z>ob=twxZXkJEPc&UZ0tn#a8tCOyrBL9bKQ!tHstoxjyZ4immANnZLZ)ie8_&GmEXi za((8f7F+-1`V8JHwxZW(;-%t0?f2Tz^_hRB*ot1C0oAu1U7x|kVk>%m1~(R4(d#pN z4TV?N&yKE7_l#mIdVTuO6$q}#@ff~4|#|27{_2f0$|_h}c?0*|yu{?kR) z@M&r2jl(@69Q)^!#aqFXOAkkrdy7MX~RDC!V46q58k~sl224sYT@# zeZTFiRGv}3T=6!P|EPYJD&O5z`9DGWXnxE(l9ZkQLXq_?MM*OG9zfa{Q{!=-*8vgwL;6=}0JZ^jA+oc`(%JPop z`P6e)TUIB!5@mclzy3D=CLVb{Jm=GXA3DCG3s~Ye=#zaqa=&G<3ZGZ}1&YXo3zpFS z8afsdkB0xHN8JA=N}sWppX{ejy+r$o=-74NxPB@8i~UL!|Jf(d`M~==jm_>9-swmu zX&DU!>G*`o$8ZTR`}o`iw4aEMODTNx{VC@8U$Hehtd0B0=f}uj#Br%K2Q6f=txhA$2WIQoX=hs=U=@a=gp8G=RNXA z$08NpJpUhkpLWHb@$}03cM^+ezZe}G$v#?s?NoUks(4NE7cGCY&oCPPK~;Y96c1E= z$Pzd8&wGmYkJ0h@KFkwT*Uj;pO5ar2OPuC?=h5f$tI&P7Xnh`I>KpAh&pb_g&*<1w zmB*!YK8-$4I&+lvlPg{DzD@5<-rn_X`d{;t=U7cY+u3Tp_46PWKN=qG9}WK}{i6vV?H>(~_K&8I%$fE((m(!h>Mu>78SO7kpDFh_On+&Pf6`x? z^4L*-Y5L5!sVOsd^80W1m*YJ?AJzh@nRL`2#XR|Hk!23K-|sFivdr;M`d5>u9rdq9 zKcD(X^ZvHX?4t09_P0hK?Qac__P2&d`&+}K{jK5qsPRP`Gd_rnr`AyUN5>P3DLp2f z{GOxnL-hS}tEoOm+mD5$kA@E~X8JP%>scA#JMTAK9bG@Ruzo!J0op%y?6&;@{_#Bi zZUIMva{p$21y*FRJ}%F*OT0&^^>=B1UYgJ2a{E?ZkECul{QdAF%U#H2zHp#C?_ZzC z>o)99!u=14m%7vNcOur`o!i)dC&*7z_73DrkRAv7rh7r2ufyw$;$P+)mfRm)ysab?f(4`W^T$pk7J7WIg*~JolW&eZ&;(GdS-q-tUoyybIDP|+0O5#&9vKUVA$ke~l4+7tMpX>#}*`VPyj!@1tpgM2US z%kUfW^^$~q!X0cM9L3jZC&;sqyOUV%K>j`AmHoi|A^#ZZOgp^Jn1=ks4eUR?gykb3 zUk`f^>6i9;2<%1f;ynI=eZ^7^6CkhpWV{?ELcR+7Q+&jmh5VgaZ0|JYa>!QY1g*vW z6X$-uk2DqX;V4hLJEwmpFfph*C;pZMzr&g2eO|Z>2YzLHAK0|ce}?y?jgT+b)IKWfc*Tf@q8_Y ze2JtV{_~KZQ4aZnd?n=S%Xt4syBv1UFpw_^aESKOEnS&rAAzZWqjb z=B^e!8_VLY?~BVnRNpekhLdQ#5}E(s{~(QL=>2Za)AhrnLhrbK(o^J!-Y+NPp>>Co zo;C>!zo>P5yi{@-`HS2illi!Je|^gw@_oll%lei%WIpWE|K=!dKAhaAzGV()D$Vzk z+v1n`X?Ajb%N%>mnX*}Ys?zV4m%Eptq# z=L1Z6XK$nQ(J>w6=|4jGrQ_yd96qs`(o08Y@&5t&r{hcUe{p>)O~<{I9w~c+DJl&MaLA1-{i-k|IKmh$#H)J;V*G0l{Xz12tS879lKKfH*tBxslL)N8v4`) z^(}Keq4J+4ef0VmcR7bowQppZ&yhI?{qY;$lRt6N0~#OOUzzM^@EPCqz}vI zYC7L^yn*xWkEZ%U$0Ed^y`1dnDD5Zxnc(-MAM6+%Ulj2Eo$=pM`Aph2|8UaWj_d=+ zf-?W0Pf3Npt?xNo!0n}^kN3cdoG!!gw=d=~9`Ysgp6@aL z5c>q|-^V;acV?>S!NFJjyTdlb{3_TRj?p>Jw?Ky;wfpzRP-M zx_Pl>jz5&X&@^seqWBTCpFsKogLb7j*Ut-?=c}Xk7t#G%Gd?NMK-3(~5#HUunEE#w zJW@jwwQmKzzi+W+jv}Rhv|`-8#xe2dUG)2B%z9~X06mZ0w>Ui7z`uP!G2Le*SsCl4 z=KV9#^`>2tr3Smjwc6l{*t}P zo{l;cU$lI>!alhV>FGFCou9*0{Jj+Ssh${ri9W@aIexo=^4E6T`0f~GpFsJ$t>}Ew zv0Obrb)2$42|Rmrv1N|&mE-9@Th+gM>ip7lIF{ef4v6!i$ZxJ0ojW=vI;>B(q54fn zJ*p4U@;Zxd8AR_#(yJk2dbyvPr21hFSszZdD7MV;sLJoa$bJ(Ok9i-DxqsXqYuF!LT=k8f>TKe5l|@eNH|!k?h5 zpY)>Z+5Dy~;PC!X?$=prSy#+O|N0fr^F&^Z`L*nqEP`U4Jg9^akxr5Ou0rDqR zI!hoAa9+e-#s`y>9~lR%~`>c{X z=-2m?`q`E1iHrk$l!J%*nTGvRDIeI6fc#jLvj=$=@`3U`0lb%SCgktIPXPP*kgtK< z?#AiKL*7&QSqJ&+1ujqVQv&&VoG*D!K%QUjjPoMn-CAfT&X#&ufyXQIeE&C+U&v|7 z9R7ZQe<#c1(N2)ddn){6xqXm!2-V0MLG7sdbpk%y=fE-1_Gv;n&W~BY@jl%4{D6;j z9BG#X<@o_=pY(g9IX26B;~DlV`A$I5V;#$_54pVSLH<3)f$3dXZbQE5WVCOaIe#r7 z&*QlQ8L#dK`O&hE`~#E|kUxRv3Uqrc?+K9S?_JoRiv9Xl zKQ8B~kYA1c@ebrU$kQli7xKlB>-RJSkjr}xk~g^`+fPVIRNPcY=K&^#t`n+96k+uN3SHV*f3dPhZHbdZ>qx%X2^p z$Yp+-f&H{GoKEKou0Jm1-qmrrT=##V9?JYEhwFO-=AqWL+)k{7{8yPbUBhw<*Y^Td z-{d);vr&JfoRhF$2KzMXb1%qGL-~t51$hTu{~_-VxrggC4f(-1?+(5rIRtW47t4&p z&94*izM%It(e|mRJeSv5i>Q5inFh`Hr7zn4`01u|*+AN-VySoNS0udGlh3D=hqOaq zUK5um5un3rPKVe_`;lH6m#1NWu9A;{eA;pF--+9^EaW-JotAuk$aPtQ^W!F2?!o@6 zhdJH==S{BHb2cMAL%4kA!+zbpEO+ka`shRc*<*3J^oy4v{jzQ$_3_c?5HH?$Z|2eF z_cY|*)qLFqkPk<_@K0y?I>=Aim&-xqQh(dr&2pNwgiS#`)R*~RbG&YUmP`Fzh;kEq zslV+mW$Wk9kjv`F*3(vkvk> zsyr>UE3YD*QcvnZz7*x`BmGIpITreK4chGqz`KtczGt+3O5^=wmmWv$)AI6L?~hxT z^YzyWjs#_XmqI*~FyHYiq8*3*;cDKLfc$tRuLt?MJ8&Jp!0n0+c>!`8^F$dx$$fA) z;C4mYxdvJGV^!mJLel;0D=e3JL&+PziE(PF#*>%(y~YRJLEIi zBmK`p{(;vRO{pOJkFHMo5n3HhZ+Pw)+= z(}jHNK9Cdmx|;xbT~!}uLSBUPCb*>89#gh=f55+y4TI~rT{iDW9cmr3_5Q$JYJJ|k z4?ennE$j8NUU;vvzs!lRFHUO_58u3RoVO(&egEnrTF;Hf*MIjoFQMl@qWTeaGO z(e(3B^1j&2skA?z4tZZ}f&MqgTWWpYSLL@#>CO9Pqxr9`%1hoSn;k&!YNca0%8xm} ziGlRI0UbO0e%LJ8nnT_Xn>~#<9ZM*E(eFDyq|Wc7itkF}&S-k)D!sm+@izI7+MD|q z(ft}}s_#+#A#{C2`M0XPhLL@AKXh+ppH$_SQU1rO`gVc}Kat8eU3FW1`B0^QE%E5} zbb_vbDn7ZNl0BWun~ralzn>LI7TOXU%5FYZ?5^OCafPTagNS^Pb$!lzXJKYJ`*UKN5kx2b)OmS?@!U=F#_M24seRwbBH_|>FZrYc~_gBlci=QuwF8uW; zzvh>~Jnic*#?*<De@xFsm_2aoE5`Po=!^7`^Po~85` z{i}@QrGH#3^8t)Y90ZW}Aj>$Z2jqF!d$8{Xc~u!_Al?+@&%DFq2B#mc73@yh<@0_5jQ`jOusAzy^@k@dn7$g9ZwN4$30TEVb{m4*XR8EI!|;3wd|(kMX_PzGr@QkRI_b>cGAk%0cGUy&#wQr_4`Mkhg}t)Z6}$*Ts3UmT~=0 zLw<(j7x_w4J7j+6EBO$}M?fy|4ugCJ%HRKzuhR_VW5o~V39|0j^fO%7ZMmJ1b-!Ds zp5y%H;pb75XMl9dy7dEtc)V|q;(W?>T${{6#HKOm3J|E$H;tp7jH|FVsCJpW5I3(uqVqrdlWNv~bN*Kds%`yVh?)DPzo{0R|GC7Y^S|`> z)vSMW{%8MOEjItl{8a5P&HoA%K05!)Z>koX|K&=mS^wtzFTc5(6`lX3wp6pC^S{(@ z)vUiV|I1tSK7+{o&o5Wq`YZFlpnP>}?4WJ)0jpwlYe)0HWYy|cbpB@*Rkxz^Kj*7z z)?b>Q;3ASJ<<<6`lX3mQ=I;$^6f%Ro&Xr{I5Xu|Gzo^%ToLNZ_fWR zORHHsn*UjURI{SW>MXZ$J`?*_4egT87O|gJ%*{KC!t$ey z;|A08K?Q@Eo{V|=rjp^oVT{>sjIP6I(Kok@-adpe(;av?qxt5R(I@HKR0ciz##YFW zWb~oWrI@F>!tPMUfW8rB?9=@i-4huL7gP)7`O6qnz|3GSx8N?$pO5-m0J^|Dy(q!> z&35GLA%8DlC(aRkJ$S$XXtiZ|7MMAJc>v7A&K+8T^LG^ccj{FPd4UFcCcaD`#zGp` z;o;?2PYZ8hK56~kVCpdNU)f)NGox*>zg#88?EY+*?ah8NwDHA+4=Pk*KkX3Tx2(52 zFwa9Cv}f*(XT1Y?9(I`@*q`$)W1=_vb-!Wm15*X&Nu(#&0r~5Q_zq;WIx!}JX`l=A zfIct)TJ&NilP?>Xf?o&R>I^;50a}MZ4s?KNpbPYb!>0Ija@1Iz%kz&x-3OwfmK zO?s2S9Q>xhGr+9qfq9_iAU>c2OapVkK*|?9)t$q6zyi=Yl;v4q08HS#IKT|h1N!@M zI0xr7eIM%C6MVkYzm^a4Spd3Qm?wT>v}pm~l!NspqxS`4hF+8%*3);GyQp7z#AEHl z@wz?PPqsC}(?%{6E;XHb@<{k;g?vxs>mr~RIU2h{8|1qq^Hh$*6&_}J0`Yjj0n`?hHD|fM8STxlj0IdD4%#1U47XR#Y21GM zbDDDc=Q1XM);#7(a1V9?Jz#C}>kP6(v_>%Ia%`Wi+l1wtSe|*4(f@=owVpBg5o6*9 z$bV(De}w*B$k#HaKWFp~g#I8#7g&JY>BRCZFaQ=ha=MeZa60`g+xc#@FumzvTwd0A zE|*-L-9mY4mpwvEc4YMH*9o~hne~ajjY94m&6sFZpY0(}H)fuw%@_c)yD@i;;&5&= z=z+;idxm~;bzrwUW3~(1d%H8HVV`x-4_(UgY=*HgxL)Wd0}O!v5xa+SYY6KTKpU6@ zI>0nA1@s4Txctbv?0*!a3(NvNkuFCfAI_8Ni8*7K*(%&-XziS+F`xVCg zO^oSjzy^%Yy=-sY#~4g!^gKrA9_S&@O=F(9n=##_7N@rtqk9o!=59v&Ag&h$VD4b% zna+&%TU<{wB`BA^EKfAxdYhia_0|X4sK@DR*)9MkE@JLIjPQ>zy1)W3H<Lc<;!iNR68GKwK4Cmx{n$K2 zx*Zt)uAjlYIrvfUGneNn9-Gg+CG@>ZbVxZBf6G{{LzI%Ef%X8*_3jUA~!nxxD52(>z;-KBx4);@V$X z{IRdO{>vLJ8x)w+cB1h2)JvQnc^||lia#)$^?Sp<-aVW@IdAFv;@pN_+7B0c;m0S~ zUord*RowfD(?{E>!e6R(>+t%M^FO&ehnMT~+2w4H+ceg>C$qi*!oR25gFQ=F-yZt? z$1#`ddHlJ|8$tiZH0I5~J1%9uFZlJFnbR;g{EblUQCpw&Qoc(+x+t8#mE5cs~kGVct2Rpt4_1FV<+;C;nU z{*?7?5Ppl&TSu~fPv{>|=da0wasQv~%e*e^KT-Y32323~Rpoz^s&7vyzMsnf(`~uD zPC$J7)?(fTe2uEFzpMIwy{bQ#sQPhnEB1E`{LNP7dww0(({fVy+xr>Lzx1zt6knq1 z(<6;Jd=G?wSf!_r%J0To*?&{m+m|qx_O_ei1Z;e*~W&Ac|kckjSl#zQsEXD;pOcds#* z_G#QYZZE1KeBv_J%lXb1nA7&L@RuFTydn54k1?12>G>_}kG7-nuOI6r{q+^^sQ7vn z|J;=vUdET-(hYOd7fF6ry~AAE({xWBpQI1q@s;LNnsa#{f#d}}IK1>{ca>uPSQd&y7zPnb)4K4leiDc|m2aCqq-AJ4E}%G*)ni{N35 zZ?V5dbL){E>use^Dz5#N#l0i={M#KdzWyKPZ>aKWrOs!yFfdf}F_-bo0b7{M zc;;9Ye#ld-m-h99E11jm+OZ*Xsc*|$aQ*0w_!j=adbysLtN4-?STFVE!uA|~0K&`n z)xGa=oyl{k1?KYIGza>`>+Db3n^WHB`c(=3E>!7#Zz0=D`)NJRT=M(liOi)vsc|QB z>920B$LZ|`|L%9Jm-he5fVtHF%D*v}@}8sYCvIfDl-Di0u>XVMzt1Mt%lx(V!_4LS z?KFkCjHiBR%3QAJem^o7|F6EtT-uZBk1!X1NyV?fn)No)^TR~uGM=u8{#@d(h54O~ z|IWUj?WMo`PQ{--iS<(6y;S&zRs6e6ipRgVihpSnEr=quJ5YqYsY5s^KqFfpM5^$^htl-sXueMe&zic!66tQb>j2$KgBoQ#p#ju z`uefVZIs7ab$vXb>i>Xq*j~n8-gK@%dR-6hEWZqWcJ7YDCq35dbvp5%Dnnm@|I)%2 z5MHmNiT?!FpGwoG=5hK;)8`RhuVab-q*_lBxnOn{``6=W(f3m0L6HkO7|-c(o#?aB z>v5atb7kmLPvZQjah`+^X0u+8TSTAk!26H=&$(R+)-h_mZpTC~`*h`f7S@;X@VO^Y ze(E}t@cA(-)EHfVFukf{A0_ehh0{-DeoJy-ebA{8}Cb=_;Q~+jQ_%E zeE&qkJBPB}AJ4u0q|pb@$Mqic&yLw|y3uEp-g=$GAAQE$F-Gsa5D#x1#`-zQ;&8~8 zJ3FpVLtkrp^YJFUeJ=EdZ0DE5^;}xJk&yNP_km>HUtA?{y_VW8pxoWJIo;DH&w9p$ z&HspUf0yO{uUxLv_gM6x(Psnf&w~F_+Xb8#T`$t~LCdh-Nc>AVo?-8QbGGsCcVho? z-mGP;fBHk(Fh{>Y>HU+r-jqM%$0v;5ek|_azm@fS?{{)Iv|X7zsdgR-aloo z3GY!Z!}Lk|n{d5uTqk*x(WhD=eJCHh8S9JZO}*Uclg;Dl_j_{uCuYip^&<6cJbg|B zj=%Br1!1(#squR6^jc&>+jHZ$VfqqreH!{vS6*QoeYR#? z??ZoZZu4-XwA^dD*Y?GpEZvk#WmIPhMv9 zUb%SvaZY0WXSLQ0HG1c(nq2?iMffG{L*C{4wabk@=n-$%JcMs?@~qd5-u@2ZS8;f2 z5{Lg{k1t)LPoE1t^5@|@8g=VmbCuCMT@$hR{c;?B#Wg2RH2TbC2#@-w{TI%P>)oHZ z{JewO9%;h6)p^{$wcj|M%WuX3>vKl$&ES1IqIXsP?W2)@#P4t6_VMiTPhM-n=hx!; zk@S7V<#5v5kNslw_K&iZet9mR zXIdQ}F7;)r^M0eP{qjQg|MB$8!lAm~BA$QyboT${iz~u@*`FLwpO5g*e_Nr7Nq?q! zJbZc)`(MyM`;^i9@5R#}EMxtHKOECF)PKPHYAk%0$A5>@pKsOqWup(8#KXHyIDO}@ zS`scbXKEsTl#joH%fITF>n<|kolQ7@(7SC|zvor<14f^2gz)lRpyRoI9G<8;%IGsq z;(8bF$M2dvcyXxjj(unF@4UwOJL~1M!uH+08S%rv`xfV~a3_BWCG%7~eNIPC-~P`Z z9LAqMjrXZY{}dqqAGDqpj#~=!0_AA`>FmYj)3NS1lT7-&T1tNb=kMIlK3QY*2^HQ$ z_)U*Gi;UjdEgs%#%KoPv%GXb-7B?Kz~s(Yt3M z{|ImONIQvR=OXq|e{h{vGEp z!1=3HV_;bSEvetoryH>U16MT<$0v5{c>MMfKL0~j_nBhSS9p;3LrMSZE@u5M2evuG z=>4Yg@WFD{fA`6~q2Bt1_YcYWcLuY*ajmh1(0_eZ|88e}&9h5>HTrB7_MgFgrPO(g z3>%8 z_4j<|-@>Jg%w5pq`m;Y^{jH-0&okkzsqypgUcm8p=s3NGQ9GZa{YL$9?%?x_u;?VF6V~*vxFDDJ&NPM?8VMu z|LC8{`x7mR|62AxZhGZMqy8p!}2i3y)l_$sKfA4X3iscei8J;L_QIy|mV59jnHd%Rc2gijw9*E?M}eU~3Z8-M8+ zD7`&^!#BFB;}v1}9`W$jAsjweY0-~HpExwG_wc^fHOs~}HF{U2&%=9L2Mq4o*65x4 z6I_3ge`_B;zoXV15zg=ZD!lJl@}Gu&jy*BFzLO^)ew=^jDh{9d=ACeUl5^wf^O61v zAAWY8iQjsqP8h$e)5&_AU}$6hNczw|?6u&!_D1Xl9A4T72iN}@zumha)XViP?E~7M zo8PGyjz6S*5&w28PJesny74A_W>dWVccDM}h@1B^daER^_dnwLF=$ic%0`_zE?&R9 zaqNHX?90OTuki71Vg6HC4=iTS$NK(ccT7#(PEV^*avlA)iin&gZBv<{V`5 zA@y)8N_oJp&qWN;bw*Q&N&M8-uuPdO$>%(qN^5?L9+ZrFP6R~p+V}HYU zy?xb=;`L$o^6Fjp+>xErgZ*uCs~lqN+&^{=(^sHxFPi=+f%^>^weQuxf$NnCv!h>2 zgwLvdvG&AxcyB*Whl_Q%PMM;xpLhP?b%d?u0XzRFwr@N(e{UH7uWXkGia%=w$20V$ zUBmN~t-$LR+Mf%%aX)|co$=>YjN92*pY!n^^*tV19JXW5Q+(Z}o@Nv~2j77$E}s}3 z`kTl4+>`1${gTs{{$t1WGtA?5{xsWwBl*>sDgtT+Quq>$<0eTP(ahuiH4U zF?#J#ui@|4o=#W~Ggq?S1&TlaHnuPNV&+tHy3;nNFAv;mcO2)V+*1Q9g?0s=4`;q& z6g%r%F5eLe=K*79H|BWL4H@ku$LIgS`IuhyQl5YO&g&0(`jD7u2gJ^5&hw;!4>o_x z_{)@Iy;Ui0=kLq@+@`05^DX~7_UF-u(W3t3dwrwcen;B+PH$I}FF+6@Tt@j`z|Er-l7; zl0M{S(x3Z<(UbCfm*XA2<=NAWzw9GCzRmr_c3b^D!2ZT{U(+Xyw>HvM7l?Mo`i{@r zo9)jC=Sztz*`GIpQT#a``}<;8gJR>)!+CJA{$Wdd*^BKy?OyVYvCGfn^@-GcM)Bvj z<$SE@G&|hV<|ADmeNZfV-mHh%ZrlgwKV)A7LzOb&n_4+7qIWPR+@NhikJ<4|JM_|7{ z@==Ac-_Xy)^_655j?b;6;>9{t=NUbI3(JK*v}elA1!0;p1*JcIi(WhW@N|jAG})uJ+I-cNshTUDoGU#_gp$5#yPo?(6V6=m0f&&Pl(t`FyZi3+SwR*c&@!#F?BzdzyS(BIyy&!ppa z))r3RQ~d|cGouiI-}W9+Uy|Y5um9W3{H@$JSvCCY?dgo?khxNc)+8+?MSB0-( z_f6c+D#z#Tgxe0AXZ&TaW_mWiXEI)-<>yY=?rQ*{%Q-JO3KC8&c=%a9!1D$95^8*twn9?&njho^O(8Pv-TX zB+xz>*PFUeC->3hex28W?fmDtJ-z7l>EU{gPv3A1uk+T7V&^Pp`!;(n;C_p~EotoX zK>HxpJEIxz{NVBBCVk#9iO?f8 zKTLX^)tuk-!(6|F?|8;o`vl)t*) z7<|s-fAx+q#S#QIBYrJ(i zeQyuvH!vnoWV-;k)oypTyR=z}iQe)~V!KrTc)UL19r@)r$JhnPr+-4+&f0_hJ+9+^Q?d+F1AE&o#evh$B_hr2$>oG{D}j{O9Bt)8O*mb?YHc_vcBk#Mrw@o}c9~AIUa= zf27}in(gPDa#q-{NcycB+^-3DE=T$=Wvq5r#k$6SW<1{eaXq68?vH?d8U0ZL{h0A( zevVfwY@c!uG0J-j(kHYJ<(hM-ZGQ`waJbA>j83z*VYuKV4zKyPko=$9o5Q)MGH!jZ z!f3?bl>Is2I((`rhjaD;!rnU(@hk5C2man;e6!}za6P$bswV#6nlRp+iy9d z&jlt8?l`s!u3^lAi@h_7?W_E{shzP;yKI*m!{+w1{J#XxjKXL!*0kc0d7yAJAMdh#T9QxnLdKZ}bk-0yG{n{hBT{-n?D_pNk zPvUj0EHDR5O=h`-^LldqJ~eqdr|S;Zr|x9T;QelyYvSSbx_9bE*30^L@*L*bGvnc$ zt2mwSe*UIs@?}k7y$dV|-m1ds_4rY&mvwpjcIH{20HP8Yu6^l z^G>|)#<~Rep)O`j0Hs{qH#q!9CshjP0r}C~4(46p7>xIy&iQiF#&D^C7k~Or*r&_jbqLEvrfcyc{0L*+rV+x^7&Xjfw2Id0UtI#uJ<5c4SA1R z+#dwB8A}xJU(E8A!0c|!FM<7vCA)>?U-5Oues40S!DlXIK5ZFe?hVG;u=C#K@=f>W z>+*>3Xv>*0T$*U&)sl8b*_1S68+;7I12bb{9dmMhh3&-AR!Y5!S;jN||&fSYK3ohaP z_c{Dq8-j4&lW)Ry65h9&JHV9aasK=d+lA@B@z$%u@pm5YPb+lb{=paCiFp7n=ij=4 z!`EN;b+`@|;5}^lBDT+iTQ!*%z$LtWJ%?}m(Y)JD{_PvNo~5s04DjBo)J@Ej&`Wq9 z_t$ry^UHH4eELXk4|2c+&^vh(-w7Q5s6$=~+m|ffzh&Y5aT)MjBiwfd-x~jF z4!?X;pJz<`+2;^H&;ur3X1VnuV|o?WKj%|Ef0yr{&6)7I(|P=rPBSJ3G3J1BUM0MF zp|z>M{Z6rtFyT{Yu)Q@H;ee?j%q6^cF0bFt1m=MAfj)3CFb@oXD}g4??foVUXa8B? zRGfT9`JeK zGT=gBY5ptH2da(#U4XTL4S>yn`vKbndjPutOY^^?X64ZTYT#PnH^3i(TYwc4Y+oC= z3$Qf*&+ozZ3xUgkD}bwkYk}VYw*Y?xmgc|Hp6uTN_5!AW{efxVFyIJa23VT^)D?Um zZWz!7P6g(Gi-7^K1en16!_xd0YH|H30a~=`YR-2NFafLwr0yluCxI=2el2dFtjig% zyr4z6k4f60d_9aWfgUg?9OFu>Er)ma;c;u{g^!j9p{G^YI`>~E^?gH(;%yU5T<1}JF=N-hcIeCWv!oQZ{qQH0qg0B zo3X!ApLd(f*$qa9>-FjWs+>^|T%dgt${A?RP4t)SUMsAx=}wGZYes)xMmykjGZ*_% zCIB5Z?~nS7{+~XX{WzxpPh_-CV+=lF|5^B*3e@WqQJ>NO?c3N-@&@=D4gddP^x)rv z-gpT7nb*Z#~u(yV>gYT zKesXSOe4m8XGZH_U^H>`e|i;{pIy5npT{N4t&h;&0OdZ$*82ydSN=5ZRG-^#7dQ`? zZou-jz?Kb}_W-T~rWM!zSAK}`FtFZg<_@qwFaw+b^neB6VxaclANxp$0bSr!;7s6r z;9}rPU=jAQ7QnTCzj@oR+~qlJL6OUQB?NW&vbY2PTc6YZlH(J(em_N7{p+ z)4G@ay6|sy=UFehW=YiWVbH04=y!ZhYfX>e-7xg{_DIa$c_jk|iQa)Vo493O# z*VHx4^FZs`xZVY~#;6ir+3h*5t*$J{~R$p16VK4p7S77x~1vUpL2o4{dlr zH{;huyb`Veo|%C71kumi*Fg@H^t!LH-@$kP*1`Do5wG}7PU3JmU;q?-dNS(+pz&dT z9^U7Wz&>qX-tX`_pO*l<0JLx&CxBK7+v)eTuy-0KAIbfK4eSI=0fzxgj$%8(OG=T8 zem3+AfmZW&VLR;9=k#23@l)YCSSrKw?*cG&0p{ZuGTM_Ez4L%L59u43=g(nGW1cVN zDft`Fr%o7e{%FQ=;Io1Kk7aqq6k`rJpfB^H;}|7;SvU{&62DueW*C33qBu^0d|oN~ z6_Bq37OC*spYWE-Zra~B=-eK+nCjzGeX8I26*(l9Bwwy zUC;fY^tXaquDRF=YQ2a)8i40SF~Jn2X=y5FaFEI^cTnn(qRF$TywD#{FA<{h@qGH zjoe?A`>itHk$H|o&9+%L@Uc!_8p?fGtrxlT1^dxhT5@H6B=?WyezMG$EXZNPv%#0e_ZaD%X~}bS?=#BKdkf1dcQ`QZS3A;A&t?m*tlLec&!*Sl%Bv16Tldq1}(=$lnCuN?Ov)OMK;CVpp*Fw5Po(Au?nD@gC+esV=yPd@>>}LQid^b>V zekt-i+?VEBwDwz{mx?h+u2@l2Do6SsjQzjozVf` z0-SIs%ezct%mbUjej0GQnIqw{sXT8AfYu$%D^6py!G|fi=tq_!p9c9_(cj7bh5$zZ zb3pN@F|YIm#S?e2-_r6|GKR~gEL{0F?HXLo{sjksPXR6ijvLE%f(gY%o`gIF%qTmZ z&t4nY?|_}eJnVKB`~ATF1t)+Dt^lw7BirpP4%^7~E>Obh{N?D|y{5gL5A=bHf#%!3 zM!ymm0M`Kvz!ISKE87$7w;r?xsC02 z7Jb;S1bzeD<#zUS-rbB_?u%nH=zHA3@}g;R`yTf)?+=vll23a9uS@p>Uh*+Lds5-iP5 zB>kk3$&*sK)`N%qi ztW%_S#eK}bh_a7CqpVAa{IBwVD_m*iBkLWq{vr35gE?G2vJN8aByLSEpTCN-KSQIe zmx%nY@_#E_Y2~AzKal+i&H-FLvTs55F?b0spTCN(&DWPkdCo!Pf0h4R;daM&>nfB| zPd@ZqMH z?fCHNwMk|kWVhh)m^J#0@Hr2;Un7`5iO222H9CV?pF0D13S;8bc(~MnI2S*KLED#m z16l6&XLL^6zFmO*l6j@iBAmoiDn7>ZyUgQ^1%&gF?<6qs2UuzD>BjLqP_*Q!vcgu6N>b>|f4wiiva-T<@yDe>e zF82C@58S=$5`5YW9$AH*++xtAj`7% z)O{_PmuyA&5w7Erd5Fk&Hl3x#tNo;L{?j=B;$Q4UF89|You!@U(&U-RJpcD^`g@o~ z_+`pZL2>b4`$mqZJy7Hthm@h$_QJ;@+{{wKFNXY^QuN}dEKFo~TwY1&`;}oQ@r*3R z&MQSO;Y!2BrG#4tc@fTc3ip=;kG_iU{|FzUxX8=G8D)gaL%#*seq=n}vi$#5efn?u z_eu$uSMn|3E$Jy~Gft88^#gZ-p3;lmzlF<`{}RQUsq>xwhWEd$TaSG(;3D5yEc$Ba z!Z+JVJBhc)PTCKF-8i5J^i}vkafxS3DRR;ORjl`QygWLA_fvB5n}K{{DSEM+U4~xl z1ed|Sq?B;G{K(_(UC=>Pt;655qQ@G--}UPIomcG2dU{+U{Jp2X=h;bpkM4S9|GBas zd=UHF9sYi6z&w1Im1K#Aao$w%yOjTr&tm(A@b|(moWC02KP6coz6^k5mvv)a5BmBl z|2>DZz7h1tsqa1N@0(U{8~6XZir*_@eFFYwtMnhE%5Snt-wVpVWCZ(b41b9}%)^&j zlI+T+%xgnGS@A1X`R}g2-?_(!Tz)byoT}=>`{mewP5AFtA{eDT!b01Xm;s;g!u2=fDYM%S5 z(x0RFqcu1^a=pB&@}K^m(^nnoovri>*08<^`omRu^i$XGXDWZ!eHsscu8Oa*I=|(9z1?R_hpjRDDRR^q!~q3(7yTy%olnv5s(q%Fot-_@}IOmhvmOyh>X~ znOH{rHJ+T8CO%M)+OJ zNdIkRod37W@PGDa@#|Hmf10}9=6=NXvhMcHd(6X^o>Jgvm430}yZp}f(tbXy+Vf`= z4^;b9a}|e|{vme)w|DiC@L%$*mvzlgRriX9DYgK!o`@@#C*}qw%F+Z)pK($x8|GZ53*ZOx<`>6ZdKUDg)zOx!XX?@o+^xezQ z_b5YOz6|{lO0U~beg5=#qTXh{K4m`sr}2e4|9ZTk{pvj98^}7AmdflFCy&k`4 zy&lhKy&m6ay&msqy&nH)y&ex~y&fNFy&f-Vy&gYly&g|#y&hj_y&i9Ay&ivQy&jKg zy&j)wy&kV={h#_5U4Pa1MTghpJFVB_J+0UMmDcO=pw{d0q1Nm1qSov2qt@&GPV4pf zQtS12Q|tBkQ|tA3RO|KlRO|J4RqOTmRqOS5R_k?ttMz)ktMz*PtMz(3to3?)to2vF z9Uo7f^e%ImUtX*91C+jwnjc>F7KfMlV11>(UirI1*-uvbn-w3V;vcEPk5>F@#cxyo zu2K3XD!q3q`+F2WQ>Euz#qU%0EtLO^vVTz7Kdktr%D%Po*G}<+6z`^ZPsNW?{7MyH zU!{LcrEj#dAE)?9D*Qmj2P=M^3jc(P?`&m%q2iNO_@`9(3F`b`s={BT_!z}+QU1p( z{j(~*smgwmvcE;~I~2cLg?~}`pRVlhQ}z!j{(!Q7MDf|mevZ<=tm2=e?4MHnX%+qj z#a~hUITij@72gtN|CZvu3jewae}g*z@2c=0D!xYXMaut|O8=&c?^~tcsQ536|DpKn zD!#Xs|4M2;`j)bPPuYK<_zGqJvC^;7dd2rp^UY6H_{plgKU3kqQ1J)k zk^4){6>p)qt1Qkc zxZ-t`|C`i$>YhsfgbM$&D!+y*{f$-lW{U5l`2LDFRsPy3{bm(^C#643mCx_WzPwuh zsH}K3#dlTw5EWms($`gdPsJN4zL(>mEKd8et+dJt@QJi|5Fujr_%SF(!Z$qBHVwK@p%W8z6Hwu0F}O>Dt#SP z{O2hBfl7an;^(ON7AoFR#ecBkT@}Aj#ot-w?@(obxZ;Hnjy-;0(1#;UxVC_X{?o22;3D*n?IKThTEM8$7b{!UT)(-a@7 z{GXxpudDpMp?F^L+g1G26u(#T2Ni!r@topMD*mkE=c@b+SNw9tM=3s5@iSFa}_^d@e36nuK2}@U#j@!ieI7lRf>;Ne6-?Y6d$YjHHwc{{5r*NP<)c& zH!6OU;{iod4#62+G){)XajDZX6scNBkD@%I&9rTB-6f28;)imy@p zGsQnw{7c2xEB>|O-zxsS;u{p-sQ6EcZ&Lgh#eY@&H^u)@yqwyPS3&VgidRv*s^ZlZ zuc3HN#cL_Po8om8-(B%N6tAy%L&X~_-bC@HiZ@rhh2r}tzOUk~6>p<>TgBTc-a+vL z6+cMv&WazRcsIqnD}I>bJrzGf@!pCbrT8(5AFFs@#gA9~1jSEM{A9&XReYf0rz<{4 z@psht@jb;?DZX0qPZa-D@y``sr})>3f2a5c#eY_izHD;57x@sAZ> zqxf3Iy{F4r?kDA}^zw3+M~D3c;lgs3|8+SlSVpztBhu6V_8TOhOE`yY>3{oCIu_CY zuMp0r<0-O#nT~n%|EpyC80oGfq$4jR zU&V5&kmq8KGzXr+m0q zSFzGmF78Q`MvIP0bX25skfpp{S;cY*-5cru{v_K{-U>VlKa;}KVblLsPdaB6C``Xf zR2S+0ljtzA{#3?Q=)B%fVd-#Y5Srt7O5aaZRu5Ahe~^6KL3uryY(3)BD^s~rn0tuN zs%!-&jRVQ%G&+V-TubO2K0@-@^wh&+WdAUwgO0!)`&F^*{)8uy?LCwyLXQq-0mW&K zS11h&>3Es`r^6xc{Z94a4@%4Z6o$&kIhD%I9CuQEIF0<1&LJCj0LjQEpmS?eT?!~3 z`(!Ey!ZcyfiSksQu214_H!8EE=sX@x=Vt`jT|qW)Qu(Z+a$HM!SWP@f2elX8B0A{* zfDYfBbJ7QS3PbIUv!}VXC~YH2W)8Xzf+^(xU^@R)X3nov7OUxe(>ZjuP?`NeY3)kq zU^C^5{`U#3Ze)8K`7`I74tFxi=x`@d{>xL?8>uW;Q~u7bV5P4m|Mye5(|i2g*_ABs zY6>@o!W>K2R(UFqPbloi`^w@c~wM^ioyrE8%(rM)@jnRICi@8l|3_JgGRl-dAFm;E-K2a>zQ z-Sa736xJiTMgM!Bkv$zArPcY2IOW~{nDTrc9h4WB@?lY$-NUFnC@$Y6zhq;RufX_N zOL0&hoKMM?!lcQMZSqPoKcKJ|QkgfbWTmT98ma8kq_fT;8R={)zqAQ&kbJr%x9NP? zpHtpQ?i@k!kzeOgil6)hqp5z=d9bMLEh-D||6%V zAzzhhdG%;)OQovP6J1pjB~`zsAJ3&!=C4#zsmy%2luEB@5HOfF-Wo7-I5VDc!kiiR zcma%!X|q}E4?{ziO9hQSSd2Y5!8fw}JD_dG=cT{$b+VhhE@(*ZDo;najiXya|l=>JJnCN65o_i0c8* zn<;O8&%cBA_1zy*t3Mzep9$;?*msgAV0*x3-$fXn+4u7OedMb~-2uM(IkYq0^*56a zo-^U`J(K}Y4Z0by`l}ySTP577z~!5MVD=*PJawK^^0WGS`u|`3h+6$4$Q%7>oiX#R9$y5HcahGkJd|r~n>z3T;)Guf{&Vi*s1yGgX*}gUBTZ-E_jpe4QZJtU z@LK&h`2Dl+qc1w;IpcR7x;c3|`#<4(9Uj+653tpLM;?iLZ|lR55wf3=N9lWBBEMh5 zZ=Slme~Lcnw`mhU4g4p0K7?@p3%Cy@kAHx4JqJJJJNtR^S|d*nq5Yo;$Fs*X|9Ikj zi8$U#+rJ0?dwD3&`u8L6-=dHCQt)2_557-%&b|-6(g!of&1F2|T|>Xv`*7mr_v+6f zs~?0{0iU0hI!GJ(eB%5K%KWRKXZ+pceTpp3K7&3(-oFLUk3{C5g-rh{;n5jR(Kl9q z5IOS9`CYq@{27l{{{!XoL&%lqv_`&w&3S5k&wd*jKZHNejAzbs`iZ3fZ}aT&Ep?Hx zZ1!d3=WCJq>qy&2QfKAO7+e33@Z{O!_t{5KSD=~my~^)8zxQ}*KaWh1``$k%t#2jY zj0v^BMEZH3ek=8d-+P}+I)4HlAI0~J$l<-np@n>a*Z5xjvxH-eI{j$U+k+SF^z5gh z`5eOXJ>!}4oHhyjK6oEPGbX)1g>2pm9en4&=J1>SC~5w4q~#Onk0$VciFk(4{unZ= zQ|9j_KS!kJC*dP+;qe#X0q+_#dxSatU*Y!%UWD5lLG$CJ5x%oNxIOs&d1N8+e=Oqw zVP`KBH{s?#2wcY6DdBkLKTcYB&VB?M_^m#n4N9KiGbbH&!kwXOo<1ZEKZG2S!;C!4 ziL(y>`GEM4(;4!p?GtVlIl-qcae=Q%T8MA;?~q>dG9%7Y@;irro$or~>k>cdor9}C zh6dOk`K${KyyoPwenx$Km9`@Ezs{RSJb3E=i1hHD{||&~Q*VC&d3+AIuYo6hZ*7~f|BSr< zN5*(?XVmpQ@Mr%wGT(wfI^F84yng^%=w{&NJayo+FC&jUwKejN{xg3Y-#qgZ-mie? zsmmj9U^D9X-cKTHo*K^#_-UK65L)o{&m*p%MF#YTXD8Id4-((2g!vKr2GX+1@A+p# z_g3iW+vo46?i`b^Q_}aZ;mPmWUm@-Mo_!qs-_H?Fe*Y`z{tdE$e)gA1A5Wd{Ikcz1 zR>9Z#z4ur7J|R8e_TC8{&rIGwhwR^u%-_cI`NSu0zH6@#@0SxFbD*;qNbffxvo9e0 zUx(NGkTqk+-d`g;PaW8dckS~C_l4x;Dzt=O{o9l~xH^1J$wyt{Bpr33eL_6YpYpD~ z30eFM1TPLeKKhh z_!mGc&)a!^mUs@}_tVhv>;bFu)CfC+b`ITK-qgjj4~385f$e<~@qaipA4B+O`TeDY z{VUM$)cz8@-oZET*HG%coFhHh6gN@qNZKy9WHtyzc@3On&psuY=p-c|f=$ zU|-3549)X|<9qc5zTd(Fe)T!R?eHf48gb8v_e|pdZwR-|`!f6_z7F`W;QI@q`EnlK z^)CkgMLh7?ga7F%&pQeK1$?hV`zQE*h44IQUj^*16OQ<5pU!WdGs4s-(D2l518WfG zPeU)?P4HjB_nQdw`8>CP!FQGS*(PD$PS`)k^WQ@M7r}il@H@co!iTtOq<40}H+h(m zj@4bBDc?7tleCeCJ>uAd*BSZRBm9iGYw(zh9N{^Geoh=`q;(bf&X7qB`r2dYc}|i0 zDYBkF0}W|EgU6iw&ynYhbk9l4{41g1IfK4Ne(J?H%!}=THC-F+!z6d(X z?sNqne~ox4v-&Y`(z^Ot;E-V*zV%1Y6aOAO&;BHE@;v(+eE(VKc+a}ObTCgM|Mf z;Jo*E&tK*H2Y7!S-#m3_X77Q=oaa~J@oPN1_x>K={~Fvs<@XtM-_67C8Sgpo`gii1 zr}l4nLtB40Z_-fzdgyspf0E}X`2Hun;a%hR9%-w87jH?+_wf6@Jfvw)((@z42hHhE z^ZT0#^RvJiZ&u;IC%k`*H}THlF(hR5m85!e65^9zKhy!WUNd*tEtXZRM}`*?V-^3;Hx2^{&Y%KMl2raV`FjfeM{ zD`Ak+oM(^U zb&;8rEA?T{Qzvb!FGvccw_c`y`$I*uVAkQa!NUi>Q-XBl<<2{%6 zf5!J80scv}HQqDcb9v!jPv^+VJcNZFh<13hnl7eDl;U@%}XE z=u2jA<^5$mSAo;c>lbM6-^fG%z4x);<@rNAf0*x2gzq1PMt%dE^VB~A9O<6Fj^D&R zCoOft%z^JgJLB2ocm1P*5v~sH44TK{sbB|l$_-f@fHebNhT_I1!n8YGT?M1H=B{^=is`$OP}V+NnozYm`NZtp92 zuJ9oD)qgN4d;N`{^NXp4BeD84u?0I1_&4>FjTz z3rQaTIKMxJ-|$%dulU`8A7N&H7h2-2y~uMDIL{twI{RAke4F3${XY_h-!+G}nM#g^u#CBf~oKnZsiye0b_o#^iC2 zyzRXO`WxW>JijGfj2E*d(t?c7D33a2drEv~!1u_{{EfV+ck^%Hp$z88Z4X(UB7@on z>7~3*{~&a{SHF&Ml>aKcXOi|$C2vxX-VK~EXQZ{pGg}6C5nkJbrL61ZZPtU1_-B{F zG2YijrXP!3J`uSz2=_^(aUJ=QkNKY_KIDA%EVOUsp)Q?KS61OML-zBx^LxN=p4yN& zD8u<@^N_F83&5$THDq}zG9zDg!BJPvK8^VP2+tpd#~&vCH}L!Q#Q!?p$f*8jC|{nr zyh+2E=&R73^2`WV|73pitp4ZF@SgFm{TIC9F+&!!KMC%mkqP{2e~LG{YyG1L!*lj= zz&;)t@H2^@xX*-+G|m4@9%Ot5uT$WAyjTBg9{BB1XZ8rUhkWYL)uEdqvpGEXh@&p) z{J%;YOeXqKEvh90j-)Htcx9<sC&319ms=3Te%Rr}tv?^FAp+4t*y-q4@f z@40=~zQyvf?^XML)u!(Q_Wgz*Hgv1@y=Uo8?fcNe-)Y}73qP~(xqa8Z)$rM~?^FAp z+4q@!&+WVRZ8m-Oy=vck_I+yKGy6WX@40=~zTJkm?^XN0ZsoD`UW4DZ?>&n@weOjI zpV{}^zMok7+ILv~_PuJ~d-i>5-!n@;v)^a-J-6@LKQ(+_wRDF+Z@xcZzu)kq=6lt? z5AFM%_Wf@Ap4<29e$3EcwC}grciqNwYQIu&u$Kc;de$ zXWysxJ+tpK`~K?RG~qsA-_QS+`JUPL+|s?n%HzHE{f1vRbbI#ws%F$eeXrX0o_(L%_sqWEWz%)~D<<6GZ&BJ z514SZ_nY^seb+5~)xJ;N+v3mcduHKh_WkP57(R3RU9wm%Wx9<;Fx;Ol+fgjrU zJMFt}!@b*n&+YqlKWFGK+V@-R`_#TyoxbAW_IqaE=Jcvs?Kgh=w|^UFlh^&0J<#|s zppxo84oGAMTH7;UwbhSD@8aLtntlJGdDnl)o!n~)vsY*E>HE`euib6ey8THp>Nk7O zH21rWH+B2%#?EBa?RQed((>-fu(+TE)tEa)zZ!9?R~%JicDk<=)tFmPhJ(?hXx}X+ zhlBRGn&?5lH8?sNl%jQTlj3t(HQVj=!=m-5$mHO%dvAJVl8~ohQ>1p#Y!xPw1$6hC zM@3_Ee52RwKT0_$=6=y^zqyI9Cl$z>%}KK|Z5mYmR0P^>X62^D&o{?qL~&f^aXjiy zimDVXElbU~u%K$(p7vXdt4F|ARFQyOAi|~PKvGp-wkM;-QF~-X%pSKeVo#y*PN@H% z-Y8yfBn>96Yqe^U`$cDM)EO_pZtgbsLqWEN)5d1MJLxuiN!t+ka5QKYD59u?1cA|h$r8%#F)&lk;Mh<~3{H+vnj(490c)>=oy{XzSrj9(ud4V$B)X8ATC z9Gi}+H6SWsvnU|wDt>@%S9eq#6)$UO$6X=QX3t0G01qEGM?LN8ka%^3n^G9%2o>!6 zRtdc@$x50QrxTM$c}}+bZXGuHe|^-X2VYv=nKUPK*J)7;kHM&2j2d@}2b9aTD0AaE_o0|eApZxUe5uSu9YAwB_dZZT$I#JVRpOO>+Ls5yMrHGUJ-cszIG+~Zsp&X zR<0)mn)ehkzS$iWt;wd0N1pLfbNHs>Ij-UR}w^i=VSB}kGvrAbAgpv!@) zYa3J*mrVyE^&UpnHDmx>yOILPdI4qqR}dXUv}VAMFlqH>bUc4|vpkRO~>y-fND> z9=bmm^j1iz2Y812&2iL~+xVVoP6kKa7OmSf-V2Sjs5pJtMTyC3yS6Bo6Zv?dJ2|{B zii!hH`w5NtEK#*=4Sz6f)8*`(j3>nrO);wSl6e|xNl!E5_BLY_U09IH((<8DUucXd zFDmTFCrH5KqTdcp)`q?3&}dVGis{p2 zetG3W!;V91{r1i=pd^JfKsuJ)DfR;dqvwx;Pq6LUKKLI0(}AaV^M5%0hA^2~O*cKtdnpXnb-UAQO>< z@-*_XAO#W}edUo%QBb#>NXA8vp1QxEl#l}^yoG^LH0%IxS%ez0*`G*-yfbKq!q$;4 zvSzT1@hrja8@)jn)hX1LH zP@_DP7e-M7mcDLlXZKJVi|5cDRD$ViE30tAIZIuFtH`J%CRE3SdL?7k8FY$C(GT^I z-!j(AXzr74rm$;c#wD3Bs3LHop)XHUh<0EU$tn!P%ZM3ACBd*q524($ko#!k?Qs>R zIULgUP-|i6lhBE@53Q&!`ZhyIC^hA992c(C z1exyM9*nkizU``rRIwtY#(a)%kDv|1MoUZt2bdvTD61)!l)F(-cBC3TpkwcL`b9fZ zj7C96N=RaLs{165>GDb`E}MpHS4xz(L(L_JWF_;_DATWf{9h@ z&CTAFPUE0AXfjg}c3Y#(Wa?C>1oA z%cM$S-*G0UByHnFnQXQu-N(hm7chj#v zbM;WCl^{~!^*gF=61nLn(#)aJ(pHU?Yv;H-X&tUJ!whYY-_4;+H$$+jLbRG=tKIqZ zsnv1U$Cz@gLS|i`ieD6WlcJRGxye=LRTOy)WEn;Fq1290L8mAXGQXyRT7#ggG)v39 zgR$tlnVMsv+p#7*X?8kA+sY)tSn3Dp43XxX&^LxgQ}%i89rOon%yqr4Xx;(LKo$lP zqqT=`6yV|*g?t(siGaqvB5F1QnJDO?3Y>n-vk$lmP{+e$+?3gbGZ{K@9ns+qTG}=o z>$2`kFNdeLKygVK1ccpvg?{iRl1b;rcEi%pHWRH|T^4XvdTNZq)q` zng5YdM)5Qu89ZyL88+xt&DP6PEFwofLnE|qX4C<&v`jTNmdI2srZso_MRTMT3{+Cj zs04J1L7etGsuav!vAbNrnAsZj2aoAIdxK-242P9sDC71uM63<~hRj=npG-GfjOMxwS&hhs52yW(tjAKw>n)U-%rvn6KeAZ^3#ZgU(g z2QH&y#z!%pZ;T=-6px#&Daj3ToD7DEK`Sy{g|S7Jro=ZDG(+JcbTwy ztVyaxMy2H?bLpWp?nskZS~hj~QE?Ko0C}6{$B3FyJ2B;Om$w|A#T)6=X%rD6i6Vk0 zQAF@83T%yD8fhFcCW$0QB(cPZ^oG*b4fs$1^nC5WAzR9x);vG(CnMqT4QRoJiBr zRkBjpG|~rsh6r75b%Zifk`orC#6Ifd*0~Wmj}_UNm5!L6x`eq+#jM6kC1e(O%)sag zRDZZRI1bV8PkWDUG)JRu!F;iJL6${5wU~TU@kz>c_2I_g$%O`^XDBn9&2)-{SudF( z4XrVu5$D#+IL_SG!7hd9lY-H4|Hkw{;|v{g3=QLMMV59(t*w49^anmlQO`uZkb;RK zv(|1`WFF%HJH2kS0O7fLwCH}wUd#$UEEBSQ|KWq^jMSxIN5Xj;;kHvu5;;V$Oh@n> zX;rLJt;u+vsy5gcTawQU4Ot$Gl}MXir<<(M&oE8{ zd6AX>TolH3pgE+H7OYa87y3beToPq0=p3(!zr$>eW#K^>{o{0Dc9CgHBApm7BcM+b zbK%isU3&Q=w29DtVEU_oc3Pryh_;c*7UL4GVQo6g5RheLXMtJF3tr5&K($zs0>~9v*vw1K} z2P|L}7d=21mavR5UcG|R^69mGEA+Ewia_Adn|Myy8U@uF)XP3vmMpO$-&)yx4KcDT z=EISf+di?bt#te8N~x#rjyGH{C~J&^=|t{(?x>5+YSh3)6tdiDq3(>rH&bgn z=k|#OWV>z;(zPMclNf86L{5*UJ%%U7rRH>8?8>sN7rD)2sy0dg9e17MVnSeMWuePp zDDOz?M`=r(g*J>m#3#jOK#S{gB{T@hbWt}4{SeK;V04Qfar>MEn?>G4%?pe)-YZeO zl4gk^OOWU$QYsf3X$P2bic30hvT0P7SF^Zwi_sBHQ_9_AmX@Wn(#3n0Ydw&j*Nn7~m{^}4Jo(^wxPXe^>hYm9MlKN>dBKYdskZ0uHyk6BUd#(Wg%FRL@IAM*9-3sx4n=(2Sa+ zC`QpFCL%G)BbsbZ`&~x;hc+b-M`(Z8NkZXYr;+_(! z>A}(D-p<Z4KO5-(@{$jQeWRt}pOA6}zM9DhLLYJqQtJHzWL#J9??Y49UI@X8ARgp<1x@PSoAvMkP z{$Mc4$M+IOd!$gYQKUA;JN@Qxd^iYdO9-XY315U6JmV#%sm_>IIt{Jlh=w~NU7m#l zWMLin1{3X2`yEtPfHo+d*kuhl1|O=pBXfa7OAZO?w2a)Kj26f$3quPp+2~=5Bxqf? z(E@ZOE*Y7VII;{TxFCPj7=|4ZWl|T?OUhhlJs3(xy#zb0ZG>v*Juo82rA+#EkLnE2 z_#R~nL#K7SCxo4B4K46V6{`S`z0JlgM4Y%ZcE>x1_>o9w-FmdsJP4|x_!Hh|v1!m} zBKiU!NErLX#=1}F((<&AG2bdnk+K~l@PA=%PG}<<{SN6MMp1g~YPL9C#4xgErlDHfh17pIg&D4PbtnbcJC1jrtVPy@E-AT~T#)r*zs7{?xagrs?^OGJx7*M@Z+oVHC z@!o|fvV{y;)K=oif{lcxM?+JZK6Raza49bso4#?k+dt7RCTYoK-8`GwR%q{%#Nudx zJ|H@&k6wgR()6LFis7Luvx~*U7K3+{-(nR*wN+(PnhLs6@wj`gL$>cIbCs`xnHa*J z9%a$*(HoZZ{tAB5%I{D6Y75v{uL_ZCUwyV>7BT8;4oDe$b`akAIISu%;ly%FJn(+a<1%e_nIAy&CWR{WFbqWj@oyVb+syr2Ynf9 zd_*doZKqIqa7pr7DI&CCo#ap_&(Y@AG^2@6Xo-DNW1WP({7u$KcGNsdEM$p$gH5eL zK#~Bdxweey(~W;oN)ZeBJQ7pPxg3&AW}cemLWDnCP4ZL>yFp7dZj?R~U3^M{mrHki zuNme5sN&L?;+{I?RO*m*qKArahYqy=?Z`Wbl`TTI$hKs3g~UebXC)P+V#GfGUJ=&W9@w0F#M8unrAQJDl# zbiAvNp(wMd>h~R8c)rkJ;+8M9D?VOz$#el}i)u-(tl4gLKvbg&50?V-*a#=`)af83 ziL(y$!{?e39u_U8spC!5!V8Vk1U<|Sq})NFYRZ@lp`Jk7Xyv5RljwGPjmpOkPO7E04fTN{ z4>8izUg@ZWEXLSOixaN;B6=J<{Py?ppH16Axqx&G4oeRdlTJch$WiUAq(U>tJq>Abm)B6zMyX5;GZAp7$_mj1#X~AJW7r zflaTlASXf)p=4IFZZ>D670>gPAxk>kyGcmxhU}J{YPblz0hw9O{i7Fg=JAng}cc6|B93 zO42f3$(~M#80AW1oQXZ>N!Hsx`_Aj0G!YB~WPtKB`%V0U0=cb=1tCE4Jgl(D2K>yup= zv4`Y4$D`b+sC+~z(^NMz$u0rIJbbg<_FVvFI_!bqZ5eV!bKI$Jh!yH<>4I zYk~>np-Gyt3~i=XlqjztH)2=TJft91ib3C$`mD87{hxiDPPpxEq<<`AmnfCl#JiNF zd}dymAv)$4O7&Sxa-$LL4wNvV2!$(*QtOsLOm)%a zxGdTbQkid`6Py;S3+fpM$6dWz&Lbdvxu&3b*8%qfkWdDC&PpG??5w@py%Tm*%{eB{lv*ug_XaQA@JTLS9w$3kjN7An5O^At z7;%P7Tpp<+N)#C)qM&81vp6!V;mRc4e`>yTXz9GZX*<-;x_$g;zM4q`ni9(Wgu@u(#Z-+>Vydgx1zDDbFYE0>JkBRS_Ic76N?|h>wULE1QKmUuDhW<3lfwB2XA1)7(k`5! zo*POVwogubP@Siz5@cc(!G*mO;zcXMNd+iSQTNimG13w2mMtGQCQU~!>lCSG5~E}= z*qOFMHTOty^^=v(c<$(tFqM#NM6yATf#ot+^yz{dPy zXTTa#6Qcv0b2)9H{xNxx5m{B~|14i&rnLyyQ5V{Hv|N zMqhvXOZJEh9fTI~URwTrl-#WGmNOOS$YnjQ6p4d|369mPz00chrYtjN8xI(i0xny> zNOAJzE3dS}gssQ8>Xl%bFE*3rqgMC0tV)&C);sM<*F-ZauV1-_kS`@w<~k>;;GgRR zm3-KJA+q%Ol`f@`Sh}M<9&vkTa}LgH;#}U>t-sP|(|hFGdKQqqwyi!sRepzZnr+?8 zoU2=Qhcg!GR=j6B&?bs78$)guV@hK<-N%9IP&9!kTG{Yq@aSwh0&-B9$M><{&OE1TvF8vt4`-&}CDawsE$)@P6*+NPx3RTJr>JRNluXau;%-`84!;k45l(mD zmyjm4@q{+#nT)7VBN}nSm<8PUHCuciIB}Bt%4}y(_>EB>HjiNXIZw*olA-a;R=HU6T_&-8gB2AzvrlDi1f}!C{-*K#{>=>8iIFKG!*m(A3C~TWJ{W zc;|==9bBM*TuhPp7)7;d$$45HGC6eC_MqS$z$@qz`jw^2atZyr5-BgwIe@ajtx<9^ zzeJ{zPeUpw3#lceh{z?X&5k6>HG(jV$&l;$sUBUrvW6kD!Lz zuy;Mons4^Az;+S*Is~O|PK$oJZq~eV`QV^+rFrH0N%c4F%kRY?w93H^-&+>f`Q7>gL@xIh;`+0=}!geWXrJ#dTroI*X+ zV{@e*`|7c@JUTwY`w-J)0zcBTyhXb~Yv~?DEM#VykkTLT(4!k4sszJSJNc%D$vT6N zyfffv|ELN2C5O#8iOFKb?QI`g%{xcxuj^s?=Y7td;0)u}o6H5eo60X|g^H?Y*ne(89Fylu*|I)>>$oS#EaPdHAhuLxv2g++n#F)7)4 zeOHQ1>}YZFY>2Fl6aU8UpmMKvf{Hm2z>ZiuWH!NwUzx>4>#JmOxO*s$r;)_QljHkl zuQ+Cscs(+v&QhzYW(6mv1`b*Z5u%Rv5GITgcp#UPxOOln`3!~>b6!_aGUZTkb%=-v zqXd=MShJhdu}DTJX?DC*L!0Myq)81uw)qYklleZ12sr9C6#TSr>`99|<4v<;cr9B0 zI+D`|drpxmduDdGC#x!AolQcS>{kX4`D1^ev804+~l=Njgcw$tW_E%zQLa}KgJ^g4XC%Q6|ixww!R7TD(vyUemtJMLe43@pr z8|yK3<<4^HHqvd^K9y2xEG}##qNqO59bj5_MLRD(isupo|iD#&Q ze49}?faPH{H=`eoOd!%tg~~(xcM}&Z02>LwO&8*YByE+V6>u&ghen!Nr%zgn-9g+z zMTrg#^Yk``D2*0kZ%;@NLlap9vyf!C=nV97t%wRmjv*=Zf95-kSe-Uq^JoxstjZFm zixqW4k9yci7eHdIku=IUC96D)FFvCwSjU{p=OM#{sv=~_N*vci;*=YiU0q6Wot9{5 zl{j4y%~BMH4fv3#YJ7FNH2lMSLy#1S6_Ums@y;+2*MO0u#6xaUcs8O(F-a)SakIK@ zsG5sOp2BO`jdvnMY(nkQCUJ`t(sA!(wZh|yc8=oQ?a);9f`}j?!DR_!gils&>~}R) zkSH~MS56Y3Ld(fU+9`Rl%4}$j{br)Kgt4WA_hAq{4BQz~v4e8wARf2$OOq1v5MUuS z!a6>u+Gsiz7sD+;U=%PImK)Q%|8bwIl#SvR%Fi`^Q*K?qr-Wqj{o8yQHEqqg}y27Oj%)K826#Qj>iQ8^n26rB)Op%abR@N*KDu0GUqX~rc}l<3L#Rg?F; zvDHuLfjIXEcs2As>TaY93g^i4<)|VioSnD&x7Ie+Ht%_sGK=GYMso;xk}}32V^ZAg z<0-@+Ht@47t0&7Zp5K&gsm7ZCbFNxMf2RSF0i(eM>yz-i<#M59P4@Y>dA!F(${0O= zcgg$Ix=rTqB7G*L3rO47NsC+HksT~PsfkNaXOMl0W%VKFmwbwOdRNa!Wb8s!ACNnc{VvB8^dC>QH*>@{OH){;6J4;?I?(AG6UF9-HL9z<+|h~ z1$>t{vv9xQ?&VJG4=OSUw|~n*(}quvzlVIxf;v}Pd4##H8K)X)sbL0{fP<1LJ>7{I zK&8|2JQC;v&rNmq)0^zVNG6tb?6KLcBzRaCM|yiI`ZFEj?dUc#FjhYaT2buL$`bGcHT%a!U}u2ko8r8<`@)wx`$F36SY zT&`5-a-}kty-`&ONB#zpSckz@7D0{L5H&T|2xv3(ZQM>8J3vGMnXNsnjz};I5(Z{e zok0myokH2*Bm3xlPNf5V?_5s8@f-jTyWDp+ z7(c^L4tZdgwY(nkHz8~fcDkKynOa7dNF6u1DlhU&@oZyh5ksZ?iD-coyPZggcE)wr z^EL@gzHc*}MoZUFpr_(e6AwfHP0xp;vHXrlz&N>5xEY$ct!NM%z!6Pm7Bo|Q6jAJ? zVY#lUy9{=!0$_iwWW=0K$tgfiyxxxUD`vIWtQ@DetH*34)65i|R6vc!B9; zv|Rvag)ZkmKjKw$7IKEU}o4 zLMDVoOAhA|$NZ=V(!5#(6WZtitAKf!8m|xMV;)2?E)(BTlTu)~9Y(3!e88c%G88F{6V@3L$P}<4rxoAm}-cUhvtDMH{99W1_gl!O&r9I&5(+Wi)-W-?~LEzWG_$4Ikk< zJ_!f92u%D1d?p2OxA|me`)1LLtSUow>CfYlCPZn)yL|#(tlQ`3H)IR#IU4O|(8E71 zZ`x7!;}bkID=rV4lgWtPU-3dlj1j$58h)~1kc4G$ZDVY29YPni9!F#DQuNWPF5R_LCN7}(=a_D;w(1a4ArlvqI$i>jJs4?+c-raJ`9 zXX@mm7PL@M>5fS5#}La`K%+O9YXg01fM8liYl!ZF7pk5Swx?5R38jb7)Y*-geVWaQ z9{4CVOFSx)6G@zS)laj! zHX%h!C~7{hvV?1#cs!`f6`Yf4C%O`*bMc|$foYW}(nEG7!Q#&v_M((63c!4o2 zI8UKeCvo|-9yL$)i~G~Q&iQ<#9-WlEhbDnawoA>aI0_}_7FcLjV?4{l7Biwee#C{= z&S|HJE5~sd(twk~LN{(%Sxc-b8p(;3wP%1F+t`dkunV%Ktidz(C zDdtiv1WHyWL-LJknWqZ+)T$l=(6{B(E=6nVo#wHUDy?1*SOEx)C$JiK4<-*njx@W2kl-Am^jNSWK51AyJCFZ!b^0!5ZZ znLP%9YOvdu#*0~@MrUd7hjA|jr*X(}B0jv<63HV!LKp9$A!Gn$y|71w90pKETRwD7 z=1F(r1Zm2hquQ?Y7|FKYYP5(YneK@(g`&)&?^v+UJ^O!;QyY^);HGD`xcLDml|PYl z=e_rrXW%uiLq_wcv-1)du(_wIN*Q>TJ|F6Nv(L~+lOwWl-1nrDeBY3LKou7g`UQgA z>KBzrkrRlWlcW7X4?9xm#if~uE8G)yMu$eB=~Hi@QNvsa;ZtBoF?D5WKITNF~1SHo^fL$LC_@kOA?ks?$qV?#+Av z;@XwK&IJ(7>@^1}1_$TT#;3&{awImjwS2UQ*i`B-z9gejoFF{ZI9%?fsUB+4u>M0K zy(7_#5)r6lTOe>5q690r%aqlJX4+qbbLwhN>ROntAQK=Zo7M^?b0w*8>t-WSe2riU zuMwt6iG#<>VH(!5T?kK1PjSwfuOvrrn3OyjdWLf6kWYxsIO`E|T3yhXVmPZXyP`}$ zbT3^JR^A#Gg*0-|wxMON!}UStWI~3;NW-q8kX<=GVLnTnu?bb7YsT>?-}s}rU4Q9C zCEAw$#(>-Z#Qc{H!xoJQBWtKn2^)R32}4bY49qqoy2FP7E}m(Mvb|HhT+WhbRFz-Z zPPD$(N^b9sld!R#CLqNnH&e!_o!u7}hs%~$vTzG|k`%ceXfa<$z}&G?3KoA(1acdN zxqkb+SOV2zR-A|+XV3a{oKtXAWn!bb?^aCteHA(91HwFuWZrE}lX`(4t zGT0b!Jf>_!t0Xd3X-koYPf%6kJkp9;OtA+*+?v7&2KRu(N@ggzzmp~x}r z{T?cw30j@<^}z)0$=%83r31>%l-+={9F9%T&+Jzt^1QSv+3sJ5A*qGs(JZF0{Txj| z=F*Ef1v*c|)W@_PGjS@#<)tg@yXT2F#b_; z9mOTuENQ?pN!^ML)wd_v02)jG7g(DUZUp`ywb&Egwk+)W2~Kmrt4q<1qyWT%Cq6gn zQjkFHeK6(lo8wzY`$e1mWIh>;FWr`zikQ5YyoF*UP9(VaR0gi}G_qkWE|_)VSOxhM zz6KqFbMO_~6cd*)CQ)^AP{|6B)gi55vBfzvCB>-V!x1)QW$bk0Sq6*tVrLlkx_GdQ*8K|$F%28v4}YD7W6u;yr}L?k1&QMN=Qx|*!1O6 zu4v56DT9b}OvrnV33jVp_#ab>YIt}K?umBrGy8l-_MK3fbIb|Y683+L)$;apuToU4n4 zb9J$Bt_I;u*ooEir=N+@`X1g}W1j=vtSF;`T!nNg8)Fir;zB0Ek+Ga9nuWt%lBFaJ zS@6Y7gX&Nsn1~xi>dR50ksIN3a*NB2x%e(4QQwKR9{y71TEOfEN%TBbTG>1u{2E20 zO#nEh9(7yC`$0jrCTw?dHmmz=u(8K+iUYPoU_n@b;B=y@0I~+nW*t`U<(33T;VkB? zN|}kcQRQ)pY4&6F8ArwLB--eX$JyiE$emWdvl)kh8P8nHwtIjD@z%;_lwX6<^X)y> zrHYe1D8o*eKZ`LAJgv-9phERkD^5a;k#3@WR4^l#byAzP1nB3(7OTxCp>Q=_fzj5N znA-v{F}I?9W#&`vo-o}|=)c8T-kp4H`I-Yr*~hb~UYOfdBA-^f1)O7bI&!(s$jlW& zr8QHB#IkT?VAiT2{a?do3|CgeWt;swa%4oGO+##<$MRAld7$9C&7q_7Njm1#wi3Bq zpX~a_YD#HKJJmNg>#$&4f;GA3r)MK4!}$p1V@DRlN1M`m7Zb^$_UE8ksmB+4ad4eu zo`ZjZb=GsxI|WDjn@e&(BpW6r9t zsrYb~#{)rss+GVez#s-=FU9M|% zAoXktN2K!fHOsYNOcr)r@+cdH%Lp2p=`F&|c4EC)qMO4g86Uc#icF{(Q4|cu|L;yN zF_A?FkeIlfG^l`U(U-vq5d&4Ng;0cc}(S-yLHgmdzmIZl)$uOK8En^x#!s zYfueRGr_d`&N!Y&le5PU{l4KOktWq*Cmhk>IVPP@M34uxhczr>sG9vEEeg7%= zQ1E%8DkGP@9|tFLdbRkGC+Q6+?ao{w%bntj2pOi}hL(Rm0X|-f2-Xh^?BHoS>;@Hf zipv&p-8AkJBj3qpdHu>x@)rW<7vt(?LkSs)$u?@4)gTSPIvM<;*7!(roQ6HD63WLUMQlaNZ*qH!j`%)B9u4UX%n_l z3Jbb`XyZ||`z+}FC^ms3Yt8nK86P<_53}XYT^mg8G5xW9=ihJ10oO`^K*G+)9lNC3 z8l4O$gJ+&xz4+WSUdkNTBGuB8xOybJe0&0rndry_*#i!^JiJ#EP{Zt@(VygFSEkVh z!8COkebSZDiz?d1q04fHxF7ix#tRodMwyAmKA&07=vd}`t*aYMh)}2c1D4o319r*_ zyZiX1G`Ycq+4%l|4L{T{j&qbVLOBOyk;dhgFj4H1Pj8Q4 zvfzI+3-G@$acYJCSiub9Ke&7~d|bN7(I6hRauKtT|GIMVxhUB6OV7pOu3f$!1-!iS z?6oN5wX0Vn8{KGf@nLjCKy$3S`n`P3A?5y4h2(^qUT9GvA4eFWJg4%t6U+)-v7nu) zpvw?;|0TQ6IS07cOFYDxo|j;@Z!l&H8;N{Gc5L~K=63QK&D)eXI%-6|rQBSKNy}4Hdim1DXVb*2TwA#mSui!dD_2&o2ScL< zq>5dAHq93IcRYJ3vI+{_l}pcExqLC$`~-IW^7ZT2E@u3$J^SplE4h>}U%h(eN-pbX zFI^;j%JS0XD=U{^kTPAlcKyobt5 zs@5pzK?>UuTVir2q>_4YC3$=>NJO68P#NIie#!_|#W%g1G7w)*Zl;XjK_Ll1HUgl( zgsQ@fBs%A=Jv}&J=Ri=g1dtqHLY2l+b8jG?SS9NWNMk^s`c}QK!lB-CHjl5fe5wRJfX-wmhsl(Hjz{*YFJNglw^t$U z-ua4R5)E(Cs|sT<#^h3kH@n*_F>1=H!l;H=g&=co(|r}%=phfsljzu5x0Bo-oKO`o zArd>1C~&fI;i<&*)6sDX=aHK&VO11bL#q%jS!RR3D4|cMT2u#9US=codE8iRF^2;@ zl}HZ)sHih8M&%?rep@M{BLkkwn)2rKfb+=YD2ns&aw^5Scz?jac~q3LJ7}PaEIRF? z3N1|p-N!du`~8+25Qa!gajiu1P^Mo#oIg6DD%ReBelq9~Iz0)WOUqJG?sH$&g~m?m z*vAd_*l*baGahhLV+qNf>KuNcLCV*;DpG6;x;a&5orba9W~j?vpw zD3RT(I6I_Da`5&@Q z_34e`WoMRdsE2f|1U7~nRoqLs{qE#3_r-AHYzd`j7nQ-=qK@9{_Ou3<(NXm-Bh6LF zWw^akpp*o?>9m5}-N#o(;r=IA)|NPF--c(sM=+I1cO`bjL>ji)_3K!GF0a(z9)dXe zn-fAjCkGzxiVb{6d=krHs59WGO6jQzWkPgHvkq8B^}6g{C`P*bsf=ZAA;%z=l64Gp ze|k_xXp->nE@O_icFMr$nvXKF+$ol6=)^<_?TzpH<|}3FE|r=6Dya6jWiF^&Vo&6R zsdC_9muUfui(&%{*cVq`a%E#$h8Ug0YRe5)LG3Z^RS2{2kzzK(O>-$icAZ{ck)edX z3S(m&PoK@p!xM?g+Py;(94cC@4ahgkSzuKx#Ti8+ltuG6s?=Sjib9j;5Wc3wJVYo# zf-y$(z7U1cppu^v%winj$Vuf+-lb(H9+lkF$*dH8Iu?Z>CvkEYkx9}k>Fj#jxS_lM zeQ!?dTjz+hm@R$>t$f2siF?sE^M{6_%rs4vNt+`oqd=CA{<$LTTwSO|<6MXZg@v zD;*M*qh$j9X1%CBiYF(M{n^!MG^c7rX>W!dzs+K#R6M6lR;JBt5s?hYD#748A!^xm$lu1k6BRh zQe+%h9g@yT-Ys!hs>DwQPfOCB*bgL5k$SB?Ei{6;-7UEE)yE#Q*ij|oX*@RNit!_y zC(bP<*{sP_kmG`P2O+jtN~ZdlUS1Eh@M(&vx)rgRG(WHvLDGAn@fvE*1j2WiQ8X-Rwr4WsH>5GQD&R!jzR3 zCKEQd5h^3PO=B_Qq2M>8e$Vg-=<{$}aZeo=4ZebpZb>uB(`KUF6cjyroT)hY(z2=q z$<)!OKPF@zNvE*5xgGSHoyO60Jh_4MkLgE3ieg3);nRv)lh#<=P^WCfE#hCw&Ar7mivhGuij6sejCcdAUZ z1qP8W;^j%7*62Oq;sWp_udO?J(?Vf;Mz4V+>3m~BdnSu(X&@=}!PT73FIv-k>Pcdi zW$UAREhL#IMrC^Js4Hcyv-HYXwYHe)j?SUR=9>CSTInUWBE3QSBnz`)@8QpFe=aTS z4|=`GLS=EAV@K8Xzt$EEPMW|t);O=Gu1gvmL#_~(NrTCS{#sgQZq~lgP!}T$krf#= zL&=^3V_!*Vcho3qIJxjQMp*mUH#u?OWOiVKJ6JTKG7jR=M9%@ldMLKEt$yz$nMYMP z;-OC)Yk|gdRA)`)MJ*%gvg>bq`%BA;V3Zr3ys@so(S#Vq+I6lr>**mpdgO{tr#gB? zQ$k6TTzP+?k)~W3+H31vP}!&;ZjM-J?gAGRC62*7cvyRYXQ58=1z$x$r<9M z#sPY9tZ3d{mQ9jegrt_bn6#IOnA?t0yesP>c*nzW(83UUjqp})@{$d~9{0zuF+_Cy znJrt-kQ_u(bCEx_^Z)-M196X(rr(#%@9Xqe)iZ4$zTsNm=L?Pd=>X{ABLk1ypT|jm z<|?QMrhG*E*x(XuI%3y`g2araFY1TNsAQ0dnCY=ewq`}(pzLUI*}XPi6?|>AJ++!r zTL(SUOC@{)eA60IE5d0D%HUQ~8B~X+6y~Ji+GR5So=>Wh5o??%SudruC>zY@(!5Cb zo`Jkn!)$6F@08Yr?S9ejMGKSNfZW$5dS$vMx6bmkILK<^S#f`+9AM{gFzvOsQJACY zUzzMkBDi0Sr@eE@3^#L_npRFH#`AiZJaw2@D3JP#ILf~C;uU|EWKMQ8)$XvNI^-Mc zCA8|RQ^41*yj0dlf4?Ox>n-x!L`5v8b>M<83Vcqb#c1#)+}aRVt@Z+>o}R)g9$iY# zv#H^3yVHYnwAj{5VGjC_?LtCM4#v#R7KOnzvm#6h#(*u?*xnMS1?P&8hg(`^Dk7a9 zcPK|zEnm4KEO*n7PeL2aIf<)Fz*co+c?jhlwSAW5l*CS*LVnynrln;JFmk2T!fTi- zDBa%1bJ?ZFh6*k^p$#m&e0-;pRg6U}?8bL5iV8YxoP*X>%)+SuCBhynFYM0D;bUfg ziiidOnmQDh0y&_{-rpQ`6b@@LTwnI_S_#ba}E zLgz=sQUH9kq*(-tIKpIfmI2O5^0BNVNA0h~RtW;r65eJwWS~14G#i`z$v_VTffUT< z9QD9)A7b%g*Q7@>BQu}aik6nW(stjfa)yH;ic8b6BT3wfX||g>STx~!EhF0cbfjlC z%h}A_lVyKLHQezvx1g9_UrwrZa9P7uKWe?;(pUr?j+Eoo>b|&oMK+n5Oi}8nt1L0H zgSnV6ryJYP1r*oLrC^^nMx%3B$f0qv4YPuXlNC5NDvzC{3%kCoGA=h)>J4Qs1S3{|(9m(qhcPe=lqu0JSECF@O|=afZiY}cObv9l zZ=g5CYVCreU-r=x{Wm!q7UF6U<;>=GFq~|n^|_ED6g3$bt}L$yu7#wU*gY{LViJu= zFAChLh!t2##_NapX?dnnxi|@+qH39)>n^3Ff24jvKBTxYMb%CB!8>Zv#698tBUx#3 zsGXvaIg~4^oGY^_Dv>8JThko5z3I$-7Eg*6r!crrGTVw8!)^5PNYe3#6$^#o1JmV+ zq+*b9hI8E2yFSi{-ofUu@@6NV`;?oq9=duMdz2;wQrruXwzL=GP3bR0hb%5630Yo9 z0vA`4&!cg5Y#~j_3q6rTOKRl8u%&GcX;?X5fnr%09Bp)EA(H%=TD_2>r?q?`I^T4@4id(c+*6W0)n>GieQ8HX!am>0ks?dmITEyKlM5pj+H~_n z5}NX$2`>zNAWeE9Hb;8>#znD11DEko#%aFKEMs)1yCixM-%zH~aM{pk4kUuN z`m`x%RoH{ny_)D z*Hl)7mD!1Kx3&G4YbH&UMvly}u&P=yGY5=FJppDmmrWVAg~YH}JHR7N+;Pk&w+H** zIxum2>O`{5QbDsFl<)hZ04c(iB^{41N@3qn&M8gbm#tFV&|`B9i2`oDJQde+39a8W zpXiI$n*yTt%>DJ|9NM_admyLt_y$aa!-?5uS7Q-7c8GI_4Pp8H(^jN#f)c-(K{m z8oV?Y`Wbf{2JiY`{klKp{%A`*;q29{ z+qv!-MX)P5Oy?&WR#v>6v}F>s`&Np#S*FJwB#%p6s}&|r(s7ZMXp>E z{x`d$t)_v3r$d<2q9c1{g$AOt@~SByEYoGS{!KU z@96y~1}ah#b42(GzXV>2S0>^TkL{y;)r>oFS%t;)wSG8dny0BnZyQaAAf9Vim2bM5Onai=RW1Et8c0Ns<){?SN!6+z6 z<3$n5_4bQ!RfB4X#kq88y0^V{6Xl~a`mt&tHoIx;394EWubm{8mFtvKtBTK7 zbjVql(D6D%+}4(s-6$rbd{AeXmXq$n>F<6{pb~vW6`(}O0%CtvKib-*q`TCdc+1bF zmnt~C7(opaYOWZV8%ur9E@OPq8s!)o8vMlP}v$)4!+);(k(22 zXQ5JB{ffIP$AsF?qKHWyNoUFzTtlo2v|XxqAx4F4Qe8jn;-~Lm)*9?SmzFu~kL<$m zLL(Jag3|T;%0POizu)O{$w!m~kIN-Nu)UPykxr;#;YXDC3t&+Gaq<=*rAw z)geiR5tNTs>IHS%mq)p-mMP*oTf)XNDk0ONDWS7;zd+ht(@ASr38NCa>~mF(55g|v zD~i5Mbigbmt{T!AjkM0WWcn0vhmA*O*@{tFX-JfLZxbiOvln10%&~UEa#4M#5|3qV zNYhtJLO`W#6duepj+Y~km{<0vuE%%bWM#Ud+r7{T)|#i2X5BAtphv^W$s)N&$he}$ zq2f9k!&6bDU2H`OCj>Y}+MIeBC3t)$qpp5wIYEj?WrR&|qH}werEupWvr623f&Xk>p7Y-Az)J;?5?GH3UdmJ+_@jRY^Nsg&Z6zt;tKv4xW{o z1823{fwdp0bqA`#+2M~bUA4Eu;Zy8AYZI(_%oF*PjoA0(#~s;9FG$&KYvAt9j(B#n z%9pHJhc9N{y8mAH42XZ31Q)V)mqoKt+c^wmfRGkPC*fquDPjB6eOW0PHT0E$G%KTB zQfJPpj;K9Kf5`43n^lfr;V0J@o4olna!x1iYFz^jNzcF$w+wSUVpsF-Kn9gqBFfYx z#yd<5SkEuRtr)zXPbw2}O0pRWj3`Y4os1%|MwO8&EJ#^)ja!+rouMSTNFouPE6OOa z$<^+liiyKjF>|y=Yv#}?5z z7_s9+X&RfDst?@Hd(C_1rZcpII`||tKF966Uk2F?x1Z`I>cO{ATe!|0Q14&y5;~A#R||d#zs>cG|I~>zWO>k;V=FvF4HId zN%ILK9II~*W|H7P)smCuKus|UGUD;HOUvvKR&GwCbxc1?m%iMWE`&+HSs9WOjpf)S zN4$^B(-f+gl}C3c-q7m9XrjdzCxu8ubK%x7bombq0{B6>1mFloTC z&56F-Ih^bqb`JvYUU85?SaP8rZL)@DEa^UU6iLb>LKsWDi*KwR*$R6Qq?D7NgSt65 zX7R_SJ`Yz&Z&Y-m8eHI5-n+vk!J~rS)rE_oQGMzZeU=Pcx5ed6dnH>BwF`AYF?=h5 z`!*{&y3`pOf*NUX7avVH?bpZ3g3qn^<97UUBmUqlYgaxeQ;ALuyirmaj&HE(tFywx zrFxEQ%u^OAxmo*8T8p`+$jQVIB#lR%zRQq=iIGgw zy(QrOE(dn^KkRh&15%YhYpqdVp#=kP%#bsu=0xS!oIX5`*fw6!}d@%BApGVnil2M-Y7DEf|m z>Y|JP-qB+{ZV#|<`WRT%@fQ#1Qj6W1-O#P>?AdnTOUrB(-zi?!GXaxpXUwq{yO`S9 z7bO?7RfJ(3sHIoQM_BZX=qurWzx8-j!37UzJF--yN)J@f{_|c%P;=e_EifQ)@mSmM z%3V4UJ+1xC_pNc>D(_mG-P&ovvPs1p=zt+#N_T$EjTuLyh0~4LGoZbVI+VG$1`^!BK^iFdW1X z>A_toh~#9Pw9iV3jJlJ;WCKo*=9WUukrpXTJeyJon~%+AMwQK5Nm=D-w%$v&}lRCc+B%Lik1L9Xu* z9}`L!6c>7# zDD7~}!AcHiPjYgfVspz#Z=7uEDBwtvPBkM3 z3kLx?NY%+k4)<7kYSU=bD9eW!>E>V;uJY`nghJc9YzGeQh9ea*$~YE=K+m4%-lL?4 zjkssz;&n2vh-ZX~7P<}7fzf?D8D==pCY?bU-_qpj{iB<7Iu{yp z>X_pmfi-behA13kka!*UiQqZ&Y|a#dIvH7fd!2YS_RE^#3q z-{ly$yRsuai-*Gwjx&&4PUb;k%k>|k+@wuFAmmr3n~6S&~PS-)ewQ@TCBh_p$o`prg6Zu{6TduSOVmEF^S4@W4V=~n#D5MK+^ofL$kk#x+EK|HhPq1 ztTDqTAbhC+MO&0Dc~0CwcDdR*`rM2@kCGdQ$B{-Z1^}FdjEy6-^>k*HsUj7~P+@$R zl4ab(6Ai{g%N=b=1Fa&lI_UPJ&NhZ2pZ&l#R=Yej?YdKNt{Z!7IwskOrvjtQ-3e+*r{uVr1ByA~|vE2GKZio($+jOs| z7Eqk8i!9f1`&Nya2};u zt}T(EV-!bpB`i#Y(7Fvp_8Hvb{M=?=j)0Sjp3s;=b-~dVo*En)U%>*g;d9H1E#hJj za$sd97Pmx5ip){>Ebfd^fEzpfait}{SwSXeVrPoWtrUz-U7LI;%dtD45>(yQYNdH& zU0*4XV-A89(

*`WW84M?-fIWJrYy^QCIs^gQH&aW^nQyBq?3FECATW*8^}R@l;Y zEQpjG)4D3M(eGlw6U{QR@476CjICN=+X&WeB))z$Oojftkf&pK!rW`r`MqR08px$0 zzn5v`;7PU8{vPsVImTSxAu|5G;M*NMbKs7{)t|Dr1@{iWW!4bL_R`6r_B+0rDS3jh z3+)v_oHwm+=!V8K#ob8Kv9#=>_hWx(mLzOQBJ>)7b+(Q~M=Z&92ajW!3bGx)3%HsC z7gsf3IoY2c_;rk@(wW8(HR(9kh#zA)cKvF0H|V|sO4g!#gHA)Q``PNtrS=!k$2KAl zX&-}QvtiDADZyE_l?C_`T#v0NLFpc&GuP~L6aQ($4cf`<$Fct{P1H`SjL zh)rlNaj!81PGdeluQ5oc+2!6rlfgJhXgHN}Un@yXudUbp8Dch$4mnc<0=!pD#B-1C z-RcuQ>lGBP{hmwXciyB_>{3Dsr9D5sA8{GdXcrP!r(iMp!o(c3?IF^8baY+ln(7$Hr8&$C{X5Tt5SxLV{TnajyW8(gqDn8#B&v75ndfR7Z8m@ zaG4&P?SWETa4AqjTAodb)0BNFI$x%h_$;Q%u2==dFh(4k;yO?jd7EB;G>A@9w1I-G zTE?X|tC2e=V;LT$QdKjMSgR{@U|l@s^nuD`8UNfYPGKn(k)5sOCJ?yV3RA%X7lwi8 zkheDP?cU$qD#aY}psMJUK9^Y_kvdcnhmki9zEn*nMtOM2gM@lofoUzP46SYLwTYmb zNty*iV~0NH(q18WrF&Yc%~4Njo)e*0=14<1kN%3N4Zcd*kgSNJCPHliod5lzAJ@RXCVr=I0d+- z$D{G0aWNwbJGb(vb3(gstE@q2N#4F$8`uc|Z>vG_ixc!rMzu}3L{%WH(qDmbLpCV& zht>6PWrqvb|HIz9HMfx?`QrUi{3t8-Ph0Y3y3fS_LD#mXZA&Z4-E(%LW1>WsB#tN+ zn?>5zZ0zSc@8^@9UuNQxKmm23+_U@88#`^WkT)cN1dvE1njh&C+MX>OA403&+|AT) ze$E10ZUn2SuPY34{hP4pyp_H!5FI4H*cRB&CcxB|X>1v#iu@OvaS5Va&X%RTsG*7( z9Y2ow@T$^apg$Ban%i|azkB|k7pnvT*JxmlYnbq)tIa--HcN$@b_6{Go#0+WtG-HQ_JY2Atgs;@U_^pd zYwQRSL5g2Nafn65a^oM*yzs&yyv#kQEAgWjOc0%no?f&x`5OsHmlKf15Wg?^M!5gW zizOxed!Nc-R*;+!e~c;k21MxZTKx@GYI8%y=59IuqHQsK!C8iMB6EWWpLnl|e{b+S zH_@fvG_*lX7J{EB-I-|@5FnF8K|=Hd3<{X2RLWASLBX|w`{AqhuBz7yIynOFo%-A4 zTmbg-k+)uui)F$pm#zjeKmU#UWAhsc>0P2Iw-~|4_ed4V#x(d!p#FpI2(}x5o@rVY zmWz+E{^^>QHH)HfYcq)l`Q8!6WC6^w)72#^w#!mem#8$-S5)fNnc_xqg ze)h^p!6ZCXmXOT7ZA2`@qGFnK8`1nhHBf}j*Gyk`$ZoDmpoW&gG>*fxr3pmak%Njj zL|d9cv;#MfY#|Z^>^Mjx+tMUrevmlHF#z;*n!eS?!7>!G=&&; z{EQGGP=XAz$_wC%bEv-4rzKzl1_(zoGLe!KaPK?%$#ceyJ`eMUL4Tu11N?axe*DTf zxcM&2LjU*ykqSPWeL(ab0sAsZHNOW4C^jxFd!j1{{Cx=C2uGTk9V66TYMaXx7Q6T)Me4Pt#0XIKzh#>Jlo-K00 zEpUVK3PBc1)@wYjU9PUpmq2^0eS7>&XtA>!?+m`Y{E9>R;7izhNndENMhau&%zu=FgDs!pok@g&je8&7>0@^;(?~vubF4KJ#g^d zZyj}2Ga#>W)IF8`ZV<7k@Nz@&C3O~DTybGM91$?WRXtKQ-}2gR40x?~uir zq7hn2)ce(vR~QN&;;g9Li#h$$IYTo3-p<3 zrjS?uO&qQ!wp6-gIY%dlxYyJ~u7pFy(Rkz`p+>wc1yhI+(JnFEn}Ia(rt z5D^*?z|L+%_-4HuAWSf#gk^?&5pGTdT=+|9Jf&U{_LO2-P-8L)ipVwAZ&p<)WGWZ_ zc_?mvZg^qh$Rlg~Ee<%b@IbzlG1l>_sc?sLPRnzV4z6tEJ@A5^L0I3cluAavV1|*1 zK$44Y#JVEt+ajnxII=+W(j3Yz%g=tU5x$tNSa3LLiDNK!`3sQ^VGb5}aoHeI&3S2# z&9D22j#f`*xIhl`ii37{^Pa66+I+l3;FE*qH6-86=W_W;!F3MeFSuCce&CXc`&(Jc zqx5I)GeH&X9eP`Xt4c1;yWhxP@*X)#0tsfMo8Ze8z%D7fN#kqqfmIlMncJoez7jcC zE$CKg2ZFn=vK6xO(r1!4{WYLKOb9e!L?7%<(bB}}1~*#J8wi_g9H8@;7@t5T1%jm* zk9}UPaCQPW4O-CDsCF#*mM{8=*aBm|#?S$Rt1SU-YuQZGBcc4<8pF{9>xZ%f@GCw` zN9G{CecpRJl4#`UqXZtawyrT;WD1wvy^$^cOtW45P3G?SD^J|ZP|4uyofDN6lo%u6=RwFP>V}=wvIn?Qq2Yt0zD*)6N9VyTZ(MRuKh*E zQ~svj$K70hqS5*1p(h(ytq@9vk$g+qu@Kz$eEDjPo$3j;;|lICr_+y}Mi``JHEodm#bL3@FOr4wAHNWd7=2WRMJ#mdMk@sTpQ?Xk5 z*~jHZZ(x|YDgo*u>iWiHX8Forx^e5PZ3htLWvhQu)5*cUJw^5K#!?*C)ig3I)nVndd)8@9WIuq>(?7M3gy znRZ$LwF2|=IqeV-i+4_)P`@C{cG=DjRRg*AbekDBlUpf5!AQf>3bMj@k#W`v#*`$1 zQbcct8zI4^L-t{i_k$KWtwi}MdOJv6e3o7mc~EEp(@ML#{)fH^Y1~a1iPnAg0c$Z@ z9feeGkOGJgW5Ul-_kiW1d$h*A(h5pJ>u9Gjr4I1C%nc}gaYtrFzF*`B8ZhbnTx6ve zCEH79;=QwF6=b_XQWE~E@8P5dF*od)^HVw2DOhf=p0J|-9I(wl)ymGOE=L19osGJq z!~&aF72IraXbuAoX`SKKuVJx9!NCIgto9-RGG_H=DOc$`&6$Gb#%VQYM$U3I2+EwB zgi3=W{e-W%v2kovfq|{XevpTQdI2S82Bh7h)Zbjeb}FVbrXab?iOlLQ8@CB9Z*2*M zFktbCF-o4DVEDwaNm-#7ZZuX-%m{_B?Nf@LxBOUBSQ)=y>Hxsf4}n+d;+afd($Yx| zM%QR!Q*4(8wolTMw9!!yJ|lk}#li|yyN=87HXzH5nQP^HS5h*X(K(j-X$QmT&1%2OXtXQE4xIb+e*mC1gC1)z_bM;~_L^Y#|!Utqy;*V^F@w5MY7GivU zW~T-2tACD)@9Srx5xTM%Vm8FgTS^QT3{z97hzt|uG!U{u3Yl=t4WiXXCq%0t%*9ZQ zMB(|Xl^H6+$rjT1h`MaWZj`qQdf})=Vq$fn4O3bzGs)P7EMRtbjkv-0c)=V8LK`{C z#s|*x8IdO&Yx5bD)Q6qT!^)B}P~nbiVhJ=U5ckU2ri=?J#1&5{`6$ZN0*cmEyw6Kv zy}SDGUUE%Dq6$&oyW;9X3e}Y56-|f5W7!oX3DDXXd*KxAkL4G-yOb;wv^-lCrW}%{ z=xUcKIlVQasV6Bh+&S|7Ipv^_O~L|{+dpXZ4^0r4uu9`6wN`^OZGk4YguZjzLFjUQ3*346n~qHB#5 zh6ERe{1M@)@^y7z${;ei^ zh|G{2CN4jbbrftN|It;Rx4BA4t}qV!aNbWTAG=hE*WT|VzQC+Ynq6gigaaoy8AH+6 z#qIf$LJsIO%ADIo5SVIZwlq5wM=Z{uK$`(Ipor~H9!tl+y zHD1tC5m(`v@tLLoa6JPywuJ&pX;G4aK*K+ytxrrEuCVee_@ag4BIv>`#IMj=ae0LY zb9kmr5@u^LO%`z*maAuBTbi%5nP@KTarqtjTo7MGx_16BbUAEX_#4{0kwH+3F(Bwl zhd=!Z*{Y?DZF1A0^_B+e@h!}~+3bHzoKQZyMab=sUf@>lroG#Hfzh&#wE9EUkI7N> z)B$HZv<*u8L(l&jJ@YByAIkpyl^q2}{-N^URplzQ^bbw{u9^;mbN^8Gzd+f4DEW6) za>W7Se<=BPRdNtw|A(%BS6$m{2Y;yfcU7~!xbcUYe^)gJS6}|n_3xu=@$u-h#ze6VVVpq3uoERTt-;vDiLr zjYAhQP%fmLPFm#j)doS)7aRzBOU_YbZ(slF&OsUwIH8cJFvM0F{uCOw^ZeP45%=Go zaU|YzI%>%>N|5=>`26EgGIr+a#m6gx1UAw>7ipX4Lj2}7E|j~Av|)=OTOl9NjwSk^ z?&P!r(3Od92s*3lhT^L>9o$l<)`7%0!lH{$tz^{wj#fCiG1Lvxho`+G5>lg}5AM4r zjY?R$v+K&!K}3n@NV?P&Qb|XeHQ^e8ZB39wY+VzG7c(bv-fzO%nm~=y>Lv&VK_keL zpfdKrj1?a>=>dgOBWe;Aex?TOf5peO^KwDu&;H9(YUXPUb=}`oe#bAlNh%b2g%NRR zl)r?oQ@9cnPr$#Cz)*^xkIpx1ycHr~oWS9@yeaTlH`1pNC4N1LgEu%wnS5Mq9>hjA zL1e&~?@$$nPmGPItKhprsVSTlY;jvv5(%V^|96=>f%!AxT?+ce(;E)(hN2I1+=ubZT9 zIfUk%@&!diE-1D}zM5Q*3^BpIZ=`!cdc#tw&h;Xzf9XrPEtl;(efCWeY<}~T15EE= zQqW!7kO-%3FA;%uYzUIjja7XFm06lcQ0?pM2spb)j^MH@-3YFIxoUyqo~K7hxJ@QsLfSd$?@oIcFfRiRSmf4Vy2&zL^U)sv8L zlAB>A(PSZfB{Na@aiFR=yKzXNMNJMG7sHZO5RtrM_M^6p%lAZ2NajC!$$y%ZNFzWTC&`1 zjy`gD(J`qnl)U?JbK?*gkajtgnAbwvkEe44&o{nuY2&NAtko!Cc$tJ7We^1y z4wdpOYpR*#6fjjbw*Vk1!u}l{_qCXQE|9CCS%968Z%qOxebokTx~i5(mwqGg^9)~JR zqV&(5>`%yvDaT_$cac#!Sl2&whDOZ-+?_7kVYbesAaC`4ndy1 zzp0dkp|1Z?U7Uy*P;o16as(?Ctnpe`Jr4rUV-^@(DKnsm6q|L`ZcdP2zFij^q^4R= z8uxW^9YylK2Axi`#yCXx$m*aD?@UB>W1czE4|tor6@V`KhKy(R*#d9uu8Ijv8s}OJ z%a!1Zn&PYT#chQv^w;Kiqt_`VI6tH+-@|XqRW&(V-~|M?91666n>om}mo8(*=^{3Y z=^|W=vArk_pe;h)%PW=%{v^H&T$WQWL;+drbItBcTMn#R$!7i+Xdt#-k*EJ-?LuKjCNO{|r7)X3M*Bh1`H&jt-BEO@w{LqTJva0Ehwr zsE4Y9eP>v@vpcz-xWeb7LwZY(C$V<8BcjH@wWTHB)ejnrWF7WJ)G9$B#!J$jgoUO_LJu z8^UI>K3&bPaoWPpozY?W315*L+9=2HF_%N|UXatRDY>0I99J7;rid``9a%J@kH}{^ zc|f*Uk|WuoTukbRUnqk-#)Y_uEIajZe0RX9C89!*JX=hV?z8pWY9OfC4a zR8wNc3GVx03RJs;XzZN)W&vD_vJM__XEe!mvz*dRY#I}BQy94fB8@$v z^qA%9h`MJQ0)C*`X9&-5X`T63zLOGSRdwg*;1C3Lvuu5H@Ne(&q7#zLuEU(v7X3z6xw1ZDdxOqR35cnaYYf)LL7nPuLIARGaopX)Qr+FCB;D6S9Njah z1h(DTrXPc9O~q1hbA}~IYzkc<=R)`yk5XGR{({BXTnE2_Km$u}ah2A{@UqUdqF^74 zA|>UmE>mhwm@BC@{ww%=Raf_{M!B(6nuFu6KPtA|M2#sRqRc>M%CK~z{)O#7mYffn z{Efj!lHz}ExU>V1yly3@L~#@}Q>8v0hg~hjKnWekbfU(j8tm-q&TaS3MnKovQ0uy! zyttVa^c=~>{{CybP8~_E$N>n!;-(J=^T6x?7BurZJDy~*L=}(|5qagX7eR)2xa`^y zqFQ4~Ft{FcI@eYSk_mLXz+y=@*PT>9_<1j@jiZO^^FJ5={Nz_DSLG zc~U*3CpYZ}-1)z)9?I2YC)=xP0Pwh+BXc3@-Ggj*gHsEVwB3+VBT83nl_A1FM;Qd( z-DDAiu08|sSKkjb0~(>1jx<6)GoUf9yUhT=hnfM+p|2UxI9=m=3E+NaKvNxe4I}_} zH7RlKYX(Tj?mk5nJIw%*>LL9QGXRqP%>eQ5X9k#TcgLk#_cjB7b#rMyGoUegn1KMi zyUC()-PLCR{$_vA)u7jEKqGW(Gvd(C3}_ZT%mBcLngPwBuNlxdon`>Q{mg)-+BuK_ z+;@lrw5#=(BxHA=BCt*~K%~0N0APlg0f6^61H`|d8DO&A9hYj|+YA8K4WfQ#Kx6bU z0|9tJu5I9K7HNOLc_#Nk^j24Iugf?wAfHP7gw=dG< z8H^Dv=nmToFRJya!^puqt$+|Iw6u{UaZjV`+@UwLDZp{ z(*O>KMP0UWA?e4zt+G7AZ&as5S@#bxx!9j3sE0NI#J?ic2)};h9Yw51Uqpe{i`mC+K=d0;uIo%+rDD4&C zZtgVPA_|Qh2S=yQH^e-*Qx3^?4>U(vhD29~4TR}8iX4aj{jJlEIhqAtwpx^S)zC}T zEJmw6bJSUmEGYF6p3D(}-KUKSzDvKzg}1)@@>Q} ziv{Vu3J4Io`bPkqsgIkG0qj;>?Gpp-=klunX4-cZpqaLeXyVw%&|$DKdzdqvB(%6Y zXwZ8|?zI{Hgpf=#+rOJc`+>I9JRsR=oP3qqTM(-Nfc9MlD5`x{0Uo^fD!`-kGpZzg z+hk+Qd#wVn6X+VI7Nqwu4G_9IVgQ`!hHjVwxbH9pXnvS7)4s!0Gi~Xtw8QH;OqqSJ zVd^2fhN*|_9Ht(WiM9<>586_53{wx+IZQoh=P(6m-(jk#_8F!gy!SBmC_RTM5L$<+ zDeoMn)UgiDRfhb+!HXBxs2!1PMK^74p$5@5yA@Z#4p9K!Eee2)=O8HxEt=psIx|s8 z-|_F?R>5!Vu-xyAJ_!xW96a@f*sQDSaCV!qOuR7wBp~oWCxf>^_9WPo+K1{h!eLwx z4-ATgV#xt5i-w%^T_EVAZ6dcgdq_)NYiVwP)}6M8IN;TJn=sIHhq!>~keIszYGRNi0YB1YDnw`V$Yh&+zn?kJd7E`pUB+ zADLj&8id=YAXw9)gYR&;m*(fw4U%1$2{4619xhjbHb7jQuTNLmfRDt`5MLH(p~=Q8^j8{WQ`Rv}&mjfnIv+Lui;L`%wH=I!caluV$AE z!gak4gskj#C%kpLC<0Gw7e(kBwTq(oq;^qswpBAWmb~-YTFr{U)6}d8eIqq1iceCr zqGQyIhga`j&?XYoE3TMM$zh9Z2mbTzyt*#tR21bqVP%&pIbY59ym6b-BiE~m`ZcaL zs|hX*s`9t5)h{W3`&vWLA?Z$lh5|zT8hsj_C8%x_LKQFYswekW^$l9}^@$h~0~$sCiiX!Lj~)Jb zjmuQi+2>z=5vufaG5@Fiv97VhF6cHdEy|3r1cL+^GrD43%zwc~T!A0YKHFcEpa>1b z=PlGwrfCV?6p*vUNnwEzFBgBqR;Td|??PXe4~@?${vRsEhA((F&}hZyxGr!bq>ev4 z6!)bEF~TAbR^yxQ7*tn0t(d4f7#ed28^yUNU>#`5j$5uy5W@wskbe>jVBG0w7K>NH zA(8arFL9p?oga^m8?+%uL*VibmlDP^Y(^>e#!;E`_fLGS(0?_`(%s;8RBk5o*)#?< zVhEj-FmPpf!Uh_RXOHQfB&LX7DviHg;;vNu?YyZoJ=LfVNjXv)=(sNmivBaz82KOy z|K$s12tp?%*O^k75wk|Sd06_zkDJC)s?@L^+K_NJ`B^}2U5Zi6^cO>kvcu*ng{4%d zSSO}sB9m#3cu8P923Yyj}`iW{sjY|EzfmP&nCfOXE=XP#_8-HDJa%p zQ3NK$eT7ffOXI%!Pmup0YEekz%N+_OsdY@3oApQC5m;PuEyFkjIc*$M>A2d7c=1UkQl{;#oT|vyO;d_Y=FW=o?|EuOk!mAFpA~cTWE=h3Hk}F z$bMAI>)ASJU4{&X>O=7ritvn=`Uin?3Oo{;a<4@Sqtr&Yibv8Yqh{)?J3VIK8FExc zKr21B5f;F_B<6#fv{C%a&+4~Q(fg2D_WenQ7~fz^s=06RyBEv zVRJDQL|R zy36MC8D83*_|#sWi!rg72&G3^Cx=DB?dsX&>;?nLhA%r!SX6qeI!2>#Mau1(`c84i zjRJIFn&oljb@rg#n^p`{{I2}kz$8`|*hi%cu#Zg_;HbS(u(QivE~Mlt9%^_q`QI+6 zr4Ja7RrB6oT`p}+A|7Y8)oZM(e^RD~;4n2u!v{qErI=LLxcu5B^MOw}ktI5v_N6xE z#1?&_o1V;^2m%?0PdaF5dvaIS<0bmc&1}L==a>7jI#tUiwA1Eha?6i_cpc)m`$=21 z)Fv&hMX<{MXDJjKBOS9-VED}_KE>POX$#OUXSye~3An6g$ox@dyLKlo*@>;N^J7id zG3z$8!8vek^y?CRC=KW&{mEfgaMugma2Z<1u9od1&bQNe))u#QmmWyjS||tT6I)wR z`WCI-Ah&&MD^gpxwxaRG)>f3B(s0V9`oxA)l)gp7HOOt>aEjE{4X0>4vEdY@CpDaj z3VgbKwef|Z^i3Mh3nt{A&~TpA_6_H0JhkCGrEJ4dWFN1co{!QJD}Xj}6j&c#o$aRR zwKj@^YnTHpL?qY}+Y9W`gX~#x8J_pyomI3^%7DS$^JD39mc>8|xFyW`G%oUNDBl*I zq4@%@La;E4G0aZBou2a;w1cfm0AkA*( z=!%*!5yEVM9%hRT?g^sL;Pz=X{iTP@I z&TUaMoZ%bEk(sPNAX+e2Y9v>-D#M}hd*BSjgLA#|txiO}bxG|pFh%4f1Y?wv+a}Xp zOch#4*yd3JE@@CM@VHpFlo%TwtN}5;!}~*c?1pkQ=Ad0o_Fj6IPOMB1^yZaw(qw`5 zU%vkSJBQ-DJ@sAjW^do-)EB)W#_N_MZB~?})rx2UcXp?<+p>l$?2=4cw$vOJ9L^R6 z5>a`nT!`&qDcLg3Ie=-g1{XKA7)l`=yrK2xAxar)Xu!lo(%}p0`gi3e^PpVdeOPPD#J`ZoY04N0kPDpgt|?(dQeqsq*zy$p z!w(k*Z39Ou^g2G6e$LZ;3J3ML80zq%%vGSZH3^HDIsm|xLLB((`8z;rOiBQdkD-kzJHIS#5vzBpug4V!-=smKxb5f46HDo~M%2%v7q zcg0BpB4eIHSe_*DK!@;o3PEv_0D3_lT@@N+^GaOmVM6=1tBcxnZ_}2672nxi7q_?N z)UZX=SVQ#<5W_tr1Pk-_wV{$A5)>y1P>4&?G|OANEk;EEvi9`a8QP>tnzNls775{G z5q5oJ?(g|3H$alru4Y#E39YHZBr@sD8E=s%8WwmJte#CvOecuC!eML~v_9e6nWkCN zKC_5&(%7cSzEsDvTiRKLw57BU^;3kLwUD6s4nc!h^+z1IFICgW#OF41PiQ%VE~?5* zBvjZj=aczZwsc}5k(%Dj3Rzidm-IN)YA}xNo29vTCY?1fCd4Gp=RVSSUNHv7=KErU?eK;7 zQ%Ia6!nDF0Vgu?RS{~7x%7lf5*kIx5)zd9F`>s!&MiY3dff`%`G8S8pdy@WeE0f62 z#r7Q`hnv|PS93T5fQ;Jry=&<(ailXG>0mk~c6)EPMui7{n@SGTsjH-u7Yh!olI6Yj z0p>g=wEz04gm6V|j@(geoDR#$;9zaU6iiBp!F138o?4k3dR+YI4t=qAquNyyjk1Wn zS&!jneq98SAM&L&=LRrv?hacK1TYBV&+z8oh$*)$Z0vV3^8Eqs1>?WsGN`X@@cT6jCP#$SVSD-Z~`J&m#mj;hJoUUF~sB1_Jj(1e^J+kObT)rg3 z7$m~PA@TQ-550F{BrzbKd~aZ&DoG6Yy@|1JXodA4HKlpzfLZ6pcF6}gj5!gCEEU2H zQTP&&mx#9(bt7*^D1+$oB{-64xWfE0j7&qSRy1!)6}$$ik# z*f~-Rw*u<_K>n zmPp>f618X9!hLYL!7~iUG?f}rxyrJ3NGm}IXAfvZe)XvtuiP8)Rz zqn;l5+Ms_K&U{xNXUaFYLK9fbTtzGz4c^YYfr?F!HnXJRpW{fZLGobO92JWS;Sxo7 zsV7~LSuL*?bc|uIrfc4EbPQoz1ITeO{YNoD*b2fiU9eDE6wjkah3WB1;erNe@2DQM z<*lTk2YHOn*7Be<+SR4qRt`B-m@fZGe2r43WnUGxOrby_Zu92BL3(N4RLxg=4`QcS=N%= z8mQlIj|#5+rh!mcaO{KDXtQu#oUR?h<;2~F2X5F>whtfI(6B5PlCrC?H#+zbK9a&P z{K?J8{0Uia7aVwq&9NCI{2duxuu&sV19>*0bV1|_c!fzmO4F0a$FsvX31aWlCOOugG7(K>|G8RQS+(aiF?GHiht(KXiARqp)M}QV*${@ z7?kl?e}shNvA6>f)Q`Y2x+_ajgtm8pS5S+F;u3BF67b>nX78`(mzvIOF|p6bE9y8w z(ocgUq?UHH;ixhneGKlPC}z0C48H~)7nj$FVkRL}6mx@yD#Hm&j(!}AO;p49XY)n5 zE*e7Wj?*YiMKcpkud6tkIV=r>mBTXb(oFz-X^NW(aVCOsh%}z1~{nw`1 z_hpyIA8jCYIuhMjB_FnNSAL}_1F;@gl{jWoW#p$#eYaM{OF+)82Rs5uxeI8eq$Fwr z6X-}H{{ZJMTX|{-7&u!TtTBMD;kUa&L@`eMQ^C;jfoC{-dVf-xuL`x%^X3avRzS*3D>9T+(k{k$8MBO7?UOI^Rn_>lz=GCG^2l&o> zyQVWDNQ(HXWYNKDRXn;9x3DR$XE4sBZE8ovV#v)0Lv&?s_#9EF!)9rDMW4+=&C$eU zG})|H*l^P9f_U${tJSRhP%OQjM6|l`rjh`}{@KV)6nHTRtEm5Y{r}@Jl8Qbrn}aEKs3m z|35{4i7PrP+M=$1P7E8LWJ)1V?=F*h1&%iByYks|ijv zn-UYlk(Ks8Yb+||>RoE80^g=3lcs8CciVCE9Em;2tMa5q{KJ~?%(I;};SRsXB=o;V z)1$h?wThoC38^gOfu9b{*k^VHaf$KKeU37o-XI_8Xd;4o&;%u4*(Eh#X zwzf;l43k@pxR%N(!_lm5sC!a`a3b+KPU+!^BvbsiH9PHseoIcjezz#sFYv#aIKI?R z93j^w4h2@~8+N|(O;?^-Dm8fq-oAxvMo2Bqp!2#_pLdPWk9XzR? zpQV7vsqo{7?(zHwR_^ilv$_F#R==Mu?&&td{1Y7|)$1nG>$61yPYX_>j2hJ&eSW)X z3I~o=wY)AS_f1ZG#+DMo?2e*(qWi-Uk`hl!8d4jCn6bG@R_Q9qH=C+5V(btSnj!z% zGUmXoVfH)ku)CY9H5G)C`zoHvtp53}xu4=c{u3>V&r%1h3`eqW^#I%oRy19A`9^s) zBVe<8=akF7XlIuwsRVdYIrDkIV7SA{BPUg$4tKSZ?6~5H_S6_UbYA+x1}o!~^bitX zhw}hjE{EGg)xeaT)zHH%9vCA8wWNlC+-_-7!FNr212cnr^P{gwljCLK2h7xO3xx(v zd^OMfqzqPOxwmCc@bicx;Fx|Xi~G^$*=A7}H|5Ao)x531HR)Am!5FUj9Z|1{KK*k` zsmyw&Yt2~}FaNana{9yG{{Gv&N%?m2eerhk!(Lgu{o&@-%kOVqO@Dax!&csG`T9?<%l$vS-GBANwSBt6DN5~f|0lGLKmR($StfGSrv?3QY$=}JGh(?_ zX>(ceujoYO!*Z<>z5EYrTH~TgHE7nY^e-HP|ALX%=5I(k!2X5RmnUu7d?IOQ!@p3( z%%J-prMWE~9{)lunzxHm!?58p`Im}CLB++24klVnf3&^&mx^foLT;n#je598vcCKe zI7{$O#4SZtZhBtTSii$l=a^sVWUE@-ep_M9d4*Z?Pv{`!n`_`(tCfyPPU=aqRHE7o z`KBTTZtL(_QE<|-b&LKkR72H&TZx4>&Q$*Utu3Bj8136m`@7U~FFDhE%pA9E*$KAU zCwRH+$sVFkzrpbGcFb-}yEd-vJ7^ddI4IQK5OL#y6A4jSs}=T`2}_C=T^*O$^v->0 zYO-=`Z7;k>6(2BeY`zFgW(xFI<%Lc0k@+Ttxh0d78}fH#tg4sutERZ~4(rLaR?PcW z``a>xJLGPZs)p}ePn}0SrIa3BH+?Fbt?Y#xUn$Z4`&Lq3Wqo_af1j*~H$$@65#@n1 ze{r@)`5U$$hORz)Sk7%&8UcEf{I_2PBxpT}Vk4?SK$9VfC*6;$FAMo)BdFw;$v9!nlOM0$F?TC|0>Tzq$HS$+w+dcmnyA`RAjL z^nxCfRe@^qAq+>;-{kL#L(AlE9kwT5YOT=6hYuw^Y{iUL6>|QXzw0l>5&^U*N?p&T zAwFHy^uGs$`xUn()+!4m$wl@(nrZBW(;F9x=UD9}IqZ945$q?7w9K9Eq>Pr&W-)}5 zKo8}0L31!vFuZu*W@Ivk&4QxB$}3#2vek8P!%AnDbB z%37goBws*CuRP1zv;7fqWHkKm8+1JkxW3{_dtOo4RZm`m!KuFbqImHrLBdcGNg)*V z*Owc?52+Br59`o`v84izJ|28H*}i!XhN4PqZHXo)OFjFAZ{!2~-A0wzB+mMal13dUjs5 z`pl|*g%rZHQl(&&0}ehg@dX1g=H&uD^Dz{kpPPjb{=1kWt}#~ufm+fU#tPI2_9w(k zm-C(&^dw`semktP3Z;K-xgLZ`ygKS}&EXu#kSz}l^Ywt1RlHJ^RK6!0a!}z-&+e8I zExi^k^{?{IuE}CQ6w&lLLQ7{LBwHr64n*tQTLZEBA#wBd5lGtZ{cOFSBkk%GFU0n1 zz69RIX5Fy_6&)l{J8W_m&#-JUv>Mg-&|#u_ya950q`pG#?0!L1E@eWX=funaM7#|J z<9?8BdB;F(qd=YE=~P^-U4L>HWDzgV4m)HA1T8nD-gG0s1OBwc8^7ITaj%JH2~+du zxUS3*XdT28iqvE}u+5#^_KAXD*9{sMu7$@g`+=MY*CmMpFBlvVE#4LB_&dAJ(T3-% zdydpr?9M@##_nYz@~FMJy74!>x=qw=xDKkQV%f{abac@F@8+SSk0uD5(Mt0N>??4k*S&kxohFcufKhe7Lr8K;l5RM^B&wlOS4)az);w9bd z#dA;iVbsAjSMkBXytabKBQ%5q5O&K!fUbaS4+o?VQtBuw0BR6Z{&NczaxXGXkp zM%Khwj)Pia;YSmCmH+_bNd-42Nq~qSJTD@RNA85E1_!lQL;)?#7B>}~`HRXN9utYb zAN?K4PX3NVmXZ&Y>01bUg{a7CMK9nb%J9YsW+D;hwvuEz7j)7?u(KudXw>AzrC$?W zu6m67J}y_E(QXbMS3Jwr!T9LxjPXfQ9BZ3;gd`QJ9uLG8Xg8}8jxwo#i1)6l*Jl^< z4N2(ipH@;Sjww2UV{kO`&1Ny#V|ug#K=Anug3S3hcfOnFqdTw^J@NhfwlYO6!lK9{ z*suUuVyl`&A2#*+Q(gX8kq={zyR!z%{=_;Hk7kssA1h^eL?D+y8)p}9a zV?IL};goMNEmqT`6-D6B3U0<({UZ+ez7#SzMWr<#V4haUgi91=_4#IfO`Z$>34(?~ zwQw}Cj$Ci*19s&y)NOTKUTl#(YLyUR9~tUUju>>_WgJz9sB@8 z{XO3PWZI?_v1ecDyWwYl(xI%Tg@z)i=TO=&$XGCex-G_WuH-CNU@JxcVx+9;_?H?4 zP6kt5Be0dqex7cMxvcC%a79Qpz|L;{SUXfu#%>F1j?$VPTXchubbAWHp{BcJG0p1c zwGlT%-`mAk9e}b}k`+OoskfbZ7?re3bQ8&ji2xd4aj=@eWS5p~%wyG*Sq1Sbb6Ma< z(veIF4KY4kWQ&BNY8FI$8(WZFhBwaJotGc2zgk-_ZGJby2+wbl7r@ z-R@$Fc}P}R_>AeVEFQwo+7gssCs@iO)Pcw_FOj0Z0@!5%LvXODu?jKRM2$TASCxDo zPHqfrgG^GCGXJ{9n=s|{@(4*2wj?^A3*7-NWb4R60e`c`v(9Mf%p)ogxEUf6 z&@coZhgZS(ux%TPE8+TSe+lAl8ny*BlI~bYO>w>`T@OX(umv6lqg4{!5Ef757)-`sEa(00PR~WFYqa_yXv}dsw zSVufM%{E1MV~@?pd3w`8$z&6pgQ0RA2;MvbE8a$m8GX}NGhCZOq;UcZl z3&pCu2e4h6^Z||A6efU@gi70ULtw4z@A8lSf-FkZDsE$2;=%V7T(}j2>li+u5R>28 zy+I2<$M(N6QzgL7dC(ZoZcpw^PlaYqkC6e@!BGGof8rdbq&V_fVfP02hdg>7VoLTy zQLw?pq#dlk9Z3{0c)D}~KMMjw{Du;ajyVh|$@fb!6V#BoLE(;JTf+#Ui)@=lMbWx& zDMzr_Cn0!K7IT!8m0Cf%a&cWBR}Z)t?yRr(Hn@xEBZm{WiEJEcYcrb=QLYL?r9Ul~ z+Ez5M)^1}i6be?Ho1!;>uQBtb^PC^jO1Kx(AbiV?Kl8)W+GqV-Z-nHloId-;^TXjm_fcck1bONknz$t3uUN-fHwNbUrUqqE5!?cz?0huK^<#`t_y88_P51<5i}%z#MyNj$!L z4TsZW>SQIBA4{CXVQjlB0xSrB#edX3)PO3w^$cOnNG(_2!#PKZV!_B9>Pequ^L%u6 ze90yBN5ERPqA?fws@TkyCvaBnM4zvhU~MPtq5RA9^dvlRfVm+Y9iSRnI-l5K>YiS)+FKP?`FK&Y6^yXBU?x9x9T4 zsB)?gc!N6I;llEGiy#SJlY;%U=zo{#3~R^6HMv z3ImxFVLpz>rlRod!!%Fs(9_u;!QKP1PFS=@MQ(qZK}C6UlbaOC`|g7h~d2S+EGTv!}T@3Q%V zx)GmQQu`hkYrE~2GHN9%G`f_JaKMU4IIEQ@&3)kuH3y4F+Gj9MO{3$>DBeQ~1h128+#Q#bICr?im2lRdB!LHF6tKUiIFb!X)+VG?Drk`BEEyKja4d)5QbNssyE0{< zV0YjUzQWO?C}6XMVOsN=7RL|r&4Jpw({j$ExMH_uipzg0%Ow(Df0oX2jifm8TJnMIp`huYNVFl zMn70GQX~pZj4S4pa(eYs0ZklQ|!>wfZ-%*>`76mT4HC{pZtbZGv~vp zlV232tBoWiU(NVq<@!3JAg(q(dGirK(v1&e|FnFeR)C%y7_nL`XVMG)Zcz+Mm1cr9 zhx#tY&lzl0q5@ph*VG<{1ZR+rZ5!>#UYSnhcVu>+f=_`i9kQ=FU=!eYi1t13 zhp3DAs7zEC8#xnr;bIf%tJ%?opw5t?(~=C_L?hFjq14A%Dp!;cOP3|(asZ)V9EA8$HZN>)!j78o-^3mVS|TK zR?G;Tzg;knl6!*#&QZ2pLN)>m>EVURgEnBay z>nfXs&Jd0^gLye4z%)4)A=z!6a2h0bO46=`t_I0kA%FDdVOCM>PDsJ;L+pu6aPf%P zKm-?29j+`v8W(QwjfT zhP4rpatDV*^#F8fiYnvENki15_jokqY=NMZpdn1{!j?22FDOTXFx zsnUQKi>aRlCN2R4#+QCoxlxX{T;lSN@NXj=eO?})jHOaqf0zNo!snxm6b-nfEDflH zL_51$C$<|Ir94dE*o&3w3zRUs#a$&@AE7tf8W6BA3WrD%J02(sXoB$m%|mJKGp7su zlhDK&noSF$n=llyq1&)nRtt)^ubg_XImF>B15}EsS01yx#^D^SJsyOAU;1ngit6 z+Zi;_)Q!KCRD39DC=_kgvoal>n(3uQDyiy^I8zNCP*3r9pd`q}detVv(J8+p3%YXR z(JFxjo%_~ySWeKbou#m+aQk?iwKrF8Nw%{8jnx<-lJRehh z?FP#WzL?xX6W#cV?o^6Q?6swItv7gzWps+$K&@M zf=1B&*|2V~p~Gct+#aVc)WL}nPO$de0G8C*y9KBNDShI=(oHL=!EGBZMC&maff(>X zN2jQa7IhQ*zefh*@3?%0NFsJ?_ea$RV;BXJ5mhyoY%5$bn8Q%@fTh-ClRk`u8MVWZ znx=3Id}SJkVN*Ysp*0w<*ak0Ua))u&CU_X8ZsD{^9zxHKN^cI;SgQw8QmCFD4TOPo zBI`F0yJ-TqA>E=nwD) zjACG%TWu^7AP64<4LTG#KDlhjZgzOMX?6GFdUgw~K@Yeq&)b*=^tw72f7}DPphO(5 zC@{pdfVdTO@Fw~F0-oT;u1cbJ`6i#HA{@?|Txm=+zhYfCLz;ufW(i*j=R8IY@)w(H zY}es>u~_9~dN=y({+l;@KiE&)?^zhYeWIroJVAIdt0_WFUKI6WZ-mEWrPiZ^u|%st_aX1q#g!^MM94^2WH*|1^9; z7N)6m5{Pp_AxODGg!tjc16>!c$xH@_QCVhF$X0jk$pj=Q@ldXee8j3OM_w$iX>42^ zXHK%+3|O_mdD5?kqr4i9vOgSUFF?WlHyq;R+Dv^nE`DC!H+>l+I-8&qmt1+IGh8y( z9W4+Ei383M`i=RMj8)jBP0)z506BK5wR?k@V@PcvxVIy9-~RXedmr!=*)>iWuIYY* zze=mYrxo4BzQSc1N)A@OFRn}6tkV#DtM>#@luD__#vK;034oKIwqm*Roq>oH&Fibk zDg_K9cw=kYK+tvr2;Sm!GiQR*yc1R@0?dv{S%1xORYfaZu%&EbuRv>nVh(5I^zm$g z%$c(kuCoc(8|2essIXll0To!HHnOg|s!ri-ylb+A$4U;SX*_+0PJ)TG9$oqTby{@k zP+yUK2N#KT!5V1{k@AOb@k{bSUC3B=$hOu*RBb^64uPC!izN~$Y92u`&KSwX8US$D zSKk`#YRj!>@kuuO_cYm|QY(7E}t-VekWe z#K1OD9)T=j(O=ImiL_j{G>Q=j%`vBSf;5a%S-%av$sM-A#r{jYUW%(w1#;8ThL?^6 zrvs4RVV=eR@4xhOoutzr;m&UOU`8fGd@PPvP@sjDnqa9bK?qpi6!i#cKxo-=j-~P( zmrVO3;?8joB&RnoDBiGi&BgsMon$ay(q9>J5Y^TF`LAtqxTeK1x0+5_7}EpWt6@ig>rG_+TZC!A<9d}bo0bGpB;&?^$>$qSr9{@}j&4jnl@RjY z&7Vx9{uV^eDZRr}tHU`Wh7kEe7cS^5{{r2VGN`p6HwT+BF|zn*!Hl~C(g%Vq1@jH; zg{fXe+V5d18Zy)r9*2FeIc(@5KCM%>UjX=75Fs>tTnmqR3)iN)JH}uWfxNCb02^#b zOuKg2?S@8EtwTN1wvO@ql&cqX(-#xed;BBMVEFa0c!w+uL8yIISa>q$mqXi>cy~X& z8DalQ$&owZu*e3ux#iQ+GKT09yJWkvcDlb?-j~7IWV0k(HBWevOI_>7qYrsj6Fc+Fn~WhE(8!ILAv(SL>Er`e z;=&r)5l_D6HHedY1HQn04@;8I#PBeLJu5>vG^e^ra>OXxEXR}%+Z{}W^&D?tg#2M* zShzs?r*}VQi=R%M3Jops5}^|Ymy+Nn*-X~Y{_|Oq7@TOIOIyTE!#=k^+fSh;;IqCn zCCn@Xwm{I($~=g%_x9O;B?m3egmOh%ABy=Y;%(s@ zOm9mp&83{%$sL?_P_n>%;jm+NX0WYFNclstu@3fc`)`tuR#|Zke2S!LSHxJnA`7&S z=aIU=FfjUqS%;rL zoJ@-Q)4jJZUv^f92F0iw3@LaJw~d6=-c!}pj~q(ZE`xi#gBAxvU3c(k_cBX3q<`Th$+}bCn+Jq5&CPM zDuaP+2gGGL-Au}Jx{a9Sh{feec*)F7vQ8g#Sc!XGcmM>wg#LJ-D;2dni)ms2s953x zF$PdiGgs1y<4LKW%GO1?7CBPydW<2AKH zodNd>e~-=pvU%C4{g8^18;w@U6xda^v|hB2&GsGJ5e9O0M`%qAWqMi*#?*Dkfl}^%o&Qo5j*GQ(AFSjGR8%PFs2c=drkR^{kTfiZP_sp)Y z@-QqcShRLNQ4G=PvdVUpclb8h^4n295Y~=z*Yju>W$#XAHk-~2((t9-p^Tp=jpFq@ zymV=QhZ}kN+T4i;U)?CCuLpRjPLd~w$Ldr@kJzoFNDKd64yMgOr_X47T*!J?;0h)g zUUx=pI|ISg2s$XI_SJ5MO!F}84222sBx-{bOhyC+IWKPuJTe->$2-XvmKkrPgh&lH zn>sbtVj;DzYHZE!=Fv!>!$&PR+kbOk3BO3gbN@Cs2n)bM4b zGjdBJ2-(lrRHX!7v;70066tf-OqnXOqF3EyXBTtV&Fm|?MGj655QpGB z|Lt(juip-akV!aiXX1EKP7~2+gS+;JWI)*a9x%n?y}JS-P+IyBrd{+9wkda59t&OI zL;-n!YOPJN*jW)C+z}*MzWFA-|?CbA+98$n!4;(gs|Dyc6R)4F1Zi8 z8sEBo2*KbnqXKz>7*vy28uqn8rVc!uXav>z1&Ft&<|)cENuy&@daCjW(Wsiwd$9X@ zk4)Vmb59wZCCC%bEw#J39gZuH~wGGL6xHh8)|2ZAb*z?B>Cu|(_ZY{Cl8)I(8beqASEg<4=x|EY(Bl&d( z0cRbWA$6tnMvF==vHMs}<-tjAr{;>zb`=V97Vz`c(b3q%vQx6-0tXdQ1{I=>!yBz{ zCXe@#@4c%-pkTVt6ESNygmF9$H#e{_^i(k(5YE&ahkAJK^oS3o#!DoB58#9e{PU|v z5?EM;AO%h_kdMjN2X3S!4ape!387SOhox&mY5#eNoUg0{NilGRZDt%A!&?Fu#UtNc z!PJL^YGQ_xYO@hbvAUUW!Znm_Y{n2Jf-P&i!<^A*ej(fJRD}?lPp~l-My@ zZE~wMHxIA&k!lhS99%oazH0o%pI>T09fOn@t`N2ha8eT^}}L{RqngYzbwP+~~Sk zP>$&&Db1JjI>pI{U_)dt|J2`rX8e=-)z?v2xpJ zm*qcX1IB6 zHATitB+&!YBGAJA)O3iDk60Nh*E^2p);3Sqh;fA$^vO-z zG+h^;%i&1a@1|O+$?9kWt1s0(XJ(S5=X$Jg*+*{Db-eIWTK-ub57dOS$#*iXl{ z8aHr{cwR^_3RkgYu#5`$s09_!&2on&Ki$$jTZ})tyeceGG)rBKf6{oa+tBRG^O9T@ zLE;(ApHZ()wJ4jVtVwCXAa}y?=HcOScrP+MxgpgUIbXfIyi8Wk+Ay|22HT7=9$k3e znEWrCqBu+8cxM+i zr5DLTx2VFi%8vG4*s)RLYoaH(iWj;4T8Po$J8CyPy|80Zw9UFhhn$zNl%eIOh0ch) z<@^9y*iK5->uP09zJ@`(HZSSvhK=kxhVKb*4*fHw$JcbtY8;V5lz5CbxHb)#hH7$e z>CDce8rra~nwA^42{@tJe`n{mJE$6MC%?~c$a=6H1!v=Z5MB+e*eG&Cn7)?RgunYN zJ**4iPs0TI(@=qcFj%B55Qd^eQ3G9CdOMVUe9WER(xRhF2Yxe@{=km+dNBRb-cXc* zUHjEg`U6M8;BFVVEV{??VEq7LXq8h`eoz|??!UczJS|cX6VSGoa#Nl=C)2Xiim_upw;+3OqdC8?tJJ!p zDZX30r9&|_Ht)N3hz$$PEcw{%*#Ar-~9l%4x7}DFj=q9JL=16Xt z5=^`pU#&N3Q}Y(016W3XI6L)1iL2QYX%uy=zGzWL=!+ILTyL~(YnNE_=XhrN0o+gL zH_nsV)~*e>3`I<{atxa%mIO(4jSw)y5P^Y#V+(wZw={9GyR+K~o)V-dQrz8oqs1Md zFJjbB`=UktG>?{1dGmOsL+L#bQflvw7VB>SV%%nfP@lxNIX4aA9X0(R)Tf9Y^_?Mn z8*_0Ra(CCu8yH2GfqKzuGa4bm5VF~Ejomr&hmhfn*G|DZPplkB2tn_yzt&q z7PwvW9dMXD2@kGU+;CFgOi$mC(6=I$AA4R;$-eGBst(oryv8~vtb1Yqb>!htyJhPI zX@jw=&!5W2ldlt;PTiIeKN*OX19kPu4lFuh-qd-zs^2$3n?o87Ti+>j2rbR5P%^Ts z{lhVQE_umQfEk){?$ArmY0wPa8xxO`7cvEr(rN(bx%KmWNyi+WJS_c_G{%kP8_Pr! z& z(jx>Avf^!nFte^q;zxe*7idb|NXp&`{P$01M<;SY8lh+J@z&P7<#xP;qxQMHDUcNV zZG4qlGJAo5fPU-|+UTS6eko>aT5V8fn`lzW0Cijgt0gdg~IteGYbk(JKcbsh^lqs|^6KQOD`14vEaevcOYL0XMZEQET>%Nwe1mC{K0JuFZ{#o*jd<)(UTe zr+ttj4)-a(gS~xmw_Yz_V1$`ha_Y%!HoB5${Ica|HNPdycP$ohV^ z950K6fYT~W^g%lGI(XR7#eTfPM|o1|mgrH1*r+U<%@TK1vHcUZp%>K==uxq#7W5#U zS1zC;AbIo-7+S)vGV@uq2wN3)uc=sqHJgH1yy3&Ao3qS2QIIQ3tW`_p1(^g>_0ThdeOH$b32 zJ4@nM4;IE zy+h6GmPif>zFlOut54p_?<95u-yMbFWyK};--j5h8}SagnSRwRWZ*= zi40MJkk3cgWV$wZ**zxP1TApI_ExF~Wx|Uj_e?_>cXqEIjSxus1)p0tU0Z=B8t*~xX+oMW0p0IWzP?{=ggeB z;xn_%XaRA}<+>4%;fN4=8hV-Z`wCr+Ji45yF;-FxNJo_k7~O4Y5<8XjR-w{jT1S-E zCP6HajHsCb)dgz0bgmv+Mjl!`fm$t?(H{-Lg5JR83<+J#W#!n1H`DB)$2+`Zlpa}u zmB2cb1xy)=SmP94ofL6Vot&R`RdIs&c}X?aXxD~rWpUcBTyBI9#R}Vv2Em{YJm98e z#$E<+)B(niSmbw*L;4WXznX^Sr%R<5;&bX4(+AC{I-_~9gN?~*8q$jTS88nX+{DsF z3^H;UXO7>Qha71x#Z2%KHnJ(PvvTsJR>Ux{;VFw20Ho|10f^~N7+_u4w3fa!O}Jq) z?P|ACiNI=FC~|DRn5-tR_GQLbzv~%gz50M&a>dVIH;OK*CBJ;oD4vxi@)MwIC=_~4 z$UUWfc5-r}3ndkS{vymJ&{KauqW8hE0*)yVd4Y~g=%xpsjHT)!WrmGge_dnaqm!lM zzKg1ft)1y`QHi1Dw14~0nLIJ(dmVrBd&FpfM5Sc-jbBA^N^)Q3yr-nt(qadU|#D_Ok^lDdsO*y#RD)cO~}0riDemAztZi ziH>a-99{j8DcLyE-8_XIcGT!dLrAE;2O~t2|ATXI^b=cA*ap3!ajCK3l1@gi1PK(ygEhDkL*GL`CS6G4dX0G&jh<)m`SUnz=7^HobpzzU% zu>WO!SWND}AOTp<4wzBn7?unko_@VyXV<8Z*3z;@?u!_WJQ)uCW`}eykZt@^D6B9y z9rPW)Q!{uru$l_? zRP!}ulcbZJhb3MeX*>q*Uu~sD5>|!%8c)_0wB-4Cdb9=3PkoU{?tUw)(2yZU8Vivj z;`ssWP=Yd>sim5I{1%Rn)SqK zOiVYlNMe6bzj4ZSvgWz(bg0zHPWEwm`+!rH-ZGns@b6>=dm~A&!(#%Q)ndr>Iju;a zgI7GHn4=&6;_!OQ8CiN+@PJZp0(zj_%vN~Oynj<3@gjAbmDQ~N-(}s`=!#AUGBX-A zSUoF6CzB@6aX&3E!XOLC7-u+l#2M&(5N2caVN1hX47e8uNE`AtE|z}yvK0Jx#JJt} zB+ExH?Y%3Tzfi?%<^k-bHA<3 zA>v@6L-b)%*EmP6OFloTt%pf z$-`XgS**pdyOwEUw0vD-E+BOJ=q{L38~Am3(;{Z@o@gjI9r_BbN8tOA|4)*w|_PgInzs2X>`+uy%`DCI~u1Q zla+3L6GkMuK}^i;`mUFZbrDx?C~nZ*^5@YBc z`+$3QxWv^y4p(2$Ur2CMI4C4`&QMO2z|~k zdP7AyX_*F_o?*t$5h(f@RDLN7OJW=@-~}(XRs9X5LiX~SDCAx}YPOh`U&(l2e#2iq zO}fXzs}pqn0k3lmY$DuHnN!~%ipSNAO{)o#>LUULObwXsl!6I!fZ!=vl=Wf*bdCoBGY18+ zjZV>!)j==yrqE~8t7`B_wKH0uO^jWN9)-Uw>M zWieaP0z>wx!jpm37jN`Y%`h?@9ZYQF&b%K*Mv_(w#e+~{>Ncx6-W^6IpO47J1D}KRMD@L&+eZ{m zex0699fKfurL*nBjv{pg(A#Guxbmc{sQmJ3^E-)$zd>1qqK8N$E@^O{x< zxU{Mfd(Eny1L(cm{`TqhGT$c!Px3{Fs4!OU_7NCi#t@s@cJTgi!;$; zy6A3lwd85cFB|B3aDiht4u_|#X*&IkwHb_WJde$9`yn5OdzJY1I`QpI;@e&da8E%# z6msVT0r*ZD_iZonO-bnBl!TJlW@U`0BZ&kPoSTCJ;1+(Z0rO+Ouk9;}o+sT>!S#5= zpyLHc+_=Dn23=?|rcg+{Ebc)VR^#56t<9QP~AtI$l``sDqYeOh!UPnrwpKU+^E(4qWZG9q)=|0 z`(P`yI>W6iFHMadoIF>0ME4<73F(pFF(oxJ@jMpGr8-SaHSUq=a^wbeO`n8NHBXi+ zeBVWEjt|x)7VoB3o|OkZuKd?x{-r?5wPKFDvra0^aqQhYU;1wzeJbc`B9y!J4&8%t zK8H)>fS{W?o)+Tt`O9H>SA3pTD_^2o0}nyn;+*u>xHT1Z-|-V&)-#`0aX)I5B{fO! z!AtM05%owXx&d$3S+ZlUL4f5@yv*1@J|ZTa92>Tf+tVi9qFZ}Sbz<(|PEC)5u4n+u z%3O=x00=i3BRM;uAWxC)e9>fskt4}Z1jHsjVF73LF;Ca$tMTSq<`vJr#@26nfeQj( za2wEnzM%M*F&=bg>lJANqxeBBw!MmyyF-Lw2oFVQMcQkLXzX(|N}L90#6?CQ zJ_8Ulg-eC-BxMoo+Jau*eojVPCIZzkFd6SKBB)mDZ~FxRqQE;fj+zWh?F!v7_DpD#_T~xdT-=(;k39p%XL+Ka2CSA;Nd<)zI*~2`O;-wZl;wO=RDMOgVabA z1a{F2{0_a*@ZX2A*^z291lQ79;s5QnLo=r0|8Z-W-?L z8?{L>Evx8I00LWg=q2Jt(}aTbrJ4%ddsw}uI}jrSU|!uu0NgLIR1=6SWlUZOak#W{ z1U@m_OJgRYN=Li75MtgI%hC2Sa5?P;8HT6oU>y|TyMF2MBkk})kV)}9O%KO(c=jXP zz!1LB5V@fab3}RaDA6AqkqD5fP)mhVawG(+pV(Z$Fel7B6GcnabQ35G%)Gd>$Z$P7 zZV22YJoiys3Q5&O7#uwNP^ zZ2?Qghv+0^#~nT-I&hXO%@TXw5-n(`H=wP?SnF1VEYueo)aLGB%-2$KFbUZ#Ou+2` z_Z=KD-%1y#J;VD#t$gd)C$KC8%R*gKEk8YnAFBiG!FuozK+naK__W~tZ$!hF|AI5+ z>WduYMnI{`XmHQYKuUL$UJKtS&kr2M4#IM0%W1I%1Bw5FA;_|x&I@EydsigYr1^TG z7>`DNM|a?+HY6f!ES>SJ;`$Lru6RJ6xc?|uYJ(i<$2Da{giV;G6sg#^!{zu&j79_z zF=1gHY>E?B6O?MRpxlUQS`CPPZIwlf8!RQVM)p!0$P3&eyufO0R?D2yPo5lBERNT) z!f43wp%Q)=|oW;+^9-fF3PFb7v)vYx}^4BtvqfJu(sE#YChxO0BoM_RUeaOLR&L= zT=+~fSzT-+Y1sFO)3)Dbb@`c=8o$7ir})zzp{M9)l0o*8G7^P^k(UZ@6Tu6U9tUor zh!pl%%(WuNo;NlV9G8PTqb{>9!tmALXBdtLmx@IUlGx{PDo1c|xoLNqFQs!yeWJkp z%$g^?B{GF)$Bkmsk%$CuxNHE5Q)m-_yoB(rdU+BcCGyg&kQ1c=w45!m2(QbvZ1cOy zlg4BLo$76X2;fLxIaY3FVd#st|!SeOFQ+_ErBfTi+2+05}0fVlsymo059H;w#a# zI-)epUZxy@q}56=5;)_2RNb?kZq{8X-eOsxJ|7j=IHZ*yUSgx^EM&56yO1({JT{8A%WK1g99Mf5mvP0iD?Os#y~uQ#IGa=(K)y5AV*8)e5FOINs&9uh+O1 zPq+NIzN;uSN~Z5c#V!^0le0-uIpm3mU7;zi?tImE2xeE{eDo830o^m$F(sa?{N4se}`$4U!Df z`!P&F5H69X2*5T{Xm_uMHo)nurWO8AWw|_<#lMr@O4UBtWbi7FX6qiXCjMV$2!9!?r7O8zjN2`>1DktBF^9G# zwclcOTvB>JqnP{_{G&qQS@&Av!H;=QcDD%dv>MVEZ+grM`C#17L zR=U%j5W*G$VbuVEfD7O>H{G{8ZMyrmeY>;Z0xBXd;JAQ-N}`}Lh>ilT<2#_E435h< zj54Sws5mO)=!gzFBR_e+-#VwxJ@?#uJAFpz&0g3{K))aw0wk6!Hg10?0>=@4(P-Fk%?qX|LE zokgr+t_Waz%^r?Tpi{aCl{-dLmkx9!+DuZ(`E_ z=-%i=WMY#JwoK^Voxwp)c=4(Vd=JMKd?G`Z*>C2#3D&sxS-^6sHCPhhHVvy?nYOyS z7#5~-+In+1`n0EH`WOA?W7JjHR;D|^!Jg@uId~%;>^jvOxGgZZl+kk_aaCo9_==DF zV!)nt!H-&*R%t3npd2srQ4B1B{&7>%3lFJ<5Uv3n&MGF^cr9s-xh-Lo zE0fj~`x*ss5Ds6XsF+)>D=!S%B&rEir}p|;Q$?=&0xQnw0*mA#2i-)J*`{F)mGHr$ zi^ebHDuH}|ZO3^EXk{d;WoKI1rmm4bc#|42ZMGxWY5eg_&ph;0FXCwURZt6&T60o& z1YFZ|)Eb4_fov-@R+8jwRImn2<7i9 zRE|iLxq(l}cC;Hqc#gf!$R(96m=y*%0#EEnrD%5xLXgc;gb-%Xr?DV0u@hcY~g^K}5PS{j&q z8d7W+HHA}x1_!}wW9|aItV5uV00acWnWq`4N?VK00CG)!Pa$rwQP9#e57lT}YL!z? z9=9N0JnMw%YL(^ZKtPal9;<7~yzpxZ%bsdPGp7MKAtk2x3yUvhpLkl;6d}pa8ehpH zWhbOy%87+jFfN=gVbQQc#;d7f+>TI_R~ZNM9p>v*o(q>dKZTc)`yj7P^jsuM6>85{i` zvk^;R3!&7^AAVmWdtsbiol8u#56e!=qmuB1B85$rx}4N!DZdxz5#}$2>GR`zUHVIM zI=)`9FN^6@gfm9kHcc{@%9*7o`h8WTL?;`^UM?j}kT~zsOLy(w#g)7G)_57q`e~xN zp;LNmjn3<-&8i4key}?-HXTVZK}YG_Xv|!f;!rGthJR{T=pg}^NYYXfTtRp3!l(TzYss2oHp&!d6y)d9aj~EVb=We(}^xm!* zEMnOuJ|EtB#Vdl6_nFiig10DdvTvf})g9qP9Uc~^lpi9rFj**dT=5F;O~-3rQH7AF znUi~92+QxVVNj&M;D6fP>^mPM4R+}1(sBIG>Og)pMI@w;qIQA!03hbOax0&+N5(Mu zy5a>T!24u-IkBhf5eL$L7?E@Tsjj0-E9P?J!`|HADA?yI;Av>#F(xs1D>2Juo3)+Z z^Jx6VS0Mir+o4=qE+}1`_Jhfmm3v3!?fC;iC zawt?V?Yij|X|Uzx&6Re&{G32(?!CsKV$Uh;zTj`2r;SwvU}eS&&cEPnkEAj=2;S}D z(v7^fxfIumnCVc-SpVGzh7fR$X4lhML&Ra3C)bf9E*&bOIL_5SUe>CKefW?Q)(e-o zrj!KsdCgNfckTDlDxV@%2)jC|YE9DCNR_Z!aw6E{*{BI@I}pakm&1qrkw!PTGT6l?YQO-c16>D_MDMm5KSdKoNtw;1Kmps5< zz~@;_VVK*jCxL*=jx6&t%FdIs#W5(agL5ADR>y`{=zYt}`C)|$0*wBn*IKy@i0`fG=-@j;n>3cy6g?C@y8x;=`Tit5sI zvI?FBEs&mm=It-|Qg(Mk-n0idI;(LC=tn2JU-u;RSIz<@hdxzVvg$(0Im?}^3WF16 z&EzOo<5v)^SN>e;OJZ3D_MltBh2{qe{g|cU=~(&gnst+-rO7dd*f2ZeY;F8&%>OsV z8+Fqw)l6A>!div%S|%{aP*bzeYD5P${Az(E(MXmXuX3x?Xs(u=vX()ptoz*islyV_;q_3Tf*U!td@)jFvqjWW8nv&b(|7bY%5 zE1|8m@A+1=I-%rnfOjDLTO2_*&17G`z8iT@5?_S2i_q% zOl8+fOKwH^nAyly5{wN60_xz3$%q&i&^_xyzveO-Yr_w6)BQ6@(;=vh@tot(13T{t~AKEoJag;=u z25wT6b|I>Ti4I|Kb>=#hY>n}j^z#@b@60j5-aW^Xhq(R6~|Bw*cG^TrB^cO=;904cCrBBFboTW zjh2i>-J0xB$q}A~bB81`T8Qt>RzxqRN?xnSRv{K1PixT-0$Z8r=&Y$sAN3q6%I~k( zB5BP!!#RYBFrh5sb5*^pgGxCaozsbfs3kUyVIJ2>a#Vnj66`sLWg;QJD~xsd{ZT>n zRW%I;DPNrREDSBOi1Ts!V;y$*$;S$XeN_Tp7o=w$D|>mND(}-i1-TYkXA?@JJ^xCB zL#Ym##XMI!QP<(jbh49<31d^5c^~AIcHQViv`Rymg_G2&hfr$0lMSKKlt5Od6xiB{ zr$l(FEJw$#zl>N|fJ69W7B(@jWwGn0s#9=<6M^chsQtcSgf;O5r9~EK)ySq35*vNCLkgV7#5o1?Zj}Y(x0iwjRB|{@xKrn)oyyl%U}VWiKInq zz9KcghdR#h^o7x_NZ_>%mDagtf;v@Ik(N$AdQ_nygf;6b5_qN6@N(;?nohYe7bWeI zoJSHM0CZ@$3)wqiD+M_CJ=2? zHrJ)bl%b0*E$l(bTcP zOh@4Y5~r8A_KKWJia5L7w?1`%PIyanbRE=+-d4Oei;zOSt;+nD*4nwHGD&zyceHXA zb_MQ16ewJg6XmWyJW+Of;s1`CJ+;CqgT`4$SE>ePB9$2wR_4+ahQNf`jv^s7fYvt~ z+)?3ZjY(yoIzobO6Uyko~Hp{dI7K$K|=M-_dAJLMO1X2E-^N+`b%J531%-! z%+Se?)Jvq9<&RATpMD7}kER0yHaa>}{kl}}w)?Dpqca9mRqGXQFw{uYOBC#aIc2Kw z{NO1QV#bBpVY^%+X3Tjr>@XHjn^YJ-r31Ix#Tvh~*zwwO(ttMr!_mHAtfNK}sA#!E zk6xIc1jAZ;I|Dd%5*1!mndBlbx?ZCc&Fj!wsCr3QkZ0EF)K_HBATwQYtp8P1lN}>t z)Zg-?#eiu6Dd?n5Yg(c|4|2SKP$Nde@}q-=*@uqt7feyWjIRH?6Fh`M*Oc>x=vwpSt)J?{v*zA7l z6kbDV_AL%K%w;GhP4(11ody<|E3B%^37K(?P1UCl{2Aj=TRwR-pl3F=@Qysr_X)Z_#Ss(I$KB$T2So_;Fdk2(LWx^Vmy)$cVPUd4Lbp4THb zoTd)F;Ux7)4yUQc#;2z;(G-N&`piq+F9B&%Yvt%P-;=2l`&y~j)=v&`GaeWwlUB@d z(S=vMkMzTIg21n=R|`RcI#m$7ugCu&L0#sDQqS- z)JR#Eo=Q)=9x;`v>QGXds2&-Wsp`@YO;(oxU%#hT;fh3c(62~S5BrKlb&-#xs*5{j zgISmiZXJh^nVj=UEnMM^AhI9ggiqBXE->_rW(m~@^G4RRB~-3Ub$vn_+53#QrG1UFKz+XTsObQPaZur4VzQr4v>^aYnG8@p(_I@Cmy)gvaFt}ZRr z32Tw!n@|qEagYKb>Cle#y553UFiWNI+O6uWkpz+KyC*8g@hV@FjvDVH@jibuxUe13 zU_znAdfFHo6fCc`s$t1Jw8|rh;lxqvuG5BA-`4|Ik)$S&Le(05Sk@qV9pK_-DZ1T=xzTYnq&(6}5E3vn?H;I!ofNP->E@Pl#1L&irE{<|Ona`A1iNl2I zxu8?8-;KZy`Ig)$uX>7GBz?#})mz%d9M$7ZHBl4!@q+=O)z7rdQJildtrCny|B72S9af@Nt8-Qn|(qoP>q+yPy})=^U>s|W4?CtPX%@Af`qTqZmWPNx6?LwY}m zpz^Ddlye7NF`c!eDrLtCK05iz6=SUYD_5*Ir2{Luhx1r+hfaz=<`^WxRu1NsPU@bE})S^h>)$-|{ijZ_)gnrR>@lfx6D@||aoSe@KG7%m4!e>_aDX2?!?&zw`F$0oiMxLHqS1&O|uKG zXLiAL&n~!)vkLB{S%o&cI+0iaR5dZm;A;?iPm&r)yr zfmS(iY3W)))BNX%6%{7<%>XPWu1#YZ_|f^nwC7-JoIP9^T_2oDQXgR8lmS!Uc`j;C zN&C@LvYb<0(NV+KWPPS-zEmPp2fY9IOnzpeEboC78y`37TK4O$P+N!CI<(dy zwGO3q2=(dIeOpYLDTK)8Ty|^cgE9VJu8h)PvSp0oIo-fC<$&?VO zMMC@zO~9uMc`s=$r@bHL zzi2UF`n)*>zgZ7o1?&mrTMQgo{&Mhz*Q5R23%X7O&DR4a(|+2gzZ3AE#?gkdSA(wu zI38ar#2vUE1DgK`zP%Igj^J6UMTqloyamTMaAbkY6~NpCbl z#{d@`AHeZ59DL5dD?tOV&!XHtE4=&PqaJ^O@{U^$dBJfvjv@S>1dpG_{c1nmH&HLl zyTiXOMZOQ=-4AdqTnUvmkff$KZ@%zw@_=3Q#F5ZiF<$HC{s;cxx>U*NiIjSyGj zcrT7ea1h=-=L+$!TZCvj5kK(9e6QQ&U4M&r&z&R0{ar%zoClr(2H_J9VZZ8pFCEKe znfsBZ6VLAVpFe#z(wr;A@MiD|WqcNJe|i$+{A3~i8)=r{x)i^cZAASe&0?f^1~AS( zONhs|3h`Q8KZoDM^^3rx8)=>e4D#SgP85ul^ptPr^N6 zKjU9ltb<%_@N|vOKZE-Z;TdsdIbSh*`0im`Kd{5Q|2@*|N12q5 z;$hU`2O*cY;W!|~$8fzHzr`?Q{%x>%qe2|O{j0}?cxpn3%ikJCZ7Pf&-vvX1gNSTBUZI%1uXk4C5XoPXpi zuSOoIQ~bTgf6jaU?Zxkpgr2dkBBR{6zSr+Yvflb*n`p$fm1?G$D<1g`? z@2O9`dKe@>Q^!p1RCoE_JiDT&`k8t6o={#s+kyPA!Es*=IKK`(Tjj%iR|xjE{QGYK zrUU&y0|)7&o#K^tg;&}J+6eOIXSgnheV~mv?Ar<265{eB|2J(8Z4>j8XN0eDUF6vp z!sKt-F4`5^Cq8=-Y-QMHvK?eP{t>6^@P5};AzE?pnX!-5xp1BFUHFeSjIvnEMulyY z@qK8cNCRPz4%$xQNPp~BlucY|M=38%_wtiGd&}pyO53ZEekK$_DxnJR)Y4~S!g!j}VUQa)@{GNJs6Mp{)2g~>aY#P&z;cCjF z4l*CpuE+DR?TgyIs2;rEpVX?p!`#pj4h$vqOQ=6kXHIM-8cx7K1>n!q~{)7!*aC-Jfz)V`%e0d zZ$sW1pO5lInDlLUrLVxKRgeU{nqJ%}^;#w%^2@#W|bQRYcAWo){*)zZhA z$V^z)i6ePV-t(S3qy1)m(idSJt$neFGwYFc%XfS5tX4fS59#{|-amzdFxTuX%LZj4 z+@9z%A$x>N_^e0b!}>PK^7F87&-}#Yv2m@6;y;O@uOz+pKL;rF7KAXb870tfYsIFL5x-S|>3-@~}_H}N7L--6$ynRQ~| zQPz)!PwHjR36vQwmt~o9$y?$<+<4DF`r!5a;Wy!3IIWk%I~yOK{(RJrCk@7zrw>mV zG(LP7hPfy0=02=X)N9gB9`HS7nD&TvnEoi~B7Kx^$_#x&${X!4{Wr=veMkDA^fkvW z_vpD7*Z;v0m4R(nAie{73~f`RrbJQ0J5Z)Au)a?J~fc4ZJhGUY$m9C0{;?GHUrarcOVByu2^qIvco9 zekrTQPK4#i)G6^L?Uch>I7ell@X&|41`!BoaUb#F zkWRu{)#s)81)k58M}$efP-a*j`NjS`pA#3-aVN_AHI6(EJ|_;f>AB{D!u2{}&#GKe(5tgy%(~UbZh7*lFv0kG6fxh>9ZLWzKeh|NJe=Fh&*L(5U zALIA34}gh~yVG9p^{1D-S%{T42=P6nxeq*F{T9y;WE-~-&whw^BY0+P!s0gpCZ0Wl z{GY>f(vw0Nork@+Er0*r`w;8D6@45WYp$*8 zTaM%9zn}wJ}mXgsbU4wt*hnDaI){-o=>C@i%$-dT#`5-vB4Z*?$aKpuI5h zcea~sGug*Kg1n6DF#U^B2IF1qYjC{aZs5g!tLcXs-H7VvR5pfsJsn%ziZA^mJFMJdnO!w0v$S0n^9=})N z{tY;uz;C|eAM2HQDZ|u7j<&H`$e8HZuCo(@j}@L$_#n(Fc@OuVa7(#E-f2DpKt?gEpRaLcg7Bx=L)>L1qb7c z#-1E0!sfxQtjBfr0muNZ?}BY;f{i-4RIY>T@%!&_`~?T=>i@u25GMc5L%H7vT;9`G zA3(lO;y43op2l^}1mp_Gv9Mu%IGE1--GSd%q5L~>m}k5)UPt+2-Sao&w~R;fo`2M3 z>O5tZGDz8^Y*Ci@p7LeNIVLrg{hN;Mqg{>CeE_n(7v*#8%frU=!;PQMXY}vr zm+_u@Kpyg*eUNCsfzN4UO}j+C(NClg%Kig=NX8WC!&63RCs;3QQLnVitOLSe|BQVw z@|gZ!t-cW3QPw|YmNLY?2K~mP^*7k&7+L!s=w^Mjzt$_8^2Og5<2UiR7uVwJyt-st zNPJrH`)fFuj%nU}4g4q^e8)7#U;YiAQ3n3$T0F-=AB*Ws9$r}v^ROJkB~1Efyk3JV z)3E-DCt>k9{V)1td`{V;oP>3SdB|gy$9u|_!G&cJ2K~F2;ow**$2!BZb_ z`YGfweP71J>F+c~xq#wTPOSgYORxOKSA3%9SP{Ubd7g>{X; z!#YMC;xqnHcW7H_U#M$~P^q-x)Hmt``*747%8k*teQ!d&ztPi2>LGnL$|`Lb{S?Xx zWsbT)n?#+Y+^{_Qb;drj-#y)U6m^36jSf)vsH4=`{nrWc0B{28I?bXtKn`&*ee~~3 zcupMI-jH_VD}5dGkS3<5J`g799L2L`KK!radM9**vQIf@dF;zDoykWYP<9B5JR!Y= zMK}gW`q<1%-<@fhmwiTF=>wVAy3s}Y=#<~6%u{aI?y=qoZ}jaRJdUN(H>Z!i;zrO5 zI!V(e{0{3Q;j_%?#tWk5PiG8;@ay%XINo67FWRQFA3@rvQ?%ixZ6}?CNxrck!F&20 zEF+2^>0*C@Y52$f1FzH}$~nuP?mU&L;v{U3370(P^{ z&&VO4lQv_IqHR6rj<6raetVb~&O5*>JiiQA!Z5h}9iACEGP1>dw;|p4afIKQd>w!7 z>2d*Are&X#I8pASJZ5>OEao?O@ScCPEi8*NO;|jB4R~k2)4TsDu0{vkcX{ua=6COb z$m93da1hQL-wmC`vnO%oZiw-#!C6 zUr0M7P8e!Oe2*wx+Hv#E14*U=Xa7W%qM<9)#=qr38ZhPt@i3?CB| zr{*Uoii7?V`2KAo3B10s$u1lOH-qfG`SFo0n`Z?`Z-0LL09FYlM2H_T&JPcQVf}@J zGegUO!iSQvprqgfD<$Gn8P??J#H>KVo;Ug7S%DO0)y$~VkVxu`@ao%}$N#O^MXA5b zzn_UT=pC=Sh?Txe3aSQHNvIr0{bT2Vt^q7^oQR2w$vZTFO~(`F#b!oh#o^(Exl8&> zEWMMXLkE*WC=6q3jf9!UhVZ?C@r0RS6%yv8y~P)w`x9m_O-$}7CCm%EGn!ABcLe+X zjK$20eL47{@8K>ZYu-LVy2(jO1HT#rW!~3Y92kg+j=_Q;dP{|oq^#&0EcRiQZo<-` z`f=MMgLtvzw`absh}WtaD6|;8qebp)nUDl7^i7!6N1`WTb~qPY)?1np0|GWCM*ji z2pb9{ES!E{!mPss*z_u~#7*P}$0o*mw`|@sl+-~nO7-6WAXjY5U~k(4Dx-;<{J?(IxEIk-XGi(!D7u)%fzR29Vgw z835QTBqSS&Y<^!8aq%-&a4@Du4CM2F)6nO z`^MnHC06f2Y?L*UxFrS8W*}fQaRaDdGw>o{5V4PHa-Ns^u%#3}7?==(-j*_1FLMN* z*I*&3ejr9Vh{)><{FgDE8_ri!6oRIqbg0CLIVz%*+>_P>&BQS&b7V$7r8d|zl1SZ! z92!k-HplY#Qe#p(;zt~N_u+HelS#!I0e~3<9kEg4ti)j|N$JJsJri5knYF;E5NzHZ zPtD8Pp)C}?Q94oR@~l--g&P|$;FHry1w-$)+>01POhSwUJ$_&W8xAK8L~FH5STbAW zSdCV{wSDI z7Y6Ky3e%har>1fzkf~ElZXOIjs_>r z&WH_3V%G-!ah-WqSddT%OfAen5DbV4Hy~cb$?)Uq7aUeB_r#ljSsWjUJpohaOwg$c zmDoEz(6?#Rrjue;tcFk;jz8?@WrRDIKooS|tBa$Bq0xb25`<0OUVaF2MgRtj1}fekGFu!zAbqV^u(0XG0gtFObBmN|UX7W0H*cG+*5L1-iho;(;W^CFBaL zA^vPBjcDmr<8k#_=L4omI+Z#vnVKXszoP`7EcRG9Y%af}*~M^-KE8v88Sw&CV-zdiiwNKu9qQg>Ori$cV>7b4BrMHY1@-8 z6}D{d@-8un(SvTKF7FaEZy5nhEE_)r|B08FK!TdZ*NpL9=ZP+!V!|*)#Lw9`GS-D7 zS$S6z)%@t81Tq9)b)ba*F$*p;3m=9agX5}WnY@`bKuLW40IdlOf56-8V!SpfMg~TF zuQ5Qwbe`F7!-vBAHgjHlG2uk(+E54j@D=76@u~tz%#KMG^Qes-Es8%L9jkjEK;t^c zu2NyNe*l7(v?eu>fw-O5N*o6Idy~vmsRV*}`Dk%QmF1IywIRd9y~*Oel_-HOO6Lq& zt6{t{u@xy(6@Q&mB{@b^5=1{EfUsH-OZ#K;rC93HVgcQdLqY3l(4~-|T zH4YobRW~qjurP5Tt{p@+U0Oma6RcWAiJavhUo4p`F-vvk1h$aa9!gLb$dtfJ!8`3r zvLb-_5)Icf<3N%?yYCEZ=Tz_7P>3se=u<;G5f7alLd-Aj6jPI%aqAgm9vhCoQG+Sk zzs3~gAa?;SV`7KebWn$}#H_rhN3ss02{0fA*1d&6uUZ!Pxz0~1uxSlG+8)nK zWKTMNFBr;>=z>1SEYHhHagUiXsI7w|!(CW<2Jfg`I8LQ7y^dI_F~MYFZMKYKYyS+`Qf2_sY{=e zv7^t!1j`wd<@ebMGYTgYdxvpQI~k;araeXaMqL=Z)`v_{pLUakvs|bjb%?=?hEjBR$=xI=tUM<4OGsoWH zLKnOT9%mlQm*a9R$P8IYf-}oSuQzE{l*%YZ((EXCCfQNC?vBo&`2L~9$m{Bwf$hYm z)9H} zeVowU)8$=~Ayffk2Hmr2I5TO>>Ozav1_VKtd}*{By|vM}E^;LR#N|d}FI27skt|+? z?MwQL+;M&w8~=0`#|OI(`jH?1!pLC~`lbii3vb30ujzqKHtJ(+O<#X0&&Iu!$3)V> zd|%w@VAXJ95*al$6`oi4$85so^09ovm|b9g!pL=b5OsfV_7UYWiP?mjXb!7iQAr)D zX!c~|&s9(o3`F}d3jJPR#J?uv?>Mw)Y+qlgdw=%{$sqtr^2}qUgsnRo2z0U8^5K%K z{0a~=Yx&BkXgD)$`D#Op)$;jba`1$WC!vZlF_EX3YzkVgy@i8aIAUfm_h9z-Bkl)vpY^&)22Dfo7pr!CWX%lCkholp&O}^ftk+I?4!D4sp z5T~BwttY2wG7@}_qm~=PYUv*>@xT@=C2XhIg~g{7VHZ-s|+YZ*ULBiFmYy9;d8=QqRqLcX37g8Us}p$m3VJYf@IYe z*qIWBJcqG@XJ~LQ)@Ou%#6*=^N)^H~v-lsfVdi0M{}Ow{3*TfXz!la68}XvKNiqkM z#SSHDoC1AFfHC*SNti7ujrcyyOtbbvMH77IgB(UUsub6sU_Q7KahHW^If`xknX7*& z_Iwy-8ydsJ>k{TlC#cbhzOjTkC;JoU#Da!|g<>aE58>Vf`=@CZ8&yrfE2=rfVJPnW zEU%6dQXCYTRC%z1B+2(_h_s%NoH%!~r7J9VZ?P16gKEuWep!p2VuFd=mH89-0R)k3 z34|`(XmMhO4{3!=oZV|QlZu%qD+$6Pqu918n zS2_9nmM23@#~ULM`-<4Zpg1|&kJ;?8N6*4>%&FndIs=jbL`b=tG3mJDj6Rt`_L)Va zJru&yk!U_qB{nk{rPflx(+q?P-?yQ)kGPs|DO=EgX$adW<#p z;@{~3oo(^g`0gw8_2u`)6{e#h%x_^}DNkHC<7i-s@hk{fxW3}IDp5rm8QxbIIdRL@ zE^{->7=q|oQPMEg8fGGS@}b~U3s|I>c-fQAJ{fyncBt?qAIb}7*Q+FZOOx1TGw#rI z7)<;fUMg}Yn1~t5kI&5IF)M9G_!9$+A0+sOWC)R0O~xMH4&@#jk86HHnG3OZod{*- zq#REb{DJ8iuySAFP;UuemF}O6JkM&2N5Oj|Mqq+aM;cUj_%h?;gMhY zL4MFL{$>dwv@q3wU+hmshVu`{t#gK}_fAZV@r$2@aeN&ZGnV^?h7#>`Ls;DL8ygQ{ z75v^LyT4R|NVI`SC5Ql@n6fv5h2mY<6m|rY!;@sy^Yk#E7YhWFWFIMD#5lnhYBbt2 z$(Ll%g}YDc&BgRhg1soK0VF{~)2=5FAu2TC?F7%eMEvJA@*0f(&SqQ@;ASD5CdNHK zun1r;cHbGvkM<_m%O(P;f2a@fCcP=nEa8L=$a`mlMr^l1T4OSslQ^+lqHj}=cb;MP zgb{-%7Ku7)b%4Z+SR1(+yc)Ap2Pa2yiOG^$y$<~u28`_J?bP=S_0zl;Ul_pNLI6x% z--v(lBHVQ`*TWwLY%Yv_`{0iRH!n1Ne0yLPsWAK>#g=YeoVYw&=tltKYt!5{?gfEP zYg8-^B^WcRTe%h36Y&K5_Dm1>(CFA?+{R#Z`sx=*R`{i@(yZw! z9j~GO(fn))cL0w-XH6Fas8m60fjr*c zdl5VhPU1^Cycv$YJf$8dDq*vLS*kVy?m^%<-?vZid^*(Es~P?RfcKri7XmWTe*xeo zTGTgH6&g-^ouil-2Qn7)RQvW$j_!-Sx9L;}YO#fHbnp;8q=RvXWCIk(@o@@lgrAhc z#9e2tC+K*LVZXDuZAO56A7+rp_D+J_vA<{#Ilz8Uk$KT~Opgv$9I+!{X>+fwB&x;ShB?|ckxmgwZBjE5R;%@_7 z2Gi44>!rviI)_I zClOfg+I_|aJzYC4I_qp~MssjzKh|H26~>Rppy@xTa~L)Zpup8cFGY#N~J8hL@6PbDkbs6wT}_X7Pc*(!$sOc<%FFR zAGYO3q`XU7O4*K(Ib>B*W`$ZJ>?PvYj{L2X&nT6a4j!katsIp`Q7+CF?j1|WC0?FI z;+HO}2LCP;m`QK7LE9H-* z{IQg;S4b{DB;|*Jyu8D5?zV82^W;sIZ2{$tEC+972Gq$Z56g8N=^|As4WhSaG@o2y zFA(oaxi{PLeYVvi8!X$&1iWe?ubOPf(!81jhM^40mR2JmbYzyvjzu`sGDqB##Vc1W zmJ96p@>W;g?#d6k@~4jciGzEEdprSpwvr1P7OIw@h&3!?t*yKw)(Me85lDqXaF8HN z;XWlBbd{mlEQ(x+!$q<7X7RXJ@{8qp&bdb^`80XGl)rH0FWrXo)%E~|4Mbs`p|aCM z0aY7C!IdeQm6-sM1|YEkM`~GG*gM5qgb&G2M~NR7Qy9wq?Nx zr0q3ARAwfBq!PCLlaha=71Jba#8!fKS8VfEt?=r|ov3 z+6B%MH}+L-ub<448CE^4m<429yNUEjUQ3> z%fm_@QTq=oNbBd7co^0SS#9|SsjiV(XqcU~Q?49SZ<1<(OwUcH@IGUyIT>ZQ$raB0 z_HMOVZjvoB+qh6Ik!NK|d)CUhxyA<71R1ccCatW>KFxw4kYstj+KGyBX-l&zbCJ44 zrW;?bE~mZ1UjsO+K%JCJ-8NK9&TiK7PB8bt`U){%r%*i&DkYcUt~0e-WzxqnQQH0# z_}z%wo{!po7X(G6tT8g~g5H#HUnUm0?G4RZpefY;pVf-Q(SNO}Mc~9z*Rb$H7MxZu zk}kkqtIl~IH|bqsfvYlGZKY0<+tukHXAAC6l-ty4xZjNX6XdDt6qU2n*22KHoU1G$ z<{%56G%J`u;G2UE!^BxR=n1&yus*L^+?x?_@CSY*_F1=CFQ%BxoPLV@p; z!p7mJybX*N;s&Yap#H%b*B$}?Y->GCw^#xJ6l$k<5p@y&Z2!uJpw>V09@gA{?>z~h zqxs5t%$v$5v`^lf$W@6BzEGkEjpPP#s!r#n;n`roye{6EH~NnDM*rheb9hh6`S@Hw z;M@JQVAP88?aoj>1KopA`|QCb%;>*ziQzZ3QkH9gcP0g|IBc=Ne}o^F=nTriPbJiK zQkuT* zGy)wk0QLs)K3EReRCs9KpPVn3pg}`>12epy4UTRNR?svogfYew*KW4v1H=(|%wZ{y z$mYYaYxY+09qGbQ!%n1C3eEJzVB8+r2RMX$m6d{dz6Ru6Elc=4p)!G;?4og7tmXp@ zfJ>{Jrp`jaSwM7P<7|2xa{I0FbV%<9q*!C+030@1*9^ICWZ2Yzp%^mQr1Zbws#BZy z-ZjVt^I>fH>6dM+@#*vW~)?pK*?W` zO_u;Kw!CPfZSk@e@tgyP2%Q3-m1(x`TB^itNVgG|TV}*o@3P&yD0x+?*T(ytc##GE zf$M0!g>dKsST;<&eN`G2_+u~#{SpJZfsCXZ00Qi*=EKa}IC8xWfV5RS1CF7+K#L(k zj+1Ik1&+Z61rl=h3JXtM+l4c2ES{RL$pT8ly;6Q!ZWo_LLa$oX$CdmdDy7wKU1+y7&kr!%P7LST z#}mT_I1JlC;e&F?15!RH8}-@I=VpDj^*N)@ZJ{Tvq5B2qp4?fjF*z?K(=Lt{91STe zJy+ZhX26@=E>D7%pgp(FhDSlsvz#m{m3RMQ%OBtw*w)a}=q!hXqvs~K%d^q&AJ;tM zoM(N*mfy5di{G;4w{b-W%C&!J%g61EQa`e-r=V?OgtX7M(@hNx2K*YL(`L`J z@Pz%lbwKD^nTx`tXk|6bwc?WoCj4ht-UU4b)VX%HsWBjOEiGm{pyCOA(Y1iOeF6OZ z2I%i?a)Y=#<<5awn}f<&NGtAu?N&}-=UQ_i_nwVB7Cr(D*L;a56pluf(VidRFr46u zOO}WaSZ+GC$eIs7~40?BZ5>PP9Or_&ZpTqFMFAV=w@bj~W-4mv>wv1_QCbJd7vUsKkwjcx)w zUi9lMh)y=$l)92t+YXa+N#;5!-!H8Nj(>H<7n{ zOLnUsHX1gV?B-xFsUL)U8N|zyhlE&ZZ&QnD3bQKZoaV0AJiMGdY`P{Twk=pK*DStN z?mk_-heppueE|x!N3N6qBqcCfCOd7_k^w;>&N zKJV7&o20%h(&wDHZxq{BJSfot__K80$BIz)rslSm7XSY8W^|jHt(L8zt=WIJ!oRkO z$I$PV+pPm|+9%|QHHyq{Yj@REy(!z6(1%sD1#a zc<>qO=1mNmeNC$G!wG>5U`jvnXrjQiNmQv0>?Cz5#oTDey=8qN5j zZdzBO>)5PzsjCnh$U>ExGaDQvaqW3(E`k_$>Gql^pq~%LRgMG7(WTAFhUQ#{YC~g@ zZo!k*o_4oQw$H(Vbn_8iQ)>Q}MPR|==3{WbY{d$kSGGYS5H^FDW^~m-8^nWF=C?!- z0f3+fo~1FxiE2x&lIXV6>pJbQY!=(Pd@+HOlu_Aqrd2#;yDJ>jK$a~O3bJ4k@b|op z%Kxsde#%&&`ms#^P+BMXfeB0QP>|uA-Kv%=JB6n%+9{Y5d7=3>A?#wS&snU;cR7$plnJ7we51|!y2-KEf&n>uIrxqnw1jB%Z0lI~0cd}uqs#F5_8!|H0oTsEmUd&ay=;ji?-xJ3AZCHwq<=9GpG|-=@aCJMSd-9Y;J8y z2TvP=r_C*{pU28(QI9Y;{_i&N0W@8(Lf>fc=Z9}ull?{mS|M_lF9w7f{UI`iCs{63-o zg(qeG&w%x6`F&l}PdW5r0$<|wP&C&20yU=2SF1C+KclQ?;nC}M8sugmo3QVMlW9d* zi0y@Hbin6qw!>|8+>AoJQX-y#=W`nt;Jnb8C+BA8b-2xrGgoS@v{pcTSE}QjW9>Ca zu~zFRhX7KiXT-fJXaqpZ9Q30xU3PUAYV6PsXuJ@9wBa5}MBAj;23IQuA!b`)F=7Lo z$hOK6kL$;fKO1ftL{%O4yp-3fv_58EcNAr=OqO6WM6bl>yYxrWjRz# z`A56|EbU)=n}H@;jmy!JEEmtD;Qh0|w*=&OJ3Vq--XUeXgw?XzL8@|WXAb<=MKUi} zI;+rpd_9etzbPa4$zi!q-JFIzem-qIWXpd*D=F_tv3@@Z@k7P~@QT=;+yp(A??_Rn zpGm9VrR5*e>i0Z@3uR0F3!U|uG~7w}75c|>U@-*)6#?OAH;MOnmU1vq0ppa>a>NvO z>xMvvo0Gq@6XjT4{O2Pd`g0zH7;I}E-ToMPbWM{FB z;gP9Co9*Gh*%BJ-94>pyuSckvj@1G+PiAa{Le(lz+w9CqzYL~0t2!Dl^&Dy2y4Y^9 zK7@AVWD=^qB}5$|w{$C2<1$naqU4ZW9_@%askvz_y_=RSSS*(`Ep{+k;z{qx*^UKm z^V6Ay;;%EA2gpPDaN27qdJoH1v27ltLgLu!OO9zU+QdT`@&IRTwGlkc$5DPC3Jvgh;YfTm zBfpkLW!{^SzsgYh{*bkoL#sRk52u=)5QiO6c&^&fpr6fBbm$|6F3h%e@zVzO%b~vT zUKn7ougahov~7WSYYszjV>JAO}JQxN`A1+a}T5O=XKztN=P{iwV)^$0= z!2jGJuZP8O0NS2!b5P2|4bmtCYZXsic)eD=@U7YN{mDcLO1-OU}Vv|}6&nJT*GVpS&i**MZ~ z$=YxF!IUJ0(_D? zrMg3=Z1fRERieCcpYlNLgNW*zh4BvMtUa>JPEU{WL zeE)`={XY$;Nfb_`~QcenmQm ztgGcAxKUfI6Xd64``=2OW$2s|kAbyd*oQB+-g&Wn*TwS1^1Tcw@0J)0lt1U#FwwB&@vM6&p=IL^E+0A0e=r*`d8u~ow85Q&v(Kp7 zgYvWo=@Of8a0k;$v;W?dnx}QkJ)S1S0D5g&!5K^GJh zeXq>RJ+jx@Y2^XTTQ4kI5fhZgt5Kw#m7Sb`Rz9BCYS&Qs7#I&G?-B>Tdp8x z3+yJw7gF}I$ho}{otdvd3vR>E8mgG#<8pMjNmn-p-q;g9A~$PUv-LSeS$jzC6zw`6 ztQ7{|q`l5w4PR6jG+d_2Um5WTvg-RW< zM-Ow@^f-NPI*cw>`hF?SN)tm&PXN~&sQa*wJliWtszPJcEZJgV0Mo+2a*JwR*6fAW z!30W%{1}`Ov@RSaRBO5 zBK;jDo&hchjjI$i5b<{S6s$&wIx7WzD~V{aJl0ypxVmR4K{e@8Y7;tLtK_ki3s|w` z)(RX9$e>Y1nLlHqcX6CjrvkS+LTUW>x2e>7Q4pd^sv)g@z@9lY04n-#QavWy*>z#r z^X#`UHgxoKmL=bWl-64ze;DI{`P3bEQx#vJOX+&_GW+qD852KxZ z5$HyAR1YEZ4h#etjqKSWbPbyM4BbhLSX1LusdPJn1!)wdEZph$4)iGWl$nsrL7$TS zziZW_aL%(L{gf1c4}UV%1M9p=9*6U4xdP{vavjd=<(V+kJ7g2?o8>m7J6&#p?{%V_ zhbQypF*q-ii*Q~nPsaHac@EffuI$8p7p6g=>w(4-NXt1mx5=HbbGEzFy4a5uI1(I& zSkAQ{VvqpjKMpx`j;N>9Vbvhge^BBU8{-0YCf8=y{@0G@Eq2q63?=)lR@-{1j*n7D zet{;GJuZa!p*r!7R-|vkk;4JM+{tlr0QA;z=u91vUD^&PI??Iv;tzO-ih!*|$U$Wq z(c-{sL#vGboicrK8^bQd$37ND0Z69olb{SwI`_#Z(43^RCtD}M1%V^-UU{;8QdYbV zPMG>-gY}CB`CYgI*>8cLkbe-9k~vN;fzk1v;5MCB8!>|=4Mzg!pElV4i68??(@q9; zK;93880(k^Q?VGKi++Y@vXm_RcrRXp6y-c=o%MtzpR`g>K-nLb^WP-rUXSMZan6i@ zfrb%!02g-JX@Kpo!d=KXX@-P7pX5U7I$8$Ipg4{N&?ng?#C!y#HfYARX~r$l=Ti^E zD11{+h;}#`#{lGf3-L>I-01XL2t;JjiOpKgGTjVG!ZBawnw8V7%?a(_2pffEm{S3> z@;)iPE!$7APKApDH%D$(r$Debr*M}!MV-13S4`g6oIZuK6s%Kiq{{fun40g*wFzlb zVk7jH{X{z3h*_)Ei1DIHlgDf8-XX74-*(lLY6^}^g4%v98Y%=HEofw+@F5D6TddzD4o(1DcuW1e>FN5$)JGudl zA4g%}|D9IuLom3IXhfrh*y#mEaOYt3pa~k(A8&hpG@X|GYl!98M5u?GxFLjqp0L-%Nw)SHLPm;1|(7`JVieXJ^H!sWpvhA;|l?} z1H{9{H$sDtJ8Y7v~HpBA)hr5HK5cYgx|DB(l`$qgANtODNk_bu+Ofm z8&wSb`iQjdlIje&o2wA8)cMsX5PL%}(^c$(*XuqF8l?w4Pa0|kL zR=f*xz6NtgQmwl6Tw{F-O&2?ozz1rJvZp-Q`vAsh#-yY9j;VWA&T0twjl!VE$7v~SJWNqeJ*2zPX}7!0yIb2XZea8Rp}VElM-d-IefGoZ9p45fFvy-;caM8wMJ&Fv^O1OE(TnlrPT}N(e4EE%%?jshm9~1Oe2psQgRu4 z`Bvp=4*D)F?OEqp+48)!bZ|NiY1Q55c{gLs{Vs;YVeT*+IBR#J-Ok82a>CEOiViMn z!jhjy%v}DnEt`;zzaOHPorTt9<SE65KMw?g!DzWq#W1LVqz~HP2j%Dkj1|~bXX^p<3R~7fbI?YH7@Z9<`WW#P z##LGvG>~cwYzbx~1Nh14DK^M|PFtT%%g;dC!)(9Sg^_x{8?gOdF4zt)63y{_Y3uH^ z{1|TA?nygyomTu?W|_XJCEp0W5n`p?1-@e_o+I8g_hU5mYiYkww#WCZ&v!0)I3f{a zhJ`}ts8O`sAjLPq%qE}lj%K_U0coIHs^mh=oesGS_fRa&c(Q&OYP!Ll1EaxIEpo2M zd{`wGY{^$E;IkRCwdZ*Qvy0VBCBn%&edag&%x6+oz#J;kA}CSID)DF*mIMwcM@Auu zj3uBMwx~~-T;eSkkeFSkY>fAynQReX0$vWa%PSkB)l_drVU&HXTtRJFg_z)XWoD0) zU%W@<*9vs|tV=PQ;Z^cV$Uq)q4mz6C)?Ac$sqBy=i4r*sj;-RUF@ZR9*&q zd;_MEy+qd>yDyj=E$47h2U51gT5Pvki=YRfqBRG{=1aEquQnP%3!Iq8r58w}S{aGw zu)|uohR4FgDPTa|aQjWVBfAnl{uTrk!FE~(*IsP3!-Tw;?bKzc0e?ZVB{Bb$3v75 zgtWIWhO_G+uJtH7Tp#8t1J&9BoZ(3V!ZL{C5`^!ymK=xV*UF8QiwsX~*SH8wlS6Pi zfyMk5v;fL#M){Ua#}AX?0WV`O*8|#(sLcj6N6n~yje=vHX`_t+q#lCPfc5mbM*eK# zDF?Jx_p81QI4%}%#WD?4L=yt2ENHI!o>JE_3cVg9<(9n#tX~SuF=q*HZR-w8-9_g^ zS@ZG!IOM|cKeB>nWDzDvW$gm4=&AxA3tk5e118I$r8@tdb`XK!9BcN5yaYaDU;#p~@2C9JbT%w!%sUi`Dr zZ0IW7jChiAjsul2q3~!fhkv%(T1B?lI~Xwr`xKUL$xo;4`_tC15W}%#!$P>}86Iuk zk+SD&#q$gWCOAQwHb4oN%DuE77X!1B9=7(iF{ zq$NLLse3IOv!zhoDaM&g@?SFQFBuz-?PryGjrD3MCK?aWHAjA}0XGjdFb*RBF)P2B zg%qrooiHjXbj&;5G={{9HW9U13c8Tn4k2>~O&2@e4-5TT$PZGaJ-K_43TO{x`b+GnN20MSbL2pRbw zY5D6k=7ubl-_KwY)1&F&{h9J?^rc%BVR=y zHuW|o{~FyH`EJEoMXt+GV1qo(*={cfCQSjas2DRHwqb zoeM2(NcCW-stwT@9gMXwb&A7fEEPQ((2H>_XKp%^a)JEo^nB%%)y%m;eAk2~9@;5o z4~F1V7|q?Lg>)T-R6AZaCaYXNZ>w`6hVrZ%C2RUdxkx;n1GZd_WZOOV4_)Um8cj?7 zL(cjq1muDJw{q4aIpngOR=6QV#*!y0Xl4T@$>Irpv{Quug#f_`d6u&XMQ>jMyOx7p zyFB+A3{RBzl``nqdN!F;d1g;%==ScY(z_m&RnCsL0Y%xvot(tfPTlLv=ykQJEjL@r zGINeRJ6lv^ve+!{!SW~2^@(=hi64^>wV{arn1^sHc~9r_2TJ}B74F_LUwvf0ylXyc zbtR%n+l9TI#Wn*U3-!!CWR_a)G=s1svh^@@?D7urDXe^gyMl3V)JQ83(-lLh#2j^7 z+`j?Cc3N23q0uBgE0?ZCAEF1Ls^uuTaFK_R$AT(BYn!;;!NSH?i7rSj} z(}Es_Nu-#X@4f2WYM+P+``udTt&T3NfCs9+)C{q< z<=-wqkMz+6^1eCpmviLrn&syaI#r*Z=ojMok39C^I64V?LTv-)MT{3r(KTk*zL zH?(3j7!3XSe0E`Kj{WEypz!l%`BbxPLbrc0d!OY;X|Du$nvPpxwvY9EOLjxvYk3<>{VXf=Nn)V<{XCny3`4!slem=Rq~*AuJHL zW2geM@~t-e+ieH~DC>J|_M>eeDuZY0aooY^-?|V>fm<*)S7J_d;pvZDxRIM|}&kIv~C@fcjPFQN}ezalTjlZE0*h(PKB z`4eURR3Wz7B(F&QRrVh(`6VkeM}D`}{(7q)i&E;f@^x}Zz7nXxrE7Yvd;^GTnkz@- z=t3NgBl1_udRobG`Kt7U-L2adn&l6mDHUQ9B<5tbfvSpXwhlwKA>dR^yv9fd>MD!r z6)u}77)aI9iWjYvjK+mH8sXldH-{)`x{+#i)dARxCd3#XlXDhXKa>b=rF+#?O$)K` zfagW%p(5(sq$zC!rD!|PX{HFSt#ga^&Bz4|SKLKIF(dacj*(i0^kwzK>YVm&aNE?S;b7k7-9Ba|zbV zDETXg9Ql=!PpkABC;}(R`_LApZf8)%6NI;Dm;YrLOkJ#+yjaV5;8hN>YQ(8w`4Mjd zqv4uDiJ(GF4$QQJGisSe1`;h@aHx&`Iiv+Ps5Rwc8laZwGbz!{JgvWm%vuFDxHuaB zFK_Q1C`WPS`}e7;4%3ro?%b$w@72{+P9TvH62TJD1Q~#G~s z(J@Df#mR1`raPH$GI7{>i!nls{Tq=ZdEd%%@H3G)xgaML zxD#%xNS@SDdTJ!&X5R z;}0=s#&9qAY)pQadDF^VE*8BQS)vr1>T&mD^yjR{=D}kWMl`{UHEo$YA|gq-YPTl} zu6>}HvgA&SV#?5Hg+v26nQ+9GZI;ln9EbS6qS;{g-O;ws~I$TSwqm;ymgR| z3=J;}oBeTr__hAHHPQaKlE)Q?7G}(Yp}fSeqk9tV5MUsD5zbi88-J#(v1(?R_KclG zwhtCCLv8&3xjQ^yTcq5AQRGZgqG|exwK-gt$2lHIsu42no?cQ zaezjszYo`PrVtp@tz3u6!>`d&R3#;g1qcE@_-Wqx{tljay-SYqZtya^usQr(8Q88t z10ly?G2d{^pH<@6D)7qtz1G{stdfG?wtib@*(yo+O*=2Azis|bzLA|*_4|5NUX!-WtDz+M_ul$lOVCUVpeHl&tfA>bSTfQ4NOnaWafdo++Uab z>gJ)koG$i+nw&^_CX?#T7V`#L8B>d0lpWw*59(R$TvIbB2h0b0`JRG2q-f54<;Uhw z*TO3jnT3I!>SLM#C+)s329Q}UZjnpWccG%8LR+2?c~YGp%Uk4H73Qa$u>(BnBTv3$EIiujv9nMyR68(k~ zPHWNb`A?UcCt3u>)EE@jzqFW1cIR-qpPg!0T}umtV0&^`N4e?5IVU3l_%e2zzyS01 zQ$(TKo6ILpq=#|SDWwu(Kp3KA7m2F)(i5ojb-=X^tKl#7X7X8ij0i^~)1rgxq4!K}fdyCO zRoSkHr)wBaCD~bap~IZrnK`DDTwy-UBR|IEA;x^q5odNbyOe`LF|b|)TM6-i!^pGm zWtwtv4XPow#hE)Dn#+!FV3*(6*I~LA7MqNsb+208&DaI(Ld{e|rVN5+{v|~ma3~@(h zO^o*jK#PB$qn6_{GF1D3=L>mG%QL#2zk2p^^VzkJIR5W)H#JF&>O$O4WlQrk80^%`!+;CR%$0lWpkw zt_S{2&=lgsN{4-D*O$w9DS(WU)`TkF%2>3OI&~Zs!XSfNIa&nj$^PIL^|V?ozF~p0 zogPIW>t!|+Pe*`{lluz@sMpmWjD#}bIx>>t9L6^{WEfLWS2!U$6n{;dQ}1vIU_cw18`g~I(}jl6VD4Kjtsmtadm}1e1ibM?f@j zj;8V@r{OlSN5nEO<|^kDWN{wZ*eQ4TyMIR}QEKEXcKx_FG^e`3((8s|W;U5MNK}~h zlK(Get#yC2g&C{XVB7!1sL6)=icRO&v*{Ez-KPwABF{3Z3)ZY`*>Dsl3}a~{<$YKP zPu*2^LxYLzu5dgJ*I+?;`cO{0li4lidyYhASF_vSvyDU5d+;0zC>P?SkdrMZx?N!# z%?p$VE^rU|6q2yVxBt#&DfG-k?-#XlxXArr9#L!X!p6+wx5_+DH}XnRUL0T^=RBIMUTD7GmpJmwv*=|Cdk&v=bn#X zVhr$+D+c(;Rmw+j{I2|aK5}s=*m4;kd0>E#{OKz`@(b9s@(#S9o_934w(^dX{*QTw z3Ta$^G1*%v)g_Qx@QbdP$e-sQ-JX98ZDkN5pSo_NYH(f@z4vwg@rgR*?o|Hq{NM49 z-@!j}Tp_<)daJ$;woF7yMbhkV&ixru2Mjx-l?^7Sa zpE1}E`v9ex-CBaHm6Lq_HBPcZImvAsa*}(MlbjOCsX8T3i_Gczd1XN^RZilQx7lAD znI(GsTF*(oP)@QW&&-6AJT|~dR%e{#cagoCYzcdhlhqj~ximJHY{*G|WtKV5Cs`Yr7qnTa@}HCP&2zs^s6w?x z!1(yP&(<<;E-DOgm@{I*Tr)NU&iz16E$A-~hr`_GTlGT}Bsh%MZd4(B+Z(V09AXS3#!8B+kV1v_~sjqM6xy7+#Jefbxq*pd?f1H0c|UFGA!kub8EZRr;$zNu3fBH)am5!Z9oN z3E`1{HuqNr*Osbzzsb}*(!Nmh_&t9jw{snA33=X2H1Nf*c_xJ?uqCV$qLLFCL$oCI z!Y2edNHx`yg&d2s#GE4rEbKXAe*(;LG%-bdVw|6nSTNn}A}X~(8WPL_QcB!B4`t!J z;BkIriJ=BSc%H(?w3>IKNeNfL0*TRu`cy@~VOM@IuaQxgIOWXcVpt5wtvJ$*;ORys7#Q(*GcG^TO>lUbsmr)&jZb3O%oj7I0WGbREmjC%kXHZx zn-#dk*Gach2;&*@Qc3~0mC**<_!KsvlQKsJ@ha?&6j~#t*h`^A38j_8MNYO5RUrS&mi%l+DR!&R(aZZR9><*7z_X+2Zp;VESO|r4C+fa4gySYhI z*S*&?b=~{*_kT9>+~3QnJ3{%>fQ$lQ_q^A2-}*{M{cOFb<6*Ds*6^$j#YE_}9Z{9~ zMt;P{A+glU`jtmawOx7u%#^ZK{444kR~sFu9uz%E*ip!YTyx+u;w zm2ou&H%8GTUzb}?tK7OoH&)$>`;kOk&*^ zS-{-;SwLV9WfE(7Y_8f+Vtvm%5GgF1+lA)ZjV0Dpxjrw?W)iC(OJpiek6vbd*G3ZS zFL`-NCD!MWQT-O{)z z^=io*A)rPcsvg@`V~m`gAOr3P7o$GR6n#Am=AnLszmi`AtONKF-D@(Obb3D%m#q9h ztFO{(-fkUej?Y}wczzz62A~nH4dU>b9Y$pZy;H-tD)RTK1}{J=_9-KOArP$gF0Hg0 z?$PWiy)gV4SR&Jw3u%7oBI#PFkzo?VTrevhBV)TM5=Hl4Z55uoER_!Xx4gMd+m8Wk z!OL*L;Y3?OzjHZjd;P^cbF-l#pl-yc?nXFH5?;}ECYUvmk}2RA)ju;I#INA|K;=dj zkp;%lRF>7{;(8dz022}gB6S;3XVwkov74>UW2^uk+cB86V{noF$WVTO;3r-ylFYD6 zjl58m=c~vZQvyyHbwry2+j269J`mY~H^&=kdD4xItE1FCf?}}UygGSByy$`&E+&vt z+kS@}85|SD$7PVk`!2VL{80?=f54UEv^G#EbF1=d1u1D_oYL;CioB^MnP>q1y?xYE z!c=FE$a5neNi&=r8U%+b6yG_vYP1DMhp(D&dDGk73Im zfVfluf}=03+i+G=~OH;*{ReR|?1z6C}Z& zls8{5Kz%j^w3k+11Rola+F*PFovT-PNsFn6oD=d{p}sPpK*n{dAGwJG<5z5sU&wgX z8H+sQ(z3zW#EeZ!4GBLkkv(6X+Kk41D#jIZ$M)M=knmC~tF3+7g&3LBVZI zC;J59JPkBzI6#rdMHR{3JFTEnfC_k-+>k@l{75Yg1tP+Y5Z58;Y&^>|q5K)H&QZFl zP?ydG{u-8aeB ze%P2tX)YH=_|j7TVy?@0;7IcX7peMM&`2^Yl;TLfbD2cy0jCl215hW9v@{h36$(4! zD!`AgZ%dG0uaN81(?8eBCWZNx+qw;`rO9Y!67y(bXPz&jS z)>c*Qbt`Xp_AYM&LBzhytH*^eBD2oQS`wV~HpFIk2vUFQxb)Nd$n=*WlRbhH>rpGdRh_q&R^{@l28xbz=E^FBm4uJw z&Et7sx6Es1drUa77!O@vHR4Bt*KBZ+{4r5{)Z!Hb1;)Z zpo5WFp6Mg}x}4vq(mPB>1%csM8obnE3l_RJiz~9ELPr^MQ)pId!im#LLU-$&ph2FS z(PQ#?&CadMmjp@VQ|vbl3(<(Nazb4`tI6Lqc=9RuWO=t{-v`7~Gw;>p?V7x_;kh5w zS+-_TH$waEV9IdNfSI)&LcBX4-QB2&VWn)19LVLmC+n!))s2?qohqzRe> z3<9${lab>*kKG1JqdOYS$5(krFc+EUx^@iiZqVDBg2#-&d4FzI&I1L)13MT#rPz5& zba8ZWHQSY%p03;H>hf&etf@mCdZb9mQdZYJr99@hH1k6B{ihY7^Q3}vB0B-zSrvq@ z(;9YGG%BzOl&DFus&>a{OiXLELlrxCMH;SKlkIg1cw8fJLg)wpGi;WZ`q*~#ls0^! zDsH5l!Bn>io@)hu2}=RlrF6j`p+AL-1KVC7>zbrGAmD$J0HO@dAZ7{|>wX@SyStr$ z+N@(uaC%!r)d4vScSZK3Yh!ta>@Ipjd3-lz16q+UwV1Vq8-ligFP@;mOvGU4x(%B`eXrfZE`XUuaJfytb} z&xdntum^T#E&Gj5ZD1|ji)Mx9r`j9_3HlouRl+K46Bd)xT%vV26`{m+f+sZz_8d-I z2GM!DYI)P|?#R9Xdi|cDfsk5X)M6nm6_P++Oe;4)W-vywR!LcMx5-tJ0uKb#RvC z=P|lUHc^lChJ40+b_MS{K+~FHYUXO=tg?+0xflOl74ru0E0`t;s7J%4vBG{ej^&2OR&zKLrcGjMR=v@DIZGljFh*E|bw>;}KLo7*i~f{U;xWlb#iDmvjm zaM-T!>&)Cimoo#7w&q9+brk1XMU#tS^OKG;Wjt;_zoVDmj`r z6}h)j)7&TOpn3>|4;RDz2GDdO=+ptCS;^3gksc6pwg$- zht1xSCT%u$#<=-qTks**mQS60Lij@mi$SyG20v4@ClQNKu0c^0EeO1RDvnd8(WbSQ zwfi>7daj{A^`c;rY_d>w3`zrp!H^xvWJ^#Su-K@MqVf{&3sz{|f@Fuw*2xd}Y9*qr zl~Im#)L-Mb2!d8zL5x{21XR3T4mH&S>Qk5PDvHs^zi(5VjHQ$^{2m^xY+n)DApiO8 z?2&Y!<+a%X)nG`hHL4JbhX!X#yr62&sLIhK)#Gf0uwhn+Inw4yhlzn=4&_mBTh*|E zxt<9-su@8!IarJ4X?-z{#BQRQt`G0Uw>=b9a&|9Hy;sEI?72}wAlWygdMfk(K1@*V zLqfzH>P~nx$H`D6l%HvZL-j(+sXRK{>*-@v$rx=d>B=W^ugQ+VvG6#kibuhu zAQ`la@!235aVvdZt@%$*it#zhoNl+#+Dl2#ibPf>59a+xvLdlT_fdu5*VOW3$s5{U zM!->rC2>PtnG`_thDua1ud|3!ZHOTSTKArgWN>XPUU0wnKqO;oTX`2d->ryb%i2tH zZT8(Yk<6&=Yc9(6mPYce+V{Oz#sa-DSp==+V?PXHCQOv4r|Mn2iV(o4*0KxTg*M)MI)E*Uc zNw$A6{t2nQK&&s@KRcA?r3NoOKHDc8^sdyX`Suy~$M1^A2hMPUsa<5u z4h6bFQr|v*wX(|6wba+S(1)+?nm)VdeLBr`N1GJ=;Hx`bW0* zsWo%6eXz2h*xIoU=)Aw)U&%sswFS<+l0-X$OALv}3DfIVLsCe4d-f$smJ^L!FFcpZ%z3yrsJ2Njk z>j;*Oay24y6qeIpQDSa>SVMPvCEI_Q!gpb9Nf;8){pu4mH<>Uy*2+SN1IP#ZC zG&`Cfn=#CIZyWi0;%rA5Dw`je*HbeD5y7`dpfO<=hB&vVrHikvfp*>i3)m=y#RhK4 zMlGy1h!!_m!$zY$Y-@Ce9gRU@S7S&xxPkC$Aha5N;jl)3IJ^NGzA-u+)fgL&X^aoY zH714=8k>ZZHlG}B+F*IG201-5w+gpxY!hzXz>;Xdwj0}r(;GX6J2U_#H^`W5>=N$W zzFTqE#vbABjXlF}HuetpYM)iy=Q){0q;A)cNFarI^;t#v_mKkg9&$h_=q+FGwg0y7 zHa{tkl%EjX21iO-V1F0`C_}IpEvbfW((PFdVk?C$s5hQJY9eGFH2PQf^>2on4MfGy zG3B!;siKL`D4(Ok@_HxArOjCN9J2`(Ti{X&wLz%ONKeD%?01>b_!!c1Zi4c=AtZ43 zQac+LW{tFGSRyuuGKQ*kc%e}PWy|k6w#x^-v41wBrJwSpEI;`O zUl1L_`iqpGKnsB_lQ))eXC8wfIeLZ`)5~3J7Ep}6P#ZmtUEbrSK!*d|Bd^NB?SgoE8Z*Q zLkf&|7gJ~Txhkh^fqII{gUf(u5?39I(E>B7cDzV-N^D&b>EX^|*VOPD&k}5o9qsem}ZB*Yhg+g4v*~U;(%Z$joQi=nT z6S-36mNcZYkgs9$X$4n8v4U&BDgc&5?E~~VaK!cwDtG0FR0g+oR!QVB;YxeEtgy1u zQcsNcUjM#_f8S~aB-Ps%g;bhKM&kjUXeLwvB>*2~p*3Ptc)^pr#fV2c_n+kT`fc6+ zpI>=jkeKPuuN9kJE3_Que&!CdKWpSo0ek62dXP(^8ldeZ3L5s|fw?Rofyb79h%Cn9 zF^L?HZ(2U00-K*d)N}n4=%G*+XiORBzdgdqJcrw46xSk!QXbbh7vG$ygn10y3XA$d zV;?b+#dsd}+Yj(7T@~{jgH%!53#h}@rde^*8&A#Cmm0aqJA7cvud=9ykl~?4$^fRw z4ix9XSi}rOx!PO$xY@_e=xtPjjM%VPzF=M;46a-jlPRDQ{ER@tTQ-T79b<88lsSW@C`pUP% zAlgwCanGVBLUL+SGbDO5o9k%lMxE)kJq^&}JPdjV1#_vmi>hWr zHIk?t1vBcc1Q*7zU5FfYN;CYNa;1?z3R?`8$h6>}n7Y0^u2*)opA_L!K{xYL6dqVt zRk5ksO*VMAS(8zZL~bkJhE_E`M(CTa*R(`tQP7Nw>Zs~}m?(xLYg;K3S4P1jX&@EX zrDxm5O=3vY+pnW(WmmNm8n_#aI6iQQ>Wy0M2vE*T&^;gJw_Jf)-52|c$S;8)WVzTW ziNk$8%ujx9OTdkJuL%qsAs0tO+ivzIgai7v3C_efql!3L-q~N74$E%7sooq@!3=`{ zy_v#FQqz*4ubuT{@I~ucA?OuM1U-;hp~B-sX~oe@A>CM0l~&sEp0ge5-`8Y}uX_bTt}i~AIeM`0G6yC8GdakyEe9tD zl|PpFBgx)@>a!i9I3P);yY0<%Z1Bu&a!~TIgdZ`3jDH~d(41w;m9!kCmE44bE!VOb zBuH}KuTo+04}Niq=r{2}7R#HfbA<`PTe)}=ovaBY+pu!fR>?SI$Q1avb0Zyv{?z9H z-Ho+FSW9(b5?_>rE#%eYWH)Z4rVD4#y@*cO38L50r+t$mtf|Fizyqy zgnK)xd>Ey~Tlw-Rkpty(Q2FWHM3zyo#9XS`0VVf@vFmV7L;U8GL|)0;m-D>Kc^u$M zXPZ2b&s=zaWakl7KxKx|j#5+cwAfN$@!4FsE@%IXXirGpTL@V>-to-IAlRJsFGmU5 z&n9-XQbFuaYt9Bkc$o34wK!sRDG`sN3>HxH7WSpQB8NMJP++*yEY2!^j9r0CqYnau z_>^cqlJsizH#AzV{1X--|HOO;FLxUrTQgQD*ByJ4lu z*HxI~`UMlnf)0#}uIK771*jWU?yRNNc1!`=jx(cC#Sq9GNBuDCzQj)`w|-i+249qt zn@Vz9$=#~&xa*X-rfH>Rt2NJ+9IT8`Hkr6#DrxZ9V;AWR$=lYTZ^$mu&V}Hm62)HN z@G5)EN4e-@oPR+7l+kukBY}@0xDyJ&Eu5E)+{{72lA*93jRC@c{Q|5+vT^3BR&C;S z$#0{ziL6WNYm-vY*9Fs4Zh({GAc0@}PBQM(+vG2HBmPpLAK)qs@OXGfjiE&-`Ijz$ zhThN^_@oLg&!>WejE@mEbJ$Ypz>JVwH^g!YD^Cz!w7HqQ^mDHtI@0*uZvHCLyjs2a zSzIml9?7_ya;O20)fc=_PTtJB_ww>rO0}oIv|ljw<& z?@VKK6IV*mNKq(_yQ~C%dZ{SS7G<5!Z~$RoKoy>nZp@udkMYAitaab`c_cSg>@CEx zJ&?zXKB`EWxWy?`VVIv|O^QN6&Z@X`DspzkURaR}Ds~AO_YmZl71aJkRS40@{EDjm z=T(`Bsmw3PA1I<~C#qvjx=VwJCVb6_{AgL8DpTb4Sa8G`4oSTm4KG5`xusCNxL|%_ z{)0mGYDb}cq?Bxzn?W_lYM|(YGqXXCJe4zl$=Qc<_R(DUFqyp_Otm8G1)@5Vf`B^V z*#d6fNG`7BmT6S^vYK6j^+EKOt4e3NBVj>3A2a5{Q3EhW2h;QUl6|ja-Ywa;N@`Pb zyWB*~AKZfvXoS$n{WWQBP1MJ}ye^CDYCzT1Hzk+W^x(DZpnJX@+K&qmZuQB|uZ2jvQrXc8{#sme%z}4JHgwFOOi@(Wh z34T$Ni~GCR?r+&`9G}+Zf-T zZ>wR}C@?^V*ulAOhNdTn%58^}i%ADD4T!$bv22mDXq7e78D^BU%qMNvdMk~V2 zzrB@Lqqa4uDKodX)6I@{hrN8lu_}>vXJW1ZQtEs1z8A}|?33&pQbiX!+F!mC@6Q?} zoBL3Uq{@(5g7a&6gSjs(0I=z zbrMfj9$BG=Kq#H6{`NAd(N`+Li4YCh_*L$ps-dcRO+We+Z9*YaI8b6#y~rKZXzo}1 zXL3*yT;CEEq9gOxBXA*+^YZSbygfB9r;tB}+&`A;PGAcrgM#WpT9H?BQC^nvow+Dq zIeY!@oU_;F?d~)qNQ#&=Lm~*@5`Fa12>_g$&%FD&t!MR^) z5JYPG1a~%K0yZYq;xJlS22JIURaq2TIL$&eH7Q60-0`i79UW5?>CB>>L87vpUX-V+ z=D8~KIkjihmxRB1}bQp);kFb6Sv#zirFp?HIW9y+j!Jj!e%bem-7OwQTsPra z8k5aboy*{wT9cri$%T=bb316E#TdorO30*7DV0v9w~<;|okgT73Mp}J&CD%^nywC4 zfKO8XuN`#z*h zc-4jthzuR*S}?B-hYloTdi~R15$TFwSnZLzn0xGBaWjUjI&S1e3d`A}+T@l!6mxOP zA@aRZdYYl*gM84lE~OQ=LOjz{h zY&JpgZ8KJ#fVIk|j|PqiTceI}b8d`{_&x?7g>8h}>wNZgBsNp4G}KA0>)mR%N1Mn9 zmHsZ*@EQQy&2MBAH_0`(2R;Y1f`MOC)RAug#z5N7J1Mk8J_8ozqPx$9dPSs^=DsnKvmECuioiM`sIX94qn-* zd@XcxR)>z>qD}ypN|-pw==xM%Z27vtUC;*KnL(M>0amh{6IDSx6e6Caqslz8OHt93 zsq!5GQ+@$~(+}#^q)LgxmJltVRh{>HY4KLu1(L+rq4{3khqd)99W=$lhzi|I08W8+ z47@J-s{I=`KRfB&QZTOzxJ60s(x-YwD;M4CpXgBvRXQqWL1l@#sny=n%Jb!27vq>< zl@I4OY-Y|mG;bWS+2y!l`2agR4f?u+*SbZmV9iB=S+OrueoyQ z;V$T_&l7Wmg!9B6A;!O3eN zm0fmm7q;UKT|7!n9_!*UKW5ucb`s9u2rqW(5dL?Uys2NFK4_yaKlfiAJIH>U!Ae!; z%Qw_vq4g}75>PuKW40q4-8O7rKN)SY-4B*x9;Xe%FR(SQ@x0>gP>q*K@Yx^)n~R1c z-A1BbYiCn;g#NcM9cAPw-RK8U86Ajng=c-F9fkS^5qqU(!*5Eqtl!v9ud479ITtN7 z2u9<429WF%MBG>TjXb|ZUEtF@_jR81IVkm_yI7&9(^ znPzxt*(=|C%Z=KCHPm230jBvBu7i-|-?bZU5mZZ=iw|a}SuC>+W<)GGYrBrS+FdJU z+6o2b7XAB-6`*stc=i-{$H_Ag)=kJaKhvpUw}-9`jreRQhj-|4hKABjaU)xg=8P=OncI*}_u0@C=?02_UkI>;y7McX#0%>dTrxOKF*5 zoAn$bb3`Qbl*KrEQWWk_c!PV>x#ys_1;ICg3vLQYb2xHiS;wtP{Ze!>MnUHe5?Dt; z*C1qap~0F$3QQw&GLTxY)L$N2D^*!w21rIyE(eEJJO$XPI!FcIkJOh6&Y^TF5SUiz zTiZQfG~jWS2Pht3t~T~KT`~>D0+%*3_+C_fqZ#1`n1^s5QtC{>cPNOw=odJ(u-2*8 z?kDp_rOdRdr;kvFFrCV9AoJvV=?}RY&;?jn2VKdB#uhXcewx+~3Agh@o~4x(HfVM> z{1zfUDvX9HiQ&M%Lcj>nRWyc)NA+Avomq9TA7B>2H|*e0 z3r`FJI1V;Td%4y0DE=!WM%8l)5l_{87UT-mKymvb=uNS%y;+)?1}tD|WCvK1G5eD& z8Om(Jb`8Nnb)?8YmR?c-SdqVg`bYuYi`o$C{DFC)MVS&!V6y&{$*Rb$PAVxKTG+NS zwX|&+4YbU}EkWj`sbhnq6v!*`EBO!mrBsf^cS;!oE`nd5%9t=_Fic$k6DJ3hDRhOE z#@-nr3o$>#xRQ%62b9(+YQOYe2*P8K+fM#f{u89vR~7tD@!`H-;s5k*&VRu<>pd41 zZv=_(+F;NJsE%3%&Xl$6jVU1wOLgR>bn_tHM^d#Q=+6y`x=M$dAKOFa$1yMcZ0~hF z(}Q0|75fOLIvnWX2u}v!ywJQG0=OaI|1_EI93i)WBZKtw;m~&1h_L0{M+u-%9;w6` zoaX=R&)3^AN74W4l?5Qg@MZHOR-;GBD%VatXgbYXB(;SK&1qL>S^wnVJ(@w4K0|>q zw1+Alb_lUfvO{L>_*%wB)uSe#9(@@L0koUAj)tIu+s`8O>&(qE_h%Dl$< zn=zH*ZZauPl;BQMrqsMij`wjWb5e*I>ofCVCTuIapqR8nCr#7yO4=SEuY#g!(|nKV zWj)n@qwmi$={Zbw?1qCdl1#U8Blsn=0I_wf*7wD0V-ld18M-sx8I?le+xdn&l7xwH!fQjgD zu61@`-K3D@bsk?;rUHRA9v+LnsntRRxsn=fyfnj*>fW`OnamccKqg#dHeYCDkvUNo zGNk>xDx--cuJ|SsN+v}s$M$maxI~6Cn#177@kq*igLNZao}mF@lq#N><|7+;sDHLU zVd~akAg|(7_sMY78SDZPT9saN1q9FW;8}{Xm2BJAeQZMm|w_ZQ>aX=mNCzuJZ-02tLQX4U1qwP5Dz@=&t|*M#OcG2h1$BAx--Q`Lgkt$B<5McLS|dWtyC zJYJL}l=N`RXd>XnRGqYPs~BoIwA*t_4K!XY>9p2ssFTfYLlp#(;89Fs+lMJ6iyR)f(Y$e}L0ocr>h(3a=tsCbp-xIm-`TSrr-`ewGw4B4j zbkrzdVJYA257l|~vP)ibZj9XP4?UXtg@3ar**SZXoz_1I7PeZEXc}}fPmf%oD~d5D z2;KplW?^g62;14}P#E}J8KKX*>)|lgsMqvp&g?*kT4*ojnA6Al+M|1 zGslFaU`$0MsdijKXLtqk3ys_1B#fWW%NzOV^?dvmMdcvjtM%6Mey+o~f1|@M24#>+ zh47H374}IRGPKGkDdh2BK9xY7y(ed$qa|W+Ta)7od8EX0V-%<r1y9oK^$=WQx3a214^1W2?9y~HVakv^dT}Tv?X=bf(P6 zAxSHIMl;hHL|5a%Z0l!krkzpdyKREGHDpdZHoNNbEdOjy6?>%E&9!jGa#RU3+Pz(r zcZ#S*Bwb=5^%HDAqbz5Z0RZNh)`x_Dt)S_hS4n>)J+d{_v$!k-lu-ro4eh1psvFt3 z^C3PSQfbX#=MH>7ntiM?^mT-Pwe(OGC2|x@8V!+a(Qn!^dWa<*EeisCQP%64y<~68 zI+@KjjFYi+Si~AUTlr5lQ9r1bKGy6UT-o-MvVDg7>9m*7QV;&de0Y~O`kgY8P9q^E z#rV`K93!$b)e#0d{X*KU{;(s|3pE*60hYIN%{Kc53!4SO2wF%H3WOu$6;{|{WLaRa zlpD-;2W*u1zAfX!kfK_Cvya9_GdsaI+TY=I^&$ge8JsN`lmdfN3O*xJjmqXb1@r6a zn)ds*?RhnVG}C#m+baS_U_be`{863E+OxN)ZmWOH7Ay<>u2=nXs!#BECLDvaknAOT z_;0n+7sxs#DG1Dx&2&uFo?r2f*U|2K16jr)G@xsOP<(|y#hJmBFeA3$$qI=DW~pjj z-Or6lEAU+P?-v}^h$s${wFG({i7tsO#m<4SIZ8eVKPBv+Qqm#Lh{0(bHaJ08RRshU|` zlRwt*e>=IVCfmz)EQ4O5>FP>IT#p)yb~F4!n>#89@*9;O`^$YLxxWPQLA_LI-2{wV z!~CHnHW$dGMo`A(o<*#*lj&+mdiMqX6VXv>IZcXZGP>sfmIXwva%}CRW z@?TT}j9m-LCRKTrZQ{&n_JXtzxH(th#3C?VbrsC$+=U`DF`pVfJ@ZeEgb8Y;_$Pan4 zN%$Z;Fr?8mSSL3}5hBxJ?u@i@R3s9ys+*T05Po0%?z{BigONPMl>zB} zb^Kta@>hF%0}$9J{1)^$%IrXgvY|Bb^B`cLVGZ;+jv*WsYV$yy2nOgiL*Uj_OC1#4 zuMtleHEDJOx~_#pFq#rHugvG>3x#-}n_pARZ*6fnv`trxpjx>IZT46OhlQ+v7a(9j*#?NUv zFSCL&Puf+-)y;yscGr!uJP}z^qjZ(x+}c!Di5yiiuM)tIUMsqni&^zl^AaE|fSF2k zU07)aw~L(ACMUP)dq!qQvqA9qoA%aL`&nYIEt>XrxGJ;JkdVwmG{FHN__wourdczK zF{|rLbz?#Z3^}agNi}B5?uU?@R!367z)>VI!Ka!(r_Hp+4MLN?z|K$)%FlPCogiNt z>i?XJ>Hek2T(irUDnzMjLr0NKsgPsfhs?pag+W8Yv-RWP(T-?-*#IKz7ruf|tE-5uDV~v+@*hD;2hFW;}bnu07l|%IP9k8m9dVo-HLNmD&Cg zbSF1NrCP3qDTi|lFa3`LBhz-m=c3kwv17^j6svSJHvJl{umUL&;9Jla?Ta<}bdY>9 z$d~al)MGju_GGS0WKB)!<#^j9gzKy3Y{Y7GPT8GRmIY-wqioJB(+XA6XcU)PTDC`* zQM`6mwE7kucNK& ziDOBhV<0%9vR9>^R@H$TRaz+u zf;1kLPrt+MP^UT7TP2Uj$2a8v{C&&JPvqZ296z@S@5yblQ!tODO6jJPjXN?-a1-5Q z%?nPRck+^x7u9aSE+r8{(;?n)?5fU2hpVh3_@B-_ECw5p9b`wGd)P{ z$&|}vqA`>Al^!r=1cLCZKMS3Q@@i<_z@OKiSBaFR@KE)N3 z8&LvFKv*{Guj}cNAown*jD{H{3EvtqShW&{TlcIeX6cnjXi0g z4pFxyWTK=Jwp(3hy90X_;=mEtwG(~5NGPEY(dF^RU9JwdTg43wg7e#SC&x3I81J|9 zzMHclVo^;p$K-8eST_A}8@*)!aVuN=h`KcdoAkIL1om(wj{s3&oo88?b0;cb`)q5l zI*7TQXPI#O*->Z(JLXzoQxRDz<6+suEOlXdj6rgwGnY7|%@$gNhe^E{7r2pmMlJz5 z$na@t=caSS%3&%ndfeU`G^A}Rfe<$9|LA^5IO{1qD48zgdy$N5h1!x7#mVr<@f6;W z!4w41--7QGsx6psrP+Rkp)mDApwUG-vDJ49T}d~se-8&b(RzC39J;mtR;k=3J-+)! z9_Z%Cff?fUTJ)!@a^MM9qg^P=Q;J%in5&bp+kE28I!a{-2E2JNF%KnfDqAd1cw$aY z$O-R`5<_xh2zr{vqlE6K0WkAGC=AKTih3`~An#2>NVKjE?PE6gxOFdZZFw=YuY|*m z#Vx15y)mZ_I^FUZP5c~JWTuU3I5@`ORH@mAg~>*kO`E|R%0lQt3XsBiFh2=0nHcR& zyG#REo-EOx%B|S#{%Elu1>JK$e!#uzJ%A@i4sm#O8Chd{G|0QYRD}jTH)#ex?Ezs% zwDk1QzLsHq!D}gwwO4cIwVdWZYgG~n@#e>5`x{G9=OZYI$n-OEnEizu?-2^b6m`W{ z#k{M(E&@lo`FhgXPf2HY{C%EuzN3VLnn1gW4RPr{KY6l6|9ZS;HqO6Oi-gD&%K&Ea zsRFGmQUq1*5m~KM%1(7;#Xo-vSPK&V`;t^wr`azLJ-ex4Q|9&7a}y}n7mn+mtza^bU|*Sa*((y8BqsnxFGiy zWDQ5_(am-9XF!K(=0jau2y1dyUQW)3C*Y=*J-(xeH@$`mtXzte%o5bn3pmITMI>&w4LVPO1pRCIm zkQZbh0#s(76Kj~`q+pRlrh8hh6J)eC^%2Z)=+qbSDycW3x9J}rrfZ?h-ieX`o=dKu z&U8A`lK1sIF*{=BOEx?VOza2AetnRN!TX9i{3Mnat-NHL@1Mo;w^+U)NhU8Zqg_-<4cR>5AL5fWMP*(gRAGhyW|*km&!C2kbym}%R22!?h}Za{>T zZWAcB;NOwsEy0CF^}@?N`8*jP_vTCYqYugh`NBi``EWk}Xr8gNVFQJ28n|3Oa+pXM zwBYIrSoir@3m(k=&H<{csmOUP?dg$5-Hh8@`D%&S5ml{~9@bhVYwWuwx2ZmT+xYPH zL^CYQbrehKJ4&Vd`Hp3?MYC7@+@WKFMgpS z^qr!tEmGS+Zc)$7=SBBL5r4}^Mc|Sp0euj0jP2V>`Q^G#A8nz1gilGup4w(gOo}$i zT9q#$e*B*U_VAp3Qz)2qhxih0@XwBy<~b5n}m!}Nl1 zAW^0lvzzYd432K#SGk~K79nDIg2bIsDJ@`|!Z(*zgrlXZx`M21i?_|Oj;D2O>GUVx zhB!NpL7mHkxJCN`F*nI*Kk_%p=JSx?{swIP#fo`FZ<5#tE2T%+vj0^PjJD$4bd?LMeA^Zk<$8=$~CgPaD5&nn*&3G|JJ*lP&KDgKG1$tDSfK< zsLF9wIlYRL}#&dmK{m5b;g`B+5C5G zpZh{rX>d_%bZ!|%*7i~G5w*kE6I$R-EYWtCuX3n#GQVbyt;t8_jd{_>AWP-ChP|P| zENAlhK5(+5f;Y;r>cLPaN0BYgG7<4JfM`(50KWz5S5u-9BC=JmWHW;~(WVx5;{2<1 zTr@uC_w|^aEWtX>Q##66viXk2bmFS`100|^=3z);5UZ6v5=Ac(<%61qCE=DTaWC_q zGvqC?uZsjSGjKDtO>jkwKy=!WEL&_;i4z=Lnof1LkR^)e^!TDrmHz3$ z8Z|vh&LDx$m3~W{*vVqtq&%K+H@IReZb`CSdEpVqtNV)$oWX~vQ=nnK#nkgWGH>@) zA8?PKbrrlk+_B;B7zT5ae^zYV$hafwEC#Cy#xPMSI5QYPr}~?9aR6PN*w3>fC>u6l zhT6ZV8lItd!O#aTlf~K1WZMBicu%9ai0u0L;En9s)Uh8dw?vi%;5#jmZL>y~y)*LR z4|#j}5P8p-J0dWpLS5!#sFfet5j;c=&yUlm$?ol$FzTLqm4WmrbHbTHDg}?WP7YkF z0ft`x9}t}j-wH}p7kZMxmcHSF_)WUDf%55un}O8>sgvjEy2O&ab%O+q`zSk ztT~j=KlS%_mK1(~kCn?hNk#&BUwJmI@ zMv0o4^*BPEJE})^_6$fTg|`k|Hyoigq<^t8z zWJcH<0-s7rB&XR}E0`ZYL7I6zm7w6JvV+gx&zWyf+zJM*E%??z3y6FYy4YEtYUFlI zmDWv5lPkfQX#}VfNuYKNT`AAo(o3q{8mW7_ab~E}O)EsAnFcFmiiQE^%nVi&Wmm@- ze`*RBnZjx&MXVH>xGr3`8K2!sxq1ga5AV!%R^bbCr3%<_&Y4r+7D+o>e}vS8SH*V7MGIONQCwagn@nmtGjROLM?0pUQZ zuYgD18qBAH_BlpD8xt-v4+3VBMW*cEqlH-aa~j!n&;_+&ax^PXHDN6}^Bd>NA+wiC=m>VW9wal24GGdo<$K4_Jz1UKj73e{8^n&|sOchW&&ZmxWT;A|TTN@8>D&z{fj+d6O2IXnq zVFCl6VS1MtTC^cuahF}f0Wp+BK2TI-9b5FEu!S;1lzqprYa8RE#g=eu)Ex~qy{0GX zi-yJhoHwF0n)flZi!>r8MkgB61aFuyLH;3luMmyKH-@9m6wx)+u&EN!j-h}W57CI) zy|1i;;>|3QL#Wo6o=_vuQ`1JcGjyL#fxVTQ_NJK}i5w-A$-csZq?#KdfH$g^L_`Qe z50%v9ghjHacLZdhRon!HU`}Ns;e5&_jQ$k)BJwF^%_Q{O#=p1v_lhpR!&%dUFR)&i z05!DUPPr?#zKo?Tp}q?NmtV?W^voWtt)z~kcD*jxmp8Pov{zPr=-t6K~4oWU5i<-0^4|?bU*nT7>+HvX;|%%x7q@$hT^D@=#HJ_dgtd zM1d0I!68Jl;oMTygUy%(UaU-2(#(tmi;BmV8jG~m@5P#%#~S4&Z53i&O)R>g8a!T! zA~_*%PR`r;WzENaHI=tOHTij<@~du8Oo5kaQkCkD2t)6$^#xQ^3!ctP6L1`_c68M_ z`3d^aZX$BPn|iueW_z%9g{K|LuTEs#{drmEq<))7n==zj!8@AT%fr&hKNRIMst>YB zKVMaJpqE(V(NKLsZqCa!Wx1g&WzdE$d_>fQS)Cam`v6aH?-|j&@`ON4VwYnfn?ICg zC2| z7s3mmS2`Meo{|GE)mTev9Vw5_Alm+}>RwiFD1u8JxDN>@<~9*)%kg zpdfr?Pk1;)RMMW~RS*7B#s7yK-tZ}pp&el%eU72#{>1Ig2Ss^b*E1$Qfghp^AAFcC zJRO>T4Q?wc+_y{W1T*N=ykI~;@os1RaJDK@7QO$KeNuZVqp_1|Bmf&OB zlKas{O*w=SqQ-+;?90X9xlxNwBm=lC&uG{GwW5on(ANy>ysG@UL}xAb1-8rbN=ZJb z$fp(JWRTf<|JiN^xje{>;YqyXEVf%Ca(rR7>`8R`A^-GgeB3}iWN=1Q2M`n)z53We zOEAYPwi`W67+h7eXZ!N*I16ai`Biy)!&c1&DOZ&7{nBC0Xn<6mz$a*wnAE%KC5SfESu^5IEK$%t|* zxM6tExcQ)d0=9Fyl9bPZguMX#rx@a%eX+_#+Gd*9Z9 zPZGzO&zClscR6!OLkKNi-5``pi6)ZlO!8p%&FY3c+>l2aJY@?US2e*g*vFgVXbq^0 z>$brTS0abxA$HrU3{ID5RrMM0V6I}yu|CZNKMc9x4a$lxz`laRX|E7Jpp8^=fKH(< zrYS?nKdRDknRd;sK(U(Op(&ks7@Gy~Q|66M%0)8C3&YgE+l5HcH)OZq8E(LQMz(M% zb>%9ouT9a*VDq?_zcsQ)L#6zq(vls6{2dCU5m#G=?h&i8EYC^v2#z=8AA$?Xlkl=e zT`v5N2WDvkB_doD9tVY_Rs!>653XkM~Tf8o5sp zY{zJ6?&MqedvAKJM49UeOA9hfll%4&)|oQjP}p$ z{pfZzM^k3NzQbnbW*A=E%#3E_gF7{4RoryDV@}I#b6rggV-&W_ZEyC^eV3<;$o2p^ z0`zfzA8Da%tCZ^Bao^U@>Ynxx>eX!Hmq^H->eotwW5H!HNz}ghjXtAtTe0~l2SP5V zMc6n(;9q5eG6!2Pg8qeEs=Dr+*ih}vx$+ebBL~Obw@OX!1b2uT;YLmiR->?E;b>|8ggrDPVu5r%)ZWw@`3~WvtBR>d7!E`Ah?tl5!y@^?LOOHb zmbpGxXHWv(-8Q$?I~22Tn)@S_KDUr5Bn>{MxjY~8X!OdpCOL-UzH+W+>^vEH7m-;- z;0oG#LHLfqOyV@hO-S@Y}YJ-Oza^+pqO-E^wgOm3sC8h znV|(3k}CHii)SRQpEeYA*)?C9qPy=l!6NT^iIWQ2Es{L;GkX$KM$y(QB}-)a(F!bE z>0Uu6?OX7pUK-gxf~ge=pS6*;d91a-0>~qC*nY!ExAtpgZ3zD^v?N^Wq(Lv# z$9{Jp|L0?#Ai-j$qY~;ojFo-l0)6aA_9*Qx>x`WyyU6aCi`~qNZis%lb$8>bbCl_w z%gDEzaBIdfBe0wKAz&nTC{_;#RY>Tyh&Sl>l={#qrn!JMtxeGUPKph>%QZC1XE;4D zdm59q%u83r*|*$$&S9}l+Jgm}fobA*-PfM*zqNR{0bPKih6bl~u8P=&rx z(0FR6u##{POf;jlbrE%;#%iSjur?btb1>!(^-AJ(?gg_}*H|9n-+$=e>;8Q||Nb^# zr%)!HPh_2eN!b#=+7=zfb`jPuBbr%8Q`B8$nZ!B3QlPTImjR>#W2Rlhz>KpX&4|kL z8TqRv+F5*=0<4m8-BaD^@u*F>VTd?tv!^^@{A%5B?Qx@buvC=*!K5SKGxw~AT&eNAqtS>B)0 zz#e$?$8s!TS&+*yqRqKovaCz)=#)Ek?~YEc@)|h^2G2+j zxiwGc@Ujs@HGA8aMOKlrBUe#e0a*C0oh%3Mo=nW26Owl223FXR)3mt5HjwzuopMvB z_3ude|C9^0_MlvvWRL$-^GC7o=M7=c+pWE-jDWeaBv&x+)GI?M@^uUrNFVElw$KYc zXsp@nY8WGF-HVzbXg>VkjJ*k*TveI(f1h*iQhQZ*SM`>3(%soZv+pXTkq`pJ7*P}j zLli|(L17$qoEZra5s@`4qCf~CEJ9>6Ao~s~vI&SFO9BK$L>6Tb6v_Mjom&Zs?|nFpCRJ$pU8$@iZ_7DI4k=Pj_I4@8#~o*i41U`?%^vBu-{3qD z`f)7#MHpX;8(e%o2R ziWv6ymp|xs=@s=VkDm$WpHwqqxYxTUIN!kJadQ`W#Z(pkn`2CfpKPP@Nqeva~S&lo62D$rlsk zgYF=3j5C32s{<)9PVy5ibk`99h@dzHlwx==Njh-CJZ&=qn~AURM5}u;t%RmZ8iyW5 zS5*lKfPapEDfWMe;} zLDF(3kCaE&{+8e1TiLedXbFCv^lb%oV8;4cnI_9_2S&1^HkcTdT#D{OQFcWvSiUT-X3%pR3iUv#^v#6hm8(yb0HX1=O~v6})FLa#-oXz!z9 zL7s@3uXX~$6cA?X^MCyY`M=p1)GP;dq-IILUI#IYF0R`-gLNbcFe>i8@= zv^Tm3q`}N0a>%zq8#QjsFeJ$DvtS1{kjIC6=;V)WzdG_R*5{3ca4BId@^vGw% z|KcdEgw0%-*X1__d@a@k8Cr5LLOoZ!)yU&Q$G~C5bN?;u=0m?l*?+95&^{@>Tm@yG z6vfR!nl3f2#dffqTs1N`y7>kG zqGteBVd#p|>DZL>HNkOAu()oLJE-B1c0hmzfMhHo)J)PKH|AX;MW!U9MbS;-=5ij6 z&ct)bhL>|*TKWhVClArZfEnOw6Y#tj36WY@7ndK#z1C#aNsqFqHlX0 zy&tCA6IV&OLc)lbN#IHe)zbRN8dQib+u)SeYS*wz?~^`wXz?Hi4^iccZ7TVpo*$C@ z0Kb~mI4J%s@s6K|yb`gR(DRw|-CO)!-JT z(@!zFiqHw^>|$2{lJZ#q^6rgJ0`v(0-Oos(L(?{h8qs5EdZgUD?+u^UI2u0SiOY<6 z0^xHU(i{iDz|P?9Z8vZQpFG1yhql)* z+0D+xmhlMSz$E!%MQ<2 zE|EdT>?w`r-PA8^y0uD0cqWYHW!&3^x4Qf@I<+>Oda<9REr_Ml2igVq407^g*|X#m5alwX09(8z|fp;@6X6fH71@$vT-2~HZ0N# zx0{@*shtu)d2~f*meq_eh(vR!(hk)@IGPSoytLRZpR%Hu{@UBU7(^gQTpO(CKU)Hm z5JyV>LTz#u{UxkwQxE)x{d*(&WjWnKT6r0LYJo+OWVeX z1KEO{Yw%rw#B|FXF#|Dem9Qsdr=!k$fx&&w9T-e{0!#kc-j-i?vK{C*xN3zvW(QWI z#T9~Df0o57v*c$Poy|R+km!Qs+z+KS-CiW{Qcc&XdiJ6?jJzX&&Hp2FMCtxH*@*jN zPvi?HQLHU?Z{V`G@qW;~p2lw~*uCI>nyU8O1!?&Tir69GotsrIlRad-y^dy6OuDzbrgws$orFIhgQsor&60^FN~*>&1Y~HRoo`BRI3;8 z?g%aCARH(ek=Qr31yGO#Be;uKm5UqeKtU%_qxL`)yUVS(R4FQHiOgj(OzUGSV?-*s zhl(^69r+Z7*B$R(KUY=ucMp6W3$jgbOGe`f)u%Wyt!xe-kz;}0!I<{_qnzWA}K-MmhqF??O%iZy1RTEf@ z&`&HyN~K~K@5_iWkH`xFK_!AINOVdGXT@31I$WoK-*Rs>!6PL@=AtV8S2i97{HG8p zX{K3?|5kK=FWMhBmRH>iWF^s7-W^{1|Lp8EhB?_c^V(s-K>WYL&tgWT02gj`& zujJ~b)>KHu)QQn+HJM9mjks^U*M*cvS+0{H#w@ilqQ6ti`7&?X75%+del^SG+(*mM z0!S>JR!t0A#F}uxrW*uVd1p!BmX)>?O&uFuRZl-ku9i&NazFzWhDb!IriG{~-bq=1 zel$7+3(RIfsCJ0%tfv>HwgMdvP62NS3;?MkwQEx{3^K}I;sc1LhT(=Q$Y3DQvP1$TTrw^y?j>a7tUIt7GBK5I$DG;fUiD$S zFpSftTvsL>29G37Sy(mOl#RC9?Z3)lT7STWLtzzuD(*qu1!@SzIfd=W+WTfw`+aBo zeY1T;?qLs#(o@2b2Za(hvmG0}vc-saRJV5S2hQzW*vXP2(eiqKigV$!0`Cdo#~fKx z1UmxLE%sII;`Ywoct`g~%R94iRABG5zgF$ds{cW!ohC1~bN}psz?_ZuPgVEXb0>E~ zd@$zQGlB!II(vj~@o%2pPf>s-Pu&Kn>h1BS_Q|g`?fUKkYnEk2l zm(~5-MFrBnMxP6QKN+OR+?KxS_;5|#e$ccNn|4CeT^zgRSY3U)-uKgWs#u&}tDRl5 z^O|mX)1BY63!6AePU))sxT}b$$kzNL$(nKPk-Fo>Cz9Kn6fKcF&4*P8FkgVK)MBK0 zgkbGl|8}Qa*kvbmxi5Lj{xkjMDvuGnBRkxp4*P3uBYxuED98V>_%;_WQD|V!Nv3;{7f1Q8?)Y$GB2y%mDLI$0stYQi#H zUrf6rZD!L>cjLjsOJ9aB^Cs^>|Am)r%sTBfdtZzD}>6+ zIB0w_fv$1`(fuvzg8&}XmV!Gna|<&jqd%f!j;qnD-ky}`NoV`JoAx;I8FjQ@l_V6x z9bL6!N%DuToYl0`6z7Jv2R{j)BCM6%k674=c4~vzkY2m1fe9Ox$KJrM?EY4Bw_}E{ zx(nFW>4jx?K1W+tw)4tvSs4}StVaDpsVF}$yStU%QS_UvPcvN+(oE}a9@5DBLs4K* z;tI$+3bM>PSUJH#fGOFE<~|6;R_<5jijJa0#zsynqBpFvkR zvCL!+TWJ!zpazxJuh{CK{&~5~Bg%hcka-5Ff3Op%5Gbs<+kk?!TfiVO$M!<{3^oix zj5T05$}Dk{AgQ)#QF$ZjA`R4!_{BCAtVD0AqQG60SnPTaSp#y}gtaU&ko+%q>o3_%^Y?BI(}m zm8-7k8CYy|5w-StFB^&6I06Mo`XgLg(*LcHxS+s~s*>_lr4nlS>uEJvgi4XrzMW#- zE&RuaqBHwK;4U899(P=*(0Caw1}K$7GJj{4m`^ZsS;Kvo8>Gk7+|fdrqiWUhEb(S6 z0m}e4Xj7j$Hh4wIuuiB+@j0Egq|=lA(-g8lON~bMg!3mB?9$%!;)Y#>vJGjuxYwUj zAiO*tjK5#=Z&kD3b-2emhEQdbuJz0lRswVTD7+H!C@in{r=p=qj$`Ur*`PrLaXGsD zAhj0{!66%K%TTx1NK}R66MpKt3fRIj5PKUqV4D)S=$Fb0&i8u$yzX6WPxabr_>gAc z*7R~i;>n=1z}dUNy2prXZ+b))P5oH3Xbcn?9)dUqa_d>w^%H;(Yd#bwDm*y_R1PMd zu~eTE^Ka@3V|9fOMAvWy(rtk!^gN8h50JPl!;zXgWnJEwz6Obd6M16x2Gl ze)O*c*Gb8{<`}!T&SnSpMaK*%UIXGwyo-a0LCc_M&T)+Gj+Rqsl7B#<%g+uVmTJJ@ z0J|-9M+(AHcAY)m5jXmkpFsf-Pc6}SZQO4dpsHaSuM zYM6Ubp&IMwp6^!IvGLrS(Y+ha{ijX8tO=XtHXOnDZg*U_J9U7W-#B2C-4=| zf$j5OOu~-O^(EiYj2<1Ac2a2Pdnj%GVkK(-V1NnvGsdy4g?gCr#KW{4np239|9Q(_ z-?Hl{cDDi|e1rw1t+`9NslvnjG<^XNo&4?WgIja)+6VoDdY z1ilp#Ob1aD9MOPN%Q;B?W$LwP`FJFFZH0n8)gI}S1NKBd(#HnunK06q$5CrIFZ)9g z!ghEGTak?c8{uwra?l1IC3$SBd~NnQcObg2$}o1@U}0u{rVanSu~rP`#Txx~Ha?5q z8cc8M0g$)$*ex9G_8xm~tmU;A-xzCG_7|RZ{`Wohc>W>8#Oz8+rTQ~*7h@;9zQ-2L zVHE3%H+qUU{%x;#66ls=-YyRC^+AsZXorqAqfN%7A%+~80g5irq~<@G9_ zL@u(SqN%w?;A_eFTu1O7Sc!*VK&v!}P|osF?$0HkA1}mtZBQkCgq?E&b`FGR7oL-# z_LcqxQ|UgO6x-OEAOAOO?TT(5f(_Pet?s>a2y8uiP&Yk!2ttdUTg1{mKgM1h1D=-L z&wKq9y>_`o&MSK3H2F!dU(pNO$uYE-P~-=OatytH(C!-y7<$j3JshsNdWbnS{lEFo zCWk<@a$tU4_OJ|KqMHdx4QWpmjbt19cWrQB*v8Thgsft~P^^yfxzw`U!t4^l?kQW0 z7VUea&@Q#zHULkp4T^PSj54ek#)#1fB)k^gIV=xFAmRP){67EVKKl_tn~q~_M~-nv zjd4F1hKyY^Y!8DW+O$;eE~K}|{+8H12zRt|huSpj;y$~G-?B4T^u=3-e7Jm`b2Lxn zBU~Irgvb@D8GcBD)O$pF9kJ>pmy10M{;-)zH^`KBv(ER*{>z!sd!(s9T0DX5pQ7Et zIQ^Pomhl%Fk$-4d90$Rbl4lSEz5+R*cen2D{8WwTJ@_i|bRqs`@lC~sPYu~88*cS& zWN7Ya-#y>``KgTvLVhZTw*MCz(G@M6NQy~SE(i7u#dqwiac;>tyK6#@H+K#*iGVku z0_sB(tOJMkb?r~D9k!ou@af$X2uo7d6NJ#7X@6(+u&vtQJFiTjb2(}B?N1*Zwg>W0 z?;W<^4hu+cO+fYD5&J|d{#G7RY3`K3J{cj)4JFL9t9b?9h4-S2df!sX6F#RbLU>Qd z$?%?lg~B9a$ef>+4q09TqH>~tT8fDzaOLA-0C;PT)k%<2+Com~`8Ts0{k&%J&S-kdQh zZpEYVC(0!ZQ8|vUV!lzdwYwZ*C=7cRAgWRVXv_mC9qK3+h?Sk*!JuoQ1He*+mb`g= zL$TH!kj@0Rs(M$~c_cZ=>t$qcSXkKNxE~}eVj`4Bl^RWI`U0idw z2r^42i=ertL}k$7+)0|@C6lr4F`%S{MN9)oY`XTTVzM~{;z4{`fklS)Llc9YjP+V~ zhjn*sM1P+QTc>B2DvD8O{qk`A>nC$7al`8-YhO+#nRk2$UF=0%)3O03C3<<1_jnG; zSqp8~TyU*K^S>6YK^S1aq`o&EC@6&)HyK`_lz0bL#&0XvIDQc|4j3jHr{>0}+Kf_i zJCo_LB3D=2B3or7x)xTUX3no|lWx5=0gIc5aFPkpy-7ONO=5_R_)v@-N-d@}8@X`ay#Vx?%YiX^4A=6TKs>jz?c;Ox)s(L;#lp+f)#X0; zDbJ7%h-L_gcMVj`O-W_?#ClWa$w9~LM!cC9Q(!-@OQvQCB982AmYp34Pm>+F0bxv$ z*XpoTw6kiBf;m)MD9f0*)HWAjY$4I>v4aFKJuTpIwv~CEK|T#-Z(l$TEQ*6ub}--(#JFtQi(jW*Qp1Gd5~W zV2%HBu)`0A=O5ZySC8gj>!N$JtU+8<>ibY4 z@>jd`MzXhQ-+?uX5)&*pEvG z8Y+>P5bue;3JG6V#v_K5#{^Q$uTgz{DEcKfI{h-;-gbz)ahHi!mXdn##rNZ20d*9Y zJQSk5|2+hA|00j|?g($@u8Om0{(~MO^{MFOQel}%ZpO!e4Rl;_1Z-bu?#o0&kgg{? zI(nHZOydB;M4Far{%6Jb!lGSJbSsH?iYG`MU_dEB-`{fIj9(9V#*4^ANRvkhO^3b6 z$nZef9xP+x?aOWzpc1>uIeJL&{IqjVApw#Uy4u-unAH`r0^UB6`D>u({3`%qXA6un z?+M|=E2?P4R43yl<+2}SF_j9TS_u9QZybL`D4bKE`D@<&+#8s8gC~2Q()cT?BGtP3 zw|#K@+GlMN!>Q=Qn@jeJ+xYWYwwHa-wv@cdZ+7OFgvaWeIFDF+3?@Wje;vl>eiaudE?Qd3U6@V?(@m9DkRlKH1M@oe|zL7yh3I0f?f5(5Za^Z zO+T0rU?@MF^UFZ(T_8%VE(73z7DFTtG@}z87n8BX@bXr|W$9($Ips_o&4i#E3cUhb z)o~IBY&qbswcx-7-_sDTQzo#N)Zil6O~z1e*d=qysme7f@1bgO+YkF4zs;~8a>FhJ zv--)fSS+n7m9@}%#NRg|#&3QhlwBozK^1Cwy+d?N8KL7h1$(IA))w4m99c1Bdc}>ZV&gvD)jxBs(Zbn$)8=c zKUFYCTuj1rd`FQm#H+{$OZhS%-dKbwU00-n%q7M2Zxvh3_4vD}=$`igEdE(+Ptt!0 z(bsN>-P2Y7cEt`%p036t*b#nE^DE=h-HN3mFy%*a{7br#23NyfMHScO7K^I(i=q#e z4cz@je;;>pHy6_fiiH!4L7);(y(0)z+r2Be>d@+WZ_I^ z0jHEF+61rk#EFD1m;+{UKs-*W41d^~*pqtH z*bn2_aI!b zhQ!~PczO0 z;CQ^BggI4aC=^osNfdgdQav7Bwc$&y#?Y{9bPt?28;tu4V+$$KlGFqCdodMKHmdHF zm>LHzytZdzQvBW75@h$|+@~7TY%?LU!lv*YY;L10bQLD>8x|ke4!rie+O)jd6Q*8D zcJk{!nS+x#|6~ruKI9-hJ;powg026?+` zuoGpiI%{vE_!eMuQ3P14#>kxB++#^avCX$DA-b_mXJ&3{EsQI>isoF z+a!G4_|M#3ZTzF8=M=xAP-ZVu^DK-Nmp{#!(IK%hWbg$$4!{xA=8gbdMf#Gn?V{lnrzjS;T> zK=chejl`nFotdCE(9E6D5PRUzJC`JG(A|K<6pUO3B5I-<1Hj;GBC=$=GKgEP^A;ln ze4p@Kx&?V)|5}^Z@f}e~VGj3WY4lTYL8D6`FTAED7WGFsgC9IvV(Ay`Q1F~AFdKpW z?WlVD++;?L;CKfMSiJmV;HUcVACJkQo4l-hFbNQCN&GwE*_5HU!8B)pHKCVKw z$K;=WH%2I_cv5W`iGq`yU*rAHVpt${1W3gW6f0+Wb7LhD-Vp4A*}Du~LQ7%Gs%UfjvfPNKbyQ|B+DsCZjP#MlI2uk`K?Z$rou z=>y*IDY<9NZWK>)ubA83EyQw=+Di0=I@nznhlK91m|G6w*s^hRVf3fdOn`-x+Vr%&iCfKPJ3UC7Y;1+X7 z<|B$chz86t2-gg|BAO$MB;qQBxka%G0k%)P5{ULWa=!e;aBf1!M37)o|3oNvAGS2r zCl0yPA=IqE?rQMJ2oPT|FWQ+H?yyJT4XCk%Si*$xOtbg|Q4+RGA*(u_og@^0kk#OT z4^m*Ui7Y-9bZkMHheF><2b7})pNkf=OdDRaGLXw6pe=O|H@JQ`L*vTOtZC#z6mpv} zjy6M7oz=bxiN|rf9kq>)4dbpvk8%eHfi`eXAhLR~+LWz?|Kx%YIaiEjz!Qi()mGT$ z+r`6qSrU>fz7{R$c*lrkri>v!Nhng3A16hN!BrrtKSlCm++m$1F4s_aSREcmRsBCJ zk|2e^l9#?q|8hd5Wy*8$EYbkIu^Q~24&p-3sQ4dLs9d%F+ox7Sz(0JD{W>gFocrlA zI9*uvA>PBTlgPdn#a~l;mx{z_k*TL zrv}n{G?fB=c@rm-ZB3VD}`sO><3MrtFt*EYxi;ab=h$3tmQ8qb2f@!%Ih$sYeWF)%B zrzBc4Elm%?EP~IguFzvS#|&C`>TcCMWW#x(^k@kzQx?id=^*=jNAy}d@y1a?@s|Wm zP;WdhqrySre}}&4KHrEIA=RSsV7K}`exi`OBB~!Mg;l~e(`cl^S8LH3d=;1RAJwE% zUu6`E)ZzfQB63Yk=syGd*)u%1!gJnNr+5piH&PKSE(s+dRGO7pLXq3=Co`&E9$Ee# z*kD4iDO_1ds6#-;_O}JPwEbGhHWXb<3Ll&49p~TH+C1ek?k+5GSHyhsyvF%9kIc6b<@#EhV zZ~Rz{#27n7yA6L7@#xU)xF02YOd6bCYJt1%8W`u0sYYe=aC-7bB+-bLU2=S_y@17V zM8*h55W9W87A+(k(_U2`8_kMpGwda?(!a@I#BQfLv2vWAPKvj1!~)JL-Vp;ncYk7c z>4kqV@f`#YsRFQ8FSc1SA8mpI23syc)47+2HBcfPGcZLLdK}FxYiP?2?+PIh!38_Q zHjPQT8BsBq&7$S3B%}B*{~zqoLh?`Vz_+!6t^VHe>!OJ?!xIywMz$(e zYq1Y}y8Eh9WJQvSJRUGN&<7sW^l-tl5CQAv4sS}%3XUy%$>51LZD&!Hm zn#A(zmg7X}NA@)I1Ju7m&J6dM+(w&vvifjT%2ydx}kI2vh0ONpxzf+)9Vp>3%ZR-qqOnApemxbxL@)sAR|3V$Ft zrtyl{F2_qA7b*FZ{6QmC4Ml+Y=P=T@Qk*SnDz8QJwu1;>1r1Pzm_wYJxt;R0Lzib@ z#G;{WVcch2P_}rD^*?8BjoEYX>BWWkBsd{4V9sXQ9+HUyX`SV+jop!n9hHDMF)*M~ zH1|WS6c>UND5AJk zzum^%Y-I(~d0hoIShAIzpr1y1LJ?CC;{uqX64GE`^nQXnFLRG%@r>XE(&~OgKA=JT zKRdWA@I>Z0Itw19{LPX28f!dfg%h1S!GQpU zfB@NZW&sWSLq@#w2?XgsN%PnsF<^tw0$PX!KS~>V@Um<_DmglSDevnUVdUz}kxG#x z%+q*aGJ7`jPm~bluRyY5hgaFsQ`|&`6ZZC4KTX0h&RY3ExsMze#C3yCvZT8Z-Lzn9 z;;2SyNcbnXG(5}u9fru6ESzaX?@&ZKh;C}EBTG6xW^XU0x9Qpaq*%T}rb4QYZf5tE zvR{{Y##-02-k`dzROwng8N}e6%pyDHLc5eGR9cLJ+B8&Vk>~ki5Vdu?rfc2a>4{e0 z!NZi-TIAhl}iEQT#DU zLmTZnIIh0P9t;jA#Y&T-TaHEsqoe2x0g1q<&lG??l%Ry57P3J^iLoFC%pV_cer$Ml z;Td%pDMovRiuzo|Po|j6sTF?)be_#dr2`sxAnnyl5KG-Z_IvnQkVK*FCJ>)(6-Fn4 zCymGZv1~pPeDck=3zU1IpSISKbQP}!=GgHct>CI>sJLtqfCKLlq1{Zpr;l*nA^=Dj z>S~*_%5t+K!%G7paR-c{*u~QRnp5v|Y^E<%v7&;wc=}P3?5(`VHqMXJeTkJD_T*_ny9;Cs?U*|=TEF=$mB@N2vUgeyaKMVn!3V=Po^+k zJ3cYC-L&DWe0c;ob5>Pv_Y67%oMbAq@L5+hGB!G+hfygX16ecRvI>if&otMI6SI$K zL?hW)iHe;@xth;o8Ak!6H(v#tLzo_OeY-V(zK>WG9X}33V zx*J(BcQ(ULty zKirHTZK9cfxJ}+-`@Ck`Njr=_HgRxm<5L5|`{oP`M$Qh@+F{v16wAqNXBAF|Jv8J{ z6A+^qw^A3s12Q7+yKI2m*nO(&fPgkf*}&s_GtzyT;IS=tWq0GoZp=(*3pQM1O;F1~ zVYhi13|Mb1|CE%?EhyyQk(VJJ<0fhqu`C4%36x43$((=>x%UWum_*n2Qun_lRv9+I z+_kP1jZ|qPi9o3t#*jM zonqa-&I}Xu`e#s6@?}uS>;kh3P0!M~sOog3*;QubHZ3zd4>}{iVjD>6z6Fq`F$vCy zh(a;6BQY+aacdFf@zQxcuz~bOTTPALrNs`!oqeKaxlF$bI@G|;K$M*y)UZ`wPD7Uw zF++*lYG!m=oXX+`%#zoYtS_6|Kh54Xipm;Fa5pxjiS#3#46rT@24DOWY}!q6z*FJ? zlFOl68oJb9*~omIF$T*-o!fNrkX1!a_YD5|3cqIDjF3S8+nnQ$$4m3YL=5+5_@v{} zm4GcBJgu1Lh){^4fiPj+VK^Y>vw4QZ=E{)RYN?BsB@)-yS(VTB?wQy<8)H&e|2S7V z+I#%R7>qF+`|P{kALGfE0i_DTXh4CF%e}k8iv(Vw@1+Su5K_o1H3zuW1BIA^Y}nNL zwF7r23K)=OzdpBQBc$A_uEN!@QdQFVsl}&sZ z?b~DUswrKe1zS=)!9{Ugvdnim^QpCC_ac;wI2<0}#RW^|`v0>*6pHX;<&$pyGNM3(I zcC394(DGyeJ*-LlSm#(itafNqFB+{)&_giP>xFWhPl5o3j2=WhhR0g4ZfA4I=DQ(Y1er+)tBsG z5cO;9s5`KI&*n|5VerBv?Y1qV?|BZ zgAeEN-|rwqQ79pQ?-~43#BaPZ5$VsNi_{Jsv)s?*h#>lGbDV6hDqJj@)WXH z4j+pCLTMY@r7{aOXFvCGym&?A%6^uZgCuuYhVRFf7p>+XLq#C7SVEYxyak0 z!%|)Rv_I}v;p?)puWZ3)e5)C4lO?q@Sr!8IWxpV$22w>uil9y2m(?DG4*|u8)d}Lw zvOBNnmKCY7^h{=tkg3L1$71=y;^pa3l*W@aqMBkzB$txnCs7fd$3ml}&ir<9xE=84 zRvpp3`K`WJsL>6v;wX6VJufM{R}8|gbQ>gOH!s?TNDCOD9%))?zaG!A z8og9w)gG^U!ats<`lqY*R5e{)r3VXYjQy$V>6-QDs(*u^lJG|~7;HT)lDc<%%{Q5C zgtQaSiy|>7ek7}pc#Z@{$l7?knAPZOB1I@F6#;y-3S&=a_~>iB(ftHXq9~nW{xlrZ zWNt_xWt}8;X;a;5(}O(u-{HA(3)}l^GsmzIN$2$=stxVNmd4N)dbXo-u&N2$?r!$Z zyVL7EI#wJH2*A%kOi3yLDfrhNs;K6Y{2qF<5upTM}m6ph>)PHYT7Lolhj`!asLhPEY4-(P4HhmB^6wgq_) zjYe`FLLWstbe4K0Sv|O(U?@8iXj;N)Ly0a}VcwW2s}(4YBefN6-cBZH&)GfBzAJrg zJKLS%v7%_S;!#`K*6vuVforMkQKFCh1Fjf*5IG7pQb)F901`H7U|$P&DMl|iD8Y~o z<6YHs7*_J7+QIy|2^GoD=z8I5T(N{cuVo3rmt%>Y-!9dFmj=5$C2Nph3K+X2X|S8G z@u{jU(x4Z0z(3-swy1HnVhYcjbh}09l`!NSkQ81P5AHI~AkPOjEKCLM#EKDr6}68r zRE)MG+U?@7$0b*CS0smUS8zWZiFRlo%h4loI+>uOZ>SpIfV(2LL*18|=wccr%U z^SHYtcIWZ2-Oe&I#EW#&u8RF6_G=xxC?(+$k$EFoVMHmzfqZ}i z8QzKs&_)N@NwcEuC|=cpmbopIcKs{)2ecM0YZR&*ysk%2%ZeDsD$GDl!}wbB=oTj* zWG5VCCmzJ5pFh}sPt@B%_PvAb*!GFlUZVWd6hH9<{<oI)VZC1FGTe4eNOjIQC>H~9= zzb!NsKWD+E>ox+!k^nU;4Cvd0KaG#>;VSw^{7NVL<>#((Gq?4&gdRYOTz?lsLt+Q1 z7~{9#;UBbJ^gzO;xTZhIlM0WZZc(RVgu14a&?9r){JLn~KHx1aJhk}P z2Vz13020Z10GdCnPFl?Tp~a1}va0vg3e~))-BitJb3J`bGpe=`wx@k;wvTkO3Jli=mMsT$b#nCekLIfMW&{gV3Dn+*T-Hlcu zQLq*MUU@iy*u>~65`M&aRJ|swOhR;*G_Jh6X(1jPUBAIsZr|WF7aGHbhLJXX9$Zl4k;U&I>lEbT%V6s=$fY-Nce*Ni=Y`X=R;*)L1+Z)8-xT~<`$wSoSx zO76I@T{wMf(QXqrF}t_OwpTDjjzlxy#W2oJHzhX*n%_kmj-U931TiT(RI0@PhQ*~= zlvQFu2V?FyAc+=PX$e)YA zvB*01amPA$rA>%Fm?^(Tlj)`U@P zsib6g)JW$~rO%Zx;1oqPi3U$b1$H)+Osqjbh#O&XFyRZJp98^#x~cn5DxZGcmRXE3 zW`9~V6x<*mj3=}jnFblM6SdV{W_G^Wl2~ey^hq^K{|H~ttWV*e*I-&H?B_n4e8Kmj z#%~M!WI*CR{J)>!MbV6S7^A`#K($6*0$(V^gso!ESPoW`HLr!GK|&X#AWnL<{own>J87O%_^Ht zj^+`)BxPeO_6)utt}8Nt_V31~d%3N+lUCfhN7=XGwhmn<#9+|f6k(MD6+ElhskcvS zM^v$%jL|^M@2~gHsLVlFWF?ubsi^yCAv&9-O7Kw%ABa*QY6tFIfmNd;=(mX{sRA$! zmd7X2T$&}LDDy3G#JVJ!FcdbMgbfHN-&|^4=gL(_czY$9J2korcxd2?gNfHQ;))3g zXLHa^PSG`4Jj4%gwY~ryBBZnEQQ|(J3H*)(N$wYSFfoN(NO}o%R)H+vOruve`X&JD zU=dnUH<;x@8x*T3MRP}@<3W14d@+&;Qi$-30Na(8tmLq8#!#>^oOEpb1GMGmHg|}( zukfhz801lnF80{{X-La`v>OnHhSf0*>MIHl$#*pYgcQATa zdKBXoThj_7pi1)-I$?zF({lNkKu4Cg8Wy<0b3|7?g*ycUB@hJ`x(3(JsdP_{e?b@C z3GOpjMQPm5^rB53CSL1IdUxskY4S1k9Zz3JU zJUYQQecFeOWR&u8?N1w8?tH_H5p*i9Ll`JgSnyHcYcW~~HtSNIvO`u7qRJy>uz)A2Ctn0x#MWGH+m4$OoAl@T&1>V*x94? z8A-I7^XPe?8w*0h0wZ5A8B(IgxT8Bn%cSsu%ThI1TN!;>v#`dB=Tt83CLt86f@eWe z0`pJ&AW;rE#MrF8N2Jkc6-l3c#zS!y);S#gMLDB&t|3kC$Fu^+9;!O*hZs`ozfsm5 z|Ai!JN!=pMP@G%>Bu|&YHp!6CZeUE?YC{EaNvOx4BhJXY>>nW&t`J=xA`r0OiYzMb zUD8SFm3vCiq%a^5M+sJ|?E|X}#`Ld@f8ZBRNRc4;yi+B-=45-# zO8nlWd7YuOi8MO4Ab#$YuZ`maLpx8Zs_{A|Ly`a6v9bLw^GA_1XZElP0J4F5h#Z(6s|N>fq!a4fS|(&}8YQ`;6N$-4 zC!AGd3kQR(h|NMbves;&dw+QTkYaO3$A$SZM`CF^=Gb0id6e3-o65YmMyZblEZ(yjd-~qu}o;VBpxAew1|m<&yFr>oXjkO2==zhf*9!W5voSuz!(=8N5mQhH^q=g0n zt|&(Y^^zHy4?nhjC8dvn(^uRu2*CL){xmBUx`{eS|L6yz%M@J!9H8HF%!v@Q5gAQA zaDx)AfI1stG>*=uBs0HNI#BnFax^+^Z~I7*$4K-T9(sf?wgr2Z`#_ON&92zJ=Cu?zVVm7MkUAAFMBXj07--omYcEn$u<@=X7avcHwkF$7!c>@rJEgH{#zH&iJp zB>%>)CyR69jaWP#(N>4ju%DB=O>39`-lOl z8d(q~5URkQ3YAKz`7U}uh5;$%M0eCHCF(E`fi&8TUZchjSb)ruM)y*oKq${&26PEe zVMRQY9r-?vUTeguR6}-Xi~*X^7q~3O#^}wBex}tpG_1gcb4B&~J32ZKi&m&|DSr$E z6J9!%9ddrPZElt!))bij8-NH3oKH#JX;*C*w7^i-D0AA zvTeVMKaqT7J%KtEJc**2@g{Cc!pSs#pC1>~{FS%>J1j_RJxFV3cFAudy~9f7;V>Pv zsDR01`FOOCq|9P;Sr>7p_(PGpX{qDgnTRibfXjEpX#TPf(mfJdB#BN~7mc>U?^dfQ z2kfpoh?(Iz?_LUOWB463~XSRNeM_E$FP)Ds(TtZxf>i zbHGk@NVU-VCer|ic4371rvHq#SBdH=TWOL*r~PqtwIt4X(8g<>*Vriz7%ZEg2) zS8VL%;8#5#PxleuDzuNqHVJ|17w_QR`Q_1|+>`FE-ulO?v}M1_mX#N7!V@*QA;h(=&^9Ld%XrzY*!S@4t6O z@{>M0t!1anFHERuV_66K@%b$Pa%amf&~eaT*76^1TXQ^`kuba;^VzrJsS4wlnsNs~ zL6>5Tnd)3P>@kd^2Q^h}P3ez1(g}tM04>*`W){ah{1%o0C6Z*Lhi`NIsvtB6?Nq{A z0dqhnB`t%b1sVQZeFTk2N2;7?Fvf-c8xP(u!Mo zt_f=bTPh2oRr*H1cF|}fT1JUU?u42F#*_KS6Rp|Nnk3w=wr{LLFW9%dji+_2+q%ZX zw#l~MPq{Q2x_hVK|5mWKG}ym^=ltiiEiAf!6zm<&0xochDlwHtfgn&evzr+EU7%o_ zx?$d^`V4QR;tCQ9Ul%0y(>$jUV6tzied~_)t&0%}NN8_JN z*JQQ=I;b>*Ka-KjxKsZ$TL?^kl~84(k`!Mi0;F4e0#KdAftD5sV7_I?FMA@cG%i&?ndIHg0Ezp<|BDDo9jPhb5nYplTA7j9Ch($z5|h< zwpm!(?q~&J7vLmVK~PWCCc2kWzczN{98h@Qx&M*MdwimJ*FHy+;sLiDk4oKNO6n`V ziaG_-{ll~)3t6Er8Wl!V9sgl${C*MTo$L0-ko~FC-s<%9+&xQ`E6DM5?u+j8jOp9R z$Z6^OhPb@I#+5g9yW2eXAtb5~*l_jVF?kl;{$x?ivHr@3?Fl)ORPORDRlEtfc5X@< z#bH!h6G|y0Zx?`|!Vtq&!jvL&sl*&d!MB-!@e(Z0RW zaro#t0Z9kVnpl8XdG4#X?+H><0t5FdrMYkQeesSxG$||2Eq=)Js74RQX&1&@HSFHn zJxJLd7KOw-x;tVuCnb+1J(I$qRM&NLdAFlr4B%$y0-a9|W_aP`wC$4V{NFjZm zp|BDl7lT4RWmY(QSZJ)8c~!z=cab8s}~&&4s9~ zkmOHw)2k_c?N{PL@(=ewF!HVOX$kFv=#m(#5DM6d&M(kxw9`A=9JHOBL9j7vND_lQ4)Vw~?1mIo;T+Oz>YIwYN(hyytu z6JMh#Ue~i;FZqhN;vP^i&ZP0-nw;d?PB4-ANvxx&#W=ovh1vSgCq1b47 z-DlYMi#z(4W#?Bopm7E3AJgah`$lV`bIxNKxVR=pc)67Zv%i}SQICHU1kO9DV^JeS z&gS{|!9xLDxg-;$H@_~LEp)FU!e`7oj~tb zBdgV|^!`Us2tPLYbo$G(Jzln@b=heIb_zwJ$tOW?zz5Y) zBo4~8x>Q$!rVj>Z<33qmPabbg{aGYp7oO3>h_JAd7DlRwJoA|xQ~bfJ6zjltWW4M zIElScaKmtShkbqE7DI|WBs_P9=Qup){)_ze|Mr~Sf?x4i+jKWl?sluqDlRp}b-iHa z=grTu-TWD@@&JipmONnYK}#Pndyvho*QEz+9KRY~s_Km8=rKh&sqX2eR)Ec8KuTIv zuwt{iLB_wP65m!KPVzBwUo@HZ>h51Fq=ih3ACK*tJS_&hd}?=B{QqS3PUg?8RW3~Z z1!=OJT){NGm_Y5+cy{)aw8XJDX;OlU09YCzU!2C5rpf!0E<8&;Am;HFuZ{f+k^?8k z`{1jYoQs45?i__MV^|=vDmPf+GE!Sb{8+Jq%JHYr3r^gT!WK0MT^ZU!mP)BU4g=>f zGNZDc7TP$;Jlh^oiH@_*ujO%tbztT^^ zYq?)^1tH!-o2KcfdUAU5n8c5>&F?Y)E7Dh2;O|18t(8h8_4$B2`7&|jeg(V3kMl!q zW=5k$8Qx0x9P4?xRV6rV0&JN3<`AV$uGt9mm8oEad6G%yua2``S@Ri?SO{G+b(!El zVQgLN#}*dpqnP)PW4A|d$iT1uXlNV)JFZQt{)Y|5JK9U^+LnDs-NAgqs-dR9WYBsQ z=@Kjotf#Xm4(m}?&b z>2CWFvQB*ahwb%`xEnP|OfB*iXG;zR)Rz75Pz=Px$%0@q!xZqk6cCU@93Ad`$WgNL z5c|m?KooS5y*7cQTmiSuPB(u$P(h8u(9diJ2>b9yn%!V-rMaIerog@Vl>oV`53y?w zNq%<-x(Eh)9+|Z9;)ht>ph1$tt-U@b?m&d^v~LpVFf7a2Sn!Iw>QCS%Xb=|4;6TJSAf4qE@0|qvU^yC@p8~&GPd8w?&a81C~?43 z7-V0?B$C5PT)89Q{CX=~4T5FmP_VWO>cyleT@B?BwlO)y_)67k6zbj4>`FwVUA8e) z)JDnE#&zkU*$MZnL~G>PrCD3zZ^h=r5Y>ekJ|!1$h(AR@+j^Dp{LN^#pXqnCnQm7< z6A#vQEX)V>7c|cgv?z8<}alCebmOUkQZW`OcOL(j1*zMhZb}M1lP&#e?+{ z6fs{3Id2p{bELDQ_#QmxyO``5vI_S8-(IYzLitzpWjBp75+{j7P_9%=Bc4Lzud%oU zupR5y+%u@yEOhW@b+x@&y8$0x9f1OcNB}o)twi@PMYB7@6qzPFWIPiowI?^VJ-KcP zkDSltxrxz3cmuWT+J6N!h!@S@Kt_=6&5_yw*ZOQ~?-H|fth8i9U){N^`~Sm%^_z^< z|4X)!0}IdJ8mf)zTJ}V0x5M{I^Aq;B|G)%KI#ooTb?(R-<^(Yy2q)tdSS*n6fVYl^ zC;jhktclSzxaJZF`_1j6Ty@xsqlXF}?n5Y<7>B$>x7rrBaDMg<(P~O@6K(*NBRE7& zJ!|m*Nj!d zUyBk4alfvT@W`SIm8`jo9oEE{E+o6o5M;hd$p&`9>q-0ubdWFnBZ=Q73?YC9|874; zV-S9Q%xh7}rZ%d6jYk0o5w>5LA+y?Nom9(a%^;8MtyOSS=7=~EGJj(ARk*7!wGiJV=ALnmKvKaeHBYH|Dk9vDacWO0D~(Ni0585|D5^bTsEdMw%AvpQq(g_ z;618})zZCIv6CuZsTT38Y5W?N%+y{Jmrm`k%7UO2CUPSx4SK~;a-IZeAy^Del1#0N z{GW4^{T})N76lyuw?MW;7XFZ4jawJTq=#|TJChFLnYx;AK;N4Yfkf&{OSvtS6dWX| zao7)K6JsShPAE*~>=g0E;K1qBZk1FaIFFCssHRJ1#XtCnUsSfE%RDn8gz zT~A|1=ogCyLcC3dN`JQU;E@%@Yv2u~Qo+u`3(MAAT(b+R0^iD-AGGnLLE!=PliN$s z=l4KJ?r>k11+R;7ct1b}iS+GIoTc1gJ=*Z0@2E(}zJYYixCLj@>;h7iOgF|>5c7;@ z(%Lmtq^+wmZC|@x?5=6}Jli#Gn|Y>3@7AfH!bd5P^!+5-B&VG>%q+(kSYi zx2h|K4)I1M8hVU|<{<0sEM1~&A^%mm576DDzmIKQk8W)s@jqU*1L7NN=~gV;SjEG! zHP5Tr0sa``9(b*FJ)n*BiES>NZ@Qrz+eN6C;5#?xhv#;KdHEi!( zjv3QlFxlF+-Qe}K=*Et;*-l{8;!}3FT0f>{-|w(D$j~U*xts;0XBT|2wY9s;+cPD@ z686-_cC5b>Du=k8e|#r~vW-7pP43ltVgJd%wzi@cxFFt@jE7KQ0?Ml|dcP`6kTeoj zLCBsZy)Yf#;mukP(yXr;&Y>2sKi3gTNKHZ(Y!b#cyDk6-{{j$(qc=Lx{o%$;@uFJ& zlA8S@1kT*m5u8nSF9}N|m5oOm%8_FyvFhpMmvIKogNP02Hp3DggrVU?!De)IPPtv! z9SEjfK)9$|17(7Ac}?i2AfAiH;hAU>Y@n}O_qNG)?k`HPZa??c`s484hv!QEIp3^= z-*vRFS%|Kna2b4g(B7`OmuLx_zFn(hwtSl&hjr{N57g~Jsx_lac?|xi)O9ws=j!%! z-OlZF%Q_*S7j(J{JE>K8PF2Rrt2)t-5b~wHs4fq?j@p>g1TIrX-ak8SVV7GQp8wEM zd#MAs6iCnBC=Dv+!=TnKq9Otc21XDsxnVs@IxRQ zaMYk)sQc&j90Hmo>)i(mbg(D;?8~l8QjPcba7yYxF*NpoU9lzcGJs055~9~41Q*0> zs@=pDNi_|RW{r?{YmqI`s`MtR>DpJJoRmVI5I_(WW>1V(c9v3BUg0u?#iDnxS!J~@ z;2)ixZkujb**4y;5}npezwc5iXQc{PcD-m7)O|TlHkKS=#)y~@DHkkGj0;doUx!|D zL7`tG_dPkSKuU^s+y@Y(H$~sL0E;jjn|-5D8Qmc{T-h9AJRxcHA3c z)RlDVvI+`KG|mbaYeg?|F&6Rkj;=EadI+|!d~8!Rw#i(E(JIPPHHdGuZ3EJJ#UIZu zJ!09zwsmxIclsaxzwCl$c1f524!1iZBVNcACzmfW|833CZ7QY#k@P!>!HMH$&_^uF zIP~Naa`A$`1G4Cf?h*wfk<^@@0BkG7OUQ_!8zWpZ<}QH(?vc&3^_-#>J&VzxdQsKi zQ?*|aDrWaoGmgBY-)z<2UbWkJ2}Qb@3LnsvPO!Buqhyyx0)|Mpkim~pQBY)dGg{NF zDlH4Gj%{srAxyIxo!JwF!P8sWnXOQ{pvHpX%Qy+v8 z{j`}}(ew)KRtz`?F=YraF4ZM@bKcEYK|E$>bRJOQGU`p!gxL{jy%Pk{i5#z3K1nbr1U(N3m>} zbgHaf-h*}Fm2QN%E*9o(nslnk)<1MyLboi4QV5LrKg7LvoK#h|=zrGU=Tz0H)YZA^ zK<6f>CWj_9SyW5}1w_nx=gsiOxp$l!1WcF_6*FQEU>J0aNHUH&X9eSo8RIAh-goV) zYP-F2-{<}Q`Zd(9s#7QIu=dJz&gO#)1WqG*kYy!SM`sy;@%sNj*pj6;DM6}ttNM4mUB?nf#)g1fbNsNP(Ft{ zV06Ug5mhJ%(IF}vsTDz%X0+bA%l+L7lImFgy6ja^0?uqB?qtpBo?P3Dn3X!)2c=e` zY9P<)7qoJ8FggrO$E~9bC8i8s7`LF%L(>s#M*}=9bUV%-K;&U%tk6c&m3ihe)hz@! zlG7pwI{~_a7_ZR5T0ssjyFzi6`-`@B^S4ogs5;nD&>km)&}N^z+Qz^j2iT=3IOuUa{0JiKGRjkL|0~RDZ{U- zE0|wXC^r_`TMFf7a>ml1^@aX~U;?gkS-gqx4dN~HN8kw%l328cyg0~iYG2T$b?h0w zL^z}5f=yElp2rKIr{GJaWX`Gy;J{-IQXe*%My_u<6XVIHCe^>;%PBeVuH9TtCKJt0i?51*$Pw zMI|SUAuw|jUQ9imYI78?kPd1b=Clg$+iGk;IwX}P3Z3wdHN~OMt3(;gK$|A40MoQ! zc%iDw{Iett)L;m%EX+YpMwuP=N;~~u9>Q`sFQGIz18Za}Gr24B*s#>wP@^3WG;fctQazOZErTMv1eyTJ}tBA4Hl``_S*73BT)nu-yk*jLtO4q(# zoqng<+*M^Ctdj4k4Qemx?!?yJ?&M8a+^d+I4{I{JG}-^wD*ITKRq^TH{n6ZVnO<>K zjl1Ghcf}LBnouikgI-XO@SE?XfN-snQXPqTChn?EAh-prh@j-V3#e zZC8^4-&|5A7njLpT5J09GK=Q(2hDrBzLqj!CPfWtWdo3n@U|;u*8!@^XsL3HVxlb2 z3$=a>``7Low&$iA91w1;ky~m+$psO@^%*(2T2@rcKe~0@rq+I4#x$Yv41TCGKUc}8 zzdGTQGN86wv{HJJ)sIYNz|sBP&12m)xvd|U>$TwIqjezOb*{=Knz%d4<#tdGd;Yd^ z?0ZrFR1D@0%PWq?MFu2Z?oU@ao+>AMn+B62-EG{37=h!TbvRyu&yfX(n{0HpTUMMy z1(Pfb+e>XECE2Z=ET!ZT7Wmvouc~DJi`>%2-O?&Di1tvA{o2_IPe97wEe}OLwZSRo3SFix71nUXet)MokD-Dxw0Fu?3*?EcBEZaE2q@T$+fbgR`+w(w~A}dt(7xt z<>OlOW39Yhi!FVwJcBBOHL0dxZKL$X=raoG@GDpGb9sXt-XXqS3p=1X!6e$-8f+ac zrVj0pljW?v9%(|8(sC<7&xM6xFcp2{>O0A+K;NBc4*-+ey%yq3?M3g-26Vbip{m!{ z${$g6!_JRIZ;Mf>JVznp;I_^iE832eeIvqJTF$BSSJfejpsdJ5UZa9M5=umqVRLt! zxH+B)u54gmyjCyI*UJm_UIMqBNfGRFm9aqLXrJ2QAR-Znp zUQVnx+BVoT8z7~cZv(K81Rc7#CRIz+V4-nJxU`#CP9xbKh)NDun$ZR~p%w^U)r$W! zl?iWcGgdOiYC1Rtk!%b2Xq`bUYjJq#P}j$E3zeh;(M}i;B~=H6n;eAO+&i(Z zK}?Aj3+;QGS_s)l2RApdR39|T_D1=8qjPnmAM$x~=4(1e*KTbv8ye)wx~@yLNwrO% zH>~xH=iVcM2w&a&GC@%od$B_06Uv&S+gVIzaE9q zM8QS`8r7PnNTV}wOjUiADm^2NKH!Bp4Sa6Naa*=QeZM8(=D_qk6L|ARJKneKVb2HL2=I!#4yq*^R60X^2-9GF(MSeuL= zq(&BfzOoGpm@)6EUf>wq#+x%|Hj{$n^QO#~O|rGozSW2XXN3TCN}i~6#$Xf*ln~+9 z$<1nYK&qC}8|uv>-q*eCx@Nh>Et*XD@Vp`Nk z8mZPS$f?H_keKqcqXT$ZxZKlh=uB7;6orMVt6VCz%IYPuOrP?545^Q|rH|9-voCqt zpL@uL9&{FVNsrokT!XZG#x}cLQ#jfy8Ss$YOqf?$O{d@BMtVk z2Kj6D=)(0bRg>6 zho>Hz#J^||S@!hVeR-hT40aB(L$%T@dU zV_@Wn?Dv-rGUpAK)x&il!9BxdflQRQd&zsfWLZC5+*gj2`$x)!&2nZl;y?9CN6Hr? zx%|z(`s>&I2;@x57y68gLc6qCc5{D#`o|`_CuiO2&Vp0Dw@L2F{&HuNJq(O?X>Yl# zx4F96-qb8NHoI%CTSa@e3D zS|4CVQ&dhGtf(zY7HXLJqKN=oJ+BsCd-Brw$ONwnUg{SHz zxPOE>+^-ai`%(<&GA5%Lhc({4$f5j`ZK}Yj*avFZNYxZ6E9_5A}+i6#B0w`L;YM`Yyba zKPV#qW_R{r^vb#1+&gk_u|aCNwZGj)#ggdX%aqfxyGFCyZiXd z*-=dS`~86g=Nj_?*nmhmxJ{Px{{He{e|z%)yKzA9Y=8TlUUYGPg`jH&u;X6t??0TB zhm!JWQXWalV@dy_1_c9I=gzqse>TWpFFC_~U9Fci+i&`t?~t+G{eAhyLF>{1mT|ju zu~~q*k`p|y-QP#1ptis{2g^c%M&U;M!|WY+LMcMUBZ@mh5@3~CYL%XepX0L@FxJ9G z3Nls6I(9pWF3BEIL*QfsXEmFiMLi6$s?oljp^Wx`UNDDtgdHA?vLnqHjO+c{9N z8R`XfA5gsZ1=<|@b+LSveYbDcs>j2M{xBc9Ru}Y_)B7VQ{Loh}O36bO+n4EVOuV7d zuX0)x&Gl6vb&$Logm?G`{OXuP$Slt>0nSeWxo}moVXkuC*P)v1oM|~!+?%AQ5nd(U z4TI<lsD^K**)noV{_m$7|v$~=N{HeMD?4ci!onaTms+N1v7`9P}vjZFx3SpR> zJTAwggwi+<=EdFWtZ z^>!mDRgEQ7g+UOb#G3*H%LgURPym!Oz!fHu(uFU`)}ud0-O8tpqmwX%_|s>#+5Wv~x=FZRQ_esW7c+1f`dPJM1vUK$|J50D)LklK{r zJFmZ4-5*-omGM4BbLCF%K!(3j&@Q45v`;UzLYLw}`#X=^J#>+RAO`2axk~#i4 zg+vqG5&SuD6|l2)u!p+O)vn#E)A8&%d%7c+le$)R1btkaa_zC(@fDhuVx3eh3OmqV zeGV1n5tX7cQVk`frEX~PFvGcd?y}>I7CT&3x*%q4IX~SZ<8=MA-&EE?6V~XzWapGo-o$jppyHyni%zb)LfJbrE{J1I6b0oX$Tf+Y{Hu5GteGC#yqWeTu*CVn(l&NugH@E7o?;dOjMYwq&(FkboR}GTC_L4`U`eE5%yL>Rl%v1Dd zHlBa)py0eA_WU8@T2(5`H%&BerZYRza&^+b3{=60wu^j#J}@%YT-<7Rw3^9!{VfAc zFMHxZC6JF*%gIBs1aeb#{0VEovfm8SGk+_~MV2uaa`h0|IV2?iLKu-{L-RCpx%T_y zA=>Ztt@0ilmRjB0^&ueLOItekWmBtMnm>^=pZF4S>5cBl(qTS9M&E<1<>jH7*M_3v zx_Fo(z+n#al*wO)>JzpNQ{I|?TcPj24D!9fZqa&^YU>Dje+ZNI=n&c6S10Y!A+b}3 zrB@6Cm#-UUZy5%%WPQydn8YiFo7-CDK`Jl~w-53h#PKK^eP8YwZZ{5>*N6FU4U;#A z*&V~cuO|(YFNOjh7X(7JMqf4$r*}JsbJYXGv}aEZ=f1BDl^2JqwqHK(*|~o!hnl|z zmM8tq54r`l^7%l}VXQ4&<)2en&AEZ=I{OGr5O6hA)Fiw4uqm5Pa>59#M=$+rjewpsx3NuYaV9Yqn-DEZ8+z?i)(P z_D)ij$<4#n54P-=J)vl4nvkc3IEw2HV|37{}d1EYh z$>PuRFCWS;vryZNAq347UNnT1Woqa8+d%nT=*Oy7IiuB_*_vG{rql#a4m8gUl+FFS z_GMF_$c-vzsV-WVRpcM&y47+u&h#K7SBBV}?g6E@Q+5zs3X>YGM&o%bN|s^CsM!-| zoKaHi_bT3_tO-c%jdO`H4b73B@gDt6vLss4vk!Ntk^_TPTtzMyOJQI90r?o>q(D~@6%uTMAedvo9{mx1CyO%gJMmTcR0c)CBka**_GKU&apgl4Q+}dN~|4R!yaQ`K#|HOE!pm3!qW?5+)5Vg; zLmJ;^@`$nBP&-^Hk&15Rk!##K*{j9gz^W0xE_NN6y#)uZ)NNu&kF+6w#qCn$yn)3M z@C;_pp6z@@amq|>6SGw?`sD8akc3DPrhtkHPqAF)aq*uKOAOizcshJtJVK2)ygueg>EFAuk`42M8zmJ803OVxr1TDoNjNzgTPr`0qm zr;6}FAJJN}`C;!s$=QU~#Rx9x0u6}VA)}ar9XO=Bw}@+dyEe65J!N9J1No0i<`EeR z2QwzelO6A!JrXkbz!3kbA^wHK$i9ZW#8bBYhK}%1Tn$0pTZa4h564sXf5(<2o&dK7 zp?zSuTss{0;F)2t-%kve#~B9OqYyjrcSuHD2>;O6eA^ev<@>&tL{!p9ddvYh$o4hO z^5sDAe-|SE(oj@M`{DVwz$`L{$|AFCnB0n8O8a0Q?`t0APX6OGWJEP`s2pZESr!l; zbYS$+zIL^uG>7stKMjm#h$?-puf4pVeQ-GGL&f}cINDpilNBm0qfj_i{7J-Ak|kPs z=rt(Ya3tdiqy<+sCh43QPl#YhDFIYb_~1tH&59J36LqfPUs9e#0UeO>JE*EBE~%n8 zgC!I^jeWuS5;X%W)x2W7{Dbjg7?xwn@=R*}y~4X{6s|Z|>8xGdSFS={(*5RN+1Gy8 zj|?`;`cp~iEwoJf=l|4CULFZ8{Z2J9Dqg-nLh0$wk$zORyN@4@^jG)wuchjo8*m`| zYYYiB-;oD!CuX2Ni6+@}Oh<^QM|~1Mp%RW$ztJWFCj+SmgY!)sh@HhBSL{uz?xK-s z@}8pKh|q!gwhw*(M<4mNPpWvm(X{BN0uAVigQW2K-X!K~%Clq@ zFF8G0I_YGFJlCE-TF}84=e2ZsjydJ+9EBtI{mt@x#=Zp6kb9futWjC!?}>&04d*D< z<7ee6G@>=_e>h4$8O2`sVx+FjkE4PDdDngXY*a7+A^1l9@fY&>(I|$Ok<&)$*{erK z$Py~M-T9}FhNV6fi7yMdq<(;!HKPR!5R*c5xBP?|F*`#jFJb`_DC?I*Z-k!`M!#_Im{3!db9D zL6wK)ebDtBr$-rTU65T~=F@Y^a$B*wlV?E;z5SLk(K`pp?SssvqhMoR87a?>l;=hQ zd;G!AMJby6&MY_vreyQg5pwrPSt4v0d2xiiGD2P+A+L>)o&dvFW9*7H2wEJ+ z%V);`B43V?r+RALzjKb3Hx7||al2EOTYxb2s?))}c|9^Sh%{@V- zKez`jZ^PYS%>cQATUvz-@(i=n7@sC!Ja>d#1Y#N1g>|z#TOC40-9*raBGimC<4%Fy z3H|zz9|dxWievo6f-ZC zmCi3HrI?~0v@j1Dc})R6B`$Iy3q1;DkO?eed0dgl9Sz>kZHx)L<+ST2i2Hw{t4lyx z@fd=UzjX`*@}tpg?jJNF`?9g}lfJ((7I-OkCH`97r$c3Iu%gAB+CqMswEP=&rO|dk z@i1|IjQyH`|}ch#?ojAlUq>3#pv<4Z|5d zf&~Y;MDeJ@F-#(l!R-nppYyxKbdxP}qgjy9S4yIk0?-7*;vSv?ZA*2IgRW8C>{|KP zE0k+@IWd?!b>=lSf%?Mmigx#zOTWf>4den|qFv@OzZ;@%UjIAvC%8*iFMa;_$klDm zLHD{gxwZ`sVr`rGTYvenzx>c&LuFRA$+>NEej5f$tw{UD2>)v>AGvI#zhWel{?ll) zbc}rdt2z3>P5Bd}<-XBfryTL+i*am-RpaHyCV2``@1Jjk>kbKiXp(Qn>P}cb9%8Cf z{&B0^(JG(1Zgtj$Z)goPS8x{o?{*;`ROJA!stHwg0Up`1X7DY4STMcGZ7h9>R3$3eMzOnA*0TDfivJQ7?hY(a_cP7Ml-;Mq@ha@|;( zz?~e8-j^MBQn6u{lMxQogzX&xQiYxNmE`tSveV9sw)+J4RpcO{(OWtJP5z4mbj^Qk zQ)KsZ8@@s6Xb~|L{%+At^xzn|eN5LCJy$o*M`PvPvGU$nolbSFH8-}I^=)jlE_=~j zGJ>Dq9vE3Y5GF9{|9P;i87wcmM|NJ%%LaF{;jU+m)YTqUIUYtzGBkHfO;Dn##?h_i0XP7APb%yy*5`HdaxfLzQLajz8b%N)f-j0cpZ-&TssLABQv2qz&8z1!<(8`8wD=Ww)D zwu!I-LZvfdKV_*|XO&geCzn+$g}~)kv9O2tNjsy_l=M0Z$Q~W^Z5200;Kjvghg^W- z0GB1(J?2(<4;Z5sUPazhlkA%ltj1yNm>{oCknbiaIc7tteiGW`SOc98!gSEDk(UtD zXf5XVG&0E6c37Yt6Tx#t_z=lPi}jW)B|4w{y{x}4Z_oY?PpzO)2sxjkj!PC574OO= z=Rjl+q#OxaoQYN=iCl>joZ|qd4q_Jb)=Wl=`2KkF_wn-b*sf^gh0zY0LK27pRO*ON zS$UmtA6>RLZJ@Vua+Ja0EvEaX1_tP)jstKpmE+(Ue=*yu$du;>W8SJM!0Q8(AyLj6 z{ve~0Tib~!Up0vpdvcNzq*o@x>Sk9tz-1B1CvHkrj&n{h%4%sYOkfM!KXyp}AU2 z7=j=uAWjZ#Wp|9Ze6m?PSsdT`$%)Xk@2 z(PcA8XQ0=H%pKC9n>#jEPZ%2O!i;mm7ybqqf`g9jmG(Bw0Dj+{&h|QK23@>=8lEg4 zOr@_1MbAzp9_Nl}?6c3ODd_E*&MAPXU!DWNK>8^B#YwOgV@^4mT`$~0F_;G|DH>O| z=d$VLZo}^ugg`XY@WPzUHpb^n<^N2~^PIEV#uYQ9l>9r&gE-pBqq4$#V+QE_tSPeH z7T%$o@XM+4@l^R_Dg$BmSh5Ok!maMwKB344JS{0gd_O!$L`Rf47#JIrQvM13x>K?>NC`+^X4$>$O z-r5spz>k#>0;%$xvfJ-*E|Ub{DYeybB$Z=tGn<8&>u@v7_cYDJA^HL`Qt&4#m9?rZ zf0DH4Bcm>WkBzbxg zH{*Tlba`;PTRFuK)F|E~m$P!s9!ygS7vRpX+n3d3u2jI)fJl9p5BJQRzL)$NSeuNg z!=(PQhuph|y=M=3cn^7%tgD5nebjWI=;M<;c%DAs;tBtqsXBua_V$vQa^6gNAtGBQ zvl%03ncxdF#}J82Kb@Z0$=>C1-3Pn(lI44&BK~HE`EG{%ZCaO2;4hsbCr`1^EaER73~wPuydnSs|vE=CZ!E8D{;Kp z5{qw+$(Gn}T4b6VxlHiGmZ=rXJqUXj+*VI{o8;z8%uCUF|}9Yg%iUJf{@Et8r*4i=KXA zj`0Z2Vc(kEd&XNcEAshRbuYVQPgc`MYeDD-osP(ZHS$1>w(lM-SIwmR+iK*#8tt&u zAvP7*XKLih8lg4jX^nupsameDmK&;N`%&`FQS$at^4?MA-J|4#qo6fcSIag3?3YWc z`H{3|HTJey{!QGAnOIlj-;n)uqek9cxVK`cYi7xGr$H(+V@%I%$V{4CI7hM4fV5WR z<4#O2$%Xi&`bYA~kYq^1YB{Rw4io{v5%?43lYcozVlpT z2=qH`Pg%LAHPl8W2{ZMVDuOmjX6>w&PpjoKJ?s6W<-3{WgIQB!F0PTwbUrSvk&9}~ z1={lMLb6nV&yqm=NJVVQZGpv{kOg$XLnHU4eGRK9=&u1^S|Fr2qI_VPpRJ=@;g`Ee zn+&kuT`5be{wb{dP$@rF{*zz6s$_-l+6P#9WgmaJ8aj+UvD#mn{j^NOOE=F_Sb1z8 zt6`7>{bo(Nn5G99HRYu!#E zFot%Q%bGYT{>EJ|CL($2#rRLI7`7R~bV7<*@Mf2WKAk+7cGMf&OZ&K!>QgcD6 z`SO1{CHNx664a5K#etryx1(KRky9pGVjiCzeQGvC#v;Y@PE~XFWHT*m^&VkoD|Z%P ziDAyUhJQL$00sv&DO5uFz0OC2a6Gw!)N|Rz!OW`mUV(gZ4UPypB~)+M6(mwiPwD5d zqZ$Ga4Ccv!(Rq2JYLKHZp|^4_m-}Yx>|oPuxqr6OeRKD0ge#Oo^ycw7?B-AA@bT%{ zjt`w<6S(;2g~~z7(>EiDx)44!jH0N`#YVzRfSK^!2)5Y!{moXk*gl&owe>TXNL;O+X!*1Q*Jh;EyzHe9F=gz%l>s<5T zTvCM1lym3FV{>7lKrG;dM51LT4Ix(?+0eWV{zKEs1X z9z`&gN(=n^hu-eCIdbD1SwDxF@a4k;%qIuP>s>E5J7&?j2h@#7@d5M++c&XY6lnx{ zzM027tUfS`5{(i3G*?G^_JPtM^+9jbOLcC33vMaCY@!x|c?f0>r1?gvJW?w49BqLQ z-xxP{#pPZVdrWXYrqu)F+XLjn*>cHj7=@>Bd*k@Cv)j1pX9LdZy=l9H!OWQDyL~7On-{OXzEN z8RPf2v&i~WM5d!bINkNI`$vYE;fM{;ci0;H1`L|Y)yjwKwxu+0hZrBi(c<3@7Ee1l z_qXsk*3tVOFyFb zGA#-m*$L{Zb4GVx!PZovU0=w42O%meW{dbZS42;pBbUrE7i-Jxem`4&1huj{37^!( z*yI#$%I?g`c zne*8fL@sEz6<_K<`}!PtW)7Jp$nl3IIW1bYnKAmSjn+IM1H$*;+ z`5(pn4`cF)7NF9M9m{nc??E?ZU4BY%l>|T=EwPp z!R7LKvUVQu@O!F_Xqs0!XTILyk_CQ&EOQ^vT;TV1R5B*_xC0L^;Oh@$pM=|JBo@m( za@+i%y@SHN$wNJFl(`;kpbJgqNA%)J_$`%U(vHSyK`=>%h^86!`@vU@?iB~b?Avkz z2M`!F?`98Yc+A26{*-B{x?oZ`A?-18}zQ??KH6IsZU8??737AhaG6 z8NjSrXRB-pIT&Aeeg0s6pfYJ3>vA_WG^t&{an!cOt#UR^_no+DLV)uzY|&JNTA61_ ze#y!l@y#cd^D*6!5!)Thxn>Y4r^kxX3+A0Q@_`PLJqA0Qm%{RQg|K&It0JHS3`F1v zMxkUjlMwgQn7q6iU+<$qmU&6I(z|;hQ}N*fVEXihD4-xs3nD8-eqKm|w(fHB!MesD z9c(&F%-MEvj>KU?Yv%eC{-b*@+-n-}P#pIQjn!xIap z{Y||G{dz=CxM_i0u)qL(aXFTY>7ShFE;w^x?(6@SZ}fso7{S4E_QB??gAFaIU+Mk> z3++aOv+7Ipp~u%ROy9JSMJJ^m%c?7Tg)-*&seW{jtXp8$E>Ij|)rs}4g=l3zS-{Dk zAI!a1E|OOl(3y=3fu(yj{Q_d^`uSY^{X*^hDT^rdtJKn(3k-N#K`x+tg6b3F<~CF! zio1be0^yler>C`vGH5(%m=w8Zlyc&N66YI6c6bzy;K*%jit%Ie(Z_JCIoXpIdc_g^ zgv@}jmG6TC=Od)F$c!>0qhl1R?YP*)^aPTMnDk_sW~U}+q8*;@@5L{B<85mE{j4(6 zbL@c}ISAbjsB15WoG!#y1<=l?uN{}$Nf|&ZkBRa=EZsWcd$N^4@bcI{*&Yy*t)O=n zH~V@~gb)i^w}^>+Zc(sQQ6&*}_tNV8q~q)M#9 zeXGGVSfcryuw_4V0w9xX+f-fS&>J~)H7XN>Oo5ZXQQ?cl^GJ%j;j8~kW*Rq>by=n) zSr0GSYzgI6W6@GF0|egrizA4c00^@R<|ihUU8C0MZs^KXUKy{Bi`KnDZ`)#>Y!%`O zJ*NI0bESfFJ)AQL&QNUhezP092mbQAaj>06oEXmKk?guBvNTGq0DF+4ECo_2;<8v3@>rkv(Y<`q6U_ zv2)EF|NKKz_bm!GE-Jcjk=?k+LR>@en}ZUEzM1%_+_+HgSlC4O;Sl{=g%v0ok025g z(%y^Y8$`I{-1XBT3LYyCwf$`m7$gHi04*pc*9%bvxc7YqOpMTmd;OEdG3?n?_@b7@69GT}|go zQ~0phf+BWz4SebGj*g|+EGvYqZR=eQ09C&kUWMwApGW*f%q+>_FYBECAa9#VvkED1 z!(lv_?IG_U$|Rq37&?*MAKnEf@1bcrd1!tU!^%M<^vxo>4<(+$EI(A19je1W`!M$J z-`q@mc_>K=aO7uzmmg-=92O~OxJk25R@xRzE{m)Z4o?%^m8?P>eQzvKH*-jM7P@hP zw590^Tdp*-E`tZhv^ctijQ8_l6qf<*E;8pYQbOc7szOW+S+B;PO3ur(gdIj!BMy6c zOON0h3RuF`pmSm$M<{2xHfk=7BI~S=Dta&vMdkBDoF|B;$k6$Kis6urzhP<0I zj9)pKo~b=d2!qG^F*e#JmEFCiM@ER^bIU+`B@|Rv!9pP!A7Qo`|IEYa-k?aU8K7v* z3?y@oK9K*ZA)1HNOwa{26V2rOak2(JvREElEPq{Wo>(mFj>zn?=8na3`(n99q{ZxE z_BC6HTv?2+VGqR$yjlW{sdb1`E`TK_6v_M^pmd`xU}ChkhenU{WQM9j9sDEMB;Zwy z<^gG2y};rH*~A0RZi;d(t5DXQBb+bgF|s%LQ8hF!m1A|qqM#==@M~UO;+-eK(^h)- z9;s>Xnz)Q|q2E)z#$j$l+t|8VA)55Obv$_i@{&Pc$Ee70^xV24DX%g-N{e&I0ss(I7`+mD4PLF{$wVR3zI(ELqU zQ1FFOJ?YZILAx`uv~Z|vqq{R0+@?El1LSgu7dd-aK4Dxp@0O#H0dx?cc}Fq9Jw;UC zLlUlp_G44$x5p2crw*4V50@7Xmn)72Wo$WGt2O5lMORmDHXFt6u0BeU!!1XX&89h5 zT^X=uNFv8+JjN{Sq!^)M{(q=4RDns!EW>tAwwwk=pfF>YMG8&_I$2$4)AYU~HxdiPoa;LHkOf9e5)x)D)UPBdba57Gk;t2ZpJ1>Iz}lG)ARBXMXwz}|IgMzbajBn zU~SHEqvK5QTVW#Fz*O!K^`F1!NT$TdOA1W+%hX%`Wz1o)mPHy?lIuB3fUlXc^=xlR zMe&t1drusT;$Ekot|_Wiyud}37!IomlKjh& z^4^j1?va`Ij)Y<5BuT>yz**pKz^RyI;{F0KXpNbvh!m#@dyMBb=%68-;~kP{P$XeH znt;M%)5d%E{)@`nkZ1JsUGQaUD*6J4p-N;*{gSIl**+Bh9u(l@uWt1b;q~JGE z)Mkf&tZRR}-B+c66O~BRi@+4cgc0LaCdUH~~=$mM?*yJ6C_xiO(NBNslV` zcKDuJ8B$E@zMJ9!Ro!0);BoMr*+C;GgCK`~I>dfl; zbUYx=1IBwR0W0owG7%GU3sM&D^#UB8X3Xh$=BquPo8#fES`K}p(~K&2U5{NWWAI}2 zhVU@nx<0HQ9PbV@e!Gz!=GtuO-LdXV>vmJllS|GKv(l04l|pPOAr_%)jcb*RuawT8 zQ|`zU*&qCKB)w81`BS-3kpIkO*r8tZ_=1Orl2}OSf2sJfwg>Rq^d?SE{p17=os>%gQ z({h$CXZ!YP8}v&r#Y;+;z}73gYT-ceH13mlagt3QGJ*zM)H6eVl=KfMhJB9g7FlY* zTgKi_6=EW^jjS|sB^e>@-6nmPvDY~Sgk`R!V`lb|(jY8Xd;p4I^)&l?cmGdjmN}9v zFDA-QA&SHS;G$KOb5I(~NOVUrGL26I9%zS)O@jT5kLA8ifS?B2*zb;+Njstw64txL**1o`7&8*A0)~r&BNUoGG zb$>^*KAFV5Ih_E9fYSo5ZSvORS!y2enw&m@C?DXa8fEe>HDV*fX*S^AV z7T|&B8!7lr@eRKVY>zj$DI~?^@u1)pWiYn~=^cP+ye8fX{Z(kepafe6vI3G@a7ym! zUmk%Pe-cOPmJ~J+E5^TMpq0};0Fmn<@=tQ(04TS_hw2UkmiUl3Es*oELW00+V2$*Um0eIW zFrV2dW;To-c^j2>b+?<057U?BTLDn8Fat)Cm7oMHoJ;ow#b)Cr)RA3*6Yf#np>>-o zAViSMs|!d{oUUm@mc$zai;zWE(AW|*_b*Y+f?nZQq|0q_72n1X%byKNYTqn)ay6?a zdZmeu)Ht!1WLWOKk2jNLH*izkA&8J<9Du;M0Tvr2ktmR02}js^W*|F52-T?ob5VB8 zfgGv2-ncpR7x%%N#5BKQ9tPY%xu2|o)hm~D59pjf^DQ1 zC?G8nAC4zOoF@OPBKNoJv|r!HZj$TqLO!|M|u%a`g=*&QQ--}#BTr{Jcm zPdS6&KD(aBuzz)I+#2aH)iG1$*_i^}|MCPy66%VITZQ4yF%ON5ig4QKw(p_Y_*R<2 za|OxLo-4sBF>ZwTmNG+%9F={hhO^wp{K;Jkhr|Tk!#gaUF?Vu?`eY;1*Qtm@byR%< zKgaRxV-htRoIN<^osU;-ca}UJjXD;OYw<)XKpUVnKBC?!5}T!<10ug&%pWPOc{^oX z^~lV)H=ZC~GLkX=BGi?{`>{mmo0$Qa0B(Vj1U`rlN}=hp!oS)bLA5reGnklu7_p;y6?4xC!JQ zL!sVXXL7AfawNg3tNEGIZ?yy$NOx})$_-dSZVcm3k%?&VY_x=%bjn~oO}WdYX@E(9 z9BZ&NfMT_3+EUq5`FNkzIaB6)*9paJ^iEV@CqcHmD=RwMj49=gzPhGX7+h5qbc~>v zvo4(SSdc-lah6+LCY~!nPRGmH^UN`Bt?Ye9Tf}=>KjTiG`+UyDpn|GhRc}- zwPaiPY0w{GDI7_<7IYmCFjg?Z834j}hIIyiPaK`3_A}}i;+Q}zXzoQ8jE>KNp9s#? z4X@b6%DK#k=WT{O=&?fF)^r{(7VLFGPZ__! zEKGY#qhyIsFeTg{Qc5>t>{{Mlj+hrn!VHh6DnW;7TyH0<^|*`&RhuGpnZ^O%PVv0) zTnj=8wwO~?Oy6P_Y&Hs^2d}33i5S-wuNHHG^yqZNUoDXfMBI@iBhH?eZCkzDa5TYD zMkO_*aAHrCJ~jBY;!|3vkI{^&Y9jF|KgaUm{AO@57{d819*h5w$ire>Ta4K9sICFU z>tz2}H6jvqwCFj9f_>yS5J)xQ5{reBK_c4(C=yylu53I0UX02sQ8}p7seY|(1cRyV z3?r@SBYh#8JrIX=uHYc(ks_RGym+fAdY0mhh6F#Wdgf!m;9!=XNh73JH;KCjXVr)k z1~iYMK3ZR`p2>Jc8E)+TGmO+ju6V=Vuib&T7jbzw0L(5=u*I|Qv9CnF zCU0c+n`Pt2ysOwt0| zfTAcEjtIQ#^l?;nfmL@}3`@aW=fc?r!U)C%I+~Y-fzT#iL9HS}o9jyOxOUThLpS;- zIX?QQlt;gf9F&^R3M@z*fsl2KBhIp3=i3;v1+5WBpdne*UVAetZ=o^x{>jz~BU? zXsq96#|CYCc{p`Kx=WzIRkj*}Q=_FzV1arBJsqdXnu}jb?o4!_>E=@GV!eQ*EHM5i z%by&p*j3%&RN7GnS79ZPc0PkdK_x_)DP1q}mP_?inHJe!_N(wFML3gKHXQ^(rD)ak zdfnnUsNwLbkFbC-Ojorc13d;_F|GPt>*@%h3EaDxdj|)Tz*XwGx6i%1c?U*{ly-6l zlZa5bszsq&7Rv!BNHKj;GBtbmLF!6b55XeTK<{qNaK3YQ72llBupXV^a*6FH+}G?+ z%AEbpQ`ALL)jyt7I3rYQgG7@-Wg3>G6sssNh=_zYxgI^*f4 zmLpaEX{AwBVF3veM-rg;jBAA z&W+SF(-5|+cZJrl2&yzx*yGu&NnE7sr7lknI1w6Sav9jZ<2%I>eOdM2&$~7z7j-`G zf|z=FF*ymuuX6TxJbYhJ|Hx{q>Xn1dMX-1pjV(Wj{h8iSq8oe+4A28O5!ND7&mjCn zUyk-O0}}%{7z4g=ScBG-uq*$rx<+-;xR3rR*5sc~m)Lz}I_C^vB16@*x+}-CHoBZ# z#R41FiIVsH!a|>HxZhBvU3`BI}R!(O#x+!=aj~!s|6f_$23}&x{w{NwIx({SVkS?H& z@Vn<3f2cj)hWsYb);q8}O0PP*4%cEcM5eunM>z-xTR> z;Abx*kBTvC4RTq!(L4!~1!%sr(QmBXBnAS6&WG&*G2+REFN*Ax8fH!axb%d#2wJhHi|vI6M}vIV zX1~zH61u7+n;`i0a7cM}{w{Vl?>?F%M!o-uFytrLHZzf%!&|87n@g1IR2~B?tr*6T ztsz@U&ZH_}IH8M(fLg6uGDR+z_M*)Kn=_PGCH^nWL zh0+n7B>bJI%Q~O$zQM^1RL>uY?Dz+@XS6X~H}aqP%Dvp5B79f>A+$tLaWlT_?IRX& z&60H7+W^Ha9EV7pEQZH1BlHA;nOkb=(d)HJW_DuG)W9Tu1u27?I4jy46u9@i2NmaV zp$vCB{|qNOm#Bc>dP2xR>)h76U%aJiNuY-;Q!_)-A^M0h)tE`6a#nR0Y3i{Lb2?o+?%}fag1emSDLAWtWI5-$VJ2V(Y zJ3JUeJ2DtcJ31Io+ZIfu9T&{NJ&fA%!1!dOv0tRztA2X z{*LxH;c>J}!avgfA^Z>8<4K0g+kb{9>Dw|1PZ9rQ@(cb5pRL+HBt|6yyZ~1|vt9TM zO7l~QLoH^R*kf?KNI}UeWS$)_{pB}yEa{GY^Pha1=ySq+ISB0xi4@eYo^dC& zs>vgZ8q$$XfL35+tM9i$wg<=zf4a<09>A>c5^;JE$6N9#Dph_pLRkZhqp^LgW_o>3 zU6ic+r|0$choN{yYs+Wt0lFJjO@*E$N`F=vAJNK)^HEBGtbA+>Gb7K5eHL@#WkKZM z?4KOPLW7RNO{3a)Q_YIHs4dzT&!m}=G2uM73I|3BTrt1H`@k}XNtYfUXUCdM_oTO; zv{&Ie|4I=LD9$g=nOVc0wM-Kt_)G<5T~EKeL6;dD+jr;oEu;ZVPm#ch+&K zj2uXV^!Z^HiR1GF43JR-2uW}!yzeATmZeH5*A2gs!Ic9fJYrmoUXq+As42#SnDUxj z<-_c~6GTo`BQSL90#m$N;cQuuc{n1si`*gpIsnG@gNF=FKUf}RCV?+CHe%pYa~!6b zls8kYExHxf!L42svCATMFTTA(u?J5zvkXstWYL%&xLYmIRj#dZZOpZX^he&z!RsZn zP8ygkf4vl}lXS;zLvIIEFMxU|d5_Fg8VG3Md8RBE$C5lxWOKwW(S+67vyr&t_CN^i zBhO_odqK^a&!@$`E2V!iZGUSs;M*%A8G4N@5;6be^0msPs2oFU&yBjFovFoN)JAOO zN;HTn=UbR2YbJ1xZ(ta8->u=N!D{%_4UC|z^935k$q$W3m`rBK8p8RVe=SZ|y7|I$ zgZYwT)X+0AnRGAm?u^;fV)jgO9sqg(K3=t|xn@bw^wV$@rV@o$;dWJfjdJa1*Xj{( z=0f_==DmWRgHss{`%T#jSTSW|wa=h%#DL)dDq|v~RV?F*3&hnp} zbw(=3Unw5ci*;r$X80BU@3|++R(Ds`x!?pvsGY@h~#G57T0Fta)@n%Woei4Ub(zxgY4Ar)KbbN9gG;JK`LpwMljv4(ZWm#GJ+9m%<<5r-vPt&S-IRe&mpWL-{y8ebiLvh z-3uA+Tfi>`qJf-DA&*|191Ak~}A@VnvZcG~%L^C1S8{3`Eq%}f#KLYLIPO}z0gq#F<^}k$v#eYU5)Wmi*0c@ef`qI|$m0n+u@ja$DIsSi z0OO|HesGiAm$k}E0_!v+?ObOSW(oqiTGss zU_bwieeCMce>x!_p{TTHB+W`ZZ$fh?yF)Hc%8rB?16{g3Y3@woA%O%^YA=tQwN&Me zo9l?CjJr8l4?9c8Pmmu%)48K}gly55m|Y?g70B;P+B=g>_u=(trhDE& zx}7A9yG;zW=363FL|Y&wiQ=?NkigWGM^O;K(RGtLz~&dsT_ai?ySMEXgJWY7`z2w+ z1H4Is)?ko7q-?Mmrb0YGGuj^_gF}9S-i*e`*XC0`H$Kr!5VWB(!##g$be3}49du#O z%z<{EKi?b_E+j%;gDe;6z2wkXiC?Tb^Q9;*cPoYlHz&QB1J~mm>(F{PfO?%Y-v$=b z4Z{XnClkG;Db)aF$XdxvM(mYIe;>kKtlUJ-!{GX)e^V0X1$hWJ@?mBW!+lZF^czWg zZ7R4fg((E8!_sNZL?8yj%sg=?-98C36-fy*rtIf85VVnj(vX}QRLifMz#{_{11e<+ zx`9udlYB~n%nK%h1IG`Bj6+X~07oXuL<1M ze>us^9GviG_6Jj;)_@neL!J23@B?ZK1)Ff|TJ)6c_Sm@v_5k@ix+o?zj-k^)qf}i^ z=Xuy9l_YgO(pf)@QjT%YpdRKE=X!d>-DlQ$s|sOF3HYL*oth`8UZ=cD`@x|%bC!2i zA*Jil9>XFKRSJ6%Vi_xky|}m2Rm@`DIOZ@p${$(cJ%!0D7L)VR{%V5t6UOJ~99rYC;RXDo4@%1o2iAQf^l(M4hO$KG=i$hcr^keM=ZN{@KdZ=#Ie=&0Uu{ zfz<)`ApdVGQ}wj`tw27a{-r)eSrLnBV}RxmIJ1ud&H1EoS~odK@do_sb^F zwSEa|Kl>Mv(@oK->M!u4WWHCHl+d=I6wKURo->)Jjo-`fL*17AZn}e^d)|d&>U6Fm z=+qT}p@0Ym4U}mP+?Folt8!`);$t)9G<^-b?lO;vdl9Vi3VX-+7AK;%JZ zlWXwyQk94Du9(^?pgRL)1!;%W$F@I{N9agYkd&p zWd_G$t&!Lt|BgX^X~%;abu(A^ahFDWsSHqcR5QRlqY?JCR4p>RH+iTV=?dM&W(Om# zbU-78(rQ|{O26x>5tRfoFTnfe7GG|qOdS>`s2AK`RY)8k0{l=hg> z1a<1_0rd$@FNpswQU7+$t=bbs8E+B4+QxlIf>Q}Fa`+2&51~+W8s1Gkb=Zaywy?_d zI8$H=Up_-MLP6?G%^7bpmEKyCtj2d+B*qWeV)j}qS3w&AH4G6zvC96R8`aA0V*1&` z#lB{o$rz&W^vQwkqkdG_Ie50avtq+A8)`aYvx?&WW~OyUB?b+{Qz_`v>G`GJ6MiJ7 zLA8L%)xO-I`Q9(Y+DGjaD-Y{YR2~{1%Iz<(+K;uW`Ow5I>VQ(|E0gI~%_qc)6{}>m zRDtTA;`&m`x~)}jIw(Qn3bx&hOm6hqSwQJakD}g}^aS5?>fQq<_q%7^$<6RmgzNJ0 zb)c(8im(MKr>c?7ACm3@haom{5&2~ya4v5WaF|l@I$FfsM{$c;YkB~pYt7-V-TMkq z_l)k-{HbNG8#WUKbQFC2-1n4%JdXZat3`c}qw4fum0D=XWf7fF>&_wtQUSJDK2sWx z?o((wsPe9YC2(XsO~7e?_}*HF)=9PUq_8)vICH34qJBxGah<<`Jl6G0($zt6UHH0^IbnM54nQnIgf z<0~H*W9D`-XsnF{T%j3n0vQil4QMYD9JCjdp8bJ zb(d9nKS#hchG23eHvT>78(gnX3|NoLe7V3E3Jgs`5dkem6!`{wm60R6uE};h2kbaA z1({V1)FytOm4D#>?aw{8hs zAP|aB6Y@MM#si}?S>|h{U6&{?#0Lk79HN*cptq!fa~k~?V6I0IMpCrey|k#Zsz#tp zvrfo(Gis;F)J9&KhPjXTXc)n>iJ$_V7bp_6+-SWu5QToncpl1@kkajq@Lg%Y7iJT` z@XZi`kBL1q8k`m_Gr>rgb_@WQqt&YXe;9icz%8nJ|9?K`oSAv{<-4-O7rqDxf(nWn zE}@v3nx$D+E0+?vR^GZTx8;@yWM*!uk-0%?re1}z=vFI9_MO+ZP!hJ?Vt#3UnM4a zi{w&tngB%=7ar7UCA-h;UY5OBubq^13+@Up0Q`)!YIJ9?Cz{n@&Q?IZ|9H9funORuFo` zm_u+0U}fm?Zu$(qpp;NKCn5v^Wv|J}BUQT*-oURW@W9#Qr2OEu{%L_BMUD#pf)^YB z1LfJ;fma8pb6nw6js_+S0{O&7k#a|xKY_4>UM?Jx3ntRn7oyu{mLj>B=N8RL^g zmZBAm7OJ~y6BzmS~Nd)Ma&hvT83@!@v7A(*Tw~RQd#I($gCVP|o^kP|!z5e-S;kA9abQTY>^HUS388Rv->KS}RFC3K zS3ru)P8SIBfFEmq2#mX}m~>uVdW(vEw(cQ=#sn~^#GRuwfID3C0UZgXa)_*e0AGh{(j%a5TCV>C)Q`4mZXMeu<5 z%+D<$D@@0Ue+NVuGDyW)%~;qYj1&&GXiNL+40~gSduaxwRkO^q85^wpsw1XMi|H^e zC8P@Hi2C|LwTW8_`=8fjQE zC2hgYB5m#nhtLE=C+k4@jx~-r4}(1{d0yI#s2=b{Pq?tlo@j*{(hTu&olnosu%M+l zkv6K$TJnMhtdce=pPgb3Pnb$sLly;QE6GJ&tQLGm`~WGCV28pH{v)BO@)RBE=L-@^ zcPDZ+78%IGPJSGz&*d#Cu*0#yGa7s}JV1tYJYXY*T-8#1N_W69Ny6V++ZvW)t&LxU zR%I{Mtvef0v(n*QkjbL-QJnX!@u0ypxxE{)bA)}n3x>d!bPICa`YsAKEbj(lf8HI& z*mBN7zOT|P{)TSfWz*h`c8k**K=s=qMr3baVcy68qE1PpKceA`%3z9zljhS66D*A&AR%4Y!&O%01 z923fpuuFwW4rCm-k2LbcT?jWZRH2T%MoD>8V2|xK-6~Nfp-# z+G}T2-RV`w_FJ5xQ)+t!EY_2)Ze%Yt(9yiu$e#4s52_Pgx@&|8QRE*m03s4wV&D`T zLb{a_5eDffl{O(2qd$ug*^kmtmxxg7+kg*yk_BbXS*H1PQaO^aeiznNZxaFAfniRLw zk{e`MV)}6Ud7|QfT}dTFe}d#QP5T%U{#d)T75--K>kT^&4KLS72qDA+C-9S&??X0$ z2>Hv}={GS)$TRRu#3DJ*63pytO>}{e5bNz8ZrXjEdE3u_Pl`(Lt|-q7E&cqrRlBzH zv#H1z%bVfsrk$fRZ+mN#R|8JrMsKS;QsYflx!OqkeRDm%sXhS==kBSech`rbOLDK( z!>c;?N_&IIRaZFstNgAKX?CHQBRPULCf9yMi~YIF*NL zGKqa+0VoU6@C_+c%5_|+B>1CJz!$OJ;#?cJs*2+shG7NefWm03aoA?QfkW709>rWwp8W2PLshV2G`Mp}Dbww0 zIcIZO<6x|3;if?&Q#+VOF1OY)GCyiK2fzT9j0r>Uaq&ImCWOz=QCu#;fB^Cx1&SWdNGSS1>@l23 zm;(1$gF(b_8C@F~Fv5OitrqXy!&}8SEVn(ENnKT!y6SvKBo1ld(nH*#*yf1hKC~`R zIZoM@)fn>Km{H#{O|A{u>A}tj?sV+8Ze`|vk=2e1q#khBW%WgPy^CQZsSu9jxV|#i zpq(h}!kP7A>E>9DzQLDo4s&t&_-g4U~HW=HB z8eET!rIi~Rn|2!7oKf4dlH3DMKtK26p=Q=E461WXGnWVZuIM<~l*CCDP5glrgPy!a z>5s(JEQgTNvSLCG6{UIy?QQyIY%h1bPsuOpvC%ZQ(`f1t(%>d;8v@tel1HS@fMUXT zcBAmahmlT6l9196&;ax2FdJKj86;3v@)TkKS$AmrSU!byL>3ach!%Str<)x2aqKtw z?heU$@u`-)0Js4b*gc@qSk>SwkV^SrkyINv%9y}39Z3=p*~Rf^N$V*NFrA}AwK`-p z;0&IO`~ucG)VVuzpNq1KK>Ik(C&}FKE&*B|zF`ERI?W1q=#1z+_y)Hj*J!gCX(=X2E4`zN;!T>> zu2S2sROR05NF9o`i96t194>9|055Eo*X%g1+FzAI`;Dr9y=re&{a>pfp5QJb6tMJt z6at#Y;$NsvJjN}wiWVr`9RLP4yFe=pD>>r_TB2nv0dZC|3Lj3_S3o(k2vo0CMcWFT zS zSw5+Vu$Hy^LFVaeD0k33Di;|F&&Y6xQYC*?8W#Fa`D7BL!-shx2m(3#Rs>raKD>gU zz`08-xr6Xw=8k!XIybXHH$0bLT3{4}l=Y7;C?h6}Q4jVT?#y-tu``@JpUx00&7x}3 z#HJgEbryjs$^ib$5s5A-tn@)%Q*%Gi-RO~B@e=F)SoNcokpjLf z{mC_+7V>)~wOf?Vftf_*`e(q7JVJlS(NxW*_r+TN2F>7|;MSu#;`9gLN%YWe+~tTJ zVD=iC!dc1O-IB}1XX%gMauc;5a!@ssSYp*8(e?^cPjMq2lw@ z%%eS;9k>4xxBJBHdzWEzn_Jq;Q!dZ#S?6PWmG>MEKHQpf#`pXlYu#<`9<#fxFJ900 zBsXHusk&%;<$IFnufc@gN$N*fNPALP^=GM)G}63*^Z+|Ob4xS(ws_tUIBJDYGC{1v zPr9SCXzU7iJAa?VH~H=L+LJZEj_B?5HkYnqWrjZ69cSj=j^*Zma z_kk^us}ac>-q4|s==a}Jw=lEy^q1*+GxhtxV`(vY zdx}RS$vY`ZG6Ga~Ak#{{BHs(II2AW6Q1KKM!F7a}Oc%84mCPAQlEFV&nF7B>5*t)0 zM8+b1BW3qANXpiN!m&E@9$ffi0-L_*+mn;|!=qNphEGM2fK%$o&G0ANOeWrO`BN)4d3kC??qa%wxPq8Z2wm^@!yiOLI!G3ifJ=20XNG*EYwqQ#N6>Yg%U z5=P&+sK#nNrJHQ3$=u;_Wep@(N*zKz&HQe3LaeQONJ^g7<$CWG3_TpR5S~JkBr_Y( z)7+f(2CD{=fu!u6EbSsLouo1^hwxJH*f&^AFse07fASek-%^Sthh{+(Zg?0EiQqse z4mKW_ztk(JbfJ~hyz()CdUHSbp6=wAC|AHgWw*G?E$LFuH(zFA`lqU@ zvP!fue!=v=7qD)U(?RxBK%+-Zf&GS)i9xY#kU)w~;Zo#^VZfeNI207DBLCds^i`gl z=%9RZ8R6pBY8L4NFqmUWxV`v_I`Q$L3#iR>3+J2?Br1|~D?3tD18iJ*C^;Y(fGp^9Jez8&s&b^GH^eF%WtUDY8q5$KHWCO7%xt>JI z8T@dP>pOdX+l80SWn8+8Qv{*%Qh2+F)05|Q7jHNbXBbYGC^|WG^iVts$&HWr;qm#VwMNYtzoupDTJ~hi{eeh2WJ!BQ zhlolnhG@)hgJWA*_LhCB9Q7xHMtCgxfIGiJHEVa_aUlE8SvO51;h z!6cVdA8Jk+=Y!p2m#xvAypqz>6qgA9n#P?pwe;U@PV35VROu1PvIs{7V-Yx7P7B<* z4}^`Wz|ILB<40h5OXK*C#XD|J<1I2jnv2`oS<;%qtKriQ%5eEpDV&r0bKC-;cCNSk ziM4QNWPX5WHl#$?!%^4cngI%t?Lp|Psy6{u)nrbK<-qhH)yf{V%i%j`$(&N1qHvv; zq0N)3)-{>j8k3rKdP~$^Y)qwdy(@WuR3l8t53+D6OUJ&R(UKRPs3sv{qwPv2g!$`c z>h*3)rl<8KqS6@nU}_ugsmwjC;tWWj*nm!@Qd(-4o1!qW2S1=ptTYQPyOUD_8sRqt zIu#P#k*W6m;23qy+sB*}-rW};T45_l(Ggnx0JbQElp`@VvWf1h2EB}29L5eBuIv33 zMMS!6w4xLa-4DpF39@%nCUfN>?{vN|xjH7M)`HUVq!YLdHQ=v#dM}@byS8Fi$Uty) z#jR<1U>gwv5bzAc6`5UwfSbVr!tCG;fs4L~)J|yGi2zZnw}s7|lq5UH`BJyILnbbW zghfa{^eUrq8|*5X=O0j4cc=VG{HN820-I51Mhmf613?gYUcUhcX16SN%jL-Ylg#}r zv(;tyWZ8YB*-NPXbj`Q@xa6)W`OWP9RP((yvo+bq`95EDmud>5!Kzhq7o>I`c@F$f zask8D7}Ui)bAy_crDtG(ZoA{!{+U+e`Ih|}zjWL`JYf$4iqsxOij%wAXAoaV$`Ynx zP-9tFFfRsV0MYihgU4jRBaPsJuPO!8U0<`eg0lza2uFiPb--mTjv52$An%w+ZC~#5 zeT`o0CXo|(;&XCsnv?J6-(&ian>_D~db(lF)YBo58pd&StLj{n)vU9|-lr$z@2Mwu zl&QkE8E@mQaeKOQ&4s&jcVBMbDl=PGm)(t-yBU_6Z?>+~Jc#>#T}mG>;f!Rr)a>?} z-Bz&#U*`n| zzN_MYmAkugd%fg_aM<+~Khr%}GJc6map%tE(Nc>WvCIxI(FnW>y zBAO!l)$M`FJ=pzaBX(pD_b&T!*UN=Qe3l1Sg&tjdoSjDn_gRlASDAm*Gou}ekGV9BwPQ-t_8z6h!NF1A;8?bTo zn=M#(g0jdhw;h&A-PRMgvz6I!PA;!brlazf^6Ue{nl!%aqp8wrZgo)Q0PN>p_uh;F3I%6Iqpl}^w2DuWySiZU5c`X`LY(7`OQ%Z!ieiSXAwxsU5)d9X;Z;HSushLc9b0XACFXe?#e%SxA0THJd|%;=oD zMz}Mg6n$-MAXYt_oFAPmYO+`fqYv#QBODAK#-MG(Er`KKffN-gvBf)F{H22 zxj<+c4yIqL*$ISIZSvc{5O@NT-!W9uZsdysC=U=!QmU4DmmfxhrO(hf|3NuoM8r^Z zD6d8(h`R2YUCIsHqpN)H7ZqL$DO)4cV0n$`z?LBQGMFu8zk!$A&}S|F2w@w>ptHyC zv(eW%E|_B0Yw~y#h#0=-%*Z3YCcA7M(Wt+ z+YtrWaXt3uZhNgeeN|if{mnMuz7d1G(T%K};b=?RTYAyPF)6k#^@CV3q8dE5T1#6d z127*MD`KWtI-|4g{^a(ajiT(KyJ5keGbEm}ibWsds_(33B4ZQ|bzR$TXxm9W%*>Sv zq{;$*D*XFePyNgud$rp9W{&rfJ*xI@T$r~i^;PiOBd%4$4lErGczMwjlIhoJA}$}w znZDs{bcEVvtDPCp4bAGNSsTdOhCwVR7ZUe`Cee=TvPaddH8YA@~^-0(EIRQg}6iasrh8jt0Slwp!VSbXp@{oMhu1h;>)aX_M`mhunwwhQ-p#ioZ2q3f18FXRg6hgHc&oGf zQoBId(52Jdwa$Kv6{=+SqTnL_YX|411gsV+tzK9s4oAsK(S)s)mtH1P+05kR7ezJ{ zWP30tD{a;aQ0*Bu_6Yf#+WZm83qU56UqJ!tAQX^)IhwwXi{feV1d=*HOrJIy^|n#3 z+XT5%W9BA6yM$> zviX6JpvLj>NA%7wCCeLjZz)?+4!^FEVtG-`v^Vx7rq2N4H(7ehtoZ4kE5XguiY3qj ziB(h%!t|;%x7nDut`o2CQBJ-FvH-*B#I0P zrv38y7g(k!Y?*3+nbp@TTgdUQzQ){If(Gvb=DEG%7G>^OW*q&50*i}>6CJFFw{%7i z#|@PQ#dN{-^00+&vb=bXcYWtt`8-mFzRbgb-=6F&D-#*NX*+Ss()i+A!&t<&IyXbU zOS}`i%JCRe-t06bjg1)UVSmxXXpxbFf%u%YHhK;IT*u?*bPi(+sb(Nv$2RAKCx!oleI_H#wXV2w z-csnM-@?JKD)AMsPs8WZv3Ri2Hdgn=QHMsrKf=Z6fOn+Qw93=%t4dqA1GD^n~n*xVAUZ#bqK!?-q^EU9%qEk4aeA%&Q4B-K=)fj#!0 zD4sq<0lK!0wVfa!yP@qEP?lz)oL{x3CuNtG;#mskB2)v#uxkbPGavS)+|lxx_H%xt7$5ZnnN*4&}w0|{=T z0@NVT&`lO4b^z2;9gXiWU}HA!rgc;lEk1Y(rpw*?qj&JWaqHuD!ZL`$QftP6-ERMI zz!;L`Dp4-AK4T)Ftt40 zr-heYT`bK^Sk|T16$fGa6oB$?B&k6o)%9UcBmAkWeUv_dQlnX12DrlyNksk~5t%zp{*nvcW zOKlf;7UT#VWwk){FM3dEU@}D|V`>_Sl`H1t!;cy_0)PXTQI{{w4#byWVfLQ&)FE7I zUp~z`K2+bbarj|<3*_@ynWO`F@RtTz`K01NLTfu5E0*Ho4@Uy5t8#1)F(Ym!XT}5R zU_1Pt$D+*3adwxFXH>d{zp;Cp`_jvKc%YQtPkvI28^R!3CuI`IC}~8~f2$Op%cDCh zOG67v0X$&4K6el@W|bzRYY$PGG@gqpvnmjz3VGhopcTNpM+D#{iq~H5vSk>#mRb%{ zu>I?Ntn(2R*ADoJ!II|BmQvWwKLDj*w+&v}>s9#LD1W`KWL$+8{B_CJat{<}!#&i$ zWaydrEbq(56N+@clHjn&i9oll&2p!U)l> zrb9`xRhaHRzkxXQyDh-MOW~DpbEJ_tE^5MfRn)2zQaGIis&&#sHc8~vme#kHRRzRY zX$5*1{QR`!fr>mi`57E3B{KwpL;`^K`lNgWd`8oV;beoV3UNtd-BJK;HjPINOKy0^j5#c{3n&Ew$Z}yHFF$8l=H=DRq|!0GEeGts3JUY7VJgv<)6$ zT;#yf{#BDVtGRr$iaiGHkc$=?P zQNdLxNBzTDjqoSf!8qvy9>A(i&>d9=ngfwp>kF*7Lq+3uw(l1E&v3(6v=@uN4yhs+D?pw+&WG@fD!A)wO9gbvA`b-2mwtPh{)%=ftI|>Ic66Bs?CI%;jBDDOTM!FZz~jM zyWZk%|G3Rpl(jg9m22oz-4|g)_Kld;o|W0zKWmiN8Ahu4u?6+yEZ%h75sU8R-?A-E z)0qH(h_yL~M2Vo{TC%Re;u~+*;pHbdq^0u4y}IFk+HkALugx-I_HL=#L&_+1O3lbZ zMaN*m4qz1`YvIMRb#@ZYa4gh3;Jg;m@sDeEeJ%RDihGQIXVr@tg1;l00}u*Wg)=)S z+KFdob8Kc@E+I^_379E^ecXQ*bHxv9@Jjx>b$3qP&aUSdp>WbFRb(wnB3JuK-L93R z(j`PVV1nF~0n$UYA&DLb+N?o7{<~hF^{lISjg~k&nS#C^aqJ8OBCvfCL}; zFz};1W(^}R792Fpc(bcQW>Xa=RIgzR2pszAHpQsA_wCClGhOGC_ReBe3S!13&*2va z?MfBss7^D>@)wH8lp0X-3@n7 z!>#aPmG?LL^kx_yI=;RUp2sMNluP0$?B7%55-VNuCxzFfY?1E*Sc4ChEa6b(>ees$ zm^4}VOI8s(_>I>6gk?zEOCdIAv=w}c0~neQX0zZC2c;?)4S15Ig4W6ki5k(!NH1% z%k2dPvepO`8Qc-Nt!G4&sTvCl%fK)v9}*XGF#VSHHAug`l6o5QX(;;ZBzZw#Y(ej0 zNptC9%a))UAydbK%7tbZn2X!5B#(Dx-|TwZ2rNN690OAyZ`yB~whn^`hWA=}lA`(G zk3cBL<3Qd;T!Ebl9&)Q89xRR<9h88_vg>8MQlb-!5k%33!p6#yW!KtPYlgb3WD?_(8@jVU zzP&3?cL{g;m%8l5E*==I`L;OE&djUh{XE_}DxiZbblE-R4$=f_-he;VW0m@#14;wO zXr9`UN)J0Ss@4{aPS&DE2+T+zH3BsGafm`BAM}$%$($V)8{Ljad6MJBvWIVz*7n?5Y7CWrG zdB%?!VC4{?m5)UUiuejY)dja(^J6_$#RxRs4%A{9iD87F>BUvIKYi*+?jY1IE`A9< zVm4XnfM^Tw%sZb#+R_k=7Q?cCSn8@bBK^G51B*(l@xm9+!TJwC)2UtQH$!-)Rei08 zIB_LDDur3<>}c8yr;XSl=jOFG0X5cI>yi2IcKl<5QiqatvP>#U8OoF_0zEQ|f^kv& zRS(P|0z87Ce*JIp*It7BQX77v<=?da6D|9u0Z8tfwt4cV@?p|E2$n`nQo9Swh>CO< zwOE_NI#;!_Yg+!Qmb zkdevHhG$px`D^bfDjlwEN9?Knj+%LUE-Dx?LMTQoWL?E zp#{y_TxmHRvXal1506HKtk{*46Is<{;~DKg0;~oS2cQGr58X!8Y=S5_C{*%HFqmFC z$V0y;x3-=`8p&t7ZzWX19NU{9ADAUCwoJ(=9TG0p6XN*Ww*J5I1D$qZfL=?xNe7jbw!FDup&>4=_YB-q|aHo$v zpRl=+b9qn%MzEl*ceER>@!0E{G_N{7cG?`EXP z#+&;_VSsk? z{O;9+A~NLPf1X#Q|HBxdA(K8~BC7T4858@y;r4fe^Aj1DIyQ55v<>uus!^?%L>^Jm zF5A!s)%K#Q<5T9IHg`5U7e|c^wlk75EyK}Lou$3RN1|5pW zmAGxk?LgeN; z3<>%*G&V3n((~YyB7X^V6zIO^+7@oAn~ok#Wnomp6{eKwm;zS@!z#O1Bmv6h(g2|d zJvLygfu@b$HXcbBVC`)60<+KBBDoolSKvhkNH4UJ#|W0{pAl*z zgt@b?62m|esMKy$h1+$&47f&1<~~p#Nm_){u+Mc<|k`P>@Qru7$p}$mCUQ=Ph$> z-ci~BukIWN)HWj2k~ck^t?cRWbuITGiLI>uplao0XS)j3bN&}QfW52l9LINlry?g* z&?%13hWT+mKB+so*q4qKn+2)j@rwmNZ@R{-Y|Z~w$+|9b{`+LxvxK1+2t ztE^F}S6U2xsjr<$t0sTxaz%<`JjOhJi8AJ_vcW;JKs3w_(XEw7`C1?iy1X`D@M2pwc(#BK#SaQZ3}O z`O#gU#BZo>SC)p1rS?khEx(JYd8XdoZgEG0GXGdp8H~1w08h54GH5ux2L5#yOKEfwfJpE|S8`MhoYIq=%tAz7D1%ji?25k_dIIl>Lz?V1 zEHwQdTqp)~xqUdQ;JTM#NIR-KrUi3sA-u7OP~mPS4!B?0sK^`1N+dH;TInu#Cs(>0 zuXVx?UN!2^yVnh>yh^r1hk8?atv4h2;w-Rn4fxv{JHy|T-(ACQwyU4c{WJJX=MG_~ zU=#uw!aN>_z@0m}5P(zVLLjEY{TqfW&PsFf)lKOvur8q$WWsoXObGvcTf6H*Au-Hi zx%I2a)KQa@mzgvPAH1INFNx(T6V0edfQ5vIwlK7nC{Y1_ibD)RE6%?Tq{o&bFGl^h zCYxfUVwrOE+L2|F8`?_iU4d(ty(ezN3V6L$ZVc&1&v4o4u6~9~HAm+X14=}nQcqEeg|c-V;{zL)?%0w1yd5obNR7o!es+8@VdIH+2RJ`Mo(SHF zha#rI2z#hjV0F>XkuNn`>?O)H>D!jcbC!*w;;$;cn&1(u|)VU@nxY+7SFqdHhtQ( zk)&>R*;PaeSW2=Aa(7*2_EE+ySW2U|T!zvAkcey|a~R{TDdMK)|To@NN2R^V}XDwjqPzp*fNX*6;x5Hl|rK&Cu_3XOHqR2ucCod%!V8|m?l@LQs#m0GLuri zyAH#N){96{d>Ojy^4N_TEy)|HG)JQg?enB!I5!MMx?z|YvY{bmm=V~DA*v>KSax0v z8=4aiRgquZeqjH%cdJu(cdFXlcPd)%h1B1eq6b9OKg3Opm9e{_6f#Cf{!l0i3@E0G zmct~lgwq;~B_2mS6M8&9Buk*u1T`kkD0_?V(pZBi&JeDmr9F0&F2xgddP+ zkwmF(qRp;j1oqpx*_9%7>co`%A9ku9yg0KZnxa=zTjcA%qoL;i$T|lWD+|4QGPQNg zI)MBPINK$O;E4)!g{lA#ev($1kTex8>R0F5H7wx^rvMmO(v~vSLtj z(u0aDOcD2w*JX0^CJ)vSO<#{>r&!{0GCxO~M3f<)Y&;biI4UO(rTJq`_h8dM)C_T( zUC>pU6#k+ye4rVAjSF3;C))^HR&khb!SBSEn%jyl4g>LG4pkjU#`=@N+1Icn(@5Rz>sCFE(TYR`x8ENpJ z@)E|2p&)C!=pi;r+(|i$O#!iss3+Oai)V=}4Mc*JuFx>(ErZF*;tiSqn+8Y<117W{ zaZT~$$r|LYXs3?7-J*ErAnjdi!=)*iuZikQ9#bY0%l_)a8?;7E>Gm6DUP}bD$iRKT4p4w73=t(2+sSdt`->A7f4`3c zhJW4@W1doO=4MV-Uk&LSF_1GdR+q8j!^ZWnvpUzpP&vLxr$q}oZ(%`^LQ9^j*!|no zp4+s)MZUMNj6TdCL4U_5E741JtqEamP*n#3Pp6-=B6*gFYyoiae=27Fdkz}O8v z3((QCHcRf)v;HT97dy{F4_=JfHUbm5GV58K#n032P&{k1_^j#6NDno?4@Fh*V<~=z zd_@I2XvG@k zhM1AqJdURvxiPHKgTCbASQdt+Q;QTL4<64>O0%6Qtir;UAAHe^X&U1mT$|+|#@SNX zQY_KN=cpg#`vBM{C%2cdKS-7*Z`crm8>MB^+d*&(-O=mSXT=27p#mB@^*qIQvw;v} zQ{htKqedlvDCHv%4xUWw>%=vUDt|3=Z>4;9(40$BjR-n`D~X6ict8U@4LQZu=k70J zb^n%@7Eu7(za!4Cu=9UvekX+O@OKF+Aj*n#Uy#C&c7SB0uCar?u|LT#vrN$^-}Bi0CgSB_I+)Mu+3gK7iQ4wjN zOF*xwGh7qSGad}Q)anRVyw#*7tyx%wgHV}W$Ws~`Q)t)POH{D37r zGI^oA5t}Zw>0WeE%QapOPYB8rck}k7Tb`{22c@I)BCn zCJTf?7zm73W`8V8oA4S%G@#T~gnud;$-PzjF0B0EdvUZbU)%obnzpTNAFzhZ`77OU zEQIo@J~xHh|i#*;{`YFJR;DvKa-WrT)A7FVwbp1A3lJK&K~q(A3SaMW=|sI z2q^SVk}4fI9mQ4aLUokC65-o zCL=g5SqiEVd#2gwsAL1aP^@Cnr6L^%E+ZQ5F?U+Hi-mcajIp3`gcUFX%&LmXpbD(` zdl)k^!fNqjn-dt1p;IxtiCbfllpMGoy1LSBd#Bl5W_Ma8ZoBR>cc-;BoZ+}0&u~w; z(_Gx<@oFh*cJ4?=Y}(^f&u zpXp}(!!?Qr5Lqo2S7F{+&eNEjLM7M@fJqOx;T!X?@f&K2D4L_I!QN!&1v^_fSITUG zJ?4lZf&Kqh8v4O{HUaluaU{F}EcA0&W*H>9+$Dyi@s{Nnq1{H2h7C2@7Ql%OiH$46 znZn1=Y@8l|1R3lE8lje>Fl6?nFvQA9m25(Ih{Gt8E41B0JIY_G$SwYm+T@Y!g^F;n zTTHf|1_RyoHyier9i5y<)^FMOnPVO(b-S9mMC%XeLK<^8kh)BSNF+m-Rn=K|3iCdu zw|BaJVO-o(qS3L7pq-;5N+rN?G%xyKax41&$RmoyVu{t*bA(R;NrA@V-K*e;e&}|-}&&Ta{al2CiiC^sRcn1y(msobG_4d1AJ;{}DfWq^Tdi-Xq zQ%J76HK}ff&P^Pe`AadmW8fwU+7^zy2a4LqGKy5& z??_9RY(t~!G=Q-n87U8e`Ctj8MCrgk5}<&vA&ywDQC6N^#$ZA4@slyk<-b{oaZ2^K z@GoDEc3zNVPuIh5>-K~)jqu|k*FfSwvmEEM=-ipRLD>x;I{c!}uf$3(l$_c(1sAa+`V4r0%{*tsrFl1C6Et^rkZ4u3)TPJNlLg%y z6`DIZ)RrhQMFn=4g%J`}mTQG70pQKUnH*ZrnD$gY0ZgGg}vuNr{8?i!~tHuhvyB5qz9+&VyM)KI( zUiB6=S63J*ZI}1~kG(<~uU#fH19gO(1>@eI9hkr09oU;J5tfadbzJe#aNtQb03#NH zVO-bDo+QT{0v0KB!xWiH(y744bPVR~#H z!68Vyfp&%KJ;^R`E$}U<2;gVFvdqw5*TjL|>e~My4bX`$zu)WuRe`N+vK_wR*ij!b z9#`X*iiy7SJAc2`9^grz$U@#FM1!GewC#bp3^lvi?^HTNXxi2oHfWA&BGb=Hfc|?O zXt({|>;G$ZrmOzY*=4TuEiHhp;V+0jIvJ%YQJRZ|Bf8w-kkFRm3qW0f z^(ibLS^N!NLxvTe4$fptQ-?+4N(JIk$Ie6f3rK2owr@jVS=Cz9vOqf7d&>+KJfOND z`Td&42vBGhFK)#>Ikt#pQ566q{cLa1E>n~ zgP9DTNFJdpZKnux6*-%HUIZuAlGSt2uD03d$$6GB8^1C^{!L~#7o&TxQPTqJ+ux7g zZwVy#D}j!1Kab>vbG4Jw4Fu1#84~cwaNIK#`pHYHhpHRWUTIfOuF=bnpcBV|u%jl< ze>SQlNfR$~R?&sROcd?dN~GWktG*q})GR;~kdO}sB#vC$>~44c4_d$&L>XK#8qz5s z?kQ)XFjXcExuY0ZBLRiO+vyQo0UzfkuORo+?QTr86JagtdC0IT^~Y^1Zu1Y@NBGV) zzOyaIv%?;;@Svsf7<^LLnb^{Vlw8YaxyFMw_z-%-j|eBoNx)t+Dj(R(3=wV;uKKz; zwEO=!w%w&y&As+duTT`(VS6F-h~(&hn3e8QeEvVXB3BD%1Vu&RGQpQEZjap78* zf&p0oCplvq)swKDV#DoZ;!8!{MI#E%1byy|F?D|m$b3j+usHG@BwBUfQ`*fLNs+oyHcREamJ|}fZV2}CVCSTE zcIvjJ3%C(T!-u}5sgTnj!ASt8)2PlDmp^mJTK?CKC>n>?JLE+Ml~`;GL|i0(AR!*d zqZBYbZ0s+kIQ9c2_P;PU-Vf}WoQX36n2nXG2w2jD4~GTrQpxR4Ye~xW?lM%=NdRzOW;NmP07sjpThL6}2QofNqWyS+P(0C>vGILw20)G)lo=N)igmI;#khs; zs`;uhSL6BgX}+pFC8fetfs%-S;TSFm%KDaECNg)L)cGm7EQ6uwUnkk!&WlR_^FU>h z+d$pVd!+dP8fjC|gWIFZ&I)57p4^ef_8)(9N)}Ge{JA8IOJ+cB3!Fo<#hmaHiJF^1^v(6>^-Xm4;0g@qA%*oAni2SlusfFShj+IyFEZoiki=J}bdzeV4-| zvWl+K07~@Y`J{pb&ncf7&&jg!S`M1;;mkYuK`n6ESV%dQ&HX;Cf?huvQiw^OA&>5P z+*6yn8Tq^8FYwR8fsEL^I=OaThUDr;hB}?v$V*>I*F^KTf|So=84kEXScCz1>l!qT zTL)0y(P)>i4W9vG*M>21J3MY{al3ikj*43!x2?F%HrbBe42- z+xFh%YTuerONMPD(irI*1F6CMrytQO(ZGLoy*U5DD68l`hqD26@hcq2XlU4W#O)98 zoszGl60J4}S^`QUESGABg$4GME6~Jr8^B~W&vqBDfElpQ-fQn8Ol=3;J3qpU{yiLJ z&vfqHl8B>0dzbAYcMX}4V!RH8=jF=-;rDaa-yX$UM> z=I*|D=hXaE=T1?(^;8#5bYfWpT-r4*W^u)aO)S>$f;|_?Gyjik*YRt(RGz{+U?z#w zb~ieIA=8avEo#9;K-H5UAsdQ87zH!D7wa`oN0i7ui(_wQ<0`}AerIIhD(>-_ZIO;7 znSppgjTvMGK7@u2bX9sr?Bl8EDt=}`aw5jhV++Q%3tl<;mQHa@vLqQQF=nLm=q?op;Pce6 zfI15#_Fw1#IFko?700yAb>1_Ik9@Q%c--CRK1S1cPn;`8guQ010y9&b=Bni+2yjlS+CCnOPL!1(P6jP9HWFMH`A#P#%IbXPAWL+mfd3=#25xsIyo+KtkYlm$HBaS*K0CkKOR_=~r6y>Uz zj*lo(Y>#}n8)4gIq_4}j-7i`4vFzdZ$+ddnVfO9!WlP^@r@RmEo~7>_%QzP9?8;(B&8^k46YFGgUTPly_vH3%(&WH=VeJ1Xn-uWx4U^MJjmeG=8WN4ms64rcrz)BD zXn!FrlLZ_GzOJf1ti$0AQd&30*moibpP#`-Yye>Na{}OPHw9oJF7j+uaGR_(iY~`m zT#?61aXUVIFnJ6nlSYEs5Tr@=)5&GMnN&96m~wJ>X|?%Bjjems!pz33c5;RaL&&?v zV>c!%Ns6ZftkNx^aXYw@=tnW}ljZda3W>5{z(1~?9Bx}=(+Sd=Jlg?HaqEpq0aPva4>l`W=P_l zIC;@_<=CM50WTRWUmd5XH}C9A$@5(j*eKBDl|$RU7c(_oM(hDU!uSJSiN6*qk-=NS z{&5=tD(o^qPK&{5$@6tWW7`B~(KaGwusdeSMM|F^L2#TKULE0+Yt8L1WGGnXa{#M9 zNyLr&WLzy_@JKXY4nIZu10J}MS3jVWmzKjN@#I>4JEt7Zj^AF;x0B1^r1g+G>>~Uct zyKZt#aOZkOLltw04k}q0>(x-M))|FiME`v9c;{Dt=bk!xw^+O*xoC2fOVZ6psLBt@L)!VRjD+X8-2}_SypX>H>RXf&F!Xe`|sL zZGpS!u=MW3>;fm!`-EOkM~*dHM2iH7=^VGc;b@M7*w<1!u5b{e@P+exf>V2O4%mvC zPj!UYpn4oi$*nqhn5o3Qz!f9D<1d}b&y76W^b#JA_1g`vj24PXLF<6}6edK5T*)?E zQ)r_@9}|lDpjY5XM+#8S|xiVRjBKM%;bTzLfK^ zFQ#8|ZyWu$%_1}QC4U9Q>Cuti?y|`Ue&}CsBbb-*7-{3(fL+~i@y`rbTQpf$<3!H$a-GY6$)GOXE-ILarhdu&;^%_S?;@m+#^ znH+!<8SG(r<^ zY_~(H3FiXM-a<56;=ZhcYppxl8q$B*R{r4(IhSOW*yQ|}c-$4YjWu}RpG682D>ZYk zkR|ypm2ePaEjnVMRT&V~wf-g&#!e=njQS82##myiC{=eTx2ry3 z1#plz@*oqo%Lpn+Tx8ic(t!)Zq}yi6T*%#HmS5%!)dHEVsQRq&C*c$$2|8^A*gTJ0 zJQhVkSj4zoWKrpnjrhn`@>y{BXjWZ?%89yP(NI7M&Z>|p?n8Ry)#H-mcsK)8!Q_YV zRNR>c43l6Vm-5vH=qwTrNN|Dp3JtaVxgOtW6AX2P8mKF^ip zKWE|U4;(SRsIaAg97iA-sRwZRq9_^oVOqEV*HHNJ7Rj1A4;kPgJGGMMNhMJP%0!zO zO4z-Y8Cw^(yUIzTHS*TT=S?Yf_Jf`MG~PH(!Z!<4YpRZGts^C44ESRZcM_sJ8tjb0 zN!W@TAu0U9b?P5jOg9Lc=zrgsPlM>t`TpESw_|n>W{T-=s>n>-c@G|a0iJNAY5iqgC(~| zr$?H{9b}|X+L5*lO_JF&-xcb3AChef>>97jA4Ksg(b6*tC07>_k*!^ln=7SX&;a){ z-*&{vQ^);!nOfD8)X4fFHh7rKD}`eKdJjN;jNm|3AiFRI-BT&urYEe@6Id3ZeKSvR z2wM}Quyt_UuJ&;k;+8P@N^;N+TYG{l+*asIDy+cWQY`K@yqo3WjK3L|SZ(Nd zYDdD~YF=_>6Wy60adRLNLkp{%6hFWBdJWW^X7XD@lg@9H97aDm|d|cTZN_)6A_sQ?chN_PdHbk1K%SiuH*f z+@g{lD-Jt2ir3+UA0xz=b#BP6^mYl06D@Xe`=eAP$V~h-t!>x>I}#D)v!Dleyr2TnlzVinI#Nar8O4XfK&AwH2OGyk3oL^8b%)53`IeTf-*J1X0>K|YCgqLOz zRei;S-9Jamw?e(zK(RXX-gn7*#3}sge`Byn3temu*N^P{AcMdoXgv=T$OB#?>**?J{eLo7Gx?Jz4+Rk*FA|3I18G61a zYj+IZUlsOn37-J%j#tQSk4$02zYN_l91dzH`=o6i@7PM?Bi|Je<9 z>ErwD>lJrG#bOSz5KblZ5#jSk6@O(cm*L#cambCp@I;%GBbQYCWfl7&-2jm8s-RyU z!%~qm$rKFHA-%ip9%|cDZMctf+V<_Xoz)(?V^JY9w0sDFw zu#9*LmO}Y(Zkj#-7K#6Ekw!TUpYGrzI#2a=;iF09PxyRTZ0M1;4*R2!nu(|M6xY=p zDSu`(&~9q_10;VxEP$7G;CBoSo)Zuaz)#o^VVe9t(6}P6akUyV(-KG1>O#?p77OaO zIRzy$Vl#-#(D(jqR2QWu;_l&AFL zUus*zUsAf=wrFTg&+2f5+ssQ|Ar!zC$ZW?y zz-Fh&N$onsNliNjJ(dk(#K`#6Z#1L@0wj0zJ5}eo-TUUbFWV|*eATXvzcp3TNKZi3 zEgt8;jz-?z!U~$(FF`9+3}w6m3u$hvg+5+XjewJ2^W)qc4`m@p0(5hYw;y?bxzvG3 zBf*|@Zk@ANXtu;IsVsm75`E^5z-81Hkmg>11B2{|az_52+@F=(2X&tGQ6fV#$dnaJ zGs9kyyt~UIppT1+52?aM$7G(zFO$6Zga{fYPZV*L9zmRS_rep{@EXxQ`6!w&Iwmn8 z&~&}w-JdD^VYQ9_ivQ{WBb4)*G|%teGcVn}T#QcdjpkNam6H`N${v(5jH}G^lUi{! z^u0eGRpK_pZEtcP_K3jzkTDNWDaQ4Is=vRAZSrFb!x7-5+%Ay&9({hIWWN!PLgLTu z%-q(LZEZPx2OTkgcRFM%&ZZ5<=7rq;HxCcxK#Bn=59XAL#4P`kQisj7lXAw~Gv+tv zZW|T@Y3tCAWIWmVzn0UN%h<8zf%;BlbdXUUs*) zTUjI^os|{;3la)TVh{IwTjv#134SQqlVXHrqndKuJG1+7%j%!1~p;DJBYPxo6;YIug%2d1E(Y=ad(fluL@_0AQB z%UFNhE=5&P1IVTUI~mK)3yyBwxD-pA8(plyt_W-eSXJf)Ez2r*qGka8#_+p3`+jAX zD;8>22Bz#>1<+Ff=;_R!%)t0HAP6|$QW-xETC6K` zyDssj3ip|g8+m)!*c33 z(`2_og?gC5Q2FzQJD-yaYaaxw9xh;5J9rm2o5<5{sc=q$19u>PuT{uRnV%k-{-aqe)9!DL$NFh0O6li7wZ zsyB}$eApVvxA()H3CRry&ICg_`xH2uTEJY+;cl^&&A6Q#w}Z(;jf~)FrV}8rCBYU0 z0uhc_l=hoy0V$5WS$H<$sJbKE17)*+=MjM3#J@jUHvi`B`!%w@*h#@2ZA7x+7Paiy zmeNfFtqu1y>BQm9Ywq_A|Dx=pzo^+Sm3`(MCIE)%zi}I<>#S5iOBO~D=#zC@R|iaO;U-7uN z7i(b)TN!^PChZtd($=i>5&Zg(C0kdDzu(U`vx%i2l?Xg34=PL_aM5!=50AVrlUR0x z0L$n8?m1rqas?t9GJ%*dUuqLdD@wMCNe0`F18n3i$g$vmPidix3Tk_2-Yti#;y ziocOX<2p3_q=W#W2|%?_km$tWvbYi0$eN)-G^Ey{tpk8A`FR1|oH$Ge z=x(tYKbnwoR0Xd!`HbRtBKQq0LO66T4A7(qzXdMJB21Lywm9BDS<;7kIrFzP>{jJb zyRBijD|iOozm7QbtA^cCoPlmrbrKHKmy|V@oUbgk_|Z3*#Z-}EyN0Ra9b>2HEYpX@ znGC3EP>y)H8il~KLh7gjC~&c1s7N8C(NTgPZY+`P^9|Ai6*r-7(s;tCXgI}vO4mdM z$tF)5mz>uxS@%zMlqgq9ge-_l`DVALPv1p_^CZhPQ}>Ou93pJya9Lczpw()d0fkVazBw z9pN*#$3%`2W|#KP6|`&<(3Bus>_xMJ-_v#%h6a!*o(PaLZtV9eaUJ>{9EFSP$yf7= ztRg{atmIuG=I&MywcTroE}tUBn%ylq3R``_*-?78B4O+|=4NCkx?-tI62qnUCFc;C zNbiX50^tcbRZ3h((n5Elg?ECM+HPO)O`MWxjQHjXbphG{)weTiR#*BYuGHZTP#KnZ z5b;U<2ouAltX*i=B7!co3?}w*8#vbdBI{G<=@^QXI0#`+zn4_U#9i>AVhi7Oo#6;M z$6f3tjd_kU2E#EhqZ7Ht5P}S1E5kA{9QSqmd%NMu4rF9LAdbNfVt77kk5fsnMazRb z%8sxn&DL3XQkE9eTiznAd@HPNcDS%L&F~stlKJ17uE*Yk=cqr^ao@Y#B=L>BMLB$( z&1GzNMd3JEjVe zP2D5JK{07fcL7fNObEx~gsy@dB*_5K@lQA*$Q%Oy+qfW=Nd{~8mcSOU$|Jjk)o>&r zz-x3yxn$I8V@D@XfHr8wyKQgqP3GSX0Ad)S>{%KCke0LGyTtYU|0D0c<1DMneDA%| z-Y1_^r*c(ScXg~zbO1q;c1xCE92J#&X+_bQ!C`d9FyqWnG)PjCq9UoAjDR304ko%m zeT3TLi#{tD+v^(n~s zX3c!R=D$>HLBmDV@r_!9jPE-4kfQ^u%@}CJ^s^Iy<1XiZ2HM#oaQwiSn+>Y* z7JS@9G$`-Rh|5A@>&>~|uJv|}Hy6mXZ4!W7F$qAv{r?hx$alOx1|T9l=0(C|hj^S5 zj|<|l9*@O%oF9*f>1)YGGc5yZ-o(1w+r6nNQ903ip#lr;Am9r@0Sj6s_JP5UB@5i| z;f%faAp0iENAZ2-##QB9*9xsqcqW>>?!@%h%c+mx;Hn%Q(5~`-TXOEZS`ZVLhyHyHQmvg4@J6a)6oc|RpI@F)J{9i zFA=YPb0g$x9z~dE?PC`021ox9LpY-sq3947VrfW(tb#w?S`BZF!jtWX=Mo%9jGeb%;%O?=0J%g&MqKGvM?ScZLn!*wbXXTel zc#Em_4o$voRFv8s^Y9G6&1H?LlrtO6tj`!65_+O?MRqOJ&m7ZG=v*8G`Hc>SPf@Cy zy&c!2#>ozHZ>mymJH9)B2LgG%X*y~h%saFz5F<413eR-7KXkD652vdiio3!w+BN^x zyj(MXfm>ku{>gkQH|q4CWH8%Y`lm<^aj^MZ&3vWipOH96O!^(YN z(+$3G^U%^_?V@cYxQ7k8n`9QVy@hS0#ZcofHEAnpGt}e~j$26;Z_Dk#5*;lpK~xp9 zY7ipRq7Zt>6r$s3Z*mK?Fv7Oo5lIABBZXAK5&r{%f>kfoPbsP@5TpTBJl2)m>=2!y z4=yE-B1UI;h>Zg20rdrRSZIe?+9q7C$n zJ2?SbleR)7-p>M-nhM@ddMyl#^eQa%W0RNZML~y>4!uX&D-_g*8lT5%1e;qV?WneZ2!*d9kG!v24Hmchu;dR|7*J3JSVV4Y^e zC^^FUC$OBw)t2XcW|qPJgdjo7a1G0-JFoy_6tVuYe$_lY6y_{T9vVzhCe>_b#dSdW zm(ZQ!%9_k?BrncjmZjMAs}=x&68H8Juz!B`0e6`BfISSk9N*btqQN4Zw0;-j{p>KA zfn*;j9ac_`9Zn$+;pFYlddyG7(cjbKZY4BFkNGctfkN)+0SkR$(A+&>?iw)n447Z` z*av#dukyS1_4r*4Nfyi#J?`lq^HhHIcRg&{e;BYY4@ig!Wn42C5-gh9*7o|p2T}%* z{`R9}=M-8+I09aZ6Oi6mo3%?4bfBOtCkn!Yk}4zUT&`&ELH4|OPVyX_O+=Z5z=fVC zbV?lQa!$J&D?*DrndM}0`4KbtF!4!N*P;V{W{$>gFsh07ff2#1nN?(2OK&tU2_NuE zsLwjAoEqnlz60h<7<8@>2S5M?1IO>slkC(;uF}elERgIo4vJvn+rUytcB*CeGU;Eb z2~rsPOys%90XiwfV2NZ4U&%L*$8vL%$4QW7fOz+iv6~1EXKz59XzZ_|-OA4m2`2Du zclRKgH1U##pVNfDx8O)2!8Lc?{tx)kNQahcM-G8QZNl7oNS-#?mqOm#8WC&lflK?~2 zb;oawr+%|7B_{;)8w+KN#Q(2eTAsgHv)UDob{xU0CreV6+QW@j$$Q~X-XDcr?`2uW z#ZC_yL=6oS1oY>QICNUL0|pr572s)@<)*YSjwQE~$hCG`$a+zjjU`%2_{9lrfX%tz z79!z98z7Hx1-6>fD;Zc{@i$xZa4?74(O`b4o448Gtx-^3asN9GzN`4ztdtQyOGY9o z9*!&}wARFTfF^ii|Dazury8pUfzEr z^}^u#^#}_X1s@I-mSXtaO^C#*ak}uN3A=y?I>Bet(a!gf*Pt0=ylHWjz^=Bph)I8K z&<)4lu`C*{FqF=r>vZ4sG`b7*JjDIUk-ZE@g=DApTuz+3@$*_?c*0y06dFdBCVv+f z1IsqgpwN8g%p6Bx(`W~@`CwdaQv@-S2BOv)r>`-b5g0Ye5+Op@sY=cPdxrhV@nQc` zYuI11Zikp+k|!CuGdPMIc{hp@MlKtW2{JNz?x9}J=Q$sV(DHicSIrg5eGw@Ydl<1^ zwt5h4D*}_=4qRybWtg$i{>=u9vEe&6AUIoD>Tu0D9gSvro}K%5$@(%B=@-lH%rfW+ zIVbrwmr9cfrUb21&wpi&I~6LY1YSoXUbN`Ld0~Vk>kbf&#Dd+?OwZ3!Oh@>Q2rXXk z%)i4RxK4l)3=A#}>$8jLtyH_H6<7!;mpCVZ8NN6aEynPT=fcK7Hjq}LL2N7={w_I| z#3~Rws!MI7a`h3ssxMi(kVv4EoI)w>R}nZ>4%?OFmWEFYs7}-5QLr?!;|^s$XDl%?=jjYMatkBKjLvov=(QX8(&19e5aTQre0f$Xs0pe&0~%te52 zAL$<@fAF*s3tZFR9PqKNM(xerHC%sVz@8GVpju8EA^%m&i)MRPl8ct5&o6Z^ljePC z^MR#|&{Ol==!n@g&m1>mjv2`vI6l9O=G?WkanDjf`L9a=VL5S7;4>q!JF7=rmGL=w zsqUP$%zt^roik$2&TpPM;)cRp=68|Dr9whOIX1*jq7(+LZ*$w;>JEB}VMsL8LVMY7?qx3AW0EBYH?CYk zv-Mwbd%U}U(Ef6eYJa?^yL(TZe^c_h$OesI$^!5Xzu4~X8fhJ)PvnM11A$r(rD$%i2X+B6mjWe`^%i5i3)m>dVdF;p&OWo&Zgk;n&Cb|0Ty zf23mCS~R5?X9>j8foaKON2cFBGX36>utYz}%JZ|l1 zi@Bv9h~nIC4I^@#Az&eDImneXSxxjptE?f0>xJYbwMx@0SUJvgy(=7CPR=7`2*ez_ zsktB}{Q8A#*r+T-3D%hdPGDmIq;hfW#B47?Y3;=P!5 ze9r8PY0(!^Ok72TH?Tb$FHe;+5?~OZf*>$RS`@K#>aCfL0Q1CfC?l_(9!!PT5M zh(w|vU`5U}sPPKlz`qbKrNF9@@&n6y@^F>elyhwC%tbt_)zM3!tGRZL>Hw88^1S?W zueX?JY-${q=<@+2A z4P4bmh=5+n<*~XPyMR|_>sGX*UTfUBEJ*DFBmt_)}I=RNGnhF|+R&CQLk+KUF zlE+Hvg(Z7Q2@gN5v)p5jQ7R+{HjjOK>#7yQ>D|fQ1Rw||02yc0? zZ+2m!!#JTc-o^A=`v?uNJ5yWSjm$5!VcuOb59|AZlHH9RPP^)f{H}kF=sTEAe^&R- zVo2C;Q{ruyDAPg$+Q%pkk<$V5&C&+wM`}$mop=5ATs?k9$*j?%=ad2=L5a%rZq&;{ zK8!6*w`?M05jD4fwlLdcWJ#At+Oy8kC=hWE39c`TA?mAu!bq*%iMXZG0m96!0K32~ z_nFrux6@N(bORZiB6#3J6D~j`3-gkmATFd6l7g>64I^_VI&r<}TqpaVZ`$~^mm){q zXZn8-6*qn5L+z?suh};q`#_FtpvAu8q^qH~V@kQuTifi)_WPRsI{TLOPEQ_Ti7;B2 zO{?=IcOw!qUPOiA2+&=_-0LZ=oF}DrFpm=)Mo9FtdY9q2a$gLVvKk6ETBt>?0^#WbiBbrFI%iYKQPsWZ_y{W1p{k>?msXw>XuggP4xx&_lj z0T@7NdLIF?0x(#KkXgmgB!|p_z6$K$nNVu}d4%#gKvbR#<}JWO_3;2B%6vW)>@S14 zA4ymL6A+sz*EfXpe#P$D#me!Sf%ERIFpP{D53uiORaF^mcoUHOfL138n>5+KkUWIp zRg~s7Xj~s3&nHC`c|8xAIl>c}BIQvh#`U{|T~RnKWm6sECn*v0w9*|+686l{5cy^e zlJ69z7gzuJb`U)FLICebCeWCV@zRfm4}cP7MhUws2loaFA2}z4GqU9as6AWP4}6 zliBM;^qA`idXc&7vhufdWyJa}Ms6XU?CO>HNraAr6LTd~v0>*K``i4cX*Gwk#jA=oe*$Ujo!Z4cZe^IA9I=m{g4 z>`{AEiABw$O5e;;V%A4Z>DOlGM~(kA1-_FLZEBhfG=Et%4;9UWMSCB{j79fI(fmrb z$%Ua~ZkP>-c;|79X!eO2v|V zvs(C*`p{(ynPFs=CVIIoU~e5>;gUT@(6Q~6+)bWP`*BdiC(QF2+)-za1{rbjIUAI< z03yn+tD2jtX{r!`WkEo&r%Mnd%SgRPZm+_(w8)viO_KfXEU8>LV|PDtef{HS`V`Z{bq*?-~{!Nt%fsES8#oS#Jl|@v_gu(-Pn($|(S)CtE7$!pI#rcl(tc5w* zeg4bF{*~yi*CUfc@4zF1Rr6V9r1CjeJlTsbP&P$}aAmse{dy z1>+9M?p^qsD`1V--?gT!NxcJuKYN1@-^ZMb=^zs;v+^AVWRJKXm{mpgf$UJrwXWm{ z@o<3_eoxDczVF1^su0~)-B5x_zO=70DB0R@yIB*=+HkFnUtMd!?Pr+M2`0me?St-c z^FecX7r#yg#Y#t0;F{DEHU@PuON!gsu|O_lALCA;btm2IPGpHuzF{2~s6CH86XK8I z{6~%2tBVN!et2_{Uo2&AHp36kj=CXJC{ugjdRSE1#U9QQr$xnri*2pMqdA~@U|xG( zZKd@jBK`s;cq`$Lc&bnUR039ZWkS?u=Q%@g8vLf@Lu!kwu-3Luc-KXcj*NtvPvxGq z_6dg7-r>#Hy#2JhoJ1wwFS9o@#m4*-aq$pLo$1Z#q-0@C9UqI@7UV(KdULO8Eb+8)mZe8=TfcCGPO8_W4L zm}>0M%~w>yZO02xRl5u0J<9O{_iYa;GQMRi8idwS)6RoZS;7KdMo_{zrE)fcV6tAt zxQFu4AMVRAJ{X@0AIIGq{w#RHXiG250)=fu!a=q~G}%}!GMWlZ)q&6-yOUf7SVY%s z>IDJ=;@*xKM~VrDnbo>CTlc`Re6qTk7v@F>e0(IrqTwarnx!?~TUEv@EP4ELhc!>O zlizXn&MtSy#N86&jSOXGH~V_y@Wow$%;gPckK!8~vhj)>a`vn4YxWOs3_6i{N_*yA=g)U2jV3#E8}6aalaRb{+DR zzL~keXlA+feMs^o1Pm|$XCXbSCZ1VQyn%zF{AeS@HBN(3RCEV(r1*-m<5FxzKi8(+ zHYoMM9C&zV-XcS0&}xLMqmh9P3sfOfWZKVn`HP}Jx$D6#U&G(_}W&}#^XS~ zH@TUHA?B2D4l1HR^Xz~l!Wuai!SI9-S;YLU$)iGEOIdTOq9&jNA_E`HOl&oCpSKzj z0wZR_8))oo`ryo)^LBSzg`Fnicx*G6EOHjE{9ZyoB^T*3+x}Mn*uEt4a>+xO8@EeA z&YxzJt7P~l8foiKGl}s}-|f#>EqZps_12aa%eTu+{FW3muVv!i5(drrytgzf!5G_b30>uUBfvB;r$+0CzP{p$F;w*1uA?@unzFPHoX zM9llg|5{~y@<^2E=XbzsBF!KrA%mckTz#G6IGFsX6JTpso1f{Zs*yZS_-hYIg#HQF z$I5^*L74Fsvv|cS1u`xqxAj7ULVlN6U27p)ZrAUy)|Ma&*Fpx~(Hq0nLs@|w@C3-# z`r;bz*CO7(vwce&hNO4Mh_bbX;u3mEDz8RzLWbYtxKLuw4d+ff!3u0muT~7p^O_tUMEOHVrWZ6osf+oHB=FCAA>Hg!GL1;A8vPix(y@ z8=tIZTc^hfM)>((VmpV4fi-anwP7JI(1xD&1Jp;K2r$(w z`Lp>^{2jZ)gr%zYOn#RtYxOBs+L@f0(=!?h{pH155BlUGRXt%C#|J2u))Vc!lV4Qu z$t^*pEg8MW%I5Fb?C8nF^~z=JWhS|Sw}@x{@BgtW=DZgW$%6R5@rm4nUpcbEuPi3d zWbhzbE(`du^9@10NTGwnlV4R#p5?X}l94qmAmqEx={M_~tOu;~Lh?fXxQma&9w{tm zWHJ1_xZJbk#dZlT_YqzKe>;Ox5h)br#Z{jrFR5Nu{k*v9v*hLc_wc;9>a%3Dza$UKen)P;<}&s2u+&BBhW}%{0m!)pJ;TG zACVlB6~$D;`A@(x$!wf`^Y{5>Ei_R|PAKd*@qH);SeJD*l!t|!_*RprR0q=w87ny* zZ7EP)L@}l0q*iqi#l+Q?*3}9sQm^?|RlEYCLac9j?P!CiL87A#w1WB56lIuR>1n`V zgeCw6R5DB@VBXQ;BH6h1i0HEKXcSJspBs4-q0HM*RS2?3mvwPQDI zyb*_G)Yg*=#%}m{!>guI{I)L|yHSWY1PKgLJ-LJ%S4F(a#53#swZHQ-?-O96ncVjZ zg^rU*o=0NwH2G5glM61nN_H5rZZEtaWUD^ICXd6@<(GAgw-&GiRM-|)1COdK-k!2U zjXx64*!HLVO-9qyf$h!2GOf0_CkHZ1Ru%Xz@N#Zz2iKDnuhrT4w9!S%CBOA(D8Fx-xsyMuWWsP zlQP%y`CjYy?XBo`~T_>B8BcoIMS^D|MKG#xJE6n4<`eRfKpSn z=c9c#{w?U!N9FQ$qI>Q8dQLQB4x@Rg$#>f|Yt6TSjhReCDv`s-nc3A$h-1aZ2iUKP58ff z7I%{Uw0@08dY=y{@3+YuuJI+l2rhz%;K|~NTZ!(b$;$jvWCCjJDQFUOW6B@f)8SjY zfipm>I3{_K#ENqo`QzbkcCfGK7pK-a{##R=?`$PnaNm^T4o}JPcZ=&DMqHh|ICe2j z$avVQP3C!{raByE6v-KR(b#oXRbC>#tJPpO39d%4kj$RxF0)`D<;-xAcHcYpbl$$l zu$V!kn}bmS<5U2b-|Xd)VXrd|1suWR7G0)=kGEO0v7RbLNxP-(K|`CHb6)HO=d{k+ z_cTG+2aG+h>uj>|N_d)N&aH2ma9#fTV(ayANcl_37A79AXf%z;0IG)?d{V=zvD@|hcJ!1j_&&jRbm?v%6nQGs z(G)+9Dl352$fSYIY3xQRzmfZNP*n%UHJBtyEvmZ$O2gsy_5oxR0 zv)po;z0fDuV(=sxCkRh!@3Vl{d%*Aa3dj3ZGkL#+vt?qwz&)>m^%Frioyfe12*~>0 zAQcms$jzX|K)dTA*|GHen)zLjAJABWoT45>DWB{Ay;}?u30niR#oWPd6n#Qee7$a& z`Y|&OcO2Yah8y(aNA0{#GIpuPV<8@M!6YfqljyYm0*Yh$RN}cV$0u*6e zX2YJzu2{v(I-95iB;vsaDhfl`35i#>-`M%v30AJ{v1mF%6RWzzQBD%ygONlR7(N;P zRmS7QMpG;!w8cGDI}I&i{tIe`LG;;Y2!(D;JVp%S(@xU=%<+STGW+lfYQ%aGV4F_q_7kvT$orjY}HcYD$Wh6BBDPJ7B`Omb4^ z_h{ci5O;vD5M#Dd5=(;3Helrm3;)h&K%z|xD3iqq`c{s(eNcuEE0a&ayOB2#CyCHT zXs?|wgB7?!Ur%DNle_kg|qElpg-@#O) z6O@&q*){nX%1dgaGlVtmna`V^bM8;f9?9UXe?jL26Gi_jp&E-vo;1F5IdHBvq~?>X zJh16M2(8HA@n5xacme8NWvAX1r3|}@Nsc9rpR<2*aPs+uQDgQdY=`KZT;L2*e{a{h<(w-T7mIc09tJn56J5ENRONZQ7Bi$a zQN|m(im_+xBNiW=v*IxTISIon{vVf_6SF|FovkONJ5jD*){bW8Ku9wHqxv?iGS z>73l$QbM|8G7oukx&ljUl=h`HE=lJkze6KTtCa=}yfc=~P-O=5GRWCMVbqM#V?Hrk z{YG}J2bt2<-h9HglNb^)0Vif$nZIcaUY1WA<%@j8n1_wI%$Q3B2L@;$pvl?BykyK@ z6xU&9m~Unh;-`Y$77KxD=g@S~1Nf7^QbQ<&cV%QgG4y9Mt701FSELDMXaF;}tI26T z#XFhsjsHs&*=dB-1?>bp;o<;w#p%Khl^lxNMx?_+K?G1!TZP*C__!6BYBXlMUJz{Wp=$s<1f z4cR?e13->9Co7y4-Scv={t_@}zM}Et8;o#)zAx51fKY4l^U4zv>+eR98X0{RcJI!e*WGZ;r_N=5+tcjyUhNcE=x|q+KQkr zIC%tYcGJw4a|ThBE!;2hKe0+#f-v1eSTunBT?-IC91nV;h_YAy$PgOoDTRFC-Jm#DR}B4Vc0>~n?3Ld0g=F!p@s zE^$p#yI=)>uCwPkPp+G2K?*=&ejbsJ0@i1GEI@CuQjt)3t5hBn->|`% zN1T6HM?mY>&OPQD8=T+hEPfIjoxk2%sPp*OuHQcI>_>66#Gutw=reU$j>8 zAq9Z}^RmOD2Eut3Grm5I`D7xht7P%zTE+H}GKa@AtTlILyZr;GA4_HJY3e%58((6f zjctD5YOzS7z2yajk|D+|aS>d-*(3-&aST27NIO7hH1SULd@>@HHKzGNcJd()zfzrL zXD$|eb^xEU6<6M2;Bzw6^rPE@cKwq#tAjZ@=y-fEgguBya7ZC^f~@joKsG)NLN7hy zN~{WGrS7OJa-VG>7%wd2k1Dgw<5ZJ99=#9KRj8jI>;=KZBekcYIu_M`5raeRvYIUr z>Kq>htzJijHOwSqJ*?>Hr(_px5^Z^KJm%D3IiBao{K`J@N_RZYm{2mmyZP~bucFMt zSZ2=&B|A4k{f~|6T){Oxj=KlzRg?-gbX;We$G+R=lX0RfJ@i}`D@KE+Q4o%s_z_y3 z_*>`ISVna!SUZSeW^682K+Y5Pa9Q zaq#=s+O*ch(+xVkIG*xM+-6kEvv#bCCkX?iLXX8zVBS1rO?>En5UKyRw^x$T%Gw8@ zT%jn*YYFvV@c*gN`<}7iH5~m{Q-U-zYO48KM{wV`nc+6`RiIDQ)rIkJsko3ot# z?>>N;|D~;}K-mEAgi0|%hF@*vA^oVo%=mBMyy`y>ao#4&G*>h@I*w|tm4vhx!Z$rA z{71Tg&QDH4dI0$K&BSZl0n)l;NQZinx3f6+VZU$p!2dtsg=*p?ecEk}k6Q8LSKGRO;XAu*BuFXI5%$g+xe>%k;HG9$aE} zv2)l)v0!FG=<#M^gO1i-q3PLiCRUl242ddUD z?PCcriljj4qg$jSd?>MPBb|yvlXgHFE;GYutyz8(J0r3?T=Ik8MIwQ%1x!joZ^)~VxXrJ-Jif;l>NtSQmWxKQ;RA?lh7*Z3kJ=mN`(FU;5%-dkF z)~&>&-vHH2_2LgaR;!?y-ITg(Cez^1G2>-U1KSI?t(Fyram=(E5#^)PZkElMqb^Eaz zcz4y#jWr;tXQX@9Xrz&haU6-q#u`jFh>maA?s=Av_Q~y%+wirIbE-AMsIzXYIR6tE z5h7$P&}@^@psP9pmu0W=wH@uVCqN}qY zXle>his`J9n&{2iQ~78CyJ80+k>q%?s*^1|#BE$DgDOdBcQ^}Hm4fLI*MsQ|axd`Vs+L`fKi$}!u&7|}e zp4%z;PFlowmL5x5(&>q<_=23>fL=%NC+v+!=Bbrh5_Uyc z*xuq$M8e z&nuT_H#3c>MgLR-*Ojd_)-cZXm}`Q&Dul1`xBxm>TJlJ09!}B9d6WWeVR!5vo~x3` zR_R_#_M(tTtizPuR+Ma$<>u34Q-8HHmu2pG=Qg8Sk8R1Ln zq5dIPOa3iIWYR)5Nc|-dp?5iduuG8K1y&|HP1iGxCO zkYq-L668`B!d0dGQF@7#_-7=@h}OH_QM3-bBNz0vI_UdC7EW`ub4>L*VuwaZb@m{E zL;8k%O7>%^_t+gW?v?#6ZZ=ze084?Y^^XCXfCp(Gp_z5L!6sX%?o>mHUzD-&Mmq{o z2%aH+*4XiN6V4m)ylVrY$#K)jA!lVFoiIGb>=Z**wX#L*2huH&hja{@zCs8i#XHNQ zKn+7?&0|OLdK8Uri^KM;^WO*OFt79c%u1yT`~(aTVO2yASBZg?&o-jse6Gs?qrl`3 zBGFmE+u>T2!o@AbP@)rM(6o@@mCP8W{x9&6u@4b$S3+M-8YtLUdcwodMl8y;hFz_- zog(pvaM3c9R0wj5CG5w-JRn`#FG?dV<^?@eZuX{a+JE;#Z1Lt65@UI?Zna0HkbV5X z{0&JtndkE05;I(ne3{9~2QglZBAZonz_96|A-RNSkXoy9B5Y9U@)m}eSZR|=*{kQW zM7djtoVT(A%fs1Xa>>Kw=;q-WA$bHk}L?G}z5g zUN9+iCS=Gb-KQ78Jzvyo^G#n2Ul+GL*h+4CyYh6pqiNY8mA(O7oZL_ zE_5%VMYd~=oJOkwmIKI0N6t-xPGG;^xQFA&36Ap}j**k6 zuW?xwZLY;V+tuqBK-kWao_e>^dq#);*%1s)OI9b4EGK>`qFasi9L&t2g;<@)Wx9y` zRLw$mwhQo!DROsv2jrVv3bHI4YH>jjrD{zq<-d@!|LyNI0Ea}+$3mywsk$1&_l|JE)7V% znV0Jsah&PoMkmCLb6YynqB^I7>%tbeb_y3EWdT&X$VEt!t0GSbRFB>&sLpDEY7Cp! z%>V}oOJ4^fdyPSmfyYNog9hCwo`O+li;Rh4RFy|o z#Zop|V0-UME~h#G_zFT?Qjx!rDiy?;Xf6SXYFvm|A9!MT$9Z5I60t?+iN(b7lA3v= zSr*q(h{u~)R4k`yGvp_f>EjHp`of&G^uyL^1 z&J4Z1@C_8RFU4Zyle-IQfheC2WOauv9?)8G0{L8tHO%};A1oZix?_-7cMM6<4#qZC z09N^o=W9`1sjP~$_UifCs}ibp*VgK*hn|8cX;vQ)-&%XdeFoettVO5{?IgH8PF{Pk z@*SN-^l=@tx?E?ihgI0yhyAV{EKaWvIpmV8MK zDwOHp%(5Q`lx)IO_XPLzVD4k7n{(6VuI7_OZ;!~E_}_m`W06N9gMcgxAyMicLWBje zqA*r#2=RfOL@59+Lt)Z=9V-$n{&hS|2~u5i{3-2g*lEYab2>`(*By}YB z7#_p6myk~4HIlgr@t)H2@p!_3|QvanZV_Da6-##I($?!}E1Z?06@GzbpH8^VIM%`h?rk=M#K zWQC+KKe-H!mIVyDcAc#Xm{uyx$vttasz$i@D#Va!Qn=nEPaD>*3GX3%O)xV|2l?5C z34c1k-*#rU*$-(LskNzQ81*GwI%FEVwP+GHuN}GC1GPWEoG}UYee9k9sd}a9&(B1} zgYsCabe?Aqwc9g-Mbl^ArS*q=c0|yDmLsK}-ON5b%jUS3dGMtLO@hlcn;S;X>0`8(H%ufGnCN+ zMk%$1B0!dt!d}Vk2w7;*FcRX|z&&@u1<*A(8m^Kwg;?{kZQz%vy#;8;+Ni+0Kx36v zG&h$DA{m0tuXQDd6M2g!ySc4xLAf7t0nq{LT5MP`sg**fHfE8_pnSb{F>E1dZ=e$! zEOMI7x9>6kLS&&)WBy&@Ic7ZVuD0sbj;A?q1HfzJJCfOcfS72MDG(Jf^*airIw>k7 zAlOp==*5ZpV~q^aR~vgYs(kBL_C|-PS}|axMYx3GWvqS%^=e!=;u!Ek;4L51=u6c{ zv!VMSYuIuaT7`PH>~GCwgu0`(3c^d4MkcS|M#e&@DyWJg7tv|a{zSl#7Dh~8fAq{6 zHS>=uB+b_(Cs;z%O4UHm`Hm8QjBq52WXMF*aL>CPT2n4K3{mAyy&2(c2sQ@!ZhcF2 z(<9YQQHNGwAR7ANGxPh>B@MCmBN+mUMhuaplH()LJHzkDMn$P^;9a_4Mb()> zpTNdMN@qGGIq`hb6^zDwNdrX-l|T+glGBSBg3wP&!2Cr+w18?A;DEddS{;_)C+w!s zA7s+9Fw;kybrx>A9FL`SHcOfX)Nr9VP$JxmD|~@;fr?UR&DplMhJYJ{ES?BTK>WcF zYPt&Kq0CG)5)ii^Ltt1&&tIAwaW(=0E=w_l%z%{XjGScZF^s&hI8&_OXvcTecu5|0 zI16n^CKU=_a5fYfQ%#Wpb11*3CCy@ys6@yri`58=0H%_x@~})AG{Eu(vl^})7%JEL z_!X6;YrYs$LTOA(E&v#5)C?gzpiP^TdI_soElB78RSjtID+cDO18A+bI3T5lL`odD zN2EK9SF(quSB>3I1~~+C;5Uj7S}Mf5f13z~XLD06k*|bB6V5MWL5rs$rvY}4d zJJWPGw)h#XmWYNJL&_1LJMxw+Cn6W?L_~0_Ewp209FxQ2m{tP+?0$vPs~)gT<(CI1XX8M(ig zS!>|!-WJ~OWS#=*8iM836m8+FX z0_axEZdD~DFS-6!x8h6z3+?Tdm5f=Ux+q0RCF$$r*+xRXT};5}!Mm&P0Y2GWTW}=P zecBeE7oi@hA3{>)c+1!YF?v#NgR=Xf1=>>Y_BE!zT*3cP(c=HT3k$>AzcGJDl}D)! zmUQ-{ouBz*JurVQ@)i<0(hxb7MEFMLKH3kV_+h!J<=xoo`Ls;vB5fJFF)55ez0y~& zLwfZ-4dvdDQ)mTIRaMh3@FYDckWZH=wJ;fIV2}$b z2(BH23~p*IOlK`vKH8cFO4S;%_ie2^`5x+ZW>&GfPcpR=4P@s+cep)_I5~6OK)PdV zc)+P+XhcS0aacef`Sf>@*U#k|2X=pEF=l>w+l(}69+Dg@rL)!?yI`RNGK2B(P!r&u z$~5biwi+6}KE%*5C<19_AM~c*(W8V{wXShb`y0rRaS+(Y6);P{BZ^$HodGN9fxBf{ zt-5A%GY41+_jRjNvO(R#lP!M^WqJEK?Zc@@4ax?*yuI_X3(t?|$h0|sG2fU6;*V_~ zeJoSa=Iw9y$+4gT0>L1f23PZ$)`WbllMM=NCxQbG6;-M>%y>@v;B;?uuxxpWv8t|X z2{#bLO0U*!4&(tT|j|P^WJ=#s^ryKOk?Np z`G8d@FzDnhfBJkWqY==kn5w7lBuIItGoPuEI3owIL%{9p)52gAJ;@3ton;U|LVPPSXAlJf*eOaBv+QrS_O1sVBT z-yKODrwn+Ju8~dJdwEw8HGU|dKqBZh?5=YcK#~z#I9eUoCB)&~Or;I*&k1^=^F8j{s8%T}xz@zj9gT07P`tLLhSZfm6bQe1tCZ--ow;ztFBL#2U?QFGLuFOUvQ(<a_Fu@jY3GPeZOtm2piZ(1s6%rN<{#66|zRa#*QZ^EdrSpYWOiSPfx_eEGu(o#{8=jt0EdLOrT;S_%Ii@Jo9;`#c%RtCx3 z)1&VTgu9Q%w(bff4&42g2@EjybnoHSP@gWiTTS>d(DA15Lt}o#fVVCmZ_E+q@A-X% z`$+maMwxV2=q7@Vt$L#P{2#i43}nvS%EE} zG^1+LIRR6o((F>Qu_RH)s5$s3gkl*=@O&4sW@Zc{av$ke%+;lyub1rACBXF}h0eJY zV+(T@h@b+0!>Ry@n=1+s_^G?3;M*dzHZ8+~WsRnmXd}phrI`^KsSkrO)6Y_vGa>90 z3HW{@vRZB6{!IL)%E^mDiK70I9wTA-50>D+k3oe-8;Y=>3(;y~0K!PtK+tFGx`?kd z^)!-vfk@+@r5#9nPBr#p<|FnW-KVM_hq}(E473s{an^h!c13H^!X^gRlz=P%eGsyK zLX`}A9#Z@@QBG7{N5Pl%hS#!^cb~x$sfy0l`&s5fLSXG_d9|pf*)Tf9O^Ynq4!1Gv zi^^t4+P>Dzj*OTWTS`XmybqZ55rnx(OigNEpgw%ZnCpzYi8oUFs87)Zr`Ydz#4T9! zXzV=oMxCfbM8-jur58b}Ce)gyf!HCyz=#-}@=+nV!4^KwOFjXe_qXBim{wPAGaTBd zhtip62T*gzakB0Vw$pBqH2Xi5?1)_f zP3Cu&d=pd+aE!BI&}|wI@fkAhS~U;B+6}C5o;=#tkEJV3?TBg%yCr)lvSi+_rNQWx z*-|>CD-t=VfMe_Z+hfRTGJ|3_Q%$FbHzao>sxn`-gEe=aEmf3;34?}8d2S?Usgt^; z4jdEcupDLr8T4f#(ZYG1^3=BEv|t$R@-M=4NOWYJpCu!roF zu{>pM4=1OBYD&U>1aJeO6&BiMI{qT5vNo=1Piz;H=<(NM;?W zMq`g@GRFDU?iTvpmjOXmMX#BH2(g;NmV#6Qkk=yuA&_mUTy9Aeu3)l1FK8Q1?qz29 zhfJlV-%^iZA4(uCuE4|4(UwoC9Cc7P;u&X%x0=tetQlF>(^%HBs2sI}aSbew$C0=a zMoprOtx>E6%_+Y=`TKRoh)w}9mL)=}&1U;qSY{5E5N<&pw6Bg}PFi#`TClB2|A%~) zzs~M1h(~wLIHG>QDGwfk`J65|Ek0=?F)q#wnFwt zzXYd1kw8640>=zDWOpIMjrHb@>kJl3McF|5Ly~=z159!rK_eDEwPv|kBy>)kIkJEB z1M9j;^-&Rlbe8JX0jMx7N)gbU^-_k_#A_h%WV@v`wcfltzAQJ|!~}{Lm*WvbOzmBy zrU$Z3$`M1vp(7HK=^&@!@#Bm;l@THHt)xo>joF;DsL9{=ZK~*sss_X|d(eufd!)Bu$vo|%&Xh$I)(|EL-8ZK#cf)=!fhm$1`&cYa7 zMNG70sz$dn6xf|iOPgR97%inIOiTxSBX?v9L}yvvjcLg(JkP`dc@fH!jgx-M7CDir zG_#%{s6<#MRxk2W8JwZGvgAyj6(*SYPn}_^q_C3WO41=&5$N|kJ}DGQ<4U95<3RI9 z(zlT*4e5{wijLw;EZ=bb6MtQJyFJ)8i`D!W9M}5r0*XS?n4jDO#)Fd6C^SNEtchx*vhgy8 z0*(+FNBl8rcim*z+|#Wbi<_AE#u3S*SY3u#X52flGm56!#nKINRmB}3q5{^}TBbEEn9@;M zQk#=J!0kv}gB{ACYGwgjPl^!*AWyzApKr$}KyTFkIe7-GfITSNXw{bSHX&01XQbuK z(rT+O!nUBiv6VhU zTYTh>YxYtTE;9B)6V@73A>k_HzsAf3=^nJgVqiJ#RwB6KQQ`rnvFhRbL)?)uC zwRyESM>?7BwAk-mMbBo*1%pur+>-sDEsbS=3v%YFd#GTuK* z&5fzQInty|(O0Eh#_s%ldGLjcgS{k#F9&}n{m#nahpE3I9ajmCO;O$vWt{=L6-0|5 zhmT@2V-M8sE$$BmInXXKcK*RDLDv$f?a57SBsU=mMY?dhL^#K#;l$LQl!lX2drBG* z?2Y1l86-yzKZ^oPQe^H1Om9AS*1lPDhu(mBs!o|Z+(23;3OkTs1rzAW z4em*<>O^Q-ij6zIgT{?^gkw7Fu^r*~4tqjJ_;bU*C@(K9IC7H4xy-48iECfH0nO?_ zIa8H~%jU1qIPTmY!F&yr1pO|KBfH0>aTN5Ik{7$OrwG+ypXoGDbowV{5`rev+(pWH z`=e&~ucp1d8Gh2VcQ(U~O@9-~&4K-=I{jmcbFbmYJb5$Nk6cNHD~@2Fa^gQVa7qg~ z73fRJ+U^z-`n_KIXfyn_X&-Nf-#6`(&9I^AH#WzS(MGcVvNW&mrsPH4;nHqDOfhFh(@4efzUV*?q2)QGQ5~-6BNo%HdC^z%pX#@8v3QRK*xep}QP80AW&^dX4*RLK3pBb>n4TKX1 z>`4RRFa7={Es|~`-{{pCqfdEt4epK6?1`gEsm))|dnhdu1OVz2t- z96lHEd72M?yvg-^e#GZ~J{$Rb4(j{se6HuSp3lvEZs&71AEGTJ#9~NrhXE>Ay&rl} zj&_J;AUjZ1?Tj9*uRMT^?HSpO*=N*FXioXSw7(s3T}ajxuOY`A$onYcDVgKS3|Dk> z`x{&6_*;Z8klKgGRA)f8iya5{fy7S}(`xO`j8h+VNt9PssIt$Ay#7KvSevXtSY}9^ zpfM#>h1)Qg^#=l9^GF&WIm3=WATxFz=wjoRxM1Zpc=Z{49Qnbnlr5UtZ_P}zu(Ck2 z0~;p(%ockHQit)t1dv(8VuQ%rgQ~b*@Q_FH3dQ-AK~b<18#dbQH(0aL<_=EpiRYya zI4%0eOmcG?sc_x_V_PBA7f0l^j1e2P#C1q6f}M{}QaZQDS+~C7`-)_%;bmM$lOZ`9 ztS=2mWzgUXHo7pv9$h8LZ|TwN%pR;U+2txdDGF?DXZSNCHE}Cif>D$`(RZ!GpKBnv zNW`K}v1yyTkq&If81N3gk|&qTMggC2xK1 zM5CkHoAF=A=Z+nK784)O%p+C^7H8y0#j%iYy+x8VT-J=@Gd0K~mm;O>w4E*C>Rx$6 zXZm?+V}{f=HOs_r37_XGP#96RYU5rTnfQ0E(~y5Sxw(K5J2sXKD@p})y69)3MHt0V zheWe>#<`*tUSz?P1L+QtpmFwKz!c0{CFwX~%iLUURg^$KW!oZXB(`16=^fqZq)wIp zQlsfR5P7Y}+z-LE5_>fTsQ5e1#5PTfVOqd99J#R%(-;YIol3Bch_~X(OaKj}8~j#` zjV*r9v$OMgFMdBPepFf_Shcc0g5{sNf5DyK8`^V89}1+w&LA)r&B z%YyvH8!)kbxfm1v(pO3Vd*nI;lfbFUK}c;`lJ8hUnhZI}7BSJU5Prn^jb?s?oQ`-b zZNMJhuS(79iGo}riy&gO06dk?2c)#@cWQS_tOpxnXp5jSV!71bIR|5 zNxwtK;FRCfC;e8PDQ&6KrYh~s+f|)Xrqmqap88SlC{MZ9AAe`DKOIx<)nDOv{%m_Z z(vGR+yC>A$J9eH%kRKn%&BtJBJ(I^IZH;^0=AN;7L&wmB^NH7AshyqixvlqO@*Jcc z)nbmFe41$BZfU`DW*&@}aqK!tXHQ`M~pC%v13d>X1HL<+6^RM($6I@uHJ@g!kI1@VouvKdm&PmrF)M58Mb=%& zDwjg98_^Dh-MPfnvhH;2cA}at^T8;u?YCGp9Ai)pO*8wWGo`}q?2(j;eK7mQ79Uvc+@|&e$ATYSjaN!G32qnD zm_4|~$F6ej3fFqiDQuibt%I8m{v|3C8*8Lhg+tRq$(MX>(-te4(Ae}&=a!l!c8!Al zg-aEma$9|)xSv6x6JJF^($%4UWe~*plOA~sHQ(gix^{1F(jX`tixfww8%!xXcZ&~A zXnnxi*2SnS?155Wr5C=^BkHijE5MRWIkf~A{3uoe+PSTHJ}teVxMzPvrdWD5n5Tl9 zbpOj(m$T>WyQQ8?Xq4Un+8@FoOomh&;Do|oCE>MZJ^H`+s$0_%xtbqr-Lj>7gSm^h zju-fNdcdoA4QmD02Ef~v4voe8%;^)Ll68|qR&zjf7WjQtgU()c~ z(nrV3Tw+PYsioU4lb(f+E&b$pi7hT}P43mk9AMvO-%ht6W?A8c!rB6GPXUQFpr!Pv zS5fd%W8O@FhaF!<(dVIb%%|{BgyhM=;2I-bc(&PW6}P(?u*C4C>x?C=BTbsR*vE=J zV8f@V{JkJQ*@oKE_l>>5D6r*LA9++Y-)QVj#xN&aE%vw#N5q8lyR zqkzpTUQNMI<2r_nI$iT>3Lh_!zwp(RWQVE8`9AyAls?^HdsqBg)))=;EoPUkkH-ho z2W-ciU%5;Pi^unxPy3$)cN>ysaSME~_yJ@7G<6RUP6Cs(cf1O)!M^oZLWy&IX0LocZCn*d8alG|=N~m4uY08urN`B)lZ}5OIHJ~zz_;apH-8ECIW*pG zQ|eyKJGSSm=-9&$8R{4f+Y5int%G1>AB5mHf2IGmi8Oi6YCJCek_;E*^Q~`oC;vBm zi_ws^k-jA&_%r5uqTuU(A^cel-R9B85|C-RKc!?yy=sZQRGeLCgazqHSX^4<7UChP z-?P&Js3mxZU{67WFpJ@SKT$1Ua$%cfn2>)j7t_lT(kGFpT|LLG_!fr6e-EM4*)*!D4* zXkki3SnIGA;;F&4Nifh0dp4=G%p+j6VqftOpL7pT&~5(CpKXs-%Jwk1eD{R95kR-z zg9q%Ca#P1=YCV(3D{ak5-sYaMdIK_o_B#Rw&Ew~<)J`chCywCM_fMXm)Kq&K$n(eQ zZ>F8&y$k$mVEuvv6e&q2*XtXdXp$yANAOQvRvXxbj$Jhq>M{*xxdS6rQC+UQS$zdE z)lYabAkgWMK^Jq4BPEXE*P2Zh&+x-HA*=M4+2ngwd7sBUdKvV>GPl(6DKC@tH^gl> zZp$2@B^%Cn7F^Q-b(EVGg|7&npkbG#OYLjS4rW>Kc(&p)4eQ^@LP4s7m;qT2Y=B^I zu9+j0=45#m>`VxFS(=HD;2bc9Gw@5zJj`=*Ph(ukpys9iso7a}1ckX++Zy7o!Qz`r z1(}w~JZOy-r|u8jSBE9vs>#5EX4V)DeUn5Cp+7`^wT{p=2ER&)zW{YbHO9X-N!S6k zx48w$%kt$B9uc*ZvZc=go)&P;%R0Ux#TOQVxRzLY`$<`o+owfCRyzHouRQ z!qC9c@{ABITpBey%ap+%RZm{1rzo|!hbXPwmipOZp+ek@C@-;c7f~b=fToCoK&`g( zF>>QBvT?VM2C#g7WmCzFmfsPNyT#*2^M1bP@fqPWt#z(@?`{3wqjhaM-?RDT-)3nOIGxk_JzTNl~Ti=fFK0e)imhst+kLsM;oYB!lc#|#z`j%Cf?$!EU*!teC zNtlc%EGc%_*0o(R9xoyQ>gQc{S#7CB05A`MOFjO!{O;(;SVx9?fqOIJb2Zjhb?MT0 z-r%|X{*qw+om7*QMLi6vzEtVU5 zDZGuC)!%Nr6h05I39-IEZM)>RDak&lHy^$2zI;s!+u9SiUFzcTH-B;4B@u)WESC44 zzwL4-tH#&**DG*hm}h1}h7x8DPc)DPr~wH^fiNfe;L(wXRof}}Q^t9nPPEGQ*o7ZD z^N(hgVU`!#?^o>2GOo7vWTmax3>4ZAr8CW~6+=-h&5B=jQvjw_s>iHz0yXO{aPE9p zID6Ywc9>fo;cQd814n5Pk4M?BCajzQi|qykUdjxalebmoLeb$36cu|>_S{y<$*?_E z-xpNh`P(jYD#7k>2<4qy9hh)U)0|O9Q}PeyBjymuuMfFT;A?^?S?fzsAFkN;OFrYw zj*R~wK=$rw4))I&D{uXU>2*vz#>SM|YNJvbwYqK=XkP6vw_V*`j02bg&qU-SzutCP z=1aD)QZxObW_~vnZ;XxY6-{$x(*S4P3=2&t{I6{cf!RJRR~)2|n$=LJ#7xn@FK>If zepyn!=pxRH(;#-%7+Q$N;Z!vdmyNZFA&H}Ui$hQYy{UiS*!EjcOm6Kh2qIqTEq^lR zyW6h+mw3r*lmGQLM#pZC-Www5iEWliZpC>m}eSbSES9*KI$2 zH99(6fpB-?1Jocq`{_2Gh%@#zz0nqa`AX$>Vu6zs%y zEMK|+Fl)Zvk!EC_vGx~DgUygPziy)G{8iI_XWE^Jg{^s-JF5r5hGEV_od~uw{|u*< zdzW_W#p=#cB zZezHJ_W;ovKUbW1z!j$bgrNa6oqfLO&MAdkl0V`gJ$$z-+)_>dQjPx?CU95@+n?g@ z(;iCL!o)mRbUQrc?C_nKof2^1w$?=O;^(C9e7@+X*FEutwD6e@@BTh|a;(LhJKP-v zmvHAY#qJy^VvMAc(rh!wA16l)cxLcoz@Q+}<$>(piB_$5Vqb)S);s?vo8Djk`)DX( z=x(o}`v3+AB!btc%QBGC-`NmS7rm{k&7VS2-UnZ5{6+uadt`2y7~z|C;7xp zJelBgm`|D$3Axd@#t*Mw2 zX_R;4$j_(x`v>#$G9l_kHo-1)&-_IbOPfoM?w>hOp35i#AMPA1GX4qnXC?;%2RGb0 z@8ffVW0FI0V|1?H@P>9U(y{F2Iw5062#|($Km5)+-6%)?Y0$YEH^%p#B0*>P9O8SE zLp@Q=OP7GmsILaKJL*M1hQsRJ%Xa#PQL=jzI>8cFW0_e26VV5G zXyX}i7lsxG0Pu9hp2rZ%@L-NVL00ucaML;VaOSkYELX%V@2R2{2Z*y2H-Ni)Fw+Ej zvEe^nc%P&|Y-{ed<>Sp@`UWDV5bg=f0%4pIze1=pR4;dDGowL_VC`HUml$#(Er54F zK>oIJaZLSMS}{}ZMqP+&$i%>?W4uSOgiG=m7wa}*k2JVPIumP>m|)E|8^`dV#<4h> zfO{v{JN4!!B?u9U&k(s-tu6c%mlP1pPqO&O7`{oOf|fvHh6@Y!e1bqpoKwMOg)?TA znSC(xo0PY0n}W~ZO$rS~WW>+`e?;pJK>(_X{*`=Kc_{FTz2C^PN<77nFy+|;2xp8< zzm^9?mqcQV8QaZfcT`fFjSpo#l(>M@qrt^JS#OnNDEg9^va=_5Lf|BBvKv$oiaTHa=?&|FztSI0~l3&$OS0|Jlrwn=Oc3x07>Mm zIgoD?b1UODCWU8)QW&qotTcM1BxkH4>5mb?xORRl8I19SlG=qB18Gto8QtE+yoH-3UE-P~>4o6-HZ4a2c_imWKet)6PNeg=9rkM`%?c@&(YfnB-XpdFDv0YI%Ml3>vZ?qLrV16fDDl_|HU>3-14L}~w zQa$h5V6fkw8Olm1bAJZBL~7B{B*EFrec6nG=xD5k*v*l5%Q&c$i0an7)5hn@VIIc&-sUiMJMxS|r^X&~yJ7oEc@0u^h?bomL?nAQ`z zM4W=G7`QGZ5tc*XIr^iEfMiEAZG2r&=v^ChbT{0 zQ;DwDgP+%fw}%XN=mYQ4x4TXJkg*Sk2fL?~q|LfN>4(3dT#>#}XBGp1i%@=SubVHz z&Z^n^lUL$u+s3hXvJPO)sQK1ibf4?^mvg{=R0@=-MEsfpsWudf{HO+KW z0ZC{$UR9KnQTO3wAx2dqi$WmS{9${Y&wBrF-u#j}!&>m>Pu`J2l;x13rq~vIXMi`K-=nB~o( zVZ<(A+n7n>lazh05tndGLt#(Uyao~Cn%IUa5Hg_<;8;`@rWrEm#BA1!@c^@;z9o{@tG)IZ$@(&XUBbnU-&b@Oq zG==*tS7a_b%;Q*$gwU`X&5Qs7I2+`_2J^BRb-z5izN$G`#NF?CsDngCQfp??>}7D@-v&50KFWGP=j2=Q;}SVd24dp97{1sZBKYt)Tjy z5HC!40YD>(RM21T7SvZKNqyD5xR%vlAFb~!0>Rc^M>QE~lc0s5Tg?DV8eoh!G_z}) zg^F2if7}2K-rX>_HO%b|LzG(pZ~8yQ!_z-hPWBfK`$&Tdcf4cz-Q0&;OA;T&sEBfE ztEn-93p3jfi}qU0DHqAN!6$urBeN4h9cn@ zinew#WWbsK_h$B+CL{qEgiHb=k^>YXU<{x5Di3j2l666Ru(oX+Iz&g_|HvJ|+1 z(Djq8d0s120CLQuyUKoF$RF((qe=Kv25Wu}2{r!;loAs;G@NP=AZI03`0S{ZvvMmP7cfSQ3(mhHadZLZpB%Xok7kgCxWLNcuE?V4nPerNMYB7jJ zSxf77hFp0^X=uoY&_e@nr3RU4#d@P=2>mfL(rQZyo7K<~dH!oN{W!!H=(8Xxy@CqeLythek1u{kRC12w!iw@l?Mg z{s{w=4v3r>#BW}FvLR-SLDz1IU3LsCV-cCms4T2!HI%r#UQmn~vsU>Ckdkr4;Ktwg z;wp9%^td@ro1FPvm;HQ~xwFgvd&&KVjj4d0d(uw$v?O0HeT;iE*VNYC3{PMb%>#Q^ zz7`-WCwa}#DKI1xv)Cxe+&u|^${urZk5T$TO~+}#6O$6Z6|)`_PjElu2r;WE;`Q7jXezt001#c~Ra_Yk ztIEpJAaKXxk{5Fqd(-JSVDKxVM3&-OTCs-qukM06Q%dDtBGO}9X-aXq8dAc@pT!*N zK&v~OMe*)r3~h(tRH!FN7UTemqoF|f=)u|WSng_xxQ^b==o=bFsK5aHB_0WE8B${2 zx*mUU1ObPrNX{b2>EO1P|3maN;a(k63nw|I*T2%>BVJAsv2|5a>%4$ipB8U&n5SE-WiQv)}&S|lB|4FhV*8EH!YB5d@VoeyL-GfqcFe|{Z zc7-3h0)7|?Wg(Qig%XSb-rx#7_=t%<4VlRXh=@)rlkI@zQ=G!IaSmN!9XPlG8H2|2 zi!ae^PfufqptDanT%M;1QQsCQqS$dX+NuD>&!`h0@!75oedI0zjxQ;31B ztS-p@%g)~z!SO=aIQqh(x$|qlq zFC@WS=nNyy*fI;0nf-j3Wuvgkak3K|Oc$$-Zevx$8CnrvO-n0_h2pH@{NnEZu;9AH z?9c-I9f2O!4L~$%vBS4=W0LCl%PfLxj2$IIYk=tx6lPK|ovH|D1)YT=CIUh)vA3Ry zX`Qg-fEmt4G{wdfcP5Te1-e<}m&2G{c$$9l0bU($&Gm(?xgNoUqyaMFGR})o0-5_q zw&JaJ&K7I7+T%i*gz|)G=Y?tK>y#p-_i-CtkK{%Aq?okFEGpKnc`*SK<6oPYxSf~T z^m0KW0}B{b4XU+g0B;39f%ScNi0w&uHe=4S$+-&L9$&^_GhC!4jUP9Pi^x`i{SbBi zo2UusmK1AXgrJr!Ci#X*uMx2B{|G?2EA7&j|`6v@dj&RCFU zU~MsY2^a=+E};xqgP0@fR=;M8H(UMwq3jOjj8IlWd0Hq>31$B4ap8RLEuK~VB)+1= z3j%8KMT=i!U+v!r`WgI^=NMK-{oaU`Vmj=2s<;q%dHhg%>DOA46EKv5=atIx&+di800bVR);nL%J# zDQqD7kwCLxLGLMfM1P(UZ%49K7DJiH$Ymsaw2%t1BNPQ9(^@bpUP1&qfMd9-I(TK( z?Vred>a+Hz zl7oZUp2pHo?%aw0Jsa~$u=G`$L*0x_iRc!}8rVonJgpdyTg{rV6^~5YmeVbwL(VhN zBL&h2kUmv1Q8swm4@`z701R+VYZyi+zXHb>m(jX}JGHK--B~2=WtVTZ_^vlFo?wOo zIpKTk2-eTBGPOy@3pR$LuWbNBBQ-r-@#q6?VHE|`0@ZjjRAWpy@fl}CSDOTe0vSjL znwN41I~g;QlIo0WPy<{eNv(!dVYO3C#X@ETOi?f`h5XW>iMEOHpcI{h=_lvaaIVy) zpsmE$xT}-iaxa`6dE^2wMh8Ui0E>)ax-~s)OKP^J>9le(oNu(#F1^o$FWI#6^l3lm zPoEX8T%F&VY)Na;jEl4Amx7tM&?M{MX5K;6#@DAPWFa`%+1?GGRa_qkc&%{ue(UCe zf+n+OR>@AXcj9JCU|u=ZeT81|P#Bt`hUQj74m*j&xL=*z;YiH$8U>55yE!}P3QW0p z9b}JL5Pue)#+rZTSO00ke6)Veu45d{EkdN$-E27%1~TRBbJpG!uKL@pxyzcn;VGb< zOu6tmJ#laT+dY=E))RMI_fWWM9{XC8XtyEA(up8gmnl(ESj zf(#U7ruILK{hcHOAa0t)M#N(MjtnU?auZ}@pd|2Pl1;H}86@?G08PTC0o9}K36!{Z z6Ly(jkj!`ZFSAo?HPK1xjVWTmKJ12SP8Y??6Gb5W}`|ES!Iosoi!-i&G(h#lL`#g7_pa zPYdVBuQdXp#0J6-*kSI>h}Ni853S3nuMNMl8w)XAD8I#X=^05zme_UZvqspgW{@b8 zr1b9K9vWEmKeOzdnf;41hf2nXLpsPx6s6)W&w*?Sx(v1)>LGl1uXvDRwE&qB2gXHa?j)GnPtw{! z+3g`RBfc{!^+xjdfdeO6ZysfiMjKO&$sSBEGyGEr=V=4UpGs!Wn`35QY>0nSThPbC zuFPLHg$1D&LHlnEJ739f7q=>0SVYPt4PijkNM0IJcFkBCZAD@tI2L!1Ra&?n=1fDX z+@OS^NH@kukUV>1ye5>zP~Kp|$5r`9BYc?(WiynsH=qEUf$0=fAG-}mccNMCT*mPU zL-8IJ^Epn8K9LmuE%wjG{uc=E%`ao~i(6k8u|TN8$4zDAKdsdLpWux4ZWlM|?K)6xhL?YaBnm z19Hs~Q{bIgwxiZ?7Z|XzKO55Cq#rWws#er=6rW5KgwEbb1@vx6Le|ONysMi{(^hXv zsy8Pc6Np-?_Ms5U`BCG#?cD)we-)?#aIC-IxCe}fIKPM0l@hKVqI^?OW)o!Eu7J>Z z$e8;im_c8)lD{ zL6c>=lWLq03w%Z!LTcYHSwI@HoTv81`3yI)?5R>6(CSj&7L%IgLPOB{IO|GlT-vkT z4V-3$xY|{jsCU)Y$2S$rfN-T;tyV_rLFrX1 ziRS_xJCYrFwsoabMchvw71!fIJdVZTStRKIe#LF6vn5s_BpHrQEqJR!+GrBn`Y>Ca*5Ve8!h3Kma;aNafIb%ii ze(V+$dr55%d<_ptBifQ)D%3O*ZX4vNRb-bIpO$@}x7dSSnR~Xx;frsV0Qwfg> z5U6Qk7hrpwq=aZQ2oqy>F}vDb02roCITwl3F&E{HN#=Rcg&cM+vMyu*x}&5!p!mZb z-L!*IGTq4?sD{lA4igKvIkR_U>7AMG0=`yA=U>?jK;H2NVh^Kg$ndRWib!xq>(su|+m4nKYrkph-ck(n%rIh6vq-nI0#E z&HA-FKflJTUwcF-J3?7`ehq{mdoYxFzVgRFlSqe)q;mxpPLI|X%J*mXfvoT)fWgCYw$h10yd1k^-agPXOZPL`mawdacNxkL4 zF3qEWEPQ5)&=bA`+T2IRo8qE`SBLv&1S7O&;{s|znUntU&D$gZo%(cSvgx$c$C*Yr=x`ShEfd0}1pE?Y7E);?m+FBU5} zP6{P|X&PMtg2T_@jphyJDT#<}2bqqlNiC^jHSrWZB&-X1EM!(?I8Df`m1Zvza%wfQ zvAjcuk{wCBOk|Z?z=R+NM`}yP`uKowZ7(?3*@nFU!~5a|MrYe!@m3~{wA`i|kMV06 zJAwS~EXE%=D`A9HQqlknAB(;;Sb8;b06n1oOp;jTe@nRjDi~x^VI~JdGir)T;jT2x=^kjNkRARUkF$i`k;cHJ2GAqtDW7GPd?GqD zk0+g93(P)1*@?2Xo|LhM5tcx1BcSBzJ>s3QOn6g>Fk~%LfqWo}&KFAKTH;A`kKtKo zh@?N%6>P5J=+X`Z0D4-tBOWL=>68|oFo_Z#!-s<4xPvO-*cF>B7;6FlL`Weaoln;H zndg)A`9@1GAl#z`NkzG~z{VS0eSbiDy?MrgpO8A`iYhDbShr?o?Dv_#He;}mn(zB*nrobqZft@?^OAJ6xTE*n~DT z8m1Rpom(Y94l#$f43Q3lR(ih_J=+w|$gAFkeu+7seiZ`)uFgznx|`Ly&qKV3 z1t7qQ48~=Na$_w$x8B-7N$O-h-uhwNyu|wBap#WJsLl2qPB^TcDd{E9~l37z|%i@TKL@Ox4Z`3g}YQ34A3y8pWgV(m6!>v$jd5F|wAK@$RV{=lPF9#pu|;Bi#14IX z(zq>Y(YF#T%g_wuCyE&(9qgPvsg1l$mZRVB`5ebNgdu%z2t~TyOyrm}B~uPG$r@=o zlXYEdlD(Np)-R$b@!p7fRArS@w699yfVn%2E4?eQDahYnnhGJ|GSX=8++#VSr=T%-3R4W;XyHxP27V+9CCWd-G4uuV z$*{qd7Dl%t6yOPnS#ybNUg{_ffyb;l+nMw9$!1Wg*18x|RWGc>2t&s30w8v7KB^5^ zjSg5=?vi87QBndmbt{h$Ad2iM<$9?W-Qx>qW1`P1;i7RUn8_LHow4rOJ5p zbMoctO(CL42F3S9hytWlkhkVWBAtkzic4`jBLdqBoHDy@-X4ru?1J3h)Vc%3ev}Lx zD{rjnJu6k%vq;O-Q|{DGAB<~cV1tI8^LnWnU5sK$%oJ3G_VQo*&fj?ZYj1w;%@f}K z#G4W?Jd@hTA z$)MPqAjw-ZaV^#eY~-S69*Uc5y4JB?kj}fX(7QuQ>M&2f}NO`wQl@(wb8!Mm2c|-#9e&B074|l9Q(`4Pc)~%;st^F<0$Vj7BKrtXe7Sd{#KEt) zoy|>2=gkQvVi|jjp1dLHM=*r?KQp)FcYjX#bskFUk0j(an_EK4ExTk3*%u}^hy~?W zf?Fjhis=&a!0ny6l@dbylyKJAGytdrBqc?_mc}pqszt7dNtk?swSazDn=S}X2)#a} zzkL|aiKQ>Ok))rBXBBz~&s-&Cst_$hek+r+a2sCAr(y&vqCrHSl2OyLcoDhMeX(9x z#nf!JO9%lT{TcdKNiek35d+cQW`cr0@7Xe~QVGj3%VD@lm=T05TMdGsG$J1tnu={+ zD6_bkwOiw61y#z8!-j;6VijFV-`L%)wV!m&K+sT)u0VuXO#BC|{j4*eNZlWtBi0>b zUA@E_RAaQ6qs@Eo#^6x@FuBMe;+((AX6#{pKa7Y_tFamy1~jSHPfSt3ro4QmotmC&|GP{p9}Us#w_5Yj;f z=;~spWB*#<_hKt6n_*o4iz&60z(AKF;+FleYSuEw9#xYhC&;g{W10D)q^&}dQV))1gUy2N zZ-0PR{fI{aTVJ>vN-p8JN#$^Qz;?G|Gq!+U^Q9 zVj!bHg0rG%_i6hWEq!OPS0R@ny|%4~oLG?%1=6Ze_+pZy?W7j{uFxA@TNF5Mbq8;8 zW~*~su*k%5!r)R{brWFEIQm2}?IcGpQa6Bsf5nRA?+{#7*9^GYL7*VX)~typEb_Mq zDLKEXBFwFkE0LeDY9^RU$tE|N!vlqxY=C@(I_L`6Ran-+c%y(1D?pmQM)U==Y9dhp zsY?o72*fT!{9rCZ`JNc0VDA)R^?li4d@n1Wox)~?#ULOC{M`?;6^ZWZbdeoMvZM;O;7 zA2!)XxU)sbbuuoX3bYlxmTL&e9Uv|jQkj9SH>-q^>I8oBoh^*CHoCl|)|2t}J@IyY zbQL0TLfKoR4QVpsiws*(=u4zsAYK&x5^flv2YBmN#2)`3#%eBmp zW7-E?E~ZA&3pqk;hUG>O{3QBb3FPiz=WRWJQKe0r^f6e&Z6K1ybc}V-0;1J)A*Aw( zNuTu-+6Y|^@zt_S9ucyUSC5FVVB;^txguCflXJdl_m=*b!g)6B3PM##)XgUP9o51U zp%N9jhgN#REkS99d|26+z+^FAN3(nrr&3;zhIs;Q#aw>HYzokraI=a2Ge{<}L4tLG zu?SMQAY0y|P*G0Uv%wCSXa=M-4*BwW`WMJiOB~UBIA`a3OI@nZ+`luysa~)$R77N| zv<59sOZ*yp4nm{gm5NYQKNp5t^P+5CjD#HK>zb%kg_sqrvuCVsfoBhF0&=r%_@oj2 zRNkYwdRz4&DFX@7f?w9j^^>aPf(TRI0WpC1sliV|*Gx8ZQ$DGvToHb~Y2;*ojK~Al zF>r{!^yxzOsX{>W6)F0-tuMc=5N;#oE(sre?Yctwp9{1|kTo;!K_|!-sXQzy8x6`$ zIe+TraF2E*79|5g6UOo;kXf9ki}+pn(o;$8`-%NgQv6}k3K5_3gdD4oB6Z_Js@_pE zsSZ5u-P<2!FzS2M$w=AiaYiJ2UNLZz2VqK=!+x3nh6)ViBM~@LC_L2U7jBhTNrR&Y zfc-uAoHg$|*hryB09FpUzp0;l0uTUi5BHRipcZu<_73BMEFpsg0Nfj(bnJgS5gK<|Ka7?c>)6C{hTi7dtcE~)ua3-dwV!LCjT)A4QyJ-%T;rj2+KyaA?= zdl<$ERFT0~>ZSABOQaD!&Knr%`%L{_V|FUcXYpcD4QZljQ5U3BO>?anUb{g?S%nH| zN=nJ2Za1Mk@>>zyL2$OwltbkK{a+31B1@JQO@X?NjgN6>hZB`iVqO3jJNCRW>+QTy z9vI4{&odIA16MVI{i0#Gz?$Am0gmK@O@n;|trA>pu|FHY7A~m9p?hRe0^Zj)E=7~L zJID~tAb5q%1+P&lgww#wzzN`i=NGzRu=0>>ci765Vspd{lV=-}K74OLU}58kbhPbh zhgBNGEX{z#s<0w_Taou|98;SHtX194vYoi8PS6tuDs@lD>N|i=3PE!-nr6mY;R;ao zvXUzyC5Kv?=Z%55wXF9hR1tWDY@w2KM3>@*DlC8j?+3%F1-mWVWXxt0x63rT5ikzs zn4#MgLQWul+$W{TX7PrpooKaRj^{vpQB<2nDUqikw{EZKS<<2th(@zyrnIO%c?189 z_17f6j-iePG!r~N3z|i_8(DA4&(Is@_3$v4sd8anOloT!#u%p>%SdMX9bL?22|6C5 zRjIkjh>fRlYf-MCW%@~>4A^5yt--W(d8J4vuVxl?XHgh)wQQsLK)wlp?AZh;OHus` zCK_Y>SZc_J*l}5Z#H8$Ei#F(xgTS@`mM3(G8A3D?1h~l3ldQ=NeXYKcJ&eZ6J&wJ@ z!=q{v!ZmSQqKzTv!e>w#d$~|D7K+buBq46e?Wvjf?gkSbONc)${t9a;_)7Aa%kyBI zutg^9l~5*f`8)-Ez3e$KOM|Yentn|4$b!loA^q2klAIJ_0Qd+`8rK(SQ1US{#S3U$ z+pv@7aW2XS)IQg*gG}a5C0T!an9iXbM+eTS7S&fppP=>7fRrleoHQBLBv0jGMfd|l zWqxLk`ml2W%c}27sBNn`pxu1BV#9g=7R|$Y6Fm%tk23{HRf%$nVoIx##SGCW>jKnk zU^s3mk}6eil=Uh`;^3W-dxjnmg`_nh2Zc5a3(1PjDBHQVo)5CMXXfxIXVE||RCCQF zGE~GJpnPe&N!!~-`^urI`CvRmNeq?37F%j>v4qRg&uCdh<72eD{hor75BZa8Y?`fg z)Bvzqqe!Ack|OjW<&#^xFu}959g%1-PYR!}qWCPfMrwgMXB>jF6PdJq>A7OZpNjUm zV(CvsM_{Kv6_Hd6BP^t`ya0yE7Bq`RRziv!KAYO~3g||;)s5D6p0fwi?m|5(a88$o zVwj;A)?8jmNS`@!}B zRFv!o^R)1NrP-bEdi!Ma&FqYjD6JXWPjJ*)K>CD7(DWQYyTjr*MQBG|H0Ic7jh&eW z>pnNLFle#Si#f7rh21Hg8w{DT1Y!}&U;w%RC@O;N<=ZV*MN3p~U{g`3Ky8~v``hsp<0lxF@xp`y(-%2+92H^xS!RtHj-#b^X}U0&b6P_rD8CCWR2&xd zTnhx#L}IYvw1{{UNTvfd*mb!?u+M|B+N}xeh*sfW8(Jl#!Bn(c#vUT?^t332GLT+j zrL%>R2|1W~6q%&u9J0Jfh=_1tg42(g^iwk4U&47j z#@tR@sY)nA$d|to*pQ~$%9o9~*2LKFHO&>qjkd5Z_JfFD@*xEKs$YC#sP)^gNP1#1In!p{1Hmx@ULV!+12s#wxN8Q`5ZU4GL*TUCua}cwKv+9mNB>HMqgj63KF^jgd-5iTGU)*m8 zdby_noFKp}kRM{#qP~awCr}4fqBGb=JsWc{hhv8mr(NfTZzV>`zN}J*(dJUKdvcI@ zG0riE$D^q{3imQ=%|0}74GZOLnQ!8XlG?q^J_;o5UPK0mPa22T>>zuvc`M*@sd4`- zC#jpICZ7YT1z@l?Yvu^{yeEvqLxW$55 z;t@OC%i=A7(>!w&a!M^Q^9zf)f8oSZx1ni+*C+dJ0mPzzp|)-OmUo?Sh{ z;_u)PGLuWG=a5B%K6Oo>9^~fV9(@jqO{4)!8E*vo3${_78z>p3iR*zGVI1}Kvm%1*HJ|wx>PQ9|I_+-9o-20o(KC3a##-K8s98w8uc^O5L#iI zjPZyr!~_g<-z18%HjTt%Vh)aybXD}@C~&qgu&11)2>zOAW}{)IZgzAd!UgRpVLJKL z8D=?*zs@-LL1*E-06r&R0EtU;gm)zSxSHAyNHVrbY{3jAUtbTA?GPxbu3q2Y)q3LC z@C1a*LE5kBi#=I0b{I2KfHt@`JjJ@wTS59#i0Ds{l7G@@7Es{Sxko}ZR z+Ud_v%y<1`U_irukiQ^}o`vS4zyEiYx204k(gJo1RW5CAXuYEcFFe$vyVJC4x)qX6 zla?Ss(Rkwif$D@FF_qfU+kH~MMR9|%P?a**M1pAbtsfOlKI*T}`}XS4H^%U-|K~QO z(X|-^{r^rIhFb)v@-s06YC{-;#dKdZgHm74p0?M%tBqr^7ED8#6{7 z^dE%8Bo7f97U{jZmoDwZF^{7?%j&Mq$Czb*G;KLGx&qG>*dOWH!Zh0t8uu~dXTVCS zWuP|pFi{D73J~Z&yNo}P7#E{R1!Wqxt5!F%Om>H7=_Ti?O%{OMuSbymHhsUxq!iw4 zxS((<(zr>U*oTIzgxv5&EkL=E4q5h^>*bLcmfJ0KI&Z4zk^MzFyjDe~A`m6Px@?xd zhOYoPaRuK>WFv>WLzv^k%^k*^XihM1a|8_j zJv+tU8kRz{&=+X-eeo?hs(3OFLin-P$batS7KA^3vEat@A!ESG+DM5CRO0U~wm%+k z!S;>Jj)40Y#_uH|bn0(ne+q-Z&OD2eW)L)fi!@9Z7!A<_2C*E%AtnObH@v-RW52sz zuV>I`274Ds49yClTWyu4c@6A%-kqU1GV!&@Dtj_wgtR1{Wj$Vw-6# zHSOe5uj?~d{qxC=9(He;^Xt0Av22=0!!H%0jch^N-k>PWFMLBS-YL9cSc`EplwjCr zTqx}(0@C;wkGWx%`B-&W;##;hNKqPWW(S88Gs{62HKS(`fk~>7C+c7}q?ra4Gj_^;jsviQR1y`el>|QBT${P6 zW;WM`Lm7uM4duR}+&h%JhBE)Pu&IWj&uL?R*o;($hfFP;Rfk86VT_%|x%JzysgAkP zdh|;KMv@;fdbAgEpoz0kj?aKW@CE&V3mc%p5l76u^vP;B3j_ z5!*>#i-8t!i9_1FoHn>sa|OOg&!bb5gP?OHF0njo^sW zKzH&_WON8fp+LGLn5)BRi;2QLT)z#Ms#vMfDa6^Dhk-E0|CM?0;ZEiOi-S}DOY=}+ z9;(r|(Nu4r2L!VJPxEj#L1DMg!vlqF^Kb#NVYko2BL%aiV4f_vCkke3!M|_%Q^V^E zxS)M1b$<-=5ZL{E9%L`{e=`r=%mWxh^HBS%d6>;S@D9zx4a~!-|D}0A^bzJk7RH;+ zDFK_PG?QS4OJC!=}u)X7nozI|NM%&Z>YJ=3g@N%`EY3!FvRTZRG_kR{9#FlsJ<<&WRd&i2M8{6DUx57s9Pxg> zv2uNIoN>?`>y&BTWu2ThW>64JQ7$F`wA>!o~PS zFD?-BWbkHVKhG@^-@jyo2$@PK^9Yegg!4szRptur{5o>}N|bzF=x==5u~RsYhilR0 zhHbm<_Qhc5CG3?rH@FWNiMybC^bKtmTLF>yR@u2AV#G*ci}C(c`%3)i#;c|3TRL}5 z&oKL%bx3B`AlD>IvvnCym=i65x)@AJM+qAj0aNwF3;Jm^A=J(kzxW8rZf1)QFC+1AskNHHO zAlg<$C41d8r zdNUgNOx4@eSw}h@Ahvt5dto}%Nia}GyMsP6h#$Tm|84`p#!FM%$3bwi4>@6BLXwQT3 zNTRRqtU{FZGFba_%o}l#T|aSYQDLkQ%=}9W!miRK06FDvc(LZS2aDz*` z2#^-LpBi^3WL_ADc`xX%e4J7{{sI6}7!vRg$@GW_Fw9zk!i=c_yz17X33_o9^@h&P z3!NkE4#YE!esOfGH=Xc=#pEV$ZuUf8h>=+O^nPz1klp|y4Cyx!;c%iZ3IlrZTL-{5 z0*+G<)b(ap4M7F7zT)B0bv&T^=@%fv!yy!rur5R_He8(TA~I7+>CIBN(k(_$uH0{$ z^8NB1YtLY#hXAIcMg~#v>OSb~bsRm2sEb5^c}fxGL>vV_YId<+*9Va)sR!m1abjk} z&`JmcQ_7nVL20DArHWep`oJC(qav}yJ_2yUgsYPr02^b2gSaTYjHtH# zTw@)xLyExSe`pN(?u77oYWp>Wqx=rgn(qg@^)QZ)%Wlj*=!l&+Q%qAJ;uv{AP3**5 zd7c$;CKhGIEJ%%{SU}v}O;m?TvGL|R-rVlpEsT5U z#y#GB$-AG^S%yRJT4L7d!|iP*{a|Z35CKYnzS8$fDp8d06kQOQf+e#uhjT>+flytv zt`J=dGltYF%qbR-;ghTZxMj85vtHLcb@pLMo1%f}wLMiax)^5d3q24=yK_Tl$m`(| zktCR8vvS#6v>Bimg(wnXBc=$BPeK(uZo3h{_YvYk=7{{h zq&?f4NPMCHX%jKWz0|6afRh?|+K!Ffk47r!*=B$6CT}+T-cY_Jlm~<|44Ji5vt=)5++o3z zbsDAQL;#F+uofbC!0rXZ+r&AsKMe};nmtKTI#A(BuZC|Xlpn)v(~UJko78%u>cgo1 zEpfnz6EEB&B?1c5b~}S2Q}Mp(jXa$krS5)_$mZq36;^N-v)sj$R+Qs=Aj0>RcXJD)I zCU|=8C!;8W1Q5+uGq^=G^+YHep{$3pw1ptERotsD#h$x8@Zc-VKIlS%ky0m()FIv$ zjVDW^X$H`YFJQyO@PGglG>`MDTVj@0mX!S^WD++Fx@1gD?%bS*Q`*; zvb3`xyQxVx1q4gVrX2%uA98?s3ie_Z3r0*F7?yqk=kkJ2%H;c)UBC&C?r~)Ga*?P% zvZxP8_dX%(A`yQJY->1c399)wO4MJ?(=z;2huuRhw9KgCJCV}RkLQNn+{=q+Fj?Y) z$h#1Xpym;i=v8(NcHYbnHs|pk>{2(e(~xo`!V`{UxJ3m&mZMH z147(JEZXG7Pr?HDAxWUHJX6#oymMfWpc(w!#4Xo$8_|?7V9=jz z3uv|p4-?)SjBL=Mu~7xufc|tawHVvTDoQ%FkW^Y^B6di=fw8A>z=xnxv>>`X5<0Vr zcJYJS^(ifh&1U5$Y|yHQN1wnGxz#}!REz-$lNxw&=mO~90Cq6T<1PD@-9RMN+CX+q zaBxAkY?FvWMrSwMTkR%mHru6}Y;ROOL&MEc|Y(vUOFtJP)xunW!4 zbgJ64m2n8mGft+M5g5p@k&G4di9U=gFf`l#og6&OJmYkJW-iRwR#Px2 z^o{BQsa}?uD>9%0Czx=x+k7%J*I|9-fi;l^2$opnY}s1uop7C;0U2VI?L5CSak~Ia zw2jRLh~S{c4`$riC3{37wW;mlxAMYvIl zm&=^J#on^nKRn{1PMb6P0L9YoV{f4G%8 zUeNL``*WAx1yQ!W<^F-h-k;b967yLZEz$D4@)x!o7;Lxv5uq-%{I(>te79}wzPQ!y zoh-8PyxpDlwTQUcGExi@^Fi5_c8gnC{XPR>+P??pbeNof1AHPL2kK54HsNx`qky=I zSkVdHvJ^qM3RBa7N(1#2=m>iVgA9)mBs`F=;1D5_K*Tx`F&Nf6)Wif-i4P*ZQ@(Sc zTmt5z=t}9guy+IXbp{;ricpe#8!UKeD0kn$TC87ZqlcxsBQA)j4)+XlCP4mY3RS}A zuLVfm{yvl3g19*7E(8Ln0GIzL2w_XejTHvD@>(K{!g8DIy!~fSX?{r8V419YO&EVb zjD@5@#<>JYetj>T0rRRrajZicbgpxM zrpbZ$k==YX9#y3+G0`$A4Wk$g`g^H#HkGb7gAz)VRV4UZ53G&`8JS}N@?*@2#S>8N z*!XBx(+MqJc?s3S)Xaf+<%+b&Xg^u@DDbvd7xcXfZoibe_M)z}X5ZqvSoQ{MsB2w| zMO{i=0PbAs3M^_~mlEF>qJes}k04b>AY0AOTSUGlp)5);JaBw;DAe5qb&s1l#o6Lp zdzdjd25il4nfvbt9--nXfQ`5TO6r#4rB=WL9b?n{Wur7+lH2wx;`S>l(ZJEs%XkIK zJB*{ciwgY9D*!aImHDpLX0R`>U^w&&Vg$%bCV$1;?XMW&71;Yh9K3v!g*Iq}@||WA zk%k9m(#Jm;Y#wpxDF(b)cB{>W#=g{$(CY!gr^22<$UujJE(r_+1hV4;?KzjO*rJay z0g4R|$;)y>Q(s7kyu*ZZm+mbU5a$%wrJGt+APxC4h`5od2;Gv9QrIXy-W^8_C*nyZ zrzUNOKxBg~2fYu0abTj3p|>5#2E;vO{y_S7XFxJk!3@wHb@o$qkly-DxG5q%j#g)^ zE)XBD;e+~+W%r4~(I=N=EviKayan=`JmN@IfDEldFhvcr)`fV>EJK#7zNK8nthv;{ zTM5jsMXWkt62rZI0H2DPz7dq3&fX--0(j`*{WW4Q^x6Y$3EzJDAn{|TxmGK7vzaTx za8RN<#V#9(-VCn>C4&|fJVgB$v0{@87*=Ypw5Y=Qv@=kJvlNj7UWIpLWTkNPCz;tsWg<%_h(}{YnO;U8ot>G)qaD zBu^yvNu1#PkCQMJ){^~!Zv89xO={mtE zf}G&kRbVUo`2{pMrJ@inZUU?=;oKe^3Pz?)MV1UoGul2M<$R2zgQE-g_Q4IJ>h_K< zlrbF;f_z^fBWDD@iI)(`$&{UDi1w&O2R!MX32wzMFh$J5PV#D;$}VO`2@TCH((q`j z7Na%@NOYM&$aO7QiSgO`bcC?iw$P8ocBqk+A#bXSs3856ic+~pVudQdr*(%W@)g>W zdHn_heQiF>>yV8h#NiXgQr;|4)kiXLhtTZbutmq8z z+vFXhb)KskMpYU99V!D6q@>222gH($g>!`WTp11}TP^3gFIbxIS`R}J*taWQMe6UB zj+A6F*vCZCZ?#-NT;UkNW^8Z4K@}MACl#W>q3AE9oX{AP!$`>t)T;cwGSFa?&G1Re z{oIcW4(?}!ax9d^P#zx2kx-VWU7r=smv4aB8f+>z4M16gm>uA=mrqI&&8SrPvHB)*VX3yHZeAvp0;KcDs`?!&~jH}7Q^g`z2xKbn}Uk-@nMKW?rwrMJL| z-m4pksKG24X!}a`AjlMUO^LH@MUogX2&^Q+;}h-yY0~!-^AyxTqZ^5+T&Q-pvy-z; zsZctToj`t5?ftI(!=lgOoD)KGBGgGY3A77K)53WXYDUdX*UAr zuOy30enze|`1ufOkJ_vDI6|>N)|EZ|=&d;Gg%g)CKMS+67vJB8&lFR`f|qq++x%`$>s+MUN)DOmb)Zfe{1Iu6;Ck$s zAfBoZF{g#kQ_TX!Gl`+&{37On&4gEj_}tNunFSx5And8;3~qUaes3ej!pr^X?AE6x zM^S2JZTrRTgO`+q=ZBc1{Zaa*U1xT%EL`28@!lZRu@(l)K^o-G5G@Vm(V<+zI0EEn z+Lt1*hiHC^VQJ8G3DY`&{BEE&OM2i**!JOi5ck0lcmn=<;!;VhA#?~{Z7rQS02Y`M zvr0fV)L0}?Z?J=Hwni<5&^;&6$obNQ0_0yNMTiz3OP}|!Sx3ruXiu%e)apHRc+VKTon%|Q zdRIRWCokSLxAl&hJG=`u?#rX+co(DAdKcr(DWpjLKH^7Kr@awpHKHKoZ)_pEW4IGi zaH*dPp2H_MK|+wn*zq23J<*Cl76V(wv-Pn*CUK zI^k)yWiU!*$LCA@5Vz`B8S03JNu`<7|G$~67 z0m8rbqGBVRG)rltkDDx!ep)Pm$OY!R0eS%NN`%;4Py8}u%W}q9kEZ`DUYR8Ku!Ujs zUpLpTG&hWxcMiSfalh9c=NOies=+mp?n&MOD*11nsjGkg32}LJf;NF%5klBiA*k49 zvtkpkrd`ZD2vgpL?XnU(8hW3%&(z}YGz=*wo~b3AtZx=SX>@W$U*%{0^j7U=`-Q|{ zNdrS*3(oaE&;i2*iJ=U(a^2E^!Vkq;?L~RW5x2z>Y$OjkVz*cfQpYv39K+J!Ap+{) zd?B(OEOxbos6iD6=(e5Dc=J>5D1YqD6a;G0ycHmqNalfpe|OnE%vjPI0Gg%!bA!yL z3RO`qyKhNS(`UZ!%{NoO0gR0;_sW3nl#@WStr#v1RQNidpW~%&(8Ln*76q?gXcV`B z4vy0pE$dwMlKoykJEG?Z>ynTGS}WhGqc zn|4ng!F+YNl9&5E4?}~Q0~t~{2>&U;ReTViKhcQv6ZK9`L&eVQ0cRd{+1(}~VHQcS zF1PkFYd>ks4W{xjW3D%qYb4BnzVm2nXn+*voViYEZ7fXBq@UiOx2gGb< zEKEGk3MoKE;Khi_3gpXeicQ*+)lny*qX1h}Ij5o6_$!X#Lh2eq(sLm>OzdRZ3qy*6 zm=RJ(em|A{`;__Bl>gZj3TsUBnFA2|+16ZW&A)i_h|RGDeP=1te?2y0znRA0P0hPv z^Luap1F}Ou@(X7bBK5J?a$(SXJGD=cMiT}tffl@hfbR6_YZ9Z7Wodv8qR9kpnP+@~ z9O#4g*3{mTx{p+~sbkt3KEcS7b=l7PU}~>S3r*0SHCy8PR>sGC zKlV#u(_oELcZpbaYp#jw*Xk#*sQGC7=cne=mH(Ksze5Np#hU&~BT&YS&$ z>z|q3!YI^4(_{Nh@$5_^*hxGjkUJITT0;SrX+r2P!P$J39Rb@MQy46EV8=uqQYPNV zZZXOeEJz`%FweX@1>o@KW#wpYu=eb4W*B`GP+&{-rn`kC-vCXdTfIp`vi~KbbY$p^NCrX zzABV!LU~{)`$IVv%6Xx5p7JsPbMDd10p!~b zZLib66S&~7?tv&Q{q=nWxZqwrt85-RM-f@7am3|fsH2nfV!067q4N>}v%2|IeMRn< zSNI@@>L}1=Js%S)u2NoH>S|plcY3(r6XG!MSGkn&o5Iay1UhjC+C413m$lHtxV(_N z=JLiNT0CA_!CewdIKIE3+!67;aO-Sy%WMc1KA)dr)8;caf0{IB5sH6j^W0SU)!jPV zKR@Nx8(j{rGwko|M@!2y5bJASdTh#nZOVRo%0D_K>{cq`29cTH^_k!HfzH5c_97{t z#GwDsm}iXrsfpi7WH&8A`+-jLew|zQp-%G=$W`~zPP0*;*L0feM84kD<*rkb-v8?T zAM1=Sp8|(Tx^dZ*xnhdCuADNL>Y8;Qo-$YK%7!U(EuZ$Er_A;Gymz*{p3qJI?dLu| z6|YBj?Ea&}{1FkW-Oy>TKt?d_%AY#o%Q4!D-&Z&KbekIXR@qu^zw&{40Ycy@Wd^qX z$6e-K#%)C^iFARCk-=;XErS1bm-${7xIBoZw( zym;eH_?vgGGk@;3|JiN-#uK-7n~gkSujw{7@O~^K@Qvw?ujQ+=pXxIg>EcGGsE6MD zw)=0qG`_Oitk+AgclP702R`D=hxNcE-S*Rc(nr~A`porx@dk;+60!*1KC^NQzDe{Z zCMXMAQuiT$tLh(gn@3tz|GLk7TWx7qoy9-sw!i5E^I7w6edby0INi_sWJV`~y!4SN zf7z71YRZ3Ts{BVF(fz@iKU?>oN^2Irp&i18&|)5chso}gMUC3hZ~r(cR>1x1g!$eC z6wvo4s_T+tdu}oQgFXJ--uPkAY`0dxT>E&R{VohVsXG#XUw8EaHak!*Vq(5O*R|re z-S)g0=G+9;NgBZLFy#6Ck-`>KLI+C-Ph8d;^&~|Sed8Q`Z|UiaYqR`UuIX->K&CvZ)Y9U zH!H)}t(JD8UM}?4GP1T0aFpwnYK9=RTv=7C@x4+%qe*+DF#ZL4Cs-WWe2P+|A?%3S z;ic$aiNG1~B&?Y@R+b$^7r{;$H#S{jw+bgiDa(Ln+QEY+WxtH49lja&B4X=}ryYdo z><#y}9UWVra6mJBUpz>fxxt{1BqWNT6_2H(W(-NIQm#$JP#8M~LvJz3nuN)u^*d@H_g)K_{8T?dOUP2LS9t^%-dBQrwlG6v;b{$3xZC zP=BB`dtG!N1Oi%ptSvre-1Srbqf`FsDf8<|(BVRp41gGUGkLgTIS=c#ERhCJijPg_ z#*LO?N#2nmEmsZ^pb|BJ7WhKqC^`u;#M~)}j{20O%FY90qzY%Rywd+53lu5a*i-}8 zfM_j#RbhwRkk$1Am6_>YI)z&+%g#Y-I3H$L0qyl6afT zRQN$Qsa#ozmcU*aB2qoOYqWJuCu z{?nLqNJwdhvH2${3mOB#x}vtwHB^LMFpX7m2_Isw_vTx1{5_&yOLce;_HUAZsdo%( zQ2gnMzkBc#$!N4GD@qa87g(7y%+rtl=+08|#f%J@kWm7hhhy`#5WK+L7n@DGa(8U* z(C4kO`Mf?q$0#!eTbt&IW~IO!OBN;uDToiYeM)8_b>Mo_qj9cc)_|d>ag)(KmE;m{ zuL$F2pGeGB(4M)27qLKNU&ReI8fY2+^Cm=<$SaJK@x^)*!s%4;$PRUlM&GU^AM^Ip zd0oFw%8{!A@=>l+_J#Pr|9<;d+`0 z!k_|rW#~r|7jlwHtB5jV77AftH&sxadU6B-MFLB0cCXzm*@_YNRCxFsXy=bbXM zdo(o{kC{6L7;E8}4xIe3It}t{JscB>Dwv#%u-#UbGC@9xghq0!Jr*5(fEvi3jap!2 z%+A7k{-qm;MVg!${jQO{cfh`9ASe;)Pyq}SoGtW<9C^S_pky_WICMQNNf0-Lw+gsr zySs|6vit`8B$%s9@kwBm?Q0}BwdOUF*0|k*CNR&_M5 zZv_RRvmiMJXwjKs8DJo?W1+`JL|SqI=&K49*iT&AFud_%Jvb&RQh9^ zg-ziT(N8)WLcwRyGnt zdwxT2x~l5mRk0Tn3Um6K?8kJMy_wjIfBj1ThL*&n2f)9Y1*!RcNg(F1qlDo=tiWHf z*4r^7m_h4m1-ZGhME2I=CXXvm>~aoe=$42tOlK}G3G}#V^1pRpr`ke`lNy{#jL3ut zOxXN);N|4Ha_j)$Qlf=n2>EZv?1Lf(E<9FuEo{VhR8d>rwOTHZM zo@0FkWsn7Oj~0R0I^{}59VOpcc}Z>Fq&<@}oV+wQZ^|7Q(jZlWg-&iG;H%B!mXF_9 zbXN!C`IA=rMXNm=_P=bk$677y!A@k+#r#~tgS-X*f*duE=K$3W4^)v}jrcWq43wg( z@PYvkIHz^N2V%rnz}oWyzD~nFR*LSlQ|pdxGWv4|jLE~5uT3=y!-aX0RV_ad)@GYn$D==J@DVcX(_4+`~s<5r>o%;;&2fdU%rqPaWPA z|Fz9+g8o%42D;XA`FpedwV983vDyBl;|tC9I~|{EcK_Q4pnWejvubJT_svQ`I-^kt z6)Xb=DmcOT1M%?|pH5c6_&_FKdS@Vp9ee&Oil5*-Ax&0Cf?{!XE%|DC4nZJ6Ngzc* ze>}w9dqN;it=XEIJ6LL-l9cgh#BP}L7`yAd`rqf&?ChGIheruBAfj@5&7D?rf2alx zJMyXfRnV_bN_U29iSo8koBLJT_euBNnw?OylV}GO!|63Y*H5zTl07G0O?`h{SvbMp z+r|?_x6OiCN&RA+`pZ!3f^Cg==-w#91Nvz8H>hyVA#U?B9FaE)F*EGTWH!O^EtaLK zrImVmpyCb?v8Bl1I5y4LureCyeEaq@`k9X-oq=IRwM)7C^_u z&_Nk69_u&ng0iu3ncLp2AW}?ZP-2&SdBmOb7Z2Uxkji|uN=ly_ym(4l)G zlj5y0^=B(9Z`xS4m6aPT9AHXE(&#LaBYUw-U3Dpg6K(y^blvcON+gjpbi)Z_h`0T1% zQ?0D7+IMj;43k1z>NWF1>?ZRo);*UFrJ+|M;@HXX*}yXSd z`Wep+nx?mD+R0DuSGjnWgdwxHfD3w9dU&;=nnBGBRsCYkRV%p5bua?as;zOU5|^w_ zDa;5InO||(eTM_P+8y@abAf9WEq~rcH*TF>IoGb43vQ_)^%5|Qjj!EC5ML*$T%c^1 zAQ5845JJ>6_ho)J0u4xh7uSf^u3@q1rR)tWAh+YRa2s6rroWr`8?#a$x(9@3K})j8 zjq?0(L=3h8b>T!jOCNih<&Q>drfzc(ly@E#P?5x5z;O(MqJh_^Kz8eLK9j@0pJ?AEMqbT% zKzFj*>+_;;`_fB|hp`Wp?)vusZ$@fCbYxO^IMknh#0Uf4G$Q-jd%_oe#6C{{KWU$4 z{~7x%`_H@MoE?t{K>vTKNqk7jsWpk)1BCe`@+k|I{+&q-pR@O-nUp@|b-$;|49m@L zrTb{$*tXw@7K#~0RUnxG-4gzA;Y_}#N_F=3=$-}wTz_7&hwwoVzXs3C{qAoWq8<4Q z?K#A5U#Poh{3Vk#cYcEBGmIG3)sDls7H&J72jinj-s4Y7(BGY?>;C+tP_AZ}6q9O3 zJ^5lVDLA(`-&iIBDkO^U!M2eQsF1i}w?M^+=#D#|sSDkfBM7v=gZ1LdRcJqdH6=j^ zBV#=wUvCFBtZTjAuWhw;t+%?ht>tyCprX%n-`0GGBwEEM%2-7;J&iD*!-wNoY0$+{ zMr4?4!WAk2te*g6E%6M}gzA$Ux=F&IUI|K!{g?a?DwbaZ zZ^iyA4D%)VT_}ch$8SKKaB(lB_G~IX?z%F50(0X>8thTY*Xwaf4WFUMiejrORDw5P~KK@w}ZCq5*J?vsx-=ONN`D-{Wl>B7}7u|XhRjz!tMB7CwSi! zlZoJB@jbIkp(&)pls4;0&|~t0%r0c^;1}+q%oT7+;*w=(p+9N=3YY&du^kj&e46?X z*qXOVYWU?j@&F0ns5XjoXVQrOQRZ%x7oAIq!b)+@;iHJrSZ2E;ZVuu`+v2_h(ZD8k zP9fRa4+x~l7+iI%K}7F z?S?<=#qftO*H$pRDy6rFdoM_1Z#cglslY_${@5Q2F8O=_GLu>&tZ*tqvBIQfsNiY` zzE2YJX|+)0CAfj%I>Mr9r)CBfjMR(nj!~%~dxpuhMd2WVOS?whh@Zt@vk$->b)sXg z7EyUJo#Q9cx#h_bf}aZuqzo=kRl|94s-Y7#f7>ij)EXNEv9GZt>~mFU+iK`jL7vZA z`*H>6VRn>>iB`6-v>s@(`(&<`_s`KaJGN#+wis@&At)9CS+Sqj>=y_k_>=6S6@WeG zVKg_dF&y1jv&WnKQ^dihgUPp>?ZW2j`|5P`!)hWmZnoo_OI_)M&d!AL@cXp_K&9z(iZHdMGy$G{yI5r+~68dYim5Xm6Pg0d5Dratkjv# zeOvOX60QRe2N_`;W!udy1zj}QZdwGT_!R6dB1Ed+K)?%vk8CN^8zt|8YL_f2_yXk{ zQP02*W%zhKSL zTNeix_~|$cg`zdERoLDl(>+^ZIkEs8&O==wOekY8;$Ze+IwbpuY~b^Vl87>>Ty#`{ zgME==d*(kTVK#bi5OxqB@%EAAf`s^823jIT7N4o0nU(f2nXGF21Be)_ujww_+wSgf zx1S*byYo8iAMHC`?x-&Ou;1VxY$1$9Tp?c7Q@GL>Y961L-*#0&abb;0K|f141M;KJ z^CirfXrt&|f?_1hS1tZW*C#WSOfrE8!G*n(Y3hjzh3K3f|5BH|(gorL5GV)0J#lnt zPvPKh`v&GDdQb!tSD|KgktW0vc7VkcOQ;Zz6>JGxKQUzmtv9_#SXoEpI8HW;I6BvL zhWmA{C$>#=V^85b-R}5qaH!Bm!JcfjZzc{){P)8O#kB`@aq52ks>{!JL37r3ZWaTc zX|vzqAq5_KqRoDZyNx~CW)JK5V4M9+$9vk`FFWnAP9N@L2<^Ex!^cX`Jk@58>z)m5 z_K1!Twb=tY-q+?H@3eD7T(0U!k@@1rMxPwqU4E*oPRB^D9SB)x{N*raF$sji;NuZ2 zfeJtEvX~5=72$RXj6-*_#ipc3=vBy_0fnVdoqjeu|2p|@$ZQvYSSGNwAfnnA-AV{~Aol`(XJ_}SyN`A2|TFLFHCW+ z8qsQ^lj;IaRKXyPWCo`fqQWGj&?M#tWV=(820&ay-t6B(vsd&OCx7ov|JLjOSZ|gd z4br2UdL$i(CSo#rk;h?7e{6Oam+XYj%>#c{pRMlW+9`dkI9@^l)^q!kQ~EabaawJI zHltl3x?+zQ(l9bX{_t&Rqw=Vof-q1gM)T(e(#w18s$LAq?wVeIH7ezf%X)cQJ;Jje z#6h+OeS0Gag!u>oakQ)sOMejp-4wHxNOvL5bSQdpAbqmep6=B^5PL?c!a7FmVb(y1 zRP#cF;F=6?#1(TAo?}dkv?TP=2rgSUh@u3_D@@Zk&0{G~Io)eSnVA=w3{C)@LZ%C& zzYi8}>$5xi?2bNncVGMy#;fDjzB*QvHc4MY#Fo6D!`B2pt^PGJUlR&V3We2pf0&m! z(cMFZQwQyw!7zru>9!ZTg|zsP?+@C=gVk!i`3WEZG&z6}8c-^05s(zN3tXP67S-VH zo*61!J!m)NtsUB9-|A6o=-qA^v^#07&Ss6%8r|%tHL+NM(HYB1dSkN*Tlr4eoEKY#4@^ z6HNknCA65jqeFEy$npTzTF5~W(ni*c0tFoTwG_WY3red=&JT<$2#7EUQ3gd8Jn1Sh83{J7KgX?t(Iy@I&Opvw|lg1wD7=S^3b6DTpd7peB*!} zH)O{TdG<#R*%3qTJ}%|gH0jdHZ&m6W)=@MIw*?>b zVYmw+=(>i8gpxARfO4c>W)dX?&=bZKE~dZw8a+2km$Ih)>y5L-ze4yI{z(KWoU&9Ga=!d%E9#L&iV`>9KxW-_O+iqThb5 zZZ?R`H}1}|Hy=U3Vi;*yzxiA zjsg&QuZ@o;7ZWSCSF5w8g}E&H+E`)rkUMS2P95U2&Kt5PN9?H)?H?Th_QkDaQ0PdVFt?35!SDzi>i{7UGf$Q~q1RM-%^72btB2fmLw2p+aWkMdY6p(8 zcQ211ndlV?=Fvjj^dIl+2=AkDxcK0~Q;P>Tzk2zrU+w&YFr|6YZeXwy-8EKta>zYB zWWUmjpW`Ek?X+R-PZ_q8huvwz$eSAe(}x-U|eKZhP^S=FJxr9SWhmu~s7g8oVl z{n$HpUhRnanhg8Fu%8|FbCo%-u#iWk1<`m7uoL5`%;Yow{>SKRv#R$F+x^4#v*F~y zVY_9R5xHyFelQB`o35b5?EH=a_1;)IsTPy7|dF{UmpRLe`ACY z=4S`ob86{;F?*4&ZEERdEMWeCvDUy`-7{Rz=Ew8#g?x3ap{2626jXCi|CKSWl+$*S zJCq1W>L;o$NqD3f{{>DD5AN`BK%uOhd22>S0L5?leSeCJMiwZGWDPlm@d?UlC2r8%Rhs zMa7q_vEI;M6m+HN!U<5oj&7S6FVq@*Dr|Et?$AN9b}VQQ1KcTe(5zkXlkCK)lNnbi zwCW^#H!pgHmiY6Vv+cXHHSO|!fPP;zioL+D8%oE5gDb{2+=SQaqPuG-ok&NMYSI^& zI6bo`U1iwY;oOUuHK7pUB~l0_H8fDJI4^XkM?FckYtrsDYzw^$KF#z{Y>Tvx`dh1Y zn`qr+;oEcEG3ZN?c$g*o%cy;8%#NEggI_sTzsB(&Cfs8az~LlmciIBSa`m`fH;%px z<$f}shbP_dN4H$HFH(~)K6g%X!JNkJliyFIk4@CM)=%aiw;&#N-7*6f;(yN}ii*@S z+c2rPLi6Pc@M_Le#!-H!(3mu|;8qb53b>|?@)YXdNZPDh=b|abBH6F{+4{IC$fP3j z;5NuHPZYLNYpt`cwMGWS&b7E8{dsQTunB+UgdH(K-;bJbw@lcr6UP4L3A<^+aqj$W z>^jy>)WMULU&p6v8ZEc7=_g45OqA}M0W$~695BUH2ryULV$DxkY<2Z@sPnvN)mD5x ziU|LmIhzIRHM8x^*?|HBOUKQoM`SSg&*+?a<(KBVS6B$-`EquL=+&p}xk)>H()O^IC+(K4?C*22>-|B$ zyjj1T+hW&uZvN?SbKT)Y+%thUO>Q2uyTzs%ddSm$83ar7^-ov*(L%ZmwbBp}{tg{)9Vg?;J~81d>4f`Dvz^z%$Ga<9 zjGi6XV!M;(y5yDbIkD4qg$P1^xvYi2jaypm0^GZ#P(m#ZQ;kC7qL%p9mgeP+9aS-2 z5@XOa;wB;Sid^K)_~2AYeZz0%_yi|`Qkl$A42Yvq2se5{stG{)84~o)`WuCBm85Pk z?P5n~Zqlig+IhOGem4p8J-U>Xq}iTwEpa&J=h!ZB8jwa(u}hmbdFRH1lnqn1DQSK~ z-2f$5#+RihVXOAWJv*|r_FD|jACP$Y@#wcx2bg9A(yFC7xL}G{8$CB(%J5T1!#!eF;n*Q1@_GTnHNIZjZajPqIj`tMDUD7Q|HVO{nF? z3u8RjMAmYibcBppq=;`A65L7GCy>_G^akfrP*>TussAozwG4#|>wj-oKS~>ww}ld0 z;;cYNrqFISsZ*tw6lr0(h?a1c1J>&hllq-N2`KILeiy6AO@U?HEDOV1zUG>46T#D4 z1*$@luYZ@UO_Iw21dgP|mZ2!_W_R2QA7m5xKinvZ0>cbY= zkqfuHrYFv*Uqex^`?j`2%=U0UB#uHqxMUYy_}fBTxrpp&V3}mn{cNE|P!3p0Xe7c>9$5`9gbWA(OyY<6isA zh4$P+|H~=+y4T6G3pW$iKTg?8d27F&vS;=5(^K}Oj=X2Rj=z|4;Q?Ox!j%1GiWmP+ z9s=DJ+@EH=>F-l^(1OtK7pLq`YMJJKr{i-|_G=xVnsR@evcHq%3|GbL7ut;rlS?!j z?w+~sNBYy=1QWs8rm?t7Umn`Lb;@p18#heZS{<*MvMY7GY|7oV&~Dy57CZWHE^@~% zO8zw0{!FYO-u>TMKkE*LHp8|IPz>-9zh{618rk25@MU< z&P_Io8Mcb9DtAB~T_}}pDKWZ!i%m8xhgYyRzr|LbLz}P@VktIncnD4waTreK^9@t5 z7>oS2+(%Xtzs2Sj+jh-MNldd0U(p@i4hdP)+IMz4$)$hA+~tAg#frcewndjM##^fh zs{ZMeyKf5Q1K4?mAqf7$=64<8-^u{cVw3xh+yWJ~egx+D%)KYqeFK?kAK=aj<%XB7 zC03~;k1i?wV6h8XB~G2NGnBM?Rn-nq`G}QOJG9DLb?%i$dr7g(FE-krINBc@?e{uf zIpNM*Tzs~%`b=ZJLo4zQ&7bZNB>xRNMCZ2bRJS@+qf<{WDP6hPl%m4E-D=0Qx;3MA z@+dDnY1A%R>@HngxOy=e3984A)?Yg}e{B!3I$<0D_)rUp2?u&a#sD^SmayLq(F;pT zKiZ=8^G5A#wSLy9UAx#_x43XqXnhT>zp|wC<5$ho86$Q!+EIpO&4`^m!myk)VkhW$ z{D^&rqa8D1w=8zIE-oA~QayYmbm$G{B=1+4l|z=69(dJ6oL{wbH4*1jdKm`ovihHQq@k-@%XC!bg{c*#IyI%+}z zHk7ZU*u6{m8kEnJ;Ls<~L}39&$2M0cy;mpF_gHG$d@h84|)uU=;{5rq{vJd>!!F?~N{8 zS~_4!z9w^@n}K{(+Z2S)IsSI71nqa#%OUv_~ zHbgfpEgiiiU$MEf&ofo~m4M=@s{JMhlgFyIURNHi+QR~mhpP5~j`vry$E*ISs{i&9 zJ9dft&JuU*lJq-E{B;x6yQVFEd*=hVD0DYK(gfX^{YKrL?lQXc8C}ki%rM||-_pX# zOZ+L`pX>>&g$USR7uhR|($!1+nk8jcW5A%w)z+QAxS6bgaZX&kn-IjHdS+zU0WUI& zbp{kZCKUeY(!$wGfV@A?w%^RQ=V!aC6uAhnq~|X2=Pl9Lh4yFXbI?RsP&|#*w+d}D z=;AUqx30==E3KQZKfSbY(USPmCFvzg+@(tjmn`uYFRAypH}7jDCMai`5Du%o;B){v zK-iC7SX#Jxi_YEEWVbch?M>;m40-6B!*H$9q9PcfWL9TiQdr3)IvyDC57M15MBcqKA0o8DvQbZWNogb0rT64rV(CwJiLRis zuXjkMsx^v(Z;US7zQk}M5Wvy#=3GG&yCasmBbQp($44ynFHg_Rc>dZ(=r?K9l+nKw z&qH5B7nx<+*UX5VFBds4>=WI!J^Bf(B)XrV%Q!xJsXu3F!v6TBb^5M2o1%OSG`>|@ zrY1XRt5OWK9>>}!T0*;WY?8)wr3K?xz^GMW)OSTsZC|(^E&l5zG?KFaXTl|tzGHtg zMzQS2BX;wMzhtSuY^nd@QvYIm^@?e1vIo6@KbehEe=vt+=iVyF<3H3(xz!bLkg0P9 z_SkCwYV`cF!XM3(yYr9c`YQJtIqLbazRFKI_-XzJMFZAowFYa=ibl@^qE_rmc<($E zJgg}JINinBPwJUUrjnr{WA1XKR7t9RR2^w8p_)SSQnjQ~Rwkw*qzr4}sFY$!rCjc- zRHMC1@3-A>mU*z-uJ1;9gPZsz6uCh2vAaFizNG`9zqZ59>c|I+x%^>=?fiNOI^KQOhw;o*jxzA#U2z;SY($!WUQ7wo&bmv=EO{tq%GbVm0R*$kXL znfxyjo!l)&%%Msvi8BWpyLyjgnj6*dp>ZPp1y_r*0U>s3|KDwV0%BxK^vAq2EJ?Cc zb|Oq_g=t+>8V9oI_HufCnmaAn4~J7+7cL)W&@~uZ7-e3Ka7tM;W*|5ff#rziP^KD3 zev1XG(R0U*tK&j+Pq})!b7!&KW~1r|f^Psq#xf7A90bmpjTUkp^Q)-%^3cXM(`}Ta z$I7)!9P1`;Bgj{I8^TEn&K%5TtybFL*J8Y9E)VS-pd+~JC^O1YmGp7A2x&o{!WCdB zerm`+I}{S1=)?MRuxQzw3Plbl-~(jJ(<T+q=QU0g{I z^6n7AX!9FI07n7=^kGkRJc!|5VP9PB$VyE~pa)y!4@;bWiFlD>@qmiGNIjKU0Str+ z7vXh@kvqFFBhK|cVm$Di;Tn^eAozD4VZR9k+47kwUoKtN}E+9N=M{Fc&w;Ob5yGLo71u4RtAmN}i)B(C)J}_Z*dKHfgxkN6Zoe z1tmMHC|!Ycs%4ekQZ|pa=#{oubllHh?`?zs=z2!|oNiLZD4{Jk7!}U}U^OEXy!i#N z+og1ppk$qo1)U7=E+_|FuP@uGvYnLJBcca{iJ{gb?FXx2wp-x0roC}BiajN>pO;;J z?~lkh+?ipls8SzH>%qHs(DAJRN3y`>kX&@TVJ>K4^fTR~JkKYhhu%wg2A7>7S!`n~ z&kw*Eq<}>DE7)jTFa&+-Tn#0Bm%=z_V-P0 zmyEKKacK>Ce&SOz+tm?8duB$}P#VMj5vGwebx;JUu$zKdEKgM!E~?zo4B|yA4V-8A zR@<39SOg>&;jphUQVfXOH`x!HF$F2Xu`oy+^+p~zm#{3k(*Cf=UDab(_SE0n&fEAL zRR1|igjdkzGXTuw( z9U$HB!yE4$Z2Z$ui0z_sB5#C|9O8|50c)7@m*Kq>UKSkA+v-DxBWgZmfF48EwrR-L zpj&aLPDpY)bD}M0HzWgVfY8b@9+}etk6I4FB?I!CXnjLJ^NA&BAHpKu`1=ZfU1LN> z5lWo8>(RiId-rSn9yDV@m+yD`qWo-f@3;4%{3J^T*15e(AJQGN9dO4-?PJ`rk9}m* z@B5TW)hV}t{q}BI!$%;fxW=X@(|709D@wJQ6&WHIU7L0=azS3WZ91kWHMUe)g`Ox0OS;3Q(6nefGUjFrfjYVh0vSY~B0or(!duc5Bzw?cfK#^E2sD0T#U zU`y}jp*MdCb_XEf^7XFU)<0qm+(5~ZzLwT088FmH2vc-Sju<{97%7AV_ zEJ7}?*}ey2PEJh_fiK2(#?VnsOdP3Qo7i_*P2u5xbsu3^@oBSh@(^x*(~4{ld+Qd> z{~z*b9w5a24FHeUkqTt1+E{BM-Lce$q5&|#sXr2ye22hsk^dW5l}>&l)Jo1^tH=TZ z8vH~)B^cn~A2EIxwb)9J#>Gcs;otDL%QhrCtxs%2viEw-2h6Ihjx4rtl%8OE88q0Q zjm3sq^x@LCCbF+jSnys--{gaW<(6){v)kW+t_1_;hJOFcetV+d{k-2E?QeLbA3Rk} zo9Q-E@jMWfn=Edd4nR@G*a(Gg1^!d~i(Fv{$q+V_;4}b(Bcppi_>8+tF~ub5cfE(f zLsxhYi$)!HG-P|0il`G@wlaR#Mo%^iKV$!8ul*i`(~WKBb>L+xH1&Z z+rmC?bMOt$53-FAR3+`^!?Zeuq_rtlN*~MrY+zT8(&EZAqh&LFpN=@CZ1FeDN|202 z3t#oi>)p+@%}=Pk5Qc>Ez|Zh;k>6Bc;kA&r_=T)bTm0Q!&$dqPYx>PYn&t2n`WhuQ zef{er;*|IfG#m8ZlO;V4UooGj#MAV<{~)~ND62d=UVeDoa8$ZwT)s!7Ga0w#ZUx&f z#*^?%%rBIYW!nM&zvaAeN82ImgeyI~+Yaj{oR<8r5J_?>XpFr7gR>W%`wQX(x`8Y{ zrcG=OEAGlRyG-ZKU)yHu5F7mUT{^|+4RM(teJ7G8I%8%sX#~pb3-LB)7naKa=&mBC6ygQUlOU{sbqMk*4Mn3ZC zD=)P&Lvz8~I0d+W@)LMCyvcU*JKOetS#k|V!sN;(({Vz#NQH&K<4t?`FqnnEy1}t;F>D z`gw)d&5sYB%KkU^vWzH1TK4}iox%HWv+eKrCi(-$-3}S~j7fx5wuOa{#2>YN-G0e- z@$STQWdQX@Aw9D@4@;@c_PPn!QR*;!_x_j5RHhnqhFLO+_0*+EG*gMsK%B~PgSam0+#^_mVr5h# z;TRPMgfKwn05OHNU`I%X1$RpXp-lZXGAfZtV8(|0WK^Jeb-r!v*RJhbF?R=SJ;xd8JV^d?kj9%^Egi5H^q8Tsyzf{ZGQ2lU$!Ij zg(t>iu@77Kjxnp*?a%^m8f&%htQ87s;B9&%1F=)~7Wd{DD(&`q&2M{-yaT}X$=mE* z9Ny;MlfB#CgbO#e*?+#?U4LyOMs?F&e*h67 z8XRD+ck99J_S?<3{w=@tYweSV<%#3594h+~Wbrfcgo@Ob;?qPOnGUJ1={_93sw0se z2Gwf~@X4fhg=eQGW?(}*_V0_n+6<1SKo5_=>5*DiU>$o!rERb`uNRZvu1dqLpN-aF zM^wfV`bUQiLLX`i?{BUr7C~u5gqt+XJ^yj2dD9i^LEemlU0y;T~&cTK2GyXm5 z5-H`Cl0b`zzTyN*V7Zjtdg}>GHH5$zm!yUEPxz;${TK?Pdgun6*wO7A;#uaxw;Ow$*(9^&Z>OC4K>6xV{ z(s;0uoR&JD;V+^!FmS3Yp?*_Yz8b%kq|x|Pe5#?;K_eDBc@wwuPATlubQfgs6z3xi zJAhd~mT{TRJ0|&09h4Lqv^!`4o7PJUc!;g1wr=Z%7PTw?4WNx2vbDH4K~z9m%r*_? ze+M`uw0O(E@wugu@PX9bb8n&JOf9z02Dp5t_08_hNC(`JA%~-{b@AXn1Lz`C_crrr zB~HRYIRDpjx+EAe;A&V)*zYkOMH^~pg`NTU^o>lP zxj^lxW#p-2gGd1{6@EdTL0t(+{c#!XihA-nEg({SqbE@i@doQzj{@$%cwGM6*rHLR z9uo64ab3)usIC&xOM=wiaVftlThC3k|aM zv*Ei>LJ6W4*~Z!C*S7|aT6ohNvz1J;eS-BphvUv12clJm=+J2YhR~h7lM18z+bXkv zhGTW^J@z(|gTIGfecNDWoPQUc9CS-c!yrE{9Thq(Y@dOLj~C4Fvck#aWC8PXk27S= zv*fbg=?-IfKnmY7rEEirt-K%9#OdK?Z>MS4ucC8DnqKurzw$oPyuJ36oMQf`p?9C+^)$VM?F#O{Jn!aZLCd07p``b! zM#Gm?$y9ctWhWU*W)KKA31Iddi+^gWv~{kz^UR&iC9X-Y`I?uYhM=}G4?%yl|1fCz zlQaI#udV`y{kH)m&uDzpk^vHx;av@!9f&G}d3iUdvry|waJ(GbqU6Le5@@fmM~pzY z6pmxW?UgI_z25;L=mi#v;*NyI!|FtUHoRK2Qox-|o8Y>Fzath2ay3Dov;=bp6qdp07Zs5TEYt3=aw9FFHs%!Ha{JBSAjQCpo<^zJ*mgk1=F~wr0;2 z1!qD2e4dm)I5fp9Q&^&IA^=8lP(H*u8kwPztI4XPbuT43pCm%3>f2PXhY&{52o_=P z$e+Zo8wE%-UOYxI@2NWx&9)H1>|EbPm>5aLagOPV0=^$ZpLmqJPv^0&cy z*K^W4H+nC>>l5K#1c><=N*q`-%tT7U-^}78{TffLvBuRX4tqPXf=m0#6cJF7M3kCT z`Ehi1%v^=>=b(%NC45(S96pY8z|oiD=qqvbRUuxy32Z-QLr5uPe558CLPjq^VpaL| zHW7(}G(m^ZspyP%$->x|PM0laMj*>tMk#VYFiVaB#6^n#l+X-)|O?O+hA=f z?_Y>I-xWPazX`=CB~4%(Sq`hH4%7pXCAqZdM<)1Z<0`r!*t%h&Z^X~~cqHxy*>BBz zG~{%G?)zA~oYp?oIO-;AHBR?8F$aM8FG{mi(m^C=DloeL=?%(TM51+t0r(4CwzojQ`Ua$r%R!zN6={DJf$!fMO8-B>3itU zROm|~>Q7azrWeVu3)?`*qLhGZlCHSBHF|5hEu|fAG=BtLNft55#N+Z~ACA65ZaATA z4jM%qkm>^gj{?#Nj7ELmoDV}2o#+A)aGziayKOopU}tIpo(u;N8E`Rdge}HS#%xS7 z8xpmY-Kf8r40y|;1DQR(}H4H4>Sg#5lNrNj~j9RZNmE5#WQxG7Ml|(lMj~ ztoKG$Na;e6tP@&mjP7i%^oZ>cjA)u=@x`8vW-(+0Q6KVn5sKEU54|mg*1_nsydOJ< ze(Xe)+jKwPoV?ZjLvgFZc={fCLO9kN^kkUVFpI@_SNR9zJz->D+mm8iU{I=gE>X*H z)$Q+$z9HW!Iv5%kpG#BtQSnjm>ERb)ko|&8Rtftw>@lL$qW%W5;}9YT73hFURnmFz zMb+pmP^aISxw#^?OAv3|Rk8gjwi{#ntDum`TSuK6YTca4IPox$TZLbDw$|A?SNQ-e z4G5yjy`_fiYO9bNFwM?yd#=SY%LS~bQmVNtXQ9Dd6uSsBBGqkcNS-PDRhde?;1Q_ctW?z64puzg`Bs&;5I4ntJ z)x|1*zXhe=~}{L+rgJSE#3zJLae8b zTQ;AwR*nQ0W+>31wP`4CRhH}<@=PWqg*?d>)>ypSJp1FMU0Gx9NoP;F;%X2f6I~Op zcv?7TsLKsWrKvcI3UJaLUbMebOd(8NT#y>~^BC1gduJZrcho&mu*VDOQ-%0RoL&m$ zi^_H})X&a#I$q?Gi-EP)?NKg>ag$uaI6)D4&}&B@NYvVtcjHQ}-U-Jk)38 zL;Yr+0a-R-TWLhxtodG&@-!r4g-X^eFHa#H*_K{C;MAB;w$GPruq`zqwlPuVR^oJ_ z>r_E~PxN%6#k;|_4`>NiC%5nxqV}Q-NQp0i-UE-N0F!tTQ(&M0%|aF(UMl`2eFH4? zVrr)%Y!nLz6z%U6ZSn^dZB@}7P>c@@`<2Cp6N{CjigkJ=4(+w5Y1f=)?FK7abj@O! zGW2Fe&_Z8~$PH^djf#ZQLfWDX4-xfP!d#3+y#mocwxn-dXA9RFp_Xabw}*W(>@i`M zq$8lY8EaGk)Wg~wi&mAhTQYZR=5ER2yEA}>zw$MBD>*p;nwN_1W!R!J(Vl7QepU2O z>#wAyay#mcEJpd_`3E`%?r2~tF-BX1YgZ~&`2l^5!{#1@1*7@Pd+@Dt{sr62=zC=? z-n`u<3$x^gXsZ$(W2FP#8_dW7d|nYAsyfurStnT1m%et?w*~MS>YWeav{8h0*E#B*BM5c7oa=2S-9=p3>PSFmAxVZ@KV zFB&dh=i_S;@2c1M_y_#igmaKKuD^Ve(bb%Z_aK_!}B>-s4|52-z)QX?A$X&t2J6fbr=5W_^xpFpPfDR@B06t-vVQ5^j@VAllPr; zWLq93;BbvhSz+v=FH=_}jrQq`?gXjnBCNo4MI)timz)>8g|g_=gQDAbQZfbq5l3Ix zQj;a=_BhK$poS>k4s=;8Ek#4wS?oX(D)p17%(2o)63P)JHFipQ9JQ$Pp`23z-Us!O zW?8VX`Iy5-8Hl3bQn|kTmNWlI0;!~xnMc8e{#KYEL%I!$haIs?X)(s5PTtx<-7DpG zg;K6~Ovg*1P$Wv=OY3P}T5x)=W!w5i6*gn*iC(bmY4cC2(I?P7d3T-JwFaqfCxk9*^eEvDF503c3e3+Exh#_j1q6f#; zFi1+^k7{4TByN=hvr}awR{*Z4Qa4KCvSciJ&Qn#sl|dkorWhTLQdPro7RT%{EB=y? zkse`X9yW=V6LshQjB?5-BMUAn%yV1KiLQv#_jyzU_E9>{RK@ptcONx8P^-i$b6OvN z)IE&9=5-{WgBmhmvwp?Z(v3+z(83YRj0c&Y!;6CtUW~l=Alwkp2wnyLDk$QDG*)8P zi)bW{$MV8cG*8l`^J0H4b6Y3r7<3$Rw<^ye4&#r?8%ohid}pMuV7Rrc<77Jas+ZSu z7amG5IAB*yy?&UEtg(Mtjom^#F#D#}$W1kJwmy$vZEt-Ti6?8{8rT*5`Zb|K~lyC@X zWhy|qW@WN*UIXh26DnUPCZ4n|WM9&FVA}^4+e^NsVNsYb?>@kw=%@MlywwPT+lr;q z7PnG9lY5G>T$Mw$LmKxm`7H>Olhu`ggA&6t_gDSaASB zJ++6NJr4Pz!C`bgBoYllDD7^@4>Iu@v%2@FbVbaVu7pjfL|v_WAEsy@cJCN-v(Og! zY@sKI4YvGKB?$WIj7~Y03M?Lz0v(Z!7^o6*B0?N(?#(`H^-bZ??;8oc^r&8u3EOm` zJ#Fnc#N5G#CCv^d%O?ec5N*J(YFb>z_2$-@4`T;#WkI{zgp5LcK21k~g+mFWK(BdR zAbd#9Bo3|R^{sDOU{kxYQIZO$gLrBfbsyzt8US9H7U%^sE74MNt7~mkDuZr-`K$Kk zy83Y#m2z_|^R@b`-&tpW4-oy)b0sl)MqU9CiA<8$ocLza#Sb~0Q zty--0E+|hG4l#eQ?G_z{8xuU)dRP22W*Ic2MbHdh!2C;_P~#PrApz?b5a)T=K2a(R zjEn_}gF|@$^kuwI@QUs!AS;>qh#F>2>1GvYmMYvviyIOV*Aifzq@PicOOe`RWr%@K zq*7gD+sI$KwPAq=CT%KSAa9TQ_6~b{8_aMVU4Saausm!A)(d>h-{=Iqohw{>zBlX? z7@|m;AKQ^=Zrvr`U+AIibkq%m!4y}^0vkY)$@PW+%o5RZU=BbH5p2^yhV}P<0t_LU z6$mGj&@irBq|1Fdv2! zx)w!YD&?4jtunPVdOk^CGPlavL6~%$W53Gz16<({Jih!ST{^}7kl>{X8D5t5>@gqsK$!`ZHI#9ZM2u!8LE?!ETU zIQ9_!o{ZvaQD9y0GT&pvEklLnbc|cw`{=1gkY`0e2*Y&aINgqA{7z-G&}8EKLrH^{ zV0am98Azi4=#6Hj82czul(!SeK5&EZMO#3otUFj(Q2%u@DG)}o5o^|Kfv+h=pUt-N z^UAnLHAjm7&9STme01lMr#jJ21jef7aPSeb742<~I&(&; zZVvq7@^bWr3?*)LH&^cYr+uQ{?BaI&K|5y4RoG?QSm1!J7tq5sxl7{g>&+?)1VE#b z0s(@*IjZmUB}|wU{G#tt%Toi2A9^|J(bMu^w*pY5-=+AoIo39} zlEIcB1LJC~R1@2Wky&Ydq6#6LvVd)9keBd9>SZG&xKJ%lFFvWShQh*2@+DVd*u@0~ z1&$WnIKf|PaR;?nhR(c+F9F8FXK*aIrO~qVDy+qH-7fQXiFfRO=#wja;SwPuy#j1m zXsUnHcz|efW@yn#%N0nn#J7Bj@129Tm`)znD{N`>)mqdH03bWomoE4u^lL=@iuaEr zzW2*5U#T&7`O5uV%V+S$Db|5g_Gpp$I&fmm@?kx-z~+3Z5G@2f>ywa9gm^kxWod46 zYxIL+?+Qo`fkm(3ol>t8oKk{I@(#5ziYpMhizKkby{-75OCAtBH6V70G367aFJIBT zJY7LxRAE=!CE3;Of)9QHkK1Ifo2+}J!|H;Kq`PJ)R9F{I41y%?H-tM8X_AH))mtDu zs90~I9s!gys%inSt_1-=+JevmL|s*W;wZK+`7dg*)lvF>g?%gg)dxTs-^BmHen@E- zOG-s}sO(oqN8%Y74uv!Z8$h;iY5FHsj{A42NV4Zmze7=7ym$7# zDP3Zi{h~sH?P`nc!wsLIy~6z-_P+4LlHKp^snmX1Yiy>z(v=x^;I=DKy<#?^c-8%}vv07fr*e5MUwu464tD zpw|r+vv}X40?&$#X#>@R132B!;&Bhep5xDAz)BsKNt|gSlaD+Ti}8R46YCx;Buq;q zs8LM;Ze;7jenJf}j)W%^Lo@bZys6202u~|b(m8=0SEMIX-_&9#2T^1;>m0RsX| zqJOPL^RmW71=hfH;wxE!7?VZ2B;Rpgh_Q$(E{3d{5kHF#Q;QA( zeyqO6b_+cwou|-hCrs1=xXXZ2%f2^OMCk_m_rMCwU5^r>v#VviwaAvRHRZkzXv7#% zN-&I*R9UoWyZ~)#9w#t8aJkNH!%K=79Wms);czZZ1k+z5X(OT?%Dh^k%#f5_Yn5l9 zAc$ojH`3>JxpDMg7!1AX7$2oK#qo;Q zX9+rAuQ!1{R_RUURlNyAL2r_er`u_OCN}k^A?oanRx#Z{F@p0Ba84QpI0K2Bm)Tm$ zo1JTMw(HzUp5-W8^DPNd5A0SMDts)SIea!UE|eaQA7f^XXo^XeB|*kiC!A|zm}RiV zu#LghFGs`0z1)4Xb3YN2Pe@vYp#s6j#WmmUz^K-VeGgH2b!Z{mk=^5XOKk5>32o~$qJyT$m6F>;;oHjEG% z&vUALAAH76v>>99Z{h)jBoYX40PGzDNS~1he12>-c%yTJi@t@$RJ2?mEf9ymhVl$r2&nB{tHNFuq`Ilzk3R~SZh z3yjKcmU^jArVUR{#hkC@FMJU3U{+^PoNU8Nqw)OCbl7%ryW-=_Z(*=-`_c2#J5BFH z`EH>N`b5vjSncsGXAm`*p3v_6zm9Z_f_#$D5Euv!M24y(lPh(jbIB6;lkkZ`jOhD?;`vY9w|v=@ z&-ktG|4M%MpR@mKOTlTlB!}r}4y%#k0XRCNyHN{xHzZgsPz-0|ZYQ={xyIMLpMiO&4D@8jO) z_bI+DW!ot_BuoFGl&pd-K^6D0Oc#+hb#2&c&sS_kyg9TH|4{2w&YECHnJqVEhd zE-683@y-UB1VNc#Ms$s5WnjsKIlJEMD*3_sb#%opF@mO9R!wy_QwyYr0gn}4UxcAQLJ{hBsy{#!)Crv40JWzjAoZTI#zl*a!qPO;KAh9++BosQ$ zB|c&zrI7(qBLhN@a9#NLdU-2+G3>LjFNgiDxH0ga#`kCM`*e#cH+X7gHE|?73CGi< z#kCe%l2)8f`Fp69;)su8IAUCpsoh`G?C*HkfcB^(fCrxHebIN&g8W$CiQ0Ab0)meO1=B~`_nk>0Gb5~{DeMgqunZLbLp7&b`>{>jEzhBRk&f{K zvHbyScEEge8ZKEsoo0!H`R#Qm>rAuM*mEC$4M7B60Bv)3Yo zT}PdUOc)dfn0>L1xZs^6>M71$kd%>Ei{+X5ZNihpm)D|S@FWB(j5Nc5>si1n{me3S zPMSS>E{!b{<;TKvixrptJP{;P5`q@P8|#AYQg9Z~Ast;a3^E-p@=J=avU3=w5wRQJ zAjzBFiJwAvwk{ct<&cHCw%%K=>#bXbCp$I3=kp|d3Qw}5xAg(;&aog%G4wWIKGN(& zkG|$b8Zlf}Y7rVnnXG6j?|=Qfw*}6?VH&OJ{)cZ*{W2saX+e-mnxW}{<3_JkCoi?R zmz(V+#fW033nGd#SOK}Bkfah2kpK|_a-_%$%I&WP<@S)2+wmA}7LbTgpBDcM&mlmo zfD$&vn~qUsfy2iAQG{Q_@!{kCuOk#YiVql#Plz+rw z7M2#Uvf0D8u#LgBSfP!~?MX1{3qh}QkGi!0LEML?v|+G;7)q}=%7KYcA~S8+htGvY zE}ml1D@k^0?9PbY>9Id6wlm|#@Hv%pux_DX{$_07iaq-y z7p{m2oZ!xjNdp>pTpFjig=+VIelFRIIDl}Ew8BAK^=U@9%Wq3)Ko>E3w8>KTkGmWJ2l9@tR5qSoRKUHx z1YKEc7kCvNSL5% zOa!q5bmYs~S408;e})`xt`Yw3^Vt`WC0qR%Dj>cq41iWQl|jOAYw)0q)p#o@)+l}< zeRc52U5StDMrSMi+rs`_{y={;fI?u6)F5gMVYXP}!u71}thlq2Slg_l2qF{_CjC znzpV(U5|O*b6lGy>|4>8-!Lb>jTGtlHE##U@D{flBE#Pa&qHHX#4ShP+(&TJ+SJyG z7Zw@4K21XtO}gTy8;oyE?MHaRrFKW!wI(IEd*%p`itz;?M9{v%*(baHv3 zEbvJ;_(W8v6;!NaQp}V(U6Tk-N*XkY&3LC4+m+6x9u>a&NM?E#E2 zSW4$4Bx!&bng!C--eqF|Z`FRE`c;`7l(l>#v#)2+!sO5)clc0z*idpMQ%@{4PjM!w`&;joHnW`WXb?* z^jpID?0lW}#_dQEJ?w?U--!v3r#15527i~q+8Vn_xkIHbYzSa*P#AiMHg~02@~5n{ z%RQR`?xHp<-u1Pv>O2WF?#=5R0m{F^qRqjr7hBMb7Yui@0nk28c$PuL zQRwO=799^vm!FzIXEPJ}5d;t@F#fHZ?;n^FqjG4odjY9|_WlZ%bhl;6?U~({wd3$f zfV%6)+3OBubo;n}DzhiE#uqa8`>cY4=#R0^l>#Yt_j~s6PeRqdI~ zex1qNz7!s6c`mbOv(evX_M2?*`M|EZXEPuEZhwIg1W&M3V7R$ zIjR^|jOL*jBAcP5IcrvP6K=ly55$fV3v(1%hL)(s>#t@Wn&I1W7=~eyxMR5P2;X2( z@gZGAd$j>y7g54S~PX83a$RR#;{{_ck+R{08e(Kby@1t{JZ*3NTSy zk|AMLfC-x_UaeccIQ<&Wi3U$#x*MCp(OUcpA+O zYB+DZLw5JDZQ%;Gvjyz8F?cnUnhK?;+K!wr2OY`++|D`Cshnd$XFp25f=j4<*=>-Cz27w3&D58KUcojiK?A@#0``&vE z5b1;>O%xMCks>XCf*=Wq0#ZZ)0YMT0v7rJAq9A#{zqRLt1LFO^pFd&m*=N_ap7pF} zJ^kTZ`St77VBLJG-m0s4YfwjU++kbH-MaRXQX1a)a?2-MU4;)RO+&_hDdR;imdpz! zb8;4)5Re0Sj-aZ6-%}P^v7GLgB@ymZYxY+q^8~hB^SzQ=TQX}(6E~1m#QvZ(dMj!P z_XiTzC+-4l!eJHqX33nZV{0!inM-AhxU6JW>-Wl%`L?8p@06nRM`)@T`}0Q31^9W{ z3rEZ%~hM>F%wY~bO{Je0X7GV`l!h~s+k3wG5sPvQ6Tr}^`y`DaV% zb0zz1X~uW^D4vCi>qqEk-F2C{J~Ovx=1-;8(#&iqnXvauGneSEzn9G4N&_#J+zTZ~ zzSzUF?5K>onqSSLqcVGBRyitTD&A^}J%nk&#Ucn_qoIC8#!N(01_!E`r!3vA@vDnp_ zZOe*gL+Y0n{jga&j1cmZwmq|AZmih9xBXw+=B2j(N86uOPQOvMXO(BHEW67~8lDwp zv#M;)E)SercGp+T{e$j|vN^q6I-_h)FGu0nhP`=UkgI-W*{>+~bNsc6G_OfhdN>MX6Y;Vi@FU2{RUw5fWokiaw zPz~sSOkXnuO)>iJ+ARBNW`2@Y8G|dc9!BIg6*q(K-8~%LIc)D8_V)~r-IDom#oe0O zTe7jV?`H1sw`GHOW#$vOgI!rRwCwAZZbQ$mm{q*poKrCuRQ!rcMB~q{Snj%t6g2kg za{u*Z&$+Kw=uO@ye@)qZr|j02&6@JyDP_|Kzr#d~u9D^F>WcYp#aw}g%UoDCE6X6T z~Iq&%@?9Jbmn6 zhs{&NVN$#>Y+llt4a4TJX>{Q5Y32z19!VNVzkFKtq&fbyEN)4m<&Mu<$7Js4tbAf- zj?DTG%gmvfJ1nz@X8t&!Q#vX0r)1_US?`&d`)cOCo|&)d*l@fuGb_UJ^V6zd?y3E$ z)Mb;j%;AC86%3?RB^g3skyxG}@n6-wZwxuYh8%c;Hhx3O~9~d*VobR&?4<;E zp+8~sRO~gR6dRc5HgVWeGyLYF1$L9lg5k|+@ut}pJ{10PTQtJ2-NKLp&n+4X;TwL- z`Zm$l>EaTqe@`(mL>(f1?dv?DVtzZnB;C<2nX#kW$-ds;pJa9!*xBt0ZQIT4UfRv> z9`Q@ErQO9+`Q7}R`ZKdVW{I+Uqj^*GM)PL7NAwoIr(yGZC2zHR(Z9W;efszI`#3!B zqP=JAo4w8KTYZ~(N6*{MJN(<@{iFTt{(ir)fAb+JnfMJ%b1IDjAwOkv9J>!jAF6u)(Vm_vXg9AbkJX32*ffj?UQ zb7ldOQB1-){XKbJIA)5o3U^hddw}7d4cker+Uf9T|BLK)shnY0rS2@-zsq5P>0@3< z%+l0uNX%g}s~(=3BlLS@YK}|Iv6As{>~K%D`JUGJr`zT)Xq4=;ZS(iG{af3-sK5Tv z))HmSp~L2gVRQH}_CVElZJKL~@pp?_9!c@MXrtBle{wGJOQ;86925U#^b80|yhx}r zMc$JAM9Plf&P2F6zuHVC2@;rQ6pE2Bvykf*MBI!U?ZTNvRl~+&RB#Vrp)~gv5#Oq4 z(6Pt*dvJvm!AQB;JP;bk)P>*myGf`B3Lx=J!ETLq(}Yy|3BP{FX5oi57%DD9q4>p! z0{G*Ge6WISA{PS;yCcZ<78N~xtCOKjDz>HeV;c}|{NSrp+1nM?*Wxc%?NL>8WVN2_ ztdFUhKOrl=T4`M~$5qX-RY}LA%PD?<%1_uMN;|1ZmTn$C9&rn!Q;&TkL&sgTl^Yce z|62NB-9DO>3QyMJldATVs`<)fJ5H;b7hba+XH?DUc{|Pw?XbyG|Fp-7vD80H>Q`JM zKa7UqZdpzDW3(1^g{ezy;rUuhvMDTvDv){uIzF)QdsI=k<~p@e4UNq;IJpBqj^ql; zX?U*GeW&WKshY62->D8?j;_9%US2hqRn5I*H^VylurVvFA^BQl&QoCO6|AA=+ch&i z9z_wPDFYX$>c-T0o~{hn@0q&#GYUh0f zG@;5yDVz#bFRr=OHNG_cH+6HdCKUg7QD+#IXXBm}V-Q99S!1~8yyj(VR z#nEy`pUNnd<({jVXTyuu)y*R{yU-k8H%HaY{$@9K9-emYn6f{-Y+yGqCDWPH>t=o3 z>@6Q5*wQNQ=d|*hRkISC5zyEX>HoZX?<#yObe7+O+s#-7o9uqjmGky7^7LjJl6r{I>3%tefB0OTVeJl>Xe9+nN@z z{<>a%s&4*NuW`ide!fCJT5&@;H${;QIIG+$B;8hQ-xHoy$PH9yd+a4Ux{vxd5V14QU9D8~ zue0NJf-FQR8euN#al`IjDcSz=&b>57*{_~~JS+$Tj4T9*^Z=edu=+Un>7as~9yNf1 zmLme+iVK4)-1JPmF@9zkk4bUQ6fbhzWJ@vOi4D?qObp3~(1MD&8EUv@%CBj(Aqu9a z)B>}pTy1j2nUrQIgX{hH&&Wy^k^6oM)6+j1X=ZjDj8EVA%u)??F@ZtkOr4lfs_HN7 zPAKLu=@zG>2U=<0y8`!P-{}&bHx?dkB48sZYWesz9!ynf`SnmzSSr_sR#vwKdFxLGvyLhtX+D8h74oJj%P@VUt zeJWfX03t`orS^agScI0ga3TOkNaM0ZBg!JnFEGQ}>dPr)Eya-41tGX$T6CD+^gU?{er_i90>+oiW- zImB@xcg)|8<$QCt&W^3ZAp_hYwnOw>zni@3@0|*#u)wkoh;W+-zm@gI=6fbf^ebbY zFgD+#zjRR3&}7t;vK_jP&;WIeg0^tvz`F%oPV^6BUNSb{>n(>XUdZNq;*NPPA&WiE zhPZ{Jh)rb5)5!O-lbvU??(gPe84HD-{M+QdA5_mYa${FOb+{EWyu!V)$#r7MR&=}t zFLTIDI?6O4Kt2DRS6DyUQF37abED%+EWU{*>_sQyE3LUI)I__A+6ffcQG=;V7KX<{ zc=D#U^C8+EryZ`q=5roR`o@paqzRhTWOHnrQC|0u&AV6oy1z1JO9wwH?^2cu7-xLY z9EKk!A+S!$FjM>mrg>|^9#wTh6dBPG$w|(f>`d5W!Bf1%23a0m*xO>RWe>eh1PBkb4fIT}wXtDZZSPj3^X_gCSae zSS<`($un)QcCXjjp0g?@s z_DU~KsIC4Z2LW?)vU1<9yq|?&fB^MK@gDL>+f(B*mvO-su%tq*m_yp#vA{MI#$*OT z5T11}*a+*WAhbRSPMB3QGPe@L5oW$H6aL-y(F!ng3kIGm?JhW9ejDFYT=T$=^k7=| z8v}(Q$t}znCQd2jls40fY9lWU_Ig%ywS79U;q0{TTSJ18?G6M77sD(lnqe$M3E@Xm zKqEVv@*a`eu<&kQ_?TSmcn+8Xll;vrJp3(*1j@?OOU-ooGdn`m{!`gKf@=b1PzF!$ z(#<)mgM=lE5eY)^h$eM_pr{mP*^Z>RW#EZW!~mMUS+)~-OGR&tcjnij{PyESqgPb(6YB??MufgHy{|^ZGRkBZw6?8xIad@Hdpb3VekA8z9R>cT7Uu9Ru2 zh7+JR-Ol7ST#w0wTq4Qiu$*Y+A?V2~fb;;~i7vsKfcw2PI92qx#gD%WiB%>!+puf{ zjIWq@IUNdP^Ga|qF}g5_Dr8k{6v}B~E0HjZ%MXA)Xa;jQ30jwGp}b3BFW4PuAkJ5a zJv>Zme?OVqWR{jpCRku)a`8DZGF0TD!ee=M9xuc5`P5)k$h^(sn74~qKb}rNGQ{37 z`WEZ+%6+0N80F$fI<22*9Q5y_X-6VEao9?-oZ#%$FCD}{24r>bnN z!tZ#Jc7*4^?i3=E9u=bP=0uxQ}iGl83sX^G(l4@6MOR1I#k zVlDLbnDB2_TevJ4M)gk|7_nBO;RNMCB?9bm5Mqi{EJjeAj$~?5hz>Ufe9YZC7w6;B zLs!W}$IUz9tvPO+#VwM6KyrJK$JbNx++Au0zEm#Mu`ojddkbqp%V-~Sf)qNhgC6GN z%|-6-;$I0DVEBs3CwPVZ_YDdjuwnn8*#JXQ|BWr&0Ql2-w3m21ke|PLd=COu)ux{o zk>cDZLZpmwZ>9NKvWQ;5m8pSV09u8_4193vh7WCF=O>bkCXhXwOH8K#CQx`gvzVms z7<6y|Cx8IbhStrG`iU-Eq?G6a8XENgBp6(dF)%SXw&WOu;Mj-Q1Ew^`Me?32AWcf| z90{4`;ePpSL!4*gv_k|hTG;_;@EeoZFT^Oz_f*^HU68;(3bgCZr`L&hWpT?2^_`K1 zO4{H^VN_7E(g@H_BQ9w zCHw@+1OICmndyczcab=hcU4g`-|NgJ&fM5X~vp8uREwhp3m}bXzadz zdXDsxWWH1=?92>N0IH%GM7r=EPS!@BS`(SIQEd$YOZ5{?;Y?5b7))YV=t}B|E7chk zF-eFpn=}AfQ5*m-a&zRt;FT3bf&{Nc{8Y$8q*4U5ZZnSLx=5CZg(x7oPpXAnNa9};FO1u`-8P|Dyt!8MF*`Xk=GUjsihtRhDW?*UwWSUtl zYfZwl%wDDZ^WbU`o}2Wr^1jq~ke)Pj`UHwhhPPePj-W_l?bl!cVcbM$r)&3KdF za#qWD_vp3(T0rqB*IOtsY79W;C z5au)!|IiatGI*=ExAC_*o0f(1>8;+}LQ->PNWK`eXjN|F5@8>a0N0+YLgJZA$h=pjC#XBq#>C0txT)%z2XT2^Vl4^V9()Oec4BovYO-WUTTK z+Gp_Gg>~c&z(8rF@P{Y75xyLC2z7VRTh+SWo$_u{8eqn82>h&s$cXi3&p=Z)4f{dW zCwPgxTtxh65T}C;k({K3suwH*QW$_W zKzItFM_GV#k_ezuIS{){1f1lUc9_H5L-S4&gTxG)M>hT06t9{#IfGQc1AqOa^2p8w zYZQaKJ04r27v6oQ@V=OcFM_p1mw3CV8d@#7fT+QGPHU#I0&J3dY5??OV)M7yeu5^Z z<_rE)?8TUXYM`y;{BT@Eqk+F$pJ`>D+){YlR9Ft;zPLXk7mD93j5~Cl4%Sq92F5^> z-`kz+UUpDu0juIDEm_zdAmvS+;NAUkGb*wW=bpQ7Ss`YY3?)VKdgS4dagn&xuzIPy zA?3^!?w;3~p~1{*l1ydFYZjQgema|blFI8q>aPI9BCdh+Va&OU*Sh@x;7Ze){1~6? z564!H!~3FdfT$o5EU9ud8;nu@y2&r5;ahIXez$Nuf6AFYg?0Ec?u@SE#(O*QxbGCln|PGJXDl79 z!RnKF4U%n+HFmkYwcqC6fy)gTg$ncangU2GJNa!HRAHVB6#P;p7!1s>G6v&*OS6|L zS2RIzm|@6sfmX@6p+CaUaCg|VJj$_46ILr_elL?uI8V3|?gAV@Mk(_E{vJczV0S5` zUPM?2+kQ8*6EmCBDZX-1Tny$ECkZIhA0xc7jhdHVm+T)bk6r#kA*|11@J z1Mk{8LI~?ps5!!PFPo=$Mf3->!1W0Ixbrg%YS6oRJtTqN11z;uc~aX<(jjIWhv+Ne z1&Ve8*^lq?22@o{;sggSc(0xf;=NqhccvWAo;)rW&S%*KG10rbM>6w!X+996VFg?? zpUDv7nzd^2?K`p}xXaKu^NUQO%;uRaeVQk@HmeJ*PvpSr#ZhL7*4vc4MiZXmyR810 zvl|AjuX40O04CN*B#XlSO<_-tSC9^ST_~UUZ@eJ)x1@9dz?`(lC$MZ4kf%SC)t~u!j^1 zJ3uh@rfKAuFbD0yHiWS4zitXAMr1}{E#}j5v{aaqQeY2}7c8V7$78m9#RtT-KB{t*R9Cdfl2@amKBS1U^= zn-wLqvXrbSVXOX@Dc}J=>?)bc=1wqV)cr&!&#|Ol4Hr5Zd1w~0o;w>Fj*Ii)I&0Qj zANH5$j?UB8N!d@j2u4ooC?+G*Tl$HNQS)|Nnnzr3;pZ_*NBNc_3$e=JnT!16q5-fN zUe33p*7uh=niXcvHn(TzHg(O!9_}HLzvA8@yqZTJyDWGob?!7mrn~Gt@MFUYvPKO zGI!AD#^%2RKtC7X6`MQbGSL)${s&1+vgaQsE{iV0V=IYPCd7Y_vuIB9p9KqM`fY4O z35IqLNG1!2HmN_fx}%w;PpKJo4F-P(C;^}$>g6KgkU2QzpN40?jh4D;8!Q^*YS^EZ ztB(gZpoH@~2b^nzv&sy}aD|2|v|kq1+?=Mkm~!DSMa*rwAuVl_Zc2pHFvw0t2u6;x ztlP|C{=PGP89ebFsl7dI|5RDxp#-c*e@-VsQF}s(%YK{Mzv${?srg-+{}tZ)(3H0% zKTqw0DST-Bc+ovp1YS=T?eB}FM~Z&4ph9x?@xb#)*H9vd2nDg-e7ML)v0pl*6b-ln zKyH8!*?A*E(q%1chTsxO6XJOIg|X$F!W2lp+l3;09kjh%xEaIu{{sNdDoz2w*NgUR z#rCS=MgV-jxDf!~W@h7GVsF;fD~sm4#r&@TfQzTRCAp|*FDwcGSC!nA=n$goOZK`_ z>C%z_AdSOHh$8!9;U+35px`>q zpty0HQr)t>aTQ+}QCaIkD&Zw(=&Ro`DOT9C45<%-PrD&2FpgZy4Qe==<`+3t`%wcw zOgrF@7s5Slt8(C}pACp!FX4zWf(cSkI7kU=jN*ldm9-9`0QMm+mnve9CYwAo*;tyN zDVO)-d3#m)XK%bew%^#=eW2&Penj5(X1{cQJRPbuimK9G;JhUV&L-&oIP>b35Zy{k zkbqc)a^Y)^FwG)&t;COrFp?Ejdty!ir~p{CXX2}2S5&=zNxnyaf`LSPV69f0&`+F}@&Tcr zW!28^vg^C^b)>#c{4}xZ|J+N#TfJoM56C|=?`2k4S`Vq&p03)bg5y6k;n}MFyL!2y z>h-IrY^yXS3OMcJm$REn4tizRV2lh|9}U^3v?Ya=b@Xp)KCUZeCDVirBK!zDM%IJR z)i4nyk)NR1LY}vldCz{7H}O2Rk5O62<=<}2pdI+9lTK~QQI?86cU4y6h& zP?^bonk^PXVUh5?koD2RQ4LiGVM9L++i&;G)obl+dg768T8D{-M?!X5Bq9s10xr9uk)Oer(dmYLk=^N0wjK z;zw)tm$mleHS?>D7fv(P-__#ZPx(9kUCsWsCN0D>bq9~1MK9Lf3-!`p>X7ur&k9C7 zXU`R|!zhdpygh;#B&)_~&_Ik2+TmC?|j;jK!1IkDQ>IRWrHtk0ffysq9|#jTb!0x=`FH&(cHIo4R5liNG@kDpW>zr2;@HtvK$uBE|KX6d7*qg zFy_jov)=3y028f4a(NL}mw2jBCsy(r-Dd%0g;-x)V2K-U$4LW@chnK7W26`t1{PS8@0_a8+UPXkv+PB;EIiKT+Migz%SiLV06IedJr*>TOwB57TU zsc#6%0s)RRXlJ4EglfStp!^Pr!$I~6IF1tz;jYr^4nipC8@gEFBsQbVH_QPxMA*g5 zfG&&zA3>1}g8C?J1r1zjcU@u4N?Ve<)Ey2_C!)O}Y;bUNEEHq}zc{5-K*d>Ix2LxO z4PZRV__esPsdb`=c}mgZ_}>Rq*hB~rYiG49!dM+adMop5>wd@Efd6kvK}6DEYo7#a z^ck6gZjsL$DS{+u`u9_KE>ijxaoe5D0Jl&}78ScgSEdC!0o^4%1}bVz9SKm(D9NOP zA+EWA&bbL1wPYFwOO0KPXDC2a}P;)0l#sTgC*x62GJu=Vn+f+ z!uy)MTTEk^9SMk|%<>*{;Y)CL=GV^sj?c-`q2`l}nF$Hs_BUt$?oxRw^HKhzhM{kD z7uVc&IXegGrYJ2D3WOmuK(G-Z7%!m%a-LNBlHewj(Gf?1ydNQodPlAa_--;y7-g8C z*DB_KJC>U89Ml;vf*kvllx(A!bh?P8=7((aaJu?N;RX~R;1Vqi0i4QQMi0V(CSPh4 zwoxEt*O?-iQn<3#^ac#}Vc!mWy9N^WnOrG|$rfv1gp(CyR!JtI(v#}IE*V@5B?6y% zS$CxlFKTY{y8h z*R2#c?5HN&=)!qVS_K^&VBxV^SFpTrM!HfdSwXP`I=fOS&RQ+m0ojT@T!*DZx07=#asWb|#RTZl=xJ63 zt^|b>9Dw+faF+STF;Bt-@r2HDGw0x9L{b2T{=(}sEWr6&de`fXR=8QFO^H;W_ zizrKtn--~en(=ZpDgF?Ep%5Ps2}=FvX8asc&n!--_n0%3@(_(9$~%PU=IK42GZi}* zW+XosU=%ltlr26b6fd6KYnpnG!h5!#L&k1{8?b5>^kbicO-in9#@FXx`9;e-(&GK& zjFY0RqvkJKo-++Q=JRtzftz2p%%d&#_BZ(6Fc0Vsn#H1@?%OK#;5G{r2ELtz<)>nP znY{D-n@E(gy?Fy(m|QMmbDH7D+&ADErQ?)VafHPeV^uPM&+7ptC_DHzfCaj|5wc8n zGn_9MtQpal>V>7a6UW%;G4ji2ONg?74?>zVI}VlNob=IK7Pf(^;CKRMo)wmJQRQ-> zy3y;T1|Y&J7`#rpVihYSor7{{Q639H8$q~I62I|6wTL@&f1%pqr^+_PrDG&-Bp=xQ z6IAP2uE?vS6T~K>smzihMEx$9QJ{s9wM({TX6LPV z?eJ21xYF#lLN=eR!rp~_UD>djFr-J&O}DeAp@;*0qle7Ht~JB&GLZ|s&IAKPuAEio+{>zy6YP1eMkS>wRr22-Mqjc4(V(kf6Tanak?M1kh z-e0GPZF+4SaGcffU;N6|1iWBP4=bicTb1%jS6I#HV$#w=Vz$otk!k_Qt1?UlPz5qC zDwTTZB;Ub3zo85EUmnL(lh?vVV74x-QE(+nM2>t46U zQD&`Y4TEVZ&!1)fLY5PQUi z#4Yp)B0*eK$r$Pg+4O>iKUXnq{{GM-&1$0qzDeSCQX?|R1V+GT{SaJA)~ zIhxKAng;uwN`?Q3Nv0)Vx~%|;@Fww*0W?xJFqYT@C!H?0%4MN3!sA5p0-oSZPnQnb z4B9xKI!J#Pe+BYzC+2C}6G`=>|D;Wlg>h@-Vkjfv(LyK2jnW}3Z~>KsQ+a=ALLN)3 zril3rSwp&X>p*(O?4859Juk4DeiZA{wgFpn-Bd2&&ki`7%lt)Q?j2FOIROL<1rojm zJ%xpzm~uRTzHg1SH=z~h@-5ch+TBw>8J`-rb*#B5B-=OVt7~xQ*yMpfCZ9Y;!`OF8 z*rg``kGGgjY1k4CnPrRdGbO8;a=^l9K-c9o=EaVAp+l0mN(c*9oQ_8gG8jTj#m&6g z?BU;R&WrWX**$hu!<^kv3bKbI^H6xxSv~eBg`a42Pi;s&Y<@9$fi}XIfu}NOB ze=_D)*kt{ySihp@A&IF z@T%8z7?>+M{QB>7;DJsh1*@!D2l6XHQ~5Yw0zioV1nI>^3APtAQrz8b3)=5D-{q6fyKr^oH{HHtV@8e3X$=|8jYwWD})gBG^lJq1cp5zf)Xo%vIA)&QDP&2 z&H~Enl9fSLSc=9#Poq_0u|)Rj9mgd-B?lW{BjZ2_@Irkq6}oGqG@Le92t2|PG)fW$ z@(>u}%>20RxbarUjFUhC)OMp(_`*}#CTmN(uPK?erB8aIebIGC64X znZu3a=x}3~QQ#%8*$28$fLJpa8XhGU+?M}j$l7H-WuR@38@FmCI)22SG7^7f#JSxnkVQduEi|d@Fv(p6O>Q=?!QQCFdChi zM#mSU@o1Lc+>8;*{h|1S?)IX&qi8O#n@i*dK?a+A^*yIzz8AS4MP`4ue8k;1Vjdpx z4~IFMq%u>oTekv&FC1D z*p?Mrt+JZ+xkcu^1R`Y78KwA~tm~{I;xS_iiP}M&Zz$U~Nwi2`#2Fl?*#|jIHFGzT zwCl8*3`6z4PZoc`LyRwEd#G!dLK>vwuTFheJ1A@zjUl zel)L5P8V6a+*Z++W*fIP)rdnrmqGHmD342XhsO5QgNVstDWPZEAoLW&zO({+#B*iI z9+G($_)^A%$T{mVpkA}1P$a~s8Uv8xFxn%BTUEt*oyXxWS)c8`E;H-1w}k!PVZSi! zi(&sw(t2_Ea1(y6B-Iyx9g+WhFF%=vbIo-bW?9U7fnM*?T^)Am((vuOe{aXiFu77_ zg;2;AonW4%92InF7oyWYR; zt5zUI80h?VY%S+R`oYLlm#vgLH{oaNXFWEB1_vdAw}MjWvO}`#9N9hq6Ow>_xm|5A#NY5ODFc6r+!DK{`TkZ*7mA)bVb`<)ArwK zkL*ntE=k#5Xq&@^{h`C7htq)C_pP~$piOgm#ax2FDBn$c9&h_!w#{$a{@3lXvvd>r z_I_VBPn6AR71EiZHA5`R&o!QS`U@%!zFWx zI(kaUbkLtX{K|#~PC$))JWGGr_BXcet!;lx+XB5OORe9R?2{!W0<`tjrRY8I5>F2K z-wv6lhWwv~I`>y36n?F2mXO^H6FDd*1jFU*xOPzeS^T@rH-`MzhRmuVfA)|&sq9WJ zkDOGtCzs7*K(6|mj1pY=V=TbW4Cyg&>#I$3>5#v8$gCdnR}8sB%kHr9$f0F>SlQg8 z#~#S=#wV-GO|#tzL*}$0f9epv6nCI)4yJmaJk~JFhwQOK{^TKl(vUs7>{gXW&Mw)Tp|~9{1|Mu{ppbZ z@etSjeM94S5Bb}N3}=2aWPh2tN3+&1GyAA=*8!^YNwH{TaDu7S$G%OHqUeSpf8CJ1 zX~^F=WS`F5Gg<5D%s!Ku@9T->o5Do>lVVN7f@1CViDawjMq7N?M7u>}_RA)Ieb*;L z6DPOWiPWNT3`=F6e{B=aRJtC<`MUcN&Jg7q7EDns^4*42IJUDm$CdC7BF@kmhUGX$ zW4G}M+^$)>3Eg9##{!URgvBh+>`zE8=T$p4w#qp%`8G0~ZagL*3HR^InOZkP$>J%; zvo-Vhu3A5No$!^=mi$TOHko|y%Q@yId$a?pXUx3)JWMTJ0^v%OXJ15uli{R?^lrFW z5L?h&?`fnx?bbw-q9nb&^-3)rX!qBKGCMs>su?bIStA{5kJdW$*wWLj2B7q&O8)p5$0*&44`+EEen=kS*GdQ=&Xim&5% z+dActXGiH0j<>JXtCg9hY>U>W+=xcX3mUs*JEyz0-!QvA-p|OgCfS|s zWvHi=z!oZuKA_C)0&;ycx5`Wq_&xEMr)>>VP>~x;nEJ@;O6WQ_yu?dvct8g zS#K7$HjOr)*j~eC;Z2?IkK#ce9C^(e5Uc~`sZ0gKWo0|zo5-g;d10Z?SGj6YW$z{ zL$I>{ZT`pqcV#K-Y57nidE{{8ml=P!A&QUU&m%_|CL0k!jU!F*7-Nta3Ba)SSmTZ} zgou=nGcFwGXZ*3o9B0^%FE{>pW5V8^Xxs^=b)vB+7;~~|on(4XF^#X73=epZL)hBoks zq{+)Fb~P4!du7F5MJi7139J3C-gvDcoxHtF>`grs4T}5WDdApd1MBg$s3yT9Zo$6u zLi=XgQiP)oDs@~?FqgE@@z>*C`5*C7hEPUfy2cv?(o#@**yx>si(M4>*J{|$3VR>+ z%@rmtT;HG+MP&X_u}c*vad_1(tLA$+Kv@G-*eFHNvVK2IL1;Gd!0(|-jfnoCRrgZG zzEG(yt46b;;V6q*=<+8c(GLtTl;bx#8M09lF;~OS9(NzvGG1O;82yZjIoSOrMi#SEf)O`3V`M@c)1~;MCT(S;aY`(eh2+T7J zA35nM>Ilihf6A6sOfPRzAAOmX1CLs^I4WUY%+YTbzID(5<_pz;rY(i0oL}b^z5z!l zHH5iMgYzSTvZ1NI0jG<7FK>(%E*2873WaWpo>Haqg3zKD&;`3_L)XEiR%rGR5+)3U z_^=gA(+fA@TvWV%6QKD|i+I;McG)ImC(VO9l(9R+x-NjM%NfbgW z4LSn~rYFYihIUJ}aMzk41xziA@X(RzP$@_Rtzl*i$kV_}&F2PRV`r(AEm|l`d=rP9 z(7K_*DrMigW4imnboS=<>9Xyk$&~aXg{BrRT06wkD-A(?Si1AsJQ0qOOw&U1z+K$V z5tT~%3YVigE)&-e62RGv(Y=%$2j*h$F7?qlE;*kp_I@0imp60{BSEkY+(rqnNgh%Z z8{S@|ol>FKSb6I~QPkI8C=?E)EfHGvKAEKOUeG8K@uebG>huVmhB_)MzUQ39oo@3L z)~vMKg#AF+`+u>|-B9wY`SH|!ety$`@oaeoz-~6Oq`W!Xr6wzvi-qS&uU@*ln*3Pl z=f9CQ&%&`s-tg(o)#Up+cXHZ12}@NmCH!;p-#}cWXJvw;Yti%Y5q(d4*As0ERlBuX zez4l56uFODuGE39K$Xz8uve;IoSg!Dh{jT(tQfRyyc2+`DupoeE1&_V05c)VfVoZR zxc5WRY`bkd7k@<|4GQ}V=U36hENv$Hlok}50Gncu!l2@TtwM9G@Br30IYM^5SA`_# z`5}R#i~So!7hf0lY1r?siZabCI5I{G>nV}Q%}nq+TM6G8Si*Bl1BLZk7OxB@ty_BR z>PcJI8o8qKHMhDRU06qmLGK66^!7aA#)Bqa3Z0Q+N2to%q}>~Y#ckR@cB0>R`p&lH zRW^R6)m*&uB!aHvGECu9tUM53yGAF2gy;V=2r2c~iGI-OyTz7owecMt ze|v|ZdUrDrY}C$fbOE;|FB9ODY9t5R&?Nbp&s>g)N0o%oLUSjuwM38vw7UQrN@mD?{DyVJ;JgCA zC|hY-UC^m8K!0I1Dn$F`VEufyWwCwcn%qC&>Kt%ueJ#2lXF-Ng2&Couo;bdja+#Rh zFX}|+b^3m7%fGSlw>tjf4z}9|YUci0jzYl~Dn}vt(7g=Z#p=pTnZzbF2_e$zkEiAKZn7Y8<^bPLv5ZXpy3<59KdW zw{TS`ZGp*3`^6O&l9Vr0;4s}#$eWDnVjY;_;=#gVXLgc2RC>ML$-Ul$t=K(|Uq83- zR8o94cF)J=uW@{wGLRnMvdde^J=D!~GF@7GeA(PcoJcJ?zGeO%n;*nxnX@-JyT6a6GRAP=yqEN@lOo8BoWOf$#{*jrxtL7(G;M>BUOHMEv33`}&s=A=DyT59FE>Dh! zs^$^>uB)0y`8AIz`uZjGi=}^%n#Vm_t^J|eUG_f$uvqqoK`SV9!dWcq_Om&I-|Yu) zfHrY@3apUol(s_4mcXrI(I`fl`xtCvaU&lP4-v>KD=#PTX z9|WNXIE=*a#sQswpG={|q5{PEUan;%zq*&aP|1G$e0ejvj9RWKZ)J zv!~XwNpxC7YWX>PHQzeDSTD5-=cY_Jb6O94pS2j&&Q}P3WX@J83Ae5%S(}*q6Y~Hy zm@4-8idmCHHznrA9DoT0c+SntEtx8c+z9xx_%=d>2?rA!ur5g+O`M2SO0gbba-Lvv zXxh9rso&ztx4QI&*qqx_yQD&vP6=PIR7hVop4t6{w+w&p4SFIH0YJIPFeexJ{SnUU z_*qShc(w>(ui1}bxtj=6RFe>{UMvWuG^Hb;OH3W651cT{BeVy>uRG4hDE8RP@|HH6 zR5U2Jz*-_q4)TmL&I=lb8!RRh+tIcjX`E59;keof&TUT2RV z;~V~@F^t$N(s*S`0?c#LNEWb*W^?(-G5)ld&-TZTHNBfwJKJX``tI6+6*G(gd8fk? zI^f$d`*@ub3)vHOlpzL|Cc#=*R$byi9<9iulx=Vnn(D}RQV+}P+9lO8+J2%KWLxX4 zH*5Iabng-I%uP8}64uvyBFTjo(N3u~d#4X&g9@ZvX*OFS=_d*M+KNCbi@gB7Sn3Z% zaCIqtrsAIMd#x4r*@}5Y=$$iVe?X;yW zhWs07Fmc?}H-HrGEb!BcClR0VAA!SIw9XC{nrBLi)u#GPvkH$8pAbJVuj5!N=kO>s5x`A`gF_8fk!2uE^`Rt)fe|%mc>`e zB9dI6p-wzDOPA%T@o-8K*2%qwJ`+Ghn8XabY$Q5#Wa8MQbjl2WQg!-~vfNZ%9 zL#36vi1!A{lVp2%WKyYMAx*O-h*=YL5&%_*L~V?Um}1K(EnuF!HbYldEc`4R&gX$t zPm0_;Xp5i>$qLQg3#XOQeIX%b)g*<;%u|{9aSN^xQvF!s zA4mSNKgi-+iFU$+nu$joJER{HLLx%s}{RL*7H|u>G_WP_Oq;|@r5ctV;+?^PoeFgGA zP%dBin%(?yX0Ca;o0IEz!uowh7X6UudwWaO+)?$nqrrfm1-bid8~TFdqmUOLEqC&? z=x@WLUrEcSr0LRWe#0=;t6KNrk`R`1p0Xcj+m3DgwGi{&ASKDzS;FxqHX9M|2>G!u zc{Sq6!4!$Ki+rC&vL6+TNX$%*UN4r1g#u<rAxKVrUIX}iJ&mGs1Fa(valklAC1QM1Qa&B;}FWEDLoD2DNkRYq)Aa3@2X z;#xtPS7K*F8p04S(j-6NM)l!awM?in%6%e*&31*g73NItr_UJ@BB}Y%8L6>xc%bwY z8Y8%wF9kLx4f}mokoyTiUFiX(=Ro&NgH=>#uPOLL#p6nLPWiQ_$$cergW|Sx;(2Gu z++Fgwl#*LZ?nd(X6_d+w7ev?wXawyEW2S*IXADP2;{s4DFDs@e4*L^^{jIXscbUJp z(0b}Gy$V~H{M$jsCRjdRfiD?Pl^|5_4(KvdRO{tgL^zI?L{w*cy$V%u)vXK{p5F*E zA;ZP*55`_@jBSj`0ox-}S-MqD#@M=5loV5F+W6`MDrnI%$2A_w;m2I1a z3#u4pkflg8cvdA^RiO@^@;{hm74r`?`+ix)ev|oxE6^&WDO{o2C?zyz3m_Fr8Y7z) zO2HvQqPdc=Tu6Tls*Oqz#Qq6+;Efd(x=&X6?N`92HTNz2h{k(-VTDb&s%UYL3$2_R z$-E!<4v>PgLo6Mmz=oGM{7h~Pd9CFPyE`}>)Yt@UHrFU@vl8~x!+v(y_pKo&CMea# z>5A-0!Y_g}LtTu|xnFU>7VcQHK*vI?oGaG(KAvqAep5}a4Zaj>%J#a7xxQjAO6@sm z^6epe?GT)wb*q$sn2BL7Ls|x8B&GL1Q?@v|Vi8ClPTWI@d4$E+txL?K`h6^cp|-zH z;L`5x)@HJ}i#$MPYWHBcY`-?kLd%$jQcbS=t5Qw2 zLO4Hii!_k)H%LjlP1qNeKB@U%g}=1mH0UQewOJ!ZG1>$?arRR=e)N)%D8==Gv*JtB z=%bq8cHb3JvzJz&gPXf$y@+B)TU47_n2RH>cr7u1A{;C+V29!$9hNmB=0*76&IzzI zD^X%6Dc`F#^J;F@m%kDk+G` zEe8ghcqVTpSUg{79p*fmz)W4>$KhzlECoMoRr+MYM(iI9-(By1Z`V1q-YpLMfpt*7 z`470l!71Y=prpRB4R9fL1s<6sMzE7WDuV;;8Iq+L0L?V>jOk0mFEpPL06I_lJ+J7n z?c8Ga^Jp`yATyQt1*)({Kn&<1m$n7p9qT0AU+^6}EM$OqlnmZQwNJ0=JTx)x=6K0iKmx$ho3SRqS0DO^x1e581Yyso2|l2QnDlOZ(zYdL_dV_a;%GQxU1JGgOaa+TgS4q;jS0PXo@i4t{S@z@E$R*&eqr&XD#++l^!MugI^8A6X(=MuWnsZH+ zBEyQ!WxXfv&0Hw;L1@+2bT3C=H5Fpf$@?oU3>_pwH5`}C9yo@brTfFjx%((%jx^Pa zXapWAdm!W=*?rq0_W>>+Y~Ek_gqcrI7nLUZwHV}t`XZzRaqxYp(>g>QjM#(mJYTG2 zD5NWypb*e(JZ^g^LsLnMH6ARoa($bn#zq1k7%SXCOVuCvk;xPg;^`$*$v=Tl&EZpSf~FR}cg@U!SDQZdQnzTyNgLPMTe(K)!)57M|c;N*3g|q8ukY-_nPl&T)q|m+vt)OLU}~Uq#d98c*b0rzkMljb%V>7ve?7mQGN%tf%P)8F zG$LR40&JlQ0<)69m^3)jFZ7N5G_i~&n`B1W-E?LSNBYHef&IAl$tG}&){|P2%!nrf zpaqvje&?N;r1^QoN*U2*)3+=rz#BmJgnfpeB>}0NBSt zw4mt+yg8@=W3e#sp~7iCKHL7K~3KAobl)C zjX9c-8|xd;c(2zpE%UZgJWQ5l9{py&9X?_rVo|GWnR!aaGi zC9so;D#QWgEvmo@arVb=t7`WP`&sNOpi&ryIrdi;mrcfDIx!}L7$JSS72b^lcs6cD zZQL!1v(&ikE@!fVeG4Z?@di^o%$nui9PQ09KAxNrM?!)?Cc#D|A6$lK!NHZWc)3YP zuhY6%AjAuml2}%R2V!cJ>_ym|O~Hj68}o`k&?{Q;slvIkS_>C*jP&+Nm%M;44a9fB z@8y@(-)+gO=UV>vE88XA4ei=8v}-(#r0^+Xiy*;OdI}wWy89Y|4m(hg{i+p zGk=F}vlI)m-n2ur!gPIv29@>E8(aeCb(21NEmj0Au^H!lgWsk8Pm2PuS#X)SR4J@EElFRpW|X^M3{3#&MlCIj-BiJgUMnYOIM? z77IV(v*ClFaTi=r)&MRc2yY`#FgcmXbXoeDgCmqs~o^B`4=dtcXlT#ePZ-k92d{^()TOT#auZ($uL0Pa4`kNN+ z4p;cvWWF)roB_UVsf<1Nik;$_=B(KNcz%JNM zXP=rpJ3k?-0%zyw>|w=j3x^U!LC*H;tR&Yhf^P8hzB^{g4&9O&h!v$JTUSGgjdu%_Hu8(siIh(B6XP3%Xe6NJC1}f$PmWB8cVvh z_mL9RpQ+XKRBuk>m*v9R)4dyvUX1;bNuBy7kkm0h(Zzx3+(2d<){}_(=KLtb%Ql4J z3~9^}2iynPg2%pyFcPai=m&5!l4E3%>WV&@0phJRo34<-ZgJS}x&n1+Cn-uFAD0fJ zBuib1$vh)83f#u&s>(PwvN7(QpWL@pIu$|=XZVQB|Hy~foiv+;| zdv1zYG)!x^FXFlnrafzz#S4peo!fEJTz!0>se6`2lX&rf~eqxLF@_mw<+bm@b z@!33rIxZ&D4GB4z{VsUk<>4-J2#ryA1RYod+rriT7cD1E@65qc29zyIQ@5lU8$d3b0Tic+hR*1cUr*Ag&LGO|zQRu_{Y30t6bcn! zkJ5*!R#W_u9AEIZ#zrNF>XMxaHgGUVPn53_kLprk6U8l-3)7KFpKJO*ldTN&N86#F zV6f#zxqG7(9?Dh2)xsu-hv0*yMO&bYBvXo*GFs`)&zt`KrhA}i?{6}(U}sP>^fQYD z3xq~V!-L2qs0=m9Ktjm(e5Fzn5?Z4!PUc0*b6ny1%j8lj*09F~ytIi~Xbx}-osz|D zI#KvxlP+FNbp&%M+0E>48}4@v^H{?^-VoH8x+?Y2jB7X?PkL8xPHRB#44YxeL8+u0 z2gAe6w$VgrL`5_Cd`hKJ0x)mC(f~syQ2GN!z}Iu&K%v>BHSU0+!gCp(ZTBIYxd$>w zVY$%bzFfCyj+Wo`+bQ^w%+~CyIeG<;hTceyG$Jl-C+uXh+ zm}4vLTYB4drf?TJLyAR``ZmE6nsp6$(-F)Ga8~{JXEcR?ho0%;TS&@!%wiwnp zlShWx9D2^!xK+wSC=eg7l>bs~uhbdG{&Zqr5o!r$ITh^`vB3HAo;fr{ifBF(Lj=_8 zi{=aVC}a5xss*It5W^BcswIar@=IY^=ct7i5`}0uM)}h{OAo@Cmw%Bf%y}*Qt(LpE zWiCZ?=q_uSb6Yr=olO~Kx2hGPe&+TqLbJ^$%*X8)YqRq{e33jz5Qp$(ydZctEz9Ek zB2C{u*)~a%4<#e`?9{N~D9sn;l&1Wnuq_(A8C^GdXYWHH$%=^Cc{AIAQheB6VCq2DYWxG37tkELh-xe6vv=ls}o58gG9_Ltax%R9>xY`vK;U*QT5 z*V7L6tGGy9J2oig2l)oZ3$WuX(pQA+gz!8dlGR@zPWb(b7CH2ksT+1z8uR^nw5(|k zZ>HBY-Mjp|Z3mr*5C}gix}+LU;V;}dY0cNfz z%ilhwH`=}ITYCUWY20-HPrn(i0jMeWcEH>Fn^DOBl0F$SLW|4n33Q5us#3MR-kpUl z+<#rzwL(YVX!Xa{ofBGxD9H%|rFm<7kb5i15F+b=CH{lcRmVo?v++4B`8hW)nPkfrCzy*xoA7FQaxjY;8Cxf8SWC9B zb7bxg;Fm#Q%=Gs3q1lGGgKGq7L9mN{*v(?uWpKjmWG%r7gVYu?2s@esUX)S#*y*@^ zvfPq0k1^Ycrn|N;n*;sF1G^>CEp>-%^lF}c_Vxw(xp7#%m) zif{+HJ4mjRfCXZX;HIg)R&(x{&K--{-pQsnC_TYB`oP9>JcUM(@H1R4S##H#2X~l*ilzTud*1QB0{_BqZ?^eng*R zAT$=(SqNq9BNJ|J>z0QqSuP9UFTRfu6EXzD;OQaMtu|R-T2J7-Szn66-iG}F>);G_ zc-$8LE3H@mWkq+5eX_gRjbCtcYEAbG@p)oxm{QYyyG_j)d?}w5+)Ep4x)WM2)|ePH zbX@N@p{6@dGNGnBj3v4%m)CSB97(Qla9GUfTb#MoS@x8Myv^}t<9c4>oVx`+!~D=u z|GiIZDx*gO4&*g(9RZ`TJJ}D)D8%poCGI`o<*Lg4?^$K<+0)DEB{@C45&{7N1PHw- zMX@145YdbG3RnH#Ya=2eC?z5yVo5{1qDUvMM~c9Z|#|L z<|F}wdOz=(GkbPjdzEKB>*Pao zdQi<#+#2;}x~~$mqL8xda^EM|6qaEt)1TqB*bnV?MwD3vU=SgO3CK7^5D8(=67@qf z$C&M9l}QG6V-}auR$zcl)ycR^%EA$l7xbUawYBUA0)$AiZf(qU*@n`_JLc9YHgTEifF5lx#cTIH~3l`Gzx_DS8B4c(r0-}rv$ zK2olXcOOohY4?HU8hPy-ibpy%Gt&jxO(3IPb4!oD zD|41l6wZz1JP^dfOyFLAv;sqcP{jjk``Z1Ad)s{qyH`nM#T{TtNuK8a%h65ap%pD+h}Y!4YgXsXnp|~tt|PM>nUDX->Sbz6Kv=Nx zkn3&O7z{%K(&V~nYa`VfN7x0ru4=fYZ_FsX}5c{(f5>1OK)}J2H0EHU7sjx&Kae}-)vPx)ynL&TBFF0lc z9-zAg{TB~GNl_tVUz_Cq@5G=bq$FRw%n1R8U#pde0H^`>T68qkfaC&O>Pm!$P%EMS4(A+Rlr0Vz_bbz0pbn&Y|sjf zOt!008TJ1#uNQEuLRxZ&WR+H^-~K!9SD{L=Jx}EJI>vpiR*@8Y<<@HS|7@^Y@o^{K zr-bFTad(6K8zyfYRyNdJzdOd$9Z?DK-Hb>$G7bQ6asjz08*LK+6zHaPZf!n~2y~Un zyvFWj7hq6BSy~V`sCoe&fyz$-oXm~&CW93Su~!lX&Fn%)sz20ZWeiQnT4UGVvaj^O zEqIcQkt|x1r?qh5#PVF?4xsUs_E4}xW44M_8T04Xe5D2vm`L`0PY$1jaDG} zU9^|t12;egEm8o}cBt;D=5j;W%5Jla-I;L@A?*X0IQxD~05F5SJvD=kf*Gt0+ru{y zQ){>N|3Wfu4q*dWFd*dqAtDY@TOi^fX-J*CB@u_?%NB$kWL}eqYau=EVqpn^gfRtR z`iNo!Bxcwvja`BLQUK6zwNPvq8z}t-B=W0eDnKQ_V$7MR!uIZ?_{kBB`{yS746k)e zw#xw{I5gR4YGEeS2ESOgHK?_jlp|VgCQKl$3zS+n;y`qQc ze&G>8$$0*&sV|~Y6SPuebSIj?+R;+o|4xUPcp%H6f5H&gW!(mG$J%miSX*`tTNk!J zF&jV*yR8!albuBW)am|QO6iOM%#x5-nC;-g@)l<$PCzm?PXEj&%mC1m`kDs%2S@lt zOo%}JNY~sK4?_LGr!=S^7K)Vm8Go0vBk`bUADv;1k?f^~@|nT7h(_+T2h#yWV$=vq z?8(s%<3}AS68wlkbTJ2hd`?Qa$QJZ8Q7&uT$5RKb>S_V;O`;cAU~6;6>4gfELf1IG zP=Q`RkUE)O&}xi$AuRrj{EFt{pZIn{xoGjv!4{{BzaR`cEiC@Xk>cJ-i~kzizuIDd zd8GVc#}|L8-#~=+l`pvXW6Zdf#b2JGWY&YK1>V96Oswhj2o5OeXsjrhH(7|yz@p&i zop#z=S@y%akuLjhnppPJc!9+%Bm0Gx2*aC0AlS!jx|^P zMmoTOyhX&z@;%NvUqX-w>X<1N;j4Z*dq-Jdm^JM?-8DZTzN7v5iAO;6O< zq-!?$({WcZ=AAFFY-0+$?aQ{x+WN{Z>^0V0xyNstSiHljmi2q6xqgq|%Mw1g#S-2a zcg#`(+t%TCX5GHR_yLyi+J+h}cD=zs4ksYGt_poY zoPzYiOfP$Zm1at+Hi|O@AW$TDgS6tj1(?b`)aQO!Ggr(u=Uev?Yc3-4Z6C=MIOFc> zaZmQ)R*doz9okfE8~MS+NJFZG?DD*0ABqH9Wxb?3eMvThD`WbSF1_S|KC=z4XsAC1 zAQ+G(6GvW3q_9YA@iqb{jO%Z_6euxr;?St@K5w_JQH?i_q;H&(zEQr3TegokDn+-+ zc+~BZ!BT@2I|prYfzU-ZUoY0O_1>CXwK~YgcfN6xvvv9RMcH!*HoRH3u8%~?IgvkG z%P3Q$;hZ&t34!tOT!JTAUCDWgu4lkO!D_(ofk#3X?WdhtviA-@blqGeb>lY^eiBRF%W==2v^N2I?eq#tRaYwevtta79qT$59bQR#gNA0%IH%3 z3c%jDABWC3a5Y1;9vMH>tTk`+Czv;w6YU$#@r*fZPWCvAFenUaL>H7D~6s?oq@x705hC zbT7+#vrkxVSahO{--&eQ1bOZ$XKsX>FUri8t`SV?X$@lXfjBxRHfP7#^JAoX&BQ!a zid4snxQ-Dn@PnZW>*fqb7DuDdFn18{y}Vt5p>pOtEPqg#D1fZY9rc19fL0@d_gy~N zFs!r31xZtWt>#~wdd0=10U|(GASf4QzLpe|?7f8vOn#37NvW^%l^Y2KU#@()Y_2KK zm>t^h>`9^JnL*#r3Wo^Ucc&Qu^6t&pV}=%j=Z(am6%Pw?HUH%Y0yn@m)c%|y3{II&S@>%XkH!QHjPe^-M|#`w;Ejh1r_RF*PSCF zYZhCAY=`MBB09~M3>Xft>Z$p`c%U$dMGAlT`Or~<0;W%YG624iz=V!eFeWTqsJ)l@ zux?PM+|TP7s`(~A!e~qZfOw4-&F=;J%1#UYJFu1}uFGW)GW#>c`?=3Mey`0m;EVk@d=bL$XLTut*B$hDDiMIL2Ey&LA zTrg#RHggkRo(5wkBzE)(*B{GD?~YePa(d4GHCukJVxFyJ!?yT%);yuhcoUf90yF2i ziha6jpGhoU<>P0I;a-&$`18W1JO{wgHoAGZRP1(=9p!Yfk%Uu$m?Sc&A>PfF%bwA)F=aoB1v7s zvBYnDtHkBPpM-IPL`4->k*x#YwvJc#65O$uS=HCsPZUUf_!D8LS+k?8#_vSn?w)e) z8E2ky{%<6gM)yV`*Qa7T&_QOt0BFhr&$3-ca93IEn-ciA}o|+tBLvXcqt~w8Cd+o z`TVN9IhRly;n^V0hTo;HkhOuhS&%DW_egQqlwfAQ3X^vx&dX`aGyBdk3I9vcSbSlg z*)#bM^Oksz?A-fJ=H5!-mc%?2C*P}>n=A3F%sKV=Q{<7G4v^E1at{0LB0_Gd2K!6Z zByv3+9j;SSdri(+RHHXodLtXv5(1`pXUc62fksC7V2jD{i7r7pLqZiu9d0O^pn50( zCD~}biD$~fM>=*C_5$22b<+sV*g*J+uT_(u=FA4Jr>JL<%#=-c=6h9A5z)c+hf4k(HYFu!)h${Kx zb?Ay*8)s|T)RV+i1>_lLl9>l2B`aDA(*1~@gLrU8kO(dg+f+uFZubpm+QW9|upJHC zL&J7j*mg9xteEJN{)og!+6hfm?6;(RAUQT+t8KPRX0IW z!Ht;L62-6P+Ih5IC@BqFg5_sC;tnF!wqS(i1l(xHh;=Y!AwGWH5XF@^ZkJLs&N1gT z_%ZABSkyUE&~wxEl$DI-WufsjoYEntn5Ht)%h_WC?QxE}1W`^n)r5^#r2TX_x!31n z+{e`A1tp*(OOoP3F|qa10*q8^=$|-)*J9DuItzWghy;ahdlWSHT1Ilkb`3Ms_~>%x zG^}t0cYt5e=du+6D-{qa6R{>DrXdQ&T-46QpB{2%&<#{B#I6qF#Xjz561U^wW+Uz( z87pSQfyVKf-`5Flgo74XiiQR)Dy@*$hQ=iGDDo{jT&8J_%ZRCU!RM!%SJ<}|GPldj z`%9vvzl?&WflM0i)e?(SHiW=B`*joMX%x1pof*q#duDJYBIWG1Sq5NkOUoR|g`d;{ zySZHeW^Gbe86|J~heXOWEQpkATZoBoR_j5L=~UP3nT|?@z<0O!S_Y*yevUJZGDkQYQ z9_FVuNOE(5*kM3t!uJ$(Rb?KP-n1Q$gMZtR6*gpcc{kLM*>#?R0Wd@?8D@QarRn$r zJL8y|{)FV!WOJF@K(zwAIb=2VgF)FCIB95|Re}t8G1&4DEa!RNe8am@e9jOD(H*%4 z86stm!kvbo7UZl04dg|lP?lSaHx=BchM7ksh{+OkfjPW?7y%41F&x%9KE&M}nTzQ_ zlXRAG+6TrIgu8T#Kh+d+;8k@CVk5{J-c9APpr;uW8RP&`_0z&3KgdJavWY6`z5ux(#(b{t&oH!o z--TKPW(#)qOU=IS2zw-HXpMQjnWdTGZmOE6%aVAQS4Xb`>z2&-4VXpQ$7{IFMrX%n zCt_)=;AW39-cI+fGLsH6@21)!rciC+e~4LDWaqu9gM|{>M<`@@ zcoIV);enQYdeq~~GJbDbUdjb3CX7~(*d7}0m@C>%a8|+~vj?ykvcqm~KkPc=z3gai z@BBz^pZsucptzGAHT&2Rvo{3aTkUW(vqEd+v3?;d>V7myPUC>IriOI83Rgm?lDF8G zqWgoUMN4Sy;!;=j&`MYR(AoeuW!tBc6L)c2SLb0JUEMG3>Las8Og2PEcEn+>oBo6Q zY(JhHlC_3Uw$zP0D~81k-u8GuY3@}dg*z=^@*?7cFtZfmSy$JYLB*KCno9)=#+f_d zpE|T5VX(EFAXJpkdUj*{nAs4Ujd2vV`3-S}e6k3_z?e}%@*KWBM|2#1G7HbL;0t3O zvG!5y!pt?+T}}J#6TxyH^l_#|iC|D7Jd71P7l6;v{1BcxpV2G;6{t4;VpwR|un~#` z_;er8`P0}aCN=iCxbl;1*<4Mn^$5PmhsS#t_6)=?V@L@SlkLQAkwrPOy97`W15V=^ zXh>)lQ3Cq$4TNzIdXhFX3qfMm9%6?plq7rk`(W2xkKD1qB+Oaobh$TcFrpJGnMd(dE68#tbbHrYNSKRYNNV8n5ZEQK zyQuh)HEKbZ$oZ3Un`+9?wK|KTUr{-_l(~)tQufiVDTzm*K_J2)2!#TWln5kW1wwAgi=E;ZAw}2C;Q1^Efb4YrYPGfm|Pgk|#F<4yrIyS|yrqXVZnq)VmpS za4;Sgr2kNfE6s9IHo;4$;fs{b+{h##5SXa7n1G%dpg_d1K*}$MT9PA|xSH5(cO|&} zL|7RG`Hi*vhCySc(i34AVQPciyDSPvYR^NFo2q|;_v#Qjj_G`Jgkrdm;id_qqzjl2 zn<7R#qMZl|tUxb&0AhOrDsgvD+|yWiAWvb>A?(-+UZ`<+^+4L(E_)!nRcJ7Bav+2i zdMWhBJiWfkYvoHZgKvc2?XLyVObzH=Md2ihO&leIW({IRYp$^7GFlFZUuxZFt+~W1 zTo0y7nohhPtuL=X*CZ^?Q9QrFR)L|Ym{ib`Gtin$X4K9Q(?tif2$mhIAgW9n2|^9R z8f0k%Hci+5P7O_@O@}sfv>DW9fi``outxvdkVykE`oqSN81qXERB|m85x}o_9N#br zAwdv!IX1uu(}kR(+;a!2H3QK~d+#&JpM$(o0Rsbt#S#~HB;+T|4K<)gI6O6lS2#6I zOYkoQPdY^zZT3k^8^$@vc$u`K1X%f(*<)B&ds7tEhy3L*v&dqG9_SoK1Q)Dm8o`FT zv2uU4hDEL}WxfgydxZejU*f z!?@gFhh<=KIoA$;LAtJ#!kEl*h7KihUoKs^eoBp!pTur{v0Iii`18$&KJT&xjK}DOtjc5J;pk7%vyt6dla^bXo{Q2uH(A1yO;y){?RD7BySg*?Nk>$ zhIeKpvK7TOWBq_;9A-ISEcHeRBWi?mbk0{AJ-#zN9TS~Pt5If-p9@(Kt7q}4rB$Ri zH)^F0GyENR(&#(k5e$9Y6w#HlAg zBEo<&_=`IJ+?YGnXw`2v^7(s=6=SevEuAa{iR)IbGIff;+!u_wk{FDQ(UfQSxcMt4g1-!?b<|7`(K*O zClM<~n02e~7|&VuOV*xYQA;>8(>_AhFo#9w04J1sN)1FAHISZLS}-i*b<7+fhq)5kb0Az{d=ZA}p~A(i;T2c1MAq=4q@Y zNpyv^4ztR?q*hHn9F)V>h3#VJUA=| z_GMf_tHTj9anR38=30reb^lIW$(d!z!S;~G8DsYM2PFI30~*Kw=}%5hvM1AdDur+E zBtODb9+8X(3{v$yIIkh#!}T|beqb!}%9{Wv6_3p~-To@Tz7IyN4~Q4m4#c6Gg4Auq z9q!ZhiXaSO>Wjn@mQGotus#OFg!GVfFk-7D#*3}eEuw}F&uEp>Gk_9TL}lx4wH|!5 zl{;}vJg{Ntb*Fw?;Tz!PNI7D?n- zPDoDOR@xnu%Xu2HD7An2&v)%S*3+$grUqn&fJU}%1^0!QKor%A0lB>a8z@*;tQc9wJb0BaB(|8Ge zb{J#nVCknyXpAJ-CGY~=aab;kQK?c*DlyozEz|2W-=o(oQs15pvO%gUI8HA)l6WX? zpPu=kOD?KSSfYH4d`_sf@yelxJKzR7LM)U3XUuPcR$C&v{?{f;MAxCP9o&SxrZ*oA zij-vvR!EwFNI&Xlp7+NKGoRK%f>{QD3b_Hz?Q#AD_vQ!yy4Jbx0chrChkqvsG@wm< zK}m3QxigoEAJJpCp}RLPfyhg*zn?G-+&}5u6A=GHJpl{}AR53B4`J}K;8^P`D?>Pz zCt7oVS_+~w{e1HhxE+Z(2plgdQq3vRR^AS322er;c_X9E1nESY$Tqhy^y&Le=+q>1 zDsKEgaB-Zg60U`qwobpSL!{)gSPS3)l@NxQWDnW{L9%|(^bqdYqhG3GR>c*_Ft!Y6 zZ*Ghs=<}G`nn7I7I1pX?5fs=g#aFg2!ee2{9SGY6E&C%aXQr+t{zPw8iMuf0qk&|T z_4O{WA)!p!PmhU1%f6`>_)YIY62$#My%($nb-j77BL;qYZSwtT83a@a&!x{3euMW} zz9ozwRmS|r@t{(q1M3VB7Kwx{r_2g<7~s2etyCYa_0_wOiAWV2f8T2Qe5uP`1bY$Y z&gYQwyT`}eLu2O2G5^F^5fPx<(_CFQ57ph1b@N2s+&|{-8#51%`M-|ko~+xu>gG|R z545rL?};>50VD4vY>2r}4l%zObH5rhza2Aok0tllS>S$MH=J16W;Rh!-Eq_QD61)e z3qKw+w~sltcZ~U;kL4b%yGQDVxVY}odh|fu{H1Ols7nm#Z&mp}R{1xLx$lgb^<(Ci zvE<&my{B$?_dRvT{_pB0L?(#4Vd=g-2GjZVF@F6GWASxk?Ax$kI@6_r)}QV)E4$pY z$c(kl$3zE0LhqwBWB%h~mhI{>DQFPDWc?TCP0nM!H)Wp(VHSO}ZlA4ZpQ}e7Xsevj zW|cj+n3;fOUMl{(z0elB% z1eTVuMI<=0h!``wAevu+jUC|4J9n45MytFkM>e(W#&R3!r8bstSC*-b^ z_V|`FM~3~LuBH&vGShNlDC{o`+xyac(`{ckH7#tXZYb9?y>AF} zT=>8+k5jx*Zu~a!Ll4)sh964dS>b^#@q=r@51PxG(-XL40yiisuig!GfD;-xg>#+k z@5*^%7LmEl#4qD|(t;zFw(P5nw#KtX#r5T0y;H%od-=}fdd#oz6H*D_$i;})@@5Jr zg@lBV^6ZN&6fzs(#;sKrkNeQrw$_(7&Oypkj-#Oyh-?Psb%DHc{1V)`uAFygAR3_t zK7K8=I!#?yZQM}>#VNTe76Pmc@?8h;lYIuwgLT6+zc4bF8MBYs+x#_TcYU_QhDi$W zZ3-uz>abuA@Sy|ige4D^#eW%0x@VcMdST4JcGx@9JMI=1`r4?wX4G6a>c2h;zASR@ z`-~el=hw_vYwmh&&E=!+vQhKJQGeBF?rSysiJBovCh0L1=I7mEu81eSSUGh+ha;Z5 zXw-dd)O>o>d}=hgw3Z@HPEgplYUa|K<0eH7q7V@0{|uWW&9Ua3QFr#JIe*lDWHk5X zntQAg&eC}IfSqQ&KNZx2ydXkOKZH_@wU&_vVTj=+62DP zTos!KBMSOhNbdPjOv23$ZdbO1xw7j=+|48A))D{15qkLBiTQjSA6vxSwB0@0?w%rH zFRz5JucXKnNe~3-g(K#wQAd=QPmcI|hLJa2Y$Ur_Y*rjK$)M}^N z{7L@g{5e^k$6|RISnnWn#h|%r$ee87jHT9J%Jx%bQYCRUdlMaRepqt1DW-rW{h&lw z@Y?fY_*C)KgtJ0!1)Yjt2BHVTS-2@?soSHxOx)!JT_b;Nw^!kSt(+~mJqmleJ+k^P z;Grk7`6sg|CLnj*sC~F(9x9pN54ztC+WQCneS`Tk%G3x@h^XUT%8C%3QI6hU=8=+F zh#6HbYKz_kiT7*Ww1JAdpAEYI88p8fG`|>39xquEV?ACnkCjM-WgjUKR1bP9*+eeE z#5`Iv57g*?QkePmGq>#cC*$n7iM=2(_twmYn!R+;eRj}ZG3Y)&h>yp0TY2CN;!45t zmZLGK^HT27l6eG_i6ZOHQt3~n-05ZW{&I<&4i};DVepYdaBaGKB5SXXv6CAOth_DG zgl1h}%7RLgc|lII2XP zcgl{{EX$qglPYId^BTD2>EU@oQAx(l^95wf>6?@`Kq|Ogxc2}sHgm=9r$Ngb!c7mT>Cn%pHd^VwSUgSFB}Yq_gy=1aBe1+~7%YQdH@rQ6YW&QIzX7yR-rp#5pN8Y=2) zxjkD)^q~+&?LLosx~tZ2EA_UXvHqm3tuyI3up1T-ZNnmXtXWAR``UajlolmHNtNi- z2qAi}?VO)o5YO+~4%U36cl%tWGBRU_0vdINH=G3rJ$;Y<~S*0#8MKz9GZgR%z> zx6yM-r$YcSdp63;H-P+rPFY!o?j*wrN^*>C3uW@dA?0Gp4Cx%&eOr&x5AZeJu8X#H z_O3i9S4~Md#T$@uE&R%gA2qHE6lPpn3*&GY6!SD z<>lsO<`wQp`wD+#{-rRDufZ(krO``9UTcr?_*Iz|=IG>g_89XzbByD+?RfiobDTN0 z?Sz)szcKq7;$FSVywV=x5Inkri*ISUdeRJ@?RstLs2t^&D8EY3zcW~Knfw0#y_TAt zrVgo&mPT@PZ?zvHvqQOZHu@i=u>;~3o8NSQRydRH54Y^k4f~y8ORNECLeWVnzn0Q= zqOG!=D-2F2#7v*U%wU~t`s|EyVW@9P;NV<+X5)9P@tdCabx~wf7`b}p*SW$~iRh49 zY6-qADd%=W*_<`#&KxwW2F*&2RrlTS(ATm1;?i=abmN^-#0@WB_u^Q%0#r$?s!1mKj7~h zun!KnzYZWemL-L*;P)Mpu_W0$2K>(l?5_vhuUfB@!Z~~6fd9^bT|eM%8Gsc7@Kj5+8xjBVIDmHAJt?w&0`7HIb0df>?wdyj>ekFxdl+ z81OUgOQAmsV2E0_n9q~Ny|eLK=a-a#$qa%K#!0nAsRa}mEQL_UH?)&`7M z!QmL{gU{iTz${C36^K%rp7bh84bq0y!r5w28ih!HriYbeJK|O(H4^nqdSVViEPEBo z()Y)6qqM{>%J;lW$QJ?uq+e`r$IV9TO&s^eayStk6Q~%j)=WPp?pa@nr?>$mR#~i1 zg!LZP_|y=oLS~Ti$A1{Ak>ehGe}{ny?!q|Ve_vA4Wjmmn_Shfv?w5JEpf!V*%2F6V z4JJ4yM-U5$#8DOjxU@G9nb(YnyDRTOT3R4F39gyhb;(W;v+I(su>GVc+jWU+*&ebs zNivo5h1?fP=^N9(U5*G?hvf3eA*M1HAbVC~(>tgPPstQ7woVxo&d ztU;&5^1>r-WQn9&_DMsi5xi`ZpGtVmY8h;8EIgaNoN<^VWhZ?N9EMQ``X#r>XE9zF zF{PpYJRI}6;5taYpd{fw4+Fqu}0OB9^$ zsVW1Fw#Q~$hWHcD1go+@lRo~fZq?S^=6&lDQw+Z52T-%x!r-Ufv+snDJ$bi zfV#@Wf+#Nf=`G;66V)6{!L1B2vo9f0xtjN{qv(~%x8*3eGO35{sx%4f8^giUO=wO# zFBNRqN>aVoC_-3Llwktx#SF%pd3LOH7(phLP(u0Y7!EQ6AH*5U{9S;0l^IB4Rb06U zPCSS6Q_a%!D#v1Ei8ROck@4%(2v!!J=6VizoM&Ds*RTw(l!yD$&FJ)6$f{ClbAz?|P+Ct)Cj>exvfowIfx6F_KUVD&_9F`GJ<4wTd41Gs$MYUpKea9f|6#y>!&C z9%XAkHwqq!lTuudZ?9MWS`W$}XFn@qnWZ|m%T zueD~K?O2P=V*8G6C*r@N73G`HT<@}%TKgH=6vP#Qi3;S*h4gsoKI9#sFQ#R*8@~bT zk#X;OU;Ko17=}OV&2thZ7#sHN=VTPAbwtP}X@q-KoG@v?t75H~=iN+ZwNm^YAGY(tHV)ex+-eGFf6`|zMVBK@ zH5im+SFCM39D}|vLEq8il}`ZCyqkrWRvhnYc7}Y|!|cg-3XsSWpy*NNFErZ5ugWgQ zY?&5VY-2ybRb-4;QJB5jxYM;D8~bjOW4r_CJcX;I0$^j!S8NxacfMFg8Eaq^8&7`5 z5Dqwf?i4-6I@EY7JbRh#B6#&x#FAnWI?I~>R`Gd~Ok8B#TjT%Wi4)`h&^dFWC+q_m z1qxcSSei8Rb%EMPelLJ)0ZLqKgJ8UOGJDFQLs#j^$nHfZ+J$DGKc+E2dXUwDjfM(3 zgQ3Y`55NF{a9+Z>m}p5n&iT%;zN9wf3ZBkcu9HVSJd@fQHt+4QCvZ(nrG^@6opOk1 zsf9+a-@EdvDHupChRdB}r@1c74`xFn?{8<;5}Dg2K>FQ4$H(QkJ0I98$q$!WD9{op z!7YN>D9QevLS>VTb|KrvEYBgnEHck!-l%&8;UcHDqosfvl3B@Z+xbqfNr=BP`h(fz z&B|y&*lriLGsAYzO_*f0uXmYmW8q;w96E7aTN#=2G&|m(G)?Aei!4=kgA%uK^q>rIn@7w0q z=7ICt@_n^_eis0n-=`rBo#}c8z1tH#@8Fj4lnq8x7b>f%tnF_Z1(plD9mjx4xKFW| zSzp-Chx=O3H}_N3zebxyiTmY|8z%4eez`e<=}g#OKZfddU)xR^@-TkAxWwIJ;xkQj zpeA__)DezZtPR!=^Hq>RAvxm~W6gMYQp@(_YY-Zp1vxXVGi@+@>&*UP8?D9g{sM$4 zSV*Fgho+S*7d))=;GOxnuVEVXdYT4JFcalY-6BF17~vI-UBGJ?>5aKVL+sVLVLKAG zJBMvs*k;3aS<8NUCf)XJalSj8iCVVXh5i1PJ9lc?Pp>WwXLfHnUuxNZL)bquY;$2- z4ck|S?T3gO3m4p4t-)|+N!Yf&n0tWT732{9f>_5Q1lea&9x?{tTy`P6sPv;W$_Yo_V=2Bt%%(Nh=dc%eGxyy^UBGd_Ps~SL;VeXO zSnU>)kHJ@(i$wLSCaYx~Pk@)ClEA=qSKvyrd>=MAGK-HU(3}J9VO?-{&OMqlkL2RV zbLO#}dmId#`-UrAiZvZcxcfP0uFGS52~kSSeEsruJ44NN!t@Wqw+_JBlKT^D^4aF* zeDU%^?mv^=AY|ZmalFFK1hg2lK8SV<`Fb4DSdy>g@nyE5JFUTy|MPisMc!PNHzRJ8 zjwY7ByHZ?6Cv-z03K1x;5))F+KSyrE#_y2EWLC#LYeS|uTi z(t=sER)pj9*%{$pS$9<+h_sqwjjeUG!g^UWd*kRD#dFY)*7H&Bh-9zq7}Dbrj9F%r z)A)^G-8|XE7bWe{qC6oWY@yOs9E^dV%4pJ8B*lbp>zLae+g(k$(%s$%Db_o1SpTfH zLUuOvRzRRnT5$V%4hC3@D|o>62IYMLj9tjhCSk-<)bB+Zx}cDq-Ck-t3IUYqM18<- zp}Wx4hX8qmI5Nxo17D|T))E{;;EJwuQ%0x8Q_4s^%cF&ny@QB2MojF+O4IDvt!73N zWW{r5NQ>kJ*Nkt8zUUQEsPFVDGUcuXrTLxPvCg zz7b#T>(?Hg&9wJ}=Z>JWa^SdZ=E{<6@rB2*Aewho)ViEu`EGvSm2^=e82Z$nlnbI^CsOpF+NU~BkB6TF0#zmNkI;e~D zU@YR^kskw=p?zs)WkqV1KsM8d<`>z*omtY>G(-N#Us@6*dAdrV(ddJD*z9E>56Gnh zJs+>Jt7Ee!o))%A*hbH@T^7#oAGQlx?pep!P{eT9cC3!+JDii42`vA7w__Q^xs|WW z9>qR1R#s7!mh&Crx{h&P4=k40tgCcl9+0t+?>=16>R8376;#7lHy z`$OYw{99u_T#YWM+K*PVVe2ly!Vb<54?k>`Lx(aSsC5E`gXDarR3ei@c;~98>h`O=jjgbBrd7Nz?-D}M6jD4(Z|5i5NN($?++Qk$* zF~?HUapvbq;WnK)yJ9b@n3IvLE{@?o;v?-{W&h`@`BT;YrD`9j#$lW8+sCVh_5CwR z;gSSHJILQ2^h5uXH$TZM$QjEodDH@o2fosc*qif(Z{tuBCzV_Bwe?{yE?MvJZ^q6L9e4R0Ws1*KO!Au+Wg6*E%t^O$5uhLj|g)!f$7QRPl45S=B zcW%D6N~OO~MjRqwtodHWY=A0AS6jr-v)rAP=qf#Xu`#cV-&Y|h(1UW2xVdWnTycM@ z*f28owlZdX&9!2U;QXY3sqsMJU3)PF#LFwgYb@T1bBcY^q^_@K1sFeQWUT9eh&mW5 z7C%RFTrV{@nfzD|d|imjGYGA-pc+*co1XR^+sf6p8ZVa&?(+y0=ttk9TweUmd-* z9t$cT?639pacyvFX^I&v3@jWj42_P~MrzZ5xanpZ7$!);-B@*6GUkTS1EPfk(utf);hCS zU=FfP76Envy$gkkQ3x}E*}Dt-i4utkw?7~gRbca1HHXIko?79Ln8oJ4P^ev501iCX zNO#rjFKc1lx9A7OIzbOIX!2bH{pFGW;A7}VU|qF10yt_gA~%Mq7Ku7SM&Vb?1@RX6 z&g1Kd58D1;nfs{%(;o9OeX?_@$p&5!=+ZkGEzF)Peh_h8tT0mH9-)y(n)f{Je8u-P z`eXrpCDlS$QWoNw%Xt#ZQJE5%dEw}i5+RK&md>@+vT-hltqa>}E&J&iwjj*oGMAGZ ziMBF#Ncl-o*lEGXatxV>2@Dh~n4hdtLCB;)){rJahu$UtmX>3#9pZNGkceNQ7FiU>Hlrb@oW7+qW%xPM2X*jU* zfs%cuXr3;*4-)qZdWy4{A4_UpAUPutlD%Q7`Id>ELFZT|ygzn4S?odzUY^L@3?2-g z$~8oAvv4shWH}ESwC}?_xR19$T5_9XDoJy(2sMAxF}LnIiTfG~fr61#QbNI4mg~x` z*yR!Ab<%B($Pz^d!Ia*=uTw6YI>>wc>XDS_UGPyi0-3dP1;Rzin|< zMV(=^!w*D-CX-zGT7tzYdc;PP=D_xfV@5YZN8<4CsOfryozNm_RfUTsU7!avc&T$n zq#65vvMCU7!=D0mP{}M1S{9W3-OpmXb3TpuM_sAudzcFlDR_-wR+7mmjex*Ec$Q{& zk23d55Jc=KY*f>R!dD_fKzo=zs5}tznQYD%f{$i1`0!a`Tdbs~U^eIQs-rh;d6 zV;VL~=}o6%ogU>1u}}#6#oLuII7#Ef5|oE~V!++!|Hxi~$K3raV9^7d3SAE~5NcH7 zDzMbe!iM}};yDdPk46ER;Yw3Er7P1-MW7gBa5mGg;PV&&6@p2x#g+nh8@W890+#-5 z_e?i1(fk0^TYN+dPJCMLnArb2dyYnLG%FN?Es4=X`_8_O12aqF-?aN*wEJHYI=Sap zgbFL5Gr%ASAfp4qJY#Y)UZES7jU!z}H3E$bvGv1!j-{bdynPcBjTx^nT?g{QhR1bY zz0!@F{&C;{a=4-oLCaYno`QxM`~+TYLAq)n7iKt*`%oOQewBGc&TDc#gZ4X3b4u}wIDhi>zzk7 ziF|GfQB$6P91N>||l4eBev9C8{TH{zvK|m47^)f%8zDx;( zFmW^L1b!vbAe2EkB-<9+d)_9063nco?Gc*x*07zvnhYGFfjB&_0a?+&hs<~jkv#*c zOsm*Vk7kshEX|Oa0Zx*sEMo?(+UBN~NsS2H5mZTmjzE?SCqeeAB?Qu4G(AUbtz4EY zy^V%9B9E)#pVoR&1oI-tB4L^L8x9O*y1@Wch+%I>Suv3o84t2*2^Z5D_pf zV9rX|wh{!XC6D)_BtLIU-je)BazcU)ON0iv z;On($n07N$8RE~GWF=Ms$#$&u*4_&2k*m>!ve4)SPtGmQ?T8u`C;=`lu_J}fX?$!1m6iulFVEa zHy?}xJ%d)Yht!cMI0_NbSCZ#uw$7zs z?~EQ0Ie1^-(`80)I*!l$gjrdQ^GWnaEwSQAB&R~Pavi~;4IUb6q7p|pTJs%i!`3(F z1V<2Eu`ga`+=C9n#-PbqZOkRq8pycb5@4i=fCULYHi6NDxM$kGvarUAuO+(qyYf2* z6AsvAbYB)VXAN~UX*(<2P(Z|SsqrJ3%q(Oe+|7DVU=Su8lLTpSm?=>ossLb9axp>! zn(REz&k*RltJ7Qyxz&oKUkTeZs{Yt&-tS-8=!G1;FpR~wdL~DT%pJRnT{~pXL1c!O z0nK-*yBGwc29Jk0TPPOJk)3v7el1lW_RHPMF<*% zgf;&amt2ecRX)nk!#v04pES%whWcB89e{;fpDJX|0l`EcgjhA7AjnljG=N6O$?^!2 z2)JO2c8%?>X;+S72AD;Ie;9bK+jQnaN4C3nC+-$PKS?o;=@1U z>;;f0$a?L0&VJO{^M!&rI-dywF{p|9370*uiSu`2)`9b9qq#^T)&bFadr4Z7R-e%V zy~*N$spu5Q3IH_dZ~EvK?`{=1 z-N3u)|J!`5J##yfX!lbeU*~&MpxTkTRJer-bkOo%0;zEY9^W6q4`9uLM1z10)JFu+ zq%)X;nPPY5FtpnlVHCYHOdlYJNyMV*SEdnNJaJrB$nt`v1K@5@^YA~^Q^H$1pb|n_ z5ph9GKlR> z?Sp0=_AVe}!W_8Wp;C5pQ-q8MQ)K*%kTEOq7$$^ySRgWa1!O#>!VMr}eB-CMON^VX zpq=y#zKWkZ5 zNnpItED622m8(}^F`}tvJbh30yaxV;+IX9Jd;B(Yxiyzz#nqq!uN7VsgL;i&eId&z z1h~J&ChM))17yCRfFmC#-_=?*fyOV!D?K8>K@JNt>=9cL`CWQr6OoJ4wegNABHs}U z&b*oK_ocZr?9x&|BB~hx!pbKh&Tq+0qY0Ij4&!eVgXak zs1V6Fnn#j#b|LFf`lO+k?lG_~BfEEH+Lew8nLzkLB}UhxxC%%F5^?-@*UZZ$C!hlQ z#O^qIEUW(O>@jQ+wX;pk%SlC$$i9x8&LWwzI9EYA_E4=x>MvOb!Y zI^VJdJ1p^w(Rt0qa!MKjM0L1l=EB_7LxcheAIzb`hoNn7Mjym{Qmhd%m4;H@=o41r zZQesKsHu4IX0K)TrA3gO2TDk`lNc93DOMEeciFvZ2X$`QL0{ck2QAHf8H%T+hXgxJ zd$7Mmm*6IPNDkW9=rV6q3)L!8^ZZ@&8c zeX~@3b3$Ny&9cLG@6NO9VoasEEw;1djSLWrW|kYG#MHZY!wK#Ch^E z8o8o~WG{DgKC=(eaEM8bbQO*gs}o5%n@mX%=aiRZj%Q_|!n8AR>?C?DNNaQ2bJ}=JqW4<8a;%EawDMZ z{yGK+GR;FrM2j<5rlr2qyoXXx87~#yWs72m8|VPRL)c5VLJ?crOHs$)-CC%G|C@3j zY?hm@1Y6c*qSjkpYo2+r_iW4k&9b-UH_&5^{%PL-uYO0f+=*Y0ER+f8j{>>unmKE{ zBexi8Rb}EkQ85JP=@sN*ugE-_7BF#H!j*})z4*O}l4SI|z(QIY#H$@edd#Lx@rAPu0wjXwR6pWOUp}3n7Hr%dc-H*)Tn!Y{0t-f0{h!I;{U2K zAMe@m5g)JHq+TDt?*-0kb~T}9qLhif=Yt=OQ0rZSdJk~~4uxPwxgT}*er-w zbQ*}|5)rzAnjkSnpy^V>*3w#Om@54-?A zYy}-wZ^mpAzFncSp=c4{;cyk2BnC`z^^x6kyEyr|fo<;&+v#B&uO-?dQDTtNz^2#u z?u7;g8Xo%7zux>!dY=t>e=Jp)=f;w5$(vvb=OzoX7z;tPAClH9doZFJ_2wZ}aAdO{k0AiAc%+jdBp=pr_r_y#t>531d&RXBl7g9C1v{h)?)% z%AJLX-=2@Hj$mHwL4VkWqi}{Kx8V$>X^(xbg|gH06L(6CN)&SF(@FQD$PLGoJ?UOJ zs&H>wYFh5|+}kYAyZ;eUb%!^<@VPs+HEgZJa(Hwyn&$*94Ryx!cVi} zedcHelaR)DNUhL>*%{^lKF9IKPq3`o6`0miB;NzfUZF^C44q9lyXN(9{vW%pln`lE z^|r7FW_m*Mv1dx=*-~z0nXO@q;;n25CeqRwVF1we5%6Hk-oQvQT4#|(!YE<`k@D}}f_~izCKjdfsHpcyL}YPTDew@cbmpL+ye|@(P=KJ zo6psAmuqX-0z-8LRcnD;cP?%(VeVtfh#E7U`+ue%v5DQXBk?RVE2bT7G7lmW2M+O# z(^9sPA3~q8_j~RAL3;#VvxH7_1W;?@87(acp{1~a!^gTGK}uZAe7U_b4zEMxrcV-J z@+i%g6`8+v1oWfVqzG$#`}dC<<)))NYKExM#D^Hwla6s|mwbsOQOhknAwaXy-HL*< zkVy*E8il4%Am3og%rA=&OR;bKB9obP+&uq$zZ3W8FeK{DE=4FhD8EhqE;Z&e#$0SH zidFoLP#b^R$opuO)}@b!0W=pF6nxA-P;Akko1SBZI@`EYP4rInjya7Os=DQ^-0|*k z<{j>x9KVU<{|v`(aVK%Sg5x)ar@3<6j-W3N*D1bEOeQ`C@zlyqm6s}3dVGvnBq_+(uU}JThl{SjdrjUN2P59Z; zwCAA!l?ivHZQCSsxY8ylz-k4vd54HgG}{_8%+RheUiNqmstpi-0jjMbNR%qBNwqOu z%Yka+ng~$|QJYj7!b$um7Y1+1p0J$}w)jt$Gu18-MG38Mo$qp}HVg{D{4rEpx4_Td z-L*O0X17AOX=ZLlw^0LoBHsqhHVDnOzz(N0n{p;jfo3~Wc2gn_nlxLzL92a;`cN(dL+1Ut<;ny4Uh;JNmqB!YDID9>KnV-fJ zw}i(6b|(_4`FP%q68sAcidZ%FC%TQM@qeR=*BYSDAC38_R?-x|<4z6+4l4z71)I<2 z70%Amd3{=GJY8oB$dS8owfTgeZ|;Y?ta+gl@JvoW&1YWZ^TaJz&Gq>QOTfqT??0dR z&Q|+iTUsz)H|pr!J)(>57cf*(kYH=gU-RZ+{XUX6kL&k|y!o4cpC*(H4@~^I2l9=z zPC!Z$M3@^f62@+Vqx+FCdxO>$y-Pl~l-^~wWg(vA23b%_pLNmYZY_4N_3K>bx}t#d zZak(!X97Uu1G_>1V}2$`ywRWq1xoFo8!!q>!oLV(J^=1-oVh0N&v*Wt%t1h{7=Tr~ z*fDMe;?fb@V#H^qFvW9X?pi+E4sjFX)%rYVJ?^V%eaFk);{5?)b=OwMdmAB6Gf=hI z%BR|S7g?8HCq?j zIE#1gKQT^eyR{8p6N8tc2-t)u_e;1>eyvY)9?Gc%o+MUv&)>N-?O{Ss;=>h zv%}irJlyrzHdL~f`APT0+rjbUy%@NT7>;8q+WB56Ri!#fIS?sYEz&-kx5=F1GxgJK z=1#Om%sW7(fjkc~=b4HBuhEM{GHe!aXW-M4HaDDW1$t`T1ebl#-!`5}pJ*PnKFhw) zZ+J6J+4g6e(_-77P?Q0I=(a!cBHsa+eI0KUbNzt1VW6^3Tf-K%b->)D-2U-U|a8vKKnau4iT)^{UMTIU~tT1a+wbfB{$~GS}DOBf9AZL z`H<}783&9i&|8M3@gSyo5tE95>Ef;Id@UeJcZV^LDas-rteu}{EFBEK`ZYx$jk^T} zjA?k=3-xyp#%%HI1j_mcp52B~-uhd&1#X|Gt+e_@!4@d)!!v=D3uc;+&8%Fctziqp zKhu0#zpH1Oi}m~2ndVacUN+NQ!2=V2?mKDT$rQz5gBl2bEvg>?#r>-;+z}#j!tu9W(Ie_EW6pBtT`TL8DFIB*<0+F zxE-lK%(C0X?H29omti8fXS9c3?)Gx~M0@-F+`j$*jI<9FBfZki2u9|`O)%1(S!m)u ztkX#dMq;K0BMl#kE(Po1t?E)5j5NBG&@l}QIJ(QmeQ7te`N{-yy|l?(myi{Zcnx~I zsa{c9^o@341F|Pu+CT?(_LG*v<%@V#AAu1lqIs5|t*Te(Y{|NV`B->0^SDu+9-UXh z2Iwr=)*?b}QMJfld!rc*>LlE%vQ4%fTdp&nU)j>o<{&MGkw}#-yrLS|cCLwS*SE>G z^OEeE>`tq*W=*y`Y!`%WygFMT79JID!zQ}McmD?4j_u*=%|(msvkU#Pbo>^o76MJD|UDzAsyx4ds!C`p4#UL4Gsm!R9469;aEl)Z1oVUes z)k`<3mpW{GIn$j`!GOi-Q`n}?-(}Z!<}=-oX2sLr)@6Pux^wbJ5KqPjqK^2l;ck*{ za3j7u=Rauu`sRPU#R*t@!gy?c&*T&S3b~UsKa&n(^HABgKApbmlGX=!kRaS!xr=*0 zz1fZ2dYd&r?j!`jC^YH3pCCyP45qc z4Nsblp2)H8LW}=(D;eA@i@WP)w0ZP`C0oMXr36!ByE!M;cv@swgz10Kt4x#(=!lTs z?8Dh@Ej5i70YlyTdHzzHJ&^uBUtL`hP4B<9#HqCXUs@u}M}aX?nvs^zKi%>dF+$tc zm(;V4=3$4{S^~d~aSyCbAjAVPE_;meqsBFI?Qu8un=ka8M81GzAh3Rv`Xi?p>-U%$ zikrS9>-UwBrAW>{1?*Y=M!dS5WV+YKnG1V_Xg=0u&&ZM!0z?vf6hF9G26OkmG33;$)-uZn~f zV2dTh5jNOFP#YBzKK03%0i2*x(|_JDffmwErn|NgEu`Ju97s)vw1n0z^{9iFCS7R` z-xg$p3<^OzfA6} z@YWZN76Lq|{TRG%z07|cF{nKH^%!%)R_po?p$3)te69cg5j8Y;8xb?#M06)Q?r#yO zf%8HQYcVInY~k%|jmfNBI7y5~d-yDc6nXQOi4k)=1KiNwZJ^1-)9$?l?wx2o3x?Y&c7{_nLkMnKY^L}QotJ+_yo}TAbx!Piv5q@ zZ+j5ZSuuy}{1rs+Eo8c{cbS#jCmYM=F42UV>2*2n;%wCNZ+~F)f2uWqQVNfjk9|(F zCOp}CmS?}q+13xx)otD-g?a0z`F!g+O56JLqvkFD)bmr#+y9y8>BBC_jW_)>WkAxM zV_r-FlL3_@Q4=q+BR4O@y~svv{ycO-4L}PP4Xp!bcSx;XT0vDPMc^^%6OvB95S5^7 zJh|B)X|@`yrqIC`#6rW*{mFt30)uVrSHzO6jMg}X{p`NlWuDwKxwK?f<7bC=!ze5E ze8dj!2aW&e2PXZ;+1I?+`d{Vk+ld4CybwCk5i}L85T#-AL{{9ECoHZ9+Xm!2Kw~S0 z+S~t$5@G=IpR9xiGp&TisM(Yk(+pyuGBZL?{G&zGsY0V2|7a27$OPELlKMOC_`e3z z&y`x#L!W7=hwc`ba$c*}Wq>KG*2}eOts`t9-1apHq1d@vvZ${Rv)5d!y{~KUI_-T+ zd)I4kt@hSw@4MQ&Nqe_wZ@u<@sJ&aY_haqd{*uk+5QnTqlMQwzq^FMXQLakdUNWt9 zWY!$Yf;Rbc?<9F@kga-@JOiD&r~uLwjS3Foyd@$X@b-#!GZ zs1`jxS*eKPL3bj?8nWL~)cn9LT6aJ8NEH;6ig&B-#vTb9Zz3k-sGUM$Rq2g7du<;7 zb>Gci4^h++v9dSdhaJ{PLBMLGrZo~Sn0O{O1+0rmT4z&hB)2r$#qZ`2uo7@&k76*| zo<%eyocUP7++b$s;05net7(iF3XC&?H2@KxQT|dt9mbw!sBe{~nRIjP2~4;%&$mM% zkD4V~Y>Y@Aon&HE2Z%fKnS)5GPApd}d^+qiSZAAih%5=zO~rDNpXO}{Amce9y5jV( z?Fw50>JoJ^-M%F3ms`$^?_+3fY+EMjYpv&*yu-uD&Piea=2RtiAS1d##77@hh|K(wg5S zJ0$IMuj~<*o<%XV(fa)%S?KfKDLIyFHaH`KW!c$^rd{!yvA`^6LP(gev1CfJ>w(H+ zCrCXTTj1GGF)Q&YLlC4*31&*@Ew)ZprXn6z_|nlBK|Xo2V;t<&aE>|{B`F>q680C| z&!+)54F7!mmN*3r1Q*&{E2!X`#L#7t&W+9HtwiyFFBA2AsJl8)Z76;*UTB^)cvI_V zWtUxWKcWlZ7zf6L%JrKXq7nL5m{uAx#T}45JIpFP$=G(7RAupb^C*VBps3pneh3xG zV4ay&3ePe8#0Rd?fcHly;r}tmXn;BsD?Y`Mj7fqw{GvKnt^lX`?!U{cG)*CZk!w|q-FZecuM^#i zx*)})umy#mWH|a$i1eWL{n6OxjQOK+&oMzvBV5;hXR?PdoGlP8w#A`!fEBRyvJyTN zN|bHpwxx*VW1olArWPnzaFoEAh*)tR1no=$nKc4xq!gl}hx%&tRC<2SM>8MQE)=Ux znCFap-uOQm{J$DT zZiO`u8vlq9<8mxZi=MZ~u{dJ|WZj3YKLG_6eio&6%3?$;tN=i>Oi-h(PUTtYy4ShCIB)EJfc)wF-ixf66>`A79@XMSM!VSMg!PlCX<2T0 z04N27LA~rF;92}1_%Ig<6)cw$&=hF4a4Uh+Rj6i(Y7~^&E5Nk^TH!!X0T+z|xUofz zyXmOa87G_9L^1UVju55)N(a3agtHfB4&EjHQPo1OfQ zW~tv<@yyJweiyU5-wh@99%e6Ss8{&CvFYB&?CN?jNB026`Qv7tkgyBRi z(zfRtyH#m<@Qa*S0h|TOB^c-HIHs(4T+6}O1=7V6iO`P&U+f&FDC&9_Y_WnM*80R? z>&L(rndG(xYlMw~uI&L`(Go{2I=mSv@*fS+FJH>SBnL+GGs#@{MbCP?-o&|sJVwCK zyBtk?w6jM!!!az5IvRVdv&T5YG0)*>?D6tKl9N|LDCH;B&w52afn<3t`XV5UNl^B@ z2W9WJ%KeUv)iofC9AQv(5JrwPv|(|b-6xneRMu*-UZ0pIUd$XfyM_!S;Ws%t8}BE{ zPsAb=yu(DKax6z%DDTk_ucI=k-`0BbGtbJ)BI3~<;}4!tn2CTu^uXq{W2-6C@oVRpb#pkfmJbB#S35N=zJPS>|8E`D z9gqt>$F1mlp#hPlvwG4o{|IAJ=It9@7H3WWCyW=Mi^JyB-{TuhOoiO?gfu$oMbn=2 z=IQeN-e>=((x39?4-=LDKb8KBH_uH}evuwPD735KwnBOVS}!I22|qsrpkQrS3d4}x zEx@BRJb?mo_j^`A+;FFL^U-*Ma!-!UsS~Bc_1QsZ@?#0z^<`qeJH`^PcBELESk|Mu)^@2BqsR;hDN&I)sx0pA#-d5zj z6@gIxREqDU_nP^xA731s-_?rz7=IMz!t4!~O_Nj33zlqJSrD@0|4^PJaCvmnW zPUpcx;ocJKs6IaAr1$|6c}!nVIRRAIrQw};X0fkb8UMgfJ2zqQ8bn7Muv@^wkuSrl z9brWG4!kXuM_zznqv#q{4fnN5Ru@Vtl?@)qQn0ICzy6G`CYvhf(ndIRLlkm&^ zj1U7#v`Oo~Zc4WY@hGTnremWZt_H0=%f#mhgRn^Ps6}*04&&IMum(X>tH}9&DIr`b zC&0hIy@I8f;T~ns3>J61rs%tu8ha@2NO@M$fwL`Vo?_4>zYMaeO_;#GCR%>pHuTn8TR z8Hld4$+tZ2UF~%`Qm##}=A;?He*$2r1IPY4dh)v_yG4S)Qt~LNk4IU39f?A@stT&MOGnP9+{n42ipN|`ZS)n~A zn$VzZErL`e>(Rw7xyRWbJ3mXHA^_?ADJ}f@^lVmyjkfVzW_ak-{X(Z&`@YZ>V;$RP zkCJ{MG}5|rV9rwlIOmKYjD^B1P{)LkP{>-FLSSvkVRuhI9bX|Q*fDZ8%};j zq*tlIoaSHy*MSH%=SrM;9+8POk^EWl;5o5ZKoa3l#nqgaLrNd93`y-E@E7N}$`J9v zgy;{g$X5W(`Oz1c;r2^$b_MDL>PJtp_7YJ(EZP>33yXCPXE$|rPXpy0JA)|(wK+db zn*Np6tg;JMT1u=^;?n4Q^r`)RoZU`|K!QsAemv<;NY0{EhvPydt&(UDpfN}ll_cW> z+sMaHhPQCML`Uc2KU%E`g6%Q+`4qw6=h9@d?~c`m(`NzmYYQYo06yKpL`JWWBVzHi zJLP?|IZkc^aLuxF1Y6Q0F?Udb88?T{BC1U_x-})L7lFnBaPT7GUUbF@S=o=$QQJk&q{-V~v>pF!pG&gmlgbC|(Eu2I z!C%14LxEcr9cS#W{OUA`={pNqTHW^tPey(-w^?tqEOzt6ekM_nPVj;{H!dfR>fbHWBpZebtS;*Z#U5| zDjC`*R)6opmP;xm9S4In83#;p8zCN*5sm_Pv$iOQN!}l*%Bf~D)6AzIudu* zCN_5Z^gy$>`8M@0ntl&yTDJ+KfY8+~#1+lnY)Z-@jIgwX`nLT(lY@x-&-_I2ZdNold!bGQ^Rq0A@NUtjQ0=?|qLB7Ew4^vR z2S86fe{^X6*9A^ZO?5MWAfEgxwX>RLn{2Orv4M-;_MFtG~|uM<8u7e;VGH z4x3r3ah4%OuEPd-NQW)a(x4)~m!(>mo7al18Ga>IOW1u9bl7w3oS?%F-{MP;4u=Sd zi|khN6jwfK@guSA>ddUk7KY>WaKy0$Df#ixaNX=l8M|&o!;(F`?|WUE2$ zN6r>>R#%U$)G|kbI%@oeay>Q^5s6dYF;5S(M(ZCk4Uvwa;4w`$YaR?=e$-gmB`KQh zG-$HR#$GU7YP03y*3`iObvJ=FJAH?Fdzx#r%U6FrG+v|KBNh3A7G3#wV%!Ne3cT!s zR?z{=v(RSqV8ttN``%CpLiA$viGxM$`7cEtp@|l^Q?%I%F-Lz5Zn2&=J4Ga%25ojS zrp+dsO-L{z}$LR#jGQm&+i>t$J(6Bcu?wE-A_6l5d^P*SHTsU7Pg(|Vkex=Itt^3!HW|5HPOtp(kHlBaKKHXG*)Sb!MGwuVoyDbDg%N zGnxW9+Z(jd^me<@LIc`=kro=*RYB6)sVRb+?Pv-(xQ1@M)Isyq*vAFBbiKy3)ImdA zrt6Xm?b-yZ=iDkl#H7m9Ky}agz~uwj1yZ%B9Mr!ok3KIdT>^Z0-dK8>|I*LuZOvKz z8ifwp|Ddabj(c>_agPo7f4=-~+b|+9RORKW{|ehCrZJR%|K&C>0g~t;K~p zDK2!-2%$ErgN|QB2OUqKgZBR`I_Lx)bb<~#c|9y!x+1~;Q8<&hDfEJ-VLlbhX{Tvf zUYFBo{Ms?m6nSaWs+`Vc`Rkv1r&PVrLHnSCj)M+54m#*~108g{fezaD=%C%-whlT8 zI_M zMmAjPo(ZlCp+S7o|25rng4slZ*+jB;JsN&#N@t7D+}8v@xdgvqC0U=EL;{EfNDWJn zgx^T1$o+G)A|f+-@+GKp3v0#g>1e5PA*%Nbs$9~g%6*l`P4Yx^Tslsbi*Jr}&EgNo zs~4L!i~Z`w?XYk|`musS)#gXyx^Vs19HkOQTo6bBkproj8&ah0z;cz@&YAQ)?|@R3 z3%AolShJMtAF`m3N;hFwG5n&bL2*0b^Z&CsL4iBrcm8K{g2Hyf@BBqMhz|5Gyi)*V z(VGSHGN$Mi`4k3aLq%^~Dtad@A*gyM%z-g-LD|c<^0NO@WiLhwIX%gmSTeO>ainPZ zokxYG`}`;IYsjT7Wg5>0rWj!Yp_3+fc1#Hz-u(Z@vq*_)?+ze~tEE^589(d%;X$!X zk4mw84)lk-SJ6G_taL_GSe{WV7bz%~iz^h%#TAO>;tIubafM>JxI(d9T%lMlu23u& zS16W?D-_H5RSm_mcu=fXf?^pM3|wd$6wB?m6?rI@iz^h%4#hGGH=JgxQkA zRwrgnvUNBnt3~l+o=WR-#G^1lUBtt#Qh1a7dbP+CV(Xwso*1+#IcSng&`$n4n&cik z^bb-!X_6V2hTXT&BnQ)TZ(o5Xc~7Zk3SM0Sr}7?y%S6iG2BGGhd(b5BCz*~NdO_|= zYcELTLSR}n$?5?-vK5+SAqLmgB)gzV_Jt<74!77Z9y`3rA}$_Vn&egffnY<(A_x^Q zMysi4vFEZU6n$1+=V!_AyLr9OPW)W?{QBp1bb6XCXWNG)sy{?j7n`5N=4Y{aF!qnc z=I60n7n?_8^t|^D zOG*5-{QirH`$FQrp2XMX_g_uim5E!O#5d*l*Cy_piMu(8Z_V$&owyq*(|xwr_$gSG9;d&c`q^{GGbQl`+H%3{R!r4U0v&%g~@O z>&o|Sx~{hW+L~Ec+bkUCgrn;@zAIceds4=(o6+bD+AF)ZDyRInO|aLL&8BL&4b4~B z%Y4s_Qp=nel}3%w3|zU^;*%?A-2NU1(y`qe8yaWl5)g4Veua=k9so`|h?v8|dbu@W9iXR)Y~Hd- zwXqGpolU0}B$0k_n?e#i;r2N@H|eSJx!EG5QEA`AVL@Wd1f653+kNfg!Av_rh>wiJXGUt!jT2(&%BX5_fO z^t$Dx?HnBv2h4Dn0!gv_aJ&-T!VtVvcBF9n)B{Ehqq&J;97XLa17xqWe;? zsk8FRbN`>Y*KC$R^B5^#f$u38^40BCCfAud_g$9PF`z9Zn%HWno;cB7~{il4eMhx~k2=GWz{8ovfj zG(%q6EGuVs`K|TOy)&%bRg(sZW*Q`#X^?29H?O~|Cfz`yne<3B=Q0PIhSr? z5I-i}ObY2{{J$pMtRmg4BHgU+-H1kCqI5G|=iiJ)|2kh;i^qUb2?DjNx};xg=qFsI z#w<phP$64X}uRreE(>e1)nG!V^ zI!!ZKOq)8dW)4KUS<^<9E-7DuM&KWN>H~3a&RDY1lpwd<;JLp(HOMJ9DEF^V4Kl_J z%Ka-+S)ugr3BOgo6{Tc;Y8y%@T`8d~z>zt`@7>Vh#QIHmnEte!T82rpu09j5x&8ChwOf~>T-LRMN_AuBDe zkd+o!$V!VVWTnLwveM!TS!r>FthBg7R$5#kE6uMOuU^6?oRUwj1zBnHZ8EfF`#rMK z;tE-*M^?&?-*%NI)ZWi7lXPA7Pir%?F55aBleHOZ1y&%mE=Ob99^6G^Rpa4{4(OGY zZh-rPoOHuUhUh)WNXto9{qM*~dn}tUY|e$GED{ZSSRo`07ERv10U_yL5?HF~3i=YP zuhbo~*+obSUe1LIK}d?wnEjbk!v4(GS~l>&T}eo)rl*Jup2bi%pVx6$Na}--G?kFF z;AuC@`PCLIT?pzatNATN^iq~&Z}?e)G!)}a2BL=3Q*?x1iT#zi$nop3`x;_MH#zW+ zfeD1)4ZLCnmrSIFSD`bIA0X=5>jkz~j2;+001v-ABJ;vnyvs>%0nRLq#*s(tVYJ0+ zO?nQ1%)((q+}f8~$}i%PS`ZKB+%}924SqQU zb<56C5(pjUol#U&Rz$_{)mlh^37)}Kj*PMw2JwQoui~RMgf~+veU35*$6%)ebAl~f z=%QidJyqfc;hd&k?H`#`qz)3U7E8A!s}UY*IJRIoLcOeA5b7i?Ctu6#mid+#e;Sf= zoSD30rB9;aDWU8|95V(%G(-NI6;ie<2Z8?W(Tf&3yB8mb!aVUs>^2z)P08@VjSa+~ zFGT|ZW-Rq{y#pbaM%{s!KH)&nS5-Tofsm=_o(x2z7zoN79Mg1k2O`6lq{Tam)h@hx zQVayCgQH;}c2X46fgRb-mPXsLF(DLzoLhf!R!uH8_A>yR8|uy(0B1y;m~@MDdxb$j zz?9kTnKf*eCh#yNuQKyj7DGI{JH!I6e4d3No}(egOyC&8Lzy-1S#IY2E%TVOI8#g+*FIqy+32e=IM2Qzx>rx;dF2!3xlHdpck@)*#P5~Ql;`!o-+i-Nqw;EeURn3@ zOXLFj`%`zM2k(x)$!>vhbtu1Bp7T=kyA6-!yWR8u=JTPY#n|-JfGKeV?W;zYfQCck z%EpAVB_b*5;bq0GeUZQNy@{Xk?Af|+`QkoZ{^Ay5&bkYnHtz*9B z_26}V3zXRN7rqWE2+M223JhJxTc?AD$GxTJq-P>^?CENB9SZPaQ zgZ%tcQ4}hkfLZC~EAs-z7t?KreE4_oUhFC_yypoYt%6S3po*&Uirwz&F^08s{2kiy-Y3h|B`1gA_#R_|5%na zzTe7{%P;oS6?CBCRQH@>qN7=cdU{KxtlwW{Sud@%-Fo!abb@-U*P0F6(ffY44&?}J zm|K=G{@D$m`|s4H+p6BHqHN}R*`uG*#;rB}8nYUziwKR;`F$ebioddEp~snOIaH5J z-SElo+pK7m6plStWBK7L3~Tu{J)2+a>Fyf=MU-U~<>hFj`=)P&cgnK$etEsobI}_i z|Ia^N)Xe=vHYg=kS5xH)$3vK2(fWP~=yN$$gsm`Z?AM6HP5*uaJj`}6kYm6vOn&&pmfVN>N2sG^F$l2KVJ# z<(VptHDIDQ#($^xt)eV_XRI5=>F~^0iDS>JCf)iG4^BGjav;N+l@LqNLKDLhrS3|v}S--r0SscAgv^fR9i3S@FF=v=|P&Tu+BJP(uE#Gum7P0jPx0um; zk{xg;b&!1%4~Bn)8yB-d9r7P1aiH77l2_~kxn9U$yc}Q>XoSD2@2{%hIm_yipH0Cn zs1RHnvF%b@6ZyN=*fquzNjg+(o(P2~nuCQ3DftT)zXs16g>Uln^%A20{D(B0{YT}s zwz(5|m>+wR{)eu+-zvX}Q4eJ|qbE!@V*96fgNL&pc>91ckL$ucWXxmX;t`_K;vmiA zRnYt_T;PK6fsjSY-5~{u>JPl!1=5##i1(jjD|Ya#=6#;KLgSBn>5zLg#6`SXt zoaf#ca!*n2l|8xV4cqBf_zgFVAyVh1@HYdsK>sv;N-$ECX2d1om?Qm>1v? z;nooPe4f6gsbL-N7dYYukV$}n3P%M*wN9Z?0XTlxad1zF^Wq*a9Nf#YB+nqQ z!SMs;LmZDZN7E2tLgML3f^Ua} zrkCQAoUSlmw&rS`t+`5QKU{(NZ_2N}Z50$pxjx#a1xvsmDY`o8e-ZKJWqF zMZ*`YUtzZi=81hWPxPCUIybuApt0;^fXxvO9joBY%%U^513@m4yz z&%11$b9W$ep)r`;;N!JpuEx6ug19>9XgKB~QCpGg;Xa3EbpzQ|qKmzoVMcisxu3dS zOMyC{qQX^vU!)Qpgp5t2yS!u??$pUhKdw_BcUpSZ2_sy3K62SKv!q1tmM+Xcs|i8 z3YL@cC#{itTERUG8@_1hNzf@0k6$41UB;<2o_QP&Wxcywxo7p{V&*1k@+e`G*3 ziHg~K8Au>r*kg>d3v;s$xgZbUTS8|r90M4v)*+0x*G^+?20V)qk&8rR{LH(sqSWNC zMbp#wC#*Guwq&gd_cPqAWM*RycXO+Y01Rg9=zN)ut?>8YaIF?)KNmV_M0;jOScGz97@plwEyE0aVCjPrx;m7?4Lzc3sOY!&OSFCNtAI_zPFuOlo8DqaCcQd5& zRQ_9GEtXvLND!NsIY(Rj`<_%i%L~U^`*bK_T)D?v`vyE zZxRel{$sO-uAj)C&BtQHV|REzD`Aw1{+u(Pbn5+Z_XTHHcJD58_JVL17N5)Uh0@Kt z!r9L_^C`?n>D}9$z0?*w!4Cs+dz_J$@d!g~+NFq*4sQwvJSaDa- z2Ol&wsGVWz8BA7i*J!W>#^qKRyap~j^sK>uEZ?8bx3j}5QN0sq$w>4(EK2<-lO2n; zuu^!!Cm@8V4DeUN#z^+|btrKps@Ej90KI|Eq8q_2@nU(X`Idqax%*8#KOlH02tL-w za?W@VY+hb866hy3MTjs{H-R03_ax{*b3}@FxB4`ZXwgl1O-Eb!vN5ku4x|8{#h;&L zGYHhKxatOeu5l?Mn^nC6Kb66#)d!-ds8A}_ORrSgBdGY_HsIi zl!`xGBp?5r_W5W_@o2Pb&fC5*(0@+6?Si3c>GW-8R%i7`d!Yc+v(NhUX*`+6)dBe9 zl)lk?Lt)qlfiqfN2sM=k3<4!alh|w)Bns1lK4d6ICoSGf0fp%hAUr!nHyFY)3mU8O zm?Ji`$|fsq8Z9^^dQXZTv08kOMh2+%P?wkwwGu_S6VEPg8Wn5;xkn@jpw16dIfkGQ z_k{D+Ys5#^{!g4w39?q1ok1O|OjG_~7VJ7Ro7IUvj6Fjvby>ADu6f8t(1T@7>88Vg zNiaS;G-?2VWO%2bz?oqigc9cyast*cGl9WqS3pr-0oR$GL%*h>Uz6E(&NO*4=Brbp z^G)&|`>vYpyi*+tocx2Coy~)QtrcC#=R;R+mv>+jcMdI_9FYs@YWL;DAWV~qI9BZ# zb|IyC04WW-yZq{8QnX+xuz^b;eoZ$w_@Sk<@kXnoD~`~Dq+n7{3OoTvU9F@Hi00~#fz!Mb zPm$r`E`zseQD3AS9WL%-$e44#3E8YW!9h3=tfS=y+|A|vO1(E{SpyyNn2ZmZT>2vZ zEG1*H=FafIm4cS5K#OiSKVwo3C+gyDsaF$Nxsb>YmLZ$$=h!czEKPZ)N(nVO>roK$ z(p+dGkfSlxcrLWO+@)}2PY`%P3;(X$ky{n;Fj`g(>P6PTEvg@fP!d zf)!@bk;Ftv^FuT8vJzY#t#HXyM(Dq+d5D1pmEr~`^I0XE7Hl~zOJHzLqs=_Bg>Tw= zC0`D6@2A_ZHuyk|S0kcfvCGV>qub@G?1v`18PyMjlTN^i2swU0uVbZi$O7vBw_J>Ay@jN*6)5cs(d>QIb z4}!g$`8O$>4v(SAaFkt3G(otxTa(_v{{w?Sm-G2HU!1}>$;TeW9LmSmn4!?Hy~8mL z&1+DRw@2qtSN1x;6AmDop1vo1A~FAH*w>H_4-z{7da!J2~aLylTFNW6A>x|>9LT+48yW80M&2=%7C?bGUCWs z6XKncILPickDm9*sotH=hDo}^`d_=sGi=WpKN0{+zZ=D@BVL6UYk#wM-}3HMlXTUj)i!63`IvdYzpi9&%?ABMrLNb~PF!&$Mo~D#hqI zKMVTG`q5-Y5@qe^_84SM&F|)nHY!@j8OD$C+8+5AY97ODqd%LdEGSMop{EB0fPi z6MPw3za#UNO=?K*+43T{6Jcxt(+#VZ#-U;>v@iCDs@V4u`|0&7cA7WeEh~1TkI(Sw z$>@agZ@^5MBzz-|AzMNpZV72aA52O;o>i;SDJuNFWUy)OmO}W)tiv?nP9Fkqh1!j4 z8&o77ACXz9tx+@;nHG<#Z8+6}WuK-mv9y#|U%1{wO9yvC``CYC^b zByO5-al70s^ZZ65xTP;yq)5;3+3-iSiDyB;Z~(e@*v#oHALW+#yC9w-J&XGo^!$e2 z0Nb?_N=%m9WtE74R*yRKxFpG6bmkH|9Gf9~mEtyNyqF9$om$iFolT>Ef#hKF6BbRZ zi3D&#_>8z#d_p;7l>lgTBQd8!q_8#SU8|uC&qp(%XM8d|34vd1n^ilEywxN!rvd@Z zET9STATxYADGB&T*I52r0CqAO5S0B7m8(Gx>xTf;DDZiJt4xZ-Vt8E@fG2Y;_n4*bfz}Z(iM@Q1wQ>2KTWB zPG(V8cY2T zC&%`t14E?WCP#PH-|j4LV3V?1)8RXMxo%!#UJWV|TGb!8%9q|p<)Vs#*&Ir8GK)l} za9`SQIw{^YnIi&96J2n{&VH?l%zh@7Qaql1c@P3N2<1TE6{3dd1MU_ z$4Q+YWhRl|Z(v$*?=ZXI7-nl$!*sG1hI!h8A7LbkoR^r*?B;g9SztBW=m}MU!NzTQ zBkELN(YNFNW65hZ{V?GOL}i6kErTxImTOiLO0jgcXn=?)>V}1H2t) zkXSyh1zn(8Exoc7fCamR!6_~<1GG822_+Pi@$ivxgnZHj2(Uz1t+`1)j!w^#vx%D_ z!s6t*{b=2uLLfsz7~*OvzO-KZT%AqE3I$Gc=udhCdh?TY@S8rpyU>HS#LU9tIN&@{ z{c3re^CZXt8?V7%IQp42)~$x-=}c2?1Ki~>7qEM9fVlpB%}YaEJ}a2zu>h;x>zY5w zM7M&BDxyh=B$nAzGV_tl0HB7eb2D>ZW`2S&CB8p{t@vsNq@ACc&t~Rw-Cv(c3k0y< zv(a4nEXJ})Z^4$B&(ksHQW-iX(VS)NWX1; znr$}fF)4b!K0)5NZKPJ3r&evrAl_1)$@8N+}- zXE3#Fd48$2CZ)7|ZyMGVjMK%l)##Vv!{2IlWb|y$# zwJAi>9MnGZkQPt^#0ix4flRgo-C<@x%c!P+5D%TEsRqFmR@-m~peq+jbp15gZ9$fD zfhFyOBY~~Zv_fTR&0+eaVx26CysB0e${BZ08LUx3Nkpi`Mw;N89xHE*@0+cx(}BHo zL`iG*k3^luQnNE3u;@Q`J}A1Vif#__Ta30js%nm|nrkZNo9rgyv*BJkC&*hllnNgZ z#v*=I-OXbb3ExmDmMLP@(}on88&N0Bd?*=I`k-{RX-D3esm4UnAz%_xm8$0YRLakUO zOw}G+H7l!Ts_Ba_q@A=|hVRc+P>Oz7?Xw(@tJ*(SK!2Q%ubK<0G)s`--My0~gSN;&T2D*0zs-Pu+1WuynI~k6#VIw3SR6lA&p~R?FF(Lt= ztazvtZn}m$!+fLyVA*uC#=HsP4TsaHGRAW0Q;jP-1|+r)^P-`sb1?PHQI^!ng6SM3 zb+2ol+KxSg6p-t-Uld(eORK;FbKiT3Qp|}}|9pjxci<}Mn|v{S1^7Y0ZPZm~Rl!Wl zdsEH;)G3|rWe@-{21_5KEBVU!a^zK?XX%B-@p6vgPGU{%F=NoJJ+1173dzlEIv7z7 z9Srv$I(xU!*?WZRI2Yv%`2FlGu@3NsGwfiCck#mQ<>O^v+5>$+B|;~|wdflRkh%yO zxqAmKf=HNMJU!0}x3KiHX!NCu^0lUNk6T?aYbx(tU1>+1Nz_>=AB`{2)$Q}{bq!I< z_-CjKJ(WM7U)TKC)%bGo80_#sey;iVla~)-%t!M1r02yYYaZnv%hNZB9K%43@gOYOaC_wdd8%xpju}$?-W14>=#NQ$C?D>*mKMfAI=U z9#F?p^F+n{Zo&^-V$4r#00v*M=9=}sv0ml`pUI!NOfP>q|I$-LM)ne}e9N%q;7H(8 zU6oX3n{$lW!4Qvjo|$XT*7Y4B4s9JyE(N<8OrZi|2uA z_e>2)<{!g|3Wepq!gnS6vo33Uw+#5pDhmrVEpo7qt$|QMionsZRyOT#Yxap6G?6*B zZvIeK74kLi67dt((|cyjQNb%*S9acxb-v&hZP7xi7M#Nckrd=V>w5 zI3~fN+SJMth5#Y`6iO#Fp}P&LS!cSL? QK(V^4hbkVQ0&S@$YRJ54IkqyrZ?_l` zMfME>Z07d3`&vWMmeA6IZ!RPTH6JYy!eHE?7@Er~W7n+FeQ~F+?=bn&8G%U}iN)c0 zVwDc_aDpVBxYuTSGR=qhrCNgRln$OPgLfI2ven+K@oqJNeaE2*&GHg8cdu(+%WjSV zH0x^Nw=g=kiI&^CKh(^A?xwoEp+1hkIkREDD)ssM2!Xg!3*523lH z1(qT2()ZAmr2$wJJpck?4*bz{BzYdNmKLL2q60)XXbv%$Ni$OI7(IJZhwZHT*VtEwmIIyQ4I~2#&p@oWgY7m1SI#bA`Kc^H zz0)LkMrg--7BF%8-pzxOWj2v85l;ka0c$m&pd-yP1o0degdkIkN11^$*vK}545$sz zw-9l}s-k%`%m#^l0l5c^d6+A#WTDDJ%0q@H#-;Q=|CmXC)qDRN<6a*idLLq$nD*Qt z6Uk@D0}|`Pl`dmCY|7Li>EF5 zl0{K;EWJql(J)r=M@=@F?j3>o7_pNZX?t>OJH~vKnYU8<&RRGoD^V8B+BbTHdCn{n zSk@&ff`0dDe~8G#z@u3gm`O{4|sh=yZ^H;)Nqi=>jr_dbn$ zs2yVWh~GwMs2BKO3+79ceR6_`>enIjSq9~|ltEvdfQms4yzNF)1~C7+YLNXsN}A(M ze4MdIYvMUd-hWU%>u{vbDu{S!x}OB_^@EWI?X(7&Re->O%NzJD+{E=|o6Fn^98G(6 zkc~fT4F(d#NO1PTG2kHaP@Q%2;DNp_#Xs=&9aje$?!HOc8Jp3+>=WK6ZML7%8+Up`@*52nqZR!ouWjK?Vb8_;<)bug(F4gJbkL+8O0 zv}q`Ys@;YretKxAXWqc5b@;T}C;{)Jw{h*B|Gi`udIPtpBo25?#Tf?j#j zxL=z1aWTN5#P+GyEV4f{20vdj!Js4lF5`~V=HyeX|ETP?9q;`{`b)j>S`aP(JsAZ` z$4-%QTPv*tx73kr_t}M7ZUr?l>Q9~z(gf5mP*!TB-~}lgHCD{7u^ zi7QvDwi0tg=sE!6Mvx`(;ISKl>HCt6uO7QsSU1*9s4>7lJYnV`gZ9~U=(9Y9eHXNh zY(PuZleP|Qs&*pl;Hb&KYJ$#y&=Ys;2KIe4Y;BfQZ3jn<2HSgV;RvD)TO#qCRzvj@o;msjAIyC?#UTKt zg=5atOy-Hb15gPfz-x5ul{O#Q4m~m5&H`u6MT0iNPleR_XK1u*el~FFh@}OzZx^7Q zD1!Y_^N{E_(pQ0Tooz6m+L=S8)}QyuuJj3bX*PJNjztiC(r0HH+y~%I_q8kK62OIH zJsfM{c%lub87uuz)cmH4RwXrUsrGiicC#;CyEAG`ZR7lhrN4)*0hfasOx;PkoJ{4q9v02(RnNd>%n zFiswcLHzWix6M;ow#+4of4hALAeWi1C-!Tx`6^q2+$Hc%Q0Z`d2xs4K{sl+`_PhD9 zxhcVB*PJGZy*V*=CJJoaXkp;rR}q^>Y%rm55bqgAT(2s48=FPl-HEx$CZ84p<3sir ziTioDa^FhK^@+uI?pG3f5-X+oQ*57MuKK-)axmB6j#6<9bzqR9Xa_7o61&_@N%mmz zpTn;jH|@ftU7L~eO3hEja%;%nVr6>^mcniZw;^bskt3mct0SQrs#Xww0$v6zt!f13 zyFz5!q&B_j8r8Ol!fI=7NAjgwANp+qZkql*IpzX$5Tgfx$)_L0>5FpZ@JGJi^mG@TslT%Q~9RX@pWZ*Q>X5o=BqF)W}z zI%Br{K8>G2oW>M7H*=^oUH~XV?lUR5kEZeCA=jUpxhGTqJ1k?|$(ch1QiW_y+}W8s zOV?x5>sd^bE#_j$r!#X=W-dTC!KcpGr%uh1(=&CLKG2t#uV?0KnYku2-z<(AW^;Wu zHqX#xAY7x94>AWc$M{3QX}SI+k?@ElczYJ#mc>^H8+|6Tt1|Z`3|lPCc{;TZlwGfh zF-J2R@M2R|7=lTDQ#!w8hS@Q>$WBUjVR~qjW`Ao$15sskUQ0c(&73MNvx>lB5(`ig z_XVAFt&$?Tt>9$ZG;58t88_+-p>J1O@J0NqXoHa^PLRmr2Z!A?mAFpdP4Ti?a~&jXZhR z4Kx#TV;YUlb-azR(r*=x2Zm!k9J94{AZo5L(I0Uh!=@e9V)w~7y)ZUsRPD)C^KB%H z)M-g0MOA8P?Z>Ohd0Yu?<}(W_f=-IBt-8yr9;qen!mIw8 zs`*;gf4+)B0VR+o{U!9ctO4ix$*dBaGkjTvEIPsNNzA%~rNyOfk{e4J-$Mh}&D|EKcg5!0 zReN34JUwyyS5#@gwLj1^_UkNk@sF6mjN!G__!h+P%ncgU_p7SAzv}N(iyy1HN2}(c zs-GPQ52rOFP#$XSAkw}8tS18fVeM}CPTpnXt%XscwP_^-c&8Vwl>}5q4lB%l0<^}F zgVB%b70FKP&^kyUG2g5)zW33E@#DzX8E0qK){^_fHI)sNnueizTpi4FeBB>cw=3$| z?`!VfoRGJ;tO^Xv48Jvm0mMBnpZqkpOtVWEQ%hoEh6)Ou{uDz3;qi_@cvQpDuZ6g3 zo@SzF>WGR|4P#C~&_%UAOuSKHUaA{EuNnhOfv58JlPF|SVOKs~hxIzCZf>Ye2$nb2 z(sSz|j|=PWg1Y$>JZ}IkYyUlmtVt}VQ=nOij3>yV0$`MFkkifuBUt zB~MTUa+}m#Ur&&5j-}8oJ(ccIY4>PZ=|}4R;ktdGp1JJDb+@J-h*XQ6e@>h(^t(~> z0kJ+6T=&m&Z5jFo{b~U>Qw9d4<+FiDYZfVvZ%+g0P^RSNE$AogLZtC&FvJtN)BleMo|3t-J zELQ5=Chi-}=e@b9VUBBp)7Hm)cQ@E0aA_s^TE%>}Y0qhz&s5CS6?0X^K9!kYHhANd zroGfF`Q_%irWOd0IdFjMu#Y#BV<*m(s2psWU|?;Y^X8UHLO6AE+xnGxrfIHfnk$;< ziBH6qvFWdEB7As!y;OTGWrc}zXOpdy!${ah`ORm;M-@%LsV#H5l>HsMwrOr{)<(=f z__bwM+}ZH|0j}-iEtnTa0?br}p>muH?c4xDcmT5_xDYIWY%W-yDwS(#U zkVP-+HZj9)u!?knDKKnQRzH2vq$CNv>7f>A=Ej!4p=DRK!Z%yRH{s)u9qOAvEij78 z56r+<9hPDY9W*KiEm(Vn?@5g)W|C8%R5RgAuo{!;`)d(FV`=?g75h-d-0I^yD&|KO zoWZ)s2=EXV9fA3$Jg)c<3~b{O*da9=PBw|#$y~cAU0m&lxtiBl)Yf@Me^am3{Sv81rp zBky*t8BDG<@pp5tdX3fbbm3s+E(@3!x-n;9?hx}*7-c>O;(s@piOs04H8Y2

x+^Os2s;Ic z-Ma~G3Q3K%DJ!eUR+`Ol17Ph*=s+c|(>gxFdiRk&JIg-SHpleE$M>NG-G&-?^XoQ` zJkmzvPDO$B*SK2k`YgRs<}Jp4JBx41+{1mW@sIWSNBiupeSMxCRxbmSZlhI)r#fyq z+k+2w`#5xw<`pj3h)S@?YF7N>1e$&ewjgYQg(+g8u{kjc(9YIaB|Gu`Qq9PahtSE; z@u1Pt(-BBCnKN^EeqOQ=^ZG69Xrr$_AMs8rX~peYL!ZNr5GirT zU*_B_By)TZC^l2XBWT)W#OTG+LbM*rpKw)+&*2Mleuu#f{uwKySW=_8XgQ{X=NA6y^P7H;bNy@=r)&>!9)9?s!T)#D14Dlbbr0y58O~*L31kV!$4`WYxplk&|#lOxf2X*{L7fH z9pp=0M)=G8m34pvoI@lKM|$S4m)1%d{g<&hNup^CZ> zphjWKbk4AjR^w#-#x4(76ufPUN13eq8)HK4ux^B2|ZRW*OLj_xX- zf{U@ti?6BCUH8;j{n`#(b$;7?qHV55&tJZmpZ%3J%;j75n09(Rjwp~bCVj{sZQ77^ zi14TuW*~vK+s5{VQ0fbG3K?jE88Y~@%_094?PS{u58$Q-q&Y2Y|E_9n3w?xCj19l? zV@bW3ISpLSr|EyFo$dkc;9cb=hvU{O*+Db%TZ=fV4!6BqbPpI)(mccgG23-uD4N)( z***OpAT(JOA_5Tk=$5;ksp(E`dyd<=Z3xgg-7$fv{B$!E&oLh~Cjepq^E=%!#-Cu_ zhpjz9q6y*B)0{aDm^anh=^U^QG|ZIL{v%qoziYLzM8Sd_h0}b9jl<19(T&=^uID~x z?D-JhysGElORBf;_KVC%HGwR$s>9g;qrT!!XXtWdzC4~uTv^$JEd-VEg`-fY_!{LTgyV5W#?nVe&a13jYDn&#*_@ zcY^HS2?|4TiwHkR@o*;0Vt1l5^P=eA85GtKvr@=(ke|#L7cdfVOPGC00BW7)7q-mE z^7dWr*j{jWIKFi?8p4r*8k@NHHT$$*X1CeA)&vU0-Ph}@OmrINq{6!pK~47%lF#)Z z{v(O~tus%Ec>~%Z8ZUMBo7A6Ag0JnMn`Dmi_Eu-V@60`}`a^L_v3(2$v@s3yuDAj63y{_(j-`azw5_J1H34S2%Qgo+pupo|789VHp|;X@Zyo6 z(PJ181n3c|#;w`I?8#<)!qA4C?I?ijWHC8n&8&Hgd6!ftyfrnMl5C&Q69EgKf%;y` z*jaV~>xm?m{dQXV7G<+*k$wDCyV@Y+M`?&Shpuij6PsF6nT=J2c;c+|C%PD`lt?bbR%r9WlXx~b8~HKx|6f7nF#g0~P`S$4z(HW66zW+zBgDoAa`X7uB& z;LX$;22TKjDhCk=`9Kwh@_eb@v5R?nON;@wHd`>uUvBmiO3BG2#RyD^sE}ln=2qWl z_JR1F$J4K5BFqI-PC|$~+h4vRPP-m0AM5rYq~Rnt%{7tXX{{EP&zPl6ta%}l=}%-K zY^_Bb(wxq8tDvNeeSfjC?ana%0|5Skd|g|_9;CTuDm3X{*ij8bUWBy^DI8S(JN<1? zrKp)2hK>>Hp0N^Ab|ee_h$LV4^{LS*DKxb^I)MUaz$e3&bg$3EfV7Y)&(u&v(M*aa zMKTj15`GXpoYSdcN;Jb{czZZLgOIx(jm%syGJnp3!TI~wSnq2DRD+B#HD-J`w~+3O zKx))yGf&X#@oU%TM(5z1C_(Lk^|Izb%}ou<`iaR|ANHe4H8T<1$TkxA|69Xzs%q$$LO@b2zvWUv|+r;e|_>zYb_j&b_xDvE1 zx>jsAe~IWtYFV9uEN}*T6s_w(qaS97p{-d$61#wVq?U}XXiM-DFSCn}h3yTzx^JLw zkuqwg?SSFjWQd%(c9wmI9jwK&Q%I3%-e(y)X!tppRqPKX9Y|Ikg-L<6kP@z2Fv)05 ztBuO+CL$Ipm^KJ%XQA8bZJbF7jCNm%_3-qp)o3qlCGDxvlWBIUZJbW$g1VXH$b62s z$uXGI6}rI1R9Cw~!Y1lzXz9?D>5Fbh$c2!__#TFkO^%smAK*dW44956^2S@ALq z7%>OPZGp5El0ecnxPDl(^#tA-GFE6xM~rO0!psx}JVOV6$)F zYpw*Mrcc|7joQt@04x=HGH#F1lUb|EFKJKbX8`0ztByKF5(v{CjV`Rz*z*%mzPy7R zSFwAVtFc9e3Hr8l9XDn6dQ>3kb(#4}7GD{(>98U{c9nI`W6I{(BqIV7aXb7#d_$Jp z!dZ)|ROV*$WTE+k1~E1(DsUG^!6&#CmC6g5tGe%G?&i!~m-$wK@*fk}u}NSiGph;U z&El_(e5|Pl6n69n+6BoTF2U>|yeDo&fwMRsfppLCu8+-aV5I=>Gz^)YabjvRY}_!q(8G+xEc}z?PrD+HH$u`5XLU`o0|WlZK&)P zsZvGA7OlImsn(75!*tvFsAcbxHSaOg2vC^#L){xhX2-8*dL3q~sIgg`K{v;l_e_vOaVqgZ6Fyt%0s&Qzb+Q>VLn5c>~KNDhR#_Wr&?vzbg^Ug@utpA7y-yO{WX} zGT|4=558!l=w6}O$pF!|Gk=ZHroyxp0ib@hTrq zw~;d%O>bx{_@x20RhjG3S|{y6exDVSv>Av1K}1MY#(W+;lozG}vc2h>)0W*^wf5$K zah3NPfVbgYI*DJ}?tnz3G2De~fj+zyZBGG)JbyHLpoX3Ei8XrR_?r2M&L0L~M3|Wo zdqd4%QZt{eRj)%B&fhiCdcfO#6}X78aX$#kmaT$jM}re4sIB>kR5vll8%EZCww7S; zKEV@a6BnL<@HvTHLPZfkgyVGxp~bA(XeC(?uQSX}N*yw2fsdX%g$SO(d2R>Ijcwb5Y%&B?ZHob$cpm12*GtzI>-Su^QIr=MB|1ag`uBc<0@o3FmS##Ic(mz&%`v?jG z3A~pR%>Y%`?8OO)7*Xk-_HgLwo&0WWAO#A2KVCKJC=2I;zS>N~{pr|Ac5w32sjr)| zqfl4fFY@n(*@6yTBv}M(E7OU#O0ETZ2XTVbl6a8cXT=n3278D`2FxU9T}f>|_WOr| zzo&H?A^NgK7Ofw0x&c49JH~0%q0L9?ldDME8^fsX6rq>084gVmO6#z2q2x+;K}eTj zyNNuHwI1IPVr(!$0pD=t&1f{IOa1j+iCyphjlb)r=gcgM@PFxfxSKuxwefe|r^X6d z)tff)b#ECk_%tB@gGdIP$qi|G7sO|xv&B`u_}^bx;z_;F=Ou5L623pdcPD;w{rcrk z^*&o(m*=tX+6beniQnI_4&&b*D`SKA*Do(0pWe62>+(ESmW_S^fnEHK&&^9I^TyuL znDN*9Z0~h9hg28cDMGe za+y%CSCP>8`fIAMMo`0Cl?Xsu=-lJg(I#lzYvG9A9V6G?BZh#a%#e!#zk9R4(P{QI z4vVg28EPq-Fs^(MUJ8@i_LVVE{gt?L1)PETG8z|boRJjAeYkn4R6x*h#Vyu3oKuLM zU0#K|PeRB` z!6}+8{E3>*|8>EpRU2@p_*ScS2HWdHPc5UT*>hNBsdR0AWpvA-o1?UkoGV@Z|8 zRc)@(A1gJkkRE`noUl&6gZX_)b-LeCr-#|5|4+${8!HTV!+BvP)@#w${10h#@i)Yt zmT#!An<)$rYv*jNw9k~KDa?lchCrG5AI@KjzZta!?r4?v2}-k=lx!{k$Zc7FO^sf` zrlEdTHSd1ZX2tzsTgDJ@*4_^SZpexOG@NhHZ*h%;{3VMS$L1?##xWRQr4m%x^9qm%LvtMqucRK`Koz5)Td`7{ytB4?i|J!lz7sWE; z;eCe!*SmNnY)zXE2~Ck{&Q8%4COyEt%IxFz#bx0^?sb^qu@6Io79{GE3(?{}%#&c; z2o(tZ!StssknbFmL7qu5R@#ZvxU$ui>?5HoXK7bUdoMhe3VT#FkPHQ$r3Q<6xNVLC zP-{SpG?+nJ{@J0Cu4x`;S2xX?=2Iq|zLlR&Ah8oMzho|w_mntr=9P#fNsU4(^D3Nc z22p!YzC;fqbtDCqX0i9K#*N%@!@`V}hbaL}^4Rm^?&NMjyN`43nXGZ&p`TO|`d6@}mC%GSXyDBQQZ#P^9Nk z0r8p{^4sKh`%HZ$i91V5IOlteW|Fgtplp&!5>ROd9qt`3o^>GVc7M zC1y9qxOZRqQutOFM~HnXL?!&|-j_8X?AlcI|Ioxn2o-F@CS14r67|uHKa1QuLfg??A#xjBgp*CpF`ZR3vW~ z3aoBI*{Xv>+Ch}?DufHnU?2kaKb&p$edBJI+fw&EV{gM34=$mNyG4K`5U~D)paN zM!VVxnxHAy?$mHroD!nSD10sg2fUqdmY+pP3$FGESFZ|h%Xt$5Xs8G9I=m@=v+PW} z0BZ^Twud@eXy=9N8FYDrKTsv;ZZbXrr|`5w5aOr=bB7vq)DZ6uqNl*$LMUr8L#092 zDwrM`j7n2?2Y7uP9z_$+V5z}qgfx)aaTIV=am~(WoPyy(rTd&9giU(9BHRdr-Fq)g zBI#ZmxfOo(r~}x(n&wtw=qiKasjy!HBZT{!EEucpN<7JncO3M?JzrD5F*>1EW7nz` zR6*PxGNV1uf=H2ujZ*)zWdFGb#t?s-@Limwp+R_48O`&ez8I3+Stl2He0usRR15^-UQCBqRQjH zwcO?H@Acm4?(~{YCuCnC1QKFIKz0Nj6$vPgI_kKOdm4y{iU^1Zigthqh=_>DqKt_M zsED|XgQ!R{iu=NdjGzMfe}8pfcRC?3{^#@mr*rGxcekorb?VgF&x+~o;tLewfOh3a zrIltyV|n}P-S1(A>{x(rV7%%4D(972i>MXx=$Oj>2oc-y*b(m(j}n3l{r%gu@YRR z2LYr)5S3wrP#QKWZFq~d)p0Y(XGm+8N7-7(z04B|!F9_IXNOx7bhedH_HRd(SbtNr z4JJ*pVKltbuSA^#C=c-+fKJhkt>ID?15>LsU+1JuI=5Gdh~w!%#s@9 z$HxB1n4i%f)BjlF?M<>^IybfGyE6AjNxl9@xly(D)&2RgJuf!r$L4hT8~nXwA1j%U zH|!-1^U+3fQ3Es7JX+k94S!VwYcy=k)a9x}pa&c9sXRlyweQNiejedN3?8KcWS=9uz?&x;jr3hlE)Wzpag4wk(VD{O# zYx!+DiC<$)&EWVibH}|D5+Pi)uc%{n)4+_khA1OOSzx2ntR}uWG0f}qBDtk_P z2pbR*PrHM=h>2pW^-8t6sL~l-WRou%cdId!HyXV2b{*RcXYkuYxc{#h_ci0Djvq@c z(b8^Lu(@e}0D3_61OFm~D#(bN)~I<5=4O$s z@EQHc{`56~zo?DLFX#gckAMs1ld{>=CQN-ZL9LT{;JJA@x+TDuv>J8RCIyC((^8dX zg$X*A-yjoG)*VIJ?w~CL5ln>uH{4v+&2%ik16wWAK$?Ll$_0gDeio4=XZ7-`;&!87 z)cpyCyikRKHydK9)3)ZTbV-4DYSCCZVmZ34$WmMU+T;?5FaS*0u=iA3Zi%^ScO`104-;!DfnUACOiaT zXrH5%7X^*h0sdj9PB0+y&xcmDtL{z}sS3BT7G=YMzE>ql(Hc)mB`O7xVRAA^WgAWfIux$F0)%8;SWC}f zD~;LNbEt4A`lU_qb-g)pHznpv#@rI@bpcpw*3vRKvoD(XTKXbsfFa(PpW15dS2SpE zwr@x4jKBs3BLt#nfzzi#Dw#Xo*i)Kae2Eth)~n5HVah~;8S~{ zb$TDBlhKd>1@=Wu2)H@b5&_#Vn{;Xd5{-Qb)a(eFeJ*RSgmYoww7PtUG3Q-9^0cM* zWz-0O4DlXzE%dSpf#1g&(13wUJCVO;x0>kJzL}%#H&6i8dlUDqL>b$DiiLu?H!*i; zUbjdiZZ9PDRqSt+%eidY=n-4_firiD8Hd9P`2^^|A__polh4;SZsWgVi?L7U<0L6D zj-FMNMrGOvCdL4)7FEbPcrNZ>Ltmw?_&&cjs|KZp51I-I3bD%2RX`Thq5?< zb1D{o8e$xB1Gcbc7EqM3ofthF6TRLLz5YCJF4S6D^iMFMY8yl$+ocd1LWCltX89pA zr@kP(=rV6E_vXaxai9L(XOH>xqi`qi9hcE-DduAQImER#ouzJ4%}jonL_&fy+KRY7 z%YI2Q)wvLY?PSe>((|dRVBzSYx0=D;ao?ePX4nCHG$_;*1+%`=?dLaj;_Hu0BVnh2=U8-A%5vZj6-%d zK~U}NozatN_MOCiKf#rS39^nr_U3!20tgS^ieBJ0AlrO7DPLphJaE?c1j^b#KKf_~ z%hSBMl{4yM7&y9#oi{f*OS(Wh#hvTjN4!1I{@waJ6YvpBMR=4q+mqR#%t!sfEM8nw zzOUEkJ4O3pC;}YHEE?++bU?0%Ij=$^WK8F{D03I!QOTCTDShA*e^N2JK1+xh;SKIr zpU{X;Z*6QY1BCsh-d{q(aEx%EPV>ers&D#pSJ}Jlv6G z1g?b)PndJa0&z$DR^sjs^RFuFT6=eDznR9(*(bp;um1tlIw*U-PT=_D`?OI` z4L*7t^#N-|ZYrzUdKNQ@zdKMT2}v+I5`C;Qml#7Xvg%(doC}y-=}6>d=L`F(Hw}OL7+a&WsHNl81bh968*&3R}s*7 zs!dpsp%2WkB%{q=&pi8YXN`OCP_j@;$iDB$+^zDs*Jw%bFIy?4@7nW>aS%PTk+rjg z5J-C$D0Bk~J(>^A?NGWru_tpPX-W=X5RQ-K^ z;_Q-AkR22)Ngq(=E(N-2Au3E0KZ_2>;7pSV9N2)ebLlWVPKpg0R{o54qtK3>$}nU) zAwO6zh$IXn+zunq!aHa`b{>E-s7LQ(Jp;!y;z1~ATm%6@nN@9$*9C;*Czxj!ma<IKCzp*%e#G<#TU5OdORcb?!gn-y1iYBBqm%wl~Ykm_6^SHqJwp&WN+r?sb(l1blqZ(<*Z zr*_~lGaZM~)-MH%dC%uqmFL$0z3-Pd4Xe--#xG`8y%@<`C#v%o`o0SFmJo{E z4W4EX$$~){>n^Ml?t-#3C?Pe+Xsx9L%TywbY>AYzCBll9V30}#5iKD@F7=zC(%?dH zayDd!f#)$ZEWV&I`D<2XG@D%;0c+1PqYAbx%v+zO1R{x&h|INf%sl9I=rj6@YYB2< zq4d^;v$%U0S+QM)9j+F=#p%9~Vd^R7I(Q$SMkK_GvV~y1K~FY4Eo~pq{dUu>q)0la zoNX_&=am6ZCdxzUKS~NlasZTQt9E0v|+8 zMethhDF0aD7I((x58nP!w1F_;ljuZfIR9Alv?S0A00ZbUNzE^c$-PBZz3DSuS)CI5 zEN~v{;%tIhkb8&Nm!%T-DOzG@TupG62=jjz$&mLB0TdLh9o<10*HKet0K+P#Wm)}R zM+E^54F2<_F_iey7?TUcMMEm)saVF-`ujgy*Lbl5>j@=SSd+jCvoJv5PTMe)WfCXR zyky+kCyaU0)W{S1xT%rB^KYg`cF)I5<$nxbGsuXV^dvJ!xs&+EY=YjTfso^SIVWpS zkrszBO+;0O+Tri%a!PPDC1{bz1DJQ{55SJ>H6}{fV7sUx%PKntCYxP6CrAWAa*mhHXTa#6f;tsKgMqZScc{%RvDqXHwBQjEqQWfQip4 zn?XzPGtu%^4!F!2qT!}txG%oLw1 zUKeTF3D5AkdR3WEhTW=Q;n+x;TWfSL_=u2hl>FhlNIc`}*vE~1w|SQxH)f+f-I&vi z-DJ!e#-17Mc+VpKiGhJN?=$vnjph7c>`RbBW*=^a=NM0%Mzp-h>^1oD2W|GSN&aNA z-5x@lFSJ58GBKj4MBjC5Gjf> zmPZ4&UmhVTL*z&7FwQrF;?a_^Y<4m?n?s67h4f5Fr-XWe79oHYGLZU;!DWs!$|;QC zbs)5GJ&q9s?K-}D+H*-g+7al+02wlh&m58?O~-BK^;;$BPq#|wTBk+-7LlkJY-I4& zgH>k0CHK<$guJ!LRD+!gAcpb203I_;Rwuv!^eA{qT5$)e1X|7_d;mb?`OrhV@xpMG zB$g=;Vg&bYoj2y+L>49MLxkPlY2}QGq#PwJwCG|JU5D71CY|ODHklwP$RsMZytcXd zp8jd%qZTo76C3mD<-IomJ6Q-@3;R^WrcfpZcF9EmG^TfuyW%4JaXTHc!-_l6k1!`4 z`kcMbO97^Uv^ToXBUoV+SYOmL#L~bMql=J^#Mkh@Q_$nb=`}2Lt@*oDyJtI5Ar*6u zC=m8I%AN*+Y|eKE8mD3|kh>`R3x&?`?yQ)PIr9PSH@i|FzBsAt33D`O9<|&bn@`pG zxP^ygp_%W&$_WRmF*-tt5B4@w5p%Li+Xk*n>9*p2TZ?8}@ts?XBt86wiEg8tSZJ&{ z78MztJ?+oiV}OTVBtR>$@;io00q^3&gp&wu;$t%s0pM1kup;7QdI?z%qzlT;00_y} zdWiQcYJ75i({Q~K@EL;k=+B4ka^q(x>TCf8l%|73WDXHo7g|jwm)LHnrBAbhp(I;n zSQx7aNDxOp=%N?rZ59m`1T^{v32L-W{|n}H^34f<(BJ=uqo6WEZ4~qy=s2d=A+U%d ztyJ5cKIEh3BqTVbd=ryp-er8`lj{%%R*&{anWKr$rT}ZZ{#rg?1PkjcF?e9b&zC!T zJ}i6xhdpniJz!4o(%Yz=#V~IQ+K^KuCf;lg3i6~ll$f0rqpz5x2LWD}ks7EAfjQ2g zz9$_CM@R^yYfcNSM*(G^@(^a#I1L9ehr(LHu_@!-kR3!s0Sd5ogQ^wkGc@52Y zm3b-6c$N8e?9gif;{G-1TXy|oFd?BJ>wXsd@5Sc((ouacPWqB6!h6pBBsTX;fAtf5 zAxvjBBkl=PfpD9QaAF4obE_%c-~*sm`mFp94h>qX<|jz@6grQ6TabY@TuWd&Td+pT zQi-uQ%UBTUAl8?ptYWzoT4Z{Hi8wCc8X3k2j0NRH6v=K{(MbR)g|pl2-agTEL@Sa9 zG=Qc7JPSuvj;^uKj&euqw?8W3w>`R*-_@gvtTbt`gahGO|5buri+44lEZoMrEiZXi zjhig1HE|p4o097>ceYvZPrb#O@XLS6CUC#633t3Kxy#yb+RAas7UwP}Lv!gOsjP)L ztT6X1tNe+)HT#g zh)Xg!55h~TG=5fdu%S)D=++tXCCtSh+t3IrSISBxn(Ju9xShRIs?iyt%=M9ZNf6{R zl5klcXvLv^NoAnvA>pfuq(GzzD9_^8T(()Ry*VXLrMW=Vuz$7!fX6%Rk=rfKsvV)U z+tGjf=Piim_ea-af-cL0SfYNx4Sp~v=)Ym(uObOYVl4CYZr9ozidW^sP@)-~8fE@i z5cn(Tj_1)v!Xw4o?E0;ih}k3}10rJr0sC@R;db1UlC0un&8vYuH18R`jereodRrO& zE?r_)N+XkPEJxR40Glik_muwuSo+8PnC9_FrfE2-6keCVa;b08qEfN_2DMo_p3H?U z8&8&p5-%Pwz!3S%=Nwr&Vg^a#u*WMR=(Ne7l59spyeO$efhEC|=zoUvGR-@L5A@0yn1qn7TFo4e^mPltI0RVCLG6KqDbwwMFx}ch*&52iv@`kf zP&m_=R8OvOmv0X%b#QeO!i_-#A2G8*_VZYV)76`Psw#n3k18|Rm5ct zW9AaEJ=JfeP760!DipDhH=j?l8`9DyQ(OsJ9D;LD2kRf)1}|U;N;S3y+bB#H^fO1Z z?695h@~(IGe?&7gM`z&ce;b-@(O!y#yA*FnqPV)ZMbTSn4{f*U4nMrzn;rh(P}V}J ztiwC}(op*C9(N2vx7=1Bggc-jdQf#1)}o1$6WmgnrfWeEcH=Uh_`oZcIi!R>EIugQ zt3-8u**eQ=25wg31~?d|Ch!WNC-I3vxRQ?~TD9DU5PGT8EMh28_DcFf{q_!U%!{e% zje@1N)jM0wc-yr$?BY-!8_HrROQ9SNWhIm!HDT8g%40$~7|I!;Tp!A=P}V|uP5wY% z_^~RKOG25S$=~|oaHc1ec`NdU-J3t0pK655FPnA~^|zAndbEth#mu_TtSzm%#EnY+ zVrE@oN8PMhoO#hG-5vT>!ynIj)u^APYopEmcSPtte-ykb=;MYLj3mPwM@0{4F!G1q z9^F$cUe;g!M8Db6Z$93ShEm};6pM~ot2>b+lfJ-qROn-#niDIeHy-f>q&YBuSJIU% zg6~b-9P^Krm(K^F2^RR6Pu7Ru2ar0`-S5o4*bOX(*ddoPE+!-48j*2o z1L3F}kzH3fYCsV$C;O#M(_}+fD+urd8_)X7bzB_c9^VV~7vF0vL_e4o!b2JEh~y!x zN1O<(IeCzL+9)22&|BEK6_<16Y-j(|#BT$36)TW`g7<8LIg#JQ@ROYw0r80C2kh5~ zSGJBBM{qU9XBD3{A7XqJ=||v8*t(b<%2hiX6~wn}`ZH-We-mOLKgV3Etn@xD{<>hd zBgq72wDMeP%r&y^?f*HRp(?n%w}QVrE5 zRniA4#zckk(Ujc3JJ82M!)L_FV>yqB#Qfi>c?VMY)zTp)7mmc=Ttq(~NW2Sg4l~a- z2eZZgYH%LQkpNZ8>KVp3n_U0MxrWsY6UaWy_rKT{P>wNKT*fk!P+|N2Ue82Q9 z!cm0b!Y)v1Yi#sx8}ZZIj6HqqpDZE$z&MdUf_~0JNYZEBsZIb&e6XSn`68YoKSl6B zOUlWxc>`Rp&n^vvRr*V6{y^fm)DF?h!{&^FZ==o3!_woK`Jc@HC9|VknN2bVOq%`> z6yNq6v+eZ8Om*ORfjL`9t_1z!-0-k7zj5xjE?pfa!g90IPrdvxZ~o-n!#-UbuI-1O zWpRj$pJUvA0Rdbm@|P?`$qU%a+MBQ;tmZcP1--)xsxpBjmP$y8SCv8*%4X__+qpyY zrXKEXhq%%mCv(SjGIUGWODKy}Ph z9ZT0-)=5T>Xa+^6o00S&bXJ6g zQoImt7n}>re1%oJ0=AGp{|_yut3FsfnMO54(lf2|J9^p+9Yti{Y5Ug@V-dfm(hK(| zNj{+*OZ|l*?16`eMw3|sW8zAwSb`(fI%oc(u7xqO_AV^AEmHZ>Wqzq$=9Yd(!pGJzhV&sZ)M|=l45{!*xkAj zFMev9d(l>f+)6{)8%n#?wc%Pi3U&cj^TiR;X|ylhl>qZ^Pz&hb1?qh3=J$p@GDFrhMj)UUv&& z1e@Vf-bkxG%R?}KUBWEs0>nF+#9S{C>U1qp&Iw5Y{@Z;VKrXZzk#k&*EBg~m6A}!V zX^_!XDOJYpZ8k72!AN;^jiBxu5!PlQRxYtttB86AO#i11;)SA!w)+~%e>5Pup$d`l z@=tLhu~o3Iy(FhdvOA5r!|rwSf;Gb(DvDTU=?8ZRX*i;TgN7(WL*&<#qE)@o;S%);kH9k2c%T%$09I6e)EfGt zC0ey8+CT^<W);6;!2nG;Z)uL60L>KXh2;h*5 zN8Cst7KDtUkH8gPCUw1>XNt*EkP7U9(KSz75!l1NG4Rc#?yg6x4hqHxS?Pf2R+F}& z&!m@V7m|5R#;y;Uc#AQY%M23MJxp2Y`uG5mXDouEU4vmHUZb>Wgp-NgpKfDZYOF48 z7WD+I5OIptxEQrc6djb)mqy0$Yg^Y8qE!RYSv=Pwvra6*s2?*NzhQ{(wI?xr%OouG zT3*}fIt^YJp3$5aROntp1qu*6kuz4+qbVQ`Kk{HviON@ao!i zk0WAa5Cef_T?*KFB(xzh+iNRUC}hx80dD$fSYXuwu$@4>t#xZQ&w5Hw`__(H8PyG9TY z{UWKnFU$(_(OFSR-CBBs$qcptVNBb@nc=6Nr%qOL5;f0QCXb%7Lwz?M=!zCGxg$~GUf&%i_;owK5ER_#(vzGYq>3~i~ub&kTj%-QI?TiW~Dlh zkilFf%#HMT&o&aVDplczPzR!ty&x1#&Vba*^nWP@S=2! zvA@+50X7Z(pN|;}MgFkyf8sRr;J0X6no_>S^ed95h0uPSp;viMd07A~J^_y)(okt+ z7!$ME;>_NNRuSs=`yNTh85VibZ;kzHUT24S(70bHHy!LGz`K`j#8jRx^sp|*{K(iJ z1Hk6JddnY;^!Od-muev2@JFO&8Y1I{p5bjjHs%+4)zll!eSBF{BJ z!RWu~bPUISspYEq$BXnYvwt^4?dWbQG}dwlb7(R0M9x&;2L?eUj95o>A38eoC2Ma2 zw&{KeX7dblU16CZS6V6h83tH{WY#)?NW4&E#gam`1=;gsPr0O_dUY78+JtS5R@Krt z11Fz@pr25Sn{509Uu=#aJps^rew|!k9WN`y^+pq<-<>art$v6EbscGd>cpN2T$2(O zT4of{N1aaT5S`0XbMjV;h_O&y71(tA$)j=~ruBZs69!A8k_4 zbBtzc@|>Vme1K?iaw4ULLp6ZArE4X7{jp31W|rn-W(=VSGF&@qtyU0TZzo zsLI|OOLzvOleo@xJ~3r0048dVkPt>zMZATBXn|jFKAoi#nU@KXWgG#ZSOG{43Rz?O zYs~R~-EQc}+Wq&BEp{#^*K`48tl}R|x-!T1cg>e4BP<|q> zGQTN*MSlF6a5=wqW;h<&UYHrJDnjQMi37(1AmuqR;OuT&ncupmtWiXvZypowj<|4> z=e4i0>G@e9IGqbzVteyhU8jSan;~Uv%=_Z{T(&N@8)UPxo*M$w!An-a2rneioPycC zg5#5y$17Gw7pBD@54j%=QN}+UG7k)yuaB806F|wHGh{wEgf!iovxdy0vvu}^7+jgh z2b;f6({lgVxnRWp zZrFUuW!DebTL%24#$GYxFCQ{j580m&nrGXC?Xy{9HV?V44Vt?K?KcO3FAP4-c&mM) z1hIHFIIj^6Mh4 z(-+swCE92gWgM}sPfK|B@t=PeVx7%6;hoRS0C(cLM=#ppJolNc<+|nS92oJ+%^49c029gBA2M?vLdx+{j)e$ zDBV7ePvN+iVRU;|g*~ zQfNCqg063%ENe}$mV`*d0;(W`7_6YGSV1}baw<~i18`n!kX z!NN#99FJO*5G_bX;#mdEZoB5jb8aAj{66a!Q7>@<0O-J~uRyl6a7h}DlNAZJZN=qX z#4}&EuU~l}2vYWQsEroI3(KHUUl;{JZcF#unwo9ts!%Qn<;qaDhVr;+WicE#LRkrA zS131zvM-cAlrIXU+e)&eRb3#gJlHR82TcUkx!jEkjTypNmWiP$RF#FhV(!rI3fIWl z9_~TF3woSHU$(xGaGE(PL&VFq4D{6k=0X@1P}w(iQZDXxc^~pGGb*eHhQ%*yZghG9 z(}w>{C1n3Zi+ymw{d~aucEJ2*0F}qz8s^5T`N@F0qQ!g)Sw}hX*C+0>8SZ-n=Dq>* z!vTadk5tMJRgzxwgNnJog%IKo^&7y-@2L3uTl_5pLQq==;s+}BrAv_D z6ibI&sM-$|viBDX1H9T}I8`1$uIp>LO2Ilw| zt#nZ|2o`9==dTQ%T!il=emRn!Dwfl8%>ugy^#Vj!sf@YyQd4U=6j^h>?Q5@gE$Xfy zpr#E)|2GKSE+J}37c|W=EdePMgc3TL=+ZDT_^6KKMk-1P1L+3VDAde!kD2P!$hnH> zDMIGq?1xNB>gS>iA;~U|cP~c^I}(J;UdGk!5Ym0wK-}Ll^p)9Acuyxi{p#!y zuoei6Xa?|y3|f}kFw&v?JcUQ;iV;MigaNgM5OmDHQgRP&&>bRi*&wIQ@sffJA^}lu z0QwEe9%&r;cd$ys-eOyx5IWv2vkMF^D3>ci6g9QM}SO}*|>XC8ATTOe7NEmo0k;&sPZ)zS4V<`}n6#3)Db^nkjWxy-_6 zl#9gqmzq~FTy^wj|6*6$IT}i~eer9*R$`wcZ4~X7FC6B4V@Fw{i1AH->>#Spc3E1To$rRWleX{>lCqu)c`y`g z`3&DlN3x)x=!A@=@rlVaEc9t*25)skp=IsVoeW|)K6O8q$GYFQg}YLMC6G}zA-*hO z2dIap+tBJd`c8}Gf$Zqe5ArPtNWif&tS{Tyd=Z1T*1yFbXpX`!F@#2*2^UBFJAJ~1 zdD~l+#Qc!kO>}=~J3}v>8HwSazFp`O)h1|pr0Ba^BY|Sxf)`LXoZ>6V`qIdQgU__5 z6HF&C=?6}TNuM7|7s@M1W78TPa-Ed`3M(#<5jcxUk+m>Kq5a9vMkMw%2IxUwlLAQU zpB9uIi%_P??t7SXOTZkorX}gwpI2y!SOLK#=MIU)TexQcUI^bdM?l)1Ic5dt2L$GT zV=3uOD@~o6PlXDGf(qx_VH1Ipm)N!FGDk5(FG)ouP}yRmL*7ZZZZoe4 zOmTgy>5LA!K;FGD&kcgU1VY|63RtKTR)>-nwHB}vKmh;@y3Tq%B+rq|%UGpxqz4E6 zT7ajf#Lc+y`IDV;oIzbc3_~-C9D`n>k4(^l)lnpiHBG6TN!yMnVBeEkqC-fy06@VF z@*zTg`$s>MpidJov9FI`OCP-6z5!{OjH4{Pr|vcwpd;|%CO@k^u-*il zoOU+FNF%gZa@>#~Q(!#99*Vewt1?V`U3mRF!|QX63?38!5VEbgTu1jFSCC|ZU~19{ zvLR8IQjNxGKNI6mM4LcHoR7KW$=+@V71n4PT>MHqmxVIZ_#z#%ACW3!n`wl>eg0O} zhQy*nih1A^Wn`al3Ex6rGBl$|4Ffb;sjTh3;mtvn8b!;9#axoW%8#2FlBxNPg=kEo zZ90PlXXh`1k@-^(`Ym(yKIQ`JNZ#@>YaVT$aX;;LkM$rYK{OPkrPdBoBlx@rfRMCP zBLT((-^iPRV~LYh-)7S{Ap=S;$LcYxyhDw-YHJ|$rkC%=nAmo;r8Bf8`BRVk2ik%a z-XINNC%xKJ8-QC9AKaUNFqKGBysHvE*#BddNN2Y9w{J6jVR+D6Bvgjn(X5Tvl-8zr z1@3DVQEMej!b)_M#ltYkJIoyF4s(Y#Ax>_=Ot@BL?*KH=DX^qpv7C*-$1yu!Iyye8 zc{`TD-0n*+iBHsmFcn;(q=*r=41sR;&8RM3*@;T@CH_S7lI%pcAwH=&#|Nkm(McQ3 zIe59?7Y0vy>cq5?%Jk4%JQ?P*H@mmS|EB3YZd%lLb|Y0Z$DLJBNiYcTAHXRDzqKdD z8)!j^4>kK%^syxZ`a_UW@=>g|2J)==6*I$E@B-T+B!`VK5$nS0(FSII@pV$sOT(J1gP1|`3pBAu()AzZJ#pn@4ZR%P4*y(?fRp7YT zlj0F{R`R{7I^JyC!;(81$`hw8!pBZK9+?0I`sXcUk>G^^Lz@EXq_EVQBDt7$ztXYC z@5wt0Pr0+YthO{J;9Viwf;^@$lYXhim6$y0VH@E0ppG=h-47N(Z$2H3`=bROx09;| zM5)?O_Zfii;TZy7Xot`BnP+_VKNcVz#l1Mxa_2ME@~s6PSC!lHTE5t4uC?au9@WC5 zU+y!{`0O2dEgvL8$3XQneer}d*ALXb+{bVTv&e(*gQ83s*G|8`eGcB{2b+Zwzssuj zNy+$0cZsdB5IQyPqRC;<2w4ItLy84cf6Db*2U3+ip6T>MJcE@ml8t`C7w&m|R+a13 z{+Kt{3IeZfW>vtZxzobZt>8637HAVv6@-26L{N*cCJcZVi_Ge=N!AMnBhn@6s56~B zhJLSA@;7KZeJ|uG!{r{c{?FDtA}6drYruBej9`8>-+;Oj$SnZtB)d5dV38vz9Vj#|W+y@YwR1relm0y_W=1g9I z!I0^RJEpQ&wduP`yo$DJeKhF_6HD zl@!#;VUh&83;DwLa`{4WyG63#rj@1b5KuUq<}cUxV3D4`--(Z zx~~Wz!gmN9v}iKkEFADZxBwo@d&>5?fW{$w59(kpO@$rSFInmnf{(KhyO$MKTr`I4 z=jm3ei^u#0W6!XXaUCI(TH|N=Uh>*b2GqnJ#6ByG7-37$@ykSK46)Y9n8YF@BWwdS z8~!!uwxfui&?o4t39lw-K*XN+1>csRpISjuzn4`*E@W$$1Pc|jHf`gy zJ6%&@Em%@o+`eRZv0KuyxVj`;JQHKcM)YeU44_*LYZa1tY@JOT(F_wbALx<;{`-3g zMuhHr0ExBOS0bIX`nQ_Tl#{awK)_Jc&V(;5UZhxNpD*K9*vhG^%Sib_6Y+Z#0G-$4 zBY8hI+8JC7btDWDAcsJJKnM2}{c|CUHT=hij*voVGV%2Oyp@hT&-exMfgVGNgir+s zM>)rpurJV5YTzZCh3L8BS|(#tB@<$f>NB$LHnx2ge&3qw;J*Qfvi~VYi9DVC+-1LX z<*zw&m+Seub2qx=7R4r2Yltev&ADt-Vy7gqouJMZY-@-&mUFj9E6SUkf-@vB8RkRe z&*pnsp+Z|;G$9s_wiKhWMzbyZM_17nd&S<{a*GT#d;ZSiGubymoB^qm?-5qR)NB5ib zksMa1dQcpDvlnq8INQy6pMzdV{w`KWy_iSB?g(Rw{G7%qFdv=QMnqj479=}Qxi_W# z8ziAZZ$$e$nteGZ^d*gGi>_q^m@|{ZBddEC-8^+tm~k-2jHPI7S#%eZjyrq?m0Fua z56wWRqmSFr8;rzcC6Y0erpqF5~V137PLI6g!qyaSnQfyC~$q_fumYG#x)z ziF6M)QWjY_yyib-cPy{cLGVUd?_6t=?J`MJ`}PU#%R5(+9=xXMXu6nVtX4V3A&%~b zn`uLIJk!XquEQ-R-b5-fGM0oiPL7HJwem`tt4#KN2ftg*dSF>L#GA4AnXt$L>!Gxv zoE^%+P!3OE!Z&t8vl}+B$GQD_L|0O0h$6wN&STzB><5J1n1Esklc1d=Gce*yaroQ? z;%+0X#4?`Q4|%|zef4avAK7t)IkNS+-gdKqV)@jgQ9*)Z?UgPQ@#@wIt;#fnwYBEn z&M#p&915{sfIe{zOz36(N}_sr{(}%4){_1}iy^Y+J+VC^z-tG`IbB=8XgpwmaDF%( z&j`n(j2PWpvTKme&xHT2AbqTwzB!7X3;LPf&B2|3ZXUyJ z_T5~AHT467(0TI$tyhdzY>0>kaf?ZC1^Y4*hZ;dx1F^*tExLgudq~E=tt;Pvc!hw7 zeP>`=CgFExXs8Q~R+t`O8H$QlG$6zgL4wwiF(*k&EDra(w&gpzh3mnPPeynQj3&-# zpdgB5dx6HvDx^Avj1|Ayz{*XeYA~Vc!(p16MQ)T&N_pcu%{$_EW^AQ*WQrTI)0oq3 ze45QD|IENJR)#fFB34;(fg7|+&ImC-xFnE7lpoeak<1op_&;wq$n%S#tUje=g5ajt zH-?LgAu`Vr;NXSkG-EkI59KFjyXn7nyF3=|J~_G}P1+h+@;hg~V;m9Ye0*18{yiq0 zg8h{-KZxyr;AIkG&-wVC#C(;Yvc}%)9Om_>S%swYG2#5~i0vlq6)aBI_A@B{n^%Zi z_gQB@51j;aSMw#buKX3!_NEX*-o<=&0@3aXTaNQK!fp1)Aj$=#@F+n>Rze_Ef|trS z!X$@vz-3MjI*Yh7TgXsaO$y!M+OCeEGfQMfVei0mw>DLv0ccUc7TKVKK-PeU# zU7SFM?&mWGv!i<4^&iis{KWp{W@*+rPcs%&K^aJ0i8C8t%e;*Ih#M|( z1Pp{G1R6!U0!U=|`(v)asv4$FQ}aNjHV$I@qre*Gu}DeFa`!|AiD((;ulC5BN~ z0(-!D&C1G`m!onLE|Aa~Sgqh$f*pviDX7PZVYhoj2vSA*&lp$RbhAHf0@D~@4&@6% zSqi0_K!BBX(R3tZyjt@@!jgP{LUS)!o6heOy&oGHvAnD;MC(S$Y95O7{qOK3!Zgf* zFha~`KP_f6tOQ1{NM&U#$e{)kQcTH!SbH8+C_?vuaah_-9rFYe$z})mJXAR!PZPc> zQ&pCbI-yA96f{$>2#`S)K#BaXGyILfsKY^I=vn@h{WFk$L{OJX7lBs>|HyU%=!04a z3lK#AC5!?6hS?{@_|Y_{F|Mz(%h6SUhWemL((Q|F4h~lp>`u(YS^8kf^AXYl$TCPj z@TCELMIFrv!6`*rNM6FoLgsg!m`2wWlRv;pFke1~1vWE5f0r&rs>r7RNsK1{K!g1K zR!k2xLH;nv>YT|I{WWIV=Pf#?V4sNXlZ=B}B62AV?7|fH`^$4b+v3-21uIKd}mqw2jlJn?;u5BgRB z=qU5I@bHsGJ<1Co6Yk4H>3_IXy`Z$$2;_T@OMV)ro4&J28lEF=5x@XmWFZQ9VGtt=nXt?90R~k=p}i$eCBe1Q zdaV}Aa3aK+!jr#?J6KNYoz*TPkk^8`#U30$fQ26rm`{+(S<#nBAcI>EuVqewLI8-5 z;ANGaUua-(mgLcS5ZN=2F$klbg{FOpxQ9S*wSaPnh>%Bga`rf_HARWE((&$ncXpZa zZa1x*ACB>O*M;SwOr~AVZynj$g^FSL*zz&ivGFnGW2?uw=CSA4b6j9+J$B@n#C_$o#u{ z=siRxNzT^Y+jaMkz5n4;@4QcU7R^U>FWJWvGoho+ zi_W47a~--|cE#+kd){VlO#BzYoZNVWB1e)8-u%1Xdb`>JBTk0pg8yuS3{-xZoRVS| zr)~Yl3}V`u2W7J$_@FA=2W-g?U@n_ds=Z%y94Wt~kjEGWua6a^(Q9RJ+gLqQ%06Zm z_8lBA8RSLuCK`h%8fFz|p@LV39F?95!ECvZ;e;eLSU6MiWCba-m0Me9;AS9QwJEnZ z1}2Xs$j*8?hRTCIy^XlFT@?0avuZ+&Il*ZtI~AhF_b{d<#MIoFw)CAFKE5aA~ za_umOZx>-Zyd9~6&yFo1+%m@`ul28qk2Sq<50OP)+jz|(*xsi_GeEQ<<{bhT>;G2u zk1Ft($KK|_YW#;Pu_8Cs8o!g13o|=#M;J<7Jh!ejPHNFLwXqel12VVbhr=6N zH+D?bNmK=0Qa~aSyjT3}a7SrlT}cPR!x{%#g{dzQaftxefmntyNk~hDY_85IG=Dev zEhu@nbnh%gVzl01phl2-p=iynF-6!gdNKT^)vfAKQ-xtv$1+LM>J7H-Ta%u-aP$bn z#EnZFLY#r|#2!dyYaT0H$g&_s z6(!~Xn7^3pQe(Fmd$}=J8!SSuq%|#9aFy=`tgQK#cXxYA|7(lQ%5AFeh~nXtryKh< zYj3yaYcwUf&ALCkvUw z>>w`}h7kkHx^LrN)0Z&8Q?{ZG2ijpGu*c_ykN*ncahV0kf!({} zag$t)dQD6s1I8ePzbkcwJXt=m{ygj7PZT`s3D5g2!obKrI(xvRKQ-BxOuE25D=-8M z9&b>H0EG%*hKIy(tJp81y-cVYv}g->tmxNq)IS+A^|L|pLif0;Iw@@s>X~5cysQs} zwXA5N_27!Z!nL9CL0slb!GM8E_&9YXF8f-nOYIE@m15p4|L0jsKZ=VP9i zkJHNLY|vbu!B_%pZ6GxMbcMHtKh@_;c}AwupihK?-c?Vtt}VozmS`u4c<`ayqqe+G zvU!P;rIQ#z%Ls-#JM%h;1;#bry{lH$TvN@P&F}R&F>jgP%+_fHcLhzg)fi3W#?l=R7BdITW6jsF<3M_6*|wLl`{#i^;X96jg~tb?(< zQmC@msfHgTxS?II<=F3V=GEH&%$Zkd|Df}&(>_l!f}#8q<<3P28Khy5OOioRC97u& zpb?^0xo5x6$1z=z}n>f~o*n zv!ly=HstzQK!7>ED-1K$%;US8NNhS-Z^7<2fH1@A2Llb9PfN8=Liu4f(klybjlef! zXCVNveR))N=T40xT9KnF_5u29tRNt#6P2OQFIOA$(VEuEyY^@$ast|_Hr|dFQDZc{ zm3QyiN;Rb}b?RhOs0Dn$s+_k#Eorvmgq?(Pz(jLePxXyP4%$ zF8amW=c-kt5PSBnez}Mlm5&6qsa>jEz&lkr@1EWxsyE)@4&DwBdZK~IlsmwV>w|n{ zgodKq;&i6%DQosxdlJaN{ve=)yu6WyFuJD-NqhxgNf{Yom=YO= zGECh}vkHVXayU&h=@XBV`b5p+2?_q`T8UYpNnqx92!=jPJP1ux4G@}S*)B&_g*bkW zlsj}+c1mj8Ot`Lt%zFDQb8zx(J12e}Z+MQMLBe+upisdfzY{|t%RPtTIRD_*vNayEBW#<`;f_QH|ac( zw0Dn&lu0S>*GRyU_U4Bzr2^?P`IIW&7sWg%;zAk1!B`-~L9S2a6_PUganp*8%6G$R zOJDG5YAl%?@&#v#K*=Ggg8K&2k#tA=y6rJi-+}EIgjVOkRhKa~K7_3|I^D~$o7MXu zs#+H-r~*nUW(4MDUttm8?OUC@6U>G&ri6bYh!Cn&VVywd43~cMag@d7m}C6Jxot4n z*rKD>#Yt)x-j-lkc>8d+;k<>d{ymbfLkD{t9>~7|;y#28kSDqm^@nx@^!`|`g%?xdCB&isws#e4R8XWj;_;kN)* z<}2Q8^@ZD18cGOc)_e$CfIs|W@_!Kw80IOS@L_$C0mx&}cK3q3p@Ibzik!yiA+y3b z)NDBLBom!Zi<&JUj1fK)&6;DhVx0H(Qp9cMir8Ek7p_)mC@t9r&Gl%IIGU_KiiY-5 z4=OvefvD#$`iflRCcCf$yV~{UB|Odmtx&CQo|(n-d0UiRR*pw`Th?Rm!O_&!UgSLg z=vCmN`RDED5Jj3B$Q4#7+^o`2T60UmYy*Al`VywxEB5jw5Q2c;n6CiDDhU$TxFz*$G!hxdrF z&>fQ<_L4pC3^2Fboz1>`%AIW}41vbqg=4K0(}8++fV@6>~?W@C}uQ zk_fiSskomv-y6hrMD=oXp`0-EhvPv0=K+w-=JgY#+xv?c zjotnx!Mh3NO<+?$HnK`;=>Tdw&sa?N_GDy;$>2E+p&syvJ>m$6x z7$H#TI0Hy_+@M7pH*pGMtVxHFip97;-Mr5gc48m1d7tXUKIN#4jh3Qn;L6*R9g3D& zI6U9dWJZF8Dq9Po46AHqRF?@Exi;JETgKmG?6*Zr4$UTc)rZYX$64h`aK0MZpf|@a(5p}>yUOiuVvz~>~6j44g z>#n{ZStq~OODVw}7n{~gjGse@J8O;(_nc>1FEBCYCjQgdjC#Zm-(ZEFqPAs#1E}zA zLQXd6q`3;g)50d~BLh|KwbCM3fM(!BkWT4dRhXX}t7s|o z+)WeDIw61Q7s!QZ1?sbN;`D<^znVltJ{Uz>NcT19D9-2iwk0ieW~+~uB}J?(N?fPM zB-f@!u|y1!`V{AYsMr?0P}{s~ ztCH^7>_GDhg{Co;Rm}s)uFb~aZ?~+CNO}-Wv(4$oT!p0@0kn1aQN8#9vLkAh{1O@i z^GRzzDznZO@$XQ%xHViYPciOym}X;ToVd?he}hlH;5!!v0uSpKx+g#{Lk%X7{Z%p~ z$Zh`ampG}@_Yuwr_Di$!O zb4!xNrRcSVw*r~(a&J%G)&vfrEa-FjIB-UyIuQag91b2nwWzo#+QPa^=u#|V5C-KS z(odWCq^f9XvI31FuoUFvKxTb2Nq2BC3GAFkK6t4#B+o>8j4Os?OGtO*qf-zU5}3g~ z+a4KucMTu{d^F82#(+WR2h~IBdO*}lC-tiu(g zSXxOM^AcNkF{x+WWeDYcqDPe8M(ja>eiOu_xP|awo(jeC!RWAa?hnP&XT86UUjSXv z-{Sq3G^vo(AxroUbg=!J$u5i?>9OPg^6omqB;s5*DTJ8G%#+*PLQN|gJ{iqGv6TfN zVo<0j{2E1lg|K%461hZDHW1`6153tZ+tGF@&5Urlj5J8hsBCMER z;!l^Uv%(A&tq7?=tf1=ZWdD`X4RLl0T;nqtAAK|RD0lDn@z?lt8XzNpTjFFJbP0(L zN@;<={-9p9c4dG#F?ExaT~6OAYzhm^7{i#_;gq{GVt+2ge;l;FTibXrj;mRct_*Yr z2wj8*#Km6AN;r=JKUbd{NIhVdnvB}glfAzh&SK1Q*H z93OpVu98+r$U$TseW-QMg(3h}x$7W!({8gYCaxI2PczvjXc5wz!z6;dThsouJ5^V_ zBYRcy7M@nu6p7m~>S_(;P-5aCtn(QN@Rz#kmT8Lzq1H6VV|ML$SlG^%vJg6FA zy~6mrn1hp7cir#_l|*x%7hxf?Yc?d&+y6KC7h$~H z6V|Vt5xg0S(dpp3Bx_9Cnvz6Iu#JxZSg1}Q{n;^*IO4I=>`cLD&$GB;#~-#C<$DB! z0*&0f^p6k}vdiCQp0`yfkn}283z4xoMQSHv8Nx?V@euN1Fo`5BT`VJIqzYq<5Tm7g zz%`SO@24{5Aiio?zYtm7z*As-1eQsqgQ4A=jnV-;_1UcPVJ$k92&05`T&~0=wCKAXUuL z?1u=Kxobz9{3jZyJ1>YYcKW|UL`hZ(3eEiP7jlz*?NqTqrm06GA}5EY1eXk0H=dQSZ1tWv497d`QpU zsZ4;+u#JFsATHSog=%yrUnR^!_3%^XNYoRD%M2aY&O;E|9E|jPy*-D}BH#M4xH#KaiECN-n&J(7Ixf12~B?2#tE06omw4@f-S)S)amI zL2jLOkVSmOGkk{LR_0US4ETI z4!9ov?Io3>!If2)>Kw`{vJJKT)z*!pB;YCkSI zhX~hbTz_iK&%h|avF-6aleuY1fZsuRA47JNQE)0hAQtfMcu{Z09LJlt*+aKlbW6pp zkeN`W4*KlrVcl`9E82p>7P&89$QOXL6tW&)V2u9hTOvtT>}_;{6H`o%)7YGj5=#3` zsG+n!6Gva|&tffx4l`InFn?o%-xsqvHs|my*hX-4p-suq$()?b40t>wZ}YF-ije*?s1CTer^_mcEif2fSp|>cX+WzC7ofL8 zRuJbEFVLsWDA?$(Ob)5$m|Al(?Hi@Vic|qcfVZk@{uDgM`9o505&A}U)sxvOMZ7W{ zwMYZh_{iz09$0w<1#gOmG>q+hf^pUX$#FnUMeb|6@P5W#R9-5XHA!56A_j5<>}Xdh8pblFmTDboe&CB?FeXD!ZAOSTAx>dqD2FC+Oc{4e zxf5Ovv0%P`!ydm+j4puwz)491X)w%3Q=4oz@ea^DRblm%{Xzk8lV6zlm$IJ4=M^bO zCe1X%Q|X{ZZYX?Ehzr$)p=QRMe$pgQnE1(EZ#hHjadL)@Hz{%wbavVnp4}sgRWqKD zzP}E&N`y%^fEy+UK_(IqIAl%4z(D3#sh#t)=__cp)hrrI%@KA$5%E*sYcH{Q7e{FJ z1?Fa|8(@whx8hdBaWG}+i6HdN=KXmGw&q=TT{diOLRT?)5SsJ(Qy*La z;a`hX$;Q`e&WCO*d;vX_jlW>y8%R=|fAFsGSi60uq;&ZwNt4jewjlqQ7eNd+CO_n@ z`i|pxJ4;~DndWV>DA@N4!!|dRE|ecJ@=F?bM>N~Mh`kWi682U02SjH(lV zfi+PsD+Uu+O%SDjf+#-T9$*I{EC#W5&)}ln@ z?Nirmcn;`MP_L#iu8pr53g4+c%V`w#+jS72wLFAa)gsmdmUD4+_#}`%vl&3Zz8$5{ zJl()6n@zgS5ABNm<${<_vA@D_9Li*27$#$X$pn_yxqa#K z9p`dI|7h~s9@xDu7HHq5M`O@;b0aUhIlkc#=_w0C0MBT1BQV&i>;D~&uxy+ zuG(QH@*oX|Z`4?fh33WXplHTq*f0#di2@nVjIjdKFlnJdO0RBJ^{xDrMf_l>TBgR# zf<>0FbGk!)@k)_IIj%TJqFb>xo*a^k11p2G*2XkiQHDU>{jVHErZFlAFhQfqH7Up* z;&bL7)Y&sOlRjo4MpYiNg<03{H% zrH(?lC!{r>33)hZh721(IpnkYKe_4bF5v|{2C;!s9F>6 zCh0f|`<)4%$rsq@PWYdWAoY;sLw`_!41AYl{DM+82f?%`al>KeQ7P@vRpDDvjKRwg znq9|m3{&T2cE_)O_&o83rtX&^$kh9(53MJ$nDTH0%{km>L8!fi@YF2c5g!_i%1+$Q z&qqSm-$L(7oA$*Wu01RvyhH-Kw7Nt-BZ;3i?Q4|i`Nez~HH4nUI!A-tkLm^z!_TFf z9yHhVT-LoaHRGAxna1N;69m;CTKA-46CnuJQ|PoxRIW$o%Pasacbm_Cx^O~&)anCugbpvLrgN*M2Xuc65WOLR8UCb8KE(nlzXm8R)?s`kS;}DW!D0v)!uZD_+67F5 zXIYb070wR2KSxWJlyrjOedsy3=%Z=pwYeS(r`6P6TX_-CsX@eN|d_q;C}<7ObxWJ8?_Z9d8gaMUr6mM znyYTtp{ICk*j@3gvTZPc>CY|HK@+SrLwBpP6)~V`7KWgW2j<)PNYrNGIah3(%-(mi8#z-GUZ2W~7S3t*x?+xc2nNICx z;G!&?=9dEV8jnPABAql3+T05EWXjPTXj3YIAnYC9FI>@a4`fSzq3RCqntb2kLPN2W zxo$pajc4YPmv9#13Zab(;)lE9iC{zYF>n=eJQ_SfX}Ez1WOhh)$IN)CLrq3b_ zwtH;JY0`%;0v%Ff$;HA&yvCE85XZ$t0_HvL>#q4_=kFt$hD=N09`m2G@Sr1)K}tOG zuYgNvw#>3(fcZA9Ui62Shai15K1u)l56u109PJ|(#rMQ z5gi?By6SqADf{6|q;+4)(~#}kk;Tf~8hE==AeE(J7t2hDuQ4z)$h?ydTs2th(Vx}7 z-`0EE;3s0cYiy})jW@=QB=}{2As;9qPn#@@c;#VRhmMapJwB^dux424Lln%yt!WadF>cq`Zn|FGHn9O;xbOFP6V2W+V;x(|A# zTDh74QD;{KcNtc5o*3E^32)Ziif{p}G|t~j9k$vH=5Hhv(cWkN!`A+rPrn$NQ^8H! zbC&>pY8%KMU0yia1emPc-aN2LDXHkL(xDE(Qq*@@1R? zTsUblGE~YdKe0?|6a2tr@&erVAY`b-afoIDfqhJXa& zg4FdYyrp^uPaca;1}r3j%ul|QR;CE{!%u?W!gkWuBNn&G2=zg8Jj#ziNruSI5iRig zB9Gdpa@d$>a@07F&D8%k4&%3ko&fj=F(p>5Xe%O6ifs%CgAga7bBrC$tY{mpP6l!)} z6u4m_FeJNgX`m0HE9o)}im(yX11NY2iCGu{v0=bL^S>f${wHJm!q^VQb}Y8Nu|0Xl znc4Ap(Tp>5XB=1Gs_yC8%rAvMDl%%TORj_lYS{P?vgkuSwyW)eCPC&x@VbA$bJyUy z;-?Zte!;>wB&%)0gh;_uQs6X53wyG_jXfbB5~|aKY32DQ$i`WVWWTQl*_7@}!YZSD znbX)jF&hlc8lQ9zP<}YfV?V*pp$vo~g*IC)_vt6hPPP|vy8x1x{@||*AI9_XaoP~k zXNE7`nP#Vf%%15c!w8zD7veF{JN^PR2z@ximgCKFKd$^dyV?HZ&+}t`7Ozvfv2NZS z%sN~H@R4P#yBdi1zZY$OkA?XN7m<(Sw|~n@9Zr9cGwG0MqbH^u=$opf3;WnSy_Gze z7Zdq5pdvN`cP(=(fuP>#kqGJH*Vt$zYgaDN=vXorPEdY@(XuC9rn8ggfhaOwb$t6T zD8h5{diqi(sfI7$TkNwG&~^UvYAK5TYWI35$|sz=Q*58u4Lu(mKu~iM!fG`t5eSq= zo(Xxf1pm}?XhxpgR=Cr7H~|!T`qk#IVS(j-dAVxrF(to|WtFi`KUIz+L_;(1si%T$ zagnSjN4b?!eyf|@>?i^D z#I_OJbR%vA8yZjmut^B(QoCLYCnsl*YFkHSpLt7!8E_&V)Lkpj)ZAe0VR@rFA2-R} zkBJ)(+sf9Xa24WA_*RPyA~hUe94f8fR65s@7)gIcNbYJDhuUSiyFAAe511IhkrI*N zy}4bXT&aNlNc1y84t;-aSLHqoquWTY&FvSt?d$d{_l26&Y(M{VouSo}j#l=)UaWBd zE>8>bX^h4V;w^xj-hCF1x_r{Jev{jQ?x6BO7?GMC{xdUIjf$-_hGbSKc#Nei*eDHC@4tA24j#Foy9~g`NY;A z(oF+!E)tpO*EZp2o`QGa$&1)W=+L#ZX{0lgdTb*4O;eGD1(4eGfJ=;gM_mL?(BpUCX4%s!IYhqL@c89+X68|VqZ^U>UH$^A$3?6%xK zl-pUEyG5yg(W7V0=<4UC#)}&Wpo$~IFy>GMb}LR7h*Ve{Aq1cfQsk%>-?xjtD?tqH zo5V)KF=`EcxYPV6BA#wvSDgoe==^QwZ>`R6sLoSbPDuy-hdA55V(MH*===?4x9~t} zcVAZ6<%Pej$gU{tJ%y(8a>D$v>xUBVF}tRjz^f^=j)LCDcBE z-#DoUeBY1D#(iz+SRj%Jb@+cy*>s^802>0Iq!2bHcyx&5($S5g?G!?|$(|j1S=)$1 zco-bdP&7vB4h>AlJb|r~=w2|ajMbAZZT#Jk@UqKV5Y^N!Z=oIkP}|+r2673iV%9li zV%<-h(MtDd)5oJnn;4Z~m_P&lC}sz75&cWk{i*4$ZutvZ^d_Pj#*E059aHsFF0pgT z>DN>B_7LV8OBZgkpkX*G(Mh}}`7Mx)FDE5g!5`O?^T{%TyNi&9iVum=>xrz~P;AFy zyEwM6btqLil= z>`bi4_geif{aCxWlSGc)0ZmVSRQ|vxGB}ZNp2l*}^|-q+`!Jxck^Xot^aBWSf%bz} zV#wm~2>hV~!4HxKjXyY^VXB@fc6ILdFj}L$$}Yp@=HU0en`FCIPqUZz{^XGkaYpu4 zvtKsrziQgo+V02Ze&4ijv>{VE_+8Weh{S}uFQwIakID#T+}5(ESQ$_rXR!R3BMIW=(Vs_HXLRvA{|n}*<_6ot=1}NbY3* zAO>IaM!7=iH(dCk*$>2a6Y~#i2T=`w@a{L>e(hl&yXQ%dz_-Y<5CKsgSV;%LwPS7) zMxOKf8H5NTGPI8@orw7UvdY(BFtQj`OSFk?>ZTHh9B@1&qEK5vs0_no=}_h(d%}7m zf5}e+q@1=IMHPS&9DYGA)xlkALL)|3g3f1E*lNmObE5Dez2| zA8LP$KQ^rQ?h7bPU>1%Flxigi6sCc?-Iuad_Q7DPslmu=N*{A*_O!?hc()CspO*Xq zC~&64>UcgLPo3jFeV#*ynAT~fnQ}Mi{8Xfl+y{Cq{-O_?g>pX)ismnj8>_5e8I$x5JO03G4H zZKD#1Adx31f%5-9D1q`BN+2B6Q39!qrn^nVfI$f#Fw0u9C#!)PQ4Q4Cx62Q$)?G7! zfzLD5KnL)^y4!w=OX{0e1Ca>I;mZOY9BT6{nbodjSr_xO|FasX6V*VY(`ukbR0Ba) zY?&}oP?HN2&Es*4Dy5H9u}@a|)KjQ4N}sNthFs2W0)h^jAu%#uW1O#^;lfFapzCtU z^d}A^wK>%UF?Q_(;DKLKDx&e^VU$5p=?hteVLh8nuc@7d#~j(c+}7aNq)0pNBI%yq z1HDf98cZ9arP;aSmd~R$VwmV+a5+%Oe9T1*0ua=M)HDlVifUC=tKg556GB`p5NV~_ z(z*-PoiP_mDbu0lPO6po;AwmoW6MjQJ3AcxNN7)3dETembzPQ|GUazXyleX~P~{h{1K!H<~#EgUnK9_3<+JJf!*uQ~@` zyLXEeyiD00a#83S#$;X;6*RYBgu`KH$jJmHEWpu8vMuYtEb;`Uh1GWo zBki~h>|yWmy$al-g{p{(?ZI}mgWuJ5;x;KxMS**PLB1k9({+UAS%4b)yK%HST%{z5 z@|Ho}xv$j~cdo0*AF_3xWUGonB(IR~QFJIj_b`rJIOM~R@<_(O1c0pI{m=k0`9}>N zhBCV44_FSOY!2A<&_bAj4_eRmu`89U8$n}{jhHfph>wjv!qMy@MUfDQct*BH^xamc zJ5|RqA;64`v$H12^GCwDHczHUF*ot5sB#Uq2UJ;cNgJ$J$@;0Mjo6&5LCcrl+2x$--td@ShK~EX zcDue*D<`vw9pY~{*GsXQs|UXVfQhLFH_#KvMLI?2vvg7J9&w44O-8 ztS;PEqfX87lAf7jv7qpHz6olmffzJdU&0Qk+^G1%=p=Zrvc8d2J)vYqgZwB9=E?-s zV8Zm&9zAs)!OT>{lScv;DS?MzR-g5cu?ko?BlU_dU|i4~j-rz0jJ%$AjZ;@eM7j0& z4Jk%#5Db8XZdk9u9uj_ek$erj*;aY0IgzJu?rC(GG4#0AfU+Lv4&iYzI<1NQwjr33 zS(Qb{`G{2nlwh+2{ghtq_KDOL>eGOmtFZEPgFC-UPw(NWHaOb0Dw;y!{u?(_o}Pt+ z>_FbPVX*TxaKb0XRnHV8#;~J9sis7k6E#WO+gw;aopE8>!IWHxLLJ%<(+;L=89A5+ zzl=E~yvL73s^i_oxLC4XS?<>$f^bqy&U$H~*VrLinLSSLNm^GJrAg(qjzv{W8C)iM zRtPuI-qN%vK1!$q20xt}kMVkhD|_61gekl9X_0t48ACx2Lhrk=LaUffE~w-0s($;x z(|f5XTKyzi`zY|hbj63lG76u?A4j5}$C9-uQi{%njk{gkJXP>wWue%9Jz^j!=wkjQ z77h2rIGI2&0D)5@E}ooa+_6SzYZ&1JdmU36ofuj8?V03BIL1}Hw=r*#t-j;%+qFT^ynshJZCnH#{s}%ZG;lw%WNoxQbGS~_EhO|AvdDD8$F%p;%KtI%?$2EyW&lguP2fWmcWCCx|EYvQc)?>+ty3meAUkw=Gg*Vr zt{_CHiGR!x5{D8-XnsekF7|+a6ehI6fDVvtTPs6xFBY4Q@)eMS{LBzoAXrdYpr-I( zeWtORaspVH>6UWd$IX6x+~xD9+IG^q%aT7K9po$gc5M0G>^PG=q0C?NT5CqwWT1 ztN#G=nbQ{lTuP>?KYyF`_jk2DGJLZOBZtj^8Q9@0EfjJYQo!@7>~0_wisymaljGb1 zn~O4iLwfxB)HbBUv2Cmea&9hR>|iazLFv=nKB!qf2_hBOL@pBhl#ZRZE_=F)KAzZDv^gvWI$6yY?4I^PDH#%4mi2i2aQ%peQ|}oDZw< zoB+?CyX7i$mxZJK9!i|NO&pvs=935athiinwYX9vv7MaJvUaRvz8?3aaE?pPtMyzW zzpNC$Px2tWou1k!w1iBZ^XFKBzKh^{trfmLeYUmN*7Ugr<(%@YKKnv5l^nab+1hXW3x1 zN^&7k7aS0|#~N!rE2!lxpDU;A>R5&#T=@k(NwOR6Fz185<%L!09W`5mD+!M9R*_ZK zzAR%oMUWQfOY+dTayL-jQJ6Q;>qbCX62aoQ*cP!JozcRC70C~48M5y0E&Zk0uPpt^ z83}m=?&@LNHuSBEI;xW0>*1B5<^Q*{-jci`n-Gc~)pYHXO#kZNtUcYDY+g^L5 z*ZsOzb(a91FZH@_^^QQbN3Iz39~gvT;Ah0Ai(QYc@GrSBVvE9*V^Q;{hNl&-X45b~ zxt!^k+OAJKQ^qv?YuF2qUR-MgXkTUXOZo-ErsoZ;go&$;7h#%WgQZR*BCTo`fbb^) z($(NKJWKN}jU}5Kwxux;+iY_qPk!7)y+NW~35x<+lvBMq()vu0K+a}`x-T8Gz#0L) zLogqaOvG}Wv;z`ci6a}>2DskkJsFtwe;3n^Y2Y}%dmZP*xxZ6 zKCX6Vw=S_Y{Z#zE%FOpx*wB6)89L6zpc67*V13s>YJOkrRl$+c#F>s+h>RUeZ%3) zPwd+Fo8bq<*v7@Jy0&V%Ye4oQw=Kda*RwDeC%Ki6?`0&{6${txsaym$_&2X7eb1f! zgNzim{G<5u>zVLhQ*n2E08`dAZHRq(+zZu+Ne*3OM~E|s zs>Rr3`k0-bCI3J!lA+u1NwwV&$OIs06=db*UHiG)BD*mkl!D-@(#GZ_ufzAmqpiNRQeamfs@AVZunbqY+5>lwLNxSk6nxR#7LR`A{3vg2j-OuCg!i`31>BJEx6Cl zZo0KiJF6MqCig&A;4Fm=+uat5vziZCcM(EQIP7B}pRfRpabu9c+U<-Or0v_WG# z5l4Ew)+8*9EgPNVnQavrA^i}0Yd#Y6JWQBKUd}tcy0h4 zFhK5s|BT2G@mL%xKMfBG3Y5)SyeXt!IMFE21o}qwMPdZiyA}o*a0=j#RuIerTeV$# z6b3A20~iOFl)N0})4ni*SO|`eSqq5wuMxEZd!)xk|8;r<5>UzkHbA8-EIlHf0NEY3 zRf6a2tuXgDj^dL2u-RsF>(F+foI(Uq#-Q(0kr1Nl>}f833b@HphV6NFg69az8|iQQ zF|kkS3HB|shs-_C`23Kic#l75{sBtq;T9q`Q~grX(EW{F|v90tF zPi~Pn+sVD7$eZ_DSb%1EF1e<^Vg5ez;uXG(Tdd@(;%3)|G2A5Ah+>nK8)136O#2$= z&UT}}Fw)>g=ABW$EIoxn09|2AW3EVFl{==wloW5{^gn0V2LT7okvoN)xcC=DSuMDW z+4)T-1(}S{DdGhYmee9+{jcPsq$O zu$kE5@yy*a24gKQZw5#3$uuMFY{e0LF>PI$+Iv$w0H)yisIT`dUT4=C97nmp&nXrV zJcXPLMMX6^ia)^UtK7eh9_LQvIE?Ndeir-mkc20r2#QYD zjc7YO9{ci9Ea1EO0SM-Hk+;NCTzlK6b*4j`fqzmIP&-8~=N5Y8^iA9adqUsDDa!Ui zKHA$Ytd+HuLO#%*tat7ZX?^?%WtS|*%Sr8l7uNKzEyh)wGBNaJjvov1lo#VbeF|W+ z5;)iKnGe+(s0s|+{Uj>R1t67rK=@qkv(aH>AB}DPPut2L zaL+j9UXGVP z+m-K~!mq6+8RV?6uuDZ`AFVi?ic+H^VQQXCuVaW-why}(bbMsk1KL4?w7eG)ktdOS z%6|`KWV*pEUN3dd-1XF^d2oU*g5xL(Y4@e>+7v{5Uus`Y6`%8!G^Kr83L=i5imy^U zBBjK1M62+Dd~~WMQZ4cwVTRYFbGE9|O0Q}fTtAPn zN*DqPC{-z6A`k+(AX{D{!;mhz3-hb(gashXQ9=rM&Cj}x7%ceuy-Yd&?8ic`rsGx{ zAZ!vf$}#ddoDv^958|#;$hmba%v6}`jN5De{aNJ0qu&L6d+G?+9#jI77^FV<@9$Aq zcL zIa?eGjx-{dasH3sGaXG<@_2-V5d=vJc3d$5q#A!6X`i`S)=~^qoM+Wg!ZnHH;yk*3 zvH2leS}cMC%z0tP;~#ZeyINI7Ep39VbAV<=WR8r--3-1@!{9#0EfcoQ!y>(0 z_fJ0Sioe>4Y&UHxUnta7rPvKCZg zc*P@cKeb?5wyI-Ix6*B|j)|fDH!xcKEpZaNtrrimaz;BGk1JIb$8~ZYu_>_A#m#}N zPLRSPOm|ydw#GY{`*XZM!)NE?#pxEI)*pu#1QPj%W(rXjh?fR8^n}{HC!K-Di@Smz z46)iRgw1hYrQ0_w1io}d`$0}`Kz6uR2um*F5npZV7GodmM>MTv!)~amNHON#r;rHh zOv`A?zOD9~r{nHTlwbt}+TF3VIUeY4da8K9j!ApU<}JbpLmEA%{_G61vC1E5NTGIbu3#wK4a!DzfQmah`sZ!2cUFvjx>@Xf81+yuk!hOy%+3=WoKSm<$ zciLkEf7GMC&(1J+re$wL3Q`fPGtKYhc13`Z*J}5R>D}I+=L>R>L2JM+>>n4Frf-dB zqZ)4dY`Z#}Q&|2y?_&$&G@doXSlBV!o=QqB#}Ar&sl&xh`*5wSn+(GVm;emmL7p|l z5XL!H7p5e^doU>rL!4(oIvbWRik9Bv&=tsVt!V)PYEp2>jww8ZK1n^v*D?Y($LRW^ z?yDMxw>d~KGUbUA-$F=LT#P)n)iRVb&SdepE=_uIKzC6ly6?-o?VJ@CKL(I$p(t?1 z9EF|Sh3V4wK9Sm;slPSNZrAt5Z5LRhQ9l>ibXBb^2D5pzC+E*XhQmzcmIX!gX?o$x z!vr)ItwU|!F|A0T6X+MC;W{Pd1yDOK& zf>7TfRW_i#ioJiJ z5VbZNumBkaR2F#UBicgYp$W18NXG!h7Z-zb9&?BiLjSZ%5*e7v4A&&kd^!T%&C3<^ zIv2tOn;e31_HvDlYENOj>Y$2UT z?>CPTZ36<(cpw8$ITq!YMwp6Qwlcm*S4`&a8U07-s`|b&T+enWSGs*)tG?H}eXlZb zJ!Rk9$sm0ncw*nXtAmKd(nkkttxYTkku9i3FvbYRN$>(4BA%J6QLTe=BWW%w zGs2Vm4?jQb4un@$|9Oh${_Lv%=yDKEE2;C$nc`?Ap*A)|*Bld*cxRh*yt`Xn|&>=4iY0#pg8Wd?} z?Wl#(E;_ZewlX4+L*q7sCyR@~5;1;;8RKM@GssNlNrqZs+KLtH3h6tdl6{64Q-Z+& zjGDUxwX9GY-PSN)8Anr_8WSU%6_3-+2&Lf&JbGcTN7B*$k=eY4RkbSW^!kAkN>+$x z1T&Fu=`{Z;QwTNrW87)|uPsA;1ViO9{l}GJ2u7maOC#{%Z~!WYG0{eeihkK>`e9Gx zRkNoX#q`VF%29M9eLW)XS;+5bS0(K2@ZT&s z>5038Fszw44D7|^0r%SH)jxp>lLY_7kf!$hVs^|D=9X0Jpg#?bUDKr}Gj~m;n5c$Q z?I4KxG$xQQqPU?-xyEja(}D8_w-6^GiazXKh70m`u1`cE&Li%6XuK^VM-(#FfESis|_N|IugWQyZU>xbbh4 zP=Ha6E;Ba?7Ss#9!K|Rws>kHQ&Api)u~G za3|D`=kf{uc>i3&bElqBr&!1IGjQIl?aobm_}%?(ijid|Weq@%)Z>tsryi#}kLJ#2 z_@h0wzIj}u+asmBt#M82ciIv@rN`d0Jn%%5FlE|M;Qz%j+cdmqcl38i{2hT0Y2 z1h@*)E?i+`#srs4J#B!iZ(@$#?BC?yNa8XK2=QAwPv%MI-mNT3{+&>_*l}eHt@zE? zZWizV{LL^rWCsTT$$Ox5%4}0GxMKXWm2bcbMO0oXFL|fcXZ)L5~8QmDIWr{)eig+>rk!k%t^p=|}f! zbIFDh6*nEAFiL4kJiQP}jX-hY&(f--d+^_9r%4x00z|&*Z7=FZtfp=RR_grw@%J%q0}?@#W8>kU1mMd|OC5k_c>@+2@M zY+7+E;`%`^zmobmbH6hZ16IGmCYF9CGN@Cd-ktrU5s?^MzuySEqp_fX(3Gu1ek&Tq z0J>ChVEe6qjFT$kG2A4}RxgI~Ly|2}N84erWlQLl#1#E9bQ&8#DC_O1 zEX@eVGZtaL$jc@Zr2d~L{Yw#s{b4dM3}M`)7;X>D#gDCgs7jw9aS&oh41jP@M!H43 zgt5bL+$Aeq#2pwHo%0satP;?3vRyl2M?i6&U(c{P6Tm9>dU9>9qLa5B?9gk}&!0 zj&uaO>sHUFPg^-FnaeaO-u>UF+;R-TJlv;nrJ}s0Z&SM*iKYTjR*XS$*!) zuIF-4g18Xg4g7%9LF;8wE3Tb@MwZA38|B!2c zTkqd-?T@lm@Mfw0%I3ITI7rDzx4zbUr}uYP*}TQnpoEN68zznz3HFITMZ_`likQpk zE7EsB`6Y8k){@-;ZN}}aIxqGJ?Qr$!|II7WSp;NIjX=NL)m}@2fF}VB(oP?vg5?-a zob%Cvpfbk99@A8ci#QV+ZRCfhuSHHr*BCcfbzaY}W=_A+J+~dthSQvyulS8F5AzG5 z>1V;GhSh!;ZUz(~_KJCy1PoIuCs7_4!|nUfJ3Mxcdzfvm#VEV>sIDO=#P4yfjT$#| z69c$~J;isHkw|x0--@TaCdI=peK^iP?t9hqoKe5Z_U}9B30JG09OaI-qxzotl-Ewc zTzx|OBwbT4VYgp&hP|}*lJ?7Lr?gM4y}bR(;uY;z<)^j(C3|)IHQB!&_F87|bqY*b zW9`#B$zh9Etcb`E6Q#X@ z>jVJMEsx|xB(#d7VEBt+9pjU8ahFY$@aPa&MSTBZK}&=YB!B2`lu8PUeK{;~@2tXs zs6I@Yq^HNwmlTBhUWY3PNP7QdTQY|skja;z0mBS|Lzm8?uM^yy#jb&2vGe%gzvA6I z5G`q4-KUJ~Spri1iJ{d$9NXY@@ixGLcvcK0&`6GEFu8nbuwc!2Qxo87C{fC z0JWtcidEiK$yAcU?K`5{oB2uipgJ6f_AL7j$1_LvcIr@}{T}OTj1UBQ4)xUIE(N#n zEF3(eRYSN*q#s)Xo0SpaM@WLVeI&+W8NjnINe*Ut>Jn+OpwAdheSVZ`D;a&8D8ovM z*dbYG#d)cJD^m%g%0~F6CK%6K0SPn6f0Q=b(Al^!L;^|g;ptqj9&l3yedM=`^k>di zLZfbG;Pd&;Fh7yIxy-rDT$N8%lk1UDk~_kPCJ)fVeuEgdU-bhQTbQNXR~fmD&L@UX zAn5jsPw$KEXl(uevTeottF%tlcX!>2Kpn_3%RbO%cg7mMvVKEnB+0xyU9|Z&!U_J6jovo-`bLs5v*+1o zlw^D_IpLgDZRG2wQN*Fi7eeDf=f5FP{1g$Bh5fy-HKlDW>^|onaMM`fdpt`TOy$jO zXN~^I9)l(r0L(ra=%RgkFTjZ_Wq6@f5Bz%E=2OuMk&+q}yEqkMld{e=nw}xFmHPIQzrnhjXTcdvWj} zY~Z3fow9=sA@93rgBfLv$mX@M9UbrJ&aHLf@7x7s65~<$->2Ch zQhO9sY~L#FJEd(d-49FqdunH9?uOFcRI0G@U8Q}pbe}HmY*YyHBH{@ITLb>|mw#AN z%N`rRDyhiczF?&#IGvV1&T=U8Hd~UPA5WtA!i2MHefrnomKB! z?wL?Fvc{o6@2xWqNj|qRl0)c|1oD^DzX9sffacM}IJA-nMjd2T{jAbQR&Aw}ZO5w< zv2AT8Quk6j8YG6jWSKNV-!T#vDnFZuCV_(KO5m9sbow)$-=B+wx3O`G#50{{Iy~bZmTu96M#yb3j|dZV}gO=8yE-r7FJ_lmTUpr49Tt! z9xz3lxx~#wv0npt&oo5|sZSpPxg4!!;^7q7@^to67#!lWNeozi$n3)$Is8VR6a7Y3 z#ddH;d!EG)jjf@@H-~1d8}1SAndy#HMB8yGzwn9|6)#@>qWr}>@6g<_nS6=KkBjB( zCGKVRLRRhM>a%;bJCUOY$(~{1>zWX^kgApzeyHOmOir&?&XJxhbc!Klb07~?R80_2 zu`Vsn#szIsielIsi=}#VH;LuJTFo0=^9~q`IAvPxJ@~1ndk&K|$*0A9g4N|n2}j8W zDaw;1+YwBj;?Y3BH%^?ph-rcKm;kWm)CcQ4tsW-bCz(n0OAsW$#E67v<5KKeuORu+ z>RMjzOpJXqxzv|nG88P4#al>3-levC@)=0c0)@l!{ix4t0@e|Yw$O`(sW7}1$-VfK zp%bCvaZmDk@jEcUJZbM;qG>Eo8V8{h5yPg6rmUz-SK|D!za3P6MpqRnF$;nr0SWky zj`;Ci8Wi}9BJ3KUxW+AvlsnyulAy77@@X%&hr+H{`FrIdI|iSijGymM(h|JUjf{63kmm62fXqZa;bJI~J3FmxeF2)~YXObrm&~5dpO4v_)co>#@x#Jj(n6p2I@M*79j*S}L>$P54E696-&YXaF z!8Lxk5N#+XkqeNLF?5G4Tl|!TWkja-rdNmhRjGYkuIc2sK%B7mLvg~pSZk;b$QvKr zCAQcP*dV1GhWZOq#l73oy>OEMPwPG3kcxW1RQXvUy)M`b2>5$`JA(k?6aFVpO;LY= zv%jR}Utxf9MJhFRW6l#XaI3H>EiwDqM0uEObG z1p9A}gVHzO;@)g;@o&a2&dv=1ii_jB^B2Nd>HQp+o*`WI zYJ-^naIM*KI2}O)Jz1fc%{}!+zaBc4FYxm*(Sc9cvO5s&+@2c+M zy$ScgHgw1{t0GRT#TIEKOB#nI_vac>J|MwL%n-pr68YqHG}6VHGGLRoz@%os`~`G=m%A@LY7d7M5C$?hgZM8jJ;mv#1HeRP!#h>D6z+k zb#er@Xd7Z$op~NFYtf`mq)943!TumYF?EggRdO91(TS6-FoS0%>FsIbr%#(nii(Pp`%m5gG;dWRj71 zW1cFsVU6uBsbqnggVjXQWT}qws%a++%I{hEU7OrswjQo~D(N)l2;7+MyQDh*y)VB4 zxk$ebzG3Hz8zs-#Rj#qh$S9=UPc26J*8tpE0OqewRfqAS@C{TGtay69v$Gu7)y+3j z#61cW(%+(<7A3h7=0rBuIRuZ$v+fkGBfk_H5zd!l14j(o8H;0+4C`DaZ4-ONjLOkR?qqk0M^VnWCUFnGnKX&ol(%4coYF4t~*=ue7S?1PSd6wB)%VLZEsc)@? zv+NZwD4supq3QH-@(aE61?BUPSYzesM-w0Y84^+l0;okg3GA+o+8CIa!U~gZcekJI z@-KMzd0&3P+vj~oLcFhdyEh6hV7a39o)9jE7*B@5vS<}!Zg!1amwo{7Is-Avy(`!P z5Lbz}BsH$wZUiY8)s^#grOmwx2ap$FfDK)Sv*9KB-W9r<+ARExv?3r8KihUnM^lhw zI-A)m@iSWJ=VgZxnpy&tHM-C5fMK9RNJKb)K&d6HUCs0SHAusKJc-&ELpUI_&drjP z_thOA&>h!v2j3?o1;>&!Q}&F|i`!qjzZmS^cen2QjP7F(wEIEJumSMwp%)b0o<+>c zsGo(>S*@R3ME_aG2=kl@pB%c`BUd{v;AcLg%#@`Dw1o%eN+G@%yZ9xCP|& zY(Q3eCT{12WDMlJ0;nH>v5#;;nEVlhlpNhtwgC&5Zwi02jVe5Oa%`I$scEucqQ9V; zr5hOP6oUv0x97bmWQ%WrNx)Wa7@>v;5vU~#z(!i@3UM8X{1%S1!{sPBI!MNxRlK$* zeifDyI;H6ewXX&LAR*fHE$l`nAW7p4oqUC?|N7KjPxx6`TlAuqigHO-)ScUcY*%qt zZ!F#=q9Mn!$~{TSWzFv-*XNKb3e7Pgub31E`B_Yfv`t9VFr{Gz+-k&?#T}lkIGV5X zg#=Y)RE_XEL(I}kFh2hqJvdId$c>Xu9-4>{%034~S*KVD1T8386)QpjU+Iv?n`{us zgIq20O~u=7V_}<$d`8=d$1jU*bzEJ|HWtIlf}i4|7p}Ggud(cEOzR&mDD;GJriSi& z(QH%|%n<=Nm!x(n>W$*E)Z++S7P}OA^O;y!B|6gP4G4o;BK@b9S+rB*2Sf(qrXSRWwOM|Wz2$JJF0 zR|uKS%}Ag^tz2s8kHJrL`WW^ zQU;#WB_{j(t6<%#*vO$Vem7UJ?t+)TDSIPY-pgT%aPWl7h*@jc+4P7#T|!}~Mu}e6 z?QGkTZ<*eOp9hCO`en83idr6;bvLeE-p4OX9MmExUQ1Mt`?JkEvZF_~a2D&c`6Nt`Su zjP-`T;E9+nL5c)~0D2`tj7NE`(WsvmvDx+!Y55uL_|)-OvSfK*-rMLK>4nSb9ZNPb zVd7#3pOZfu^Fkv$i}j}iJY2&isu8dOd%9S__Ha+SMlXck`uX^tIOff z8W4Q}*BG5yGqBs%ep@rJ#^T$Y*oLiW^_M(q$#=m1%G3+T%ume)L{Y{T!?YbYnN5z( zZG>4fwwkyOrUi$4-kGWin%+2zI*ST8x)`NUTv9 z!H4=&Oa)wQoC?)T(J<%oUX`uZAej=tO7;|{B!boSlyFt|3mcB#T(nUsr@w< z-}+;_D7H&xwEM>6-i__SWXW1$B1Q*>YQxw1b`v?h*7J1* zGi%0G&S^EY{o@L8No)1b?-kvc{+_rg5g5Q9(K@bKT+gLEM9+=qbd7ryN}-2Of4}<) zfk5|ip1UJ-Ox>*~i=)tcg}j8c;;(BC!$P4|iHR%mt!shvzYmzXMdgRBvA8~)Y6#5M zS}-EX{t$)zwG?x~O|X_f14!6Xd)DTfZK=(PZ5i7vwqfcRB(*c(gkh3%RxEs(TMv4hx8l^QzL2!_$?b*;nFs4gAPkyX>;Ed)@>zlTr8Dg7Fwe@6) zDrt#xMz2AYrhlNc_E+Vt^*J>5%G};3fXveG<^BO=TGIEhh4MVwYgr7s44g&Cq3e35 zn~>J5f;BoO=!xA!p)KFU{blQgM){-VM9}?ZCJirvd{gt-jZNFsti?8LY|c-XjBt-K zzM_+&mm)2difnmueE9ORx^ui%~ z&yZb=mBL*z#88^sJhlKzv8}#qI^!4~KQh$%=o{cy@fqSlR1zKz`kn9XOxvATtW#9rjJK<9N4ijci(hkhSVespcQ4y zuq1#hFSaT~t-m^Ak$?p4!)^#FE^5dCQ+plI)}?GGnuz&35L!ebuz+#(tB#g|{c&DmjHj$i@ z6=$V>Efbpl(HllmMOWXS66oeO$g*U=No`--FYF<{yFcq8ip763bq|3sq;0nUllm=b zz6_^`lJKbt*0H^!PGJ=A(sUp9B$lVoP2%G$bM*+COLi!MhS9XmmIFfAOtEj^Y_VI& zg1bSSl!AUDSu&nnnUzO)vu>G?xL}$$H|c9DR=wD<4KOC7m1!I zqwo$5MATlKUUp(^14uk=H`pH$8CWwtIwRQDzg`$DoLOl-(07!{l8J}RQ$h{YoE7BP z1OKn$Re#8`Z$mfATTsMLZ$ z!>bs79oqz}I~@+jj(qIHCRF@(H5<}i!&g#thW)%euN{-mVR4f_5%>$Jf@(O{W)F~? zl_9QYMX=EM+wicLWJmsRkcMUD)dgSJQlNaC&olAtDct3St*5|uVegmj;wpARB%nLu zju2X{IuS2`4tZzHH%}O@54W?MiG6b)9jM^|Wqv2Vv~d%L}GQ*~s2o)v9+J@Y{gQtO}yAGZa*V@sJ>g!Fv5-_8z)Jon5$ z)xSL6@sjK;Xf^8@&MoZxqE`x&w-t6_;oqUpAsUzB)cBq%5fd&FhT;{1c8OAZPBBx6 zvF<8?o)l~lAP59{FJ4`8B~}IG&Nk;IdW4D!gt+jrJNT>|29Mo=ZWk*gvRnkEim0)h z7#Pd~Z~xmfI5o!K=n}k$V8z|>oF_ajh7vNPku9O^&?yxnB5stIdN290iU(EF{_5I+ z>^|6EJT|~xHC!l;0?8llUW(_>LbpFT>1VSk=JCa;*9$n_4&a%<4{s?uc^ONXEoD-7 z0Qq=7i9`=>2R`*G($_&<>m-yW9Zvps!yHg2s29MXZM2CYrw+5yp>ly;a(r@IP2~fm>EE&Dn_Y)h z1;&oUn+k1fGO)`IhJ_4Q+3)Yb$tNYZwUwSf1 zPQwhuJ9a@6Dgzu8y)Udrv(G*FTw@vp!KcsPnA)b)Z^ZR=$?KEr2n?qlGaaL5y_>DT zohcBM{J-7f9tjvqec0#rkTV^m(4Q)fqueX(ohM761+^v$rSgx3`%@9MRgeLu0=`BL zFOVB~L}^G{!akdWZ3#!j_I4NFF87;>pSuaniV;|b-<{90fwhJmr?wV@7ALN>V{6AW zM{Uo>F(qyNnDkiF(a2oN=XqW!uce%$yk2~b%{cCGpN;nLawGcA)Ny0R@p!U``aJ06 z!9#ip9l~Mt0zcK-3OC_H(L#|$yoe|H&f~8!C)%o)TL7zFYW7ZJsc`cCnR_ocg{4!U%jvPy4vem48{Ws z_)`0cHO|-3u^iQ3@8-q}{OaXa9%fGy!1jH>-1X*To6_EHlu2F@c^Irnvs®tPZ% zs{I_lhgYIFn}OOq+s3{~K0(zk?Y3WMTjP8(ZvSqtv_tJMO=WmGHI1Hy0Fu&|!?bqs z>~rm8_xxy^G z%_m$r1L)=wGC0Hs4jtXF(v)S?Akujp#v{@K&G3oo=E~0T80nDXuSIjL{M z>uH=j0x|voLZF)sC}k5zPw`LS14seJ#1d2VT)0Zu3sN+>a%sg-BC=U?jO#Ku>eY!R z>mE#YuAgVXGxYFs?Mjp1qhviQrLE9Pb}?IzeQ}^)-cI&Geam4&VL!jGo#?Nz4sX55 zEAxYm5-H)u{)Ob>pW8Xbjv%X@kee76b|nv!l{>o2l&H*yW2R~ifMKwKV6r>3y{K;U z_u{;4q@0OQBg}Jq9=Cxl)=XF@SGJOigdjkF)A=a0vXRgriGcH3$u+tI{HZ(8*Jd}_ z*3{Nyd2DO3-El_S6OYT-E{yFqGtNxJcJR!XY2Q?pYB&|6BIN}&Ac zBz8+1DaMMtzCB7kXM?$QhMNm9#;(m8XJvLyW^Ywnyfy2C%+|BocS`@E(w&6wK2HJC zMdR=)LNVKf)BmKAZlc*+Yr6MZ{|0ORt2-GL%a23jN5ODp`Kj9Cr=c%vew@5xh{%0o^am4KvL5i4&s7RYqsDY@t10!Bg1QnLJxZrRiQMD$HiaYcT)_aV#gGdb` z32$Rd5sCZUHuz-D%j>mqt4nwbV;J;O3NMah#Td#5YlEfgSe31@W~WK-lm_V?Z~}({ z!fzukhO9cO*r-iZq{}2~fB`zz7g zsl!BAxude9GaOYMz%eViAHj|ddvoeOlDeDI?EPtW9ZB+2cU@Xsn_{sbvjmt*dqTP; zc~r#|C)MOJy%kPG5Rep3u1m1(0)CwD#2my-hM0Od0Ng-NC3W9GE4keoFLS5Jb2cVO z^(WRjh24|w!5>vO6da`*6@-fCTk`ET=e8w3BDUMc_VCy~Ut{Ck#Mdl<@ zvwH+fbcjz(0BRW8m?j@cTqPw}_%Op^?nDFGFC3TXqO{5_2`X$g6rS;uusm>4gRHJ3 zIS0X7^iUZrM%}(m6jE4Mi)!Vll1I%>Ip)-c5WwdE~iA|^)iC#@p?aKtFk4&F1E9`!eQQS?MrO@ zQtLyCZug9}DHYUqU5MG^-|alBv+o-1GFeN<>^UPXe}vLp#CG4<1_cydK$hTSgU2li0M=qwL>8rM7jD{`p5u@tcIGjzg`tx9uKm@)fJDW1de4S$J z_;y?qJt4Nhc%n1dV#E#6F4&CJaDnxl((^I_aCu74sgh1WZ>PAI)%oN+xSMz@~{D7Fs)HdSO>aX$mxqb>pCMt}bDqrWfuovwkY3eFlq&0U7A3 zH%J&!mY^?^AE2v;QufyRW)s#!E`8AJ2D=zXK=>3N+d*adyTJM$H)e#7;^^Zr3o`JbAk2Q*2SI`?iDE|tT>&lNu&r}DeiRQ^ME z5l*waU8M-ta7|}?KL@i05Q6Ch!`oagKe=hkT_&x-PAbfJCH1gcc!6DM#PGX$0( z7dksZbOZNhX_4kj^qo0t7@kPJ_V`&tH%7%osxS8S!Y<1{+TqidPoIGz^WOl09Egj?n;0FAMs`w;KKo9Ju#88BorysYB3`_QZ>-!l z`A?vIloS^6CjeG{D3X3Kp8Rs7ngS0RhD^fcu*Ul3fZ9|uO8U)}hb8~8{AOwIu0ux!<+k>4G z?5%+;x#&6JG?)Y>3Ib@@!2NCvSR9~yRW;4BoZ;7&?FDSmv$=R^4U;6EtE6g1`Z-8O z8Qf~20U!oO$X4NfDa1&*Y}UAW5m2ib-+DYILKiRwbnD5=?UoXv7~bpLw%O8j+oQ3` zv6*F~m<j1o*M8}V{1oH?g}`Dtbsp$UmnTea{R#;?dmasj-V2i7e>YVK7N%JL@8rW`2-*! zbOD~D*waSmAAc^zBS3_NK%h^@T2S8yE4WWN|7m?&^@T>ju%8r=gv=vDK(ge%$5v~nC_{zEpz{OeQ63%WL2l_tcftGz&$Ili+*kfZ~py3;dl{^7bw7zvKNk*}FC?eFVDB60r-ICiXN=S6iolf?_OQwe1TWbb6&e z(v~}tH@+$*tNLmxZ6spCK^H|b~qT_5( z3m)B*3ajLeYzP8L&1$a7!-EhX0a!o4)qrJLS%Y@jQbxQef5a4IKb-{Ow-^;3Ap(rf ze#+8Mn|<62jTBDaC=~4H>ErSsJep?WclZIx#=s0+Qr`)<5Cp=Y={Y`ELLbnmpx8=S zLu~##MQi_fY+oMRU1Hl$s*+@7-*o>}`0U-b>qXWVi=YL3$_ZXL-tpc0?HznuFt~EToE7U z^5VaXmhq2%fHmw_nV%KOYD^j$8Q8o zWPPms_Hxc(jQ-sY7JYiQrTX8Mf|jD_V){UZNcf1hUF$_!26W2OS3CS~=i?^C{;Jld zcWI-C8snla*d6t0W%DAS zna=gV#M$e{%GD9>Ij?8Q8LkAGt zBb5~P82o;wbYX&IUn+{yJ5syV`~Q<1~#%j}sNq z0-FzFo#iTz@UK$B=>}Q#9f+v!^uc8Hw#gd739ND4*tN?pSKVA9T_!?tRNr)lF&g>s)B$fA|kSYD5zKy5f#fSBIruX z@AEzPoymma`rkj{p7+jer#CJ&sf3sE^x3g$3PL=p? z%=bgBoosPfB8LV1Tj1m()n##BScXkK7j|aX8-muHX&#hYARm&{oesblmvIbhx&xDD zT*{hhqtwhxm1enw_M+5i)=TZpHtYk_RsM3j2?E)_Bb0F{($)B!G+HA|b6o2_5$E|B#mm44Cm}goNEBZ#Ou63-H4=kC2W4wAFMlBLh`l)6CMdF-pU?_M7v4r zm)gZ`RDvO&JrGB+wL;sFyr=Otbf@%iS|(H&3J}|YxMGy?YO_6y=ybZUZAR=tHc(bO>2LTC5eJnsfUAi= z4y6I2P>`meWn{PQOiILZATU02hL}NFyiL|^6Q~)Rd|C`movZ9&aVNhLm^0qK8*$;g z0qbf5d$D}uOOnK_gEZZW z+eIF2QPQ#79@mFC^m`yWm4kUu5S0-w>>I7ODN1uV;5v@<>f~(N3cBh;f9zv9*Rz7Q z<(>6r_Bw+<`A{hFEidzatw({+1P++*>hj(ymB)f&exq{}fADUg<&R{Ue z4nbix9Ug0K)rdSo6Imp@kG!_^_kp3wtz?3|b{3{}3eX+aFl=JB!ElN5yq8Wl57sQ8~Sryop%#J5m z_f+P}mKu{MktO~@6FmxTfNvjSM0K=bYqQI92tl!HzCDYBBo$)ZzATZAzs=6-@H5;@ z0A+@sna+ZQxXk-2fd+vNKc6#|%Z$0&n5$G~SD6Y)6d4p#T?F>H9Y#`tB7bK6&$Y?h zZjeNBo2d}D0~YYBDzmRL{GcP8z)SBi>yzZGAX{@6M4kI#ZbRrb^Qb2dATo{z5_e7R zrvzU!B>wr9v6leG@Vn?z>;?&O?~1wV;lIDdJL1gjm;GSePUuCf8o)tpF^C43i=x8#!tJENzUv1)O~=W$^&Ag2opwH{W8ti-W- zf@#2dkYB9Juukk3T=9f=9oqG{qF`1l?5N#?jm>XOU&QkK-r?tXwCo>bxSB&g&{I~^ z?kwO+Rtef#?AX##HUiApadO$dRtfe>*wR&}H^@Fpieq6E{x#9uK#- z$Z*e=Qto!(sUf0%+9HyGS`H)A<>6$gHVg`b@EggqVKdYf-4hQ)cX0?IgM{PgBBF}b zWg9_7q7XcW$fC)MHS7bDc|$-cy`e0~JP=NMRJ+_nUrV5|*;e$D$X>(rVor~LIe(>n zOi7=Q$65>htqd{@a+q4BEQUF(+g+G#LR3n6Dz*}vb{lVWyKiu2qw^c!Pimvso)!zU zOytpfP?O4>fZ_1&XRf}{nV&fGQ|D)&raWdnjy!WJI%OI z?uDtOhJgK^MWOKd@DV5`2!4fXC-HYmPLyI&rnaDCc)+;(jl15s>!>t?mIKsjl6Dc! zx*;`-RdFzlmnsCYYVhBcgd9p0hyn5>U7eZClIlSIxwvyfXG_VIhF~Ic4sa+-R^&8# zPq!n2dn~KP?T5hn>&;>7v@5giSECg|0;bmuCa>zb)l_aVjYVc*$D(|pTU3RkAH<#d zTI1xfg3O9p#xhETnsd)E(!BaC7!cop>1IyH?0nqKY}uB{QxNIDqE9mWN@4`5(MGsJ zs8+ElbQek~!6Lr`fp3zI;AkwH*!B60b-7ueH`nFOXaz`hDfyuW6SrN2_QR)x=KV!$ z5Zr0#@Bmlf!wI4l4H_nupg{(l&7=pbik>3CW9oJnv6B`R^Es`CIbk5WLh085-c;yA zh54(nyk87}FIpCzZSWB&DnCJ5LO(7C648u#0Uwy_}K-( zOMs2J6BGjX#MEVRC6Oc>K+!nl=|Gt7G^I~N09lXcDo{FsJGiG$jw9@ucwYxCPWlXPDrD^ z{V;PN_~7~*Uty7#OFAS9WjPW4TQX*2_K5l_(ZkNX z%N_&$s3pOu$Qy>y+$Qv`AeRu;RHDA}5|ZfdU{=^2c8MNy=?5*AMrMCv)NBlCr5}~? zTb@c&r@oB5@N$3d2CG}vI(g5hb1EiJP!(h8Lz`+z3-jTBUoB70PINEq+#n2+(*bD zN%KH7F>!#~ts5gTSJ7EtME9|00AgUK64qjjo^|^xKLV2ySgKy}RgDvp=yFuF zGUa4t;eUwpB0H8Q$RkI9f#kO(vfICykXb13%F?EUbfc@xus$?hMeCDHb_XOuNRFPK zYYvURMl#IWHK-0VLunnpJs`M$SZBS}q4yUZ+Hd_^2_Eg|Xx3pN zFl0u>5Hte-lC%WjKUO>7&p-aF;)hXn-IW@abI7xiU?NYkir67gogTJ2zQEQaT+quZ z4%)Npl`T;NCLmKr}A;@ddlVvva-^ckMV#oS9y;9t~w8GU3ll&rjCvEmdGkb%y z%E<xD4$WlPk@BB8N9WG8wf$f zP?SJ}Y4ixya?v@eTDk3ENpzPfU10{UH0}yhz0#N~jK7jO+$}KQM|R_qVh;z*pV01x zSekcL%Cobp#kqQ2EO`|o&!d?@&i97=m$$Cb?O5dwse3kmemm3Fd4kTHj} zDcMFbgZoasYW8-o=Dl4Jn(UJ8Xm>I|24vn2EO+?yTtc%kl%MblF+_Q3PBznIxv*8m z>};0#rKDglIZ5uBk$|h*Jmzvrt)2OWHV=AAayS&)KCs_sZT3lk3d7SJJ=&D#%}o4d z#)8TnYgcoWJ(2DPVAGHtN^=kyFzoMQ_2fm=FEMo(B8fFbkQ`RkRt#t$8YKT{$L8gA zz%%jbI4X8WBBS+JyBZ4=7gM14pFA3-NASx*j#^++td}QX7PJY18C)OBjU>&|G$9Bw zdpa9^16i%C<~m?_$Gfu5mF*!oEYDzDFp%R7OBjAR(hg?RIy+KCmI=`iO|zIomU+u8 zHCg9~a6efJ&e$o|`Erx)9qD?7D<01P2zg9(>gH<#kL8% z5cABD(GnJQ5I`_Bj_PTVOc#&=fqezRkOSJeSfS`vo368kNSB3%3<}Jl%wc+l95%Li z?y06dLQSHk*BdDDgqk|{iT7>nWA~jG9m%KoX8Z#mKZJS?<&a4p@b-QtTv5bYxYW8{ zSiyRWLK&fhLe~;B$#$osfCk*JrRG5zz;g-n{(4P&tlcz(z--y@bwCW=&4tr}GVVyu zjT0Fik;xJr4XtibPwoUvK1}u+_geEB^V+%5`HY0Q)4O#Fn`gd4qIn-f9&6EKNCf*P z>rKdo62cB|`%~^or7B!7m$ASHWQ~aNjx`u*+-(*`SL=I|ZGfwpXoGf{Rc45U+C)9} zLI$ir9@Kz@lM*8GqT&(yL5G`A48_d8JB)NiNfGY`(L@iTGJ~Taoht4a4g)6`V_Bw= zmA4WAT-p%S3iwmWWdp}Hxv5L4f+tJ0#TslAW?0(S1i~f=7TFb1uC(&vdbTOrymBNc zjb-1JQu~tVNnk~>do%=4F1*fy6CK+D;lZZKLyhQEPz$$!1i;7m!cg(`VW@m4heA0b z3jb~dWzkRBKXBIB0uzDdmbqX=%Dg6SM?wu-0uAqk+)e=|_BPvR?eRWSagkl&cHZLe z-NeYXhV?TtO{SfT{fK?FlBlFeV(23iDus=@fk7}*K`8>;;P$H+HneaFYT`*up>R-s zr|w|UsR`LFItkb=;aU@u$Kj>%n(a{u6Nl>BqN`x_!b{4CKz{+p)dqZm3q|ulvvdUlWh|6O%bDi^Ge4NM7%-x?^)?Phni1KDD1Xc+ zt=&>|a+~6YGu77KY0M=iz0{bm8*?XwEOiADioc(#iT`5}qOhk64YIZr>q4-P@&S#j zj!q=n%^>D-_9*ax1b^s@^|Bq+J5vQJ%W$HYm8Op8H;YsB8Psx+bi&vB&>L8yJ>7ow zXi*6)DPvsBzUsPujd6pe)v&$q5wRqh3HqJROcSIU`-}Aeu7TVOlcs1TaE|!OVz1T} zCTeHRQF21Yay%e<1HRjXsMp7ZU%_4pOmY=7ui|?JS2~r|C&FOZ@7NNTOOFYY42w-w z3?(fv=GAJgBlQ_!HTbAMY%Vm@zOtQ3BOP8}vV+X-2>w?YTr>#Kxj$8R}rn z?I1otIf4fXf5+hqT@Mmic08=_)=UboR}_LKAs}M9QxzR;i3|=0VNH=utcE(UurLTCvV4FU^G@ihzZi0M3Do)LJvV z2JYlxb^&~8XqGZ1DC|j?P6jNv{)oQ;vUncvQL|IJV;ubeQgtw?A#o9d2Lt712N=YA zEWn&a%y{qyI~g#gG36W+s>$m`swKC<4n`-Nq3*Gs7}ETBT*L@?TP&M61N?zlDhmHI za!@4Z3J)#njw6iD<;u^{qh1lytRf}j( zuz(+5oM3`kj*E7DAyr=o@=F+}p^;XpS zX>6*Jn3$pJxY#t;4pqF!jYFG~{IqqxKjy0t$$o~cSku9zH|=Ztp!lz8tT+6=OgAVC zT*OFsVX-1{!xe`6Y&VQ8%DDv2HjXbb{(5{PguOUGZiaT~nsw{% zHRhYf-z}u*L5kn`|Mc|FAeTgHz##mZfEz%AX=gHQ) z+OLeE7ol_?_8qOLdD<~E0yOb#%_6)O$tCA^{L;et2zwn1AMX&tXGj67pc(^&%vcO( zkl01}wiV8>+}?q3s6=z<jmDaKOywt8JgT8fa zKF>P(n@Lx*v)R+vUGKrZfP;pz_=MUoFad9mhVpfxWXQnwMd_!UqxJ4!!49%u4`q=w zI51J|@vpF$lBJ!?=0sn_Fu9yzj0&-AE<@wjS}Ew`QUW@hWzw@vXQ@oE`a2<$O5Ze1 zREZz8@#(A)M_8R=1+22>+2DEhH{+3?o(rC11)OcoCqj~kGvqXtTmjn3#r4^E*FSBnCFd&7Xyz^2BJE6Qma65utSs(4dL9!ZPKry`f%^tqH&KE?n zeU2SskegCSrfwKDO#`7nCL9=K=0TAsWuBRGp42+4%99KQ23hp(brwkufuHsBx= ziYKi3z4d>v<~Pj{LWmj8L=_G`d zLklSm^x(h|!F$#satdE3HNqgigXUZaN^_~Pmx$H8-hiz6tDs9WsM9+P*7%ngdxvTE z_C&);U%E|lY0#MMO{|3r67!S9{x~t8M{^PXIx#;>%+DdXlAk2mkD<$AEPZw~k0kc2 zq$wrw{FXvT%A>-*r3@%Sb0|dfK+v5bEL85M{bcOG_8}n!B7(qTupvVdfNh}baMy=! zlGB52mV4+@@}kDpq63oqT1{AN?hQ@Y8Y`uOzL6GQ;dtN?WITk)neNb(AmR}xE8q3? zW_lR484|odV$yZI}8od?n;SN=a{lV8lg|m~}_*?c)-RAmmWm?i~kyr#V zp}N8-pdhl;V=Y)u$Vb*C9F}%6V#O(&DC`#J3Kis6nK|$|KQ@)QUexG3+tb|D&8sM@ zuQNlYeN{ET*&qFxCu={lj!={k3j71+fZ0a!jDPUDVa5VSzy z{&%=2_c0YQ{P?3L{i?x9_h@%a{Dull7BYuAYQqM9hEvQ3+cM^G5UU_>6i8E~vVL%z@^h=KhduHhyh$AL4mV3#2ZTpY|S_t5jIh1nRuV z{wWnW(MCabiZI2P4YB-|v!F)gSV-D2teHP33@6)ZM?!%fZ&0#sNm7U?Y5xp4|U}7%!?lSK_ z=j}BNVwf}A_6&1dc~+mwH%U4ZDWv79~~V*GuG=~{;Ar<;7Q%AVbLOoDjMHI<9t zK4gb3n0dWsFC|AU9eTHU&s(Cq*fda}h?b)fQ`N3Pv{v{xGaW(I0u{(WQv$V8y9O7_ za!|Vl0h_}f5^j$fHgAc3sxMa<@0a#oR*mQkL`=BL*1N{_Fv8s3j#eVKV+MJ{r?T&7V`F5Aw1|uR2jC|mv|8F(_<43f zCLwgOv*;aNtdNfGBp@9I3~>#*h#a%y0=G${GYrDyyIsCZS23xfQVU~tq1hk~jO2%H zREC-9#p3;FIw~Mp-j>>zkTxW3!>tWkq?XmqUX0Lkvx3?1<=2zxrXD065QVU|h+**? zLsR5^{lHMWBYuuNLY^C%=IH>JP?en6`VdG@QFBlDy z9H&l(BFEMEVx&dSKwx5QB;hdhRwE5|q9!hQ;ljiRrIHZ97;8RYEW6VtDM||e*r}xt z*AwN+TWhjY3~{yy=Gn)mAF%$LP}NI7CzD4K56NB@9ug4;5V++s&~9EViQ;-WSaI;T3WjWyd3glnfXE2sKQj5{!dAV%Wz_?U2%;erZ7a%WMs_ zzpR1%d|<-hiyufz-(mN~JW8^J4to&TzlXp~lEJD6NU`l`SVw|&&<(_(dn(Q zmq8Z_dfj3JXhgeBR&Zldx=9r*4izl2+fu=jq5?H^K>?m>XyZkVTnRz~XozjTcX)R% zJIcHJx_q*68I3FUrdH$Re%@-FBIwdM8L6=|0dMR~$V`7tWDvf%br_FrNT#r>+opnW zg6)(|Ok)t+_fbvQ0Ix)gd!;G;tNVcaSNnnP4>vzzKAe9f{cvaaZA2>rF2C5j&w6*c zx0h|k>$3cYP5=zP^14oFo zlg+lm@`OH~_>LR;ZQ-(vu ztCxe1E`^*CKKdN+f^BNF2Wynw`g4$g2W0%@Gb+=&qw^RdmsKYAS9qY0tg|pby`d}^ zo@wDcUuXMJa$IW;tF4?9jnvuZQbp;EKA=;3i;S_@|pnNToyyZZrRP9uf{d-GJ?FD7zr zubbsoMwnBh4VGio%3*}3o={C!qqI0m(lMeB7eUlvv;vegz#9x(9kwNG`^tt>kB-~? z?5RM(#CumWg4c8KpSREwWE=p9#imda(h5czgd};Qv>>bO z!ho`!EVHjRV@ytObcdEmnB$;PUQt=98E$QW1*xDJW($;A2<)t&SufrR^`hCm6B`P& zgowl0EYqPlIzMT>EpiiG4B!Se44JIvZ*!QB?$AsSyxJ=!EZcU{?#586vaI+Fkxtp8 zA*Cz}q;40G3aM6tZ3QYp?JPT3O6UK|gSeb(DS_UlH4}SV!+I5?9BgSm;G^x|6d@6# z&9(RgF@pNJ}M$833Baj_*V&hm+(dcK~EwEjhT7NB&FbW*0bfG z;hH<46K8=X=LS|VK0f&jz+?kQ2$CI)kGJO<{@g!0Q%Y>j5H?s423qHx>E}&Suz@%n zOSax^v7E+h_(#SfBa;P2xNd%F>?7RAk6DsHg@IteK=mH&!3c&WXw+K--k$VNM$9V; z5^t98MQYUnRo0*K1lg597o#YJ0#}HXiuK30v`q$5-iCw-YyH*%sTS#thQs-^i7QRI zyOv~d$h43XV`)0hIHIlbBi%$=)tS-{l@+<*4Xi=6y{18knc|cZZ~b3@9H*eo41qZ1k=;V;sSteVnbR z9t=!h`kYkE6VJU`I}US$8e#~$hfZXgWTb_M!ZsZ&g1o^F))%(-_J`DjiTf%`x|cO2 zwnRX9{FKB`4332}84UbBwg3qODg8L27luPlkokng)VNl>szrr~LWwEyZ`kUPhO zpFEW9p-eVmo_oIOnKBK>O*ucdm=XH@PtHiWH6ud%|C<@X3e~2}NM|>qr3pxvnb}K6 zdg_dfgc(66_cG5-osp%?2&<)|H6!g|M#y8l#f%KMH~WKNE5Xr7M$DUM2hM0t_KjT;?0!3nP zc~F3i30Q)m6?|$Cu!RU2R>FbEmU}AQoslp9HK{zEw6NQRodyc5;Eus->^BpfPf)HA z;*#*Jwsbyq0HUA9KhT#xf)Vld=EL|Dk~Tn`NbtUca>IGBgrxagguuSEo81*jF0wHG z!nBQufsr?)vP|p`!zK$t$6dXQijAksYx8(b&VxCBw?{sIxEwzOPP2;foX~{O^@KJ+ z+^MnHy##UC%Fc>ViV*?mr zr$umWo+R(AdUzsvEn`L%D}u!vw)|7Cj4G;~&pJArO-bggJ;C2+h^*8AKxJXu0ISps zIfKM<4b3R8P2YfNUwExMkQ}&Sw%%Kfnt|5oJyH$?dOW;G9D&nK*gGUOf zL1}|Uc2MYZtKmEeZ1A9a1dJl}~TC{tkj2p*XbmChKp;ISe$&e-Dtd8K_4GgybHV zGi$igX}_eidvylg+vq|a_>aeunhz6l%h(Sj4_f;jo81e}0dl@?Eoa}h?t98`!jh*2 zKbm~DHu)4i=PBtjkjlT}eZnA`A5Pv)?o-cSQq!Z>eA6ZOx%fXZZ|+zR(FOLJv3?CG)_aEFMNQOsDwQHY33Mq5#$(2xSomCROa0uVE=EO9@&& zY{pz5>uDk`6pG%&0t23z6Hi4$@@8R!;3mJ2`T;x*;bCT*IfS=?*wlCU7z8j(T^CJe zlVssRUb82$3L&R3QsZlAJup*H%?edr{GDK2l3;O!N`!C34rgyRMHV@ERNDO9fWw_7 zf)L+~q5*Q1?r%e|QO(B#Ci=N#Y1j|4x`-)>8cQbL(u43%Ml!Hpu8%0HAlz|Sw+@a9 zM-~n?mw6awEW*DdMAw6YF3}F~hc#y<**OV@d0C$1v)Kx?Lf)(`lXQkyQE_D+lRz|N zP~A2g3`X*CD19hL0`e1jK7zEnQhS(3%h`rtObYs{CHMh?b+M}&F5u20s%2{kY7{&l zsR*h-9*&@L$m&5orM?_fApIk#cfIKkL|!wL9icorl3J3R9~FB>nypUp3PvKwf+{tV@`uf{N#z9r7L$!WUjGTKpqoHTe=H@xEtLpC z^!Fq^2QouSMY{Y*$$z%poZ0SgDw&%yyFSa;W%puTa?E~HC*}1Skac+}z62>Q z;oVpk~_LBgOA(T`#1)naFyhvT8GMtO3qV@5VTot4cbYdlh- z1#2WFN<+C*D7Tw(WpK)QqHI_G7-on}nv8O0j*8AF_<|N8xlJ+2XD|{s1N)#am)z}S z@5t^fHO?y8b1;w4Kh2W(Tg?$jffxOW&Ft1Jza?WflJAn<(flCK9?Ih+{etTvewpJN zRT!(UxWltkhTqB!4XE$x<4U- zn9ca1@~8cSj3)kf@_8XIdW9euy|qSlh*4SPj@V?8w=E21C6sMj7recRdxZRvjNp3br-GGvDx2G6Av@zHXtaalqUi2KZ9B3O#;PigPUFM#|Vg^8mPoBt|L5PXd%%Kkf3o=F}koBPWqlwA*^9RUML*0f>D zGqp#{<`?DoQMO9^t0~vh$IAAX<@#e~^UJacW&Z23IioUJ*J=fMqVCNt`a{|NzFhx9 z+5En2LYY5BJRACT;pAJjTPx<4iXF8h{^APA7{7e;sj7LVV(yMx+s;+e{(hpe`Mb^Y zDvb4~rd;WIs%-vJHlfVVL%nZq-(15T75?&fR4lb_ZrcSF`Ob5y)-io@{I(7 zvlRfLtJ;A@AXob?oC5_HOVOQhb5i_dV z{D_~0C6qE(V@N}u9_|qmX}X=EKnq}+>w?94d_;?dpR9+N~dT8FsWB9IF*8E+M2F6r82=9{E4fULAn^YHb4W zYq!?yH_8?RFhTM9nq5=d3};z`c1xP?RRbUEYo>rJ;-iB zfi$%f9H^KE!~B`pO=6G$mR{OZ#cX?ogK%v0asg(@eV9oS2Rd(S3z~t%;mzu9V1GG; ztplw$IZ6<(qM2j1`7|_ND)DtD&)C2Q)s#9-;&8AbudpYRwIysD*zN^5YhcNVqedGe zYpTRwWHzzx2e%@-+$9Jq?=EDQALFMQ`B@m5fXX;9@MnLA6lE7i5)EyL?I@A8>dhk3-t-?hop_2ij)`h3NF-r4o2 zWXyH-iK(sg31d#Ihnc;xZqBN2p0BYB>ISNv&*`JvmcKgrR^tf&xrVu>VXwu%j_sNr zCpR=)xjqI{b~OVH&i@rKWPi=)>Y>4a)VcLM)JAGe^U*r^P5u0(a)BzPzHluTbI|L08g)bnXx*Y0-II8kW$jn$=!F(`=d2GkRe0=*z3KP?&6yCWm{cszy$ODgqT%P*iBUEC&~7R1b? zLusm=tYtL7665Z$#-Ms>3*|1MoUsO4F(Ftu&~P%bL4putnAHa+3Fq=5WQ~M#X}K{D59hR) z>v%IAV>5M9-r|{dN z+(7WGf8w@G3!99E5(;}v5Beii4}ZZ{WKN?)W;*p~gMOqbV}B(`-QzjUX-wHeRt}ZfI5kXG(A#-2hYO#L%imkC$GaHszDYpiHyJ74N>M z=*PAg!MF5ji`>>h9kyx7(GHL`-!;2GGccSn6lT`>F%tgsLOM~QaO zRREN~Xb>qKSShTQTMbgFs=dXyFB|)9?DMvKu70=iKPr-cT2z@}aD`NuCjcekQh@l`8by24ajt-e=sICO zqm62%HhfDUZKT&`0F~sa*wC44?&8SA2NVg0-72&k$Y#?9~YTb!8@O`3vX*tCSs)jI! zJ+u!Oe1)-`PI^KqnS@@s8uJw#{rJj`lV5rBOTS|MDpS+)8OJL49Kb28Qh0VH8by%P z7PbM5bCJg8vLQ;5{Rr)?G9S`wqbCq@t45nO0B_~k8dXWCHnty{toR~BsKmuu{ly9d zRcoJtmO@V?;{xWKzZI1aK+)fUrso>!8Q!5s!$4;pRvklGY-ht51XaI9g>Ud1H;99Y z^H3=Jr};x^{BV(&Be7=?S61xk#K?rUviofv`f+^G4}u0x^y4u4u`K%QL_fX`HQ1Vm zM@e6Ubb3lh)X<_MMZ-6D1mPZyS4SlepX|tVIv3Oy@TC)xjxkQ#&iP54s@F zxz|Q4Eo?hGvmOv2GGWuzQ3An`*3U>_IBrL1h=iVV5&2WGd98VkCl3TqifXX&;%qQc zoJTeI?P?S1KEDmGU}LYvvjQ@2mZp#l2LX!;ld=YcAT>z@YLFRks=U}T_Cgbq_GIY# z@8e@Y+P{R6KZadjl2#(AYO5(ji2XF!IjuK^-#kL@GKf5_Ai>mF(;6u1b`0Ybvj=Sk z#)GoPWkd<@FX6U2Lwp2{ankRaNa%`A99wi^1P7jJa1B+c)6fy{M2un{t^gVmQRIS6 zgulKV5=xFdBv28qVDji^%)lY2O5UMKv)?WjPUHjsH z;Qxt##T$b}LHbAk^i&^2@<00b&pxBJ6}77L5;U?^=1f-pKmM|Tz07l%7ZDB>UFiSv zSMQ?|gy;C*{i@#oKks8Kk+67yw&Y?#WB8X}QNRf34loI%?UY|ACecUozyk)@rk93o z8ZA;BC5`qJDNvn*P!sL0tvlLDM$4c~8$Z|qwTrNcsjJ8bO8qNxV^MPWcJ`++a4#50cyqHf}AE@h}749$avsf__PVng5i}+$svBDlQ0`I4qL?u z{{^L{mn{{jouGiyYdppPvNE)SVBBcAQjrxGPG($-vvXQ!d2nSZuHDHs1n}@r-(L5oO9k>Q+jCVvv;?Nc^85A`byjK+*O9ttj@``2<` zLxVyF62z6+B1(*>3_#W!^n`2xC3e8>=wl#xx2xL?)<%yML%q=qZ)mAAOKkp(HW*&J z3s?cHoXE|i9ap;=G_v-+#bMtohZ1iDK0bsp!0y%pt5*~BNme%N%|L)wuZ`e=FrZkdFlIbiSK6-7g{7hK@g{mi5TLpnjLu^S z(bzGFpv+5b{cKtLo@mWUwz|oQnS!5aWH<3D=z8t9| z;$_RKTI9|iEpjKRVO%lxqQ)6nLd4kUeiU2m3dZ34Vg?nVvNQ+0*blZKiehcsW`JYG;zt!NG+X%(m0hoV3sM zVwu$|Gm8}ajy7zkfUrPkT7q1#b67Ldy3&`FS@Z(Neh!dW30Z9=`ZYAr?>Le&lGWp) zKZcM&iZFdA>rVN16&eQPl>ptMfWQ*`$3hh+K0*C7!G1`;pi@)1C0*dvs<8hh*ARh%!DR< zIq>OY>BGkZqhQxUN&wapX$bD|emfY5eUZ~jBl%Iut5%ayNWaK>PSY=BG#pQC&Nf7i z=l48vwkliK{eYpQ-!LbN-mAozjbb&=v!mSPsgxu)^X0fW6XZzDlS%SgFeeD= zi;~?4i=<5kHM<>BXD7%#1a*F-7>W)*I57~;D&qbs`y~@r(2)3rWTS~f?onok9f}Es z025`02v$ZL=eN^#C`!;ebh$opiZhQ>idt9?ZO@L6NC z4FBqmAUWY*o1fp}Cw8Hk(y$ z&4IjyR50&hE1;p_pSSPEZh(09=1l8KweB?Cp4Ex`5`Tjc6*KuK zMIJ(ak@^#DG>;t+A{IGk$wBK|;;W%*n({A9!<4@_-)qsnv@y__{Dyy1{+_rVekFLC zd~%CBFTVcIKUuuV0zekE`3UVOz=x?@*_g1!$M#B;oKmrf7Fc<%C9aoEhR-3s00iNK z@j3Q83b!VN!qnM^fCw#*;$Q>*2Y&&z41dtW3a^_G+9GK{E;mxga@!)Wq=6J1vyy?`m&P! zH@d)pwh@7T3{}}8c=nlaT@`S!4m^Jq*ir$p@GNx?S|XJ3C3a&?0ob$wjsamw#3)H< zzzZGRUC1$);{>)ANElG`MP`jNYh77c`R3hhkWUJ*XKEaWGvaoMqosLHa)G2YeLjhV z`-1>k*b4DeA)%qfF`ZXQ^njd2>V;Z}wwbV<1>oUN_TmxTmO<`hfC5d(67RF56DS&Q z>@j>XhJ=2Sk2+KVbFgfgBhbOH$yl5TiGoL`P^+Du+8>hu(3nTLH}#+Yld&hFzyZ4# z;5c93)CPjgk{bDb*uIKSIt+(ly=QE)SomxmO6+_fM9Wk?H&kzC`#EOz+~^)4hwMS^ zP*)167PqyFWD%LdUpm-I;{hj)q&%9??g5UBilmyrQIhIuamNV%J!#wmvOkjk1me2` zN1twurhFBZ{q|iVUSfrkhT2DX$T63#SrjT81pzOhf1jH8Q=T~3#;}IgC!^~yZA(rv z;U{@XIV)VLuS0F&=bmkn0Ko)n0glU*bXiAbPHA?`k+hB-&Nlfw;&-xqw9L*TlKyk( zSUZk1M`(?Mr8~?EjpW&ed9GofZkWF|%71E@e~phd{}-pKBsm4}%RZ){4BvOZBahdVor-mNMAO( zG99m!N_kuB;e$)T~VxuYr~Ib8a;C8;rftk9(GRK?9Lla&ke@i zXxt6P+-Tw(Og!;MBU~||8;lQM;9cILHtOX1FW}wwPmRN4`nzd@~LAX)SLgq0NhLV90LtxkASh2n=2kO2MumA*fZmjZ!l>}vesz>mQa)uGu3JzokduyfqQQ2HRRZlHPdi#(|A9m&;*Yi{7ngr`^#|*N=zMk61g9R1*C7*2cB>8gX1-}#IPaq8$ zcpMzItGl(srS@5fec|J}-%3`a_HCeqrt3*ChH^8pvd90XR0qkAm@#5M)fx?SQ)N5Y zR`H{aMbr3l?$@!z=4k5aCxmvO09bqsr9Rz8xJccrD~$*chUU2vo<-P6!mY5K&{=ju%0Bobv;T9}BTJFY89?BmRnL$05O(WA8HqWdt zpU>wjt>1ALMF(r$$;cVNkUY7o&b#qB0y|hsU_N%`jV2mgMw{x}|ad?~T zMNVdyLIa3AnJ7nuk$7vYXXpFn3{>kM%S8Vn4Z}7L83R@=+d~?@Cfkp*gIiiweFGok z*uICxLFV-F!TBh_mh&FjkIZ}^@2Eoa3MjDiWEhJc<(IpoP(H~MKCyxHjY4y6y}yFyvpvMfbe7vr`I$Gk)1Lw4s^$W%uo zdQ|SDq`fik&EA=!lm%-rI@mnVn|GDTj!T!};F6?s)n%Yfq$Jh{Y5aZOhM-Lbpr^1j zL@bD^^-~6-B0a-oCt)c#g%|+5S|P&=yBDO%ArxGlHU(&sw-+#LN8AjI!sTZo*~WN% zi9oE2@z{m&n8%jCJ7bYz%#5c)nE<}&TEAkAf>+E7W%tB+HOhwR=?FdCfnzZ$mo`2s zI+*6)oW?gOoCJULeAi_6n&~-Gw(K?{skX>KO`qrz&0sG_KgV$ER{<*^m3z^LkK6Z2 zDKyNpBQ$UayaN_PL98+(A{M$-lvFf{laiGVb%z)oG>CvT2%j>7{QW>J^7d-A=L@L` zzgMILSj?U0ZSczFVWHx^1_OPdF*75tob^y)d&lyaCBsRg!y z!s877EW6BQ=eZ23<4l)*!eytrY$;!Pr+u$|pL?%)8?W5pG7N2eev8Y#=(4wP_gI2; zf5d&*9BXeP^u9d^A@0iLYHP2u^{cJ9%9>C*nxmbU>S#lHy)(BsQ!U-?O5b$Zoi6*j z%bs)D=f#{J2Rg?Boqr-opm+GEl!*EdaRcsf{X@P`Aqi?RNpCKat8BH zn#w#9fpo zu|s8;_GTR!A;QtlArU8lZ|64X{j_bGwVK>wI=KTIW^+4{0?J)_b{ZOGDfmIwiY!asL>1#gI@Ua&r0N z2X+{yxI2toku!nuW49HWDYN6Mk=@a~coqkMBVELUTJ@UBPig^(J$AqF*-uH+87r7# z6?p2zR+H2nlC)q!j~*uuH6sc_+cvt+U`a;Sguxj1&0oe@Q5=hIuGIry1tzEh@tm#DdoK?gMz@HlG zrwAWV59gHKQFf*Fdab&_=*ct%4Q0cDj-(2C&x0Z|bcK(Z@noD$k;Pq2*W}sFX?9bZ zU7ltOTQRaOr%MBOB@V0ByV~?MdgsAR zwZm@NGYy-C4*^@zzO^pyXJ6;{OJ9c|zGkr!y*0x%9&h@yj9&m(x`6zeb{D(Ay;a&d zBD~I*5LxKyN)um(D>+LDY%cmka#3tPkrV%#e2T1-N%#$cDQs zzesD3rtNoby#=+zu&=4^drSX1;!= zX|J`l!@+`KkdvfMQgzfi+x;l6@33z9YPTjgYx8O-eJJge^Ww_XyTz5MW#^PH96jZ{ zct@CvYBGODZazVQB1c5f*; z7ri_L8Owr2G=n9lqy&lGnibM(cMInPdDChM=ZfB35%~~k>3QKAp$52rwi+uUTwA1w zayj!LDGoPZKjnOE@rk4gFsy$9B=RH zrMU<|7P8vQ)#YUCt^1+2ic&C&G`o=Bcd)d-AsP+B!R^I{CMi)5bXuQ>VnZE*%Vm&M z+Ij^u3sPycX%#9C$s$B#36yvOC7!gyzckNQ_>qKcX$gdQebkO@4OvZ8?|7rzX@fHx z-I7r58p`FNYzyT;D3c9Hva&mD^aO|mtp^1q<{hBSWTs1rA=x@>hvN6RckYzzh-!Ax z7(IFl#Y@(;iQkFq|L7M+GsC<*vt4+gXEh9uNHq)oL&nIQg#j| zAo*>@KT$Eit+?O8Z)e}HWRFy!9mRZ;Crc~X-e0v3R4L5^Rr72mdx87NijumX;A`xx z1d`G@*e^()re34gfCu!3@Gwh_J(q|Gt|DV~{uXl*m}PFa+4!nj_PJVmO)XnnOV`x0 zb86|QYT21J5(u7H%bu*JdGecTwy~OSsAi8;(;rr|2din>f1{e+rhC^_v#a&@h1KkA zev?zG*?(8kzg4o|Rnp&9vIi^acPiQ4Fl3eNwo3YyN_KN4y{VF2TS-4($u6&?msPU! zD(ShE?956^Fi|PK=s;I!Jc+LDOxC`OyVB(LT6S|Sl^;VF+621M@X2p$*`q|lP99RQ zy9s26`xM}>+#3nx89_=4+yz?M8p5E+8i$|(1Aqd-75GdN&NdFz4J}7S#YfDgv<&o1 zY=5P}79$Il5iZUN74ZnB{?;o6l(Kp7BXXP&)X;i3s)G?2d^m)Ln_X%W`6Dj(5a+%^ zvWyrrywnbKj`!e|)A?Jn#}Ms8WcOCSJPi}38AudInliGKHu!lK_mOBdQ08Hkm?1S5 zHZFt_u-}DzI0Xli39DX35>Z7Gk%IT+c~A!LRY7=V5MC98mm+ycL3n8^JWDz!o1G$G zNRcn3i?W63R@vfDm?hb(W-iN?cEOZ}$gy(saEHPAiskK#{<{wKPJ2LgcnY7&Uve-G z;$(}SEr^Wi4tx*STM;`JA6LG^GDOM~4@QIe8_eO=x4GBXqTTY(8M9rL z^al0|n0~4Ld7*x{1{1@<&K2Gh$1x26qs7!gW((*ZY-^$b;Ap60GxMudK(k>5I>(Tpq09*IKIYJ=o` z^7rs=6zJ7gJFrJ-R(H*|sSoS#8hdj+?!C-lkc-?4SqQJw4${zA$1mysa>-XN|2v%APWogR!HnKJwE)VvKevc9~Ahx^LlprJTA(yl{rD{*D+Ec3Lpv%5t4NSps&v<=ohv)(#T4}Wk(hz!849S+rcTvJM zPu|MOh@~aZ;KV%1S89p(L3g*V9lI&lKlH$&=-1_+oQpa>RWmS`Upb z-)i3lWz~&+>TPrgo14WTh$H<__ zz89WR6L>ChdvkxU_;q$aH-=1ldm?`#^tuV>;B0PdvL8jLud!~}_d}E~K(gJI%#06~ zOsufz%iP@{*o(FOF#-_A_V^8w7WIbGg>qUbb4BGFZadbjAl3FY=4O-JgqhL9@=wEF z&%@z;-nkpG&88bjqPDG+6wtjs@(rQJF&klcIDRgKO*Q7XAr|+egU79=^$#RW=PJE*qRTHz7?Pu2Yq}zEONCen#+ z4*B{arc2yA{IG=8zys0Ov?zTlAAIZ_8z3I$l}nI2Y1Z%kjipJjoY$lCSk6 zM+A7q}-9+uCcJ_VZjoG+2XkmfmN^#IyHz zo6DN!ay@=U(|oS!KcUmf+r_7NLL)bfOwYzI)cmuxbZygq774k*qNn*7mA+zwe>BHE zk6*OfdPdWnIaxsy!NW5^yRH_kJhkagQ7chhk%hbIo@zJ0&;$MEIsH7}j)BVK?S%SM z5GU@3_y68b!%$??upiXycWUWV?e;IIzo4Tv)nq7WHn4+=XC6Mfc_X|^q5JSM+TM>~ z2QiM=Iz_ceLUv-y%`6S)gO&NDR&J{iCpj%)Fv_b9KjuCQm4f)5<-@0Gv)YY0F}p1s zAm&5xVE8Db_#kosNCF}6U=5U=2ftzRF1GIFWwI9N?#?v_)ivc?gxw=*s?x!sZYhTV zj49zs`jKusbc>1)!&O2s;=PMf=4e*I+IY^I*sP6r3+29{+%A+!D0@TM8A>0@cZG7# zH3Zup?mW^SVUKi2^uQOGMRwtw-V6^{HHkE9(%29-AGa3mcOV^vGk6|y6V(oDDaZIt zzf1tbp>m`#*nxhc!OoI%jw-AHs>`)O~^_wiqekR~c)XkOfiIi{uiu_3Arw*rE=KC3;p z)`fd<0IEA3_Jq*+|RWr%U$y zlK*tcoSM2#5Zv+ECHE zL`574D1(5C=qQfLnD6(m^K_@#=5wxhuIqi%Id#rcXRA|n*IoC1|Jj<~k!^g84*Q5_ z(boVs^mCVA*cLz>tBG1`K-748BCM{2{brcfh{rERc!7*NKB%;reNWv--B0XGxIo%J zv+lM1fH)%P(80i`bu*!ki*gAtpA?GTp& z*>A&qi0UC(4iW%(Q#K6giR7i!5-s%eMP?%T3^5Thn`&sxO# zdm4Ej&B~%jMnX{}Z_NFWfy6&&_T|j}KI?gj`8P3a;?@KYEKO}WL*^ovK6HJgI>+Fl z*cylrp{ucz5e#75&!7d&Z91Hsb+|V*NFIGB(o@8QSlV~9B2hR*;-Siy97S11Vg=d0+ zM+CN=<`Bij&7nKcDnPaMB+p7GIuROWt$o)TYu4H$eMwqWj`Oldp1XsK_%7hf72h3t z1Q@;Byo+cyzAot?#TJqnFdj(+ZYX1zUjxL-lL?}#z6>cZE7%E$t5%i6)n!U|8M(Ar ztWxu(vbjn|-EoGfvL_`V7@f`;#%#p8^A*T8q`VAoIJ)zmvc0!V>F$-Sd~fI4sZ0|5 zP{NBbn2A;v+$XjF8{%TFAPZ~9%2=7EuyW`ikHhF)e$i7B>TeB%KpziQR@pfC9ycz|sWq`TN(AiQ&vjy4)Va z5C9{+?6>2)A39}7_^hIfO^ zXp@UQL-3k9S!4@3$6W_;1Q!0j9Ht*J$q$Wr)J*xIem%x2sAX+-=zeCRk93$4VVxNs zjLG3buQB;pD>no~V}enS@3p@sS7GWJf&x%aclcv@G+}r}YvFH4;L5x0uBbBm?FdJG zGBzTpBX2|mfjETxWHt!4=lYLhV=Eo|6yjm-3R=g})AKshJePyBHXKj4FP$&*`fwdS z-GfD#5O=X5iF4|8&Bm9}#in?Cdfcw?Di?FrwN<)U-0DLl@2nSh=c&)R^hC&xQ-U+ti-4EFkJo*`Nglx z;RSNqryK1G@hEgaHvlphL68pv%;ym-y%>K`-b#!kV2u?CSVtSoA&VPYHhni|@yy)rCE|C-8ib+NAtAd=@qw4{m(^n zM-hvIaouFz#v*{=+}u-4?=PDBis?f|M|j;wi{J}bI_xPy6l7{vzD;m0|>BQVyh!MwGc0?uoAx8We03Z~)@OU8cf%26w zJd;>FWc1jQ@sXP1@TI(NwAT)L)mai1&|FFbAQv{ ze*-v__Wj1$mk2Y$-7Q9&9nK88p7EURM{i$WI$v(%%dT-{UZ0P7USC`XQ83iaQ;5IE zroE;qL^{7US;gzb?+g!f?=*+W5?v^aRf}stG+YyP&UvR-f3HtX(Rt2>W8I^^&Yw?k zV`Mr_8~)YWzghQJ`kC+lPiQlHhO=incZM@7U5FjB4NpcD#@j9{WM9B28X&o#U@t6~ ziwep41%h4vJP!XYc9@Nk`B=e<>9#QSjoPEN% zpUK4x8Tijt&%HL>6E^W|XU}o&Y-cWV;ki0E5Pe$Iu_Xx6t3@ohvhJ1ZO{)ho8R9rZ z&o&SY<&fjqTR*}w?GUN0(H?&cJ7a9{oN{G+F|=2P?u)FDeA?>Jei`S8(0w^HSA}+! z$gV3x`<2jM7rL*6=2~4(-PNJ}T4-+w-Pb~My?CRqQ5$iVzU}modQeKj(f%FXLs^Zr zveP+`+%iu0=FLn)g?JUV#^9ZZ|C)s}z4xL#7&RNslAgt-U`Vq5W*UzcfSMVPQgq0&-E z!_N_rb+oP=!Pa#wRYvRD1u~R$?wKG9?zLuk4VZptli<-3+NHQhZZS9`dMnIC!xf-;TaJz9Tc=$Z+hmSC@Oy#DIY%8uOVipl`K2K-LU>U;cw*2 z#iQIl(t)x+j(s`c%e>By}x`!%zhu5-530$ zTuQaVs2%Q5v!6f_$m-REjxolOg_`?O!8}uNQyd=2=9ZGVxdiX!((_B%YMcf||D{9z zoMPilcb_E$KPx3m;U5aop9)D7l16~8kE7p|)8`aUuRELqA^f94`3X3<=uPg84713_hOxbaTR8+=+nMrr{(7hGmiJbk!t|OOP(T@N0b-3dC3q zOhsnj@4M|1f=Wt1K4q13Y}mLn?u&xeD1GaEYYmSph8hkBaxBA|UIaz2Xo&*Z!3XCD zza-AyBzcz$&5h(FTp##Uq{pF6@0$g+y-Dy2*MY9{>jt<^OE}j_Iq$Rgk=-u%9MKH> z^e|l2^CwK;2?N+34DN=lAuQr8&fV(FEnpvey^Fp|$5gI!4tw|O!09Xn^Wpm0H3jYr zxvV?rJH`Da7L$OMNd&G(Vihjw1)nMF9}tgVS4k93Ab&ck0}I!K1Y!{E&7r`3`%V(U zejWwU)x{HNq6B9-$df!Bl4oR{MN4Ma>6mVkAl3{vN1sa8y@L0i#%~(IFUgjjIXN_I zc)Cx%$NLy}p%T$T@R~||ILw>87;jU(PC(P_yfTjxwvnIa)(6beN31mFBx6>Z_#|Uj zn(9eBB_tzrSoloae7bEu*N#8iwx4TPKihUM+3@!mF8NWLn|~MPbWmtk+E;~=l!6Qf zRSP=LV36tz1>QJHu_$OFJ!0uO6aAV4rH56Q@C@1uvgYQ0YZ}t9SNf z|0|qn>>2U_OYr+58pEdz0gs{!P5pWkeN~3e@HU-Hg7A|>h{dW6MvBPyK$5ydODl89 zZ&)??edh*^ytv-grq6@6U1g%x(o+|ZmD@?Bex+`_*M|34d%rE+hlgTNm&>>w46=Y( zkdNV0W~X)HmPghByvq@NpagTa3>8+wrw^=G?iU>>4HPf8|B^#?8TNbx++KoA@)5K_ zME1jhk59^TKJOpB!fw4@n!rWAw680-^EK!UN%XWKk zUy1nKnCy6#!+i@ib=bj5?!;p7d5}Xy3PVIE!d-1xRPP6ruxuuP>t(3c`9`un^Js$1 z1@q&Ut%JXLj7LTN-bN$~mJp|qMS%pz*f^v}WTYTa&~~R-9cu7Q>QGTjI3a7i8SsoS zWs)srb|#M-EDmJB`^5<}Dq6aAlcqTn3~mP`Gzb!Evy~PT{ufhqw_WgY1m3aT9a7UP zui!g+SdEI5m`yS2B2kgw(sPgvzyPBip^atC+eDaVmG3C0m(mDcxiwgX8inx}a74TT z`HIpP<^hN6$y@>sgSC1*2HVY-6<;RnEYn;rw=-EYTA2JutTH*cLX43#&ReG25DTEe zgHH#*3#U;KwpNKTP}-zu7DPZXXaoo$Hq@8@)JB__5*w~H z1R9N^Uz%VguMCHBTG`$RQra+fLfoA{4AjStY4q6fo7G=Nr|QqHcj5icOtBv?z%``y zS<-`qjTfAG-lc0p^KEhFskxga7D)+5_lD-K{OJDB+>;+Y5Ssh)qlZHCo&4zG&^(wQ z;avFL{OAYX%}`HyG&JAO>mCcu5A&m+ga-G+pLn-KiMe$hu8Kd+&;FZtT2$SWq4~Lt z60pWPnE8Kp7(E->gTq7I0b&6C-;dt3m&WzXhF+5D{q=wUlUf_yYt1h1iwFbW@UQ)b zwMvEgzj6Klq@{0s?ft@lc~eG_!_RDZuXP8+j`du)5syd=b8t0dDOOz??q}~KS~Kz4 zS?&14R=)EHLaHcKV?G$;z7Q7962gm)*FQzT#3+>|)IyJe<50)S?@Rc?5<>wC1g0f1 zQNUix<02B!+R0{!5@{`ZP9`E6imBKo$>>QIH*keK7cpgt^P{clnqCE!`k2x}_59g6 z;k?nv)SXzlZ;=)p#FLbZNQI=ifkX&di>vPlY+bD2&|Am zd9S{<{6^Tm1EM%g9|<#DULOcE%|C8FEh%-mv&}shjF8nvg zKuQcc(SDBv-hPHN&;9H+KMu{~epcsh$jobIGjsYEn$uC;?bxBFu^M1#2rCzU$vEn= z@K;D^@gjV~zd|=|PrVuI7nUWENc!uH?5t#J=N{~{3Erxl>hmy@m~S>k=&_kugar^` zHnI2G67R{UFa;j>3J@eXCZ8KB@v`*46dOpP@!zm9V}0?)XZ4gfKJ#YvqgTJ#`8TWo zKfO<1tnVCsj<5uagfBPLjXxU2BIE1c`1qf{@{RA$YxNYriTU99z1XKRlxE3_<2kg<{Vycm)F&J(bPgMPj>{_G9!+eE*$PxCP_Bi`dT^TO$ zHF~tg=Cy>GkfswC>*;qG@kIKF&uGX#Z3{OLH9kItcH)@2jjgMLq_`wtD=z2`vwI#B z^=cl#Vo{dMs0nxy!4}mnxU`-S6cP8veFz%6yMnk+T33=`yO~@rd?&)*hO?5Tv_vTF z6o_%h^QD&2@aS@2OsO=gNK&w|Mh_9}qez>y(KkFBOsC-~ln>qvzTY!jiceq3%|MA! zq~s;}tw#%HSJ=vWF|Wn-ct&6zVxrl>v9vjL?=v5S4MEh17)no{fGrqIbro3xHS*}3 z?eA~+`?Jxf3^Jd@s@nFz<0x7zR$foK9X*% z!eX$4rGq76s<0tINmduum{kR{y3oC!sAP%}fj&(*u z^GVvmtmV(_=BI$I>>fX1qQ5dFW8RpB|LOrEpW%CRizb?YZSNdFhU~V0exDf>a}IM; zvCS%X6mVXsVC4T0Af3zHypR`nH^D9wl;>!z<12j)L8amF2*LoNQ_08xO!GX>6w9rS zlK7)s?`d@t#S%C01o*Yd3TEp1%&aI>eOX_h4Fu`%hF-73^1WHVyWSo=D^L@eR*7JA zwmHQSsJ1*D90e;Z2}6?1raOm|drf$kF?ZA6vF=1Bwa~QRpgZ5z>CF(A7X&+?U}O=* zOLB*s_gS{EGhQ6W(Sd00k&_fqMWdve4_SG-iGIju4`|xtJ>KHAS_`JJ9q6agH2NK# z5Rw$TrCk7X+JFWv7rCc|i#!qBhY>-7mFQ4_4eU$by zc+9CdT`>hDfR#LM<5gJMFOb5#jNJ;vCaI77^xgz42&%7yk~hOXXVl9gN#yo4Ze80+ z<3Whoj_8TtptGQBzhHv5Cm0(r(8PI{xQEouxT@m@H$5)ku04oHfpi?O?h0fCY=Qm+ zFnSM?v7U&4&fvqy4uqqcAdKZ{T<&t4dnQ%#Wvw5{#}J~vL>eD2RYIRmoGIcJD0FLT zXpX*`{=6G|&O0FwbwSvL|d6jaVE zM~8hN6ii@!kWe}PJQGB9eqJlOJd6V|YgrVj(Jz`-_P1O=s(wROlZjPTcjKrDwx%{5QoBfjkdns{I>#x-9s4ke%91L)BzGbPE{$>J@R6-6ujObE~e4B>X7$wDk90c29*UISaYWE!B z6&O9cKO%V7o~G#^v?-*J$6NHG2G~p%`yLj`*RuvA2X@u>VxoZ3xY@)cp-He-scP1m z{XKp}0UQSe77qK9&e$CgT!LQ#Yw^)^(gBPtP3BzNj zdh|!S>O}le3bwV?g|eBP(&MWB#eu&qC%WgE3D_OBe4>E%B;7T2_Cl|-X*Rh*@@fz zzP!@n3Mg?dm<@TJk8VP=Wu+7eOJO2o>Dc_nS6|DL(#4wG-i=!92~! zhh%9sOQ*I&4X`vD(cxb4WDW=XB1%~EfmI1vzREmn2rUX`j30pbSXx@@sl9w5%`uHg zH8KQhaYx<9Unyd0^MoNqL?`~c$%ccwStrdgfXdR+viu3-7L1$|&Ywc}OqSV`=ISiZ zg!w0;fh9fO+ocYYVJiMP6;c-}d~)pfpc-Bsdy%B-#^gq_j@F-N2^eHesSVrM0D>1b z)Q%qiubhQFlCVG`10)orTcmYFQJ}q&bP-~Qqp{Yi`vl+rw_D7TO?wHW&_*so4`XSi z_R@0SUj3q_qksP|v{*ehzQvs$-q6?oZ?*X6k$Wn#K?9d~)=6@QjcsLsKf}>~VtgLP ziPoHBtDmywEL%Opnlo+nGy+Q5@+sDxYO`}8td3_v^0FUtX6DGj+h+K+PwVSszrMyY zuqu^NjcP*IgNx9dft{H#;gRXl9`}hqwc%mry)2@xxK=4>x)Q+z8^{N!14W&0psx(R;};?M0SU?OT3Y7e z5EX1LH0#B09mZldYlQHkl8&sk%hy=5*0$C_q?9Jo{9}XP!U#nLw$MWRJj}|3Usq&i z+F9r=#ls?DBV-5s2;xA{5)uWi84(~nU}lmlWQYt(q-i2@JV{^WTI~tf)1lr(rw)fZ zLisdgTY|?sqBNQ$VbZaD_gb@;?@c5k<)Cy_a2*l`UN6&VIQ5`-2x`kjjVjz(ND$cF29yAi~u(MR7Ky7sq!Vy%2nH%newu*;| zC*;gTd=P*#xn_0?L?@bl>S>n9y-H{<1ysz^${*=Jw87L7Vi2)3QLm=4G7Xx^$i7%4 zMXhXeD->0D50)_4pg-&-pmzg~;VqGyR*sT zGOyND)Bd%;+tlH-=g;+Y_S4ymn(5|#EJjG{CF?03+c~ejo;Ut?Kfk-!I+mL?`0%OF z=F=ficgAX`3H3THYnIA(kzGKn9t@6~v^)=H5gO6xm#WpcBS&Y;TSPV${6x%y7_zYT z40D@Xa|ZH>GtB1<2UuDBHVjTzaCw+bMyl+`XC58=a%euyy?E%+sSICeXXdqM=!t2G zd4P<-a-9}j^^s>^ZT36dDMYz4d{jHT?N!cP?xtPAUD0bsg86v4>w42Oni%*uBR-=? z$oL*!;KJ8Dfw5d`!aw8ghFt*3v_%k^a9`~3-f-AmIP9tX(Se@IgU1i+#LjOGlc#X} zUf)7jviV`*Q^())M#HFK9R(F#t9!fIbghmx#OgAHrT$s+HxpT`U`-$wfA8r1qt73G z=Xh{YA_gQJa_t$xA^ap`rxS*|VVBt*@%rOai;$E*hfZ6@>!z{IvV-OwAoXdAYS4j= zDC#PW`&C>NDsW^}mChF?!W^MY6{B#}#-cE{M3R8!$%r{kAmL{SR1-@wQ$+TMd1bO< z@Uv$9fLSq^tREnyQp+QYQgKMGCC!qVic!0%Oe%K;e~L(8K%Ouf@Ri6@i)X_7XpU+V zhEPv}VDM@;pF*W+%!{V<57rxyw?9kVnM(5K$o(xce~qxsr2F`v>!8qX)0yBKVtp_b z%TawC5vnU?AT^5(d~Ns086Z^k=|og6ub-Eb(U^ipD|ED8%9iG6u1NIziGX6ulM! zjY96|dF_SVgviK)B${8r!9QXT*8BN5cP1V=pF3bV7`n34G5TZQCc^LBZKx3Qh72*i$^0#)u@#3Bxm3Sv6oyppC(>1ACn}%!B zD0r+;(Q2rdenC$z*mFjIQB^<{hn5n8rGP@wvxzx_rB2u2QLy^4pwdu!d24!`t9p48 zT1Ytz^M#+#7q7fQW)o;O@;n&BPo)4tS0BHQJP3} zp<1&goBU7dxr0CtTrs7CIih5t2CRPMnz~tA57*R@I3qal80tO2+M@JcJmbEsgLQ6` zpN{U8k2?L4_eOLJ4*${G&|DLmn?ie|H_b-yrhI5&x2OB^QI#o}8YJYe6>T<~5{^I* zP7U`Ku_dX6@XKZ%zi=0EA4QIvY>J1hJvTtsKjkv0)xdjfI)Si6w#0t2l8|efBb?> z)C1{)I!vxdfhaXVoY0tnrzMiM4VUy`J-Tdh6kJiNh;Ht|q=Y`D6@0%_Ohx@Y&((R0 zFb3uV_xZ^DDm0S~D}gniB~%GiYsp;1;!DCPyj6TFMd0N$*8zF!-owhtneBKwAjpH! zvPnQ7O$H0M!NGwR&7nnmFLuGhw8&4eiFQ>q`u^~-$Xptk%VaYU7y93zt!pCeT6Sdm z-5;56M&?en@U$qzF+x4+b}*-owUQ3gj8H2F*}ylJCFH}jk~2NNm1wPih`D~WBjS5% ztYu#SEt6qJpRA}IyS{2x)Eeunkm`gP_Mo{jPzWeti=n+qMv!w$)?FvoC*n?l?72nr zQ@>EY0^gL)XTFkMjZuQYi-4Q>#*t*<)(C4=m~HK-mp2Zn3r;%tyJFbr)I}$7NO!g z!5o$B67HPr7Vett5$>Ms@AgZ`yqX;34onVl|6$4xGZn;OWds%AuU?*rovw~txkRaR ze2b)}9O+N{Dpq|Lht=l=xE@$EqreU<3c8qIP=!J?ud355%j>wJbKR@zbQhY_ML;Ez zeU8y82o7_09cswmox`U8rE?O#=1*CqJ787(q>t@8$ljFMZ5e`X0U}z;$T9&{Z z(SkO@VMXIAN%_|j`)&Ax)CiWhg7c~|pC1J)(eL>W6}_d9r%Zer3dtb+f`bj|ePnvR z!t8Tm9NdYiPp7l;eYp(X*y*aJ{^L4#N_Mp~cMhU(La?cUZu)VqtwIbE_FZD_#W+zs z3{W!WbOsb13ZHX`BLa|)_WYnaM%;40p;n=vTEV{oQT~yLCI-D|qUg&B* zGx9N{B&qpXCb*v#aE0Qyi1t$WD*!}zm9EN3{06lPRE_e4$%mY7Gd_&u#y;clPr%h6O zYI;zTk~S<Ell>12_H1}A!8pj&Bu(n%eZ~wcftPE!^a53N0M~~gSG&@hlUSq>DnuNr=J@a&WxU|tf3=rb++Xvvr`q@mV=t%Q z2rq=$RVupeD;JRaX zvzwcDg$k*RVi~MR^>?At2YIUCO>YD5@)|{%5#S=@LOQ|vfE_JZFTwLdUoXL=YOpJy zMv`_DdlQud2%6T93GUXw8=i*d(3Y3NGBambb9j1$@Ng@56mr}n3l>T&JtQQQ^;bP5 zC$qV;ZEoyINtSyno#4LRjwg3k%I)*@nb0hDhE__RsJJ=)i8fkoyH;5Y7^e7g(3i*f zGLN2E@yB`4()_rwN^trwV^HYEpb*LI-J?+W#tt3_v(-47^sO)HI1XAKrt3vbgFqy(4p>9S>z!(CvBUMKVIcjGkV0=CgJ@ zuwChe4oH~+klu1AUrI9LRLiBw!p1mODh8Ga*Lnv;P5>l&0SURqejrB`J>h1A`0vA9 z44)a;ojnL+=3!$?_X3#rUwAJ~@oBq+>=1k85P$>(jz>fts4<6$1bC1PdWG2nj+wYs z^a-Z7n+6|^KVmVTD}AVB`v|7n;k=+(X{JbTupcy}i~z9K)^7zKD1mR$m`f!nz$}?f zd9?T7j)F$UbER`mf)yUKn&qW`;c=~y9IN3=O0zqrrS_1<_v%G@5bj=LP5Y&8md@In zih3cO8iEEBJ^O_925Zb&zQkpaW8@l?7&a7x#!P~jI$6%l2=0B&rFv8wkG^F^4BdNzpTs13 zcjjSh{+gOs()dA}Dm1a|9`=LI!{&~T(?IYT7l}&b{X7b48v9$As>zAl1+FUW_v^EAfkwRy={^k4u1PjK1hADiSHW-^jfg2AtwUtMrCugC^#9p znHW#{hDaUe`NaG-F~5(aJn7aeN^5|a7;6Jo3R!S&OQOTIs9T$QLL4#X8(A<3vKX| zGFDKr`AVsLRmnUEJr_L!i3%3=WCh55+@GT7sy~FK^v^r-`#PCh>0gmOb_|f28PTXa zWWN{&;hr9l9QcAR%@N+{F`Ud;{hJ*NK?=+3UU5ny)^sd^fPl+$U6NN&OUV^*Fa8)~ z6ppc87^}~Z^~BgqnXKm8O7s?z=OqMsIt$B3fnFJsBW76&(rkb8mhuj6ze?~_Ir~Gg zlMCR*q8L=~>+c4V;QksT&=XdD1bBFlBi1On+ed4vR9YPmrKR9C89U2si!}v zyYJQGN9y5s>gglMy#eVo8u5)l!ayVD6gYnZNgs~^sHUMAKTzraSuc3DPU9nShGIu`pG0G5EB3TQ9fK>llRQI9QGxdRjR@B>Jg zgGK1Agx<>1HL}DGGYqRuV--l^kzO3>Vd`;qK;RHO-mArE9R3DFL>^E4^JS*zE`#RQ zkdFHc!@jVV`S^RIarj(3luZL+3D15^O$>sCXuWdge_#}U(*{eYSk7;0KW3;9bcCS;!d0 z*i}0~EnoVfQbBR8W%*3FaPDF|R1A>nNDR5-j&E{pH? z<)AO?zC6*F@AhTEm(#`_PxQyV>q5d{?01o|7n;UJ2Ak-fi%fW-!gf#=FEYgo&ABeP zKJqr9d?0h1Sxh2Un<;_;9m-q>1I5JlSP*W`GOz%YIsvMmH32cDHQ$Ksw^)4yCEotE zX}%JhJ0VHBe-Vc-5yCtUABl@=a4q0-N0pEd0WS7S&fJP^4B=tSwXS##58#_`cG(vl z#+|$ue=X5Ei*TfzZ?+6^pbWP_Qt%;rtocxUELRK8r8Xq3U$qu)6%+Y#NpYe4q#*kz z*jnZSnwRAdcfE}3rsYI;*u`~VugBVhm%pv z?FNFY+ijEW3RAKfX3w~h2f`=2Q1p8j61HYPo*>E#K<23-iaUq_^&5N;A^zEl@YNYH z8fSTFah#nmO$btati|8%4a9HCZN`uAKPIk^+rjeZ!o#9>?ik!p_^M)o;O8U?rH5qZ zfwxjf2!+DL?A`JEOfoPp+Qzlq-nhAy%ybCB`MBk6#fFyA6@x zxI#pOJ)d67UlL(}j&TOY=n{XF$Ucrgqo7Z1f{5*{^|7?*Vqb?p_bAoF02B$;q<8RK zg3e2cVBJ3yax8j|{)PK<1nT#yIyl%#(Fg`vNXFRO4DQaH*vrjut-`y##6S0!*&0~- zqK&S$=FnHwZi6cOZE0`m-DXd-mvj>BiT?xhCm4sD`> zO6GxpoVyo$eJ#^KVktTR$0{PJk;Ng7CSw24i!aqiprLkf)+mdofIZ-Of?)~;<6-qy z?^?C_xbiR7|JcHpTl(@Vxf|By`ED;?7y0rSUvB5it^Vn9+t*#WyRVt<%kjtgGdK42 zm*lUX_ zPiDU-R%%w!*yTN#iy7aM-aiT?q^8{)4o~OY>ZQTr3oHF4eGv}nd>w2 zRVBor+sPlQB%c>#hWeYO0sLZ;K9Nbv_tPZ38nss8cWL$+n|<1vtvXDARmOfXEnS_apEuFvNp?e$3}7ceDJ`JSMV#PnA3cAYiN2j=Hzm>6 zfGGQ=w8%ZzndrtOJDDrPPm+(*KZiB;6hcjp`-)d$^~@FaRiAcs;(nCb*Hqq|xOJI% zMk}$ttzvG8NMIkHlZy&HofciF>;d9Uf!3=b+rbn#b3r7h(es&mL2MZ-YxwsheI+q} zPt1!6u^l$HJNpO^vXAjm9-@oW^pezEoEjKZ8R?08v5l{$*-dG9W18NSnj2GdZ5n5d zRcUrX8lIn~7o@PY=A1Mmckv~o~C!C=JwS59*dd+(tJ!yJxYVJwRN`?em?nyJ> zORqckgVg;j&3u2p?i?9s*QJ^7_tA6qsWf5~M!)2y)ZLzDelSO`82ieTQ+IZn`C)(E zxieFDzJNqhk52#La=&33UzEC48{Tqt>TVR60bp-<^_3~fq0xHz-gW1t$W^i&R7D4C zZ2Yr3z5mC={XNM%sE}PQfX+}J?vlp9>%a?zqbikv}=^J-E{#{3Lt;o6I&OOsR z`)&R8l9rSRv1bwO9n3LLlqw*e8uW^NA6(q*Txv`nN+wj*X`J?mwy*_a4k zy~y=ukSAwGDg>moF~%09CH>>M3Tz=Y2$YQmJhO?&sBI2B^27Efh)dh@L` zT${S#Y=f82;OYhuf2YSIwwL2{akAa89Wv9y#?E?Cp+0=j(1A1Gnj8l2COKz;&=xXk z-6DDx^LA8xT62IG@iH7qvJ1O_oiIA(j>^3n)fN(_dttE)Ki@=_Cy)#auxUVnOLmKF z!n{fTxIY@oCP!1`iJA^a(_zYID(-qZBieZJMsBuo3#YlM)-9auF!*b1B31@5g^B0% z%y*lnB=c0nnC7B%p`yn0Zjo&sZk3Wg-fY>kO}2HoB!$&9i^J{GZOwLO+iYpNy;;f% z^Okt0;T_#hcE@a&bZ4`R**V-T-4*ZtT?>1pyAxM;_iV3J)ajm8ZynQ>RQE~uHv5pM zaap>rS!VVP_fPjT`Pn zUB(x2wAd}l_{M@+xV77A>XvS+jIZa{=5Eo{g>F$szi`!NZb3#*G5O~t zL%URRFF*^reId4P60U6IB{KzI=5pm^6@C`WmtmcSN2A=(q5aUK=V+JIgJOK7jzR0t z919A9je++PL|q5(&IG3fT~K1Uq4re6ww;m(3kmIJs7xhvnXhjTRb)ivSf0^g=z`lZ zzN!OVS-VW|VziFmFWMjEaq?~~m@5m2!z23zIWAmRz-6==aU8{~3*O4l+R1dCv<)B_ z%rudzHOSw?xfH$b2zNTg?hvhtITl*m)D7VHmiNUzz85whEepZ&$4qc0AsR?b@?Ve# z47`M5h}4N?*jQD_E-6G87hW|u7Zu#9SJ(VWK{j&{neg=Z)dBhI7+#ZNZ4@VZ5uqfo z`*mJBVm#=xt1GH5d9K^Pd#!SrmOrlN!UjhO_t_NqExR>F;f_*gKLLWgFDLE z`V4_aZ2wa!o&cuFzU!dz5%!gQ8T;}rtFTR7j`he2a!ADzP$?0!e>Jw*i4U3soeK6+ zMj-;Ipv4(95&5(pmWbcGcn$jE^6SC(J&*DL<{JZ zWh1rk^i5mf8-g~e5%VeFFXMs6wYDg$&^$M@Cl|j z_!D0NbIXnNFp*Tz!Q}&tcou|?Lk*n@YgCX0QUCmy^cuPXbh zxA?N?%b|5>Eta=hMX<+5nr%MiLr1vxUt?N@W+k|+sz?pI#=5_tu1{V`&0lct5tm|I z52g%KLJ+x5c?t;1Sqx|VR?BgUGVE)W^r{Nhy5z5dm+2zF@nVm>5jrt*rr7U^5iws( zqX$x0SvZBFfl1-PEE~vCq<-Z4<8X==Ei;TkK6(zID5^x2M)IrDcuqt&xyMWHS0(!l zY*GucKtl32h*j1+{Vvv#Bk)JRpu&Y%gC zbL)vM!zEAvE&kdp$z{6>d2G)#C~|ORb@o;heFk4|7^Ae1Bdi-Fu1dt;@$(SRXYk{uniXrL{SL$f&v`Mfz|eaRWi^h`I0yjNRVp> zcQFO?Rbk)YrW)>Ur&YU7uTmOc(s+{LU@i%$FyCDvj=PGjqL(qFoS}{M!`Ot}BUhx% zM#YZFmmQLVrG(fVM~Zka6qpEe{UFeyeS=$T*Na#^$;?WfZoj$a%gp0 zfdO>zcrClM>|1d`S*=)7U|}t>g^NK^RnWT^g9Oi@Tw_GPQ#B7)&3CKr;i~;^VjscG znZPTu?ohrnjLbBD)E{I6_KwNCNKA@bsNh4vC8D%}WE zc4aF_cgLi+m(~hb*9sMHGtHbuDt>V#y`*9;u9(X(lQyT7#smSf!Hr)f2KZZ%Su`X0 ze$jo2-p&~#cU~pBxRUutj4?(RD7nm+Ywm)YU$DLOZv&cFoDMp)h`$hO1f&n7W(I^Y za?zU2^6H)d+T?sS7oZiDNKm6u$}Q?fz&%a=oSZ3;G=^{k^!5r%kOG5U$L79V;>+Ik z_{+Ix{Xzcka64g|)kEzB!$hQZ7i#x4gKiV* zHpMT-a>b6iBU!>Lx93_=Lo~$MX}D2LXFto7+(h;RrrXVDztBt}5CvaTW-*R0uclX2 z&E-{dHG);5J`6jszK5hS0Xq02~VBJ1khvWTry>cgk_QFQ?tN+*+Z^7Ze=q|gGqOxRmSW{1YjYY;6BfoxmTD~K1n&& z7syR4abyZDuPKvZYKiso!^~S#nIbbjSwfW?EblqOzQ65hJ}o$`O`dHsK#xGY|A^ZAVF6`w-NfQu!<5Q= z-`p12TOw@fa)lvu4b}`lZZw~1m|ryv<&PT{v(V?7*{_@Cw@ve#rhO58whPN`dqy)o ztC`i4EFs|_cEdMEwHtBzO@7`;A8k}~GRSGn##XRg(s85~cNyGtqitxyhtDQb{1lmI zD4fl;CPd2(SIg?$WLiml%#SGZ>onK}`)FcnnG$}|t>81Q?8i0V#gEkTWo|RS%$0mu z^Cb}@S@@hUH(!TBdU-ePU15T&TV7-h+7?0M)-=UqT~W8Kmk5=f*8I_O45(}&lBU%mEa^;#?oK=EFSpx;wQ zeT945dP_K&6MR|ngr(t^EjfJf83l8DA-=g_ zf32^1l_>m?dUOlHJK{SK%e$xR=`ZW%>AHEW4it!_;bxYZA2wT$HO-HjhVs6qd7z2U z6;G96B+we0QyS@thFM=%rgK|tmikH;LLqz9acZM*S|jskg(M`I%h7d5TKIT9db*zZ zw~j7~>7%3mpE<8cb|>GLSZ<8Tht(lH$HRNuIfx?Z>& zh;I<1e+Iaw9}oR$kO!4Kwkv_{LA>-FY+V#LJ+N>CzYXz?sTFR*27x5rbagKV9ExX2 zO_dhY7p_h;gl0c!ZLqzeTOMg=pKkbpIJ2RFSZlUgBVIW4W#UU8fn#~nSz|5x#ld-9 z*c_bHwkNj%*FU!^FSh`JPj=Z$y0B=Ti4sb2Wcd(yjgZi*z210O^GxjT_+=^=|1vVN z(ilSGhs|tmr!mpB!l{sO`ZyLD;5u6DLFGLj7+z11?}sH31o?gK*8OctL-~!i#c*6p zdKlf_3h!u{+goN_>F#V{Ul5vmTH(DdN;9r>542j{5a??4Ysq*-%CJ$xqr`mKZe~lP z#QZ_8Cg$s&@D?`K=8%zM%FvnS7?hA*_fK8fH=2G7Zh&mE$y4bIC_PNB=6EWN{W0pN zu^OpAUVWBp;7&yH5VY=~BgUaPkw#{0#+XA!IIqPGUj|#^Sl36QZ~RteUt=qKIh$$e zciId)&qH2*Nd4f(As-HY&|OT;*Sqb_-S+F9n{&GjRbQqr zDEIh%-L3n(&3)bGfo}VqZuItdbQfhl1elsNZ9KkWcXxX#g1Im*?i(ivN042_(+CY* zPF9-uQ@MMvus+RYn;=pI&}0V$XI8QgqF#c)M^UfbEk=)_$pD*NVkNsz@vd-97VSpo z*r#Y7yI0}EFvZ1lkPiy;J9c7qV5+paA;SVef$j!f57xp1VAr&;z1;zq{|TV*iPUaM zW)zvgxzjN55`8a&L^b{h&_E862J(%$nl?QW7a|`K#***@0a&CKAk-t6hTKb~l-2cB{# znRR3ZTHZP^Jm}cq+TOCj=#1X*%w9F#1MyfP6}bLDhTSQ$UQfSrb3`a_WNG&8IYmhs9v~ZRo zE~WFOJ!e&kheR;K>SegE6*_ZW9}{+c)oK3>8A%S`PWu~Y5HOn=2{Nty=-f}X(_$~gpe`NmDNnF&#hWPDBAg2)$eIw+zo7uFm z2}HGQ5Hi-+h#-bcSY$1W&30zn4Ob*vo1M&#<9^q25v)0YOui-LkZ%hWlOn~YI%Yk7 zw|UowYr~JbkGY#oc$$D@C5i*Y{GT%8u7AJXC&c|2qGJDWUpsNmTM7y7AakxSKMFX7 z@n$5|w&q{W$IZt!JegLlH0C5Tu4b1qIfx2Fg^XuRl@%Bf_RZDCe0lsuG-i7f59rDx zV^V)M+HSUMH$gMU>#DZy5=P+un=bl)H&2(U4-|-H|`Jwtg@8{%wWdvnB%0guW(%K)JL6q?aORwr#|e|hXZb~#Bn^33`PS5 ze_weZ9qbzD3C_VR9&_M(y`3vTn4%1!yi1>T;XTL*?MuS>e>WuXt&lm@x<~L|RsC8Q z(OhD=KX1%cB3a|`7cPCuh2Q3^@ps*C#yu#n`U4cVD|a5`^dNrgLBBN$c>3B0{Y`Hi z{}^B4A5!W(#PX1bT=YX@{$j{bJgy=4yz1r$jr*&j{88Qo~&)pB1g zUt&|7SRZrk2hC^_MFyRu!+5nQ-p9?dIP14v#Cc#=v`Dh7)R-8f0_SxAh{c^@H<**a z1tIe5{7wjNCtgJs`{)9pMT$f!M!?cJ=1jSaCz`io#wAefB>qZVTO`zU%U`J@YJEI- z&PPqCmxjNFOb-5vUZA;As#fPLkoDv)t`9F??F@PnQJ<>xfyiOc^T-P!a zfvGOG`5Xgd!RK9avxJ=y9?T9F$&V~3fGMAP*eIRi%vo;Q_lzT?;O$zo=mY355r~AI zf?|`Q2I32ifk?PXgnKSxE{eCLWE`O*}BTl2M{b#Qk%GItK+K=bH%W zX08T(vzJ@AiTfB&g8g88jdw{kn&@VDK_uyu{P#}t-#Y}FL8#b`7I8KT$=Q58SIIdh zv!}zzZpCVHAdP&=G}f6hVN))C$^}o;N9#c%g~Q|bEel@6y8wplgYgF_N`EE>RKDxOw4lnnhs0MB3g0XXZVf$A*w?kb*R1Fp@a0d8-_?BiC+By` zxbnT@PS5qnEnmK6Tz%w^^Sg%pabCv$cxC>E{73Kl-Y7VMR3w!iw)Sz+@#Z*l{MY5i zkg-)}WHSe6z)G`~jth@ZKM)?>W*>hbI{G@yhL7*_Ka!(OBdJ#_*u*!=^$F~k_7)Eg zkMH7W8jqRDrg!WB6s7Ek%aux-&3n7zE3cVVs=RGReP$9Ige1&{gK2+;wW@Sx#H%8k zS6U~~G~}=u67}V7L$W79^=&otjT#2w=u0tyWbG4u#-Q^e{qid*HIRDLzz z>lSfm67v(up^dq#B1HQOr`%%o>+-s5D)9p@`i?UnF(1JG05%>AUIIs>!-Zy^6K;&a zix$3&5P&EA2Sz1mseYUPdLDaTbEZp9bJF63Tfu77+&7rI-M1NgIUm9dxsnc%xE&_{ zcAdGIf@CB>9)W6L@}pv!@JH@KP#votON7!DK9s-Vt}6n)kco(%+br z2$?6CBYYoXFj7m+J{C)^@Y6QE0QcVTvo^hi@X(OK+|33nPb5+=@X7mKNL97T$IM64 z`)qnGj8#%M_4s2%!Kkbx#B%A#%2CWw3e?`XIA^1o{ zoQSW8b@c^AYVH{ccWa1u??R8>VK{X0kqoD2U10>o(!x6c3dAqhF&p3@lRQcUIh@_t zlViEjo=-xE7n168f(`+KzXMKyI)JBhFQCG2X|*A?ytf8J?e~(y0$p7B>Kd>v1fqNh z(8~Jv3wkqbSd4uOwTwQ$Yj&o;4z`o)ruv%M9LIJGj$unVF4)~U#wNz!g(ghiHnC*a z1-lWS27^@%mEWVL_VPF1?!vEDef!2tWaf4dN~(#5Y)E=vW&_Hlh|uQH(WyPSSg%yc zqB3(prxZzvBcpcPvQV?LrN)Y};I#a&VRI!Q_U`oi9LVnW`{q`x}YH(-a# zlP65+8(umUPw#16Gy7&to1M;@xpBHtd6RU`4rE13=ck(vFQ{zRUQ}6FL&|CA_0Nsw zu>f!WHcX7NGEVZz>g-Ax(67##zFgqTa9r8&$9=xs$CsP>a<(sre7TV?-?=I)1ScGG zobyPp3f9_jMm%%warSs2LSf5I!Dz=HS3ABoqd+^SLU~u%GLomh!Wsa$@ejQi{<|J^ z0Xdp5<=xDFDBK1~c9B*AQu`Bmp&0YyVtj1$A>CH`sjEEhY6wOuT}*&|+!5UOUUOs# zSvNy{YZQMwsvN+|cY>%IkeDc$e>ClXDOD7(Nbj#?0PH7A=FAdm@GgKv3lE~4HDP*p zSOe@bG08!!q zoTM_VL_6XzF&93#E$0VdcvGZN#}{f@jr^(`*QO$Ys<36% z$)a%xV2}hS3tvf42;Ku|?Ve$0Hdx~8 zsTN0nv*y>JWqD^Bo~)TF9UbcryX_;fTR?MYRT>32P+~|7DAON;X z@-z^KZw`1h^s%tP@&r%jkpK^$+iRZCyMJZvb?p5epQ3%F|42&sqT1+emSNX)yN9^| z-)mEW@N>070PHm5dOf=z&;?&>(s4pdQ_b$}mx4vU%xA=GCKd40YvpPM zm%Q{T+S{ru^djC9_darCeVsc7^4T@gE)`f zgfx)uyRNTnB~9DI+4d%VrQSpPVx+mrgdyycvG(gp`bQ>ZWK)-g;6&!4j|&E%Z55m) zKN{9EM>cLIoW=r`7s)bhT^2yZKB0#IeW+j1b9#nL~fC2#;Rx%!=@1zMQolJD3&b1{tQDP&A?;3CqHaWwk1R@Azskxg!#Y(5{qb4Ox6fRctL{1!~&VOJei3e-@(ziDBP~ z2)Y;GRFb7S0+}LuW=#LgXp#>If=!gNO0-QJ4!X@`Dl0oZLeui`i%=}dPDsGJHr;-WBGM9K zm2!qJckyNB%d@QC5wMoic~ySg^Sb)+SGLEUt@z`KH8@^^}P0Etf12r^GY=SumO%)Vem3#_?}3;({49e5Ydtu%kWT|K(*zw#&B;~! z$*LF>B%z~bCi`PImWGWAD?MCP)vd|WyR&$So8kIO_Y~~C1$$4y+zU<&|07FZ&f@N5 zP=XJ-*K3`yR7PlGe#7EmUm!FS&#l<)?4CS}B?P#hm<*y|)&yWfVo69?t~u77fu5Xv zICJde_|dZcei;XZV-(C!QS*ugyA#sOIjD0;6vg~*Nx|ySyZ17{WV|QTG~f^xCGWN6 z^NW=Wit&U4q_DH$*UBYlu#}#Bl+no8iqFVc}3hCINwdU_>wjzrTq($Ti zpnnu7k6r@iNYCdVgObmbCi7L0EunqDyMCDOIPohmlle9M{>oWI*ojIFYy@T~ya-<7 z^xIRJYbHm49Y?4CJODAaj{~Ii^g?#ItsC(rI$S_v5BIoXGtMF50K9Wf=0L%*jB;9b z284Qq8|f4@dNuj3Kn_*x9J6PLSfVwdSgX|9<+N6-bx-Ll_ok@pM?&0)M~an3UwQKX zN8NkJ+fkK^-}_l*W_Fp~PCq9ny-@-Jfdmptf}ny*5fm;c5dph^qGGuc0)&!?bfhNq z-n#+`y<2~qFwe(oP{PGP?DaMrJpTtDU_7(MgxDh~QvQ|}0{q$aFTmekB} z#^)2)KAab2dYAQZ(|*og<7bPcYp(!{``!RwG>KIwU$S3R&PE4^ays7kCC4)ft5JCW zHJ&L$MUM?R(!!5Y>@XNLcmPaxoq^^zmE@^P0!61-FKdC0&>)QbI&)iUZcpX5)N<}* zc|t7Mmn+Gu2>gxu;ifM4RcPqv(`0TM7Bc=N>_e60(F&H^Qz4IBbqC1UBU5)&Do3XB zOlqFivpk)e`Kj3j#HxhQ9Iaz0hwKlm0_|ox)qyM0cW}dHX9|NB}l+U|TYWDNrfF0|O(i$t$3t zc5bQuE-{t?v=i+`)#TD@^!AV+e%p}OaEFB|Pf|q7%}#66x7TPFUuRo@2plV;!w_)2&o+wrU5)KtxINv(u>3q{zeERJRzLF9L~pQk%h;H8yC)<;k-lJ zIa!23)%Q00bl6=LaO}Me2to7frR_S`-8P(GFoorOWKiBS>ea)JkZ%Y9Z0KV)e!Q8W ziE$M+=dvc>rPx2?rLCe+$ut|f<)weFRT-#$&?f`>z<@=!DEpOhpim}_4xSuWS3?&a zONr8+K63lX=1|_eck=7=D#{{gGV@w{5w|()Ol$<0^*HcQ{q<|ns3N*} zqya_{l#4Tq9Ph6c)!JjpS7i*@Y4$*ByIar94tb6ZWz(S?tyCNMEC>;!PEJ*zgG1rQ zhn+|#8D2^m!%1AlW?c4^Niv2*k$TvVy0CYvofT z`^}~n(HIaqgD}x)mOEuwboPa?{4&TjvOeD?Cxsx zjvN)K07SDcWd3iaBSia^|FjJ}k0vBsXK`Cx&2^^C4BaVllA(Lv-fDgDz2i*b%w|-_ z1shn8hK1K|rKb`UV00#HkloSd{q$;2)5MbqWW_iUl?xE(iBl42X!Mb=Kg5vFDYZ^+ zw8$%1DgYW0snZy|PvB~D=&s3~g62vPSbr-3JxHa{FJnR zW32!dE2u^GnO)2$G@g?hqs-^~P~WWvHQf!uIaX!eGdq}<35cGu8e|-WMoE%69cGT` zU{Ne#u9$nQ+B!O_aMvIMp?}n%66y;JP2;wz1#20>yAxFM_`=c-g~_rnN$)bUg$9lI0?>jSdKQ`2JECO}pn&Z8G$%=x_y2NTR`(m?_b zR%I|9*;ROlS_6~kTG zGNEmnvTrNtI29L}v?jLSRR3TQj<{Ij+h8N6c$P7KxmjzOk>&U%(Q~FMKRI0I9-dve zG2#57aPA1_&xdm{oZHi6Dc5vfJ>}-hptnzw=?H0OeBAA7KW=tC9FpW~77tS8gUVmq zfEN{x1HNO>KQZ6^d|^+Jwz@iQN0Vcl%arAR*cioWBH~#*nSf| zBs34sH~t)hOw(P2u!$vXuVK0d%DrMgO#f~Tj?YMIKVr#Aye^#RRnn`(UM==2k>ju= zVlUcXfgk^~Fg?eRyad2dtXio!^V#GJ*to>zlgS(+;NqlBw(#44Xd8Q{iiiav?j0H- zhvW}YU#MQyHd=q6q_6^T?|ZfeUcvQ>@5osg86$)J5RXUFF@-_2W=w<6MEbJX#*p3?ht z$41(Uno5s7hD9?S$UK9UtY8SVl0~7Kt=@kXP6cvzyk4p9F1=IS01U7K|8}il;-SuX zeQgu`DuYorKF?y9vjy98BWFouTXn{s7}xf!OxV+tWHhtEVePI4ah6F?;wTO~qAhD% z#XRfhh_)zat0J@d`O7>Y}t_F{m>E>A@&w&!M7~YBGu-Kku5)9(X z?nIN|7O!lNHpxQulKi?PZ^FH{e~{!Q{aq%>GZHNY`~!lk@dCZL#3Yxqu9W^(41?K~ zTyLVeyz)PY%|~*K<81nzanFo$#&ZMOF%d2-CYJ%f9<*av0iRH3+%Wa%&62`>P;8h7 zgiuS0MY^^~*TNUCfyh;3JRBq){h-u1Dr=xa6VHwISm9-Ebg6_N6V5w_^OUwH7iO>H z@TfPOUE6tlThh5%(vfYCpg6UT`fBH~EBKI;WLJh_I-2Nvv5p<({l4;y}~m#Q8#kX#Ylab^ys6Qh7SMx?R;@aMp$(=dDKSR!@FahH`!J_ zF^8WwGShsSbgCBxz}8K+21S^`0DHW%CpdS!lQ*nA(Ix~!9-y360KrQvCAJxI>Vkw- z{ECrFjBGFW3TksR8olr$cC>2~iJJxfq)mPo6k8Em>}}SDQL#fnZ&9*YOD?wdWt+Te zaf^cYJo671-NW82H%kvds6;XIL@q!{Na}iT^gg`-zjy>#052re5a?wCehm1EN~!+X zG{48RO=ETIcIe0OtVj&!7n=2Fn-<@BUK(AcCHl7@l;r={e2~j9>sJ%jV!A>Vz7k*e-I(xejH)1ui+(MX6n3lX))M zvhkrtM_5_f{Q}VBH_lZ13ZdB?XPSrK7517tObMQ$%&xM2oWp z3SJ9XpUBI)**9CJP10?u$OrcF6}THC9H!g$1MoO@uY9F8&}^D|t+!d}8qvLOI-%Ze zDqW@SdPjF>+|}d8bar&O?ixbbi6$AGQEGI@>H;)yOQ0cl1+q?Oq3?x#fT|H#UF?vG z0fZ4^Ozb$8{TNjMs9C7!v@_~}{{|#hq7XY#FvRWf12L+*-nA{3?p?icj~#(%q3)?=L}g~QpkAbLPhwhX8Vfk%eO!0dWW=TEhz z2dzmcE&@_VKy1tJ1eUfg&1%~cZ7tP^WxX5EI|>g+(V|E$%%C-db7U@z>;?LS*~FLv zy+OlLws^$nk(^|sQ!V`EHOCQ|abf4oCeZg8eP&I@Mp0aJ)$0+wR?&}J z&pa=gA&b}x*j9C-fo|(`g71f`R)Dy&09FU$CA%AA+EI9mt zeu{19dbrNcZkf~W@1aYfa@mU|*`DhI(FRa^Ae&AEvUXyKnIJBm0I3X#17(VSD%VMM zk9>%lg}Y&L0l)Yrma$=1_x5K|@)+vIEU&sD#uPEG!kO=aeXIvyp~0-ATM>8xc{N0^ z>iRJ|t1wZ#eK$(xf}TZVv1X=o;Ss@{#6t)f5YqCadoac#N5>O$Xb{FuTM3Chw;GDHg5Hg#p)T!D@nx?jJ8 zYbSs5?lqXq%40n$LEyZ;8V?)9;ZB}xe?%**_R@MCcL8PJl)7$nxO3$U$v3bGgtWQA z)biirg{m=hy!r@4r?KJ_d!%v4V1vP)quq?}Qd19C--lhIcj>VdgcYOeWUB1^4t4y8 z%fMUIn;V;)M9NAV))*D{yjX)!@saaNKnPNGRbE#qHM{U*NCKI_dhkQQ>6mphfj z;9fuK9j~0*m0yS_%Pcr1GmI+L9)IUB*)oG6WlnSi6%Nc9x$XmP%w}E>ocKx1ybXvl zG0AO(_P{6M8F!R>t8qt|Z(@JVd>civkhIFXL17DK8T#L<@ZDb;&r)RNhI6q9A(hsH z9!3wm);drGra&bv_pqBJz^W=JX9~)h;Ha~v+=;MLfu}m@PWsT$Ky>{Uzk(J@{8;qF}+Mz$~Ckoe(Jm@hGLfKl*3YyiFKv4T_Mu zNLk5Qnun|AA>E2A(XXoVGFyqgzgoYi$_ylcCtfs@ON;DB&ijixL&sVYAS##e4Tou zIj#*fX>?xYb176pC5I&fu@@#yrVMul?;!TRm`#)chnWbX%eomGn7>7w>`a7zR!KVJ zDSWH79o62hFlD?gxwY(ZOdBv`S)Z?9%?@$0D`eMLb0t92)_|lMQ)bV=;mHiK-3o)5 z_T+H>LO2)O&L&(pR=~hLNx<&`I_W`JoP9=V{%jbNgUa0Npe>fdv1JfcCg{)@ypfvM z@dH3`9sCfN?SsU8!oOo#-a#H^Urp;Trwqqmh`>Xd7O{T5CZexndqXL?S+z_N|ASHK zYC!ds_^XUC0GC)j`KFL<4FjOWQvz;a%{;o1jbKaqv>5aN12d@;YN2j6@#`le+27Ss zFl3n$&teY;6@VSec1K%SA@m?Ur_N!uVSqSWqX+p!<1v27l#EVO(sbtimkVzYxl~0pDvqc!*OgxZN0HC~EbC4O5QSoLrG_8l zrzW^KP#{V_=$r+2*9}^wuB94Uxo|Rk7m9G3_1p?85}^5G*kDRtl!FX5eZ5?LqpUi@ zo_X6$?Fx;g|cht_4zDqu0HD%e3Hz`@uMi+v^ueHR2n8(RHo82>PU*Zl5~P3c0t7)s|tRn#d3BWodssD|5&^6 zl=EY{P1Ml6l;+D)*)XwrA72oewZnYT&#fU*3ur=v$D`z<4t|R2mFO(F8+p+j3WUc9 z5Z%;H4AWt97*|YTP&{j&0k!bt(rmJfYx{^FsQH&dy zc~2#|zXCeV2;OOtOr6N2a*tsTDM-QX@fCMm1%(n-0-}$U*`he&O~zd5%7iCRZMNSN!z zL3oQaAA-ZxMt4r_SgTY-bJ4iDGLk5tb7Gu4rZdYk>V-{^zT>pKREEFLT6rC(Izd>K znNBHHgUWGG%F-~oEJH9|Olb%aL}BK@70T=H3Fl}96x5tl?XzYMS)$AsoEgnX3e!O* zA~}xkHfN|p&G~MP#wZuD)_Tu$k>YcK34L|XVn?&=56&ueOc2PfqTuYKS%wW_k;gG zf1k{Z`Bq&4ZbqvQ?97#Y5YrES*m$=0w_(M|%ydRE^jS6^J^AfSG#SWSW!>ug`po~H z5f>g+e3&%~F1C(b*lsw7ZC6v?GP&^{>ZVhG!gbVNo8UF5!Mff2HW(spZCAmz8$t+^>-YL}x$(aze5#*7uFsfN z>Uu;=y|$(YECqO6vE`rFbXwuvU<=K-cakI>O+SaJO0}|w`fI2%c?bf!_Nra=8;tzWIe4yB&@vGoEw$A@iFrvJweVVA692ro zkNN$y98lIK$X0u)1W#Gz(~xUvsY52JcLXz3N_|8KLsN_|xPec}%EhikJ3Z11q9 zo@iDr^sKDVi(5@SkGMESn|hU)tHoWC6?(M>pmFGVpW9yQ2x=G=I^ddNr5?iQjKbAr zKK+j%I&CQQcrhoa7N@?J6O@v45Keb%p`+}GVNLzTutGl>KL2xhnV*qgh9dVj2e{9N z{ye~c-u*=UGwn^pvoM#=bUfMSDdMIrttn|0nH?p(e-n}gT9WAZL zs#gb79?A8XP}%~!YBx?m31EzUUtZcef{cMaoMxuG4fKsn_Zzq)!H<>eH^z<(rA<~g zekkp#VWq*4!fv~gmln%4p)`#3g4ZE6&`aHA#y?8PSkmG72;?zqPba*;%rts5?y^5% z%Xt|kdsip}^3rwc8^o+GzAco2TW#9&>tSt}2>AF|o_n9*{;=I1?%ekN!_~Kq$Ab4oCr`T+we(Gs ziyTJK{ukChW)C{r=EMv58!htYMO8zZX%)LQ4i`?eUh&NzV%vot$nLg3+8zCn-Q)h! z?#2Gf?&bdK?rDGh)9FXBylc)h6U=`rwW6$;Kcc9O6(U$G_Kz&zV8*PZm8w=S2B5pe z2=qFL9gVAWkiYb4L85`b@ku&$qRCe|WdG;Ypo4zqQ>i&qwEnciZED*3`N4+sn>*zK@w*!n=jZB?8zlYn~f7CI{_b1;+u_}}upu?oZ<%kqZ`p7AZ-M&zLq;k2AHMi7@rOe_&>x41O5{xSFfoS< zclL1chlx8z{LwOq*fLB1eeuU;>9Nd09l=z;uTk6`(uHHie_tH_dQX(WQ(Eay75~F5 z{iz~^1T)oB#r#m*sbYU9{#0>ii$6;SRazC=bB;p5&>psx#hU&cG3SaqN9?)c&k=XA z_={xl(pLJ*#9yAJzf9y(O@En~%f($L_Hyx;iMy7lF*2z3t)cuI#NU{uN7jh5P*c4@ z;66I+(h*(XAnsOS{S4mTO20(>(k%TFk=r%>5;05F%@~&SIErxhioZva`$_L`O7I{n zi23<|_y@CGJiA|WJs^g6A%%ZH;-88Kkp8??%P+({mZkrN2#!ZhJv{N)OFkj~H)5WW z=t+^MBz;nf%Y_6+mWf>s!bfe(#Q#p~2v5v1F~RF^^*NSy#7_8vq|ZzJ1vJMn&ySVv zeb2ubbNJR@o&ql%(C>5YB@)`R|9UV_^qE& z=CI(hap6WE?<@P!^zS=a=#m9)6;uX?1iRtC1G7%^%?!F#K5l;CfbtcB2*_8$oBgYJ z1T3Cx;e+_>e2MkfLM#`;hh?F^ z073YM*`iGqcC#peS3C)-qQOF@3J^o%xs0>xw2Ur@dN;__ufD|eSSl!9`NEteeDO-! zSHIpDp720$IE1R3bBx&kUL)mu5PoE%c@$L8an6*{+R5G`pJ7x;v(gI#X;m@m(`cUp zjNjV}gD?D4TX1OB0sz3**VuwD*zZIKdno&*^7YntLDM%GUE$j#P&Fx|s#J;oRvCSz zmi1cc{Cn}Ar?a$lmP&F2$xOhrzKX_q%lX89cHexIJLF(XIy{Md%6!tHnO=btbiT7+ zkG_(V{;Cog9K*tJ%K9Tn;9-5ZPpMo(8EaadY9qmVVojuolyw04<*73E^$p?=%-#e2 z&!>H*^=VA%63usQ?22r{-l`p@=*j+6|2ddHpBZXCESFV$XZ)A_Vzf_~c_x>u4IeUQ zUT@owkyg!vKJAEuT%vXO7opv&I?l+ZR-gK#3JeupPDKRWkvd5-?w7{BVEpsOzhNwH zKF}$|N?+A)U`D|(LcuBw;b?VaGpCm?G|KKNoRSZYhRHO4G43_PSzbdk8)OEZc-Tk$ z1hkf~PYb7OFKi#??RL0qZw=t^9++ykiDv7hSuXg#pvFOJ^r6Q2d5whM5%X(Oz;EOi zN|6-?--U5ZwN4CQfLU6nit)d+NOyopV}2f^O_&jyu$i5*#wLun8%G-k`K2z6aZC+X zW1VIsg;!v6#}B~V!^b2ka6Lv7b%Gz%ngt|n{D%|V7hYv1sw%F>MZ<$u9MLzL8Cpe3IRG8qTYUo7#k0&_aQs%RZg{P31 zxSNc-ogQamAf16h`a2#Xg;YRPXPvYfJbXa0V?!Vq!i&12jYE=+V>`I}xa3~~DsLhU z-y7ytk$zJana=@3lmiAI0D=G%lx?Aa;YV6E9|AgvKXl_3GL4ad!;LrzT+yDK_hTOm zdBjk|Ca`+P=6wm3vryT~q-0~5!uOzdWY@0Jqi3s+%E16Ktl28`Xy$)7;~eH7Lj^~m zSDDGHSOgYfs;gQuOl#H;R(lFS(BsW^0EU_86+HN%0T96oM)(l{bC&9-;8B)dthF_E zcrQ-CIW3-p&(?wXv$!`!2V$}2T`%OFVmBN5(BXNvu~VaIzA#twJ}K@6u_*k{3sVo+ z3WMY+tWQlg^>&iohRXS*{nXIZ+uo9nsR#1t*nM^CvHU(7?V1#R$0~GJi38ceLoSI*KvrB5Wl9Mu zC|(`9Munx?h~veb89MgP{9{X}89CFSq>**(@gm@>twn-2k+XB6yVAjSj&_Ik?gcPO z#9g53Y}&(%4cKsWG9cz$t>SC$d)i<4Cknk;dRt?73z59Tu88^@2xzMU+_BucvVu!Q zy>(H+U1IJQcbCZBp@Mk^^ftn4M&WR+U{^OsE12J^;J0dz1BRGY@QBD`D=YYo5LsYo zdfX?L#8mLKm}kU2E%HpL;5o%*qn}a1!-#-np7y(!L-{T<)YRcab(*n9nP>qbi=~HunBC=ZA%Hr-|+%|2)m}aF*vG z!j7yt&$Gt<+(f@5&#NHNZmOH2b-uAJbN<(jeFn^${Eut?kGl_HP9N6K z=Ua0hz?cB=T9}gL08ud{0T_a6isnlR1B+ z%3l0moIl=H*B0SAYu3KuIt}OP;k<1)Zxha&h4T*K+}^5zWf*+TDbAKzZnou+SzWW! zSu%TYmY+RlmJQF`tU;d3?7ms!xc9T&tg*BExp%Wh@ay<3a+GFOh3DWa*|oSuyyCe&q{;x)54um@#shvMs7i5!6>H-*|zv#vy zQa2jM{K`nKh}yi6zNvh0*G^RPmvvY@iOP0!L~5XvKP^rqA2rpds8lq zp?pKtzx}ElDErC2@Fg|apKjNLQjgo5r^1}L-@E&~d(cZ1ZEY5KcdU2sH`|lVnTukr z{huHKa6RM>^l?!~b&rD&nNP4I$%o9h^v9V;BKOnCZEp^X+0CSnNIze)x9e{;7pE&2@=fF7_yO=r9=- z^FxiYlzu3vPU?j-iz0WfraVr}pNV57_Fi01U}-wZtY?l8`)$GvJ|yP#@HG$Okis4y z`^)#5*J#LOxvuEYtT~T?LmQK!iHdJU^0!Eiip>;zV(d;Km?_$Sd)hk;#Ns}#&mHVO zH1+&T`G&)A)7-Is|VV(br!tvO-{7W)g_t?(7@+)Tff4QG;d zUvPI7qZeZHd@L{E`4rG}R(PCrs2Ore5}jEL4LTD4BH{?J`;(&ITRzQ9-IM9)?JKHw5KTs#&i*NOuf)tg z3~TuUqD)^ZZoYbj(bj&51^aeVIXgB4czrNy@1bpVx-9f7Z;U*oDXLjI3st8&|4ZVC z__gZW;rYnD2sEBzF^$FV;A0J@2goVOWzCpvg70Eev~w5 z+tEWan!vB}Y~+~LXR79la(2+uWM-i!RM|-0j^gXl$ye?(>o0XDsF9NV7H!#%hc)6G zNWSJWmYk}dxIJnwLD4=$T8+(?VY|XBx>g(gkkoG5kmL*D#TXcGA0ua`RI57sk=Zo7 z3C?<l=WXtS=lM<{$63;K1SL$6UTvdZ>KZ}2L z*BA_;zFc^obd2rgoGvK>V-6i^%!H2s0R9-tlJk_JR>`~>2_l)+4_m)q~-g)cy8kiI9741XL zedP;TS+&nj#al!A6`t+?zRy)YtMJtwl$UlOUKI}1CsmI4tI>h}8}93y9xUH<-_BED z%qQP89953^;OLwFd)XWF9r-S#$|1e;_=xy0X?)jy0*g}AFc}F@%EE)COn2p7zXT<% zImv~(o*+Fq+tzinbv^W?a)A;iXpUu1{_XSn|3S(%wPE$TR)4alyq&AHv%1=o|4rW2 z%cpJhq{6WG>QBhPHTCeHrSQ{Ia=e#Qyg6CVHk`R`p6=~w#B0gV<}_b9-OFj-p6P97 zpb*ZOEL4KYf<4>Yv%Iv<<}6=d?Cm0TF_E3kB41hTWs$cRdV7JF*4bR(>%_3X$V=;N zF7lO&yF7j zdzY8i+1%x;_j-Gem)6?%7#DZ}37 zB`X3)D*cDKbjU~EX(-#R+=I?SnhK=XI8&ogYgUF?rXv_}-4m`eZ=jq>!xNsb#n#29 zq8^Mf@xpJwe|FSJSIpD4uJyOUk%Z@ekez6o+It-aPewaI`#@zU8h;P{ieSrS(7;ng z-Hc}@Y-<(JFnpK6GSI`|P!Wv*>_fl;h_i%^O;MG2^)Tq^TZ}ODK+88+{aCCf^f2{G zO~bBP+|KD?n)L>lI1+k|;87>Gm>Or`tf9NorGX|j*n}RYx3fosGV5Xbt9YzOaf8E* zs0_#mJ77o3pdMyq=b(kTu7??29c9O&OTb~qc8;;?go(q9uaCoek;9I&6J>%PW@0)4 zjC7J7W^!ecOtzC)@g|;|EH1BzWo{CO^JiAzw`5L7uJx611ypWQ<50OurhPd4c}weW z`13WV&ouRhr&ks&?tp_v+VFR4F6+J1|Ese8^LI(!mU{U9|2$tyJ2UII|6RU+)mQI5 z9`D^(nKoX#ww_8tRZ^;BTN={V`wX(c`^&5#mH}($yi%rD(sH=}CWKp*F@4;tkzArK z^{ZgBe~A1Gkz7#9SbG0K>OU$@eA%$l{~5{OTctzY5K^1cU(+i6D-fTz(vOYh`>{VZcJ~$iR~_-6$8tnj zZq7Mz!C`2)W*j<*%`S zGj>;%{LPwc$*^3liJaET^}|Gd z7IK}C$nl9kA#o3t{9~Hy$?~cT;`T&tY2~^(k-vmo*CujJ;wkXA|9R>^+32{699>Bl z5~%?m?OjQ(e04JnoRu}Wx63ePQ=A!HlMpCd>$x^5JzerBVclMln=0wem1M2t`;|-y z%RFZHcom^q8-?i&tI(rg$Sw-QbF{mugb(fs<}>QHLO0qURPcHguo2a)%9yqMf|=mr z3n{PK7YE%I=vigtLQ3*PjVrM?$0tyDhe}&`*C9>bse)SvDuUTNw!rXiK%Z2(!#n0! z#c8OIZ7#eUCu4cXoge4c_G`Iy^Kz%idg1*PS(o?R`6-z2XL(^Yg?Ah4lPY(9V~eg? zD0g$&BD~uil@U|r&TsC}sB+up?>`_P4DUZ65O`JY{0H1e{YTu6`TLz@=kR_f96XvT zcYY@nnLpu%G^F2%C+fN$!hvlJQ>ogayZq@%0i!NHO(rX4Cz$C zUdsCwO{&#&s(6An6`pJJG>cSr9=?q$eOe7qU~mW81xlrewiI;>BBo0fBU4pWw5!mT z2v2wp&j+Q{(v^z8A*8xma3WK`8hI$_u}4>Ebv+{0AJVWI;SYQ-h>PzsL0nwdxM@S$ zLs70h zR*qQr?Z5(=RT4z-O8G_a!rxDR99f&79;<}45qlE zQw9jgBVk0_fJ7;|Q#9x7_7HQ^W6mQZVgO$6!>=SWTIRUDKMs~!HBBH%T3AgRa- zP)B7*omd*_iWy4Uz^T4+8+Tr82so z>ZhGkauwBYNu?@QyE-B~Y0v#<63@7By*BfC$Z00=3$gvp z$~F$4#E-^ySsXn@|DCUS-p=viH4XgjvHeLLJxu>CB)R)-(Y;hu)r(`XQdD3ZG_sZz z-SVPYRup{w292x}wCKCC=G=+R@0y!)d2DZsqaV?nx3H9T4;0;FMb!)Vhu;0N=$=yV zVug3O^H;;L`r4x#+MWxo7g63!8`qS+&nKH6D= zWY8k5G)Nr%Y&UCEF5Mi@yC(N4&D7_kd2VM9%&H(ER4Jt9O+p$wtB*1y%JFGSN+g)0o2Rf#O|Mde!8 zmgUzKdAwpzUNzV9ia9lH6YISO@`@1Pma@CG9Q^F7ys;%P{Be8Uxv4xMk^^`9-*Rqc>CD0w55 zYufMssTTCCANgCV^1+8I<}+qDxJSPYx7TWGx0mf?x;+HZ#p&4Jh&eCS5BWz~fDoX| z-){c4D$A>~tm=lz0Iw{|6=k`;f}?k0uDG2ruP?ei-4uuNlk0IG0M$d8oo)3_J>6FA zeLPi>nYl;nc4tzzip=~pd7nI#BHiguAI7b98~#tbha^nTNI(H@2jfEg^BrDg4%RYp z&>$N*+rnINv)QdTG@CQBY}x)slZ-M`38+fBEvd~jB>Y!Wo05-msZF_v`x9ep5C-tl zJarQKyJ3c6JYtQ~H)>?|JQZG{q0k(a zW7hB8HMSlcJ?^>8$Z``eQ;f1UvyyX>THK(F6Qf&FTzI?&P&Gyl+^0@wI?=oK1>UnIi=3V`HrkT`_s^)SUs-FbfcT05@js=)7F3%3jEuVC=zA=>_ zR~4Clr^F^$d#B`XNZn#17nb1&|75k?r#PX^0Hjk%H1Kim*v@z`TR0Wj@sffYj`;B|1Wc#&uW$DtE(^{_`<8pa#gDq+W1T= zzYo=1T$YQD2u(wRe>Jk17X5&gR;Q?dZt()$sT6VS%@&$+Q{Uar|t}x2KuF zykkN4yQi!fshV7h*`LHr%%l{m@1F4mR#W0~h=Oygt0Bxj0=5H}GRz*8J2l|TEw567 zyXM&cD24tN?rS}mWds^k1DED+C+V{a_q9=WVn)HHu)V|E2ovLpyyd<&$*qTa-Q++K zLN6+^dIBDqRF2Lk z&kMTJZc5#qxKN1-FY4W0se3Re82*tve@0o(EW=MTr$kEpt%s_Q9H*Ty;NG zM?hndZU_t==f{}ys_sIUH6mURn)3uvD}DZ zH7uI)B0RODC#_P5jH{7B>=sd3CJMr>G$iyA1Z&*{RnhlmEntRleqGM1XX>Qyo#uBH z`{zm-4=5Q4H8ON~JMo$E88Y5N_Ph%8LC3nsTui^K!nfi#%EcQZLup-$3pw-*VV;lK z#`)U0GW!sjD|YDQ4q-`@SqcJ&AnT9uhp3m<$UyUf)GYN_5;j=v~W~2jMfe5YNK^&Dldo8IzJWm{qs}zi!fR@ zH{_Z|dTk@QrlF$+KZEu!mvxq5H-USQ%%m-twH1zSMk)^EA6{BsuK71={tq?(TFw8Z z=I7P%ET@J&W3edUqLX>97CCqm0*z7X`O7WpnN7b0s)7O48aDoxs_?md49oiMdbSxq ziWZUFlN%$mVcnZ@w{lZ`t0wQ(bxRI0Kx((A?r$|Ww{Cx(Mh|BGeaw4oSh>b<-%5?) zN~SSfqT3Nf=f+iQ*lco*;W?d!i$hxps!=$dD=^;x?H^9-^PBF7rk#%qdG=q0$P1eJ^qt?7 z?>GJ4^68dxow>Xb{h2fjvw! z=H$(|tL6HFW*cPxV?%DJJL1|Z5A(XF+}M;Gnr^7>%2N&dWJ7+{V0F|WU&g!V8KL=KnxWTYF^wGA}^a$G5ssa~!?z&p^a8;>}#c00-M1A}s`e==a`kE2$ z_eSjhr|JJ+34si;ACK_)O1f4@or!d0b2ws;C%|8}dZj~NtZJtKL>{fmBh}i2Rqf4T zjP-w+<541f$Z4z9Lb<dI99v`M-LP8jxiwkQmiyX<+!S)ZRh7S0Yj0Iqjhg#h&Hcb? zxlgLei7op->b|!j4~5(dYqFqLTUc}VG~5#ndrfWpud1v#UHzlOR@{#EuTC0dVuAl@ zn*W3hS)UfL#{1pL-b}`BGOYH;_+5>}z}2g6!e``R>zHmRK+L|JWOt2Z^4$9)a=tA?^? z-y#GDdu_&^eW2kUY?ucc@?axl&%UF}>59yp^75KpR2zaY!#B87aH?60&ae5xGg|(2 z4R?D(Nt5UD@=HH%%I!_LEuZ|y)#T`vg&khAhqWeDhfoxWG241aeS5lhs`7SK6$b8X z+7oNhLf9maY87Bpmxd~S+N$EBrd-;TOY$nn_t%vW@ztvRb8E(T$UIsbs(7(#eqVKA z0csUjHtmXPM06?`x=Y>mOR(H_`-QVz!+BcU^XcJwRyfy}xHgt-S56tJ@kon0QiG08 zj+kQ{$5i&;;RYee>}ZyG&*v%o^eLUi-*(9I4!5ksF7MENI%1!8&v)ee^zsgSSktjv zvm3>KYq)t5=wVoakHPF=wwpywcWx6Z3Y|ZK zY4_PtSQ>0(ZfUw(^=Z@mqUnCs^qXTmumh$7{n!m0z`m$@ghzKOpKv+JP^|BCYnc@t zGPjdgPHyUKQ-Pn`>9RLQ4(%jGHM_HZeJo)A+7Z3bVV>`h7dk@Dv)XgM(81^G8;l3^ zx0>Bq-vxp9WJkKVV>Nzc)8I`5=3}74P_qDg%R;hj7OuN9G+`z_4%g%3deJYOWcI)- z^15+%QeMf1Orpb}8(M70t{iPp60JzMv|5fdB)IW+X%!N!6UFhEP~T(WJh+~zDC@Zq z8A9*-7!FkTq_4`6^hbS$VREDMw}V)Tp}Z%9RgPq2)DagmEIV^wFg94gpD%?QyI}V! zs)eCLAnuTCGh%pkbewUSOLQK-P0b5-x1Azdv;oRSLy<@oUUL-hZBqMC)DYSczL`UqqYu07OsD64>^3)hOS5oqcR;{i2e%A9hgLF9Uu?)_U5bD`(~zed z{+Wh5zEcsf7rW%?uJoC%pRbzK?TPn$m&4Y@^8Qz=)7Is0s zuiAl96AKQN#VV6%x!qzJz8B_vxbO;{h|HXv#i$QeUxp!A{a~IxE6kBBGL%LY?#QOf zpqm8Iq&)7~%aq4$mf?2ut#0+P47ZlOYea`~#}_ZzcPX25yex zY&UoKZpca=Sf-rj$%P;G6irtZHw9S{*tds1NXY8K3>{pAZ1?mYoaUM$^bI|j*?sr$ ztx@4@vV-Qs)7IvW3Wcs)nBNm?L=qNo7qJ5JqE*33yEiKuvbYDbn0HE1j4vER4XiH; zdaZk?NJ!)lEl~k;c?l83hN!lVe3w|iaAQyWE{^DW+=^~H7p6r9+079>Ib@&PW6$ey z7k1HI0BRSUPBS@lr8_s6uLKAZhNo#Z&sM>7w57YVRgjTAJFTR8I@-V&5t13@beWaq zc4j!+aQ?&9=cfQt8ZTIel3~M5rP}HJa?_*l>1fR;lnr5B-Cv zs>|{fm@h$m>%ZT2%a!;w>XId0?oM7zQVsP_VU)$@!^bS?Y1^(2?~$K$xu27Y4j`}m zx~Gk&axEP%NB2;ieM)^hGw|$I#=PF;-t2ONLbzUpT;tr-Akn&{M|Y~ZK_;jO(P-<2G=*uWQwyiL z8pa`M0ceH9I`Fed#R96D(SXV_PmLSh3sq`IRSK{50(T=Eg?OySvgFlJW)gP4WBLMo zys2|F_;`7z{JBTrzFc`W5?dytvZ%|2^C=(x++TY0<@Hjp zyx!Y3m%MqbxBS;$d8=10=+z1O786BRS6FYq?sa$f=DoN%4|_u&7PfIsxDSENpy3in zycZ`gxNuiYb%igilLCo-Rxv|UM*A?B}rQ&}(R>1Rge}#xON1MVKY(T7tmdmrj4U+-y#Kf7acybd-*)Z7^_V3p#<|`Ut{%|9 z3w$Nqwg?-b@TFZ!)j}_f#+v|sxY^~^ng!46^bGyYM{7EbFIR68gO@3%W~arVOt3SZ zK^KH_v`>(cu^EC(XpbVkE4uZtXwfcX=Th|2V6LfSA_TF(yg7^$T?0Fg^7{QO21Qc= ztZ;%4mXd4_i(I>DtpL9^40zx4aIGQ2MuvM7g{fgZrVRV?WVe~w(rje5H%TUN88DwF zo+}viPe+6`@ovv*YvMH?4d|M9sYm|Q<6r7=XY}iu_~U>)JCHs%kUTq}YvQdQWcDR| zWBfaYw=evPx|oAk4P>&}dA)Se-+PAO<2ya_cF#%%-j)4w-GK6{-sk~P*52rGSM=$|;?39CDcc+<{4QHD30Pj6s&lS&;-?CQF8JEFUpfx#IN-I=ov-or%(jxT7a z{GgZq^Kx9TJ*;=Ajl6d25 zvnqvGsKwqiBIpZUPk>)@#|XK5gxob^C6yQ-I|5uFk|RbaANh%Xb8^2qr9VpDih=m8 z0YKaq?7rCRDl|}?j3OL*#YnT=HSrN}*5WYUq$PBa)Su5>nW61#9)gr=x55d7qzKam z;-W=6=_(DvRmSj8xfcp+W|LZc@Q!Qu=a&v}30+QFR8QE#G2 z-7)X=uLYPhUB7w&%n0bpd!dpx2(BSW!GlJ5;YG-H8=5>yn_SjYhn~6)DHo0%8DYQX z2$aEU1pN69qMrJSN<(81C$%+W@55Y9Y4{@ePQxPza$`wR3d9oDHN8B|** zoNL~wsB=#sNLu)K)UVNG&lFr-jgIuIq1y=aYLHHSW10H<|>>45qNa7I1 zD!gu(Vp`lfnMpW-k@Tpz2_d`zL3knA-M6vD+FV~mvv|W@i#^Osv>44K)^=I=3(LmGEih9?CVIg zT&Z!9M>Gawq(MyNyx>dZB0bY@Az+&Mjn#cpOve+pN28L08;J&>_WhD-XEpgAka$78m0v zdaHxcyr2`iRa~1`Fjaq!OITYyK*vq_w!ScQ)J^8uVcCFA@WHXsMF~mFFe}tUH9y>M z5YnjB5S`3$&K6sCoju>1^h`YF2b)(oZ<_F>3;v-dYhUtJbtZEct4p!5o@8{qGZ@=s;zCMdYLrov@NE`(|QXZE*|cZLtVV5?2fW% zkWP90It4PyJ!#$gsE$mOXKaQl5dIY1P)y{0ZzX&4N9%rX?Tgl*>ilEkACl6;>SH(> zl=0~pMsC24#_tOw0*YD$YJu2MiN0?QF#MHq1KK5D@nqr{j;ETJ^rqAI9O(K`uWZl@2q zmE~HU1O68!zL9$mvv%H^1F9?_ke7|#RSp$f#!n^e6J_<(xlcr&JNlO>L+Bt>`PF{c zUp+e7-gWbQDm9muAAn=ym6YFOyKs z_pn)&VX#!zU!tf%aS8BzpAnxZ@0Na|xLZfz1hOe$FYS!{>0{|GVDx)+WUByAZK#Zx zs25nqKG<-hOzL@;uL=0i3!|I z?PKx{Prr^0k*OOtzLD&6j5;wLc@JHMXPg23!m3Q6F%&(s@C@8@p;#2TvtiKrQUmTY zPO87FrdPaC4eAZSLKd6N7a20{Qh0$_*y{L)O+G-vBD0LdF$Osb1Gc09MsWbFQb)W( zz-=|vVPbtyKTs-)gC>X250eXrv#%-Q4G4F|V&0KGSqEyXBkdiMm{Gc0Cz~!x>VM@ zm^Pr|H>P=WHUnx`_bM0msRpm?QTrK(Rwqjhc;6`8Z<0=sZ7qEl_ho!9uu;o59Dzmf z$uDP$BhZJ2Nw~~7B95~!C%=^Br$nOQRmq?1N}N7F#atr@NZo+y)s?G8x7 zl7#9bBN`ii z)NzC4x(JfRwfP1xnuX8+(@2@}-4M8@$UE>){N0Di_^B`%{R&u7bLDU;>>XE%G(^iS zMdkx;TiM2akYV~jaXVeC?BtQ%mEF4x>q@KG7mu)vU=Q^1R8V@SJkf{hsIL!MtEVOvFyNnb2KjM?$1^R z*a)>6*9|xW0q7=Qg`IB`kW^D9jrPn!fbllqU_rbdJ|&KO!TtJaz*4hK;Jx5!xPu1l zQ+Pw-&6BPCmaEnLrBL%m(Ru;GS`%=kP;-E=9Rw7mPKB{b$M2f`KnJc?Eie})UwC2$ zE#EOKSch}v4JphAC$Hom4q>pzY7Jm()A434>K>%~;mTj69>e~aI6FjJjmmoIi+L~2 z0ZnI-_T{}alknW5%A}!Yr(4!Xhc$O^dWlKom)b8Zv0yX@gma{!^XAF{2TU}h4xCZA zmWCsqB&v~_E25CdE5(nWDfw2IW|ymKH%WLjtvGZ8V#4xq>+B0EJf=l;wOvC9=0wxy zXnjULJtONLD$$n!g$fl_+_MG8z5x=^F@?fdH=Z6MD2wX=LE0Q*i?nrG!35SVhF^rN zN|Q02rvW91Y8?(pI#cdimkF*O%@@ngDjR(jV4^#K0$L!i`o5un!3cylTZ>z>kO0*g z##zN)rUxvPH8%jn8?o1DI%&_saaJws3DZZ)0GC5sn2Tu`9Ah?9r!#EHxb4QxvGz9$ z>_aQqKIh6og^O_oZM`Zg(l{u+00XAx!&2)WGiH&&_M;lk zm2mD2=V*}`Teu&PS_EfQ#Xes|T8}1wDM}lTMO5yH?AWZUKLt!-$Jx*|Az|PN@O-XB zwWMAx%y-d|HaP}LCFg*t9WTrnwsKpR$qg+l-Gx`71frig^K<2f0^NNX zkWT~#xEF;ig)a3EEHSWXVYX_ z;HDse%J?GNEz?aSUPiOt$y(0v)+q|UDTg+PGD{IG_wXqh^IwbWHa&cnQDFN%Si9iN zezDzqk%rt%+s=LwLd*G5-oL7jBsuH>2Nr%5mm)dLRS$P^sFQh4-nRBA{H_;aDQcli z4;6$^Xf$Z!|J6irV?Bm;iLj)jL^^h#NtFUxaK+{}YF(SC!%#UA`IZJaT` z7XJio0q4yBwCUw8(5$SJCy)kkKqqrv?YbDA11}X|RPkvTb~^nqI*Gmje4#?UV;GOU z!bp5JoJWOovdH!n=E=la$!r7X6yvnXU$41(A8UB?!kaLgR4hnDW9{_{+tL(f=;vts z%?{;GqI#NhV7ky5OJlzTO?Rb5nf(T%dvUKwWNzyy7%N*I|O#hhTI&hN}BH}bc;SnEL(P>BTYzw9g(b(rUWM3psIXg%fxLT#y zp)sms^1_JQ{wNhU+%^s%*V9^n7L5{3V8Cp!7~nS~o|dC){L&hG=; zx}5Kc%DL?Z|7Gp;GHZBR1{ImiY?-t1UUA{uTNU2#NAj#P^ejs| z5gYnChG%~?(F-Eain?mW*x_f^{@lt{cFi6Y{uEL?EpnG+D%m;J7N-(`| zkn|;6`8$g9t0lG1*>)@azQ2imLFzBJrfsI3*={J21sBK+B{EH^i~!9HC^C6UuGP_J zjx;Dt>Re{4y#lzRzdD`(02rWvR%p*ofO6(mfoDYp`U%L2>RHA)-PwRYfVVWJA1E$? z%!zNVDOeqS_&|fh9B9(HO~nHj%erA$YKsWcx{D&hAbRO$h5Hl4OWAC4I61PH$8uS0 zu8*r>1B88q0}m>+?0S@|bDyO*0$PIBmq8sH%++#{w{SYm`ha&pxo(|CHe+Hb`J)(p zgjP(Ga5Tovm2rnOK%k}<-h!tBxquKx*Rw8(%E*Mqg?bd+jRYV9hepCL;;Ew72xO%< z6f#+Hm#KP+A*JSqewe za3w(BZ^8wNXewXbYJm8~Xvh>hJE6~9@nY#fr5&O;d^Fi12^e5LRlzXmkK!6Z=ut$@ zThkJ{DS-hkqA^>igqz_IlXyvMfU+`lmr&qF*Y}Vg6bL_lAv*6;>0F}x2}5k6rOaki zJ0RY_S#Y(Zit&+3E!X(aCKVx5kvV3s@*JA$=5&Ocn%)$5k^bb%BdHrgbX4-Kky+z= zm=kO#6-Js#0J#oS8>v~-y%Ms853JFwOgW=*NOdmg8&hbEepEBy2*tuz0qc|42qwce zRmrV}668lYF3SFx(@8TthubSHY@s7XN2?=;gdj{R8d3n+p#m*+q}WZ*L$=r*S@_Fs z9L}!o8U-BST>P9jPkVU=DirAbtoQXub-1G#-6^=f-GmYEt_5>2TnHMQmb#PFr^yQ* z;Rs&5pYrxks=K7~SIKNtBAHDSbPUydQy<+v351ChG|7IAADwU{-7U;H_3daNF*aSH z6h;_I-H{w;6$#c0x5`cH1@IYMkWbB%9pGhfD)OZ6gwM4|n)8t{0C%=T&{95f-oSsNTf zI}hH0Zi2c5kajSR>JnKME^*OIzp?|BS-hL$3@ZVJ-Uj%&nVp&`i#WR_F$zuvj`rxv zVZxV~KdLu07C0(Ng98c%EDtcruZedIx&}UYoK$a}6ii ziQ|5ekg+n4886?_2@V0!i(4K= zv!Gj`;l6{uyrQuK`Q2FVCfM)dQh}?GPCbEYy7z^28{~J>xtkHvO(#50Q`V!s29bZ3 zc3XOmpmPhwE|6%U$O6W+?Mn}moq>T0bay9|+Baz-?!)Eff0bdsEQ9%cUc&*}8Y23B zgX6nai#SNWMmxR`iu*2Y`MRPWNNWyedn?n&Xr#cWY1r2#{#1^{{+S7+&tc5#5>)RP z;Tl9)<-~xIkuZ!9J0~$Yl&f|fGv=#QoMG2Z0YS?(Et+;8g~K^fG%oC%s6$l%dgJ49 zlM6p04m_T8;s$~OsUuV>U-rSmU zlaNLTp$Q>`-aCTQtGMDWveqSpDj+swQP4z?B47nsV1-0PL`3DMpscGVy1-&x%UVDr zzvt^S=O#CSeS9DL{pa_~&CHoIGiT;g-o1SmqhxE@GDzoPcAhFAHO*rDfKqjBx~ja4 zqFK=c6>QA5e7HHRa(L6biHNL#80m(`Ng|O8MH+@wd)gtcGE2g3bPAZE=lLa@ zoLS{-VgIO^@+A((qhVhhr{TER`%SPO|63cql4gi!pr_c!D#I!W@3ikon^iy!Q1tbd z-OJ3gXW}M2k_ssE>|8S|C=9CTLy&+dIOK)Vg4b873PS@n6($4%qR10|4zCxKkr&MZ zzIMDs4Le<(g+<)$SI0A*^m>+2o9r_y_S*Vhgb4IyyZgDY#mSL5z70}nl6Xg?eqada z7!nwckI+mBnb<2F8TKWg2}SQQ*i<;*ME;pT4kS1sKdxe1JX@~r=NQ&IF{7KMQDL~wIQw;weO(pmRXw}6ZTiR@R5c68owqhMWyPI9;MUo z2QyM(vOu{*q!{1jS;ak?G=e1_S1+G^@S3g9CEQ0U;a2w@cx>gEZth3e!w;bdvn~iQ z?qBX~aJIhXiEsj&bWhL zj%%THI=`4>wPpdoWkC=&sd3XvX8+`fdbrZZ6+BvjZ>lB4$>D84PUlm6SLaxEGPJad zV`}ix3ZHC`sbdZkVeD_4{?|?Oo2LIMxg>=0Bsekxo8v*(8!h8qgCksL63Y2}%rwYN zK$;8toWg$kWX58%CYVDc5Y>%5~tuF#oIeA#K)fGRkN|(p2~^tB>x`XHl4Sf zzoi@L+w2APw;33SxgdY5smIuV%)XPqs-DP->B|hn%dZ3n+s{h)o}K7s+hy@`bIO)y zm5&D#Utd4j>}Jn3r>7Sn1uA}_G5I@i9284ZM2Ur4el=}@*)J|97r3`M!dR1z_*~%_ zg4FcMd(Ja66Y5C!jZHoX$qXc;p-Dod!R8r3;OyRVPtCli^ltaA(z$f_t?^Pg2Xbmv zCb&15^OF|Msv?-yX$-M~$^KKG!@|RTE8+a@?p8Pdh}e z)8x?by$p4Pgc6&Z?7ZcvN=@(D$s7|O?T+ePirH%A_JeKno2K%&N}*6?#X>=*Zf|hx zG;n71K|t!*=<-CLt*S5eCUs=X3M<#N>6 zI;T-C_q%hN<#NRyGQo>QbsmI8z(shyh0%c(lh{HkpR{KMUQuaWa-hiw3aRKE z)y1(_aSfwT0~LvC7md80bfzu%tFb7}Wzap=igJD$}z!7`Ea# zi%O+)(CfR++48LsCW|;!d{*)M${5N_e#zLERIHlS27yPNc}w>92>EZz&Q0EwodyAX zLv}2oI8l%yjz*F}{Yw4eXn5^VXF_wJgJl$gq(rC~=8%PuswO#0f9*O#NFGliftK>C~s%dL?*nR;9Gd#>W+95v5a&{ zh63{3OpVTTWk*8cSP47SkvAhj(c+fmnTB6S)rN+BP<|225_oRQ2th&bFwKqt$*mVn z_9ufm?RA98B&EeXn)noWHcJ*-e>*Y561qC~%67o;HqmEc1mB!WrXMx-Yo_N3gfTf# z)i96=vxYHdwVEREbNCH|Ty z;c^^xRHrv^cyK?VcX5{WBbn67Gi!B|mwOv^*ON55x`@1oQdrjBlJF6m#psB7?9oin zo>`n8fcgy1AYJt2-HtTff>3pV{cew7XtFs&FQgBb)rgXYF zK~q4b!+qky9>+=HP$@^H>_}Fc!F;|VdPXMyBa_2e1b!vHv!sR3nNNg7-@=tL=_cQY zHljBo3StcFZM()9F+=!XsUdd`eNhC@N!{R-&{dnmgM1wi48|>l7D^n&X~`)KvBV(Z zN|(LdHr7~tO?Z~rzdP)wp6OhJrKYqn$Ry>>l8ZFsf~tynBUJERvWY&ooG)>%nN;JB z=w6)b@GUej<~B!B63Y{fo}Trwz}GiCEbJLY9V2%7yIq_8;3|1P88Oa&!DSzEhUH*C z;5cJh6qfa!oV%B~!vg-5t?ys&49?Eix}RUOzAA2(H9mRSmFK6QCitXulLma6n{*yM z3<7Adn|{<@$QlxXwTMO&@q8{Lgo8q8M;*9Sm*iXl zUP9=gSBM|F_R#^Lj|kvfcmnqg$7G`BQRyiYu^cv(Sc3>-#lvU!@+<7Vejm37kyM@p zw1zG7%%m$rOPHB>fCQ2!Y6JjiolQafPLzqYX18SPZ#)w_sP;)tfl=bphseB4+wFf$ zGerkhB2+ZUpdL`iQhy}kd+d(Y3flb$13V#Yn8@880YH)C#r8q>9fP`ePod%s1yycY z*q1h8E3I^&RXWq5P%ed0q->_rU%r4@9f`gR^`v1MdgDH4OHWer80LR3rgO@saa1m0 zyfKWdv|l0$CNeoV7Mib^o|?NCe-S1EpN^!Xl6;v$tnxONPZRn_$Y2f6F+;z?@m8kDkbjOc3g zQovS$ZKl+pUr@=*BT*%1g_Sz~^ZY;}ff9Q}J*xFa!&3DyhA>d;{Y1#mKRWEo>wF&7 zs(ZN=rGx7Sl@GRr@Uk769MWXHEq|y#q;+s5oEwVvMqOp`9F{r4H*Zc6u+aR|4TI`F zWNavLuKm?Md9TI`v`#Mb1QC+SJ!pE%wR9P)v+Znmf~~pAA;BEN1K=5*u!chbA}9;; zbue4h{TQ+2P2`BC%oip9A+S~^k`lH-Ohy_#OnXV%)KHO&)T*c|g(aqB0V>db^LiM7YeYdBD4W0sH7Af$^b}Whh#S@XQ?u z2*oACKyVs7aWPBhWSS~sn$#2g*sTE9#n%=c8*g`Prvl7t!~@PYb!zhEvW1lVYC2YF zcT9ou^{fI^+jN)c4wGL<7v9;3p5lXLBuSXU5)Zv%@tXqq1CHae7nx9DH=fnydmRQ4 znHPx!!uq5%VxQb->|bPe*p$WFvmm0o0b@&$u6RUWRh)plI=1NW$u^o63VV_CZ2ajTwoB;%# z0s}aGCW{d#w^`Cf&CiXQ;qRavt^|}=@)efl>+0mI$RM7lGlj1Qk{<&dM*yw$ifl$c2Dh9blB+JZi8Ip@b%0l6-X<(ZSp#&{@YJJ+iP6J4{>`lO+nN7XX1jX+KLk25Cskq%%S)V zqjn?mricuLVr4-H#;U}p#0xB^FX>AVeC=- z9P&bt=L)+b=B^x*+ZRpSki9$r%itPZYNhj|-+=CrGz+Xr6_Lv9&AFLkUfyK>&b=u$ z*f7KSpsLg`XXox#V{U_Un@<^k3yoz~ZZ+-}v<}}DLT9ZV3lp&fR}2Jvv7bj+2v=sn z+qj?2Aru8g8@B0OUy!9v4mj2d`^Gv9_ND7hbdyw>tV3g)qE>I8A}>MjnBc-kV@6=r zM>C42P#|cjTfiy?WYxfbp3}fQ3*#1|xH4mhnUZQG1)yvxN0|a6fU?kP1p?IUqDX*% zxyxJ-MSEijoNO`+WV*Rf?dW9>xhTS+P~4-^>}XX?>?BPNP}oyEup3~LpL~xlT%z&r z6fKPBDV$19+>ept+`i)4S3kicNa?{;8Ddpb zfZResyb4D%@u)IwW>3#xr!oo5fQ+-!n;&#_F6`*W{DLrg1vn>s)B->c)0Abj24>7W zz)xw)aS#f?>Rt9-ERD{u03k~mJI1q{u{>n=B&wVj{e+(|TERCW2n8fa%+P}*t%Z(y zs!H(^+#Z@@?raRP_S-yLanhXN51a=fT3-O%+sQX?Fs|e@Zp%I>y=jpWgq>+Yx;`_ZuPP{r{|E9i}qP`lD&IktQ*M7X5pOnPA3i=HFF5@L%aJvYVs z#=Si(AzVsVay^UIr>UG+9tVP@*g8Osk?;&N4ApT&bFs!1FibZC3`1Y8 z!@UDsGn_%RfI#py4F9%!p3HgeQGxIk8^mh2y?Mg=ujTHs-2N_=B})hrK*G9Tr6pz9 zKW*Izl|?R%?URW)&UFhdsVQL1MVVcVP|7pG4b#ojm{uN90SM>|cL~%y@xd}~OCqrT ziyBHe zv%p8}cjQMAd=|bP)_yYG;K{`~?%lGLTfzP3DjaFwTWrXT0>51Y^ zA%k?BlSnhdrrIaa8HTfB@ff;}G6j0RB$!)u%u*s@KLmsfoy9h`r6ncZeRJ z@7Kh3ZJfT9wL1!}qr@sI+YslQD1l`5a<2`UwhfsFNtR|7dr%ENnfhze+HKw<$x_QU z&{`#liOw?kj9NkW2HTUwdBR)5qyeL^HDyJqnrWIt(f1BZh>j;#sDl)bK(3AZH)(>5 z8;ft#kQ4&ekF^}i#+{vvdS8VH$I1bhX138Bwn1?Q^TK}Du%D^O>I#e+Z8RlD(SUP^ z4G;*)VKHMkz?k}slv>W&teK8MNOF#OC--SJnzf3o-IV#&xw$A$R|EKr%X@QkWnOyD z`xh89k&?&1&sgAxw|9xKZfIOMOL0kC$Wba%9sTk=mV1%h|kvap~L+= z&7Mf}_vZG>Jm!thGwAN(+`W>ul?4GNb&Xy%7rj#u1|5PsI;Nc27KJ952~@sWxVu&( zheIbQ(3F_Gs7stq$Oby$i$=5v4q^;({AY=u*UUv`Z{aYqvV8-u64v%S7yoU*62oY8DNv&Z%KUyVk$}G^;+~3f?Ci) z0}&Dcy6=^;jYZtS*GmYXw&R(S;Rs(9H_=o&3fPobZM2k{2|5=RzAy*c{bYuZiBv>} z40kmjL$F4^?4!2TH%sO*h3y&pbsZ7TrQSNkW%apahXNgy2US5Ctii$&;yD3+!efM& za9w9nM7ASHBJukYal=z)p)$ez7-t5Fn^6WunVG&221j$Nc08ZKtAz&0{wY4~goM~v zFqD~J=0#}E$I5OD!L+n-|Y7RjMHPY>>8rQ3ilwgn%zy2bN)}ntpE|FDE}BY1%zr6fYo_03d}^ zLg^&6g-yYS1Y(Nw^tUN1V%fe7MnY@o2y6*p&@8#DDkGs2V3GK0 z(fr?+<4Y^>D=}Fqikxz6C%H%yTiF6yopEiC&ZKp6`JW zUgDkdc{`-GF5;43@%C@tyyEk}dH0G>{^q+1M4Qgr(T%3Tc1Z}3V5Cr0xzoG!#J=_u z;Q+-GYvGBVJn@p)UL2cC;@ZWr!(VZ+Qj7Vq9pi~8ro%jui7B3_Z0scD7?YT?O&e>0 zS@?E4Z;V;!-rN|YRI+7l;FG_NCindz+THud%HC z?QwD&?Xo3o`HIfoJ70-Irc6{3Jt(tgsM4NK>PO=#qwzq7kp4h&nc!vvtTlwW4h%xq znCkoJ0YZkiE83J(W-H>Kix~jSh^eZgu!ebMNBhdt>ezrj{zMM;#Q6{KH3j>!hG+XS2Yb>R>pMIHgJdoN?{9@UFySYP?W{jIx&vk zne1lvU~ZP->r~(OC>0=9a208d#CR3IsxFamhklR6UgN(cm*; zyvQ8K5pLTlLv^@mDO1p&ZBB!&X%Kl+U}j{U!`LO_OOabtjE$WA)6jIb02Ck_t;$#e~!lDSQFHFqSYXJ6TT$U?6y>koN-*;=2kVO6HfngOo<55JWf- zrnJZ|g9l+n*p2?k$<8>6+r)d*e?-eLU90T!O~$Nh7r@EF@le?NP0%hsLCDPE(T&Kt z*qM8aTE1xAyaaw#;bjfxs>+qvvj#731y=2d8^!fHlBW5nIv=`~u-_x>XN3LsVSmULr->m<=x4-l0 zpAfz(IGz7y{3|B;62R)aV-notxFJ0CYm@gqWc%*1$(_2+;?qn8YJmCH;IH{4$}L?uHKF{Yzf;mtbru0kkB~0G}D$l=5`V&XJ;X4P}6Bd z+??6_CG>fK28XGp1c8Z<4#f`B9668h(zJXnfj(79mp#NNBZYO$fXw*Lq!nxAF!ne5 zC3vTOeFMp~czQgT4~P4PW)4szpRK_3;-~1jk5KrRvHgn3c&Mn^N z`{aH4OxarGJo^C%)g16=7d{sO$>3;!=z4S)Mm*e5g<=pm00MxwkE(JEnqy9_S-KD1 zl3*aciK);ryy{j%gpp?vMD7~+AMcv_i+AS#z%$F?7xR+S39x&{*!SSzxYF298+V5S zyh64wqek3SLC(KpP%Fd0KF5%Tp5O!Bg;2ZMor=M?ck z+L4Irl!nn{GCHCcFu`i#vZrnSBXEl1g3^1)?p5Wv9G~5vz;NaYvktZH^WbebnP((y z@-`?EQ(Jl>i45Ce$vrN5Mj`C1A1BmAsw1w!E^#Av)C@t7jQON!g-suT-{{Q%C(cO0 zGJoE_Xxz)3hO*h1aWojKLeRZ$zi#BuV;aCIaL_dN+Jp5kW9L zAj%$uCivtuEw*Hr=TvB;C%md>JLW)d1_t$hWDSWr#Sv8vilcCT>u<^j;WPo6XiqZa zTFNCL<~c3wi3?rxEEn0u3##!+Lm^egG)}^m&WI{ug0oAwW#Cqn1eK+X;-= zTy1I*;xXAg;X%8H2Q|Y!ncNRYz5Rrs5S@a(wx=@6=P>Eoq7g8i%syz$LwfRuO+pa< zRfMhR9r+a)fhb-njq)Ny3zWyD|W!Ta4NxyP70CmV1(4888%=k)TmRIUl#e?2cJkChdG ziSg!~=mT|b1hV_2yUtDXWytB!ud%(SPH;H(#SI?h)s| z#*{(NA9dzrjQt_Fj#PSTn_CZrV&O@c7n2N;n|LjNWj|=B38YL#+~t$t;m#xG=N%MP zB`sXGQ`0{xPyJeq=%5jMxlH~xqGbT^k_^HDEwp=Ohn!n6giZY z9w+tqq(eZVkrNr#>8V0<8@NZ3Y%IE-e1OX3@Sz%m$&kAGMq_TLk~@b}o9M4J?!5@e zo16GcM$WBPCkY=`8A9 z7JW7t5kx7_2?0~A8~h9KOOQK=2p}tb0Oxr4?P&4aQna2Q$||8kBj69lbPs-nS(guC z24jvOG)=`>iujKbo-Ktsp47*PWDu%vhbJgAoT!MYza8YENZb7kqcCoj*-tNZ+QmfU zT$dyjJPuf`9111a#x$dP@Ft=MX_8Yt;$aLw5a;LMF~t1H z8ERT5bK7UCJV)4yO_0FL*O; zVaaMG`-9j#gZ4vs2*uDCXAlSVW5Vogs_8fB8MduQJ=a)$$?T-~7x#W=Hd;56?{F%* zz6&~lbxXODx~F}7ug6P}e4JL{R&Gcg0pspBPO??djof-nPcd?5* zoh%^*GQ%`Lz}0CQ?A&|}wL5*CqyEboVc)TiV2$p5$mM5PiJsA|3&}2Gm7C~eINTJ- zVwzo4+0_+BTW$QQ34HZghB7|DoFG3AFR)06%BeemBbM?|nib!}K*=^nMj#`P8HHy$ zw+{qo*S7wClz1={nFL`v)vly8?pN4(c6W?)+$B!B5l)NWhF8T@s52xG2rtM)PnVHO zfXJF|jKAzQ#AajM4Exz(ABTNU*dHGDM}+-uVLvPEcMkivZr~~1T?zondrF;gi)AXC zNe7d5e1JrP)~IHKx=(VO1d2GByUqCz65$DhY$kulMMJS%Ag=vslYfdUWCNAZK@}IT zBoihzR8fedM?~=$TZ7iMJ9(3?*M7<5_Zst=Ns;P9A*1p=gp2^wOlD3%$p&acW3@A* zaOBZs1w;H(2u{akQ_lz~er=N90J6T8liw439qp;KV+_28^g)orK4htV+SlK&CNp*C zn0MK4z*3pqHcnYGDDGsA3Y@ZcfIG1L&RUWw1B_?dlT=gJ9+RjFOvcJ)J5c{a_l>`YwP&PYdCeC0KS6`vq*q>K}mOhVlMs9rzd%nEy8 z_${E7zMpqCuJ-;aAg<#BnA+p9ycvFly8vsMexIP>lxa4?rP4UG42foT5f8`32R>yuH-h%Y6Ev$v?t`W*@TVBeqIa z%+&muM}tX8el_a_*lyyO-*4RqZ1#TOifC#df7|h6GACYf_FIUX@iWf89~WADrFUQT zaSy2fA#Z0vSxWB5F22!WpyBEdSr-5b=C7G*K>L8ePexl+HB0Rk%q+mJID*__2?agn zWK8XT9Am=^#}@?pGaQa@68u7jPQWs@g37mDP|#)h!Ie>aj2l94+TYKuUqHImc?wvN zP3(KnR+t~N-NVYzEs`_v8K6^+W#x1`dHE^fCV_+z%#XqHX)sOA1?f3YOldgZaY1jn zr_tYHwYO&9=sHE? z?k)l^=^DZOKyqTA+ZvXCwgnSp=e0pMY1CTUG=hTBs*vfcCnW$Q>@B+BUp+`nRuL8RZ; z>|_7z%wH7E32_SYjU?NzWAJg0#pauF{6y@gL45}hBEP~}+B!)U?~BdNQuzzIbQ7@^ z&^~$DD0Xh1^X7-h(E5A|0~!N~zr}nG!a{vR^Gl#2f8Lv)`|e+8?|y+tjD_cZ9-Cjq z=9e*bw>6}{iS1{6@>!%8`$3<4$eZ^P6dhFk%OIqE%rH!up2J%3+fGizssx>_Y_5X- zx;H~G-W;=xn)`&Jr~2%Sw{SVQ&^n;*f`3`l`90i@y{%oTu2xk^cY6S8Qij({0s3~MqpRBf$WY}+Ce zrn{%>NDBCKu4kYd9UseyG)PdXrNo zXf@SlwoK5kG(nn%y17X4gx8A^e_S#wh#mx&Pk8fDdIo+J2=@Z_6P5c>Jdr}WmnKIL z)rJwXvKdlr$4e@X3h4NVp%2 zXw=YPsG875WY#De!``tI+57yC1;FhoG7-~+vnO9pLudjSRp?f4FPflm7meb!O~`Wu zEt(J%4c1^kg zYWd>mXX-Nz!H$C5+0|5`7|0DNS$<(Hn1Ap_WPT^7G_1)H!JHJDBgg2}=5Wk|B*Bwa z7g%p@Cv=aD9c5fg`vRmqM`gaRMqiJjb8$LQX}s`dOPT9p%TJ7>-(!H_hq7`p;z*2C zTSgU-zZ@YnP+8h(_b32iS3<7~Znj@wX0@=NtJ0*ku@HY+C$(e@Dkz)wZ-DSPA#CPF z%_XQ(|BFE`VGlb(isgJjevN#{!+^wD^LyieXZ&Ur>tOnt5PlN0RV}|!#yZzWc@m-v zg(opuly}R4j90z?FJQ6&72FJyOMJH6FAW7564m@qYb?>rMQnII9GAj=YRq`o8ru`q zRQZd+3EOZT5 z*nbQ<&ycF-%aLZx46*!g(|%!VK7YG6A6D}V-heQM{O%XH1K~4cK+_DrjDFqe%w3YD zzu?RlwLY-gxC<2WNMudIjIgm_Hj97f%a8dOhoc$5BGV&cjx0x@DJ77rSNIS*;Otk7 zNUze7FQHDF0t%#Qeit&d)c^z~HidWxlH__(d9V?c!vK)_rU4KEJ~_Zff7mBs5A8<) zIxg%-!oGl?@*04?Hm8(sHLx>?q*1!_(rwO9Z=YA)t+Z5TPJd58sH11V$tOeX{$}=i zeQGY2bD|i+)mru9V3Y_3pTuFJTHP;Q=~-v~!{winXuiU9_@lui3>u#76sENXcd6JP zZ2i=4r{?}GPrZ+)@;vKq@TF@tWyP~(`08UkS%w6+hE}kH=+5)kCT0nW$~wHZMar>& zfQ6T)$T~>Iz-!o&#RXOm*fV+1JFrp#!Ay#~$wrZ~bEoo#8Nnozrwo|RZx(>3H%f5M zyv}lP6pc-R(W1CD^&$%gq~~+$vA$0_+H}0_CI@7dETb>ao#H*iLjt0S+j&%K`4=b8 z1o1#u8FyW+Q#@3j#n-z2>U+!FJuf~Tw*GpookMPW6;6XfTKl=SO?||F&?xC9_4gY{ z+*V!r-*>ANJwNaNn{KrYUUVz4nb56L^s)JP1r*i@y=S==;UdXW>Ziin;`a0mPC3?e zO>Fah>J%2!SBY-gdrCh8`QK&&x13{IWPW_r6u$NyufF>=CNVI6D%EKc;p?ORIwN5q zWp#(>iGBY6(K7iF8E6$5PuSeV@L$C+9S5&)l|cI*0%uj+E&( zfe;!oD|{uFFN3u)gyQ<1I;Gd)PE?t6##EeuaKU(^`uB^Juf3JA-u5mB>*Q>Oh`h_n{ulzS#OM35NZd zW{f7#MfyeZ{YAp`_7OTHYc~*0L>Yw4up}7a{j*^Y_LwkFb!+k(+zoL-J&+F(5 z);z2+zm3!zdo@jrKS_iWdXJ4U9bo055KpTqog=Teab5eDwseF(J+g$+ z-?eiT`>kyy-L5O^9qs9(vKE!T1?qimP*(~m+9cd9%vRDd9Nn0v6t4XVel7RQL@7wW z6gBu2avtJBMEiCd~P%UW~6O8XI@^sM{YDon3VMl)kBK8og;*Hw* z}su;~Jo7n30{;o+S0F7#j3=hJj~cTSXFOgVmcI+5Qp2iMKGG zhND$b6dYBvf?tWK%t0toJVDUx#Ax&j0zyD$R3EVOqSX8=)uho`Gs9m0SpAUe#6Jfy zV0XF#fgD9GKy{Y~wCQGg7$wmG!SrR^Xyj{+w7wu(ON{`es|S7h5bM_}w%JRK6HzV} zWMDZv#M^dZxy}$D!*yvHJrt8x9ah7F?Ea*2+I=!x!iMVt9~tpjyD%C8*OQ~?DdAeO z!Wj7+g5l;fh&w=DY`&P8m79TXPrtkka&wZ2DFC-L-T`piGe*-%YcZ=n!XVU2m3k+y zn@QU%9Q85O^jRM1S&a4=M2$OlVfX=4et;lL+f^Z5X*-mr!Anh|J#gp4ek|;V!hR&| zr(S(iI8)qNkyqQFM-K-us4CoajAOvhg{(kW+{)-`GCcIJmtst^>t(hNy2T>eV88iR z?!HT;C&_xxq{%l^`-Ju=#a;0*dP@yQe4o3j1myibCriZqCU45&#q-e?lgfYQ_J!QA z{{_#`YIPtuXvhHhxQm#>Jb~?oAu7zv%0EuYP0?^EKQOf?0p)^r#R#g(d>FWjKb0w+ zZr^GKJdu=z|9ph1A5nB3&@muWj_k^^s=ysgIuZPMw!9Kc<3CPjtMH-`Z~%kZUh z6~>gYBo|F$3M|V~za&bUxr?k_O%9u}tE}Q@w>Mw3<|4c&YI<@IM5{@`f{g|r94z=? zRpP)>rz%4`ZMpVmL+JA%H0Ds^$>e&&?#BF~vs=ev1}}@X6TiDkn)O9$%d>h2gpv4@ zf1?@P3wdJyFyKmgpcXM2t+TQ-TYG0KTf1hvw3cM|>7~y7+_o;rl7#g(%jPrIb>WuZ z+pXk@N^KVKYDAa>V7h%qXW90$D_vikh%F4Wxq`_=e$>P^!(`IKMZ|$Am-X zu<9u7l_&ZC*u> z0FEkV>Q*_cwW?X8RnMBOj;yoQq6|~B)urF<97M3Iki8JWb`JYXiZ4fQK=x~X-cefz zbyCrr>i;?B)Lg=oOW_bSK=l&jT+i)k=2Wf+cuoqYg(V5nt4%|MMv9!+iAbT~W+75N zK9}>ynEma3W`Cx9F0u{+hdq|rODNE>SAG;Dzb_82UCm0PvfKxlQ{b3Hps;wzN|5eB z(|TO7dOP!`KAzhZFLsOE65h86-kNc5H_TFn_ZHDM4tOs$?m0S%Z6C8gQ+oiJXjEN- zH!qUU?kXI{ET1R0oh9SAfalDfY)%33v*b*37B*2hGHIH&#oD~%OiDLw9bp0`JxS$z zqdpW5K8%2dCNVU6co{a_Z_d}#yOH!XAai*~YR(IAM;%PlDjj9d3>by90pW zkV02tAQfxC%6r&#_LeQgPWUcEHFeAKditIo>Wpcn5zHL#Ux15s<6|M{Wrl`=b zeaEZseog)B`Z7dkiYl|$|6XaaI9rP7!)5CXLE~fVSO=o2ici* z{XA8ybJl{z7%N`MJy;&`=M#@qEvKP=419x0zzm5QhWq*4gHMreW%THKHv@~8Ua- z^Puh#o9vqMW7I&c#iP8gYvbyNRXL^{e!)+R3)Zi)$8Hi`+&AoZ+Jsp+VRa@?fzq&l z+(g&VIzNOwll+L1-07_1h)EEAn`ZWOGuY2IxEakb$s>s$LxPbhjsjDGBkbyygA055 zAvdhR5Fs{Dqb2@~QH@=t4}@c?zUJ{9d&tyZQw7;m)_Xbc8FTZMi~c|SvK23J+dBoV z2rr@Q0ES{hI_&;-6#x26q$VyRUAmuP90aA2yyzruX9^7Az=MKATG$xYxF!XasZ9^?qTH z{nq;|N>pWEk`9o~gIhc&2NtfBKUJ4l_r>4sEhY78{;S>-bJp*^5ZCg6#} z-#q7}Wfb7-A3#i^L)~TS@8B9i0A@f%Z(t4?Ww2BDG(Y~U@d{DYgO$pS_|Jo-1jr?# z+B5V>l^RFX(*eW*;UP8%thY}V?lXNfqka~)_ucncR&A2+RP2R0_t#i+mGz(EMGocq zc9N$4q&tkgK9KF-O%b-SuK=@0o{{(NJ(mL+>RMEN=c?Fk1=tnpLG%vU1E0`BOHy2`tzOZT_ z-hZSP^I41FJ>I@L0CIkgq$vi7>jR*`-O7!$GvfEz@-^0dlto^d4QFrE6kwg91JEn= zK6tqj`}%O^h;Wsu)EYbEu*jWJ4G+YaGt12J4R-+23J;Rvjl@78gyu~FQ-}ggE-mF# z0>tF^*pa12zCRwAu!ru(`7W&0CvVSnjwXP%^H6oco5N`=Ap@SzrKj zjkg^zdxLo+DW7!$qY5~dlTdLVNLm!k2m{mtn7fn!JkHOCqW%Uh?NanZNkY=MBoqnO ztvQGGogIR9Rq?n8!XI77vQquKMi1KxH3_LvANU1zGGGVoJm7I!&OG#kIp`XSKM?t9!py;M zcbhJKV$4$3l4fwgB))ts%(2zucixC$9O3A!7hH zH$b7s@{}TANlx4ZVQhpaP_s+$3-->ov!uVn9%jJ_6fV}r>vW-6;1&uXJA8RUxRTDa zGtkLKp*$+5MR_DjuwqdIcc)`-7{+7bamM$drcHw^6>@>IPvF*2n(z!1-B~1twh#as zX^+NgnH?!TxqMT^;7JRAybYn2Aad?{5=_X@z;V-?NM zu!QE(>k2!nW_$hyybtG8O5;c`u)q%Km~;d)SYKyvs}-@ZZlK4RhD#Ie&AXVZcXMLXe4H)Cy{AO<_vFa3ISL z?P+SIJy5WOhWJCh%lV6~xkRV{Y$^H-BgQD279W(iMF%nqO1D_Ap8z}!F;CLp8?uekoiTs}i)dK5Shi%w5EL8??b14Khk zSDlEQUx4XoSc%Lma1hm8!GB{XthaV(rqeu)CFiXTf(IDUX>FKO7H-tPqCU!5B!d zB>10D9?5&M8zEKYaf6FesJ?34!}Keg!MVZsX6tVh*$mz1G;2P2%puA70q|x+8<0L5 z_DCE))T`r-;}~3eM1$lpcpkQ*HNS+v?V)ps`5HP%PbRY_wLjjC_>PgD9|>y%=+!Q4 zL~F4i3%V(pCeZI_FGdKlMW!J3N{WRFY!JnRDmibRlt{9Q+55?35;zwDt1Y(9=m9Ny z!+>fGFx1q{#X>SF0o(Y;arXVV7RTR-vq$2Z((Hv?)44adOPP2eqK+NA*DffkA;oUF zX9e;;y4D5dX;JG`ir*gF;oL~s#@Vv=ae}w|Pt-hw|3R++7|g~%RhfD<;$I+<#4hHV zt1K|Eq)zx=wqqi%v>LG>ba!#svyh5(*CVZp(3UsRdl448=mSSWBOD16;ysF|4E?#8qijd;D%I>{Mnrb$= zXRXIBXkCrX5Ozm_jcPZ?UD3EXe2onz&aI^S7jY8MMgUPrDcr@}q3p zmH$`b{+pFD+IswP1-q`H#dQH%WIm(5fsWSh2~9+7A!-6mOmI}De5{YVIaapYT64@A zQRgJ=n_*v=p^G!~C$DxynHDMp0V+vu@?CJCEoW00{;cSq%RvhhQo#xOR8IRTa*fszS5u^JV2Rlj?8#lmq6s}H ze%!~ednihan1jeqRaM?D;#=4C{ z*A^b7bM2tYZ2Zo1%ATt z%dQk!s!wICQyB&go9+~S;+g&CpvKNMF?$DBZ^809h3<{*gKDbVLVO(LLGBTy5OJOU zqf~HFoR;xJnvuv)z#;=Bt3SPxjwzqnayqdCvM|q+TqPpB0hg?DYlula5GuzLA)u2I(ZA0dVO)%1aps>r%*+4jHFQLz?4RU5b)CY*A*OCOVR(9hO zZESCsLR}2@&!s$GE!oj5g%SYOT8o%Lk&L6X2ySy7Zo~`(;Y8#0W;T(tGE4Gc2#lh+ zeh_=Z9#r({P+eMKOYE*Uew-r)>@_KVQ>MqJ|f$$wEOSTjx@Q;P(`<{c6NG;rGElblQTzAqDA2tDijwjBqi9+ zMnT{%#Y8MXaE!vk8Iv3)Cz)`S=@Y3vMmu{Une>M+g%i;mRh3A>22jefQ{;#J$#|UO zW_J1!#06OHIFm_#%}6%0f=EWF(d131yV z=k`3quR~Tf^8*5ruRI;U!bE6fr`qS#KB$JbDCbLvmfy9AvdW0QC93b5COywZC=Ae>v3l zPP9CvTC_5OfY{1xG*Sg*mwgvz1zAkq_UZEv|8vX(mxtyb@t83u4}7&2Mk2z6~2rxb2umJe5gc(u(Hia!fD zXZklFGUF_x1i$1zs|>Jyh9vYGR=+$9>%z@!){~EUA5W$OkU0oe@SsZsj8CiRySnOC(8ox2*8k)(3+PhtLuS>t=vd_D; zkxaB5t%n1^3`mv2fSd`BM2;dWJt%=E%^_Vh>cv0-%F6-W%@W<-tk2E5+=jtl$&fSR zdr3)5*77nHTdbg(69vuyfH=4K_JGTJ2~T9!*>kL9Z`#I#QKENMCs9Q{fUPo(?m>Ah zH>yOnLse_3Jk?6T=XQdcD||5x=Lyoj3x?-j27|NKY`rjWEg@X^LnJb!M&YA`}%ML4DvQSDEY@oLd2- zu$U8o6emSYhPC@bxzMjh2r>7zv#8)Vm#{2R>y2i3gTZ|}J?wqh#~Yy8xAI(6H;U|# zG=jxnP648tuyq5m661wZ6hPOLWnSn|fgwFx?qgOr4x={^l($)7@22dBrW|t*T^eOF zWv4qmdWz8(OMqBG5?Lgu_M%T-^5&1!T9dj!RpdI*yEu3NU={?V$O@0QLo2}e?LyN? zI}@8?GAsCEVNGlc?Fc1&xGcomikQJ7Rq)Pr(=17YO*Cl}utw1CflA$C4Mo^1#0c%;s zzvP-PI{T6{|K+I7y4c+62xQBjRi}9{#E#^D%o2dD*2)|^${lSMS-ZmwH{0ych1ly( zgXJ&qD}o*kK=Cs{Fr1A(=N6>^AmpX+s$ewW4x1bu&K5P@N2VMX)k%wLX(|bTY*iuL zUvWdz&2$iUDH29^T&rwXu2sgW3mYM~LxG5wqyEf?F-{pEd@c+Sm}B=*b#5r08Ye>^ zsZ4CuS{Ld-UM)pmRFK!THT#j8xvu6vTr;1lxsMCfYRPpqcXQ2@O5UDI6xk`}4fz{` zg;g9tHJb_Q3scKr*XT5~9wh+wwE1O#pxo}&Fnp}yBrNI#gS;YYI#^Js8GTnaE&0!L z8(ev$FBS`H5f{tTs|B|Y1RLrnyAi1kJ<;uWeNwEu0sOIh$z-Z>x82;(PRe%|PLCeo zHe_OEfdXnNA_4 zAo_lsemb^y$W!v9F@H@Ocg5z;*xV(7;4`tki>iuJHfplk-x<3*V)RhG^(fv74Ijbq z*|0U8)q552dr&x()r%EH;I)uvg&`1(L!*YX;Q4C}2XTl&p|4cfL|$?dal(3t_Qcjn z=;mOZ8G-)a8v9$~k>af#+>TLC30b_cxI@%tR-iC;gK2F?5T837DqNasZ|@QdW?cB2X#hA_G9z5))_;m z`!mex$n0eS}fecI%M$GrwLEL2Tv)MJ` zrQuxTgC<{VYE;=;Ys?yRI)N)UfFnNpP13q88M;03wSZB``*M2w9 z_(DU$4`zAob$&PVI{Vb@##(4vyI*)SRk8=`>*tO8+jj7WAXNf%bGbG`qHl#% z1Wy=&vCMVbC2x(7&2l#vbKR1Bdg(X_eh-C$j;iih+ArCzo;SycPdZo?VyA(#OoXUE za?(XM8=_Du(jne_TYkv2(cmnlUv}(?ku$|kO>_F11B0X#Wz%4s`aaS(UK{h{X6*D) zH{zpJ80_w8bVm2G-N*I-+n3oMv7OOlT>C{O)WrRRIWgwmmreA;W@R&!O+wq2&l0b~ zZaLtlp^r`r@BmP5GtyH>{r~QG^C&87@C7I>L~NCL)LfX*$R|zoWOoHLpG3dO&kFt) z--Zrf>Mvi^_fi$=h+iQS`;hm!7~7Ei+eG_23q=*vA2IXD3Z+qg4? z7s+(XUq>fqDbri;>73pi>>PJvX1qSuIfpBA+m*bgwhCV^a|-7UVEgKJaUDS9dT8rD8a9t==plX{3m)%4ZB^t*n=f>;uo}Q?%$pppKq~xD!&wv{ zS!iZqnVJl@=@VupJA!=Z87*RzRUw>(7gdZ((!lXjsvLeF=SIQUAnjCw#*sv;W$1$U zppgi2I09tV{+e8ONb<{Z{*laFhs8ng zUwVIHu1!iGPD)C+_@o@)lT2|DjZoPhVM$_l2;buNH}KkzfZ>24=kpK{T_^fAFwkTk zpW|Yn{*XWg)SU)i5;jEi3ymi}E<%zC#MLDm}K69=x-FHl-j;p5tZ+HV8$E$(Te^1#7<79Z_Jc zsm|E-q_xh@R;61-U8u?uex<5Ta8kwl^!N1bHUKpp92)Kz+F_J8jFl)ilFiEJQKoHT z-gFtKJ1}RI1_z?YgX?3`yM)oD9=+gm?B0wl)gHW7S*tNZ83!WxJ%z;$vs?$WbE5 z`UWL*FN@a4rSI_&0L$L*9W^xngF#4tU_IVz1uA2>W}pZ1jV}AKE6p@R$}0xgJSEaV zw_d8Gct70V`IhugE+;+KQarIMxb2+X8_2*o%YE=hjtO<(7!J)bS_;QS$UQ=2IIai< zWQ-#>7~rf^@K^$0AHd)V^XLZ@Yj$vmrcM6~-m(pf@NN zcp1a8-fq7RVqwcvXvDs9h4Gggf1Z0w7LF*rSmt|lN1}`^$Op5_Up0k}o2Lm3G~b4* zGtAU^V5K9bpF6&(^G|vA17Ciah4cbW?IE&(xts>oP}A^NMAQT(LIAlx6&UJ}QqcpK>-W<_wh7)^F946VlPUBeEj|MJ({&0p_?%fp zf5t1sfviR>v7$i1bYpt-hSY3KpK#%$9`=i;92aNy3uk78{elhTSwI@FmYkEHeJDc- zMU96+sk1rm4BX^h&;_UxqapGS>*tgB2hI`Beu6!oeCgxooDd&>P>9yB)f-H7J8>Ju z@*irq`7Smp`@LH7FIj#s#8ZYNm?sZn6~P?OmdsN^6D;SoQ5-ji+^} zNK5?J+`pV74Slc#(|HK3%e`d%A8h;zQBPRXTc8anO5v*mYX_JE%SqPh*9jFKa`r)I zo2|Q?()|;6tFt#a|Di_q$BO$?#s9HlUaq*mR?MF&$p;(e`iA@GhQCqw#Gg=6ng=TO zs}=VM%QV_=R7vKb#?{5fe3GR(W6Q!9*4`BN%xj#=aQ~M`|8M==ym6mtwCqePV8ciCz%o&I<6<{=W`A9ziB$BNRbbAPn^ zH>9Nim;}{q22427+335Sgp^HKA`x+&?Mrl=ZFaFKKAE63*$z+u99sdbI*$b!;|6OK zb;X#$En@kc{DVCAVC*!2R+wEZ*zVV1Cx?^*Il%aS(KCo!43%K^)|$PuX77^q z^CQ^Mpk^qz&=r9&T=F%H*i4jHsvh7TBmp8M88?Xq>mKT=PgtS(-EUi3_vSS)Pb>86rjh2n}M*4h_+{<2oO08hntb30G2 z@M&$pkU|ZsWHyy_hoLphu6$yOjSBumr5aLKn!QA^!-o{%1YHWZ`OHU1YH|JjYYwp8F~pN)#=R4k)QMPmHu7O99>+YY<2eY%`IL#{`R#v*_#98ox&iYS{(w7M>y0VyOY0Vi=A%OU(k!MA-qGOGYL z$#9rAEsz2^0n}qG4oIASFn1U1bg}#3x)&Oiujz=nh_K zS()kkWO~VYzi>eG7)~SSQtp?FPEd+lUuf3r!!vd)o&kq)Q;z6r$4fSO(fXIH*}J*U z+4at>bG_@GU58HfGaSCD&5N!ehu!?E5cR&? zJecG2%D=-m`*=Q#jEFcb7+J=qF=UX)d0-S-Ra7t^yLm-QftVF|6rO0)wSMUuZ`S(V z!v3hRPs9F8E_{ryp~BfIrEBa621uglZm^AO>|o8@V6z)7dmGN#Nn_+dFjE3rH`u+x zx!R33yTM*z4yhb^33m;?5BvASps}IK3{-zkI)OAco{kZWe3PS|3slehg>cYt4PU6P zJ!{QJympK2~E}~sthw42sa{o@_`37 zDjCYxvoz}sq=}SK^iSNe<`W&+4ISKbnW5?>2uOiN+C*SCCI(zirABd@%cq-#=qP&@ z-;bnfopt;0gMEpPIm#YgMSbD5VdLS6d!%muS_`3i$H%Yd1y!?OVuqTt3SH@z zGD=s9J#o+U3Ns=MwH1zMg=5L<6l>%-wl^_q+*J6QFg|QP-VTZeX_jd!wQsJRR@}82$nxu`ehL>XW_>M-5(xesgLESHq5Y1?n|tXdNGB=n~tXl9ihfcwzg3o6^o$W6fH-bPamNDQ86=Qk`^^ z480S109+wqR(KmM6Cx>MHuCH>lcXy+lbt;{bW#E)0yl0RXl)Z;D_41#I|=08=q5He zQZ#E}{|iEaD$yxR1Dd1(HOc{*u=8Wt2iZU(_MUw>yClRuk5ADbxrE zr%qx_SPyXWIcGi#yqYY7t$`gVEW#&N7<#F&hj7plo~Rdc^HUv}o%s{L`*{G{st zy=uPd%nu!91^;#HUvKrze5}KM6gkaM$@@8CMQr>W^NW@(O*-4s^@Fc5ULdL#Ma#Jyqn@T`cWqKZTsD!?HeF&chwmAIM8)4jUc^Xr3PCd%wUBJw!AREY-Mn?g z$DFe0tgbWnkDlN`gpUD{@UNuvUrpEJDo+MPNt_CD0CNb#&(axfeRwx0((5=07EFVR z_W?=NEQH2$U6h;}bkQQxq4$qYSr}bO5{d{7dKJaN!P3+7Gf+noX?=cyluO1195C#AsX(BaaE*ftlLzU}MBkiKX#_CVa$GPGQZg^PBE>_;;2_FstRA zxO7NE{-Yz0>SJmpiX7 zvqu}fx?Jst#@>?p+qLfO=W=_QHJ94*XLEBW9F8?o18gwuVD*k9D?rhCF8n92iLzdo zCo_rD^eiH9OL1mxp6!I?MKiuYXu|z)Qbs+{Euyfzm3>N|*eAmGHicu5fWHco^=!KX zSd1+Rd4If2bz|}F*!`T*haLt4N$vk->^Om^ekxg7DV9qfScq~0TaD2alp9O339qyNS}VysTuY3{#Ux;^ zWJ*&-Zdu?O_8u7>u0TD75`i)#$Y90?_lV0A0#%haueTLNZv!HlfQf=Mx9ZQ2Sd{MI z=88QR zJ+=9l_nc@0euCUrAiIe6dwx#l3C>wnt|NR%M_m`~gs~NCb~uqNpXFk&-a;4QxaA;< zK)~ifV+3s{(IMJAor9n$%~`S->6=vNHjX6{}Ab?Fl^C!bfpg z;9r!^sDYof4MV{XQt81>iw#%qmv?ZnCCIes@g6V^&HDwX`{6`9xCT)nf@pOKIzq)*Dr1-0h(vaF zCU~FmFU5!>ry+I@y<+A*f|XIsRD+)t%cu2S zfm1}_r|qK8?BzD&yhmi#suvg5VFqYNj;aEXBpep;#)jmcUy3AgQz^2N5)gpWSmW| zOkmPs2XUDbS%x)bXDG=AdOHL}nLwB#Q8DqqX!0FFiDGC8_l3`A5#gW;;dpF&+j(&r zLyn)+A9g8Q$%5Bq|CGhAqAq&h3@Id1TZFqt7+5WuN2Oo2=Pm z-0kE%xKav#aXkVF9s#oha~8tRq4H>|DUl}7Vd$X;b#%`fPbcJUq!0I5XLY`-^=m!< zGBNmoM@pyNFj#Yh@gidTWmfGNWur@jb?7^lo=*60jK_bi>j{ZNR}lQBNj~bmQJZY6 z;yia{B!Fi<;EA|QCsA57rx8X`b=P&JI3!eq zC)=eAMsRiE6C`i1JldB2zAe13&7)FBb~aBYkN}_sa8vR=s^c2gRKF-+_lvSqP6n^E z;Fb1dt5`fp`0=i|z}a7|{sSh0^@$7(UW4K~#;|WC0cp{&VtJ~L;iQT8yTNFyGDA54 zYBrktz{JhyF;ykGO0;3Sl666MeW6;bF&3+bV(y! zcd&k}N=wq}=l?8fBL2r+fZ7mR3IysFSdwTWWwUGVl!_ann4Z85u>|6^@TdcWvRZUJe(ZIg?9`)S`Mf7enFgnRPq=|;e0z;-x{96J7p4D-C6)(v@uXpe8;05S!=f98 zMb{5AS{w)o3q*a7$G4jd6cnShE^R`k8K;xexj?J7zwl*(BfU$;a8^&~hUPS%GwjsB zg2ItVvo1o?0`o;GJyao0OtogOVpH>U9T^#{WMvhu(*cVFYtfm>PAs$r?vF`b?$Qm> zH`n8<9(~ofjlP}i+h{%7xdHq1konki$oZ5Z=iR}E9{d_WE3AH51);(+Y$=Go@XSA$ zh=~N5@ih+XVc`$|;QLR$N8eDGnQFt$sRfkBS-0J1YI8?r}BG#XF+X(@LQe*rVKj71k> zlvv5sdFS;428ADiT-I5^_bpnBr|+LJNM=o6WVIF%bj0^bxI3x=Z|h`*rN736a?SvGMs`JD4-# zR^9LMdj%Zr!MIVMO)N7xIwTynd%Q)9?(huo>w-W-mU zLs2Bv{W=-uN(BNXXek!|kWT?W*C9_L||?4jWN1znmoxWZ=;q2|DCMgqJCmSSr|HQJ65Vo`MV( z$U?;05tbJ1zdhrPI2wuP7_c$E3x$hac##_$Y#u>J))(SC3(?-`XWhQrINEi`%+lZKlpPjnM zw!?TfU5hFzqnS}^r8ap?fN*y72nYXf^Ijm@NbMJkhXT3*s;>p7jg-p(yiI%dM4$U( zpZh~!_(&gfC)7_0A6&1I?>;uH$$qv*=Y#8uMtwB6X$%h50bGET^ED4g=b&(D3a%NM zoKQ$kEZCw#adE-U%bXk&=T(2$v@EkTm9(N5`Ln{=nVpq2^EcAay3E#QBQI1=!yH!V z7xOHa>!fDY9nCqcG$R1OTit(a?UXeHZU%-)O9#g)n2C^8Sm0{Xq>Mv!Pzjh3!hjd* z>T;K0W~7hTpb%jP^R!2>d8Aj#;yGF~SmTKd%F7rzU|A347jVDxRJ@ z@8U^SKjQ}k?=si!rz6}=BiOL`2r;h=q(S+sG~Wn?Aq1yD##+BKzRG*xSKFiAL#`lo z*#RDs^PB~{uSaGburfA|F9kP@0$6VuncY4T8bD&xWkox)7%nZ^l46N@u0Okoi%I7? z64!WgjeNjJ5M}&KO1loYSEEj%iIkIzF_Bc0EKPGHj}X7j#7S8|8#KyL>*dpIsU9+w`2;YPuNr!q;j^J~%1 z#@Wr|+|R~^oBDwe7v37%&*H)zapC78e*ghLK*pwRP6;X#$#2*Wl0Pf_oZrIKEb@eH zjBmXkdNSL?;EuJe4e{RtHfpS!Qt%r*qMnEZZK;1s(yNpD4$uR_SY0qhfx6k^`k?3w z^8oamq~|RTz1pE9KU*~hH~v#oI}P;=@G3h>GY>2aUU^)D9qRUD2$zF{qv(y2^`foq zT!!#vKPkB@N_Kgv^_r5~$9<6n1+A2#>q_q0l3j@@#;z^3R>C_=_KQ;9w*0Cz9yqn+Fmk4~Ja8l^p;sb&F4gOaP{wlN6GZw^~D)17GKZR&9?B=H8E z`okAQ{s@;*VidescnIhlxa*XI3|kSA82uwfa+Obsqtz)tgpQVqpRi3fWur8>c|>A| za$pVEb{7kaeLgsUYz>Q{bnGeFgQLP{NAV)b)#B&rWn2x1v$AoGsrgUV4?ve&BsE^} z@NW^bA_P4W&0tSDO>9hPpJ$~!Khor}zFyhmU%5l@A+Ka_XcMm=9o{{<0VRH6@bTgf zd3F-4fdJ!HQ`dahy8;!1MGN7X1^0Ah?^o=-O1L>GX>YGc#~tH(E*qMmSi@Bo;~eo}8&)u%tHPcExZ zgU83&@7a%yg6>WBuYR%6vs2Z5T>4nSo-Bk<6vC&Y@K;rPM#Ie>sJi>B;eA#6t=jOn z)yf0akz}h-9J%koc)Ff=2V$@H6_BxwVW&FDK;HoN5-)>1tSsPj@kjJD5E_mQME_+u z@JLGr+Df=fuxvbCe!bq_s!w09PhO$RYsTBTO>R|_y1dTjdYnDex7#SDdAdo9DAC5E z?K=b;eJ8khyrME|$f(m@VfgtMuCmfkVI{ySS;Upgt#FNfgo+lAW#8f+Df6rl4g~;& zfB;V1_CH+%&m8TH_6tm8BhQ+ZLsSC@XLrV4v5H^~mM*#ughMSck8q^3`v~ED2UNSs zEy_jmTin^ zt7Rtc+{w|9gkj<4N%q8q@W}~qr%Y?;(w~lNy>46|RG+^fs1%b++->7pH;n6v8iu#@ zwc>`DzB#t_{jsfmy!x@R4V%ZdPOpk-^b;zQ0=7p3h3N{?d&c`o809_?ikIT#y0|y7LF=M z@(My5G-c^zC=YKD^%$0sUZ>4YooF{rw7KE6l8edC@vaJsk?`ddXV;CZE8<2nc+a(C z?8DLamr3^SB--SCDm3sQ&(}vcklMg!XR9>A+(No1MOUyvSR{wT_oWl6Sd6*wvvN#DxHweN3;(D z!IfOjk}YLcpdR@fTqPAh(BN#!YxnQNn;oEYyM$eh%D8_%jTuMT~11hQ$<%( z>8OwDrlG5o?4it{pUncTTqTldV4-31pEKGrV=O{dKavKUsUzHrBgEi7ifS zGrz5OFV;7_RBwxtkSkkoVQo0wKYb2X>$U#unaS$KGm}xlUos{A@f2I$UVMVm(8;>V zNF-X6e&+B&0q!pBHZUHZT5be;t;(Lm>mL56YP(VMxX7o9i8Ljhsz;tOoZFm|cY#_t zhI2H*OM*cq!GJHsl4GHjHC8=@Y<><0OF z6Ssm--kRbiW*!V0M^ud!GOsrgJOtnF(oJ;-Y^+n%d*^ol3vjK7FZMcpjs(Nk)1N)wn4T%8L%VHd7a z0X*7YvbHi?43Jof`9dw8RQrNcW+sbfxWzNoPnu>^w%{Tk~Ts!(~auA z5#Uy;IB2C_#-yQq>HF9t8yHt5ST!>_XNFrj!!G#fvCDaE^y{hN15?91|BqLE%RfUa zi6@un_{SgJKoqgkXrfZ`QJm!uEv7!ghnD^g&F0NB;m&TGVW%}l*YvSS>vfvkq*JD! zrwxX4YNqf55K4gSVHXPh!TGTT85Fb5y~6EyzwxpS)*(lw#}QuIRf5Mz)24{h?*tZD`;?{iMgqJQgG*- z!lE7g`#$vV+hj+1ZnWOF72m>{dhxl?UPQWBup$%c?5J-A>qob z4h{C3$eWlT+Q4&N9g&4NUcl$`ow$b~J(ROT;`%)J3;)JgxV>q%Kbixcos^qqJEiet zG23y`4E{K$a9-Z)vv=^blKXIHxg8avF0^V#5Eeww<_GKM;u1Eu z@I>CZM||ft;0e24vP{vph4t|H>Q^Jt3!YLGPa(v5t}Iq|C}NrPb8I225$=3qK4xN673 zoB1oZ)L!`o|1!)2;>-22oq`Q|**SB0*(r0GDX=ame`~ByG9Wl~ z4ThZ5lP^xRHzy`lz-JJXXjM#(>WxRXDeOqFsAEVz9*r$n#c9*DqnM@*kehzUX)m>E4@&V+mjAPB_7 zj9JD=vK<<$9vrPRP*jurXxNeAWkAu63`x!qVk>sTP$Smk_*~GRxKUb};XYvnfl;|~kx9p<;h*@xwXs^>VpIYSMVN_PTqh1Hf|(+{&J=Gjy8()$lr zLaC?b*<J?y(VVsfn}4XZC{h!$ zSGiWRI0F*Q#J)!0>Aa>#m=QkRqx0;ZdG5Y> zz#w*Sal|;A)iLLKK~22cHb;2k=~ew0Z!sauMV-t(+(#cdMg|LyDBP1z&IUg@`%hqV zxcvx6Jp$I8@{-L(vQH@qfPR2Q5QZ*?Jh}J)L}ic{;HLqiG!MrRL$c-wZ5AF+RP1je zGBHQ@K`I$C^(g&gP$M`}cFfR^bLQ z#)K`$C?y(jUvcO+WZGdXGxNL`&~J+BYPN-c@}$s9&(LxfLo{RtuuINX*&hg(WXcjZ zFL?KeD7f%Qs5veoFf%nWJv?`=owCfiGo3BN_fc(hrVEij_e$?_;oWkxC352ncC}Tm zv8JoduCc<^c*ImyS1TX=JgUJHkDGggtkJ`8J}(hwBj1fu&a*8D4N?@rKau z4eg%LUOOC3!0U(Oeg4+r_Uhq<*A5Ry)mXKO?0|o!%PeIXi_s{?Mb0NNXj3$N{?|k@ ztt?7dgzsa!7>r&MTxJ0WK$mWcrfrODQ}memkI1(pH+r{(@hDd6xKr+n+8D0-J-oiB za(pv*^2p4!XCA&^IwISA1hH-|M3rP~BX{GWcGIB|Ej#B|huJRHu4&Xf}t(5JM0O_R2(&qVf|#O({)FQlI0qk>If&T-pE+q$ym369KfgV?vDhn%x9K+&N*@SglpvO7gJVkMBpuIE1yO!tzR2jWfN*eodw z#?H~F?O(`6RjlC2MJWj{`Mb_i)*L4a8$u5wGqMUH5Cub-WiMl?_s}}8*@rdrpKK|~ zddX-$5q-T|O*`(|9xXPzA0XIKb3IQ#7>4(-+o1{biBPIe8NkoJps&ibu^wS;LL2T+KunB2mF4 z?AbxQgAkwMOvm`)h4yu6JJXq46B|eizOD0ZGUQ5WYa(}%z1VMag13%B&c`Lk?_LPC z${ykbvEr!g?4z;@(V6)DcxRFvc12_V6+{QX)nn^Wy@*}(eK|TBH&GrlCwORnHYJ{p z4v(&E$Zm>#?aOxG_OH(*ZY;7{$qsH-yhAw4cIa+tY_*o8l^Q`nUdH(tj$-TeM$KZ5 z(9PYVz^h$@$LD9bRiYta8e>6$sFP<8m>g|x)^+ediLwOVawi0nQm6WlE0MdX5MELs zfWt1svQAiK6$ebbfr<{P5EGcG-OU^LN8{ze@{c`QW=k z49r9vKX#}!h~wevmP763Ls`Hrhej_S$_@NoEKnX`zO|&c&$nCV6MX36ofTWkU34p3 zGoNpZ=g$wvz<2y&zilj@^a=h2l3o~4?x=Ulex-?7 zoZ~H9dA!Bv#>E+>(dx$~g>X!1EUrOgZH(Kqrj^v@mF2v_GpNn-1@{O6D`F&BC3A?O zkl8+9Rjx&%(PCSrC_>4^T|%rwy%|UN{U*HwD+g%mj?Wp~vIDzxpVDVZi95S)x5lJlQu7`S9nyv>H^>LDEc^Hg zGaCW-uR4r8R7p9AM~2TQX9$X5*N|&RNz6id50{e0xZ*nTVk|b)HP52#YpI=JB%d-1 zg#{&Na3JCU%(}{yuxjYl>>8Geh7nO;I8qBl_eafOGSji)81qLorjf!os4e|boJh@V zj=ss{L713Km`#){{;5-1`OV1Sv}1DQmdCz80McOjF))b_7ve_*4fOtT!qAwp3;V+# z_q#Ry6!yBHAMLb@&+kX|QB2S4hd)@JpP$uV>Z&o1vLeWuz|2Q2;#}mXK;<~_TxTBL zSPb_C!$Kl}OmxNyU}90@a#7l>v_M{0eO7!vE1R6Fp(OSe;AzFr>Y)6xx_j!;O{X1A z>u9&mcMFe}PXS;(=svA?2oMIjeR=sqS^H2AWx43v4yTXtvY?PUx}&2rxc!*I)*pFp z|J@(K>=5R0I>FpT>J`>|P4&JW4sF74FUZ7|TPllun4fANFM=UP5?<)!6Ji)!k6qX$2hq~~n zT(Fs*E8k=-lN2#H<9)%|$1=>9j&9m=G_9k3;%IlnvG#i_JZknZ`f6{1%7-;{(uZp( zLBRIFSBZO4O6(4CpC(ApPYd`V;HFa!KaEUtg z*Q1+1IGWaIUp?Br1sEVP$)H9F4qmN!{-jVdIR^#xs^sKvf5owdSMxVN@89h80#kP> z1RIVM%ENuHT0+?jh2@AwF_2ks30_C}|BCB6|6?kzM^qmhC zgD$EVv=R$N*pkGlOz_+Z60Bx=$6yT+SDyD&t~HWG%hy`{)^aPz!_Of_%Rz*G4yn?m zJwehv^ds|(9fa5IqG>B*je)fwF!m~x<}P{q7~4W; z?d44@x*R-8CGPD{=N+kAs7DK^qPzArX;{#roi#;~BO6megGL4VdE- z$!Ws?udLERsdMa}W8pEA@XllDdqYiH_~ceradM?{P(&%nR)W|l+0C&2u%+eljEE3^ zN_(V@mB=IZ>NWi^Q1!vG1rQphYLL#t#(1e5xmu=km_IBE`OWf`ZwuD(0|^lqL6|Q& z-l88LZ+0038sbMSE4_x82Z%4H`Z3fTp9UDv=rr3J+CbFc+hi+GAFy)?GbdSWdvNmt zIA>%SpZSq(Bm*GA__(AmY!x3G+jGZ-uN{Z?a~D9Rg%w;83eGE>`=Uo=Q*>3#x3oOl z$y1&Osk%p{!f(#iD&_kn*KbVmy9Pz5t+U*fjw>E6=fyuHl<6G?ik!W`V&=$_;l!uZ z|CIs1b-Z=gV@r1pe>q<8(O5(FGS{*zNVMId#ryl~?h&Re--_3Iq^#Z1Pak`!WP-n4t+3q{97Se*@Vn!S_dh6V1!E#1A1SYwGw=vNj^P;smDDwjOUy1N5Iy zr_VR}{E#J6Nf~$wV1%MDHO=xTL90*06Zm5h&B0zY4sF&D8dpQNQmQuwSDj#mH;?l) zU<+k_?mU4R^coC={4?{^<^?a_$G&a%ZOg{UM+_6}7jrZo;tnkxQaSW{^P(RXIS3s3 z4I??}*zlBN3m1~@d;L8)R+YIk8l(LpD=@X92HWL zD{JjvDHjqLKRlBZ0o*qDFSD!b(`)3De$ojREk1#VP=%S)=ibhGGS^Pw-RP@Wjy|fi z%;k?xFh*2pE#gm%sxx!2u{r}&jY>y@$*8IqlS3XBxFzXUgjyX0xc*Hvn%kiHp8;P? z+d?wF1WmV_#(AClDg(irLeEo+(5cW?)@(%7n-7(5z66Ka!Jd~2j;Isfh_erNTFw+x zNc33vAe%!bQd!-3u1sryp;k_+R#g%Fcd`}`IbSHEaADp>TfXM^2lB6sv)>P{TJliXOL4zZhpE~VH5dQ&^A_u#|>_j*NjKLP1b|>mM`>L4rV~1 zN+?OGT2vHLfD&b?kbt%w8Y`xdDG)nV);-_Jzx?VAR**adax|i*6mg{9hBIpdBUql%= z2%exN3acn~yc52F5p9*TKPAeE_4f1vt8Q9g?oZsCgikJD+0!RW?uGf`89%U7f53B}pKmLU&R|$CIL2Lk3{wj? z>Yj$4#)O;LoBSR^7J*U)K6n70F3Iu~a#XRXQSo*9aXYmxm>9ND=L!hCXbJcWKEfHAh7k;zCHWSxw3S z{n|yiPuTg3c&A&?9iGr_Cw9BzyH(^)fni*JyDR+PF85BC``<3&R_5Y@`)pVEOqYAU zE4-)6UgE7r9)Er=S1ZJ`9^k$u`f_wIzOYgklIoCq+G;(M=YM2vxVmvOo0dK`QlLx` zXpMhfWbwskSbS=^oF_R`hkZhvZ43$pls$o$ldR7+%69QdjD5Vmj6E*G5r=D&*-v>i z%?`qUeoJ94<0|456NAew?MUr9xu&9}vaV=6cTNb$lUKXT^`RQvK527fNQjYEM+In1Z#?$02EZ5f^%pz3eE_jOQ1(m3fu%KddnRM`_O^*v{>I% zSH$wI`iULQbxosNH%5BB2c&`O7o+Jy7G&Ji^w3a(gRIekui_g0`ZamT4 z20mc}kyiy;A7K%~O9e)CVKuAs^xpm!6+Vn;mO{Z&4H+ z7{9JFyuQ=j)akD6tkjh%4bwcm%~x48SLz3FFteGs&-=BUufJ40s{uA9!*ZsAbFCoR ztS~~kD`h<=bvJqj3KeM&G5udA0*tVq{6RIO-Y1!uxt|)mb&?fs1V3sEUB%WC@8E94 zJzTce=2}V>X@Wb*gdY{ip5SflVew=e!lr}jz(8Wr@h^jaYzcm~82(WM;9ny)!)BY| zu06?`uRF=~$-cjgR@b}jO*UO-YrwITQQ^wAYwkieqG!RrBD0@xT*GlWzqpm} znn47+hC9H^0nre0ei;JaA+E$^pqN7asG=j6I(l5M>mX3!)h|S^eF29ipq(_snhst7 z{S5#PE|Xq;z|-?o*kFZ`mxI|YT_fvcI3Xg7j#!VX{{!sZOV;RwUeuq&s-za>R) z@z4s&k|~C>TtS(HhRd5Ay;7$lPoG`qAnQdw?xGkxda{r%bM9U(xYtOEldHekIgo95 z2I2s{l0nE@^c>j<^U8=Ck@X6N47C{d;LMA3D$4y;PqWkY;!;D`IxVr){kjS>&xUVYeW&$iu)o7`*1%U23uTc&Z+wTfF!heFX%Z6p_HpA)!e~e z-+YCw&ul|B#kVW-KMMYjl5fw>FXdY4p>5*7KE}7{;19%U3$X~hvE`u5q(Q2Qt=b>D zru}iVHjc=5%EhpDgUm^S{@N`h&}saGjiR z6$mGc=_%Io;mM-h{@eymQFZK7Ptk_EgzINJ`x$yfvTjYaA1~IndzWqK?D)lk;tx6h z%SjfV&nI`@v&8H=9-y0VJjKE@c+Bzq&AQmvL2V>~5ppLbI+a1(R{+93_irRq@+%Wg zX2lS`@fTP61C@&|+Jty26>$E8U$c^Sv>ngRI*G z^+@~EDc1ZXFMq<1v6gDIz}c<1jQif$Z6{m#W~S{%jwJpgZ4(oH{}QWz zln4HvL!X+b-z>oaL++Jq8q71%)`PQw|0fQXql-dkY0{GJ6o1Y}8I^+Uvr3I&$MeF) zWXU^;->ZXfGQ1DyY+_se(!8~pjgNH%y-`eO{W(=?Is6u2Uw}(-GknsWQIqH)Zayi}_iu10X^~kJ|S#Ffy7X4?L0! zI^xBU`;^62&k@!VZKT)kD6S2@hcH;rgN)L1gJ5#>W$08Bnv@Rstd?jt^t{zN;+e&6 zrN_pbyXu-7W;J&cVUV_USG`6QS8WMCn|$6l7=X&N8**#Yw2XlnC;x!0#SuA zB&)514e9F4I${bRji08(&Az}Yx4RLAkDXq0BCS-;d^|cj)V&n)^_1*N8t7h3^Y)3xuFh|ZdZ~)|7*v#8b)&o7z1k@Qs^4pl`AUUaKQFr*rC-{%)DEvxzo)qKnux`a*1Cw{c zDf#4mpviLtiu7)DM1|D(cG5;{r32RaiiKX1DZFf%_o#0V7M#Y~DvZni0zP_|12t6K zRR3qrT*BXo96vc(;P`Q0`KwO0`i9!BkVWUK+zx0p5NlXMct@ixWvJ>VlE)09Gt>bc z-4La+^TODLwdC@Q+kHFFw=KT?Z{OzMHv99uP1X`8gnkd^web}RX0-UPWsNiLcJW=# z?soNi5+E>OcrL_;ln-sj8GTN-yUay*k^cU&(}ddphXtp;)=~^TqI|G~;#Gw+7E)oa zp$OJsFt~x;%fuEajN+m{e<^+UOYDbhPqX;C)6DoxpXDL!;mtVL9>Zh*`la;4FPTDN zsv$aw-l)Boy^Ptq>sD?n6yp+pzYBWA;K|RqEmkLGx zrWmn!*izWUrrwCWFP20R9gVQfS=@^nqFjyT5CvUC9bBK*{SbEq3d1y6Ytm9_FSpHU z$0w5N%MQQf?bgdn)G1i?(KA34;tTyOgvfEhY*nHZiR!43kFWN48#<{ah0(>V&;UHV!Q_tNmE&@vP>%Dd#oK39_x+)s4qIr zqKjEieZJ326FVq8vU?tUH~W(!036|3b><%^88KY{{Z(7W(|*F!{-I*h&gR>n{-572 z=i96Q&u=?IAkP7LV#)r&n>*|VzPtW3M)7}qe<9!X*!utbj_5w#z-%-5Gu*Kz{y6dV|A5}}xMQ9tlh8s8x83m=EZ+25 zTLhw`mB;d}zm=Lud}75Yaqq84K9QOi2>-;D4(tBJ70@|(??-r#W(j}!FIy-%GS+J& zKDyQg+)e((H9U&?XMS9{>~Z%0!}Wx6ftQv1&=SYbWyLJ*h}Y5f<8ox zdo?zA+n=3j&9{;n`j#`zZsxe0<0`F<{nU?=T&t&BsmarBR|_N~Hf0;1saO?Rlj^i` z(%a7Bxgen2@_zdivogTV3>KWla?AOPqOEeXuD}>5 z*BX9F(L1(YNs_|#k?{COeuqdl)jsbAEB+JvE6ZNtkWO|ShyI?+_*;YX&w|GQD1YB$ zLtPNhUxpCTBR9mAoITv*9_gVqS_T0wN4$_FEHlTaE#-3wpH=q=)AUd9eSZT8 zGJ6LiG$ES>6rB!OWwsI!>$6dAR_$+gU^D`r@iT7Ey1n?Oj5Q$o=w5KkkiS@T=nl9C zTm`UE212f?gidM4U3!KQZT5F9tV1z)?JQ9N zR2V6Q*}aT8?TrDmmq@L#1q~U-O^T*v{HK937^N(sS_g*NP%hVkK#|@&eHW$-ra|JZ z%FDAg;fhcelNo`7NO8okG#Zib!(AMq7SfKu>JXauZO?|j7kcip$+y{BtZxH$aTKgO zo9a5u2HHtt=ZgH`-;l9^c?4vVwez;|gb?ow->4)Eu6c18E~A(}SK{gym@~xva(EUV zZY}lr6uERuNf@W#zt7yapmOMvNp{%-YW^CDG7kVH_;_Ns+u&bOYt~Ysc&*!e@JfnC z)O}&wU-%liFAx4D&wsbm;9uC5oZEBoufoXm!cul_$$eK>;xF@NDUi@6L{ae}G->YP z)E-I0m&m!EPK3C9R!Y3qD}0~ROG|~h@k6ETe@bZs=xxRjAVsEyy(33hGJzGr#S4A6 zbE^cWg375mL^emm(X)I3lsF=ytWbN>LJFEjwoJPW$CN+jiT?_q1N6S4Zy#!GkXgC~ zrFLNw))q2ll*W*$gqPJ-8?%NyP1yLgGAQQ>!g~C9ZeqrHlfDw8Y2UFM544*Ogdst{DWqaQKQR8=glxk(maad? zvYYw5Mb9{QV!Vo^1m(LA1VkK3LQrRF|7Oxh@=3Cpkz)2R^z=uZiGlK2N?C(aR3s5F zqAnRmL=HouWP~w`vT;VRYRxJTI7*mIJ7D!8w&oBPJqjP}go+sZh*33xj2`uxS-f^u%f~^ zpyf>a;r>VZ4gQw?JI)zQeg%~lzm%06M=OvjY`{)%F-7-jp=1`T+l_Lgp$WqQOVga=j!)TP;FL?mT-Z&%gB$LKEE1ClkuO#s6Y05rNlX?Z2 z6Oo2`I@9JPWWLlKH#LF=)Ap>b6O+RvCi_#$Q!_m2{WxE3{qbl*#K(kaJ&O~%l3X8WV8GZY+BnAr%U41b zJb+()s9i(c!$HyVgYfJ)`5-`hFyBUEXB=crVXxhGP<+!t>5cks5bSR~$fFaHVaTl{ z1}Uesl9NcWO#kjVhyd57 z<`#X+7JmyHOjMo_Pbl2DfAP|O;1sP%WixItImbp#6$S({lH*Ap>P_-$jZb_Wo!!RA z&E_7dd1buijjU4Q{)2<{Bzy_0^2}(@uq$bz3R&p(P)7Vh^jj1+&ONQ>;QL;L#E^A9$2mDJ9WI9_eAFs^3)D`RlQuA`LeZ)qcgfR(SQCTy6OL zIbLlz5Z;`ZjEuHA>PIGntt>IUwdL2+YuTH@bmdAEBpjDi~Kv5@0Wfdf5q7a%dNZ) zR0vWAl2^F!4wB3Ke5KXj%I7US$Ain2e0x2|Rm+(lAm7=I%X8lPCJLpi@+7KtCJtQy z!JWveY>6;Eevu^`7Uu3^Xgoa@%-c_{TwiVFWHcOHIVv1@-g~V7ojf5xEKSgo9=N#< zRPaskNA<2ts?s!iFz3M8?TT8;*AmE}QxUGX&?#24tks0yCEW!F-Sul&hSQ#apE{ms zV5=QHhWa4KP;LGXyqr;u_xiTax4+K6Rw2Mao%!(G6?zY($y~z^#Cq<+ilbT4iL9s$ z(_-`s1=wFX^>D7$P55NaSmFG$B*#it{wp=pTMiYuOR3JY0=a~ERxfJE{27woWLXfj z$it?oxM#3zm8Gu^n0pKUpIKqFuO5}>=Wwqb1)T4}+||4L=iBe*hd0OW#@KF((`zC5 z+4-94bEN-x`dn)s=&{Z7b!G!S?#*FtL9@MqQCp$4uk*RJ!8SufA7hK($MNtYOCRC# z{g{_sgeMXjTdWgtD;iC?^wb+(3pqfp0l4lt+AdNVgd8%j!5;nN6M4H`MvF@W=uD{R zCa(X)U%!g8lfi-m{B|_z;B)v)!avf^*0UBKV{cMOam^|#tX>7P!8xC4?bFt+GPjm% zRd)tw6rsWG?NhdzU$5ji9e`iT=Z#0(4S1{bW@?l9XMX&iGn9Ra`g6|KVzZHaIrROw z(HRj8Fz)->?^no9nS{SvVd;DG?Gc^|d%GcW_bAV7dY2~qDoysK7T*!OKg1;DOMef) z2H1wt9s1^H7Vqs|Uun&6@K!$a?f4X1F*G=sa5HDlo?#PXlHmNYC~H0XrQ_!ubg+ zthj*RY~>Lj)*kb8gWKE?;`?Cdoomr@;3RRJ)vv)yNiV>82((|U#8*YhE2yS&%)iSg@Vl5=;djNw2um=x4pA?c(7C4XzNX)B zO^95sC^Gr!y1a`AG0QlviH;7v#t^+~W>>BwN>tF%;W~r~V66K*9N9>)O|di$^h`br zr0C2Iul_%HG+4f-ouuuP!u}!<)S%)Xi3*;KWs)pN1tnF>O^eM!e3E^|#p9{YO~gsV znNmXM*&Npb*5UM4*a4zB882L zW0vNhJ(udfZC6U&1f%D_MfVuBCDr;0ceE6Gh@|OBhZTFX;qeI9XCuscuA*zPI#$ge*a9lRNJ|fl;o>TNqqu@H2>1bvA2|t4#r*XpQv`m z5D}#+D0T8oN?CnWtH}hz1*6`_!tgs*a|8@{dvMzMR(Nn_PThaA(z_uIR))xiipC{e zX@#I?2N&}T=iXaot?$#-mpNYJ;DLf9DpN$&iideA0J%)(2H{uM%?UY7jzFmGM$V*r zov2QBkigsOpE5hRj2FCI>(aBnOX!flz1G_%Q*2NX;Kd=_haT{2!F}gjde3=y0-UFD zEgJhVP2YWd!KGgu=I-U}gt2bnSXx^M0Fd(AJ}mPLjOpKHW&|;VJoLR zaA_C-v!PZtGtaFOO97_zJn~qPX~Grf1TSjl;XCg-&zkRGH2%!p&Do6%{dNu_ZQM!o z-D&eBR)$fr)F2l@I{$ZTF`H4a#Ye$sOG4_F8aq|!GAd}l8uDLaQ8@@y4krX}@$wtb z%>m^q@Go5+aI?2UN?p~r!+cxxZM@ZozLbFay!%$%KR9Kzy6^1i^Q>tTo%UyLGiMKT zJj!u_kL8RrJ}+Siyc=$y4zgMJjI?9y7Z6&ptKqwTS^j=d-^`b$ANcP z*uEhE8%%5Kxrv)-`lZ`;oGR!fC*eg)t#D^Q=Ztx_Tlti#a63N$q;=4{RAQ6! zj`v3WSPVk^J(_%7w%{}D9o|G35cyM*0ygb4=+(ouWw_5T1Dm6Hts}ASKEBNp>z4d^ zoR_5O_REknWZwyvtx*nS@P!r^#s)g%{`Y)jtMirajWc_jFKQ8 z*WF`Sm>$Ps|L_v4;ww^)8lt+(zmr}0otPX_wM7uQ6q%K_c;9>>N(4u_PvNAivdgm z9`#AGCkFi-WqE8mWcu+q^2|lOpJh}$gRBOmVvfK;Ns?G0JimJ2z=o5^Ai_g{8f&hu zfoFCBtjsMF(j2y7%lW=1a{F(T`6=;j=359}#L(NJ6=99&iEiv2tXYfexcwF-MRfTg z_C5pFlD)Uun%?33&m8)Efny5?Z$b^^%MOu&2kmRA1ZC58MWGH-lsr4J^r@r()^Ai{ zstZ{_q@i2CAomEsLA@G&B8k|-y@QJh%zNRy9Iik_a)jC7zTn|%2j5nFi))7#QKTDF zfTw?$4YZMcT3(Oj8x?+t+#+hBov6^EY2RKv(gM`qE0?+l_c?Z}1 znQh?gT8L- zkxxQh0J{JWXYt(NHF|&T`S~ndaXz!~I2TseE-cpt7wx~(eya4Cdh=?GG)7c&u#ChD zMSFdXwY;$gMt%(mqTI7ia+tz9!k~=6T1Tpzl(6dr z#4R3kJZRgSKuJ(Qe5JyLH(2|6ANW=veQRUzz&dfH*RQgMn^v*B-hwQjDxSj^JhcPK zUY27F^Nqn%>pV1^=@oGrb8Gey>Pt!1e!Hx7*P=Ph|bqS5zSc8`_s zg*Pzg+v&lD7f8QHXUR7qC+J4Z53jEk_F!Jiukv#~oj%{IEMm@4WUE zg66e9oy;B>rVm~sBByh#dn9DyQ4@YH2*8&TiUb-qoxbkcRg3++;{tw|qB zfyZ=2i}0M&5L>-FwTEXWWwU->T@1eBtiy`DH8jL zmk!gN=gcQamPyAtBXquNaN>nL!JZlChG)I@{5VO?(US26!sElG07S<=nU@WWyi0!cp=A`yzFefEWF?XYdXS?G(Ze!`0)i|bT8o?Vq_OF zU74LhZY&qR#JP`1zUDm9*(}PxJRi6NjBG7%=oeivMpT$pUFOE3cFg1E%div0~=zqr9qa-ya?Jug_r>2t#^rO2m&c)Gc z7v4KA`sFygZ(Q_S-3Et?M!sH@Q>lL9l-v)9${_J8#>JP9bJvWEt{#VRX;RJ4MO{BZ z9!RYm|D2<%Q6NE#T+Dik^T2Rez9N!j_KU5<1unWI9r8UzexYj+HXIc)q>m!Z%&^mUC}9&I@n4mf&}E zJkQ_|x>nQ-m=Ji4k1oXXgh>f;c=tGcS`uF5!i!yYk#m`urh6}O6OIcF-OLzPD zkZwD6dx7VgAx!#PVk&T^%DPHy&`oSGP8xYy1C+MF8qBB6Kzc-hJZTO;(66kO2k6D- zKRV>zS=@dzB7}3k9l6;Ro^~D2yYM-;$MbI3bM7<2hk!6!K+5)-E}>SxP7fY$!$5oW zA=dF^f}1IJM}~+P6+P+Q>oBHd7>fvi5_@GA@f7e>D|1~%1jWJ-S28I_Ij)AHx^4Ln z9T*hu1JcGw<1y8dXfY<^Nbg;ywSL45d^AC%kBiuT7y!NrJz?g$0fbSoFQeT=_gNOA zsjjU1sc_+rcermM#H;AFMlh1O>0nLvfNIn-ABLY1Uh4#};iA9)Ob%7lERoExxbJId zO`%S%`Bm2)Y{vBsu!YPr`L$rK<}E6ObOsR)aD$x&2V|VxJ(^RN^Vp&w(0)^w`41iaG!v(-+b( zFx7Ai9?V1bRkuhNgM+j%tR99EafJI70?HDR;a`bb;}374uLhjWhFVsHady1>~^>CPjZdoYO~LS(JJkXZk?vyq;{ zm-92+;q!%O7esrr=xj8ml6tN~&i!o2_g-my5J31OVjjky9Aaw2#wMQPHFWF~X2*PY zc;Pr4@@YZY;Lc>m#>6%yqkLQSZS30~-+t4#KeHiUe0#`wyFbr=T=eJ3kZ<1d5+Gv<u zTU^geZubN1I|)Z)K;7;?g~?N){V5DLh4HIdDGsw3E+yakl$4n-f8v=bH0tvLMV2_XxG8a3bfnH?wLaJEDAbz zRxw)1gX!y{f^Er?QnEAW5OUfKaLnA_3h{u9I7OIRnLS%!%jrAaQX!aX2oEn#kvo>F zdfmC@v?1vv0t{Z8@M0+jfemB}Btvi*I=J+!yrtyBUse}BhMrfEJ(Rz0oXjm> zt0?hY#BrQoXYMCManwI~?&>a~^oqhEK{u*h)DPG2p)RIwEJJ{O9R`NS3yKH{%!_Uk z@A$M$Wf8`6i-5R-L*THwx_XsSpTwO-jP+>;l?J*F=&88jb*PESj{&xmZ@@cH2!&3F z62z49*b@vPwLpyW`JalIL!kI+K|)jnD#`7ij*DU@3t9slbo_ZiJ#S-NS za0o_lI*Un;ouuzN+bjBcXm^C}=i%(&p>leCOd`nWnzCJ8b~izX^fC`ZNgdhU^vS^UW}`Y}TNpa<@bklySTyi-o` z_G2cZK|9KHI+1OX?72W*)^eE9oj2|v5?-Yp{`!G8g@5g zyO=XTSWEZYAYe(n~cH z6$nIEK>^@5SlQNdqlQ(Hohx%E(^BO8oT#{*w+<0hq0K)0cM1TBxLS`!wu04{J9<$i zD8r46UaF_O!9Kqp=78iOBm7HB!DgR=8bi$z5TSp?(#TGVii>%L7z-Y$xkKORUon}# zq2UGORlR~cy4?u>c;$E6WM5S3wxohD#F^HB^D%90G>=qSsB_9qE3<|bI34p)|F#c9 zdoL`$A8s2^?sv&b4INS)&T_RODJPL16kJD4n!<2N1q#b8IMJ8D8T#!AGEj5}*K+RB&zn{XBs(1L#sEi@->wkAjs=9an%2RZ4sxn5 zvqosT5M~P@6`3(*M{#$CTX(U&P-SD2EwnVeI=9kV-(HtAwl^fr?M-<#jd7@}_D{bY zeWkqAH7!9MyIGMee~Y3I>1$yH3e}q4YPqDhe2YzBve`?RZ15iQIxJ~JQx>Y?nGJ=)RcCJKDGTw?FXbGktroZ^!s{q;H4b`3-+1 z|J5Xa{`350)BGQw*_t&5D_!>@S2@ug@t)cH*7Tma_bt8LvMa3FpYOH=e>*qKU#WiE zeXDS}Rj;tI?^)%2i{7*BeVgjP+uh%{+qOG;{na*q_eZ}Qxy0F0S6brIrEZ@k@KNNI z)0H@Xi$Cx7?acSA<$Y`j7QS!O{j)0X5um3ljqjWD=X3m>ZN8sf!K{L2TEv(L*7yq1 ziD7XCnDrx$PzcMWvlNGYG z^e+{ntFoJ-`@8P=1MOLV^Y6l6vWQD|SXGhBE&GtGEvziDj80rm5h#jr5pnU0FrESh2qkv%x8 z*hns%>D*g#nkd1vfF8gjlHLpGWs1HxrE4O)IwG}W?V0KiFLd#f`Et?;eIq017ko;1 zV1o;jeZToKngb$iaew4h^bP+U{W|%j{>%%Iq&~@Q;+?3ySj26Ln@HT1>aQO05F(m1 zI|`-!6cXEra5V8|9ch-0l>30|wLX;Roh<$sGXJE7?s79mf=Y3E)KDqcH`JA@4VB@I z4dniAXzFThb=$QKqvBSAhfx!s&G^+;JG<+qf;WqW?eTtT7JmXe)nMfkV*=paW?tXT znx}RIET#hUczfE+=WclH4O0;}c&kEvwQelDu)G|YgodejRPWxNwCzSFph9=GtG6c` z*4A4cULE233L~riBS+h)(&%*5KBL_z<2)Xlj2T7>+HRz;nd5VR72Hn7hvTMC&L&mH zN8?aT`US|^!*P_3YM29O2k#Jfi8i|(Ho|eObh`%sUI93j(~#LlVo!0w=MTA)W?`LY zq)7Z$o85Zdq*7Ak|0g*rv>y+Q(91Z(ARa5-*x+A)LqLX9*ab*L3F=voWdk;*vR|~n zZs`v8Ler$zp(l_aiz|N3=rMJpj2`L7+l=4zxAol3T#YDVqH?7T_H+`+N7yZ7)b3ijw{C51ULVFVe zN`X{a`Db15Hbfqhc6}J%1bK}5*|j5VSfK|q>u}C{WWRs&?LJ%KI9IyKIkEBW_-}{b z+Lxn5)y*qxFV6lYxF*i_?*VY)@7clVjI-N`&_UiQ<9&t#z`kwg{N1%oJ|;i`>*JmP zMUb4$#ec-bI|{FI=|gTh;XDldK^H&d?0y$NK(Yh>)xEgFM}-a16i_=i{2C>cvtPRS zS7==Q4IBPH(%uBVuBux6KYQJZ0i2E>*S>SrL{Uu z4fA%Kf5ga!qy@wQ1V_T&Wz-==LV}>^Y=I{c?mDvE4^F}9vcb%qfMgJ^Qdtov5weqz z_uk{adDD;H;HPf%{h8RN`p>}PM|}4xKe$?aom}e&S71p^D0ET%J&l2q@>?q!oC{?D zBl;Vm?xx7@TLUSYCVFN<*GCeZ%TC94fWYNUq9EWpy@WHNhl?ycP(Ko}Gc0VJ-xJtB z$HBh!aJpzYPB+f$l3bzGf2p71;c_Rjhh%GGgJ+2#Ld156KMvY+Xh=Nf^9?EjCsTFk zGrClZ)QfMj?YJ(Mirc==A~$(@NA-1e2x_mrm0*L@RYCk5Z4xuMb)n{wg{AL8;ds)#x(0&T=x7PWw05mBc`B6)X?jJ1f7;`8;_*}Js5GotQ@kfuqsxOl(=%y2|q z8zD`T@-&z~riLPFZG<#UiV1-?U#u|XUlKeW9mVWjI+A_?eya2*k`m1j66`x_caale zQ!Z%_?|oE-<8x2J%xtN0Nk6SJpTGzq_e5}A$OZVSr`)N`ys8B3&WGp~?Yj-uR7~sH zGDWPTGN@~NEnKoIras&k^@#PBNLN6DaV@Md2Y0y+53a!z+G}~&*YoW)$1k|BuFTeNtW*Y%Gb`aW)vimji`5c$vAaYqE>5~`pL88vseD`@I723wt9In9D>PXZ zhS^@D67Srkkuod_8=*4KsL12kDC60QlX=o9a=;RtFaK!9O&ayn^;%u0-Rre`omST) zqsZJUcW>r|jdpL=?oC?V%w2RbsraMv3)|#_+I>j74{Alq8B%kq_BVu_JmW`0BbH=G z;LJecnuVK5-C2=hfZCZ1J>1LS$%gRNB6VfZB}$;k94ll;GZn2ydNCIV0pd=vI@TC% ziP1n=%&ZRUure8T;r6(6D!WpcL%a$*#eM$;vXnsj%)bD4>=u_n0Q zfosbu3eJ%n=oZfIv)YVU&n4h3MTzfXXS+rp;e2ZJye!+#Lx0%<6PA&Dzc<3k7m#Fs=e=d4XifnH(aq{UZU_FeupXDW>BpE z(e++%)t}wSt61+tG<)9R6>>txe7&q=v1Z5Ikq*MxgLQr&tN_cQN^^Lm$avP{y&!|8 zL*z|ah{~jih(~tph0GA^nMm;*kqg)I#W1VhK0%4mmvc;!=nj^+{C!AV&aq_W44Vcv z4IP(;4o#*HYIjf{$i%F2F7x7LRtmk6z>I)OeNwK|6`iewxm2gI!w8Jni7@J-;Q}c^>e>&-7FcJEK~&wXGB?04GV6)Wh32dsi_q<9 z3*LK~+A;CUD!QhG$^~k@L3U1>CUx~&Lg6x0FFYk*)v}U=%Mwv>cqeo=_ORH0(CbZ@1m&SgtMOV{hIAoZI6QqDbowrc z0>x0s7gM1)J(2frSN?I{?vB$Yt)SmpK&K1v)qiLOl^O^c6F)E&Ny-8JzBmXoKhL9+ zJDv52M5R_BA39L%G0?&#nfh`s)WLCIWWIUUX~?$hZ^=arp0i{Te96PqCpd6tY=JV) zZGLdG@7=_9p>Fhp>o`L$&hSaE@Dn$ornUsEgba{VII^z@e(}-1T2qX?ThcAWkFW|! z&wN`bZwzms;dZ9~#naWIG(v6va03^)Mn|zSQWS5#TddLd=#@Hm2D)A68Bkavh*eo6MZUCI>vraV zT-$%FBATc6R8xqLVzy^;2gJW3npR{tcTa|uQBMlh9}{&oE}VJQ#W%dZMVSTa$M_r9 zy8^CNm=^z-F8iqvwEmAFet)3;GdM@7v+>v}1NjeEzvrrdxbAzdepr-cuD)OSf?nKq zCX?Tynx(hE++Hzb=Ij5|8wg*ZUQxLhR1sHrh_HACzKefE@6yq-+k}@cnw|@ zZ$vXIC#7XoZW0WW)y=#NyT!G5Ig{(qR#7Mpnq((~ER6|2K1>m=1uwz6@9qvOJa6_A zYjp{98M)&k$UJRagf8ot;nF<&X%>;UbSQf0!*Qs<0BF=Z2UwwkJ1K3m*Yz)AuChi? zJxm`8>wM@}4y!se?Ytcoc?h-wuwua*I)X7z2EP9vQT<+&REr$DS13J!cQ=Ikq7xuA+0`=S_Hi9y7)k4Tuuu4%}8psd(!$Wm*!8;8lj>dU(Z$>hwXHVXbnQJH#OwV?~^( z>zOt2T!f9y=NX-y!u%RiwQI$)cBV}WYdL`qHv@taJ4yGz0``%Dtn`;$+rdw2!=22R zvd*({uno>k)GuS|oR~U0rp}1DD-o@NsaD6#xiPhxjmexB(?eiO^{{mAVP;*-T!~Qz zOIGu{mu5BOFzhQaZwP@92Ep^OnsUuS7z<0U@{!dpo{#M}1j}qTOmgl^yYZXNW@&A| z2lUjfa?Z%JX|qk;i8PRQRVK+E&Zi;*q}L^Lx;Wz2tIblwVNinXI)2_SV|XIJ5@b-Q zN?3h_#_oJpbQ;>y#~`suBy&^}a&r$s%=%DUSZ*tX_!^zCMAURkL``w7NKpSL0`pAJ zg@DVv6)TOiaT6I;8&Q8zivG$rJc81PlA%r#czBnCS8=2}2{8*3w-?hfe3hXaQ@g#W zkUA!9%GnA*-ib?9{xTd8Ds58?A^V(9rR?9w^H0S|wM{X_e+coO0l|ad+@ge=kvdrI zBMaD;Psfz3x~oOV4>>qciwfToPmLI`>@3xph?(V44tv|BDsvep0!~au?~5`6eSIfh z+hmsgWUb7yNI1VR6JCYoBPPOj|KQ5i^%9s>%n13^NXR3QHs>_sjSPsw*Bc~zf_Z>8 zhTWBY?jTM(g>~i^yD78pv!X?p4LxYI1qEn=aRu=Nx#~|*9S!C?p{t# zZwi9TS7441zciilXg~cVdSOmJl5h)uC~{>tLmsh%6beNuL&ZBntYLD&GjxqiD2aSB z#hnG`-3fD@^D;qZg8$mg9s_!1IE9Zl*6p~<1I0)zRz!o#IZT9PnbBvquC&AgrzwG369>v9y* zolO{QW)uEN+uT+my%|$j9RqYK=e785nK1ZRC{mP8m{?hOxngb# zViA>Y(SGZYj&z(>qE;k1q8xRrMD-jH4{#+-k7F`tHyWCMC^Fs_=c07xN$gXI7uPxH5?oVOyOz5w4X-yh*nHR3sUzskgO8XMZ zM(XfY+A3{s4{4(vpb`YT5jYN6t6W9I^^DLjAZ3wk;eZA~Av{?}Yl3KQ8IIxDealaip;Xe~HSgEmvdn-l9c)<=mG2GaVk*|z2&lC#!( zYIV3pPmB;gfpE!#^qBj&#C7|=FCI)}vPR{I4fNPVI5uZ8d*yRa;u!&8_~HRtQC3ZEiL1wz}Ky zGF!=Z3r0NZlZ-kitIx_tQg}Go&FqelP^3HzrdLx{&+d|u{ZS0&O%JZkvAc(4Fws?6 zbr#zkXm?8{^~(vj*+S4`o#L+$EW^HSgQEJd;=+rxNSU$`;lu$;@D8)9-j$`55YsE# zjI>9{cccYe)1&-j+S_XugB(S+@!ebqOM~+U5ofx|Eb|E)-fWIw)2XVe;#^402)bbz zp^`HY5toUqJ*(phE%SE=HLXz5Y@R1k0%g6N0VcAUT*|puyd<6H=hyh8+NKaGXV4pZ zB@zpD6175Qa~Asq-Ele`ZUwD*)Kr>0fmqpqX|UXpvr2NN#ES3gpjf6d=_XMu%u$no zb`&eU0X0t+go{z_3f0D0y=Lb<*Z>4eb=fxTIbblo7JUZ@XJSU+swj>DS%r9Oq81&L z#~xP0whuBBz6f(TcY7{ucUs#E9s_4$4%bXMwXZ7kimG{4MPE_Dt19^l{dN+ev$4Cz zHpMj`VIgPTr*CE&_#c5R)o-ZuzxYlcHsU&6M@Aw`H%8P*WQt9{XVV>RT4&Sm+SIpc zX;QjY*)P?T(xSai*mR~%Yc_IHl(FNiP=MjOINQJ$ahY;YRG#J9(K6kS4>2_*tGk7@LIs zCS{$ObWH(dJ=+#T0lXVlB^`-dp%NGfh&NE@iXeJ%L|=j#r*}49S(Um>^c?1eh`Q9c zo3!_SL~V(v&5`)WqGCbXR9qRT(~*5KGFEY@ZcCfrruFUOjQoytorpKCNtJ~%3P|0v9w*#h#e-@=m8 zdS8$t%ZPA~UF!6dITNQsVg-L@DqU1VDRXH`Unbu#O{FhN@xo213}FP?S=VmFkKUiu z_a&Ln8N~luflwqksOs-2+3K61l!LkwGdh7Q;}|ENme% zf5z`=y`1HZE5QS}i2#NUWqi1f>fZ=w$A(!$Bnu&SMeQ0Rs7D;%g$+nblkp_Ru|!SP zGlM7_YMxCG`Xj$@RSRnN4pB%A zJL1;XM&!7P3J1;z=3jIUdkcfh>4R%HH6?UVS@_0HxazSu)b^o5dJECzZ$q(SU@Jt3 za8m3dNpLXRB6Q^nJabIxNaK%(;{28)VYo zj^(hqBLQV4wF?TuZNqy6agP;XnG3CnRl@iSBayS5Aw-|d0%{Dp&FfaEk=cloBUYXg(6xC43TmnQ5!#!4D!oV?nJ>fAg6;p&!zK4ppe zWU=~mu|9c;I%SDIZ%O5VxSVE}(K$~>P%gG(n3$jSP%(Mktoq$*g!veH$d2Tiw_znX z&8_n6PI9a+hR9tc=r{=#YDJpN_`ou|lP?_IQwN z=(Q@6V&go)d6~JbZi$yR#7PzDRdIc+x;nuaX2CI6p`Wg7qHB;$P_&c;LrHsO6{ZUl zp5CQcX5#0z_c^S1xhLB^r+cVilr}OP(4|OAD1t`kb$t?(iUdxBE~2?9sD5(cLrWlE z;^qX|%b+{b&g+TjRiUWj5-X~}yYC4}nHmIyEm=xMXdJ@_Lkd34*<3E`+@AE$Oqh!j z)d>0$`HK?j!h~9%&{yH1SzngW7su2ERwi~1GBB*-H7?&cgXNoxVqSZN6#|ig0i-n( zlnVy`G=|18j$*_i&JVJ)|n?!9?Mq zgnlpqm3Jr3o7Ejs_-!$DL(FWT*J$K~e<+aZ=PH3%C#(b&lm3(an6=-GghTAy34tj$h5VFSiGzsHCw@_CgbkTr6*w25kWfEQpt>NU)dhq^GgNy-?XVGEV1?>)ZcZ^d zVFKXz`4f|JX*rrQ5ygoqA;z7^00oV)=7Tpt> zW~*|4VX_K302c@o)8?I|ohLEYml?^76@L4cR57KlPH}<~v5;(2fv&|rJffKI@;lSs zCv3nfIgg^wQRZwSB#HH-Nhi~&UYLR`L-<$}t(MgByU5#o)aZTKsGn^Vulv7Cn~SEa zbEm5{)BWG{sNo)UZ4bId%$?X!r8i_)84`BgXL}gH=Njn&F~=9P6{1a*vFD9#Xf)R~ zsgGUnpRW0s*oFLhrl&q~-A0&h=B~NW3l^$(RAy zsj!56Yx4Zaeq#4c&;e*cu_wp1a&lpxT3^GeLWq$N;U*v zUy?vb#SU_a)t-mOn&I`vZT6e8@KWF@!B^|mkAffTAL$=w`0VVrdwtHTEb_-dy_3#u zO6xZupVX^qvFnC>aq&W{{n`mUZL;}lV$w46n2DT{9kWq#U(i*51d&D;8?kNsitTTu zzC<^_1YL%Wzg`#9SH|@9vBW~=&|W-nc}(5vs<9L{c1aYSd(bENllE)G22S*kqVO>@ zh`?vmGY#gx26eyCH5+@(|2$Ry*5h5*fTw>#+o=837tu!HeF6EaPOlVhS&+Y=XF2;l zfpj^c$M|Ebh z@H5AZY!uvA1*W?<7dUngMt*(5BYw6E*Q!M zaT{rqxDN=jk(r1);SAJYN@8u+P`^5`vhV@}YUbpjn-J3IT(k3To+12QF8@G|RFmGC z(<4}yqREztB#r-9I;~8vfL&UU_QhnlW7`BTS~WW#O*|NgvGVQCoUG^v(9)OK+ugP-2TP3?e86t3%1Re!d#* z&MAfHgJCu9mTF+!{_V9IRw1&goZAafbZ6zOR^|0s zIEc|J@@RlB%))1UnvqEZc1Jk?^NPei3+c5ctIp0E zYl&b$=eB@fQ-qliG{t)n81k=)rUuhaOO|TWRw2Gb^t@<)*f9(ooY5ltwHzUpjzKKX zjwh!$`EGV%@diaQWP_fL1hNOgWPy^TH#+Sy6MGIzn}MV_d?bTiuXlu4<0+S$iko{{tQTIg-rNXOC) z=1Xb)I@)|J9I=LFmy1OU8KS;Gvyq+WrsatG#Pxlt}5 zo9a_!4KXu>+J!>G5NFc220`3-VB}o2m?^vZ-}sTN z@537fL>I`2;gl^XW=UQmk|?f16YC#y-l@uYne%=W%76M1843u)9J*FcNPNwvflbX? z{Kd$s#Ee1Rf_byDm}!>Zj9JskSZV&*$WV$ za|L$)t2Ws2yrrwgAzdO<*gwR5gKRq@PmunYz$q=V1g%ar3aw_TMp<1nzd*x4$4MAA zL@bE6V&MHU25nif;IWd&3M(n?JETGd%yOQoj-DqN4|mxGLDVB^cj1msFZmy8{jqlc zsohUGX{(zDOd(;imCeATP*@lS$sz?yyUH0{OOwR0uoV+}){QV!EKuF78g1TlH!8e< znT;%<6)L^0^Hi?IS%sq%s`i(v{F`!AR+0X(O1;2Ya(1K{i(n^$4U$pwBl1;Hhk^w!&Sy#dpi-_+TBNgM6hK_(j z;nv(pLr$X4hoiD2?CT6-#W+XV+S#Ne%Z$zJ>o)xlo5o($vX-l8Xh^o|+@o0CmF&v< zO?Ce2YOwVU)#}D-b$PX3UoEO=i!ZBEe>d*C#!P3Hun5qSUgi2981oW+u3t84Gh|bR zzrmG*tYc5VZ`@O;tH?F*O;LmEM?NfbE&iJBHnC!5>4_i6iTU*J%pp&=*g1sGFI-*I zanaDMccm9nm{v6rv6nYR5%?tJ3^9ad0?DLfN8lhQvLp=BtzlSoB#JMx9T1+CIRzcP zylCguMMZiX6ZHo|mt*s`x>@>U$Ln?Hb^0#l++GtrR`ee%>W7L&5I#G`Fk3J)mMLCur8tU!8^)YpA|ymfrJUv)4-aY(eAy6+x@CEB4s4IM#VS4hGYN$`ip# z{sc3i)=1=4bzJiJ(^P}o=r;5;W*ZtCH{d8C^1kwGQnd;8-hGZc`gn639=MLy$9cyW zo%OZB`!(i+ny4?HvMPU!A~;u$E6dbn#N^M5%HCP|msFc;^e<}F&uimnV3fpq!hliJ)BbV!^yxu)c9* zH^MJDN4f2(Bi4Uq2l8~?)dZNeHs(ZsppLx10IKQR3V&!nCl@NAyx-QEJ8RW-waF2} zCU7o*HOY}X-KTf1?Tc@L^px0tLURa@AR0K{%jjE6=0oGsg}0UTb@jon!mo+2gg^=J z)x~~S(hrsNgC#|}m4hVsN8?qy#(CZSAdkaW^k&f-70f~P?Ie`LD&Tj7UWEC9+%0cIH9~4tws(8jMQmk2)bc#a zE7^F7XUKII*AYFN-v})OErBAD?`iwnTzd_wCMQln@eJDuzb@fQ0)HW#ws_pvSSRuB z#5X~Mh_1r7=>a)GB|{_mV3JBYg}Z`233bP_*vz?7GY+R%k_}IZ)GrJn_7lqL<~w#@ zJRyGT5bspPVi8mVutr6BvGV8Yh4JQ;*hKDR5h{8NSno`HZl>PgjgWMj-+-v28S5QC zUWog(JMb*(RbsIv34LA_58Axyd?^W9PUN0mHy15TqcqVO zdQ4#mE?qUNaZY5%n0)9ncju}?d@39ozYrjUaAo3jFSoDTz*wSTG*3f;f<4u-;i!hY zDcuu8GsvE0HAqt{7;Ssb!Eh)fp24m{gbLLX_5!_N49f`7H=rjbN^)(!AffK7O2t`( zxp!;bnmi6C*C_Fl(3or}V2Y~CwM}1bp>_~cx-HP{?XgaFAdcDVsf}G5ba!N$KRwlx z=nZD<(|B})*K1va1~6kGozC^A+rH0i0>Iq#0??dIyq(`D|!mnf+O~_jAZ9p<4^O;F4>WvZAV5!xrjVMr8sQj_P&#I0!Kg&4JG~ku{(du9n!4cx_ zN2|?a)i{rUlpr=y${u((dWD-PIdu!3uS-^D&3{;rk?XDTt14W?_IEMLI7j z^)!Sl5CuUtZ4wppJW*PT8h<*iqw;w~2sIQDiCQzg2wLlQ$C+tYkb--{*weOXjEU$l8s*`zy$5f6h!v{`xwrOGJ}WrJj5H@1X;Wx+W}4oMUR?vt|=p@THcgDsfkpR zzTc=nY80Y9j|}Sia`gRrOYy#4FPzkXn$5u-QIAbX+CJZysB^nC=B%dtswPrxQyOy$ zx#M$HeJqVRxq-%<*r5JdZw4AL{|YDd|JD@oZw<{UP2uRBD9Nf}WASP6zqLu+Ut2XI zqB#>P6XV;e>*l6A50zQ=Yn$>zO{6+ZQ~Fm;ggh1E8J?)3NEEf%Y!J`t|G4ynyl(Rof7m%D1Z^H@20R)g)nUo=Sf4XIt$ip0>jDV6(7ZfJ5dofpa? zmsdA6YYDUx)p#v9b;2DOE#J?cZ51s^o%cN+8u4Kyc+VJpKX zE&>hcoxDi}UK1jnv}wX6ZP;$>oO{xAhoQ04o&K@M`wk6t>RBy;+hq?;fT=n~GcT;m z8s#EAuHME(Qwh3I6iA}{(&on^zKWjWu8pdpXxygFldi)%cPo2q=34x%tWbNfkRyCl zHfyrcYYR@+(HQ`Pzah6IfG;Pee7=9)= zXrnodr2o>wVVR*;^H+hFw33Awz_DLfa%w|aUc*d4EUs!IwQQTHeyBQl2f7>kwuh*i z>~AR9HoEG5SKa68yIpmcs~>W?moN{z>H)1zf=rK|hC5r~&Iq`n%d0c03vqpVm={xH+VY1Ru$94dDiabrZ>l zVEi2>-IIMv?(Gkf;x<_kYLx3nF-js2?vP@+LApQBidcLOzn6pzcMsteDP=F|WWjM6 zD+y@OEs(FqRdUEj0^H1uwxs-Yfbg(XC?&s;kJU~~61cmWwh$7; z+y~pffPR*yLt)V}siQ1CwsaARV=TmI#UrIgFZ8 zJr%)IUm=(8M=lwwHnpU6^ZAXZ71-t*eE&cS|p;vkeabp_JDYZ{TDhXwumt^6OHF37L%}|vKb`M$YdC)1pW9)lX|wSzpQa4Bf}|6TmQv#$363or{BeZUM0MLcxsa;Z{sxI^WSI- zUTBM)Sw>9@nW!Hr>n`=1vU-4X@%TXB$?U3XX};JylO4LV9`&v2Tbj*n&0v=AWWS}p znQ4SC>LOaqm$PA~t&@RfE4K@i$_u^0wNTBV8;R1yuTnfiUSJQyodz>D14o4_oF&(! zy*yco9`=0JW&>?(Ve*JiFJ1*pVUp(KE(z>l$%6<;-E0@Drqg3YBURDa%s?9(#7Cj! z;+O!vB^?@X(2Xt5#~l#Ce{MEoG(eO>`i*A2iEECYxsfzy3By?rj%1@2qgm!(5$Rzd zzsUKtdiGy7Te^Y;tWAq^34(tu+v?|zj+wP?go0NqXZ;k+ttjcK*~AT-&LKJ7tLU4W0QLJJk)HDoj&)U8lOd1Ke^E zI}WS%s*c3E4!Pr-4IONb*R@~)bA5~cO^dpwMd4P$oZIED?oz!P-^J#uJU63Pl!#K_gcQwY{GT_Ql^- z2IF14V~I--cCJ!1z+EE7^D-p~!Wkr&3j;ba=R34f+?VOmIe@^L{!gU+D3Jq zz23`SOD&@s;e|U*78E@(o%<2}J&VL?8RI6=AX!Oi<~BMg^99g-Lbpe26J7icdCjRU zK~m!WKEXQ6(Qd?ExHPOv#Nug=b05$|Eg`bZ%d8gB*WrDQB`RAq_2s_A+!0h|@J7TG z7HfBm&)sN--Jf}s2D2qSh>92@B_HQJSw2xdULhl~S9NPOGwrfz(L1_$dqbD`ZI^$0 zmw1>SQuTJ1b!}P<_uAc^SEoX7CODUyiR19DC?PXyT?B8=?Ndea@~>>Iknyl+-RvcQ zo|=DQDyg<9Wc8C%)#IJa;0KuD>WNNER6o**kG~K}+VQaCLNi4gVtZLkF)elyc9w9I z|DB7KbKxNe8yc6uBnpzA%)xSovKtQeA5-)1O(oS~n$mAh6%Lj)A-AY90`VyiTnYi)5Z>5jhEX}Q)HIz_QKr1xGcTeL3`7uR=F1WHhhwW^^G zaPlVS`Z&S}?E;3zjKnkvRBgAWLi>I-+gU<;J zCU*)SxmJiG0!z&?=GFxt(SK*pSDbck3MxO#KhYOz}04s+^hGx%v)V%V5)m=xA{dkUbGNP)9LtB zvVwy1yNjG=VL-qr5{ZHiiEYP{s*ZZOFcKR(%)iV584TeK*nDJ9qx*b`uZ%FAA!M+? z5v=@*kxZA{#Hy}`oi6H@B-9(HPOJe~lHq5sX6dql!9EKe&t`<(UA0#eq#|>jhkBwb zr`kz$%2XMk5p~!`(UT`_8nvkx$_`_eQ5I=}&V9%E6HeFEb=`&QyGa%4rQLcR!HgmQ zX&(VoVHP|hbwG`jd{-Z*;@G zyxyI8ty{g`O@|`_XQ70tcqirSDJG1JhveEtX`?*RdT< z92*qM3uhTJq_U<2gED-~R~R`~g+=&{++mUD;1hN*&3+p571kI$9Hc1jgbD8E0uvhz7<3%O{CFMGtqBX$h+*vsp2 zE}6k5clmU*S~pKuH%%vXZz2W&f7b)hbUoLjp6}78_NX7Rg62VIN+bAHsV}FCVj&RT zU@qfQQ+_g=dICSk7U55r)YDG(8cGy$kZVr2axJc?yGS!8HlJ z1G*QC>e&QAdyuE1HI*sLCk2DU=0W%2NSgUohP3pOJ+|F5PY_x#cv*G4q}-R4dI^Eu zMPd$rEd)m%S~RsT;a-_g>o^@kZWZ)lD*@i1N`#W2LePiw4r{1J*ulV_xnx2UE`a0v zV&=7IH=Yo;%inD9()GBhbvEhzef+K;(6yoxAX{#}UeA2-MoiU4qeje+HX_@zMIXI^reJ18shbogL5WnEB3#oB`St%zEy?aD&9e9!(I^e+;4@h#CxkkEO{ifo zAV+%fN*!vxrlC-^i(lh=%Qruh9XP|wM@z?G%zQKYgk2adwhZ`rNG(}Ac?L_Kqa(Q} zfcygIPV;~!S?7|hma9QS@-sg7Dw4u{Z-Jg~7NqC93)FnI;1KS?@bH!-SBu5OfUe#u z;#Wkg-hg<~E>WJG>m5>^7UAf%%9Bu@I(~fYICXsHxGKkuZ>oyni2v=X%v)8~_=P!5 zQ=4foHO{(?;xhmpBtw(IiyL#R3QkWS?k(cyIz}>-L%02F0RaW1YyltTf)8_QAfMWr zs}vyO)yA-ZG{i0r$-_cqZ%IiZ$_DM^P>EVpq%@l1v72kRuWDYYk8*jAK1&dRj}W%5 zX_jhb4)>`NuZm+yp?LZV^9p?xQAZW*G<bMdea^vjS4YSp4 zvs1UuE`>!&hcr=N7D2XBU%1k+Q&Ki!$B1g$D7%r~s5}t4qZ7v$kIOr5;@MfnKg_D= zk;EPs)S1Q_*eA>iXrw597*t<9H5NnXzUq0keRdLRoE~xRQijkqWquk^aY32-ggD0u zrM~*OYU=uOTECRW-LW_%5znf9k$C})w;d$NN_2F60rx8XAl#>rvtl}8vqK(MdAyoQ z%TO1XGq{*gec6d_Ow1z-xl_#@J0=p+#9qi$5PV}F_U9$#=hX2Y_9*^w1;6NjeG7V}N&MzIuA2Bw5U%}Ab}8n0 zipEvu8bn(-*Kj@TF3K|3=(Ze#!)=T`#8lNS$C?yc=xCYBh>e6O3VAd`U%_mnjZB=P zf|HpqZsKl2d${+YZ{|S|vS=E`N?zo7<%rG2D_6!vRBy~0HV9eEvE!4+WgRzmQ9gMg zk{gYgRLhHs^H%yr<&BBCeCqfr+xG*7s;&7-e|Jopya?%;ytoz@-#>A(W1!-v7m}wH zG3=0}cBd~@r!AG* z^+hW~$;1eLq0G;f`GxX-4#nfL3xa9iQ07(TzapApf<22jBe5C~9Bij#`hq9l(ZfUk}kzVPcubhXetKc!S6+GFyBjV>zIwZJmrXb>{d6S-F{kv&ko~ zB3oCnNbQ?5xWPyi*o8T z^CHrjV$WvXBz7L|k_{6-2?os{m!heoKV*VG@k8~@5A#p|kYc47LcFj$LlcMv97tP^ zc^8CHA{GS?VC~L$;E1)zcwDBlmx3&})g(?$>_pCMbM7OMZ=R5?b|RE2r!1Xd}7_>j@)J#dpqO4l`-#S zFe(xn6I#dJI+1mQB{)ULPvfc(0OAEpcmp}MNa>-2p}dq5J4Gc=Rx-sdOD8W)%M=6g zXQTy~A@94^xVYz)X}z0tT>Rz>@baX@o(M&NbQ7c$jxDP;5_wy?Q4~$Euyhxma-fjH zBWjiIEOG`R$Js`@G_H4(?o>Z8|0(!_2+LrID3e`A=bQmtG@@3ZE(0oeLTZ2=)(%kt zEXUo*NbF&~F{VahwKmP!v}R+B-BI!C4f9ZQR#Qn-?1a!_K@G8R;uoiIY-a(K~g z`|P*M(=GIZus95cyJu*1)BvBcOy<*eSeB`hJwe3NaSp&2Q7+c%BV!Q<%$eo zc4Yjmagpf7GYr-g8T2;JPH@;k^dn$baWay0CH&7&WR5*2u!l`|x9N5^t+MGOp(F43 zS`-{^ake+BW44+V-6y^GD|*lP9zD)%q7w?CjbAd)yAa~H{Q^&&i)SX>3B~YlPan4t zOB2D8_Uc{O<~BGYj*}fwU50pOo{4c^3*9L487!`J3v_o9egv7bn|W5pUlSslsot`$ z-RKzbWQ0UgyzKoICO=uD1n=d zCS3^q-C*I7pX+qG6sIs*d4J)ZB7N9FsiF#h1Z zXECygN`Qd|6peLQ%_YE3;J)=p(erj8D(qREYgCk};Opb53JRyY)JCmFbi+nLO}X8p zyQ$rS-FDtRvRi6*f45VVGpZs;k67U+WunV6>cmX&NxH<^MY5@J@*4R>1BG zBF9!PA~og=BWbIP!adSp@to0PqT6GGK)won=EARZ_^mit;~a)WM`vNB&}Q+lC`GnV zZ;9Y3hkAF$WoZg!Iu&9QW9)_2m($oR4(U;E75>NLj@xyG`r+j&wIsQCiL)P-rk$WA z7w6CG8FV}Vp^ae`%Y6|9Gr{4e(3{C8X__!n6=st|(6_?N`J*QJAcN0?NF^d7xrM37 zig^N8apBED07CJIBAkDM1o<<*+0AZ<;RLMZY0=pF8r3(&$~DH?T+7;&YmxE5`icd2 zMz^3o$}`A?c!6Lz5naZ@zm_naRyMmvJ5HDWp?F6kl68kgi%|~2uje?J4HXj^seW5+ ztX3n{Pq`M9CZ%b6y(;|udahie$--f$Y<}OuBpr5X44Bu0&pG$=qm9z?#X%##Ls8m-z(b&FqRL{;E`K*D7t^0=Ax`%{u~rrOn?3uF?7# zU;o}$PjR}yFFvBG?-Mz}y()JXdl(qy6&sQ5DhVKTv&!8-WY`+5hJE#GzlJ7=rA_iK zow}Q#+lIx>0z!d+KG(-!Gn8fshn>a034?l8h^Uu~HOLfn1G4s$sKFlz|t^;GX%BfVo*P9%2(o9f8G#gwRa1SFy zPFYupCM;4#O9?}kPBgaa`ldvAhR!#~+S0n6C51KZwCMC~k5&h4A0v12QcZ5nPK8}O zeI&n%i8rEIrY=@4@OGSDq!wvtH7G#jm*J-fMTY)FB$_()&D;?aIbr{(-J^25&e}eS zY4PN{k>gUkly;Yenzy81Zi+v3^6!KVZdtAT@5g>14#E=Wt7MzMGqYUmZk|KpGDl=v z*n%}c5G_|WtN@PG^~2N;}7QEqx*0SDLp{ zEhx>8?_Sz1zGrEV6rKkBz5V@>WA1xIf9Zhufu%3Szf$@#Sb1r^vj$`-M#}F(X9zLB z_X!c_Fp;$W&%0~FV8Rkos&XiS(PwswB$}xuJ3p4ZvNMW~sUF>wkO7ubZE}-Nr|k1= z!*m)aN+3op!7Imq<#<(`9f>Rw3=6J)oo`{lrv zVHO2^i#Am*JPMVxBNT^Tt-Y5dAK#m#WnG6T%(Apljghfv$v`fR)ftglt zVp$&m4&6Vo3vzq*LKh~Uk~k@og*dm5v&KZSWY=OSdF2~=ZVkCvkEYrXvVmtKSk@ry zHl9RbXXUtFJM`UvDQ^`r?a|OPziYpcHo+Qj{F9ARioXeSsl?f?`PU);{T4kjljUozRnU4z0+JQ?E}~7{*kYln65D z?JC>DWt-x8AnrUMTOofrx>y!O&i94n;Z!|el4vRrQXs2tu(ImeHjUcU+lXMoO+jMU z?URsL(|gZ#aCgY|PJ33k=49FST#PtYsRUkl-c?w0db--r)Kr+%F22s8=NM2SSidO3 zLphvtl*n-mw*QP84ol`j$1gDRGtS#8dXJFLcZy9tRhnk2G3)bHYINMWRA+zZBuG?d zEiJQ^SeiHrVLgRc_NpBzd`J2TlDH1jy+)|?E|1^lVhp>#GpM7lNl{}`RAkKE@kQK( zxZBxup-rPUO}ywL_uAu_!cWdr&OYcPAQ7Y!p+hOS8kWj{OXV8=0gH)hm5NCi4-gk* zFAD;nHbz$mv3PPO@=h6GnF|l0ffA;o$gaR3fYzGTF6{HIQ67pSq<9))d=JD3lnQeg z+hr}moH#{h!4WnM*2d_Hl}JM)Cm4A56TFJUw#W%Fx#DXeSMCYhAuAOT2Q1z)RbTlM z$erMX_J#4iYWZ+tT$O?SeEOV0l}JPpl2CV9ofsuY@C*ZC6Hkj`hT7i&A;8@0GDPdT zR!kUJ6cHZrW6^$f070OjI_hzS#Yn7#J}y{$K1)UUOvpn~Wh1 zCFsSYFh!AwH0(h_Hru8ubcXd3`v=JMO#DZ+hewZ^L>9G!OTGYZI7>#Dxz6l^ zm1eTYb}XSsKySWS)v~fJ+|&Ll<{l!Z7XxKGQ*rvL=h5*I4(Zl`V>_v3QJ3&%K=_-y zUzIC1a005xCX^{RHpI&?rLDyX6-lrX-vjxsq!Hk3L>%-B2A)tn*#BbY8pBBpffot* z8N{#%RWrE$e1kHF(ZEeT(U~Ol7iaRyaRfdY}hl=bICNTSSd}~GxWb3zP zkV@wz0#G8sCpqbp;#xj(Up|@mY68`WdD3*#W2UPe)f~NJmf%b?`84rS`=`EmAh4ZJ zQohFu(ItF#h;>HW_&%5GtmTEgt5x>bxa_Wp%BQRH>`dHB%X3klllyhh&cT%EKaK8`cPaYg!h?&n-#~u_QMTTcxS-gQL0ZF zwKF{)-hEf8waQ?ts*rBk9ACG0a(E#!gax|ZjqN6`iavFTd~M!z_1>HVmw>B8kYxJD zZgXpHkyQaVoJ8&fL%+_gxxuK#jYDqDg|2#nxy%0fB)C2rOoMut*@e9A6t`x%9RHIy zJ~lNUVqF<*F*Tfj{lFC7!tmOpQpqg*vxfnTtTV8CBqPV!Bm`JaJz@$E;seu`3zb38 z*4s_zbtw1q!=~msR7)5*CWITJUh;dO&M}3vVCblii~LYOsl6w#)nGf5#h}$w$U;rc zRvjBKHJ^yA!;NjxHGkLYU0rwwVohqA+oVgcu!#8W1zqz8TPt2HJg#G-y5>=B2tf$S zkuqe>ipvMHD&tCA&F~JQ; z?c{CAWrD$xds4Rv>R(EsY?}-3;ni6P1I{O-?5rRi`y2Zb=DMFfCzj?SD79=CHIR2L ziBh?S`=B<~8op9LL0kX&Mkw=^`c1Agyu;^u0m8-h?&O}CTxaTH>qv#A{*=pjZ0i@8 z#Q*j-y|AstGG>!Y7ef#cfkSEV3_B+gckLfCM5a-SX>~io2j)?hT{-VI5F?}qb^)nl zB}FF8Lpad}<9aAZqHcpeK<_!Fc51xqfbM`_#K2#ojR_mlKz}#Dk`zd1bEkzP4~3v1 z(i&?xyi8zXM1nFm;q76~*+#LXkj+#PaRr$h;%JLK*(M+spuair!pQ`YOfva#a$X44 z`eKS9(?^1a>4t2%rfFtNO>jY?qpyW4qRTp-VK(3kvgR!c(;ES+hrUVPUV>J~eySi7}S& z4Z3G5V;Q9~I)>AEQ5#ZW}_MtMk5(e;j~U!yFB0*_e=d_+5)61ooB|Nq6GJXb@f zV6rLfp|&q5tjQ z$Ssbulpn`|4l``w254H~5U>xp1vm`sztwR@fa`z*nrClwoH5|RKnYnl3|uN))^GU^ zYq1D82!FK!*iTnyZs#8GNZ=rF1n590mhRv;;8I{2_yDjUSiX~IfX4twfuq1O+Q`n0 zJPRBEjsTwpmQjH9j*uV0t5yIPp$U5cIE2zFa~JJEqg7q@&XP3oqM?l>;(=(ttS;4#24 z;3zONh5mVvynqA1zE1iFSni^J4{;B;0yqX717@Zn4|tgJddLen3fu%72lhR}@6#P; zHLw>r4(tame3X2EL%<>67T^f5|1q8ct^<~5(ALMv6L>Jt>7@<8rNCg6ytku{zzkmK zHvmU~&J#Q~liz^-vnUrh0vrPl??^qLq}-jz6F3aq1ni&3^S`Gaz!ku;K0d(md_GT+ z=K}f%I0zgCW_F{0p5~ps$O||C906t)G9R9yeBe^xBH#nS!TrhS59A3v5;)w?^T2VS z^DNJOiS_}9fy2Pefwb>Au7Qhyqrla`zOV2OaPT1d?|I7j8g&Nt9!&oMhk$ebNP7;U z9l)iB@-DFVFzWj!@;aP*z%k$k;NT+4`ZIX~`+?0iF?f zBJBY7E~g!@a1S^L96gC~0`{HE=T+WaK|a8-pYs6@{(|RUqrRtc4IBk-lJBSU{Of!N zt_Jq6YvZEz!BgUVEIDw`78MY*8zLi^3LD*elh(CTyzQj33M){-`=9XfQx{GSMWS=_-cN8 zoBV#m^T5GtxCa~>=J|KX?>e3bjsiCT%hyxqaq0v-Qs51|3tW05^YQOo0}mE>3;6&u zx027hdr#EW)D zfCRg7@X75A_(>y%FVXkl%n98#qg2$~hP~m{88JkYj1(1cvv3i-4IT&jXz*u3hp6 z9tj+;R?a%$K#6PMqFUuFq^xn^5OAPQIa`3`dgb&}_5g4la2)7J*^QJ7%*^I_U>Qgp z2xoi_&j-|JC*`aL_Ri%VuslyWbE16TLpdveL%;`sV?fS3Ib(ZL2Vn0)>I>`#avZ}M z0Uit--kbb@nSIC)xD>dMHkbDe+di_Fx8C8_V1`2fd&Bf!3|QBI0_A57lB z5#VZI`4Hti3mgQN=(C~kk~c8(@02Nh2$a6;Ka%=GgbV_Y0hYf<9>4)$Da-HQ=RM%q zQtA&J`5}4a$nPigH*okEKEU3eQja|KIF@&SL&wptz`^6WSD>CJ&|Y9>faifjC-OXF z*uR{*07rovfX>N0!&nXhj|65`P$sbN7d+3{W=^AjfW4>l3~*qO@c|B>#WN-HT1EST zOV6gxz~Nufz8czkK6L^1UO-v$eXVjjnG;Kg_y9*PrhI{yQU-Hm(R#`U4qnYYVDEMG z2XkcT2A-F>aueeW9Jz%)YoMH4=|5okHu4vE2l+FXhDWF)aQrUD12}Rw`809w9`XSW z-cLTjr4R5NYIJwZOr*&IaHpP;gN1bnP4o?C<4y;4p9;nAwhY?8x)LAz<(Jv;#N>?3+V9 zW@={zaC8US0UVg6o$^jRx1)BB0geJkfkQjNe9h(eFKA}~I5MC10s9wdr*~)a10DnH z-;L*iy}R@LE|j|m&jT0j$@9R0FY^2sxZh8?z`-w5E^zEC^xv+$e+c~t90G0vj(?qc z&g1uQ(BHt!VYCZaUPQb4$md(M3)uT@`V-iH1kcZ>-oO>Wk?)c}FmojNFW^3K5ZJqz z{N?-6Jii;|eV^xnOP5lAVDAs8|L)xXA@v82{fPPlNB)cY@4>u2p67uhK*8ZlPbL37 z>Gu`n4|Gl=e_-DkdjvgN@@IuBHIKGzourKwznE3!4xrF%u?7xipupjwc!F&LY0!M(uS27R3NPn-> z&QhRrm3AHg4g<^kQ~r9!4>$-M1v#`D0Xz)|1^fnz*>DED9Dd0^&co(B#AgTuK03iStOUgLS-2rzRv?RcH% zflGnw1a4wJEMngLmH7Z12Mz-V{zg9E=RKbP zHopTG0SDjbd0_bi$~}U5ZKhmc?-uF_8~{4s;r9>82RH&;4P5jQ>jIehC+p(7-2a$$ z0W5z)zXC^r<$ve+$N3NP0Uiuoq>VES90vy9 zqn_BLF9Pqj*1RoMV84 zzzx80;1=L$%sBl=Q$E7|b->;P_4z*eq|E=t(S3$DQe6!Iz6raFx6gL(E)da8AG+zH z8$ubn>7vF75QgqhMQA30Frhj?K%tth00GlQXr>R{)C>^7Kmwwh_VK;SulJeJnRCxM zcPOhJNujHyx36j}STLQjKIdg%KG35xGvla;D!_+ZnfzZ!#`vRfKY_9mj)wz&( zRye|(9Y>hc!V82dGd37X`Z!WvrVNX!msLhbsb?{BVsmkIv9z>4TIV~gf3)~ztj{u= zOqSEfG2)gN&v*s#EU&2VUU^s6Hw%t1UPWEMbADBIv9$UEVVGgf1wwqR^Xn`S`WUZg z9<5`Rj+1YF`^5T&>TMlwtls0zmEA0DVtvM&TAwL9PY};uMw=}Va;9C@KT&_IGT*}b zjJLG@N&5Mv^_j6?&d#dyTbVDz*3PrMt@9^4zk_&Y9Adnq{r$alSY~x6ud~h}hMm=Y zisNo|v$nhYkpx&eX?Y?hhu1tDD(T;{Tw&qt(X-2U$BtKYx_B zS6!_B&V9z{SaqGnlz@2iZK+I_Ie8 z9P707Nv%4g^Yod?&n_kx$j5M@`;Co@?CYP^eX)E@F41S}_)>kIt8b<(U#3sSm+SL9 zaW(5O{hM`KUZFl#8K3XbiHtv(6gx_r$aOzWBe28xhasL-`qfr0>hbF=gdr^|Jb3`7U?-iMm++ zRR1hB-3M&4ttQWB)?>y#*1s?}hA++MZ}PC01qWIGO5Q8fF=`I1d~0s3v&rZ?`L1*w z`&j=;KKc=)(0P@07Z?*VHhfxhgweP$q2p?K*u&hns5e+{8xz{Dk%wJuu#X;O2tzEh z?OLz1iw*YC^V4Cd<&-g@G+H!R?6})-B2HOaU1J{`6kw7x|wzVE>G8(P;I$|PZu-YTAjB# z-bS4*w;dA-Mmw4RZSpW$Dc#j~`xcsAMb5BVzUVa7Tu`{?Tq_3o=L zW*lUFKYiUP|Ni=7=>UDP#`Zz^4;0U&$9FZc@f+*2bf|joa_lcKdKnJaAM>NEf44f1 z)*s7ku*vp&t_O<2l>Se`;+S=}u_Z;y|SYwR^zAvyq3*w#7nAGcVdZ*x9@HnZmN)8;<(tI+Q=Xf}GrvVV zLtXqs*11(Yn;c=~ANJ#6`R^SQ`dNNpjK9Zl+;E>i;`ImB&w`^YJv1gHkLrg5j2||K z$K-p&9GE<64y?0n$oh|oZ#g7ClgHJ=8ap3X&r|AQ&YaQH_Vo$-`iy;L`C0qGMs6RT zbo{*ggUJi-SEetzU!PLvOXkV&sywW{VV+OR^OkwC{I+?rVEl~w-Z3A>|5i`S5%oMP z-)G`kRu>ijn!HPkXY5N08!eZS z|8;X-R(>`)#OiWmL+K6YmmeE?*<{Xm1?S&1Po}K1!6u`(tiPglnXtwxo2)bbm)BSF zX?-RfW|bX<*V)U0LyTf|y{%sMFy$a?^eIR0p_nU^mB)sDRyo2tJKuG`tYRI;9A?Un ze|w$1EI7z$Rr%hNhuzFLz?`FuR-P))LP;J3cZ$_Apz=JXu~>{~zm%-Hg|_zsxsK?|<~Ou{pQgR39uhH~;^d z?-uef-AW!dwvp!(`4aUo-a$Q#b`<}q^>-4_s9QWUwl}TAZq|3TE}LvJ-rYPtv(6s& zm(@MZgH3jPZr#1C!^Ym~Y8~%moiE%+`^v{?fB6^=FkhxjzH~hYSUu1`H(#0Ouf(%p zosAxSf9?FQ_08xv^0UV18^;IfgUv(q!Q?P~jH>@g^{_Zf->e*MAHL-=_JPT<_JPF- z>i^CSy^P^|R<#fADAMOV!UNhgiByo}c+M`Q`Gk^%wOX z5DU@Y)c^r8LpRyO?o)T|C3mk*^Sm^ z?Iv}O6L*_BnX=J3zTLho=+FFj*cWDZ>5HYi#kX1aUhyp0VD&!vy(e0EKz^1Q@-t&; zf;|6J53`5V!x}xB66!3o$v!q7_Bx|SUL*~Tt z!}y3c3G_mcaZ(aY}hnbu=>%VGC9<5%@N%luhk_L_Mzyl$Se_4$T*GH0EYLi`-_ zWrel3#WR0je7k%fif8tb`Lp_8^O~!_Pt1$ePtA+TXXZ7}`OnRZ$rt9u=u7jOZyok9 z|5_d<-^jC&V^&!D*1D|GS04KRcYU$WI>S%uTtppgz2{vI%{l%=u zf+K9Q^@p1lM!`zrNXGe#+mlzkSEa+Rx{GExum+?9)jFys*2}fCFvaC8dz=B~p zaZ8U2-7K@tlx@rFn-%69q$i?6w1Pg_!wMU$vVBGIOxfTN3;NKE_l4BY5_6W>5j($} z`dQ*I%Zyiceg*r*5(}2uxr+6evC0wFSYFk7>}SE&-!X(0&2u%^v&NL8%$TgM9}cj| zu!jCu(m!L?nXqk5*RjHkgRC>^bUk|*V*Rtk_O;Z>lobxK&eGcA*~?^Q{jJ!9FId>YvTk^}n9DHTBOLN0>8SUp)I*>lDw%TH-fQ$GYN~bA-`)=2f;1 z``_%ZWKMW80?c zV};pv=GZ!Bw3&HtFFzYQ$iuL+Jexb-O`eu}$-`(LdAfL@JZ%0-9wt5VY@sgpF#fgu zXdSbCOZk3dKUm=q>n#1!_3ULh$bK+l$5yUqmHEN;gCRBFt0)_U zV}nh$cguT;IkL(j)-E;2UEMF2iD!)itkleb`4#52o4T$vXO`Gt!nWOA$Ar;Uu49=a zY%tzKzN@Xr>NV3a`F4h?K)yLn>krmciXWM?_Sz-M)^|ZX*9QPM@hq*A^sc)9qc7VFr z#c+>0nRA56Kh=4lIv-Le8xO0K1>;{?=MncGYmd7BSg^B4zmLho+SBs1j-Qq1*Iv)% zVe>h8Sbg5QztR7T`eo%M{jy+kkmHy2%Yxxx$FJy@IqOV@)qjXQ>}Jk>mS43llh>^K zTi3nrJnL+-^oDg06~{hiEZAg6s-E}!^Tzmn|Gc%1NA!D``Ft#%5eJ!YgjJRfH=j?{ z&Cql`W7b))XdQp%zCA)b6E@k$%IEfjIY(Lk!g@!lmpzQXG#@rN%;YQkd6c~DV*0i9 zSpCL72h129?Rs{z!5T}W{=LF7+mEq6D=dF&zu4d~bC!C=e`kHBtTXyieQdJpcjErf zTv`9g9GHa#L+7#Lf96}3nf!dgFwA&?1w+Sit{<~t=wEg0I5r*B2@zD6a%bHS}=^V;TPq*PSpRD1w%i*5fMgM^2;urCs}v8^_k7E zKJ%H@uUdb$^_k4EKC^b~pX~Ly)@Ny+d`vmSjH9fxv^ z9AdnT_5P?Xb~9s*4UVw1to&!`gT1V>!8+T{Ru>aimNQ3YEEp|set&X3Q&u?08pApA zvx{K`^)lu#6P7Y@>|t2Zd>C`Ib)MxvyPkb)GG}Qe`*5yxSYed|th33So#%;Xm8Dq! zOc$8^?Hkh&PLa(o)f2QnblSS)zRsCP2PNpn4$grCJ`>oF| zhSjadI!9SvL%zRQp93tfDgI)4*v+KVJX*(WyTp3zX0(=gmO09VotLVYy-YdCjBS6l z9=q9KKXVSVVEbkAwq9pg+v|)t%$OaQ%gbJ-9Au4cHF?>^2K$(Ei19l1^>40YFEi$B zw9a4Q`gNUWvYzv-vg1niu*xQfSzh0MT;)9b7;j+Sth4iK=b15-?H|i5U!zX;GuqJo zz_5`zueB~~taFq(lL7m)v3W9O>zR}WyBTj{{;gxST`&Kp?l)!}WW1UC=?3d;?*3(s z4aQrT>y6@>GT{(2mTnTqUe=kj!H%1M9*QmX&*+!>XTtavb7UW@ELdY_-F3`ZaD-ti z{r}zd>}QEh#_YOPJ*=_9QC6AUrthuw&*t{}zujDR&_Cmy_0Mv*{{NvKR#@R68ysb} zt2}qOZZ~;Y*+4HZG+D5DG!sq+ zCl6~JX6bnQ(XbBtSUSOeFlOh2=ET-BLj^~eoM=D(DIfb;Imv!7W!FRI&l;N?WuNy`6ZUDnQ#hiN9ykn31u z#!=?y%JaDM7s|utMe;l$@5T0yb=H|&V*j2Lce(vz@i+U&#ufJODeGTl|CqDE^lEc| z+I0iw&YVLmt~2*%%X#YYUy<)2{j$j+mL8UW*z4?NojG%Myef{ZXS%A7$j_YdYw|xTKjUZB z$>KTvzApZG{j&P1`;6gj`~8M_v7fbf>^GBl?f0A3WsP-?vhr{H{g(Lm>^IBr+wXrl z{=j}SjEHCYk$5)PS;+ITby@zebs2wR-M7`te%3y-E=!+V_Z{`Ij~R0|8NF-$FWe`r zvcd4Bx&2#xOj~|!ZcM*%zrCmLQTdpBuP!z@%K8ubc;ERStlb zHo|z4{Qpr8`&qEbbh7;aHQy=nv&vBh-vOO`;(89Sh}uHVWI$pdj#9zt3gfG}AXL^}=u~c)vEvU{b&4ab8&A&}u*XWnw zTIZQEj8`YSnGKjbqwCy<6V%HJzIDD%k#SSiRq2z!qQFVGD#lxuzIt2mTr+} zvN_e|XYKFSW#v}uPEjAb*|^iXjP4RYReyIo&-5PWS-#i#NS^!D!v+Ui$M-uwO+62a zXX9bSz5q_0M$vdHc$YLo8mjud~FxVqcjK z+gCQ;w6C+(Ti92o@7Y%tpV-$q=J2_FW$~qbW%8AM^-N~;jeTW#!SP|Fb=)@IXHD(n zwDF;ziQgh=GM+I$bj=etXMCu&oHsslPUwG6r8sJk0Wng^f20Byl0La``+93UpVd>@6WXI(XRt~ zk9B+~l^nB|(U#+V4#Y7#7PbB_$N%pc(VWAMD_f~&F>_@f3l>bb9`AD?>R`t5Hu5lG z`x35Wg((M^vB?@cmz0lHHaN_jrKOzTR$Yt|{jtetY5BI(AB*kv$LbE&>2TeS`eU-Q z{+O|C8S%SVpV_YRFy2j`W!1U6JWMO{u(Hp1|IRc2eeEAh``JHcY_d4e{w?qIU)euK z9A$~g3hLzmOFihw3|)|1f>C%Di=ag#B3A zbw`>L)1%~Je6&2P@OSou;aK~@Dmz!TJ~LL1b6>Jxc{O>CR}UL(vV4NRRu{*NRgN%Y zyoP;ZAM4E7V8@!SJ5fDMILr#;PRA!X&jyEE$5rRoa?Cy^C#$n{%+9qP|K7T6aD?e8 z>Rd-Zr>c`xHrZs?y4E?({MqCv>!+LldggwH`7>i!-*tWF&-hH|S!3A1dVdhl1_zk_ zQ63gd%8t)c59?>Eht+e`v!UyK@V%eqKdXnu`Rduo`3uy;21i*i+1UMYk^C$%Y@!Z! zv(!J{-=m2Ci~Dm^@vN|Zsr!@Vzq&s+)92;xPbM|@XUi+x=bP)BDXUkyZx~(WzUfj2 zdl_AApIB$d7V=$VKUn85lWXnAmeyq-n*;WP*>(2gm+HJh|I9hk@<#n{CC|SX-9b(z1Q&K<1#k~&!*RwuJJ)w!em@2IoozwHxqCOf(Q zefz{RThD4YKd?`%eQ2MUH|^6d@_uEXSRS=ctbS{sy1o9LePZ#0ePZb+`?Ra`Oj-6N z%tI{s_WaUr>hcX{y-a*TMb5^A387)6ZM-uh>{kNyvIKkTP}793_cK>xo|*Ma(H&ek*R z$*=U^^wyNORdZBS952B<=@JGnYlB)+}s)0%>7XPvc~vIb7$i!b5F%z zWA3cqVD5*R+fC-q=w@?ge2cjs?)p2-o#9?{XTnj&|CHwl=O2=XrHAD?(($9#Wx-nO zct|`mCP%scaq&!_6o0gNJY`+hpS3Q_Z&>#j*S%$3mi}d3HVgaRE6)e^o8^z}H!J@! z_urW(Q#Luo%BSXjth}F@JJZk2o$(juew@5tnmcn2GyTfkk9YoSb7zwU^KZ=k1bIfy zo%wg>&g2JkKaoG0JHt=r&fFJ}be&|*z8$8e{kKmHDdV}m z#E_+V6McTeb?jxb@WhZaTV!JBI9;AaCwdRadP`3X!^}D+hWHHi_$9nPhGi%Edj{9D zvrpV|6GO(z@)N@dODjwa>;Dhtxw8IQTSfmYR@MI>onKx5 zE!WWhSUgmWP{akAg zvw6C?vv#JrU#b5;IM3uP=UF+=`K!cTWItHH*nY5dsdcaBW!7bbIdgVgqmIk1%Sz3< z%o$&+{_Est<9hj--{Aa!`P}F{D>pgMl|g{1N+gCXT*N6!SJBIKC~Zfd~82h z`j7qir+Kl$@_+3I^H1%^LynvFqvdDzgEh84?Dfy>2a_+vv&rZY{eLN*$yef89u@zn z{NIUZ`n`CDAH_eWUiL8kNj%HG61i>2d3LenXQgUvaI|&oXM2*z#re&*ftKSYd7och z3r_Mrzj?5~b=)?|_dv^Zj5$ZxWcev^> zFIacRq|nRy%t`*=-hDS~Qs{WmF{?~wPYT0qGJeVRbNs*~n=D$#?fQRNopbfij3cbG z{EB(7pUFJ^w~pC0?0CNZnRAroh4lZbdN{!H!uo&BF}qn_ME|X0w!N8EtHiEZF&P^Z%teGUqTG zTZw?5K%;7=(v-wZ`w|rRtUz+oy`e!nv|F6V7u7BoF=%3M3`u|!y zD~z7eKdX$sF+cV&Jga}^Y#((#*FT%j>7V5n^#85;UerGu%$c*}JN3V$e^y`CKTEIZ z|9kxo>!0OU_0QUy`v1Z4Tl#15uKpR`*Z+^skLaJthx%vbBmMuM*FV-jbB?n7ss4Xb z=NI~C?MwX!pXvHW|15u}e>T3?|Id6D><9g`_(}hae>T|<8TrhZpR4O(I(D*umwJ7{ z$)SA#pGBK6IixM8Ob$aVrcVx~F=Ik)_GF)b7~}h#Ci^=j*Uk4m@?(AWYoWV(L&mVgWPc?mKU<$YEiX0M_d2O>xyk-r z>@#>POb%V+t+V1}|4w#(rO9EGIm;8|kHxcCSv;$&h@Yr#)>z{xi`6FkyDRxtpB((Y zsn6F>_GcJzos+#!Zw_lu4s}-6ne2O<9Irb$R9IPWau{T^zIjb`%pT@!(3e4l_Q;$! zG%uDnHZK;No7Xhww>Gbq+n5&%c1%|XtMng^Fw7cbpN%bVXI@OUH!r3;nAc3#v6t1I z`8Wu)eFlSlV4*v$==9*xXZJ414Koj_dp=YlWerFV-2gb02*%*-u|A_Se^3 z=MT_V%irjW@nQO!r_R&V!}{s!X?cad=BxKgeKB0EFUHsEYaw&HPG5|!*B8T$`tljz zlo<<-uzZt!^;zKbX8X!2o2}zp#Q(ya>f#yy-MWl!wQk9Nu*&o{>$3U}>nx3zxpG`j**lZoQO$l8qczxOw&l z`jjxrY{nFSX0|Q|n9LNzwI|EzDJ|8*R1rGKW|=zm@HY^Q&w zthbJL(EobsVuj@$_0M=G{jaZ%otbJo=41)r6F@<@`O2VZaz<$BlBm?k=5tTu}i(Ln}L8e{WB``|4VVKFykPbZ|i?6@$cxL<#+YZgzZ~sX_=s!^>`&j)@|BOG<|90lVjP;N8&*VS)-`=|H zXZBzHv-XMj9bEsZ{A@PmXZbVvcT@-an6qH%3-LS2%dF*B@-zBIojbekJ9$|BULMAN z<|E$4`7u*NAH&$GpOCCSPn;S?Sg^dS>n2SN{mf=e_1RVH&zu^%c5{B# zRPXVaKO4;1zB}hm^h}jOH2*Dd&s}k)Q~fDObs1-n#VHoFf2DU z47XfCoqL(bit1!oNgg)Y`Ywbtp6WYb&4aa;D^K;ly!BR{>V0|ZbAah)Q@tzC1fj@QCuKf`wNA7FiUv$nnb ztz)(ws18vWS!rd8^ba7mszjvO|4;v?`r)5<= zzm@Oz>S4km#;2&~P<5X!5A#0tFg{m3spGnO*tkqgT)v2<0x~P7ysnz%R8PD zd0$@ssgd{P)e}YDm$&}Hk@w~Gy-4JJdGWu9ye}_q(a8Jq`dU2lzPz|4Bk#-m=VqzM z`||2pI`Y1}c$SV;S4ZT1d3l$Kye}_q*~t6y;+BiNFYi2uSzbQ!zI^L`j=V4LIu;B$(W~^=)g%L&@MWK9(>o$u*KO;6-V%MqSw}^bdl04f)J~yD=ZKIH! zrmn=?nQUk7r>kQJb7#7vxijo+?q@j93Tqr-*hM`pyVcWYJ!UL%gta}?bEbawQ4gyJ zs)x}*>iL6wzf}*L9A))T_54visr*bC&QkYb>S6P6^{{%Rdd_w|D=dz3A22?~eefs8 z>}J+0KTF4{hdJYO#GPQ@nV;l7X&v{;m$@#pKI1=IpH+rGTbJDo=USgNww-H^=UJaQ z2U$Dc`sbPB1@1$ZE|j0;e)r+|uD{fM$mnwSVe7c&KD}7U^`-hcl+z%Ie{aW_} zV-B-Ep#F>Gz0v)^aFhA7&dz@I-faGC)b-0^(ER`6_yO~0=|S^non03@|A_f8JZ1hZ zpVt2+>g52_T>qDfdr|+)U(!G8FYEuW)_X<&Y`&_0Mz87rGV6Y*e}-@L-#Q-E|K;ZM zt^S#XY2K$drv;~ZpI*MUY2K$7KYp6`#JxUwn)m6=VcInB)9Zh_cqV>`ymW>7=7?vF z4W{kWLi?4@&zPu3 zZ~e{G$*fDAtZbpq8=T)#ovi*+oh;aKqq%M^Kf^Ziv%>f$uWu_qYun4;I^Id0H;da- zoviMwPFD64e~W$EUpygxE7{7es$pE={d%kx|Dtgv8lsQkC;>qz-o?UkSD z@22@KOMM|OnTcoh0`YfP zhZ!3e%Fp^m^51C={o+}@SUd}M4a$3|{4D+^KTFrhf0y`cr};_Z_K|gN*K#r*-^*b?;S2!+cnI*nTm6WLk*s)9<6!VfAr!vB9C% z`6pe^;wkgJ-`trpJS{&Xh6nV`F6PV_KV$yP*!oVQf<3G~t6o-edD&$7L3O`if7pD< z{le%~_t!rizpfwF-_k#8|FYggUN2n7Ci|JaZH}zGqmGBgy>E`pKCljp5&b$3E=HEdge@~P50csePET9>C-*8FP?EO&y4Aw+n0w0qnXn^w{H&2 z*x(4uv!?&=JDYO$KQHg>>7I2rM|Qm+&z$L@#*Cva=Zb&Py6k5(PdpQLy(BL)mKKtq z)rG~sY|e{_XZ0828JEPr;{2lWv$UA}%$W?^zs0AA0VYdK|KE2;)!F^3K9`&x>P(iN z9@<`$XPN1t!s@c-%f@o%`?|c#n=h*?nlH;MneQ8}XUbA+zHGAersGx2m(4ZImszLz zzNMaZ%$JP~%$M;N=KC+NZ)LuWwlQDE+nH~nZf4AO)ju1%>Hlr@?5=-Cd+49>p89`B ze%4soOaCmGyeoch{j<8S{{JoS0s3!wp#GWm=>I)+|3?4Jeye}xhwA@*>l~(krbp|a z`7!$cKs_hwpT)`gXLg$YN5uV6|18eZf6H_9|DpBI)jyN-_0Q}={eL81zy6tCqJLJe z(*MVf2lUT+UH{DQ(*J*~dyoDZJ)nPOZtbbO9_5Z1Py{>;| zZ|a}Xzx3af=Ux3X9??JZ5B2|<<8SrP2J1||lb_l5>iJyU59(+0NBuJT$$Y+Wo)tF2 zj4;5;&t~|ngyS(YLO1JUXZXy7>)2$}HY0R?WxWYALY2v+8DWU&lo=uV+FaPva_WrG zU^AK#+P{%!+6B=*_x8`^i^?dKT)n@3GtIr4{Y_2&Ylz)(a-5H^u*?Q_{Y5f_Y>qmK6 zV`YOGVU(df!*}*tmjlcgeiFZ-Ik3(;D;v!SZNYcNY&;`WSmhuyMnCgiG@Iz3bv9Vp zRR2F0znT773iSi$4zO1C?%kmNCJ4rsaz8kK>VWuoi*4L5d%*s)&W7O-q*732fo5B<93mYt0 zI?;7g)xio2)>t~pzA>Kkw`^L(d_HDX)Gx1E%vp#b+8J}4g2Q1^vdjG zlLc#U+P8(&`IdcP`Y-#&+S~SRVe@-O|7^0!_+9-kqV9j|pYe$PnSZ4JU#NovjQ*ql z6921zRzKB0>rMSH>blSL&n5?1|6Koz>HjPJv-FMr+4xcai|c2~%#bpjH!}>eywJ>0 zT0;E7Gea*ci_Q!=%S+GnJx1oY>`c$$TYsgQp2HWn@=Wh(JHP78(8pr6nZ8>{-D}JY zol85v*36K#TxVt&VOV!&D0irf{Y=-J>F=u5wf@Y|wT$anV|9a>zPCs}<(bx$mjkS9 zIMaJ)@@zEI_ZGRH_15vmGktH7ePipp2h&Yv`rabfFto6hvTMe=Pn({~0r&-NAV z$L9KH(xrb!Tj+l!^=zqsHkdPPrT}^+Y_A zoy4!oUBol(7SC)~@vFIhH}hg+ck^O%PxD$`-FurCi;8)%w6A%sAunqz?`K{tn5=0| z`bW?^`5RCR@r31&h_Ov!+codQ0w?i^WVUAXPH0Kv#rna zpR7|>$2sz`U_Wb_`LXSBPhJrS)05 z%KDo)&t4{1+kY19*wpbg>Sc7T{bY^tX6hWUpNy_k4|8^IuHWnJCu=u~Zynz(zRUjn zT|A>(#j{}N7UFIb&&uugpXopB|CZ{#L%*yJnjgz|oA)nWcaM3qdY^eSzu$bevL6le zX7-?c+gjcKv~R3FWZxJbHqUJwKk9tTC!BBjr1RUV<7xK=^XJTw;dygR9KT?Wj9(Pb zSapGBR6TiFbCyHl-LyRWP3T=C+lU>YMV~wM%Pni{x zJ=Gn}3InW9o8|AM#Z8|Tx|zuB_R$C9ef4pWc&5xa#5zj{t7AXwvc{YZ zb{wM5{bz+LGY+$SfIPpI_dt1A`nCKl4tD-fafdk1^ceYB?v+0k_dEI7V9w-N`46++ zaq=@dRepxk&F65QWj;*Lp5@=U^89I5=sd#tY*xrv;Rv&H&G|^j=bJNY7nnceOYF;0 z&i~cEFkIojVvWhs?t?4cC#+oUJ~>AIYs{C?wdULMI`i#y{s!?ZIKalu;(sUqE%LKq zjm^6JO#d$aSnJ&?o>hk99N#8CYpk<$yZpzigB8~PVgJ}*bb|Oh>>o>Pu*&umoxjt3 zS?3U|gXVjZ>)FfrF7s_2v!m+zyUmyJJ?6{mz2Aj4(e~{9nZ#BA!(i%-DIEKAE!4 z0T!w2m>xFU?nxwI9#+|P%~=0_(s#=X)!*_W zeP8RkYxK?dTJvr>VBQ1zzs|f_zTUi9xxu`zlkZ0JX1LkBS-Mr8>s|MNJWL)jKNc+A z;PofvVa7p5Puc$)@pJSd4|`dkJjb(I zu8-z~j@!JxnDeYNZyhh;{O#&tmBo^C!Y~_4&k6BAyxuV<^s%}6oKUc`#+=Z3hu7B> z&r+v&Mr(<`(>&J}&k74R*g0tZjpu}{;8l-*^6qIiBAacjO$;@2m5uIiBBFKkE#?o8$R? z`B-7;#5tbd*Ds@o)6eN zbvD`dqB@?lUyL}&l+jD(!5(I8u*UY6t;>`R4zc{a`Me@8ds$)5lpVwJu*%{^`^6?( z-=W=j$^FWl1)J=AO}#Ih7aJU5c*VS4R|or99o9b^jNfqmtJY=xHS4nWrgh)c&s)}I zg~Kcr;@^^oeJnn(F3TTT_g~I4W9dKEW%a+-E!6Xgbs2qXUFPh1TYS^p+4$VtTgP9R z`#WCe0MoC{{aty!F?W`}HFp-@JO6L<`oVdIADwR<|DXMM&-LtLv_QLOH|6Ij<1y`? z<5%y3?VjV;*R*!e@jIT~?m2#W7Han#zxajQJ;$$}McO^bZ?3R+zibNuoz-|oFJ*R9a*IevNBV6tMn z=lJDe%Gyfpp5x~#?VjWJ`l{`o zoY==Ei`Mb_`u|Lx4fIccQ1cwWJR9r(bMT`;vYGzBun(K-pQSGSGiLIoIyk^8 z!&lC4p?@~E(m%7U_5ZcFZS>E0Tm7@l=o@vhhtYQWXM^pd;w~?lFlWk+AIxnhb7QqzeumxU|4|%!8TJs*(w^e~&vmS_ zx|evC_7?w>JnUoG$DCPXr|*8x4>V_1e$s}_srq2X9(J>Ky8c-`L;o|)r%(UPILPQs{m*jD9!7uAKdWq? ztzM=~|EPb4v-Cg5>+EGXTmOvN(XKC6S#X%?pY%Uh9`><%j{aH6^gmBuf7U-69AR>< z{^#q5RYvEjhbhBC);V8aY_iVk1^Qapbr5FNyqZr8tbpw*Osr_*QKocmVIT)u(Uq^rGHin{kM+a)_;d} zSYh&x{u%#U|I6r;JuH2of2Je)UsnDP#k2a6cqSiq5Keh|-`?XkHqWs^e;KZ#$NVQ%PU z>1T68&dTJues|gWQ|9{JW#^~P^(>=0*|w_u(Ol0ox^CLsFxWa~w3__vVR`yo&+v<9 z`|9${nCn@7bKwwkwyhyXCH}tb$lS!9(ZfOo%%?-7dTbmb4 z+s^gfvHIN2JlNbr{;gcUul#Hr=sfdZi)ZB^@mtGxusnZ>QfIooDS{@vPiuf3|o2e&<z~sXE6?kfDU)5iK5Pz5 zUpJp_eZFBnEdR^=nZ50wuU+*w;{IZz>3(MT+WoMb`TXF1Xc^{t-=9C9=Y4}|QIJZvl`zM`KF@ho$QrDd$UkNoUqeOd8LmlMCQI+hpD(hBl3XS|;|tvE0A zFp*!|pBHM(ILffb zJm0-z|2V*~=DgsCpZspgJn!>!t$F|Z&5mg8dEV!jhZWX2$guuA@AF%C!+GB4*Ebss z8_9E!cy=@0SRUqVJJ?({m4^)uGTKa@L&SHP)5$1lBdXF-9R*n|Wf}KZO=NRj<)~kLN$EyDreVw3w7ALBo)sxlV z%QMx_@*maD(pl>Nox0DKhZPPo`jb4zI%YSE^XxBc7h0d;BKeL}CsWogmXFaTt~=g3 z>}HcSRxZ^KOMlhR3C^>h@%8Q_W;f{PM0uF8&Jjj8ia*Kv>}R-HJS%ngUDfNi%FpU; z?mLFt-FGLuo&#*$>wfvYIX|i%mL7Azuz1}4a*BO_#{I&^v(B^pocL3ne_kHuFUrsC zW$Uswte(^4c}<-RZtpBpet+wYn;!;RE;v7ou+H+ItiygrZSt_hu5-+BygY17kcT;w%R!c`SX1az&sb8?{fgIUu6FOe&Z(o#eAOw zaNVNweGWkU;`4nDK)p-M_c;LdaD>TH|IgaFz)Ll~|9{PNYECIlbi=vq(>dy-WGG3I zDdBWu5OS+Q>PQ$$C1gZIVMug0kuD@dB9sP|Q;I3kQDNvJl}v<#(*OO;-s{ZjeEatM z^7^k=pFU^Se%5C{&$HHE`!aj>CjLqI-;DU-P%-htakykM_RWbOj=~ji3Jy%+^GU=H zr(iXe^|v5?I0eUG-^s*34Q@sJaH=iw!|_hUKb<^vre1g;^>FNF>YqdB7V-iIZY3{p z^mg(RrQIFm1r81-FL2~e@-l<`3?(mcWEgpY{r8cV=dp)Nh3`iP4vs?S1@s<32lhXR z4xEIonZ!K?9XRqhI&f?PIxphKGvpVJz~yjq68W7)|4b&ou$o4Gg|KxjeEj_a|KVUc{==y+@&8roH?m&X z`xE}dkw5T1#^*oLfsqr8jo`}!PJ3dj5Bs5qR2 z8^29nZp+~wf8rRLqsn0K$Q-WolkZVET<1sUksPk`lb>-pT<3?M%;7pe>wSv&;mFg( zznJ=Wh#w9tA%57ul=zp>{(a(y10NE<@N(jRkM+Y7;m8W&hr?yWzm(-Gi68cVO#E;R z4#e?$4e`U)C&d3gekO#1BW{BplvJ{AKvJ zoA}`byaSH!A^w%b@e}dG-o3;x{0s4~LKlw0zJ0_GM{DIWc4-ICfm020`EDA!Q*u>r zIMF7T-xjmHZLZ2*gTL)^RS=H0&sB?H)ghPfzG!z=t}2CN9dlJ2_MVfg8h^_2PPr-s z$2;e$GC0;HmuEcD{`_1u91fP`sstQ|i`NoQ*Icf_Q{O#TmBVq^w~lyv|c-neR8>gHTk0SDp07pw=K zBYX?-Z$Re`;)hdk3|51QzntYmh#yW3BYrq^7x8~dKJO-e*n1E03y&oJjp)N6I1na& z*guN+H_`qf;)i2nh#!tVO8lE?_YCpFA-Ej2o+bXTsGmgqa0pJqktxLgH9Bwv_D&~$ zI6Ra1zv1(2;)hdni64%=O#DgKKcD#F(5u7`$700)E&jbm{BR0Z-(mMU@xv)NCZ87& z{}$E@2Vvhr;+M~F6aQB7@(%ICp?8TN_AMd)ZTP>G_+jt+#1BV5ApY&-^F!i?y>J5d zFC+f%v0qO7aB>Cl!-+ED-$A>T#1BW-5I-DTNBlovzn=JE?-#@mt8(JsNglo=emJ?2 z_~Fzh;@`!3HWNP_fGgnSSH!;?|Gy!A*h&)rkJNuh{K8v^AC7G!{yq5fJ@LcbRI3)j zaoGD4%l8pK9Q}>>VgH}RpTdv7h#!vaCw}>SfcW>~|KG$9Cl3-ooKShZrX=W=+(?q?e>w6T6sM4i};SpQ#;^5ojm3LmG*V>)I`{- zm#0#2_?SFZ@*8=opNEh1Q-eHJ0Vf*f@y=9qkIz$cV3m`nxX{eC#XPR()4m{&XI|32 zNuJ971AhziR1glGn5P!W=cd&Ei9f~I!{L+Afg>%^ht95~zx{r!AC1$#Jn zD*AA=H980IzYY4Z)fRpE+&+)z>66b6d8$QyYJ8K?e@Qi{NM&ojT-U6gqJ5QFP$gSagoY4>$zJ$B{>v3wB(Cgr6dh za9{#?l+TmNqZd6m467;V!=Y*D*TWCE6!u2Zhr@8=V~BeOe!@{W3J0Di|MD639ZUNc z$Up3xN&eyRi>#+U%V*&i9G-(;aBwc`X+Zr;=)r+ki5s?FBkqRSEg)_<0xyDnZ(!dD zzu;0h1;^poLi}os-ka#dakvb&-XiYf(0`Y>;ouVDhC@s7=Xm0c6E_@Mj{mU#6Z}7c z_G|GU4t_#6>zzI0^Bl?AW z-h(~t{|S9K3R^`ihkL`RpUJcEKJt7bas5r6;gCCDEt1c$w<+tXmaj@-m6gx6P}+O) zRpVy(b5uU>1SX%g^Hmw_t&`8aC)mUCPP%wqbl_0E{Pa8Ml5k0LJ|COUxe|5_^0|+n z&yDia@0<%BpU-{#=;!2fu0-6q`P|3P^1OWR<0sDieD32%*FqnT7N8F&3-R-0?3?1J za5L=TU@`WsXxAKj*xv$s*xM5OQ)ma5!qHa51;>2E%epPQK2-zaiuc_T5Ro z+Od2n`GS4J(1+vj4*5JB{WIym`_YFJu)Ool_bB@9>Bq6?!@+Ur!>Py7?||KS^o6IO ze-`U|9(_3aBKd{=v&iq+_%oaQ!v2@YFPwzE9kG9z{KCF@-mtlVDBMLcHh03F3tlpW!DQTTi?fkgqR@7Y>#a zFRV5aZwbpc5igwhnt0*Bx7c^3AGTl*C%0n{s~y;PgMYvtj>2)+zmxoQ#|{p`(Orx` z*uR_c*MraSaM;?z_!CYs{(7=rI1C5&G5+B2ujKJU^8Pz{gd=bg_WnU0FQR|p2psy8 zJj!R-AHbi#$QSH|Q*Z<>xtP!U$rl`fE9CP5@^uO8`x_lNt}Im!d)*e-?a{4faowIc zs#{#Q$Isdp*X^mVV{zRcyQ3|x+fxtwFT;O$A{?)4aowKv)U&v5kNzl`dzLq_xNc8=8d+SoN3XHPb$jyE40|{Qx4V+%#g-ZlN8m+p60U%K%~^gG z`GSYT33w44KbhsQI+f*D)6Z>L4hPO;IUH_p@!MJAILlI{a2Srmf%7fZ_*&Y-@{YUE z1(qs@nrPuG(Ncn<6xNSQ*na1u_!iD$|8K;oOsIDsQm7$@=>_TNnZPGg+F$?1#}*dHZtgUHVd#19AI z3OF{C_-`SOS;P;=VRbA1%qD&~G>7+`5i|4PsuMFT}yu9zK1zg9+Pq;U%YEut;;p_)# z2M6JB9dzXL(ddjuw=Ozx3Xa3EV_5G))E~=w;Y5Q1RR;SS7VxeZKEuP|$Z-WK0jJ>N zF?>G0fa~@wKcRr@_Qac0z_viVxdmLeXT5m^T(^g<0^q8i39eJWj##_ zxNeVb(*lm=$zyQ=*X_v@9EGjs1zfi$-jfQrZjU}Z2adETQ0j5wZixFa0zL-RtNlI07eN|Mm3KWa@9EpI}u=Kgs8t zh<^(H!ed}Hkoe)qZNxtnzi-DLw(elPa0G5Vjrzf?7mnP^dSNvZ|ECk*{rC@uM&Umk zei;9sqkatj!?8!vfxXY56Xi1;f~{xKfg`XrgZ5MKANEege>gr3|DR_)Gw}zuUc`Sm z0Q+7b53k@q%n!kM-ahfZhW|6sha<3WA^itOVE>EgzeWGS2{;8u7tw#SSkGem54M(| z1Bc;uv#DQ-4jd>$2TrX*XAb@F1v+pFPQjs#=*-3b8+71cl5rrPw=fP~V)-7%0qjjN z4qnDTxHoM5%s7DKaP~abyN^7=>NoNTM`3w4Xsm)f3jawx9Qcd+`FuV=JskL({K6sC zg!}YqpVfqGXZTmW3HRv}e+}wm{5tuC zL$IoYJ?w?O3($c}VLu#)gK*LI}Ct(l!+F}nUVQVp;&uF58Z~|Tgr`qA?64uuqKjBm-{Dgz&Hc{E{;a`^~yn_$D z^P6xTpU-d#4qniN>-ca<6RzWZ?BOsRnTHOXd<~tC$=3pO;K0HrDhd1FYQp{cEMG)^;PBhz2adl>e%27j zV)7##C(m%;L-PCy`fwBuEhEoxdzT&dHjs^!M)+&7Ia{1D?00G2M6I4ya@JwkIv_O z-iZ#Z_Miht;l^JOe~SKqQ+vrH?Ei&4ZeTfFD!dPSIPxp@<$Q)iu=N}Ea2U3}#Lt7| z9S*5NzT3gBMxiR+$of2mDhx;Aa^dVkp6SARY8I+7u&;KZO2V-^g{s|V@^*Bgion6T zg=z;J_7-x?#QN$Ls)=y$*g}4!#zssfHQF628S)^l8;nggfu z3l$F$SAGi}I9!MhY!#sc2Tw%jTk4xq4_nQshgEaxzoUIi^x??K=)>OA(BH!H*4V>A zxExOQAg-;frzdg2Q8)?50@QD#-Nl7m--jcV%D?H~s;>27OrdK_8C7)=upE7OLTJBuKx& z;T!0eUF11LzreBE@gMfyfzEFF5AF>|1{bOroPtxZZz%cr5nVV6`|c(muyqgl*u!Ud z4jdUtzJJ2s`>==oVeH{3oP^aV`ZxKPr@z3}%4u9d$ zWB3b)A1ANBupT%h`~-2s>Ph0>haNl}w#E}T9D$2}rT%GjU>>Ti%H=cc`;B}|rr+S; z6#Rsv)9~|mbm0h`g3Dnw9sLUO1((7W9D{vu1ssc#k3VQXgE-(goPfPAu>L=>gF|o% zj>BpudH;*$a0m{~A`UnXTl-l)n>gSgoPZ;6@d4I5hdAH}Tn_u@635@H7ajw9Um^}T z2Ddv%`SG1fTFL_z4H*lW#Z)m$pKu(W1BZ5zXISl~z8>p`=fDv-C7*w!Uyi{Zo(Ko_ zP!9*;65d&zf+KL~C;Aumr|9?k^EJ@dNvRp`LeKhu|`i*{oQ?R@nJN!HKa0HIS-U|AecV&m+5N!Q{ z4jh3k-kt6HlW`8m;RGE1i}ms@ZSQ{83&-JdIB)y`uGF;dG~hIRiq}u0e2Dm1njF7sS=C$;VA6S!XA#p{sMHX7pVwr)u0~s!G7M= z9f!xjeoqni^|PMrBIPSYA07k8;3Vv?S)_`Ku!qBNv{n(%??LA%>Q7`j9EM|XIc(KN zrzz`&$H3k?=)h69T{HUmX#5ebi$8Dz_80TnOI&aQPQjsi#MPX*;V2wDriknQd* zgg?h(4~Oex532^)x1b#?@BCH`v4`WZysJCjs7TF$Lye30juO9)D^k7TFdUQ5#}{$k zpU-d*_MK3~b${By-cyJ(r-PPEbsP?HX)y_@wr@UId4_5ciq%2V4q= zN{Jgz!j0P#*Gwd4OYZ z61Hwdrz81+hr{YNbl?zN%sbALa2O7R(1GKy&(Heqz#ljUm%+io=yamp5Om?j^2sX*Uub*dL}Ij=@$J{CI$RI0To$-iN6_k9JR@4=3OR9G{EM z`P9FJJsgG0VDHP=UqC+K;cyC0z`=R=Q$l+<4EtXpE;t7Jx{{CiKB^AjLEb@!=?d zN9p3?e=j%|WTk7k{ncgh@Q`Z_UuVq6tbvYA(p<)AzM_RWXe|G>Kf!1BZ-xN=0 zo74Y`%{(5pm_S==3X7iTdSTJfy65qC3Fr>AuGmmKSEh-d7Q$j<>Usy*`dZgdn=7!< zx;OB5e{d7fx_3Idk5MNj0>sADwW{dGXmdL@T36@cS?aXz0!R0stImF*7u5_Mt9>Q+=o_5;blli+H=m508W&BO?^n`nZexNTfb$jeQ zG`~r7ufj&_-p1eJ%Sc%3PQXTVC1>w~kH81O)Q$JB{e6x`G9IGU*p#r*y4(5tS5N`8 z?kkRNYq%rm2F?SfZY*`EZivP!4BgwX9|`UUTK7%<7JoOw+rVD18<@J~AG_1}t@O9L z?O$=aDL)^=|Ms9G(7H>}5ZzCx+X#LIyTF4zlzJM-_AtoF@4d5>It_bXFRltk^J?i0(yHTm|}r>wxL6ugdrhupORb=*sxLlQ#DN{-ul7?d*IW&4Q}ivA>FL z5Z!AHUD=Mu(Qf=mG0aoE(A z981yLg-nvu3|+~C-j=KC>O9EGs^-B~&FFg3eYcA4zsYZ~N**k9%MIOso8M3s-2l4# z4Bh`<`Q^2wsv%=skKg}a`Q^2wsx`W@Pm{8o&JvUXv)@?c`B+!*L6+}3%zkJX^)m19 zT1I)@d!_Ct)c-f#&smlPQoe_eq#Htah@t!c%ilPKE>(|;gXF=* z990`gX#gK?yEvnUlZQH75-@agux|-`KOL^q+Bs_vc9$%p3(FvjJ$lne2d^T>U z&01HHdzt%m@vy$5K|LV83J6Jxy9;#}0ew6ol|c?*_89FiXL(<6J&}^WIIiAg;+Exf;nv_I_OqYIb+tEilh|`x ztKt^a|Ip>OQFRbVsp&bq?(Zyu;$B?K| z60lQ5w?4i!0-^`$e7Z$fYw_3-HQ8vFgP|Z_8c>b~h4P1EV;#risrXy;rD*@(H0;xJ zgk#UoY}LE`U2L0I+AlHe|hRw+%~m96P~>j(uI~PLKtTK0o_Zn+$y}5h#J2 z;!6?x^q*O(MzEeehIQ_Qcd)~i;yqp|9`CKJ^e0pVb=AYL*ZM!Dov_!NZb}})hs(oE5o4UAa|1fX_Q!Oe zh}M2XC%p}z!z(flIhT-r8Gef0klVm{FK7Diu>4<>;|D=)Acbj2k@CuTlLYhlVb*)E zVV@pKtT%d?_41f7yW}f17ahy_sPiz!(2;FJ#zX8dI#Fjla66=uc~$2@=BM#!Oa?Lp zWbC&U9X8RosapotfscVaGE|>uslE2`Z}L;NKa9D(PPIZS-4}e`hW&2v6VSTdMHSt; zOr*y3a>`zV+Rq%aMobCF&Nf5(}0-o#$dlU4WMjj_RHDV~3_|MnKK-?IHkH|F)+ zss=kMUOS!M?-;sSZJixhCs@yyGB(2`Aa)|R%osW)*zhW3#cPk#`!3PB*LKi-6?MF> zxKcw_oT7al2BH}suQ2BMvv!`Pz2t=QAMu%a7JsDZ{eZtI+w);`hZ(x*`STy@M$q*g zMt7v4%aHuX{wC2q*U%k}-Z&63bm_9R?z7ZQ1yW|fN6Le@8T(m7cNX@N-&cT)W6AHW zB19e*QYXh>QsnsWNV-9E?=W;fq8+bHRvUrV9d1i`_E7f+NEhaV|E62OIPo(^N zP1bZxyF<=&$04&C6OCZ~s#YGjpy-B~tPiCKH=(fYRCA4`TXx*zF-KxiPPjf&a z^NQ%{qWw*v%hy7c^MmYvmZQ1S(Cy>suAxqs1b-jt{3V61br{{x4BhMgg|4?1=W&P8 z-C*d-btsDGpZM!Tm&>~Ukq2J8TAAMg|3Wu_?v;k_c{nEX4zFpg)E(&PUYT~={;vc& zQhvke-eTzLe&hA6mAbe53xA{Njy#MmuXnA~4gCwYlobO8<>uuev*3@+b=K=G$D~@Iyol^{*05(^HK7i2DI%{BxTl#5`p&^w% zz8L99dqPZ?!7L90c^#&nuWt~Cn5TGkq~bM@>U;1tFntMDxh@|#onv;xFJ23&4uWGi zH_-7-wRN*jrmh{})ryB`M$wF;b1wS?-LDs7!z&MYK4-?i@6+xnkDR}}2&Bw|i<+cw z4>@FiBFFNwpNgHqF(|sYlrC>!&ny2FuivYTyQ{s(!zFOF%v^t!in%n-TbM@(?%&UFV!bn)kz=SFEU^K=;!(vVhe*Kpn3y zQ4@gK_p5iZJ)yG!onqTo@p=-)D@7Eq_o(#mRMtoF?4XXushKlwKNJDxVYGC~yrJ-!#)y6#rgoe8vWvNEd^+ZBDF$G2R^>JF|4x<4*+ z{F3vxXQ+P(JO|`jP!PzmNa$qydl7}LoENIjN;O4SkISXl^C*7BWA`g{hdH_%soMsm z@OXUg&&cTN;~bf9`ePqK_e?{V$L6a8(w6gZ?eBw*u9vzKfRrNmNV?v0xsGAvfydq} z9#O9z1zPtpN0&!LtH;5MU<#=3<1s;P?EaaPvd53kgZ~1>?|W?@A(_*Fn%ODEy2;lmw|YrniDT+=~UZ{KHO&!cSB`9SOLaC9%E?ouG7H+-b+ z(1&h?p{uv=YYp8W9bMVJZ*+9!{Om}&L3C@?$&9Cgw3`mTHFV`WLW-vi7kK>O7Qkb@ zRPWi$>p-?|i;LSbrevNQigDyZ#t6E)=L_ zP^kpqaZi=`eBRda@F*bF6AS>?fcnH%1Z4XTIM;1t`_96FSb%xo@b90FpA_w{qP^~C z9t)%%0aHaEUnTvk9lvfMfU3toNoW71&o_b>bNpcFCTa6M*a@_5%F&g){Y<@- z|Lgmb7`hJ`x;&1D3tRLB(7KWrisvZm8UiWuef5#XkM|P$+%MZUSGly~kv!@QpmmRN zbg!U}$LOd#z)is1juHdx`A%;~F?83O{lg0GrOo|7hNI-yS?rec7(MqUXqRL`vE$De z?0MvlS^%^^1MOv=UDW*svbnbD0{h6x4)%32j-$?a`GNUI#Mrmw;ISy_3u^lMVa(aj@Yr)zn?NF$DRnWrI)9U~=aDAr6`=iP6O-cE zMIDddNEa8^SIzx|_$)=|FLVXt-q6j)o=0LR9y?K~>x^B`4brnWRK%KgsZG5zAKr9w$)Q-b!1l z8aygM@fd(~`yR~Ai`v`1=s3kJz)4}u@Q0uOtLj_;I}Yf1#%JqgT}54g5CYQAGX6!Y z_ActK<@+|nuP`=`fcZf8!_Bsi=SS*(2K>}l@w4Hyf4#*&=9xs5{k5+T*CRRp)g{+r z+<+3G{TpE~%nDOC7CZyQzuNd0TgiB4%mn*#d}Qd%#U=*gK5^4W-4?JL$hcc{ zj2&l^Z7hk-D0F1JjRC)7Bl*uN$i!KlaY^wUPu+dOp2|<+p=7fzE$TF~qN_)XfHOgBXzQ zfXD8q`!eOsBl`PM>w2y?Iq^#4W!QfSz5-g8Kc#r;H?fc1<-?i$?F%wTeNV1d*xGvB zNxZ3k)EWME!oDl$0km#FgwX9pU0)!jKWxr>qA5lD8@Q3NOA=%|l5#WlgTYXsb+5J; zdH6|=dIU&`z=zX)oxF&@adg|GtMl8Ax(T$I2(<2PB80#EG^et>@LzPbzX^0NHT<1} zJwLrsOMup0WUG0eD`Y+Z+;Xm#ffVDZg85bA4?6QqBgVY=>t$nm!tl36ky701sm=si z_iIO&TLRU)K+1>0%r$zx6>TZn-{`FzUmChAvHuKg0B)K}-fOw|kFO(+I+1JtpiK$q z3Lv#SGY&*gN@xhVHS`8ze;2q1WU*ZI6}LEB%wWi}Mh4b$J2%MPM<|d33f{_m9+70KH$4HqpB| zMxedKA>|;;t2JX^2(({JHBvlHi|s#J!6yN8ocEfYmG)0xH&G+`P6Az>y7#eP237;D zd#)|#*-D+1@8NdDW$H$q@mP6WhVHwE(XB9aOKiC#>L#n`#?k%U(DgRwJQJJ&w7)kw zx&x>i1crj!f$6XJ!p!w>=^qtle`e^8#{Oxr7--$}{7rtdPhyh+%|RY`gKc*ZkbbwO zWpw5H1~rOvb@nH^biw{2a4FFJKF40@zMr~r!0dP51MGKbulxN5Y{z3W6==Woe)chE zCGWAUKYyFo%X%=j#XsHeQFP^+M7GmrrK}fxYv?{`FY}0Bbz1QFeQ-RG^UR{tm724h zaZ_4loYN?Yt{zX~Z!6lI0kpq!9bN5jUzUl#=5h0&UAC?sPyW%IUmLmuX%hw;fYx2@ z=ziRidQbuOfT!8MhkWV8&q;>l;eHxPzo*7d(WCd z-8`@m$hJPSzO9+!*f;rzQoNeFvY%F9vqtQI_U{;5$5T$-79eFeZ1$7*CPnvCJ(7K;Fx_43cAeauugD$7C z&ttn4KZ8}~sl*e^k?87qW&!pefaO5z{^jV_WBbnmt-y)E9B1QpIop{t&f-rh#jA%a z{q2hVKrj|)-BV>EWIUv(+YgSx&zhhMzV>qbjT-(+9#Z4^P7z&8T2fkI-yU=XTK7gr z_i^ebftSJapon88JVZ`X+MhX~a7Y`)9#(KjwCUd%;54FFuX{UAjJGw{mH&8`4j_!AczYEcegJp*9EspMX z>h=OD`{BcVFQU&Mt!bQ78@kooFg8Gap#6>5QtA7sim8`!3T)ThM7FH z!Cw45-_T`uD#auIh8*4dVDWdHbZVJB)=XXBbDSHXYe`GW80=?*1wi||%wFWFdphGE zv<6MVL^K<7953hgtBvjT0QQL)*rBVpi;mcL1wDb*ZNzp>@yK=QYpIvgA3j_j^!_IJ zJTV)(w_$$|2m`IF_m6V_+cVTlc@933u6lv%phkY>_-GDoUN&?mq?MG1slKwv*wgbt zy2G6Rbiakry#rlcmY}x^tOeDml)2$fI}cf{+p-@A-GCg&$h9Vs3o>>iv+y4s%MrR5 zn^N!+&~e|(JV5d6JA?n*vF(9e(BpEhBLmT@aE`Nlbj%Xg8=J@X7@S*+A5yNzeh7FN zXx%pAD7p)&i-S+VN5IUd-#KpA$1(9&xHf{W-fq9beg~*=W=40AquY_Xt{?zp{xb7< z4L`&YzjZz%^SN&UT|KUDqRnt{AJG2F{psTG`_!!lUxUwqnbVLnAL!%FXpDW4;qNx= zcY~jR);+zttt;cTX8ZKticKx8tFKLouTr$XzSlSpG5qz?t`W!sTK7^%w+nR_fy;s1 zPi~(7haG?Qc91}qS3Oqd_gdQY0|S88y~)v)`_XQrUdn&FA5Hsfz0P?y@$oNRhSKhS zFwW3@#nF}f02Wcd3cL@@^Mu{T@y1Y;Qs{O-v~oOPp9B>^`@7cBJ*|Vi?so9$z#I?3 zD*Ln40>-N0??u@623G^E+emWBc#z}T0n|&m6&}v^RlmM{z0>Zs%<+Bb4dx7;1ndW6 zFZmq-%Vs8vj$4^sC8(Bq3mDo%S|wN{zOXITC+_#SA#3LU@3vm<{FECDgluY_$7 z$b8+nru}`CY*+IAX8bMgK{+ln5B!O}=WOP9pmiU2bg!rG7H}^Z0&-4Z8)ROQSp7fS z#|3hHCFjH5x7q&C)%`OO`xn4$pmjG(V60Bgla^2~Wf^SF13Hg-JjKzy6dyTq3!ROHTFE9&%c6;*{%)US=he@wg9cG@8i1p9FDob z6JQkB%66mkQ0CtJ-o^t8C`HJ(7J==5B$A>x?4dQ+$BtI<^E;qv)a3Dmia~Q??^6V{xa<(@^cMtY|gU8R!=*k==x;@Y1o;olT3K@Tbz3(~osSz z)*HIx9o_xi=%?=d&IdFCaxFC&8zpcoF!Sr1Q$@F&dpZqW$?vVS83A;D$Ndla_5EVc zzk0g}e93oMhHg~+0C^W?bf0x}`&`8K07iqmfSKRk{zLQY`-=HJH?zN}72x%ZU;)tn z){#-hID6+}at6KupMj^@-e!XGGNr<8Y~5+hxg#*PzU3OWp?moy%-bLiv~DNI-#M2u zKZ3Pjx%`dhGDo-mSUbO3H?o!O7G1r)Nq(F5;yY`gb-C3b?eCGsZ@5Z+ecRY34c*@8 zm4bw!JIm3naXHsIz{#KhnECa!Jv6_m9n4SYdebg3H(ZSZ& z_p?|((7%SiD;(W5hVE)xDy{pup?jq3mH~91Mz`2*t2WYZ3;5m8m3=qGbNm%-6QDIH z2J+jAWG#E%Y3b}QGcQ^KbgA=dL=vFwo zGpL&d7K2y8ZpN8>&nCLQR`zk9)(!9D_y%Y7_VyF@zkxr1_BYR+4#p$*>(;r3V|*a@ zE1CTkb;g-K|4X6U65VVk4=2#37_bzo21~ggk(*-fqvrzB{-OXx;ga?jY)h0V$*4!`)vX z+ku?7CR{ExtstZO2=)`eB%pQ69NoXJC&@Q(4HMJ_=6=UN+@8Pm_Lbu1p`@W3z`i%= z3$(60M}XoPO5Oe7QP7%-UXAD6Z{UpInQsw3qTV_CEsmjTzld_9D*( z)LjCu2Uh}f+=uW{@+kM?)yE(m zceUfM-cH5og@*2N*cXEiKdaJ za=vTkw>HNWqOZrNYUon!(apBos#9rmF1Q$IU8YJYo_nZ!7>oyx15-D6(V_dnB)XTP zt4kF7SHbJRO{I*JruIVj2I{^8I&`YjrO}({u^)&%#8PQ?DpyL-iZ1P>WfwykQ0O%4=E{$S{+6uQ5mtG6T3{Smls&gkB5=pILXGhhMpI@dTy zSD$Z&PjIOkCuaO@i~ZT4JJ9}y9bGxk>PvlpAlIeM>wWVKe`UX9<+#*IhVJdO84ex< zTK7puSMHyDhWcrs>hr{sjAL_LhS9wQU7d$Hw0RXQ09tpFqx(K}tAUiyVKWb+DMjZY zh3<8RzniiD0sIKG?zy&pd+Y8rNnWTy)f&j z+c@?B&4FB(oLOSWzIbYu^0sCFX86~RHk+~g25A4jcjhI!sj_sCzmB!HAv4BE{36x;wyVa5va(oTtQ{`B{!1r2kV)k~@vKCu08`cnRpZ zqoT@c!hzg9*}xAgsR`0M4Wj@$L(wH zQgWZ6F7;@00=OP%|6(kxhg``W?DN6xKpu8evE8M5uVU`FtD3UTVgH7XGQhx;}7&vB;5L-Kr=8jo(aZLStzzZ|RtTDP0M%yW>sS~7?zjo|uhBo;8ouObF9 z@s+sL7Y#G>mz)P!EGPn6w=}JkIS-JJ|JU!?1L(eD=F!W!6Pl+3Jw9Hub+gW+Eo zx-Z1n6f2ISv)It-gUyYg6zKk3Y3sP}q3&T|?(bqqQzlQb=N$vC0 zy#h=h%$&*J^~HILx0_3CKaB1hhVE+Y{#V_`j&2CugNE)R^gaaZfXhmxnA$KDBQ05|@&DXLYGRf_($9 zfbOS8c8=WgyCEO-($8{_qLlEZ9LG1e|Cayd_Xg6>XBzES*?y&eZ*Z>DUhWN$lB9i+ z(Y_nYuL6^R&ihsN`m+k}<~|c}29R8JAvdXUF13Se=l&~P>J~$1AU1b^(Ln3G=jgmZ z-F&bR=rJof+0DEy4+wpziq10ZJ_cInHA80`^}8LNiR46dyf?A^p!1fYqx13$wmPrV zY@NgA#k$hPJKoYMO8+On?>RtQbx)>!!joQjJkMUCUl09B-MpwpRw7*Mzep0$!`oDkbmv;c? z6Gr>LSYNd;=Qlv-h^HLtH`=_4WKC6VfOF2o4bR0(~k)4`xx!-V);X0F3^6+bId4N ze^8foKj)P|u0Izu&xjvUwsC8qOWkeg=y~!qY_%WqY{G{B<2;#T_bKm{;aPy_W~W`? z905H)coOLTzYbl7kmvGI96x{&;0{p4zSrB9?@L#br&TQ$3LMyQ^ z2h|?P#2t6yzKOaaU^GY(w}qy}J-k*mmE>Z&=j4okuVeE8SOv6yMI5VBJjXxCxfSRH zIsnOG8z+Z8`cgEtfAJwM)zHxGiv86f2C@)icdYJrbi0fuK5!$r8kloI?P`afuSSQt z)EIPiiDCaPSO=<6DY|i6H%rd{sy)PcJ1{l9zq(XxxJ$|X-ntZEa}qciXq{hb*!|ss zy7NI#VCwjtm?ds077cxOxm4X&T9s=!w7CX63$*TEHPXwo z_M5&1$cp}+tSgM`yIH1$oo?07j)Sbq)SumPX%3o)@4^BC4JvV zf9i*W+kkmrZ4XCRA7}gTbMYMhjP6+MCxaJ()}86-%5y(Hr2ccT8kqhLGIZs9*c*1K z&4%uG*#8brdnBW~z|q}I-45_K_yx%E2HuRD!Tf6roaPS_dqw>ZlYeL%{$%Q;kUN(;3 z<%8QmDKK?od+mOc@mqkscMRvT=;|^R`Xn{J>dAbsgPbGRs_n%u8=oui?z$$G~!Z!+)eelBJC2=Fk_@jYO-_q+Yin7=^> zV9xW!M^)p!vo7_nk-vV}3;~mX_D^po9iHS_GoTOX1?0YTeSiL0q0D?C^HA&=ms)#j zroZpQeg-)Hsf;e0fE3T*@!X#ZqTp%p#(CWTP-1^iw)N6O_me3u8l2AYk}h%{-~4Hg zD}f$|`SJ%J4|g8#o9a>*)86X{w544SFahXzOYQcaGbiv&BycUb7?|I)EIO~s_aOvR zYWNk!el}PPbbrWm%_&)5Q}+W%0dqT3zcE&#E;Yr_aXrIjD$osRofVGG9O@Q;cYx@u zuWOH$G3Qs~y;gibgO1+be#GWa(BaukoIg2n%I^X{rhWrh2TY%WFCTh*7JtE|nzqj5 zFK;5}{@??k^C!<2r+8XSW_@4~xCX4JPq%_GtC%l3+x;r@pNyB_OU&18GV|YF>}yQn zdI!+DyBytN)QtjC#>2k0Y{x*h6J7NFB{Gle+34#2o{arV;B%mL?~_09clT7TVS!_& zkvAaw?=I{>MAQ3RW_%CD*!P-CZ83B^Vc!$<2U_=lqbuLDd_(=uUtg6;vkob*Mfq+GECyQlMCNLWXBBmy11VMSW3aQ3Iexd^;rJL` zpKY!-N9l9WVMa!`wWBNdy_HkH4}1%98f5a@`vQBrsE(FA58L9$0@WFVmZRI@dA=(E zgMjw;CP#M*bw7bxFEB{~nMdP5e9gJt&citzhe>{8alXewSC8LQux|&h1zLBSqbt7~ zeuMhAf&6aR+>Xnc>qTFWvlIstr=hFM4%*26d@t;#k}g*ZY+ZMqnS2Ivj{9Ja-6H-0 z{Z4zalTyI)V$cm_@mcC`cgYgv2~#%~JOyZW$o@R$^!s(JOX3ZG$n|k_b^p9Vo3}vC z7c;tHM>j^@V(#@HDJOZ@tg@*0|>fZtB-wz&> z*6b0CSv;bGBE>gPb2kKW9DYW;^~) z!Tx!$2xxyFb9Ci*0^d=e0zUw=-&S+%D0$HR7Wl%YPUg5;>mJ13J)hst0EE&QcR%|=(3_i2*=os-Mwj}?(Akg8Q8C6Pkl`;nEvoY$A7%b)M!i0EmdX_E<#$uM zKTc!$Ip8v&V5TY)KW7WqI z(QPgzzunT?LH0tnAz%Tm+s4_S^hEz`>U)CofvFpe9=e}W+g<8?L-!i&2Y}%~>-rsC zx!-Up^`C+=VCqJz=!W=S`5Qy`YwWjy-9YO;;^@kCw_0y<{spT3Zai9LJotC;eF(Z% zI)$9)(xwnJ1zPt7NB4Z{E(TJ1!wJUO{$ts9tg^?Nj9)np7|QbC4}7P_ai}hRu^#{i z0+EyL#ra)-`uV#6knBF_2G zrR?X1q8r)8@gKUDtb}p~_C3MNKiy%hcdW@DMQl4LQ2vZxQx#f9N!H8#}s-uwM$k23j{?{$PG-{4QexoB~b)((lq& z(%%V3SFYPhev{lcveEGOLhLUC{eadz$e=PkL%_)^-G0P+2KmE)->x{dsq{LZJ%+hCcY`+}n@ z&ja31eYK_b{ehKb(QIJ0Qe0<;3!-y&!KE%#quNc}b7GGOY)oOw&G zXUqOOw4d(>4c!6Q-vRCfT37$Bdpvd1!94JyurYq)RdfRf*q@_oDZ9u!gx;dfDxh`y zI{wy<+cw9*b%CktjoNw8{pSDMrJ8lh z$b2Sw2pr@(GP-&^jKh8gcoAsbA62@f-*35zdMVrE_WTgZVQj3-%-;%4+1|p;(l;6Y z{)l}A*blVs(b!Tv{`c7)z}27^IAnfs@~e+)QZBc8tWzcr*JD2f3ZQzp&3V$!f_=Qir5t|a;8wFbW%4i&`**=Apmndar92HkAa~#tPz+4nlyls!^N`AN ztK&Lnwu^q)4+W!v)_uXzeVw``U?o@v%<&L#=Chv8d={(WR&6@l4dwrG?6-le4>QM~ z+w8?z^1VVA>bnEe*JPA?Av|t15FOqBgRr?1OabO27$XlmZP)E@lVeGmQw#ISOLr& zS~Ctke)HCKt2*aq;@*yZ-4*2Nqm1s)j&A*r*_VQDzz^=ZjO~MYqWCeF>i464d?5Q7 z*-nG}cxiqY9W}>gs~MZ%bD-mPt}|p6U{*o>L16mYgL9Y^I%^Fbxi6{q8hiw9C+_;p z4V3hKNzJI&xs%EWUozAq!DwK{5hUkH zbo!v9$L-VDOa~tW-G9ABknttspvNbihkz@9#B)yrjzK%PROAEk1YZH|uRh;xv6lO5fRyv# zA#B|Gdc5uJ!GrRg>AGwm;gj6zS0fL-u)hKH2U?fSK}!1f2@&e0j6CY_{^~pgTe#KH z%t?BjO`=^C%rx<6gnHa7ho^r;SxiaU5S+R`(67ux_zlLx3Q|a^4#Ye4x@Xcp?mqi z(3RhD-jA*>x1$#ZuL33Z6K=7UBR$m$ZWv^!pf`&lH%~xnr()anC zM!gidZhNG-lj#0v_;)7lx`1ne_HVi4--!{fn^FH9Fg1G&W*^OuYtOwPv;8f`<|FU} z(Ef2rTT1#pkxf45dvPG!wW$?}+VlLG*vj!~f{Qa(qN_^@Z6xlCVLe{BBq_x$zqh-E zdYwC|44>{+chcS~A1Fgu{s4FgXuqo4`!UaZ)U5aH}y! zymFt-H?-Lav@Wlbm*Ua)*&P3ceV>gPub=Z$>0>=FB+#8==(eU!2hbU4-7Svpt<>EG z9tL4x>ZYpBV`nlBjXXS#{d7>UA)}iu9ge@v%bCYO0CWKrY{UCX?EU3xZVDBDbsoZJ zyHx|0`O-}}&cS{sIOEHVuHL`?tLJoHIO{?ro$Jlvb9BDUl$5=cB^ja>QY9#wO}*Q<9)HM zn?64KmHPcak15fLu{|cy8DRM5*~ItqpaIbSl{@|wQr8Nc4yu0ts|MOWR{D3YTg^b% zavVI1HWz?yKEF^`-h_&r|x60 z0jvY&xbiyw>hrwNd2Y1{-E7C-udz>o-+|V>SpJ}&O2S-U+RWGhErH}z-+O5}x_UbZ zp6^x%41YUf-wW&oTK7pucktI-F9S2cB#=Tk(AJKB-m~_2y^irP6eq$xIgjefvAcLg zIqMsaIl%y+b^BLOFY+`>;xA|mS^?AFtbX=*y_WTeZYbbZlh9=~>2hz9{DL_^>%Qmc z=6=iE2hIX*fH}4z&h;hPzT|i*a;aOrimu)cuEahF9s^pp19K0>^8EzZh(g&)b5t?SFT%K zPMhn1)}7(#>g!hHSSHu4%yBl(8NYg*MbT}7uFmfy+ROrT4c+%0UHKi-8tSF2hfQ7C zex>O7E5(mdIvTp0u-^f80qyTbM^}F5#lWlK_IlkVEtMIZ|am>AQhr4s>;yN}HKr1E}WYp@Y5B ztW$Tet%37^jNLH}Y+n;>6M=qQ2k5R9;-@8`qweuj`zs!Tao84+Gx^`2p z(_sH7cm-(PYig$Z$4YInLx*R zD*2#zUjKl5i>R0KCHypfy_G(f?NAr_o-QK!<9bBB9+`gGiv2!Noq1O4-stGwPu*kS zc`zBsI2z>WMx1dZ{gS?)WQbd}Kv(z6r`T@=`+?S7?C8pIMf()TGT>Yw+qXQwGqH*} zgmVb-S&#c9Kjym(T`L_O*9fq`3)~B|?lwnv{>$Z2NSL)eEo%)4m9)VmC3Mv%_qF0}rWQ6tpNsV-?!k(G! zX$we(qO24h5ek1q{m}55O9Ip(;U)k@) z#&AAxVI~iIv9J9*n>Nt8rLsYghexQJ0Oo?}z?|<}GOi@Q+Fv!+t$gU}d1XEJTR^Rf zjP8q$?mFte27iHHfQ;X9g?3(J$Jxh&y5IbdyVZ1bt+Y$z{15T~9t7Ioqva3${gS$^ zK%QZ=3z+^UIVTW*_3wScPq@`qBM(RY$#?GHM4)vqu-EC4=g?e4eP7TUn7V#vJji}p z=D~Qxty~vn@-Ptl5nwLRy8Rs8tN-G-7TgC01DUtyupP+nVErpIl)Ze&$i2b?>yLJe{cP0WJly zU6^xhkhxKGb-!89xK#vQ%WkVmX)_#r474tn=%sibKfpW&V&Fv(WuAQSM0-4x{9@-p z&XZa)e!P?1>NUgP27j~d0}E(fhE!VDM_mUXr89iiv34%RPhI5u@^rtUyAEAelP+Dc z?+va7S{Ik3q~C)#lzJ&u=LfB;w~O#(uHP8CQ)qK9ZO0k9XQq{uhbbXlma=R*f0^Sf z;Ec2X$Jx2SSv7_Ie|0XMB6X4yiqnly8M-J%qYGV`R4QE@3Poii-4qi^C`}Y0HM%HW zG$|FrRH7?`E-uNGq9`V!EB(LE-s|bCIsbWkU)JaIon33M=XajHZhP&u*WO!oQ(p;~ zSX#p0PqFy{Dx6ICOZArEeYa;S@k`(hkojMkV;5i3|18*!Q(xu20=jy>_?z_QO%c-v zG#>_A-37fEvlgMQLT!+C9Ia|~*Zbv$nb%L|v;N~4y8N^kW|RJ1cpp^vLaQtDzk7-Q z4dnXP&4;+n2d!r*obRq^m(V?i%~=7~ynyPqu)6X-wI;-OfY#vZraQV`eV86FH#xeu zlU{Tm0M*4rLgZ!QW`TtHY+YTg59)8)%z){SuIBF|(l3Ecpt_?xB@^is^fkM)y$xL5 zXuE>*Qqk7~=2^$zfuw&NMuY0^wYt)dXA(aTR`C6Q!?F%?1mt(~OXj(y3x87)<0Q4Gl2M!Fa+H4 z-*4+{P0~vHN?#N(6VTOuF_ZMOUU;bT|^;;+s*WKH+`RT*!{dX}7JtiM6Gy4C%T^uNQIXC`!+ zYL*aroVaHp3toZ~m8qAMPwpp-->mMwIvg*$x$gu_v5txQw1)J#a1_+v1y)z)4O`O* zb%ze%_SY$_Z*r+M*5L39aQ&ytNRUcd*Dy_72J7)*o-9I6#>%-U9G1j z&njX{!^NPwo2~9+#7P*#_6cy;Aw_E!tf%QXb33|P&)y(?HY^6!^%^SqE92TV#7p>; ztvim9JdhyuLC)X#=#FyoVH@dxhohjn#iav4SLP#YvN)$H)C0-i-DMbqtf!pr_4^gg z-_#GezlpAfR;2F&-9Qds@`ou`34y-E4F>nQr&vzFY5#%|4je6 zhr?Zq{QfkyCg8nqJLY-8T7A-g2HQdX4KwB^ zL?WemrVFaW1t9flH`-aW*YtLN{nvF7IUfg1sFM|CQejQm@?Mo| ziN68l{e=Z|+xhiLb@Mk;YdRF_?Ng3VH@Pa z3EFuH@=(ewo5k$1)qF_fq;c16iTF8zDyZ%dt1Ii6IuPF#l67n3`}wT8 z>0bxTPINWY=O1@t*9Sz7Lp3+~x`8JATl=gG?DDbGdeo2fgJ3AAKNC16g7=;KCyAFZ zmaSVZQq~fryk-2C`c1%`(JfIf&y#)%EC$upbq?}OSmu+=p@emG>)tiOp6}e|7#w-?J{-@_t2E*U#O=c{Lr~XVKNL1HFB) zAH+>Lj@SL2A^A?kKN5dI;#?(q`J7{N(RtVL?+nsh1T8@QTkY#aA{C05GEf65LtYiu zYjDmIt(vU0s(moqh8D8?Bq78=G0MoYHp&%&+Kb zIf?E&*ll!lvy*gxFGM%RmcKqn6Wzu5hj}9n;fh5}3AhwAKfm#JhSG@Z3U`9*>u}0G z3my4RfVQWHN%I&i1U>Fc{pu5ue&aXdgBQ|wvXyzYX5^_HceWkxX#bi11M9lb)#I*? z&6N-b)&14#%I{0~mVH~ntw(8x{c_48-v*H;x+h@ncl<4P5$80h3aVS$w*T+##9bry zT@7*`k@f6mE%U{;KT>~l(4Fn*{y}|ij!hd!w}RCz${4aV+{Qi~PiCMi*Sm)K8p#jU zP5CikKCs8fUUPm~`WvVUA}9H?!q105E#ex0mXV}S$NmS$uRHk72jMx#ud3Fs{lpyx z6JvhzMcT?4iMfLPww>#HfYSE{OgqN$T0XV0X%53d^O0Me61;iv@^q3{!ljUp?%sO- zaqhP3S?-_$N%_Qo;eG(RF|j1{CH-I+2C6&R>W(IE0!WxFTdr-yXURVejnDIS^U-~x z5ZzZD-6{V;H)B8RrV7!0!_m$954zCGB}6(C*ApbjI#H=pskE!|_^b7* zrqwlvSRYV`ZeK@N@2~tD-8^)sIl51vHxVW~y4QFfnux4FokzTc|LuFgnSTaMwxjzN zb|1njN0%leA+qUS*0osO9cEP6xE|0S?g7bP zomYz|jZ?DdzqdHLBSSn*Ia*hK&~UNBHr~Y2m4BKKY{#hVC`Et_DlWU0eO#lwE@;)!HnN`yYZKj(s0%FW-mDNU~4Eeipwk-K>Ns*gp<3L67SZdt7sfTL2Q? zVyk}WG0Ao#Mp8d>(OnWxY}UuzDoKzFb`C>mDPQuGGlhpBgmMyuy%5_9iQf) z+tSg^!EP;VaC8S)-4B}3KU-ayM=C6TbsQNF1WgY|cL#Pq!ahfLu+{xjGg=N+@V%JG$y`T}Sts z|KM*fx;c)n`g^^jJK;b0n;s6DZya6qx2>al>bmk@TefEciMmOHyho^ zor!j#{@&&2&iW7eo5r}VQX#qn9Nh)~K{p@WhK1;kaCBGx2i;62du}O2_bEr0;iQDf zziAgK+}!Dpu7*kIO@V0;AX3I7KlwWYZHurjhInni5}k?tNXI_g+AnkLf3x=TJm|_K z`+V#tVXx=Qwe0^Awt}{IZb?Y+o}>JSc<(=#6`bGIbBm52bI+h%I=WHPSA-N$-Q`xQ zMF#KlAbvPJ1Tt@|@1x8(k~BUzGic7bE8*|+q?dkr3R~6PZgrOu_Z~>dVOv-~t^TI) zKDMgpYFI=1%~0g>#C23q3WQ}#@?2sc;s=4;Gs~~wUq@`@+ANKA_H`Zqrjcd|>;v_$ zvab{ASA%&<7z?A}P$kyV+WDwPg9~2grEzcn2FJhmNiY5V$81GT>PdV4QI2K8+ll)L zWE?Ixu2QI&dD!1({fd(%zq0>tFr2W}uMyU-8pK^KCvrkFw&GU_bhL~|$qshQ_#wSa z&nwTfn54LR=Ujv>92`y{ri`Wwbt#7pj|DK4{Lp+e_;GnTy>!kOLwo`7_(s4ILuxBow`(;I*B6TP*Xu zplN_k%v0fgN~Ea{?Lhte%=-5;aeqOv6!h3uxFB)=B0Al#??EnBOK}L%3PJM}y4o+v z^FKFY*9ufuzq_H&|A-68&;Mv0(BqDy`-*eiM;P!)|8Xn2Qk~?utH}>A9apn7YyXLt zb`&QcCFEki5PKab_G7>JH4xOVt9|>(Bg8!p62`EV^3Zuf^-ITzv5GF*(?=9Hf26uqgKlxrL!JDrhO8jG>Wg&X0 z_-ZNzO@(_B_2eniJO?j<`uCeXZn-X6MEp{a>mn(S6O?P5{xL1-xHHhb0$mO7WAia= z0`>1N>z~}`+DZI=koTgOw|!+C&9urv)56jDgEWDwSd#*(6AOF45^7Ie54Z>158U!CcD8??Q2VXeB|$UG@vlOJYesB- z0`+gS^{-VOziivG)iH>a!+g$FY1M;fspH>cq?rVhLH(O!{gZWv*~Cj&#mKq|UFIQ||W{a;YLZbJ7$Psw=i z|7k^hC#b%PbMzIgSFcCkx0d(m+5TFtBgc?Fjg!Yw$KQLg833a|{cU1(rQR$d{yj*3 z?_P<61>0eaMdhV?C-UJ_Z1Ug-P+fcbG$Qx+&#dR?!#QjvfAzfO=RzVMQfdaxrH<}J zq^|)DKy{Im5a~zU5O@@l*CXUpM+@X{t)RIMUF|2I#bze#2GyNpb>;b(xmU9u1m1^e zN1ivyp|544=BuXu{c?T2FyqRg>17=fbD1Bi&yDPcp#9`WzHXuKeaORJ))Q#x#XQuF z*tT-)msfT^=#}M~C%z~*Px+kn|rah19JVC53{j8(Ap7dWs zv1<~#U9IkY#0`SSU=VG-#+fl8E_`fIZ}Qn3;DiJ%!= z0~4F`ymCuFAI^EU;JSs3hKwiB)qI#l`nj+cR5xJ#EqX2eC{%+ObRdt5UG3{u7~=ao z5+NCJWiA=qY>pbjZ@*l!W?B4;EuS*=)efGHKwCDX(#BYW5AnQD0Amvbu^)0fMb)Vci zGTGM$jjVmv^Fo=%?n`P%=b>2F)Y0!M)9PdQZzXGp= zo)@n7bwl#KnjGRk0nIb=xx;p;NN4LG>??CJxhbJA~HCFdE;^u<{z5giMeh&D5)!*EfL35X* zdu1g*e-}HtUszq4AL+<`39H#(n7^u<(kf^MIl3E3|1Es)=x+C=BEcrCjRgs1*`7`} zt#!~m>FAzMdZ~|<*lPa%YIWs%dY4<>WBl7~&zcwNZzj4^3emmF(WRM5h=^_@NB4C8 z=Ayd@U9Ar{a{(^(p{b+$r`7F5+?~)L`hfT=&+Q)JoRNWs_^kCItxeE;Nu2g;50QR2 zWPtV?M}6I*@_zZ}h<^n%4|=a-d}cMJt(&$%^P{6TUDCpPp!r(DFQQ1&YP^qUd zwj1s#(Ei#7O^v?xKWwr{E9*mMv(^4+lfS#r{v;Fo#*Y1B?B0dN5{?`L1$obPgg{JwtA zCpv>Qw=DkNTiyE|?~mLkYfHK`(BqfBj}R#AK3O*Q4?2Ey#;zya3+l%J>xZ1@A0hs6 z(0q_Om?h+GBCRunt?kXIF>O;Hj_S6(b4|BA8CfdT+n*A z*VlH4hnYiEJ5a=;T8D>+jve3EK(K${UL&FGy=0mFW<0cwwapF}&{K&z+ zE%sWUO0vH!R0l2hTYURSC*rz6Z|Di`I4Gr#KaMy#EHKT9^&8byhYy3g4dVkj-YwR z(H(}3^m8K}T^==&5RvaK&mev-$oB(W-M*akq-;**1G-1i)$k5BAHf#Ta+Z74gpfRs zQ>0Z9a|UQ#5v^Ez`u97zUcX;ecut)(m%=bm9c~qPI;~nW{|G%n`j5?&yXa(XV{CtS z(Clz@mXl^RdgdYvOIU%;1}Ms4S9Slfx?_o(3`^h*sL+XLVL|jhy34 z;cwCxyNUI`Aae41Eb*5RYD`=!Xa{=Cq8Xb-8|X)SLnrDbH6~7){_rBGf8BkZNZ@A1 zd2l&Yfo5FecCgpEM;PCV<|zF0u15!Of5Xx3Mf&?815|gY)qR7wC9n$K12;eAyWy%^ zg5yqmAZXrobiXA1kI*(Pq5GoMt=^9D4O|Zmz-<>ZsduU?{$@TDH2cxj@G;|q7qEE= zwEn#0>jnn(q`eWZc_}urK|vE9kjTfm?0*|_K##Y0*i(;uL);$N58sks?(vSb?N!&U z=Ac{8(LIh$QRbq~1=YR8>dJfFuOa?=kmr9~-E1cx<#@9O^NvwR_ZDorLoZO>+E#ZE zaU)?2G$%7$-5mOSHy_i71kJtZYX3bRo2f7xRQGwSE90b3h~EOAg5>8ADx$0_o^SI( zulF+O1SX@a;b+oI`5a`c<@2Ji>y^(*;|Z< z(@#E-@OLLR2Vq!;gl-Y)G$FFGBli>FSNIW*P`*~$41-RgE;EVYuaTbKWrqQA%)6*RA-tKq87^aaoew0!pX zx&gU=(wTV4R~bV}NPU8Nf5*N%`+LKEj(w?!moy;Hn@=KM>x;yuj1HQQ9ebG{no642 z9Q$jmz040SB3{c?>@%_7hP~F09EXD-`O?hhOWs$0f7gIK zk@lbe6z8P}6S_Cu#=RMMA2h%BS>4%8pe}|lU_Hd*^dBJgF7}PDtK*9_PO{gdtKrga zlo8wrdR}VdA55q}aYG>k^q566&Uq;q?rm|g8DbW*NOD*&izRk4^KiP z&Y5i>)`)L&C*2Qc3K}^_I15=j?s*yaXR)dIQg~_yDe}G)UTb^ zudLF%&#?#hN8nzNvCj}Yj~(UxBXX=-9%+-fZ}0dune;Q@Ezo>xQryec&{pDh!d`Im zZT~^`zd(I;bPki|B%H%VrRv;lb>#a8or%8(+&T6v@;~RrpxNN)3@6QKSO=;z)#`MK z(^ta1;2vY=cY-GNGWEmBuSKL;4xfSMS0nO;5IO4(o&$yp;aqT!Eq2r&2b{{USAyo^ zfu5cBr!MImL2FR|?)G&<_Yl_?27s%X$8p9O)U`p!FEXA#B+XMW8Pva_)<0RVy@dE2 zcn93`U>bhZ9|PWn7J^Uj3sn^sqzdrT+(F?bkKswbW^o^0pCWj>}7 z+No26=03;YH%R{;tOE7-d#fw`(XSHUlRkm1T*nlv=&wu8xBa^Mn=&nEhC2CqCN|}u zE@*z9@O482h#L-%gZ6cD>`~4EX<4kda&)GX<_*{ns#E?L-#(^d3YQ&zpiH9fZ(m~ z`j+_pAm@k<_8d|D%Ycc^XCBq@ugtx)N4ON!zxKY4cink2@tr}&vF>$piQoO>*ZP%% z?jpy(p4jw<@u2?6oQU`*@14pOO>pZ)CVi%PlX`)Umdg>+6uFQ4OrZXavHm?t+;i|c zOaqy3>I+gX+wFDyNQ~rsn75F1ck&M#){(wk-y-IZehL2;S^s)HKpDd-h*DOWzxr+e zP>Ha~TOBkT9i5^N6*1?*RiMXNtEBfMA^GmvBg8)eTJOZS=q%P7tO=U$(9wGGI%(Fz zPSE4L!Pkjw7{pu+?1LXb`e3~eJBsUC(bRItTgUq6!Hi4nan>8my%FdIs+(zbR}l9J zY=bW#hjUkqK1B2~@lm$NXZrbp4^BTGuo+teh2o6o;=Q5k0qln9dY2coRG8P0)-lxpFqNDk_gEV{KFHrvu zS^wnz`FX>Nm>49#M*OU?;>9(QeimPEzleqK(DKLmmElWm_AAU!e+*K=xWF%{SK%&BH`a7 z*1tW(9R%};pMQGKee!!jleLxkb9A)bH789Q=mhHD4(s2U)RW%C-wzBI3Otuk!}be0 zKVp~z>ggQ!5NsZUm7u!2t*(53z0;%A9VmJ?_g1LHqx87BF7b|^-;?H-{&mnycl=$5 z%?EHnMnbo^e=w0f#2tj_NcO=*G-q1f$<6)zlzu|4Co;A(&&rKOtw-Cg^5=P<5?g+^ zJk+qdav$Lj_DlHR?qB7gTh7sy_bWe-=2S=bN~`ybz@>x%rU#9qYUt zUG?`%N4Nfe@HYe9rjG7+=xRsCxUqQD{^ur*2E?-fw+|u%brjw(4 z`QzLJhd5~dK5ccok+-Xe-vsg=XPMv8_eFH8Sx`4?XVCOPS3_VF;{~V;+D>2fbwfOg zX*v=g2lxDwbCmJKF2+fYf0~~UlUDtk>*VK?#7p?!%0C<3rySj9v6}*4I=Tz3?)^_N zPK5FBB*^;OW>x(BT+97nxBPSWFh7B=)JqAno?PZd*J7vs{%Lh(ePbT+5_El|^i!gz z;naP>AGzP*=mvj`;xT@`2*@^5aHd$6L>GjCk#n z#VY!9&|Ff8eUT><_Sc`rJ{S9Dj{TX~oeh_R9>*Aa936>^!#$urNPf$^B;&sXP2ADR zAk9;-9#m(7)oC)Oh`9;6f_wh!)tmR`9pE0Uqw@l3dPSzRUh2Xj(ab z$+|z8*C_jR;RZ|wc9fbSi%U*^ebu~olDTfb6?YYq}xu@%2! zwC5ok5`BfPbI3sV8OJZp@0+pH{2pihlKk#Uyz5sU_DdZ5UhKago(3)V$-aGPHF2N9 z*Pt3+JK`Ou*?$Jj7wBldA0o{^(D9kX@hz~&S8*)YJJ1BK0l99_=NCHPSa5umbtGsm z9G=j9gY?T`HE6zdE#qZXWCp+g6Y=>V_y5GliDmuy?wCDq*RXGpsV3HHd3B} zafy7p-PiS=_bx%a9<$hFW8c>Cs|@>P-thvq>Q{g3S6hDb65=J?&PK))@?1rxQ=VGS zqQ_VV;rJ!b7s&af7WP`61FT>2d_fw6F}EMd!2Ut(W8SYBlZss$ybbDChV^UNv*aV} zgs&kdPM(33X%^QTvX%GM%XlyQB-aVf@irOH*c7^g>K?GVFA+BfBrIbq>oq6Z>vhqT zpvRk6BxIJMtL@?g(r*&m%!F=9+7ltNl743q@vGroaNpyV$#tvf>bN&Q7&6-(e=AJj z+yFO#`rE+jo=bV;5Wf!O`&H83Qq~F(P)_L6F*q7n|guq?I$vQLFJ_q~Zh1jP%_UD|&-V_g+ z=L)eO=-8J%jeRQi^9!*b;@FovjeQ38%L}m|?%1Du8v7jVzb(W*!?7=a8hdj_$o!3c zbz2ceVK)ul0&32IO~x_J0C^* z@$ej|?)z3(){!3~zUU}61sV%LZ&&odfvGNn-sViRCkZned;jZ z5hFeiz6Q}PPoE|2d^vM-;cm0A#*jt}nIU?Q6?f%OeNfO6Y2+ zk4-8x1uc*LzHUI~f!Y)AmPa~^HR?L{a(#LmX&wYEkK?|*cYQjO_=Vt>hdIt~#zN)> zM`t~0w!scioiN8i2+4CaW(xCkAlD0~f}dxp^pk0oLZ*k4PiK?nJg5PhPvv}_h^&Wb zOMC}7cx9r$l>ER$`%pR6ybIhFOUAn%)U*Ma0&|Fj%4FAkZhj(;t&X%B&E3IATP{`Je^ zdKN~4mZco?q4^<`d3nezaQwUQHOA8LFsOeGsH22PgGt78na(pba4)#mZn?~Vil+LP zc16gnaQs_MdRZ5?fvwiFr+i)Sd)Ge@uXRLzEFb&N9KU{Mzg!m|WvhN=S-)hxntXpp z!YbOO+s}zl64bBkl#tow=$<=+IZEgas{6jxy>uq`m7oi>gJQJV;hp@tm&H81=r*wZ zKx+Mv`P0#Th4gb_3#e{+u89Z{xsR4Qi)&!G5!^D_ZpTTgo7;f?J|oe;j3E6ecm`DW zI;$(+CzwXOzL($4hqUDRHXdR+2V`c}{f)ZY)S?jOXRgtKRpx8V9ao1hD*ZuRpyF6x7f21!E-QH#Va?|_3xy?dmI=Xs(>`3}JJPNA2!s_b#CbzRM`F)dL zS%00g2= zpt|zBJR!1zxKBVrE?c+UBrharxuv!YnQa+~eAq_%{czFzgl>1vYlO%q;=Y3t3pn4i zmHP)vu4eq#h_;usUO2sN$ozyZ?t7sjHm%?$P=62jyS(=>jUxUTNM0WoqpgU)dfk`G zx~H^}iTXPQn|ZJhR5#)udPLTt9VGq;$T~E)9mlP%&YPy*95RE@ZDIW_{w8zr@HD9I z6;^lSLi$Gd6Mljg)wu@@;-~r4uV-hjHf9foQ68$V_gEr-D}}k=luem{>OO6C|0z*c^s>ixU2lof0Asyv2ACUVyQ14*f$6#NYB<56D)_E!K_phOUM)7t?1$EzogG zHGgO5ZsG>O5OB{yS(I5Gi)k05qy5*Dqd}U{i{n{%Lbe)*%t7RK0F_tJV&P!X?nl|pvTEiO9*8W_X?R|57(3sIIv) z#C@uSzcOC#i0uQ8?nR!G@y5#&h<_PmyzJ_>8CLK*FB9Fij_x9CK7g&D{?4?zvM%TR zrPN8N26A7h*u{Q*&SVbM^*6gG{*FrcEBV_Fo9>{x?^<2S-zSOB1nJ+TP3!wFW1JsF zSI1Quy+URky4wC;!)6XF0oDD<>Mm!@mrMK(*amJs?6>V&#y@f%jolS8I~-m4KFL08 z4uI-5vAXho5_t|!LjQ*R;Tl@|A@Nm$zHcuJ-Cxkva2&lemeChLkVvVgwoeVE5Z4fH z0MRRfuDrvq#2;Me-yJf?9G#m<(+%RFI=A>g;;s95jQFR(Z8tU0N$VXlC7)1lxaJ|v zG?)q6Zsf9!5ZF%KUQi8*%(*vYsyOz)vHx!vyzbZ`#f>eU^b|R4PNb@0V06m_ueinQ0^$5Mg^ZX$BBJYwBo!k!0*`|lgN_4b7R3J??s0mtbAKL!7 zJ#lx!KFT7z1&B&a&-T?<<%GLgpKErF(d;7iT{{D9KBsLZM%dl6!_K?qk_gKRY>X)83N^w0^llVHI z@7L9Nq!fF++Hd8f+tl$(?zdi#-94b?oaVdj-EVz`_&Ff&fr(f5%Q?!o-m*AR-s$+a zhBUcQWJSWiyL_F<$jkWtAMqWaHMs3wz8fd`rvBv)37HX&e-D#B6S6@4JIngFg}6NU z3EZ-XbIjSpLgsCBQv9?g@ILQrh4Vo3uafnzK5t zG9Pfg0CJsPpLtI8H?6LppE{0-j|g$SpU8&>*tCMzKy`;&-8&I`0A|2TP=7J~Qw#jw zK%3-#n&iXhjGuS#n`S)oS5Nwe@;7BQ>7fy5{XQw_sJ0(fW1fn5EhC9c!@f-+_BT5A zJ5OVug?&%#HNRS8cMJ3eJ+1-rL$upJtMeXq&Y$DiHyWh=>3gd(ZTrytO3CE9(DCbE zzt0}~SFxw5cp;15TM9=&{Tk)(ip>6qxlvdTIpD7AXx7Ki@1~Tk^e1Lw$ozn=_J0-E zkl)Z3RClG-{fW3g;H-}+3y}AS%w?TyB`m7eXd7hW_8THO(YF+ySav z(JzOH`GoN$TnObLm+}8bu31Gl+xFM;UG2L`pYt;Fqv&e+Tt@n9p#`Y!qgHntaX-Kj zH~@XgJYC;1-?kGepHiexofw zm;$QX)jyWVXT(X^!uAVr{atJ8nO?8vq1zT+&4+(TAKl1z>_BxNu(~6O8x7CFI1pcT z{F!6htDa}`n0uXuu7-uAe-~DRo-f+^ZiZyN!I#8;1KJl!9mt{XrY;DXwdiPl-AS6g zP;yhkKYm(*_a4O-#NP((K;|n>&D-$)2)AD4yvev~T;jN&B>l561N6Alt**>pttb9V zkmuvv{LHt=*PlE3P2KDvh;e;*+IIG6~kd%2vDD7O>D1?U{gz?m?r zI(|X)TGk}kd=P(MCvp7kkZHoeIVP5b%B1fG7kr-3B@8Tn~BeZuffgVY&(9aiLas?%L$ox(A9Qvob)BP zFpdJ%W$CS_+nTsrp+DRX?seEk+mFk2SS5DF$(LNmUwQA&U~Gng>N17s=|;;LGg|Zr zjcXOW-z0e;LH*4`_eXTK{*J?LBE0D6rdeHi?qIgnmHCz-WfJ|OuX^&N*5BA_t`iFJ zH`~!=%1A=w-}sw_ZrSmP{9S?GYFO{+a;fR*%KgdR#QzSzfSbR0wq5A?Eajt+xddI! zhohvwYAfep5IKo|%<4W)+!&Yw){TRDV@Rg$*b#$eF*w4OX-F%x5S|75}?T@a8 z!`PgJ02hmTez4C|hvw0bTt$3Ckp4W4vDa7Rf$2xR_#|YWbaZYYO*`les?*5#FC@<= z4I_RexNWP%A#~Pq|J>2ZB#qSj$&OAdt0T{+E+_tdt5bqTn6Zw!fXmXw8|n54aDp?lELJwp0YUvYf|s(YiKnUR-> zn+1zu0l3GSY5ONFhfEs&xtUrD%(0X(W-B(|fL<^7)Apv&_od`we`O)|J01I&FJGvA zbW_N*#@?ned$BtV;ja_Nb+K5gqz@caLXav`rClt6WyHLkjX+9_r1`a^u1v`sP0~Wm#6y!@qfV2;Ogcd_w!*S zMxMW4gv?S$H?25xli#uy3{+S84}$mI>br>_1pUC(O+8UiH|%x#zY=e@Q?_i~abu>VE9al|%^|Z7UCrOJ*i?mvpt|*~Zi@_#k@!bJ-t+CwdF7q( z{nhpp-@-bU2?^cju$c}2fa<1M-9Ns=7B1w1Itnt!GKGpH^&tMOKYr5oW4@+8M_0>j zFzLs_K2Y5ctnQZYSwjiO;1J}a8?Eo_)<5j)>bzRk_K<0SE}92b1=%O?CaK`Zxg>StnJr_j2+zfa{S$m%~3F1)T{19tGj@>ci?mQ2-a3% z9EoBaXL!!@$NRH_^qb%qh@2b;kFH4w9p1yY&3ks^lr*pqJ%KT6h{%9YNfn7uFw4M(n{loAG=y5S+D4>%q;hBOIIP%0Wo^=KT>ecR9y7 zl=Nd@A!vU6;OmCY+RJ@XxCr!kMJqa&=Rf~sp1{%3`ga{^Rp%NzF3|e-K$1=h=knNJ zTn{_PIf67}U@U03G z3wJ`O75D2v^rH0&=3m;;kog{6E$=5uKNSK$C;WZY>i$98Nx0${p5ub$vzdPa(Jj{2 zug7ve^6K%)kU8gh-%$PzBmGk_0W?30`G+2Qhq$%y8A!RzXWUx!L5dH$qPo}^mMB44=8W9Gw z)V*0G$qa?fOXzm?ZOu{A2M#bl2CA#$&632O2NGgzHd-HdS9EJRo9 zRW;I!zbR~0cZ~m=-gjTqi0=iRz-{kwTW)&(Nh=yQThWz%P{RGB9}EwJ>Sp`9A{jGT zk3_tLm)W}SdGJf^aMW5t;vuR#7ijtYY~%CCGq|U zSJylLaFgkbiHZ6fCB5ig#8!1TTiq_i^#Tb4*otlk`d8JIt<>KkBuP6XY^tKG`8$O4 zk`GU?Rox=Kl=mFW4C3E`H^E)6o1T=v(KExQ9=gSRpUg+3{}N7s>Yi_P>%GA@Y=2|! z27(LyIk;|Iv&nT$)*Sy_WJoLdkjb<9tsUKmuo(rjKy@#+x<3>5Clt>o0_0jyb!B~+ zkZ=>*Bb0yU*F59y^fOxixfg`Zcg}Hl zB>i152=ur^v~NOaDRHY{6S&t9(N2`#g<*3Z>w>jCZ70o7@H41?lYE_se4oAKA>Y3; zY~8VhmdmMi6c>d}eMh%E=`VxYpt@_Uu6)O$HSrxl?jwtjy5=F%wo}#3MYj{W+D^M- zlMWAq>hft+3Ep=W<`cgdlE1T%o-`iFi-pa*j_yWmw!t@`x*^F}bmjY^$A~Y&AUyfK zda2~8V9B#NBqTD0kUB_$aPuuOZ+;2PNuystIs!| zD1uZbkB&I`Tjnrh6Q~KQ%OixIzw&;bj>LC^XZW{UZfy?v`r7_es)tR97ZT^0e%OqH zaiF@LtggJjWghWMK;K`d<(6g3?K{ph(hsIw8a5Tt)q3_GHtS&%sP5D1Z_JpZ#0QQr zZ$w$Tb+&Jkzwyh$rVhH8c%d{l7eX~qU2fe;h%Dw* zKSfu|?Fe=!A$T+~fA+GcWQxjnW6mO8{FLvcq$Ks%k}RcW*z9%uRK2quy%|Y*a-Ey3 zC*vXs>FAyDVj@3PuY#kuz)yNw|5Psf;B7dv`Oll1g^X72Gy{z=G3uQ__! zU!|aPH8gbea=etLsJw5vDe-Q-)H12*7nh0E3Y+$ho}BMQuZ5%ciLd9a?~(T>tDZ!s zVc#2jEl>3)&9VRN|KN}G#}YEo8-bn%)$8o&{pIiU{Be&@qO-A|fPKtQW8&ELfqtO% zHs~L__g$=^#7mI(M##9JYAbwj&X3pFb{4xLY^FNLDet+Ke0l`r^dRl8j4yR6-wUg^ zS})SjTY#RnTh$xm=w0xC(95#*A_Kkm3-RwcN3Y8NK~Me5MsI_Yk5kZ*Rd2Xgzz|`j>~^SB`(GH_y>)n53umOtoaI^){M9yGJj%9V~M6+BkY{Jt<5t6}`ib zp8EHWqu1Wi)AyK*kA>-_qj%=yL_Ji!m5yE)NAG{iKP}%Z^v*|5>x=dy>m0p1t={SS zO?l6;ggo>vK~MAVOLU~2Y(ZLEtd8zX!=#}?O68Y7|(L0fJe5WhVH1u|%r}abgzo(;D)F1bs_B?O= z-!z%%{auJ&KS!_BztGD?@BEkk|KlC(=#@{>JKga{>oBe@L~o>{XTS4s>UdA5mxkU= z=*5!C;~7Vr!98eZvds(N_|Q0-$i#2>9gyFO@x8H>dO19%V84()wRbRk@W(NiEjq- zJ5Et?HkTVs8|@ z&d#bIHk}=RzsF`T`~s?5+dp*g{=-q?<^ASz|6#k64>?KxMjM1pe{^HMPfUPvjRWU^ z>ZV)Ws>EFZ4WS;$_)6FN=33px=h^&4_X%`0{M+@u`Pk2Oj<*TF*AC*KehsyL)uYZ& zApT{L?;N`2p5DpNM=hVshGFv&x>`Q-uz3sCf$C1My7IlKUx_!|BRtGj>b)LsJgIzA zS)_8*IbLbsWw0v;()^@-P4RWTbriQVx1jZ2@+brQ3tvh26~pcl=mF{%Q=}3i-x0SL zj=(_>zqFp`jqvADq~DQxo^oB-^mKG`GKTj)$@aV1zQ5^B!{%$p-+M{_ zAmo7h+eU@B9AhbH3?q znE!VCokaS%a1d1YQ9EBJ?@@VyLC2dg1Kj88b06__#b249$+;k$|ZHO+j^Y ztnN(W7Q!0%0Oa|kzV-b4?Us_zmHL~S7B;Usy7{Cx#R8@^s4lPMlMs>jjg=@KFqgmu zu=xt!Yijc$<+FsY_?yx`Y~DjxLx0jg1QS8zB;W0KNM%t1I7S{2C;bIWth8Kk&Ux zw69EbE7=>$l3x423$YW=BtN;eB_VLeRsa9{pmMRl%JJ)B?CL{LP`}Ff_7Qnc%^c$2 zg+<`i%L@Mc-^Oe$cpp5wOW33p;@3{>)UPVmufo1pmyi8G$FEaou)haqK2a97w=Mo06n6KQ(DouK}$^K~NYvpD|{FG0SqQW;~m|w`~b%-_2f#UbQmMuT=5NI4@~@l8rS16X{Ca-7paPftwHVjuFWV zeO@;09?m0)beh>TxsyoE$ z2G8c)1u-}evbZ<5kvtYZYgkdo6J-;ATkn zH_u)-sK04_!scFdW4^7q6`R|kC#dc?tNRdf8ITE2f!i6myrB^9?4?~+OI_K3!A^t zWhm!`W$0~$t&Z*}e^;blnSf~mw?R9|sOa~_q8GLP%JZf7^V9kLm}h-0F zO@-C44Dx8h+o@}!mu0WlG#|1DhRqO1cL(Wzfj>e0l_4P^Qo0;tPN)X^iF4~i%;tmY z<~f+s^7v zsm3}o;w8*w>%ON*@{{$tnWYWJ08$fk4t?o1?_~yYXSPI21W&I&!b6!lj z+OH4NKS+Da8xb~TrYGvdLDC-qb6!H1N7N;F_kpVtUk7S}dp|tY=C7;l8p+2_mInB@AlPB{qlIx7vqhZq&UA>O$h+c2#>*(&Wy7D~X z)5J@t$iLnCpt`5l8)SsdosRDFR(HCi`?u9yLfnT|_fxj=eDO@%-ja1wN763PE#`T_ zcjrjI9d?7}Ls5G^UelAwHXZ^j1^X5h^fxW3{^p}Q9^IJLEr-oT&=^#=l+_(d+$eYf z#)DgKId+_M>U{Ay^Y4!C+oWFwAA{=hYB>qsI)<&pOJIm<#&Tak=QlJTsA>JBfP$o`y#uM%~SiQ-3#7Uv0a1@qE^>P*<`?hs{oOHGgN2 z{w-Jv>Mx(flMvZK+&+--J6ks&)L*8hJb&|_44Y%<#=O+510j9c3ITHus4i0s5+X}0 zl0UE)^1$3dU$B?HU>x^{%K6u`>r1f)m&DQWVRP0D&(8ZJ&llgw2bfy}Kkc3O7W=vZ zdA?ZNLMC=CvD5tPN&3Dp1SGkfFUyHzR3UjzVJ`8@z?}ha)$j`GXTltioR{Bg z>FWlT61NgWPrlzHAs_o$&T+11|Cg{0G~e4s>|GAEg z&7K%GOC8;^q!+(pY*n|5)vZNb1CSu=zzVC^zs~mSNqkb+tVK8Gr8XuNyH4;Bs4lM_ zk`R${d6)Q4VGX$TFE-B~XYQqJs5T~*k9&THuJ&`kkp39_1FHLo)xEea#sv zKbJR|d3toUy>kUA>w)0g+J_qBCYb@S1^#L=BZ`suI)RQETl`wdxe zfcPR+0_HEaZa&15{EbcJdIsH?_)aK?%|%cJR5xs|>$?(nH%Pdjtvn~I^UIP4614x! zo)$Je(AE6qaZxh}n_-~3Rjuv@iLimkB2Tq-U(x@o`CzhG-!d~%AD%^T7QF4~G8Heu zdk^Gx;(vhmxCn6bA=i%Ea=5-+XN3e0tK+t-?87b*ohM^Ff}^mvWm- zx#iI>tIz6hYc+Z)s^v}w2RbPVe^lp`(atW$A?`dP~C>!FPey4&(tIS zTKMALg8pV&-OtGv(alBo>RE~WZHA5LwgxF7$@gZyZc$lp*ok;KetE7rqoLn+rOe*K zPYLnY!=@d28s4sDOb=}Chai!n*WBOf%>#`gege3CSe!mA51j$%#Qd}-i!}4$1JL7b z@9&G;aWU81Fc{>!KCXXrw)x}K9UO0_q@T+=Q^&vV+y{IMyY-;{b?|jVvd;84@x?AF z=wCkPs(f_Ldp(hF6-iSaE(i5*r}arXLh^Ufw7#FJm-AS+kFNID)mpL+8{4jq zE>&5A_gusi#E*p&KNoEG*|wk3`wN-#xz0vc!)w^ghgG1*-P+&jz2EvT;{OI6H;A7y z6PUGt`Z6cc&ZE^i_dy)=xCi<=k^Ywk%p>p|JPlID`u>p;!~Js6c5W7jO&dq|V0Y%C zvH1>E_amz-^N;(8m+-$`SEr+UC%PK)u{#1KE=wFY!w(6eO8mK;_^ZG@?kwt~SrRsn zpp)VyHRgKKbb}S3{(b51i=d_md zUkpb;b%$Er0W~Sg`m32t?rR(Tz?WT;S3%&arEH6u z?SXD}FEwi>v1<%19bHa65+Xf_>jUYKtea)~&69S#mhpbr40rq;jLleh7u4UTRyTG9 z{Q_JM^}xLkm}kdp+WzAovX08}H%|KA&=*wqL8~jz!9Pa4gmG-;xhS0<^>e|04qnEs zu^g_)&@JZM8fjk>v6<=Uj`8zCAN3^xwW}^EAy4qiD!Dc_) zl#=lGa;y9PZobn%{C@ZWe3>-l=w8Gijeb$@YH*c@{Ft$k&{G=x5&`OwSi%6m3n zBYr-}dram2Vu=R6Zd(0<{Y>WSuqiv=W&#^&AD?5l8OSED{?+w&2b!?eejp3Gb+$ld zx7vN>*u7MUT{d>NwDfn#-|eLT34R9ccR%)bhh&}IU&J2={awlHZ0cSTI?;`dsU+>~nRpXmdFqF^}u$JV=^H;0Z^^%Lw{IS(iA8_?N-e$-$rKny?90 zO_aOl$7`h3@{#LmLS$-X-!b_+gMSy+UUmK?hVBhj61qLHnT_TWNB07&yNNgn+t_{q zuI@%#PgOS)-Mcw%sPwtNS}0N735UeJBl5|6?F!k!$+_c@9M8y;468 zn`!83{lAFx)uAq^uFORdB5jH540k|{N(JXD_b2(AkM3T_-~QMPhGC$(*IV6D#El0D zndEOH#%9`oi_a3YKa778Heu?H`uhTQ(_lKNZfmP6_Z{A}y2;~6Up4WZT|T;%9o@IF z`vBHBy0=>0uZi0Yhu}AG^Ec1tua;Z(+OWCK(N%wsW2gS!@gMw6T^BZ8(AD}7sS_~e z;BrviK34Z8;<~_n&ISZZoP2l_yJfJ-(H(DfcN6y;1nV+~z*g2BP5Pu>mo zzN5R;>TV-$7aW2E(2ahuug%}Mt!L+PUVDi2X`it^+VQtYJ+39;63~2DWp%p~*Bb`G z0C4j++nygaA2M=zzSGgwd>DtF>gN83e8@-ldvvwEy@=lHu)xv%-s*lx+t4{+`tUgE$HHr%x<&n_ z8j$v;3xbrd>Xxy(Rfwwv*Fgi2cC7WR zjrCXSLv$0%^Aej)BJDAs^68c6=uU@l+uqVxRm-NTGo<$Rp|RoJ}l_qA3O-Hukb zQxoRQiI;E_+rs+aQ~lZ3tRqEN^S2}Ed%!)Qy0=+f$%hfdj|S~`v_DI^zhJ+Ui|$Ux z-*MQy0MkHq`&iw@#C-stfZYFf=c)T9^}m_laGuXj_^a(W*U=qhb^lF&X1)!Z%h1*9 zh;PvQ3I1|)*H~S7k5sv9==0$skoKnE0UK-mHI$juhuC)JQPI^wOP?F!7Ip%+u@l)?;-1x#--Dj@H)+q{)K!K=bor|3@Oj8Zs{dbKn)I$~e7OE5Cg9 z+x8{bWmRz?o)*WG;!c8uKGZRHeX`V92|BNrhwegjqrR;v z+KBsha2}}cB!6E-zAM#`_+}vUJ#KwXwe?Epd$M+h&1dM=x4Nyd=?d?F>at8kLS)jl z^gr+pEQIQuW2$os@HeRLxuhdR#y6%Oz!I1P_32}pf!v?U=h{hhkI!UX9EmdyFmHpd zwu2u?|2qUy6S_BB-E6kCh_4Is{aW{0DsJ;Z+d=BDyaxkaEw@x`ZiZVx{pIpjLZlCI z13|*WY~A`GzDl@&a%)8W&G?OVOz7%)XEf<2!(vcfs=kECQHi^r@dTU+?s=!%2tOan zcJ|wC+CkRsJGxhrJ{6jR>h`m`GJfntyo4TX3-dQR$Cy<7P5qs5DY}}!eMmnL9tG7M zX?16D!d^!F2O#f7aPv3a=I_rKNk5bJ2jfz7W4^6fhs_tT6;ziYfP{#=AATS45~RJ! zbJ6FV!+)+z%zM8|V()(JA;zU|c}cxL@*el!vHc4~LHeVM{N07V$32dH9qhGyPGVQG z3D=9De%X16NXN0fJD&JIU?0f$pY(p^TH9Xb{c+96$K1cd=3#WTf2q@yeia&n>UQ+~ z^48nkOuU57Y^8jX?@#IcOxj`AX*|=gu71k&X0)$>)=!VYA)wPwqopirqD! z{&n$nz55Wi5Z?vdx|4-}S?I`pYt6sjqIfzbd%rbNqdz1 zdz1aqV_|dd;)HI=OL>u!9Zas9h9xW&R zTBzDQ;V)U?=@!3%wFpoXsz9?8)(kc9eeL|U-_9$|@z?vClRO`bu9nj%(mxL`Kyp6F z`pN~wF9oV$f$NP->o2c~_u6A7x^YMMeQa{!FsN=L>#ux|GVMnC5RmbNn-6)EmE^DL zW}BjB9J*R=W3ZV4%RzNFSlxe!E8YT4I0s~$5(CkTGOm&B2*v?&pEDy+)NFD5twZ{z z&;wNWAFDean~#a#1hPKd^*7UwTh-s3U{P}dT@8Vj%s)X4w7y39eL`qDac{vUkZ-}J z+(Vvmk&{KeeB=1rOzydpFcPxFEg?i#ky4m8Dq-x|$DVxm*)*m%M``| ziLS!_cI>tO%KH$%!}dp{<#uFbA2eBs{+@UVhmlvdPmny3pyPE8?${W=N_Q{h%ixF` zJ_k|kms>O8X2$rn3ij$3#VTRB{HlVx2(FAf64r*-32kS%+OAF{?kptX0=BlEMn_v% zKdfFm;hH^D7(Xu|{YW$lslWbsgAP1wK7)7(efT?$yFX)>jLUAFgli2~Ls$Oz1l!M% zR988Ut!Ph#l3x6yWm1ma4#w^a3h^s!mLP36-0CHv=_Y${e+%l0)LxEnDs!3|ID_D5 z{n(Q<2cQd)9@hoYp@h$UOd@^;nuesm(D~;)-*0OD$gi7lryK5{q)$}R?;_W1CU=m5jL%5%>!R9I1D2lI|@e z=F*<@d0t!Z>aqJ^%ffN}KAd++Wf+^V)BO4=>bC;zLTlHY3w>1E}s|)xo!BjFYyxeJC1TcQTElsJX!Ohs$IhE30H%hXAH#l zFr?+nCq^WM=Nab^e+jaERO(Uk9nR^7Bkk+xjC0qMRy>jNEor2Er#4Qw8_e zyYdU=n~)?i-`$&p?VHB010q-c{_fsq?33^A+4h<%D$h!|-y3dU7;~`u+Hk8p z*Yn*X_Vpu9&wJ!J>x#G;xa*89_^aH%4ENY&;O2|CwQxJfaGUQ_D4&y;fm>h1&BNU- zhP$rep0NyER~^-DZ7=1UCfuQqLhTF$* zhb;p)UBs<{`%DaXFT=fV8MqZi+&Z{#$8f6*_ts_LRu*wx`-J-@hI_K%-n9(esv>S0 z?n>kS*)GmA+y|C{TV2Ghf}1kjb?Kmo!5m?@k9zL3Rii)T-#Y%b?O58f1g*arxZPv8 zHyN&=(aMyUg(B720qv8Ir0^AC?TN$qA!_J1g*mE1phiq{^ zY?*NV4Ojim7_NWRboufD?olz^D-CyLDTn3E-wL=F#&ENStM`X3PybK_cVrRwR+x7g zZmS}HA7P*LXLi0*WAay~Y8klC#BiT9+zpn2TU*4ffm>I^mHzid!{xL$}@8|AL7a_ zSCqdgxIGN_3wU3n1%|tm=gN8ikLyJTHdi_i&N+(jlhg678t(pvtNHMY;r1=^SIEWn zthn5=a8EN_nXmr^v&nvi{%oM<7JnzZv2v?}JHl|)-!%<)-=+B57`Gno=ooHm!##Kf zxaA!a?v)ts#)f<33UJeKKaJsTX}HI)0JjqEuQA+h4foU);8w$JIpLq}Z3n|0x&quR z+>SBazJ`0@3UF)T4v69IX1JHH0JjeA(T1z%h5H)r2+utrYQo{a#pj#F?aieVE@Qab zE)Fr=8&|;J6x`cmxJMiA?JK~ofctC=_awu;X9c)baNjXpd7e=E$ukW1Ar{rTxQk=B^@clb1-Mzb zcgApkGu-Jbz|F%QYq&?!`kU@w;?_WV9x&5$b)B~p`;L)b!i@73O`_`+I?h*iPPm`o z>bhlLZ1zCo4EK|;r*Xko++cn7Af9JK!;rjxa0rq^^V!Hv7WzANIPd!Lz+QBooW~*zF;$!_F?&pAW9Gp~` zG`f$e2kCY~dffWleJ`IzHr;6OBcwO}J+QiJuF6|1~(lvuqS)RS&KU7=J${ z{a0uqQf|s~<-O@G59a--xc*h%B4z52;?~>d5`a2 z?CX#0_>eEkhjJdy-ko{3mfHcOuR=#5<@$OMzIXRz;w7BUwz2PcWZ`aM{5_BK;_oF$ zx%-Bs$lq&;mvA#%eV?zJQmAWx=L6h6G2FWh_ox-%mhYHw2gB8TkngrX0`qah9pbt2 zz2|YnOUSXEvM%={tj+b2?F;vvNqvy>)@ryH7=K?U{d;I8(tNl)B)RZ?$$t?qVWmSF z)(5F;5;Pwwc1pNw;A+3PZZE!9j?L;wx!X#|Kz)$=Ew*641bKgTWBp79?kJNF>xds{ zZ^QNXkpyFi8;d?g@1U03axZQ=DzhG3Ocj6kCLjLfchkKR?lI%9yjN}uv-kT>(4g-s`#YeAgD9+lg&lKG2PYm<&_6-LVtTq@MWaLK3TS4Rlw-8_a@Y(1J|7z@&D%T z%z00)&>qIZ`w$&|Sb;m!*RLK&a9s(ViH<>eF694=BoFf2MdPm2gXWY&MP7Ys2N0Jrcs}ioJ=KAn%ij^HJwWz(?$Yu>xp}xZ z!_{y)1NRBoo`S?PskhwzC!yq0_MJ<-UW=8;^j=)=%oWQ0QuYr=S)}!e+iN8R$6dnr zy@{WJ-bR=9V~&8N+&lR3OZq`sC(cm;CK&GLr2h%EKB~a|%yWk_L3)z-Dd+{1UY~Y@ zkPk-zIR{Nf%Kgf7-x&-K{ekMyQS^gjnIne&e%#RC+@T5A)o_nIjlN-JuJQ9E@mJba`Qe=3z}0s34e5VFTO42DKIFM~pTKin=w&nx zh3}`vXYsfF?+xd(<;Nsk6farb1K71D@pXF^2n@=T;MUE|TzX`CA3|ml%I9He4Q& zkg#0-*1~NzvCuALo-`b0)^NEMIpi+SJgFY;28P=e!bq5-40lz}?G@sqf9P@>w2P=V z6y`JT1lB)dxc3+?w+2cGmWNvox3A&KI?w|!A2!_0J-7IJk8Q`r{-)vX16S+Wlki?f z3k`S2&|MeE_Y96anKdnRJhJB{qj!(`of>?2nDlkLgz03%eVFuPQR0+Bdt=HXA=rlt z9v>0^1DcEEy`6f$XUppva$R-8U17N5Z?Ed*_jl+Xg??T9tvNN}?lxTU_e{9U?eFys zxphOhU(;|W!JLj#rxx;IMwksQknaxsL41qTcoq}wL;oQ8(Ebgci@Bj8H#IcjX24BF zw(dyM-;A1{Uf`bLxow9q=Rv!m?&wCwiu=&4DV+0ki^ffT&L(|c!u?>lFOvQ(wEh_d z?q{Cc;!M^rP*=1Snz9LND@eF4Z;br?o100vm0u~;v#UsdBN~OY|Gmd^<@+#C6F&)! zN47sJ_w!sG=WF4n3|IQeH?eseDfe^FU9NueqJ-PlaNhzZ+z$-*N6(dY%sl(QVxP?K ztiP!ue=Fhc4_C*F1=#$BO3o_ytMA|VYy-a0OuU3G*|v)rFFs;U5uSHlOnZYX{g{L< zr0W@!aBh zuI(SR-_iMlyCmUmGhF#jl5j6GT&5fn0(no$b;L`!m2F&qrrb)nPr=px>@Lzji>4v< z_c!?iZnG+`zYZl|(T2#Lr;fQGI{(#rmgB>Eli||shT%y5?TyXuNVzFLzK!r{j$_|3 z{B5}d{k+<7Y%(uNUzTv68-LHl<}!4R;r{INLB4PF4DmU%&&$r`Oh1|8p%ckteGf^^ z<*eUKD)>7Mo7?%p_Y8LqeKjFqdF=1-Z|j)u3g)SX`zgGy&^*Jf_uN{zzYs4$p396Y zH_e9}+#!Z5U(aNYPZc6?8&6isn{<9(R5;kFL>w~Z*Nd1*t zyvv6r+=qs{CF!?8J0Rs|N<&m|5^+P(h3M*?8`iT*UmvtxxT{&efU807gUw=h0}{`~ zUq24T-UnNOeX~~!_32LRoF+V)v zHZ$B&q<;uaK+0Xm=c7EIJ%{*j(L~~GKAu?Au4+bbKM!0jpPba4^LXbOQf_b0y>?jS z_uy zgsc5lS8R4i!;$(s#QR%I-!uWeho&OQ-)t4%)8HAks=C5FLAbdaIgc>;yAkzxYgC1l zJIQnB6SoK@GUynd$(Fq65|!J>Mb}Bvw{Tt3_$%Y%Q`~6N4O`{b@cV?ICvka+Bn;&5 zxPCy^*~H%r+_K4qa@&*iRp=C?+|Q`6xE6ds+*fGTi`a+c{Q`ZS0jjelh$I(?@GQmiF^z;CN7O6Y{xdvU**=oZD+VWNWTkXS#P$=O?d8Z#O0&E zmzodCt%uvuaAkZH?jXZ$>$w*Ye+iOsC0p^g__<=`R@~0>TyS+9ljr>8xuzowx0A0A z^8KFEGL%1QWZl`eiwZwa(Ed9EcW<~Uae{C%ytC1bhI@Ja(RtdMp~p zc4A&5^`YjDggXMRhGrMD=7bJH(hW+vt&X3BrYFyE?p5NaBm3TpJ35uR>QSs$7|vt+ zIky`(7_EFsq5r#Z^^g;O@AS{-8$Jh6gZ*X3eogE)Ky8fugvh@1k{M0hzQjv=@4-Z` zx(DsDCgJWf_6M^6Bs2`^ag+NIq~J7p&HoS^ z_2(GFdH-MhsfJStr?ugzKc^Xv%ykJ(m*!6fPBonFaI}8ElAtZXxB_W={5IkQQ z?h*D)CY{ZrTHk+Y{Vu;R;VO-PCv%P>`91+#9Y2%I&k14u$*^B}5?O)$QO14}c5kAO zke1s;l91z*^PeTeH@z&XL;6nFEXS9FbE@I2Mw(Wrdpq7Sim|kZbA*B)H*j5Heb1<3 zjxV9+ex4JHvEP_06)gbj%^5J?U-JFyIw+%=rk@~8~p2zND>~vmoGwJU{4aR3@lVe8hLP?Xr2b{Sf5qo}apUCG7)d{#fjhZ~do#>YhI>yD_hI%GUoXk| z<5q4R+)s+Q;_nlN`(zPUGNG6&c_~4;sYes;=NRrd!yUH*+-kV3xW1_6FZnC^@Ur3N zinww4P*;=>Ik-LGCgtD=)8M^_78vepp4;XMo~cILqFwH9xXzzuJeGXWc3Js&!tG_a z{jjM-`yj0kJ4ZjcT-W7l;2s~tJ=Ab_T>)-A+^gYA`;#E+LE`UV!}a6hKh~k+>Mxg( z!unfLOaElJr@=c1U0}F}Mn4&>%s6l*@e-DHKCRpuxVaeawT64-3UJ*M373!I-e$Oi zSAbgucL`jrXAdyV4Z&*(c9g+H$Ke;?~3M4p-aTyJCafm4*7-$#dm-k-dpO7F8j+Ppsbejl(%Fvi(Et zQ;g@v--}6qEqVxPyWsXu3Bjkeyq1gjpU^ziuB~%J+edwFm7nkaoexhl&otb`Rm^u$ zU!+`?$wIDtkKqpDpGFTM;hwm0#4UfcVLPsRHsKyH+;>R-G0G$5KH|CKnsMC3OIXBK z$1XX34eGCZj{D5vYLN4ZzpzORD~vl&MX087eWnHR>mbW?r!hZ*^MT>W`%hA&+X#ud zwC8Br=3>9Qp22>$vG0Ig7jziX@$X{(U>ekPHgWanHzf0h8#({UNS*$OUmDN-Nt{1w z|FQAa%x};(Nb~9F=wO3Y?&CfX;txQ5F}3w@s-G|D`e7a1{oqpUVK@q#bI>fL+$TJ@ z?FiP>(e9`M$)fJ;bX2Y_Z;HC_n zzm{-Y!qsx$7v3@Gc*EsU@-QEgOaxn@-l!XDzcOv$gXsA4Mdwd7(-LkM!#$ex8FV+& za{JhG<-U~PiBDb^%}?Z9ekfzA_?j=OPwBT4?f}Ex44bZKC#2ltXeWeV1aa4+JJ2mi z>y>cJ4~*6wwA}J_yuSr*Dl}zni}Z7UYT~A_Rqjdh2i#|9pLyaX%q`*yTf(`tSE{3{ ze<$IdG~9a9|0&Gt3*2#@`|O95FYz0q&akb&%@;@cacpZpzn;!|G+ergFl>%ZH`D{E zzn^*TaN=%3_o7i~Ed8IJKXvfj@%u;I>h}_EE?jMI&y)Uj^chla@8~Ckm2coZS7;N| z8VPd}Z23;&K+o0py;i-?y1}$U{&pjMFSIjK?mC_;=fiswFX2gZULECZ^sj!uE)RG0 z815j$ZMOp4su>B_7H-$b)*TH`{5{!lx&2E*c%SXL#7kJ3zbEV$Ht=NN9raQ6vG z%g3#n$^8N*AFhFS+^5bc4K z>+>z_r%omQ5_BFKLf@hHS)}~9^@eZvRUa`<8-L|Fo?+OGK*~KNI{xrH%0BmL=RxAL z1ogKT?%O6`Z-O`P_a<(X;U43;FQ348Li~q&HwoYOXt`E%QaY-A0ru-q<*a%?QA-r4dVy#-yvIJlRG)TB>m_pJrL)%CQhD>+@7rj;m>Bi z&o@yU9C!MMgnQc9?da_W7`v}x>@wKpjNR^}Ka$@#kgXmMuhd6xt`57o#_m>ccb~E2mEEDA_V+4&;`tbf2z#C6~w(S4~|PTN}#QpIMY}2btg3{$nmS5~QA= zFMl+3iOLv;?^1GQB z|0S;_RKTe1PHQj{HnZF7ISw{Lc%7P!DI4;oOY9^gp*7&dO3Dz5f-zvTkbp zmo_3AXVPy)`c?Sdsu=%?3PTRgkA|c9ztC{P_d38C9K+FatAn$0orkg!-$j2K&S3GM z{H%%L==huZjrD>UPMexSIiBY^qhmO$dl#zVYz-&v|KAsTQGX=nuc13V=L+I#W6Gx0 zUbLq_IX^L+-$*OpgDJf`ny(3GF&x@!A|S2>s)JQL&uR}#LOr~F3~wFM3$Lx=F|;+{ zP3h zgFG+m0(Y_D4f4FB4DS!m8%*3;s62)@*z>BJ1nzpnyTJ2?8Qz~oyi^QtsOMFd1nwTg zyTx&8xb+S1BhUNB@K*A??}%IH-V#?6!&~fmRm}p|(eP5Fm*efo zR*yGP#H)?rCHD@y7nfN%aJ>w#v*-0Pyr77ei{Z8Myqe~L+t=`R_Pjxcmn`Dd#qipD zUT&4Z9cOq4d)~>0S60N!$MCv(UVhcUoo{%jdETXl*Q|(FAH(bCd9FNg*BRcGo;T9) zRxaYn3D-aRKhI089=Q7rZxo#M8MkT;ues+vP2AV0Jcc*e^D0{e?s>zjCw=9*+*5O3 zp}wr*c?sgWqf`uUsOM$Z2;AF-*NgPR+sE)$E#jqPcvpH}UCY4DHoQTecaq_)R>Z4_ z;oa(ahI!syhPQeVuPTN&R`ah_;5IS5 zdp&Qg;jtVM_Rm@#)iJy&nt$sAZU=a+qUQYfJZb_whK;a4T@#+1=YC9FYTdx?5tIIN zlYR@6zMAxhlRoA7GUpXFxxc{g>C>-FT$c2w8os74H|hJE^tGhFi1ZpX{raSrAoC?I z4Tkf;T+Eznfg~c-!MOoWdGue`9{axh-Pv%S_M8m)S$tlQCV#@PUf>>q!zKPO^u}K1 zTe}-h4o(U4+ltly$!YI7c{ne^(fO{PFFb8H4@+8^-^Lu*JpcRB_}?`7qx0<-4QD(Y z?(t1rNk371TpVf`%HezvaWWF2oXLih^ZpEt;ph@%3eMLt{!BHT$#9aK#|?_%)Q8^C zp5QEoBXv}Q`ZL{dCYgGwHLWzbGdCVw1juNnb1#>|e8MX#_n3B(mwklPhJjmTI9l$n7*10;_x0tNCVjK_3csuAUo+_w zCVd6zH;769rb(YP=_^UUZA|)iP5N*xLF*OicZ*5?kx5@}_|>F8DklAxCOvBge;;p# z^ykH-pKsE)GU>CV&&H(x-lSjGq^}|U{W0l(GU?Yh>1#>uGEH z^iznVyZbx6YZJJQO!_Yv2t>D%UfPAEAK=s9F`7BtMuFRg^x7`&Ax?BZlJv5+L~{uF z$3MjSLp%Bz(zlX0f{Y=ev)L!<-#}aj3)7Ezm~nqo#&P4%Q^d)G< zKerP%IEJ+uyQ{Vc+$(TWl9uoQX(gW@Wh+5AkH{Zzs_0LO+gCMl60&fn8-L`!5%K2* z3}+7Z(r#xGe*Btl?<;>TEdS_iXW}CWe#toHU#(3`eiy^fes0 zo-M~!8^a0vICfXUx!Z7d!(P_^_c5G}J!fnTr?dB`2F`fH8G^m^KNlKK7ta|T!`a$% z+*W~m&v3L|JZLz(dd`(GoGm>k4d*Arc@Fy{*_UH0=F+~ZJZA=R3@`s^W3syjPB4o< zoXf~RT7Nz_{tWe;Suu>Y#fXrD)4_1GADeGD=X*{*hSMs-xOzAT#BdfFPWZkA%6(o8 z$KUhjQXQEO7>Wb*Hzx1`WTMCr^HpkdDw6?f7Uh}9t98U;o=yMzvtRz;Jjiu znm=s}=T^_j#gzLtK7VTAd~G=D&z6QW%5$c~aJKcFIykFwuPGrT|0riW!@1XU>S8!u z#E4K2XFJ2ud#e%GgbX5`)8lgBQ&MH!08jixyf+C_bAg&lQEnf zJf{lI2{D|z4QH13r#$Akc%?xYvT%kO&ZF2%e>j1ym`i^+-*XlaR~yq-TL`IB;6@tG zPo7iqc!9H?bRis9nz*qsoHf{84d)TV*^aa_Z`s*!4)>h%h|9%rTEeM^Gr@2!@to@o zhg;b~e(C7pKm>id1 zW6G<$=j7qIkHY_CPpgXIHuPgLoECH*A(GG}g3eSI~`j8L7fy5n+PDI_hHhjN9}dT?g=y&$$N$D{hC!p+*-K%8SX3Cyo)|W#oRB6`ws2Fg1>wR({ihexD{Oj zcX|wW;c{{_Mcg{Lw;1kU@T7engNprK`3dfWM-p=EwEpUOg640>9XLJ}%U= z^GLt6e7KB#S>Y{1KIGtT3Aa_4n*Q0ZclP7|LXdc0f5ZE~Z-E>4KJzdUY3t$jF@DPZ zz|uzL`!U5leNNrpb8E{9oGb4ZxC7y}m;Vs%;P>TxaB2R%zaKVqT*cQf{W}a?1-#SX zX^`<>)EyQ1SJKPxQ~&s{tHl1Y|H8fs`y2lY`)ceT#-1zZ<^<@W%fS zdA0DS{TIiV!+!REVPA*+@Bf8;9{aUFiSkzdUEcapkNsxAsGka0S0{-NA@xK|sl-nTmi zXfgLqiSR-8H3N;Z8#&a!=b_@SzK4px!+B+6T)9x#)^PQ?9C?35F<0MjAzazt7`Fm$cf(ztd}yqm zRl_~ma5th{mR8T?eG!}ca^vy7M(b~RPv)(LyES%l-|cF-f99{e-?E42_Ga6dzpfYe z8yoHb(l_SszMgwYizbb7Yv8_VxXaT&G*)i4aOW8A^3=1&@}UlH(|n=+{#!fFOXVX- zJ8q1d>dpPR`DN!WYyVIUciaC1HwSkQ!)?tXN!)oVhaPlcUr$*?A+%!KSU#ltaDPd@ zkPo}Q(j#Si}*d6pQZZ;t}9$Uf0BMj z<{aPc6>;C@Z*KRKz=0Klq*>Z|ws0%p?rHp0?ilZ{aAn<=uuR;GI7s1E!ac!oWqgyd z>v6*!6k=WQ6mc4g-`}KjYJCq|7Vf2ntNy-dxWheHnX+A4J}9>aZcR-7PBh$amco6H zK0^9-^;NdIo>vQZd<=K8;m%(HZVv7&xUC{v_XfNV(8q?`(Q{XNih7MCv}Rk^B09gy zPAT~OILB}$`||_14r}tac>W;up$S@x`cOPykE>@o|1IA&a4UaN@V9urW4Xoioj9&O z=U)kTOSpPIDD^?>gtRxy?Zy8!EUiAMzg2K|Hr(RzSh?^TvKG4I`v-a!@Fp`s1gLdqHXx^(&~fu59M%I z@9*r+Zyi9E5NOQyTO-* zd?;=glD|Fx+k6uGk&t{?ZqkX|0PnWpS=oR!F(*9S*iPYoKd!}CP7>%3tdj&3UxF3>U z)^@&TtK1VkcM)+V&qV&N%+_*+EkU?)A5Y!hf%_G1`^eU{ApMqTYoy#>p8F_qkE=ylwf6OM>`dQB3&;ZmM$r@ij z29la?Oq1g-|@(+|L=3YGG=aC;)mliqK4Jp1f9R;~9n%84aQ zMq^kB_x2d>sfN4Lzi@L!+%a&g;7&5!TITn%)_FEuDIxKI)YFUS5W!~N59qfPYh()w@ZR>PfVxVK|A0_Ft6UE;YbkChVR z-IQ!cv94UjnXig!2`(v{YIt*z(1e7Hq7)snFXc&_7O3SUL{@eVmfqS9h z-h#~o=uxEHbv^fO;yy&5qkq(gqWrBsg!?ZHcOEtiQS)&HF0Zx<{q0ZOKIkBH&2bI= z9lp5XeF2q+vMz79tyyy&g56n2{iT_Q+^dMY5#5etJ=W$!eUZPla9hkR7wlwb4B5oG$jc{9qhe{s7YxX?v z-A3x~AVH|x2Bh&vKV z=)yz)wJ>vS|H+m9p&srRCLg*IC++QI!@bROg{R@)+M7F^`{d^p<_V|4yC2On-0;3W z+QqsPxHknIjSfHubDgw&*Qh<@o`~LWs`I7tV*=j!RN#J0`uXTDr1|ix_qW{(eA^Or zLB;C~Res!6ZY|tf;cB_{$L1h(3{q}cl%>JN#En4rqLIjc=X-}2aq{7!QASe}D0 z+{Z}Y^F`kEhLpRe=eB<8hVgEXBOKr?f9N4s+>%M(<8oqZ>8!~>I2-QFX&H5|08PhN`d>Z=dMj$ zd(;taKE7c)F85p=&#O=4{Li02i}{Y&^hX1c9DW}8@vY=w;*Lc+mb1&%VZR0T>etEa zKObd~9`B2ikouG*?rtPO`YoAfRISml{;OZLr!#*s+%edVM=v7fz7djKFrBzBkp%fp zNn_l)A@tvHwLX1^UD+g_Nk__^?YY&&or^9-)6e3aq8MGujmV>4rHr~qZ$t%>V{B<#U*JfW^B;nuMaaAU8 zkHXdU(c4MGVH(5WH@Nd_V@^ELs)o@P#CT<(-d!kYzIn`2b)BT7$4pk!= z)8#&iEMr@mI+nSJ`TchV|Kxi-;_LaOmHt3DJ0L=EDRDQU@IRbi+x8XpUWM=9lwTaU z)^Ke0#-qJ?MyO0PkG zisaakYTP=ax)iY(Sln=EJt0doywOAPLW~{kao;edmVbsh5B2573Vo{`$e{HJn6qi zE4@+R4)NTJiMs|pita~apW)n)%$IeaKQ4^8J2B^f3Wq921a1^u4L=;h8a_5NkQ^~x zkk+e`&xor>cAi*^{b=m9+!wLG>6;}kKytid-y0EvjfmSG?T-4Qdgfe1INy?o%&Uvq zecd&T--det>1BQ9D7MPoE5vf1Pu#WWX4Ds_rJfF!i)e5&MckTeS(k|6-fg%CEdw`O z#I5E5gC&M5jXz4ELNB;AY`&2Dg=W<^_0D z(dUMHuIFwywZt8Y&Oj%jOQ}mYBDtP*VkW9jdOfS|`oR4Hx4q|%B>lr^8q#w6(sO%H zqt2tp(Sv9z^)!!!J8P}zJap#X(RH%S&4Fvl1GegKp7e`Q$y){P%2CG}Y(?BI=pa;y z>er$DQ`e+DHUBl*>O8Uj7Us2vdkpE%K*Nyw%WVV_f;q&Mz0ErcP%G56Gwl(HzqS6n zLGz(}WZ*7m zv*fc;K4dR0l-sQwd*-gdC4MaUdm`!2M7JRI_XE!@na=YtXm8XHow%=aPrrmuufcvS zs%PS_Jhz*_J8-SwYW`kI`fJfhq}<;6@BwkZ{TyYAtk2kdkhUQ~>sj^v+@Edy{Sn(gkbAGd zJnIZ5OL! zmqHsM<;vWK5cDPPC?w$|wsBm&4w-+5>#>F_@8Lfao3oK}JB1|XOT^jt@W-8}YkNyS z%zHmPw}K+M7TZziDZ|~xb7v9v6Iy8oWx!V2afWfTD;GS*P#;E;PPsLY@ID&28f3g! z7rQn{=7};M?}|8|)~55%>o~qmL^Wc`5{kjq!u{EBw}7_|>TbCGJ-0t`hoYmA&J~1} zE8=G1R*qqO`>D{cpNP#_=p3Z^dn8;!kR|SJB;jGUIr>TYHjVlzTj_u0dqU-p2Cf}k zJ;KOVh^lzT}?a=|>}enk>qp)hU#t88r-dAQrc)q2)+CiM($gp^wol3Z{;aU;>a zXcT=yySBVjccaMPte=l+xn&;b`6lD<^VqzOrXuCO;kk2&`vH~CqCK;<^|$Wyh_C*p zYPm1SbE`PSX4s_A7D&0DdG3kC-G!b+52Gyp{@gBF7n@2wyIj(dx3a!c`()rUa5YGM zco~~1NbA{bc1k^4T7AfSUp0@z`tTI@1sU$*t?6=}Z{psADw`mboH<|anva#w25!*L1@0nj%0A@&A*9?JJa+(b$D-5F zLE|W!Havgr>r;IZH}@RR$H3KmxDcCb&_hVM{_&{bPvTnf?X|7Yrf5)S&J}0};&1BU zXk9|{p=Mm*x-2aC+n@CNpu>=IfA{{LOWfsX1d{&SUWcqdy{Mi(&%DuaZ^hVFCC!hhuZcBiF!JQ3P`;m<~pFaz`3y^Xz z4Bc|U^~61d#-S&XZSUotE8~@XH!d@Q)9XbAe{-aN4ZVkydyVIQPh6`{xR!w0Aj_@v z<*xqLyb!oG4R<@z?}P>*BpltkaD;5 z+$F@V^(o&GK^r2=&3J#MKb3r_e<^Th8-JzTcfh6sDR+?P4j}F)_DhiOtXr;>nFP&; zbdLAW8ScK=9)(UZTuz}S1lJKa8cje?Aj|Dgln?G@-kSth!)v5}3%!G++K1=i@(1CI3+>rZF$!U6_sxk5XQC|Jd4^m3d{XF7gbO--#&hmS!p>|P zoA*?_f*Xqq<^2fjwq)tw&plc09N7+(fUvacf|jJTMu`f z;TGS|Z}Z`J>cAPkK3woG^&#^r{kGv2KmTR96xHA1-=6=fgZr7`{;%qpo6K{IhI=)~ z_X+wONj;N$L|%@(3|`N3UGwv(z3tAnk~zRW?W2A*>a7avvgxpXUgHW{iD!No&)b_kT?bpuhu=MSJK`ktU@On1T7T2N+>~4SCD&;TxA=Z@ zslQRNHOvQj{-Du*67EHYTij2wd=SfBT0g1dTP@sY47Ye3VmkddSH_{fJ})LWjrUi% zRkK;&fUEUkAa;ABqmbn9Q*h4?bJYds5_dCt0NsfsfAxArzG(hc{uS2)4fi?HUqJCp zW~x1Y*Qb$v@C9+dqLObo z7j507!Ff=nFAw!AJ3nym!cAd&BD^weTA>Y*a=-T6Zo~~jC!=H0eFHehnn&L{kMH@x zmG9Z<^`nY!12^Au#rzD?XV4W$xxaYsUBo?xa_Bj9$J&K;mTK>>aLeGQzvI2;OCm%0 zSJnYu!{#ld<;?98jjjV^JX8BXSqG?t+W{`sD-0jP`x^a#)ZYZZPY8CI&zOQPLT96= z$%jcucv;W>`~39>uKU4Fg{Hhm_$02+@vfh%uv6|9o_iBRjn_-aa z^2!ijYB<&`oM#PZdHtxz{!{F=ULVOXiFCdz?Lzz@G#JnRZN6LelO$&z5t*CqY5 z9MFpP;~BWy8gB9RcI)r*TyJ+j^S%SaJ%c%rj+I(ZG#@5WMsG{~AavZO@p*qO+_T{7 z_%H*z`RIEjT!ur(P%9zm!~owD^+VHdX;{`{iq0Qvf8jbVTn$HJb38f=iDwho<@c6c zPTWIC>yFq|FXH_pnu7n=vi~JC5vhG!Nl5+uin#J`xwjNYFI_*1%rE9T1ol$CT`_3Q z{vFYtNRCT>U*^h$U<7ev&tVAw+7hY1EG&Yn6 zJcPy|^-t@`OT@i_B4OT{z+W&hlrcnfq8lx?&sbl z{=7JoNV=hhyC!yN)EO!FZO`3}xIsw5oVM}xLEAx9kaQyq_b}|_IpK2*muZ27;3nc~ zJom*q)}`D2^LYg2W_T&>k4|n_*XhdK>Q0XkTzRal?^>o7fJGnV;#ruYTpE zn{E7+^V2)9xeKYkmwRrsVa~_C;`hJQ(MMT-)6J9a55tx7)5oxT!f=Nbapk`N2y>y~cJ$okY5(P`CS7mi?@B*$-3OH;EjO8Ci@zHXw;7Tk&m+ay48|yu;Cto-SOyrq}=`?$p!ZkmqV|kt9ND%k9m-+ zi%EOS`+8QBh}v5{+{GGq&pQ(Vn{%al0b**XD!c zI;^w{&ELA!lkWZ^?q1j)ZMa#_y@a?L^bnGKIFUT~17)|T-!J0kS|r`L814ka9kmSH zR1vp!jij4qxF5p%9<92tkPmaUJ`i^tIu{K=GA^FC4nIn{E%0;APpE%#Ug*|Lx`lAH zUC8@XMq=|Q()>+DRV#RxxO%h*$#qSeziIEU_LF(IrL0BixTy84*)Ij|J~Hv5U8IRy zT0N`qTwTYjTPw-8p$lB~cQ?a*$8%37ZfX8j7IACWPP%q*Q=vBl- zd#m#P%K5yEZz($R9Zf#W!sc7ld~t!>W2MO7e#9M(s?q#kqwCySA3At{bsVo+FX;xt z)$vW++qH(fhvzod-m1O7+TJSHr$00PYJC`QxbppVLZkH|Q&esht&{Eq!_|EF+;B7W zQNnF39|jiXLuP}dI|ptmG<8nPt@W>k{C&W48!NYLQMqL|Ou8$LzgmBHHQdiUx3T(L zQR6P+z3& zjavsL1eXw3gC0TBPs;PQgEx!%^%H%4kn29uPu8|ex;nV+eSLTXn|IM1q+EtB2|-gX zP_#jti^g*~nY9BQFY?}BdC#-7n$P?oJ#RA!w`}i<5TT%x#)U{kjK-{tDWF*gB+Hol5xhrk!^LMkPYYVrl=U#-(FmxwUZYy6O zUM21eG#{PXh5amqWLZCI#d>wxb32Xp^PSD<*WqdqzUVJ(CCp@3IHvMm{_?UY4<#Qp z>=>rvpDmJZuNY2~KMNc#yT~GBqE!rMC(kK|b4m=Sso}KroKy^_!gEq^u8rXYhO?IE zw2$HR^qe%D$6`3gtP|y#b5V>mqy zEaY>zR7gHo#Blm}P8FP1eqmptwA?S)vT$7C`Wc*lF`T}hQw^s_3}+Q{T;VN0a4KUs zJ9|zB&VezUa>EIIhchUKvy10s;S7!8v^1Onk`_)?45y#x)WEqphSSP$WIa?kgJU@T zJ*O7Vb1|It4X4s`s$)2zuk6pkc`t^OG8~$A=+Dp?j@*kZp$^V3F`RaWbGansxH2)E z@R-@1hqGo7)fGvu^{2hzjFcU4u8iUACi@8Wa5}|sIvCD9k`PWdhO@iulh7gQ_K4x6 z4JW*Q3Fp=r&K?nk@6RRONim#l3@2Q_gi{m43Cj{z3eMFroUVp5Nq!T~=orqP-k&s_ zM`JiW3}=St)W&f3@|+4dQ)4(4hV!B4jE&*!?KzcjzKP-VF`Q35Cl|xn$8)OS1pY=- ziPHMh&v5dxgZ!Bi!`WB%5vt*A62lp2II|@ooVpm!ezK2{fzvmJQ)xK!Bq5wxF`WHn zA0Z3ps2I*ZhVz{ygp-fq43d3>8aS84a0VGpy(EM)FNSk~>?72|xjTk)u;DD0gmCI( zI0wo;LJrQ0aLS|qy5jrC%yq9|dF~&p+%oAtHr(R(m0GU<>2M={G!UcbI>O!$2qS)M^Qm%EpsSY zH+rxazNbe`p~#f1qM+@K^Nxp%!e5zdtnG&-ZmcS`W9U;jV{Gd(;sXbGs7P z8_D-^ioZWp>HSr1W^2Z4xb1yD48ZQPRhqa1JhvyubW=1u1m_cX4H}7VLZi+4ah2z` zrDNbvSJx@&PBq+zNIwZpLE^7)@AllimaqnahM?n-oTsIbr!KU{tA~z{akK4&ePy(eb|ch+oA1|a@#~x zv!I%|bCHD0*~+@V`YS$5(DTZwZjASado}5AN3}?~6`tFpNvS&lU4yPfEg#_74J5n@ zf4`7&t9MGe*3Am#b}#9-?annvw#vQ0bEgpZE|Tyu+Yb0E>o~&IAmtY3Z$;8=1-Bwh z&GYS~{|)_#lzWZmu2E9zHb4@#VB0l@`^iM^c_Dx6d(tl(ZYR?BM&C3oa7TLXgR7Lf zJW8!v>RO@fGpupWjr={hb96tUa;pX>-6*))pOvm&>efJOBlY)?$i?7s;>IHh@3EEV z4<{kb1KH}jlRGKt9)hdw_+!$~N8cgkUhKKeTa>yDk%Ufc-|I-9#&NfrQebO)t35gC zCc@QvrtgaGhMjUpdhY(7c>>#GkgdN{ef@ozen;9{_LL;|92VAxHhj{#tI(N-`=aMQ z?71(p9fuk`$K$zrA5?8M=hrJ2xYJ1g)VfXFhiui~-#vFRajUEm{bwzwUZ>QPDeqv8H0cGQKn z+7`HF-rps}t=cZC4_mGi-LEP4ZD?JPt@LMd+hgW| zq`L>MwzsvhljmI9vXyS3C%?T`#7%T0t{2jLO&7JZFxi7i_k`i?O8V0|rv2EepW8&d z;5g#sJq;3u@V8yRQD3$FR>K`{xaX2yt}9<|xTkvV^~BwYBpinS>aQ|otK2%cZyD}= z*gcABk@|bB=jMoe7fCpSlY4tFjIx!R9nHK9uC|Adu?wFcHrz)%ccCyhj<{uP%i4|yKb*0?yr4W(=*&nLlWORBCZGfCCGDAjny;v2=hp|?IT+^ z3fq^^8-}|@v@cjloZF<-wLmK)y-p)_p@Z)i`to~RR&&*3xSs{C)`uKxjjgdsAw*&GN5=MEEcNoWZ_Y>VcG5d;0QJ zZFEnx^fh(x&WPa+GrU#*g;!;Gna8*uY64ujUD^XBgg3JQR38=|>|i=blj&4<-=z8hQ(T-?`!a5LrK7>AbV*$)tM~ zuGaVWv6+qLBIVxdx$->gPsB@D+I;}pPvzl$2v_UnZ`d_!&-j6qdynU?v>D?r+8Aw! zZ2K(lQF!k~*gl`*eiOK<&*#`SxDVc+-$+^2_;K(S z3Ooy|d|c8sqXKGupM}j&=oh5rb$!H5bl9Bn8dV@$-!mnVuTtM@pHI4t;c0z3h_q5( z$Fo&GANPJ%SgH7Jogjg7NA9FuWcK0Ym2xQ z6OyjGIsU{JrEWFU0x5U8=XN2kFWMc+^Gr5h>wLa$#X^weYvl_`cQ9Nnufwq!j7~tx zo$a}&6L$fUa2Z?W>KG{7fD8@K@8F(b@?kjXZ$LLAAxQr;L4JVtI-0nr(DO){QU}JE z_Eq^J*JBOu71B;eA0Xu|@w|gNFb_mSMW#;WeSact%e|a**PG+Mmb4F|=a3$EM?b$j zXiLry(H-apRB7gcO;7Oeli~bM#_8Ooq`4Q>AK|dhnR(tMxTQR3Z6{tI!B4ZSi ze9RSuhMqyn zz1DN55%($j8m-B=I;b7znQbF~^F{t1M64c++>q)mdj#q@Kooh?_PN)k~?gG!F1ztkaBPG+(Dg7-BC!w>1^fx z+YYeRSJ~?PwQ^?Cy>Ig2T+&~UZbHia+H=PdHwnFoPKFlF~^KDCAIa2QF zp4*4GN+jVRw$@+G3q21n&vW0T;T}W!+t8gzxz~H{BVA}0=nXUp?L%Ld_5E3uFE`~@ zf0cB*8}4S?G47xfk#axs+&8+Gx(`ulH^yJK@AY+VHWFX!{FpA>Hl)vgmvo1~?HV2( zzmLrw=sBd^ZT!4uv+Y^KL%q=s$etgho`~Aap{>sxSBo_hr8&qo&_}&;980bl0=b*0FNmhPBO2q5f31 zD05Gle7Fsp`_O|({Z0CM_A+sGNJ6)+jpsvpjWYKNTrIbm*!_xD=~LjY;kkDa_dJ@3 z-bR_jXh%r=%=&ixce$-y=6-`K{mxkI=aW9sx72Nf)Zd*vcL;G$p*PS(BE6WMjO2XzJ;uSIQ#cRVpv-M=?C)m(W9SK_e(|#sn$9Gy`7W$=BD-dkmISFXw+9^U zuiKL5LG%dHe0w9rxZuxzrLL?$YxAfDvhBX+iKw6YJKwe{b0Z9QThjMK1Cernj;c&B zinxc-So9>ab+f|fojx~~O_#Zm=D1%Xz0Bv{V5|9dh5P`|-+fBlLc^2D^wwqWUSt0U z`%4FqKS=F;X_vGjZd0WGhFxKZ zrGJq8rQZquWM6X6D1VFJ%dh#X<7O7_4*v&kdH*su=>NdY!aeo>z%3t8=7#+rxLLTP z40kokr48zUv_5FL?L^!Fv?nUQ4^T{h;O1#FA& z1E?wD*1??)SKIM5*x2t%Xg#B<1vjy;21)3QY`f5UrtP?DV43^MaEsq7WVs>G|MPF} z6{>~%r{OYmJ5fJm^MU`;8x?}=Z_M93+|}3o-{r$1eoy+H>fLC^hWj!Dhe)2qYRq5R zALp-f>)`f+t6~3g-VIIJJceW(o5e1_x8ynErXV|xRqW2VZ|vnfr8|0?bP~kgUmr@S zotCj|e+yGF4!}7*hV!oBg!k>SzSbD03eIgY9KHWf{CPC`%|=f)WO-__%|PW`99=gjkqN{6SqIodM453m1XYi7{3lBo#tB* z9e$&J<%;~OhkKpjR>M1n-y3SUvK~SR&Li%w`=kHl9XSU?Ir5=7HyMb(a=v^H$62{wnOh&O26=C; zJiB}-c3PkAiguR_Chi~a&BcBT>|1%jS)^G=`it4BUypmgt|9I=B;g*maq9#6+;!Fd zWo~DvJEkSCoY(8) zQNBM^ZVIISoQ7s zQG3HzF@J)~5E&2ISNQDv_H;6%&rSO#(9x_2(dmHrnln<9iPi~O>kRUuKm;T zNXGC~_d$93Qxmj@Mrr-t~?q&ai8-lwMZpi+~3h-nO zeyqc_TTy&9NIMNl@SVc@7^qWHw~DC`-dWqfe=GGokK=`Kzi{l8eKayIHF3C?TYF_6 z&H0uam6J*xOWsGLehBr=FlBgVO8nX%Q(922C%W0;-e9>4N&5~3u3}t5>78gtOipB6 z%Azm18^$j$Fd?Kq6!*(8RqeA_BpV&yt5QyIACu>9kDGFQkMz$l=cXuM-6?PP!={}= zTJrp$kn&eK<+Xo2hrHr(`TN88`=#}dp%ndNxZ`MBDgCYZUkNrbd$qa^g8{w z{mse^56Cbd!`1%cW%9p`wj;IoVasiJby?FD^+&x?Yq*_|*c!3@oAej5KakDjW*6Mb z@^8X@O(j%3n zqFSh`#7=*lXYD(tB`VImOF;D`_XLl2y*-R!lZsaGLZWc zaJ4=xA^&*(<_covvYaKs_bzEM;Sqiymh+awc?)sglIuh6ZPb6b>ffKqKa+C%iIuz4 zas%C2%R~~Y6PGrh$sUwpZi5?&=Qie4@^?U&Bjx^XxeG{Ji6p#DoP)jnk=iP;%->`_ zojoMOjCA~a#|eywTo}q6Ou771g6~6EG0XjyIMwyEu2-0$8D@s#-<{-l*W08Y5H9_I zZ$HQUJv@8=ZoMrJ?kb0y{QW$adnn(}D~9{if56STBg0hY0z>PWd=K7on9q!@sLu6Qnxy<4ISy=M>gsy+(IiX{1XAz~cl-5s1;r4;6?cylT_rd&dBF9c~ z>O->>YqR^5JZ9>~lHE_BD3wlfP5ic`1Oq=P=k;2&YoDL_I%T_C?_~bhwAB9|_-? zVcNshaiuhS<0s1g=V0^O$C*P8B2NVFjc_a5{H1Y|&r@mjJO}OsxDm@ehjX04oK>#5 zq}zAv(M24)3=O8wPo95X*T`!tk-IX?4>cUw4jOk0jLkzcQwEigvz8{hF6yHvbp)!rk*<*gGP_RA4OCe*duf7j7N6TE7mP zfA41A|6kY(w*%a|@w3dj0p~l)3|scWC)cwf92;fp!@b0vQq+fybG&wzJ2J!cb^MS| z{(n->{_gwThRq}+96ubc@j4IgJcoO@#_M9ZZ#dk;HD2e8&M@CN+=ub+lc*3SpEtQb zww&}-_s1fGy?Liz*A&C`voNalH%hr8v>^pIM%wpiANm=|^TLCXtlQ*Ru0AhZI3~l? zhpX*vE=C@#KTFTSUilG(zniy&#%7on{{c52Za2pdM`KepR0k#dw+U%2&=u%1F$q ziMbQFPN95A{)g}w$L0CdLSm_pQvMtd?^{FK`$$5Jcz)Ht?(3H8OIdFyypQn;uGYuz zrCd+ef{=3CSZ+Pi+9C;^iRFEa$2<`)ooKlqllB#o@Evj8V-j4w z*6vO_5^nhZ40El+{fYeY9KZjX1b3R{7LfJ|lJEiXmR8)mM#_{}xv?o3<`%f2cy69c zC%>#Wd_%1E%AAYf+eg}wy*%zQ#HH=WFL*G++zVIxX}zX8k#fo%>(!?c>t(e#p6Ytz zdFqJxAqsaoT&bH9lzWQ9z4s8fdfzn$cd^5508iE_FLt<7Ew>M8Ly&|q#B!f6d3{8= zS<@I_9j?@ibJ2YccWDZHB~NMV>mj(Gq~PA~a9{iv+-$hNzzy;DB|HdEt`TNA+!dDl z6lqJ5gqMkB9Xz?c$vU{K6NPw5P{mNm4@G@xt?)W+c+ zw@Mwx7&-G{cN?{BSM#2caQI z{mUtq;Cq0yIY`3O#MRi#p3D_LbfZ2*3TVfU9~P1SZS*cuZm&4W_>SvC--skMA|BU> z{`Aa}{?&P0WLAcG0j~BBP5SX|T>4+>gO$slm*6`GRtw8*M_k%^WHHaX9aMNKPfR!PQG4n{ctGjds%a7Z*X-yo{PPy&YSpm0mqgj2{LY`y6$iBxHn!D z!o2`)D4v^sq!d4_<=6+7`*;2g&tshb5ByLFci?~Ehsfg@X5xR~hZx+a;cEZ;DgOH& z?LlsPlk-;Ly0WGUs*h@-s54&V4ffizw&UU_GR#|!z2}nuVD+Ij$F4*Yn(QrOHdXxV zJWAU|{*xId=GdD6$ z4KHB7h{JuCGMmshNVz9i?jF)IZs0ltB|n$m-_EO*8-@Fvl0lTkyYTsx8UT})a> z)DYoW*s`aUxIs4$>@DS8E34s1b+5~R;(vuH!{?;Sdsy`MNr7C- z*Q30x!)86@en)$e`2SSOPm_$)!*gz8TtyOQPA>V}qyD}F?=g5j!(0Sc>$h^-QC_)k z#7V|?&p+VmyfCsn!}M~v4^md<^Ya}p(_9Jh=OMO`{ww+cxz`_sw!R3LZo}lgK>dar ziWfCzB%4j-d5C=u_j1c^Jb?MF-Ham7L%3Y^!xhZEq@8EI#C~#z8>X!EOPz_;4|iGa zb)?;bBn%-w7*`&vDO|yRK)8C|^2jgmOS+d>x%rm+4r!kt3GyA9RNMpS?UfAkgX7;X zC@1!AceqO}w|*|ikOX-?CKXrDn|U?E1ZySiJ)d&&ow07jYVT*3`!s3KS?&trR9v-} zm&%#yaJ8PTB)`<(FNl?!B@Kc4(BNj~&Pc++>aW^cyeh*qb+{K$t_$jigj*lxsd19= z#Yo$Zen-2IYj2V5N0l3SBg5R__#t!)`zcXdq};Za`vGY&^ds7a+<9Tt)?b~!h2P9D zlN@`0A%Eyr)-;fE7g}y$U|DkiSO+Y2PCW zKNGw5YP-;JD*L?*^RnXybDLMssuL^s8CxGNA#EsskJ1%#R)JQ1UNE zF{IqHvBdtwk=k-H4LhoG&v>1g`cwX+xv^y?0HK4tol{yedAU$NZPq)p`;Us?4uh-f|3#GD zh_*P~lj0>5KP2 zvi{??H?6-qZwbSF7p`6}odNFx)YakYyk$6PQ_vzb7geGz)Ux#a_vawjLo%^%^NHjFxol-tH~caYX)cv&+N-G(Yri9$$x z&^qEhmnZKx5POS?Gt6-g_a*XgL4P3SuD0BqJE_~~WArAfL?az&>sj`BUi}sBHmXeS zFBzsTTpcfdC;u^b(XNnkcUo>=(jG=j(Njpi3#scKIkp{ZdkgPo-;KkSIpLd>c^fI0 zBngT7kjL-gxLY6A*?Okjyx%iSTZbE!Sup zxz||k?WEm}#v{32a=8VzKB&FnGJey5XLYsTk@t`ee~mek!=02=A3oxkJj>&9i>-gP ze+Z@dO?$XBn|PSQ9Q8BG6+2w2x&&X3k)99w5X)Rv=l$Y?T-%Oy+$=2XH`h7bSx@lo zV#-ZG>Ib$q#<}UEJZ^wk=7uh}Ac-3(=QqRQHucJy6De~Ex&kTpY|9-unthsR5t@U} zq%XS|6|dp>l0Uur`w8P_G3_|&_nTR8!H=jB7wk3a8 z)Ez1J4qMMQk+uy<_?@_z0cu^1l73i&y(=gfI^J)-gB$T|GG)edJ&!6OF~`vgh&iny-h2R(P+Qy=*3kqMvR zH|cc~_2EbI|A~&hC&B&Ba*w>1^(u5aIvJe_SNnA{#Pf&Lhm{zVQ`2vbf*XokVN6T% zUxj)gwYQRY%vVI(CiElv8oBMbzOBC}a^8AjM{#Yxxzgd5nSiaR3Q}%Y%NqdvK71G4)el7dFQ?GOy05=Qm>+a;IDFR?>DN344fLuGX`+I(vS|ZR|JO z;cEYD^0|+ODkA0nV7X_Ib{@JIwV;l={bZJ{XOoZcxZx&#Q-xxqfrI z!+nDMD^L`vy=0Z(`;4@0XfN7@Ty9PhH>a83+ygh;D{BIidA|WV1u55#QNAc?AEE8& zOC;PmH9dR#+x}rM{iN6%4g1ZT4tF>C%Rj*U4k@>~_cwjxNqZ1Um`B`+vG`h~Z9?K1 z)W6--%h*MnH@KmAZetda|3&mVQZ7r}5`15iR*cf7@cbXK+um}mA2R8r_QEZ`*l&(I zHNnlKOas&$DfeQ_t@I%MIr<+u7rES+ou3K!4GQL6;x|*^YP-0R{0pz=Iay-meqgz? zNP7lJknc3=ckWz2=zKlDo!>kOSKIM(l-q%RN6Ia0`-eZKa*t~o{Tb?jTze?QMjKv_C86uZ7BFqf_snUp8hcN z6*L^>qA9FD47Bwji{}Bv9|zu--_>vY^%C>n$H~7Qy@%A^49+jXckLsL6=({Yh)VQl zi@koGW!CsO6zc9bm%`QawwnCzb15=E^UO8AbsXD(B*chITaPM+d!1wNx8#@Wi`~Rx z?*(x41cbdu{vYpFLlWe>)ULf+C)D189`p}zL-E|QpP9OgE=0<`$~)#8MA~>HA;2GW z>w|K~FqX=AoZHiHMmgMk$~}!1AmzrNPsZMDr0qfy4z^BvV7F;v^oCB<)w~rvw?_QrUYi#)!Q!xYHf(nZkQ3&9rj3PsK^bcOz+ok%Z*;E23;eE?}owN^;giXZyzD;FHJcIKo_D1^n`IcCs{(eFJAJGn^ zT$-o^pKk{9Hgpn7zTcH+`!lT%MQ}e&!L3VK?~#C6ElGA5rjoM?Z)sS8Kj{W9`=fOa79n{c$gcX*WhzGxI0 zhGgGPlyy8gzb}pO`YD~q?MV}$%zHYg5A7dQ@rOjeoBzE85Tc3w=#cr9;^}&t#w{yJ*SL=IY z%Ctq7A?23y-0ho8+T%z##u%ROZ-*ITGLQhPUAZfDZ2Mg5TU ztM2!Y3T%Cs>jtqmZ-n2>OTiuFaAW^~TbRVnzT0n}huhTJI{{u~9`F5pJYg?gums=n zPqUVRo`fwNdC!!fhxjB~m1!-9exsFEF z(VUjtV;<*?uh~7lSl{y)dBkt}!qxV6F8N!bE=akvE%yf|jDtwO1O3Lo-ElT*{j2RQ z>;D+H;FkBw#{Deo<`Wz)Qx*xn_mMSMw#&_-{E%mx|;aBU=iudQS!VX-*UA+Xd1M&O^#wX}O(A%R$#6 z;kxH7mUP~V;XVX6BqxM$6Jdkbl!(KIv>x&79BE+WK_+Ruho z`OVjkALo#NIeHyQPa=L?=5f>alU6Otnh=uTpFhuIt%GS^_nUo=eK&FMsR{YcM*6$A zTKn=j_85{NEU(3ecN{YmBa-uzykx8jvkrtlrE58;L^_Z9NLkD9MaaOv{n`{(Pt&YB?l z8hwVUJ;xrHF#AvGPx2dh{glq{!<+o(-ZK;az4Z-jL35GX`-=R7^A=dc*n>_+Cn0IK zgKK;3HaN1Rf6ZpUSpYW_FUoh%$lnDGK+4@^xr<4A8NG>CBlo-&+I~j9&mrSh=yShW z>2SX%|DUMp+5~ri9S5e7HV-XA&mfn(&H7il#WBCx>iA&?`E%an+7qe0wd^=B^DVAL zP!W0)Nqxx1$5MZnb@0xc)`$FW7`GhmrR!j#!AQCNEq5hp>ydT<(L++w)r zJAPPBnGeu$>l5~VVY%;<_Bq;)wj!51u4Bpbmh+R}+yqzKaf^4jkAZGO>W96S`z>jI zqGO7hx?-CHhhgW)pSkbtaJy4RzPs3uSnd7Aaz~Oj1wD#ny}}*eBDOv#w+L?8CW(5s zfHKQb6e;%)%l&|~&(RLF4Q*l0eZXGJ)%i{CFMe~h!`)4OvDf!rf?M4?slMY#tAQlQ zd$8Sl=C$HPI}ZKoH}xFuDU@r1&PHnQ`Ig(Bv}@3fNajuMIFy^*F5q4WHx##m>lVt4 zMH7*7ue01j(q2R>k$msMwYQzE586)_{pL4UJKT3EBiv0$xdSctPtr2p=Q(CP(!5Lw2Ziu ze7BM|22DlxqgAz->oLB`IVroqMH>e7d`m612zEdA6VsErez%+BXP026w-xkEm<<-Ixd?QGkfTp7d zk;`3|#Eqo|OnZmBko>}ZnOM1fEVr1nG9UBIHWIFT-nJ)k`M9*{1GlyH!zq+$i*82B z9cj5OHy8mKSVefjh&a{fAeWDH#<7w`d-=h(ZI{JYSvNVz;JBf;0? z6ZWv8_NXm#{Sc&s6t1*m@o#oUz)Xc(H(r$c&EyyEt;EWG*K!{tZ5djJRwJnoT7T8cOU<`;_nGpjCs?;_q8?`<=X`{wxM}0ru+r zp%Uff`l2SWa<8yl`3`7f(wiZw%z*jI;eJG!z38}22`*FFxF7Bz?Ljmf%|z~bTV}8SwLgm;88G|dYCA3>f9S(B z^Et8FThm^@e^1)4NW!owX=Xa-EqNcBzUQ&HLcko`G|`UBe&%s25-ax{%MFv(9`#1u zk+ci#AKE3=GjmkHoZ@iv$bTQ2gOq!L+DqDARBki=_|qHTw14RDaO3@NMfwN0>W334 zbGnp6%6-6ckBrgZploz1TE*Pt10?OC(zL|_D~xLSXkl3(Tz7ZEG>Nz3g>T6ZL2JKwvN_lk!o=n=jD z>c0Ez;{)baxLNVs#`L4yP&5K5_f5;4OWG1i?^VXEAa>s;T439S>{mF6{vlL_e#hay zPX2e%W+~6#lJ$`<)DK^>XBpK;HIe)N`w8~EDYqEz1h`GDy=O~))B`DZpXDwgZ8h40 zHlaE5=_iqRV3@t${BDu=o%qNJ0W%w}_Ge|kVy_rF9x1n0X8e!(8jyAl>VPgm<(X4f zM#3$$e$f6b|HOb<4p-0HmE@m*@{w}eTJCO1+rl*flIsg;#|N&nlCO_X3YhmC|DI2o zhtaiPC%DTj_l|G)n`jn#1U*4J-eTtuSp!P0>lV}rm@RNa*1toxvF?cOMQU#qZzGyd z#^V=Be+#Wbu_}prHs7AN8OM9;r1^CN=4Xfd5&1tu+mLeWTW;`M`Z-hwg^A4z0j?wKXSQId)~%cd(Ei^rO9NIQjjdYrJ`hw6}P_etN(RbhsPH--AAB z3$gm)Cd)njd+G{mfx^h;Mv}Ps4FYDA!|hLgnWqjRR_?8qJD0RV^b&d=x!k-YZeDi4 zJd}dFfpW?n{13Q=N!;9q0rQN*RsWXVp5P8o@^5R7CHps)#4R`@U|w~&>fak3?%hfL zeSu@i{>@?xk@KhLE!HStK61EUQRWx4+u@G2+_QdQofUOJm!RFWHy!6Ab{?huZ*Jp& z`NrW6A-`N_jUiV5PPW{Iq`iP%N3SCHyoHnO&217edmQfP;z7fV_C0U3{WfJF2?9FCU*UaHQOn$k4F^gEa3oLgvX+lw@z#x%4{@_c!uOKe?Y+xr;6LnBuY~gtE~os07#c>slY8aBpz9YHxGODRN0oG>+b9%Zua>B^N7R!p8PVO z+(WGPzG1mdcQ7AB-O%M|KKBA%viAdWlDI|91E$d74kN#ux6#DP-C(&3NqYgkiB_Yg z*xTC9cVdozrT;Aq2h2)`yOI1~qOXy1w_0x5ojj+4PDM4*7UtAH+j(`d!@D8zZ&+_(8`hFs<;x6Wd zNcP8P9pjBdAFp9;)~-)>#xIy|axM;-qj=Cw$K(2xX^#3J<<=n|!MB05FVTbg;0OXd%7bv{}3XZ(jQLdtD!x&25RgodN!b;N?C>&@&|0doo5@?Kdp zl`_|`|7I?+a=G;_!FLDgFOvQooyHFKG`Q=^<_>Qj%(N#y9!KF`%Kii8%Di3n5j9+# zc7WT$a^rOSpL2;_uJ}NLa*N>ha_n7&J@2BA9eZ0@?oQJ7N%}9|`mf8aZ~HUl=Cux( zvr^cbNm=#7<^RIoe7Hj#d$~no2C}E;D2JP4xiv}a!tv9IdCb3rAKE3^TMYMkhkG_< zTB89-?S0B}my)&$y@h0b-StB(X?;D`hIJT+JB|HGvcGt*!^M0FzIQnG36db^(&c(* zEpgro+Xl>T)+e<-G(5@kLzQ0>_HMDAE z^}{yH9Z1@}XfnEDqusyV!>hMhw*KmVlt{aP*#|ci|8>SZOc~j)`vkFah?}Q{j zgf9!2Y==9JwKQqRr&3nAS(bYaX)RDY#IkP*Zgx^TE`r-41-GNat^5zT;UsR({{p6m z!|erc6dLDn>y`8T`!;Di(UH4Z4@VDS^b?}h#)Ou7)yeP2zo!zb zA9`7Cn6&?)UZ^XQ`#3s3i&%U0{T+oJ0_IVNdjt8!zqb)9_cqHNOWIWQ1bP&?+#H8n zfi@oQ7%(q8e$ak$A?1`i%yQ%2u>UzX>AWSkw+P%%;A*>A3{U*K(&0Wpog(cxbIU&uCqQNR_^I`Jl;)O)jixRL(-p>cgD>eTOZ2s!-Z?Q z1k9;$33>9LhLn-_2cAbPskyY_vx$%8Jx+Z{3vrK6?h~p%C6;pea4v+S{=AXA)mcLt zkJP^SU%{Ub?npBOIWFN2Vs{-`x!TVZUKuc5;cCA)iu_a1W60$m3wH@=Z=m%kzW=FW z!q)jo{Xt&WfVtV>wtv0kcLHSIFZK#o-WPa1k`U#Oy8VpWd!T*6z1QI;e~;Pa9?JKa zbFT`R#~rSGFXB^v=a(qi5BlAiA32u%y_I!Iekg{!!r}fVygyhULCM@xNo$OnA$gyj z>xb=0-2AIqhjX}x%MV%I0%rSvz|Dv2J3mo>FX1XKip5+bo2z8ja)xO?L1oR zZ+`cHspW82kpE5eDN5#+-Af;ijzP)X;$&_Q+8bPL7pGF@EYuPub6=$HmOZ(Q>Bh0- z^A=1iIX-0P(B2%b^tb&fHw>qx$E_H#yA=%#hI9KO-_ZG~>#BROS^Cta*%xenac7dzm zA<9013XzPAH{0=tf2L>dW1SFnM8%A~eeK+%fU!K*hx@m1q;5(WMxIIN0VMVb=l=K+ z_Cu3a_D^rz^%J}6cxvB1#%q}u6kSifa=2OKua24^<-TsY!$}*5=Aj2st=fs_x9i*f zMYtbJ)$1QHmxUAc{|)lrNqP~na@Sj~%#XH_z6b3>?)}>W+YjhH@*{4Qan23G!Db zR7uR2m3`}r-u;fOTLR`SxO(0)$$utlh13se;xD)j%cPqN&=u$s+MngbW;sU zs7XA~xn5B2@n?AbLSzE_Ti}M{x!D6n{?@1+QtnX8trfsVB%vp9;R(E-pl!)(hZE*< z9%wQ-Y%2KH;oeIAk!UPZZjt4l7ECucp*ztaB$aMHYmQ=TsJAzt5$=cNFP=udy*Sa% zmXm)idK)R%+b+U$@tNtS4w8^f{PEGoRI63e-uFAPKrJ8@@g+#yo3_-0HznE za*N=e<8U+K&7@2fhg;5aWxwvJ*4~w*$^KWpPENKr|IvWy3|Hqv$?sQi&dVXaUm>R; zU~Y2kO}<`KuJnJ0a=lmtcdTP?BYtOF^sx2ANKVxS;$65Mvgl7x9vX<`xw-6Tm;=6T zOx{Ruzs-Slh$jOk+%n;ZOOHx7IjA3!^CsLc;#lLGM%pa&B-&b|WPja`lPf+}ZuBYk z+rX8u1udb>t7r|9k#8K_qvW75#K+2;TN)s>NhGX4PZ=~%>Y%8*3iu_wbJ~rge513ib?>`n^O>`=f-!FDDjgsKI zl(a7BTGRu{c%|P_@1OMhbDrV4HU;-q$|?7pf54rP#EmQnm~APzQys3>gVOAspTrF> z446OR>iGHsJozs28lz?UD18!^01xf*~I?pDBS9;65OvTTa0!& zTsdz9pL{RG94+4QeD0n*^ATW&e7$ z6Iq9<#<|<()mdR5Sm$~+V7kH8e)l}ewM1{uswYnuiPRHoZHaM7@|QNx&4>FO+-&=I#OLxG#qZ+z z>FRguew!_R-QNYc)NS>h#ErUm<7ffgcanbhgEe?hg|$w&l|5L>@A>Yrf2a`7juf1s z)_=m`vO_|toG6@QcqfAf`JG~qe;)vXaH_`tRXV2#&be^3T`6a{+cFAtb6;o82!__;B2JPIN8L)da# zlhzyEj`}0Fo~^U<1bMFhd8rRC1Wc1l65NU8pMjo0%8gj=d!+q@${d$&ekFFfdDdR# z#$IGT4maeLHPt9n3)Ml&9c8%cw*c-V zaO+Z*TO6hhylc@N4tKKUzC+rt=!nX!5fICJZnKz!3})RRJDg_5kgo;T&HdpNzZ@{> z?f8cclW8pChkBHi9#{I?8N@QrE#!Le8IGr#C#fGIGC3whUJ01f;D+M=$NSvi&8@>e zY-08AS>7>UXVR`k68aIl^-Q@pvCmifhpZ^;CvdeL4L0ogor5NDgL68PYZLh%_}%2wdRDk9U|zViDq~T;>%7ZKR%wz1t|K_TI|je`;?RYi|th6u2^8OrZRBcx9^5mXX>! zM*aZf&<&*Bi)N##XiZJ_vanVq_J)s0GlR)=JB2XK6ucEM-=(m39p%*Cd-%scwO8AP zS;u}+7P!RTQP^7qZwCsVkg)evYj11Px}jT9eci-9%ZosW7Vwb#xqe0xc&SeJ9++*4&>Qht7 z$Ti%T4p-6$f4Oc-Wv_B$a2q&WT?hW!;mY3;?xCy$7iZU$F;Ne%6y;#)ieBMz$}AX+4|vR@;5|Hk#hgfa$A#j zB|5KmTKvA9`(0moIjOW+-S1ZfcO6`9SFP#%_Ol+H<8Zl+Cc(Fv8^676=eUP?T&nL! zANVfyS3KYA*gNwR#xj@#9PTp9E!xRm2hzu&A6Z+D@|)K&f0gyhEXUp;ekg$ZtK;AM zC^H?sg47SMSZ*dSz^jXzp+-pjtIr$8?DdWKw-4ON*8y{Khrjx_C*_p;i{+N)-;iT3 zLx73FJsWPae+M|+Bl!D-L-KEEEBD_Wd)2?UJ6w5=UAU$BH`}pS{96b&$Kj5KHx12m zxK%B84QZdDpU_t1wu}C@{>uDC>TmQLp2u*wf0Dl*eN_hU-BbV8vs`)Z{TR|E$oh4v zbv~`X;cWr)23+-THOdLMj>F|LP=fC~(psbTsBcm~8Mc1V{x=FY1~(KhYK*)u_U+S{ zuRGkkEVn;tw<8I8#I7H7eB3~NmimzUE&JZ!%5|xPvE-kQ3Xu9iuHy;5m8AWSB&63) zH(Oxnb6^L!9X)RJd!BoD?Co)5nyEmUib%O~uZZBQLfRc1m+*Jj=b`O9_uet#2j$jw zxDWmdZXVn-;Oe~QG3x_Lf(FEVsq}_@}qM^w3?*v=VbQ}u*5HMH3mHH5&d_MW* zJ1KLC)ejZCYa!ncq?wb``Mww93F<_j)|bvs=HC(O!$cTbABuir{tZ_{P4d@Ajga7E zyr{%Kn1_{bPFh>k9_d^`&R34DV|wnQlnECH%mZ+7j((oEe`j~f56?I#LfSi`)_d7pS$6etBbFY+IxfLp1}V8rlhw< z7a;fj1o_sV+Al_a37806?PofYzZV*Wlsn6Ex01FWRXhbhpkR%}zOP`qw>~q1^Co@> z?`B_;oIf*lJoTt`&@(-hMD_Ez&r&<`;{BWGX#x8YF`U$-!jrR zpr6oIB;SS1x9zFE{oPvc%|6y4Iw#K4pX4uBkMo3-+s$&%Agu+8AbCH9yMEu!a`n2t z5N>t2I)3$_%s})2Qts`RyMeTA=ogg3jc;jB1=ijIu1CcO+F#}F=RF;8HB_okTSVuh z1OD_5rjH}G1@3kq6&!K;QJr^n)G*7|R zc78RyUMSb$W?FkcBrX2EZ{pvOdtFdq+l{pI_;o?KpxNU1_sG-PbMiF(6tUV%Rtdhw zq_ss7t{|56-Eb|>56S)w9}zS^JATL^|BWaQDffgp$@rcm?FCeX)*|^H%N!)k{&Wz1KU*^0D2*}9VU18L2e|7kstKNq8X$SJ>@Jmq+2XC~5mZO>2o ziKI0^O_B5yIrQ&hQz7TPpnTBOhokkfDQvlyuJ=1-Ja1-ypBH)mm2242#>Yaq;Uumw zefzxUL4_;pUW7xszK+7}l!7bwo%MX!x^+n0BDj4WuDs8(IsUp3DVL`57ki72EZq;q zaPNYv{o2>~KZ<3%j`9H4)&k!dd?)iO*6Sr~Cr&lrKd{c65j0aA?ocjTK4YU!V@{0P ztK1g+h6s{yC2^@-Kkd`uuKkwhlAkf=7huU;c`x@hmb>?a)Ym6!Z*CxHUUazPUy=Av zxL3hfd*g=M|H$!F_Ws6k8He-XZgIG`(M}HL-&Nd)>Syg8_&0tC2ZLrGTJKY?MPC&`rnC0sD5`){x;mSFyOSuLpnQKp(X~MDO zc2S&UuQ@Vkat{O74EOpnwKoKJ65NnA<~(e1*NLUyk$NWQttH3WBMDuIQ`NKTojv~+ zzkku=R4Cr)6aYO}Q>5{rfoO7ocQ&)xXbkO#JI!S7h1qrvA-`Ti4-= ze_uwcQ8HKkyPjjo+>qs}f3uDXn#&xn)CaNm!xY@o?9EQL7w#a3tM+bA!7a_+aI(D> zgXS@~G_!d42HvZ66VKxCrjEaJ3yDu67Zr8Z<=?_i(j~ zBDgyp?%|r3gsTNjCX;>j!{M5jMB&zh>(+;VGcSqSdZzP|@CiY4sl(NA{!r&7MV71c zk|^9;9In*ge={$MCEI&q&^+vL)!sv$mlP-43->j+hkITUsvb0(9q!?pm*m6U>u~?= zyd>MsAGH6?4F%1KS10`NZ|5aBN%NANlY-_VxNiMD9Q|1gZV!ihxcal48bNa>Ty4ky zc3xuc^se7@oHsRtroiF;+j&Wr+<9pySL zS~qCQvtX|6swruk2ABiuZ^D(eHo~E;XThzWf;$s?)Lwl)$isX8{kwI&7~Hc`aA!H( z)*kAR>@9}-KZh&xk9#_$nRyOZ@29!;Ca*Jyk0t25ChHW|_2BCI+f(pfK`R~Zg=+7a z-2X%psu8>QD?DoA`*j7U22CE^P&~IWwaI@bx(KO%yLrcay-B+p%|Q>Na1YjNkTCOY zKOp16d@5qjX+bj^uC}Yi{6 z^+2wFL$cQ zpv#bQ`&#ZG(ng?ZXaaJt)3THLt^9L?<~qmT1>~3Wwv<@8H(Bm$q^(C1M)QoKd%r~e zAoj}r64Q+NGu*IbCTyVG4`>Hc?m)|3d^YPaNWy!>S970H?++_?;Y#W*`3u59bC2VP z&E(&O(#}b6S6c36q+N-Iqg>?LTkO=|ef-_r3xj4YTy4i$taVSP%+pBiea~{=CT$b? z5q*u^`WtrYnfM{=qM+I6aAm!FA7#p)o8W$Kxw76}i}a?bA#%C%ZNGjy#t1hC_artE z=y-7v`Fo@Lk=pyCnd@+6Jd-iyQ{JJW++_kf)bR}N(iGh19D85*7u*=!wQ#jw=`)uvINX;~ za3#Nc{yi@w#)sTXgXUwnA+N0Y7~Z5W)65qRcfIBAlC<-AHUwp&Gdb6~zbRtttF+rC z^j~?G(ci+=enz=9DX0GZ&~h6~8WL_x;Nbpr5tu~Ty2Jw;MO@4Ww#f8K| zka^xQmfMN69!SCs#6O?(S3l@?Ua~p`P16+i-t2Hs{}=Wa!)*aKJ8lKlSslc;w=~d)xUqt@%)>0dC+u#8;a*P<}&KZE0lQ?2|@gzRDEgpjqH> z$B|#|`Ai~Kdna4&4ASN!2?v{xwX^pD^Wa9|YCV(vuwwgShdaY^Wk2j{%RPCheLpVu zrRDc)kmnxb<75u=M~AzPa`K*+4;=1X%dK(|`+ksw_ZW;)^^;qfSH}HtP0)M;Hx$op zOgqX+J?liQepqF><4Burxd(f0_b1EE?iDn9;A+3~6y<)uv8-9_a6hu#eWX>onCE%W z3COi~yB*ixuu@l#Sh^= z+`omZ{qM_^`4Ihxl)KMze`vuRq$TrgbOowJ-K~Y>yU45hd-d!VeE1**i~9#n7r5$& z@5sLg9oZ_uWqZ5?-vgw*gf^mg(U~wWMp75Dc)y&)oAE<`?8vz}Xl{h7{X?bJ_zzu< z)ZR-hw?Z4PGf+=-CF%t8T5E5mhdg`Vp6mG`Z&1+O4Oi-H7xV%7tGDG{CrG(=%dT%O zX-koWO~f%6n|QYTZ^H)C{<#6x%7u+Ja zC&LYSZZXHeI}7zgdfuKChfvS1X~(;+&~!8zy~H@5!}V}q_Ct2z0YdS!wu{Kfpy>}+ z>)A)-FMlcf+b&CR`2<*;+oS`1en;N1h3-P3kEs9kyn6fdo7}r1?C2!#o??DJGiXLQ z_P)@WYaCSm@&tDtb%NkqPTDH;E_w?||Df;FY5keU*KsK8v7ng&SL&|>eP{5AS0uRW zEZ57y9yE@1KyJTN-}Vo>Uo3kT^~~`@AM#5-ImF@e=_?7o2S}TVBs@;+zK>dbAi=P& z;THT5nH@CmI`$Tle>qx*^t|yY4hcSA7xv1b)6k>+%9OZ%k3HwLxqB%j{aNC+55bjSh=rQ?jNL8?2eDo31~Zg?Or73tDUV6$}M;yXd1!QdX~KJN%}MX zeB9nc+4mHAF=*Pt)p6)t?6?B;LTazB8OA^ z$lnePK*}9&?fr$ctQ?+4KugD#-2YWz{VQClXSuJkF6(g5p-cH~mDe z^5F#e!M=w#&DyS-duJQhEM z-w&Ew;Z{zvw*%#rdxPcPLYmq;o`1Xc7Fc`beg!yF@IlZ_<{w&r<$lmq$~^3Fxpb0H zqMrTT{h%n^CmgQS-vXHZww5_yZ>>1CBv;ljT|a1j&~{M-_Z7GyudJC3a}iqV*jvYP z*OB%alCYK7UH4Y5u0O;+44RJ|dtYTR`-w6;k^1*S%k^KEZmJ>)$={DruG*XTQPAvg zxHTwu8afjx_fpH1@6xm+{Zf?7EwtxNxuFd~b0ll`S|2)5rWd*iDR-ddj?L%WJfuI2 zPGud)^>6S9ub%0?)9jDgx94!@P)7XwoWs@m!^fn3izMtMPBs6PdKRz08`<{+SL@jy zmrT=Gcp?ICR~ zC$Zx7j73QLb-li*Z~Y+lO8*ev%(^{XskaiUlD{EpjFdamJLG$wv?Fiu{7{E@R4e*Z zq;)~!nu|Sd>?`JHa6?{M(}?^RqRvRUpIPoZq-{n&qHmDY-;0s(+F5($y-Q4=sNg~K zrQ?S({g~UKN=UhKf0E#9OWJiv!Y#zo{|;n7vv4)Yx~JR^3U6b*+Ob#G!MjqQhC18^ zmaFUQ!X3qNx8G5&_LGrsnU^rP()z31Ne)-%QM$e^+=mas)&4B!JLc7JLz0;w^YWEMA@0pi4_AaGd1^lqW;p#f=TcmBW+;55Z z)=oU%s9ddQMQ|s=wQ~ehO#Wbh&K=V8W?vWLdy2H@(EI2OG>1C21m!b9F0$=HzF!;V z_l187nztQ$zajrVbjFPd?r7_WcS!pb{fHJ!WNn9Q*6pbLTG)0yO6|?x6{P(n&f8fx zdG(vi(EM)(3t6NYUP)S?+MRlK=RdX`@KFI^X$(w5@0l+JW5t!u@T( zBizB9kJ$d8IpKyxeK_V8`bpFbDHl^D_=fR*xm?o6pgWK|4#n*CraqsYT_)3905=pb zYRok9%ewN*#LAuQ9rKmHmGvl88&yNDy%QXJrC$$~%`~In+H-EsB>(?VN2J{CmisVi zg=h_0fyBSV=x^nG6>cuNMqeW3wzBoN$w0<<;SSL<26Khr!9SNomc$ba;0 zJX?*FJKu6&AnkSZHTo14GoXg(8=9hZqrB@_`5x?TJ%k{8S|Uj;zDvpXv2G%|?4%ZO5^SeCdLkjYf6ZOq`P1{3TQ$>M;&5kBM(&$FMXcOQEO#Ag8<2$4j!#QDKNGHu z^O0(qrl-UGox0wX^AvNq_MK?HZKUn8+`Yt6#?9n?K*FU-n?ks^INY*Bd7hl~0I}N3 zE5;=FCbA}7lk`)OjGL|>vXcCecLMDgZYW-q=P@ZGez@4--fp>lNV^$H7)tDNwQbzN z@0I=`=fq6&h{GL0{^{sZr1lnA?vJGXjskbEzJXjnOtAGq``>UV)9iP+XOjPw?P=zJ z#L9ivaz~Oj5j}{`J)LhnGf&lhKp_TKv01ocZ{bOq%oP%TSVEZ!QDa^uR_;5N`#ovD zq9cZ})U<|;{j2sC)W|gFIox9?lZ_f7<$htg*ON97-GSz;ENO2}lD(!@rum=4 zeS|V{-X14buKg5_PtM!Bq+$MT1c;sz#qv-}XCMe4kg&+oBh} z^OjvN)7<0u;WhFXqo0v-8G0r7`ric;Nf=2i`!&j+T(V!+`cQmYrkU>eA^E#rvL1!I zBplLry>d^_G%Mk1e-`Cg-0|3wkJMh_9t$_dv4=^29L+-R_m{%_j7)R4!&xDo8bLm}ka zo5kE#{GjJ8c4nse5^l&E)0X@)kM2yY{$1!D^YtTb5E_T>LT-J?wdYN)TV=jm+&I(h zhO6yBxzi}8T$;E9-(u2~`*IRDFR2}vvzUjl!Bn|#kbfH2UYi`Q?qdtx&A5Oh3~kQ+ zLFS~|kBbi^X#2`eqWR9ci>wpn@FBd;aJ4=( zBL7Sv=MyXU1j}vnBj0r)Jr`Yv-1?Aj{fqnJ?IP4H(+seF5TouQzt}sOSh>0`@H%N5 z(6=at+;zis)?QsV%sG$w0$gp!yU8!_obZiGaJj7`!Pk(q3sD;+-yL!7-EO(s-m=cm zG>aYXYK&Zix;R|jU(|;*4U)&@>bX;H4&3!{wLcp`*;{MyUS)?%w)tKNzB3DN|zMo+^F|XPRwr^*ZG{ z@&`t<-iCxA{p2zkDBwm&y9!-{PFllv5syyHn+ohaM6OfB-slCHrrm%9cPwR2|A6%d zV&&?6{;j0#Kzoq%>l3h5+gpLHXTlwez2S>8&24Zaag*3eSg-n)Sh>y1#ZC16Mq1ixL+hOD!Xl@ClYX=CO6C(cC+gi)^2>GX zTw>*pj$`@$3u#d#;fbrf?_H0o!Meihi9Sy2UFfPzbA`iwn{u)*{!52j5GNVmPSW;S z?hzA8j{90a4sfq#odB-Z=L+Pnh1w&v_i4*rPTG5D8~Pmm+<=P?I^1n%l`-2Yd;6HQ z-W7MtG;hMq^2)~RkMuELwfCOC5dTKKVR*$SX&O7+*_4y{&jN>A*>Ybe?On_L_L?nP3?{J$~4cw4aIZw zo-a@igs!B=Y%^8wTfh0%2C#1lx)S$*4Y>i7}6Hq-p(*n2to<-GMKR_Ld||9<9sp1xO-6=RRw>r#;A;74=2=s5xt$Hc0G^a|k4HfDRTWf6+Cp zs~ql1@>d$m*$`5D#ePEIwTD<=qtH06$D)GEj44EN-(c}e9#{Pky)ny7cevxpzZPBk zu+QzD?s5M!o@>h!xOWiUhuYJBYi4GcU<=k|mwNZ7DmQmTmdWG+JRKi;Jm!sO{fO1x zQeoxj4U7 zGnxAz^AGJe6g3JKS32m-c%!v2ueRDv(FouW$S*U#O@1}n?EDVOn|HHA;IV^w2tUaQ?7}eE{?EA(Y=YU(C!hpTX5oK>l0N z5JYoH&ht}JwQFQA}{yts`5ene`4p-aR?dPW@xrK1mUM~Mg2*`bH3+%;# zzq?PY1a8AyeC|kiG4znbU17NsNqYuK;E}m<{j2qs`+Sx;-{C$_xw+^Sr1t(~xywmg zha_w!u3wMqPK-N!%UPEcypUyXb+|1kBk%3JWRLfLO_}FKgfjEItOd$z4{C2A+&Ekf zTd^Z!8uzv#85V?@?R{LKF=;JP8|0pQmeLjqW@VYz;Y7$^f;y7tB6Jp3qk+!S49yk4a(ssz_^b>jK zsf6f@Ste_s-;dI!b58?00;ydmSi8EB)(iDP^1iY%=Wn*(@9Re|F`mHHb?rdP4CKtX zkXX5ASng=j9z_!3#4YI5sq2^xQ|)sraJxG8P9y(I=yjyrJhitqejxpGbkgp!_nn$T zZyZ#63trAL*TI#3FX21Nm>Ha5Am#S3+!IObg!-UMQJXeApM|7fcCzaN>G!hlDV)PN z<@iCK_Y^<>tScWj=AZ{m6e0dIl-?Q_IbGf%O79 z1GPlM>6_z_*t;f$Yu?E++Z=8m^52H0Am#pOxm9Oz4+`px+M=m2XIpM*ZdrSS?`E0W zxBB%qnEb=gVx-*dmfNw2=hD!%s5k0Hd$Xvi zA^i<5Und!UC;!1Ovv)x1hmhsIPTB@kaW>b7iThJ%IFfoR_}%kE!t*#rWM!7w=y1Ow{|*$G>vQc~ zA2^G&uILJMF_Ljo_mTbVxTD!queZJ4-(J~2v&q)iu!o4G?9on>#`+crYH~D?jO>kGbsROS$ z>u{O(ioed{+P~L8rSK&3%ew1nV!_G$y@9_}=X_=cX)Rdv>AFku=dR=a5C+nq{10I^ zc^0C#kbYlzwulh;oU|=S!W+fR3HVXhukxKG=z1{!n=F$HSLeswlnX7Oza!;xlp!H- zFKLs}JoEyR_P-PfFaH(q{7&mV{#};246gVw0%#TaH=(Mp`dnUxlH`sgZ3bF~66hVq zmeEZ12_DdlR7y9pZ&yD?He{JXxEj=tyC|oAJPqz&`7uWM=N&)prCjJW`a9C^dzQ89 zB+`1I8&F>)zdKQCzi+V6``+ru;*DA61N(i&w!!2dhQ=c0o^QGDlC}!1M{Cg~v~itZ zXWDkI-#5BB%WQV+J(4}vd#lsUcFHN2%jXgTX|MB~8j>LEjQ!PnG2F^Sy!z=+`C62V zqVth*`&#Y-(pIBAs8qx_GM;gy{dfGH1@+U&sh{|kEYk$82CbilZ}{~y0PbJaPwHa{J)`{ zNV!uj_t1s(S0v$N;@z~-BI?Jp)caH3cX1Ha68(*F9xhET8BQnvRcIzs?#q_jIl*(` zs4tTHNM_TS-$6p`#6!*!>$(tbb^ekJaUF;}+X z?B20i*R3VLbA17>j3W|ulfUwtoVO$8R`Wgz+(_CGbRT+nRfef_q(43_wf$<>RPQ=r z(e5mBsl$y^rU=bO%C%dQz^|n3MG}H~<2IJK7xOzlkmpr}DgCM?^S;b?rGacS9!Of#=oXz_sl)V`ZL~8FHd`1Yo^$zFVC}R=N zTN3ZU*5JFIy@q=@Wq&sOYvw=-7KO4+(B5b&h1K5flvR6&lDN+L2&|}N}2`6E9tZKG77p|UPNc|10%D6{4 z<%)j^2efaG!@bGz!{_k6Lq8z(!=)a#LWRXVpNVQCx8Cv!SqD|iHVdgIwQtU`#vDn$ zhDhzZ%Hsr@lXkWII5k1`L+&^u?LdN_D-^@M&-vbMDcciW>~Np7+<~NxL=qk#-Y@r% zY%|@dr}5;UhMq-gFHKuQUoyOXrL-e>KG zCL`Amui5!Qxh3_pO|ipWLjI4@3Z&fjmMiasIOGHF+d>iK+S@e6-e|*Yv%%q>O8y?` zBBb2&EcXu5o=0z@`N-wQ9REtal{U&Y6$}0Lu$26_bFlw4v2riB+!{-Hj~D8R&OSCay_Yu;E0^0>Bm_p2_A*+8-asz5D1{qql5N_S!!4$qa@p?xEjN+E zjU1D0`jo@n;BfEy2i)QmZqc#XW@I_s3LpCRj{XPSk`!)kG}}DqaO=Q30iEPVt2qL_$vQ$HmAxHN@ZdR(@7%i&HT|3hb|o9Btuzf&#uRnnFr2`8ZP#z`G_3Xjh= zs~zr2%B`=)Jq!-_dCT2FT8%@!|7pwczCUi>3E5_o!>ziE_d%crNbP;qa$A$u2Mt0u zBG(UQsOJyu$BE|ICS$1IFNTp{=CiTH%GGnOsie(A5>oG96CX=ZKNOspZE88(Hz+6m zeb3>3V(nc`+A^+nN%%Yenv=3k)ZwZhHaOfh|G*Ey6h9Q4oNYQd+?j(jlJEbl@R9F_ z?=3e-T8Q~e!e3qQP(PHw?d5Q5QC8~hXovf=<+dU1T68Z!g+*=NJio;Fwsw%KQdvjB`(NnX{Sy~Qvlf$k354Z&>+``;!^G!M2^kTo>8vFxpVG6hOjBK;39BxgA zd+a~p#!|TXXR+>QLeTY@w%=nN?#cgv8&BaDx5+jQ%i*?hxVis;Ta?1hJv-Z+=5Tf0 zAK*rP?HA`*uGfRTKjnJe3i`CH)4h?*`&0XI9^CG5b-jKOb`C&=j=kM2cPeSK&;qn( z63@ff>taIeWY;?_)4X+RydCR4xNKXJ;XTTHi+)CGukJ6~f6R3v`%c=W z=vH*Se2#;vA-O)Dpgr78KJ|k+FPnS4G2RBbK7Q%|u6a^Us!R6S#Ypz82XuWL?z4^` zM#Gzqo)dd4H|qVffwiPFv-RJVWKKABypV>)#T%+u>?I?n0Rh&?QK@A?zju-ipu;NH0KrXbUd4#B%jqE4N#= zIc#{Ei6o1%$EM6EG!7~EQ17$AtE4SPpQDe^CA2X;Z!Wc5ef}(ZLAE&+ZZOR&{2lo# zu4KK0GDwnon!F%{dAmWM7xOSt|h{Y}1mzHPgmn!)Lw&AsS8?;7&TeFrxaYdf$_ zwc>50-JQY`o&;e@D1i4iyu9T98FN4R$DzlO+Q+3H2^F6wZ9ZCv`tqICzR|YsuY(!w zk!^m2r{lm9^6o~VRlXlFS3=;n)m$?~PooK_TQ|pZo$pj=E!?|dsClZlX|yJ;((rCD}{SDT$v9f$ZvkE#Fi4I z_O|ps3hX6KLgh8IIW&obL#;D&m*;1>&sWYF3i@*%1UFX-5UP{^cytm{?rGjffwV9A z47EY0B6%O!8hgDwXu0w}uv6s=4$d}b-RpA;$bS#I4=MK^%kB6Ta{-dDmAHsTRQQ%> zXR^}xL%6NsMuxMWy3gnSLH_KuTxUSa<ON#`xUkd;7QXxJzw67JIoJ(8R`b zUi`4nZAtzPs54S-L(3gb+S5qF3&i3Ff^8|d6*Ze!+3epHN9!8yZt-`)$ze-*kKDVJedLLg4s$4J8G#Gg9-!nIf0 zL&@Z9^T`B%d^qV^?oqC2lFv>mmuwOOr}W7%*KmDJf}9t+_Z2C3AcdqK7f!(saCLlG zhaJD5-;i?Om4O}Ztnawige0sa9^QugGLSMQJ{CW;7e73kZJIph`}a%o??ii$a&>%I zyq>LCV$fVJd0wA_*T8?~hyZTDEx;uKM>2@^44GkaBf=c=|`K zry>dKiDi74-KMO6b$p1t&3V8ipZnjA52Y#NLvRu80j`b@KViooD1C!(uZ|B<(sI!S zs3Vf`A%c?icCEL5?qtV@$P)bUxNq+e^52J^KBmHLrB6TVi_NlD{KimJ`^p@HlyL@CUetn$S?hP0kLv* zeAp;yn>=m~v9yQOdeiZt_#@8!9eZn#zY%JJl&j-IKhj1Z31f(5e9(U3S?k#=<3r&m z*`~*nzP*o-{~7cwQm*#nn>MpmdWm7ujYwTwOoM$-kG5u#L}@ zE7=HvquBFJv)m%${f#@taG!O!uaMvUeQ|faBhNV`Ecd;?hRPJEXMxfD1 zuh%FycxG8{8ARN-u#x}!}8YahRn(N z4emjcegDchT;pfz9HnykC>gGqo1T1kfXj{Ac}Tf2xJSc{*fNdaHAiQa!yQQ4Nc0$b z5V_Ay=Gk#lpUWuyI@@$|?0uR13G@+4wRbmZL%XJ#tWw5?x61y$LZRiVy@_?K*WofW zB}3}(cxrzW|JH~1Pk-Mw_Z!w{j=k#N6JVs;YyW9eo_}*w{F?{&Rfl^zypE`QIozvA zD?kI0+%M$TTV4vc6z*Dwy9guiqTKx`)!sO15@rxD;jbJIe0+oOOPi&wgI~kh?zHJOWq8^&#~CzZxHEyqj)f|8RULf;$gx zqyGnP;g5`G{}0^ghHP`={{y!K?l^~gJoQoDdPl~| z)A_7}tv9*9Zh!5eXk)fn09X6PzqNnM+r)L11K_r0Zz!P?@&49lrErg#`tREt+njBN z!qt$6|1L#WBC+g0{9!+>a3^U`pczQ+cUr+9DE+K}u{ZuxwwdT~W|K$Wr}Z|ma+ne& zRQQ@S^9#?NAvw=#%EDdl#je7QgoQt6o0V{Me639$d9J$wvGh~1=cy#d1g;`&G%!dMnWr3Xg>8J_ z!_rJuV&(R--?uGkJyCD;ORe^)E;)NLX4z#&5s`>?6J`_$z|Gjhyv&-i|rLVkI_)pNwkmEXA~1U@2d9r_t9 zW1TG5;Y#S}()T;1*qd7+$8>=kO%|oy@1URa_sYb|4O2&iz>ym=%n_tFLygfUn0kG$ zpXKIp4kqgslbK`gg)99@LJRWCdkW7WRxY<(NC;d^T0b-d4L~~_?r6*H@T$ixh5M1i z9Yy}}XgX5vFP6KGv@~v7J{~nhO{uectO{8#3;MRky{i185wB%ZS4ds-3wB>FfZ5xu1I?n1ipzSwWCC9XIxVtIW z?sxVeNVz9i?)Rh}zRUZcJmP+|-}pUz*ZVx*y6rbsHOHI>HNL z>u3`j$RpiI%DrOCZ6&ScA>O?LQeHv<l` z3!~ii$4uM42U1A-O)=ajD45INB}6IH3FRT>>iO|z(lYiiub?dC{@zTnZ7+I$TzqJb z+2U}IB7X~XDpKw}-ai?bPTEQ&VI6TE>ju3Zrv4Z1MSPcNog8!6bl=`hOm}6GLZJ#X4`|Dr|wiMy0oMYW8Ja zor%5D&hn08+;_NVQsz9=4JlX8kMAUH9Fp)jv3s3O+k>7Tn6aQ6vNZ$W@4E@5f_Ym?&(2+>Fwrd93kai)Ga0T)H+C%)<9P*KwdF&lMkpAu@NUU6vlW8e95X zyKX0T#{u=DjsvA||8VTOoBU(Y!$|CsXNN2|{Rz^}tL{H5B6bx-b4=h_|9i>%q~t8; zWvBczl9BIK-ut8yl&?;CT#*d(D7O$TL-M`k-*4xA6euMvEj`mzMODzIBUu|E`A+$E zo?U0#UE;VL^9Wp~_~ba-a$QC#}2>VThATD`~4oHV8PirW+z;ozvu8gi0o&^QBJwK zPdbq2K}z8Up7ZC~De#2*oWp(7+B=7|tS`hn{01xgBzZqXA)T`Td({VzAL@-41?_T7 zZMafz5)zb^b>e#t_wD2-CQz9de1Ar|gmuLA>oGpo_nE?$`&DG$5pADiPPX<&_~IMM zpH?x`)JE!u_bqokX&<7W&^lDWxM?~uHZtMIW~Un-D=}$wM7pSncgivE!_}be;gEpe z9=7t~{@O!nIqf00bB@{IaOJrud5>>hhr7vg|LQp|GpwvHUZb6rz&&`TZ}0P4(v$t; z7`T!u{Y}o~2Dn&LQnWB;kgE-h1cedPm%; zZ~a~EQn(SgTHgt$alEh{vjdIm$aFDE+Ix;yXVE)j`TUh{BzmIINX~lEA4Zz z!{xHEguvaTJ&Yts``jP51a2$G50fY-WAapoJK1uJNLy&R<7tHMd&bpX^+Q2kjyW4{ zM2sLTw%m^#?(g15fz_nFY#)mLa>G)UAx6OGuribGP?N93+?k>wckp7f_ zdktLq?h<~3cgQTBE9Hzx+k@u@6WGINl`47u%^}|3d=&4RWA1Rc@_tW|yuVYp7Xwvp z@*lQ;$miv8<*Y{fX#(z3aCID%^`4x+Ji#VIxV+NJyV}i&m7DLO0#}iCFOqOWTG{JK+8*|CyXBY<9X~uIKyC@^q+z+nobJ;dZ2-GL-L?odtHo1OCt+(K% zIi^2c?H8v~t|RJ-)Lv=BguoQiUP14pw~))twf>d8sN41(VZx*CGRz9?7xpD*ZNn_BMSQFm=oY?y^W>JH1r%&?hlrG zL{6rOqPFNXG>(n^R3x@mIoRvR>fhXJb4*`{n@|3G(0xd`SsBT%9%vNIG-spUs0X_8 zBJQ6-!Y#1vSKCAU`W!P9uJ+@>>=8T1(X^kvqy-+J8O9B+J3 z|C)h0W{Kn9UF5G(HPh5X%I&TF_yO)WBRwDWKyH7FTYL3=ld(ZL<~z6=ZYKW_Gzy7Z zIgS$(B&5GWTBKTLau1x(JS-tHB*$!b$~WQj=ICUk<+?Dx8gv|ssL!tccH_maAhp#Ca=dkox2GB@=_+1t=v z4tI^M-^WOM5iLRskULK24fE9k?Kf1ZdZyV++`JCYgwW@O`{|cnd(2?}BJC{qt{n3+Tpj;q zd^(PD&5?4aSgsd&e?C>O{Cd;-;`89Hbh+@_p{@@19n0;^y8SxRZ$a|jJ9)oL9=?cS zZ?SFX%8d@oF+V$g7)F`#=w+n#F09UaSnUVx%lD(_FV`0evt7$|Lz)G8jX>dfP+1w1NM8t=5~4BgG?g%v88}URje_Gt;z2za!=D zvfNt_V+=tX(Kjf5I%n6Y6h&)!?>`;WEX{D6kBN?D9|t#TxwC6AZ`Wr1a=6bu!}hmE zN75hAb@emN<*3q_G}v4l_#w?)`K8BwwmWNL@)sB7nBa>(_t1u!rWcxz)ZXi4Kwy5I z*C^9`g=#crzl_>U$G=E?aODl&zJAwK@4Y06c{!#fT=heL@=rm}AmwtKRg&B9DAq$r z!dT*d)Y-Sxo%&Yd;j36bQ!w&+j_Czg$LYt&|1_G8l)J-nSCFS+St=&)D@#WLb_$IQD)+{vBu+ zQttPbTkDuib1ah3l34m%AwC}Yq3>huZ^a+wm@nW)lDQej$bUJy9w|4I??njoKQ_}W zMmy0KG!ABdsW)GDvVLf7&rNba&oSHJYQ0r##(IlAVh*u#^DMUpX$_Ht?}7#(m0M(vG*v-bVTPO<%X>v1{}}X2s-qH zOcO%U*SIHQJZIIk-@ad!UB8rW$T5$;?Av=Q`6r?$k#fJX+_udbH&6-s80BBivD6Qo zZ!Tv4Z`-fz$E05bf5|Zm;Ocywb7H1>=OErmORQXdF8WN;IwJ`^iRJur^iePw*TokS zbo?#(HOH)htM%5K{5?*}G;@)18`<$;d`srNQy7!b1QchC*(C41YnX0ocJRhYt+%4e zL3844pLxq~eC zWYW$=9nb*gcy}FDn9|OQvV-Pkhnq*4Ug!#>-1{syM%rW~VLGwP^}Oq^ONw)XoH6=- zkaPYyl$no|yU=n!A#Ef21r0d5?D!i?u{RM6nit_}d)P^tuBTCNNV#8H?mPT2ODGpV zpfgZD>+ee6dh3TJd%St*e8%5f_&%}2gXSm4-u2|)f%YQhvh0=+sBs4Ca?}KkxtczK zF%cv_h^N?FP$y`z=JMp>%4JiUwb$rXpVNc zA5x|SeS?%c+j5(nnQ2ZzXQ8ZV^ap337kRZTx43SQcWCi%)xIp&c;GZ1bhnVZK2$iEwf&hojoGY0yRHV}|y(4-H=g7D-lluz>vEIpRGD+P{=% zn3fPYnY6P}7xdcDjC~yWo<+mFaZ%@mg5!c_io?BtGIG5rpIG{paQi2*CeV+xJCOuw zPimiVH7K_f?#mAMZpuB3CL!hCX}K?vwg@dn(qCO}BE{aKv z=MU0CXJ?w4NUq1Z>yP3TZvF{Dv)bXF$Fo4a;q*bNT&ee~k%R;vy4+Mh6vN#Nw|z1< z?K#+zU!5ls{xn z5cRg@1rc6^mw{QalY{0|huf0;?YMEUBe8P1rB6bjFKIU;2?M$4-JY|=)bFj= z3rkxB%|&o^oN7UthnedO9q!pl(t%yiGA)DVF1W&zFdW`PXrjZlbr{%8+8&hCp8kg_ z)b)P{p`YEK?qR-@{YX)(p!o%EBuO=9&M)bvph1Q?T*{Fq{=L%sEYO6slaK_tzFFRQ zt=BhmPYs&t?A5g1)W4@W+(DLmApVV?7BtP_ruz3{hkL)}dXe`h!@++0K@r?b;HrOn z!RwEPINS#;cRgvlQLPS~lcI0Wz!yl`?r6uq74Ti*89{S5+=y3ppMQ^{ocedZp~KW4gk-0*!FE(=B&5X_Y!MHlV{%!E(ld ztGs%P*?D&NT5tR+>JT)$9PUNrAA%+#^#j{`34s$kWt#I*f7A!feuZpY%Ij~szK@<8G&jO+oXpLz+L^fpHAc$i z777W0OGx_<8i)$e)QG>&nQ7bakIlW`t2jQnEn1Y+fOw%np0Gt4Z~B`mO9 zeP6G%F$uzz{b}BLK{E?(u4E>>N&aHA5h?cu%Wc^u({w_YqYF{-F}z0w39r=F+s&`q zeed}}^RZ*^kS!Tz2xaa;${k_3vq*abtw8UgZngdQ3omv2P>F&C-GZj->uIK^XRcXG z{-03hxjuKI<+dX29CQh~0Bxl$s2?_2uC#}K;;QaJa|~SF?{$1Tt<3M3%lPYI`TZ64 zIRlV{JBat!9%69Y!zC2TA2P0q()M`egK*h)B~P6il7xC`TlKwAg1XBGV>w)s#uc&?~ zfLnQ?zuu~b9Y><0k@}Y-DhYveNxKMLj(U8SUgrH=RVsV+ChJ-8Z_$lG)1(~k^$xdP z61#tH(-dy$O+nKUuGZT)ctz-Chug$*KP7D&N`@YP?cY6QDC{3J`Ea$Ji7(@QcqV{x zTK&-0BOQ=`V{k_}eyEHc)lprfT=hdo(z>HQs26hIe-^Xjj`;T|rjC*ULGwX5+<}x+ z?giG~{c#g1+~`2gtKkv~<&Om4Rx|~!gu1|U3C_4MXFu}9jQ*z>@JnI=*WD@t1AHujkgN59u&l#w9M zew{9Vfx}Nj{b?SqX%9_`;dFC2pHg1*g~LhC*>LzNzdvy*Sx)r!pt;83w2h}5(ft?s z_WWq=NtDAGX(#RCJAy*F|y|^ce&n3w7aQjTv*Z&7@ z?n6N{!12QbzUz6|auG_kcL=}X{R*G=<}G__$!sR$iLf}Z! zPC}<4eQ%MTN0r)fRqf4tnCogBTB^Nyl#y#QmlBH~gv%|55(2|Xdlbcy*jxKhf8Qbx z+ep2sy^%+PW;EPLvMA3LQDzzX5-HcNAp>{K6S9opgS>h?keGf&6^{?z3

MN95X@ zZ@K&SIdQIkIo#gZGpP>Gl{(yCQ@CPxALMeyRtd^2fSd7-?}rEBO+b%3+@rmrW}m&6 zUBzAmxqgT{evo-EHYI3|cen${e2`v}xB~VM^F{4g>c^huZ*NQ`F4i4k?HGG@tK}TQogrzHqqDl7BX;(bu=P ziWrC=J|}Gh+Ksj%_dd?tUf%s3YHws_&^*4#Zx1(I%{^)8ex!cT@oY9}3(-et33A7? zeCuBwAI$SXvlp(eYd4U8Cn|3|`(%Dr@@vI2$af^L>D_q0=DVAZX605yB zo}EHkXVe{~j%UH)WyiC^7dYpF8%es5=Zz>c96f}TtK->cq-~P)Yxz#Z?szsSWjsq1 z1x=@S{qd|WWtyO7NVz(m-Ep1QFdiWugWP_c*UlR!bv!GXA2bu;>Uj1B`In%jNV&Ca ze=EJg<5nslMs7PRvRtV*>2HaJLGzYlZjy=$k=B18jtAeHlTP zPbhUfiwsX0&sKAO1vesngSI4u_q`izjKndiIXI1VkyFM%ZENIUEz~{;{ zfIm=XBT{ZV@1NY)e&v7VIY4(^t?f+PZ}9V=xdX1Q`?kT`i!yHWxjIgsekb=1APL=x z-QO=(?!IyINA4ehtNy)={C!bBr1o~Sp7?S;-Er$2;7K*do6-cNyiB8&Bbiy-y&QtM)24 zzB6b-OMO4ggf|y0aJYLdx9>eXLxv<267SCsF%vSm4tFH^$D&7&+WV^K{6NmV^j{>Q z5%K=Gd4Ul3koxVc8Tl(z|C19d4HtuKX)WrMuq zhs$9SDx5)DSJVs1drEp9mU23P;D48_6>m zO+fm+Z?*QlPFgYg4DCKVtxUbd>^!dXZ9&bD+39e1P$uwTrU@hE4!7K6NjnX-N2%uv zORYbJTbHk!duYhi{m8E;d0$a~eoy9nhx?%APUKxhT@R6e&6u<2ls|53J;mX6gsXn& z1@C$^+2M}2+-;=Q9Ghv5M@>+IIx0p|?>$au3AbNbQh` zJNDiPuMj=yaEmN=HEE@2FG}ObzsH}&yA!#HBK8(u@1FyRz460C<^#Aof7N-2Ywsv9 z&bN29S;7kJDgw(&MTJC((zDL_pDRSrABI^eoUyB=t zOb598{O}mhC8I7#x#w8!6Qn(l7NWUG#zA@ii1gdwJa63A``PmwhfIICjXfWlZzVsv zb)wJhZn*;==Y2R&@GKm<6jkHQditm|eoHvRj9cdULGDAyAb)W*WS)Yn03z) zJ>`$Jl3ds_WNx8+#LHu9@_9#;kM#SV>wOmZjkNQpGX9`G$hE8eVP*XoX&o}9&i9>2 zetF-*RAS}cYq{@`wj4=VL+o;;&q~-ge&>cv^-uiz`G)+zpq)s$vQ{GmPJWvEU{OzW z9&+~|CC>O#7x%@_2$^AUwO`#petG}nEyT*5Yq@dKG*sPM{`+}#d?`9JWTrUwzD(J- z(Yp@!6U#mH8TLx(baWDO*DXy`+EcV`$h-wtgT9xqGvy>z>b?J-6#Ks#>*o@yy`wGna?%RWKy+Owy-Yib4)?}?-CyK) zV4VUtB43Vh2W9R@4FzZR`W%AIPt z6{d5xhU%ijP=s+*=c!J%zV$w_!j7y{;cC6f`l1D#rj%2zU2X;rWWFweTXlu+-xl!h zVBl!yaQRq5;C9kRqsP!g$i03Myulyuq(22ahs3&1^S((qX|3XKiGW#t0T7K`4SqVq`S5xwwh)zdpUjy$S zN&f!ig{0>rdC!bHE{?YSR>#HCD?_H@Du0~0nljVTW~AH}mV5sTtS!-7XfAU1S&?u3 z`BnB=`B#U`{cttNx+-hd0q(P~H38>YINI*3!#Dz+j?_NgXI(+s0CWdR-Dj1i?6YFm zgv?%t`#5E0qIpQU!|nH8PTD%O8KqwLSZe*L`>cX%L#EMczn)|qY|dWlf8A#l!#xkK z`e6sQ1d3RnA+>kBwRa3@GtfKeHRSHI@>BZ1xjtlWE63jdb)S_F_Zi3D?buQAMfUGV z?JctQUP@YjGz_Kgv*IcJwB&}6`53PD(2*_>UVbC5fa&wi=w zKC7rd`)%Y-R(Sc<*VGM`WoXenv2|hR+~@C?z0jjLS_bB9lvxuIN*I&%J>zzJ7m^4{vGu? z*ObwlNb61aS!cb$nLWB5^+xVKYe&ldpm=1+OQN(PwD0}($}Hg z8#q^?Qgptp!k)sNZ)2lErkTUtLYZCYriDJ2%OVm2Hz&An9Zf(FBA45f0mtPQ+#fPM z;cBS;Ci_Wr3`(76E+OqYG#I(_OjB&l9TPG`;AlM!BhP5`Fw*a>^Gp$Gi_lV(`rf=w z_Iv9*lN$?}=ix@Ifh#Ff?Jcg;BjvX6J_~Fkt;yTm^N(_oJI~~QS9YE$d^lv9f9`*8 zX;+U@=0MK#`eADcob%x5_!@^X11&*npUyLzNZX5Y-eC=f+?`nBlMkXZ&->q+LB7-jy~c_uhMWOg|AJ_PeA zv_Ieu6Ka|POl)Lxxu&iat?74=2EkUP%==a)Ut3(nv?dW}EN+)w_;QN3k8 zSLc~OKBE17%snk=9dhTHtt-pUGr_qblLuG(g^mXYI?v2Z8NUkP4uh-yo%0FrheVr^ z`eBkc3K;QZ0yHs?JwjVW^iQm)Q3ok_a{^+r2(m7TvfIourVD2BTiu7>L= zGY}0%vh9*{otwN5(}$DxB+|1TNe;foIPj%k{uw@>i&i6@r#PyWP~ofQ=cLD=wi-=zw-T>&F6JcYovClU%w=++Gnf{ko$ec_sB4X?}f~n9;Z;a zavzE4GV%(&J@v~{DS6ZH9i%15-_^-Kf;7=cC;zcF|0$m{{(cZLS2+34AWhT;DSx8P zf7lnSPnL$vO-_DUBZ%a=ED2)I+42|eGY{r?dvEcnhI~GkPzq;KIh<m_-85Y8;p zippW-VpDWk$Xw)b7LYf|ayYZ(AB6KgX|ZxRmAF0-gLA9HDJ8F5tKaT$mRQaqYq*A2 z4u_gFC2+<&9J!xDREIM9y}q=Z!KB5@;V`Ul%`jwMD~EFrWh6*_)D$4!D^gDT31W+c z{Nj-L%He2zedBPtSWdJYPEE@xg;TLaGMYm9V{b!GjlT51*C5O3@D=CI<*=f#@|K58 zV~3;9gIquvDK6iuwRAPU*Tl7)DV4)%45xHu$eaTwD#j3|k@rP32PvmBEa7}!!n(d3 z&iH1`51)n1KsY)d?jdjBYo2XH$_b0h#U9e~%Hj0o-zohfWG2IjkX1s0yzirB4(Arj zd1@W)kZX_H*ENpty^6jHnI(=r?~r#nDnasfcwL&|HhBqwgTCRs5YhPlUYx zJ^C@%jak$QSKnI~UmG$%!_{ysW%}?xClZUhg=?QLOwT3lJfw4mlnIuEOzp4zbxuz{ z?}hpxu}ge?l4Qg#dB4|#q>o3b@82#M?tO1@uB;nM;f{f;>)zr5?-`FL9d3^#)&y33 zoMC#g?~owB_q@ORk78ekr~{w-BE0!%p~Jn(Lj_iowhsM>xhucRG~c10QR;en>Xx$W>G*oa=XGiO+E3B?O!F#Qh19=` zEqCK>6-@Q-IS)d;cqf;8eqChugRKj(yVVc;KF21~w;}OwwIluKQcB+N_d(*{q8~$MEnIC6 zGIq-K%is?_H%LChzUy(P^O@vv{SdeI>UFb%4cs^Ot?!3dx2Gkq>$G*aM_FzGX?LP~ z(cQf>OmjM%zL(7$Tz0=5-xxB7edpV&&nP|Ra8F6G_iH|rYtpX0d6pX|UFt1&6Z=89 z+Rn_6tk2LWr2akMa%XLzt)LIlBIK^q<@cs!9MJ0|r9X$vEXUr#8#z-%vygJfSZ=dT zoMWP1s0XUw2AiW^{~ej*argQ6myr41;l|1TEP4qkx1Q7*_4f2;dCw-cqSukzkE56S zTxk!H-$LdIE{v%kPXCGai@G4?K5e-fKXXQc>Y(b#^>59+-abhDAl$qioa@5XeOqrB zH)G%plvD2Smg_~{pM8kj>l@-*2|Au7;2!e5&y_WV=pMKdy7TwscVRv$EN3p+n#@I> zogw!-%zYs06_Vs{sp_eysSfdP6xs?05Q5d;edD0~Pb=i!CnVorg8DNG z_Y$}|zQ*CrLJJ%&!?T3I_mcJt*B(&TJeEM*I8*IR&!_qJ+(zFATe>%7=EK!?Ri82^ zpp%f=dy?h$Agw==FqC*HMz=WL<9gQm&;8}4h0Tv}Bgx#x+)MrmD2|kSmiJlU9nwBS zThO;i#z7soi-&uBnHQ(0vp!1?o6wKGA9j#G@GIj6Qf_CN8 z2f0*Wf)&E14czwD-ah2N1-*%s+ud^S+{*Q5Gz&e2+8@jQ!Cp5ns_l(~@;-}$sNra( zuvr8*V!3~iKe&zSFG#ryE%yY{PDfLCp8D56GE9BytoAA1_b;)2INi=O`3Hr~R=8}x zlA$x@u0U5K*z;4oNZ0DK*s*k!+Z#(!d<;3309In)xaIYY3%<;ax?)>0c zyLaz)@(Q;Y?i{!}?p#M%`8O8UPuu4Q@vm?zEKc8NZ*~6e&S%PPtL--&HhbXeJg&bH zb&F$f#Bzy};dVZga|m}HkJx$}Kn}6D5blkee0%?`dNWnStcm{pdW*r`1^3^sw_vrf z>AczZ!@pH;akvk|{a@=Xno@5GxSu-q{#*5ydq~(E`qRH(Z^dx0fctOPTVD0BnGN^f z?r$Yn%5>-b{_KY5lYBhvD>5c`xJxwjcLrn7P2~zJZ5%f5KCwOPERQ_G29% zWL+xj{21Ksa5cP4{$=P(B^bVE%&QW)6G@( z;hVqW?pu!^7B&ZO@$J3np|m77=5TMY+|9@GyUb^L|MOQ|T|ef*Z3WSR?mUkIRovVG3 z+F|oS3U@GN9z@FBo5EdG4mW1kx5_O%JZv_>)p}Y(nQzcf4!5FL<$;5LXKjF5qvpu9 zcN5=T?A7*JP$z6E|Kj)4bI5-lx&*1cS(ZDPw25dInu^?cu?me{xUvqFeO}=aVRH;z zwRbVkP}QYBEufrot6Od{X`Den|b@OBoNmX6F6w<-3- z1bIHhZ4Y9z1ocCH1Lj+}I)6>1+~Pz9Gu7eVpCp;U64Ji2+)`rq{MMs-{X)K9UL*R2 z!`&skJ?#CE`eBOYHX*G&l5jrp{ zxP?cD%>{5H$=u9$+G-@Bdesb5shWSjpnlNykZ2M%{TzEg+QR$2 z@x@mT_ZZ9ljkJUQ@VICGL3^lY%xv~N?(fXXxFh~8Xc{(!4mV7>TBs{hd+mH1SVh{` zXeas!jmzWyeGXx!w@ou~JFY9Y@L1X}TpiEa?BzayRq3V^v2t&=_FhTa4M@T*#O`^h z`a#;6*qhrdY<_~P{X+H+cT(mqq}(FQO#=C244?(Xh#NxU0zjDt`EmSnZ{oC%M~5t7NDc z0;gZ){+_0A<+@!d++J|i4>if(0JT8Mtz_qm0@7|n!_oRg*?m+j#Sf9=!)7Gh#$p8F zQOZn1yO47GSne6=6-@yehla`D`xsM*q|Wkod*|+7)G}rSRa<&e*vy5i?cr;F7p+xB zMRN{P?t_;5ENQQx1lstYj57Z1L?aNsjytgyVY3WwPtRQQ5oK1RIu(2_TT%&urKJ6g z0-1at;vseY`!Hkn{-cB6;@TSa7PktUop9S*?j@AzgC--jmq+uH+?Ew9nk&%lXaG9( zImRs{b_P3oeyG87viHJGv<{n`ZN7isC;#W@`hd^<&2nEnsG|7+)i{`24p5Zm2Xgt* zxfbv94))>R`MCI3+F7(i*qjDe$KNgFPpgbANbTjP6q5G-Kw4Uril!c_j!KJYrzg{> z*e7ga9V~WhKaO=`+;O-=$sa=xA?02uf5E@ck@g9a@ELJ={aCnFup`)oamV3)Oa7f` z4^r+YmfI++qUnbu+(tZ=ac4FXJ2fabdS2K(1XugT-Q*vG9z@D*lq8wJ6Qs>RMM!=R z!kypZDgB}d?n1Z`udI2MGGC$Zk#f(p+=Fu}n!`{J)DE?$-h%D$?^(2eo~ag_&!Hb* zN>c2?u-WTy2b2E+G!`j$w&l(w?R6wU&I!x=yYM-Ev}137%H4_Xa=3aPUGJg{GsbfN?spOLFAH=2>G$J@ zDc6b@xjpW1n^=3(zsN9CE%$2ngXQ^GpED|f`-WrhbCi?!3cuuV**;4MEFo<<`UZWD zTy8%*{yxe;6Q=zZ^a`7maCMygjr=u26-_Oq{>@Kfc`k~yhtVtO85HDxy$F)FP-52) za(-SZWK7`|VY9>Gt|R{+C_U_R2V3qWMn@xwT{ zk!$F0a0|S$rZV|!p(BuTKk+^b%pvVfv<7{Ga#{0rvu7;{+uwd#&pZF&#;y;Wi{Q4f z+)d=)h5kUw-DbJjhg39&A_?_~YcrSWa}jE9@MUA}W-w-vne&;<^+GVPoYSV-D(B%zeJ8-D1E)K-bbzfVZL-4Zqv9D9E!|3TF&nkq=` z9ca1rNIM2eIEnbL{2=ST;(=i^-Lbbd`Tv8iMaq4^awpYbZbM(7k5B~~U`-@;_N?u# zAH?3Lus3l>*sO)C{h|vuk}VmLVJ;(9E|*0l1g4T!gx*C9QH!(qt%8$0Zn13-+Am6m zhRqg-yNdjqP$^PwnqBXlawzLTB%wF)m8_u#*>#lIEJ3-k5n&VHz)9`Bj{Fa!2}rq( zEO*Xf70rj}ceDvL#Rs|854D?neh}`_e7}PG!lou%o!Ag=52}CtMwOO30tfe6g8Wx%wW^ zR&^?xvr$iUE*e?a@5iI@I_)fPT-aRg*n6|&N28H)Td_VAd$*Fd7u7m~S8*e; zSKqG^x8tPvp(X`mkA%$-$KE#N7w-AQ%587C&yhA4y^r2R3HHWI>v`>Wrep8(*jqfF zb1sMbHTlKArNqj;&~mSgum?iJ(QRlc+(C^!?ox;Q65N7`VH1bjH~CdLk0t-NXd6=Q z)0UfkBS+FiZ!^j$=P>?{9kRTVv*lz0<>{&QAY#t4qkg2ce&!p#?p2tSa2C)$J5US4@4A<(LEMU#g{p`mCv zw#J_Dc%_cLZ%Vzr!gU0=IzD_%{;$v$q}&t8C*1Ros%S1l1?WF$G;153he~Wa)A1qk zYS;{O{4jz1Gtl!$xz||kaYs{M=rYt5^*oDrG`9BYE%K)45Aj2glfmFZ&OPDk_%M|G zkD{lLa{DFCHGvbF(yq}h=o*yEdM6LX7BR2i=y6xS#61KQ%v~HdYv9rhlVJ+^=b`0D zxdA(_zj+MfB>D-Jpm*}P7J{UmMLT(Z{u=u~B7gDwtk2-)TW-B$E1Kg`E2LbG9wh`0 zX;#rRMjg>Lt89K>U#w z^!9HhtHS1FxH?WwB>z-23#q-hNkZTQ($=7#(D$e>?O_lS-X>>!koh?O)39j|H$Pd_ zn4RRWc6>#10#a^8JFY)S+T&;*dI81oukMfI)(_H;pT>^p7h!WDTy1A1Gyq|taUC%b*Xee4K z9~zT@ldIXPw6WAK557!NVr z45z&RpJoyHKSZA(t?z;fQ!jgSuhb^uzd>hl-N{{7$Q=tHDj znhgowF5(WLj0>=N8TU1X@Ebl7Zhy?zdr^Mx^Ag?C(r}n7AUe6v=%LlRNk}MUPB1ndJi} zw+zk~4yVrFa0=jLR|uF}%HW(5P1tkx-*8Ic{+C?XNg<4?DqN!dxzdX^9OBe?N4jUD|yqXXvXeTj9vBgqp%XCmd=KG^F*Tu*cp z>a>J=s-wnqN5UO%>sj9n|GX&844B4n)!y2Sf$t&T<4C#ft-bQzi3P;JgXFyvD{zpm z>&~@YwYR8pfNQ3S^I{WuwxQpVa=TdWan0zrQ7aTh<5{2{{;huSafotOEgcHn?s?}!EiW+q(gLo=0pucIa{ z5?sA6zIrRJQK189H`+9gZ&SBPH#50kA$zcY-)_IozJFsU2Fx1A4=-Fy|BOy(li+?~ z?ak!H(2Y?))Dz`>&UbWAOf@B3C~VB6P;3@INPA{Z4w#hviTTJn(r-gMkow^S|Md6j zw9PQ5BMD8|PL0sLunrz2yZb@TlC0dpl>J>Pnh?`Aa2;WDI<;Ef}0I+E~RjJAyH2h(Yj_I&KOOTyOkt?1N% zxdX1&->aTG3 zu|MzhfT?;QQGXv0-UocU&f$I}AnbjOxFweRA>01Z#B)7bZokkDT}ww(a7Msf3RnH` z3F&`8$F@({yDW}1ULkRF&{8xX-3E6&l6EJbHdokUZ%+!OxM9G=;AZ=Ixz~=le+QT% z;-vk3!Qbb}y7_U$OE`(`N#7FXAmu*o@AB>-ZVD}m(hPG8nu^Awxlz_8 zkZ@b%`R6gifcSYFYZ5U19d6~zSg%E=A+>j;jZTdfa6;EhD19In(~ z;f^J45|XfjHq7;d_*g>IsPEt03j$^=T&=&4l25+p^Q^;_dM4ai#Lcr@8BdN1B*tex z^}vD6G%Q2qhob?r60X)WxzBjV<7wtChb!eK+<$Uke?HuBYU-cur1VFlDYwNA_jb!= z6A$-LQO@HpcU%%3?bo*<#EZQJa3{iz@VA7+Rv#`5m`!DH|KIfi?kQ>igZj`cV1~jy z?De5}z`P0fu-6B;>FNLf`jC@cAK+dBHxjQ6hp|4i2$(5laR1-+0q$nFhrNAh88D}n z`ybSYivnga+{0cU;4Xw)-j>^c+dkwb*9W-iEC3$%`p_z1+Lyuof7b`NPryCw^`UjZ zY=L{&+lPw-roQ(-s1I<5!u|i&hrHza0Cyo=^~1aLKT9+GcQfd?B!xOC{oBXv`wB_; ztA11GS=ns@<_pIU1w8Xs)_7gGTVP(#ehKoO%(9+`(|K$RZaHpHbp0UbhV+Zc_6ko! za=)qlyv{F*;hqFHVl(VMgdb8nCj4;V^h3{w@-7LO%i(JMtxVpS=f77x+1~2x8)g0Q zSK}R>zeU;x%wLF7+1Nfor_5oA(hy?d-(r-j1NV!i~Zuw3b zCV(O+uO0Wqp2pnQuDjK@<7VaNb_keZ4)-+joR2O=%C$?C-kZdIfHtAEXtQ(Oq1euM zge&X0MVAIlC9Wwle2j;qF3*TRk5QGaa+xMch<``neBxUoc@Dvyr?$2B>O3{~GTK+T zTJPJFrw6(gDfdjjV7&63S$9B~*-kNuQ4%L&}x) z7_qn76|9q>uBZcQ*an+(eD2H=f8M9u+%5q#hIKr(SL)HjHIT|eNxxl z@om0suasK=_XW7x-_9b>eDs#XW!f#lJEaR{irS-AsFZTcA!9xZ_$%-8?bUWAziYts zr7V?uH|d|R#r=0|)!wT9Y2j68UNN8e_t6sM-q&AX?G>)fC!^g0W-eT<4;x9}U~QV& z!dAJbS#I`~8Kyb90<}jgPUIeZPAIWA-<}s*A7b5UU*SgLhsxNW^kdL{NV!cd_q=Yb z5u#b>DU^2=JS6p@pl#y1q|Da~dNUs9K_vCVG2QXq&NOopTjjoQxjl&+ipHXmXmX8& ze{*d;Q~#Fs37AoEsfzLNBI#pj8B*>imfNofYf)$$`UVwXtC4q(%t$klp8mSq51sK1 z_U7gV%nY~@A8Y1a&6?$!RKAa#;AZ$;i`Oofd$G|-Gz6tz$~piN|27$)@WVp5(ZK<; z*5NjWFM6MR5_V!VO-uaxO>Pz6m*P(iy!s^`a^U={49;qYbF}5;mBHZ^7x9n_=lDvA z<5YXrI-H>8vkI*PQ2`;P$b+;yW^*i|L1Ynu!f78A#v+3 zq~r32g!Jo)dlEg1v|kW^6jB#*htS@_(e_{^X^PQWBtIpbZgI@P>+pkkz{+h8)W6yu zWDaK@;&5wT%Qbj(15&Q`JIjbGM&F>%kgPZAx^}63?m)S@BLe1gxDnqb6YiN|PDa-t z<<7SDeoEYTXb<`cx$hC`OKP!M)<@<1%Dp{cY8>rXbjjYT7we;3V{Xq@V#N=ud~Rx2 z;(8;!KTM)yH(4&JS5ndCuYf;o}bC%DC`Q&+TTz zeFJX9&uiwudlM~jxce;E>z!e$pp(&wNZRlIHT|}s(Dn3S`8BG zV2!`>+rvRr$`9cb-4igu%!D82k$x$97b!Paeh_<4y^ekyNyuUAj%)P%RX-GuWnS)Z zJCk1SFTaMZa=DF0f|p0!ok+sp-VZ;Hb`CBfU;Y_QKGD5M4*vk~`m%%T?^$$G(Mtc` zhS%eqEiDL`u0BUnn-fS|7d1rjZiRN$cc)43Ox$zmWu)h<9B+&=%A3S^9gdCzE~H*G zCgV$y^!RZaI48w_!FUt)^8P#aOX$P4ta?A6PD9rFis3$826quQi2sCZ``mxT&7K@E zOW?-)0x(P9oy`%ibGSW(Pq|dNKEqT;r=mJY?A^ikJ5rr6d`d7ePdlj12{?{>IN zNdNH5spbl{YOkzi61*|QO+=3&d0&jnEpfQwhunu5cUMl>JDWU<(HBU$W8zrjHNJtl z3F?7P8JKQrNm(#gDB{9Y34N86nLhte`bfakgB$VlnsMYA_cHTTw#wx{C3qv(r zO9(YfG4*kt_AkPe@JVyO+%g|yUCy!h9rCS32ORECagy<_>&G|?O+fb`8P9Kqlf$?r z!q`H%>fgL40%nlIT~2!W9@Y13)!r1p!g{s)bG;Brh_aRbP{zl@S0v5_^+W6_`dP;h zmqpkIWe(Sf;1f|(wKwk>`frDOEP1kTOE)L8Rjyn|B6v~a zEdIYnV6oVSKsv-~CWLor=vhIz2CXkWZd@@8@u*TJ9v`oRAK{ufLr)3Rm6_Q2JcJ z41^nrXXTqQq`w7?M9O8i1h3LSo1bAkKXA0w4 zjcH2yUZ^ipZm#97AZ`Qt8Er+osJneBH)$UtmRolr&*))q&Wi!_9^Af`yNC3P^JrHF zCAc#z_uL_jBxAv|h&)IZXTi7bMUHlK@ z?IA90xL?mIvK9a8eL+%o5~MSUx9h6|CKOEgH$?hubU9LPSId2oxET5rtwz(C(@f`n ztqrti1^fK=LHv6feu#b?FlWQne&~o>c$Nk=Ldw;4z1#@u89EVFN4sk99x&QTu`|!M zJ7;q^@$+~+{iMUafb=8KD5TtneII#W5m)_I=E0~A;#ROf{V?9vU*(p38Zfirs(+=v zH73t_NV%9L!Sgr%zcQ|K=OALUgmbKaOX0478?l+rhj%e*>u_uP`@E&RsQC)wB^1=; zAIDywdf>oy5#JBFp9RcTxSS^Op!W7~xTpBL4zl+dxMkG`Z6EUBmJcQ7&pqMYhz2=a zibjHW8*yWhgaWo@Jr^YRrL)v(vNr|H$#AtksJ)Lk+_U4PgZ=wg_G-CB;kJQW&(CX~ zgf|P#b+}SS1TVyi_d4+sVrxa(}G;E&M!S#ykG~S@PY+`HGZ#Nu0#vTS^JpKIA+a;tw!dQUO8i^+ zRlr2TiSzAi(r-oEk^13o%U!%9)$Af(!UfD7%BsKGo<+W9JPtP!Pt916eC2P?Fh?Tg zVu}RsSmNp+38%3=)cm0o?)8qnGM_w)Jm({RED^m%!_c zu5`FhTCRL2^H$>TLf_TIiB9_vwf@!lPUdFXGsoU>5Xj4-so`K5_Sc5maP3^m z%p%Wxv;Zl$p5=BO$vPt%js_yP{f#X3$G~#F$$TjPC)RP{YWw><>F1-xNV#WQ?l;8k zK)<3vXZho59Z%-idZz7fQ7M&>1#9)gV2^KzkuUA81h=2%>hs{0*(c9~v!wQ?e{(HY z+sW*oS%-AERmmg$P+fxemavnbj)A3zol>+m%(l3aAkZ*h_Ao@r`)2Q zj7#8Zxpjiq6ZLbrqb+v=aZ}M8^bE?OL+y@ayk26*eNt{zL1TUmn8^-z5$RWzVjo z`yrVJ#P$WuZ*W`3U1-b%(l@+^>u5;1?_2KPG2GKPHp9G)7NE6^b*rvOH8TgKnihO$ zM7TQs&8-wPv#KSyFN|Z%crW+VBjrXJcM`mo&t9#|6z- zaP@pkyPx$1;-xKDuI~op$@_&WuwO!eZCU-0UXL-=gC^scR1=M-X5K{l#;6HWZc~4s zx1P8k(E+p@&EtIQ%i_KGHP4=J)t{y9!jRk=LDS9Q1|FciQ3Ir0vPkeI61NyhSjo1C zlYbO*df{pi?kKp?xTu0k+}?t!G*vgxj9uD?SVV!i~mL8}kt9=b#sna)bUp?`z_A zBMGUK{Q8hwZc@)Kg`0Cu&{R1t;fD&OuZF54<;Jh0QGc5dcPWxkDTVKn{Dr+zA9BtM zaxN#vsa?r;9l8lAx4i!jP46k~NgZ+?yLD1amaQ_=WNVydxakHaAGuz?L;u!x) zeUNp}6_i_P61N2Ic86PO%6cB_JRtYN zpa~tH=r`wM?>lHMO6Gn;+z#|BO1_UN-}XZ~Z!yh+<{Y@X4pZS>-Xn-XeNZwti+web zgcv*Badoo2d2qWr+@(dSCVdL?LX^y{Lfj7acVK-`?h|vl1=e2mgJ~W#Bg^1c|A*Yd zByIuR>5hME!E1<`m%+V^xU10hs26g_4@F7-&1(@f?>Y7!F8`Lm-ByOZ!?7pXzlX9n zmSk_NWzbYU;qcdo?2Ce?A>2s3_OMKjoiQ(Qznomp((~eB_Dw?)*+yKYpRR38dk1g$tm2QO)W4B7LGz7c@2#XCiylD9-08% zO6K-ho@&ChyXUYk#!k20$+TQ;ckY2lR2fv9{ApRe}E-mLVfwMaiWdEVKg>b)ixbkf8 zU)kH?%T&|yd%quA&ie_=;_l|BW!#t3A!yQT_-VzOucz@YTaNW=Bt35f+wSmZ`e5Ql zp$W*n&L5#Jl)^a;j{0*dX=H9Wo2?wTaJUpG!P~+5>2l&Fyu2wj{v9^GKhgE4ta}&b z1kFIWOp)SY75RQgS58fEnJP%|E`Nmc73HIwQQooK2od* znim~=cay&2qg?YrYVUcLJBYaP=t=Z2%0H3o2dw+b{d0wOo-O4s_U2p>G_`9c{IHPp z2@$Njo+65A8bWo?~tzpise9Y08Yjdcr}`3^UmJfo(kn@ia$_Z`cf zLfi}_LB5wJ<<^99+rV}6jZ@E({Sdh}Xg+|e=i680lQP`laIdi3nWxY~JjpwXYNwhy zBMvUNTx)Mfe)e_jErJ^*T{71rzuMdDKj7viaicwhW+~i=92}tv;s-}dB@IZ8K&`;_H2<;H*tpQM^W?9=z+=Cf!Z zsc1d7eZY9qt^<9rhG$Et-i6(K717t|k6FyL@$j-SC2E zjkyH>=CQ!B9B#yNePI+Xf7{QM)xRy_bwXD=+yU0! z3B=7tZ=*L*|9Xk{GMK@M@uRG#m)sOI?>qL+VC`ljc{U-n_aV#uj<}e=wiEwupW9Bl z=dskY?7=}3I4R*@<^Jq&W&M}n{S$6J-0pDYJf4dFfLHz*=B-HWoi0?kZ@-;tjwN2g zWsUsreJx{MVCFl1`#XwwZmaXhAvZJbhpV9``A$a-k;KaQVWz(`wGDCIkd8s((L;jf zJMu^TG^Q8(N1)M2?Rv}K=lwujwdst(Q3I6T%$Ry-_%>zR@~Lavft=w%a}w+H>yUDX zTW*CJz7OlNZ9R*=(di%iSMlqU`Z0EA(9DFZA&WJbbIErhlGqN|OSO}b+J?9@*{|1w zB{G|a>rKb5Zlsa>dit;xyCi=*Nl5vwyd~WXC0@dvY+bvwKI(TlGe-u^r*O4B9ZUL2 zXc|(kUayw-YcD5$HInZ!xcB{zx5umFml)h%;5Lq1!F3?=9QhpYx<<;~=pV564sl!1 z5i_|L7Hv3=_Sjx0$mQLwV)H)6RC4|ljS8BpPf3(dYtmnVdLZRmH+v(9y9X7Z^BVB} z{Th^aFU}7J428D+*XxLp(af9SYAELZ&WFiYh-ADi{_W@QPQS8%?+|w3otJF2jKs#d z_#hw7gfcjPZ^hqVYkq%Q@~3NdGbV3Tb&xvi?h( zlVQ9K<;*c`-SRBJ*1QQpQ?b6EP|P@vmvWv&y0egSrdUpQ_6@Ql%09Qeir^H(SpX*@2SvD>G*6+7 z=M(kvVxQxcdx7;GbTX=mB2n5#BxS$JzL)uY+n+`s4w~H#cM|ERp;<`l`{S1T0dZT< zUi2&KMSUvm<-eaQX4|hA4YxdJQ8bnMb!y_eVcLs~r_l(c++RsY@bZheClh^y-a~uV zGoN`l#jIewQ}T-6-nLsr{iV$-c`|4og&T=yHKxu>w56yGQtnNbyJJMUnMnK$^b~U6 zGgpF7#UJ7au{Ty2G%>h3-d;%hmFRnXgtPf|B%TfSh(`umHJe; z;%Py%1up*?54VtB##6H9p;bqwS z^K9SVVz|xV>NsY^8R-XI&vEU&n|<#0RNeWP?*;!mD&mKnXM?5>+z5Y5SdJZkpUn2$Sti1j)6eFCn3-(%leB;l{VgRASng>ZMl zZ5>Z-%#YY{;CXKEq0%P%VGsM#Uh(U11-7Hgu($FC|9!c6GlM3~4cA)Fjwbz5vMxu- z++)~R+j8YPTzTfq$>)VW_n$c{XwHL6GZYWW_phiQ*crcnrM~5!$@Wn5q++<;9qy23 zDP^7)!dkpX^4oE#S2E7ZeTn%uTrH<a3o>60n$ZroBYC=`K}h}Zk>x&4+#EC??X8{?x3`nE zx8ThLSNu={_ce#RiahJkHl*B*mK%PZ`ijm#@_tD1LodtCeCuFt;VWDha=5?Hr?w(r z7o^;+mOG5Nd(jjm+<7&b#~^8AqPE}Ba?72^b?(y>{++fv-OM205~SRpE%#UA%Fk!6 zi>mi$>_a10h=f~^WN+T9tW(0(cDx~Z?l~&OG-s>a6v~0%T~FL7^eCEyZpH_rP)z2z z)LG&8<$M?a#$F4W!4CI%(!Yj2K+5G+baCzxZ%}7Z6VwQmUPW{dTOStq{#9=N{GfT- z;a*Dmu4oWaZY_uVBJnXK+)S7e%PmOaM&AsY)ed((=}XYhNV$zI_oO!&FQD$|awPU< zTW;Q^2isfn7XA7e3I7ft{ZKRxDOc9<2%d14ppTJo#onmp7AJA@7Y5DQaP@rqmh?MN zxwjJBT+3}rTn_4qHcc-VuMb7GK4dODxc(L|3YxwScLaIHp{YpirE13QEhcUoN?E|Q z8Mbm>Or}0ayOUYzU~YCSXr{o8#IrKaBTpmL1Sxlbzt8JM+(`5=DnRpT8&=r%A=h%H z-8l<83YG*-UAfoRa$hF>60{5{SKG5q#BD58q8L?-&^h&;vPj3_Wa?m z-z!(nH3_;NVwQ0~n{fTIF;A1P2rWiwAdfPt#Y|7ByMl~U;bIlzAn8F z{w`&q&R1rHalFpmN@r^CB7f_v;7UZCwdhH>%HIQIGopf1kIWxE9N} zo{74mSRed`q zgf`QE1>c~Z;Q^HF{i$X=iKN|@`c?cD>rjqeH;{f9x(i8p$#I>AKoBk`=10XD^G`E;E^x%k)H!@jf?z90VfJ~+6U=Op|qYY7#|TMdbSKgC|H z5B8_c1?)Rl_H)fj=cVIq-ze+KQXlfZ<-QlVeQl;rq#ueNkbIUa&)yQe9mJjV9%BU5 z3~ef*+>88j%f8HS$98;Y%(>VS*%~xc;cB_{CjCul0#dGE#SgC$w*;*~9p2%)3lyl@ zTWH4}avetO&HjPwJ8(5@B~K|jfTaFzPn27FmG^0fP<z`i{ZTQaLy)8 zvjb`7LbgH|4!4L%NWF%*$I%`bYdgf_OwLb1Q$l|0bF-NJB`EC!|M(>T$$nC=8*z7_ zN$5T_mbsKZcN*bl53y;)O#eCh$j+dt&pjl19=%5TkI*J0eiJTJED7ENyx*(*O4d}- zFa~jMInPZh=loxSrj5frjyz3Kf27d$Wz?8WNbD@I<-D~a^G6(2 z{9Djm50~BXa3$$SpeabXS6c3u#HD`7wGea^nv5S-*7Ui#)(^uL`F<$a!+rN~b1e6C z(hozUk#cYF_j$Eea~&Jqfrg@9OSpf6>y~0~v8_M%eeSRCI?~G!`!Kfd~MW%Pc-nrxb`u^xL)@dmdd7IYJaLdb|ge4TBk%iP=WwG!h1+jen)#Y|30v8^=aKkA!k^bO zQ$pq!xS8?P++#L_^-1!ie3Y>FVSk@@sOy>At-U&akHM{QUZOtKhWGYqtm8P`_pEC9=ZKt#UN{L86ndRj+Xxp z((FU^*ZB4crdUPEA~Ff!wt!&HP4wJ+5c{{0ALZaiv2aR{`AK4!1FkDcH4!=R1@;w+ybF z^M`s*Ss~mpa3g+RvmZN;`h<7oAmzSbxt9@l6&iwWMDBaavhDez^}481$UN(CZzKJE zXd+VXa?71f+#6^$T7lgA6{1Pp?97l^0+;439@dfmTU2pfg8Px>-bLI5^b~ps9p{Wg zhJWL?pEpt|rjfp=a>#rQSNtkrIqA=erJ2=i)n2(?P4K=YZa<3u$NGc0S|5L-VH0ja zm5|x(_@NT%2Ujd-s$+Wn9y9xVCNe zUM22B%a!+89EuwWhfE!~gnaqu6Y_~R*!)#7kD-uUjL0aZVbbH(ZNVq9Q`(de-H9{TR&J#{JRrbL`U_?~T9_USr8v}(xofUD(n`v&?rGy@5u8r%!wB;)05 zq%A^&k=+01a_iG(ip}yIqm*k7mE!}4dp~)mqZLTGR7DA1mrogApt)!oa^KrnXvY!4 zmG%6h8X>dG;jSkAm*^Ly+!rnPg3p+vqnlAb;&mKQaCrL;7dXVx-)6 zEcePy%vI6T=n*8`z07&!_{~t?57o}#J#rL#alMe~>~QbBCfzI`&os8m_56bKz9#NR zbO8N^T>oY}T=8$=$szM1+^BDpiNRcrwjt#Fqs88wj{bQo z&xvw*-jvi2nLiwRUn2biRE(6n)A}L(3+5{*f(Evuccsnew$C`XzxA*9AwKSD5Hj^H zNci_m@-#yiA?4aO$Qwx9C?sJ5Th|X_vjpMFdk8YKLna5V*5ApbpM@48}cCGLqN~dNIG)OzMvf(oK_Q2d^XKlm9Iz{~P4HtVf!8+sSX&RZ_ntu1*PM z=k&+jJ6WV1htE=6)&^7~p2!LRTY>WD5u^WUVu|AA{_oCqcMyp;L(KJw)? z37IRSiT3FR(vLuQAmtYLzv*o!?z*r2@*Tuh*11ai9NhLv`xIZ7Z5}eC;YO`fZXx~s z=n16U6_&f6xIL)iH;nbruG3OXu!YZ^YsCC{wMF&u za>hm)zZT(ZbqB$W`3Rlm=cI4@fu0wLX!nJLhH;1?dXeD|F$@3C=A5gCC7y45C z@*Gj-Wg&B9vjle|>AyzhzDsah$IUff9&sbl186LAxkdJRmT^Yrl zH@&!bhCx|bZyK76 zq@UW$@k-r`EKlK`_#tx(96c}IBF!@N0aCwJ_c`9r#09soj*3oxi*Zm5?5W2XlJ+YX zmM}3hUQY_H2$|;16a7ei@?3~IA>p2ieN!y=P2#pn{8rW{*+%-Xj)jC*Wc$;R&5Rj_ zV`5i^%uOx*@pThGb;;8lNXtZQ&9~=@{4JpXTq)cptrGR; zSFvLo_b?;1H{05qP22^jHInQ3dL69Z`}o+h<$G>ougMLWhv7#2yru(rx}s~4a(i)n z1n+j@#-mB-9$x0ui({7hF5E)PJ)e4cJN6dC?avK~c5H6sI}b6AUDs#D4?dRlpD$C* z>oRaAoYt`Haq({UNxPAMO~}0BaFd_Mbh-b@^O(xbzLxom!JzVw{z+C}X!=s#|PotSgVx99M^%dgYN808{ zo@j5zpN?H0v0t7O+|E|(>0lr3rEljQX6eMRJY);-%L`h(|4`ziVYshy8m zJFEW4eL_e=eYWW#KG$&K!FE11&$qLrf5_B=tL@bpq?fkpLdi#d@!zsI$#_>1*B1>$ zvd-X+bBb&`sK;M0AY|G)+@a*zh`vV3{m^p1+QEGQsLD?~r;qB@;@JE6_7>WHQMq}; zLgoRuber)|oAkfZK%LE2xj*;~me+x}K4<{i@HXErW-h4LePY($&#WJ!!$amxxH|qB zP9Ax#XAE2AX86_2dxba&@3MUx$$RngVYgnvy(bg=GAqX-h17${Eg@6CeS*7`tjp16 zhg;2Zk0|xI)!9}>?)&!4(Fb#5a4&MWdhhQU@Bh7rhX{AHDngU)$<|aXRgzrV~}#MvfSuS z<|C*p>V({K8*lx)bfNFx%u$S69qzwKpZW{WFd^j*wA_n`y9NzIH=iR*-VqwCwJ9rT{WVq1UJzj+fvW)0kk{05;vdE~qp##XuC zTJAXF3Xz28*dFTsuIT+-$8oqXkbWWBgp~V}|_u@~xrE=OHz@xHw}zJH^ce*OK1 zITDXUF(6|dKPTabe;v#Gg7}f-Q*Jr=gL<}w=jg@}FQHFO*4;SY^gPyahx#EG?)h*d z@zneb`CdXXhkK;uX6)t~CAt)~Ml$c;3}2W93w?WKJyz!Z1rtLi7p|QHnHx!e51NeB z-eKk9KkDrwE^80hHqrCkoHR3>uy?+FE_u~Ae*Mjz95SbH&!XCU6?q1tJCJgZP zBH~t|btoL*J@;IH(Ro*aop;GNRO(sbl#qFkYf*Z>Z6#0EzqodZlsi9;HD0^lS?fe= z&e0eS6zt zVEsQR6Ep;=y{)Xho%b>yL$lE|R7xAZ>sy{DdnHDm4T!G;>{5SKG5__&` zT`K@r2=~-3iGJvN?D!Wl`x5qA-+Q%*Ylxbn(LK{lfvvyd-%`uIzZu^y!N%feLVQ0g z)kLkmEy&Xm-H6l=`kc>e#H~VKq7A5j1N?Rpb3FV|km=Xo-xqOx2v+X1Av560M19yx zdT)P*sf3iP^PQ84I|oJ4sdP@!8h$+!d(FN6`LUDpcm#fk&Ip-|u8Hx(t=v!UJ}2Rf zzr0Ua7iTTsSV$bXY7FnBDL4<^X3t^6s<)?FQ@Y^Qr=5S z{f#`td-1Hjmv9Vn9%s(tdN*7RpONPWRPI1xT-M0nnSK#**P)@Plrd2uZTc2`u${G& z=p5E#yCshMLDEb`&mgtWo+I9T;#Q;2(K;mS8oJ&cvG&QlN!B$=;O=y|TS&hLdHheg zH~IE?&50X}o<&cfXdnEFq@Gq}axPn)kHuc#x@GsoanB?DLbMD?^DX71_a%Nz+*U3Y ze9l(J$`vy?x8CF2$JXo@L+0MA66La)G(VzUNbS4J+E*dPGu2Q{B;&>HEQAysXZr(t z-1)Sb(q5Io-2#`XXFSL{%1PwYbriAhKUqhKynFCEN@-EZ3b^Vy^UR{NA;~ghmcV0&3lP?gTw7Xo_^>ihf8tB*FzgKHoEgV zu0^vi`TcY`)(^5MAYAh@^Kplp{5~f=pTgMtPu~BR4>xpm!ViawYhDSN1`hWwj#0R#yO?cR_Ud_C_-e?Umz(gz;hHzc;J)f`57)Yc zc`anlxaRQN8-Y8=;okfm_f<1qlQE-P&!j%cce2hq%C9#q*t+vnt!IB;&w=}a!%ePd z)#$FI{u2ICJu|O!zFnK}!{MrDMR1E8?%}Fu`Sa-?dM4~WT=lFF?ynAa1@-;f*Hg_W zY^B^TqB-EUED7Ee?seH>&kOG)yiVJYJkCFx@{o2i`;Cw};BbFoJ@RqJ!#_LR&2g;p zGFPXYeU_V%$`2lbTL|~aUWs{02ilXP$P+|rZxze^v`V_UwpohFVqfxo;<>h6*ZYAZ zZ-&eza3g+Rb3A$KjY~JDINS=Bdl_+8Aql-?TZX+CFouZx;Vtf$fXk^65B*3#ZCn~- zvxL1YO-S&bCT_0q2(QSt3~ujpco!S4&tJegwqx&`q~BD9aTZ(kL!CH@_u~-vv+xMB z7nWUaVsHGuhJ_(B!m)P``Le1r4oypNn_BK;#65>36v_54?A^!yNR0cO9PWJ5zk}XG zYVVjh$#{Wu&zyiH)R*mFaOJ*+g11BFA&1+L^rMdD{y(YJG?<51C~SH?5p!s-W6P?Jc$3vp!8T z1Bky9-HcprTRYxSZr-~glhQj;e zUBUfGaI@oCSwACBEmRjN_e{&Zkhr#}JGvZ|Fd)=%bgAX0cJt4R$os4VIQI4<{mz{{ zKgL$MJeniHyN|eskc5w$@y}oQSL}_!&4F805)+;v-ycEd&knbP<;r*7zOdZ&uMk-V zcLw(>#K(OrxgH5u&*N{&_dD9}a7S3~Z60+6Nth(tGPrUcv#i8-$U^1@hdYh*?*v#A zVXJ;9wp_0~bp=IGHRN(5!~AizaL*!rVR6XZ16RlEt0L*~_rEnIpK#lgy`kmGxOyY| zCH(F8fIi~>Ke&;2YVNV4eZ8wjnrZ29U$9&`FJ32p1&See9jn-`FX;MB?wXJ}sZYYc z)sOJZ#i$)pdtbNQw}`7$!80eKYG?(8t?Mk=m;3W-`L4UH6J~D=nJIA9-rl5N`fNFK zGh5}J>`%MAxx~GPzCa(N;E9R*z4L5;q4pMi#`UP{;QBV1Z%HrTYy6$9%-6)fc3C6i zvLik7AexHi6(#!2{`MTsm*4*)WZr(**V;Lm^b1f7shy*IZbtp1Jo7Mm1_kAsH*0IcEU-Q;hF7$i z_O@@LUX_qmz8~mSO4wOs?Yy11_t9tQ6C`%_ujw=Q+I3=`Pna!S@4X(bwX@36UVL4< zE?dF1$Dcf^FCpUs;`$=@nNP7(g4ie_XDj1hP|IrDk8WO+e_VN9iZ|mvk&;0Kk7VY4^^nQu*Z$|pA zbeuidYWbhq}~!T z_7HbYz%y;o9vY=iImEKqmAS}2M@u%duCR;eCI%$x(^As@icdaftH<5b=XqO*`xlas z8f2`*Kreaxsr9MoSH{b5v*W3`-JSH?wxyb5*eZ9j<>nJN1s*xKCT|&BVQdR-t#0+fL=MmMVN{rW3!e$wN0IjKvO*HBX1f3%)Z#js*r`GMApKz^BW&spO!#2{>6h(KH8a>M_eaY;znW);AqivI zu7tbE`d7K)2jRw!2%D?ms=WoI|2-wmJjGVIH~5{Jx1PA|Xg}JGa_V!hVNKuOT)WOB z_R6|q^vJN80$1DX6OQrX?+I?eHkIF%_GOLF&1gd0)u^oY#SWuQPQ|b(hNqzq>GRNV zq&$9BLh8H3opG!ee|Oe!)=4B3R1TZHT943m z*=UupX`Gj+N6q=&9_S_{gdDgh#!1Heinxs9Jaas%jNJR3x7+qf?{|-c!e%*K9al9W z{ngaRj%<}nRgvKJBW^U3@E}|FJ^WhUGVkic0p^CoW~XEC1Gu0ph~d zeQpi56)Edc_*0p(6?^5my^^Y7Q)^Jd4-H5kMVBDuK4G~di7P}eqglxHZ^f`bt`e?{ zt8$MCoBJK^64HN;HY4TEwcPOWoWtm1)C_fkp?zwhGoFxj<(%Wf<|l`nL;7K83{vig zmisetN1x!CY*Z8Nm9b9^!BU0`l%e}s)XI?;Wqc>4rjRZHxj)&zrMQK-ZPYIi$aJBw^PWmlq z8&WRI6B4|>d~;8}8z7;8I_CO8>)Pwb`rKl;FTjn&QyZh)-yE)-Xgd)xKJ30XHD2!&EeW6%KK039bxT_oJxJ5{;GfHj$o|_ zAJlQU=lj3uolczlUEV|aw*YQ;$6n<&ak$MbH~D+NWpUNNra{<-;b)1~jDQtc%gS+`3aFJrBr zHb(AMOs>Db@>^fAUxL`{u4hQwCqeBkxIAob{2y?$JBQ6H{{wC@+`az;Zr&AP)1Dh< zbUv2+eIjXJqCk$FphG{PrQ%Eo`2FtKoan_u)TwvXxljc4iyqHr?NdOF1>sPWv_B;D0NMF@)^y zv^#M1JT6bZ%BT)fd%3kpg4dO}k?0|GKib=jdr=sSRb`E*$l7~3;|#e@lh-3`cEgqN zRv!O7hxE(QDx}_5tLIvB?75~MDF7FP+XJrFhuRmX9P~V$aJ4>Myd%{#>;W4bBkrMn-(X&nDm zzG-kK*IZB+B>of*Q)~&|)x^EgN9e}1y}Fz!z!}NwW>+on>s8)>u=yNr`FLvonX=Pv zP2*-3zVAq!aGAzR@DJGk+i)57kncIjhr1na#HJU!g}V@naydK+|Hjqvh5Ayq{kTtQ zD2WT;{^9sjxi@gk62x9RuS->q#21pU!flCqBl!l9=k&0@7%KTHN$ppXT2_wc|~IVU?1u8hK9`p4tFx? zpFy*ca+_K1GU7f%o6$3Wq{Z)B&hF*gn`_Tqy}vkb82t`hxBg1q+~#mEw_JbY|0`+A z@~@6R^Wm;=?EMMe1N*r5*5O`ZxhI^(JO@c=!B)6)alTuBb$*aHB5b~Jxa~LQUx07cwT7{H*h2?fQmwJWaV=)dk9>u59^sv%_L6| zQZCh0g4gDYRC7~4@36Qx)%;F6J@3U2dA6QueTc$++Tk_^5Zsq)j_>V1KPTDR)c%9>a7p-^vp!Mvi^O$R(rl=7LQWt93`jBt^ z5O~@5Z|)fSb+{V-;3n19!7{|4jj;OFNIs|*n8Evez|``J}vjJ;v|zgVPTr# zvo3%7R-&VmnRk$1&&y+)rI>Bx{RydE(k2tUNwxW|I{PKad3nfo8t!-2u05pR1264- zu5%!@t8<)W%0-B4j$|y9%dzS)zR&)`DPi*)oCtpxp*Ezu5_Lyv-xtIaybbTBn*PK~ zkp4-YYkr)98qRoCn4^h*!j9Vu9}1f)cP8rPVDjCD?n7G6B?6)eevY`;&{8DdU1?a? z=M_=Tg%5{K2RM;c8rHEY3nj;=~knJzTAa<4OM%nue6C>s4zFzo%bgQ8mG4S$8NqM{wHcjBx^Yfa(MO+g@-H>wSJSBJwiF*%yi9SK)BZ>RR=2~u325+y! zjl9ftBDh`RSsCw={>)aM8Hkj7tgR0pwr1>eF?9j8M_taM{Y7&8nT!2#$ja{-tI?ik zFQ)yleptYNk0$-&Xf9Ijr)b=wl~4ZtUv&gTz3WWeHAq5#wsM^yrx$H3;|i%G5|%KTmU>sTDr{=n8>S?^j4R|_ z#DmGFb{%K!(s6~3BPD(2hhcLG`K4Vbf;WtOlh6#LcI8^TDz)P}EgFOFL`CyiKHBP^ zzuRp&%K5t#PIOJ!+yqz7v)ib*kCOg1va>c4c?*&(YQ2Q6`^cGXcW zq-7vwSz_y(+<#Q?W!RhquZy+sOwx8h9g!Yif6KdtxV2~|ZAbrQ^zYavJPG?@Nyz&q zY@!bDYtn}=^-Ke#_KoqAW{e{4$ji8o5p5WoVj6#gJuvH22Ffh@A#ARImy3OU(RrjD zh(;mh&9pqzk@d1p{1%FEW7+)8lm%nW65EFy_bvYWHEi<4e|{d*lC=HNFqBSQZ%%{H z{e9j`#O+5%UQU}`6T8UN(9Q!&tq*5nFF#}Q_JqwMxZ3ZB$#Xi2BIWM4+`hy;f##s; z$n8hl+GCgR;c{x3l7BIe9mDbadCi-oUyXi1di(*u>@v>i%z2FZBE8oxhJ6d{9QJbj zo4q$|ipD1TjR#0O9~C3Dv!2iMx?JIzeW*?s?vX*u=m2x^>js$pt(|pgTc=@Lv|N^{ zIxaDeIG^+_QCp_(r4mSL_eJEGQe}!;Qfve}kba;!=QiuDq z<(}5nGnXL=1K2j9>~oMZWh?%b_i7a#m1SBw_TEPN@#uM^erWHz-zz1qTsQi16h`TE z^2_G?+>&8_TcGx4XJ(mUa9jH}nR7@#1T98dZrA$UjQv+J)@Px*BWn9Ws<}Bm&Ezl! zD6nImyk3k8YG#=?;EjzRl`&6|b|#vG)Xt}^ov#r0s~rS>d;?<>+ULfol=DG-@G%X* zN?1j3H^WuGzE9pYXr04-#B#qQZXc4+X$gIGP3kRNWyuyNW#h(Lki!iIzK(xI!dh7TXrIwHv4ha@|90twnzIZ>(;X zdDQVk5$WGRF{JfC`epIM`^0^MHX+v++16HJOOWqOMCxUkVz|a+UZbAeGl#aI5>(3gH=FTq&T`rT{8ESC*LqvpD9cot;2)@DzNr^&Ai4*sy%FNY z-dS&?o5nM#r|fg>9dFxQ-`WEQ){-&ztSob(!+nH2^U!>UJ4t?!`Sd#CwxQpVdp#wJ zA7W=`na*&u{N=sV5e7o#=n(a|PxLunCUGZ7M@hKo)xT-C8fTg7;YQ-Ac_$h9PDibf za?kSjc|C|5jBZ1BEaUsH{NT!(e%a5p$1c}x<$jgSbFxf98QcdQt{n#*lDjgAYtGFw z&p6zh=zCv)`KrSmXYE}=+*tM(v+aYOGpQJQT`9-bExpd13wJ>o-1QFk!T*4po5YR5 z{S+|P%T*D}jAcQ|srccY|3%IOM6&L5qpwdH{?_xvfazSHw(0~t%; z_Jk|vyM!(9enGpD`j6!-3GwGN)35jIyCjEs-d2Qhxs{Pnr zM*9)X$udjr@yook9lUGNeMpagq;IG95plavpg&_`bo0gdkN&XyYGZP-H3wVu_~mBb zt&Y8NUAi%Onjy9KZfoyfU6(FQDlfS%T?qFG#PcwQhPB+f>%V`dL&^x z+XCuk1ixy3vX^@PE#ps_KW28xGPU7qka_Jc^6Wtp%V}lk=HvcNp4V!9a<^^eAPISJ znT@~ro*e}2k!bh%$T)-B8Qfve+- zR;0fQO+w1uYq=={IoDBZ)C9Tr!_2o_9k<1LWSNn0wI8^h^i$CjNV$4G)XC!-3F?H} zA~|RK*7w`%B3pI_TXCg{T$^R4Ioxif?}KhY%B8!N;62IMbr|sy{w z3TV(-sRwZ@R`tiP%FXYUWxjN{$CBq{bShHrBbIvwao3{jQHb?_*WUU`+``^jW+&Wi zKd%`k(jRd%=Mz$HU(3xW?lJTdnvL9XL8OZB2k~!D(&yipWx6}uPf5QO{fLyS^X6}d za83+mEf*bu#^VDS$B7>vZ{%Oc5bh@0p3*^C=1#}I&y)U%Dk)|GTjfr(_HH8Xd-Mk? zMQ*v3B>C42&N2_e)%M|@jpa;*VLbnalsn6E{Sx3A8TQFDf-bket!H{3=iN;I1~+0e zT}z%JXt={&V!7AfmTq1q{%!OYa^F`p-g2efWFA^PB+JAcdmF=Fh#NOL+%-w|iU~A3 zhrFMs1nwse_j7prQMKU-KYV7nvx$2XeS%h_PV_nYTv;Lgrj(!h!Q7H%_B(#4bPG(> z7^#1ETkdtl4Mn#jd4G|v`3Nu1*0c9n2FT}Fir`jwFwqZ9CC?jZKT@vecYoefBe<^^ zZA0H6>F0ad^P=^4e*M*cK5_^BCS0u#;$OK}O73k^Zf)Wu+TZy9{+)lb;hyJmZ)MJj zx*+A&^-aw#4O-iuw*%p&3?RHfZ= zx#D99%FVqq%UtW&`wsbv(ME@ReVk;x?}^)k4j{RXCheJCKQa^i^I5r>`B~;Zr`$4b z;~FPA7OB0JEVlu17b6LmvF+<{eQOUKkaCmpL&3-_^SEQLyt}9yd2T?;CA)+_%k2U7 zeM~yn4|9{w6wrQ|*4nbyCj@PjfgDHxSycEFALdCj?`Z-!bT z^~0V1KCc~dU66zdd^dnwlK-6VslDaa`P}F|S*9Wj8#?ZLDK*7hP2Ld>HyS4;{QE8Y zN`&vsBTMbtzn*VqOqR)l8}aj+ig&PPhK@mMulk`EaknA~quDN?t~I>CuU~4fJSST~ zeaIV^Wg5V3Y`OQ7{z)_qDfb?KpSOXyUy+3LJ893l(LNw$%J!01jQI*~=6Kd8;A*)^ zJ*!BbEF?o@8COq=lMFS}4~^L8_L(bf9jh@LTj8}&;?m{Dw``xk)N$sMUlx6O5{5gyC>(OUOx&N}#NMmOJu(DXqK7b9KUxLYxAfklIC0dX~%gP7ZEoExxUmg z5?@Nld4l#3p0tOX0o+OY2hdAM?c{bi3EuH{<7bqI`l3#qcpmo~7$5VS!w>Gq@(MYx z;A)U{$}!~mKibX)PO9nu<7Xddk*uVhon4xqs3b#C5mt&2ZIRG}Eg>p4k|=BxNwla` z!lDw2?LmcPQYca@Aw-j>lCUYH2mRmQx#w%nPGkGkpYwWs+?;#Pd%yR49`Cv5-gyA2 zovk=Y6H3cf83iV>%N5Y zSGW(c4{+tr$hyj*Ta!(@`(yP~=IK&TbzRo2rv-Df%y|j&=R5g(RFc1p{5LrHFCm}& zj(j&K|ApRlC0|Y40Hpi1#NPaQ%*V;Eez}A5WeIU#APE~e4roL@IJjc{ zP`^Z8%`(#+ZaL}qpsFKd+-IdBQ;sJQ*BM=n3Pd#-Z>hf<%Z$l$`Y$QR!q>CRB741k z$v=Sf_oEk(jt2~X64G`OSK|ThjZp^cVgndsWbLLMZG`IPg@sw>BgcPIeyvG=5mKJi zMau6V*VX&lHYfF2f~>0-!Yzjzw14Juc)ihZq<*CdNbtQz+(&3F8XbuH&bXc5>psW4 zaF)p)8!N9r$df*bH8rGMnpc$Dn7B5m6S@c$a1pw`oM-!6EwAE5jI$26C+TlRqmXjj zTJ9UfEl1y?&rlKdK>LYsHE(_|+>ugm7IS|JH|XUxf091qLDv6}a$mLF2E<*23Q%X{ z{vLb2tp|EPUAiR83~{)9NPj!J11XoHmf#yp+$^*Ny^dUNK^!;uPL_EHu8e0AR+4@# zT8ES?>v#m;AH-FEi09krDCCakg_f)H@Pc=<*b^D!wgGS=d72{S{t(3)uYvLJJkE*F z-Fa#)W)*$5`ZMjv8Sz|!QpmTXf;Bt%O>T7sqmKo-7 zPb80&+o>Ft+sty$AnvIto>TU9{~Rjge$?S=xt-^5TUc(?0ro#qZk7GMY#!XX3Ah(J zT-irQ@EjZc=f2l}^5L#>xEkz%7P*M*{!e?{Kf*53aW361a~g z;6CVZ`y2qb6z=Q<+=m^m%+Cn>RUg7|KZNV|io?kFipRnn?{IIh+!EquBMB{MR-89! z|Ev48@>XS;jS1|1&EfJWUcz4XCaP!Z-+Z_!iX5)?J5M^?RR>^i8QfV8SH55U49qDG_tOO2ef=(bIo#!Nwf}8f#hBB- zN;c0s+)b7%zcYC?=OxH{L}QpU$sElsH=Rd?K4JeFTy4isvK}=P=6r{{-Ezx*O*VJb zPBr%;KDTJN#fd(Esu;$~^iwkDI)0XR2v|Uv)k$>BHo25f!NU}+OriT(zhZ8pq z%|yC(AZ0!so60_8-xnO6FNaC1 zYm}%XQtpXSlJQL;ZWa0gtw+`Agf_p-`av`HB^{9*eXmj5XUUh$=bY=7b;ZmvsvdntN53+AJ2WCyA_I9!M?$5s60=Ov?z4|kZsc-{$ ztx;QtJJxcqC+J2rb&rYUos85E6FFJgy2tVa z>mJ2$o5R)mFbm#lRA+3=4^zB9>Z>-6=S^q;>WeZK(yv~?Z;G9b=6LIDT5gdIS>`_L z2g$yH^vRF$jy_WVuCv@W#C1gy`pNmW)Q9;M`ziH9-X`wL7@*bOBz{ZzX7UV{eB_sQ zKzAg;cLe9c^nWv%D8%o#E{nU*DTX`U@x$Hl9z%~iTplS!xhsfUha{ZK{Z4W1*n9t4 z$J7tO%{<3)xSPqBG@dm_r1omNXh~dW)CcuK^k{xwTizE%TQw^M}LTP9EO`)|HUjd$HyI{4;HlcnQBWNsiv1#@nmhGPs6~%E4%A zekYlH!fotuyL#t*=M&cnNm#ijE#dmJu9p|%{KFk?SMv2ig-Gqa)^baUdlyMq#j%`< zqUEObLAl{yv&HV{5<-_!rvJW9sh12{Vrr4kJ&4AlHi*jN-=eb5253b``n<+u2-pli~eB$9bEPA zd8F@xu0-kwOo(!yA#Mhe(7aZ1h3E2GZtCCQpIPP#xDrY@%pu=0^c~XvmuaSiw04tx z=2|oaZD+nUhv!q$r*EeJ9p$Alk)1q`u{?>A{`@i0KaG@!`zr0vs|37fQfjNLW9RcC z?R}14r@?y}y^i$yPw{NQ?dITrC^zU&C^bQdq^bH8`vH*JW$U4@4{=54DfAfXy^6Zq%d;!fmc!%-@9d@-ljrlBdT{f+ zzinP1{W^3{NsO!g=qb-oM(Anu5GsP}Z|&8^uBUkOfIAoC3ofYOaKCvFZgb1+`Yii$ zQ5dPcYV3M<0iZIu27Q+pCdCd)^zXQ!fYOnSmJBT~%Ip#{p zkEGnIuvRVmPD(8|L`OsyWx_}K&ChVPKE2CJVfp0gg_PUC+B=xIQD`h`gAd&~d|z8n zwLTRe?Kg)$5#vrK&usKDQf_0*{foFGr!g-_GQV)&2?;K%_&mS7rr#V7SNqY^$a6O8 zgp}Laa&IE;KJ*wWMiurq-R5zJk}3M#(V*X)4_DjkbkZ+G8Dnat zQeU7xpyN?pl-Y^T@o*C%^)_VvpyifV&u=E&9V<84m)4d%?U35r%yRcVv88H_H=nzapRdVaET!&5PrfH1L~m;O2RG%?@}eFETeo%B>}TpK?2%xHFK1^Ei&D zK;xg^Ts4R93(^*Y$N7!^wzZ_sCw)(J4N|UcGrrr1n}{Sl%dxWkd0`q&>XWhhAp54K zlV<@^F56KgctHM_spKpzukF;v<(S5?&de9e{i^8u_&EI-W;TUXkoc^iIexSBKb%bGC;dZ7bHAwrSL<&B(w~CbA>}gU zO7Qh3?p`z=jYjS`RABpGop+YC@SAoCxHHJ7+*XY#PV<|q9quZ4>(Ca5tL>ud z%gi0n$*7Tt#-#U+(``FedreEf8SHRnJU)j!U6I;b)!JLxcpQ!!k0WqL!qs_+jGH&Z ze7INY-`wLZ*E@Ls{io+{I$o5+ecIv5xOpqwM;&|B50#CZ<#FR?4!CdWd zPqp^$&$yX?rr*qS?3HnI9n1|5x5#oU8#jw={Z&7hJil25H|XUxzro8aWgdpK+*mr1 z;Oj@+Wb`_kE&paMFdX6Dl=*dmouBFN-<6#0H{0N9yO42c6M23{%4LX^uwUa)UK_tj z=LT2ngN#E-GuaD>l&kks9zO9nv^;Jc3d7BXtMDxL@N?{yBcr6t4E` zZQxykx;oqsEO#Jr_n~6c{hTxt4950}7u#~ve!cKqzv&7$&&wO#C;k}u9!KhjGRxhc z{x{gxZ-zScSK8aNFy;HL%H!9zlS=?`;2hg2JJ{U5GHYejyspYm~lh^m`;Rv_$n>BE?{ys|j zC($!V%Z*2(5_~g=TZkmQEk_0@oj(X$f~+ga{ImE%zxgo%cZI{1vM1~px8NebNq%~7 zdp~oytqy=2YVS8WaH$5-AocechUqi#)g|IrjbpZ#&xMaIdi3jM?m8 zLK1=;6V+cGH_PE(0N2)0(}48NQ6AEAd%|)rBknqM6O!-Jx&1>pu7Akv;5WCx&GYg` z`vNV{_K zKR`ZtKKzK2{{iY9A$c5eQ;@s9R7n1Jo&3`|FW-Tj?d0zk&2EyH5w|uz|HXduCHb|T z{L+PQ6p&^UQh#^w{=Uz5T8jCLc;6iS&NnaJ_xj4Lzoi_dokZZ8XJUSldRHA^R3)zj z@k$Q=!mZ#vIvtC3IbVJSg^2a5jlpZA05cUjCOesd9Ay-wjaADnJ# zI9$C?vB>B_+Cie{@Vd^N33t$czzxEE=0D(u;4c3UxOs4Y{|~tNaD!}Q|F_Gn0B*VVEwcM(& zaIc11pb%=!7Mo90!@?P1u(%*IQn=-f>%J{M5_sO#w$!Tf-(f#Y}f4ICj&CI+k$@J!^ zeSy$6Gbb&8^R460Ol}t3-xU##*Nqt8%(K(XF7~NONTbiF>~{`KSHIceaKmf)U4~hF zzYOVhD>njngT3x@j+JqP-TdaT$+7zSHZajr6wi$sYX2kWE8`ZzZRBwO#Ez3MV=n)7 z>~{~;UNe_<2PEOBzLj%}y8F$U4)@>kZ}3XLxzyp_(kjV>F=;dFJ@I~6@?*03Uw=Qc z3~qmiTb=97g?lH(9?#9;+%eVoeT5kGKIwi!2hHExzG1|Di4K01 zzk$TQZM?hH8~YkzU%;_1zc1tX6wf~SJM~E46g8K;{J(2oHt%ttL%i6Rk9}qDdiISZ zT{)Z!;pqIeJU_+AJ+8dxtp1aFUxWF;Zq8+N@FrWbUo}yGs`p=o{ru*t|A1Q#cYwof z&vjji2BLU-pCN7rnv3K+ZSHzfJNzkSr1s|B=r{Mn#XZsR&aCvm*R|uhL-{S0i~dN9 zrdvt6P=foJ+Dy))^$6}{$KE0M?J9C!gW|ba?_a?uo?QnnX{?F}V59bc%H{&Git6hZQo(|U^P0hX=>gHPX z4T`T1n~B?l(qE%b;C*{{JwIaa2Xwu;@Mgd1?HI6wht`PO>?P_!cPM-oA0| zq8x6a!xcZYCf|h#xax;X5^&3{z1rRi2l&lshkGz}zp{2woAVO()h;4%r#sv!T>L+A zQwvfIpBnt1xcRsE%_6u#FS+T)^Ob+n*N-Za2dx$r`Fv zU(W>ojRN^kKhCw~==Kk0wAZJrf6YL@`Q-rZP4vFJ+FJzo;O7p^50!Dvt$x$+0J)X< zw+QZ;aCLn9x9USV+>8GMdkY8o%{7j_2U;I8caC^YjdFmY}q?%#);DH2wIGVZdIL<$TWIox_xgW~6e$aVp z{!qUe;@JED?SF6NxslAH!f>B(xMQ&MpXPl}aBe!1u&?=K=x*vS++Z{{@B95f=Y2KV z11I*D!d>rhXJe11*P-`wtS_s749xH|tWg*(~d z{#*8jhO?gQaQ`iP!*KV&)$#3H%H${XSAue@_B#7-P(#$BdBydTQTD!2>qE)Cesj!= zF>W*R)F56uNT41njYi+Tm5=D+*9EOqh9;>`t`yG{N{3p z`)}1V`Pl7Xxc_!ND}wu^!~M7FnR(D}mcZ5eaG>K*aoji*g8M@p_dv#>lDKiG0Pf*4 z4y-;T8ZWfpiNI~}YaA+v`+&p!x9lw%?KiK&)%tLtezlW-X8-H&s862;Fa zayWMkk}!^Wab>?#8yds@9k{x`A?2a;gj`U*Uc3YQDC3}nNgNaTTlU|`{Jk9RdWXB7 zZ}}#gzmqh&e{(D6B{V^axPK5Y+@i65v)kdm%{#Aqc;VwW%F*=$Nu%$>POz`A@Q zuG*VF&TsrL$Ncbb*&Bg-I$Ukn%`r^&^yvQ0-*M;B{6E2StoQGI?%ynj+sWZh=Q^b> zyuwlZAl#ZBEBgKG`PL7&@W36W4ssTa1X7T8s#o>xV0?z!Lk(7 zknCwFzMdVZNOO`9&BN$`3PjPggQ!7{NS<&-;(*~~A zhiRmL6TO3!tKSXWNZeMm1GPTB;`&sno&RWk$b8XnIy>B?MeO@Pb&zsDxAwLnt}D6; z^+98p-1TbbwTo~ZH!_{~KOFzwM*4*)<*gWZr{%67?hoW&%sT-{zW=e7{Yc`2f=ynz z$?vdAxrJx>&BKnpeMtW#nt{~b3%udbS7!;|lttaqr6}B!JrGE^Lmu|_YpEYf=K4*< z;SQDb=u4#BLdzZfHv8Yv60`s{<(YqLB=0X3+I?WkEqvW?s<1Yu_2H0rcqfOwMFEb= zU17PW64wSvSkbX!xrxmZ{;m%T{N`A=T7NGh-~H^dy4>Ni-Bp6`e&WVh?&BQY^-7N# zYZoPN_@iri%6*3PvIl4eN458N%YB=;uaSfwI3^m0^!~GKA@$7hLpkXW3-i7jN97)x z5=Hs4-sRm~B;l5ZmHQ$5rr!*7xW|$2G_(RK_jJo0zKr*I&{7meOW0G_>szmXm^3&w zS})YUk@x)O9f!O4J^B{pe?P{hYD(~ZO5AU#?sCqf)nVS(Vt>(yR{Zwz6<+;SZb6x! zefY8Y<$a`o68(wP-Zc3K+{_hx!ylcCPD7Va$2P;3{;c^TkNfrsUb`@#@_r+G^R(QS zl3x0Xk2os#BFmk|2bR7eUc!5av(8PX`0s{GxygHH#c)r6tNqSq@-4n5-S}32+z* za+BZ6Ql=c`zCi5FU*|Wk!`1qr?f5*0JNp3IaZ!Z#xE%k+&r@Z7#-EoEou_`vJXS)4 zW1{{+>qGu}?(ZG{%03o%pM`R3z}0;$!j0d@qP|f7mcdQuLbaa#TipDwxu1ip?V<;M z?Tseg?!7Ol_F|&15Ag$$gu6Ia=7-R?%&Q%4{C$hd-LLx=xSiqVN3GyD7O-P1nuer$ zrQ;xdFLftz>8tpC1#~3Z$cvBqz4X4ey)F9EnEq+T1UL9iKe$@YsveqT9%rs~BKeeC z-}{@M1MJ@)PP#<%zMpVtbbhwcZ;Ie*xEXp1v*2j?%RSX7zKi%dY1O`d9;bqRa^K~$;_Fq>CcpVG0e8E@op}J<{2!U`!VP+P zO^uJ3o1;9W_I_r$>xnCu_|?qyIJ)(!IPCRb(jN@sT0%cFuFr}2;na>P=4A4q;c%M+sE#MwINW2SSX05ja<90WF-F+x zTbEk`cL-eJNoWhNGwSYe8+fODLy0R!kD(j6*h$WF$WgYw?q+Tz^{VU_`r8ED$qx6F z|A9L_j%&8~%>uaE4&H@V|Nd07+TosOx%U;Pm`%h>_=9799p)i*EBZmcJ9Q1d&-~SI zK2PA^A5V^zn{5aC<=-IOF(JME;i$3A|M(B1_h+J+YjvM}i_l)3bJD$L^=h5UDkCcnM zqWhJHuvht2&R=z9+FsA+m&eT)^0)iVjc_$Q4R1Pn9Z5fPIDgQ7hb6U)xDDuMRI4+oB!3p8_`+&1%e4^1Oo9Amxf}1mDqX_+Ah?9W_N|tg!^C=+a(u zn|bq4>1X8kLh^U{O%3{24MW+4FMG-h$*2Cj+B=yl-?6%z`1_ITPg9IP;nahJ+oD1G z0lChJ{k-=w#Gg#TBvkO{zWS92+%pq!i?KnkJJEQsFK#*9?r?)qlX!0c-ucg@m}edC zjovxmH^lvnB<$w6l=k1ZcE$ED{*?A__V~?h4!6o@yq|(GM{-nqAC8iYuMTl1Aqlhi zK4_?Z?0YlX#|pQba80&(1g`d1r;+b`bRAOecu9tTrx5oVT8lQnkV`IUg*(L%~BUnHC1#7mIhb9KMlBuoj)ErHt!u9inJ z`6i$ehbwEy1m7#fEkz$9`QBtOfWPpW@PgJ}{cd4VX0{pUa8GEF9JP0i!_A9gjqh9H z?s?nWZ?b)7tpAYj84Fv2aOM8O9G-37gq!K*HSM`R@xw0;_X5jp@NTl{*~%-ef8xq| ziExYH{^jCzA(rt~2vbwNs+Y$iFSY^`7C_HO@h9ux#?Qq{b0B!-?sSbC_ )L_`>17W4W6-cQ5e|BKc0PmXp+@P+WhQStHwg z?{Mof-b+6)#^HW!x$^tgQ;C;wCh1(Ruq9}JSPVDwwOD^RgM4$*8xFU%M;w7}9T1fw$cO?CLTkn~Q z*sJZT9PR*XZw9=-$aln7%%_p|FZWq{L&TkiS|NEqP3C`VYIuEYKK4!`o%Y{(+1cz_ zjI}FyzAMjm&m*t4E9nQwo)Q0iSL}-qyJQkZD1rNn^+P*&7s7jt^_QL;)!uPlQr{5b z?n4raIm$X&T?WrJ?|U|v5-cSI0 zG;z4!Sgt%5Jd1b<`|?Ba(b=Zaf|wuLlCLxB?r?v$+^nGLIMCOI``}alFn4n^R8l_4TS~D#{Mdgan+mH+l(#!&FipmZ#?>;PkSbVmMzU z;QZXtvr#zFItC|{fYZ}*O5kL@vG;W)?T&G1)=_)vC*bt3oKiR)6L3-;PW1j0dx8l# z1(p+rGd2N7ev?{)To-LFYERn)9D7e@%HX`0fO99$5hVzRX-JflpMZ0XoFhcw9K3LE z|J?7Cdr%T`U6&=`Tr1}Y<#5_2;2eE*%%0kk5Kch?PH#CUA((B75^&`Ax+TbU)s=t1 z@h9M1Ddz~8aONlA$a`PSZE{ z_VZ|m)6jATCE(b5Xp;wLSOSi=iw0bmwu_mV=LTy}A)HGRaI{?rN882t1RUGFm?Ai16L4By;n^shT{O~ARy+EWVW`UD(pU&5(vIr9^6 zZnm5-oau0)b%eMhnM=6OZ=CyepWF{##d!&GU+k`1YyYp~b6%}%^BG*sh=xa~*OSr9 zNXJp>2`EF~nXLOQApUK1wZPW zJnyfaZL;C0ecL2|guV1g>=TaOx3Bvs)m%mV4M=`3!+qXZMm=)p52kLmIUQ~ghJ>-? znT*~?%00_F<;%D^)f~B=H98bT?)qHnO%>PY%IjsD#~gd*_YofHm2P^HPr3Sia}V$R z_iNHswoVy3Cfm$*{JDvBl%Cwj-vC$rY5Ts^yNR32I>Un;*HEWKa_zu?8JZpV+) zO*eE4x&gWCekEMH_(0a@Bz<}PY!g@<4_U77Ywj~pBUBH$>wcwiT+@*8 z!r?X{eJj)rY59+{+6$hw4W}5`s|o!=gmvOWwd98`K`1jYd)MlaZ8d6@C{+j%rWjtZb9-aEFn8 z3>uG=JI8Xv#QlP*Y~&sl^N|5cA0w(Uo`rLD8g7`Pgwf08*#4*4XAy|$hw|9y?b zhspOE`8XX7@_RzH$kP<5J=*W~B(5JCfb{pq${CDCyvvw=d&Te9l(b(O3l)B0Se%+@3IL&1ggYmOh%exwga$0FsvVeP$= zxPfRCx(kUPbbYIlt@q*wUSBrB_SuZBF+V&-`j^oon`2zR*T8+Z{mggy(P!u*G-nBO z>XAui{Fo#Y;)+tT{s+%5cu5xNfzK&?Zp9h^m-dz*U998tK}294>N#(30|_1Sk~ zet3!WU!dQS%txf17JA(D4!`mY8a;-3l`_Vi!rqj=e3QwJak3T{z9!ogIJ{Y;{Q#{+ z${TNa>09~j3)CFdJ(1r;i@?OEMYfFzuk1S33*l*ha4u=Pqg#+(_X{4+C-cVVh<^!9 zPv)Co_;VQL-F7Jns#lyh7WBzBBON>6CeK%&q?pwlrPq?{e%<4yZzAqDq~+7wSsxP{ zCFJ+bHj^FRU!>38#+Z%ty2T7adN1M%(O@Lky`1`#F_!03?|FTeFhe);+{n4^M@U=! zH|E_)ulsY4=NtY9-$z0@+j-`V8m(bJPb2z{2FYegr;7VEOK;CMKRI@OPx_sx+Mh8y z^St_*-j2A=r~pN;oxZ7zKCqo_OJZmF;B0do_bIqK8g3=+Ludxl>uzs(KM}VbCGCjq ztrcJ9+p-Z}@g1yR!V6knkhG_v^O0V6FOTP&*MPo`_8I&Z zOyo|U6Txk6xsQ-%HhK@KozGf3|03?7o%jT;h8Yp=X?zRPmd~qFpfGBSkaqprvNO&z# z9`z^U@JjB@HW^EkO<`>5aZWtdAdGR#Y8DjNSCbDktp%}p_Rr+MvCd?5CgEXX#8EQ|T|*n=|6 zDd=>h+?`Spu(uO&1?W0-f3=F=8H`ZbrGFQ$JZC6)Biq!6D`6KZB+p&wA*A){=;&FK z_B?T6^e&S7mv9ZnO2&*J6>prEhPACID z4aR4iXB|7`c~IS~RC5Y>#2eDTAInkf{9m32wex&$3gEs6H

()B@gX=zXN+bSUuz z-`$5~n3vI7v>FAkW-Ovk719O@@Tpw8&X>c>vdyn?_WnYiUy<4?>m&cGz54tr_+GZD z_nue&VwCtnxW75v6Apk|1oyH8+`-=`{cUgE{oqEONBRid2NH0Lm!$p8rE34reo$`S z``Kn8TyISdOKVB=eS^!-^7@0k54hiZKh_@%BHsiw3u!&;Ww{3(nqg|9=IA8U^b%u+Eb{t+ z5?jx*cn(R?n@}0!Al#sr*K{I%cXTyU?kLNBfVgq!X;d(hcX%4n*0>iM#PiJQwhdon z?G1m(cnDY8g@ozkc^xf8%6-~9<*UncllO?1aLTTvguiz%-}Z|ItLVoau8iOEEbBvu z`+SsSyot2;Z}YM6xfj%R7-?$~lv@aQt;7AC^fjx}uOPK|o#p;cT=HQVrWQI9$$M7= zxNhN@hrOdb*Nysj4gL7?m>(LF{sPn$Dfc=qnc!PS-1lfF`V9$pnB^|F{#9<_7un`` zxFOFbb5Lf6IU3bN${lUFt%+-ou14L_7?_ih%>OrA?(PM~^rniJugx|W!PW8n2GZY; zW+UZ3Wx45WCai`+=vcI?9^*+vkGp+C#rwg$b<{JsL2GYk(%*^hM#{DG8Q&b@!brkL z9Oq!`5^JycSc1O4Q@lRgjD}lixnGg~kZKubE>i9eYj3sc8KySsgf2w$zhmu_JqcwE zl1=p|y!k?CJ@4R@D5J z`k7=~K3rdEa=@hVVxZRFDWv}heSwra z(sH}=aOD-W7L}nvtY?h+!Lv7F?-%}lpDQC^Iyl_lNS}KY^GBpy`Tbmi?*-!Kq7`T{ zl6$2g+=qy*1=bJpUZ&g+gbxmwJK+YSS&jLL^vOr#N2FYRudN+%m!VtGb!hcj)U%cz zw+b)a3SZkr{-FUg$KgIg`q}6wq}(d<59}RXGs8TNUPH4`=xSr~knp-b;`vy)=J0@7 z>G&ZjNFRwpNV)lzJDIprv=Y6GE}`6dp*-qZiS>i77nM{Gm=-Hz^&h}m-5m7%kZJpMy`LzDJJjJmO`hp!Hd1aDe_yz7{+w(U6aOBP=XUPgHyFn)%nF$2 z;1)!)GOm#)vv!8L4@vcE#)a7aH~qFc)G_o0%BQ?WQGeyyCpGu_vgy?2;@p7Q08iV4 z{H8$1x*29DQaiU=J3l5)!gh`wJEvB7ZWXcRrS~VLM+Hn~S&VyZJ?^K_gGjl4+uvjz z0}ow}w)Rh{z^(pd?A}kt%S<-qHioO^HH|zo(Hx}Q`nH`lYnf&i5-;H`u3;nPwLR#K zKjH@oT3%&vJHyq`oPR7M?{b@8{Ccv-O?S)3qwLwU!)A=Yxf_noYs$#~Bl;PseQm9M z>BnZ6EF|HjTPlAJ5UCX~rH((3Cf~8BF;XtKwGw=HGhb>$d_FopJK40R{tsh5C-bmI zwjYpoRRy1fYX?jOZhkZ?^`AVip^uSr@3h>(Y{Cem^=K6;piY(8I#puVWTt*;%(ZFE z8|w#5`w#c_XO#xt^@=~={~!FB2PeS#lln80{54T+r1rI!e{kKqeQ73S`5DarrC+66 zj?QDWoOFH^Y!EO*urC~Q0MZ$ryX{JTHDw~={L!1Qyt-LRz}8iC^NeVMq0=xwC? zj`cn=KhEBQCIK_i;r?6pM&Q2TaF3cde6I{X~1lBxIdD2H_B+V zw;$GwO*dJ@AB}q8B)9$+#Q7n3a==tw9jj-5l~(+{0@vQ|^#3)iz2P{n_Uqa%is7E@ za66FqDz5Vy6kl$_y$(s(7uTE;Fg+Y@{O?z|_Ui982zOupErL7xKj4~X0rMu@zsJo} zEB+3Z+KY+%^*dB0aKC{Yl#3%YpsatSpKWgI!_#zhGIt{QIuTcZB=nbKW9A%4%S?`P zzjqBCU`dOB+39e5@LXETY>>lk8zmXv&79Nk4Hwr=JZ}>3llG4!#?U+0o=w3xJ8b=bg>eA`@=1T+s@$*gV&3X#LjfM?Jf84 zl9+H7>CYaJVmfe?b|HQk8a-otj}kW*Nm#~l z8~Zj=nII}t%Ps2PvjS!yT&ZV+(1)a#-^}`oqjDb<5Pp#Hc(djH#xa+B-ppDt?t3JX zaZ{c{gyD{Yt6}?z#>if|zsM)CVz1q&mHgG3WOF6+XPuWv^Rx<>$&Mf8z5I;E8Rjq~ zc0Ege{=9@r-^(wI>$eKvz6F=5Wi=VR{s{?qMvuJ8MY+5}AXPh#y#-WxY3WSE1H+KXw?c}=D7jmOOs!f;Q9tNqpr zVDkMJ{XIG2e{1)BiXZx7NI~0xS?6#+A^jhyYRL10aOYd@OT;Zg5wr?z>p;8e=+*z= z2j2ROaAn_3=)!tTubIsKztrK5pV;-#&a?}- zgd+LRD$AF*7q_vTYG%rGZi!{LjAjEt{(XW?=&E8Q3F!5bntW z+D~pKT;E#u7fqA?|Io9LasU%PooHhOQ2noG)YT;v@1z&{m{eib8_#uT?3gTFVSm z7v&(gzpYX;wtg@3xy)+<<`KBsE}D_PHM$rn*N&mS)9a?1>xu7=Wc||}&l*{--cJ|9 zeGjhsw}?E&=o_S5yFbBq+vz;ZLi5mzXawuZrFMSQwS~8?DE{3^9WJ^)V0zO&m794+ zhB*dZigp>nj8JdlGt6$J}!BsIh$o z<#0DU-1zksmn-wRl}LiT&s-U|!h5{ZpxgRw{EO~b4omSP@DbrO7!68EIF_eGAq za8NGyFKVkCWjt%ml%uqebp^P=XjZn$kbXH@jl{pg<&q`%YUQy8jk=%4=yfEfW-!&9Vx4$VSb zj#&Z5mRo8WaX+CQNXCi!wLE4RTMGsT%rkIo8!|~}Q76$6NbD2No!%euC4Z4(#pIFy>U*5HsYt@B z9M>-5d7^EHT30^#&X~cp!^rKdN5bvx*k5@x%&{ve> zJI*`j3$`)(oo=Sg35KkKq^y{qH5YDm8s?L^95Zn>wl z$uMW3i;!8u`aEr=FA{spt5@82oA&_g({RY_mV12vtNB4h@26YNX7D1M&9SonH}j!@`N6Sk zdc7p`8P^m+((l}J5Z^3~l1%D0i90959E`MWi|_hUmkQt%d>gyIBS=#RH9^W@NR^QK z8F97G%`k0{q0I6>_R6?`d!5Lm0rP;v=|q}ts0R|d*i$u%G0Cfm`yPoc@@+c_CF25S zwv&Ix7QVkjnk`8DEBUPtlNWMMufseRFrSc5{rervFf&K}oo~q}`3b(07zcO2ld!Mx ztrTwZcQI~KTj~lLy*njZpIq+Y#2S8n$%fy+`t#ip1X! zV=vQk38|}?LbM^i9n$NSavMdtWj-D-FD2k~ayZ??u$djhWBZwdDl54T_3 zQn+)F@G|Jk>G1goQ5lD#DT6|JPZk3=nUP`AT?KuK>mBW3MJWrzMk(NJAT!L?PgEX^__$_E7a^GtrN!KJxHL;Q1Rsb|{1%lq~1A`yw+ZV755+UPZne&@D*peG0Cf*ZTT$KYJhX z51~f2D&7}`9DAicl)}y46r0D7CC^ke8!30T7)X7%tvz%6e0+hLprx;}w-rerU*5%= zH_3X;0vddHLBM6<%alv+8e|t@ICL(~K{{USCtG_rlejwO`W z%78iTrx@>i((Xt~F`YRokL`>S(#H~4f}Te%uf&cu!pr+GVEVwzu=Xt^t=yM(FeW8Y7ty&@&a?Bb|Awh=~od7ZguM?I0ti!feES&MzepHR;1qYK*Wy^*wcqWh8Z zn9fCc9}>3>?L;nb`6Hfvyozc{*6;Q2-d~n?7jYBO)5sl9{qz@N z-yLg=34R_hYlJ72j_@97<(uXob5vfY{DZ$(?{c2~pfi!nYx$_ho6r5IS<5=P!|O^~ z`OTMWIVz9l7v;TA+*jy3la4?CZjyY`-8MnpHsWUcxPGvH7D+D)DGo4?IYjz5mH|=zYdtI;c4hX+CFG3 z68oI{qSWogjqb`E56L{NfcrHmhw1n)v>{+_aX3%1uCa=AA0wd)=UvZl$@3^b8B5e> z^0$k~|A3SK^$qFKcJ-x`KTi^Jf482v9Y}1LT-}Q--WV`XJNZ+(vCpMDw?`cH?|1k5 zqvSfoHAQ+2QNpHxnMZza)MCx7RZ{f1=;@@BswMX8dhMfBfR>z>u;ZlI^HK4Ih8CQb zaVQV&M-I0&yo*s+r2ds|mXLlEaYNAvq`nbefh{j#njZt^M~A0&jwX-V*)NJ?U1Wdk zgqzI$T)vmrJO=L*^rOSAVYzL)v!@vKL|x_IobXc@J%S`W#Zk(Ael^ctVM|zL%d223>n(6~T$@7rs`OQ1j>^5qaz7{TJ0wA# z>&p6t`oW{_*`xa+!`lMpQHQ&keANpw%pFL%Gc5O>9;{uWFq)52*(b1s%@o2M)YY3e z9Gd3Uugo0*^S$HWBYSe6gib`t-DtTziR*_3qS6)pOqzK*d+G2+!A_pH z{2KG`J>*${b|U2_k&fV-dR2yb1Fc1?Mc6uQ9_v>;XrE^LYGeJIX>v?sxa!}-dQonu zHd1a&Yi|?cEYhCC}&NtzEr&1~7e$2P>{!@jg|{N9`(wTMXCp zy7$~wzM)sk;g(sh+Ac>~2dM1#C&F;&J6ze1RS#x;hx?)B?$5rd%#0kf&f&_sh1lE3 z;Y#00*lT^EGJAt?t8R-uHweLNfm%A;)s}k>ah;Ha?i|0Re~Ev-FV7pR(I*xjlw(dz zz`fStes%!d@WDCe61ef@*5BcNcL07UgL|99)pEPV;eKzq`%`YgLvqX$4p+~McO0B&a09COfb zvHSJi@Ty$PJOODtPV$yqJKsTZ5P?)i>%1PdMGu`$t@_th_>?k-a$JBJVLrGtZ zo<_<&*>XQ7?t6*v&D=(g7xEbtBz?njhuf4XM0sY8ITNndGp!GYlUKPHS?>PShrGjc zOt%EwBOPx01K<|JElR-6cDVTmz>UCt3T`l3dyJ8NtJ41-<#0P#?*8ms&99bY7AD}< za=0B2fE$MU1>6zV-g@vFqZ1u2OC1t?7Z7(Tl5nk@U&K7&3a>uscrp5W_Q{aGpn8t^ z1+LbI0RNaflW#IQ+}@TO1=|0V@sP*8hfuE0bBo}b-(%d{KVlD9FxA}TaQj-W{Jzqi z*4}eoshk^u>vyWf?J2bZ{#1` z;{4-xG^*6{?s?S5OTdNU-st#Y7rgUoW8C$zc3fK;G*RN;d-;t43Hw?<59Z{UhvDk{ z`3QKqsHVfsv)pdP4MGy`;V5&HQFSWTv!SFFKbWJKKRDcxq@RH1Ai+ibV7Xrtw*{s2 zVJ{j-_x+Sc))ya=mVcUpqjSvZf5h(p0^~V`{Z92bDp!6xm*6{%xO0$%;q{YDm~x9B z=Wl9Z%t3HNHFHb>T=m0+7{>T9{vK4p&$>BDt$%xiGHt7D%v z>2IbyUf`%)ex_1_Z_xZ?GmCf$@*P8Yo}%+ytz%k$Lv?e^j}BMH`LS?cb-34CZWL($ zBlF41p3~&PJ!*T*-Z$XgRg_|q`^D^S;IVus5O*q)aKzBq?_{X0a@2k&{}{$MhkF+J z{BV0Z+|HIeh`4(!_a?^S%KC?rV{^>q4tFH^o`*GSnc) z%uB$1*Wo^S0NfI|UnKCuN{741a`(p%<#7LUxF5lbpl=-RTbBFlwv-C{2HgFv>IW^i zl7=~^_Mfr(tL3)A;VwQ9d*PnzaHYL{9ZWSpI@~3e>m9uRN_(s9e#jh`V{U*OvYCE? z_XpbHa5s48e3>`!jxv%E;#kZYL&UD*s2}9H(r(U|!yV_?+l=&=p)N?+&-T$zDd7+R*tzFZhdQSXmL`sf9Og+x|Zm; zDSZqfy@0qrDAm~?AxsHJeHl%|{O z9gvZ4l<;whl0N)K@_g?-0vX?Bzjx;KIp(^ZF>dcWdB%!{Amvt1iQ3?Mmbkaj=V&Dg z)?u6Uv7TQ?*>O#|rG0YDD7Zn-e6xY{$#?P21X8YjUgA5RxQ<9d7mj^#+Va{J=Oa>H zVz23!V`ezqUZfw0?nlZ!#o9ZMxTPq9K0;NfW4cZ_-P)VNCLHXbP_rx*Vy!cUgN=en~U)h?lU0W1@aR>qGG^ z%*)^gqpA582QgCYKbb_IUJMgXcxaBf z0ImkDPjks9=16@Cz^$x4g%j1MyK~H)aLH~m_&ZDBeT2S7YHtH;uNltz3pyShgXF%X zh{>ebTW0IOjx#0qp!LCHK+?q6g4Ds4i@6+3>RqNt>!t+Z)HmOrdVzz~YB; zOn?Xe?QMn(0a5Q6;l6m*(6-TgUQUdcs4+_(s^jPu1!t0D4LTYap zYj3U5e8&}CfjXifJ_sSPcTkWu7#LE2Wu6dvj`>mod#8|3%k7$d*c(bk=*a(BPlmC#B0Y|Zh6l$Z?s$)FFKLF7rGfKSJy4x zC2k|yfwm&I{|&*N1YWtp7jt-(K(5#eT7Mfn9P`7g`|v|v0zZV|{tQ?8$x_DMX7Ku> zwMgyVVEwS>5%#b@N;^X%(Ds$Aht%iWgQ$Kj?|HcPlSOlwedvJa0?P}!*CZm+#Gmv-&dET^l~yD``{8%S`*hEm20dz$ zk#cXa+)s(ygtnqi>3mb1dkcNvsGaRkw@mT+|KbJQ&pLkCMV_>A>CsRD~a2LBuqFYDZ1`EiM0o@QG;-IbH038 zj_Czg#t{jBkgv)_z7v6zJI6caYt@?H2q#{`vc{>=elBRQUAa2H%6~t{+zYq4m)Eo+ zUq|$Y!+pm)=Ue?a^B}Yf{fhGF>+3FITzi;rgf8~_jnnvl*{2j-#eNL9o)wIzllWd6 z`VgtTA6Rb66O0|GIXVGJ|Ec@DOYFQ&`_IhPIVLG>Z|=q9Q||Krfg6eAnon}LFY?Ac z$vzz3EcB|w{m^o^5_j;EyibI(Q9gdyb)r{prM4Y}D5xq_$bvOFW)NI0w}zzefG$I7 z@9&noG0bm#5I+{RKHmGC2mRh=*m8CMQ~77?dvN?98bX)6X#vh(GSSH#BHxK?<-5oHA~@YyLyy-yHWNFF>a3KP9bgu+JM%frsuOhguTKnh@0<} zADnB>$%ysWDKE6D>yIQ9bKHd=Hp9)QqfmRLzn)LAlpdCA9)zp* zP9**FXaQ30jn>}ZiEA)D!<>drLhgNGah$y+hv%BG!_6oCwxoe~TX`@NAj=bj^A9Ky21}T6lD>IqaaAy|-u6YTr`r$az=b^Kaau1Pz;NNb<4MY;|;W(|H zF-2F!>aKA0zNPSlT=SJ*MdOijYgz6~#4SP+ewb1DbFg42*Zl6-yOewxuh8Ei z<(^=<-RI&XRC^xt95m@`+8yr%<lvgskx@bAu&I+B7H}6DN-)4fl2VK zA@0!mp1lE%MfkQj&c73x*9td$My}}zSNqA@q?dAQ#!FEzJeq<;WCgp|9^a!ZJN5lLu*-2OqzOoG;jQn<4nduNgFU9Kdww-1XS9xN$ewD%UKBtMjE?(w~T0BjxV4+(E>RM$e*2C`_Bs_4%;v z&wiZk`8U)$*VJIbpnjN3`ZHeR`36$%@zEfbluCBMbq38Y+^w+eR&al25}#eBaUi63;GK;Hdw>sk1kTysLzz3n}V ze9C?O0PKyxErhG~o(pd%dfVZyx7?4GFeiB%-=hDc?Ofojn%@8aG&6O&(UeLu=ggjp zN@0{sh=xKc8gi+K=|)i+gxp7N-DHqkMT6uP4TZiaqeyfy`W7J?l_X(OD8m2!oV^}r z%`ry4|DW~xe%ik4to5GfJnO#p+WQ=UTF}<&eigR2^o14e%^i^;?|IzoNxuNKU6io5 zts6H_ew*hm(W7WG>Y!`n!>dVe`kk)aB=P-FdS!+jRWD(0oyD9VK{Jr~f!krHn(=`EVz={><8O`VAQ}%HyWJOP_IAIDfQ7z#^Zc9 zzU8~|&A2(EGvqCg+lln!(PU&?ZkxvKZMKwg58a7wM#I>b+r++{e*c-tPNr zET>OEv(V#cQeVc2H(Y&Sm8^E#JT>pG44DGARoo=*y_0@3`T-fYY4};B_IuoCK+TZ; zK9fFo$ni?q!Gddxv-gB{;r?WBcZTd!KjDXi$|8l%-+7Py)7VaE&3j9M!H@trvDwK>yza6HZMcgc-%uiV_t&J zM#k0m^Qj-k5;p_QM=v6O9h{fshn&|lWRC_3{~Grr@)`HS-QX4_akJl`o_X9_OiY`g z&d9g}9QO|5o<_4#6W-15j~8W0+}wp3GQ#6-AkQ{byDY)ga|P;$j>Pprm!cua_d~#8 zeyzXOj&tA2kRp#ef%NyHB4qa3cyaB^DKeM%H_%qr>uPThWp2$lRJVCWZW-M7JnmBR ztU_hTxZRw+dXDBd;;VnoyKR{F8F$Rv^lNVZVD=U+;(>_0{#TTRtrKu*uG;~lD#F%GUO+ZtM6%ji+oFwl^a_&aeM9j z>BoF#=TGgtVS19i!SW2*pMxScURXQ++~XeP`S-8-hk%Ds{C1IEnjvRl9 zo~SB&gCu+N;ojzPqwo$y=X%@>XKx8{pQ3WK328fC${ctM<9tcdc%1&<40+k(T0LvI zAz?4CQHlFk+r?kiv%Dlf$f^w4>~YNxXL{T{cca|0ll+hecb`UoF1LXmx1MM3UzJ-? zlD)YfW=J=WYxdsdaet4yA#u*RvU--2WN$Itt39su>(e~0wjn~K%6>h`-qMdUWZbQedmnML(Gs)(RTzgjfa|xn(lyLWJg)g+Gx?0Wb~pTxpJZ?L zrx`NL<661x@VI4;TUEK`CE1(1HbWlrxMuGu8x!`@Of~#*p1!hrR*+@EE~Lw@nNX76(z_l@1Kw=l`x>~$H^dhbMiFneF|xC{*%s%{ra_U3NL zkUk#Q`t^EWCj845X`EZtI3N7CVm-^*m?5`U!tLyFr|kweFNvGIDMMcLxaQxhJnrp| zTh(|`l;qz+xa&Qx`Qc`dJ7zcRElRStU~`7-)g)1F#(ml27VHMMEQy=HB}0z$xaNni zJ?>Sz;fLZRKa|29;c?C0pFHlU-LN;{Ax^*FDfu=-rg>bmx6P(RJ-c`}>@7*MH~20? zmU-MG;GKwud)#4;JDa#tv<`iQwEwmJlJq7O_aD-?X2`D|x87H*v(S;q`pHWicLZ@0 z&~!8f`TcKRk{{ByWk@gYzSW1!YZsDd5i;%-ZhY(Z8Nau~=Nj~V6IJaa=E6OApF}-d z4)0S`=5g7|(-7HlRSikm%zJy##Gw_rrAhwH-p+n1+#sHnbu@W)qPklW_7*vAN;&r; zQG>1QN1@`4+*@I>pZ8s=G~^JK)(73M*7-(J?Wmm4G-0pqpFQ|J`;W-{@K{Rxk50|h9?Ec}tL)n&!`$pxBqcSuj zVQ;q`H6;%X{5ipOuY-&X`i=XA=pHm4={`X=i-f|wROuq&dfwInxd%sOqQ~t~t(KgN zW+U^%DQ>@HWJ)c05xs-nKw8gMGf#b>V>Nm6@{0DB9vYQbJg&9lj;RUm*^XOPzh0D7 ze@hOJ%BD)V`5rfSH@Ia<+~N*VX&g=X*X*6{aoziJs@#7FR#ohGijIo%n~@2w^*gIQ z?qSZ}s>X}5q;?@kM`gIjHG4OD+{1Un4|z%UmUoQGWRKf4t(LS#9g)?CgB|x`;>MuK z=q}{9i=rfM@v%{P%j23KW|PmjEqB8Y?-Z4Ck86H-!{fHv4L_76*(=9IB^pbV zoB81fj~ncUAA%1m)`ybgqteymwy0i9+M~Y6{J=CoL*#Md7NC`A3G&B_yrl7>xJy(n z_qgVVYBdsEy3shds`i$bWN*=lQ7Q7c=7+;PF57l-Zk6>RiJR9oDoZ@B)rUbImtiW- zt*TuJFDtAt&cprD)A=`%&ph-29WH(mXR!AI!gn9+%tfac))qElc)8 z_o$riam^2JdfZ>!cu|!ff+TxO;9g$|cZJ7wj|o({ZplvK20fzkw8yo1NtwrW$L1qd z%}at0E4GW^)Tn&qam^2nBZ>OU>&)W*t;!EMN%j`NO{1YE?;Ex7xV+ZtZ_6z&$=>|j zs2l<}h-Ve)4et|lR?URH47YLapxU)$8d{8ALp``Zl*i^}t9MeQY~7xYOW0hn9IFbZ^FM;Zs(ED`m=8xx2kf>O)9ryxa&RcFnHIY$Q}tU)4I65hwfQR9z@Te z$B_1ubJ(ZteP=b%wd10s`j9>}Dm^RVPV%_F zI(w_)mL+lJ;;7u_am~MrJ?{6r;orO@|K`D+R|)q6kGpL*xcN!k61YElT zNB`pWEdHj>BO%drvd@poEop3+Z`kiujZpGGF>DNW&c8@!NJeQ&eka6vO6kiaR(x{d+M0Joq z?<-2OH~ogF%=ftK&P-ipT2-aKxjCinoAA+z@b$GyC1E!loOyKDgqz znSNhXj@U0zZZ;03W+b>jJ8o6ud_mH9kq`GWk8AVw`X2We$E|9S3V$y$z zK0(Iq>$oQbwd6FU;aslz{96~K^}xa({rVG8IUH^fPtE%ZNIxFkiHvLS1NxM>?MTC~ zT=iU#-A6a>AN|f#QR(N|YyD1*tOWN_XKz*gPGM62Q2Y$zkjL!~?;Lc4$K^HP8Y1r# zm(q-J2-QK_KTP19SP?h8()(5HcXFO%{@`&tk-jVHj?51)Iqp#6u0tBWnUNOfmQZ%N zox`@KzA*ph&x%S@HgK%mZYJL=)vC$e9`{wpEhcV>9(TFp{*(9M9@p%>)Z@PSzt}rFDj!tBz1HK-{U5l6aH9t#+VQRMCZQ)h z?#GV%C2>EaJzLb0RHXBV4eW#WKDwF&*C*zwdY>Wt#i*PNH$Cn(k^M-29vX|RK5TK^ z!}qHt{m_-@5;Uj}zx}X|`|~_MH+^K-pV@iv!vEk0xYjOSB)zt`1ze5WwnqGqMm7?+ z6=~R5!n4>{7mhsgd3ZU@J0M4W}>-^H+Y@yC4qRpxzg&0g!z z_V>759Ji|etSqTN%bpjNJ|5ThQ!nEMH45T)jAA;ATGS1`L{NW{!d$QwJ zHGe2e9^c+z-R=2zGrZqXoq1MbUz&Z1(K6 zb?h|qS$)WJ+<&r;m4#8+t8Jn_bfCa>9s7*OwfDzaUC`^_tz*;S9t+p%gVo>XJ+3|X z`cLX_&YRpP^Xy#(Pwzu+@VM7FKkOjR!bZ;J>UuNzT)fra;4Rhxp1oFoYqd)Fcbwx^ zReuYT>RH*MsLb`aR?klLxOX^iRrM_RxZ*suba7P5JU@(vHx51MaVI%$332bC576m{ zrO0Bho|QU#%@0NIunssdQEp$8=NI%lvU>Ko<381z`-Mou0E)|6BHPRIc*u?N7du=qhB~-i~{h zV=8w!>+Gu9arq}vdCar-QS$0M>M4&q!g2MzY0o>hu4Ss)@5@;ev1o4 zT(#4}-}RHZaJPEg`SA8gsV1vD?rn~H*oaj5#o4?1*BZQo@6YeGdW*9Fx{p})X;f-+ zBbOqGhq?!G9}*pb%)c*&Q{Kp>#9fCpjOW@Vn|Vpn^KUcix!US9od;vZ zedKuv8JDK2A)@2ZtuLj>6MUxoslAyb>HBiD{g$}(<~#1Wq@XM+y*zuLBhPziqsKkX zaXTDbOU^_?(Cr+SpU!g~bC7b2T)TLOzZLhxI`(Nj?v3ObgYH6R@5PS$EOD=)x6ot1 zq*kcEZJUSv?;r0sSRa+C9(Or;K13Ulaj$UPM%k<}Q5SR!nsfs1i*Wr}uA5I9xA=>w z%!W%5#KUQ%ABOUgac^9rvVBN_+Y>uDhP%Ts-T-(o)s~?zyk({H;ZrzC0@F*@<>>7LcT_?9mWuPTb+>7^L$OzhBpvX={77c}c;qQ8~xsb|ueg z=nQ1s^BniNC#uT;;x$~uZ+7YUmOKs_H}AKo+zOW>h=+^GcMZDL<6ank6nTWWIp|&V z2HHWH+3%dEJA3v0XCtWFWxq$|d5_y=C%+v=p4G_g&3D`l#QlIY{Kj<`{klCLt3J@6 zTwQOLN=&}+xYaw*rcixk+{+xd4RMDf4JUA&M4vM~X&$BTCupK=F(oFM?GokIgY4dM-U2 zFY>C#AKH%@234;X71k4*hg)1LCSSqD{CF5j`jKc9 zGVTT8N0CXyO+z!##Gyob&j}Yh|Jv_A71oYP`}PTY=aFYADn-Wa;<%fM`vJwn(aeAK zUQ6?eelUB>>%?UEzrfAigK~r0JG6qAzO&{=`y#Vt!oMZ;VzTsKus5h5li<+*d;bMEo~f=ZXCQE+Ur##&^4 z814LUF>yDb`_Y|fq*ot`oF8_s6X~SwxM@r-f}7{K|5xooGGg+de}S79jq$8{qJMaf z@3#;wL*|DEoF9%mhIWC5qell+mkrFLnye4ov7U9%K27@v9fyLfm^6p`f9;2YW-&Sc zU*MKCkI59cfh)I%`HtEyp5SVJc-i@3A#p2_23=+A+vWsXK!EP#-c)0 z@Il4rLCV~CVD<)w$7CGb9GI{1cOEBC30jSedyW1<>Tj>(S)ZYM(0DYG`C@Ka=-(3d z2^W*-^fyKH_tSEZipfm49W*E5InvKVZy@7-9DWq}FLCS8S19}V@&EFnP_JRlKvP} zfXvQ0&d&N>Sm&WZ=v?Hthp9)0tt6BEZd-x#tNm_U{_!!n4X*VI`u=Fu zI%M`bmsPM=Klj(Y=3nay|7!2-zEdxg#qkJ6cT!Ebq8V)6-RFpYaRd8VT$ zk#V;;ZrX{AhiE?(LjmKW{oY8C$9;$InsZ`Ix==^VzlXDFw~+eh+=7#0a{Ir)4Z6nU8Myzy{w+@OuXKw^+m4BLd>Stk zI+QXw8u|WhOuIOb&xWF()A%hdKJ?phiQ`&3&WC$q5_cfwasm1)u9jOe*V=_~%i#9^ z7r43IX~+Kpw*>BE{{lC=M@-&M;%d2d`G@6JnpAE@a5pA#|1Zid{gjy0J0?-jx?|@V z=$uOIy@0qI(QRnNRVfwfSy__3C2-q#+&jo~A9@HS*R!XIdmSx8hwrL5FDp;t=A9ao zlRfS-@~lN0Q8ITh2LmTni^z8#w;|`7weHIO?n)3g3^P z%)N;|te$(h*8^F+Zy9Ft`k@*fqjC%3J^`1ZDjrf#swL@YAC$~(PTWE0Fw~z9%RFw6 z=HWhw?(-`*H#a7q!L|PPSn^!Igx@4W$=s9p?DwO?e)5WTeCq8pSv$7(_m;v2(kmvjJbUMpe&_I1S?h5- z#z`Cj>Pml)dZMn#f6hE`_S(Fp{EV3V=-E4f^!K4hk@`XHz1?vKp3Hn2%|u10;W@l7 z8>y}7(eVDv5XMujXT^PEaz>{_|L{BMYjm-1kKbR^^_9NgILFy*_W|?!$K+Md4@Z)xE9!}i zd!gfAPh25-5G_Bwq91aT{7^7}ejTp0iz4#e0OwV%#=YHfzaXw0{e&((ykh^G=eSmX ziwDN!pqzvsej`t#Q<&Ew<8E}^p~Q_sFSjzc53PRqmBJ_)9FtyfE!LO-DQ?R~qf zUD@|7gFDIlzF)#?c`Exf$hg0``m~g|NKe)gs1+K+x^F5{X1?>I)&HVl%*)}{3~iF% zic`cNclG-!wlVhVJgo8ap;Jrvqm|if^QH9RG1=*Hb)TgJwsb;j?*h1V(;DKx6WW9L zUP$*@{Po`iSO4`LU)o*^;5I!jQSRrGrvN>lo8U66Xo&3Bi+Kr}fgVO@9K!tZ-4rS0 z#c9*u4xdNRa|X&SDu_u>xK?f_omoo;pzD#@dtn?aky*sOg5E%Sp2hb=_IY9dV&xW$ zj>#n+cMW-dM<L8WLh>iq?4oXIIta1ZaxZXT{zjhi)!NB z=uBkxe(ktd5jPs$f!0h7&l%YMSV3~T8XuGCaIHS*`#c^X-xOqA+aJ^Sc{CUo)(3r` zhhJ`1f34ig;m-B!eGH!duIWo2x3;UltIkW6rNnF46|LC67}xwzG9f1KdfZjy+l7wl zld!j`_4LRZJE?DTh^is1J4xX*6o-7qkJMEdoWOViYlGM~?^ zZpoCG3?iSl(++TcCw;?y%p;KbuUR~sL?#gT1bPjsU&l31$aKCZ(GWC~j@M@mNIYeak z_H_0RCawV8i|#~fulxrFKT3YHH@zq(zgJ@KO!Aq%Oiin_x405}3*csSPL%iD z?$zXt*J{X1aQQm%av2)hoANGkpCcXr@?2-0^9bXzmw(x)6xl?YpOD$b*U*r*$N8LL z{4^o~SKp5_>GO-=bnrL_lIBo!G}8RLs2`4#a-x?%_tBViC!h7>9dAmJ3;FyKr2f`? zkLX`#e7L3>zr9H07gwf8kAjNxjS}Y@^LG*45gzv%cz2?y9`|v_T{<~c77)J-S^uT) z%`9JnpWOV@xcQIK&%m|$+&c1nhkit6Z{WB(50B(A-$UQEuV`$Cn86yXrA{`jgq4LX0(^+v&y zF?q$auLJ3iLp_k$m!}n%`gQ?v*P`2yZ(ljzGw-RGlzN=INHY~pL&nL8VrF zD@vrspjr||?NKY_KW|pIML5o>y}Tkp3TDS-JzOjA4%l`gc}_;gJ;rfEv%`P?Y94I$ zx)83Mn6UR0c)ig%9=DU@4kYdpq`{uE(((#xQ26gF%v*H+mpv!Oo^XPD1^L$8#C=hZ z+u3m&m8Z%rj;qh9R8~KYTL!ly++;tD_qg`n;kcphpTF}%Ioz{6d+&s&>-UE}?y1h+ znZ&(}G$i*M<_oj8_{Er9>T%5v^F8jF|AQa$UW&;CxHhlU=SHvUSzQ)*+~(nTkH0@= zSZVlw^c=50HkiHId1<+oz?}`(`upRWvEG9Fj>kQ~aqrlbD(CaLh6}h>R&LfW2QSCu z1GqsvHRt2tZ9-c-?#S@7$RmSU2O|w1a;YZ43FE1^o3{=GVTb+?KOlw1EgUJSDm*9 z$m)V#jhp{gOy2OgkCFacbjZ*Icdg^Te-VAh#XJjzh9jMCr5zY@i|O@!Zq9Q0v#tq0 zyi5A8QSD&~uKk|UXyV>SU!(P?jEY@&P{_?azv6QrCGRtDKRLl|IGiyWU4qOHchXN2 zBDF7}pF>xmi;;4tI_{_~6}kDVVlvs|&L@4JVLh+0w$eOovD{WBXc z(p`P{!@nPJZUC;ew}#|77_~>nL~;hWa&z-5`yI{PH8E-3O-JxB)a0DO2J(G_ zG*;&wzlGeizJqGS=c7Xkt5@iA+A_8k!^wq{ZJmpm?t5RxWRX{KOX*t9cfd7&t|0yAs2mx$rQ;rO8RtIG5R`}3wBVe)EB_qF zwf?L8^O!vEeebJCUx02#`gQetyPxw-d4#xPWc{pu#j|=E$ zU1j~Bm2>*~n51-1)c4wNSKMzhF6*qw-h9>!X(+)NRrMzYaGSxk`tvg1=S}oqWPa3U zuiQ<|;dkj5|*3)Qxd-Id*E!-HBX&%@7Fs~ABRemT);^u!D zlX;#W9y~u)R>1tI67F@UrOJQ!{9e{=dSA-7x6pB|pOQ_SL-M!_+1D(C`DG>C2KUyG zZ~43@zpawYElT1R!rh}s*dVoDehY7>GDn8}$tkp}aCnilEaL8FlG~1}wJq%fdeEQb zeif6mJx(XmoQO_AN>|Qs{R0e1gNZ9ZwzKR%w@H{+t~kD$bR^FS$hdZ{qBn6DqN`Cp z^2=qITNl{;J$noLb#VF9@o;a;G}|bkGECa3Xx3KP2Z94 zM~~Y*{48?lm7%SjxSoD?nsoK(aQMEQ31 z{XFikdx5wGDADx1|SEjJ6+E+s8fUV&@%>}T>Fa~<<}r1mQJ%s5FR-w?MGMXqP> z5c&O6zN=@}KNUwZWxt*YKQtr#U{rvNJIQft-@se|#n5&Z6+XAfacw@9TPIU`d)y<* z(-rkX${m5d4?6DiQ)VnaPnLE ztRi1I`U#m`ZCpL=c_Zr@GzcBkx#Ic6nXccoexRgYrfl`>8bO{b(alJ4S729&m6XZE znN9iiGbIBXt=){?m*)}>OOp>tqxNXN@oG5z$FYSG`I`9e(c`@;ma7($)$)DLwzY^^?4Kdh?GIF*qp_rbOP?iJF% zjowAZZS1)E9NPxsHGIR>Z|7ESezLtHDQ;SjEs zaf{gJ)^iE@u}o=tX2Ra1NPiO@{|Q{pUY#NlB4-je2x(YATxDEak7Z}lFToAssp;p* zr}g)Gk2}tBTilr@k2tPAr&1aBY}%F1$4Ua~nP=~%oMTl#JneC(IPU%R(`13;7Cu?= z{h!JHEy~K2Ri3>|$oDb&%;P@kxSP@<()E_m52tah%w9dmrsY=F9RHq`=pXu!ek8gB znSbAN+@`nk4tvxN^)IiuZYoaN-znTbQ+|MJ;T-Y|K_igHYQHl#{4nhf;vPj$A?vsF z`?sa;7Pick?q?_F`7e;>Kj<}N_I(g?B8!N7A89y&HeX8n(tYo)%M}I#cGrwf40R-ft-7wuHS*`QqGa?6aRpyoSB_Z~wbn`P*}% zIc+j!y~jO;eDw-bz%OIxMMu7%_sf|w+wDy zk2@LO^XN^Fd#%Yw2qJuJJJ6s#jI+Le28iA}nJnFcw61M_v zMr)CtBj^x>bs^97w>It+XJ<-g-vqYM}0D{oHZqPvD;H9o(Bn?a_p1INv~{QF}{PR9qJn zok0I}PQt&NNWTN^Ju$(x`zr;+O+mBJ6Ug^NdS1o--mtDZJ$c^DRY5pxaIaB6)+|$T&33?eB_j1RbUC8?{(U0g`l#73NIsZ1{ zOpp4*xOusm(z;*5-m~vwzJP8+#(lJnl)P&qV`~ zaZ4O`EO9f?Ec89^rBXlGJS02G59McM$~|y{u*_uvc|Js+Amg&Vts#s%}2>O7e(dLN%Mo%hw^hXWuoelq)f`ZM+njXRM!b-nLX#m(MOEUc>&^RBo?wb-qvnw-?;5VP3hK zd^6C89ygnGgvi+YIMaikK+}*vk1lof_rBF4`aa0)fpm0viTbdb^mXoMZxETiI~-RY zU@ro7M28~3{^nK-+gmT@k2)VKyD(E`!41?R!nLG79e>Z~YTQqmBM>4T9;BZ`8hUW8 zY<+L@^RkOF*%omV=Xb$I_oxz**o zrapVsoczX};J7Q_PnGtLtItVQ_Fk+bgSZvEbAf!FN2bd49=EW5{KLqv z#MPP-+FPG%RemVAG*hO)wRUXWy~$_xe&M+H6Zdyq^F!gtO!?UJ!(-%CKfK^^JJD|u zBE4Tpkp;wS_&YyjUy&(I1|~dc4i{AdUJ@^K}k0M%r|( zE@u$0LHFxR_DP(JG<%x$b?XqgkHfY8YK}zW`xyN^Za2pr+9Fj3JFcEX);V+XbJ@24 zls$^|u4nHhepu$X6Io|&C;k_t?=$lEg@?KM;wN-`+Wt%7b{Lf4)}F>a0CYGqE{_^( zh>R+tk3z4a*{Icj7!Q#8w#?0wtbS&X%ar{GC;a;}>1#d0z8Es@p_r!JPQ-OX=b~Q7 zuSXM{y*3`?j?a{{;97gityN8ik|!S-x2NOYM%+Va7Mg+lcAEWQV*FA++UGH$cDdnIzrv;2m-23pU&&|OOPwG7Nk3Thc&5yTYvZ`~gD277UH@c)yToz-Nk5n~BU3hc zTDdXaB`-7sj=DQW@Oq;aWd=4t5Mj zBaxMxb`Ond$JY}#2Hl19IZ6LqdX8&v){YCG$&}|j?gQj`8a&S&P0y%6*8w`?mNe5}Eci=TFd^=ryzjpRMc^`nTbuVY{&U zTQ)0GJ{p?v?_tlden1x@<92cVrzpwe}ax|p8{1x*E)z9Rk9g*(u zrtce;V>+Fy*6Azfh}=#3yjK~Yhb6e%NMCandqv2&=Q!?x#C1WZAbpxmo4 zb2}R6af{+ui4+m{44Q+^Z;-}sN+j;%7dUROPnla#u}bH$PXsqN%qy>xXE|Dd%-&}m zw_P#)5Xwbe(V9l#97z4J#g*G#>qSnbA4=zk1rE3*abNs)(%*+3MaH%J69+s`dq!uW zlTkTsZ+o^IKU_P}=V;4m9)%0Jp90tF*&XaX=>5WJ%AV=HhqtZ~B{=QVQ4lH=75}@VMOR*AUsAd7$-g1#s(*NciDdc!wR$?*(zy zcCGE=&^T6V%=#u(UMGHKTh?Ppy#3hmPlkyK-s1i)Jp2ATUyR7Rygh!R&1tevI5!@M#-QC=%azaDKyW-m=$L*&dC zQl;)}-tB?xeuwSjv~_;4@+yU!=W%1?X@fc-<9_Y9cixdEy@@{`+5WNBukxh*p`v#) z`3P-6#-%IN5b69Pb9OWg4My6J z>HB_ap1?UKcRt|{ZYlePaBco~59t@7kCAcvJMQq8=-<#2Xc}rrT^RNidmwNNU4Pz} zwQk%G|K)x%dj?j|R+9b**8iV#HSV2`yG`R>4(ss5q11hMKUCX;1*>PpD_I}HwSJ)v z`SwSLAmdJT+NU!u`U$)$C_op5WFY{SlHXEx8(Zt>d0V+#uxsZU+bCf zXUWHz(g$vEYuJz;hk$whx8W}eyPLjwj;_AP~eHqjY>F=vIEat2f?V#BC z^N)G@M)up_T01z2^mn)C{ufu{+W7Err!={Ycnv3TC`ji^uSLWDR=>XnYX`E4{`QK5 zz1Nek_o6hp+vARO_Rb{kRkRE(LM@u{{QChsr?iCo9b_#uZHuQsa&fX ze~Z3mzs=(&uLphZKUojvY{`^Dk9+rj!*hhZvbI9Gsl6Gy(j;YG=-)0ZYW;bx`Pb@0 z5!_`Sw+{IlqD-XqVK(JP71aF-7lBI9x^D$ZR&T&?+>-$!~r(6=}EEF6Ee^60!ND9@BTu1awCBTqYY3^HzS$1R#$ zQ+g5K7qutO=MGDnhZMto3vLkRmBHk>0^Nd)+uw2bdY!h2hNHo#3ms;9PUwg9n-cQ{ z^+WFVO!>yM_d3$wz(jN+SL04`+!u&ji!|`~NPN8=(5G3M>DByD@I$8TF)Cs2Ch~n< zpL;1K3GU~P+n%^CNW*Dd3*nv*m#-5KgWz3-Zu7WoDQbw!C2k>l7bU;XA;(@Ub zPor5pm(5(n{2O8=a>`cDob$PcFS%CLj?3Xb;n~}t^|}APBW)MT)qe6KKD#tq{~G5< z_-{MkVLs67g-yfyoBuQCkv#4-Rd8?Svx%rG?jQS|aO+;3D7Ow|8n!D{%K4tYAJpDB zkbiFF1C~Y=%B|G(lhgQ2+nfB7DM$VvaLbbV-*mV`;RgCQ2*=>Rzw+;ie0G}iZ*Q)Z zmD@7*kksBnxc7V9epPTU;4>??zv5cENZ-Y|PPkTo@2P@2mCydpUaJpTL8oi9liSNjm&}vdO$)~NBP{MD+vG%!~G@qkQNb99?Zir(fQg;F8 zF3|CxMWh*5eLihai?IJIclF1{t@M;Y(m4BW+%}}o=G~Mhay72*I}sum5O*oka06GJ zAGLz5^CN4cw*Qr%8pz>rt(R4EVnKm(9*mpbk@#2x=8;|m&$COsh%w6DfHIIBrs@_1M`kQpBLX3|eU zPaxwq(T)l~d_>$Pv=ilTOOr#p;a|5-D02O(_1F1(1QNL};opXDF(*L#BjdJm+>XTM z^L$lzuF21JPjKU``8R#fK<qvZ?yuO>yeMXKX^1A|o9O-dqk$yh<5Sf23a@^XB7}HQ26rfu??o8+3 zGPv7mn1%HNxeTtgi{yE&*Y5w(yf(K%AkV?I`gZ;5C2+rmn?qg=UHE<_j8OLeKeJc4r}5djNQ2(z@ZVQx<#yr1u-x+Z3Z&%?3GP7h z<)f=mvb_bwjYoH*M|o$rUmwz4J2q~1!$A6b+y}|?40-`2bLSDa7=4W1M@}WvI?%%2R--e z&yUQ%xHO{Yehc7f7|XS?an{zoL6bo0az4Sr-K3w2>Mjn)hkaNFgw>Z@p6_7iyX>E% zZ02RTEhxM1c((7&RDRTLj4JY%k$-9>`M>w_N2|!6-8zusr2O;Vsv$pm`5%e9 zuk!my1>}D#DgV#p{msjNXjsgZ-iInB|HoeblqHM-s2;NNo9XO7hPbK7pJ$XDNc;El z&*1Y}Xb!Uc{G3y~{68ixYiY&&a!4R41&MmthR+W}MlNw?FsDKn@^(5U24D7e3eZ-2+_pYis?zC;Tk((#}+w zL%fF9xjxW3an3;7u7-#93Fiw1M+I^m+*V;;Sxov*(FSDP=fck-doE{eMu(w;P?z4^ z14GJN>c$7%r~8%k=^X>P2X0Nr?Lzts`0t)vjaw3a7Jm=mImBxi#I>q^ya4VjxY=P| z8BY2~H1B%}u3IWbjxS|?f;OQwXz3dIstZ%3!;}=6P&0fkU!TVxrTyZGft++xVm@{+ zzfXDb`#f*Yc_g#&6ne1K^r}N5IrE=LW9EWy+-?^4(Xd zGLd)xTAt^ZF?9;<%3-ot99(V^1V`6&GK$?zD*mFo$xc~4m9BYrDHKbg-im{q_ zzdh!YzoVCbE1y5CsaGcQU+eOx6SptY{L>q1B5cSF^GEhxkt!Dve>u|U|NQ#_TU@)f^+V2Sfz0#l z>HuHo#5a&vgO<}~XP19YAeVeA$cKyLVKnL09}jXhd#Dl`BELPKD$h9H9IlnM_dn+G zrw6jpai`X$KO_B4)O}Tg+skp6d_Z0OkogrFhw_i8Ca+az>|dNBIk%_EPVClsh4yFV z=LItH=7hbnny~~mMP~2Cj(Z+)BhU?Kek4WGy0cG@wBO5d^DDDAKQG`}r$oK7eXX(N zwR*+T3k@lE5NCB$eW{_8{EIxhrjq6v^b#_=_-RZHkqyMv`G|EU%0PqYi$}WtxNgg^ z9ojr9dte}uTN3^8!KCkm&O^ri*>P787x|d^9%_qL)2`QiTTM!7gXQkrm#yDqNFW2@ z262<<-$~yW^+(2S6}lrbzD7hE9-JztT$uK!=lXMBP-L8uWK+T2MBJy7JL7(qQ+y(k}|+QO^(hep(%umV4agjysAM<*adBeI8WLCH>WN zn+0&+^z8kDeBYy;9(SeV>hJ3|`hTCKi)`hnw{M(v5 zC!hhy{9Eq0L)n+Oi}=UTRJ4IU-qwYAuAjH_PWi(F$%1R+^IX!ujXpreP1XOPp8ZH% zwKdFDP&#VCK6G1bR=>7&^~~x+!6kuofNT9gE7IqpX~?*FajZnzea1Qny@M8_uCFnE za8r8|^(^?d;%_r_Aq+uf0GukEWRhu=Kz3JBm@-keC zHXiOJ{WEACGVUpq10nKM8Snl@pP>&>?`PTjM9RxJCbU=kQGE`h?8ZQrdffd!XRe4k zBjYZ0+|lz=q#yA^(Bh%t`?>=RQf|Jh4{x~fxuhVFuRK5KZ@-Kn-(?>6L&uF_Z8yaLF>@Rs3Ucu06z@r#NLs!cds)- zZqChtWWxQ!zyC+`s$#fZJnj^}!dKX`1(|=-Tz{rtPeYw`oS#MhcvI;4SN)KGOCaaL zwfd{?-ETmi#>lt><20%Az57LuYvY348ptHLL6}!E_D+-iVIJUd>$rA&FrOWTG+fY( zHrya_9?Qy2=kIE79^3_RZGPDfUSD*Z$8G4iHPVLuWn#j1^e$)`Tow%jwEA%-k!neiFza@^_ zrhVv#{CfiVq!K^K#)Kcn`~yF@?JC+ zr0dVhoxNwWo|}gsO79PNW;1LTJo+phNPiOQjg0%M<32~+Cuj#MMFDaiK$c_3%UC{ST@4Jcfo=9W0KFs4kv_727XXl~8$nVRFU0bs{QHXtoQv#U{ zH*j2QZ->A&d$&4!cc;D8cjZdIUXy{CRhYvswKR zzmNS8`wL?eo~0u1|~4cX}{C^3PS;x;CeX{ugfGxP8f&kFG$*Jil5yv)so5x!3bU2l(maYlw_{nd7=Y9lz)r#qZaxE|?$k;J)bD+XP-q z)W+j#Sr8(9h#Q174Ci_%eQol*Pwl;+8tr>}ARofD@nR(D)xS4$HGB1}FCp?6ajzf^ zyDq93KS!3#wfc}ggL&V$M15FDz7^;LWZZdawaI-wFo`==` zcbCU~&vAF>JnSFOMLiM732?2RrG3r0Otd#Ldp9_4bK(v{8tUI)LwdKs=Kj<(<_}hG zXC3AG$tMFD>~SM|rN#Ylw8xd|aa5$&=c&@!aoc9GW`R4sb;bK{XH%bbd<&imSx_tFo(*IIJgfJ$ws7VjMUmC}!$Mx9KXD_`SfuX{>C!au{9?Id{V}a{;nUC;y^xvOBWmu55ao@E%mrV9rw=fQ{^|})4yRZ zh0^HXp2+^Go9EkpU;g|+j)AN7 zOT%i?YyI8H)wru2_n57$rO{AyK6;J5HFtm3Om8#xI{$9{Kx8KwgEs=13^zxigaXpv ziS9wh-8bwUBGZU_8flo#br5~D)id+&kqdcWFBwbTV7=pUUm^Vm=o4hz7LMEgd)Cxw zG`a%i(_S}ZhyGpa>{agj)Sv9P0y$`6!Vfb@KMyTH#_j328;Sb~X-L}!7kh&y71ulF zhr+i5=?T}ynLS9KfjS`LUhTNO3V0_s@zC6Y;gSV>9yaw-!3C&Bt@!r9Qw7^ve+rAboq(0U7so$Gw}lS*Qff zMLsvUIULJ%o}~9z3YM~tf*WS#xEJZQy`}w-;4XFC#>BNmN1|-x`#}%SYMrooUG}m- zR>HOV+lBP{yXd{Rn!RfrcQ|pQkcOMMUROyy({f8?Tq;@~$X2)%Ydnl2z4mAKa5e5` z$DK{wyXZ6YU*w+1{t@HoKb@z{PU7am?FUzR8pgtV3BB!cZ*<)3owOHp5gLFB=qF1V zGv+cM57s8?nfB{Bp9XTD$DKg>C(*OW{IJw<5BQmPE}>h|)u=;r=Hp1~ZIAoH{w#kj zeKP({Ul;HIXTsijq+fy7Ami@A{E!f-`3q|cbTDd$7PscM8!E%z9IrmCrtE?Ztam)_ z;iNwS^+d)!)^RT+?lN>W()F*_Gac*I-dvBX^{n8FKsv*sdikJ(Dj3xyZ9OeOJhhay9OGj@yQ~j!44^T(y4~gLFJrn>AQ_E7%msop6)u z@2MX50>{<&VE$G8%}c6h5p1s>i-}pEBLuBK@L&HXi3~4dm!aVY|qq zvOPtfuh2ffC%9KR?qcHhk=k-9Iv$PW{?CvHs>$LZoO|S1oa(eQo0rJXjGOl+#^e91 z^BMVYhyDxPGPqO!1#a#yfxP!Ga7*B(+(!d(;daWS5KToItLy8hDL1;rnlp*h@CH|Z ztzG2mZrJniTU}(#*%e58xYmxBl4ldzzgmKOYgiv@d`8^fDYc~)D(lZ!&YrQpyRy7m zcyCRgl`fJj>2ZIe-@Ssgx1xE->h1TgeqEkgTW&`$q32LJ%shC(o&2`-H{tkW`!>bZ zvt%q>s%$*eOQSv>#rVzDxH}wI&o8zmUc59@q9)x09c4Ctlwthuj*rzpA|Cebp7rT{M)#yUycT|1UKX zb{@QEJn0|fp#Hz3kv|`>wqyON)XkEID&eMi+(Z5cZaUnOO1L#WZu`IBR`#AXvp0bI z1>7cluK?fFg%?Bnd)#{+_juy^qs!4nsGR+!0y64-q7v%-EiiPS;!3Usdt^!V2NU(l z?h{=@KC4gn!`1%jukI6-CiPExpD5UqdI~pQzp!@`ez~^;fUz-`u^jq}bV; z&HsN6-aF`hWcI$}?EQhbpk{4phgu`O@50ZLN(T-ws~BuXUD)bKP>qhKV(pYz|i}$snplPeX?Y*yP>OJ z!2HmZyyk}={s%vl!@bwznjd0t%@14u#t%7__@S_Amb?X5+nb(~X$$W-bTTqOG^EWF zBIgq~3XMZIA#HDSs1sU$b77CuR4}yumhPJ+Yb&w$F7lbZjo~(~qJPM%#NPajEJ>RZ z+S`@CH4WZ#=p|(KW;%O6AnpfLy>@N+jjP%_llrUnPN3ZG^Xx5)W=SSo3+CT+@|wL_ ze`9ZcCH5A?vg8D~+W%^Z!8;Q5LuT)z`^GX)FA~2DEkOEytX)Xk!vk)9Waokl zf-D*0{GhozAl;|f;c+7wAvgZ7`{!t`m5p!q+-D)&(Qs`Yy$hb+hl|un*gMN{^*-EQ z#B0#s38{>0>xJ~JEScf?p(*+F9M66p_hrX@vQ7B=Qp)8vf2GgW+V6rD!CeA3(8Pp8 z9Jiy#o$t8%zRIq|YtZ*qR`$GZD)m|GZ+5dR+3ESWC;6H@Q&Z0OxH}y8L*g3Otu05R z_NZ_sdnnI_IcjR6KJ+5J?vD=OYW1N`G;U&~fVlgShWI(kKAfXO z{{1+s4|%P!WB^eL%-&hf-b;wP9^HX%L%RQ=@B7es?^0JE?0!#Sn=CmEt_A&l z7v0Bvn0!`l#UVH4Y2sc*Hg3lAw9S&+$*=wQ4zz&J-$$P#vunPytInRx`_LgMa#PLt z^RVd$g#BR=-+7z@UH=^%m?f`zc9|b@$ZLLl18!CQcX1_tEP?y8vseGzsql1N*N3aw zyV%(~l(+(+L7l}Gz^)&2Rd$c{n~ONx(`i6Mfk3Whk~<*^2}h1aL?gozWA@$mwqJcc#r!j z>1QqBd71_Z?r_H)NZcg!6naEIKY(&VIu9tN%++pN*W`B0lCR)q$GyOFKcv@lGwU@U zaavAO9QPOEYU~x3e|@g}H0hu2lMa0{98U_5$&$T_5`Jh(dcB|7f~#@)$#M;mvxvI_ zO+dGxF${ujUA-zxvbVfbmK+T?C!UpYg7ovzB4k{BzZ)U4g}C2Q{f5+4)CwOIE)V;q za>u=4VQ6pq@mX@7XYV%Vmf9~hA)j&i3C1}0t!q-HK69gwd-2>qANu2}JXF!Y1#suU z4dPivn!)RUF7>$lgkYRIvN8SC-nHdebU3Q}A@j_(DbhQKbImQo=f(7W%{pEeoth<& zJ(_6$HvVlRpY@|BX72v@H_}i&vt%V)?O!$g6#U4UUgq*HG9dED&Nv*cO07HmAIPhO4HcG1uI_pip)z|{q9ZyI!5 zErq)duJ&ic`2U&k4n{f1?CtODy@$9r(OUE&>cU*8>~rcc`-kbbu!l%~>(BDf%#wqr z|JmMM%$sUyD595s3kIc$2xnv5jPlJg)Tw9f449X z@ckQ{&3vd5dnb_3?B!M5Roa_ViM=Ioe{uFUp^i+3Hxs>v%-*}5y=BDxf)2d3rtFWj zo^5Io)`#4r`XGI>r1j%}_HT`S6ZTH}8+-FAv9}2B0JzrP8p6~2{Fz+M-nGu&m94AG z5yWfwyWh1a>C3q8ai`u^O^zk+i5{0%vS^4b;(e2Sh#!jbkiV{}+q&X>KL4C7nd`VE zWWJ2_H=$dR*}G@J2XG%C?kS|<1+H2jY`;p|!!7!OYsY!#X31xse=mBwhWv*-Zy~D> z=|NZ@?&Pz{=n=FUyh?fbN&X#0#xl6|W+eQ(8s2*J9Wr|x!zIL@1FoIHc!84dn-@5? z*_+!hOV0GT8RTh$x+CKTj{7KatI#)SBPwgoy5<1>B4bgRJO5(bg7dTFdbmy0B0|Gx zZHb}-k#VmH;vYrMBkoo-1x-RDyHj^xqK{d|`02{c#`*OAS#p!tufKSZ$kXI`0U39U z<6ip-znMmSDe6QzzdjTu`JoW*6P_QQ+4cX}I}h-viZATXmSmG9z|uqCWJ8s*AYCw_ z2`XwRDkvrN4iOMhQ5Hl+M2T2HDXxf8)hM8dh>KJeH7M0aKtwU0m;LYsLo%8OQGktEeh<)gNYvP@)c%Ofg`(dj&MR6dn6E@Fp{b9T*0~m)) zyz%HX04;&GcfR5sLfix3IeDBmoZg?l^_UR86s|W5<8=*A5YbCR^0wn5h-#U%7k|?^SJZb zBNC{Dq54|^zLww?6Ymkl+a$p$Mgu>XC$X7xf8zpOF4* za0qBWuq7+SUOUk#I)QE=o4z51wkGqPg&(|kB#0d*p5dgC{n!13xFtBQI}IM&^TcHW zxn|P$fV;;fh+jI9`f50?L_cA zQvhCFf2g`r96flx<^keQfg_*=orSbf@@A{DmVPT~A;04}51@FHNOO+#T|DF&XkLCg z$lJaKdyC*WI0QNb81w5|MeZ?+D9fCs7!JxSkNa4UtO%8dhkORFN2(d};{n4D)^I-r zf9-zeLFmi+C$}%s(0dFl0@|*hm0fjev4#R;!BFr!eaA4aCtXJSPGKHY0=x7)C~Inh zXtpwBmu@f5p(kVIyzO_xt^&`~C+O#o$%^@c_n;n!D*oFO-u~t%n@71DSCjr zz*w+g8SM?o@^SI3Q5ouf+4nH#5#ZJBa0}_>eIcI{YTglw_jlsT)#Y3ZnB6`9dyM4G z?vgjp4>LGlwJNl(^PtlhGy|H~r+Duq&IhDCN;o-*ai^xiCpo2T!hV~De9TM`4d9jb zX5#lbq(2I7zaqr@x#GQ}KAa8MCj?i5+WUAO*(pdexE8ptvA)&zx}Hf8pTMg{m)oc4 z>2ho9DZHO^DIXXXt9}mNm{&r)y1iz@i}}I&a2$L8X?vX~l@Ryb1ko-6?{_BN@BRaC z2E3Eu)$QU0eDMuA9|d%|-Hsm+#|q;1gHzypP}a*ijaTV@X&2d?yXMl*ympa3pM64; zz1qKV-jIL$UBtic2>x{~NDwt&4UMN&;cE;!18uKQ+50eY&w!VJyA99rq8*9>>^X3g zg*!X{JTtH`K}>;{ECppFI)8#@jY7OnD&8Y9uCY@z2Uh?$KFFYdSXYz#=i35tdBE9uW`T?MM-&210 zs0nLDU}?(wmT)(1wsT|Sxe3#I8T$tUTP1JyD+!`WW~krUhR&DZ6wtg~mA%(AV?7SW zfS*Rji0RBTH@rz(<~qN0`d``JbUPNA38EjowUxb7(U}S60L?pA@h&B9J@^D{1#;i! zDY)f*+^NbBdH3zU&b}|ay8kVz>#s85-Jtx?#l}0DvEw`N7tr=TqU;^qoW2~a1*;#tVGIj=y`a{|HJbWK1G!QZ`Ie0`jAOk zehuGAAXlTr+5st_jbN>fG2!QF@~%UJ_ZxV%z1O2R z8r%)Ez0R6JD#sDxZ0)F@pfs3;51KHR$^2(2Kj`^)`WNh{u2thr6xUyqel&O%Xx@5? zcV>I1_#EVd6QD^C>O%luY^6>xADza}>*opw4kn0nc(v$xLi-M(c|s3(WxrL>zGu3s zW72LZa=n?BogkLLtLs@0_|m}wpzR%??CsW(Jq@r9ya8l;-En}rO1XIl8TCP*zYl!N z^)m3v`I3~gq;GW%=L3M|ouPQAUrQdYbBflW3CP^cH9gD|vKYs5TO0L3k7MaM?3cW0 z@J^t9+;u(YHo-GM^PX3{+ll)Hl;}b~3kLCgngx`d^zW!T7BLgMBrm7ng)5iw99}Kj zzX|BcSlN!FYZdNaUj+YV!dvz&!`>$RPGk5wf$l)tTMAypF^RajU^(yux&NTA6Ul)4 z84~IDL%B~Ui1zSm(e|!EPupAOBKG7gyTK`XfxAFD z7{pv|LV&S>eS!>CAGCjcOe$YfydChvGSa^bJ_efiO2uox(J88f=Aa%}MVW0-vC zlh>?fo=Xs$BjkUViLcW|@*jwh|I|McMC{u}{^{bxVab=uIY^-MKSSkzB5_ZEm%$>i zld|ee|Ja%Qiw7-YvN`{MCWyB1YSH!fUG#MRGcICpRwR4>N)XePy^{MA_`U{*fwuQ$ zWpDMa_#9jh$VO|6XMgJ1H;OH|F&>B2l~puO26ELuWF`0Gjt5#e0Uhdfm8Y z4LtBrj5xN8>u=~E_sjkfKS*A^zaZ?1;w$Bc-MFbYI(LCbf%d~E25;0S#Qg+BclKBa zmM|C3h4Un~KBOF(8;LZ8a_DHjT^I3db_Bm>L??=-@XB^TlNL4MYYX}So&RGh z{{iBD0Ovu=S9yL8b{>&)kro!7N5uFJhiS>>%LD%Hj9}&E5>~q1hFM@p)P4+eW5B7QC z{S{t4FVg<2X5wwHc!RlDfA$oQ$#?&-&bNM_VG6uS8$#u+{pm6Bc8tI~`(ySiBixUh zTus#@cyEH&t?Z}`UwhEY#G9&kpCRrQuobKa)0?vYO+A;kx>fzu*Oj|VCW;yGlB=NX zCcV6W;~=3f|GtVhhq$O)jQaHvf%&|b_Cww}5#ALh-Z=Eif@(nX%I6Fr!S@6vJWFUGU0sn*}eF0~ zuL7As+cizub%?mr;1BTXoy;Z3>*R_?e(R}zX`-?#yL6%$120K}Qt4Kwr~w)Q&HJq4 z-S=&jxRLlX%l2e{8e9k3fh~+D zyMZj9Ts6P!j=gf7=qjHmmcW~$?7fTh@_N2mgqrtT#rq<0ZvZJZDY~*{z0v{RN zlK=kAM)~Bz`x(53st|9Zw;L4i72+*!{H9|zaZABAAonxYv61uCP^(C1E#A3?vEMm< zHETQ4yDBG&Q}DVCU9o={@0J|Iy^yzsc;_hIp2vAl;7FMZ>}zBfqgWS&uWOHXH|))X zx75Z^eTajv3TSNNU8;B|5;q^b30?)$nFoo(#{OZ7s{eW(lvOQJEQVM6VH@ec0)GPS z-;IiQ?(Nh+@HyBH&M#xV#eQcx^TD)iqYcRC-32H&pC?glgx788iqoWz?akgg(7d^d zcOr39!CWvCnD=Mh3C8+UU+?Fy!MOTf$PW({j~0FS{T3~GPA`0NoZ?X$QXFlP>>>c4 zlwl1b_b+~8)N8- zj)V#5T>&(28^x=Vh{hc2%$NO`E+Ok}sw%ho_*V8W?pldrD7<;?m>+0~UKjWq^a=CG5J&T%sb@B%l8C>$6*AfrqQXYRU*;sU&06nBF&q`#{#<0#POf0G;#;Jqr1 zy*ThLSPPaNq~6xz;0UWI%QaTg4^PsTr60V_62(1%P&vi-rww(B79K(wE6b^u!E0|% zTn}(NkYkK|rnuxvqfTYR^PGw2k69Kmjs9y0X{Ea4>1Oa))5gb$l+B^%%5qqEnW9?)Q5 zh?mEjOK~I*;`#@$6wC*`U!@$_7nA!d3+5X2Cu%SECXzn4ZKAjiUM;z#KMP_8hss%P z7ujnQmkh20!D}d(>&dxEI^~?!K2eM{@$@3i05BRzc3IBq^ojjZ;ueEffSJb!PiBWi z@sx>Y9cea!y+HGD`GXYuS>j3$;d&|{c|>*WV9mG_-?^_z6mP?$`^6L8ETRGFngU%8 zIZ;6#$GvY^MMvVh0EU-*<&c3*vM%fSTqe9fnRvUS(-#Z|n)j^YmDfkzO}r1t=X{yV z*B{24)+v!^$A;{E7@av_DbT!4@FC(jK->}VBgkPAW9Cg)VQWCu~K=&6EvlOeFxIO;RI3^?0(J#3<+${%GbeAa)G(oyt-d_8J#b| zw?Ol5QM|Tv`WH|U$mgeN`((WcgtgyvF6t?{CB!Ss(~Vvd(7eAXUL!F6%K1p7eF@zj zeDKyc@z#c~4Y=9FdsgvIA#NsE0_FqxoO*j}!w&&9Z`8c*QS|%p%6XuacSyevoCVt6 zTUGm=bqDi&@HzMh9AiG>XFekR;I$e3Le7`GE1UGLafxCzyzvIDC_9dF2Uh~k`=sLS zL)-)~74%(h5o@mDT=tvz?NiEM`B(3Ix+f%xJ@9%J?<{n-f}eoqeMj+j9q$wa!DC<= zNF6|Z1k&D2HGk^=lCh3RpOh%GE!w~S`x8Yx zytS3Ra$lt>IxT_bO^U!<(7sB?rN+E00B>V>-HLY;KX5gi?M%F?T|2rHr)2=2Bi%oh zx34lKQQQcxZf`?Le?M3aw7oYNWR4dmaxcqW^gqB2>|L0b0eL<;jti=!FLZlzJ)9^u znEX4E^!J0ur5=28Ua(y8zDe9BunX(}=5Z}Mti9#VNEDT~h05(K(w_u1Cxv*aMp7K> zh_l>{y}%86aK1E!lWLi~x5=Mv)MH(4;_*ar3%pv!4RVOv(dh?tzqgBn^jk5@-?obu z%R=*2T?fozhN4P`cP6~rzun)A7DM423A7(1FA`k;Jxsh5x&CX>693YUTaY2mt^J!a z8~>X8@YEYTrxe|JCf+}TOd_xT?&R29(wY4mcSU|bWMyz(V0);Z?L#N(9{Mw&?JcL; z+bZI+!0#Xz%vjBOr8C#iq%e-C`G~$R**7;))P+~}4|2SWznAe2^aYx?tKyBH%rS5; z7!Bgyq8(p{KaO*5hmbq{29y4Iun%Zn^~yHK-BUS_ z3D$sBUAhB7)be%FryXE zjmE#a?#fWwEBmd$JBcC-UZw{@`IYoB53-L5H1Bo4ETYIa&sSMZBKo04Kc=Z-QN6xMA76!!`|um<#W;> z2T>1)cy}t^KE%xeFM|0%UeDYEKgjjLEbPsczm5-r^=xyZxE@|Dy4>DGPsYk}`-BgL zm7CX}-^a4ta^W2fuUipsf$vlBInaJ6DL>$cgTy_+J4mI-=kiCIFX`*5(zkFP8eZK` z=Av6_2G?~1&8yzO>KI7eMDQ4heb^S9mk+`ZazD?NX!uvJGXvY$FE`n{3>`U7SWBpR zn=5-iChi*`AEdnr_&8&GqSykj?jLeUZ+nFEGC=d*sCZ`)w-me! z-UOU-6-VwppC}B6NFL9}f`1$Nb~>8b-ga>u-8iqL=`$t`9|Z-aH%LOq0D; zIcanw_H+Zf+}>06KJvCz+(x{VYu4uPAGD9Py(zmA#V00v2ckCuOf&H^#7J?xNt}Ft z$uYu%Kt8`jKQCdG>UZ=yE#ouR*YN6kc7gP9kJ0vlw)d>!ZB5+upbzK)WP6k6)8x3h zQ}O1VNBTTbbfII>{iM7vVFEfcf#$W)-le^T-Iw4eQ_>gO557H#VlceY-Xn25e+0go zkF%}>nzyv#&B$P02CjO7^(@G2NPGCnDgw25e-Q5xDN8zSum5nOSfn~aX_VaezMgv| z4x=aOWPjGr;ElNNos0e!ll}tM8l6`-#A%cME$GYs@Sm(hc!X+D(!Na;C*gG)U&XKR zO~kGXggU>g4Z9o_=WtyTkm6rv=iV{K1W^0qP}`F46?}{GbRS6+4z3As2UByNx`knH zbM!UuX2rXiV~->n30Y=D>}U0Oo&j$icy+nY`P``gZB4v86z_cQ-Fq2G>4!~)&Ku$F zV&c6XzMQ_gUiBgY6*xQ)}?KGr!Xp$a<9GH$q*XT;C^(G?O1bLFWu;_GHKp zzbM|fi2EGmfTLgnV?r+Omb~d|+|}=8%RP}OR>3R#!!797dW!2D!3{wB*K3rq;|b!n zfUm(GaHJP=p2N(+*(W?nyQ+k~UZ04ZMDaPiTI9HU8l7`M#>#rym=A@GyQ%(AeUP@x zan}!T6a%*I4`ZHYUIDy7+uKLkJBGOD!D_Gy1Q@fXtNBt~J)^zq>!dSt+5do7_b;E3 z{!8!!(7b~bZ<%NCJ7^DXJ;3`>yzC1q-g=5xk2kqL(GQq-yPS{({Nr}fn0P4-_$;r3-J;9d~`@t!Q59Upepc!8=~@mW`#|p;v4k*Teu_ZpRew)z4C|z&UUlxY!t* z#eT;Q<{JxG3{@oyy4eXp@3v0J&{=c`|bm0+O;0?Af>gI6x z9tF#QwzsqVz&JZ&KI>s17BF@bPUqagmV?Y`nLD3kf5vUH*B|W>b#@sUmCEvdBsq_1 ziZ025?I0aqX>a&_INGP?aGs`ectlSVul7Srcx4jVzT~{NFh8XG|LzAryffjIe#ioy z;OhZ;1MLTS&jR8YN8F=eA$SVN^{J1JPp)%(e&hLWy4?IR9`RZPdzYZ6?S0`Q_GZ9o z_J!6SxJT50SN1#E zz`BTQI6-xwd4E^DZHVg$`h)Q2(D?oNpYsza;Sudr2LNqeyGTb@`}Z8Yh1G{lIK%y$ zR?;IzD0`2=dp~@SfrUWZE288t3dcLd?FGlcK_KtP=%My20)FE;xq80fb9uyL5$yd9 zJ#DWQ-oot7f-~G+cPWo}2VPlj+4MtE&vOqMs0*~c^1UgDqc3q&z|&wR5LXD{8^C@u z?Ksx%JzY+%v0v&DUq-NZ33}RI=SA$zg)`h}D<0YLN4P`s0fdjQM=H*K|wH0qdMPm31}AM5t!ujmom;7!3NU!ngbI=9PL$(HC{Lu1b$Ncpej5%u;M<&|j>qWDWrkqBx6o&Pc_|DB2J4x~(86D@Xl zSlbT>wf!Z;>->v49&sbQDa6en!)fT<3uXb$TUYT$zs&WGpfgAT9T>klGk(QC!JZ=Z zQ=XTaNc~&|S8hFzm;kR9T@UU+Py1D#hbqhusS*5;QQsq8g4e5vr@%KGJO{MBeU!b< zrPOIK38aG!0nRaT{!x}$J?gEjU-EqPM7UgyJ>n~qy|YL^AFKx2-mwO+{UC8Cz;8g- zgLqXBy2y(-uk;9uny^X}?@{aK63asEa-_jyyLXRW+}h0+Tys3d`5L+Zl7@aw^gE+# zkAfS$S>Pnl`Fl#`@7|TGhayhQbkc>zs0*CWegl&W`nJacvPh@&REz zep7XHwOy;>rCcID|Hp@Z@%^Fk`)OBna9!p>U)$B+@RMV0RFo*kwbN4MKD)f1ApE*? zz0WSXdqe|xNfMMs@TGt*Cf*T>_u&ki7)1O;Fd8J$9ynbQ%x{`%!$T{n>;8LCuJCF< zP9^=c?d=|s0 zpWhi!_Ud+)+s`9Peew5lYlNP*x6wuH&5B@eMw&-l3vX>%Xh<@AlfY9z+uKgr`y+Aj zYna=EE5I+b2|G7i1g}9>4c_PX*rt&8y$r`o)`E2l5uz z+<}+DBWvKTDb_Y&3yFb&&z$4QwFcK_hzeDN_-a7`g$7IIM#Dt zhRR2e^O+|7C()PVN%%M~?Mp|$G5W?&Fn2(2GuRGvzI-ZQKghVZjeN=HsQ!`=dajes zo31w*<2|C6$u2!FDv7?%*S#uVa$Z!KcyqpT(f6UR$D<^Uw*(!5w(AP^`4Gq5#4P}8 zz)H}y|l1nxWGXbOoCCUBx?|xanXXcnnOy zdB-R&$$L_@6Fq*oC({qXtLw8I@8rJmi|A?IEX5oAVfE)?`^H{)_rWWRwgm;f9KMZU zvx#@N;w`>`dkR1c;02;TdmzWq8)+5U^cgi}MX?y|I^%wiNch^Q=Pcz#GU;yveSxk| z^9|l8A90TYnfD&7S(c?3bwJ+#mOaHIhAAGICEZ@0GVv^gM~*|`?M3=ATqdcem*Vq? z^$~dInRxdYyb;@>AN}vqmwuD^TY#RNC%i?d%foIoJ;y=fa)6YI_u7K<&E3?Wo~#3; zebeLtMqX|=H{G_Zu z72(zP-lXi6=4iRtb!Hjx-X4Lss);wX0N%o`@AktxIRbBjiMQ*2;LU{hd3bd{bxi~I zYgjAPF!6R%ynfD+n9q|;D#N}R$INvqub+B=eF}KBX#duMSBe}L`WYXs`w1naq@cG0 zJ=w00v7W3?`u<>%$)3?hT8DU#^H1Pa@G?k0%D5736O(eSqIYus`%rRcc|_BLA$xQ^ ziV1}3(Jpukt4CsS{=CY1l>V4U41ibruM&I>!IePQBc^In9A8azh>pZd8T?Fi@cz>9 z^Gv0f7fj^1?{Us6nC#Wd5OPJnYj=eLC7t+I*rFSr(@0Q0=kg)g5$U)LA$ zgh%{t%CDYZ_e4*YnyfD_cxCw%^!zDzSoz5LwIAMu>`?g(k$mjy-btwKEvxKZNZbY> zWjmpKpUo)tO2g~7JYVO2(jz*;tNr^^AMOo8CmU$qhKhH1l1=^~5pNZG`KKC?%`atuxP`urV8vw?FPq=SQ&S!r281~i+D>wgB9x)MKH+hT0q0`V= z1U3TAdyC?AZ>G(Gd%!r*f&EkaZN@q-jXjDbaOmqU#5|996J9O4+!msz%WV+6g_T=M zq;i8dTlwK6{$2^+Hn10Hdq*pKom*&cpd+{%$n}7)j$v;X{9sW~`9RzIxU%<0;;L+AEdc(oJ4D&0>`OBLm-Q@F&G)o_ z{R=&!6TG^f?c-jZYtXw6Xx=4?_c|^_`I+NV&Jmi=ALu%)>sdCugG_$V{c|^XrO5Us z-@A(x`ka+qcpi_ybBl>*R;`(N;gX5WT*q(r~M zv31}Pj-73rEb)kv zM~a@;1@8^;>VD%H?0gBlW3qR=VWPvfgL5jqlI&FJ(3+7DF~Z}omrVi3orT=czM@Xj#t?pkjV5{H52rFuv)yli|99*=bWwQeV3 zsYkqG;@w2WlHd~MLV4x!!g$l*%`)+J#2$&xpfG#o@xpjB;5}vH{jQ}Drx#j-^)8$@ z<7c}_!|zfUJ__Y^FY}1l??Ux#An7HBfN)--@n)4Z&BInhxg7)!0TG!5sOT`%|;gN2l>A65$vl;k?#vH%nzCHZZh%S$$1dDe}9r< z5BKjDZU`FLP{#6{Ov1u=(^h)KNfYn3`t0Epz?(9cdk|GUyBM!)75&-KP(8bYe3pBS z>-ZVt!tIssv-fQ`Ha#98EX>|qcpIB|AMa`_uw6XOu`$Ztn)l?tf6*8vjO!ZZeKo#U zJmOXp@4ehddEphFj}oDO*wfc0?ozy4+jCt(>CjkV=ADTDQ(vW?nRpLPu@`9XF^>HV zq+F~WXTv+s#9I#k-uID3ds`i`BE%@+2S9p0F@p zS0?j#cvFI@d2ac4c5y#G{F6}I%M?e7W7|=?I0AkJIbcm0>aoX|C-~GjRP`m!RZ%wH zwH`6|`w(v;1Iz>PwjtEK4=dh7#3ddx{OcvmWYV&vgTbrIP4aTdt;l-UBTk!mTakVS zyf+eRUaE%_#}eYU11Y-*3*+^^&w0dSq5ah_N>Ue-LWk%|RyNc;+Pe5^liW zH68NX`#JX}(-nwxE|yql;%#!sF6O}7g;4WGkq&XpB5p2N4*Xyz{?&6GpIUcF-j1YC z`;hv3JapfG4t6{Puk}QT_e#a24F zRJ`{P_auVq^*F*TB1;Q1g~G4mp01aVHJll7zB8 zgqNGFXI~dj+B9HAZ&f7)9KJK+5BU`>|KI3&UE&URfWq_R%ij zjWKKz8{v%9uQSbcIJnU>i{nz{^FHNzG29P&-sjuTex8YUA9}~ZFF@NX-6eT#A8^e% zs0hC0CIrc=_qj4uJJ!7E2R!0y6K@ha1=)Mw@>tP`<7Rtx{ndWZyzVbJpK&tehsIaL zMC^a1y^|R)n$@rddFA`E3tKm5!dt_{+X8#812=Sa-tPs&016L__KA7h`M$Ap*9UXC3Ch1u!KW?p2nv#83)5AWS3-l8g>On9F+@g5_8 z=Rnkle{Yv%iSvL4K=x1Oc9|L0F8zm?f17w)qH{gy2Eyyp?ZgcMW5NFS4c_VOAItl9 z17W=BhdtsDyt=(lL`Sci!s}C{c67~=U^Nu3Z#}}6Q}p#I1K!&3YWt>OUu*&GeFn## z1clk@JK_=7nd~g8`sDh~BgUF|i>f}k;hhJs+iS%pE|@QN^Y}?^&ESW z&&N#bf-F_;`u@C3cz43I!vK0AN1=X-vt9-sC3o|Ny|*M?WmQ>t>` zRYBuaHe+{Fd0!SH@6Rf1oC?4@!ennT?9u*}`sRLGpLf)@$uRq%OvU=*m`6NfvahK6 z=}dT6n(QlUvQOsFY~Qc-9KuqcXR_c@iAZ+3k9)+2COeC&U1h?1(!^U-?J59o>D*BL zuS))Pxl6yB^B*bS3Cm?F&J&;Th?el^`cqWp6M%Q1$-V(;w!HctUOpo^7Jnu8bVn#3 zZD;xq%$rPh7FGF(lOD0g#9LJ5x(slZsQX-&20}hzdW4{MZrOcBA(t2)9$`|9g&^?Tq8Rog6czo$2tkhfj|e z^1Na$>3;*^c4|J`POj-Jk^ek;J;kZ*%z$^C$<8=*=0mUugpU{HIOYas8}*p3%RTiM z=A9-xYmr{ppKv={uz%y_xRh{zYF=$;HoTimcFO!q$oxq$mwRFMi8CJYgUP<4YBw%; z<9`XwGYjk2B8|6aHXHfy!*dlpx}1xuUvvNJ5&cc}U94YgPPr6jXD)oxO?DPlzvlbR zBUYGri>iDw;N5NFJ?d34#pmlTE?D~0#QRq!=Pb~Rv>xiq8mqio z2gvnMVf~yB-h?xu_1jfLZF%eKaC=KcBHSzai8^w4)E&wRaEw7 zz}x2%*y}yVK0dth!QAnDtU~&ezLeSZrv9Wg`^@3{jC%bRfHwnPS-+(0ej_@s-wXHe z*BqM-q#PrxUGDF^aa+`NkAH9-kjW1L*3$*`e;;scH;^LN-I4k&U2d-PT%QcD-p}Yp zJ_}mM{+??WHrB6FZYNz*1V8+_!LT;~Z`of%{>{LL1=%b8J6yGk^3>%>*RN>*rv6Di zGx7emmFMk5xDG(>=N4YXrWGLE z5AWdH^&H;}Wd9)d3u7u7`}n!a4|<$+i6oJ0;$02jR7S;U29i1e)n|O}RsLIC=Z?&_baxbd#$%MBHyt@5Vr9JBLN;WC;`XW;Mnay(; zq@=_qiE$?TimH49@Xj~cR}=g6bxGmtw1ymO1q!p%?@SV#Om-Gk`FM*Zi4!K?qAH&h zcq^O>mAjm`wk409LHPKg=h^4%Tx-Tb^E^8>yx)Yk3A}ng?%$lxF!SnpcHa4nxFpfd z#QSf~XGG%7JD&mXc#|KxlK;M7bOb-h^A7hD{~!oI?~tbaAfN3uk$USYo+L6%ycy^$ z1gk)}f93h?cZuHu!q@HTVgB{Qn+dO5HV))dboBm0IPW2jodi-Ab3IvM^Wn4-Nn)SL z-k;Hn`h@%IK{#(^;%b6M;Ke_T_onn;i;?rcmLVZEuJ zKgk18#AbMPeVvS-2BH_KeO2K4&y0V#Uy=^*33%Po2;>m?FRXo;_X#4^E4?lh<&%Wt z&k%18yng`ar+@cvMdIp$Mj*WZ%2fWc4#Oeh9B_7dKP)3#t$1eT}5~mWcZr`o30I zr6lpA$=;%>zdm@&{1w{YS%Y5-DmQuFb~^cyvWfkuEfMznO22E2>)y&qq8YsIU~1|I z`%DG-;os~>1>o%sZ&CLTDOHlh6nJ$#D=PnH!uyhmx2WowyK0iyY2q!aePutqznOTQ z@>+Pxu%Lc2k>gS>HV<)COA<9Ng#24nygqol!>i?8>hA|23&>dWcy@}o5)5P&K{n%A z<^bbft?U;$r&T>ktTOS)`^GobHtu%{TjxZ)Z#>n_ywIdC&k?!#omwE=4sB~QFV`ev zhlCshbFm{Go}c0IszfE(Pc3NPTA5=tfs|)xk3Axce`VfK?J*&VHe;lf>Hls1GSI&| zOuqqnYZ@Vc`uQ1i+8fWj)Av^AXxvlJPybMirB({gy_ixHic&PXPgMJ$Nbu3)Nukz(;Rq%aGuX7K~c{RU?;y#l|n?b;$mh|?Hr@w4S>GkEyuXBf@`%r)^In0!P5DhjZ!DMyv|ZO2|1=@VQ7lA7DC=&o^@|G3vg)uhU&GNjR+`KQ2edT>qtAWJ8MMuh)5I6UU{zO<0&8 zv*B&1c&AgP-Y0!7uzzOQYvDdCwx*>x?jY_Ba1?w47Q82fudY>ep2hnp9=3?*F+kg! z*(6Dfg;$I8<2iIJpBwF7@+NQ)KgJ|;>^jgDnA^Qu`9SsoQUVl!yJ?bm2Hq5O2l3qr zUv+Q=kiRB*yAmRfn~56??gx`V4;;UfvX}1;_QG8Ydt2`1+59BVZJs3FjbQIg^khFJ zd2b;s%-)m;_NKK+65qkw6J04!!}k(+8)$on8iyQ~-A+*)Gy@I5F>D?6GWUM*t}H*# zNsvC+%}p}PQA}%{By6@&eQ44?|Mia2UR@u~B-zAu@JW&Hi!Zc3z+2PA8-D+qnOEMw z*6FY4Aa5#RVZ51bl0;h*Z}|OzW?p%JV0UHjHuj7P<4wDod8LUr+}_$IKl~eey~#-; z$HW_MubKDX*qaS+8GESQdXvu~U_6lZp$_)e2zr$Jm5B4~;eHU%4)lJD_ptF?em~DO zN};{U`k>oedI$O~c;gLSaVzPMSF(tHgqpXkQ6U`1?zW5aRg8o~67%oJNlq#Hy-->3 z?t*t1$=A^khQhZR4BH#x<#HG)j=?nEj$iN`xG%Y;9i-J_Z-NJZvfsJ z(M9hE?}#KZ;S%ur;oW=*cwHltL}^FS`@siq|4YD|3vcEn;7uEqBraS6-Yj@;h$(vi zri|u1$R*&-g!eeSTJ$_Dmi|HZ1hSu8N62^_aUSL#lO(Q)E&BXsz)T0NBTSPfiePU`P#oCOx8^AH9Oy|`& z=^977#IAzYF|t1D`PRQ#$7I4=r&x&hJATu=?lbc?K4=x6utt)y?;_6&NE=T*HSwNZ z>kv(ugw8%}&6^Ksd&Qd$aU1!SGLSIReCs!kO$6x^lEll3Hl-=8m`)-+e@VTQeH(bS$aT1UX3DeZ%l0mLxy&m# z|JLhh^IAdjnWc-Iv?H1D{%P{V0{EIUqkoG~+pE_%pAq*pkkTT-xPE5T6(M^if1bT} zao#&FWbX;|&VmwO8|wvWuc}Y>CdAzg(f~_Q5vTTEvzgNeCNVFDCoTA^?AMbfwzNY$ zL8$Ys){D^}5cfOC9>979?8iYm&og+gzZCC%Nn!wex}G|}Vf_P9bwZurfyQq+9=O3K zni4PNj%J1DH|_o;F&$nvX?dNSXop@OFv`Sxha{rhpCax>uo|oavncb<+-Ea`=g+1y zZ{?CFv9+NPv4knUByq^(N7>K6kIn}`7NPW`tatSD1+EiQml*X)`yxPoGvO^>-0pfZ@}Fe%l(oDHue>upi+1F4l)WxEEvV z3x|YwiO}^h>sgimr+6gFLDK&Ly0O8hc{?fI$%mZcLs0xM=WIdBBkZ}`*h64{q;omD z;8pWt!Z*BI5k>D4N#Yc|x_*W4=SX`=F2(V0`oS!CyOofZ8PXp+?f@mfHT-)!q@|2Q z_BcApQs48gC&1$Z#DXPuk^8w#^nQeAIxJ5TU%^9_3yNG<6#v;O(pW=CcFFUYVW8vR ztSh`LxPImm@cQA6D_QjRx>hEMq)Wi-gSYi1;LU}%$0guRTg7#Qmw-15-szWsH|3Qi zvEUN$X2QGX67afTW!%05yczHwxdgo8wIp%=67Z(OTgFxN?Km6Wx|e`AHIsSACEyLf z+w&6edS7Qh`4aH@;eF^5@Veee5(_T@uMgfe@aplmBa0OI%#<4m_4szo82TJHlxF@T z7cq!DzaXy%7m3F8;Cim4=djz2QGWyvIj1E;IaoUd*HGME_>jz2QGWyvIj1E;IaoUd*HGM zE_>jz2QGWyvIj1E;IaoUd*HGME_>jz2QGWyvIj1E;IaoUd*HGME_>jz2QGWyvIj1E z;IaoUd*HGME_)!72QIifjdc&WbNHx1!^aG16(cWE&nGcM$2Mxv_{s)NUa))bdM$3w zgcxhvLz^5&_r@Ph?>}K!Y6 zf83U2}vZtXj_=|66C zvnGnI(@i&ZZ!@U>xVwgr$%|>*?S^Je+DvTGz^kGr4jkWR=JUTLKLr5OWwoY(bmx>(=4!Uu(=7>5|bf&+_G3X0zHXHn;j79@bXz>Rs%D$4~`D=R@0*9d%RCf6v<9hHGQ>jn$1K z7yGX02`*fapBpmbG`w?M|1pEcj+Wm?_h)1p=N>#?rj^J?|XIz?*x?x5gBHV zFP86n;9j1|6g+Ai_fN|IJ-e0>pQpbfzkd3;{QMa+O6KDeTl3rFI$bH>aqpUH`M&$^ zD4+LTnWcX^81H{Qe|rCg{NK|zRnOB8<~Qtn`a}7TyNbK>r@z{j?|Wc&*?ix<=kkB| zb*h-}xPOSDZ?W2AT<#QK1P#*WKHS~PzBI~;r8THlwa;V;F4*6V; zl!ZqBle$^;LmwHTU--%kx5aIB+uU|{lsnq(aL2e~-A;EycR#oA`rh-u=X*ZmJ>P82 z_V-ZoXGX^a%Ox|;2z~KJC{xSh?%GEGky|mrD8GJ*!SK_BS0;S-T>nxVtIo1k8@%p` z_ZOEcf6qI!`Z^ZH9^SF+O8H3AphylWgN;C>tFWNsa%_-N&~dG=3rfxpq*OHm{fv#E z?`2f@@oFj^IjfS=A}z_6?jIHCoZUkla{rlfAZ@Gf3IAh(<=Jap72}H8%Gyg-FBKbI zxrC*BwfLCQBC2YI;$^IroaL%GWNnmGs@h)IaP-_OODimz{`1=(zJGpTjjq~$a~hfU zuP{9smuC_mq@IjB`T3dNl?ImXxp%>Di@!PhjUL8zQrEmr+fOuJGxPR+dwPAd_v_bM zOs?Ga+0~6UE%x5_$$|TyYaG4xWYa4iceQ!9*av6djc?wnch2cGjb5C)Vqou&25&r5 z;j_GouKlC$oHm|@8EQ}A$BnzO3qG#fYk2u+eqA35I<9}OpyLkZfr5_f{-mJeJW3`g z1szwB_i7b)dQm)nmLwMfF|J8`amp1@c)_ zQnF*8p6M@fU9HN|VqyI3^DpfYA}#LS1$|FB#G)>~m!E#nA*!CYcW%5+Tlsg^ zSRv=VZt);Jr|j#q%L)2+i>U3c!(Z9QdwH0V@D^;RFIfn8Ti#03PKbLth?FYVh}6xU zMAC$7eb2;S@7}PnOYMn~U zqI7`gx&~?mS_?;@R6w`b{47#WbNm;5yb@P=1fH+Bt5IQX-?wqW!A0MQ7CzTCiS_65gK~v7x^O>9+~f*Xcr<#wQ)G z+h2HpqV3Z4x*&a>t|0sM@BCAJZNKI%NWX5a`tIgeHmy>pv8O>oje1S0d22VSnsi08 zlJ<&aVl9ZZ{7)Uyhqi;yR^)pYaZ^eTj2!Z=xbBiCS_ImfZ9fP6mi zBrq4O1-rq|pcI*x4e&-VI?NGUEZZzro}~T(e?PSB#x3) zzpJZnuI>^)SN~b9ydg%sP5dRj#iL`>9_nlvY_nS}QR0Ov-b&H2bsy^d&~L>pmYA@k zi)~hmTWT5-u5T?=pQBQrE=5bYzO7Jwj!OL$tsR=b>3X8;-(&9>j?(VZp>9u_LC617 z{qQpUxB7bC@?U>nkE{PxzqnRZ;lH(C&&xDTLHfG=YJcngMf>@`>T90=s;}+2n0`U^ zN6-8ItG@Qv#q_oRbbhs-o|kHU?N67A(Dh%ZZ@0ntr1P!gwZ5Zdh3d5$HcyT&QNCKu z2F+7ZB*X44l__aST6~C@emx@1pJPiMx|F!<;elVO@fB&!iT2F_XFFe%8 zbv@SQrQ>z|(4qFXp5p3s;k^30KK@_%|JU}b+mE(`SN;S=&-1ihI@EUSQ0wXA;r=Oz zU)!zq3)*MW$Nw$AeD{7M6=*u zHjWs-ADIb97HV4IM-A1B{eJjyz2B+pt3JP^Hx~3hsGg_k4Rt*~%G>wV#GDLbb0|96 z;fRTejdePU6^n~2Uc5w!k|kZPQl(0lE`yXs$|2>E3P?Os5vhb!R#HVtRVCGwR1cHz z*|xT$z%Yhu6ygP^dPE9OUr*!Aice$+zsM8;ktMQ4E{)7>NwK6^d=|eYV991LEXC@x z2CTy7vkAM8{V!jXh-R4(BVv8L>e?6QD-LNHUQ`h8tLUpFDnnmG)D*RSb@`d5q6I(I z*>@d3ce6Hb@<&vn10*W0@P(erWLf9d@) z-KO>axBh({uYYHdI&e%8x@s8_e%`MqC8eFg;Cp7Cap2Gy!?X{dlUqbW94Xc#_}j-O zb`I7(*Gs#P$$0;`8{=i>Qtom6EH{#T+m_t_nEa7VV^2TiA)MJ~QpbV($e^XG-^?Z( z#~!hJw$HR{2a!SwJC{L;Dj=TIc=<%9w`hl{qdXQrDC{TYle~Cc0P1L7ymo9@d_I!D zeaU4HT1rw%Fc7_rFlDw8&e&{(BYqEtzUd9x3M8xaz+gCi-`rq0_~Vad=Q@&f^T}+r!ZOewegiUIm85!4J__CYdy}50B;otw@txYDj7dvhoI(pN} z_M*)bX-9ehRtpm9FnFXU`4MpXT~Hz#9~c3a9yu4>OI>C*(g$ z@loSDU0v|In5yDqRl1oZk}|^xXkbP<>YGnSp>Krp3vrM5jL@b+s|t-k_?+s4()a1# z^Qm;&ju?a2=M^onEiHO@v_<4bpXG1B@qxo4hQyAIwTPp!r}#U>Io8Q%=PCXUDK@qk zpNpO1?~u5$j95qGPVsk0@v*#R=4kO#{7o&26>q9Y%8Cd0>#Ot#?{Ukh^elfY1p;QEx=8k9x@Qx+6BGS|Vcn!|pA+qxY2CQ*}?PJ;{4I?&-S6|AiwvF}p_g z_urrV{?zvuzPIPNb38e4p)vjH;PMtb+uehR~l=?wwfHVvW zm+Lla)cDFgX_6;Rn>It5OKB03mU+@DC|89^>o92}3Kr@4378kKTe-MBfxJ@&z^=U1^xUCWmQ2xo@bU^Auv%9Oj7g|ML0z<+I?|RLX>2hQi^li zWkgvfE9KdBi|5R^oUl~kN|0)txp9jGktjU8dqz%RYVls;I-;()LevxWd9P7JrZbI1 zW8SdaglSDP-l)`q{<0N4N^9PZd$mXwDWbm^AO?y-oPQl6hVuT%;bMdsDMpFWVvHEe z)aMQ{PK*~5#GSm~=Pu3<-7W4Bw~OAQkLWA~?zx({En1ZL2i~^+bL$Eh&^{!8Mo!1s~)HHShm{-USm)xiEh=%r_kj{}{|* zcx33i^v^}zf2k7`ub|dV@zttS;lCkyAF4z&GFAfJ?ALxJN!efH|~smpkAfQV{Y?z zyK3+1I^Q;&-hG$&(S82)+edt~vE_ALdVZR9YQ-PzZ@BWoj%Te?Qd>X2tm}&@v*#>q zac0AtUnICL{QTL+ab-^So7&kx9x7e52jaxad zjZW(Q{^zR(Jn{6AyT9MjeDq^`Zd^XxD;(zjw6Vu(r?EFPYin zo$}pcN9^%7d#S~HDK9RX`t!a&CV$&<+WdwW9&Yk@`wm;aIsQ?Pd!K6j^821PIbEJE z*6!Tq8)F*0Fr@Z@q&FtszWCbFkM625?)P5@Jv3}y`Q1x;@9Z=9%{imeJE#8nAx;&fkvoj9Rz4V{}%%YO_B~E@|EL+PgMOnVBP7JyELdO+&vr zG57SRpX9DNI{uke6B56FrRjr*&o)0+<<5g+I{j4Xj;eLGJ~H;L0VP(%Z>)U&maDJd z_SxO;=e`X5HR6Y57yBpL#8smON)q-zqETt#ItO-2Oo6jSn_T+I(N!%@g`;9QW7j-}shBe>H!? zPxCiz8b1HXL!CCB+O}|0$(=96C5^3jy#9*m6R!Jed1mh~W}Kbat?SkEMsI7hY)36y z=JqMQ-ut;&_O?z>EZx&};D^ye>(nXxp!2O-EfOj>I6u1H4P}bA=-y+zbMQOP5fdAH z@cz^X>y4Z9!m;-jUq7VMvh+u(Y+B-%bpHrf`AthI#*AuS zt;=1VYpvao*<^KGm6q47*ni_qA1=84=dlfjb;pHfH=V zDs{3w-;+=IZ>>FWoh(n(c_~ZkN(bC#rkKoJ-WH$!>Dz(Yh%|& zb#Y8_PP4AG-4bh${nPef>||Sy*jH>{Ta%m#vDvmb%b8+Rtv;tWdRO$h7>E6gBQENx zxY6P+%gkbDV&iOeWA@ocM?V!^%l<`7vgQ3^dz|&GXRL9~D`P&fH?n3q7svLqB{~mU ze{mMKcCf!2^Q2`=F{}MT%w6`GF+*bJ+xI!^S|{2bja?tzDSCswV@x;4Yg~nNKIZC} z&Gx&B&9vNUn-x1EW-hypo1)7)55#qIPP7hpq}poQr#Lo=sQGcXWC)t#3Pa#ZIy| zDwZH>N8Rfv89P6!mTik=QB;#+e?<+5z17*$`9ahS>x$T#aUa*Cl%=P%9<_9wZC;4b@+n0?NPw)J?Un|?e_D&* ze(3aNvg<_S?CwLqPR@${XxM}I987+>)m{;`DLc7D*3;D*y>}@2vwM;|+14FS9-81k zw|eEb$&Uv4jwCl(IIrfk>E9*qoYCoVG2&=)566vrt6%?p^7!^6B?*B5gD1ER5L^Zc zK7s@rW{@Er++BhOiy}dS1_%}$0t{}!-DQ9PIrr+W{NH^3?Y;kfoqf)=tFPyMSFf%v zU3XP2t+D3m=Go?W^G5R)bDH@WkI{NdDN8v^1xpvpSj#NSV#`v?QHzV!!&=H(&)Ue^ z&N|IH%NlRpVm)L%YQ1TFNx`RwOFfrHE)gzcT{gL-xSV&n=rYE2oa+?V)vhtFhg=W4 zo_1~Mw$N?2+ex=H_eP%WJUfuj-srj4bD!r$PmdfCx%+ty^m5JPowsYg(fQkZcl7S$ z-NifE`;hlF?|0raeCC%a?jP&F!GD|oUjM`Xr~NPb-}GM;5JJ{=Mmg{DXJdblJr$o7 zUm~$=;y0N!GJ`W~XV%MX^|jB}pZPv`6r>|g zol~@t%6D~8fy2t8s3;{*=OGI7D4TLrQyR}Ms(5j&sqA9*GU?Gh7I(YcSxcG`E5dH#Be<_Z(9)oBD$wo?Vk zo$XCY&HyCCFYzC>a3u~!Fp}}<_{&=Or6=LTx*MKT<6OA(Eb48P=kwjYr_e3vki;m}9?7un-8iAOU0uP1|WCco-J z&a@;s&$8q$8y!;jva(&-sq9k@ zD<_rn%2nl-@<4f_e56B{L3L8is*CET=1}vf`P2fcuUbMarIu0s)gZN!T3M~CR#$7Q z_0`5|OSOaAN$sk3Q+uj?)o^u~I-EL(NOhb#L7lA5P-m%g)OqRxHA-EgE>ok`mFj9W zMvYUG)xGLI^?-U*J+7WqPpjwD^Xes;bpOfJguHkw^4~v^ryfi`I+@;`hr~T`Q!FcV zPn7WuR05T9d`2s+#6jf{pEN!P*}Efrj`BII98-?-IlRkuR7Pq#|9Ul*fG)5YtKGG{t-zGKca%(=&Vr|u&F&otRG*XyS@jX2?Zhwc=Bsbel5hW zzWh3lPhsXL&9B|{6ALts2`ki%&tg7dg_ae%NsrHZY*DkbT@FN ze5C+oCzU80sX_TgJIX@(QtI(DWgR2MSTRLR7xTp;u}nmZl_F7W5L?AAu}2&c$HiH3 zMckkc>aKVwo{KDDQCyT9N-k1sUM0U$Q1MoLNwX!CQi`8aRtX~AhA54AVriy?l5%@1 zLzGBmHnmo9e|8my{W*wpBZlHoK@{YIpT}^(VETIzSyr3LT-2CUs6HZO&EalQLJR zYt*&s2GZmIwCfnr=^<7uCz^RrR`hQ+=wwS3jtq)UT>e=cdc4 zE2t}?E2;C-mD5$E-*GElH{Af;K;2;7P~ES(5xU=WlXO!!$A8JzbkGLyHgB;*+%HWoTY#^SNhxZT>KA z{!q=QYVjGz9oyrh{of~hxtfpG=1bLer>)$%|~Sa zu6+;Je25kwp?!auy&YP7hW1@($17BuFFkuc?R%2uL$&_}YCcT!(c1t2$>mA7>RlAo z%|bgvqt)O{9jlv3@2a{ee6=_mbWVE3)yaiK>t=N}>FDL|W_42a&Wg!k(Ycz1(Zx-1 zay43bs5YrCR$(yfT=XV4)yb7!lxCx|#UNZw7BlNGI9ZL(I@Qg^$*kzy^s2>ZVs=+& zI>Ngeb#4a5#cHzXRVNpdaMtOKX4T2yYPGm2E=Ikxlba4QRja}5YEsND!stwf$IVq| zF{oCPi>rsU$PLrxtF|&N5v#?rBY^TnZ`-5KRCX8mq#mQOE>?TmO z>J^jOVstX-+$gbicC`>=Vta*~S?BCzq(!^M>_Su@&f4dlbj zZq6#Tnl3uMD@Vo5zL~7fwBmGe6&9V7v#T4cwdzbl&uT43y-wv=lTs=7HD6+I8UI+Lr}jdd7Ulf{Lepk|XZ8|bR)tS;=6(aoZm zo!LLZw#g%`GwIn{3%jFoP!(q@y;6*37dM^3NmX1;db5)Y=TqpICu3M8`!hnwFAjf6 zdpBve_nB1H{`>b@gQRR9y~7dzA=;6$vVC;NYWDdfF zBQ|n3J}oYDAATmj3WMU#eb-<4hP!P|rIu2cC&iXZXQh|&lQNL!#1T9rMk*7PY05%n zr4py?ussAbh^l*}|&Tl1VZNFBxV-4vedqIjZ{_s&G_ zotwFLrl<#bn!BK8sCU%|>NE9~`o?~*%v8Am>b!LMb>2E(U1?o_E=X5dS5wza*IL(J z*GJb^H^hFQoJIaNQMXHXMEASyvhKd_vF-)WNpEz2=sxPc=!|-+-d*p*lTlfoj4JDc z^>y_P_09CH^qutG^nLXG^+WX|^%43p`ic6v`bGNX`WXFs{bv0({bBtn{aO75{dN6K z{cZg{{UiNr{d@ga`hhzeavF*l${H#fY8#pv+8VkVel!fEmCQ85EW9q4A~htuf1}nhYiYAFFTAJFLI+!|{!b}59BTS=BV@;Dx(@nEY3rx|bHKuse22-+WrzzExW=c0* zFx@oWHr+ElGCecBGW}utV&Wq1l*_4rlaEsgr!r3EoGLp7JJoY)=G4+D%&7*^Vs@bk>!)MRoyCCM!9D|cs*H8Id#+BX0ojE_Li*kvb=J@q!F)W zndQd$xvR-?%Zi?di^;Ofxy64tEz2*@Hh#5CmSNs0*1WPT$4ovyb&4#@?6qX=Fj<~C zR4vm%mT8W;8`EEwYyO^DYqu=hOm5{dMwV|T-t7HcmT|h5-{LRJIfH#O7s|5EyRI+x z$nwtpGvaE_? z+&pioEDycWW5X0#CK|VCSan%0dVl@12C{6lTE)nmvV63s?$x)lj8qA%og&LgXD#@U zF3U;-_p~<3^3pslUtEx7rta~*+Q@R#(mn&*$+FY!XLgmB<)=r|&u*7xsI3xfwUFhg zQBS7dlx3-L4Hw^%<*7Y79z7(>RF6-tbw`$~4!+Z4m@HdewSChCS-$#t(~Hisj5VyR zR~=c-Iwvb?oqp&mtKnd_t7kvC+yYp!3@Hp#NrGOY&Pk>#&Tx)pzX zs%D8}rv0hDhqsSEQ|L*?$qS>dJpScO|L3Edk9xOrfY+deJ}-?i3y-WkG^&Srz01cv z`h~$B`5P3-QPDCm|JYIbBJYM>Ejg#?;gAuD$I7SdIhxD7XIb@cUEG%(>{GDL*4j0X z@A>`Ex21pU^?6CD(H)+Cx==N7!11nL?He?nnzwG>TGWQpW`*3ZA}*1a;QFZqD5@w= zE>{&*lm^(w!2b|>n`b>#TDIi;@@%=)QOL}PsD%Gm8B#n zjB6cxtxv`7)1Snq>jMm`D93e75z+34D9JhOe$w42qI0~*Vc5SxO%7g4mMYWwDjLgRu?0n_@e7clB1q_uc~tPxPLFUFf|W8|S?NyWKkl zd&K($_LBEi>_hLT*bm;HvBiB$`|8PMmiIM@F}}iYlv(jB>&KS+b@x+655Jy-`}p-E z+}|&p@F2gTeme1+Uj*SuzuUw=@q31S?H3Z$?AO{cePifP8M6pmI5r@bv=Q4M`)lkd z>~FE-u~TBFVQ0q9#?FsTz$V77kEI1t>=x|S*n`9ziakvDXl#0{NnD7%M9jNb-z>j& zg|n(={qwtaH){JyA(S_i=KlJg@CWj_lnarcRYeeYS1F05*k%hSA$iZ2%(Hd74p8Vics#|ifF60 z!*)==Bc>DiV!i05_9EPy{IM#2pah4~K=R0{7^Dtm38N^>(TPZP4Dl1xso2@-Qq>}s zsr7WRLgsQ+F^pWUUX0X@B4w`CZNP5Sox-M5=Anr5x{KJ$+{sn(l6yH@tIH%M2kN_$ zhP&x)q}wU_z1aQSrB!iSUzRl7%`l!cobLXKH2lRq3;Q9*7t(MZs29@< zDuo6MU7>lT;l+hk6f%lcg|Y|>ZxdU+gu}f@ zVUxT!V)uBbdwYsA-WlF*;+FS4Z@p;dJJz?72=wbpYV79M%TF(Q`}HL?4)*(*RBH3P zL-?uRbL<-$nYakWznDL5*lDh16t!adkxFebi?Kyw%aX=|Vh3P{$Bx8~j-7y=8apf2 zAm+r*B{eRHjVC5K_5k)s>@n<_*o(17aXI#VY(dc~F+8!L7?`-1_HebbI%fT&R4QA! zu&4G`fZN12!s+T6&X=%{g%(Ds1ZPk;0Xs%No+CfYAX~MsGRQXUy9~RrdkuTB`*_|_ z#6d%dainNsY)LypN1MWMV>sap#%tK?#+%p-<6Z0no@W&C*!T?VXm_YM>zox)-MKop zrt`Phrp`^VEvPM0M0e*NSVyZwoAVgLDb6X_JuANOnyZ_uikjwNGw0M?8{5F#5ZlDu6zgc?7;YX&*k&GsO);ln_m~f151Wr-kDE_o z9qk|$i_XH7fTIygMSQ=uRSejznTH0Y9Z6zZuV+ilD?8l~Aj$n^jPGT=wE@B_p zTT&`kGP0tkwJx^4wIQ~#wJElZwJp}s&N9+EhVUNiL2R1!2=<+gb*mR$3K8m>IV+btjRywe3pd!izR>VdHF2XJiT!#G|_!a9zE@cJRis8D^ zy0N;ph7N|wJY^p?q#GU^UKrleGQ-8>Zfb05N^8zk(nd%+Wc=$Q%(lqH@S*0eLMIIhW*)&P^!+O@IM)Kze~sG~vtTEuTMpxI)XPj!>@45#Hy#FL=N3e(n9$+uf&xPoPg1pKzan zK0|z#_(c1p`lS23^bx*Od}sO=FB(#`anZ1%gUkF}=8fM+KY0cA_Yd^%<3GSZ+<&zH zeE(?w&H)htZ)gU1zU<4gnPrv0`GHY^S%HRfcgnpiH+A*Q)oH8KSC@$iiTNW&aP>`( zy&U^G)-NtFu1Z`;+_1QaxI=NLXsetN*DAhi{K)tS+A80Ouappy&@@4|Rqm9qKjCV^ z`vfJ?nCP77n&^?3JJBccC$8Gh5@jo9v3AwPq>E`6k6+BZC_eW46#nV*Cy{B)%#rDp zSuk^d=DExnUuJ(f{N*&CmtXRJ_4)eK*YK~OzbaXKvi4^k&N`NLI_qrK#jGn?*RnFQ zo@af|`tobu2Bm3xlKn$rUJ`tZO zf9HFq{EnAJRkH6ycAp)2Uz1fpa?GQANRF>b#e{vVFFB4Ei=@3J|3{xk3$7`IS=VQ| zPJZ;|duP-Ibwy#Q8|sdFpq{7~>W%uK@6iwFNAwfwi~6DdXaEXF1JNM#8OcFMzMxFj z^)kyOKI1EZ(Z3jQj(hOVOUqcW1?Gd^`{A2ue;%*?)7why!!K`1ddo=6xW_{lHVAko)wv6U` zFxw%QKbrr$5g*L91hXx28Kc<_xedYM72%AmEZr^ig!x`&nb@vi@tXLDc=k5EECHz!Htgpnc#qRdb2)nmY1c2A06yS&P*&% zF30qg^}-3ua$_0tILLD$=l3RFZok}Dxj$J7$4h}H0Z+td>E7}?=RwE1Sx%OY?a{Gr z6YDauE)(;Z_}v7Hl2sGGn^=yC<(N2@GOSzEqet4RDpMQQ@@7S>!OCuw_ zLyjJuG-1@JPq{vQdhw(~hfPNtG?;FA`7-nAx^=bs{QmpqFDqAeeY{~qy}e786!~NI z>Q?gy4Z8bXYU<+s+qMlY@a4;`^>gOb__28LZ#!11)VEe}@NeoLf4pmc;>4Iv6)JrH zcFmd=74F?TPF}RAth34FTFKM1=;DZo{55XhzFGXqlaoJ`DADZXrcI-l4;}hAzq7OF zvN2-{4)ODAACNC!1^@i{D{boE|Jw789k*mgM}POeWy{q~t5y5C+l33;PHo;iYDB40 zEo7_t5z}JpEWshGwCSQYK|#IejvME3x?a7>?jJusGE#YI8$J5#%h1s1y49=yvUB>&+h&9965@Yefspkt*%}7tX#05;)%L-Csy(D3TSum z;GEa7v27xM`|Wds8a0Nm?b-9lgU+3ImdTs9{Jg<~??;D)r7pDD@}6nleA(!0*H-l| zT)6IUrAxO?oH?`Fn28fD$LiOgTI%`p!{@ed|EYS;oF#v~aU=G2mo6yE8Wd z+^ktuR?nMP?wh-J&rSLD*VpAAJUCzH$dM@>zWJvAwtoGtOr17O>3RPA#(@sZcBuUIl<$U|Rm@0!<}G?{;G_wH`vr%Z7f zH+iz#%<%Ax+BIvAJeZt3Frjzv6PFq`oN3T8M(F$Ze|WmO7FgA-+y1TW*d0N>sje>vE0{AD(PpxUfu+hezREKm2eu{PN}aNzAe)Hy<8B?d~KCD>LA!5=b zXD=6*d{_4F?N#pn{flX}YmeRj{rBkyYSoHNii#>V%-_GGhnriW<4H*)KDTZicVXAA zepPel_5=Sl;6E1pqriU}_`87rZt&j-{;k136#RAI9}oVcz&|hecL)Fb;J*m`>w*7L z@DBn13E-a#{GWjTQSi5b|5Nbq1O8vY|1tRQ1^++5e?Its2mbrPzX15J2mc?zza#k9 z0)G|!n}dHR@P7;b6~NyK{O5rGVDN7R{=wjH1^<@d-wXV^fPYEwUjhDZ;GYcs&fs4O z{1=0N4e+O{RI?wz|0MV?2mk!wzYP3`fPVn^`-A@`@P7{enc)8({F{P*H}F3N{v*J@ zCio8o{{`Ux4*caKfdGF$@DBrjZ}2|?{?EXF5BS#y|I6Ut1pIG+|L@@c5d0g1|4-mw z8vO5pe{S$U1pXJnzYX}$1^?6F?+*S(@V9~gOYpA?{yV|{4*0(T|F7Vm3jUM9e;xRz zfPX{q*Mt8}@ZSgiL&3iX_y>Z2KJa%1|9;^AE%@gI|BB#W7W{94|4Q&b0sd9Mza99$ z2LDL#Zvg&l!T$mHmjVBI;2#bC3&H;k_>Ttv-r)Zm_$Pw@81O#^{-wbG9QaoU|6jrX zHux6=|DVCX8u+&b|7PG{2>h>tzZv}dg8yCc?+pI^!T$yL$AEtv_^$^4Z@_;F_?HL& zI^f>{{I`MsRPgT!{%65I1N^su|0nP_fqyaZ-vIv8!T%BX=K=ro;6D-k2Y~+|@GlDf zufYE>__qN6BH%v~{Fi{gFZf>v|7+ks9{k6F|4i_&4gLqgKLPwNfxiL#74Y{2|5f1M z75qPfe>(Wj0{`~lUjqDhfPZE1&jJ3M!9NK6cY%L6_)h}=Sn%Hp{)NGRHu(F1esYe>nKh0RIo*9|8Ve;C}`D%YlCy_-_aQ1K^(o{=>lE1N@JJ|7Y;O0RC0MzX$k_ z1pi^+KLY$;ga1$9KLGp>fqx+QE8yP;{L6s%jka@Lvi38^C`F_^$^4LExVX{@cL+3;53g|Ki|Z3H*b>{}1p#0sa-h ze+~HG1OHdxKM?#cfqyyh&jtQ^@IMUxMZo_6_~!xt`QUE=|3%<$0)J2Nj{yJM;Qs{t zOMw3-@E;2P&fq@={QbZ`ANc17|Nh|L5&WaUza{ur1OE%)zZv{Xf&U%wFAM%-!M_#w zpRx15&+i%d`+)yv@ZSjj-+}*+;C~$a_ke#D@NW$M--3U8@IMXy>EQnY{1d=G6a4Fd ze^c-e0{?N~Ul060g1-v>qrpEE{Hufia`2A>{}$lC3j9OBe+Kx^2LGktUm5)0fqx44 zZw3D=;O`6m?%e*@IM0n-++HV@Sg_$ z=fS@j_{W2PU+`}W{%^tmG5Gfa|KZ@j9{ioae<%3=0{+FoKLGp-fPW15Zvp=_@P7#Y z*TH`X_+{!_qzGWdsse@*aD2LImR-w^zD;Qt={UBSN__*Vq~cHqAQ{Ktd; zKJYIH{>#As7WjVv|ApZ10scRL|7Gx>4*qk&e=+zM0{;o%{|Nj~fqxtDFA4rP!G9|F zuK@o^;O_$dd%^!c_}2#i@4>$o_(y@iKlr{(Zo|4ETQq|BK+C2>w5V|1t271plMpKMMRmfqw_^Zvg%;!G9h2 z{|^2u!G8nzF9HA6;6DibQ^9{5_fe=PX70)IRI;QtK#eZc=S z_-_RN@4){@@IMazd%(X6_%{asZ^6Gk_@4&D6Jor}ye>3A1@b3=(v%r5I_}>NpU%~$Y_#XlPZ@|AF_)i1>^WfhM{NusDFZj0w z|F_`(82o#I|8VeM5B^TzzZ3j_0sms)9{~OZz&{53w}5{d_&)^y>)<~G{Jp`y3Ha{@ z|0&=<8T`Y+zb5!6gMV-EZwUT6@P7~fuHfGd{40WgJMiBD{^P-aANUsp|7GBR3;aKT z|3dKh0RJDr|1$Vb2miU?zZm=rf&T>Xe+2%gz`qUnmjwTt;6D}oSAhQ{@OJ_Kz2JWz z{A+{%_uyX({G-6%AN<|GKMDL>ga0n@&m9{3wneL6wSGIW+AH7uE+wiS|0%KBt9nbl z6Wf1UJ4bh?Nt;iBKJWYeyl&RdCngsAR(<#R;=SevMEV)l;MLG``s* zvRuAex)EnKZd&aPy8^^ByX;*=0+t;~eoIbYyzF~OYFZE_j=zC-Cwhtp(7CARK z@wVDzP`er4cS1`_cWRq@xa@j>A4vk;q^v$`~6G{edTC=yv&L4GZr^wvT$`m*~ zdG?lHV_&{Dy;*Vlhh`_=Ug>pykHP5Oy~3HvUtfA`19zBOwYug_m^sCm8q?)y!Xr?sAC=zMVCfZvV!tX!dCMKc%X zUGQl4g%fo`CJkG5xzFy}50fTNnPjT`UET**pC9?PfqCSLM{c!89xXIh7I}&><0G6K zJ}XmxZMa9FW`jER&CGKB(zfzAw?gOF*C|?gN zJ!m)Q?EL3jmd#q6zxkomh?PFqy8X~(SjXJ?JPVk2&bfWBbaZN+jKb9mdgb@DxH!A6 zJ3Qs}j<~Ju?)GEX%3X9H*Q0hw;RQK-JkrLFD%yD5hZ?)idX)dTzTb*chntjL zR`5;WrfW}Ij(17Fvu{}6+?&fp?Ff(m)Nrmx;hv8g9Gv<&E?@A==u3X1+P;_>RQPbS zMicg(seJkC%41V2S$dm1OX%VgU4QIe|F_or zb4+{xaQmg1UH+&N@NUC_$UPCAmyev2{zJU~`2m|fi=D0(wf0Ets66iddn|c4Vpp}R z&c75t`8H37{>rvRPvaIfAJePy6pQ| zUehi#{kdt;5Q5gA4tpcDuM}xOK%h({q04FsXWuC(ld7pQ?DY+qP%-3f!u` zvBB^TH_t8!9{jq^ucvb6+564m!H-Y=(dxT|FI{sEsBv@r)w;jf-fTEl^WgV8{3;Z! z+%;s?{y_%{=C&5Qoie3S^^&#bxnvDg;;P)(Y;{_4cgv-j``Y&GHum_4@U7*~59!pm zVeNL?)4@L!{9A$l0q~y>{>Q=p75FED{~Yjd1OD&9e-`)`1OLz9e-Zp6z<&n#Hv<1Y z;J*<3Z-KuG{+{5!3jEW+{{i^V1OF%Be+B&g!T%iimjeH~;Qu4|{{sF$f&X>z-wOU- z;NKhkw}5{h@Q(!lFz~Mj{u$su7W@l<|6K5I5B_<)|1#h|8~k5_{|fLw3I6B7-y8fVga1h41OKz&zYP4Fga1nK z{{j4SgTER4?}2|E@Gl7d7VuvO{yV_`2>6c!|8K#67x;ex|K;GH0{(}`!e&F8({NI58Q}9m*|GwZK1^%DFzcBb81pj>C ze+m3wfd66e-v|C*!M_stp9cSu;9ng4SA+ju@OJ_K#o*r^{P%zn@XrDMQ^9{Y_)iD_N#Oq+{40Y0Gw`nt{vE(S82o<)|GnV<82rBj z|D51I9{g?KUlaWOz`rZ_4+4KH_)h`w^C(@b3uz`N7{6{8PbyDEPkw z{{`Ux5&XM?{}S*Y4*ng$e+>Bhg8vWTe+2v|f&Z`IUlaTXf`4=HUkCnc!M`f_HwOQ5 z;2#A3qrra|_^${5Xz(us{@;WD7x3>3{vW{q9Qdo?KLh*|!M`NFnd{I`JrYw*7f{xm8$zX$xgga23XZw3Bs!2cHbKLP)z;Qs>rmxI4M_)i9Z zBlsJ@e>eC~1OLw8|2z2S0{=|#e+2$@z<(L|*9QM7;Qt-?KL>v^_`89>{gp%DKOX#_ zfqyvo4+8%z@UIO1=fS@U_+JM981Qcm{sG|M2mBv^|9tSD1^$P?-v|7gfPX&l-wFPu z!9N51y};iE{11bF9QenBe_rr6f&X>z4+H<=;QtQ%tAYP-;9m;-)4=~T_`d@GYv4Z) z{0oD>2ly8S{~F+59{g8;e_8Mk1pk)ce+T??gZ~cjZwUT9!G9|F2ZO&K_|F9YX5fDY z{8xg%1^jK`e*^rdga2OeUkd)S!T$vK*9ZR`;J+RG{{a6D;2#10bHLvp{5`=x3j9Zb ze-H5A1^&N)e;)AP2L6k{zd!ij2mjOH{}B8agMTIPcLx8F;J*p{i-7-J@b3-&Dd2w$ z{L{hz68J9!|101h3jXiG{{r|Yfqzr**MomD_)i3XZ}4vc{*S@GBKQvh|61T*5d6=A ze_QZx0RG*;-wOUY!M{5AmjM5x;C~PNH-i67@E;8Rr@;Rk@IMLu3E)2f{I7!l8}L5} z{uRJK1pE(xe=+cH1pf2DUjhHk;C~nV_ksUd@ZSploxr~x_`e1JTHx;m{w2Ua5&V~e z|0nRj1O9>F|1x8PqN{Lg{^4Dg=~ z{{6teCHU_F{~_R?0scq9e?Rz72mkiqzXSYlf&WDC_W}P?;C~bR^MQX0@NWhFJ;DDc z@V9}#6Zkg-|9aqm1N=vT|6uTM0{-6M-v<1jfPXUh9|Hexz`rE;?*)Gy_~!=y)8PLr z_?y802k^fF{s!=`0RA!H9{~Puz<)CMzX$&@;2#G51Hr#P`1^wYb?~nV{`bMZHTWL{ ze8{t@6`2K+t1za#j&g8w-1 zUl0EN;O_?hoxuM)@UH~^X7FzZ{?EaGG5Du~|26O*2L7JlKL`Ay!M`y0=Ldgh@Sg(y zTfx5&_`d@Gq2M0{{-eQP0sleZ-y8fh!M_gp7Xklg;C~nV%YlCn@LvG_Y2e=&{C9!> zNANEN{>#9B6Znq@|9#-U8T`Y+e=hhx0{_q8{}TL1fqxMAHv<34;C~GKe+U0};6EJv zw}JmU@UIH~zkq*N@c#|`r-A<^@UH^?2f)8G_@{&adGId={%gTM5BM(u|7zf09Q-?g z|03{j4*v7NKMVX*z&{uGKLGy$;J+OFlfZv9_y>dk8t`8U{>{KY68vX^e=PXx!T%}v zp9KF|;NKMdUx0rC_|FIbGvNOt`1^r>FYxaI{!_t!0{C0NzYq9d1%G$&zX1Lh!G8t# ze*pg+;2#hE-N3&9_%{Imv*7<4{Bwf;VetP0{J(&I4e+lE{u{vmd+;v`{;R;hAo$+~ z|LWlH0{(H}Zw3D?;NKSfkAr`C@b3%$+pE&woYtZCR`gwI{X&~ndyD1l7WIE=u`K8% zO}jz*1TjA?mT6l^TV>kTDhIG9m6zDJ$~){Q#fuiCq4d_K4IzDEX%k4Rb=rYzt=6X! zp02f5ceMDuK+F|g25o(B+uQmo^x0A98`&J&f_^%R=t7HMS{3RKVh`(&V2|rhVAJ(a zu+R1Hupjgvv5t1^9cU@YzBt+g(^imHriRy8Ut?8ToYpcn#tt-Y!|pWh!KNBdVbhK0 zu;-2EvG>hFTm#$Qxf^zz^Hi*CQ%HXS=QwPFa}stFEevTZP8&l-TywsT zz3F@l`_lOp)hW^6KT6KSPSyTl@bo|UvLm8}w4m!%}Ov?UN*-V%hZ zL`y~X$WjB_$}y zmSpTU+A*?sRx52#U9Ij|PiskRX=@O+BCQ!|32g0#9cP`2oo=0poo$W7CRjIPlWEzg zh#S@$*bM6}tii=fYgJblcdVyNNo;8sIqpQ5OE;|I8bC|bK-WNQdDkFph-*Ep%{3A` z)^#j)g6l-=WY=kQPMzzzfN+%Sa_kD%RoFGIv2@5xa7`q<&UHOD$u$YP$#pY!H!USe z53UEXA#M$5t=h=#2dtxKM7sNXcZF9v$(K5*9y+YS!-RGAaK>6ZtXNkMPi%FM5D!&+ z>k*1=?a_vq4jvt_T|K&CdwBdnd_Rx=ga>&H#t!vZ=21$-c_ex;LWjo&Vzzl~$L{pl ziB0j?kA33t(!(G=dVC}-=$p@075a!3_EzhTz8JDKWOgr%P+G0az8K%pcD=KAXTn|W zZPz2bw-DZH?}g!bm-Bw_gT$oSdt{{3hLSeh_I?=|_I??!z26e^UhADv+?O-rTLv58 z8-R`Qjl`M)O|;kb4J?JN6*w2WBJdN|P|mL$Ti9&Pax5(xvH!oC=xs#58iNQd$B=U6 zor5aS@zzqYl2v?DnV}-(=-`gs5X_KaO+YsPNGb??!$muv4n-_veUHd}Oa z_wdZY$Z}q^U+l|Bl>@@_I5qV@2XxA=a(mzz<0qHwm$M z%TW3(ZRP7Ww0(!zxhrM2{a>;Ts@SvF9;Vyf|H!m&|KHDj;P0dUUChD1pW#s2VTKd= zpPG(}W5Tgr$4_VrIQcgVIz{K`e>faE(f1kYxbOx&SS z8x)56pa`@R2`BD6$O%{TRt6NRGoC=4w}spvj>g5D#+(^5qwpTXLo_Gl3L1 zns6*iKo`*!bQcLmwQnFtwnr;b68eAyBiuV9LI1>hC=~TY;b;b$jkY4e7-p4FRn!25 zB00uc7+Q$3$2rU3Y4C&niI91DW~+(hm}g0vQt_`)7K$pua*J|YQ3R?}j2u>3_7}+) zmX@smU#JZdb=lAQ`0v^8ABjhD{Ivc^j-8f{o}riM9TGpWKS++JHVlnIGmw0r&LSiR za9u#n(F8ODEk_yXE_#VxArY>KKPKQOa=n;}p9a(DghrspNStMV zQ7SrzGSC|&&e0nXg`r=N_?`8jm73PzV^BOYpI3wns*f6@H5bThq2m|ny^E$?QN&`j z0g0Q6D1?fk(nw^`7aQeArBNX2hh}IJw-n)r0#OYVg5-F1%}{eR^ENp;l#1loc3GOz z?>fe}%XmQV&WA9EE+F|Xq$VhP%)2gxyP|yo5^>h~}VZv>t6l?+{(;#X58!sjs=lplT=tHAJDP1qwsGkaz=^ z@5!N~r|2d6jOdOpenunEWt4%QpqHA&2Xc4F9W_9qr~?W^6VObw4ke+DNPJX8t55un z`k`<%5=EeiXeK&@PNAzP13g49(Fa7gY;g>|L~YzL(W}pytbIzP#pxG+#mU6*XT;^WSK`vgWI3 zzN+S{X}-GVYiPcv=7TjKqWM~yudVqyny;(*dYZ4V`39PAsQE^kZ>;$yn*UbwO*P+4 z^UXCMs`(b0Z>jlKns2T7Hkxm%`F5IbulWv|@2L6jH1Eh5|0Qcop4dKh)qI%dyJ^0= z=6h(qr{;TUzPIN4Xx`rNpD_7q`y?~haw1=CpZaRPpXU2(et_n~H9t`EgET)_^RnNc zob357yzJ{Ir(q7qC{B(^X2Ino`})aggu^icqhlJS`QJ1@TJsT_w`o37^J6qWR`cUD zKVI__G(S=ElQchB^HVfGRrAv{KV9=PG(S`Gvot?j^K&#mSM&2UKVS0;G`~>uQJP<* z`Nf)FqWM2(lx5U%Ek0WFD>T1S^Q$z!TJvi(AEWtL&BtjzUh@f>Pt^Qc&9Bq^dd+Xp ze3IrjYJQXEH*0>2=94wQRrA|4zg_b?G`~~xyELDoc}MP@th@GoujW%Vzfbe~HGe?! z2Q`04^J$tttob9FKdSj-nm?}j6PiD%`BR!dt@(7#pV9nT&7af!@0vfa`3suAsQF8p zzpVKyn!l>~Yns2V`5T(Qsrd}e-_rbT&EL`dUCrOq{C&+o(ELNqKhpeT%|FrnQ_Vlq z{BzB}(ELlyzta3`&A-w7Tg|`I{Cmy+q4^J*|ET#-n$OhyXU%`n{8!CqY2H!uT||=31|E@o|<&Su*2=+xcsBhiB|CI zD#okjIbJ90a3(WS%U^0~!VD&AX(kdQM|4oc3wte%kG+1TKjWgwS{g^4%sr#KxNm&U zFpMvZpBP**)A*TRvyAR0-tc28OP;?eBb@Q{WjbccD~_8U6Mk!Yhov96lTj#)Y(`$s zsi>1#lynMl$|GtywI?Qwks^3LbDDsi#E20*V>%th>b07mQ0E+4&Cft*SsL)q>wmU7 z>nSUc^*_5AUBcrZ)B=5WHkh3$KlmT53mRn3DF$nGLH|iDP&(zz8RlDNo*m891_|T8 ztP!fH)d+R6c#F=Kt`^3*vDXLAj5+0q7J4Daw9r##;iwmqHA3kYgSDbC z{)2iUStC@#+TNO9bYP4Na{7#Mp%dMBtEBh8sUzCW$QM4;6lGXdam)J3>dqTe^)8-& zQ(sioh2^_6bMX+ZUD^`vYOggK?lPM2I7Z4K|LL+3yV+$IR*ssX6Z;rBL)I+)xlT!r zjFMfaG@kmR39gfgnfj+1q`1GTHAx?*~IuBl#+Pv!KQld zXU+qR_Ca}^=LPH~PgzFenIjigj{ZS@CC3b`V zAB|m~e*<=B{(D$aphy9(4+WZII~Ev)b&Lr0=dsNm6?jVg^8#0 z`HCnc$A$XmHEeRUv#@_qx3-@Vp$`1Vb!!8CM*FymnT%CJZr0c2YY=AN9KMRk<(t=6 zFY@^oBwWq8Q7lX(Jgm?XDlWSjL*7OvmpH zR{UwKDLK*!=g~jH-$U3K?S%8_AA^naUyn`l--_MtzXzM@pN2i+e+rxKe*t^R{|2^f zfFbaLVhk)7SVok$kKa=*uzH}Z{hN?S6a(O*CE#CA&j0UMrp8|$c3mE*+7lA=tk94khaI9#@uBN#I&b!EgD z$~ZHVmu=t%XjZi5|epP z!BjErPg|IsQ~guapMsA3t0VU+^R9o%x6T#wc;A5=BP%ZyQM~s+PK(8ofAS{ErMwqG zR@}-rQm&wib`>R6g~V#H#{NHPyoC>AH2VqL#MsY{H(zd|M5`pf{_Wc@?SJ?`%4`-}$fpRccG{8urGlDwN~G{U6glk?GIyJb z_EEQcfOn)E5@|wK5&z$B|D8u7JI~L7{(pV?cW-H}CI0#?tH^WH|MlCyj}ngZjqK8mzs!(b&LLyY@ZJ+DhInU>oX(Te7UCkQmyy^Qg#4RDqCuG^g-)G8>=G8o+dY+(eqxy2E8^mWZNmm(+pj)qw^7u~VT_*5BL?2KaH&z& zvN7t8h#ogiL{FMzYczTC@LAJ4D18n2Qx!PDB+Yw}&xgeiy;llj3sHg(AMT;V(3i>k!Li0If@S@XBHN<@U*?&vC!k#hTDs>t=vn-^r8za+@Ed}V-b)v6%d z>eWHE*th^&LPC%&aczKY9S<#5;bWUME6A3V6kyx9G03)Qa{$LG$dZ=RHhHofZIR_~ z46tq56lB}X@^7A#m=f63By^_MRXVqfHX&)46)`Fj1WTjhV8_%7#5 zt{=7dWPQ#R+h<=E-u{1CBg^uh`A?j9&wt94cm7kg<<4Jlul|xH&+0E-YOmep_RIlPBj2U}U3!?EfrSpyVNaMvos~E@a|FxqM&Sy!rpN{n2~((qlyw z4B)&*$#Xh>e6fHD6Z``vPMj4ud2-&eQ>Hlf$?+cp=jhMttH}E5vA!CATHod$O~Mex z`H7k_LqtbKRS8+NC@OHtlA>joE^QjH%zo^O*tYL5aL(nv7qv~CBDbg5pW=$!v6Yj3_Z}T4agYvnQ2UuwZ71B}@91h>tHpx++OpIMjr5 zNZLBpBsuvV$6Q2i+xDgY_U$5a$BvS=oumyq>tSEEgWA0%H7d z+kyoghlTQ-8*I_hFq0AhA98!(X$!|^XAtK_rhiAgk?pl}6B%F|!@iB16aX(lwkc$Y zoWxSj^_l>A?TL(B7cfTK|HQ-q+uF53wso83b{K34iHTjPRjaDhSiL5s zhP=L|q?p*=_!`kG&ed4C>Pw9TUZ+d17ehRmFWTwAo%BJr4I3TDM;?FqJ=%UA{JGYZ zwM_8umX#2JX*b?p6 zKm*q=Ilqo;rSt~5Y>t~e#(%jUZ`;l_i*1$Z&So>%;Kx3%yoMWvxW|9eY>UM`jz`+k zZQ>qWrR}gE1CbUnN8ICjls0dJxR;b9(l%ztPZ9Sx-f7cf#XatmY4OCf?`aWZ#69*u zZQOEkFFIPJt=J{*aZOFzo&CQB;vVTfEo!s42PPJ&sq}tVD5-{Y_h!9X~n7D* zzcIG%rY(ExCT~4ncl-8>b$9G~Q8y*!UEMuQFBOr?wDZmOa!dvHSP{7i?qwQ~4YK7uDLN*= z77I5r&FG1rG?itL2IRBER`x;8Yn$ru$y{&rq^o7iGn%YgmCLFSjqteyrtvILJ0}asX*F$ToE<%bIDh?@Ofpnj@bt9M>zECX@>PepIsWUy`4u9FcF zI{STJruF1x9cfc#n{+nGsJtI%-<#xm7;A_7(z!p~mx63DT;Gydmb{KeM(W@NhL#oi zk#rqJdX45GoLl>Pl9S!aZr^UO?PmG%zjE1Pk&)c1IIc2|YpUJb?DrMLzF+ohft^>?o(3t2me&-Xy<8eBTjte(@unI?M|Wtj zVnv?@D_0I}uxizq2CG;5aqU%XvGL`&Zu!~PZ{ohgId+!!5S|t7_YTf+!aCmuyLOdn zkdiX1!S3Db8tmD#yn*8yxT}HOce(D!v5xCW1jlvaLayyAf^4gJ7L2i9LuFcX%(s5M z!>_epBe`a-3$U$U&$X5BGL6l&KcDcNI>~XqW%`%bu75Vsu8;pQ7j)CMQwx{G%bt-k zzB;ys77oVC_VO}5yUn}|M-U_1&D+D&kjpmlGMwEmUWR3xcyA=bvR!4xXN8@F?dHG-2cK);$7J5hfc^RMGc3y_F+t16e%wWnk^fH{?j$Veti5srP zkHk;b!c+0HweVd0JT1Hczd{SI!f({VoAL5}Aaa?z@UnfreZ59U8+{oLAxkUU>C12! zUbfYj;RyUVEj$6Aq=h%(WjlWRa>)M5w)`?I+w+?dS3_aJSJdLa!OM33avs^%U$*nt z5e~t((c;7KKWpI#yzB=c=g+{u*1~V`vOj=*y=1F%AsP0;%YFeeoZTlthW8M+SBsZf zaCa%=&{x z2QT{&$Z!N+_9Kwt?7jo`cry2!kPK({A+XOwcE1R+&qH=S9?5WaKLYza)K?ckGMwF~ zz?pD4;_7MfvVVb13(GzR2eoi^KLhz+ojd(VkPHvR57NT2-@zE@OkXWuJ)+TKFLT zkQSExC}jT$GAqJ{UytPcm+)7$u{lV@ z55dbm6*4TS+b@G;I2wOb3upJQknysQh3sD;&#wzU2a@r*@VT{c9(-9X9E`7pzyI&|d7u6H?q_|U{j9au-e;eE z_Bm%g&mw$KECve!eAu9s0OW5A8pUTwp*#}wGQz$AG>X?i=ez_oirYZ`C}x8WKxHQA zUPRd$bQnMh_F14&9EZlg9mJ3TP?-*zL6ljbO^LDxXcW_-4?bf-Pay0QK~EyeA)uEL zWfc2?Vm`c}(fgp^0#N&;_V9}h0G$&mXg&bhg`mxeGKvYoI>0f59S4o#L6A>$XKgG9 zDx-Li*8o)3#e|@;iv_%<0FB)p${s{{k~TgB*-?y0AOMwTfUYLW7eJ#}5l!7G@HzrC z{-E86@;J~aZUotNu_GFvP#6UO(A1r-jUhqh0Z?8AaD(z{&?uG!*-<>neWLsTG>Ru_ z2X-7ZiYw8ylc0?)(Uezc<4aH(#hAPUpz;UMO+@({=S(-b`+C>;!$`|o&Xxfr69X5HU*V+@hKX6l{Q8Nm346{sH}@s(fD76_z3_i`-5Ij zl(Rvj_!4ACF(xR!1ig>)K%*EFR4xEr2tYnXp!X8xeW3RP&^1#H`Y>TX3YvOF>rVrX z;#5#w6ssa8%2Lp6iLwD`8=`Cn+L0(bgLWs%x)>LDs>S+(Z2)2S1C8QdQ2QwM*aD$u?}c>w58qC5k10#Qx`y_F~zfUY6RM?j;P8`K9~ zoDC}LVr@`a7hj|Cxv!0_L1il_qxc$BMzJ+0uErHeD(D>mWJfVJ&xtaMuXzqYb`)dt z7=X%8K|d$Tzk+^AlwX7XPL#2y+L#>VkK%FM0H}=Oa!}mOZaBF(XfgoVsi4~v<&K~Y ziLx#}2VEP+U{exy3(!_X*#<4?reXv*-;Lh&}J ztc$tP+QIN2`5!$4+N-3w3uFw9bJ31cL00rEi-ZYJ1n_Maok#+zB&8QIV7>>?h?*iXj(fEzlqc?Q_ zivM1xEHM?Bt$t^dHP|3^1!f0Ncn(l@1UO;Nux|Y?x;M-xj-;<TZg z7~Ze&^EyZ=rW?<&${{M38Sz?^cKkD3h>;jCm`FBX|zhftVav!?UMa|XFIHPOu zF1W5S4IxGvHE(ZO(vCm61M{i3mb=(J?0#$GKj8JS?L6RI3!eu+M=$A&k938<(b{*{ zV|XV$`S}j}?K=;Rd0G#np0rl&J>mIJ-xqMizhW;jG$QGZruDa)b@S)Q@0H#`t@nTr zFd7Jr07Bzpbc=R05MpRV&_9;%|9JCXzV(myk4E^S|Fn^?HEXnLeg^t!`rU%?cdgCF z|M~4dV8j+$sG0@Cb^PC@o6p7XmHznrNPe%{Pum~Y|LOX>g+HqQyOsX55se+95~ImE z#5iafJjHgzI>S?|3B=0x!ptxW3_X!rVb<{MX%Ekvu2>)VC9yAto+1ar_%!sa7zj^* z!7%DP6rSKB;Xjvn7_*%QpX=H1iCX}ls^u_>vIai=&cWEqn{cOk45J?3!u$6-M!}gl zAD7~7aYNh)?}c084!8&25BI}|;eq&gJOq9a3d3jP;rKi}5?=tBEu--mJQk0`m*DaE zQak})h9}}l_zFB3Ux}yUtME1WT08^K!Z+YKxEjyJH{*E_Y*K(1;zf8dUV`t!cjM)F z1zw3);nny-yaqplAHz@Jr|>iQIs5`%i(kgC;@9z;_-(unzlYc35AjF%6Z{$e9Fk|e z!r$QUAe-h#{4?H!f5S;68i`F3kffwGqz+L4M0H*v z-6Y*3JtMs$eIk7!eId8RT5@R&oh>7kM{%4|y+nA9+9d0QnI4F!?C?IQcaBJoys&8u=FaF8KlZ3Hdqs z732l|0YfhJD0B)FvV(FdJc@`SqqL)Rq8L$nQj`=ciao`J;z8+489*6A380LojHgVd zgi>Zv=1~?>Vkt{0Nt9GdIwgyurfi`UPV8^d#jRC#Yv&Uh~V;>(txSd(?;2XVjO}chnEmkJN8ehMq*Pqn?SLO3z8pOK+&& zIK8QQbM&J067|yca`d+96+tf3!+Pf-bLm69Mm;KxOKV5#L9?d0z})DAXrpPNv`AVU zEs3^nrJwkMwikJ>1K3Wx;MQ)M8=MVxY$|r1@r`Z zDt#?oP2WZ@rB~CB(=X6((4W%Z)4$Le3?8EmqZ`8%B4u3}zKp?)QH;rq>5M4GQpR$a zb$u0MEhCGO!`Q^gXKZKeWRyW{Y&GLB%)Wk>QLCAI{XWdS{+#ie@qzJ$@fBua$CxB0 zg{jA6Glfh!(|~Eh?7{2>c~X^33&@aa4KuSlGM!;|c6X)+%+Kz_?8o$F4q*B*2Q&Sd z!zJ9$^~`LhnwiVo z%*=yass+qKW)WmlEoGKL9@T1Q4f7cD6!RRjmU)$VlUc{CXFg%RfZ5&~nN9E`0hPss zIp4)B1;>#7h{}y+$FmdJ$?P|%Bqy8`Bj7Rb0%@7a%OO5bLMdtaH2VjISHI3P6}r=XB}rfN6p#H*~%&8 z6m!Zr6`U%F-#x-P!8yaZz`4x1&biIG$9c$k%6Y+g&3VuH#Q6%3G88VI%jWXAVy>Lq zj%#_hp1m>a?k<%V&?xslu`$SE5Salgsj zG;TUKlbg-W<>qkK>Fp5b2LUglnhnBRNchuo*!7u?s}_uNn1ukbSx zg-7SHd3>IjC+D^0b>Ma88S#4XOnDYOYn~&|o#(?F!1L#g|d_G^wZ_7928}WPb zE%>&4XTAr&AK#BZj33A!&kunJ;&A=~ek?zMpUhv)&x9!AJbod+lwZNG<{#mo;$Prj z<=^Jl^Plox^565n@G*fNr3tR zh!!LWQUvP+YQa`Pv7lUVP4HOoQt%2g)V>kC6}%I?7km&j3O))xLAKg2f+oRN!8gHo z!4C-5$AwfOL&y;dgc6}b2(uCh+Y37iI|;i8y9tej-Gx1cy@h5%bD^csN@ydr6FLZ; zgf2oiVIQHV&|BCSVx9X72MPxXhX{uXhYL}p^Jw8X;Y8sS;dJ2);Y?weaF%ekaE@@U zFhV$AxKOxA7$b}mE)gyjE)ymRlZ7e5G~sIDT49DTOSnOpBitz5B-|p*7j6@77w!=5 z6qX9hgyq5tVWqGN;-pcO^bz4P;Ys0X;aTAYVXg2o#7bWm-W1*z)(IaAUkbkn^+ZCE zfyhLp6xoVAM1G<`QHW@&C{#2<6egN23Kz{2MT!=PqD0Z6SkYopyeL7GC|V&(5v7S% zi_%5wM46)XqHK{`lq=dS$`frB6^IH&MWSL+si;g;E~*e!imF5hM2AF&MMp))MJGk4 zMQ26lMHfYvL{~-EMK?vaMRlTkqI%Io(PPn5QG@7(=%whj=&k63=#!{P^j(CDDPo$K zDdvdzVv$%XR*2h*+lvjwUBpJ>?&4lzGqHu(N^C235Ic)K#opq+;$OrA#D3zzVt?^4 zae#P?I9NPcJXJhh944M4o+q9!juOX+7mJsQ6U8gUDdJV)ba94wy*NjlE8ZgBD&8(G z5|@a}#Cyb*;{D=-;=|%&;*;Vt;`8EK@fGoP@hx$k_`dj|_=&hd{HwT8{6kEa2qgNF zPLduHONpbzMdB{;l=w(~kqnRwlK4x8OGZjYOM)a5B$FglB%zX-lG&2El1RxyNwg$R z5-(XMNs_FTz&s<8b&@Pewq&DZvm{?qAlV@)mh6(0OZG~tBnKollB1FnlGBoNl8chd zl53Kik~@-nk_VEiU50(v;1;|Fp#>mFWf@PB- zMt{02Og2Y0Pc~l`C5w?QmMxVf%2voyWUFN9vJBaJS&l4Mwnes8wp~^vE0LAS_Q)z_ z`(+1Zhh@iPCuL`3=Vi6BE3)gdTe3RYec40V6Ip}oSJ`XXJ6WUbv+S$vhm0hr%IR{J zoGTZ|#d4WkU*1mMLEcH;Rc!BOxPB860;P_$LFR~Ra~D2x={6}=Q@3JZmm!dBs+a8|e}JQUuFev1AI zKgAHmFvSQ(pkl0Iykepv1hN;*P|Q+gv`p zDiriD3d9toV#USIcuC0`th96swrf`eURJgU-@SV|4s%dqdn%aVx6^Iix-z$n%xi90 zS!x)^dDv~k1}j{x&cHTq%*S#!Rp6U9|0izcKXEJC*pdG$aVunOV7Fbn{x)`nj0J8ct8z3}RrMW%ndPzS>IoPc8;wRIM<*(&sj0W! z)~;=bZQiU_Y}u03HZKoG@$EbW*NmlcS=oth<>gMsd-h~1s;Uays;klce2o6mrK9u{ z5;iE7EjwxgF1lWq?I4mV)IhLB5gQcaFV5?TO$5*dT#n!A@i=}65#4Cwp@5kw& z|M`bv(JSG2w{?WiL3Hibnj|F!L!Ps$wq<1j`n$`KO&iNYf`1pQs;0r~g>2fm9ukbD z-nmmZrXS)mG_e=e+?*WbVTK2zZui2j}UpdwQk?_NYLFOOwZRO*4B9;Kv&NhvK|05L;c zN?BPf#0?24d-g~v6&3oFy?c{2F+?anLl*-?#;%jrudgF**zns}AQ-PDD=yZ9-)ueM zIba-CR(1!gsCWj?0NQwjza~zI4g0f|t>4fe&(8Ld<>chy5LXFDPu0YK!sv@f(f_Ms zhcvM!Sd8YH_)koYHqQ00kEub|54xBBC$zRImK(Nyy(gBP(;v&t9fEDi8-;BvoQM_ej>F3LuE6%~OU3qU z+E|N4$8E%xtaO9>CVYPY8y&_tM8~;dic0*i1iaR#soQ*A0ey zT>vRNJBpN(Go7SXPbF>K_z}-7Ws`R8G9i^!`om|U6R8U7bFjXo=+%DIHEa5k($fc% z)~*d8ty>pG%E*{b%FLWf%F6nPuPo0Ok#LfTs751$uKB-CFHHk}lm;747ytS>1s zaWHB5@&Hm&QWR;$is_`}g>v1<9Ys+^C7(VKdeocXx6P4{=FR$oiSclLPD*+LoO)Lo3XPdnp! zY*O^Q{yFi68#~zS>?YV$vL$^(&Hm4uX(Qsh#2DMJH3_-z7y0p)`R0)9lZker9aqOK z$QUv3=(gxvlRut*_v-CoGP~oCmu3%l=Y1GtoMAEO1ozoR{#~Uwq?dQe#o-x25BCp@ zunxEGmoaf%QPz;LmnGb}>e^>V(hux^*WqSJd{F=518Z-2NN=Wej9jCX+b^9tpl`}e zh3x=4N24Xj-)>2GU_STh1w|q2cy-w8HlJop?t;Z{?cXR)y!f%-tjlHdK8J2j&hKV6 z^iA)1dC#-1Fxzy-dXIS4ez^D1n>IcBCA|HzX6z`7SH=6gulF~oPa0q|W;ttLt>xeu z;gVM$ufO@gObKjfb$QW3dCwGsi0AiKE|_n)Xx+Z&cn~+jdggt@h{EI;cjrqt2OPaR zOfNG%?9A$l1B>???|R@ea7n($lgbi*$*%R@(FaGyRsXLyfelT|Ybc^+TqU>#k-=MUx&K1x51qQ6`^SHA6 zVY`u%c?~Z@pB4@s6lm;A-?yWGW=P8NljGbQL+As(eZ8vqR=9+`BggjQ{VkUt?^{pK z+GC!>_-YxzUfkJm-i`D6SNc?0`^p@B>jvla{xpdDYS&Wc7ta03Yu)C*xWQ-;4Cr^} z_~Gn+G5Gq zLYpcly?0kJ&s@7;?B45^S&jIT!|IgGvu*1y)~q!8Fxe}3_e!bnnCVCFG=3>5-u`f) zOSirG=bn0|zo;8qy-@C&A8yZ`=k2^=ZnWs()EW=>nQ5#UPZS47cutT_yFRAfJUaK@ z%nw(d@pw5yLlevo1ciD&F7x&K!Je74CBr&;^jY9zBBZ5uzk`!?}N2hzwU9d%Lmkb>OJXr4*vt&>FCHeJKA3AI&6wF zgBo_|OL`UGZzlGvY1Ji_<#!6rr>W>nXx7MM3g_xH)8jjrH!PcXEo1j$(g>g-glNv%#GeNqpUipq|?|vc^wSnj-Ktj|5a+IuDiMp*fYQB0;PTC&bAI0 zIzB6^VtjtRg_YcP!>9C?S3_&=_4PiV`@rH+5SYM@eWf`W7X63 zj&Ho+m$UJ8Qq$C6lksso#uc|)w;{yyj9DeFQVXg)&vBU4V&d$vcoiv-)~;H zXW1gX!L;Y^pLS?-PZ1S0k!*2cd z+)bPAU)g?L;X%v6M+*vRE2~ah^An0k&#J7RI%K!o+sipm>`ZgSGGjlYWs!NO+GYda#cr*m~i&{q>ra^BtoB{JN9oJIW}!57iL}2d%DB!#`1>AZ^xdU z=dg46(32Y`EwbGvH@Ebw-rsgWeostTl9l8>ZA6*b+u&2d6UYy(_S8Qcm|byxQdZl+ zOFd?sv@#yG$>^@p?qc`Jb4tH2U+>$g#?7V1c%>d`MgNFz?-xFj+@16~g~~X%V*8GD zJ09w7UL6x#GQ`Dc_K(ZmOg+x^p}&|U=%jba=J3II?{#=l0C%0gYTei~`cwz_?GTK{}Kgf3aMvpF{g4`gZ zlYPepAG$hIy{hV5$d9yv4O6#8E1e@d2L#8gw37#ne|WGqx?gw0Jlo1)zc^$s8TY97 z>1A=bdmnxmuBl+>#_ebC8e5!t^)w@=@ZG0Mn}muZ7k}MFzB`KXYwYPGtGhOixs-HZK&}hYfn8N zWw-s7W!9YwQ^L0{6`zVL?>gFbP-D2-xx4)+i>sf)5}%Eq?K01|!SN2M_uP4Yu1^=O zbq;FRJ?*mo>n7s{-{mIj*PJ3*58`iJaBA|6`JIQ3^Hxn!mgbKfPg}5P{=^5&El$TO zu6x?dnEh>!I^tf^ZFNcePIE&~?r}9Qmv)POe2uZq=IosprxYV5brb)xCpEctDm?qy za<~yqhf)RQZ7TL(niwUhZ(PRtntsXfxp#3>VN{&qZ zjIY0TsgHESdHd@8-o}Jw_{O>R-Me$l#s^xZ1YF<0&`=-^F+Y8F#;3{; zT_%qUoVmDfmkamYJkvVPopY2+pLwHc81G@A?0(gb5vnuu9M&jP;=j`bQW{r!vheQl>t^5aTUQralcT4R*D?(!ee9^0 zMC(S6JhS!c@G)~dKA8+{Fgg7B+esCk+O~d)NbvoRg_*g zalg1Psvr~A)_V_Z8j-XytRGgfS7aCmo08obFDplSbWs)VWr;=JLIPvCAJD zZ=#>3&+$3fy~_rrSDxjkC0i9Q)_%&~n+-4ff?lw1}OS zBW_XdO*z-?`kN6qj7SHv#_hV_W&VA$qAq(EuE}vt}}L+ux8l% zlrw^hpNzI|M%Mrw&R|u zlk=~r3QvwmxO{)@_lVsI6;r=Gr1dlV7ssl4F@>Id{Ok0CYX{EB{zQ3e z@sc{RbHA?AE3tFZj@yeo=4_sSe^-53!N@iOr%i!f+{Dskhurls7uHXJf3h^tJ$GDUrfzqCv46Sq~;8%D9E-EywJx$JkBDA2#ca zEqAunypHU#sg!`q`7@sR9<(U!d{}rsX7<&mcg1x=+SWr3)Q1y-sAoOWJJpZ+zGTF& zFw9@P#`Lkp!TrxH8@82<+|uv7TwFM6SlpL(Pa?}7A2_y7vUp8&NZqfO=wtI{M|&n6 z!Ml1JHNE$)D10&@)1dv-(U1nF!`s8|Q5-4ct8zUe>9Io)aHV?NibB7-9@YDe*-h5c z3!z@Q!v^H8ywm>K`~m0X9eqyfO?%{B{&nwLMK$Tkt2-C?o2Jc}^kk9ZYT1Ty&lg{w z_H64$>F`gn-yEWIJ^GEm=FZf&4Jz>4ypcg#lp7T z(|+t9aqGzCYX+Uw>XeufPDSU(MD=NmXp?FnU!7z&QhnutdGXA=Xp15Fl}&+bBIaBx zK6at;dJVU?3qq&Q}npk^0Aku&utjx zXiK9z?VK6f?t=p;S)NX{M^QGVBpMa-6N;OyQm_P?B}eFd^o{po}%!k-}8^N7O(`DXnTy0R5^8< zlE)l8C*VX+Q)%5pf98x!&L@71xbT%cXzA62E)TZo{aU@3vL^ zK;YQB`;*64%-!aX+VLPX@PYdQ=BMX&J*?7=zc}<)_BXz?_zK3zJ0ZJwiMDpvd8@bn zOvXj~OKzjaJUSkE|BJ$X2X{LE;EC!nCiBT#_bppEBu{9+E@Se*abGVlD}7;XxFUgm z*wOcPzQ15Wl02<5Kj2f_d87%R?yN&4yGD&l-ea)ME~xQmH9C7u?^X{e#oq z2i{Oqj5a&(9IkkzcJ4NO?Ns-sExunVc~5qFE*ciPWAoy; z`Jx_!>qZ)w&gXsQnY>RN#WeM##zq`mNUyOucB#C3=8UZiJeYN=n`Kk>xz>l1PZ|s= zju`XNO@4db`6TM0B0(_>x>($1{GHeR@6^3eKK1TowA3o8>FE<`!oJs6-q&2`c?~FE zGL049e^x=Cg<^W`pe3$B!;d_fq)>O~I%8u#ZrZDeqOr;2$F3YAYnOBM@j)-MIT3Ex z?OwgDotruRKE8BCMQHG(cq5B*Pr`=KuBHwyiCyu0aP@Qj{0TL4Y1xj`@7(e9jvo>+ za@_Y7L0O-jF55UxG&KB>bbJ~;YFzO8sGQ&jRxhoe?%eFT?^@daJ@vQlZqIIi@zCzK zA*cE)joXay^V|4vICu7(70ipN3$Cv@wa9kRf-|>D;%9wq3Ks4^e9@yaICGNzjNP3} zW?t?4%fX^-<~O$s06^+;?2mW3sx?Ay}4Uh#@7M&i`S1j ztV6MOy>A`HmBOD4}2ZJ`1q6`y|**h1wM8y7%*|9&HHz2()%?mGD>u9^p>fl zYgTv6$w<30_vEM1s~T?v4?ME1t=r06@`wBv{OW;+H(eU|EZs;NeJ|oeXQK`yte@?E z#6DI$+{d7@iB?)*JZwVU)?U`jkFTKA-M@A>@=SWbpxJh>17~Gz92#}3%Rw_+rrPEo z47vEvrhoGpfDA(|;&@TccN*k;*Jih0_A};Jn*kn0{;t4akRpa63R7Ctqd_h0f^v4! z;I~i|FWdos>FfxZ)I0x0$CRpxFiwMr#=ngNHqb;lw%7C_9rC&BGR}9>YN0q={#K3{hk#rrA-GDccERnp+`{P3)f(*ll z;{o^x!iM1^@lg;}HX0xE^LTW})l!cI|6DAH@EP~B_P^?S8IMo+hx%lG?d z`rq<=Xv^C3{%6|wD}7t{Z!kUypA6B0Q}C(ywBJ2*q4;#L=Ru;WnGjLfy8fTzv;N8Y zv++5<+n2xc%=|NLhyU#VT!^b|KK6)zS#R^ONPPakthImUnEpAR1^7aEZ$|;pvgl_t zX)d-HdbTgr8%;?TO{O#p%#3ZpgfLFDk2jvN zUP?1Z3oA1>iy>xXEj-NpEJDp9E#{dmwpeJk%wnn8DvKnubsFDlv-1{*%`RJ}=Ui*~?O{GzO*>XXQZ4;mQ!p{>m|gP3*PBD@++@ zxj?x>W64x*x7?*XV!2tl+tSMGgx7xMImv-#8 z#MA)Ye$~L#Y1Y5#k!JJXSky$F->EOP`K|YC>rX%Ts%@<0dVuw1Yk%vT8jnfVFVMc# zAvQB?BCWq$M_W@g|KlWC^KI5yx3$@9ZG_rZhubW$IcM!>Q()c8rrg@rrpnscrpDSs z<9ouo-=F;NraZ!U|}!KR`n9C zZsGdpi{zM}$tB_Zu7K9p05ElPzqS2o=Bk+Pa6~38&Uv%Pf2{vb`nME8T05ZM|JLrM0D@(%iC#(!tV9>1wG`dRqD@M_3M4POu!OoNXDXoNgIz zOSNBYE3#j1t8brX+tEHnxzTcsGS4zcS!}8FI_!1bcA|ZS@}y;r@{*;CSFP7w+fe(* zwsY)%wOwF;QCV+!OZmp~m6EDzP-3d?<}NB*^8qSL^S-M0wsH1Nwu$ym=0jBO<^d{i z^U*3_^YN-t=Cf48%tKW{<_lF5&0|$l&6lbY&9|sFnIBe_n%AoKm|s)vvv^`TPgMq= zCBmP#&U(*TUPnwXnD4dmvMaT5u{&<#Z+Fn9zuga;WV|v+rtducDh9sJhsd z*&EpHus64>vF~BG&)(MVr2S0uBvpp_Ue#*zovLW_tlxYV_VMoPJ=$lk&qSX^KGS`c z`aJh~<8|KqiT5?{SKfEMKY9=K9_9Vnk>P|p37pVX|IzE4m$9RbqnV=%O8bjV_nzau z(kI_%t{d%I_Y%Z>7vs! z@ag6a^H;;R62M-^{oaSY5BOa3Iqq}U=e*AopPt_4-o1T1eXM-?`#AXw^|A3D?Otm& z!F`v`8-U2nL4bX7QZbQHMuc5UNo zx=y=BJ2K08I_Zvr*m8n%{tKRPRR#?nX61EYG zd8$%4)me3fBb)DSF#S~=N@2n!Ty?koqeO#YbzJ0E*_|_U#$i@g;`Ct8e`=G zeSc*C!v39o59_Yh+njoOaIMAGO6$E&xmN3~QmlS)PqezNGIrPAeBU@(TYrZ3PkIZS zqpUpLi>!vbS6Z!ggS~@zI>1Wp>wrELhroIy;z-n*^&hPAOyToT(?ViZ5)9bP#! zYTH7#KW+un_DIu~+S?gZ4+IPcj0H>v%mhRLVgSnkselZ?MrWv_-Ug)-z+S|3qCVt& z5^NWpZ@~J#^D|h#asCWz+=bzybNXkri?hy)l0nJPJgch(ty@s*d@yxK7ffvoFay{C zTmasHfq>zFv4F{dnScn?jyeWP%K)h@TUBScYy{gjml9a-bvXp zK9XaS0=AGIMh}75wGz*1%X6b?g!7+FRltNpfuP_|)eSLu?NW_FJS^kg# zD;qLfeG!u+S0(yVr8J%0S6-({o_dp3N6)5v(BmOHqAMd1`k;jTgn^Jzu#)+m$z*k9 zsaOW=u8?sroSnz6W*0%9YM@snP5>vE6Us^F+~hF1T*yk73pwSOkPXfkdKm@%glug!aCYkXZ~1xxl0Ydag5K5%q(T#+2lUnia*iDnamD)3TU)WO*b(~d zC-#AU2SdN>#9WDoBv?`ky>FB-rTWsYa18O%uCjXB9T`tfqM6Wo(Ng8ra+N|TGn8r1 z^1o)?NP9{9PMa-rrTfwS>4EfM$SN6552Y9BCCC!#$@DaO5j{dys+UQxq{qol(QD~S z`b~NrJy7w5PGVF-Y9j+iSB43rTF;e%UgbbW5F?lo%1DOSJCjksC}osumbJ2E-YLi` zxn5QWIVE2*q>xFH3t1%fnTE`+kV8_%w1veLaz^?>p2#S;Qev6G%uuEt;}kOuazbW8 zZpd8730cnEC_`Bv>zED9w@fZ85At23TWaX?<$m4j5b&hpTua5PY)xdhms?|Fwy9x;|xoja@ zpPiymvMp?!{QhyO_y-tXHp>&Au+nWtX!n;R>o` zpMq;$=yOasj+`elTTU;I2gjEa#CahL;RJI0ISqPYoJdYA zC!F(AFO8GQc_TZ<`NFx%`Jz|PY1DhGS52>>f9E{r`TW&AN zj_Ah~%l#o!Vj!19)8MM*Fc`kWzsSsD*hyX~Gm>BB_LSFi8@M;Q zja)sRl*i<8dHOs9o>ETYsd%Bu_|-=cPmL!)#uZ ztcd3!ujCDopM#u-FL^`cB)%R$0CIpC@|AokO~u!z>C=26&tU-MISk<&&|>-H<=Onn z@&f)jStUPIK1)7NZb<7&d&`f7EMZF_KiC)kcUbfUy=Y8th`q4PrD}f7v7c!2}9xicpq|#B?_O*(}Yd309t{tNEk#b74~Aih74mEWI?Ni z41zDk!8DTSJLC?uWiUlt$RMaM`XZM?9zmr@B{GCeg1(|4Q7@4zWEAuf{g5*tCs_zB z1hNW-i`pq7MO;NFEsQ2o$Q9wVO!(}}h0KAakTtLpG6o(KRf}pwr$mvoTF4L>MSBc6 z0Sy(6qA#LYnx42^ubZNWBA%8=Q^GwWSR5`kQzVL$#d+dVv7KUttQxZR9TTU?YT?e| ztmvaKkeEo65+8*MGVp~;!XzP*C`qiOzam+ZOiPyxR;1C+Nvc_oCBqft6zQ}~8do|= zkxiSX=nD6lUeYjyhcuVwFAaixcwy2!+FV7vG*?E{1OSOf;~3T(05eAZgehD7t{_Oq z07pPlt5sMm@Ff~sb9*rA0gZgnh;B}RrNChdHjrk&Wx!z!gGSD36&lBYCAIQP29AKF zzgA%|x$snn_30qDq!xlR({)nBOuZF6#|EU;6a+(D*}#yMCS+p zGoxwpK$`vFUs0@-@VkLFf$Rca-pX$ea0Dc`vuJDy$E11SFnX zg?Rw)udy|^Hvl+7dt$#x4+M_Tmarl12OPnGupvDNI6^O?A4m@dj-VuLNc#ha-5;$@`ft%51kz)GBeZO9EO3Orgdfu5fFqa@?IAq@ID!%3 zhjcJ-1b4!Q^d#U2y@|R=PX>LMKm9HBe0U!>;%M=&95NG}AA@Pw#~bTn{;u7n@bvA_}Bh`t~l2OOcL-%{WR zCWIf-3BVCr`Yi*Fu#eaW(uu&A1G*75q?3Rn>?LeSrvXRkLfDYr031O=>=)^5;0Qv( zhIAjY7K4fF8|fKfn+0h3o(=~-x0P)^@P&X_qAt>lfFmH8tyNe&a0Db>wF*lBo=Dib z5Vj=X2uPeYYxv9uzLKyx5w=v|8wpzn!nO(c_Exq+;0QN~^Mdpa;5z}g2piHRzz+fL z5jLc2fFs-|Y)Bsle!7+48Q=&92|uLI0zU^hMA(o%4;%prU8}GQz!7Q)Kcp`LM>tN{ zkiG&O;UZx}`YLdQQ-lraYrqjc5jLc+1Ahf*`C7gPj!;1OA^iq8LNQ@O`Ymt-8qr^* z-vLKRBK(kk4;%r>5yJifID(qsNPh(WN#oai@B9oL0m)~=_Y3f*R({`r{~-KypDB>q z3{vlF>!P`hk??`{CTtxv7I?-6-XCy}XdmeTz()gG9`6|7F@TomDiQdBRh- z3OfY+Tr0oxz!BaM`$hT!@ZMBZsrek30r%I~psz%GV}Zv5E)aE*R>OqzM*&F2wAJe4 zz!8w>Y-fQZw6t9S4q5-PgDkBbt-k^P(aNTV-{D>XkX%@+)vtknAZ)s8_Y?4MfR@)< zH1r7}?TPkT1?d>z2&)Ji(y_o1))O|QX zXao4Z0BHI8W&uYyO1zdxX9Gt-;!>~eiyHV3jjj3GfC)e`8bEj6k){LZ0g&jzUt0MM1dh=1I0gYnXlWY) z9HFIcC~$<9w&B1LTAqUuz!6&7Mgd1?X&Vh3p`~pM@QHwy`_{I5j?i)+CxD*@AklqxodJ%3MAzO0;1>Zc&((F{HwZsn z|84@mMcCF8eYpqxKA>gYN5BzUehxkb{tSRb*Iong=Y%bVXzvAZ1SGo8w|Br1S{~ac z;GYS_#6G42M?i9)@S6b~0ZBb! z3j;n&V{1OPIl#kP*`k0i(%9fRlc*aF{6Q|%C-VH0+Q5;7J8%RfC4{XII0BO0 zglz|K1SAItTM=*sB)ayBfg>P!N%)ljM?i9wu$2NwK=Pch?E;Q~L`^(1mjOSlu{B>; zM}Q+B(Y?Qp0!KiypJ?wGa0DbLh;x4&_$iI8xxEX(FSfE>1%9KI?E~<}R<=Rg;aLDM z=BxJgL3#%80|3)WS{u?G;azO2c_+hfOqw3S{S9~kpyl&DM!~Ue03=3)pAqo>fa}D) z1nFYnms;q}F9LEj-kaWiu=u?P0D0I;Jj{ zV|(_{v5Eum%?&E;K(mxamc z(f^OQ_l~Ff{r|_GW=|rc;bb&WNvSh9#xYJdr9nyq6)LnPv>bcyz4zXGQBgjPLFH`{Va>Zs(lmbv>_rJs$VR{qcC#xo9ZK!4Zc#IiVzH7aR)a zLvnMYBJO@D>YqtPvT{*0ub7IIfbvWT%1(1bIou;q_DGOk1ILK9k8=3^b=;M?nF~URa276cwZFC1of_c?HT|3Fk-NN<~D^EqRzMyn--Iuytjta^#fZtAM1HYY)g5QG&V_6wH4Tm^j`lbN%4aj&HN(x`m zg$x{$S&WiOpbV9wq_Q#`)Gbm)1rDiP-dk1#2Y%}vCAqrd5I1anafoju6^R1vV|tW{ z;(;8KfE?q&+(?CGC?2#ysep0<<*^pU<_7i{{Ej&e@r3gVLGhv3IKaBcprqJ%6r3rA z;uDim@Ei6hJ}n1Da^YGPqnJ){!nzJ9$q~w-GfHxCMe%OlC<&7hxF6VBq4>yXloS(# z`!j!R-CYe?FkSUFWceeDl{oO*_$-6iK7)BMz&V0%X80XTMdIRdTN2=!fILQ+mxkMt zPQTtlbHpJ|SRF7X@Qq;>UEUGT6vd_EPN`2yskgo>0x zJYCkwRdC)kDuQ9E3l(vNc(u|t=o_BaOk2^B5Gn%J6bWC+hkFFqBMt`+RG}g@ARpif zI6DOg*O5MB+gER3zr_ zb~15DRz8Xplv6Q$#^eL)N)Sp34n}1{;22|3Bo4|+5h_z$j6+IFsYn@U7vLH~!^%>T zVAy6h6`a2Vr1qI=d z;7}CkC=LmaLcw_nP)3&bBZG=$LL7p)0@`NBWF4%tD~{344TrdUJ|KC)J@tJ+qRSYz zZj4a3zzGVBwB>nX`$U&ZMmLD987N~Wvw7F0pbg$Gj#a`SdP$wQ9K>nAYKE!ppTalACKaJ zo|1rVz$Zall8xeX3QgmZ}j#N%(MFnP^7fo5l)MRRg+;0!HuG%xQu znxBuu+F*fg#L8sZondm0ML zQNej&pf1F_zc3!dJ-R(d{@C*sj9Q3b@9`1ji7Lk|H84*rK9t zfVD)Cm{>bDtnGMUudzB{4Qd_O;GEaM`jERp{*x~yUQnPfURY=?UQ`q!UR-P~UQ$vb zURv5EURL&9yu4!luF5K@UDY+~cR`MY?MD(7NygR|1!+$<%nrvnvd7n!6nHyp_X0ISmS(Zgirr`Se0eN56Jxm`l zz7O&sZi0H4Elt4z{)UqB3vfsw)L)Q~;US%SKyFLHd>K!aQB z?H1glWt#%pWuz6L3>deBoN&2aFiwbJh$F-c&=258cz}HrfMp~XXzP6d&w^uy`wDFZ zKe9&PzS-@KtCk_#REGi6BGpOYAg|_e zDX7N~!&h*t0>xt-d)cl6S(c&E0EfivAn?=RcvkEMv;mOwF5@iLC$xXLC_WGJH)zKo zzl1Um`oy@jBTB;h^g>C%|HJ>`WU$>R91;z62I?_}S$}*DJn%Iz+=NmE4bQ;VJbG^D)dj8k#SX; zAE%msoP=6@9;X`2!??kUOs&`_$PFMCW}*1(Y?Sf`gO=qBlXpya{=!f=ZciNIwagzF zpl)KiuA>IOnGhWvS}^DSkdX01p;*3zK0RN;kml@j%*Exop)1G-*2OcM1M_p>K11AyhP)PFXgn3eIcB(iFdzCeurDwkY~8VPbl&ubi$h(7c5Qj>vGKsR z;2uF60_tKi40Q_1C$yc=uY}{pxDMS;WA)(pq0U0N1@*8#oPgg7{Qx)~Xv080E21_? zb9+4K?x!~BAI9kw9=FXa9@-r29?%!q4r~MCyW70-%Q(Ftw*9Lf#0TgW}(ntfVqJ9 z1ZIO#U?)+a^C-j!&^C%;e<}`5gSG@4Cs==2rjQEaK``b4+O5Q)kXr%F0qx+>P{?EH zYlXu9{8d?t!}`J6gP2=|L#u00EH(n{4N8OZ7l1HYhqi6)|8*(ZrPs{cm$PP-w_!gFHs0pNR(7VWn+&4dXu`4djZAMnl;h z5>O7n(b>!rF0b&X4AeO)ZYydmdJxtc%JRMLD5cgpR zJQQ_t$8i82%I@ijWAlPIfyorz?f_ZCgLQ%S01sjah!fmD&p=SpF{Kv8(>WTn<-o7UcsuZsAqVn7@gV+6_ir)#1$|_|o1wfy99#A=G20YC z1?S76NDv$+-46nBF3eAp!TdDHPoNE6#$pUZuyv;*j`Z~goCN3=#4jk%9wHFc1&?Wl~ZGO37UDt^c-x(1tJXE5@Z9Pzu;<5{MV! zF?odR25ll-x0SdWoF|Mw!G05{NFo>`#$i_WwH(FM_Z0AuWv&HeoF403X&>&33l78# z{@NEP@1QMg?2rpVo3fjOp3DoNqxq|*H2ju}qBQ(g5AeJr5 z4rl}FEVPkOX8~sfu@Bh($~>`pm|dXTAQ&G8d6=B~qxb;m8$-;FKuM8NC?4=i7{35< zMjSXR6GZ?v;**o9ND8Qf`D;*4VVsoCg()ERKq{f{Kk%JVQdKQIMvM8st|;Y?4#MXJ zFs==KB^X~?X%EKHv3NI#pM;|n?D+%qp&)loq#{Xqbos{CZ8?sF`E<+ckMW%q9f2|g z_YLYel<|Md0f-F&Ie@lzMYdOb5Xf!8crZD@>>Y^3LEMI%G6N-LuEcC1_y3EhV{5pg zr*J%QT_B%_{bT!x@h%|881KTKlVDg1{TD1g{pb0}ivI#)l6*vl1U& z@o%8rUiNJ;tfZkbK*j)8fVl2ovW3}6xJS!z56nk}HgQ=Fuzg$BbIixYI5FVh%Xo?1 zhkOEz7n60^Hn_$x5ynB5Z8%mgkIF>%2{3E{eg{~;kQHtRbb}rTp!2uCd_1h~KX&|| zm>=DSlk#DF72?iHznBcsVIfP8>s;2d=9`3T*XKpmv#KoLBth+Kur^j~4z%r{-#_{Z+ktWbeODM;#yIE$CXDZca$vtnw3Yo>Nna9UQI>t^UHr7ETF3v+FK0ZJsAt6E}F)=|TDJerFIT?pf zNyU-UAdY6^KwOr}R8)y0L5>WbVFHc;<_dLxf_g2WBdapB`o^@*d9PVfX|4Q$A#%A#$D*LiOC>_ zeGn^Aq^ySSt6}*N$KW$X?AaQd$AihJ{fVKeC$9!&RqoMAs*sNv!!p08m zO(i{sj+KS$0_(5L>2GX?asz$hm3d+u8KhV0tn{~%kLlz;YlXGHvNkLCmdApv8^jyf zehsvFe{lh*4=`2-{ZBYHXy4#^|B-*7w-9IGb0{zdPN>HaC-Vkb^YgVp3~MLAB?9nk zM_7xCr&&u%=2=Ti8CI2*v8^gE=UMeHZnlh_7&bxK1npt=Bxc#aUr9p@hw=;3SlM`b z90My0`-XD7a_^6RK^dSjIXD2W9!F(@ybQ#Rn4td&{tXKD?<^SM_i}mUUw?W-XP%ok z5&7I11=08}zlCDo(_55l?AL@i@US=!M?{}3v`P=m-mJSYsLwiV^q^rr)?c?QEbXK1 z$uAnQxAJFBn?we6)Qb6g;l7OtYfpree$r&zCWtq4-|kJ%>f2eTL`h)XnW@Bc@)afU zmEO$6s>hNx~ z95MgA-!n>Jm0d8)!_r;xuLGRlzrE!6UNdZ7{dzLF>0s%$s12xyfWWV^a33Aj+wVWj zZdo(@eG32khP`dRh4*D&sf;5=B#D}uLvf*}T_({&$F(1&Xtuj^L^P_4&hPSiOL&&I z_3%^5fm_-7_$`YYycFsL#6@eh-v#km%73ot%i+vRl+3v-)NR}Evj5y?y`GCX!&Gn6 z8?Kg_q^k{q(FWt{@ z`DXaXC@y9RgxHt1_ZB$i)fC|MbU0B3%sDVOB`pjjGzNvzB2#pRWhTe-YX`uxK$+ zbUW$30m1E*YLMy9*Oe@zXHFc5&HZ)p#`^LalJ8D8(Tc0@)c6&c+W0>n8x3|H+v!3&l)g6KR0V977*8sT<`5TEi8u$5^Aro^A{zE^EK_+=`{QGLqg z;MX7QZ`8(Szxo2J#y6h zY(Rlc0ULQ%p)E6xPt;KQCNNq?&5BZXG zfM*&}jDx*1OJZ@!PgTREC zA-*|J88XW970hw8SVdLrto5gA$B>T|n!FB3Wd5A0YIE}9LE-n-hZNn&PgxRQtY3ef z@qOk=y$_rjOp0zTs;5qJIz{Re#7-Z1S??PBo88Xfa{#StQ$!E1^Nn=bTAi>f=Sl}W z&oBA$rTFnSNo-YMU;Gd*L0z&kY>ikQQ!RP^<6U2SL3jI$3pConM_(Q;Z5F%{Ct?xD zD4Ve5T*>*ur5#BJ-Lh3RRNR`=E6*WuOVj&>GrHGrTsN|qQLJ#LLCNN7NpotBU-tEw zms@pYin;Wk3g6|9DB^2aFtTG7^LW0n+L+J!mS;#p;GOfEFO1RV zTgisJ>&6$f%J`|DXgdbW9Gz!2oJhA59OicS&o3V?fop+gEo8j@DM}w)u+kK8M#0sz_MDHt& zl#1kzyOuvv-gs_*GgF5NL&Jm4MsGpU3v)tkhRP;y2khole|87lqDBoVf9L(V{zN$W z*rEDk6K?E4 zYw*rBVy2=}$Li*iYYq~$$K;2{B5KKlAvOYT`ju+|hq8=Irt#BAjn+ zMVrU)-Z!KCq~Yy1AKdRh{%Py2NyDa&DWyZpJ;}N{rMfs+ ztdaMG*5{Xtm#3fTw0;>p-BXp-EpgqnX6Q%s)ESabQCnuHj-5}AD)`<6B;K{8=|8r!Xk-fgFczQ-#P)A~Yzkv1QZ(b5CJbV#`Ch@i>JI{5LPEFXVcb%@2yxX+l zy2qmHP_{TjyLWU-#@DOUEKYgYml!>rpTelv#}#Sg;L^+7i0=MgU8+@3oNFD%?&g2 z4vlY_`(OG{tDWB7|28ip&m-@p)~Q)=I7>9t=$ed%*ZS4D$JwI#MPKDmZh5nAoOt8Y z^J_cXH#&%Ma8av)~HaYtn=pxs-FV3*F**xK6UrZ^UI8XIi=Xm{Dy?#Z^GrKmtK2`Je zf$7G&h_0abo3kD6S{y#PCO)TpFY;8TcL!ghpv%w{589JTdGL$1V>-IUBvt7025moo zZG4Tp`X1F^zukz;2YpXO?oqt8ek@u$_Sa4&m%-^tYUT5n{5w?pAK#tv{di(xSG(9l z_d$Hf(b-mmjrh5nRgM+5-@O(4gCiO8Hp`QBEk1mDMa28AlY96ua>vt#kE(qM-b&R4 z5$@9!59W;IxU+T+m!+-eI`cjB(^TmX&Cc~V-=^^N@|G7?Z_P24Jf$VTbBya%0N?p{ zUnIwBksi*1mXPLcRjqNBLCR?p)I&!M6&5Lz=Jjss<-AA4d6H6!ITqM0nTvKBe9XF0 zIqLFhN~7;%{d+F{S-)dk=4KbRj-LpMx9SZVm?_$=*P!^&{35>G{cuM)?(4wfTDd1U z<$as1^lvy>9SZYa2uJ5*1KWq&Mh}`E7OBaTt*uS+M(z>Eq?qfd8wII;2@NBXnsn_G=_UVSVL38iF2@Y7F zRNomoS9VLz@^tUza1n#G(_uwbKg`y#xsr4}YIpB37{3>9ant^U*M$eEyC(Vsj~5?I zu1}xjE2GM2_6DyVx^(uO&bPoX{X>V<`FSh(gqr3nx8&QLU%zR%J3jRD9$uloo9kNH z&panix{4YFR2~yu8yMlhSx#HZOwTDPF~2k(_G4*~`{S9qF5ZnIA0C>_Y+6wAY4M8= zdz>}--KX$wLF$|zzNX0OS81rFRl(a{!xuk_?qbd36njf{C zyAwSd^_n}vFFkaP2`t*fQ>L9$mVRed-IME|Rc+4P?lO&eeM2k9v|CVv$4mR}B^Kjz zZf#Mo6*z~&yCe*wo;3cD%V?-S=)*0vq-`hZXlGmO@v>6(h=KTxcdbv)Oz(fnO?DP@ zH=8Iv7*_wge0S%rPl^?FE8ppU{+Xb|?Dn?& zzWP-y>Dwj6X-g&vB~y(3K}8-vb@2}+&n#6pzt2|bwB3?jb>Z&xJ{hImKRFL%I`0qO z-}PXeywfQyDu(A>rN@vcnoU(psr*no=v&v`y8rN= zWTg$E45JaR2CwgRtv4Ey^=XS*bTEBz;N-!o9agst-LEw`-cGzXb#;!rYi^Z=oc{L8 z0B@IrpOkR-<4l(EHtU8La>vKie92kHBad9%Lr%n2yjfTL1;_ZbmnTZk`qwODvHQk^?Q!)r zy00A4JARfh4Q3i^O{tSEf7+j@yXtO|YD8DB#4!$P^0lieLCz8%TkKtQD=fJVKR@u) zw|uxfuI628TAKRWWMs?z!tI;WR^^1fXcFV_K96+z?zOdQ5`AXmxIpCEOG7+vP>%U8 zNZ9{KtXS%hi>K~7)M6=VH@~U)X?mTTRulJ+_tmZr-DTSoem8f|NZf1b;Uv%Ge8fJ1J**f^%P4_nAjJZnwLB-lU*XFyenk9q1=1~{V zMzrRf`n)K>7$y8hSGB&OLP&GvXRZQ^33C~EzJ+B z+(wY2TxXuLjFgwwT%;6cj)*YqTq~ofA#y=Ic_~52b(LN@>w&4 z&b3CR`vqN7tzceLi02hrU0PosXjSeV@~K|)s?n~mjRh4I6wPloHU8iDSCdO~Wj2YW zYx7TajTPq)`ZSolY1~HC?cZ`~wb=VLIWi~jYepY8*7FP&S#`qa$m_K z#x(Qt&3xxBsS$?eRE2TjI|O?BmJ9?;pT_zMSgz-)o0~#CmG%lN+*iNx{_&fR39^J& zHdmF!GojaOPT2&wWjfF0bdMC*$MI^6gpXDRkLp%Ea}_;Tq|qHNqLeZso3dlnK0QY+ zej}&-nC}`>ci)1cxQ=X5X8grAy^S7<>*psXlAO9FSmc!((v5fdo-E58^-k8YV|~gy z6DVCuAeiZI%YXi2^-!(X>#FxWb&AINe7>|tYhrS9TP;q#y0WHBJf zUhEyF*3564crSEKO}&3LvB!tCHqd4LqU9Nv&{sO@`U<#5rFX2xqg~6l%O4mF%RI2j zI{D-au}A6W@%!U?6uRD-1?pU9CVk&Y{q4(?u0EV)f9FNxd!-1{i<%qEy5Vm_jU%0H zd;RLDs+av-&MSuAoqdpUtix(DT|3!sE%K>9Co+7;EtB*RT|-U-+=DavCm!A}(7Vf9 zm0HYwWz$r7X5RI;F&728M?PO`VO|?r)WW$c>)h7s+XIC;e?1Pnb4H=PNx#aRGr!d0 z@ZGX|k>%bjlwIPFxK9r$7v==|p800;f%#PAZO;ndmzz#V7s^cO3pP$uH;!K@bU(lM z@muvvmPyik0*pU1yjvfyK-r%yBGUN2x{^XT>&1_9vKxhsW!VT^%J94=o4Zxp-&?BY zQ>ec16TR`iCp|~^ys_E$>hOJQ>(fWqjS9MtRJ`bFUp?I45ol!|Au5__`dM{<<3Xbk zH=7p?brLz9iae_)DbpuwinY(Z36mdl_U(E;>X}n8cD8@vMa}*lsb94nQpRy_FH{E0 zs^x4dXzOmh<`&#GN}KtqvVV6`tAh`tPKKS+wgke@bw;zs%87WD__YQ@SJ)IOky77h zJ6e9~x#OlQkzEM<2i)FSYaF2!2)`MTQ3AV)qfd6qKk9iD+9NaW>*GA8}J zUX7x91-oY}X-M-llTpi;4L^UpoVfGm=HYmuFz4^pse)BA?xXxXfu`ti^)T_$iw`4& zT`6ZnO7CfJ8f%MFIk$(nB)7^?DA0EzFsRuI96TO)(Q&yQX-* z^4dLvFV*&_mbod-KXpHRmF-!zw`RBO`V@1&+25I>PggH}$rI3VKeWX_%aA|WiLn7q&Unuo&OO7bqw`d>d=&pPwiJj*&DV9tEYpQuo+e0#Vnxb78K z!NYp3NTpO@*z;Iwq+RudZjkzDZCoPG8Uh;GkJHkV4^r2jZFsrMY~yh|N40LAAkP}<{}8~Khp94k#72r^uoW>)aZYw4>$ZfEjshB zG%YmzU+FE?wf{bJcoP7(p13cR=@ox{n>+A2ukSzu4ydNz-ogm2K~S{U`08 zeT8W2>h>pHe4Jfr!^Mn0>4foTtpfEgSJLs$>!@dsoN!)AH>GU!uSmI?vXYkL+A_1l z?`i9Dn(E9JJZCGbDyHWZvJxzSlC@|1)?49!uEKRkSKh1MVE_C(UNG#1Ax4&rp zs^&YbVGpo$r|YIIy_!}0^)*v4O;cUBcE+8vD1ix4kHDp9};^YOI3xbPfv>%Z|pZGLU}3@;v(T#wRL``DLNQ zAe|_JI1XJsJ$k2=c;657=DC}e*zfgih~Kf!un0l6-E{>Yn{aNg)Bu7=J`((TH)mw3?b2Cx0Rg7P}uZ}(~?$XnELH^2!S~P6CwtD!t zzgh5Z<#XK(YQm9x+jks_`+mQotV2#m`Bf$@W^zjSN#WZxGdGWc4zI2O=Uq8p=JtMZ zkDJkdW@6*QL82T8idx)e`Q*A~oqShn+mhC1+xpv+YO)8_C%ruV8|(#lZ1|R4^7O-$ zmvz`Nxi@XsUyq0JCDjU!Pv|O)HN8;rJvdU+zkk9cJu~gVj>SivghZ2Zi^-dc87-gv z<^&f-jn#~TRd@0F>iv3OL>`izD1BvJv7t2ddU|#58CAok4ATQIKRqY>Rv*##sMeMf z8m;9T^50im6K;}~W!dkZoSeH>sH9zD{Jwl>Xydx9=Q@!FrtcHwkG2g9ZTdOp8Gr3e z*?pg<`**4sr);C7=QMfCn7z%fjs3Q^Y|W$=Z_h%W>+kjb(kix>h~`pPvYz~uSXU?E z$$T_)m3>3KR_Od4*`7TW*{maXnvHIf`W9#3xcS}w)SYyFT%k%V-u%~_&&3hhQ?4AX z7dc0X=%Z>L5#a~j+3Efa4gSeW?B932$`a4#eBH4yVE3l+!gjk}vn+mJ^of(q30y!q z!?okOO2-lkTFR4MuNswXXFM>zy)BE)=-0>R#Vn@?1`dag+MZqAc+x1Jw)IJYLhe}; z^H;LIHq+L9@gp*uo^O<^W;)Pxz3Xz&+Gdd_X4H7iEM{Z5R&6a?gT-6XN@Av*U(Veb zdeADdpD!lZ=@7BzMG7TS=(_T}kl`ac(+~H@RQ*~IGtM9zvzy{;8(Il_-4TJXSgsR~&8f8NJ zMMzkmZ?4)EPw~xSXInB1RjX=(i2U7b65qGqn3rE?y?*pqwDGm=+vQIfd=g0AZLSk| z$;pNzf3MfoJ%fpDHVp%%8m1v!(TkT?-;VHV(4Pq($~LO;*jFQ}F{UIPSrY!v_ju{< zv4F90gHK$Kkq`I73dcW4Unsq>bMK(uj|Wjk3h!=CI3gVbb~Tq~A`G^ka`ocLOro)# zcs=HJ{Xy=?^o87z;AdMjudew#;&YAPbb{ecf1mw29~L%3p@FOaZ^?b`>pSC4OsVg$ zc{`CNspxnllfQY@wQOAMrO?s#htJ;+cjosqH(Iv6XNo0?L_EmhauHnbTBh#Q>#Vk9 zS%;6%OErnSV1_g0{kUOCee$mT=eGL~&M5iae|qSOvG@AiL(>j1Q!-p6quLKzHj|MN zNtY8J?_Tw??t9K@LH#TS8VB3cAG1wQ_FDB=nTT&-j6d}B*Ft1}->**>#O}Y*VR*z7 z7hBG-{^+kkheDwv2@F!N)El=e^Jm;WDZJS%g-yBM>-DUGdH#rxtu@JbmAd(fuLiKkn_lm?ItD9(b3nDmY+3%TcGkjn)G{R7z z_K82=YV*R3iuJWyhNj9wd2UKM7jRR)T9uC9L^bt=&Hla$kOM^*dHn{SOq^7&Oon66rEl|n@z!Xs;2n`GVt=SVa4vg-xBQuKaaI7^86ARU_?_dKd>`$1C2S~N zT^|3-<^IX>;Fx#Dx3r=^tg>>hUF&DC?q2cx_u!TpHBybc|^G@ zf3@Bqu{UtxCC{76-%@8o%*)vJ+%=54>m1^5e&4BLL1mk_Y_DW}R7%whlhM$H`QSa1 zt9a`R8QZ?w)>lpp2ngN#uDZV>@2mrn&&pi9=bULDaYJmHKxM-3zBVRnzdQGW>Y`i% z62)2+I~!DWt=^or&13u|^|bJ3n(2bYT6a_~E9Y!JL)~P|i(lnxxJ*xU!^Yc!^|_Xh z)f9Fdk3E{>c`7bE=|*JZkNvK;Ki(hBCP$bkrMhjAkL*v|P%L0^+}U2f%`Zdj>4UB# zf(ePLm(Air2x;hZXqMJj9Xrz<=u^bE?#_(f`zjX; z{T;GCv)2ZE$DY~Vyc`mAE~H$gJW}1ea0e3B-gH~0sxtNH`F#d?v4jUdS?)xzTA05W zsXMZ#mud4{&4;ngs_*v+f93o3)4gH)tsKG6JGxGGWKQ(&mCw4F|MIoS`|F<5uKE^T zl+~}a4mw{rCgJwu*iGwmW-L!!de53MWY38eUux&As%?}>)2*y!=q8f}pSX0JUzuZV zx02*LtIqgfEV^hrquQxtM}G#x`EJIk9nNR+l{ehbNt)SLJnO?WMo6(e{W7X`%d_jc z3D(z{FWhepB{eBC?p$(x={(YS4&mCTxTVHC!7lD1`Z3A&&I8?cor}94u^BvZ^nCa1 zSmd=dwX=@WQXG5c14>s@Z=FeXe|on-IpIlllvMJX7eRAv7QcH9gFNQ5>Q!`jAC>=b zV)?BrHTPxq*f;HKUbjiNZ#2Bs+j{5Q$qnJB59~Sd0C(u-tMri*`O^ZmR`2YtPQ5+H z!?D)%n&3p@c>1nvwPR2FJwJIVHnw-V1P7A&x_2ptTprxhD$S~$Rw2#Y|E$z!p(&#B z;bS*Jp3jNG?9{;UXrcIZ5xt!S>;e0;I9`<2GHq)*Vb7~=q*EgFip$wFS9OmZesfq~ zN&NecVULdAW22`wGy2HrmXJ9pv5Hj*$ ze>F3*=bB}W5&iCY3tlLbrSZa$>Quoi$0vzYwt-v9M8tfqxp8z-FQLCbT#gmL}v9Y z_HpeXouPm&9=k@JlwA0Z>7)zIoiQxEXjwXx?0Y^Y>kRKUa>dLCW!52~){SSBE${B? ze)!F{b;K#RgSL5&*Y(to1BG}#Ba<4FxQIyxd17N46;5oD=bHIVJR@aKq7UmZBLt0Y7$2))WrezVkggTVLw& zf}cN{S~}^QVIbln~i=?a@p}CuGCi+Ws4>y1Q_Q?Afm;9SRceC4fQ}^D# zj$7ml^)PcU8~q(F;9u46^L9RF@yq>8vlK;p;hW<6gk!cY3#psBjCZ*Cg}BRgy_K>I zIxR+Wk)FQzty0{>^B9wQxz$(uR2K=Ad~@qv61$y`r*cnjeqr@<=bCdnDbZ1dKQ35& zU+d_1r#PVUo0gNZ@eIS`ui=SSkz2_b*91k^+FVz@wO8pWH$$e=OK*|!?oQ5^dW&m* zXHvC^%rfK|)|AgC>kOM?yI%AuvNWm`Xtut;d!qPGBRfm5uWMOw{~HH^zN-oRbK9K+ zHx_@r+ag8?7#U9Hj+#m`*?e}d;IA9^zME`r?R+@qEq_6lVp84K_fEqt!%zAIt4gK4i-T_yHAR>!1iw`eF0*;x3mXF>29cOeMH;* zs6MZWTTnAOx^sqdu4ju_b_g{`!;bH625Gvbl4~ci+feh_vn=0ntp~4aJ2Ot$yWTvx zRCny_#nackC$!e#H>dLumadNNakBE>YjJwx&+Q#q_m>{){US7#8P@ms85%h08`+no z-&L&K_MQ0R+s!@6U!;tOI|N$At`;gi!t=^KFLF55X3|cxb7@PvZM0vNw)q^fp_*l| z`+UQZjbtPJ__T%J-*##A_)sNN;~GjVnvPWbp864fsNnEJKQ~DemxY(4xMPuZ?N>Tp zS9BsqLD5Z!#u#m{EBmTLpV=Iw${Q9Q80N{}1|3`({Y0EIf?o0$E~()CHmM_Q zHTon4VRw_X6u#I!X*oE`H7wgkJ>Q8qtxx!9{o&cyoqL@fUf8oe{BTn1jGiFv>WtiQ z!lQTl3!)27$y^jU<5l2c)KbzN+s_nxLtS~d%!JVHA3xjGoh<4TS6#{w_)r|!yD`=U zC(2^w#r7oVy}&se-T238n=UsVns{0+)0-%_{sKBHMq=)pzG-j8)SRMvnQ9jKRf$D! zkFN1yRZ3fkyZ`S8lYWh#Dz@90u4h-N%9K=`{=TsfFKOFir1O!jWOSf0^v0I83i~$P zD;+RCaBf?KZMEO$O#k8M0^!&Yh#^%_s z;^Q}JidbAJNi=~kL1}7v*6D580}UjW^Z4l-)$U$CTFnfXyo+hB_IGpqzF165MNt^t z@&>nDjOz`zN@WorcWkM@*z3J6LB&G}bytz=45+o+>XCoKBiQl~yG5s_7J1KW^DM#q zEIaiN-p*`My$!MlRbPRF*f==Wy#6{&HRm77A9V@Yx`xY-$M@>y0H&=%hGuDt8SNck zL$2??V6AFyTCCU(;$Q#&{KX^i1^?(*|MTa6R|2cJ{P926q~G~J)&IM#|Ie@U@g;cR zH{)Mx^lKpTpEN`WVuUdbc|sncm{37LG{LVd6E6}k6WKMnG&g}15S2P*N1)Cr=6)FI*-DQ_%ioo0&W4Wg6O zU7`=DgCLQ_EU6;mBdKI!iIlWvlV%OERq7eBUrJu{x#lo&QR)?O3@fLu*{!)>vrm(W zw2{n75+U=F#K?G(I$4ioPQF75CHs+LNwws3QWH6c^qiaqo5ue3?v>h0*&D~7$S)&M z#$U;QMW7kqMd~Je0EaFIO7W2-$lR1Tc{WP8JRc=ho{2Vl=49O{6J$hsfy^P@iho0T ziGN4hPdF}lm~dLMkMLP?kT5RkN(__oB1TJP5G$o}iFHzfBqDh$Nt*nG_*!a!_ye13 z6d?_3lRg7*hLDD-=AwqTX1v^9`FnC4@*6cXH6!H27j{bcpAzTq`I|C$V5E1&dI|J~=xtS1;&uE1{(tI)5p zOZt0e-eC4e81M#v@!Irna2YG(Dw7XW@!-8O?|6`k50v?5rr5oe`~UeJkKO&}6RVBH zZ;<_W{hKm#e|p7g{OjYeOsC8-nKxL8i!x)d?|2K@YqGaxe#_8gm}OyMZy9dca2Y|_ zWEnBocKj{byRxk^y0V!vdt^&xR8z(nFHAOzdHGo2-x)J^3TOTr_HZi z?ZqVUA8A{WPP$~SoMXQioYhWpV0BpKfI6@_usf`F;Beq{SmVG2Ufd2m4!l6R*Mj$d z{KbQGJpI®^>uK?fm+O_*jvZTh=Ucnvk=fB)P3r$q54m96$$Z*TqI%fq{W_p&m# zfAaq86*$ER8`&24^TEmdI7a3H)sEDh5^-7sLbcLN+r9Gi+l3j{=b}BntV6#39@a z>?~_o5|9ie3)#SOiXjK#XHjNUW7K54&*;DygrtD{9H6UeK-*x>pPv59Me+Y~@4poO zA2Zk)+>o^l-N-fuA7mE;fkA;mok5G?0K+DhCk*-w!YnNeXBmtbOc^dPSTI~>xW(Yh z5Ww&ioIe-Cz{?WPki<}elQJRRlrkb- zmAWOzBJV7>R^CgFUp`Q7tGqukK`MlpDiue}mm+F5Xug)aAYV>wmZ~Sdl-j4+q4`ek ziu^}8Yx!}xyYlVC_foyYX{kvf3we}?kav;xk>yB-$x@{K ztVJ>=A0(NQ4M|pH1JV`pInq7y1(G9prs78SB&Cp(NDbsdQU|$=)J?9E{3K;dE+%se z;Z#>dFBmj?yfDo#ai9AXSiqNqJ-%De7NbL``eIma(>t_62Qw?W@{u+FvxMHQTg4 zX?1H&YQ58%)jF$S({6HhqjxRkJbsL>q=*p?3B!v9F(}U)@z+mysmgw(N57^(Lpgt zD?-adDMBegDFJI=Kx?a(hhl_cfMNpHeu`F>)DiM?zR5q)u zR@tetUWKHhrFcY9N7+*OxU#kKS!H`=1}Ym>LFJT+hKiZW5tXYd4;7mgJ(SawgOm%D zW0b3u8>Jsew<|uA?g67cqBX5Zmc~nqO7B&bmu^xy&e)_fsM4UeQR%JpRxqrCZe6)n2Rpke-s>r?gESuJuppBT6@=*DMSNk(QKwBM}7L%lDpp_(gwLbXylO657= zVW<#?kOtxmI6qEEP-qhL;x6Qe9Kl`z$N-`TUMDc+WnW?I9|6U39g@R=2o~U2J25UEN}#a6ii?q{X9H=$v=Ew+7LE^DMAKqv@w8*=R&|^DjQXs)Q{AQhjrw2J-ReK6_h|QN ztF;HUhqXtw$F(Q5XSBzC+k88GFZlk(*W%mb`v>3azBhbt`5sk&s{V^wrFl~`OEXuK zruhfW4xe2679!)93Ge{@&+1 zKHu~CzRwSQ{>kUZKJWSbv(Ha`e(v*2pAURC_-ypq;$x6bi{5Yjo)CDP^6RnoQ6_0n8vrF6fvLwZr#E$xx^ zOK(XBrT3)|rH`fMvPRiW*)4q6S}FIF8{|>)O~S$<01E$@-vmp_z0 zmTMH#74sDf6pI!4ib6%DqEpeW=vO>cJW)JTFq9mnRH;<@DW@xEC<*0a zWv;SNS*&!MY1Jd8>eLikbs9JwoB_@RXMwZ9Ip6|t3Ah4WlLEc!2IN}6?PSUEtL{p% zpsqKi58HthE4D)^!`RYIe3L_HDyb>OA;qd-Rrt;foJp&em~MwSYHextNU2(9qE#Eg zvEU@I37iT}1E+&Cz?tAIa5jmgItRG|aEU9f>Izd0>KaTf*tVOxu(g2p)&&Yt`Ca~I70{@f*Ve>`{j+@H=}Gxrp00c#Pf zpLLUUi#5o)&-#k>khO&MnDvCUl(metoVALzhP9Tpo|VhWXBD!FS*5IURwZjctCm&I zdd6a~8(Gb)R#peAlXa1GiPg>OVO?QeV^O{7;3|2SxN8L6+*{lpZm}1Gca7W0eIh6n zgz-l8_l1iab%nYex^mqu&2_*FVKExZ=p*z$*YDTo>oxlS*2U>(>euQS`h$9dew&`BFV*kShY6l})QT=jMt9&Z zKG)QJ>OZRgU42V^M}1HIC-qnA|D}GSeyXNvm>Q0T!xFQkERAgx!))W{9KH1DjOfhh ztmy3Moaln+lIWV~hUk{)_UNu?OSC-ZkI@y;G#vk^*lKuq})sDAyBO4)$<>U^~xE%9$vYa8k=h#_xWr23p{3f z40=uHmCMWdUkUgcp~h1q)o3+-8l5Ic6QT*zL~3F*@tSFxw=|D^*?w}rK)*OYlizpz ze(1N-@0`9z|AjtBlinTM*c(C&WMb#W8psuI~V?~nA-3I5teXq==Wn*hkY1!ExaIXXINF( ziHI-56`@6uz0tpp_&S^ub11AO>}j}9s377G5rLuAk!Qp1`u^2d;HUNr^?S{4w%knc?G_k4fq8|wQl?e~3EK@mZhG~e;P z?enS6FMa2Q-VQC${#<+1_lMdI+V`}3eCKHm+7?Zo_9xnXzTeRrwI?+{@V)1=&TpCK zzUF}@IOsP%|Ec++{jT;s-=F*bO0(JT&pyA_Y|~!V z+}22ggh4~TG(S&2Kfg%7H~r@NE%wXwJEi}vzF+@m{eS6MfnI?=fuVs3fjDZa?4PZSdxJELO}AFg)%HyS)}m zOJzMWhFqywDSjfZ_nhIiSVkyTDh0e|ey+INvsl)tZ1z~?xn3rf`-x6T`(+xTKrHoC zdg&E&c(-`-`T3%?;!B?Qy;e!{W&33p<@Xi)l^sfhWSPWIIzzTbaY;E_eo-(@cui6+ zec~A{i}V#__)r?H3QqAA4L8ekEKlTISj6xzsCP z5+*H{UXw4AJ(I6h2>8?Zeu7%hui!%|%oolPHHxoD{Jf^|)(Vz*T=XavPWM{mwM?2X zU81;E>kL{j`?~889Z^`NvesaAqm$yhj9*Y~X?H7@`hdhnfO)o;SM6y(}M)FuOC|NGW zvlSV>*(WF48Hy+cnc^Fj%K6F#%2P^%;-X>!Zw;@IH=DnJU&@~?SSna9s28*figA6o zUwBE_BODZ#iY|(7i57^Jo^hT_Jx_UF^PJ-qCYddnFPS4*AXy|?ELka8B`KFQN-kjp z;-u51^QCK~m!wZXOJxGNUcN}aSiWCgD}N$qD6T0q$~a}YaIWem&;dL5CT~A)HZPyo z$|LwofFVvm2$l#M1s#HC0)dCYW15Gb&>&nSTqCR(b_$z;=^{`uM2khsL@Pz*qJEJ; zEEel=J=%(qS>!n%DAxn0Skmt)_R@GQ_FCgr>DB3V5z)Ej_0WqWVMyX6%|LroqLhY7 zX9I1RY=P{OOf1&`?R5Ek`3zuQAfE%=%Ygf)T&$R_Sf*$N`o{`^Qlm7W5BW-i_bu-m z-kvH3H;fy_Em3u%ulv5~yIs3oJD$}#zjD95evN+Qmtq~i z-kHiA^Q-_S5BIckrZ-c=^koJz4NPxZ?5x*jCC*NsT`+U|%(9t#XC9i_H1ovF3o}2) zUHk85em3*Q%r9sD2l9`=f1OF6g};S2ON6a@mjA5aS>dyiXSL6)nQ5IVn{~`@8*>}8 z%uail6zykHRLZ1iFO#C}Oo}ctDf*a6(Lp9fHB5>QF)3Gx~DD}DohgMNSV``Yi3AIo3nuk!cxkMMuPKgEBR{~Z5x|0Vt_{8#z^!av`CtN#xF zO8;yAcl`h2KkWa^pBLaA;1duY5EGCXFf$-E;QIk@2P_SEH{hQGvI5owydSV3U{gS0 z!1jP$0TluJ0}ccn4rmHE9dIt7E8tSVrvX<3{ypG!fHmN&fd35muYjik3>{ae)CKFp zbkVv5U6O8w?(cN-bPIGDx}WM+>oR7&JF9BeUuOMv7GpMlw%2U$*^doR(GL$>Pk!rn z8U6Uo){`fG|AT(;Z9UQZhoCR9w%(AaC`1k{6pDTz5b7lGt+jctn)MI5AHeG}n_n_o z{{cVJk!byKBvxYBcE-pxLZ7?E==lhd(Q(Gb5tHAIM8pw|09!0dW;r796CH`mSb+#_R5HLO9t$_If-winE-{Ajy|6cze z`M>A?ZU682f8d|vf8YN<{6zuMfM)*_{%2t2ynmVh7yitEm(_FF`3QJ?0;u^y)seF# zbyn(>_6t+cBGB7GKMu+c+7Prks1bBI=u*(FpnnVcJm?wde+99F*}>7Exxq4!FctsQE(IJVz4FnNifeqH{1!n6YOd5G~mE%NHKiN zFy9blh%+oS{J=o&x#5Q6hSP>3!!E-wK_40#47G;egQf?4JLvC%Rt5bes4S>B=uFVQ zpq3zO(6nGpaC-262KyR5Hgp>PU|4SWGMXyK|qGx%)qL8q)lhL+Tr4_5-Mok zG%YQOmP~sCa@qL1mL4?c(!ih0D`tId#<>Rk#Qd{a|6|Uc#hUryEdI<*v&1vE&N4E; zJ?q=E7J_zz{tam=w)q&SZWIGzT;f^c_2W zmr2ngCPnE?ihjVP=!Z;-mNF@Nhe^?UOp1QOq-Z6RqAVswKW9?(3noQtnG~&KuEV)t z12fM~rf5@iDku%q8qgYmVBwd8XOEj~Yr$J-MYQd-608`P;kjc4)`qLFYF$G+fTxRx zY)>AWXvb(Rv=g*8jAJ`S@;q%NdlhB}IfCoi^|+tB$nIueVfV9dv6pdHa@OGXypi*_ zx`VCdw9mbObr%jx!1`dWn3X$s%Umf-iM4=Y{K1UMxz%&)=N_Fq@%j!G1J_j^vc`3# zhb+6U_K;<2#fQw{JF&EjwBOO{%)h05Oq+o@yp&C_XS3(9=d+iwm)qtvB{%+di2fJ) zztKABB0Aonq0jg3qhF&d>1w)x?L!Zx-=R;ZC(xtlUGy(#zoDDyyXg1mf1)p-FY+e5 zzfb>1?*{rmdH2(PO8*)C0PO?%JKnkUP4q4FTl7ED7kK|WeK&m#FdU^{rhiI5OMgWB zFWO&e%f0`_K0<4YEfDZjvRYblOBi*VeT zgJV}Bj(a8?*V1vEdWV^hYu^jZ0_c(YG}>A)jWz?@vGnncWxGC&mJXSBY|{1+(MWyj zN9ogOS*W4>k~DTpygcu2-W6UP zub+35w?zGf$KZExG<-k4fgi^2UsYy??7(AY3fG$yq5}FU%Dd3M++!oLijz z!t36(!c)QytYCBqyRn9GQ+Q3-FT917jQhgJ!Y4whNFxdp%@E!7o-T?K%@)lUEfxLQ zd%0+dXp!hXXO(EJC|9&b^cAO6R4E$rz9f1gx+Z$Uxg~ncdC2MHUF1Cz4T|ceOSt4( zh*WG4`-ugrXPoKcDDfQe0X!JvxL^EO%<)uu3OvQ08c#pZI2FTln&)&+F?YV_9M9KOQJ%{^ z-&C1YD?M|u;kq4!}B|;9?yQyuRIs37+xH&#i}2w^j>jZ zO71i-4OhdR@3jD{GfTXdd--v5y?(6P@0F=)^y=~M@LH+*nd+A+J=ef}=#{Hlr`oLA zrh4M_48)N{aRrh%?lf+>%1>gzO3rldUezK=wW?0FMp7)9!7Z0mO7=?{RmVW5Bn0=Q zs$0?{nayoi-IvVaK9M|>Fr>d$X{7VH25FRZhI9e#Uq@tiK(y!#pxD44dDMvP)FOZ34Qkh2f zM5UDZ$>LB&6drP{f~;TW~rBRm&;bl*2sL-Yh_~fO71GIOs!I{ z;Z|bpwN_RyYn3(2I%KD?7S$=cDC?H3S>Clij|60 zisg#+id@CF)y0ZpZn@%n>QZixqSIqgk*@xcx}00d6)WFU@8_;ileb$@%2n#w%3AIs z)F;|OfO@m(**!9-wNL>*g(Z1)(u zY3j8$q@!@V_JTvg2Yvv-YA4wl27}BB802PLcoV?3VVkUd@(zr+w!s+L;F~sXKgye< zYE&bth5ws>B%TrI`2LQ){Oa+rS(@s`&7D00Q`H?$4mL48=r0scx)AcVfAp99v28zr zK*VB+@IHv^c#2-|Hzsn*KDlMF*)S-4L8C7xZim9QcL#2Twya1ra`<43V|fyQV|J3N z9LL}{QSIV&l(`oe_7Nj=I=MeXQ#hVBDe9bwWR5D0CqG6#se_*jT$_+fd8{m*&8BRT z2*}@9LSxra^3QMC&sVwEjJHEhQZK|8!|-$1vgMs+vPk`eHZe!(xpd;z=fqtrM*KS} z$Fw!tKg<4B3k@k0MmIZbO(C`WCnZTMhyuWCPG7E zwwZAVIsCi1bSpR{T~ddVOVtZx-L%~b;~Jq(#w_X_M5;_p?Rk5)&fBVOfh0H7ZzMw& z-3oSrk*kD}^7xWt>pK)DZgohFnNq5@Nnmex;CBpH|x;mNZDYZNOyLqz%n39R92)P*_tw8Se!s%ht!$vnuq3nnhl~S>E%yE}C zN042rp3jcBbBz5+#os1FBCrFqL7pH;R1+yb>SwdzT4x(js$#V8=(4NYRPtlEpG=K9 zxRKk*B&qf(ij8ZGKW+yF(4$D(Yoj}GHe>ux!M@gG59_)tf4dncp>RcNi+;}Z6(HSQ@jwXrs zC3{R-bPh0bBPR@J&(V4E_YyuR#c{XebZ7i(3dZ>DW3||OXAjVn!!!cg)uzkDaIJR9 zQwSl4j7#+kWl3cyypj#qVH_U`338;PHt6;eiPw(b2IZyu;?i;3LY8ZJ)SiSn@6u|! zXdG3h@tr>`o6&4{rep0Yd@dHg#bOS*nr2&0a}nZH7P^$&vV3YSj#Q;`lM zR2JIvWDQkz!-s;+-wSNqVn0@4lmkSF1nhPUa)y=a3zoS^BGI^LQ?a#4*lKNNY{n=p zyRM7IgjtA_?XJ52XY$O^!#gDy&FA%lpHocateuw&}h6 zhV(d@_T&;{TbO(=pL(RHG#w}yq$rMvb`3kUDd~?dSWZRZ!1&h z8nTBY&7(6|u_qhH7Z|PZ!eyIj8^%dCFKpSNFhTbmBBRsMP)4`AmM03ItDIYYlmpi` zvdOjlH_T7b)uecy!#k>bQ!RcoUH#j~kKBn|p3zMfpOK^OH`6~@8EtX$08h>Ja|1aw zH4>4X9d|clC;~Kz)AeLA@)TqY@yV)4Tdp#rW8&UM4w>uEwRHl_?qYTWZk4VD7YDA@ zljos0xsbaA6i9c^&joiem(mNnaM5*SsUA{J-7$|?AIUg$CP`mZo-}8_g|{O>@&ZPj zuwOuFqz~?MZlroD8Mw1~d{MWq4~Uu6bJ@wZV?UeyTh)?#qWkX^JqcXj-5D$IW*@Gp zY3bZL(NF{9Y@u?__OWyLZ{;++$|(>~_Yt@k#<~QC9xo2rjyZ1rU%lXg1EXz01=nqV zi}$OZ+{e}6E6iN1zo7RxSYu26Xd`1{D`LgS?MOV8oxxqkX2!8WlOt1z(oE?R1*t3TkWD#lb^a-r-R9I_K7l; zUC^~(&iA1vb3nB_<;ToT%?Ks+%}3z{#>)Y8?0At8vj0W|92%3Qq~2sQZvlL;Mo>Vb%u625H0x zRdTaE()P_|$SByZ;V6@nx+4V~V8=u;sRf&s-Gsx~n3i6w(g}*}(s(GwtA8HxrWto3-7=%0%~GMr1-=V6Sj`C z2w_{nd4-$&k=@EYGa-~IHoiQn|KDmpx5P1Ng<}%EU`h_b3#zzGzftpWRM9cKHv6N! zI2s+P9mHrca;iIXq2ey3rVAzo-dA}kH1>s{gP$%jisE za{T6z^@&%=xLSV&Ga`}O|F4>oCHzl6x1alAYvK|)q7zF{t*#?KO8z+mqYaKOkJ7gn zogQ2c5VkgF%b2;BKQh{ZjcIkNl`r9)>hSmu#*PnlvU{!#X58vVm)#mCERcXE6mXnw zU4?8)%R{ZdV#9PbuS2 zcX)JHB5Ox#M%y1<9uc&?ed;VtJ!iK#apK%xCzhL9tK(AgAZsh0)fhQg-T3ZJhpgQYo}oBLNM6@WX|q$g6GTUollQy)9?NcO|CN z*vRu}1is&phFm)Sc26c!(HFeGDQCfFBEDhb@WZ3|!ecU+OtV3*<#njb8Ig79e3R)r zZvF~rJleVuf9q(Z|1x|DD;X&df06A5?U7AhgzsqqHNI0tVj)R{?^N+bS%qvT2@_{(`Ew5N0h{E@JWGN7JvKs~IgZS=kZiU=|f8 za&OnzW(bd_rzf)pV30e~$9jjPXut9F!$vuJ- zmpm~y#-@+IXk%Swo&Vy|7f-*a`l1bTj{ZzFy~ElxP&9y<%({UQBIobb0aBMoQ?MgJ zWBZ+Y?No70B!`(YGR~@aRs2M<-*j8bxcfJ`-@kagziE80YA=Q6_)F1arzt!q+4fc5 z$kvu~j*M$g#Y^!}@krtccs!oaP6DJ93y_lJY;n|#N^KTg>qrX%S6`Fl;ePZFJ#b(n zEm0VFNpuswahOK8*>`%R{5fzsQ`f$r7C43ONKz0Vw;B=#=CMmHX$^DeIikFi@&XF! z*jpT%JZxSmc~@+X8fQNoImZY%`Zcm=xnXtIQgRM^4sC}uXG*n@7;L)aaTBwEhG>#= zfFd$RwlQ}cRnhK5JyDK~g>5dm%?Baukn_{!8tu7WIDEUNlqU)eH3!{WT}v^~S&ZRP z^oA{UALICRd>J#7o<+~b2Z;;tRZQF1+D4r`-#t%b*Sg8YlYS7XQAGBdtaGHJ+s7-9 zwnUbx2s$)u7UNN(`e~mFwptq!Xpk*$IP@rOs@;*Vpx4k-XbmI%x3!Myfy3We*8sNb zgLs=3>3>xJ9bQNa6Iph6a81YRe+#hkXzjMG-6DKD`#<6jSVl0tLDf<<4v7?67u`bd zMZZ&MeX!?Bt|&(GWbdfB;K<-A?a15a->Cjm8jhYg(h-~E#MxcTAx&w!+8{mKP`Qmi z+0PLyw(+;gjfvTT-EP$pwJ~s!WABuvEZV$~ak1%PYnycrypGf^YtwerO(k{hDJf5t zU2AQYsrgU(9zg;b8xktz*~j03-X@1}qG*6_wa+H_u=T^$gA*sY|GUnAhrVkHeLG5r z9@R>foOYeHWF1xST1Vy`I?lYqrZc59zlEU1peLSM`CWDCypMF{cw+v-SUXsWaZHo zT;vhyVY;K2w)73#);WJ@5{5(Gi9$V6ryI!`M{=f0`z)mXa$eJ!& z&G@~t!0M}0FSCz1fN$w!F^8DLOgf9tlCab)9qTFE$ckmH8%ScASgEWuRyr$#mC4Ft zWwUZv1*{T$`clEEVKuN?SnaGXmW9>Jde!@@|DF1dxJ7d2amak1eNE$4FgR?IDY?hY z`oc+M*|nXLne_!J<>1lnPHig3qYVJ16+8qU2GgP7VW~P?K=f9I{=1bNTFfUo7W@(O22@vN<`N0-OC3PBNFyRiPCaMs zBBz^kWkkjy>rDGbwFzjyvU@^G`6jm`{!^F(9Eq*P${FGebLd&vk;_x1<&cFYgjFe44?iy1hAIgVaQ%mN&@D1T0DR4(bcK8==eFLQVXULSK7 z?&X)^8*`ahBdg<)M@x!KTnUbDGEm>bQqeN) zshXL^wKDo(YZynizu-!iM02L3kxdi*8+PLG=& z&+rAn8Tcmkdf|Tj{rsE4XF@;xrTitL^`iawCiP9xGm)Qo2L8IennsQad4?N$?u?j5 zqsEn-S>(B_3vea&=@`WLxCb$VZWisKls+pw&@jpth*1$UTZuL_b2B7`-rhb#z&D zTeKqbQN-2gN70Iy#F&LKt7FPy+G4K8Jc?1oQrD=7u?u5Y$Ckym#a@kl6sw3!j9VDD zI(BtjSzKG3b4I;fF4v>bqZHqozZ&-_P7$AoH4=AP7+>Sj;L+mI?$PC8@#yvF^BC~3 zdJK6Cd(ee^p+vYkzAU~i{%ZUqkRl;5VPV4RgtCOTgsTaU5)_Gvi3<}~Czd6)C07e3M&cU_-&p-d z*&Aijwl}W6@#qc3n~85Od~@}iZGmNPw!LW-#tM^!CSj^DO_(ms5M~OqgxSIzVS%tj zSRt$tHV9jU?ZPghMc6Cs6AlQi!Xe?XkS@CVCSUaEO+|8K@~q^ylUIY@PA*Em%B@R2 zpL{j>K{9)W?~KG5^Jcs=W8I9h8BL&%K{r5;W{BPjek=K{g`gES+Vob{TWxP$2H$z> z>06uLQlv!MNFo~cqD$&h_!5akEzwDgl8;kjB}o#K-Z5=5 zWk@n5S(0o?j-<|1ASsblNIo{*G1W-eL@+UnXpp=^v`E?|!(N++I^tvE4$&pCNO~oG z61I6jVwDU@g3ZGcx^!5=mrA6w%xbAls+PWE-egXp8Ktq(B&kW7LQ9pVNz&zumo#bQl9diXF*fVRS4bm2AyR=Jck@iaaqyti`bVxcZrOWs-iA*ii z$&9jCS(3~oOO>U`(q$R4Oj(vJTb3g$kd??PWHqt|S&OV))+Gy`>3R(Nrt{b%@HERt zBX^CCV*{CzM}e31*~_rVY_?$IRmgpBo2>M_CsJdgK9Cv{wg0W#o_;YJGkIPj&rh?) z9DC|KMz-^Duj0h+nx71FuPh7qqSTd}NY zS<^C7?lQIyY&p@pk?+oXcg4F!?>4=A`P~Qair$NSk4KvaUIE5?4(~PDQs}(*;5|`h zWahlg6`4huO_`T7A7qM_M=qbYe8uvjH zzvBI(_nY3o{QiUYMIS_dFzoDh)QfRS?B!x+lsz^gV9h{-a#5PNjjctyiKvANoP}C?I z6fKH&MVG>&=mquJX+S}dRYB2^f}&vsMfmKXjrdB6Bua|ZN{aANTpJmc6vZkjN>U~% zO`udz8Ymr<0m`&fmXe}uB}F+(iVBnk$`U(OC@YjTc4|;k)S_%rwu8ogy6IXwUfdn^ zT}p~9N{g}=)CWqX4=4weR?rY=*iLkBx;Gyr0jWVckP#FMO0tp3o1|0+sngWy>I`+J zI!m3c&QTYrOVkzW8hq2OMcuCMQd`u$>OS>=+KMl{4Xf!IzDA-^Yjhf;CRQ^Y-+P;* zS)f_0S*pp@WN9+Ivo$8~Oz#{`fu_K_L{s8j;a#Dr(X@LvXu7;D-YuGTO`ms{X29F( z-K*+T4XK7zbhS=xRL825>?kan0(GyZPt&d*(2(9^HCD|Kybjy^rg^7()3q7id~KF@ zws($qfwx3kxhKOhJ?gG zc~8}vRB76DZHAWYQ##@otL1A<+DvVhHd~vcEzp)|E3`G*25pPBUE8I#XnVDN+5xRq zJER@f(tY?o5+AjX&d2Bz>yxfc@-g|O`lR{1te(S88t~-Uqe0c>AO`6^={}R4?a1~F zpG;7ePqt5vPk~Q~PX(yPr@^Pir`@N^#{%m0=>rXbtUg1aVIR6L-IouN_^LrVkkL2R zH_6uoO7%_iP4~_4%>-rnX8RiPJ?kWYiGPx>)yL$Y>hGBGRO(#+cdws%DGlHJ&hXFl z&+^a4H=+ytOZ+SRYyBJjTl}fs|0;fB>hLeZ@%Ey>p5E<$#lLCYqP7O{rmmN^`2Y!N_FJ=dFxlKk6d50z6pGJ{e$(Q4iIzS%~7NEy^S{%NSO$5vda70MF zAuS+k1CKXhV+2q4H*H=8!`iWHe>~SwIMqoC!hdS0;&MjfEw)763`ORZl|t* zt^kXjdIKox3!rEqfFditeooR5Xc$D-(RF+fE~{;Xoy$f#`_HJOC{{;N626j7lF3e~ zI*QVC6s7AZ%D^|lNy@ZSmX4xq9Yr}hiVAcEI`U{(qNAunSD~u`HGo<`WQ537yKesu z9m5tiLBK^F8C^1>9scTo41Tw6to1o^Hs}%Uuz)N6a{^vwh2)`sT)-3mE4qIFRh*a6 zzvbVryYGKXC(4P;$>7h+nU}L7rzoc>=W@=298qp$?!4R;xkb56xtDVv4{}y)EZR6Pw`t?$jSn`8Hbrilw`s+uqD`VrP2kI*2Ov@Y zyiJk$^YSC}S72KNY64vb&C8#c^B`ZeIfWLvdEVv~n~T6rn=fyE01|BxZHWYt>AWo~ zwiIn?+H!fz0~?77C?e$^K%1m_nV6h;wUr>mv(ct8hhO5xA-Fa^VA%#~^BT*52ZJE%++=G5o-Tik7@} z;no%4P4+Z-Yt`1at>}X-t=gJAFmK>(@H%^%JaBp7#y|?~sV%*Xw=v1~6dKu1t-wEz za=dzna+CO`W$_MzC>T4H*st9 z4f+;+yS_``C%d!NqVLtSxAp0Rw+-m6`XPN9cUV7bTPmL($PbhRssnX_#=zLXq`(xK zDKIrKEigSWBQP^CEAX9do3=TlHS+Eh`JNKqShBrmM&@`tzRSB)jyIP0Zt_$uC2e~z z%VAd~uoLXLlB3<3+TPcq-jQ;=2}Y%k_r1{OlqPcyd+xQ)l1gYN5`){jTTU5g&UVhA zp1$KA!j(uFvg}A*+a}77H`-1eNRaE2j_3T;({|evdg?s@w3BtOQ=Xkz$$DqWX>UY} zv^qkR-H#+HP6?NmV2&j?;VBF4e9IT8wM|P^!-1hlj8OS`TQ9s`$ImPd%5Q+u6~?vC8Xe;1k&&5xV^le{~gbIUN#2*58KPlo&Q_oHV|mV znI1Dcu)_9U%y1w*h#%C;8fHs^)Iqu+V^C~RQcwlAi`mOE1*Hb11-0PKne?DcZbs1L zI;J3Zoy?%&z+RRFHoEW(d2;>#=X3n=_`ff&zs3xc(s3oX9s4J+QVF~ID>S9x{NQxxWdvsiX9Z^m=L9=2Oq7yK zs)-;nT@7+AKN@)CRvz76x2@}wa!u~IPM>m0t}*ms#QI#Cf_8MG)K9c%aJGR}=4EB)d`Tjc)Q~h_3-;3$j1~&$`26qHs4DJrT65JnrEBJn}b7k;h@RMMML12&? zGzPsP%n)aoZXgVE3=0g44NDEn4XX@m4Y`IwL#d(CP-|#3v>G}LU4}lq5t$^l7*sEH zjnr!xmh>6=SOW$tEW9`k8HNq?kg4fOLewF;5N9+;=Zqn-AxR;okkpX0ko1s@kj#** zknE70kb;nskcyC+kcNi*Gc2rJOK~zapMN~~xL)7{0Em7@JT~U^(-l!Yf`=SP-tWiTz!%_5TezYW7 z^r1Rh7j2Awx;-{JDcTgB8l4uM9__m$dB@v3GNRY*$c)a4&W_HBE{HCPu86LQZisG) zuG-Na-4$(#?v3t?9*91_!x}vlePhSd9mCP|7=DZ-MjfMzF~-EkB*plam|~JkQe#pA z(qhtMGGg8?$&AU0Syxh3k{y#1QxH=UQvs@pX^3fwX^-iOvBdPooGG{$drTkb4NR7o0NUAMDp=lPu}0Tr<~cGko;#J0pf-Ps=76>Ev@jqQu|-8B$vjU9>| zj-|)(Udqe zG5$u`(=y-k*!ZM)Q+#TCa(P<(+vVx;8S(4NtI9Luv*NSkbK(o)OX4fyYvLQ?TjJZ} zyW%bJDFMCleenbF*7%|L^X0?w^aOr_Bte~Uqx@;PF2R@(n~;=XN=Qxc-JO<@o{*7{ znUK8u?cG@k*$Fub>vk6;lqAf1uWENiLQO(L!uj1T3GE4836_N3guaA_Tgl0O&m%bPE4+F97A3urI`Y-;I20X z;xaJUWH5!8LQO_fm?_*8VTv?GnW9ZGrdU&)Dc+O-xYJCpnO-*~nWmfGFuiF?Hq9`- zWlAxbOoYj7nu)h$W}Cic`a4sqX^v^G>F-VRO!H0OHhssGW?EqSuIYQGg{DQOe=vRD zlx|vV`bX0bOmCZ(n0{#bCsT%Lsp&_iALAXHWu|vc@0l`9%S}Hq{j+I>X{G63Og}Yc znO2#8X8O5lwP}s%7p7mDvQ2AE@0&g_tuw7BF4htk4-gmYh>P{a#iPW<)5NdyiC@Sh~KmjA8#c--bq|4CNAwDE*&BMbp!FQImEx_62Gk_etVj*>>w;9 zgr$_Q>?AC^2um4ZDJLwu3CkYBQbAbu5|(|0Wj|s0h_F->mKwrROIQvNmV<<)jSWXa@lZ2&>u$&?+@OXx>v=f#N z!g7|doF^=uME3@wJBR4bCA#y7?u|tE7NWa==q@C>w-Vjki0&eyyO`+SPIT`ex=V=e zQlh(z=-xwgR}kHMiSB(wcNNk75z$>kbk`Ey2Z-*2M0Xw0T~Bl$BDxO~-3>%{Bhh_? z=x!%|w}bdy3DJXpr9{t8qGuP;Q%3ZZ6Fs|$o;^fQ1<_MU^z0>i_7Oc*M9+Sr=Od!0 zn&_z^dJYgh2Z^3KqNkqdIYjgvCVCo(o))5~mFPJ^^qeGm+K8T0M9*oW=M2%)PV{sT zJ!grYb41U1qNkJSxj=lff%qhc_#~J3B#-!HBk{>5;*)&hlg-2@TZm5zh)?zrpHvZ_ z9w$C+AwF#-K0QHvdXo6GjrjBw@#$&e(=)`U9mJ<+iBHcFpPnZ^?Ib?EK=f`PdUJ^0 zT%tFR=-o*4ZX$Z~iQdgb?-ruBkm%h?^ll@1i-_K0qPL9bEhl>S5WOD}S4xO0M~ExO zh$}6`AGQ;Js388(Kzz1?_^gEZtd#g{C-K=X;kY(p6ps+sj}q6Li0jS7^<%{KpXU>wZzew9LVR98e7=?Vd>irk4&w7t^B3o~ng{asnFls* z0Ox>n!Fk|~;7#Cs@MiE9Z~?dw44sYJz(wF<@OJPHa0$2+ybD|gt^|{|tHArg)!-U% zE%*TV5V!$+1bh_S1a1Z&1Gj=tnFlrj*CzPi1b>@$gZF?dz_67MjQPNr4{Z7Hoe$sn z@SP7|`S7zDdYjR<8M>Q+bMsCxY;1=9W?cFdi9)d&3@ZyZLrbzWH`;x%oEwbQ^Q^Hpct*mP6*-h1KTU+j77s&9{px z%(u6ffsyXmYQ9}^#C&@fxU33X14e%LZZPcaX*1uh>@?rryV-oZ9y*8W&9@tx!Kci( z8(Yn{kDf5!K2~hLeS+jB@EP;%HuyiY0Std<^1ub)LNI)tDFVaq8Q3{<7z~{=uyY3X z&OrAJe4aTEzF@xH4*hn>wgXoO`q8lojQ(|O1MdQ(PaOxrz|+xTzI_%x&!X)t>dry; zT#5PidB}94z7u+#@ZSm9&a>bP<~x|%cQza`-^uAP-^r~o-@$yjlUE5oWWKYp5eymZ z<99Y~0vCaIfe(V~z>wK=7~BAcjZID9)8;$*kk5zRe8^*;yOUoAE(h-h!$&^!^C6#q z6x;@e&gKH}Ht;_1esDDyZChY#OCESL82VeP!0@*Pc(r1{QP=x#;*R$$o*+gss(D{OCV2gA-b^lcmLYy-}1@Uad3-*yHJ``gg}q8#u> zFyxA0r>F#sI21vy2zZNtr|38s{)%9?=sXyD#fU-iDf6A}z`uP9xDbp!ZvO}jeA^Mj z?TEwnW8fAr^mhQq4#@66pLSp@cA&pIfMo};mcU*K+DZ^}>}PjMYr*if6FzqBFyASI zeA!Mg`cVe^<-lByzLz8B<*-u@U*+&o4*A`P>u%WF4I6vlvjTmpfZYnjYcFE37ctxm zpZhRw`%vBo%=^&C{aej2F}BfZ9siPz4=Zf;*Vp;ouj~Tw9|a232|ux{wB0F17kDBuo>lJ7?WeLaSXc0 zfa@4+9z&duVcd=b({aS|IBXtA-Eq{lRD)4&!Q5%VxV1nB=ZHJ4h`|ZiJ&72%!Dbu$ zwV|(VXm2}bzHt<1F;f!rnQ=4c88L&Y|rbe4Yoc z^T?kEhVx_@e$E5udDz1-;Z7&!Whdg&i81IzTW6E`&IR=60{V9W^%oG=3+?7XoZkjF z>@yGMw3!EUbHKaIgL#|5d%!j3L0pFnZrTVY({12la5eZaxWhb{k2223gE$5Z<{tyM zng_AJ58@m@xcQWM5c~V!7U*t)%$B|2X7EWc+6uOUcY>?TgN0?_7V{wX>%pzt!KmK~ zEL+ckJI#aJfN9%#^I#F|6hXEK{)^6<2aEH-z*F2{9^4L`+c$xMWe57Sqt-lFg8r4D zt>iQqc1kP3@V67PyMT8WuwXwMEGq{e1)nhwVjmkU-wmz+!&domFk-bEws*r9_OZd; z2f%ga!9Bpa2Xz&|jeTvf5`C>K0mFYK%6s#{(BDhuVS69qu@5r)AdmfUa35l_4{_dy z{!}5BRj^kDELDhM75cXyKKDcJqZ44nyBhiGkHC%Q!5Z|f7GqWmnOf*#KOH={%RE?@ zYaXlzzC(~XjQnBPY((1;q(|WIDEfXBn2*A56XcuFpC-sRVa%FNgVEP!_-MvhHN$T+ z{5Av2F^t!-A~1B0!4CGR!Q;puKWQGsxqPtYka@5beQUj79z21$assj^5Tg_DjcdKZ zlh8Sdew~E;NsL7su(x4s+JLhSe%lZqT>A~4hM&_Ir_-oE4J>DnKZCJq2c8bZuLH5g zxpD9;#_23%&m!JuVf!5LoC6;0AA{%6pYxEvfcRfH0`4^5-LMzDAKY%fn^Oor0zM8t z4L)zai}U4OT(8~5wb$J|gY(VZq7L)j z;*DVF7sGe)R&X7-)qHn5@Z+3vcL(Y*zweenrUbY!$M2T3nD3SXZzuxN8i1(*bq$bf0Nw`pX@q_w zVtNEPk3jbb#t!?<-Db!fgZweX^cegfhdrE6?zRAb3vjj~UahD*f%F7$oj@#3pr6p|6zrUW?rDtcY2Y}6He74oJ%jPVwd37(jPE(b z{2XM?BR=Obe&>PfJn(j+trLCd#N5Jh|86J7G z@G-qt#Bz4_iY)Ng~IZKurl zih#9fCm7{o_$!9ocHqNx*F9W!-P^todq^32hI0(7l9EITqE7v1N(bmcMtkqf%=LaVA!mHP0a0k6-UhXa6G?PiTvJL^Syn= zVD!HV^;PK0e%RO#TOZY!?^PoPI4oFYk>PeE_gc_m=5d*L-qh-aR7P; z;ja#Qb--1J7~?+mUOnPd53GlP?GW@2!(Jn7H6rGXz=^qd?+DtCK;{UrAAzkS&^-z} zM}esc@yB)1y=KJd_-XUK7T9b-A6m-5kZnbLTY;e!eQHHNPC)MjFrL^9hTI9n`y^5v ztM9b|R~zOAj?MQ@A&#dIM;x2)okqM*qrL(aj zHagMgPRMq)flryO8xDZan5{VeT5}q}t>ANJYwmV%33!*;iu+tE?rE*KAGPM4H(RmJ zVcodbY{hZdihCC8rbe?B_Z`;!lVI`Nz`%ubx)tYiYZYXx(C?4n1IInnCrgX?eW0oXZk3=C`sQIBhDE9Q;09&xEZ z1U_oE9?AtHhKD-M)&|53*8tWduz3XfM@zwVVAwbcoJV1&sS=EG3*yqU+ib>V{Fa?Qz!I0A=inq=!EUgYV&>E%iOOlFyG&I$b7%* znEC$xV)Ol@r_J}9O2PHuLo>gwI=7MBY$7-LhfC2%?WHaR*)MqS5Sb>jw)>Spj&m7mDub>BYCxBuJ!+8?>*qEI=a1Ku=n0#EXUrJt{4k!ilTz3fMNv|1py`Y5*4X}4HZ$a9gvP7 zAc!C$CeauKMX_N$)>vYVMq`QJT6<>kUgyd2;Ny4i^WOJI;$Ht*v&yVlGkf;zefBwc z1KA$}Bo9&?8wh%_t07IXL4iQlfWO$ac|bO!AQa+@4MXWJa1gr-LW|uC{$eq-*!|Gy z*i3L5n+s`=&4V<>`9ipH8z5J4;b0LL1$~Q)fq>%D!AsmB$XZcN!L))HIj7uO1k|d zU1LeFr=-_g($kUjbR|7KNl#zWGm!KQB|Rfaudk%nPtr4%^m|JBy(N7eNncme*OT=1 zC4B=)-%!#wlJxsZ`u!w*W67YWWYAkO(2)#uB?CRlKwmO2kPHkZ10%_xuVm0qGBB15 zdrF4AB|{y_sHbGqTk6|W>f2lD*Hh}(TQcq`8TXb34wMGBlLil#OzkDT~VlB|YFR`!yWgJd;M8ZJr0hfBk)r4hrV5wj#~ zNwPMRtj#6s(UP@;WIa~087A4xl5FNkwnHV`(UPsRGEsYr?jaew!SxXL9 z(pWQT?09MHRB4=*G;NwRZMrngQ<^SG)19U19@2CuFO^&zq!}}$8E(=H4{3&{QlDmiG?kTy?mpo=k9-flNY{_G>Pw!c zlBa{@IaczVDtXS7JUt{&PswwhG}}~~Z7I!mkYcBUs~uUEp{{Os@u)1s~#Nsa2UX02!|0Iec|W_hcO(ca7b{N!C?-E z1ss-e41r@P99D1)gJU=xBjB)x!v>C#aM;2z3XaimjDf=r4tqEp;1~@4(C48;FxFTT##or)p@Sjv_5m-SYS5I*~4s_ryd;oX43(u1G>zeX*Oeq8yp^H zGePbO{@vW%%-q4BJCxl)&mHvKp-qoDGtE384o^1^Gf$}ZgnG|89%gf$r<*MVo5gVa zm}&-xrCCAlY_n6jrf^uo;Q+^2IA+4(0f#3X^US;I>U0C5i$o8JJ`w{YhDeN%^hMGS zi7^m8u+aILaUqHKoW=~2uZN{ zr9`wyM2keUNJNW7v`9pYM6^gmi$t_YM2jS}NJ70N)JsCWB-Beny(H92LcP7{bT3Nx zV%=V>OGevdtV_naWUNcUx)fASK|u-%Qc$oT1^ZF3A4A)ZP3}jP{n*fc)Z34G8K{?m zdKsvffqEIJmw|d2sF#6ynW&P9Dw(K~i7J^W%|z(|^l|`I4xq{bR5^eu2TVX5*(k`tx*V*_!MYr%D_sS3K$4N900AkGAS595L4gm}`Cy$7 z*7;tFLjq>LDD_1%-%>v$py!KvzNqJidVZ+qhkAaf=Z7kOkOLr~=ZAX!sOOJ*{;21V zdj6>Ak1GDC5`fYGlm?(Q01^Qt9SG_Lz(@lDrvYdYfEIyh5s1=2tP8}t)i51PgHaHS zu7c53FuDpxSHY+pjIM&wRWOaOzjj0FBU4|Q`pYyxrhzgIlIa?m zu9fL}nQoA2h)g%ibQ4hccu1z`4L%-1n!OWg4%)(pJ4g@hL7KZtrf3Tvw8%6WX>bP8 zbtr`oB#?%zL%J~+>GthNckD)*919dqaHKm{%M@FL$B0OkiAYn?8$9Gf8kCK6JGKsw zPLRf6Y`?`|yKqA<)BQ-(Ba!aM48RRlHwztD!}Kk5b)Ye%J5~dQ^~6FK*EOVAgDabb zE-r0Iu?8163te2@fXZcD+AMT&S+mf^bq%Oo#zoCS7Z)@OU0l(C%4J;6EOc=xv(Ux0 z45(a&c=RlAWre7K8pza8rbaUDE7N{5HMY=`BiEB-){|q_lcUv>qt%n+)RW`XljGEr zs+RgQ2x1^t!t#e4 zEPx2G+%b?r&*P>E+JgA^G`Q3X#=OI%iwLqLvc0?3t21)4z_G=r50z`<&YJ^`>2 zmQ`#B7FNc|t;6CfS2CZ#VSuo@qB*RujFX#%<@FLo27r~Yu%c@K)Xc=?Musb!5M&it zSg{fTRA7ZgJ6KB54uEqggS8YnSWOAQ4mps)a*CA*P>B$1iAyO|;$n)NPpKa?i_0ju zMnD+?$Y9Mx4gok=EupsvScxE4VmE=qaAlAy2TMRit* zodyoBbTY6?OgRRG0BbHf13(!9w2mOlWG}EHvr5@n0N9}lGl#VoT?3#wth#6i>#hvE zl4%B)hF*#T82}g{K(h$Y2ZS^R=`zU9N<&Kn7#GN3H3MiJ3|I@HSD4BbrgDX;TnUj? zu57VXWv^2mOt)0!B&l-38AMvD4y{w=K;>a?RbD25k18($sij7n_65phQKQ-W8f8)p?G1t=^xn*3qDlvrGh^d0;mv3g&-=dp~Bk# zQb>^ zWPg58+EgZmQUuH{sM`WlM~=mJn_PDo>bA+Pp#@Ac7CkSa#V*+&^ax7W#s*e`u=ha; zddV%Im(Y2%*vDcSQ8pe>7*RG5P#7VrLJzmVTT;Np%N++k7Z`91eS&jhe;EVm*d;R< zJ9d*XpaTg-n`GH$4KcnkWs-fe1k9A627!SM#JdidOzam{VK_%*HO&49=pkkX)%<{A zsbgvw?;cqK5nuu#P_Q||7_fl|HkC0r(!gw&)u3}|BPUai1DM?u5%knIlKGXx!ERwU z83U$r5YSWDv)#?za3?ncwr4hQ_QH9X05A95^iSzF5|&&)1rqrz3LGitxU4Iy84sCc z;~oV(t~+I;1$-u(YjPmUk#Zj74&p+S1Uy*o!&J#54m|V>W)+*~;2-A)^f?ROB7*|( zku;dNmH`hPfuuqXan6n^3m?{yn41Af-A2 z=mO7L3~j$N+jS^t4O7rUrJ(hxnkp_(6`Q9mW87B2XB*e%VI7v%ZN<{2Y&cTO zN;zev!&nMC2**CT1bVC#o%w8IbgFhuxz9e(?MgLO*sTq!tl(zm=Hm<37MYtbMLHcR zZk*yuH**Z-xFy??UHY^E@ydsY7d}M1?jho34-v0=hsuO5wCZM zc)3Hws~sX@Y){7B6|(YX}@nDmY2-)B=t#uei)N9Lbux8IUFS zAwe6?SeVzacfz{$t1F}#AE?YKz6WJV2k((@0AS<-B!hozWAS?7@g;7~yK*P{> zIC>e5dc#pV9NQR<{T+dV5hxgebtAAJBT!(C0&8r;8ZB%v78|Uy!7yx4V1s5hObWI( zXlsMEBT;!IDv!iSM`Dm83)dh4i;)=VNQ~4LRcz7KXe^CpB~TfSDx*@gtw!eAs2l|4GJ$0qHu zNqcl8u2L(=m0^vA@&N!UdajK{FR@(gjoJf)*|q zxC=J!g0?Q$xC=J!g6+DX0~bu#3{;+p4rXGzGco3w=w&7ofbC3d(iNqySm%nKUD2s4 z2JXtFa4xuFKyDb28~SrY3pb3)4V}87Q#TCA4Q<^pAU6!i4ZXNwRBjm1EHs;iD(>jT zos~ev9UF4THat+^fdUV#^T5t{puiIao*1MjI-QNN%*MLe7{+WA%to`>ObWKM(RMc4 z&Ozlls5}QFor96i!KpU~E#_djb1+;lRPjPrbFnm+l|W@Ks?0@|xfs=4bUF_^GY?hf z;k=)Rg89hK$0W{2&+{>s`6!*wq!7@2v|WG!ExjM#0QGOPC4-v;=LJpzTsrUW&>~Q3Wr^Kx9E$c9kFF3%ySvU4^vB=Tw!S z0>A*NAJA=bIhIqbIf&(BN5EB;<5}RVO2H_np~49>txD0S#nLbT$3%=WXt;5vVL}v543nhmL5Ur=~acBWeSm;_Q!gl zSnrQkK%v*C{n6@l5PCknNtSNH*op&W8iG2-A#!;$N{i+E0LA((a($Rw4wK7UWx5U9 zD&B^z07cDha(#sCVV7LqBTM(l(mlLp4QR&6+5ej4w)oLWDqB4a9IJP%f{L={lLpy*rZuCrgzx znMh0I9+k-BUJ{QwB?(C7?ITjiInXWHRp5aHl<|>MQI#vJWV%VFTYy$IE%bqDEyF4X zz6^p80+tEYk%il_ zSp-?OojKdioNX`MlLIn2mXU?==yNm+Voc$x!oX8i#}uw*um{1OwQdg$L9_PI0=Py1 z*9c&auogjX7D1ND%_0YVK!9eQSvL_t1p&H=04mOGI1pr60F%kX;an*9&zZI7TqqBR zGwTBayZb`#2n@Y0W=Z73j%nV!xDl3 zeMf-2%wet(WCJy*;?00NBJJRE9uS8OSOFS>|4Dx3jGYDG8xJU-k z447R|I^*^;IKUu_K@NjM40sENWiHr<0c$TffN_BgHZov+2##W0Jc9%VNeo!)!L0S* z3|5&5Pz^3|n~VIfBZkVjrEoJOu^L?NU|E+jl=Go7zm@B6<@(`V9})EdA?=9F4w2P| z%JsXsJ|fod=6Z<8$8a8z^Rit6Z#RLrn+D4`#+?SsHbd5rG?K&R3+pp^dx+>Sllwg92fk=4uX059iRMpiGkb2^Ci3sBAvpqw8-d7Ry{@F{Uaz%?A#aooUhBgf4g@v#@`Z{-M&&14>*Ut;}E zj_}k{=HYpUjPMXuMtJ@!BRo%#QGNmec7PFRpU&$6Wj>ShfHHrO^MEpsPwvnT5P2Wo zK0ZyydiKBs5TE@b&)buq#~}|-MdW(k4toFs_3*GmuIK)J4s#wKIf9+PuV%jJkKLOMwEGYYAX8$lzE;{_J{}UdECCKTo0coqCcK5Upc-Sc;4d- zxxq>>M_37GgDgu{_j7oJ1=M??ykjuMds+4J@O!vq%np4z_mCJE5jhATxqZe0=GQXH` z(Z}BWVxnA=j*-LGea$4Kn~RW!%5+P%`K8p-)aB-v(vX&VqYlvZZb&zHA`ST->Ba>} zH!U$Q4U{$U7QnnTNG{`zfcX`?1u(yYw*VF}_HZeXVU-MUApj2MA_KV+W-xLv1sTYd zF!_KhU0b?wA;3BY>ltid5W-+%Y3d4#TS+@%FJQ^9ZI&=zc#X5X2${IJYBWP12TOAw|@m^3VHqf$V2St&;@S*gJO za2l9KL2*AXjvyPrl=T0~?fZmF{e(;VgiA+`wMSmcIN3{p96_0kxP9rqrtlO5NC1*R zB&(4GBiWPf3wL%2z80tgcWy|a4({3#eBqw0$QSO|vVBLQ6mHCrfXYbp0(W0Xwxfq>vv2MGw^HU|kv4w(6c$`R++E5PYj7ut6gBb6f>f@THF3Y_2WJ@8!=4Z*>;Pp7Wjy0#3N=v1 z8r;7j1#dvH9k+XYbHVDFK4gX0GkwT_ ztdEHLm=V;6obY<4k2%76`O3@@_2sLwj7(p?A~Rmrhy2KX5K$kH*E4;{9Iu!CV+K%P zM$kt@eLz`XMplpd668)_g46-oT%-@bBNM5u1vE;gdt@3d(-@h?$}~=<@iI-2X`)P% zWV%b^}*W1NV$n{a*NZ3W(m^9R~6`zGAp=TS%5Uv18xug~O-1pldmzkGo& z7I+osuM6#^G-Li7bN@=AKey&Fd0U}N8n8a-d5n91m27DsvJIF`22V$ zjOVPjEIu0VanD%)>+tw`oMwDw&Kn5%S$mS%_uz76i(Yd7I&=P6Q^wceykbQkS+2w5 zapHUk=arn-TF1@<)pDkvT8o{pT2YKw)Mva?zOL-A2G`d*#QJM8mgSRpFM;>we101i zU%N^yz)781e`x-U>dE+KJbslhA6E$ZtNc6D*X8;&-%@#d#OHHK;T;lyh7l+a-QPx8u#h^^X9x(?x*vI^h3COe`6Letp{$~7~g^N z_2Sd{ME0rNez%(SL^a{lc%BKo*2GWc4Fp~x@Hzr-BJj2X?&Jf{k7t6u)}&AUH4u1(!0QOSiNMK;RVuuOsj#0&gquPMr7V@p}n! zZ-Ecte5#s!DY?AQ9d>^8;pf*SJH~&@`3J)JnB~Ict+{-X1>?WreDDayx8!`Zz`qmx zZTgkzcjx*8gBefj)#E;lr}c73UB)-$`t|G?Pw_5n&v?52JZ;OakLFyzvd*Xe&I$dQ zFp0@qasB6ld~|*0?>jEf6Z+R;6SHs6<<+V){eGNZDCDb~5dR=Se@`=JKZx7^`H}gj z>;GgSU)GUK-jVBXd(L=0&aV{oPh>KA7cQ?`%J^QKFA(gjo@4T!T;A;+;|)3AU+90m z8B9*|WuLH~_tawjqy7vy#q{a=?^v11`*Q#0LcS&n?G^v=>GeVRTW|S#&ds&xdZC!| z>G`T7@FoIpEAUR7*Q(CqrTy{EI*h0J`Ex7A)AjhGz&{jtim!pwr|~HSUPs_f1m0HQ zodn)X;JpPtMBtU2f9TEnqu>dB-GuRUz3$MR@wDH0B&;t*_nBOW+aI0Dc$y!73*&u4 zINzQK^RvxUHr{kULHSb&@jnxIEoYYBQA1e!w0`xS!gxCWtL|ew?LW!Cg4>^}$K*8L zBv)}c@y`Tai{D3*yn(pwhviPp_9MzMrN0#lQbz{_1jn zYlQxr+pzx8dfuxk<7vF3IxwErTVElcqW%s+KB+RZug}}}d->`48zS)3UnTF~T32RI z^M~ZcLOxWS|4o)U1wKUJl>(nC@c9B?EbuDM zCqHEQqU+CTVZOMeF*&XG$1X6Q_B%sIGM?7M7~ypOY9*=gR|uA6n_3=dZ2nr}M!{;JpOi zTi`Ck9==>-a+E;OYQ%9B$x?XkK{%QV{YgshlE`L5hUn}=6)ZoA9^K$P~ zYUF8b-KfqxL`eP1*Cfwi_Vm1%knFuPKQC9tFgeXzLb9j)Yh6$iKR>^e>pm}cTF>O< zm$1yg2D!<9lYd1X>pwjYA*A?}SJ=8$lUERuTy>wxX@5#ca;LLQPWM}cB=;_4?GNM? zge2E$I#NEb;KN?|KSGk5yk-9Vc?BWKoh+ICdu~ria+62Q-iucdmdOj5oUX@&WpY0L z-WS!3{{zUPwV?7HSLpJU3?_x_S7yR%}=%c(_o)}f{pJ0 z?vRlDE0s)6pYIToyqJGZK=+%3B-c8~?CE}vkmQj!*#3w<|Dk$9>Yr-UKj%-8#r#vd zgrdFHeGTWg2D!I}>yJkHz0bGrtx5j1?SG}cR_QvQU&=dSnOu#%(@A!I(f*f^ z>^t-Q3;*~{i;(1Y8um9~xom&dl^dwNK5IHZHOXI;FF$JA*KqyP`s3^TJ4JlnUeo&F zt!Dgad=(UxPi(|9^XYztX?Aru=E}uhIVe=jTs(`%mXB zAHPiL{%f+=AkVk?{QT5({;9iuQvSVneLlaM#z&o8b%b4? zY2Qjn2@_fGkRd54BvNz%TGrDggB)K=g{?dIJA<0!5_IFMNEI$-CA=y9E zV6XKHv!~As2+6*I2Dx_#n?Fj9`5eiG$$y3X z^ZEZz=ZD6h_6>w|ef-+^OlE)n{8H%ud-C@D`3K#15>osfG{_Z)M#=jYdTvZe_W6AO zM*BiSk~iSb7wNexA;}dQ1B)L+H zqa+La7e}fmJ(4fxavC>n zt~gGxFPlGtob1W3I&|8|-%nJhprQT}|J?%(YHG#zHP|2E@5_q&2(=EsQ1EQ?=lg%V z0efG7z7MfBl)s-@&UOxcWb=(a$ME|?dv*QKD5swU;e8sDa@tY%hW^gtr}x0jztH~j zZ_J*)&ocB2?OzG`AMu6u_Lo@uv>w=gq5T-4|8`$!e^B^-{J1Z)em2n=ss@x7ux?U^ndji+I!w%?bAL#`U~yv2=mYN3+;=A_&vVRUfug_spa&u zR2aV^S|oOa}Ip)h{)ztDcNaQ-d+Li;De z`2X;Q_SGAHy8efLq5W!M{3tJb%jswG0M3`z2~9M!T+uHCK2U}zxoxHf7*Yk>!-T<`@;AA)#<6Lcl^rv_wp;-R~Hx6tN8Db z5&ti!&F}p9_tTzn{reo>Xt2L4e1HF6{A;Li`^x;e3D@uY!u`-jVSgZoGhLA93fFTz z!M|v)j_*|7dHIMv&!O+R5UTSl*1KI~^Piq)(D$+P`Ez6iHy{-4XAAc9{cXj0_MA+I z8xV^2^8|bPzPG05sN~1Vfjv*0$)7I}IxS%JHF!Otc%C~wWN~)l4hg;AvwC{}laRZk z=Cu_3eMefq3Ck3O0zEIGY7Y&ccafetRMcht^xz7FirXxHy00ZHTj#mHijNcBzYwZx zPpr?c&)WCo0z##l_GzC^enR;3YRU^C$u*(!EA8ihW&VHoO8eBWjNfSh%YPnE6rrZP zk=&^fv-jc^ge2F5FNEJaY$n_f92I{5(BlB4$xE8LIl6xQPj!g?#7kDB5Y z<>A8jxxW>@4}V5@-nv)#J(BoB=20OyruR-vgy+G6nif4T7VC*m(BQux_pff<(d55R zK5MA~HOP_9Q%&(}Y1rTQ(BNNF{F?mhaQ_M(H-Q*G%{P*lp@#T-<}*!2nE@-1e~O=; zp9*SPn&R)0&nzha3hsb-G5-o4?>G%{b2nOz^I3`f(fbm_lYcso$iHnl{dW-jD_C<{ zKc;Q%i&YuZ@3)7Ws)pI&?% zi6?m(Y8dC8Tv1VGzzXD#{L^|TsA-)4^!;6mpV}dw{8M{0zU7@ieRx3RPaI$JPjQof zTMc<3efr%wiu?a>|BdolOZ0q={L}Z3#rfHS$6em|XKgou=kJ&Lr#zGw6&KjP-iR9zme*dBp3{G`f91dFpXQOK_`g3T9o;1>(`tBW)|zpEwu-t^bT_oepNc)WDqdt2Z$_;UujKBwMh{iplDn!@L^ z4+Q&$f?SO6%&AZFc~0Oj3jAe(zb5c1fxj*AV*5eD=Sw^0v(F3Y_tr)JkRX4!jOi=5 zW09}*!>97@0zWXB$vbm?RwB3On9os>#O*n%IGXI`_8hg6xjn~xj*1j+&r$5(Jz;)6 z5cuB({;9yf5cpRD|3={73H;vzuPv-cRRq4ez#kIUhi`=WG3FdQziE9QC-7qYQG7k6 z{amcTKNZ%OY+*h6QHZCV5KlLO7so^7oiuztNBWD)spnlzz18K^3(=q#bd%+u_JiLR zF`k~I%@xkST|&M+h4CJG{?qYnE%>t*n^pQO)tX5Wej<}-}( z^!t-@g}*N$uHP}unLhn~`8{~aa9)9)?23jLWZJbzp-*dG_3R}K_D-zgU4Gll*v7kH(xKdCc{#Ydkv#0vcr z=Z7YqekWFwUI$u`H7IJ+##n=1Rk|Q(&}%~rl?FX+wGS9RHPkNWiwu?3J`nuWUsrlC z_G$J1I!?vkXa3^A@6@*YuR}Clor5awFXXe>^}=}kD(na53hUc3VgGYn zkS`O~A8WkD4RK3I@{Bdot|1-_HO7Yp&n3VgD_UlaC2Q-$@Woe)of5br60zbE)> zEL_i<3H%1Zf3~n+>ml^-y@vOrE5?squjo14|F-d?bJ#p|s`PUEh1>z_Zb zr^}g2&G#bdvgoA7o-T9wYUDJ2gu8|PPDf$?bxPQ;>WE_==&D!|71M%vw`qFsJNdM_q)-8zXDHd`|V2iKV476{j#%Qe@Ym?1Rd&@P?FeT4D$;|+8f_36Hu>Ng4Z3*;wX7&lr+9`9s+I�OISvz^>ruZ3 z2=S19RpGvi>Msk=7pdL%!g)dKX1+6vhkl2a>Te0-MtS>FsHg8`T@aoNQv6mz{0(^n z=Y@Gd^{<8TTE+Dj`?EOPamS<|uBLvLkWY%Uu_5zA-v^-nO2YV3eU9Lt>RWp-KlD5I zt9V@B3H4>|3ir8G9}>&_bmI2p-%<3#<%@)RUR?%O!v1K9@Vx5n6?Xp8bthifZ%r50 z*J{H3r-v}#lLh;9fzJ~7PQv-DEv)~!f_@D_-dnI=CG_u{@cz$H!C!sB-*G{HQs7?* z``JGQK2%u0iv|0U!g!n)fVItVa3;wMH`DlT^A^6)X#M?#4-yK2!oek3pgnW{IEn%FgexWcw==$rerv91m_k2m;M3^VEZy%wi zK1Y}*q~Ag~A8DQKtfqdYFkeVNU-;f5)ejT)Ep&e{K~4P`;rpYcZz-(fbRXfUrrtrg z?vuX0@O+l;H-@UI|E?#y-qG_{(w`vgPpH240H#mxkI?;!rJDNM!t+Z54{Ittfs!Hu)igJcVVAH^(Xf-|MdAaJx718 zrapE*)9=FLA^nBftiCI+r|+xpQd56EpXqlK^iLdT_1$?reXj6AsORE;0k^))p7xi+ z%o$Ig2TajsJmuG0(Er$o$tiyOu8gO@H#x8mNcq_^uTAEkgS#LO#rdc(m@Y{`BGf+amaL5!zoajBk>l zZz=q}+*v|?90sxWX#GzS;*GLna$2X43GoLD{uT-IFIbpA9fb2`o1kwf$a97Hq9csQ zPQm{W!QNY#PpN`_s4)Msg?K6pe=oX;!2cxpOA!41BIMIih;O!#|I0%AjfC$9y9@Rg zg!WGe`EvP*jUTPU=^Gi}p3g6b5satnc8x`hr}wM!g#IiS+TS4XUV^_8!Tx~|Z!_Wi z?kx1zSD5dSLVNE8f9Hh$uNTI@zu^C!kZ%)Ve%%-1um3H}53T=^LVP@}Wgz~(%8w;% z{iFRgUB_&jetI854t{^Y^l86VPs96)WJh>a$geiPFQEN2>8ZotzS2HO_5 z!I#?qA*}zRJ>_3r`(eWPkNU#=uc^e&H`>qb5&Th}M5O$uwfgjY{XzJ9%A%dfj}`tt z_m|p>-!E^;p9j$WyV#z{i{EcAuRZl&{C>J>h{x5{{Nz9kX|I6=F2!H=v)A{o+zwaQ}YnuQ6^7|BO?8W@k`2EZ8LkRX_e?|UZ ze&0f{*VKQlm!Iy}wZ72aRXBgvoMQgyIT*Ds_MiGuOK9&)?GuFgBgRkTr>=eR_YG=1 zW%1Mf8riFx|8By1C)!g#)!~c|Z2h8lbH3F6pwRy>wf{x1Z!Gj*UEEYpkknFw|>HMMg$set|-|+Wl=y{#uBK!QF;vgiuuSJT7+9!X+|4TfYe#e;{{F|sQ z9@>{?2!5BZ_bJ+2DzsQ#PtLwO;jc<8>nP}m1KY4|$?WdH9X#UbX0_8pgm zc$EJ`@zD82{i5f%cZ7IU|3mT6`9$&1b71j%otpkG1lj+)NY_o#euwbl?ZWAyC(#HAJz0% z)ca1&{j;bysGNGH<J3q&SEroMClZC{{w;*-8|67$(4%@0RpRGNq5|jZ zm$G>!t^=Zasy(}Jq0bFOIq`p~8F!Ks-cmDbPWG~{Xa?a;=^XPp!uM7F*T>r#=mGN{VQdUt>sK`J?1}ZX8k%5W~RAitc0~Hyl z$UsE~Dl$-!fr<=NWS}Ag6&a|=Kt%>BGEk9$iVRd_pdtem8K}rWMFuJ|P?3R(3{+&G zA_El}sK`J?1}ZX8k%5W~RAitc0~Hyl$UsE~Dl$-!fr<=NWS}Ag6&a|=Kt%>BGEk9$ ziVRd_pdtem8K}rWMFuJ|P?3R(3{+&GA_El}sK`J?1}ZX8k%5W~RAitc0~Hyl$iV;J z42&OPC9&u$O<1gL7c%AaS)Eq?yHjRgSa_+PU!=E>&zr4&e&)9uy;^uPrKaVPy4~Nd zc-(i=s&$8M%schKxYWE|bjucFo87XtIXCRXAGJ2cPj;N`XXk%^&C|ET9ZKgI4{(iX zYaFfpTj!;vmL8MV{`6Bb$+qa!_0zwPt+t_3n}Qlk1MPxRCbiIUk50UnQ9Zk1V56m$ zexny(HA_5wVByN;^^9T`s*anaDHrbU`Lai}&(TH+c`c`{zrG{ayNfi`{<~=#uJ7>C zX=SK?=Hu6&2_bI&;C1WQPJkO17j+kuCz;fH@*15tf-QdO8Y1L(B{fb zivfQxtNq8b3EjUPGL)j9IXHG`=II_24w+Ye4Zj@y@ndGU!Dq@+SvP9>V(s5Q z)Lm?MK6CK;-Nuimf3J0B^=gaAW6dT!w$~oHNq@7kYsRc2*RPCUI_Ju=D-CCK4%a#p z@^h}1_9oM6+q6bK&&)d)b|Gt)UzOFtQH>fjt=e#1OLLDiZ|5%W<=ND*;fs}n+;`8L z;WKXM^{0P)xbfTTm(PO=OZU0&JN)Nm_kCLrzrN=dk@CS<>9#6=rdhzORVE6h+jx_2 zr%W0%c!lMw!WU;RzbL(ZWb9_2kh{pQVvlseSxyfZh@tm~n6Zr}7T`M%y2k9+&AZ&>LV zPVgwP4Rl-T=lR5W`+C8FqzLk6E_aK7Wqh z)cY?(?`hrXcg3{B28TVvFGm_TD^Y~T25xS4rH+&J`wd2J>xb2u>HGfLK!3|N1ugq~ zS_bXj)69EDM0Jx<7uPx;+|YiR+1a;&-(K>m*5>`Jy3wUG9=s~7qGOun{KHz)%12+{ z8ooUKM&O;_i(9q4*J8=b?bEkMmP|gnyC|bxzsQo-u`9Pa537GgtJz%d7KNQVUyP2+ z7^tyS_9_2;C5Nz_n)V+%;}%&2m)lUb{|+tle)<^HG(9TrYHLXf$Q&clt&h zjv6)JU$o}nxWPfYikF5z**wEF@WO^02@Q&7_lS-@_Hu2Qs%y2C^Ynf+y8mcFQRKoV zaXarPRetbf;y~XjcOvI48ll&*!-m;5KB24Hf3un|>G~Jr4ew_bDgTPAT(9a5r_>tyKb^Z!Cw=)K zRg*pL`&-<-sACw{Z`iA_LE~#z33#Y_{>RY^AD!Dwxj4?=Xwc*-1NTS$p1O8#UYcbK z8HBsrLA|ydW){{=I1*C zTJ7KIIdF@Gp$y? zV|sP;3>`;kt{bdUp3Zax!t|1v|U2nP+U4Ome*1CEBg7y^CR+UiKr_Y+aIJpg;Nc z0o!l)MND2l*X6wZf>-I;RSl94L?-4QFMor zbKMIn#q`+xw_=@gg~yQ*!#6csUd`s`f%842e)s2WZuY@+-u;4Ti+2Wr4xM@}wV8MS zRJ6q*gTS!Py)N%Fo%6qSo2*=ii<(>9V5HcNvZW`}>)_t{w2M z>y;550-_(3#5RdJf5EiotOH>_n>$|_(Bx@R=n=j8e`vQV?s@Xswcnz*4cay5jNaV) zd;V%xvMlGP`PQ4aHZrZ(+_`R6eD`g?EIi=0Fxa#GFOO#4)g3f((Tum@v(EI@Y1_p8 zWaRF-YfjB~?UL`-rF!8{(OrMu)%g9BS(b0EEl_^<);YN6kmG-jSz{A6=%fDDLH1jA zn@?4ZDe0;7)?N74RZ*{?*X2axm;Jn#xsTncsBo zcHF4yj!F;Wu60;G?0{(_^Fh|Lr)@v}-HbY}=STjvEpF-A)*l+oZRuF=%1hrt{eKDy zH=Wn-a_TGH-6lEX2DsF+w&^t?W9@@}9+x|J^!d%Wp!4deINL?%-i%#4`gYq*_g2(s z@j<%YVDVb5F)Q`&1o$gFerV%r<2*CYf0a!KW!}q5=TF_aZU0?>&Fku|I+#q@-fPa3 zH7+m0YQOqrt=5O+nwz%lbZ*e&>S%AjD+AwTG`2ewZolPeY^#*GwD8;`B_-3d&gJjQ zxq5EI$nQQhF#VDGJG9e#>=<^*$K%7tf`t7kX(>G? zI~upCzGvd8?`I#LnE2+G<9#mwc`MVUYUcAq^H+tg?I(YCD5~Jvq+74f6a=45*S~r? z`_($by9;J-EDVcxkEq?G;6cm8^M)7f4O^X!@L2e~`^qEpdi*uSY*I(9hP`bvRGiBk+Z^bN%DjPiMiK5+|8Ukeyd4vO1snk zt!rvqCnfH3nB?d<@VoJ2CjGYCCMhv#SdXNWy%XXio|pxgc`U!3U+~Ih!QEGeCGPII zZ7M}vupK-3yVNE|OFGpF>3=xP`s-dC)%y6m-nGIzp&O?kd^Th*%8Ykk${D<>G$S%3C@jm#PlV=o?$+v;xYy7*zsq=LDQUAjKq+3fFz4Q4uYDd_QN?t`@c zacBSTpOTPL7^HpByy0H5M#)j>S7++I zv3WCPuUko@QAbQ)1s%RMHKFiw^Y3q)4(hwmcvQfs`kRbQ2b{TF=xAm*u}xIsAL){MdUd%-!R*IEFQSF|4Cr zVVHi%RHv>RtLC1}((g8T#e&jK=MB%+Y-u}XMdk45TE-q}u|+A#@qKptU*79hx%%cV zHBVG}Q8mr?{N96ZP4&{-r7u#{Z1MAszrL;3B=ptu5g#G)tlR_r$2&|aX_1^Xr>jNs zfL_C```a~9-hCc^!uZFF--X0>Dma1AQF>$2^vX6t^s zP}#WFl`8vg4^8#o)BauBinA+RlSiD~ytzxCKZm#Ve;gOMXQ{1cwbtQ#8=mlKR{dSw zjy4}cN9Nvm`)l=sTS^)%%l@V5^lR-yw#E;-^?J*n@A`jJ_kc-)!Ks_^rh7NUT+VCI z#M~g`d~ta1V6V_=mTi`gX{zI9*X3+mtMS#+?#D(o7$0ubx!3xo#i=XK{^HSLjoaj` zEVl)jZw4-M>sPnu?KN7h>diVZ_WTc~p`DBUZmulu`f#23;eK6A4jTP2sprFWP3|}C zq;t7ngy-SsW7b&jn02LehC%nKQ5ls&xAjRYa!(sR28-zqI)I z%<)C${alJ-#TYC8Jn}{ukgCV4?i?J zbUg5Gdj9Hm-gmpV4O=v6oK17b4}Cv&cpQ1;M%a0SRtx)dO>)a0R>$?tmYe%aZ#ecm zZ=I63-r?AC!}?A7O?tBAr((B;)qXy?IQYl&?Vk?STe!A=e2f0q8fjU?egDH<*VVs% zzwOkz)StTDpX_(M=w?5=v!xAVO=}h>w0Udidih>Gw{4FrdzdLQOkajTyi*M2eRXF9Fb`df+~kB=Q(-PmJ{-j6?j=z9KY^)Yo$ zU2S(Za-#Xl;r+UQ9KB@N3JcxL8Kb-mEqe9W9c4H1gsz)mfzzbop_4hl?yw%86P% zEJi1F-`95Vp1i!y7yH-EH2(Eyc2Gha+e(hhJ#%|rY9fuM9KKzcQ{_@-hE= z<=z8nv3W~>SU&rgZ?i7!__1c>#A>mLcSldN%$;r7`|ix+k7lmzlYHV=@7@cPoqmnk z)9Q4uS$BL=pBrwqyA$2w#pwaha}tBH^ZV_eI^(Z5;dQsXYHIlW@{X>ldvj`4**1J; zOx=>X6Gm@2Z(KC`+ThdgQb`H7-A}C(^(GsBc=_=1gWJEISidDc@uwSwFLH8P8G2lLnpp6t z$K1>QaNE_}^Mu0!185jOX+@Q(hHx3`$ z^i_7Rp-0*-8h1;%uwQ+hPB*dx+peDdCc)kDL9|=u-K|EU7p=OT%?KDj>Sl`wqdfz5 z*|uEvW@S)tr;#P*6D9o}$+LgGpz~nY=$bL_H+ai{P1q@ z)wOH8ciH3iBGbzEi0w(EaV5nb12)FDE|wf$6^5UWGWj;jEo52o4O{Cm?<=<*;@UF0 z>HAjIp7#BwaASl+Op}>IW{uB1AZ1@{o!}PI*Y;&s>mIv16fq$GYdwIdm<4VVIoK zv$t8nR!RSM=7LIj-oLG??^d;L>twf(!O!L$jI|$I$tW)L3;*kJSBVb6Asdb&$l*Y;`RNLSM>OJDn~on`mU9le68UAsn_C%01m zY+tuk;?MR8yT{&2oIfQf{QUN$1yz^yc-PT#O`n*V$C?jI>Z85bY17%9T_MQ}2AYQM zebUeN?P4phmEDpbx%&6)TV#04Il6La@-Jo!B4RuS{_Rz8;kNFyEw66n?m0Njev#FnR~+QOin=XQj11r2A)g9?=L@6ufvRK87Ur?OQIsrx@TlKUp6?_ z{D}68KeyfKYtU=5V~V9tm8*NYkLufDh1HVzv6oZ*uU4Br{oo%BtbJ4cBG*;D9X~dz ze%{D)84tVP-7$Yh?2<}#=jH_s8(ls5!o>E^miIU`^So(BUi$1P|CU{U^Roz zb$+e<@SHTVHhEPvP;ZME@XqFYm+J*MeRyS>RvcB|Xe z{#E1C4O%Uq5NTF)ZIPYfnI+0^+TQwo_N>I29iC3NZv5Q*?ROR@)9c)scD`%dw5L92 zk8XFF{m~+C;^w2>N&|ex7&xD@uXQc1>z}@pZHl69kJ0}%&aI_+r7XYKf6TXTY}M-5 z%3V)~uD_YEe(&=FRqE>@d2e5RTWp@vaAnw}iBr~LFdQkl-hnz_60uH z`KdI3`ERxKdD;q;*$&GOO5_slW_2S@q>@~ecl{7vGCb< zKaO!acyxOHv3*^t^wK+*)OzCdP+iY+6ApJSqG|T*J$Z7t-<#B7U@^|ZJKp+ zMXjOpFKl-mrQEu4_|`5R0;h+SOzpG(l}`BZ+`_wozj|yO;rpOor8|dy8}jV??*?uA zvoLVx!jM-D4nJ)7W8{s_1DpJmF!;pF;CrXe{W?FXpmsp1QM2mXQfy*fZ#sQ*II?8- z%%7_HcK7PM^_#)rPnzZanAq6mL+*$Bf44mUq52=a3f*dk9_W2I^}+bkwVkw$9*k5h zP22dsd+Lz;1=Si^%&9kj`=OsNHlHC5()rb9w5zda#P+Szmskx>wL9Cf`~721mX*D(MIQTgkZaP`UuW4R zrTVT(y%5uQf92y@Yb4VPBd^rAJ~*h>pH}%dRyhy3IV@|4ci`)sbq32D)%n3_NY#`3 z42Gnq{_w6WIL z)U%_*Vx|<=p1Sa3>a&~SCBFT;M0L~Z5!$<(PH4NTJ*JF4-e!7c*H?}0>!0wyCOH_6 zUXU?obmI&9Z~O)Yyzn+3cl>9asOt6e4M^>C$$wCWF8IIHYRUki8gd|ms${B*7kk9g3fwQuw> z?PpJ>^nLAOeSF!rI-L)1*%X%6&tz=fN=FyN?H&BZUz zRDYi_vePlkJ8R#+aQcu?|IrbxS`8kL{2|}w)uz9H8ri?@m>AA1#+SPN&?t0v5 ztMQJ>!!xV6y)M3F+2Y}}<}=m}nKwfBWcu?fuQgwk{S4C%xat)N<)-=e+Aer5oS4 zw6AN|IdNnY`vUW~Qm+HgkIl05b6PpMC_Z9Wwe*6~_Y=;YPWq;^iwnCBD#r`tlpdBq;?C zLI2|Uii$QWTWhIjv?$7t6s=+&gqOOGp>CS8X_%rsLK2XP>GL(N2${&%Wa@|0Wa_?N z8D6bw&7ifxT1Gn${T>`bC1Z%FYg|3i&}=;soKDN)o}OrEW++Ew>4}cWB*QXMuarth zWK>2+y4orLp}Wc;N9~A)4yVy}epwEb9SCbM#?Uqi6G|I8-P?roFl26Hj6Rc|)O+E2 zxV{Moe{;C;9SfycA0-DN_h^d)kyO!V9AFCIIE?749>_Crxf6J9K&x?rYIM>I@2_uH2jj=S_^Oel=T_wk!=NH;0 z3Z!-!fEhH#&~JRiz1n0MW0-!eJ3!d$F2dl{iNh*2L#Gx+xEM~GCe9*eKMqll#Gs6j zVhRzN3~>n@N0;{LKT-Z_`@ZkPvQykU$I>|-#Kk_uvH#MwjE6|fIl8kj4BP*7XZinwI}04Ce0LW4_yTtpi8*u~ zcR>n=cW2QxZO%+N6O9yi7uU()I8b+76g?wdqjW`;n7zNIWxOCpWljahXMT>#oC>mj zo}dbN5rJscKb+*N1A|teWG6$X(?$$wM@ZP;cn@UT)<)RC%odJ{_>K#c^dHk-7)OG3 zG5zYT0lKmgy8b-YPojN?WUa_#{rQlrtC7X}Wr6*W)lI}_l|j^Bgw8I?vW=Zh)XO&O z6GT^ea=%ISDNji5D)vFQVHD0%Cr=3bSqL~wFPA$dqPz^!tx-&qAQeXqozC%&L5zza zY}kP!evFTMmJ)H))t0)7)@#oRXu9?W;)D)~&G_mnM2I^76b^IQnSil>3Wxs`4qX|~ z6K=+1N63Ya{+A1f!lH8-kE@HysmlK(EaR*5IA0e(y*g1|+K`hR>enr;sQX9Vv)yiU zLd`$pE-JtHl2+w6otYcYO)(eRJG2dWH0DzOi^3+bnVvCzMoqlzl3be0&AOCh_>)h> zH91}plC#MNnbC#gBdlK`d>6XLd*DHO>Om@#{a7=PH(8s3BkH4 zY-UdCY21U73ruy@xrMMoHp``bUkEIX*OeC)70;!ehN=o&6q8Gp4Ad1>@!FiI8xxSZ5h({oIx}9A6E(TpG$NjI>BisjT&Yw- z)^Y&TuMZ0?2Kx^;0BtEQWj~xvKciE`MXwp1B0O$tMKv=ZbS>cN6k$U5bc%+rQ_Ra4 z)_I^#5!W{PIz|5@Tc^m`Iz8vUQ4sGa+&amys(50L4(mR;U;=Sf5oK+Au#vZdg``+^|fD9IVqSL=JIq zK!V^EADIw2)ruP;H%tuaDnxErR)`!r_QY~odZh@7gcKs!oBVF2`_28PH9c%qp@ z`AgI#gZY^dxug9MIm8qV43Qg>9U=$(l-FiLlEd~*C`$|AVe<+yM8uW6@PjWn4uQjT{hI1mXZ@5Hjao zm3EwWzkfi}+#?f;jM6l9_QD60jqy?yUvq+{%{U`k+EC@j+XO)WUS!r_EfbkF_|V9# z`r3GvAO01{gnz*&&I|wIO!(L6EXC*7K&mRPYj7s63%wm0*EQIW>l*XY#;MzNKki zH0W{a=cNth>Q?sSP;M5gkJGdacmYI-2^pRA`FOAyXGBHKfm)#NcukF~>INYoUTtQ> zoQY>h*eZxC6MS}vACmfnZxx#Q_CUug^@&X7c=hcscCRv}X3Zf{oT`CC-K$jU=iRGR zy6;nz)tl=N1rDs(HGsvJ@_@n@N~Af{c6~^!}HR12dCfYVk~x>pN~%x1NN z@S*lXsM~6*D=U?E8z$4gPyY=6u0`QgXJ}egg>X&_3)#DMA2a0L>i=C$tEp_5t}aYI zBQv>YK(eSV1Bhq9DbKyiU;VCckTnqS-TS^KrRPX2e1Y1dD%G8$g@D`JFGgybZ;e39 zgEY8$OQ3U6Rs`qQT`xW9@t~&|wsz%%&4bXxRMrPK;Yo$9FR_nYc zCkDm}pO=C4eD%{`a_1cKeZF?s$CtjZ5E!kPSyLIEUhVFRjyEHKrbN>!%WC5(h)C;+o7LwKS*b zf>yGYP2~f~h|BPDkCeX3E&YL8 z`l_tbOdn{t5K?cY=138n!)2^+xkbch2+z^r!&YI3tLX>~1~Y2oGFCCpBVQj!^ZWb` z7rPzyx*aag>X3R51#h8S`hZ({VOD9T0|H*4m68CD@_=*c5n+(4Q`FqIh>|I4>RYsk z`-}AdqcGVqF@wpwK8MK=qn(y79mo645kd3S0YyabyMfrYo0?#IAg14OD_rMRcqgj@ zT^C3U9ESj(%7S`9K?drL0yVs-Ut}AU6XjSKlT(Jt81o4i;D72}E36bDex^jqodta~ zz0j=Iz}+r1FJwW(bT`1KL7g-^5-K8caE-2N4zHFR(~5t%gK>wYflL}9s1KS`LCByi z+Tad%i|W9n4b51%Ko8Q2+17(Jy>E$XfN5lcix9E7NbWLqTOR5VV!7Gops_hXTbeDw z?grJSK9;sL!y8;0-4KvQLlo$xiWHi~D{sva1nknBKJKR`%OioE!3;7l({b@Z8dvHU zfCt%J7{WaQ;+Y&|yT}R`6muz!#Srze6zgM~zTvP*@ou*1>yAl@HVLJp!Tsng2yTX1 zLHs_JJIzSx=PKG;S)A#(-Hkg1a&dz$8ND>m!Z<%{QsZ2%E86<7NrL>bfS2nInidX9 z^COs9TyAQcWVEWep8HjsFk0EBc@}PtrMVWG$ZHk2squ@NP!J>pf;>E75@TF4tB>m^ za`YdlG46FTV|-626}L}g4spfsBLSuh5a|sLT+qoOaIMeir$&EnV`pZ%|0Y~}uuh(LUwA`+0Iy~{Xdc6qv>YrS z+_mrmUREknK?M2*pYS5?r#Y54YxyOaCi6!ONiSuacuxycQ#WWDV=HFc>zDIB8aHvEDkItniWCW`S(uAD&KV(?Bqc%T)JsC}j=*WCtc&dGp+ zn{WYqKSx{e_RSJ7KQyF|p&`9!mKrdVOhfh{OWLM}NN`HL2&2|x2f4q{TN4W&S1l`# zrO_(fQ5&p}J$bBb-b31riL{nFC4O9=K>Oxn>EU*ZiL75hS5hx>Xy5#lF_Gy?a*VSNuWE-M}w3x8VkYPT-K14oz?xAr*)I&&Xqw-up!xV8^I#kqri;%EPE$)6}SL zNv7c4_L_!K{bq<7)p>OJY>bL9s~|mJU~g+CMX8 zB$$>Ccc@*WanNI0VOuUQK+we({GepFBM&3lomj54JP!{<4rY@cp4ldI-Ws5^py%=c z@Po@waAU%Hp%d0CXh)zIdAK%%qsEm1rpE(&*fy>TF#URV#JDm*JCAjY%LDY(r4~o^ z6F)VW-k2X@Gs?C;YCkHQbUh^uW_owNWn3OG=~_xiQ;%T>bDOfVn%ah@jB~-r-<_?l zo=J1ixHdp#2wOUrO*$SxTpOVBlyNSDEN>ryD_QgtX`5U#0W4vX&0QRoS^nIQ?l8Bp zy^d4pxLZpY0a{#<^4`>2X{!Zgl+ZOM7*r3Nws+h8zD(w|pUW%>`ZcfnT;_^Q=JlV; z+?&b#<>xXdgxqHH=w{*9*#Ir@G842K3}-;*PA{{B?ij3wf~cjhNJ9#7SOf`#l3o@WW9(s2z4HK!n1f36JHzDtoRfX<|O-G046z1XAi z`W!F}H`O_ACTMyH_?A?r1~*whk%skBZHj4_-xt-onW&TR*YSI%DnCEi@86t$zCXnM zGQ=L%f0ftoMb=Hez|Gd_a-qAl+=8P|#EE7N7vSUN)omBImIUYlG! zfEOwM)oAUabKWEmB=hTWX8O>=Cj3NNE$ z9ICO}xX`<3BL4?|PYlN#G-gkub*^|-SC1{GDf z4u@cPU(3$crGoDaoi^gEOtT=W8-rQR6RopJ@5^0IJ#6bIQ(Yh9wXPWUd83>z140cQ z_sT4Q(4e10>b((pOvDUwbXg=EK>(x8^rJq5Xysw&{U*#io78L>M7JNd$aN?pn`;~! zIjB1e2fLartQ4juE7;<$aS#cfATaYq(kocX!2&$?CA%Us9n||}*d~2CFJdq)F`O2Z z#N}2deKsFyiwXO(Zc^|3h)G5#9!J1)o8I#vlYi!}hH^^dDNUC;u>mac)@GMiMyFNk zxU-ZYrp7P};5v)h=0$o?og8ZZ(``|w2UTgo_AZW26+s`P0s%q(PJ$J7a#ZoeP8@-O zXlRy10C)&D6CEM|pulx?xExGpdUOsnGK$%NIOx)p=e)1|+=t^Obvgv~?QWL0copX9 zttN|Z&~cTTPFI62qsCy`)!dDy?+PgTqyh*o=J{eqV(-jHT}5*aED*Yl-W%1t4&7ve6k zOh3Q@((P_P2$wPcI@1;C5rShHr#K)8$}q+|HQsBKQ+#AxLF%XskTm#<9yoHloEL}4LYP> z=?PC{>X_=ou+Sy77LFPrriVnK7Q*Z7dIAsbxeq$$NBtTpuLjl){^FGM@58Cdl>5jf zIWq)RipbJS@t&wYnwtB(b|=LAcHfrr?Lx+8=1u5Kfjz3@P?14Q3zS9>H<~R_@$>XL zZsKE`Zb@2z$bGN5B<}a{diG1YlQZcz7?53*{tt2P7!;^PVqqxR#?iH$rljRc;v6#C=yNjo9Y+kOwT`~_xGq6ee)NrJt1mxp% z*f%#aU@v36iFOkfCMHM%-)*L+u=g=NC_F35Q5uqPlmbSb;g+(^YuKg=i|IrMx9QN- zKKv3f#no~~>9%Go$g~?u^V?;#0=7Ppo(=Ret!|DKvrRjIi~Eob5?o{oyE5s?M(j2B zF)o07va!t|dVqn-#~V`y(S7*M)HfS@R@!ughtVgqBOIkC{RWwy3Ur1Iq8lLB7&^Ts zjo!ko&~_L){ie}^W(zOc(>azgf!1NQW1e-#m_YY7;n7}vZ`YNSsgD)@b2M}nxO|HK zy_@641lo>zm|zg*I=M_ipCd(+;A$w!)T4RTnC5n~4Tk;Nyn(x=q0_+tq@T+KBRXA% zo|j#O>Cu0#?gMmzXy|6U1!`{41dazUa+HR((u|ZkeV$5}_EANO(9y$3){Pb zg&naefD%~1>u0+-4fT}!W`9kM-|sWhZ<@`tm$@^O`9|J|A?=Equ@sgmirz98_hJH; z4N#zqwpUFJF+Bm?Lk4(2Phepcp>5mLC>V>##=(IeqpD7>PquGw$^4VT(=F7`t`i~ZU|7Q5a=DelGNTSaWgmc(TCz}GqdkTNS; z#y((+;`1K;R3%^K&-FXsD$Oi|)k-Nv0qQ;8*>pjs;0`I64q?nU%#9d2y_x3-En>&9 z%_Sm{nOKiVP@#_UYYYuWxMvLWfrMq?e%1bawfXaXVBSF=xK?*#z@!)Fw8?mEOw(P1 z9O!S2jRD#_&oW*M$jjljq3fs7n}d3haN`^SYLLSP5Gp|#FQ6yIX7PU-+^P!7p^si1 zf*3FB@Qk2AlM&dEl&oZ7jO;3fvjVY<#5ff&A6D;=4BjZ8iw79YWs|dRBmCs+tnQ^ z#zg8A84aSn2`Lf{9n*>}Wi^_RrJ@?by(yW7HU^kh1srH&??(%Q-K|KF?;tvP4eXtv z+5PowtJqi{pd%e)B30u`SoAXz^ZNI@nMhZVBZdDm-zq?(+f3>7>A4YlX?{v31*X*H zU}d3#j(8*}^aHxUD8TR3OKxC9bVR>Mu{N4qZcKz>B32mW1X_Mdfy*`qb>Hkrknz&} z14%0;mV-hufJKAAGUSXA_yN+}1p$Iu-I8ZIOh2|z0y0Vq@|k`U6~gX4ITd5rlN-tK{Xo z?^NOgXRv$Zm#tXgI<_f+ilZHx4I!x-ANdY_*FzDm@UiKUB+@r&jyiAY|3wca`{+^4 z(NCrg$-WqVvRoRiUqo~IXb>`H_r;h@hX4y^jW)_Flz*3OFokrYL*Gj&N!5M&zhf*gF)6O;_fJ%W87!_@|XKf zKOu*@r3^u~h@&E9;t19bn=M0;^|Fh+7A%Q-nk~@Va{-6TWVy*8uLEoO3WIzm;Dl*d zf~q1M7L6L!r9jJnmfZwP@2W-%-Q0<8=%s*z-1m`-UaxBeH(+{)9R@SFUKRcw3N!sT z%n zg7nTvL1_Usroe!xS>CUV24R6h!eTDu<($@k6`HPaW=gg#@&u2r6@E+4(FiO8?DnP>0CU!^@-zR411&1| zvkh6l03w%?vevR~l(SwycR;xV8&`<-;x=(WP|Q;xL9JywgtM%zlp0uq`hJ6*rTDOam z%%ik0<;t~d;Ir5nmEudu7VeVMuJh|H&PXJ@+MAVFCjScL16^`RiQ<|tVtPtW_a$^M zwHjp7Q}f%55~llQhM=iAq7xo9BOHcPk8OmR){CfDq8HK=dA&xM*5k=i-9fcYup`L; zO-MmFiAN2hHL4G$XQ&n2Ic9|Eo_Q9!Kd2>)FkP0kj1rTcGxnf^)$=TDJ2qO}X($%g zqLy^AdKDiH(+r#vmGBys6cAJthz}ZJ_hhP%+V3~@xZ4yR3HdeIjgl~c8(~@_$Jk6^ z+Gb#Pq3QmsyAy8o@w^Rq1U&jEPoUK=z)=I!B!_tYG<-sJqyKO}+cY=Dp+%n5QgXB3 z(hF#bd$?bBRcIZVyxo$QHp292V+5FJ5hkaHaPfGU=_An+V-}WiFA_d$v`}mjSP!NE zI|?vDU77=?Zc&$(V3W>!HP=5l3Uoxz`wvt9M zuaKEg=dHQe)3avhIvjPdcd!`)C;-rWS(XTRWV*wBQ=*0K#T*ll4opc}|0z+g#f3(g zdeA0lu05|GBY$zO_lArxJ%stAE$m2U#3-Rn_|5bfKA1Lfq990vuw3q)YlYF` z14?lNOBI;P94I$T2syjZ21b~TwM{at7)~6}x6HSAic)0$-Fdwf(>l-c#=i4Tf9&^S zeCX*GcPJrxP3H6#8Odu3`rGiq^a>}youk5L zq#0qlZ7zHuKvzZy_2eZycE6&{FZwQus#9Kur|!$-{|))z1!V@&n^;js3Eh~N04seH zi-~T`b4V$ag&~+E)(;baQd0gel8C?gja}t;eiLlHVAybz@&1DDTfyeh?hly3q2oB< ze!KsgYu2n=#qF57(Q4=1yV`I0F0_nc6&JEy#KlllzrayhQkjmTl7@7mPLa}p-IM!i zw!i6@bogy-@8PUo00ye!t##K zlh^s6zHg3T+tUfhGU(H~3PdrsFQR0G@*Gc?hqzfBV>=SoPuw3<@Wa>ZtcpNnB(#}HcB?kHlorlB8R8)N%m3+tB!_6K2KcC|wJ_xUyN_iBnb z0eX;m80%{#x-w;-^vdev4tk>Vg!}!b{xDe05f&3*(beTjqo6(0OZq&ap=P~bqZbXj zv#S5#pf|9w5BMb(3;_#N^NwEK0j$iFSaSFhV;}U}Xgz$1m6;M(AHKxc4SpM|VWY;2 z5yDe|CVfJmrxMxemm5O2A2EX4J&&vj5BUw;c?1U1&QOqHLD<+@)Jyrsg>-#AWU(+q zDlzHZp^!P##)V8b)WJSJK4u-|CAG)jwL@an@m|3X!iN}@HBKGRxhNmBe{ z=u@p%aYyuX952gwT@$ zDFib36d>`r0IjbVX<{q#@J#w1=m}8Sq=)LejS_ltXoBWPI9lW(F1E=BNw)J`G2S6nY>T~f|2vhr+hhc6`hQoB?PJXxQ#Uuf>Sz<#RtRY|eO{Jz&7 z3G$M4vIKyG3#$3D+c>z`Tdnzz`!(J<5=t$$&3xe(LXfIg{Ui}<;l{p;?WqRSJ9U;( zLf0MP4E!w_*hq~My8Z~Kn67|Im`(bNY(x5q`-2=ckkEB$gwqfzb|2Xfh34ItqQMuA z>@;k;`%*a+yl^Ch;uy}t$~_HAO41>4U@(xMz*ub2C4;5d9kS;^uTj|`Ti2LFX;J^B-kLPHN z06wNB-nRjHh&qM!ad@K$`YN00hOgebj788Ulr>k zrPVqZP1-y*T(iSR*KHlI2NpehM~K;4c>XJl(cX>C}w&e5*v=6Os(w5 zSt80NT`sJ|q`x&rxJVWgRB!G^F2-M1dj^uiLN_6%%f2ovK`*4EdpU~vOemRS?6W?A z*TMIe;S&($;8Koq<~F}fpKM;0 zO%{I9k1hD3-qK{5^roJm8=E5ZC*29s%}o~T6X{QSudH%gKjq_C0eGwoQvd0GNI~Kb zrI$mZpla}t82q_tPPcxM8q99#>X@pd^k{dOK9QbS$n#*F(5n8-<{j zGUHGk&cR~9Io^cvLvY6?)kHkBt?2Q4`Wt#`p_k_Y=XAS~fLoUhT{HeK8dZJNez4ms zYJ}bX`02Am}*ioJkN!lp>S0HmLTiN=W<^%oZUy`o0Br$Jg?n(uJO zGPdbwZCQQ4WF&OgFgJO(-}eLPn{Jvfy}e1TOgGNAOnQqt>4y0hy+y6~S3il~qF(Br zZ!z6Na4w;p^Wn(XP6=+IXP}M1Z`wKE(odsyDH|!EcIrfuuvJ=338rV~TY8A@Q3cSJ zYP3VNBh;MIU910sJwAw!3y7D`Pto7VVQQUkvHoNFkYGsmnf!!T{PHiNypW<24Z^JA zkJ3UiLQx1w@@ORK?Xl)X9=as@*Gu=x0tO-I1v3BJ5K_V*1NZc@^S}N6{u}+Z z(3_1G-6Wd?y(#y&MxnLw$fo3Nzx)ToN6@pd*fFgh{_P21ccn?5@gRKCbH;cHKry>-R3LK`Nz)aQ- zR@RLjkN%d|D(3>|`@mTiF-SK?tD$_Lop}zm&$YZM?G{BIjl(g5B3KQOLCR_h94^K$xKcLB-m)hzLARoy|4W}lf5=03`sD#^YWgkE{=qpKn;LwG;eH>b zGRq^`{J;9cPmjRx-NhlC1WHz3^!P`5u zkNs|Tp&JiH;^D zM$k*MKo4}sV1IdPpZX2Hj|LG|_se+}y)9$DUX2-Or(X|BFkL;*0uD_2-Jk@$GAF|H zTRcMYhdCC0&>!Zc5X$tsK|(MMB38{;BQH_xU}vx2-@;=>sMtQ+qE0n{?d|~9WCl>u z=a+9eb_Ct82Cy|VfXSHx%hBNH_={?Kr<}v7BdPeybNDoARWgge90TEMrv^O_wY%lI zQBG*|aO{Qpk7?x~_)Caw2I&rzl}$L8xcr2v9xDvOH64@SD8(#~=t>UyV|;2P9DYRB zQL9EBbd9M@?^WO!0B4vIrfWuuF)0e)AE*_a827M^Z5r-5suF|zL5C(zYVxB8B0uDC zv!g&#Z1V*gmZ0W{^6e2d97@_uqV2%>7&0r(1JgzhG%ZW8Fju3>Fb8(QK@J6AS44{G zyPD9l_PC5HqII zqAF?mC#M%mK&qtjxm+`DsX;XJQ8--sjSbS%Pz)>K;vnngl-+_aMXO6Zk=V} zboWuMyMf*}a*sUWGXYdx4n3@};mjZ7F*>3Rh@y?v~xER~hP#NaVcX)#{Xs!N1P>67UB5pcz^S&!pX zjUp&p$i-~mQ*4wo-QgE~*(>@}yxOXd1Eob-FM$7hl=UEqY>y_~Ubd;K;O*w`YCje$_$bGn>OjTjKhpJ#GTFZFZLy``3hzBeuXlVhkkiqU0#_Iy6rf; z`ZSgHRP+J;w~hpD>;Y}CNw2-Biv?7s9y*&4J8Xli-K@oPwPMal#YVCn=zFR zpf+7U(qWng^xPMqHXsq_T3U(~$n>$6U~0it0wQJ6f`DU`(DM*TK!W8a!QKvP6-&ckK^bWnqDURs&& zPs4l1K|)+6c5|$FxfXzX=y*gEApw8-51{ffW#O%zAyVF|;~W|T(JIEaaiM-Hwa<{L zNBS{tI`V{>fQ4SDpG^1pWx(ZP_@5~n53J~Ju{4t@%b5l{!Jtt>S03jW7t){5C;du( zzdpmFTG>;Ff&I%Mt1Wm#V%Rqv*NV0tmkR5TbHv>%M$f~ahv~8=i>Y6CnC?+~RQK!% z7*D%=0ZIB;dS-T?6tqz^K^>9@T`NZLkUs)g!lD?VaiJchPCp&pKZouaxNyJlNEaBB zN!`cwi=yyCrZqD}XRt&9dJ(`DMJhDlc+@{?l+g0y{Ns7YvHq0(6^o6R7K)sX=}~+J zv}U{ST2phIcU)L8(y#cys1ZZjA@P`0gek%%Wp*e>LbGF;*qp(5O}K;Zn=Zu6wBlvpNYm6P`H@)O|)${{y(@&u3bzj`{<5^7`c}#zt4q*y9kl%gHK^LyV zV_&9=>(B*)CEyNt$Fw$i%sVP|y7L{(Ju+^ark-MlFJKA1C>b1!#bHkxp=7O<$NONv zsoJzzKlh;5WyuMC_O*B?54blmdrML7WZ%@ZFH`QW(OE_`ZM0v%Ejv9wlfEiDeQYNE z$}#F(kcL(K(K5(ndWiI1ESmd9J7I?AUA!4^5-^X(RtwrnoA5T6poX_#jNkuNXwL}K zQwhjeu61|K3BGlA?AQIGH=ro@FnWGa^*qt-dE;pCLf2evV=bGjfQpROZSpt#c5XsD zMwm8bclltZ%YB)mx1uOQWMG6a!ql$1o#b}=)My8XU=rRagI28Q2)L}wN`>V{3V3WS zE1VisulZw^)F0XetFSqZB`5Qzk5!s?{p3Cr(hF#QpLf3R_@+OC+W{85X0k`{SY`y1 zzvUNw6h&PN5k_$3z!AJN+JRRA;yhjK$p}Qy1OyOr^g>8ih<(uupkmi2)2u#kC*FCI z-_?`SRS!o!`_P1M`+1Kc&j{1?*$}^xfp}-8=$k0YX_#${Fuh%hRbQf?Lf4GxHNx~> zX`4|(*NhPdn7@_E9q5}A1AR(OS#TZ8t zu?F{1YQ(xX5nLSQ=mGCIwBi(haDNtvvj)B|llLz2j4QGu9*7v%Eep3V&N~TvXuu6r+o?~UP-d9e8utrz@8+QAC zFscs=9NDBgQY2D8jj<;GDOz=l^ zSul(BuE?a{6~ZZ6S#H58qul3hTlo|H+G`Feg-Nj4H~z28lw6OJVzPl$DNRDkwK?VS z-3e#-&Hgn6c?|(7)Qlw)YvJgwB51KgDY%S>giE0_aMKv4#aE8nM<$rIK|W>b#NkVd z#MmxsSsi0~4uMqnjC5l1t3WKRPX|6`o9+>V*nr~HU13oh3U^mnavhA0mu3--sh{*Q zMEX-`a4&l|4PfEwQ{ck~*+O|i&_&&ze6p{|AKT($ETnX(up=fe$FgAHfKu$B)iED) z6ssA7(#Js4p^x^tY%J(EbEPzswXC}`c`b+LjSczrZVgNK!fN19)_$HyK>FFVv$z)t z_njwf>#1TVD3L4bQQfp@dDCA~>^F8h8Uw38KZ}1n7D{+70(&z>??O>dLqxj%Jb0wN zQw+mVG1DvOfk2h$1p0rc*umE?0hprP_fS;m2GebGJVdko5Zx7FADp{YA8K<48x1b< zWERcwORYz#UD&5@L>EShK9Sxi?sXH45_+>Z0Yt1+hOY>5TJ9@TV6~yje z6k`=MODpvR(_Yc%{#M7a!seD(%JWDXdzN4M1-G)u#=GE_)VT*o%0)r8^-&*R1a-?i zOGr_vn__J1Cz9Sv^CEFXJ%Y&1qd~PwLjhmZG?3sLI$~%0eSH)j!BO3Hv#!acFLPY3 zvf87ju1w|%Wb#tsWtcru8s}y!#I_zHqnDE1#*+c2tIicK)(^r7l_3;zZPKn+xKv1&kX^SUI-wlgz1j~$G9QDw6a+! zaV^#Wc3HM1dHZ=_wOHT)&ulkJsIwR~=xIf?q|Bb(kh~Nj&Ydoth9>QZduYIS&+fvEK#L5pPy3e^{o)-Pz)PrsTf;Z zPq+}Mq~{K8l<649^&kHw8o<$s-VR~^HVYy0tYg_`X~!}B9Ho0JEv9X9mq$O-eU;cA z#H*G{+W-jkOK%>c!w$0M&AOkbP`PM=pG0Y+mdwnaMx!!M|>j;^ruC3Uh=z zku@5?KlGcrO`1adf*knevMGHccE`nnZ7^V@%4SNRRa&(0 z7)O4%Tj`V7N(JlK%}~5OM2(r1@FTygwUe;D!#hsoGY(_;P=tz9FaOgofO@i69E#|Z zb6n4s#gm|k!eI@V6Wgl^XHxS8{U_@1mjCa5FR#gf(oB`2{ulw{{Sw_h7ca<^dsCpH zp`GGP6GfCE+p^Puf`BfXZMAST76b*`Fv%p`|t=?|r|kIzWBKh1pl^b7$Blml*t2rBToLZ+VcEo=nPYS~;Q zPGd$1b%Y$FkanDJb;w6TKH4C9t8XX>%?Rq>dyYkoaxkKoChbAWo#&wcWwLX@6Z?29 z7&(%nd;A4~Fc7XNPM(_NpxV8tW|Wxp>wdjmi7!HLJOr5HMZE}?X4%GTs`Zvf-REUX- zF}-(A1S*&k+T}R-{{8pRV$hgMW1JxNtq&Y0#8Xfgy#fe?Uz@Rm<}Ny-m3}QIHAChU z1?cX10G4nsT$}3sdEDHkem;xQT4juxhk?iFq-*C{^cu!}iz@QEk&s4hrdJJEg{B56 z&`I6H9lB$trH-XIrVf{+yJlKun3sUXL%skRcTn%YVOD8L0nj2q?GQt38Du$HI};-W zp9qR5{^L(DXCm@aV>fNp9mA#vHYQyXf1*pn-Wa9(A?vV77pK5$kL(1Lv;&aWNXk*2 zR1=}2TB*ijibuTj{>rH*YOa~6>r8OXEZJF$`gv==nrmf8rp6Yuwu^33XnVux6bu$H z{kN=WA=nsgKic6at*Nq@o)RZP><8^k>#Bewm~DU{;WkgAC@!#2W#~RB5~9ZsNq-RT zAdo|6`EqDU+#kpb0+tb`1GqU;q}+vPLP>GpZ19MMXC@%zE;>hqhy#-n@H~%-a0oyk z@4M>038j9UJ!sPi)1vQzc3pkg$Br-W@5~f^`OFbR+R^>TZucQ-k?ud1?pw|j9mzp_ zFfBSW5p-qjqrJ0y8YE7){D(!|b1kMT&v2N2_a7h`dhJv!KKQz<713O_B`z*vJIH3g zI%G~&-lWvj?S{JPg*cr7!Du*5&;Kxm9ht5?qu<+DYs&qRzT%Gb;5k_%J&+k`{xrX6 z>MTV~I6yq(4t64S$Y7tr_^3nS`bB&&U3XT3ZWPl_l<8fqpKg*5ee`FZhuqlde&dVL zxDloov*EaXl8^DOOwm>pb(JNLyKofJ0|Lk6=$al-NPik1Or2-Cb1sY+NC6`@)kQbH zoSXO3=zJ#J^4$NQ~c!>B2oA5GLhYMNvEb+TSa3~D9d92#U zyDs*--H*1wC2{Z9*;zL)slDQEM-hWQ8a4;0zFoSlxx{axMV@bFktEZsp8|&y88yHI zsk^<#D*Y6Dqy;%YZKv>W9XW?q1w@QR=j1s!xtc0gf*_)}jLMpbfZCt=~&Drox z-*=`H#Mc!-JpJWNheMQ1`DfF~GyPs~Lgn@rBTVZ{AowFD?}@XpYS))w)jn|+Lb{OW zMM<88Ag+yPt01n2&IB$@)PLk`@D;NI^9r-TBWF6$zXuT>w%*Odwz5Uh1d3z(FJanx zro)8@hQ{>hXVbo;Ab4Qmxh!i%4X)y)2H(eTLA5`C4`l#Yf<~DB0Ok$Wb2k09OT#Gs zun`y3sQ}NL5ku&af#Ul%ZWRH1_pAEcT``IskT&ATUvx_YXNK;2-tz0#x`tPvoR^gt7j1QXs;3Gkq%O`?~-3q z>w|C|+7i2y(6>o}J?=%ngqx=W3C5k)w$M3_eHb`mk>!VeGHBQ4C@z4o^%xY{P_TX) zvqXKzrly46R-#g<}vj>k^~+>u!}Zn4v?vCibvPD z_e7k>muwVi8uJI~SE3u)=6irTDL9r|Py`|t+!s(Nh%F;nBx-Hk2`*Piw?bNWud+a8 zxm+5hiG$LN654l`&s;k)Wp_(iFezViq0RMM z`U7&?4Wb2SyO!w%XWmC8jv}%)5=HJCOiEKGV~lh!Y^r#L0HFkm+OUuuLrtfI$aJQdjx?I4?qei3DH<)Y0m zaWEJRQnMxII26qKqee?92v8;%ruURZ`34jH!(Q42sFclc<0cJ!>}w7+T2q*6MSRU; z+L$TQg(7y4hB;HH&Z1G`<$&(s&4VKrF06xZi3vYFcRiB>kzpap?6vqGQDEN7PEcOIq}j=2%R3V>>tL6>SgQp0wyi&7nJz z7QLW#GOa=$({p&P09SD=m&On;Le<^g9d{Eh^PyP;XuwWQ*eZbl{3#;GFhe2ryT`%V z6sSem$vQ|DP5w2cV;B1EEkt_|654%BQ6}?`DwDR(N-?b$2Pk?19#hyRY5m|4 zXTcLve#`Hnay!2)Rz(1UKUnG{%1c}o!S-+_0pK#9n7$r}O}icjUf>)QqSc$VUR(}9 z&mz&?D}KKjOPzK-l2jYrtwLEs+w?+n#0b-p5|#7OSq|N!B2{n=sLk>fWpvXmq=^G% z$0n^G;_nNOByouk78>29+5}|u0WO7NT7d&7Q!1FC1K$ULWwW1Bveg(&t%p+Zb2s1afX__}hZMvDA}R9X+= zOd92Nl(ayj_yaOls~_U5+-McaBBR-s0AN6$zfcPW_HdzbK6QcuY%VoQ=t;n?21}m} z14y*6HbMv0&lG%R)?ov6Pv1BV&a$snhyFoB0?s%E-#ef$ruC)f*Q<<@E}2p z@I4y~SJHcEkStwb11BK-3sz-fhZplNX2zACkp|1E$PExjkYb$Av=s9lW$f{nZ`B~C zMd<)kFpPVIBi{LD#pV8T?S53)TtD}FwsP$XKl6QLwsSuR5z~!z^7=z+f$AZ;=~1mW z7&pT7nkb;@mPa)?%DpY0OgBHOB|?ySL|X&rSBx3NqF5C$=(8=-okBHam}kMe^f5T5 z>39PJmIitZ6JXMVdICRblfH*GHCY9;Nq6XxCW|)eolK9S5Iv+Pq%RZK4+%v>yhnc~ ze$cft!|z~Pua_Bqx7L#NVAA|HXfCJ;|L7;vC4HW_u=6S(D}M!6ATGT9u*gcTym0kq z%@-JIuJ%j+YKtIv8ACq_41^%~&|pU>{$k4rQ`-O9P)Xv>^1+fO@-%o;9sp#j1lAHz?6IAo!3mnxMrJ}* zv$&OExd2+a9P5Om^w)1Y3`45uPoYCvx(Y;(OPHb^BDCzr6;^ch&m(A+KKwQod<+I5 z4B*2JiWpfhq(xhRIHo?hVU^$$xWX6k$fTSl`!TF$;C!Jw_ye0Y2k5yBgO9Q)XQ9tr zj4W{r^>5Y^bk}@~okTf;kV@LTOgs}K%C;OK+;)&0B5IvHfUptwkF zFT$@xy`&Hx#5SX!M1fvL6);Qqk-$l19-hllCos0Dp?k5qlWVd%ly}_Z_xUB9z3ld$zeezMwK}$xMRz3*&rfVP5G7Y}1(o&CUtwxwqida6| zq9vGqW))x;qNYeev3MMO2VoraUqBKG;o-VS+-puJC+HH9Uj>f@U1mWrB=^BYRs1SU zspAfET67*!S)n)%Muk)3rZzfe4_si=OsBWf&#aK`1e?J>U#5wv%v!5dBuhf<>

@cnJ-rQt3GAMlJJY?Du>2OiTB zT#S@0-kRxD7y{RYnO`ZOY5xV_rTw(5NZq@@*uoza%z;FK+75TRpWA?aNf1Pz#R(};qbup*#l~m&_lHFajiEPGs4vOfN+LIkE=Oe=;AQ- zxY}v!EX2Au0#9)EAl%VOm3<&kEPotW)sYml`o#hbwPF#gmHS~BR>u%R)XKP&7D^7d zOjMYxP%xBbMYB9sVCg(Z8tS-KB`^< zHh}eo1F-&S7S{9c@k{rjbi0sC4-|J?Bxi`NE)~-Kk7@KyH?W>txU#EUEox zsz)Fn|0$@7Hh#CC5&O{;!5M)rMbkjMUJh$FVj?EdQ0&Sm4f8Sd1Nx6?m)0wX0(eXj z)8+>O09YPvl0iaY^zr>jG!TojbSrcZ>Hz-=xQ}rmo3vs*mY@{)>`&B54vy=Itxxm6p%OO2`2zFw_arkzj71jjr#rpbPBdO#5bQ77Zx zF}yI@O1SY7Dpz5;FfSHxvB*8RYSVV-5|%g&_&K;b3p1U?;K zVce)l*P*+2RCjQZ`*L(agU|}vp&_mG`d6$e$WVa(R?V;IJ|F>K9HHU_W&oS{61H2g zL(H?dh;7PqXs#vtlek5@(bty@jEq+LjbMUpI=+>u95UuEbw~nzTm&a2Rt#kE1IY;1 z;2l^JkPpC4UV1_l_(;EXeG*=jdoWnhLxKgCvW?(g+bmtQh;6#SzYAyY!`^bEZCJJK zyqqsGE~Hf`8;?;DST{WK`tUnAb{t&`*Go%Fq1V~ z!Ay(`X~+FZD7#ld1Azl~6mDXW)TlOsAjVJbhg^YiEZwR(g=~jJ-9btJ-h8dg2lH(O zW?KA^%zO!B)W<1akTYLc?`j`!SJ;ATEee`6mjQPm5t)e?zHmVcri3=si`U7P zp$RR7}U#Ic7CP>ZisgfTRTok)M9KHl~9|IYC+g8y`kO{YSO34X6cl zvxB3MHh+J=ng@vIySBPA>YewCeATm6f!HpeOwWNJM8wZG!n6v2i_9q!x>4dW`?N`O zeHpUm2_N#i6mVJY8ap$Y52{S1uEz}Mry`0EhL0tlp>X9^zpW=wyxqhE1nH$3OL9QxW&lzDlh$61$=Z&peg1)KV(n8*1kM8zko^2-a2f<7qmp~NCXZI%~ zVWWi7z%$iF7zc~vaMpHnSgr3U9g%>L`wa(vF@P1ei?5@jh=LX;))tQ9Xy8_L8=Y%0 z49D4t)EL1&3S#&E7Rf9_CxGi}#B~u%K*UO%Zop1J2bg%s3z1ug9og^h}@--TNxY~DrlQj0vLP1iF%kV~He)8S6h z^#eP{yU{j}a9C>VZb1y{?lxe7A`iVm_(`K>hEpnhoqZ0GR2sFJ}_O_PmHO2U^0df13e2|J_N&IP^V{M#xd#sxfcCc6#)Iw&m+RS zLAD^p0$>F*Xh!LEDDq--)BFh28~Ba&cIA9)i7AsZ-@<7>_7kNK;RnxaVBvw%0pr77JyniT0@X*~ zw`f^V90_f-tCZtZ>kbIM63^Y6OyOj}bmLRl{G}oeG9a0YFl@R9ea7fHAaWIcvrRqH zx7%KGo9-$~_wDuvfB8174gf^I`8S-7=+y#H9{pOpB>^p%H>8Qh7h_*_cfj6bMUxt3 z+qBB)>cTzr@xNL$%PIq8bXB3l^zRnaUH@h=y;uM$=-?tQ6AQ!b2))9n7GLb12VcgZ zaFrq5c5f1|UX?KQNE*|f|K@h<1M+e91sW~x1sV(w#aG1f5%I7126BON#cC-T0m3j; zAPpn4*}h0EZWP6l!dyec5>yLLU@t?IzZkA;HxZa%IZ^21HHlZO(2;7ZU5FxT66Dkf zbYCP&dho6wuA0G~Bq=}zU@_sjnr2%cdmxUZ6=YjmKe;cg&+jV}l0zxF_xlz-6%KN$+X-5SX1KlQZ!|PnX5Bj4l0SMH;6|<&Yrr;K37*PM0P7lw^w!zI4 zX&?D%izSU|I5Z4`8#%?1gLf2gF+!7Q`{?TLTlBsxgBG?+gSew`t}+6q*@p3AmVA-v zgi#H60`5g}TT4s7Z_$25_gE?>Ocrtl`&!7p;KaZWF1$DLnC|P8eFG*p+!#U|x5>%S zW_q+U35`AA#K}VZUgUo`I6%Lk{%vJiyA9W=%NuZL^7+mro(=OKCn{muwoOB{N*+`9 z=QFoUrrJFA#|!C^f3@(wYnomzaCCQV#oqMS)+(&6G^_9^{lxu#Ji)$9nMK>N{xHp8 z5J7tHe2cCLIrNb7Z+^TGceqnC)CRQiUk_ag?`5w9_bQoX^e9#z)UQKZ=?>M-i-q9v zbl2xv_~Ny+t1!X#)Sw(0mdoZPW(|=4k{~Qc*s*NfCIyEQOfCM}Lm;}9|Jp{}%4ynA z=y=hFHShR{xkCo+22#gEY!`-PUqAyu1^!||!=~k(o-^`CiXy>hm0H+U3`?*sXsR8D zCeRBOxSR@bMu7e%%ycE%!Es|E?b{~uAX1DHd_gadFBT1vGGFzWNWE>gDC^&TN^6CI zUq~Lav{oqXvd!8aO2M(f<`^B=1i4a15Ttz%X9_Qq!i!LNfme8en7!z#dzJ0LL>HZe zKAD!?i@Nklv-I5$g0-m$3tlAKv^x?6D10hIVOr8Dr+`nlVJu9S%O}&{P&dTIRx!is zYUb0HA6fL!1c&WqQ`x3XKeA{ezLo;)HKK6GJz$V`vGthr_K(o7+#x|7N(Kg7gCMMf zrF&l{y|+ng_3YYRfA%MO2PPVN(aZ6|fT)FvjS~9cDb2Uw#`gK8_n>q;;~`OtZCdo> zZl+I9hd4z`ehe;W(`Ton-!zB}>m76Z!8-6&vA|rlUFH@kxa4EI=0JwVCngumd(U7< z>9@^}h+($}`n{|vy>+v;hhBXc?Qx8Lt!C+Oz_=JYmTyzWj*mJqMVM^*q!~Tkv>jG1 zrZ;fYFwWchNp!<@_l))y-~+x~+PX@niQbV}qN_1U*h7M>=0W^s-j+Ub0)@DQBir>T z2iBuJdoT^^E9j`-uK zXO{*S2s{C`$|9*YbQBe$RuIl*b2l>hBq|Nv6%ieN#?0Xq*uBNVQ??%4x7{l z8xYht+)r}$5?&w0dr&YrE2_Q*IKmtjQs0ri^xh;aziZYdp>XLtQtU6wFO>!H+292I z4fc3!E!bgNVLA@eqS=^;_Y`h_*{O^tuWr*Ex=GC^y|Y>Cr(4tqB)L~NYppUbHgSR) zE#f^OJC26gj>z#toG|dzfRGC<+pamZ_iT&luj;)CAX@Q)YaZ41&|e;M3Fpd3#Un)g zHCWM?;sV#B*au#*tHr8H{TRE@SKl_Rla7R86m+t$jW zU-q=-;FN?KAFWj?B2!Y0FcRcx78LS7nUx-)~rM5#O3C!bQs>u998G06(^PvNU zhh|XhLL8(S5zyQ(+Ke#Ws1}Indt3hr>m$mPKBt>*7r;N#eSaEp#jyM^6DY>C2_4#1+&B!-!`uR}xPWwK5& zr0w7fSuEi0ZJL93OuB^!Fx|6FYYiG&8p_N&i7pNP@ zLwW-4EwujOqy>d99+Z`nMU0+NOmZ`T6e50TnTpui}hmD@8LY)7W_aw(076@}?1JU}PWdI6MahYQ!* zr!^-iY+GEpQA$w;Lf^xfbl2 z_6l3#AYE4ZC2GP~fk%|es0sfF_O-AQa>IdR9QlPG0<&0-KMoAg!O-QfV2Q9uw``D; z1>6?;VK)cnZMIGN;}&g?KI#*WV&v4=rn~#+0>dBjDhxtU3dP{>JcJg~?x(e0FA490 zJ*}3-T`qoKdRj~18RsaQa5Sba<@maJ5_qZ)wFym_4F|dTzKU66zw@=yxe=zj2H^X- zEPSublzv1?WA?!%uJJ&fNbf$a$z=>kGD_$@7u1_vP~QZ$*PUZYWGkXpY27(}v_nBE zS_~6z$NHF0Acvdu4F?+|PzK6|LIXmKiA#lKC_nD=@YrGcL)i}C#OjQj7K%|5vc=p8 z4tr?JT#ct_)IudUgj?;cV9az>LcQAzSyxbqJ7~m=!*sX2_l#KuL%?wy@9AL=Vq0vN zxYM8RfO8RZ-C3m=?r4DM1d&gPBO6Up|N<5!3sR zY5jB)K5YDF(i@LqDSj-o*Dp~OU_<@Xejz5ZpProvlEKyaHv8+eCf|qSUIoW;QJF6l z^9ArBt%E+nldt!%UPy26&=5U!{TdjBRP@yAJ7mkZ*~Gr}ODvy5x17CCYmk)nN&Ahk zET60rxaI^5PU*vy#GhvUhcpafKRAMcw$+}7H4_S{Z^-qST>Jsz7uXi*-W-k{I}VKi zG<)NeK~5Xtk9aFaoMySQ`VXQcih;PO{rf4Sgg)7!Il=!QZSMkKS5+nc-+R`{?&%8x z6p#_5j5x{+O7S!~H_35+_5TYhGwLv-Ge4b6dvBVXme@84NlIUN6mCn>Hb{BSj9kujg)Y3%{>@#vb4sFb> znbMw+<{3vM!0!NfhWg?)<-274{iR*9xm5a(ap^brDSdWX>9)T~+CFPDhdx+%Z#;k+ zx$XhQ*n5W+V+Y_MH$H$liSD~V`6P(Oz9>0(e=5YoFM1qr_=8C&8?+PCmnTH+3t&4J z5$8C$_`cMh)v*FNe`Hjl>CY>GH^*gW$rYPZ(@KwbZ7*e9Nk%Rf&ZmxwTAd|VlaW04 zRawdyCL^gaBgruGbDI;z!kJZ;bF1cUplj}QjBiG-b8Zl~?RUAspvfkOOx)SWV`d6)rwbR2NP8#}l+|aL*hJKzjG+4t(vmTAy#-1+im4{aHF zVp`PSWw!Arr?Dw=bnxA%1Ko5SUY6%Z3TY+Wq&fe(lLJrx=6>dFHNl`Y1wxvtEm1UqN6PK4W@-C{{w?hiycPG7CBVCL9apo zjVq`co-T~kK5m71+yv59l!FTEP}$|sJ{7oepb^OMMz@||rq-f7loSy-vc94;q;^(+o%d-+Z1O&)R5GApLvyR zBX~Ax8#IA)luX^jTpS$A0>7%xZ>TbJVk6I$elmwIM2ikK%{4P0lZp}P#!WB!J5sO_ zH3G?xwHNTN1ExtB63F7gIzI5uB`Oe}z5%u)s36tWnjlX=%T-sNu2aGy^XOmOzw`u-n+iJ6v8Y^aJxagrQAO?gje#&~>^Os`HdXGAgS zpi@MQ+dCn|!tMz*2~^~eS5pJ{V1gYd89}g_gGr{?J*g3sb5BBE7o*^gQrp2a_or!W zYy$Tr`8eH|qs!Qtwm4a2i%gp| z1W%7`{E2FJm~+MOz)E18MLP{BE03!TjdSdB;m+z-AX(YmHReu@kXXaKIo=o&eINcX6f1&PR4*1X!D`i#A00 z=Fs2nNpkgytd>SRL2unf+u5jgHpj?>0$jgrW3m9Ic=A`t+7FTiHMk+Y4%0 ztr5AZhgIq6ifm4en5?OjZK(nIL9Z{{@M`*j(hAwrQ#80mIj72=o-rfq*j?I6^z5VB zXMaj*4?F36eY_kTc-iG#BaBIfkv(doTmH6P<9N7zK#-Zq2^u zE6)$FpMZ=NMpkQ_B)S;WnyRm7m9vi0LZcB)A#rTdBo5$LhETK&71 zzqgpcB_i+jEBJbM=w0Zs(mv?%OmOX1oz2pmJD(z)R~z{O52_j2zO-n3{irp8{BEh`S-b}6 zErk``UTWjsQX34u+SDe|?`M2A>1n;OJxN8V^!Wlc|{XYtssc9Ns1HT0BkCMq$>&eyWyV_5Z*QrKFfy)(m(0hB84rU9E;noEb|*t z9WL^#Gp1U53y^i!K@vK%+*P2qfst#LMRenCLnt&lWFnl83omK8#>lrJsjhgkCC%%t zSM~U=$ZP2^R+Fs2rph@+1YSeB{eBZ{%~TCTYO7d!OIDG6VMmr{r> zLnduAIroONkr$Ri+~yiZ^otb%nV^0f+{wKm?R-w2T{`B9oIN2_zm1&xFGgN3ulDt% z+EyM&kwf>m{~`}B1^U$!LL_wV7maL$UjRe4m;v{Tasiv2&zFrpL2cv_rd*{R_#CoTR!TO2Z5-G7m{mJy)OAdBg9 zF-QgIfAM7ZimYp-&kf6f-lKEamvpf^?jmA+rvFFM#dAEt-S1Y&!<{~it0nhL>Q8tB zS`puy6IUEAReVD$;;rXa$esIFzM)ikw#HPAbSU}%lk|Kw&sIkV- zMUTprngy%M{Z1jfY|4oHv$Q^hk=IfKU@VXX`m$A>dqvu2lfYIsId_KaP(e$^mzyUGEl} z+L3PVJ>|NvDnwD?ce52hztB=W03i++rC(X!lPLkhZW1^H1 zP)6wKddJ>uXUIL>KJPR6c65AXkE`rjIM7k?vj2MocXbxsg4{Sc-1j|#+dDa9{!1WCK)ZvJ!&ron znj4L1AT5XLSy`i+1!h3c!gWR^Q|r?sB*lU$6SxCuBM)}bBU|3;0Ww@3P+EA*HmFD? zPD=j8=|OF9`;0SN^uKP{D6;B(6!+*D*hc;|55AG+8LGdWU>|`!n6km5x57UQ8`Z!ezeJX zZS0jU#4HCIc{2sWN{1Q59%8~;+V1Q1OPD02h<<9KB658 z0-^)9*S}=I*YZeZKQ^A)hxH}Z1!JS>n@T3|_L>Z$%%7(T zp$UY4R2^3|d~9-cQ9QvwlX+XtIm+5&$N<2<=#Az`Y8B$Qb>=+z9K6*4FM0+xolelV z45a0O@=pFas_)56TTuJKS=_9}Jhe)xf;iz!*T`Mz0*%TI@CbVL-jxOqMrl?Pbp~)F z^XiIVt#hI^VD*sKc+#~6HqZ&H43en8HJ;p+-ezlaz}}S(?T5b_UgT~)QFVR?dm!e* z^O{RlA&btM`VZOqgFU`tmIFsStfwT!%6no0$;b3KI7#CJlmOpVJZjruS3ylj>ppHU4|Sf6f8bGYGC;qd1D(T3Qo@P22KP z%de%UnVb`8%vP3Vya884m$hupE7?;JGsUq-v{hybfg-Osw1adoj!dDadt!>d*07k> z@Tew}xT9)mJ+}5OJxNQe7|@pXLO8pP@ghv`D#83)XWFD8`3Ls4%B*nT05*dqcq0oh zrwq!8#Izpiwiqwpbhg!5`94;QU4xjh;$n0>rEV%Oh4Xy6V2fbuR>wUwU`bgk2F;yK z9?=<*pQJ}*tNr=o^axXRyB;n&;`7;+oJ^OemF?3pJF_@5of^UrSpOV=*ES!w4PBld zk>A;$-?wf4<~`c{UOI%vKHWWN>9A~Y#yMzd-U*3-%336CKVci(kRFjI?a%Mpw(on7 zw!i&ucr$N1r`jD@c~)sD&OzK6DMg*>3{1n%q(kEg&bX@O8#Mh)-#XI=IopyNumF3q zQF+7zonz#|)E>nw^kC*#(sFNVnq)#FpY;vzM#_9&1AVfAG*V{`dTbRR&Neyz%tmEZWdgyt z)rJWq2bKU%yTRnp(xB|n$1kT)HqKE*ic7U*34+AI8>Nsd%?bx8vrP_jUHMpO4 zze)g(+`mp-U*=?a>Kxc+`}~Sp0XF%nPfJmXd?jmSMrh<;d?UwGBW?p2;@|SctdW_2 zHup1GL!3|}KfwZ;;3jI2F8>(MTilL8%vGN?a+^BPC&}z80Ak8+#N-0&Ai0Lt?tWKE zQR7uLp3!BHYpLC^>KI2?Z$=BbV59iO;0g~A!7hkY8d_BU{b5Q?I zqp3J+Q+d`$r?G9VAKn^6*Zj#D=^ivWr!v~{=O&HGooN5f084|+&acz^AS(j>bgFjQ zJL>HcYS*`P`KtJ&)oB0280lZG$gNsTHg)-)ayy6PDMe!(A^aPi&W(@*5ISlnBah*9 z{lhL_e>lg=54-f4Ms}0dtC1~~xT?$7AI`CIl`TPTCX{=~9qID*hjXlqkelP8O=l_- za85+;3?%$PeQPfR?hNC}A5sPPoV0wW*XMTsR0^?OphNyn?-tpk;s{yIqVn6bKPy5b z|LUt32ePQWhJR$ynD*xica|F`@1dK(1lP{g=I-iLx$q5>nNYc_bCB!)%?Nfh``4XV zFUX{E=C2>mYH?{Hki*8M#d-i%79a%ZAXQp=NolQr0MD~l z!&_VNL}1pqB7d4h1OSdTu2$8GuJAXcc*NyAC+$h!V#JxdCxy|Wgh1t-w5(b@25ZZN z+RKyWL=wBs9p5N*{~P!pBOe$s!pqF`adzAYbBbWA$0oF6DmvfXh!>2rUT zHqPhd`5vxWa8EU+Pe*%UI&^=PmKS?`TDvekV!sBpvkUboUlUb7Ql5QR^^0xw=WX?i zZT07)>KEJUz~EGgv&RI0;mPlM(7=~NvB5U;gl%SnZRUxnnGM=ZqX|rn`x#lO5qWYz zJDMuQEMDrnrP;rS#fNipxhur^ROcU@JlQjD%z)CGU$HDbr05MxWuqO(+25o{oB3R%8 zhe9+^16^6F6mSN{aEk}`kFxJp$n^yjd?930Vw0C2_8`hRCygj^ zYkJQ*_ZMk-q#GeS^Kzv#pw=l5cZ1b1i@FqS^#pq3A#as0560Z#${&~J{T}As49VmH z4S*V$8qQ2lJnju#Y#6WXPKPQa=LeJ7*Q0pgVWOQkecq#0HSX8dNCrOc900jM{RzA} zOqg-uK4SN&;G9?UPR*&cr`F2iEbilreIt`tuRbCY7U2@Nr9$1Y==!hhMc>J2aE#}6 zV1Eb7gJ~#s6+(D2Z9o@e1vmg|XH2k7yiim)Y|k>%jQ5DK0$lHLMJgHU@%4vutPJ&x z5mK$f!0m(HkZ;8wdbW6BC1)xz@#J(h-RmZ}tgP@YT6)r!g%qE()BSo{E=iXB75t_N zDSAL?w4#y!Oep$}jinX2A6c(u#;cJ$tnx(t{hrIpWjC7Rt+hD+UiLwFOsAeRa% z74Yx5F5fN4w}c5&O7)B;d_&YW0MhkrSJsK70Ncx&`Vmb5!ejfa&Kl_{5|o8)>KrRQ zV=>v=eoYCCrx~rCA){f9Je;XxUd)OTzyYWO|H|oL1yK(`W>U)ku=yZpc@xM#49JH< zY4hc0y+tFKrs*^_8kv`sBSs)?GIz|#hek{_AIcOMTwnr~vW{b={C44yVmE=u)? z5m?qhjw~3d4&_SFf=9>rP%A3EpH$;UA)oan6Y_W6I5FE~=7^Dh_B|n1>dkN~yyT!m zkh{5+qSDY`a=kMm-^KP)FX0RdhS%iepb_X?jo!;Ku-TO{IiB5?4i(S?H7od#>2ppL z*&~fb`R6RB4v!hQM$Yd=KmZZ?h`f;6qs6>E$QG@$0`sbVP|QI@g8MFO=(rClB&OTZ~l zHrp%x4e4yyk!fky8??5V0&;y(FFe&Z3|qpTrg!zlZ$!`%snS97in-D5@`0( z6|Cd`O)qp#S}t{Y2pKvN80UCdG8S9CZ2x6xv7V_=Wm;e!j}Dk&wfcH%xh~b}3XH%o zTLKn2tAeHcY1+s;%b70dvDMiH)mosh0()Gvddstb&{u)0UXGMq+M7M;kT{*+&rU>Jz> zj$7Rdc|6*qMt-Dd!g9Quc|wjwpgNc*xoEg9AkBTb^_~>&{ZQ zH%8quVYttJ3%ersoV0VIJX5g;*Cky7d8UG33FL}1KtHm*qEO3vMk-ZUDd!`QLMOvn zc6=(GPEB;S-&N}CX8LmLDe6lY2}{Y&05C=#pjFfwF_1GAb%v z3NgLP6BT*j$>XJD4V%SKK-oV0>V3mAqlV z;DKcjB6-88|JnumWw2A?#EP&tG?puwKUjmPV!%lAAdAHx?=cg|)Q}il&=6j7%FV$P z=&@C`DneoFaklZq%nA*nU~+Kx-&~P38n%u}eh!Mx{`Yt*8JEhjc)?HEQX0u!827k! za>4TQJ@Q;dvPX856<*_U^jilc_MdvwBvF3b9x63^QCt$gdJcrAmg=N@8`RK*XIPG_ z4Z@koxLu>ZN^DEX7`{6p3SZuraYb%iifg_lId@G!2BK7MIbs0CA`G_0$j@mxDtEh< zlOPC3EsgdeJKDU?^x0F_|qjL3ldIr$+BI{>NX zvnJ!#%Y%;YAXw72e%0*jygTW_3C$;3TmdW6-+7R%0BL#-?-%F?15Ub82> z{?!qR-0l^$H^0AHXYB?Lm6kW;*2_&x5;%T1KQ zXvaG{W>9858MRO^kCV$q)UA_UFy3yxbE5PM6n0ma&ol}w0(r0q^+wK;0tES;?ra^O z?;zK*10lxgST}5|2e)Mcw_eUfMVSz<^-m=|zd_Hk1YUzPA0LyC<37{w8v+8r{>l{g z`sWN1_~cxYELeFE{yRZnw_a{^-ZKRMU24Y( zk*hayBf3RBDlIR(QLkUQ|2qDDJ(j{a;ZA2mM#hjUpGm+Pz-Pm)k?W9V(96B5e1yW) z8rjo@$3~68Re{X=B;;#^#bHK`yda?~a;<`*^@mW?yAzD8=^AwFWtW6}^s{`%lebjy zh)A~1PNF-($;dk^##kLU%v75i?%B}(t$J?M%a4|3A@Pq+fG)cA^24PpYR)((VQJOx ziaZ#Vyvml`5| z$9I}O=kxOWo-sWiSOOFaI6Lti4B0H}!T2|*a0zg~>u^7hCh5uUtf}*LRJQ=DGIegf z+@>bdk%s;x7~8xna)TC`YgM_w=tNXdZ=~pge>l&4w}(ZkH}Y~C>r$?=uE@DPj12Ru zI@9NzB)q9Y|HvQ#Pu!82uE-^E<3^t71qMgr zz{q^8ytl_qpi@rZdrVgrlN(p$>Gv%12t_=(J2j%z--Au=GtS8f&75O)S}!n9tscm- zVwYpfX>vd<=kS}{2ObxvgNSOs=#&di1S%N*YXc=lD9Ko7s@WhPHr z@xnM05Z!ut8>Erq1j1U_Z_o?NvSkc*T~zFKEp|>)>~&jA7v$Ut5aYE}H}X6zi$dlc zC7aX;)sxfVw7iVUkaAeUVfFoc1BJb8S=Ov`MMh1?nqH|k?y?oy?$*nXm%-mL0iHZo zJx&*%eEVd=us>bRaq#Vvi60RECV<(~)$oS(o?P^~giLREJ=yS&sw_ttx@wcjG$fuN zJ4$6XYngH51u+fJzmYVsQE#$+jm)bK2yr;8Py-h-TFVQXk4cHlPXAL<_2;&#mZpYR z72o03mkQlRAqZtuBBEJ4N+q_&C7iFKQGi%{(-5!Jt(Rw(Wfc+NO|buH1jBTTIwyf= zTnLCzC+GpLHsjT?M6_~O0_H2U@6M9`YNQ5`t=%V&ft>_v%0cxbwFn!1jTte4bHwhM ztlnG8!2UJx`h4!>_JlNof`3vOM7?vYoVOgzdZi*XTjkqJ{b;ug>3|h=-mPKOWCdlN zR^C$)x)t(sHBrYG=Qga@5w-uklS^eE9o{w>1X;7($9t@*N&{MR&4mwOBzi;C=j&vd zo_s6x@8$drhcnB_N}U1m}RrWy_{!wR7}+6h-;Ml zIeC@4mrfN6p~r8(y)=O{&;Yyifw$H0+O(I47veT_6n2*sa5S3k5ct(+V|shc$B^`H ziNyq|y$R%3aTj;%WT+%ksF!<75`_(=qNBE`k++u>S^u>LxxKpIxpfwDWog#xrSZ*W zjCwtHMtOBv7Me$;a|}Gr8_OC08m2_WT=fQ_mpwB*BG;;V+O3n<)pgG~A`0DDZ)9x+ z<E;2a}zx4t6o#K_$f%FHDi z78vIXoWz!fc#xHx@kC>)LU{lI?U<2UFv64DQ=!~~3Pv9}lAC+8BTl?S`DwnAUYpW_@^zMy_)PcvzKz zj#i0PbjyQUTyDTAf?`JQLg{}M0?kQiXtb4mA=spC)EON~&XaF+Cyw%)4@~gzCGb$G zAiNMn#I2VBlL(#KDwCqOc_?ZuJ`#&$03>wlo#W)y2@ultOA})G@&uJDqP5N$?~oc! zX-Eo-fP1h$Ss&WJjXmI;8L|`*JW=jjeoZ77-TWTHCFh8pGRPgq$iqDBe`Pu}&MbKZ ze!=ppo_sMY(+4Ey%SS@Y=OxRo=gKtE$d;u(vu#~i+?J`OKlR$Wq*W5bM*7GIhAOvVJVFl*!x zjg|Yy1EBnJ+38s$zds;UaRMlPxyIxAe*m69VZTLcBd;Asc)P*V2jxF~qk7dQUmq-) z?aR^1c{`s1kC-n8>r9uQ_l_m>@vvkiUr=dJ*~>1e!;?G}rXhwbM~LT22(h75^qUhA zLXK*<(VLd&A%s~F^SQ|jVe?=POkkYpZzoWEAnEDbRCnuSldgM&dbdnSP;b@|Nzn}x zr`f0S`rXgS{YXy?cslGAjNE{!!Cj~O_Ls-4{IljSn_DAy+ABdGY(3b_gky{}3aW20 z!zOq{nH8CWlm{7&O5UnvnNO?@D^1rY$UO}wQt+fZ%e4t(TQ&6Bq%+0&ceR%U6Po}o8SIe+3dGfD_MD<5Ip)mmU@RA3ziq^Fp?B70QQ z#p~LIUAXOdoGJVBy@`9=I=yq zlV_K(i=)-N&`VY`e{@oG4@INywqYNVqW-yb%ABmwJ5M$m5ph&H@VKSr2nodnJ?W9fbdHMh(!n!rQF zr13iOc^Jx=+*Tt859pmWK<0@KS}MYr-8$u?Nl!1+lPu!c=NK;f2aD|eZXnk!wp;H0N!!dQaq>QI zIqMC_B@LWE8Jrq7UY=|3-8zF%qIx6OGlzs>S%Tp0dQUl%IY#K}`P+pmUIFyZ(Gd0RL^<*KY%Gl7Sr_Z8bE~OW~ zTn1Gfm&knNU~N1reSde3`$^-;lS|=Qb4XwwF$2z5>=N!cAz8vJBOK6>1o=!&B7vDT zBPp|;Wri@m(K*Q44oue9qLGEw8QEbacg~4&?qo>mj!qWv+{t=>-o=NJb(70UDQAc^ z=Z5#=+>RDSH=Rdt&r*hS#7?=t2gXpYO&5&Z*W>e?-E|tk%8Kjyq%DfP*_GJZtn?G` zpJs?i@fVlT=v9*v8a;hxQgoc6M(!k#%}ti?x}%59CpV`DjQpCu7s)N@Ju1kN^d9y_ zWZaOiN}4%K9i@lVrH(0jAeo|JouUVr6&bcu^avkDuEeAq3JdF?l&a5@Uf-E!muzKq z()wyzckASl^6~tzY&=0y^qeVd&}+@9;A?*#jIvhs<4iEVvfy})op@{5)hCM?Fk&i@ zt2dd^R87+C<=X71rT-{MHg+PT0SA6Nc@J!|kt-RNJZ#5!4If4}PcGL%Z#C5_yH;0T3`0kAAs!9Wt_bN@Fjav(%1;%N@ zc;Ei+H&`F!xd_Mr_h?8o4~-#5B?c5k3SC_v2n3gg;HMu_7i=|GGu51~@*jhV*7W~N z#``=|6PY#4;V5=bmR+a_jI2!$xF^e-sB9oD>vYrMEI=>}%Dp@(#8Kd)#aZ&-nQY}W zeH0n*K)eV~xodLChW>vi-98g_3vi>E)$s|KFO~TE)M=B3PzkN+P|CJ*3epg7E+@>V z2nWH`u@0Oaf9_u)vJC|&tT*ADYVmRFRn%hsl0Q~?VT+w0PfX_WyZe$@C1cIt`2q)ebkqnSP##7sjoP~v}d$Mz+4EBt% zD%KC&%CyxWDTh1AGFI7mfTY^#wYLph`%|K`gk+Pt;b$M(fk8h)B)qk z_6|kq4O6gOdZNSU7wxP%Ba8ADI#~^dR@zW?YWA>mA`_(RZjUqm&Oz5vOB&}Z;}F4+ zr|-RYOyE6Q<2U+q<3XL!Ka zX%aR5Ih(sOFp4hosG>T6K`((w4h~hRALvzj84RO)Yud=AE9^`t7Co(QK)B2VM%J|} zYi>N*HFcXo_324}JLYEfM&2aJcxtsv1&p*|e3)BLlArKe&XG934Y?KaJMhCGF>FwZ z&(7nVVQt|^cvv&J(sT7H7(vn|4o~)Iurgu>+9u{?AO&zQIyr$i*USd8wNpd2Y2Db4 zy>Sx9vEJ#H7pCqp1X-1x?fx8APrfrHbiQnPpY5k6B>lbvtnejyS9sR!B=6q+^EQ<7 zzJ36V&=5q0b>oF+wAVl3TFGH4k67!MCOh z_%~X-GP-0i!x=kfR5;+EC%;mIGLVCVGcbLtZDJu(qB_fPZ^4lF|( zW6p0lpN#O_{YO~YkmUVgDl@Z2FO@hX-4;(WTL~0$x@Gg!J*#7OwQ_D!b)!}tH{-FH z*~(Jh@GSbdNN>Woj+*z5DO0=cx6uPJOft{EH0KlNH%`w8KqF zCs)!5<@-%gJ5x(TJ=U0Dqsh&TXMNVZq{7Wqz}B8k9~kGt^xLcA%kawiNwHm8%<7Zp znm`}VwKcYKX@k6NH9)S3*+9Mdnl=k$Q=^7+R(Ub@C4JrFO$_S2Ha`zo>0u|INPA<83OO}&w;HQ~Y2twiJ}2ZV?{etSB|=?zohD*Lk7BkDc_ zdQeEwv?j5e-qD=&zn=ah;ni&k2tQX4Zc`9`PLD=@!-uLqU=3qQ=!~~x*+%7pq|Nhb z)2)-=sPvKJ-mRBw4@mB1!=<8^P?X`{lJtGM_I(T8$?f#b?|AsX(1hEf;ooG3|I4!B z|I7~mj?(aNL0nyomw$L+0_Jz=+i2XYIvGj&zeoEY(f;qz?#Sa2_j)6D99TG8H=fc4 zkAqxQLiS1*sVGw|Ngo^N!>yBB%BNyeX(~39ie5p{tt{tH+E6amJ0F*a6_zX8lJWoc z!1C#ToOOi>hQ6D=A}6l&Bp16KtqB~1H15{Pa5)_Jl;Bw3mK6ONMI(J5&&LI-UdHvI zj;GyYEZwW1ovG8b_dvvSXIe0gJ_%SaZ?F>5e>%*q>|zmAMAuigSUV|VWZ-nY0_;41 zvGKlipUJ_Hs3_V%o2;`ICI!%dYol|UvmIv3=!I95G+8d3sM*|H&Bj>OzOo}3;2V`3 z8S=N|M&|m=p4xkI{gA06y0{Rj9P)~sK?!~~Ym8=5w!@RTb@EhO)~%QC99S^w{;ND5 z)AO@d9m#hTZJuiLT^8|ImF(vz>(`Y!lX7z_hpnpOZ>=n_>YLgIt(5aXwks3*Gw`5f zSZL&zniWLhzc|W0B#?v}QX!!kQaK__1VYQepj`cCtHq`hQ^C3oR1m!w>cR@_&pod2G&Z_1-8w^5gt9n;aw(<7&(LUMJ8g=z?oBF^NvdzRXFW!Yu@o3Kv6q14NS__W1HtdH`fth4RYThFe;1R({;$%hq&gG@!(F zEL>HDW_eTsqZ&pS38&mZTnJw7*SC1&E-Nj%k-Ny9Jw892W8P9$s-K$RWz+;LApuv6 zCkrqH3RvA+bvB`ejMc4|pX)?>a;8c>yqw&EX|pFsZuTer| z=ag%d6)>+>29YztOr$CDWVbIHoY2h9n4n*JeR;auH}YMSxKHxyQh3Qf@M{CV(7&s^(P$>nwyjXbWF4=h3;iol>VekDd06zdGu;Ldc})`t6gy;CDs z9=OM^Gazfuk{n{jpoY}21N_TCVesh7JI)PCi> zI4W2m4rioY`aY(OgOD*9;#Hv54B`DjM;xZA&AJS$%X0RcfPVr4k;-bQw~-%LhVq|T zA<{=H6o&%n!*veiTnwv?yx}0+9O%xGa%f?uUg>1>qUc6< z<-tTho?+ze)-AS5PmxDe>gBS^g0yDc8S-)~(x2~DhIslj@6M4je@y;nfX&GO&@Zs3 z>MZl=_4Dk8k>hb7ufm1xidJ0=BU>xEoMv;m^v#S{e)>S%hmpGvnP&1%6ufOlk|FmW0%ksHs+bEeuj>&QIlMreC|WQ37?lr% zHo^%=F-PisLNrM2h~aH+55`?4jsq$%>#n|rW`|Ai*xMj>q`lAwgvih$s>bl7aWLLD z>%Sg@Ef*__V&1es_8L zhEn=H<>|wv^n2+a9?=jKB4>u;3+Idr5*QzyHZ7jXS*6rxry&)tGxE{`Wp*x@7Vijn zjXEPQ@OF)=J1&|A)AqswcHc$Q;Ks;{E@9_R<5u?CLVF2Ydk73$o>TtSh01Gra_%7` zwD;OVU-1a^zvK`sgiIhWwm?WgNlz|3Xq&uD+_Ccy3b`q=lMzf-7`btQHOj7+JEw(K zLA;QVMC9F&}gh*A%g%3?3A^Q)OTQdmaK;?=;ESv_0IX zY4BTi;6`}I2Oq&h3y5-Ek?&|j%ul_MCl4x^oOg&E6v}lizFd0{fz2W4JZVA79S4Qd z>Cc23b3WKeLBl2xuK;O@ir~naF@a1Nky~2)sFEBM8fo)0a)`Wl#TlE*Vw369?HHXFWRV=Yr?h#cB6hz=aqf>xbB0J%hqMn!eUI$w|_4I zrT;K{()Cg2?xf4>wadG;%j>lV@*)Yr@S%x>0QBnfd?(@tQWhLN3FF zIA?ORW6C`J2MIVoQE*&&3&9opyl-lVs$+M%Uf%n;7T21q5=GnkQq^mzYSvJCCo)uO z_uJaLZ0$8u3+zr0u7FtE#8mz10`?L3pkxqtzq_VsnADpfM8rw0mpi7Z-FKvg-*9;X z?&fGp7#TO!v!azbN#4}-aZMmB;{JPZkf$C~Syu7ML#j>j$>z(R)L zycpv(mR{jfY-2q+$CtgHn7Sev?eCdkw@%*q8apw^N^cOb=JfI4Qn2o-6T9H?xP*}*thpXZF^BAS` z#ZCFM^89HQ&5rZB%;a<2zSDi=;an6UX7 zz3PG%SNsA2mOIP&w2c0e$7hWU{h}8qqz8YQ3Jsor(gE?RcI#y5RthT;%`)=9rWDe6 zBiG(a1tX7dN)>9&sTtFtKucQ^>aX&;2(VceMa)pl7te@uDrBPu3LJ8`BmH=?XUu(> zrv>GE>4I~bWukY)g=fiC2P7Dh8c7;2ORbDkFH7?v+$5BPZIC-Sj=;SvjcwfIJ((J} zLQW?gnezXX&hRpLF}I=`jo@YJOibh3ei}iZl?aK=*)f7_it&)FP#-E6HWcsg={J^I%tS0y+}rxHyiW3h5+pNj`t$WR z{1` zd3mjcw#DcUc*-xz^FEe9xxGByOOo@qmK9e3XWHf>eyDPj&3B(jOF-Cj!N>8aopJ;s z6XNaZtraO?mrr7+kEmEC7L>ELyfL~yPmYE)=T?Qalc|v>!LN+;_Y`GUIwV}6iV2lG zdC;YWzD&js9ELh@E=up@sg22@xM4MIcrv>v6AL_f!5bFf92THz=b;P$r&W~+Liw$#95|dBx?3ZSutk6q$_J-DT6A#M|yxtD&R-{ zi!O`OGqy^4i_$%|Y86hmk(t5f`$|*t(sfvbKF~}&du5>sZp;eukybgYK5!_ zNT*nmHRHh&NIzGQ@RNPS1n?X=ljowS0Sugv?>50q50{*&pr^Ehxzg=ppwZ>~b7wl! zWy#okXV%X24mQZfxe^Bvj!L1(6mLlH6)((6m)~C?6hqy&Ca;aP9HJeXF>}CQ&m%CGS({({ZF=TX04fPpJ8)8CV9X=8e_Qz zetjtPq_9Pg3{MDG*W+g(E6g)Dc1}7utr$PSk$+Q<2;blb@ zloxjozp#xVWFq-1eC=#G549d+qaPcr^Kj`#aF9nlZ=7R^X&9qv+wV$!zYvWYP6}&x zY~^{-_@9xrK9M%^$q`dSRM~&_A)hk`ba7W1>BTHOPtO1LO_0fuTO0Cy5+#k)A*^Eb zHAsGY$3kcN+nG}{b<*pH&hh-L4vohnyYXslc%lT?bfyGUeLK1^=*quF%M#4v(2kC@RWKV;0EQ)9INx| zoRfEpipb4sQxpBtlVwdBuSciw&H(l|SD7KODcb(ipDA@f91>{vIG5x)vkFV(z_~EJ zw`NXtCXcsA;GFUfADQZmb8P&}t2X0Ru_%1=v$nlSeZ=+P>6MonoKYW zxVO!nMR%55tM2n@XS&?i$ub3I)QNo9+nz5~Rc4XnHHuO+{e>jY3y>GZW+cO<4DIj6 zi3O8$(?;$*gE#wcQOhTUeQ8f_KLhPqS&!>qj9NX=j5|y10H(iGmKO1t8{)J+Gmgl{ zS(Ubh*S{PUCc_(#Ct><4HaX_OXGGuS{cI{vi~5=Ms;#HZG5UqJo5~$!>7{NqM9FbC z!_ha@nfr@A7|;C+1?K)69|Z5C$#}L9PewGw8(u4QLhWb&hTBmZ^~pDy#z3)ncx|so zC7HJVXeuh-C`*jSvp!B^dO(tuQQ?(;iVBla8c^k%HaQ;9^yqsM7kc2HW#pv^p_`Tq zmSVjb#M^T?Dd-gJluh4ho6;VFn6#}grH*p>QcQ{n?yOx&4KLscpqaL%O>sJ-jeJaI z?2gKk77cbdPSY{v5;pp+sHi3gANZwshP}KKj;PqHVRKJT02BwsBie zdnrTrS*c@F#8A-m(v)qElH+c+N8jb$w56k}WmC2zPV2KI;`xKO$sd>b@-3u?kcx{I z5me*eDHvSj+_YTRk&StjVeSN}RB94YN?tj5{Qq7t!kSrP0Wn0Bkyj4hCcjm;sI-yW zJN@{k+ZL5_XUQuEhx-`&MA1)iZ3i~}$nltpZjaJUV8-L?TB^d9We>$y zzse|0XDV8x4drRkB2728uI=Yz<1tOzQJ!8}q?HqF@dW(nn}Z@}Mq^@A>Dgi=Ly)6c zTUnNzBPdR^-=dBZkf3gtF=kOYNg`VaNVO z&zuxDx_7Z{G%C5_p#2){OG@tEx8x}Y?^p7iq~z=SmP{S8U&#xSk}vIBayTx@){hW+ z*0iWLX%RwiiqjyKWeC0g(5NiwC4`=Sm`#omdPelEO-9{pDo=~LnRU3Wr_GghQ~80i z^inq)qU5-n;pn@(n^{LhRm(>crvbHde8R?T_?xIK>7{OFe9$Jx-OP%1~IY>mlZO7@oG;$(l$j^GZZ7{`Xtr;jJ%Y>Hm=r9%lcm5{fxYvib?GGz1dh} zU3pa8j{6yTjmHN5Y>f!{HzYN5v*x%%#CtOTp|~{^D6@=g`i(Y;R>(PiFRF4kEw}$v z9cvSgCVzBUEi)Pqd~+!ossR0pwq<_$dsiq+E+mRwN_EkY%o_T=%CiVXEfwV~;QWBZsqdeT^YAZ+?@(i(4jR>(PN*g%4bLl|D~ zv3oeOaann6S!rlmLb6jINpi}p>Dx-FUK!O){asR+nbU}|vKFI4Xxy7iFcBb{>p9*C z39pl&oOv|AoA#+ssMTxZ%J(h7+lR{6d(cgkYU+}vfn{m!_9Y00#^aN6(?|D#C1-skt^{GGm%~TvB+SGpm$Lyh+gK^Oh;A{QIOr*^I3$rH--+ zoI(+QHRB(W3OZ|LM^|}{ovDcE&N@D+Q`X6FDYYEd=^sxDdth_a#)h(FhxcgcqLbY9#Of|Pw@6l*Nof-1W7eYE8Aa>@*w z>eeE7RBY+Y#?-1U%GCP+A~$c;$=kV}aiXoKS&0M9`XtqzC2KdO#@)2+xwY4iN4O)& zkJn--#+nVEe6I#m{(Gh7OQ}Q5H+ExI%6wxgm2y%WQmIrr{r05EsdPFGJDEWviCC^)Lz`a@O@T!KgByb z+nZbFcf6k}e{Pt~g_ef*)6Tp_?+en__XT8AdwXm9`>IjGI6EZ^n>(A{&v26ZbC-2C zH7uO}o>;3G5Mkls_J#a%QmN)Pen=`6b*SZr!n^O+3ZMcqX zNllxC;7+#F&zO85~+Vyrc4a=iAPL7PN>*YcOu zOFH@-g?(S=;ke1RLy{(WPms5jsDYj(AXpV`0z9TUH#ljb&?A~Gm*OeQcZ4BWz z-Kg_9_ydzu6LQ+HN-JhNumW;@RH@Gd{RZzRX2iqRdM)zoWknP8BX(u3lPparnibp<1KHiI3{j-F?9OBUQKuz+$%c+t57AinY8$ zHfUr53!gCos)~l4D6wcJpHV~*cYtnGVC|_UftlHz)|qjRmA;-a*OlGr0{sjy8d>D? zo|Ic5_ouO*ypsrvuH$J)Uy5gDcFz~EUkV1iR?o$&2ir&GQ^X7fn@h81rqz&A>Ff#IMYI|bg+r6`8VF+ZZ}5tvk%q<_aveE7bSKKc zvyxumI`a2nSsawMqY{9=A3b-buSRH4FB^bEdb6(4_w#YSb>yBD4`UlyHt57k-w!2? z+~JfN$4oD!-xoCwChm>7)!|b9BR1dKyMu-h3ktK?+zfK;+oIEHxuZ=-Cp|uGYr0cq zsM}}SHalRU|FoonsRz{waT(WR@|pn@Xw5e@3_Mwht$T z=~I{Ve!Bo}TCVAa2@T?r&p5WMFs649fKh)^%v0MYhy`4yeB?c9KI0Vj zo2ayNq_mAOCFGUOR$;%{$~jfhY!L~a4!_wwXzLBwDRa}p6MXSFXMHpQUGyN8+k^D|)M##fKc(j`p7;JLyid-F6aD>x-x4w3)ckX- zOyg;bTNX4G;1SZPGG+sQrj=AlrF3Ncc2lXg_U4vOepjSYv>S6K@5OXXuuR8>w~wZ6 z7>2{qLt?a*b1P)#2&nY=tRZ4xg&ZDYoL~>wKKZ|R^0OuW6(ukiOLhY))RTpH&UUj!L4nl#?J^M&O^G#uzk;&A$f~q#He?Nx7B*v<(1r4LzX-6 zKXZ&6u+8MuUKA!OXS_#cK~WYb-rCbsb}zV&JP21p5P;W^V^H^8H3Z4H6ZAxb?;=%C zNwq-jNtzihHS;*l*cl?$nLA))Hl7Jbg!)^J*kN`mjQo+fJv8}E?MQp!+~ppSnalfT8ALZ%GJzY{bRh5O_ICiU9hXjYG7R^`JZMrLPgnO1@h z%ZJCM!N-&zqx$7rOir?XER8$GO(yyxn6!RfkZRMYV(c7#!i^?tGPP!ADaE=Mzyl;! z&ayc14O&@|H2FA9nm%RF;LSKU>~pV#Ts+mD)=HRb5u{{H5Sa}UwZ#;%3Mt}g2(oU8 zFYz-bB<;UR`+%$*&c@iBeSN7!(H?(H>0GAj`krX2?n=YR%9L&K)2{p*C_BheN~0`8 zKMsCkx*bstPt3VUZ7*NpJCR*qYO<~MUU`*Ko=QqsfWBB>wSNDcp%{wFDYm`$ zsIz~Ltur&t0d?qKbwK?;kzB}T!@~Ix(KIrSMQsi3<##nrT-?%Gc7201pT4+ZVQzj? zOH+GuV^r6sm2P?I=<=6qXkS!4Gb*H%`8^5&#=gl58#+3*v}WzwRf@OtB4LX=K_>84 zb>6{}^F@^x<^L*q(YxoZF`5=Tk`9Ij2sPwzy>l^{3Wk=<&$%{}7F=Vj9zvi51o>M&NvA#!=4l(-Mwin?-=}Qy#{mx_({22atG+h=y zlIhwUr%%@Dvgz9uL+1WBBfZX+(whBdY)1^kho)u1yiP#~?D`)guVKhF{Hi0ZR{ze_ z)$pH=I7W)?pNK~3=xE~P)Co~CJ3ZPQkLws`|5ambY+u&a*?L09X$>c)Qv5x4VQb@p z+6LI+bc`ZBb>5H2dtGHKEsm^6> z`qS0>+q|ZBY9n0Mo18&t+|b$Fm{VyqdFfa(9ep&lG`1GvoYs~`{VA!`qQ#v}ODQ9% zRH0Q_k@9T@THgAU#KGW??_C&IMC#%Btie6}c=P|ZY)d={A|+})65 z3T|4~_9nrPr~k^f=DI?Jbwb?+PUN@M7UxZsqqyzqU8oBuY&iQGY@myErpBFuU{NKj zQ@t%Py>FodG#PiQoZX8=txRUN^7g2|sd7A5*0mZ=Wf*B8roBeaN)O0hy(P*yaX3ub z(UX;P)1fRKlL;Y#Jmu?nK`hDg2yh^a`FE=<_GMxyulD$|J{_9=jJ%=kAbPiLuT}@X zTI_sx_4u+m6^f73Lx096?X4*7#L&nW2KC|;7`Y?0%>+c5luapwV_g0``Lu5WcdGp3 zfZ_cVQ*9W*sHx4U9MoQ%og)Z-i93LYvxHva0xnH@k&yp|^}Y=C4$`+bKoH)|sn7^7 zu9>q7R@Wu7 z`KKPAFGGm;z{pAq&vo0Klv4ZS|1TvA3u^>-h5xu^^a{kwHl#aaP2|#jqrou zupbCbkZs&ro#FM7yl(_i`@uYy9uR3U@Z^|+%<*xEC-g~cRy0JaeBK8+0%N@qJ=|7v zgl~~UY5Hu#)jY_1a^!$C`gC+?v>1O=kR~4y#NnZl`MOKH{8)irxj6y*Av6>!SqfaY zR+?`Odgn-H^DYR_tdLa~JVnLmzCKUYG;$i)kXta(*!GMrSsr>7aSZ7oDP*Uatk#nwd z?A}tR6ax!xaqDFeO^^YX&j?r#EWG25l7TBdmlWg#B2zA{By+w%hOG>1Ptue*OxxHfu zG^8hM!hVC&3T+Ie&RHg}r1nteyk1lUp1h(J<$JxjlWb27n0zL27pZ(S>F;Xr9cc*- zdvXcUyyQ8GSDAqJ&H@gNcok&K+>9xjzM1iaZZ5U+4DIy06XjW`Igo!tMQs^(BJr6s z8COuysI+BJr7f23(yf$Y|800If$SApw9-Dbr|v{~q{nw1`H8lxGkA?VQEu+Rw*-0U z9V+viC^u=bJ8)U!ZG)G4=-n~$O7Pjh8RB9hw3}E^uMpDk zwm|r&$3o7=fxyMoh_lTE?nJq)XHfMfH>JZEi#Bgd25}{_D`33_GQ`0Y|J!I($F-}w zXeVWz17G0TfVhBT-;92bUqou{tB#NV%HJoQ-K+YxwRrfMJia$ZUME1=dFjw(iD~Xm zkym@NuA?U}Qrp0tBCjxA`r@msVQL>YS$B$T$5D=1>raO~=o%R5OBbRuA9SB$q8J8HK*H zcFbFO6o%3>h_+sO2ANsk44442N}9^AP2SWg69XfYH+A@@BbDcjv*(EFp27DaQ!C$x zB%8p8)n9nZlu|j<1;mi*;Z2p=>4@B6QF?R zfW#O(Fdh@>i9@&wISoIg+-xVZddokTboyvnrwF7otd_1)E}=O5ddmdTK4@er2e9ay z2gHqHi@y^$n|dK>c2`-m*P1@F)tx9WcPU|!^Yu|^uixZQPT$cLVbsl1;QQBIKDzf` z$TqVP^4JCHWiqZKmk5uoSYo2je4*={DED@4aUHo{2Fy@$_6@%nLE_3k5a8wjjpO6G zMmD7eT(8#T+ynF-Y65>N7!x!S3c>E1d_6F7VwKGO7#3vY!=Ze|mqS8O+b9Cb(xS+S zbo+CZ=qMWX*MjAPc9SM#bTBAw5i(z9px$%|rAL4)R`OPUdXUme|yJ6Y!!Y z%@Gmbp*hw1yOyKXt)%ma0%sx**t@dSmu0>jQjmpySdDH9|IuD26#l0gEc%Rz*+?eMnsrz~Kq*w_J# zRA!@ip+V!+N6di_?7^`Z-{Te8DoKa3JS#%L6;o|;m@Ws#8<(&7PUIAF%4-RXKa5T` zYl47|KrcR9nQ{yRGI$^c(&g9HWHQn{SlxiuX`yf9+d~I+!I)G46R&_@GM+;ifkt_@+oLUhPYZ3^&l_gg9lC@1Y{=$&se? z%;Rwr-lhnQa}3Um%8-dhPZ8p9`ze0iDl}SH+(E+jk}~TK24;@ta{B#Xru2=K*((oUtlI-kiKDrhuv4N!3?k z4y=(EG33_XtPmq3KpQ6AmTo(zj8{34?MLb@TNSH$PQ-)QFtNdLi*;&??2R{+ih9(0423tK2Mj6^);kZ90y<0305hkR3!I+4qFzZIhD)etfp>n42t% z-u5Ftff=DWcu8yvyeczV6)U10b%8~D6GFMfFRW*JMF8ti5 zd21mku9lwSY||$_zV6$ds!DnDzh|4ZFSeHBO-q|QPit*m@V=z5vB68+NF@1vPg!eg z>w8CIJ8*!+en;oTtJC%jV2UwNbI`CaAtdiTmwN-lzwHMS<65otD55ZHzPV>$&@EX-fr`MzZFEluysApOtjG%(wKKgbu+@b^;%bGFW=rJZf-KjUzqdCv6cW`#1TBr4RGId# zWWX>X)KoN*=|PKi_qw+A@eKPxi(e{)=H6hdO`sX zCy>T$Ee}Cau(OB=ki*9~NA2NS;TNRWhcX*ZrT{8VH`19+{w+(ihYHMpU;>q{p3332 z9}w`rv`3bxUA1ztYh)_2m%x+8EDuw)7mXaOH+bF(q5{LnR3{zBfb0*z@)!Uf3oDvj zm7W0DhTNL0$S;uj=y_(N2&s+n93#BDg>Bjk-9=aAVyB?|zgrZ2WsR-p9uFk)F&Ph3`0}l9W^!5=y^cUe| z!`$Y-s9w}(%!8fq|L#1PY=^e?)_Khfo7$TiTif5)hKl=6TIyK#z7FQ3UbN-@;tk!{ zw)ig^d1LDtP3;Zyn@(u&Y-~N_{WS5ep)YFaJnj87kc>S#t^aD?;xAeGcocutEnt3Y zp=s{o`5o^~Oy{c|L^_qyLXb<0Eb%vte$Q*@=v>%3|C61#>&0mcTO0IV08`Y|&{kNy zNS*%Rism&~6Nz*x6@mZBs8*5{-7iQk`xtB%!IZy}41@sB|i|XkK%R^&I45zS?8hU*%N*mTt*uYnb2EuyA2(W9gcs$?dK4 zq9VzK2)mNp98q=J_oR3v-KG zn$=t-ol3QKEiIO( zW$-$aaW$s4yiHBmuLLtcw>gF7NZfIbZxpNoCE73T=wXq>tG;OGqsh?$Q)zaFsq`9Tg zx}+o5(YZKY9oTj!UpZ=q>kJm1k7IlmP^BBIKd!Gx(j-iaMe3!56_?G+(XZb4$y zRN7cn#Sp5!scBJLT%j%YOevK+F)A2g!ai%?(Y(B=abZJ8M}o(UW}nH7YU>2dE#cL? z#Vw6(y5vqO6$$FV;kLz8n2<^>YHVuhY%gufj<$xDj;0vSMe~}MDiTmSok}%yG_=f! ze|KW4mTPNODjX$7t5XI6DG-IY0KeNhlk%;L+Bi1j+}1^9mp46G8nHG`P*~3x3VW=^ z8BOhrnmXf+(Ac(k-ol3Y@wCLB`>to(Vu%DS=%TdFDDn3YigO> zp$CX1#AReT+Q$@&C`;5}`;tYmu_4xpkUZTv^V*x@A!%aM(u6j*x6WJGFt;f_fT$90 zF1Wb3VMVz3qp7n!_U4a>Y;JBV0TrSG9Zijk+nYO=<(fM>nmQNHZ;qQuGRg*1hCv<8 z^UH}>N9(*!Wgrq%u`E{H3v$MNX3nPUKks@BHn+qh>{zy6rDZYEjoVEU_HBGwM{Z&B z+?WdS+0xq5*49`(GhX7(hWRC`Vn0g95)0DQ(Wz%Zyk49Vxt7*>4GWUxX5xh9l0o_@AB&2Mgv2cs$bOq-@~WX4^v!VEU8o?k)DzQRp&Qx&NQ|X0fnK#FvgtdRC}@C1 zajI)X229TNj0|Hp;0-%R?6m)PKj_KDGVcDJ@nn@dF4GF~HQ)WT+;&Emt2Z$PjV#!r zH|W$vXSm9?=^ih;knSJ@?yJ`OX=Q0Cqr$zvgCY(fR)``fc2NR@>!p0KU|&_MLKyP52t$|xX+S7g zlVKLf!9=u{FYWsAZr}ZaT;XWfsK%LcYv15Z*OyAAV~{mKK(x{5bMy_qOjplQ6Hw?2 z3g^)Tr)FbWu{x76eXCR@o|@>SW+g3*{h1VEqq}uv#hAzE~$ymrnr^k_ljn)>Je)nX# zzSnn;Hgb0=^e~B2IW;imYUDx_%8Sb)Jp~cyv9nU;0k8#;!5P|6BaO^^POMr*r(b%L zH|kfO978?{$8wf3A1}^ul4q@%`XPvckT4gr;2T_(`Cm@S~WTy#n7`>5RUL@2bvmkEGdAIQtN3vhC9j{>KT zmivHUoylmcEPkEIHPkTQIEtusfx=WoYE3@)P;D~ub2qdQ$n`6-M(%dQjH($dR63)) z43NplO)Gr4H7>32Mmg@`a?^6(Jz7?$L-%lXCXp}e;-8Ir+zeUm!?jVHIe^sW=UA(< zyb0Fu_O7Op%^>aSf4W#Fm!U&qdtYKP>K=S@Kr$29{lNBmyC<27yO;`tc~(wGX>hIq z;2idwIg^vcih9KDUmzFsCKazyXmj!(*;r+E%D=?;C67X&%n%Dd(3_2?{f2a6_qw4} zb`YY?Gad2V+mV6deQLfDEeP{x$qnp9uaVok%dpPK4>@byW3jMzk2Z3VPQ|u?N!jQzB@Om7j_wNBR8fnAlGL*C_XBc|hSyL)GV+z&*;yFS?oAAI3_F_jrNVn{VQ0Xc8qFMQKaAchK99bXwA{zg$tR}50wdq-2B5{@ z=S~Zd9tJhFs&~z}$I5Sf27UAAedhMpytIH-?0NsJDEw>FyL&E%cWrC@h+tep8P-xe>q7w&XNRqv}=ok|5FA3J`4V} z?+yNQT|T;RU{(d4^!jeEJhU&U-?E^d2dJ;apsxRGKt0BSYGixYpu0dW)1{cD^ZG+6 zm>OnZpKtf%CLP7-=Y7&{_CtsiD1PVXv+|juyFf;h%JQ;D6W$94$pTxDpL&^JHsaWGe5Ea%4TRq73omHaCUcyuzLrUH7s$F&#!X~gx9kuJhHv4`p`08gLlgvSo8R zI^;NASZ!NXfrw7GANBb1U{Z`!{%}gmrPzZ{)`ft3T(9QyN6Sv(w7aRxk59W@0$LJV zPX~}Uja=evS2>`g$u(H~fdV$2PO3Q&10!c(4jcgNUrroK{KD35pFwlZ6F4rpldYc< zqqO|e<&@Af?eebVO#7ajoN3!jWiPFWxbEv}OdyAZHEQK81)p7rN8Dp8uvUmq2l+-t z$j7rV21dSF5t;z61bZUt#MhXhriR5(t9=V77{`?Cb1L4oybQ68Tt-{&vGP{8B3VC& z6_k6Io?%BD`MNVMrxs(&`{8^v-uI*mJL24<|NK8+LOPPD)Os_gQ#pWhe2Mp6)>mDve40;&W zJB&N)9xi9wIo(?U)dG#+(;;VEoC32t_PZL6o0~1(G_GeHj*(I88szcvllz^tk>OsS^oLTR zeh&6#-8p=jz<7-2AqWR0uIepHTf(}`DnuinGYS6)j{3kfxF2OpQu?}{>`Wl}t#UpJ z$xnL6LX*2jC!pU1yf=Ed+?Og){MufQ<3~--taT5UPloOsh#+A%bMg4sH^n{5WjJJ* zwbr&pu@CQ})wJAiyvMi=WokG#XTzn!1>jdeG9H%i%HYcTwW9I*{8+NkHb3f1WD>qH z3*^C698wFyJT-LnA|dpdagV|^M%2lJ+`CgDc5*-M)v~{_U)T5grhhG$v<8+{+7lC? z8K)(k72#PC1S-eEZKMXu2;UTW=5L4#H%yYhnFfL9+AEXBDeTGL>nZUs&?WVkiL0H` z9QDy4e-8;EUR-tFa`WC+v&ucj$k)As+aSA^_;?V5C`TSmW(Q!^0 zH$gS?9z+oc9F0RjA?KX5Ox$D7PGnBT+v*-8Z&&b$P=k@a3> z|GI~ZgD1yslMlc=!?jP8J5xHgn|ihS9ro*H`}LarI(CMdC4XY`;R?hB zk)<4FI7~m|=TDBR-a7KQNCzeln7eGtYi&Po+OKu?YeQNg+TPV+Q0)#rTr3z0?rT z%k@2~$6gJ?v5{eQe=(FkPZlTAo%L-=|7R<&Z>0*9Q@p5=-ef^{#AO==zKWuhnWIz? zVD&<>*GMJe)0~kuzq(?$*m+#l|ITOh4o}%ej z0Xcls5bxtXmr_b~2{&Y5m1k2U@{Qi2Jew*wHRA2bSipI7)C4l#Q!Fx2__O!j2EN9t z4Pi2hvNAm&Jz43_u147sxky#6NY-a7yGhAoV1{HG{2{olQl~5*Cx{Ac*v{kf3foCD31a0HAK?Yz;Xu{GW!B5H( zI*Z&e$`Z~f5>}!tVP4S&dem+u?;|HG%U)QlE&D9)(|u2o_Z!O!FE7e5q0CVNDaUM+ zY!cpjLr*e~s~lpy3>ln9*>dBeLLTQt0;Bvl{9RSn&PVk3kup~|C**=HVdQV|_;HTg zC0|L@Qf;de2(L@%R;Dha-i+)kPr9?l#-3Cil0gzbj4PH2q&v|QwXIIt`86ZG#sr+2 zufp-UuB>04b^qDO4}~2A^)cfF(J=Bmt|Dr8jh09k9C(M!98Y@3>MXB9D33l+3XTPw z7&UW@jhxDT_3|`NTjJV2ic9HU`_nx}UV^j54EY>AWmm^H+qV7)*n<2RU?X>a9d4xs z{Bq)0ntoI5<*PGp!N~8@5M!qXxwVH7$4hAc<~31s^OT;$uE?YjGc;`QEwl1-y*5;t zT2rO>VWz=5*aT)kc{(^U=a`*fBV;K0Np-!W;P$DRvLG9uQybProz9noV2x`a&~;2u zEq_5H+6Yk|X3R`kKh}QXw&e8*d9vK;@8LrLg~u7C(9RocRoad4sgAu z3;9q%4l9&~#0bo(veeJ#-7m-;w)Goq>v!4KZ{Xtl0obT77#Xx_H`)?+CSL|;Hl2Z} z#7(xu9kzsCp4IPG=J|EdS6c~9!=$s+!*X(ngk(0O6*X+M0}e zxI7C~ZnN>^n>}MvLB9o#J8{F%Am79XMJl*TB4V&bukIOxA7I<$tkC~H#1Gocbv=XT zY||HGYifNo&}J@2llt6Bd5h7x%|`lEmsRP>ZZ}j<=uAKD8T-E>Q$wLuD+b+48ABXH z6oBAtj%|o{;5K`5)`~Hca}O8$c6_su-y`THKhFt)w3wU9 zi)MLB_dEs9o{j5auOAy2q|Q&k|32p-0Io5)2D{g!2ay}ImdhKN%wYk$2uIR0sGB&a z>=R_a0M*PkSTULhmITt4EX|AyqJBK(dIF=jq`=5MDV8K-9ZZo`EM;#F_Tp}|H5Dph zWF2NkZ`7nFIvYxDslfpa!0}1%mb!?lc$KuBMq>mY^|<>b9wgIyHl=f}XXM3Hh=`rl z$jA!IcDGX|ke-D5-h5#K_m9%kbP5B5a)*EVRZO4JU`4(h7|Q8>#yweH?g4lrVg)$q zhn(bjS&_T}IqRaN%{$=q`rXOg04RNSuP<+N4L755Xupb5ksai^$&!vn*pu3pe<+$nO6%11mS50Fl;jV#xZ1%1?d>K<@FJ0@!*1Cj?* z1$VNn;fSZ=?k=Bj+{id!#sX;_loiFI$w7)DrN@qMBy0Q*`^v0AT<{~+ge~^nW8{7T zIYDol{I)K`xXCd>^SHZ!U2JtVo;;l%P_86jZGvhxXkd~DvbKLS>1O~g2Oh%t^4=JX z_1{Xe&nFv6Qoe@y>od+t;_OMn6rA$yq~xWP)P|1V8w}hkU%6;5UnG5iXi#sR@V{!0+kNRb*+}7orVYfdm7pC!FC*F@4;%38YU4mHC zgk*2)jx@itlsB6FU9x9~rmob?n_UW5LmhD_(<7wSpw!D zLTteNh&T)~$W+T=T-c4YLP7iUMm`YIggaHXt?-%Li`@a@%ss!tHyLL5y%u1M8HQ?6 zU}x96mtYlpbv7Ic^4B1Aza%f{VeH9zR~KT(3INz=AmqSG!5zp`O16W(7>Fa)kATF z_R=oj3W3rhu0XLOwdNObLsMjbG@^Z2)n^87iVTfXh5Wc`p#!bA{6rI8TDTQex_78J zumGG+|FK$>x=7l~B*~H-97UfIpX|3K1DsFN(0xng{48r?isEOxIGB-FYOqrf#_pgE zyl7exZbe|1Tbos;6)-j;Z|qZ&?-@&Sbb)sBO|y?$hF0b`+FF zCdeW>!5F2FUY(4lUVfQrsA$-t*=ETy7zJuc$2SKHmC7^d2O+|@ArWXB%!l{*v<@xZ zkZ1%F^mTKS(NMzo+hRb2%xv$czp zO4h84tL!05GYt%nzkV%F^poh>q(wctYLuCtls{D7_oKo24y_q7e)4y$7EMBnCTWp0 zr07_PU@x*N(DfxO!eo_$BiS!P0Suzsy~HOQp{G-ViU!q4J4CtyR%-s}>P- zSHhw(gvT?caJa5JaEHnhnUdnfS489xjRKVa^3Q?$o?I|Dp!jx-06qcPC zH`~8Q7BuYhGTi1iagm!Lmv%6EE<&O53$+ou(&tW+tsV4tX>HW3MVVfZ5m9S^`eTFv zk>a1Z(-3WMOv7OZ_TH&7HOS{dTdQF<*M3k~V- zZzFf8Tn&T_l$-M;++8l>muVh?H*`=o~Fq{-hlRwVLT)e3Y3s+m@)#TM9<26 z-ZI?|f?sJ5n$wpy+uo*ZEV?b+1rVW0I@Sb}$zyCQ<%Ftv_l~sY zBh>68RXEjJ88i!;jf~M^D4fqGcc9AIvu)tphE()uanT@mg^ktrUh(aPnrljk-~M-| zy_t z2Deq4eQYQik3M6=-<=tA*(L``z;Sz_4&8=Rl}Bd;?nmGaNM*j6e__q9&P2R^syonP zo9)g}wBtWgSu+DTc+5O^V9|4k-C?3s+jdtvwPm}(je(tVi&?>*s)GOSw9uIpBCjTd zbV1U}hH8O}C{StPXmk7wEKvZXWI0kUA%6?Q+dQP?j#Rm7wk4L~%`@iy3?MFYQ{z5*hqPLNM!uqc$K%?L5@(Zx&dvH(wCNjXq z{7PO!jHDqD$^__Qyk;YLzI}&B6j#<&w8b4sukKKJcL-7nr9s@O4i}!I_5f=1YYA+&iAk7^}32JNe`O8 zi&#*VH|t6o<`v=MB$XRa3S6V~W=VT^62UT`b?tC7QFRm>q7sY%(Tbs$2`%|E8_i zf7&8ILo7GF8?QN8-d%Q6u> zfPiXa0@LDg?{w1ZSuB;xDH_QabAsk(H(p-14_9Ttq$=-bO7ITl?jWCUlFw{lc{fxJ zjR-Ge>D^S7(=(WZ?BWYIRVta1wrFvI7EQmH0;$TinMmHJjW)%{0)Go|Jp=3`qEE7s z8}C(8Z16o2xv3caS&BJ3@^V^tL3RKKexo1njKbWmOh>wr*){B-=Xts{GbIB&+gb)>dU%m#xa%9zIZ% zget4MR(^TPFHyixIaHaBht!v*{2F{CGx1aYRgLM6b7XRo3%d&PuJ@h?No}3G%cM8? zqAG1E1GVw-bk^_Y8tTm)A%BfF@#sM1b}+iiuZ@n#(X2nqdq4%&)Gj~>^PqNZ%fsoH zIqV*yKpgPvp-$v3szhr5E8k)u7YF0yCgn90u96rTX>ShCx=?c!X{WDHewr#&wmy*# z^a1XH@hZorW)2m{rX<@;7bR!qBDj`NTjW^Ai0jAh3~UmY62z!v-jiu-FV!{FuLf0* zSGm5OExn>HYL@x4F_o&k=jT2u$}0*XR3~vqdqnctkVw8s+lxGqU2F9>)tNzJB@Zf3 z)LAeC&0pi+JXsg1{Bs~-)at9PXa^TO=c`@r3vH=ZyUf3NzAln~3FO0BWt9t|Fl}~< zq{1Q5uvtE7FMk=DR)ewGGrEU5O_i7GOai=^M&*^dlFHpRF;4oz#&NPSM%s$aWG3hn z>{D-nd!a^6Z3v>?07j!1wv+RFs>&;tvN-Sg29>jYE1WW5b?Y2f_HvEb?r%osL0*#o z3RKqh1i-KskMPa6r62R=32S{N#?e)? zf_z?{vw6$qvjTZO7s+L_0(p^FvPa1cm9k2GXgXybD-nDnHeyp z)9Tqk3l-J$)DE~$Qv~gQnRhWI~13*>X{sC^43iD^X&SFSqxZbzp9Opp5@;H z*%Q3=l2v)7E|SBe=CSc*4Brb>4vY6m^fjd5+RI+g5z%BS6djlx%T<7G!YjFPl4RZC^g3^v>QtI`=g&9lnY^^r^|s1*I_ zJhqT9D)XLA&(o%a*`J&zpDvUZC4qbaQK9WS4e}E7xD1x?y;*_EId!Hkz!T>=qpxtA zq4!dpB*&+Sdiry5U&jjq*MNGjc#n5T*K5cJnUlX5eY2gXo9;Ps$E&>E9Wa#cz#VTQ z2cOTXi`?<@PP<3qvw>GXNGiPAT>$8Qp0Vn$VX2r}&?5ELX5gi(%vDmtO_j@_pYQ-0 zy5n)m0y7oSX1Vcl0-f){vO}V#s)PgXc=;!@!cMaqI9bTCp zEbd%}7|MLWxSgRYS6YutJYxX|3*lJoNArR8JzfVbd5Ljz;Z_lOjE>&zF7WPLSg+k- zrjWnDJ(-dgRHk~q7u_Ner3}_kJ;N@Xx7%W2IAAIZq?#6hk2UMA(J^Q)>s~7p5W%o844-r3)w-w7NDVR+md6 z#>;QwW=|(>>~CVc_*NGkpY4+sK%;K)rBv>gFVQIMGr-BVq{QjQ!RFmmx!nNmc$GVR zLy#%y4)RU%7ZD5gDub0%h>$>i0v3^xCA^~kGLruaWLVS!ckfK%1n^TL!A7d=u=?{m zf|N#VP6lx`IvyDkjcI}IbedZd+!6@dtMrkyK`D24+6i6Hs(M0m7ptyvt1t*RReEh# z)^!B16FCq%L4wD#VsdZ^VHf3c(X!E|Smf0hjHdlRqR(kM2z^f51$|EYO7uBR2cyqv z`lryRFYzVxISo|#CG=_UQrr>_xM_pY=d>@O&;HlaiQCmge++$o435#rO~aemO_lfT z_ZfrG=jN_}9qurapjPF@Ye{^gm*?t^us2&Q9k&2J10WoyIMN z0PB~K>U&I>dBn#NB>}VH;=E#lRNw3_@a}>?3#qPQAS#$MMr224e#Phl;Qhnz4k?J&gv5HjRjhj6aq>3^I$k^IN$Cm&!p7HPZ zrBojMGE)6VIFs>;PC!HYJ~KtC>kZJRsodimsv^|`4XHlMU}fQ8q^f4f=nOEbe=p!_ z6q$lM7K%I$1dwXHH(ROliM7x-2(99fX~Vb{w8{wsbx)JqI)bFZ9|Y1E`+v$$12dI| zy4DSqPk~TV%-P)rB;`OQf{ryo!ME7<-9y`UI&Nl%RPL2Xz6XXX+YZv%Kj1Ou#>)#< z<4h^ZVI@o`gAnk6-n(@4yD`X3Hw3J!4L~h2)#fZ?k@M_Cw=`*Jd%q@w_Q#`*>9&}w z;h~Z@GOVP`t*DZR*%`$2Zm|G{$W2$d7~F$y%}p2bBm-10xEZ(rll3Kz^sZN%U@z-zzkVF0`} z$yAA#k)CqCReH*>1+UgH*G=7&BIs$nMLAe_!>9;N zIA-Gcv`N6ybd|pukRLD(<~3*(ch_uj(`9{6G7jl1n+dHSk3Y0$U5qKVg;buMT|L*|~fnHu`g#_>_7@dtr;&W4Fi~U|PC%`o)(q=a5 zo$dl%oeeCh+{=JeZZ?uORpqVjLLMEYdF}FIxoz_~AQLyo2pB$`}^Q!+b*c{*fl`U>sJ&03>I=`I}m0&n<}Q-wh0!q>~-H-HP6o zjmEUZ)5jdjpl>i_6jQzFFzw)yFHyZ0StGDnJZhtQ(^YU-E+$m(S-(iv{8?1*G6tmb z^6yc-M}ME{O_wz+~ zlMGhQ9!&LQj}pB9KBj!yEozqbKSJ|i2i$nMfn8F0KqBZY)4QInPz%E)ljeh`hzt zhHan5uysPc+?;8^WYwP?<{^6j&#K?EUgO8BdvIXU z#dJSL*3V$7rgu!0Z#Mo0Bn}U^!5Mc z*V!rD|vn zvd53G`0!nI(%)TXKwdJEr)=}(n(Tl)gLk;?$_KLp9L>6e>aHNp9fH~`H-gp1lPfQ$ zvCeD&4Wa`}mdRLK7UQE%cu!8L*iMi?kY-AY$ME^HP6LvK5^Wjl@`PxNrSPc|$&Ytg zfO#%i06EXaHcHvtS&*v){yNW*!izQWx80VOub1mwgZb;a29?!41$T(NP{W}_!~kEY z%qlE$4MOzNAzaN1DB4HXS>~i(v`cH#K|J9&zfu>u2I=bw+#y0-5*daf4fER4BClm5 zIjf5mlsDn|e%=+BF4u-?aoru86zA|zbIqFW^aM@%sFxj#*t>PLwh*TkD6+7U>5m_{ zZRuDIpOO(igVE6d5(&o|00*#Yx&%n7vHHsf@UBDQ&3wdP%eoNCPC8V?rt)JC~6{$BeQ*7j`}(X+aPW6kXz zDLJ;K%482YzYg@J2vRnh> ztB+XJ4`v41j8!bue6k?J>eDVR%K@Uc(~EUXQyx~O8MpxZhh%{(u_1tg%n6Qd!kf{1 zFx%@kpjm1w{D&tHb=$y^IfxAkjg3HUB2*|l6U)sBmFrNBCGXgZwRF&kTExa+8r zL#NuP(mxBh-IxQ0vDGzdg5G~+j(X6W=ltTF-?5Til-Fw`%{CLw%||R;cz}nje)dqD zbW4~znvvc9eaPPw~1&^DfLs*Ca%$(wbo3J6Rq9iq$yib$$7-;_vOnp zUN(V9u7VY(DH3?d{GNc=dJwI=GdGYk9HK@agTj6uYEwK0nO7U>qC62jWB9rTCQ|cx zd0}oK7dgOEyfM4a)*40&h4EekDnASOHKPzadK95YM6}BU4qzS>F-gTm48M7-#YcpF zvc0=ti}7Uq|5+@`{WX!`Kaj`&UAFpuPb^^4q#~>|;T2r=I!4V0qQzg@E}zZ~%oPZ) zKQ=&S1~^%>I`4D@xCce@N;i)F6Pxq-y!0Z&seI^Hjk&OBgteW*b}p-p`L(z?wYujVa;sS>~L$0rlEGZ61oe6VBq?!U|-d=GF2g_qb*5I3(V_KAb zN{hH7Zo%!=9jtQYOnm;|zzM<~Ea%PiLW9gMC(u2*F>UQvv_?J>HwCTOesD$yOT;ln zl514?qy}1eu*xFxO_7PpGRz(-coMO|dbbPJeex5dEq1)&09aO*8J3vU9;3Zu6TxiyKIy_d$t-~QwJfpUY5eZ*Vb5U= z06ZH!F&Z?4YM?d(*r=9fkS5{S^OmLTV4O&==&>v!BQR{6uH$k z$Yt4hYhQp!^#m1C6L4m;`+dB`hxY>sw(Q?w*peC0FlQFMnT(0Z=qW4ZuTVS4UstdJ zXr!&J@dD)G@{4)P+1x8LgM_`U`O9?VYte{`t2DVrS!f#KgH?7IgWIg(F|H9@#JiI- zQNGPJ%F?9Jwwj2*)%jM@ihNqhvKAVT&=o$p^qK0miOroi8;^;g|<%a=z`5rcq9i=&0*Rvx}hsPB7 z`j@Oq8>qEU(JqayFj4I_1C!(j_+Oe!gtkCjk$v$2F;dy9p#+i2zXY6FxgFHK0raB1 z_7KX3ngMRNw{^Os#xBF&r4=X=uq^BCf>C*p)nIjaN%Q+g1i<*?+Up3zNw;+xyE-yr zDpLbxfUM_-01fZ3@HTqy*FU|UD))7k+g#u;F?C1Dg*^oZCYSV-$)7Z`0#<484CJtb zVsJkSIUNRIQm-(l;HpLKEw}RoXV-jYg=9_8r zzcr*qkW-dzy3bfyLxpSlv}r8VRE<4+lVc7F{=B2$j*^S5E7N{)rZDReeH^FJ-I{dp zH4A06pfBQii?Pf<$y6jVYg3+9wHL*JbQGkvqJDyHaom=Vc_9%uGZkrFR9+S|Uc?YL8@8V2rFP!1M_zJqO$t|-AOzsPb z+}B7Z3+^)#F;?^3!%c_+RLPdcGB1@0aweEvEx}Xv&`-6fz4&JJ0&^xO$gD#4AN-cK z9TIncQoW37fJ|PNtq?Q-CMDf{qb$#Io@RjaZ|wqkuFkuYa8sEflhUjl+Xl7sc8xXk za>g3E7?p?iyGGe+q|YJ{mUk0@fND#7-AVGlqeKZiZ&WfK+0Jr6G9QoY<;X1KX2?x; ztf+T?O*B6nQ<44uCat*y=y5aTI=hv>m*GBMLwMkIJ%Kw(epORRAYxv%(wS7!#TGEe zAq?|r6vVcq)C{|bTo>s?sJ4bpCMztN4>SI~D)XI&8`m5bk7<}RHPo-}*CK)i`>)w= zf2PGg(4xxMDk`&gz&H+p0mu#4qGpdlyz?{wqMQQ(&E%+!`s8pg?m{ zzePjRCD;x~gzoJNV;da6Ypup(~_DsnbOZWvT#qZN5&P>~BMvU*UFr>)4A zK}Cr9kgEq3dBKW2IjG1rv~ua7BClDIM+X&ILy?~kD)NpMxo=RBn<#STpdueykqv{2 z+)k0opd#C?2#Vum_wJ?0f%Zgkw zsK{1|ygjIu3#`a_gNnRC5e&ykD_2>O%Ag`2Qe^X>R#sV&1%rxgr^v4c6}jGue1T3W zhUt@Wpnb4fghT0%7_znhCHF(IOx=}KJBw;JR;$Tl*7B=^S|%Bd^bacXlofe?P?5`M zuF$VwE_9m`dB}x$lVm_9aLne6&V;*1pf$`H>k)$+)w1X zK}CpHA{}+QAE#Q8D+d*MhK@eysunrRid-_N$X1HHJ*db9R^+@vMc$yutAmPMW<{0_ zDuRcYJU^(&FRaMoK}9~Im8S+3>9-?C8%Y8GG8MGO8kXbndYH00~VVc4SJ$UX}($Es~fs z_fI|zm(515A!7AA-4yx1ftSID%Of>#D9}FEmJz^&P?G;;Xay0OpLrtQbNtHnYjILO z5BKaZdqNK5&y=kBd_M1{NEZgW%6P?-Wld);tZpD7VAFh%8wYNbmvm9{qZ;nqbZ`}WJ-7ITiUJY40;8ba%z zRv8|_c*wDV$~PiZwQ@MZGh?`<+T1o9LmPc>vA3)}w(kln&gPO_Y)GUsF(9PJa{tSB z%0&C2)WS-jA>0k&QafWAj*hrJB9i?hcEe%tkkyN{CaF zX%AO9cwN#T-Is}&*;-Pqsr(9=Qj2bQjIvof{#K*g@P-z=6%MtIrG)1bN?)&U$w{3- z)tX37hAl z7B^h_I!yp4o+TfXSh)YSHTUJd*SrW@XX5Mm{9KK{qI;uQ*}-Ye0;6UE%W*H zn#c{u;7-p&Rqm)E?wB1Iho%vMBmT0*)I-BnR%A+Us9=>Iu2`k1=8a_O$LG6tN#i%1 zgYAaGf4JeOPux&>r<>nrX828_@Zlh4H`H`2!%@2=efTb5Ox$oe)yi+Rx}`y!{l@Q-w+91nIhl#{8A^K~?{}A>g&fEnX#o*0 zz`@D?tANha5xO*t;TW||(Hod@Pw-3TrWkbef24r>nUE~0*No?j&75#df-IL;OWhxr zVgUICi6Fc+Deyh{+gP{If-%Unsh)1@MDm|--C9`)VtFnI~?dl2dGf;#qA~Y6h14%Nz zBxNz!qEY2WONTQ;(2LYOBvLcUf( zQ~I-&8q&x}D_}}F<6C$)JW{Ez*rCit?zD{jmdWXAH$qNz5HgM(`t~AvZw{IqKsl8S zmR6O5mpK1s>)b$hx)E|Y717kbJ9nj53yx5EUrH{JGdlwwE^(0}P&awqAWBnSpK3Sz zARql{-QWoMpr)eo0}hq^&q{{Aigb)Dt~@T0d>x(yvwfq|tzUw7%n?K~74THZKX87N zT_X`&R8>-eVh=E&+S)f&&xUt=HIK0+&)UyelSoxKr;7-nv>iB>vXYganT(qmjax$f2(i* zF5f)AKCPb$=zo8oy%gYq^DtdrZ^f%258Jx6Fx*3W`! zNd~x=>#Z!UKU5ou8Um`|?#pm~t0lwO=CH*f%I$RfjIqS0z=-7hRiVakf~HL9kySkf z#{FE{@)efE`aH|aCEto@`8AASK?z3F>_4roZ8n@&(#bB@sKiP`sCx7AL3g^&<9?j3 z^G!o+Zw!Rqm<=BJ`4DhEIQ6yJWqSJg5Cf?n1%n_8|2@)av5{>$dpq1wo~gBqzq!TvvudpS@Z{F)8l6FpbrA%35CBjjv5K}Y7~EjuARA_=mU>`W2Od{JenUF;)!B3JOO=mL3vc5E~mq4KVR z9nYS_k$zjF7%_g7K+zHs{(1cc7dPOJ+FWH*?TKU}J0Ml!5)tP(#-AB5v-*ob(h1 zWV4HW{+K>Qe6xMzHOwy{7tFa_{rYQ%`<7fVw*aKP2Eg%zDgZ@(cKoMcknG-{#Vm7f z6Rb>N-X`445I%DZgVJYRqx`0wtlOZd)A2NFl=tHoQZbgF{mp@6n*H!%-$ICd&U-k+ z%u1F(uSFU@7nmm^$%%3%-vWm!n`Zl7WSyakMwN$?H*36&0IEtiY%@$lsM#wrejUb6f=3Tr4AVC=lUkVWt)8g zT*xO<(cCanD)egT+Eq^NqQ)oK;BES$VafHg5|VI_@jWWH%nB5qE2W8-)f&xxzQYe` zk>?<;0_`q$6-;RIf+6H^EDIQ`-9!1iRQ6cj{^#SVVYA}0Bc#_n@p?ITd6|=9vbJi8 zCAbc1AMla4NtMiva5I1s=G#CZaqzkwYgwUyXrG*rQh*7w)9!Z0I5tnt$hm2EORFW$ zBxeQwctx0!J3C^2@_xTS#F^kMI6{d$VJX~^#Qgd7y2}^3b?_?6?X?wcLD<`~0>I(U z)?U_QeEaK?_8;+8*ZQio|4415z3xb4$(uTM_zrd*MY2s#)RyF!azd8&??`8R4Rg^! zvc|mXx994OKC;6)^jU%rRwz$`Up3`0bnk09IM7k*mXb zpO0%`6*MI54rcmDv2=}mxyM(##aENNsiks`$Bl29OC^d%NN}BOZ#SN9QJparAIyl? z^t7RH=J-Z7s+4EJqc*8LQd7cZBt*l-zXw7yF7JV)e{92nO)!%k`2VdLe7KLS)c?g% z$jL$0$@&=_mwV~Kcz~OePVeDWoAGcQ+SSN{eH`rlwwuh4fl+EpWK=y)<%StwVy}*e zyE1>ABUN66iYm~`sj8EwIHM{L*F~;Q?VyiPdDJkLOh14NfMVXr$socZsM#wTLja7S| z#|UL-)}j#X%@Slx`6z41+itM#=Ym1Yr)Nm&wxJaW#^5*DBWi#&0 z9v3!GlFK?l{r!-m9|UfSES(u7sPCqs6*m&UdJO@0q!W~6d5K+lXQr;aTj!;dym&WL zQ51DDdH!;o{mEkL=Nwo16;=Lvr%a;BUI2xBcijd4qd5CYDiJc4&BS|dgvy!NZI#5m zSiXvJJ}W4HpG-8ZcXljS46&3Je3Jg~+hhA8UbTEQtaEl8oKW`U3{X=PGc8oLF&pNm zQ`x~(0Ni8U6uCS}(Ktfh*$oQN=SIm>GvEgBtYD=5Lo2RPuAdp7l!xjmmObKQ*C^{S zGgIb~dQQe~k}{{)LtVC5+n$EJxm(4JQn~k}xRv$0fu<(ADRQ=zvhCe&FVQz|vX@kq z4{TU65*x!8C$fuHcG>O>BhO-xYVzi4g(bV=vO1+~hXHcWj=(vRNKxT<9-sMVay%nA zwH!@pGjTM9jOD%Qa^8YpB+E%y+!T4KI>c2rL_qkG>gcXubVfrstGQ!vM_IzfJIU!j ze`wT3z8>x6lAl-z=HBTAt;V)P)8e^*&|VvOG( z-aP{1+`r4UCmP!!kMAD2e^*&$o~O3!zv7LoJF!fV@d%ZttWLAorLpH%C;5p@vYBuQ z3p0S(pAETXybV;l8rnyy~rH zNjldX_wWKsUW+ElMs*ba2a0jmO+1FQA{4dovuZzhe)Y zG<&do-n|1V)hM6WRG>c#dkPGlKccc254#?wIqH_4mbPfOZ{nWXh&n_=>YH-9i^Kfy z1)hn2Uh5j=sDP)#ORiBS1g?oJntAl0MKH#Jeu2lGo;H0kYDii5rNQ3Tvx7AXU86)1 z?Hw89+oY?^SulfJ+|aP4X+zi15VjQD9BhQIMxO`4OZuusSd!Lg9YOH8nAQu(B_ zz;mD2ut|2)8;_-tzYzX3(r_U{R0>yz)5HOv$9GZi`ZG%2|zq9iQq&Cdc2WmHLu zfxw3tjD(uU!K{g(f0@KpDOF{(y(zm4vL8ygx`UhHOHF15O+RqI}!DJAye31_CGxekX}HXpYi z+nvK8^etvKI6A>}8>Hid{w9u~c!Mo%) z9(G>5SR-XjLw&DE#T_6k#0WQv|B-5%lrW9KY$HW9>>r;hSkN4)Gv3OUcyUfw024%^ z*LdSgkX+I~t6X4QkD#>B?6*HwyG;FdhZ<_6%KdXWE7`q2o&?vB-TPC8VdlaluarT; z?w9YT0^PWur^EjZ=o-OKY6l=Q+*~MA%5CbSud3iX?1Hq;=hRVu{qH*V8Zk!^KXsvN zRB4XfNcsQZU+)`YNNn5efMI@@jpd+?MHKiR?wn?Q`-X>-(HEvYT#!Mys_1+iP7fdp z4p(9&OM6=v4dq5^k#D!}^6jqp8*^drBD9XfHPiuqG^p{UWLxhW0+FzhiPu!ju|cAx zFk@JRTm4BZNvi`M1S&$6kVf0kU*T=!MruxaGbIqB56O}VW$)UN{SoYHXyFZf;}qw( zD2qh-=$o_%K)2D$Tyhp@jtEh?r>0~_poqZ^e#t^bh+6dDAF3(II>>;1=YG;S=k~ni z>GC}aHPT+DKp$}HA#?tOi85y|m5e0Vr@LIqo8ksmAN_USLO8ft9x4IXgDRfk+a=e` zrPwU9u<^Bdia529@GkwArem0AIIfm9sIs87g`5-%yKRv_&P2_O&VWd|1DR3Ij`L8^ zc3IlQMKl2utX9tgVMfZLtQ|8ru)7lHjRWQJCS;#=--R)-4w<+aqZzuZ!^dVM>39*Y zjAM9t1y49HX^oV#vK5(9PG&v*@^r{x z_SJa77Ev3}L{ZTVqSGfSUaJwTV#Knb{Iju+X>RkGionP$mN>}rIzlVDgIz=AN2*vR zXK6230_Ch=G)nl90RNYaqV_#=G>r(dW*UXZls*y zy(Vl3Nbp7xrF{p6YDGVRUkTv}4Jt3!L^KIuNKkK z`Q6`!69$ehoGEtyr1He_LhY65L4phVuS}Qpd?*GnM~gLl!m=}+6wFuodBzyYi#iLg zQNF=3$JQ}Y7H1NqN^b2hmY;mK?vCRK3tr-Vm@HX9% zYC^Q^L2g-sv`gU_=i@;ZOp-)SRXi!g3}(iyAoN2+Yh zqP{0%z(|#?*+_2hDafHEV`kz@HTe4z#|8`n9tkRnzs@#=nj-Mes3PjROAZyjAv^~Vz^Ll>>*0yvaaeBZnh~r z&}Be%LzYXta1XLjFVb;0i(RD~s9b8Y_DCf;fNL~Y;y+{b>WA=fO@%gY8?<<*nq~N- zPV0HG{E?n<^tE+cN1h3@hW=SS&k5w^?y?Jn8~ye#`8IZD*t4=C<@*m~Imw(&JD)i%LpuO7Gl2Y%{*QQ@z<|}6-hk06d8$erk)!e|1 zlxMIFxJJ39t8BWtZCxCc)3cFll1FNyuh#yJG>_||=XhxA^0_99WbLdI$sS@)fYJ8Y zt7!2n6I~;o!2Y~39fGTkx4X9=YS>yt%Mw)wq(8DwR1X{y^_QX>T(l z*wMjk7DAb99-nn^QWw~BoqXXTp^N~81VD{XsG+imB3!Q9Z2SiN`R&#vr}SR8vk#6@ z{=S@7IRFv}PjW!4HN}EB-IdZTb-`|^=5cBXUX?Fpbj(PdNwvy+T2Ll z^si;3QrFJq%Q&WSUabXa#=4K`{b@rRstvuGG&E9H)mF4vQjRS)<@ZNpr+0PoAj(0N z8<2Wkojl+lZyIC7?LPl>%a{tJ?w;?Owd#&;kDuaI@h#S=a9#9B%Ujg!rpRg4k=|gG zd$BuEer2rZ({7}!&VnrD__CWKmsE>hVMPZSb~mg^=k+G@atF%Yqan|?9#xRqsL3_T zhOQV|V(hGPdp3g9%j)X%xB7-V;peeu0$Mp)RJ&72c+le6&Jd#13e8?S8Z`gFQRskK zt`|mg;UIy)Zlo-?*ODt?h(A!S`+r=@kt)3o;@!e5F8;k)Ky={_pXdO%fwZ-?==UDB zSs4s`T3eBSPmlVBb?I7Nhm%6TJ5Vk%jI|BDHyys~Feaui-AMU>vyIS$#9H)@rD`}= zTOdn_VK8hcO%5UzON_>z?kc#0WEI;;!@n|=a0QlPBc2Dz8(@4sJ=8_`;s+^fcuf!-u%5QBIvg3`<_KbP4 z$$eLTV;JJJS{nr{W)WT^RNUY;aeo1xw;}9=%(Z9Q=PI6=JNP<=Ma5n+Q)T&F>?I>r z-f$7m=Yd*P-gK2VH&fQmr6%kWMILieOWw`I>g+~ZX119spL?OAq7f_gs=@e3x!Hsi zqVQp;xWzIe22?{em4TZn+d2Y<(UxyU-;#F+$(=(Ywv=Qi)R+nuB=)9=aMG%+wEs@B zv#Yv-Hf?Qk2VsJ8BV|1jr*+!q4kD3IGUx$bx`Q-F?xN(S<@@uO_vhUqaz7iRz09$% zBXD){a>mHPeO>UYcK(1xcOd+uGx@@el+On5jyq76?h4fJ`Okl6n|E|JR`ImmlAeaw zurvZCI50k?i|FVeNW)|e$LO(n4Rft1p_(jdgz?c*x3h`OT!O+b=F-68%!>?q6y%wjlA1!#Bot$>!R@EcVP`+kNM;m>3RD^3 z*vK9wl?HUEq1=KvE%ywGaN6w&jM+X4Mx#;Xp;`C>cviSC7a<>vHanF354><5Q_VW7 z|Jsm9hLK-6l0v}iTV|Vu9dU&Pz=i~A;H(p<(8`jQs#RoIsa45gt|k*hUoV=KV7*`G zc6y5@e%9p7E*o=AgfE&7=ug~76W#)|0~gA&*=09UwhxKNRFDIs1)!4yx%^Ldi{xj4 z43EaP*i_`}{35{x=yp-s@$Ygw^JH8Lo0(ECO{)B2NJYZ|mg?RbS|^=_7-!n3^3tFc z`(OwXKn{w}jh#iE%$W-{%*p?uKN%OGzBFxe9Em&njuZku!4_{2kg2dVZI~j_Wj@u8 zs_*`qNPgfKNcyl>=HxFTi&Y?`Ayj!~76<42TqNJ3B(Gkd6|hcU+LQBh244AmOI~GG z!55d^q6n!MP7sBUpFli*IliFicR>(_2yMUaHHQzV)a`XU^V-x#JTi#eg|cDRO16KI zxzto6o0s2}PItXo$H{PjeGWuraS~$@dQhbeiSpQ5L+!I~8nhf^lnH{xn_QbtBufcr zGarm$do<$C$|exBv=_3GR|(YIllJ)%R(G-mz6N3?%fD-o$2$UpDEI^Dt?7vxk~)&{z0MZNN0vBlv<|rBD8=>>A_+#-?)V zkVqG?Wn(p5BMk$1=%SZukYCR(FqCtLM1aNF?HE!Cq%g$Cm)Gh_c!B|}D8^kVOWFaa zx9XxvE|l|lP}ywGFs2nkCT~#L*b#u+WqECsfb@p@()GRz6S%*b&NOhe&tyCxweTz_ z6%^NCk5RpMVi6iE#|L>i%%W4K1j_mX4zgR+<_;m*v}=&1a|35NA6a;F2pg$#kwh&2 zW1WFcb`5ez9q7~@A~&~FNAB&ZU+(LWsFfzRARP!^)r@&Hh+~&Q$>EXw z&#tdj4vfG^KZFj1^53a{PX7bxqFjQV8S#NXHnF#vAgBfpt@}Ti7QBvv*fS1Md8U&@ zi9h#~87ZfB1g!oCwKmslXM*%!a9|%8)CPI0)-=OscLY>k(Fs!gArlO2#kOVbn*N43fcEeN3AzP7!E>HSgL1ooSZY zi6^p=A=E~> z&7SY_ZjfB6)Go`jQAs+|o08^jO1s-`-FXEmKxD8WcXpL?@*?gr%Z;;WSGg9JmvpBA zPuKaU$NkeP|MV;WL^xHW5?V#ce*f(;|80$bdelFy^-o+)(XO#>tG|cYY^=ed9Mn5x z=h#*aSH`Ea)tm=HISCHIBDjvnT7;ANLM_U)0Jm52r5bTFe890TLh2Y4nFP!^@_pa+ zQa_Mw=sAnHk#-jt%ITSk+$fPnG=$0tnC33r_v>VRT}EC1T?BV!s^QT-GRKGsnFOBnVq~})oE)73T%}7Da6N{(88(Q zhwbxx*nVEFu+Wj`e1#Pf>9J^eU84*j&{5(Jb5|Dg3%Dt1>wZQ}S6Efs)t=>e7=Uj@ zh7JAYd;2lhoWONxuC2uk?js@@5h=ojT??jbElkGTgd0^(>9(PtmMuYz8mVvU{klj2 z)(^V_md@1J?Oe@-jKwCPNSE7sQqRTdo6`l`NKlZXCXq|B71`_!j4&PX8F|Mf+jf;R zO%RizcnGjkCJ?r(zbvI-CdWJoU_Z9Z3`1iLa@9%H}u2FtD*XDAEi{SC7c$E$H zY2!R4OE&2|p*Zp_kOAkrUqQPUbp`CIYnMA+MY0j%Dk~83-P^$t${Q^S>JF9jyRks! zB3%zLQ9BugL#P$hOG$a3_kr?4cl}~xy<4t{}o*y9>bl1{C(irrp*Mz zdKZ>|wzD(O8i1KnLSFjJ)@+*9qRb7Nsg+-BX@H9IlOSO>Pbdrl803C?xKur zK&iv3yj>f)5wf#3a_ywF#~;g7Q%#EJgCGB|oEf`9c%QWJi*o{$v&$7)ewFbXBj6uV z_Ox@MVBn53=F!~bd|(V8o-iG*xQaVYZtk%CgvEBpsr<47*);Di$k?@*yebCV5bqZ&%BMLAvf9mL)tUfws1n(tTY^FjODw(Aycv z8Qp=rn~C^e`Qh5-Gf|bJJ3u?{svb`hF7M@S3jV5W6HcfH?>;tcQZ4*Kl$n(kcFSqHx`sYsn*-Rwu z=`G&DQTh{XyRbHDMlb%UU$8#3xBnZdqEY36*#*~5 z5C{(s&o&rZKO|~uRj5qPQtjusoj`gDWKl<|Xwfk;jU@d!&k^kZMmmtY(6%Wa$`Tt* z)eF^l@PtZaJAyvCu`x!1_ly?P^X+pu40zHvs;uWhh1*Nn8?`~e(&dp~mCB*YQ7e_0 zY8eH8g%2DF6GF3*`G4^Z`iH6q?XS~8{=BXMEOnuUX0@xlu9imQj5$UK@Lz04(}&P8 zCm>mXYnSsq6S^R&v)!n6cJI#^aNnH3rdx9DICCgj0TGcgB+*kbRm(TAETeMO92~eo?OMrm`UXYW?f+I<`3kf?kQ%GpM>5f4^Q}ZP=Xi_l zIvWR(Z%b>{%dPbdi(DgdU4EoXa^_kFl4fGtXO$HSkdq~*z@u{mv~9D>uAS1c$^qvm z^1Q^t_O`dvfxXDU=;LuyB($rn@_iuRHL7fxg$ayP`itC7V`(0=wpZmMSe2w@>8=|3 z-bow(FlqdkN#l3rlCj^PHhddnC!s{6+&;@7{BDz(y(-R{juqY8-c6fcf{MkBk{d1L z`xJ948l_kwjDWAND=mhPaDjX=vz#KQm+GKPmzv!2VqJvl#|3h7GP<2PpsLSINPh+; zx=~3D0T&ZPRIdOB@|NVH(N60TS4H}w*%I74Ua7QRP>)9~%DL;90 zN>7;(XEKLciBo$Du(pWvcq$V=Ah`jzkFnrZ;*8Ff=Jkhjvui|*_wU#C1Y=#JJeKse zwkN=!r%`^Byg9umNE|cAZA*vwf=Ox`w&n`0I!SxoUuaG~?%L#zkW;gf`wJlhSFwK~ zUh9nboNS3OF)?Q`Oatg3&~FqGs}P8GeS<)-%bV6XPy8_n>Kyh#I+%+D482I3+!X0e z+;mttiHs%5y8~JAUUx)bx~J4}ba}OnRfF2#h9Pr_IzrAj65gn?qPyUZz}n8kW!(Yo zZ_VIaf{VuuL6nEhS@vD~u#Vxo{p_C0BNQiMH$_(1$i@Zk2w5%>v?kVbkNYs4$wTod zCb}u|OlJ^}LgfaDyg%UAb^;x0Ha3+8+I`K+S-USXT9xUWU{u$3($%B9GV}f#5V1Rg zH|{S~mcrfZPIm-(Pq37@1LYR;ej^H5KJleho}>Yt^&ug!?Op>d6;ysDa@A8k*)mViOhAz=DZi;+tR`Hok z>3rkq@lqQ|JC~&{N~N`$kJBb@u_iodyp=Yga)pa@Eju)qvo%HK|CHFN=OnVE+!OI3 zrg_Hs7P^U?NM+RO@Trq@h^nj4lz{BNulG0zOx15(iz z+&&n{?2L~^(A`JPs?1`Z(OFix%0+G;`Dga+jLv{Z?1)I7P3YjpBGip}j|?Tmq(bt1ldx{45ZWS!w} zUn~IxH+K{g2!G1748D+240A6GBbm9@n=1ZGhpX(&MD7Tg*B$V%Y;HlG z0TIZR=t%4*%W|GD(_)JPF5fSVq3joy;o=B|&+|>ck(|Ga1-9GNb*o#{ToGTlVfqtq zvch6_lEr@-!3WhA-4QuiD?5!spQJL-5x5y1n0wt3<`7WS&~R{bAH3Gy#EoV5{+Jl7 zWgX>Y6~3T(x3`ix*U$8n+9+?LAH%*4V;2A&A)hiF7T6?X(yh!6U~=7fxindvtx_VZ(Mi^&k1P!Mok1ye?fV}@ z7i%DXub3YDy4inrg;$luHpd6)NoDMj z{X+|v;~%-Uy8!opxXRTuFVA6HK`uXrJwyE#c>;z8<*UZ2pAQViWi`P*aiHS#-O}b(C(WVd@y*EuYwB;~TL8EXVD=wuYQpYmQ_s~m zc=p+;O8-nlj!cpreY8H}?X|GNDx2yf+}ly(hcKp1m_U_!Ik|AxiZA+#7x;>=)<>;i zZhte7Q3mrLlH&`qmmSaJI0X6XZD-^5h{rcm?j{O`m3uayH=Y#^I8;1LVwMx+(u}oh z452rCqo-K`nt0Sl#n}4L1|gn`)vW3WGvRA#M{gt@<$4LaB7lY=2~K#-d$P<9@SB=O z#Kl+*^Lg}Kwv7mvxe@ftrV=>NwNi_6e2|dodBf6f`jBRIkuJxj6WQSuytV~f~*Ikv1tZz>BFh=m&0 zMopT3vRNm&qwuE*?W&7!pg%lHEe6qE98zj-A=fD)N4%vtK>e3RoU+Av+A?!f_6Ru` z6|wxFTz&U#`Rna8cvM-2RNN@pyP~~m+{w0vbg&osT`iigp%Pr&iaLVpH~wLVpLtPn zKHF(Zh*5rF4by*7eFM;B=X@gFWm{tohsG^vo|S?Y==X8wi>r*2DHe`H<(3pl*$8{ z(jvP@WL6_X~Up890N-4B^!l$?>R z7~eaQ(i|XNotsFwe!gELlI+OjRFSuI?{pRq3<70B2;+mJ*ZHeqH8#b>h;!aPX~k#b zimfU)bmPwVO%Dz7Rd=(<{h5*sbO-W~=`ZZC;A$e-3Y=f%%TZmUR6eoof@`r$q(BUz zVzXj$JRxWAWqoG9EyK6dv0P7fE4-*jg{BMsEBmfFIcA5fq(RxYWbxQ==4#hw;J3iF zp(w3~A_#WED_om^#0v&T-Xwc(SJ{n{9hs=jTvDjL`g z{-j!Vlx(f3$UW?EwU0pMS$p#s=+BVRgBc`sIRPRLTC&IE#mx5(~UvQ?d)C8&I)W(6t-IiV+Vco8XTAmQd99&_MjR zIVZj67&~UyLlmky-1L+ASBL6a!1JOb>s4VL#m|NQ@*@KBaUKg2?87OD@v;QEk zok)_myOH^6@G3$$X!2Q8tR14VLYUS$_2z0gO4c*(K9#+#pX`8Zkn?PcZ}5RW&)dC3 z13c&2R9Z^%AoKW)%~9DlH=bc2e;dh_=HOyI{}xG#iWq6f%jI_L4w1))l$4<_8-jP= zd$aMJvs5Xln4xr5=kyNVdjqs&OL=SpA6ieRI&-WuV*@aNAfGp665^94I2|QUY@<|e zFx?~ywjHRAV1uZ|WD%E+(wxQHSf^!wYbc4Fv8R|Xbz=VkbFGz~(5U-5jApH5y=)bdfAO~B zt|v?BH40|{7;Ribj#9Zfd3&$D#lqHL3IspqkLPemgx#lxPYzGLN>cB`+6t>7|5=9J zAu(}?!U;k7woXrvvj@oIb^`$;X;8vm+$f8i3$@ub$d=j!3187sR(yS6Q&^E^4VTDw z3&+R_1({s7Fa&Vd3C6$$^yF{e!O{*;K2mFk02fgC*nXed5omTM?vbRwcH?DbH>dRp zTTvGIOoP&Q5-7Dt<#`p1?pT9T52RHXmH!U7F79tRjs^O?67XPcSIvceVV;MmDMjt= zX3imRu62m9Z2Q~-o%i!>N2#o~FwRYsuw-aNm%2td zpBEqoE)(;bHimU_PoW@NIbJJz&=<2q*w@#Y-FcMCBU#YuEe3^u%Gx!mm!{;?%w5y(P(zn0y>M=Mc|2?0krw$#PY>g;tg_C5V&I=WATc6~Dxu6HzRcQP%kYH@%Zu8C zH?F*`rHQ=Tp4{38Ed3n>_xT!7!4~sae~V-D0&pn{ zwiW(DCz;SR5{9ojsuobjodGw0; zHX{1t29_QkkLUljfzhMm21d!Vv9Y2>!)w80Zi?Jv(SXJVKUj+!oL6$^-4r>+%s=Bm zDRn=IYpV)Jf~?JrRC-T z_J&RDPhw2PdsB`|V^-Um(p7zANP~Pd7X+}2~s;_4l19cq27wvYk4}g|5(>Z7_F~&Lv57U`}Ysy(S$ztm@X!IM%%bx z0c}rL>M}mBTJr6hhWcJcr??w|1`=57<40Rw!$4d~<`(#Oc0tDw#a0J*V`f2O*!i`{ zEO}<;iS6Xe^XYK>RXRRTPLwx3PO{v=j&24&C3H$83lZ3lKExgiHS6M5c6`93r!3iM zVXH9+HbR2u`WH!i*+!}1s%Vl5^jKVFf%cL?BkytGyq0tv{#t}Z+GO!rt)_lsa~6?} zww0K?R+^g+GR(qF@r1;%Q>u!nw$`-0kI+=WA5ku^MTV-)JWOh)-MsoY{GOyC#4*Ib zq^P|lB;pxYlV?qY2{k7xGdr8*N9Am+{ahZUUHv?kx^3bJ$8Dai*JzA%C5u3SC9fCB zjB;uVp8mhmMjlKWX)!Or)H!2gwaCUlP-OJ~PFpbrz#=WW<5V6p&Li{}%Z7wz5W+V_ z)+hRnP@eETb{5RTYl9m^d1=O^c5*x#kiT=>-`rdLZ` zK?!%9lr7Zw4|)#Us6S=Y0O2CwxSGFy^#4seTupQ4!T{BTz8O6$AoXZ&M4N_XueyxS z^44nA+pVf)dho6+woTVy<29n*APeeCBn1y;)6Bq4;dVu5s;W?~7Gt;Sq4Z7ip9MLL z%(t8z9FH*2X;De`AIj}B%OF5=!Pw0o#9#C$6Qm#Q|_3jnEoZVi=^C#Z;jsK7i;7JCc zePED$Uz@h#@mFNPUWwb^Mw+&cG*ijE@5D75WG6M3G;7f_2o}U`8Mff_jfinj%+}wg zVQ#pcjESHN{_+|wZi8{oQG%sgxRA(K7D{HONh3%9W7gZW8|0qa23gpXu#YMSl;p=L z?HJy;e23(-f zlsC#Xoh!-tc)9Uj4Jrc}%RS8%iNt;cDmNvPScA$I^KZWr60UZ`)REGgjbQ6DCCR`F znTrlM8(;5*#``p=Y!~SGsfGvt6a&$KcDmU$%B`K2_>uNit|w!m<|cugWT@#EuqAg-2^$7KLQZE1cVva<4^M4pv!u-+fe ziH}W#65C+52a^TB9GKo;Pusd!2IzkU&B)ynNiK3na<#l`kS8+0lTFE@=K~ zbeeEW6m@M0{T{qCxhDcCcf>&%4Jw;$<>1*Ha3xTGdGU2_iacrCJI)|x8!RxlhhUU% zQ2Epr@^pjXKMArbtFn=sBFn77nM~?r;vUuCcLq({tl`)^af;w0;|dqZUqmuFAoXrh zlzR;E*3Lwvnk)IOXwguQaZ}`NYnOq_Go~V2?54<7_R>0ghnF&Ji~Q@RCi$C^bOw2u zPIC)!n;XIne~YB6EFI=`-IL*ip%koXT=_=;6Vqlx_?`e0> zS-EYUWmzWZ9*P=nhK7QXyFuxUIG>j{xk>q4iHyluqGjnRYwwuEayI(=X@eiqAml^7 z9VyXe{}IU3odI{4c#2}uf4^zQ@req7ZJ7$EBqQ;EloL?Yfw8mueF~~u!k`of%ib~B zcm~t|Bkk}Z^O~R_hFtPU*U1xefRTPJDS;cJDExX6K=K+cU|!7XX$^Iua$#y5E#IDs z_HF-}Hov6KJOJw}DzjF~5Xv{r1)jxu} z`Ot3BFi5=x!-2d^Hj|5b0v5$WxA_s~VK7d$kueClv96@yeCB{z0H=jRqQD#&j%GJn zR34(88+rpxLF;cIG(!(`Ox&|I4IcC$R1=#3sj~QG?ZG%&uOy_+rK-c z;x5X_7dHP>>Lc@9{(Poo+B)5oB~4thKy`q6qG_GeD;65$Ee4=DEP9{CgUZQiLw3^W zhj@H)=1#50ym3)RKA;)S7a5|+bT5V~DOH$#5FO82@3xeyP3yzoQi@W9N6E&TF+mZYrI7USNt``Hcd`{8~G7!Ufj z_a`op``XK{fk*`x$gS;xjD{0wP&wU#6o8cmIW-MZxXUh~Xi&K=#|}RDX9X!VsC;C6 zku3DC*eb7d2il9-B^7y(NsrK0++1{tei2U&DTQnVau?}*)=Md$&*S%{GSA$8&S~fL zzFFrpbf0bEjtwdo*GKa7tUx}s4TIq@OLly$X;J0QIf2}6AdSF=rX9|Dv#j8Q z&0$yvCa@|nlk9DyeB!tE)Db1wZhk{FEkm}*OS2%s-mU`W&X(l*A(839?q|KcseFq@ zvZkY8n5(GrC}fx`Q1bF{Z6xbD0(rQ$WLL@oG9SfJQV1DW=0jdKd9#o_iw;l)SKU~2w7N=h`jm`GsV{AL@%!g-rq3b`jZTZ zJUb+6YBp6CNc6Rn0x#}0$ThnGB}Lv|Za0X4BIVNEO46Q225v2-Q@Mw!cz^PqGGR0W zGf)DO@nRR5=kcL5%bI{mFb)DVk9a(zIk~H53szQh_&a${3Uptv?sCeUATPhu3b*ad z;O17=mza0<)10{Z5$am@AVw+&7*aVPMJm-_`+c&QgZ5E-S!Houq>E)I$hCSN-Z56J z8MYGkkwMETuH?Qa@8SK-G zGG>Nz-EO7So3L7cJ$Wx!62wtv!3pOt{oR~5*J8@Br1(x{e`;AOQ}32-$ED+WB59Ay zbF2b;*@C~A%q-iB9Nq#sL-t1}3b(CO(iU#DD<>AkZZ%Dc&J(%A$gVByn{_5np3`m& zEpvObE^3y`fO2`h-b@EDBwJ31!xS3iojsUevm)A~^Ec?y+*nU%RsKG}emlcYdSyKs9d@1d|B^dunD#p-fC3OFtJ6N$kL=eoZ8MT-pAD;k zs5en%6spncR_{MOo%ag64pF$CY!fq>XYE6g1N1GNW<%xRn zTs43>AILPFqn*ac3q(l0SI@-ekHqH3d6gfhe*?N%tl=1KHO(#+a&kqo=Fj;WD}0S) zX3tu!C2oqmG@aF$8MyC}C)m?~=l7^^<}W*;dgh^iRp?w z#ELK@(}^Qaenls5`cBRn)X5vx$(hwo)_jdTuc^*+^hs$K*AHu`?{!DY6*J11I8~k- zYOu9p1`anUWQInrP98rYAl>CHV~qK~=eveOuHS=1oNu29LVDUADc7KyqRy{<8|V5q zNX9^X7J-IuPm5RSNV(749+(|;jXfD!aRGX=o9Qj7i>iCyh18Lw0xp z$5X29+%>GiWDJvOxBw+Qc|?|-XlwQU?tm8n$y&f|9jWr`ggBhJn=Se~C)u%KG0xmA z(r___-aiRvE!>KHAt(v}96%wX{OX;m0pJ`5Oq)(YDaygjVa zMkvWYm*sCbXrcT|DhF<8PgnMW;b_no8PZ4lX>sQaacwe>`~47atj^>9-J)i>3;z$f zb$3HdUmTy};d6$Ur!uXZ8lpf{B6z)g{XWWa;)&11g9e&1oT zCcpA)G6g+$n8vLRo1M<&PMe8jDuQIE^78R9ST@LY!&w3Lc6riYgRB_tQH`ew?4mc_ zNLQ-s-B@k?GvB&Y6lEnrjb=p*oA4xrD1SUX=6Du=4V})VN$-QU?V{nfZ7(06Y}*uu zY6|=Yd3AWj^ZvPEXlcHp`nkUPufMYT3&Tlh!@xMOPz9s$FmPvnMO!Oj;IJw{}YG3YaZ%J!6$fd)9zh|tl!QUstEAx&2`Ei;LWAxl?LJyxBVU}_f&_#tt zg_@JyBFhma!%9BB%rd9R(296pH8VwF^~+6%9#dhw|PMc5kbGe8N5&uAU-aRNvpr z``8G_9gxVENY=U4QEE2nPFHmKNDi_YkLFt&Mr4cp@5FGgf6TC;8fHW9NoMFdUsxGzD=f#~IE-SMJ8$}fvV7)9#LS`0n(pfYD3@OJuj_VnHO4{d(^y%&7Clp%o2q+>N{w+MEU?(NM>-L_YRVU+| zm)S0#fXNj7hv%i`qppHV1$BtZhh1ekH5*YohY-29$};mUIb7wvOq6FDp~}CR(&nxj zV3X(J9N^Fre)XyB70FK^!?12$fi-Q@wAPnoBObxuW{|5~)az0D?e$zxoVb`ny)DY|KWS74Lb(o~>He>Gr97&>6^Yvyt6d70TjFB&T!*D-F`4;Y z%&R1uy9+9pWk^5Kq{YG}d7T&i8TO`0Lpi^zAa7)`bcB8>X(*J{Omri2R_Q0aL(cEo zB(G(oygb&8E&G&Alw{1@usGej0cQ$kBba3zL)Uf|$hgp^(vN=3#8J5qgQCivoq^mV z5zcrTTGA$rSQ@Ep#A)87YdM<}L{ciGOHi=c)a)O*p>vbmE|Fa9$N8#7hAdzr7x_0Y zXHb?`Ynh^cqS~GDq4agHly@@~V8M-6dAqwndUGFx_@04vsDa8^-J4Y2&RF~i5E;^O zfc4L^6NL8dEIiELweCPgZEm!@(H)p-`vv$SaKPO)C3#FD1a}(wfW4Aa*^WvxeOjK*M2W0fb6UFc&pYsED@Qc+GRsdm#El>q62^sw z>A5tuqDARs3%i5l^lhlNbFxno@Nmt5%0N#+z8$q0zN!bdb21#~D`}17Ukmt)a56U5 zn2RXhDq7SECl;rQCM|)bav;*8TI#5xJnBlaW^SPJWI{^h#By`KwMB+UB-)pMvd)&w zEyr;9IX#@h3aY%uWrJ#g8=M~k=R-R%a$!dxub2{-b7ClmoU7)3C>La+7E(36229A$ zGm*U3T~N7(02+C+JJ4(w%t8(f7pg_K$m4A16`TQliJIgbP{TRm$*S+P6=pV176UFS zv~&$}4i-U`Um+Hlg2W7^Apr*RrYp%S1aMqz5KZpUt(;y1b-9coZOue-Nq1l=8b_4K z`nXBuL%>AkYDV-dBO;wbD76Dz3_@43th?OSB>71Gt-N@Q{7s~C0=eijk(Ky)Z6s?m zyOzS5I5EJ)Va`Itu8WZHexInVS>EjksI?>?$jq{kyH##y?F@V0Vyj8AOj*lp9%10O zWg@wg0p7`?s@%r!doq!EPFs#d%{H^hz0hS-$7RQVjM1nlJu5ZOKw0~xwH4(zr1J#y zvxk!`%|LkQ09EH_KDX^PE=1)zKbTG3fn3XAHZhnRGm+ejdy?GhM{^gWxt*okz-!K@ z{9OQP+)12BqN|#IZhDw*HP;JVh4+@{=5mf!;J3lB?w)HS9YJLI3KzAYw%J={gMYmn zuU98!H(J%z&ZCam<)%E#%sgxfpc?!ts^yP6&@*2B=cU7V8Bg3klEuiW9XsSRTZzS% z)IfSo{0tSRc>e3f-DO8jWLM?LsU3y*0G;I`U*SCuM`t^uT3D~zG#{Di0K-%fk!u~u zw}&`g^Y$tdYOOM_16>RGJ!+#oe2@H>r8XR5TVhj@o#gmZa|lgSm#MryH;}Vj#ha01 zN?&t9I=@SXG}N!|YjJza`*R9(uvD?DPF&2XZc@2xP9XOx#N(zp1&xWvFFg%;XigB% z@BtUeqjLg;7kSK19w%8Y!+8-5>YxE+&v=4;@d|pmYtANLUF|-jqAp|+$=T`aFg5CI z^mX`n|C~Tp5(X>%ezTW#2YyxQ<*lx=J6zsm0sH5Ixm)FR&=%8r%l1m{o5Oqgx4^$+ zSVhnMt}*A$b@GC#Rq}2$jzk1FG9Z-~O_|)QW7mu&&z`(sBMg~RAh&5G z%l+uC?2bV|)nm5GH(ZoI6^Mod1W-yPhPQV89nlMDk2c$o8s;PHx_4n@#YU-~d z;>$yI)^Dx~E z16hWW_|@`Bb6oofzax~^?MXBt2CQY7h~~&NFmVusK_Vo!)#?h?&L)FcC zl9xfN+gtj3cHP;h_(JJ>tEK(BlN1isHU6Q!-4gl{UY6#57-ve ze8y+N6AAlH4c-%gdL8}K!8wQ!#oZ$o^zlD;zeX`nJi@q49>eI?(r}g z)8IzSJ=w?|rgBS{A@2LKCDEuw+D!rERc?^~!w_%-H(DOVGt%Oq&6DOzfRYkkDc6Ki zk>uzMr%l5|oilD_+Ufm(v`i>l)T$l>Vb8n6J&iyx3f*Dy&miqUuntrCpauiv z;fgxQ(8+q+L*Jvywwegm;Itka(lWSP2#hatUQZy`)7FDs4DZF7l8NMOp{+}iH1h1% zXc~|MDs=e2D+tEgBDDhw4VCX=t{5#xZjv7sDDZ9#@{jHR<1C1*H=gJ2Y{eZWS9gJpsqHhGI0cE z!fqYoR%OUAHQFy$i-u#cLTIS`HS5?(0L!gB0Z#Y8QNs_?adVR!Eq9o9pW3~KOwd`g z*i>)RQ2xT_-xzBM%pCGa*oLsa4^qv-q~$NH<@v^{`Ig3d0t$Vnb`I{F-xTW8n}>EvI|#E4t!P&KIM>1bvVqji8_o=2FkYHosSlWp2q zZK}`i^6!l&N}WmQg{g=X!wY$ZQ)X}|7xx6@0;3mmSRq@xhL-V-gFjjpW;{2CvXoM< z*OY#b7$|42dJ{?ijp@Wz+i22VB?kxK&uA90wa;?~*tyFyT+C^kPN~W97@kPJBkAV- zcaA*6LFG)@Ng5{R&!sn|J+Ckg?-D(RBC@(CKxG3vnJ*^?spjUm^j`7;8=w^}uvXOI z&lJunu;)oKS=8zk|NYhKzoGh~Ce6bEku`QNdC|_idF`f20{6cq?QjP*Ak~xIc&`g| zd#U_76V0cOV-mS>X|=|D^!C(P>&DCJwk~@q*BHk&zV0RIn)PWPr&2=~Jq6!(GipLA z+}CYkDrehW-!lWZm+Z(?T%%l{6nLnc>jSiyhjF_#dHW06vz(( z+`7)pM81gT;= zwx;7))9Vqu%$#E4i))l$?=lc>I= z-Sf|E&-+v!W`N*5+jaBR$GO+-r8z0>kp4^rsLTD`o6sGZ9y25qDz;|;87gmOD{hMB zWJOom8@BesE`W9UW!ndv>&sc}p)c`5WgJeCd&w)=2r^?@*C08a^ERZzT*B!Dw~@c6 z8Nb@SRQ?e`HV=uE1>L4>*#U*0ABSB4`~1mgrpGs}v|Z5$W+i~HcWffbPQN4Vi6FJr zR?U^PC`6!EROWZDbd4>3ZRjMlYh&V{xn?3c!In8-8I4tLxX;4+TIb4SA+9kR0wKpW z{4%ER*W)TdSk6sG*G3%a(b`MgUo7#ByDJ^q11th@fB6N80dijT>tptH3EXUuj&)hK zB<~o9Ez7b9os*5;uJNPWSRLK%Hag}FsGV9J-HqJv)?Fq^SL}(u^zL*-r?>_U9T%+J z+C{9a5?8WX`-Gbnivg$??YHM;+IVZH%0s_Hw>+a5-L#1B&ilr??WMB5lgs)nC(F|sB23y2A?9nv zR)fQgzB7B^@5XA_tkSs>6=V*_)ne=zbW2H$3Ej$sREAVk{EeTYku*eN0jW#IYI8y5Ypf*51~5efY|of2`iyO0 zA*KgX8;vz%O{hMEf21&Oh1;{)k*uR=@QhuoBa zf$PgynrGlqH|y$!+U)j{*8#vWAXVuun@9gAgiA1@)7eUGZ8DfSvoq-D<=cYw7|^S9 z4hAx}m%Jq&yw2f>rr{CgI3Kamg;9(o6TfZHQi8x3HE;|9r=xnTrHfvS+dbs z)V+4Q4_AcmB^T0y4fK_cO*DC7t)-G(Z7NmJ<|b2+=3S57oO2PKyS?PnwB8FHJE+Ht zXJmF7-@pIE=@1^PCC8&7UY5rQggV$G%Zw>yKm@zBiHq`i`5Ab)O^EEW{}NBz>5rrn+0I%TcF`6rs=B$E%m6V)*;XxB zHf_Es4&+T{3QY=HkrM)om z810qgQ?YB@W9h(F5Xwm|a|VZ8QUuDghkmRJ2|kMNx!8YWT45CfbVtgy-FOi0rSf#f zoY^ov<=tKw9*sC!J6?L=Sk_jMcwk8Foy;lVh1d;%zJz^-UQvopMzw{YVvGE&VIMgc zW{SSv^nINZQ@9xayaCmlQwC2M<%l`ErN2rC{TPGRUJQbeS+CQ@5$EwaK|-JIG?h7o z%d`xOV9Gq=W~HIM{7;VmlRKznwEQlu$noXGmvigmX^(dgjU^OhzSa5pEW$C#a!%Nk zmV1N$+pS)Ol$ocoq%rXbB-rN`fi!a&@Dlam(R7CY7RX#<( z&0?K7#5U%g*@11$yBtqhJ`4J=mu9~$3+lJi#>SaM#hxq1n4L9*-7Ds0v#~>`5nCbC z*{h2r0_E>9o-H-SAHE(r?JVJj17xi0926f!ixbZm_nAc{~X= zqc4yD*aZ*!ut7j|d{!sI9EG}BrkAr5{K&R$jyv63_HC#)_U>5|fR0YYnF-INC7K%Q zd*zv$id@wb$mW`e|CJxgFFNW+VoM}9(kscMHTJ(oT2Qe>cPmk_q8&D;gNJJ(A>T2a ze2BJzg)1ZrH~QJM^G}h!4EsuV0klMaeh2s$(ExXeJUA`3;BBndF2iG^i?lFYPgrR{ zmh`NYOBfOEB(k6W4X+E10TITzLAGs4 z+qJYFa({POF6i+F5qSU^7JH3cn=SFb%G#_on3FrZQ6kH2nNnWK&8gy(bIKjvJ6QZ* zq8u0OQ*#O^jabv%2n~;Bce4NMW6O~#o2=Be-J4`nCenP1hH`cHN<+H#u%f#l&zWCR zbIe6aeJuB9|K*eGxZObZA_t7EBJE>F%nbutv;VQ(P6p`uwCxjvn*MXBCJ zG2!L7?gY6T6`FVQVzXSMoRi(+CcxsemIH3~WfYm=8~ox!z%Vx%|S?FHD=-2^!` zYXPHp6a7;l%l%uGN;iXE(G$??)*8EvwZ4`HGqztd1PjQXbCoV?M!1FM-v@+Mo&`E?UWS|cy;?Ds5Pit{rOojj6_m*IVO zQsrS*NB$2rw{-=rav!UhK6_ojnZ~V)zuDCw%U1L(+l`#u&;uL&v%sz!YLwfp2fx`n zy8@z_$uF4f7s9C(Gm9_s9|JZTJ+jTmeHKu%N5mkZW^y10(%CjnH$kpqv*dg6kVn(PQM_-0sc4iSl0yyl zFYAF?FSBZXde?OaGKTWI&d#ArVJH)nMUvmL9dau}=GFIU<}E*Lm2-Lm)aW1{k}`Q~ z;aFzf7@2VyBy*4kido8RDRU+W?t^O_k+KW&33i4K@}>Kf&= zq|yhz(y6}KN5-+Ey-5CGwv*qBl}PoR1y@>+C&QmAzV2Y^G?eJwc!+sU`fK3WX2oQ!g3o2cu`)P+8EQqytWz_8>r>=T_Zm$ zXkU|tEo0aws60st_Mo{o4V&=IkmG^;Vjk%Zv9%P){^&kTuY8D{IV5ro@<$RW=T%;x zP0HmhwUOFo)OnRRxLs%Xm)dI&n%yCC(GWC?4Vp`@>d(tS7ZIh;2O7HQ$iPVPBs zkY}ah@^VEd1)h`0?-ErFo@;GYxbw)M}g*xJKOe_T{>Rro6V1^0856^g!Ik zQ9Ee9mN9e&Rw8Y4*avBwPXH0FQKpuynlwZ$ZVFzKo*8MBD>4=Ty&^z2z6>V1@3EB9(SP3vm*2U z#ft3w+g*#i4RgSbyw_cJ`^sBTaNNQJ>Zc!n33YM%5~lG;#!~{wft=A(AVnaVR2t=D zD*gZ4!s)b7qJ?bE3wOgls?L9jGsK!b^y5}p>W8+Y6UNW-mNCSSF`y*-xgo78=b$X@ za|g=hJy4{5RbH;yxggOSjs7_95FZs;aB; zAJ4t(u((8}sL?oq&<3?Fw%9{*PLk8Nv|3v0+m^nzZ-*<~3tX!Z0#xE-c&Ai&c7}-a88)D)Ya^qltM{C%K8Du3QU^>H9=q* z_Q8tO$>sXH)l~N&MA)?)j0DxSR(fHo&IKuzjcv%*KOnk-mgvi8BFICCwT;7Tu^7C( zFM(@iRZVpdy-u=5`yIW-B6LL zR8>`#GhMiQ%fbusgmmh98?UDsOf^_v?7Dy2!Ok4pA>~g=Q&mpV)Sg5;VG@HnY4i&k zONVJzS*=N9$0m(!EH^f)rrHRyH%tSERE)z}UN5cC?Mj*V$H@ z`NKAY7~QF!PBd$z$*1wQHdM=%ht)9YHhR1}M3g&wBh_dF{KZ#qyb#`pE@AWm4RKt} zGq=(BRT`Ivsd^#Z`szWRuT%xvW1bR7pGc9tHB>P%RZ+@*JV55462dd|)5YC*Gs=a|ppbT| zSsVcd?~3_TqT{7Q2P6$%L0J^ia$%-`T3w)5rWa#}#`1El-hWdo%2-XsNDa$Z$K6%S zU8x!%LzaiE2B|kbZ=M&!*sR&w84qaCHgPR^JcUghCyvdsZRO#V+U)zz{+zer_3V$0TQ5J(w#VyJDl6fi&QQ&ey^pO7Uxue1 z49+7hL)5|8?UN2d^KPd$fsw_tq#L93%I}c0WkJ4`#fJ7J z(Jh6Kv5HPd)+EJd8O?ohYb(md=6$|cMQf9HN?JSf z^TjG)8?kXCa1tB;w^{Bo(+!cqi;SsBgM#|&*#xzt=1X_xv7VM@MwYwHy^cAlIpzO5dU zGbCt|qii!sq6=5-mH@PvHAL`{g%p!|!c!k})k9T*GgNLdVFNok=JTQQQ!^lNNc1PJ zGgN*=6eOi3U}dzr^~UDH9SYiJ@xv(jVc}()eL12S4`%u9F&x$MtkiUwf?@XnQI{Rm zBPOs(4mbGYY9sfEtI4N;?~IYBt4itv!u1^%ri07RcZSOK9c;$V7U?)g(Y38k-1h?~H-Qk&nCg(W>>z0+*XQEM@S+2$10}K+T&-zR8AV!lwZ$ESf%b#1VncJO`GZb$n`r5iy@OiET2|FE<4Dv!$yDoZ! z=h&9jU}BeqDU9opFNdo9#)iJPMJU;`iz$Jt|Awl}G0r3%TVy9wCKa^RGRJ>#Ei`HZ z5eMlX%1)b|p%{wBXvPjcQ2CDr@Q&6pL*-oyzNR8lW{3(Mr;S^oPP>{o>%<{mjt6sq zw!vwU*GzLcp5b1_5x&fz2IzL7`gv!taQ*X8l?6`8854sCel7(+Do;qjZ42dqj|CqlJy-w^-Y(HO1z?qL;2RL~)urBr1vFDq#$uhM~o-g=h^U_+Z z*k|)~Nh=TbtI=F6f=({Kzmm8PLc(x{BF-SbLAXYAldLOGb7)b%bD%sw)d>Gm4Kzb0 z$qnK2Wjs$rGJ#CFabsFheWD~;crq?YS-oj^0ud>z>6yv)?3ekIW!GBc29+**5@^!wd$aYae5YF(8NgL zCGwMQ!i9(pq!^oMG{e1nC(vHdzoxpA-4i5HDqAH?uzkbK;SwDsVtlZDL>dR2nRVyJxO!W(ONLPs!dK_T8h3*EyjY( z{I4h7J`9k~FnJqY6J(s;UPS)jOtOgxU7;!ju{AY1LuF>W>5#_XpKEjFqHuA)*j7Z@ zf88KDixa3JM#*6XC14jlP+qgo=U^DWG+n|+c<&UWjyudvV2wW0=E~JH%ICYM6rG{O z&DlW(RHURqpq%S)-e=tj6gsD0z^n69gy_L_7?G@Q>n5!?@jT2wKRr<8trq(-RPOH| z7>;CBThYz}=Sj$A2H5)e^37h2*x84+b`89#@1jyRhBBjp|jJ7Cy zVnhfP7Y^VGF?(cYG-_Z)cx=ar!h68*GO$8VeiN>zY~tpt8W|72TJ9fE1HZtvkY|E2 z;aQeMs-K*cl@(?-=!73*AM(^|VzL-fv$GclKWC^sXp779<@jVis`e{)_u7D(IgLz5 zMz QSpp+DsD;v4gLGD6aY_NPDrGZvEPZi zm-ESH+(r6KkQE+ahKazp!Ga&&74`0kv>RGGhpL>KS?aXN`^USXEsthOn$ygn3ADnN z5OOm*=M7xN#$j}c86_lU-pix}Et}i~t|ue7oV`?@I3APlP?g2iC1)>!+CpLF#_B$8 zU_oR!i)8k8rae7jvvZK#Kizipu4!;sCWSpS4_ugnhe{6-8qOGblzm#|Mjj2;CLJYO zkk%}W&eW1M&OvhfbXZdlP(0`a&KS8Pthe0iArE<)&pQXnU0+x*Du=ww4a8$Ku7AH> zKK!v^aHq8(;|zl1bxbeYCJ}ZcwwjNbIuquwE^2x)dl%oF+K`4 z)JyPr(l#~px-Q(puF?WeYo`)EI8?421sgY*hP-$M%ap1sr$*D;6Lz{HYKjQZp(;06 zQ)5(~G>_R(m7BtH4{u#=Ddl20VyvHx{4(MZyDYfSToaa15EhSN1_H(6LyVaRcZpF~ z)fhKWEstzd(*<+VXUe|y(ed{Y@#)5fkfwoj8P;iMgvvwHO<=v*!5%w89x~bhp|Eod zUt{F9xaL`&J0m>6DEKVT4izyFw(43=NY0o^P_CV%c!=5>`Q4`vL(gEx~H@!ckl4>bz`=ElQTjdo9Z@W3=5d_zcjKSE{s6rA?&4+?1Xi4!bm>CLv-zM*pc@mb2;G|g3?sju=W~+BcJzon_0ZhnF#@xkSUC@%>~5 z@5Gp(oyMyn()`Srnl_dTnN`(nhzR~kiIG1@nzqyg)p+KwQF zV4+?P{oFTbFj;(2 zDa+G3t3A%sIzm256`U41xjjM1J%$iWAvPId+VW?&3BmzheVm)#CzRl;eiRMxfRK-o z7oe6vZSpLi#>EgSY=j3W7991E{=l_-TGdB7RAQr#%F2EgmA*job90g|A;CzB>qdn8<*k?{ zN^mYp$zAPPs@;|4`pco*idYgG^4+w)PJ(Zg~fEp=MxMoJ> zSBNZi@|S+=DRu5N*Qt|7nN?=VsgoP5l= zPKja{P6xgRVv73GuwF1dgmqdH2tM+sF_6`AP3S8@6;xk4wUH%KR$7vha&y`on#TKl zi2xEp!dDl<#ay)49UjGU}U1-wSAlBwM))d32;Y(qJ8 z4rds0B#V)+Q}K3Otq2m}^>fCD15_TF>V_f!^>dsBT;1k6!{nUGlZF$_*)`D_CKt2P zs5hsTmOn;dP|vwOCI6VhtRKJ~JhEPM4&Ib^%yV;qoFYpp@@}gwuCt`izHVx@**r}G zpRaF+g*79mNhx%uk)8ut6_P}bwSYXV5yMnYvy+1dsJvy)Qu)1Y*3<0wpX~P(``ux` z@2*-3!8m876ppAl=Kw4+u?*b%iwN^Slwmd1WZTklPOY4sB9icjDt~7&sSIAv=HRnY z;>NQFr$u(e0zKMwekj|Oa9P`w>oa%}ea&TN`|NOh=h>bVGH0DHJeT4WXDw4Ntn?Pk+mZvWDbQHWCe1v*GudNrqv7uRe z(7k)Y#Jg6yOYZDQJl~UllMJBJ6O|tP!K^qCW>dMyT}bQnXjWjx15`eJ`08az8N;OI zIpWAk^N69Y_-)e3V=z=&2$6$!<%7>y0r_20@>%)764_l})*VQ0ojj1CtAV>_}RpXKxv*D{+(xT0oGD%2tgiCI06JlM$$3$@2fiyf(m>&JQ{EneD68W9hD$jwF}iOtg1bm&G!YpA(f&?KyGKgb1gE!h8WOC|05QTj5>#RzVtKr7G&H$B< ze&ez+F5Ej%8HXigfXd~+wRbaZYmxPP+ih&~ehn9|Qx2Kr$;Qc0fqb#%WUjYfyT+R} zl4P7ewu~VeNd>l4K_Z?|h>e>3VC>`nKy$WK$=SGpSP}pnfSz(1-V80vI<|!G$!-TUjj~f9f6RgfIX+9#~U!;+%I$VbHOqC@D zQMrz1v$uM3Q$wKn2KD6H16_Hx0U$|gbD(hrqN8#e-dkj`LngV{FXS@xwD2UBjN1}Z z`q)XyTs>%{8D3lNHzpi-e!0XFND*F$C_{)Mh{M#u-}aL*j**3=-8Q!?UG#B98{2Q4%5w%(Q`gLR zAXUFFw{*MIIAi4OR7rixzT0kU^C@`9oH{bQASI5GQi_XSa0S;9dCNxpMA;Qwr_yVw z5p+ChgqSq24bh;bj8W;ULMg3NDTeOgPb%EOb!3!;56-lT_2V$U^K48qv>;(diG0B= za<<`iG?Hd7S!y(oV>W$KN@k}vJ9VU1in(;};pL&efVVspW7*Csf4IZ-7_-5=9Rqq* z=oDWY0t%3j_yD!UGIbneHx5wKhbb)mLwTf%u5YyAW#o}6#HTtImo;1I{;nenwrgwC z3_7O#eMv{~1+8uobcFF9rO}u78LZ>5*}t)pDa#qg~GaD%Y|U6culEIytDbprYih zgxu#emv%y3dwIpT$E=lF;?&JzaqO5>gr@KWYf9<3^veAUZEe`@Zo zE?5O=Bjh8%&_%wpx6C)SeG@JC%@B|&KIS0w<((?z9qh<#)icZlE+=4u%gBKBO0vfj z`IM1UJ21hniZf%>$?a+q&SMyzojO^ffm5sU$<$~!!m)zxYC!?@FfxmFNaEHg)rHFZRmbYx@J(R)HNUGFhnGWTm7m8w;_^Vg(C zsZ+Vb5?tIfEduKfBN=F5k`_%GL&#kNn4DCPxyy-3yK{&?n5#Kml4qY|6F^OUpsU7QMWTBNg#uQ~mhUD}v{jyM4>$PW{cndBTSA51|S@{I`KJ~FdE{<}ri zgyH4v{@}$FC~j55wX+t3Sa>;w_b+Dd#8r|@0y+jwOGfsQ!CO)iLWFDx8(KJxRKyy29Ge%+_DdM9u)obqZDyvWElF3^+nV6+JmMF$3rah8YFWlRFPxwnqG&u{ z1~B6u5njlop-u&pcyEeTu}Qwu0!z?7h_wcK&zD~oV+x<&mh`#W`ovWLCq~dl{>M~a zD>K7Y$QicJbh9&9?mChDEjOBruuf&36TmtDF&(~Rk_d0-I05wRX@*6RF$=3)+HTI| zRq!{t4$GvJ6t?@qmv8rtsNL4&WYWl!64bm0 zYS6dmpofCu(By3hncsi2Ge$N}hrS?Mr{D~hOTmwZK2xVoE(KAxlNiLP3!WU87%a!0 z9CuYMrw^!Mb!P^#><0Bg0uP46=;fG7bS$aIspYlp8R=niOMlC%V+Cb-{~+}HQ};sZ z%8MZt^s|7JoRW}DLKNL*-KN2yMwgpV#XQ@<0xJC%2usd4sZ z9YS1aFt^1T%vtss#TOlsJps(YneY&MQF>V486e%Z>p++w>Zz)Lvo{x|5IK$h!hv4_}^to^Ff_JQ@~o0vTIWHWCVwlX|vO8)8-7vMI1RIUO{I zFBbZ7V8HxYF%8I>QTz-C63X;(!XG85SqrP z+!@YU)>IH51zBEh;!>K3)T?FFBFX3Beh~3T#~Xt*$G=l24;v%Bj}6)C`=Gf=`RgbjVkRTKQ$ra~Hii)@mk{IO)X7FD$r&S;Ox+SaG4-9!7`YVaxD+}jb^ zRqo}E_XcfnT)JmU9?&vE6#ctyN4fKC5$oAUp->&jb*x!0o#7ZG<+QI2yH6=fTf z1kXMWhVaHVwlvQ=j)2NK9*rMgdAx0T2JF0TERG$f)2lgl`o80e$xt@fP_7JzGHaVa zew#(K*FJOzU%duD?SC4cm+_2gh3scm*rCc6J3Pqsv98Kz$D0RhW_2LD1@gPB$Wm=` zT-ieNGRy8hKl#hyK-sEcDqCWyeAc$4KWL5sNQCuy7ANFu4yzp2bLJhBUG~|_3iDdvogim{@L74t>POWC-`r|Mjg%+p9nANwl z)18MnE%M;8$@Dz5ZOLho2Y8`u%O-X_vTfi8#Auu3ffWkU_hgFOfE|vTS=gnH$s%B= zM@Wktze-0nN%eY?W#DBl*s0@Qr$wHzsb92BL4Jx1>&et4US`4R3GKd5qP2P0QB?kV zVwU|x<-&}~lrNrWk_ks|PqmD9<0FCnmTdwskk322%MJL%@JH)bhp6jR)>ji&cr{~W zlc|%}wk@%VH-=%}IVwr_aUl&_=uj@g+s17I->H>r?VD34o!c^9jZTZKv(}g{d1Kp> z9aL+CU7rE%sJvxsDmqNkz=u=A5AWIsv=^q0easF*e#-qdPMyr$E+Bn|%A;nlv%g6< zlrA99H#)BT?bz~a_{TOZ>{AWD?{32xN=m%NCEtwujBQxO@55yN^D0((JcGrk#|XYo z<*5+&H-xy41?-h}l0qhh+Sa6OdUVL*mY)^RWVQTdI}?JMphpj{PPi2F>Jb9!=|lqM z^D%QAeRf<^<_t9BQsXS-8>&V>jXtu-8f`MyXkNYO5BUECUk){3k{x#BIZ1;{X^{IY zVJG>3@9fPr%V^DYDrebxr6W`^=8T`tW@oTmd`xl- zrcPy*rD2I6wgtHMPsZg`Zr#qLp8Nk~?Dn?pXbB_3Y))>p$f{#ljYdlE4&+cxt)}Z_ z=`n0cj}5e{OSoJ+et7nnkoq&SXkZ^(^EH1m5oyDAEagsX4Puk7ub>fj!CmgNw?u-< z`9zij@ypvm+$yIXLlCZcym|fUz&<9-sg=t^qEy+qT_JSydA^J>WXtbeT%Y$^v)0*T zHh$}DMts|Soyw&H+3a!W7E~tMGFBN320S@B$;`CrykyE={0irsODH7oLda&3?%Kxq zz=u$}${b_jjm{W(pAvRFDT`@pwnfw!I4f=JxI{2GGH3j(##peLy&|Nn&$KyIk-6tT z?SE!lMqYC>=b9%4%!RkMk7o71Y!|UZPHZ>5jw>^dDetoD1{Tz1Te+UeQh91%AgN%y zbY$%wG*8noz2$LDPN`~ zr%o1GJP_t`IWe0d_~mF@M1jPf2xm~(rlX-(!}29|%{lsO z4suvHnJ=SJE^p_A5^z5gPTh6Kg>xV2>8rL0I3pBWpe?7lnh7@&nZ{m{P{adq_5_$f z`nw+iiqj&`mcdwNV01C3bDS1AJqDvrxVn9MQQ8wv{@4XcyKm5Lbb^oUI%8DcHYC=m zoE_d8{yMjY^L);>p_w5MLLiIl%Dum0y~`ejF3o-Bt51>4bL!d23|*sv);nWl$+6LH z8KdBlZ37@i=U~H@dky7xe#I92-8`t=W#&&iJ(9_XrsmuDB6oi!fYeU4hmXI)HgM)- z*Qu2W%wV|bzxb8FS5Kz9*aQia&v5)kxE5HB$g1qZUdF2E0iBQ7fQ-&R8B~&ABq3*m zE>AFamR(yO`q_GC#Z#3}3RKH;+Z*eb{(a8=^2yNv1JTVK1UvUc1K-(8*09BBk5ebt zWb9UBam+f-QuNHrW8sV1;x-;NJ!@GtD(pXC4(Y z$n-FIYkO3cm@CuB(M2FqIb~4b)XHYNXZ=5{Kc`OaCu$e^iU&RR$^;vo3!ppBeZ%YkQ!1?`W(h|9QUF4K#dU#s{3oj)a}v zvIBqscR+~0JbmQn*2!f%5Sr|?$kTsxW3A=39ZM{VGvaH=>r|H5ZR2P;Y}xN)cQbFD?LnZ+kSJCv$vcLGDF(juzf(- z1{YZEEmv**ev~nc*bOwcyA0aK8e855)LmQm{lE@=T*_6eAlu}$XzfRWc3 zLag5py}sP`do?xGUU;M_Wl!x8fKq;Bk`d=)Tle?5WNuYX8rW8qGSRqTp+D!dg| zIAcfnzMp3UI=6+w#!Gv5La3e(+1kR5_P;j=1#^5{>YxAc&Y%Db6r17iv$0ub^uicT zvX&7c(imz`_k4ip4uN>upk^6Tq_@kgSu!x_%6Dp3<-8EFukILd$a*Hts(i9j z36LJ-NtW$o4E5}=;2X?<5tGW4#B(@#Q8J_+hSa5vaAovB8~|8zTjaGP+z2eWYG<4S z4CHEqiUC&0(jW9MsLUZrJ5n1@Q37!k9A>%f$4^x4$%_*Z9@x2teKz)U&nSOc2XHFe z*?M{F$Vdx6Z3iQVO-Gst$26W6u1=>#Zb^VvSpzN`0usJBW8`jlu1aZ#z?T$v?S2dU zb@{_9`v!7rAuA8;j+lbft@3K2D8J(VXi$G<^rYnYvTt)CDNuev3kT8k<6kuPrRCXnQ~Dl3m5pW>383!%Mj?Gcq5 zB-@2tw6h5>ml9F#{6ysrhsc&%mDN+BH&Q2f&QIhMvdZ!jxepoom!>8{<(b~w}e2v@Jwp5wI0%EO~Iw{?Wp!kwc- z>FJ-6LKtHaJ=(ta^5MqZE~EXG z#&j;-IUr%Hn(6L|@ZI8_gFM;05(axntPCtKxAzy?b86)RXu#G0I1QR$xK6FCI4mU8 z^e|bx%Vrc>2;}C&A}XwtHM?K|x+ta6dssAKD|P`{TTR#|t1y~9?&)1hJ|~}%7{b7! z#1K|>XEMx_cV#1;8{L!=UVPc3i1_5^Kw>ma!G3vQ!*mL4yD2jS^J+;x3?x~|q{ve{ z2hQGd`5#f2q}(0kaK}Ko`S6GgX4fLqpJWU3Kg;xY z)_?~b;WD8Zi$)voO2*i|t0gnEL;)?;j`q+rF!q#z z-w2Ee@Ne1&BfQmokVmTm$CEb?D>}Oq(O7h96XX1SyO5ywk`@rTpf#j>ds!>@{yuuV zeOKlbb+t=#@?V9hty-ffM2cM^e!lt8?vVe_+6|i7d}!3btlfYaRnnJ-Il={#CM3a1 zr-w;*|IJRVyc@go9^R$oH>;dQz}{O`T+k=GaUu7Izc)_PZG)QNMsU>08M_23M>7YW z%E?B!mD|*UfoxMJ*>JYnrp_P8HkFaT5;YZuH^`jBAhGUUOY+}Y2?E*QRZ`Hudnm7$ z*g!I<6&coghOno54~>Xu>+LDT8n@1!l+1_9>ARLxZU_U!*|uv{p8kD=wQj@n1o_~P zh(D~3X9y@_gZw8HoxdwW@(0#6Gaa>j3oV;FIH88OlQObpw~~7DQB_GgJJ1?x<;@^`cLeVeQoIC=gQXdVhB59LnYlX(^ctsD-YP^@&)~v) z1a(%vC*KTIe&yonTUszKCTH#%$c8jTd24}*yoiDM7_-=PV6{1&K!vNr3My-NFF3XG zZ2WlL?k2-LLv5Ar-AgLdvnrqN7RbCq-6oY~B!GV~9W**Eazz<~Z_tq~*B9(1b0ser z_L}9@-2yr9P!~TPH(@+sSLOYCOx|l#cnO_tNS9;<(I6@(b`$JkL0pd=R%Y+>cY(dX z3q>p3)&GJZu9dc22zN9-zvc)K_8|sp(q=Rga=^JHwgcK*axC0m1h)rUv%&CkCu)S=XdfJAuFdW9pT>_;Si&;0u?znPF zJc##f5NCbaL0q>JoK246K#EMab8%;-*ULw{eG$yF4>g!CqVfL-^XlC|eK%Y~<>}qN z0Oq;B2h(*bAM6TUo_-jQUjp$r*uE$ytO9|Te=#t1Dj)3%$GqxL*V$c$!D;H$*jE^B zc0;^!Ry%D15tS>^*BQc6@G8C=+vM72J$yFw-(#E_tXtUEIJ!RH) z_U@&;T>mA2ehHvemUOH#J41_nQf}lSmZ8f1Rcs)h?92WZ>U;8iIEp8~w*Q&qA!oNxRyDW--L;V1r?Z5X)Vl zsh4b7ds=dS4!ohf*T5$(S@CLvc4h6``X}VVB;#!8B}s*p{U*> zl~FK^j2y%c-ATu<*lxt&ql2}V6|zpPEMioVTzmQML9_gI_kvu^ZW4_UV+zM-q$ZG? zcP?qfo8`eHT)AoIf;RH!D)lVI1SVkFD~K+_TdSUIzh32-tSm(Y)n;`O82_~l`O0nq z1~mDdD|LbVid&WCf3iwuk6U&x$lnQOZwLqR{1L8P688W65#8Eop{hAF&Hr&p3I4k* zx5Ub?Kmg~IZJr282su-G&CY?`ctlY)>|T;- z?n3qCl0uT(V&mFm?&c0gCF(T70&`N@IZ$8TJ;JrWptpKZ_1+QPnp633XFCcmMKBki z*MEIyprR;%@guX2Ko$SglJ#adUZV zCqU`^gPZ>E&=jy>eFDlNKp`+7=bWH&<&hBQI=d|L1eL2Y0TFgbRyE!Rs!;+ z+ph<$gOD|!nV4Yf*1@GbI%kgCUeX2U4=PVi>}EAC8XRCaX~DP`CK9QWWrM+?w01i! zaz+`8i^JD{WY@i3JsNp?BD47KS!M6w%?+evY>_?#$foImu&q}1aSz>SX1Sl`gG*g3 z)c@&@Edd*!Pq6bA*l}uQ-NeWmbk*Ph&iS0z>R@McK`K8y>i0M~$!O*smDnY=M+;k+R=)jjg; zE+x6nxVx+J!Nl^O@ai7y2^UN<@#>{L0vyqn7)@R2UxF?CHT-tz9xzJQWe(GKqk5om zfUM@Ttmb}?gW!dxd#>fB%GIIpwkCYKW{)5|>cdbTnEHkf=apx*dkES`uSiMW zsI@lkrwx@y_J9l9R&>qbt}lJHfiWH5mroCK<;7ZrOlN>R5*EJ+rtJ)nhr`EPMwKvm zsJ&ujvUrccZV!B9LIV8#LrQY`{{i?nTg=heqB46Tcaw~Ud)&?dma05bTacxPn|;O; zqT!^cBRsm2Yp`j+C^Vj+MgymFkHD#w8HWPmm)N|Q>{^KZMO7~)Wb=^`UZdSD7WzYB z7@a7ctOlRJM>)ua4HeYW3}&u5nnTrkaRL((+GI8jj{L5(%ea}dC;VU*FUsMF*Y1gb zU#7XrVPMP?**NwMd~M|U$F9mqqr-II%g>6ojLoR-MgJu>z6Gx&)3U~Hh%jea zDd|IQ*^@a)4)j;Onv{Q@@{TL3Ct&`+b_m?aRarEl%mKQGzyVfIFb=SA2*yH=Axv|1 zLbFEEe4dY^US;7#h^kKIslm|Ljkf9h2O6`;rcDUE&H#CFqIGl2U``o789tmfm=t#d zRBoE+s%&S8fNmZfSnvvj5atws?D**->&^tnmGdjcaJP-&?r;ou4*3FgX+aG6MuYt7 zF9SIPJ=f$nugUC@k)?d{YYBdfLkaJ9W$^@}u~mi^sPx2;QiNq~;`z}bfy#=*vntCl zl0IN^?4~KK@%x5YFu`ph==U>0;}Bk(9akO&NK>M2+#|rKdT;zffcU`EUyaSRP)NKZRXID`OjWJp7?)WTThI{y)F+N ziNN%*wLTggeX=KO*>Cbyp7?!HIiTv%~f`)4p+97;U~u?M`6UgBK>8C{N)r%G=_+Vb356(Bec8h7yYRFnB0tC4kvl{Za(hwMXDP!(C#uih1&dF5SA1?^ z_6L>j-@7W0>>9|QD-iHyX7%B4R0i$eky-xDmT(37bNrriDG|Z24<&g20ZHyN&MtO11IAr|Oe8jtpd&0YCep#q*-6QCdX-RP9 zuG$ns7CVL7Os0wj`Cwp?Js9GW^n>{4Pmex#Im8|JCyLuo|LF zWI&(qH-bCAG0H6%&&cQx%B2^4wZ?Jf>>P-eT-lKJ#!wdjxj9?>9}W%Vcvq?ma(p&# z^AQFnf~qEd0V)6UIxtYaJ9<(TUcl$<*09!C5X z6M}AzYT@Z3=83j*Ob0Bpc@&S)4>R%SiY9LVf+o7+Afp8z#e}!P^ns$AWolM^t+$9KS3Vmbxv)Uw_I(xkgXbCJi~xDTWJC9BNC)V+@_|7x*E;yrX75W(*WE)f#@ty($n39` zGzBcHsNED{w^tuQwcG zlX~^n0_N1y^smI(?6}fvLtODyNaki6-j|rsIYS|H-`L6yirv!FnVc~xMyCOz{rSdm z5xyR^_HX~P?VmgpRB+{Zk`<>{5>&hsp#lb&pA{R5tDLw%waggK^#~}+3)R?h2E&=B zYvb9UkrbBB;o%H$f*27C3u#9}oEysp&#;2V5YEU(453awtt+t$y`05O5Qb)PBJFDh zjz^k-<$^=X{Ne061ZeofW7dT6hmY#8s{yUfVSAmD^xoSq9+O;cz0YAPKRpEN`&sSm zwJu(VXR0O-PL;{f*w@#g0 zTxUG@jcggG_typNQKS%*DXyHmUGV>){VYZ8jVue1_Vd=Cit0%}yPo;n@@!o*Jee&M z%ft25==*YTF*e^U$RzXj%wCY!9Bn{ELho>f%S(tNa^;?ZGhCj_7aI_6<-9%DD?6<% zD7IU+VVRoP9`XZ{HrHC4A)D-)71MNYxxn8j(1mc7!GdsNviS_EP8xZGGEpNo5TMI( zwAHuk_97NEqg-P%HGs@=v^ZnrFGeAFslIA4tLsU?$6!=0hX_j)(=Efdg0p|8qx47<2=quoy!V4+O9oJ3!V%5Iya; zEDodb8GCQ$uu*qNk57L!2vZplUR9fsYk#+e0jwDw#TQ;}5IsLUaCVk+!%+BBPjt=0 zbW}E!pw?)Qc4`D$P`(d4Yz3E6aXA5QOA z654xp@0#i!1k7IbW*fGEn~|YVYi5voZj-N>wPBi0#pgtcPG7yx{0TbVOi|&KGg6Dd z=CnR}3M3Ukq?t9A-n&yyNHTeD+&+PzYae2DAIlR$6(=Z>hxP(ZOx^^JjAx?+B8Kh8 zCY22X1C`5;arpy&aYksDELpX@tREQ2RmWf*ZX`31a9I}8x4by8AQv6u%8LVyT;#%I zcn#n(e_$Y+j&W5^8wl5YI&AU!(C)b^e7Yerdp{LEE*uq%l6R?d%KxQ~$_K}|Itusq zuM?(w<~}epSuwE0UJi<7p&c}rCo5j{9D^O1lZvj_`hN4mJmzR+*tfrzEd*tb&TA6ojDHfMT~UN&Pjol zNr8}r8~qB2BVR7}CgtYRUn3q@BPL%?9T;H)x(OSQ+qVnkzCXDtw{7=-BBKYWv+Vy; zN9B<}g=BP1nT%cy$>@{q%4Bph=do3`RJ6m zjz9UPf7k{k`ED5+7|kvAWpHwwYxFL0C3*Yn=B3MNE=ia;qx|)hZ`4$GrbmSe!1D6@ zQ!C$B?V7Z9TIKuE<@fU|-_Iz&KfUt(^78vLD&JS_mb8D~H^?ATE1Q261rE;s2Hq5# zR1WOQSD|5K!Zir5wr6H!^Z#bUJ+}09ie_Mr;cmGuJXdb(c%phTx+vcZjIFL_Pmnu0 zT)C%;Hq*mDgwja@u5&auL5;_u>5=4gl2a=e{VLo+<%MsUD;@~gf-+R^YF4sUNW1-R zE3z?m(mCC9sT^OF!;7K(o{|0qnc}LnxkzYp)tlKQHA|({m6|}ZNV0eTiao0L8wJkD znPW|ka~rf%{yTBWR}D#!cJnt8sIY}d2fWrsOH6rv#-B6F*IweN|JuJG7e_~S;A|iY zN9mgW_-8zoo2IN%d8B{AOpYoyPI2X-{sr}YE?bscrWEA~D=LeCoxd}(c*+))$NQIT zC~_$wo$AT-BsJB>Jrf}R0+4fLMt}4qIoYm=VIEn_@@REIZaO|I9};N)xcr+1ixuA7 zxGP~(vTkHSWsXG_k;6uok*P%GAOFX8)EdgE%$v$2xolQ2QKU~Fc53%Ak`55Lsv*=& z2oVWw#nNxRdd@y7&;K$ak;}db>-8g_GD5D5PL}k@uSQteM_wkkIgHVlH0QU~^SjV? zB9(kWT(VA18o?xY=9x6-7b-8tU7i0;yWivGU%KqkO3=f(-z;$Dz2zI@=FW?M>4w&t zI(g@tP-~7rg7x;+SkB+RSt7znbE9J3JmYH-jCFFu2yXw+t2Y~2=`X)58?5`jfpAuY zSXE|yGmz8FhJ%D=!jE*RHQiHIE$T}3vZx1SmmyR3H7d%e_#@x2F8VPAf3BF6d%ww>yMAf-dG4DWBqABpz6G4uWIJt83OvhHU`LkgjA@d$ z>4JE+Ubfk~j3Qq`4!(b%k`1w-VzF%Dl?=XmGLX0ok9lJQQr}1C?y3NGl>yrWu(I-- zu~F%?7>;)gj&r^o94g)W1UgYJ>&VK6s?g8cHI7?73up_b2*K!5>tXNXofer>CY9Oy zLMpKI4BQdP?tc3^+VFVtz9nvgLPYb3WgKKt`7{SrburWuX@?rg%w_X(=~S{SyU7_3 zd96B-1;@K`tlKC91C9jAdHn+U+qA6wxu1zpY&P=pv?2oC+;CBS@Qr|{k3f)0*EfSE zd4Wqna&*mzJh|XFS6(8&(k?-WepX3+V|v7=jdJ1fuDlra`^p2{=RBSK{k`H_pJ7Nxz|(7+ybTv3iFN{1`Q701i8tjdwdFqS)Gmr#xv zyKllf?%6l=LBDq}KK@fj5KaVRPUJk>84esg||-*63XBOy=}!F0#^bq#)-EkmbKz1*duFn*caKZU*J{ zWZTVyJgO|?t0p(VEKeGjR~RWaVPqA45l*G)VP<1ReXoyM>*R085Xt*ga$LFK;If5s zxgp~oL&lvUGMH({mCIs;JZ}iO_RAvVY+K+`PC_zzWCB4S5a6icW;5a>pJR~@P&wz9 z<{Nv_2AqM^(v3EjD4whA*W52@Z2q^{qvqD*fj{ei5;nZW)|Q=-!Or)jePLb6`}>+7 zAUYDhvE0&yw1h-sQ9BnO97%((jUd5e`0|v4x4^zs-i6mTTlS4+_V6dCcWX|1or3)( zi!6I)6A@{A-PSf{r$tULgR$gWEPjvLEDhO}M>(u}qAxE?APZWvvXQthREwO-#jQng zAjt=R5pMVCVY2L7*bN))yxxt>oywylg1r2sC?^!rd+)9yvgngvfE642L*Jqo&R;t% zQY-^=cl3Tx5ZQL}^PCoWwOnySSio1AJ-sNsPGCuS)5mb?>zZ;q7ly@TVm2YX`edf= z0l(UXtd~Hd{E3Kg zr$rtvxBk$#EtleQ#xM7T_`GAR>0UgV809rL;2^hA@2*@eQgI~js&BEyWCn2-*Gk|l zKl|H#W(aK{4^KW%O#D(<8)nPT<8KbgryJ0)j0df%&B15PfoA4tZdPNx z%E#p^GUhiY)V^dS)GksoN9J2UhjxwNuNcYF=~NPs$^n>ta(Q3teWV4rJzrU9w%X9% zrmhl2{u?jH6s0+kyjzb<6Z&wP8c^|z+wL5<}LF^FmzAz<1la_EnaPuyVhD+_+?s|<~@^RZmVJ2_=!lCov&_mVj)?pIUYiOQeBe41BpJh2b1PLkH_JhOz#v|_BE zZ7SD0pL!_WIv4!1orVhs3RH}?+@^8GB-A+Foc*!am~?NQ~^aEj)n zKQOb2+{T=TD@5ft(fr>}oSUT&eC6k>SUPgvNS0iBq`dO;EzygDI3lTE3eK=#L3 zWpci2zY_N)dMcA~EFhU+3Uvm^9dv4p2(=tgyrf4H?dd57LDO%h04sTysgY_J?Q@3X z(eTUH_A8MRIF?UF|2UbcGrq&h#)svzS)FcmY-xYGAA8vnn)UN}yF~37DIig86(wHq zHFW+oPEdK6^OjFnG1xA-p@Zx&Gk?agzZaQ&Z{H72e6RHaWGW|rr$m6V(;_R9xAKqu zN;04%?M2W@{52CT@00e_sZ*Kz9V3=0tKtEzI4~K|Eewd81dQ`-VS%X3f*7>RGBHFl z+UTDoRX&cZAUAaxwU5$KO%oBMJe5afsod0>g&(V|wBXCUuX1T?))^qBsDu4rYC)(e zxL{xZnd>yDJYE%uWf9{Pv8PfOkjAPb>(t5Vw(7YA+T-hNkzh7yme6Ukab|+5pCx#C z7QD>0Y*7z~Al$NFV0mBfF%%H9>jv^;){}VaRSt~D-4Mu+k}-KD`I(_+zx66VkB2=n zke?<+_CvbPJGJsOqYK?|*SmJ^5CUj*=~I`gsECy9O$W_wYll z`zbTTZ3*9DUy$iZm|=cwQsuHytVN0PN}{G8g7uaCG0NtL)PqAIBnM}JJorsrzT`G{31QKeO zJevw#E;e?#u01^I%~0i1V8IXt=ZZ^bX;F;vS<^)+Ud)(p7dFG_(sksc_Efg-YvOKQIC~T&@t_NT1@ta?r7FO={I8F3l;qDSRPkz^ z{J-%KW$Yh%b15)A_D@+Wt{$(y$J z#8luscM7mFJWY>~6SuHB%=+OWMnbmC{W#6pdXZlh zW-~~h<_H=B1=7ZHBf-CCIZkm5oS58JGWxekldF9e64mP=JN^L3oUTFFTJ|I8_aPs> z?4p#)%|FUIb@C$V!R?&0Cp!k?oC8(Pv4U&-09hDw~>F**$ATOtfBUd?c`;W4g zvh_jhvC(k=x=4IPD642s20F62g|*%8b6ag@je3qFxBV#V`~SM^9ZA`VqZ~(;|H#cw zGBZuLkJRB|p9IkziFIc3xCqfXc{PP(+h7;(ALpVq%cYQ@JkOq(D)B)*(!m%pG;(z+ zGs=wMZ%3lGSrsxk%amjX9z8zk<0Nh@oYNpDVIOqmj8q_<^r3VIuB9M-un?J%3M@|@ z^b_~Oyn9?Zk1)YO<;X;55Wq`3m1DCqohT%<83Gw1H~Z*Z0>iawr07lyoHztg78Nk3 z^0#(V49`skPL9pUsjYW%jq+6#gT@Im#dT`s7p{}zcKm5>_|>V^dM78TKwb-L&$rsL zDSSFDe0rzdmGjMW$R);DcfS&RuEB2voGWgU_uF0huqu$9N)6HjB2->9@k6eMWv)s^ z=Nsy^5j3-!SY56*8}nI#lX0tL2st$iATU#$x2Cb4*jYA%@#5f{ zL8bB1U;j{Mwr6?GWV9KsRoYwxcoRv-qlT8`Q$C$cHiQe?Z7Cng3D%6vNcT1Slh9+0v2CO1nEtWdk^o%~gSulQ|vc^6<3Dsg;kBY#pxan%x1{k!kGewJMvdN=}Y= zb{@7+QN=0v@_IX)|I8FEeL|}$MIKb%Ylp{D{9IP8HTt)9YxLAqp%GTCp@Vf`Alrr0 zDp}WTNYTPLBs5YFIy51enSzVlsg>K95J=^|@P^FC;^(EQz_(LyN%ZuLa`n@wjw_R0 zO)gFc@>&~OQARFI_ci&G@aIK>&Y<5XonB3?u4azgEpKyWd6;ki$u?IOr2|>hR+MYP zSC2Cfi_?LuZp+HG%m=q*WM)rNX+p5cVVNLnis$6yubk$~%5~IPk`82Xn=3m7vXp9% z^72N@6a)F@wq6iM*|qX^J1gMK6y4dh{40WYYK7xxVC&3u&>YCcHt6(Vc{Po^*DPn* zi;Qea_t~96o@V5DC@CW~NOPPZz!wJT#Z48S z*30zbIK2TvL2HqYtaKD-sod7~|3Hol(gh9jIv#M~Z+Z1HPs`J7MOl~*y7B}+hDvFt zdYRedpPemaKLv{Ln6sh9K*`8mZLZwUs5{(Q>Tx4rWIGu8$B3S~QnDvIGnQ4%sK*na zN5l{-I)_GvyMweq#-@0fWLr>4lY@$RgW%f*=X)~N zwY~^?1ga5Bofa-t7;T6PpigN{oJgB0`g^6^-<=(9DAabPWM#R-MvmZ4kH|CW6C)`t!s0&-l>(V(*chg+Q?_X&}%i_ zD05_U=y~^?{p~&zqzv{ur&iX4t$0*}X>;rT@kQW@7mPBo3nKZh_LrGQEDZK@zu22A-jyE{N7?de2^d|>x}-`Bgs!4Yh4NZ@DJia5AEUIX zXhfAm4pS+$x$+U*y0gvI3|N2-nK?jVx?k`#9Ec|wOU;=o5MJOw$qe$HUFSge6KQU7 zGk=P(`%Aji=nS_A0)pnpsJuFD6~Kn$x51H%2jQnReCn|O3@KmmH~Baz%#nx)OmwK8Z5PVBjAY-71oHIs#2w3qwACV-- zm|0NxG{(qWvG9SMb~Ygb6Ty%HMQCImq;ZJ+wivV7>JieF z(@q+}{e*Iy&y4Sidt8I5(HTR?xz1@~*+vG?s*{s;S6aJvx^51-s}ROPEIqT_X&aYu ztmC>H5lRQFm!n7z!L&tRk9J8H(WqQ2&PY17A=4!-S(Tjvt+#fVYUU$Va86I5uyN;U zRS-VdJ?W@#wr0~K<cy`F0LDn zdH3k}TQMajjF1Wc6igw`V3h}^8}h>I32Wu~>F6ZIlxAe-K8QlLTaSe)n)OW;Z9M{K zLcercBtEQDOobUt(a(^_w~IEpVoLk~CCfmmF1D&>0Q3n3BH$1iPfgMj71S>Wik}S2%9MSGaCaYKti{dN?g*k)?>p0ypt} z&~bA|)(UZ9drK_5u(OYe?rEikvzc3!*Ttru@J)MtW~&<=?L8#M{fe##)KgMoH~Drs zrIZoUDJflqRg|6m1eaTAJGECm=lYZi+5t5Nl>ETCJ|&Y~p1!hvRDGk?Gbx0TrRzQo zq}`P?>nJ@%a3qn!PmWHuk>0sc@86f*Yw+fI@fxSAdY^e<(8s&Z_cPvu@(*mFuVC+?t?|BMejG1#C1Pi}tLBfcPLK{)b`InSH{UF~2Xd6#z-pk0f3$OJ zZlJ6IlPo2#nhx~ghQ4M=? z)g@f_KqZ_Tn!6?1wqO^CT7J2%S}U*4aW2`gPaa|w_-B}m=BTMoWau9o~4iXQ-Pe`VaJD#cG1cZ zL%V#ZL*Cq~qI_C~-8n3HSyxq%E8+KRfi+&}vaYHW;$R;+uVaf-EAN>}<5UFiQ#!hx z7P+g#H35dEPBW74B0a6=%TwjL@3))D_K^7KTGbWNe4Lt++e~Hu#GDoxnVDLz#)D*5 zikOUA%`6NDgDDh7wK?%NfLfEqpTy}PSJJz^>~D%E)Jv$7Q;V3QIX-1%S;rQqMb4~% zdQr-DD_^z?PzzbDEDEF%^x<$%7ZN!W{O7dD`4yF~OBJ{=a(NYp03!}2+xIzj^5Kw> zGUrVwYIBIhIw@>nMIQoFlwTz0uI67=M#s6hR& zOO>w!JEhn^Gr0gCz)n%`d5^hYHuHFr0Tpjfv^Uii zZ%(i`+f=*>k7jOL@#Z)yvt7lTW9`kriZ|h*>91703D21is(538pt60%o1?9r9V*@& zWp8$@cypw^si}DLM|-nV#hVs;vvb9pKiHeviZ@5tn_Vj29ByxRt$1^oz1gke&F}5a z?iFtiwKscIyeZh5p-UZhR4B$W$2;vN278*xa}iPGsTjK~AvrTH48#-Ds$BA7HAWWT zX>G34ij2@dAdA1JvU|a4mmLD9mCNxxmE8(7ziYrcqxuhx8(OJ7kdd$jMy=XjA^+=A z;lB#J{l5KuPZiDGVS+uDEcunazSmyg-eF_fJ0Rqkn)g)&azxoaaceS)XR2zPR#`?f zlzX(QKo~{`1Nf!;xdH4;k60y=5qu{AAdXd7?We8t-?}lp&s3G1R%~bT8-wE6svsTt z=QhS5>{Ts)ud3;S;b0V5qn!;F0S2q=Rc{<&5-7HNtDK)+4;$HVrebF|f^IU1YW;kt_HWx!GvnQU8N zZ|qe$swmr+w9zWPSyj@`bSAViZM;1~;a>6}dz5opN`6cV@V!)grm$;{Wde)%r0FJ(LgK(|NTkS=st}+|V0V=1n3pjgepL2j3iQFDCnmq~)PP?RoCf{k5cGuZU zE<-9ELZt$cCoUMdUIeR6CPdb#}01kTqD z&RUfj9YqQjP4;Zsy3zE0GCQ@I03NxwqTRh!E|)%MFS)M6<>8sCfF>4pbkC6^6L;i_ zyOPf43E;P?6x)EmUFE9DJRk84!RwSrUEc0?u%Kam|GmthF|v4GI) ztG5nKl{w+E-9o)KIIZ#$o(GkKY@K|s;E$5STqV{*CKt^Tag+af}Ljwi8uVn+9|I5HzK8ko+HPM!~baA3h8G^-yEk^PGYcXHb(t7K+Gdr6J^7* z-~I%rRc>o7hHEay(@7FE;y{TQ)%<(Pm^#a;acZS+dPJTlIe|9mI941ILT6*_+T2u8 z=SrtW=jb@BmhiwqVF;rQ(R5>^oQ$vXkR6x_KOo?-82On4GWY}Ok^kieb~vaW70jjs zgY@2N+S6VH!>HW+rCYm`)^uKSc?syD&zMs3_s8Rxcb%H*&bbU16J}1+U;j|OdN~m* zsf8a~o}cGbY6?msFK&aMiz_KpfB{a)Vrlio^12}UArb^l=5)pt(5e|Z6| zSA*G)c!-2)hrVKgThQ}I%49vtqDqmxD zb~<|@1j71j<*KNwbfY}qZ!>{%i+H0MnLv!|TVF@TB-z`qIl!}*$^&M8u9XX?8Qa?g zghtxHDW6SchKq{Ae(iL_!Lv>QE9-}RQF)_Zpt5U$%YJL+bq-ZJd&xP>?Cl+{99(td;vr{XVgu`9$ zF!m3oVz{Zqp0k&%H`JY$%E-IZIbgaM0VJ{UVCZAXid?2dPM9L;MJZu2i_I@jq(!*j zy*JfZb2eskCEE^E=4)UF;0%YGrjk<3-j?7iEOO&l{303C&IOI6OC)PqJJq#jbbOsX?QaY>UA{$osWHC7VkaDyXldkG%)(fcx=^jWB(dL zw*e*7#pNb@%QZG(5W*TKfaP6nIgIy`wc+CcXh4_0Rnep2&t^frMrUtXkks247Q7@X z_}B39;^^@u@J>CV+j8-Z&qG(lK*_595yEj3cuBnL5({&~Q`4n#ToG#tY}+!6=fKWw zQ!v}Yd6?d}`v)*aoQg(55@@gdcX4JqiVn`NOgg-WXeZ_5mWP8Sr{1{OnY^)F>Jck7 zj})Gi^1_ML^bhR_d~(KS=&GdMXRXE;w!0uHRV^DcCL3g$o&9Aqr}hl_vu*E&$aMm` zHDZI;=*L^;$fx6Os^ycg0i+0awYg*uQuzv!1*hr2%{eE!PFu+E&l>>S@zs;x6(_2E zJ%Id-28mqZp^v`g4<;Ab^`UmY$~MM%zHMUcjAo6H$wf`FH*b2T4AYhYHFolGwU2hc zI!TW(_Stfoxz#mY&fao$NYyH9wFJFgHPtxPI$I?@e!SuLCN<^K>WD%h0DdqzsytnV zgl(DiJXukNc_)OL)ysZsqf7Sqt4;4<_u_j&rWWI_R;-SJl$)ZiD3a4qY|r>02p+H) zU1x9kbP94Ha%{1`3FAG!%5L0JG%mSseZ`u%&1$)?+GboHASlCAENvLv!9Am4ZoVG| zD0docIq8?^B4gK<+c|$5YiEtKx6GdctKVcqlS`;%L0tH>l-wO!okJwRB3GthE2pJ2 zJyaH1z;woUYUT0xbw-x7qm@XjQ9`DaCuo##oYkl-hd8xzf4S^!VcC8qr&h4<=4jcD z$JYvh{q~&RM`nca-lj3sY7___0@2xl1s7hTocQ6bOf81i9I0m8x+3l`Jyg~k$Z~X% z1#4cm9AT%iNVz^j(~-gj@?8Y>G{ioF73Gh+871Yk{Bh8Rq$U>~K)$d?CWp}bz&H+` z>Ky}T8*r&MmDBP&Tv`KMvtq6)t=KFfL`B<7vY=^KFlDvU8S$hhVUAW2d73>Nr8erCYExl ztwx^Ul`Lw_I{RegF9!3CTr_=*4Prxk))_9BF^F*dx#)CZ1_U-T^hQiDSec9t%oe3U{IU=PT1BwZ~IJKFEYzL@lU2|epPqq|I_dN(ki1i!o+ zV(GLr3HE^HBVNc%v|48$xwyRu)}99z^>xBrn^Z{GXU|B$Q|U)ik1Lb5RL)5k!T^UZ^?HHo&9-b9V?A6h1++EFSpnlx zz8+#Gp{%nHDVza!lYz_W;D|NkUrs=HgmXdj)VB?dapWw9Wd|1dk*@4i;H1Qo_M*z= z#(54@j8p2-?(H2cEjy)V=L2M2yX)7hXLIF~cI@Id5OTch43~G(C6>P$UEEg<;of#U zLT-)RR~^vY9mj{qZn92`thI`xsrVKZxlT$Bbs@8y51Zj=bl}ujXx4DKxx*!IwaryA zUC<_yvc1#h%E&?p>gp6(8M-0r&}WQMwZRDA#S@kdi_Ibcqy9_WlQLYMY9}-kAN(q4 zA5~Q4_gTgdhh*!vBxNlsi`!Pkv$X;1j~zz19$-~YkpiZ`7L}L77iXjcU5H@6kF03R zf|uSjXU`_&)tjiZ!!M77WjVp(`7)v;$GLK-E3F*QP%Namh>=J+1xM9u$)w*RHPt2~S5Zbp8VY5)kjTjm4H541`{~xzm8cNdJR%BY78kyz5mLG2CGVB()yfvmymA^|# zJGsN%X_3F$dx%fvW%88NsM)HSj0{XWV@?@f-3Hw4C=9EJl_iGUnv1PPNUZIZoR#`96pbH)ImtT$~a@TQobf>!QYd4 z-he4(Ug*;`-2{(Ps8i(Zt(RAVbq;48Lk>QNgb!4?YC7FN=fsEY|72O+q<_)m z*)k4H?;~eV-{K4xG$BkE!&L6Zjs~XgYDIM7C@8z)VcGcF%o%Sb@SWPPrn(0n2^o48 zGU5>cDA#pln^j({3gp5L7j}-Is4#&j$USc-+Ny(wFal5~A(Qkqa7MFBF@W(Dcgd?QV0VPRspO zrFerH{Z4|Jm-^Mni&%mgyvo(7f~;2mbQ%XhoBYQ-xTm+*oe% zB(0GL%1th_%AVZao^6)fX!pLj-HvRt+|HYOE8pC~n|tCnccx0QEZO^R0(N(7DUj59 zL^DrItQahEUU*EKCBjCFYCl|MOTR$ez~Zr9r+p3BybeYs@!0vBWSVGS&~|IauZKskQ;$tN~7*7L_xnxz2ET&x}O!zlj}vK)SvK$w;m18$SF=Al6&6FCH1nf!zKHP zo#>2IGa8etOcNtMB(z(7g4u~QG|q5Y-49`eSQfb2KVwB6H>B8>V`x@Su5P1nvpkay zI3B8YiW9L&o-x6M&KVcP!{4ZcbJcSWQh9bNx2pCqJJ&__U9Z78NN$;i0Qg`(OF!eQ z{Jq_k`>P6aE5R~nn~34rDOL4!JpARADm+AJmlEXAt-dZLd}@eMjnJBqSsyyS16? z4;ka)=$K}cuT8RJVDz73HtYcWC$+X=bRLd9?xXNwp1DRjb6*M<4WQB=e4alrgp-~E zNi&czkFxZ~x{-r=Jbsx0Am@kLAeGa*eypCRe{04W^p5l%ay&=bq<_g$QXFO&HdO=r zsaXgs47$Tr)+%krTlfm#hw#SItNvg9FEvjVHp9 zx-tc)PHyXPom%c0aE8ko$~Xh$h%L;-&P;YlYv%5@z#rWA7N2PIJ$Eh)1)_yqJ?)ir z$(Lh`azZf_5ki5LW%CQ;ir)ycBZLQ_#>Kk9MoI9sf$@eROr&Z-EBW{YH&V4!?l3bN zo!ikF*z0}v+2s942cU_`fPynvt~vqpatn;m87_|wV6%ioCDZ4$$QxD(aZAf!M4^dLMA zb~=v%aGO(fxO9HG+!{_9m~oK8UFzM?z#`b#A658pm8&e41Q8Z4*`qnxx1{omB79*& zXgC_C+>N6i#e1d|Bc4NM_hx69+z$hD4wqY}x%O0`*ll<)jKG16@#QrL*&VyiFqNy} zDHMHl3VT2Xn~~=XlO@oiMoaQ=mELK5PEQRv!o(!6#@GP~D$Y$I+WDN0P5{2HUo^he z1_`E0AjyekZj#^F0%>H>|%wkS*hROU)AhSBU>37D|tTRkL>z`O*y2K`iicO7_ zcK7u!hetsJjS*%!yvz*a%Wq&bGUqr$yM63;+F_q=EUfR1jO z((MeBhx-RjAm($UlWV8AjScu)4ph0ky-0=isemvq2qPnxp^s8cV*y24>^04V2h zxom2;l5!4N93hZSUx>b4#J8V@lX*0D++ivoMBnC2EynenVe)?XHcnL^Z7woCn~rQO zBfpYRz{yw^SO}RlK0b&=(0V3lZbXGWP-PYZC}a4r-Q^^#XdjagfG66{R3@uIQKP4) zWXGE7PBMH(VdvWy>SZ$1k49N9UkMs`CvC1 z4-dPUu~WIo$ruT>%kDgwrOg%);+dErpSGLLLQYPtmyg@KXKSuL=2ByKF1K(UEnxla zVi`}+dTrJT+M`}3ZVb{A$x%nq(UPNSJh?gi43mHvI>q@&0+!(N6S)X7w^`yPX`ki^Aat?Otx>tTZXVr>bmjE7~E_S(q-u7mzQOH#@3u0St}z zFJqadXhuFut#=M2MUUE-SIb{cO7}TaNoW(wVHD>yOjvGiyK67rPlfv&kk2;LdtY1l zWkx!fYcmP^n<}5Qb$_aTtg=s2K`al<-=jQ=1&l&_jBC$zrpoL80rrfXk?v!yccC6I zsu=c+%m;Lp53LDSVv_-jrIn{Lu$v-aPm6&?mYgbYwsn7MHRh)|s8$wWMh-~;R(UC1 zqZZ#aft%4u_Gb1~6PpMu_SpGJcHX0t=$0S22U~}y2;9@NDmw=1$sdb2Ts0?0WMyZ= zs7OJx=A4OD;n900eM>LVwb~h8EJ(ZL_Xzjj@D3sv5_~LgzeuiP#X_OK30c2jQ7du+ zh7{?r-|gK{dlRHDZT4nbH;q@m`Q}oNy|Xp4>`-dd)93gDYUhv~>x-!FZuJ(cCkH0R z=^4Y40aVLJQq#pyJzW$g(U&2%yQ!D8ELF6G3{SF-T|x#l^s7<*1FPXt+U4x8a#kqo zI|rgO#;6(_mmC&~L)y#Q&5-KUTVj&Fugxwa08&;<7zv~v_ely@-UdV-rXNcQ=SB>j zCTdWJwaA?W9(l3kOhsaF&Pu7gIyI}gMMm@!ecu@t_)1_WCb$N*_)r^?S)o4b7c@!d zH1`mdj}WL-x~64iv;FBQU+N(iQ$a#2PoFbP7MZ0@%*hANqo42)rL^NMo+1Zqk>)~1 z4(OI<(oHz1;U7Uh(DOYirg zNnMBpUFu1D5kkmFDUc819x5*k|m^@s+B$;HS3y2PlZ7vhuZJ`@g*ljKV5 zCfXc|3wxavFubt@*&i9T&mi3N@DHGm^u8Z7Xhy30Xrf=>Ta}5c;@xY-H;mEp+%4`X!^Kom)d*_5r z&6W}?#gm^FId;?>$WLAQPC;b@oaVArARF4dRW47JAXJ2aabyRml~pVn+P&W5PE8V^3m0%LbsBCB@kZPDpj}&-b0f5dh=^?f>BtwTB zMm+>M=w)k#oj0aa&fqJ1p#PUVbZ=7s;G*X5jp;f}tkTzA1%bx(vsKVP>xAf<2|YB8fZ>i4XaJQD`rG#3qcVu^>T zylIvP_T@9oBoG~ksm9uMz40vRNbr*nqSnq%2~31hu=bD#jJP}L8x}b(R!cU;RnAFO z%UP+KUcQ1rzZ|nQ^1D&9l)0#)Ok=ssU2CD9~G3;L3D3DIh62e6g|e`;F1}GK7OhS3vkc z)P6kJ(Ul(`4?lJ$uurc1x+-e9qUQyb-`7Xq89YC67aJ=Jy_~M;<+%MMliC|~RCB{9 zBk&38%~CHmHIMy%1qcYLiZL$pxf?ewctF_s5@~jS)9C{(5ezTDCEqbV8NkPpp0|s4P_zmx{k` zto*tcVb9~J5A*abS4$>u(Qi58Ahc&Zh++WRH8)v8y<;LM|7>u~!I}rLt0mHZI~6#8 zPAU01Li>sgDj~KTd!{cxNkldBM^QsVkxU51~H|La%u^43!A$sh1Qx1aOxpjfdUUZl&9%a?EVhmg*@b= zeoLKNM(hkBT`uvakGB~@h#i_qz@B2toSu!vnfV9x3;kt7IOSDsx3@SX89$?Ki!%hJ zNWG+q%k1l% z2B6)9js*_q6IY_-0h?j5Vb;NtQ~?pwX|Z$UmP&t!oQc{J=~lQm@KA*!GXz7Uq4)?j zoh*I6HehqKpU@0ftMU}GD&5{dW75B%W=^Zcf$t0<#)wk4hbVZeDsWox&xfwDA#w$< zBA1(SNYY#ds|5ap@o=MQ;ua&mykc4+*)?7%Ps!C6vB8l)OVITc2qE1OGY!$lXT7nK z*^362LwZ-laLOQl7B#>T7ML<8=kzPc->114InL^5z(39LWQKTsrqHm}x;PW?8Q0nZ* z)hvx(%cIuCB&Swhi(j29fxL>mE@w`6RlZSBFFjPIZjoQQDm!4iMr!4COg^CU zY^yM^pmISQuH1zkAh0_`;L{-j|A|^kPC>EOAha(gB&daG+k> zT8xNH`8>Ok$`M60M!OMx2xo42;_#i?(6A-4(ExgfPq{lu*kVg&Ogu18=+m~ z6=qP@$F#A(>$Gt0fs}I27`dGZq^X?^Cr7wOoY)eEn4#HIe}WvEoUNGrs|2%i(PMO; z)~j!IxNZ zKyA^1?(Ufu${=zl3{*adALsI@^8PfVQ;rALiY0!dYNu8GC>)_Bo0C7fvFI@Lw+YNI z1M{4aM?ObOa%I@;Iv!N6!s-k$tVL0#<%`1wV5>pDcsU$URQWM{=sEHdBEC>@UB9K+ zv+4=|FrS<6YrUNbe*t+Xc@zlb*?s|_%h}T*{tAM+!YuC-neGp9syjoN z?m4=d=u(*Qc=a{iNWczcn}NpBdzq4O|7xTZ2P%G_CD#Ih-m+&vrtC(!0OtFQ6Ud%{ z>hmfa3v2dmq zE=-3ODql@Fms~pR951x+FPJ(sfgshVfN#^lmA|T~Y#gI>h ztnm`Hll)t=s;_VY_-NNMF{Z+Mr!Fw`)yfn){e*$26sO|IA7T!9rhiVz?;=UJD}4G$ zgGOz1YANjukoJ?(R9WpRildAwg>egYQ3I(cWR>$H$k%ZX&BJ`L2+51^}*h#QP8yl_Fh8AsYYmMDW_C6>3w%oV9+yd8LZl|~R9^jn7K}7(m1L^<7ZXekB8gdKJgh@^_yy$MjE{KG(M2nvZkI%z3ps|e+9 zl+ev`e!+W%JB4^SOeZeQ({d~vUHw@-(@iprbzsQ_MtW|^o7#cOq4WO{! z!b|KMR26*++Y8LVNuubfV(MND%YLv6uZ$X|JC}f-XujrrtTNL}z}MZUm{wXis_1oj zLq?W&Bc_I_M`P{CDI8v zPKM?EjD6UV$I4G%j4Hxzqj85_>RXJpM6_lf^Y6Djnv(}t*&DE07jXg5{Ug%6)-0sisfR)(Yh!QJsoCJwG*59b@gRg&7duU> zkyor(0sCnYwfZR7H#6+)FVdNQ+zR0oT#i$jM+@u(?&~XLUx$6-L21g_o=e!+$0c!K zgOl+g12THTM=#zND( z4WWG>(5@DZzQa>_6}!{eak)MLWt6X@?uv@9(7vsyA~DQhlzXz|sw#SGp-;(AhY5&I zz#cib)T3J#0ilYB!>6V4r1I*8pyJ!50=5#Y zLfB4PWYebt+vW(io27CemfXL$uh+!qOFQFiHIar@0c@-0j8nK^k>P<TQS7pQ#J>}A(d>}N~7L=Vx56P{qLx#3RKw1EhZj@Ug-qn%; zZZryNw%iirGBc5`Q-u)YTITvaPlT+ttb;rH0VbY_YqN0*5u&5AtG?qufhbMleX~*t zi$dS==znmyzb1^qM5c2LFYdZxp|Q7zOAg}=m)Ue-aNVCsf5#(c+=;tr7j1MTC$J@s~ezhDQ_P-6XJIp^j z(*%nhf&)=K5DwffT9CWwU0L`76q%AwWV#o~gc9N36g!Pt#+PytgWngZr{{$oK8Y&? zFhguX%(0)cla?SDi7$}C35>!QJypyP8(lRD!HBjlgXnQ0)9nhARK>Ij3SpSDW6tQ~ z65((+Edd(C#vMr>t^}{OMmyw>a9hmKz|`QVi!BJdzegA#ysNC| z48sFvh=$UKwk0Y)EX5J|9+YA>(`Ac&N7;$=jyAP0GmzQLrpE%_U?Tl}REFk;#A;jA z+(j0Up2`baE`tW!qXtg~hlCp>SV&GKKxS`_S}jhMy`E9N(x}Vt&dr02gZhlvvobHQP%8|JuFECa+8lN|usG`U zgP^hDHtLv8fXYL|X-&rg=7v_R_G_ zIH=%?gFv4f*S1>P$w>Gd7;$48l#-KyJ;6i}HJC+M5ZMZ!pF31+xpI)Oco z_OG$P!b6zv@a7eOm?mH^IW#{R#C_h@7J|NA2H1eEDdU10!$pj0<$V!EG+|0630LR& zwzy3?Ow(<6oFF8sBew`RmrS+QBnkZfV~=q>j3ERFbGSoIqMMe0x}q>{{_MfmmzE{r zSyxKGOk$)z6q(du2m>O+WO2_^;MG1+rqw1?S6RoV}x4{KmKgd2cddjPlNJIw&c zKx4F-AZj3mG)aJw3|$H0gl@L9%Ehj9D?Ogg!{+8D*e$C`)Vnk(vMYF;jG@ONB0>3U zAv9q+BNsjmJNd+Lq1M6Ex)U<~T$1Amaau|nvoZzi4KGdmVIC}qpYWgo-~vLz&k4Wf zul&;>0>b*VP69EMf7BSnC)2)F!%Ja%rDGbTT$+fhB?^JD8)MPCs$Wf_lYtW?Rzhzm zPfemVND#^4GBFi;55F<2Gnc{V;cfj%r!7meIdeO#oR=i2Jj1mK*QHr`7x5NB0Kp9a zg)#eb{#a?GQ!*Jafqg5&Sve7`NhjgaJCP3cLv;LPTV73qbVim24{$kb)@_$oeiODb z)E>5ShqR&%l~%#@1z>MS+(lfS90^mjA~+<{E5mvLV+8brRzmCljEwZFV-jXDd9VRZ zaT+m{z{GVRfI^L?W$BKvwO3SSSucG4+S!qdCm2_U_UZ@q*w3Tm#1<8M(P7w`!Q_nb zv?I{lH%G(-!sp1Mz5AgsuBBsl!Y}){B&iH!n60uELCqiwoo_?7iOSX%u4}Nd#P!?( zE1UM#&o<(mMDGE|nJyQh1wFm22jUGViS#yt%^pGs_@|PC3BkRxbpx8-xhyYq;T`e< zuZGF1!)aIzS8g?iu1DKgh{u<~m8DO4z`;H()CS0jICAd=w$WOkAk(AxU`YPeB;lUN zFkPn+k7<+OVUKQSq({sorU#eVEJmQ{Mdqg**5I0=>rUynd8Y4!!aI21Vy-RwkXgvKvciXJz!gUj+9 zTiRP>;|A;0*FmBIH376fESF=d+UmMf;;vTdVk6;B2A8xrd@s^)VR2|w&?~4$9BMd3 zpGwhxg_6jjlfxCc6zm|cp2BWi9R<-GOxSQxxu7pk2|1VBUngh}d^Wexf{=y{oDwv| zbaqK4WDM>!QAvSt0$=c7O9m7;n^*rns6)@`&UqRR6HEz+FV+YHv}S-uz@9;hCGUV` zH|P!KU+HVQ!lpiurV#Zy2spMGG*Te8`ru&aGcw~&4eKn26m|`!&rNem3}Lmq(Ilpw zSYqZC*o}6Mi-0}uSQs=A9(Z& z8-9Lum>%G^c@XUf&tRB!t^&x|y^z?*>Cp=e0pcX+v|K)*H+^S?v zM(`9n_Im$9O}FR3Kf=>$ZQRzBy?04K*+Goo5T4~65CSouEOKjD<+?5w0i4__v5l-{ zCt>w8#?=`5wVhCt=)0cXz;v7tt^>B}AAkXWu@tmVf#%tAiZN>a!e9S7cmf=SYP#3I zE&)+_fB_QIO^8{5o*(oRZbUX)^a#LYG4sN#U>dD`Se8dbig@My48x+Umuygz=t&k1 z0J#5-hz%woyfQpT$1q|qYg9bV=x4_~;6nzYV0F#^pqV@A01p-d&>q1naaG)AVQ!eJ zA-RD0C*ttqD84JU6;bCj6yd<*1!3Q*uAk#<)83F~Y9kW(kx^O3!((yuBM&XN%_FD6%*GR!=fGvHgFc2(G!O@bax zVmb-Fm~4&-lysoh=;kSqZl8d@54wrMYou@oCu$S5Ld6>)rXHsWQp+U)^z;WnAlUiP z;yfze0RLw0!?>RVsbB}EG2}<$c7pT}ukrCJ#0iZ8@hX?8T-erAG&jz8+FXr6WHv@b zfk{kj3{OU>(TOd3242i+5?wkpLtsUi=!SQlg zII?>zykF44=0TV+%&aIfVE~`b2Wzn{D*up_Ukx`5-pmLQLyzEHthF)=z49SZjd?{t zvimL%d)t6{l)C}$k60F*0$?*Kmk%z7=_|tW&!D^-PY*1DT4T?I%U6f!$RP>#uEGfMGJ4s zTx-w;i+WJ;MJt2%EgcS_9(Pp;^@o5OSOsbgwDw$a7(h@D?D5qA7JtDn$d2F2w6sUI z946lU!LUHHl2bH4)UXfzA?)BPs0P${`gGxEPs=Be^5Hn_x>_5Zs)FfJEc{Du3#FTjOFfIqGUDf!(E)TIPtYURr!x)_363q>RE|x|50o@O5-a$D|87RG z8(&5lyEdGYP5zub@(__ zb2B1fBf=n9P1-~v?)TZq>g&Uy4x{Te*bRLgZ=g1NYTTC-&VVtK_9tiaXLJG_HqdnE7lj( zqV-UXK-r7ZvJJqL3nVNHfVmLt#RTNBEI1NBZ3FJ%rO?L~>BG^Y8_OULflVtmn?D&9 zYs0QC91Y7(9D|^yeR$1e4R8ybF)pK>9v6h!?*ObfBhUL@c!Sx|1PXH8EMQ_fBx6&l0vm`F4C8K1kf)<7YNcV@eUO_FnV2Dm<%X|o? zCox^)SC;5XxJ&3}Hlu$kDR2iZ4)0y#Zt$B+knjKR;6~nQ3b+c7nbKkw>~w&8GeFP}w+JCFt7) z%7&_(p{Rksh@_VGF5n4JVSrLh`wpuM4bUx>0Egn+_}-DTRfD(ys{|dI38;W9bOS@E zp@nP^Sfz#20V$S6`uAilB`^Vk#q@7Urno1l1To*gs%M!@BoL~gB5KbQk%kp{}mW0VG0B$v#RJcqctpX3xumXFJ z79~T%8Fl%z2z=2vjDNhYF`FMseH~<#4gyca8KGtzf*@1c;qA4uRx$Scq{psc~|Xs!)QV4y9CXUg9X8I_%U<0GM4e+0n#u zWO9vXhKEeW2ZCn#seX}=vf8>*N-HTq#&dWAu^XK_ft`Qr+M9uc6bENaXP zkYh5c$}cCU0s6&M$UrUnFuRlP%;kgINnfRRS^8*->>Y4-+MruF2W> zdNKwj{(UgfR#j|WtLo@>XxdaQ-51NKI=UsdLDe$VWklXZ&M>D`Et}t^qju_Z3ma@A zovzT%uEqdrBGdcisXFRimK2T~Z9Nzp*4^L(k9yTvI3=g3TDmue!hiq|Mb$F0rQ?mN zj>cgGwM;+4jiE~Z3s@`-%C2w}*(9hsTt*P?6e1G`3V?vNcRd)+)^d%-a*R?Av`Qs_ zvjCuu>CsqbMQ~HB{&QICR9y=+1UuPE0Kryaa!Eg?iVI3-!~@jP)>tONF4kZl4-L<; z8}RWJjQqGHeu_pbAYpYh8)Sznz0WQ4<_!642&ZeVS6m zRh>w#kXqG`nZf(_m`V_LcBWc5CR3`8GIkJTY23r%*w^W?$7Pp*u;MR{4RCr^UBGw^ z1vT0Ys?wa7s}gjRpn1E2%?GBrlVR5p$Jr#eazE_T^1t}~U1CU3ClKn@LjUEojZ>?g z;0)k&n8sod2*6v#^qlFT&hj>!!G0m;Ot}ZUfGS!E*|E9=0**CV+*LYI;XOktoS6mH zSVQuGSFCV{#=AsS66=6dsAyK>foBpkmsfO0=Y!*dcMC&!nF15=a@Zm*fyz(ROyEfY z>&!6nQ~J?ph;QTkXF?FaB|Yr$(-&xLUXuA*0U#J)(8uDf5pbcG8bS~mFB{>m2W9)h zB;y^-6l)W-DS?B+z~lb%Tx!s>SX@;_e~t-6H%hF^1HjGD_a#MMC6w9e5P!w}0URUD zr~vzFFbQ;_hW7;@-iv^H+F#^dlmUKS)@I8bv;dud;%VL!lXARGizD81gAblf4VA%J z+#MyJVN`M|@WJj2)5~bUsHDkxs*%92?U5Vrwy+GsjH&UICy!km&w}w z@~YfMfUUK7X&&-Jo9VR}(A`Q?Hhc+lTh!RC8U$f>EH^BX-V}CspI-v=Y(2#(7LLsa zYg<%qP?z%{n!!=?3__T2O4)?hE*yrIVPxgL&0!}`>&koXWZagp+>5%L8cVN@@-Vvd z5aSb~#+PA)qHH>EF`kutc~k)~yt@!&!qTw~Wm<9KSesGdrBMaOE$i&Xp_p^rQ(;Ge z4|0MbCJg7LFDmhY1_c3%5&0}MIEq5Zk9#_7Zk6xGDVJG9Qc|UK4pxm`^=|=KwmPOu zATtK2{_7YJ*Wk+La3gH5eZz^GtzpaO7Pq{bQ>v8O(J<3apgGzVF)uell?mDulBL)) zVJlbaR>Uq0V046w_aRW6t#oMU)*aRA(Y5rdxnYTlXTx?8y{#7(YEP(II!74nDqoO< z)q7qV*F(4HLl~XN*gJ%PnYACU2w2+5&xLK?=J)DhuYg6uUvR6xNBizHH2gPQIWUli zpcUp0Y?adu4_OyVpOC9z$2}-#5%Az<6U-Cbx)TA*XEU8j8D7C1lntXkEguRYkw>So z69q}a@o$StY=PmDE}`n!`~hv0gA2Q&N00IF?Yt7t(*zsp`LOlpcdh$ySavD4Paw{L|}g5cfs5f?ZkhV*6la)3rK6u(c05^TDrqIqd9YSOal~==Grm~i~+k`w)YDrs3xwIfwcCOu+a@YsVU{vo#UC-k>RZA~IYcXO| zpQc$ROpP2BJ}z~=5q7*yH?S*ljlCI`kWg&?8(@G(*|6)l#@qq<(5S7diXN0n1=Dbw zIm`~P!OY#;mdDK8EU0X0)45|pL16V;Vf%matzpvYAbMb_t!5(-Dte)-#j)3AXQeCf zf#;@N(rsY8#B+*x0F1*kIX9LR@En$}G9C=t?l$n+mqhG+8lpBByprkFYpHm_K!EZh zBlT{BX**uPS_T&FG{cjWQ*si%AM$Rele8u0fm^BBCIlk*4-mctT?2Un(`K=IKOgRJ zQun3BdBl%G)DLle##r!|R^u}80cfOrcHr}cY!cCBFT(FlL6%&=Ai@1z60Qf|>`%~C z8y7knvo!%5EksYQO)xGQZ1wR91?~YP2DRm~f@DjgqAECfiJ$@GVC0=UlGiV_G3*=~_N-3)J`Spd z{W6~;y+=Q^5Y?zOgK7(bGa#A?PPoXmxf8!tw_({T(U8{C0ADQHZPcj*$BU$y8j^aI zV0X7v8E23R!{%;7bQ{D>A(9pB_ zA0RWpQQ1FCZ5zUq>!YaYt%5MUY^{d`B<>3jrwQI!5R3qI@OKKLd5h{`{!X(cW*D(j zvuj7x(5XIunTFpUy!Fu({8k;ds2QiWSq+F}0xJv9+S~Oo)1flxyJ0dZ{{`;3u0{Yv z5U)j)1feWo2387SkXO(`yB;BWDQn||gp391IR}wWaVe2Ei|T0~uaWCuR&q<$hFSL> z1D$pobcX>=02F{*u$~J0?8+q@MAig21g(syXKjSe2qCehG!xbxBR zcCe1iKZ)k(L7AgfTrZI--)5j0)0sjil_29j z^&76G?~CjixD4}Uv?3;hvg74zAFUMDM@-WEwO+pE!(j{e;+hY}k6s2FE(hd~`bAjz zMFpp|^a&_3oPZ!%Wm>UJn-t@u8;bRhROZ(*T@0x$xC6E@z%;^&fSBtv#8bg@7RTjj zU%1;kG&>m}@MMEy24455rB`E`&l7AjK-Hx&f7Gx|3rFos=f;^~<(;Suva5M>7lIJ% zcNbX2@Q{W^gdyrfh)!)<4Bq@LIljO+Ee3x>ltJoCqY6kncX3iiMa|NNiw~NcG<}|J zGWq>qX6U%~Yor5AfmOlv;P*hZ!R7g^NQH`O9H55_shShHfu@T}98p;Y1qhW2)dI&h z8#t?lRhJp`O%MJOeJVkJi+M=s5_(PV;07;e!DR*<8-yAo5=zmDfa0g^dMk zCJe+F%#TcdsKQ-pqtfqnrEM`B3zr#k=S$Fz*ykP$ml?DYb>;kb)MbW6f6ex&1f6OC zc(!2A1`?-cQ>Cw6z043TCJfvP)MP`c+l$1La<=IozJnZ1>7&2`j06mrJQmZCrbyg0 zc~n&NF%;zvaWnr_G^HN!u2ap1$W_!umlHnX<3UgLa_O)3E}^-=P}l{G1(lWmK2? zV``ucfQWkq{QvTY)IdMRh#X96AV6%Oy2wSeJE(uT#^JsOkZ6#nkO8kK^Sij*pl@oF z#6}Q0+Kb%^?nJW{z|qSMrlm>T`Gc!aUr__!)dWsa`20<{3ZY7M~qv5YEbOJWu!7+3r+ zP(SE38pIcZczYpawtu{VySHV#iBem#t*-K{`{iyy)c7sa5DpY@f zb34hv^fFwY(bG!*^ZjfO4qY%yClN}}r>z0A)EhBNuo@^(Q{GI2$+_I1kF$9Q4}7c9 z*xo*hjMjv371v`uG5c6)Z~{OQE6!ru;Hz=FqFVJ3{g=}Yq5Gf{^f68^-bDLl)M$F7 z%~s`flj*6^^nm=nLViCbzweRXx5@9@<@eq48v^ywB3&xyvD_ks871Yac=jx`uCqG2q;tx0); zQ?qH3zbQxK^*v5;ADx!7)$Cd|T4*+&L3%iwRHNwznFh=pEPl;{wtK76fI^bY8nA{A zR?VR~VXTYg`-aebsG&*WB*FvFD+ps3{RBQ&Qa{I`+F%i5Lu5d!$)opMllU*!qvK9B zTGs$CD%DPuR!)>zTWUkpuSU~3GI2P=uusdGb|S#B3)eoh%V==-hCi08dpVs%Ffbhr zSOT0F_-Z(e)n+BGvlvE-&F@0_JBI0TBLn6>!49Ua$JgFDwGGz;1g7i;Fb@*kVtTFI zzDV3dv9KL3r>uI{JZigmj!cuW!9Z|{OplXr?0ys+=Cr?y{5i^;9|Z zkfWjcDu=3SG#!kzoUKOVze{X28b?JrCGl%4(vHEe{qgHJ`1L(cmAmwxpmw8~YS7PP zyY_>p#c3W&?}d74{Q4$-W%0|vuf_N^3cvmbzYf8#qw(ti{Q5P1ecw~%jv9@rhhyIg z_UQ7gO-~1sRJ@~PV=-u}aQ(E`FnnAgThq!J-lUlt^Bcnu)nM8Xk5eY93%kt4+F}^* zWRt2KqFQ>o&4y+UtO%sv8qr2%qkO^eWD!9uXfHWrvBZeXVkPAEzXHd86>`jP3`^JW zUy*;nsfNYnNkS=ntN28dj=CotFAie-#WNU)+Aq=$h%5jaw10?x3E!D%F8%|vW~;t) zrK?Sf5i16T%rOYi@|51kmqIXt5okRL8basAe?=xmCm=}V|6qE<4*lNhT;R5h0^HVS zINLx$W5M45Dfo+y(YO`^qj|Bv)XmNX@|07jF$_pPmR*eP&?a5T|7$j{ z%1Pd+#M@GG=)tb*I2+~)GzcwwF9fi-V~8QxKd!}a>!W^vC!9WV1uq^ z2OXbf;0$^Z(xt!10$80G9etEd28@08xBY#3l|hG!ZEgz3qijojn2W~rb4IqfubTWpvcG;VuwT>ypHVMTSx~*{$R*i>>KifXu{9W z0-FbV$G9uk$9c8yQAgI+D(#6(SDQ!?7psX(KSDT`KepNW z_Arr7p+Xw*U3^UFj@thavS?+~Ie0OzI)lphPytfp+gfc-;Cq0ey{`3Nd8lIRS~XFk za6lYJS3q`;Vq4*}%N;5~>zBzj(PnzV$f$|*f~Wzm0FC@yW${F&{XNQt7S==G^QUMm z?L4xwtb?}v7NKCYli*mUTWdU3N+10eJ9#YA%EPv*Qu-L58`W4^d6?&mtgsw{50F7& zmyJkB!ReKnpCI-RE^v7ags|M61|oM-N*L zfYHbw_E6zzUE%Vm0?H2^Q0&sM?eK7rSHPtKo{Fj3UBF80QDkL@EdO6ag)850duV*nNDkkkIcEj=QFt2V0@S6Fu)_3o|So689k9O6t#6WRYoQ-+k*6c>FvGr z7Qk{zgrd3tg~r)FAp6oCdsVt_U`8@IO8*o{g6E>gGlJZ)qxNOG9xt;nJ+{PVP*#M$ z-IuACJsc)s19E;zbi2Rd^Z|*riQ~ zg242LWqCJ1*X2>E7yMF#x>RpbslKSx>wc*Rb*USRN)1J&w)v&*)}?MQDplcx1Np!& zb*nCQZ&9gnQK?V;QvJHrLq(-pqEhWfCHNE_>4E9-qEg*ase)hXO5N6$qEdrVsnh*Z z7wb|(MWteH2>p3}sdIFxSBpy3M5QkAOP#7q{k5o6Ix6)CT?)dkV1q1=a(lzvaNWA2 z+#7e#9f)#o-#vFI%Ds2@+*mvu;X}LUR!6yy@18p@%H0g&OjqFWLTnJ5Qx;vWJh1)j zcKTZq=VdR)=?TORxkhFPqTiLcJ;5j^FqpXnq z@J)|_DEBHq7j^yfaVW~Ye)rtikHd-RM=roE$iY;>^d63?2#V^c>>qK`!aU4f5gve1 z?f_aE31e53`zJrwK?o1|7k0CW0N%l<*aDi!LIwGTN=&Kr;{ssO7 zzNRkSpzOe=5qaIX@qySwDr@^D@{LSW04dyoaA&e;zyPa_m=@Zq`Hmv1=CMqF{W>l} zrS#%-8^f_+9;^8^g@kC zFNlGb%di#w^CY>T-B^n`Dy0YY=gqYeWbRM;>z3LA`nXSjJyMh56acyKKl5rfy(>M4 zMbsOyLO4S|!3_OW0QMvhiH_}nEr;7S4BpYkF_?*N$PqvHV%*5SiM!)cDrAGV{ad(l z7h>gP!QDpbAJd&(RwW7D$y)lX)q0DZApYR3F|lw=67isEI#qDq=koYP5-Iz zfeB$NThIzF-~k%4BxzCl!sZ#ob=cLJT8&8wrB-Y{SmF~eq91tho*uaToGorifsB{V2fXH4Hk zGWsE)mF{IkqKhNj5f+gwz+?^6Z)~$hb1Uip@Mq&!W%Tt*TqIq11Av5;(pknZKHtgO z>Q_u>=WJKG^g6H^*uxG|&k#kM8!|Y9MGe}OSmy;W`0HekQ&i#wM@Gd_QNNzN(bvZr zdHNTTJA+bPMWr5%Nik_C9TXTWD)8){3RL_e>gTVzfbPf%4Eb7u67e!xZb%;uZ2R#l zv1nkf;Sr`!1d%-eMPQ&YL(cX-D*#I9jMJj|g@+y>Uxthau1kf)QBsJt!w}5C6KS#h4c4noK9ykO#vtps!v~ zsoiB4{_JUZ$(M~NjMapo7Co$h;6G*ljwbCyI43olo{D+OqAO*Y&EH{DrP!Go*wxhh z(jS*atMw5K98AHK4VxS#)N4C6(gVN>_-^Vg_xJ{R4rJ1*lx8r1Ff1x6LuM+1L?S zI#v88Y@<79quXy|>uzma6t(e8Q5(PY+c=ya8&z4>3!Z2O@Tx*|q7q0&j#$uxRfvV? zfp};Y7}goWMr$8V(QoohH4u$%6jkiuOxNp&iXX!R!s>9kUfRJ5PG?7bm+9@rb{tmy z4%ex}scTe5gr`u|T)D(nhtn1zcg=YSE4b|24gGrN?rnC4bv_=X`DuzHT+ir{=fJX5Mm#!FBq;PJeq7% z&ZasKlnXMZlubXgnI_2&fkMP(2maZ#JZB?-C92{t;KroNcy9~FdoKos8Mt-n20fbl zks=LQ6{?cfEwj}GkVSy^X3WFT9ueAsRi!n8M`_t6*eZ@cz-`7>I_+}?_yz_D;VZOi zDI2SZ&9yP+(FEH?0D&bo2EG5tu~OP*!8K~lBYUOc(OEg0X%-ZQ4%5z2ESH3C`uJN_qk~Ks_Zo2%o4o z0W3D{W^Sl=>E2id4_6%N(M+5E8%B$GP~|5c&9xCnB4=w4^Lln%i-uwaK=^eJAwH+t z9)4Y$vr&67m^7w4bGGki4&kj!_m8YWi)`9UsvTuhg%{MCZBJvz6}u)V5kXo~Oy8F2 z7}41=f<*NaT5*4m>A!6?o|cPN=Ws?~U?BAH(nubAI8(kv-YnNhBy2prz0}_VrPlStHbGK3*7c;i+l$6aHa<(<&SzjAFwdDej~Kd2Eu?NtB<3Ja~C7<(cw(5 zB0v{)!$R|OnQ;IIicuRtre9ytRuHlqu)kCC$PsXZZfMJMiY_+_?d|PMm*`^@Uezhq z&a#9+_C=@9xO7Nl{|-4h2$xkK9IIVX?uWbQ4n?`E%0*4%2o1&7gdSCPt2o+Ocy%~M z*)TA2S7X@iWkJ=Irhj4$1({8BDI_nQjL0gQrmYD0mYV5)3r%#hE(DRc!(?rxR$B}m zI3;MAroYHOhYKPKyapp!rnO7*65|1EM~WS5bG@8nh|XxirNm*%glFQwjIcNOnssPC zP>cU@565z?}-Hkm(ma!NSFk4rl3k=zD=onu{-|1LV;)%A+Tj;dRhB(=Dw@L~_G{u?$oErt+|yOgFdME~xFZ zStNHG9*SXA&3?Ix(Yxf)N@GoC} zQ~qSrwXj}%MoQyU0R0fj^agnd(ShHGt0pjYEXy-i?LE~GUw*=U6JcuZig_?Jn`gtC zOil19W4J1KMK=x$t7B~$|0~OEdnFy>fg~g1{eQw_@R4BQ`!zbP;uN}ke} z73x(vT_7c2G(fOmK3&dqp(H(Pc-+9X>To(gsQ9u}gh*BITKI{U)A`xFYNEH$7ntP? z%LRZNR5_h4Ey?k%4SUYa0`NBt9|_M8QkR+mHo6)$1-o! zsPxGappuY)f@-20{o-j=4g(EPxS+TLMWD^J#ov#z^)4wMp58LgN&XCWGQMFM6tLhY zjSMH5nluW4lT;Jvr6o3?>}Icl!3Q~P(dyzhF=ZKg2U7-?l`}Xo+idPYz_j4TJ}@`L zjaPA#B$&ooc#*Ckq{D$HwQ!ae*g?(+35Ne-6Wvted4y*Okl$g^dul7w$$~yS=TnIcS~_l<{6oS6>&;^tdjFE(M0P&xq6R&W>00R%rLlQ z%hN`mkH}`aRt24XJ40VVs4{G14CU=*J5>{6J3*1T7b*c5Gbmo4p-DD<%kzXVVgN8p zVVWT3p}Y3&X7uI&k&tMXJk8}e>NeEYGR_!a5R9nG=_LW|0tGtir&75P0a)Fkg$}Hi z>(a@&(8r--VThqS!588*$IZE%>G{Q=%NGfc)u+m7NYLkE`Ci$eY4^}#(0QGPGS5kw z?ywAKz-nP33Fez}kp~C40ua{XAIJ^?Z9Qv~W&gjDC;tf9kbkBG+{v~@A<)-qp3~!G zdq5N>FAj5WL@uN%x`1y)*B2FdKnkqpso36|R|IA^HA})K9+46qiLi+8eBT!p*@z;r z!0@@=*wV1}7Gwbcw>_wsj&ff|Zc)X)C~HTM2s zEK70W)#6M&0eaWTO&vSJ=Wkz>*tZz?LQpttDg5niphLmm{to=DYNGFgwqq84iLa|= zLN+ZDE*B%M$*8$-bece?=P+|`#e~nrOr9P9{2u($s=Ydt-!TepEe{Crn-a}D;&1`a z8Hj&#U+C2 z8=fZ*Icde`1gKwFIv$Q?*hQSKgFL3W@_aIGTaYgRZBme%Ccu9-O;>Xf$XGRT!gcG_ zT()_Jnn2I_t9p@k_JI1X4He0jVV|Tng@2%_ESynRiF( z?i0h&e`%z~z`qDmH%Ca_Oe2xHIYR2@-I2PPcSGvte}L4@yho&N{!FB9<_M{q|1nZG z|0ASsMr?*Jj?~S?NZmXFshjtN)XgK2x|u&8sRvht%Q^@p-4~-(pF(<#_D@{&k7$*I z#_&`#-c)O%dq?5tZHq%%_3NZ+ro%ygug}`5nI6+S5VY!g*`&=(TX0|YY1Nt1z{>%x zTDdE&YG(RqaUR+m)kLR;w5pluja_Ngl|lP&39b4Wx`(6Yo@v$l0^Vu^4CJoV>b#(( zmy9pEuykoI38=KuYZJrh6=bWKX)8`6rcWUj!gsrw;kvl?cL(u$Uv6N{Olu&oQ4{FB zHXEM>Q^-8T2YV)7)t&x2x6nH{lekw{Wys(3p#-_Teg#DuaRJwZv5cDFpN%c_TFgV@ zp8#T*_#w`cS>T|^#W7!WmSYU;sTPi7MPJO?s)Z!n6I%Ibl&4y#W3fKTpOI_(RhVE$ zwcs^JoaD=7R$4TF;i{&F=nTgN&QmQoD4H0oa|F@VxbKu;I%p}*@F%5Xq(b`Uhyy+4 z@@@m=@^RIa){Uwby50aPK8E5cv^XbW~nVKpTg^h*=df8g#LF_syWdPEjL z?+P$?O)O7BwBQk5cIeyScajJR0)%6DQR|uBW@NY?3=c3G4@@oCpeE4yGQ!hM+{TCWK$ zX`wUC4Di_MTn*t3y`+9^l~hbbLM#)%S(8 z_CT07IV@3zn2@WW!4%sjECm=)a6BoRYSS2K#2or>L{xfO0D(~$>xWzj5x@hWVyA_@ zZifDVJ20R(vPpG3mQ&Q0>~K2>OM4I$r7x=V2`Uxc!eXZf?Um70RLN_2sxK~ZT+_df zO%BR@uZ-R+tyGn?X=$Dwi=j95J*GjGQI&M_GFz-7;Pm@FraKi(RuC2Tq(!!>q~!<% z$MVwd_n218<)V`5JAplK&6$2H74%oa0@dGv|m;Uk96qk+Fl8g-(bAL4``%jyKF9%L`3SV(0j+Rnksm zGEo7l%T4nU^tqDh@^Vp3KCQp$%5vCBj%RufFJYp)d=>>Der8F%wA_Plauw4p z^K2BmVlPis(Np-KEBEr$czQBu;Tqk-6C+#r1D*h({R@zq!dl>wp!j+X(ZeH)uSIcQ zt*YoDJfDUxZ`sR()-nZZf}KpJK)MbML1lz%Ey_J4Rp}R@P0I3f{Ym*6y(dfDDFT}$ z2y}IcnfCPkN;J1URY-b#IUe4bJ@=LPzc5e-E*aAD4l zt14}EjuI#9D7iwfI#m@Ru8m}zniHXITA=bPpOLfiaNcZTunvMa<3M^%>}njC8jBzT zlLRkBYJ$>F3&SXvH2si`Ll=ZIbGpzn7IXLqVxBsXZh@U8BU(K`yEOp*UJM2T5)AY% zloYc8b|Ap#lp9%XfpLu9vF^Sw?CW;3Qj86&aCbOR)T8LA&6bXEGBMqswRN;&u>7){ z=^pqa-v$m)?h**&52_CuS;e}kUkX%*1N`46<$_hk^aFhUT5`TGWBD!o<)IRAOy6Z- z@nJ?ZFZ1*8sTu(Z~8XUD+ZpQVcOY> zICm)+V`4JTHTqQ*U6aeJ1EIkLLY;7NbZ{_<@+rrYR88}zf#DShDb5K-Lb65NqjFwx3uGz<(z8K5$v zxpywJRW;>HGA@~!My!cT)4&#CV+hsPnj2w`;gnIFp+780gQS!^)m&Tv0SHB3IkGic z7#|l4A>Dz{PN14klf7aZPk>3{PmeB=PrG!C9G4XIZJd}I<#>djN zIF)c2u*B#v7qVx|qcW!gbJchnFbXVB^xd6Jg4ep;fXBvIx;tyr4JDq^uOp77Cqa&x zuFi@(#wOW*>o8_aGDA;>V|6^;f*dtgjv}=8N3`ejYGax1%i5i^RPcpblY#!H``Qrp zHTxsPQx$a{n`iNI*ii)_&O8>#F%}Hx8&x}8HJ%=Bf;q@ordz)%T;sh>0@)p31;2N9 z6L^e~p8PBJF96NX3tAdOuTV$J7JN42?tpCi{I=0O^k1`f)w2L6Su@UCC2xn_(ZgLPC?}0|I}~eiT$iLf(GdRDF}88J5bq7Z%)PHjb%FXTUh#cr{VK+-Y}4e3y*zM&J-UbaX#TP z+6bPGUf;{3v*sn^bm26cULI3e4Fwz|AYe8e?sMtJoSsMw=UB~6Q(Iyh`}{t~GoDxee1qO9TO7Q83;`xx&}!53+NSDaxi128%nbhGH$6+ZJR<^8aNDft z4@Of)C%M2hj&9)Ng3L~%3}G~Z@6whg#%Y(GmFg#q_=c|PG{kiOCrnr5?BEboxSni| zg(DsaTKp-!C-T=0wo*V!{9{fY4hUQK4L(>ur5*pM-0~ZPa%FH$2XeUqDtCu@)Mkv2?tV1U|opr7|&2sA+>-*~B_2MUPP3zoZ&!~&nwa3{U^8rjjvPtf! z3znniwy^Kjurb40g+{E)qs$ACsYcT^t$AGH$qNe&Muo0Lp%E=q-ySxUgBAU@xCy0!$>^qaV#wq>eZWsKXaKm5VfBTfCc2BV+1rY;H%RU1 z$%k(OW&6RS9FDSMcSW^d*u8eTDBE|tV|oZTZg%K~oZTMqIDJJ0 zuTYh$iau_`xqb=)cjnE04j;4GR#fs@YzhQMt7S(W#Ps^IBqr!w1^ko{VNCmYxZb(e zN_kR^U0{2s%F)BiY{Y5V7uUp745kagI+Dsj@(RZ^z5hGluEHteVc#mo;SN`grB~&y zxJhavFbfFh+BD6EUul1L9vPN;BP~M5?k}=f6!J?Okwrikz7t30gzPhPTt=6}ZZVM#Q<=f5JJ4S(N z#Ki00xTX>;%=|gV?c&cxE$vY7MnYNq?Qqkp)oi-AsKD})4E-{k)%5+*fUdxv1O_*- zAMA3>c62i(Aa1cgMTM_JVIaY~OTjS$3r;EVRtC3~aY4~TX+JT=Pf48R(VFNlk8(eT zs7*YAi`OLnk4VgQ!0%~sX)X;>oYrc;_jLtq3GsuF=p18MW2P@<^Apg>8eLumo@ z1NxREq{;ydo3tSHjBH*c8wqw$;SGe)#4r|t0jp+uKk6BWG_u-D`c*Z(7iB=;d?>9d z$uRA=RrHG~aICO5{l6RVdrH7y#UAq4XbwFW&=l+g1ZRYl#}QT=CYmroMZ-)F$2?w# z%RD~mCz;?%IT*FogR{Al^z-e|9%EiGV!BKOA&^wUzN4HL+iDJ!M2=d*bhm7EIN_w( z!2*lY{e0{QWWNk1O3lHF+-&Sn6X+S7u*i8e%DK_lp_b6Y(h{^>bC}iz#o#79hx+hI zO-Zmv3~?ZUOQH2WPNc2WHb0aph7IsI;se(y$ms8}@Tk4e94rKODldG6ZAW zxi*Vvmi{7#6s}CygI2L)t)4=ycCAR{G!o+En84JPO=3OYF+3{ObSs4m1|0{aP64R% zv-T&}KK}pJ96HB@03wuGh+Gh1(UFAUHJpNIz>iTs`SEZw+%LCx=voA8p>M~?pk(_H z4?zM9L=}a|o?yl$n0jBEz;}!3=_i%JRrqaJZwT9N$E_do>LpB_Z4iAB1ezG&Zm4e2 zyXICk0V>oHlWx)b=CE1AZF z8EtwCziws%Hh5s859T3dbWNE!L#KkP$xx7YszFZ(v7o&$J6LhhJP^k@d8V&{u0rUR z!a)RTf;FIH5L2NPNJi4EZT6h#Kbg~mEyoQ3@3HruR#E;c%@+*#^zkvH*i;=+X-dZX8BS&lF;_O&` z-0O0^n~bl*=SHTjeeLFljPCWtu}sS*pakN?9h|6Reqc8#&)0`_)R?0 zQy~#AbR=X*t1L`gik(%s4uaEnt_K<9t-uhFdNvP}16JQT#Dm$4OA6OcqlaKf(9;Sr zoB3Tto$kP7b+GfKED&u36{VimJey;7pdY9aD7SOkG+)lVnd=ct2J7_%J8a%ye28L@ z5bYn6hex`f3Fqs{5?tW0Qd|%3_q000U^;7=t;R6Dr9AmO9iPys%GcS;Y<7_Swn74( ziEj;?i1{3ydk^M^>HZiFF4~yQ!vF@E3>z3;fyKo-x$JDN@0*7I3d~50(_&1P1m(r! zAQvvnbB7uOFO7ivt$8jOLm9oum7O#pi54xY_vCG7m+4mMY3|9|xKz4KH;Q746wHV^ z1IMGwm&i!BilQ23C}5u)mbX8!m>ukVnw-LPt56PPcEU9POALyYV5e3%7+0T0!3+-# z1<=u}kUzo>7or3-3B-iORca%9SR{lQPVvUT^kv)gVb%3I;GUirdAkj?> zP9xWIy^K69a+x|r+@OW|cBXwYQl_|sQ%kdx+)*1utZ#YA2k#6+UiMXEXeWeef>!N| z@)*<(%hQi6j(o}G41+#gC`wY%}!~UfV^mzuHemdz6=#Av+WYJx)7=s?qqUG2%>R_f5O>BMe!VFs6Ytxc^ zcoU!eM%e6~Ap4@8pzKb_l;F-suYpw&tOJrF76x?4Xz1>~KR>zvZ`bp@Fa}osqhBTyf43GUEUaHZLi5g1PSs zN?l}7rJga&jQEqNiwt2q9KJ4C$8ACJiFB`MP08$s$_>}7TVw}rQWNRV&>r9xo>CKW zWyhs+9}mA5+u+DRj3J5sKIYtiJrb`u$f1?Nz4Lit7<$!s+Jw^tRq+VIcm|t`aUkUp z=};)trKU%3w}I?eF^z|iVImD8%EYtMki#U3f(s@7F>DNIA|B1#*yj7FPh{#aA-Ue! zX5(Xp=>ZQQpWd47Vd?I28PyDQ@;TyZHIZJ=LWXsE)ZedB&ql~0Usyho=~U_RZ@SA< zq)VJlHUjkPK4FqsC7x-Y-)FlTKTS=<=rB(&g4abS+q3e$sz@_4k?Fmxh#uaQ=}D=H zbd~9;*>snPITqMzB0WMP?SdH;R^%PoKqz@CPOUSH0@q{etwu1^Zk_y|u4dDR(!@A4 zv5^X((^T6J(KX$eO{$6XAiM}MPGo_ftd++mP=c)$Y9hU3dRz}ezfzwJs@9^_ZK(Rb z*{^D8bz5Fdq&H0uWpFa2RD%AlV_Z*UTCFlYt$|{Iv(e&n#lEP zJR@?ZwgQ?crg3phAyk$r?o|`%CGv1XNU;l%5FVC1#h}hsFv-aNkUTY({sLx2O=OGy zxL_v=+%5&M1ouma*mNH+gSxasO_bP}*i@D9{u|G9vlydHWZDqR;QQK~VD?7%!{~2# z2Tr672&qvDGdQ{8G<2e?50(KIbcK+cV75U!;w$tlu8W!$yQgb&w&cS?Ya-LDZOMSV zw7l<=r%7}N*bMGqH-VrZ;0iSIecD@_35KTu0LjF<>An25+V?Y@a2i_{kKec^9B0|Z@%1?6R<~X zTcU%Qw#%C@Yh<&GfFCa>AYml4m0JCuP50#6%oWaGw%9q5uzg04x)t8*ht4i=Z>O#!z$?%#_V zV5=iYABfV2RRz=Q*ix86fO}?Z-khxQa`Ml^`kS~n0Bayf-J3TB4F9$;EjYz1Cir#t z=3ed%Bw^$H)ZlyaF5l^0zFYk7vasE*-BSAel%UOlUA_l*`5xNkd)qGGv56Y;puGzJ zJA$DmOktR?`eqGjKQ*Y=;(yBwgetIZKQ$=Tx6AjyF5iRxciG-N%)_A{CPIVEt=;UDS$MBk%?N z4ix9za6^q3Tj*M8T8QZB(KeBY>&0m$H zr(gZkdP+6Un^RvK^;Z$}x5lV+%zH8f9O9!vqNF!XUi*tqqW8e`g$%(T<#~*P}u`Y@(9W~|ig^3%#_N360loc}&Qnc9s?*A2JPK#+W?Q7Kp_{}drwpct*T zbXR3`k5SnffiNyie^An8YvA98wChP@*l+$HGJq&SliR|o?Pg`C3@`y4N`yd-mWm8x4(N+k)i3k4PT;ieFk>M#9p4d;7`n6L!j=-$+-5pS*oU za?3t`@&_Z5C-3Vge>ft!=Bs}4MUtSB6zvd@@JTke; zPyR%bJ54?K>izxXPe&w=JHSsKrphwQF<1D-(_i<~m#Y!!%MbL^PaK)v{S80;q>4lN$gNOO)U8*weHU^tL?#9E-@IAl8m7i6jB`9$l zN`w`=g5-OV?8bwg-WQ~IjjF^34muk2)63}kQ6MTHOuny4?E4yqGP(s>aaSS=%O8K& zyWu(GC&XR&OUfVV#5d^}3R}9ow6Zfg2V+0bJg6(n9&l-M04UQ?9V7bIR+ zlsFJ1K3!H>G`IdA%xHt0Pqe@1T6_MV98~(i50^sz$n5*9>b4*1SuUfu$~B)BogLLj z=sX%O509^!AnnA_VcNJLEkC+wr6vcd9mT2XAazY~YDjqpXWSj)WT7W`DYqTiW{#bz|mlNu@(!XpP|v*xxenD76dH?uLbXx^;(Gw1mUv!oQG^~Sj;pr)UJ5#Xl+4727+1EXRwEESTi zSTUfbu~<=?Gn(hLG@meQ{;cM?C@LUiW-i8Lr{*@!pS4f`EyE<2nk_588N)rktV5Mh z1*F*~42&5PlN4l?bMwJbSJR=3DihwLWe~PQbZ(CQG!U5$L?!NpS`j)&Wcym$n$EC+ zKbzWB3vLk5(ol&<`+79TcBwT8IlY%D|5G4bu}7;?=c~5R=i*Vo{fpfh|Co0K`??$HNGJ zmPFuT)WqbYwiwAnfG#PJBx2r0CG9YsZxrJ6Viq3T5X8r!aH9*P0Zcw^19*s#e4>Hb zf{?$1>F-$^OR^*>olr>u)=mj_*1={+th_kQ9cJB!Jq@(92jO6P*+CiVkY?yrO#0V6 zrsI-nmIp0Kp0`A~tuQvj2pm|_?d#DA8z`HELYL+z*;$$9PVO)d+W`wBk_VV&WV?^m zl7U_;uE?0i$Fg~ZthhgGyP0APcrKfSI-hPbwrZW<9@+}$?XD~^?XVqGRS+#Z%=>OL z5FNy$T%NAW=BYU;#i_L?R8*WAqV}I~x#A9}9nNaAl}T@y68Xqc<@87^64nudy*>iO z0@%uAS}UqdXrWeN(U!sb=ZRoj4MvSWf%_8X#W4?|{dZ8QM`nh8m8_@3J!-QNJO}~g znSKdvx647gnV~k@Mr7k-ZN%2DXJ|M5d1YrUReChbR*JR*DmS4_>VU=qqj+C{hBM0$ zlp{zwMUvh{35a!IH=L+emF7dvS;UPdM&mUA$ZQs z<#pUR>yAwu@%v z33@nl4pr2<#a|$gNqs$7X4rs#ESLj1yt5BbI7=X@50)9?m=fWrvrq*Av=p(_iVWU7( z^0c2fopQ-(G&|27v?TO1t~oEn(Cd}fG0LujJ-xLLJ1Y{ZjOm`%4XTQ+gaVd3lu75q zY!pThFD)bXJ&2y5l!a3qJLuu%He0mRrnWb+ z)@rdlCTcUU0j%nRJHo zECkRPyo;(b$Sv&jfUav$aQJmdlomz~TNU9a^t)K+xKecCGMj}?yreSH zR4+T$Moj0caY}D?=iR5qty&Ln;3GjX@)0$k8B;434^ag|W zB#7|Gpmt0va!EWal3{Ak*_?v*{S3o{)jQ5GyMdRSSxg7L2fsQt52NqP zJi@3*K9A6!lc5kd52J(iXQTglfc`w%|J+Z1o-Hg!Umd`DG-$Ve2Rl}RcE*N5{@>2R z3~m~jAo}(&HRd}#%Gz#yTHbw~mfk8~ZoNwfdDMsq42wc?HrNwl&5JRh%__bgG^B!^ zX)aSuMovJBtPG-w)v-fA8m3?78{OJO96Z9|QUIflvomG-r*&W;m)f<7DG8p&E>dMl zPy*pBV5wGDn?Y4Poqyz3mwsn6ReO|5;@`zK&9xhtay?iKd1FrIe;qrpuY#SbgY~8V z-T;=CE&67GX{JpFfK1WBUIR_D=?6H4X|7$5DAV>YzZP3dw6qBhb$m1gH@zrr*;n^WmEAch%xwAh|Tl^MLFf+4_26HH#KFsf;78s}2lrfR^OW}0SO zeRP~n`+GJTx;y57}!QzAn#?-r)!eB5uIC#~Ktk){_axkl+(l1*3bKA>Gd%D$q&?@~G zQm(`4{9TW-d77W6ntqI^aX0#fOXxt& zJ6IKR{KLi{#%MBFpP$?e196v~dRHzb33lM2LZ&CAns90^%~Q^v259_qbNNkv2NvmA z7LE1V?Q#l+ZZ5~0q1Gq$pt#8y;w|7X0#$WzN{sw1xsy;^4}u2Z2?vhXvuo{3IXiwG z?krZX9T?Y)yG8dU`fbs!7g=5mMVXcyig%cZ8wQusyV|f1L(`$ITf=6CaeCet7QDqT zh+yOYYC!7t>wk_iJG0>Eqj%M2Zwmx{yQ`MI>>o3bjvwhuPVb9BK@KhT|7U*%rpIHL z#NYmNCUMTpIWwB(?K+JWwm)XBj3*$V7AZA;BpTcXHx89)plgiBp&pC}pZEhsYCI87 zyPg|&)2P-cRor-@!F6?m_&^S1)R^fqL5=_WPaD;o=7lqN9adkhKdk*LHMl+i7=ZlI z-@(5CHu@73(z)^(*Cr2fmz3wL7a1k5~^A13sEOV^X+!iJujAZ{$ zsLrd2>mioW5B8MZ;%C!6G10fn?87p_PcNefVVMwjU9(baeh0#w$B;AP4~k(g#!AiV zM9}XmVX*+@$zk3*eqPY{@-Sz+&ItzJ7iQ$ebRo6Ve?uyp4i+UqF{HJ`>ym5z&sG9e<11M zK(~b%pN!0?{z(8to-l3lCzzgen0c~f2Hh+V(@&H1utHy$agJmJ6$Zof3rD1H3)3%^ z^su4oaRFF=7@09S%(z}>M2H^d+#oqY-`!#QEs`GgF&JhHjLeArYtZ@qBQt8kjE6^N zq{EEIM`m<|880IP6p_e_FF>7^u;kNWN!Q)2 zU7v7l3rn9;610of@K4mVPmEfA8UXR?u7!j42ErmAp@<}R$no;4f6nnTzv+b3E;f-v zexI~=Wg?zV1dAK<4f|9rVnE(H*L2R7jS~Cbe<&qZB(-r+RW3h#e zO$&nLnTzMooF^(iip5S?GG~6{bm;pi7W0SsBmZ;mtYh&@6_+w&u~|nq&XBguSgfUK z!7;Ofwt}v;{&NQEqKEm7GiLk0i^qbbl2~l!2{Rfs#HQ9P4jij7T@encvW`yN;h*q52;a41-+_S?<9~j|b-x8)W?Jz0blze)Njl75M#KPw0VtUec-P^`-v2DZsLS)2BfnI1Ge zua64t1lsH;QkvpuYTLR-ePK8yE3K!Vv} z^9OW87XIcW(RfhLrawcIkBPV$UV?z;fh@wE3kAaVNT%gC#G;4=M)P*+Pzej{8QAP1 zqI0M@Kb7gg0y}i3UW3HcJpg(}Hc4+|=IOL-UVd2>bgDQ7(>q9Kb1a>a4PqCP4%tj+ zWNmsgfb;AC&e!(@Cv|3R`oPG10dPC$=b@sZVqu6&Vm7@>A0iL70Sj(3vGVs6{T`6t zcS?Hr+ZT*NV2NxG!oQ2Ew8*oWTxsqM~bC-v;k+VtQCwTrWQ8`qOW8HB^yp zK%Ri5ufTnp1wv`{DZu=;VPmnpb@Bfr3QTqXbIChtq}L`dJJs@bSw{+%Q^HIzVS*a z=0p)Tzf!bL&xRs|x>Aa62)`c)*6STGEdT#$o-N1UQhi~+j|iOoSsS`%eI)9n5ugI! z@Ro%?U90_X^Pmse5v71o#!1*3^k}RAGI@>6Bk6zdl{Dx(K*n&=L*ER=#Wb*Rppz-9 zzH_B32NoV$GRJ)^gijV=Dik@WMG2oPX+{d#LdV5_5#ipU6Y}*;2tUl7O5sgMf`(Yy zI!@Hn1w_xMlYtvdqcJ`-23-`j^uSIE!S#`^hz`Z>`fu53-+^ls;PoZUPfLU*j{96YR%LVPEmi%6dsHhwvnXrFpb!+8!3u_|Huc0nmw4Sb9N zc9=i-1gl|DdMCpg@*pp8v~D>mZ0|b1J$3~KurNJxLywi3gHf@Yr5H$OqMdPeeT;v- zPXdm;FCVseuM~y)KT%wEERH(V@C65hE5aJjq6T+}cC8NUQkuaof%IM}`d3?(fQ%t3 z45$mx5@#N!3fv?joKa8CM%il2_H-L2CWqo@Qkc52`lqnMG#D$;YqB%x92~S7AIrDG zO7OAl26hlcWQ0LSdSwWyFz?*KX>G+J_F|Zn4g`I{(!quXxF%LFb%Abja!1(0vtV4& z0;;w9#eA*gP*m(SU2HvLp5>7N#BH-u@+1~;-9mS%23nL4)PUPM!%ofulZXMy04)s_ zXba}d55m0^m`9zObbJu=8rT8lNijlfpqzvzMzm$ANiCWnw}G>NQHXX4(@VI)u5YvH zA=B%089|}xMH7)FFe7H|FwL-oN3JT?X!y!HcoW!i-emrV7HD@zIm6iIXJB4otNUha z5)B1m5ULAd3szCcg03iemxilP))kbZn}qZm44yCWwy62X+2j0di2|V*O?>8ELm8L` zMf7JGy;vIG0Om%Nrm<3UC~B-2_Q|Jc2#fpJ<9$RKKELyL)WLPd9c+tQD()ctd)(jY#e_{b(fVF5OF+IR69FdS&3Jjfus~*qi?= z+1kRHb2H8JjyR^+M?-uo0&{=ZtA7~@)@L?2Kd@QrHMDlJDJj>2W2-%wjqHWZKyMKykpGqsbQ>2m6U zt%Izsj~#cToLJ`7#s}=Qf5IJgAy@G2u)mF{&7E?uwKmHE;J|!LG0eAYH5ws$U6$`# zAl%WBc(y_~iBB=A;wRZvquHXL6d-lNIbwc89^KLRJ4YIzpXS>+rAAYOXD;_WVV55g zLh{m{Ai?p3YdXNetL&p;5g6CtT!a1r&JzO8_4q6T#}d{VHeUcHjux)X`4Bu5^UPL# zB(;1egy8Q6%<($dRTf=>3w)m%P0ttw`RoBK)6m_sI0}DdQypyTJk989lR5LK?HaU~ zG~4m@0?mZba5S75`q_D^7QBFCPN7?JJL$fd2a6*=1n)iY>$oLnC!&WWv>c7>3_Ey; zl>R+WZFz8lny&<3e=Vkbn3mcheg0lJoOTngF1@m>%3_NK!1TZ_%u%DwgXs0x4rO6h z>u{A&qb&&)1>q?8n{_$6Zg*G)Vg;r5^h)5hu4%d4z7qZToO8<$!y4UD4cQCu z3tD-7QQ4X!!m=wwR!2sG!Sw)h5x5mM7wbB9rmxUj=#cHcohm{-O8BxBQ#nOhTh6{% zsX0&+ws>aPqDGCNO>nR>W_S(Bz*pfdLqA|k_mN=(mr4VxxlS;j5u8!V*TLfQ7~9o% zh+3|15I8&8sg?UvuvcV`aHq8DWk1t=yPhgMQW+>Q01kq0;f`*tVu$8#2y@$PuAAnH z7{{t$YO%!-@qmKMl+Du#dCKNja))jDrf)-b2=jx0S?xvJTl2v70(l6Jbt^39)n1Gs zUI>VvLf)z1`QYDS^s_C3B`wqtfauW}3@I(z8Y@6h2h&g4{O7cR7QKpSUJkF*Sul|U zs^(zNyq2limUptiD`DW+@y|jECCHQD4)lHzM9zjsOOx;tI>x4bGkW6SZDmuB;n7zL z@Cp(?rcQQX9gHiEj(F6YtemNt>*qP??B7%1oIhTP6ixr5F(!mC$Ty+1ol>ph=774$9N4i1MdCu~RFVdq|<8lb6c)_!9COw+Nn z8y0&4-F-%}ns`|3Z4?74U=pG71p?u>G%Z^I@@}D^^#Bj&O7It}h4T0>Y(-h$KydYw zb}aZSSu|q<K92Ymo+^LI#v7nXnY ziz)i}#uVC=(b*B91f4VmdjSRHL;h9gn*ROh|9fZou;~bwD^= zPI1zr4SB9(mp0yQLmP-&IkJs6+BPI2xxgw3mAtaR9n>27j&%RnA7wdSN-pc|l-Z=C zlAN+!rXOXvzRne*1h16QOshoJk2|BQShQcGS&qidsvIwLtI<%5=_3QU`?b2C-3xU#;HN#2T~v>paI*Iek;`OhjjCO~uWhe-eKpUKi)J^gp(% z%6;V;aL}y7FwB43&d!r{YBZ(u)=KVB<#a%(E3O&;O|Y^tRW2`K_wqW7h0}T+;lM?I z)G&|AX!>$BK2-$Ww#DIA{Xx;603T z^ANht*g>N`Iw6UFr|$&2gm=!dHG1JJXwepPj|f!`67?4y^#lj-A=&ciKVJlmiWGmj!`$GX}l#Q%1@&vIQX zOxT`w27qLG8bpAumr~gL%p*33E;NXK^o(m)V6KZ5uIyDS_=Xh%;u(>+E; zS$-Vh6ab_943F0^!P%kGV+8AVXum=|!^2avhq9-EitovR_}TNY(gcx!mO?7-c^-Dk zJb;EafTACtrvp7&lw{~S^pr3Tbn%jvLm}7<^Z;1KU)v(ZJtm10+F^%|Niu%I6%af! z9Vn=WV1);47i7RR!~gKGt#xs@O@bA{L?HX9kH>CiEbkvCw$ zW-w5gK5oOwq^Hw7gibVfD2qOA>)|^7|FQSy@o`jF-Z)%cr*u?WYzu5iAdr|OBq76M zQj)49Rc4$?X0l8c=SebI+IF|3mO;U`ge31$x1nvfY#MM_5<*x^2uX;HCEE&Ha4`E0 zHUhjLz(5EPmYB_8%UJK{dv4Y3mMnu6=gl9_@5ziz_wBo!d+xdCo_n?{AY&h-&3@1? z_ka*vaKdA(7{|!+E2O-x-uE19k>ew4{63BdsS8Q1WHHt;2WhQQ2n{IH`%vs={9;U9Nj+5Dxty1AYGn}+Ew(Tl<)0`ZnH{g zvDpuDd@c~4Y^0%IZk5myGfU0UWm)`qaT0isVG{QV0yZb5?e8PFasED)eqR{k?S|%v ztA6l_kR!s>-_4yu7^W}xPaC3}>;wk)DrMUIA-myP*$q3SmZw-~{*ei6vq+I3bC4e- zff##;{)O>7WKeHL5PeZf4~Vc(g>}PXV{e#Wd*3R7#mg$Chmy|YwY)7j3%1EhxHw>J zdMFviD;P^s0ZQoCW+FV(cKk@s`gytS;08l-Uq{YiAke1Aj09%fm2%j|`TaOD9HV#n z%K&PO#Pny(7W-bQdrMhzWtY(Hrtd-N5Ge9D{#XzCnq)y(C$p+$2}?fBVL}{n95`Yz zT3aRbZp5#t1YZ*=3#$<@xK(&5I@pm+&zN9vR{#zm)heN9O}tLvvsiwDn{_GE)l%ia z%xsG+-#V#<*^S859~8xUz|3&|P%;YA2V(G=6fCkrDD?Z_x&W%7YfK+;ec z>kXu`-4zMFehw`3(S6Wz>J*iiXNAk*dBwZG30397s-O)f*l}|juNt9JdZKt z6__wfnUL(Pq20F>amr?{1>Mp8roIBjZFuupBQX`$KmpD>-UXn8yIJu?yN*m~;q z+;7I*nXD|{=deoYaU547l66I7heE!R9l9D!L*NC>*r9j@tCZ<4z{&{Aa~UqvvZOJ* zMiTDSorz%ZLi#LpTHQcZt?X79eJ}*9iEPUWiCvk&!=)ZdGv?1CyDFioh^d?Cb~$>r ziL2~1-FL;XzJYA?4LxEvguVm!Y(v6y*9Bb%I-H4(y6eA5Xsp032eRr`Uf!(6e=d&f z!iC;Em8?B>=Q7VKM0BReXyb7u!1ZOS(tkkk20A<0g;xTdmgZPRXo=f&h92@w+!9%C z#4%}v2HRm!2b~gBaAF$=r@&8m=5iULDKixs?=y98J-BM18Xk`^aY4Z*$%qvWVsAN! zuh2Y>-W4@mXUee(**qR!;bz0`ia>FPHum8X+Kc&s{lwT^fw1%w zupOJ}J?uiOknlo;jc@W9a?5)im=1NyKCHpbpc+_8j8 zkm-EcjeDdPaJN5aFJxLG145RhnQ1yMTZ7y=LTYrT!}Jl{5pe)9L-4l>;aH;ZfTFG> ziXMyjaHWMKY`a6pC#pcCRb0bMvXIm00pS;}Rme77X~I2hY%kN?DC4MtVrjkuJ~uy9 zD6UoW+u6~D!v%OtIS{T~w!5uD8jSenR3ic{;-pr+V90-oo9Z&~#vkBb7|1W{!3jh- z8-YI-GTlQEE5uLtz-YQf*aozJpG?5B0|#^syJppHhdry5&P%2_CKG&3ai|NDB1064 z6Gt_KY=T*WK9iMu5eRxIyNJ`pP6IEIL)2Bb*}(OrluGT=id>bJcJFf;% ziZ*d+6ZRl&+TpE3BG~Id> zGND-XiC7@`1LwldgUZcGU3tZ)4v3y=r7QivE<%6tS$g)*{C{uuE0^fm-}LD`LfHb2 zf0)NZ@)Jy28dN6wxU&4(|HS+RJ1xU7jT0k>67i&r-=lF5(1An$j2NkrGAt^PV4RN_ z#=MqVNpuJwgCO$A%vt!!Fn-#k;)5c2K}h*QM3f%_M9QqcL%X?H7QJu6u3x~kIvKSr zdPI&N>V!r`oS5(d?^K8{(3?pug)jb@@|Cena@E?hIT7Zqec>gc>;hfmHX~1sJ~1pO z`i$ZpJxWSR!#!!U*~{Ymw>k+Qt>N&jlbbc<4OA)`kuAC_q4K1yUf39cnq}&lQ z+*558(rS_XXlc^%tOB?P7K(dd0o(%%;T~APa1X>sh%8vR@20Iny1<0b;B6ApuvS#l zg-m~pBq~X{?81MS3MB2JyG%p~mqqqbk}J+u^tB-wwF>Ea!4$+#9N4Z!=opAmy3s^% zu`rK=qpyuSjlvQYLY)@F%>t6e9T6Y?f^?tJ&l!vG-jEU=fd@~?ieu}BpDSs`gV1PQZlC*PDz{NJ+ zT!Lm`Kqc7iTf`Cn-(j;dzewgD%cMVvcF~4bz?`@@bn@NwC*5$&gyIam9g)Wd3xouV z*C(T9@ThY0EkR9-PQnlw@Bym;3S{V)TwV_yzh_w()T`!Lw1(l&g1hKik-WFmghEq% zgYK(5=xff0^+m9Ag}u#5^A=+5sXgV-y1$yBzu6z!-`G6pW~09`s=qH1$m*s;I!X`d zkJ$Q;Cy!QV;b;i-fKY_KgmsOH#(zqO+UIXX!p1Nyb7nRKWUMBFFDBgY5|Tui)JIf1 z6$zCm;r9qK3#;tLWEe<`JXfVsCb<2n?8fp#USeWPo$FIR7KmFWT^OoFwunLxX&A-U z-f1C^9Znm=e0P0r3tKzv!Fum z$6MJYTw=j&VU%riXs%NUb;nyBlE@7AY?JA76QYSHwpIuacv%(ZblF^;V(t*nCuH&y z1;=g!3RH=QR%6oqE*$1-aYByoMFd6{otcIU$0Q>Xc}Zl^|Nj54hj85}Nw>afc3W-Z z?1q+8W;V9|OHz=-dv+8-{H;wbqXKJ`hgIEzwuaUbcW0PC9Afo{kS&JC|rWUDEpiei(P7iX3TZSRfqLV_kn-`DQ+eHLy zP#*+k$Pad=+zdR2j0w7$cn*!Mov9vj8HY?KQLVC7eiycp5>J2(8>Q@`+!#W$edk^|k^!|-#7@MLxSO2C&S%e2! znZAq9k23=IUo{3Tlm078$+Sgq>$FVjPDL%EO$I(+oQlF>GVt{4`H05gd5SN|G@;wV zAB5fgQV9Zsp9o>A*RUygMuR_OE|&>{gZ-1S2g|#u-u;)=RX9`sDoFRk+)0C$8B=JB z08_Jpw8QY}`+_nM-R21p7Op=eVWIU9mpPYJKOQn!0NY*#q6-MlE-LAy5Hu6dPi_g3zmw8gp7w{xW{ z!%~yp!5FkCt-R-Gkz+A^4F(~T$yE}8_|*xRim`FrvS2idu~|Xp9+o@;%=8pA@zM@g zW~LHf#0;3=6}ljY*!5h`aIXmxO~h~uL8BW|PC0J44;wzdu1!U&xm@1+b;amL?O-}k zF+2{I-P|W)2oU}2VB#h{1#U(c3SXvI)vF5^riM)kxkCgX1sH&GJ;WQo3PY}%6Ck)S z;$v2K51Z94@PBQ@q%+dW*zkHXYMHPAbF~-)**xjpJW8=)6|atS#$vq8FD%KmOi3k^ zVVmAeB6BosKvN*AZj2<@HBa5cX|b#0B*ZD$$z&(^eanJdhQ{1P$(CJ*+qaj2YUcEv}S<_wsDs(k0~^iaQsC`%6~)3jC;DfDo1*iv92)o;Ve>$==p9FV^LD7_A8 z37u>iyOnW%lIuM#w=8j<)u`HH_>~~C)wmHb^+Dzp*ETp8fTn^Ivg{9f7JtSpzyf>+ zqBYZ79bHfySte{#m;$iS7v33qaxEVYTgE|0uOl};-op1ij#*64cSO+**3~lUVidc) znqBS+%}`UiG!XshiY`i~N#&a`Cz0f!K*xEzHTX3&+(Qi@(ZTNn(UE@3Qcf;88o$I8 z8ov!@SkQd1b9V`5;W>vbPp>4061kbcxSQ03x=qQ=#f9^_qz7k(0O{ z05EPOdJo*(0r3TMzsFSiu~#Hrt=sw=6BH@f z+W5=h83rHrZ4fRTJ!5S5U>@fpSq9TP3gzdUh*gp%_7>=9uI=apA;LF-`hYf;knblc zkqa?ftEY{36osN?yC9N+ZQ9Td{&hPEJ#5eb-Z2b7k5cI+;c>Zfo$eE_7Q#OQgwc!E zfZFvvx&Z2Rr$$(q@`$!6)d!X%w^4v?e(61$ukXT>dv&MhTF~<>=Sr`}V!A5qc02hX zk~ISHW|BQ9ktgka`tS`qvFITjaM<2v#qJdzUgC#HhbY=V1_N322I?@J?kd@(&-Kzg zryLJ=h_2KoD~pi-g_7;t_L26x#i395V8T@Fdt8a5j^;a+cy@%vAkAKtGSgxMNhjUhK_wsn zhy_UguI}{a#C`12cprg2z^W%tg;bO6BMBJO$9+#3?uBEzTuotWe-}!9z$01eugj?= zrrfA8B(W;+ty5aVTEwphc-)k$^$gBm@JcXckpLux3ltVEuSK@)MVMPwv=S-|E9M~> zX>QsK&W{~81!ES{i4%%5lCgrTs_4vg$YCeltZUqcb+gMX3$t{sLUR($L#(FlTZJ@F ztSjS$pg)(e3hArBJ9sbCMq|(_go`}1nlboExzpu7CNa=kZ0c~r;a2@g4|kU_!D8}p z3o2y#vW`dZ$yK=n5Lp&oB$_4js9i8-yqZ(6PPrxMt%#mNRh-1Y6blt&-4G_@^70yL zPA?5lg0c;|)%(&4D*=}yw6MK$a3N7FxpJgXM;Ck^-D;Rd)LTwmML*5)f?Ri(V#PQs zW`<5gcfgywz!%D|q%=H>;VumIy-R0DvP}Dg+j@XNhUaQ}M?@kzCnbc3`G8ufJzZa` zoUMZF^q{y!4*lKmnO2I{Db6mvjwdZHkgs5Z9=0K}e7H>Ubclh)w`vUB36P=As=?uj z3u`5^L-JGNSmkUzBo@AF*#I#gSyE+d=yOMZ4!d2EB zo6KITDDaN9<8m+dBpVS<>zMfiGWS4gGLL#s2+0Kp$FmP=3&=germ8`vFFEYmRM1<& z8Jo^@LI&D%y9VlM%*wJzJllnCit0kG{FFO%HIb>XyS$}lkoI$B_-irUFT{~y0Trr2 za3|$ggM{>T9^tT#vHfI_ajlp|O(6KQirGf~H6VVz{JSX4?sCpxGeG4BS^Lfal45b` z%d7xDI#Hc;>IPC*HpH)M+%QbNSkpvfqngem9+8!GT? z82(!zxR?j`RGNVlA{K^i)5C7OM-Tf1hUMkF%!9)shRv{RdzH8YE)$_;dlHYe{Z`!? zwm}L4dao7ZAAF|ZtOjs0>V&{axhWa7ieYBN7YTc?im6A)Q3aN588*f}VYfG_VbtZM znCWKh)P-G?OyejGc<+Z1OZ4|wBX}_c_)7Ls zo96k(0|5jjVwb{k?kx%3G$)5YmMWrzB1g_VczN0M@xiJDj;kO?<;Gm?-PkVJtM*|a zLfmqGh&awtw&|onkZi090z+U9So~lBL@9LwofyYbhlqJ9>~AdGPL2BSgNlg(317i| z8lBekf^cG>!xqhrS3RV=y-m8MtQE#pm9lvpy=6d1s0D~zQ^_?P6&z=9io%D~1>+*1 z*Y2sdN~1Je<89-^LEmHOvp7vH_&l-*tc*}z8}WcdBEhdyYx9heu*T5u!AdQr7r<8+ z0RAzI;4^%t!fXxzuB=8@0L0ZS5QDemQKU1YA?M!thz8-&5Z~U?aP1~-(!XxOuUW4O znGkV8hY;SH8i1c4Lwf?LndxMlz+>>X958uinryMRQ={hg#^#{gA{q#A%CR2yDgu+e z4Y7Nj2Y}_c!atyW@B^GbF*d)ui)Lz}bmgO}zhVU9LI+`fWiZ@jllzAM=40(MA=&H< ze#h~k_*|IBW#VJ1C4{PSwCQI3+e-agC}-^qN(2Ub3l1qagkkaHxt@+ePYAsVUWVQj z{x%N%<4?x_V=Y%G)mzw)%oNrU}0F6tiFEwG?6Mc%MX(E12!}3AR4z6ej@tx+dE0|zkkRgx+P$}0c z%ox)EQ~@Ggw%66XiK{)5vfc_;|2&UG&I+F(3r9>B$zlDH`iL$8eGqt60k#?EhFnZ~{R2m4t`t;x~?lSF@VkcMJXNYV680csIuG9l` z7Ks9AA6^D!I$M2vUwvCjKFfosOQ7(AItKwrxHG{ofZhjiXQv!I!iSXr@d*JX;ZTil zfF>sKbQ1V5AS&kRD#N#pCr#z0%RpG(@vO$&HYC(w`h1q@e{nO3gJ5HFO_qcV`>No% zUc-tCV)ABOZvBx?`qRoKVdrzY&*x<|XksrTQD96A^>!J33`bihg#5$U-LK;q1oIc# zamnf7N;_T|w@kW<29dw7(vxf{5-uOcn0*Ve3g$_>bD6vSV~E?u7LTEZFF3Bvo);+Hj2o)VQ&hN~Ffl2|h6JMaWphK_}ZguF8dK!(KP zF>S_@e^tIACq^}jhkQW%Iv9)VF&=G{k8!3S1WlllK|m?pAknty@)30TNB#9aeEnWf zJAjWBB!Y^%Vf+53YKnyA2;HhVWky<(ia(?W1~o0?jW; z^^shX>2q0jxv~l(js#U*nL^U#7@dNaaYp`6x+>L&r>65O(&Q-9sXl$%@#s`krEiIB z-q1%c$(=v1OAPkR@WK)LR0Exi-r`K3_tp2|EwRV*IzM1PTY?ce2EGFkwU53K^!=Fw z^i8*j{U5~9BS}-~+Em|uJ^ct6iZi)B%iQ*x@Rx%MH3zD2s~|bf$aX*C3`Usb#{<}_ z4+Q%asi+xT?~FGz*x)3fP)-1OmPzMHU24zWW!}^v?E*+R*<8lckHZxzk2=Gv!IZaDU)c9pt57%|!w_6AJ0xqo7F{20igp`m`L7&HIIY(#iSJxU?))eDU`=VWW%{`E4foCVDTiT zQDjbjUw6M#0Lt|6KK2+20nqAQ$pOoh6nO5^W!zmEwwtt1w|fEYtOA+~Sr)v}GN~XD ziYA+K6&FKwM)JP?Ad9>Q=>eP($FghhtzIM2W;a*bozq3%1ARf9GYGAPu}4L{juR74 zEp8w+)COtN10w~;o^Hq4JZ`rvm%x8jQeI2n97$A{S52*hxr%9ygO@xOJgtFn?4$x~*~6sD_x`Fskk`AAr={($v;8nyIudXsxfEIde2k>RT2x zw>6#8Hgi;d>p`pe6_ct)KiFtmoL<{{`e^Dmj;i-rwQZ-LQr9>t2}0Vm)Xq+f2CPwf z8t0++l;$%U=Z|J6olUs5b@r*FX$Y#D(=saJX`J2G&@#JrCfbgA-vtqgqXAKm2kqP` zP0ekmoZd96p{}K2-YH*hm^rf+39C;z_2e%&Czqcxv$3vWenb5!tu6JZG`2O2{92j- z4M`}1aD~~}bod52s&SV5K!&lfX?8>7>={k?g$!dxQ_HMcsiwcgT4vNcj#CxXn5{gm zEyGX`wV}AcFdCa%YUi~!srSz;e8&&+A^0KI+|=4QKPU}0b8OCRRm6s|5YCyyURFJ= z3BRD*+C-vdNDuW*vt~8TMvX$lz+S~i+b|S);Y*QWV29yjv0*eySqU;c%lt|WW44-U zsbSQ%G`2~_Qo{iGqsFMyi0N3$3Mec;$$MLxo*?Aw2FpOEXXJ}Vj{1QH}Hz7YZc7h>N`o_bRGq{|% z5ID?l!hoR67dS{w`(Qe=&XlQgCV;MP^TM&|4-!o6{Lst!tU)cZmVp ztz}vWm0XyZ4>}D9Q{A{y_ll5;yEPtya;Y?nKvWf(X;Trz+4uolG!L+gl(D$F!w?J~ zmp4Sd*mFph667y$X=rUd<+O#3&7&(r42Uf3u+Dq;)|Y(u(z&@Lj>0;XgQ<@Z)Tzt>;# z`91jeY(XJ<0-F~u%7y`IL0QBw5)Jt^PMDuch$&xjmZTemsW*E2^S0=3S^mPXo8H(Z3ydt=NWZ+)u zDho)b+>>1Ppn!r72`Ow5C3A#$62jnPE{(_)JK4pqz112|UxP-W()TLSKb#vYao>%G zF1|ZW#rG_eGPuGa<}SlExM$&pglwevZ#BEC>$ry^emT41X*UmOnodKH-Wpt zPD_LNdt88fVDX0@TpVlGNH{*m-kfycw}o4uO;@MVmPtydin#8+LU`EH#by?+#&ZE? z9k3~=REcfy*(PzDGCv28X7eQa=pIa@8jfVJp|v~VR}W6p$TfL?GR-ksgT$DdlhB7! zcchNm0#4ZT-9`qmU*RYHAF^?1Lk99ar1yvoJ z1PE_v==~^Eee|poJn1_yRBkq)dKAOF6D^!1#7k$Itk>)^H^N&HY-={H4&fd zK?~LF+EwN;G#fj`WjYe8f}Mj9t{77X%H?qjS_SgkZ=JUau9UHAsaof)vd(~_y^mvX zXbT+un*-^`GU+ntXSkaO<vkrgldUC)$$(i7J)gJy=RbA!f z{@*h#BNBP!D8n#~mDZssID)IdXb!~2@M<>7tJ1$*LB61^p?2n|Zs2-q5IskOlhKP~ zdMJR|)PhI1M|15A*Py9Yh8|HqX_!4%U-1x)*EFZCAy(Ha_mT*FLqw&McQuI^1|spA z>v6oBhOsc9j-X0Frl&QbV)#>dhs^6RH-xqm;I1ox?tjHZIC1lgyhaDOisl9gAV@T} zG^pET5JiY`$Xh{vAo#`1%ZC7$_SnqEx_S|1AhXSHsL!9kjG0aKXXHULqp9VL=0^2u zI9@AKmv5kQX5%cKuS0zaa0D1oL2#YB^a1#$`ZMZhwVswgV2DIjaVUQCtH+w!PHzxV z4uRT@vl?Qvn`YFWAu|q2TH9LY)Qj*IQ42SF&a9F84LpcMwoEk42ld5f*S0mzmChh! zwl&GEi-*a8s3~-cDcboz7HJ10{6tDD?BP%?Ynk*B_!knJ6KURnIxj7lbZn;A2q&ox zxh=x9XOWJ%Y%WyoEt8H>Mj}XwF%cnLc6%!0aCwX|6%Hy>1>2PD2F|LCuQy;P`}qKbz+1=+G%88n(M(&9V-o z&ktHB$I$0xC~-Ix;XtHxaY~`2&QZE@Y|yn$WfO|K;oybY#5}7|#95F5$Vc)T zNm^Ls5f^+5_|t&#C@b4C9!@G%eljb^>3%Q6ixvt3c*H=&zznfK9&7JPqf9rr_JbM- z>O?{?ip<5iMmIzzU;*l2QWMv(xHtpb6BzlMVi-1TWgMo5x=V%vC;}Tk9O$o;-SE=2 zIM;z#*%ay`bwI$3r*C78ptsWpn9tCk0bNepBoj@M;)BZV2TZ0d;S zR{^eU`ksR;7(S!fNPqaU08p~XjPP~~wX85Ghj0z&kdz>f_wfT-#+vI)a}7Wi(( zKN9p|(I){nVw-TA6p|8{T~DOs;OSt@1d*Gz+z*KQr*ejx^n>7wKdQUi0CPZ$zX8#` zQ)UGRr5MMThP1LHto1QE0UHv~*=wsfHVuaCT8`1ebSMn3JSyn1kX{0h!Fm9tTk+2d zAdot`!i(v|qjhCTwG5g8hsO3DWOsRG07@(@IY&IhQM!HTZ<09;%_Zohqc6nQ&X4I4 z&XRp1>HmQhhg}hjjUsc>F}l)Hsf6F%-0xFzzpJOmGPT0JU@v9lwgD-CfdvNg*&$G`S~G;AZjfRdVo%Ra6ft-~OlKrsH~!ljwx;#{@YKqLcu8Sh zDAi>KK1VxqgKZu%SWPh2pTn{6cp$o&jIh~4rY}M?V5a{JzTX*+^QV!<*)dU1>KQdo zjJI(GIyyNR36Ch@`=p@b?*tv69DM(7@Vx@*9+|!otrFh`xJ6J>2U%}AP7ky~?IM@$ zxK)VA0zuxqD3bW8gyDc6Q4Sb345ffi6uy_pFidUGVTD)=cK*gs>OsUoM3xAHbRdGk zLZG%)$V+4;pm-<*Mank_A%(nLq84%14ePk)Q@Z14K_QG++n#H8L_Y&)uO#0ZL(hwE zEHhnBS+105Y@v|uV0^mYgqT*yw7x@)D~~DwC|VnqJZvgp=tfD-%&a1`Av?LiD7)4e zdbYibGZxdOA=2PDop z#0T{BMv=KAH`c?PzZdRV7!0E`anJ4i@Z% zO`p*dd!SGTi!mm4|AJtwZls6=Rv%svjI}mcPS}i)KaI9crJ)hNM z?Xz)sV4?4t=kRn-=!9;ueQONeJ zC9j4hm(CXq&I^`vzMtcG(FU&P_!ECls3rwrdp>8|MX3{K*cW zxP4xLn++my7cy-OOKuHI?txy?V!F{5D3r#M=68i8Hu3X%)a#37RD=g*=OP{z6MVY} z$T@FrfSfmrG;+$ULfSqzfa%Jj0H&ws1~6S!6kw)5EV()?d48^7hUqfAdLAGqsSuN$ zt9HTPi&VBjIK%y$2Jt?Ch+Pmcw1>j+*M#Fg5{}jrj`mPkvMwxnOpQ-nMfo_%F}5iu z>H%La5$bf+re8fLnC?Tx!E}4(1k-(_IGFC5uw-*s^2a&Bbnh=7VJl2NNe{NCR4>52 zt>FUPSc(O>AMgKTJ+3buvH(AC3m|&FB!KABwg93RN&<*32}@oKORj|P6MEZGlHg$r zaBr(xfc2$%0T!REL0mOPKpaY$J6i)F&M6ImcoP|_Fu=K`0T4UGlJmoocUuK5)LELp z3OUMLd5Ru&8%D*lgGbyFjxbfMPq~KibtNYu)vr}Z9}lFY zXdu3PGLVX<1ya$(;5)JugXw-Y5MSiMyYTP{9QHO==t1`CL3*r0JVProQe!a4grL(i zf(dDWGPs;H$o}oyif}HdswnEEZNl_tlM%?bcGlU zw&p5Nk9EJ=nlXJOJ{a!K?r?WL9PY)n;a+?&ELk6x^r_uRtA{Y2WS99hLmnU5Sfxk2 z06PU2R(3_SK&G5RdOzGTd&6PR3J2L24l)?-n8o3aS*C_%+MQzx3%bHrYd898(BcgI zbU})^f#vQ42eQ--JB=c9%D<~7#ThuA!ta^sZ6g6C z;yI}(R{m`X!g3*ev1@yH3P*04xOWcwidM^1t38H4q^1=?EOc66>Nck+Bu<{J2Oq>x zXDPE^4Xj!$z3maxU3W;tJBKw;sU=3JtJ0^&56Pz_&%87Wi^9LiIs@QF1J>vqLbO># zE2~E}R%leIZ=N%ncFM3Zq6wxR<~?Y|;RT4#vC|r6H?%a0g(c5y0k^=wY(aTB6wRCy z9TfKBc2O|YXsE0)dcOG1hEd-<2cHU!Jiioscp%2g_mt;`r2Zb4?PO$!AxpC0$Cg9y zeLJG<740LRDBb}T$8zW-@ZeDdZv=&!pDDXlca}*n;yAF1=w@*^xZ6mm1552cIC9u= zjv<;g#`Uxy9V(^D_UIlq;L^<*xb#1o#4%yQhX9&2q#r|binm!XFCNhF{8ho@ z0iac>+$hM&gLz@J-%3u87~4xjQDM^Fpo60Dg$F1>!6)q1Xf-WlXPd%`M3oRJs3FZH zA)6l6V-T_uqEc8sjp#?1M2u*g*mVs)IX&tc@r$Cg|5Qd)^PgAV#txQ7L%{ zQMCMO`k{lK=y+dV{KCpEW9Uth!ROOHAwY}~X&^ya@Q3&ga><_La}_G~;2!}2s!>-+ zNF@?LT9|#oxTcaq#10W-d%B1zS4s;1*ddwH0|T0}c)>w^sL87pdknsVJF6APCekiS zMomyOMi-TZMdNhQ#IWcvT{I~y8n25cheb!|qA6j~M|4p7vD9(SPWo zm0{6chAvwbmfe#p>j}#q$dzpj%ihS9Z3@fIis=4!gk=}!%65ijm*>ii;4G)#<;u#! zvTJi?6T`Awb7hmlvdy`&$zj>!xw0u?*|uC+JS=-MSJoVsJ(DY29F{$wD_a?sy__ps z6_&k{E9(i%-U`aDmgc8B0{qI}a^XEwGqj3`Rvf&<{WpfjqQhS55+p&@h!4M+foKN&iD6OoyL zc!9voFnZH6@{*1qDM%Q-S!fs_Tl4~Ng1jh&ysxpAhT71VB&?F}+@Ju?cW8wd#DdoR zI&Ft&0cIYB@9V;xDPg#xm)ArH~h+Ag)39 z>twgmqgIv72%Zd05Lo0W$AP?DKH1`RXj5Ak$H?~0$x2As^q;!xUrFi*v3t7H<8V%e z+d`)@9>P%kxxIBfm8-cNmQctIJ?xrGB9t27EDrBf6!|t%PX51;xw1&QAn&5{7LHf4 zkC$J__8~beQ^Lq}Z#|qr=Q*CpH*sZaHc#3^DF<2kJiGu;$M{vS(uv!yc;)kYj2m-c zg2r5zrKuxb!7)lWp={Lr-+F?(1KfFd;IYVOJX>W$5`l2CM1UGA?zyf!)9H579EaGi z>8Vstq9U*0qI4ij5hn=A={g`pxyW;M4nfsV3-!<&zp97+m4#<;Gc`Pwks#SEXOc&0 zus+g74MVO&U*i7>TPBd;bH!4w5$Q;+Lx!sv4v`XpYIkD&(FeqA`!Hl<4AO+ezplr= zS;p>G&_)@Kl`)3jiHnz_EuXtFW;a%onUoOPdc?gk+vpcPd ziu!>_mfUJ~C&$cV_Ic*<6xoA!x-cDi;mSM;9@Kz$x`%4+hkTYtkMZ^sMr(oeiWyE< z<&VQa#-3%;9~MQqQcMSspyTYCCsElRrlvHjjS$M6K#f@o7$gW76<`ae!{ItB7>uz? zKZeWkLLI%@{2h(H_XK?zm7CgQw`Z!yPeT0WDjO43>jD_j5#vGy4 zRW2(oyW82GinW8vVhOwb%47`^KuOyCk=mgr~p*djkQ1mW0wkh*{1U&vio)> zqlo18trEINoME3geCX1Oe0tLGX->KlZpNVP^CLd}x8opf&}PG@-=rL=@$-~JO%B`w zl-REJ8J!#P886m(?7m-Xe(_yOIdndpuir_gnO5M}0jSC7$OOhJ@L?mxyj+h#ACDOc z`cH?7ef&DnN1hCaCdfCW1jPjcba+`R8WJL?+EWAx8rXq3Nw_+Fzl%=HZbNdJNPi6& z#tkXQD#0u0`~TD)br1pYNnApZFcu8bRnO4RaAbo^>2AXh5&e5V8K?uvCvCd$V7h)(x?Zzi*SAOdD^`?KaTaTZ zw7xL*seJ*@7W=RY-X;v!1<||86UAQ6-Oy}xefX3PgrS!rgN*l=4@cU)5~g10ys+F4 z$~`^9v@>}S7&}1sEd9I4{v-^~HGLXP4jdx5bJO5zJs!B9P468HeEov<-|UC>eynJF zTDV7*P&|@;rJ}w~?;K3O2beeQi1b&`f;8L(p)JNf#PiS@k6n5**>`9NSgJDWV!1jc z*A;dPttTJr_CMH(B^rmu_{d9v^r0mz5A3@^2*6!QoN^FzJo*)KjYB7b zSv`|KtCt1zJ`nDZcF@uU6Ms>vTq?60(iJfFbugT6rs;<|p`WMH>J$RQ!IC3w1ARZ_ zUxnaN1H!L{@UKouoz|4IZ!##1W2ink6fB6CBR`fvT7aLexY+yX*vd-cQgI>p?0<=l_t2{sMY;e zxVlAHU3EIGi1gACesDSsU)^PLBq?ra)BlXn{Jcmn9d@AR=SBwQO}lQUuN~@4R;SW5 z9mG$c_SuEED1FwilH(tHZVNQUhc*V@-;t|AH`7-Rtt+OURGOwb)$E#|rp7}%L#Q*A zE|*k9cpHLheC!cWo9x)GR9eKP>r>7L$sNO27U6I3^tI{QRGOYJ64X1o4*n47C%3G=r$3o{=u-(g?cm9nzT0P+)EyZ#Po${> z|7c^IUNm}zjQ`t#yg`?Bq-hmpmwH_3dE9MJtD+~{XLlQgD( zafgEkF|7~*pGi<+vl6E#&>krJHZc#b7ONyql4RneIajq&4^ z=+La%RgQr2aek#-_snDVO;g{Oc{Pj}U#P^|(>zFXwQ|Jh(DUDm`S&1qhN!KenS$C% z{}-0g#~@atH1D4$6`X9Z$;jGL`G;#=;4Jsg|dqAHcB!ioY>n(+EH4WOtdzv=HV`+_7WOS2E3lrfdxu zdU2NKz~vC%$ja7oB8&&>SsV^S@4#1gIlJajdvJ?`7Swh*mY`HxzIYt7YhXKO1ch{& zfr@D7S-OYw>jYf8i@)~4|(6g zw*Fnv8ar0EzT;q9U#D7c1;Do+Y@@ZR(NKVII@s1X1g$|lg5C88+j4`nwCVIT)6s~# zhe{UP7aJ1XYOF8XxmblTu^1kCl;r5xqH^7-gysv&Yge=TB%DX2E!tNB1s5n#ZZ8B4 z3w;#}0T=q39Oi~_ZhLSlo}fay2bicdvkNP1B6HPHGbyd96GU2_KxZnG>JV7d|=9gcT4 z=%)_T;dlg)?U+ns<9xldKQymvyyt*&L_HwK*)`A#T;*Ys5 zgrSUt1sVC~aAdSvf3U4?N%oQ4fuTX#!mf2RZI}EA49C%?~$HLIhihQ~%sosM5f{%D%95$;CYG9wv z=)-f4^gj-h@5|N}r49&HCh(r(5Fds_ruPH`d=5J`%ec0j-4$F>!_g|bIfe8*2cE?% zbNfOXc^c>MK+=&dl7r!I2Mfb8VVEPpuzHj*EKfOD{=XjViT(5BfOze>>L~o#K_NfE z^b9sLNcFsfZSj0^0J?N~S^}=kO(p^)yH<9`HXq$d@Q1A+Qy_R$sAO|gfe$XwEWgll3El)bMM%}dVDT7kWWHU|!BH!!JQSIVS66=d=C z9+8&OE1WA$y1t;FYw8E4Wx6`nx=DYJQDYb|B${$&@vr<=lmzanW3ECII-l zgGNBs5(L0rfi)zF%)pbmC2Z3zbJc+s>BUWt=~j8yB_?hYajS%$pX*4pJJkwwpOg2t zXaOz=D`1}ZwbYk~V?>1u(fzbL_;RYGrn z$gF;bp@HSA<_3%KToI1ZRYPY&PZ#wI(Z#G1di_9s2GjWk`cxSdS38Gj1$*3zh3~rW*@-LF9AjDI~)Rx&l3!tae7QF@h$Km&+JFW_IVopx!M7 znAxzd>4t)S?m^$Rv`zJG)64UtR*6kF7y8!GbeD)8>xvRftfOT&;@am99~kSl7A`O% zL);h+Evg*;fT26p&|9&0Y)VFD>maNhpueO@f5a(5RawwK4b}2yn6-{R@C@hB;}`}k z@wK_p>n!s_Pf8J}kh<$4(ASH}G;WSDrn3VszE|X0+^X=1#tNj0XzIdiYxKY3ouq`U z?Q~mGJv>(#xDDgIoA8!DK2c47!dZG}(kZWI`5a@q2Y%bu(e&H`$VCSwV2E6_1&7K# zNr(P!CU`oO1}h}yq8w^FTZE4T9w|%{66|(SQ^d5s2sOCIo0li*g0 zhpmO4y)H-Bwog+T1Nb$GO`Y z$Fyltp9l862JeYj<7o4uF1svnX4KXF8K4JoWb;j?czj3p+u&RV=Cxp*RGvDjbpUP34Q!l z{ZWs5aRdV+%Macs#ymlHwtON6Bc~EbTa@TB15dKW*9WaFB-6>Pla`r83y7I|Ec(BLrGNIu(OR7wy#R$?+Hs@-;f|w7?7}UNL?g7)nTAOmUmJ9gjNTwb{ zawy+2d#lcYvGPVe{xztJ9_v6qnC24IZSe4B*-g6Y8dT*Bl#F0a8p^kS0c`<~nwf zbvOu#Y`!HJuY_)nOkkHi5p<@3y<|k)dxYW_3AoO5=I$_};g!0zaL)|GZYFNf{q2+f zK)ckNiVA@|Wb}#^>_Nka4N=M-F@(hakb;v1J+5ZMF6^}-Qa0r}?KABM)HX3lKXfz8yhgPJmd! z8pMgW>fSbsV`(=S)_RA%TI!cgEt3_&jOlV@E~4k~^#2tmn6o%n!a0n$!%{d$@3$!% zb-ynJ{noPE3(>->qkH%3d9C#PFnV5>>ltHLaTVRu;n0m{R(2jvyz$&EztNdGBhISZ z^niat$STe+q*vFl{JRL6gozi>p`p-!hS5+e3+ts5VPkOVJ83%E zhu0Beo?_4kxV(fum6#UV6rn`&(IIzo1WtXN3k?KQ<%Q!pQ9oFbIPIr@T=(0sWg2-2FSps z9WK_!d-P)MLloT1wY~Y%O@Gxb1vY?lK zyU@Wl#AlAt6_EJp*#e*K!$n?m#UpVGW362|E(J~P=`c*W=YBo(wK6m^*twt=ra4i` z!VCv6k6mPL2jj!TpuN0QWEx=ecu{rJ*wwWd6*yW3p^Ijo!MrW6D`kLe*J z!BcTKl)&Q~R0FqAaYJrTL&!t^JYwpdi_|S8ObhiY#M-2 zE2sUk|4_G0e9 zMmk*azoUeX=wCV&54|=!nhYP)sQgZMxcD zQLU3AekH}gMbxDMQ>paWt&sQ@e5Q#Yc`&7RM3O+XV>$5$s2_GcG?C68)5Cbl-Obeb;*Q zTALu3l3jF3iC+Vo8G5K-8V|Dq)x)2I5sVO1WDw`+Qey>ovw12NLWSvh9)u#yxztB| z1+YjrwmEcLiO(mSNSACz3F2S>8o8(LxUKhSU?>h35;M}Rb@51Q!A%d1Mlq8 zm;_qmJ?n2fmF!Zh6Yjd6XM=f_(&KVp!wYzEs}OFSk~Ha-lw*}*AS5_C11DMx-;c%` zq!9bBpXi_bobC!wJY=xZ?8;F5Uzz*$Zx~2CV3y&I4{8G+>$D2t!jV7fBpDS6`qZcu zn7>LIK#|@~cI{;l;WBK4dSGFB;tQ<&L+4 zE6@m3?aH2Hui9fICva)Pptup2VSL*|?mdhV^{^J;%d@(WXxL8B*aQe4h(hEQ76me- zTMk_lwK&!6()S0%Vr^yB1<>=VxQ5!h;$9UZZIa4;D8{ajC_L{Rc4Y9H^st>0M(v8fE(tAdh>AaM~Rs0wh zWx&XXqJl0-Ih0awi-S<|Avp&fy%IulL8^~-it2#QPYno!E>k8o5+vIJiG-e6K7e=y zKwOpTV@H5sDjlq5Up~{b!vMNQ0knMt)X2>6^1ZN^D^q=l;{z0zr9=&gVO+$pba9NH zH+p%H-by-Vpww!9N3Z4MvX-D>#aw`RP&KwBeF3`QZ#j5ImcHJHO%q&?K95tDe$dA* zo!m=bb{LOGyL6NfKYDh}W2t6O{t9T4ox>Op{UMcct&C37)7YSVgrCeaf!@M;uN#-|nN&%V<9sU|izm z^Zyx8rw)B17)=B?`UJ2Yb$m)i@n||t?NjsKEyV44BEKw%GR6UlkJAuPeYBE3Ju(E} z>yy0*G|^A`0E5f)LkvulWNzO9iF~e?;h08k@w zl>s&diFAjN#Zb4T`Zz{4y_~_hZw80Kj(wT}*N8T$o7YN5fD?cAc2J>BsXokpE96|E zO|$CVN+7YSvPOg;nh!ey$5b|J$nl+pFqfCIIhEdy$^xZDz)kC_AZ<(sQ)^Y{U|$YLDVXtee7e96X_QV2AF0$bePXx702+~ zdz9rttCaq-$T7cWNKUj(W)Tt)7*;7gEfXdy3zIzn!(gzG2wg+Kh_k!Y>wKaTZz6-% zx3as>TU}Wt$`+6_7A_MUy1IdvYtRofK^HB+Eo-5^W$my67{<`2px}otz-4=1$QGs) z=-=;!JVnS5$lwT&r9?ats3G);-OfBo?Mc8$rtbi@HvBuRP2iNPwv>57*r<}lh1p@( zVTtlx=-@wC_{XHnR9E;6Kv|&%ZKu5Tr2_luYd|1Illop(z2`0i*;KBGI^b+l<3AEm zf&orw!)>a`1`*r%Rc;e>SVHpFH{*b(g##chdG?3Cj{`e+SO{`fjOoghL(dvM(-kQv z2gcC(+7d+19`w*?GZr?Z>PirKWmWKh)wBd?P+cHD9HmndVS@t+kg`qZMzS=^Aw*#qY)lDvqWKlG%d$f0%b|j-AD0F$mryWBfW?M@8hxT|? zB!Qw~Jzf>@Y3tC6*F+L18dmXY%&Z~`T~jhmFDIk)hLK>2ZCQp{?G9vQ`KIDD>A%vn z$4EGgkZ<}VP7~3nx^%2xjjQAzQ(ep83+{o}`I1WWEKIBps$}BR|(^QV;rMjB1HEkL%HxhgL$p`kl712Kv+7f7vwFU zV+0r-C3mE zBblZjJ8XWD7}tH;t)$U*;aU5V4xJ6$zCQ%Fi9E{6F{O8reUHu@05;Pphck4lAKCyX ze?&9gmxBujcDO93*raTTgMb=yqmHy0m@&&fsvDjQ`4Wc4^E&#F82b$qr*j$8xhf@! zXN{pZO<$Eb>I>Ing-Wna=o}b;fEtkaQJ^uv0moiqy1WAlyfJLkb7nu*55Y^=3&-?& z9&WHM(?`(;Kn8u$3=*2vd&KOXF2py3{l#a_ZQx1+F%D)>87FbN)kf(^-}9=g;*h7u`$*{nCkf2m z?q(<{>6P{nX%~r#%sh$4f|6uy1=&!7(yEk0TfrzV!|{ZUXq%Cc)6SJNBG`-xJo|&r z5zUSAr__{HP$)1jKVGHae%eTcxkO||gjLe%ee{upeK@Vc3zz+d@`SuRCG&qyb+-z( zFj#D%5)0`VKTE z>6=|4MejVWNQrM_=mMBqtunf(W560iub6&VXUZpaowG&6Q~8&`Kw_01JeS;k)#ajA z0QhM$&?-;*lx}x!kVOUv>t?qT?E17Q$d^t7F1Ct@(#@_9&m=l5%QQQRSD>cTk%Kh5 ziw^UZD2ig-9@JAf=uRY@7`wJ=%2`qQyLQ=hDjB2&T{zJeIK;knB-8Kkh$Wo5$IvD) zay+~Emql=c8^iP~E5R;GUR;E&umxq8SUzztZYLN^IYowG*02K69v`i7YB&b28LY+% zy&CUOKgUF41HBXyZXsE68J@!wh=i%Y52uXj z*+pO~S6GUvC_QYwI1IzB=Lnm2w__Q9E=igOtTK9Pkz*YJ!3x$7YYZXFvWpWZna13# zBk1*Zbon#Nnm>4#%!I>BV4_xpEF^7y={_G~`Fir_^%VXL0?l~L2I z*u}1S9CS`Gw%0<#4FN|vf0@VbLcoo?w0ccY-7cd!=`eJy?BoENQhEb70;F%M*u`EY zfWj(S3EU!JU^7qHHIk;Z4aPx?zLZMfQ+B3)BF&f-^lj3eqAXR{5jVKjP(_028xGUo zgux;4sIUq))wVufHutge?{UEQG4zMjfK-<#c}d@KtdBEMwX?UgYfr7F8GRB0=SmD5 z`$}0?%gGCb4tT+L$>HSAK;0Jx>Ejv#?) za#~7ckdbG><}tWR2M`n7-o-#_`b=YRx}j}wqp)F*9Quz06kB^hhKn7y-%CY{}uSnn97cPW4Pj3LT=>@PFv*X?E?$ z#jt5F!9yeVax}7OOcwuTn;fWYzu7mPTB6ghoBBSviM@OG?xmj&Os8*0_tC?!k*L7U& z%E=nYvtY9C^d!$Q8sg?G>uV#yaGBCPO!|_lUyS(VKb2;!61o#+H!H%lWZpn%Oj#UM zv&PYba>LFjWmPE@e;gSe_q7#i__;nt(f(_}GrT zNT2alS``8QIGYm6S&t3V#Ae z?oLph0XHhS|L*5{Aa+=59Nn6Nq1Kf!f!jZzzPKs`_3>0XTy_Rj_U(&=J!9m*0g*|E zB8iHc$Hk{&{eV?Mj}}5ZH;&GpuQg+nzOMF&Wzws#wFhTW6nx0cjchAE)n&B9A+gU|&j5uIfn3z3*S2NvYXE{cHnx_p$Q@*JNlqV6)a*%~N ze=tmWx%gY^*AJRAIJ|dEmGrS3YM-MW5&hKM(^v{}d@NUZDDZzOEkBj|F zx!98*%EUm~!g`JU-LBGpE+re zT@j*gACC;}b%*p|N`2Gpwubp_#we16tNh!eX@Pv&qp9DzV0Qgz>z`8F)-ng(5TV=G)-{fpwmd*CvS|;`TMm?jDxZN2n{$5UGCYt{4bh3ZxF zO#uyx!2S!$;@aNLG05vLwWmXi*yQPoQY{lPEF3d!#RaD1%72SMi|{HR&z1i!zq~nD zzDCLe3+>8W`8p{NUbkMAE5AO!yeC&K0V84kjk)rTkvupz<;q`B*Z(-vG2k3^f%Mf4 zzr_rze9XrU!mvWM(Gb3#n4bTOsy*A8&>_1lu75mBX|CfUp69w2|3K;5o*+GU=$?1x z8mCobwg*E2BVMa}zF2AjrZkUHYd`8fp7*p>WVaeH_wOqrpf`X$6^XgECm_Vmi!uh-&NhB7v0 zZR`dpoHmvqLf1EFUc2d+GSP|z0fYMpAhoIN1rfZUUbn`?+6V9Zc>Z-S*YNut>g~1x0 zO0UWqUWjF*S7p`k8<_~jr@uGUWRm^9b1pu$BOKWw=r1j-v|2WpoA z;k3IwipGFwF@g9N2T%_P)WPM>mIP@`9ms3KgP5jG?{;7|*l-xwGHI#lEA6RSv5Pv= zOh;z1_Ch5>_!r6WuV8o{ZYG7Y4}(dn*It1 zQ#*uFxuP&PnXaKhXj#8b<{p|24I35r;{Ac*1O!M*M8ns~+VXi~@dR9-}>0 zR_1=bATXWuAj-sMx>>El?NIDnvKf2ZqZs;jA%hM0**uZD5U}v^z!GLAc zMp^S5hr0zz-h!+w(fgd?nt(z;B^hk59%+yxNS1f8v3wXZQw2@FIIn+2i2^(3YUJH! zO-zDY2(O#A-zc0UhVy^r_ECP21XC8L5%l5L6=z4M$H%Ub>U6<`O;Yl)a zz8?86!FoYz$H{}7U2NCG0s9HBN*#=FIl@{#2fWG-yUg0dvY1aV!XCd$t?Uc%tL z2fRmDkOg@IRiYGxZNjqH1`CzTn8`M{mW$)hRYyL=56iO2?1%1*7HZ?nLLPTS7 zX05C_Rtie1Zfpe%g>*DkbeFY9c-_s5Br!JA;)rjVbhEJF+pyJ`{wyrOLsS47%2Rm1 zhviQ2k2T0(t?y`eL{zw4+HMe+2D=9`&#jcOO6k!>X%IPHcd!y6r5k5yaJv95*aZ5e z>&1jTi0p>TLLyw8t9%D41L0-WTn@Iip#xb=4#>9dg+Yelzi)m{2P>)H=bnUltwc$MP z@)o!Yh>QXeZ+e_k1JJVnG);xQOd{RIIR=j`)3*;VN0=wIiC^l%G&DtHfSowU>^dB9 zV||w2aOlZqi^GAv3gzmp7U)`CreY||35a7VhDjNRoqc#{uWafvwr z57KTNRvhs8$>*xtISwD+uVitqB;fWd_bXYoUrEn?B^!ehD4FuGG5I_VLeTw`{Yv6N zNpXe;**tO3?2M>$YRdVlvHYO*(;IW8TVzePgh5$TR*qOdS311@&ql1DD;-|{f)VTI zO1D59!`s4B)dZR^)D5->-x&-+H7?S>3|aTZL(28K?;KLB*FAo5UN2zso8g-$+O;){ zhN#MR=4t{Gc{21Nc?}7UKM|&Niw1zqkaT)}u6rKLJtaKx7bBIf%$06I&-4e#HdX{S z>&J1pSteaaxX-w}#fs3Q?M|2rdGRHA9omB--X{GruUIeqs{GPiYHZI9nH$jm^CWhYHL$V+lZ>fM1w3XG=P!( zhmG+{@|+nnX4TG))it)o=Cw4oHONEA5yJ?7sBcPy4I5|YchlB%MuWWd7cq>+Q2!W= zG^4gvl#mg_NHo^B#S#tZChWRtw$c-Rz~hQo5rE5hXt7Nm5iB%}#PW*VG8g~xJT?ipik^!*OSWzy&>0UI#=>?EyJN| zLRkSZF3?q8o=S(l1*@)CeOtvUjFYo)x#ep1=;#DfeUz)13MKFh9!z+6u!6?!5f|A6 z91r45rM@TQcnn8JJl_7fecqFlay;>uNZW8Vuw&Gs9VHgu7>sJsc3d_f6pFxGz_>oB zX;~JV$J4Jxel4J<7NsFc|Jw2)D8;19Ftz1Ter^7@?B;r3K6u1gwI@=Kk4cQfr1NZ?B08^ti$TbfK4rBr9 zldP^T_drOTfJ*jZHE>{r_MnF=l|z(`a7~+j+NGS6k!=2>y?A%&WAM6y=R%u9{~M)K z{c3pPeAQt(QFJ%oi!z-8xZq<1ZI`GY#a{&X<8@GFxQJyue2_idsG)#C4yz?hi=u*a zq{pA1UIvZed?*FeG%vkk8O=;Dle_`3X`z@lQPa2lq%Tq3RJ_-+i>N62X%mhL>y){CDUf!>6^WF5H(o zS$=S7Z7SNu8Csyz!A-nP!?hyNsRAIpo)$xX_;+g@IwA zI<#lR;yXe(nQoADNZO3HQvXCZA}u~7p#f*=*F*O8Dpz z@I~nX44A>R8hnB$IIxZiu%-xO#n1TB-tKXw?MXC$OweK;1HK2YG{&a}t%i<2q=&sP z7#6ATTvqRG0HN?uE2bxsT{5*!u0&`_1rW?$C%YCi{T6g!KOv#^w5SVU*w8cql`fkq zg8=BY&nhN_KxHhZUq$@zn%Vq_hT^FJ3iBv@s%Y>jyn}vdd$aY&jqbd#M8v*Nm=F2@onM znsH6HLH;qOxd^`fw_fSXoJNE=g{@c>TF}L=J)IUI+q;is$l^|jK4R=4NusONiqoqq z*!AEa-WMVWsYB5t6v|xlpS|z?!OuaC;Abc%C2j5&T{8|zoWXDpTY*y_gsv%Gjx&_(|3T#o1)TUt&_WeHe5=7r%;R=JCvz3f z0)TwgBBY#Urf0;>znIQxcX$f|mzFd3emtP}H%OKZ&+6LSgaTc4GH**d)Cvub-HHBg z>quKnS5tz6$QfB~Kw@`7$n;Pt-R@w>y(BUqegs*^l}G>x=c7ZsU=2_F4o6|%?hNU@ zid|Y5g}bzetOsk^jicLl7Qy$>vzS&|!1f!9WSH|TUw}Y;ne)c>G!O|8HXtg*DyEIF z#X%TKB7FdUk(C25Rx$Op1D>2HI%%J`&)df}7-}67)VYEZ!0?KMO)%5X<<4GAFE2_< z%qWhED-iAyuKF;Ba-BkO91{W~0agTA545LgLE2)r>1K;?*}2_j&x)8Ye@<}_!87gw|dV|_8x`t~$Byd6VXws{Qg>=>{h zw?mND(Z?H~)ze=NHD{+)Om`|^=oTw$71JF;SvK8*MFuvx4&C1>#k3Knf~(CPAe@-R z)U9x(2=LC3NNxe0a?l-ZuzZ@}kATaWdAPKSXj zG2ISzzQuI2<%bej^9vf=>ws-u&amqA-~fqX&qBJ2Fdqf}HeQ0yr;xOr&CX`ZkPn*(n>xzg`eh)_tRBfjl!ofz_v@mCF2?{JKc zN-)j9-Gq(}!6G}0tPR35l#jpkh68YZpUMp#c&%=HNdx&3l=hg6I7ixLfpKvKi!pUO z-k-9VdI|>R`srI^2}u?~KC<8CK5DT|s|yCzZwSr{?w)eS0za6xr%t9Ah>_{40(ABC zLNE!J?I@o3UZB{(O$7;SEOLUwSDa$#4Aubjk&&EDs$am2zg)l4vrShPWaa2~IRmuX zTXDsA?QParjAt=jS>St>0xW3c$^xIKhprh@Ue$c*rExTJ~zeKJZjT= zpB-IgxJf7(fRbq=Va~gKw%1CcUFO$cJlHu2}lLFX;FB z4#z5>`%Ith9{&F^_b2di6<7W^di$R0qbu3Qk}cLiV)6q*53!B-(P7DIu*ai%QEf6r94HlsH z`JTG>w!Dx;Ff;E@($=ket4^IdRdwprsdG+2Ax}2|H+lp$(Rv7XDohVbJvT}{9-^yu z1@#EPKX?EK1kpdBI1nAsi2gwU3Xp4{It;L(TtBxD?D~>c-!YNRW;m1M|)~^eRF+lqP;%(O_*6Wh|Y9h6>m!_ z)d=k<8L%l$dXIQJx1(O$e~z;5REE*i(b(RAY$3yFY0zezA-ry{pO4pM1S(t#G_DE= z6q-A{GYVzmY#+hxVz4aY7_apY(ONFl#~8|_L2L>=%Sf&J05=t;Gvy=11KPD*Th4f1 z#RC@wMep&7f<0SXIA%@8AE(?Z$zh8%!&3p(R>4p70|XO~@zdDKAwXP@?VlAbzBqvR zl0aO(iQ%ga?s+6#mL8_b*bP~ZjuHod8F1$W)Brf4K1ENl)vE(Ay<%LH6$uoa5@va^ z8Ge1IToPnF2N@yNnqU2p9m}r4*2Svk>MD-WB2nLkT`brGaK)pZDC+<<{p`+Qi{@dU zNQEw`L0@qP=}iO%B5X#%+h&N>3HBUV`W#*!;aHuB9#EHHe+d-|LTbXV2ri8j+KzA{ zAs0=AqgYI?ZgZ$H!-PGMwFNG52;5Kr^eAo9_$0O243Fq76iRyU7uhv7sU)zP?l*tO zWJ~jz7lNp6@Edc)5E1`nlBA)81-rC-k){^pJ)}9 zS{W?suwKlBL?CMiS5|VZFnRpLF~rn!7=h!vhMOoxlF@)4kZFo!vtlG_FA&~D}6P|l&b^ScS>HZsu8D5nGE7nvPH)dG*5u9B9yy5Hpiv| z_pNKHvE8s-XhXBX9)H6%9$1SMjFdf{`;Po;aSQ83SuCF*DwbadavFxNRSZ#zW4zT{ z>@*_K<({~EX_fHgukNvHFa$cFX$Hw?9L`MoIkrApgAlDNI7SPHmvNRl^U*Zd1ehHs zOu}LFgirige#+Y5>(}&Giwd^0C_-wVL~uPAmCdsuz9g}6#(o|48DJhG_y=QKT4k-# zOEjA&?t*4Hiw!-5)t({X)ft4ciuVU~KB}F<#WF-2g~?_T1VJV7jA$E}77Y3J5z_{O z!q1>E_W9TpF0X>{?9^od!K#elh#>*Itnw8ywVVX!02ev~?0(`I!mh{Kf@YziJ(=oq z%%~olr~9L2+V#4&yj{mJ7*$zc@ZOpKF306O;>%%+*9RpxgAn;6{o=t-`?5O!#d7ke z_r*7iDgdflF<%TA82t$n{D}YMlpBLWqhSr?Uu&asP&1^@H)C3Mz~bt^X;0KP*86X8mbvN{jj1weG1L_j9@3=?x@`Ir zxuO$4v&*J;U>mw9gQY{31AKsi6NV5KxKPJqx`tdFPl}FSN5!>DDytBJBaTz3LY#u} z(|CgbHP{hRy)6QnV__wP<^_x)Ch?%;sY1qQPXS1G*yl1x0O`hZOpg5eIQ~GSo}FlF zN_dn#ZL0rpviHz9`g2RF^JG6qTZ4bN-JTfrbgEaby|rO>thGMbQRknICu;rPO4Oy= z{CtT#96ScE^FZryQX+8Sp2*d-g9JTS?IrdXoQ zPt~{1^4{0C&T43$=O^a3{2U$4 zZHZa_*zo(()I##qTfM1#G#y;Z5_V=XG-5|HH4mvt6n53ov=Dw1YLa)M)d0uvxdGv7 zD!E!6&6V66QI*KqW!FTYeb)f; z@NG8KAQ4qrqmJbYUWSDfixZ;%PlEWUj)ix>#LbzHL?IS>F9$;fQ-!RRP%Noq5!B34 z2jg}URY`}V#vYp%*`^o4VdHHc4UVHHux6@B^uvrgj%r*r3HQ;))NxExlaSb|(8z1j zwmObBOX#I|1h_@8HIKt%1=@%hTcQR43X`CrNe`>z=n>=xO8)?Dz$B)Hfbx4o^eR)d z?0(7}9wuL*Lvc9T3qmil)mPA3-7wt4cO26RuKJ3EN`+H}FHUXni658X&`-x=uEuDlo{ zUFf@?Bg=6J)^r5ZdAbUE+i=m^qiJ`KjT&E;8jo@jIlEFFO}|K^)SHO8MK`2vdI*LM z4TzUKEy{EoMC77t;pC^R$^y27D_bN zx0k`?yQMYNKBj_otqWS(Q)R7*=HyuF8>0ne?Sl6Du~E3Op>Av?^1|jA22fp7%UH61 z22)X)v2#08?e)o))`n);J&dKw)R@Q?jOvzDW8+w=XwT4yS+YNby{NOpWcbHXt1`jJ-^UTkBgIy-)y_ZyQ`_ z7{Y9jzsN97txtGz_Bg|6txqJ;3Hg^H{ZdQf;|wF2szbrBVSs_)alB#3ILGq@fjvv- zJ}7SvJkc-~!ZbM5nyhcFmtqsqrgkAmv0>CT$(EwTFg#8RXqI6l+Eb$co@5xHw7foO zG)j|48z4{64KsvPNLkUH2fW+h`m1O}Z%*3+c$!G^IAkrmov$wX^s zLvt+A+EiH%oDMMzZ{{56)2w6+^Dx%h(TvH00vhxztdCxa{M%GZeRHy*b?^6n!9({B zDRHPF<0&C@4id4}S#`FpLrUN!>TlBxJ~)pl8ira&Az#}jy^w$7e^1rTNwr8N0H9L= zfc&cl>92#m@g_UjgJEJmx-q5l0)Rn$A?#tgSG;%4^-|Fx`em%eN4*roAd#=O)G!v- zx29s)W(WJtQp0GkZ)!m&d*C`-6M~iLjS25H26ZqZo9dGd33M$!9&Q-m9C$n;|M&&N zXiT)VWB)lHgDxgVI}I}KkMy>UgI~nVFpBzB5gnJ4AY~Q}83;Z+o=ul7muR#ILpqMl z@6a}54?Ai+U7nU;HQON5!d>R{bVjt-YYa$=(>}MF%P>S`AcM6{Ff ziA3V^fBCOEE`K7CI3B9v@reE(^5BLi{Dc3h9aHL7iS7%V;`~JkpeCuM=js-??<6-GASi zXq*FnNSX!XQ5lA2-Z-0XXcEpD-g|Khk6xUD#OTE-JZfoHn%4%aN(d`eQ)5H3)SOR^ z&fufS`;=8EE0S!GyFf1|9_k)Zkh1iLAR`)<7YPdY7726`E2kC-3O`jOD9EF*ue1PE zup~F`B@$@R`-%km?fxQxejlujaI*3m?5*MX`q4%a_f{dW$#yd4KEzMh8N0` z?X?5BM@gjc>1vQgRr#{yIKyAV(b9&-QENClX7n1a4zz|NQ%Wms#|N@y(f-2X_`I+< zArKZlIncvNY$`gM+hfg%_J(=V`x6agc71!N2s-UnLtVD_NR^Bd9+Y0~ zSFiI{`>ay*aVQ5vilg^H&p9GiG4ye_UdD|=bK3PfuTu0IC?PouSNp|jTPZ{>2%-e6 z&gZ)r`IJLDC8mjD%7k21L1(9J<>0n!yjm3+go@3JR=rNNQ?LWG=)Yi77lR7Xk-JH~ zL-LmO9v&0u2&huC(0}5*Sk3`qs*T_cG%=?pP>T(HB_cAxxT8&Ek?;Qx$Rb+Zzqc^5 z@m^0Dp;AZ{@}G!6E2Y}l=c*P?lLzE{_<+uZN3Kz58uw`=Bm~->Is{r)cjbnzjHr_z zQIN=He=cfQ&}#9}QYM{)F@a92hKdoYcQVCz9S7i0^vBtubkxOP83>XO z-6#(zpod3~xH?Q{hw*@WNgUQMq0ZG2N@yIe&zc)t)fm5DlSi3!*Z)89nD$_R$N6Fr zl;x^w8Tu4*Ys$I0T+UJvhgcd{ILP{vE4p0!lWgm;YpM}FIYX0N*ihhr1@y{Lca&qJ zojD&0s@mvP1uBF1pcCkcXZ(=Hua{$t(|+Vb+W&^UNG}(QS$%4Oclb%KWpLG7S9Ch; zLr&0#55-g#_w}sn>;(o7Uz`~(4 z%Hjuywf~uV%Y9a7JRa1#S=TC7DR8gQK0NGV&`1;-X_+>!iLv}A z`Wcj=W8bUo#WcqtJ%1)`Fba*Ol0ek{4B~xI){{s-RZLH`r~CtRU;VYOz@DtHYiLS{ zyoXN>ZS{?@#`>m)R*`!`hB3?Y&-Wf%7j$^bUqf4H%|EbDv}?^CtX>whWAEaX zsLO-$z5YXnAqs9Uhrf(bUP@EOsAG3TByCQ5%9Ol##(T+K1^L2ZEq3jEGMGZg$P8k+ z!VwmguzVI%tv(!mL-5JqtQ+58Bq=uaat_Z_o8jsuI0r?xpV^Wqk;EYg(?dUjwHM|# zm|z6AP>Ua|MN_;J+S71IdN9B#6E7W_*_UA$s2`}*A#h9tyPaOmadj~Lmkkr~Nlb-F z*gZ{h0qVab1Qq2n9HVirZxlambI{&S z7~92>!s3qx^&)|lCjx`p42%`=^tBw*Pi<<2;a+!!ELUzqKg_L0lf+I*5iYZhDibHV zSjF%~Fk##v3V6W}by%nbMvo5DJi+YCIgin<#Je_l3|RsRSFHh)Bs;a`uyN|i;Qy%% zXK8xCKu_5kP~#d;2toDuiEj*+S9t3HQ;UXrhLnlWo-igYMH8rVNZKX;Za22Gg*!a( z8P|$vIRR~24(ucB$k{lpPuoZXILIfahIoMta4>~a{yiZ7P2$f<1oJG=>gpa0kIiB= zi&X{-E(!t)j5&zh8d7crLsF+`hAZhbZ*w_uv= zSh)WeX393xmw*;q^fTSyujeY+g{g6zx`xPfbMUl9<8gV5Tn|N+fl4@{g4&Cea4jMf^HHMa&$VVBo{~1e47e#LW{UqiF)x0gen6t_gw+E#44_^ zN~(AXbql&H=-;uY!8qx{U1Ka)+|Yz1e5*Oe{aB{8MTjH)5~cC3zMqN|7NUY+>M5P$*Y}OIKt5-61R4v8m3ch0!XeuYxg_GVPS` zs>F14ItmtfAGvsazQi~{f58lg=(6;XGR61knWl}~7O0Rao zl*B~*Jl|+_;MIU+0|)|xkS5q-qjT^#QX|t0yEbbQ%&%=GYf$jYf=(4+4BoaxI4GQO zsBk-1l4lW$MGBsEs+29dQPWD!5lUHZ=teEr(m0B6Q8xl6Hqf;QIHP6PSR}#~$h_Vw zdv!X4By3roLCS&jA@<{3~SK`uVAt@hBn4{{<>Oqp`=x4yd@;EBjZd zxI=-8=V~gxCS;q$be>T0bxD>o4KKoY`s-A@Rpa4P@w%=)Wzz0ZRP@Z};_n1hT<1G4 zAXdkePkOCMPpVSd27KspJ({;I8p5DKA4Y>d%+I%rG2hs1qNQO(L;w#{rS!}q)P1pX z*>bd*u~BHo-wW!!Me5a8tzmqFc_4P8-_aZWtVCZZE$Sao6JZBPxVZ|C8K&orK?E$Y z6|s3TN!{k^&?ml`-qc-QT|{3(z*5imOPd(8z=6 z7wJAlRO})OW(tcBW>GOFB#!QsxcOuiu|++IYiA-Vrz*8*i=6ZkU75ChzoVVI1H>vs zT1AZ;eGCRP27mYz3`zxqIX(uZc?=S|egy0dYJbwJ{n~u(t3M2CzsHjut3{5ZFC}q1 z75y+;4Ur@PS?rc{lztD)kX$N$m56YV&IqrJAptsBFiFKdRg%wRR9nE$;>iAk46H)g z!F`9qt2Se8g-xXRYp_M%M(jZIL^=$*QnI&0E@4F77bZ+c4#Jcj@MuLkL%Rhd3_ua{ zAvH=A7MPKiGA_Z-14uMeJKKc-iVJsd-hPQ(d7d0}Xya_sPP^z4)4VjsqHUXD$B z*AaZj`E87;cstkFbzr%uulduUy{(~9+mIq^LU@t?3_^i!G>Wi&G8ECBo>+^15lI#z zkMMf*z;^LNhdyh6q+sJfDmtZgutkW&ljq}uoz&W^KbTTOwcV4pg#D+JJ?@=LJ+$k za$D=f4IFl?xPwt3K$b?o{kKFzK_%Wm3E*u6JX$bB6I@<`tgF&68YS9d1iQ3TEFW*C zbInq4E*7Ix2{%TscMs9I@V1o-&}Y@`Smol>?M*HC=s5I3w@nuX850-;XmcGN;P2;0 ze{u81Ad!COL>N`Qr$S6wx;cMi9^`!d@V~8A1%ej4*76BhUd_wGmjM)2s}Y z=BR6!v(y#1s%8l$G;v~;8+D3qk}Q77+1Bsc+=1ENaa4?u$4kZXqFHN(XW1qJK_EAcm*;2 z&h5$2v!Nt)B8t#5Tyy4$M^h#SsNVq8m-{R^q}@h)G7^Hj5?eg*0_+DP64?dt(F&OR zds4#2Jh1Hm){CLmj@FB$gW)Pm0l8o@lQ6oe1yO1uOiv@~;IB-VUhB>PG)B?|Z~=TC z_Z1A7IiUrp18N~i+>t1c1b&o=0v2bfC_kS{g0`)oQg14u5JFR+j&1tRJ49n(2kR1? zhxbCW#)TJx>A5pZndAXvI;=DTu`4VAb|i=eMk9WRy@hxTF;&8Zzzwb{5iX&aYq%Pg z;VxHmdG(V*z6cZn)2ng%LY$6khQUVQ=us~QCwy+cRR?0{Oy-Njdc{SXL#cn+8 z!qG5+yMZS${Xil>oal17gu5~=BLvPkFga+zi-vB&&1uQ{0L7u)htk6@Sy1+DS@Dh~EqQ$TiUQ9t)=XirtOFgZ~bkV}pW z5u_T1{W6K2Uix`P$1#+N3%}E!*c47eRY)gd0jZ71eVbfeUBM3B3iZsr5;d!uso7>| z4(S%Web{iJHSqG>2R8IOyuZ%?uadlouLu{1sV+h%Qj5*#MPcR40AxFgfCgB+pDA-BPzOd3Vs%^;Ym z7mys=QFsiyHo`o)P;xMMxv2YsXjrr`1C=SzfIek@9~C6!sS|Vqf1hX5amgyCpAOOS zn436$qVH$u#~FtDN@|8P4eMo(>Yz=E(SoS&@4DfLU=&;w!Z4uFAR@4#1a`OVp!O~L zRw0gHjd3_>dsu^X6J1ThBG6Z`Ym3sMNo?v-6cQ=2O$K^dT&6D{yQAVDNkuLK2F}gk z#LUYGa0U`XeRZtHd<#&>U)f`}j%rXZs-K!DhFe4MA5{#r0 zXk@Uqir8KQrZ04ZTPXPU8YapyV#FX8JK_{=rRj3fuu$DVioI{{nxgXXNy~;|Ez><9qTaZVQw?>6A zX=?Ryr3#oX#LCR^oC=F(60g#kB`Qo-4(;)O*k}|mwIUXb?7#rsd)W|0M3*8L8QL-` zOgn-FkR9-`;ZQP~;VdL?KTyKrV*)C^Z(_`6t0Wl4I*grhBM2BV$7Z07ih%lZcB~b6 z-)0M^!>!&`ScN-nd=sv>oShjUlhs#aR^aMX>aiod#3KU6Q3S2Ta)S{Qa_gelL@0Ng z4Q?6HgGzc0)p`*SprJGy(JfXjkS{9laxsF0PO4~Mc`&oz)-WZwk)u4525VjDl zgMLt2H*o|(cEOxw##oaPC#p|Y$v9WW+5Ed*>Vf7 z(I#gQQ_B;dG1fd{!2*KPLs*C13P^B*pcgfN=hp)SAN91Em_}01dPq(Sqk>@;qeTW4 z4q0YYt*Zcyc^%Eo1m5x|2B7|^+bUBI+lvfinfZ4c?vH4`7WH&su1XAFjIIH$sH%{4 zD^i8ct%Xpihb9rpShVY0jv>J#HcfV^(Pm+Goi+&bG=qBjJx#qhCiGF&YpSi{ndmOz zb0N3RhL>pAwb6MHZBU&Vks+e5zj6E#198;6JS3m4XGCe0h?pp`De_5xKvuappIBfc zlxDXLsWn~^VP{iB5sfSwQU!FnauLy_&6cAZo9Sm6Gp>13-1hpwqJtp4qt|L85sWbz z^z2mPY=UkylaSeWbYpjpwJSXY2YUyhql6F8i!{h|VV6xWQj)I3sScuiJ2E+|D!8f& zpsiwb+$O9`Af<@Lc)(3wR)n*3YLw|^!nOuNYJ~gg(yk#a7NB2@V{|IcD)i+oGklh? zvm$83lh|B=8}8eP1yQD*A!)>wqOdq%BQAoMC%w>R(?yh|H%7IhfN7z~+vkh4>0wP$wW(-k!coWSS!p_O{ zMomqC?=x7*diDI25sWIb5qitO!e*gk<1o6DYyKUM(@My5^KC>FrR@f6i|86{XcU5c z1yvfBZh!$l(tjg#LJouXUv;p|{WH*^ZUjQ6Xyv|Vqc^r;y`wWRI_aHmTb}3%Ipgk+ zP-iAW-@uL@KB+iLtOg4K>ys+MIl7~mQY9|Uwkt8U8S598dCY8T9fm-Rd~>68kPBrU z9h8JT)HKW$Q1U?n$DqeL-;a5|g|>!rFol6!&i$L1-onmrmFen-y>YGHE9X1Nd8X;= zhZSfQ(+AyAd5lOeLRCS3>>ieR5|8?JOFREQ=&V< znKnD9!jTmWpJ!q;i^zt^s&Gh>egRfiAZ4!itdJ;W$DcS96?e#O49Y1T9=4 zCJ@SP*aG9-;39(Y9LtB|!NVpezeU9mhCHFZN*0(?vpgY;kLZ4%CB4 z`0VH)^0?5{JD@z}UjtA}6j(FHG(d^*fiE!A-91q-=37G8lcFV1_AqthKpIP_oPtV# zdccMyD$HrW<-tCPEPzJF<1t^+F8b-P-hCjr4*eK&10#3@L;Wj*h#U3e;I@Wo-wpsx z_5mFg0ID6WGX>`a+WF@1-xkZ>tgKXn-a?=km@M1H9z34QL z6cg!i5zA#-iz*w8V`ii_J`*gwQyOX&`w4K3e#GR)#y?)g&Ps+|t;5w3+}cp!Qcng1 zO4}0CYzsScCE8k?kKs@AzTRpz>g(sA7vwSA%LIeX_`|iE>;O(_*X#U<+y@)8ein5Y0k- zH($`*`J({RHv>p)$@X{3cA6E1+&0BUtcUa#VQ50BvOe8M&v)77nz35N zmBQ|+6TbU#9< zI(qiZ@E&I1Ru@?4I3nMGt`4tM#q^XI0w5ykM5ey3J`7qgHmj-vJ!FH~%Ci)1RiHm3AS>GEstqA69+XjBi1s#8-p$(G3Yx@9f3Vpo<=Z1m;a-tWu(Um6-r&pfl8E-mQ$# z1vp~7syEZ~GMo)eXje@SfPU-&t&$zKg_(`k%DM{F&{OhhIc6YKn{=W}C3p>`fEMQ} z!GrvwG#NVsEDE^EFwPF$k{)Kd%W$!8a-R^*hJ|NLe%wdtjVcpb!SFEpQ>_$tjg&qwCU;NKgXDy^clm zhnN%wzmONRRO{8q8e!U|}9J zLKbddf%fk>H_jA|R;glo52I8MTkH zea*1Hte#RE^!EoMY3Y?pNAMA%ru+Q!f(`k+OT5i5a^Rp2yU+1Vr;FWIjD@X?m@eKe z?zS=GE|lp98%l?v5OtWEAVUD>YPiHQ?drm`>er=Bm(Cy;V_`G0z)sCk<1Ko>YY5<0 zkXylyzSN`-H{rmbioTs;X9g;L5ylyLP)xr<{^j&)m#xNwYs<+r0R0wtRUock+7_1y zi>#4y%t=&>AA~8#JZcY5M{A{}7CqI~rzX<1u=?xA02wzTf+dzU%=dM4gbNiF?oy9w z^=8Z>Cx>lsw_Q$0B&i;)cj$;5o-m(cs0`Xr2MFR^LIgUj?#iY4!xVO_a0yX?tBM&% zUF>icO~8-}yHxB_eN?1K>2JBo0Fyfg=N>FaJ<%l1vZ>Vd5@;kg9IzS!0{yUH9~M>t zCL(4gF~p#zpG2u==&2!TVvHU0NSZrb#e=Zd(@272&4@%knOU{8>N=U>`qYxT``9Lo z3Tg`sI#-<*jJXWNL1*PVkiN+;WyxO zP5Mfb^_?c@)EH`n42T_k8{S^S$#+CqzUw{=`^krQ1Fhc+zlcU5C0B31}8`w}bYswVKP1qPr z=8r~9oc-NJu-d=yf5HptfDTA?iMDzPCLykGA;V~_Z_meL+w05h)66GbEnq?6{bB=$S2Bk&AK3`d#(ydlRPoaqBL0c-B#$i1L5%Gh_ zZ%u-f#rh1*bRjOvp^vO-)Rn1VZC{N1_<=WpPf8}~^iYnSB`|@u{YarLbA#%4<*SeH zTm1k#l@VT%uN<8&<52{GPs0EUuV4TPVqXlV>Rn@TTY$l@u;}(yaYdD-SH2iK%W=vb zStfO1VuFOUFeBr_C^UEEOJDAlcAl!jo~0ao#XNlX{^h9$$I9hC2%{rF;8B3kJpfu^ z0%?N=kP|=bg_uf4K!_%r!FJlfMzDlIL*lT7?Iz)-3u?q=WQ&5m$FzAtJs+SRbSVx; ztGOIZBE};Sp1Xta`Wy54GgzHQ=SPS%9-&F@sQlA91Ne{W{4vp3LNJRBc&nsk<7x&y zlhL42Ge5$Ph;0XB8uu>e7!JgGAa~8P#Tn2jG*{2pb%+lr<>-;4Tv}D)(kBy9$h?c_ zDid~c=62{o5!>OFJJLgPCVfp3lejYcd4eUh*3WAEtnscAwGCtPacQq_9ZUYU|KrGC z9TLt_N2nPKo`L@jyjaFkuvG_h`3yz!b+#>N8%z1NvENYY`Av)|qH&srLak=}0jWV&99Uuk!eY z4GGGR%9R9q4P3}mseGt@c)`!|qAQt(=kKJ)B}%ek%xh?CsBNsD)llD<^jFw?0Hyl2 zx`YntZyE*$B}&14e?e1wq816uFucJhVQDSHXl{@|4TXl0&t)1$OJm3E2JzRo@~yBC zm`OWzS#Y&V)Jc2=%P^W|H8g7iA~9P?WEn>N{Py}*ABvAqkzs&}Et1I`<@i;EPywD4 zFwPh)AB-~$>7mgbGTw8m7cwAD6GB)?V%%t{W1L~Mw{i}&Mft02iS{jpCey0s93C1fjn>`zBWfBgOU;3G?1p0iss zzGd--E?u=UMNa7F6+9D%O_6006*jA=-`t~2+K7uaaO+Y|ne=$KjT1akIE#^ym_^{i znlM{*o;k=)H9Q?wN>qSnpp-%KlLZ;NI0Q|Two+h^tI$kc|64yeJi&Z*I|TX!vSi0w(}NZ*4Gk)j_Vh#2FK^e`zG zCoQ0#Y(B7Wz$rLXbNXNlRp2eI@0pun5%f{n20`P3JQ`D=jM=P70d=LOjp$TXVEdJk zZ82>$asf%Yp|pgc5@$zW&xwSM<5GnDUltMEE0~@bTcNdS6nb)WAv6KNA+YL`YnvVO z?{-I~W5d~p7Bz{-VmV|Ts~1C=ei$6vtjY!*G=dI-xK^!qnA9_bfeHc3%r|WTng&o% zpW^VI9bV5)uk095_nHdVJaIR5+f@~V%Y7FaE&BNfs@}=V>g7S*AB&9zO&nxuiDFQV zAK@(5023S#W9moZIBLF8j5}c9Z!$eH2c8DBz7&J}4|DoNby$HW@4`rkL+nPr4Cp{3 zhD+PJg%+zrE~cK$PLw@6Z6_7q1dKlJSm%tiFE zQWaD$kZZx9z@A)XT9i#^G~i8?R3=>`3K&Pfx}VUPYr1SIax1l&ufqdzS|iwI4}dN1 z7J@#2!~hzaUBzaC?x#I0l)G7!G8QdFr=S-gZ*yERoeP+7SUQ$oH}|kpshEBV>e9i3 zPfh6;ZAJ@(aj)69=3I^T~&d(GSSqNiniak1VXUL5! zW*UZyyeETMqfF|hq#4m0z{TeTFt&Oy-r5Jo3J=D6d%-w64`b>tJQziEFb*^Nq4ku| zE}KeS-0pSF9Gh1t9#u*Y-%X|1j>iBCg==zt zewP|-IwVI8w(e3o7#8jpSvV@UA{?`duLsC0OA!{Yi0K*VZSdv_nSlaw`y6`)L$xD% z57cB~W-1(^s3NXWirybD4O!7~(RcpoEuJ^cJQrYFq^k zENc9ua1a)q;6!-;VTVN3SyJSR@(-_ecSMz<&eF9G9Tg^SZT8A|ENwNyA!pd1-BN$N!M%DgY(ceHx+ta!}J~~!nCZ& z^~4L5-6t1b-H)4T0B))H{CePqMIc;i1>i21lRnxcH_m@bE`a+G`5+nyk*^)+;v57Y z&gp>sc_#)NQ{T8G!inR|!kOtADM-RQa)8>qh-syic~y@j0Fsrah|cb?MU%Qj&%Sqy z5M8o}>6Lc$cAfOf-&n;f?LaFxP8y0TC?wWhRWemxZ@1z4rIb09;=^>yIG5@7?J}|Y z#<_n>6J8bgD>Zz;WO`SyTs01>5)Ln*hXMAl0x)_aH*Sz=V@PK7b#3U^#gdsG4kc-@ zN2}K8A!PrXN%=uLB>Z7bN~Xs&^nuaP9}FewMh#sR(MRplgO86-LIS~(Rm8L&{DmGH z4{FK1h=Xl*xu_B$zgx9R(dL8TP*`NqHFJl-chE`lqfW?BSfNBuwp;{$wAY+UQP>&( ztz9-GO;-DWSTE{D7c zL}iIs4%Uvg@9{ZvuW$h2%)7x7*rMA)Bibw*86W6Dm)#y$?Vz7qWTZ@ZkVYhX29IsC|!2- zWLJijynCwAW@|IWQ38$9mLcdZ^Y{_E2FB$Wa`%t(rqCJU|9e}n`NwZjir(Z^Wk=nUO+AESD-5-+^%!anuGBa)u6T+RwWi!SXPVp*E>3%2Ny zqV4&_>GOty#AaL+-EIdK5my94#B-3~MT(FnZax(KK+ny#BPvWsKpH*Ou2Nx|g2#mH z6u+n+WkS$z4#QSw{{nF00?klFN-!m)2xxo{c4_EsTAadeX1faG9tt&HE`AU!&%)HN&-&q06~EU`(PL9T4H*{Ne{%wXqX z{&*LdK?LRr!X?m&xS<{{7}#^+hea&Tj(I$F4P$c+y|sg#fy2~{BWYSB(qmwSIqg~> z%_6u{07VPx1!npeoOB`zg-+n~lyOUlqnQ5DR^!>CzeD^15w&WO>6Ubc`+eV|;yw>D z(e36a|LQ6E_rHdyieZk0EX?%B5KM{B??GK*rk6v5Tt)BWg#8V8hW&T9O)JQy_q(F_ ztL=k&73mVPS}6qpTv@0H_e(K__H;I+TlmnzL*GWEEjil`0xg-O7@$yQ7s6%YmnZKG+mBLlJvDfdPu6p`5x2jg*iG8a+Vm+^;0^a_3G=2pm@ar z-xl08OdpA+ftHzW1y`XrH+4l7Q6ILqJd<95?a%q9%dqg5Ex`zN1qOUA{QAQdfpLhj4BcxK~eIA2jeu zOLH<&H_1_IoGv`bWqNV0tzTlL2{XO6NTfgE9>lbTk{F;DVpcGnKFMWzWiIe=;KT@< z7$5`YMJkD<1#SW66wR_?GfW&{8-?bE!2pwY1rs{sph2dO=b|!{+2S?YoJL)cm9QfB z!t)EDd)Zkr-MW6d)U{kp+d*LS7@9xq+YPV2F{ojU>~w&sDp=2nCZ@W z7;R7D9g^2Tng=*3Ky15_0J^=+o%D1%0d^#n76zcSh+3F;=J7k%7$<=-JL;|J^I{j$Y`r zX}q`=BK)`;Ih~MHNzM=re5%15>$dgqJI{WzL@C{vkyN5H{?=bbsd%DZR zyBG5dc;D&ceb<5TrgfbHna(Xp{tZpY(!T}Do`G9~0X8%ax2A{b)s7796vm%WL|5oR zMO)@c8R zk@vx~9wU&K;ZGS(CZRFY<9{JU{3X7@dGQ@VW$&S~UJQy&3&2mZv=)O};h^R24sRIW zF7@0Er2$6u?Xpp%TfGAJqCg<3i-1P5w5;{(n|fzZ-{SGw<{Om#oiyS>0e+)AKcpdJ z*Fy`2na1T{&?YiWH9I!U62bbhEdYS?u-5@bFY4jL^hf{AZwfI;_H^_iA`R0S0E{A| zGw@nm0}9hp&Bj26mHHgyPT7yf717=PIJg}=h>-@IWq{244!YY>E7P6xyg~Onr0MtI zRx;>rlJtW?dQj4j1nG5tJC=dg|2yE>a=gKHztoHAr}27to~`Xx@~`$n)ngw|d*s4) z4-CaT>H~K8Dc?BuhZ&5f`(;Oh%kYrGPLj#R&IWQJxFF|GWS zk87;6<|!YEnWpsIdqH@bV1fNZ0S>x9zijE3c+xB_lsH7jP5_Xr!zw(Ick=FYcq^cZ*Ld4rpIx1{yvzL=N<|HXR`)?h3S8GA~IoHSJY*fRwnfmj)FsGTtiVnWArD`aX~>17`+~ z|AEvu{M2`ms_;|a_R#$RYHP~MI3I=Xf2VDkpRn3BPo!6kJwBQnw+Cpx0QCcBvB-kv zn-}7g3#!a?(Gl)k<`I>2eZb3(jkH&+1nMRbs?%)2(I%CZykW+Ze@;QuLs31aK7m`msycwMe!+c!Ama zxR*oEN-M@O-6u>vw#XjdE39DtH%bRVTn3&CXwYAPC6^|RFtx#_Y75GUl@XgihC`z{ z<3o;#`U8@XEOZvB?c!wxs=dDeJ)xNX5Oy)_;V=!OX<^h3m4$YV1|w8D_-kQVd``ol)nQk*D1VYQ zMlqci%2mmMsTSh}H*{bYo-lWa5H3LQPOsmxl-nR9h>KiA09# zv|Ezk`VW{3dQAX6=p?BgK;QOJ1bUakz6Wog5ezZlfjq{gVp^s^=dJ9}-{t66X-I~% zFhve`<-T!BLR66hY(#1`>BB`)_6UmNXDhu;m%Tx|Ad9Mgxxj`651F~AC=LGjM^5_1 zrc-w`!~M8@<$GEdIOUCivD{K9GHfl=uREjkl0K66jG5qE(ur0tvvj;~!s!&;lDlnC zh`Q|R8tB(HORj6Mi{UuPhzet%s!vWzq$oX9h+XJSU3mXzD4^G#X46^}f1;z02Bf%PP(=@PKv%ni zQM;d&=|-qQm>%rN(E1{8o;=WDv#4}%;Ms;_*ci>#^E(~&&-Hfh44VDh@mRU~nQqiJ z1`o+uyXDZHjwoey1=(~)8}rZy2<)~rt$hR;k`;tEZzLg72<0FF$%LYJ442N*Nly}@ zkF7yjs3LSmr%j*8h{WWh#fWY*Y-u|krcdVCbcT!z`TEg3o0jM#?U`rOGVl2VQF+j9 zJy>%k=e;49ZUY152;D8>46W!6zV0$}v_g(;B=dVRM64>8Weq-Erin%hBDDqLyX?fV zzP&#W-ISGzAr3)$cbKjUZKt=TzI7gymn0)b8g}}Qh9cZZ;B1x8m*n9r0CseHdWgb* z(MX1AYL5OXpmmK)-?k&bkjgOy>H8VZTJ+8BGMTZv2V={Eb)G9^rq?^Are8@a#`J2Z z&1MI7|FF4DWcYe^BFluU##u-H#Dt{xyD~DE`a*Jg?w_Lj2k8F$&Ge79Kdn~n4#wBZ z&=7LgI~}8(FlP}TBz&ZFgxCq@q-OLBwxTPsz*ul+ho!8(tqb4q6KxP{!Xd2}z`y3<6L^W-R1$NjpLFu-KJY$|3LSOYN;Q5 z4_z>pAIjh-LRJW^7F`TZ+n_e_|AK*wIIIJ@(aN?A(>vomuKJ^Z)bWpe+?gJoP|7jN zw%PRIc-P-pE)Fu#kl(n|+oH610wf|(O)yDkd_KK1UtpVG;KSA^MtgRGi$l{&1eK%5 zVKhj0U^-!fJOOq{YtwdhL=0^lUaO6s@$nCzp7&iL<$65@pKRN2C%6*|Wu_eLz_6Y9 zGt%_m9heD7|4Gv41?g8LeSVPsBX}5a7=kGYtwD7vOu3FIqb&mCZIpoxnkae|y@@pa zB7pI_q<21Q9F-_^s z@kzf2;v;^a(k`zDFrn%BwkTZ;FZLa6HdEhts7vYb);_vU`V!(-nCZF>ER~P@C|&pI zq(4`RVF&;-u{d4M)qrsYnug`~0Nv|z``=3GbwT>xR#+(OrR`~8c~7fNw@z^5QiDTx zx7u{`1PplE)+!4V+Rb944IIE#%@gUO@q5gF)^gB5Xq-MVSM-boIm3X8gEdDYv~wci$*@n*r~>xih>oiJ$8^wgLCZ&(?bWja*kdNIZ_Bi{oay| zPchE7V2TQAd>M^ zdhsCaq%PbW*XIt}<1ZmQ^AMi*A-tQ4(%HqnwZx;X80UjB3UQLUg&q2Hs*ma1VvkF9 z1zd7tL4ePjKD3upQMxdIwn3oX6+jzchjswkC8MB~d*jjypq*7Bv`8>z4}t;`n;na8 zNt+5*;-dK1RdYt&k!#?uq245r@|#YwTj z6DTy_W<*qg9v3CAYXPt)CPIxT<4(rjB+M|(6xq)q-!G)n{fA(6ga+ht2P#Gxx+x|5 z6?#aw7-#O-9X^qgSwxojY~8_e0~X<A94w+Lsh3>7JBLw-viA54Yms-jvOBTQTAx+@0EMw%lACjKFCV z0{G|q@V5y3tBVI|GvGs0^^gHY)-w}bx;bUj{Q!S+Dns|+;kJ}`)4XIP=^}FJJRKJ1 zwPqZwz_|e01etc`vAdRl*i64m?WJV$MCQ&C2GP`qx-54`zMOV;?Wjw{ab@HJ=x-~vDZj?bCAAL(p!V{ zWzepm@^e!dlCR6Cf;!3S^_tbcce5TY-s8J0Z7dF&b-UlJL21?_#V)N!vzRs(C+VDK zo8ByOSswnRM++Db21=xg=}g4neWOIrecKyJa8fN(43+KoLogEXzI=^4CIvO_@M}CrYJ98MrB%Adw~LeXZnI4} z)QE=_cz7S@D7g~I^R*F=a_=qHm(ET*jprLcOE48|LqF_gd;SyxGy z2*d$wN+2oA!UJiJg)ydOY+s;3@_}&j;j9%nafW^tLC5aY*wF@=o~EF7mWD%bPjtaJ znxafMX}m7p53idAUN7C#l#w}%Zt#EmN;j96>@lZn{$(5%bl+OU;6sG7d%+s$wi1`_ zX%_yqp(II{G}-hl%*`%o!i1!)lW@%e9!HHv!pQ<+1#4oB`l6q!U`Y;3=wN0{Jv^xY zVZZ(>rT(oYNxDMUzpW%m%bRW5h4Hk!IZE&2VFd=(F6m1v&eYqy7Y}`>WRLIC6+a@V zAvMUk7F$!YUQ2M^!B!>Y?ffKMpbB7ROZGl9!2WGYUtl^JMVl<`{ zy-UX^7O%8yN6GcJ=D^J-H6r(`$EA3O+fIj zNiKbG+5v=uWe590!HPK|6ueQIjIaYk1A!wbAK?aK!OL(|pbJV}eNrIG3GiZ?3_E|3 z3^3rqt{;|~4>)~yi|2z_+A+Q5sa=0@uy9wJrKQxBbF^geMrqP-_>M0I4Zr9RPsBD4 z|6nQR{Le-zg4?0K6l5DjHjMq-=S10|drFgZ+Z-AD_m(DU^_(a@SL!mYo@3Lqc(`Pa z&GZ~dKo`&1Yus%s4aS{ubO8TW4}L8K2?V|iGktj4eoNrPlZET7YxI`D2au+lVZAO} z#QP+DYmmNM(su;uJ29f6Nf%X*Y-X&dvF(d-tOEZ;{UJW>%K>g@!U6$pzrR6rGq)Tf zBhR`%LeF9-u&9h=MSuHXm##kzTb?*Z%LfO$w7S8j-yPzjMtUfpJq7^)+56`q`-`#x z*(?8j66K!mK1BS*aDX0Wy5zL|w%(VP3Qe}o@n~`}()7HTa$xIyp``y9q|cZ1D?$1^ zj3E!UQuHhgP(-jclUz<4{sxqfa!4$4Z0d> z+Tz25r>6?jwjeDOre}ln#X>P+ilU-pU6vbA|4V#*j&&(DOpnfs(#Ap;olBbwlR}}} z3SBvzXhBpjYKhWBm&@sTXuR?15S?pNp^Jl~)${t8o+!k*t1#f@It*_)uJl-tp*~Tm z#+aTf#HDuZ$XZKD5poa#5Nk-eFTCwIE*Q%@5A&wG1Y-xA#jl35=J$8WxzlEjE#oIK z70UPiXNEsB79Sr#yXLp5)~F6$NX=)FJ;dLkUN zZ8=55?%MJ8wF_|ez7%N8@Rcr8z7_!7mIt_|nz2^wYQa=`PaH2VfXf2Y`w)GY&YI|g zj|+0(h8|GQu6}18=$QM$W-Asv zdS?Qa^8x83R^0C!T%RcvQArM5 z>D_wp>fMJ!7(xg!5jw+6Svhk2@NiiTFud@CY5}7QKQiIN7iRicAY_jC>p9=82TM6~ z@0?Kd10F(*G-jB-DHn_h&wfqv|4QeNa|||z^i8m@`{c{5M?h!_4N^~3P=h=o1)rAB zSK%`-Ts<3pBWolK%XOHNDok7aEPFpQKG2{X+XTAVHx3idzG}9&2)%|h{YoT2EJ80! z`hp<+qNFbh(mM|G4rVq>rB@#zY6iWtiy;UfkQ}PV_GX=r^MB8d(w$-w6TEt?em{}n zebmI1^)Y(~X*w&Y`7KGG6QrdjbZ(G-O=Cun)!U3#*I;_SheAxfvu!l)f20a=!d$64 z=fV2@2)swvc!uAM-AL2&08PP&azXlS`F?hgep5GtY3Xc?usulX5zzpH^DxtkLivk) z>Rvl)L|lG|7^2V%einVl9ZJ2hKdLXFYL?EHvw-K2rb_};woCf*ApNwYuL{zefE!>x zFV+2azp=n{L{3LSx}zSVxMKP}(1r#I%kqP=vfPMaOd*%xp2u{HpY{IHSs#$BH=%tl zrETwrx#F-@DJm{OWxZeeQ7MfXxtwF76@s0AOswx?OgE=9zW?27CtxS%BIr764hk<~ zFaBQZpU=EjMS|L{)V0+~ zq#77u%P2ujyAKl%1kl)Fc%aV`(C-T9?u@s@(V{+Iks7ZGK(E)J#rh0>c?eAm?3ViE z83vT4va|)L8P!4g+jRLlvD*S_?_v=JYG-OT3e@0Ug$DrsAc-lz{0rb>d*SBl?fK3; z><31}hAt9K6jK~L^!W?h>Y4zXKlmlq$fAHL=&firfcXuHSdlz8PThs*lMV8k5IL-72Q_*r+`UftDk5dx1;$zD@$flK%2zur{f(t|olHx#(^u=jkG$jP+I!|Jj59##Vd0amO2KEUeF zK31>P_379cGz{V_72?p!NiMxF3Y7@ca;R7Sq@zjv65uU*xeks3MY7@DOv@$(RV+Rw zsNx;JiWd%8#i~gz?ewcS3st54FF%T29te8* z;S!f_(EW&E@nDHdH|3M}m$-C`m!$hpa)-w0V&H`M2F;?&*i3hq1WjK3j{!a(`S@(n z_#B|GA1`pJUw0W0_yhWSP0-hm7Pz!N|MlSlmu}EWda%HyoAfi&{V29g<94}++wbOk zxZRuY@g4sZ;CA{^kgc*z?{z{;oO_~f7#J;+nKF7XTrnw0+q|B_ z^(=b$_v>oK3gHQ0LTisqGCff@1e5+?Y+NwO^;tm6zXZtSd}JP|>(jARs1K+-CH5Dz zb#jib6zzc!>1u6GHHaE5y1y=yk7YnlO%6JumSmaQwJG2ZE+J{HDOoMs^J$xkgHP&@LGo1Qw0l2F@a07Kw&WNIu&V-}T zyHF4@ZLwT&OT>_1>et{l3OIAhC;m7p{%!zot%k?cSBF7OPaT23UVzC>N0ks(s$iZ%m&ywV zhiNF|uaZ;$BS7IUABA6Q6kgRR{Gv{FXQNPf6)2pu7YfxC@)niD!pM_=I93gTjB+gi z5uw+7WPa@-^NPGE_vyRwKLcbQ*YsU4w{9dbmdzw=jVokVk7*sNVyB8es*O_Ss3g-z zwFvu|Im*RjAAOMTwSj2?)EydXEyD#5h-STwKA!akUkuY|XKjWZIvap@)(%njsGLtQ z;|Bq>H#9W5tU1c`AZ(ftd*`o)Gh#UdIA5~`I8H!$>d><#a!^ATH4n2S9CqkWXcg0W zM`23PYqe3jX`DCVHs^Um{D%Qf>%Is!-Ou#CL@1?|XelS%Zwwc8g@lXp03pN+ z(*VHG&^WE|2UtirVxtT2TRpl8H&aFQ%XZtBRu}&yAVpO`ib+XTL@B#cDLPvd;%(gK z$@FIvXVwSO`>P`Qv7N8+*c^0Zcp3HeWK)0Y)kJiYLRHJT5>Xo@8k5W_C|j&5UJCCf zFD}G}e4#s4X+A#3#{B!c01SfrhX}Ah=Mr3b;kJOY3z(8w2>p^bg<1Tw0FRBML9WWb z-wKe}L^=ATgy|#GrKR1r937_n%u+3e@5~pu98p3bqgSgiozrdq??*Iw(R8b6aoR>O zO?D8xgdxta^^x20^PrvU1LQC*s&LxiFcl1f={UmJ@E*G-{MEx7+Z8KI?Ln;Df&U8N zJqvjBVpk@xPN3JY_W!12gkEjV_?=fA4NANa#uBiZ>75?%ZaPcWNLXZvS2IH!z(8BaSzV2kER7 zST}0+w63Sj42rLi;t%QKH|pY^YcmS?cfjWh{6Y%ctqa`oX$9i3poS}oN{jljTa_(6 zm^G|dEQ}%V_y>*7EbOD*VV84^Xpjl<%8B8AEf}Xe@=`BISvDiIb3UlMS-fShko|Z> z)WcMm6wS(Y06EXrYF7tgNb|5PYL0S00tS0?Z)YOF^%>8`9aTZoiHjlVnNE<8E)}^H z%0R83pJ94unFF;!`L{-u$FfO}bw+{8Iw>nQjWc{caY|iKY%v&O>%A6 z3eo9;bHIa~N?n?h*)kJ~;cA@G&!o~EFU3~9d5F@PN-E9a_Ug2)iGy<;cFN;bOqW4X zd^=pC9yDEg3f8FrTh0_pRI@Wet7$o3m2#)Ib zfqZdLo0T(U|2~Y_0pP1;#q4GJc)vyQoOGW=T+*s)f2sUMx-Y*>t~yl%D$3ud@@w46 zz*)bq1{#7KBT#caOc=xBOdDca-<46v(ot|$SfhhM;qAIE9CtXp*=HQ4>0u14@2~uR20I*T7qT3pdjHHAp*Hy!w!li zfVDP9R3VGbgXMtf*|x4~4k`_Rd)9~{I%_oWy8_??EWlTe0)9!c%L0740QXGeR;Bg> z4y-(qJ*;&y8W^-RLbb8>QenC&gzkBu$2Kj^r%%a$TPeO2aTTI> zWx_b@_!j6bUaq1nk8r;)vFz~P=7oWs-kYL{Ji=LjEjVSau0KC|`+d|lc%LA^<4CBV z9UrseR$U4^4Y|h~_>9+B4|v}Mq8SpDg+X5nx6-Wy4D3P?WwZ1QxhkTsy->jb(K$9b+lilxRP-thS-OZ7hX?%4!$1*QZ*OW4qSHqhB6Ks? zNDk{9+}{~d9>5F>slvs}??zQ2A+$dd-?LRA!K?y_e@6nI^ho>%5+~zqk(mAydnWV8 zyVy~MY|%po7Rwvc$Rfc4;FXxps6wRVg#0GB@2;(6SdjUWgi#)#S%Zz0TtcYE@i4=+ z9nKy&`LD*olZZW#r$_zJEJkQ?VuUm=B%D$qaT?Ho?F|xhkiY_fL?aTDy*4!29HIG9 z1j%W#VQ@acm8y`AM^k=k1N!kE^q<=}`eH<$1b#HPPcS3`j+7qqS@+J0ph26BQV~)5 zxeC83pHEM%46?k4P$4Uk3s-4Dp7Ao1NSD+zuz z@X9rafOGv611%s;Y2vceO!Z@mrKp>NzGO-$F?A3t%G2J=^%=qQ5}?CC8hg0 z0)bx)AgX+km&H#EH^AazQ2$0au0nUs>XVHjI{IjAEgdLv>Ntk4^pAGABD%|PRU!R@ zJl`$Pr^xd?@_c7n68B2t4xK=hjTj@G7j`c8Opf7JFi#e4sR?ktUT|K)u>7;)Xrg%n ztai;KKk|piVmHWra|i>zU%Uk4LiU*)!7j()8UzP|4uD{|f|E^jlx#n&NoY|OI~M%< zLN0y5WgSHZR;B2hI)*gU3jqPq7!j6X3p;&GroRmXA>#lej?5sDRVPs((9aMbQATJk z|Eyr=qh-sMA;RRM3?3HYUbLoQcWs>FwTTHHOL`8&#%chJgL?X~!lOF5>$TR!bs+Q= z^v(d&e`I(q20n1ID2&;Jo!ZA-1=;^dx7x0*WL!3k_=C{4nMd!^TOY}X?WT*IrC!b? zxDtX!ShJy62p;njLj(+7v*w;owxrl{+7;NBs-SPCU&An9}Xetxq{Cl1T{4@jp;@( zy-JN`0j_VZOC=kcXP4C`+UhII$5OPd9TLTuY8#uh)%JgLGhJ$IP4Gp)mQ-Won8>xD ztu8UPBJE@B{)SZ9tT8pFwqw@Vx@1;U`53N7j&G)OIv9CiV4vxK~ev)!AEg5wNh zVXC>lp?Q|jXPjZoO0_m6+EF}6#9C+7*|uHfzeN3Qn$Fm5Xf7fRLyl3%*S1NI=HK|= zQ+0DvEm8>p=u`k8|EfVcr_eAuTN~Qz=b>~V?g$*+7)xH@6NoQ_9dAglc<-9)r6Sxq zh^}h!QMcpRghANU>A5_)U)u+lx0+(e~5&Zz74aBo0{yE8Fzf2gC zS3e}cne3=Cru7nRBc{Ui=sX)Qt}k>Ug%8c^Q{k8@Lg=tC4oXy!qr&C#ucL~@JP%hT zK$|NU>cuz#)iF%@xLR3szu_VR9KOU>;njGRbiN|{2NXK9A(jcf2qP-q{82O72YBh{$47)MJkkMw&<30 z#+Kto$k);~6{ap5PpIROzf{NdQpclS9oI`8k4hcaOC9nIBO+{Xb?`E4D?|GZ`xYu3 z)t5_IGESul4-2gj8~U)`5FDokNlSxjGt`KW2p zk3j6NC!v!4NuRRl_S+M{2N2S+Hq6VDAE(VHKu{uCnM*SDEx90Qx#WCi>96qmrLe(rd9? zVD}>gUYM8z452_Bh;hQFH^QQy^hJb7dVKwGSQU}jS%#TT#_9O?F&1UaQ)LpH9n+<5 z8wk;-5H@T*6;W2N3iB|UF!|GKM_~KLVcEt2U=LzChnXt3BlVq(GI2WVu=!PJQ!J)M z;#;A@P`1QmhsbbQvY3jt?>qfem~0sr2xy?f)DB9FOS1VrY8^f>CBx=1AIH~mRW-Vv zDZC#{#>_!)(gHx^BurI46&fz!GxsP9m0^dR#mGh^OCbEOfrBVoe6m}Om*16_Dhz!M zF84nv}<7$9f6?3&;(Ie)OInB0dC4)XuwP~;Ao5y^Lv!Q zNFM*G-AQfoJ*mRfmI2^#Ikd^)QM^vGC~5=P@fb|hJ}g-=fZH$*#*3dry9^z9+a5)R zqED;V>oLHxsAWhr8gZtAoYy^v{e{-e0>Xz&>s*?R#Luw%7g`@D^#8e#8hzsb&<}^W zMDjVY-)BGDzJFDiI)aF_t5c2Enhe;M`2Es%E;sOAOneHiXk` zsa8bav%!0q%CXshT6>&+2kJD;#O*r$A459NiOBP$e)nT9&sXG~990;F=9FIqm@F0j z%U~4M-)7-B&-3KIw-ZE3AXs-&x{|sx_-hrzk<_u^XLh_xH>Yiykc8hGoZ?vT{j-tk z2v-8+8Ve3W7>a-Flr{A?I!@b5av&U$=+2Iu=|Y2HR_tQX5MIiplC;PsCBd#lX(h_F zZ3vkyc@(jEBxTc3dxjBX3{8|%Z7c^49kHG1RJ)pE7NRfXG6kq0ELf&Ve8()32n1;{ z?dq|i+6^#Ovpz&Z!z=}iN$=(Gm^L9uD8js3bXkx6S8LcNHCzs$7Tp_y%WbEyIhgDWpwr|2oXV;)+GFgM?{_AKGY`N3Bf`$6f%Tah z98x&iBh}LJUM&(g84D_Nhpp+p350QJfY@dr##t<=!Y0u1y)s})RF!asb(oHV))roI z4%3%lni9`2;i4C)d*w&543Xr*{DRaX#%y3?=E)y2j5X*Wa0nUr=^=Io!HKc1>$3;c z4E~9$WHrcMh6 zm&zLQwAiWu6$i|}n&}N#M{Rx`6^_!Gq>Ut7u>Ty#W)~%3UpS(4XIe1x>9-J+)icL2t_|q2 z4m?PWr{e(EJbYKRV*0PBZ`iQf&sac(;>3&|f+!f{_)`B!(&+P`70{7jJn%VnIHJBv z_sLe$Q3Z5sH&!{eO4Lbo0*28RfY7W>6+lMUEDB6boy0h&K{*7Iw2{6W=S!+UVw#uH z=@RTZs|sjD3^ZUwJfljLQN|om1x!mZMlt=y!R8a?I^-14qbLTZtki&vIJ4+Wlsj!$ zorI$ql$ebxR2|K2`z@Lt9BsC&4Vnu}UG;UEI}C}g0MT@P`du4aPu`Uk4O< zHBS++guUqjoBgzn!{3uwV;EBf-bMWu;tQSx8!)dC1#A_old6Dqg-nR-QrsTaCpKD; z4KnPkWEi)_*{O_JpmdfN4*Be&cpzxdW03jO*Xf`A7Da*tbg9KzbGWQ{WE-bG-H%yb&Ad3{R3CTG{ej4X#ojK?9# zPWn8DB+d2hW$pFz+gn;w?PD@N82f2S+ZazI#;ji-n>HTklqK3zO$~Kp(G@q>k4cFS z?d4d&kMWFP%&PUa_SU*o^SrU3;u}Pd{4qDk{Q|VVqj; zYZya@(bCXT-`LPB3RCFBeN8w)i9Snf&PS`zeVsUIP54Q(QwEgg{(wh1WcYelsLLCg zB$KtbvKji?K-nB5VxyGJqT5Y~*0^AlhO|%{$F!lMBEv|uryB4$ZZDnacuyxPxu7*p z)gcobMniM5em-azYfiK`%oEfmXszifwVGw9Fd-HPgjvYIvhq1>LBw7ww%Vd|;kCia zR_B#7sj{@FpZ<*C0=Q2Uw~j($*+RcSMDgvcFC>I{D;12u&>T{fRzWPt5GT!JXu+_K zqJ+1Q*r+=y&KnHQYiZGjx}kTo=uKhlSuUsZAT+!tNjj@L>hEWFCIie)hb#hsxkxqJ zj7Y*bi$)j(i)k$kwGn9ji5~E5*k?dgSR}hY(>t?Ktf)0e zY3ORO(-qGW!9pyBU`xOQTur}A+c3nSf0Wg;yqf8o0QnD5d7yu=nZB9x{>lfNzL{jR zoq8a#So+`|dR-uFVPSNv>zAyU&Pxv50N7#B&-$iwmWo~f$bI0nfYf4)u^i+LBSxoe z)+OS-!xDTIW~df>ac-^2kXyu_-GOP3ZeT>2p7z6{Ck=FxiJ+*y0nh3=L1mALuuWDD z+Vd@m3Ub})^snd-ho6pX9|kH&(gJiPh(`hQFA|APnvMZvkXu8ipW0M4NZ*cDFn!g_c$H>WvbPtshHaIvT z!Jx)`(N{LpA}9w}q%%CCO!}H@Mzj&`s+M2~Y=WB(RJx^%dtW>W&1Bk5E@%SLNUI9^ z{L$JnHz@r)O7lvI=m8*>T&~W-4Sr?H-LjE#c2*GQSTtNGU4zxAj(b+ZXR*; zR)kU+-=b-8Yf#bY@Rsqmg)z@24W}9$(NqR?X0kKDpU@9%aNYGB6Q&Z2;{Yobr>+bY z-dTD%DNbQdw*mKH}6xu@d6FmyIC7^Ug`ra=8kpD*@q=m;=A zk4uaC0U1}1Sai%Ff>yy91^g6)I>zQ#chgTZ{+4IsJg@9{rmvN9#Fun;`0q{nfUp;$ z&>$~E#}9ytn;Eh0zZIp3%k(V}1Di;#I5ZE30?EP3CVS}3Qa57s&`~axxHK(_&|$~A z)NCUOC&(5cDY`k-BZHulSL*G32#Y}k!vw5PGarFjq6ekkLs!(W6RqJIcD;T=k2%mp zTE*DiVz1|zQ$9{mJCvtMC{=%wp|84`8gof#n8`$^Pz-Gm=)nWF$?)isj=u^@qIwv>DJsdJcPlpN+PH3!CrtGhI{i?VQb*eo{gB?Z_WoErK(y92RzXlVW{ z22KrQ+D&{0u{4icLvx~a!C1so&$8w~M$&l`j*wwAh=2%l==N0Nq}eHmk-9b{lwoK< zNQAXiHV}9fD3kskjtbK#^;zX%3E;sT4DR)Z1C|_VC~> z#~61^_~8s;imC$}su10fj>`7aC^UD>(@qr1q?J&UKpRIPx0?H{Vqv?`FH}XuN>Cc5 z=P(!vcEtEmXnO=a4$>ZIwN;iBNP9R5z#Fi`UUsN`2==ddgK~IFl%u8$Q&DaPnYm@O zFtcpQQd$60&86^z?6iFcoQ(?tL^pVd(iQ_VGtT`ufw(F?Oj`}8RG2OLvoQ#Tg-N+| zA6=ccX|o~fuJ;V+KTTShjw-nUrBd?PI-yMS+qAA{*eaIw32Mo>v^C4lb&Bb+P>v-U zo-%1&Pt?cF=n8Oq0Jw=Gk;7K8#;iX*jO!M_SJtNca8!=qYkD2yCB9fcwL1VJH}8R* z)+S#BTb2KMy@+60PSikza~P-ixa$>+5>P3O#UH#lUS7=(^g`nO?1-|Q2O_u%3Bnl4 z3e!T~nOH`sJ1T#gqE+nB1I8fY7y+)e6&e-Cq6dwfcetcoCapn05e*gcp?>T)WMm>C ztyqo<@D6X#_0~#usArhEGa06i3=}F>Eyw74QJR{h|FoGZl1wE@>Wos6yOPp2Y}_a3 zg_p+NaSKiYt?hT!q(KBk8a7pm?muH5%mRmr%QOZN)&=Igs*=dDb7(sZ$m z(G(Yl>v9$rL53^3@$~`It&CLC$u50AS`PDmC~gtk;4k0^gKCEvR%)wDXsb;!jD}eWZBkwc2;c#&mE(XS+yK~O4 zBD$y4MdG$OHa;GMa|2oo$q=jnUv{*@7v&Oh^^_OSw&Hvu1`Rk-ALgktX{aTN@@pr; zRA@62bglkm(NitMJRPhIY{x2L+&CKHn0bC^_fmL+G=mdk4+gg&ruN`wkEtv4G*Kqq zt<~b0mNx0^#nMW7EZ@?sR3+t~oX(ZTdozlG3JY{08T<<2V#Wlt~A{ ziD<5VqIw(Qs*`pcQN3}5<)FgxBf8;>K!`IY2)#ZQs;HzgWy7XOoEbnb698|K#ipE{ zYFS<7!$UoS&P~2T=m8vSh^l2dJJ9N*uKS8|I#iM_a}{o@N3c>C9ti^G3b}A?o49Rz zxh|7e&y0R`iM)Ewf5mjA4jlDpQPPXWq@n#!Re;Uy5>x>a9D?IvxNaaZ8H+F6%r=+i zTyNpT`>QeHXbT*^{B~S9M+VVTrTVfpw+R8xlza0$RNj{kTAA;l_dq??5x|0m?UM_Y z+}~kGuwz7w30$4g;#G4DmJmx`0}{~q5>!5jIEBQKz)2zxnoNg7nnb5OZ9$~)aJs4! z^XX}EYJ(s^my68|0zyH^KX5MKw0mCkfH-|c zH4Ur5meXlz(y@Rw!-Mr@z{&%L5Vn-Fa^CA73X1d>)3tnWNW-vUXaFZkkY%cj)ul86 zM8Rer(i|qYgvdgUoD6GX+BG?c^zufT)VCg%tY(Mlx!HE)|Hs~+fLB#r{p0w#XKhvv zNfcDHYQ;i}t=d|GH_5pPN1#PJ@b*oiZ~Ok*Uf?F=Mo9${GcYB&K`)R1QE_UW#j(~X z2@-7;2h@tz5sk=TRMa|8IDsZY{?BLabN0P~P>JpD`~1Jp^J|~rJ$qbx?X}lld+oK> zR*H@W511)&Qb*utiNxSJR{sQBV7lT+t$TdnjHI|-e(fGux?t%;y1>tU`hb!hYzA5r z`v;T&>z^jkHFn7Yot~n(C9`A|b|2>H`)>?0ISVFpWnk10ZYAw#Abh5e{ci}uxk>Ii zHN|T@Px0~mL*V(>e(CQEo@u54a4e7wxbO!d8vnIV*a_9tG9l7`W@A^mqFPY>&upoz zY$W(f-Bb-lV(jpuqoX5olS9|d>tqT-;&Br@^h9PC{;NOEp4;hvDC{ZE)po+fb5t+= zBph7-?)anGhnC^c!|-~g|H4(l0fkhunHI@%|3oi$9h=Dp!vrZ!!Y~fBp{3H$`+XZ) zUMmWE*e#&Ji35S`Y-#Z0z6~zv+hA{IhmRw;gW?}$)9#wE?MHQ3b(uj>1SgI*D4VoC zXsKPPk0pyP_se9`KX9zW^x|MOJ%U-zTa7~nBGqVpnS&`UX17_?48{6vm+9ezm+4Vs za`B8fj}?aS~j{?zF`P5us(89}Q-2@|C9JPhm=hFl&x%OEfFWcs^H;U9&jE zZ9%~6xncURC>`Gt$%n>+UE+>{Tnc%4`7p17f=!f|#SwPxWA;R*=NH(McQXRHfZlFGpCCDiBk z*477u_7bgvRW_ZB^$LG}5hOImg;wPRF1qVhxpG*oFz7P+cT}|H6$5lz9Jt{IF2x6O z(5oz|nK_X115_0c*e%i=JZCJZDT6h_0HShyEtkrF+kdg5T4t$VGSrS^Aka3&6EH?MfPoRIym)5~Ycxm*@Q zNpM`Rt_^B{2vAvB*6fm+(%`pNsek3MytHC=Wl2@K*+WLkHO)zRnhj1joyfGTStZpq zDWeIAf1=&iGW=C~i)LA3XtfOCU&UFzzkgu{j5takCz8SC)reblZDrF#AuMvah(rK) z*IAagW(myoTnCqLa@vsdWdv7|)0pIB>N+dJYi>&~21OxuPnTUMx&*2&$a{fF#E8|V z!*1!YjqP9b3h%H6+v`)gUxtRIMaD>__o5aUvjc0iwx#Sk?fJYqXn$p!Za-68?;(cT zzwWi+fNu@9i`z}63GsD-u<6OcFB|>eidnU@%c@8J&yW85mR3QgwK#(KWLMq?`6@&Ckmz7zj#qwSm+?jIDu`_vMcPu?4A9RWc_&kdE|F z1OuZh&n%xmkh+HIORDFL8AwHhXc;}bc2-S!Wl7b57TOG0dmi8dn>A~8EVaDMudh|R zbXIxUoB=@!ZuPagnq3hu0~H_0CMNvmsQj8T~aRL1k%JvyRR78 zN7^(W>0+d9S+go6#v7!}StasQrCG=h3d}5Em%E@+o62VBum~p30=`qhDqw&cj~Uxv z{h10@;aJvwf>j_~gC7E7?>H7ZEXCfzDs3yO$_`Lx*0DC}$zX-!= z0QdnXQYk@xg$X!;un+pnG>}Tt0VkaFfD>YaoE~rj2F-o|l1vkO0=Sf;0)l4vLVSj? zfd(G>#Gc5&VYeg&K}LVEC&c7gTP}Kw z!O}BJ;_<3JiKP{@XIBX653?-n325SQ%hEg&A4gawTH-fgrY&$EX<2jhN$5z+Dyb^3 zkwedsmIYpnT1Qz{*?fWLC*~1@?`f-?XfXa)J18%c`oV zsPUHz@(#DGxn))5r;7;`a8g!P6RWJKkdfe9a2L~x6jT_dkDBq48Zn_d22)*KSuv+t zCUOYKrQ*!;;G?pnrlzb)dK!XAH~!XuhK@``uhp}J%T#)_v7rbLsvHTD4VWYC_TK2R}JYkt3%mz zr|$G0x|EHiKh2X;Z#f>l)(~a7Qp_u$-M@SuwEH;I${DkmR)pfvJhDrfaZxUJP;Hc< z5u73CVlg<4;YseZLvdB}JCfH@!TVXF~zJB`yC`E6&sb zD7GyMhKp1biNH&SgCkRqKAsYaAjlrceKLp7DHY8RQjx;N1ecQMPnFRYq-7Q z2IQaOEsw@<#K+|tB5LBl_Wa%cJq~Q`fV9#Sq}iqimtSucR2FH%LD zq-eVUM_-0|gC>?Gdf?f&GX(p$2%Rg%K*g?o)J{BIX^$Ym2O8D(e=y=&i%A4v&A#Fu z|BbS1kD`ViL?9U?Hj#Jh3to^O%_oN9|;{Q8QXN>v6ujGi=v>%>z7?z1qQSHE@Wr9>3d zbpzDs>bFcUBwAtZi4o&30%ftX=|;4HfdHTPr7-_@X%GQgGMLt-(jV4?5d)Lduv7E= z*HqER^(1v*lGu*W6Vjx8($2Jc=+-`o+5(JBpl5XlNA1+}36eGN-I8j2i!|n1g z&dLKQW8MGN?NlB>ne7KqX7c*4?rZ%4l*xJEtJ_(10A+?g_|@$!3d-2@Ds;3G#{$Nq z`M4jlPus;!)R?q=Rcz%$ev39CFgBP5OtkGWSeXt~yZ>+f1JCbRs-x8B(j`=x z1_<5&i+?@koPQo3NMHKQ_5bPfaE2TR-UfI#;G-Fb-TxcZf29Lg&HEa|17|6-26lO) z^?V23Jj#y_;tzZc#-&fKUrV!CT~kq2a>jr-M4%`XMdLu|P%8okq>MKggnr#w2**1B z%>Ey`xzhJZ{W*~t6*CkDeYK~0yFogoYa|r3B1QqGi~?*>x^_#Yt4eYIf@V^xJLo+t z=9mDvMjeY%=3_72tTw_&iNR2n`5H2OJNPy~{EU>A42sCTl$)9R)w?39?ml`pJOv`U zB6O0K(z9i!>)8%U>Ddl8dbS}(&o(4Y&juZ0$sEyI4K;eUq3L=ye5CYjx)oo~HbgYi zvjw9=()4WTU|&7k5CJ7n*y*}j1$c;M1v$0M33xIWD1TD&%8I=Ae7 zbZ*({Iyc}aUFU{x!DE`v4XWLWDx-572I@-b+*G>GO+MtF6Z3Gmm9BwZP7YHLPM5ZKH`dL8Y0#6V+O>jw6|CeW$-UK8kab4`%e zn|~*O<)YUX5Q8s$3xF?_)?Wf}-z0q}Fe3dzb2%k~LWpw)a1{S*fB z@_^h}e?sB<^j)UWS_|voVHQ_B6h2ZOw~2mSgz zViX!?57n@343cb$8~G8}xz+$G`r{l(6>WJr zkQy}ue3u00gn_h=^5$IeHIU{bK~MHT8W@lrjF}^ZR_IrVE~~1VQ(-nan3?K5(Ii2p z6>+am_%O1pnH3dhno=RSefl-`zVwGR11`v``^yO``J1wBS=D88(#>Ae1d32!fkJ<# z1d24^`vKu}JW?^M&m$SSfxQI^ow~0;k#@pReeE-pzI)Yjr};=f10WIo`;TO#AcZu5 zF>7H1U&zy6GPJ%)TJ(Tba#isDMJV+o$sjG7p$ezTh=T-ex<+hP`3!N&mAD&o%S!un z;$uFT41ltrI#xQXqHKOyPzygzCEK#9<{1eO9hJ_hNzqP5aFlY4Apw$VbeNC9Mn+}G zQu^T)vaG7qGbXTL?n6}=o2{jvl@DrWs4~;o;i@1}(V@dQ7Ced9o=P5;X2QEZm3(+w z^2Sthep(~jQ^{^x=^d%$NLsS>kPlwANIgxEc6#!d#9(3tfos-b%wgst@PGcbOW)(zW&y za#@~G9?LPB*GZ1&$6V$}A@)fR!Rzb75rh-KlYLZ$t(F|bNRii+V;rMd27pBm8vw3^ zdOZMuZ-@M{0^%K>o`y5x)I#S=9xI_2aHDIUyQ69onQqkv)7q10t(h)ul$+>62nP0W zXyNGFQA3#i5LUyPUc^Ju>KKXcBu~WQmCPY{7*a0DBFaXDG}}`{$k9>`rp!f=OntFR z)k*!G?SC&!a!d`OL6E&%#GMiXp@!DLoMFO03`B(z_*NpGG5x?3XKVVODE-hw98=uQ zqbEF#iQ7YoOvUs|+{f09d_xc#gO10)e&@fA1`)`%KW;kTgU%V= zJua^ffSd4y$;ac5`{b`m<+osVePaMbq&<}B7$F?`lifw{G=w>qF16!KuQr6~Dw(ql z4FDH}8;c+SBVuZeiUa0oSblr-Q6h|J{cy*Pu{|8#UlFC~pNX)_rYrn=jxB$Az}%a;2A}qp6X1Hx1k$N3y%QBfOTdPVWN))-Ij$55n4S94@1si zR~^LkQy5BQF-c65Vz}&~xg#~O?WyKAp*fHXsTMVaeuqUd0~NmquQr72mcW)Oj?e-z znuV31n0^&jS#%N(Fya_-TM+YT(K80nrqHm=Hq5}|RtFmFXvFg-*$C(Y$LlOS_rr@- zHUhlpcekIB30aK^QCaN3GQ~`XfP!JKZH%fcdcu-$c8=f-K~r3IzQjLxBgJ*-ptvGB z#6!PxFeG|Vd9s`1c+p&D^HY z!vHf)^Vwg6kw2n^3<48y$x-Iykpj4tZN}q;=U{Qos(qVRtZH z-VmlYZI4yg(@j{O&M=zOnokGebrGr$qsQ$HKGv_X1?|t_ z8@QoX8{V%~(R?~IK7$&uI5zBbXkl>rq*Bw?h1?!#Jjnpa*q{dqGRck8(_9(LE}zj-ZD?W=!i; z2Zk!l$e}yyWHVJ+4*gxlgZ*dK3uZivHOwh)WObRmb;tu?tLlZ0UM8%}Y^GK9GCTAd zX*$y1iuYNfFH;{nY66J#xe}i3h*x+9?2fkS{d1+BCzVAXDi74++{!N7qiA7I5UFVW ziw3&;0bTk`jCjeKWnK+0k+1M!^!x9LTuCif*Rjpb2K@MfK1l!j7 zmrZZNbNfKD-On3InLd}>7V1(XnQlyA)Xf}2!^3H7sGD6il5S0CUbh+QBp@<ZAfU!IRv(OLlx0oIwq^Dj--cWy=1dP-@|LlF?FOv|6>HCjc=OHGo|y? zV$69S2NQwqoS&O&VXiP?-pjyI@8>uT?#>ByY^gU$5fHdHG+X;>_740}|8vTfZsb!vle!dce_f zf{tou)XW;tN=z-L1vOE@dFNSSNuR(x|!C(T?FUiY(nw`g>*uel~n zSyOJ&x~>K-(=z6onSISQ1-u%J3FXh!=9*bBewiaK(%F`U6CMg2)c4N)p!7R;^O1h% zeo$aqd$48sj_yN@^=HOD)}I;a)}L^%Pq+TWx17#m)wSg{+W0YKS!b3lh?SgCRaQ2; zOpi46${8H*QwE5i0-Mn zbY@w+)`V*+t*k}np_b(zr|<>));T(q_8`mBM>M1lvr?g+4!2Fm=BUvG5Z)~Sx7yrwoFF|9|S|u(@Sb+)dWzM>9AuVkaaMSlmRwctvVerz<^>_ zPzgseB!0uP&Z;dh#m|t~y`HY)jKiP5tR!weaI``pc=1=%Xg-1Oqb(~q^WgKFmWAUB zejc+gJh7Tt)nmfQ>;r6dS=C(KQ9sFj>S!WB1BU?E@>xEB!&4-i_VZZF(%b|UhGWsC z0Um;dsE$lJB71rLUyZ^fClvQ!VYtn}jw^1p?cbmcW?R8{CAh&qs7IX$c{VV|SEiqtmgMuUXYl%M6vM`#G zJ`o=d78?+4!SZd(sx6;WGv;KOv~TN~(v25i=R1`HD8V{_`vcg*YQ1I_zk9lfT}mb6g1(6UNvq?1EZTi9633XF&HENk}Z<#Y4`@+=FZE)#rU z4o}X@kCjx*6rB>}#gK}oQ6L-WnFbNI1;W8dbuqy$S6*ww!t=*?Oa^KHQBzy?9h*cTFR!v7R5A1 zU;V^zYcvA4-pmxk^(ky+h!KCnMcVYVv~T<6rlqBQ*H8N$0(bj$zvrj@HZARVKkffX zOFIE+gPE2$M!89y$PQg=RJ#d(J3n^Yh_ zC^rq)_)^B@&A2yFx}R_@@K@wVn|1p4l3^bU)*-5Hw=RbTxIFTHxtEyx?K%P zSmUb@o!=M^%q4Pm>6uUk^gxCf3dPvT4{{d;xwcpl@n#*1!fpwfD3>cg_Zv8ow#dbG zo7}>ugR?id`}@`p6ltJ=5j#!|$p93q+waGw!2+ktOU>O{D(_`n%Ob!Cf{#f^tGfKvIgel{)^RP+J6&mz#=8pL6Au_6pjc@axY24U`rqq8xvr|1w0+` z^Ti@1vKSYcm7ORS;n?(upNY{+=Mh*eI8=`KED?^;p>diWv;f=n2#lv#0WrP8D#8W9)Fz~;_>dfXm? zyLI12DJS$zD`Z)gBI*h{L>Lw?3=>AI5ukG{UTY0Uri+>Rp{*gThS0q*cTxxA2~%8Izh`+Ik^eEUelSx@W0GApldh?kIQlIx zr{OJXCVdRc9x%#tLoI0gxZr3?r0>2RB_cn<>(wvcU*Art-ux5ljaU zKXhB(SXf~ERC>q-tSBBw<3EjZ2^XkrWX@(7KB#O;HilImnU^aG>7HMRnDYG=9<-#1 zdmE!AXdsFPnnP${S7Pr5_P}D2V<_^3uC}%@%EgR03@&$a6XT(a!{SWSqFyt%`Ok^c z(U=qzqc|*CVO=m?Xnqvd4cGx?^nxq|8A{R4OaTh73Aocroczdb%kVypQcNFbcKOAK zsk;V?6|AWlkx=mOv}Kzj`H={57t=JDD`2^Qzn5b8O(d*gX+Tv&GYvTe@g!CYMc>Y` zwJn8N6E^mPJf=g}OL~jOWuqys;LxT~yr}+|?In znLc|A2_7b6l4q6iids9zmf|0w4Xh9+UuP~6aA3gzEIKg1Lm^#pq0rNGY>b$q%H)1X z>G6=qW5X(&y7faqOF}*^R0?qpiE}d-=BjMOqe4SpS|Epg>FJ4(XNnQLlKf`sq#1=~ z3i4DoLg1sB{ZZAEA+G@Z0(4cPh^N6A*tJhZ{3G($v*vtQ(|E1HkB#I|b*ErwsYU=V z!CZ&BC=o8qWvWY>X&&mMd@$t?%M`;Vw*?}_&!Z}v@?pF?m>rq~4>PBko7g@L&RA5R z46KiG4lykhz)}O8Hy_I}H>muJR?cTU{@|h>rU?(neiDMD(xe8{(nQ!i^SV9Nnz72Y zN5UQ&;(>E0OkRZl;)M$sz92A^(=7jS8Kyu&^VHDEZVN4J<+i*)#9Mio59o<>9{G@h zfHv9w3mHR)nF1H-0tLJpCGdytV?e+Bz&1QtZ%NxzwXQhuTJ49MzHX3OWVb9~x;Esg zpVA|BVRa(ir66I_%L!OVK`4N1;ftdajxg0X$_vK8x(i_m|Na{PI!*aOtfL9vFs0k2 zGhN!59@uTQrwTkqX_o!#Q|XUO#WwCWlyjtM^ojI3BU0&4NgWYA?1G(I5(8*U>_s7s zyvn0YBb{Gac>U3)@XP5vZcL?bmqyZ{9Pv$4=auw2IjQv5q>ext;ni>u<^0G}SxxXM zZI?BKG+xe&liZYN*K6Qc9b?-5@awiQ^jOnIkKYD1?-`)XJXnURpVImDAcv30Q)B1} zSWQ9Jr#tY?^cXUk?jl5xa{@O-YV2eewEo! zz;Lc}2seRs`qN)L%JlP^g7b`TDZVI`{s#JoAO*$&Ms6NZ-JUA+eozPn!D@8D5t_&* zALsX?=%3mQk0XUDn?AtPHcB~k5%;RmOfN48L!@iME_h z_4R&88xH8SPaF7H81*mP~*jjcvrI^lVPZD-42xH$@idJzi z(}JWrg>Dc(Z^iVM)2mLQtL7!S4WF+%adir$OgADTh*49=Zs5`TWPm^CyFzbjC z*MTtWOyih6d>1#F;*}AmuEd3$n919b%DXT9Wfg0zsq>1|5wM*ZQ?&{JEYm)2 zCl^BA_B9eaQdM@N3ebS11)q)O>^jj>qBtH%wmm$A?T`)S5oTknEL-rnP53nkJ|F2` zi8-ejU@j5GSd&D+V0mZEj_@)jIf5>ciMX{MQT3nKKS)k{ZoF`d!M zZUG(9B3`95Hw-gMaOmKyiQ_LRPZiKALc9f+)+EC8 ze8}U~^p`}KHfp68wIy0H4VW<61T_acbbA79A%-xl-gKrm$WRU)*+plB=^Gy5RnEgb z#3`n4bl@RJIufTQX<`yYygZNbHerk@4;dK#QjQ1;@EZcGJ#v?jszU^h8Dl!iWBO}j zkA9!qp_v#GRKrjMED*Q{bdQzsJb8q&jY58s$xo(DX2k?C=9u(}*s1 zp`97gB0`*K zjfD}7Gg}de8V6jv08mCy59M*Jpu6E1mkx7`=^@Lb`y0dbpyfrN zP?O3It+!g(wMXuRjyVyI_*ih_nojKF$myo};Ru{`bfyjulMjDX^cP`wD`?h%tN~)r z2QlNX*pH_jGb`zC;cxV01LTnyQ=9EArTNimbYEi>kG(%+wd68o3bVs>QzE%^@lsll z%%@lq1-{cl6-frIyAGYyLJPw>RJ8mX#e36rVT4ToSX^(KQ~{kI@|eC4X+p~Af2HKn z9u0~!{RUHnwINL+AP;@V^UF>RqldYjV4X2y&;GTh*Y#5jz22{&o@fJIaZ?cfWazI< z{!4frv0c2`sJ$+#3g}rSPQ16(h1F@ks-=)~!Mh$;o;r;>>V%O!jAGQTKQ_UKjtkXk zbbGy?ILy4O3g|86iB9V@x)dCgV`%7U;U2hff_nbRx+GA+v@C?c-AL+5K-(82IdmBb zZ)JKNa<#K9mzQ(Qc}}8J+XY{MeO^Hxy}aLCaMKlt>)ncA<&&6O`@or&grK8x4Kz4 z8H&=4eyh)+Ri<0T|Gyd67EB+dfrG(0^h~N^t6%Y6@~}kc$MCzCF+dLwPt;%xW3?3O zGCd%5dfDxFlI|i84Yb!Kf%O(_X8o3XT9>;#C>P&Pdz*B*i&2i=SDtJqPJRUYi8_`3 z(tsNk+N1w-j@=T0uu;HnZXtcq(8<(f<7A*E3n7H3;g83zXs59szrZeU9jkG5ejQFc z1x!0YEqcy{kqDNTg1jC2f;{}BkI7>i>|w(J-^Fk6oI-31@aPL8x&aQhv-@BMONs*tlGL_YcD`ES1`qZzrB$d^zYb{Zp z{G@YIS-W-CIm(ltbg5)vkCy)4(Ak$M4?i)BEnwon$Hj=mK*eFE?|B%Y0OWETenWP5 znv195DhKqjQgHWnU6lT);xasXuC52XtyOvSTwR#yX6aZu-K;ykS$Xo4ZcJrisp~?o z>Oxm3Pkz!pQV6|)YEg(D&_(W3p8TZsQiQ7Dm5hQd6as6z9G4HO0JsG2`w%k&TIM$4 zTMm5y=$Yn1_>hJh!JXyT)4Gdx<;hRlBdbla&eo){hdlX7AEvTa>Z}jRlb`fXD(e!R z^$vORlinmOGP)FkrGqcwsDh?DYr-#s+3!z;#jlN9^8cc+XR;?kLlC?06SlS zPe?--t2kEZzijASH1xjT&|=V+-ez5~hv`M~3Ue7h!3Vm?qwDL#Oxv**=$iU8G}{IAGIjGbx>gp) z?YYu1y$J0x+Uh}DbX|QE{}m@|JD0AP71JNO9?{#!b+UbIP}nl=5*g?L6?f>_`fx6_ zYuKKa{%AFM^bBBIE3iG`!{#H1ZWq{|lCrmJq)@wt?b-SsX^ft(kD^hr_i%ChUTtpQ zhb}Z?*?G`FLRtk2D^?L|T?n(BKq6v?p3*FAjm$*9gHxV~(cfNYTD}G?<2cSP$Wj=5 zGrLaSGUw33++2}k!i;?fXZ&W4@m5@U#&~yd!(%m=#@hYHiq)xf;;4H(mdV@z=a<6*Hh`Ug{{08w{WU}Rt$>U zxmu{5oG{(qhi-wLX51o@6>7lLm4!|P|tTM}GYt~%SmvhsWsd7556?&r*ll3o6)AqqQ%%DaTqRC1++`W1N}{L(iHsyMVr~>7ubzLh2T5$ zH*;PIlf0FJt%{CliHwb~NH6Shn4#Ap5Hl~2a3N2nGcfMy5mGJGfPkf8FLxR{Fwl}C zqR@;(^TSX?_Y`2kW_oA}C)B~%p!UOALTF!@$a?yJmLxMA$wH2SX^bF-gtP zOZij1eW|Io7S)=4wK8;ALXu7-LJb(tCVQBsj+ITO%A*<3x^Cg6{wWpKBNLe(m0Kyu zy#-A7sa`da9>t!mCjebU^4PUcqfj>`qGQAM$S=aO8lqZv>#PwG(Owp!H3}EkSen3s zGgT~)=)L(p`W$m!Mi;xzZH!Z@+y!bizgVBpH#;1PhhgT2s86n_Io+}BhUuU<(~EUc zEy~G30TbQGgMGXfJmyC6!pL50$W%E#dV=5%t5 z8at^zoV#@CQYuTTiS+lnRtQcGqW1*4lnLjXiQ0m>g}MUrJ;)maj(`rb^`x#(mAz0d zO|ZaRhF*LrpQcyYA*fi$ZkSux-6j|i z;E&OZ!_K8dPq41mUtnOo7}q9~q1u5f1e=4n- zFXyaRGO)8%;Ua#UsnXdh1N=wC$Qve>4+q%-TH<)N zZv;_!slnl9;Lta=ayD~}5v!(v>7J!dOCx%xG0i{*WuU#W znsZWi$1o7Y2Q? z$S?bdF1uEjeIg@{j!~8_%7}AQUa&SJXv2CJ8@K9 zK9w$UJbGtdct$R*aP&=at4w4QjGO2a-BXL>(I=)==dk8{fg%0)Nas4;{vt<*%cOhz z2lSEdu*Hd^h2>M}JjbJt&>^Eoj6-!$T^Jat3v-i0zY7ODdHaoKGXFMsSSHxm_23Ty zC*6`^IO*h@Owl!ihpB9tUCnM@c6x)Dq$^XUUKZ-uEMr{(s*WH&XJZtB7B|}-*#BP| zqS!4<*}r+hv=l?6Wfvr{)t_B^4Ba+Zh~n9~Ad1`cJ?qxFVbH)mjZlkzDDlq`SBh57 z75+0=^rYvZ2Ghrbu}m?y4m8}5KOI`oqat+wTwij3U@&Yd9pEU&uyBb{Lnm$vNYu`_ zeBT81lPZ9;+S&kO!qqs8nCbbszHo7av|=B#JCJ;~|Ha_#9K<(LS+o^17NGb#$Dnx8 zEvD$XDB8@fk2wef_yU4l6`>pE_Eo?>;dl&nw9DJHDWn)d3NA^9xk63*7&=R!Myx@m zJLiGN^e00<<#;qd8H~WX&46_auK5|}>rltK?5Pv!jfHxZ(}EBlN^K`a-at8 zWtYB~+gCS3dmM} zwbmf}ZqNX14ARMETZ&v(rHXDsQHbf&!5Tib;sJO45`oiPZO8$XeJW-XIQp!}gk1J% zEf&;@OyROUwQ{HM>nK$tbwef=ClCk87K~Cfil;f$S(~J4?0%3Ba#aC6nkgE&ze;Ou za?Eq-680e5pT-V7SPRjhfYGKpjqU*yAZGg{2$fh;^jKzGokrJ7PN6i~g+`GA;bt_h zBMO*q&O{KtQY0mW=IM$_OQ=fq^24?5?SWJQe%eGE5~1NZ&Zc zU=tIZODEiAz=T~(An9JKTetJ*C7JMRWr|+X6Mi#3n6_o6ZL18qiQcGfWtT3?Y@s)5 z!*p4um;PU-%LQH7hs2DlfL=v6bX8^xe$iE#UZ4M7;}^GcXuc7lN31gduLZJH7TuhV z*Ul7P*QbiUA~rj190TfZsqwe}O<5q1Y$T`x`NEz8_^|VU?PFBn9k4L=-%MZ%K^j3R zOgc*8zpy{T1qC;$Ec@5=Yz^ibGiZ;YyRu{(+E9aT?#hx?^CUi))&c|;hKJ6uGxb($ z{zS=aIrsR`+GFS;-OfFzN)PFF9>NFHy?xt(O$eY`F$0?$y@!>_bWygaTS0k$%ogl0y^Y%Vgsguxe!)zWpP*Z*aGoNDqJ?+sCDU;aTdz@d7Zpnk2>o zIt6a%T8)jDfen>Kt7>p&6ro{j7PfLx{6jNB^xDNIy-1IG3ZroV2iCT`*_pc_y%*oL!=62|m?!Xu{e2I;YvouMj=S~P{)W9XAC z><%rM1o|W^j;-S&d@y~G71v_2n0(s5_%eWC<)fz2NBt3Bls@9OQX|g!hbj7$^xwvi z%C@L%x*z)rrfz*Ut^+}(hNiHFF(&*4bSCw{8~s0l3I(7Zy&WevjzK>&A4*dfc6e}4 zXo6)2b^B3!rMyEiBk>eJZs2gx}67bfTE~#T3jtZ?@y=6sFA$VU-Ng&321ACG~+UdjxH1fHo8cn5-%87&!&(p$cpQvf*;P|KHI5pi&S9L3oqa0) z(fS?W49G((@lG|4j*$jS`)`n`tVfNbZ%G+kE$*jNVWy&~0>w6~DfC-_2C4)Pg4Kr_ zM^;nu@kSZfFmlywbz(r9F`_w9w zN(*|3Tf}WqjiW;_DGQQ#M_rAh`Qd<*Zr^NZd6y+i!xeK|7Ny^hAQd`?LE1SMz0v7O z7H3OL29zXDK9@kb9bSRs8fD78HjDvKeFYS@dT2ymAg7ZW#Hr&Otb4lSeZ@OX~0s8^aBfG z8we}GTzIR2vLZ|uCU9gdV)_6sb4=GIqF4)`hzBsvLq$wW8>6%xW*BOmeGFaM*wd_v z=vAwQu1U0d^g<}kbVCA*b%*7l{P~!+r%(Z^1N@<NF-b3=Wi`&}3rP}?cmX6IX-LjHKuFX$`rPl~UjUL;YA-lCQy39cf%vaz z>{Lb6+!&@#01PzX4`b|$3N?jYHCadQ-Sff`u(bLK!z$}2hUQ# zUuPVRfw+Z~B8V7G)GC|?0iMS(tq8TKBDj^}sXBJ(+)yvOs>qQf@tVf9LKD21^`8d8 zF1A8AAC7p{(Apy)VO2z%5Z@jH%tiaRpfdpc0tji}F*|F(5Ujjh7{sVwqDS&Ju+%#*=SIY6i~Zk;>cHXUDV!#By%gkrtpr;n%a z4$Ew7kLW4SYT+g|o~gDI9JiF+6=LoxG=a;vi0F}|NSzeHebW4BzBsVM^osLwZS^O) zVSrW%jlDZBtQ7sp@$m81yf8wZn>`+vSvYxNq?iwlmgt$>`vo z;?Qx}6^o&Ux=?vs0JFdt#_)PwC$^ex@T}{kL2-4WeG)BG-D*6&Cmx#@skj6bKZ3B-C^6#Iu=;3vazWm$es=0~V(Lp&_0+IS=jRdZO1AWfjJ@Jq7>ufFE|!kOzOl zbY&yJ*08;5b#s&b^DijJWBNISWSF>O|B_YkzSV^TV_}QDJLGaCQdr2YecTr`%43S+ zP6?ZJgsDfi+3`$ELUIr&(bV^Oklc)95y8lgQmo*G9#)|SHH5$xdIobl|?$p`M1K8k0e<)q%&-B;yD)wAV1PT_H$T zTSUW&HbL4T<`~EFq2C0~iaB5C8R^Abh;j~nw-@jBI5=*=fF9bJ7&{OaL07U6aejw_ zpXrT39yr{1rpp#$a1K*m3+r__2}B1jC0M*P5uou+NLAWTNISdXEDGN-aoqQ4R15tw zSx7(hn2OMZgQF=KRYlm!a{R-b%Bb|KO1l* zGM?#5TdqXFPV`Y4LM?h|#y{zQ7&89AbaeS3>?W zA~0)8(dRf!VP%e@W>E)XG{hYso4bH0w57(=%?+K>BH-3aJrtA;+lgVbo5=80NH@j+ z09XX$?O)K3delT3=c)1bh}3U(=-$R8K(>$Xr5~GTS5^)_Mu&zOPs@FGDQp!k;c*`r zkO?Hta+;qs;!&ol@);bTmH~H%v8!sCE`@?gjd$o>+f%i4aYGb)38HL>-WrsQ$zIM) z_DGn!WP7aD?VwFrWiZ>7FfzVy;>cF*^F}&Le;wPUR~9u!_3jO(_?CWJMz3eOs1E!K z^DDO#&Fq%S7iD)rVO}fmr~~A%F#-B{vQG=E(SjPm^vEFCanU7qiyE(u88D$5 zPnf%hLI@Hup6Sj6C|gG0P?dnHH)HYd^waCW#SrC6jx2i0neIt+j%AneERNpx(eS+? z5j!42kQt^>ms~43bf>IJu;%f!4yOOmB;uPRl;vqm<6%J2H=Rh=hq?gVn=-^3`V4Wi z9wJ>m$YZ+P?p5RIk@;aR;Vs<8d*lwJi6K%-5U{hwnjN>82of`%X}1kt2W;5?jU8BS zmq4{bulNBgJ}v7wFy&x}KCkbTr4&;q5+oDB02hPkVro2fCN%AXO3-jIymy+}g;iKx zIKKpz?X_CrMe%q9_uuM7x*0QovViJY%mUDNL6Dki{=P&fcvAoq%>FPH>r)tIx)=4~ zd$=%3)1Wk+Z;r#;hZ)lEl|To}v{O$hmOwLBBo?Ax|4_(@REc3Ass+^{~0>8*@c0Ej?$zr9ff@-9GLyqJbzQqyY3q!~=RG>1?&*t{pO z{R_IK0VnOG8+8%}sEPmhU7Op3csuwz{wyPtU-ChiNrFm^u~sBPCjs zi*a_ib@5^YA3|Wq(LTy1&@ylxZYvNA6S&s~81CgZt*b(Uor6!=2<_b6 zFdh>QnvX%?nw1)(`4sXTO0>eS2GnYw0%bZ7GlN|O)5_PjjMF3bm@kr|S8RjCWXE|A z)&OKC=_~MvlV{X_rcWvD-|_(-PjAB&3ZmbMBzLYtswtwJ)2U8$$Y}xbirNat<2H6n zLFXdcErgK(;ML!MgpnX_X~ia_Dr3>zqP@~2a8~c&nnUQ>{fQ!M44MP(g65n zIuSzMT*$HLSf%LR4D6^6&X0x#ftabLz}Nw*Li#kBQUND6(lmE5m*Q zivg!#mt&OGK{d$C7ZD6?LKpynX=E_v!FwL6*4zvUo3N=3DOFg#$$esXZZ2ufg7ONFsphQ?q;mQxI z8WzbVpw9YMDT6_XICuBL9$#F3co4*5nhT%XhSX3QA1-cRv6Vwoc@;I3$QWD!= z=+4D#!4p*_dcb`!uDBbL7?0nf_=q)Qp}K6B??kyRFkU zCN^!f2w4Z5>VeD=_cX8$1)}^>J zWLjDWnbL_YMJbT)=i;NiWbfoJK-Cic!oM}k8Dk)O9v0fXT+}dF-DBC6vjlLDk;BS= z^8%l7{Dr|VAkPK)&QO%f!#~a;`P0w@mh6q#m(+OL##pj9E&$`FZ3qcB2IK=e8e3|CkGdy2=ho#>G+S5yaNnNn? z>s#qyYzmmY9ilM8u|5MbEE?J{zmwg4LJ(kTI>1Qs#9OC zJTc~OW|!l6TtJP{LNIk~+>l|kj#1_f++BYH=*=mh|6+hn2scm>dNF|bdO+-}W^N12 z#&SH<`A4uMN9Wj43c$wW09874HWhFIm=yBmA1xE=ZcJ+{Pcf}65nJ=yc?k}izU3|3 z4}>rWLS)hk4hfK#0tId5c+c4d^k|(ZxEu%{#aJwm%0+fk2Q$4_7ga;(Eeo5#!Cw_fbG&vYh+&pu zVzwv;Jn$kH&AaT{5SI;~mVK;cez1>yq9;XUDF=!J9RGd~c7th9Couh4$O{eT;j|7p zPOszqh#F5xITFg!#nb>-&0yCK&BqO=eH!pBn*t5-RSgply!A`=ZCzrgI+O(=-DzeW z{Bfy7fP^rq3>CKJcpf>{E)6|e8Zb>UM6(Kb z+i0ON#a%SX5Ly#VaYZwREg7<{>y+dznDK-XvhT zm0x5yzG%%T{sOW2QT zZdflIpgqE_Jz-Czi{W%F^q-arSs@YTFprgb^TPzz#q?9drPddjkv{^PVN`}j8zRm& zcA*`1eQzWH0UCUPVXM<=`|bfFCYu(v0(WQutOiU5;cvl9B1-~sq&-#o4U~pF2bSR! zOf8%Wb|?b3R9>b=(b)?j7rX3k!IIo9J`Y>iMW7?-;ix`2m78)6D8DkG>_X$bjGcUu z+OU#T4%@$`ix>2;yG^$9)eX=xLh7`T-U4e~T4z*3# zp_u;3acT$$aVm3u?ZYsMzBeqhtyv3^!qq9!gVPgYk?B1H#}UX_jbR)&fL|PT^ z4csKn*J*HuyWwHJrz*2XBlAF{l~=sVGh#9v^vh?D}Rr-Ru*e~#EblW4t#s1dFZ z1%w;?L2c~HX{KF8Xfou$NlZDo8<@sn6`|kwN#BP|qT}z(V3(+cj`o;Ju#rPz1zp9M zy08b}vZFKjd3|G&UY1ngV>|4*jyl+x?oB$vw_9o=Lu-=CowG#iS_$u zi7>ue#Dfb#B3-Bs@IU;b#^h3V?Hrno)5|o*nG9Y|vKczl3M6b#u$@;j&6ybD!oC***ZfuI<&RC5$kOd(@sm~NjIh<1j~sg zWPG$&#>GL6_mBZ)!3zZ8AM3zZ$W3CpuO1T&zJta-47$HSy5Blhh}RVfQFanABn|1i zLD5T46wR*-nn%0$)I->!X+~nrDKXRVj@*dQpbn6(Ck8Tc9Fk$B6k`~U`mrtT6L9p! z#L#e*8yBY^wSvq9*sB8wUIhd+9hyf?`(Rly#Y7*W#m(Wr!t|Wm7 z$SATl@sEQrq0qlMTTl_6ef=(0>S+*`!L%427Fd=&4M~jpGw=pzYVJ<>t7-#a!=W6d zu9nDV`&+#qg*+}sJfn!S1fnItzCF~UCSh6-b?GI`qo0NmX^UO^1j^_}G^Ou&G_-~3 z4%xnBU8xBDF+j~b(mfOjm|_T-^n*AJjpGPBH4G+FES7!uB8XXqthuK#20`$-K`Wm~ zD>SE5-znj^a3a&+Y$(PjF})(UKM^%iqQkgql05V0b^6}y8@(8XgoW&M{gaPxL31lA5RN<7&mW}(FND|Zf)=+1dz zkfMD8b_cw&khTIGxCLN1E`mSdO4B}$GP@XYXh?Bgj0sn9;27WM?nkR4|Qtc7WvJq!~bF#n-dK^9#Dl|H!nr}eT{&m(Ud z7vv%3lX_XtNK&=*k;&KhR(kgYrvQDyt%0?B6XtxmI*}g6O$5gFH$7bJ);fTTicn9z zKfWb|dZGw`<{PuBBD6Rd{+D|Ve|eCxQ)i?Yevt9eUKtk$8Sm|taaoY@CNi*e-!?ZQ zlbG&K2n}O2m^1vz0e8Ru;;4;P6rm4;!CfGOlc~e#e`lQrwpSJud_u+=vdI%V)W@1Qp)d5LT1uAGX)$gPS2Y^B)A++bH`93&X}N^PftwF=Xou`IgXrI0+1 zg0cvRgH(oNpor$+#zEp{zqWY$pUSjelN{4}QItk->-$AQgodu&=lPyR~xq*<)?kr z&Xlv}g^s5t(fjkFqU^#U=pDO@Y}qZ&!&Mf2YNsnFCs8kTLDN4giIWG;8&d0Cxr&2< z!5>9Y*-yC`tW*fn0QkOku?2@5MLXu{-32;swCV~8$c&=wM8*`#$zLZmt2KFKPJhGN!XnqkS1LQ9Xwcxu~X7@Bo zhPv~yXMZW*C&;&&NW1Y#$q>4hEQD`32iKK=c`fFgLLK?k7=;)Rq1S@$uafSmu~S9p z%^>kIod~|eTU3Nz3No&f3bJ}G@Nr>76!tST3tIj!Y|kiG_nk+58*BgI%ivye<5ELv zuY=UEq`_hI^_Bzf0T5zryOd?HcGV zG(o;HUlSxQ@nF^nDvTiDqF#y{1=S6m!opDXZ>g$}&hJr*9?HNKG7JE+_&<4T9*mA)O)PPlLLklp)!7Fb&$m(q-J7G)pvi5S;8t z(EZpWcY?g^;XBcuJ!Tj8Vb6V8_1sq3-;Sr*#=2p`g$8Gz0B6kw>`wDp4sIzO3UxID z(P9KEM=7icCkimYnIIuzMKp;7s>8TCr~$*nU`@m=33U+WXO`Bt}26Q1$?15$tLpTRSrsavh)?=ff#~@*V z^WRaAm4tSI%Lx7vxtHzBLi8lztSxBA5y**-1$$}@7-8|12HdRx7p;JPAT>hu2xX2a z>V(bfOTkGjmx3;DgPl?W?!%Lqo(*-V2zAQ+*d(Ug7`s-wSWJ!B5MBuiT!R7-TJKJD zrVd_Rawv!^rf!TJ1}th4y*?;TP8@2nxHf9)S%BO6NlZzG#WWbu@?v)FWA;RV1RQuS z3VFZ;49g}ly_JZn5%jDbr5M;LrP7&*D=uaP( zyFEvat4Mvoj$;?jVTT@TNK(DIG#YW8Y2$FTp#w`7sv=5?FavYq9NO6s&POxnA`NPb zOT^^2NZWoNsK-SP!{cH{Ll0DQAl(=PIC~^5lcRD><!Jn8!3QCF#IClIW3o zbh`lahz91H2ADsjfcYJS!~+55`9ze*VoZPDFDRWFl;5U@&Uo>2d4rlMYtcVyqLm@)-LonMcXteQ_?}%9AM=-0cNuXCd&W=FE)RQe(nQ< zt>9k(gJn%e$prOSQ2g`<%Ik@s%*~DI;JljXk-F4td6@0(fQNn-{%SJ)k9Pa2XaKN6 zPCdXuzn_TGu&)^OJElzELBE;kk-EX4Uzb5oJdi>EC%b*q4ElZ?V-8>e;E_hSRuu8S znCcG`L7BdzelO7@b%RmABcr~r@j!GUh(Y6Yz;GYP!Q&R#Q3-KpIHHQEBbbV>WC1;ZiN6A-&oxYI`@r;B zqDLzFFny+BnrmP>)GN@3e(Y&C1^|5(8@Qk`S^&!UpnoR7^yfr0ZLe6-*n{=6PI2e9p#Z?Yn(AXNZBkrt7k?rmG25&#~03k2OjUnXYegI!~z<}S> z2k`y4!Kxg(UIX6T2XO78`KkcEDh=>g!DVTc2K>Ja;HPN7DFfQPeX)%jwF6-fxG9Np zfF-7<_lxb@HPBxqQhUKMxVcb8xS~}#)RqQxrvTlifxbt=REae4gtaf&gF6bS#wQZo zq8%96r$%wzOEp3x@_0qRH1Z=Mk$3t4y-rBv?}F+38e7TH&CuTk==(I#x21r75kMnu z$>9ERy-oCp2f~wnlaA`q`=je$3Cvsj!2Ao~ISe%aGDXv2NNm3pioP-h?t1_i&+DX- z#-NS@n(yQI@dE}o9>ubt`GVsS`vbQ`zGI|6GJ#ctbj9VVD+%HNdwXD8SR9ze4<~B5k149{~%n zt$l!92w(>R*oIbU>=y#qLBMT;0Q*70Fp29zXdeQi#x8a%96ZyPn94!rBCXK}2z$%( zqAdJ|hv*bE|6mn-=&JCq4^G7T@~f2oUP(e|GMBvSOQ@D zhQ~U7C$^VMCGS9%6Cz(|fezt20Q=rBQI4Iefr?? z7VfmMBVfm0X=4Zb7^h7+TwY}NLTBU*McvCCHX^9Ia+ul#q)r3U8XKw~ht3FN zmK@sA5KcuQpv&QU{CWtA${0K={@?ghIdp#_8W?94KW2`#&%uKQ7@30c{m|oLRYbcR z!WhQcHgM(8o`&dGogbVvDu=%3;Xx0Y8^(Wx9DjxZ_p{I)`fdk0{j$M`Fb;jEOBK;n z%=^iZFDTPfMf8g(&g>)UJJ9%u@oy13beG&VHK`oJ3)3+)b(h#H^hs(nNzaM+7ZITF za2WJ^DEb`eiSBR2RWxvPVmv_6_-Kle$xFN+uu3^{<|Cjktufg$G^ z14Ea<(3}rs!aNeF|EvKkVuetPT%M<>MdeVgKnJJX{xGC+=!a5ZiVuEYF5%F^R+U4p z6#1;d@Gpl(_o^cL4LUxt1HZ!T&<%+ml>-ftDgrv-8<;~U#{+H76om4d!-#@>bL8ebTBP-0A6qJJKKA?0KhVwuf#vyw5WiUsTzO8ZW@p)u3{VjS;Vv_AXlcdLtY6&%*dRtOJLi5IOJiWF*dP2 zvo|kXg4#IS>}UGU(j{Qig%W*vl0)w#z&_6od7?_VE72Ng3>=!>3BB=9$fgxoJ40I_ z&uA9zGkpDJ}}pS(u_h0T}b6 z!6aF)8!(y!F!Us~;5j&1L_4h%NfoEc!)3P#B&dq$P^c9&*fd)XQb+nHldn%yR05a} z4`7~YV4nP@f%%IN?m5gd_@jaO154Pk?>N_j{YDC$=nr^r!jWxvx)j-*TKEo4!RUU71J0CIRRn=C*mj|fFQRV;1OJ_& znT>CofxjigFay8U41AmA;YP?Gc+NYf#@kZEA2?nLhc{vojJm!zM`TcR{r2 zLQ74Uz8r_H6AZj47`RR_@M0PUz63<{jKRPyDGY4<$Y4Ofk!>*Wy8r{EpUZ30A$dtd^5_2mB+m*YHyB8^ zrXZ>O(m?W20Ey--hvUu#_@14@;Ko$xkAu>x2-~Uzox!08n=59%?LPqbV*>ZB2JUB4 zaIfEE;C?-TM8o~v0Pa{A2O5808_{dZzY&yIMRa;N6_GPn<`m6lGzYHE^}$Acl=9q6b8=W{;x#5 zsT_JM0ABCAYeOB_m7f4eOm_;{W)ydY;062d?aJ$A3hp#eJ(8M&p=6-?5OO7qWL5_& zb8;$kCo*BBh8W`dau3#GE;ozaEH#D^vEkr2Ys2dpyKti0#xbfkVb^k$>FaWslyjS! zLcR4`^Co6cQ)o{;jt6HG0{rf&Zv|eEh4!?kE^N^dS9mu}v=LPj25VXmx`^UxikS4f zeWE!dS~XZ2AVPz0Ynwlgp`C8A4T47(=_AWnM55SZ=^D;rv`u?LagNX*8^US|Z4_RO zJ@#T75%1B+2cn+EI=j&JxDYXz;9naN-(Q@#w3T{2RN)adg+9Qb=|YmA$nV#O`+#er zYv(Za*wCXE`{wEL$PflcEeHtJ2bjMIFtGXpK+i!FD*z!!LCCuj$+Bg1qG;8a1#fg9%axa7=5Y(dmISDNcffzUx)7_BJy--hww&%z98&zJSi z56QK^ORqupJwJrTe7omD4doTmETDW{L-|X;d-S=%WZo^Ag@6xH8JOOw1Ja%`RDM5s zcsg=OoT(zrkWQF3Se_QH>3wv+!9w?F_fykuORC+6q%C?v8cr>rm&oH>E~HAFYxK#K zeiUYxB5|hgdUUlMcxd@vlemh27YfIjei5cZU>EoPQ`n)_I%q>TDa2vFrB0eXH`GEO ziz-6a^)l!kfS??Q{wnO^Q7lEqD1n06P~T5sx?8vlkhej1#~#xOQsgyJiV&hz2L);heWvkwL<{Zn3P29i2Y3!jp=MHldJ(|85)C02n1kWr6o)PV zSC_`~*riii7{)^={6`d~%N5>q3mE0ob+{c_gBfLds&4;`64hL`BfHt9BU<2vi6%dk z>QA9bS(-5Nr1f>tKB##IY?!XpCNKr8847HMQH5@kel`7LO7_UY(oM;9fObf~kFWuNvqMk&GjfgK25%#$W+W+V zZGSVe-Jg+b^o(rxXXI)$l({kWX2*KYe@E9_f za}Kr1Sg`@v7Xusg-4>?*4I`+qnnHKhLqqY8K6G^Xzl8c9KGb&tYSc_`MNP3s(4F;I z<7gAWVx7_DjWXHm04&oDK$+N2V+8l>?)Kho^n|v{%s;D(r1qYF*j{n|SP7~vd^-`P zL2#z6LX5dqbzO+TKe5JW6UGJ>sLhmRtJinev^dnl1W)-XOq=S%2tZ=z^r{?s#rA^e zF{_R?64BLzhQadLl5p(nQ|Z?Z8kX56{-f&}L>$_Holz3*(wCMSY!459+|>Zd$$ZEy z;uw0=?q*&3uaZQM>e7Fe(&i%yhWMo2i-!KHoBf9kEa731JCV&m3$pK&nsDt)hIE!g zPwAR>>Y9(4ns?}$kLf<|kecR0AUuH{HxPD652_@y_Q(zi$wo)N^9+#;nn>8g_D&^8Sf zB1GmeZ8lKdtZU-kL_8TIHO+_gQ$SlyKQ~DdZPiWRB&E%VlpaajOz9gXiMHv|H%e*q zA-H;3Lv^FZ)e8oy8<5TPg06Xk)HEN`PXWDX`ngV$=tW)nIw@^Fr1VI7$&|iUlISH} z`dTS%KBRO3y=+Rak|YSDz|Sfv{j#nqS@sDitn;u@nLKa@+HOX4jc#YVu6m8MV?LzU z;q;2>^=e51pFqo3OX*j1RmpPb4UM_0bw3Z=EzQ{cjzV0XVs@QKB$C@+Ts#&0nsIFw z{E1%#Y#}50?eJ*QxsnjcYcJNJfP%uN)dvS4btCSB~!GCFH>(b!H zb;oxN%X<;5jn{}`K3W(~DGm>?yZ;|^ZvtOcb?uLzd)CRyNeCDbU+ZhN z_^_?5wIV0UIZ2Mc;%nb~?R%|1c&~la1#SX2Xj(`@NFWIr&NK|B%E0!f!vkbLXi^ymC(-1(m3LrNLAX^n6%hQ0Un4pUO9zJY9URNDG)Z{R2!MYii zvAwEGspSBBWG{YN6N8~#D^R_xpjwiK>KwTxr=Ys~u%Y?@{n15-M9;g(XZqt1^t>hL zxl{moO93*_q?hNi3_#W%HXysukxKZb zUjRu6Ao~>{Z>9mcCIgTwLqPs23tujP>^USLN#rwKF$9n$7B=GnoONl51^-fdISt4S z8Gu}TI1H8Pa`ZxP9TJL5kk7PY2ox6y6z?q(nQb}1pjC&{o$8IhbazbW$weYp9Yzy! zFJ(|1mvHoN=M1mj*P!~iMf6yqTkj>7_%VIBXjn%kYhpOq_NLCX{o*nUPK56*!cp-e zG@!;!7h%|JQ?EilY{P&{L3X)7_O^!XO1al`hY>NobSMh4=`u_F zc$YP0`0-v0?9n@*AiBzeHzmEgNW}W%>0&GJ;}uK)`X*@i;D~!4s2k8P&0J#$FFCBh z3m}=UktypN9=gD1;bxR!T9T%5o30Q5Uk89}ew(fkiXUtY>}IqJ8jSne9FUf>0fLzf zmAEB3fus)=wArTLwFVSsPQlGY95L^~=mKoo^e7^0$*3R1tk9-KiA?ISfHmzay5T2! zhM$w**~&!Ti$Mc7;*;5leID3*0UFWape+>wM$YHpO+QDcyjE8WI-XZRSrh3dOOxBW zf7ZY}nE}iT8NkpD3gLa>LOd96BY@;Gf#f9($?7yD-9Ock+?4^z2bqvupdk6=ZvaWV zK(a$aa&8)u)&Hs?=?WoH)ZHiZ3>5w@EA5JA{F^S{9hSGwpk0}euTYTp{|z7?l-+HQ zhP*uu`Hr7!NN&l1Ij`$4+eV#B)C%`U)n4y z6pk9C7lScXLvncrBzOJo(f5u3ajphpM;d+W{#^rceg+Wh4hF)SNbf1Eh{hR{b}&}J zs9r}IWh+CGq4OGglOitr8B7mmveZ5ZPD(EimfA=CY7!B>UFpUb3$MFS58^G-Jg)nf zdJq?84C4IEIsFijFi^V`tjK_5{lVz`>PzAU0pdyx#H;B! zE&Fc`#9-DC51>_G@4``bTABWm>(VtA3)mg%euk)#0*CpFgu%ox%kP1`&_&pj@f)jn z$TfhP|Iq-PBh9Hj=XV+N!L%J#EYQI%F!g|14koA(EVJhX8#icdY)wx|SyTh_=}>H3 zqp|UZmFXw9E?r~y&>ENN8t+(Xjy!&v?qQEEFGIT>rXZ$`iuSrOerqB2Z&rp`m(5LnTtvjd~Q@Ge)sEUH;v$ymbcM1E533_?Uux-QQpl-7S#c zuOVNThPCWh^6D%0Jn!3_V9gZuAf*z5kN25_*+P4!AWoWf;A_)|?~x=Hv9_L*%kv5DEW z-GCgM2G5l1n*6qHQHk8^gm-{TcqAX;Spz=|3~#@His3B*7MU6XPm!dCZ=Z}C z$eGGR%R}o9&u+H_@2c1KGQz7 zyswXO0o07?OqaFk2{=4_(`Sk~OeZ2BC%cGAOegwK<(2{3cN<`jAJ!qjmWXUqncn-N zGW|eh5qs1j6qxooxfb++ST^+CxU$^Ss)LGLgt{)xNHC4}ab?$K+`@NfFg!Hg|Bcy;%UC+7j#&0^DiWkP6@bO{e}I6k zfmPr^isGy%(V!STx(tT`muaq~%YAq(Kx&ZhGWoABDSwn$dXlhRoJDZPpGt2uB&><% z&*&}gmpE+!1zanGRM-QpEfds$rUE|?I;)fEC|T53H#*dar}Wu|E&5va%^xV6bIlyh z#9qwKutl#-r@xVnxc>+uC0C6_O_^3oZ~^Y;4|hhy7W~I7XsX_5(9V?RvGn^a=zpFB z8tM00qW@XjC_2KS3W%q(apCz7m=D%%3Bn`2viRZ@1Nypt)g*l9|2sC3tS?s9~!P#6lbXvY_1oi98n$8EgP75-k05c1v{5)QjQJZHF4Naz4g0S!Xb9$Q<*tGUo1Z z%+H2XaiN+DYa;Ey1k!a)a(p{j{f0a7S^6M4exY6>=iuq0Q0N6%fF4JvmjEf#%}`qn zH=2=1yfQ&ID9Egdv^7f~Eb|*RBzwZeYMlZ8sjjc-4aqZUNQ^}q zlJ`SMlni>aM)7uabojbrn`P-59}KPWJ6)rX)3W!vbomc;d6B*U08(dqLJd^RGuA{} zuRiZnR-Fn=QrpLXmMUOjWBkW(&no8$3fECdv5@L2LrLgE9YZ+(|G6kM3sX)9sILIB7V&zU4`XHaJ z8XHz9VR|DQvfVx@0}@JR!vR~~+pTlhEy6GH*g|+x!1l2&dR{i1@~|eSa=^~f!e!Y9 ziCNTx>zok1Bpmb%{DFg4T2gp(AtZ&%ZAc1lNYt$_k`&I)hNN&nVR8vD$t35}E|J)5 z`cRrfFzl)H!N{81&%Szpld9#&AVLGl%WrT!bQyP(U8Y(QB#wtW*HqfGFu~aQd$Y9{ z+RAhE5ofy0#$ZemCX zyBB6k2-hQEL((FIx$KimTe)iASr`bdeg)*afOC@vAn-Uqf=Dt=z({r zP2yY;*;*8|uL0o0m^cVZwdhhs zeAYzjV=Yu4$TWh&_R5zJlpiWKz(D=vj~UJ z9$m6Wfhs}{X zF-10_R-_zoVo zw5HO<3t_{hH*F5eU7C$u@vR07=+bQ26`x%=%&z!o4tB){fH7fo>vM3w;|yDpG8o;} z*?L#(zEnf~l%hh3OFg(CW81=-xB;u5mBMMb)d}W<*ALE+(Mplwna@4g78z8U_ENYC zW1`nEKAbqO$_|*%V+$eYfNI`BExNTqc0Ow&U7M|$&+5zd2sQxk*w@2z{9+-DzZT35 zhcvoT4R9MINDS~`=pn3Amn!txf(*|2I|2Ts5ct(HSHPbRQUkt%CJ)~Bt-oEbzsttC zR7J1egZV$aK?md6@P&i^qZ;(C%t3d`pf`nszC{MTd%@uy#{X8mcFLf!k$iI4LuR@I z!=XPPYRKv^{v_y)>Ep~X-!5a`7>@b&kV=2As5I2qJS;-L^0@=sIlQT0x>w+O^o!xz zq~Llx6Rx`iuE#^T?h4`R917Rd8F0Na0(?CtaNYC8aJ{VHdN~uWhXk$%Lbx6Z;ks7A zg?rJcMV)F@AFR9xzh#x^{X73gz1nvL{3_|wB3+&%0^%FkG&L8q zCekAy1*Qiv*5L+aO{Du(k%uAjqR6mpbiFF{=7I#gSRpjn^guQ|)z&wl|GTq$tTSkR zLpt=>*h%^p*G!8Zl!NHDnjaQ|{P#o+K9IQl-OdlPLgS!>% zPkaFm?o>sd{z4qwrwZ-J#6geXV1wYGOK)@{MTe{+9>5AfCHkIiGop~cNsUpF_ zJ>Y3kuyS>;SN&IA{dH38!la{bOfn7DLsnK2F~-n|DG*;jb@3CTI~uFTQD-VG!9>wO zjvusv>$h*$17$S4wT%WOUK(o3_1l3QU#{OK&=7LLwu0-oOBNtL8XCv_$vK^jfM`(T zsj{Dlh&Ywb$0vO_1cpz)5QYx}7|vHPd>Fv65)Cmv;TD4#x9NO=0gVHO#!e{N0K>%# z9A%bzDo2aai+eOFZ^|8)D={eWv|mRNyF5$Y_rlAxa+9e*a0*2Ph!KC`PZ|CW{?Z}` zvkN_88FYpoyFhfO`U5P7$;b+8A<>SO%_v z5xo4nG}IphYGFd69F=n$>f0TPLPTvxqf;G4T;@@etP|J|aJ~chOWB3_j}|5<%O{2& z28tJ#<>9flyo5{SO>+!a9yK(7g?nH#*HH4o|relHDRuXEdlzC667 zbQ)2=SEz^>hamGb%Ic@Ni6x;wb@w{m!ei1xmvnCyli82>bb*vJNfQLA^4) zf6n#U)33U@*w^I0!c=52<9j)~qEcT5Hea7GgVi|kF+G4cfr({BV2~U%4-pH^mw-)5 zRWyZvw7CvD%0>yS0ixQJ3vNt<%S2 zVeTf`2?|0*=ej3#b8DqJS)c-&dPf>B<}cMQ7*FZa59AGD&e;ABTV-^9E|_zwE|WQ* zmkZ{+vmPp=^KwOHw5@I!=DZ>g%z4jz$VTuHJ2x+2&Pd8&&KKos<~;Ek4fR9e7_BpC zg(8pM&>#-2(G}o@Oz&c%!G%B1sF^N88G0`)a`1scG3uK~$%$lndW<%$#IWci3=vwR zmAU=u=(D8`k{VV?Cu-4O>O)q1NqP|DpVfnSBx4Y-0t`$SpuZZBF}+m}G3~2qq%V-^ zcwIwsS+3p`@}JX?JQYF`SZ$udw0xc`-H`{Y&66rKV>x0q0(KA?mgEh2SWHknPo*d3 z%RP;|@_gDke<&}xYm{&_bQ+TQGVlx9I$v3HuFBKW+3JlN*cSk7$8e@Yqc<$X)R#BJ zbeONR=xx;hipy>Crm!d4CC#Om8&)oOjd?tzBP_cFyS&t#sMh`X&AR52(Ly*qsTH8bn@<1Z=u;vc=qk zlry<@IlBNz1+gn<7j%J5LLTxg4h{yw-wU}|{-PB^NC({(+CymazjA~?v2J!Z6)i8b z^Qp~&H>E#Qg!U44H!W#j(vCv_7950mY}lxRemr*L)DkUjU(%kKwp8{OTA)uED__#c zcn(C1>CZT2%wq2V7sJ1@F-n$?!v*8{6rdEFB90Kd*?rL=L|=Ch$w&F`4~KxtYB9*)p7;JbT0Ok zTq5;oY>HsgR6xiSEW;GY>4_S9{qj=t+xrTN%8JZ+GytJGwo~lVtj=z}9drS~ z7r`xjYE5`-yEMZIBkz>Grcrn8m?7(ZY266c`!2lbLbud9q4B~zmUfQ9R(l=&FD^9VoKSpxe+6wEQ4{ zk-=#HC}kl|%6b}R!@?TdCwpRhSmfaCR;j-)7%kiFt+g3;j!!W(?7+%G{OPEEwcUOo zVeYWqu0kz3FhAUG`$wfWxi|GdZVm@xok9D-`~W?++k}e3e@@+3-){HFlniP}4x}M5 z-qw(Ghma`7_4Yi)xZ>aw;72yC8WRUgu-S673T8AXj=qmjs0o-_EZ6VoACa!bKJHn0=*qyMbC0VV~>W{$z~H zOJ3L8#D`UseZRerLJ|Mg_*(*e9_l$CJ}a zESsqV0m8V;I*InSIHjs=Pd2*oima39*#?J`ct1Lyi}cgo=Bbd1T%6^TU?@sgXCA9k zZ6FW=yG=8roB(w|ioZmE$Dlo1Cfugn_>38xqaiH9+8+-uH=a?jvb#EidI5cUz5U%;l|!IGWMkLkPy* zf#Kl%w_L?xwHlfJ4$RTcSJP{qWYe!Z@u>$&&Anlt-Jj|{4`3$L5-8^$cButhVw_`T z>ZLRp7&cU#?qlr2{U2*_4bX{UfpWGo9y^D+WV!5}-CX3A7nI#mC@;Y?wf34nRifyI zJ-X*}&2hP1SaA?#;?SNiiZTWB?(#NvH!Imh)D+evCWJIU2`@q518+t8h3poDGLbqq zTPM-0hELSvm6lU|C*{ULX76Ri16sPO3Gb}i`6Y~p#;t#5`nc7xPC}GXeC!B6cq<0x zvTJYUQtO}T5{Zn7C9=k|PGYl=-j4KhQeK%i3wO(QZJvzR(p^YgxTjafkaP+~3bQQg z$Q4rGRSyy&JT0!`3qy+un#4th8254v_!5ewi&{GsMZOsUNlIQzFUPRoUEkWP@-|D} zCe%9(rR+P!MUh=XY5 zU~YfTU`SE4vmUgO@^Bx77A62SAo-*vfU*OvP(yD!otc2jc%c(cyI~xs;zOTAuOZkn z=ohQkW1BX&1k3IwRA+l>D8{S{D@~$rDLc;@^jQRxZ&L!(0i)v&2dzo;Uk-kr=x2xs z(3r**7Ka|htx0roYm{LfP*T?z3AOb;WDAUDsE}YyqJfquTK{f5u-{EGzey`w2WXBn zi^f38*#VenWoxH(27MGsS(9k5dLHl%16WO3lT^4@3qdGel`Ahv>K1K0Y)X@u>J!!( zxBvwzw`u+W**?3pAd07?aHb8omU$Xj7M9EHl8OTXS~$agk+n=U2@kxPWoo%~TLIoD z(Rb8#07>Lij1>#F0G%(Bqstjg`-ex^VZJ#wiSb6pftCU54B9K37?8KLH5zJTSF?uY zAFxfF!PF0IF)alIWk}c;LS{{(b5zLi#o+jun4!~X-yy)CjY4#GSZ3(tA!r2LL?)?X zOh57w-W)NoTYAG(T55u%%mV7iJMw^HZzP3xwu0x~C(#mg!gK(@4F}0#*|!yb1_VC? z3O}3_9G#txqdqxTNH%{U^~c zqvp>QAuJxH8-6lu*t3|93mUQ}Q7mdP)A%0MVi^QmsI6`3)23s4B%xs5j*6!tWd4Ov%#DtuqTsM4-Y+66_6|B>tH{88CmM>Wv?(>kUXJ%eWot zYGJ%sJ8yH#%VaNiYT@~SuorA*eg7%mA>r*KTN-fvP)K+eIynIiTB7oQgzry`V&Tw zaf8Z`raY^KK5TVVGwt9kyfx(bm$fa%vCu)}$98R~~9*o5llqgW)k{2FJLB>3?{!1} zM*uIemqHlx*gle7`}SfGxXS=^3C9Y;{&#%iOI4e4tgiN!koz*W+tGYJv@9ZAo2Tso zfHX%YWAtEpKxG(EjOn=T!GL1)z3q@3*!&)LGu-`+af~=_o~5v~?pwOMkqmD9@n5Ri z;(yT9K9Rw}3&N&vK#50jQU3gu;hZ{TIKPIF7NfE8kn`rBpgS1JaB9B&rK)ZCj;^*} zhOf!EJ*B6_&dzEUDO`@Z#v_CHRkynakE*? z^Xl(wP)C9o;|E`=TG`3E+6BT-mnp&cNcxGOpw!0?D)sqAc9{OzVJh~S{soz8QPT6q z;H+{kq<>ZmQ+*f8-S;RS9_5(Tz-B5!QwMOM!to6cWdvt890ZemaH`cyp7f=v^*OrQ9kOq%fHFd;0Sr?}15CjV zj!tU(Y4_Niz88#cVQ(nf*A(iO9+H;20_`iP$zG08f}UW7pWzjGqF-fd83<{ouSnN_ zM(V3`Cp5S*(6vyb`4ep0IGnk>$-Fe8$gbPdRX2~Y>f#bz^(CR95NdlCyU|i~3e^db zkBv()bQ|J2Br~}D zT~JY>i4Jz3_ss9@bIfncKghYNGP`!6czjS}Z#ed{({&5a0|?&2CDtf92d4<@$L6v0 zS)|_@#gu4uFq-`lA7AarW)~@g5m+fi4P=i4k&4)Z3B!*qo{}(Tq6AXq#}*OPK;6Ls zd#M14-M^_f_-X!t8V5owul})y@&%038b$k>Ara{vWg}o_13CjMEuEC7p z;Qet{%K9PG`Av>x(FZE_#!Pmt@pwfkPOmH!{|k!rSUH#8Y&SW-r`u}X(QeuvQDQvTbyF{tdkkRW}ykQPRHvD?9Q}FzenFC2sVE($nIvD z@h-e&GEKBQFnr*VbxWKd%wzUJ71Y7*)*?_}p(ht;xR{exF+{yvM!}6KW}QwC)`>58rHO#EtXJ6N8R+N=G&&<(0u}N6-^}5P{82PT*hA)kHRbN*&Vr zhw5!W9JBGynn>rSUn|U?8lc{yec%G2H`v55{rhy4CBn6$dLhqZ>r|4M&g@#>r{~DG zPGwrt>{#Dt+D!34A2>tY(GVD=m)%Yd*!*NgBQ^i7`UUI24%x-}u5ovem7C-)wzpu@ z_Xfi)bMb$udI+2UDNYXzH9Nr*2|mg$yIzsli3MDna&aWYX>d*0bJ;J`ExZY8V_MRd zVA^j$@0$=xbG;%uVEA-Sne-saKHz&7B-|UtqVa2=>)B#}jGF<09{X z7i`zUD@&k2#ir=9yCz!1NosMzW9a1nRPe}flNjv@*szn~X499MpNWb8%BYu-KK{Q| zIxzJA!uX8;Rq*5{dC)SM-Zp$I0?viOhSOS;E@ZpBB~0!cK6y%1Hee0QWLa#szCm=w^g1 z@t{@A^r(e7-eN^)-691^;4WyybV1u?+YNc@N}X<QiRVA4!FZ+S=;c5!IOAIDc08oH--uq_V26qPD7h&M7t3Bdb_fIk#dY9nEj7nsrKf zUG?0`StF@dSv``{=wu}2_2_3d)>V|xIrB@^t*omq|56oaSI()ZDX%+gPG#MQP^qD= ztC(Lmr~1o59ni`t^;MM(BZ0oQzG|dXui*{*tZ1m1RZ+EIB;C%RQ#q?*M2lf=dEHs3 z%&e>%Q6&v(^?X@=hT)gj*HxccQB_e}URM$SKZao#lJmnet6}k)Js&?z!@!vEGZN%W zI;r$nP4)cBh9EaULJWvUl4Tem1pJH`MqOh~g;b5GWL*s3as8EL7_;NCx%G7w(uVp( zi@%o2S%%RhEyd#HwF@h&V&%1Siwe+%WoQ6~$(kLHfe*xjNk#?L^lYrGphh~ix~8Hk zURgUd9n^s$Y#2?|RTY(0v#W#8+10gk%j<@G#A;{HavY~POho?ys*{9)4WUr*6SPuaS3Yx2MF#33hH+*^-9i~DfT@^U zlR+01RRVTMU9pPV+A4u8Vi>gxH6i2}M&@sHr162e4HdIuRn=#eSHQyHHf zk5$*5RS^z(VR@wtXRe?vtTv}|=B!B=fE5rm$Svbr4G0M!)yg>)LLRyldMd9wE6B&s zAz76*LrA@*wmMu*wdD(ERaXhsVXPH%s~1$1&zTcbloT-pu|x5wNI!&6id4?}O=Cz$ zNNALGqJ9lc&#V_$P-Gj%obvf~l~wVI2J93u;RlrhVXk3d1HiaO8AemMx`$a@7&D6H z8OEIIGx2M*VU*XU5j(~(=FYCHDi`uT0$8pTiW!|=394Dt1s&@3WEw_Qb=Ax{)w6z6 zS#_o!!5BkUexzZ{s-9a@URx2XDzB?tAd{bK7+|`nJ=QR0&#J0Rv$`?hTFA*V zjD@x3LSA5zwGecWew1Nku$~;lSWrHvUYgA|4B$@#nFmIkW3GH>6eiJ-BSu%KHSN=> zmWgN=F*c8f{Q~|~_`N&hzdGHraYO36{TX??Ic9yA&GGcFqSqNe$T2IzHk??40BSql zSN&Vxr&r|6ww$L!QQ~?_LSUwR5EL5KsA#7!yIz#rQlM4d{o(zZ;4x`UVEPDGgy0o+ep?h~V0Ntu^ntOR>Gx3fA?F3dN68!79DKcKK!dOa zUA7H`hl9x<>n>MCFay%|(+0dpY^I*DZGI8;S`*>KZAjO<5%n+u$)eENE&mhU>ROfS zTE{az6~Xbxb}gIkZgtv2hpg4<(s!e@g5g9c*KDT$hJ(xm`fiG8eJg_O+Dt1WGO&+Y zWnkw=;JZ1EOF2f?c4~@R6X-pab!jAS<C&rE+L}PWf-|knbW$AnqhBWQ@k8I^QU=z( z#dJA5sge2ylvZq)U@?LoM3qO2pyb;ePVl-t8lD@_0}M6natFuSG10;0iphI*smG+$ zGM*v2BJreLX@cTh`l!W$&tGuka9z62KB=>uZ%5Nfskn??Rk^t{R0q`T)1BM_WwN4n z^Ls@7)D|r)g2|!T;gVn#c5`e6Ji_d?GB_M;&7TVv3C_0+4q%=D6LWY>UpJsT_$<(R z0igB}fbsiv{%^y0W!no125_=p;2Co0=0=2}l^$f17Q3-vV;%LU#GH zP`N+(Gu_cL2{)b;Hqk+^O6YT!uEDFceRx*>cZ93(_E?z{m*W(X+w)tvm}gLx1J6eq z3w|9az_*eK{V^D12fNf5MMz(2bvQN+;j>@3twfY@C-=jdF=Kk9YDa@Y>;w6ziBX(OfJ z1sN{2f)a1j?M8xjClKz1bU+UcQ_toQYpc@*)(d~6Tc9J}3U0x4yD*(@j?uQ(0qE%9 z(=t)$0kj4>J3(zM8?eI|hG9*h$EBSvrn+cJiA8iX7!GZ0bqd)-EK2MYMT9%_O;8mQ z`!qjFqf#MAQ%J-7wBiv=NBdx*zlkdH1(93%On0_9)&wwIy$kF}*WDi0&62u{4yt>* z)ZJ*L>#nqdesk${8A6irJb$~~0CBw%k6r5*cp@J-&!ra)-}(hz-eK`Cnj?Ab9Zmwti`p%guy(TNzKgy$Y?30?uCAB{>CcuWf%`fsh=n{4Y*p94=_ zP>StYCT+rs;Xx~x{s1OQRvbY;#TQY?VS(d$SQ-=9Jdr9P8I=JBrsF_y7#lc;OBN9S z5xTc>X{8xwZJ)w!lSI%lj0~RFfr@Z(t_EqPu+1df`PKyLY;}f`DCTk_=#{CnwUa$7 zmo}--u2zSOtXz6deXb2Y{~YnHLZP4KPb5NIB@xm>1h6FkdxG=kg7^pcSz`%)=KQ1?k z%WANu!_7a&E*1%wdsYq|3krkp7u*KQxyZ9lWBMP5T`Py_dp^GY)3HuuGQ~bht)NgE zjV%#EFiiQDNjbiF^P${=sAW=~&q?b<^F*3Iz*{Ju0H2H)Ip*Xe1E`7#cQRA#;uJxx ziI85p(*-;ws#A7~-QqG*5gVyZJIIb78?;p-?r$B|1bPbC!=~BO8ZbU<0zDb&LDF52 zaYbk^Lw^AWUl%s=9l|~=(g@Hm_d$a)8*)LjW!%LbFku#Zg+MXAE2QQ#4K5g9*Y3ckDmz^XwmR^$<|vn-bI5`o_+u1* zim>e#V*_^Z3k7g@6@N*B7|AaQ>w;f+&|=76nyoqWt-y8q6Y&jWIBno1 zQ$$TTfo?r`XN5C;4EUl&1LlcpdtLptpvNCku!z&9;Jz?#vP@iuqHfrYiXc8s1~>gT z(5j4w+B!oLT>bT+(HttuI5KsI$*0kg8OiI?$zwB;H>8t~LUJIe?g*0Q{H!dJCk*P* z9HTnU$gCL*`OE6R&&UMzneo9;WM27=j64xe(`L@e-wg9Cn{E@E5)Re06*CYz(ymJv znm}W=TM?QEPtyrZN2b`dB9Q)Yxa_tv!(OQ{NfN{`Fm-wY3!~EliHnB1bO?=dpZX%Heb34MFDT7toaNhd>0grsrluW)4=qM%0v(GU^~l zBWgoB2P7Q0XH5QqE6fD zEIPv(rVcpGVJy1ul^TX=JdpiW?EbUo){V$?*;q9qqcqB&Qa)l!d2Pk)ITf=;1evyW zmd~sl$%rcJDr!emendumWeylg`}I|ov#R42r_|TY_C`{#dPKcwY{5i2l5PT<&`7G( zjnKQIa-_~B^XqD7RagDLXo8oIDC?G~4F52JQHIgWN+)Nfe_Dot1&G2K zKiSBzufwi6t6c1a+Jv2LD945QsAy+?M2yhTNrr(ngc3QbLn)nmkS_wh)4m9!VAYFh zv$Ohy@m2XEj5?$*LY{F*UxYkq6ALZV7eO_BurGp2{k$*2C}sV`U`7x1MHrolEB%q_ zi-5!oUxZO_k>a6f{OB=bNBpN`N%(2g)OpYn0L66z#sg+C$G2Kd*7 z|A<-0g91TR9xy>zlSGjlgXVfn3SaXaE1!-@LEQnV#>N%;B%3GWdYhbXPuy=Efna}0 zE8qMkH4g9$xYgEnvW*yi_6(2mQZU5YWNybZQqaSRVh5K;oKkglb%tkCo3oR zX)pTtrl?4xh}1zXP7##Cv&z~_o&73eMS&XZI!_5iR{jtT!G?5l3C{eXaXFp*h>?-L zBc1*yNe^@c`Ncu&dE_e%7nI>#&V{gfP>T~U^u-w%1GRjo)GLb4l6ar?QS91fOtlV; zN#T7B*fs2K1_B15_YKu1CAx<#Q0YU@gSsNmcOaY<$+^me%d(Zn=F$k_wqP=-O*F42 z)JQ64A0ijHmI>VxOWKkMH4iKamCR+Q>CWC6(ixf?(ixQV(!jC^$&CY)&>gDhPz4wo z-LE@^mvxXD>0c6#Fn_Y{RBi@~q7BUzV~oWa`n|>Jl1qfD73JbWgSvFI++#OUqAD`% zf@6HDa%NI943LO?ePo=WE^%=anj*dg$eQOM_G;W&QX*F=tg%d2N1#f{V|uOCvBuIh zqKV97>T8W!W9i!P^JV$GF8tiy+G&lY8^X^QS`*e-x+&sw40X4{{3=NFFwe?zvejcZ zMUi+R>>brkfe#Rh19x*Vh}uU%kJin_#axE?Kaw~L>$3y@fn`FH#N)gSv1yi4TSEW8 zuy83Ssmck}pF4i6k^P{cPz3<$_86D>Lh2PbsZmhYavg{9PzSh9rLWG~SBMvBeU}x3cA-*>%=anMNUZ}kze3}8BGHw>=KUAB1h{m|5dmBTF^Gd zff3dEH++w2793byyR68@3aW6Jj`10P%RG<1hA?sthj^e&v9__$x3twRnnQYD=(R%=8aF{VYmWiWWIkl%k(GOkc;Yg;h~P zqy1?fS;7i}r3EfpEJpI|eq=dRA1$~O#9be?=?I@`7PtyPE0(wo$|QwJS_bYBNud@k zbf`6o2#77w4lIR5(R9muQ3CB!mhVv802*(N;zmCaBL1KuDF4j@Y8}vJbK^Elk?hX| zd~Cy!0)qe;Es6b06No{Sq!~^ag<|y-jjsFP9YM=w2UWB5DU4Pi@jI;|TG!%$4?Sq2 zcdLl$X=ugRcC9?>!5)GQ8|e?E(;rdk#po498OxYDqzhLj{!_!J6>Sb;oO`qr{n2@C znDrfo4klXj?;6IZ1x6FLq5_U#6@uSh*CwECHGF*C9PAVHq7i2YDYqg;vC(Y^0hLdS z5@xss|A%gT8yW|{AzZ~{hYPGCrdMS+TdX{~PK}0Tlxp-y{=G^rrpC0hPP);zam^Bq zVzJI**D7UN4`E2^-UDN@njw@a1(rZu_@Qpg-Qlr2!xL8x`mQeyO^Wv&6KB`X$5X{v zR>C=DreX9EGiVukz?1Cua+T3ld-u(Fi`vSgEflv*y1Chr-zoW>u<~$QWgGQlm%mZ+ zwo;E}(oM}#E04C3Z<%ysb1z6?D|mJ{3SQj|f0aC@R|%IGO(a-(_>$jIv^tkwqkarB zDc4#dyfk5>VR@_jIG+(E!ll2DQ-k9(*t-jMBfGw^FttZx69ePL+Uj7}{3SIH(6K(1 zJ9NBHvmI>KV|<@Xzm#cCf@Jn+lxn9GF|w#3N(6PrDvJcCySvi!Cgh6c)20n#1&{lc~gL^Kvsr^wiw{*gc4B58o-Io zAjV0LsuO15U%76YhG%UgLss38PVdP~?@OoO6&WX1$QDOl*$^`WI#zRqnL#8M%#=7w z$b11|*y!DGFg4S4|BquZoRk}z^qLW;t7O4KIQ)x&`>2~k?!I`2uJmkB>54Xo=?x=J zm+MMz4z08@rYmigN+}ijL;%Hq%;R?hX9yt^1m{^T(ls8H8p|2CZgm$K!w9tH`LlH&p%hVOWz&*YhxX-xTO<*6 z*ve-5V-7z5w7~I-;m{YHuvVt)e*kA0R+mOgX+M@O4MuusacjcLHczId6 z!?eUqF%7ghX!x9Q!gZM1tD;sm?LQKGcMOavQROi0Y3XE_&NT;VPm2R56Z~bfWgjgy zedT{5cc+=!%@nmJ&@8*O#O#q@=Fe!U*`Ge-RO<2W5qRm7mITwexUkLiF@R<|&x|we zY;mly=JC{F_FJa(Y12jKpfwf?M`RFq^U{HqPHa+Vn?4{u&;qev{Akn=fClrM1C#=+ zpR_nE7q8I;xVvlR(6Vft#%!jm7dloBZBVXas}NrqE$+eFQcUl+MA@ZF%|Uv<#i0vL zAAbc4E73B%iGr5TmzHBpAGSD5=bIw?dR8{Q3&f(yRaL*7aY0p->FpM|2!4s_ z)7vdkY4##C`(7I0RcMypZRurNW%~4{!qgSO)axkqKZspGm*E?E<}c`Sv)`k6iSUMf zeznGYD%`5ESuQK57QOW@P8Yfag*XKf6+$uu5YTZMh%a0qHM+@bP`ZQie07w__u-u9 zQlo>Ha8M5n-Bki0#A3=*so471M(k17#>2|QKBxz%Q{2&pn2Qr5Y9pqrtR}cHfoIz4 z;r@(fVexz0>T_1;|F4 zIh8J?e%c|lN0+0;H<}%@n69AyQh4~%6_ldaoAGx$Vyp*9087cHD~}d*ymDsLvuyf% z-^!-5oMOwS@!%ubRPJE!#?tP`k`;@iM&1qSk-lIkjAX?LLFCzXF<{lJuhdjnp7a-t zg07(L$TLp{kh7?-IT`}VbQSf`YbeEZEe7*)vxD`&g8E6chu~a5z{Mgi;}GF8y^RWV zF>ObdO&3$V6j%sSqHNPuG)V8E6w{?>_AT9PL3jq<(X6NY8VlOY4+a9^`yo=%T~ zT<(Qr&*uNoW7Lmj=Q463_cJ|)j_*+jxKHp37~yHPT|p@_WHDUC%D_%NC8U&{t?~01 z@B^RCY_{n`Y`G%X4NFHtMc9&#Q$ImaOjmq84wx=IGa85ni{;s+Y^I)UzZ636`UVGJ z(KWyp^-+)fGEb*hsec8#xKTs7Xk?j{M=vkJ*ih{hA@C0|#PgdS^LW}t{j@@8U>9M- zy$UW+U_-%h}=c5mkx z{e$dH_1MB+bjb9bX=J8k3nS#9Xhx+nRjCF0+OSznH3=X)BNya|#d90Sma=PaD|U(t zC9IQc(-GTgzEc1N2(Bh9LbKgv?DBpDz$k)FGL!%|tk9UaTwFA_upJAf<0&bfdC})G zH>5Mq`CR6db9K#q);MvVNjJM8oq1iBa0m>|_T*eu9C)RdutXA}BcVd zlVcf1vxUkWME8sZvwv2U#(@tjTKIL8#;7kSL}kB<1?v)2_xC|wJgR5eAoPO_P{XMucTP#-i2ePTaw*)8p5Q353CKOOo3 zG~C%tPaGwy_?0sgL$@WS@jmj)9m^JeBAS8%wLUkm+DfR zrFzExc8SzZdR7iSgP+zRh3GZ7p~ZF?Ed+UD7&vklmqY0j$dXeo)Ks)r2&GE|5Zkq~ z>2ATx`X*?~W{!$RbtCD4w(=}QBh_jc?U^bRXyD32YZ?pUo9M8c? zcK}4DbF+N$KZX$SA(H37vY}cO~g!m1LRpHo8S;=N-`lfmmAKSpx+8 z_6SUL(9}mrVoDw1i?oc^KR^;)m|5yQNm`|nER)`m?@RReTQ!0drpK({D0rW24p4bX zD#&t35=1TOO$kv!!HiVUD|9k63DCx%p+{4fjDc&n$@I>=V5`1tj1)lHoAVr|i<%Pb z(vGY_x~NHS)$eEdOc!Eien#(P_0z>b57UR3u=9~TRgy)%vFWF{V3W=C^cYBFCR5M6 z1XdS5v7DIhKu=FJb%KBXAT#3eYytA=A}O z4m}MB%v0&!tbTg5$)S4z;iE{Td$UrUz;F5^68{AFenxj^_0xKcjp=@X{cw|G(}P*t zp&GxdDJnnLVZ_JMgIUEKV3gSqDzmO^8wBv{{wjemy6^im@{ zwdp!eF>P;j=vq$E%Z&;1WV)XF>BUBeuICiJh%0E zQ8rs>1^2Vmyb!U0QS<#KhjukOl!CqclST&$igURit#9KLN?eNPQYguBroHIcrX;6m zZ)22_yd6o%!(S_#-Vz-RYS0SC?nS#Aq0G%TPp7x?KumvVM7hV&G+j2|q02Z$Z{jF+ z5pM?$Fa-4aJRHfH{vITo&C~^4BmFr+{T=vJBmFaI_cqWjZOs~_+nS>0*^zy;G0Uf% z6w@_LQFhJKX=7GD<;3wx8?#b$dy~WTT-J8l1jQ&97tFU)h; zJd0L0361u_sCP}1L$73|Xf;T`5;}}PnerbiMU=k1n>()4a)lvc4X!UnyAc{fj|%Rm_P zIQLj4-PPz|;?fg%3HQ+3D9CgLh~;$<(=TW#2gfb}vdxKOIydunKyiz7x)tCubv8Ot z`l!*N73}+^j_UbCG-w`67jZwmr}BhcFXA40uQ9RynSYS!3AU5Rk;HW8nDlnC9!YdhW~oOd={}WY zne?!HKcK(Y&BIkU9K*)+V5hd}-rw{xti8+#3u3#L?pT&lxChv>$3Xk!6uCI z@vK2wCzN(N-Ivu*k2E=SUsj490dIaNq&T)+x)-IG9s@Pq12R309v21*wu&wM39#Eg z=p4FBRd@=?Jv!OSrafbPT7xdyUbR17N>1&C7YqWr}T3uceu^jN*TVb_yVNEZ# zvgy3BcqVihS?Z9oSmxo*GHE0G66%1iQ66iMO&u$Q0e?CgyK_Jt%Opwj;G}%+%Pg@} zk_J?gWl}=E4~)k0JO|1};RuK!c-B(`;7IIobW%Haqe68kUa~ zECUS*rlr|&>K7^aU|3cNEbU(omLy==n+Z!t1}yn^X;?nZ#xbxbi0)ciK$l1Sf)aY8 z)xj@%vlW-Rg^G(x=#q%f^d8g?1tkSs%0-@2%Jeqg{HIH@Y{G_)VqUQh5h1vb| zenY~h3$x?&euG0dWcx+*b)VibeA?CO(29sZjcI4AgZDY|s1x;Nhj>2QClD^x^#;nR zHivzf;XHOXmB3eQv~O=(zNv(!QSJ-h|x^0m5mG07jcI$j%1GY@{x`pVqX2P^2ZaCjZ3W zi^lr&h>@aA3mw`OQLmTLn^@R%WmAIwl9dvo^#qn;fe&VGyIr0Oz0?|K3nL}V75=+^>H}g*_gnwl4&u5=+Ze&pfx#_a2J>Ow0EIH zowm;u#Sj1o5Y(E0@||5uw)K|Kjs}NT*nSBnO#EE}CD~4By3ArqW%sK>scfG-}G0e1L9vDCVLl~&s z1ckl^!A?pFi@rc&NRot3D9L4%t-vT19gGBoC~cyx(Xu@I#b%V zHFlioxrL57m2R{9sdr(*rrYc|^)7U%%l6r(F@v-qA!jIc%bw|3X@l-S)juzEZ0fRm zX5n%%;(MA*uPi{_IBi(yuuXro2kB|lu&Ku$qzwzB)D>i!-=iM8-^!u4APgS**nWAN z+CXtRmhQC&3BN?zyb}|N)&rwgj}7fXjNt;bK@EYdIMY)AfaxxP{P;qsTJN^|fkJwx z)hPnTj)X@RrcVB;^#H=aIR{Chw*r3SALc5c`{!jN&tUCQKp>k*5OkZ8=y>l!$E>H@ z?EaF15~k@_GiVp8vxK?W0ZDeiDh$DQ$mF41qnJUK?@SSoP^U> z!29$6q>RuS7r|cBE$aVlbm@wBej#%>9nFvwH} zz@ZfM=vDZ8deqUHpgq_sZp2nW)^@Z5EEDXZa~vAwdqph7WfihbceV{!4%5?y1YbZH z5C=x=Uj*R?t-Zj|1uDeG2d$3f;3ES=@2V23REc+`#43=K zvib>N2|yGq@f{mAM43rctwde$QzjWKS@?-9a;!qy)#3n4D@<9V z@`Vqy7a09TvWEq}>93Wy;vhQNF7SRp<6kQexe18JJ4xomK_WkNZJM%G!~K3plK=_c zu+}7o^Q#N63wpTNnnXVX%@xIk2A0V60z`M1M5jCKdY8>FqdcxywL8# zb5Wt^(l*^G9TBwmYa*!AqA)ZXZOSX8{3UHix)6-><@%S(8W8uVFq!wb4W zgfziU1#SmZg{(?Ifmye^tU`>(wF+I!QNoz*T7?Cc!zETB7g$b-mx7IHzoM#OcJsIE z_Ej`0Y0<^oo~280Z35aNFyh=ekX7%1LFD?DBR3n^b_+x6-HxCo*;CAQ5f$bkF&>v? zSJPLyEIYrhwzBHXkzKSL(REQ=mmSF^K9v6w*K312rX%U8GK{Sq;9#V2o`Wt&e63aY zi&&-QBa3QXA+b2W)F^R@c(`!M!RT>wD`FTk>nrEPb*R}{bE>NFyOnNvP1EK^;jev)Azd3*Uplts}aMPQ&Cr^dPZnqMVF|< z>06K4hH++XdCgfWwrdtH-JPj|ux5Qh3|QMR0#u_p8}~fxtLkD^p+|fV9`@xJJ#6SPuYNsfmuVbcLjq;sUWld0VqH4> zxD2w`kWM}xlc@HV<&b^YwFgSO#ioJd0ijEKEGMbWF@rNo-y6Em@XjD8YI%WNbCOLT zuN~V*idG=>_Fnv^uJtEi0De42!_?17JYAXxCm=Wjp{X*68m&5d*!Z_})enQJT!I}w zEQ`OT*3l3Nq_H0%Lhd%q=+Pj7UX|)2l^T1)*c%hy*3E`Tgt7p9pmAFPyCrg{6Q=|S zq!7Yzn+@)splU8qhzSUA@;gC;HXiJSr;qyX8vBK?^sezt^K5=9d9 zkZPUqkkj{x#>5td37iI|7h|p|PGjQGtHHzrHO-NOFkt`M!NMhY|JFy@p*$Za8Jg!n zhy;O`vAYy=h#&2^P9d1F#xPxj6T4;8Gm#X0n9w@TrS$74(`Y{^VcGOrqzAq|W8!4` z^y^OOyWUWNn{ID)tTA-83Od{u!38Vnb1Sa-(cb~9F0~F&TY}{>WASnEI6RYI-pC0_Y&~b zzoH}Pf!QtyFR5-wmwi(3l4KPf&)$k5kKSIuOBiYon~saCYl*l)R1;-_&KQeEln8dT zls(Uf1_5UaIj3|}E{>tUcA(%r3l~^%Kd_i)M(JzHF61uhD8{ypDJ)`~sIl6;5_YLN zT0k79e7_j}#9#NB+M;r`Sagp-4EC>bQ1}v>p8##Qi!UAQsX6@=@e8QlDa5tpc}{sD zlN~pMU19Q|#?PRPc{$^iy=B-P-)?avX7CLrToP@#qvE(PzR<6=^n#aMW%^bgz1HG% zsINGVI5=`*pCGoWuceoIg!ic39FN%2dfCHmy4u`s5iZJEHk-eoYfQK-5YsCyVxYV8 z5cO=j2Kzbn;bx&tj8i=n@Cs|Ca(;tu*QP5a;N%@>b6bmZXh1F%K;FUygvoUJfYhdU z%oHkK41e@iu&$Uak0i%1*0(Tn-13Whh2Yw8%f`(GTxR>A#i5n~%cjc^oy;*$RJ;00 z8V!*b;knkrK^@x?RKUeW>~S|26mu~{(}c|%D{C!S2)q=Q;zhG0H9O%JxZ2V!?L|u~ zFwwi&Ww&DnjrMVrEMS-N;{~`k05U=|;DUIqvxsv261o1IC(<4mIP&0{a3|J32&$Xi zBF~;yQe1ckwK*k)USWyE-35^YE|=ZJQkn}J2HZtuk9DA%iWfi?F0p3ej)uVeK?ux= z*#Xpf^nHjg>S{Ce?6-H|UBPbgjKn2P0D!H}VYieO-8|1Rs&{-E8^>86=%Kb~8N=Rd z(_-?iG4v<6URilG$@faFG4xZp+v)UhX-R=o%G0<+{>GpT8c=&1mYUqj4=+&2wbxlN zv34M1(ZEs~HAoEuyqhBrCIDecE-zuXV7aO=ja|y_;dX`=80zAW zaXGP=X+@4tpWwGJkxc>^)jOeHd_$gYHIU$WCZ^xB#?k{=aoAuU#tLQMI*R@m4ss^b zcd)9llL!kC5J149=(Uc*rEHLKJBof2_0H^M&nqnD0$lNzHopTu_cB}y2GD4M=0~lg zs8Y7H2|o5(*QT}se|xc0;uRG-;I$|xJ)uSs04?L94$xQV$%W zIZ00~big8MUcwqfHv(fIDh&KDhL1+rwaW{zqye%Eu$Y<>;jB(PQg@D5=98S{GH`oE zlGte8g@49*O%8ZJy9I*urF7)>jv1gZy94j-OP?qJVnCh0t(t!|3x^>a6c&Tx(w)w< zN<>dPA1Sm;^v*V|Xo8cQs#K#9GX5wHlX}H3$z`zCPy_*c)q2N&Ax!~7PKn^wfYad; zaGLGh&4swU3g-qmCP48>%~5(7CR0w*0}T!u0RM4|?5)dr3skAc>}L1*X*0MBp(1{& z*ksMo8kUEGJzcQ?+@|FYN7y6;Q%n)|B$9dQ5}weUz@}}birBSR_|z6nfn5X>(ZMCK zp~)GI=dDe6 zMbOWS*iWP5RGYB9B5)H>tm>q2AIv%M@2^YXk4u!I+C*?MO!l)@2f{3F(0Fz}!&(T| zQU;#P=E-~6g;H~d4w|s??=@&Qs=G&AjxqTd)LM3zAz}dRmO>GmfxK);^`*c}2fMT& z!E^)`2`Yd=VdLv;i5_alG^tm1%-dR`mTA)_`RHtcckm|cnzs62Wc9FX5p}loS|(m| z*<|(ydU0%<9Tw3opWm4+6K@VEAh7n#UmiFFC~SUHZP+H`rVy&U-kiWmU~6m8qH&O7XkB_zs<|AF;h#F)%^yTrO4V2 zt^E;9+eMx?nQn@Jg}e}c-VljfN2GJFlh2pK+-oB}))CYfeqIwvSw|rLi}KaGDgwtK zEGiVhd4=*)7SaCZLrOCye6|w=EX!<`fT?En05d(!n4$ILv$q@)!kub_$Lt2{c6nJT zyB^AIZg%LOKpSe|iOg0A_e_(1M-bUwLPKoRjWh@e7sM$fbFFN_CevR~jOohsR3LQ& z$(-Uf*O(gLQzTMCjOOdjweKJFe0*Xly9{UgK#f@B|P+6{+u9J|`i#1p3C+-_f|eLarJpL;c?dx4-|Qp*#Vo=??*J zpU%YX@4g7P&nw*C1Z2`nA#QKQC*0=*x01=o9g5pEUj)!*!EM)2+^)(J4)QKMc$CNw zEMAr+-MrZ}1h*Gt4aIHQ_cdoa(7)TYAfYtLE zyLTwPTX0yYIZ3znn6$P`-YI~huSnJdPL-w6QqfMJ33&s<2h?t%83>S5exMuKD2+g4 z7~|)|PP)_iTP0uos23)1IcW&UObftWf$5*#ZG#jmm->dKdAE5SCz1+6!Fl&*^A$kj+}P3}R0%~7(|KWX$Zbs>m@1ts&- zc(Mo#k8T!`%vKI2Ouv^;Iz}F<>GbHog|U5iI|2sL!v^Fgm+6~_#o0RK)I5n95(4yGrN`bRaWM}tBAQ3kawJ*WU%ed%VG*@~pp(o&lM$5joR z+Ir#KhU2j6#f9Gj4r`^dtw4;%q=*9?G`yYDSTuw|)St;k8vOh1aiT0XkKc_Q;-&`p zFZX35I#XE@m%+~nE`^a)xNucuLmQ`JNJKX_;M`*#Pp@TTuSSF!7Y>b(vEhf)Ac8ue z6fwi}xsY+fpbOg6b(knrji4+jmfkS7%X8wiUruhWifRI%qHMpAnxw{Gj5yPkSPr6^ zxuD^5G6bCcZDDp7LZfAw^nNz>#0xPvHh)DQWcOP{=YPH?R<1}0cz9z5 zg9u~XA;JSzJLEf;JGm5(sVh}H*z%0BGIdsfWb77Cw~xfjLd_1}0mS35fl_dK4n!XC zVrp<|76#jU75in-8{^DI>k&^Y+qK^;rvW2wyq&v448V`oqzIh_gKK_?1$i#kR5 z0nl)&h+~NVqs=%wPE$h^w*Xr|RbTUebg~MDez#b;x)@QoZCc*~$9^4)e&UaG=|L#7 z{=sKCS3*jPFilIlOQZ{SsmCbP`wp)ga6_mxzKR<{hAsWnqG8R{^>wpPErR_P?_(J` zX4zx~Hlm^g=XUT!Tz;&NGBm7EeIaO>I%>g|B+rl}kx18r^l77!Qa(4jsr|1GXkN($ajI+Z0D&vHqctqy=FJbT>k$Xmf_>p)N1XaGe--F?Q{_*CS z^|NOyCxA$%QQlN0d@-o&kja$kGho5{Aj2o2zHYVzO3qR>%jbykpAAC{_8G={om3t^ z!puXw46F<A*^^b*YP~; zHZD_MoDm)qJWJdnp^2h}QMvL%{}yE`MwJw_&qVzm6hGLn?qt_IZLh~O@PWV=G!J|m zd{~Tfl$Ubgoq%)4bYXpD`QZ-R_Y*xT?fR%9uSiR7is@e*rjitwgS*LAE-`s+`bm_N zRH(!M?f56%((fZmCh(MOshWBOjNo1fkn4?MTKi)gntLc6PSE%l-QMHU-ZD8EdvF`X zMhlHjGeX~@`RU*#Y|Hf7N6t_T3Fbq{=#Ct$W6zg*;%%SpdrR2eq)}1?d)bM>Wd2ll z`U)PxgEm9FD`QM{%42zyE>3Mx-~9IO86qmQmvsxpKo9vDgqWIWD1D9pSKZov8A6iF zXaiIWOx+PklpX9A(X*`%-2tcKKO;+iLI%Oc#mjuCNNF1eaYMw1n#1$xdga6kw*xpI zfKV{Kg`6x1M^g50y6;P%*XuylFeb|lkrb;oZ;r&dT;RtzJ$Bd7^AH3O?>40T@jus1 z{Si&!Tth1XE7OihycmX2&%=Xrpg>5A_K~3HB#O}9h!0y9k)L*$KshjfI2}-v)|Hq2 zLU;5sI->U^cm=I8e zC(`3aoJ*FJ6wqB*!H*kpfGk?iZA@$A*y7ssl(8Mf%jii%>d{kJ^uLE8hMvR)a~zBU zq)z@1-Tzj5T&@J)v_XXBkoXWviW*_iEx`ey&2fsku#{aa8duqIY-;Lc*PIBGd$7!9 zS=a=HKKl$)hzNX8Rw)4=w{jYlkpnhd64}1I(3=s+>}^pTWXxmssPFPZuSorpv$pxO z-GC{}FE46Ga<*Rxda~QSHTYz|s5odU+b_cMHBZ}JRHn3jF1ycfYPXj{ysUD}xZ*c` z|D}>7KSIJ!bfHPn(nZ zSCO2Ux#oic=?G|Js}pcLf1ctvXDev4eGv3oz{+m4lx=z|4ee!VXjd(W(z`i6nLf>T z5PTO#7J+ru0tM^4If7*x6{q=~D1oRt)+oAgfn!-r@8|gHgI#=nBZi4aAfOvsGyoqB_1%|6_x}NB}a}^Ik1DFo{wXs{nzsFw6U4`KiSvMo7lubv2 zfptKxW;YZt`qJ5%err3@$)ERIGsh1c%5g1FSQurxQ;~=i&UyJ4JFcxx*V2w_6gaMB z^A|d0rmTDKmAmhjLt-DQZ0-dz@QAdBZub1&FDc z6*Op4A9izKW;wiDu=U0Y!@Z<0ES4kqgBh7!ni<7?8V@4OiM`ew=-VxyfArO0-I zlM0xby}Y>4PZZOCI`kcABwWDgBJw$mvL4vqXP$Rkp+CLYDK4Ohk9)>qZ%lkft%;?a zq*fJ3JTVVPtEaQS%F(K#YF2f;vg*uJem%dsY9!~G|A${@3db~(;mmBT8`%+aWCzCi z|NlIf75IOTWZ?5h{$^5elpRSgL6uW*z8p!t`Ex2~jci_K%^mSeOQ+1NtQ!%|X3rhb z3ritAo-(t1ennBiNXDOmw^NHoQgP%Okh$e`U+&GR%$ICN0;0ar5m4<&Z<6TxSi86p zGMr{4UV%r{0u5MrpTf*|j%w&DRpA>{=8)&8!U;q(S%&%y+f1uopc1J@k!-`jLseB; zBU4jdEzc*(B?|RemTgGty!zm=EIWPMBL|Ng)mD_xMOzUgL%|U-jJf6V-jJPfBgDqT zPjliyX((#+D_wQvXNJn3c=arlP>tjA625XX)jyo^;?bx~B~hNC^+KZ!V}5mQ9qNt= z9#Sf=YMfnPC5?}P)<#q`D3EIyv(#HkM;L}$$VeWOer+iqSgtNd>Zl>wqr435QJ#^0 zcz&y6Txo$D32-m33#B$*OdmVW@9$y+F&=- z>2z;8IAz1#FJmyUrSxJ?g42J!W6d_nsY zgW7t-qC}_zn|PLL&B~*V&=hyTx>FRsM721~HG?ZBu+nVC1x1E~cq#3JEyFy1&vei8 zaG(;8W~Q^DKo%iCK`q*()m<6%WzwLbY(gTznfFKVB$DY3g#C{HjV^h4D6qj!gkz5Q z9^ur|I>0f$Jw)C3Il9UXsKQCe{lERX9CzuJGab*$LMUEKLh)wNc;B)RvoMS4IQ0{b z9*DAovlU%>W+Dio`VwM&p{*^sI1RbQu&7*HFQEzIRvsMNarSk~EsL)Gdfdum+I}Yd zy*$seOghSkfSMGe3d${mO^d15-p(#Mzj`5#C?=O!M9&;8k+Ep)uN{kE@p|Uye#@k; zUu!Gtj=A9|neNGaJiw?5GiZ+*zD&nf7QOy-IdZ{G-7MTIU{G^sX#Ad7t*a)WyW!;w zk0aR4<&7KI+2E)i+N`+@SE}0*9COS-!)ny%CN4k|9q@FA^S@GuLG;Dcl`D6n!f=|q z)77pS<+1xB{8B>`35H8`QCLag|5N12;NYy#jbDXK zjbSFv?Z6)j7JF!Y8pjI4Gscwp8qQY$Cnsr~Lu6G=o-BI&Mr>7i0a$c$#fP}%du&phO@6u10u2BJD`?zp%GZY z1;B6#2C$W5Jd0zz31{UPs)zWl*VV5VXRsvGXasJwn67RK*UtO(Aa0B44`E>YC#n*b zMRyBha+tCQp>ts9%Bcn_2PjD3j&=u(gmB1f>-5D_$*1GvG{d3ed(h<0<}kG3 zCt)v7+lV4{=R(l{TmmmUIHzNEU)&NF{xmH7pq*aMcViK*=|HoqhL$W`rZg@F0XO~MYi!KBQ#2DW<;v^s>bi+wec}OZF-VLD*rj3Z9 zVp-<5X|3HaX-EQ0w`YL*EJjC{g!p`?Dr#AD89tQ-_{}QEJZ6t2MWA#NAl{QwQc0D1 zyGn4>mp!NoT;JeW7QHDxsED?BIjq_=W)Sp1GaOuQf?~^MhkgU6J}C2ILb5i^Pk`nD z=W9|Q`_k2@FY$#-Xr3cCG2rYfw_hyFrdPA$c(|s)=_o=h10=m6Nta6!{a?-Tx|=ki zeWVC&CKR^Z?K8E?ApX*vuuQs@FwalsfC3&_0EHP$1};2wpaO!^JEr4qimP-)q_pV6 ze197DDP&z$jj)Us(;Ml~;FniBZ2pwq!=*OiJ6KdUU1auv9nqrBGRvYnkM=o+Jtq3O zJx01thkO5viOppTtj8A9TlsMLG?`XNP;!fDPrBz0)$?Zu^?Z)>4AXZxz1JecTqO@i z(K{`jbg>El4VTSR>0x=Ig{GkhEby>Cp^NPnR;kH!{UT&x7%`?CAnzt1kLl?J4hv%M z!nb_7PAT{qwRO^{xMhVUpHn4itrUDtY2I$n@jc5_U(_vnH$m@n3!FmBLbhen#s%R@ zhkg``MnYf#yUL{2fehu#8SzEOL-XOIoR`8NF3JG`mZ^r{MZa{3^n_V8>Fxy%ecz{F zbuyjgL+c(xrSk;JT??WGmZd(Jc1hiv@&lY~3rXM>nFxXVc}&Fddt@eW%I~pE+KhPd ztx>p^&yeI*(z3w8_`rZH6A6-iK@Nz0*McZG8^#2`+I!ogp-;}lvo(ot5J%3gki;#E zE&?Q&ulGYDxfm%}rA+Stwx80K2NB6%LL$MiFV9y(#T$S+J7rmPe%Ju*td@!X1YUx) z%<^XzpvG=ls)6`(30NcB<+v?KXGW)!4W0~@HWrtOd7wP}#&&^;G>@e>6zyH!=9EDA zETE5@p|r)Wf*H2yTvF?&2F`MjQHrr+mO?b5o3c_E?$#!VcbHEL9$Q#RCc<4ZUEBs< z)t(KT6-w@F5GL7??c-~=aI&k$0VHs;zx)PdzJ(D1i|M*HY|{`9%GiaY_is8Cga;0S z@azD>xt*3po5x{Uk-X<{7(yh#c$mVKPPLs}PRrF=-L2MYM@uK|QESyal?wa8Nofvn zBuf=Sj!%!mN0{k8!ax~s8n8?ElTQyfM+wgyK*pzEMVU^D!#&@UYmFEiL=cyYm%<lW#TIf2gB={ zq2X72!?ooJfnW8~X+AYZu{r*uL#L(aAER_y9QGRc1-Sy0%k)c!CZ%A3n3oVzfn+Rq zPC*&=IIIf@JCf(xlxj=RCq|qcpEk8ROdlBj|Hs{*z{gQtdBbpZozhWBwrts$#ej)D z1MdqNz$PViRZEqOGn34gWQJt2F>SjgtK}uwl8|Lvi`8vtTP@oSIBWqzg4q{&k&P0< z77PhX0vLn9vate&gncuc7qxADzyG;a)h*lPK_Pj*_jxj7(N%Ttx%ZxX?z!ilbM84< zPl#pht=EGfAP~s%(K6o{3=K3y>5Fmto&}fpRhU=Wr@Kt)_9#H(UejK^hdSG%v{&z^ z&UTA#GhBKuVc|E^+XQxQrkPeJEvEP33}(xBrgt^iw{ZAjTAs9MK#S9v-nR=wm?(qj zb_2QI`r5kWw}dC&(FVH1=oEWf1$2NDtI&KvIM|6G3LTs<2tGmcQv~N5yb_F~ki#(( zA_LaoVKYj2@}BB!x*twN4R#VS-MkF#mY>>7d>}64ZixP^04lYEeo_ zrRWXFO-vx5xtN4agM~UFK6z%0ftx5(7gXZj5F+#pI<#J1ReFYPSWtu>YJi#hlU*qq zn?Xv>2oAV0B_cNASQRsfIt@ft*O?w@6R8%?7q~prfF&|6gKTtBizXv2D()ykCWS*a z)60g7r^`i9d>5=OjOWKWCkt+B0eiUx>lH2R(&J43uhhB-RxP~S(Ey?A4n=c#v8}<< zC+roQP>dt%0P<#h#kXgra5o;okeHrsuvni$zXW9pN@5`P=cxx#&E}@)xPG;i7J6*`*()nEqSP z$7PKBb8r|WN*93EA+c?04Ls+jUK#91J@8qE*8tmcghZ6~H0wVEx4?+O5E5!J+iXtv zkCfe41rQ!(TuqQj{el+cg#}vvZ~+bihev*Hpe~OF6{VW=oYBMfT86Ko@EoTCeZj3+ zgXwAn6b7_6B$TwlbS>T}JimH@#rk(?&0zpv0zfb19(EnrVsJWLj_N=5gU^3>K~(XY z)uX`3rIGh9u$Ts0v1ifJf;gx3Gik6jE5GR~Pn?QjI@zQ*(F}M$jm92n<4lJZI?+LN zd7t0qm3X7N{0X{zrXY9xeo)2S6w^uQ0GS!Wo+iS$2@XV(h0JzXDEPq^^jBwkj3G6w zLsVdT9r0w<5fTBg45q#k*+`l0YD9U1>1n)Ct#>y<8Zha}5>GDeY4jxGExAG)5k>t0 z1Y>4;Sdef!JP%W!&~KbhuOcb9I!T+9klYtGsr7iR4*El3*#fE@ z+Twu#4(}?!RV#djR}Fgm5ITZkZuGnM4Bn{zuSeGqAm3(s(Cgaf)~x)dJ`b)xKfmx{ z1MFrwy6Q}ueXtJ;SZQi=qg6I@nm&Ht+83r(z?N#uLNE7rM12xMl4I*JnC?Y20D9a9 zByUvz9zp-^P^jDH_3w_>to$a4WSxh)`}6y^Y%I`3M-xCmrHDR*$ZHz>zRZ^%Dr;GB}m0_?+ zuOxbL@rm8AVNfq9kZE0uidv3I#1;V=&tScr-Yvl$cAe>Rf&YFE{%XLVL!Y#wJF@15 z7%na7Hw@a+g7fxzA@WDj1o$;3T*QBDd(;m8{W{+6ZuH;Az3wU3gT`p9IF-K`Ca^)h$`?Xt1Y6GP5f&mB8h|nla{1Cu|hpDwotz zt5%|ScN;pcUmFIk1bj5NN1DsvQ8o7w29uSztF(;4X1`0>maP1y%RE%4*`)VnAZg4# zcJjQI`@oEWwH^VBgyh=|o%YIU?=rkY%kN3cSKv{#{5D!nwPHV$=&ot~yOe6p%5Sngn-bIgS;Oo9KXjf=w)HvRZ80id{wKqFI(Je7<3V; z0Ky@JLdOV}VI8U>S#<9sFnXqphnSCAvhtf&=5=qcbPp3D+0vCO6@cAYB4?zRv5~#6 z;Vifg4Wein4z_4OkE7(RC`r#^M=)tq$siJLJm2UY+Hs8czDLzb5LIplDD=Hw@OmpA zRmXQ=^o~zJx0m3T&h(;`lWMBmPPBiyBI9>FGQPYuE5AvSO5~C8mSB)(DTxn12Hfq? ztsW*Ip4ujDE9hs?rpKYLMYquUYyLXA4sTQk{)i5&QXTl%>%gkkto)`MJ-D)i2*-H% zc&sF_wa>h+Jp%(e@bfw$w3D#i(;pKMD@=M5zIGf#09&*pL1w=yX9uc@`nNZ_HwEI= zemDSjl7WQ`O%_ys*h(Fa5hT}k$*=5itcI;fHER^D^yCS4Hq^}^aO0*E)8@RnMEok?(%ceGur1e52~#cR3~ zkMumdbZ1MHp656w3)8w?!@h@A5~JIoQI$8eHVNfRr_3CYk zDS=ZaP8G|fs|!U42Gat516i0_qrNd?%H6>b`Y?nHN-akhnX>G1gy|wILORGUe$hd{ z!(-UyyDAv&%h>3+5(&js7_6AOPA9a|f{hKoFffw}Alts+3B277rO*@&AaO?r+5s|5 zI>wtnTH!fs%pyY}Gyf0GI~MajF>{wBYXww>gBIXBT02lyjgM zXn;<12>-=CTFF7qgVK9FfJI`C+rT|c!4!bn^zks@r5y+l#f*zrY>P4ST`j}5QB0R5 zyNn`wSBtZ#(paB@b6K%|DqWGpuOgFP)q4EngLZ$=-gRQk>f;T*XI(??+i&m<-PKP% z1Gigfd5QD|2DK<^~6ox=Qg;CQeVtQu|cn&mf`l-~C^{@7? z*&H#*d?F}Qqx`3b;o3OFmG{Qwk`B2Nwq%n0C$2Po88NNztpLZ28BLBm<=p6GU=wmjI6x=X2l&q zII1_a5|`XD@t?G1Gi5OJW4)3|KPv<#OnT+WI?Mbz*|zKoQ~@6`V+f&z$NYa(m=S(! zGop*cWbeQptac3i0`1|y>O0-7`xW%a9;g+{=!y0$u9n8+qF+pziw$DBz7T4lG{<;@ zQAB;smd`3y|31LNEr{?0cF?_CuB2&2dg&9v*E^K)2nPy~-JPB_u84N|2EpfiD_arW zycp^=s0BnTaD{{qOB+S>z6hFTDX;(=HLUr?j6~!@)IUk$gjK|p)e#Q{vU_LP@!iNH z9IK$kK*Dz1twc9Aje#b!h-ph5NJ4k%Be*q~lnIHE^PVx_5;3EQo~X0X(mU-Iigy5Y zu$rwluntx+wh7xPqWfTzWI6~kq7J!%e1L7<4(u2v?Jqv|twFZ=k~rSNsV!q=?Z5B#BRRYTjLhIUmjwD;;PSdwJ^Ul-h7-JF~Msago5 z(w+-O^+z=-!=Rhg{Qtq9f0df?v(Eocng6Sgod1{9{O^$Yzw?Ose_r}{XKwz%c8Zyv z^wDwGQP6RbkB&hf9Sqj9^;F0DGF=Kq(2nJ zGmbE_MexBwW`x5?UZsiOF8xFPE~FP4bnv-VViWYr5C>$-N$y|P$K0s z$#ZuE2-*)8%4wsR9#RN;km5!$ZB}hDZKfVf9Fl4mF+Eh|9*M3#!JVLLd8?obRKc_p zE-N0Y3{QDUPZhfnFJD;^y#UJ{(@ll0_@1DJ?RZ7E1IyTd@oJmAx-Y-H8yFv9O~1KVbI#bfC)U63wJ+tXMltY4V+tmgnJYc9up+|%16TE3JG^B zEdQ^NaIqlax-l*~_5!4u!;!EhmvOI%UWQ`^T`&F|LI_MR;S~dfOeWy`0k7z${8BsR z)h+6kQA9iB`8NNVsjm)5`7jSD7w3?2+n4|;&(;N(+Nb^@K*}mHOpBoRyCqTTDE2*% z&>dwm*&?PYhs%s2q)RjmdXimZ3e$57Q;PO?Qok{Ut_I@hb*vSpCs+{kKugpprki~f z_bU|NgpwYenMis}PZhYdrq!Zd1uor$%#!^w9=g7@*FR9{-$Im#uyGDu0dOK_y(3Wx za1hib{9_IBD6>x-8+|;SL#ZWVPr?d>A^Rm7fLwAitx>$^CGZ}T9$gY;y13Z&H4qf`8)(llUax`mcj~dct`t!6;;nj%X95j_II#)`DrP!}jU45{-E8L&{SJ7$TjA}| zRv2?}^+Z2|epLX>Vb{`6rN;vOC_;IMpeq_2?dE!mr6hf28bBv$^%84n{|?IQWzs)k zeOY)bAd(%r^|9ZrRYtHEnq!P&`cOWAMF_9?5o{8)8k#k94Fg%)AfnKRZSv{zaiZP9 zheK%KoV~OV$by(|(nswWpqZwW?)p2z#8-nkx=d1)V{7~v=k4D$q2_1Z4_C-m@RlO` ze0k{N7IprYE;s<$Z5XtNgEiLC60EV+uLYgC*6$3{voIQANb4I=-K5Lgdy(Y#9bLr2 z=SD+k0I{&E!3z5)f~Btqpl&aK)h)dew^+@w5Xp~q8Ad=sB4VG)g>^ZqP%_mcv<8){ zNed-J2f@T#hX0(OqEQGrsmzSF(k*U7r$fA;B32y(xQ`(DGHy7MnP=ek_)aD&rKU zdtgdtdVI7x$Lc53szq6<$4PlRBOOo_M^OZ#mIEnVeT`FKYhK%$RZyQV=mDTRTP>(u zb`-ccrC~Y#8t9NQJRqE*(mTpRRbXSHO?7_9d?O~SimhdHc|<9IRiREKZD0ia#BpGh zceY7gAZMEf`n8z6H(>NkkBs$ccWP+KUqia}gJ9DCB&({A`}}1x^~2!XGBIzj63r<_>boJaJ^@C(?ZV(C zGZ8!xQo0vwpz5Ol!es(M_Qi*l-ni?t9i~=@T=144dnog01tok`;4)pe2xs^z#oga0 zc-~zmdzw(lyX11GQA{_WU89IzD(Ej?VU~8t5nD)4<_&X_cA)vS6&otZv7y%B5VYo# z5)Hv)j3TW=DJHZ8f>|6(~t#F zrAE9<XxhTARk|`Fy&y- ziyf*Lm@h^s4H?05Up2EeSMU`fDd{kMC_M7BE|VQXQr8|yQhyMVx=E4LuGTJ08{Mo( z>W{53SwAO6k+d4_TJdQ_N(dHaJYnxf@&3SMD4DPmyi17=AeyvKG=o0TJgkW30Yec@ zDj=Ez7kET-eU50hwf6cV;hNAWe`&oh2+JdyQa$+giG2GsWOnBY9`Xuy%kma7tHyEe zn`*yXN&(1=VH^@Yr7d?lRtPr6vq`^b=>>x~ibO^~jUJUrsbYH@DCq%qjnn9n7K>vL z<{#i5N%S)nmFxFIEK zGRonqP1A(d?!rAG@5cQU4nV9^BLG$3))MHK(OJWwTU#JG?|?4=(`o}srVlaADE-qC z7#WM0UWXxdBig|0Ck+UTkkzX3ac!HZpF}i)o*UHg0w{^;HOi=ysiGeoS`2y<=Ll51 zY6)cTx9}*SiA2H zMgv#MMzZ3bBXs7Rzo zTyIz00856UiPXzs=mBfczlS*bgKX(`CNI-#yan)p1lJy%M(8Czk^RE3*2{Qnt$-#~YeHeGl6eRbe3t z{fs>b)@k-Y%c`HXhi%=Vwk+(f*m2OdnZd|x7DKmwIwi9ml+5O?@oPo;>e8Sa|A21v zAVLhJH!Nb=v0*_p5gGsZ{)j^hqg3odV1!i{^a+DJI@#81_b!j%m452MGD=v~VlBsAzqE-vYaWx^eSCn)v&R*u zjge%5JhAx$SS$j1H;$wK%$>7N2x{!ZB{1EF8?_Ak~%y;0B8&HkO3G#|cNi&Br%oP(zMO@L?B? zG-Tmq{f{owuMyTnpFn2~(08qJwAA4snG0>!T(;@KWRz8kFo{(o(`JHRS}X)_p#~(_ zkQt3J-2);u=_xHko01ltl|jc4E*3e89C&m64j%uV^vh(5=_xdzpG-5Qk1qBeFbv=e zb0ZV2PdP{%k`|j`wq^J%b%h-0aVOlu1VE11-}jH~nZ8@XdCoxK8-AhV#h_Vpg_Q^qQ6~|1RI^zYpM_ooAF@lY8BV zo_2DKcQ|wod@5|VtC7seVaK6wT5P9yL)f=+pL=Qm?e|q_Vwm#u@9iz8YbZmnEs>?e^b-+` zO)yLSiV(deC(7k)^D;l!OZWd9fO!;P(2glrBQOk9#DA$hrTf93PY)_Qg9=IpG--sY zgeXZ=Ab07o%plTZ$x(spKQe~sKcF_`3n_@b1OCx8W@!ET6khpN#2%`cC!jNK)a5-w zD50@1m7*P0;yE4RiViHvQi@zj*#&H@m3u%mGdQSDIWriV+zO`u5^T^pYG`Q1L4FnS zRagNpA9$7c=Z)RD+RI>urhihkzmlsBp9%De?mKFS9#I`y5=e)u&kj2DV}R^*z^VOh z4PmNP6VGaK%oIM*l2jPiE+IvrJ@pSr0eY2p4c(~Yykw#d5b>h4i;hWYyao$`rv+Zf z!0FN?_^ePw{zsU${bzayyo~85L&3i}6g(6>Ik3w=1K2$Z?8QUDek5RNRo>gxA^JqU zJukrFuCf4!4+Dp^%tP7%ElwBdE)BL@OiOk52>Mu>0HcDD*t^9$B^-cy0-%5~EOePu z!ID40c%ToUQ>KnYmiC8oRCkF$l>*HvAM$QHfV|U(j8I}7&U$Bt)8UA)rP@KGK|AzE z_?OX75?Oj9uZ^Ec8`Kt1wpJ0qf5V55`|S!Oy&zC-N@QVy3)aVN3Q8hC&XD0 zOaCb-dz)YO=&eh|TqEcHb5QXYe#M{>x>$O;JCUVJ^Lh$T2?^ikA3ZmJEdTKRK_pV%_H~m5Z9%+ZPbaOIGJGDIGh9jM{~R!c3yq04`J;q6NUt7u;NxwCaH>4$SYhCjL zukkTBJ*PcY*Dd8C)Tt5_Ak2Bqm>0AtiWrA63<2qej4^EDDq_^) z*h)F`K}Ui!w0~q+Gc7=eq;>mi0$hS*b+biH;3Y|h3!mOG6ck0wxxZj zf6$G)a-&g-R}$wWXeI#4r1!+jWsHf9%0G~3)j^MbkMqS4Z|2;$J0Uu71|e^3oG#F5 zkgzNyAcrcc`ix}hS_MqN&alRW*w-v>W<7xY%=-Qc58dL+3vrH3@ zw<)zA$RfI~Xk0OpeHs&)-dmJ1oUlkGY8g%# z3(+W|*9+VZU%fD8wg*+GHA)Q&%3~8Tb1BqA&>@pKh?;@~bQG5d90wM76SB)Dwh_FS zTMNCeOSNO`23a1R^Wka~;db}1H|+2PQWS_aXe=?TrH2cRi} zPlnXA`|*1G0R$B>Uy@6wSWeY= zl=uEO`7z`H*wV-EG0vooLS8!4GvI~BnM~^!Aw|uzMSx|TNo(ctsUmnm7-zCccP)}r zYN@&(`lJLS!TgwqkYn`NIB*7((U)Aw`A>En_J0~Z5PY%Y;KQr1$f5PZAH0zHbJ*^1 z%ol>Pk85${V6vHJV?)P3Dgy=3X#BO=Hm4zlWLt_7QHxrlDfu5IQp*rX{{oBC*hW3q z_RV!3ffZ4oF;4bT7t99x1gOVBUKX6DZFxkXWN@Eu%!$n^&6Efm5{8Gcw+dBrH+C0X z+Yyz*^vH5dp;)O_;)agLn&2UgnuwK~uw+)nw`o1wv>590x)C|ZVs|2Zd9;##k+hIIS;AC9GNqnmiXD0NtEAP5AB>M2iy*r} z4HaeR&SVO%5w9$^VBcaoXlC$y(PD9x_(@TmzMo<`qlZ-z^92xTD!8lyF}8p!Y$s(E zWfgE_DRjd%aK->KO++y$%&f$cG+h0(eGyt@Aqd6-i$=!rx6ox1IXm@JF6Z()7;#Uflwt72iKzbk@%3Q%CXI+m#? z|LgDND^xb{Ns&~PGa;-L50-7h^(CxnP!1CRV`0)i^wYEywPlwvrJ>xQ$FvMk)|jfH zUnZmB3IK!px)RwZ=y?t4&n(GVVIo%nEQ}9pam*`ji>07s!n=Qzsb<;;J+Pbf7o)3)sl^nI^Xcm!A5Oe*&}d2cc$S3(V9`e%3vU|;w$H1%KXM>mg>b1qgsuL1ihUP>~I^Lx@61$F$%`nwtmY{bBhVieW42%Y=d<2qW=@H@q zYO~6Dg~RqtVM!3Va03wQU8as2Kma7cYorUpV8KE$so016z{N0!c;Wx3X1yKK9Nr=M zNoPnR{`Z)a$fg9Zi4-~`j_~Lkessed(bnGyQDc@$;xHG5dg!}gV&4j`eKe#LYbZ`* zPX#5bbW2*1zT5j`{2h}uPWl2xA^KF|7?Y?@%!D>RGj@M$sCW=tE=qwM6)OW*&WqaK zuuL^-e1Yt7#T(>ml0IP%!gkc+d&WXYCu#qAR06r|}X zDvHwx7nCUk0-<09+vRFs0d~vTrthR^vI`@o8n4;GHvMOc#>a(HY_3uQPZb@XVOUt6 z4Q1HubYOsV*w#-yRAthdL>Ijja%oK>O1nZXU7xTpf^yw~pX^aTF|C1+jd(OKhFl(` zL+zISLnTLPD+9D%7=l@J1De8`)X&>D&8aTa%kiISSlTut+pMkkgbq}Od1shy{hko% z1F!*yfS{G2Cw<}|1b3WR*u@eqa-9me8^pLzA3u=)#_V5+M~jWwH9|XY2_} zqwA5>$Lw@4%?KSL2+nVW+AIJ)-bE3_^@(wx2?JhZH18!XiK4?jLfNC-T>nD@G&f46 z=)bxR`sezRqC^&AD79yoE#twkey&2}#47?c-gY<|_eY(XmR^qkL}f&{#9H9~uf`Mw z;;1EmMES3R@nnaNWiif~Al8B&wxzo%^z)mlpHr?0`uTwLlR`bnb%Mnf1DJy`)MD|O^k_9Lb4)uil>189U&@|pgForORNsZ zE0_z~u#athX^3WJ9iogBI4;9sN@UGtnC!!Pinv5r zld2c`&q7}H`YaviPRB4txDl8bal(c;)STr(Y7e~0SFZ_hH7K}BGZ}6fcnn-G%V93W zVn-&?Xk}Ri&!mYE8T8Ys3LESfTsEm1zfN&Yiua1goGBWD1Tyxjtsdtb5jW-pcS4rl z7&MG4o*S?x6-N@%d>|xyAVKA{1POvzj*j{{WDMY!et~>9@V(8F_TWgUpF;%$_~m_< z_!d<-!?AwqAd#zoSJBASTZ87`^2QBARfnUlQk}v`s%0BrVER%IM~qYG8w z)wc&-ng}KZs>KN$=*LMwNZ72zA<1P&NyJDh6$v|B?$M9gZHD#89upa6keIwVaK9|= z3ZRb~SG-{(+r<55C#$>S=xYT$ZL*bCrGCLaN~Ayn%Q@ox!v3!R0@TI&($K*SE>5oH z4WJ|sqnp{UpQhRfS?V;)7P-Q!V)BMgeigq(rhHHIE%{- z_Fh%SW^pKl1gZ!XkJQ|S5(@>%vN{f?J0^39>~g zvbK%P;I8Np*&+`l_-8?n2liyh6!PTly+e8x8lVJvCHt;0>Q8A>_GC)tLmZ&z_KQAdnIA`+)O;FCvJ55my2N3XgCj<-$ph zA{|WZ4UqG{S2`$tK4@4;2Wj@C1270M!y_bvNxo$8b@_gTWH8B>48AJgKUFfAxIJk4 zt-nArP==<@A{dyI%)0u-KpN1`#LuuV6?nqzxi19W@uY&xpOp$q*&ZSln1>vF{2nu` zpD{!#@Su112hbrEtWZ)x>0weqnMeiy{wY$yiV$Mq=<%Zo1zTSV;6o_D8POLC5K_a| zuLvCyIUp=T!MP#~IAI5Rrx{`VGQ$uIJP~Q>D;~I^0>FfQVXAcLU#!n61;~C#`jAci zZ!HC=9{zbz0L(!|ffmqcAPNkvdW1aSeLk8{Fkj&KLV;IMNh8A9JgMQRl7UAe+S>}} zB8-U`_`^UT+j_*eD?)?*9j2}CsEP*BMJJ3aRz{#z5*4z5>F0~12GLer8eAT7X&{ke z>I}Km33qZ;!14;%Ojjw1eRpUxba7+KC_vT=RUP0tMgs|J zxDA}38kqRr-?)Lzhc$rH60OSN;qxNO8%M+Yf8%kK?p6&Hr=e!+^Q~q}z1Ike03D_g zV)|v$@-r2#_C6PtUtqPILEdn`+}huc%QHqZ31F6PHHjG>>FMb-$>z3ocTSxgV> z{hT&J^gz4i%Obl!@Vjdi)9*DXbm&UmjgX{`H9~ZC5{mj_ra$RD4l>L#qKP^nKaAm; z_7qIA^e4R^NpI+N94)Gt#%$*Ti}YcQzIhkxnwuSHDV6Vw^xfNgZ=-uAKCr+W!ctM&$k71m6! zo3vi#t;2nXe@4gCzLIb2pZ~mc|DUZrW#G?#-j)AZFB|l`+?@@oZOA{Vo_iqgJw~CQ z3i?X>ym5E_cmIU@o8ys&@uAO^hH=)s>e<7|oi_aRZo|u0*fe*3?Xdb&r_x9cr%wHX z;gp^=ziBwd|393bQl&y0&S1O*dBdrKY<`nx&K+J}!@nRGWaGlR;nGTwx38go-n`+A zX+g_^n(E-Bdq+Zq0C7np;14*09hWjLDm-zez4bFp0Bg*J~k7Q|liO$W&O5 zAEarR3H&s?9EEf1qzYuiFqd-B)U-v_^A<|wp_vHhHmNLzrWX7x z&Tqajmk@BokqH4u1WE4D;K*ENy%J5ERo^(jS~_Cn((H}OCFL817V9NBpn5A=`_#{w z05?k0JZ$3$z!%P)7gyaKscE&%O|^}6_%vD#REn{t&6+!}Hny;CZZp2&IUYQfYTB%L zZ2rQgTKT9RQS?Vrxm43ykwbDpEMDEXcy3*+x^aHE3}D7+nn#^S`!p*a1Fgrr#exb7 z^sEIfbv0g%+}rww+Pe7M#-Z<9>+5Rg*3FWs2Eqh;vBp_7mSsgy7cbGy0yVJGT#qmS zl*sGk%mwkGZ~UkFn)B-$WL5w`Jp}-AujV$OLaC-LZk*dxyU2SO+G$p3$G5S_=h%o} zz_;M?7&Q^zySiFwbBuZ!Yw$nX%B7mXO9YyeWf#HuzdSFfM5HXT* z4N;#+by<$l7?-IIw)8HvjB}m?7(sGE=l;}lx7w>>Mk%Epr3};4xjb0%R24uU6 z;C?_i1x2rbzC+?k+Rz`u7w{vbz%fSP?2BvVIQ#lXTR4aP7*RH0F=D6y;1y-#*i(!& zjn@?p7&^V(j8kwQP8}v)QHY#RZ#P?BbbWo|ph%BSvvper{@S0v!-cEC2rDt8g| z-Q37F>+O4STg{6OTUr>@M1UzAM=@G;4+_7i5Les>Vf>6S%wUx4g||wKsil`pIAbD? zCN1tI9S_tFB_=+%^)j57isNk32syzu^}_#PL=P>BhQmjWc|3;(TTsYBZa_!v!KcH}&!7v90b@LUj0MLg5^1ekVi_kQn%L2?E*OsE&u?@4%>&SSf39MH(6SEsh@@m;qq%DZckw~J`u{$FEGi# z(E*tZovF{j=}kXp58keAOTqdn+#GbC_s`Jj9wY7(;*?_ovba?kHyTI_HPQZ1T(lQ4 z+7|+icO>BLN7+!EmZNzZ2*p{C?1i_C?^Qi@TmUa45js{fLJC*WrQ*p$e>AYj1_@54 zyv#p>YsUt)KK5#vc69muJF0u)z)QHAmt&0 zcbW8Jp=(qy-2ifBI%|+Fl3r{ubf6Q&p zPNK{W^3?D;u&IAj#%g(c=0P#g*x-I;zOHiOiNRQ}Kwx!F)3t3;rq@9AOxLs_R2#i+ zWayeUEaX>>c$Mc}xz@i*U3yXg<2nU{XCjMFl|l6L=BTVzlU^z8XPck%k+dsUNrKX2 z5U|@%ucSoEXA@IS4j`|Cw`n)L;jqU+`MnA@8Y0(AcOZDtPlGkLIm=WG3<79B+xkTK zxHU(q#O0ZJo<&no(q$RjV76^q(kpK~fTzRTK_+(!HD<9dGA=V4>c0=Lnf2xg zxWwjB73K)8v_v9kSlrfuKgd=P!M0a3Es7!~CEV0(&ziXwS}+8ta1`BdJvD&e1?PJR zaAr5AbEGOT<+Px{T_}JnO;yYNuMN4cH+WJ5E`YqVBxQU#Z0K|a#e<;nTm7=7p4%T> z89%q%A3N}wgTMG^_4+$&4{X7P8>V>{9V2_|-_-4|%JIDZKs*S5X&=>$zV!?b(WPQW zPkWM~{$=nJWjj9&w(d&!UY!MBy>`n9lOD(7sDQT@BvnW|;Nn9p=J+5l*h_H9tEU7_J|ImZZu&O1v261F;>@vJrjOs}8KbdeNj7QEyUm5Y8a{jC z-A08Ocp^g>w0*-hc+8Jd2=MYFur%-bn8WrqN0`2Tsr?rm{#Z=-r}G&0EWx+82Hyf?PLBs3hs#VTteHKXd2%9zhB_!;43^9^j#rJf(#X!x;}QdwN4bQvp` z)hc7O@5WXNH@4AmV=HBNQ$c1{hZ242V4kzuxsjmNi8xP!l)%2haK#1!t+q`R)kg=fE=aFg~a_;)ZQ!p(il<&+h$j#RcVf;qxz@Gj-;Qe4AIYVAH36i z;9m}a<4Ov)Qr{3g18Y{Fl5dCp+94R7Jr9vJUr|-XRuc2K2kjq&w;N#YvRQ9Zrq8{j zG)>p8(KRg;I#4`93xz_6dl(vy-ms>zrJ<=lrwyDx@BG^NcG%G-*Ecjxo>M=+c4lMk z;>rJ9J8xcfrhk|@%D60xe)2KtqRJWk1x~?hKGW)w2rm$XF4rAIVKkuPB7QKDD(ExPFU+ugWArNk`kqWTiJeMH#Cj#4R+(V#A>WZNv2od#QHP^h9( zdHq6q!#0GBH@4Xc|G+Qwptyd{ehX?CrxI5aF0kUNsZ|&X>r>^jTv41BT1x#>!S?hB z>y!2xCjByD86&CLGEADFQoWjVPYAm?rpv+w=iEFx?0rI7_Z=ysTyD=A=daxR--b+jK{J6k+7dF0K?=6k0Xh z*b^+AIY^7FK=JitFkKJm<`p)U>j2p;FMY|OvZUAnJb&{R_+ zlOz-29KtM=y13_sA`hw}mi~_!Wr>DmGu4I9qf+g9$nb=;}g0K$@L%Xo1YK|(2iu?D5mT15N$qkXv3#0d{aVoU4rzLWx(vMCAiAOL zXso5rfG%wigOsBKcAYeWO_m~ToZSijqLBQiFk){#{#5Uw_uDPH0<41RZ7`WDb=gP# zE)rP0iGWOLY7V@n*1Qq)aG&&0QQa8ErC<7}-Et5U6L_Hc7Sm}simN@}q4Q-gpAIfc zoAkH@YFVi|@JYKxt8`a3aog#X$)i`>EhYV6Dp8y6(p`sMM`!0)Oux{fG?>Xkz?2BI zTNB?3`nXcGwj02`WMFm`RBTX>HK$ZOd0?)B%T~g**>ULUcF^t6X0c`q;?ykNr{C#u zdbV9%9^0(D?9i{;RmLaTA4Fz*RF{Z?_6+j?B5-)dFzub7U)SgLD^@kbTMX3mDXqV% zXXuyh7X4m#g*DsQh)r5}^fBDrZXpy+mFn+{dWLQT%3eTcnKt8hGnAG-S~0#;;!M9D z+G|f$H}Soo*NCdV0W>)s`NIsN`&&ViS0A?B-BXaGNjesDw1Z>30y+<@8;GwWbzu#_ zV#7>bEy#JMGd<6)fkzy?b$X5=G}$O|XA3+AVNZ|}ceEf{&^F;Um_nrVK!DIi7Is91 zplMtQatCV?J!U25y**0pDTC>@mXx8>^PDjlu%)pK$7n{7Xn*zYVBq)p8_el0GKC^> zXcYvenX;Q#K;|@&z`s&>V6qDxP&9O7|vKcPm$qN!2xnO<>wC9foYiT zb@S?H2jd@3zZ!?{F6LDUngjZV(_}DB)idWF_4q%pUc972nie?z3~{PzQBEU*OC#rLeib3Ay0k4v5xk$ObYA4z5SumI?hUPN%FQa)~;?@!Lbs8Bna7m)&exEGOn zd!!c;9CAEGgsExy4n=0ZJ67K_{K8lZo z&Z|h@9cV%got}Y{I($*!s?M}N8CX9XrYeQ0q0<^sIQ6k<`(nlB+{cFyy^b@C^bxY- zIGi>LOj^)OqvO~-m+5h%KqM#|M_ZUd5iJ&XMmn)K8#)e!h?D1=bT@xBX!`d8J&iCtA3LT-eIYH4TYoU}S(~l7FElVuk!1M#y!3l96>Aa}= zp?`H>85Q*))4+mgCB}^Js^L$acKwbJ{oR1bh*XU^hb(xsf1HNW!MnxL@cm%Th%2DF zdk=h>h@q5rs3qm!SlXQ{d!Mk3G!{dQ&2_#mtlx|k55TGzk4&mejYUP5qkQta#UkPB67%&Q`98mH3#;0kedFK;mEWY@P` zt(g{J;yqy1Aes-~Cy!~3iu*ik>hz%CYbZ$HLa|_~Vl>(Hi|)!5{RlH^QC+uJN#*{iUnY<^vYm$noiX+zsRm! zk)L`+(2^AKvH&$y1yDXlr$Jum%;|@GAX;@$;uEjLr;yiXucX|P)SJLQ)n%VOBXK0E z{9M++nG0tP@7OYYo7nK+FRYteQy;ILys&B3uuPJ$cnwSD_WbhyA~79}RnHsNVbYhe zK2^k*nK=6>S0@(KREyLU^6h)1X}}K3dZs=REekX)?{)!e+T6N$Z8N%>=RsbW?_+K% zDQol0fD|k8;v#R&y?y~A$O`Do2E|;1qk1Qd9sM}bL*XsvDOu<6w1H!cgQz@a6|DC$ zHn&5gge4oB4)IBTVjD@9=_e~Lx-}m3VmTH>r}A!Q2P{uGl1X};p(eJ?Rl=@19OPF_ zc0nqsz@CiKHW(5@<*Mkt(7B(C;Ea8=IFPQ^%nI5P51CGw7y%ss#OB6u5`h~Y$Eb9W zX_ex`@8hcP>X6H{I{}?a8C{}2e3*zL)VoU`Cm=>~Ss7PwCtKnSp@a{*KM9B?l&J#Q zW}4UQ>o#}K4!X1j!{;k7NBlQW)az~vYVFLAA-_5I{zE_m zMFK_z+c1HsC2FDpMEILpUlvOPP(JdRo}Ub{7f}XOKh8ruKk> z8x{rq_^D@xLQu@YxQQHT6J4g5u#6Je4b3qK#d*F(<6N^7D;RV-7fSGQS%LZ_2=xlB zE0k^+t_LGsv%3l_m8$?dA${2he?xe2U^#lIY-kQT^NQCQG&r&!2kYCCDW(@Sml4}W zV#l(_VQ9OC#e>@@)tJRlCMrCRdWG>+vJDT3w7O{n-`y&agXsCQMf$NRD zLk`jOjxa77M6c)=ce<<~8x~)>Fyz8AvcQ5r^hh_tbU~Clk{0ck!XL2y+EOu-a@EQ3ZO2o$y?gC1|q)!2tnD zPd=pL^MfJFK8n?f90)O{u`&dh>hP;mqPwz5vr`TwEjmd|Wz#LDzl*c?)$EW9K{2L@ zQa5I)UqRjffbau1N!z4vSX36@K11>X1MW9u(Kbr3Txv|I6z4()5~t9Xt4cQZs%piX zgvY|GNW5!km)65{Lo!N#0CFT?V$kl5A{&qi-ARmLswr|A1|b`vf)`m3-q49-fEx&K z_=knRFp5y@jVbyc9Jh13sFn^0;}6(Hy)zCRdS+njyikIsdI_^H)Lp1cs8L2(($BV0 zNC(4v*UOQC1!pA8 zLc!JX66NMAXMa-*R01%-crgh^!?@x_QYg*uK&hrxH`Pn9g3&|jG*tc zau5US4h`we7){HcSO6KDuWJCVkD`w_MyViHcdEUl7>1tf2z!)1VGpNeQ%V~pY|`b) ztpBXus=vkvHfgXuE5VygT2nZP{fkvlOJ!@p4jRKobi^}U8zEUjBC@%Ub{lcrUo@#x zhY3PIeNfq&dM-%$isOkH)l;LGHjl*hgg)h?C*jGwa9C3=_R8qggPjqlf%X1zsv-Ar zn@SLlUDg=QCcWD1d1wB#u%Bb*Mn^AS z`jBP_9FdyyA$cjDHeVV*+7C$RG4^915`q&XQ572ad9??l zgf3PZj4JHzx1QiatfiM`TTqIh=o$viu_8tRO~lV?3##u^)K77urJtUVfe>)@>=f-E z=^7>U$O8Dg7<9`CF4LCAC{mP0j1s1c7g~TFV}-w4lU+hA)Ti~!D+rM4O)uW?b9$LN z{Y|~h?<>=F`M3Sg4L;DX)gTOFdK{uG`uX$4a9!PcECd#k$Z_b#uW$uga0R@p?FMe1Ejib%W+XE`$Em`fK#eT;9t zI_MR&>xh<66;`?NUrO|PWVN3U+3g{QlVGDZnKjtAO}s;GFWK>7%e3Q}giWf-O^%=G-C zD60Mffau{x@b06NUFbC>+FHyoXnm1eVU*C*i@Z+0<#+Nb0sYQc>BHp;YyGg&0 zVg7zmibg9Kj91W@871@-nqk@mhWYNe$=x~9`7se__Lk!_|f&4;l2)6(_Q~%gO_6op- z&fp`<=qJ;I3FrZ7X$h#nQ57C-0@CgqfmZ_IktS;fXd6m#xuGmH0@bU{X$*a%BEcud z_NWSX`%ONP>sHTvD-I2d*$?`{a_tfZXV7KGFS~$FwsR(lwx|Y2wNm3Xfu*WNzf0&qz47dc!kgUX#*&2(!KKfkwupGnoO}DpA4iK ztbdu-6b%3voAkmWXd3c9-GWcp9`VO0q2DaBf=ppI1`G5)9Ux@>5(Gi19I#u)Jlxe4)3j?-h!8WRW@dMPNRN_3N!t3v9ONk z0Elj0Y)O`~uS=#TxO|2J2J)mS3}oYCkAZahJ>Dk`tW*o{69s&Wna4o(;T`Re7H&mP zX^Z^cxHv^2H*6SaU9k_^S!anO=y$L{E<&dQ7PEIy=$aKAr=0wQe!N{5AqTL4O6y zTcpuV*txNzK41(QCG@8@i=kLD40?A7e0M+6;n4~TfETlpM($Z?Rnl?LaMngyLfx*w zQek=tdR7ed2Ib{p(H7n z@v>Y&5IzDmc)&)$N}%|OsHPuNgMJs9s>ig9VbJe#tv^#I_Jk)j7bYZ{YZ=9qLN^!@ zQ8)&k6%#uh^rp~$*$lG)Y#=lyLl;=n(0jGAJzsGsXhvn=KO7O8hg>m+h=`FKECdCo zn&AzC99ss{$HzjZ|8(9JgN{^?hivFzN{|k4haYa-(l+n`Ikb=MFjRhw&6lZ37&E;0 z(axpEL2&;p?WA=OSRW?;LLdWk!uk}tBAGP|`VG*Bxx8BD@+~~7xxBK@5@G#rIdg#T zJ&wCnX#5_t-zcHmgr*FpKS-uSFxnf0glUd+^>{FU!=x95H1P=x*QAj5e!INCZLx4% zlMZM-CViue%6dRP4-&F-;NzT3rg#T32sTnM>b_(gj!!cNc=NKWUEA zEUSW1OX&SZ8gz_N^mFkfFJ^j$;~W+-@?|lytPoA9Nn>T&=7X_UaSZ1r>_K`tjsZBB zK>Rk-4z4%g5!axmLsdsb&;jGd(Ri1s2C7Vb0-T6qV%c%H!xQXIJ=DoG6W`NNOHPC^ z2+?my30$)fASRvMqkhq(LG^>_1mMqM+Ze^viQdbs>?O>vxgDx#bS=h+!%%`X26|-R zYE=su)VfmapCwGs)_YrIH{Ph}e!5=x$RGpnKvxBq(JWNEQaHq062CyEe~{54=KDgR zm5>c5$}qen6wqiVws*szH(5CH6)he|zNHjhEMeMNkNLX|lghMRbz={M zovmoWmpw9s6&M239>$Ve*#aeL;UFzZ$@;-MlCCZD&_aU*=Jttj?u&&gAwXkkiKqb` zEjsnmI85_rSw?pYn)?ic4zyw4En(^~(8hbx#uaD-2)_g<#n%sgB(KIBwdUWiw^%G+ zXuIz-xBZ~BjUx$m2!9YHDge5mi#yok3nqOch+ElWiMLY+dh`ZoFj;KUB`wH+YnXIM zN?iz)>L=3`OR`1@bsBC(8O?_EtAiR+#whAslEOePCpor(IgLlW8qsP8y#S|@RCI_m zO!Zb79TP`fcoc_P0R?BMY{e*A#r6zG8Im_JvISm%&D&LGTuRDWpg?{^Ejr0fg~@R7 zl7*tZidX~1OzSCr*wRL>C3N{YWP30ba`AWLJl7M1 z;x@D=6PP~4Rn#)v*_n>re>$X^auGa0;wm4V9Dgq@=r<;);_}IaH(sB7GwI@jxG{k~ zX|;?=Gyo%%aT48}j2b2Mw&5C+nD(|oxWU}vFmw`KpX@bC^!fC*F`(>pX=DcjuS}*e zyA<#8J83ZO9*yCYFkRY^Qk?Ch(O6P~mrR4jW;ZA=2crwB2BcXU2igNE=RteThjxt* z?adstt3Cx9bToM|`h6HT`7mC}!B_(rVtvGb%l#pZg1Vwy_5Uf9#)ww<_4nhApp!Zp zl&tY+8`ODVRreS;7Sd_>3z%jEuI6cofs^R%Hqc}V^^)ro+~&sug6ozGiD}X6nS*K( z;nK8?MWRRO(hH3QeTT;X0c`H4>jNd2%t#gnOm*h~pantMToli&*6<4cP+3#7b`?TwiB zYA#1JvNu7O>brnV-Qs~AL6g;y7_80Qsay~Av-X4?nXhg%1VIYJq$vJi<0j7{uZ)*_Xdc08#f$47C2kzllHL;<$4(kytIyG+7OCk{Kg1cw^)w; zNW6~^n?x%zXpEpOqR*^g+u0`loW?dR@(7qDM<1Z)RTeNn5Ysgftfc4$ARUa)>|kY7 zFkvxAuT?lfkz-x}7EuZ#QPQ$LelOd!xC=em3(pr!$2G|)o?p}8mSKz_<&@XS$UG;m zH!q+$Y+?>W<3UeraX5Xk{^Jlta1+NiSp}$s*7Wgv%duiz31NyZ0PooJ@LPJW|Jxz2 z>$CtT62#ckPaA;48QqV)Pl|R{-~^_AD~~~3F_jnHY$HE*Oq9qX^x{1-Ad7D&9=v#d zot|C?G>X?U6xHXMNIxxyTpB6MVF4y5(kG4lxj0q$V4{zwQ*yvO5@z@}Nsw%W{qSG>W0G4;R={)?;Uwo0#v zdEQwdY)h@nn-SPJ_ZA#r1k2kr>&xWAlv*Ni@U^QeApF`Q&BoX^pN|MeJd2I=ZEYAk zOnZuH4mj0Be;Sw;;^>Ax#jtxFa4>V@6Asc23_H4(MB-6t?&yZnM~)0R?KDkrYbPi9iJ^j`XCYYH~G>Jf`Mj00Gn7u5>u(PuwP=+B#zHO z|5b-r{TCK-I?<&$7Ia8zuQf4PtQr%2rc&%#@(8s@h5WP`_svd#uUaKjjm31bi~rI> zi(~7d*4Z?TO@}sxT--r?IsrnrnNG(xOLMV90GqN`Z_jgTMfvGchc8VG9HHlG)2=u$) zNxDVLFo4t7hQO}Mhphm^L0`GR=@aO2ZGhG#Eh^{{zDXB`GW4&O>h(Q<`#3!3$}{xM ztQsS-x}FoqqvOyy=q_EIOi8xmV_nE(w%X{iv8pWqXbG#(`wlXR(M;em&2rKgpwIx= zh*3*(EeWG%lEz;>xM#S#ci9Lm!;6cqU% zk^T~x+^#ORs&RpwXRspyn7qaIb&;mQT)JBf_YMvh7#9#lPb8uogCeXGM-ZFoH|;=@Ne}38K{86x?-QVQ3T2E@ zv{K|NtO%%(J|>apZKnN*EygCLAyu2eIxvoU(h#nu;Pb~?05{7N>VfACh9bNIT!Su3 zq6hjULi{Tn9YE*^2rVJ5Ot{Iz1=evnhM8wN5yuAUC$<@$VT__%+AIg|3_A>r9^Yot zx4@b!l=k;|Q7-s{SZBoP-h?IndkEcEJPZuYRPu;Pzmi)p;&0q3n%tpy{8+FOlLHMd zQA(^{>8He2(!WMavZw;>pp7tHROsTe1>vt~C1?7!UlF~bp&P|?Ir<4ae`3V(ys}N) zIt(A_81w3J-W@Cf-Pzcm;-oj+a?vc#wf4&PYPT_wV4eYa18h`>4Z4VhL^wm z%|POm;8TeIg}m+vlqSCyqYz7ABAm&n&N4>P%TNc&R`+xP-fguQ zeqPwhX?X#>aBPXCr|uC*eh6a}ZE1~Sj18z=AY(;QLOwZCF{vwQ zRU(2E2LBarbIBo2@3#xRd(u-Epv8=0deVp+Cf(I$(Io%w39jFPMSNy)D#qq> zY@6E+lYZL{61Z1)hl){`OGK+aj!~kE7D5%T(Dh?>p#Y+Dv%xGem`4R&c@*CGhDeDh zNuhrH2P9L4+lG|#iv#=8d*OdJZ>*|9sTkYz!zj}k5T@w+y>y1FpCPfjQwVQ&hRbwu zyJ9%6g3)4%K;;nbO(oo$9z(vR9qiDgohT=3+NP5RxeGVrW&JY!J;s2J=@INjx=U%; z9NDBTsHA{n`VcrmCUsRJ5DB_Z3ig(#<^IYF4$E$WGiaP%mdyn71Ci5XOEEwz&$J+@RWV)5u76(MJt64yhv;-CBc7ym+pPBFmZ1##SIHL=lTmxaq8;3-ZRRs&f&f^8rKSf>^Ua8MrGyv{#)Z2dyez&)w~k?*!u zna4z#uJ(}j29Rd11fd4HfXx$wnte!iu?f70X?05qNGfJ}kK=NU$iIR^tKs3Y*m9^) z$nAZIsG=Z3utpq;aCVSetHL=v9Mj!}8HR25v1Vak_ZQ;04J6QnI0b%}uNUGypB5TbvvSQ#KN_tH+9%YZ)Y&5f3XAH|}Hv6J`qJzN623^IW+ihify z{i-ldueGVw@KYmL4Ojt{^c}zuYB?)IJG@di2BnNq^gy$uSbe-O4g;3bF4J7BB*Y(w z#!PLv*Mn<&}*ksxQ+%jzv7t~So6dvhrb5MQ*)0<3~`<}vYreB(I+K(d@ zw7g6owRMU3bFu2t3btu=s2@ANnW3Gn7TsLnvSLh~UIl{)Is@-sV-&6SX5XX_3UPYT$3xHtzu+_o2D&afIbLINIdt#oT)&_9~HU~b6=J@{y^Yf(hT?C4w>Wq z;#pEmHx|0SqwDIk03#86#QX^A*88Y?CP1AWhakYf(t?$7 zUAvZy!Yq9cT4a}1OUHsdZI z!X&q2X$fm9I0D+@tV1i00&R~#6H5>--}xKW)^mb>K81d=QXN461`0sO=(^fAfo+_v zad^g8IgdMRuV>iq%ykUill+6GHb%{sKf6^{(9FPSIc}2fa+E8rN=tktjbrKI7ATz_ z@OCqY!w#FJ#<5JBTOdT;!*M)33Xc@<71}H%q11M8MH1#rXyPdmRT@Vka5fyMj%9ka zA!|&cTfq=9b)eDxqg~(-A`P_ST_CHTF5IleQHz$SaE)4#zPem>Ud1h`LIpVqs^BGB zk-jvXS3r(O`l%|<0ThDv5g)%f=pUSS5BUg^*3S%Tu5t2e%D0|S@NsiQeLO64JVf#W zWY3ElL~!p%7#aUJN?aM=(`K#k_&EL9#sb5heD9+1T|3WdYu159G6UTH>ZC$+n|Ah1#URI0Yr^E2C z)J>i_x30ReWjJ>=Rhz@9k;_AUv>S!z*VfcjpD!2WLixEfya!x#!{tE<-9x#A8Tm;% zyrQ)Wa;Y*xn$}o5d+q}75};ptNP*^hxkeY#{9lHqDSV*O0!>px#RKC@lUxNXR5$9@ z-06kqCBii|!=xoCgW&d)W;!2-xNW{osC7@A9H^&R{b zeYWX9l7>GY_izQ==4@USsZ_~UDkJQQ!NGB)->{72cl7^Y6yza5s8Dx-1SX?L(ZZ?- zS8@ddA(avQKTz<08xpDn3rf#(1P1OB3;+S3i2v(QKv%U#ImYz1?&_uzf~MT-ah)P6 z1bZ7rsZho!qJxPjj!74Vpfrvl6ByEz97?3{_QIiW2i4o9L*FhifKYFrAebylqYH^_0BV-MQCgutH2-VutqTbH)6-1?ecP*m|}2_@C4#rIk0^Iivuyv zf>mr+fo8cq@R4KbGt)?Hhp{zX*OYnVE}2d5bU&)5%g_24^nI3SW)_L&>n)KVs`{m z5%}oxB$QY5p@yJ-)Sf~H0%SiRoQ`Fil)HTT$p;znr(_gf2Gga#QBdV{fOD;B%jw`E zUDI^!n9%2uZk1Fs906`l_1`jwoF7`_Xu_fDD3BianKLNU?{1qJ2(*_hO-K2 zjoSc5h-qaqYDlmnzo>}W?|#prk1l8VA?gUNS``oy!@mQRDESwo58j&?h@SH-MwgVcPMTxPH}g4 zcX!v~P~4%o7AfvdLvgt2-sj%u-2BL7O_ooRnJ?eWyR5{9s{7ywe_(B@W%TbN!nhpc z+%sj6L*!$De-@W>I71MAFJkuaHwuYq)pidWIh{|(RTWpq%~aJ1WdR ze-nZG)A?c>U0rFC8}cD_BlE695zvEo_k1JzJyJ*p;^E`AGyW;MQ?Wj4gWb~8sW6-j zXzdwP_#++`(yRJ-x{g#kkq$jjMr$Z4^1k!Vi_&o}xCHa|BKv7)6X9raO?R6f-Wm*$ z^>xLwVVU|7z4_~NU-qkqJY*tSNxKL=&7(Lf75^la?l*-LdAVpd-&Dn1BZ8{ayfeN4 z!}&7Pxu!%x8Lq9*%pT+IY{Od#^NQ^_mTUDUSUR#O!xX>T0LRX1QaOe@M|-g_5yEtv z!b7|DXj@vGmuCT}g&!g9mzfBj{B=+xBMAu<{3IqDo80za2~^+}^6OQ!mOVc7RsEXN z>0E6-NPh&ZVGWTl4nTmb_lU<#I8CPahjlbuZ4uNbY7QRm38B=G_-Y`NdudOZQ|k8y z7bG;ky&}y%QawOQKCza7SI*5ytfwR-UrcYaqMT}(AEDu28VjP^Ux>u%jE9YmCZah` zx`Z-{l*tE`-oC1WfY>;>XwWKPc)lf-Z5up&>` zQ`Unnq_(Po;`z7diO=wApJcjg#!rIA2`Sc6^mmg}z&8Z^+ZdA8m8BD?-2p^7oxe{} zd?1-3$?2GCvRUzncO#W1u({<-xnNxo=A8LO`pP3sPuoMp9qtYcK-X4}%y02JnicJV zD&IDF?DpUF(24WX`(5EhuLIgOp+46bJNRdUl&(h({8JYX!BB^X)A^kke4p+$s#uU2 z%q&EOoVQDpFKxs&B4x`T)e(D=ttU?XiDglI;)mLZ;k$nQ4i{AvRT#X3SD@L8r97eh zQTwYLcghUBNZlm&rnP%?9NaAHtS$rRnKwUJ|=pz80U2T z0ove|yMR~T<+@=uK$PR*T{_{biSL5SMnsbd>|8j2)>$YfjLb5QTL+&iQE__il#U_; zWffnXjba-@xTsrq7cSW5zNI2C`&`0#^O-8vJiMOtaNV zybZ-OxRsmR5`Y7nF$Ifo{cf_Q3TB?Cd}*(^1yrf61c5iIeH5gQSMS>SV6j0Eg1kqV z1v$t9O2%2rfAOAH%5Ccy_GBYB0!>j173~Y|IaZ}3M8~fQcWgbP&Dr49ny?%w2{=~_ z_HZGusU0{U-PzAuD-B5i|i z+Pbv}fmT?;5l!KDFv-Ar0POE`v+>DcZ|mF ziBJz@X_W8JC!XUEaJ}&!n6+I6nP(I&Qc>WoOaJKN3o=yesU9S*+c}g*e8a3#N3A;f zW~8MhK9wLh#Wbw(E688VfC|L5W1JXIF5TE}oUc!^d=c-}#|soRez z6(v=1;1=_ng-3EsZUMdX4n+)QSfX)@@xqmBbVD)&UyQc+N6>k}HF2G`w9RF@UQLf4 z#&FjW{?DduEd_HorMX|?1SabmJpDsIuodkESNRQSUWeekACM%M(LY@$NydjOXPmO~ zl>4SH1JEIS2)JS~y`kNHmDac+3RA6zq7t8m55n37xerVMhCM<^;cs+cJ35d}(Y_65 zChl|ABm=feM`_yEjIC6h$UHNy&RmvuT0U2!k68bPfP=e*50IP^t%3Od_;Z#fhHUzv zt?s(0~^`}Vo;m+?~4+owMW{g2a1bu z?z(bQ)2r1)EHkB{!L^HQDx+mED%O31wNY{`7}epdYwbs;lF}dQFTKyq%`R?Oe89&vuO@L z$>rcvkCY93y!lMU1(T+1vQ#&nf`@suI`4!UE!`&Hkl2}WYT_`T{?-(pk)hDu$&BWr3eTH%%bH^ zzW`KD*7!Z!>?wWVoVQd0<%#Y@8NXQyG3TI#BG+zk%hP^$lnwofLcz#L6k1m+~Yg#yysdW#q$c$&RN# zOs?!k5S+8Kmr0KwmhHBrO#XTh2Ec9^&xHw~9x?rOgH%HxMs?6y>N z3FHzY>z%!O3{R8Rix{IGu&Jhc;ndhhw3?DC;oguh|E#Ld(s(OlK)wH?H`7#}iVok| zTI@xr9Vdw6`OJ2coMpVlZF%>zQ)$R%mdL`~&Cr*B>y_&<+r{5^MePulS>W5ZQO?0x z-)4D3zm#GRU-mm;QzY)XUga_}Db~+y91f;NR`%LPpDL|w^!{W_PR&N`h_Cwcm?WiF zToc`?)>=oXa8$?7ar5|i;v*Upd$Q~n!JtoaSf-Xkx_AodEwVemGHXc;4Ihde^SC;5 z*|`kkiywa`4vF<;bpMTNm(!Eh5#WTU4!5D5kF^j)Zzm)-E4WhUJj|i-IJ7ZfKFby%1t|?`EI_uDJ;K#N#d35=z;2MGqH=_j zA`AwJfcGksZIo+G7}{-*1Xavr&p^!Go-8Brt^yUuuBuh~Z{^FrC%(wPZ*wsuEgv&* z1guU%^bdof^-GKk540OBH0I39>xc&I9P@+AlNzL)1s z{Cc_Q`Mdtm+Gl&+YKk9bn#Bl>apl^M)!M&ed2-{&yDQtc&et;C&eCRcf>4NkhRuky z6r?4LZuzwd-Nws(q7bL$mcKmh}E?ez-Oh>nEG!nmoV;O zu50PQ4N6GM#g3&&`f+C#A+6k9(t21@I?JEL*Z{)5+)RmPcJ!6wQol-#PWr@Us~4HQ zDh=TxKd!~Racs@VHmQQifnzSOvA)?(vX;ZNRj!-CdFBfi>7A*;e!nx+*Ixrt zN^6X0?qkQb->TEw_aroG9tx+5>ZlJpN|sjT=&4+Jmna<;q-d1W1z(wQPYwlO_uTK? zC>^}vb=$Pq{jm8Ds$Q0(zB~g0Immb+=jcEA>rzYDxagE=JC*YiLzEj>ba>%!-VnM& zP=97vnqV75W8XEyx?n(T%hnMe-Wq5X+b#u=j?y|gzBJP-ax#Qm(2Z}sf5=@2{p203 zGKgC@&-w`>=C}VTFF>XAzE1(2lKR<=OcD(hW8R1aCaB5w!Sz>m>N0i!Kn}gI5Eos1 zs5s2UQn7%`&wps}v{OZjGD&s_1e;~58PI%ulzPNsGYo|^60A5BtZG=f4oQ0lRSD?@ zBky#%!EbxjvUu-8!7ugmH=ypA>jpsviE!wS0u-!Vs`*}(7>1!8(odK{t#Wpv!IyA)dKD`1 zLe9lg6OLWfP=oLYy~$0VC(SK+;Os4hFbh#uK+^k=7D4fd2Q6n7lU<-)BR3ks#fWF@ z@9$hZtHQX5jY`)xbvDrc9AbSkn&kU**z*=;Aq^poQvsYdqG8@S6biVr}3Od`tbF;`v9A0;t;u(w~{$kO>^<)Ixs zVah-I|AsSHEiMX*JaJkV3GH2rBdiZg+X80xcHq>zl&CLFdPn}OgpPTBwbEVMQZ;ah zD{OSz{SI9Nt?F^j4W6@$opmI_)07W}1cbvZ@$n*mMSKUU#Mi=~VhSU`uxJwPYWKGZ zdb2}O8JKBMhrB7Mc4KoZchAhZ>By0VD8|%u2+MBx6Y}#d(z!TaoH02!F2{3`RD60L z^L&lCd#v`n;TBRRlTp)JSlyhMF=bEsUf`r(8UwVLo2Z;vf8KK-{u?ZiswesyLgk;I z{n|=SOK5P+av>wVJx=LZ@ig#0#;>3chMmvy5>mbn@e2FJ`ihRcBP5fba`#Cm-!w>^ zSz1-!U(3%=n5Aau`mXEmUw1pRi^{ZVLH>i*wbWT)QW26BVYJM5hcOP#wX|gj(Mx{5 z9VJv^`c#_^n+V3lSJ6T-#}CTy{<0Y)GVe@k*iO<-nuGpC`IeoY^~uMJhfj`~mxT3V zEmj#;ys!&2>gw=tQ0l%yY4q?^)6YwK{shrw`Y{~lLi0T`oiAPL)?!tt|k?pf3um7KgJE_%@%h#ykG zhz;xEYz_U^)=S>jGxn(_snBqgy$@wQCEayDIZDTAlq5M;EnwURjEXARJ9l&on`1%UAdaq2^4qLm3;bcwcyR zu~-ETl8b|3OpYP6xqt*S!%cC^4hr%{voea0)8b-%of~Ze{hn!McE9D;taxB6Pc`B_ z6Q@V?$oQVR#b|xL1cRWBZ5(#~Yp%^(JL5oJjp8%9JYsdINJ8j>6r!K|TIf{_Hr^2e zgJHco=G+Fcavmt9$X6%zk&`#q<~*PqloOna-8rbb#)?&LIj!NBXMCo@a@r-P8lAN3j)uktWWZ!;3|$=(x|`~?H=zyS5|XDjne6?vFFQ2;mEwV~F} zOx{XH+S+By2Dac+C1oE zKA#OF(8(-IUjL%pdW`Q8DtS$e2_lVrSs*w|kuMRWiG0QWC~pOxj3Yg1kP3oquLmg7 zp#U6~N5Nr>6?u;9Kf}Y!2e{DWm0acuSAwPDkS&$S)XBK6OH-(i9a$KeE1QA4pBBK4DcBT_Xr91Axng04!d?%_z6G)<*>q3^b;1$89VUEN{~q)u_tIa(-O~Khj6--6hJi06tbvufX+M(Zg@BM} zY^zk!nTmQ`_$wKb{sI(wANmP?iZWedlCRt&EQdQ*PCZ{@$ z7&wr|TW+>{&&OfmLkD z$T9D3x#S95en>pDG?Y^BTv2g97bdOHYS6LW98jc$v(?$PQb-g8NN$n8%`NQo(%D+>!yUM>C6dc}$#d zEebDkn9;d33mu~oLUGQUwYLe$6{P4@6LT6E`x&IiUFcFaQS#Na2-0|uI2t^!^SV)N@--dQ((9ij{xfCyAjP^Mora z7Y;oh@-K7Zu_|ywqgwkoEkn@zy+BDGhxsu;1F2Mfie!r?ghsO`;~0-^bLt?yL1ZPhhW z%}a(Dg&K1}385H$sgWM3ZXZVWvKJ71hqXpnwhP8EA5+_SoGB8>@3YgQ6Qr^*hCf|I zk9{(tmRvVeO&|o6ZgRuiP^Hk4hwLnSp%;R1^v4-Gk{4vOa;MtDLPx7c-lQs*a50{@ z`>c|nIj@t}Ask@>1I}RLIB$nA-itXNWTB)ai^wMcqFs)E6^OGhBBTm0BIpZk+K{~~ z)9;0EQdGOYc(=%}x`*?EsxTqkB;H1O4Ojsm}K-So`Ll4A#E-W@0F> z?60l_N?a6l{W=mW3bM)@=>}^31kbfKzZFl)^w(o=*mrriz{fogi7sHSJ~xq!@1(-; z5EGK=(R)Z!qzVBh=@`YRoV(@Z5fAYNOW}!%HoYv1HiN#{P^CU#TcKf2Hk)Hl-sinB zU9sCS7WqOy$FD$`iWRbbYh^m2?KGsj_>8vb)$!LC2E6fFJh-zJFMhPqtCz@vl=HjH%N%kNYu5Y; zTN3MY9YU$8?hkcH`gNg;Zbv~F$e*zweAWqOn&3R)Oh+*el!Q|k)cTzYVM=DTPZgut z(jd2V=$t!6x?t9xKx_364EXz?^5T%t&#*GC-UQ#zf751di%nSyGRyMSC$)=Itc2b% z(gct4*oH>pzAxJ7DT8aeN1`UjUGRioFX@TFgPINl_zBHS$f!YlitZ@X#X&cYb1;CW z&;$%mF~lNp8IBRFX2^xp7EIjt%8}jt?-0`%UHK)}8oxyNt*{PuxA{7kVj{MAJ6)kk zuWX%9X+Lc)9xSkYN0EM6?Ww=Xc%QS7jLv(=B6Sp@mDUzEsj<HO&gC`un;cSp&TeA1;}0_luA*-`a0dizOfL_Av3;pf3a5)x}jsC zShf+pVCkycZ`g}Bg8Ru{d73>M{-xeol~T4rjfL=@hHTf=PN`k>Sb~C@W1Ue3-bQnQ zzC(-*Q7*RYu-;m+Bmk9~{P6Qnc>Gjiw5vc{LZYxxBYHy8=n|?kN9rB2woMn@s<2Jl zG|0LU;rm9zSLs6${*$=L{)*)ogY}Szcq!IPa8U&C^BB-SNdcabV+C#7xz(_28atf&KRs~`%meWdDrs}KgYSceW-#fy_1DY z&&OqnjH}_HJY>UX98w44lLf;>Lgb!mT|Lz4HN{8saVkR$qDE$%tA&mYwcAkv{QIFG z9b!Ynvq{9y0t~Yuc=e&v^z;4X7s9kf^4`4K%Tdy?oBK6m#^oh0jYh_ITj9VV(6DnO zc~FyG$y2>el+-jvsC*9en|^;PGSxy6Wbs4=KVf#$RoSCZwQSjGtVs4cKUH63=(Q-V z@K$hrUWAi?LINq_-140riBbYt7do2v>s)^Y({U$=?$fJ9ehkxRzb?Cc1(p8eWG`0X zuqnC-YjJOy-Y1O3;49f35#_Uxy3m^JBv-i{;7iH&8srwmbs`x4xv2s?2K&=0xs=BGl1j@eh-QgKm;`;rh1m)HQ>&A2@r;YLB6T1w?Px z;ms*iDP^;ZOj~m47$x<0>|!qe4r;H4Kuw&%Li-LXb`MbL2LHqE((04pUQoByVDq4N z>2H_`H6gy;Xes%Q%_+Ho-f(w2&?^S3h5Zp;2|iC2$`@Z*omzvDbh+Cr>($|LLv|}{ zg!V%67xuaFb;q=_;e9Rw+D6Qfi7yV_N;>I}A7r|mjWstmL9QEYEY!}|@dq*N42HL# ziPt6g8|$$Y&t5`qTAGZGhxGFbVxdFwN>{O!7RkuSs*cYw9({2K7xEBPj{Sr(H2b_b z8}lac_4l34)2$>SnVN$~#g~;X*aXYQK>$t8MEu+&0O>P@Wt7QfE0ydTm zkT%UAP%&3LQI3Xad>D};MSdj_k&~hjp)g$D;2HZ)!+z~)yOM4s&Q41LtuGoSYH+lf z=~Th-O^9-5Yjsx;lJ&<6rvcV3r5bXH}|*ApaBQGS}EQ~;QON91C+j`P%QwLgN&L7()@PP|rBNnq|k zu^0ye>ZWQ;`CQZiEQ~+R65;hZNGgrYUC9_RE^4!b?r>X~?tm%HPMP_9G0SDoNC z=G7u#JreI!cV?WT&kX~X4aWL;GRFyXg>L!&C^LO2AYPqvJ(YlcwYrcAh5QaH_KV6n z-u)GTyg)&6UTo-^k6_Xw0ozr3lEBZ~V-B&y5hj+nL|t6*5m^#B$qotd&!a{=wcGIs z)v-PCP*4ki$>}#O7Q9>#9YoAy@jeMc#OE_?ws2kIwwH399+xbOtdE3SYFCY%5`fBD zJ?dT~MC}bgkA;*Rar>;R91@Kz9@t=rds+NDl{bl!Ig69mp~&!{-g(l9Jv{V76UK@`Ggy?S__#KF|5%y^lt4m z5JFS?`wnw}On%~YkV(NV;$fT}-EZNnNn&UC1*){U$j!pzNp(R|G3o%Fmc-^eVx&{U zq#%b$v`pMi*I{yb)eva-vx63|aBbg`@gx#TtoSPrm@4$^1BDD`=A8haX)#X^#gzc_tp{j!V7VdsVlDOT-X4hZoQ^4Qwcv;W-OrRM&i>e-(GA^m8(ltMRUw zr_FOxsoPOC30wym?c=G4`D_d!kTnk zDnaQe=4(=^2Bt&6ZK90fQ1G42!@aqn`hb>y2wD`HumaCb!ij^#R6!t7);s`$!&d~# z26KE5?79IjN#MH0fUMek>zl_*1|LyK5mZ`L#Qz!I`zXx2R)@P%=3UuxcoXV z5Rzhx48f{qykOJ9$yfcUv6;1M)Nz-r$=6vAb&9vLC})A2OJ^KlC$qke4Yjmw}5 z1N~vf@$~B_@K%B{{V&s$A;+pUcXMo;550?sTrwH4WN?3Ze#tk>G%J>8{w^4TWJ@x& zs%M%)6CBcd`~`(K2u+LO=5>eJS)hN>gt>CDwi(rpow09}*l!o#_hay;+2aSw-PU1G ztSzb5FYO56YXvkuGRdQqLMWvh)7$5AKr(&L@Ju4FpZnB=J~9}mAI8sK^GUlT2peCp zDcMqWqJ*{64C#V{#@XDYhyObYxt0Dka;AlTS~Znvd_2O^R}oqExVXHyXYtE({kMc9 z6H+6azzH|R#f}D1J;Dd*&8a1(m*6E$G&RWL!eq7ssI!-)bqKnuvd;>D{a}!U&@|cd z?>}>$cDBKXgv3+*2dz6kU2AaH+(eo)K;RuA#E|U*Btm z=`pPb*wFG}H9qDP8E`FbNPw)GM0z&zrnR?C!h}_LoE%?~bdi6ufp%&(q^WN|inYEO ze7#B(Ieas%sV_yQv+3x0$$_%}Vtq~rr6>q;5x@8jigY)w+1hA*saSujbT=;BO0a#v zS%0Iu>^12ADBjAjdB9rtYjPL8M^r>IF&+NKPxjZB{3ptmXcN*2l1RsK2Yv;czjByN z1m(Wc64{3j_yWU3q(32v^&RcuE#R3n3jNY2ieIAceLt}^ajwB71U_Vb$g$# znEQwAhMqAfmZaia_h0@xb^0h{%e@#*2`m2|fU&zar9bX%LXu+I@Hh5bQA)O?_bN=X zWd=Ry!Y8NlQ`*u>YHAB@33+SMpcBIB2>Y`H(1MaZ(w0?{ZQ<9x8Z6gq3m!)l+SD-{ z73ZVOSTma%sqL&ScG`{1Gy3(!h)ao$zdK=q`&xKc+AZQLm*7%1dYs37EjtFCwwAjiJT(eE#1HpO^Kq4P7n)v|%;}WR zGgd4clLUe-vsJ}Utjvqd&IA=(f;a%UKOUs1ymyNO9SM1LSf*bQtW?o`?%8E_1lR%a zr-D>wU3+ta?&v-Vx$?;8sj%Cs4T*e;j3p2x3=TRIO=;&c>hQxbLpr^oRJm-rNwVC} z+J_yo+=L^C!1<;YLc@m9^i;vYrz?ioC%{Q?K>V`S7Wvp*gDc{xd5L+XF8_*Ps#d)h zKbanTQ`nz!tE=Y|2GY97M_iIu6hEY6$+}<))o4ixV4-^wh3bMPZmycM0}q#nze71E zR6MnzaWb>Fq#l2Y9zQhEb84wb11QyI++$j91YR2hON53)ONktc^(jxtQhb6zi!~FQ zB|*~JWvGcfcw;Vpg-c)_Lo3bM&p%xVjIea^eehQCuYK_hZH7>XUM%^_I5t5v3s0}{ z7U@+<4hREyJA{w7!g=Ak!Wy_k1F-_E-1G2<*yFH8qb7uSxG`O%X4hvIPwQ`!g zTL1nQG_wRG*hoFh>qw&w+&JFIA7t?jdngie;SPX!D}j|llU&9i)~OLhG(}y;EK7LB zN@(lR+hk{5CrU9%M^h$>nF=K6d?lM_H{XpFt}3Na7Jh`6FVrfEbb%QfE#{cSl#pr8 zf64iqYckKeI-#3>Cp0%Nb)}7Xpk1h8F48^o1E=x7pKm&T%DY3<;9`qRN>jd|<7eJM zx!BrVIT{Sd!+)WU7d{@68aWKJYM>xg%V1R94du7rFtEo`X|un@#8fBbP2b1gw=b@S zm|1kH3_Q8%Z=+bRZj|fj4GARSwXO{4#AKnwS@{tae%_w+sXd5tcn!1Yi| ziWuY^pW`o(ig~0f8+{M|@ZjTh8e@pfu`K)V0ZQGu%gej(wgrKbYTc!_{d+ zJLjZIcBZW^3jEsNsmOxW0Vt@ln{`TwkDt~X1OCp8GAenzMq~To}e7#y` zZbvrv37h!Kn@hLNbU`L}IgYZQrdbZOU3b+8_-wP7MVKGe`I-{mAR5e8zVDOO8-CZ8 z<-bXrc88K2sNhj+8N|E$7AC88^`ORT2SrhS452n~40Zo5fb$W2^Y&-=rSM|l-5$m{z~OV?Ds5x+LCk8DPEZ5rA`sH=4v2tr5* z4pYxGc}FsYM(l3(7Gq2Uy4|FH)*pC@N*s=2n`b|TegjjeKr@m!96fOlD0p2F;2FU= zv^yqh7fyyL5o(jESYSgNmJ+56<^+7FpE9Vq*eXlxY3T);qGzTtc01MtB!mS+BQ2_1 z=oc;_MBiBX&SBx&FF+q8O#5i#6_B0?^qj!xG5y^z;#0n=zkIR;TLI9iLiQ@>yFXX< zdW()i-3aH6u?GT-BUTmJ&gZr@W-?m-yf`k{(%${rgR@Rc2UzZs42QbV5(kC5#D%`s zM7>ctg?z)&cY62u^Zv=}quQwe8eI2;oGy29@#vmr*T9|RhiV$SRJWb}mJ#f+w$nSM z?~7wgOs`#Rw2Ht4XjB=QOO#}BN;ARX3)h=gy)9)qnYbEq!SP5s6^9|jCcemQm-pt1gES${g!P+?Tvz8KJ<}ZlXnYEo^e+zlry_rzX*;s6mY2ST zYiP*aP*icfX~Xx42XB9=+(P49Y`8xcO8uj=?p5LJurf34xUfZgR|igpmR|JZ<95Vzvtrk8lAhrT=zwIru4Ty%BFGcc{a&d}55k9r;WwiFUZqVw!iP6UuC2nSH^a<# zq)-TWCT8x%cUMJuihb>v90L=rW>*(shYp9rAM)IP~dU9 zD#JDgQ?Df;!u5++ewCVWzLmGuuh4c0+YR(-Q3_90{4$sqw7b;-jIKld5~?(D?ZKJzV`=;b;QS_}5s>@@^_v1XYcOK~^U+s?} zF=0ucQFFMKNTF0jQCSe4_ka&(7v6Mk9ao9Z{Lio~4_$AgkeU4$Sw9HdmE&dFrsm7K)jelg6z3H5 zH}=Zf{NXBp6Jcn!k+gwwgm0W(31d0F{jhrt{``j1*hBLR2JKLnCKZ_btJqwaxl(}u zi=@u)I378vF1L(-#F3{I^-|ICvQtFNi{BH;#VWUEZB+-c=R^b=g`dS8X<%DC9CJ;x zQQT`b+Uh;MmC~$NWxq^m%Cg>h0+Du7BD3Q&8^ekQi(@RZ%QtQ8wzf807g* z#MJ+(Hc&lMtATASIx=&H)|#P#HKXa3xOSg_$=+{{>*aG^umNG)0Bgm*cT-KlnIQyh zKpcuBl3h90up;PVB%WN0%WMMb<*kpIfBg3DovGeFrIOQWVNc(2A`2zA<&n{#xRZY2 zhvpzLYBnVVR^3zOLGaNK8V_08obMW=f6`G&a@^67_FT+ODL0uI^GOJO=>m2$?MTUu zx>@BSHYPA0m^zB!M%yN6OTnQGbW`Z;L)R=ctg^_qwbB{Tk?_}%K(YswLbeRK;Paw^ zAXfp5cNXR~WT9MAmQ_W-<<7zI8WBnT7zrc*3$3? z2O(qX?WZz+uD5Z3Xm7`0%{CF6g;>O}Zst0K6HPOV;hXyz0oPyUdXlf=K+)0=L*s6Z zbr~ZYun@mErJB5`?8JT4KRD^->x=O@e$;!<%z4>6RL*Ta!EQOQ_$3LbwxhKLw{Hwl zSu^!o1#dQ>Ywzrt1p+k_T0F=HC%BdCcdLpHM6}l^RcZT)?=t+EAYj5J-h^@B@lP`B zy>!>dTzLF4+1Y;cH%yYq3eh;iwCqWSxAeuM>qx$OdnKGjS#H~;_7I|=$Dh4 zCgWnO_G+M9BEv)%$26Hg5-bV7tsxyCLJ-T_3mKA|6I+5RN6zfYn2dDqIB?$e82}JfNZAR4o4<=J<`AI4^l@s}q`VvtMz* z8ISQg|FG-Mp-5!Y`sGQm9Kr`r?M{7Wu;djXVZ0isV5ra#>G2gMM`s=I_4JXo{MsZP z2GLhZIZS?6CJkBwJAar#6t-_LDLXqAAK(2ED%62UhqqOwkFF7gaeny^U@f5gjM99B73dGuY zZSzXg?2pxShb+9_o9+eHpZ#$j7R~q${EuCS-l{nn4pO|cl*K7Z#6K4dmF`o|2dU+9 zVq*JCim*M4^E+yTcWBqCCXwAkaUiv--lUM<#%XDBF#gEz6Kv^A zvp?=Tfyct0%1He%^ifzD?#(np#}@9(Ie$?195hVRE-Q~?X4a2~X3P8eh28Pza0f@c z^_eyb5#4cVU}-Y3A}wI0zdO5@uXKsMA$p_eciUhb@gFyg5qZ}ar}`#zhSs}L|H=6I zEo5o!n*g!tiC?hcwQ4utOl*%F-k3IeB?kf!7N@^6i!Uo+qBA?jIk>J$04)>K$(R=S zRY=C^96CBwR%kS07}FS`gwQ%@MiXSu`dQ6>DCIYm#&U=&v#o$Uz{gv?#Hxaote*Gv z7RiRv*4&A4DCNiyBdng`k1u#kZLQoyuedaaGE&%|-#c>a09{iD(bRRHuI;C z?gmP61VV6!?br<`FXsFh3Ei#`{3Dq*JF3s`1FvpT$wHdXW1jS70d|24#`_$d)7vi6 zK^&2YOVolW?WFG}Ic%Azfq+nb+PfnKXX^VMbBIbMXcZR2ng^DnctKPjL8lrt=4d%q zbOK(2%q9t!-6!m>p^`UsXp3^eh|k8G1_Sbo?(s56@ajbK2l^JC$1G@*v}`Kx69yKD z(@$qKn>W945#WPq;B0{mwbk25SvC_xT|n;XV4K5_09;sDyrq;T)pIh-F0r176=Q?v z9vnG9A4D zE!Oq0fth&lkq1mEafRR)&G5!w65env?@lH62~jJe4yjih^jq+pg(u?YWExB+1Xv&a(SEUjPD7DU!avFiJktMF zlL8L%UmXAK#I||kxm7)H5U{NKueU8=-OpoT-$tVMJ}v!kPOx}6|39gf=dsVWZ{~{V zQZE-BU?Q_ETEqGuE%VYCICq|#pH|L?1{N*4hX)ZOx4t`-A2(S=B>c;5KC)lxis(#B zvsN~D%lqB}3(9}<+5+wbS>n9Yshv{EpOcw`Jpe~n{!g|yv8JY&e*<9iX0AF}!}}Bn z!DAIV_g@!^?-Oi$J{bYo{%bwlv3j$XLr9zE(G12{sd*@T6#72{421Y}*nzmW$U zCaMciTmqMh_)pf*db+>v@S5&OFSTK_w7i28N=8+7KNcAR*f9wU`&tc7YB|?P)xV~> zR>bPXr$kpVQ-L?ImCF@Z^w1K7pP0mea8zR8ts6nBT-(}KV(FuPYJgs(|x~7MXl?0 zeH)MZ;4L*g-yE5OQ2r7{?x5|GtbZgU^rCCXAQuK17`K?=+jmgLy<3gH+rqM0-OG%; z^l9Fx>acmJte3w>RLwjHcIYhVqPQhVAy26f#p-kslC0nEUQoLh6 zW8!2lE@?FK3S47u#bVKz*d*It49Uhaa5@~cV1cF8BP{F64IHdc0#!#y%}15Ln?E*VDYoZ^J3PpmW}2hC z$^}*+x;rW}GMM`X=ebKii`8OY%=M?J2XZ%E4}HSkuaJu?pxndgaR-C&JNA&VimS#n zgF+p7sNHQyT61iHH<}K1LD_^1spjTtP_Qr=a4Iw#Vk%g0!b;&M))GY(a@r>hYXXoL z2);B94!+UhZD!G%^;ks`3&kh)|X~w|3h^(1L_b2lxz;T2dsrAmfeHt zHYNZ9O7sWZ6WlfWxIjd5usgZ5fw{E618&FzZUV}RC@5z^wb>3)9s@jz3L+fmYzsXE z6V9Rs^1{X%U9t>$l#txN2Ilnfk-4<>JiBn@6?s;1pg0F1ur$Ls1CnrmGjWEz71#ji z{;k7AgQ`LlMut2wFH_X4Uan# zo*A@Inq|_7#Dyl&qTa1b$_Pf&VqTbIaGnGK1AMuoa9>`j(1t2*P7=B81viFSg`AkT z?jHz8X*H`GZibzzjspRa}PB@qf}Aq%c_9K0IPlxTE^D&+br?nrN{I@rZ%{}%=F>93Fz zBvYPQLkv<$ffG?e@XhAFv6cYy!xmI0&$C+pLF_NHLZG*m1Av8Hg9HCJ*Ulm&3cYur z3Ko)x67J3-Our(I33vkB`HE>$tqi!aUg)nt)PnktR>43>sfK1$8Br|e8*9wda|%gy zcCekDMGLT_Fx$krV*gqZCBjE;ivJ3N{h{#91^_s%cQlklLRGM!3TRRP(?Kv;@z2|T z?>=(>5Ak2uOZkGJ{$D&hi={y-z!~8Guf1A;??16l07OO%LNzqN zpX9+Fp7k$uJb_5D=iZR=$* z)s%h>H!A4hY;FXb;HAzgYJespI*gWq;HMcS8r0zIIRqO_glY#@FeB8r>T1>F@?=GH zh~>0($lz=gjHOWWO1Y3iT9|!BLKq%#e!jHQn@vp#z>)>nZem~_SgC~S*gpbB9JQ*1 z1wx!eQdm+ID;Vt6x%FSjH5woc4+Vk^=YRYDOJG(O#rra47m2F^_bC43fjUJcA})+p zC#VnxhQ}K4|Ey;JvkDepAg@G(NOS+!h^wajM}`rO#IP@(I9J{V0fz;Pt|FNb7}gCs zSqddmRA33L9EA8@@*N$Z<)Q~ujP%XsM^}xO1oWlUlAutLnGJxaNL*73xsb^XYEx5f zfCy4eO_+w0!3Kd>^>?S8(BRL;kb{xYOCxjZIlL`{AXT1A-#Zm%Lr{L$JB;ng)c zc@>k7s643P?|!h|rp;z>ibj0H(V_TMVC=dQrLBbkBNb=-&s{-g#$9dVy0pc`0s z!N8L_$SOzAvyYrzcG1I zlqe+lYGqWz-!>T{cwL;=VxRz;Q4%Cm%i3bSt-R2B1GG`xka$c?&m{5CM+!9xB}r8V zHRu#~>x7LRWA^be9IDLkH9%rE$oLGZN_$+^? z&k;nbWGhmi3`0n`n;#kgG&m8Ewx21)pyR! zkXd9K&3ASBQnZy|sk@wm@dU5C9P@KgdYcpVkL9X1Ssyu+k6#j!?MO*=+llWeB9&;` z{TntKbQ!b$#8l1@K=hdL$Kd|~SwN=0TYTjQ4=Mck;>J_)SNY-t-vKTprTiuLkqT?l zWtm4`_o9=BN5(sR|j|75d#>@QeIOhr7Svmk9m- z5CP+wxzN(@l<@6eA}Hh3WN=;hAav2k6#SMjT356Ryk`9d2`k}KzA+>`>CcLxO8Qx} zsuNjj@Y9N5$~TsZ*O^eiHlb2dcVWlgsYB3?XqBHTA%c%23qIt7%7YL1&<*ACzbHQF zJRDZT%kY!Z?38>`VUVz}vpaSC`-|fbh^u3Tadw4qUbDb+=W0Kl23@3%ul(R)xJGJP z6bOd_Lr?hm61CJP9Xch1#6s50k5CO+Jmo2!ty^r(qVI=1B8CJ8rD9^C2Rw)3z`Pd< z?;?mwAS7ej)UIbV)Au?sO38?q_+L2%#qJ3H~~V?UNZt&#$>1>#N1owS}#@mLhpG_ezsjQz3d&jO9R5G-yvZ`{!CZv<~8zjH@g9oIG zVv+o$vM36C{Q)l!RQcmK3to_NgLTYqRS57_CFZASe`{W5@;joK96s}|Z2oG4qQ>Ov z{DwKXMGI%n8jEewf0DJ-nRADf2E8CmmGo`{Ox;MgVS?M8s_H~oK#bDV(Ns`W=@}#8%tdl?8J`2 zxU6nOW4B6cnqA|vh12KG&rP2{mSmOD)*&0kY14IcIBj}Hi8nLmqIOT4o|!Ry&h+^+ z#e{1#0*{8-x!UT?oVg7PX6sOmGZ%_2SSu14WDAxe5yLEk+mTY^v5=q=yG)%=g83Et zKo@;S?0tOOwJkahh$+@@vg$+&Vh=V{<(OdTKkzTY;5b~0W*9TCy9(+2Q7uQ+kv5zM z48L7KGMkL$t#z*$*8YMhL|y6(B2JXZNjf05SgI1!)otCjMUTUnqAaKNmZ$8ihQ_Or zs>%)I$c7{_fn|Mpk7yUl=&=^$^`%Holn-SL56;*Yy&MT>V>BrC+7$^z$52LpZgB<4 zx1c_xpQ&K*?656pEJc|h#Iah@!LC6t2?cG3Mb}tCAq;!4|L{gcZ4pu=g%|y1HpDt%}07hXV!cR@*}L*4hre(28DiK{TjkA1|L1_eIA2 zw74%d?kB{3g}7xrpgG;48e`oa!^9K9L>Hr%>}bH)VSTKhGcqb*#qBUX*t*}g&=D~} z*cc{o+?(2<^YFPAgHf6{&=PZCc;}mj{BDqkDXAZ}=534q77f&fCD2B{Vyen1(>h4n zv&;3l7KXA`GgsYQz`wZbLVQ&NAwv@*2TVWlTUX0+-8=`m=$cxhu z1Vt2ZJNuFX4j+eK`e2{D?U)=QkC*Q^NnDR5jPI+gX4zNP0%@r$xfcZy4*g;ql5#Y; zxawFtnq6!bX52O6K}O|tD1}aW%Ovp!SonZF2T#hi1az zHI(?=Zv~j7iaXcutpMht%2+x8#^FC$xOPR+MRnVj1o{TNiaLpO8r^@@Ao&Nqty3kT zU!lGq5KI$02EX!-L>>x>d<{X1h-@`P{tw=a$jSDQ$SDX~L}Z5{@&&vZkw{-i4Z$!^xPXJ?oUfBN-YOhQ-+t?R+;aG#e^!+GC zDA5qm-0pBQ%RK8ah_(^imV4cq-iq`xt!$I)zGWqicS)9X7!Nv~AP(qB)q%u=J8_ z0-N%=3=ZvaD-OSrHP8S`iHceOy37e=r^ogH~E@#i5HsIm+zMWHuqMrmA? zLVXk>qn;HY`a&Gc9bfhyFgJZ#?mGvu^c?#d*{pcSez(ocx$h>u#<}xn&Y3ZG3S)`! zPXdB8+%TXh$9&1pKr$haG?p9>q%lO>9zz>3Ichu@IOLxT-3-jRYF{TTm^mXd)D1&u z?tn|8#c+vLP(tAyMItlj;sKb-w4B_+6^%sZ%w4F6;}x#wnZIn_LJ7wOxUakb_c=vy zUwI^gi-QmocMrL~JXR>*qU(f0pTea;@X^-^A*O!Rz5`;`oOEJo%%y_9A=PLk61FQi zD>2mJVH_R`(3El(K~v)fz)Z1-L8anDpi+kzG}BRpOEIl+k;sCXQhtXjurpDirvav| zfPuIJ_PGH7%*u z#L%Re;i4BYF=xJQ>5V2U!j~4ci*-&EE9`qgP)f7s^!U6@wCkR;%^uHU7gH8M&nlZ^ zs~|11hgUyBabj)sca*d}p1v<=+>?4Yvep;24!)reHP4mF8Ss;b<8LUCZZ)S23e`{rspr}{o`gGT69cQ8R}aAee#AzIlG*Kp|Pl1{mUt5 zbRYP#mBP-74*LjcOrZSlIr|8hR%Tim#SyG3rtUTueh=eX<^8SQViVQ@PMk)mDNt9W z{0yw0DMN5rT&N$w1|E!I*+!b7%BV$+Lu#J6$Y-&C#TohHw(UK!%Pw!D?O%w*b zEgB#wmP&~KjPxCqj&C0!*lsa>s6{-MS1?9VOz#DW+JRWTASyJ`=&E#&lvuxzS87d^ zv5%1TTM?_TfCzopO>?!+-fGqspe6H=(mjyBY5nt!lRls;)1sTvm6XWE`Pl2Z=fwGN(muxQi`2-q8Aq z(CR?#14O^D4eivz`J6t^Y-8ybkEm|ZGgU3>s0;|Y^b2%QnrlpwMm}WdK85DYtJ}fh zrZxaN51??^c}#|^-56s!F(#AMrO0=x&bxjUdYT?Da5HI!GpC0AW!;e`fx9IEbUY7B zZkiN;|2x{yBP^y$a84&-WL8~<{@bl)?=c6EmE*Bz9Z8G!!`9?_mAoE1WNEvkn=?#p z`z>FU{fd7x$xc9b$N5O?1J?(x%3#SKS~c6Ehg)1_KO1R4HG#vY6z#vD+=!ha~1ihi8wS@>k0t|ah3R( zi994%mlxg-;%al@-C(Y!9AgL%&DEAdfI(cH@^KS+NUnAk-j{N9i|mCu^ ztpnxsR!d&(mt1KDfQvHFuPT%H)YU3SLTFDkkioWtYYL)PhNr^H_P8#)oL+AM9@l|; z+^m3b(Ijeu_ixHfIsBty71+e!pmpW z%M5qahJ#b$-%ZrNs<$4u%jOe|h=rGT5eD%MOsiX6rXR|T>oNT;f_V+;(d*G%7;keS z-m8e$X*;xYISR!(k*s~T4RZtbINA$~7EtJ@D>>w*d69pxEJ)Zky%h~?hn{UgRHw%t z$86BCXzIolgD@n{JXmlK$H^_`f=TVUcZ zLos19T_VV@XOE+OEsBRTEyKf#PZ}Op*w~0`JM@s^p+*XV=^6?kCpol360pYaPTMwJ zW#!B!fO*FoA{`---zg#~P2@Tuf>vugw56EPFRUE3RP#C zf3C~wst#N(;nOq(lKM{7QkCe+X$Jf{5hqm(jFvs$r_;J|#edjZ=bdO+cnU%oMWNOd zI%qfM0lYg2xwRCR%5H~DSf%8W2`3p6&w~Vtz{E2Tih$3ncsng~!zQ&w$F1z(4PyCT zU69>jVexSL=K)s!xa{44l^AVLXq-8w3Fx%hW10;li7`)*V?qh_ww*EHV*jK`biu;; z)8@`uJSNT-0#zORlsSgXOO0Sh(pVxD zIJ;r~f{=;%hPC-6Nz8932_{8Q9PhM+a~r^DG!j`*xGi1v9r%|;R9-W=>C-~#DmQ)F z%z~^nZ|+=q!Gae`{y8GPXDph&NHW3&)U><^H7&wb*@#l*kx0YBxigWr6N#w$6HY23 zky$w{WmOhn%A&_1z|2BMP-HrVcHf71==)5E1~nY`?AHJ{pyqBXre>E@szd6u8sd~a zk zfj=;R(*h$aoZ=bU3PCN&L8B45<%Dici=bRgO{ zWawH6-G$H>L*3urX;;uw(O$ZuwT>Q(26Pqnt0@eV{sKNB&`Vk$N0z?j%IQqhh5(Xy z$a02LsC8Owq`?)RG9~nP6s{7*AUFa*!>Rz~k-{=X+L;4tfO#NV(R>3ePx-W=(@j-M zdc<}49@9KmuD=O)b7c!5pA~_gJ^Og38(RRsRj_q5-DmB?;zpJ*`sDwFDTe%uI+AdJ zehcK!Q|MJ|H;!qBwZE$n=`4uJD9m=2x<7zx8wA9%%jrzcp(d8q+2wSWG}=t3Z0YiN zS}l91)R&h5>@9a_9d|J;t|Lp|*`^fpkUr4_GcX!pSI`^F%%X0>|1zBK#HBF^D!su; z__VLtWslCX1E#l|(PT^;gD3di)9fPsPFq~t5x&sW*4qIQcIo{Up(q;AYJ{U*;yf>W z`;<7N(wGV@?^?T$#uypPMKr`-q(c<=^is21OKa?a-e`8|97zz&Fd7hWSo)rwqZgZ9 zrVcxxH^Dw~HaNRyf^kL&WVb`Pb|Cz~?9$6pQqsA8WCy?;?Q(j$S-MPg$V`qR%D}84 zN-cfg&e7isIl~^nH-sn2+iE+9v$o4!uJL_(TwXv?;q0KR?0_B<54y(AvE&3j#zC>O zM;y>{*Po2rx_sIy0d>q@*a1Bx9<%{5w=`qbj9ouHb{~6b_5^wUPCf)gcYWS8u(f4X zl^tDa_D)U9^p|hO2=tbXh3`-C%=D;@E5hnHuqPH?tDTY_Z^F`(J$G_mI+hwUJZz=` zHsaACIn{IB_C zY$#aDna)Da(E<*j1<>o*TUik>wpBvGSHaX(7I3;2AbG7_PVLPuve<7+z2{YI>PKXz zGqrQC9k7F~^5yUXk+EK1~@G9h>XJMb~AdUge!S!S5qtKzjwycde%tt$&^Q95rGM?6bM z)C07>p;nYFiWwhz)LBZP+45he@OQ}NPhIS|a$_w}f9v`vt}{weKN6sQb{C^$)nr*$ zCw%H@PjZH7X5h3p0lg5N5`WvIGfGC7_BNT>?n8Zc`Oty(A!Up$w5qHZHBhTM1cpo0 zr^%nMN&!PIie%K>p|!NRwVhlWPQ4%MHGUpT+P!|Q-~lStd|3dAMJu5;f60{KE-68J zsi~Xk2W8N+(AwrY%Ap@*b)li~-dR?-r7WPoHPzMFpL=S=#@bnAM~ZJGhrFXmcez{h%yBGnA~MPtykjdL4J{GOamyz8{R_=!Z?1 zv{36{9)0K8k+>f=%7}T;vFQ6{fsbxcl8@asx-dk)2KIOZ`_vcdr5SbfS+}RBN94hN z_GrFK_eKKxKtLC_xlD%yOlP=skMg>x&4vCI^Jjxgr?_-z7tODeEfXN>r5U0(>e-Q- zt{`ta8!*^up$o$^M)zqc_8RCrTj~7)P%@f@7U6M&DE-+Ejc+mEt z5(McitwU#kN1=r?eGJ<=%CT<1;4)Mf{1FG!s%zeLm@a8@!7j~pskIIabkG;waMqsmEi;_#=c>vLC{mOzbu`t{R7i!Fa*pZAWf*-& zzxcSD1NzIduv0Q!#ghBSm!UPtfYN5hlg->$6jlXA$XdGsqlv4=#9z`Cie@cc#sTeI zhJsgHC+KLjd^%9p&y!_$1*#vJfF7Wp)iJdVX_SQkylMJ92BNBc3Pxr@D#MQ4mS>#}^TtAGEmHZM5LyefG4%)pp~?E52}9pC>-I?v zI)x-1T~*nkZhfYB7cIh|;!tI8dnj4;6oLl0ArO~KQd>TNbcz1tvR942Sd3vU$yu%m zsI5+&S;?{sIE3YHTB4kfEi%horcVUwXb@fq$rh}QNp{a{<$ylk%k-@}R;qqqL_h9f z^kI~nXaEpZ0TNh5(=j>5otI)LyHlD(huE5<*RWq3*Ni*#-8|R^(jk{Qdc}D@3F}eF z$<`UW+fkgy(uw=P0D4>&S;R_{X5x}tDDtSP3t6aUl37Ansm-!Oaco1M@5)eI%bri; z`Zz@k5ziqf$F!^tYuq%`@;dg2f>VPUrgPBPz%p8}H1xMI2!a!ODH71R@*bJbu-7vn zz{`<96Bs%GHaB6(iw224R3I%|YnaY%!?~}agS-|A=<9h_S3q{MNA-13L4h;tV|2#; za7DasrfF2y0kGjU0=r~G%hOiPZedTfK^T#o6xcY|#Hu8Dsq;-lxFL8!$z_84YS zdunI@YVJ_oZ)Mm}=bf&jzfWC0j@iz6gge+d5JG2(Stf{E1(P?PJ&9>&q>DT3_pvpR zE^FJb!2!fx&#FSK2?xTiIcc_`@M%y$FqO#q5O3R32y(C9OM@09M{6b5ZTS#<%FfZ$ z%ITP@$_^;kFx$T!rna;2{yt3#J-`EgK+9d+Enl>Nkt1X>4O+p|pTlx}ng_Gn@N z^hSxNf>L2rkYsCcxiB`rZa_ zwVg$|lY`4Y#e0Q@Dq7i5Wbc@w?9nD{+6=Sw7&dOw)oppE?GflrGhDq2V*u8i)LIv2 zuV<0T-a52??f{s^RF(sti^V3(F%7hkR2Molc=@CUEgn@xXuH|^lGH{ZP2RCS`+E5*8L2vtjF}%NC0!wv_J_lg?5I0fpjL~hPgcp zCr{`=YaXX=X(e7^{0P@M^OhQxZqY1tvS%Ij=7AjAB6>R1tLsxWDeDUdR&$zHzQvi^ zF#3ch6CIbiZ{Q9A5v`-&q$g$WQ3at<9mc`xcB;Tat9m6sd^C>>qP?MCHDHCcvT9eB%rx<=&=q};zg}`k*byvC{@e;`c&HWWg_<4vsnkt2%9?eCIw8o4$1Llt7dN= zl&quoFosr-y+_v5^6q>s{ke53%TeYGOC#uvsz$!9)#UJHT|imPMW2KKVSTJ0!^3J` zwGD;xSruhVn~5SPa`krOk)kBdG*MRJv?%s$$czct@croLf?n46G`*W0`UIeMAnYzX z^6Ypu)_?1g?J)6H%VER~(s__Fz9LC8 zP3)s@pfk}#U~f%zOo#SED#c9y9ha8YF;;9B$Tpq75JDWQGoPORGH5nFkd=GgtPXR{Y-nU zfT=@xK^FAL_=XiYwM^^42ko-@wux9JC=$Bok@VK`{f%_49B}mLlsY;g$Nn^?=a;** z&C1cwT9n+ri9I@(a*Xq5=+8cT)0`dj#XM6jc)&1$vhGhaJvlh;sdOm?bX(hgx+@Z7 zY3K4frmG}TX#^4zB`C{bT8?5+XV$l4q32Mv%biSZb!b;gD#n~kw_uXHClYWq2TXm- zm8yZRp&U-s%XCkhn%%}~nL9V3q5qr$rbm|Np$YOHql>AFo?7nG`IMu-EzdKpARK$3 zRV^;Gx?>RI(rz238($YW8zKwYlMz)1BWl%-E^a>pQ29M zk6kfX8Bm*hYBdmG&$R+oKcdgl=;BS32#xbjcGMRZ*;jm z6twdWXhKl&XW2Pg2_~ejz?JSHdnDF5h()hbZ#(LdGxV0EKyRQu&>CrxT9P>v@cP@$ zdB|zf?xkV1!*_4w=$g#JpDCEqllN)HgfP*GiI}uc3-<# zNk#9lj(wA!Z+7X2c0kWIyL6%Y%XmPKNHMfEgL>?bY={_Qc*g0ZZ|C)ZVLB@cP<~|5 z+3#aK$)N16Z+5F?#Uy#V5}+UTHFwhw?7n(>rMZsQihc(DRF}#gRBb29iC&^0Y!UM= zGA-ioP^Cfeht1TD{%ZOavI)Wec00iOQYH+h-{?ZgaE9q-JLutB`W^hJ*Y4}7Rc#TB zZ)9&3ZEnufO?HkqREH~pofn%VZxp~)ZFpCGj-p*YT`$8d{n6@I@UdX&^poaM2!l_G zAq`04T^w+hUIG{dZfHeW0L?+2ZdGTO1?y39Om89xb(Zz5R|X|&I91=CV(T-z*`sHG z1)o=jiTA~(qSCyuEXN)_+mxptmIZKs$^D>jiEyZ~}4Kgu(u*;{5RUZFl z_0to=z>mRS=VHly1Ko-js3RpqWxt#hgLV$afV(j zGi#Dlt}?t_uXssuz0Y1f(+VUY%SfGae_4>_RqV22k@l7KW@}VQtfn`X)zNAQP3WiW zIrQqXI=#YLfobcCrct!JKcR`B-O-+9Ma=|_kX~Dsr#0AE38k5CAnSf=xc{}{p4Tus z7`*_9<{S>x8j*m?yd>zF&dWLu8fxmfop!^Q)^cA%O?~aOYEzEvtY**p*#5Qj#Io*m zwTQnmN?d|W(v!>bbP?xPv6I0T<*aJ3Tdpp163 zb!f*jRQ*Lx;HfEnI5nq!}ZXfrC6`_LU4Pj#bx?4+QXU4T%FGk4z=2Tj*%cnpjAr3pba4T zxZO{?gj`2?KpUYs-+?LSnr4@7z|?e}A$wMNKrgq{ArKb8KQ71Kb82#H{ThhL^d`t4 z&j)zefY76ogYd)Uu)+fqQ= zzbOxVAbK7)e?2T-SCj|nl1?wv|G{KVKUkI@Ty;;Q>X>|#yhdfHJeW*&u9Lo-SF6l` zn(8J~ERatEq6?-7rM zB;WurT~053*QHEeCI~Db ztkiC$T0j#Pp{d=6u)QtI)=^mWpe)tOZHFL60bJyHVwC@g$^VEZ3O%OD^}?9fxqIKT)F=n^}CNRC!Y zPslEs4AZ5$LDIRBDgw%+_lg5S?VKffrYG^p*0F#cI;3cDUWL6Hn(5ZmrpV6Ghq~z7 zE`)5V&$CB|Dlb}CgcC?xc1^{?0EcyyhHL1eot=IzsdS zMexN&TMhCO&>4B2T#G0`C*E=bSw(7v7QMU}i<&)vP7X(*Fuew!a85tFBB<{IvOsSx z#-w<5MSs1Rw`=eGrJUNw2nrGQj6A0?)0X8h=9OWzmjDix>2Sq`e2t`x7WB`U^ZXj~-oG5m4XaI;JZt0(w?nbue8h zzIk|pv_p?C?xywFee}>`)UxFH--TSC#L?6HwFuIJW=&5?QEX6zTe>kiV%QM56&B?j zD3&-7^_o;6xP<_fK{jy-2!SmHsK5_d%&;drv@kk2+nF{kM)%M&!UGa%pe5^V^2 zA$2Z>Pz9_)axCRy9}80wb~)Wxf$h%N{_HV*Uka`RFRS%S-=)sQb+oS{XCF)VEye_Y zQ$@hFs??a4cH76&R^zXZJnvhCZm!Er5$CxbPbTh=hT91e|dk{q19jd zX_M>H^Jv<@Kq#PnOY*eZ31~NF%nJk)DW(31P^SPm8J)EhBQ)h5q1!KL2&TTYJWFE$ zZ0N`q1N2*Xb%~nYXYXyl0e`rjGcpSl6xlFCt8uTygF4r7dRD7vpl!f99$SN*1V+PA+fPDY-A)p_8AK9pmcp2uJ@K8N)1^!^;v%siF_RN2e4 zD4)Urii(m6JKjsFfTp+<4aoZcbN4UsaaHC1IJ{>+D^GUPTM8+1X|P;8A_#5S$?TbQ ztM#bI<2gr9y>O1lp`Dpz0*$muNYc_vGEE3Iv%d?(!d+sYwGhF&-2<=3=Hf*G5MJC44 zNG&^b2W*?aY_rI#*`hm*0d^^4ucoE84Fwb(g~Q~#G4nG-yI^Tk4&n=2T$_yz=6g2T zZX>Z98Gr!#mQ0rJGoZ@|qXwpbk720SutPIl31xR{1{)t{HXY^C_hPtEMPorwj3tR# zkAG1TD8#}e<%VY!i*qf9VNbLlpdAw^P?q3+=VA3=SFK99jr3sJMkHR)i`sCh(ygx( z+HgMBeVpS%v;0h`wms(Op!r$%{XQ*3SuFUt7UhAm0OJ9XEIV{l#-`IVF;xz#LY5US zv)&Yfe*`0p0;rsg>8zT-Db12HRNY7U^+LSMaRrizyiV0#ZDDrGaROWSyVBjoZL z%+F!4G3l&2g8H!A;oD=$A(ws( z@Gml5o<&cBD*|Z?(%j*-%5(fpG956W$%H;MmNT)joso`()dc#RHo!UqEnVBACeW*) z0llEN=yR+GEzuO)eg`+*oV_ZH%u`b+QO6Vb;RykhBhmXNqwe?4R7@9dIsDIPH25v zTebLiQ5GRPK!(uJx^N`hNK3^Ax+{xyxWtBhTxOd2Ie*(+giYPCXwopuS|GNHbL~nP zM%ezLTD2`e?MAHt09Y%b0K#;7$4tS_xf^>D$RRn`$M}GxRhn>k&zIV6D^z_Ks$PTJ z><2)dvBN_M4uaFMA`ilkBkB;jZa>cE7s%X&`rb10ROyD%H7wr8u=z017zQB`N5{Id zpjN0lkK)gDW0x@x*-?*=r~$`b4hR1aab8Cx@O}b&40IH6jZW&0$ys84a|hqRvfdvM z0Y4o)fI3cxmjcsfNcLH13s8ypvL5|a<@8#YttQ|NnERV9dlDF;9Hx%{iC*i91<%L@ z9gCV^K_8Y_t;%UTs$n`y*UEHO2n%!c)YvArO7pSnhHGgG*e{3+=G&Meme4wL(S13f z6M*RcE+GyNPEiWeLAn6&`_M7^TUWNnV^2AUCaCrxX1mT!@0QAXFJw;VH^#Rpt68)ZFFi2bR>hV6(Np|P}H{o@Tst0gnqC=DD z8XA&$Asa}e47x?X>WZmJbia@O%dx|2zEDk~dw>&;8k+@JylNUdAVPpp=-gX@D`XtLl=e-(7Gq?RpoTHm+^Uk&U8!H2a1fPFL~ff zmD6cmpcJG}X#Ooq0dth=>c#6TghM?Z3uMQy23`EM($qROc?-EfxKL)dIs(yDG0oH@ z@09LmXZ5-WTdUdWzc#`W03gZEDt*J^_civaUJI)Jy@x%b0XP?7|MUK}p3t(PO^~7~ z3x|@LMCU2LedsQ&K2%B1_!B}&%o3$=J-AU z7fYlSoXI^>A9ggTVo}>I;IlBpR5{&I{8fLH5whsc<*~e)L_de03T<1SRg>ta3QCi% z;2fC!j{v7{2u>mLvan$^Vgto;200CX96LG-QXqbU@Y>+#KU1aFh!FG=-+I1 zK(i*08>&pRV#7IK<5K&c{6avTLHEPMRO+D!k&+5g^dP-6lG_t z-it*so)Zg6aizH=gQH63!;^-~Yq02lCTSS!XAGCxv%cNe=C^~kUXix?ZcycPyO^K{ zi}AKHIel&$nRzCU4Q1XlqDrwUf z^C)Oedl$D!(nez!{UGZ#RTrf;pVpz1Z0gP;ybj1udfWgHG|cqxwpl~9BMQ|;VfF*Q z4~VG5U@yY)344ax>*W}xL-J_G!bl(CLWuRL6wSwq`_VES3efUF9zrlUA=jd=LC_uM zw|Br_OZwNqjx5O2FcpEsw?)GSBuY3SB)tPQQxu5!5r}RiWl>m7qW8*%s3V5+e5nl= z>2ObvYy@DlF`sB50!^-7CblW4;e5vQMVCMXn@5XfxA+dh3<^rIP2Hm+5MUeCua!YqE^}7T6d`FM< z#q5%qqUC#bXN_kyi)AuT&yo9471!gyqU$&%TPFGjPS@(gabbRL2Ro~=q&y3kVO{c0 zDT%?k9;KOX851$lovNIEBYCkhy}b8GUL2Q9H!aNquXhM<8t2-%j2pOR8^2?zy%ON4 zNpus#hwT2PcAhfWMtkAmKo6rV^>SX7(?fzN;3ef0oX`rFbNnplrkX(6Up zFrajr-)nTX9%DGF+vW`F-Yw*})b-b(nTvU-2-d$AV7*XaeFm^Goe93ybTNxw383Fy z=33}SIsJGUdh>|*{#~{VA&4f?&1J5+!ZSPEShX)=Odx2DC4>J@-tiu|k1+u}n_%V^<_c#{8a9p<$tMxD?U*FRK6b{?Y$MBN0EO!bZPL?T0Tj zr_yU@jShic{GV(P*UJoVg2b3{@{lK(FmEYy!iilmEpf=q+VC|qZ-V` z3)RO7rliA2Pj6bVpt-HHxv`Y!I$qJ{md>X3mgItu5X*h+8m6DnoGd+BQ;7kgWDO!9 zPANSuEy7cYF{e6~wv>c^%ew}UJrNd~#G(4omk&2X%4$O{6V9t)nBd}seMq>^sBQaNZ>`cF;_lS~p z#}s?csbS8}YK^d8cK-M8j*6Z0B^z3AM*n!7P4xvhF=n-=lJLb_ll*#l<=>C!X8G2Zf zKO@P#-NInAc87%2RwTCYT%zUR=dT8-2&TxQ^3=RuhC0Zxo~T`GH`+W;&UB-~+*lnz z_(wn}5YVnWL-;0mas^CmDwyB#84|QPPNSZ+6?jZ@=ic;w;x>X&iyt1o`Fv zY-za*teOQgvQtf1632y+%Q5&t2zBIfS9i?xVgpv~@4-(JxfVkB5x~IPjbQgt0BZz% zYttVUzE?x<3*YJK0KORq2Jp=U{+HIJeFw5uv9gNS#B>l(&_h5L~8z`bP+U{ecb(6TK^=Y-jBj~eGsG02~6+Ok$$in(PKVuwH8AQisviNnKC`44~|jQ@up?)z}d zOEBD(bR0k%Inv64tFL;$To@&Ph(7};3UlWNIu=avv_zB?kZZE>V1vM>gG~a@*i-%I z;&gdd6(RtZ8S#vP{q5&y8WUt2CI}FX;4zZ&taR%lKf_;;0TCv#7Ak0&vkpqGjwS7SjV_>=Oq}NX0>ofAKE68g|P)6I6wcAb=WP7YNt;0K5hO!s3{sDkzWb zO#}L-2iZ~S+tRBIFfx{J*fh5-f%pu-L{$adfQqCSA94y* zFn!wF6z3_69yanQ+3Q6i(Tn83Ju_U5c2*XIywaYnhTEH;mzzb%24 zo@54hz*}(y~HB?GkLN3&XQ7gpS!Zz(k0vmA(qKp1`2ROv;GoUxgFaX2A zRt$LK2@gf4S9~;GxK!$WhFzdmX5i13Vg`CqbwJ$;0fCmDgdCU619s@Se#z39j`1Ow z-vrRn!P`SPSkSHJSV-7i^h@?Y;A zkrj~M&v^~wjZ0-3+*j1Un{@wFg?IU~#q=ksuivllLBBpI?m+$k3%ZxT6J3DCbepbt z6MOUP_NAC#m_4ZZQopJpsrhpM^)bI@9~lnQe+>Ghrcocvwew;11ELJm1LhqdyL9;|e#(ya|~xyCj&qm{n;PdD%>H^$w*6eL}wf3%Z;{fENvq z<@j~Z{-XY-2sdwvkt^WrVdq*H=d;yAq9X0#)mD`?FM{EH3nmuk=XNM!xi$hsx%nA- zJp}2%XC(PFpq!}-{=??LJRZi4-Ls7lSeEs(gC?=2bMQBSi4=GOs~;SaC905^kEFIQ z2Lg^B$PI9Iom}87l*b0pW`-^UFOc zltAz*5mHdiC4MzeN;{)6R5I-po0JL(H(JT`qd4NK% z`;e$gdZQQ+qP3FY?;;!6V`{I^Q`@YeVQUwE(Bu~B#|JC`MYQIW?4#PK3@Sv38~-Di zW4pnR3+nL-cB*b6BLMEt>@WBvZ*7%Kzjn9vtkebgOXoVNa(K|^n1GRt-W<2gS- z!nAJ%8T96e{iHZUe<6e34806}zSC;rKZ6?df*xio=Npn3bApf3QG5(88(UHMxYMYP zgnjkh`a=5q0R8Qt*7pIrkl}rR{!UQC7(jnF_&5g8g^weFE~LK?(BBJceIK9;8Qurz zs&eNHmnwTiZIP z&u?AW)X?5^%Ji=^Em)9jpMGNN^qI#TYadfPeL-_W(~_pf=^gEj(@R%`21v%89IfrC zrgmd2jc7}<)4xhBWEhQYi=b44mQSny3JOgU4J5vjXA0sy1G#uC+}j4-XQ0r;OEOSD z2KPm4J*TJ8goe+*sD+9YnrLay+hT6@TpwQ%z3QdCI)rTI)2gJBZuD!jr zU0$qFw>Nv*z>VHYQYKBP<~Xj2^q1y#|C6E~Ij)3q7N!hzAw3An83LWT(Y37UFhrF@ zAh<9`#50X<#t!`vv<>|-6Huiy=IPa~6rDRp!xlXs za_Jn+WV{p_iqz40@&a+)r67it^tEGOrGLvOGmtX@= zFLcFdE5guSk+FFl+p~Jq{&bM56wPuO+=bUwxcE_?UKEv`I0*dN7*_k!zuA%%4p^6F zY)wxBvi+I90YXVJMa1M&!Sl~yr~g`Q>?&sKme?VKhl7FIrx4JtSNG-}V0Mr)yM+SI zK?LLTNd3v6Lz~c{HBpv90sRrCAwf<8OcO1(5sL_f*>F;N!ZOX*xa#)c61+*^LqIaSO?xDNQ6w}Qh?wQV^ zJYsDDUkEG`jwGa75iK$OHRRG?fclFv;?MSk%rLq%(+j+VweDqyLT;23^Vo?fp|`pR8Qzf?e#%2mp+!#sg{)L0QYSh&nG$h!;4grzShO-TXq*cT z6#UvBSxv8Z$B@IJ^9r{FCgum&IGw7skU;eMzFvBLA;VkZs6j(ry>uz%neLPc0)C&S z4&howdO}jP>2MwcDu%^?3;`rAB)68HLl%pE5*h-E;Xeh{u#Deaj;_)0%aCG&g1Z|% z7ILE)oe1rcfe1&UbR{M-Z4HU*id=h+yAqk+?#{xT{O)v&b9528{;fYNei-!g3CgB;7bAAvU+{?7J3tDRx%3xG(a*&xyT8j91eJgo<01 zg=~bLrIf$wH!cX^x(RSW9FJ9zU}x^UWP<5S(0Gc!w*E^O!(ovyT3gw{#@N5t1}z(~IukkDJ=SZ~qIYB;jm>X)hS(1EbW?pvOvn^a!A z3j7>9A?A`Ao$o0ZVL}=S`8$K%$#uG4ok;>DP1z@gF!&~Pe%qMmk zhn+C}1@Z76lgM0&I*!&SxSk~n50oJ@ymH@MpNX-iC=}(I%W_v;oK5Q%JrzgtINj z4*g_#mUHxy3fA!U#Q_ixmQ_~{Gp$-aNT-$!G46uU#~ssoWkVxh2R{7k`DFt-Fq!Vp z$9er}*#JFTK+VT_J*R8{u99bmC*-{;FtkKV*HZso=@{LD5xx^2ej7^Z0C5`09x?MazOTl2?7nnyftDdN*nEFlmhuMk zeK+5W%2LoXb05bUn<-5asIM(PN*lDqu0>IT>*{Ig6=;g*zgf7+Bs&RzJI1rDA zKpQ9H@$@FSv<=?V`qW^$u*{{mmSwq*-c&(+iNuOvQLJDDY@vsj^->=?dUIj=O&y|3 z%Us$lDd;m+CY+u?)%nN$wU{NBn=YM$v5;xvug@ouatTJ~R;U$F6g}+6w0LltrqglD zbT?)kwlr_@$&?rem9Asam1P5Uw6fc#2jJg+y(yH?E>LU;B@qtmxrd!eC`6nMoa1m^ zGy-ZFb!F+HWj1d?96^yDK4w7P&|v)kLEfM_*Y<_s+G9GvrWzn3vKrS^R;`WBapSe0 zR#`|(_-T_0X>ESmzSWV)2U@hl!0cA(T=U24Tx&~bvbm+Hefomtj?OVTbbJ7}8uH9j z(+A3vn$6D!yZWvTo+F zrARb%!BR@hPj<{NC4XrK@PWLf@goVPI!m%r4;m{7XG#zTY19=}(%RnKQj%0mi#}*PX5%W~fhlT~MjVF0?SEqO{x(9eM4FltfN62*~O&MNp zDerk4X^VYNKvh<--{w(rn{wO|+^JE5LL)4}E9AaeBv%5;3e#ArcML@|sk##P_U|_G zifC0P27YlD%mg{k!yJctV|6Br)VD{b=5*>?BU5{H>Q0@i2;TQPX@`+#E3OD({rw2H ziB$m{j~a3ldjvOF?d6AUcyHhF?Zf0r?xhXX5smsTR8>6zj9T7>aNe5Tz<~<~L!n6` z9G?rK06a2++Gy>nh9!_v%b0$$p^mx*`*hmS9l@Y3deslrRE_c9qTa)}c$ z?Sq$v-j{vxvcP+(#xZIhw*y*Zw&;hpBH9lbnk576msge44DF#spMpuoc%}xaQL+Lm8*{&CZ%K7hztjEoH2e-9o{Gw ziNsV77{4YkGR?xvSG^Y(FJJLqzJQm1_Flv`^2^@Km+)n__p(uJEN=%RgUUBSeNPKx zbbvSu@8fbmL;ZPc2I%=6GBe%kgL0?gN*HI*o(}NyZW_WiVtNC@Hl(ccb)bLuQ-18H zyoeNjM2;Gx!c3eN;O-vaZXF0fnIq|5Uo5h*kmn)f>GAVi?dO>GGx;w6^>^}mr~i6~yx!rz-YKuY^Iz|h*G>NG-O@}#RnqN#(!G-7Hvjd0 z$zhR1lZ&&B0K$udh2MX+*JqO+hb_5mn!+Gi5g3}%@?Q9-L?Uwcn;7mUs3LEzVcHM- zkQHHyWku=)^Bs(<1ke=z%UbJA1HDIT&kYMRFizzuwlFWP`f3@W(C&JmVjh1Swxy}1u{G7)a^m!4N6XAoC@oJvSQ0%aWuS;FrJ>U3 z;}RW2+nY{U(9~FxXKX+eCmWhe>PKf&dnx%ln-?~fl7CT4b7O0&Y5Jnh6C$OQE1k&h zSkT++Ai&lH;iR?PPQyPVNr{?c#qeN zqSQFSFt7$t!$iaIrVZXE8Ah_RwHc39(#lM6a-M7$4U3u=q?)ArlYoiVq?f6wqcPbg z2mppoNNT@i_*UyxgKYa4M)BN2(!QDj-M&GegO#v69mL&+yoPhiWQz_P0*OS&XW_+r zn2Y;aIL9z?|8Re;K8~vp)62`UBCj$k%(h?X)|5$yRl~SgC`~Kp^EaCd-;W&Y`;1G2 z@>^jFNQj0kL2nR%UBup~5=40)qLyYweku3FBARg=sS~`^Wj3rNa9wksZ~+;Z_LCZ_ zKB0y@~tD)VFGr~#>5Qn)} z*Vl4=1R#JQ22&<6(AJDReFvOFQCl-$RQj6fDwFcv z*;x#MD}6Jjn+=crXmh;~V~(H!Vs1tBiU}RBgF(NbI|gR!S<{6_a-o;g*khSMz-a)z| zE-#O<4oxPDr5%&YqQ#guePH|Gt_v|rTNSh!QBlMxRVps|H3Cv|QM zf}1!kwf$AI{<@(3&B7(>Nd?OtS?}V=D^53&)x!Eg714vgEcYm>1K6UU3saU;Cdp+C z68jw7&=#y)I;EGEWdpR|5VZG-*B){{hLMKid|%ylYn{Y+c60Uz|lYC@P8N*mtr<{G>sxa3f7*fD_C$y`d;~H!MEVXVwysv6W6E}8)_?&Au7+hY=MbnOM@xbc*?v>d zYF?q$oPZ0EXVtPe3<3mN(koUli0OB&f(V{Fz)S=Yd1#2A>F+7LRbNT`nr$zG-Vb`VcDevxl zv<6V|FMedY00aIRFw~a}oN26xUzja~Id^ckLPu8O5&saoGNd-~NIi^fBJ*&1VJSJXGbTmc!Gyqm zGQ%yJmd9?;*jrR-2K$@`GlB(Muu~9iD8>@gPOjcg$2cCro!zos;i$E?0wFTf=(t|r zJ8;hJL1#9INkE^P!t{qpaQ?^M`m44_m^Czcm#*#PtZsZN7ODel z9G$*2i=o#oPx4ys$TCDH(B57rLTh}MrHYc*=bR(EJag>bcGA0LXx|Y zZj!a;Fz!auUz@UYp{|l%47s>YbS2YC**c~*(9fPK1^Tc8CU_+zBEC)<=Ovwz1y)dP zxtFpWD;%a%&9LLa9SSnp<8p}#9fh?cM}v*2htoG=eeD#4!J$;RW4=gAM9=uL*3TqJP%`^m6*7C`X&tFiwXoBC2}xgQ?j&AmX$s zG(_iTVoYBL#e%DcyorIyf#0rAUn1xw}1_08q{ax=KdN?~mo_HNQn09*o^>g7elzBM&G1k|GaEUEFn2y2n zlM*)l1H&{RCr_MhHH9sDO$~s&SRTU?dRc+EJ8QYkeY7vOHEhXPDTptM=eXcUqI)%eal*kyVHWq!@VkyA>N9Dy)fWZc2yB?+(%?4WBeNgzXddpvr0^->9# zek6uW(~LM!Bf!1m$;as~(EJ5+Fy-X?8fhh^Ge9k<2C$EBaBWkC$l3gDEK`HtdjAw)1dpz3&?e>YDt=?w{ z1M$>}*@)T?F)yajd^@W4qiL?1LP=Zg$8@lM#AROC6W7C0m$zVKu>h>C9F|RfiyS%* z(?4TU(`^#SlHlwHv~v#%^y>vdH?k}4n{@xu7>r7V`M3j7@4?u}k>S<0Rgb{mCO{Qs zGMxcJ5Rh-5fB_JQPH(r75A=k9LS+L$Am&vZyD;j*p^iv2Q<%2-b-nD@b#9@q7us!^ z9C4@IlF7GQg2RGO?77+X3i= zxuNz!rdnVfiw)bFo@2{_CL{&t(+(P-(_|K6tA3q8uOx7x0R2u5DclF|a1TyY-~*H9 zW;HHtXzzg@1>}q|Yi|*Q!YzzQ zI}UqX_4e*PKz}1NK-zBIVcH&YEvN+LRBq1Fo!wwR#(@4NbAav_Ce&eE0BF%c-YI(` zZZ)uZm`<{NZ>G&p1{BCk<(irXB8AGIh2h1OfMEL8mAlXcPoXVkDa`$Ia4t_o*D+!Q zXG3AhSDBq2$NJ}>=6kH_N>RFjg@NV^_qXE)b4p{zUgIkcH$N2=d01{=%k>L}t6bS( z_CPVN@InS`4YD)HZ5r5n=mH(g_q-Fc-cFgvc&aJ%=PDT5whzvVuoFc`Hg}1T1HCqr zJJ1jwtw$Kk#iCvuG8n^)FlkZXnsQfF(?t_y6QIAGY-3p+O&3kvg|E+_jA_5z_IJsM}Vgl)4DPF z3;O_MM_7oUuiv-nvw2+7Q0*deb0!E{-=4L$Za) z8|=DeN#ad=*$9c2)MBkT#~Wmz*WyZfj3UPMT|icig$h-yr+>z#=`bxpL(1c?2C;OFZ)HAK2wbhqK<>$GWbHD)1|Qwn+Lol0So8)4#G2DDO50XLHo_J3aFnWDCGbhS*?+F16uRzgfD=2#Rfm!-(T8l~wGl zncL%e^6D_A>oc~hv1qfAVhMj2#q`15Z3%02PtV*0U(YK!jvd{LPO;(dL3{54 zOxX%P|DHee(_Rlgi=@;Gmms@xbO-p=I)A9Q7qYH{IfYz12*cD=1VFg#n6R-lD$E)G z2r4}f3kIBsmO~^YE9oM{DUd)STSS<+RdC&k=<%s~BdFq5h$A5l-3}fNxHvNeSO3@i z*Yy$x1xzwA-Ij^bTv*c7&{KxQ4!AgjSEW__iV$8+K^BPIA#qun7n7JQagFJ9NTf3} z_G(3BrjWVY29NBh8d@D1Qu1*gKIZWeV5|mwp#fDx&lwOHpO?|WtuW>35D-Q+Om`c@ zLczx)X!=={L-Hjdl+eLHYgh;WaKYL3vtUOeH1OHk6r-iyo;$!ClIx0Ev6LHn{fT1 z)A0h(X}bXl8y-1AJxFG;CUhi(7rlY(OzSdPWTEXwiigAghE(-#(AFiMFck1v{^dTn zT6MeOQZ5sNpCCi@Ot0eX{RxOIxgQeOj_Gp;+tr?kvnU-4OO6+f0S+TT484H(4_Je= zLIaG3UN>A?m5C9%^f&ZL_mL$%>wyn3-v>ANf{~)1WNfsGMs=Id8!meORlU?#3AaAu zYUqVDHUS9ER{29RBNXJvwP`ud;o_FK&{uuJgIoJxv!1h(S7mB>8oA|&Y{UCP@ z!%#J}LL|$%S?>U>ccs@kRYR*}U=p|{z+tnS{%IHbOy38<=4a@LT}=ONv-xE@dY53$ z;fFoe#iHd|e{C3LL2pji?L^;0I~u!a=L~6wzKwdJ=}2%NujW4CUvVH<5jw`DZy~HN z+fRh+B6V9Mk%*lLM{)lf^k$j`1HlX~2As3Q)=Gb%ruVKixejW!v6&*?Vlg!vh|!`LOh zhnDPINH%M|f_LK6Eg73l*GstzDZfF=*XSNyS-sbD2vY~^g+hns<{c^4JBLA+J4~lQ<`rdx%R>Yj7d$^*nd#*|vxWwZU9;%QOpGn@%yz^z+k0kt zaoAi+j|(i+lntnk2>}l86dd-cDRkxJ6sG=`maMR77F{_R+(-@6)|Os1g?>5NW!iR< zjg;Sj;r>m=R#WIy6r*=rY>`gIIewj*Lc>)q^(?Se4RXTBRHjs1?w(Uq=BbB(m2pSW8fum30yM9Yt$54=V!a5HKwM}wkj9M{dsE)lY~VlglxA|c(vWN z*p8xU`gMim{Jm%kaWG1#SS%TH8VGbP2(l1yn&a5;Q11(&yqZGigJJ*`!t|Wnb_u=4 zzv1fkDGW18RYSLhT!<|!$g;qT;maN6Zapfwc?l$=a{s=npX&a#G&4MltA)~DGI9()hq@@$Dmt&=(g7s2?uGg%i*UGW<@nNleNN~7kUQZ8{ zybMMSn1%H%=AvUIDKv21bs)tCLbRa^nH%z?0-OMurUhbVGv(DTbWqjMKOoOwv#odv zT>-rls0fVa??Wjy1wjt+B6q%i3w&WNBAR&C77`X6i*_x-E%d0_q94Rq9H!ALFtj2< z020m*QBTnPPBd@+EaW2)H0p1pc^p-~@SC0u4baaqi)ou4TBcu!%qOBu!qVm*RYU(0 zV3qzH8lcOhz~g>_b2GbrnNs55U^D!UR{jBIkyZi0=XXH?<838xE(RAch3Oi9<^M@m zZogi+54J*3F@y|nSuUM}DbDn0tIctw{TQp3{*F3i7U*^$f;XAYBJ8$PnDP_Agw!xS zAyr?YtKQmb2Ozx(HU;Cy!`R-IVjqNpt^&jVS9a;D3YX?#7UUsb6tNvUQ^GvdK;JQdwO?xwpCX8OFX1%=xm9rnPUe{oKxl2>iSP>gT7ZfBKneZtK5> zswr$n=@It#oC{AJzuXupV zbToE8WC?IVWhs{RCPqgP5?RCSSp9-Sryj>44n|55^k4hA)Y{90If(n@rn*dU#$jxn zX3fqu9Jf|Q1s(nk>)1=MWGpHl=D0Q9c3u~lWbv9}igz|;V086P@hvWx6HT4chmnd~yu$*5|;8&%8#Nw~hvrbCCp zOw;U`I+T99999T5OdHh@RI}q!&?(Zt+Gdv#f)t|d$Y6JPyr4~geBV3}M|((*X>7HU zCp$(%zgU3%#9RcmgHKU@PkEtHa5uuIv*+ z(X814IX6sKKJSS*l3UI2VyPfgkH}G4yfjWX#b` z&{y=GEKL_H2`!++J)`ukEaB$())+gSpl;wO0rxDp%l|2q=YHJBjqzslc-nx=K5^;v zpF=JW(|g^4WIQo5z>StfY3n0%A=XwB9JBfD9bs+^vtxdZ%w5zSON4nXwi64p);w-U zB&v5zdHvcBcCMxIXdlRhpi1{lzWyt9?M6*SRfcv^8dVl-K6D*fO7C7tnJjhhUbzx{ z>-Y5hn(YJtUlGzGL`-ZAaFij(=>I59OM6)~&&DZTq@*yBvVBtp4yF2rOIKR^;Kt($eaXzOmNJ4WN& zXdTN`PTLS@4`a=Ef%K}r6!dafOQ&!&)M7O_To;~#2g@SJ|Cia=ipo-W&?*dwcYzIBl4bMoVItR$_&4V(LAx*c^9LZp z(h#EFe+R*GIMcBM994(WzuW3?s!hp|u7p)5#$FvJJXj>64u==wVe~Z{MP>mps0;gv z**Od&zkk)UcH38jR;OXoLVXG~0Z{({bTa?aXOjmK#dy-S~+qiSxobzrT z#QN3TS2Qu^d?Tn{d5Xhf1Xi8JD{+uH=J(!<&I?CnBQv|+g_XA47eos%bNB*rhRKvGsOdw{g@c@y0Uq68+u@{sAT*< z?!b7>toYASUD3^a(&Im4&i_g?+q`Bfb2wD3xap~565rO%>zJmhMz&rRqBvNUX)b&g z@W?bRkJKzX{u(sC6O^1Qq{lQN=7v2@ZWo_c+U5Oqz12tcpLDdf#d)9qzW<=#x{}YrJs6iK>9|8{OIgwth z3#GJvhsV(J$!)jrA*f+E3aAiXu!HN}XW9>f$^qIQRiUrY`TKf9+E7UEWAi_TeU-zUc{)EkB6kLJ=C7o`Vv2GN#p!v+pgGxLj1}U=2iC^{oZ3Rg?s%rLN1nJuiD?#pJox*KnhEii)?|m=Ga53C6C33T z@qN9Cf)$-3xz*lWE^931a;0RO0Fs- z+qx-e&Y+ttW116FCfV*vj(e9+%r%$jZOrHlYTE=W4c6D%fGDqLL|Ubne{r7RD=-+Umx-EMl ztQzntURtnB4HjG}_C+g$j%jbnPj zfW$^VB>5Y*>VCHI6NUG)|Pi&^IX9 z$g+cGp7xRO3yp-K_d$XQ3k{9g`Ms)Kq%fu8;W)66;>aL>G#;Xu9FyUsBC;qbyF8n0 zHIXg4J=2R;X?jXc#QY1M#vb+QA{NSFYW%mY6dfVd0Hm`f6pBIwgp8o82%+Fz)?Pj`LPB|QgoNNF zUKAmryikp0cyS6m^es+$AsQ+QAsQ+Rj}?efpu-guJ&r4;&&C(Ol!2~l@0hP6ELeuo z>_;*{h=m};ffqJ!PF!;qV?-@@EHnfXh%VR8-H$=p3|8w8-9Z$B?OthR(mzC2xdsd< zP9PDsKOurkQAM!Bsg*bpmqX0DGO)c$=;oD4>x#_VhN)7!W5t=eK-o~Y{1u*TjOTG6 zB%n@)J78QqwBugBM1MqVX{@^VJ0ERG0gqby|A#o4N5T_*ndomvF~*u7eaYu(t~MHtij_H zbQ`7~Br5R)irs_=1}OG>UF=sVb~K9pJ`Ji0roMN=CPBd`?siyfz~vN99-bB8oD+i{ z+_z^B=EX!quCJ&LdQbyE&`Zn$2<1f*57O5QJ4Q>g!s^4cetuB@);;V0&luATm+8A8 zor3yjqJHWI)w>#rqP75e&2u>}TH$^jw*w28y6if?GwYj!#$MaAv0uX?_#T*{{#FQH zR4pTNuYm?{0FIYn)FbSymaj5aLBrG;tMj|H;fFy>KZh$nNNLdhzzV|Lr(^0Mrgv~t zfHGm?cN*GyUgj;~o7`9dn z-5M}Kcn|}!{TMFDFrhGMJVqfTJSg&Pi3W3Xv0z;AzS8f}wzi<{4N%;S>QN6SL<-{t z2Pc^jLfuOAZ(&y{aO*XEW zXPPf1KPCOf{6_yVlyFU%s%Ju?GidrUv0GTn*B~2y$cjP7^Oi5*cJbvaAM~;{$YOD z>T}iFBe>+#m}%!pFm9U4w5l4*7(L&!nh$Xlq8_c*2hL$oLB`?SQXKaLPPW(HWFEF<{|W+u?try;n>NQ!Hfm1ysR@L zI{h-68>JcZ5L(Z>utp&-0>EHh4`SN1%w~WNhaK3XnoL)TkrDD**tr~D4Ha`e)bz-u zE5I-uL=P{6jOCKDJm+Yh?F)o9EcY-}P4_}})kA|gcrjYFvKcz+5T>)!S^fS`Cb^70s&K6yg?xk%~fjI}}BJ`+O{>5R==r55u-9eKNNRt?gKT4r`5CYPX zl3KkY-AhkM1?C*OEWMi^@>;z*9T@jlWr9|>2Cd#9g+5}d*Qa}Fvs7Ttq5kx4y4!2@ zrb4TmR|Ks-k5)OyLXrdshVkJT2%d#mB}mu<#hcQ-h}4Sns3%0Xr*{j@0~_zV(=mTo zx1AcayFLU#A9mzb{s_#-24BzK^XrPj*NgBKCl|`Ypq8qJQ4@|1$D&r-^bWu|D}x#@ zK^e|r&0v;tVwQxqaM)T$U0K)_Esgp5S7TLB>{lp;1s>tm0aXyEZE&QZi!(8ZEXL`ly#xY$wBeNUNW-Xm1B1N!6Ikp~`uA#@+ zq>FzzQ-v^qY77=I5?UiJB53V>1m7wdhqE#<-|cQrPtd_X!7c{XH(=YxnT>5s?pDWO zUMCKt`eGxjXZ#?*f+-h%*Tc9gAv*>342s)O<7x=Q*uy}_Vit%mR|7kAqRn&=EZAj! zeMmG}l_cY=PJ znr9~*p>JtRRVx8Hog#Uj7ZppC$I)pE^ox1G4!xP0$sHh$;vA-~-ID!NAZy}GjW*+I zGIXyC;~S^Wz%g$2k3AG0M6Y9s(G~Cpz;(<+OqXY5SH+(CD7-Nmg{oVNsA$+9XVK4u z@TI4XA@31*HxHqwWVhC~b3&Re$qI|pzrSIa#-@soBOugL!0Anmsg7hRA}CMJQ^L&! z|8GbReS9rg(A?J9+*k?%{GyGilCHAwiUnFZeZ4^`4R)4> zZcItXtWg)rR2W8Uiw^QzByS&Q7z-D5HZ4I-l}6D8H{*@qf}05=q6kkc$o7jKCl%r; zSB=ybR26Frf|p`#K~P(|zS;EflOq z?z*slJv8dEBT}m95f#{x(N!X4M3rrI$BNzuCA~+KtUGq*C}5B>qNGgFg1tv*)M$i? z)glu;h+ErJ%`J=3hkc5IK2Py4B*2IL3`4JeynL$oWq&^wH&Q$O@0iqk#?YAVCTgnZjx^9lxcL3PHplh7_u}q#@mjqld#g|hLLKOZ{xIW$;Q() z?I<@9{X*C`Ym2rn97+fJlNsxEE@jd^pfH8Gi3^ir9RBVDzC!CWF=f(UO)w4n zG5s`ctAh|dKpj|I@)BM07rNx%jg&I!IXFPoQo>jFoG;l#f>hyIRd$TA^&`w3JA7>(B?hN?LeaXBqo7KF>5c23aN zUm>vM9^m?7@Ghf<;gv5#q?JtkykE&8WFZrJ1MVV;ev@(hR79)nO-qm zmRmsdbw_1Pd2nvH-n~mRjj-_&83N2XAoz>0RQDi)|0!irf7hTYqX#K(B_KNO7v~j> zoEAMeU?LscH(Q>4zSl`b?@9EYzNmQ9^>7&BR{+85CY<)RU|yiYe7CJrA;9!f>Brp= z&h%?gt_1&xtdAKfRfecquwG$$6p=Yl_Hx}P@S}TXKo3_kM6wV$tb;J$n{H?fqI9p< zEYkJh1&i3~!6m3QxNoZrKJh*vuSz>-`>!wgfSdtOXh1qv22B5zB8QNMumGqDrgSTP zgWCQJJ>;2mnii!Z^iQ0_0+qvkYCl@t9a9Hk3GcDnJgBgWS9Zs+YId6`-U^wy7jdZV zXF=7cqiQviPQzWd2sWmU4pQx%*U8A5jejZePDymjT(@z7akKsBsoqi3S zivWf>8HuW+>8_5w!~Tp4F8khAXy6}M&SSyeBd{Zg&1l$vC9q>eOs2O-!M5O{Hr^Eu>ve;^>28x6#jdjrC_B&hlYRINTs0|5e$;5zj= zdJEx%H1gDEsb|^Vad)3`P0z7sbD{1#r0z1NmzT@od$tDiyk7=p`T{H-%jojuvY$Pv#@^3NI^4z5LWz341Ab5mrqAKrhb&6m zx}eO%j`K(qaDBBGK+SShnHZKbxhd3ra~Yz@-m}aOE2dk@Tx4AdErw#+S%y>k&SkcI zO5qUeTMQRmlj$^tYx{}m_2sy>-(*@VDa^E~YzWR;h?VCsoHz0H=4HTJ8Pit;4i|8I z1xNW8Qi|!7(lMk-~qhVClc1+AqxBc^Skf+9s0D$ps{ z5$8WJk22|wW!Xn{NOte$h1pjGQ|=tBSQM47u z$vPnRXF}L`>LtXGK(V3jY#1gFQY92li*33!!S%h(UHroh&qbifnin9#lMH#BsySK z)l1kj)c0eU;2N(o$PR>DF>rndax|k4tt_CRP`<8E6oz z>HvPOLtJ4H;Qc^4JJtYo#p)?-a}F!b3x>G)dJVI3v4cYLIE$B*_>aEDzL@$9d8R?kl*tknsZ}Oeo$;y+=r;@rlL`~-9sfk>?> zbxN{5)p2ZH-OQ*_f=0p<=#5#=^b?w!7L;_QUdR8Yl>KAL0!u(qfE&nXTia8mxlw7M z;@GmsqA`QuBxD#Th%g10oi!DlaP(`D>yRlInk`7SFI==hL@pse!K)LdY28UQWuXu8O!^VzJfL^r9T~?q_Nc`6#47Cr;>y((X&vaf?SaOU#Iba&X41w zkn`-&0QAYjEXKGOrL(>-_=ewVP?Sk;2op*-WO@-HHpLDC#L{Ni9kPFh(WO^D#D~jl6A&`@J>ygSY3=QG)+B9r@BnQiKO8G+oy$6YBD|U0V`KTL3Gxjzh*-NMO`uJr?6v&qE-_Iat@=Guf_C`&B6&p zOq&hdih;W{T*NGuzJQnZet@_iH=QX0!GYH_&;iWgB{Qf%+|urJuda6mqSG=B<2ZfW zhe?@0H9M+`X^mXBZ0d;iaipB>r)-4>58>hiY6yf5iA@1=5dR~`>qG^OQxLQuF+FQ| zqo`=)2#$I+y~}&GBHE}ht9}<$eLp0O-mPC=)j$QRf8;|}zej6`GX@$ZXm4Agy}?oK zp-l*92s8P8yK@ERXD|hbelG8N{(?Z!)-;d|cx>E;JAxiQQ`Dk`8^3c=J#)Sd^ClDb zh@tiG*myo>KoQsxSscprOp;LX;d;s@?hGn^St^F?6k$PRzB#be)aJXW)Al=L(hHb4 zvH>^|MeqTWCLlXhCcP}dfM&__!>o_#>U1yD!=XGo z)GNVF9oiCd!(pahVtr!nU{iQFB=Tp_PGaMV{eph!EzyTUDZjrP?+zMRD;E4Tet@Y{ z#HL5@WQQv9Fv`X-#vyjXmVh6mT*NdvAM_Jj!-#!m)Vvw>YSL-oCv>eK?;t}h`a>Oo z-Q|jV0~jy*Ffiiw4GrI#BIr!8QHo|2Ho{R5je0fF8;l>iy>|d-|2_cldJW*X0nc7p zBhkv^HjF)xcpZ?2JEM!h90kRwR}*(V>c`)HjD28gvaIf@ypBDS#^ zrh6gtrJK5Kru$G0-Kftf8A$VJy4S=SgH(W6b7D+)z$_Fy&CU9}xI2w~$D-e0wKLs? zB6UTA7%M6y~jLjZI<$-x15L3knZXv#=pC97*qqw3+mN+$Lsm5h$ zh|$3=;w9<8U1TF3R6iwvUuwmwj>G@fVQLxlWgqJw3fjUc4VDksY|vyue^}noib#NA zIV$eSKCAPmptKhiUh>y7O>zNmJ`>XbKZ!$-Y4Q*z+GG06?D0Qt!^c5HhSsgZ4$EO$ z8UVRsFlgp4vM=k?G6I`jtSf}_=Iwze~<@7`-&j`95=Q!@9*T%|doHD^}6)PU~dSua` zG)OI#rGn$;8Oz&=*qzKm87bXfAEnOZ^dTni%0|eZ8bmr7Qn(gzFo05=Tb@ z<4rEV1V@-E-3F<>hn+$xbgcMSXhavO4~lqg^WKZp*Vi`y|4h%zQcloMO_!dPXNxX0 zQ+TEeO*cYMciXhVL`-~{`&bc?2))={IF%mvN2HQ=nbkmCF{o7OMi7D$O+Nm3gUmC! zL0U@CA5E9~<=LXmW(v==8LeL1ZPUZLRcR4(H4`RGQjE2jnTn)F3e+>zr1-<< zPGA)3^(b5o(K;vS5#`cV@@&x-mBKS^QEr5;SZ>o3N=89&q4D8G&)V>$j}ICUmkQd6 zm9!>}H9^mZT)IF4j(aOW{>6ofBi04ecr2Xa;iJ#{DM-bykb+$i2oy%_a(@j_olD=B zWmk+wT#?rv(`$h?8Q|l-Y-2?vM}j~{rucn}>s3LU3{ZStaRA``5IO06%X&?Nm!l+R z;>f|TbXo2qE{r8=jWh&{2ZLY{}K$*o#oY$NQ7!#qQd`uO`q|!pRbYztZMA&sBp}mw1SB+)^E&2KV$IVClvqd z{YR5s?L4I45dcv3n-W4sP5Z9qX~!X8;MT^@4jIJc$7D`qoFe;1Je zU;^RlV6Nc1Xp7{b5g0~@;!H2D4U1@Q9EYH$l{^k~7s5CLf(fD-YPXq0a$#Z}f$CYR zshjcvyexsX%cA!T+y;quXo1ZRwW0cnNB}?r%nkB{qqPF409}Gmde;xldC?oY3W^~7hFlX1?Yi_Fk_RC_x3XhZ z$qX_-4~s6zkX@m`RfnQ6#ZS zrs@|eAq38pe*!{a*k+X^&@9P7`F}~K>zQemWYjHXO6~m#3rjNV)uPQ(&^ubB4gg^~ zS4tfnOFJ5q3rZ?7y4i--_U4w7(AimH?2guw&=QEz+geNOSaa+26Fyo?9u}uBYH40l z3iiiF<@An^FgsxhIi~O*vjPHh*cG?p!89Z-o=Uc#(%gcSke34Rz4u}nedp4)CL}Av z!1Tf+Dr*sue_3G0iV$ADk#d=?5n0(VnwGSAPc5zO3zJdeZJ1KDwg(XpVZNwSYG7#EKpX4Ki8(lGqq z=t(f*(+R0KR!+QqQ41C-5TmO;0Xwu;qLA9!*3^<}Zr>xlt+k_BX4Aw%YYiPTB?|4M z8oaeOp48eVgN)qzO}c^%g`{S|?nI1Bb7zxGmWhTjvh@j)fO5c=U}BO+qnFmwBn?&S z!~&cXVBwrBG&{jCmNm7v#v5DPFy9t6c;!7yVr3Y~_QqN)uhxY{y{|Hij%JyelMAY> zeY84iyMJ<QxzHfFBGuNl=|-^!0)#>FYs}`-RWQL52c-U|iVDcUGAZt=p*jhnG{) zcs;wo3v4kx8>lw#vxFiAV53Y4X1t+`i${7{B1U3{%nV+niDc&=1aar1_iGd2Eh*YW6qzV8SH-;6(czxb%^OiIi`Xj4F zey|TyQWU3ju&ef=q?pDYnyQ0Pn;m>kfd@ydESS2{Xg`!mflR0EO}$?zlWq`IPaim6 z5w`cU1JgPS(gXxTimmL}8mygbL$KKf4Uylqe9s@XImw$XHld@)BRnWt*cIGvlQx;qFU4+1{pwH)(C9weGI0IsV_F%JD zUP^<;*JrNvC{9dve2+u~uG6n?A#v(EINR_v6Uu9I>h9nj77pfA_)LPUL?%^_6;kyi4Oh1R2KdN5}?-Yv|jU&=J zWcr{)%xUjNwUJa@H0cTG=NX!EwCu1BgJ)4Y0?b%3e4Xf8^Egf|wjWPynmGlN37vd4 z1iY_#6s+It;1d~ayH>f20_J*-Cp;&YUUsN^aJBHf&~TJ7Z8u<_4a(u{3@-e>hU3%~ z7!9tVpdGo@YMtgQeuM8{2!9+n%W*5k^H?s(g~FUNwrd`;12YOYy2sghbl%*#is<@ujMwPC(v6{1z&&$- zok5xXSTpo|cg%yol3u`od@$t2)8+~j)3FlPV1W7QZa9X-K~{o3?Te^sbcOC5(-Y

;B-Et7Fx&gvIp{kethYQsVdM!-(K@zp>s5lCc^rlDhUz zyuMi!uKuWV*O%TU3YFDkXM&k}Tw)a9+EEFWSO`rpIcWGN*v8aUrrqf**2Ah0a5|M~ zS6cQ_whBo<5;5S|jEcqB1q-1G3N73LnI-4h!oK_Bi_Qags{OSr z)*sbDbbj|Be8*tCwH#u>J^jZldd$>5T31D(tsCOsvuhobQUmssgPl5Q`H1BB3h6@(|?+EsD_;mwsr& z4y_+~VKJPqGKMt!(E9(PdlAJ7z#QX zeRZrgpqM2mqL5l^qkKG^XV=sTQN3)C(pk>IxY)B>>OVN>)<)QzalRMLJ|fLtEzMdz zI#_Fj&8frem^C}~0~o=s-M$KG#ls?V%v{<(SJzL#dif);vlpvi#P~A;R38f z@*Ial1>*n2d`z!D)GKY$?}hM(L!nR}R)h}6O~k$0&`{7uF`JzRD$iSM7>2wKsAoDp z%hw>r3h*fxdSeCx!$`ot7AW~REL}lovvoC*JBw|qaOq?lS39@cG~T5STpkYZ6HH)s z%p>TO-E0km+w?VppYfn2s>E&x)BZTuW3rnPOsW>raXX04Vs4%DflLmk)VQDJ=&62!i zgzlUg`Z(R$OW4eS{_*eGE9I>JNJLN&?6N`b-Yca&kG9y0E^Wm1fFK%1UJn>D48NbC z2SNo>!Nj9~7a$pX#(49bcjugRAd2Cz-ikvGeCa?Yi@oCyhTIu}7{uiu)S=J8QTZVF zE+pNQ0qX_H%br>+1+#%b*%{oChiKQbbwt1@ z%U(d|XKaoO%wiDnO&t2fbSwl3=n(QJWk;v>813V7ILx*5gps1lGIqE=5~2UJwb48{ z3G~t9NkMysSy;6Bf#%EyfT$n=i}YsE2>fV}SKi?-k(FO4*eY;75hL%W<3r zHBqOd`+qqwiP&&*fQIu-9KTeRiqJCHA*bzxuC4{QX;kcCAx(PC1vzb# zmVotwiipzz3=Duc?(JQ zUVB}hwbrwq^{nT%91w3xbL~lT9i1k5^I}lnk|tFrNyny1bCaaErb+XXq~l#uvyFAf z>l*7nLsgS2NMClRXu50Mb-qIuG5>^IuCAf6E|;szEm%;OTTox0H$-*+>fN98-=y7T z$}|67>$kdGy=$hfZb4lgHO2oXKd1i=m^k%c9se0J-DKAZ%l{jOI)>Ek7fsKaHQjBv zd~P~>4Q7VZ$sg~9(>jk^nd#~crzhW7I!%M`)~2>nOY8ZkT+qDwRg`?rfO66ucm9N7 zsZ2F2mC1&s>C0rpQuKk&W;aOHU}nS^Yxx4F{$jJn-etAfT4I=+YWaAX(DGp)V+O?> z^&1m)AIiqWcsKASjCX12mFhm~m@wXTWgU#pY{u+NUuL{(aCQPFu|zQGbr!|G=Uw39 z0*B7}W#PLVzc>SB8&}bwOS`C?$O0~jKL@^;G3HJ-eT3#3?!x8+cKmfc$huMdfn*(O zpkPF0U0H$>jvPnDGVXtL00urc91pYQV{O8vTRCF|YK<4!2_Jr3N#0+m%_|q3Mk?>8 z&c1S9j?0sd2@GdA+qre8g_)8DSfHYvAd4Gwx{I^=9F@;hu+2w-FORHsH$})Ya)z8O z`nB%7&A}Y`NJTz)qsm=t3$*6;;V36xxBXmxhNqG#vA$~wVSkR?>#{n(o}2mD(2?$j zwSg~(48qEUb|fi_lyK+iLMmXtOVX1iuGSLq^AzRqF^X0?(mPga@K3To)-@BL>{vaymJQks%<;aEwkrzTx6R+3p$la`K^MS6Tp2bQuNLJtA^srroyjE@ zH^dj5QH8sn*nb~)FpJ6|BOGoG|Ndq~<=cJP+C(o>eNGD2ZomQ|{dep|Qg)Kv=ld7@ zVM6?O6PGuAYTyhoz&Rjl*vPu-SO1#KbJCHLf`*OWE1bAdwxib>_T_WNJR$eJ-u=?I79;$FI+9Ut#v&V5NfCJgm z#5-MPvONoj*Cw<5m|`*$yUwupWD@^|ipBbrifi*5` zX*{rFms@(1O^w0Hmi7zWZNz?ppvB3Woz)KQreL*9md8v*NNXKDR>x9RpW_ZwseBV{ zv)yTCPUa>KSQ>|yedBH|%IZ<98y3m9o$euyVNou@YJ?Zjap@^=EhlP&N;4MNHMyJ%GSMd>Rf13Pa~I1SYJ6b~MBVlZQ+1rs9` z=O5#y69$!kjC=7fN!x6A!yv>zL4~ZRL-Jr|1X*`o$)ceDtY-)AI(>__hJ#h#DZ%}tqS;v@1een3u_{(JQoR?Ri|#?Q^m4ZY>GL=N0}EpM;h z)YLDZ&g7B0Af%CWVc0jxsI@@#0@lDZd5R0^hHb ztNi`Yte^H|ofY2a{M|-0L-}P-Y?${Mk~Eah_ZBD~|C2TUM?3xNOuoL(x@_x2ED+buH ztC8kKHV$RBVrxKNnYcM3pzVmnwLBO`&S7dTrm-1xWb#{m`&?cMC^eTO^q&m0rpEv8{LIszFS#) zCQd^lyF9Fw!+vLT|K+}Na;$17#>XmO=CUM$Kt5zKa{p-C1A%MrV2c{UyWW#5r-cn^-KSye6v!o+LHRLU4@7c%&nVUO+iH!lz>KsZwvg~; zz=Ur|=ukDDHP*DA%_p*L$Q6^mY(00|muSCVrN>^SRIivc^rs@R>Ln={x1x&-0@>pY z5+*P3WqVJS4*r*K2GTe5AS`qSK=P;$CHfK>KTORWF2fkd>ldV5*au=kpP%jy5qwcfHuj$EZzK&peuSiC!n8_1WG!9L* zttsx3I9(|QE;1`G%gUxK@v8hi6FDw7ZR)0=Qr?w?WFVPg6ZS{6UwR5qmcJ{pLA{gD zMHAD8ShYw`nU8;uP^%%Rl)oeXlZ3Kxh}=88Qa#56vMj_R|GRyKK;#Z@5cX_9IWj~+ zrF^%~YO&ywiXEE;cC|ARL5?VXhl%9ywGTH3l`_xeifflMv(lRO>yo8$&nx7}!)9=| zGSk6})1gOe@t)i&tX<~Y*IlnvU) zpvT%~YZdnHDSLD<5#+yO(j}ehO>FqS=XJHTcAk1#A~cO-si1F6_vuQ{Th_RuvnBPX zJHNTLxueCdZwx)};;yz%;}WVoud%bO1z=w88lR#G9}k-F((|&o^mV^@J*Qrst(xN2TLBww#~R=MUPx zv)L{V3Ulo6j@^~dXY$$~%7eW_@>zmjKG>UucQygr=lV=H6;#NBUS1>l5*eOEuEg2$ zgJ$_;vVcE&`Rzz{*c*GY^*IQ6ZI1tQ8r9_Lb%9(-I~vNbd#RIv|Eltlz5?g|r@T`7 zCeL^!_K1~bSa4ZcFrRd<%HPqFcINwL@hBYHd*Kn--@-gmGMm|7>NRSpiyh(*$`%`aLj>zX~Ra&z3xf-ow%$mo73VAa(Dg%L6Uv}PH>a`^O^lNR}Ucy(_$z~%BE@Zs{FHB(@Fc2v3r&BsQD%ObN1E; z74qr6LgM+jXCmv{IPVR@lz~WCmcvzUwnJOwK${;_)Wg7$Av{l`dx%(UP^>7}fy=A% zz4o;62hmdKfG*A z_3emLTzwOews$7=BssUlDQ3^G!%i!BPZkq~)qI6-?kgwO<_o)A4S_GeW0uXn*H}Bv zA)IEhF@p>;H`t1ax%_vU&+tN7HI_7d$;z}M80BHzsrRA5eS^wIqeb$FH;D5af*t=< zd7e<5uk!Uo_f`EXGyBvst$$?|C(0OAXb9Vh@FBjZY3`a75VOzMeokMZMc%Y-pvI>= zgz;@O_BsdMv7jU&a#4^INp+(O0_)G;uaWxmMeT3R7;lT>w?zy@Qs7#R{Q3K2O&QC= z@tTDEJ=~o_bhFX&erF^)Y{b^=v*i=EHXG#Ej)eA(CWpT*J!#=POg%w_uf?FTBJprP z);mbVJ4N=(QF%E(t7W=E<=0uy^;ni#&ja|1-cd zh$6)zq{xLsT%y6rlKX=>zHZ}<@nbQcSFd)d8bMu_XI0L~s;Mm=_Da%1#MnH!S;2O} ziT_UV_k&jcLg}*cJX;QCbmgkvY*38_gdJ@=$~kejYD9{cGy&0V?$EH`r;&f`e(5WO z^HsjwH);UNzk#tfu#QH$lkR{<17@Aepa8W0zPQ;cnzfX(F$lmogA-T1tSd`C@3sZT zSB}ri!6jKWWac{mmZg~xYnI4ATTa?vvgCGC4Mgg{PkOBW_q}M}58aZ$qo@}6M%IlcqN{}i zNzoZ{NpKh?-;YiPPt-`|l#yIs-fao>d3mHaE1%(*r}A)bA%0{fUB=u+DiITAEB-fd z4K7Mrl!!a#Zc5rx`ABe>Mx2SA(`!U63-FUw7P*CyM? z*2~fccdAXHEH0?rXg;KJhlwh2ACpA|+Vkh{ubD63>>iaLNq)qZ2a=li?32Die!hIY zdq{pN`Aw_-g8qD8}f8hpn-J}gQaB;C(R-&o1D_cZM_WaTZTP#S2jByY*f<=tiX&Vtk$>X8 zdh2vwc95TR+}^QoL*0DmVJ}h|+<#*k-dK@lRji_8^b7}W4Tf71p>K>E3DKyj#4=ne{h+Nl=QTe6t za)B>*YySG2f9gK%=UVT)Q$c$lOQG0U35pIgGAv-t*Wtb&X|#!5#=LqU$Z+ecvZ5sz zTR{gCKlf!N9Gs76?Oj^0_K^rN_$G}O=kVx>q`M%J59?)Z zwtm6p_4i|T=fjjnDm&MKQ2xpCV}W8~@OG^H+}mGM3tzc-k?itD5QrM&;M1{ON>)?L z(qFuYr7K_SoBZ`BR`%O{*+udTZ?|0MKKjQJq`Rh;4|0z;4j(IbdwHY$fG+#i4aq=| zSJ}7D#&}aOBG0dzJjMf+`FnGI-9Fpt<)TrTWO)M90QFClD+xrPVK&*Fy-%fse_N$3 zSECR2L1X$IR=+5NX!32n!$IB;w+aivi>xImLi)-}Sd;Lv?boE0KNMGPdlTol>r9I0 z&%t60zronyYT>8TB9F&Kkl+#O0^5UXm3^zR*um4RCh+O4W|96~x0(}-uX0}{w0om` zl~GYpFYx8Q%90VXxQ;Cr%3pf22oQbQr*=`x7H;(<)g@^Z-d(>*Hz>!g9G#AjX3>VA zTKdCsiSHXiH}3UhU}y|dUgR!1 zr41_ASC%w{67!RDSI0cD8lN6t;S=gqzA%x6jTzx`q_S)Ekh_X)2HI~jkLFaqTA5dW zKCZQgs(f{IY#*|#(p*YAQD!Z!L(!LonEgz-{Y5(0??dD~`zoMPPgXL5w6 zUEI{md$U0WJR6iKE@J3NeI~Ky&ZSt{w)KlCNH;j%7AZw9ga#@c9%hmbj=yF;gk)eA zM!ihyEt&D466t`_yk93}ViEcb#BKjV$3u)C>r6GfW2uk-!+6xTE_~}_LqG+*gkdg$ylF>QpPe}%;2~ol6hcHzeaM0HzGIo6>N+plgFIk z7r+%%_@~MjGy55iN*T4O`BEltuPXQPuE;@L_!u`PNb-Y9Id_QR`WwRDsLDt$@O&o2 zUf|E!8&t3-V{wIdYSFYbz9qkn-ku8zoKRwxR)N(;x!sU+aiH@19zK59N16j7S#NLX z%gSQLp(ss-WNV%`lp@Ff{SvT->(b6lmVhytEdN)NfL-U0|NoH$Ox`$t00CHt%C0xd ziNT*^5-_BDmGAjn zv41Nsa*LmbFsjNAjPa~c`Ho+*+&}0k^ZG--MC|LE$gXmSkKZ{&tDaQ}PTm`++(Kbp z5>&_zO5<8#mog@%lix@oOr>=6`hODCJpgIw8A>>?h2M15%s@`(?C2McCu~IZjqYP0 z@ZLeQzh6hTv{H=O1suYw?RT1G}@S}wg2?% zK~NV+ElOIBCbFok!yLxrmL~blqK0lP};P@Z-UPINAE@6v2Rf@ zM|M&!5PV~k1Np-5#GnSwM&KZo7W0BWYA`rRF34&Wz7zy_t=N{|s5yL$-t5jF4=061 zo^tZ}ia{gl16tG#16ss8qS&xV;a#_DxUCMwbpbWE(}JVTK6Wsuw1ckCO_9e8YQKMF zMk@CsDff?-iv2-le8~2X-OG$w->&a-wIIz3Be^Jzv9ROjxI+?Oon&G>F!y7b*h}Qu z-ZGtt9i`^ok`{PjLXx==&`q~}CP##tlJOzTUpblOhm(R8sVe*k{XeG)$EaXO4C)aw zj6jgXJyy__d7!hmu7dHa@DoV~+rRHhnHiI>a!CkY5Jz}?6$rO8Jx<=5W(=91grHpV zD(T*TPR{S2;G9joDu3(5hhai}BqKKW2Wf!=3fAwM^8HK%?_m+aV}fv7f9NWhkg)Ga zQ*1qOu$m-c_ilT6{<(3AqZz|V#op~Mqa`0e@V;%z_X9w7$sMkO7_veYb_?@fm#@Q$ zwvOqUU@dEDnqKhqX)3Qxw~OBLH7S%knwKtXZt9$_LfMqUcmrW~a3G7X70rzuO%_Co zI1z0f=Do=y!-0;L=eCZ{V$1UOWySN`PCf0^ICU+_OS@W4`Yf-kITby zQ}5z|LrJq0XfDMc85bS{&s*8Jtjk)in9%L&#AAi$tytCA-tJr}xMC)0E_Q_28skzR z@;96vH<{et_5{{A5B)jBK=HlP^M2%TFA#YUvJe*tt2g+uz0%esM@4*lLl6?66zlfk zDoy7_ZvJ1k^dXLuGS`7yVx!`x9+fOd}X<8M=aLTc6y(?!sD6X4P?(0~d zX~apuqTlH7e2sFtaRDd9cn7}ZXK78*VKauY0VPV#oIEz9a&dtT-I##xEQy_2PIKm} z;ZA`$jq2qsBLw5hX(VsR=j7t7^Oz+VrW`VEq5{`GItIzt=Hw!x;-IZXz86;4Zj*}& z=t_h{NOTcPR;en%UWgJq)2J32rA->mH!XvS=jUrA|5lb$I1bxkHiCzfe<^bnYojbL zGygeRm1zL?2t9jUwQC|)UF#$`x|mK2EGNn(zZ`t{^f z(~AYdxt+o{Wb!pyA1+!vpGl`Xz2yXs_MIsl<1ri(lxkQ8kMGTlitZ zyK;X6xr8y7b*R2VH1g;w2_#DuE}?=P4Jj_ny+5t&CWKPD*+PhN#-YL-i-{veUb<5N zc1)D{5oI>he08Hn^T`=YF&I&phz5-YOvnJ)D?)M3g7Io0gg!JTT`f3782o z(8EaU7uVP1>M-_o^^np4;>WLbD(93fe}hKXg)&g#@6%S-qvycDXTFPJ*CvTf}iEhD^@g~KP}t0*!J*1OiElE48j!m z$k%;N*;$s$XzJAS12?sy-k@WLdZ20@-UM}GMnyF{x9wgMf*i$G&U88b0X(~fq zW%npUU0Hhy<=U=M8Finp?JB!Rxwb27PoZ4bH7dVwpRemGyGOaMD{D`oT;DY+KX;$6 z?<%`TxxOoFPoWHVjmll_^Ke($J<4!b)}HK)F2h}jI51U+H5U!PO1rn+bx-T%zg#8T zyO8gNR(lklXYcY~GguBX(uIOMbXh~$-c?ZfiTiY8{OKpu5?W5V!+rX8{OJz&iJbC7 z_vw4_rysgc3|QhU*_;+@ZKV{k{zu!61S5p%a$drp(;8PahM<8{cblMw@+>k4 z8-EYE=` zPCk*ztNeFwR_k-}u}of`U@y}g!71gb`1!*|I|wJDM3*Rzp|UhW4YNgC{v7NrfSxz| z;g>&GR@?p|y<>jN_IkfhLH;#E$kqV=L1ALD-e)CwlYhjhQ0lD%|5V0`p{R1k>w*5% zk``s{m`atwPtf;9km5W;8K%k_?zVFHXk&F@Vef_GeJwz5eyr>}zOQfM(x~d@i z!@LZjt=UJ`H(P5<-KcI{RZ#hORbJ)sl?D1?sP7-Uk2X=PzeCnn?N@nb;;TQGdmgOQ zF7sUOf`~%dTs5xa^4!X7V!yHN*|gJNtKc@mL1(x1tn2U?@fch+rt%EqbNQ-?fe2?(dxJ*ua5$p!iHT42IUNJJ;-GOY%AU^DaL0Qt z?fR2uNwCu$Vm`9y^c;}ft}p=?+)CLcYxr1InnzDDwK*j@SUin7Bcs0v%WwG4k&loxcj_R9y> zq#f9?KLuc{@~{~b06eC7x!M76&A}!4sRLld0q|cAfRC;jk|!-+IV+zAl1E}lK5LLX z_-7z_!XWv`niPxXy^un3Zxs{@;b4RW)ajrjvoQ$e$<-X{v>{hl&p-KqDH=oBgHQek z@gRJLr$L_f^0bMk=Xu)5(+fP6c^c>GlFB?pG$&829+fLA^U4A@v8!R6zfjhV`JL{B zyX~bE#BZS&hWIYgn)s9zR>K}f+5~4uT;TP(fIn1k?-LKz;|G+dELQ`%&^Mm?9d?(nsQFQU^*pvAd}ake71K?? z;0Dif>zb_G+&6?x%BD(Q3$oiQseIlQ-GO27EtY3&YM#7hk=@>)JY11i`OMm^e7=v; z98-4o4gtPw?1N8`moQaS`M$SXBl%WuR>r(M(8=wYQYhc=&C1Q0lHAmnmD{|0{GM02 zr4PUKOZo~bPnj&^FP)vu(4!6hE2c+Y^hUG@$D6R%4+rV={F2)>!3ri7zM`+3cas|m z-3XiDHBxE!wu-dm6>m7q{IO6OjAL*{^DlYC?bV2Pc@i^&EBl5Ndvuj6`$l2aZSN+Y zr%U}9-ybLE+wv-BXH`DdS2im?4Ek)Fk;mck*}=~qH49Yv4I7QTEiaGv+KXI;S>>-u zRk_LruzPfZKhp5pgv;4hnRei6I$(qLxY>&n+^Z;^1(k!0-Rp#}2(dM#XO9?%GIfvo zG%cxo#4uO>&}-BF3B>;XTzLwl#zr;5>eD_wOgo``p|9{~xj7I;i4dP`d!g)MfY8Vs zh|8nnGH+1+_o$9Du!(SF-oa^KAAbXkvzV38q%Ed6oA}8U(EB-Z$yJZxTre5vvXRAo_4mHjx2rsZK}WGlxQl`mN`j z?L+#(=Ecn-p{yyJ?nfDhkc&FtR1RVT=D;q0m6xVMRykgS8w<=LO@e$VVVcA7Mr3V-r$?%{UPG~qI2tr+SnNc4Fh^_080qUc|AQ^7$MSX3Q=m>N>nS73 zd~;qF6&PY^cjL*hhH_ZxJgT_1-Y9?Qq-fTS@&;K|mNjKvtf*z2IbBYpdvl>wVERkl z4$!s&91L;vF`CL^U%j%d{nDKpoXtHV9lYPz;K__Z1gSEAx61#M)jG-MYn14@4xbtb z)_XEZ=VTFw87;J>5lVo#i-h`I-pF*F`LAA_&Bpw;w9rF2c3g9Ec9x}GH7>o`I>bVD zj6fB}whH4@uj{1?iw$ED6zOPDD94PeGv7HEJd`-C69t`=vdlnLFmBqw!d^p;DbY9?55Wvhs-YgxTJc zm5U*NkNA1{bWc|0QGA1N^_55c-EwnJ){(zQ{k)|j`jJP`SxeuL-j^(I4c8|GYvG&H zc7HaLcCkv1jA!F$!5iZm5^&K!N&3c;Q|?)Bb_Hkj2fqBxp!_W;Nmkl4L;OGgAbq3Q z+YD(mf_6_%0P!hLlYc?^G>`O-GVJC@m8f`th$TCPEx)nnsZha}?MXLXw^!si0eOSl+Sl{0;@$L4Pf7>?k?MHlrhr!I|rI?@? zX?K)2GoFYvIrUF|4`*s|s0L?3Hg{<) z*h2BD{OW&Af$c|LPrO(*JV0QSXfg`fAYyUHj)?*T6tLFTn?J)3Oz_g>S zYkEGyv7B@oUe;~XQaP_!v!bbS*|Gz!LIaky_i{08W2w~9oZe*Wg7n4tUZE0Rv!YmP z>{!)e5ri||%cQ$2)|eugxI09$M7djRTGqH?g>y5WlD@oaS!YXA;|fzTXFLbq%W}uv ziGf-Biq>s7`6^zLx7gL{!aE0^=Xq8B!oPC#I0)ouV>}hpqbf9HK|uT-qSj>yFA*qm zde{YH31@1Azd)D=`GYE1OSFz0>X4vnK@gM?Zn1I@0L|H7Gk{4b$2tuyP*+&9WTduc z(|lH{v9FPajR@VsZFcKgWJoNnqgs^1#y6HXuHUd>qxQ?Xlw+KGd9~fQ6$3UzfLP~G^=0HzD5`oqjeF4XPLa72OHb~^F=BHfCvU% zSO*&00@@#(7mk+9{z7gE=g^EZb^;)NAiwOzK9}4ek6dN&Xle2KjnYxpu%GCG?1EmE z|L_SZSRZGtYCltOIi|aqTRr0qa#<5>C9U!k*sV1|AkTYw=^Gl{mTNTU6`UJjHX`MW zRTLfHu#qwBOUJO`ZE0IyqAjBf&^2f@lq7`oz?4}(>KW5x_oAUB7q(OWT-nq&rV*fx zdHGz93+^<8pM@_!7@1Ni7o;tobyC_Q&WHUB`k~O%%YrKTK`*zg2P)sma8Xm0a@Gz4 zxy2B$O65P34BsaM?64e)>|t#o7N$CtYi_{JVx-{!l|X90CLRYg3SX>SP^dEs>BPjL zbd+EH+giT8lxr|*9tx)PA^tFaX@gYl-XMLWvZUa+&KR4#c}d4qhW*0~eNnUWGr-jv z7PXj8{JsazhCt=R{$Nh-=*`M@bB9O8R@4Srgt7i*c1zNql8#@Vbo?qG{+t+*jIz(d ztWtS0sq#vy{AF)euF90OU)dPs3FngbF~1a4*+uqDR*f1gR)3(f#i~7zqmXsEXs43Y zA9>HIN$2D?v{2fwaQW;v^sfOa(CiW$-UTSyLzaZqKbipcK#aEXQ<{`-_KvFjD3h0O znsh&z2>D5}-nL=U0g_>9*{?El5SYHx$EB;UHa7b}3YEVH=j4Bx^gM6R|G+bqu_uBf zk{^@{%5qdm`i2&`1)el0pYI(~*WG?K8SvMftmrp<;<@l&*p!pNA@ z$JIS7z{e6q`!v!A5Qfu|U9lX_)cNomVRbzvLite77#aSM6!}O}MtNFI&sIl4`z|RCBLYvqU2~WL(bAGKpbc4jN<|kqeDpar$O?SjVITsd$=p zGO;WEeN1MR7RkD-EN2x@+=Se)ARf*NIcj24$!WAvi?xZY;;xBI@t(=Rtt7??~&)P0FUx^<;_((3Xiz@0iF`R~v&B=koq8t?ECG z&+T`>PpzwcBG(<0axJJyA&aYfILSCeVvj3^Vu?mIH)sggzC@$K0u6T=r7@H0#X0-J zAL$!qeZm3tkLM(VCJJudcKUeN!Y8J zjZKTCyHFs2Uu;jd_4IfcX2@Sp4AvyLg<~&aKK#6fcLQZTqIUpEAOtWBWyhpy7xXgF{11yM4(H28YH^@x|FPBae6AnJ#Fmf1=dlxoT=S z_&ZrOs!^CF%eiTm6ZUf!JE#P=(l4vB$(d~Hzr-*e?3T_Zb4cB$wea2`w{jnVof-+z zld~03+W>+TIW9IDdFK^kU9``(8q(h<@2$6KgZ(s#Ei|z;3oCqRsdn+Ajf)m7B3O26 z@_*OArXZd8Kd%ayiv%ZQ)RbOBPPQy#>JNaR1~E|c&DQkVEcSn)=`jzmh}tGlw>mF< z1?etqEG=FXqi<}6wBzg;@+vtrwvu5}f`9Xm-3QxxG+{Mp)1R{!8S=qS?Y85bxL~DV zQOhaXf7goxki!JX8awY?uhI4fC`jxau;=}0!FkqV5q$1fhhb)CVYqgnr)_Dzmp=h2 z0tEjlq{Lj6ZijTQ%HLLW?ak;{Zdk8y;{9BZ;a1oRm172jj5L-tO0KnNXn0wH6JfI{ zaCVjgCsN>?Nd>AKUshmFQlQxi%%Q+i3Y_J}ZrjTX9G4XM4=ZpS1qzc2oVDa-E&Lq? z>S}`waSAI`{*Dswq(polyVvxx5nul}k1{It}t5ioxqWUSvOO71`YJC<)NX1<U>iP5PrXz$gIm5IBF(&`O zcK`ggQuC6oY2B%{VomF`#Jl7_rLnVZdCRnlr64Y8S<&3oIi2xHi^jd1*3iUr;I34d zPOEXj(lk9iI@Yvzww=;Ay;9As@hxK0y>!hrH=S-F5}!`~=}G5f&y{H*pp_w#;dH=S zx_nwFugy*E$h_Q=4pXO)YPtyWz*^Wfds^H58WR^Cr%re{-Ro^17?cV z?aj-lQ_l6_l#UgRuc1^*ES9vKzr1Z)I2E@Gxt^%T-32-g5qexpb5rRx!QZ@WSxb9o z%d}|gM3d86+HtXo5o;s%<#hTAp1l@oyyAk!Y27UuL)u(AZT|dIr?r=^aPd5+)2S3~ zFKAvpopLKWJDS>BS5Bu$+^=b3N>^)3Q(LL|l&;RDx#_evsobfx(vyEu)t+S1wlT6luxrD@&zokABQ<8(bUD22+5B1v zL|Oy8ZC7W@v^JdsLvhN|md4jMRnU%UhN({B?`mDqxU~6#=EhQU#{spEZ7t?5`f}Z4 zdt=86=M9?iyqKGe6?$#0O{T7Vl>o?r=Q+le?CN=)jZ55(2bG@J-n_h6YPRv=YsT{& zx5-O5`K4;|%M6r|ZApoh%^fXE6Fn$95QQ_I2Sdo?L7ry~m=(jpUTWF!2G47;v_m|v zskx)G*xuHbXfRt^t@cAbuhm_(dZ_0$cElIO9y$@-xZ2~6I$Eioe0?4-6Bot0T1qV| z$#P_4aO9QXGANEkfG93VMBDdn}Mk+=GL}Ow-3_b zqzO18i49HqkrNYhq|HBxM|ocRMNKPgy3jDMXe4z`I{I&P-;=1>9jlgib-Jr=gIC-& zetw(xCeK^Z+?3R2>;%bg_Pm5e;Pq$^q;6bpn((7N^4a(u!%R-<$y+>cB22f9vI}{A zkY{nWt@vEeskrThskxxhnm^X_+MU9Fj_1XDFm3;p=Pl`KSypl@ih-c}D^H!z$2p$2 z+IDMRfTqRi-j7Mf(OP*cv$CSC=^}C+WHo5+Bo?CNJUI01O^V+jHyE|y!#N%o1u?S(U zwU0?kOY6$!4qFqX%y;hfl#Cza)R&c2OLla%a-;!*!1G=;N_*DxCX-pR*Gvk7 zZe2Ug-~)vC5fDy0?d8CeG6i^$q1mSPXfJ*#GEg*fw6mnR9FDZld3SK`Bs zT-fg%N8wnUApgKM zxfa5aAdGAHUB`VTHH!Cfr1B31A;$Q-vr<)Y=;5WZeFE{s< z`RX74+$MBZh)f3-+ehggLpkuBzA}Sp(;drTxRH}#d}HZSiV_u?kZcn7hp z$Z4%EKmv++#Y!_G@lVEx!(TM<8;`n3M*2q4g`S*G1nLj}YdVzI6Z;LN4ax`ZV!Bv# zKe~h1z=NjDPOgaG98zI^=4<44?Z;*-Jd6ogEz8I7HztVEN)(I5MTx!Jw)Z7%DS_q? z6Z)Z(`t3+Q`EoD1vw`5q`4)S4yB32ZRc^s6JUB|_w+I-6Bjw|LSze#=^2s%J2px}9 z=judUmjARsui+5UkLsw0m`8;DM z-Y>nWDyIGYDV)D&IAZlnxG_~BPGorIfEms@H?3hGJw+|DQzQRqIe9<9nL#>!8c?}a z><-S5Z=1!+aNx^rtFplvaue?iRc>2wXzsYbZOIKgO6Z-(%&x+9!+-U!~OKPd~5<5&zSj2-QzfqwN&c`7H0E z#RA2D$At)$H#r!kCtIhBP`dD$Lc3JyR~F@` zXpThGgc?WXpNEq0Ar0rHBTqC%Op#7kxx6Z`^)H|}g9OynQ}3RYbOd2^m}%DbBk%iAI8;Qhq2?pki==9BGk3T1z`a)AaBg8yD37BL%-RRt>JY^ zE-1*?d$aOx8WJ-a-w5%X&l>Pmc4l@%a)W|w<6_WIZqE#&bK{k?_2_fM4PtfqeYEzf0W+MYF(=Q6}j*5RO_p zw&wq2&D*nv@~6xoh6@oLlJcw9H)h>gZ}puhV^*KLFLuo8vu6$EC4j?g!h(f>E{EkM zGvW}CS^qdB7MI)ALu~qa`BP;+luK7<<;BW8e^ms^F5LyGn1qRunO5K>aZp}nkD8-E5lVm(LU-y`=B+5uX zB>rabS6G*UuAlIC2Y)4xvt)U(DxJWLM}#OECmao)18Gd%2B1DkTj?=^*sP<#9<#qs=B;3TTxH;VB z%m7{9KkDC{_3(%Lmg!w?S)Z)62g{#aVfMPE)y42%iN?U)P-t zPMh?;-2&%-hyam%e#R;dM_uQ)B*}YRvh-m$=%}n{D4Ub-Po>`zUc6yl3d}bEMu)Wl zo7l_DS|`NmK7=ne_txXwG2yK4w9KtGfWx+~3}tNqgVSVK@;=a9&>u-ra!5xs)+B zUMy@3W;JQl1UTKCM#SD)jI|azjj3LW9$j&s#PJ;B;8`j8BQpr73VkFtx`$}(2seeJbRem-H zz3!Z;8GakH0g~(9FAn=ct4`%iU z?~y%b3ID-N-aT9N*j=^T>{eRFco)UwE4>6`?J?X5wesIrFg2)UyR_vuAG@yYp%|@KR^V*SCEhWnk?h| zA`%Ha)l0@21u7r#LEH}{E5e*DU4|it$_L7z1HqA8kMD9^YE@j4R55P2AGy=ti!kVZ&^C z3?QO%NiX*y1S*&KwvkvJi*m~GCDL3lOCC!)aCN2>%#z>23u-MMK@|H~CLhcqyntcR z$1=MYFmLkT4oQ{VlZD$WVgI)Me^>_D7rK!Pb@;S@* zOUw5~j(_C4-{pgtz~?bC@<>liYp>!8eH*J^J3LhRNSgFilJr56xX&YY&D(N$3h#q9 zb^|F-S#%BjGAa6bD~dHI9RkQQfH8rf14JYo2t-DE;z|9g6%hBkYbe<}AHE`O@)|Qc z9#-7BY&(I}QhPF7|CKQoZhLQ*t@Y=2It^5=?=?mjJg@_&WjAm`*^;VQ$3S0cR^1DBA4r3^Q1fOGKAeX(v)-TXS5c(L2)h{}se6Vg}6uQ#eV zQ{_&>UNC!>%1^or!A$vv7=He^I~&ZD5dnjL(p|s~n?{0}vQ0)9x3BdUWKA}i`@24p zcJ+sGSD`3Hjcx#prxP$PlieH%DoZB!iQq|O9Re3-iXVnQoE7-SD+I*5aG19*mlUrPq_XC!GT%d-Btjuor^WD4<>F0ht=;$Hc0FOTG0W{t?~ zL76qEat+ctd2?Q7mE>)M+|q^Y8y6D!2A^C&Bv$Qd8pqe*cydD@QhK`MqHOTQJ`cp3 z5=Tc+3Q4fa?-AIj%w`+5-*`RAtic$H&D@3;=`IlQ0r*3e%S=3MDbKi+M=TJ7rCbx3 zJ3tVG^3b}`k~-g*L5=viFZOT%f0zL~C&;Y*aa$(dKPZp+^to~_cG$xj`6tU`-hR2U zufXMN1fD%LFH6fr_29e>MG9uh_idEjSr4T9-8#->wBw&^;n-|aKGN;>nNXf0e=yTO zSw3aADY=py`+$5?_h@j2{MKFQ{&A{8(D}r=vRqmPcu^3@HS2KuW@t3J?>tUww)T=X z*P=_IkAg7Rf1fNHbG@XK%6xL+PLi9u_6)+u*Ch~sEP#~;!lwnR|EUmuwtF-oI%(L8HstSdbYGP?wl zq0qs(BWqDV&PYj>_tkWbe7NGU8N>LMvJEn#Av`%w2$AlTU(k13xRaN7y+;02(OcmW=c7N?qnxsY=FHF zCTv4y(A9O&pv^OE7m-w4nSs8n6Pp4WqbsK}BWx|ZoZg8<*Av2=aGg;U?m{r{YW=Ba zTB;(eV6XgroceQ*w;NwpJ8X9@};@vT0q+ zaSgb9r*XKJRy-k1fOxA%##t&itYfU<+(qSqgijl#6e-sx?cGU=8lfV${V-hd&>JSe zU_AzpCRbU*{`TvwKm?d9|cNMu1VnDouqs|N!ev7vsCU}#}S|95rB&clz@ z7sro`Vwd_5+6M~Cin3qhcJ)iXn|AO<3!f+M|lluVLvLxl%BxOUAvOh_=loV%BKyBqUaoJ~W91#Ip`j#a9 zNz%#r(In+JNy;aZlwXlDQRnB@#HX=cE**MJ4c8|5?{w*00!sQfll(uJn*RMH|Ba-R z^Nu8CTat2TlJfZ^Wp|SD=_KWmB;^xH%5RdCE0dHbl9Uf6DZfurE>BXPPEs}|DbFV< zmy#k^Wp;CWAbcOw$Ap6jOuqjPLrz{vhLy(;6NaKeBmV=Lee-dqwhaDBj_Z+MCzE!i zo3wCugTgRyvY>$kv#2G!&rRH&NttV1nQq%G@9$0J*js;^Qi(qiqf`kN#=>BLZ(SWv z-OX0|KFm}iU7wtz8-5m-J=8gi?^L4i|F_4!uPWc|%bE&5I8++5L8X|KEvNom=@_fS z%~ZJ^sz|tzqWW{k5>LNN{wuBaO^40s4|3dE8dO>U&>*Lw95Y@Q94goK75tndb#w1b zOT6W<8U4GUHgDI^Vm0^UjXKmv%r*unWaZ%H*dIzyHt~Y=?n)~-7RS$>T1@J0NItz? z>>}CirbJ-YSc~ijjD3jrKa|_nE@SuBlJ*C0S2=f-sCh5wxEu2@XvY?8hQ5f{C<~oj%gIWsN-7JabjkN76L82R>xIN9<-Vmw@4cr>OqdN1?rzm z4I!tBP|R7A7LXIL-5p$V%qg*CIk1ngui^pB*o3l#ZQ$)HN5Yi($I3zb+5hN?e>5O* z3TEr`=e_{bVQ?QOHxOt=Bl;B);g29I+!{$M45N9_#t^L@OE&r)_r-Jjb~*WFb36C` z^ppAfPMY>hCO#eWvQiIEe5!icr-cus6+Z19Hlu$aN1u>bah-nbIlDEk6m-3LyxQKWFb^rP4@4 zROe{TtmOucq@yes6;%GFgfSzGE)!uBnqk$33#76n9ff$kTlAMm2_D2YBOFCEJWPwW zG>e>LY%>p~DJ$RZD{HZyE^gKu_P8W+Guh}1kOLuEu#hZBS5^+mZ>!DK0vm$>1#`eo zuz+MwictkWi|aL#Vj;(M>F13qq&8CErq3q%A;u~a!@tWj6!Nt`F0--EtrOVu+hpqG zSMdQLr;(ZO^ap3i2l}$Xg~)li7+LQ2OMK;LQJwnl*;_|k5C6n{^QCVnPei49+A~*v z_dArnG3m?Z(U<_9odg;+lGS4}159)&wy2={qT`qWs2l`Qw{Q=k$OxtqnG!}XPO#gk zQE;Kk6+L4O2G}ltP>a%=O{9S#T7)rAwfuaIllVi;hnmy~ zIQEsSWB+9F_p2cXAxaa$^q!~Es;`QxmLQ)aw$ox7_+(FEQUgz|!RTR;%2N;vzTYH+ zyx-QH1?4{C?0CRWPII71F&CU6bMnC=x!uZCt9(JmHKNpG-2{y~DU_X3qOzq~%dSBy%In7TyV(o(JIy}(;SN7-s{BY2W+smZhLTXGb{7aEB4!;A$ugUO=QoMOu# zQz)J_$?iV zk6T~akF`Tw)5iF_4&SnJV(f0VvHSi?Gdea_E}ZR*-ZF!-mI?Sq)nHI9ql^g+sNCdg z`Qgeej2`d~Pqg`;E3*>jbFy}5y*0>q-M%tgOT?R%6T|hTN~UagJXYUktmMZl3%UTR zd%Z<4Wa;>)itL96BW%8ZNk-l~<_}~XzkSzpDWYCy3WH&r{%|(KjU?@Km4_h2LA7iR z^1k!=JUGF01H3)%<#SqR z>nOx@5SlM+j}5H#I_QWq1Cb}zWwoDHO>%t?qMRD10~;WWF<+kR$-@2g`+50hW_M66 z-|Wt^>4IL4zw*0N4wMIYnMp~~MI`69?yle1pzaP?8YfpwEbNL1`e#Jlk#^$9TWBAe z-i^8%#td+FSkD(VVm24#>ozoMc(3LV0z|UY8WnG*G{4e$ZWZDg$z?=MFo*?e9Uq z_igC3P1CFLjV26Z@@J;(nmmNQ^v@TgvG1wzo8Rotf`h>VppI*GIb+;l`-%-|NOVp< zNj0-2?L$tEO<8}gG>s)&l&5KjpF{zH!PyNO>1LB0HJWFKbmE}0b8Qx~`9+)KBUBz@ zlI61%d0rk|%Rp|eCISx*z1d!Qq8Q3pApV9j=A% z(t z$2_2LIYVXFT2K*#_-d7}fo3X>Qxa$d!-jy10F-@W*jvN+M=Em$EubhFLRmA4V_!u{ zISt`c@kxwk& z_>_%=0tzCM2a}4N9bn)=u>rMEpITH|o|T4z-RB0WZx~aV3v@6IE{OTd>yUpTC5+@$ zqekvPZ9OTrh?Is*afI${Q7BKXf_{7?%tJW-Osl{;r7a*2 z7RghqpdrsxnxoVSrC?Dq_o3{q%*#>C|0AojTFZQXA!D=uKrcr{=5!q5>OQ+Rt07DtR~ALV z8SLjVTpl6Ovn8ijW&&51}2+0SB`ka@G|tl8S)jY0~qIn%a@p{s9Ijs ze7dzBn-2Vs8F<<;v++iaYU5_ z&KAnDzrj%Dy+}$Ub;_Fqj8ieEax`O@lRw&yMse9<3>5T^okldTvI3+*4b8D%<@aVN z7`c0hD)O=-D}Jfg#)ujaIPsS~9mhd*-ywr2|AHKN3Xb5x5UH$~Fl9O}9hj#rys5i7 zIb?ubIM7a&V_#+gEUk2Ay*qPDIOiNxaEjaqTno*FZT`wGv$FyNqWnPndY8} zvLC|K5IjPPauzOhz2#u0Jd-J34IcK5qXD)rpf%V^}>q2zlRFc`t0(PzGIMrzB7m??i+n{^6T zw(boTgTYLBerOVWaJt;vUEuX=lE?2R?#Sm8 z2swt(kDx*M%%vueOk;!oG${^LHWxFShN`ILtS!ogwr^Y?H&*eN_WfX3XoY%(C zi~@~ZCwVjWEV4{WuC1$+v3)Eaw5#2qJq$>$)k%Z)a0Y~K3}(qyNmqUym%GZ9`|qUO zHIvIdnwI-`Qtr2LxsSVYzo8uTjy69wx$I+B*0A$D+q7cTBO_Ml*3nxjjsZt+fqviS zURAz3xu#pJCRttfmpS%w%Sj2&sr*h3Qw(aJHXG)dDwq4nIcCYzJyVeFG9PR6JxQ() z#<}*mqwN(O0)gz=i6a$NQReqanGeNfe(%a`Ny^YtEP_}0pp|(lDf5xI%u}w+hbaRO zMq5u!Zp%We`+@7=wv&?%%7+7|{Gg9NOKQC#uJvbA`?$^8{z;N+B+m7do-DU(SLK6B zW1ot_WVK^nzLv=cGvy1uVW-NSJ(Do!2CMQ%NtNG>tNf9x@(8N@pxrqzwaOc<%Bzwp ze>SVgm zrxn~_1!Y}1VW_L;rDOa#JsToG(8ztkZu!c(Y;czOxMEGLB?sBU zm;Bf!e8*{N`RD9NFsxsH1)WE=CVyK}`FGe81LQunJ2_;K4MYwd=V%Uq;Vdo6{cE#w zT}2)>(8_XsFk2p9gRbjJ%_B3Iuf<@t%KkNo3TNWHt0hPo%beXjl!NoKvRqHd0JR$% zF%s{1R99~bP`P3a#%VKE{(w28-Oy1)HBeUU z*PPy;o#`JGu%R!|_pelNtP4bCV^V%!THdITI9ynQS(mIHWt)9CHT(>LutC02;ZC{7 z@1tIDYi=d!gg zNV8$D*{}VUEtIdqT*8vZUVGY&F18p^z9RG#V?4Q9%fem-6) z8RfkW0X%s*ut*RYtt z>b*0qGc?2cKVT|2CT%5TcNo>0nZ=(iOut>-UmUb?L1{0pWPcW%Y&2y^V^ z8@oE&&TnpQ?r7|6F8v)5iYCKB*rJT*t!ixPJinu@%fgKL#Pn*lD0KcrWGV5yX)B} z$=y)y1anWF`e*MAR5rtuVv$`>k*DERl>>Ia|IWQMp#G_3TR$nAJl|VDmU=eyT`o@9 zf*lv7z~1FSICtuJvbg3gOTSsn9^me=a1y4a^-tcX5oxx&nR+-7tpOw-^Gm$>%)`cs zw6`g@q5vp6CpwHM_foFb2vU8Z@rC9796BYJ^%Z1wIl=e#))bJ(4aj~gl?&8)Z~ArO z@2B1BJYqYGId4|UweX&BOmUl=NG8gU-8VIH*W}#xw1FpSAW(VA8z+|o=c9dvWSQ@X zv)cNAR`_RNp!+rVz@!M9srZH)6TQ;qjs4Z=$ES+9U(~8xej(0rIEWc@mBz`rFv&$K$GeR9M zX5e&+q`)s()A^bMy}rhXxoO)ioo&+!BEDi-OVczXS}&g#0ewxOl+*6CHg|S*G&W6Z z9#*tZ@5T3S0#JK2yzg9WHuT@U)YoX_rS{f}{C)bIlp-MBJS883+w5+AM zwKKf~G&ALb(khpwt#xUO#f-J`9gVA!Xs7lb1KN^&D7Lq)XjvVn#;ZQ9e}?C+?OfB| zydsWRTWsuD?jnTF@I0zzPNvk5!Wh@+Mm()>Y7pduyfg@MC81Lx@svAwQnceYOb!uy zh+Tt9mP0-7g0?mb;+pZirKKrAEG?xCR@=vv19q=mp?aZJQ*9NpqaCP+R_(H7#h4b+ zuB%`vynS3^BDJl(xwX{N@rv|xgeSGOWQ8&645QmosfM?4Gd+(oaaxNlhFK}9 z&v0KSx^^fbX&WaXJ=F8oHg~iYo7&pf6jyY1Er~0lM-!zu!WjH1Xn@@`8R?nj$zRvT z;}wtc6pS=N!lk@4MnL3=fX9`dx4f|_PG4$cU+sBK=eNc6#9v65G=j-%F>W-@)!IBc zg!63Ai$R{!u_SF$b<5k6#AHM=p4U~f?oKHir^U@t!z;!uNsRNb?rh;tZqK-wDHXdE z%EslAIc5<}Mv)BbYelD_0(7Ibv)M*ALEI^6uXbbdVM(mzDKNA)cQUSROB1Tx+T3}5 z+vIl_`Z;5PrdK4B+1}EgJg!*Q+-&$?;d$+CE;x@b~Ad2MU&JU~jiu;W5)?VT;{ewKcD9PLMQ46 zWXc`p8B@{h@Qv&-GR^V4wLCYru36gEYBMs&^O}|$TyJEnax3FaMi7lbqP{nKUUTaN zEn{F?jWk0Y*S(sF1*;V=bxXKS%m4v|XkCn{+K;(tq6hz|b900tef*zjfI;rs1hP~`#pI}q2p z&uSf*TI-{KL9Gm8JcRM2ToG6KN2~IRsa5WNO;vt0uJTE%@}o(WfyliR^-avP77e0H zlS%y~a=zejrggyI??i_tnzG9%Q4YNNgY=E!j54Ir2%7!iALWZ^D$RH-n4_Wm5&efV zK$Ev&iQ>=2YNd#qw#wheDgTDq$bFbXofN11tEE_^T#78_)inC@CB)jK?Xut;5zCtJ@KT;IEw*U%!fJ<3s5!`|$;j?LBFpnTPq+PP25$MB=P8 zamz&FQEB4VB=Oy%Y2vmdvC_Gk4x70YU0oD+(La@$C8Z%w`9J+vIv6F>wB^SAH@kj5 zd`aASU(oY8s}#-YMzg~Tb9gky`r7snq*)KeaLrlVS+r2*g}Ln4D2Ee`O_pgQ{o`KH zXmh;^3OR>`qH?njR#8GOwGRzigq*#T@OC?rkG3kVK^jh(j%=>ZQf8FM`o%taa~XbX zxm>PhTSLQnRw3$7RL~ff7O^et!W^KuD3SNhU7z;ic;w@o%`Ve!C_Q#B7>(f!?%8u+ z%H>#Skq+l_!5ctc^t!%Ue8kAQ2V7rDzG)fl7B@^jCM@l}R zk-ui|V*fpRH-dkA_>-ukFhQbJX@=Q&g_R`Du>2z&Sabj+yB$wDRjI zZ;{{asRttmxxjv};Ww%4HIh5M5!vdh{hcWm0V=LauED9}A&;=t8cA1~oQ*ZPh6a@f zJwRqL?8G!~2~X=WDp(ZE!J2?~nSuQ&xZMC;7iZ~NFvqM7Uh!SpvP$rwjqacJ*ZT1i z-g-qkSJ%YekeHkbL>wKjPi&lQF8EMJ{+o5lm#+tGnDWDQ%))oYnEhk%*b8OtXrdQ7YfB2=H!&=f7@cvwHzDsYkLFd;o*s-1t6_dC*?-@#r zZ+3hr1@z|*P;jTr?FhhTVX#QC$XF>33}i+Y(1&O7<|NAt3qVgqsyFV~O`2o}0u^2XKUC!BR zOU2+ey^cxXCug_KCY2wO#QS&qu!ai$nTu`kND9uBJdo#HI54}a9qTBpJJ|^B z?hC>WI0Jw>YA`FoDee%BGpXH-w z=}`>M6t1F)H0NBh+>o*26vrgeg=NX0?WC~TJNXMiJpUac`x`Y{q-JBq{2BaeYyM{3 z4na%+`hE`v`(pbPfgUOfw z7)mPI`o*{j;+Vpe^)o0o-9Q{Z0JBgSRA!N&B?(e({Zd+v?olg~aJb5k{d}M@!qF!< zTz=vYV%PAEp0VI?`I*mwOGCM#XNdSw8xod(>zC8o9sxzA1>@>Ul_vhGx+UhUB-z?+ zc3^p9kLX||xHKal>dlh(_jb$xdKQ?iGJAkv*~3oE8u(7R!!0aCQTvGz{G*=O2=sm| z*8F4lOJBjCv)_gU-xxRYAZ4=&B9JZ-3q1&D_7Q@=!KiASySlT4nMTY9pA!kb$5uRDJd zRrx^%h@sYoAG5T0H@2FLLL}8A#;sHgnyW!5-5d*LwXK>{Ayari%C|YtnwLyuq*DF_ z&*Yyd-}m+tzY`kZd@lStobr~yi(})dDp&X;auR_rG?e#Ry~cOYktSkSLS&i(mA#p~ zCAj&!G&4w_`q?}K<<1x1BJ^;4l1fKwS)GQYvd&^4MZ}{F@xQ=HkEO2i{_G7ZCzRkr zWgV@%zRBV6d~zheZB&?n?HUcmSS39}te42VIb?+H16+R}CPx&IcQZRV4l7)FQJ+^r zkxPAE-d(2G-|_Q7lZ?mU%ZnBYr6Fh{;ssVdQqGicW1-d$^cl*Id$<}msgx(R!sr~J z722nnA7}DXqKh+1aouJ~&Sz-otAh6{P@6e$&Y7%=P%26U724icmJ_IT$$2s3<)6p~ zlp~I0DGNd{zt6I^Cu)${5C^%qAhT?8>`P=vms}_J3G=ebMa2Erg7jvCC9S862$Yaoi!;M_P+IS%Y9)ee&>K($4on@cUbMKoe`FYd=Z{Thp6VM8FKi2$! zMvHUG`PM(0pum{pZqsO#5EZMk2|b;6V_MZ?R#m^6izUMhfozhiY^^?mpwX~6l)f@I zpf$;rX6qg@fCC_k%47au&?KKpQ0W0bkAqTQ!H=1_dEZJ~d)~DM5)Hy}+rgBo{v5ih zimcOx`3rlg0lzly{ zwhamh@i7KliEy+&8N!9%PRnk0TT)O(DEA~{0fyvApBT<69i{3tc-L(CT~9WsLgo?7 z79DTU{!J9@Ne?;=-$`r#zN?+Vc3f(|)^BX!xU_k_bpUHXl)u^zg%-vv<~J|-Zd#a2 zE4Xp2N~KtsfI)uKljS6>Q7~IBcZ9eqltqAZ+(=L*1$)ZJPb$Z;Ao{6*-DHU^Vq_;t zsAI!Dq9}6&2suBm%O>J8@8%Sor(Iuy4`kFFaUzX4Ak0?zRc6GUy%~R&Ep(K>wlhNf zz__3ZK{)vb?I>m$sSIp`JFrT6_#7z;Tgn4(SxPP~cZ)B*<*$pvWP^Fm7ph0JVD zu&_R|u1rG|^XKdhLMK$w@nE*OSHXBKfJ(|+2UQj^a!`?N+V8(xbiW*4k_)p!IHZgC z7E0x~Mwl?0??0YAtQ~5vg(QVV*=-{xU-C=XRnOV0wc5zjwV82^UerN0Q7z_nSpY5i zPX!#;tYI0kzZu>ZH@ZZ z)u8aGhml$$z2!u7`|w1uEpcEJXa^KLB%C0yF?`mhIytUFmSp}dV&*g@DHA{QaVJ?>=P#ylRB z&)~E1p6RyUpjqYf-OSLEV7A=nj#vQCser!#h!n72{hAjzutf18fyisRGia8tTR@v) zFk3z?CBMN9>)Z#@VZ9~bM7eXq41UYRw{P(+XqInvbL_xf><5w$n&qan&f6rtFRyxc zTIUGmCf9k%#J8n%>6@?RssU(X1|k`LC^aMqDk(w4}oaw1mW_RF-j zkHxK}(MqcyOuyZ4tqs?6EUGsgkRt~*lFW#dvgn#3IkF@d6o^H%T?s!I$w4F1S%7V1 z^Q*O(CWZsz=LtVo7lhKml#Dwi1 z8N zb49`LtSlR_2*kjMuv&|>a(xv7b2*j~l14*}V|Y%=4aMbh!4a~-!qSCMmAc??*>9NU z&t)L;4OTn!0Yo!V53^}OqhWz@b%Ylp@or|}aPAvinx1>M{w^M!!{t7cwC;2&2ddqy zQE-gr_>0{wUIh12!tbF%Z7!(sT# zWaSun1=9)XO5@zLKa;le0fber(hl`wD)J8Xo8|5dE$$2sm*L(ljdG48w1Pz@2Ch&{ zYGS$XOmpApaw9?V-;;nydB4TMh#=Wt_5ys-GBDj=27?p%&CjVIn+e_#%8xR;gTv*U zy@lW%@}o@Np5g7Vws}V=_hiN$Tq?ck65qBz?d0dIlQ$T@%}Qm%iZpbAsH~fHTJ+y6 z-D4p<4!mQe{>35$*r^)auF?r!!tDp4%GWHI90cxgC1`^78=V*@WLJ~Iq)5*Y)uxBK z!||IZ#kMXOh#VNWt^lQhp4{XksG}PH`~GtZod=KVk+}2NR;ORmrBecb2 z7K0c?u$hpW=2F1O>>PNads z5wcrYxe>4yrMDoN{C2DCQ{9A6+Y%i3D>Eu#&pfopL6lI5_Y4 zju}zVDUzzBD?)}WO%}-3-N7QMPSQ$bkxj-Riwq5P11*Cx&cqEwT;c>s0zo38C?G`9 z0BNBApYL;*s&o=&;>`Q|Xp_2UKhHVmInRFF(ee!N6H`R3wc6~-Q?b`#J(Ji0A-vqkbU4Y9`rpXND9#5z0E}ND)Of!v&L&j#w4+nNzF_)UalH}$ zNZf_*gLK$~NJ$yvV}|^o8wjuvtL>WKlD6?0i{mEBcMyTIoZm8GFm0{t>T?s($F=fx zLD3y8Pxbw^IQ0RbG=w(i4Yj06$% z9{_(g83+{;svV@-UFDf1Elp|V#h>%CMws$nwoo9e`XqW;TnheiuMiDE2tdZXIq3i{ zG{r5O%XF;~1tCZ7)UZeV3S#0nRbmo{YmK|FYyN6R-|#kHth+R1%{_z=|c z9C80qvfp4&5tFpj+lfL^uJ7i|@AM3p82H>cu8ZrshuBWPFjEs#fL@p(7T(G%GPf3a zAk(Yh@Zb7r(6mVt(!uVu!4AYyDfT{OSJ=vJ7FOgYv;#XJ1`M2Bsi;K%*a^knprtxm z9wvARmWx{AGT>5tJs-W+PmH)G`NPU$n>$+ioXNY+%g#X0yJsbn>TwnHYkypc!0l+2 zi(SDr$%}l_5J|lbj0RB#v4U3XExjiG3GKZDIg^@*9XeCu=g52)+5op~EubZXA`i+3TJvSD^p!$IIb+WpUM} zbyx)o#TEgwlOaBAFNr5Z{Fe)oA?^sGAzpt0L);lK#9J;nz!1@$L7I$vnKLG@bVcus z->zYf81O5yIeEsb$foWpvLRTbb`2-D<%vbU-79eHF6++Pzss`ntXELEs=KP8{MxHX zXH~w|orPfpCIro5nS!bEgjZ3yr@Ja|hX>$r&+O4KdCZnV8r;E~e^Gpf17$9@ZKmM? zGiM(xGm!TWGG?9;Ai|}t(N@{iU6d!hLE!oCjGYzZChQurF$dxgN@rGn;uWy;mX?YV z{N_M?2e-_2`0{ zcyxJH7BHo_Qb;*oz?XSd1Y0!?%1L9V?3KT-ri}$MJIflz*UgpWWCO*Q>l{dn&j1$Tvr#mf^7E&Z876TmtMKDJ1bAK;9u^}8ZjNr%xfU` zbZ7muViC&+Qe<+C@!z$Ggz!&#J<+HS)Qem?3MmYa-*r}v zN`khcRyv_l{IiBqGp{hM^aY0Fn&p9&*bv<9D#?&0onGqE1wWNEwC#OoH|Ag=SJWLwT5=SCaj?yu!2{y1 zMFvd=5|ju{{$N+uzdR#9^7hK@-9?kT#M~}(wtFj7xgHI+3a9~M0qaNqQ*s~L`+K^J z@FP|yDRSRVPyUJYkQ+&C&9f8t99>om9q4aV{p zdBlUq0u9zwZt1o#`ouRg3*3QHVpY6;p z#Sov%_$SFg_mDe7e(IT(#Xxtp@nfqyLmu%eGAAo<=9)pOEv!gSMDX~hpe|?l@>Q$( zh{u@ZU%Inav;Lv-bE6(0{bgm9%M+aoI)VUgh>WGI%2%u=QDi@tIlza?FQaDIOY1b0 z(Y-3CWnH^`z&7Gc+m#a`_2`M+&9u*|x^_7pIm#J?_iEGd)*Rb60?MbY650DyI)h5@ ziz}T@rC_qvI^IAz-D;gYva7$R)|9x`-vRDf#+Zl5@JyYBC{j)vN@(TXA4yO3+grxe zT)V6sa({&gh{3u-qUbCCO(D_Nwfb;UWWq<&B9kamb7x8^>yD3jGsJ7i6L4&X2Spft=e1&42p<_!exu{C4OYms9^lX@Z{%qyyO)Cjj9$ zk8VfOoSb~FbI6UAVZY#VvKo&TfS=|Xn+XiBda2c0NsC>8+G?3poEO$;)xou_9 z<>W6aOoga9ZTU7ha&DGTI{QpCZIV=zL~zboz?-^xLLo=UpB)x$I>7!g3@XDUkm&$SW1GjYVotaDC^Ql1=j1_HX+}u5s zu*UNpI5UOCmcqg7d%T98wb z;zc3t{fT?v;o0U!p#W#oeU{j9L9a#xQnQ9SBdhX(Sde0w`FuW~4Vn!r4QMzpWWXON z+Y;ol{s=-jz1kCZg#v7yYmd!MTv5(~rD>OU7hDkw7NDNj+;)`}MWa{~wSJvZL#+5aRr34S{u^qy zQu}Xsp8{14&G(@%D+IEvnB^fT1{6L2v@FXYM-?=3+wof=L;R1u9;={$z#1@shU&lX z74S|OdiOYJm^j0=b65YwLzYE(_=9)KSw&Zre?@FA=WIPJ z8YQzYs(iG_F3{E)t{vaQk}C%CcYVpd>{Zg~y=HJYLLRLL_A1Iv-NghFo06iZ$pgrr z^)43E_PJ5~reX>;%=P#uzI;}srVR#QeY>*SswDz3hdK;0K{#d4XLBY=Kq3wv<(}`c zd)el0G@MN`Qu$9b^~eNSFqDYcn$L;Hf0{gK*IL74gdO#~C&}wEDOO4Q^X@2l0RqLH z=1-K1Gke`fRes)OHm4J+kVVolN>hv$W zs(fFHC_*8nt)pj>7AA;2;AuVJ;u-hwn7YJ4O?_(_}|xSZ7$#Tf4F-l9=oRnSvi<#m(oYz}y4C zkVsk#{}^VE(YwLy`Uqw}ti$Z<_}FRy%y3;QHNvcxF_`V`;x@6N*D^88e%e)JDqpjy zq%eELI{K*Y1G8_|dw3|xPhi#?_3)>951-Ezc&njZUPmlXA7TIaMWcPN2;Vqgd(9`6w{XR^k`|O@y9kYqTJ^+%=gN<*|y2r69wp9 zHiYfn;%p*!p0FsM_za$MwNqs8GeKs9rMOmO=hsb1(Q)!3*u*9a%}3wCwzYm7W_J4G zxXE$yGUXuOw?YR4C(Q$P$UYvcWzy8%!FefZxMrQ+D)*b0K3S339AGXh44FmjpPBTt zdP&;PrSXOqdvAvot9SF!v=`H^49C5k1a-AvC#NQT#D`|n(zK8DrW!FfUY;;xQA4*< z7BgZt&|J9if)_?^FUz*GJU1I)|s`Gf?Fx8yZHS&<{Na5jaN?FS$RHj zY(l)S)*cZJ%?9v%&K2VJQ59X2ShbcfNXLaP3>fps5YMJpFr3s-<+^UOKZqc)1L>4h z;C@Tu4=Sz^XzGOU>qWfjsL1D{D$X|FhegK7m!ttsnU>u+`3VDGr@>_TXBe9O0r;O@ z0W$#v<2mE*5Y2GemtY)?4w9x`9hODu*&A4%&gJ_tj8TOhM%-OO$I2d5J*~ z_TPt>*?khKyl6A65A~KzK_2eP%En9~8B(~;* z588l+5jaQh3O)GIngIFj;kPjqcier zyjJAu?!J+yLwJ5%uI%Pe{L)KLg;!0Ryg^!MY`|cZ!dV{86eJjw3#t$$W&;gRLcjbO zI^_p{U>yuw%m7Em49oHe$zRYXH`=OGsBzArJe(<{rLgcxIKhk3Vclypm8-g{Dov?C8&q?s^TYDi;|+Kw`rnW8jfUDZG%Sdtx|WpvAUAaPwQ^5$XUQ)-YzuK^&s31mhDdH@ zn+=;c8&$s49r0d*9orP03Kcl83UZ)LOS3>AKNPfARdkjAx@xNtd&nrbM+O$%71zl z!hhS&4OM>T^{U*C1O(AlLK%4LQn>xzmaL{>KJP?s|3I&Gj1}-K|N#1ND9{G!5-KooSa&PAhc3G3ovV>;BTl z?w=QPVA~6{X?(2=rZ_%8!BC3?j;>t;d3_bD%XPv2qIgXP@xMrFR{<^C%ZY6026z9_ zD=?o6VWg?>I^RZr8bDrUZ&=|fE&-r!l?yXOj$WON+FZ;a9OZsnS^2i?lyB3ed{YWj zRldh>K|j~?RVNPYw-OD6-%YqbY$Zf(>C7R60PjL z8Dx%cm+wLTsSrCsmS)}CRkj*Vn+0?+OBxpi1LC59odK;ffUVu+ZfpjewFAJ1Q+T<$wjsCvUF�QA_uKX}CdWQ0toT^<)`QE6*YW(Su8%<5Y$Z#9}a@=etsC z2+Cb(R0M0&aBEQRxqZu34_5A(`kHxHB=e3}BQm1I>P9lX)z)G=9APM#%9J%z5t%^v zvafe&_-b1)V+s zWi91%!P;5v#hG;C_heqrwNA~pXQ#Ak8xkLwzu1<|Y%u~^ks6FOe>rXc^BE|I0XL4j zZ3Ni=FBe%1@egoC9srQ{cBh+qW4+Z&qE;EK8z+T=9hMeT$c&Ts9k3~x)<8ovUz-l_ zD-mmmxsQ?bA#&F6c3Qh8Q<=(M?#^Ohbrjfu5!N(pM|4iOoL<5kU_((uGx1L3d;ykI zCgsz-(E+ehFlV9iIK92lD&57&8erxbnY0)oB4^;ngTL?t^}R#oq0&^3JSVrJgE zK84VWChzN4R9w-#`ZY8Xbrh0&+ciCCH&Garxvc`&J1(YBMi1MMQjT*RiW{`g8hSel zl=e+omDLgtUt-n}ZN~K`HJnQF|GN>DcXCfD?P0a~DJso-hgsAgih(y}@<}bfaD| zi^*%~pGPYzs%DE4te+emx1(&}p<2saoX`ymcLP||*>Wc`mt7;yS6 zDeM!WLI9g!l>{0rQ#sIFCl8Z4bl|Dlg&j20tBKB|?>F0u(*4S;A8S#j-xhaYlDD7< zl%_$$jXSh7qcylAmkVYgH!>I+7W}pVbNokpTq&kH+&FC~3|T1<$eoHS8IjqJxdw&I zq6k6R9nr3^x!LW4A*N0NpUN_jEv^wZHDK-Dp;8m91GL|j~?v?t6&J$4;vG;hHB+uZ`gVbxev*|*;s36B#sq2 zMpJP3Ck<(sUWL=|P5ZtA84$BI?FQuo=}w=T@%}9>c#F@m>yL`CW!Mq?8$~H#_u)So zefh`8!Xd2%(p*sqzzegB3GK>y!0Q4|#Op-k$*Sz>D$18K1q3fzk}ES6_1BpZc&PFw zZn;RXmdBnh#FrPjoQa1K@i&`mF4uNb^irY?h?qz-+*o4QfDBM{G|8RaXySkBVJ$b% zohAFHMp;?5%0M@=OBTO|1%ZW}n@Na3mJ|s#Dt%7(OKE7@n_%o_lKoXNmbE`A{a2Abp%nvpq8TU35NEhSBMyDYvfCQA;zQuP9Oe> zW-w$Ct*P1P4EjA3{oZ4pEJ}YwIyu6F;ZhJqFXEVKStbD19uxY=|m#1rSNUy* zUCgk-6tcK4fn~>oaev2}i6p#{*+FtovL!UT0VG3BMu3IDHq)-u+_9!4F)G}(%TBM? z9UDn&LPZCT@TOFkVjQb-U3Z^rXBTP6=e6ue*DhPViaSfp>=G^;9I)$Um9_;m4`F^bClWoyaF9QSbCPuouW99rTz9B#36uXIP z<0)S{3OI%NK0dwjB7xwad$ddD|CruZ>Q99y52`@f-`WSK_j=9mRUEE~v4H1lv_jbTu1) zmNxsH=z5|ZQw1m*%Y7TSr_AC#@=LGajs>rAd}Vo`Q+buyf&egsWO6rjo1ir!+@VrK zYzA`1!q+sMsX$1!=_yjenelx9bVUK7ara~t?k~UzveL0&eK*kk_e5i7a)t;p_ z|MJXBTBG}tKAcEt%G=V)zs$5?mKLgJlvx@ ziM>JxFL~BMPt*ZhxAq1dOnNl!U{D&wYSZfJcNScb3bl*QbVm>z4 z(|<)>@MIep#vD>gm?YcY)DR`Lm^?0!%ueV)Bhl>uf&jpiH3DLc=Xxf}#~#dJOL|*E zewTNq$H`EBGdV8oKC( zl|%S|%0^dlpOzO^nk*d(;bK<-ORPVX?s%Lv=8ZQ$k#=|o9pY>@A&mh&X;+$imut}! zu+!@4gN^Ca>*@E>8p1k4V{jK?Im5m(Wl{>Z-SryxHMZpaBJKJ9#`NZT`h)afs}pKa zOD5D&FYu$rimU7CKWG_h=LnOZsR=~_KH?=`J5u~Fc+IZ zGQE)_cEv8(hkG>KJ~C!DH!pvzB06=trgfLj)N^pxQ|XNF zZ5-+Ddiu8*X?!%n=jJa4`@x|jMoL~f_mPKh{#DxMj~d(DQcwRG>5y`X6nS?&@2RN2 z648$nJJ4UJE&hSrbsD-4s(hgnwl4sg{oXIENzl#adg<3F4YyazC2e_6r){n0@$0y) zJ*(2T)=9msjrG!3QQA$En^r)JkCjJ96`&4&w*qVG#@l$T%Fjm?lo%d>=&Tr5tpo9H zvkThDtW`tFnOOasbVxVS2?H*#NwH*-J6qmih^wS|cSIz5wD_-xun4Soda>+|kh@I& z8w`;0!>+75LggMyvfz6UAmWD=jBVAo{5I}-w!8l z5s^*HZ1pG0vLPnXRB*H9f-F;SDqt8*2aAgC2>C*$GL?_t8BNQ8Yn1xD%{mx>n+@G; zd4Z{Gjm;XW*-7JS%SG9p3TZa|E}h;7TE^7GsBkH#$EErCH2-1fCyqtk(Y9RAy*R$) zL85NAVVk#R>5t) z9gjJK<5}jZ@ots_O)WX*o=4mSShigpu)QGS`B}Vo#(5fiz=nD;+~q;YquTy{OOIqiB?UDsQW3SV6oMJet=v zfQn`0;FoMn{+bNdvm<7v2ZeUAxEfRcP6Gwj}jaf+KQ>oSA(yU%_X5F%0Z zPhGfc%qA)*JRAVU39#G1pvVD7bUqTSDjiuFTYyZ?YB*DtW!JiPnFJR$TRxmk2%lXq zrTcJ)$>?i#TGwe9Owo2=mX$XZaIHWE5d+hV3OSIb!iY#M4n!G9mEr|4{0P`c z8WYFOoMLfWZAs|6tYNS=Te)MVO@UZqQJl%(NDCG>!#JW)R&3RhEYBuLal)(VApRo8 z(5x{WGS}2iKI_2Iebl0-1#KKm_d)rpZ&5QJ?<{I5XwLcL_rSw4m^_9qT>dqj0Ag#D z)vu-9e9*cviEkLpU;)&h#52k`2Q3K&ZtTGb|?l(jY;PJN?q#aiHzbGi>?5msN=0BUf4wOfjX!2pn48o$*EDL>` z7G!J%ZqTegWi*E|l(!DceC(Qu0kFupHM1LQ=TOCeU$p#y*1AT!fY??5QP2$yF$PP!&pNN;Rh z`y6=}1lRet+G9-1j{Js;p{!GMb}D@_UgP29*iC;^Tz@W48`aESoTJ^wOxNe^Y+xu$ zvlk`UaZ6I^pSTnH)zrYN;9(+zdc31F%$0#p&?}2^zd{(n`N35&%Lg6s0tA-KT2Wb! zRANALhTD2HrwD9Jlso?w%1RUJuSf%&*YHsrCUzBN4c(b?Z#O4yq|4`WHh#^Wss0J_ zal)VdliY{379){`Ov>!TI9M0Ze-NE49?61WaTPG(alMc%qYLqdCwj}`6!kX0IR+pt zb3xV^d$?UA7os%@pf@Mw&NX8sIM-54o-sR4HF)SOSy=bpYl)lwkR1)Aj zm-A1O(}vuKEYii&B8V<;dJ^=K94$^qD2EMK^KG~~b{C_vn$@XBT+z{RjAzzFarlAd zU3m5b&aD_yjwnfySNnX}K5wNUxm)d%*OEeW+aPV1 zrCpa4`MG^wpM36QC%f?)5UPSnKW@)j+qvMyZSF(%pcDkgY-LPx zjiJ(ghxC|@|0b$4yB+pusfn9Vwn+kX4VHJhTLJKjG=zSzwXdEP6y$$WVKze9tTR zAH+cvPwhZ%=uRiN(_@A^4A}>AT|MIoGB}!_Bd7qwz&a3yX1vNPk*)SVXJp9s(^|mG zetN||dy>zW?DG{9dvGyQE_`G$j7@NyUh}+M+$D9JhCe5-XS-O1wwm%Ro2auvk||%L*;fEbo1n43J|5UV9SPf z(`s?sQ|e~(SY&-`Zh2%=MQ@cNnO_r&Kryrmkk9i^kXb|V0}S)iKth|}m(JteF#PY3 z>x7-KUz6urlbc31dCrj5)+C(m#`mXe4f0iWlT-ivBHGvaj~l_-($^Jb_sEC&y&-! zLMZ9>yt@Ez3b39#&othX-Ew(vfsapT-BE}Tt(A!~ZZFDkE7sTDQ0y~VtD!Y_9+3LS zAuWOIFuGzc?;?1K<=9KKP;IhwYr;29%f=%=Ro>H}?U`Ores9Z|eh4`1_leXW@^--= zI*XC%+eW{@_Y0lHJoc9%E9qLeR_@{SvLOv$2mS6;xvvYO?t3FqU+7N7Bp!7@i^|;` zHN*R5sB239z)F9kUi!YSEV!^i$vACFI{H&$2@U4?5MAk>8>p8tS5Y{*b5y=z*TAV7 zV1GlGQ}|En3sBCbEnidTr`=dq@l=(i+5ExP*qdHZC9K!CNn7L#Lundd!M#LFBN<&#xO|0-IOpV+9xQt=9#-f6K z{2YQA21ypfKL=tK;hHu$I&3bajU3$)*@}LHHFO_moiJ#^ts6&&kE_{rg*uB<-Wrv? z(Ub)wV9FWok+S=^e7|lqKkb8Q5AS4gVRr)F15BM9w$w|U$V4{iMOYZ-*D-epn$A>g zF#Rvn#g3$CoV7x1rMZEi4jBt z4~E%XvJCmenBLp9AHVU~$-8`N+{AQ=v}c@)uVxBvy2>xQ&^q;C+HjEtWe-@4p!tvz z*0Xj?Xz1ZYZn5d3ai5E@jN%aM9k^~v!JUc~o|fDq4P@u4qUPP1W*TeQuK|-Ug4zco zzPAF4A0hHZDowp&C#f&FrMCQS+E=6iZ0zeYN&0#w>FZoj`tp53M=gR=CGBnd_I32g zzP!In`(g`&Xn7AI()66EQsVWi{n|d(+CF^WwkP&RJ2ah#_p!L`#U?^<^L4#jjA+-L ziGXNCvx{{yDcOcb7wg*eirbYmcDfwja8c~8C!g4GQLHDxt~*`jPo42a@p-?%_aFAZ zD84@OqBtF`scSJcDvxQp)8+oIK#^-^ZXh zT_%G|4&*X6{o`qwy_}5rqWA>Y{^=UXBsv~(O?cDOI$y@J7ezB;OQjC|TGDoqQl#R= zttwJOMh>m=G@d+iW!Dg1uHb+uNRD)}`-A8mYZMxcIZx52berH!=2x4}1A zDEA34I=2D0KDWUqA`{nrZ-Xdm%E;T`6VL~V7=LrUhw-f7{Hx>o_xQ=ZEpVjo7e`LMMDWF7PRNTZ7TcqJ18>TOq4f%*p zpXu@=4cUWfEf56besHncXYI5G$=({H^<+s+alL@YoQ z>EcDAh4jNHI}9g`FtAQ5?3)tOD_sXjJJET>Dpn#ixK9$=Z7XujC5Qv9lB~%3l?W1> z&q`(;|R0&q_yfb==Ip zJ8^STV60pnU&R2ej`6k(8NUdB=AhWAp=`p`08RzaK^=R2J zDoyn&aR=ii+i=dB51c&Vz{!&ioILHo$Q2ud6z+cZodxMfZZ ztCFmaw+fyswt}r1PF4cHg#L*#YVYK!vb^Y8G!UHm_Y+1$OlC>thhCx09VN@Q!k-Ox z7hOqy=2cvaTwwyQqZH$DH!-o&4%0dsK{$I3-^FnXBJ=eqVYEtBv01ubc699E7ij^QlXE$_c)IKfHYyr|WN z%o{Rk?u7YKE?WbR3N?3p2Xw?)Rx#udQZFXAS1*X_S?0(tCoF8pL_zK6hm(tJ%YJ#^ ztB=Mzuc6||RyW0&?qkSl`gOZTiX?_+R<>Uds1HolKPG=#g(pz26x_$;#Z~4B)GGkZ zvQ=JORYaz#q2AWBPjvIrLuMSajjip+Pp8W$)Wrj=+y_OF+oBl4jGbc=ne??`A2n9x zIe#a8*c0WKS`f|WeFpKxP8-DYwsMxO^2N?%5FDI`-e4fdfay58{(Iy+WGOwY$8>zo z3K3e~a()Cl_c5~z(S99>aVJOG)WspAMWL43tgOBut#kq|U=c{EKa(! zq7gvtKv+a(2h%$JE%gK`G%6J36Cw zMyBtsrw>DhUMmHcQF(He5v6w;o05?qciQJBsibjPDD@+5JvUAU{he44**vg#54Gen z_WF2~+v_s&AZ3$_ds=ronr)nO-IgGpE%lsXO;75y)dwwK&uJ`w;X%vSa~jKkE-fFF zbXzja&GqDMW=(&$o!fqaWRx~TsC zIhzieGhuyNXShDbjrG(KV{G0Km3MDA#%Xb>{YSF-pgGfeqm~a)r=BxhALFFYAFTX9 zoa5ebjJx7e`}bG=!ogbJbiyys@Eydh(Vu z*^QNFMh)9-l~K8R1pzG*6#2!p?rQ?(p^94@`KgcoaVSw<+999LoWdq&j{#= z0oUfvk-NJwC&>hIcXwZmiDM_UOZvyGG`u!|wR! zXbAh)**!8Rt+V675dfxbV{`B-Y5NpG=24r4dsm`;fY$UC4z_PH#OuZ^{#%0 z=lR|o@w`mt)iH;8nM?+=;EZ?j!j2^;pR;iOoY{-!TyXM-=ggZoYthMPFFg5__kSS! z{>dlLn>%~XvN`Q1FJ9Ds@`5=_PF}ibo_7dU7B5*ecfr|<52csS&R=pU#pf(qv~bZO zb{9)}RNU;R^&Rs_R|EPLQ<(fM_LsvZk? zVxj*c4TC!;>+Q)w@3=WFh^;Yv^nnXb`j@nz*=3-4Ia~?I9I8dqlz;BdIv=Zose2>o zxu!m(yX%!>Ic7-Ze@93)vi7$z_<{wZeypb7d`sH-lZbn9!^3L_&KZplENEmUHB||O znw?smiqFhizOz%ePSxNjmOCGq6-ccfmNv3WSyIpmICZfBRID*jxhWHcmI<{7jTD__K$Bk^ z#>F61x};J6v^W&$m5!I=NrRNqJ?Un22~0-!MsBcpdH3bnhv&ny zo%7t+egDpN?gJPzs^`NBxAMIS|MKQv^cV5rkWFQx((2B4ePc>tOu*6CG%$3seTC0JezQO-+$G;{5buRZ)8#NQ_X*CV$vg& zvtmF0O0L4x(G0yz%nW+@dxBTA%GXdl65bycV*F@0(10Y0^t0{8B-|rbV_B)8y~rM& z${lfgh`XU6XrmFQF5n5lZNZhevp--i{?)(NVKAL5Vo{%%f#~*2r4F?IL@hZBa5tt| z-2LL!X$oE3%{e#_lm;ehW~z1m%Q@H?mWBehHz>dr%Za;jat4n+u1bB|qM9lLnv2_9 z98J?z^$V|jXL0tRx|0B`lfJXM>h){MJ(m$9)dTKfS;^Y<(yF#Pm9*Go?x&W-+CBN{ zwdN2E@wrk-8h9BVxf${Vl=6%%`9XxsY_g{vedPrD{4B~W`g8i77nPRs^MaAe3eUT} zJyz}oA5aliv~=1-a-g~0<#|(CUzGD3dzYCP$`-^>>GL9>5;mUP7~p8FvsV1G!R}{C zYStv39bHK+bD1CJvO?`tQ9ecrwW@%C#O+Y=WU%HG$gQQ(W5&G-J0v(xrJ>V5oAiA# zBP_&Tjn#k;o+GtbDf8qytu`XTD3*s3kesStH5B6Pxa_TcKPEdVwXVS-vP?}>GjFh; za*kwMXPUpN0+wW;ep~-mVd8VC;jefynMM9)jn(ppRUscw_2)7&p^NdDb>Bxbo;hEi zJ>yMRKkaZhrG6o}A`;523Xgkp^6lNMVT-demw~)-d%9{}p+T(T$1JLTbNZAc`RSqS zi$GB-PH>>a!Btaz3ecEOEcfU8JItb#>=3=peI@fE7>J_gpk_8#u;%%IgPA_zY35YZ z&DI24G4ra+mgRjREIOcK#g+tPCySU0gmwJV9?8)f; z)6WoFp3p=mb_H1Q4y~Ye)$-4Jmch7c>0{|?=Qyu-8S05*9h_Elac&C&2fE>uc+tdCTy0yV1F-qtmjkgiWFw)guY;&(z zGLgO0tlAQ&pL23oKrpt^k)PgsJh3j;B)Mfb8iC98tb!qEbXp^;aaKB(;UBOo5f4=F z8QNl#yUf8IAzfrCv>3=@@*7CV!y9O6Ec&zH@pGz4t|sk7r_HZP0*#NPC|TBhQaC)r z72`>Az;(6u>Tg7pEUJy8k8|j;4HS7-UuZY>Qu)RH@aMd4sd`ZO9ugy)v*<$>ZQ6xBiAyzZdEWC{TotrjVQTpyDfgjV zWFNC|>Z>#DBr(!$k;D!&ed82?!PV+Yyv(w`e#?Y^QhHTE#Ct6_mx0jT+c@s~<(9uy zCjzH@g=;>zaZw))zrB1N+(M-Vp`|-gpW&&1d~d9ib5VJ${`&DFv#k@S(_ppMSggv4 zhQ;IiFPOWy{~N)3fematzq&}gcJwR~Pss-7h<9f9d*#wT_5OTYl#&F&t--IXFsyh! zfaZ*4v z@i@UUk`-Fv);Iceo$C|RVC@CIs5B`1Jyp-c`%#U2DhiH^pmU-LSjrJlryvzwx8l9M zQ7iGN2xy*H=Vqp6B?ymOqNw%M>cqUgg9+6~&7~XwJ8{DkLN-{< z38WsKB7RUVk;oTeq)*44AsI)%<6$&DxPX1kgE{nZmDM%JAoQ*B$L|5p?7g)v*aOMe zH}Binfa_$1A`PwGCb6~VWG$c5tT#mpOJObSI*7uYJ!tm0??$-fMCLum8et$N0;QeQ z&}vvJ$rpeqTOixi7Oy!MuK_N$A2gBH&W*1Idz|ctj+1W$F@AV||0`F`l)~!@(umOR zi_>9nkb~bOmS8b*K*3x12~~p3v1XNZBvG2%2k+*3{^wi0_LXaTb{Gc9JH$ zrxLO}ulFe&+;H$i#kt*dvfi+>wO}U)_p2h~2sc$}=*2>g8S7v=Rm;O6Ia68ug=OFG=Oyh~m& zCo1OV>Lb-68xvei=X_pwvahFUKV;$HV+%DeIqAOFNCvaF&AG!mp7pR{MP&XR5kQChZNnKGn=Q>q?4shYAl%7~P#K@+tZD*d_4;(_5y{aGzI? zi<^h$G&!9^WRi=O=}UBv{(|mCwa{|Mh!(dzc@X;@@#~?ClGWur={;k8FD=KBpX2ER zQjn?Ce&Si4*5wNI?Tx$!{uniu%HQKu^2zO0(YkZ}{#cdDz%$%8gc>gY<&4RxE5j{C zE=8m7ujs9k+>PxvPPfXszqr^(Ke_0Uo$WH28BT{!1 zFz|CpU3^65UVb+0QBmt&{_rO*{?b*i(>}DUSo^U!5ITMW%qXS=eF^z4z1M=^kyv5> z!s?RC3LqtoZ@T*J?sxUTA)pDb{ce4>tz zJxhEk2s(%8Nl@gR4MYTr>h?C$dR=zkI$RbY5S3)+JMH}jlEe87sWkN)e6b-Hij$i< zgbibLrg3P%yIEvc1N>0ZZoEKo7f@|vXf~ED@sRN~?bhlD?mN8&&N9-_29Xg)$ZVHevWhg|>^3#SLlh%Hk2-3$V(TEP(shi%FGF~km%u~#rEvbr1P1@(A>}`3OwCl_vzlN!XoF*Tg+1u= zY@I)7Y2q-G>E*n!@!r29+PIXo=4Syr|EOBo-mzxLe;5D7n43Cx;-bgkU*4;a2lX=| z;YRMCU#7jVjg<RVUN9>q4Ya>;x6<&Ji_lJyl5FkS_+>u-pE}FF__Sd^(=WFDdwk-)%FaLe z)=NRppO^G|7BW7f;Ymgh?SxG7(10`rkIU?{^TOHV|3dyfDLWn7h1mWFpL1r>QenQA zTOi+H-|_j7v+SIn5KtA6>Gp{gE;i)jrX0eo1u4x`f_NOSdl19O)@vb@=_~q$BzxY%`@0&PlR_uY#v(nU#~Xe=55&^(^Ilh zds?0pm?r44M1PtYG)TH#eDS=HmC|m8q@_<>wyny*ZNV3C8wJOM^7TX zqR4&w_wUt~w5d&*oA-Ii13*`9VI9KfZ?)`ob7HO<8fFvluI-Z!%#w`m(tPLLOB&ms z);GTn*Ql=Ay$_y_t&VQ$jhiG|&xs9LaND8u<@*MR0D@#_E1~pa0sCO1jWZ?;jGKBjkS4ZQ88=FIOP6)?Y)q7W0^bZP;FswS-UM)>kHtzxQ?ZK8-uz zwzD+e?W~7xWo<zCa1(etmhNEYS4(>(vU*9V!6 zn(~>QhTd2y&9sTaV@{N76B-}f!?;IG7lAGb%^{xp6f`al!{GM}ZBR%X7%BXjG=N^_d?%)BgK0lVO%ApPRV3(6?=Fn%dpnBli zp#90zOfuZ#!RFrsE*3M?z@c9#k?il~SfBtj{lz^9E>96lRQIE}-*q8@L&OseSlIlX zI%fyJWoO?q-zWCVH8>9!&0XO|AO_dnif@)8o(#gs8r{zL zfO#~t?O&ZYZS;1~&7Ih~ZxvcVwOIMTd1JO8%hi!n_YOJw39ZuFvcsoa5bvW9V{Zz_ zUYU<}=oJG_E{?btV%sCsEsY0_U-?y*eS#rw^eu$}6&#;H^$6T}83|v~;+DS^=tTF+ z!Q&dP50m12wN|?{P>Hz5-=Z-CN)&-INnh0Kj$B*(zES+)GK*2a5-Cx@r=);+*|RUK z#B;H-;5PpJ5uKu)$>>)4Q^D-LVuW1|giV2Vng2W?h57B_ELc*xZBS_w$@uZqp*d-m zehR!)*3z{_zpSf?KI*N1&EVE%IJUf|%lR=Xyr5#=CU5io&ezl%v*}iCMIIhWQd`;~ zbZEJAYcYFoarPmt0J4Am&+^!@meYV`soK(nkBZ!1)9Wg4u;%S%;q!wU&L3!pV^ua{ z`l~iS&ChHN<#}TsNxJLjegOl;4X~s+@PA*`TJqYY#ts!~qS53J8s^?0=;IVxKd6Up zU6#ZVUj1ZpYj&1=k~)+`{z_r}eVpAdevtJc zwTl=bbjvn@J=!ua4i}kNEr!CJ5PI;sTj{!7bu{n3n`f&9pLtW#(5saEp?0Dm`?7N< z?`d??ZV~LA%r6N(R+$AO_}||DoT1;pbc(+>;C8XSQgskT)D@5mAIPbJ{~XJtG7Gmq z(c9wTC5A&AQ)0QbY_M~BTlmg;F?Q5?{B&JiaUQr7n{BDT{% z=**zHE_6%ElO;24HY)tccE-}$ieRJ#opL;4rz9f|$koF=weLjr0 zKH2kyjMjF_`UQ{j$>z6lI^T4WG(Wi0RWp>@>e*UzJD0?~Wma68O8yhx71M0(*PbYw zEN5ANC~BMT9+xrsjbXU?A$>9XdQ6?00MDmH?qYfod!zfjzZAgM!0l@}#J1eIp8nAn z@4bX;zx~eboNd4HV~@P;pbBG#z7%bFtaZnN7uX8^Af_R%1HfMkpCxq6^T(Je>psL? zFTFMPoRgRv_Simjr%zY;ITCNi`78!cWl>2FTSFXPd@X22psuT2iT$py!^?m*@L~Ub z$i}7NB-b_Gtry~>Yr2cLhHv3o2`gIz=efVstEpq9ACjYO_I#6E?1HR({LFpZMot?( zXx)4zNVmympdxt&a)TZ7-F(uw?`R26+)fa}t(qg7G$cJRcFsREHHyW5RyQ4-7@1t6 zql)v z6kuPLwa6T%EoJ$cPqzl}L=%6$l71kOMic2;kZt0;5FuC3KIv~z*HPH;RrvmKCs<# z^!49mt}Fc8M0Na>Nfu%!8=6J1&b}6f0qgqOpcO&2Cd9LFZB3Z(IrQ?%(TtnK7&V$E8Nh zG1Y-B%zg)dp^A@3o5swTd+=B$Yb`_2MDAxIJ3Gvqd zZx^LExrA459mmzIhPaMB`TX;~s(OdnuF-8+djM$@d^rQQJ)Q>2z zZ~uPEcCX{05_BM>JuA|sX$tjXy&nGXjt66rxL*9Ixkk2_edwM2L6A*Y*_WH9vY#oF zADn4LwtL`Sj(emcpcuS9*rRa_-#?-XVf_C;&jI5)7vCBSg27@4uG!aC|Gzr-1??K1 zGMZYXF&_%R-!zuqlQ5znbKqjmk)rXCiajIe3p#9&a!E6k80MV1{geMakGYx^s|___ z#`BMS!8~|V(Et7XTddw$8UALszAxjh4FPdtSH5oIN9ED*QlRha?{%Af=4WOybKxS1 zD3f{D_O12~v0?Mn%U8pk-v&iCw`FR0&BgPuFEVI&ndij3F_{!Y>kOJB9~8j~jMn)o z=r5ka1DSVauhOpsdg)^*YJX2W3EF{0`04_8b=GXDa~f0`lwY_vukl<6_d)TS6xCK9 z9P@NrgcBt^X`GWq^_yRw{6+okROGlz&7vBt7nLj6ILAcD*pQPE8cCJD*};V1lBn@| zFvQzBO|@%__S7Dd^n4xo1;CU8KK!;xhla${Q+&mi}~7f8I*U^ zkUzoP@S@01j4Bx{G(&$zJa=13Eawy1vw944;No5N-anEt6(?fB_w>RTLpEk5-&{>v zB&&-2bhAaM*~?tYKJ=%be~Gn~x&-yfUSHxF>u2CG1&UMZ z*WEGqE9Yd%5CiAVrH~JiBL^UmUkRAgTc?fm0nLhTx~Jigg1?+c^C%<(s#yK0wjp81 zn<%hrd%E&rD(?gPBPr7~#DM_3J3QBba}EQZ8aZ)iaH@Gntgl!mB7`W$V(KObOwh9r3w$#^PoFlzc1#6D>Kdm_Kp*?>?9mrX;g%OGuMx&V)uFDmSmh)m^==i$~S;BP{!s? zs{i4huh48kCQ!YlM{&I)xLh?907)}pU)r!G<)VZno*RN8@^e` z2hVn`PrrO0O?kgEiPi14@#k%3n}xB;nC3y8R_I{}Dz_k~p!NKM_U$)#ZwE?P?LV81 zWOlYDSI*E0ViBa8OM(5QR}yRLk;D zy-0HZZpV;rcl*tQa6TVy1I%#!Y5f#4wZ3{7mJrl<6QB*#v(F=ne7stA-vg4D`GCsp z$}qSj99!HzxFuKTX90?(@R~x|y(?uqD+O2XAsB$$my|rQ6R!_%O-wDuE-RgyROBSc?X3v$F$IkG`mkdTaWq>NDDVYW>98_Q2bx9Z&S_| zSAVbkkjipRK@TQ5uJGTg(OBAbQEWo-A7a$e#`vaZU(E#oMd05_xIC{3zW7yRfv8b# z8m4;tmC4{rCO~wzx03~%I@i%8o067$=5qy#PRGiu)5=Crr~D%$#-k=d0l)+Abww%! zR=FF1hPk>k6&^kf$f}H}`Bqc5HQnOGgD*y<%%TQGKUF3Xh^@);A(SDnbr~WLv~?Ni zBjIOrdrgXTmW9^0EJu}}88>enRN$dQA0-ovYIHl#=?k2Oqf|L+wnPa$$cC-M!IL}& z;jK0R&4V6?=H6!sq(Rpc)C_d-cOtnE#v8R0}RD;LgIG%46Z}y6;+wAdGdEkw4m(P8bMagp+?1)aBY z)NThUbUz1+5u$w6aTJf(a1=`XehD0at57O`@cDB+nQV(olK5!~LJxkW$7DM%S@YGj z5ZdkxLJx?Iwe@%EcKW_dyT-a(7#9&_t$owY*xGIJKmY?A>>v8k-k}*!g zu0Ecy8c7a4l*obvfBA59^7jH%pYD=u%F4wHt^}1gvBPJX%%kVe|Gt1g-yz{D@9YL} znRN1V>5+31)%?`I7SHh!nbVvha|UH|e=f+S`$UJu7At0gV7X}B*M#ue#Ct-!Z+W_1fM zX?z}Tt4#y?O?)WfTXoX5wQbP#Sh(};>5-)AH^etNf%qc=hR8VO~xog$K32>bN9%g*4hXGrP&^dmJd zgZS?iGH+{@+*GUCftGCWe>BVF{<)`|X`Rq=H8poyn*dAlHs0n@ss9izp&wO3(C4 zZjc0Y}DIN?}$5nfyWbWL+EI{om@!A*(^QrvreXZ=5|2-!=io4DmW4MZ`8hRJpk zO}p)|g+sSh2$f7t37C0*&Nb1iv`??iL*}mrsGzw{zX^RdsI_d#vc944PL*RUyPYcH z>G_8n84xKc{*M*(3+jswWL~$O-TrIwa>F?8i4-${+nkqg{SM}B?sv#-`y1};}kSF`X~ zi?!KmZEy!VA%|+{6KwUa09FDIHc(mc+mf3KNa3uV$PQ(%e!B9lW=rLkqnzmV;PFHA zIyz2@ybW&59FQ01+@!nK-gJ{wEbGgAJK^QpzVt2BE+P*L{DM(&*iN|qrq}t(Nt0pf z{1Az$zPSp1$+T(Tv1Z9d)^SGeC`!9(J{in=j9q0Ny}Olq5}&^=d;wfb``KC9^LPk- z{t~EqQS@-4JMe^Udfv@k{fesAVj8gNFAeZE@>Aa@N1rwWq&#upMwbs1-#JRuFKBTK zYv~wEOui}LEusf48t_pl6O(*zK(NCJmy+A1?jjcnM;2aW@2=X%yo=Ua`0Px?lqHBl zk2Am{E{Cjid`T#f&4kltH{@}9!B4WQkMiRot0k>-o*505Li?Kv`I#+-K~cMxDdx86 zX*Jvq#oh*2i$E=mGNJG?@yh>0*Vw$*oLYw_7Gv)fvL1u*8q zJKji}pd&dR_#H%Wso$H9i7SC_yKN5V&NG&yA71k17?$oEsBmgi#=JA)n9vgPmpBRA zAkxEZw6Bj9_2Bd71>8cTD8694(O2W6erp>5++ z4#|wnMgnOo$`6|G$K=qV5oA)B^}H)ngrJlJILUVmi>X*YsF&3h0fgtS6-FmiwM8K^ zFGUT%NA9KfY;@?#DW|OUzjA@RYyU`<#aHat@@zma-`P@KCn45Ts31=H}L4A(dp!q&5^<;>D9mvCppXA z?at(ZrA5o1O{rdwz6HK%jYx`U|9r8pg&^#rQ@U;iO7aIduc$XK04@k~p&6nohVC}& zEk)QQL{3MXXW9bW9geP)oPej@fT8T!(wRlM;fJgyxoxblTukh?;pb{I*zG;9_TkHZ zm~y-2;N?w<&Z{ids&m+j-j{z7Y|&CfR#9`%di{KAIOtspE)=zKevyO{d(Sm~!ZN(w??|9W3g?Md9y7vjSlK|s z>LQdz{&arssWNeRx=sm#D-DC61jHKNECN?kbQ0yeB|>2C0o$A=lr}*jT!FzTW9ZOE zfAi*T8mK=tEtqa60GxSIQDrYh{kOGTr(+(R;fu%mT4BlAnNDe=0@spL54&Oc15DEg zg>L;^TarB@z~kxNz(%Z}m+yf|heLxpi%%I>I~9|#ls7SX-I_3ll-oR_WJZ5EGgM#G zjF*P}GI)7>tE69m^RfBz{3qBb)WiKA4GvWpb`!DULd)#syXA+Sq`963&I~LC2mx1d zPOI<@+lv-l4%Lds9^@ex5BQ*MPX7JmF(l{G%DhkV?X?{zAERy;M6xe)3qb=MhvHPh zTE~(KYaOezRc^KV1cypxA%^`(W-Iz5D-uS!_R!>aB=Gy z?_U+v3Y-?KnxiD|&}zdqWQRn#S!DUSchoc&1>4ISnu%a!2co??28b1@rKSchT`?K= zx$$kPm=3;ZC>oBrl{c9;UH94SHZj7WjB9sv&e2it-PLYO-#Rzts3f`h6TVC7q|-E}3`uD%Ry%$PW@N zzth56)icN&uw$y$K8nb?|HQ{OgnM@NW1d1;*~uk16q6~J0g02ojj+3&$}u(<%{;Go ztr_%Rbc!I!QI1AX#%^6S*X5=_lfT3s@2lwUKTo)OU7uSwrhBGgk{Ge#_4s;>y#I5w zDA4Zeq0GicKHZj<&~*?ey4Fn?&_REC8U6EE93$9mhkFToYSCbD5^n0L@Q)eLb?05j z&+98VUq_#;^xW30=xYO`wZxlqO?Tn{j+3;~fSI@tK#eh#7?L`v%Iyg@05^F1eTv#% z3Q*_!akCDLLp$+H^{JEN1e!O&rzm?uJ*dJn_}B5R$EV6C&=qF)Wgav~Ys?SM%Qz1C z*nL~Re-J95FYPXf3?S#&Kn~q(*n03QiOx4NzdpWv-NwEC{SobV{%RFgy9;44 zB{&7&ll-90&+wOFXzPIZ%cr96@F{GX7Ul4jK58gNZz+)?Lu%x_Lx`D3hr zt=ZxL%JKh^HzfS!Kz>8PTMgV^$2Sa%MMJtegKO}eSa=G6zUJKhMzPv{gLS0gE~9;% zmurHso%-nXu@A>$Qi5_U!jaQ0k)HAe=;7isx}3R1kLI%U?5=_}=)|IZ3Y4b&H)WM`3ITZoLuO!>95- zodGxCJnjsR+>nT$KcEE!JRd|Ik^7oDA7;HW zKR&;a{YLETqRdy^Wo4S8iw^s|NE>V@_Hj)Awc=1UoKC(XU~KiD)5@1(*#IO;{R&sr z%xNt)W2}*!F5=sKAMJ%BjfV~)Fg@~)5&Z4_Qb^&ogor#DGQsxfbsPMD>(rNq<&qg? zQjZR0>Rr}+V}mxY@Llq27Lqn)w*g7jXu2WP+Av7dg=d8an_yzHgoxazk6#Nf_ATL( z)-E_#4C~%MzriZ{Mz9ULezKmH-&uD112+TE*Jv|4admz-4y{S>@?}H6ggLvBifdL6 z&M^8V6#Gg9(WC1w2ThFD3(Erm3NP=s{SK%29a`d*0Lk*K4Kn>(qxE~LGlzseo-YUh zj`y|lh&V#XL6XgdIA8P*UY9K?c{AjN`uG=$754l7-n?guS9q8gQ;&~f;h%a>^$i3r zs2vU^Qf_CW7nPzj6dG))hu99bX?+P(MV=eyc_y%uO zA!HcXSu9k_=c1N~+3RBiB&Uk+itl*1^=hx5BWfAoCf!Shix#y!&Ms1Gagn{TaB7VF zoN=G142Vod(i}U#yK*yJ#fu!A+$4-7}OE*K$3e6Y&dNu|K!I#&k*YatlCEM4RX)orB_3Xp-VNSTn zqLf+MEo$9@7KT};Wy8D>lTr`C6yOtZ8lWb1yWORb>=u{it>M#bA~U)+h$n?@Tx>d9 z|6=(jBXrBT`f3(!ew^S3j1Ap_{^Hs_dWIdy>{0Qbak_{MT6r0dt#d1N-j4KsBGrT# zK^Fqi^5o-_BEm{9#SKBP_s$n`;jwNHnfyd2=)U1q1z5@ zwR!q%s&4!`iuthjGU!)^X1fp@aRed~7S_*|D4X@!*%mbx?vL00=vT>q8Q79Ie!Id30b(}1k`8*igqBo@X{zEFf zLtw1#e7iRxNUmRQOTG3%A2zzjj?hP{;GlB)z-ol6LhX~Us@LY3B0D6>oI5?ImGkc! zzomW@#`ZY*?}6piA>hKv#SrYIMV1Za6OuY3B>S=e$Fz4?_~d-Z=iaU6!k(C?Z5wO# zBZg33d~%R5lrdLVzXHPY@fwO@a;V6w5%^Zxc-4krEMc7KjjK*2fRrH zVPbCW2Z5#KdvZ+o9`WJKSaBiPZ0Ey5thQ~AMZMGh{66^j!^p8ZXPovac#ca#X26$M z=9(P5a}xq6yzf|1pz|*gs=zCi6TXI*&q{#szU+R7zzE+L+HSHa75J%O4W~@tmSMxO zr-^G7JlQ49qTB-Vn$WoInwk5N;2VjINq$Spzu{JDbJ@Wssbw&q6^G1WaN9}qJ-mLI zr}GZh3gyeykpHP+e}6DGn!kN9J$jzloi1U%9(yA>zipk$2BdT^u0}2F^Qr)}fZ@e; zx)Y&+o(ARj|5yJj(AMINqKir%a!v+bq>#+1gLRvd0P%o6)SWsEm^U!lkRtc28-_I8v65c=_Effky^ZMJs<)hw|C{ zVXGgs;W`O82{enb*hA-gCYg&t6|PwFur`k*nQ!%XGd=*dvugX-*Zh}P4mt5yv2$n_SKgCL(_UeTEY79@nU)C zAK*#o_SUYlP}`9NIkZnEHo<;Dg!i#3;yZ+D^x6nBp_Oina8qW4UeQjznG? z9asQA>4+&lJ(dOif61pVHjL|_P4cR$<_WH#WBQS`N-ez09V14>Pi<)1QKk1u#2))g zq+hK4`Q2>|JkaLe zoSBz-f7S%4kDX}%BRy4!YI&&HzLagnDaPrE_~IwST$r}>B-L(ui}+$b<|QMpovT*z zPKR~Q+VU80`24j=dnPL_MkCoE#wkcy|Xy(8nA zf3Y}Eur{T3@}G9?pQ}REJm5JyWm!F_7Z|Vvr^prw!JpUg<=VyZgYo(~;we^2_8UyjQt}y&u1meM$kcJKy2!)L?(U9@}?3 z_TwTtAy+sa+NU17X!IjnH3ZO^j#@naYGf+(?`%x+HsbF9IfDFJbweHR@+aj*XNylg za>Na36S!!WSDNS8aVK(Q&mq9) zAH;%%C-n$leO!2824LqprC4Jmemh&2;+^}n9J%dtZ-J1^mq-M=;F;+A_tLQPk|um+GJo{B`*>~czFUv+Ls;4I^zn=DMHNEUIC9#8f4`WuEv2hvq* z4$wyRCD6#%k=5DpXG}I~XQlVpVN*6=?3&hax{3|lh||8*c=F|)m3*8stmKP~S2tTn z7uuPw6S|AN!=86a!}Lj(!Ti`3DLiLo#5DQ3>z5!{g99Mb1iIEbGeq^LcDw?ATem*8 z6qWMw$-HCJ+VxB@$AdGM&wPfgf0hC9tpBN06N{r+s01BAhRdZTVqEI4>2Y|Ctl3Q} z1CM52k4s=4$G%`PDbH_jI9>0J#mc0=x-{oQL5VdAHbn`c612bype8JF=Z++*4oe>f zKhwpFWHHs}eN_x(ky)h!N1QAC6Cqinge0#EmlqcCOOJjf37>Dv_05Dez0yN}yry)$ zi&P<&Mm2?Ty~B{OpZ!KphYf5Q6cfd%a7?_6NVRo4a<27z`?9qk!4)^YsGuFbPg8!) zL++BfFFI6l_J}UGwlz6xLg18CJ9Tm&N4B*k!UY4kDN7Blu?knYe7PyLbt=epj25!5 zJ^Za^_6x4h&{r@E?!dVwQ8_)RF8)JVQNIJ;2`Rkgaf)r)tN;3M`*@A9t{#^=}+zb`TJ+&yi_P6qVYyaceNekNbvkW3MpUyPPkMD|^}y^x=S z-ve%G5v*fU;w*U0OF#}>YJmsRrm~Sex2G*uwr@{g$-v&U8s|Ce2Wvl!zm>sS>U&f1 zfXmR3BR_Cg0wfjndmQD5#>q_oyoR9NM&VW%1VvoqlM#t3k^qHlMB z2Vbmu%7C(0Ev))z1my%tQz&6?6UkxBGuQ-n5W^mGx{5zAM$;sccqr}s53@ze(?>q! zr26L{Edc|%FWL)7ny~}_M*jAFEN^@R`Vy%PFA(hszQ)T9Ya*nPicz;AJn6sJ`%XWa z0xq>~ujW_?gVS%SZjp5vs8!>CMf@?>hc|AGX&?H|1t-MRsRnLk?HcOf2-^iJKGR(@ zkLkzpBEEvmm7e!a7f=ie^O4NKpR`TQ7wrV67O%wuWxRh#LuG@w=96+C7|8Opz2}-+ z;3}dN`dl>OmoK*6>Q@Wk$OOzL;- z#f@I=Tm%yLm?U;C6?`tQw zseQpCsnh?jv=q|imbQ_Ava+?hQ1vI$b_1fdU+T}sL}??PKx*w6g35(Na_*v)Qy`-3 zQhMqBp9^s#>$+5;Fz4frO2RMS`Lwh>k>)C6ECOZK`*GCi(|YMjMcL4VH4=%T`K8UrZkFW!mJroN> zIAPZIr+DCF|J|V-x^bZk7ZFc+Z`}_U=+T%PPd!O?-@AE$I^LjsT?|Kn? zKs$KA%$O}*Dr!`f)UwQSbykFAV?YJy9l=P7g1jMp{ zF;1#ie>IY~N|<`1bKMS+`weeCU($JCioUX-Qz24Fw0xEK6=QkFDT-mKaQ8eG7~Wu| zm&97#R>}oD7OF*5>SbJ0;ZI%w_ed+F`-67{5_>JaZo441h@_=AgXekYYX5fo8iXoO z<))#6hPk&0vz6@Mj9#6EaCyTl19VV6b_!?gSil?3*fnYcPH2SONN?WGXy3Skmk+dJ z-6P4gi9@P`rVfQ~$h7T*3Tn?Bwt|IF;Dcp?u$mkvqMgx)ZBdBx@etqXO}pun@C9QH zi67XyF}dkt;YTX?5oom4eTR$x6<&Ej>gFb`jvM0 zDP#;L@p1QArr^`f#cV`b@FfP5`+1y~Kb*l_x@_!aEeoUZfjx_>~sIcjkrA`{+o;kAW>O(_rHKwyik-f=PdT^mc?#D>CrV>Xqj{=gK>1u zopUTpDF95WkNiu|)P=F+zT%3&Ir*F-Es z8d9nPGZzbo$)1ulwQ*7_E>ellB4;qM^xHaTx-`$4i#!vE&nSWVP_*AUy1dAl7N%U) zzh-n^^AkGi1tsm-`$5N@=6*_h{jZXp{3^r+{)A7JW3Kd ziyAHVh&+E{mV^L`%J5yG%X4<|r|_9Uc`n1O~TB`#u2A7~O=QIErwkci- z)j^wulmi8|-v~Vd%IDSIr`RBVlEWxf{*l zO{q|TZn|I<^?#Rt1o8$nq?IP26=81&4>-^qZ#>*H;+&gv=k_=YuHL`;ihFU{ zVo6Be6gbntg{0|G`E`U)8<<|!<<@?{(_RroJ4U83fcDaI+b_G=s1mEAN!f8d?H0AAwt$2Hh0Eu`BP7}$W*n3)>`r99~l?rfL zlOikkg@U&VJwF|NWKAK(K&UL;6?*nMJ#etC#=4MG*_#zFv4%UAm+QkmNUrq`kOeLq z=cN!VW_(cnRG&Yg44=DN#n2i_Stho(&*%|xODR&8278-ehQs9tN(m|_Dy2{Ip=>9< z25dPnOkYl>cLtLCspR?8Z|jqwNnmyGliZ*C$JA%yFJNc!CxA)DaRLN0108PW$dx(L zmhjx=UfYV;i;chc8bB6|*#`fj7KC3onRxES1^LZ@sR7P5cG1-#4nwaz2UrZq7GF z`$KI+d$Rh8J1u@^uAKd$YVd`P*m^je#1vVA@Lt_N=!8u(Td?AKC!>=MKIGe*fJwe;BM%B1pQ>IHBDH=xmKQ-S@w}YDZ#XfR@l)>x)ow}<*T2X zduu574|_jHr;jMPeYsMN+QAmD;|+U z%%!;2Aid`$`5gU@{#;@eVm8>XVkqqgQl0y_&0|fA7fUB+sjlY-ThCZBmP0K)?jt>a zv24(ssVlX`4282O!v_g65H8?rtHp9Cl@!gp*^mk-J~Bp-llZ~e_}Jd9bEmLRhfH_^ z)mu=g0VF%Ou!%l3jL#PvkS&yM>niBB$2#EeHulSG-lD?Boicic(keN%q9^@ieEL$g z7*s}sgk;j1SKv;x{e$w8kg6r9)rc6tGtN-=2vT(10P zwNjNm^26+b&-tBZz((ZF@ZaDMgg0^9ceiX&5NRR}MhCuvHk0R;MP!V+4p5_DFZcY? zkZU)F(L0AYeu-I&6M5Qf++1CyF;_zLW5&f1m~u7Ismr2e&tzW2gS=#LRsX{hNN@FP z(>u04z#*2ufdMBTBaecl;pEF~BJ*LE+}GO~^WuU6502}o>gXkFz~aMffK#Qd*GtbA zlcHnLu$S%ayV;L>0YWkeEr9*y?I(b@wEqfWy}vREBP!Hkn}7v@Xp}Bm6*P3Le;CpC zR7!_ZMtL&aKbPOh3j%vV&2uU^$_tJhoE-Rl#*1cW)B6_HeNgiNFJuhE5nHZy1^0~5 zBV@UlNe*i`g`Yc4EX1eIP7Kt<61XaJ?v$>s>-`y(F}ZU0kCf4f@fj=m@3!M@kAG{a zdZCD&k>?BWo>tlY?fwZW@Eo#O2%biZLC;WSK>&FJeL#AE2CHpAzd$^}ibAmqT3w+Z zULqcNv=qUt*z=TW?>18^l6Vch5AN?B^J=`8OynWsCGdX&WHwD|(ycF~!%7e(D{WW!$ zC{ce}EXqJqM(pex3R`GgE5uua8R0tmHKl(TEQe(KX$#fo`_^LQ{?e z=3Je3JFd4CjD4#UY>>%&7p1Bm{odF~h4W!lm-q|mV0E77MffH(VCtEZV-a6_$9JHR z!ga-**f6pZKIwY>p`nH!_EwWE*9+D2GuDIPCtmtW^@w*uW%rycXRGk^UJ?CN(%qTfei%yF^R?x@2WXQ z17;Pxu7Zsl-YeMq7xTRNlLCrWP~|luYwLw2l;>Tm&D7% zM=_OkW)M?>6z4u2Rh;@n?I}EUK8B#$cr1+X(Sr7}f0o|v_4!<{^M~Qd|27RQeq473 ze!1GjH)*gEYM(}aw|?GwG-082HBziZXs%P|5iW?ohsu$WTqopyvd5C0r4dztMs|65 zTKzd`pFFS~*{(2QY<=k7u?zF~hr$Ac!_N-uZ}1`qL0wW^jNGiO??xRSe2QgZk#GVy zS{QV{e90@_P~RE2p50e=55EI5P^asvMEm3NJ z`Vr(D!{31J*LSn%UF$DY#s!?+>+XpncyZRXy?*-_V2X|Xni4^ za+U6P3qsAQwe=~$-Vf`8is#F|{7&F)WV}Hv^1Szty_jqG1pSb){ViCW@<#PRq3mX5_a5b z>x4Rkb#<7|wa^8xE7i$9pN?2U=pVYoXE@C1o^u|w>g~g&DT(`@wemY>fX}`?-Zt4f zMmScwb9I;~Tv}^f_gi@OnQ-kgvZdo1Zn0k%cS}c5*^n^pctqQ}3@4P$?(CU$HL9)m z)$8OqjZDtD(Zns5zKc{Xm8!0v32~V?#P>`lhU-N~c z{ESgfUr=fNe`l}OJ7DAf+k>f-&#%;j6m1wJ~2rj!fJC>}KQVoZSLZw8?W*E-1fr1;5vS?sfka@WO`s zBVwYFcOuJE+Apt4wB33or8o82;0*#&6vnrxIUt8_hy8p<=rPP|Fvs&mm$G)GEb(Wr ztHz@uV@9iE@T;ED_Fxx6i;x>ttoY^!s}Ozc;L5n1U2}QuQ=ATG)0(6i(@(w7ui}C} zz8sH!xO-twzB`HS26;t2!Fwj0F(MNJUOfEN^lr_|r$^8X?YdlAAJvq!|6qcP%u^BQ@~({U&vNu;K2}P)_#);&)|Vv7rVf0cqg&P7h`^RLDnN zNBUl)R!L*%n=nT}H=`3$QFWt@JhKC~VAS#cIts9F<)C9-JC_87(>Ldu1=)a|(zy2* zP4c+rw1{}ZM@X%>%^RuC$K+2QZNOh`h9L2rnfHG_k_(-3i<_=B-mdB9i@tRgPck&P zM#H_U?ru$)H$#LDEqyk*Z|KdH*r%k@^gPsDdyPJMdMc~BhVLi4;Am6k^gqJC>OK0i zh_7a2WPxe&{}3~L=7;ZppTG6?ePnISmg4NA%msSoYWHIt@NzjT8qYqfyQT~6_>)ep zn1F5@K9u=_xm@VVPi2#azB9ap!{|77iO@j98Mk{wVMF4rDt6>nNsio7t*7)j=dK?u z1RC;-X4mZ7T5FZ=`>^n1b~&#%&j^ zdsMJ9AdN6LJmRDc9y4Vx<6hDdmx-9J> zIoF=IlU&q4u=jWv&;R+!3=gh9!ra=nzvZ!edzfFqx^c?4(?*|--+Cpn#&^MSJ`RE) z1tb<^h2vfmnjyPdS3F-9>!9@@wA&(% zI=W&Gr`s+t=7n!jJ%!#6YeblOE_2U*tGQNE#Nk7r3*B+3GY4i5H}cmuyHha>D)7ye zb!(a=kB_v37?dGDh_m2n5ljC=M&0;A_L|2F623dyIo^BrHARsJ?y%4BQPFOl$oP_x zAfwqnYTSH(u&nnkG@g)6uEN3Wnad?3z$er-iJzDI%f1V@+M7|j3Ph4?-}wGi(nl)k z9_OpbA-?dY1L3Q}rfX{)u@(0KHKeibP;XS#d;E*+UEoPzUlu%Xt%tPwYqtCAK`@UJ z`eCJ3x%$pbext;S2Hq{)s_C|1n=WMBm~PKs^b!MP&c1wXk+C7^A1!>`>frWM`UXUy zNtLL{JLIxgb_qIi;C>&gvuwq%AUy|CrMK!oza;zZ5M;`N@xfkAgukZ(f3W=qb?;Dq zRSsB^t|6mIPx0*^oBe8AHtunsZ)@?f*0_zPeXEB{C;CB5lMV(yCPW5x>|i*9`q+t zU{(X?yB8Zc_ZL`q1l;!7N=f%8Po7dmISvSzbMG@Yq3MWiDPL=~^Y*xo z+VAZid!9SJK124Ba%T~NvG&*`($jD++Wl#DdM>E<#bTJP+Ygq&?w>U7>hbb~?{n?Q z#jxLHjD}AVU)Oej>qN5zaV8-DnU50!?;RZ9&g-*f=7QUevJH`76#(3gvtQQ_G5spv zy(2wG9B0i|ExXq2LNsMd!11#g%LmTuC`#-FcXA4FVVh5{bf*9J%(N_w+jBFz3&6t^ zWHh0>*(3X?^bY5(XwZ4M7G#J1&vrvz_Zm-Smc%j}6ZuvcKA0wOF|Z`uw~#-fRdPSQ zZ2LYRa`mw=H002+Z9fb^4+Zz_+Sr=CiMT<>g`|!2Gu*aQO^166a??0~m2W(No<~g{Glln-N_4jUYwHV#t7aA6CZ0`P)|%ma z%-3G*aV}yW>3A4zvD5)`tedc0Z322%KnHvnN;$r_UrA!Gmrr?-{(uFq@WsZjgCsWI zQ1MIZfsDJ>C16uND~gt#I2+kp&$-d&5uE)SS-4_n2F!C^B~2%QZIkK zc`3EWs<+^t$oTCl9d}w=U@ui(vT#RjHh%_2G@(2$!*}Ujv!XoV*MfcXulGP1w4qu1 z5~m<3qKw+4g{`-SunJgK$?SF zw$e%wfQ()aw$hB%_Wr?TL$VQ5l~QSRSn6TKo75X_*UCnhq!GR=zhsV-%kK zM;4SduqwWK+`oXo#+cXnNW6wSaM+8J#_E~XpXD4__ZmU>&=D?9mIdB0P! z?1SW*!g47+O|My~6>^Lp9hifV{s8p5Jq<|5n)&OrRl--&z zNYq-325o|}wm)+!uL!6SOe9ta4JxTWzLxLwn(dZk8Nq@c7n>ql zIK=0eKM`LC((n~um#ED7RPgI`=~KDG9*j{#%iP+{3bO*I`B%+1s<&g$JtCG5eSv(~ zTN{;5;8j2_LE>ZCJVo#Z2Xed#Hu_LR=F0G=0Ed)(wi~1Rtw%9n1^PSJV(G_MeyI$u zNtEpfb|8hwT-oeC*kpyv(g~^SWDfjMX-*-PelA@4-XJW29}jHmjsLYP`r-N3uk^6X zZjV(+3eV;c*Ey3jgg_%lkVBJpnsIM$f5DDc!?szWowYsQ=M?dJ11$0~Z)S`25p?ax zR^w^XhCCzhe_F0N_Au?_AAZJ!BY-$}4hdag& zni4kG`DqiSd!i*0Vpockenq2sI$Hj9#WVF-ga0D<_AQGar*std z@+JIpiM-@W!R4F)ggrfgbsxR+qR*a$_DamzF>mXnT^l_`9c!BOR1)sGVOBHq|KiGs|Z zq}d6o{Dn=c=pt?tx5YQFZ>5p{&eo&}a=J#8Uk@?t*1A)taMSP5fPQ#jWgI4KP zMwttMyhk@)zPc>2b2g6Pk8lKhvDpn9jOMI7Xu;#6BxLlSY0)Lq6F;B9#cf|jJR-_O z)3T3>5egej4y=FQHyFP5gi9~@Z~DQ}i6Kwr70o1VMP?}UA3Prs=8n?+ckXk!h^{|l zS>gzKSc5YICFz@KPe{z_RAg_q%xW}5+~I<~!?}KyLfP95lpJZ$YEHjcz!3N4zX)Lf zoh}XLOx_+ZLWEXCMs1VeD*(CZZGCVgDhYRBfPJ&%f`&{pJvv#lw~vdrXFNjPr+G(e zcPGQ$c~4-GY5@#YU@aVQ>7bk7?zExVp2fd&E?G&DPeYT9V@&@{ttn#Ar=Pn$u^H{R z3M*^UF)$P%hRIMK<)OO~D#VeFo7ITh^;u1B)ky?ds<4pdB#obF zS5qs<@}{s;NgH!?+;NoXb7Q~{q(-wX5xL?0-Pk4V` z4nPcV^E84t4EufEb0oyd9*=|`yS-g+7cnZ6z56;ivdduhYMUX`=Aa{6N?)0A220J& zwh{Rgl{Y;tQNiu-{qSSvfStnGt#Q@Zy-69q-Q+;#;l(`WagCQJ6qjR3n!WzJV#AO) zW80J$*kh<9&f#0~^Q@wZi=Q*sj6szdmAPjy0dYZk<}WdfDUWfHZi|W9d1xi(9Oq05 zwl!?JfYIPr$B@=l5i!jRl0VM7{kPXta%Yu`Tj_UzN|!y#)p||Dhu4>}K6*XDE@U~M zeVVtl=y44}$~XzEuwP?3j^Uh6A>Bcybx(7je3cH!S+p(`#`XF3;@v^k4Vp`@{1*&i zQr*YrQ><%m!gb(>RxU7Ivhep>oIxX)UEm3$ak2}i75231#_mZ>CFBBp81ts~9eY>$ zN`?LDzj1+WBS`^j7f_lhH2ccj82mGj-y)Ls47Kv>yOC^*j-&;|M(~1B-du)bZn2aj zGFY)A_gu%k?y0r%t`|Yt6cR^YKx=C{&+Dmu15EUl4NFcmKO((m*vOK6F6SvSZhXY+ zeB;qVPcQj7B;BqfX(xrUe@u-1YiKr&7Qg%b6kkSAcpr}WJ$oI;lN`40vfl%YzUvas z_-Wh?N%`KCB&kCcR~26S*#voP2iz0i{4J^Hd=;WyKh#v?sbcVIeTZSOqQsK&E_Oe$ zq1zyzve#MTdPdFt0q$^_LX!^UhC`44SgmGAoesXg$)XuA0xo!q=f(uN)qg7TbWb}L z_h=F_QbDZXx$cdtysy;t9aU?R)VokDTpTgGP_!igHqy$UrL|T!TJVOOGj_VS zzxT&O+YP=f`Dr4~+_5gZH_`OHU1<$bqkkym_}#>R-B1_(%zv6Q?b>2So7$5?%;PYA z=!U{B;pwqgsWtg5sRAedW?@5~Z(#%9+q*Z^4qS{dqEkCIC+LTDltNxrR<&zaadkkm zOYiY_W=i|D)qQGcj5$-f(cI-VP>{l}196ivSw1IAX`>q>`p=_I;=T%J{6hKUcDWf0 z0(Tnga5ad?15H2ClQB+z+gDyc_~!=g3r}a5(5GUVGhtbN<4_+om^ho}KT2N|>G$R9 z>2{@RtWR;kJX0hPkx-wP|a;+UO}HqMW`9RX~JNm!+PsM z?Z($l4IFJ)PK5EAz`^EX@h`YdT>7?aV;3BddDph3B!2X@9B{lbUV^63c8rx6t}?5J z5ee~M_UjEbSKat+%H9vOtI>}Dtf6U{uO(g`Q=dG@sgYF~{HQxqzBeskQ7tDoXEZQ+ zw_eI?pZ1K4lXtmkC*bcWH0T8M`9jU-WjUpsLS1xzRFC4$TwjXkLN!P z&E2`OD4elwte*Df+}_q3RvCCEG4Gj4u}&xjRji$&t&4wEU-;qB`xid~1K*%20tq-( zKzV*^?k^c%N0I%ZH#b(eqANn5{f`m9XZC-l-F)bKZ{hVV_F0}XYVPrapf0b%$k#lH&;5VO(XY7^}zk&P!)PI^PZHY z4r_Mtx>olQ*r{+UssZBzDcZA#?$Ai_*{^llEk2R1 z>&yejXLNuC}D{;q2O$q2qj|C&pcv4T~7EUPP4QL_l%QZAPJv0`v>`SqD_Ts zaXGg)de?{Cr!1mYunOhj#DJRu9KlS9(R?Lg33@c*wD)l)X;8@wIu5?+Uuqu_AuK_# z;NEa##4GxljLf8x6exXz*|4Q@qZ+@uCp!^#BF~u2z>+Hl6N_=UW~af!U$z~!v}-V3 z{#(ksD!o!;Z;u2BeZ9T(wGpETuWYW+UpjlRLHE}2MjGkK`xkf}^Y+(w#z3sI!&0~C z`}~{!V=EbE>KGZDFwXokGjXEW?cxg>uR=E#Dgf&IBJv{E;I~948h_u=7`1i_HaTib zFxLMnF`i<#^K|^3;7D3{TV$1FVQG6OBAp0~0#p4EoQbw5$lDq|@J z5%!4tTaDCoQ8qs>s%aJmYLSq77M{<#cv^GINh*Svx+H`?|PS*B;`v*kS9_hIzr|(Arls_ z2#-A?D=>!574*^&#O!F42k?0QtH;@;Eo<`qqekUwkeP{%e)H;Cm2FW>#Ly0-2T@c^^X3)9Kng@jK8 z)|psuum%zvM`=~H2OEx|thWdMi4K1UG?|uj9h$v)@A?YCxx2Ot8bI_~0v#DFYra?1 zCjgcu|9LUpbee6(3QWMvqg0JL(=JDnQ}^M%p=x2V|3VjQ-)&Xli@repqyB|1C|NLi z^WV?D!2HLyV?Ja)Ubcxi3*Jpo5x;s*(E63Eu$Cvq5Av|3-ATNR!45ajFjvTJvZFM> z40Asgh~Z~`gl#bxoZ#f@b8)D zCl1U=hyn-gog%N9y;eM)PVW8>>=9}xD|A1hn=V#H@9Ym#(%zlXbnXW_W98f0BkCb} zyk+sQHD+v*$9V&6Mp!%0@nx+%zyE_`P;@aU*5eRZpMNweC_+}4*;ZHC(&PUXCC~pW zF%Lt2{LKGoWbxOP(18Q+G_VV zgn7NjBY!Y9y35*XUp9p4bS;7G*mM2%vu^u*4*H8+(f!67e9OUwiyNWV=5o}Vx%#$y zSxP66??5XCl)Qfa<6=2gouYc=$5DF|1I9mo%uw>Dha+}d;$S6GqW`&6zD&s6sw{J9 zbOUwk$jv|H4~DV6(5BwJIaEEVlXGt7#xaFbdt7B?0j>{SE7hlOUf+4^v~~hQTe-H6 zL0Dhc{UvJo;mP|$OjGRV~8uaZ`9{=SUFG^*87$7c+owZ z;utrM=aQd-blR&Y$nJvuRwSL!x-f9hJljV~OJ1-1iLT1iE0;)PNa_#7el57Y=e*x% zDEFxbaMYKUP}5b&cXNE9ens%>5?FSsz9454deV2@ufo7jx+p*=xbACs74#`9tE+y^ zuYbqY^8C7+vh)t+C{H zhUUoRN(Il-uk%h_$Bd4%Enl-vctG~%ml%I)2!S3stDw*_WN z!vD^$RS=GWBJ?Jkj^aeJ71&ZSAV2uHMN-*!`di|S1KiL1c~1?wRRy5vwQ;U{WBXnL z+9>Pnk&vgWVf_I&8x{3IWiK^A#Z>R~o$jDwXZA`M%m8_5s89!#2$Mi{zae4Uv{Mn; zkYM;3X{uhE$D?z9LH=wVBDlr${FS`5Q{j_V<&?r*dT#3;Py)HgsDM{+$_W+&9;eH^ zk>tapfPM*jp-X+rnw}?bkgYY_FLGA~^a8i*V~a( z7XF?++BWN|I~y1E1Nj1!zN56gZJBohz%Q|rK^HRk!xwq0enNkX`gJ!&^55TF{=Dl+ z{+S0Ky=Mfy2)XvwgLt>OaN{Is0zA`!E^SvJ6_OjKFCeyMeC;9+0IiqTf58wc#oZ|~ zSH9$jEAq#__`)#ee{bm{ee{NDfylH+D!ORmGGf=h&2H&zet2Drb#>sXUqVjxj`(_( zU)7mC%^7@%V#vt93S7^C-#<6XjGPdn^EP_mh?KO-9)zm)xZsu-=zVS`YX{mTx?L-=D zj@|MH5x~;pqVG&$*ZS;=(403`_qH2|=HW&2p9%NCVa!WSXJO1sd;foWisnyGg!(ba z4Cz*6)KRMFN(z-_LL%as}B;E z{Jouw8NqnJ_5DB!N!^ki<#>ym`F?=`EJhnA_kbQM2`epE%?jNYO z083*JQKJFY{=~V#Fb+Y<4=twr`wC6vWE!Fv6^2R(_{{#3cZT_rai{+PdrI-RMC$`U zG(Dqq8q#%cI-Otm7Hx7D-wv()#7r25U}K>A)X5R(WcUTm-yvs>zDMbnKBnfl%9Zs_Lf29UVR?@!pkSI5t+NrT;rzb}0-)#fMq4Qztm`G`xV*(62qZbijI>9zJi@=?23-rXeZBMinxxPLU|mmc=lVd+-Lm zpme>gSA~Z&zU0IKJQ%{t0XCMZxS5D^2oK!nE1&n=_VYz2B9{Se&?UG*>SK~We)-xO zUiQf^!KE|ntaBfbzyF5c?U48xg+5dGrCab*wNtiQV(NF1;YR9^xycu91@_wl^@U%O zM8rK=lOMw*Wq_HRGB5ue^Zj!51}z=DI(1adU1X21g2p1&h`e^g0|t94)063d&Y^cK z-eJ&PwtLh2$zk=CAI*c(z|{0GCM`FCB=-a2x{9)tMd;Xv|eEy4wX*69;*kv z)|eD#4)*f3+%id zlY5(`WDc#L9`4NH!d~V}?e>gj>^@s(^um1&m-ZAnI#U<_OeiCkZWSZPH~BtXwoE-{ zSY-{M%l%RV*~QXqi$T>(m&m0TGq^iggNoY4m|&}Wx|e**sU1JJmQs-Q93Ucd12?rN zk%IiblLX>ORB`RtQpFeTc_=XA`_3_>4Z4gYVnOw&DCHC&Iii!8p0Y!)w=&cuZDw3H zw~pY1WD^g&NwkVUZtnF{dxND`q8Z}|aIS6bKx8=UQTgut5<=;U|4yQ}O^ zt5tllhlt-<)Az9g^WF3w1%H z3E~==U@*bBDeD~B9(GsSh~lvUzpm63CVA!PH_7gCU{!7h(aBKK)KY9#AXdAXHN17f zK3bh}zy3bc?^gY)3=9^)9oSr9ez_yR**T7LZ%SV^d=^2?zbvcrEb6~w(Eo~P z{_RL>>^wq|gdU_NfwoJe&5WX_f4c=;@t|B9PUaNitB5(f?atUeB0`AbxedO7nmhFE z=a%ARTl9(ZrGH?{bM}r9Rk|$U9PVb#cTcw^Vd+=bWGAet25^~`4%i_ACUOX%9q&nI zgT#yy)?Sl!%J~6zb(g3HoM66<={}`7-|x}y!Q-f?dpkRu3gG0(nC1$j;};)Y!hR$v zke(nUy3i*rKOFj(0EpHuZ)P;h19;jm0lqDVAlQVh1Y~)*+H*PCPI>-=~Dt z7iX*(I>15o{fvQ3rPl{TW6eE2pt7b`uiR2Vowu8nKREIlqDmmUG*uuPXPhs?(m6{7 zNezWMC1FaGld$TDUu|!}`RzC4o9DgvuFNUX3o?03<~PJ8+j(Rc`P@t;_l|)kE>>5i zd#9PV-G{v<&e7APj7FG6Jn{n$; zFv}btCsHnB4tL#s{-iaPuc##vOJZvW=UDom%lvb=va=-Gom^x)-T67`EwbD*@8PWM zt4@BRgE;1ab?EK74)6y9`OUBMt|u6RS^!~A-;6cegv(zElt`Te`O@8$1d!JmL6IhK zNf_);^Zr)4(e-f6trI2`yB(l3Vr0)@>Bevp=|Lm$K0_X8Jy}LIH)eX@s^6>PR|q&E zo=W}w$ac`v2shj_*4YO}H0n5BA9U6~6JIZSyXB5H=kR_V;m+jB+IE~Gck52RBnv%iNc9SFoH`Tg^LB|) zg_EE8!T7wfU@nViU*FJUGRS;iaaJ|q9m_fq=>qcx&2i-U_q1~zo)ZSUOWd%(1w2IL zU05r6iE2u3It`e_oQJyte5Rfm3GzXWE#>m-ZEKO4VF^u>?=S^9h3MF+8zV6UuIZct z%(9A@RuMAT7^zN_&lqp-;8fu_0x0b&4;E#$#+I(Y!I~a*FW*ZkMQJd|$^1AEAqQ1INbWQpQ%EIuh z9{W<0G>jRta#qrbw}t!U!v-))!kEhx4Bf=W zH=0}y*WW&HQ@YbB9aS}RVdfDYf0cM8ESdH1PMeZ{+(5ry=D?+y96K9SZTVmqz`FOX; zdsmmbN4sWzKWeU%cX)`sf!1mg{9OPZlAOAKf0eLoQ_y$^P^r~VLUhqT9GB{=N)CrF zgrdN*-DfwuTMkZCw@5{j+j$RVd*gHuqmC#Z*SWVTzr5)h;aX1(JO1W)oZ6`^>%A-e zzbLLO%f>TCk|p?vd$(@+Ea2`%w^QBPflnLLU|CEjq2*0>-JTOo@I~({$U0JVMS-T2 zueA7JGE8pkPv`cCR9dNh z$UZT{zDhZaeiBimf=>*bh)9rFGo;;D3tcc=;{%4Eos)j+BIi#Zb=iu5+}Sh7x`qhT z?fS__5dH~Uw81-jb;W5V%A`-@Za^W#Ufo7(Kkshy>#crb^0D2#J>7jDK2Y`D$9cEd zCOk`DrPI*%Gq*S&_1Pgz{RsB%GKve=on?&}WOq^?W&hyQ{oJ3J>+Gj99wfAZ?5n@e zX9%$Kv*}pOh;}X4loq#4ad?B_+MmG10jjC@)}4+(QmrZZ^mj41sa+wA30o0i2DlBm z#n|X;|75`^;C109xO80R)}#;qA5Rf}lh{@D9>OU*2TwMV14szDZad7Nbn4?6=B!`n zf3ng8{}X)?M%$ct6`M~3y;w&OySx+4UqRLEOoYsAt-dVD>j;Jc@e-N9PtM(8UE_qe zDbaiy2`6D<-iU;`nohpLG)F=i_|A}-*FPz=kP}DsHHMYtVV_?lYrhqd>_Z7H@BtS7 z`k){WK5@^;>mo+`El{$rYE7~WmxujxZ^q{$S@W%g&;lopdc&Amo(KPX&&c;8rubq; z-P+RY z1dkc3cU1oa6jJb$L1QNYn=-r8ou(WQvF67EN%Bv|8puzK4u;+8)(!uHJUeB=)&iB@0I$$f0!~kH)-av6}Bcsqr9)JI}e!}pWvN= zYD4hka)M^Tqieubq7~ae81ODKWOcf4Y4VtaZGXi1Dt*3Rfl#ZpYgL7{2X8~6cL_1H zl5~B3njfb*DT5eH*$_U=5{2UR#HwGSqv;^r$4o*342L0Ohukis^PaU z?ePe$*_`kx%L+H#xBs_Se4(IJ%@iqkLqq+U_=8tRTLS2Zz-Rw{PrP&iSCxNNg0Q7m zdaZP4b#<*<{h1fR1L&$&*v9JT(7F4SV`;OU;Gi}}nS$cuk3wTe8_O5at z??+_5c15H)&XcOuaSF^D=&)da9*KVF9}s=vq?l13j-L`Ri=u+h@rF5;!`AF)nj%N3 z&gw_EWq^Zj>K?-XjQjH4q31Xk$j}V|xkZtOncD+%N>WU z8E4@6ZwKF2o=tz$*eJ~#dXwRLhtYv?f_Fq3ZQc_nCL@!!65wYe zi@SBh-lTJ(Cyi2pBn<~sqxN9qHw&nx6m2vc(4C6NI(kpGrLx)w0mu}6JE0#bv1YyJ z@sGXvM#F@1pJf<>CY{{OI?A;3(qHVenanpG^!uWePEP~30W}7ujJk z>Uh#1Cdy3t2d;_lnVc6X_FB4|Yrp>4 zkwtB!b;poRmd+^Gme{-Y3n-j=^-SyscZd5rECG8k^4Vvn&p$5z@#ypC3i7GizHH)K zno5@H6NOA2ig93jK+*T>E>;XXj z)jl_@T>sYnf`+bWewstrHsAQ^F=n($xvkGQWAfPe7jFNocrQR4*%>aTtIioVsgOKy zbM5&dO+Ej80q9!fmjCeO^YugAB$wZ6k*>{tfzio)q704ce@2a*5=`1C4-xn_*eUP#=&G#6D_k zG^}Lug7yf{4z>dk@fE$EL#Tr%?udNm`16?>Ko2zn&s0peY>E?VyeQNa2bDaIonBNt z1T0CF>?OP*w|Kp~J8pAoPv)?VG>9ir6DrzC+3sFwei8ViIB?@I^UwH%yxNY{re@22 zK~12ITrxeEeX;Az{@iV3vsHKkZi3I0X2x)KJKF^_egV9B%lCFB6jr;`G}JI#{x2Zr z*ixkrC2$M66pC}FZu3m?>bR>dVI|jVv=(~qDO!RN%JMRA?cyJRA4&W* z-1k0LtnAH`tf+p_PIElRa_%@K9UN^dZQKu5Hrh~}!Ct~KUfJu~EP0)Ij=keT84j=B z!xKpF>_>~pF|h}aUq2fXa~1x4>+V^?71{pX%XmzSTjJ)d>Q-Ku^lUmak9kA+o$L8C zSH?FsGepZ|I={aTd?X_jb4br=)j=C57nH@Qt9#m%MWei4-fhkfsxl*{K-Y8Ddprv3 zueyE|h)}6dCY*Rk67tcu&#(JTHyQSsOWF=)o3x314hdriv<5}XmTc#L?f>@nPY`lD>AwFSrngmgs_=xwe#)c& zD6|rnH_i0MU4TVMi_D_x=MJ?rpljHusAv*ZJ6YztfJ?Snn`rf(X+purtx}VbS0YaX zL^4=j;nuSZ1;oVG;k3~9q;PTd1MmB${DvBB(A#M`gqydkUl@A4>gYJfc)9t9PpH0p zWmv%QS@uhR4!K6A3m(b$qH;=T6HgJA&6tWyP)1bKyxW?_DyH>cId3IkW;jxDGX0m- zBs1!3!pp=SyXccb3BpAVi>rExFfG=zVN2@oE-2!C#sSr~qx-GRiAkIeNG?uIr*mFi zg{96u-KNB98W3*l2r&Az`c#ZrwwEV_TC5vE?^+h(+ovvpXD}(_IyE+~vG( zN^Y3+xkhON&9SbAt0_7i@E^Y>Yegz^FF^8+DCj5t>fAq77WL^$9}8_{!QKr;yx&2- zXKax7RM$e;3F=>@J^DXBKmPW*epC`K=Ve;@%(+p_uVX?G*yv;3D&G!|iEE5nN5ke` z;3|reuzPa5)&b&eIv055B+wdPZ-!e0=l%?JkBlZii>%;1PHMK-rBN?5Yo;IlUga=? zcm&#r6j_1Q9mP4+^NRAY#Ct|t)F2{e6#!%Y*y|QrNROkgBnEVrY4B?lt*S;dU6T-r zw9l}Pc{S!~wDHdwf3xk$$Tpvh=6ToIuwI=lMnr;5Dzf#`eD85loW-|1 z@w3+EXx_Mu-RYv&Y8{Am16E*|`H@6hh@90TXDvV)o`S1EEaW?L9dpb~JbQS~7^GPg zg)IXm6-kMXCAVJ~q=ymh-$W9uW5;}9#sNRq@?|Efse3l~YnmB(&eUs~W8lY1Og}W- z(c~KAOV9N#+{>%9ViwzzxeX!>bd$xr(w*|ACMs;|Sn188C9L|^6Or**=pn^d?3d-^ zrf;EoEwL&E4S$qMLOLfRBP_78=Q4ixr`U7(?!aC*iB?9pe_Ob!sk}$JJ*!n!eb@xA z^()_BZGsEfJk7B_T^W#YAR8#9*#JImF5H}Twiq?R&3=&Lg&G{-uGf15agTQY!N&Jb ztf1}X=M)r)Ux)*Uumfp-&5`!7_=s_4R}GLA(KNXCeWugCKzO5gMYq~aiYCoG%e*z8 zdZ3v4%JLfRO|M*Qg&*=G%v+~@OxEAXtW>P@R?go8!^gsu)7_U^9wVDT13_-|69cg} zN*xH2?S%Px9g&9hmcom=vyp|~g3P}}0c{U7?ja=>$B$*DmQDhh|2)mP4{#DJpD74B z>HxT)pzyey-Q8GpLza;-I3VFP#Q)Rb|3t|X*ZxOmci(Ok($OB!1;AIiXFbbJ5zN?&8KZh^w_xDkYbDQ_5e*&0>v=3N+M4<-nQKhpKr zi2-&GG7Q;CA*QUW|L^$if5%j6&4k4`b1K ztOQ78XKShZ#7Bytg;K9oi-H-t7t}pePFMRwu%cs$Cg$ zGw{GsDsw|q_v4`ozymxeTXjOeVXhPPJNZFzT7%S?`^ z^YPss#)M6Ks*!IFpFj}DkMrhaX8eNw9w!LTO0ES9Zl>H~@OBteFHq)}dP0O`LOH*{ z#C+w&VOp1)mG77zMN8TsZ@Ap^xthY7t>n}BSu{wpdQo}*AnCsj-KoKQ&9q9XJW-to zfhRZ?bFaKGdn@z%t@34rPhwb|)O2lwtjkj0*XZp{Y=ew`y z?zVXWjQs5UHO>;PrWdTq%t`YFHMsVM@$T5(Rb)Rn1j-;l=H@-~04T#dX#y)00j`xg>mKY_ufr0~oN_o3p#e{!)Om9Y-nTHCqNB>kc!U_ctnjc9ZfC{h^byaVP5dMfYY# z;^0T|UW_lE;A?VQ8{$NZv((x1QT$U%f4OGm%+B8!zxUl_k9iLiI`6hrs) z{c8@@hQyvbfdOl}F43ps3m%+?vd6Ux!ZfK&)dCHri$ulV7@7?PI62mOaIA|@Rp-}& zR81%Rsa2o#7a(2pIQSSJAT9dtmHQoMjTe*n7qJJgA1mCL42zE%Y|NKCtWKQWW+X0t zKNult5^Dph;QyMe$(oAh`y-n&1pkC5LIdq*#dRixUy_Pj(#U0_Z!kUHVPNjCJzemNDj!4mL?^Ua;1l_uP3DEuJfO(kW zSu+cn;g_1I=X-S96~}K9-xYmR(i3#IDO#ciX7beu7o)r-BK5WL`H?2oe#mIx>I5^g zO3b|3y)WWSC5zT!%)U`7Hd?L^q1ZzKU^L(f`uj=->GTPwe#H&ie;c;xwGUZJ-%?Hs z9oQUx5Wg(NS9p+*2-FOFE|DN$OpmVT{;w{oT04&pZDZ8hz5C z<+`&PfESFdb*(>k|7~kuY8{pAn2r2H+(CR~Pu2fCA7}jTz8puu4*5@1enSrG*19=` z)UxQg4_$Qj_(46{$~?KJ)#* zE#s9746F-{zu9@=JUQHo_wI+jXa{qU5xoP|yZ>bWQ}c*PKc{noFuv&P-tXL`%bQ%@ zPKU!~vU!oE652p>`uqN!HsYN29a+EnL^D_{QhH{K(9#zM?5!cP2LbAn0bOvDpQB)H zaYIno9lr?~V~(ElS{*9zOWv=%PAe5oPs-ltm5lzYsRJ0;D_aVh+|ISieq7mDv4flJ z6{c6Pc$ZWFWsR?PZ;Ylc^7Vb(jxmv!ZCCXn9xr@Zvw zf5)TOqxfk0Z?^hr4o=3dnt|@W)4b>lou?~@0ZWL#^Xaco$8FgJcWjF6@h^vE8n39p z*{14v)BvxarWw|42iHL2#69_>rcQabjlcI|!->EN{ycF!oV*r(LVR|v_v_xORW;L^ z^%GU>S4sXgDS64TXEOSU4(cx28YKao{^S62=$riP&@BwnMYDTn#GD;5+v7_VS$&^P zpUPXEJ+}>X3wb?r!nFBm4t1ej7fFNYp&$EA`!8uV84j;DfOo_9^e%SG0e(`Rvmd^# zwKmIaO;CA<@B_XtQX$(AV+|YWfIiKS1_35wm#@0iTtYHr1)zdGoK*1v;qm;6VCl0~ zOSZfDY_@8^^j^5I3*fVOWi1_uFB3Et4 z8Y_HTyc_!3c`(tw&U7)FS&wjx_!)W6Yb)F}0WB`)K|QH9snzCE+2tfY^4#b<$n;(K z?be$BN2czBnmk@X*xG#~szfrOYiq6@`}f}`CMlS&QzaUUhVZ>>_|6U@G2U#N5SxZR z5IZ5RbMFN%9N~<6sqIZ)T!CJJ6K%!@KjIJ8cLL}o;@~`hB2-aX`y;XT!%SwujaxwN zOrZjSDahE$-hV6FlBe%qhh@y<<4w?xx&5BG>MEC-FE_W`zDuDaI8D5c^_5LV`i3M$ ztJ{CD`Q~Y8vIUj+R^!s+Zh7?uB+4*+QKcX}Q8~&{`P`tL33p@) zWG1V7WqR<n86bf-~DzRHJCDU4yLeoPPeJ#eD#a3!%_=t*$ff!gi&=3oRcQMrpO|X zu7jhM@HtdRqMAcWhg<0-dxJK#1H5!hc*(>5+%3j;Lgd=(gPrd;eb2075*}6wGp9FE zxu>NH8Df*Xn%c$`Z$X?F%WcE$?@GRl;Yom#H+hNkuXZiL$!-{A5Y%($*xOjZ6FW=B zIT=Gh;IFWb%>nKxl;r}LmY#3@ub-P^PwQCU=}g?Ex_voMcF<8`%#$*HQ=jHQ6Q26~ zFx#xA7u_ieK>FN!o57o4amGcO0IofX-G+<~lIynU2)X+?@4>p!@Ektlyls3o$Pe|Gn3M#SN#ss3%_X!Sqk>}VPG z|AbNv+&4k1;5+8$-MqpS+rz`Ir7BjPW}51^f)4GZSxI^(iO54zWl-3QC-*!n9hq6i#(9) zJ@MvVl>#XGfjpDl{8YL$6VQ3M=C(WO>>1~3j`BUyt(FWJ@lcq3c`;JGJDMx5u-6*)Y=mZ-5Z*H#`n{GgjrKtEDV@RX0U z+9?k+Uu@^FwXoTWE`0MS_*SBdvpxdLP#ZEP+!woIBPl_9F?nW|Ztb5&?__kK9_(Gh z8x-3G%kH!bx_g!(o+@~-@RHs*vdzQQEgR8^o>!F9b;N4B=0><^E4PhpyU4$5+Fh$% zZfg^SyhRH|mwHKGeVzDciIiMQ`I)VY)P#j z>E@n+PSk$Ti#;5Wx<=#*)nOjnFkT#3yx)H1&auMX8s783bYS9FX$bq>UiB}N{}04Z zQX2fF+dmYlpg28|b2dd=PfbKV@(8r*8y6^v^J+0nak{Kl_PFyLP6FtuPMkZh?HBV@ z@Cx-2KGGUb!YEW5p-yn-9H0G7Oy6pV=(787J>n!EDhA{n(!;c<U#EHD8VT9?w_76RpKQDOJijfzR3F`?cF;$1@tJLKOhSrJQYnBCEo_$>u|1e4g z-;$19OBg*$Q_6;K&3(y96@!OoaVpb2~m_~v$( z`dX~^i9zw`{pSCrKP>{zKI_8eU;eBklNnA`32T4(&EC^Wn0rXfV6V75zNJ^is@y-72$Iq@m372a#Jf+xw^-eIKHqEQlo=vrf~nsPI%($ALnzuQ|vf>|B@(m^g8?UjOj z78h@jM}6|J$X22p*&1{sI-SnbqBQ}ns;xXmG95@S2%{;f}IYS zFehY%3?tfm*kN-U=C^k&Os}MH$X;H@I^Q#9J@JXJ&KB;I>^Vw#yRuxA$(;8wVwXIr zyXT)An&)$`7bD-~$o?{+?GmG9Pe6x%}}T znCviKoHR*1o1^_|Iu1|*ZPG<|;6%y@dOnbbPisOrgUvy0zrv|#10UkaFu$TJ(I(YU{AWMCuA z6vu$RdzZft3_iQoNvvDpcJPgsyP&4#7F2xQH0&K?UMi^9U>#*Chv+}7!8kGo3;nLR z<1SfEV9DR=U_w87&Bs+VE+(3jt=aAd#75NqBelT})gQGmarWKg_OJZiT-W7x7<2cR zly)vCAJ)^9G-2}31$X{#=_9xZ*c9D?y;zo6dFI68v%t!2fvAktmihOQg1_XZJHAVv zIq=Unq<5HK)M1<6CVAevr?iJJbke^6+`e14_+0Pz(pE3;_d8v;GEX-7N_J~kE`yTzIuurR;wwU?MD#RTDDbNCJVq>8)r}*mZx}sny$Iv?^3eK4*1XsQCc*U zEeeFF(&D`T9VN_lo69gd@^i%REeIZRVURQAIVXSy-jE`Sm9dfs@fb*<3GyAr`+qi3 zq~mq4uSNX8`HXp8&KZnUeifG`zA7LP7Fxu&^Mt03&>+@NfhPK?2OY+w@=?&*y_&-K ziu`9FPW*kkw2~@3>um6;Cw-_^m8gu($eZJ@A;701S1ON|)z1T?ZF$UdT&^u#T-?n1 z-?#0$9hA`e!^{b=)Zk6f!zBzuu}wDU<^E;a)%!mrb^ItYN9%_$&|3GxVxV46W%vh& z%(bfBCj3M6I4?WSuV#QDum}ytC*wA3Jtyo2^7oT4@tHDr|2Hj1nU_Cn_v` z4Z!)++*aDGd5=02rOpf8k?ohiG_hBpS^ZVs!&@*)hwC< z9x^4IOy%yRX@t5B7yV`R)JA9Q<9FxY$Z=eHAC@;-9BjZ2y~A(=&5r`@F(Lzs;0$Px z5~ztcgvmq6`QYVsPq|@LA zayZ622R{2SBpT~QU(yPNhY#Et!}n_6Zaos9d*JsPyy*8K@sQWILO3NS{&V@XIk)QQ z!T5J;{q)R)@a`07cVa$K97N0+MRdr!gIpYKn|!4Rtks`o*i(R*mN@fMVNCo(x+v4= zPRWVYo}g{r{SbLOfbv6UkiEi)-ZqhR zXln0%4e;6tbVpu6b|Du-l18Goud)?c6FmQ^btsL~2)MlRd|%?1k!he7&?@A)0njWN zOpv{bYQ>w1S4`E6rr`QG8@B&R-sDbK{;r;}k-E2%is+|$+1$5LZ(%Y}HWgHbneS zWuiO?Zit^Iu8&Bhe^a%Y1@_JIyRJuH%Vg87{)oYe$`kL&iF`_As{z^7W5HMT#s`LJ z^scrKeg06uP*6c z59a>-q2(j3c9TcDrd-ixcHu`Ggz&*o%>Le^JTGb5XR-~5{JF{h~f@5*42iaZeLA$9E64(#>snkLv(Bj+v3r~VT^pkC} z8+!iJksDSBxzkSqDzpwQ&?&I3*T{4-A0<++jxzh_SH!O9wrtlu?LOM1shc?qpX_Ox zuO+_!qktNF_icF`5~7cGmF`BPG*;ht<{$(&Y~Djs*^z`jsPTY)7(8MMaU83$at88t z8&rnd8AyD4Q*cRMqw*?v6Y?68g?puHbJSbhUc`I?&khXa-Ck|e% z{7VwcP-jj4pA}urKfdJF7`E0ev}K;}qz>byTO4P8!-SSLem!(}h;?Si{@+YcgM2`E z;|zaF-fT?V?V-1?4-S1r^BvcUhqB5uBCXI{q}`zZDThTPYE69i>wmD^Rz`^YEQ=Yg z(7C0L)u7$#%o>bz!-yEM8vGSFzxwN(o$DVq#@fRhjft}(^G_I0sAd0K zNWi?7!y$4Q(bRe};R(wCj426WFFLSYnE!tdg>9~$PHD+wRAvF&{fF?6izsJ#eXS{Z z0q3zYJJC{;g4RD>ulGov^Z8#A)YwQRykgYgGpi(kZbrCD9^Qvgqt_2~Uz4xvCUEXt zL0swn(=VPZjuV|LK+XiDf?sx*X+N)gIwV+V&sc4Vgn1(ph*wW=lf=o1iBb3d>=D`z z`9bqoDj$c;{g?=TitKT~^3Geh1<7Z2cz`M~X+hB2c^?^(%AQcz^)stKxBX>r^p6Wg zs9wCNbLEt7rqJdv;m~d4pY}i1J>sDf615u*s`ri(zq%ggnhJ8oMX9UFe1zZ7`4D&> z^Bc+*&67c)R@&sVSg2THT7WnPb^mXiiU)`me> zXST@AdoE##V|8$5g?n?Ie2Mt0^O(rnjMK!N_0)X-WwuDSa!Qe7Iwc>RgO{^)W8uFu zl*g9UuXNQUkKX2Ge%Q}m7pC{kr@}@F%6%E%peeDlkQ4+Oyz1S&J*y(3+kE9`0b-i)8q_}@mwCx2Sq+VV&LX!BWz=C3BU2M>0n$KK ze8Y|;Zq%@hW~L(h1JnIN1Dv_S_#Gv;tL{g=cz5X-_qj)(@4kPpGyXDG;&nl*j_Qf% zACnFS7O(4WiHu&ezkah~?s4LlQf<_@y^Tl}SJ2a}8K(<%3JcF?N=`2SJD!F+Rw5(^ z-0@~2R(6_jwCunXSt4VlA8=w~9U;~<_Ac{kiJv#^FT9P|<*Co0TKOm3o_RKH zLlQ!v^iP3?N=q=2Y7C(@M;}E9?K#n&)+*5MDj9Juqo*jIQ&P73#GcdkBT@6isRF=W zexI>43-8Wrk#A!BPST~8B;bO-oM!@AuNFCqwtBjUzwYnV+r3wm4^(Hi1)d*}5Skv^ zbe(CMipuEE$&Ypz_$un$CZ;(XL#MZls!4GD;QX1Pg%fp$W&zi@9d3qd9cHt^3f8G7 zh-p4&3b&uIgo>ioHA1(XHYzBs|1NR-cZQp2ymPrDmL_8tfaCEgUwZXTR3Tvk52?`(m~gc~a- z1GaS{cC+X?xtC_879vnLYAt8K(C~bq*k&F8URYY^rNECeVjRF2-N(#(Jo{hh$APW? zr3O?C4Ww;~+*}~pb~PJ<4sD-;?iexlr4Ljpp*dBsTyZLL{y>2OSdl+F$}|FZ;rp4+ z8xHVeKkt+7tWt>!r?FSIBW6$Sh0z&v#4vmrJB=DL2vO7#0{+0yn6i27LM*`OW!`uw zJHQg_LOo3}n*IdJoD0V59VyIRP7K2;mak0D?-5;FVzmo0#!8VM>=d}#ujo@}aXTM7 zc3yjM@k*FeymUS>IO*db4r7~r9H_caBuPrG57*K7Z z;2M+*dxjyK%&bA^$(1gdZQSyrxa*{?T!v?&ezv=K$Mgu3wa?9&m>fv1@Aqut+A?RC zrWOP?jrlM~o_I&#e1irnugLur1O=FAh;9c?@kGhaJj`rpzZktAz8j41C8rb9CjWYH z5>}2}=PA7lddQx{$hIS`w5{vy?O;Bg-G!I8=!Hj)Rf+pntC4KVIBc?TSK*-N_x#(C z%ON)>3cyE>%Lly+kaF`Na~D?nyfXG9{u97iHe!Vyaj+VKciy+E?CzUap-w)9_~9nw zbh>((9D%v{7xNc*=z}3$?9XvHUklj2OBPebJQ>ptwS?4RQTw&j>8Fr*aYX5n+uzHK z^7T???P+Wuv~*sTytEf*EKz((mJM1yTI;iZG~hS^Ck_MDj5 z5i~E8ol@d98zVh2D9svvJC5zA_f8nCugDfdU?CMqRc!f6EkjYM7G^|?CyU+T?J}$S zY9)701pAIcgmdlv{OLRu?@Q&$%YM6DLp>w` zA)K5}Exu8qWBimZ*l(K^$&O-v!I%c=J#Snkj17i&*yaNwdNr_5?Cv*Gl(g`lOJ;-X zdzr$npZwbeg$B$ro-K$U8p_6I@I0sWZ`D-(onoH-X^c-4M`u%dN&OU^L%{Ql^GrwL zpNKoQ`j7i0Fv0_Gy&)B%MO$6U-}Tu+j5k>1>+iPrn)4uh zlXh+F+=9J!bNnPN_YFeYT-cn0(Bu#ie)!Ho7>>-~kv#0_sk=a5!ML}cJ86UaNTXl)?74H!{7`m#HxSu{2nXe62={ywJQ>`aKi4m z0sn3E&Ugs-yzIPv3F8gSq5=5Njd!!K|FQyF_?Cn2T2?0yJE9a=5^o!X5bRPT4@~twzxx^{2zx@-jeuxuF^<0SpApRQuib@=+}zXVH*1lPP|ET ztm4gNeJtO<~9s)EYk7OlZsoGRGUpdAA!b zHGi-z1O5oWSv?HQN#uG_oxlyLlwh@*_(i*8(MF?k|@DsUN#b0ZDYk_U`8Vf zrzEur%`Nh`Jvj83& zUtN`Yi(X-*yD2X>l!%HF-f7;W+xyY3Z69(~(r_J9Zo%0r#FmcUJ@FD$PzdH+G_NJ( z5F0;s;j4m+h|l6b&7D1VMy8@p6_uxkm$~v(kECR}Ac^q5im3y_w^m--|Jxz2L^XWD zCkgrs{XEGGTa+}V7l4`}{**aDAb3G(ywVR{Qhw)Fv;Rdz>YGQzdHO~{ z>K4-0E(wE_4R?lD9>_7{VMMLYuF3r7SW^9hr8Bp_w0?n$>g@K34&3N4fxF{9X^aC; z>nd(N#2Qgj`-@hE@mJ_vAr@1B^bph?cwh6+5WIr07sN(oJ&8kocTVK3(NGW~a;tks zRzth+iNxcv7yXBwP{+;^t$*LCh6R!}fwy#0@&=ipK*=Wh`bZ*CqyzA8;n6~5ZKsv< z=KgpBBjV-3Ng?)p+7Af_q`TY;9Wn%Cr&Ozyx-Lc(Tja;aKKP-tj7 zs)hAD(Z9#PW-p?CktFP=^-*MfJ$}jlD(`PS3nQIViI-b8?LoC->(QMiYt5R^-sf?@ zO1T8s>rOH2>n_^)Xm_cnGvJ{^hsnwWDE`gxCBL zA#?1rZ15mbW6g==Nzx^|Diaa4y_T6{ZfDizt&MML2yfIp)LGt`MKwuQ6Ade`<+lsEF@=Dj zTUz}PZyh$gJ$AxxhE+P{SK~EVDQRC7w-eZ8PuOW$Iwk#A_5E@=Lyt{5kyX?73caKK zmqjoB_f*c{ZB`c^o3wwEsmA=MHMTudh_JhlE(56H-5&3mGL5d;t#q2avGOlqJsj#K z#f3GObNZdGrLw`Qr+OVG22s9#>$Tf|=LwaYJW^Xk^9%}5Jk4!~bpz9i@?@pP# zdiZCflB0t8X0QGdfA&|6D*;t1gixXp9CkB zhC2t%w4nOtFZh?oXVShP-Wh}0S)|ZQi1bSKiS826U%9P{5WMI3K&b(FISbbB+s}Sn zL~O$*m>)W4)(}6#t0tw#ZMtq`w7LGdALK%LK$dC-yjB*)kG5@vuA?nh9k%|l*9-cCnh()^kO}*)Mkz|ae9oi8vdhZ#GNL%TC>t1o z+c(bS+`U}HQPJaMG%LfNQao-D6I$<<;_Bm3j!0rwWu7n+;YNR~%%&@@@ocOB<7RmX zDtaH6tMIV#k3SahIic@9@h$E;FiT?y{yB?Z@XC(FbAv9* zAr}NXM^Z0*TfRPd=E1ik|7+pS!nV6(ZW7HqFR@?2yIQDit8K+Z2dez+BLgeJBh10# zSTwNRrKY~RekH+Usk^Pw#^j8N&~V6I!QrcrW*$rQC8&`T=)}9atBAlzja5mxD>`f8tvhUpdDcqh@a+TpXL^EQB8`nM z24x~l@MnH39+O#gShDihLG$DGGt9a-_sxI%%%o)Z4E=KWvaU^J0#ZCjpz*v<$t?#4 zWe^VS4lUb*{TD@>Mx4HXBt>M_)-xO^3`o}UuA+zchZzz2nfbTPQdE(kkaXFbEsEe;yw_~;S7pFj%db`gBM>=~ zX6-65B?%F~~@28mfel*+NvaDUdc&nf{xE7v@kS5xF#eIn5#DwEnn z;*G^#Y5CtK!-`5~S>irj>nopp3F3g2?c4PNpWu^s%~_I0t5zh!K43h-D}m<;y;)rnot2GNufVY|q`Mw~G0w5ZM;mG=fl6B(AlDxJR2$|LE_G2_ zomqVOw(gL@7A|Rki=~Iy8X!ekzi|odF!)532uY`8F>h%$l0xY7E2mK{QqCJ06{p7A z?p3Q+7{WnRE#l-nqLp1<&DZCyM2sVGRs$l105WDt_{E)Q$^sP&|-jl6*X_8xr zcy^=b9wfMO@`I#)zQx7}OM`fYHHcA#jt>3k#mF?;lg>pNTU~`LAg|_8@($6H=4gT>=gEdKZgE4-Z~kRVrPk*}4O#c$KMh zR2g-<=cs{(GuSrI)&cqa5iOiJ;#r22z`KxB_jI2I?Y+K{b@dPH8z^; z)@fu7f{P6$wV58h#yJu>o2R0d8!pvV|6817H#!36w~ z+I#g;KDbUe2xSs! zraRs59l+p+%H<*7Ir^5lOAa7f>Auf)xqaUWCX!x$wGV?v9ww@flBX>{?to_DkFI~s z6gqhodB6D>uq=hvBVe@Sxb-G4$ovXi=CJV4m%|`i@^;E=e0 z+ngUc`NO%LUDH$(TUrX5H`TpSzuKcz6ICK3y@dOCVVYGI?y`*?RCC$y2wZ*}IpYwg z(1K}R)N$>bGRc34c4szZjI~xp*zq90h*=4CmXEsj%5F&^79mHnZ4Iw4Shp@%A7#Nh zKc#7Xgw#5{UY8ixB)(ZuBm?*MHWpI2bX5DKD}a26PuPmgsq}`<5{}{q z*M}P_a1r87Z0qg_hLJ93T$9qWnP))n!!u)O!8PmR8Az_mbVQ=TJ_dv4os-WfIl9pKo zbc;nt4)gEYyq}VWHkznl&Rgk6T$|?!E{be`{ac$q0N&9?PcSXohkr(HzlqAA-5SDa zS0*0iGF@YY57ToB(l1%>`j@<|T`+EGX@YptvgEPl0+RbqDRw(@S&@sVtA58L(u+w- zsbe*HnGg~hMc0nqs`GnAtL%@33% zdwJ)y#8M4*iBth4UrL}VHMVoZNPU>wIB;i#Hxjp*)#=LRzD}%|28R9>=O#S*)*Y+2 zX^_sk6;eV24n>YwezI&~fdc4?xC-xr9SuL;+B`*IYG-`5MrZt3Au#|EN3j8UY^5~} zGZigbPva;Sr;KE$uD6Zq=x)Rp#}#yu$|jHp3*Ik#|>t!r}mg0OZMH_dio=BSE6hWHHfF(wDcz%f5eQ%K(5S? zv+BHOYs%3()^95bU?J{ushH^;a+PMkEiJELtmIJOV*d%h<6co(Kt(n{l84YuLB8_9 zl~4T%PNZ_5*|=6{{NMMJbx*r7&jP0hGua(p6wz(+6y8emJRf6!)^b4%6X{E!(&Fp1 zccz~qh}Lf-M~~^BlF#S+h}p&@QivFiJu#zM-1V^>^?L}>l?#OKL5~ixQz7N_C$*2j z{DAdv4*6<%Y}0{e2PzES*Ogsqysn;Zq!@a9#B?ELBynvg#jM#(_AJc4wl$VhCp0CN z${CI7Tq=bA{kV8D5s2p&0Pg^z#&}6{@K#3QT;4@)H-^mNc_0ah zi@XJW3c|?)&x_?+fd8vsB!p9Z$gV@thp`kL(F591HqY2$l<3Mg1TQNZ)~>-C*P6MC zs1LlyDf;hlCkDu-p&FZd&fyF_rCn4aA<2>LIx*A_Wb+Mm+7&kbx+u`b>tg1Ak;eZe z+Q8a@e85zf_RwI7T?Yp?L6WpKZz=)Ya9IX^M%9r|LGIL%3j%Dp*?PgI$P&+7?7=Ns zbmwFu>b2_l!RLh)V(e-nT4ySt$}erF@-*SJ1AHcP&K_E4&r=9R{6eOWWVqAO{V>pee{U za5QwsD&p{N5nHjRoU^2EZ(}_lV08fNINw|70n#CH_H6+RT7V*^V)T9;2Da1k1tM{j znMH1dJL_yufDveK#Z>18>lC(DD`ntJ9pDmM#KoEMAt><{U-8uC;#Qo5&H{4k z>hYcuOtjEmfs_bWl2b_V@QPCNl z@#&g+BOM&R;j+w&h)9@?Bs+!Iw-Fp}NrmPGM+*5TVU+PBAe0mLTa!E0p zmUTo9Ya7MpCSR>q`aN(Kpo=z&nkuE}=5rD_Md47Zg{_YrILZix2VN3*0UjVbZ?$?^ zoc;)Yg+fqfSEzD@4RiQRUKf@sFL&h52yV$*>rlVL%uYSvl$dZy9(mze{i}t$m>zP~ z7HZBmQ5IeZDTD0BjsTbJm)SX@K!?x9Am@%P@dMAHeu7=crOoIMNvCG17R5qkBV7L& zJpAuBxv$?5C>y38`&iN;6BC_iT~gQjWU+KG^-jX?L=2woTqZ#>h4n{QCRMSuEgd-0 zQzPizwUZ$ufn4*=WCL99{#;<|4BC3*hF4=;1!m0GkT}veUlYviGB}iz9JiAEtLSJR zGUrzFQ&t|6sU6SQ;KVG#pYF)m-lRG!7rV-IMc&3Q`L1@us=>ag+67=8PNs$B1*c3qX`p#+t5G7_7h@kBBs}syNBkl& zrwY^F4n1xPHD4$Bfh+s2eBKUQ%uqkwxH#XzE&T!7?IZW{Cw%-@57p2c^YDDiQ z-}K2qF<&DPJFy`YQi3p-I5%+bYQ_U$7lUtBbvnZtDcf8i=@1;a&s8UDaxK>Y-G{7& zalYv}qw#Tb)@1iUYa1=s-CYW#S5vADY>r2YKD9fnz!$Ei7g}SH#~Ji@(-_z@>PNDo%r-&#=~s4bK?xM=HLQ@jVxoL`&9lbFtjo<(33ZLDWIFr;v0vpHkUKKw{VXgP6B{7A@waewCqEp{>{O;o|w zA}|UE)omFFkCtf4MBwY#z3nspXYV|#<;y;t{l`}2U&Z4N2ErI!4z8^8YJ0Nj?dWzl zP>9Z@mcW_B>N64T%c;V$V*bBbO0uJKOFt%|l=VWXTeEAA>ucN2wV4;SoaEK}C~c6+ zuDW`A5U;Yo$Q6Q7R6eyt^kQprDeJ)4Wt}mQ0*tbMwk5?g+vcxAtqmA%JZ#s)eraaw zQ^RSPfE$eo6rzMLXb1E6sGe~z4OI3YP#~0dW4I;Xm(U_O^g<`B5#bVd@}Or{)Fhy5 z!)2qm-6P}&R7+=hYh-VB<)d&^n)}9r6t6OO-*rJ#FhgPZQ9Z}pJGy#YA?v{}j;*IN zQi`)(WEIS)=EQ(!P*9JxV%6os{7a9V4HUFNO+9O)43FWj@W?^hWo)tu+zfBN zZ%iDRI)m!e462o;LQX{XZjWU#{Y!WVo-@bnV?)kw3Tl@cR;Mhk@pXI^k{1 z`_Z!P%PYt_`Q-fE1Vi;;eEaG(#7NdjwPUUo1KI&B1HuYPq2=%1kqNIY8rbQCePQ9= z*MmR0H55-3&Cfaf_yID`-#vagX!n!(kI5aD#+u{G1;u~07R06P6yQr6n)+0Mu)cD0 zIE>*XIdfPn^8)qQ;!Shhb50tM{tDreR2oZSC$Ie7`^RqMS1#y#ci^V{X4{!@hvSgp zz$l=n0zNQKGAS_knVbLB4#mP(7IG!+a;n3lnz3s#H)8xAZU0x2`~&u`pvlPZX4UsM z+r!$ff|s!$%IY2|JWcdgdiCv=fhAYzV>e8L8s+($p4Z1yex5k!kY&()aR{{~IIv9z zI*FZ?MFQkS<*y0y+k9NNZ20M8|M=w8C&$MU#?An1BZV(v^Td0}^s+TnHJ;(cY+rK5 zQf3@aM%K#f{nOLRtC3w8k`I1zaDaX3EfzT-BPO)P%Gh!W>Z2=azo=B6owyJJaG zmXX#&ivNeGZ;xl{|KqPzDxVNaa+z`~iV$+!D!I!gA=f4M>)h|QA|WAE?pAUsa*5n8 z%RSfSemBhhHZ!x?^>@C%KYqV|JRWl%kL^6RbI$wydS1W2iNF8+II`L(ZRiVl==$Cz zu<*sVFE4N!y9X(>-99K4+xECE?jWu@L06lNKl&INQk7Bq%Mlg^6^=N}K8M-)zC?L# z@_mFysTO5GyWRS`{_@Ljw0N@r&b|p#9txfB`29wm3?6`PmwLJg9isL`9zLw7IHrFq zxaTW)t0naL6HZyGuNS`+!^M+!-ClPsQ|$B);1rI{G`R2M z%}R%(wm)>mD#Bg}Ha3~(mx!*Y}cxi1cmuS2d4P8`6Xeub%T+JgHzKRnU9ZMuji>X^Z{&)pwG2u9n=? zalhTN8QXYGlWZMTI!+E=er93m%9#=zbBbPiI@3GkTUF@iv6Js=6y49LJkNi0Y$*1) zN^<<4@m;&$h-GRowpZIS!jmC)RDD%O4}RDy&ro>eni)X0rf40Y$U?PgtFH&ZZMKd= zC*Jbds!gg`@^m?J;N_Z}=^e*{N2j+`Y0dXGyQaD0P&b6$LUisJ^N^l9Tv@L@51F+u zbWGp7(`TKm6l*2Sla}y0^`!{UJ^tdS-$nm>{^`Pllong4!w)T#<=>B&o`V0=KT&q@ z=bMX9gr&g0k;oTX!7o1T6JGQs{FkRoyKVR@M7;%9Q`-FWbV2L)iOS|2-&M;iK-m#=<>ZKMTb-P2)`IC0*&p~A7%^8G7;xz`*)1gL1zTD0s(Ee zQ75PYsupsq2RUe=j=Jd87g0qHS_yVy{%rXJ~5-L zyF@~Zz`H>+P3N4O-{$SRBq29h(o9438U;GfXip8-8-#9ksmvx*O7QV|$S2S}pps1% zHHGHAn8I;%@lK)((9McH7Q=A720KV_dJ-h2$st?lGo6^*IP>NxA~$Ka3?k$*AJ5pI zKi6HKm@rB)loqwCMc00NkLh`9-TgSbd$c7kq*yf%FBOkW*2edDZdV5z?RZl7?A72= z?X0mt!=B?iL~O-NRck>cZG56MzV7>a<{1rp0b6-Q`A$oY{6$*wy@e9o)f=6}t+7M1Fv7fX&j7do78+vwE`xYyOE zhGf^w&J@_~*0z9KnJb|r4<7I5Hw5i{u_4PC{H0ocLVOZw=oPCLK1cOPfXx3k39`8+ zjmIq)>p=Qg(T&dz6Ejm*U7z7L*)dhO7Vd0W#00yI1ea~c^P)U9cN4pkNU3Ofp<7Jw#TOb#Z)%kowiEf z^D5xQ@i#&LSYs;TW&*~|$bGz!c;E}dLo!q-?j3xa(ji`gJcu78e4t{WHBw_t8jp4<5G&ZS0n{18V9|rrL2NRjPtVzjN+s0kt<3^AD=e&aWz|d$=)I0 z0`wYpzi`x`XPQ|_hIhROaWw23>cnq_g z_9l59aXW~!rF^pDe(|HwVqo=T^&d&Vsl&*IzW%BQw2ca~TbRsJHQkJD>Gwfy6>CdE zAkV^%d}lv-p?piqN40dNnEk^wu7#Cq;9rx=+_}tI+XFj(G8m6rkK-j-{twZHepCV# zi(6ahHl_zoT{ICS^I~~Xw;3@31pnGB=ES-sB>bbU{ftsWg~`H!|6sQ77XiWz^0Cr6 z4BYV|W5H{+tS5`KcbE9)GIs>NQlV0BPQoV=u0->(2-(6#s4h9cf8vMT={+1SkwN9B z!^wlQ!qBUz8BonRNLjy#Vq~_)pUIV9r04A}2d-36R&!!-Rl@!-=pE_$J0M}}eYoQ9 z!2w=gcY*>%s9ig}>Q}9c)yMg@LCzPqwXZ>$@$Bf;SCJcSCzoEc-F2yi|9JkFEUjjJ&qi*(Ilz^(u zp=zp-A+)}mu60iN+If;q5hi#!r~oe8ifYEA9xq1Df?X*QZ0?ICjRC*{n7xRc~G94&IJx9y`@M* z-s0~1Et|w;-Y{d4lcTMY330-!Jxv+=aaH=cNMdkqh%D&iju?nfkF|{R?;lCCZ7`s+ z9EK7DUkk46X{8KBvO|~liT-YtX5&L!B`>eBN}7I*dP>M$e!^48rg&>AE1qMPw;Ccl zRLtDMaNS)y7oo5L`~c}$Ia0~jm062s_)zachS6U@3lHp`MqF3UY0lbDk-5sDNV+j( z|KOZ8?Bb8yq5rkpqY9}-RmCE)$q}RE2Q{tplYq+^4P<^)KQKy~p?#+hp9F3p9s2Nt zOT2J=J#bgEBw_N_up739vX!E2TQ+C;9eRmJ`Sqby6S&;Ej z`J1^-Jdn{&?r4N(cu7h0WKFguP2EzHl3O~lS4s{fj{?|J@4>adM)edu!YD+K>?)A~ z!4vGj4~|t6JkCMB&no96du=#g)>lEdr3P^69oT$k8=saP^Gd3*iO&4jNJ6}&50Kq^ z+0Vnb_D#H%ISywduBFZ;!9N2RPg=#_=(+1};!nrMV^}A-WwtF+A^4?%rRC#n$Mli( z37>A>&zy;;5ykg|mV-psf#OlW)nj(JQ4~kQa7b);F9BNs$O%qLe^-Q^LS|hpI?kQj z*bi)2?wVVy_fUVUbO|0KWVwm^kHOmB6Wkn$cd zn(ppM24;kpvUe2g5L@eqwt&AMWVC3}k8xwFEMZ%Z0rHXh?%*?aaNDuV#DKxB_k6XK zi#h0hJ@+c>L4^6c+cQoeZTUZqTeuA+fVQh(5pcoCR&DC;M>8|`Gi@@!ZL0SUV2@E$ z6eqZexn>P=w*u1={y4qkxn*}rt1{ki7$Fcm^y5_K7_Xd5~cwGAz^aD|a$wi4o$o#CEZ8;1zX$tL6Xyz@h zcOe6kOv2E+gd#f-2Ip!gcx1v%A)5?kgT8OAlPbZOWFC5#?pA_P-c{5p@whd4Be(^O zVKt@?Y=5(v9zI)Uy>99WnIOpOEU+B>_KX4n!a;u(HN#joyIlaqF0WqoHG?qj&0>DG zLR{$1SnH6;263`S^G);wKMOSad8|3bcDHfSn)!%KW`6}44pjOQ zdFnKZd}_(T<}>X}_N&*%ni1e*xh46Anr4{v#`byv{(F^i%Uv}tQq$~J@|EaBUxIIjDNdz*> zXzqp(Dc@r~QJP0Ti9XB!(}R8uiPYvp`Jd!Q1Oi*ROhD2(yTAGXKsX;-vCIGPtrdFM zoOXQ1ikZ^U6Bq!w*sKrkHj1mHEe)~Pa`zY=Ya&O2Bi^#}VsL)Uv&T9VrLjwEX>RaY zj3%YQ))b{`2f`Xt=p!$et{}BYZNgMq z8#>mp`Uk`@?vZaZZzYTDM;jVbuwD0OSM@0&7NKI@4@}sa!sb%m^63PuR{L z*X_@^{DJ_J4)ZNq*lrD!#f(pr^BIX^%7=(ARg=dTFUe$cxG}ltIEe_LRXk(^1JkL*9oygSw%g7UEq{5UkjYZec(+c5cCn*1{H6I*_ID;%Z->xYk>vp z!eUUjooc_?G_FIBXW4onzfhg>F!2*ssH6Vo`NiVnCQ*x*!Ef7?X~s=$;>}$gTrz}; z+Vsie9>1ISToj#4sZrA5h5PR0gfZaboaIE>gVjPIndlPai(OoS; zLBVF#!5XMTUNgCBL1jp?Sjwp)sNLBnKMtR+6ARL&0}0tnF8Z)-@KWc}I+3*K8}~Dm z{gjEIh{H>hLg!Mu_)q3eehH(E@8LQ0KJuuIYYsOy()Pz3poIRuwLRKwh=?KsrZ2UJ zjLo|*bniYc>sgds)@;#lFuz(-I?-J?-s-h|4sV`=(_qBVorZ?YAM6G$JZ|Q$jTJIA zWrMASWFkta{PrPZ?+^4T0Q3KOl7CIO#o}IYy%S}HZ4*rwkI*#vuUvdS1o3B8y2Y?C z`J0L63uD@?K~#m>tIHR0sT_!`fNFmr4umP-sH=05trz)Px_Gu+UA#4`IrH#?!4F*N zjb3eqk1xUMAHr<>lmK6*MW>}3G(5O>@tyCLtENltR;aeX0&c?G$8e0@bEtesKMsZjlU$+Vs`< z9YIs?i}=)2Q^Nu;HxrJ_0Fz<-ibG&`5-mXiElIMdpy!(H_5ADIj61%@E76D(pRZ^B ziAdj=!7B{cU@owbKJ#aBtFK%^p^k(`1{ZWIeT#6XZiXnvh$wITFGLCcAAMpAyaNC| zYq6A5&XZ3|$;F{UEif^`ewjk~EWN+e@U(sBI-o){mU@0OJQx_OVXHZ&plvUg}4a-!CP`^6XX6ET&aK|houP~WM+1>87$;|c}Ut-4(i3@6)) ztAAPvp2@Nq5N(GIt#V`?0;CkDDpl&# z(xPw_l%zO+{pp+>^Y&~aTz!@pQEAP-`h!Bc1P+JFSww999I;%52X6fIgNWwa%NB10 zwmt}>QzD;QyYKo>!*0|uW7B;eenWVX&YMBh zbt7>G%i4&c1M7AWkRL*<7W|ov*CU)CS5D{>F|xsg@oYP^-pG^B2#p+IRcw8Fyfk_R zpN`FQG}Q$fk&ZJb>MceG<@?M8HOw7Fk>Pwf84Einnar@HDLvM9y9n2^BVnOx}u}^T% z-5Ye%u5Gow!~L)(ql-on%2OK65lGW8{|zE%l6#&flM`p5!e0=`2Y!7h#~eL4UC+Dd zTi59i2%P%v?1tZ&=}eB;*?uQsay8I`pa%4^okrKuz z#(J0DDR|ERmy+kcrbz(MeEWP-o8%=0Q=XU$eXHO=&A&N9E?f?B&3!g^@r4rFt+g0^ zoxUR^$Np3i7h4$XlqSF<1I+%>^X&Zu)jFH>JHKS~Iz(+MLNi)EkS*ekl9^VOM0MnTn7U_A?0{s{TO)s(8o z8go4i4YVuty3YTO7{{5UAFfU4bfLSEmA*E?Q!wkJdkby)01}okn0BB`8DRe}uxeCB z#O?}gN_q1)hi962Q8gyRD@e5lVl9ztz+Oh*AL3s-o#Bo+@>_NayQxs!>Ws>!u0E*g z2WL&H#Nl|?(+mgX<-K*OrxlA12fKr&CnKz^FKq279qqIk*xMHIDo0WsYNS7d9wWY^ zze-g~FBcC>0Vk052vrz~JK{iIqjdtt)PKQ$cB+3L@hyxY8C?Y^Pwnf`llCBnu^3J> zdEb&@wBFcI0k^D4Rtz8y9uog<*QtRjUQ(@NkoRey*RkUTsMX{8H~ZFk9I==Ha5J#q z^1!+Lzr)%(h<*b8T01NhEC3%p-lvo$RYcyJ#Yh4l4ZIQS|2=H={|~PoeQmv2u>^j$ zu00J`1gYw^!n8tL@>3^&bigJU1KUb&NQ%@P1Z2~Fl)j#RG;?|6;bza#*(D<_C}VMK zi9ndB>S3!$_D$-^^{X8eZ{r*e{-AD7K0)}n5h(4#l!LUp{s9RmuJqUF3GrZ&m+)-@ zqKNMZz_5X61CM9roE8lG`sB5pA>-c)u@k)(93CoYsyvtcpkv=-MVp9u(5$I$csDzw zbQxqjRzl!<-GD`b+6$GGST*6)-i*1*3RU$-G)sC* zlx20!+`k-;x0Wk);TvrLIFlU@BHbCOA2qzxZ5s&wjS6AIV_Oe_5w@E{W*z3X@Gatc zLin(t=WM|dRbv^=f+)a)PPR8fVXxCfm^#}t^w8z?u*bxfv@w#=Q^-E$`{Zrrb%wOJ zBh{A?vhc%7T|M6h+XDsaAX%qr4@8cO%d(#%drNztgh8irv~orZs^1ZpykUjf6d-Hc zml@@&SzS17d!6KH;Un#$w%gyDd-xem0F?d>Rhw6qW>8fMVI66I+*;Pzc?rBlmSz|@ z>umqA=Jqt6a$CCcg=9-rK2A78)D2QhDlOojl+R5w@_UbE)wM5|ybkpXyo}+&1vJ`2 zP1ZMMQmv{4?+Tk^CUcq-LVOW0V14qM?Q2Sw9eGYlMHmgA-0 zb91?eqGNoTKP1CivGy@LdoKqlA;?z&bdN6Q)9o$&Z7?n{w$8k_<1A}edUe{LTYKBh z>4z`##&`SpvW9Y>x{j-k~(2;0ZMT-5M3-{SZ?XafH88dEZuJKxYVX(( z-#qjovfCEGQ<5UQ+3o*o4qfJLJ)0ygCpg!jd7-^aWN>m3Fhy@gq9==qfMW85ch36s zZm0E<#1=8myqCl8n$>dh1c+1SDBK|t$6Aoh67ocN-r~`Dt);CSkw`DR{YX5H=?ggd z)4kG94%SKDjiHr&)Hut$#hCL?o*s9>j5 zKE#SWk{t;yVKKWLVr(Jy{--ZPz1%1q`ceGdST7W=X%F3fT`hms{DFC)aLg+qNDT!Y8hGX$xEK%6xeqedy+HlL7eMB^p{Y z!M|ylJa1d;xIL|~R>wUFMwcM{304B)<&zIpyX`j6g}0J2BeTf(`ICF2ipe?^d8QLW zIS-)udjU}pT`GWEn*!r<=TC{vbv-YZgb_T67^z8hW_keN#hYRWz+o4L6B z_&Yxlr6)6U+I>kW=G>oR+9=EI(6M8Z6A1bdmeft69g4;T{zXp}(s) z3^VEI`5dzm)Ee@#(Y6A;?3Qirdb9kb^Tnh`2S~-9=k;HUS)(8y2e}-MYyT56LY~+iDm_n9(C}bWJxlQw z(L`@b;IgF5iaFK%R>0e>?+>4yO8)+(;Vf63BM>6|@RI?@1vl2*!fS0|=PA!ldqk?3 zAlm5d%L+oDO1>$&HfdEYlv$Y(g34t7ZCxBYtG7E^fAr9FOVi|#AZptMQGp@tJ19rMgik%v62TvDF&7WUg+7cNINkFmxtB>1msg*&$4fBzd5c1~ zY;~FkKpvl>dFY$BMbua~>C5?_LOye(HUkc`2dD3-=kALqrS~Qj<^+~M!YFI@S?*r6 z3k}q2Bp|jBzUOBlwzL$i3!?$g>iDTm2`zqSbOmeQdJS zN&QnN?mRC;P_^Tcy~!;J7dw87p5SAQ;Vo9YRxNAv>_Aq^%&xPu&^5?ZeBqk^vEY5i z2!+WzvjDLfDV5}-+cxn|Nc5Bqfsr9c8zTft45O zLBFk4Q=?9EJt=PbD13>NEs^uyi}NP=-JwI@E=8@haa}Sw^|4R*<-b$%*MuMa7;xzP zZ>V}9xP@wsDK8mQ9-O0K)iB`jvKmBL%fyUFMN2Ru9E%B)?Z^C}>23$~p!<{4bGzWm%|8PY~Ke0CFAcn`})yypK02#kyf! z4mCQ_TQC9-Y1c3RJ8gcI?>#>=8NF;_V?G@3ccW+(cL0?v)e%ZHsw7w2u>lXRT{a@m* z$OBxmpiBbv9ztKm#uCDNMe=9jjlHrx|+0CI|j9`tIionx(DwNxP&M{P+`|;r$SFs#bR;I zpY~@9p|l%I@(iWW)9UPN7fEXdg7Bt8ZIKb^m*B4(>sw6yHyCO9vez5vKgbj;YE*RU zA$M#tDXtzR-6d15V(JiONt*~|?N25QZ0_>9{O%K4`@Xc9`?gFqev+Lkk-g0to9Gwx z)~@@a(d&dO;`bl*l`O04Jf3~7^EKij_=0B6FS*a3kbV?IRi3P2T$p?rXrQN1;eBMs zqgR(M(+042MtTn`dRqd77H9*;=hjP(<&K{$JJDXy&d>TKYPmF$yvWvXkabemx7LRE ze(#CXngv+HtumMk2aQ{E!@yt}?gQ(k3WrKx&vDpbC~mqCr%e}5knEsC3WFp-V@>fe z3T4?soLLm12dBa(AmT<-ko0zX@T0?Q+>HoxP;T@VcjFd7c*tZxUwo^PS`^f$V^Fa` zFjC&oPT&bUq6=aE4b&JhbzaeBrKNv?w)iqcwaz z;&14#Y}@^CZ^17IZxN50J?9dQDurLF-F;W>^eoF_9wN376~cDHe>nB!qG<96e0vcM z1T5K(vSi$uP^7vE@q+xx6tzwkXyTzaTWUbq)D6*ry0)Hp=)m#zHPXu#hg_~D$mV_< zYT2tD{v9&;f&6km{Y|o8{JwWAxx0w$dU&hdg%mqQ@pwylwJnL*Tp4ADeLD1SHFp}8 z0(a2raf81I{^zGv_tH?;X(fTLDJc4hC2RYZ`Ud#^Ud1gOb-3H>U4$mrLI0LBWW0iE zxvU-fH8}3?Cjnoex4TZB%vc8q+jw8|m7Y22+&iy|H`@YJwq#Hbnb$|q6p!H0vq4>} zC8W-!HKOc#)DCQTm#)6)%s4mA+Oqo9P+v_-o5OcPo$F3#pu& zf*z{2Nq_A*Cx0xLd6WDjcOD_iIMr}c<6K+RQjQh@c@9ZLzN|{M*-*97nmjZfgzH2c zi;WB(e($~X6Sj$HJNQ7AF8|9zQVOKqw+n*nZ~tX5v4fyby)lDBv$8xRljMJm}Pl>>I+Gqv-!{xCurw&HIwwP~Ag}4@R>%*R zXs;!G)Lxw_8GX?o`YZiN`z*AFqDkMz$3o(Wm~%dVac;n8T4^wFWN_cp?mJ#3ntk|o^DnHE?}N2qy#sUBbWhAvf@7&; zD1ISzkQg1jBZ-4p+u!(M`TEUSw)5W8kIwu6S4({Gc4k^re{~oCdHbU6f=&olf=dQ# zk)YowMwb7_TO_h7adk(+P5?&f-CXyb6(9b`XYx!XX**3b_~+}-JlQshGrz2%c6fm5B*1n ziUYCe&L|5&hgZXi-a7*>z$@F>om6u$?xmIi0oasdTF#Y;=*b!3H{`cVqXOXH`lKoG z^Ksc8ExF!^l4izeJaSMZEsgWTwA+0K0CH8#! zZj&X`punCZ-QpFZVg0-<+UTv_)S*Gdy!xXsJQ&)C>*gD-ep1Tw!CjJ7>P6VfXb2=B z0GPDC$Fc}YJ5NaKQvXD=#Z6w^=7bjOa!g;N7+Uxdi@mRDKb&#T4^?USaQG6^WMdf2 zJ&wI7ERuK{?vCED2EL>_xbzhAzs6HqW^t-Hc8{C?l5*02``whd&-&B&(_}b1fahxw zKt%+tuou1Fr;Wq21(&=0zk?w-GxZwKgIxCB3qqC1J{%-D;3cYo&e3AjG#r`fkuKW0 z>&YpT$Ne?;>~uEG88BW9jnAqVM~Kaf`)44d^>viCm;rF^JnBWr;B?5jZ4IQ)5B2m* zD83;xDn*++-!_O}W3Jw(Ig2B@Kb{wTSw88+=WFXGHGe+elpBAKT+twC7H}V1kp>ly zI`SVUIN^>M9PK^FX$v^RRF3V?;MoUdVzi@#co4A4{d|(dyWbZ7qkYTMp*G3) zKoyfljCOJM%9%NU0lLOz(I*ob>_jV|z-q5z4pqsQa(T&x!8yct2bg2tgAAG!uVM8{ zLkD*}R7HK+g$4_H*w}7aH%=o@v3v zSjkZbRo#P}L+~8F1T&^q42?fJ)L~U=tM$!cRbY0a>W)TvuaRggkn}X+UG0Kru&>^6 zyaahzc#V{gK9Ex#*;iO1eeUexvO92E4#{NHg}uY@gq}iYlb!aTeokKkLY3xT{A3OP zMVD)rsm3(sfsT8J*!uRahQX66f;>av=E%Zb&iRF!pOaO7+Ih(IQux(yxcS})d;{8Z z@OhX&jBzULIY-754+!3CeHHMe2Tw=5ezkJpa_vdTreXr->ZB;n?qL|d**+C>>)_-; zHvbYhZRmO7yL?w5Qx>Z_F0DPX*DNS*oOl3p4HPPf!}#$yF22&b*vUF7e;cw5)KmVM zd0w&r%wK_&C!N`Esp$Ek$xzFt+dNl9yckDEHPZ*H>;EoH;Po$q(xYRKO6p=UB2*g zD>up>OW-Cp-XE-|(e%J&0^mhQ?$!xNSLl@0!hUCUU+m66y7F2W zE~gE9=vILC(D^M3_1YDZnQiRdY}C0y>v&!|>+jNtpp1VPD9s@vOW4R=mAYCeE6LaU zBFlevwa6LUIYZiQ@3w$zM^sm~;-FdPIIBwC1(TmaB+R2QK2Sz8S6-w;SXQKP^N@9X z`Z50itbP8T`RQ#-`OA(HJR-}36}aK*;#KITvQ1HsS8J)=QJI@FX7MiX<&l)KRJo;V z!-B1et1d?$x{8+A9Z9JC1)5e1Io@tj;?1U2AnGxW9f%9^uHKR0`}*8!=#Du(tSd7a zo=NVg#4SgB97^Wb@X&I?lvt`WCM`c0pikurvXxcOpoaXa;fk%pW}u^%5xXt=+0FOu zVd?M-h_1Qad{>#xm%>dq&PHEbGs9;aqBAh0Hdm2fg0%9TLLl&h`f-A3Wj2l-Y*U^r zhG)>O^jrHR`K)*ZWV244&VK;(ahEBDqn@akNK;fSPwC+1mJGh^yd6@FuX`bMqxHYO zTRSXD7)^2r_z#uJKDE9B(`md9%f^Y@3E;kMuU%c6na}NczpHFMU6D)TgtdiyvAuik z5xoeIWEl^k>w4i}o2Oi}S=#iwaGCzz35%kNuPPTO;88H9^pZT?JOA3jtEa8*J!4Dq zO*0!E5tCt${Ja&#>7GnaXsCkc#hE9IViN{Irc6>CFOm0HxRUeDIrOnxL=3b!nQYl` zke+&mrIi9DM^|v6$dR9siFL%%2p`tI(r2$(R?X~ACiefuuWuWv^61^%*Wae!mf`Oy zJZkKASq@=aySU>J#Iq*pep3~3$a-6*uIUrxh~~8YYMCVgU$}0S2dP-UZ#!nh5E<4hGF>t9g2x zhxG-lwG=6dB1W2NcbBkXee`;--mWw7Iet^BcxN)H=pX1=S@F+b|H>XB*_mVbM>z>t z{Ji|{l9`G-U@dYeM0QrQO`ESDHXsWh`PlyN`$BGV8jvC3B@iuK2v>Mav|yrv>pFK3 zDDNI!G+inK@&|4IPD{igKY7@`AuGUwuAwI+&Lex(qhmi#hLqSk00|6WkMteg`54%s z-jS9sq(6TBd=AG5oh9hsvbne;v9tSnrewH>}Y5$+XVCgOVl>w=# z|HhoZLcveS??D5}=tz%j1I?5PdJC{u6`iUcf{EZ8nm>UhzE+?K{I(9YW!;5z%g@MN z539)0&{Nx2Z_>9*eWxjE_+@pTTkNul|1Im7E($#5+%6tv_~6fsWRm==5XvUMYb$pP z>glJ~6*9NAqXms$?u|1}4&<+g)+7 zr`>Q;^oHTs?pJITYc4<+Q8VtGn)I!~BTQ>*{H^deb-kHp*5jhHHq2G6a`8(DjzT z>E3mS^FgW|FC1M;bLyOOC5EhJ5`m?rLB6 zHDgrv9>_Ca+uE0gx?{4k|A%*|UkU5vpKdr}V;XUF8h05S(Z9Hfk&C<)zrx@P5>+&l zI3$4xG$;@(b%cS(^8=BM?5SRtD(riKmjQ`bQBsf5OZEJg2Q66 zu1b>N9P{38m2drS!TE$}_>Pb;yto`U0IKEiR=4;*g6b#+KnCm+!#1I)Uz{t(>uzt7 zx83$)(}=fxC1aCcT96U!1>=_R&WTZ7nQH5Sp%k3K)99C)Ip282o1;=OEiNDIK6-$S zM&T_Bh6lIURIyebj{C`m9IClrSY4GqY(D41CkTvAiiS>8NkA6p^BxYJud=%DmFSI5 zpHIYjNQF+TKLReCkA!Ta`;AFy=mAF^+F9jxS2*7}Mq=JrW3_8_&U0orZ*bxL@#*Re zOgSNQT7iEma{c6q!#j}cAJ@SwCvuyd1>TY_STkc*1GWwZ`jn&UPd^2o0TrSa+?S^w zK0p6~g)#8;$?QE?INCXMI?Tt259g(6`aVIbxi@qfNN7PF)DCf15WV{QOkLcpFYkPy zIECbebomceG&9)ojuM7~GPCRQvpzSQzLqmP*{OZ5vM;rK@{L_HO+d?j+GvNS<;Hlz zKXl9`i-?Ag_kWWb`i8E-sZSeaAA5!ez<#cDFFqYfJgTNzkDn}hHOl`buzq()sHe*L!ziKG1E=!Rf8Qd$o+Y%JrT~qGy%tbzB zDgU*;!Z@KxSEe=iiTXswog#^(e@CA@?){D0P6M1a9|G*h%;8rFpmxJaaA*~`-DC0@ z`0Ln#`mb01aDHWs%vL*e^2d^;FcQsz)HmL^e#X3tA=vGF>;}N+gHZ|GtmWgwf89^shXr-Wg?m7=HcOV~YyCXTB&i%&_n+9VOtaZav9- zn5{t@_N~u>e8kwMRxq|h;a8B}N7e_?nT2ob-`A`cQ;0x~u@nAZ<18=_7+SOQVyAma zCkBpPJzOP?$6Iraz`lkFo5x&T`I)o43R(h`#^q{!D-DkNf!rKe2-D z=zp7f`Ptte)EG)G^x_959Ev3FU9@(RnKbUUh)W5&=j#qsa1G+#~63Gk7BBc=lA zC8{R$I@zM8?VK_^jp2zCWC!)|%k!tnn`c%o(ly%2klM&@ z+_V{d-xGUjTu^svhd(bEd_L}J$hstAdjn^QsYl!Z7q4(LPlek@$$$S`({1Lo&x=S| zHRymBcj0H#KJ+d>C}|^SBnFB%JT<)_H@{;!ILd&ecuSCZ6J;j)!eJ%eZ(qvYS;;V1 z70&#*Hhyb*=1u1jrolR~VV6OWsRmc~SW5Pe#|D3@%ki2MKmTV%lPg}wZ`c8MeD$`1 zi=LZNR%p$$aXshuSWgSrMG3kbC95K~B$=`IKOkvW5)xFaaw*Z#9rU#7?1xu>rhM+b zkKy`jshYdyf#B#**w(F93jP$FTlP~&KmI>1VF%~rRpF)g1(Bnh4!TMdzMXl3?P{(| zsj6PeM||{YJxl8J?{YVsSg!N^ir)aBIiYp@FH_QwrMI9n#oa*c$_E1zIfHnG5BXNu z#OJALxe}Fgw_7`QK<`V8(7%nwBK{|Fb6dBV3-~_wX2fBNKP(YhDZd7mP4N$%n??J{ z(Qiby1upe&Mb^i6xNJ-0Ld*6b2(C@%u)m%fe>Kkj`l4t2aD*w+)4!)wS%{C$(j*2K z{F}8QQsnKsRjShKT8FoPE#}3#EB%lYUV)CO-`l-{ciGM`--mdXgCYt{79xs#b#A|Q zlsl19@|ma;#j)EiBD^4dGTHGChg9%My~2|}eJ33Tscqq|ofc&3$iT4&pN69DHNn7K*S+RGK^vXbL|6+!iZOzcPZ}`$l_gvJg6y zi{alkh_c-9DmmOEqc|a>P2qZbylIGvpL-hawmM|pdB$pF%&k+gy*&53S0s3p{;b@x z5&VEIo~}z{L?9meSA!)=>KhgSUq@|Tl|rY;MDSb}cQQcvNvL$PK8GjZ{|{N`9nFUO zhW#$pR#Bxz6>aTRwQ5USRg}`&GtpAiioGTM*4{X;3oMRkv@`(Gn?(6zod^+)65#+*z8@&p@-5f06HvR)wSF~mv_b*F+vB*SJDf>?6 zUFdY$Vc#G1%Y54}``R+PDB<1cd-(qRr2$%U{U-R`1K#}G-*cm1v#Q$&Me@CPOL&~1 zB&ppYFZRUig)eP#GQ+LQ^xl*OzCFs><^i{$T?6Y5*sdM zj)qV|Lq;!LU0@pIKf9lNaXsG}pYNLO79B>LjO3Q%sWIwU!kcfI@A&(cE8f4)6o z7krc{{QWeotB)|V%nkV9#+s5k2ypx8_vQ1br3vs>vK4(wf(6AGuTeI{}+=cMpE~z_+?`kr;-YeQ?P5Fr)>#5&$#7nO( zA#QePK<#*XzRz9Zy#HKktgjb|oShlc0I#(Fea!>ELbs*1*O##V8S=+=H{fnLIT5l4 z9CAnEXGl_a;HM4!x1}V%%gt)Q>C1T_TZNMC9c&&i86HCJQu6W8Q+K4D+wcrU(;OvI zasR&{zR?{4K*Y>=+rFm2LedK5*w8d}$lWQq?V0G8d!?ydx&kT)Jk!MRo*gHXf|`>H zr8l|d;CT$A*weZOD9`TaZLQ%{fy>T_Ba|ZIn_Sk$e~73|4S|!%Y3qf_D>MJ}yng0% zS+PbTuG?vfkJ!Z*A{UcvAaM2_3nXklo=LoSNRL|}JmcO|nd%1K$(*rvL(E&GdO(@m zaQPI4l$j2jXY2*-QBD8I@3;Bu>!9l#q7}#d6;Ij%x!(m!V~Iv-eW#g!+j9$FZ$-(! zzhl@9)vv+Hdb1Q?V37i*F3-iw!0$D6?Eg7y+b?%DZc?|6?WnY$*Ph;8?sl=rCGdM|!v-?X9?APry4f%S-_afS{iioHmO?POX+ZtHVFSe0ci;Ke%19x3E5zNl;Y4+QohU#U*flWI$)%w*TSTud13+ER zaIGmZL3i#wafrwl=M=~1HO75$vTxnnCP{-BlQVbpCsUHMxK{@KZAjM6!7)jBuTgu% zhiiEc9f^nK`La>65`Y|meAuL{j6`%XWQST9F^^Mg@et*FVxMZM@ghmOSe+WE-z1d~ z*AfG9KQq9OA%T4nB+w+Q;rkc8MG20xo)^30R##GI+sB8D`Qcg^@0k+w^VxZNuGpXV zjWh~|+(knZWCW}}c#j#=wHN%keH~G6$J4$?elRqAz`B05tm44jh1x6aGJbX|Vm=Ud($IgHsSs!jO)Gz+JPZ6XRT5OS zMUCZeSeK(Gv+C;#h=ymekWHu33!h_is&DV=JnlCO%ofQRGV@di|M5r%`T3mb0qqyJ z{1*~~T^mQzcRbT6+^ez<4-3z~yJSIYLB(F7Uboo3KDX8@7K7R?@P=&Iol|>o252W` zS1ks0|9ozI$O5-q(rTUmhSkZtn)<@5o3}xHMp~0rs-UTdCFaQ@`&8p7`RnGtlbdFL zHs5?h2U*!2QFp%Xy|`<{Tx=m8P6*EjKy_-GKP|txqrAn#_WG)mWs=7pWtUsNZB+o- z_MtszGx=S2ODhsh4u2?qaUvMrbm$r+ZfT1rU*BQ$%`PSjtGrSu!5d>yih>^CXckeehn@`QY}Iev~qse zmiOvn#mOCT*UdIcF zP2IM`Ki*ZxMNZ|6*xf&mK5h{?4iGseXdTO}Tpb-fV8|55c5h{-S`S(K{zQ}D5ocdP2QqcWJX*Hxkl{SL2Jq;1$dmb!Ql+Bvh7N6>aU*kZ_Xv} zl%Vg_G)MY=6onR0Gh9_ce)=YCB7vs;v{#6&&9W<-FC-+cBv75cpK+^ zUrrKYzTI|UJ1NLG} zeZ1|neJe?0)BzxD^RxTV#vLbHyv=T4y>}GlI_Q7JTQ%gV!SVKj-~*0*quXzfXluE& z6xQ+n1BW4FSykYHn*r+I$RF%%V$r9V{F=8aH`-fLkJu3laveH9;6Rc=ASYE>UK)`v z@Ctis_KfC9>mi(HMN*OP+2yTpB$_{tN)?(!q~TBBH|-tjVC_ zW&h`E+xw}_6P|2`-!|TViYo{z#He-0yQ>^3dbP*mPWE$q2s#dwZp;(Q!%-lrGY~w` zGTNU|&C@da(QjS>WbHAm)_eOt>w)WhbN|Av_M6U|0@wBydWsP{O`-yapfSWQ`s@3% zUkUuzCN7lXec8E9o9^(aC=DF$0=213MhV}@UVPiq9mP^f&+L8KGDd%nd{%4;VLyrF z6n$58WeEh|;D~u~Cr$Jjryj*((7l?(RMOBd6U4~;OUpM2EeB{t;<|$|djx6?cNa2L z`I~T_QmncgJH}{0aiFNEOFT(mYLw!zId$4%QjtAabVwJGika1;9!~FZK49zlU*uWz z@ruXv&>f9m%wCN=n%8YG8(#;!hik?+bsEaiF^s?B-HQTuTfbj?V0ev_JR@cgrXEIX z$o5jk$Qp&BwPnr@au_WN8>X6EcQ&1#R5lIE{LoF)Qi z1Vq4ba9(&X18m@?dEYB$*Xa-0%=S?3-9E49h|u*XYMumYx1l}<;5EZ|M)}T}?)Sq$ zV52aZa>vkc~sX z-TyssUPE%D!2CyGFuEw0L#gy)y*7_KY2S(BbA9!`m?Y9Oxt`-14~`$?sDoFxh(D%6 z_?4>@FJ9dGx&XRIxmReaW823LvIY(IEwM!SE|hk}}P6p7aC9-;Wf$KGUS=ZJlYkdmi7K5pirN zv;1g2DN6?o<<`VIyDDS&JK$E^9MRSK%GYBIjO%Jxex&zFe)t)1W!j>Jce*4X?6ihl zjcv3SBtF=GIj>56F3xVx%2??atC`#lH0}x)S~~}SYTAwlco6-dT(}4i*bp{vpI@-Q z37Iq^p8}gV4;bY58y55s7`J`(Ap|J?SICEEwX0p*u2L1x0C&a%>i_pXs=c>?3f6$d zjvgN0UKncOkZx}ok~&y+2<(f4VOoYg%Cyc?Y`&MwpqQ5ospj9N%E0WtT!7vtS_sw8 z3@N5KNs#OJ5%Ba)J5n9$T&*w~b1UDhF!(L8JHHqdh2Kr6;Bo`9jzv_`H_YFsgt~J;)?g0x?%{pE$p;mky|w! zWXO$CKdH;7?z9zh;88hv>sHl^k$n?#{jMU>n~+D+U(Bl@&$(ZGCP`q4sF{XyJSw?@ zoaw?0JT5t(%!lm?ECV#L{J2?B%NE09kF5(+X!#FN5%KX5La2Q~(T!LNIRDX#J`vs- z$_2H6O-xqwaL<6Z7;8e0v9I|+ivU%;Y0%b2$sT<3$drN>s|Y6NDXEzD;5?~t6A$uT z3O044q%~n1+yJ5aBv*>@k>r#va4Lkb?fnLNaFk9NKUF4L8^8E3TLa79mT=1Y*IGRM zsE?Vr`Yr_FLW*9z0beGwD-iGE3tRC;D#Nl#WF@7G;G+(%hfl*m3fvh0UHm-JT!H(# z)*9<*czW09Ecej{`Wb)#|BH8O07Ng|gomgQ1pbY7=w0|j9aQdtACo?xl(q4KPJV9V zsJ!l3GBpv_Kq56%Llz6D#XzVDZsL<&@iYM4wmG<#1+o^nPrSR=j=dEZ5G)?2 zB5uAn0AGT8*!dQN2*#{PQ&J`D5B3k02m-?vc38Lnp-cnTK|Ca*2uBB{&tl5~(rwYW zt|Of~bmke0j~;v*=Dh7K9C(}9+S%}**e6>lE6KTyd}J>kK78zt4PsCTC{-YU z*^j#)3gP;{%AZZ6lL9ov&0(2r2QxijVCQWC|WShcdfF}3>@)H{@G0xT0e9bivEjB&>={8$l zZ_M7>!>+HXsqDqWWJwjJ`%8Y8FWC!}G!|e8tq?4x_Wfhj374?!e?5ZYK0cBiBuxly z+Q!?`yl*{$uXY}8?_3oH;O{9jP#I>jHF7M@_Vq*Mb48ZkGObwMx7D#;1liIXkn5|S z4;M6ob1?a&+QS`a1lDX7YGiooBE^1ls>EX@m=Hl-bCk<{x_EX*hjBWjvL;UFuYWnDp>qf9K2k$HZPipCmx#EiZlK+U$j_ws6REzBLMn zg7ai+Q>I8>R)&j)6iwHSk(R)Hk5ct|gk7$$ z&zNUC$pBq8%B<20ts%Zq0DkEg+#Ym*RpMy+q&}^HPoX^J0?Ya{K8-Zq8TE(VVEgIF%9>%fPHjl%=cYHP zE;D@-o7$$VkV3MEc?sOg*Ff6qfxhlG+@{KLmK@_7&%!-R7dg zsDU+pH-vOIvqq5A>K{kx_!jDNuCDLsWBK15(N~HORk(Fy&pkEU7v%LJL9N3xT5c2m zfT5Wy#da7#WscRWZJq_q+R4A3FIQOJb*dA()~j-%_B9=M$_GMc6fg<_u@A7W>rDar z%KnIc;y+-6+?C1T=0B_`m6ol3CJ8@PF?#qaypz}T9Uykj7vFD7|!8qYrwk54c&3_1t&+e@WhOubXEdA(ZX%i79oUh$P#f|OPeua6lYg0q7XS~! zZ-B|ZbL(5$wI^~BzuK#jg}+o=4}u=7tIx%5s6+P^TEoW|9<5wa34`bXG~#q0s^q?KV4tL_pvqn!l})fsb1nzlWw99I|3GF`QyFSC33RDZBVgHO%JJdI^u%xnKT@ zqbG?qN%}wD?ZxI+B&{msiO8uL4Y_3ps%tqkqmw2 zb$RCc_r%r|ue^%_isTHOW2@%SQX9K1785SbrBN)bA$|B~vmJYq>&}T)1~v;tc%9j~qZ&T=w~cj|4ZdHiie4|#Bfv;EJ~ zpDXvD3d>GxwR}EEHmSdD`P~I!Qg!^v$P7cgZihT||Dd$7F5Q^AS-oWiJyr|#WEoT`y_(f5AX-Pa63VvBE@Y|J6iwWG!Na0+ZulR3*T5)KK{2Sh?1M`^A z2n?n6=&`J{3rc9W4;x=$=O->R+&fjV{=Zsk^mAxIxEHs`puE+qUxHC9b`jfqH}L)Z zjYuX!4s3byGX@fsSy|FCk<^v)SUKp=(luIe>7syz0#O?e=RzCj{W7(ny{s(N@mai{ z+lHiga`q|8v9c%a&56bUVea`1O3Dzc;m2}EU}JU%n&Vk>ab=NpW&+VZ-9R;HzF(6M z2spd5>$y+Xi`Lh0bMaTUw_aB1a3XFf+~+CtKxB)LO!9#8ea$$oOp6fiI}>CVhA@BO z0|;3TeHJwdB5g`i8;=(FC8#rPfFH~%Ja=O^_&y$ATIhsK2C}mv zF3p;r={z4@*bnphZ6hNYYR>4%04U2&6KLv~qFrXNdAqOW(o0v=gJr48 zx>S_VH%c^NC|RKM>9i)@LKh$S`bM^bLo({6>SU%<6>Q_kXmU3kw~v-MB2m@uXMenU zW~j=Y8rfZJ^&PP=l=p68tVTcMe<9z*#Kzga3|I@D4&EFHU4x`#_RwX-BvMlx-M2M$ zZ72Ra7RJ81PQMk-TIO%}iAXPMXM$!lAJWR5^QYX`rY}DiiV5V*ciV^X>^eR{n8zrD7}s8;95r^4<0xd{ssHue?x4qbWj#2TP`nY( zlh*Y0VYFlM3RQj(u4Pe2Q20}3Z(-&RV7wnnbtkyC)i%TMOGd4UHfLfVGlFPc4f%pE z%~rT=*u>wC_icLxeV$V-vP&qllLSNu)O!tQ&zvvyz0;B=z;D0y3JL!1EIpd(JTW}; zWy>m9mo*q@K=?zg7eevldVc3HeqMERYma}@1cxgX2-U!qLifgE!*5L<=(^kEMQ5lt zJyX&naoz^5%1&yW4Jx%p_!x*Cu}i#Rz+7#YHnFd(T~w}Nxy>09@`f{Vs!wBjm3J;d zU)3zZ-EA9p1<~x-C2%Gk~B!Az@(LWWhz^b$1U8kPqEn zOr%eNVc^61&y-2JdvL1TdH$lVJ^Nx_g?M-zGiT?+jwTHZaX;;y4cMJx7l!BxooP`Q zUK~TCc<|!++3gW#5Wb`_WYcaRnmKCq2ot?A+`pD*Idn0;ch6pmR>H5-FJ|1QGOHA} z32*vo0*#IHlTup0)@SWfX?zbbg; zYLkU)4>d8C4wuOe*F2eD`}FX~kYhjS7R#UuL{&Nay*z1%HKnFS>1tgfk(arHCLk88 zv(-G{n%vZf{(v#mmw_&b+xVpuYavO*(jjr!-(vI}Mhc>pYB2QFu{~n*HT3qqEA+4S zA_x6k^EWsrh=AVCINAa)ZS<{)p^?0dA@Rv&6V)Y%A7KC@{C$=bKu9UJYjxT*2YlwQP;7Z9`uUn)XvVp zmK<$q5_{*r&V38jR-8M>`A?BZjb3HTc~V~LUGYJHd1r*n$>)EdF-9w{GgZkxJ<7;K z@doQ@N>5#W=60dGGGGDlBi<}RInVMxh_va0{NJyQf4S*w)`N72rESl_ z1r6BV%*cV__KHWq-{Sk*>y@uoUmFL7$H~(9TX~RHIa>w^qtCgRXn6)0>^vs@gO2h< zWph9X3xM2$q0Z>@-cI$q!L@6CKVCV?Iw<+uX$o*N?yk3Q+o7aPZ}?qb<389=@3ISY zI-yPC59iK4Io@}lJ>cu-l>fC^IQQU=3VuLHU5_mzGC_Io@@{?W(!7YKz&fi4>zY#{ zXfJ%Gc3Wf4QzS&1f~_ueFE zy`7;VepS7JN|Z!2XG@1=RSt0P0fOkM1mQcRlUS|khVZvQXRFd8cYc6ctCNy5*VBii zyZbk}5!MS9h49)Be|Q^pWf)7kcO9rb7S|_)I)ff%a^WZJg=ezU6CH zq<*>d^G6%)feQabb93=B<5FM!nQl9h46lQgXU_8V!5JWUoirN~$09{moA; z#A_pT@R((h1)A8!zcMpLYe39+#T^w#G*uc7qcU%Rt--Ez3pX3Yet_Aw#lquHy7%2K z?{%c>Y@0WZI-WX;3haT93A9CG!f?}*S)c2GxyMAO@T0aC6Ihj37=h;AwzfDuYdsP+ z1ExM5Kkq8ZW!^RO|FU(2^RvLbBluALF~Ewm(b^puqdUuM0THSDiktB%8_^XFO3FukPKaPIKv66a6SZ~cwW~<1wBHKj87yfI9zD~*@y!KoL zF`rJQ`OdeD^A}FwPN>({{uRfo2w8-B>mTUzSw;q*+u`v%wJQ6uy8AN*F8iaAteXq}i_;yt zyZ%=xyV|9N!AChsq-Gx}!1#f+tfFzQ#<-(-Vb?J|69RABIO#S9Nx1An3A|a{ zDO)twB~VOGS9;ONt{~6QX7n4_a#4?wy+l|M$jDzBg@LXd2(pwm&e3>_=WZXf@J3 zLF{zvbT@c?YNQm;KFaJ@u0b;#TG zDFR`21Mw`MhGJ#XQz`vuW-{yyxQhy3LqFF-aLW41sP z1i8|c>+v0MoQYp8dw@3PHV;$Sd*4(Pr(n=4*K`EB6)7^=>hRn&{@;p3`Xx~ME*mwD zUO#&iwrjWu=q1+=Tpt0(T;EfHXl5rgDdpxqgxqFc$jr~N0(HKy_VAOmysUAGi!w1S4ucaGe~axbn?Ol zh!O`Nj`_kBK)S6suiGJgb=4)*AYVg^njKZ-Seo%aZQPMQuW8MSTUb6OIsxa{R7cC& zCLBOzC3xK@ZH7m}P$GTE=#a4@e-6BuvrItzcb*$K0P-7L%DDz%PW$S}&k-uC@S<6< z$B~N{IK%jy+>9{TyZ;yy^MR}B(vRK8^$0HZ+nDEVs`;7JEXo;!62Kg^XBO&-7@Jz# z+NZ_9M%cY!Bge@0xcJdWb06-mWvmbAIFJI6c7!Tha3}2>j$>bBIJIa2TY(0tmcPn2 zJUIhYX4%h9=wR#(jjthFkm*#i_!*fJ-*O(Ga#}nXLcI-rqNAU}I#-7N)_?IgaN*Nq zrU$p*LzZZm@RQ$8)%z;Un{%neWbb$(l>L)06MVl`6aTxC&h%e_Ow?U9i-I?oBUK+i z68tC_e(5ETzQfg_4^%{UeLl$#`ubfKC6KscW7egLhS~n5R3`r0+JY(MZzZ(Oe~H;M%vj%BjnSt&U6OW1LZ^ zOjiupFZ<43c;U=N+w*}jJNzqAq3{2+a@=2Em5g%hTmzMEsOROGd(84`2|HN`^L3<} zn7?YBO=UO4JE`o{9}5isQ(B}2z)EeD)6V65(84=sgJuvgu zEj3iCDfnZ{-%JuOMzYP??dEiAi_H=Te-zT7D^&qV z3S`N*Tvw>#iE?;;>v7wMVm4ManuPa9Pv7diQY!k(DR`16&6sUmtsG5OMky7evQHXrz%?!F?HB35H@}LiP^^3NT zEVU~kivWB#j&}V&=ICUoN0_zxxOU+zIQ5o|bAEiSk@&Xv==SfLX4(eQ%=Cglkh0f> zXuFNS^P%nFO5!rNt%LHq71J$&Jwu5q1j4S1T(>|`oG9aC0%8$VU46u2$ zIO4;Km+E?N#o6z?8~0aC>@If|d<*kuDh<;YnYxktKB+JGymTgh=RBZ+7T|U<#dy+4 zgYh$Z&Xy~l13gEb1)V{($j5_2lrb?;weYk?PZkR+tz8`fpQ~kdf zmMS$(N7m5Z9>p@XX6A~sLy@NumZtg-|9g#jMXkDVPt#6~UlZ@9RPCD>3p1!z z?y{DGz933dzL8h#)09Bp6Sl+b4bV4CEOFXPEfVm|`RO*qI$BZIk3& zI)guVQ(oqu#XS$DTQQe+Qagdk#|1aHc5r@*{V8h1!c#e%xFq{yYM|#;B@HliGhA=; zRvz0=mlV&Wt6F^_N%#=a=XTA8%Tl-xhzdv=J>r8waQ4Y7cmPzis zC9@Qe1^CVQ1egKva0;SmdH*t9KZ!%pXos(llP{^ST}gO^8;9QRlOKY07X^*}z1(`( z^aU`p#vJgUf0{R!7x?1&efbw$m!aVrNbB{i`^g;TyO_Vxs^1E2~&0-G|DNSg|VRM1rwixZZSz&`+0rLause zxg+RJbGoH5DoKh#=+lC$;;+`KwWLTGKC&6s!JY$&~<#eJ4@ zFwr43zoPSm(m!h$GxhCr>*^Lht#H7HNjzJiSFJV$74=YF8bt+5ZhNjaG*k8WKFdun z$?d(_R1LQQGPfc*@#0&jsKSIAm3lH!^nI3Ud?o)wKU1-e$PE15W7o`_8Rf_|w!D4Z z9F*p%ph<3^x*5MS{)q9GIJ$RW?I!8mWfCFwjxc#`R`;I~`$SkY)8pYmgRwRjVvsz0 zrNZUpv`i3gt{tfN`Vw?Yf^F)lY0jk4ltT-9836V-$~Zhn&~zvzn~eK(#P5 z%AK}5f42%ts`>Sh&;47E+@<(my|+ULJJlo(jDGBWS8}Oc^RklvbEo;9{ezvJR~naF zi<9D|Tzy&1-2|r}2ag{;4GY-6tL+){3E4+_)F&~jpSa>gzJn;l<+-Hpwbqo04L5Db z8-wsC-msE)`!Omho*M$%kPy^jDM<>70M)~wv2L+vVGpiq`Z)3Y%<%Oos?Wpd_y<0RESAp6+qks>4!HF9X4-}L*P?GI@a!EU}FK||xTpMQS;e3Ya9_RyQ2XCWo*LTX;4HrvysfHu1h(%N@k+M%Io@e}>^sez7TcSAE;&7VBNpmPW2 zJ7HIL?-T;8|Ft#!tXKr}oz*idBMB}Q2?d^C{4_kB7U@SH_@I2SD=!q(l+i1%f*x1} zgw<7!Nx$Ey_~gxzHpUnxGFN$Ccn zmqUgdRUPeC+w`Mto|L=ZM6$Ve+-i*qF#9W||H?O;(UJ{F(2Yplvi1+Eswnj>i%s6! zhot%LspDy;c z3lA`*&whmMLZzaR;6{y5r3F1iDC;{!3@7dF64^1;#MhG@cQjp8UPGx)%K5eXy7AN+ zpqgoBoDFouznZw#$f~hEjlZ^54Phhg^hx4}*)jFM5I&8kz$=HDGR&%}JeD<0O_p~60w(DqLIpXsZO{u~GYXB2P_ zE_?6%Li9`~i@tU6dQJ5j=DxeJDu*m1(lGgb_otAt_b~^KKZF5SdWO|AIc)6)7zgb* zltTg&3HuWq%uqx*8rDc30FaV(HVsi$f=cgV&#JkuR(k9!4tC&rS38WBOc3$E4nDFF z>#=(|{}m?ccmrZBd)m$0;n`;|b9r#(d!E6*S~DklgxAMj$ES2%9H4e(r2+W3RAfUk zQ6xvL5p`wf;fr&02VPE%$g%UiV_D^0!Q< z%Y_&xPM$W@Ufaw`M1!f~Yv`(L$(cT)>)>1i{%5FY5u^N1E4l02FW!w$KQWNVlpdc< zYZ8q3TaRIfDadw+kVr(s)*3gDK_?QY=er845N)1z32Re)LT zx>(Uowy-hX>Hn$%#5I*UYC=D&{8SLm05mI88$=s(d^54?u5DtgQ*n`9WzBIc;LO$t zRNB*C`D9wB%frgZ3-9@RP~ZiRvu5UYw= z&#w`N-UJ`tMHXPRX2A_YI=R7;TQK2xf3hA@Y$aNs-4S~d^FW0Ygp~WkNrF)ohedH5 zg8T*f`K+xc-c=y1=Q&CjD-F`Q;3M9J=_Zo3uzn}^y7wPQoL}7mTj}XWKt|W!dvyk zII_VsKw2j#++2p??(zC?$v8}r`d3yLKTD*aKnvj1`ex4Ia6yE3y8>7DC#U_Nf-?&` zkK57k2$ncbpy9xj;6#u_*t9Q$%cqv`!@kU}4Yy1I2jT+Nr;=i=O! zc^-BFIXp^NTCoHMlSbEnE#j_Dyqj%mxN82{L4C~YVtD6P`H9QK@@<_HCC|nKi>jWM zbzn!^rlH2Qs*}uW-32rs3An7%RJ!!n^(e{YWAPOHmf>gOER=P2VH@VsK=V>~V088K zhLk#i5`ogd^`POC=V)GYi65PrLnKs{OVg$PB%SEeuPQs z-wwb!nsYC(fQu|<6`ZKytZF6IMS)00t{KLs)=6iz7j~zJ2-t2YO-lupB>B9CkU@q_ z<311+8xmb?Wl^G|ZkA3~<=qc5iuVHg%i`}qTTY}2f|!(llM|R3g`5>?kt1yEtl;C7 zuaSu!etGfCDs8Zp^Mdu@OGIM4SUoiN#kJ5sr~f+P?{$e?exh8e^Q)Ou zNLFm@t61%zX_bZrQv#p^-e4aUGoL%)Y228S*a9<7Cjxa-b>%X9IE;#wfpykr?9|Ym-2;kLCgk;OXdo=hx_x|IlC%dGg+=*AT6-eSkaB3O-(^JKvk&pV-*3wI_7nM`24nJ zYNqCCyyYklVtM}eH_DZ@!br}v>g*9$ zTt43uoM3mRHsLsXa(*a0m>x&olF~Uv@A6D4z2H9RLLURgil?l9&wn&dMsJd+4;e4A3nT5w7-i3cMCe zH3d&5h0$6ef2zckobIoZsYAf!b#SRq4OI`qo^rjo(2mP|E0(J_fcw+mj?4A8W#k2V zbNxfg^q*HLK044nn&e-bJeiNV`pj?Aei61XjW_EogwSIDyOvol^f}&Ow42IcpYW5) zx%fO}pj#D=wejzB?{!0qu(SQT{7cWbjh3lG|6J=lS)hdElC5H@b+-Nnq~0pZJO;_ag#hq*#f;8Bk;TY+&Q2V@UEz1c_hFde!^e96`itpWajN3CD+te! z1%4qv5Xr!?GET?BVmv=96SxKCq)hBQx*q=sSPWCH5PqJTZ2Xzd>}0Gp&3Dn+j`D#u zOg2aw^Bdc05d9^b(8azI;GaEf^PoYxly<&A+fjz>>;`Xy9;g$CY^{be3M1AKsD z1HNfCZ7;x9rGSgPS#$*7J@CPP=?*#eonSu z{y?pvuz8RAxLkPbIrZ2Di)e^Xo^L%wDm{Q7>jiP=`OK{`10v4uq9lAI3r;pe`$*&2 z+jjvq4W&Vqq~o*U`kMf`>IXqM+A@)w)Ih=w_yvUu!4(UNibP(ydvMfD5+>h;P(-?6 zL7^ARc_34Q0O&D1R!H5P&Yd*1ZpD}E7GMLvKe zF7v@}E;7fh7BawM3kR6HKbM;~Y%@O|y*?^Hs=vvX+IL?gfyhZ9lv19r?8vaPh24Fi zzF&NDWx{d-e92=q_XaKkG-`C#a}!wz0i3rLXEzTG(hIV=M1Q@+odjZf+@6E){-LB{s{8S zgXic?A*c8}=vTW#me%Q`X2;*9dmLwxsk^ND>0{!xBhS2_qyC$vyhMJ+7@&Dgi1>&F#!n$K}w`H1EfnjCoLiS0n!T<1qA^qX@-J=bV$eO zknU!5!!~kbY9j^lauW^eXl$G*Go`?~JyJkO8Wj7fjmq?Q`rAaOF`pYyoaKjBGs+#!Gx)#q#QqUEpWCj?=D^q zRw61F=al2;h4?MDi`H(jPhE9(Tn9iim?l)-xuK%=d)4*L#U4f;cHlpuU<+SDes(@p@2rBFTvlvd4R{b`tIwN++@j1u4Efu%i zk(VrcSL@9e(iccBqq;}Uwm|vcqaxL9?c2{BT=xz<`4TX5pUb79z*Id$vkd7*YD*oE z?c$*?-U}_(tXufl5a&9ozZ&{aIFeWUTb~L4^`~ryHM5)BsLpLX)Cw^z`ZgjqufXr7 zT!bNJ?$~FgtDNOUkUNj4ZQSK%hF}HzV*Bmh_wCZ<7I*#he?OT_ou4}2xe(p8;xp_y z!k19RF|rN)r&NTAxBR~0+Xj?I{S0}nB!2faajz(@-DP+3vC*zxH-FEs3#CIls>vm! z0-oOVneiux5yvn&hOdSQ+21eai8z#$!+cw!CVb=f;32#w^8hJW4uOY6O>9IBA1bWu zGEHjpRDH1rG*ayb5A85~;qEVDdUv;*GO<>CEi2-2(p3$)pXKP#JBtxq_&TI7K=Zhj7S!9Erewb;5Pdd} zc7vvkB*`zDRk*w5*sY#8=~?&PXV)q3?&eNMyPe}%-K z;%$Q?uK6AXj`wI@yA15(d6-UHkqOG{V5`=J|NfB^I(rUqXvq5l+wy&Qn*6m^>Yl8w zTOL=?WW(QR?S~-I;s4sZ$}S5xdD-MDSiY&f--U?ys47L?g?`xQzZ;N>f$Z{q>Wx2U zb{v>nJiPzwZ;EN_u$>^$=a4%$9ey^kx)t}&+aO!Yv&zlv_0UqOutujQ@@|u)fmew3 zbEhFotmxQ83_0)fR54~M&9+IW-8e#ZRZxMX3FcfbfO3mY*8Zr2m!K-2`k%}ywJVQ4ZLRYJz6}7HpGX(UWjpZSo)QUUF z=&WPFs%uI0@zDk5)G2?As_k;$lZ4b~or%wgTi{M+HW`uK(c`fj+<0ImG2r`Vi(>ys zTVxKZ*UYSQfi4C)qacbIf&7yjR{p{&=vv%{S?oMFQE6EhT$#U)f5gKGCt|)Je(VF3 z;L2I{pqp*aNLIuMO3S(QUlK{09hA^s?Rtmx^_2VNG#BI;?~%oP(=8z0pJpK)up79A z=ioK1Y3He7YiD2_{cQ z7D-njdAau)R9wzonuqXJR-@!v>$l1jy;@UdtA9z0u@^A!h(9FixlA|_rCmKQ`1U@p z5Z?ve7wSpwH>`oCWX9{(dq)1zvYdn7K-R7+jsx!HA~sdD1JmZygS{3oXeL9NDIDEu zj_EJ8A3PpeA}~FUUtL-c)XSNz?dbZF>!agE@R&!jn_q*eaapSG_6Lr0@0sGFx6#KX zxAHmCPI`q+IT89Y&sA<~yi4$N$vZ`TKaoC9;=w$QknySY;=he_d{9?=qnRf$H+|p4 zd6!0)J^sbnKW?f5Cpw}y@Rq&o#y!~)G2rOyCBMU?!)^vTh>o|?;vrb!e9~a+WsENy z={AC&n6wVKeSs=~J5;bYO~b^ANnHbWi$d<-53fIuS2agBZS4z_*_8@|!sg%P-`xJR z(M}B)P~r>c7Hv6+2)d97?@k| zAf&5*Z3Y}rD)ri~+Q z-9u!8{}%mrhJhlFM@f&*nX*7hj`rZ~O|s%+dq-;0zxQ5}La27C04m!qI{Vp61cp~z zta6tMQ99F3`|pF_x3%tQ7FOx88qwtNMTd2*X>4aL3o6(|Rmq}_7U|J^9xJ%WUU#p- z#BK4NSZE#TK}X+4ozb-VnIE!*67+b?o$ za`#UY-Ac-3WM`x<@D=ipp+m|`H#~%%cv9W489iP4!8@lnY|@#u(nJn)Pfgav zVzb%;eyy(WUgY9?_n!5R>Hd=_7sv%WBVtqxx>F_yt?K!cy&dR0ACn;!W>3tMU>;~+ zL+{Ae^b`*nor~OY*(g>RO?NG<90}l-P!+9QiCl5@OoSl`+qq+i?4+KW?#bXeEvFKG zJ$v9!HfKinBVOkFzuGN^ABLFqnl%7z6OZ2gx6xhCOG0@eT)S<6 zOX|KeQYa1D()`JVJsqkF&J_~OnR8{deWj-Nj1_JQijAQKL_hQKO&ym{}R;)TT2 zUuJfdpPKdCY3S^yzt8Vs?P1V&3ML__HUsG72v8C(q&x?jw?Fj5LB$y0V|p>R8h8W-Y(F ztQXXKvh6;qPtu*$O`J)*+OjIF6gFhp>s-cDBbS@J!;J)NJeBJ-tr-7(?NIaZiEjf; zVcXf~MuPD}%WU;1p=PkG4<61AIwC|HdfMCmUVhanFD(_@8tFKFvxAUNC~LXDF+CXE zgY-nY5JzEmFm8L@mYeG5qle$Lz}5nzL9-RtW!#q@I&p&*&AB97)7qW+n`e6_ox6+R?vjzAf_imta+`YBwjdFdCiF;Mhybx{cdsTdApYw z@p|z2%Qm;ew}#td@~lOY_hHnzaX(lg$WkMIs$z3aTA1Vx1MOA#=9~w7;$(h$0_I&c zYW#;Q@~z2(@@3m`bM#AZZw!~2L7=*B zBxk^`i!Zgp_U6PHHiNxCzx|`3sI0X6bo64IlmES@kyLxioMuh!!_HkCUfM zegZ+?1EbdXe-DMHnVw1_N`POyHK z4WoE1(Q2HpT=2-&Pyaj-OgKs-gfhEcv$n7mIE}nc4c$YFkOS5;_x`ood_gW~z9MVh zo_aQ)DR|GIE)RV(J}eV0oLK~@o^EaC$|(b#gl=H1=0~|AJ`%GIZ*Z+H`-gUTl&nYa{Hb9Zrs~-m{d6z*;bW2dy^9$YZLui1TOQNvvo6aU zQCyAi^_~uRTZ5K#pY(#&p+uFY!lF||-|4nqB=-5pTfhhJ-G%?1BYRS%t&y>f_08=r zhf^DDsx6b|aLkfgu1aAkM6N}}NL#+XsqFN|00KkX(YI4hNN6m%{xpBD%Iyz@WJA4; zGp=m&^Z~b00atUNM?lRBp8=BTHWlxv4LYgeNiNKE$}Q7xe!h^wE)h9{z6j=lnz*> z?O>Y_!0ltHCZXBNN>zHhKR~ZosAOn2AKqXxC8mH$qIi91< zpj47mz{mKkzsMWWA}Wr@c}3C#$^Nj|MxB401}eyMIR3LMmTfPj`YOI*C}fMHqlTf{VJ3qQN@YhCb`_=%zXd%p0>pgR2*#y1CU zZU5RF7ao$@H@laA<&8z}14nYke$vqlBF>v8zy~&&g9INrBDz%933|9P3-< z?tBdh%?o)JG=?L4(OGPw%Y0@mC5`6W#f|dnQ~BQXUT!T#e)<|el`l$t?PA>$Dz2w2 zF&3Me36P4_-4ND{U|?5HZ8|>HRlU3qTI;YLIH%};kw zIt#LnfpCf>T|+>Elcb#o+%|Q&%~8j_{Wv+&tB`BwtqZ^30Z3z`=Qjj)wN9gttv3bQ z@(60SXiGD(@tC;wUz@+7sNZ4!YfN#ZnRD6g--k0pnSG#|rhiT?>nmM%3ScV@<|A$6 zZaDgc>(;jQ;kWbKcvbA{A*;LD5)Jy?^V0%N8-)!CCoY?8#lUq5`99N?Y);MR_RP(Z z_|(Q38+)_$2m239!arTP!|QYfpYo6!Kkl4Ha$lSUQ+v7Kv3pfHqx8=sbT9YH> zpRA>fpRv<8!>DT^g(?3C5vU%!!MC=N^>rXfkovb zwzrANFJkD7@)u+|AA5Xw<^di-aYTy^!?K-jxNB)Nrq> zWI@f(_D7!D(hc3+dv9loPj9q`D_ydD1=ZWq2A%1S+U&8c^Dk5~^h}%?{q6Ch*YGG~ zek!ybTuEQvY_P&&z36AW__o<60vPrFr@7p4-914P&)Rt>PX#ZM+ZmAC zGQ6Qo-`L~se4*YnkvitSQf$b;gi>o{wf`s7w?bwlL0=T0;1=KQ>AI>7(35gh8HEKM z8gptI32~>=`?3YbQpXg5`2R+&%m4v^r}%!zoY|r3j}im}fqKkHdRN|Mf$g(00~4BP zQ53DonrYy)$lmS8?byKN>hJr8=Xo$0oLp${F(whZn5N?spX+;^Umbs0h#I<*+<|oa5-5>(2XFdK zaQq)JBAdVFW4E7pj6^fw7snWp`-QUt?GMV%n$%hY(c1WaZ`r)dk?G!=dv_!r2btr; z+}IH^gFLrJWv4Q3nhsj^{@M^+aocT#qs5Qu>0iJpEmUwfwjt`!W%If4kMWc}d2gzm zOXjqOAS-oMQ^rxCPv^7r)m!rwaNwloS)DHNhm+r7}&&y zYD*`D9V-e=k>d|_@Zw9h#W}d6&&e~LlR*z}P#YHWfmPk*B)9)e_qSDSMW+U0Zz3>M z1>#j9jFE5Af{hN?ZFNNj{yd^@V{FS4c!t*xVi?v%1!ftZ*TIvUCQi<*8s4a?TM(l& zrY^cmtC{Nm|5Q`u)hl?r0ptA-$T8&Fi2yX%)fUI{Za!g-577;xI4^cuJJq z&N~~WmXIKceAKD=}bAe)qo?88Q=Lq7xczkhnGGTdN!q8lzvNV3)%g z*FbYguVj13cOh=5)(}}G?|`p^TC;MsbI!4A7H)ct@|aJ)hNtiL@7Kt9>Q%ZlG^Gn7 z>QVb|i;Sge1~yEHA!J~5!!S!Ely%1H&Vxpres20n--Yk-C%%3@)$5E5rG$nKo4>o` z(!9(mzlX0_=W$RShl}q#iRgknlKEsp7ftUOn_}7OO-n~N4p0T<(TdA{CENkkve82x zFPWTUM0Uy*+kHb$1`;_gUZ-P=(Q~3C&SeFI;=ph4N$|o#-xfo|<*^r}6lME*?>`cc zDG$#%RPteRyW0EA=SStlyc5qArP(P0j4J0i&q`(^&OTCls9aKthm#Ff)}hi()GRjO zDmYzn=IQ7!K3c6$I>6IKjf-)>5zJ+Qzi7zg8N{QJ9`U4cFErE@(RJ0^)%KO{vAP0t z(;2&Tx>zN>NUAf-vn$&T9s|@{zruqE0dbawo=rhOu@x~BXHJee@qxrnb6Q?%dG&+8 zf_4_sAMiFlfX`+#$!&t0|36ymD#mc3R7ksn(!={dW-Q~ozs^wgRTPWK1YFcWp~3gL zO6Y%~=G=Qq1b%BLri5gy&#s{sXbckiPmb|^TIUVn?b0u+zc%mBGsj0*ftt~eMQ0Z8 zhhlg4)81oQW&sW=FHwC{3AY{VjorWNH2$(qm?>N+t}-c?3I_-;=mAp5lW=rN{Aspy z;ODwkYm3Uc&q#vBB-iwo*Co2ou^k|hx>1tJiW}jv>unfUKm_Bix;=}x7_?f-i?J0* z&ml^j+6ReU#GksOG;-Qb9f~G@!g35*5xw=@{;EdS+OPzQ0lqVIn!#}SgHC*RZZql* zyo+j~i)g)*1>#%{q+d?%vFhmS9A4>Lq=I1fpHEwVvn_Y)r3nph*5~yMf4^o!EY+4i znX;9N9MYfMZ?m#f<&q)tB@PDOe=Ute`ICF^gFb@Z(_UQCvthW5Zmx00%z!^ge< zMuaDdw!gaBSK&5p@XgYRJGpzu_u1+!!KPZ`1XQ=ktKX@=d)ta$ud$d)9}oSyNLsThR%F@wG2=hjnOX<2eyG z!M2fZA$oA?o$-UN!ZzM{{l@*n?s=RqnRQ_Wpsl!yumQ($H~f8a3gWL>sb+J25mNiWgh~k)bxpi$$?(Ss zB)RZf(DVNaV;gsul?7Y&+pGPP6FvbbTX}!SIvC!1D6ouJZg9(WZe7X% zzjU%#2zXC;VPH{FF<3gZWE~?1b0C>t{IvQVod4~So6|fUixX+)af%wFfT^nZ&YU4$ za?t7I-k=HU&gH#y=+vgr3d9U~d^}IhBzY)G!CaZDbL{Lny7ls7?C=a0)X}}F=~YTT z5N1Sf)YNCPl8YbQ9`De4tp%E4seWZfIX2EsK40f^`s@o7CD`n1a=;65FKs<`urooa z`rKzDTR{E;~Ftm#lLQ3oDvD?`bVv5C{370*^55?xo`n8o7#u{>>V!MUOq&Va2@Nj~xt7cI^1 zvtupbjoIAfKKc{SEsPw=q^7StUeJ}clwe71hByO=1mU#QSK5`#5{)CjgBVb3EKxsV zrpV@#ihp+1qHu!p9dCNNJj;3$I*vADsS|mKE9kkQPo|0D^s}kP6J~aZ89B3vq0|SW z7I!QM8uJ1XMrH?FS_PcaQ2J-Ww|D&L$2m4pT$Dn;P-QpGrFQf4wrc#bI5aJ z+ZT8dQp>io6i}SWGT*~ux3^=i08rqzf=z95IK6bxd1uvB-&8FpuPo~;iiu_3z1)|W zaN((zgx!Youf_@f)DN`HQvjeJ!NQIItRysH?qot2J~eG)xR2UeF7ke&nUgzpy{;Gh z2ECfkDBhzG#w-yM<-3VBx7gWJ)1nhB<#^$7H;#+1&9_(j3sQa&GIP_{9cEHK!lx}{ z{U)ZqOLgT_ibCGfe3$#V_Ui20RAo@iGbOfA9lAHYj^8lQ4u>B0%HTpU z^OcwlX$H^493JrAu34Dk%cd4Ky-UVzy$7BO0y&%I&Bbf!__D`T)tB!T9Wm{fEy~=B z(^TOl1MgXT@9k?ef{lYmR^Wi86jyqYU&CN%J3ab`QOuj;Ie)89VD&8aaNTof7R(lr-JVx*-eg*;^`31h5 zM`CP>fdgYb%W?#84SBk~EM`WAS9t z+R^l~;@H$^Z0B|Dz*=?8qenN5->-bqi~HT_{D(E3_rDCSDZ`b%<>1ML&gB8Xm#B)r z<&gKB#Sy2}hYXV6IFRQ=T{YV+|zQK@*vYIN}P8lOtG0aIoTsiR1yNwF4|#-C_K~RNUwoa!t&F#7lb%70+q=RjY!`v|@o*}N zE06eQ4een=pmr&LCxAuJD8|a4;$5Y}_87Q%9^Iil^Vx?YS6ga zUJCGh$79tr*ZA&$vB)R6mSt^(vFW*;*~L$S#NJTBk$eFDE9H@0O|fDe*&M5!{OB!J z-#ma}ingQa^ZWQiGyU3mcx?IK8|pu5_ilO09Ar{+TjLDHG2x+>vG zu%*7Tk-&xzg`*zne#o`rwz>2lhDIUpK%F%Lo0QA0lhpAn&hc5$MCNIbm6&jFxWx(k zV{}2@5aNeUna84LEzSW%1;*=&7!ZVCHzUf5SDLOkYzKXnTLKfdPs0*`vIk?_y3~ zIC3Pq5Z3?zaTX`@BK`Q}{j+#-0epOT*@V!EeH{y=&35mJPhS zk~*Htc1&gFpVeQs#5Ji4W%xJxC8NV_p~J+RbD5oS17M9Irb+UeW4<

x&7>jkyv1y?|NHsyQNG5t5!{Do`) z><+0OwTTQn*`T)jge}z}V`9gn>Gy6bTl%z5rMM7htE!UucJzH|^$}T}<2vM1RDXaA`D^Eww)rI6zXD$nX`R9; zv9I)20{GvUK1v^&fc>Om0#9kZ9_+vyscyo7AKxAQk*_`9RGp7EQw@2<#c_-1iOHmdxukzL{>y0ZaCz&67+kZbI7w8 z1TAY9^SS$P%7Zd~cVmBWIyuoIazyH0se1PvX;TDoH#3X+_bv);HLPJoBN(>x0z|`j z$LZl-Vq38Q9lT_Tt0*I#Y>jRxkpY!zcfSF;^B(~izc_RO>!WM!Zk_zL$j#!UmcFb= zBDvz67MeVCu zJfX~S&rd!EuL_5hpGx-D)UCR~={~$>+IW9YRxC6I6GTW~^hvmYDjW;0dHBYyG1ugzpW(E@N)NarzqY^o?NBI+9Es7>H{P8 zTO>0Xg5?4)RAcT0k5hfdn}^RZDrC4^@Yuo%43cmbPYXjJH|84DV9aA{w7#hrm71)M z(?`3+V&sE=QyH;0l)kU=X6Lp(qv4|_ojv<1Y;Tr&MRGCb1w(cC7TxM-ptEmNfuhDm z*tIggs}{lt!B6*lZmit2diAC^^>j=Lo@)O$NH*h=D7l@Z8y>gvh#K&7Y7>yum?wP@ z_Zk#-T@Ad(!Z#9b7`W(WcGc2Z-&i%O8=mf}9n;QP70W3Hdto*gS>056L_;f1UBPcJb`G zYW~ZGAyov`e+*Ap|6wHC;5eqdYxk5|ghOz*?fIYk$u4zAFd|13iErk}R5Z^j&GR*}UagP@m$4^YgWNnx;C~6cg7AdOgFC4Pm(%=WlQI=9 z>dos^Aifi4QFyFD=V8A+@UuJQC2DwxMG&r=Lt}O*lQOP%W4KFs5qo0R`0iNrS5*v2 zm8xZ{&B;Iy*Qy)exaqF%jn^%H(HQ?(UGOrmWUsJ;cCViJ#A|uc!yK&Cv78S4R`q=r zVQy)s6|4H>`zx=3cRVNjP|rB!58?%AobfZ_1V#a8EqAgQsNYAAFn^}`@KUp_1Rfv= z6%EvfJI4ExckFcWUr~u@b(784>!k+#zyyFRV%M^_`7rmztKJ;O^`G^a`jea@P|lPj z@%4c(?i{sip~VyPc`t)}eI{#uFfLOKd*-g%zsmEEyF$+MS|Hd~_LYz6sq-%1_7Ddl zhqO9#R%fl2u3ArO=6a&dONk2Zj=VvxC_d8I)!&tVVlJzvu}(M~!$Flthi@mE#+D*T2|- z5D&J05s#NK9r_S#;42}WMP1-ckl}?F>7Rf}ioz0+8MzJMSosFody^zRN4Pp_c$9U! zbe`?}t?~-dM;1~J;&g!R8Jc}5r)KQch|7C(U}(QFZ%jB$F;TgG#FO)K1AMV)O@^b& zn7zjR_na+}%p#whP*HySsd~Qk6*e0Xho;nwiCVw17ck2si0E|Xfta~^x#Af@_{CVy(tAqeTw~qUxmNB<>>>+y}`4|sU*?4 zO1xd}&V7F3*2wy7ok#Mv>*xHMz7nr%`%x)x3-FNBfXReDr{1wTl;LW!fOMVS*Bsmb zx(v;#{zzH{oPJde=ZKGLPq%_pW1JZowBD}D3_Sd>7g%ISmg;+`()?#Um)7MP&5tLI zU2cMi$t*5k&r_*tVHGh?9|>H6+OA%TT27doe$UM$umW6hJ2xbX@T43wbGYy!+CS7b zL8DMYmLT_8Z1ZtFMr1(}{)Vi{qfxE$-U?{H&g;LZ*;u5uVbPI_g{`loO7}L+5c$!{ z+ovL*_(@7Y%au64kfB^roiZlY4BHD#gqYw75kx) z2OZMvq>z4T7ti^eG#{39JpI)NIyWC~)&DD|6P>AIn0)==vBevOgPePF37uA66erro z?!9N*vYq%<`1C=<7t~Je5I-==ZvykYGsS1Y)9wu7qBavaMwOeF8`LF|lxex-wW_}h zmf__(ZPpU>fI9W5WY6!gYnKd!CZ$*ADge|AFZ-^|u2S6orNUcw4g;^hr8kU}wsvca zP2@>MxX4-CI6s*RzJ7D&*}r&`A5-ZFs*4q>*u+fidi#ac%i_IgfZuEM745#=(8+PS z;=(2WBm0_@rq~EU&Rhx5*%Hpc*%@cV_Xp%eN?l+lfIg%v*?LU*ya{88(G_@0{%|Y{ zYCB-ucy=fBlPV8cgleI`%Nx3{D1iYmC99^8IdP+p?F#+y#Cdtb_>M1&HBeJ*Y>N(^<9`|*mTi)pa^ktH-W$Y%R` zYb+n|#OWS@x8@a}zLHpQ#Og{MD;1~lsA;gld;SL-iNM_E=G(U6&ax#OX@)b)dy4NQ zzUFdBHHN;%FRx`iKz0r<;`N?)xB8J9d{e7%}-ucYq z9Vcf~2W{J0ymA|40$$urrJzGy&v#ud#uP~Y+2WDZtjm!_O)izl6+~Qsf4Q&O1m*2U z8|AQOr!MQhhVHF3A&tVXFF83AzSuf(7>O>6z8UpY)~Ql>@Q=g09i?HZZJe)L$|ro@k0Bfq0IjL602g}>9+#gS^YtZ#J& z*?fjfln0j$mdujeHicRm&I~FHERjtE2WnGithV4LIxH8?fp4ak2a5xCoKgaf_suJr zO5)9kwJ5&SVbI~e?PcXCJUh_pDzq{7Z?Jl19Os}B>XG!-#&nR6!ZO_1ep={JbkKg^ zyUjhEZ@|30ouJ%L4(U*i(^Ay+Anz;8{(+AbW2_d8d&LYk+b=v>d5j@TKVKj+83WK#WlzsN^`T5Bw(1 z2WOyPeI>E;;0A#u?$d>FbFa;l?gfBY{``FJt#{vek|X47pEOkZ>&J{UfAslUO4p>Z z`YHpBP+1C1AO`4nO!7hzzhb8!PSZv0-ht6$bua1Jxf(9X2SpR$e zQ+Tn{e@)lbJ%-1klQM%J@Mac?>D6xw1M^W`U1N#tx2?AS7429{a=_O*4Dg zz<6DL#?L?NKsXS-0Zs8cFGjYVsLJ!zYP>%!S*~OIRt)1D_c0 zq_X!f`A3J}L#+WSaaI@U zU2d}cB~7&Nyz{Uo4psB_*XBYLa{$G?jYHcHX2$Y%=8_*cPiZpAPnu&{+mg0_GuEEV z78hI=dc0`Sk13-k4+6d{mIGt9AqzBP0yls(jqZqC?WS*~+7O9P=@m7$5W6sGEBblV zL+<<^G{iAfQ=%a@cD%B@weGRa;e3zqouPq;&KA|+P3bjQ8uBv0O>rhKy)L$&X%Gg{ z8I=8qQ7v-CS%n+9)uz39HRgnwcA!YcEIm#`J_XG;YG-nQhq0XiUzLK;&OoEvfi&n% z`mOG?Nr5R|i(t@?NWjFWV+NB|>_aG63YXDzvK3a7I{JC^+B@P4|9?&idQBhHKt=&X;pJ9>jYFXNw ze9`N|hwjj;(2S-;+?EX_CISl!3p`}Tr5bpfqP`(%V1bJ-_x5f@lN{9DU!^FvBC2Y% z{`-dT1#LZu*Dd3w!yFRXYr<7FtKBqzPFsMX+2$|Jt_llLmMsc#dGLT{I{Y+pSyji! z`_0RzN$yxT!W0wuqX$vruAtdtV(oz@-R6|qT=(_a1jxG~$&0s7L!DlzoK(8+A1s>p zqA|Y(a;+@rMIJ>T3fvf$-qaKkm{k2)hkQ7Ak^b)N#5Gs(@98vJ(9rqh`f1h&UtN>} z(td0bkGO$MI+(^3Z7k5?0a| z`@l_Wo8N^JPRG5QuHNBg*i5Tj-jpa)(d4vGZoIaRre>u}-wfUs7?}6Vo(f{Pa3tOZ z_=}>NtGZ9K10e6w&>$cjn3GA;2hhOv0OubRa~*ZpW3jYh(tn~XcvqKIVn4tc8=KAJ zs_r0ikwCg2iz%VQF{nh58Cz^LQ?vSFfpiue<1-xS&-Dzv>-re9XGmmhNglR%zR|5B z8?=e})O^m%f(~Xs)?JemG~7rY6I|*$*eoOY+_RrnZ0q|fF> zkHz|1fZZmona@iTK4q*-wMC_+iI<=U03Qs*{u$oRq<1up9`l@{2TfkEVONO0-)inJ zg0xDgGkW{X;(LsJZu@4lRi5oZ^@i{5ZMmH=oMp&XV3d4-3*H?}4S&tc1cb%( z`Uio-6x*czU$#(O`_4Xi$wpJQulosHC_Pi z=(=RRnD=X}S@+aAEncNTYV~=>hiut#mtRgT-6R>5g>v)!5((dSVbhRM+K-wnH5@=G ziLPuS)DgB6k($DEohy6e==~Ws8&D$T%&;YvRNI?;>Z+t7FtJX$g*98BQw&lB#+8;} zX4xi?w0nk_Tf?RI(Z|y7kRs^tDwPoT@OD|Pn!=!%;vNL4+4~jVC*x|;@WFE;L~$?zB3_ z*nw#SR+DMS%=UOWso6A*Q=ivZZ?Zb73RU%eC$J2#fcOBPz==>UUliefO&**}THT^nPF zvAzPW7Ew0;J$N1@X>Er}xuyZOczeC{0|GMf9&^?-pBGRJGq?W<*)O5UBLAJ#c5!HA zT8A=T*dnglcSjNK?J#N5VF;CoUZreBPx4F8Si)2B_48C-XzxYRY;srK64(sXO$Ru0 z%YvVhS&m8g#}t@nEHMi5$z9{$Y$WvASF+1J6^CW*7Q^JH=o!kdy({fCza+a9Pp;$D z+GAHauz6%QN=iWI5hUa60)NItsXK&_LHGuO<-gx+fvJ1Gm!V7NzZ$luqsc9se5=#_Y+?;eU9It&kkNs0h+a7xY z$%1*mp+!wEw&sNs(!Nq|G#9`QP&bcP$^Df>(aAK6b8i242{i?;^4^9{;Hn1cDFv&# zHMGEkp+?*~9mo?(>lwQ}gWr1hy>j&v1!8xJY}TzPA5_zP7dU!iHCO7OUsg)!dkXsB z=M}&-K_Fns(Dy3CxjX_n?g%8_4IXseHw?@M#Q=6$f1l;i0s1&_0Afu>Ck~lf`mp)< z8P@OcesF9wex%>9Wrzt)0F{GpBQ*(eP7;R@;U|QEfDb6T{RD!>Wbva4B|_(*Mho)0 zg*-eBJn(lp_Ck>f9gW!SINQBRNjV4kfoV`<;_F}~_^dp16Y;$w5@&Mehz&dfv!lH7 zDaRxLZ*u~8og8Ry^$($x9#5g;y59kq?C!xXoPWo^IBK_Spbu+|NPb7f2!Bz0yqrMb zF$pT^Uc8UZYz&IMJZW(gA(DJ~u{UCuc2evgw?EV3wn;Daj=^PQJ)lOx_O>TVd@AN#L+1^3O+qnYm54&mBG zq^{NiWbr~aW=&xN@h{Mb8D-Sm0cqsAu%+l6^Wy6NX#VdfJqf(0@`Et-hg1LtZ11*E zym0LoY={bG!$snWobvX?n@48uX}wKSw~d@~Bfa$+i10 zo@a1b+zb2Q+!V576ZpYH=pE@5ir#bUOo)J03s}v3CryCc)?y7=>cTlB4|AoSqhht2 zR|DS<3N;tmZ0%izLIIHz*Cl-teX~r_UE|)@pvqG;Lo-UL+;({zk7zG?mKK0|OS-#; zbQn5V&B2xCzwqBj0xcH5(2W9Ev+dX$z!C|Tvk&_dI&2So9PFTh{KuJgD~fZmKv$o| z->vJ(OMst0cqLSj#V&zaMWBvbElxyL_oPt%Tm`-viQtt&lQ#tz7r`q6HMqC50?DE_ zN7`9=Vt_g8DC34Yo=u;xW7{_QeR--cIG1Mu#V;V1Y$b$pkK@D0|F*jeM|RCV@>+b{ z&B*TxtQ4}{U4w@F6$|sNy{9r1?46NjnEW3p^1t5Vi7kJLw&?q9RiQ_Om%!&?e*Sqc zDUs)1QD{cwB3*sdh2kx2bTo1M)YA}mGkg$`RBSng6bNxCYsP^ag1*GT5Xu_ZwF zpW3PRwU{=xUk}`NEfs-{V_4DM<1{PY!;YHwJE5;nbxL{*wbzwbNgHlYOK)}d1%MgO za%8zOCB7QNPK{6V{1mYZ1g}4`rpZzD;C_C%=iP9mj!uXPe#RVdoS~LwUAhK(h{xMY zj}Lvih(G@1cH0ufjVtZN^<{JSs!iaJ7TZc%H%9`~+s>v=Uf6L#!awUpsqTB*KYA%X z`}=*_O=+=CHSjj~aOsOsIXyQsOjF-Z6o_Evd6o4Uc`Wm2=;8U)Z5>fcPYE zB;zD~-{181h}NM2Od{$H@FkN}DMK>QI@-Qe zXuej+cC*|Asz{Y03zWWq7_pV$)D=Sh_jD5e|I>+>z%xs3k<}nW*d`c{|)ZX#0mGQR62v>?^3Q`xLUsB~bdgi$i+$8&(}%X*WdfC`x}@4 z_x`}m*-`4h;7ite+RrZi?{&wT&ZBYe18e}^ebaHD;FClr`Bb4 zDFJ{ziA9xQ`MF z$0(f*W@>*p4P6WFN%b2HUEFWnRNGa{^J9I;aFJa4(`PDR#oX(euDq+*gPI3i{%^iI zZ%iHtiIXoGMJs(Vsuh+M%-GU;;S`XsAX6SPy0&wV{!>@ptYml0N4|y$j~F3ZLwnR# zA@9rR?{kyRVh_4~Ihe%_Id{Ua)u@ z#pv<6@<@=xI1PLm@I>Zz>~!IeXYkYt@&86?K0WcKT%PRmk~okkJ9}EJ<*&i~F0~*s zgwEAlnC{Z5p^r!Q+qcLPb+0Ub&0?xDvFsyA;Hr^;c5U3pLJn`^C&6uBs$Z;{fBn}X z>F7D+;F}lK$uE7=!B)zs?D7+aOWH%UMm~b?KGj{iR&vFt;O(UGrTgH!ial;|1q}*A z3Zl-q)ut8ZsO!@{qM>z+jMekB4L$U8uPdbG1zXz$rfIKmxWi|Nn1{H> zxB`=sn&)$oDM^b?qX;f7ik!>fM9?Iz=CxP2w$hB#p|rcG{^Y&Os(Lgw`qxf=MLF)@8khBM-LHqGh@*I;%*Fv+yJ`jigB|NiCAUzyo41=qc!NYUWcoy z)xhxItSCYx&r|8}x=DpV|G1jmgk!@)(wKp4M?tnDk)4$XRCQS~U`Hamg?h~5ts(AC zo>W;qGZOL9Y|KEnRM0Iwro9Ny5naTNxfZbWy+aDIjm*aFRZtS`!|al%#;)1*@NUY$ zzd;Fs4eEF8Mc84m$M6OfpIC^ggZm+H)9N@^6`xRp05TJ?`g(J_YDTFfw7@vqmpd^}AZ?%bzq~+K%Dt4F=s3a{ii?Y~>CznTPdH)2rDN_87B*enEDUvM9Cr)=w z(mt@-URfP%hirvDvbEVrlYnBAsmlkdR+>m_J}};aQij)l@k$Hw5pL=aF&oc09%4pr z|AC07PVo`;^oMv{M-jGqh#C3hhajegCr|P*_V$OFt#M?5(FG$16#M{P5cAY2K1Lju zWs=nL%N<2H&cn>eshOAygO%12byBXFMMJ#NNwhk4DusgFg7u5VJ9CVt@Q2;gj@<7g zIQkp;$q*n}3V~hzzj^ZJXkaL2HAkxtkH#a_O=iG$hEiqiz}=Z+M1oa}Tf8P|R|+`P zBTALIz=X*eW62;KhgWK2&i8@U$LtQM&(d%q*Tt7PJ0DmmzrM~n)cpD`Ct`k`mV37O zbw=)O2M)+UQ}jd|>Ym*1<6ucVoapFC#GxKabCIQ^ig6y4Rzm&iWWsUn$b8#vOGHvo z|JM4r?w~&GSdmC=n;mYaZ}?WJvLry~SGF3EhP|q7O_SAP0~~ntA=|)-LyHx*HQfcB z9hxN6i(0V39NnfT0IY!pR)<}^$hIO!T23_XLaobFWpUuCsf8#@w6y3!b*KyH3ce};_gfWg1z=)*vjnowO) z(OU>S;X^;7qzPc~KMLVP%}7HJDFXiPLvPF>Yse?+9SzWE6LQ2iTmeF$h_o@dvVq|b7sbrr821dQ3g2sna;i@O<}Sp>Ke?Pc`vBH&_)mgWItmJ|UG zNpuFI%Zh+E7#05N486G~B!68Vf8d}3gI5&6=c~uSsVgeh6#;t{dHmgzOZyE@&S-p| zo5*e<9-6fyy!s*Go?KPs0zJ7Db+(VXC-)C**X5lqew)iS6)ge^E+_($c@U0&OOEd? z0yY#Gy+3RC+-gFiKZH=M*gHk8g5MMYrGqtwCz!Xb2>1grd{qP-XINWqLSpy?K*h*n z;Iks2Vvk~AbTLOQfk{SC>U8_`+P+;#DE zE}QrXq$>4yiOIXdgyiz~l!bDhCSdy&12>s$4<*5#Ay{)Ua6Q4YM35HRkeU^<;_W`y zn|27+)VxVOxQ1Eauu-3qC$N>si8$Bi=LC*V_nQ+TLK8wj6gRI`JS-m}vtgajCDf;$2 z@v`MH+$F5uM)G8}bcnW!yF6B3Acti2WBwqk6(xq%MgzOSV|AlRFszC^y(aDM96+pV ztl_W1gcQpZNKaU4OO&dbe$ta@(@sIj?$eUpCz2gr;z@R&FWEyPS-b=&k4?hE5<9;H z$p61$;A}&Eg9*vS{rsb~kgv7ykf(*sC8ok&n*J_1{&n*S2Y}4g4k@o(aBz830k%;XlZY%Zb<58&}l)F~zJUc$mj^p=I z69*0l%A2EZ3eQOIn6ZYW*bf}y}kI{;&j911+)LoPRvzt4mm zv!j7rx+9z9px~gix?&thl9k?2($jC_E}ouMRXcEox@~$P?%UDRlj)WV*oSW|G@#^+ zz4xE`B{RD-9@fJl67&fy4-j&JN7QrE3-OIl$n@yqLl+y+r6wdn{zo5?(zIBG9*K=! zxhtgm+L%L9d+y}PWQGF;96N-H(lTuieY)rHkYiUBaLtje!*HV-!7mNVaAoyGM}y$4BPrIX&gh zXhXqShL0+=n{NmJxKs%!j*501Y38wv>_2jVd29xOMRw~36X z$v+5p!(yx1N7&OJV*8LnZ1NB@@>nL~^dcua!iKmz(`j8~x29y|Uz(XaJC?AvmtWt~ zVu#&mmo0#c>{h8Jj|&F*qpypiCUc*iPa|E9NV4ZI<*+t`%-ykVqwC zQCC*liDT2?Oe6x-)M3S9lAWy{$(|K$)tRBpvh>8WV^-_pL~^!cb=bJ$fWmAgcrZoBs)t?W;fI=NH`r<3|sqFfzDrMH76Wb^0TVJ%TINLTN7alN+kX8 z&``I`O09{98y(y3NNQYJ)nafNM9o&XU6A$5@v?BjNu*rMaFy3rcMWyRB6f3UYpoTF z9XUxbWz~+M*BYMHI^S~Ik4iZ0wj+Q;`bF28uvbSl=UK+!VM??nq>&3AJnqcs;C=o6O=pk2Yi&nb#84IBQpWmvCuy(iMPiStKzy9uU)%#btYJ2ZQabp zGnXxir>qvcF~Joo`id-HXQ0F-%(s#aaqbNhE~~*AxQ2Lav18k5u&gF$z-BtF^P*h= zu9W2@j5j;3La&5P`atwN2LBshL&~ z0MS1M=#&A07P0j@?nD7Q`TM{o3D`;B2R75KkGswq0scuQ@cf1a6jieNX~&DL=2^0nKN~Ykje7l%;Aq6 z9OfB(*v}BrD6iHpQZ0jcQb;H8tt28|rR!pPbIHuGgJ2b99pnQrjgH zfHUj5sAGrS2JSux_N)OJLFUwQ^Gb4ivW%SB^Q~07WbT=XeK7}e-dqQ#P3$v}6M*{F zu46U1mLm{ZjY1%}pc9eYtQH`-i|p3UnB~m3I}*+sfyy#sfm#xG60w*anH6pMi37@N>p>vSsOf`m&2l(COnp4kWUmyM{~HK~HQ~-|)uZ$827;=OM?9Es z2ZE_f@Q4*pS&^x41_E2y(vq^>#fii`%W2h4_UnOQo04`oYUydkwk)9e*6K#r>9A7m zfmv5+VaB}JimkHNq-I;Ol#PPiGP4cKbv~5|Xs5zf(r${z+fk6at!R-fX7J=g7F>HY z9-(=1S2wx(2m$-*j6x_B4C!ZAH?=0k0qlg%cyu*Jc9j4#9@Qt5))wJ^I2S18)7ybT zwAFJG##N|R#cjywq>Q6=8L%@K)lOMt#UqIhpOGHJklx7{uDIzvoJ*4L@f>4%u{8tF z>~P$T`9&Lhv`p#>Zb$8)s#+qz4m5(gQ9xC=!;7{uYB{YY_ncAQ(wYGCDBnb=9qyf1 z$kQfP3_J4F&h4Xsa@%p@2_4@$Dr4IqgReQZ%VQ)_j&eM%@LP|&5*`;+Rc(os%cWLe zJazR6nM&DCS0HWVXrQVs8h2B`_#s=1Z~zk=o6%ifEW&Z4h1(ja@f6c=K3;R)s+q0k zKH90Hedeprxv&VoH*xGo*pEgJEXr)hwtZPV8J1_ra;U1ZqQIeKaNnH=Mm5KZHo1=N zI0?tc(364bcvWSG6A>FYp7Jc%SwyOVI$m$AI1_OK#aX#CkbhzYaG@!Fi3zD%i|e~r zGLyXYw-%OBS2H6KT~YX*uAq%3R96HR#3$x?)d^5wl$s5+sDA#!BkSuL771W*fg>BT>FD|M8WtSBY{A0W3+K&S zc$5?fbg>!mOBOfGuV1!se#2q`8;F5L%N8Ck0D1J6+&K%leA09fh)1bk=)z=Fi= zL@btARWI#~mU_rQnr&ECJ9ELp1q)`*uNS5UqB*l++01$K1Z3@YV2+$Je`aH2y{>|_ z16EYSviUQcv=+{)7=X2UlAJTMVSx}Gh$f*HEaYTWzyd?#42@_oU&_E3qYqe9{h}iq zYPB{3!{)wlM>j35pD$DcBjr9|3y-W{G}|+szz8@4a6!W|^0uh{2(7t+Oe6k-E4Hb2 z*_`^ta~3UIw5+M&nEGY&XD$_1pB$)-=>m;MWfyn~Xvj}@hOt=3vZRv;+o@E-nUHc5 zj@4>o8O5#ng#mR*6NSDdM-7aJ}Fx}4Cz*o`}n18A3WcC(yFai z`Z!QE88O{)H4$qbM8X=$+Z`wmC*m&abTTW$7fBL17HG= zzW3f(3KZHC8b!f2W2UEcX26DthZ(-0vJ=cq!cUn0PQSJ!F6y|Q~F2N;x>cI=p@$@eeHJifzZ)Kqww#8B=F5oO-P zWyS^&VQqiRC~G@H;J=IkhCX1_7BWO7dj+#6?nTB2u*#IxWiL+5j>c@C-5DE$Y_ANm zy^?HSwHHtov0MvS$;ri&dkO0`P%BA}V)(yx+i`yOS`o&K1@^IxZ(B_drFnE%sD3Ok zbm@7T@D>w_2q!Z8?y>x4Vc7}5B}TmKO-O_0Hz5`A6M#MUCfGR>fKr2Eh6xF_nj^cL zCIE8;?5+twoyqnn6KoCb_fKpj7|5}ul@|1Ap5D5B{i*yv4I7Dwi6>1+B5WiT{~9(D zaq_ag%9(BK&_|13Xf0{0I!&b;$Z3i#gF^>rJ9N|5x;8I zDP1?>^?s^Yb$>3W^%+LQ3!N?)zhyej@TY=s`aZzWE$3_Y=3Sse$v;Q(Z{EkNo=Xgn z2TYal%NwblkrlMp0s0iG-cw0X`JPsizIL?o$C`MJ2}$WJ8UmGnuuv-OzxN6UyTXFp z?i2JpnJqwlPm%eZt5J~KyHTH)E-moJ5~N^wl)#_Ho%bBrf}_fchf7D-nVa%>kB zoWh2>tXQY5e+Q=2B-;>=x=|~}L(QHfq20ZM0aW=pN>$nMF5nz7oaIz}dR9&pW^lDI z=|&UrlwTZUf@rSu&|D{IuB`xyH_@3#aia(}b{xHm2!i|N3xW=&v_$P#1iz9(4=|zK zUW0oj`u@Ipn%-%fg!8I^;^jl-yQ;?o>EQ7IiZ|+{rzNS9Nho-Fe_-2Ch*jN4t9oyL zz#R{a`Fww1Ml}yN5hhfkVY)q-u4)7G7-6zh5lKYjQP&rJ`Cgt1*u563Ea*gv+V`P0 zjrE}UF-(yvz%-z-L>nKER#xsgjs5tzX~0+HV#+jNH@Z84Tx%eC04o4_=mbC@n|(+Q ztVn{B<^bf+JV<)le&^4u?Sxdb&+Lr7J*ESb<}VK^#4{7jq?wU_O&H)5$*~6WNu9c_VnzG{MK6GHhW$)ZZ}$y^sY)^S&cw-+`{$@xFB~- z{|j}5^w9~fg4}6OE|4h%S0}QNYE5yaZbqqI(0lT|l5C)*)27nlT%sM$frE!-1I%1B zf9ezz$%DD&!>Qq%GXYHL7@WNe12wAb1&P!pfl9W|AO>tfn>R5%-Rf<4~ zcAPFP9(l1uYjwrAd5P9m+o`N>bfR%L7Ki#pjw*`*-|be6vM%7T^+klunAmBSI15{t zwBw5sov!WF3Y*m}o$>HQsA?5*9V?#VHx`&Mc)t?j=u58!YJGrQ3lD9W=~tOg+l_YE z)iYaM3jMTFWf5TGHJt1<^OO#`nAO?R!ra$#1D_jY4;Kbz+OL@8VqHOc(6G0vvXarm z?KQyO?9Wc6>C1-*d*6SvoXr*0VPC51FEu=@n`o6WVmXHcJ51#4L$rb!EO5d^V4tvK zTdu8@h3d_>5D&)&GBY}DB9OmfI&em#W@Paq9kO%S7eM9J)9ICNpANh_5g7C6bl~mD zz?d!5fwn0eIJ_|(*kFqFn2^P*=v2R)4y>O7RN`Qc9`Xwgh7a^wQ<~N+Qki}*@ECFa za835LCozagzGk=j8m4=_Fh^B20~hA73;lYEH`dZ|jP1fjIftz;LVo#v^pXAAV8ys> z6RfPRXBTX_w#6Riq%&5$Mfn)xBQwc^* zWxk`Q{J=4^g~OTFk%-talW^Wtk9)~CqelPtL@1daG2Q=N9=76PJBH)t+**W}rt-`6 zPRC-;j^F0c#6M@6-DSmU-PJgb%vJ9Xlqa1;N76-vjNHhCn3Zzt9S3L005UX9CU+}m zB_eCE=4ZE+;F1G4I~~SJyz9zYrL(1feuU{Eo=q<2vdSVml}yA_Hl9~&ZZARpfk;pA z{JhOM4}DpLLz(B;D{-kR+p`3#81*`n`tCKM+uI=_;igVrD>HrM& zz}*$5SP<|tVY#3l82Y+F>%*1D?UZXr@VS`be8PN`4MP^ckSaK%p014liZPPxXmwvI z26kgTP@H6PP~De{QM}2#Ugr6_zDhEd&ZbpxE6P{eCQ;hOT>X)M`STA3x|e9dN{`eb z-RyF@+17)B7iI&MCuuiJ;2Hz8(S!uHJm4X7eBgq8P5{=9f&etr=kI?4|JKRhdGdEL z|7srUj@GoVHzCoUR@2$F`DQlCM2Fq33kuHz|8`7Ups(fHP}riB-m91d%hjGcrH-+M?zKf zwgWHB1@iyd4h%BDULk&A&UsKpNer0V091~Q5l1KntZiUr2(8VgSlw}&CTU$wIMVux z#e|lScpwWlgT%zcJQC2D8x6)@6B6?#4|9~zY0Qfaj0c;6`5M9s^9J)r2KuDq(>!0} zVg3`L)0nqq!De6{et4Saf(K(juc^E1i0$PV&~P|V@p=r19Zu>r$d3%wxhJHl|BCRm z-usC$V7*BL^neMegj+n|e+PhjZySZb8=!AYNMM=MeRkU@jF<;hetX*}Apid^8HtY! zj0ejbM-Cz45toj{5rp04Qf^LbYtLV%d2KWyG2caK^1PAIhG%sG@cMEMyrZQ<0{@o4 zdz=8=Gtbk*;;_!%Y(j#~;%9cmIcvV(Fhyy-253e^7b0-22Yg}{@Nxt6YZDT<&I7(X z08FLT7$;k2LITTeBvjff^NqTum|PDkLtT>=NOK^zr3TJ5A+i0PvJ=~R3k z9ycK|mwCYd$pYSHfEpcLh`?h!;L65Ko=-ABubYs-dwIZC0GPVnF{Lrqn2^BZ2~6Ey zO5n74k2WwKY=(wEC#+}Q#5~tPM@&e}6FkgKN0?gkpq3k`44mCZhz+pTK6N#pi#v5l zYzGk=4SVRKObx6xK)og;@N^HjJ`1?GOJj_gkiZ9fz$d4H2bef~@0dM%8qrcM)5>Z7 z7{1pZ%X1pBK~Ez#$N=WvMc^R221>^<*32r;mA*Z{cah92&Xt+P=slbM|7R8haJj$F zUE><9Ce@eAlbg2BE8oK<&N`Eq-F0VXt`?}nc!R3&n zesnG;`P{7?m`nSp?Q^;DsGu%i`0=@%=ySJiU@pfm-9DE~mj=1Kc%Nn1%-+TX+Co^P99K`KRagoOm-g= z)a1W^bSC%rnd=^y$@#}_pUKt72ARD7*uLAldP(yKbmD=7EFMFlQ#|eUImq&Z(N<(G zS?jUER3G(5pG2Rb!A{3*U-lWt1!X_=xF6e!_Vby;sh%GCF!{#!FqwaRkjXi`YV!Yc zWwd`yPVt#b4$R~Y$8TSgTQis(#^s71mC4CIbG-vI*>S@5nY`qLpeA4a(V3j&GuOCw z!0O&_+4h-SzAVV()ysZto1Ew~*E2Aa-+m9170ZK6wsH+@uxu?Dl%=P8^6UMR>U+oX zG7@euJVD^}r}X`<@z=Il?tD9y;#oouRkx2auopV<&JujH-0W5|GTF-5txVHNx%{ps z^plXm8~Q%Np)p)QJn*~pdFr=EZTZmc^?u);FjFtTMJZjPOF{P)rK zNqEN%Rgl#Eca~s8vmwPuLvx1UMg5}huvRa!T_tLl3wERsWeK5B1EIfKN1Zz{pRJQ9rT9nP=@JfD+Rlz$|^e6s9c zHGBJkfOkvcjQ>ej=BnBBnVU@1yoxW%&loH9t(s5|=cuwy;K><c}G(yQ-=vuvp*Of3jj0?|baLPQIOSp!QT{4lsq+(&e>uxcdA=3)J70$~?zlLC#Q=#UUGjW!$+^ z+*?ClB2`^0=K&zsXX@i*+^x=x?)u0i^@f7*NeDe2$k<#K3s*9F&?U~7aMheqPD+XfY5h|LY#MDq$> zCW*SP>q{_>Nn}@e+nq-rEGBs;$ZKxDp#*m@g_7@1$=_v+MFh(@ZtY&u0u-&l3nkk2_SWqIiZ5g_ay;{)7MpJ;O$#W@o*r zwma9spBpw3%iE-@I048%L>H?wAr~~q5{z2~zb2S$%Plt;xU1K9TMigRnT+XW*ILGO z$W&aYUXxYk80O0+z1M_fgXgK`80JnOSMQKIFQHHM`3S?yx!z77^;a6+H2`Gmj+j zYXKay+jS-QhIhFSeQowwIjeP?;Y3>lmJ@>4)1qM62|$lg=pyQxXOm}hLoid1*+K>t z>B+f#uBPTGaz_&nnUGrGikc`gE;KR}xyp%}<04IZbm3!MKF`BU9Ig78H_%FDoZk+3 zCq-H**cK*uo5{R%3uvWsi#vFmThxu&h%!EfajQAyI#$@#$HlK@5}$c!b*yAt3J$+C zeT#{~UGvjq*StLr>~o4!QysJ8xLfwQQM`@S5Jif|DsG9hN{}M0&*DJw41TT=i^lB~ zI*E8nf)(c_qKkm%BxqmAzvtFbxXI}4SrbxrnqS2;w~oT^6F}ulw~hiT-{4tJPk=b_4z3?bnvziKOYF82KhSKtI@rfRC_yzUDS#Tz317C78U@ zLwsWiR;(NVu|GW#wnKgB__GmdcgKwxb5N*0eF`I#lKTXFpnwGlXp?KXovCc7l%$$@ zjbYy#$!3W0{Ea2J#uI~)*D}Sxxw+}#rRA)Sc8yQ!9xUNZkT^zO%o8G0w2pdq`CkU& z8bMy%b|-W?vFanOSf`!h2c}TB3{gBa7@0E^WjwkJnsu2OQ1W-6KVRJutF9Pldh^O^ z{R3*fK3{m>*kU<=$ipni6_VL9R9B4Cr}!$XXWMSLO=>CyU#J7MaJp}vET_#T-?^y- z!yKL_<7uQcIvy1%BFOt8%F0#dhEV6%pMGkW5Z3g`hstx*mPrEe-X>zpB0jeo!IVpZT zr4K}T1wv*>sH=ycPXV(znA_}%YWbCvr~Zll7gOaewjIHoKVDOU(oUe5mnq6f<1ngs z4&71yi#CunY@K03YT+>Ys?Tg-b0<*otqr_N_<+0WUl&J+ZFu~8Lq7`aUnF??wIvwY zW&8>w^Sd(q3L-b>$RTu2jM;G7TsMi4+ZF&bDt3}?ninZn|KO*Kq zimiC7z0ir+jvYC+HBp^vvywI)Sv1~yJk)PeR|(vO8! z;4%4uyl_XddY+w1^@D6TRkj+KRac0zH9GXU>98YFD?T%_!V24QcVK=V>7Sp}DqanxEqVSVf(Ns2&Estq+P6`~w0NsC z<}UV5QrAVTm>d~g?Lw8PvPjg5*$(jjsG-4PY0s2CgRP!Ysw@o1%A0skp-?f7M=p6* zu(G-~k!a^ykUCkoqW{Q(l%<-=>M+@tW0`7gvB2^SXaF{4$0K^nUOuo1cJd*MaJS@Y zC-0fKN3F?r@ZqiA(iR$?c^Xm<_Qhq1l_SnA++_%IMt^O=lQE&=u2 zAm0eeczQQbaqLVh#w}Siud=!!HM2Pd^;EvB%A7cR0OJbtO@u{uMD7@Z`ZTv6++|%+ z3n(ftL8v8;&vIYdn_mL>+&QN7EW6F>it>T8?{XD6GfyN)uU*Dtst|iDi~3xbd1Vc# zjEu`tZHdlUB+V2XR9Q40j&(+i#WE#eWo!VZ780FqYa$xwWzVg&x=Z`DZ5Gg-Nk&81 zs%viGM&q}xdnsX1^I{Fg)avXQRHK+$e+{Ew^OQ1ErA(%I2E?qj;)!@PY{fJke$=2o zNPqgKA=B6wJJAmBrkRuc59G3;qN$lpwG9nTNh@qaov-B8w@vcyh+o6elop&9=^QDy zsi_MFsWQh}r7KtC6cxygN*Y>5b@~InSNTnc_vD?AmIWpm`i=Z@1!}3!XDEuJ*{cL_ zSB#SqdXup>5sx>@!&g=$!pOXYe%#~MA#$tEbw>qVQdu|*yf?mt&u!c`gg*j!1#3$* z=GxA}PFFWIVuX5Qh$@T5fnO~xL7BZ;EYRTLy7*s*FeB83H~U2yPl&IVuSbZF?9ywd zQ=tB%xm|l~2^+&oZBk{tN2ZTzjri<<^|vyW)$;MoY$wq%oA21vwspqap+3l0?6+<@ zHlvk_koAtkYYM%)%G4tTs>}w?8(D%-3co3!>t!eQn}VzsCZ#EZo>|xPYMN^K3-y9V zvhcXR^tjS5P#)~sLLs>K@qJ;FDZ4Y0(59VH=1g5?RzGDToyk}ErXmflt6*X>S$bArp3=?|ENnH(+`si>{U|Cfx4r_zmmgS zBn`DVxU*!-;sL4YJ38ibOUjA}BKhi1JyoL0S{#5y*^Bb+wd@AS)o&@%$icn?wXQg= zo)(-_eDN;ZqwAX*)WyRB)buS$wD2;=wo-}sK<%oZttsG@ktO(Hc(svL>al_G_gnk5 z8N*GJ?Yhx;YfAK0$*bMH6sU8D3s?WnG`9W=no>|dOBeDU9}V#dqBG^9v%m^Kb2M%_ zYpR>0t@Uxv8D#fj;#9vQV+8~Z&5#v7ol3o8$o-DVlc3IHEgSW9eV^WCt&t^3>)gze z2Tb9-!>{80XwqU*Q0p@b_Rp-Y4Ntc^HjD2H&UUQU4t|%a{xw2$Tx|8%Ky-=B^1L39 zpTB?7Z$U9^UJaSzQ&CVoMS@-os7o7CK&JxDW3u~5V|0bE)Luf8l7muiMneV zEywDpZb3$~0rZwMLbvg472GodvRv*NAQ38!_!>Iq(*Pnyt{w8HJCvIHex z0hT@rsz#NywE!B8SH$DXlc&R4t=XH$mqmHofO?kg+c;7DAo_u|u_oKC?})no#RnVo z8d+gMi8$n}4#y~|4>Ni=$!jdtha=NYHq?cWL_z;QY|Xb;`)FG^WIs{94i?VhdiwqU zgO<|rrf;az0;1X-QJo5_@cU{}2DGi6Zi?%Z(TJ`NsZ8njq3+C)rAq0xqv)0;k5Du2 z%F!mbJqtk zfLhVK?t5XJbCQR54lX_jwCO1Pk<>@K$h={f<*d_89T?DtIv@D5T?Kg8Q6)6o%jGI8 z4tg)&wf|O-j3&KlVqGF`d-IGl_Yk)x8*< zU`698s0SKVStkmqiZII0u5d?m)k6 z-xBHRQDu-`?@Z;-;(j2(u2`bA1!_~HPp!;+MAW#`Zk6HyX&i$dGqNuZde7(XfwPUo@?L%Bu~5qX+4(`FrJzMq>0J{$V3k{ z%8BaTUQGlEnl^kzW1#haeT&Wvvlg>R~Is z*lM@ai?^Ql|Lo8-KcSS1&kwyMMn4;N)+F7;1UHuA3VtFUudH6cCynB9s2lX^{db$V z7Mn^quIZ9eP~A#)L%6LUU`Sk%l?q3r^(#BA*la$1t!7E@#r~ko+_DmiVm61T*(GC% zRh&mJl;t2jCdsyjpsW!)%mJQlbu6-Zy|NquO?d+UIZ)%-gsiaJX*Mc@+yeO#P?xdG0Pptd zq#fq+pc{>M%BVrw$4hB9^FzmF+EvNeH06ybWTXaVE4@qNjVxRa;B+S#>Ou>RDZr$d zomA0LR@9wXQ$uXz4(c~_EmBj+s+B+?C*j)uUdQ8^CCcb5q#F$oG+NPypc@Uh3%H}D zh4S6TZQHRz;Y*0?LMf=bGl};lsmFHW)Q*a}ZKk__Yj=NWw>^IKLbP7EsjQauqo%N9 zx#2dwJ}dL2o9mTtL}gybC_S3)TI0C2{)}0L2!&&~Q+@@w6s0zIws6$Mn{QV`Jz1|Z zdSOyTPBAiuadQxy%)(GCOWs1!Qh$wtf3w zq3~OR)n!Fv?6ILfOSgki9G@5J5lMC%_Di}}yAAbix)7&+y3F>vR&gf%PSNx<=W&&u z=0veYQB6OzImyD;+(gRN)&%vUlEqdwmQWOL7-YUkQdctLYLtn-9N|oV+&0Vo}%5D9VYgnImKs&16a%##}2EXd5yWgxWe_`w(fQ zVBb|D8N9ShkgqoXvO(z1jjBI?mQz45`dF`N=?fql6RCqucbYmS=E7JMkd`jywyeR3+v;uxcG%0Wc9x) zIner>bYD<+1eV{*^v;m!uG(5~r&)sYt=#a^7pHl8Myx#r_3HrEP!x~IniVlVG5|)c zX)TrP3QSzRdNI2{y|Q%eL22L4t?}y)8atz}1&VIxBv{r{t}g!P0hKE@=a&n1zz+;q z#wfuX-&>Ig)Pvfgdc%9VKen*=OlH!(hC1UZE+Ics*k4QQ=uEk?syh+tLhZ;{sIL|1 zZgiGoTkU~v^u6!tMzxeyyWJpjk`&bIdGu;*U zzfIB@jkg*(4fvmjOV7Qwp9QaYf~R`u8Myw$Q$A34GzOP-MDH13gof(w3)K-#rK0gx zV$`D~sLk7l(x~q13uHSyMdmLcC2c3Yy2E?#)MJgR%yEH}qkY@Z_6kEikx`O2(Qz6u zq4mUQlItpDi4&pYWUZ111Vbm{_$B`ss;90eWX)VfJ<-&H>@_l|wbO&kK#6voHa)Ae z(EWo9u^c_p(repL7w8ub?pw`WIPJ^&`E99gGMpPM`Oc;DT%+iS#4rmBKR z^g6@*6fy{84hJ8UPLyM`snD^*J*wvt zv$z0gIcuOc=JYeKoZx@DB-7UFz`V62JC`K<@s^7vozWGdl`$NhT~SA2s|{+uYX!R)P))8m6Y^> zs0BQNN90nk4lq1aY*DzFvSg^_lBAPz62=!U6PdlK@mgwq6?*3Y}ysVk_C4oSWYzADdYG82n8MYeXMwd!$)a#_i&-C zBLbW=Dx=}ai&(4{_8qnrff_5?dFqT3u$_|GvVK0%^POUtC#1OX-8)E?xwhVC32?*8 z_s#`2@&xHjcb;vfT&P|plz23-tixK7aP;)OtRouteh2Wx6W&OVJ!icv@sR9S#A3@) zMbi|rJw`_El2oS^i>;Ywh1*j^1a(Vs=6ne}j4U@wQy=W)8o264Bj;xfye!(mVGIM@ zMNpwP)pLS(s+?H`84i| zai%_`lvC+>?5|n@_K$ZUGY!m%3}KdXP2O?Lzkwb&+=|Yd>L2okrv)} zNqYG$Slj8wpKk8D{7@->>~;<*%dtP6&1mXu;k!rbszKs<#r{(|O(Fp~J8Y-bZnFK= zj0c9OvPjAWE||v=i4!(1qaBY7VJ{Zy>cS&K=sN;Ps9n8GD&7(5Jf1a-0SBK`^21tS zOEex$waHWB>0!ylGR;so$B9aFa3nOtPAipC?ljlRea z0=%P~p*TC8m3SrZYrc#SByE;RE32caMRqb~t*+0+F&322Le(DeXmdsF3hqb!F>@=ksZHFskP_Ga1)f(!;8$+_1 zMa4Mo`OaIkq_UcyIjVn-Qf1*7@LeW{0WinzNOai?V-cHP_w@jL$d1`~W7HX)LlH`b z@z+r=96&L-U}Qz(wlg6XP4Pw}?)Fty&$2kHnyn(hK(~=vR#=&bt9^mnFjhH7sg>PpduAj`v1o&3G#Mux{J!hWEdu zImoblQntgU3#{Z)g#&ftFjW@r0QMHuu(s^gZZ+rIiB7lHiu00n_3$87mb4r$S;!Vq z0Btyyh}(<0r-Gx;m10)iMYNI zeItrXAag=)DTrkx+Mizn7GGLa@v&1C;UIoAmBb5Allu zy&Fh#%o;x=Z6_5?Nplf+Gkb=5OCLkNX5$a+U@FE1Q1MudIRxMh$h4rJ4^wFi@4^?u zvX;Ms13`9r4JmS^9rPPH{!oxD;;%VbEh3tlQZsG7zF`RJu@S0_M_04^;~<4-%Ij{L zRz=-#8`L{#8iBgYhVx{C`XEP{lg_Me?<1)|QR#*(J<^hiv7 zf4-muA`tom4GMcvs6hf|hl zywirdc!zXNaBWhOYukXXn=>YTW}K$GKp8<-R@;1;TsN2$)SuZ=JHS)Zvl#F#QMBw2 z>2J|fD}MI&1BUb!$7L?t9i?);1G6Tc)KH+aAfi2)*n_v=s?|;0o`J?+cJoAsu@7tGf?)D zIL{eFZ6*Lca-gGKV6TEYL=D(SHpHVY_(f9;+wXuIJ5z0@zt|Y)gS((V+|H(3VLJF| zJY_rXkruB5;zQboj90FVlkV+y;u-3?8|>lwVWXdKgM;k&bz$}$0;33?z2Yc>+xk#9ZSOE-<_=UikzBLbv7#}X4)5{> zf$gVuT(Q9G`>$9aoxh7K7hEupMwHD&pMF8EvA$$vWh&te{3W*k z1$veK5?x)<1K?1#XSG~vFtJ8H7%wz8*c^J&jldcF*4+XQWry_~Bh4nVXcbTui*`g^ zlknXtpnR2;nq|c}42+(oZeaihFU&W@mt8AWS=;*EUGHl~jv%_WJ8*6=!& zXjj6;v5ddTcw02?;@BsiD#D>B0_7di2wY}%GG@g%JmSX)K7N<@Z)0Z6ZgF|OV->;Q zVTu!NZA&Lr|Abcy>}HTA2=$Yn@V114-J?KqH-H-A#h(=mqQF%@0VL(OgJkIL~E{g*DpA1x78wI)uldd*TH3NS4?p0X##9+m( zQ+~}!#d%i`ty5EP2JSr>V9l*llqz}(z^c=+4UB0mo0L0G0mjGeRXR~=FpTd~{>PZBNTJx%l) znEO4Lwx1f98wr!C2`knhy>m1X=SJ3m`tbCyby`p#PYj@sC)DO!f%kr@_3?xVWS^m} zPsgq}1H4kD2I>jll=LH(FP*{WxQ9spjFc?=8BpAd#)oKJ59?6b7~&XP|2_lz~e#6CI;7_aR~>3obA z7pcz5p_4>;4ChL@VE~ChcE4wfRh=)9NsO$(1rj-rk#_t-A|uuT`=sPLj*XIlf?K0N zjnTo*3aI#N6gX}jQ1K6Xwsk<|f1=_j3Ax-r?&3lI&4ZkC4z*60O$KIH59X9}47Rrk z<7-skDl(ulJtg)|4=VFsUzwS%@ZOx6cK|!D*UG$CgnN#$W|&y(2Hp|m#f~@^sM4Ci zy*a#^HNs4#FqJES_s#_>rmO%~{v4>7z5;ldKk8NhdFL7143%k{ z8qY$syTUt!eCzZ3u5-Gt(m7v8ZDO_v#9}V&#SfN zhXiZyW8anN5s3Gg_s8o0?gXy-g|5Ai6+5zT7;A=!?Rg=sU5_6=k^ET~0>zDE*rO*R zow3doK3BsID};R^u#aQKBZ&@trk;Jb2q^9u!{!lLjW4C>*(_SK7f_yX+iVA4N$PJd z6gdlyXc~#NdueTaPOlJa8i}`Eph7A<*N|z= z6~pcV23@4pQE}1%g_zGsGj!yPi}>;$uNXijGjF^I*e902I7U$37RA2&_1BAl;$Dgw zMKu$4`Xw)a42`8SFqtXKnBohOinh3QiB$u)NssyDte_Ah`xj#5%Xa~{{E~ym1Td21 zDlZ1A+M*^lk1;KV&i?7eyl5+ta50L|4_(aXkjZVhEgFkpXC|({gnovjuB~Hr9m&;` zY-OdF8jD?csRx6-HB5;0NSL2{FwYUjmt0RizpsDv#GX+X6u-EC@rz%&8z{a^*I}`+ z&PX$K1;2IP1 zaaJn_$glaua8#H?FS7u^oq3`s;+#TMi}DJFHU z3BBBkPa8*K^%X$=tJS5LVRGwC$ev|*9#FBZx)fJ)6rI2EqY-|#hs*;53dqT(atQu8 zgMi8~@0LUXdDt3A7rKvobf@j7d7o}V524WA`1 z3gA#eXBmyT;>t6bZNBNu0B~u((pT?Zn!lo6E`820I4L{HFEeAB=2S`9%e=Ce4P17z zUuI)Q*~`7Mmk(TaieIKTqwE!4*((MvyPsdC=71p2S9)cy9JnlBN-~AknowQAMsSTX z%DP705A9#k`}u{o`h{ujn+Kc1n+IoWy2DCJ^RBM(%k&%^i%ySal^(71uk*Cg~L>@eLQYVAe=xb-T!v$SESFP=-;@rciB>AWD%cPP0G^A5$U=+xCfmF#%mArA>p zV?1J85jiY{JJiZ+i;%dQo|`WTt2^W=VT_MNBmT0}9jg8>g}`_&psPFdd5_}d`|$N` zxzP^Y>gwG}u$+%}4xN6u<|S#uawm)j)rRsC{FM0xx32&;?E_G|GO2xmKJXg)h1XU< z9?56cgY2ZgS^*r+tbSGCOW}#XHbV_PS`9s-hKsK8YNAIZ4r0`*G-Vr))a)y%{+E=h z(l?^LB-(rP8fw!8UQ#ReF2TXqa@6W8^Cdz3*|okhU+PDhF9ntPl2&GU%1tDxZU?93 z@MQA+*NLsH#LGf<=j+XV#7Wn|mATPLg@O1DfjBqvugv9uA$@=G4LV-TM_$W2@wxi6 zekcyP9w^>a0hIgK{(T`9x9)l_#d&xCeJN?rT~8}4MkP@0#!@^Gye?<+EkiN<26j_9 z?Z6hJzSm92DtwEj@&`{S^;5qIjf;;4ulqJa!C^s>}!IpZwLN;15j~UJ23fI zz?hrcfmM74R$kTJWD3k})Cl~7d{y_kUjg!VZ{Cg259?l^1y0xQzukMvjX(uo?Y-|t z>Or)()`<3hlJZ>BuIMo-V>#4%r)gIzk)8jxKzao`{|Ps-a>vJjO^Y;h@0gHG4q@@b z_)sv!03K-4FwdHhFt3xS7=V}hQ6QI#EV}HDi*?xw6SC~9epxSdZXowH$)z#GOzI^j zWZ7~B>FKseobw8s;x;=wci4IRmt zPYOo|N%rKZD*Cye9QwJMTj=)K1@`1{OZLoLJeoa12t$8u816M;xfOF4EyiO~d@K1V zp0O|82cD9o@>^+K99+&@s_7<8$O>#!&;!krZZ!kk#U^*D2{Q(oe-LCGXnNU^EU*mq zr)+lTop#DKCs#btHy*E9(OQ3#Dxl17ip-ziN*%bsn+cI4Ln^HH=-3E+L`*QWB zIP=WT^vxXUOv}q@1(AiFE-uLZ&59ySyB#Q(D)Kj@@Hf$Jf;nL;&OV(8pw752ee3`ZzmRjV%|GJ6lXH-@O~ikmDSYqx52?r|n|BXlL|8cLVv) zG?wD;hV?H^NDAGCi4`2DX2k66B>0bA2*NupY zO!5iTEDx&W>l|ee;#+CL`Nxn3{mbdObg7|}CGO(1TzcHn&G&E%qzWkS!fCk&jw?Y4 zN6P_sp6$kj{zRYd#ti!Z&M{NCe~U{A;z`db=Gd-ZY=#wN|Lt6vQoNnZDa9{0166V> z^4rqcJ-68(@V=ePL-=N<@OG|zfKg5%|0Y5f-y8HHlP3E&_Xu-S0@jtN=ae^Wr^ zFFDMpxGn}vxEH9rol)fT8U8N~=pe%_i@Z+@;gkIFd(D&lH{6?PBvx#dwI(&&iluB{ z&PIblx9AUYjk-R_l|D6~_D&}zCwCSm{W7J^L0w%nNL6_kNMALG+x}bb^}0(PM{zGJ zHqnl22NC@N_i;U&7lN1%53ZGB$K997cDkY>37_V+c#!r>gQT8b8pL|)y^lQ>d$5-V z(Zu$>U*awgX{(n8v8I|$lFaO08pQfK>wcNnCD(Xox850KRh8g2CI|0%{)e#s&i%kX zoh<(0pkGy$$gD-*S8YW3yND8dKyp^5@b;jh@)AsbP~xd1{w91Zeb7u;IKi2i?nOAC znW(2Tj=geLBC^tO-Qp|%0Dug-eOWe zG9fdL<(~r1d}?ZRN4Yi!klD^2J^pMb|NR)SC890nSrfA4`2nqQTah zkYL>e;|wYiz4T^jXA5wOk*DHG1Hirv%oW^F;jnQrZMr~(*6qy zJAmazsC$UBrUQ7FK+`&an%@BxvpRsL-vMLhbpUt$j=kDZ9l%CY>P?pF>Hyv+@cA9U zwT}T6S9AdXdeUU!!$F_! zTY}OPO+qS(4+mB3TY}M#o0OF)d^l*QsuF}so}!66u_o8SHt_`YPXVIB(n(sqaVmf+ z9Jz(+O;1<)2sF_bWNU^%uMAS;S^U~7!b$!CTQsxc*j^db^GXp)k0bv6(05qvwsCf~ ziDcAveEc=35V)pb6qhH?^}{?m*HgCuW@+d{aTND$@%5NThlz4}k+z6~f9LW1_dmBV zNiUeCur2S+suGO-JqPDWdYiBEsMOPcAMk_s=AO#Cqh)+eM|iJU$eA&T9vYwKanbF; z-_vQZ;fGdgW@}7H>-w0=WrN-Jd*2BZ>>me`M*E)J zmGs{{rI#E3D_S}1X^y0D=3we9|1`#Y9h^CsIw-wf5*=G~wQ(?YHG1$ftUpoAe?>8& zpF9)L%*mOGIazv$;!UELHa?RQ!C8aZv)=uT*Uy|ixM%kgjCd9(4ow#Qts6}J?Z=e0 zOi8rk94XYnVLh>jw*d*9q0)i07Lo%rG|P4pWR zlIUoH(TV^1Md-DtiGMIoe5nERz)9C%i@N0x#)SA5A!d}B0f0XGfiAmZVdycT!U;Cts4Lidhi4T?GwQi^f8WFpN&#W+3b zSRC!pXYDJi7x9x-k9_HgS`*r&dxpruPtTAY#+Lvu^H>+qGla`P`@hU1T~Iyx<0`#E z-ZMnF**-lP$M}TSW)OT{6MS9>Hj~6DJU8{c#w){*7v%S)FK3M8^@xXG32S$@Fz%0j zJOZUB1toiVh^o@xBp+_C0bZ7m3qPTRW;)9>j*51D1*nqGzVNanEPBP${mVnx&`=!O zkN!=@LhlHkb+2$}CJFCK!nd!`(t6^+^+w#AOvpNVg_ibK9GL#c%*u+U>JlARG``Sj z>THh4PhX7W$w6_p4N+C8&Sqen(2M+$Rb@U%(n&pk^w{2}*)F*3x*f25Kiv?&APaY2 zw*z*26{vjZx*gbPR`0bNHX49$32@$CyJ7XKK*hy-?S_|M1qyE1Yd37&58w#yCC)y! z?6n(eUIQw&?zI~>$YN!{8d9zkO*VQ1Ie{yK+e|{7 zN{Bzc$$7~%mg4P@Sg(lYq+Xdvtg?9tAlA#i%GJH>D;f!!vcoCdAV-gi-!f0i-5{Go zOSp|vyoo^e>ZBX(;C9=}Ym2bjgX5&FccgxYBaV@(`Pt@5Z{cvN$_An@sNGCL}qZtLcBEh-ChCll`R$3FdRPj9@6Z{r!=s zIY4vO!c!nmzCV(U^8JxG{ck|U8}E}`0}{X{L#UnFJemp26CSpjtToJHt-bcH;o`Q#VQjL%;({Bf+3&!nQ^*Cx7>uRlv7BQTDsvqqosG=11hN{ zApeBVMq%w4x>UCbS&9@YR(>`LhrSQwzkA+roMCc3D9&3h8wBJJ-#8p+o6Orx$V_e+ z$&6>CNxa^K%#=FizjQzD*jB`MLn+_?l&;s@De*&Fl&We2wkVD)Bkzmrse#&}IEF*H zHzI#d;+MY<6yrd)Ro>J6ytLDG|KRZ~CTd{3py;l){KGWV7o-7>_<+o1ai=|G%4ezf zydZ1Ot_W=ve{U!~B`81J%TCi>~W zL}D@6-s!rXEn)vhgRR|W40h}WE%FQ#(oTGXeR>;BiVb?H$#zXhFyCPRNif}@OApqB zjwYesygw49|1oWP8h@}&ulkQ~t-{+3Q{)&HA#ZzrY-Ua2X4d!%sgk9BwouZ!;lD z`yBrAQ%`%3o9tc_63plDL4s+Q(`yKEt3_PS>ynplQknNB=}-7PqwU$KwWZtDv$?9O z3wSn{mb&9-JP4?JYdkB9rkg+Wj8}sopJS~mgJZf0={a7c)Eswx_?bWM!r({(K1H1G zQLZ>E0UYN2$01KU?EGO=lobd|}11Y|n%Xg;jn7)b>?}`kke$M%Y6^llg zvUPk3ek&=AQN}NQsu~{U&7CdySoVVt`~n!do8j_t6P8EuzRcPG#v=p8rH9C1;}c0* z`V|``Q=0KiG=1qhGt0X~l z66W9z`5P0!MkDrS6A}TB&sE%+0Pg-87;}FD_)iuq<_kpqEn#gUE*~CNaxbPf7PaG& zi)S;jR-7YawY-*Q$NwPFk7^P{O0 z;i%2}Z#_mY3^ID5W;DGAJ?Q{=d(g$(Xb9mjE|U@`eQSE=%cRWJ-*TK?b11vZFdh<+ z&j=E>VliwUTzzy2W_)LgJS0Wde+LwAH6^Z<5>NBIk=cx%0jb?xI3ILlfa3X@g1^l# zMWe|oW7dD?mtvY^aSMB`$(q6}mP$xg?t)VEn5_MBfiZh8D8;XsmCr0(q7^*XgtV%C zJ+R;VU>i;LYbGREjR*D-!G2H;H@TK0ui*0Qb35W!?~Kbe7y9_@i@KU$lxiLZg(^9a z^5S3~NSV#JuI3k|ny-bTt<^xiD5{3m`yVCp9TbW&Q;$0SAoahWQU>G?0^SxWi@~Kup6%qO5QtI)_{A3v7abU@?Mer zZpO9by(0OamC{yGGgBK(gjG+g*e!==9eNT#(qy$V>yrdV!Ly*UbF zN|M07xhzGi=r*OIEVcWBQY^?-+HzJB=KJf~Jj~h9vcR@^Bd{lC^17h7pX-YLTq^p) zT!kvRHS2Sc?it2i;Pb(|Pb|UEK~x4~-)$O-x;>x}46q554t9&9rV-mkkl) zx@-v5@_3#CwR_nRYWHi#ME94;+SVR}nHvT!8$xxTIG9Z)jLYS%(<>QE;>scH7hYzB z`#qP-l3~#hW0Fauh^stRUO0qWrb%u!S=TX(Ci%<|W0E=f3SyEc(98ve)1@ zqL~J!^)M~bjz?vc@6QDaY0#{K6I2B!OE?6e?~`0?+fuJxSm{C7x6MkEoI! zdLo~^l@+-_t;tYHuPiUJU1v>Q)Qb7&L$`Z8vblXl`(>AkuZFNs-FcWYLz1tCZ~;F& z%%l00)Jvc*pXztz(pB~vS-+#K>?+S3Mt!EIDZ-U`33OIxON;HuO^7~2y>|o|+mx@W z^ovQG^3Rx6h?{*zHsx~}zmL4hPPfi-GcbD?XLJWK-@vKVf77ugC>^fAq1}su_t@c{-d@Z<>3~A7u3pTSx+;&m3D=6@m3-!P z?x0W{st3yBcui8i-a+vUjwLtK;Z?yndq;)hjfCQ71ibx@y*a3{FErjS1n-VJ zX^nj$@-5rRqwHylxr?7 zkl2#Z3PWEr3fS>pjkcVGajpPV@QZB>{x%7pYL8F7V$YSpnLB#~9+CGDFtow|EH~kp zeO3aG?5r@RW+m{A#(2<5_?T+$*M!v-6_J&|p1XJ$U(gr>YLX_Zujj9KoP=X8Jn}Fk zAMh~Ib-Yucs%+q$f-T3DV3v>d9fNgn$4dAmLJxW+D>kkKUhtvU7IK~g48C+Fe5g%n z=*L$ABX{*UKGTDKZY6xE&_h0MKAHLI*z7rZn_42Wau3FYuseI`E-F5Bs>2 z4)Ajy`VtR%4RIT&ttM1gR9xu*_xaG9=Mg)w(H!06o+HSwSW zA5+b*efn=Yz(Kotnz*%C@t{F=-Oc8s%X%lVy2*8T`>%&$#v>joR^IK!vW$7Vpn1-4zOkeLND&XKN2LX)sVR?&Fb&?@oTxjZCa3$bsB#SI1|< zeCpdC4RUd>=Hg!A;)dN7s<=FZdxdc09&~tY_9@%NgYpUPSxoC;S`-fzZ<<($cbT${ zDRHaA#_vRc344+=6tHq?=4iI3l5JK?d{3nne*=F|@$U%u;J@>Yc*7rhJU%OSN&!Pc zs*kR@k{yuNFS7i%UQsIgAH_-w^^amYrg0(mufRWwr`%bDH$z^1{zFtipI3uGWFjNF z%iwtM1Nps2nL_@S%3U$tFuBx(q_=`%{;6_TtS?ijc&l<(Jk1}URPKsdV>m3j=pVz7 zG&*qz{O~`9;Z9}+s?xZ15zndQYbeI68`;+WE-2$$LseA^@YYbieDsGgo=k5I)tsBp z{@;<$q<7m(v7g%}krqZM*fs%Ow^xRivzUk_!GrRKe12%}?WvEaC4>{7Le0}bO4~|Q6`w=iCdgLo?a|pLDx_sjgxV$w)XMx! zeqXb|dT!P;Te{SNym?9|irO)Y*2ygZf4jZ3= z@aXXdVQ)6VKk#-Fjc~I;G}{7#VsWEB1 zHjM2oQsdRlYf>GiJ-lWrss(Qk+j2q)PMIjJAtJY;zB`N^)HM@1;cYNh|GpG>bs`OX zlnZP$vc65F)VsjWlbE&81$s=@-lEirRW(`>t{izf>dDa6<)z}rFE6EO&X}T5m43wT@=|)ARZR4s#sV%c<>K#UOy!h` z2~^f2Oh{T0FaGx_p7Jg)W&6g^8HVsu6N-1fLNa&QZ@Z$TxZ0U5k0(dXE5W+IUdwa6 z$P?W!L!RqJo~M`?kmq`lXYZ-XyV&M>k!LOwL>?V~bE+rL^};gBTVhsg3ip=sP5)!| z=kn-b+-t(^r8I%-c}GyeFI=GJFm2j<=K~d+UEnFndfWw)CTlvgK5~J*rzuo`CyMx^ zxD$A8TBd5!%^>3uBwr01Uk$3`g;F*Clw=L?LMhtN#ve813r)y{oexF1|2mKhJNQua zKm4FffsEN^NK?s|13PVBnBLbkYTdrS(slfmke+goLKU|Hzbf5zR0)ncNMU>z)K@}F z@BMrwOrW@z^%KR{rJPQlNHz*~>jWAP*Y$H^K6hysbppROI(*H9^k6?>6`kG*bW6eY zoxoaC@FZs4-wC{QkV3_yoxtY@DHOca3G^C}vkCH5C$RH$1;}`;be<-=j#(U!O=nin z#)4e+F9xg4jc*Z>0 z@w|ID74yk-4#AniLE1e6a>T)`GtpL$Vf!v$TcJN`%F|6qZQV_wsI41WQnak5%`e*Snl%KL$4LMpo3DDK~$;(kSr^1Hf#X@=H? zCZut0BiL14z#Fp^Du30LHIPheSwF=ors^}j>8&91XO1v>K66A`&u5OHo_DOx(({=k zsON+0sAu5J5!CSVI#0uAjtFS@%n{Q1rFn&VHNgrKQi9<;%_X@tCUv?AnLC2H-(#>^ zG+&y=cQg^*;4U5^F6!bDY;ZkfxWihl*P$BffQv_P@bLxH!-1Q-7zPO_Mb?W1((;yE2QlwCikz3f9u1*@R*9-DF^P zAq<=E9fX<8b0$6v8a1AsJ(%6*CQ-jm!{w)CX)XHJmjk9Ga~_xGD*k+r1Z2Gb5*P#(Bt! zW;IR92Fo6!va(5OuryOiEzPVbdn%iJKi{?2IULZe=lAjl`<}hlUejKC?eXm67Hy*~ z+D4aX8%4A;de;(dqlor7vxC4!5v@)iTeOXC(KZ^<`066crRx9P$1>ixR83Q=DZMZ4 zA9$ZzBcvQON>m&&sqw*`#s_zb503A{pfBIRqyz30DPCu0P9TeGc9L=4)wcXQjmSQk zFe>>XO)29r$|#PQre1Ojy@WA0aJ2;)-Mw(N?QJEBb1QM^ZB6^CCk^jQPl{SM29ES; zU)fqDeK`hL+t-%qTd}eB)Jid*avuPISC>Dh5U6m;&~fph_!;%fcwONZLXmyIM*o}Yz+7wRp4Cbc7l8>=w#{wJ2a&ud`j1pCQ_yace5ay$i{1Q zCRF!Y5`V7{s8XE0#QCuh7)X#1ul&}ic1Tka|6GVOk|zu?QhmjOY+K8&)wZ>t6ohzC zaSjuQ#M@nGA?f!*pzykyrmlAN^iWQIC?}$ww{6~pAB`*~#LRB2p?r4Z>dt$1V|L!J zTvyB2XE$c&z3zI}g}oFpyD>Ys53aXsWOid$7e2eOsh+SMwU#udLyx>cL+1gl@hVMe zmU!O5=klCtbp{U^YWiRCB1ey+|GM?rWb3oZrOzhO=l&aN>9a}n$-9yI0GmXUk{fMJ zHn}y~WHgcC#!@xlBbrj`UbUn?nZG;oJ0jy}x9D3Ni^p$mOpo74sT{jus~E2CO`-vZ z8(SN5xH0f1TiUHgK=JZ2^_7A3bwKPOc4Ee|4ka|+YKWQ-7#&bamOLV zp*t21Q+IrYI47+e{n$$;?>jZsW9wG*i(AJ%cJ=IWRnH!&o(4CIa@8wed!%~qV0IAL zBh_>N&9=gO+|{$kXzvnDE8qmVlJ)CLU32rZxHLL4&mV|5LRQrf_DfSCUpD?`cq}}( z*qpxz2%6U-7G3jH5=3qvDUFZ>cf*I4b`5EHnLqY&(+UsEBx++2)17 znr6mPFbe4-KfDVp?9eqK)43?onO54IuFesde7xQR5-;wH$fHQD`Y=`E$fIz$jy%fwqbLGgo~0v?{4B;Xj(f_q)GC(Z zxTlghGVaONanE7?;JBxuU&lR_TF8uhPV}Gt{)qiKDrry2W|Jj7>@Ud94?EV&u+9ao zZaUeDc&lyRD4-(#n(SCS>9;+nBHoyRUx&P%H;Li>cvkBBehZRmK2)o@sUv8k5%?f} zb5tzmjj+r+{lQ2GZ^o~vAM1KC@y&QHl_4hXT3I>!P#(W$`B5YAVf@^BvFJR~LQIhG zQ6oBFujCCMNtvVTjs>c&HI6*E0g!b0Sm2(K7Fu;33#{gkzGH#+Mq2P)I~G`@E>o^4 zy{`|;Ef@=YFEmdyq*yPMEH7WtsPgRC4fWmj`#Ik1Q2iXwt}%X;g|?$c06)ib6kR;Z zHs{ar#>Hf9r1~Kh*3ZV!dVI$`B<796Z&D#2jIxlJ81nP=VA&=4Gyb{B^>I-Sv%{H% zgZ(25ax+6X9?vT*7D#qZz`WvdchmCi4e&PeMrHf6FsW(%=vXA*W+8D@7S=R9;*Z75 z+bqPST*o!}Q07QH*_81jbu`y8nduLO@Lp4TLQ*bw9?O|I0v|P{Hhpt-pZFue*hUP* zR1t;c_DJsec`TQ5y>93nlFZ)BjJ|Iyka4}K%ZaR$cgF&Cd10CFDkf#v&#&N*TfQ43BTmSJ0}Y;%c1wt;-QmoIM| zFjPrguPONosrBQ4%6tn+8^-}%gIr(TJPtUZcy+u+UgA$B-8$m|e~@FC@xYsc&k1ju z;$>(`{BfMhL>G4UqCzFxLkdrI_Tq5Rf^W&&@t9J_&|j$OEdY*i2ISlDc05Wo_ZdwI z{f1D;`!-2+P2nO<$@D-2K;O3!oa`ur^OB~-`MMTPKcB&QL{s9tQ41#~+Tf&XN}Ly| z;6H3>?st4HahhAuX?7>P)=R zB#lF-mgK6CTat;P|C1yaITEBdFqXDXw8>|GBT2omTaupP|C1yaIuexswIq*(Pi>QJ zH6)3RSSLBFE)lqPM0Ou3w)v;U*cP#(**I~7i7%_!oY)|9BibP6lF>i)$PopBtW<}1 zX`Ic3u06G>B_V%mQ}$g?M=UvHGOa251E06RLTe`?PnC!qcZFoSqW9FMjNZ*|XTF5$ zVgziLDlpl<=At zn@y=Bn(~85zAx@zQyyM00>4OY#Q7%F2%1vk!wYh9v%BTz<>d#=%(H)H zFzj$lZ|3Iw@|jkfEI~#2nS9;7deD5>s(C};VYxR0F)268+TkB((l@S}z@swE$r?2p ze@fm8$^^NBW>!rkusO0vfJRg%4=TgALOJcKA*UiD66=N)0E{XP4Xol-_K7 zp>9JUnm43YXvRx1K=R)7IKW(q>Z?kR!&oIdMbpaxzGx?ZwRFqM4CDuLvNCg3KU=S= z8v4yL`ppul`c1T*!91g3t|FIdN_FOlIx8oNI&-ApJB~UWhSBZ~ZFN#^G5VcUo!&}z zsHQG;q(*JFFDaaqPiItPjn==bW{vWGqxLC|ex+x*>+#;RtVBOnh_oZ_vf!&+(-aAc za*?Ja--6`-yQV41@3N5m$C{>q@A%Sqv{oD&mlp$d=A{~qmc`?&A`8i9EsJM82<7gI zouMfyC)kuD#IB~?TXAgM8kDyZS1BXvbvlfqr=4x14)*f1>Dg#FOc6>nZPk1NF!OG{jx+&y`|g^y))hTAB6$%YS>FV(*`T>C95r9p zT;2p&*ZkQLvG~Sey{@^ib~)~eWOGeYTLWlJFzakyX3Hph2cwV1lchc@4YmNtJ&UOCIq_P8V#kXO!5 zpKL+iPrhjKN$c3~;HyA3sq}Mfk`(QJqtgAN(sh&VI=H`uta`MQ`M*Ahdc9Sk6`GR9 z1E*52pFiMM%;sL@NMRH!QN>Da#U8X3+k7g;cfRdRM-QWlKR`H(pto-R=a~ESxTZUq4igdpjO#br%Xk4S!n=OYO;51lkJ$QCi`vb zDNL4?n;-UHmlGK!@y?-Il59&x?T#4R9TD5LoJQO6JnoSeT$dWfXRZ_Bh}h|AKJ%#E zkruo`KIR^Mi#Op`(_>cARKAo6z?vk(6YPV@^ z!K)kv{?fF6MI_ZJTylXC(00oVRpEQv#-;bDjb9UKtA{EY@r|@SLLrA##lLKl)OKM@ ziJJ>sGH!l(j~!$dwq!HYAfpcwWHzWkF-nCrnw-Y&`7aFBBjmD9s>40l)|SFJPl+lP zPwTfB5QmgvRcuHxtKlyc>z*I*JG3g>*)%ED14gq4M6)r)wq_5sd)W^`fvb9$uRBZ;j>}xb}O7=LGE}eS3aL<${UoNiT9W$HuKOsGCPz9n|@F&t6Bx9;PWNKS9d8 zb8ixBR%{QfkThUF1|0dzTnH;ksV}VVVuUy>#2l<@M zCn}ye$5z}x830mYZMwYI6E&?ZjVmiZFd}DULCEhIJ?yZqy2LHQFBBma_~r7bi(}Ep z;qr^fq+EWvT%u`WR=&Bh2*)nxZ4*UvEF?xc;8($Yo=*kAk)h0Nf6oABUU667SbRH& z`zINI?1LN^I_J;jK1c;K^RVoSomaJr5%C63D_{>m7-3}Txi<1)8NW?o^n%IGB}9UNuK9i`m;_dg<%>J(16+_sG6GOPB@<{B$= z+9$bv=J5QGJ~+kFuCR+UaIL!1YMT`XR$Uo4C>Br7x6J4etFENFzs%QRPK8$QYF00A zr~}+qVmQCw#w^7%Pl<$ag$pfrgZgXJ4>o(_{FZ=>xE(GDo!s0usX;agyZ09I z`h_gKeC3zzV$pPwdO)d~qnzfT2Mk#BPd%Wv7;ay%B`J3#GQLo$>|P-|`ohvh>I)l% z@86hPM!|#qw^Ma>2Ae}^XSeQKRQD~S`ySH%pGj*~+JqCXax>U!s&bpq_a!a1z>f}5 zF1dj`8}DDq&xFN=7Srz-ZL>@(*hbBNp8%{{OhZ>oCE+3*$;lc$u16>&6&o@$Du;DO z7qHA4#k-5mvAtbaa&jD8Vy20^t~5&k^fPOhI2szxQdOEMy~IEO z`vvinB^np^3+k^+C}ULJJK^vSY8h*Vc~R}RE6Z++mEJ1_)^ArbfCEvfcf!FecS)y-{@#T~7-;jB^q zJU_6wBSks1l&_8yMC=~Mo+qh>kW$1UPC7!&`Leh1ITB82_WK->@8kd~mPI7za+lK>#5vNAt`WOYxICE`B z8E3BT$S!)s{k&hne>Y%~Y{yvZc8s-RUG3YS7R%rFfaSjI zsQR{}n2c=sLPaDr4(mIz33MpszJ(v){b$c~WDs0fDmVBPg#Fogp(BR{dreLl8-=mw zL9N(lgmwQ1c|vFe3Fj9CvWJFpxbTYq2=L{Hc)uEf!!xr+=Z_eH%^jcZTOVgV%xl>2 zM#qu|8=x~EB6)bTBS+Mk4_llKLk?&x?QZO~K~p-!zgzO=&A^>N`bbkf7qNPl+zEX8 zFyl&k97+_@nsJ5w+3Ane=+7!f*67coOF1POWvE?AYVX_$T=a;Aq>t_d?s|m2yZ25Y zMRB?ir|v}HdD3k-k@tO{#sBaP$ptp_G-7Yjlw$W|5#BEKEvflPEv=V9%hi-(T2TsL%Sf9H61dOt(hc=$IREpsX7H{#Pn9<#?E--vt6c+B=`1J~ic0ixm1zJ``h zX(sX}HGUlXPoehZV-^yJ=Vy<@uN@-?8zSX#b>ECW#Y)617Qy3ig@^B(1d2 zYS2Vr%1S!hshV+iPG+v_CQ@6ylKw_CqvpMxq#@tii4D2wDvs1SkGi)L=TVt_mbi+0 zJ2CtoVMaJ7Ff!MFP1w&|wz8X7+qSs36WgKM;@(aY_)RE&K*~S9TJ?NDV2z*PEQsTO zJlttZbVDqA;$%Z6S7_q&x8IO%GDmYXsT%0kXtmL3wP+NzhDYi6;W4Zh2|BH@WnJB= z@7#tAA0tNM3Hg238q3V25^_v?T1TsMCiUDJoh8Mr{U=r*+HZ&>RNP>A-A0Tg4t2QUNn584Vnwz82C=bx zin?0>FL&bXYr>OS0WS;iO+JPrnGru;>vZSxhG?nqmc(uu&Y6#tY025Jas3s!oIQP#AWBSyd@B4CR$Bj6Dca44T#&EtqP z5ARd9*hk!AA2DLH`}#)muXxHrVqWH0v#vKZb9kJ@(0b#f$+5iNvO^qYD$%8 z%G&Up14wkA{YTKCH`*|YItOU=%|zg90UetNq$#LA zOU2#=#6D{wy6Ig&(P-deo+j}CNhx;$1D>^zeA!*nQAd|GZH96!dm77bZrTiUSavJF z$B}CQp+H{0%68> z3ODdPau}$i`0jGfy-4deRNspfecfF^FxS}abW$$93pluebw^N{f*RSNmM^e{lgA^< z;~7mUo1pQ^rE7>5J?}1{N+~oYg?H`(#=c;o)rWTh>jY=_T>w8s$l_-b=d>c=0|7NF z0y6T9Y|RN8Tm(eFXra}OML^n%6f(OAD0Sn!R0L$bXu-GQ3NI>dFm~IaX$F9MX!@_N z@M0PX7zvs&Nq5t+e5rQ>6ST=ZqbYGZlGOEg0|Pc%Xq9<4Fm0n{#*0OYR;nq{rzen_YaQ#v zv3iY&r3+c9PQ=2k#z5AJldM72>{hD-vl%GWzADQmRpyYSgd1vIm9=)G9qTs8wN!G7 zQLoCXp!rT1uvGI+Za>POnHw3^jh8w(e9|&)70B=H&Q{y3Ea2_V9B7>N5@&WffsyuF z@jDWm1~J=wxeQx7bCowDIgzp0+L_~;C4A%z`u)xvqFnwT_pGDCWz~1A=G_gHsT=fV z-7UKtcI_Z75553U*8QBDd;wWnoa_4dc`ttag{5?!&h`q5@6h>b(D3R)z&v|gsrzR zYn2ktB1@We{wr$MXDQaH&Dt$9Ff?*zt-T za;9I!*$79!DX<^JMhUkY8_l3pv{9>9X`_TmK!$>55p=^OAeW$1JKG6zx-DaiTJDH! zEW;~UDf~<_qGg_YRV}kqIG)ns2K&SQ&=|L6Z1Tu)TV}pmX1-YF_g8Jp%y(NxAH*`{ zs{R5>Max|H+TXUd6NGWAZ_88a$Ys<=$W*~XGNSr-zNYGz3dd8bpOc;IAHrFRt`XbB z%WrclzgLytE6Tt2nuT`Z2&}zVEzRWzxAQW`cFPZB6@)_mKxE+P%yHN+R{4_rPsu)# zpWi1t*Jf@r4UTp*|6Q5?F3g*6wk`j=@PO~C(;MTEiW{dT=?+!`%%{9;AQZ@!`2p+V zZyJp(p?Av<1RQ}fa@~T>Ni&UQP8u7_h|QdFV{Y2IM;qY2&DP&|xrfTYKpG(tJ)pW3t|@jL@{yHyF^^cA8>5j+KI3>H| zb?vdrmFv0GoUP~$5s0__BW(QsyabpNfv3MgSqVx}P_hlWh#(mXB&Zzen$lC+k~jw$ z*S&F)K}LoWbA|$Qi1j~o*kX>J`5|)@%Qi&Z?QVnoss{O04D!?)_LA7IX&go!dc#7) z)NCZtQ|LsW`hV&LF7K9n^^C>Q9h-ouXKaRcK1IyKVbSB^shQxVGgg zapaR>^~yKxc30emB07jQ#hX(lbBQFFzcSJFaZ5X(6MmwStu9 zb}=n&ZWp$+s4cqEQX-Oc+F~K5JS>N8=5^umn+02F4=$!4Vtk~emeb0DWmbM(Ugi*g zFf)`H@n;VWW#;7iLt%$?T!CAJd%NhO<-J|F5dYj3+Z^|H;qo23##C7MibO~p?H?D$ ziY|O<;2VmU5@9Cp6Wq3MbF#B|67WdGuqe}%x<1CLTR91+BAUy5MrMxS*`Uo*M%a-O zV+|=|uorD{l_F#C zisW7LE|us#8c5M{-8H2W8>s}p`n~X73(2EKbFvlv)FhxnE94b=-o8n|^Y2=S{(2IC z0#jAJILq`-2EHK&0j1Oi#l2_oLPHW9TWF-M)HDsKCVt_2N<3JbLJlbi@zUC$Sp>Na zf099Qwm&O0L~bTAUAA0a5q2seZk!RyRK`c@y&sA9uA@wBtCh-lJ8i|bdW?*tVzg<^3zs}Z6F0!Wf#{LWu@^qmm8nNn$8}PL%rlKn+wfwKn-|640z@?3ugb| zfH-$LpUohC1aP3s`+->8v(1*~fLodaMw&Ke25aw9ly&Wz9}Bc8%*=`KTp?z3J-WRC z&}LL-AUoGTSYM^LTil|q>xji1J9K#?#hbub^gE{M!>^e8Pwx4`J9!7 zN4m0~X||M)j#sP_uSnSUfoz3_u~=xI`JsiF(n77GRl*{j+s0B47;Ea`{Br@RhsOx| zUsOzWR7;rTR&j={;tXLaTFo+A?f%e0Yx?7yuJp&aCLjIXVAN@V&N7$Hti)k@l4Y)% zr;oDp?FMBRN}24I_=~RQpw$;5W#}Wj6@MYNU&Uvq6@Sr{t@!9i+R+~r9XoC3M(zY4 zL#;Z3R^__*g6-@NP8ey_l&>(96XA&%nf}_MmCSaF_=Af0gNXRtc3Z?BM8t3S?1=b- zh}imL9jRGAh>SOV%vg|xqvAnD@{}Klz0*sw%D*8H4YuQ-BJU{(gd@rUI9K zW+C~Esr){}Ia7hz#PlUi1rmx4--|RQ!Dm=1eJb!S(PU-_zBC8mRf0G({7yhOPX$sG zw3(nuQ-O}VDF#7&%bKO$u|c=)wvaSuDlo~xDbeB|+BmP+I0po0*;JrhaXukV#Z=(D zJ#>x}))W7_A-cyWc3Y&=kjzP!)@zdT2Y7>Omq4~00y9l1>=3A!XKZLXg`?d}{5jf< zvHzw$cH=(Ujj@0J9(%-aRGN$qX7qtui9dE@r|{Mu#-u>D8U0~NcP@-sd)exFz&a4= z=MP|RcYaN&wd6*|;=kP(hllYoV$U*Wbmz3;dFJ^;AN%-R z-SuRfH8|fkuNg;{E4yDPB*&3m_v<*)FMR(dK4<0nGXn)dsY$yhu~`%?u~|{ohq{}Z ze5gBXvUtCSU_2y7e0e`BBMT3AXMGJkz^&hca730uc8DZaN zZXs)AuMqQnVWDBhBE!ATVv{C@{dlMQmSM3N@&&7bqxmXTzCCTt8hTnlT#ZP$(^SL7 ztuz*XC7_sTK#_v3XQ@l30Wn`%Na{Qd`1hBjOq?>s8DirEY@B(5L++J|a|?0E-FuM5 zg*!KK0c6U1w{XL8syFz62q z$<2@S%-6}1E;8D+S+UgE)3ndStko>wu+-bugWUcP^O2QY_mJH@4@ z^?iqhJ^q&?UQ$35m&{6i`p20rXT`{y39 z!{vuP7%m^?GnZn0eXw-yq^S)t5G)+l0sF1f!jSMGh-*JI_A4bbg7-nJ6R_K(GW zL72^qaAskTfjx0VxRx!tWsSnd0sy+bI3Z^bJA{y|1^d6_u@%oj4#f3(nW>HWsg6`Cg6 zO5s?KiTz@b@e|iWQ}L_Nocxopxtr%Xb93_x{n=R9^UKcl@fgu$?zmiNe@{d-jC z{b?Bj=~^&D(=-4R>jFZc(Q(S~ca>J#%GGX{>KHn1jgiw*w@X)nZlyqUyNi#j+g&d_ zt1DL4zvbq{%!t!z+vLjE+GHtrQ7@~l_OQJe-0wK<^sv1+>_$w{G+B_pTORK}&X4$v z!QH*MVE^NB9#1EH45kYFYIqo(UZ(x-G4`88agW~vl){5n^@@R?h$I(wwN+U%G^Iq- zh|deV+FBmSg;2HfGy}{hB^y+(OJ*+cwm-U7xZF}zktt~+1b6u=jg%&u%H)5I(5AswAJ1; z%uJfNkeSs}qLGnNgs^wNq>cyel^zq(gu&|Ct^YK{k zS)F3>d>s$k_yYxbes0~^TD{!ssSvlLXR9^oK&YPpEIH>*3e^BBSuAfG>{n zxL=A$agHZdFV$X>>MYfc%9UZfTmc`I0ME_hz#=-q3@#V={9K{C3Vmu^uFvYq_2fZ$A+00T-d4>dj{8t9P0Qts@~< zqj*}-;yOD3td~~Ntl%^cd~K%#!DoyNpJ_^sFQmrTOb6DS=0Q^CbYMS!w_-^F>*I#FBGeZ8;URUf|gr+eVzjejXh z>@rO$`Z@oX8cB8Js{D_!hN+R4d&zZvFZVi{&;Wx@_sH9bctc`bvtmBij9RmMGiari z6H+t$l}J7G{^fKJ60>u%ujzshB_5_U@}Tu7z8UelGlwWMBluH!FDPG0mLb<#)=!|AUIW!%>Y{Ru7 zotfsjq?rdvYo`M*5=pqHD9%n>uOFHz_q)zi?nR2?aDVnpTd%U(pdAE3_L_@dFphLY zQ~KtoHkaSdR4&D5DVM6+q&A&pb4i$E^l$|DoFK!cN~sTh(QqS|T{f3GXDgR^XDgSC z+N9n*+vZYI8+4c;W2Fj(eP&Z`*jyhS{w8`L5535rM^N_T4}${<(!tvWlc-vQdyhSzLqwZxRP3G z);&l3ji(FJ6;@?aPC7>^KYNZ+E>aYyD!*oN@#M1Fpx6YV{L!*_R4MEyQa-#a9;pc) zB>%um`Jz`3_ab4Q>2C|!^xhlp#ao1LAMR!6^iN{-P4u2>&^GLF9ZI?jv?^Yf#Ty5{o zG4m2sgcBrG(p&=XsErn_t-1KMs{r zY6kIw#2bS{lGCg;X`2z2Z3^Pi+2FsVDe8}}xI)+Yr#gH-v|4t45of!PO)*k4<;5(7|L1v6G0|S%OgwFTcLtYW?%T3c9qGy(DgBJEUaReJz}(ZFf4=Pj<{7O=ux(u?_(TudLIsdE=}}6zJWTu50_d4i5?_|bNvC# z?87c@L!w7###)gx*fDTsT$RZ7&1Ijs1*VQ4YOO0mNo_cI9x)NP{x!onTT@zg87c6c zw!VVK*9U%4aMEUjPP7$7#T_qEwhW%SQB$DAfqZ8 zM?s$v)N}^$#swa0iWQ!%1IkSblHUPzFIK*3c=;Rb5dX9E&BqXIsvF1(41; z{t>NIhenZ$)+y&N@-=AIrfkRPNn9;lzRebhRr!;)ho7YKCjZN07WD9w)ZSzN^6mUXyu;)g!m&c19++(u}@IwWX1^<1RMp@ceoXqd$sY&xmU9(f7H%)`?*)s4aGrA zu4asF(B6Z@k@=A!{)qW1@Vu+J>e7wbY;iotJv1jzt_+)hH4m?p5EQ|Js~;K=i!JRv zh$*7VBe3-9Zp~v6__xPgt+iBG9sIWkF)2#{?wdYvHTO+3Qap$$V^&rerB|~NJfA|H zV=9*dZSqHBoyY(*EZ=UF`b^Wr5qML$d?y8>8KBdb8I!g+ACRCtD%EphP%g5u?6<{A ziB%&gOSu32eZ%ph4@^q0Xlqk`jb&=T(^X!Z@^8edK{3+YR()y&H{H;O z>a8D&x88N32fXO+!>g~!j>U5q+P?dtIFy{d{7Ag!n+rYQF)AFrnkW4(y+}Rps36{W zkq5r`nZO!t882x{BjudWGsLua$DKx!VVV*;j!M(^doEJjH@VmYvHh(V)As#m0-LqK z7EM|39u}N16PPAwQU8uZbJet?rbN4y+@-e5)Xc+JkQ`T$5o^1WKUmurT*9{TkK#ei zGpmk~Grlw`8e=c$YeK_!V>6K^yP?b-z6R-j`C+_`$KrRFXs_wU8+^|h!7aWM^g+-`?3d~&K!RzgcU;c zqRRwdzMIrD$Df-W#)EzN*{bUZ2uDI;Jk*zyihD0}uijpM8UEXMQzX_tR2z(h@ML`T znOZ)-RO?$;Ps+!w%(ea;RK8@M8xGxLtaaJvCQTeQ0#Eei=R-DL;XzF0a!y+OFIS<@?4eKi6x__&IHO=m>9B@UbSH+aBC+I zT5X*PEbioikKf*!waaK%swv$vr2!z{-r8tL?VJh3-DgNWNK&!0fJ%~TItysp*@ISG z3A&=Q(rmoj&`i>lG&|cgZ|zL+q)80>%-~JXlz8{ENeFbQ=5ElG(B%SUt4r5f8LBCv zZ&Eh4x+&!0``0WWqTq#^5^WpNZkYuR_; z>RCX!!avfKc-u+#{aHY3;)x2)_ZbB{YD%>CLos?*ri5-0C>5;K+yj~tdYCMz zU@u`ow59tEixrv@?Wm1*o1pa`4Xiw1&|cD%Xh`I@Rz?G}ukzG*KO?AhEeKo`EW1h< zJT6Mp1&>liaTYvoOXG-hFEar2`@)DZR#S>$QH*f`pe)S;->d+Tu237R zE`UK?0$CdX4pvi}qIhMRlA6b+*0PIID^{qDRfF2KU0A38m#8Z1->FVat*Q}Q*Y~yB zjsez7XqeN*gVqr|)0ctJ_Z&HTYhQhRvQeFOna&!H%_4Z%7}13B8U0KcpV5zD{F^Qs z#%J_n7@yYFbfGi)F^r$n%?{%;`kBbB%g?yCA7y)_n+Gv*_w#zU5m?cWq0ZBtYdXwg zsJpKpLv@?(w3&Ec!a*ZSYfV`LajXG4>7B%j=Enew?l-}s2|<^S1xmYnARhxfzlR6% zF~HtEG_WFI?tbO>VVXf zG+k99d(4qgW+2R4?d&%{cuGp{xOePd{+*#}>S0YkSu|PG&tDLWDLp-C8wA$$+LbIrb@jc?rS3zb-XMQdOigYm>Rg6*5WL1S#P#i&hXL{l%A+dBL@VuXBJSRpuQ|s zZ#Gb`j|WMo%?2*$<3Xzyvw;WtkPeY66?uS-{AwQ$l8J0EzYt7LNmELVFhht*VSHDU zNZZ*!k%F!z=<3GIpc`z^7J^(e(Z8v;>S<`IQ*{|t##r`2gi7TxlstG4p-Ep4 zTAe>24wv<%msGdgQ!wksYxf<}vTy9Kl|8>ddC_S~6ggl!&9ed`iwXsWNQW$T2#BTP z9yTn-+NI(K)U@g;763FFcF&AtIyMdd>9%cge-jCd`!f<=IKYIy;{IE1iN$RLY*QEa zH=!>p3)B0vr9C)+?JkTN{TVTL@JY_o&Fs&%+h?E$Z6=0NWa#+eY1Ko?tjr)jJWU$3 zz&Ui(=a}?|c^uF`#vdBOuS3u2&t=QugE-xN#8rX!_h++SJBT)jDS4C^CHhAc1hVm7 zf71P8kl9n@+l3K)(4X`1l)+s4^arxL@s;%5LSTfQr^nIvNbdd2^@l?EP!#@nFu6a@ zox#j(Y!wDC4>n7b@8RvL);hL_9Cdt|6#oHW0VUNb@E^ zugnH=1w`ixDrk%CTyq6P*Dg~~B}>t@pT4F>Xxi>z#;rE=ciWi1T%(QP>}$0V47rxV zkYk1-@3T3M5zyPSfl>wiL^?ms29^nk6V13M42$mxqA4n_Ra4ZvPA8XRuk&c!E>Yw! zh)h$QeLX?sUZtRKZ0ECoC>C*wql>TfYQ|n0r zqSloPnrmzQqJUT`Wv$VAh4gyIX9GdaxsP;`=K#9}nRMcwG{_Iwbeh~m?(~+Rg6?N2 zddp=usSn(Alls5{K_=63MSg_H6!kfRWT89m52OAtO{xClHoI?$CdR}%BguSHpfOJ$ zR?C>E?siTQAgd3b+hH*dtz0L&e%J8$yM)Ky!*rndyM(>RhUp$kib_+r4a_m!-NR^? z4s(E@g6a{JHOJigmNa?}aNf-xBo)j7?!1{1<%CU*lZ|d2aozp0EpeG6xjC0ws;r~P zn*DO=&Dt-&CLI5ffy~gU>p1Q@Dg6YG+d}gOm=0v#0Cpf>-mJS$^9C?{r`%%qAoB*W zqu?4y3hW0iVJ?*X@wccYD^%u2l$oA!-!1>xIMt%Mtz%1D?nrH{gEHemXOI)EQ+|tD zXTPxgN7m8)RWDw$SrpZA8|YCr(4%6Ys9Q}B`{)2(yp(*aZJ$R6I6drR1K7i6+{(4z zGJs1SPYmFy`I1}dRov|es_<`-Kx7C6$){7JI>0evD=aGRE^yZSqRPW8hJ}J zWt+Qzl(`VviRDjN2z|;J&shkK5Ub|0?IkmSBE@SAWI)~mhOi@j4bjlI_o`0kci+>vS9d%dxDJ2_6N4;EhChg8LN(jb`}5(?fiep!9D89o?bwG9Iw3^*$z7O!R7dK-o0K;m{$ z)pv)V$M^nHyAKBNTgh?j{ah%08c59c=SDJdZ~zCqA2T;?JqN;J9Fm|JJAy4oT5v_Q z5vYy|Mz;%5d$!=z5uOvbbV}D7A9h-B4zX&sU|w)qp?Eo@#bZnvFDa?&loX#kv*OIAlY+nS$`z8|6iQ?x|~3Ee&IF2Y+fyFGb(Fn zGo;Om2b#{TcpyEx-ALon#RGpWh(-EHyJm|AN~|`qLm%kT(+4te-!oFDMg!U7ZQ|32 zT+APM|G-!@7^VGOn(8%|dWrXrqW99`MG9I>(C&%AxKTVWJP}wiN=^SVk!8B@Qj8J* zGfk<)G8W`$aUTnEJEU%oj*$oF&}jclow9IxtyTZy;`VtDk3LBTBQD}wm z{$DCg=5VI%JA8t1Hb1KRnAX>0QeQ9UnEHBb;KO5L@l%dnUyqsk;&9_}sg!oN$%HTz z!kU5HR4u%XP3{@f*FH6neeLhJu|C*`r95j~@M&tqKJ3EL+J}wikH8FI^=KmVmD8Xi zKSyM~mipdkk9jS%ZZ1uKg8JQZbNm50F_GLpJtr{IQ4o#Yif$Ze&h~5+7i*i#acE>L zuyG*YhsnydRo$qnCh`VSvswS{Kz@|)kNYwe%pAe*65#cL48=#89W8z0bH+-qQE@tSe1Lr-XAQZh z8%-I)z*O)Pls<46Z#eg z7>ujUv{lR$_m0WNRxwR8PrgUKpOb|T#Wo3-=X>CLX9Cc@v7w!zDaGAJ<$0zjm`{hG zT~gF86$H^P3kW)~3D))lqlvr4*~*l}8^(5vD^d>clLK4ao_uX_R`8@Qce06vMriRs zn?cH%HY4hGNSoRrX=<&5rl}noSdusJN3(r$VBsl|VBCrHwhkDuBHWwPaINe02;&gU^$3hC!D>F1Q z%#5pNHo(gfJ!ATj%F;~fQKmK&y;Ckwr@(_G9yaY(K%Y9n*0qGIdCt_9utF(YC!RAc zF0dQk$c$x4T``Owmn>7xqGN0L%tJt_49MET~z!9)w&Y^Fq+ zT_TtYl%;d^S6hEsI@e?QkP55JrqzBf92#-8Xg!n8QwS`P0_z#O#L&5LFjH!kc+M_C z?$5coKj-HDoK0)oTsSn+99l_h<^t(sY*W5wV?W2E?~0Y7skI{M%elZ87a_h&_N@&m$4Z&z?w6(zVW zawJ`^k~|_tFBoU@Jd)1Q2z3o!m`9UlY9$({_ zw;150yiDKkaH_1q@o2i$T6)5x=^WSjmW=1vRcwtP({JlhA6v=C<#oa+kl#HgGq-n# z!5qEut5QLKsQRN?4v)%KZtW%wvf5IgNrR|Q>pQ5P|8`)~Ao_f-J8bh$8su{ONrM>9 z&DpZa0=VrCGr+`T0d2X%!`+|M_OfGC($U@?N;S9lm2{kuQXh!s=iTC7FrC+A)?e10 z!14(kShXnu(iJp?paCVoe+a6rb|jRUO`u- ziv_FPAijz@bfUhBxqvt)*D2rVWa|=_=+@~$Tc-y_r{^ZJ^UVSt9K;TKCm%z+1V5Wu zY&vmwd61ZugU1H__i#h>;**#T4~e?3-|benLx+>7+u_0*>L$EsYVILh-P0!7>ZZFv zRJX$N0CrzsxUI4ed9CU;pJ52>!7V$JSLfY`%Yx3?cRp;t-sC2J_5kluSl3Gx+Hz>>t#3ZbOt!wj0>~L4TGul%6pa)_!S4XhX^TAsiUg zdWU>7Du9E7c=Pm!leM`W6vXSMoEUb@kWS`@?Yo2>MdJQt>}IVP$pgezTR+f{6Xab} z_?A;9UqS&IGRWLK)t0Rx{RfHRP!`VE{Za$0pXxzkpde39AfgEe0h6b3Q#KqKIxgtP zX)880Km~Kcp)AR1%$&yeNL0`}J%4I!M}-qJcph=N6=IN(O=P+A^QXoVHBOo_A~$m+ zmRo1O&;V#ymSl9q1tv|*!I_k?Xw7b$d0dX(Wz#+Iy;2|O*v3##bCr9L)GO-)H%(^~J7d#n$k+%W`4B;u zZ8{AT2?CROZH%%KEAi#t|s$B)nwQX(g|8MP-%1@#8Yb zpOa>k0DWe9kTkypcyy)*$;&B6@Lab^CCBDB{oV>mPS5Un~ndlq%XA`&%pO z0(jiA+KytvgJ1_{@`Z=7S-Az-e&+dGH^AAmJZLw1Oc>VVR_AdXCXSNd?~jcIV$xp% z-%?&{v5Ofz%e?lt*5V1>b+c%ewIx8wOOVrd>j5NRFi(`M9#N{eW}a3sJnWA+A{Jff z7Il*mb(4sC(QFUe76O~B9ZxpEXg>LE!8eKEiF4TC*?AI+J>Jc8>;}KdG9A3wD1_HU zn}u^cNGuHFWh-i31H8*8>iwFiy=^Yv{g4S&#Gid_W^NdpMa}a|Jcy|fGmOFO)~0c> zD4NIBr~kD|>Y~+mvH0tHO`T2p>KpI@|0O34RKTjYU3>!{%C zH6@z!k^K%tb2}Dq@XF`qH+H?0UTnA+SE{n)@MD$ecg-SNU$8YNZLw|rDyw@3UZPR< zG7S>M&lcm)AB(wLSPmo($DdYdYJFU|ge!I>%<>1a)APfa<{{>_%!jYjJmC6^j0Cxw z(xp8XkaW{LU>xzHiB?r^j2UHvRtPA29+2{i0YwuugD>7LVH2ANgqM2IYB^Dsc@X{Z zJRsv0;3dEWme=P2BbKo$2r8)!nodwH>&bY`Hd?|ZMs#aM4*(rx1EvF;(VAtpp*nP@ zom_1t0=mUF0>hO|aYL?IeTY89DS9fbki-E*9hwKHe0nW9VWq<Yyw|qL`K(5_8zfzzlg#AyU9z*g z<>!yi@jJy6uCR^l1+ddAR^I95bm!K4IizFe1I=Tx_+D-y;8S^@_q}^{?82eRi{9#? z3(vQ`=oOR2q`WGNoV%n%&lU9j-t&Q?R{`?kRlZ3%^MUmQ$@%l%ml9r!M-lqzO7|B{NuPJt3jKonmHwLhmHtpgv(anNKmP#_g#NGxNZ;l6uF8=< z&8d)-o<=Q8DrQ?9shDkbSTWllFumEfI`o&AQV-e|gmVHT(?j`V#{t{wFc36&PB7wK zb*Ys1G1h_1QPW>DzL@UO>w<0|Oy;O1ZV)GJC6DQ(ZG+(O)}?aA8N=edb*YLtLyLi) zkF(R{`x|XLntJS~Dcxuq1@MLD15-K~=t@lqWmk~2Xg)CEaSvMEHy>!Y(t}oy&Ih`! zw8`29wzvvPP>trNIIC1JN}9w^agJE&K{P+bnX(x*KgId#N{@buvsJgdfR?K$TlACj zf#7CioYmCso%ukQRURaNG9P>yZXZKhnh}1KrWAf7k$;^JEFrRI0k5l$Spd99y3)v} zDE2~4S?VR0Vk6&8G#RawD)(}NI9j0`GFr**ZrB!S%3+8zT6vh#3ffWyKcgwpoYBe~ zM1wDW0dRd6LoQcS%JV6irYr!yTJ1s7Web3WCp<_>TL2V2!7A{K2GY8kg4b)xfpSc4kc98MCp>5!!fpuwagEk+FOLm%XB{qz zHM@8X$CaBIaM?t9gosLKziSQSF2%*`raE~PPDI>9(BFxe>WG%o%cyGt(g$@cUQ3ka z$A=Oae4jfaeIO#m)md9(8lICK$m}wtTc17-p^6#=7e~o>aB&nvMVqyDZ)Tp^C#J$$ z97UDj>%Ra<*V4I~()tNrASrhNFk-C-ts)D6M=6AFV7wO*#Vgj7_~#LSO1u}l)_RaM zE8dGgiSM$B8A9EdAMiWkR`qtp)J@Wq zE<2q@U9te^^gQDbdzeQQyjfGC&E~)8LAL>=ZxSwZe{ zS!1MY&XdgXEqOa0xdV+16Er2@asHD_aY___C39S!dD3c>kawX|i5Wu-WfM=HuczN( zd0woj74f_n7>S7?Y^WFbvH^BK?=fmsDd!h#wc2c;D^*V@{#t|%4(qa^ZV}3C5z5>m zlo=6PSJzxwy*}SHfbY%?9ue#*c`V(af~9ORMt#~A?2~^iSgIpX!nJO}R@s8BatpS~ z2$mSaYLOuR1rJ)&=T_IF&-pT5@F05EHD@D3C0b31s;)U3`(9w5K<6k?&=UmFIkI1@ zX+_uY)IT=p&x&yW$_yE88Ov(O`fURlsD9gk&FRG#?dJ7u0|_R+Een8-T40o>^queO z0!ft%fZZ>8kbFpn4L|Td#}lqvb;(nh8Hj{s56uxF<2tw5ju}yoi71^n+M*nDiSm(( z04+#S{;);K+*m`DDSs_W$bWl5PRJi-EKYT#$hh7u%`-7#m}g>Wm{%QXo{2Gr;p?|W zD#b!gDb2-E`ugoLN;7J-48DlGL6Np-O5{syp>y?HXnoB@x)4jAAFg!+(>4mAgnn<%bgs#yI`H* ztuA*>fd5r<{`)WX_VE3-04P=$SgI)vaW5yGO&0?7USTs#SO^5)f`GV4oZMp}$ISf~ z0-ay+pw;yYfrZ5Mtvwiz$PLD(Q#2*Pl_a?7U_8Em#e<}ZgYmfYRS%MOAB@Lg{`l%( zJknnCz)<*6OUB%2D3QWCQeeR>LCabQBq_L9Q=&afv>6M5a-u~qS_qV?o}x(A$d|lsQ(mD+ziUe5)ise95ZUG5IpK6VNmhRDkiwiu z)+qHj+rx3Ik{#k@>mhy*^hayX+xa5g>#g* zc<2-_2*QFbYypP@K-}8^%cnAe$!zZ}U7Jo*(S~VCA8=-SGu~pQatS-4;1W%V=FIli z5)Hn}HBIq|4cC-RCC2`DO;c37=i%9m4}eodPp`)ym(G}ZJ9f~e_}Z>y#gGL0pdYDyIxO)q}O*0fB) z?`ldkN7L(x=JwpA3&QzXqr*H4;SVLZ&&m(*^odhzR4C zaPg1(xo2~5{j@v+XK#sah}d^MkbPW#I1HT4s!o5`cJs6UfZ4cBf&S(oofA^)Y>ukO zzUx6EUzta9#>xBmn2WpkyU}|bQDvgYnmjocH}feo59w=~#$qEMM)9KD=45yFRvm#i zC&@SXG#r<4*i(0#dp#+&!-e?BdS_JwSmyhQT(8`2RmW>SA8J(U^AoqeH|C!10E-n^eMBJoKL=YP_#Om*7$f)AgDuf$ zy9eiee6S_ZYTv<@_|JAaCFdYNW*a|h>^CX7T*amly(Rfs*8Gpi!y44%#CM-mEFn7)^}u33;aQ^_mhoo}!Y-^<4fEkxCT2R8yiA5{-6# zU(hJhOA7u`QwNRIek7XMq)ctHh!kj(yj^OO-Meboq|;}NmNX+bD@jduj#y;%XIiAg z?ixjw?q-qO4!1aGVpe1<<3}5=9mf5@ zEF<3ql&@KGrKjeB{gZVe)yKb69flPSWA{a)QaT*W&5883Pnp?>3FF+7K4@9h-HI8n zXH?q#ih0&`fd@@UZZ|R10hSrcn@%NAkcYIGYZM%B$QTFMY*}s1 zS@6vkWA+*QZM8OwaHh2-X_*vF$$BSQv$b8azb0$-gf7|Hq2%@h{N|vQ&9z7=816sQ zuE5XIt={0U{@Jj0sZ)cdDrA#NnBZ3Em?3#gNKSA_9y273o&S)BH}<<`O0BAU-dukG zt=Q`u-B+r_FKJ5M|FCtx;s6`>C5Kv~OdGew@o&FFEphV!50b7s)Dq7-IF*WXoW+Ur zuP+p5>K7-qp*=Was>_Z6M*x^WXTRULNp8JA@yHPA6A#Bo$G)(~NS}Dva#OzKGw_Lr zO*h0R=LFm*9*$d{VIEJw$;ip%{q&{n3!ixGu~RCnPdpsjATbk_l6}QN_18)PlpIvE z%9+xst}R#&Y1X@jG=KLjr^%*zt51`$X4Xp^svZOV#dAztg$_G*h*TFH;t-#CLD`Gk zDWcCm#8FW=qL0yEu@+*|-e;Q?wB3T{|4Px!<7Y&RDXqrZCs>yqRxI=Q)M2_tw?#nJ z`zG#RPuIv_1T_5GgQUVmKzII_x(Jx|wKnn{>de1rN^hHEabJtwS<0QJ%F{I^bULAk zUbqNI`M@yBBBSRP0dJEmd+=YhK%I$(01LiIFJuG$=4+}g6oN|O7Mnuc5szsG>35lu z!!)fy;ld-NAR-nke2u2WBQ-<6OzGcf)4$fHzq!cJuh5j`90|t>{S)H>+XW{N!OX}g z#+$IyC=#X^awg6!@Ji!Y;AP`jal}1ISRjW0UOwVMJ5%M>0m%&0Iy!snnK=z?Dmt|hg9=;-2!u=vazvNrH zJ6I$o?&dSuEfQ{RzN2Yn&|CVUsov!@9tXX5e#d+54z|R3RYafgg31raYhF%*UiqHp_(+@P+k&$72TnYS z?9BHh^U9;t$VaEo)vBCjrvP4|(~DgN8JgJX)<@g5_4(DccfLhf2a+TaMWP+F5Ofhk$mUPl?5gKbe`E0pFNQC0dd;_%6lCn9M1P@l?+d8aVxU4noT(%aTFf=OoB1Enw=wzH$a4p2&RqE&pIBf4Kcpknps5;eI#IZHDJI;rYZd9>bBZ zwePa{wf23-_}b~mK$}p=AHEg4tu-Ba+byC?nlwW26Jw`6wA15F8sV&8X{SAlfieYs zLDPOuo3=W>7#QpzR4Ky2nh1*>gp?hI`1dstJ|#kpS4?XgzP94+N9GrXGI=Ua6%T64 zTl?KH_-_euW|ry{PkX>^$M-EuH-z7}xFOu~H}3R?{9$}(y>M47#{VX_a_$g2`_g~s z0u}ezKCmYCi@`m=^AiYS;ty>TpGq{BbL9sj<_q8>BcmLdDoWk5{bB0(58=J(ciXT2 zFn;A@GwQ9%6l==n_$RBCt?l#Q89aSUfaXfKpQc1>z~FgGY2`dxZ?m^zOIGx_Tf#*i z@%%*|dj6$a;FF%GUgY7?vA6%QHCQBLMZ^SmXb-hSGORz@LzUT?4@+i1 zGAr%O$0T!?WXA2(;%gw3G$T(fE|w* zQU^38smqD>c>s7)(74qR^O!-))s$!*Ex@&V@ePrqC+%&45@pwgbe`SY0+YNH{Zwuz z6)C~QuD+_ZhNC4N&ikes-wI<+QCFv)JchqZ%6JMNDuP4<{Akpr_M4v8Y?miCJjef$s&xL1l%4 zE+UA7%EqTN@Us;5nodzk9L*7?yTcw zgVtPAqB;A`+ldArTT0S@3|?GRK$DO=4i#PekafGrCCQB$IECr&Q+AKJ_dx!nI< zh1yu`|D4PHXEvjDLan>X9@Lc7wvs7774WJ}t+zsLtQypg2{o}5J~y^1B?a0ll$qB*wulxRlmN7R-Tno{h?$dqFLL8kn|!AgbNST)4H z=qxe9O-q4~6z>rglGOPAK~i`rF#0SnlBTk)o#lmZ{!(B^Ep(DShwW(nQs7gfZ(ItT zezq4$uP+5go=rxsFlA;j=H#8-aLQgY_A$ZY>$+CkJUL*Duce_paJCn1%^73OIlZ~7 zfqX3uB%fI!e_s9=zYy1{2fmi3@?LX0w3%llj>^>If8Ptjzgl>u&5S{nG?oD^yol!g z7%5t--_yWRQ-RM~c-?nm+}4tzgP7%t$%QJxT-CA$=B4MbqUptlUN!nQR=sKRpGlt; zIOZJn;v==NXlcibKO!-=T}R}M)vq`>=1hOxurX%+v##aDlt0B}9TL2#jwyeN9dZ(E zLm3=)2&u6C)Tj~(W#;fL`ngdIN~H;2@Jz#TDf$VY#*D#nG2pL!V&_^S=ew^v*Nd2h zeL&(E%#(xUcbsdRBW|;iv!kXo2Zt15ju+2Wb2L_{ja9=O-;>zM&7nh7+ZbtY8fL~Q z^P{Y`qy6K6`65$ND-KUp$9qM`xaXNCn#_;lAl3J6E3etw;NHk#*Io&?J`P(9?Pda3 zr1qbsDfK#oXJt8oJD;x)$ONwY8%Cl>G$k4Xo@)ZvpUAF8lym_nNO~QPtx@F`L&s>g zDN4Q-u_-ETdMq}d=ap|oY>MJaU~FsCZ*0=z`vKTL&46~3)Yz=nUUL~(`aToauA=C? zT%xQs+q8Q$TP=-y^lA%K^p-7{C0A0ED8V!myjOI{}PToQ-<{%e9+`FS~kk=^qPhUfab4er;|5w`SQ z!_HLEA5n6s^^YhHdZ#4X4et;0>O@Qm83q0R(L-{^`|+oUzcG=MVFEbSxLeL@N43#` z>D_Xs`>8gxVe|(}fwKLk<|fh52bTiBwece9yQRQmNnS*EdN&?b`+*mErjOyL|GV+{ zBFT%SYu=4VgfEr2RkNLrm0^b(TcOelw_=0Jg<8j-= z$zF&iC89~g3sjR5!CZWS3Yl=g=rNCaP{_O6s*v^DF^K%O0J!WuBSCLX=@opjKn}}h z5KWGpW+*9w^vG%SA4`D}&3}q4S}y}8xAUS^`(?mm{Bh|r;0qx_e8hq$i4vltJ51PWRypdL zFj20eY$Xqdi8DHAn7E_^<&EmP3{IioM@BenMN-Bx;D!!fB;_mvCUo#3`SxX;BaGvJ zm(gVcQ*$ZRrc(Z~O^r(4>$BQs<(dIS7`WHR(e$Ss>>9Y&M;o#RQeoZed$wPFG3{y} z-+G#QA;%TDc)-UU{FQuS09q|*M=$i6p<|V*dJbQN=g=|!A{{zbs9r}j<(SYJI=*-j z>xRSZnC(X3Bu$Ct3>|k9&Ga?B6~0nas@Jv-+3j*K zs(lrQw~keMqHlsdyjKy97pmb*K?4 zYJ-|_oEz;Q6py&WM)AA2dMk}1O>>H<9UDhBk$t=ONa2*0Z_rBRN z7T0v~nrRF^7vc|e(P>Oj>uw(_g(I7|uJj%6fH;&X=u7T^oIfBAHxlGhyhfAG%#HBK zPQ*VrzaU^=4(Z5LvEMD*G0FxU^RhkP*_Er@{@eLXqrBPGHp(&YZRc}!ySlM^+1O9- z=0$X)fYyl7Rg7+6arU-R+6kpo7S1eJ*m zrXKNc9uMp%zJEM$c~37|1;zvUJ-vt?J03_m0$#F?4m8KA@xc6^9M?QO9%!tjnrr%; z+=&Dv?H><3OSHeN`;cAbj*UuxF!C9uu?RDvNFEWdV#TM4#S+O7Z7|t5y&-3E}shcOec5rB+L@~)^Fzt3WM*b=WbjZtU55@o zuAdh%Dc@-M=S1~w{k&+{@hfB0HJT>Mi<|4kYO(!wlDJ;*Z|F~VLQDq91o?&%j@8+A zc`WJ-;Bdkp4&u0!8_4Gfo8Px}7${!lGyBf(gwC&g%*w-ebq4o}#qdF1H0=Goks?=9 zelGrAS?U5B+L^^J;Qb1W34Y5r`0-%f`)0=Co$y_0#b-1Xv1z0Ql)8p=MGP!>1zI&aOj1J_unnu&eL%O;<3)aX?6kXI(J ztLo<;nVB`Nn~9e;Uc!$yKDA$8S8oJaUzd@w@H(!Jn*vy0ml5_kW-)?by~NUQD5cl6 ztFKFAS)+3TBe~VCE3`JBk{bM8t1FYE*Ca^9UC%=ZoQLAIx}26=c)i_Oye69@g-{J)~F3vOC79N!^_M! zD?it%0sLtMGc{2bC4;82C=QxB593C;8IYh%c4Z2f6^ihYqlV{)@f4ZAuzzL{&qr~b ze2h8yx!KHFFFC0dEJ)HHE4HVC6xH*@HU4Z`Lf z;+i#4t&LPJL~`cP{T9|C2i@fgx{x3ay64`agYIYegM;o*hz$2`$tf)4UKh2ERO%SZ ztc&7DNQ>YisH}_rN>-e6=zm>DuBHaNY~SZGr(!+GF5*Hp=FQzg5l zO44q%+s5uF4j~KqoFB>=sU)yFiqY}OTkR%do?#D94@7OfDi&YdsuQskmH#%{jOJ{f zL345pBIPRW2wRbVXR-}Zk+I6PTvJA{QMOVei1t@X8Jp_d*O5NW>(=6! zshwjY-}Fqoc8-aBA7*mJ+4Nh#i9RictH4z%@OWF`k;84vC;Znq)dX9}`w4Oh=?{!a z=j&EI3j$e=IHfhjnq5zCVVYg<=h?A}w;AQpDqYV!B{aF!ncBW^3$TpfSRcRf!(U#YfD6_mCqs%MAy=W_w0W7b_*}yN%$jXDY zydKB@|IYGi*R@=>QEJLigK;tspN}2ar64lO;aF1F&H1r<7D?l=dgsrH#SoH~3tk_S zYsL_h{$C)m0ITc$*rOrFWO)%&B02hI5S|dcWm#S%2K6l=tdYDOS#0pWx^)wIG*GVYu$G{iqk)^Vy+~R<8d#C- zMXOb#fn7vK)Jx0Zs9Wl&o%HQ;Ak}XNgbFoc)X&S|U?oJ^MoLP$7r4Q}{y=oAX>xZU zD~JpX$q1)Ub}lz?0uHGnha}bi*wpyPQsYbgT=Zql0Uy_6m%q`^!=_qdsHQR_{6xh4 z!B6A`6{p!cbq8>$p4B!29I6-f|7d&j_^7J$fBczDglvpq3s|>E zX-F(!ShTJ!5Q0!yB7v&)W63acL#CO`40k33)N1MmZV(ZXB|unJ7LB`B(YVl})fzWy zT`F1^RM1d&QTe^z&pGGj-aB_TfBo_YlY7s3-p~7a&U2n~&Rxzu*+Ua+y9w-L0yozx z|LkLhnBtX_Iw%SFgIbifVro zoM_sb0KAR{PyPuT* zK;|7!daWU<`9T0)3Gq|b5Iq%k)1cq1Av%OVe!GU~xQLr_5U>UPg|pk_FI@g1g7?H$ zqOlP-aTNO_ZsI6z`N0&$7qt>CL=}Yd)(HAu!mzTp_HeOi9%-~T4G=`IXihLojl=M3 zwVICS$wwICZQU?DmFtkcmNWiD#)@OgKSa6;JRPp{Y>Y1}c*9)4E&^&2_m5 z&AxNH^Xj@>q^gVNs=T@`mv8cl;Z`1S4%;o+7E4^9wjm4>xxmWH>;Oe=%u$M>a6p%I zz&&4RS31{#)+_qedGJlHVnY|zRnY{-YazoUOEomRjz4Wv^+7-p*5$^=n z{3zo7c6g9`LW?}J6~?0{{1BBrVRXz*+!M}=$)3h zqpc@&pbC1ze`4@!!o24Xc!P!q+m98>#Byi1@Z@gxWRFH%_=(LY+MSCFL1j2Lwqn?w ziw8L_$2oq(T%LHfQ8{~eF7|Ig%#9$|?p%Hxq@W94^wG0i)^DC$%rlEa+4l_x59})^b->Hys>31*zD1j3T~{?9?P1O-ajMN7 zNolV+Ua=4zsYT*`2V7*5Rr6&>+Q|yc9jRNv94%H6n?E6}WB&mIv$iznC(;?ice5ngq=h#Zm}$&!_L?OY3Hr1 z#@Wu8zF@oiTq0#24hi}H{u58_bMbsQs5&$!r4Lox3h8^b9K z+1oEx4ZkSQluYFTyQN$b&joR~rOzda-O|n1>=Q+q1}o~qClLvq2Wa@@qD!Ptnpual zPvq)Xfpz{ufx;hagg@4>KQ>;1bs*82JPakT zWy5_z%^RAaheNTc;n;cJ0QX*HUeyq@nIGo4hR!AWFz>a+9y<3=ZW_kIn67=ue5s(& z<-%IvBlbiqbmz`oLsaDwaToyO%hnJ*&PABfw#lLf6g}BWly_O$jKzfw&2R8*DK+!` zT=4yz`Odt|O$B=ey)T*Gd?;di#aV)0BRGKGYAyo3h%8!wBG7vbMHYve){(x(Xe``- z^e|&`wZW#X%*<+0KF>53+T#NG2;Y8ltR6DzaTCnMtn@q*pNM&p=3A=~$7E zD;*`xu%dxzMLw={T-2=k_lkU^1ex~a0rv*i3g&vspZSLBTw0lrtAH2cP`snPDj%lDD(E_kjxFLA|%L_(8p1R}B?jXp_Gs-Aj}Q^O3s_;X&wvwk#LkUEG1bgs}X#$ zAW_XVm>(@%LsXb8iq;%KRJLsmZtmvZzlLaxJii$took4$WDzVkT4d3#BjooO&mSg? zi^$(xa{L~nDZXWp$m^jhM8=d095L}}&zHC2+VlC{hPSVA)8Hn$hg+*KYaw2TpnEuw z2L589J7GYo6#{ybp`v+|S=9&m+?76Hu!#%Bdh!ESbooMF7XFZrW#I!0-PEh!2qF5M z{4EL5!F)_Q-v9|ttR2!rAHjJ@tR;)!Z!F^qj}ZPojH2u{L=hBOW+Em(nYo+CBt^e? zwBVwa-D4#B$3VxJblW01COyk)%=z1-CK~hVG57~PcS2Z~Q+tJ8bF$Kc-hXugJL~EK z@U?DYDjos9rl8^Pn}xq`W?NgX6aK!LZT;JINK#^orTM9~0GZ@RoQ(Krbpdkb z3D@J2du%@4Qh;&c0vsZ7tS&%~|LgUM`;A}ofmpOk)3Lpw!K&*_GrO&V-BxCI+6`_J z?<{RCz$z_r18#@W-390#%{OAn*29HMfsYoTJN+GJT5zU`9%qBP#OEE*dL2DPM>z@p zMC+B}(mY@6I`eeFV6yy$m(J{gODDM4MLmV!ulWqd-g0^7H$8~>x|wc3-EfA0YL&l$ zngGE>VPf-&IWm-ETFw%fs#8X!`f^x;$hc*(jEslC!x9-0^ZjD4-_OIp#48-9aKZAe z0@qMpArrmDE&aw~44zmUzQs-5X9=bN-Uba7F#gZL2{9_tTLl;vmoHHR>{}cTy!7+Y zzJmF~deQSs+*Beq-}$KEuFd(B-GWf6okMRIoU^1CO~YY{FG^SUy6EQ~T8P6~5T8*9 z(T4^2%I#w~6$yF+Ap{bAP=Gncmo0AURs0)a%2o2W#7`d=w5-pkbMYQcuYO}hai{#{ zi2ar+UVD>y5FR;#4i)^c#6zxD%Ki5mqBiO4S@89tHAMZFVtv|cT0aW@T7VWvop9nD zRQvg~estb){mAgbg4ea-Dqej84pVL#_~9r*8?-ITlU^nSEg8XJWb zz0c@JRmON0v$}WkuQR?`j5jOSxRKzfpeDy9V)ZuGeamv(3cSdR=KI5E#HM(eE;l)TYNikW-re0u6(cc z`teE+y}yca3Jaul@3QJ+SM%91I2-ZOK2|v&CkrAt2hy4kk4n>@J(!h$wi@|E-2J`Y zg9-SMR*YW(KVH-@zOu2{gZ+Geiz4xCWs9fumma!;PrySrc`zS*7{_&foj&V@_kFu? z2$l3HoA&7%C=;8)V+w!02jR1BEuOvTCG9xKb_`sH8Fd6FBVPJ}&FJ8B9XJ=Rr~5q^ zSJ$>7lCaC({hqGV@@P*R+h6FBXpegkZ===&tqmum^t1<=QD2W0BhG{%y~}g?IUc%Y zy?8%)7h^w!D#b^-_?-&+s|U}Lyn%ByIH%FyJkwe{bm|6ZB+A7TmYqk6j=dHg8&6n% zy+J-yMsW6bu|+C zNg%_wFL2Qj>xf!p(X^gKr6;W;x)2{n%KgWHfz)9x`e497TDlc)+zuE>FKuC3Hd%IP<7Aa$K9oOIQ6@q026or0T`!v_weVcXnPddk3oq~VV-0|(N_+Yp$!r>-OF zl3J=jz_*U*MBKv9!512LORzE30u(OgYUP&`dAGZ%^wi6V)^q(G9~M#Z`J&#L@)z|_ z1}r`a^RMl0D*Nk)MMP!zz6+J&t1tt1xTzFhg=xgY0eszgh7`U*{sQE5fb>79kaif= z9+ky!%3oAFwI`AIV(RV-1(S#zs8D(s&c_#7^X_z0*`9iMXKy{xxI5kAyR6UIgmBTs z_%3VlU2ZBX3PPwjNOb;PaA3kYCoV&TLm_#8Y?Luzj=4ziN^*b9COguQX*0>D7^r!c zGs=)?uTfgzhh-kuV0#L`g;RLWDeZ0&+Y>C~lwEiMPROK;HvyLOxG=B0MkvhfxkOTJFIZSUl)%H{}M`5f%3peXIyAdi^@0)pxt8^yYO$FXL_L z-1Y$jsYcd026dhuFp!4b;}$)lrLT)4`%E+gdc>pmxTy?1f(CszU?81yuk070du6}K znI%kLA%89XV#~d zk9zT!dp_c!a`}C38XWP`$2`>KJb#~CMEMx0Y@hsv;KwS;hum+8@;}@!qkO!qtb{C4 z{t$$^qx`e`*&PWM;I>E|VVKZtnBt-yt?S4*K|W<;>67b-COzP$(q~crfSbx*T89^_ zU&sF@*O}H==jnl%rVq#5V}c%TFwhG93Y8t;;rlQjd)BvjXxjsLo{i{Z&*;1I>D>q1 zyyM!(o~y>?k$fiWW1a<2F5XJ-kdar9$itiH*E}fSL@#*AO{KxveW)FO;Fa{3AA)bp zZH$1|)FWhN!*TCZQES)C6mL9IJi~S4k$8ry&%cjJ**gOj+!EDyx?<<&8r79WX8 z%%q1gsqyM_wAfg`U#BH}_8<88U=l8|V00orS)+K2sWtOymweip&zQd`KJAQ6C3SQ+ z|KG_I-g7kL!vhMud5&30$C0j~^+X+P<4!|I2P0ES-3WS;|Ezt)qNLQi!bam8 zzuaJPj(p)X4{d$~H>Yrf2jD+@M6~=5NAkR1Ts8mUNX-8z_kE0lvgHq> z{N@5Ty1zs$6?xA!evl zGMkq0U&TA+W0mFoM1u6>5b;%%ik+B6;1*9r78T~mTRbR09#W;yQ5ZJinGX62Q zc*0}OV=4SUU2BYLegm;RXf|?yy@~} zA#S04jpHb;=Z{+u)1nyUVp@pk$wCZLW1mzJ@gzsY^-rn@7RPuFXD&XV^Di!b|4BT# z20!|+@P3(3Ui*~tZF@5gS!Ysy-<)<;Ww;}XJ^Pa`C>Xl!CQT#rl-1>>~Zu&zzj z=Erb`ZtcBrd>%c5nqEA$N2_`xlYWj99072~cWn0yyWCU|8Q@ZHzuw*(my^EO#djW~ z5!%|j<5&+3_^X>YaN~EAb}`sGoWk=g6T&)9v*P`|Kf1|7PojulGdj@wh@~DH{5N@Z zXH9Qp&1-O|`{=gbzB4`a&fl;d;!!SQiOOjcXn7xy?XerK<#DI{XD*($wW38Z?wq?@ zjysD`iQ|8-tg{Yv&}!Sb2(32eXd%8?wc3BU2)!mEi#pJ2a(~%I)F#h74V)QmL_MBC zU;wB17=iPo!np)RITP!N<}VQKx=8*aqO)`HJZ?SF+GpG(KLwlXzlo?y;&27*m-)mq za=*;s&&vHWn~!l>_sjG@3lOzq_gI&mk;Jv5n%RvW}{2!F&OA??+wRV`r9ptNPt=EI~#ay{`Sw!X|Mc z`^(<)&6F=WDtkVsl7={@l0X#E_!UR{q~~xm&Ks~q`_~-pIn5~J9{Mf!(86bNBI2V% zy)mB|=z1Qvh>mrU$Tx>r-Kp5`DRt&x3q~2t3$HL+@L6w{n1Fo7kTlp%uuBkJ{Jfh4 z!GnU}-B6xzj4-@sTDmF45rcp$P4XiRnU*Wfn3M6++Hvgt(?DmU^LD>T@;rOFF0nzvN(yp*bCjEnS2XZiKQFp(%aWqWjT zdjwy?)Wk>B*$2bdMjVPDq|QDVJNv%ureVAdf~&I+22TDhV|;4_HeuX z_GNKFKNbizY87?#uRgBv9@_J=n|i$`9ZPrlI-Yd8~UwE7g zGad2W>$e&H;#V-$!FlwEgIswRDvR-A?{O|}yfE4r}jch?NFB zN|XD28&SIizZl>rtS35ZkDE$QTuh?$TqL^MYg&BHLNj; zT~rc{n{EHAFDBdjUz4HtI%^*MFBsgoo@l?Mm~*3mhe@x)qz&td^w-foZqu9l&Lr)Z zGUSMcO#{(GcsLwkSg?1AP$e30bzeCguI`HoB#E(69w33YvWYo&yzVAkcxA+jI%?}X z?~Z&beCKrxCjs2EZtaUF1CDybO$9aZMYx{Y81u|G%(q!bQ(NC>7v+(w)p{}X|A%Sj zF;`hPrbSulN<3&A&1^v7eNYGNH8_juDKM^Ze^+T-8dP4=_RsBp3 zmj?wc@Ep?4HT}?0mcM~F!S_lg+iw!MB{90CADsO#oZYEJRQm?H_IwyI-lj{1r@#0S8~a(9bngBYL6Vj1xUH=uPE{ z7y97>q_sasLd0Jd9zUlhPEGH{&% z{9;PrNl^o-{AST=qQAW)5MN@%>)%vs@Rzvl?ld&K6sI8+ju=p5ERb*aOE#^tx zoSlQWOI!K)*tgx}kfk^EN0y$6Le2@5x?WL7oBQM8rw4IX*P=eUy??O6Lr3k!g99hX zTm4V>M?bv;=Q?oCPtWylS)Wf`I8!{xMe@0+&i?4IecwSwr(VI;q79ywze99w7VYUT zcNZ2g*L2ljk~Cu0YJH5ttaY~2ulHe z%m$*jz+~_S+ypAe|HMw9!NNZ)Q4JjEKNyw4A$Y6M@V)#6Rjvj;zxt0`-sH_WP)`(* zbyQsot1#;FU9V&*+32I)`FNz!O(Sar7d%3Ur?_e)?H55CPjTG_+Ljz=8A`jsv92Fs z;5DyRxUlZ-{;r|AMs#=ocPl*f?0-~dzPmqqJ!S3C(to_dS zRqkJ01h;XG&|Ep5}cXGZUnGw!>Y z!gc%5qIdLnQ=9ZcIlS=K{%+c_&rO5=(cev-sAKYi&l?Jd0zPk$Z-**!ZJW%7Kcq-L z#3WDrNRfPqS&;ZV@56lH{m4zZZ$Ocxa2hCl(cg{lOVOb3`@88^AA=@70!}3&3eG|u zeBk@WkKM!{0sjznOf4Lj}6Ybq8cO*}*Ez-NdGXCnN@7$5z2tO#BW= z*B@ZlcOFoIa)6zN3|mg>fL8!tb3xBfr8jAa2#649!J894#c%}&J$;8zL{cFfG#}OA zpq5XigZ_s4mNeqT-s~Y|K+QIRuRJtHQVBx`JaqDB(nIHCr+tfu=3E?zOy~Kb3`_;q zwnKq}?&1Nip>r-KT0G#STRe2hXCg2c4?tkv^O*?D#REiOBGg=q2Ow!uZr2i`h~#q< z_!P7d9o6NgLH$~Y&c`3ev=A-nLg;aP$;vBG8Gd}9i@4=wQ7ejISI$9c*DrbJ5Od3$ z?UKzd4caSzMYC^`hHYZQrXEzyzDe+>9!-dFDbgi>U9H0|i>9e@YgWo&F6fZj&b zl|h?`CVr0Va+`=Q{~Wm~=gdt+?@2!mzDt;mJzs8uK+Pti2R?UG z4#;*n)H)BfK>PrsnY)>J*DE>AQ*Ld$Tj&wdxVWb)H?WClOquY`670!x`zE5nU$|+| zJ)4Lw{K9R4o~kv;%NyqfVs*Hrq|J}9j?o5ohoJ+q&#C>jp6iM25)r*4NPY9e5)XB; zSY%{HFrcY7{jGD=Q3i0~y~>C&MASacHPjcyZ{fKfTKa{WY_yL_(3jclSQLYgj{g_mogXJi8}FiOwhdz_~0O} zpia?ZUOr|=jWqa|TkP06z%{h47QK8)PY=Ba*zVpU*22T1VLhhd5vbpJgN<60@kSUp zT?XC}M0BsqHPnlr#};@f>nk@6ovn^`e5IUnul(}US05J9rBZy6{ACbCS>C=P`uA6E zD($nch>rc*O*u#JE23suO;z7ZM1`pS>wQHu=4&^VPS{sOmw$~QFj-P6+L`^Kmkkew z=kNvQ;oSODwI)#@EmIcJedZRw$Kx8t3Gp6&+};7*^!8pzYwC_EXDtf7FkLOaU^t{r z)2Ddr-ed9gfHyTM*GsAFj5mq=JJ- z^c4fR#!12N|KOqLl>!`erKg};i}Cga3be%ud||zZ`u^Z1W{-nYe!%UlWbC8b9R4nW z!Yh8r!7D#o?x715E)FhDi7Ol*2+aVfKv%zE&ls>RPVmh&9$Kg1aqwiC@UmTn3I$#M zu!9AsFZa;xk2qM8d#i_jP>OIc`f#d7!KgMv_ZnM98^DE+Iw-(? z^7V&hi{M~GN(Go3YU&Z^bF>8paF>&U#F4|h5Rz2o`e)_E*IskO!I~)vlIT;X_QqK;L zZSpP<3KZ~OCk21F)k7D$vPDY5!DZY6Bn(S;U zY4S$Ld4t}NPxI4QjDU~HvsGR_7KjDH*db>e3Pv4~O$CJ~;vt5>oVwU#Ekupbx$}r@ zDhX;Kd~AFIPUQ8-rjkftUN}Y<;CLpE>jI$|U9jtK`SctP8v=gPQJB>;8}A?Ln%9rf z+fb=P<>@F4YPB&#ncg#-3OYdo2n|5k*%Lc@&ICC@6k2%>iXu@WqQ*yIu(6r4&$L0aZXliBu|FE%DBX>S;ZtR6F z)!OCU+N4^W8`~utd!eh>%q(ij6Ro|FC)X45vw7+*j#_Xe{4USDjx*fg3tdOg%%UIi zvuRjJYZ8@QLX3ua~tp%nhUZi-cglv8@<8*p|_Nm%3m^j zYe!n%+Tl9mcUkmMf$-K2HX+nYcuO2Tgd^byc_zz~O~P9{T$6s6MZfS^yd}iok-Hc! z;$e@7gHDg#Bi`VoP*l^=gQ~;&|1xDLe?(q0`{k*${PL8m`n)VEKT`PRDK_LT9L*+q z^Z|}IW}b3+&?3$wAXE+^MhG9&`Do8@K{W>dl5U%KFuXZXy0I8!{7&%DlO%6ezx z860uyKYVH)PIZTmuPz*pYE?)V|HIf(zNvZL?4zr*R8JepJuOT2v{BsC+_I-#oz--H z7JbxP_~_~^B$+9FMC^#82#(lCS7%*)eiq%_C!2j9T>*_3|JSqv&^^EgY z9}b!&d_aA(;ky<-xCjUJbQ2%!?VC+QqD|gNlos;|%DrGSQS&g7-v;23uE#bLo!Bp% zN}t(Gw76e3mA$wbOS`u=6W!epyRt&0U5X4;B8Qa7tp3@=m6!KdmAho+qg3U+r~~}* zMY8fGMdl>gp!oyd|7C_$?G<^;+-~<8?KXnjt-BR59A?ROyEp5k-)GUTBGGR5vfnNq zAlgkF6%9lWCwbK4sB9W4oJ{v-{psaAs=^s=*L$zn~BPYi&hv7 zFI8?K`rxQ+Dy`Z=H1}uORC@jvqDSyY%@(2~j?Sji;1;5)qv2$T*GTb`mH2AZfq2O= zQhYA{fcPW$1L8f7RpRYZ{1hdA?y*w)oA?9Tha8tpEUx1Zi0?ek+-kFR&0AmPjp4Ix z(f?(qla1E)wz=`T(n?HSuKgEgQO@zA@wzxuT#cjIB#$1&5jRek>$?lH==0;d=N$*R zeo=8Y70^j8DrpG&8-tBe`oeYOZ}VtsaW)Ony&-?Nfj)OV_kKQ6K?KzT{`vG3i(UiK zy*L++)oD6?&1b(YhTACj-0C7~E*Gu-1^QU9x`>WB0ZnpobrH48qOWo9-53|^G;id8 zocWbw>0PsfOH_AA);UThUz9~NP7v-bK}uVLgV`hwUcmu7x8(O1Wzp9sWXsJxXb9>K zACyf)f?+Df0hQDRs0@E;gD?V=W3xCKiW6%FA>9{`;4>p|W*^QNK8XfuW8=e3|I3V_ zd{O+q*(d#!Pm)ck`qlg)izfaYv*a*QKhRx*qq=}Rx(7$R$B8`g5l*n1`i1_GMa6@& zsYDNlW7HqQ=MT=Nf*Kg_s|y7E#CMTCADnF(?E?Rg<^;<6&JWFA$Vu{n2C;mvh<<{X67;L%yhozo* z$z3+QnwLff-R6CYwtT@BYvL=1sO&3Wu(e?F#T&kuGGjqRGY}P0z|op*%>TK>ADpw_6qLUtxR6dtzY#E}U`%w~Z zE^$qP8%?=OX)WI0t)1OWbtwdBvZFWSk{p)bWHXTG$BPq0YE9xFb{b;zQk)$K#je|+! zjU=dq+YWXc8zowG^k*wPRB)1_w(4lU2cO$@2~mf+OZU5!=%kafY0xp364m05UtCJG z`J`;8d?M8w%9YZ{e=W-Q{|vV)b{op~|BRLA?7D=gQ}P_FM0oFS{Iewx5rQbdjPrCEUymkPPXuA#u|p{5baohdYMT$=G!ASu}9FDwS^U;|be z@CgISjA^uj0xSH|1`5v4k5rt~1T_Vy0xY%y?_uCBe50nj&7@O9;qd`)u=`eNRX7|h z8-BJH>po?`wVPlgLnHkIBRA=)&`8spq(PW*>Y@#^*CHLu@bT>MqH zrz1p<^lZM+L(hS!&xb?s`V8=C`o~b9fLrVe;F><8v&p33Uvl&8$LE_ zAdUIDo3cjK3%7#K2VjghXCdnK$I3jqWVCW_FI<#$NaYrMR%1YSSXI223mtzk5Z@EC z3X@OR<5lXmq8fP=O!+qk<6+#%KdbRj)=fEIu_zt@o+%P@<6y=5#Y7&?t zhj@h@Ib<y94(uheLZt6)$YB5GPeL{|ra~>WqVHk!!>7bIru8PcYDp#D zVujgOaD0=82K_2c7t5BksxX>SVHf^jam2O)++IJFt`2TSs|Mw11>9j(u!D^_?o?$& z!Zx&`8of}#RsCq;y@Po_lsT-a^%P`{6?zXtzj11O>&Y3Iby}pPyR0bN3VgFX^rJyA zMgNeEXVGA^oq~=Ww)kLM!DoN)(74l;Yg4G;wzDWuO{Z|j&|ooKw{j-g!6s-1e!`}+ zfEq2Q&^qm?U_JW`#X4P7Q^P65y;kHsjC}PO@ePNk_sp|5)$tTS#APvjTfql@58ZB1 zOyOABc6KF3^C|Q$yOx=L4^^M39Fal`H=kXHG-O;b?5m%kd&L7tN>6)Mw^g-7pZ0-O z^9+5T@+%cUCs#9@9uvJW4!JiFf0jrX+So%4#bmMP-cJMhPys z{4C`HGo`9grmIDuv_Vm7PDCl0i=vewzvedxM65(UXW#5Hh=eJDga!a{iuQ@-anmh_RZZ=3KGPN5wSOxd@zHXdioi3`W;S^$v6?qRM?-*xkI7u-SuJHQmG=D`nBp#|# zZ^tW)!W@I8c)pH$4xNEjv_mEch1KR^nqDSfKT%N4cE~%cnz|wHtP+px zI;?ihka%OqKq}6)k!)9pkr71gRd`27uI}2aaHj$XBgob6pi3}UoE375T|;zzH|1zN zV}-&rm2i2UjixR^z*0AnRTZ zobc%>AKDb=C^gObHgjCN26~8a5#u}<(}`FPk?TC+aegW4yl!o!C#KX_CKM_~g#~tn zcr>(I;i2RTQ>rr@suXp^DHIkq5k2FAKT9g`s1LcGaS4yIjnB9;G*Y#rQdiZ>*-k_4 z)TC;SX9+F)$FdUniW^>QHMH#Q)&>n9dpcHnXkG=boRQo-cx8n!^J9<6%#X$9!Iq9{ zdMQ`w<6lEbi^nhx%HCmRPe56O=p7Hj<~^zG9git)-Vw@jTh|e_n`ctkLixz?*;I-x zluySWudgGzXgnq;W>3_5>%|Mgdf9NiC~PpHBMn9(!QDIHV|49=(lbMtof>ixyZn6) z@;t^M$7()2*u|H+RHjt`g@pzYTvU4@x7R8UwT@Tg!3*MwSI!x8iB?FR&&gk7R^*>c z^mv@@b$E*j_fjo-mA)>k-sU_!Ec>pZw>ghfd``n$qKUFL zMXgfT*4tw4HF>A6^3V~L%G_(D+Ld#Oj*(TX@qB4^>S_<&Xedr^f@oG? z%nucEhh0NxwTE6jTbY(ZL%acUBBx+m`dbXx!3KO`z$cu@SvgBk1y)>S0|n=U6BXw) zK~0lG1z2tc-owECCRrwjY;r}Ix6<+=VObhZA#_-gwiPt`JoKwcied`4$_Z$k7NhkP z{Fnh2Pi!kVFxx|y{5DNB2~3gRR@j9DEjo6vDQgWK3BB#WULq^RHajxduRcey&ji`j zmJ*bya$}ArKWW5~)?10vL0Y#R1z2M)*DSjzKfMwzMlTno^;6 z*tM+kdFZSu${8uNa8ufKD1R&Hwx3xH*};a?7%CF_+m35o_EUi6N85nGJv>!$&j8re zjtX>)6?_kaFP&;>N7-Sp>lkl9RW`iBkZU)|$A}i5x?;75?l;ISJT-4iK2eVUQldIp zH7I{o@-~m*>}f;M@abAqiy1nKkLh+qjA)zd%{d-=ZE7|R_2URzOC2&aZxhdcTP9Oh z^fC#rU|a0)Fy**uX85twl-J|ot>D9}W8sJaS$M34`CJh98(4GsCMEA+wAVDlOwOJ0 z6>ggiAIujz;TMiJT?@Ree;?X9%{pkxwzj>CR3<0WN>9yk-S$MaCZadaz&l9K!&wm= zVLIIf0wHp}afWF+{l*z`iDFtSu;;{cMV1MI(o%DLw>FKaV-nuknp&lrqGOVnCq8*C zQJW;zPl?ERl9MlWZzviG>oK+88W#C#^{rHPDx6l{%k$mIkFWO7TtoG{lM`5z0Gt=) z3bUrT+pO6{+ZlUVRW>gxw@-!(AB3h5xwcO>xo~?Tr44~-G!U9&u|w+Ul)tW4xz@O^ z>c6Xviz)XxT#evHu^*`iJRF4tusye#WL+`>(taw)yp0>aRhH`T4s_Hma*f;_Q37RUYV(gKubk+r@Qfqkae{{x{w3~NC=&N8gF!BE)obHlRSrY%Hc%voa77NXF(_ADXZ2jnHT$_{e= zp}?948=zQlFLnK?)k6=#m?+Uw7y9fQFeOB;rTpwViHjIZU2I6MuZ1XLCbG1J$UQxq zN>{ZI{Stp{Xd#+9-AYYvu82;6faQcDpYTwO=obbv5xcIAOGTWNlZ!yinl79m)f!r1 zMdC0wv?jyUbUmztAl!#v1A$5mc0+Zd!7sFW=*H>UoD~KmDYQ=)*F*-Ja>8KtnPsrU z4o0}%p!qS-dHn`(he^TgAkmVX&s#m@sTL}hY!EK1HYtdQsNRHSpGWV*LM4e`;_pK!I&&;OMM^_T<#~)9x zBs%hUcB-=*YXjj4!NzEvCB_Yckx~=UCZhd0RG?WLmxsDzkmHwPmbT5uU8SRkQ|?rzs`eU za^*8aO}aN?Xl=G??_}+H=cyFfDIQ-l8K-Mr`=;6QqP&7CJkc;n_Qtt48$Bn;5of|H zJ<$PCjoJ#TISvwAaLZ944$=WoPInYe`7a$QVyfduZy5#$gvjMLbrye`%sIRa4IOcs zn7N@zPe)TI9IFe2I622Oap_HFKm|W$sNJBiTMVB&WIzW*6_*OCJr1&MO~#Ni+7wpt zFoWjMP?JW~*%zQ|-~^6#wOP4&HW>>>CwfEvXq~s7mn-(&mlbGRH@Jwq?E{|cW~_Gm zN)LI4X6%(LQ45C~)CKxBGx|2(k;@v_iFsqu6>f|fn9X+Vt!8Ys7n+%Q9bbnHVmB%A z6?Z>0TZVVcg0Z@aI?Y!Ptx8)}xuKP^F+QB+jfEQmzTu@Oj^k%< zCx^X$2&jsYZUGSr;u{}1%fB2zGh|tf{59y``{e+dWN;i_dgAziH;4fkzqq2J&5uEe z0xut7uo~*aC*xmt-@Y}Uf)`}d(2y2u3hVVmue*_Px8WQf@YSMGb?)N}R9<}DjSq4n z0gWKn>+bl}5a*}bu;AIgV;f_2S}4YszM)NNX&Y&C;v?+D z$-l?6wu$HwcjQVBU5`U=W2_D*wh!>ov%go)d&I$CrrSMb*nukUYiMAgG}9=9fyHB{ znQ=Je)2g)?K6q87hhyRKTCKM+7~`Iy#A{424tp*+y2TKp(cK(~5^I z6UARRq|-*{nRVDjrdN+ONlY1kG7FfLCl{!mIF<#Hkb;KF^Q)#!pCJ@YuNXaRbh)9* zS>yQ1YKu&68Zx6GR5`xVBAAz0aQwts zgD&2`o|N*Al#fhL%#m_oGTF*B+}t}gfFtQXY05dW?wedVwWs@~25@BEKTSDD)Q$mB&5i%@OQol$c zWeYP7meJ*gJ4-SUpWZzjs@52ikvNa{O0+ZsQPnc0o>`WGxQV!O+ld*8GjS#Q%fzBp z)#r##f*F@#>CnUi<0n=o5E_R@X3kg>>jp@o|05Fsn%a56SaBR zuJ+XAvPt!xmQ2>>VY}Yb6U!PpRreZs@+rLPI3o=iTih$rGt(3`(Nv;mB^I@}Ddq;U z!zt-sCzUo#O+foM$)s&9>!7^DwAyr91#$dWjGa-+jr&4G$y` z|0t=rsSOgyf1E--mHtnX%A5QT{SJL#e(UXj%4J zhI9q(?R;YrLCf^fZsg*01sx4tlCHFqu`QVh+YP=ci7+B6VXH1pSJ2Vmo70tcGI&`g z!ghmiNg>QRyt=Y-yqZFAKrK%qWlkV*g106Ygo~Xuu1G3qN*Hk}SEi6mtaDXTNmGJ| z)44i{q@ho><=GWe(8blE4=&yo3I2N#xPrQt98GiM)gU9ZBTlblcW}ccv=~y;4@? zn!D1GwJ(<2GZ9jn?oJ})*us*j>R9(=BJW`7z3Ix^t-UW(L1pp%Ndz5Z%4YEcnaDd> z{9wBBc8ed%R8U#`a0)@rT*gEk+0$n7Uy=ZrrzFWR9!ZV?&!^<^Xi^N*Ae4;A&J-YN zF?lR0h-pAb#^muNAn`0rp`jb zz}2Y$D&Yo4-(8cDsNH!BGm>`n-lB{koZNS9#t3%*U6%@>yUVGei2Z(j`f_&9-H^Vh zquXvwU*5@Qi!&3qJ8VfRanqGui^p0r5VX7OrVL~qeYP|M04Jy2oT0qkYs*r}o3gUS zYqw+|X!qLk3}hX>c54OzPF`D)p}gH|E0f7rPp`r-ASO(X^jD=LZj7jo5UbOHz>{M< zZF7WZO$UO-om{vk9cjY`Cxo@>AQ&b%A*@RW!LY#Ih;8XeTP(06tWO8QVu2lDLox(R z6D*^o%?E0txiMo96AiTt*^~+dbkuO2fXL=dF&uboNrhob;|XZomLY@#i`!E{SOddh zRdz>a2o4;!rb2Lx4cn!)Z5e}ri3Cxxu{{-tQ*1~K6{I^d#c<$pXDSSb*s#&KD?V%lYR8PHv?J6f#bdm0Bl{9r>OU5C=dJ1Ym)~uly~Tz z4`v8qqM$m(Lm5C=W=;?P1VPn{{_+!K72!vc$##!()2)x5D9w+iFYVy0oynxl7ef-M ze=H*a2L_L)07&V8j*I{t7(9_2z;q!dkrSTG48eiJQ>hSAy5Z@}5F9w{N{(P!a3u1> zUo%5+;PAI(2=Vi-1bt$6<|qzao=JrQLdMK10hMPn2615W_f#O}@ihUF=Q71`;PHGa zjMQQJj|?FkSiF!7BEFy4#`w?00&rmPuM_~L``Za9yq+0?1BW+~AtcV1 zHdnlvIf?_9w^E^iP+Y#WQTcbqAP!94P6gtWFKtBjW{Tm!P33t?>W;G$aR3pQl1fm(mwM3CMxbm#KhEIXsz0|IB@wPIZAp| ze#{uefyv?IAf~yoB}}AmRi0`7bCSmt;9KMFg@NM0<j01`Cugx{r{4^v7P76~ZrAukiPXcmabZsi2METd|qw9VO zjsu_TQ{kl7$8Y!vFb-^POa^1jB9g`O;>-{nI4ntqfERX$Uq*>>> zWRgz$W}V6wwxum6wMn`4$>b8*C}lTfBx}{XF`29pQTBG+l(v-A#=hB{ObRw8)VC!g zNwdz|l1V!0lRmvYZ8@n;%H5Gl4tCo6hLqi!k*rnkwq&v*KbYob_V(PKv9#3A?Ybjl zX-rWoDwN%KW-KqYOTXQfTHfff6%)r+OdLD1d{))8$rp?mRerL0__}>%lEp)u6O%&S zkqinkkRGG0$zY7|P0D0jG6-Q*dO+Kg0l}p7V0I*ffmhOlxib!?Qr9cOer;-)mRR>e zRPmpup@M{ zeHXA=-P@pvFO*HzLR`mgF1vrcH|C|g9iog&1wfQ5*h7pfROx{*+XM)EfN>4uQ{i02 z9$Z{Sg%;Pa#}wBveZYn**#k;c3WVlxsXdaoRDAq{%k2Tg<>t@Qb1i!iaV>O|K&X+w z8^yKaoEr4bfln(#Dr8@^?xD(ESkbXu_j0C-N&@9Qy%`UABr8K}43YpbTVsb+uG{rlg8<^7(_o!yO zF%9-AhYugNhY~)25%31nK$A5^d(H694w_WkT`lX15H%PcK5XwzezB$S6z_a|_rPvX z+!jaCm^Y?j184fuApvmdw1^hsQoA#_bfPy{E7JVru+JOhiuNR6t2kY22*-p%JA1Av z;26mszB1VC9y@w~WshR+U$Sat2!2Jxa4kDES!;@SenmLsYt(fu6cgi2l17r58LCr&KvRvHNB#)F;ve?607eH zM&(nm{LUUo4c__ZHfoI;zS4|22uE#{>$d#EzM)WqcYcM}hfg}=YH`5lmHj7rjSA2gzVoiCIW_DT9r!l`EW2rQ7cp0ooetIP&10^{sRETu#y5E9PYT6O z8O=lvtA8eAyX~B!HH7sAT(P;?+(xrGkjH7YVO`_M3WVl3HYPqOFs`u{!QWkOZVKre zqchL5;tWW3FimpJ-iU%pYzFRfU5=o=nf<aE`9^J9Inz&wqM>>}({$e6s!G7|%T^!n;GKNp&E5f{oPgPbXNqoF`; zt(xzdqndY)|DC2|?t3yffhiAh;aI<4pRpjKO{$tF8q>tnQencm+7sZn~Qx7{&9hjpUX_X;4PDYW(=dB`(SZd%Gho**-OsMEI~VosxX1#g zjwFiZY)Ru%b5aq-)lGqz4_{_ct3wg}@5kX(>0uw1LZLZ~^g&#OO0OPVFfkB{2?X=b zdP3E-b69e>$rMpgp=;il=I7F9;!1^+W%09d#Z$u}v^5qLUUZpBq1rGb{oSD|oS}t8 z5_~SMd^*1DXs+^nT$L)Xj-lLK=^t^GrZfg)0bv?5c_FUG*dRXevtYW0Z~aNrI^$|h znldvg3*U_^WF?WHR#6uY_+-x7ZVFbRe@b|s=C9K925(3U#eyQ4?ue^4A>0@e{=YM> ztXk8i#p*PjO}Q(scxo6QZV1!{B-8e|8Z)D~^e|uOzB{hi_)g7q%^%P;pJ<(XjV%(16a-%0>Jpz zxOynoqA_(z29;LCRjLSwYJ&mU4(3Zu!hL6lLxRStILXOcXim(^VRc-kDqX`@I0F17 zWqj*L1YT=gMM<_Y6br-_%m{}kd-XX&^O{6;XHK7lieZ_g*2dMF8HzSWM5mjg`2$`I zVC;r%0bwmz=j49yz}i}2hK zS2!gQ75zZ7#PR01iZeqIk(0qlc!_y!iK{6-xyOaK#T5#slG+Kh0R zGd!c*5m!7jGDr9NH8{elhu74S5d;4JM%Yjm2iDf&La@QD)<@l z+ql}l^XdU_C^j}24NukPgku43Ob#vI#nqh@iZ<5P27KZJUSZuUJLjRenv=uf$apQL z`IJ|_kE=2@9IKokLj($HKgQP(-?l=MlV5ZSgsm}TzPiypug7c|%QuW46T@;)$jURm zVpOT?)xJ8dL7NhYHh5#cI?8o-oPREVa0=hVa#^a*jB1m;QRDNxl5QHCzQ>HM>H4^!H&kB{4%Y{OyVn_4)Ah=^jo#qYaBRZ= z$KIcS$x&7PUL{vbi>8|c6rn{bI(2Z+;h*pg+Y6{6_(gFWQpZguU<9_5^|MZ0e<>?T^|euZB(4cs%7>scIB)yZ9a${h_sIMgg3V}E@LL0 zWBH23O^U^3byzHx^^yX+RDs=A56smoMu+ZthKriFDA2t=GFWJfU6nd&GE%a95w&PeW5`8@69eAR7`Q_(NN5zOtl@Uqp0UC%p#k?lTT)Vw?y_Ct2IN>BfIl`2U<8i=dq4wN`K8AW z8Dd1ExUxUj(_hclv;qBOo27BbkGpZ(LxV zrGj*rDa6ZdZ%|Ir&Vf;PrB^~Vih<5LrpqjpIpC^5A7}s?gqFat0^i;MyyWV6qgX9> zs)&@N<>>|`RIGLpzO~1M?`QyAvi+(3{Y$D=eu?d_Wf!eQ+o=kCV&34UTI{;3Y!!^P zEsI*T9Sy|D?FH8Yu(IW-McXyp66>rgyT#O^(HEqm)ePorC=<0TwSBeb*P@MWTu8Z{ zas}A{W$DHRI(^$&W2|3b7Y4XzQHwUEaR~(tLyGB4i#D$Te3y)Iq7R<9pBz-LsiSO$ z>#j1(iuiSFR#1!9(Euw;$M!G}@f_Ro43H0t-nh&f1uZR8xIEFKIlC*j98hXKSR~da}(=yiDosQ17c$d=C8V5x}ciOHm7fitGQ7j0 z9gTvHIKN~=lY_xH>B-m?(_c`<0!RUV_%v$KK3?BMy2g6ei^cu->2*I5evA3s`|p~_ABKzhEvw_>{>M|f@&IrSND{6GFE|nE;$5LEpt5}@QGAv z(IzOU&VMIHOpd`?_98l5y0vwiL#~v4B4K+ zBK4|3UDJwGBiXKtT^IU{@szznHhdV_>NhVGplmE$vtkKg>JO*|yV@1PY_AW4QPf2! z#$CJ`sP3&_9aN;O_vdX3LDVt_dtfW;tBPeiOg=lqpifC?^F0ZT3c*MkzNOMqdt)^ak^z;iJ%Wl2`$ACCR{y?L_@-& zVlV`ID?PmvhlEA+#E|`_fN~gi>I0u_xHi7A%{C}KL3ZCiO-KwM!7QGe6C@b1k)4aVAf zEDV9(YF3maHq}SWGh!Z2V?N4u)ujfji*gA9*oD$R+p;^{@D$Or>%#@YTa28-H3(L$ zbxns_w5*cb>5CHcrT`woSYvS12yRki;1KT`!*{U;(W1?6T%yA@It;5|S|zkT9IM^H z+n%2+u&i-`l~tH8C6jaPgtZHJ4AfsXI>s+NK`*F}8zJkJI?F?uh?;33H@IF{t$gJ* zhKGR!)=?3{jZW7$5v2(KH4C?D`DVqKY@&TEy*$|)@Mzo_UlX+bU}D9aGE-2FmfGoj zr7$^fR~)l!q)YbH+_dSr8HcA2@w#=_D5PDZq<5L*4S1ya1CDW6&RCyw>tN5Vvi;@O zGDCO5Dt4%<#Ag1&yO`twh`CF`!D&IP>D~UUS8bul-Z2 z4_%qU=$`zp`hSSpD^nTl=2suN4F9GNflShN{*`yVyX(cDoxTt4_rXY6dkw2#yOT?H zS+`0#+bvD*pB%(+C&Mvuumncq7F{4{?A~cs-;O;}-w4Tw;Z>F^8UA9ccfK`1qDy2f zy<&Jo|Jc|{X^Zl^=Co3}S1(gx*PK?O5tiXkuDa!1wtd~26>T{TkSIao>K2iy5}mwc zZ%TI~K>h6&l%?7d9k%O}_+AVeMuEPspodAb>CNw4xO_Q@+nYqjBSKYalYZ31$}=Pu zvta4DY( zJ}(YA#hpUp^Zz||4AH@-xKnbD?^2gJt#w#G%Z5c=tg>8o$e+#Y_SNW8#mf$L1p&tr zQQKMjYD5QKyst)A9h;&9uiRIo9ml4KT2RWFDH!s&B;>5)6lC?d0CJ0hJU$8eQUIAb z9z*VMn<3ktx2>XCYRhpiSz^10)CA%i`@y&vPZ5*y#S+meg%6COzT;Cg$u-skoKAr~ zdGh!aNjdXR+)pKO3S0r2$k^%2fP44&6g7{WJeKBuKw>{mT_LHjz*`tL^^Wca=^g>8 z&W9vY!7YCQ#(1{Xbb)~N>HcCs2FwIZ?ASNX%=JPw!1V$x@X!evpoj3oxu(=RKss)uq6yO_E2aDJ;7o&jTDXEgeN@iU?irm zG@*G8ETRFpMP6gj#r%MqL_T=9q_Drn*cWxD=-`}Q^=-f6oG{~1dRAcc5}Tukv+6xO zbP7GEAV=>4C5~nI^rGmRvNx5wRTVFC6+h?(pC`@+1ExdoaN|G854YiAFP+tdZuBWr z6+5d59p_XNtwzmr#!sM2$4TLDQ`aUUdIIgdV*CX9wwa=~wc{rc5i7#3*-{bS0RR-? zWm74_!jn=&%`aavj>aAZ!vu+5f#{92HMWQS*C0)#%&mZ2&cbh<*vE^v^W9 zAIOGpR`R-KbNcpt^sYdgoR3L5q-D9ZNldD`Wui-)U`HnQ1by$)CJ2aR*Sm~=9N&YM zhDY@c{ynFhqS5I&f+qIT<^1I7vXbZJEYDm8J>~j+HFD-i0qqA??%!9V&s0)$;4k;p z=nRH=c3+LQD434{=AC^tdI>Q9+E=6fEZ~{7^8nheU^0MNvGV}3trTUC*?9mx2$*48 zT)1Y^84JWQ_JFupu4@vLdWo)Mu148`P~>lAY_DS;skvpGMw^sSzpt)n+-ackh;bTy z+D_4di^pknrVV9SJx-%r6x;*q3iz3T-#AXAAKNL)o-jbtg$VBbcp$ax#p5;lGKyU}UZcI;6lJd;uh9zJ+%jIH zO>Thbxd72U~ll2L_D9=H_em9Ya@E)!ZCEo~)+m5SN4;;3zfB3hTN#hZHMet@lQ? zC%u}{y@AvQdo_DdeK7lIQ`3*2KIfp2>q$D%!%Y|MJBB9rrf3om4D@gltjQWYU<0Ba zoQwyUpFG@z;rUx3k%89kfnb73OF&vc(v(j`q*aI>Z+d*+G4v=ZmZUw-r0v23Nt%2x zxi3Xb+T%=`)fdQ5^@C>gr_GA(@2e{}ImQ5Voy+>5Rm0~%O^E{f1G+mtnV|o6Z^C_AC0IWBM=e59n9juLa%s-or#hGdl1K{i@$;11aKu+c&Uh zT*UfBz)8Ttyn_>*yxxSKaMwVnB(F0kFW`aXL_Rq6WW~wr%*nGS2b=``A9iF{oD*1( zkN;PS*n%v?4J^p-|0^YJ#gxNqY{gMuQnup7FDYBG;qV$;k^XOGD}MUlkO%F#u|S5f zSqg=a&W6GuBKY@20uwuNin0^|50K*}R;8D(Z%?huSnO5nTZqqkB2Z1+ngAkc6;#`(`#Z=Hm zu0D%fB=DB0t#yH2qhj~UZB3jKi5)m3ej?*6$d;iIUDcv#=d$g*cHp+BW%O}E9hUR4C?1L@{ckBI=|z1^Omc}nHe>53ve9BZP3NC9vCE|cK9C+w z_n(rYmTluF&@QEZX&8$8#!sL*UrtfWFUL=y^jxWRbMfZY@e^p%ms8aCC-|8M{V)C= zIBLQKdJW(oo-lzr6!`k8em(cGiRmbPR6hO_)F%HYjZ!%v!cR-(9sr&f0Qa zM#;38+mO371xgABB_}53U83Aq#*CrgY)H{WUKySO@s2q)MJ>l2tkH(KvTrd>XC183 zpH59t^Mwa%be)peed-E~H-hpX9<0%6{{uQ(?mbweZ3@Goc=zVP8eM~TJkvhzKT_yl zQdgk(jGSo?gFj|hCQ8f?vY9F1a9e`Y&q+={2TnhBT3~NJhv^PYwCMAgD9$@AMWc7j z#T3IZdgxe)A%7YiyqT#9G*{tUrmi6M^Wdm6HG!UGurEF`j@GKzj=BO2v6HO($T<4< zX(?(u`H^wpa5xO|YDmCHM=IWQx@1TPL%kbQG^vY72gNsROwl102^}oaA($!(9djWX zJ@yDe6Rjz9y1={4S5q|6?Wz{(G=YMPz8bJPLCBgup|0Se5U~0NgFVppQJSG%uTWQj zodmGowtbYwoSvfWi)|mJnWv|y?TxmN5@p|R`zU?!bku?<%Op+KvO--^ybQ_I6Ad;B zu_hzmAQZ)h5fX*$vsVZKDO@4YIUhbfXyhxHJLTeE!Lv!4=+I_<{MPAeej@`UKR6>* zxB9Sa1-6E-Vf|C~iqr9;5R>wKL@$b?_ZdTXb^nv=o_1Csgnuf&p$`Az`*mm4XO*K3%Z&bYrsapo$F{^7JX6l4 z&TN9(d=GV4B+hJl{D{_ooipW3YV;0hoJr?$Tz%cyDVo?t=QhFS_n#f;!Ci`h=hPK? zFbYE`YsuPXS+lRMc#Aowe7i|K-KMU1I}T-IqxoP%!geDZ00Q22g**a*8o$xKZaiC! z-}v$SvlHWYYB2UC<*L^$^HyMvSyNo@$_NUF)i|w^F|nDo_S}*CjU7Y#pM&5f(YYfr z@*H)J)Rc2aZd?$G_S})&nd{Y@ZLeE*TU2>a>8{$!my8`lr9|at?P2A0tCnMOP#cX( zu>4{QEdK!Y@bLO#3TBC#|9Y@S?1MZoqPs39kt3DNEf(-ih1f{`9IsDxcCdWQeG*|V1EUT5dx8Pn*jDXny& zXbd_3v!9cIr|kx?V|B;ERcjgNxkG|>PMHUb?)&aAR8siDsqVk23~ZwDh@h=K<73 z(DT%&Rd2sqgktY_f32FYI9+yOAViaHlX)$2QS%_G)9yI(OLvVBbV{N~vmj07p5*cs z$CEuhfSH$o{g5E~V?z$F{k90pIPa4~yK4E+V&*d-U{`{WkJjh2YWX4H*!H+!z~!Je zT!;e7ums>!62fWENtD=E5dEQHlUmn^u*|mi$DwHzjDF@YsC?TYaSjit*PP2yCU3^q8if?&L6LN#=omq~=nZFws9EIQ0mru|d)@j}8n@v(LE9JY$4=3%J>r?mYcXUw zJn1Ui`JNEt*rgJ)G*IyNkIzOnvQflr+T?CW&-c){6RsDu?I$w0F!rcD84M+YAya@5 z2~)M?vhCd*L*gCRd=^<7kC-G55gDK8zP-lmCPyd2R2Z=IikVkfJOhUSg+O}0@v!?U zt`j0t@QIq6$55u1$l{3EFr<*v9qeXjloHilK;`)|L>J&n^GQ~8x15gd0pic^w-6wHhR6ghuP(1LWVn=^IDU%5Wxp5R4;DH0Tq!x&FKq zHLp2LqaBKzc?tu{wHb*kp+J+>b_;8%` zX=0K=Gz}!~z?8m7G!6Y>{Q1F*ej3M5CR(%-)2{tDA&acgPe-CBABA7YL;ftzdEh)& zIudIGeRzSG=}MHod{c@>PdQ4^L{?9j*9CyuwF%iQe>qGe=P1Fv&L*Plrw-TX@UNw) z<>mLgxbft8+CPcL}2cM`IVr`VPlCaCF!uDay_~T%**bili+HBxJ2{sUqo#%YbK( zTiY{qH{Z=RC~Riu5lOFLO+A*I`&cyhahInE&3!DI`@+jpbcjjfShOk~Qq!d3ShN;R zJc)F?Kk|wcHE)`$(JrM*@2D$ekq5uq=W3*10Uq{LVLRD@TNHEZ*aHi=*pc9-%-od0 z&BiNIggm9PHMNzl1%c>J+Z6ci>IwpV5Qy&d(C`VA4&f@J58FjUBDOyo7wr}yCb5sW zg;y*72!e`4TX=Z6Vl!6mG0TGT(-t0TMtvjbW3mXf{C2KJ&O)g;SAotC9HCL}8^ac) zq?5-VqrkKVG>+XKr}c(_)*G1C&wN9w!VOI8y;p)(q8r#px#pYr6;gf!Q~vxn1Ii@= zn1`(b%M;_a1j?=0xD{-mzs=c_qU_m+YxKZXO3UB4Dn$ogakxfb{1)s0m_KomWd1gk z0`u2>D@ED#5gP3P%<$AnN?{o$G-{?6aKGvCxD=iVr0@(&LAyG|AlcEEAkI-oRwJAFA%SULm@>(zlQ$~v=a|m3q zNKU&}N#c_01DI_J<|QD6*#3IGf;r)PP=k?Y9-)P7jC~?tC;RgwG`jnH0m^iC=SYqI6#^e1frr&o6YjeVUj=wKJQ)zvL_{;hBzW~RU|c@_{eZ3+T-8KA z`Kg~~;@8&iV>xY@!?5xQ!@4#-6|b`>CROyowu6eW32%KrMU#w*>0`qzQBiLlL!Z1M zAgstv<-`J&aC6)ZDLGibqIg=Q>H%lG)XLVSLp?oCUGdgJyTQ6#cmtYiSd7Ko)K0@+ z9MBjrN56?Pc`B5a=u`nhAG#sI)Tsho{Sh?}v*P6JSuYgA+Q$AiUh560)rU>MYoe39V6bFFBm0ij!g{wnBe@itvo{^7(XY0qu*g1z#2dW8OwE@( zG>+a>INniL;Cd61YQFuUaYPEmXm!O~M2qF>Zt!$c?Co19J6yF;{j5#;k0HU^p9=^_ zeeXbUMDMUBzOz*>#l6F(bj>!bhmm-P#i&-|-idMdh@xz}x&q&O5a`y2#?eP^f;IfY zOXKK7rBMQlh|IY2tR&$}>WYHM9p7`-9NqQQNg3=QuQ41q3a1wC2qv^_=!e*nK55{yKKar+=zetQHn;-TPj!w-sIvRyaO}3?jpU(5Hs&U zbC^eI^d+TXXQ(SAX`;=)eUwID02r&oCe@&2yn;G>?GB|558wvskiAo>!^%6AIyiSK zb=aXwhwAY4J5fK>;T2Wyo9c@CLv^_QPNfb3)Xo@GOda-rQ8rLoU4aDZAeAL}`iYvi ztSo;5CDd}mQ5ub1CYAPV@c8;s8h!AtJs68ZYns=~Weo43+H8A8QYSm{`5K#<^EH^6 z<#!<)xJY!qhTd1cD^Qs8HB71BrxMprtkyXN-8VuyI$n*_vsu!!ndupScR;{qNx(!8 zUsC$Eh8dxLw^G5YHLSRA#81O@>Dvr?+C3@hFn##<5`9`-(HP%@UF3`7@UGet% z@a+!uloW}0y9IBDXH#865900tlPB(!)JWmqp$YWXJ2cE9N)pb z(jl&a{j?s^9QV@{9a1Inf(Ba>aU$O;PF_%)cs_s1=@n+@&Y!_b zc+0BABE7639Ps?lhU>V;{99?z6!`(!+un$i|DGiOJtkk=m!e5kqW75KqwfoJ@I4K4 z2e?gBNxT;jOsA#rc-MU?xyf^53SH&Df1Z-Q!&)V#6V(+e)Pz4R`}}1bZBTuDB%ZLZ zu~WgltFHV4e}<=W-EtW~_WqE`?QaFlqVcDSNf>@<>Q~+xMW4JsMU#*iV^~B}5p6pT zPkM=_qE}sbf581zutgKynh@=@o>61yZ}+nUJ016j{31mY^HrA)-RIM-^g93QGJTH@ zg_36;NYO+u`hEKYDH?sRRM6P~5D#MSV8?Pn6DzvcLj_>@1RgfOeUwHIsAi)#B}d?A z8WFWT+%c9mEEoLXs|a4c&@q;%?KS*u-qkUdo>L`vs_Pgcn$JTxwZQk?rV1Z}aT^}t zQdG*XcBoeigIA~E)$kgm3&&~J#Fz5n-DdwRMrD${qVDJ#Fcp=9`E_(=YRhPal&?>Lsr`K&3 zeY%q0&wm&V;p?tX-(-Re5g z?loMxhtVAM%M|TBPrV2M9KAsI@%t`DF+2g(TiD_mM;m{M$-yG6MJYi(4Dm%vn9CJC$10#O3;=aQ#QqHC4{-^il#U@Xo}^Jqk&Vy35Tgs)wJY# zaGP2X4oKS)k|-u&fq2|Kdyk=uLV{$3uo=tpTmBjH9&Ky`|LG@GVmy8v^F$&se$T3Pbkkj>SLFIzSm8Xi zb1hgqMUJ#sG-eE)7gUTNKd)PHx9)Z4hNFMUs)H(nO>={){uiq5CK5<%o&IFdG59CZ zsNvLDH7$Sow3T{4r}E(2U;7NJ9M+Z1;L~lqI@XD)3ykN6^e{0LjDDlA)~y@OPlW{;D#N2f9Y>7gbH6D$c#4nD9S#MH!w(QB z2L@@wih|+GA@meiTX{(0jPqpTu{?ZzICPZtlLu;D%@*xxwFeb$;P7PO`MR&N+rUYu zlm~{I;gwgD5PY=2gFq*&W;=Id1qW3QhmjZpc(g5A`t4PW;=NytB0*QCq6EV*=~pH; zdu7u%Xxy5+1Ubj=#bs#ao@5lp(Tu2vJjbK5Ox~IhrdMV1GLtS=1}UO?xs+R2_1R|> z9~Q{Bg9WGqwM_TAJAL=qlAZIqb;rm_LXIbq78%cQh!!krr8F!hW#IUyO*#;`(vICn zpX4XrdUa#cM7)ZnlS*tX9mzl_o?Ts1Ofn*|zrY5I{V}8c(v$G4t6iS&l6nPTeefv}tXE1}@fhR64 z#*^dmghr=VAp}Y5@Mr@bjoyGKC&^Q}uEvw*?{;Z)&KIRbwx}y+-v`23dbCTU`=3cs z_Q@`dE_)U+L%@$cLBfART>-x?fWPZm1^>?w{=>S2&#EinIR^#!r$49Q-RIySKi{R% z_EjMG^JsvoxG_PoR|c*Ug5Em}s4e)l=FD0GQ`zWH2=4(Z}AQj*{&bckN` z4LAH%#0jV*%F_F#)jVOUKz@^c2cI#bB~W(m9BcG0(>hvrox$21&^DzQ(1;YH1>)ma zb{`7iW-7!H z*{A@$?%NeJpUbu{hv(83Y!8`l6*~KLhr}lY5)ofIs+BT!xj^*lk*$=mj6QkS2;n-@ z@@LIx%dN2W0*PCiM5an~|2&)t5Gv(RRCM|n-^E9x%6pN%r7DdSDJUN>m|Ih|LK)5L z4zz~*jF^O&2gz>#v6oWRysD&;uT<+1bp_}gM1=mQq~QQP{(BMrQu8-U8tqVUZ>lTc z4+H$e66VNbiB^WWHuTHSLjQQu1`YC~<9A`P>usbLi-DwJZgXnvxk(OUKUICVu;N5N$Mr=-#4 zsA{inbb{S9jLLcpu-~h&NyD>H?A@;(d|O>78ng&Gkax4hT8Cz%Z1YE9TeOSRnf+_$ z;oRk+3%FHlscrIU#OF%4Y2p0FCC~%eI~gmF(LVOIF(ew)u+;YOQDz+0tQ;zN7WB|G zDT3xH@I_x3)v$uK-L4Jab(w;{ZT#LO=>CzBuhVN-jSrM9Reg`f-TV_lQ|^+}z_A^B zVpXtTf^jt5SisvV0G7kqBCC(-B;kupKk<2|!!2rhT+k6Xgk;N9*pOkdm-Fo&!&=OH zq{L-HWc;bJ;*38{-lR?XhJck%chNTltlMyD!Y3`=`SPsUufOab9*s8KBjt3Cx}wejK)>U$n%+^h z1~B3;!}0bw2r5bMqZXggqrD;LC?LKy^m8G>NgFix#vTW(6L z!J#V5xc@9?+(*8JVR3i{kiXc7#^KpQ;fq_1-VS zeT`c>_hjITqltV`OiFCGV7R#zhV)%YOyvOwOPevkSn*u8y%Qg~Z_6#WOnmcS%wvoE z)5Wcn@eHFtbj}itR`pnj3ghz#BqoJL(cGn7JCafSiTLOu86(G zE(2;g)79wSPB}y6L!*A+YSjNH^y|p0ugmKiqiB}PA(zgSU^6VpKAO2b8r*ILvfbih zaRfLzrp9rV7*`3nUi}l6J<*KaXvUB2O3~ePB~{6IAJcg ztd(4PF!n!iHQI`IE%&<`ZC9^)@am7QMz8Kl(Sd(;HTu|}Q9e3^ATEi#?=H9U91%(l z34zcvC>uKj1dgp?fOLHTS4svbQoJGyday^ho&wXjV;&Hp^v14xzkRTNYsrnnLEGFr zJ!m7WRN&}=2m3~b)Tv1njT`@<2tlXp5kUcp5$cBRWDXm(rjdBIMWkKBqvu*aa5TDe z2;(i{FS;);>Uo1iEAJ?rJUtGAjAexW93pp0E5Y-{`a-gv`Q_65#hoNx)kM0-7@$9B z2d->|8F*cua~)P#(R++eN4FN1bFxT2VWG?Q()`Go?B-$XO~Ou;Do zI?+a)Z4}mlX&Q-%B2y-E*8;Ob$Zw7@{q8lo)k9)@2$6B|fOfG74S=hbL{tpHqAKLO z2-Sfxn2-&<%gkllv5T=S*J1Nxk9g~)Pc@B~ZST%^LoE2zBp+-p{x(t|eEXR~oPlDs zTqbd@sPiiCRDCnwcRfEJ+BSmIC(Q#fY zrM;3vPmi3RZ>97AJ;yh(QAvp>**MedM0DN|9Eln5f{+eI8X3~TT^(u+Ykifec_5-D z18Rtg>8FQwOUwe>E@S1D!#X$0+0wP3K*e>SjcJdzax%s4xSAbe`X9;5<1tKfcBrJ0 zL+c@Z3423>Wug7$V?O_C-Z7;1bJ-;pI zRK0Ez|DMHKc9GvoVc`TYRs|j3ZNuJw66ZF_I&NZkSyuvF;x;OdKS1Ii5%M}fj_r^* z=g_eJbjhJ}4+Zs&WT--YGBEeyFvb?)x4i+=C>Kclp-D7Cg#Hf|>y+*Kq-+h=EY+?_ zje<4B#yt}EHj7M=Xy4P4y~xThw~BVouy}*^uSeDaD%<&<)tzg*(PAX-X^VBAf}MPM zQJa83E^U(n33MXcUP-nHX5bZv=#kS}3ExiHmM~6+)qAutpkfFU_yaBok0TO+Ya}%T zo8ILDCl>i^mqZ6|Sl|1raV{AQr@R}UmxMl+p-)XfJ7wLp76WsD(`Klv5>PrPL@k#& z*Fpb++lsB(u)-P7ELm_zxuc;++${c{Kw^LJX zI(IgS&+_MZFqB4!pbP6P9p9uf3bSUAcvlOd((CCT+EZsH(99sQCIlD)&Gcy`b|ujK zS=;-pIGT#tPhur21V(fy1g`0BlUBhQOqXqW6V;Gyuiz_g@^gr7xnkMY{aG`XnV!Ge z#SDVPS4Z-8`cIwRN|~-|zQ^!AqEDZL#XQsV>GYA?&W_Vy`g#`@19{6B&PU+kjA-N> z7_b4Sc0kiP-0z{kIwaaf1|Qd8Y043xMy~#VW8~0dR~L&UEAnivu~jW#$XEkNOvrxS&a7y7gW|3Lw*`TWWC6nCz>mfj&XaI?v!^0RYh1 zXZrbW{+ct!=uFAh-F&xH?ug}lx1l>c9Th(oB3*IlUIFU?9e98<384h?ZfHZA*FEKj zkXWFaTw;QdM|-0uY1M`ay(3+{dW!8a?$hkd`|~X>=8>v#zGD%y>wA4W%6hVbd*SpzFsOB z1tkMnW{#2rbMNaVswtu>SZ6t0#nA#Ijq^eQt2D6!3=a9HgV_k{z_JL5~m0rwYk4J4s8(L_gg8Qr&Gm;zZVyZW%T=< zdc`RlB<>y+^B2P2YmxYmU}@C&E4t^G?b0+7V|dPV>(Q;0?q7iX^bfLy?_b>S(7qcN z;PFb`l9C@MU#--w&b(eWXgvQ$s?mOopxZ2U%iE~A)kwgp?uhCai7Sf$^!)m&=^7Oh z&!*U`c`lXoYFwji=pF!GO2q_Ia_Hq6$ZWf1tFKhH$?55L_hwZoeiwtzfRC^I4QktU z9$E|iEEq3jm&`b>fdd5#@@GV`H^60}4;Kfa{ zg5=!7wX0U4>UhZ1gM#o43obDS&$z~V&gPSegrTk%%-%J?08?;FBU2zLS&97(;)nGJ z3Hb8#3g-0y`BO`Q+StrS4$hYU4bnm1e2i@4)Cy;w?>NDysC*rIuR- zj=CIe43KzfkBTfVqlf%HQ6^v+8i+a353bPLK+5*>~t)$`z#V$8`&DjaCMQyH%2fGs5)$O5~+4MR1F4H~!OD?b8g*zl^$MJ^PWpB>~1{TBzQ5Y&kn zaMYBYzXmA2(*LPhtf}`!2r|%W!>$;gk$3I9{`{5+a<1 zK|p?xrb-;miZ}#&#(e3Ek+@C7yh(nw*uwF81wuwhO{p~oJ7^@XXc8H}OmxFt2$`=GB(~RT}pNjjv&rQr2T4*JIMWawT6*)i9nT02*uG|Y%Nf7nK!gtqQY{%-3ht1 z$uG^w*^o~3SEd(cK>O zoSf+aHyZxJ7XJS7PwtMxc>~@Oy;}C+49{%f40{9K8pAW(UYpv zk2I(6rsm9ORTbZ;Dh`{z7T2n3Wp&jjaY=+JBGWS}Y92~rgAi)yye$#L_l$~~ipujv zr7M<%$+qX6Dsg|)!C*SS6;m*^NGWsvR58pemoDt+W~cI9!TxkICQ!;RD50-+ z$Z0Q$`&wg`KFIqNFT}$dMcZAh=es*dJhqR>xQ6c87SRpMTPahth$cUnP$NakWSwKR zxvW||lcUnL(x1TES=27s_EH{pEi_j@Y!aD1q7xXg!k5dodjsA|qXNOV<;t9veZ~kD z^Jgyw1p#2C!?lYiASq?xKps((cjC-s+xb1eT4t9EIR4hO@~**OLl;lTMbN_@hIix| zz4TiiC(?OF^^E+1kPJP?_YoJSZ|jT9UJ~CQBQjmpV(ltKreNgdR4i@M4P#*a(k7nW z$oGxLPZgn)B3vE@=%TtEEMmyjV-iPmBzFY@OA!QQjEqUwE|(1#GBnz9@ZXl&dYQz= zy?B`V0oX3#$yfG5*e71TY*uhK&*EOg^4qX&%JoTHrtxylbK4Q@K)CD9rEt3JKB@9i z`FQD0U#~dua!GtuuGH^fgB~`Z-eatY)yD87-iG7RsE$&8k*ni7{BoEy+~ox{|D(+! zQ_2t0r$Nw5TzC$)1$q2#m}M3j1v4)L5|77=WU7{LmOESn>!323Azdu%C6AtH9+~g} zOJ%#uty=4ys^6vi`EC+_ohCB!mX`hs(vdSF@hVICg&pgZa&^KHY|HuX0GuxpH#YG~^yysr&=y(sMmEUhurmhs zc;hhbx!j*}sua@D-pSy8Q>ePSZ`u}#erd}wZ*U3?Evj_!>pqiveH6yfma9ikw!hFv&JzpZYxDf*B5MvLh z8q5L(I8tOV6ZGU9G$L3pmzR2-mhO1nwol@U5wVdEA&ipe>weWE@g%yk%A<&iC-Nq^ z$8C&K7m((X>~L&xYjusAlS5%&2u0JVQ!SB8tU+i2fz~3?dr!1d+NE3h;2?@iG_VV{ zv%*AN<;}<@Kr!+?^v7AJ`pG^Y7OHw|Oi27ZDpe;Xq+~m~+Y^j7H)tY*#V45*l}UI9 zJh54${Dlpl#u(x1`Js|h6z#B>10;SJ(J%yuBz8nFSSfgMR&q_j^g6p;)9UHOT3)AK zG)SD*BJExh5NS)2%$(^Rqnl`&>Kz8PJOPMLP)LAh4rKR$SE_^15y0%Esvd9ZZ?rNk6}v!Jwqg>aW;EZFKajV09fJ%KjwCF=U-jUEyLO4h`B4?S zDdhMr=1(LpjEbcD>7v%8TU{_Jx^K9qUN%n-Yq@l9c0Q9cHuQSw%lhuXQbUC|Ul*Z6 zo9=^e_=ae{B`#`&j8+%dS@~{*H-m|LV-ZZ6zaZhA{Xmha-zdIWfonn0V-|x(A9j@Uj%P@9JpL$TLy`rX)zC_ z#yK}V3G;Z|I}ds#CU$EKY>z`s@8WozH`|WHUAHaA8)?LGyfSu_fw??UrPvCkFA-tq zYD867NKFYa@AgR+>+42v8vJA&!!_-K)Z(bH!+9W`)e64fLgH;1@p>RJ3(B^mdvR_@ zTaMHG(G`Cu!(98kfV>-%=<8R;Z6<>)FZAciQ+WWc55S8N$T}R1z#H(^SyfI9m^&y| z8POktA^bVM0iazhA~7;oHau@_w{DTR5WQe3(QnqYQl>mj-b9$eHPTt60hM_&uQu`G z#R=fqBtoVyRv2)N3SuwGe3>s8X!h+RIogB%k;&I&5`|yyoa_yFxfOL$x~YnlPs6S%`$+q^MzK~O>Y$&hcV&dd7%tnfIsY{ziV()O5f%r z4sBxaiuld8>q%-PuIFRK^wJAVO1e)Eg4sI+(nSYj>CEfB9DZG-`+5*c+J=!PPju#v zIAK*z&`J~_@st9EtRw7R9IN`67YZ5{Qjg%00KK&Xx?FYLrRX(qy)s4H%^NaC5Mj%< zhxnEtZqODjlmACme@lYWK_X&PzS6YhQcxFho2hsvq??Ad`l*x|OxS&V8>gT~=8SNi zRngrZ62Fc_pTQHO0M(a6GwfjWE5gVf7MWP8rK=5E8j_2(k%yE|61OSV(MTCDkY`jw zC^NKdPUhC_P-P(b2yzH<=`MQ-Psg~E5nxCxUOKEvwT_4H3UJE%##WI5Wwp&yr^G7T zQmGGT)-7au%dH~sC>fMcWqPahRU}69IE}G6V|rAh@20l|M&2Ot!4NFYgJVV>MO$Fn zH;!VD{G44QfrP{But@%EVQz5aX zo|lX{$}-(sWp^reNZdMx+XqviH2ccv-E`X+^cMKUKODn}7u`DDJ_b&7Bti}Z@OY1` z%W&QKvZSk`5+VXxm)PzCUfMVheVVx52xT~p2*=cD9&v7Lf)l)!R)~AHnoOJ*o9LW3 zXriNt;hZX_oW$v)BL4X~oiU07`?IZ@(#)pn#E)RRn{Y zyo+c+k6=x}5^JuaTUD$b+#uKfuBmA+IJlhp8!mJ0#{_@nmALe?LA(0y0Ha zIc!{H3*Xs{CI7TXKW$D5k^9wJcVK>y zFa;}+o=_n1=ZFp{B`OC!?=U?H1Q1SJ&XcxciD?x&9i$2IayE(6_Ys^NB=t07^T*jG z8JQl_D#&ETG*+DG%zYktsg=_5fUdLl!MKt3=$w6$1`YM$)9$iDkHigijnj>xaoQRh zr)@RH3A0;>h!^U4fqfV@be zt6oT|7E};?snr$+{UUR8>LAE+*=8BU(LN_F66PjJ8cGVSXTUNN@=dykk)};L6c#ap zSF@zG_6o2rjj%YsMih`&?QAKUN6IA*@y1^4KqhCB5fmCl{Iebm3!Q6TYpV&4OcCN2 zsx{0)zq}hnI$D!>^JZDn$PunV%MFTAo(`)0cbIDJ(R^?!i5DZG2NS}GA}By0{JdO2 zTNpNIeekGZ(YaX%-iWG*P)CB4lOPhOHP@9l_Q?QgXSG7&MHoLmX`QF6$azqAXbyfB zRUxV8zSV|l_EOjG%N45@_N-S$qBXo~WqF~WM}N+yRS#qsji02^eO605(#u_uJCSx4{TR(6BA{yAA#=#q<6UJs(G_drT_%$)63@mZsD8T!YQP5%pxzG;TM@WExNcmN@;lJW zgPNqzG&n{8>Li_>;V-~45Guy{o9z8z`QAolS#iVCGIJURX?UgCSK4pbyrOlclK7jP z;W3jQAEJ)fF6(F7q<8s9`*c73gLlCV)i^9SZfD~21G-2e92B={;DlJ7UX+6(<_0pZ z@hj9pi_ec3N8JRP~bEk}6<|$GL$AdU0*gU`hGQ9gR z-zRD_DrI|M%;s9|ElQ7MLevZA$2`bd0Vbq_ZF7y)fTTi%3%uovy%xqMQ1J6Lc7>}a*pfb zen+>$c;3%bNVW1i7D_K5UWJL7p0O@mg7n-x^8EN@x+>9+cePRm`A0k6YNfPK{lDK8 zZ+h6*zF?l;gF8`2;`t4j>UIz88xJcdkTq%W9WM20i8e? zp+O=NI8qI{W3O0CN5e*59YjZB2JfcE(4icdWZwx!5o0~i6WIN06dacEH~9aZhTrKx zymAj{kPdK}T+@B%ouo_~lw;1Njo1)#>Bp=!$t$#pHfvCF;J%SA4e$PnANB+@DGWDD-Y{u8kvb>Gx zO_s+m65FHV5g@u_f-K^!1};G`;1WIGMx_FABuchn*E$~%f(>EdM;a8SmkkQ;1|vzz zR5d~*&UgVRWvB9rl#2_8o5a0xRcpWh#pnf}iKz5;T%bcxtk&L6(?Xi#)v0Ye1=Gvx zNYN;;M}%EI5tL>4j$JNSEm^I24C+%L`tm<6z5~7A+Q(tc2HgO3nx`QA2~G~jQ#EiWc8PQy*8&&ktXvWT$jX#`n07@ zt7y}ysZYF@lo<<%~c7Ld1kjNW5Z9nP;M zu~=k!)=MF!jb2*vWP;zsi39jW;uulqRDeF$^bPsl3KBc8g{<}n2;D99o*+{`;J4!^ z(OeqE&Nq}HVytA#gY_*tZ#x4dZo`fyoR`lT7@kKzOr4G$<{sUi!a8?M?6ROdk$;)) zb+|?mr;m)~=qX?`j(o&FVg;4hqhxYr%o~g)2#cn1URX3fxQMsF@2Pq2Icb#@8^JpH z!Ox0T8l|&s5m?!@deZHRyxa)|qiDpCz$B&HhZ;x%9+IVx=K)HY6evH@8X6Uof0tfZ zglb4!t~ET@D8L{8yh3uk{eQQNYTd#gMbeU1{6uY(ir3<Yu!~xqC54G2z+{}_{uY>?v_rp;H<)ca z(1LMJreZxTAG>-VZRgvN9G>s=>ev(@86Np|nlK9bYIEhKe$!{OW*?pzj(U#{&nADK?)<}^-93EU`JV(xBGUZ^AxHKri zwq(aje=>;w@Tt}aKoCzFGm0plK$aYuh*ZwBPX~jci1ctVsATvZrezr};uIt{ORpDk z>pH-e(H95CHmiCoDHz(@Fzt;4D}Nf8;ZMKZ8t=Q=b|)mFEoT>H>LQ6Vq%Qzq#D*X$ zUORH=%qBLgoG*7Kf8m&4d{!^hS^P6jzGGsti7g$b$26(5(q5wLM-8HbXJ<_R%4*p+ zg93jNoZEr|82jYtU%wWPx%{;re5L}F4szMJg_!=oagyASr~(3Tw4b7QcLbPTg7nuS zQsT9!h6+W`8;`|#^;DoxtvID1|NAc^qb%XH-)(v&dV3`FEJB{wtq!@nV;8Pki!{0x1@?CDcy&Y9XZr(hhKpK27QKgx1H7Q`oB1!1#woUk&69yp;b70!VlUU^&tPIb)k&E-xiqs}ZFl=BRp01`e~rA3xgz8C5uN_wQJnK% zz}Zv8N1b8|0RQA;xv-l`K4x8yz#`lgyHG6~7z}$2*DMZ@_-RV0B=Dc5kP#lia)!HH z?!8usv`OC+P{y=LTlpq}RBP&TL0Xo$Ylx-fJWgi-Q@tN1>*Gf7-F?K zb%mYhrPaly?NQLw14 zo*4-OPP2fTWK65ua7|xslVeYd1@WDkN9wIO5bMU7QdqQA-jp1=c_vCm^kmyhs=ZV+ zY?Im)dJ~TJ<-`{tlhYmCMMWX9BKQ5vgg_em;^LOs7&NLCqTfu2j|)lI0w?ru z8y2i%d#2ps`J*A=ot3)R!?QdRTc-1%v~xTNF39;s zaW;KVe>|y~*d-%9Jj6;dZe9bUj4)%(Dw=2bOC{u$N_p9TVaOy-HVGIC4L7&cR=zul zXSC=vn~1z5a3_BYOFUT@G`4@6#B+6-E!mEoDx6RuaeKO+Jdkk|CF80InqEKxx!lOY zz66Q4M{>>>_UZEpQm~F0y<%5!Drr*nnFqM1)U>=FbW~S_u~Mr zU9}3+NZci}01*yMdvy0+`_yb{Eb&sNqFRNiKg_YcJ3lfiCf4M}brN?)`!`|PYFeyd z%RQ#iaoG>FvAR6cmTX$t8+Z`Ewx%;M$=n&Po6h$<0A63y86*PbN!&Ojimfpe+iFn6 z)PM&ov9qo!U(p9t3$s^S*x=txErdd^7i_p!fXm<5!0&qXrs>Q-t`tX5uVJp z9Um+-t9bOS2cX&a6pal2x2X_%Ezb2He?K^ZrZ!ls!oVO>ifnB*M1%;%g}$!;_z2$?I19cq+K!$S@&3dMe#Z7c|uxQ_rZZ)w{|@Th6s@e+iC!MKq7- zBiT_w{z>xOs*Vrq zL!hZ7KEzB8jb|FcMQDg+I*EOEgJlMZak~MTN!Ya6u%cP~iKB)fhZ1rO8-mOx@sZuo zGKW73+z`vuDJ1sa4VtM_p~kzxGmSr}(73MYgk_w@Ff&M`cURX;5}Dn>%p!cxsBv9~ zk~nC0cxLlgog3qsLs&m*2s3R8f6S>d%v2Hw?}naf(+IoQ8{?VIF}lVuGdT6QG0aR7 zpV}Qgvk0H!YmDbm5{K*#&ul))y)m9SBqle;Q+C6RIYXHg&*3i>HmV6b2WK{}sE7k9 z2{E@uh&;nT0redbB7=tpuQ6`1LxTu#aB$@C!CeUMsNutdwYd$zONQ^6B@*Apxr)-Y zPJ23iM>{By9N?Ly<(B8;h$DNMZWTBVg*;glW=A~j)AvP7d)%&bA=6)8vTPRt($nS2 z7)F}3NoS}b$Mz=kqHh`SmR>%=4B#Hfr_>xhIem+b zE}mI}JPr-zQNLBp653Fm%9@RHl*EOmJG>4q$<9hS`$u>IQAjDNS|=b>g`(Myd#rzn zr6EpPhy~{4X$_)~X@*&w9D3t)K?ThcLy7qjra@cqRqS@Dq*et0zCKE1JfFyIABAiz zx_DGlmNg5-9+h_WJ~{CsaS_ZdKEJ`*obiZ1Q=aJ}!Rz-9eZ2{- z6qOw+O1QLacj;x1#D8Mp!TM3bZ2JkA+u}Hh<-CocK;q}}Lv>H}$2Bg?X9wd{a*MYM zka)U@w+U1e-!u#H0V}o@5bqK!5~E{b?BFdVe8AU>O<2xVYr8MWA6?kd{VKymoH}`) zIP!kkqw$@LAMGVFzV4O`KS(Kx10%X3cb$pLTbRmy@i9CfCP9u*5p=e_cK?KHcy_s0 zemu*<4&Idl3;(>2VsLHI!ci3}gCjPuOtj2Ex5g(KkWhgYqX*{I|QSwNuchnGc z`65D0*^^tUYj8n5WzreIV6S9&0bJLBsznn19-dOxvOMlT`s_qak)^@@AGU|6frGl} z*)=s}+e>Y^VBJw3WIi|%5Vpi~9;8F7XOz#={qe_%xkBC<8Xo~^+oz}rcKn(5Q%>>7fU)^S{2kG>4>TcJ*&-c`Xik8qfY$YRA=2$QtR z?jtcf=8?p4b`AJAc_)FcSB~Qmu6O~Ncq%`=gtHDM1!@Oq+Oa*lTQa1(F5M%?I+K3P z&l+J=pB5nT%Q!7*haMZUs4^Hh*_6Z?fV-@ zz719>8t!tMH6ub76u_GQunQS>u_0`H8&!9~g`Nz~f5-5eKK6KI3&aWX^^%9jyoEK$ zYLDo3VJ{OR;N_ACAYf=gwFoq(mB5bhYgihYKIw1+R~W0lXCB;Hxpi;nVvGZOusGrQ z|6gmy)a(Brt(jO)L1RY!pRJP`ia2%9F_^{umeMj>u|dYMDW_?}|~BF?xyS;UHt(EqO%KS3u{2!DkQx!zz%tK!~7tOGcPEFiTFiAw{47y#ZGvQj zCO{y;*Zpef_&ik4?}`Ky-xj?PGa4Z>MMr(J(aA5s7|EN}O{%f_!KUowHtK7dhTKNY zg5qIQPRKB$lWuo3Y7JtQy`o&z%G;F+&S1~5D=f~7dt@%YQJ3eFCm===C&Y%<*voYm z2451z_--D(%aq?0mgKw>eNh;P8RyAfsapd<>cN7l{7|;IrbT3|U{oXmj_9+E;mC$2&$fpj zp=}lz)N)VvZqC) z$`5`kJi zYmSA54%=dnYG{j6Ad(dz2O_AT{Ww&h@aAxoEb#}JbQ#B}lg@*F%^%<;@m?Z97%xI? z!igtQv+}!&QQ>1c#hFbA5f;?WtiezuHpbJ!&>1+;$&t*fnsAC(-wK>(iavnci-o$Q zx{_t|0qR@wK#_u7?J9GztNB4xmFacZeTKWbxVFzG@w_~hBsn?7$t}@d0h~1++hC?* zN?S)Fz!ag)wnrt}a=y_It1NdrR0>1@4nsWXytH)$2lGw3Np6bs@HOX~M*Q&0t@4zw z+jvh!Q~^#F>E)evB;J+V0}#AJ@_RvLCgHqk&QVJD(m&+>J&(W3tA1F)O&9&M_5^do zcg_47PGrvI?TS;yTnwL?EfS6T0wVfVp9r7ZC(R#ljKgxq`dp};`EEVmt1)YnLh}Hcuj=_nJghcg3jSd%m+n zn0d2;`d1RscxGih%-Qu|rJHCx)aVhgatab0_-Cwog>dSu5Ywgv#*Z?_j{@VL6vmGN z;|KqeB8icI4#xQZW9;4I%G@rYwfkK=bSyAI(M^0HN}WLk42sbk(d1xPf+qah#X>z)Q$ek+pq49`Cl%Ck1+Ur_=ro)&*Ld}27afkRpi}2=Y=R}ZDW-;n zx`u^Z!v()+6Bcp}cl~0Spwp;fkrCg|MPtzoqk3}0$E|FOPP#hUOUrZ>%eaayQpGZ^ z;(facohFZ$?lZ3I=b~f3#FHoD9YMw`N`2=Enh5+ zdjg9MclJPoJbqdc=eW{E_x{G}t>lsikifCNmsZ6!^mEZ^N8>3mC_Y&r)Z?Xx;;?S% zb2OgFT?cVqS|11FpFtM68+G3Eac5rTqV(V6$(=cfD}8~VbUc8ZTIAreiw4%~#P2Yg z+zrSM(<^bu`?=_jWAT(&B&Jpip&iFih&{Sqw5t~ik9IS}qw07(QNq?TMM;AV zW@Uo4t4xu{W@QTNW{b51td?jg_7(eZcTtAz{6^78H^Tln?y&pg(5GA{^h}~tmz=N~ zVSgMyEcm**>Qva4`tiR>?R2>*#eNzGv5Q2k zPCadl{q*0&cIj-3%ZL$sFpk9@jDy$%B37sVvc(?!H?fy@wZ&z{i2W*##eNkBu_=FO z*Xq>ue^{>l>fgj-DzfBN#mN08j^%z62e}0zSEpv!a=$rQZeEwp90<0sx)`C~#j()u z;vjUR2-T^#ZK2=&P3YxawNOh~LyXYoI2PI*2cbtqs80RwPs;<%CkxH7gj&LyV}$-3 z$3lOOgV3w~)Na+OVYbkp|0c9k=Wbe6OIq@E(V_Kg983K*4pOT`s!px4rT%)d)NmxE zg<8UDV}$-5$3lOPgV0w*s7`%t3;q3Mp}ZEi$rBVL-yxE9s??S|LBH(VG8*{uyg5GKNLyBBk7#YDC?4jg zC=ByUAhUvK3fJ%~QlqTY$94@-KJRoJ$jX_!O_ACxHy z$}I4)L3F{7Wr`@}M46&o09x7!^>oazYslzj`J)37y)hRgk?e65K)&nq7DNAA#x5cS5zsWDxr{5sff}>l`Co&XA(IlmMdB->?WarO$F?U zaz%N7wfN}&X}N5BEV~|fr>NDb42{FBhmX>lewUo2M?5lppE)c*j?-2+>k2g(&y z2+M*L5>b01>hKF7M9zc?MN0&Tk4~^GYK6$}rN4gP`>TT+|R}Qaauv`-&HnE`=n&aRvld2tAZZp8O0t z#2>uQK*}HJi7(RV8wR|nOiFBXgRbL{SZ9f7M!n7uC4!%5& zYAl>oi6JIb-D>d z)U^pL1t7t7U0u^KyD}7;QAQD?=y__9HRRS z*4klGI-5uL@tz>lVg7?GTSV4v+BogQbz0Oq7L~)If<)_B(7Qtx(kMCT&%3p>XdzMC5$4 zLQ!|gw3y?G$i-cU`|l88=L$vHcTolU@5>3M?!rTbYoX8q(S_5{g;QN75hWa~P?RZ3 z9w4F=*C<7kTqdQQK1$KUE|Z++jZ&0xvsSiJC@KtKz!)|P|L5a>856b%WS3AtkAR6W zp+Pb&CREa6LPBViqM&Hv#MHE8l%iLm$;wfR(pqC?SwBiqyUOc{nq<)nv50I(7UGz2b^mEG1 zeB#?BQFqc;;%43HqI(n3XcdZvN~HxtVH9?u+@rE3(_*&tsDz=T6(!xNJL)Znypf|7 ztw=N}ZS-hG`x0TNJ!8Slj`{j-OL4qqZ1U}v(WykVO&O{AL3&O>N9XZGlT!I%Asg}> zFPD#49Nh}sCiQR(RClq4+hO8tSCDUR;H$7^^wo7k+2q*29!)jw3_nSW4`Hhcg zB1sa2zi|+*<0Kwj{l-!FW|B!BJkt7uljeV-n7g`#U!olTVX@n}U)Y6+Ocf+AtHA(Pk#|CLGX1VmvHs~@19cSI;u{)b5nL&-s@ ztyw6b8xW3|oDWH+#f%zCU`vB&vdgmNyf)|zqZOqN)C!~tjj^R?8*~P?rCB6AW7*}x2C>7hz8EhZXQ zcS|RR)CI!cDimcK;S&t6>)M#q0mCbi^Q+Oy25pN1zhQvu9;0Z7a`V`P1&Wqrr8cDBg+j%9rVB}2X(pT7)mBY*dAn%Y8ZhesQHR0}`K!V}Mc zqCujgEV>X$S;!w{xgVWj(j{SEkuQg@;nJBDRlVe3DZe2eZ*$!CKRw=k5sw*oJ5s!q z|B^qh{vGLXX->E{&7rd_T#wm{Gk9V`+h~BSKNOa~u(mrSG~4@}IpQtNV|fL;l^Faj))Op6JX^azlE zMR%iUaw}}X`rv~zWqokPStjWR2}8BA_L|_XvtS5UJ@YQDb zt-{_R6lJrmvfCunVz!j!T}YPbR0_4R3mJQsNxTc$0_AuYQYDyiH~~6&7g9g;@4Jw{ zcVJmPTAiXv+Cq;8JL9oz-x-gY-TD4mrnYK}V73W`RX0LQEZdI(g;v=rTnlBH$IJF} z&W3+mw82nf*Il2jJ+5_FY4Y-7r3oc6&z5z}N)zjt8AzhPt~AjPUPRJfwy!j?Z2tmz zyxOIe{2oNRcDh5pXre>*Z)cLbX{hc^^(TW*Q)W>~z zf{zK?Ld8T~;9Rs#i-CF|4~)QW$~LQB2b*3s9@(Ki>?=_8|NwrmS5(^4?^G;8zz zUI7}(7MwI9+JDs!Lqt|P5X0}IjvS(D2jcx$Yz5U0%n;G@jpVdcnN#2m;Jpf~min05 z<~g`C&12!`oM*MmJO>W{DB;R6ijs!wpg)6;E$m?6L-J0xxJFy-Fe03S*ub=la)PCD zfnbM@-rQUpo0)4FKAec|9{9qI>a*?7Qcg z1-FQZIz%5BIbn6BAZmmH zITCUb&{=DT8=Ny5InIe=6de}a38BD`h2*Ek#QN=T?zpuiaAS$>i&~%7-(a6i*Y7Ay zH{g|XlQGu82h$BWp+~Z*Z5}Fq%oU2d=R>mu*xVqx&V@EHc0%u<)E5zovJV3J|J3_> zzB<#=H<%x-_ZqGD8rFMtvZ>qY5rZ4b-8`IM7kJb_knK;#S$1xoF1}4Dx9ejX-@d3M z9O)k^q{j_?u${O|P{~p$qqm)eeooLWk$B^6xAYj2b;oVJ`rGPe6zF=5+wd5ExEoCrTKijcxX^3+l;yxbEX-Fev50GQC{xC%TY9)|5|W< z^zxrY;jQOh1-_g@TSk`Oa*daWl8tnnR43!C|8?6p$@qZ_?VC&S=lAEE#* ztB~NE`>%i!ucVliuw#s(`g;t5?pJVM_?KMH26kr9Yir_e~x^@oGr zNKOG!A>7dA0+ZU?hxosW?k?+YpwWo8?^18-T@+l942)w-E|2z62O#+yU6XI8R#Ca$R;wv?g0!E z(uAEZ6flgPEUg=0|Kg=iwqu&4sKK4=9mBw()*gP!Q{pj6+q9Q$T7$Ihop;$guOSa! z^at{9q<6X8cOEgYQ4GWb6jmv<{Xh3bM15ylm0G+nOC%~yj*E`_$^3@b-R*G>DV^a4 zn##xF4T}16m3TXzP{QjFWtdC8tdYfbFpBZPQdP)juX57HeE^@AH zZ&KnGLF^XF?~Kik!)y01$C{X5lJAT70%f+){UQxJf>D)F}G6DpbjRbEv-~kdQptC(=Hk? zD(LGG4*Nn8e@Wo_P^cu-p(7Jpspg2K9A11o-$?fnoj-80i|ArYwewLudH4ktNYftC2I+MP6InR)ghv@`@w$H2;0?r6zes&js%*d!4^0=Z_`u7Qf}`x^dAS|JX3n%ZNVaPg)wgnUpw0 z_(7qbGHPP(1y^Amw|lHEenhCJj6UOs*=1LQeN?iOTS^~fw{B#1*GgS_s8CNCeQ6}# z=%O!}ohqgJ3H6lGA)|7ti_Yl|;f+%2s8CNCeZwD=4`gpd#Z~jGW&$onbE?oid#L^MV~UeN`x*K>M5g%ap;cMU5C1} zq@@N3^^{RHkJID8J}DNxfl)c7;Ydj_zV@&KF<7l*!r&obdYqB&Cz{Se-p(}118_LP zpK{LRnv;4&FK#)?@gr~n?8q%)!O77=pN`|Vv7YAQQ+k->ZuapG<=Hq42yJ?zm&6}L3M*%!BCg<$2iq?v)2ZUmP0yk~_ zSVga1Z&J!rV--1XFe#;BtfJxgw%ib7HZ};pkZo(@G|OhZ(6a#^Ec6G7Hn6d- z8%??-m_x<}?S)J7{6+YsF1Cd(DdPJfG?XqWDFm)}XY5b{w21}uxX~mJ-i_H9hplcd zl6fI`@A0@1H7;6vqe;#^V-@Wd-zSSc(58z)l=j_NMQ_|_Qp)kMivH6l#uol5#=P9H zZAIM-%bGl_j*a%YzOq|Oi;Cz>eDOr?tUQi~vk(ql`U|r+$<#wWu$UOaKJo@$d zay!$dNDsdmiaYVtyji?-LMRlx2vZ$iKUo7@Li{*IO$7$C@Y-G4f5$0$?Piml?Zzo` z&eXQKg+jn!2)J^bqVH}tDW%6aMM?coBa2G%Yf)SaM0M#WqAU?f|0HTqKM^&GMLD5W zTLC5uMcp3M4YA8DvF(H#jj}a-!4mtK77L%GNrA3HAr?C>_KA55)`>TdQK^|wP3;`1OR>G-63b$mh;NB9vbc~qz+fhYmy zXGG!I`A|fa>Mynq!QEl$9wQ7zM#cdank`+QN3DS1Sg$8GF z)d<%s?vS>->ketVvG@aRSBGlY_ADvYnMl`CEz)a*M%9AgtU4J#Q*gCHqwt8R`Y1eW z^^&F4mrw?sD51YetW|n{s|z$qsVo2|vK^bVZ5P!f|TcH(?Ax$Fepil_yiiSp)ejkyfne4IS zAMou;6#tvkF7IkP&2m%R6u1duVP<5IIBGFU!qR8(2Q2*vBG^$@$+ag* zh;kGSmJWLX{y^~?215gTjju04?BfqdO7cVA;tqJKS`x~&-9)u|L^hWDiO6rH7ZCY- zPj``juu18;r6Dg4d5HYgJzX?&FxFsv5k^mcgvzj~+|-h%Gu}Z)xO9ZRFoP!Z*7KLa zIBY256MN)sXKwGKCir3d{tYI7DIA3-5qXVEEbf zhH^z$4>3u=|A%k*3M1PkG*2s}U4sW$1w_-0-qT$)e27Wu{8j&SU)<-iQUCVMM6FI(gwCS1Wnvc+6DUZ+l6&VGOjt;6@`O=tY_`txp+m1E9kxWjlQtw81M_H71kw4-cF#wl{;#{&r zN8CVR+fZlQ;v~Y8j|frW)R`{2@otmS_2DBG@)yR);QLkz@lD9AyRFfo5Fd+?hi^76 z!N5@u%?c=pzxUj2k~;_s{4^37zapa!89@s5oaLhQq1gXY8Km4c6v{i#7_X?CDAq?P z)C(daE*-BZG}NTDtH&!^hypyLj&uQS;KYRCCb^SKxQA`z zoYlz5LQar28F2lFC{Wy)>RH?$!)@m`i7L%PVaoIy*C2q1Mp}e9bhR>=ivLCAbb5IMboUSyvh&mNONCKL=HliiG4+&`A$^w(p@YU8%fD$6(6*UOv5HNGbEBXeQ z#p4xS=10h^9It5kBHik1g(74=2L9#oiiR)@qP7UTK`3DSod~3jVHl+CTdbuu2?gwH zNPBL)q6hpYvDmK}53$`P=6VYS{xI;+)_t!@EPwF5(6({BqCSGl5(@nH@M>&ahw-Y6 z9neinwNeHTE0f|3``N@d>I&So#%UbueWNhtlbcEn@)YON6uwk?^t~pzGmzz{sc{$| z?}tMA6$7^jZzmKQnfM~9f=|is)SuD z6tLFTo6XgdX)#d|R&#i!lw753N7OE}9Flf*m7>q#=NqaNomFU3+U-?}ZZC{E(8cj< zZb<~4r>|W&>pn{ebXXE+q!-|4jcEycsHV`Q^l~qLNN@>%Xt53%{M|Dk7!f`z!#U|ld&71sfNLjHp8tODIN!iDo4SyFO zC<#VFw(Q23Iu7VM4saa@rH%tE>Fj`2hfbr8L(I7&@HZ1WU2Y4`xsKSe+;-h#LJ6))vtCV6-}W z6h8~tqjt@9(Qg5~heotVAw)WsSoZEw=y&=en7#aWMTtqClA_#RbV%Jf$3^cV9q^U+ zq8~YZUeKh3;Z=%)rFx|N$RMJGg;k2`B` zn^WMk{nD@^+CRsYkeb6wU@6CL_&UFm`4ass|*?=?YDmxxIzw@gqpC<1>0Su4m3@D?eMGb1K3_&p1p zTpH8Ec`1zR6Z)6?@nVqWtL9bFKHD76F||4V`nfLZRSJg@ZH`Aw%__CrvzeWML;Ox; zY>tQT$dls@`|hD9nDu@sMkwzHdXkw#%d{_g(VOuH=elS&(j0O7IDJOBsmq7yvG~gQ zF3K*)X0iQ*eS1$g7XorU$FZRBdwiGU&JE{FCPsA;`a-7bJb|8fwTm~@7Iic zZL}4?KXaPK=rqUgZydi@NLDX8%GKY6bV-pNzhB0y1uhzk+^G0H#<>T^n3OPhf}#d- z(G0k#n&Wrb1V!bQCZ(*NplDeooC#!dg>Janh+iNZD@{t9Jwee~V@={xxnrf&#<8)1 z9Px#`5#K#nhlgz^*2OsUQPW7z3lKeOPFvujx5pwJX@iM9kaL_#?(7PVt&Mz*;C0BU zMNTnoG7-(o$DziSwk!>VeBPV_Y(a0~1+-mVQ%toVnMO;Yw2O-J!VzyyVF!KlxdqKw zYc>15MAX(u=U=L&wrdu+=!J16r3XrKeMD_PTi~M4#_5Klv+(0s=6G}qC<;g?IvbGg zqO?~gD0+0fNeKrgD9RE}A8?LMQ1k{0 zQNpPc6;%oEZ16fvRMd=8T_-AP5KcdE22E7dzADyjMS0M3khjR6>y7wq!|4&NtHBQF zZivC%Va@O=JklOzkcWS^;F{?fE?Nk&Y(es9Z2>=qfdyNu@M+HkMY-bCh)~#RTk``1 zXlt^fCBPcb1e2Vw#wuAECBe3=?*z8Yfp3j0^)qcV_$#9#2*P*3?Ib^0C>E1=F_te4 zn8SkXWf$HR>}*S_T4y;1WiB$(i-;}?4RFz86ZCf<7vp`z3nrS>9>+W(y5f_CIIs!h z`qis3Bot4?25K}1cP27cBclNsWz+-C>@W$I@cZWr9@W*=2}Py2)`rM|cZMx(8f>lj zbA@{*$>$2oCBIH6ieu5tAIVJQYHT}8`Htnzwyfk0x@vJLa1 z2dM;Cm9($M;nElbFJ|yT%uUs>#I6E&)8hy6OM|xb#+X`a4Fj)C*BZ3QMa9)71kL4x zg<4385fm=+`GR)&j7`znuF-PVu$;#sCqGDQ`F+Me{w65GlJ4>2g1>;b$QISOIa=fn zEpi8oym2zdfH2Vxe&y#4B=MEX4zA+5DJG@o`nw&rYV?vE?W$|Zx5d-r(bZsA5+GX1)znNC z@2%u&-m$A$*;36auBKfLRwae>2s5s$F^L@66O}DIXjhUpcA}!P8cX4Yd{zh!+4Vv_ ze)^r+Kh@y#5_F3o9aHGx*G%K1>OiC)j;j~Ow_D_*h?Pf6M_Tpv==5R%4IwZT=1+GyF@7P?*b38%Wc>eVQ&=**!vdjYrtBDsPL%p9lo$H zROaIw=zY0%tTsHSE7LK%#bMw;Vv7UM5#6} zNW;k!q-VI`{k0~!v&Q3`)2e?jGa72~7FHma-s9wNlEm%Z?{iMax$q()#VLxm6AE(& zU}uP^xqYrlX^5H#~%J*pRk7r1?*11&^T{GEbu=9-{MYt2=E1D zE$qdQ()20K_hotWePP?Hwa-U;_ppOUgu@Pu2yNz>l-OJN1wuW3`jfxoyBXXp$+fvf zo#`(JI%8lSx)6Wz-7Jcq5DEh#xH;`WxuPSIX)!T-VGSmi4z-gC(uJZdt~zq!<6n!J zEKHQex<<-oNOm`&C|iblQMRjOTFj`j=w+5cMFK-vbSv&)?rv9CC%D0FKHepM|?|IlW*S3zl@(3fl1(->3V9=pChT;C}ROuiy=j|&~Y z(*j-;?q${hPOx(Cb4K|BJQ_h~E-T5)qkWvY5t)t14A94%(`6xC;oOA*Q0iVT6fV|#lB9l@))r!8w z9~sq(PFakN{8iP8vIP7XtZr$VJa0JSs|e!Gdt1Tgmn=J>=-W`#=M50O-5NWU!o`-^ zZ@0!*<)N+tb)3_UICoVLl!01>U0u~NysmYM%veJ8ud zhMmz4SYe*0jspx*gQVtMn`D6&m>Qa^XXalMNhiu z;^oMt_ZWTuay+e?fX`(xhJ@)&u6plsMAu){in1r@esPt9DCP9Yihcu&C!>BZYwd>$ zMSx%DAnLFOcYX4dlqD<+f~f&hQU}~6h|^0<_O(()PX`fC@MqmG6Z|#zLyhFgiW&qn z1enVwE829wNh#M&R&joDh&(S`cyuKn@ltdUcXxSK3EFLxci+Bfu?{ z_7j!`Il0o7&Xjf$Ho@Ky??yUw)IG`yq@-81d3vImoycblvlDNB(naq+fJu@Uc(W7H zzs`HGCA@S_PvqF)H9Hag%lDuaUb7Rqe|ds5hs)PIh)vi;tn;o9~w*aCfBnX4lLotO@E+I!gO`T~jL4HN`gnyYui|{g zUmc~bj2=GHL~OuUZVR=9W$)ImHdyEnTIdff^yZZ~DJS~D$i2!%6OkH$RQs(idU~bR zsy}d=JSAm$)XaZ3uQbWIZ?d8zqOkKdT?f?0X-L|ilNDXH3ST{!DLN{s_$U;xG98}3v0O`uCQGz1=b5i zL51gOMHPx(2P^?=I#i5mqhfG2% zeHB;N$wfCKmFo{hM(JKIm+l81GU*ciYFU^JeGxxh67W?-ZY%SJ@`~__s$a#SndLzF z5FdhVeF!^ttK8R|7xyp@*C%0XKbn5ztbPycFZX}qRLR3QQOK$W0%ONtoI8(m>yTUI z4M%9ULTyhVHz7DhQB$>UuL5{x+7w0mz;c4MTmrUED4ZL#V8;PVn1`^LtR;lNdUT4S zbJv)Z_Tm&pJ=U1yd>b-FLX*%@c(8zG`g)3@+%Ku%PCSRi2-?h{@iP+m zdE#vffnEM6PMjti`pfV&3O|jq@lixtGct4NF|O=8z%r&7{9tDUJ;kDu*1-wxY-IcC z1+J^hI+NTD$SI^(xae5qWK2a3^g8F%aXE4xo1!S=Ep2{xq3~Wkym#7EMK7&0b*NgX zZc(VM2rxwSmn@FkB2)Oig+kh9NJCpBtVbZ8JylUZVP^>i>~X-ZnW`wAVQAT*!p;^7 z*i(Q(S{YyoH%wKOJymy*tzZqCs%YtYlTv(B6>Vco$y7zPg4qU4^;AWl05fN*qPPuM zSlu^OQG;OC0JCMPqANF;l(ucEqQx6ba=tuO(E`zAiBL?0{{{TVsfrS?VWI@g7tJE- zdGLOntSEn@sV9raHcEIO5D^E3!noJqQW%%C2`$S`=qB-*DHOO@AQrg0fpZ?8s%&=e zDA-QGQfm}F%rLlgfw1d@LfY$qL0Wu08mME9ve~PnV7mbuQlscL!2Y36-E8-z#&>j3 z={|LTJol;l&+(VK6O7I^bXu?^(mc}b2Ulm4z_}ik{{p99c(2h z-a}#pt%}Fk>we56?qD0^(ZOmTGs&G?gJnIf=1O)VCmT5>MY$vB5ia!&Z(!ZE$oA7_ z7Mu9EN$zIk6w+2MIv6?G)3`4^$2qefM@#9x)ctL3b|0Z|ZVFB&(U&$qE`7;D^^QW> zeQ7U?8veiQ8avk#kMG3t_KX#aslJM zR69-grHjBqUz)l_`qHC}L0@VROh;g1)Z zq(LY&dd=zpjZeU>ta%#=rglPs>jtsV{Ol)TAUc4>PL6`%nnmsac?^ReBEqf`3Tf9^ z(snQm9l&B&N5QTI3?1Oqr=XlYsbJW@F;r4+hkest9S1u6e|B>C|LjEg7eA%L|7Rz{ zzxFAUJR|)1`MwbS=0vpaL=M7|zrhI831=w7MSmhU;48Os|KQxqwqg|%z1I1|i4HIX z`FTa&{4o8+LRW4zDPcs7qU7nibt~eD(#F*&+PBrDwCOd9j&GIdJRt_h@6k5EKXrg( zq(Ap*7~W!eeyF5;a7n1pwz%qDT}oU2Rsvi8Rst;V@wB%5tpr$Jfs{6f=&b~7!Ztl^ zQo{TiMcJa|JE-=_8b$Ab^FocHTH)*lr>RC!!ZRkN9I8>2`HV>km|7YHvI7uwft+U$ zAy{v<73GqJ!nB=$!L((75jlUUQIsKoZbAWk7r@h|DSD11U1d~TUDGWN!JXn-f;Lt++G+T8g_%a9WDHyA^k*U!M2-b#m9bmUYjWnX_llo=~r?$KUOu z#XfD_Z#(|>+L8wp8=sN3tq<`pqM6oUSNhK~w-;FZZgAZ6T-gv3c>s~PT4=E9+oa`8 z`5)UIDN?>uZU3YINi)Z-{DpN7m1bKd{YZLIk!jX)>YYL>-Q4QMhvSWI=pw&qcA50L zcSiuedz5}@x%+aDhXK7+jAUwA>5rPEkiV~dit^! zbbXT*MYM>g9^yfY*j*#P2(3k0VflxE;R$sr6q6asCkD{< zmoKfn?b0pAk=|RN*Q*)WpBupN$1FQQ zM8He?oUMQy_ z=rtgd#old{unAtL*K5tXv6J-11aSo%5V+r}oJc{}y=0l0%|;!GYINl}cNbJmYZlAm zaY`K_u=VQ@=(SQ8<7(&Hzz+bo8t|-{DdHTL?8SZEFw)K0DN8tMtIZ^F|0M;^wF-tx z=40GC3|~#(gbpk7&_BQ-zy8#Gq~i%(?@^dTVtZYGfI@rlaNGwXO-`W?>^n%wYsPcW zh8e@w7@*@**lpj{K3f5dZ{Lfzy_pJ+nW0LHTO))odhRqT2F5+yAa=v$R^2$l2y9Y{xjZ|E0PvAR~W-Xd~=c`D|B& zF!A%K|BS&PtA$(-uC5~=^g4j_HhjB1@;-J_)0*wBy+xA2aaZD-V+dNnRIlAf(vIuwpTbRK z>?UBB?;Z62uzEbMS|J6@B$F3~|nKA2Y|SNN}n zqR_>ctp?})kX4GPXp;mIr7Oi3^{sg-p9;WXp3U9*X|8Y$c*{zx11_4CE6>x%=lo1Ic%=SvX1Q)Cf?@Dmd{{DB^WCbJPmH1s10gSQO=`)k*hD|agqi<4%mzr~lsf3Q*hFJ?XrSK(td9phET2DHfm z&7P6M>yabP-&xdf6q=#D>P^{pjf4_A6W%yMlZ0@1Mb0|B`$G3l2~(S}K4*Y4Ka0CO z(TXGuMixDKq>>uKOqQ5noplP@?=mvk1Qe%7{25A%E%gnh%1rg2f0n*$)446942(1A z9V`j$Xf>TV7=QW0kDsa;p*`6yn;s-?|VdJhbbGFsY zg%sP&7LgiUFe(rHd7xx|{LcP>_y6vq;I9xQrpeKw|3n{`Dmj{~RBcXq_UZ{5JIpHR$j^fhmK-lmsP2Sw# zL)i{Rw+u{3k`e=|X#(4~l!|BiefAx zi%8<`9fyv8KoaPP6G+sflnXV?s5gNifj)&#l&A;6yt}m7(%CGlr zZZDHWpACiz3Z-Y9X>hHdb34=e1;Lc$Rwe!@gk>0fxwAOP(RlH_Oq!3d{PA^`qAw>( zRE;YqY7mYe#*-5zKUAWGuBhMPYx~Tm#p+F5`|Ni)e*V!U`h1sxJYZFVGs88Y0<34| zOq(9a&cTV@U;vKx^@m(jN7((VA`j{SE7u9*tHICdtF`BsB9N_}X{T*;cu$ECD=TCn z2ZP6~-FL{yig$QbLM!-9PLyC=cPkSxGqxs0u`B49|6pz{@XhlgTSX=_iO}NnL(JF) z*rdFHG^iA8YvoLP@?ccdK>VLo^sf=pQSJSf&DOjs=V%u_FShP;nZ$Ku^xs3GGAPjP zKs^bVXn$5E0aozQUF!nkqdi&rA@*hs3oL!gtgIN<`ZE8+kl zR9!>^~G;LsSlGKwr{E= zOdoY_D1PA|@^0&bs|WA!sE_z+X=e|W{S%Y@^wP7v>C%g>zsejH?>6_^f;k28zoZ~Y zD$tIj0A$_;%Ot%G5(BxS;9KGqD>{_pEJ5o;S6|&Fl-Fw@e~X%%K8u^apKEp+oYor4 z5E5(Y1xrqAn?KJ97&$EMgY%GxBfj*vOP7*lkNAg@1L5t16ba>s-bRD-eTNV4)-Gua z^NH2|1Vs$~7-=7D-o=pc3f7{1D>>jF^B7cpLow#^ATsGwe)aa0Z`4sS*>m>kMdI*n zb(UrLvMjIS_?(d`s7^vh?9Z=9M8NEAwCT&;-NDMpT}c4Y<&jN--39=tg#(_&Xoyt1 zduW+nlxG=CQjciCvEUxO`tTTPD_!rqm41;}dYS z7^XRp09bri@DT#?=8qA9GC(s92D0kQU^xm4c=f!RX0`THHW^t)6!UF07;|F;g)F_@2F-B&q=bls(KM-~^Hy7?JR-k@i!>QjoybI;Ifgc+@u z2NP{MOcD;?J3u-R6$+*(=yw0v=E}q{nnttmS6f(hvSEWkq4sGb_EK5blZ5)&tS@2B zeY8xz;R%vKHGou&NNY9iG?sE3O}6Y#GET3dPG`?lL?ljp=HHO*up!EWyRtrsVV&59me!`8RaXotQR z_nOy}%9nZvgRvqQ1N{fv-sXg2_bCOV@HuY)?X=I4cjjrOe7fmc%u9oD5mQXAH%@`z zw0p9u=BIT&V+EUQSslozy`m1@wNlMv^St#Tq?ZAF=Zib6SXpoq+EYMVch!bbTq&6_ z#de0xMwQ}_Q25p)e&$zhg9*?e^!B4(e20tC!>#>>0r1M%U-0J8HC1fAROnB!kC7(N zJ~}k%SDD{X`*y_8GNTbTDHH##LfZRw!q@*g5>lC=NxtGjN_y057+Dn8Txo{jBs-pj z0_koYONsxXTnxu}%JL+^0BZa{zk1XW=H(T9Tzuxw^Bf|ZpcrFa+cDO88Zu#UUKBC^c-j*J6>Lz@?TyQyQn^xSZG*;zeVLvDa>L6ju4qSH z*|*0Z+mAcKiQB)f4!#yo3YEP_d8uA+c1vozaiw*|cR3$D$GbR7@(*d6{CbY(X5+kp zbj4Kp4~0J9ilvT+mu6c%D9n2?*cDee6A$kd&Ku9S2j%2|j0=9))OYra9K2`LMY#%6 zlAcxC_tv!{t9EnP!LeXvst`6p!zeYl>s)1^obY*`nYM;@+P=%aw+Om8^&JS&()vjUz1k-MKOAIGc@Y`$L2z@)Q#EHjo5ZN;X29wZ2o&OMNu4Yp+ z1v9%3F>i27GCxyntmcT2qYN%aa;{e)a9&QPUj-8ADFCmUIy4L55AvVW_Y*aMdeg3| z(=FVDUxO5uCe&=>N>s=d8a!GvCHWt6XXf4pO?>buXwz^eJ>9C<4Ch|iJ3);8%KDFe zk0jQ&eY1GaD%M|Zhw%lg?M7c=62;fD*ZvTZ8vIx%6~@g?Asw&M@`)HS#vZ_YNI3K-V+l_t5Olw=M8xMz8N4aru*b1iUosiwHVy^% zDfXp*!d-`jrON!*4XUdhI{Dri1R15scO|4Zs9{8{TyQiC7<8)Wc{!Y-sPdTpZL<5- zzSDQHBl2H-E_N#&O&w=?2x{?zf+ts-udB@sAsuQ!vzs9Hxk&w^NZ`EXO{H(f_6Zq) z!O%>Q%;uvb14hIOGYRj@8?+$*)kqqm8*^xgB(7aZpgkgTb|*P{t%KF8)MV-jyd?xM zjLBju7T-;M(rP^xc&*ZkUETz~;`i4w@XAJ6!7-s>e1anUQkQwClj*BL-q<`UG=HQ9 zl=O4cdP2qd=>N@~Q+K|e8ygDt3c{QB(2zizcN}%*WNku#+0WhS$OFXksi+#^{fpLh zDfv^bR&~^l)DM66%7rq?YmSCV{=@&{zOw@W*h}V450YjzScbt zH^!3?dPa7VwhpJk*6Yc)i^(qF&zzyG@C?X?wf++Ys;TK-5%F@3_G)*wJTww5l}A!s zW^-L;FSPUg+xzAbH$DAl*Q(xgpD7){rp-o@eO@EpI=$bxKjf8)kV2=%^FyI&_B#wp zTe>*Pu1hq;8*onN;*axye@TBm<>We%(1@pUySp2i6xt7W{bZZLy%jfVBX$<3gZ_Fr z4y3WA{3$~CsPHlKKzd46{ni8Ys4!J1r#Tm>M{6Qrththfr_%Ph6?Th6XHvNz0?&>> zHUa$G%+C5aDu;cYS^3%P@>liIK}@Ceu&f>06sdOp4*lNHK@1*%lhOo|VUh7(_ex~l zbBZ(&<$S`q|0Na!1KkY}a2NV{3G#~-X52FBwIk+bwYkv9F<0aMcC=BVNnL^*Z+F~M zQH_qgLS(7Ci{#x>GtaTiTWO3r&I#_`wFS_ZW9sby#7hC}x>TSpGpT1sR8yo@_2Few z*_E~x)pOWrN80e9yvmYbag5?8QDe>5XIwSm(G7BuJoL3{f9{wTr(2?z=qHv%6yTFy zcq%*>Brd^XKxDh!(GYf`&B6 zni)yRPc*vIq)mE8!X8Is50Lq+*ZX&=ROEJ6!Ih4n%Yup7^*KTFk*UukmwTc>Ntr7F zjwP3%SVytF%S)~$0oshuQ5S)Pe(BgEAE>w6y&_?NiRKSLfi9G6{9I^>JOBr^++bc% z4ryzI#_X9Qa@UCG$PeK<-(ZgDr!bC)jt$U}8P$Q;1JnHrUX>TmD5a5+-dsf@2%|ub zZ7S*urGUn$zneJAqB7`-WE@|idof?z{gyK}=_Q~3UR7f!?QZDW1VrM{cIjTMEIogu z*~Xe;0?4bqCQ7F7s1%9*Q1dNtl^#D#N|{P)X@Wc38yXJe(@rZPU{4x+eSu1ULa%bs z+@MtE1vTN;9e)ILjRjL5&>w%i)@UoK8E!O;$NjIQ8@!50B}Dw*vBS(Fbv<#JEEoA z`AGwlcfX~(4$(sjw?Ctgpz+xSB{_|y{v)`4mKSY8i&0#VH6^WT znC%Xp2D}tfbFc8oRa^QNPA!R!bXV~^HP6e6-6gu6Nqh2P1Zzu#{gY>Alr8_WhG&0M>F3VN!X0$4|M9&aUx4(IK(oFhvEK^yz)M_~;NraTV0KKDy`-G#VD< z(RM9Hdeu*O1b_uERF5t##`SzvD2|fM*Xc?0rg|piH*d$Br2W$eaLb#qD}P86HRhCy zem|fczg?1sR|fBYz$o7P8sQeu5fiXVFoAO3A>L*?SB$5!%-HXRc^&ty)#s`6w5cR_ zZ*8W>fP*A57emABZ^&Zato(O!9j`9hQpwv zw6{*sk`{9N0Y*D^dvzc0?=Wju>bh(jc0Yk$S#svF4+`W)VgDNHr5K9~z{+i_DL0W^ zT=yUvlRshg*2+_EFZlA0ZSPC%^dp%L7L0c%lM{0m&@0Y&JcEAQxf5@Q*O`}C8q z?2ljWJc0eW;^Lx1A5DmUk9@0Yan|y31N(pSfH{kVOE44&XW;Qab*m%(7l&6PDGMU| zmh6K543G9^Ab6culEt6k)ngl5eBj2W%SsYNX@7rkG#cWWQ*4`4O%;K>o7js-wQQWy zW(aIe9yjmmOyK33893Zxy%|Vz5Es8Q=*0>+b;SLUsrq=CS;TKD z^G0Kc`2Ep~)Zx)L)gIB+SK&CgjDP%I3w-I%G`<2V2S z8@{a1bFpeqhBDb|0jmI|d!4wA(yo#!*&n?;idZ7MVCn4%C^xHxMb$LwuV_a`=D}TJ zKPQtdBTyG3R0#w|8q$FJM$9DG6jhqW!oP~0VKH_|jj1EdR-9w<`t4`l4no)LQR=w& zUW;B{i?&6sfzsUEt4^)?o)gxe4rtr+mNxIK_!4HqhEfYqM8);~TxW<6tvtbGkHeuU zPcQo~r|$YK<`rch%NC4*e*!X?0CQ3LErGD+Q$BK8Lzb2qgp|oIw~50HCR-#!`Xd$u zQnwryIL1pYr+2i&+8u4#rs;MIfFxDXZ-3X1d2t8*389#wcEfk@!*N*kkM5G(u zWJMLYCf6O3;;G=Gb~L-)pg2_)^>9#&z41~$HZklP+$VKC?cqbTHA7+m@sMAbL*2mcPMcFsyPY_iM^{;2vBvmH}f8LaUwf;#T z32<1mE&4;kCrho_9=K*-H_9QIg0>@*JW zTv|qS#bC=KJvkhP3WmK$eYR?*>SkRXxv#otx`>o06?iTs(etX@ze`7V7_=MzgrS8v)H)a~SW z5Ym|wj15``6a+h6- z!c08OW#>;RFS5kk)O&vC!IQ~fA*)u?Rk}&-N-L2Hi0y|bw ztb4Sp8RLOL=RDY@mE?f_`B(h89DW}|RmW0p>9sQl$trfYzisST$W|l8{}L-47&z)e z{uW>GG=xs4)dzSMt68Vu_UFuzvq&tHiO${5t_Cj-Pl>bgN>2l`iel9YREA6+oKLU@ zYYK%fjC+Xv4VWedW)ZrV0=zOgnU`P<3}5z=&g-EQ?nTJoV=LDZ0T;`va`3xOW5dgX ziH4I}hn4<@#uF>S0(I{rEA7_UvP}%jXQowfgvkyvtKWeu*Za2DE+)Pxx z>u*{^A4!_UwzbEZE7yYq36)P2x6wM1qj^Y@IKr#`Vt=h@XNe0cstR9TsG^!*l^7P? z5k?XWYjm~mRpYF{i*8R6F6i7n8cRNQQm8Bl+U6V_krDeb@*&M5+o?xLjxg)afaADR zKc*FB&j_)0|DXA>(A;#jpdWLB#A%LEbUce>^lJ7+KjuwXKPgvIifV+M9D`ZOJ_FV4 zO_xCdBA4!uFNs!dEc7|LItqZsiSRWQxMgHs@nCol*-=^$1^WhStRz1`$G%&P4@cdW z;06zo+F*L`)$=M+c5zl{F~U&}iOTF9N0a%_N#Dr`WUuL3k1SlO|4oFl#v)0ZUi%2i zkBAihr|{?5Pl#QdBMFWvHShZs3(-%p)lC3M{VCUqL`VY;cK>QL&-N|Zz0%_+pXx|R zDStiF)i(jw3HV1M&Z&2mo5%jx>8~j?ZPgjY%qF;I3HVqE4+*Hnx!1dPmp8}fY{Wpe zDc}+z@)!7ZR+Wrkr__{1Q*7ACUdiCH?ap^aS8_SbfvqZrqJF9W*!k*wrl-z^Y3I_z zwE~Qkwnhj+6CLzTYcKl(`S82MKJ?@u<&r^--fA89wHPEi@>-h{Axxs$bLK6&vZG90 z%Atc#y7L4d&&`--Qsu)3B@P+)ikrlutD0!(lZ>EA(x~J^4MxtA3VLKkNzrFvnSmYG zeS9&c3QcV;>$lSsAYbs(m!%SN<$Huv=wH#xH`0UJe)(kZ9&=D~n)Nt)0ZQ{p+j)sp zf@87f5|>Fb%C-|JSiOxjl_H4CTuR!1!JT8k$zVvs@vT!YxWjNZ-`v2Cy0aH#=!8;K z!c0UU1H;`!ASyB|$z9W;upS>ez2Z@^2jKZtE6u>db`3Rvzzt zr%gVgs#HI-E;_Fh94#+ZIMo$&$kHx22llW4oT8}qsdgoYzC;TXeYf_SoNE#vIsod{ z%LjDn0?Y3yXFb?;w>x`fM-P7^gcT|wJwHE@KYbdJS=Z+ zj+ExB(GNY`6S|U;!Ox)Z6B#BP9^EwNBZNnKB<4LrQ$)JaSjD|Z1Zve)N}9>X+F}|G zg&pE-3E}o1sKfpii?3%>JBL_?zPUUF(`@|-DBWmN5W5~kLC-z9y3Dw@rIddfnl>?_ z8-&i%@sMO{>8zLOTN61J?=G1B|R>S)ikD~b^D!m_AHB0lrIy9 zw50{1h^IDAgfi@a!pA#kdw{JVw+h3f@l9KoCsVB}!fp_b+i4=4I9^U~;_;sat>M;- zbsxJ3`7$y%CscXAgDZuPzpJD`rNih5%`a~8IrHX;+_V@iVoY5E+Y5;N3y@&wIi;7t1_2oY6JYGg zaktEqP6dVgTjDfnUVGFaJ8$!Wt1(S2p3Dz_Z29ShtzM`R>YJiNU&b9WVDak2etd*n zkr$(vifq4=5Xq1Ai$~mEMYe2@jPS_|B{n~bhA-|#t%IC<^^&H;8e{;f!o15yuTkPN z{@hD;l5qD(<HvrkV} z=xDMIJa18aptOArDxV0KSj3c!IhrLtaYES|DmTw`Ie#EVhdh}3fqim<*UiRwF<2@y z2yJu5Z#H-X%i7pQ2?q(yr!gV05ikkg^C9n-1OBMOLJO&KQpD~QPKQLRHW}C~@JkZ@ zuMXlA#c`hFRveF7m8}8-)#8$C;2u zjj$F)BbTF-v*7MeRN-c-)oy-$Y2*5P26CEC3|6UJ;tRBPkMpl(MjrN!Qugql<`2~~ z+^rl`uCpI9c#^2JQ?=!_4LsE5UXSKC#N?(dXam!R6r9;&EE);l1X*G_2YF`d#Pac# z4pdk53$?+HVqqLd4n-Mp%^4*;qkIl^G-;zet_N2^MXrnAQfaHgCu8r=A|9NAM?p*- zI$O8>s0y^l32o~MP3?H^@VT*eCRdXF5ntRfI|4hS#RE|#gxEK+CwYTy$UUhxzC}P9 zV&2|Te10os$`Isg(J^npLxQ(am2TffjK@?IhDPU)SvY(bRL zH`Cvzq?$iu5mj9in8QWtKHAAd)j<3cK1>i_`x#@j8*lo?9@$~U4CzKpfnGb~pe$SG zyBOpiyzj zD8EyQrI9na-ZIH`f05}$)IhaL%|V0xYTn*{>RMUvH}5Lebn~mX)92B;$}jUQ z+cbGLsg&1A_^@UIE%8SF+yDFATWc?%QfC4|p+k^_72r3s?Dl)^;Kfh_(A>Sc&MDrr zVM3#YHXmSg1~6)crgT>y=C0rUOR7v;FoUR5#&p5KW#$rmUC9YexsCA{?k<+zy$R#?9=%4AU-$&e8Z=lPCOlhcd09xec!%;b#MP-h zR6yLw!n$HMlcb=qg6lnCvg)xMvMH`9ajxKuyAas%oi>=hDnd$U?Z$0_PDW&`pbRJ_ z@4f_pH2YVVwpw!s`PYvbf!IU_EnSjBN+XbS$+5W`yZzBTl!5U6n)M2yW`EWS#dL7U zMjKEtjV=}7VZ!yTLJogc4Lg%ul(B9w~7_a|-wW{DyYK4rl|SKp0+I?YM4 zzNmLrd}|b+;Bq3rz3+k0s}99fK=WPQ-G8-Hm%D8o(=5du8D=e$#*Y6r=*MIg6Jk}Y=(xf!<&H*DiT zU!BIFX*6`tAR8ba(`37 zMCzHiv_9?B`heiQ4yf<}o_^rwvwahTzWPU2rnnYTd`XBTS~V^Bs=Ru|_Fx73^BEWk z4T5|WgBENN?u}_%01o*_jNMt%_rY>lyY<0X4oN*8En{)rOV`0A5*)&lfF?Y@ZjZ!8 z2yofcb2!_K!2r$VjQMVxoWrxGvx*=5nQ_2EPgi97M@ve2iW znjGH{vedNX*rPSwN@&}ohKN5$Fg?m$Pe9zYJl7X|Lzw8Wrqy~9m3R#M2V zQ;MCxuPIhsT8AP701&+9NtWNFJWJ~|$cwBbW&x!XXzDR$_n44F1@iy}a9SAH6Pc4Fbb|oK&T`{%STupU^EG!{&xvrA%THkt0gfkd$k6Uo)u& z7I#VnhZNLAeC8Vo6<|PAfl|jN-Yf5I z49?k!FTwl|G*mn;<`@@}^Mloh@luV)$uWdFrhbUy6WY%7};PV}g} zQU>vR5}?FR`tiH^Vsno`zJHMsk>*LwEXVm zgb&SuqzQG`zz<<(Cf57Fw4ymtxM6$^31;H=+vXSXBvROze)mCc$Qx&hWCx?^3L znsxc2uYiaiwq{Ivc6bzEI1Ju3fxjpMKSZt925y@@_aSu}AoPcsl|#DpDobSN;}N7V zQJ~K&d=G?VEsWDXgFux1s8RS(NNypcTj7)5A0KnrXRQu-<~AAN?^fK&u6%JV#Pz=w zXAHRz95T%ttu#=FAflpEQ)e!x`g0|0_ZD~HUqVnHR6jU&_N5xE-20L7U~^P=m)+jK zYygK4fYX_HaQ67FS0>nWJ2Kr>$3och8+Qc;azNTx6z}a%|0?f^YL&PKuh+kfO%A@@ zNEtw-PNFg#0(=6sf>rCfCS~(eFePEUi`Mb|P=HCc4a|z`Ezt-&8L0t3B zGoa(d_>x4FRRn0&CjlbYfIpzp&9>Be95Q2pmZY2siZ}X$czPdnfPdc{Axw#X799+2 zj}4&WqyTKbi)3*;kO0jPq*S}*!~7E-E^1jAt)t$ zGVVU6azLB`tMWc38|{*cH*c4N(&YGCTIVHcp#Jf<2-!npT-gaPtYU^u-4TJNZbR7HKNe ziGfBzM`-jFyq>6k_a&r(GvGnVQz?iLR=4u{&_Zm-^tbZ0rnW?D&Lxx$Ph})kYm&RD znM3QG71^;1gns+A)TcZ+MI-C)x9Jp2?fyd5@1EI|fvA_lq+$!&!u=&@Z_3WdZngG8 zC|=u<@7#0T{S3yCnygV3!kz5oPv06XbyD=B)oePBHPVtIwaF9t^1~^UB!eMu4z%W_zZhq1J2qi{wLsrrLAY!b~DJmJMw>7?ScPX*Tu&-9#G7G3P z7dfwQR(9ingheGZlPX&XucmXAcMS{+u|HhN;t~Z??$ZwqB8L;BsY+Za}ze z6J|M8u6)Ovf9)=OH29U(07tO@v}FMF_T)fp8xr9xM~b+Hc5+bH0ZLn&sq?W zBj;;I3JPtpzCA)l zjnF&$45Gy&5<(07y7Cnm8Z~Vz~AI@F6OsuXMMjqwFEZdn|Yq6n~h}`x7zvaF* z%q|oVRyazW?ls*m23aLlR`CmewRm+RR{BW^*tYEz zUvE{_r@->OClNQ10rt}gof%BD7nWUERYm&pwOscKmQ#S8P^Y4v$L#Rr*`y}fk%8+@U|Ckq;(qps$VI1nnf!PV0M}N3JtZ_q2guj9q zIQc=-5IuZGPGK43)u+2K76}}bBI#C$Tn{J-ev!&7tPO0=2zhM68+C}__PVY7J&`cJ z5>V5tyi?;@Q|R*}Hg{AMPv!cXGq>0KGS;?ogtZl%A)?qzz*E0QnFF(TEZF6egiyMoF%uh>IiXReAGOjW!AwFFQ|K`t zNp@`~*#Y4d)VHIPbA`wIC! zypbsYDg_B7N`d zz`jYG)V9JyL2B*yIMQ7J2VTKw!+gWnv?*0;f)|6V;lsW^=+ixs!iqc98~hH%!A_-L zU+(dY&a2oxYqag6HSPGdYLT7Y%R=*vx1Wm1Fu2vwY4xRM0q$C2Ce2=rh9xC(DhO6R zf1}Jua79a#O3&27WhiRLA$HCdYnVeT;32^7 zVlR&Z#Yy(<(J+);MD+Eq5cOc+ir863W6Cq#WrGjuxH!Hv`WuYdDJ}?bDKIh3Buiqm z?{{C@t3gyB&2I(L)X+8JPTvs_2l=Szl+^`D8n0xqGOtslW~Rn9C`3Pf?@c+4sBgG5F=m0CMECqe8i;$(c8$Gfc7m-k z>O8wl-bCR}XX0E&t`p^>pGr-V6$P=dBW!qMM-<>EL0I*U+lbR4r4eF~6(py5wk`N> z(uTF--NiRdQGaf&dXwLdq=3T8Xx6{H77REW8!lC62wIA}qXf>N3ezS3apjWcUy)g7 z38D9fN8|3d5B+b2+93vc2;CPqa);t2EC?J5>~v(*1-!34n}i_qA2Gj=JZr^9dm|1A zulxOZ1F?xp#eYAH5^97Ha_b59qpy~cDTBz3cGybg(s3svIwD{_JdNu&gD3FJZP4pi zfQF&9Q)KLmoVZNx$b^> z%b=P1C8FE{g02b5G@Tq|-9Oc9wjKUouHRnUpm+U<@kX!TXJD+=vHfFMewi_3SYG8$ zBSG~e!`7h=((`u^Y<%@O_iRB96c2mP)ZE2zHW>$s=``K^hF3xEdaGZwFRjWqV$!ZW zp?Pq0&R~qf92^nysq$+MN&oZwu0hN3OL7i-%WFA?dt&ZbLVAw!Ln?NX<sMJiF6o)g=mUl&7y zuUi`Lx;G3_3J;)$s2a+3=S+QMJK=4|78#DOFfHrSDK_xhXKFHnc_ z%focB=(+nx@Z7JDzSBpGeAbFx%lV&De>a>Dbp6>M%n$=MSSd<->tDr5x#%BR&|=qS zim9TA8XvLO9+#RemZLI}4p;X}(23f-{3DoIAWPOZx5L$R&-458gLSS&-307tu=}k~ z>}Tz}5?1LSx#^;(zB3?GnB3|O4Uu&%CS-`DFF7CQpT>_1Z_9L{21Xek30Cp=>2%0% z64o~!d!o1fe&se&0RuVofsv$*#mtEbjjlh3;eb6&avhPZThS{P?n2wV!{_pLhlzK;WLUw-h#b+_j;WU5wd z;`3tn*Fk1}v3IVppLEh$I#0h_U&HSFSU=|Z?wE!uRZwbCw4gQ4W7&>B zX8+^OVDarC@ADtXwJGffBx)`iw`&KF36xv$A!C06hhzE*$N2u6wvhVz*{zBRgFVUh zak^%bn@a>Us&o9>s#7K+-!^_`sl#md!3}xGBse19E}p;0ct{=`+`FR@JT9BRJeg#& z8Z_+NL_bECWYOhT^9$diXidyObCSYQhYE_>FDG)lH4!t5JfKGOBqwZGGsCQAChAB;G0aOVWW z)}@bINQ&cW7`s>M%=KaEurGHizm_w{bjsaR>+H{c-51u2SzMIKU3 zOg8Afdo$2|k?RvJFZuw8T^p|NbLf(t=^?js+ju}rEa{Bu*b+O54xK`qhH-hxu!_P( zdw+}kK>AZj4(;Alkf&Q^!MQ9s`1WBUBXFr8iOX6|eQ%HIb8Ux#NljSRDFl z%f_|v%@abQlHmOuAMs0;DhUfVkdeeqN};$T9aw65@GT;!KkvS&zR_T`iy^Codm=1L*4Y+-Q<8CaQ=o8 z89TCcG+F;X)Zg-|Bn?^R*YAh{qEFRmV`Gad-~0Q}H(r^{CGXitv=(8Pw1Jf~L-XS% zAOhcYd=Oez;U7J)r5DBk{;`*E!-hbWa0`5LOM7qb^d@_UrV=IUccM<)IzSP2~ z{vLB5A-tX!O2B;W=om>lS}5u?ZQW=U0XQB!STi6G5zYKx(S!zVlT2ZJIPr(mt*NPuNJWRCf$omPQnWrPEJo4hxL3f z1uVm!nX`>Ltej4=-RDW}x$?8vkmvRb6ilGPKj3LzRqkUoBA1GAMb<|%@#8=Ltj_il?-+< zdhJ|-n9_vJlbDO*7{CJ84KV%V_GM%k3k%5H5GSaLZY-kVHj$Mw-$iOfx>zY8R!1rV z8ZuE4atVWz8zKJ#?8NHpKbp91&%Q|ZG+j02 z{BTMaQWc(Z|3PuUq;eUZFOGKmmd1~w>MAn1Z|e46WrH={$N-u;58vKJgY8>WykoKA z3l;K}mRV?`Z%XHEMe3K?ZoaYH2A@RRWS9-@7tw?JeL?}CdSq?U5SLCe2i;f3mP6{a zS-)hM6=Dp#gzlXUw#1T%&KBa&Z)AFD2B}w-ht^tcXhP)&5%iy!<#`90A*#E7K>u5j zvLC0x0eN*lF2(jdCeOUEI$r-0;N!z&cux-aDm!FIw=1wG;QoIAtw2)0bmq))&Y7AB zBb8J#6rxF!ri-rAJxV4*5>1383|(}?NKs0DkLTKJzvj&3^85BjYtCA0uf6u#Yp=au z`x+1NY`ZCY*;r!(jC=!*%rH^liH%-rJI6(tdgn*CE<^!zxpCm|fkOcdE#8>pBF~5y z6_xGQi(JRS%|Oa|JUT zg~1f2%)^5J5f$u*iaxGWX`u4_e>S;sbwMEx-kBG#$U4lm9|GuWz_I(54k{9@5Kl< zN*f&*>&*r#x(|C~21n^*$BJ%V`e*^#q50hYn5F%^0Bu9pblpB((~G&Lx$b|^7;NU% zQ}sDgZ@mSgO<SRv;2$ zJ1{34$_ep#UMdM|NOqu1V-n_Es)e)-*@aAqUUW8N7#R-3+D8@BOqj;ax5a9Fk zl2MZvxhNg?!^Vn=3Tdhsx znGn($4x)4S4pu~-gM$_2{^;PfWIjmO_CfT7#QbH#-?kxt^!LJchEn=>^OJSl8FDa_3)5hfo znR8!K)OD$gG74T&l)uzPtzs`J8pX_*->MVlrI3T!thpe$OJ7oyE!mU^M0UKfd(}&d zY9Mp%ONxF0|KFP~j>jO4y;3JM@GIqj-Bz>RPF?0AeU<2Hw$qjP&E4;6w&Sp6E=rFl z{Do0^i_J0v2-}t)GSj+cF7m)qX+LWdyn|A~IzKE!JVU8!q%^6|Yh6&P_rZozr7m|- zTa=23O_?Kd_z?z}^3vsE$~%{fDF-iyA|+QKbo&a@&?iKg&kj3!Bgkz=ae4?uf*a7Eom#4BG1xq}PgGR*vx7<9V9!Al6@C7i zi#!`gC`#L}Dc6C5iwA#4BAqHnXgxqfs{g(&sY-=nr^c9{`sqNI8iqrBG?8SD}Ro?5^8W*C+9S?uv zCHFiOpkkt;T2XQoiuT$>MH@l6b*!S)U-XRRohcYk)a#FnT4F^-2|K&eisWpkAltk_ zL{l7?>1(;dMVTS|LZ5Uye(@D!o%6fIf`L#16pZyptuP-vQ+O!FuL=z4QdAsC1Y=a` z*xAWTt5@hvE*Iiuu086Cv2rz6gWrRSVtRv~z8aA4?z%7>jL;r_;qkcNzU*IYpPK1_ zQ*B>??JO1b=y{B`Rsu~2aMr9JE#gyZ=dW~;FB_=5NP-sgX|qw_BVQC)?B6er(}!I6 zxhwTD(feE>HDF`c9U|2>ULaf-=la-nf3ewWguF{2*yb4zKlKg7c2sEf0qw^hnhL%0 z4cT>2cObb!bu)cgS6z-k$UKCE-@T})FJvBH@8Wc()~lvy5xR|Ta`3$W7A~Uu(Kq09 zc8zRu6u#%B)TO|}MSf%tY4O7w`hm?@H~NCpwO)lt9t`)tJo(LMK-IemO7BPGSboWyE=m`HG@ivx z1S&$$F@4pW@`hxc82W%f6r3O6!UKc{-$dB4V~lB|xag+a?FoP0fVf@A=+xxGPIWrk zgt3ooy7RFF+S3Cx&&dlgzMO*&!oK$T&ZqA7Qj2QzVwhqXBGwus5Q?0K@nroBX{nG5 z3noa8n_{UHro{pwC&L^XoGC~{gajQ(-QPc`Z1i$rLt2cr$%BQ>ii5fhT!tP8i+oCYq}b(CZ9g$(f@_`y!B9$IV$e{`??)?!;Brp8y$;2J=GxaGdX$+Tz5-l=m&n-TsPm zhP0J4dKs89U<^Bb?ZkajpMsF}n*-0oe8)nYy^WPVVB)loFU-6jnA}6SFq3DH>ztwP zURv@tVrvDXqjb!<<5n*@*SN@6r&0asaW&ypFZEl4#zMpS1myfRNR)FJvZ90NS>@a0 zrH_H+e;zzgPe!Ib+dH_Is#8ty4#VE;q3<*L6`sXJLh+qaeErMB*h_2p|vA zMvrJu{M7LG7QS48_58fGZ;YTP2t;1nAdZI$`viOPOkGw``ITUd=numby!p05GWD4Q zL7BqeULf|+F4GNkh&1==MADUO)x+lfuHuc_^BpP@m1h- z3~mr+Uuc3LTnlQ*?BptEp(!6?8mMRx&0+z!;HN#`F`IEO0vCmEl9llDd(Knlc~++%va1lR89xu%&?1#<)=Gr6Uwi8CHw7S~GS2U4v`WsDnoX#T`pRlwP2=Oa%gb*VaJ#0x8l3~H< z1H+cK7`9+w8z@&qf9lGA14mFtmn-^+OMu}_wxrJyh%DZO;W4)DDo8_wa6Pn1E?W|Eu4CzIi zlq;kjhO`kR-Orl5eMvfOhEs6f8G_9e2o1kPo#H!2_P4YbT33PK`N~vE2SFMl?4T>f z{Xw~+K0+=Lh_v5AejOa4?H~Mq^GWu{%^2Y;7YP3KCa*F<8e)B3$qsNsxHw)v;bZE) zdQq|u{M5veSsz`0;^5_}anWBz#KRmBuLBiHI9k&0oFuOXn{EHpAFjvPhvwB_^i~U; zj;S5Rd5wcE*M+GKDA$(<| zN}4o*NP84^uWzd-3(2s+nr*PhOWG=jluTg?2!x!2@L-I?4%!77BjoV{!S)9iz*t zx4D{rjyarS`qfqQ1@0z@l9SQ{^rs8EeICH}C0{jAQJQ2SgXcbRQ92f{g0X@AL}*}; zCh^ha<-p@kq$yVBr(SBY2}b~EnzimzFO}ejJc+;9!u;1nVB)mIda|dNHsVJ=TEp_s z{1m=~dAsdm$2x&9a(mPT=IyTiROaoX4(%j*(wP$;kOW5sBH@)LVezN3&Fzp-4HukH z3-Kf-DXMTdIpIA>IBt@nIY@|e!^0Nqnko}`9ubXCk=X)rI@zZ%9V>?k%Z0B}AowR^ ziky|5DcCB3hRUVfhCDH~K1ZmAD@oZ&lN4n+o%IRVOj2~lXZWh|d^kx_u2UBTFSBIq zpQPxT&s@~@h>q8=^U5WgD1%eHyuruaNR@4`49*ieI= zzmOX2DpbRjRD%$*QpjrsLda!C$Y3L+TBwGrUPyiajqt-4`r&&e7_&oL*-Bj+9pd&{ zj!fnv+H1YG#YLsPm2`xWPR@rT3oGovejF6wuGoXXu_Z1K|GEiP&nUZK?}6PO;y9ZRKdM7P0T zqU$eHPMkD&f>Gs+ssSoQ^Z1?BH^0OqpD9kBhNmUW_Uo6}OVRIsn(s!)JAEsD_U}(s zZtP#^h@X7N+XCjkZ!6koy~*AmAb!MaCRVX^|TlhO7aH)tW69`dD za3T^DWQI6Ki0I)Z8l#6VS*6ve5{R^$I2waG=1omqBM{VGOhvv~)tZ_u5L85R&esq- z3b8)lfDG zFhP)ph;QzD(AB~A<&B6*Hqe!8>B?=6(=4dmy6>cNdu>DIq6!YYrA?A4ra|r(;B;2K z8DW!tydG!lhg}(%c+_@W?bYjXt95C&$$FgrITJt4@)@n>2V%bgj$7DjFD!rRc33_! z0CWVRp4)LnBfZ1&hT%tq-sK-FwqyIV<-|<~YZs-37g{IjS1 zVW6UZ_XqP5wq>f{(`MBc`W_Y{`kpiSznx}y`8}6@1~Q4$FP!7g_z{t3ayxzxbS={Ge^OA2DPI~G>go6nh$vkK8=D+~7KM;sfIb(;K(@tlrtbnf+hTcB*rMqkom|^O^IW(Gw(y)?p(w|reJu`O>sX=a);)6dWZFAA zkJbWl_2g=N5MA@QqKPbl7g`cRFA)e&913ZJ2PxVCK975}qAJN~5=a+}R+P2ZMXlP8 zR`k$bI5l3lnIpxi5(rVxf*+z@w%JH3DMcXIMjE!=VDsEkp{Q1Jd=jKcg`)QRT$C}e zLeV4pT-2(xLebiNFhm1~=u#A^AG8~HsCr-9Lc7KLO&yv2h~8&=@7;%XV}7;z0nc9>0C29BG}`%Teo%J5ry@b{ad;k@?)dP*-hk6kSF7%-*44D^TT7~}5w5pD%9 zf7*wdosQpV1QUJeL*D-pZLBE*SSS&vy-l%nxbH_4kUURSD5_|xtNK)&fS6UGsQFJW zYV|>dqDy{qk!N#-qMD{oI-iL0(5)P(P;|#nF3L!qqNwC27rF0wO3})-+7#6S;T9Kp zh%)dH{KTJhcAZ8k!cqll<`iU?J4(?i$SN46=v&A_eNPY>C5Ykt*5iCzr=4lqG+CDkg>V33{>g_N`+7<5DuCRh0z|S zAg6^Diq;?}w1+BD&u$Ojg9Yk&4@!F|6G`I*B1^kHv^&Tyz>h!GihXWGc8kjt<$&l; zOi{E_($)%uR(OXvWBe3Fab`nvG34z@Y_~ufn#=QG^GuqeDD?zg5lmV-MbT>qU6iqE zilVLfW8D-*f8vkLQxvuORieF-xl3fD7?>XAXn*~$CgyR_aO0%5bMriUDS7&ha2+br~2 zfvDan?1y@*0-pzaDzYUR36cnL`50Bz{tQ*I`p1lbWGEn*NR%WbAaI8O;zaE(8Lh z0nuUx^W17sJ@NYfjAn=yiQiqsokO*hbt=q&&f$gMvAuVcqSwI!Z?9{vLjy;BSn8|a zQBZUUn`^YPwF1v4(roA)=!kR<=N>@~aBNCT)s=2u7I3gh_gg+3H#vGB5l$)ya zwL{W;2t*|M7=IuV9Rx2TQAMh*p^Hoh*y5-}qNAc3iCs|_b|gZZH-y&0fSQDQ+GPwj zBZ)c#%EJ&>A4SYXQ?HS7F$qom-lNjgAHyGL>XrBdP5nKnhC9WL%uJCD)$FFe2eKRZ z3%0lP4n+p@nr_^76g(UVztj%1sF|ag`(s@zn` zOI`m!`;7V{w1WNXA^f&Gsuk=%F9V0(bY(M)npXdT+N^hEJ|I^VNyLi6;b2S-$l*KP z)~xgst&gL8AQAnnAL#wrx$pxoZU2Ki%A8;<4|jpk&wPBsI_9Fx!eC)hY%rg@10ElW z=O1&CXa5vM(G#5nSvVG)aqcUMdLDC8#s#k^ih>1)kt&4cM0|_!onT!>jqU;=;6E6W z;!xVSVmXdEXUOBOY&+@w|CDlk1tvlQli z7LtW42;iIrL4TC9d|!QXjh8-!Fuk&M8`Pn7Ebb5z@Cn1zle9M=#L6XbjL#PQP%Ox! zl1byEf`7abzC!TB8sTdMk6mlYa_B!g-w}=YvjqP{BYdvlM<(I1$WtcxQSds}yjn1P zcmtYeIV{qf1L}md4jcZEI_+d9Ph9AYQEptL)#Fob@@l zXW@q_Sa;n}Pvk{hv$OCcrv9QV^5vZDr1W4(0*|$`(~o7br=?hwUf_=f7&n1&SEaxS zQ0HYQYtM6F!ILpHg4vI1;{ZsfXr5neOxAIj+y)qyESMvl3Yh9-3`e}J5`Jn0!d>=) zpPgyDVD||G^=GESHBbIr`$!9cp#Fy73fKI<6miY*C@D+pA$&Ok!SCc!k3pyw-EMIb zd9J;ODD70;7~X(LY^zxYNyN-Kr)q=b80NCZ%v+fGUz{GV>~b_V?w-9>7fHMQXDJ+w zKTAP0{v*XAeT3^XuEF+AEHW|pvlPs$4FIM%9*o^k6w3<+h(1ffxcQYP7KvANHAir_ zvWRy;L1>lOOy96C?E*R!2nUnGHq|ldGM6+f_$QS$K~Gu;TwZZ37K|h?(SorBJ;U1$ z%0Ma#2W<2NMt|+H$X9);lhQ-Lnq&-bCJk@V5=sO@0`oJ~V2AxB)nLnxqayd#Uuwx&0wH+@Yl52~ z>2zH>Y}MkMAh%hTUgbX^{3+s(*B^fqzGS=M$HnB(L`vtY_YO4aQRAhhUTxo3 zT~FQZr7!SHZ{VKgn!D0VCpNYCI9S7d=J~-0E#;G3)NJK9TFDl*8m0&M<&KwmpGIL* z+y{dFfd|B#M+L&HJ79DV_8(@Sp44qe|sH~}_Z?Ee9wHBKz5StG%4XL*#h6P(9 z5Y*(YiS7-rC~^z2wLoP3J+>7ceFdw%DN_|~L3Z#5!+WyuVzK8Lo^oshLi3mq6=~Us_jrk5Be-}w;!r# zvt)sH2++BKi%(D9IaJXkA6%3?cTQDQCz+mz>?TcBR1M+KHtkGpHW#*{d1g#ilp~m? zDA~5DinjSI%J^ZbB3Cm=MLXjY932PR5^u7IoclMKhtuzqKDe zdfx*BK+{;NzS9n)W65$C$CBkP#F7u1S(J&#+_|$tSGnAU4(G>a76OPrFG#D|>pPs# z*z0*{klIq4`zjO=LxZvtFq}e7WDDtPt_h4IqDdHL-eNOX2Wzr;9vOX|pWJI=A23Zm`F74?y>sE@d!CN{T7ZkqeZb$2%}tpXxohB5RJ z2bXW5Iu=kjiOM*NftCal-HHNY4_itqLJ-(ew&@Hz3IyA2xMLk`q_+-s7g~ux@bosK z+SsCo3Asuj*zPoJH-YV6Y@+Xhv29SXL#w8p_A70Ruh?CzJBwGwy%G~z8*OdKFq#!miDFlj4!p88p;$1mk?hXI!IBL z@aLl7cs+j}i^7qWTpq!>*($2KAicqe>;Hq# z$e=&UlZJvU0J6}JY5LoO{=i^Ck7jhztlLmGtThklNKpiO^<*jyXTnAhiUtN_{`??~ zWYhtm*woMS@6%4!+3WjhCNl1-lPztrDzOhXr)f;>ahTLJZPnF?nf&BHKq|UGAYAB8 zxDcKk_`j3U95|#)LMLY8{>7=x6qqq|#ja$7zwc&WdfyFSDm~d4{CzGAwR;3ZGx?#5 zzK$u&Q1m1}(^%&z+<-%P&a6k#pkS^B*poe6?8zP!d*G=SWrpH>y=d2>NRVhU z@5y@cRE$5(v<^IkJ$Io8jdr_Zb3rG!9XdzX z`y3R5IkS>5+{vDD8oCC&#I|1Qt41JPVi6}qGt8BQi)~C<0}Rxt7R(CF%wlZN0m3h-+7XOA0u-q_n#u5M4VKg{-ulCycw7OPRkYK&`PVf$ef zLQMtPfHpWAZ}}3_d{|C4(k#e<`Sfy zX;EgOzl7+p7w!AfGtu%*iH>j)?mtsX^jK5W(3CR~0?^%aN`@kS<8 z$(l2zL`NW@zC`(XUF=1uE;_IojnjTg*8G{K4m+GqWnIo{se36Myfm&8aQa(%Qd4|O zm$Ym=I&)`KDsum*Ww#axUB_edVdqLkK0zAd1w^z4XHT(5-Vj>|6NsDmV^x2^Wwq8d zJqgnAanM~Da|B}WB?CGhwF!R(>K>!Eiw21Ua%sxqX zNFd5M4YieQ{TzQ@@b*0az+faF@BNF(aj%ySbQcrxtMJJK44UD>$PKDiDRk6?x=FchEzyeFcJwTMyf!XAz!& z&=LZ{vl6om;OTZ2y0%EAqV+=FB@k@2kb&2i3~_T3@dYCL^(F8a+V;QZ(F(ikv=g5! z5V`DxASfd|ErpgL5Inn#!uPYnD0`NWI|>BbyY-^#JtpJ)xbc?cHDlv2sr$7`4tAXF zPDf^(XuC7~rI(&M%fbWW`<#XSy;#w!R8-yC$vG9_lXrHdBFb1)spxa4^-ozw{jmXd zRu%h`vz)3p%c%-kwmaL>2U4?yr-MKguM#bK>mWt#1!;(wUB&S>(gzcJ3pp$hXulLqauNIFM0l z&XBG;8!^X&W4XDK%X<#2d*X56^PH?ndlyr|qF4QnM3z}IZ<;dR6cK}Hj_j}f1zS=1&J2>R_elj%iXYN7PJXz>j( ze_jGF@J(S{`*SQFELEmp%+}`|ETf!j$~^T z2sR-Q(*7{k2n*5>Q6g!9u*OIsj~9rvDaIOO1Zjv!tF7@*tufCYbll(%g@Xax-*BT) zeb2Wzg`>&h6hxCVTS+vj7T;Ve5c*sOeYOr#v__DI2z`>uvRTNr0+IF#q(#kc6{H~| z?Z1~LRvfX_j{c^rS~r_&t$sDDe_Jby+Qow5pg$fAkfWNPsHb+)kQ~+Q_2esqH^*oT zlb>vbZ6jt1a0{cHG+6EiPLNvX}b1!=-x3xv~M#Je%aKqN)tr5PHzW4s39CFQxDMd|#(#&e& zXq|4`bykxHF7uLmdTWbXM`<<}G4(6_$_vqKR-y%+2clnZj-Jh0q@4?#m|-^So^iu@ zR+53Eu}}o3$Oht$sGzMw*^y++u4v-OG^%vYCtAUhLg!gzqp$a4s4J8TP}rmapkUL_ z&$GybL0C>39ZBE9!V~7HSqFZ=87y*__THtv^*Fu0 zg0`une`*8hhPG8Z^H#3ui_;17_il3 zV0S*w1RdeDUjSLwR(I!#I6dpycbk_^%s_Wu19U%{==%INFSQ3sMqk^-f^`B>gZmK- za8p7!!_vdBT|zZnuruNw-^cbei}M^oFMGaru3sSG)M<(qg627UnxZTr{R&d2X^KA0 zuqflkX^MPpE%MwxO;PUoPP&MQ?nO5~V49*XZPAF`iD`-^9M!c`ArN{Vfv8c_6g>z& zH*%RNx@ykGlIjQ|hipsD=whu^DKNJpk3WxF|!GIlIYRSo+u8uOxT#n2VqGi0Gme&q5SNMf+ zGN+vhADT*@Uen@?=t-kgyQRr2RH2?E;;O*B_DUlnX5pm?I=!a|Ee_ zO)3>qACuQTHmOQT|1+dVK&mf!#`#fyI9!z1)t~UU<*g}ZI;M+15(owG`hc+nj)!9y zqAhaF<;W}~I_8@4jhCieXi;*YJ;s*V3v6L15-KcaPK+&1xCpkxtlu&Ief>qodOD4) zjmVGm5ItjI+gR;I79M+fh1aJp$g&748B}SZ0(#*m9~XI@f4`Z9&oTZ`TEU2eS%}pa zI3eLVE#YFs+O$k3rQc8UIEP*rTjYxZQ$q8ZX%3jkR{>l>aefftzf-@Wfk9fwVmBZ~ zT{5+EFkCcQS9g+E(w@H$H*$@b$hi+07p=lihEBH@)d92t8L$w;V zl>ahzmTPAy8Y4xm6o?{U4a@jvC>nN|MXf?J6wSEIqH|+2l+1#c3G;Y?@RV=hr`rCx z7is%Ip&CvRCF7J0U6lDrMArMlKS6kmqgqKrhDmwRai!#ogspy8wEplhgtWSg+E)Wlp zD3xdZd2m=NvW5u+IS2Ydr= zD+-Gv;eG<-2n(-)g*g(G2(?NesB29mIEW0%y>O=Q?Jy*yV);?}M z?K@!hK!5SjK5mbXbg-yRJdb{GM{2!9?6<8&U`K%{X(37qzn#)S{5C_VnF7I8@3#iq zp%In+c1;JidB&@XMhd-BAfzOa*KMyV`W|`pPvV-8#I=lT1`84uf=!kIDlfEnuS8z| zV9fs5N@Hx8<4v~Z5pByOY|D06iY{H02sUEDlvAJP}>UxSG}hC-BCZp|wK$0jR@duID+QY@15~b)3$A9B@R77e_|^(U zUPGZ$)9H$OfzN|*UM(>O2Sqt@KXA2-CsMQYctXc%mbp)4)d_^m(MINYmRa_yqUaIxw{)Qh-&2&EY*Y-0pAC{34+ zsEU7R9`Rt}`od^puw9JOrrKe3ai;3xOy%PA?$bQFKD5$Yy^;Hvj`gRN~uWq;qmCGvdqs-m=ubq!C0aoDxfMEQ;aA$K}* zN4VLK+!1aJS5^`ioL9Sc2f-%G0F_<)lxzOVo?QxxBLnnTn~vSah*4j%F)O{eK$NJ` zyQS7k9j-}=)|Fn2p6+2(a1h*!D!mvoO~1y{h0T$|W^o6eG)0Rqfg}Yu-kE%JnkDDI zMDkz4`M=MZA;iUrmT><27$x~H;ry?-)*=rs#4DBjJ2lQf8miy6(Ux0>oo)6B$=QFS zH=KOq{j=6f|GO6Z(IU~}1kpEMOtn7@Ogxx~7KI`&|244p!;Ve=w(Jnfu2@y-qzNB7q zvcSzE2v=KR@FZjCI0L+dEm5$wLGh%WvhJ*%`|Z_=QXRZtKi5DNj4E+w7$$@dK9-z2d}$ zriVLQ)U2h5ZZ9x>FqJcN%P#o(!=;r&gj6aJm6eXUU|e8%T^EZow!f^XANcEw)h`tB z#|G;!tlm6AlGL>)C6{HHQ_r+~4R8nds#%JB|IzYW2t+P<*cbDa4huVHDe5S+Y=PhjAcm;fin>CUr}k+@ zb(cDsQ~;8%Ows#Hn)0-wtoE9O(?}UL*oxi_vr^~GRumOd5mMbWThSHWENXSfY(*=( zSz5>SlJ91L$RF>>X=)#_P+x(d-iWU=azp-hNz^rooInAQm(FOv$!M4G?h^=6^%)%y z>>+`m>Wu0=Fynmeu^1TSN^FZWxpoa!ovJPE;F_H3aAf+4raBI;^wKHaF}E4U&-+$- zsY7?P;P$9NzQ%K=1NAxx*g9bOROkg&l?$*AdTk$(XN_Qw3Iuhci9CaPNaT5`2SSx+ z%4kK|l3fWl(cxmM=oR{!B;1rj`gG^oGqw3}JQ71k2z?R*0PIY81G3yt&sNl;rDh8Y z1Y0oOQB{61lMYK4C|jI*aCr|@@w6QoTv2^{jcf?{oNnVR#YvFn*<^O zUg>oYeMr$7K^h`PtMg#jWxOXb$z?LmFT@o`=qBRDG4vVHgzYw|XX}D#2VJN42d-1- zzW=W$=AWbVk;3Wf89h;wZoxQN4N58XyJ-6ToA_%PXSPHDvn{m-RdOBrfcP~xmpSVD-ECpW+pW}A~ zpQ!>)cph1@bZC{2po%7!f;Zd5wKYX`hHwoY1I5TDS8(zBTR*QUe6+RUF@f$*sPLaTE z!GWQH!2r!kfn{z(ieMygOaK1yV1nkdm|~<#z1m6XiNYw&OWAR&m)^P+8sq{QrB%#! zh*32f^%0}aybYC1zSNFRO4o|)U`4LI4U>zk*-l3EyA5~ET*G66?>X%VU`nrbVkvbu zW3~g6eI4{*bM9qxUUWO>LB29jLZr95MS=BYW!r{Z)&z!I!D#9vXzZQAhM2Dzon4Q9 zk@v|sdSi5oNLAHIw+LMA!jY=3Q*xx@6RugEHIIeeeK-(NDNT2dK%8rk0}+s!EBwoF z-wUE&YL^BwacH8ti5O}NvX91lARgz(0PChEF0IyCVJ3Z##^4wG~% z1)|Y4LoV2TG6j6@BeNCN3VFLgu(jr!6>#-lT34-J8zjVyb8C>0Z_>*VQU*v}rYLHD zhedqL=l|}oh;Pz+_zsKs_RP|*jc(Gb055LOd>enHPEoY;4ohDvi<5VV4Jqg7^5asv zi#(*SmG#`ozQEULR*3>_p#ZMY%mmThW{#qclD4}*6zU!>)NOMV)k(rT8S~5>MQPXT zVqOeUug_6*3q*O|o}(y7NVy<=H%C$IPK#PKnXBlHJ1x?;I!Z2E1w!L1EFv~oEhO=6 zdxl)wm<`t`UvL)+=1!ffD9}n5JS-5(=Rs9>`?-pS2zHo2P;os?#{cFj>IEt8yt#_X zgj_BVZ2eJ$(Q_5Oc$YjT8?w7Vl#8!`-aSH5;%-Yc%@J~MfnXblStn@v z!rc~Sls={Cb@1zpEGmU@jzH*K$rW_kSViMo>s%`Zg8C`c?le}>SCGORa&ck=Zhk0< z?Mtv>*pu-;DEHraijE4|eV&#IX zAiDx#hMXUfU9+)@ZszQU4pLMjWOsUfcD&;bLE~|-c~Hh$sSe)rhH^~@$({MMq9KxU zj6h`M!y@n9Pb>PF*>dJ7njqwAfnei3c!TCC>Ua-|kBp8A+1I9CpX(tFnLfd6Pcltl{T&43ArQV=p zS1CM@17nV03T#YiG8|n4aAS=Y-dbDRHn&^b?bh7dwyjNV+qP}nskhwPwqO1?lev@3 zWG0z=lXK7c&W8}W`rk#h*kL*%ibN$i-**TiL3&b6&`#RCo_vI^0<9iGHndH5^$|`i zmy&R|$}_?$m^jtkw2NN>h=*|$=o!)TB^asY%b@`L`&6KFjN908;v}Aj={3T8nVeJq z1hsZe4n>wa!f$NXD#5>ZhaziJ&l(?uS9G2i9k!i_Lk4c?-Exbs1ZLu^YJ%PbxwL@JPnxfn zdk0U3?)!TY(&C3K09Zj4B?=d7{Q{}4l0ts&2qktZNzJpzmX_54fY}*M==qI+#Rr-G z^SOW_()i*&*^pR_kErq^I)5U;83o3=(pGu%RD^TycjL{uu|Aq`ouhO@-`p<$P z@^ZI$-?JxzN3V^}4Un^W0$Tl!O+4r+5VQUg@oc7!dO7)2v?;aP3(WKZy>cna(P5J> zgQde1MB|en{HrO3l6mkFy%MpOh{?eY$FaWKbp!v3Lk-TxE`=rvu_$Jx{(G_Yl-pSJ zx;$8mSxX$%dJS>t(ND5q=#_rbB3fR2-*I3WL95$dqRPP%D(iQ#GWX}hD0SK*)BbLQ z}Q#GPV{J#;QS5-5eZ!ff$Jzpnb4$F4lcZOHE)>H8*2aoF05-(N=UXE4|;+` zr%iHZc0pYk1w|KZ8zY24F%I575NqwY3NAhZo=FK%$by8!*Uqb|hdOM8I`w7N&pjL+ zl4UV_(_X{ucwV~_s3&-Em-HXlx^eW+vh4;*3xoyVLLam3)4b|>^=6`X@hCyUI;rM! zO2*shyJOnV%_(2k9&uX_@BZj`-d*9f!5E2?jZz%Awd$vaH19j0_nvgf9}M3xp|n?L z94vKK3|Q9mHfWVCjW3p=AQsZ%bbU$8SZtqQp}AgRN>^A0P)x6 z7K1-V$5cd6(CtbD>5Q}%TvSpMRgHr2>IO%uUTRJ;KMAr9zQf#hk(u0P?MRK^%{%88 z$sS$bzK^{XN3pE{Y}_|udDXoBNib{m%NRgG%uRe1tcI`tv+EDa7A)CNL|=~>j%b5Y8(a@-v)39`s{RqljU*l zZU-bv@KrJJ=W5&h*AZTVyQP#UG;rj;w;U|9#p&a~VS^K>{a*!G35& zN(T*CQYeHh@hF&1A|&$y0CxJQ(Hz6TzXE1iq5?2c)~f6d@q_Y`XZKu$ZkOJhGMB?~ z1_)tTk?!df<=HN-Cx<_U>%A2WwpZ7=Ouu#=R^?G_K20H=u2k#cc)MJG`PDcG@f>>RoSo@S4e#0MoZd9(i3387yPIX^In3z5y!G>GG zhNtT&a3G22Tf)bR3a^4vGuuaO+%h+D>syY1KCqT$vHcFSI6Uq<;CX{2EwsEnfDr|@ zVuj4go1Y?HfGC}E$OfS38faPaA16RjnKRz&>>9>b$?%$~C9+;pZpEb2Mmc7x%rA>$ zk(q#EZAiu~{Rb$ux7F^wJb_J)gl$)R%4;v1fnpH6J0W^>cow7N5i|@tx6%3wxMSEL z1M=)R4rZ{5NJV;>T4Bs;6RrMQ%Fri5VuH5j-sHAjphdgh+O~$$Kv=!u$-{WdyE3;H z_4+M6M=>vR$x3|dE&Q>}kH`!|9ymN`3SY0V9;2r%oWbP6R)Y_Q)7ZI{F8MW9q|x#f z8muY$2&lAvnm;Ezu#4!_gDM2+5C~x*r9%x}NyMXg)h`cD1bYs6@d4bKy3eSk)XFTV zPHy1VL1Vlk#>x)Aeu`^J&;5n`4Cz~Gv}ZwlZS%c&)RO$7FM$SLaDKV=Yc*K5_l*S`lXkcE3r{w` z_c09IMz{ye#%j1LhK*&&5kbYM!Kym#NlB3`%37*ps`aEXUi7ljB( zVrJQ~rV!plgWW=k`I{jxP|0<(ZlthTFWKj8h^X<#@7SE7jn-(SXFDsryvxr>BIj=m zY*~a7Ju8kZ4em?dA$vTY2}z!J>bbhl@mQv1Qj^&}+n$OKxM&Hz1&*AlCS_1Y->`Q^ z7C3SC2%^BS+HQ^dg=mCpSzh__7VL&an*4?0N?04-3>L|5vH?KtW=XbDn53gw!-xt< zXFBj3+Cj`{2ws1k6}WP!9rwNW2f(PHR}&|dhxa6CrVE*@JrD9-JW8mbgTYb>Y^OVS zF2k4Jjo+>Mh+_tf3+XN0Ve;SPiQ}Ff(q{ylfD)j3HL+b2h1!&0Rt@QYXM%_puzs*h z4L@m`L*+9WQ7n1MnHl1J$29*fu+!}a3DxO>FYGL+c{QNr?@WS6gYj+Vjd)TNt!l@}ixt75HmJZ+nNUgTPH+XSY1XP8t zUVOH6yY`D>~5W>n#Gf1sB6z-vHo%Q$&KvQt>FZP&uPa#NwWVbrQ5 zXml9-$DVTUs8z86GA!iMma*%+amv=d_g~|&#O@~9hQ=cn6+2+gKXh301-%L5teOA` zcWbSoyK9H0uuG8sk`bIpkI}Q+g0E}V7|moLp-zN^)%JYH)fdn6S(zyq2|K~AQK)Gl z)u8Y7MmHJ+)47Pw)wM1dffcbzV)sas>?dW+Ph)_pQGnf5)gT^ep*BkjZb%cXrZqll zJXSlW&VIu5r`N*(+j!P`lYECIQ)8Q;j#<9o`|gh>ZnhyVAS*)drz9Xk@*}>hPH;x! zT!PoT%ra(@Au5fn(u`?UH`cZiUrn(yKiL zSM3*zEg*ak`?PbEg^;yX8#Y~h5aspq@Ye`omLDEVdtZg(_f+BpsPbE90MS#)KQdtP=mds6c+V52aV+l&X`6qUWjHePUAe0%IIsV{T{_{;6EZSIf2Z z-J{sL7Uo*tL1KJ4Hx)?#xFY~4c5%8nWD7_reYf#1<+^WSXZ#S%PF#M*MYevAg25Af z3UZv{0{-;T^Wj(A2+6t0LoAQSqLi#M=L^$Nk8Uc{%_5lfa1p*MWSa@Rm_m+flG69! z$Qh0vwaQPwgOiCB~^u}+MA#d94l zBJfIFLtaFiv$NifffIev!2RW{oLKM_(aL1$z;bv>U!0vsmBM`)2*@a?cGFm7c*fCu@CxjHS+1#R22b|3~Lwo@|* zq1b18nxF?YEDznRvCu8>o~c4b!-2?r%jN%(-l~ZZJyCaWakco8U#U?W$PYwQE?B1_KIJ2W+R-&5aMbveN*;FAq5Cp@}Vl zUvsbi9Sl+)+-SSml_K)b*ABzJP*Plqxi?b|;-)bEGWT@f_{!a77~Mb1q@R2Djed2N81Oyd3O@|VeB5R1dcRDV zqXhgQ*rFZ`gaB6pd!HM%Ph&RWkku^Ns6;phtzBUZOc`RLAu?fjGbb;XnD9lx1W%V@a0_=_#C3Y%+6* ztb|I%$7NEWcyjQfXbDZa3JmU?U{l5Y$EOoAcNEx`69dtB?RA^Y;4E?!o4fDp{04ii z@6O`*RE*gL$A8TkCp6f_foSi5)_2am|Fpw|p1-N`qNlcgHFyegLqC~E@BI#HQ)Ij^ z1kWE7IsVf&7yc4MylISM#%rgU!&K$=PfhMGK%G4pA~ot~p;&?1xzgQedVI|%lw{OPvbEYwKGz3mO5rl;}`c6{5QwUq&T{Q?gLeMNAV8+zhBi z0hKYhPdWZgqK8ao6%|SD{QZtsp_L^V9rfJqyt7k?no2Wthi$47knFSLl&7C{>kFc_ zHIyo^`{g3%z?#@ACAB}~FT*{9{@Yw*6-Qalv1ZR>zWHAPu;z$Fr=GmuWyL~CZK;v> zzq-!-I+RV-z6)2y{>;#23?6)g=dE7JAQOLOCf z^C{Go4e|@HU+Jc~Sso+ErC5(?6%=aO>fmM2$F?mJDtnCsl0HzDMZ{Qw$ZjBVJ2?~C z=FeO1Tqjb>MhbV098a}~ieiRXvPe0#h>mU(rIBSi@z^dCrOFGnh?Fjq_@mPN2#1RA zFMLr7nTf>eO}Gq<#OjMaWKY<2hQ9#Iy{aT=_W+00Yv@CnOXgiAf3Zg>JmyM^f3_+a zCl>=^Dj9}Jch_Rxcoxz>J*1&f%-#!>>=l~s(@5zRItR$sS^3PyFkKqh@hpaINiYXS#Nr{fTy5*jhlJ-vH&zNFjHnGg)PcNN56!BiWEdqls157 zhy*D6dMJCtey9TrQw{x6Mg|WTi2hF@5+cMykO-L8Ntt~7c8jM!9+%SV4ARBHzMkcM zpnY@jpk&bZ403bSNx(k$+eahSiP|!_V0;Imp0GPu7BxJ~c3G_ZWBKO9+qusRTW zUr6ps(3ky>iW~JMj&#r>sCS#fhu!`D&Y$HT!^l&SZHV=q=essrM-4has)_auMzCjPbeuTbXjZM+u6o2fBt z(#b1(GKpkeV^F+j{j}&60+uG$Cv^zL*A5i?=44h}#Pz?|%(HNS;yLg`z<8pHyA!1O zItCX%_~&T~D9FZ4vAcw!=oAI-kFjoOZ}%waqe!!4+b{fOgDvd72|)qhJS+_*0-rlB zf~-gea`=LBLGEKVR1s=L&C=Rc+g&fsvjEM9B99LjBAifr79;)h1w7W9WNuk~F)e$L zTx%P&Sf&l`b4YHn7b;P;7Y5tp7uLLgS7v;~-_~N!t%cvw^xx8bs|nIHSo8?w=a{qJ z(G6DD{Y>reaQtGd6D_X4LSzGojXG7R3yNNC@uaO%MuX$Vj)iamH5jn8u}*7f z;s+q;3WH-j+Nvl@W66ZG0+8T9Zp}lmg!K9;l;=;giyTTc)abp-4@_Iu_XbuSo{KkT z3`Wl6#?YsSo3USUOv?DZM+^nPk)Q-neFNrR&1EjV7szxsfrXKk#CA={)ZmUBnX#un zOE_mJcr3}RO!cwa&Cxba)Yi8ReUH)rDL$-YV4uPLuq93&4#WP`?y382k!S%4GGqO7 z!1ns|or(zB!42F0t`a@L$&X)?)U<-8sdE@}%Q~cVloCU5QgbF&VZ?xnr|H*uY_dJDmWX`Ep!xkW8_pfS=9v510d<)#uhJk<$fO%d(-^YjH>l;R z%I72SDJ!6h{{q`LHmM`H{lWSieqbEVu2#V%N3vfV>Vt7%0s;c}&kB}!V`2m!dW?Oj z=xZ=ndK#79W+Wc7@70wu?|-a3Mb}Q0w>yp7#0Xj3Y7S0Rpn#A<0h-!!1|sAo>2g8N zVxp%Cah=mzJ@J0UQ(dy-IHCP!T(Kmt|4cU9*_d2Gq6|Tfe~)5SGztEqRb|&TdH#QN z_p`*Dfa*mPRO^`K4!;Aod;0xb(kbJAH1|HnTW_G@QcnfMc4^emXiMpCis4pHVpa>0 z@d2wT&$jLb1g34^dD(r);LAxD;TWHM4Ly?y?>1>0DNubW5 zl(S}d*QqY6;1$!?w=V3(GC$ew#W2k#YrmA!IR$(>`GZnK&7-=y@2IfZS5D}ZCj=l1 zK5qFVq4oj>U_0^@YljYyQ!&}MPGnvkP7LeIIZ?7zAA*--#Tb_4AU1Gev}zDn?#55)yyHKOLO~tIsovL1}Ij2`5ex$WOIX(J| zA-cx&9)3tbANM*~iD#_mogIxF`YR&>$7h01v=j zQlLl_-#13>^p=U_{}8j`ma<;aB7x_}mKeZ3S27c=8^>^h0T^T=b3kU|V~wHwkok|Z zFG6{JG&+dD!n#Xp6nxwWi^d|R(VXC-w?5P%}&~ z&;1W8Z&-d9{8S_7W|EA#uC0%Y{n5`Gr>`G+V?c&Se@YAw12QE;grzz^t+Z5(wpQ4; z((xv%a9k1bDMqj6cYJkY&h8HD+;OGesvDGXXG*YZ>!!G(Lz(EkvO{^{fgP{QgRRRs z3Hb}_8&obtvhQOMHK+dn>je7%HFUmzar9boY2g@@Bo z3@T4kNj4TlS4CU3;F{c6>|0`2E3HLUnn28{h=a)+?=7c zY-t86qNbl%kXbRQlVB-qra*oG4iQKh;0cfSdl(?Wz@i*jAykq!EwWZOe($&+EL*yV zZ1O6=YI;ZqbjUN@f+1qz^q(d(>)B`j{zb}Nt%!Z}W6%f2qfaM5ki?TLjeGynO9Oi^ zPNchU2YC{FqSDQ%1SL(r||Y=O#K^Owj* z6?#fbZ7kLS=BnTCn=+=9$JUcCM>Yh65?KQtqUX}~FF;_mk#nBg)2|sxD3NMhMC(?0 z(0QMrTw%(jl5_%Tak(dDZ@fTaJ5u3H_I6Mdh6evf;8oZ^z9`)y3ri>9?@!4d40Wep zb^K9BOI!bPw|mSieUYS6XcKT=yTYBCht$aVVEGoqhAm8q)J@-te!VgP;u@h323VS# z|J$(}q=)c&*7HzYH5i2k>E=}Bm*`-l0>P6$(TkW5jl#v+lR%Gzsac7Ad)3bPuo+w7yT(` zy2R^>V@K$tnjioCs>Q<&W0!^;pnBTN`hk_2*$uF8xv(%zlu2BL{&NunjA-aPt2Z^naNuRJ_IRu{0vuZ5ZSNY5pR8+}v zXzu6mv2Ge`v8zRV20Ck1i;_K|`8=Lnze`;vjA9l`GiHJ0uLhy25`Xt1+-e)j zE`mmvw%=4(cCWCJZtK%7Zx@X=YW}Y6bmSpD)+Td#OE)R^_xjjtCDpV-R;jT4mC@a5 z8cS6RiUz+gA8<^1=QMN*3bO)MOOUt>**zA`zO#$4GL(?A;hdL*3xHfnjzrHRMzJ;j zlPgjjqxA0-hsM3Zi)m$195!(WuniS8v3#+R#$tpu)l+*W)>3zQQ|gW`|2(y~>e(yKu}K)?7WT7eBd+&PALxNOzQe_n~DI5_xo)qtm} zhyF=i)r0bK1i9Ij-`NHLQh#52h&|ney z*@m*P9XwG)E~=@+Il^xdl{&4MbW21)C`@h}RCDqj;DC`?O}^STt{xdw825*Y*ngJn zvFS3QE>sA%EtP-z?g(7G94`Bzor@u}MG;u_#4s@iiWTC02b$&=EUaiH0gdwuVysd5 zQ4A+2lCjIN={^TJtDgJW zBPEp>GN+Xn-$!bD@@8Y~U0x&{X{6taIAhaFg;%-c=-}YRAI#gkqJolTr*Ga`vGP*0 zuOyWAyI84mC4HyyQ?6n19Z}xFg3yO-Rxs{&V`{SBHwC&7Y-|ik=(&s}+HH^VNLn9# zIess~Lf-1uG4&rPQ};^nI)02SX^6j8#yMH^w?cJ{F72dYEF14WLL4eQNg7nzCC>s% z;DeTIS|$Ypu6D*nMeX8;+z3y1&($$81EwKb=*GvW)L7xv^yw)!#>e~;+J)(An#ZP4 zU@?an<5?=j=(jQd_V?e)4zMQ}RQDAHxihJ%(!tq_55Qp4dBMRE>z9?;L@~Jp_c3(+ zj12jg-l?eM8lLzg(oyX*B{<>~^2Q&j;CGC8gqzJ=OR$$^V6#^%33l00O+3+K2OjM| zvvIRZp+ai!LvQxD;R0!KReSQ&lxQ;2GumngZ-B(WcFg`Yowb5L8{YTp#?q1OO8mA5 zj*I2>)oMwbp>)+QVX1Xx77;Jk!W+Yn%#4TRMN{wvn1*Zq^TjjHUB5qIO`+a_xf}z| zbu4SbX&_4EDO7arq%0nkh_m+WF+#@%l3j=b+m6|@7&_Zvq5uaz!Je4H^#>TIpAFk` zZ}}#9edPF!7&-<_s4LxziSNlMR`X8b5!VkHdv?us1md>m&$jNiI04&$wu@It!T>>} z#b^N9YgA$kGzYWTRr}nvz+mm0B_pwJw+}j=>wmJ=Z@m&*h!pq_=mP0HkiG2anU~HZ zOk0q_BJ~~tm|BbbvNZdZxFFjp7T3N>ml znXeO%umc0sE%10d~bA(hzVR+`X*el|e4JQYfp5zvifI(`F*|k3t2U zd(nJrp7ElI(mO`a3SJ-!EEwS7zeC~E$I%z0K|=LFjDH|qO8e-^I3NT;?w29zQNh)6 zoVtFFnD|Pefp%9xQz2D>276UEO9qK?4^nZTW`F*qQG~k7iB)9eJuJkZR8z&s+8y3w z!DRm3Hc-dK!|EQ+Kk2k*u*%lk9@)0Koq5u8%GI-uf5UDIhwQik&tF>)OR$M@>i-R2 zRZ3RfKVu^{&1G3%kfntKjit*3qFACdq%`(D74_FxiXB{yb0L7Odx2a4nu6b>4?HYP z@gu3I@NH8g2Va=52VU-)f9?2Shl)S7Vl^9A@~{Q0-Q91RS+K+a0b$o@5OD-?>~VJ&z;) zDMp2a{=Bj@WPqHY=iLgqZq?{&bKFORe3h<8#&GkGX%SnOZL?U%Qq43YPYl?fBTHq) z>vMb86q5vpFNT@oJ;$&)CZ0G<~`pquz!eVk*bim_+yF>eGl=5 zAw%u)I)<~|EG98L_acj>%sd9?k4$MnIf=Zv{~v0pBo zbR`7%_6MJf1Bj?l_UT}S#ivYO zju?L@@hh*O?t5QqRh8Kk5s5qv=4G^Dr9_4$gOkELf9N1=*U_YP{1)pD)-i}sasPF+ z>72T1@=jr!2vWep@i-q5ZVq2~3>k^*25(7{Jd>mdf^~p8#~U~lGh5B*b86S{tshoV zKig-Mf}9~m=$xo`@wO?BtHFgs;#yW@LunQ#d42=!@`C#{{I1<{(BYu=+$tHet7OWx zk2G4zf9gd&q}SVhpUmoHoRG!4Jb*OD>73W#liHc8W`e8yz!rMeWx^uS{J z{Vs=u@DB5~W{pL!1%;RBBw>k=YbCYKdJ;JtAsw6qH4hk$P=3@(5(ee+^%HMVM%b-I z^4Z{XJZHrs$86)kQQ9EzUa5fR(unsOCQqTh)8iP`TFP^P^i)79j|JhEEovFM8q z|9PNyj{%kc3D*JV*6l=Gk}0u*JaA%G_poWcvo@>;o3wZB9%a{me>3mfRz@Z39;E-Y zs}dcHE-uLn=XFn35S1;8>05tA92r2a>4s`WpBRbuuK-dpsqpRm58)p?&i8L$|Nd2> zB7Yeu;k%UZ%>p~6csC3_#e=c0Zg#=Z-9(~|k$eg*Vs@nuEoky`_!b?P^n65bvM!?I zh3bMPo+~zo#~$M~umpJCRlhGi4>SIrCh@^!u}m9rOz;)V!^aJ*WDDr;iv}djq@DW5 zSn~Td(df>-?!rbhvwBoG$aYQ3%znaVddX`R&^*!FUcs++?xBfy`!~=OzM+Xn-i&U_ zVEBF}idRE3DBbgZ;xvc>{m&R~4Fr0nm663xzXo}|TwyD?uz%AR372U--2x(zWP#1j zQX~~G!W31wWi-fe>Coto30qPXi0FO>W(&#ZnrOWaRp9Ud|ma(u+?NCYm_0&*EAJRI8!^8+!0 zgL{=pW;pLII%J?}d5i@m)O@~cjY;sWo*U1toVU^#4?2jq|64iKrH2DSR01+lrbOG! zDdngYI~B}}+c2z6exJ8(i*|4A3m`i{t}H}XF4n<*04U^3(T?oueC0|6`5?r>rS}Bx z-rj1n4$hcNul0JA5ElG$xmRQHl^DFA1%^!6s#PBaH}VJj!1SHt%v-!ym0dO-{{eR7 z*l@Ixi$bQ|M4z0~H1j+~?LvAM1j(w8v^9W@F3c`m>ai@IM8X=UJpp^ot-^oX8`?*c%4xgl^w zBO`Pk;TE8sSDPC}yjQw^N9O~6c*6aRoYUje6?fF?&Z#+lK7TH_(J z_=>bD@tay8>=-Ay8Gp6tvV?&Laza|}-A2Wx#kFH^Mw{ExhUqiVEdUZ*_ zGYFr0#3&s$)+tTMiEP|9c%jR~ypITWdz#Gx3bI&rH{TTVSFu(w**K*x%&pEq(gHPq zRvU$%_&g;&A~B$)kKZD#Rle$}s=LY8#*oL5dB$(oKq@yW&U(>}p%#LMQJ+u(ocnAK z92Iu6l>t&a#S7$2XxA|QoHeoIx|hh>p9R|wv-dcJjs{ikD~9~byr*-4-4db;OrKo= z2HNRJKTeZp<~f{_^rgg(xUW|qJO^97PlS(6dRk^hNzzQLhXhT{w7)iOQZZTs$7GO! zFH)+$IP4F|8TSopV>#JX6Um^5qN#QRow`WHnY%H;CvC-X?}DfcCiH}@e=AM*95J!b z%C49{^1GNu(HV;#)9WVLET~*oOl0}Me5yHow=OR*2feZIMOob19|WOT%3xj$6N8S~ z4*ORK3xYZX%ypA-{G9C!-E9oriYbw?#}~!jbH*41{W?mx{)Zcg+0^6^`b$6MqYmzV z_};*<`218R{-K}`D67@ZXS0)#=8XVA2&Bpc>CmCrboYlCxMMdOiVVb!Qq*q(`i}as z9X5Zi(L9kDf?QEXh?>JVt@imU1=H`3sQf)s1u};=tm_Nh`=S)CGt9F$9IoMVZmx~Q zJL7t1$ZWIZ4>&d1K13U62vf75W*1bafrmN_D8U2Eq2H$#Tt4@ht;F5XZ#ML=$lQE8 zl}`_?E!U1LtT46@2(#&XxOuASnb+|xW0@rT&A zILgajA-d-g*}tME%9rmsq7hAM&N|32LZ(nBNspclAlEbMQKn%>b5DFGm50DP7E1%3 zQrPCh2(1>Sul0;-hUNDy!W_pt2`M1P_@*FMw16QvPK|* zA*aKBmk7;V+ii?=xuFj#S0)jQ2C&ljt?n#)3PywEEX-=Ek6kT3lnlL=^^rMAv zzVbCj9s&U8FzD%CtEB86%69Gd z+|{RPgr0i`EqRM9Q2k%N#EtG*DmK!qv6g6|=EHyqVrOH>W>TylOb^uB>4hmNv-pDPDc`lpbWMiRF$^9BYD z0NQ+N!ehLn&W)s;yn2ET0k{4?Jv2BXY>9~kVhQjD(#LA@PJn}vtGKw}HIbb|4Bz)N=_t4{x`G4WDN`6MHM=A0FDfT7e6&qouQuQA z875BimAUET?$3HFaGR+PX^)mnvmYQfV1jfOtu3qlURCw?u~suf2f5RQ|9ltZrkuB9 zkSUa9fh~D2F0{nFH^;ycHGsm)}iOQr3YIV<@cnM!*aD!8RlViB++Cr0<1;N=U)8=r-~^z#4lupP5g>3Bg~ z_~UGY$LVMa$DTP#2Ms zaDq^%BI4d_XV9J+YPnS595^Q4oi9dYu$$<8F=ykh`G?Sc-qD3-5sd?Gx1pm#Ycc42C<^BB)(aiuP2o zLma=ZwY+e!|2uFZ3bI6b!;EtgghrA}z=}<|hC&AIl^VSD2AI+GFnJ_jOh;&ys&6mq z`6hN}B%NNLW}Yt*32}G5Qfa$*R@_whVbx9qdKjTLVwGwY>B|@A>i0R1otQ@projsi zR*E;Jskm83bL`M^nnsmUo*(X1lSXct-dTHz)@9U{f|5Y9w6y-(z+k^<FX#>9S3Yq8w8|s3|C~gpLuWonvXO zNYL`LMoRIY9pQv~r{P~(YFwEsx850qOF#W7NtMksfSyBw1&bKuK3otqBMR!=zmY^V zAL11Q)zz8wEvF8HkYxjCYnJh4gAnf{;qMEYfiRkZ=F!j(33pzAxx8kefm#7Ha1}@Q zyIV^;eb!&W-IRI%q3HRqU5d#4sxNY#RNn0GkOe!aoIS7}j5+OaR9-=~1Pt7NJuXmN z0IU+2^EOWh+ms9L)Y?f(QcDacIVudY2HOtgWqt#LG+MK?^48W$AG38Kgf`t@2 z^NOfe=rZ~%b~$W{#}Hyqp?p(l)$6KaLI zq}q}0A|SQY9PyZIrF1x;B_3RwKU=NV^i6Q9My=_9ElSWpcWr@3jY%7_w>s745flXb zH3G7-GL7H>jhFd!d=grreDNdG3@#QtaxP)nOgs$z$j8_HClaCXzg+pxzEx*zEmMIE z6bSA9=(4=vpf%jZ!liEk&kZ_JZ5;eo+``b4loH|?K^{tQHt5!vy4y2khq-<4Gz_)Z zt{`_X)a!3dtVM7jZZQ}wL(u~QHx-KITV_)*s1@{B^GN)ig-wg=dq&btLXO$qrMG^x zOZA}8I@#FsTPd&&QUCk-vS)fhh8vdaq*jTAId3F`PozN+EG&Ew$C>Bhs%t<&oxu@E z+VxXw?MqJa{^z?gi?Gt$qs28B8y+Z@j|6QaQ=|CtzbrB_$S~bv6eHW=9C>ik(5tOH zt4C}5leNF{Wg(McyKg5-)tEXtv_8I&B5N*tw>BX7K8P!#vNjH6U$cfwWh)2@RpYzR z$HK00f_ir|3|8ksuuX9?#lGobnpV!&8*j?v*OK3W%dVbdC$zJVPt9>G%=b?EQR-ed zmWl9PfU0;-|Aoh(2V6UAM)Y0U0u37dOMVStWMPlkX)YS)yd)TfU^&Gl7O1hAyq0SriY_T?z zhfE)<{=QCKZQ0Sx)Sq4cf*qpK8V>m^MX!pOvudf2${l!V3bFNRNEFXmPob^A#y@oo z>+(u3xX7lMScx3ygdBLp409rE-cz^aoq!7H!io`UA`C%kvPlfGYExW=)ro0w=4mWM zwM7W7{Y_G2%Yv!`;*kZR5Zd}*SZ}4Dn|gQf_3y1;L-AP|AZr)$XfOI=u-6UFmx$z= z6^}~hMh8W(ubFKcoi7uIZcK+Cxwjx#b7C)-%pA9Hb8PPoIJyga23aDfj%vQr<_i3~ z#hD+TKLAo3+~m;4c6La4s)pmTgYS?@dNyHh^1wlQ*3!lnqXgVqP^%6?sjnT}_+oeR zIR)F72}!-JCXEXHcLmv)W0l&5F&;2O1HpPPS~uLYKRcPnld}a$1;g6!e1AR9^lR!e zPQ*MLfuRk>)nJwIa_@IA!*!DFd%C6<+?%_0H*P$K`L-8ynQyX1ht=&Y*c!GM{Lp1Q zZ7I?L20kn4*Bn389Ao!Kq07Qeb@o9V>R&@l5zg;~dC-XY?t$AV8t9NuJ&NYK)))4(uwKv)bdJUah`dSge9p^}@Xp^_gd3y##5V~F@OZSq1sU`C3TG+zyXCaUtC0TaBf z&4HN}D=ew4E;^lmv!x*vHW2Kttz(&Y1ec^Q0sUQIO^LwLEK?xiW^YZ^g^3D$K-mtGah|;1 zNQNxN-V`AJyIBTlfVHV)2RoC!8@+*|iS7pCrVe~Oy}g3_&itS7 zbrYBNF3X(_Z!Prr8E+!?v~9`?rd04g>AR3#V_eqVgyl>ZET4fNv0&djej0kgMXj{x zR;9s2K5ahVgrnx4$-*wmiNZT>tp1k+#+~;7X9nr&BuH0qV26Tb_*|AFl8x))>(Kd~ z?~aBgg@f%>s^G{BhkNEvf|B*W3PhGnXki z=Zm17+_5*~54E=l;(>cVnW}%3PYw5c#;G{n`Xt~tBmVToKLJ!S#6y(mW#tF$5n~(Xm-Cs3R>THs9uaPkT$z zQOR%ME}cK}{Tbh=xDhu-vhwlQ!n(?_zhKM*Bi2#RNj;QLxTu zX_IePR461T%TTZ^)fbg?y?IHS_8a_wdeeBG)SI*Wpk~@Fi@)TrL&Gk^`uDVTx&fT( zmB5(lMPOXnCtj9Qy(r7a`&g7AZ#ul+i-q5PeJtW~>L*b#L?CQyiP!C;oO+xu<&^u9 zvz&PBk|zl&1R`MyUXzcpi?dz-J7c6xH>Tvi?=-2>>*|hJuVjm?vFBUVT@0vX1JY&# zr6abIjcm8{d}oC~+f!gmZQSFOUY_d84(Ah<^9KMp`|m>$Nsl3#sWon-H%2@`kJ}T8GL#;BSPKqLq;;^BGJZ;?^%o*6%#q3Cd}Q*^3!Sb_l#5ONYfJOAb6Q<3+~F)7=Op${1A=qYC~#=^_gsE(*)z8+gP`6Pn@1`MuvSvdG)?67O=q#0k1w zjKzuVZz;-Ipo{4gyyOAp78b<$@DTgHk7(IfApFNvO`as^a)BV9jddWLiyMZF{?5EaCJb*Am|L!hmNf zTm8kw7O~M)V&lb7^XOZOtbUe`(P5J72!SvLwq|x*r075WEV0=yYYF&$RtWYt&9448!I%E-%(88X@RHfgtB& zdD;SXwqUCSg6d}~jyR8O)S`m|L7j;ezIPTYDioxXsPp~1;K>h`gaVg^VkNd0^FGrV zxVULtskk<+RH#i!|F#Itm43SF--&fJ8uo>*4aBl~hH-q>h{@Y;#N`pu<$ONn@*zJZ z(^LqH5?wz1%4DkhH_vj4LV=-y5app6{s|~vy`!rOeoCf_|DgK4?8a+Wi$VeNy>vL4 zJ_NFAAyx?jWdnR+eLF3hi>@M2b}cg&V#VrZyk1=2Drj8jCg_6y`h8mq#pGInp!^So zb=G)AHG(w6_!!fJ-J7&mw6x(^B-&S|_-HD4Fp`xV!@`3zxa8?wr}!i&a>0>V-g#^k zmJTRr^4wOZDE(C}c`P=e=GG}Xl53F%TX=awIfI$4*Xk54{9L=aP9TKejfC&lDf-VP zvX^hTY8_nXXd&O!DLe5R2bqOD^R}XKmspfJ_iaT_KpNDj7dFF=4}o@c6Y&OO!zCDL z@!@?#-z@ylZ8CJj8UI%O(S~BAUoKYNV?5s={Cfn#t9o?LKCBUb9EsLG6omK?Zb^Gh z`w%=?a0T%VoWmmVs6c2Wet7e3MM)B~UV$J(LguJ<6rJ7Q(i`rUPz@K%Z;d417G+kwqbLk%aAvx&IoTgU!#cTE(L6XQp$^9@TS?fiW4T}I zAaHSAd%(q&?E%-2pOWe0{-$kzz=aA&`VFiPy0H7^8i0<^R}_@z?s(>^7pA2y*RRPm z*v3$u3$^h!Afog(3wvOIMc&-k@KX4I&(DHpxi-Y*bo7aHBBH`C36 zZjHj}OkwM}Q0qBxg`yhK6Q9-1%wM5s+dzvlLn{>h2qtSHto=e4c0Ym0o%t86P;~00 z7G*A4p{Vbr7Fp#_DoPSICp*7!@|`CY-3<2iPb#|eQq)?mI*p>kQdoql^BhFsV(JJH zHBumQ*0qQ0W!meyn2?a`Wv(D}z2q8JC`rQtA)*c$p@ywNM%cy;3e|8qYS@YuiY5tp zia-c?8$!NXq3A0kq+F;@t_L}&SjbEv&lU(Fs}KNCphA#NqE72U6%UJH`0!vBYO_9w zQ7Nga1wtgMau%vG;=9LeMOBh^HxXssG#is(Zi5d`?OMLe%QCR|AlE}UC2^J?e#!ESpZfoLFR zw>k}9>40=`>+_{u7BJfGzkhk+#W20dz!Zg>2GRDc0EhNM@w{jkSDb7xLD9(R0!rLcnDGd z-@(#%R*Prm3xppXQD5&!$Wh-U2uIY{NYYw?5OEk8A?hQ@2tl_rr;b+l*dctX_%v^MhIRIFmCwE z?vU3&F9`-q^0V}PAv-9m4mg9e*3CDCYTfOBOQyYpP1jiKM%PB(#zh!q(lUNLj&QNs zyrAfy7(7&DLGuf--+%rKiq05oxZje6TkWL#F_J7S zf#8{h5I}!>+GWt&USq>spnLYQwJMh+uF4KXwAJm3b7r0$oe;e;P>eS|XYmZrFz0?_ zOhTerE%=i2tQNSfeBovK=U}uu|tuwP}EnPUZJgw#I-;!zLWyXKgTH+hQc>$XgCnG%%KCw8Y@*pZNX}ahqSi*iA4X8SCKE=3@>qU(8!15IiSg%^zoSCSNWm zz!nPCaA7!%Nk-<8GDR=yT(H*LB)Ob|KIQ*ueIyVJ;Efo@>a>GS>tEE?zsT0Fhjr;o zw83-P;BSE%2^0*T#|F2)0_D#JS1r*7_iz}Dq5dLqah*UI+!F@ledRY^Vd;~gwL&#q zFc{t@r>%3&}az#3&6dSk?Kl|taBby z)J{sSyFfS_CB_F0-iyo9Ccn2SKMDFW4ypk+$AWJ58{& z1%kTLrlv`;bP@<^1E|h|EgRLpbTl^EG1@Q%8I}~T%DzOnmOF~*D$nv{%D56s2{F8& z=)YJ;xa>;jG6l=h=Shu(h6G7-T`te9^@=apK=P~TY2w<%)AzCyHZ+;oLW0#US9Ul| zwAzIhAsE-@U7r3d0iqzkjP|?mRjd#27%qv@7kv33?J8WxibbNd&V@zqyRX8I|58-B zScJB*g!iw)35Xnn{F0HsfibVvO&KMlZlfl~b^b3rfv++(@iZ+L&n2U7<9%te3>0TgkGYYRA{e$JaWKdW7qGRX}7d{}n8Mxw6_z1|)u9a^m(+~gUTHon8 zMd4*`cHPhD$p$>9=;S<$GOu_}(fN56b-U>~MZ@u<;5kLN=UL?0Q>FOA#MwkNgWu9r zrRW(XI##7V8wLNsw=l*(py=$YEwcJPpeX4F zT_%=5aAkS8EXng|h2r}WC@u`6G8a`Sy7FqwOJmO|>LY1}3k2J5uw^!kW42q$luaJt zV8gI0b3&P-N5IA(p&q$hdxJ^%2=$VyE#i;hE0vXa7i#YFc8hh+S{mLQMB-~(C897F2jekFI<%!iRcN$ryuaH z-;*hD4VFZ~{iF-M)igYhr+NJF@->zoC*uYN(pJNPx|3P%X3RWbcof=Ba_J-x{4X&T zyMWyVJ3=6+JX^&s;MQv_%EU(CA-2rXVaE-N24cLqeQ6|O8;qNs*L`lf3wT$y@<&Afvjt__T(H~G2hT8q5tZ{boEkjogk`dUm)(7)!r zrA>dsHg3Kc)hH0gVfM?$eR-`I=NX0ogM!WLxc|K@e@bEaLfSx28_{> zK)`lL_$OyLyyWIrHN50rd@Px=hheMMPxO)-QJ+7|qKwgg1VjyQeLOi#BE4J+HwP;O z?7Cf|Q=>pAoQ<(AT=(`c8Jz7As^Nm>@E#A&4x{AtTIf;XJ0=j)dP5rKH0NB02*Ci! z5>wg=1lt8z<%gHNf;2?rrM<)ocM@`Ufk@jIX|Xz;E=WT}TIeY&;r)a>P$1G`dx!JQ z6{H~|t^`+I(-j7R}BDQp@qz0RVHGV}D^br$vT2ZAvl{h&ELQ|lGgNSXc!KR;Km z=*;Ua>h@v1qCwYNbmkWrtauL0Rn#bq*JBmey8AUnvkz;{cL{_)^3e|0zoux?^%iAq ze2s@IoKaexn>mh{jGi?atxC)&R~X|Nc_o9P0uwSi>dff=iHgFK{wOk%8F> zS%>Ns1w~wqK)CF5H&NCp7++dgWD5()$+8eEdh&Im+65&#a*e7vKkM?>ZEdBfT*)e5 zKir~hZ1e}BIJrlaN!U=H0QE*pRJ2d>J|+DCH&AJiCVuF5J z#fQ50-iXz}IyYsEqa}C-gI+^>-|-8w-X5N@i_914mVAZh6rtS$H!?4g&Y%z zv=5rJK|va#BP|-SIYQne5NRKUv1r6b-(u-@tV*a(F0^BrXve1C0(V*H>GliX2#3ux zAq}ouc#F91^F-I>o#47-BBq~1jP1Jix5iy}Lc=vsLX}K-HYJ4V9K6^*Qb07P#ch`+ zQ@>kHoX%;1srtis*8Mll;q|=5xBf-LbxpwMw=*&slpoUtsOv+e$*7Vu+5^E{Au2he zo+B+X8C7ydSC9OQjBY>;v2!vxr2XN{smkP3#W}?wnRBY*oHpWFOse9XzDvkS&)sA+ z>%+>mEkEbDGoP9kJe;j*f#GbL&os6*Ezm$yM%Yhl_~&3BmLT78Q$}&1n3lIdXL=`a z4Zs!ATddY|z@)Fhx`LnTIN|#~gpsbqIe+M^`wB!YtHP%XkdV!$UA)@j2f#0CAg@lHDz`^zo zgq#qeSdO+`+i1JEKM|egveu&|7i3ri^U}bFr*+SN3Q>Q98o9 z^C%hLrlYKKN13uq=T;bxtBLnmwB-HNg$>plXZF{qip-K}WrC~T5X=25CWvz8%%t#rClgc*-k{stF9 z);EFZhAI7%>5#4Q4R^pfu}u^F_y#DKj5uZfFd!rTDFbeiEKv!86~%i&D`| z$CD}5Z&43D2e_R@_J~DkMEB##L|)vL&M2ZgFqrD?M=;mp4cX)9&Z)qHY6MDe5sAvo_3s7YbjUK=4l`GMlI# zZ29C$cD(q3ZJsMQ`BkeFWx92C*#eOqD=3-4RflCi(J`(aYPrP z^cNSHx^odv3gSYdfnr}iAAq!zr?)g>&^onuW#htjdz_x%1C8)Ol#D<2P`?rC7tmEO z&U0IWu#%{5dqqIGKo=L!gD%HLV01z>KNcJ{A>=C#To#I**V|@J_c(K(<8oz#hiHy# zY`RfK?tcB7qn^#e^4bv#3X?q%=lPavMQC@cRj-IZC`B5DVppOYL0i z6LMeRa-Dl8?l>>d=~(`2{Clk}|5cG+Tr`Rny4ItGYw_Lylh4Rx zr?pE=eoM4W&hIVG?`u21<(kI%)jIOii{_FdkmK(Doqv#uLsVJVH)^3zXaEJq%pKv_vN94X1SMb~?sS?qM_@Y(4?{G1(> zNZ-q%d_jwP_+qgLedt>9Cx1`?H|qGs3ZILWCDA~PesrC5MKX~!Zk3{uVnSFT!X4|J znp!2;8iAmG!Bqum4+{2}Ku{0b)JdZGOo5>O3TnK8u+5eEx-m9A)oJu$Q;dh%OzNJ& zMjmzz`!kuWg+YtD7l-KxyCiK}(CN?Oa3sKW$VhBRko=ZgyYnq~&9G$pF)=BpGyLhT zof@royQpRGc4edUY;xSVXdo6XDzHfnHVYha-0i|kDUZ9cg^*Q@yW%0DLSW~ZIzRrWi@N=^O3{Bx5?b1@FI1F2pfHd>rk8DQ-bu#nbRz2O;(IxL z;l)uUSlx*h5cO@kFqx*7pnD3_f6#=lE3wG4=5|GCYuto{7b4`XKUOJPBYxa3&`>ke zS1bDF?-c57D^j1J(B-adbaHe#RLBi$6|RH|g<)qEC@9H;ae52Wh!}aY)3&j$?D0fn zF@C&1Y*F?Ed06$gg08bAG$lwXV@YLgE>EUkMN%2W=8XLdm97mID&ty9`m7mJ(9ubSK~@}_dz?2Y`bcKN=rEvF$t-u+tM z{VZWXMC+%YCjOuDuCoP&PjxCdOADOE3cfA^XR-JLe=D%REhOzUr!eeUj>ncIyw^3V zwR?h}u8;nmwf(QNrPL%ydseIVEUWf_NPCv0t@&GN=W)>*L3yV;)q7D3e33=`E&}ya z@7Uj|cOI995!93*@D(la6&6?&)As78s=pUFz?M?c&Z*u?Eo~)BdtapKr`Dx^r`~{) z(6~S(W{b*c?-aR7i`>K_FD=yy>ZeHwBDpEj-#juBDiS1prD~z9D!iNON|`6>4v33Sg_F514{w2 zo9@Bed-6n+(~t%3DG#2AL!?s`y@B=aWzq)=#wH!57$r-{PS-Os}&CyWOrm_y&9M#@j9O*1zi}{pC?w z%qeDpC+Lfs=Zhxc$kJNGH`Ox7f=P%K&!1Bi<*n7e?`3?y1vK{k0kP+xKsdEG90T7s zfsK7uE)s@<1fQL9hxqLBJ5Y=$pE>E;+W7(@3Og;1@~IS(VSyJD$_Kmfjgr@;#`h#{ z^a@0(=Dz2^jRaz)k&x~7M&Wny;A>}+Dp%}NCQ9G0Hnc8}t*&o;$BKjlg>;EOwo;&cm;B~z7 zc7Kb-LzAi7ofc7QO*fsSNnpk}dZ0yV%M^MKWM_`?>`XQzyX&+%yT|aN8=}WrAiEiN zIh)nYvw7%tRZ;NLNTei^)sumpKAa$DOAFVzcZPF_wzRmvWeU&abI7%&1%`(Xg)+~+ zm95X}8RZjyMo80HPNiOM$;Z84ZV9DU-enOFeqZKyb?nDuG)(kzODvVQD%Y*jLVk_O z+2y#{1b;=Twq-`E6dF~IAj8UZTgjrGKm^%;aE28t(^JbW%EWit=G*-JgfCYh_;U^a zCd2N;&~tLfpiaZgLOaSz*g(%ojdzNaO^NESl&v_uqPXO=!-`)#MMjPx2jXiKBgw+>Ik)bd24z;ckm4^+1NsYM5l1;ir3Bq3Jyc0W}aXIz$h&LSAhq zqqeh)vL#uLK$u);OdbS2>yvjCRSUUBAlN2>t<_pZMPU0ohYX7NiUYPIax=A-Iuf7L zU!Z-i6$*XkWQ)!%8AGn;_$Y6}?~H;l&25F2@HOORcS4E2AB3!AVsA34$Z}>luN7~^ zzr>=f9g{8U9;R10`?Mo?$|xB_^IBckJDJ*L-(yjRe#gN4R+xn4-(%t3H?Oqf7aF7% ziYb!>X87qPcK9pzApWdp6zvmIqd<5KA5;BHXJPjL-*6x_pu`{Wa~MT!?bEWIR&3L5 z+s1k~+#^%@ZLLndCz*Oqu_$|-FA^r&)(W*{{}dBv+q7RZM*FFONv)=$R|$*`MDU%2 z|BfDwrJ4^~VSykofs}_8o>oP@lPL;(G*}Wd?4L0E+fyy_q^(udxK4K-4T`AS1wc%- zs9XQFiqyRpb-QA%q73}FeyyU*@MF|kMZtS9whFCPl(SwZ{{W#ibFHF>!1L%@MQ`B8 zGiw#?#*dfRDmr?vr9(LVZ0(%x0^#L1(Dk5}cf1dw?m=r%F2X;AaM1n@8s}Rtq)$LX zzP|e`;(Vvz2lCyDAIP`eG|6|iD}o z$IUuqW9JepjFjll4)(I;HzefqqUGuyrFj-wck7#k#G`#NU(jYxKUW*4tN2oU#+GO) zryDrka)dr7lj>>`RU{-{&Uw~xTI-&K#1ZY333dv~ab~;Qa%ICOyDc;tA5CY&-?Pxp zw4H&6e0twPrT!C%g+B%(Yaqw ztpQ52G|9d)7Yk75pDAP=l%LUVw*kC6IbeK3;(?7e%=3V z^}x0}?mxk&oP{{nWI(;Na=KLB^Kd)(5lgTZ-;Z`MpFU#oX$SEb4a;M@GiO?4Encf= zm*~1rAng1!35!j37_o)i+uIqICuIDC$CX_iEr=5R;6Y*BG*cVd#96bEhxq3mGqsUF zvWit;^OxjDe9=Ps$piOmora0@5dxu8GQS+@FLV-@7>)9JJ6-a#$JL!R`-K&=R?oDk zdxpvLXAk;>wBLX%2-7{uaDBU84`|2TlZ+xtFT`UMR6J%r`+x+W(nWC$gX|5bqS z0EyjqYq70f(|JxctCD42Rh7(MK+pIbYxAC+on+&cgvjoEynhKNQAnR2NHa78iRc3G z;0TaDY~0q61dbsh=kPo3VSkCy+31@7Pp8hilch4Uo~<90!QF0l)1`RC;7&hIdeAgq zyT!m@0N*-oOvcjhn-A*O@>3%d-@~IRB8^IY$Rd1xxF;@SbVZk?AF=4y6-!g(&D(&h zgGB|rE%kmf8nm2;&;xB)tEfU6>l+Y6o&#$Y)eB|}-cji>m0(jzaR3{|Eg1nq@#M8me!n98y_5}dGk|r{hiMaJLeIpzw=Yj%w2`YV1(px>?5ZB&exuf z=T{WHWK>BcmN&o_pPP_PLyF$c9CCz|fsK2VRi{TS{od#QJc{&oImFGm zE)*^b*@mR|a~jm2=Z^hh(5Ods$X*InpLi4}5;o&>O_&Bjy*h|z?J7itOO1A)FbPpo zFC_jbJ+yN_Cfc3#STpUy;_On^&S;hMFQ;}lvvxP{C`+axkKrXiP^gG*MT>h^qChQf za3F?Rg^`l`Z>O{(-l{Cx>PexEkC}KXLQzphfNuMEXfl2G7NvHKNO zLL%;Y8Hq8Gh9~N6nz06-9PLT83zO+Yd)TZex~zNmL9S+3e|9#j)IRY4|@tl<#~FYqMEJR zWPTIO%5{qF2hFn)UAK^4fnDFPQ?%zPi!y&+r|7Ioi#&twQk1?;r#=H407Z8x3Rkk^ zKi4VB6WVJ?e&%{bZ&g~y!_G7uP9kwl%v>S@X19r!4%BjRN?jgrml=s2zW)+?GLtm_5B zt8ZcjeCc|O6jrTQR4F7UOA}bW0!!8*{3WtdHz+zRtle|9B9Mj`#$@%~V5A-W3l^N; zNR-CfrZs3U*F#&+fDMXr1+x+u&M{YV8ZQtzZiRC390v-?$?_LD!W&_czY!AH10Hyw zO3=-%?nWCF?R*+V&%>@J$?2FtD6<>B!H&RTK^h_=69X>NeMN)h@gBkkt?|zY66`Yc zd>=j3!?<<5ORwm`=vVjw$H}1(mels z);W{SJi8d|xKlnl5{MRZtEyW%lUw-!XRh1K( zxKSXo=!Y!Oh&{uapbJx^e@=$#;xOX2tXFwUBnF) z2(kDK7mK~-Ig7H;P7=$^6dorpWM-kk#6LTx?1C-|<>wWJ29%WYK+P0x`lZf7UW-CT z6wz8P=hEja%9enowOqIhpTnmb0{OJgy&)xqKG%<%XUTNQ^A>r2z#{ApU64cI zzD}bw$HpzzhqUSBJ&8q=8yDI09c_ zbj*5=He*4BPnh*!rq)ixyoF!J=-@ zzpm(w7cAWY4;2Z+1foai*-AR#-E0j8IEFmj!S;_R9p5)WuK;UiG{`D5wI8~R$473HmBMm zE~~eyEjsfGE;^Lhp=wmy)e9BXNZJ6>ezs82#V=aat;H*fZh29>RU?76NFcm*f$`Se zFQV+>Ekmw#u>F(0#i7x6koLq7fhf`fRBsGa^eCmDkPHh7))^X}A0{hm5`_!-umFN( ziX@vY5R!{buuK!AAws4OmUtOPOQYdH$UoRFt@10huCB^~L>P2q#k2@3Th%X`;wgH{ zn}VaHf$^~$`32}voOu5JMT;_`v665&;HPmIxL!OLD>Kmvq5LbU4D*s{HezrfMB_1) zzF?k(Z_|#C_NE;?D*j-;(PMBR)SGtl8Q!#)IOBl-67bS49{yhb5;}qRaGbuR zjJ|L|mB}xmu_OID@a_`SIRfFN+t6cQ{h*?%_uS6OIcD+=@41-=hxmCWzx*Yc$>+SE zz>M@G-`C8mQO_7tA()+v!KY2?1*6ZuGUhT}V(SG$r6b^H>Mp_V69_8aok12(j?(2i ztydtZcx8r8Yp`t{)ZZc5Q?)@+lW1};#wW`*C`wvjQMXka6kWAIUY=ek3Bv-hauLI< z3Tqnof(vE`h3e$e>l^$A_FY2WFAzdXjgW=~;?DFvPIo#YJQ9eO76%5~ft_=uHdzPw zWHd{>fKQX%2y5$Q9P`=3K{cJv#s2&<8VnR>)gE0w7?X4HG$droUZD>(PYPZz24}Qi zD9&g*RA<~tAnbrM*az(d=_ERQP%h*Ofk08||oO@@a7r8W% za3Kivu~5P3&_fMYd&dYnISckXKlm4*$9-`I}IG!A{s*Bcb&F)3}W;uo=l^f)2} z+o93dWIOaR{J?hTGW@`H=sx_wcIcm+{PN|B(mvDy_5zaMy~%Ay4^}tEJ*MdLH!VG1OP9QI z1R~vkG4+ctlXn-AVKKd`StfUw%cp#TCHZ4)3+u%OZRsU0glB*BZpF|5!^kPZFKqUV0eIVX z<~bV`y|f4=Z0E}UjfU;)!U@?HIc_E~XDDcM)Ff~gH|HExQWCUz(yn(JGuD+oj%X}~ zHz|vqjU;C=I#}Y1*(t)qohc`9$_b0FNT!~Pu~13V`55lR#TNCzhh^?_Ey1|T+lZJh zq-R}VUA@?%)Z81ipge&YMO4LUe}@W`8Wen`z>Fe#nd!Zj;NaLk?#&i5reFzN*vyCd zCHQdUcw6D71jTU9r&}<7Ra?tYSH7FjTTz38x6Aj~+1aUL9S`Z9!&9Rr1d@&3}{@C_vqf7BM9 zd#f{(nfkr0Gr2pzXqiRXfsmi%c`TlbeUULh6sM-p{AG@gx86$ZclN99%qP{IrxBpC1H2N(|zpH4ysJ&SrMx{@*CCdDG zqN3@@&ANP&qCG<1ClG8;f-O8r(L%5h@m!@z7>)`A;|lcJ)`J@r4IQZ!pCk~}4T#C- zH!9lpmPOrO-l*v8It)+1Um23K zQBkH~a|D8#ftmcajaaC&=FV0$RA|Emf+w9NAjhC!!vaC=3~FLX$WpEeh>C>3*y3_^ z&u|357G9{eHo+{rTJoU%0y|}nU)!ZaYnj{8H9iMe?oGNNNW%XHA>;|cudGEldmoHmlW7CYYLsRS1CCZu}WQ%pR6ZP zoA_(#C%ujR@i1*x*ibIPQ!h6j7uxuVJE$O5}3XTF2=z^RJv=DaWj z-cz70KOhi!;X~lO)VfcQhKO7+_;n6dBKjrF+MmXcRb#YNUDFOI?!OPPpQ_){L*CyN z#=PI-sW?o3YWJ?t?k(p#8$(v$g!dQD{?D|1fsYA zUyx&SRwSgDpXACYpjY`$&%@0#bR_34p_iRFTWpOm>Lu8 z6oH`L3Tiz5e8FIW&+oS_&kZ|0Fr$TD2AjceYFoNWRt{@L-Wq|Bdq3(emQh`+W%V!p zQ(fP$=c8fVm^Mg)MuAB95FF>|Wrbu|P~Q@2DK-Q1MD)Wj#zND1ute*9OdupbYU*&4 zAPo^Ro7dsdC6P;g`GtdR571aGQAg0?7P<~EZh<;{#cH&MpSt;E>r!^ki=$*@2nr;%*Z8e!Tc5OS6qlhzB;5Fz33OuELGAF$1-N-%4sn6;A4dU3Ty z8JIV&YEj%ah4$ew7zh=_3TYL;?yKz@%wu485v^v-P+;gY?TIz)iO1G(X#a1sqIAh@ zpg@?g3Rm_@H!E7U#-eWbY*us#2|N#NR+RIZ4w*FwriGgoWxZ=r=90~deD6Z2-E0;F zVuMTkfl;Lem--^X3AR;D5oe(t#N{uXQ#z>cdu0Sd^~O)j;nM9d3bj;=RDs z15-r5GUlwcSgiQmO&MYOtp%oi_pHU168+91mam0MWX)NnXrCCkUmy%U0weGqhnv>P zdmQdqhZq5WchNIPAozbX{8fg3HS=RZbg1x+7YP18z>mezFTob~4Ug9_hPJ&?GtTLi z2V3&OdUZ?m`S-3zkw;4Mqx4}*xZxcj3bb$O&6b$uA6YN+Uz|_gr5Bp*g=cy8j)b2= zw3mhS*?@BsV`wj56uV&q%wDlskvFOpX(te7bikN&|7MIvnWvM`G6jOCql2emj-p(l zjb+-AgDE%?p172l(G@=VpVQYv`6Si1tQZv=sI5C zqNpeMthz0V`oy%mh6n`PzrePAi=tv?`w*52xlkb3E`dKiKWtG{^|`K7*CS!dRz(#X zE%KbYRS~|62i`5r3k>boyjy|MO?IqQ>o7$i5)6l{<4yK>AsH65MG37Lb_J&iQ-wgt zxe?h|(N)y?q8!ohX4vAbR`ktAi@LR~R+P2LqHdk56%E~l zU_+O1SY#zh^h2f(hrDob;b=1=bZ*{$<`)L?#}wz|tcP8N!e!bWx>xwMr7J_zr?f)l zDcNL^b>3D*xFdyO?jA?pQ`lO}xea#zUs4YU_OL+c&1vIRy!e_xG!PjVxNM-UP-B8( z54Dn6(?hMWW$a9vUs7CL65{s3q)D6TG^(Z4~=lqHww}ak&!NXt}i1c#_%?8t}i!jmikg5lz4rq7mThi@$!lAYs-yWYbI&q zb<7-ZrN>mq*@Lq-C%*joI5)5BfyHqKqT_ra^5iY?BmaN$ZARWnoNS?pMr^^03M1&{ zr(3VoCDUAR-xx;*Z$wLN1tpwf0?#Aj)jJ|j-GhqiMN}$|*}nInqW)Vg zvc7##QG+B%x?MXA2|HooqU%GbJ3JX|`ohh4oYUgNhf|=4rlo(W^*#xbF~^>_RVIpu zz{LCEdBT$f9-I3BT(M1tw=*UC*#co4HpbaW_katPxI(B-F6}TrC2&;8Si(e0 z03r4c`SNYz>2lFVCv-&6=tv+iF%T;Zl$At+emjIJ?$Dmmq4ccf^5R=?8AY^`m!W^$ zh8nPNf}+h5bBzL#eR@*DT2h0M3=5_f32RA^Gf*-dCJ=H4CdwHiB*W5NPP|&=`=X;G zCB?T6j^_J9IOR|pwJplI+iBTHay9ZJz8ZOT1I}gz{a8Gwk1QOmnFdrSQ073bXu#mC zG(SHOjnc;!cEo;eu!uh!(kME1x>M%@GlGZ!d^V*0cKNbK-gdpt-6kjni^huv5RL`W^zE+O^lc1zc_u&~%+)P9FVK#fpz1aRnN-ldi7 zD-g=sp_6A4HYOz0nxq}jn)r<-({_j^E07FlMskJ4*5nsZ5YXj9acDBBTx&8@V1gz+ zcN!BC5+;^+tfatE*rt&4CTqDmjDMBF{*?=R^iD@%|H_3u9Vk=S8dbm3oU-_pOPaim z2i%kqp>n=>`O8jJ1bi?iC~-AkAWYx`{wVnlyQJi=-i4BnZB>->wRYiTlsuNLMt-eJ zzAHB8v4l7cyopxg?Q;lbV^;;<216wFJ?_yh{^K5uxL@AouJ&ZEF4Zwv;%4jF2d=Xf0`nPUk zWFsTjGqS=Uzh&gNjBGH-CPub?AMSxW(+_I*yo3mY3wymUE-VBOw~X!X(WTo-AY8rx zivS#EVUe&H7#N2Kc&}p#M4(f8IR&7;u zOt5KFw7y9DKI~evRnfCBhflcGh{BDWaM=@zawI{aKqTZ##?0R+{5zQ+8Py42gFx_q zX);=cj9A+}LT(fYwr^PoY)O-tY9|oXCQuXFSBXTkHYdF4_i77ueS5;=>MpHmCC8uj z#rqa@mjKi#Z>K%ZUFCdVbnvi}8}OxuP{E**$lw9C%=!e`)8!iNbiPJ=?H(QLMAMTI z=;QDl3HYKVA)sElESa9zW0B|Srxky~4L0Fsbmp3;71colYs*$e!zBy9K(M9a=HP*? ziayz6QRbnoirgPql==HsMQ440Dxe~zp;%FIKy=U3$2qUf_l0bgtL}5UWuB-$k5wP? zfl+-Pt3C$Lp+FhQ^MfB)#0>^MBnO4gX(4x&%D-#YV7$)yaBJk19ES@;!faecMOC6K ze7mYy(=vEaq@-As5BP1B!_%CqzbUG}$*S*yswF``(VMLLFL-9v^>c?0P22fqGI~j^ zT7#(6NhCwHzAcFUMl~F)9ab2)frlqXoAT*StsY5n*@gW`%9WQT)4%MTA4xfNJ$7>n z!t^NL*o=K>k;k)5QO$R{h2guVJ+>)Y_#xWzDUT?cBdRSD2o+~T#mYx8K8zQ+Q#q*$ zMzypVP8A0QJ(#;`x{l`-tO`x3%BK^HO7XfiTE zqt{a$TL4bi3?5)}HYB9jH+X=ZUG7Y$ z>vn3_?c^Np+h>tA{}Dw^B0NcWk;!gkQvZLH`VB>~!fOM?K0Al1ShBu*(I8&-4^e`-(F?_u&gJb6{f z1)s>Ux{0iPGcm>~p%Xd)s(}-KEU#x#f7_AQ13`C(aQ3)+n_r`8oeo5fOI(_%gyIP4 zNm#K$pYfuNQ^PW@nk{AeFYexC3-PwoJ6?NhYjxNxgPdy)e<)gxLdH| zL+sV=M5f)T$#1_JMEgKf5Ue)HFCUNLfDIy`wp!x5JMN@MC@QIOkk}XS$wzu2$e;I< z;nj-bzzotvG5`4>Lfrc1pIS-P1%$wF4TLSG+3z_`C8}gMLXf&1?oCXy9NP%v$v^xU z6otv&GD&7n$eUUcPLw%myHKT0y+fp1!Ks>;E5j*`H;6oRX!C~h1FO1?-*Mrg?!A{I z8XKR%dY307)|#6AiT@4M!`7G)<>0C0R>C7d-D^q@enk*x!lHPKA#AGC)L}UW)0t;V zkx}&j=&cESLL$@ppi*73eW;wiP7a_mI50#KBF=S#2PkUw!vrjYBdu_LP>dd&wl9%~ z)DY1V=pOh{<@?xLH!}+^Ka}4dN`c1)zDGP|Pn>gfHVnVGhdQ(|U+y`*3_ds2nG_Da zQ6yMfMZgxGjqbZ@H|q$@(!lyfx2_F@g@Qcnm!prv3B^-mcE$_i%0lm~{ES;VNWU67 zg8Uox4wc;OdM8oIt7a{6{O7)b+oP zaXco2L%1TCRg)8^CaV7bFfiR4N?S`02i*&tK<-7D@QBZ0Dq=r0xS12~St{9R zr`#ry*!fKpD2(-2JL>HH%%q`Vqa=!k8l1_lJjzNIv&l)*=u%J#IIdA5gZRP3BJWhF zl+rJWSMsY#7S9zXq|b3dM*GrEt-)|OIgxd?69Tu<`a5oqD|j@F z1x-+{htfXqgE9yQ2QEwOs=t}x{ol%h(psVGxSl7+^)_E79kJcz*`y1DEKU?XdGnC^$A4SN! zo-|p##xG6Co;M4yipTlo+4FhP2;*>%MU>-j0*Nw0G*5g=R-KV-3W@<0EJk_~n=jeZ z=8r`>k!|lb5$!D@&x3<~R;fDe*NsB*?o7yifg-Vy?NQ$m9*YJt{Qn1%WhXWIoK zc7fz-{^KX1vVVUKqiKQYd0?*M%)hTT=E15UKjK^Nlw(>3f&U;~GA4psCm#4Ptx%i( z1OK9jh^@}DIEYKvWyDL89!I^W!S%)LW;8i%EdlVb)(~Q!H zCCJJ@lM-z{Vh5QEl@pv*d`0rcIw9T>`)?6;#o}Zcfj!$KRv}y@TW!YQb-X^dYH$tq zf*lHN3IdcT|LRZ_lD`@r<-{7~-;_0Iv|89H_Q};ZJ4R-COwSfwB#eT^lJkAv8i<*t z^1)y{F%SgLEG00)phmX8%~)e=K?`njQ(1EE(L7xp44UX2G~i~Gk;*!o?~M(HXg`a> z9%QzL%ubJOf`%wWW^%bA4{(XO)B?n52i?2n9bETRTyS0>TPO@d92Zh&3-6e^IO>^<{8qSY z9|(I07RaRhGQ;*+GBlRHwmjl0BTA}1h#@|MDaewz_yH44>Yht#SrReB^#cvKDfN=U zq(7Ipt353cabD1BUf?!1-SQ3ua+1`hIQY6Ov7MsWWWV9*H|`DmSx@3|_SY$8`w=nF zti3M~()^zc>EXjQC)Hd8J|PamwLrupzT7LArJYu(*Kx=bGOMN>k;N8k&FaQ*K5N=a zL}(<5S<3ljE`hCS^}Qqdx2b;XCh66tReQxJo*L{nA3>20ttk;7fzk`0_m;LxHAQ~T zulqVE&)>wz1&NrC{v)$8E$?R`;0fe~Dg3EUxqJ`Snyyfc_CC|jmNO%(7HtUCSuS+w z62>#59ptT4(pR>rAhc%P6tJPy7@OAOn z7_(Z5N$?RH;oy-*JcZj8&*@jzyM2@;7~nrQyqL{1`y~KN+o+oRAUMO14T^-JN3x8v zl(|&ZNxrB}dc+SYq!O{lQ{6)t|3yP3&@z9Jk`)7$Lzt(CSY2cn;Z>u273otKaErwdhL;pv2s7m*qYybXY%LYM1V{B$ZeYw`P9E)sQQZ~F1p{W;8EP1k1wCN+;peU2~6nBSf`?(?Qp6g zqg*>6Ar>*V$!{p4RFt0OH0017ckf*S;bsUo+ry5EF{y75u6711xq075@V21xTAM8W zAw|-JJF1u|QaHlJp3!xC`bV>`z(Tn7L{Mf|lHCl-XLN7#Rpl0W-Xt4H`3yt9c0c79 z2D%OGf}z#VT>sZsL8W6_x&dlxVMHSBDcU!s@1On_8An%%o31Xfv=YFEx_1sxz_o1x zD8v3o@Te$C=2mJS8>uym%inQpuD8@OVC>{mO8k@M7qE(T1#`7X3|-iWjDy}~xuHOG zMl3gO7)A-*c6Bl57-P8sws26m41JRZwB4~njye>Nia0^<0JXoPUH8dmZL%(ry~RpH zFlQ}+O#G#*X<>-X&m8koAo6d@kX9G_Ze|~j$9IW$ht~Svfsr{+#Z9{;?c^;o#}i4J z^eq$UZ~?aFYs50@JXmy$P4{Cy3p!ZUS!Km#)mh=j#{*JVWr=Jv<;5nF9*86nZ|?mi zcQceV6!H$4;ZK!9{=XRBt1eO6of%ejzqS6#;$Hl$n$S&VD)0FL>W=`00Qd4C+Avj! zD~3$A(JaK<5^lpaIkm`^dGh*qFBD1t@o5C`7bu?9alPtvXiiP!0(XYLe!8_=lkLlb zzON-QMet&SZ>44h>9i#6JYlt7WDowO;XDAQgxMG?k#o+10J4ETIO@4EmiUaNh6(;a zVLpwJY1;F@JY6|KyoW`R+KIwK_u+GDB~Q36Gshich4RI4fe%Jxa(Dan-8fV6T4W*6 z;eDb;Ol%cw{;0`mk+;?v2t)!&0;^t%;y+nmfNouk{ZvU5z*pw|`bK0i>o~EORD-j8 zj_|HKx<$+y_3#K(pO&1%dp7$fotQ1h~l% zMMRp9i@)-sr4Aomk%Px4L)KuG9bZF&2S$OSAX*6=F^g!n30ye7OjRl~2Jnq4Hb*@8 z=+srr5G;6|3PS0zTSpTA@XWnHhs(;6`~W8w=7|eiPx+FQ;)FIBkc$|4p@7jSK+HcX z)>%Haeb;rEkC6ZS!hipd&hLQ%6zw9YUU5LI41+SBEQxJRaOI} zF=Iqm0ivkra8^IagFkGQxkBZZNHF&jC2EQ3cfMty*FSL?z8X3a5LDzDnZq0BN>^gS zoxEedYy7*7ykootajY@PgGIfQ08`_De@nO_sQpQLs8NaqIMBPhmma2xEp^Rc(Z1HF zt)BILWpGYk5%-Cr)+C+Mp>_^7s`#WZbZ~BZ)bEwC3$2{4zpnI}D^io>bs`TO6NCWC z8(t8n=Q~H-OKn5Joqkx_HmVK3QP zsiJ#Vy0|J;&fjwbQbz0WyTEpqZg5MGH9R?ZQJ^%npyj|eiU$dckv$!XuRS4V-GDc- z9A<=#_WR+4T7wzsnF5&A51Y0;L*yPwazzVfmSdiifPzQoT0Frnz`ZV$Xx|VMocVEA zt1Pp(LzrgFP2>gz`_NF~<_;w8Fdn6H2+nsfOO{6!F+iD6zZ=?fEJCfsB5ek&=Abhv z*~lx&Ef8nok}r)ASMdTH5FM2&%q#PJCqkvyR-^7QZ|3H`#F7ZHL?+X0T9A)^P&y6Maf)}ZZYG{Nhc?ti4WIzZjo z3q)jY8yXRa@AQzD+nrZRz6gijxfoHDrDM6D5@H>BUKd#{brDIN*EA)ia5MSDe>Vfd%&F%OMH=!1Gi8pA|Hh z=+lv^adE zNMtXX^^5W7&lE*E2Z>FIu1#3M(<@x z8WuO4v51odqu~Qb#AB~Y`NER{MH-EwEBo&bRtvT~`q$6h2Q!z+LL(P5THe-Fl&>Ze z<~iSyn^frXAnuSO#+^X1c1G7NYgF>iHUX}^P}#2(?0YVI=Pyvjm$t>MndfHJtj=Cf zj>}bUc$OD^YP|BKwE>)oQZYI=65ZfCur=PB|JNIH1Ub87vH&p@u%ip}cX3#JC_O=5 zgd22{fhPOA=!Zg-NLH8c-Xjdc1%mpgBDJax0Qv?zFRBzsh&a966qoAnp|3%B4(#r= zPpLVt9hJ_&RgCJ6#VPTWTrom&G(q;Oe!V6^@R^_Y4@^dGtWZt`(=tCYmge8Bdsr~w zPA_U8E(X29fPVkXjq>ks_Y=U^is`#R%};PAOTbCj%V2qKH9y>^QZUf~ivdX@{O;NG z_aX`9%^ejn)Z7{+oPD?2j)Z5EP&BM?C~b$i+6GC)dNr-b)1qaqPssf3ha0d3BYrBk z+i#=lSUEmK&^jcqMz2=u)IcIl$c@f=4Fa*E-NTT&w9J?ThSgA|s}c|ixD(@J_5TL< zDY)+syz*39*ruvuxK-?vBqc z432ODNACH(5+_|-=kEF25+GWDZ&TrEXPP^JFjtT z7CU$o88&wVI!!DrvrA2qQxyzGB#Ek>wZ1j}oN>B7T%_P?9~VkxQh)`qobO3H^SZU?Msr1BwJjsS(?~VYx?bmb(hC;XdswQA()j+vof7 zgi83YD4sfTH9?Xa&98-~_MAiO_aqXL&G1kDY!}9h)FjK0M;FFNK3`}YJmYbivpu5O zS(yG@;{7tL3mMj<{*#UW%ZGyqH+(5FyM>CD%_hI*qo*P%a?C1MI{MG#`G1~uoGaol zp=aYL-u(-DkjegRLepPeS%siiZw$`EQA(ux)Rv3xl6AMN56hra)MBhz&N~40r3=Jp zv_seWG^jt^h1#Y3#AVDPuTbxL``IKyWZ%AemiJ8un=xE@4zEF(iKcGuyjhS{E|z6A z>INn56R*hH+zUCo$_4#F*zPsfokNfFGA#>U{Oo1i)J#F8lz5a`9`m2^B4a7(AXAZMwGabl{fIJ!gR`jLiuaBx_G9$wE>{e zjn+hfR27%U&CB=dSy_-5{edX4YHnG(I;DQc^aIqyqBIO1sj-86MB+8w#Vd#ktQn@kn?zeSwX7>dN_|Mr^o|Y}yCgWN;nqnH89YOxZ7Qe9DOC)cIe;^e^*& zJ$TVagDgV`d@x+I*qRg()(4p?pW+ zyWS{1gYuTzBt8?DAZR93C$2KdU^hnvnm#B*!MnXsp8soxYF@bsit7yexrt@`n+9zh zu{-kT<^V0Fw0n$q$ugTh*%OM>SG$eeVb=di*+bBeKZwKa`Bw5BrtqtRdR;Nlg~FE1 zX!Xzc(4{&`=E%?>VtTT*nRNb{xr?Y!aD*(g?t^Xr9(YH&I`v%uV0I5fnc~3-T}iK_ zRi8zcde`b~gVt*}z{(~_>)MPzI=eUA-y&@&Yj%PcM4|Mjg!aOTiTM}!;|-E#Jne#86Uc4NJy&WntBpxaZ0j#u2Y zQhNN~_mEJJoUE_S`{(L+@T0g#C@6?pq_9g`snX5X+wOh%3!P%lvix62gyLwF(s8kp zimj)NGGr579}wN=rk`2qMui8&H$HCt^FDWn4p`~JGj_@UndMolV*AbJ)1sSOC}`dt z)05P_$ZK+$hqxWfgPPH2lpPI3bW|?r;T(ybyfzmiVWD~l#puvT_y1yoaqBYK6-Loz zu)BIHc`=;^%wDj&cIfp(Xgd0(ZlyRaNZN~VHV?n&7hH|KMd&cGs2o2pjO=egaGw0m z4G=yW%eDicY+3~M*sDX=lcNkyc+k4Vs6hX%7GDpw63gA&AV&t3L?Hfee1kaH{mOjJ zbrs$UBF-&dCLQb=JDs}1LJ9F#>#1+05&tmjVOr5PScTA_baNQ`2rpG z8)ItBl2%gF{cu!UCt<@cCuoXiaum~XGDavpdbIgH`5^5I>73ws-V(pOyEy2RW=m!a z-41|s&SJqYGu zY}PXthujQ|9O5*^;B*TSTiQSp^TCZ`|J|dX6M=Hv{H-htV}5@jq?@OQ=vff#E`pz+ zAHn}llJTHuQ>k|{ah$>)x<*_2fY>iR&Tc$`{Q(o{%x2tir=C{xuvTDzffE`yL0hr2 zI6+gvnibx(cC%SU=T)zm&4@-*ScKtv=FygSKMXCNNt6q=#H)fuYkxsmR~FTRgK9zE z#r(B>%!PW5p0We>_LFwl0=3<6mGn=RVwldU9t;=Cu>GiC#OC48m?L|$i=)iArE==8 zEDeY!^N|xYc0N*0x0rm6;x<0oZuBg{T8JlcD{RC$auMDTKhIbZ12cL0C`a>fFExNz zB}vO3z9mt(=$U!=1xIS$A15bG&Q7bx!+eE}yhBvHP1N@JeSDNbpDz1wM-2|NaS6>D zw5b$mhIW55!Vf3Yrcr)tgI%e(rCZ(B8i;LO;;vF)9*i z<%jXoFi{dV2oA3V$BJXJ*LB>98s?&4!&MzBNMucC#4B1-oEC1o70HUiQftVQjgF~I zEA)N}tBZ+tK_n?7F3S0vLd4a0h!z)fmqV8l|H)x5!D})gJjaDgIL*P<-Pi#a#!}^o z#X&*sa%8iekjOg?87R9APPvPSP+HE#mH$@JoK?jsyYbCvD7L?BWZAzZQxeYeH#JrU z=QQ?VOBs8vMPM7nKJ#YQ1$sP4n`x?cJDQ_!n`Gu>JI>g%qj>;oSK_abac+ z2S-W14x-L#*pk>CiqnhC>fx<&M{R&$dA`&o8EY@>9XhtGJ;bQNt2w-E-qY% z;eX*FD)Wxu@-|=Altqhf{~7{w7Uq9zH=m9RHX%?@AxEY>z0$znj@B3AUTitYqTHGG zq?{oACpltUHN$zLHq4B6cQIlQZ`U=XT`NmqW2XL!_V^OI<@F>!znR zP2Bkw^a95dEEkq=o$ABY5@kU3LK>E~y55kvm&XrOa|v(q^^TZn7#!Y~`jB=HT6&_t zTkG(b4+NV7;+yLU*iB;)!ks0gVISX|6s(TRx}TQ6ulAlK|HpWN1r&nFoi$lqv{G)v zqHxgFW177$GRW9@9Cu+@+{?MwvOTOc5K2^`K~1LOnx!ZLiOmARz@qLJP*F# zcC6ANF`jg2>s{I;*b$m+ubcz*hhm(W8PktAf#_y>=u|2a?jg#LW}JW?#V2GRto&C& z2jRhmUL|YsPC)bj1#!q&WHqX(1O=oC+Wca$mhOK0qo%TY@)P7KRQV!ga2^3AmLMo0 zsQHRiD;XO7Y?ERTehS?oNj8fjPJ7s5osy?NV#9S#&>=UoHvODs5bpaUvldU2T(EGv z$3#1Y(~SMx^9`?4e5U@#u*ovRTerzF?h|N$tvo%%o&7w+z79|Gy|bxEhK#EYx3B6n z`466(dMel*w!Q<8r-^VhVq$zVKj;t+YKyoj?cM}}&Gzuzo@5m1LbU<%v}co&X-rsV zh;^o;h>MHIQ$vO%%@cRC+0zpEs^@&eN$X&f4(xbPM+T6wcbcFTswf7Mx{}f00ghZLEvM4~(&p!>6`3 z#BC?#**qNk8HzgsV581$wtLXaiJw4{4{*LYady9O4!u(q7ko>Y4Or1WybD}=!LwfA zLgCIUReeQ{^?3ggeohn=drKvTo9UhZkNYsQ4r-v{kHg11nO^^0elzaEb!T#~ z_aZJA5$tiYe}GSNv-n?bq~Y$}0Dh=t)o%q=`&)$-O*840MgNiBm3YZLs9HD@Pxy=J zn~U%Vf6^oQ=pVy;2^lC8Q)!V{k-|S_$DLE213*2HXrF_oQKSMw%?z8rM z&oLHYe)p>_C~u;-UTT|{qO}qF+lDsnChKuSOhYGHBSxO%eAp9Jg0U6G;^A1IJ|YLT zj}iE2k?i;5lKo>q&_X74yUUBQiCu`-V;Zg(=1SA*Jo!mQYCK!%)ijmP1i18s-bso~ z?-L7CtL0Z2z~!}|ouK0w*i=gk+GLpzQR3Xa@>606RCOI9&{p>h zpd>yFJx>Vg=tjZwNQA;{+VC}fL3U(7{TdSvBSV3DN-WCSr(XSi4ct)wR%Ft-Y*51q z?eB?HWZY>Pnh^N3bUW}TEe<@9S04i;yzPxrJMpi@#XIeD~NfRZn|{_AE& zwCB(RQ1hdIRmB|mZm=a&Q+-@coV`H>jqZ;t_2dnCqWkr-32y&ffVRxt1Y4gYAI9MHdH@#0%1 zSCI{n{xe=z;%Cv%$;p@>-zqU?wrc_AiCD}1aIoK_6isRuTOe{D(#HYxIPoIo~$Yj|bhA>4kK>E*m>|dy` zio0U$!t??dj2XCx+ni!-b40fsfb!>3aEac`TzFOhR#sIc?hO?Fqx1@&wW}UKE=|=uv3NuXaA#=7_uzyF7=O$;64QhPYqLrW?W6V7=uE$Y*HDC` zK|8{?e-rc1@I7DnqI&14(3}p%V9&QMzyk{e@Dhv7ZCD51xdsbbWV4Gosk-Ny4MMOB zNIFQkPPu>@W?6ODz*wY+kzsjh!Qm<~3gg<5JT+D`(5Vsm3Ahs*aqwMSO?i4Oe{Iq_ zcM=zlrLD8;tFivp88vBJ=cjSP??AI}G~%9=8SazLfC3V1xD(eas^8TCU=V~yhyC@I zTF$>J6HrjeZPoGrIcj(zV8u-#Ini_n_ZEtAMXE`?Sjf$g!F}9BOdEwE>B+@RN@1w* zUxE>tBI=HbwKBQj{BIo* zyD5Xj7p=^x@A#A|x2KK5 zuhJLK!W27)?Zjr^BSJ!V5Q^c4C#8V#Ko4PD-KRnsw3S4|D#7>Y{d1l~W;^MRXzyWk znY6=U$;pv;r90AZhjE;c0Jks6r1K~< zx3@xqs{fGdhW+w!+jLXl@ODY|i%EY%(w~ zGLCJXbv&$2jpz9IZuwbEY@XHqi}!r>%?fEFI`$T;%-E{XENjcxc1#mKpc(D zL5(D|YtBhRP&p*Oa2VzzoqW5N0}}4YM_LdDmMI`Dpb3tK%~af*bNs%*EdvK4jNwms z8U*u;zsf7jiS9&NJPMTPi}K~K$N^-gvZ;fN=+SoP(DLvm1>>z^*)&mqXkNgEmq|F< zqy~p*q3BzjglMUv>Jje-w_PEhK6{v1kMCC7@&yy@$+xCfr%eMG(MMa%J9$?;o< z@lpop7iFh=tz~`d8qBaCpx9zg;X$3t!D3pE_QEu`?m6?4nRn)US5cRrkv8zZvxOKH1YR^LyC-#O*p-b>UuQDZCII3GrhYRK606^Qx!NLLYdWo9{KbO0#NVxwc|m&3h>lprF>*$e&QAN z4n-(eSRNWwfp%vNLfhpk0cpu>EJpDBX0K9-ynB_MCXi=Z{K#pXtu%gl_EgUF6^2-R=BWhAPSukLlI9;>mwy;ZPSMO9h^pC8&mvGj?06g4v@cdAokK$y?;yEj z+gx~64gMGU(>W|h3ju0{lC1B~^Os>>7**DJ_A6fHP*A0o9Fi^OAlA<>^Ohu#Ex0@p zAi3%tyl(F*St=1Lf-!?7Y12|n`TD_lO6YqGI2l^5E zF`#*;tRK}{>Hc<1ZIaNVcgp;cvwn7dwUuG^`Z>Ws#*sX9RiZ!Gd&0RDV&#DTWc2_s za!kvBl0}3GsWcA;FTG8Wwj*sUrgkiSZU3ijDW)<1e@3xid_j;;?HPa2bGvb)14kwK znUF<>T20exSWFjfobN`9{(4u~N>X4W%{tk`yYlfOC}}1BMGkr2fn#yre-T@Xj(j@q zyEST=DSFRAJ@SGS67PiY3l!!jrPUr0Vmr)kR#EAAvV`__nUVzJKGx9Hf6vd$Bf)7I z(xS8eOwhEC1rYT`_V#Dj4#ev*qG|tyy)tP(p_6JZn%K>>gFsgrQ=!wHQ=PsB?((D0 zQK6Yto4m}IeFf^`F49?JWQs@-ZRc2CqgU=AfR?{_K#%D-?G?_QQ46TFBc+|G-1QV= zapzdtOO@jD>l4Teyp)DClv*+|(*80vvSKQUKT4ySa##Q?B2D~R#qnaDvIT~<=Y zmvGsb%zRyzGyL1W!dpq$AoQOJ|A zQ@Ei|Vmom=-VQQhYSI@o%e^>-C*VhJjYn*BuZt66Je@qPn&)XRPH`^dWq>-HwAbu( zQVdY`*TJWZxR!ABkK3Y{7RlQn-aj$@SYf_9g9v>JGE`iJxJTV+~RpVF-Uy7NMq zPYWIFQU_0{cXMoC=OMACsboEhjf0ZJ>}RPot`sP7%M@EDeQ+yFcSz&V?eXRHi?LT7 z@k1~#`=urOr8PV5Kv-B*9Xa^l&DQHl~5=v{p{5mPLN3Ct~p$h|NvM*DfBuw={-c%c`NMJ)c{?hQSGmPih=s>SYy z^pP$48wTCcgHt3SVbIzFa#Fk2f1Qsq${9B$S!3-c`I z`M2>xI7bg5YpFTw$n<&|mFg#+UZQk$^CyS;U_hZQTh|Vca z5(9;C8c|`}WLq!HrE#H8oLTs|mybTVtU}vaN9gP6tIqpyQ|V8e*Of`m=Gt=|_dWTK zhu?ATgNH9zRKX&AIunX-O2_H)Ts|{W6wMNN*3ka79{ZZ#Vv6Qo1|OeNUB&>DwbKK& z3`^K&D;^gH7jmP$ar!SxUbb3I9a#Pu%z=j{;EUvxmtCp(NF~2W$Yj{yNs6_xbmM@v4)~hvUko2RlRf2^z$adtxl}MC5!UuI8!& znxCFKr5s)*n2eeTfz_rA^o{vd^E<_(D9&AHh4C;{mIuZE&W-Ho#d5_+m#Yv%m_PzU zoW2RwC0Be~m|i$^mb{aJ^T@lyfBOlGgx-_nu@~+c7>zfrKi^k37f6oqyPCQ>A{73j zqq)|7`7ei@Jo}rUdyB|Eqho>kfK0k~E~8uwBiFN-`ob*X9hKu~P-s-&XUr_nl(LPz zDl&v+-?H2y%#QUBdzFXQi^6t&?+Z!TPPGlX&GR|pLGao+BKoU0todY6Xy@iPT~D>g zRICi9$m8GH{M`13qW&CI$2q33JXzo{SqgBtg9m|1Xj+J96E4*uG6(9wcz%i*IGTHt z;JC)LAUxgg4jV!oj*oU+Bs>T_lwKf&sKB8*Sj3yV<8R z%FT5sDRk?eV9*9ZAF0v_;TiJ&f$QWl{+1$MY@(g?6d95NV=NO9Mx#^!Zz%cNpApnG>M#?-}U6{N{%?bAk`yNP-Z5dw2G;O36)0t~@N z3ae91h zwH>&C03r{}_=6J!nZ6*F$YA$nb~B)tw+7!i0qFg`Ua=(Jv?K#E9XqF2W}GM28VgQm z5;Q@MHwDXsthzG^Qdpf>UIIal&R!o8Ew4i-5-o!Vx`{w&pY$2QgIfw5$QX@t?bj)u za^V?+SwBvmX7`4*PO2&#`_n}tFJES~PF|~;N?$BE!D6ZCkL)Ya6pGcsu*PqCW*v1Z zaZ)ZM-UKSEKM1kk2lxf@(PWd*dlGeUwH~5fa67RKC8C&?H!g6-;fT!%bW#$HoBN}& zz8V49)RQlzWou0z~!mP_&w>UzfK2y6XTP7-!Y}vnO*#n~S-GBup)=-)Q z@E_zG3!Q8LkvjcH7&=|70#Y5v(Q0X!+Y_ne^@nqP{h&JyEcg$Gf&fwe_-IiE)<>#C zGl7AHch1GEG9)`C4Vd$^SQ`}_HxKXJJTesxnA5|1OPZ#HGLG*XsKfme=#ftv_;|FO z4vln(W0SGooE;31Tk@n%<)zll(OUxC^pIXDO6z4(|Bc{JznT^-m(lCnN331lbdU*+ zKr8O42cnK|O?88wG4_wIja^}iu(CHFJ~?0any(LKloiC3ckJ46?}wbQr#`r3qt4Mo zE!dv9bQU(p&3IX?(hG!^Hvtr$P-eUX1e->R{AGniKC?ZL1IEU~vN&HQ8t%A`I+?rf z8x5Bq#K<<;UM2nXakC4KrRcsxZz?Cw$}}G=b+?23z*o3g&DTEDu*=Vm=Q-neZ^Z%m4M|So3k(oHrWShRFR5;eD8$Z)(hjmM z2WCsxYKdjpwgnv2L`pIGNt#0N}I z!g9(5`tV~ArFm}6+#!L|jP+U8^%|0vyy+_sWYopSnu5(U&)#hA)3G;MDrr#{hofWb zp<9F33pJ#{v5~a90=nq$c=(sw%Nja^=55Ps5ep;&YcLs{1(3bSQjIdx_`ywb(o>m?>Wraldd5Wbzm_ zaO7_u$=_Y+S8FsZIrnSoX-m%1yz4cXrMYEUM3!onITQ7Um#&-qdhI7gzv_{qACE_u z2Lm5C%-|f;e6c~d+AD`jAR*>S=#l)1vt5Ccjuj`Bvh=lAz*ZHjRj&%i3(C0>{fRf# zy<&^<-~u<)$r))yHr8s#gJPG0@{N(QU6(ETs;hgTH=%G-6>1P(f$=v&u>g65j9m2Or-r|K<;0?>M}62*gXqoQu$f*>YBcKOUrLT@R%nBr zaJ*be7;Gj^D_ZO8N@RGMMK@Pu64bA0DuDORa#l1u0O50 z50DHnTipmbJ)XM~%Ex@L&L`#BM>8)R4+efZ%&w<%KMp(&PUcSE2Zej1NcE-F9(+)# z++#{<>~wgi31s^o8rrRP)OM^^w-R!8tg`Z_Wh?=uRxkCRU=k-BkkXh5Sa9H)qd7jF zvCY=U<|EDqa}1})GK91}(`b5liF-5x*vIono5IE=L9rKi93EtIy=rZz41_j;N^HH( z(0g!3P%JoEbSO@Uk(eJ}hh13b2Dg!^^5Cz@Hx_fCve0Ia&C4W`!nUJM%ulKORJpR# zuWZqqw8=&@O@2E@fuWhK?dI6=>umEKsq#2!BKz04IUXK1lwaG<$qoYFu>_HtT{j9! zR9s%ru|3XcfEOV-KD@d7^LjajDM6IFt#P{8wC_hKRdjQ)AC!IQMy2Aaq*5>1L)x=v z43xs&6V0?)gIs|wniYrjyLc~2L6kq`gL!GAPp)~3goTJ*G>BF(f11T^e7*>`G;hqd zX?YBj(w(ByGu|H=txoKe#q9asK>f`!2ZgdKZ}-VLU2ofW)x8;`U}h5AES_s0)^yR~ zF)+GFAFpfA{t#$=x_VoQPOZMfH#zBZ&h$neiAR%hB8*SH6kJ2+_8}UXC9BciN1ED` zS(74W4_M0_`gIaK6GF_|f5(uYA$+`CGcb6&y@6GDgqyUXsx3{*dHsb|#3T4Ew3+T3 zMqXLO>FY|v`dGOloyapq;uA85Yyj5*1(y4ga!1*3Pjnj7{16k^FH^1yzvh_v?p?+~ zKY?{JPA`-p+3%-|iUQqWL>cu8yol{ioApC%&i^sf|HJ_>W+ z)CFhKagF1DeJ=qn<|u@#vd(}bllWCE@bI5;%s0DaWi>e(lr5o(Pjq9Dh* zx%&qw{1Yb|;9yMAx>$DN8W~>kx>a|M{!8=X0{t&|D5vUnu~9v0wyJd{QOSM#Mje&D zXWbsKD~|gpzC_xQz2+SyCjW*;MX>;Guy^OIw}Wr3lBD(CjqcEWngrMQf^fF%YP)5^ z>s{|Mh{sUdgG^*?r=k^bRuEt-n_TY0CaqSFw)@vN2#4D1BkBN3VM0uRF+lt5^*X$) zpOoj$fm3v*S1e+7RYAD{g-fu06l(~PlRrzfpyr71zf;5UjIWMG7+D*1?z{7*<$Y=}5o{Z~ z4D+Q#6kyKXO}J?2H+qU{$(!T97qja(0iAqp7Yz`eV&LZt6J}@%)uk=6iKm8`gY>;+ zL0-*ror2^EdaHpk@D4S?QEcOVy?bTK+vhiUg`Z_6~-xp~sl3 z5}ETqIsn7;fgQ`bhU955@bIk)+8sj-Wf+zIHNS?nfgun$<}X7GvW((C!iCzX@mFv7 zL9wJD?A1OBVOpF$-vvBuix{O&-n|Vz#DC;1d1cqF{c}rC^ECo+^@j?I_ z&r(5E3k{++Cdz*g#I8LXP`%iabHMp8ztG)PP8Prv=5~^- z(x<#izP37S#MR;NMo|x;Wzn~ubQ@jNW;(m8AK`+1?3}~P6?9wz2&PX9poUodVlsxUvi!ODBm6Yz_<%F?iPDM^=SqPI+`$G(T3cH=q?t(j6im zVkIwbe#!7ESKK-5U>ovINWJ1CicTeE79W-Xg$Aag-RxT~`M^`7j~5!l=kgmb|05Q~ z=~#<(*jSPK(o_f;-^h)o*(-!6yNOM`+ce&BI zeAP!x;+iDLyx$Rrm8mHKi}I*x$#~5ExI?Y4j4N$?xo(!(C7Dg1V0 zz`zF{>Ve~XWmPs^S)1k~o)&qCL6gq$10vM>hNF)_yHrNoyB&DF4gMs)%K4@#cLcokk+caLIBa_5dn|Ah}R#haMr;S-KD$

0tSI#`D%VH*u4JP@Lq zFe)l`8{qGmXf)?b6Pq_6+s7Ij8^K5<=PS(!MN`0ne$ z7N_JO$vE@T!khf%tj|QN>Qom{3ND#^k)7hVzi-yio)_K-+i=Ys@-|NJ(hUNh80&7~ z7dPEw-@w<+XYq%=(ZoQP%CzkG$_rFJ<^&JgZ0aCDIIZnWpkR< ztaOC7YE4B!Ad2&5bp^5id zrrvGSfY{k2`18%PL7A3L9O5x++Swi_j}n6XNN3KOz$^ibQXez)2mbn2Cvo;ggHqj<*3U)9I0|xI!RTXW@R6 zwNC(i-vXJ9M?8WV5HTfp8N(B+y3)2L?Dk84TwVNq_P81Oar3QFkt5G>cA5t%e0(T|dzxm^S zl%~iBKvh4%Y7r1Zbx48V%N!3!7W)pTRI~e_gnfaVM+1s&!cQb}uwzN!XNwJW11K#o zFPb3)X0h|QrJQ#{^t2&*O0JD1fiI@mFLD#|0ed`ByKbyD1J#5$O<)4#na%Xt&^bXs z0Opn1MY5Ss984=z55+(J{`}FIm-MfZKeCm3xglgUWA&()_d=2J z29O4?hZwZ71vT+gS!`Z zcX#&y#a(-IzUThnC1f(gcxLuqYi&OJg+%~RYNm}%;Ep5&8YVs@AChsvY8i_~Bnib` z40NpcVX7iefMjK57yhv3D%2J_uG{$aTS;VxoX<&WQS0r{lLl0cnz{BOmxadWCt!ne9&JD>W-fIaMOagyRn{^ z&Tfzpr`C5pduX2S0dp%l%^WGd`f8_g)Zgm%%)u#?YLUeiKYX6P2W z7apUui1Fce0HW#P_J^15iC*LX9%6%#%&%O{#|o#Tw{T#u-zDi@Fe_JL*~w4Yw7oJO z2fctT&=JYWUe*s#^Y;W&m|gDqb29>L!c04VkId0zK@X(Rga0jQM2BTFkkW4+G$p%5 z%)-|i^}FAEXMQE{46b|0r$(NiPtR64==$dE_o2+`HCW~fhvxeqIeqpE(}e3ZMNJJP zRMc0=204lkq?~v4zj7w=kg@?EP1pqyoPJ9#dxrn1R9MWUUQi?y{2%(<}IWtW&+gzpcU=^xth}_^^>i zwqlmvXb$(-vi!l!BPvRC^sA{7|6&w~E$^@;j zy?GD2+9!T~ZL;%EjQ;cQ9a5mgQ<%mxi6`bofrRcPlUYn7usua|LM9tql3Fs`(PY?D zUBS5`j?(nXclR=o#S|gPH7sO9up6?#{A8&Si}HDq^gZZ?H2vuUeu$c%uYy19oCE=a z(ANJA<@jImUW>P7KvxIhCNr6hj0ts5)byGRx_>&6U|BOBOUSBxr7S>UT6>JPRneY3 z0L=37uU~?YN={G%N03sxd}9=Em{C*^Zq@uxb=1D5yg`Y=NjqgRJtzTc)J*F=A5eE$ zEmDsTtDs7jxeqbXC@{S_y`)|F*E}-(vS@K|u0;TBB9j5vN!`8gki0s@FDN9jJ^TJC z79llCTj&J<6a`1eUg%~4c_5z1hm!}?=EhK8Duy-E%EgwG0p8X~=FvfE>AKdOTZ+uD zVCP#U?LS~ULY6$XtT21>Z8gBY47_AX-9$vXCLYY!EY9l+r1vSWGiSC=u$J#dcA8v{ zFjua>QB+oQSisjuRvR4GtXJ5ZN0pARsT#FqJ}^awe}m)5V?wKK2joVnBgmk}njRv- zI}~o8(x-l0j%uvRK}3N+??iJD8|E1zGOYh$EwKnof3DkkT2ozP=GFM^B4rEX9`*P0 zGj1HK@^=;9Gw6CTycHwSQ2m^;+1}i;1U-X8rkA>e{H68-wwv%RgSFO)^oB0kT4qHy z5DeUBXd`PC+J8$A3R2r=n||8fHJDm7;p1s$HYspUI18dYiFi?@-&Hsn_Xz-`XKl45 z3=3pIMYO;TbGP3thDr4)vsgUN_#)9);d=S?8(IP6ADrrI<4ks^Mie_P(0krNf&?0$49((5zu@(wTk9N;*M(e%OK ziE=6=sEY=A1}E_+y^Ji`|F}QayE|+|=w71v1^2rO)4BYM#D4o5fC)1o!KFCEV{zaLT z{abr2Z#!#FT#P;vw;*7P4gjx%f|7nf@e_9{8l{jZlVxl%axJs?&scKIR4ed-cQ!K# z<`b?hT5_U>z@5fzY=cYQQgi&NzJS`_wITOPsoE4vnsjHRk}&z#U1K#)(q#l_VZFhQ4K*Ir3jyvfVj$heB)EDnjdCZbU3t zr3hNB7VOe&{)jJ8{(NoCT9o*K7H}|lKp8^dCwI667S$DGc)z`6F0kT%2W?R=IA=*; zo5S=7d!ihXBB5fZWM~(K2Qpl%T?0s`H+;Zg=8!wRatqHW7UGorRRgE=po_vE8SeQz zEj3R7?nyk-l(e#sQ8cPaUc6lUPn-I7ntr9em*5 zl7}C1_=!{!2pL#d6jW(UFr->elmDdV+$}gY?PI#^z;DSc8Z6tIcuuJ+8+f>XH_JH0 z|J@^1{+h;vv96vzC=0-!hoMJm;1HAvu$+0%L6krHF*j=(^qk%WyS@YVA>7;X=Lz{- zF%M3&#u~ru zvt3j#EuXqq2xeRm_(ZaC+-Tq6YJHgpj$lI+0swxbeV*Wx-PmDnxmWmfW-RpymriDI zeaFqRQKRW$^JWzp%qaTGQ6z12L#XmJiugV7c_h*Mjpe&zjxGm=R+Ua+v#ET$C;UK( zIy{$Vi6N~3lM-U7j*$x96^0wrCyN~S^fSaeg^=@o6r)e05SDejsSQQl21SU7va%6b z{j6bSj=xAuC^3SeS3jW=I!u-9$HxyEWO^)6A!J2NlXTciy@OR`} zeMPZ(=R`k1)RIskj9h{*UxCGvmDv;>cwqNY+=p^p=ZNAb)xKp~ENmSj3Q|NUiE8oB zmPve8zaDCgsMst=IGT<;Q*N@Vy*2!k%rr$eb{!W+ui@G3dLZ*s2<!D))q$f{n!2eo&xqT#QUUO#au0bsoR$Iq@rB_>apd2fP3cOCW;%eU`gNE9%e zr$c$dYHKS8g~NG=O@@+zxn5$1vrvpE&4x;F$5Bkk;L9>mq;H5B73Jx| zeTouA)+-tD>L~flv->5Zs9Fc!TS!2PM>Inj_$$Hj3sIEG7VfvkRH zPZ;X~4&jC{ISDmElEj}F+A9Yiia#AZn#NXl6fW45DpubXM{mIQXidtjE(*yNa_j$H z-&^jxj|3vkr{_bJ{y-EUWOgD0e0q<-y$qJ1#6_xAkm9?s8ndp^zY2Rn1Z<9w<^jJr z7dFrwbNED2kF@zmLFB*aCGuHCLHJV$6Wt^QMY66CrBcBk*af&p#$r6pw^putFfy;( z1B9gB!p`}O{B(-DN*Xh|3*UZXO*gV>EvWo3+vi)?3lcOF;6fuN@+!0b%;I|@X$yXu z!ty;Nac@_;woxg=zY!@~0N?|2j8aeEsp4;l2gf*n`d$e-L*!>JOb}Y+$>~C6fXG{TIZxKxY#4>qvAfKO^ z{|Yc!$d&gN(pJNMx{UW-`i+;R)o`|Img$QkhrfTTyxdNN9}HE)-RiA6w5C+vO*|py znS}ESMre?Y@D+%h^^Lom-OXq^Cw8*TQa29}CrXW)rz3ls;vUHTXh0@q9)N~X@r_HM zUU~{UKqk!{Dm|P@g|(zsO(0q##f;?@bFS)BBdTBM+Z5&d(9ca^HeAPE&+;C#^Tvz7 zp0!MzL}Z9rj*u`o&XBXG@J5&7{ZFcB#5DhkdB2yi*9WM}LzLzFQ!}mhKy}X)F}Y30 zw1e#|#MhzlZaF+p8pEd;1oo^@QK)i;+M}ZHjZ6=r*0QUkC)uHrt`{Wh-yojMQ_{XSLLlGWLz>dIj)au9;t5@E z7RN`iMr7#%^f254zensYT-k!>XmzxHJlwHH{0)H_?iztD{<$Elc-9C_cD%c_)`l(J z>T3#_-+RKLWc=>DX;*IJWY}T420LG470x&Vy4cKY=gC0B2x*PU;$O2FAY+XOz$ag- zPxLl31fW}0Yj$XxZ6w5nKqv9BlQqJetrmEi$r=&Ej+1#rQE^?S%$Qg37atBhqN}E` z;?FXo&o4Xe_D{kgvH~wwP8%T&c7o^^Zw^>tq5gv$T+ENYOc00dA^!dQN;O0|R4_g5$g;e*RhZ=9~jrBr_auU_LG!>fy~ROju&gYJ%SS z3Cvm=nbK;OAQ(k<`{M;JjZCG)rMGJUEKt;Yt*VNMq=1tHm`tRs4uf6(Bb}LX_N+kJ z4#|;gKl!?ANX*$`|COt65cCdE@Aj>3S7G~@4(rocUwZuBDINo9+68Op)t1}u__#$z z-U5}?ogJqmCxfS5eXrF+CskfK-PvC#zK8ged5t$3Q?|+&ej7J#4Uj>&fT=_I2lxi= znBlFK{y>;)W^Ix9$3<4p4Qlsw-Kp$3ct83%XofCM*|TcJv+Ba@*7cIjz_*lob*0vg z!VdPY20@~O4eGi4dCJ%hV4{*jY(6l9kq#Bkg!Uy4X%iGdA25#oC^Kb~5&r-=yIF8% z`cNYN-HIK_u!sY1{Rhk0Hd{{S zog;2@E`UK|5a6Krowy zE?~${`iO<^OJAiFM&r}CWTO%Qx+qE6H!sCPkojmVL5Yu4iz)Flrv0pQW^_K{Atk0= zqE_@H=chY=mR~XLrm3!ft5|UF=DJs)mi%XIGuSCCSpCk8lbJrx&pdo;fa+;wQAKAF zxPWN)E5QU9zf!yCmskUa#zgBI%%rb*N0g?W1XX3h%4jtvPg~0QWM%S=D)5L#vQZiU z?c}|=Mtgn!9E()Kx}eF5h4e&#%$2@|$yGVR1V#grQMc^t*VwV5&b#BB^za+}kl9?k`s4Ujy1uNWiRRa+ zbGKR-;!KBBMpaL0i4(T3U>Jd`NXJ}alr1}Z4SQ_97OUC)QR1dbq|;_EZ#Zlny_y?HZ3qWXMg!ZcSbsXjxXjzoWUs)% zLOC!}Yj7>U2^{wA^;B)GoB+<%$K2RFkXi;`(B8%%017z63-p7Tg*MF3@L(u4Zj*JZ z3S`Do3P_3v(6bYLyKYb{6weo#bZL_rH1a6`2;6+|{tdgVMp&Mm3w&_6cA6u)4wN{h zBv+$*8ebVsSXKrou!ML`%gJNIj%Up^g*6AhA(lDfL&(JA&+`Adeq6?<;al9D63Yn) zDr^hpI&mR=lxYwcAWa~4&{+LH@*?R@?^Sb@RnaF1U!WIDKqSCdxQxM>oxCrkx!(q zr~LV?2Gbk8f*I6_4mV3f)ye+(pEQd&ez4IO(O>tXjI2+ zSk4ond27tf`1^A9VgX%+RCrx0usG4qEZCO{3Zc{D68`~(5Vk;}+;T9hTMI8{-dGIC z451@;AGC;gCFI&68t3n#lapLErDuraM5a#2H;-_Ms$W@74MC`+I;5R;srqV6G*lE) zIXa;P>8;XEB$JMY_rZr{J`$_cgS~fR1WFlODlO*xQWJw(^SbC<0Ku!;dFKBo2Qc-m z(EwL!&AuQK=afX`Ia&A=6lWL1+;VtjQse4AA$GY=MvSFi^VIkG_xSri8=ijNeWq>^ z4k+#oJr8p4eU5lTL>NX#R}5}Z1<%XB-a7h^B>Z64}tpNS% zy&gi!Pvb@4{Q=ZuefP-$&-cC*bhO_M=g=qX&_s|=Kh5~B4!BN8BvGcn6w!LdXp%DrEm@M)#W9$?_5IPt9Dvc z!Wbq8+UBmbxjUVjYdvR_H=AAzY^@t`^zh_z3k=&;jc^6`dSYxI6(V&hvD4S6>D#M+ zbhs$pFY$QO(GxXJsLug-R^cTt9k7MZUL)~Nll5tkm(G7gYsT@bYJ7kU~{^k`&fru6cum{Y;peG8-#3Q?>N?5yI=w6e4T3{n8 zQUF#T|2N|HzXqV~MV1ktR|C)&j#uwv`u zsgVzF3bIveBigU_E(M5CjAHlElN3@*14OD_#vv1fRSnCeJGk7FOqs=ee(A*t2cp&E z5c^E=sc&5ttMvMN{T$xbs8-5tn&n|YQvAU#VryJz`wg$Zo&L`?l%=xGt~9I z+=NRoqO@)jS6x&J*ZFj(^U6t;g3 z=@1MD-95v0rD>*a5+?U=t6!Y zzK5N(Lmjf2!_H0DA*xFIRl}fdoBV3s&r=-?eC?noCtcXqHTCXfXlxf0fzf6K4%O^m z>IrA+J0=PxrX!)Z;#}!WPh=j=UFt)j>kP3|0ZON9=h-`K6EPd*(yWisevAL5+ZFvF zWSDQHHEF@#kpQ>SLyH>EImVyhw`Y3LN>C7ck(BsMyPI+E=&?CS{ii0(?cCU-PSfIa zyIg9>dR;)m#z*_B*D}XbBtO@L5qgqLbyUt}c=12OGFF93roL2CrO(oN?fwn@EV#He z_m^neQzwlZmBk$@gLg(qk=K8iy(}6)R>Z6LzuO$ZX%?-J(9OXVC)W;YrkTF*u$VoB z2tzofbcBdAgKyv!K~y}A3W*CUeT*-2pZJlb&6>NzKBZxDS%2?$-=C8iF;zM*9d7e( zh49*pv78g~?+6#FM*3jCCQS6?^#0y;%&SVb9^G>=z<8q`e{nz_XluDu(1Dwth-HK0$rwiWiS=dJXN zZ?7dN2tJbmtL;8VtiNK`x154t5FwBa8>v>JFw6vIolK#wK38XiG6igrM~RCy%ienL zkg@Z9j=n9N{2fAa-<^3_-n)#@MhubUDSAIN#H^WUDDl3j(S)37QF4EQfSIic3wK-- z9U9#<0|Rw>1gKy_PVCG;;QOYNJF7C0#G>wIGBQx+rwb8AAI zRiVi#ZO4Xkj;`6-3v2oj1`+Mx|BH&7+1cQW)BY6T^R=BteG3&kQ*5t?wq;LhV@AHU zp{sD2(q#5o)YfYjA=$0&$ij3FK(i@zDlv0MW7+kkXKteislRJ6Rl95RD!#ymR%6bR zeQHjQ-f)}51lfr%`feJy2&-rIeD)*zr{R?x$l_BDMp} zyjLl|(mEMjsu~~@dHr^?+HDIdb9#KI^+{x}h2%r5A4DkoNow{)OokF+z8A!kBio_$ z4O|<-phF6}?T~iu&W-m3-A~fs7bHR5wni=WuKHNh)AS%ppfBBpe*bDx(tuv{s8QUv zb6>tP-3y(qRNp?L!-CWHKuz)?W+!i;5QtxM!gDEJquJG&&c zzWNTPEEKWuRpRi%EtgX8aDkmgfs!IWTOX;|O71;rK22rp>@Uj?TvzSJVn;&Q>wGyS z-n}=Z$y57^nK!rWlau!$y*I6>m6x`Oe^+l-EkAB)?_>U{AOSwyCXpTy;WjP>E3utM zhqLiaCdUVP;93%XcFz5t)j`b83LH(&vf#4tYh8v!ef3K*Z*qhv4{Oi~?>=LCk8ebD z6yu>UkrwusU!Lk=Ck;0Y-OlQL)7U;=4(LFCe8ZE5`Y zs^N1%H#rRYD~LE7@P!GVJn3^#)bic278vAX8Lzns=4eIh$V9a37Xw@L7M>f2YCc zwI6IF=RueJ_04i_k$`g?-zex~(Ky~TA4XUfhl|nQ_?Ptf&JX5&H$&g!f87j?HZsGI zY6^)>cCC9 zNV`3EfglYbmXW_=3ezLxNgZ;3l!wp+g$0Sph0fA_5lT!Y%8T^V{}p22CS$u<<$(w) zEBAkwdET{>U~errRBJ7LkK7uo!jJwK1HKLPA}oNrWl>>>!4+;jKaRp_cFOB_+P7vW zpK@w|ZA#6XcJ;-+^>u=kKlM|jOG<>kIS~Ey^+Gb4HXw^{aDsTqfX?S2luN`6UPT1$ ztjatXxiT9^v3M?nRz5xPQw!WM`pC8~Qe`Uh2R~GFW5mjnK^|LF3Gvzww zGS#Eum^5Jr;dLK{2+Efoa1(Y3hsZ>4@k*rPM+&b5#OBd^-96=3U2t?tn=HWRA3*Xp zq6%N6nQ*e)x5Dz!@CLWph*^O=_w^N?-nhG}M#%870n+CS+j_^QC#}4TP=dQi)KdS0 z8*u&(N3g6!GU_`)EY@NcWGDJn2^`Fp~+aVP|NmC1P$1^YcJ2vJZ zeT`+=bdyKT>C&+PmtYk9t-y6`w+?<$31I2i|3hn?0`?`zO%uJ&hWxA`<(2U36#m;S zwLy-3@wtG+BL?FB?xuF4S;vx%&jD{i@qM_H^u4p=*A_SUquOMrlpTdyWdAB1q^Rz2(fk0`gb$4--69JZ;rV;{d zUYL+4H;sQ1Y{sc~c<(GmV4yw~i(k<9JKUPPqdhCMH~#(&`s7Msc*cA$$f@x!uu_s- z&jnXsC%3j1X?iUfoGPE_K1F@>{*XQgkp(0}WQm41FL(6j$hLX>dJLeSe;7D|qklj> z`Wj3?|MhC~6ss~$lWDErsP91wGxxnoOtMx9=~wInos?a2FQSP{`P zh!7PrA>?6@9q{ogcT>gsMV2)Gx!{~QddNJob}^j$bc-8i?bQeBBR67)owY0nJ)DE2 z%&;B$7ALicoNO!f3inUaY*h;BN%5?&3eQVb$* zF&$FSu#{_Autf?lNiv(+C^fAola{N?4imedgjUyQH`I@jN5mkBKGL1y%6)6G@c@oI zNw{*W+vf(>&qfzKid7fzpQ-F&`@j@Z23$MHBMykHewHSyNRSxHjkov~79h9-nXq4^ z|4UW#06PBzwh)&xnK_?fO#EazI*Y{70!B(v>JQqxE@<0=(6Ome-gpS4$1 zS?&5UK;H2Fp8@%t>kQ%~RzHu6VMSA1nOpsP3U^zrmjWM;t~{^*m;1F^YT8aLyx?bP zI_x8N`Z_B-tf%+nuvIQfjWv@}sgwutf)$Psf--aA%$0OkmgKIS2U!lR!bP(Byx0x~ zAJRb5b9Jd7lXmk@#e*MsQCHY#UTa0xdo3T_yy%t zexrD>WUcw)p*I~^x17RCpWvNd-L+V-)6IkU1+v2cQlA@hz~BN{?yF#wnd%+W^&~4R zMN)hvd>`2T?^}yobi^GotkYzQ0nz-!)1evHs3&(ic`%RSCOxd;XO4N0_nlz{KdtLv z?P4RUsYh(6&nCs6W?pfy`>){zFnzsBSm`TBu)0a&C_XDG3fUAvm=VOpVcllXm>D75 z8DamENvjxuu9c{*J6&aaq#=Z_fzqM-U$=H;Mr3i$pv|Mf2BrK^dnH*n&7%l{A7}&% zCXr%)qYN6B1@TKYC>H}xC2^@lVr%VxxieYgv^{H_iot5Xk{KZbt34AKR(hkHPUF7r z%C!)_8FEMRe{Gu=!awDoLDPtVRh?2DAUW|I?e^?h1TpE>LJsCBTLXp&m}v@U=pBGQ~x{Jj zc^vj%*?Qfo|ITfX%v$jpp0n@#VAjV0*AMQDrMxp}zhz-nr_NiYXFc68i1>&i3jA+6 zI448=U#MV{5eGJtS!hmM#IzfoTJgqlUU2@)%iD?<%?Ilh6IiOU%*v?lN)PN`!*?3} z*QKv)AjD`MsRmfV6?O6MZ=4IKK?V^CTGpKng#QgUqP+lp5=p5#EN^6VeP~0yg9l88 z?Lqjzf-Bsab-(2kmrB4EAO+3o7gkoYMo4&I6->wa^P*ED#NHn^R@Q%DE&dbF_xH=6 z`V)K5ZQv%$e>0(QJ>me?*VaKQ5-W5vD+H{kG3yaA$f7z2(Qyk1uG{Mos&A4dHFXC5 zV848>nOql+;B>AIeTS~>^D)6|!$r5Po#w-N;rlicmq}57<9~xj(yqK`3?e^hMN|6` z?I&MD<|X4Qd)JT0<>0T62=u)lcG)9Y1gc9^YfKg;I0Y(3ot$)4<|ym{@$Mh_%XU>Z zMZ6jz#DP@=NZj61D;et%*0Fktf)0%krLSA>I9wYzZ+yjL-#x4++80{~qZGLGLvCz) z$Tcx05CV;;unIQet&k>6YHw+mRpycL^PQ$7R_b&~0_zMf(BY`iwE)_U(f_|09Y7aW z`diwY^uT#ru`s~g+tg{$dW2WA?Z}I5BjnkDPe+?DKSrGeh5t>bCRsbN!L1Rp7T=O~ z2LJ5PickKo)g~4)2X9=3b7vzC*Xl)IO@BuLI;Higg^7v<ebv!>97Dm^;;)cmd-adpH-UCXx?69#}v8N3)6kkN(M zj5K#TvJjvU&UXZZG1ZwsFM;n7dlAA_TXIVYm=e$6r5uoE13u|pL?S?3*WkCsZ~t2i z0V)z{xW(MsX(3YzFEx~uL!i62ATB&R(zvAPVnTf6fegcEl=ABq=73|6S5i>Az%5;B3 zwLO3LH!q*(3|Dpthg(ZN1*<>)Oybo8&|XcCDS6}F+Q8(n8_&C6iJ2W@A&Juw$QNkR z2TdXohCcu&mm(3)2qEjSWBcBC&3em;cPkX2kP4NQ*?&J@K@$E`heYw$g_>Rc13npB z&Yg|2(_l_caJ5P#W~aMkdi@7)Jf+BFJ(Yz`qeRlHM};74S<6?wnx4+jQ(CkgpSyH= z5;XbDi~)cruJ;uO$(no(8vq0f{4hm7Jx|Q?U#kZWFNoJxsA3;8PX^ZKvhRSU1iOek zYsZQ05k2;L{&U~!!%~^53F8aBPOU%=Eq3kaI%2w=x!W<)wuT9}RqYXJ(gy}{%f3|* z%_Jp)0p&>|^7wx;a>i|KD+vUu3L0gnd>LimX!YM~a2~H&OCfJ$DJ8ZX!JJhY zq1Q_7r}o>z^@a+U)J9@XqlC@QAw{|XPgLwLrh9LdN&} zLxn4RIs&rx2`0<{uDib}sONOl^eEEXG!iAKjT9?oz}xqyTTaF8-@WnM$A1p~3Eb$A z^K+t2i_ z#^hbi5U#@yXiX&5R#plZ6obFkbwxRv63x@nAJF)$d-~c$e|TO=Y;gQdm9%n2rpGiWth^NR-kf8h%D|`wR)HP+Iln ztkEJ&_v~m0Mi=-;8FnKlAjj_VM3GszAHE6u2;!Tun)9hP=x7_M;-LKdKXEerb2atb ztc9{uYjoo|x~2sYmuyJGziLLiJ5iAT_GFuH#LOG=fY~!1KB`o>IENwP0Q)^-OUy7p z3~1T7gju@*2zuG^k9e`TM7-Mk@NF*J*dKd>RndoPFreh50kRS=8%+~iOLPA^d&C?= z52+$b*+CnZ#hun!i$!Dk6GLwT^`q}EbZz>y#hNQ7Q1;Bo*Qx(PwAe3F`n7*DA9&uBo@Elug@u|5h{ zHg&H3TQ=m@rv4v4s&ok3V4fTD=2r7jz5|4oG}sg^>p{A637eJ+N``AXqRWcuY#5qh zX`v!^mY*50?TUJcocaD_S*99>T>ob`6nktXw}L#3?A5dJK5z^dNicn9pC~~>az;NU z6$ z-S^^xv>U|dhQLkz2G8}7Ps0XuQ(qr-#dm|3jyUg+LqjU%0npf*`d1sg0>7M!b9$^L zv8~2kn=K3em$hxY7xS!k&1rM@t8q|-D!p}RlLTKoQfJ*o8G|3H^qN-yRysV1$mO?w zv`VUqZ8D5bchC?YkIrab`XLP}PA#rC!7^%f%03z}wag7)* z;PsvI)M+aJL@5Hl9>7u*9-VI@e|Qn8QW^FKoDvUG8GRNH68fRZwlF6W_L7^)ZSN&l zx-y$f`_KyDLON9&Hkv5VVXnLYyA(duYxxgC9tkn;5B9qb(3yMdSbyFJ116v37PUMA zrgE7VhnQ_(QOMWyVIZOOn&R=WJxyeIzTR%OxV3&5v-n3%81t{zEBL<^+`oUdmJ0HH zG9~n)*}xI$b{{y-_|m}XHRjh^@Y!$U;PNQ+off*!am?*~`NLeVbMS2Y*pOMc~1jYS?a+L>jV zKFB{P^_iJ)UJyi3t*0oU1&HI#VLs(>zM@IS+^-FP>i7wVKwdk6LtdJ866c=;oqEHq z@b(CB`ZOkp$`)QS2G?2TQCzNGC{#$IMp}<%km*pLaO{(ZU&xi6o1gHEuk>3Xn+{7R zXO?zgB54gqyaw@?))O=&}yM8x6}GGO0G^2i-^3fZOG$N zN1dGrc$wu27U4?cm_?&6sD4TovQ8DH6kX*gsXZwx-QA^Kj0 z-!+3JIveYTMc^>Y9(9`mop-~nUSgw(i^@C;2B`kfSY+x>9a)EK-1q|;vuym{g~6i)Yi zP4%bHof#f!8z!YVz?-iY{4=EShnYqN&n$k%Hy}N30|gA8`o4;vil|%)TW`cf6r>oU)}0kAQ6f>Pb*oi4yO45oS+pLi$%lgnhV2ir}jx z$DqX2T8+^FJTlIaB#v2ET>Sv?sjV%Jq=V|M6<_?{E`&+7UN)%=Ac}4SSg=GCtd_HD z8K#l=wGy}h)}O&m>t9tS2qMPeBy`;2_DMH^E9`i*hH(3i!upSN?A6OHc-{Lb@Z&(a z*)+A$Dr~nvTP$d4Zj!%WUWV(4yWO8MdXn&h4dV{SaIwu~BLcD`6?I=b;ANOg3-c%; zl|pAznbNbWQ?Rz?4k}hbY{wY=w{2Mu)lBz_m zDbPZ8Y;<_-mK+0v(D+A$SJL|XqUFTDsinX?+hBHeZBnlX+2A6SZOY?}KrV;Y@sDRG%Ya+;+ult#qF&nO^g&>${_fb2H)Qez@U zJNs1G(F`==OhW}(N3Pz>LtS!u-D+BYxmr7q`!LzBk~w48Ml7|6{vH1

$2I+sfk) zKgN6lP(=J$=b-q`vh58l`V_#(-w=B`)r^Uh^!vBx<=CG*C*QnjGA+nOTY#Tq!iLi} zt9PeRLHq7;Vq}GaAN7sFB6H)?+=(Qf+<9~1y6J40ocfF=eXH<#;a*$GnwwvD)w{H? zj8A*&Gan1Ti;#=_Ze8~So%H*n!{v@>6^|(1Gp}HLy5B+poq`rW2LM*z)yCsTaB->Z z#wF2N#)`-JB6G>5!`E8kI<~12T|JUn?EQ^!8}Umq}o5{89+Pt4Q} zcZV|`koa3)?ymwqORk7u!9N2F3ehgUz>p)r!;@#A?Yt4{>|*n z-XJbgYHJ%_)XYmDcRZ30-!)A4*%1zIr<2@vhVw!i_6j&d=K95X>h`atn)5mj=;$|3 zD5JT%bpIAFipI`@bwG7-g`(lKWgItrI3gw2)crA*4q!PFA@JTvKrX0Yv>8eGSDn(- zGeY2BVY~nfei9hT2MkY96>GCf#MG}yC+e_t*u>Dpu%Lw6kWNsW&gUA{Vnx+@v>3xD zd?D}3-%TD* zXjcNDoteJWVwx$_E zOE8(XJh=RPEYg93ggPsmo3)*U?gQi@Y&Yb|MQjIMFX0B>v;9cY9DB3b`exY%Xci2I zV_n3!N*d7R*dGP?<;98sDC^a1D4`)Dn^;Xm?P#uK8agZcqE z{z!v4p0A;P0_8BOZg`2@;mL187BxJ)^^2Lav6Cc5uY^c(L{-+81~_nEUKkEfq7G~v zuqsJWV!y!XpE>}~?2t#IY{Nrp)g-1BdZ-AT1Rv(4qfWnDgf*z~5qQ&_3iBx}V+TB# zT2aR&M_f*TeX+3M@imBd0h`Bk-jn$fATT^t2;qh}x92fWQ1@{_+AbpkW! zhHq2Ald=R`eR2`U9vVR)eGVM!)N~Wc#XlHV_iQNP-;_0SW`@qAN9SQIx6qxb5((L~ zNGKDoj0gJ3mL}XWqvh?g*Dwm+8jH?8D;q-%+>jgl09QRX%#D5IISM73i4=A*Y7eLT zi~1E>%n>z6!*Ihj7mHu)<7yuGTbE8Dq`#Qu) zr68Jb7eFQ?AxbU-8{{}Z^!w|zKY{=05pHhL-OyACma<{rpacMf!mRs$XfmXJ45!Y? zN6RlmSO;LWg?2~U+y6&)^=$AGo-QZg-xJdHG;wPgzolv7ZV1+js!&grUs##Lvm_0vfXdllXG_oc z3X-2bNUZc~(cL+&=n?0)XV|-?ju*XPv!Cup#S9@OCeQF^v>kD+#|a_QP?dV@;ui_G zxeu89*@(VnrCSv1u}L{kq46Xp$d6^p0FlW{M;Zoa=h&|G-kS;eVSwSjqc=7{LNJJ}j z0jI`JU#Q^*7X`(jGxd`e4UT43!sFJLE1KzZ_%F>cAPs{9rCm>V&eV2&=g5q+cINU-d zRu0+WmX*T2#6g|V(=iS_haCFapT=xgP3{NI0xnm6pt`C*j4lg)pgUf^0+Bi+FvE*Z zn-TboHzgv78K^IWru9Iwy@G1u*z96!QW@#o#rviMy6w58Yf2I>gPKp3dumQQ7Czn) zgOnueuz4(m%Pa{jCww)h|6LUI);BtLoFfT1KHpd|qsX}0A>&^P|p8WtB{S23593P`pSH++Bu6r>eXwZOpzV@9zz>&2n zEnQ=aSsL>jTIn)$t?4P+U!qg5363Q4v^Z zU*1laIb*21{!OHyR91jTQ1TmbS`pefQz$)<=TGjku_caIYOQq${qU6jlH$lQYFhF9 zc6@-(dh5?o)RBmunEou9nk06@(d4@QUa#du4T^#E(&THWBVkhkgA-@>{v^Xc`V8|t zi7^b%*GK-Wm*&-pC)Gb#cj*2yu;lH1BI^+i*O0`nunj*P6Klixm6f<`c}s_|m;@>g zkY_)S)`_ix6U=s+Xm6KD{V?K)ZD)SQw0Ai)f5u>^SSl{<@x}At5Cx^$`@*d2&Ycj_r=`R{-klI9Us~mpkUN$s2p8cGc;j0Q%A00kvx1`fu2zZQ z#jDr(!3CGaUWoiJac&Cpod%!7`;cHua$~Vy{MZ=^mt^xFkE+*be%FOE7ik%#*TQw6 zxU-%b-;e#ZQTwE|WUTBU^XXkPPu3TJT4f_+uliOtE24)2^$4)Ytb*iEPNaYX%vpzn zrS(=m`V%*_x7wSJ>DaqnDHVYfAH8*h{{8_4bSvqnzJ)|h5(0wzh6@VA{%9*0+#c9y zppW@n;c(A<6^1Cg^)@Po0ff|ZV*qtAP?&FjK!mAM5DpYEflmis=`)pLIl$+S=`;3X zIhJ^DXo5M#XCIfGza~NGR9_SVs?8t?R))7~Pav&GPJ#r|crd%`jpE!LBA6PohdLzd znBkI#i~&aJ?&gdSx@N9@LB!)l#;95v&u|n6cWf(U=H>0HYlq9}j_9Ao9x~}NAn>CE z_qeHS4K_!R*w>fzQo1ZRpIN{{14F*Sq1U)uRpMP+w<%7ZAr=WqiNPT~YbqBb7_cI# z`H?%LSLoN4UFa~oUNyf?Wpt~kpP}P>Lcd^?6$uK-)vXi7bxN0(=l`Io%c|4e%-7?x zXz$foq$;at+3RJXkF9SRvYDmiUry8)k&XY=nR>0}AX8R_NSzWJ-bKN7+milRDBGoX zoF`HNV}zBE2+ABNAG9A6ToejNYcV|gAaua#4)$ZVBe;k=&{gd3(Z)5yXx2O`D{Fv& zi(|s#z#KTJEDGGOe}T>)gR$j!_s{Mxi5J?Wvy+>!?C zid~M@*P`tBsHxSe$@kA;Qz6U{{X`E9TD90*H<!%gIczQKLW0aqHD=4H-7f#2X)! zSU7ve=T(^S!=OT%OjtRi$LeV$+V+s75034jM9Pj7YueR}RW*SO8L4Vh40!|=DfAry zo;FS+5(r>QV*dvNLHxen5WnoXFyIh2#xeZtc4ZZnAui}=_lQnOv^y8)8zYeH^MXw+ zZ^Z$5fPQh~h@``Pho~d9QN_me^OhdnpXW35A&_N)^+P^?fBsg?2x~Ax z(NYUrZR=4)>uw@ufI!{TZepJALBaJ)3A(UkfpF(UGy+ zawU7_V%X3Uyh&Ds}B)@$!D6V<*LrDA;0 zOy*7A;Pu#m2zW~ziMf{X0VEH0BoBsUA0&H2U?kUIEOe<;u{V61uMqEQWG~$^^geP8 z<}+w)#GkNp@PF(iFn#RKLIXVK4HbFlJ{VGW2Sze~pnoovApHfWN6;kKW1W)d7oN0uhZ!8-+YPU>@7 z4qiy5xpHH3PF{UQ@>teQ-k{@O&p%`RJlEwodaete9(kutS^RYkqPc9#p*wBrB%2Jb zxi0N|-ePc4yd~k9>%yJ{rAPQd=3qWO!>7yt7-&;^Q11>bbnUy^rT}tyqVg(J935y= zMyWr*S^wiKJd7c=OB-1Y&s~4BCDgoP|Nqi472Ukz|Nqj-RnK9}!tBgnKyR}#{qI7P zj0`LF2C3fFqjM6yau+5 zCG77a8r@GHP&C@s;V?@e8sB=HQonLQ(QWtGl=36+V$}9~#i*c=8S*%zt{H3-8#Qz=jB?|p zw#S5ll65(de-!eu88``a-1dW_{4RK#b801LyOYx(IUgV=+KR)GVP!+_!(BY`NQ2Ey z(S^CVgPfY+O7{-+-{vdo5h~)3u+j7cTm)J?7|ksbqG~?R{s_s^vI2S}q3tM}zC$v~ zcU>V4My4g)cb!e`huGA*MhaUekUvKqq(_EyRAd8xk zUlz#s;$w&YKvCGSqUuy*%_Br~mMe=Vqi4O=If>>C!FCYQS-(xR>HQ&QOgRhd4@&nJ zRMOd)#Z1e?Z3p@n{-5jR+e+nu2(?4H2>0^a5DxA-V!3cIQ;Kq$&!G@r?YNV!Arw>*I>KMvA!x#k|UbJw(@_m@S;j zmQ-6GROlk9JAI~5CjFa2yE%nbw~4i_rxe-~g^p#x9-^KobRMU&CAE`NT_lxS2o<_Y zs&AoDff@fsg-e{Gt4@zqp^vE02PyI4Q(Yu=N|C6*l$UX;o8Y`f zMj=MVC=~Kmgd+&;NNLc~BlnD0eWtoZp{XutG!2qVh^F$ZV3$C0m|Ro&J9N}(DE=Z; z$+8YAYLvLc_dzmK4=OsR*rv>j4l0_7KV1(hD)-xzcEv$OWAKD^G!jLvvM_ED4jmcs zIclYx8LM8MQLm2GtGy35%!%q)q4(}HhSzDW(udM={&&eR`|k=1l?QxIA$4tI3tMFh zTji?0IEiMI=)&|@_}(iqg{^XZJQnvY0`lJe^Gb27L901?TPaqLixj1Y=wlv(7nY%G zR-y6$-i)j7078jKnRbGsMDE$K%4{&oY+%dRl_3C#HgGA&%ZxG`*b~GvmTcO^)%HLT zqrQKXuM2&{86R*4?PJ# zEM!Hk@du>ybhJc; z(ea;)(GA0ULAhK@{V&8tt10BIK*(ty7~4JP#`<896=B_MuD&irfbKvJx42&DiGKhi)TxAjX+r3 z2}Qu-b1UL1+Z!qKU)s$vE$2Kf*~J6-L6}H%FahCKTw$8fK@LWoVq}x+U;?kw(gXCc zns}{E*_AeB1bjvGy&%d1^m781yQ3;GIsOk1*T1mK-mA2!b(MH^kHGYN`jxL!{Z@(H z`9L;R@u%L~jfCRSD&p_jVLjWEX9`C1s`FzlcokPqh+b8&;F^)ff>&8pvEWsH%LPHv zCNAlRk(i5k!x37=w@KQJvMD`SR^bcLR=$$;Z)B9xex|+~i2xmDk&hrzHjECl%oU?- zYCSboE3`sjdVthtf*>mp4Vc{_P znA!91hj)^5!0)F6tmb6Sb2`#Of1n^xUQ*(?C^b!MrYrj|?W(^Nj(k_$Z(Q{kiy|B- z=(h z>CYJh{iDj_QKgaLh_95_XO4tiN9F4Y&p5d8K$1r?a|8l~?SH%_%FXAHJb!9%XWL#NtQF~+7&rDY+cIpULRp?@%+3bBh& zGX{ICG@^W}P2b?ZFAxchqzd59djRe!r17p7nrylgsd5_U>h`@&w~mD)o>z!IDE7Fk z@jIJFAh#a5Va|Pm8F!Cm7qyOLYNclhOfT^Ve6oIRSp_xl-GsM!?CMa3#o`UGD@uJq z@lt3!o@WTYj>D93uPZ99vdKODbwvg3wbY2fSww{6CnD>o4;1AJ`#emUTYsqNK}dGb z9HS`fLd`iqAY930OU5XgG|r~9|BX?!6o2Z+C{p8XO8a1pqPxZ$-Ks_S6OL|~m9HyW zHePh|NHujA7^B+Z`;+mjE?;HP8z|^mRHm!msqBV}w7xE`^aTm7EWFla0gqMf zCYZQikT7f-CM_HTVOq+`{u3~B2^G=8gh`$xx_e^mtf{oDpuEHv=4GBKBc*d}SrZdo zSiOOYCMDiI)26W#ZAxx>K~Z{X0ZmS9JIbb4Ct~H6rTMZtX~~IQMY%8Pph&^JfKdk( zt(_$2l&0VUQ2?P(g_91PL>`0ya`zsqXqDji3q<8#h+AAk#)8^ER#EH8P{`?qGCOKV z6?h}w4xJ;u%1Dobf{^2>)Qh#nx^^CK;mRs4AbPw7T+?eZZhsdRcI|?kK(bLU!K|Xe<_9onwxP{W)y0OQ|x_F?0r_O z)f5~qdrN({6c!S_&-ZRRPBE&z-vWIe@#l&Mij<&0`0E3Bf5*FuZkmGETfD1iBL4jJ zuA;?Ln)$AGIX*zniW!9(J+U79wS@~ew|;Gbp6DY-hhMc0>4gFMqs7Kbo4%WZ#|Pk! zLO=c4V!kJdPML}ZITZ^PiHl>I%a~a?^iTKIBNt#0_(Z;enr(xh7QZ1>zk9IpqHmhK|kUT3hMiy zbXd%wK1_J*T%$_sEs2JnuwUl*A2KfCbdLuoH@g$)+xf8DiAd@ zrR9825?wyS)J#yE+!F3b&D_f<_`g!@>L!UVO#Z7@PC&Sfk%+#YVr?3YP~fAJT1>=q-Ngd-Y5_ zp=CCe=`B8*zW!mX93vrregs(62DC z4_p42P06`SxtHD83d^t!$f-t-FA$+^t$Ji9(XYsBKprH0&63WZ1r_t@YcBJuS%@m4 zk&x(4k`NQD{7WPz}6~m(2K!jWP z5Knj|ON$w_@l>wgL7z>~R>v&VRHwkh`wlpAeO z7;S>Xih|xqvD275-L*n`BwXv}Mzz+BM)ma*rg5y*Uxvmf5BVZ(wll9HtL*Ekq_s0 z^b9Nc+tbFlXSAZ}A)3ek+SSNs zFYv!kbH%>)Qf_yF(EJ5x4*POJ1^coD+f5*-9Q(p^y%8BlIyUC^}=Hkj|A<;cp_cc1YbCS}7-jFX4$<3!cG@Dk>*t#>V ze_8Pw$cqJ!HT{sHrkI?}m4_5{e8#5C`a_B$&p=bB<;L76a;6h*w_$?6<_3w@Wuc(Z5{E!xMJo@F$-AGzoW&>*TrNp zFU@!1@qxR6egNgpXY#)*H|&cP_#)o?;xN7x<&;<#SMpLR`M)gc0f^QHF4ZM>(#*7! z_nI-W@sMjN6Q88q{E*KZ@i|2_#TE6c6!kymelA+ntI?uD5t5^M%EH=}Ugfju)gWe* zUuEW)qWNt;PD!a($JSoG%hf6DFCwaUp_h)~G8Gi|l@vNmxpB;EU9L_Ma;3!;>7&oaz!Jy6o>f+pL$KKnG+Js zM=B>IAoTY?ixnQx1l}e{s$#RRUeyI-hoAJROyxH}a{ISok#C94oS!3odVY_K47Y)c|q=lLaNVP9c^Q$|Wo zY_-3U;9}u#XyIChB%(6P!+0C1!y0i-s$iJe57BM&OX}0KDr*50MQnv6Z<3)wXZDb3pfQd z#ucTl7tILbcD|vN})RUSBE$qBQWeQ^5#eUWdTP}Rh+=bl=NATDYQL)4gk-6 z_3WBN)z4vlT2f%rE1rvOZg^jQs4SmfRAgk<#mTNwu1>{e6+|_$vTM}srzg>7N48EM zYcAVcqK7d@CS4orh=t14Dd4LhTBs5(N+QZ)J#`{=UaX$nV?$F%f+tSWE5-w_Xf30a z5gr-NkCZs%!8o?}m2P@#6l%BEJZypchZ4Q7UY&rEy3{`u7qIAkEfx9&kwRn?&C`x8 z@!`?K_to%AYW7pLxLqVDI+`vuF! zQiwidjf{Re^$T<neCI|8XPRGjdjlkA{;Z5lo8x22*bw4uKBuE)HoahHmQdWL3{!jYLro})7 zbQ<+WS-uelx1Mo2t#OesaF_pOQxkBkn8=Jqi5c5V%Ok!@J-IW=<;E#_OqBeSrMFpt zh#@+r7W`#n`VqmadXA~Wu1WN7;JCD7O#8Q`@z`QYsg5gca-xpUDT%P<9!M-Dnw-d` zY2lL@^|!*y11#Q%ZHN>9u!yf_84rl~hhxP*%&3!y&o2uUXh#{DDZOHAbGFESiUm9e z+4)4X6aUc-4P85Gc4Bt7Bw7s|J8Cx5{#{`eUVp?XBsZ?Gc~aPX7O)=z^YPthU05mp z!evPd&7@==6aHOE=)4P1+>ns3#2fLC@C^wBog%B_ie4f`FJ*}rqlA2-CDEdnaM8Cd z#DWh9MB)gLvB^uA?e7!``-=kJ5~}Xif?P!O|1oy;@l}uS|2n(ZIy;{e;~PSll13p~TJ%Wrd%dsw zx{8irrB^4qwBXfE1s7%9ZQ9r% z|4=jOH>Im=E_lXuf)4r<5%YC0YY@>13lKLK6!@dDNLhX?6d^jn(VLzCT8EtuaZ!IP z8usT03IqAJoVqx9r@0{Ev`s8Q??=#S4(6vdVIC{tX}Ha$`WoXwXrJy%kCjA=h|)14 zsOhNpk|@8ROzHW5dg#*W9`gAk5$XWG_fGdvyFqjTOK*`Stf~b10(1$sUSC6*A#@3} zS~Wdx#aj{>nBN^AZZc|QTx-;sL_{aJ(y_ih0h9b)P==RCoJ@JmGzaZIUs0I;gV(w` zz2@QT?ptCX_!cDam?;zL&V9{8Zrn>=CmL=S2wA7%;VJGY|Khbc71Pc`i;A@GP!=x8 zi}`!{^TM{4RXvTi_||5F%atA@YH(fQ_s}8er;8MjtAP(zlP?yf-(61^dMIOtN8g-1 zNeo^i5RCrBqgH)$cJiv7ic;q(&Tr--O8b4MqM{idN*?owqHMuDg4Zn;Kf?RB$ghz6 z;n+Dr{tGi;RO6Tt@kjl!YeJE}JXtn&q+I8;XKsRixo9q5!t=olI6Rz5G?!PM?eD~qdjk?(WpVMST1-l+>xk3&p_P+?o*T{<#Po>qkHj6;J-1VOpXPZ|;hj9%{?C|3n7`kc*g4d1a=v?~x#iw(pQvP{1K% zI96Wobon9NnHMEGlz<@lcqZB(Qi!O(W_t8yc*|b8NUA^>^#F|GV5yd>9}UMNSUSHh z!BQs}J17Rd9v>8LU+h-SXddeFfeOG>!s8lTw{+Y?% z`w;IbOYRzhz`g&sC(lCFC!;RHQaf*g1N@&pOZ@+paHtaq zIkx}5kxYX$>`^~oe5R=N4cZOa0#W!~oOYCs4HvstiRe#YTJqWTifSciHI%u$UQzxm z4<+v#s&Gz>d3=q2y8V5Mo`VT)+*UP9L@W>pQJbI#UUsXNOoN1;^5Pg5t`qzYfhhbb zq~qoHC2$3}Y!|GdGF;N|@_QXLA@{Y2U7;fh0&rUkZu#|!)NDB2-jkK!jJEGk;4ksV z{5J%GcJS2p(caTB`4xgFO7xXNw{YofTn|c{IHZOn!RO(&a>w)9Y>(~-GUjW4?SORj z1FL|+;E*errh`VMV&Pk+ zZT-$|iSQH>|_^|nV}^N=N2Lxs3PFUF0u z?c3~sG}5ehb!`$3eV~zE3l8WEa)iSGfskW2(*BZZkcK@P2O8;N!Drm0O+n$4a6(8L z>G{tn3XAB|5^&xOO|eRH&V({(icdn(lJ-1Gl&r)+-t!de0Mx z0_Q;VSM`eCp5vi5b@ht=L;>z_GDPHsYzsP^T*>K#Qs{6x&Xo?QQZROhGi)wu9vzNb z^lK>)V(kvcE13ofpPLQ`T|^JT_Y{c2T_7D@#EZ6!0fH_T2(pVohJNYixgLEfN3LKE zl@Z4MQtCW7>K`4BjPF+#6xb~^+)q1RcQ@M;d1&9Bh@t(~d3rwmD}O#}98$3;?M_5b zv>YjJ#e2DJ9Og7y*gkGQx4(m{Wfmyi>~aF7emca23*PY%C8GvY7V01v4AW3+o8OUI zJ8?c}a8e>mXo3lH_Mi(;?!O;bG)uaeYJu=l2&_&WtEfN7l6TcB3JZrYP|3R#-9KN( zxHVGnD}i9ZAMr#podc?G$`#cL)=r1ljGh zK{5@}@CVx`*D0zMwnM@8?RAQ7eb+;67Ohh>4nJ0{Q}p(`rdci09xcTTn;Tk&HDXJB|4?<$* zE=5-^@aRujro5*Wo&a;XeI-gbrU--}JpSaN*e#g`=@f)>9a+K-7lZM6k|7H``h?6| zu;d@jA=!7vhxgdkP1r z=Qcx&xpE=S#`9{ifZpd%sto~A)?y}_&O~o5MBC&$`sx;I`=@iKgnNxsm*|`Y$l(cd z$}L)-RDtlw5_l3Z)@G5!SWCe=sq7eQ)dUH@A%@9>ga};PL0zqbnM^DK#NJy9PU<1Q~uMmjBuR#+SI2W2=u!{-SNo5;2qzMut#6TYG z<_W%DAVj?380;_{REiVcf@nCw_nwE?2|XnuvIU~F?S#qi87JfjdVoNX*-q#$nFa|B z#R*3X%NTl66N#y}=?KsfxkeM*hS_i#ing2!5%{{1}(s#Y_{ZohkX6DS8RhKOkumE7d_s)7-}&!L_qz)}(EX+?mF~ArXdAoVln*pN-Tk^T1s^MwSt$@v!p!0G zYnCFIKxWWY4zdhIc9o&8$QJIGqH?(@d2iQFZY>bZuTVtXlmbhouSgNBlge&NEt?<_ zsLrO8A^2i}5OFmWlX)XIB~dH~qS2H-foSw6lkU(8RtZFDyD6PpgJB=DtU6&}H=k)j zKSLn2lID|)eX^7fwWIm&GAwy|*2vaJu3#KX9(M3v;dD?SG_jXI4?&ZoMs+?8xZ1&7 zbC)(px6Hqj^jPtC62^*`YfMo8orIun`GJQnh#<;cI)g3rBB=H*)tdj~8H=eHzX@W$gNc^(Ulx5mB ztp=tiS&}msf-pV#8-94?enl06X`G%^NQI9U2nFrw$#BUuNQjmoJz57DDGS;Gl$EOu z*MU^iTnAE3a|F`8%S<5EaGlOrW|q4;t*3R>qU%OQdequ0+8h=LRqdv>bs0JuG&NjN z!z;& zDk;l##eEBI=t~wzff|7jWgD2V+!(k_&}#&O%rtf$#&jgGi!q$9Zrg$+H~0~5H29Vu z>)U(thSEsR=!J|}lq%f_iieTHp2(waY>6!RSRN`+p4O(bKv;V+9ues7nRhSx3$9oo zXin#qI5q}`Np7`3V9#viz;%%&b%7h-xPw386OZ1j3k$}sh@PLAidZ2GYXyR_T@h;} z(;yuc!Hu*^_+E%T0W{J!dF8m zR)sHrBAYjD1#76l1;NUjH_KtNo_KZ^x?+Lgc73BWxdqk;ZiPV5^oK+&_jf=dwjq}Z z)=)trA~g-$kk$%JVYmz3D0FoK!EFGzp$qM>Lb}ke1nZ=-yU=f%AR!9I;ZH@%^!xvK zG{%pR^Z;gBK}vykpKcA4ljXtc9J+e0`CJ-dlN;2Nu!xETBiIJZQ5{DRzq3 zAR;!fh?7=n5gS;<#jA{njVvOx3NOQ;zDoRsvF^ozk^=gkDSriJIsM3$c*1H--svaC z+`igF-mFg)`R<_|oKv|P!P^*Fjho|vU?9ei%#4baajO0)s{YBU&IRLIFb>ee{I$AY zn6VwP1M>>74;8Vw)W&gp%;RdL6&+;1?B=py_u{;uU7$MP^w;9^@{Aa`gR`0Yi%8%(z%u?QsfXG^p4) zYh`q<6-;AgHVDsy0>Rs^%)OFnkWl^~m01>S;&J2Y)DrC%U4!ei$JepPe_qQTU(X)* zekLxamAZ}eh2C&+_s_(|{iX791wu>P#RDYMAfctx#j^xoClH12YIHHXEJH-+LNr`< zACy8{$Q2A+mWC%DZ+~VcG<${NVS!+5`|hA*8YHy(hwu38z?B7q^Md()j{mN<*D6Yl z_j(aw8?tSYKS)nC!vin(#?L&|7LT5DBDh=ZY2G7Av1-L#IpX;Of#C2Y_HP=8%Km~f zBv21dvxmydb}1Vf6Jqdr3i~~bog%qa0)c&ovEQs#R4utR0)d^JK;%wZplF_C8ssTN zbR3H69%hB0zY>VD*dpcObd6*hBsABKbb0pFaGQyBffq@k*o9ai?g|+F0w1ALAdDyBDKf6+D80&(NK^!9X}h_i(ggip|*uV z+I;+lD?OCruhm<~>yF13@|0R<*A*|WESTTf<{l0@!(m;b4u=hVwSaqrsrZ&c)m9*k zzd63*TM5dLzg5BF{S00J;gfw?L5PppG%^$&yTigdkl{ z|K{JaV16gt(3+6b;KRz*)>oPnx~8NoT1!Kie+*qk~3v0!C2m~4K*i3ut zenqhj9+LfKO5VOpQPyhhd>qnkgw{RyXbwJ4?KMseLZK6KFP`i*!x>Zb$j*he;c z^u^ir;{AgHQL~TZn@rg83a?g76&wUlPSG1NS3QTg#M~Q`>Lw z-4cJ0=o=mr7H;%Viy>mqL4m#iZE1!*@CGPb0>-q3ubjB-a}QAqLl=u`&b|QEG5vj? z;}VltgzA`jIa3>|hDPdbOr7|JP;X=EK3|}H#do4O+~<$jCQz9cs2xAPxhxaMH%F() zCps(n=eRFC>@))vOcSQ8}5qMO0tjm@hRjTL;I zKnQ*k+n0gWir)UhL+zuh!Rx`*iYkT1(1KS-^yQzdRp<&lGim2>e;^Z+NMF2E zDM#H*#f9T4aAkA;to6#~%`UO%0IJ`v{j~FNDgf_4^KL==4oKR?8{*iEa5I|`?(vsc z)Zt5+%}f%SX#!!a8#@}9&GXEOaXbg4=0{a@BZ^ zWEv!xH0^o+VNgC^U9&Bz80?JfY1*P`Y|(}<Q+m>eR#Oio#&_-l9iv%Zvk`>%?!);NksepQr-ue>xySW{)M!IzAUj&*Um`n9fp z=vw@>*3}PPE5G)T8^VSO{X~J#yAWdw1b+f5e4uW)V8;ps75vB_s5`>y9bG@{X7m|& zk3T=AU24>;8RB&55Uu7AsM%!`4!_|lCK`gde;LxTp<$vSxQTeyCJ(vq|5Z_{aJX0? zRKf*>Y0v$tXxk>N4X6LAs6w2{sI|W;I&Cw;#%``oXErL!E;Tx44|VGLfY$W^=z7y; zTh|Am>(I@jT)ikoQWVNP2(Lo9>VMbZT5~p>Z=csLGvZV3bE@%#OY8B33yS;%MfAH1 zPw;);-M>L)kn(|53}{p)Ojd95DfYwrMRDj z4zsp(oNJpPx#xYSSA$1b-5ZdC;WQ@U%}YG=ET^d=0oR|L@x4digB1Sv&o#E73awU} zKwn8DI~I}e;g^38Z#+Cn(dY*?*;s+_#u3Cgz7F&plU=Y|(G)>f3j`Uz)p6r)MN7W- zP@6kq0-S6hz>2y z?rzJ=9qE+2Ov_!ya?kqF^i9ink*{xBq0h7%uExUr=o0daLZ~x;Bx ze)?zcJTMP8qeX2kvLAD5Q?IqDXKkMT3H>ErDWZDjSBDHyub$O=Vuwdw!_pMw3(~Jl zyLX3&+T|C8V9Tz=jkrPFn+sL*i$XWj9;TV_GoHQG;(pUY+M8JStB2}-hW=%=HxU=0 zUi1qNSAif}Nc$3BZ}89^zrZGPuSJTg*6C_0!*XND zB1J2I@#t$!bH-@n2MENthPy0@+_*Y)qGZny2>yIbc>N+p?wwG?p0R|{#PEFv+oN?) zXy&?w;{<0Rc2LoYp1%__lnBO}%4U%a5A{VxI3IY7R8x2IHBwJDn|7H+HAs{M(EAsL z=ozLzvmTcaBd=fxJAeH8M7z}#lW?Oi@&hTL+zL=E#&Q8^46}y=tdtg{>Uy5d9&9m z@Zy0X5^aC>;qinz{5{FEiL(TX&UgJJQ;l6yS;9qooQCMU+xM96VyX_Nwj9HV z>ld_T5A5+sX2B48fdf0G3V9rZ64@`YfR-Qa!NvB0Sd^x5#*eVcZjji=#Z9Q$0e`+f zdQB*z2SKBAMx|5tADX$+gL$QXq91tgDET*|s7|Ak%Opplqn+^GLp!l%oF8YZR z5C7)eQRfDCO<9l?vPEY-=~Uq_TZO+^g=t`yM~@`ode}8c43vcLcYctqF%`f6Ly*1` zs$k$yTS7yeq>)Mbwc3$M7(_RLTR5LaC!yy#VXud$cLjN|yfBS#j*E45Us)kM#tKB7 z;0Dm9@o!7dw8hp;a2h&Z-+DSe2|Z!Yy*%O(jc4(XAz2g(6d(hU{J~!C=z@NFE(vc? zA4Hm0hv9sBhRa;I5678stMzm2tC#nA^reP%(n?Rj9vsi#Qa38r`xNB3N9(v;Do`fIq}>Wzx$179QIQW%9;2~Pgr+zDWuS*XI0!dm zvp)56U1M1|+v_;@oh|v@9r;zyE2@zEo~Xlx&nx=zphus$tp5BMo8G9OOztu=ZHF~H z>vY_3>@60D%8i9IoDaL6amaKO!<%CW#a=xPZ;rXDJLiyx+J^IKB;Vwk@;H*wuq0c( ziDXGclBw&EECLpdGHoxC+SwzU!`B(7{DC7KG>XMsiUeM|yiqcSBO~H3%?ktr!6K?; zs@M^Yf;GYN^r!%VA8Iw7m zCydoBuNE1lc|%CjzyImMC0s6zjr*yMu`%(nb zbUfs(`&>^{S2f2(bp(V4%4k({OjKL`KEWdX-ob|n79~HnNKwURl+l!NJOb^IQlK6@DXo4%2Z3vFBoN5dY$C_0mUG_4h$h( zHM58)2`ZY_TxqxJ_8_J9prf`vZnblpKY)juzGfD+^`Y77jFjs*BPhvMC7jePR8ypm zha`LI#nYWDgZbR0=iqDz`d3?a_$6I~dhC0~?Mk=1e|pA^{(nw0i(8;>kunSMrf%W4 zYzl_ZBscn6SCT~tK>eZYyOS)LsdTH-{Cf*Tgy7kVp2vJ6xjO^`ivz@t-tg%9WB$fd zRil5_D_V}W<4t|7@tbZ;Yo1B6C>`s$H~CccSJ0IEfoj(L_~w?r(3q;-X!#wHE~B^H z=u0*>wLlgw+zagjJmzVfFh;ML4|< zyTO@b`TS&{5OEpbYN4GI@h}8AxsuZbIgORf9bH5(buw*VG{pI2rQ3x8a5X!@eNVDQ z7yxxLsW4$$D(wWL+e=3He`fZ4y|KBr&$)}{~6@c z6cD;z3r+rXs-jA9P+zzn-?sI}gzDWMy2#rWak1*l{XHD6c6BYaOaDKr^p;=C8 zv*latv-w->XP;;ppTEgBT=z#tQ5e??mqn?X@68^Myz;z2tS}UzkCJg#coVXOXg=%G z_CJ<(eucQ@O}GWl-~1m-kJIP26o>b035R1QUMkHoSHapLigG1qIdU33Z?85&v0`kh8V@wYd16blD;3$ky#}WJ?am572xlWF^j>+{;55_ z^iOH{%>Owk_^+18;=|aH`lhpHcT3IgW^GGPwy15XpBgyjUW25zM(3pLMviW74Qz0$ z6Hl?ojaunETl38l2%Q^XENbY^Q>2D!>YO#i1*@dMEP*I^$P`>`3O2+SgqP~1z+QnU z_}9OC>GW2Xb11g!yQhChL0M_oc1-nK+7LbXj_|nBqp?9mBRpT+>7n~tp}XYobI}OT z!c87pjGTda1xRC`XfZ zCY%Pt14H};L<{-k__?PUC3Kqja~3htxR#A!lTD-iU+ za_2bUw74qc*;DtTP~*hUXq-C7srf5jmmW1DQKDD8mk#pKoYSoMwceXvz1G9h<1H_i z9zUOEQHvaL!8C!se0tT3YHe}4MdZ!?4hu1==B(S1)!6Juuqx3xxp zf-hMN7xz>Ngp;pNB5H$=J`HPa5r4p<3iQdKuaRoH$SUP%kh@18bgravQVQptG0r1=_6vX$t zvhlD<#ALnW6jzS`;O_<0vxkSAVNp8fcU14i1n+rUegn%t@l1=CwzS{-*%cObIaAab zE!-yxggW=etHX`>R@{*t>EI7W%OZYTms(q=ONsEk=5nITF3hs%wlgil<^>;Npv$o0 zc=1eAw@iUqVJ5jQ<7)Or{Drunz8g{>Lbux+6y*x{5`mD>4Uf1-VdZyLT*GlHS;!}h zjc&Q~jo(ndhYLFPfQ}cQWfbD1qt5@qO!_UusxFS@)vmU_a=QBYEQ{Q+vn+kec({;E z6bRYZ;D9<#yy;inMhnW2;PerWg4yTyamsj>5G@u6KD~(aIr?hJG)Qnb>Nzl}J@K+p zpB6bi)0aKd7oK_JEQ`|pL&8LT2lVn#gY78B5?)~~+!pjuyHtyOrPuYK5Vo7|OtmO^ zLW81)?-c8aLsvW>N&Q~YWkg(gM9FV9D9Zj`)8L6=+L{JM6H+Z||3w3@%t25o1Vd2a zAei5TpjHTm8iKD=ElLB2!>O1@ZEaALvQ105A2feADC%^!MQO=<6y0*RMQu*lqi8IC zwBDoWE&OP=N71gcP4QeQKFkzPX(PouwUOch{6O)s_<`c9+L+>1Qv3l^e3vPna*hyh_in2@tP(mLxO1gN4$oGs6rt43^4H;lT3pI2OX~*>!SrdNg&GJWy+3~ zOoMDJ+vo^6^m{EC-si&Tn{lp1 z$t8OfWo_3ni|=oxjo+iF=v*XS19TkS+i%m@-o|R|#^%Ph)7WOi##S5KXlymMv$2hh zZGHRxzjJo)oij7LbI<79=lS7@n#o=-0N1*(u58aaCp%8!!Ygk3X>%#?6PP-P>@?J5 zdJe8~yiv9bla5lg?rc=ykVr7dtRu4{l$cFTdGllqY{(HKfcx(LYZaq)jLD zDTjfYClJed*%ye!mHLkYwh@Uw{-KZ2n&MkvS(}TDfNk?fcrLpscX!&yI>fSF`0|Kk z%Az0_Sz1*vhHSuY?5jm|eq83@fLcb)R`!QtSS{ zMQQ&DKr!*9tn1gtkKc#!I5GSCXl+$+6K40at_!62CxP;AkBNcj9sWwE56xR1w<9xW zi!$#%lJ6R>b3)e79-zfM|MO%ffj#V3VYF6ap=q&VyR8HT&4|b(?ZU}S45B5jEcXoO=QM;y$(8LgC8~IA!N~|jkkH&MhZL(d zC3;OIKBKUI^_cn2u%qI!4JRW%@@SDWBha36t|~VivMcJ4HdbWffcHQq%f?_0lA-ZJ zx;4^xWPRY-BoPZl#tRie^D+C zQZ2vatVNp}kB&;~33ft{uuwZy_Z$(`&PS+U!vYQ##rSmg#2((*+P4(HT}?h-Fv&&R zpO`AQ?*zgVx%&3Tv|dvKJAW%-2xMPsq=kjLi7Li2v8uDnV!_LD<;(6hFIrRj_zA~) zA94e=MY$jqQvHL&rXVP4ylm0wblx+Y%vCgltg~MjTCEtnVk9%}5;va-ZDop$&>cEF zZ5OAE({4vQX3qe0;9$&6J1U4D=L{ODUA$WPN5&A`B|BZU`F1NH6WJWlJ#x zOj}RHnnd3ysi)e84pIViv9Myp%?q6duOBFFY7sZVT7i!Zh^^Q0W+S9%HYeBT$HYRi zCYD%k3pa2Rsx!~Uy7-IPvf&}sHITxN{KQvLebBoz?&^1_8_|qCdXEDIDIwm?e{*m| zB3@d&uCnzj_kN_0>0+1VSlX>3p32QVdg;SCl0TfiVBg`s8U*|plY_`Ye`_ieK|nk%_){seU~%KJq}UhrSU^JUCt@vpss=rLHN zIPyT~>#{6oB`_zT*R&*{MihVArwCZPhb*X!#AlykA~XO6zD7^!ZK{$h*ufi+32L3kyUCLd8T;&3DKkE{{@?&BZ5a-B8-Dhv zI8$;&FY;@*__pqAcl_ovCOoh#(TnkxgQXuD94T95X|+;C=`H~>7oc+cpZUziLT*TcC`iNE~Sc~iH49NGn*kC>M>AKydZ%N!qc;| zU^un|(ZJJ8{E^RPtEpPCLoz~KhUWcpW8#ekmMe5zO6CfXYVn2G@W58EEh|2^oqK+*cVIfnDPZ91Ul6y3RZEWw&u`^NkR`smyNn?_x0Xi;=36$Y?ZJO^yrY@>)j z&xPV3l`a*K^u4adf;zC2pyKKVvL%f$p_45vGVz@+Q&5?v~vqdwYj4x_?r7qRi4L)o*y7 znvwHvCS^;yzHj<6wC&17%>;=b_VLVCvmfLbhEF*^!tVGH9@jm=MoqSl_O{c!1$ds* zMg2=^#}*v-#z3rUM?@OY$(sGbR#py}0sDPo54R$T-jDis*-}#1;F#Q+qH9}y`KJc6t-E;}dCT?}R0ag1y zR#4b|3qGp<5c1zZDH_`$m3`PV7r>C(XUNC+xQ-a?e3`7_rwwf=R4if{i2+WF&dQUi zol`XD3a_3jTduV6VKUY`*EI4X^bcgr^L&mDEtsZI=tWt%VNu|#2?M7yqOZx2!J@tn z!3Q5bx%CK-bJog4widgR=QA&PJ#SMR~Mm<1bxB@IO*l?yt3ire?wUQbYByT~c80G+tbj64G zBGWH%eMnwf&joMuOVw2Q6XJ6Rg&2{X1;yNRs=;n(5RX{ct~_6iRtSAzPSueIN@Vua zSz6zx5h5n^rA>UEOZ`)~{7)J&K7f}5;^fY#x6|_7LH&DC1yf5b$**vQKIg9Dz(DSN z^3V0?+I8p36{0Q-q_kyjMRl@8D@C(RRcdp)*0XqmMM4HZ>8K&vjK6gW&rwd1#2qX7 zE={q0?__P^>^-RH@a4K(OmNG`49feq04ozA{>WN*pnqXsW8Ta^GOa_PU5~aQcZ@GV zPB0yrSBBl&w)J=mdEP6Hvt#n%nhk@q$40effn|d}DfXPA$Vd+D-(;`d z6x&tbVw!sDzqc`Okt%+(F|o=75A<}fk$p~SG#$#M$Z*T7eYFmgmA0+Qjl2qGJ|8IZ zc>(HvDEYDYK*-^B&mY1X??QnK<|TLj5qJKqYRJp0sg}wy_SAkK!QHZc-#SDAF zW-d0Ldt|N!0`zk52a$3G58tAtoNi`%7fpKCaC*+1Y`N9|-N7v;j%@!@jqq^Qce>8$ zlN|lFh7+51$m69uIdc;>00WUsvY_d5%8|qtcR{^pRbH_n0400Jus9W%Sa{)G9~M0PzMT%9~HE zF@{6-W;ch6UgwrCJw$*mANZJ!SxSlSw{@XE&NTO@G4~%f-<%=1iD&UG{Bgr371Gxr zm&H0PmP4+tFox5VjOkEwn;K6^G(7Yb%ag54Vy#V?G!eQ4 z%vi|Pt)|;uHz;udYf{I_kg@&Zlk0oR*2!2wKDPbiq&A#x$cToy^FmB)wokcNV@lY0 zr!s88y9$IHid+#yAIkx3;Y`c|;F*pU+9@mxMV#hn-LpX!$|4PjpP{s$e?2I^2DIV& zMt|S5kr;r=8%Omd7`+AGT=+_t8nPhAL{l~Bv&}G70%DG}3tnD+*a*m}O_4nj|JKRy zS~zGH#0)n==-=c<2OahIPj5rzkA)HP# zFY&DoZ9=dAd=S(HTf?lZQJtyIT~S4OMy)EzW70YH;XVj?G?n|QMh{GK8lo&+P{}&% zqJ6F}YR8txqGE#%#A|AKO7A|*!@oGI#%KT`o8B8dyl`o|Pu(+Pt=fqBhDlsC73hDV z72>rqhU#?E5M?LB_;TmLb+mncBrf-9vQs-SUn6h_`(ONo2&>p%mjGdIB}2v@TQDWpNrI6MxkTudlo0}H!oBrQ738cx z9{6W8fN490s{505I{JPHn`Kn)RBRj@aEiTzqo*jx!J51|-13yO4=Y6f`&+?-eUYyH(B9kNL zZ-qE#wCEx~!mgfp;7ts)O{IG>TC(joOG>*Exf#&!`faa<2T9AJ5t=094o5c$cGGes zP=c%Ackz2jUETW6aqQt>0?UbNDV*Oe`XO3~kh^`2V!!Wx3kU6iv6e-jB8GFJzt`$QTCITUafZ@?J+wuQXOvmxTDUdW+^$@PG0?RTj{fv=L``c zr)Q_(@&d6R+BHPigboi;B;~IjwR0(DF^E znhl!lmh@(M(g?k<4vfu^8H2d|Qhk23E+NMiv)BElq)EcGQbDP5ygM(>I&{_kB(bywhAnRcvC97{oBK4Ew7WJ zMw%Hz(uxn{yId}Z-?hphQoBH*tD>8XRDzNm!VQ*{`9Z7JytHyVt5iV?%JmIhMr&4K z_+o@x81Rj`r{L_HhTOjzu2nImP;4Fh;N%!d9%SjJcw<8}zJ{}ht66y8Qio7d06cUK zW9f-Y7)?4QL~()Yb(wTVh=1&3X9KC5ybl1L1Peb>64f6qnOty|Gdq{QA^JdQh>n&-ou?HnhsO_u0^ zKkDw$fA#Mswxv^q*xqqq@5gSvfRHnrx+*{u7`Kgq1Tv`tF5}9#eBl#X`r;iOH#6SV zrxq#LCvo}8?WgoYZMrh?N<;qCh)Yp!9p_#W)wg5Ccw?3P^Wc*>%+8k8hUpnsH@RUe zzbL7W?DPA4}T1G26T=Geojq-#t`e% zEbzp;k8bnIn*x!cD+IXrG?ikm`{Y&veKII#6Zw~+c zts*N|hL9^#_5_Lf?jV10h(o${d0OaK9h(6g0N)bV{TVI9;ghz>UeW=|_2JfpQ|Aqt z^^o2m*@RegTq6??UrVyCcuRM~Xd@wQHBKtG?wFHk=j*5WwqprzFt-qtJRHa(`rn{K zLi2VmxuF%!PRSjij|FaE^g=sq?EQ4Es8-$fXJ93&C1WUjFy$k<_(~NVwmPoP+`$wa zHqocIt~66!m8Y?eWDHd3IM=+qTpn3#urj~c&d!}c= zNE+`j_{J70lyFD(^P|8M@vXDy871+tlCg3-X-DDJndknC4Ada&buqE%;@*&U98adY zb|4$$@PMx5^12n1MKw+Bu%?F81@+cWOcxLEg4%WCEYH0N_hWEDOQyu%mD!gQ4-pGNS0l7}Fbva=;dwP0EvpBMpa zW(;gX;wid!XHcd}cBM5E%=#N=Zlll~7IVfhjM4R>q*=|H$726nmo(u{d1V2kU*XEE z(FREL`!rKCdKz+fZ(a!1)5+6}c^CLy(so2H#zw9?!@ghU;nW)v(fa1(UmA6A_aRyQ ztJ0+qw3gOk;16*lBH=o9x&XxzXds{=GyLnOhw@;!JcbT35@x>iLRz&A_~F2ON&Lmz zWi8%tIR-&HL^$5WynpkP)YBMGl1((o#B<@^TqBTE(YrCRLkF>XvJ;>W{kIc4;gaun z@v-c7+YPxFM_dVoc#-2B3Wn;)cyC^$_V8hFODnrXVfd7W0tfM+TW?!3*a%?WFeR$> z6MAHM@uy`1BVMzTb7%2ge_kWS7 zW4UC)HSnk+os3u}>s@Hy`2v*kW+Gc5@9b5C2S5kN6J)$nvMb53H0>vZ=x3dXXDL7a4DIT-Fu&o!@7jGxcB z#kMlBWZoNNea6qh;l&pbJje$MXRu?@Crq#143^}6KX4oTYSQ@3q{~j^%6xNs*2>gd z$G<2q<=4_|E^_!Tt{;J-DLP9Qt5!6)w*S2(5VZD)r%mt{RK7Q zYX{1>Dlq{$j-VW3G!Fd9lYH4AhhGQ;0H}~`9ICu34=&XqeO!x$6VTdK{2wmzIJKz~(E4v}%-9&pn?Lm9ym z?m9X)L58w+g9euZ4`K<}V(aGShD0I=sPaS%S$dPxHHJ`FhbA!yw8Xm>nHI^h4^oCv ze#F)Z?bRTw z;u&+UJGYvz{h&QZE{}K*0Phz41P2CV07bju~W5orr$Fttluyn1kY$9v?@DmMT>B1H2_7M+JE|ZYiex)M# z{)L^U5nt#>_=%CWo8c9Uj1jVw^mOPC!X4Hhq)Pu9Ha0|U+9B%8fr`YG1;d~UR#$Qe z8{9vvlsHd*SZyV5*KJ&P@F)1Pk4nw+uV^PtRwya>7QPD6Q%d=7qAfl8zG4tpwy%`( zJOwjqyO!_%U|nzbG@0-N8AOIw7zA;d9u8y-F#mZ-#hp4@qPEw-j3K znuJR1V&6&-Nh~i7k&hL%#W@fnGC{EV6L7@&)_K{~=!gdAI}7`v0)4@Cx zZvP&0UTKJ(ai+ZhRfYV&)cQ$^Nbr94II@c-NeRAJbi)>p5Xbh-Bn`vtn;-CCCZY_R zj8@9ELMQJ1XGwnsU1`r1z>?ZU62W-uI^s(1!rE#{Xuy^ppL^kc|DKo$G2gS!uL^Te zV&zuOWW;;OHD(#cMaoGPzD!`v1?_H&0GGpwc6ZZ1Ece|`3408_fiM=S^#%5bjeNX= zJ%oofMdL07`=2NnuJdq-_xEBbY9tdhM^k)b6+CMS;PR4Fr89ab{$e$OUdnB;%5ja`V5oH*cG!mN(;Px3T>Te zFu6SUUYRw>cahNP&#dl2b41;OSm$F37_ZGK=ru&A zZnS3)6V%7F*(pDNS$HY9GFOu(#V^prT-mf%zWy5Z4{ri%1Te-NHO*e^h1$WHz?E_% zTgGcNN%`nG7qq_&SU6`wl^QXW|@1C#2ZzVuU>RJ1!)?w9iIxgXL*Ur$<^E(XM> zyoq!D|Z0pcKDlHXkFnR+f!K?2Sc%D30f31rm8tqpo{y zX-e#gEHD~Nx%C&P_yam$zs9I(>K*4fv8a&TZIU(nBT<$FugtyvUKgJ-6H? zyfKodaDnUwWK|a;&NW%D7$aJx3=)cL&*?HzK0ySt%$oTEKQXWL8h$I{ziA;L)wwOO zWx7rLf#-67IUKP3E7R65Hk;Eh4EQR?>}Wmq2Qo`c`(>~%%9ztD!Q*UP#IE1*5*%i% zDB?-mF<7Q8H2{^WI;m_~u$sV{%68x^ae@#p%h=SsFe$?)CT3_AjawEdE8sc+9qk)* zB%pu->x(2d-=`F2a@&dTTNY*>#LJ(W>}<^k5lZQa5ygPQtQU%id40)=ozaTpxZQ5V zTPXWvzA5=syq})sv5Hg?4bma3fnBhz{RTuWonW67B{U~gDMsKU_#W!%B~da z1Ngs4Umc?g23I{0MS=g6Fz%EM7pgWct3u}^7n55(qR(NN_YhhKi!C{VucI9;f2JQy z%lse=`)v`wCq2R5Gld?mP=dTRC{zeevT)=X6N}```!vgp+lB)rBD6I9VFub2ci{;y+YGvqh*Z+6PQm5>3Wrol<(eXh(T=FMm$%=eNnY zZlFS#Du;ZLx97Eg5GsCr+vYLDy#C`LP)f7UU3rKr_sC66&osm)qO9ed9m<%7h0PeU zc0xt7A~ce}r%OfiS3n_d_fX{Hg_?%f?fU}CR|VA-FuoNndj6BDI18J1T`4=_ZU2r+ z>>}OrjzNwsM%0UaT{sfSrnugYi-ewBKQhKw>hrooaXsi;ap}0N)+}UFHz<8o8mko% zvl7)+>o0n$_(OS)1}yQ0xX}kV&jpkLK0<94VktJWyB(XKt4xwd8%At)BKJ=uiJ!FC zgiy@R$ZE>KKXShjMlB5-=@l>*0!YN#m{P4$hGXdQW)OpOy92pl5MQ72p!^FNz|B#? zt9W^yK{Xjj8HxWymtQl1M?v9jM&I~cB7p$$;dvD1k!hH1r9{1G$MKR^k&IR&nX|0% zDE$`2y2lNB;x=c*$$3nipiOzzi5~{Y+H=Mg;09KCzyl1J*fuaOJu=?XmD)b>_`D^v z_KW?e$%dlhrI}hEpV^Z(QbMxD41-hXX&9fhc4?09`vO!?a`ZPRmN#W}PezB=Zu>Lk zh$ClI=q=5T$V;_<(G$t==5Awg{+*>zit|9ElDzN4rNj_arUd;c?EM-ym*7J39*_?a zmEUmQN~cG2rFdpEiaYi}2b!bC_L!Vu{%5MwROeo+|AEHrR@tvJA*8~lEtYT#XRjHhXpdeH889QIo>R#nZ{hn)TvU4e32AsHrLDMpx3yEhGX zCd{*0)?F3@*@R9BHz!O1;sbPg`tN7K7EJ#if)vjN+&0&MAJ;5i;)F|hFuWTRqnNHu z!eSC5o=24m?Ecs=U0+MU-AK-Zw!fw|5rYvbm1&w36DF_2(Ci1J2_&A6f23 zOIl=S=;uq(`mWr>L09n(pHILwZUr}`TSb0D+2DV{su7Qt^Y?&#OLYwyaufUQm)Jkd ztNAnx+HyWGxWE$49&Wze(JvTXt+5NKsuU|NOBEzcj;`UK-$Q?3B+Kj|iLE&XEXVm4~?DVNHZI4^$vEem!40BkMzqA4Bgqc9CJgbhD947qxI?6{aWH0&0lFZVH3!_IN>qlm& zkHPfusVLkdixfRJqBHyv@K!|{*YDlGZ0IM`Nk?pUZ57T{c zW!1v(RbYk`F*PFthU#EAW2RY9ab@FJAtnfmSTfDAf64>}Fl%nMayt*euU9&k$}cCn zp9WuuS9A_%xEYf_k>5Jjdsp307YaoX+luw%#BVYeB(G|CvDt~j^9tom`kutu^;>-~{3C>zkC|Kkh8rBF zRG4=~%dS&C9RDNt58n9EkH$RKhvN`xKBGgb)e$OKB4fn$tf-IHji>p+rKj;zsIHk{Qk=<>biQ{R^&I_Y~o3eT|JpfJdiljG?pzT zPSQB7`a4eNhfG~7LP2?C_x^$sboJ#4?LS%rf+6#keHmtQeEWd01S?N^cC{@u3OL21 zNk5!Q7-jyo9iC1!OMIidsyn4P&DzgVlP&CW1q{e{z-r@>xH8RwhzMwb96PF@LzC*! zEgk}Ph4MV>SRoaB_yY*=LVB{gyuj96?}WEBA-(WtlV>DVb}lff&tGzXj%@Zo88CE` zWvaRAn7U!6aegpcG)u%Jjs+un`r6-8He@M+hv};3l*QE6W=-}N16~jb)9jd45fZi& zrn|cVB^yhk?10vQ5({D{)ao|NxIv?l>4m|h<*=!@kcfNAo{SVl55XO#-WOer^?LV) zu_9NuQN7J^*1`gk&K{B|=+mn;bc@vxm2fcUA`IFN{Ymqk-Z7cXQ1-WV_xD|5Ix&t-yHiGaCte zjJ*w$z1Sp7RcKNB?qT1ZC1<|X>*Kz<(T z7!ho1HGXSlk$9W%@IXj}UKf2V;{&;_ftHMokkIi=Ac>pEwSw+-0aLi3eR1lVE+Ps# zJgDU|252$KUpxX>B`+^*z>V=xM4yeT1!txWE<-wJ3a?abe0uc!A$$iR4_i>m61h!L z1MXX)*4wuz1LG*a-@rvYP<`Pv#6oDof(Dd3cQVw#%hs1H4dA1&{XLxBU*}s*P)}WZ zdSU>G^QZWJ0wa3wGI;v?gN$!BISqrFpt`YppSkYzs_7JwImRdu`f$KdY}|YUhxg z$Fkg}Nil7GrpdYx`9?M&NntQt0_(y!&hO=!kV0j#jX+cLq?BPIWrWm-I8PH+S1tm@|Bgg zkA)nJG2326fNs=O&PxtJmyb6Qp#Ek9&Dj(jPEx0pa>ia5_I%Lerb&}YaiyJg=0aEa zLRZa#{^s|p$B@R(vr3b;{DKJv`q(34qqeX#)cUk@qhe;_3@hpkytl=p&7j^}Qnt5g zbFcU`Vz05s2Z%JHV1?Y_*xxZb;OzzQ(p-v<=6l3^e_=8|*tPrtdKWN#&8wZl8d( zUe}o>8akR8+V+IACBa9kaV!T08(f?O8}1 za*O^=P|?{dC%;z5wVL3l@=gM}r+QaBgAQXj<4NTzYbu3B!jCU?CNJ%( z)VfMqGFJY~98f8|Fid6v>=&sU=eG5KqfJ`%2ve~B8ww=wrCjS548)Vg|kXvM# z{Wal=!baq6TdZ$qj^ZRI1Srm7n{BrJue}>G_ez*x*hjR`^P!OIGAe$tE#73A)!?4G zs$N%vydcf-oaCAjX6?pr>pY?rn^|YuV<&DgGql|aXAOgon8$C8G9az%3TvACWKCMg zZt6Kkv3KebHRr6r_-%>_+eAZ}`XH+SjU3wdS=^hLC%81L!B>+3tHW5%`eW{5f27&2 zG%Lf5qs?*GOUS6}(&h(QiOOzCIaG||({0&^%2=cr1vzIYIoU>?L+7x1ZxPJf=FOj> z+2lPRQV%?8|9dvFXymBB&$?`K2xbk-c|-0w$Tim>oIMOVx*2zkpSb;L&I>UBVXjNt zA7u5_PVF3*+iXgMv|Pj)M+5&G23yFHoVi!GIq#pe?53?$!H#}&4T2$8)Fv;G@03y1 z45Uf^5TF`k!s9GIrO$Hf>aNk-GHv~(eZW*h^As!=NoTgu2(Q>UB~SLTepS70XvFQ_S{i&yIvc9c` zn-jw#LGno!!y?<5ALO_5=1J@8R*xy6`zZ>^hTq4?kTB50u+PvLXNqXdF$^jE&iM<$ zPa(LC2!txPg2SxU6@cp?C>)IKk&BC4`7AQBnr{ z05#q+9_okk3vP<=J*gZIll!_CT>U~;u-_V3eR|!1ZY9Ysv}^$LXpjc`MsZBu{K0nf zMH?d_EVp`8BrkB^SwoJU|N0!EQiYYkqj7U^7?~{XUMXB-_i1%XP}7y5ejqJnE;;?o zAlLC=KkH#{e9VA0bwRBC^J_--&k_4&$O?q0S}0k;cNZW|?@u9&Mys3F{K+T?wzA~{ zoM*h#4wmOO0ll~Y_s%)wl&j$B&{O5aTRL-nJ-^artu^~q(NGN=L4lTSVRL%MEqZ)o zzKI1x`%Lx*ojP2Y2=<2wh9bG=3qKxyB1W?Bwv&C0J%>y?ev8cQt?C&1S6Ti>=YU|^ zL5>FwIrMFAwZQWcRcv{TQ!c2y`*m*5n;vZ7t}x6*ISij7X{lGXLL%%$N?+p$X*t%% z-S^TyPP+YXXDb`9^HvCS!H;Ah1EUO@02KAw^aB$UxC~OG3O^}G?OV6sv44c*AMvOR zj6yG`OWL@rdPP6D^a5JiEF${VtFTjP&|xwL)#s(4!Uj;u*t|H2LSU5{V5L7`*%mQ>Bl?-w6&oMRiMI49=#4h?M*W00t( zpnXNH(YemunIp5wq%xm*v$^ywyD_r#E$6Y?H-pjC zMHqIIxLM7IPva5d9fU9MtLx;9wYgD`P1bAIq{BVMW_KS^G-FPnun;g`IKdg7+W`ey z$b@*^+Hu}t>o-aB$tQX_BZ)7!xUXofTOFFg1&a6=`i7kj_KB)g$mH4}c6NY9l88x< z8#>p-=4-}cQyU-!w*KV!d@pq1Rup=fXc${u@nP}!*++4Rym5J1Z~5RbMW0UfjRQ&~ zbt8H>X3A$g9un!w^rK;}KaPiqOk$A}zV*LzzZ25!yd);v6O9IPcYB!CY#n?rC<0X} zd~GS-Di=6eyY_L(A~=+H?ELn}GHcao$ux^#NN~poZ<7A{e?~A4n`FG~UUuKqCK+jF z%?;*NCF^ZQrNvv+#QeK+WNLR?jmiXHbe~@*h*!Gi`0Fh)3Hd=?zM#696U0!xkY@NG zn&o!b9Ai&i<$Kx)=t&W1$74cZ9)8qJ+m!OH8p?_&HkM)@LcqbYSmeg%3HmT>Uv(4Z z3=c>pkorOLhy6JEW5(wL_`$0~Rf1w{r2O(HJoidRY2V z{a9YOzY0qHmY?pvD8;}Y4u}>7yj}-1Um5|ay)70U3K8r^A z;pyL|_#`xv?3xPk*WeOnOg^n4A znN~KS>5a@s#oqO~K=={o*XCV0au?L4w^x#K@}E7VBy4cph3#8d;d$PSqd~vFSxq@eccNOZ*bS5vBr!6u;eXJzE?-V`I-{Ua6{1UR=I5d z!r&mCWM`32z28u4XDZHZn0XQ@`7x*o7N?2I0t3V+U3Kg8D+7`LwuA98>)Y)ds(0jT z?I*!QGZ~9D+fyejW#0S3OB_xa$%pAvRhJ$huUDg>)WIPP(hU@)_)&PpNDe;M>|2Vr zqZy5UB+07a{sxQ9oEF-8^TLeFtrzIAMp&v}04A2lD+q!<7FT|76V3DLo@6qi`*{7% zjFo|i!Jy#;uJ)yGj76$3!s}Y*LZ~M=AlaDMy4Ux2c^!ziW8y#qDT+gjynd9qwa39z z-qu!TRhHd{#?Vu%O6v^VVG_DJ9brzDwhahIY*@VvJtKC-m`T|;CvZD{RnumB2j6?Z z(F?d8eDitxb`-?3jY6LxA`|-uL-4wrZbH+O5)7JZkh!swzzhxO{pa&BPgGqWRdVZ6 z_p>aGNR*jF4EgN|Pf%X^R;9iflA7UYe4HV&fA=n1#_15awTDn=rv%Dq9_GaF562)L zs}4eC!;P_TBUua&Jvx5Uqz4{g(mGXpr(>25<*=g)u?sgHNY_^{Rf{^S4fo&$%2RRm zoB%uCcrA#ueV`V-{%p(l(bdza!tnl#;k&$+M2xa_IK&*?r@EN2aAW4ep2NlE*nG&` zGDU(A0FP2SD>3`4({?iHf2!YF;e#JkQORG=qXs~re zSqc~>W=07!@c!FTSy0|R=mu?gsF@&b*@ZvC=NjqMW%}kk*})UM$IKeKhQ+ES*-p~*PsqP_QVv@%oxMA~-iflKq3EXuuDQ zj?q*$F6}*f26|ef<;`z>L;>>hDaMS>MJu>SxBE%Zk%0{vDE?MhQ}PP(a{#&uvoD>{ zyFIs2FM+mbMfR}30jr4D%8_iH8zHTLZu2GFmZ==W-+-7I3=j3u9@py0l~0)$)Q$;-(U5IeF-CH-!68hQKC~RsEAvMjp5iqIXG7n zl$Czir88-gg6T-XhY6AIE0OaLQnvW4SLkGa(cyO_FJlw=%Je||Bb6BNr|>nr?Rv`A zSwO3_{m8aO?Y+)?JrER`E(5yFh(ZhcKa%^L@TGgbT3_tP?)#UM{(HdXmnQ#TBsIQ# z+KDyc#U2dQ>B4_Ux78@T$8aWJaUH}vS@z^@9?0T`Ob zF_%wxG-^Bbo6?l$CiuZlLDQ#fMr+63;uO7}k6AjS*a7i%Bo*j`DZb1`ggt1%SF<$3 z&Z<*0JvVUSWX-7To;T--c0Xd{0Jv*K5b6N*O~3hN_pnoBU5FT4#dd88K5iWuRIPu0 zvy-Q4xhQy@+afk5MYeflP_`0VD$zVTP@c=+@04}jl{3j)|5l~sU)1NygLa48l+`al zv$bDN!@DwZC@5G&*FStx+a{1wD_rcgCX_E@{qX$%aeNEu}z9Fq->IWATPkLNU@78w?U!X{D(7!%#w zF#jT1D9AWTX6KC4N4m0SB9mci)lIaPLu{_IYUbEqbXc5&6=vw+C`ZdcH7pQ`dX5WfD=c>AFoRV}`}O&(XRdd#*&Ld$@VGz&mD z!}SdJ6Q}sftph^NGay5U-S{>C5LAGpe_H^wii^>}eg%;!TcCJxCT1+3l81A3?v9VnIq@@|$eLtsGP$8VFob|I0roL1ftR=vEwr)8%1o zm)5<@5ER@vUK(O%s(eGh!e5z+5RgJH{DP>_yc*ypP9qohtykCKw3~?6g7SJL!oAGH z%NzVc^tC`$)#)@EwXQJAziBulDpN_iNBB8Pd;N;7O~!$;?hAWhm)I>SJV88L_#l<& z5zg?400AagdUAPWfB0=k3aAaU0_cG0O?laV#*(DvJqKL1il}6_vF6;lx|DJ zWidVyLac;8txbC${+jfCrU8C|&Z0**4HLfKRE``?Ltc}CCW6@FcL~GVll)?MN>`4g zn^G{H!^k}+EV6uI#%;WcQ8SoPxv{;w@5} z&vi`&IA=!RTU?WKAza^PEXcrp{eVHIEIP~J0n|Dykp=&T{6pb`)5_k|aRAJjyShu9 z#`1u`yyD|EB1ka;Y>N8+cXd%Q!qQOljPvY-8g;z3UH)ap%1XuG(bT_om3*_i?p5Ud zYs?mN1>xc^Li}0evt!P0(^bSaq3(A~zXN{Ekwm~)+Oqvi&TfGl;^eVG7XTsK?fx`< zl6@^=kDaCL+NP}HOwC1OD&$JN+V9#$2J`&$ef%~yA*S%f85X3*#C75^2l!dOa!us2 z|HTR+Wj)l#fyfyy=~>o_N0s^v_@cnWmEaSU&>rcsO2PFMCW&K~RbCKw0muDXn!h3J zQ=wLK^l=6^T4aBqVD}CGY?pUdVw4WvRMw-W-)go?2yY z=O@IXcPP553;N5HM-#vHsd5DC2XN;_n0eLhR}k97^n?CO*!hc}C2qz=o@T;M>j2u? z>wFHkS2jdM*|#>5NXEFnP9dEd$h1$0pB8-1^u+SqDVQLP5wjCbZrU$ugr61s?l5V@ ze`}didWh4-)%S1bun+7}mw=Z$lxDGF2)>-b9N`MjmJv<)`ZPwqnH$RyFMP+|>1`f>B` z>`!wljO=2h!zv=M>(*(=F$1}Le8ZElc=(e?qwo_rt^vZ1d~l}dyx69vQ#wwfaJgGk zv@)+xe=z+OFi;_ils{)v-(X0X8AQ)88o&pNZJ9y@vF9qmat7u z0>Y#Raa&om?Pn!xk-s`&@zGFcxHJx-TSo^|H%Q4RBD|P4@#Z`=?*{7G2|yl)+k2?U$L& zU2OZz=CFNxl9znN^m=pn<@DxW^2KOQbGYngB%~vp%dQ#2IW(_1I>HS|1n3=h${@Fw zeEG%x{6SagiQ;@NkdW*pZ^rqU)|C_ptL2P&Guex28&HF2EALmYOZJk_Pun=*_ISxl z7wG-iV%f*d@}N*IawCIi0`Dc~d%Wa}_$4(G6vVS1>Y&|T!~PtjXeMwRmtl#M;R4}y z-sVKO*Lfr^&$lElV?s7eAjs{w9B3gflP8Q;lpzwAL*o3=ie_70N?tlzQCKi*uvo4e zt*F-WQkxy475xq&r;k#!QJ8-v5JHxL980s4y@)w1!M6*#ULeRm0T~vAmw=3PpSM>C z4hjU}$JoH*rCYTq^a)hO(rtj(tCwyMfd}_KEh9Q=S_=e+RiH=j^Ri8rA?R)bL534a zOtzfKpm+~K_Y?@S_234@cQ6^Iaccx!D-dLxKz896MOF*A0X^dmLGKj^vaKM4ZXK8m zGcbcb=pd^D8G7xZAme_&PQ0-VC~&A|6sjfVLLEihfkIXM2^2Tpx~TR~y+G*D05UY* zj>k#kHB>tsRCeR-cN_=L#bXp@iy#*&bKn?7LxAJ{eOQ?mR3Q+8nu84e`&%Gue6%ZD zSay?dDA#3m(|;HnV?+;IXx~2`=cV-gyfD((g-br(i+z<0-0F_bKo5Lb(s%R_2;n2K z%Lx~zACC%wrw0f+S0Knn8L~c1hM{V>phpV?*<&DsDG{5jO3>2;g6wfa_B6}Mx5^#RK@KFF};R>#TTsO4{D`M*IzG0{er-~L1|Y55!33B8fl@;9=`*oiFvb56W* zViWnYBrxLWKBw86wAf87b}bke6K!I#dyrzYH!b~Ec0NMS0o+rn_Z^JtuTGGI71 z#Br|h>d8Qb7p0GD*~D<@1-(rny5H&2aZ-337tU;HB6ge?3!3=$0ZzImg++=$-7`%Q zR#TW&j?HSyQCOAugLdhRI4-YyUGU87Ui4q(Hj~%A=-;4^y*n>`Mn|sGP91GNwZn{o zID&XN&mb>HZ~0tHU7?qQ^>!rS7rH!{>7foM8Otwc!!LL+l1Z-1@#vg<1L?YU3p{kw zNnY|+TxcdV{B4~LCwa+NN`0`%+z)&M@cDETxTK$qxl@MBo#+-Y>2os13-ab7J3x1$ zVB};k7Lj+N%naapa9S*WSs@VC^udw`gVVPs%iv_FmI)R5S$lADodRJz+^!LPtw3`1C>L&P}0-@EN5WQ-Q zqQ?xkEWsKoa9a;5=ssI>gN?;PS0d2n2D)0%@pNajpvMXX*CW#=25u7tJw+hM z?gkmSoqQ^6P|nzH?1wvbv_qOIoS}U$hxWY)?MqJ8p?xnx`%xq~wC_b|e|oA3?Rybm zvN1dif!A=V*WMQ%$;Ug}orWBG%!OFJ7l{Dx`VOSAoVBPiD?m?hk+RdhBvgZF9G`tV zfGqAR^apWIYwaa(Y9<=O;6i$pITawU7I^`BiW!YUPPij)s7&QA>n&*wzvvCBx{f-Q z9)Z)a)l>&0x0D4BB=H{XGJ){zNRXkWs57Lc7^)hfLcQB9#diiI@%G6I!LJbrZljpn z7)95Cj<-)zI%$0t0X2M-q6ues^<$A$OoK|#6`IFDgNoh-6iZ2YQ0v@MAf)iC0g#e{ zZ({H+Z(BiU3AD)`9Ifc~Grjup-UDZfjp;&`ArLm&yS!E4O1#TkDT=W~?DBpDR@mkJ z5k%ZDYlMd1Wx%p3`7CLe>1Rp994=xi1wz~^W9#6v*riw!ED-!+fv{o%=+H(cgO1zC zSAyOy5M)z92G=dN$#NdjuFDk&GXAbG$hLxvo!=n(^9gI95>la1Tr11xr<**2bnR?6 zud0W&0N0cdk#ht&H&r(?9zjA#a5KAn!+$-rFx51(e;+}RXQ>o6d$Arv}`WrYAk04&b zw-N}g?nDG)1Uad#a7z)ap(-Jw^+c4mVz8qBZ?V== zG?X+=(}Ji?b8DcvkV)Edot(^^Ooq;!Gt8w)xdtduE<%NZAVmUHLB9#$|##MqG~@ zO5;huRibvR74pgA0_OD~=P(QwO9(kFE?`~{ax-ATvaNfNhI+3^@7m}u+P)rS$2?WE zxd%Dm^X?%2bQJBipLdJ)lpf^F&!fLk3r4C3xtl|z2iXfxD$s+p%=fwrw+Eraydbv+ zIUhm39^~7HMJO_k60duZzalK#z_md2ARk!}p~#%2(t|+tl?#v?ss|Zes>+hrgRBFQ zYs#kcXK@d*MO@g{>p`wqAT`BBZ59+tkylgh0yEXNUnlgf;)0vkgZvse?m@PTK3)&< z>BFT5={p?dADK?el+IocG6E9!AQQs<6>%Yj+jI0FH!utbM0N{#&IKxo9t=ImpAMHE zr1A@r#=Qb3fg_Dx5Aww?z?$5H)QVy(5j{xf7b3(x$msxa#q1Z1*Mn>U2780|K^)T8 zD`NV^g*dNbhQ0`KTt_wu{c>@U3aX6U%44e6;7Lh!`#%m^bc9#8QuH^D@|F%-M}KFb_ChS& zVl)4C{^s}B7oxf?lL6bMeCFu)2+JPD)$L!*=zWCDIhw26`5dqQNaUDQx1&|j{WIE6 z4A|Cy$kpxs3uke4+aWIO|0ZA`(kCflf?xLUs7P$YQR)=6a5#s95#RytZs2VV8a@nTgz%MR}w;7@!DvSCq_lo zwXJokt`)he!Lx1yQ4z~?YZto}t;q8tHRUhzl>abP$kCAcHa71p-_Irf8;g0crmp+_ z1fBu<3D>hpcnVII9uHxQsZfs!^rZ?c9$dhHJL)mXw>D(V4FeWk!|eVSs9nc!l~TQD z?-HQUkf>WjcP)ufr2ETU#V%%j_AEi6i0ndGimqiobq#K;Yq>fPHAE=V{}qn)4UYA# z1{F(%ZRfB@Rao^`c~|;Dj`<45LJE;yEVh?c##xwJ6Hq zx!&cuO{gv^!1GQ&PYjsf5p=~RsyKn?1K^2v=@szAK8TAVL4l{&F70;|Y{j*sQs@zJ z!HxTR)Q&Fz#{=ergnWp&fN^JpC6^9{p{~sn@&a)I8;5S-_Wh&Kg5gA8c6kMp9t8G*NO|+b>N29??()Spv#54MO?sc z!=-PgUMHbr;s&YX61r1dP;Y_=X!3$5;x!>piVGNzxB%P#%WgYdeVIyEt+;^k0|J0G z1ICu=7t3(f1L;hVILSj2Gc7JSY=S;W^4)+jht0y_N|3=zW%dN~1eadVmK!qm5|i z8)cxhl?O^sBP_duTlMq#koVgNscz!N@oO{Cs~ooi)k6)Gwl}G={C_Yi21?%rksG8> zer*;Hlop7KZ21RZXpo*+A%h4PHBV5eI$neHS1?mUm>QwiiVJRd!G(t~A8A4bLh~>o z`n(H48*%2iNt%bN8H2K4-K=!R6C+#@{s9sXlr{S{q z2RmnzK?BV)P`a>L(zs1vw~LE3dIP1~!Bq{ExNI!5Tr#Rc06Jg~?X8x_U&hba*A zDa7QaqW4OrPrtZ`xY&hJEAj<4-!2@Ea!H>9Nmo8R8IZ_WnT)gE^6J}muil+bM7Llb zvLTBfxWm!f8ljp5E@0lmrF%2{`CJ@yjER2N8d0a6x`j)xxZs7Rj?c-=Yl8`9o%^og zNu#f;6os}fxc*&(?6wFkxb)5k%{ICnx7r!;GG=Yd8|s0x;|8z9ycA^1L^t0 zj1X_%_HM{pM1SB!oqnuJ6#TXO&!#Q#v%UYo{>X8zqJLoZR2ixkzVpO|wPGmB=@El2 zIBu$H87mV{=RAGrYQLQ?Fhc6O)CC4g@@>aONIej70blxf3jDmBQ=C?sd3l#0?=kBR zFul`|R|4*2CU1eMO>`$;Mfuz+-dQcG+d+dVJJB_8niVg2JjS=U3Vwlz-o*z-;9#TW z#8nZh$ytLd%)I4>oh@Nq_)Pn|3>#Y&p(6%>38?@&B9EV!bW{-?;S8MCZYQh}gn%hU z=Ct6W&;_$#(PdSsm3*;u@G^szw^R^x*i%8Y;Ec-*dTdpM=3aZ5L9ecgQ03<@Fz94S zc)z%~qIl~Jq6J^Rz@Tska-;I*%M98k^zGt;?d+LEVZ3~kZK>dh=b@FO%2S68dPs2N z;<|XgS6BnSm5rAgR4=w_78kHXW)dyvzto_3M}!s(U24!bJ0eth`lSXf6PO>)S2?e{ z)S!m}AG_3`-C&OfQ408aE0-GF(CHM~?Uk09MeTC)sP-CvUX4}p8!uy3Fzd%U-O@W& z#gjoc>6n;w9+ZShkM0zcp3*5M-62Zu6gObf(N37u*W@Lg1Rkhcn6r|pe8I}8FP?ac zja}<1h&S$bYU5J_;OXI z3;o!;2CbDe7sLhQe<0^7pL^G!dJ(uvTu^aisPgu!4O%9Fej*+_fVJ_M$U>i!ouSJCD!)UBeOI=yI3;nl~%9cI>L=rS4) zFq~8jCz;`on4$Mxz_4Ymo5FXEWv-J3!DB_&xjG{E_8!RPV()?I4BzN-%Z2j$Gc&*+ z2e_hzoMR_jb4~_QJhr3P``Mp2Csb1;I&Tix@89jRSAL9c?L^w5^XEMAP!%kH-m{LXK`@#&yR>D2#i$yQH?xswA3w@P!7Q zbv%X^d{?_wRphEl`N+~u$46)`uG{{MF|XZbP_1D81DJ!qW6+TlbSa z5OLi2NIVn1NO z>|PXAF7%4C%aLyIa~|DKa~|Dqh(B}?g!3;y)NgC47j8>LX|&wz@TG9$J@n) zrZ{>WzGl>*$<_+O%nU1>bMF{6=#Jh9&HeGHLC@mH_^3gDgNh8UZYu#F+$Z2=_yO=q zeK47?Vaht`v{zT)iCVe!R<7N~5hY3Ea&bdjKS6}m?GKCJmBm9f5@TYkYbC_ojVE)^ zi1B>uoIbY-><*o~2VEYY!r9Fq3!wSg1pR}<@9m4w{NbdmLcaxHE=hZs$APOcLCfJS zoy{}w&Q9Ix2yyxCZmS^3#6ytk37S#S>2h&F<@pevi9Bm{M7(nPh9#i z3xrlHE-+a3busvC*m9w*5*L^Q@Zr4Aj~LV}fi4k=QV-*NSWY&joxJ5`;nl$<0LGi$9OclevMnol%M8Xxg0)^O*@!=J=Cjy)SzxrrWu3O zFF$I~Lu(>5_xMK*dKr*YA2n#RKtdQfo%N_eGuB2ZjE@_S3k54BL=;BuPfA$3&P@)l zZda*3PKSS3F03(A1uw~)Z}KPo1xfk~ob=DH^^*PqCoQ|yO}g^yr2mHL-+}g=bkS&y zoAk-GlJqZ~5Ft+bm_WRwp9%_+J|Pr0>0nrWtV&|P%kaA=2*cecM5q$Bd`B{|LQ;hC z>M{S=i4m#<^SxzIBIv_x7PSh^mYGPWY|As*_%^@kc8lqDv*`v-#1w~UH-FCadiYcs z+RcXAbs`FMqL$ue`hF*&LGrv`g*WbG2Ao>jGXrDDVK0O(<3ZUE5s)hk(&a{Gu!{c7 z-Xz^^;7OmVlflmQ^_?7{!xC1yVA5UcQ;__|?G7>1E^%S5BCLzXYwTocymkuJMWLR1 zjn{8aM&rfb{XA$)+2WJpg4@|#E#G?Apnl1{8^C}6JqFD=B|-~6y2qflQ_#Nyvsp0L zxR`Y=<`xe#E|~ASm?vD!dmd&|Ft@sxxu=S(6HbM!3&#zrK2DXUdw{ul+@P~hjnLuu z;-1Ld-;En|jm+@Ij(R;63utNAkNc|SDhlalo(i;LvOQHRHWXwVyA_Vf?YAggD> zcHE+jvrAl{huxNRrv$o0SXi~B+zP)V@_^gljfF54ls#VY z=y^iY$(nX5?s4n?o-%i6E4~~4^2|_;Mf5U%==9v_?sm(a1T6EE@S5B>> zN8%CM|GL{%)a~NVAGGOfvoObf1z2YteQg$cg#8j&2*-DjZkHt7BQ8Y#54TjP-Oa*h zqqv}k@n$q?Hxa*V>hS95a%Kl=OT{N!*OK(FQQ!d^6x8SC_CPwqTRyL&fsWH-kWI*Sc6W^Y8gotylA6d zbUc=RH&!A?QiUcS%iYcW@B@hmRo?rwK_?5eUtC1sCG=;XHt0pj3g&ihoyUFb4p&J; z`i2pzDcYDf4ibICcs5-{A0#4FgAfE@sy^3Jt1o%SM$nz@Rn&qYegUq>a>GZDt)d|> zT+J!J!E?&XEH# z5Iy494F)Y2(v7HV58PnT(StG&%?f381quYupT{KVHh?jIK4nnm&!b?dEj4Vev$CFn zvOn+}>g$Ju@C@FKhad^J4x%1p>83;Q{omgWM%33s?ibVU5Ell<{S>O~Od7OPLg(D6 z_`1}?FHNF=Z^6H^av=)i^fV`YSrWD9XTt_9m&o1Xf(g#~9*#y{P2ie^t zAH!TpDqnckppyl(Rb24-T!iTG{dNJw4dXVU-6JkA^Pvc4RJTi@ON12FFtW|(bf~;L zEFvo`Cnd$^S8-a<@a`5C6=G=cU+D6Gr-WF-f4T6-kr~d@dtLr-fxpgE3s1VDEj96w ztwr31&$1^I3Y5f2M~bG=CsVJK48wC_?+M{GnnqBJTWrb`kxO=`RAE&(_hi>^&%j zenM+DlHx7`$#Qksb+=-sbh>07>vZWn=(IG2HPUn)cPSBT#hssa@`Y}?c^(@1_gO`LVtT8XXG~n=?MFUB>WS&ko`!mS z(G>% z{+TOLpt(?Xcb2?bw+V_+<(t1YXt~&KrMOVI6;ftCZ_xLFtGsQCLES<=SzN$YVtrTo z5h*yaY|e>yr))1v=ts(KDr=rvDlhUswG=C3RJmY}LBAH3lj4H))1M%k`{g|b{qwX4 z&28Ib(8t$tG2o8QFRahbrfg44bwJ#+OGDwFJqC@6^>%@4YL7w3Xl6&61t*izE8A`Y zTu$8qj?cRZc;1(Qc>-B3E)wtpL?HpOb&`PdKU}6thbt{Btd!Y)FX`AW>39*$k&bU` zo}N+SrX7VzTo!Zg@#nzXOZnppZ?pP4*SX`Sx0j+EQ)HsE0=Gv0xfFHfO>pP2%O+80 zi?~SP-!bN6svA01N)69{&Ycx*!pEGDU zJvEaP4kyP=#_~XNHZAjgP(Ekc36DPR;joH&UPubwd3;=)ALYJ?ALagL8t);{#i5_I zRMB(k2!)@>8PwgS7F556UE{ZN232MvH1{7lgT9&(74v9|WR8O+<8t zh(6=r8?kfe9}JnfWZpxM{@5RD)YW)2B=Mv%U&v!87d#e-v@sN_iSxK=BQK^7Kv^S& zt^m&|yL(Z2tI(HCm<7v=UVEQk_7$N}O;G{piqPMyXVZllbo_K>Xj}DcdMFd2%EBIl zCPjFK#71hr4vRgt$Do&i+VP@6yGu|X!^8ZTxYA1o&9Gt4@;Y`}!|isw)3KYaL9>`H zcv_B&yb#;>TFeGYp>%C%>!)Ya=WW*e8=Y@UA-y(WTQ-q02V{v4?r`m)>Rw~}_K1F%|jEXoH0*QB|+8m5pgJYdGx zc~M6naPtK3rC(bSazEj5Z3PDLk0WUX`bGuDjuj62D2Y%l?tGi>s6ab=w1ZA?9oNhrMLbBluDKl0kn&MDpII$2o~~%3hcE zG_8F|@e8rW?^Z~`f45?Cy|>E$ZpG|VW}^Z+i6Zu6_EemEMHS7=;$(vrFQgnhPkT8+ zH^OpOf!^aN%R|U<_BeEx45GOo9W|%}FrChvm3PwXtxhvzVGrL;G+0K2tzl^zv8I-D`~$ppdIb^~3&5 zzFRAFH!HNF;HLO)R>&&2DOPlUU6&1fe&AN5xmI?iWCPM3D7EqfEOV15^8uFmeNU#M z??a|*a#z-PK<2nk;W$hCr6+BirTw2LP0^LfHwlx=#f8ZqMcsnQ4=ReumlgviHxr2- zb*KRrclGFg#BcE@l?{lVWNl6?y7qmNwW0Z`1pShKe!Uo>`GZc;MgVHQu|7hTZ~e=l z2~lddxKM8A4B0MwMFL$SH0%$4RqeGB2qr21sTNg+-!Uj6dRL1JDbGU+CP8x~&?Q33 zX;|sNBBUse^2{_QkwaUrAfnb#jhRT~h+2PqK^4tkk6Dh_ zA!6Ip>%x7kTo#i7R>!%ljsvT9W)&EZMy(89PB8P97jyZzn|Dy=gpeq60yw1C<3Oxw zC%Az)0XcLfg0O@zm@_ljtv&(m_uXYhNxD0V1#85UGZv6{sw?kQ$a@~l)2Ts+=u`-M z2Y$b>Qz7gV!!S{4x5IGa;*j^1VJxVyyg1~U@cZS(A#daVgFH`6 ze?Z)jD{ctlz6173+z`Y)4ZmO95X8OB;-)PDZvMG~wg&`eT!9$~oHG(3J~nfffHsMX ztj2BS1>;v6w0HzXJ7`ebDWnEBtV-nzjfsROV^X54f_pnUe4^VcR-Q7OR*$%~{C4i| z)t10L65$nbA@_@Te_kybPAhZ`b{1V}+aL4i;N42<-7Iy(NQCe>n7x10eZzBp{qTcXQuaiQp`Fd}UI-wZqc z3WMf}v<2bO%$_Hqwc>($zH5pF66g}ODY&4#DCF0~MckX+xDyiS5=-JP5CheV zi@5j#4HvQ+33Q1%?zDxh2@m>tW~V1rT2%r2{>)HK{tTk;^P{PUo;{^`w@uqAq;Tp` z;w=|f#sASv8OQ%ZBwL(AN1x^aJs|*@H zJ3@17t}eC{HHwhQ@waRECNn^a0rb&pEL?-3LHV;2ekwZBlB_KS;% z9{&eD{#%8-O;NS@w(}IM&U0lGFph{dB+PO<{c}861 zsdoYCF$l#cf!+l>J-^pA$*n=vzfx>eO?n0!u?3=M_%cY#CRJtNM+GPBwhUvtGZL8GHZQ??)K5R##ur0a-c7*=pLf#=RU~6$Ki41R~KCq!+nd#)h4s)H= z>EP32p1$M1_UnC_ft50w=rUdu8oI<)>@xny9lrUVr7I1zOLt!qQAJ{TR29N2p&N?C z>z7DQExuHW#K~fugt*88FQ*>36iGp8zfQ z`GOf=7xxl2A?VQ7&lA!1Dt`xx&-=kP1t+h*8R&v1pLeZ-HQAFwHTYI4o%F!M*<@T6 zk!v(3rRu6^KD;*3Noe@KahYqcld#N4^Ru)b&CaJTk5FVR%40PL0cTu}9?DZY>B@KI zj=kWQzjH>YCQGz)1_a)CxhrGm3?(E@PjNB&_2oPO{sjkr=4;qDq+iUyA&+COh)_8E zhC#cc6}-yafxN1I!=Q~kL2qr2(A+P*VNm^5tfN?K#kXNl-n6`>lSviOO0&Z?{NQgk9m&sTiu z>?-Po*B-FvnT&yE6?s&m-*; zQVP=qkbdrws#mL6>s->o*Fl_@7p+A*-s^dszUh@3Z}l`)S+5wm0&1 z0tz4d6N9=%-`8gl%{}iY1~q&WEzsV-8Wj11Qf!{MFvc55){;qsPWWbo<{md`&@X@w z|L>UwZ5LK=WB4Ba`%?x@NXTChYyYPWnt4OO7`PNSShVAw4JIa($Pkyu*Jp)lV2juJ z`w0)+;M(H#S;%?f?^9;!jaewFFW!KWO}uMxkS1qgO85Z+M(7_LckWi0XK;Z2J?rID zW>X8isl*7q%g`Zsp$FE_&f5>MzqK_&l}+y%)GWD^6&FT04MXw%_Y8V;YlP;e-ZSWx ztr4Q|@OuW$S)=kl3(Pm)Gw9G8aZU6+gT8!Ygy!Dyo9eU4S3Yyl2pDH^S!V zz@8B?yTpaK4Z)!|6vesZMV_@L|Hp5&w=4Og)Z3LX(o;8L<;Bb;tpY6hb|qSeqrVlQ z8Z`3p0@2(2$;*x3iqH{UenM~ac`G^sGT?jlE2`Y3E!&~LvGiL8tct; z(W>)w8U3SQrw78U(*t4XH1j6bX&`M8JrIVqn^lObIR^2tNPBHP$G5x%onVzr+oGmeFRoWdw*SZmN#Bg&JiW@Ed*R8(75!}v`j1;86#gxS<5JvSLwY9PGw8%y zBDCP0_Y6Ai)(9;y-Z$uLw?=60{_h*K9X~$$zCpjjkI%htPV%hM zye0DS5Pt;x<4EOm--%F-xrmf+H~Z?u_p#eSTP(h#Z;#LscygHI(Gj^ia(`O6ocsGQ z6XyaBHBX(5afayO>a*5Ur2kbn51`OIrO-SmbP5YyM9MeHzB=*U%D%cgceT&7{!@hd)71_5$eAltB zL3|H>S8T}=lR&o8R?e0UYdxL&|LiwruhP92y8nTtE+XX%eNS{}+C26x z7Frwo>cw{&`<96BYwT+f-}di`?#xB$EM6Y|pnOg#pM&zh zVzG-z`TohiI`Mt{`$)=S@g4JhQHkRzmEZ#WKn%};ga3Rnp+Qh@&7%Q^K^`6h(I@kMc_AYk4JK0JK|OkhIvYPW1FN#L@8kvjn`eC0s-h-%hKDShzQfQBcSfl4l5GZ!N@^Yw z7hLCI0(a{+gSOrop~Jtw4MVRVZ8PYRJJDQYYiB~xuZfFTc=5&6*|vszQyDAgsoVdi zU+u?dglbqBdYrZS;7;^SRxafv+@OXZpuE7B8LXlVd~CYMXPnkjMcd$ygL-_TieCMJ z+nGH+qiJ9^VISiO{`ILJN-{T#*|vxah1=Bho{gL@I9a8!tLgTD#xLPn5u%rvYx0M# zyqEZPnVagImA49tiyOQ8Eq}USRa{l=LSlti*$Q_ez_o%3+T}@lRY|J+={AFQ ziE6Kii^QCUJp0`?gZ}=*2rc;IHiJI8JwglK-e%BI+c9>Sg?w|>S?;NI1b>OQnv1_K z@x!@P!SYa>Jv83(ToU3kz67n<$=i`}L`yJII2Rt9XbGz0E!$livDY<`?F@VDCOx*_ zBBDb>N93(Ig$`v^jc)Cs5`_3gHQ6(rPp-uWq`)Sdee?uo~Z0U z-0Z;M;^`To8qRne@gjPfP5k(FR3gi!U3^(~FT9y7eVeZeoPHN3i3PJz%&&ooa&%9{ zk;`Y(uMl3eS!SXwPk)|q;nLaEcXx#5XL_C1bTL0fZ!(87?p71YOqPBe!qI)6qHAAd z&RQIEPSaxs&N$z5H@?@@En&alumgXDu+0d|Wa%wtdm6&_&rYi3Z5J2xzcPIc^vF2k z6q9ss#m^dM(_=rvYuG1LkbB+Xrxiaxwu)W@lp>=bWV19;k-wvgX5Ryu69~&>>E4;p zeBnJ2iYz}73r|`4@l3#u2aGB=UTx4JZ!2Bu#f8E7g6-C;G5-qVy@AmaD>%FxHL>!M zw+-4T=*z`LtZ$-d|M%Ml6~LXDZx!KjO>zDzZoG6`)`WrJK{oN)&Wdc+>ePa-XQ4~_4r;RYQ%l_OO5y)FkEd%Mc`I(!3xWPQf(RY zjL=>a7Z@BLpvs4~81$BeR_|6yfqD)Ojr`XZ92$8U|Ee-DPY}z*1rnbdR#H|;Xur6i zUIN3S0eKG!b9RhMqV9FGqxu0f6t`b(&|4Cdq)P#ZpL0IK8sif2A@GDXnjR2q41$C1 zrf^u@%H_SDsrH{LVWF+Z5z&sy5TBI7FRbTY@&NkVIByv2sKk-wJ05WR*&U4F%aKo3 zVgUG?2hdq%>2Y49diQ~dx&*{NI?Se>DuULf|6~JJ6b5kg?hK3=hJS*4#8#d*@w)rZ;2q?nkehfb_TN9j*_u-- z2&aS3#7boCgAt0AZP`qo}+wAQ1mZlAV|3c$D7c~OqL;@1|nLS-6cfuSFOy{?C?&(N;c3Q zB`w6+adW67KW=8b{`x^TLv9Y;R98j+deF_0n|a%uRF-^)r+Et>!Z;;Cx3d*Dzz0)5 z!o(LI!XANV>F+Y;-w$CpFpf|l?_!xpJ{+OQoLB{^Jh_*VgAcv<3&_gWzE`)JD zt^Bu2%gN#bi}zJIpH@krpIDMl-6as%F_2GToOd0KRWN5f$;yYTwo8DQADe{F7IDGH z&5sIqGP>XkxP;nc+UpKq7*$(()y@0yY8$mEH?%cWlQZpgMAw9Vo0v_%e%LkpHT;|^ zh5OGpXd+g@vCtzeIQx8q{sn{h1r3|2g1X4+iE>43_lX)N`e)@kiP?1MPhC0xRIAP5 zFTG??ZNE>>i7y%S<)6a9z64phoRd>&IecLYJRbk$tvX2-QTL(NZ0h&KbYoM5DsMT< zpch4L@=`*%%p>U=)>Q+|NejlH~-3%?FJ%4a7I8^Me_Ei4GvD1g%gSU6bsjD|Eq=gEWMBplvbhAx3*_IdcxwOI^!tGRZp=C_%Jy&TstOOg{#Y|3%&EyMa zu2AbC=-6NZ)XFCvtV}NQ5ObzQ+SYJ3=fpikZ3&X|;xF^tzE?$`d(>Z^O>8x|=>F1J zaiSGi8)PbG z35h9t&_TeA!)Mcv5D+hnz>iAwn}8O!HOyybJf7Q2cxy$x;t=;h9KHgD34=XigYnZe3`@_;d=!ZlCm4>7uz>tf|No!&oc?CiSMbR#~y=WaOv*oh1Tf{ zU9PRTq2uuyTjgc3GESB#^v031>Ca$zx_Cal1KSZ4;wcE$DmP?P1F6D#D_(GN9>dXcjDK4*n;vr+!}IFC7{_q1!DBX2j_C^{RTO?) zw1ubU@qO5#8)w2hQ)wOsc}yqEF~4lqZ0d5E!;{+=bKeNdW4XMujFmCGy2fP<&;9#i z9aD$dU2YTQxW0Z$75(1j3eQKL*oQSTaWg(-dE9Ed%IJE~_p7M=39&Ie>3wqLw`5&r z+siS%q@#+ibeY2Q#6Fouv!;!UO&-IEa*Xfos-pK?#_&`>xeps8<112yyvL2Y%h=$! z?kZ|}QfvUv;J&!#vsOH1ragAc%d!2_@l|w*%NCwHrem8gXcnrTvvUzCn#G^;4iJ&^sFlop1(~e(3LY3A2Mx^*_`9c7zaC82k#W)z_V=UK1`nN z#fC+c-yL^ZQcoGy`wh;flU&yDT(vLOyn_+Nn5oCgJH=dF#JKFo%87Y)>1=w+6$1}F zwT~3oj=Tx(G2C8`@!anWrEn|T%49=!&T$b?Mzb}?aUvYL>j+bNo;GwhW z_bzLAKJxQ@Sb^6!0;>%k=jDB6l*d{_`_IMB@TB)8!Q1rv*?sjM&&}m{f4pHfUFq_M z=k9&+#z?d=pSN;&eL?lOF6Y|SWwh_Ru!^2{CBXB}z9e9Xkh0SryE$vhu*K=vPdp<= zfv4x0eV9F)YI5vC&Pk`OoX2yl9PbY%s^}7zH#~Pt$NQwZQ`|;XjqqH4wI`I(`ROBP z)3YvLcxL?~pz~~(SG<<6mT>bkzx%IifBpNTb@(C+GPzZ&3xnPwM;%=vnwIjG5vz_^p>8E zDAO+~Pw#2&jW%^N;)+s4V^7CQGvD6azPgbC=am9F+Pk9-^=ns}`HrZ91QXrV-PfIS z2BVm`aoWdq8nZcPa7A%&jgw0>IYqm`^iiLM1<-3I{dBDYFN3XelFH^>9LiADQs*l4 zP6{)}%R-kypS{j@hE?X2VedHU-dxe**fXX?2D#Nt=M~+ULT{Ip=@RIPO$lk{lt?Qr zRdQ!dq2LxDlDgj%ib^OGbqRB2H92;^;6sF`iQcu|$_=KSVP;b`1=QWdgxOOF+>!8G zJy0$(g}#zkT)C(NFk=cJy1|Zn1ly!X; zi8jW%I-|XfO~-L0pB1BxvA$?aPj63SM>o?(gV7r5ds^C}EsCSB;f4|``&1M^JE*3> zem?DO$M?0XU|lD{qJ7mWmg+NIw7spjr%`eC87|u1*0@4arYjprK5WJ?n zvuP!Ve=-p6)qqb0g8QO9ix~as(%?D{{)`T8?QV&+clI{7t4#J4AE9)uRtouQA1JM> zyg0s+FBCV9uk;J0uT$CIR{@05)1m;CKx)4Vl&=y9rM*+>=Bo!niT17F^!Td6ioR&9 zvAL&lm1^I7x(TJFN44EP-9YJDajZ-6>4w6O#?K}B*fFWGsmUez__d>b??HR5>b0*S z7t-qXp58v8`gjV}D+E5CEvsXV&B}JZ2ra8)%`K~1rZ(6ut79#lUc>6MDPpwiCfAp% zEvsWqt6CbBb$s1I%j#H9OH)gGw`ySgZ7rnq^mM5xqrQ<2Q=}x1`Ww@h)vfKTdSM!s zmUDGU6q4@D&Y40(Bh4AUUXwbzdRt<>J?(96Es8ekAF#!`dsm|0j5YPNH1@W{np=7s zn^sC!wVz7oB!7sNf*4xw9sXB4rKs2_xBi7T>*51__ThSP8Z;GKMh&6Tfb@s*@ z>Y0nLaW4y=0#hrl?RLT%p=VmFe8pB#qPWD|h&zs;>QwouaWl(QZCW371(US(e5A!5 zUu}~B*=4L70Y6!B*3E#Q*2GeE!=`9Wj=kQ>6_}}`0~SE7K~Yu@SU7!l zD$XqQfJNzq86^|+n5AGQdHNkSGugDtUOjRdC=Plv3*QhNJ55d4MOaacud^hTzUI@T zH|Fz>(!`fJsf@KUl`nLpY!%%r!afVS7;~fqQ|q;;E+$$WD8cwjj_~O$W;~e6ld1Zk zGWn9TV2Yq#r(mZ2kXv-=9c;IiVMo>9J0Iv0k3%c?5Sq567LLPNR)SXuG&3_^NUgVG z`D{u}tTo(6iN$3KtjA7Sd`8yv?Y)1^i?}4qARNlY0T$S`2trRM$Gau$3+Cc zov4wz7JEpDV=lTbyR=>_U+7_)E{lHJ8m8&`?9p&3n4ZA6R3uB$^eo1u@oo)s)l(Ul)|gHwnWU#PK9b5hJ*ja?>Sa9a zka1i+y>S82V!}~;^&H2gasH{or{_5VA~V&K9hcT=74YmQsy8F_w8uqoe+KjEe%K{( zrvj3$!Y)Za+^V$FGa(nz=Puf!pHca*2g5Fi*OsuM0x6%JQfZj1mbR3bxlpa#cpm_p+(@i}poflH^5 zd(61inzBIYIvfF2garb&qxUM?xI)|A@p;hiR>n}0ga<%2sk zt)XvR8N)aZ8zNn~1n!?8ZF)R`(v59XZ#=1M0B<6tWhkmfXg-s0MC^b0w?llxU5}0p z7d3ngYi~3gIAoe0jEK+4s)a}&c{cQJ-ScIr0i8z(oDNw z&qg$#kuxb9-&=2WavfH|Oqd08Wie@?uzD8vjhnMp4kJR#E^tZml@?v626!~zWa|lr ze_U!NT1}i_Yqr!Ndxlrs(;ls}nBhd|Y6ZQ-Hqc*ZAI>>%=&|ypHsf%-W(#BJ+GDLZ z<3+PzCDaOz7VZ($nvB-iBeZ0)g$>M37yV4i?sc*&&HNA(^Z*)zY7^)5g5sTYqP<{g zik6S9V%!o=*=j#b%gCD0E*#kD#G`d`;w2kZK5~k8YLT@jYM}x$4_}tX)6T$v zl}lJAFG9e)cRqaIKVEwNUoV{M}Z={EmL zQGP={K+!o9&lUp|ZJJ~}K+suZ7u96H*~;cwjy9UQG=Vjj6=s5t4(On1s7Vd=OxB8M za%AQVjI}l4Dj&jqeA}8bW={1pI#m){uyUD{ZQ_^;N7HFa#yia1I-Hxba-B{= z{h$*9ndiqeTZvR0=Ig-I9>8f~eq3t~^LNKK(;l$A(Dp>y@*p}RlkpX1d|fh!J3UN3 zr$i;5gL@Re5=>zedt|@p-drk~v~p;G>G~3;Lc85Va7%pz@;JRSF+a8_>@ULYa>)w0^SJIhK!&84C}Ue)Z)l3uj$1Xcg%uG2U5ZSUz47UH-k%` zT}Y+nE)dOHS>@AT5UqG@Vg*{EL`p$)Dg?s$g~&R;grchz({&P(@unOYM$@=$fu>=X zULr}xRS`kC)__fid3l(OcdfT_3WO$r8TiK_3H2cGK)@ahlgZ8xZBFbEm;-R#kQd8b`M%Qcd+B@ zIx%$cA>1L4o}OaTQU>s4&ro?;81#$RB{XLhaG7zWkTd-Ny@r;v62-VR1)#?;8FPe( zf!$WFduT%*zd0{_iGT_`5L}t+ipPP-(^2(RG^x#?A!Z)+i4F@CM33BIjofdmrR>hR{2ULmbTbXsD*S}KoO6|b)SivsYmQCV@g+Jb?1UJRP{I*;rrWD1whMZmwi*ZW=*VZb`e}fnL@r?_vR+r%Ts9uoe`SXsiG%Xwuf^sC7C=_qCr+T7`QxOz+EiG;h@Kfo}nL-RbQK zlt-9(_ut>0)J>X`@3GS6NJq-vYc`>Ks9EcVk;i zwEaYN5>$_gVx7_UHs+!S8?k7+K4j$+8EtQi>0?Ve8=278sY%-OOp3Pjb~HxSd0U@S z(P+$n2uZm8EUChSLy_rPhx9xtZ7x3cjj5ESs2PW!Q=2agbM3&mtht!Ua$KFMh+ERyGQyX+h2c^}Q`$1^ zqGF|ulP?T6rqdoauSR=FhA!{rEwfz&sq)2bZh-GT&&uFcSf&o|GT*lL~2Tzpp%Hb*}$tYa}2c2Bp zlC=eCf=2KBqs~kmo`CjNjt@CtS$kc(TC&kh0OB&kvrrG9+ZzB@6bDzJkb}h~fubtZ zPDd?bx&U2XF~?42umIkhimyXY7`5V#trk?quf0nbUcLc|%OFiTcQIeH@rd~TA8KNm zczU--6Ju>DcjT{YW~?n&v~90*)KU}yi3o~_waMGz9Ih)=tZmp#mDw+dwPl=KYM^K* zVnw@{w^V9%?TNKvsh^qY>Jn?S?9*{V&L2nDmsnfe$!_3RFF2yECb71xdTd7I=t>e( zd-6z(FGp2qzy#;%{MJ<04(FE={l>VUl<2p}#!_NX(Mx>f?sECm4JBPQeA(3P*T9$A z-N`uKhvfwOa;e+1wyqp}_Xy>+TRme88)=gk(x}67UQTBbg59+PJ%U9;-=P{naM9(A z2|boN9jDdjU>;im?o&USN~)tgOwbWPI4+e|r7wu@MYqXtoixDJaRTtyi9&d@TDW71 zZUjJSbNGAsh~mqZXex<=E?CE7lCJ{=tlD2}lCaoV05hdwoB z^Nu(3(luIR+RWSl=4`l38|NJsD5%TOhaj<1%IN*PZcd{)N4*5=T-AuEmA2vqW}usZ zQgWx$WMa7oO6eX`qg>;Al62qc2k{8R!{~l9fRV*f52m|K4MV<2F0oF!#|&bSJq*yj zqy~8FK%OXFwtXn{mu*eFnvUe_1AiLh`*t-T`rMxn!tc9-s?YuTs2l8flVhv76F_>R z5d`t~pI6~j4^Vm~B8YOo6YT-@YD`(6htsD7HIBzhYDB=@xuqghe`wckLK-Xql4hg5-o5lB-V>YF;vu5vBBLLXF^Y5KpARLmw*C2t90e zw>N>*XACtG<8svH-XUv^lS?a0=%a=j*oys`?wk|nFIPEd*tp(+7~ADV1^#Rq zlQ0-eeK>spQ8UF5k+G}=%hsn6HLS@U)iXjLM%0M5Cai<@I_j)2a+jf>_jJn*^SV%n zIid=PTH?Z_PB(;kZeX}3KGW{s%2I?LE54oyx;0h=x6?r+-TdaQOs_fM^{0Du6Q8#- zCKpgy5~4Rsq-9L5J$|y5$+Oz#)>S;6)15H~x3_h2uuoPhDFu%^G2L$DQ%P?Cr z`BYN1R_K^O3UGpHT!L0ej`uW{xyp|dL(?B{%j0kf4Ah)j&%$)8?~&ZC7T=0n8s{ke z(fmC#j*+F#P$$%6&1;8Fo;o2{=1d)Ln4ZSN_pGT74frHxOuI0UD&!mL>A+*FNS}bT zMzX$e-IWZaOn_m)^G3WfXq$FJy?kJT_ei77taIk}xqKAVRARZLgC>qgzABZnHmY4`_u<}&D2PF{<%~>zP_P85KW&N7OFq0uC@g;nY0peim{9~ z1?*d5@ySZi*dl)q!ITasks8EWJZQRoWZH>NW*}uTMRzzEb7T#^frB)&`S7Pr$Ppn< zT_o1=c(s5xWRTTZRm|%(JYRe~%jLn^?LkMQbi2+dtV^_GDUt8hsm`&^tGFJdt5~qg zLyG+2x9(iZ#u~Vj%Tx8Segi3b6ki$i1GJSCpcfDBBBoBPA{<*pF~wmIP<30zs_|xC zw}nHU1F|`(%NrwF*7LgS@?-tGE<)jjdu18WODxl}o?E6o)3mx0muXDW9luA)rR=1N zr&ALkazs7J5?WJtpSrQ9;?4?YLyM+!blq}D^wykpdeO4u8<<>Tsh0H^mE|!jNvimI zDkbp&;H5{7KgEuU>|~kht2!uaihC5_UuLqNHwDP}Tn*Fpgh}XajykQ#R6SKHp}MQ7 z$Splx@<3X7J=*bz%~rmUbChlL1gVtlm!RiJ9xCdlUpG`9jR$tDnw}YXM0Y&_n(HZ% zM;5UwtYpe&jGhgZVp{A3BXnm3aDp68mMzZPkw|N?OmhQ(6b&*brsN49ig=~ z$wUZ=)YBY~s@dupjz@I4>M4y!;wib0wVuobA)3GL#DWM{h@QE4bgap!)ft~8J##68 z)8h2Zr3}oKsV6T2Li^5bG}}v0SbSt01Y@ecKqb^RXO%f%DdJ(J?F4p{dd>Vgt)ad= zCGlN|tDU|*crp|4K~vCWANr|^zCsQPqT3DsY7 z^#v*qq#6!?iPyKNJdi5!y3zIxKUDkGGJ+4DWh@)_fq7dDI_W!A4C{0>j7~1orFcv@ zW6?)5Y--PfxdsIIW-u(FWj#n7o9VtM#(Z>UjOgc2;fzuxcku!6y@Dq3H9$+*_#tdO zSr<+L`4t_E=i~<^T``l}EZfSZ;xsxn zyO`eU3&Fs)<+z%*Hy1nY95z2!0hN=5=I{svnqRx%V|BD!kKyE%5r zSil4D(gfI-j0PB>7L8P^#-Nfhfq=MuWqQewfaz+|;bk@Bs)8*J<7BbHQu>yR4;aC7 z8(!st!X-llM)0BHHuME~#*_>b7^7Ax!J%ZBz|=C!L4cMF7TUNYcT2{%rIO<9> z03}630d$%HDJdb`*kDP5NBO>ro_3O{xS3XT%PypBE6oL1PszEUw>h|9^7C@q;Q|M5 z^MN`=HBhly17e9%ZSGKmi6w=$&B6Ma8?&UqV)zGr-kz!tNUltfOB4c`SDZjA(FZgc z{4lvhBOkeO0FPbTR)TR_IX=p4;(1_w1~FW<5R+Y(!~?qZW~t0z~qEqzYuJzoO? zO4LfLEiPZ?Wz3PDqOJ1ugAX3lbFO?bmvfS4!SdEQw(G=bqN!do{NNEyET|O@&8na< zKB)z@(lPF{tXklh@P(*#jY(gKTF!V!hh)tBI>lTcaz>b9uLo8LQv!6&L6{Pu+ZtGD zk(VZ&0)*Fj;hG1+7klA8bElp7KxsW0N=euwp(^*4)MDOhwDP#nMq#7svR?kq4nDN5 zA*PjIXKmomZBW=4-=Z*Q-%{nnfWbOz1BH#{Usv?nU(EK=i@}TdVPovz9u}4O$pJoC zJn5`7^L)UdgTlt_viii6PHaQUN<*8S#~;sZ##i~ZvE%X~g|IO$EU@I_OHwY=P>;7y zd%HT?o1zppB7!SPI&gh?i&tc=#45*3AH;R6(_^r%tK4x^B zWo7$pJY5nt4k*h^J#55PsWnr8(DA!fQXjn7&s5<$t)w5iT%h-D^XGwF#c1p5>FVmG zuraz?F~lc(v}p#4sdQpBZjL%Og^kT+LGrXx9e-ws!bbJKJ2@K;deEPnDSME@#!g8e zJ`SzL;Jp{#f(jeE1aS57{)7SGvfm}SeDURF5jc2>BDGtk=U)L-F=ZDPHK>=D!p4{+ z2cH_&$^f&t&c$q>7SmAgV#dp2^4w&)m|f*DczDZ&Oq7K{oka~;jH0m7KU0;ZBNzMA z4#?sybFYS&{{2whK=Q$RXgz1m43Y$%nt+L zDJ(TvabxsL%W>*e{uuZsZz_)THq@gU^mle~QCk`)YB+P$X=W@S=44el`1eLXpRE7h zMcDYQ6grG^bwL0qaQPx^Oza!TyC7lXH9=xyL$f@?$~pXLldwSo2+BzlqH{}^A5zR)d}EC-eTIz#g-@{IQ}<@`*yN_L zQTw>6-Z+x2QzjUxeRDpH4~30QQh4a6er50pF1}adJ{c7@M)y}*9~-E~Si!~JU5y(n zi^H8-H8-WOu}pNj+4RdCOxY+d6gJ)w`qhD!2Q1L?iNg5crcU%=%03-gdATax-`r8c zIZk1taynwRkVB`BcIavq{~Li8+DxYhQ)qSPOvi!mCg&DW*f?48 zoMGb_Nn0?_lJOO|=V#e1JlsVWx_lZS7I4&_J#NFQjSU!e;@hfvbsy>PTyQH2MeZ{y z^S1xg-w~+$y||9WyVESTc^a`isex?3DQqkh+@Hq%9PvU-T+W+@jq4<#)R8b&-#JMcvE*D-$1l-Xci;`AN(&r$gwS* z$DfrP`u7GuBpGkgFX)Gju^UxEI5Hp%m*Dsrv9K{Jt>V>>`qS?MpvvgSl+EvZ(JzN@ zxgkeBqNojz`QvAXEHfJe`8`Fx;Rb(Xz6*wGWVwxMZP`Xu-A3X!Jz(PMt5jk-05Zw;8S8H7#TnE*{~n{IJnq7KKBm?RMVE6=ZC2h*YFt4F`;8 zw}Psk7Uey6B*+CHhE$Ip335~#^1dAbyW`@F#pAi#rI>xI-*1Wm4cJ~5FroqXmj!^1 z;W#q4EI{A0p|G)1s@ImE2ed!~2`i=2cmM;yT1k0uAk9ZXg1}{EfgRrcqaKs_-7mER zJ9fHxO{DUehIjCNUv+8&)90|UT!w(bZmRSW4~C+*spjddpxsk;*We&_GQmqU{7sKP z%{aK=LB>V9AVe@$Khb4(jEMcvuu+c3|&gsYa<>gb! zyj6&q=`=ccQRzR93qUX+NTo7+#8r74&t>C7c!?TMBO2GG1X-;Sv(?Yr=ZGj zrGUmhn>w>EWtTI|`NmsCM^?$jP*WFNbBr}4;+blM+@gzsuLFpL1Nfpoe zZ}^8Tevo>zIBX1wf+Yzvl%4S$BflKURW0%y+*T3#>MgWWAatk*OA1RyXA{4`PEZSixF3RD&iB=9uMN>%` zp!N%|V9iR#`Nl=qSpLa!>1S%nPPSo!B?@03X!p!?+ReSqqD3(G06-m?kkQ)7kNDFI zYTCjG`8tUmENr0TwXLu*A>uO!`Ku!i)Ssa`l>d|j{wpvyh@1H&W}L-*V|v)!kjIfq zlqMDnh0?Ec*1NZzWhJV%wFIdRX?%!foVh}G? zs&o)9|_CBgoLCl*cd6$8hLLvYz&n}@ip;`bYRso zB3^#5-{RJY6&FB%S%7T~XK?Et4-_f-tC>o3b8)a(FIc7vxO{~)iN-4f{h4L22eDez2rjYU!^T)HLiOIQ<0LbegdUZW2z()0x8nsXUx=O2P>)h4sR)kQ88E_ton-->Nyx-4sw90g zV5G!Aa=n>DDagtI?N`tFji`^3>sGPXn{Z>yp850wbI20NwYb`sTEmGJxt1MC415C6fS8oUix7Am^!F{wW12dh@ zN&~k`tU0Ae!#sRU*e5^cuid!ImhHeJG+|?05U&LgXoYevWTGr28DGIGCTg`tCMLfX zWuDj)FjoM|H(M!eY%H6yR6d;=$eB6p>U>lX!L}lmH*@i!)OytTkCjE_t#~nKVF8K4 z#vx@Pcse_8*@-MNVM4SHcI$xTtoV9{yjB*Hw=xCH98qjXRozZ3@LL)mLl?mKsIYi3*$-nZ~8^ILwybxF|m?r}mHBD+vr*T(L3A|hc28VR`0DRaOlYaM) z0VZ)gv>PipF)Y)rv{9}`Xbtd zn^jS$c)_1wMyNNa!^Y^X3NjFYWXuspArTl1OHU9i)-EV+rBJfa>6XRu5CBCFDc}PT zVPmBrgS|$^DhxU5WgDCeX~hTAy9Q1}c1*~+!21C`Q+A;-m+Y~4%>A114BCUIguFbE zF|*)A2detD=r?|&45o7Vg8G2g&-7p68Q*5egS*}_;QX$#tmfzhF2A>j=3jXo?pOpg?n z81jwD$i+FzU%Hc6GcD$Q9p8^sQ`2yW-&XOYgNX%?7BB@B8{}&JHi##k=!VRIqdvA0 zHkOq|aSC`rs>dpsi9Dc0f8u}qq8YCa+muG=A5#%(0}J~(YY>-3@t!ZnpZ}@s{bgnv z!&W?oxfC|mh80C6f++?5?n^9>2`a|+QxUdPNDXd?$rP1Yxf9-@sc3i8KRcC_6KU!Ir)6tOsDHGp%x~= zH4?T#-YTeJ-e$ol`?!J_Q?xzQb!Ab*xm3YgVdiovE_=1IwESH<&^PY$7UJ-nG4g7w zC?A~JB~t5C2|1$^HnJ&|=fTx6cV6NyevOtzSv=2FZ(@Xvv9i+KQzO)>O2?TMew867 zmr8i6f@6~Us#yn8h|VXX3c_DyNbE%)eetdeqCK;C3pm?fK?|wPsg7q7N5-8@HkG!j zlg^@rb&D6)f2sP2{E(Tm64f~?X%3_)SG2iFF@)E`jG0B*wH+txv795ULt8^cw10@` ztpm$w(Rz1<2i^StO^c+HbUYs20kNAyLT2tkd8F-SA@#LKHF%`jigH|*Ejp^sBaN4ZENWQ1tlpz_&n(Ag zaoy5}Wgcy!ET&<}lEn=kEjvr|@v20wP#a4+9&L0#jUk#H(m*qxLEZ5Xn*+$ReMpqJ zT)_iu4**9)J|x!&mp&fAUFyeWP&GV2b){d&@A!dbO1z6+9*FhSG_gF`WB~qb0M1pZ z*J;AYo=0f49(;M&uiIrae2L9CIg9I}2O1B6e^3@2vz>Thq~Kw91>(HukCV@)(&@%r zGVg(^tNhx$<_8yT4R>ML0DP|>E@J}D2NyPHe?RvdjSyA14J~ks8a%|L4;jELuJdTS z4$#~Jn1*_fw*A0qFlZuN;^sNifY3^~#O!}e142XL61zV#4T#$cms50||B10zX@GULM=f2nq|O6O ze0myS{i37l7uVK%0IHq_xNPyVWev51sJ;bhg(;ycuuCH6PbZOl(4>D)|G|;++W%Y}f)Gul9Al-*e z16{JXVcF8!qZZY9kj)FGfi9|Dx_D{*qGfgU9%k3!8tZjJ87qVH3ybSKa`hK9jOdyQ ze>{v_U0K8lB$2%yLPXbjvuI@S2OelP4-WWdQ4Jiwd)@IgN?XBCP$yViqL z2cYAAXwtDpQUwoD8$dqcN5+S&_`1euhlg1nz&`KC28Wr8JZN_y%8gaN#08_&dC>kq zl(FffEcT$;K$Kgjk5cbJM*~rAn?A}C54t%J<-5~IY4D(9fhc!OA7!Zr-5!YY!|9_e z^PuB_D0lm#c!T=YbslVTp})XQl!f!j0$iqb!PQ6lBmK=k&g4T8>O_ExstuqwMts?> z(S06lc>sRV6uABnvWx5v#Mm@_j6M(AABb}KG*R%`V;7kX#MnH2j6M%K8i=x`e3TU) zY;yp9-8Aq%4>uOTf3rMZecIWDZVyD+I$ac`+l7t?qHLQcN|T4%6~Nz99`Ant*#%Do zBHizgl(Y(iSr0H706rWE5#97TJ>Oq%*$F3iWYS5Sc5NB z6_mZ`GrSbB{r|5;0FFQN!?-U@<2Ev{zbxc`DfHNk&}S{87xwH=`-qE6K^)1A`m~B_ z?A>S71z%vn`0mx=3;F0t=(`o6&!&m)pab?LHxSvz8HdoL|KB3#@R%uuZmkG?HcfN~ z9Y8em|Hs_hfX7u;d*f@8X*(T~X<}=d5!3;>HRaZleiB46{iGENp)FFaketk%Ob(ej zGd*XfP0{LzC{lyvZb7TzFYE&OE=a*qEJ$FqvomZ0A1>6R~7U3FYA_Uu58W}wx|S>+1F zz88y~owExSiwHaYXS>xR$IkK9H*c=+=`D6eXGZ7;F#yhhJ!hk9*W)N0#XcX4ojpn% zH|K0V9Y>7f?9vFu;NPABSJBGlDE7%%>}*Zspdv1-{>q34S!ctsJA@d>j5$-LPsOl* zGeZ@&Ymoaju!JdRmQeGwTWu=U}Z+bQ>W$}T=Vl?|(PtFZe$AjFiKR-E_UU+v4 zJ?~XsHxw)eKGE!^r{)Gw-TO7Ju9Z*CrI7&oKNR%zbajQp9LogT_XCXdHjTN7h>kVI zmXz@aga0jex6l$FU!dqNd{s(owyoQ|8ENRO7v7_(KE^O;!ZL~-Yl0)67CHV}C+E<; z3|qOUg%;VKN8jh%{$nk)h_$>=KWnN{iXQku3qAK$UN;^Nq)*13!*min%K(NdqeV zt)S`Ul%(fA&_d_rC`tPsXd$8PfPz6;;mI+T0(vwyw1RA@7wIz(q9M#>EIOp5^O*V@ zV=2A6IkqHAbayi*mCrpOQhGOc;g25B5_xwsB$N^^CFv|xJfr@CDa`B-P?DDX8132f z{5I-URQ@nVbj}x_-$qv;YlgA4+bTFXA-2ZNZL!J$rQPq9UJ{BQZ02$GU^B**_hX^> zL8kc2KW?Gt?JiPGKIwUb7&Ea4h5YAX#KazCaim2f^ivkXr+(Z*i?WD7=GwWBs~hq zeX%56@Nf$yZi*!-@gAWnqyDBa&QbZDu_V3hVWw;*#mE{#3{CzSbX?#21i5tJ?eW-> z0@3Yp7@+_9uokl07YfwXraZwe@_R| z+>Z@56I~Z44Ea4o)OGh@Grbb|_aSG-0TQt?eQeGtQU~5|l6;Sld=Ha+?=M8=?qP!G zKGH&o3m10MepT|r>MsVyl_2@Dg`M>AM_Op{_Jy4@Q>4cLq8av8-MyGV18Pp*YEt$4 zIFE$i$1xIq^@vdQd#38_Uuw4gJ`VLuE#23#Hxj2sS$ZP=lE1dnoD(bpEO>`UMDX^Zbr8JsPAI|K0w@;XlO_xVxO_qh@V^0O zMp+7MY0VD_=A;>L>_a9a?wZAnxN8;|aq@&P;;vbcr}_Wgq6Yq5vzQT!@Zb;KKkHXd zw9>8r-GZ_Ez$|d!34Hk1(*wLnDF3R37LC%6XF)Ii93Os#9-8%>zqZm#e%(SXeFbLc z&l&i3q>Llwx(YqaMehGK_%-+2u_SdDgO8HZnJm*M1l}kP?f|>Bj-?l*Ku7Hlb`LPySLfjWvV~g4i*ab&^MS1Ak1a$ky>6Tq;bmxAjl}XFTWHZ9`TdSR zwoprVIgZ)l0%YF73A^I7NDSGFkoYJ{lt$yfeiPhch`_BJKy<_`-e zj;p`mM;EyM;ie?L0)TTr*_5P=7w4>(AnQIPoNA$SV$Df%PC@UV(VV1l1#=N#E^JQH z2T!$7*P7-eeH}lxG$-j8SN_8?YoY<1GXAe1} zL=MKq1y70!6J)%U{))&l&cJ~ z4+$5|21wg(PPvI!kk`&`rIrJG;oen1qD;Ri(V4|7^Yj5`voY7D;fQ^NEK^%tFZqwd5D z0XFxG%}F|@Y8gY;iRL7|i?c91+g>RI+=eU+&m)Oe>iStzl1?UCWt`5RtE)S7SgLzm z{Y7ZHHGs7$U=4himVyLsp~v=9JhOpy?AM%R2r$OIK_crjeTA!kG;Q{ zh<+E3EveuSX1$|xTX`+|JGK;lJ-3yfx0_5uB=U^gjv=uNZS zWl4H7KC;W{&75_Z%brBp>G;-aDkB!o%;*Vjg~%Er&0)w9*on=(}9?Pw_Q~FSKt#Ybyq6 zH^sioG|$|Ze%bPTIrHi3{tRlbeL+((i-^uUs8FJF@$n6O3_J8s_;{qXm0B|U;sh>%{{+<>_>yK&Y>7?ujV4s} zc3^)M#OE9AgP{a`T`dZKbj_*QuO(tcI6IBz;|{; z)%vLVi< z6k0QcP$6hQ{4}69uGxGU1J9Zsm}3cH);xE5xSUliGu)jsAcX7Gu?lNp8Uydq8Q87h zFK6J|Ow1m!_b}k1>8Y(0tx<-1#!PUI%WyN)o3$$ZS_WUO!dRc7!?IRot&i|meWn}97d6~QI^AslIE;LP;+y^6eQR%@Fcu4wJO+_rO8MIB#-tAcY78rnM=zex~FyIVb z&w!V|q=#WUr*9*}?ohB@)5C_XWXv;hIs^w(UIO789tGmmKlMkac*SQHrGzSIxNa~4FNL8zt+b_ z(o-`x4&^kB<6pjli~ZAMGx7w*w(;5lm$~c;qKce;3&YX@$rDXN@{9kTx^Y!#fm{Z9 zUu;I2+5I}VxkpJ17Njxy7|dw4W{rzM^S>Vtj`K|H`(IrQ%2kJkuD88XfRC1$2h_~d zmH@?Ts{u&c3KR6jG2%u1Bx-lu8xKI~RsSn$i*_+v@GdVQ4t-zAaVrP_q{%qE(j@1M z7}2X5F>mP);sI#pe@#xYwA&8RU0_o1%ox#Y8&U9rAI3?8_WtiEFk=%1I0t%6GG@n! zYK_SF>ks1r=)gZlM!#3-?zg<+>XrRD%iH6)0m52@h6FB`C26v3?71%IEtGuQETTbAMv6S!D8jcF%_17K z`$Dr%&NYf4T8UXgLzW{YmK!B-@YpP&ArC}KY%)rCmgCzvlwlUr;Ny{ER~W^LMXPL< z(2$wun~k~JEKzmq%4pcWNST6J#>(w>qKwwSwMe0=SqO2lOxiT);Yg8J8b$V4g#vCx znx!=KM5NRk%u7$Mg`2y@NHGmQaFNM~x0%I!CqJ|&7eE|{Kz_UqlEpHBX?uZ*^>Zd{ zHecGt5fh5#04x&$|D6f$J9+dv;#`@n)+GiaB_FRVY4Jn7W;qQ%7Ag0nS+45iAy~7J z1|E+T+TUdKjK9~Jh5pqnG?aCt4b!l6vDvUY%rd@{w+eD|+ayt!=#G^9u31vRCm=cd~S8{g6$_}$* z%*3QIWL{*l?zPQ^An6v#wRp3zE;JA+{06fyD@n7AhOI@)yu&P`E9BUqlX$2L9*!V5 zI5mQLrKSrWiy*jlY6Mlc9)&J?GJ@hmjZkDuZf;T=bh%8g$@N21mPbe=lZm=qU!?q} zrYsK)ons`PU9c+7ZoA)&=o5{o7D0IXln5nIjY*O&ITpe2rA8QhyD(%H)kTg)ijFrb zI^q-}wdx|Y#$@ET8Ws&AJD7Frg541W-)n@xw<~_x&N@R*WX9Je2O=2mH8GTY_&ocE zt!za3=n@AaC6AjWp~S61ApoJZW;=h~0CYzHQ;xvCFK!HJg?4oSb20+^vqrEx0-*9b zlZbcDnvA%AbXE6-0o-^5{x>GPfW6 z_O8BZTqs$&{(OLq)|)Lg?-luoh>#BeG7-Q`jGp!aT#06Bv2z#GGpep{4cSZ8`idt0 ziv0|t`FNKA0Vn4qS_b|iY0u;(m}?czw+5Z1d8Gb-yN*&D@%V{K>X4*M9XHoc3qQ-y z8+KK#0(Cwe=Xg_>tYgPShEMgTwpBdpdKG8%RlMz`?Zb{wOFoDMde&sZZ|zP>8&@o+ zIq_uX&(g&CQAnnAPAoCNhl~_gdXSPi)u;f7<=hYBP!dtcP}R+LMOnNg6a8MC8AQP4 z#Qe-*DKrm(ROEkGf&6EkK9Xb0k~E*uHN|3Msw>myamxDTN$MDIgzWTr=-(bbGv6^` zOJcO~h>w27&=Adv&l=Osk8)>kSp!TcfT)D5@jQ#)&!A0dmJvel?=9OzSL0>e)P{mx zq;3RcrWD^+qaOLd+f2PIIdy#(q8wkIr1W`{2x80~W0^Uqtd1T$lg8Ee&P>jF$D^LZ zD{!7NDm?*`z7@$jjv))>W+kK}sh)t1RX587^rXk1hh6MCer$FnZe90VEmJ@0&$tQ3#3gkVCoUfS_KJG$*i|JGQ%yIId&z7LHx#Q_rlP$@(OU@SPV-X47VgeMg-O$LUwjK=1h?p=|JgUSP0C+%$zkhK-k~$FD0{<55Ty~8)fD`QlOGWp# zqibV^g3gx^(R1k>=o5XgHq# zV*)f8V)J_B52*aJm2)&bZv@n5%MrHI<CtTsq8H)Sx#*paw&1cML@iL$fXwkStof4+nE>D&*ZR)O?&Ytk+KwxVCif5~Vk_s?d~YOI5c*e4LO{JLGG&OrKj2NIb!_ zXR4oC_(xqsBDH+)q9g`L_ca(HV(KN^(FFYC@xi)Aayu7hz9pW)BGisw_U=@I_eY9r#uX1}lEQLTC1#go;s#@_KlX5MOM57t4@C$(7n z#B-!^%)d79S3p*;PHq>XkGnamJX^EDLs`P;}h|o4*!oAIL`vFQ`|9 z64G?7Qz=?y?y08s0mW{nsHsjMHoMUxZ5K~guTbKT^(3hy^9GTC1v{6oc^Ve#~h>WbIq2jrTzIZ52cprYGZ)HGG?>*GbWbn%7Km*f&{d9$v9NAU zk~(@nCp|Z1BNcEWP2B=n*UEmPcdRj*i&*f}5U9{SK3#jQf|GYwuOv#>W^AyA&99lR z#fpRImkz256Afvv2p6)esSY&RQ4gro24Z;4j5hSNiojXd2ysfbb|#Z#wflHEWy7rq9Zb zSb3Y;MEAhx%G){0$b#jHLXNs5=Zdnmt7=nEz1E4=@;at-=L%M!n0U@!kEu}oTVP~h zMyk`z&2(12wA3HA$~NMBIW9~~Rdg!TQhZd4DKUTIlhT7~0lkyz|Mw`3Laz*fyx zGOx&N)C=&7GCR66Q;zh_Ze_*t%s$gX+o+v`JGv?H(2LOT{~&T1%F|%~7Stdu~(>TPkaZ7jH~b$N2sA1A8*y;ifb#RkAi%9U{7JBNUx)$pc(8ctp;@u! z3AHu~ps`uF-Y(OegcyxQfDh=4SziC@d)hiKvFMrr&DFci0yEmx$oVd=8vs=~3vz$($( zL1E;sDHKX#*EcF(##!O)wra6q`|zNnEle^O;<_vIC9hPiIIgWOK`S!jlUq?4_DXwr zaWN;JoT%^4;4!PlBzmPGIdj;k3y{0EI<(cca#P9#_Ae#mJSrIbbUl}Bgqg2%3>i|fku4iVE7w$Pm?!_hoXwMTbUYWEsshLz7L>G z{aE5kE-7uRl*%C#qznH7QUfTfQuTocx%d}_=<`Zhbnp}~^6Ce=cR+P-V3lBQa@@T5 z?dYaUjp_24T9m}kqCgC<0ef3mnHz40P_IXd^&)d2D8ctuaoPKVSw7|%??b0MAqN}GSqv-0D^sLK_)&xaYujDu9+a?#)?R9t# zd$UV=t%d=)LXY85tmwtX22bv>^Sh49cHU((ULM6O-1i@w?;;%wGt8%y4{tmV>5 zaUQWA<`<)>hZjnc!a-hoFJ;X6Y|-*YIAxQhlpWW#al^-_K1t!(ByYR4LrjmHvwKUm zKrfdssf0aP15zz!2!7^<2KR3Q`Ix$G%HGE{y0)F;mYIU( zSCq^cbq3{sbVV}MpZRRTaYwLLn3HJ9DC;Sd;JozmyFn6NBZD>|NoB^;y{{9qG^9s` zR5R*Vz)qiGEi+9}-1geiDDZq=YDoD+tH9qn@&nM10SjB`uu>HZwKD`3ehaY-w=K0$|S{7BlcdkSTU%tR_syQai!7r`oigP zWg&&7>Oeknr715;I_Zk7Sbn-PodmS%+_v=iw?vS_z7h2z@~bX2CuBH`IBt%TiZi;| z8dZF&LA{mBd0QB-tH}nu*bcGq*V_mbxN<%pZ2O@HxY20nG_Z!8DaR)>7b=#;(+v}eH*<<}+2702Dn=B}u?ZqcKxBRE-W z)-Xi?dj*iARx+R@Z|d6~=or>IC*lIg<&h$ACsi_+50g*DxqK*UbsT!yHh8dCiziks z;>P0bbcHaM-Now-H6ij19oaFUL=grDrdp^L=+kKojv~J$+)*2l(I0eMk&dU*6HQ|D zrSjcGZ+S#xGE#9=fHe^%BtGxOfBVpvrfAe(s3g6MSzNt$5aEcOAXAJ_%0*zE35> z5V_iND)jy)yu$>O$IdaTB8@2+w*B#PDBae?Dk+7k`QYH8Ch1fp&|%fUY`%oIbZifX z0Nt(-hpBX=>G|9zWy>3(@usiyhgB#P9Cw6{HsJ+LO8MDgJI5)+qZOKni}XpU&2EJrRw)pLJ(g1`(uufWF5-nuwSJ?M@k#;J zq31*}56{9=s+2+@IjuWHM`zu`P)I`+bZnNaK#;KAv$FQK5vNSYIi!{p|L0ki&t!;Q*&?P-^#&7r;_lmQ7Q{X2(KF* zY84p+m;Ft4A%}Bihg-!n8zj<^*8Lug{$RQ+(s#C-6I7mGL+=>-zNpSXx#j6_b9(HL zY8n!8q90uYiuog7wY*%nI%Z-dEz<^nf7x=KEM@+i%k<~$A={&_Q`Z7Rk-DY(Se&Ux z6s6+2Uk)tXy*|3LoTX9eJ(Vz6rdmv;R7NVLvR|hCDxqw7R*??GMC>fye}Z9N_32Y`WSia}lhKebIXN1O2}hYG7P3Q1UqX0&$IaSwSZPG0c;Y=c68og70qNf*WQdju zco67|b!o#M#`Kp|DqnruW)0bVP?hdf$;`i!J7gg&_Lo&+KI>;KH&=3P`ie?oUF??& zPKCazl5pW=H+)Lj5%}zV8dn+6EsmSN!bLKDErxg0VMD-ixW@Of7W=va&6YfQ7~&f$ z0i#zgb<;OvqHW!hdKQEA;9Uw3V-mH{x9SqM*}Lf5F;SKpNIZl`6<@uo>so^ad%IJ# z>6l7{G~2l%9gm4#!c;b)dXBHDVX{3D+w}s7mha@VrJUVOwI(GYlCX^ShY95@9Z+9o z%d_3eB@lpvN^R#$C3bf{@HD_maMma}wNL4YsW$|xOC-;fPVWaCY1 zvY-em9?w4=Z4!OSF=>37X!5#|jm;;IPBgv9#ur~Yx_a1l>15N5*w#w<;$lOOpHh=X zz4I_fe{0$`jN<4t%ga-(`MvxNF59%f`2!Ae^I7R*InZ{fnJWlYL!LhJQp1MkVM1-` zu^Z!XX<-&oMsI4^QHP(y!UBEh8dJ9@74Y!2N$Ti*pGI;ecw7#ZUp|^)<<0)v(hh9xBM!2)Fa(LRHj}@^KA6mCP`*>1b+aI zOZ3TE5Wi5W@b?Z~Ad6Uoc8&&Qt)7 zZbhUryVDc<)e-}$ULuE6%EMX)XQd@JH&*(p2U6v%R>DIrb&CDREX#4V^WRq>k~Qss8+XODEFftx_~qCAxKh>H2$-+}}UV z&i++={6tYtiS*dNsQT1~HSbuHK1slRJC(+XTK{aVl#Xh(i;LuGE z+%_ExcoLk31r9bmX{R*}&)1ep1&XD6_4ao)@Q_SJ8<)7tGMZL2nl}CcXl}8}9n(p)^gKU(J>Bba4(R7E6L8 zpj+~Gq%-F#t(w=%QW*G@kU6Ogw{4sJhLceXrj^}Qbvzq$5-Ep)ayW2O;4I*y^W7{r zDaYT&Qxr)&)|6k(a>)EutP^=UL~w)E;=_^omU^Ua%$2OHdc*IEd}pce${cHm zXxp#NLUHM|VLUZYOKco1xwdYRE_uW0-dTztyX4>@&ErJ|w!C-w&h?a#Z;a?R8KmF~ zJ6}76@+>aV=jbeHHTdCCMrLh*@MGBuJ=tb7pI$=Cg(%|3^jh1ideLJJRDFsNYz`)+!}P*iPGP3@z!*{PCG4w}a@x5xM493cW@t>-T{2nw_AWj1R($Gx7w4m$ zY;SMk=CQ=c%BT-#!i@F`{VzvPVz%pI<$?X4%q#XS(Wf((WO$2&k98atH8Bsh%QXo$ zYUG;4^rg=qn#5I~F34kYeykB&eVmecljInD9k#79B`(c_5jftPkwwP!2sdy0bS}4$ zjl^}8QLXEXjte7?PZ)LMOcd#)Ev1|@Z*H8dSWXyUE$tI8 z7QV2rw>faiD!_+GuOGqeucUN=;*<%1dTLGWnPWypnXnTK zR4*Sf#~wxp>EdtDa5K9EI@hkR2H^2Qu&|2 zI-XvFfs~)}=4$U}q&f58OqbhueDb<(Bu<6kEd_(OMIIY^Z1%Qm(X)3!VGxZKF^Rwg zN!-Oh*)-~8ono}b!15QM=V@+=jYf1#B2Ck-fFLO|nEycY(jq#1`85I4cpR^-4$;YC zlG@tG&5p-e4=t1d+)G_pXl>H$Krn*rS}mR)pC`HDcA(DQWLwbv^ldjuG1$wLiNQG= zGDbsEn`gRqbnV-!3g#{oPdqKL6efP|q7(2kL7Q@>+HPOTm*B$b+mV<|Ku5jY1=Jo-TYZFg3hkVh_?C$jVuo^2vu3liedVOi( zv2eNUm+}JVtG8Sa89az13(SyF8gp04;L{$2F=gjCQ983)hJRSb$iBra7V&BCZ%$9V zQOyRni!g9L(R10&?YLL@Hr4oz90P5n_E5GCJ}NjN0iThK5Nt;4J4(B)0=3A-)^4lNw@&py z$gX;{Zz&tB4qX0}u1#WeklD1$D1uFaS}aZ9QM4hpg<}<*>oL`?$5r~2Y;q=!?}Gdy z0t#*FgYQ2j3!YJy=8!^QTxC|FBESfm-gNCvaTq|9+R=~5BYOXNM}L2*Y~iqX;`gX2 zL>MW5a2L9A)oM)UJp^xYPy2+~1wBR=RE7~cj7jcJXKtRUUI(q3R4*5A_fYN1?y%af z_TI}4S?Qt$JC}6_;x&wolTvnsT?+r>C21_yK4!ujT@Zn7OOINq?Qv&V8Oy9g`Fj^2A^45z1b@F~X*462Utmxzv; z8k!w8*f8j*&3M@Zn0xq)Ovm_U>Ce+V!uM5-#U1f6C z9d?Wv<+J6gHhgZW;sA*H7fgW$z{!ild!LXjy?igZl#p>=DCMc`(^ZUe8@t)mxpEIa z+-i|`x9RxBJk(Q#l3$UOiHpTcg_JeCB}OMCDddVl!~Ckr8tnLDmapw@M7qx19}DUB zh~BjuB<-Pu{D2>Q2v1A4ty_Z({mpY^r=;1XA-#AW&Ux&4k%^0)G}L`NqpVnYpEjN3 z-`?f+XoUtwc>hhK1gXOUi@`v(*w?X>9yU}{{GU~0#zvyzjB!-@%$?Vu>sPO&&V>VC zlGQah3G>6+OZ1Wl(HN}y6T))i4nNAW?^yfM;qv>k;DCck;~2keY8=p2-Ezql78ne* z@r$D;V^7Q>oEtU4AP&LN(NK!}_{1Z9We>df$3%t$2~?Sh^ws57KjS) zU3(%TBB-A`3e@+Ax(!gP%-%HVjU>+=!h^lihxX96O3AaS^PF z!NIQfiA$s|HUnnOq+-yO0GG@4cfe(x{h54 z2^~UYqs5=%-~DvnpZ6ls1q>toXLcoqqfDCohf(yN1TqktH`Y!YjZQPwWU-X17Hr1$ zoP8`wb!;>Mj>gmDC#GU%Ijf9vx@aG$1d(7U4udCTn>bs=M`q`%INvF7UXSFJv&iey z9|gC9MI{YO!VofRv0GG0HwWxUJE%?|6df1)$NI|}EobSF`ocrYbXa}Z{B=ZqxutT& zVvT$;iO6syhu5vQnIDp*$Uq!e@C~aDpz>qGuSCFG5va@JYNFqXj}n<8@`?8 z>)%J4mb@B0gYsCWVY?jfyHUzPXE^SNPp6s^HNcd?`~%H>@=>Vf9k+}yFo*cPP)e0Y z^23%tZ2NSi`R5D-1uQot-O=fYNc%DmXV^pY~_{{zh=bF=gZd7Y#zg@a>e7&Mswto z$GD#0T#OZ;+UD`@3A$_>PWtOZKj=_!QEL_vgFw@$`5ZVZ)}R9F79bvVDoO3#LwTtb z9WhJW#X(DV?qh#Abo+|saDuB)5723PioN5t=!_|+X^j)aLzX%hPN;c$;j2SWdPnCx z_z7%Rv%;9`gSk$lNTh48QS&I)!FzEol1U2u^~Z&$K~;z&U7Mp8kh^i#Sua^Iyc5z5I3&$k;!_Ce3y^rPW&$1F6z*I&rWlSn zLo@uvKb@B0;ex~(be3|EhR%fps&)3oA?Sm>4-b47QXj9_RFh?An}wXX?j`hz`l4bO#Gg=*I1qKLUCCs>s_z z2HCG~oto^>77CE`+MUC2oq-z_LW$d1R}2l+i7Ha*oVU%R zr}wsrZGxau*z*nAhGER!gtqj!ElcW@*h2K9QFtz))ytHq)A&KLpQ3To-%D3LtCq^p zd}7ruH-4|ku(`|k)Vso!*%ZrY8TvbxYc;-O1!Y)is$hyMmb3Eem>80t40}2_vysMm zjM~eED`9@l9<1iEXK}S>l{x&)&Kogv?GZ!)+a|-crN`c)x-;kaWvh}M*6Ip}LLE1M z35h2l?A6Iwo_^zut`Lw)5&9$m87fF@99K5#-_ zD@z`mQ_M~acCOZiQDh*x_M2@m8lt(z_iYd1W%xom{pNL8i)5us0!}vo>`K%3ly2$g zQGgwHV!4L*(9L=_HiaPPI7fg=>>RbpchT{E$IYuaQ=K5wU_^u3Uj4cxo#jlhhp}xx30k)?iH71}+7ko9+9maA$g4`$mg?UMt3iGg*Vu>t3G>X(|^i zN3pQ2v-bkw-xMVA3Z1PFS>EKLJ(R=7q?x}Un!g}y(_pFU<|LE=8gr@A9i!Uux>wsB zI9m?(i`9w4uGId$sJGF?weTj1d$Mdb&1#AdoRWbzKU*r63-;*hmDHS=H>U4ktJG!7 zd#JgsZ9*j@43H9Gbnk&!v=MW|Mr`yjEA2=Rd`6Mqw@wdQ#K@vVbw4Zv{Dz$BxoV?R zuqgxFl;Ra=x42WnC25Z&$Sg0$AX`$bYBxO{23|P&go>ZsWLnOsl8)@hvU0A-dN3k5 zi_*PnxtUV_ZTb(;YGAFpS$f<4Bt6sT=#dD#LFwL)%WBpbid?-PJ~Q$LFpj^(U()j_q!{E0;FJ4V<9KR(aViwg zH>KYP_lG9Pug;%(F30t#C=Qf79Bc30E)8OrOA~xgato`6ev9^IRGby;$Y5rYQITjL z|0knZ0>$)3*gw4yT1x%XfjOjDynW()qgdSFF_gz9$FG#~o>inTn>YeiM_~>mLygTl zB0kaaH-f2EH)o4?vX&HAr1mA!en|+q9L4Fpk=+}sdA~uzB`T(wXBJkk6SOaIsF4+NYF?VqBGsQKx8|`S0(BBd?(7! z;03T3|mDtY-sc=QWLhUi}tBvz*x& zrgP!gH>43jsZFvArU}+r#}Kqo>@Ir2TaEg{QyFNP0cC7z$we@F>02SL914~a5@2?n zCngU~675gf03D+GHP z+ykNzqZBg`5bmXs12{FJhaAr*u_@VRlxqtIA>#TfA~(ZxBR-7jDU`|=9hix60tX@k zK|8Z9;AF*8D+IZz5M|Ccq)|?9R@53Fu{SJBMt`?)CpezxowYw_0iYgHl+mC zgBpF!yd`@0Z6NXTb(U}Y+SZhhwoBf@x;Zjf>k;mH+a!+Gfk-{qzWsD*6nI1DHWbR& zy?sX71kuyq-as-NktqFm<~x%0A?WJGbLzqdcMTZ@H#KKjUS;83C$Cduy-T%VCUKdx z6%FOrxbs-VLXm~`(IWc{mWQcCwlO-9L*{qL#Xx%N$IsL)4Jkr5Xo~KJ7*0WvI?xv- zYr8dwGqhUA%)?}{bh_dpSXgZ}I4>BmcO5EeR4*!k{R?U3gc7Z$bJ|v|lqn z?6}tG{2H(TdA;8B>fS}1?zA*^d_j?mKEZ!x)6#I_x+JJ^V2l4tnTf)3-3A!bV03;|zo&_wU= z^|#nX9ImBNvyNcI8UV&n2?eXjH#y$-zJ~O1-uQcGhqgyRA>^IZC7ZI9$K-kI`_ThI zK&kN?;?h}Gp@67SJ{`Rw{;S&n!v!|GSNS$g+z^)oz?`;ik6o_Ni5ueU_`pnEMm=G2?qvHE(l0GZUCGLNnN<>I6l{Ti+vQ>$(MMqv2|4%NO z#e&Q#(9ze$7kva2pa*0K9evkqj9}cdLtq6a^uW7Snu;qlbl z9J?i2cB_pjse=U-Lgz$XcAq0};@F<>=ys zSV4C5JqlS60*ns4clHksfdf_Oj2s<#@9cvgW#dUtc_QYwA2jw?Ab5iD$PMvO`u%FL zXnA{yKF^}g3d*O^#UB8E-^P0s@XykzTAX7nK{a_5s=X%uu3G_u9moCp!(sV9@R~S% z5P69FNk?82@8>V331>9v;0_U&$0`bUvr-PiDKke!3IBb@h^|f9&)wpDXB5(YFq199c4Zxma z%)OH^2UDB0i>as?2mT}(*XfCRZ^bSb`p|X}EV}pA=x~$45L`xa=LeGE)fEE=v{xMw ziFyQF*}_{R2=X?k&Qx51?YK1=P7W+3n{%UN#N7>Aja`JTxPk9ut3>_7IbSGy>SQsk zlqyz10xPJ9hgwQAETbX)2DZ=M%HAy96G4t4H3-(H?caNTHMP_PPU(!s=D0a~H2BgT z!W#dgAlb7eoY>lYCNPK<0@H=b8;mMFD6_%J+ulsYGW%#CR2xha(L-ZsNLV-Z$;_uy z7&GnQ#IPG8O)wvU;?DDlo?x^^$E8j#nkrhOluEK# zB`<4Jk0c`=)TXPqP_O7R1c}@vGdgRL%j&GGD4`dQMLU_#-)_rVyVpT8@jwWb#A|O0 zmBbq+z@^4YB9zSr=nOlG0}c`UpG^^LQ0Vg846QE!R|IkIl*|+2-qqbpje>soHg0UN zh>}o}oeL+_Spq~2)#Ci<7pG}oAdPsNf)|w780xcf5yjI_wC`Vd3Y!{!>AygaFN?g} zQH!;Yo5A90F0FmXA+##)rXoOmt#+)@tz!gCs-(@~O7p zNZMAmvyO$Rq6buJcG%92!2I&5b})Vo7sh(S&aFj^#r+3m?hSQbNZHFigx=tJs!m8C z8nxQwX$@Pfkw;r29*snv`(69Zli68!+} z3fmAZb;F0zj%D2|dsbZ_mpf(8sy1o*to4R9w`=3heDE|TEwWhH1baA3h{|0WBtsEVSeGZm(*zp0-A;WA6gbf2zLJ50r(fhaBy`W)D{*%kWjza^%sU-8o7k2TVslfb$|2{I&_=M3Wbzl0-FZI^ruqr>&=@U$@8d>J`LRH{(d z{^#>luDg3uMdKee9X>-QH&t2}_Efawvp#{|((EnMu1_baqt-8nxdP6X>EwoV z?S6Gs-pLUkW9C*PVq{#5e0E$Wo{4K@5*-pwaed!L#iZn<1p@Ey^@}B!Q#yb7X|SwF z-RVL8A*}3Kz?>}0SMS=x$V1=`LL}2|PFiX8g?0agto>VsoC3KIdM(x5$^}>DY6= z_gP>OHvkX3!l#GtWrZ{(1O`RqmyAudaIPI8^=Z929=y3Ne+ z>9|T&F=uIF!+O5_ozEiBC=-7|Fopx8OgDeRm>x~(Op^}3&QVWUWcjEsL5OiMjg!GA5kcQzT*)IxfHL9t7+Nd7u4iSLPbgfk#S2C?Zkr;Ja2}@~eV@`yqLljSanyuYn zocHtiz_B0|3+dFgp|N|%G}izY>|C*Y5*if7<58xn%HU|+s?~vRi`)&awY86%z_q3Z zHUeKY4rh259uz*O-Infsj#73ELwp}OhFA}{TtRR!1}){$hC}*alPxqvRMHFUyBAG& zun#NYIy5L;OSEUqm^!sargFh%azp9|t3#P6MjrLBzZ2<%sw%%z4ubXrg8;P{{pZ^KjT zq~$MTi+gaTo5MM%NvY8zqT5AUwD;&Eur*Mv3^rjlKSW1Rf261s&S(^+N$RLC1s&c^QHg z95?W)V3)B0reR#VT>J&tfFBiM3N#95b*D4`p$3AS=2Yqmx1N1UVEIR>wK`y;8ypZ$ zN^D46rRvtVvGtKwwXI(&% z`Y;F?P-iG{&PWRk=XNg4s4QUB^V55#nTW$)WdtKfaAN;{(Q-jMIV+LwRq3fpsf->V zK7(!y9fn(XgokS|V}tqY{R%yDgx4YJAA1rF_lSndgp1oj;#N?~(TBK3bRXNQYr|2i z%YWnuxTcw+Q&G1erGfJrhb8OZwO)5Iq_$GW0-Ur}P7~iG5}`=7V!?irnqE?T3t0_I z++^(JWp_w#1mDJg2hr-@K6Fj@g;3GzA3p%66b8MOU$jyG!p>t?2G7!XI_W-spGk#g z4S__n!JzIKzgrY>(8Ymn;K%pPvpkl^u{UG_g=LWt%fh&i8))r1EQ`K4<882@tJoYd zb>lR#B4j5c(0itiV>E2l82ugXa6W^F?&ui%f*R1Lsl#P88Fj2i6!I&+gt2U>Tq!|V zej4;>@eEvw@MIeK4V;T`Mg|Z;ukkqzco5;i4o8kKfw^Fu12=(x@+D)64l@wGTgeu| zCbP@%JE#@PZmZzrNVfCCE;kFGbEkooO+=>6gE^-DA!;Wb=Cg2lY{&D&6=E|Bjoyh- z5sX#Ew7P9l4&v8Aw~O`-WgY9)-L}}68@_Cerhs=Qgt2>rC8bRqq{*Pm1n&GY=u+e? zWk(=9{_yip?jh{j@GiM=It~V2nbdtdRul=D@WiDnS|j91fZ|nOj%rqry526+vo4u* zvb#y{*vc)kKo!d+58eWO&Rgfv)5qII%i$)N`xT?_48XSZ#9KvIq;huIC;IePV07T- zS{3IQ+rA2s4>W*La^+Wy(*lZC_{p)lOtkkNm%4c(Fr~C&p-}*Q{a2FuoT=tEa@XYZ zo}ITU_L9))E7kJFoyC-xKcF@n=WY6GlIHz0?x)lRkUT9!-~Ni((~%Gd%&}d+ink3E|FA;`eeF}?Gvi>WI=rFYII~!O~5odIyc>Wvmz?U7Uu}RmljJo?iGXA+1ZLu2j`-*2(hko z(wIu(0~qYcv?_Lv4w=btUFGPonZ%Gj9Wm1Y=F_+&g(4IX71I2K*(S}YFndS3_YbN? z?3MJl?e&S}j8ZZ)?)9~YgTG}XR{R{PBD+TbqJqmyg1}^i0k9}HS^hAfx+E!|-QxIJ zB&gThN|j;T!?$`B6YaR9WEJ$IZTalFy@;pEZ;R-6$W$61f5=fEqYhayxvAeDH+ol# zNQ4s)$hZpShtB&((i~xvOFj6t21sc~di-(4)l1gNOD62_QZM6OvTnVdbGY|8S)FLU zWZmXmL4t9Liz;==x<0GwOOjfLH0Xy?RB6TjiE221q`oH*1%&g10mj%2I zNd}*eK0CS>X~QUQ42~*7GhlkdQg^J4+eg!wl|Q`SgQKX*Lr_X+C#^ndUY!GGEo|0F{!Mi`r0;KqD zh#VFmH{h4t6I;EKHDgC6FRO~s7!6R;(~mY1^f2F8{(AD>8ExTki+nZTX1dCbo7c@B za$LtBHqw>O^(!TYTi16L>ps!DA(O%yKlJSxnWV)uC^U_zYJTW4flkFF)KRl!a++S0 zby>udl-J<7hp+)1h7d0=o~D1~i8%Dn`QO3Nk*gO0QDxYOYC%Ht{7)qH4GR;Z^*c~i zVG02*7};UxDqrj4pF_`PWWo${3(jsEGY}t~=M#wZ@OJ>I{sltc0I)68dgh8QHL6r_ z*qsdu|J!$<(pKPNs1nT1s18FA|5ZkIS2TO3@QyrkrV#|0Tc_`6#=(6Xbh$D3y2Iv2 z6b@tN&c$frZ5my?0UyH25`5Q_z>MXjIuOdc zlVN{NDlzz72%Bz3KD(~uR=iT7U}KlGV$=EeG^%(q+|lNcY58oQ?cpF7kvxi0}cvW|gwU;{OB22dA@?HdGpB-o}Dx2iv@iI4D0LX|SMi&xVGj5Wd-RD$6&BK^aaP zl`=?^UEQqZQY|KsMPnnF_NzqGI6I)yP=(riIjA0}L=uqkPRm$K7#CQ2oO(#58W=hK zuu31y1A)5db3`Rteqia2tE8N^s6hfN?xQMaXz~vHF_q=y3Ve)|;g73CI1RQTR3N8L zsHYtx_(7rV`Ip;v&c@@z(lCup<s4ij8|ufS_8FWQ9~; zRdEW=^;X3xxx6OY-z4$`=E_RzKojnfiAZj><(NDjG!n5~q%oDiRx~dv=}?pOF()z5 zVU;3hJLQ7P5tUZKJ(O|vsBk}ht%)yMg+p`ln0oNL(4;G`88TvrWTMJ4PR|UCDK&_- zpGt+)XnZ1Wrb4BAlFDN8gBUpP{m?vmT4s*eO^BH@y^al)3M<;+6O*{%lZt5AH@_6a z8q+7{JeW7CQFyjL57|U7+ur54b$jQrBz26{Z;j-$lK?a6@u#VEleztuFH@Twc_~sS zpV~1yV-l9#$4pIGfHKpkPjL36qgQRw1`})k%Bo$p&B>Hc<#2FuN>vc{A;I+<)k+ci zzFCz@T=9KF6Ngg;tGaas%lGNUn;Jsov+MeOSUcSlgQB7_k$!4 z7#y|4qNZa^tp@(SQK~JY<>7!8J2SJ-Tq8uP4o9t8o-C)%e%Lvtn94cPtAneK^qLPhwAUP>wU`~#*!yAA**KK-n8ZU- znyX$NV}5F2%=5H<1$#f~Gxs&*3=UT+h)gI10Z(wmL~WSxvB|jjAv{X_$9))P1$(GM zZA*r<&S2dQNbDF$^`O2Qv}NIejN^54nL*x zSw#K;umnACNT0&~hx-hF850R}2i|Mjyf?-<`pqvGmbSOssyMrC6?cb^V{|_Eey|D0 zS8)I^BcE9uTyvo9Ew8MjgY$FY5(+XT#@E$8@_;#( z0veA!U`UH$EunYAfp~^F+Ds_bCa^A0)@Ozn#2JnBozg_%Y9E?WulBX(Z4^sfTtka9 z2un+e`4h9nur^|aIp=$|HDXhz7*A!(!~XwSV$_vA$HfX=xevD5 zxO?zS$*b@STkHC+(&xGLs@^Zyc6lpSXVfL@GEP|+cp<1e`(F@Wh>yp9iFKt?A%}=A z!#w3mfo+xQAeH}>L3Jx>^pEmAT&jv=m1?mK{E|z)gy$N(Qr4&av7Vns6+o1eZ-3Ab z_OfG8)|0M%R=jShkuq^$LYxDqgfyg5p*H!&7nRjdTvV2yw&x)(QXVlYRNHOOyFwq& z>K4v9Zn+ws*;>Yl;V6jE^8@rb`zqOLSuDm2CFYO6P>B{8HjiyI*_4jJ?fWmi=?zIM zSG>g;{mXcWmp-d-L4D7|+bJf(sY~7&U}w5kET_c$vAGHthmjYObnY{L3Ym8}_B`x_ zjxiOIdIob=QBb$!wDJ<$g0~`+VHSh}xG1w@1>5Sp3!xB9v+saLiyLs>{|C{8xhugg zcC4VxJSa`xcL0ZA5$gH^DhOd?0=1G>0U*?$5IP8-fM6R2cobr-gZGzjO*V#5ArKIn zVM+{qECUfC8bIj5AeZZKoxcH)#kb7$CGvZ2GiZsZ9`-_WZ7rc;;c!e4lH-+FxQSDy zNfK~WCoYC8NMQ^c3wRnT@!U<3&E+7QD`=g7_-IZEf!MfZ1Pp--_zl_;ppjZo$#hVR z%AGJ;9IE8V;Di?IsAXjr!Z@eJGl?>}2xewb8zS?Dy8scm=lra}xtlcJE&4xln>HXJ zYzuoC)S0UOJB{YfFdTIc@HRqpP!{2Vn`UW+t8_BVZb6)@hQ#D(ZiEV zLSZZOd-(LdhcIq^O8nx=sOpzRsuX+0LlB{$1uBnFq!EM6qMv=p;84iKqUFhvtgl7R zg-yRz8xX)Oj5-q{t8hM>I75HH)|du`%q}w!qpXCo5NO87tTQo6ukB^BUYlA>1`-&@s@ue9ne&`L1K)g>l{h3wuR5 zqn7`|=(9kORHm^b6s)*-*+UNT(7%4RjhllR;`@V#C_&a1?&gfQlf}6>WH=5x*3U z5X_)Nyy;di`hrese_nV+kT7RAw4F#*FVDRwN0?OA1mPF?7x6X@|(?86{fdVlBMbug%* zm3R|rnx0_mat3)As=8U2@k@d}#$5o8VYN6T8ID_7%xXG>b=(S@^ou8#nr1>?#!;!T zIrCxsQXFPnnqv(sHzP~J=P>VmTaz>o{%Zc8V)|*|`S1VMD-;IS|7*~SNmX*a{6CJZ zC2I|hU@BhFwZAe|V9+>@)zwi;%IZ2;ay9IONt~HP8grwbs!7EfHS#2rw{w|}nu7}* zZ~C<{-i(Mk$is|UtBbc5iK$bxi^`qR`lx{>m<_{xGbo%KQ75|TQE1JfYNhJgl=%?8 zLcJLlz4zC~f;vc+U71X%o_YD>YrPB|?dvNpoYjxR?7sFcAz!7)WZR%-sp8y6F@!@O zTO9>x#6VPSn@XFH1u7mb`%NPrnGlBMsVsHzd@nI}2+SY6ccNZcwDv%v~X9h=t^iF>%8Vk=K#tmA1q z zrY>L%YR01;buvQyPhIxwzc9#2w(S8f908Wb_E6qcum4i>4R>v2Q4T;|9yH5QK$x@@c%y9pZFih&;RZ2rYldGz5RckoPSL9{9h;gy?;Vy&YG8qR1R?}!k;+A>*}^JFNf*Z#%qJe>Y<{EF?$`E9<}+Wyx@l{bPser9Mw`COKk z{}q-n_Nk7ZDf6aV>Vv^Y$Ig65bbFx@#;^VbGibOR#ij^TB^P5?ZcozDGo!0GtVdbV zE1QrIN2URn%cVkLz3pe|gMUj>$3U+#IW=>G?1N3|+O_H!K;YfxP7B&#St0q5R&1-_Mud zc*%fXAiI7ZDwM+?DX&7C&Kd%{y&R3HljVB=a404=?4Al8jtR(~9PMwqk;BOOc43GP zG;w3XSwCq12mfjc2$I)vjK*2Fk?Qn>diAA2ZG!eke>27vurUYc#9?Le zKWzNzP8zXy+is;o_dPLP3=<8N38mA?*>3RG(2!HWMp)vNH-UK)>Lw{0tV8l8N9}kq zrEhXov9h90ew;Vtny%`)Ck%_L;uLLo_f3{mcK7>sR^QpJ?Cy8*DR*c>bV12qJz;i2 zkbLfwMjtHLxjKwBsR;pR>ceT{iG)TRY&1wJ!w9`vtXAw%I=(AO9TUFXqC+=egsC^N z>sHN8r@NF<6hiqp7Sw?BoB!fJ*iRv2pm10CJ*l);>ETuX~kAsGR$cl7uB zPG0xNj{g2q+1gdLDe?PVki5XYH;eCi%IM^(@8sd?E8&i7C}{otzLU59fA-!qypp53 zA3hppk>}cXu9&e#jXjgg8jUfsq|wM4OcHrEd+=-?c^McFQtRH9)YiS-*WIm|kBJV&il2c>SjaHe6% z>i?1ez$pI3@|?NLNQ2(AJV$E|b|x84)5U*{x-C;@87^&eMGx)6PI+(5@wPc5I(+x0 z6*+1x9u-3*Ca@7?n>BNyoI~aVCrXYiI6p381NV-MAP`qMdBchy=}!`2W3a9PA?L2h zE!NNxpKZ+X!5UaDF+nQq$S_ZL=%bYbojTYU*3J(PLB0`Cc|NioFVM7%HJed0McLh? z5hK0SLP%nQg??Ck{0LH#AG5fan@_U;0OkDG0+h#%#$4(uGM9v3T53U(gr*8fJ(|3E zqEi(MyQsh{WRjF~e4`FNLRJhR(>evw41Bt$4G}uys9{!GVta9(kjswI#oJ*ayCRrI z7~?3%;m|pp|4;p!-`Afox+`Iaj&U=*>nTZLqFGWnI`9!UA#7eiEHjNUCz+?FIqrI% z!}E-KbLRS`htL3M^dQJlBF?~;%tE8WvRp2i{54nIYcgza9D&z#J~#Zhz^>?&Zft)a zSC1Ud1&8+JOzjxs{#&yT<^cqUpe64ZQt$DNI*UD^YnyP!K&nC={>2C4h|Xj-C;eAo zMq3Y=YzQa9n&Xwvl5QBh7$p)hFFSUbB$QFIkswNxmsO*Y6qfUkhyL#BIRu%MBglLU zop0#tyvdA;TXLxLgoX^@GNaQfhb134%#rmW-g!yo{{m;RJ4u4+g)HNJRj@a#Sc|xx8=Z$|-oJCXF|nnM7-%#Vro)^R+$O|3PFVbEFi&6;W7nF+ zvn$!Fl6prJn=^LL#*vXykKNO^v1`-juHnsP=k!vSQ|>GcZ{B$N=1rxp?#|8Kn>TGb zz0a{rBb~!tr`w%fn>y`br>nftIsFhlWtBNxi{4H?`i39f?&Z<9|CT&t;W6VRtzTm@ z^0C8vlaW6}d3L|b#sL2qov?#)r?Oa#PhBe7Cok(L6=n4sWb238#YC=BCwo!kSp({O zrVS%QJ?9bDsD|kjzj072(b3e0VO$$hU!w8IuZ17ExR8po}n z)V)o8ipB!}NLZ`55q&~^f>Y=)j{HmGTgPlKq#0F!`(ds(y4S;Jy1hx#1-KsaI7nJQ zxyYI99#(gLDMDA zVIguyOMcnuq^TuJLgYgiU7dWWmFODvOCn^>@?9htk+dLi1xLiY+kc{|y0WE4a7@6{ zRk`!w9mdNp%_0u+b05-es6>B97`jbPI9NjP2tN;=0B`jtyolN*6_R|q&~_s_HTkmN z_Qt96!5zGd&rWfpJXR~P^j>hR=Dj_vr zz?V};Je6%eF%2^aR<6zAHge4VnsNOYiF35sxgAM3<}f=?(mbv0hHaSKUxo;11Rr_{ zW^$$B{pDCr1;IPk8k?2C_R4;ByNBdni2%4dKLy+V`d(kn4@t?L>y@2SX)j=#ku+|L@)i(f@AZ~;O zno3Y|3mm1iC^aA_L-Y3|Mvnc_VH@6(uH++H?B*?ygbL2pAxS6;)uWV<&KqhdjnVV4 z(8?wg%ULU_3{r_+(Q7py-~PuiqB<+0j@KOA)X95EXdT!E*JLgNb<0tvR!jiNzK7et zSXM})|9N&>cmiP2qlOEa#*9q~0yGQJnuA^4n-kR|tM1jA)6vV3$Kzmkqvx~m=v<;} zQ<4eG*&H=7EVYx|Evr0968|6Sw1Mz@gerGbqBeN)Fv3$$HYYElGS6f?onCv`*bXB= zU2>E3(R+5J3St`F(s+oRv~>2mM(QRhHzW_{q&6j3`=19a>Kt5usXF@5{jgo3Hcnz? zbOklI-3L=$8THG~@Wkj&uSB;rB**|TvP7j8`Q}wUANvC0a3gei00L^`^1B@3VmI*8NFQ3 zEo2~r^=Nz1J^@Ct;SiniCk6?NcPZ%VkIpfmvHbiI(4Ox8A!;c;@{VM2JqEqXqP*=- zFg&rRL3&QFGKp@@mL|S3W5c0phDG4xAJWvR>L6hQ{Z|TULy+dx)M1JhWGf|B3AVeq zp@%0|8Hx_kjZc2GgA)_<2&ZV8w{-*_d@|U%=w3?wLp1r6N7OG}{n5#p#5KOWuts#!2pgs$EmppY+2kING4sH(4`*IvkAr8sBG@m$UNY4j!hko`&}$@sbqs z;OysLaW+Qf)uQ$Dwo!R%j@Ago!M;PP3z!sBO)awBETa}s@KEs%aYZ`)p?^AVmTBnu*W!J{8&eO%F>*}Sqrjex!LCsV) zig1eW#>P@253$0_X7nazWh_g+pFH6nI#qkW??*`}Sj{dMah;FHu&;QG&iGTKkM;NC zA1*rcUM}hqefLi>$7^<(+J22c4G~n-nZ}`MuEOVBqh|D&@*W(GBsYT$vD7H0|CEo= zI#tJLx@#qkrNfrTOv!Mn=fU$z<#wa{Ycvz%VsK~bH%)0oziCPlAc&sbZmRN_F7)0Q zuDAhrv=kkU=qG0L$=p<>#_V*x5ua$u2Jn(=)=f}V=fi0_+)=K1-v*xDBDjl$9*((M?JeDH}bx7@k(l) z)-KX`=#c?1CG%l87{;gQ_<=Mn>hwPhRphGpCoqo&OQZf^C#`w<%ToR#h4=4L^r(*y znEgifnqFpHA8mf7!H_ivb21$z-THYR41yPib*tBua3|uO`$0agG9f=*+ITK5$~jHwnxpa z`|3v^MsV7x5R8LKz&PVao`8xQM)aRIq^zuzbi_ah-UANTJwrR}Mu(kJ@`m72fKgK7t(yBtrE zosil9*Tfggc7w<@G4cn&IeC^RU9hH-*~f{I(Y+#$)XY>^zes7UUW-$q^PUG$vq+I%eH$F zeXj<66bhBd8+k}ip)BM@fYc{%@kCR1J7`7o@w%*D$-_pdOSGZe@sb^s{!|P6Nc{ZC zvcl_k-ABI7jFU)R)gKvQeCD^niMr-&K5(LLpiRd=f?{>4JMZc709ks~W^%L{G*YHE z=6)VI#!tp^7st+aQ8C~%ax_Ac?4SXAxUaoNx_@t?>I8M&C_A2CRbSH^%ShkB;fng` zi{bU*SF)0r7i6AF#^g;v)S&ViH(&bs@iUU2SQc&F- zcP8n6lY;C#0)TN9cK+RUyaaOy25+u8{q8)6V-lo+gDN)34Ms5z4&1nV6gs!gMB(F) zEfXX1p2+>18u2j=V{j4om53g!VpqUEF?sj`QR&D)riB726(W5_D{7eCIZGAcUWY#wyA* z$0k-qQaw!qsa!J}>FID98U0KV@n^jY9o)u4jFZT&|(Tn=PFq9pVDU##~YnGkUf35Ivb>r%()nqG2?J(SC8nG>9!g`psBh?=`3@z z5V!WtG_qtn5kml3#;Ak)Dc!&FVH_8$jlg2){ z8S6w>{aAr{nX}_2zIm&OmR<#;3|UCqwo4XJx#@cita4#zakzA!q=M{|6#MfiVX+A? z6Q6hn7DG7}!VFHMpybSH74`=5AHFXjr;ZJ({Ohw9QYj6(;36k_tVZBNG=?d4V4sjK z_RfKBS_9Isk{s+hy-PVEZac#e$2CR)U~kCm3T|j>zyZz6#ykL;20Vx>efx}sxDbo5 ze%^t{5)74lF>%xfryEYb_arr;k~hVMr{Lif-}893Lw+~|L(iKE@K0{{a4ur?t%ur` zvff0U@?wUF2#zH-V6;rGV1Y5SUM4_G{%k=mNPe#_zDmpjpuTM6KjulQtWtP&Q9=FH zO>OEgo#0Cmy`jh?omynVOv)k2{2w)x8OU1}CSOl!427{e;wKyR9GGahLsGJ=c-%!s z&<;!W0y-dWRX-SFiq1UK{=@09HT~20LZqQJ4R+!M4d-YSJ8s-jyk&teY6lCqQ9&Lzz%vyh z>WR^pMIHTx(Jrg$3j;TDQg0{oX9-~=LZn)ge1dnS0He}y@ zqX~fS(3Xi2pyAC?QVPpGroi(fr+DlT^=&goZt)>F|2J^|(*yUY!#5^MQ8CGZ{(ln> zGv;xect^KjV_`rXu{maj~O6iWQ2kHtu`}#1Y4oKSIq2J*Fho zs<8D;_}cUeh$US=gYNhvRFsQYYC%pPcYN?LYC_`Z><9U8UTA4(d?yMtN*G*Y0*jTD zrkd^;XYaxP$;uy#tXzH;`$)R`cw4To4%B7qZd&YyhYjN`6C;b<@W2}9YaX)M#pxgu zqM|$M++}IM)pn$*Q}(psjyOVWT^qXyrpko={y90*Tgm0vxg?%S_=vQp^7(655zXPL z%usZ+rsy&DK#W~|X3CoK&e0mRU4%#5lC2JdB?u;WVYSodFBk5%Yjnyc0_AN`e0(?dEAhcg+~zNKWj9^>ol)2K zB-N$SQ+Cm*D~#p9uhno#v{`Q8mL`G#*#Qee(8xrkLT%q{(4vXk-d)E8qnPt;Qf1Ga z9I1nC{1y*)Gz*KIqE0~c)jc_Cl4IV&ANJHUC|i%3_sqjBCRw329IgBU5HUCSfUtQQ z=OCG($hM@=ylM}ousjHa4UELWE$|KycX_v6<2UW!zIRRzYLK%Nv5F&ON@H%N95^1) z%w9N%@m7RBB6rKwE?14A>Zp@%E@%3qaX)u2I?#jjyvCbQwlP(H2A1cAJl_VV2tpkn zN?SNF0+V`s=mj`I(7q6M>dzHVj$QD^ZhD4UVUdo4`R;4t0;uWZ;W6PTGKns1Id{%s zF`&#DAKAmi8Y0VT1OD6l>ZX(gQoI?>4IP$q{#p`z_-0~n?4dl zK*i2&`jQk!A;|msnV}bi`_8S`!cvmxVNj*@aL2h2v&EelMS<%f!mr}cCHv>$V*)l~ zs7_On!9%4MD)6@e4V>+d+1!Tx3;I-3YTKx5uA1y?6;WjeGopT74|gVw>i{WtQ`3I1 z2i{8MCGJ+tXV3A>QJ7wf6k8>VXpm*mBy1l4Wrl(UT1#<3*2hdcbM+L?Gd3H5DH}=*8bs8nWLcPkD#|_|QK9^7Fkp<@q7UH@dF(0F zAzQ=PF17#Zttg6xrSsFb@b5}F{?sR*_#*C>iPE?e>D=PJq)9xR4=eLW0-Ik#*@^Qy z%5S0P%6FH}kw2t;+9&G0H0G4X6}|)fDVc22$EO^~ktOp4s|RZ%^E%T0SOr&|C~!;r z9lOle{$FD8@YF<6tD^j+K zQ(3HU2=zU53C2qM2DPf9;!2r5s$#nM>Cd0H3kNzGkDLW}B}YS8p^eRG+zaHKDV9Tt;)|10Z&yU=Z$z8a%@R>u#u*5q91HflYsa&$I-NxLz6KrgY)<7c?R*1GFa5^=|p7 z4x|nzh6PDTOK~N}XgyPZ*$Z+FVsb|sgeBWu$xPK0+s2HIziq6C3mVi78q{?H3H+5< zzP_Yy$^8hsOCPV>-O9C=w>PT8^?cB!?B;yXt^K-A5a=lO`rz!xx~F!WQZ_lhus79n z#O|&C_Q<%j@8NUF?$KPbH)$@}y^2ZTVV@||pj)rz;NjW)@sEgyBVEOjif^aRUK1V` zkD0GEELMOTv1X);I{)z;m_9+5Vof71)J5Z$s&$|2U=;BJ8;#KZz=^V-_=ciOu`=*i!;l4e-mrIq{jw6pxwi$a08DicK&T7T2RLfaDR?x3mO*O|l-tCAA+J`;f4& zwf~>Q6iCXk@Ja$aLXdgu%za`MClC(O(2KW;PWY>p)H?P2h9GB_%YhSy3i5J5&U`fs zS*zHtCyDLf3{-^xhRXI>X#=f4*QK?f;7R?<6^B?$!0^Zn4_=- z&WL|{@y!3_O+a0iO+(UpjU&<|m)7S$fX;qC3sT)XKv&`MooRKsX?5ZNmK%!|H*)w| zQR?ziLnyw9Wki?^Pe!=2#dph$_c=|(TxEPDmF>uAH;5){4mDf-*Q#f-*ptB<^rks5 zN|ld~u`WWtP+%1CNcoYoiHft!75zouTQ=NY59M>yT;R9WmUdEe_2Imbd zKTT5x_rvyTEt=fvMNWYH^z<1}#p?r+K~)3DLSlq4e|ab=zs-(p#V~`O2KBERo0Cg2 z@DGmUsCD{fA}-b`Jm16Aw7uN{5c6JKf)~({oiBz8Is5{K%)-oFcAci|<$Hgz+4;&OBQrcwJrbQj0E?GMvY4 z8OM(9u-e(s`a1A`xj2_b;`v+s+9Z)RyHWGhtfY#cI-aF%C+-2z4NRO;At~ade5p0v zh&l?2NO{5P|E$!;NL;*~^^3_=0jpKAJar&N7y3atqA>FD}Jg1tG zJGOUt7&`3FFt>L!BNt^v(vPDu8CI|F`qe%BRxlMpeyU#BG}H zhuhKM&>=BQcu36BqBXQmb_dGyA-)cAU!*-BV)%*Y1O6ORo)7+82r@B`dYS_rTXB1J zvyZ88w=hlU;O~7_{~f}Oypm)Z@-Z&Eh+fWrC}Ile35na-38?c;C{*n-rvLQRVCgAA z)>KM6=9lB3pQO5in#Fb|HZo>z&4Yj)M~%8q{hM0Gr_ao`e_H_cox6pyHdnJ zW$Y8=VE4;_27h2hScvawL{AD7+s)7P#p7~BE!2qXS-N89alBjSv8fhu+GZj(Ie3Nu zsc}$y;zRU~5Smi`W3K)G*_yHS86-MB!hBk!>^_#AE{19Z9y_FT2|AO$1d3Vn7`E{2 z33I3(L&S!uFz0-I44eC06wh4Id{DTpGk1x~5OqDL5T?YNIuDyH!7Ny0#k3#}9eyN& zm?q_4CbBYRmeVz8Vjl5G0pE{q4ZUBMzO^yyQ!qDfUrB5Fw~Asjj3 zPK8@UHl{_rEiuojpkC!axLK)g{F!(gBKY~?cv$!1iI}$?!PqEMna4)HA8!p!%24O| zvN)UXl&{iK$bnuO7{#SW7Ib`K%tegmb52e`=g;rgV@z6lVC(UnPh)DNJauHuc z$fEQsDV%Duih8M!zp~fEvnxbjnM719lJymz`ND+M7?RGMR7-F2Ev{X(rj7_V^R>V| zY{P|tKoc__!_sc0$ME_{W>P(z^#=Ko{QTd~x|MaPNG2MTm82>9@jP!=Tlbtmy zgSe>8nmSpTPDhA$2?&2g(B^%!%n=g;a0?=83KT)F>g0PvCD=jH3STlNcS3W0pI+1quzyV3(>X8SeREr6QXTLaZ z6GO>sNNl^naB7{6l|5T&$onwthE6H!*lkCpF(;sHFM*UU6&@vLT^Xo$Yqmo`@`s&Z z*bn1(Yc82^oru)qgbqJANK5z-JkObEo9&lC-<6&6M2(6sxPtAYxrq7a7qelJEk+_c zYFk^pS1ibfUXqv?ul_4wZyaJ)*BilO&2);S$Ronyr_J;80V>~G;DmT5Qk~`3QOjs< zc*z`QNS1i+IZTnM{i@kPqE~_s_+!zoA7khg#^bM08o6PMbQ_vzq0-3D)tA8>DNB)>N13Lg18e#g4XJ3HDbfXsXCNuY8rRE{@}4J=+A!cbRJ313f6& z)O&mA}^gm5n~@jhV-fl52{I0VeZnel_|(N$aOx3TfK7ku{H8 z_kD8O0r_NHXx$|50rt@;e{FC%ekV{WcDKx2Y-q)9Y=4>ST6MOq9s0PaE4XO%=9gVf zb1d*Darimz6n#`G;yg`Plu)wQ*I>z$sdgSG**aXYRPk{Tu`+cg#lRLA;IuCyL~LX2I%zhH~JCmez zu77$t7BL=gSTedeRQ2x$Mbj^oQLLv(gSA`zUsM#y;~w(*v>ol&= zrP`$H&`8VBoYq|$`VX(f25Ca00+?n9u-LSFv5F*9eRvS>07 zPwOcO_m}2@_FK-Q;PohCJk&FOhIyq+WrPe(As_S>57yT3s|_|kTx(0W^Qfsv8&&`M zYE0cl&XdQ3NuRt%&AI>o8>5T<=HZ%1kI^X5g@0>`!sF0YGkO>KIMp&DLK^Fp$AOS9 z{>{9q;y@?wQ{?&|Z#d^U4s`G=T%$Esg^>?LW!%ykc@6f}UlqHjeqBoglSO^92W_Bc z)1YU(hJ&5Ll3l5EZ09GmRx}?!bE2rU)uAdGa7S^7woIMdA#QkXTjM~d=t}Lt6HyZBLm?hj+37h^4?XquIE4L*Ac{pg4O6tJO&NIm{2(IU zXiy&aSE!j^O3MfK5juYQ^_UmywM3_p{LAautP~pd+egx+1#9Z5%EO9BrnrNY-(gqX z$|NsuWRAYUk}FWVzc#L>U{$O-{xaNFZEed zUueLD=lh=QQGoGOpv*_#2`~m*3jQ5CCJY@snrhR zQFz-M8}foH9{0w^tl+|nt^T!&^>Yh{9gWeFJJL-Uy<hyjI|Um z5aH)pgm)~nGJ=aE71tZ5>#ogF>-2pBUN=`_uy$+ca|;a|r$^Sp%(qmQijeQYyj>=L zpNYSh%ija|dxiXsmZm!Ke*e+C*3i4 z7S~^6P2c=@^{An2#*}>HS`LX8Eutr}@+6nF^^hGvy5c%hsEr<;=}pp;-UP8$|Kd^2 z096tl&-02ddlR_7&n{EpX8hTZY@naLuCZie52m&1oKlT`(~{TB*#PyviAR=`y2;4q zTlz0FMmCQ)3}fIgO0SU@v5g_@G`fs-S#ls*t+SKaHN*hriG``%TD)`~Fj;M|LbMJo z1lLPYo2}DzaD!V1T`!`)`Frf=RcUPPslsJTmlgjYgq(EnkQN11*tT}|(^3W;c{>Q~ zNuucQF_xHM&2XVnzGT;^z<=_Zn?ZuA&1p6+fqSc~CkuJTKNzC5rU3v@^|jD})PTY2 zlM#s*^6Ov*OxFzqu3ZjBbmuhmlCl|N308&IG+8r4a#bzA%{i=N=y&6x*`>0W?%|dCSepBcGTC#*RoD(2q|uI`RM zbjpXV@rQo-u*D6cvH0UY`N)BKL+~w9o^v^dG^T%zBXd2~{{7Ek2=Fr>6bWf(IUXIA zKX-*SJ8D5!2WNLB6~?KTPb_k)5fWR~{l9S+~v%3v~B3R7?4MJ=hf`7O}m{2OQe z+~Fr(j~(W(%r@e(F>B^3;YG@Ud#Ud^RCw0)kk05RwcpBrIY%SMD^ouoeCA7bX$+T| zgNXL=P|KI?UYQOUzhV5*&hlX6~Qdm)Yu6g|&!sde^w z!m>37PuGVmz`G-@eRtsfcsWj&!!&pb`wod`gALN`YpBg+iCpbV&d!J4 z2BmtJppmW{rTeVulay@jvM2qCNL#vc`A|I-Dn;8XPLq zSB!>!`PP^%Jb5m83-)XG$ox66G-i86mN71*TUOB%hjKzQAXT?e7|?&CN2Km_NspwU zU`+)|y0#Lnd^R)~w_rmGjP{es? zGqvU(ah!22$yMJYxgExtJ*X&avB)&zd)C`?bTW4wsZg<6Y!3oIph8>!b<+KO7zHGG z0StN894Xt8y>l3%RPKm>7IDB2KG|G$wwKo?r_4TW^m?3BoH9jd|5te7=_8 zUTive>zsK7csZ>(*wdK`mYpiL%sgBA!S?x75$1y_!iARU%?1ExsoKQnYY36}pxLiA zMgc-^+gpmK8C93jNwI%I4hH*-NL)-eZ6|S(1%sKUowU9V$5-AF^CVuMu6Z*y+LJPK zlyIg8@a1pE;hdUw7E*ZeD1)o?vUeJzh3s8P5wu0Ig-o6?OS%#m8I&*1i9^aUXoZ%c zG@^AS--GkEqwLsaWZT_-wB$Htrz}E8)@()Q53Qdc@)z&SCCHp=Et=dK9VN1cew?iz z7jAu5GS_q%>@j-d^{eQKv&&={k-IS_5syQS2L6Os7SzSxF5{V(k1^8FyK+brp-D~G zc*DCeY`Uz3mklex6ucWN?e&5zZY?R_XU+VTjB8#$Jw@+)caEA6A61a>D88-Ih6~=5 zWVLSp_`9KS;4kj{$6c`LO7=wPP(S}jIGQxI;@vq~t0~ltaUO)3HCaCD;+MW#-H+7u z1f63h#r)otNQ&7k+sG&0kYmQi4e1Kf()7CoEK8pg&_m#qZe}YpaL0sKqUP08MH#T9 zl2P1#Lyj`lwt^A5?zK{P6|*HCJVB5E4&Go$SQz;=j9r^G`^3g$0|+W?|0W*^Ynt&9 z)!uf4!Do9UastreC=SrhEC)=GpWYybuyWjYge5z4XxGzon5OC$jfio+Br<9YNFDUM-($`PXq*5@ z?Ds-Xy;bnU(;+UdS@bgz#k|N3>hK(D;w3GheK)=hVy`s{9l&~R5E7$meXmIX+txt# z4G=K>X=x-TATI8-hFTjB`M5Jl1;m)}Jr^qSpNT-l7+#lZSJRo+e8qQx3!e}{uwtzd zs~TvU*4d3hWKy2iIG+QO**mj9S_;2~c`?hsT-JLby5 zm~!jVIr=C_n^KY+lcoyWtbR?}kpO}IA<;c)(iG&Y->*qiAKKOqDbl|FzGO%E_wUNj z84kvz=`la)CUESF!WS+{Mt?ubrqqN;mr3L2`SiDdnNb{Pm}H3VL>b*9LTn*xS;_Yf zca)uyTeU0n>kk-)A@XXbN^f1B*z#qf{LPSYuT0a+d0UY$*XW(~!q6y-|nzJAJDtsS7@oXoB$%&E4O1Srp4HdV*5dyL@84QT}XKA9-Q* zEIY~pP3yC@+dhP`%zgzit+PWyS(a>O*Q_G@gj*7_pK{A$$==va$6GTd*?WMZ|r7bT8p-jZKZo9&u83l3!FL%BF#zx=(DDlDsJxY>Lt4OAJM1% z=<~~XRCd{s>+O{^_noZPGE5abrCV|d{;hWC@Q9mPT`^eF*Rx&S>upXYvZ;CVt-QkN zPAAnrc}tGg=+9+>%l-JYnt4J(O_{7)dEwC&5T1I|pL+A8N$d z=Ujm+Gpj_CK~0@`OO95n9E2VAusaYrHSHfS{!)f0F6V+3%`1l9Bnl!f*g9r=+uVSi z>e({R#fWGml@KYeG~iU6QlzV?f7U4(V(4ohhQ0OzA!S{*w9}d?E3;nwXrIEMXh(1} z+&h%#^AVK*HH^mhj*RG>2X5euDlM8H!QbslR$Bp*gjEhM_)BM?&FPYgtGrMJEOukj^}?9%ua&fD231cR)Q`j{4`E2Mt@K`FwV zOK9asSawXpBBzFMbd`V|?9+;J--pp$V?~)TUYsg?27mHL&4pPvVgMNVtq1sZ46a5m z>Xp31lH1a(@S~65F&FjS9(K1mVTo>SvifgQyg~>YF=t^*Mszlq3>Mm@@vU~Ha%M0Z z!o5k?GzqT)CsOnL_3BGlb1M~^*1u45UlfG&&ZY-G3I@2|erJ@w+|;y{56%3ENPqf( zd`dBziR}5PDFMT7bb(Wf{D8K63_SY0%!rsrNqNzly-ZmLo=TLMhnJL^u@f^IN~)~- zn8Bpoc5Npf8b6vCZB?Rw{up{P27{K`?)x|^BF$D^<9Q~U3+&1SA2l7<%u@}i73NNz zYJHeJVj81Vo;LAG(~Ek|gwu!R$4D~Hsn` zK0f#{3;=XMi@!XKg3umubeuLlcz9)UXM^>wu6EcB-H0Pb=?yPlMJLXvbf4ARa(qQG z77_-eqBZqi3i}(eRC?p&86IJkb-<)l2px)HjjT zYnHI+>dOlAH}&_QFdtabq|0bRI;NJiOm7iQ!p3*AMvI5Blu!zE_l(*3suPX*WgbuZ z$qa%)1~coW3DIxO{7wm6)gEVm>elFJM0D_Wc26@hy($K!!vVEpWDJ245NYopW(;MO z!dJAWKBf4YWcUxEj;UERu^o&e)JHANfzrxttdtY+O(9)=d#d9m{UE<}TJ=0x&m>3j zH*U{SYw;a2HR2g4Wu4ZH&Y&6*T__ovHNOSqL$~MXWc^7kb@nI~+NM=H(K(LS;+CY2 zJRUV>&Ade|cV(X zoK$R_5uXFkt2OmPC5$5xvdVtpIoQpy*u_t|DbY`Ab$qf88Pk7PlSNv#BT|X3M*_R* za5AgreMiAUd)O^g;Ttdp5*7(^#&X#wO%Ab`izA@tQ#h{6uJr~$n^iofdebQR_rJk7 zO`0?+{wCF==X;rK-#4RWH7?1()Wa1LE5XI9k>s#}bHoqIYC%bwudhAvDYN6U{K2?L zTIMnJ?8<4P#gQn=2|C1G(Pq)J)O_ED+~Sfql!N)ynfBHsjx4!3)&SFgG9B}^wfWP~ za#x9#Yvuf8T1uF!$(n=_-H=}7>ed`gvQ^9O%r#ti=0|nqr%je|+iaY*_WxFNO`?dZ zb`8W(<<3TvOqU)=MsT;PKtJk9O)L$1=ACBSwJ2cpp9A`-)zW%AhjnS6)&Du!{^j?n z3UvFX6nYIhokFkV`FWqgcI@e%(1D16}z*~3Z+T0te;qITZ_tdvO0vjE$j&WAacrQBcw7F9(J&M zlg`}rJ9J;q!!?xdN1DAcyF6h_K3+9XP3wD7YVF5z^vq9kTFG+4GqEATz7Fy5M z8D4A8QGd&tmrtu`5`$jXvxCWwVRv-9S9Wa=_m=cOqs1)L zukM2EF5Kp4!KGq6im^@@sYV#Cj&qzznRH*d8yexLXoN&h)xmXIQ`aaJjrE5oD@hy@ zwXq-?x~Z=A>+a6cNp(aFu;5@M5|<6=$*X``?`|wRjW^zjBGio~*nWoAVizf2Z+BSx zYS6g%kD%Q};T`_0-ZjMFvi&b!Ry@aTBooIg@KI>U)C)8gvcm@1HQM1 zRy1R$$^R1P{0NR}dV6STbDm4;Dk8>eMf33}z`r=R0ls&B0hagsWn!*Tu?;AI1zT*H z=`b|rQn;}nMQ@X706Edq@AfauJ^hTQAs@D2b0e+ODuRBs>z_6fcI2+43Ou9W^(>7wEhu+~FX@Z6&vxEsTF(C+;b zPBdO9lOW+(W8h9}=7Tar^AgU_M+l^^|1Fi(eZVh`%TrDAdtcy3e#x)sN9NErmK=3s zVXN)=9zXrmCUJ`>r{XWagt@T?!RGBN@VC*|#?)>464fVKQO0rlh z;0#aDTx?5O33Sb`R62HSqm^HV0rxp&$i`!(i~FqU&nqsCl<~@f`(<=xU@{yjt21Ey zfY4Ig2fqy0My*W6ci#)B62FNBGC_ zOi>oSMpXQAT^j?FLRYgj^G%uJdDlPelo0p;4rNhk_$$7QaZRoW?L;Md0E1AxQv}zz zKqRGYYll?P@V&_i)yZq;KB0=Q;|PL+)&ClqWBDxzh*L4OnDml;FS%196}s-N>o7RW zF8XGx|68iZD$EXnVZ6@N^*XLJQIF6Gf;~=zKplt#S+k17Ea%d2cis#BrN?_)!}!?X zz8tmotHH?#cH-1 z;2ZVwMs%lQ`yzKUUaH@>9B7@QGW7cS{qEO z5PQZ|SWVa}Ts8BRcq>0o{exctlNcExIHR@bh?q;u>Jcz%UOlv2OcWEZF8#_^w1Er* zsurn-$PnDRN!5a#v&wLurvaCJb)hsA8JXI#jh~=&J-i0l&859PXA}=?EJ|PajmT5f z3zeOTHz7`JgBQVh9Qa*Gu#WP07gDlS9oR@k=3gG~LNeKP*a|W>(kQA0e#r?#E?UtfM^UgQ{ZJR-3fe@NvP1o_!*h;6k@LRaKxv)ghmane z8j4|c-`7CQr-hhx+0vpldx;1>mIl<)_6RJh0;GUweYV!$e=SFA{g6XYakYhyS*G?# z8w`%%RFwSiYY-_v?AYac<3dpQbyyq<7z8Cn{T9urRUk3*_EyT;%G-mC{5Yj2CQqWl zxxEC(XQ_WEZ;c!7hz z9c63BW2>nSzW&B6B5g=m^x=){|&6Lp6?A~Ri%BrKIX%y>`z2{12Fnr`DE8> zj#p+`*w0_MZ`|DM*AW!e;98y^c9h)!VPiqbt42L3c2TV*x`9TkyT6+ zq6 z@KPQV)T0*qO~pFG`sY~N*)ayf3cav{cx_iJ`bK^4@<5LYi_0I_M19f$UQ98)u>{kf z|0{$^-Li##i%afmvSyy7)&ky8awM4!v*Jkkd==N}g)#UWE(j{}jgEXc8B!mv=qKu) z8~pyiKn=irP^15HP%v|+s6|kq3!1Hd^;NBBGQ?1^8pQoqAk*T8=J^UcF4MP8sl*uE zUqf#~ZMM#;g^vr-<*d$q2lW2zRif`0Dp8(Ld971(V9@IBSx{KTD>Z_VYSdAy(Xwx2 zv-=nFb)EY7r6z0W+lte0n9$yDLowOGD7Eo56^SNO=kxiIYDE1vGf3e0LD|8ZHaDqc z?4!P4J~J3aRPy`Pr<(1#CEoUaKz(5W#HQzF_2Ech*R}^iNFOvl)cB%bNVhiexihU- zUtBQ&U|01Ooa3Vfepv(phip8kD6U_ zXqo)D)CqiAE`OuO1++qbaFq+bJV;7x}>aVP1mIV`D!?-XEyF4YU4N!fBSbh{*dJccR5tbUKrAPyhZgi z9`DTIhT*@($Y~6s&>CY%7Arm;p}+X99@Mx~A${+=&}F+C^*E$b5$9QpY@d9%3n0a$N zwH)=&LNy%d(s$VobW@QXgR~ZCclbfoju1#FyBEa_nQ=+sTmLO5@Mrmm?e^l^g|u6^ zCl!p+`*e*o{8S1)E?fRR49~%?E*4mv0_^=B#zy^HW0qjwd?0NIv9QJF2@Yx#kN?fz zOFU4o+H%62mvoY24b)%OykuJCLtL$nfsgt@IOdK-<{s2KC0qD;lWx9(<9efE$Ce4V zQg#A*-oImeaGB@~+!;xg7Om+QscDAPb3`lt9is;=rNctS0!R=(+t>w&r!EY_+AQ=s4Bla)tofD1F0+my6F4RA72n1 z6L&2I@=nB|sJ3e?cAlnS*ycm(fpND+g~Q)Rx3M%0OpJ`UNBQeD1+}aQ$}o+(>HE-l zIFZNjK}}2*TKcyOHLUTbme5*VHE?PbJHDbjojeuHml@%zQ01@teww}N^Rs>c>3oii zRovJ8y138kU!uC{BM&x7s_B3Vw{Xw=O4-9E{?}JN^P<~z|&{c z{^1|;7}&!VhdTHD2QMm`;<$lt{E*dKJjdgLwzb7u#WM*!dR6>Q>)c*;D@HQvcuwG! zG+SuZkD!-S2v96gS*R~9v1auA4SYW$W)7Y;9IkM75t1;SeATYiU@z|D2o#Dtl|^{l zf0)xt96ua2##V7aZvj~}AaZ_^W=%a@+uG?T%T$N_XjaT?QqGO+{2#%|qy$I)ugED_ zLuaVTbcFc&r_r|*k8GUe>BBY1;K@@|%?#E_#;?K9C3u=uvge$Ds-W|7e(OEx1 zP?(CQN_wbD3l_adF_WYBb-q;Omy#_&LWr)j`X^988@tt2%`h%NQ&UUR9VwD{?Ab(; zQ*QHc09KIce>cId`U%fD8PcSc#Hv7hTFid-W6a{lc4&$MZlR8Z=l)9^+0FnQ+p@8n zj(fo5xsamd@JE;yMWvEfLnUI#vdepGw5;g~L+XN-b&8a1k!{!P1CY0!Vvw#&6}4Oa z%hY-WTZ5AU@NFq&2IMPt$nHfdRvut2m8$;c2XeGVzQm$|r#$^bpYUCex3`=3}O z)IZ$^WqwSfZ04BgTh#|u)NYZ!_#WyWxYQZs1nNH-0sZ zAHUpQi4GB-Pi>vDMRKi%{z=nFKg93Ax6TgDSAm+AxO;yVi9D7AqCiO4f!IC>UJIxBAqGELo^WCrEw}KV{o4KE`O+= z>hry-9fV_c1tG_B&~?>+VLE?OSd<~F0r=Tg|5`P|95txTDIg@d@TBiL6nPMJAKJyR z;y$Rer|kGPXs+$Q>N~O1J8W0nattl~IQeG?)oI+1#x)97|BP5JLcMS~OH|{!v;UQ& zlhe#s)2q-{e7C~dICWcw#}nQq6Mp2VYdK^UH%jd+9cj21{wzmpb^9IXIY*SCKjSwf03isbOHa8)K}EJdZt^1 zV^!ZTbLyVp4sa%e&8+lqzo?&zQZBmugTF+0%tZ|sI}@(Yn)C=?2MZqrT zuv4M-RnKMrY!nPQjz^vMHtMK}(*9+R_%3H^WfwugdycCvUqvSlsf=jwz;miO#t<{k z@k)DZQAkUy8BdX?7*C4vVacvJJeZBwNR46XuM{<66CTPe3Nk`_UW6;AiX^B;2&UB# z6yn=Dq!teDrb2A?mfn60IRcK8VFg0;6HhIqpzyKSGHt$Wxr7#o6JqNJ7qRL*obw#B z$(60xuUXf%^SyAQR^tGCEFGf5|Ml0{klZoHTGXII!J7Gm^mslM-mtB={<>bA4M1Mp zgsK<+Hw1e@s*D=#R#D3kf7RC&2BJVY7_t+sam&yC?|K?{%L48oBEt`8!*8&BKUbKL z5$OzQr!}LKxq;T2FV^KObJMkc0CLkk!c(lKm{z3D|p09Bcz6WqhhajV+}F>ZiXmFIqv^65*~T8)f{b7+zJ>M#BMyux*$ zlV8g+mxyH9d^qZ!muN%XXZlsOYnoTh)==+Ko;G|pAEF?VksUyzj{S!QR$hH)+tyCs zC=(kz&f7s?PZB-<_c@|`#rgl(`}XiSsK+I zdx>o+h5TYO*_|XqH?v`9woQ=Yf}kLXh+K*y<^xpJ?*l|cyrAm^MLw_isrY$|H$?G* z7nGawd*1V&GjnEkc6V|?{DXa-rn@ueyyv~2_q^w3@~o9SaV<||dUZMoI>QVuD8W0pym*9%B?i+-KYnV4tD^i5o98XH zqie411?}*fD=fnOdWZ(?_?jzFXjJm|b43oa%!LypJQaz8&!E_}rfX7g3de*k)+lXL zo@#ejMOoIDx;`%-ChRPxLupryigVfC_i5e4_v%5+eJ|&Y3yn|1YZkM0j5;v3kZ>d0mtg3SbzmGrkv&D>7O!z z4JMn4!`_`!rmamq9}5xvNT<`688wWr3adiKhXL03Y|Jpb_(&l$4x+yCO87MjT4P zeGy6n;cqBJFkeOqzkC%1opxrQZqg9)>ME)$9wqm|p(-{&_7fF=K>*rU;0oF`r7KKpeH3_Qu=@G~&hu;1BweMx4NNh$r=mZ{C_~VPCG=%2;J9~;lWKb`tfn|9LgiI?zE53S(pB{cv{3V?jwx%`;^k){aBQo|Sp0mLx=k7~7 z%oyPq@eAn_NQWP)4l#y!Nl_h%F!_{swrqmsaUNeH>%@#AerJwU1e*AOI--B=1WTon z9OCtoV+!k-U_%T;_LrE-Y{RjX`9&;%J!<{*U6lP=KHx64c80Uw=3dfShx3dv6?~K% zyCb7T?4GZ~(M|ANa?k-atR<)2@VPg@TwDE2c zxDugpmDAAvIFw}{*Qw45$cAOn^6j6obNjY=w_h5_XG*s3EQeqm;N9g^JWNWk z9D^su<6W?sY@cSP(zv`BcWL~Me3G)7y+19jjpDl26XH>*S8wKv1w60D6!iff4xDM> zos#g@9$XH9E{iXvr6pu3eurU9OWHXt4Qf{$wRQxz%%q`_erefJ3ux&)CUzv``gS&8 z@l4L1$l1`3m&HIVXQuQK7^+h6P^^_6ysm2CsRKUy?h*uBrfOi>^b(qc#e7dHjfERQ$L@UzK~*TQc_mZM06J@RtLZ}|W*hXfEB zkn6iVfc}c~u~msqL#x^LK4~h09?)1q8V~>Oz zanU6}@C~f8TGCGBpi`VkaNl|kY{8Ozsz)xde0fyl3!Qkwmp;$(!+k2!^e}EVn4v=d zM3z(FI?gS}d0BI8MOp_2*;)lQf9leW3UoTam%r&xkz5N0ez1mpa5=W zbfhDPc3hlp4~}XiEa4wJ5MHn)D68?%v@MxY+$NY^5rz7~@5$q?kX|$@GYUOLAGtLL z@XiTbqK5~jRFk)ypsaWNffd*onlz{%*@uM;tH-d|)oRBdjSpa@1UbF}C!VIvP?z_g3=)KeV@(3q zTUSc`T*b30409AIKoR2y5&!5)YIq4;$ClelZp~QnujNzS9_G1cXG3k_F`mH z5N~I<4rYNaqlq05M>Xi#EOgOa7w*%*oQBwkg}-s%1G^ffA7>nJ?}LoV$tp$?^oZ{+ ziM|3Y_98-pBgbDQl)?|W;b5z5v6N@B&UCQLLNhaBSm^=&kksDJq{deDKZXsWB=Ex< z6z}p8S5D6Cpf!;;Y^ZHM4%L2Ko7}~w_>A+#wpE(o1C{Zfv9XK3(ncd<1nwGuJ&Y%e zm?2^zUdn0c7EI6^oL1bVrPF#!ESF`s^U`8jL7PB!JrY9X?nD&XG1gK)j7ZerTn_7p znR4^P!~B(=eMM9{dUeON3GdY)n#Hi1NHj{Q(uE4+O_#)`*K7=+4<`<$eh z9bt*t!Ji(7l3q>{b!RZ;BGVBi@nZspH+=v6mDqW>t9a$2_Z%lro8y~_EE|0xkm(ai z*>M{54i)9bMg@8ns;_2ZwDj}%j6R><%T___B2ja*C7I?=S)US}l-gtb}=|unUWMo8eC+Zak1EMPG z_1wkmoWLU@n0z6&C6mC-#_DZ49&Ox%lt38eSzkDeWaHBkF$aFi2^f-FRK|;z zC~~MuQPWfa7oppHErA#s}lF|U~)@md&~4JG~RRqGNeZz z(Q;{VHo?GY3Uux+`_nZSdQMUB`TZq#KS^%1`Az_ef{Ldc^Fe2nxV##HVpUIu@V ztI1bgvJnTD8M<@sGX66s_Q~`xb>{05e6kbF{lDYIG%OkV8rA7YR;i87caaN$>SGez z{AWSKv8M^2=YH z5CvyH2{Q}EX1QvfAMohof?RIdiHMFzkuA1e>7f&GoykGM%d4VPpi6;olAB`^1t9fc zm&5UiaW?+1B|XUO13Kc>hYSHbcw!(iaA)rlSUS`tIN-sPkpVw1CV<`f)|+PA^5M;m z7#EBX1TP()a$O1xadMc2Gp_=%Zrjd+dJ8J94{=ZJ-o|1t#j3^|run(8>MGYur9*ou z2BcXj2rnh$MuT)Zx6p?itM;2ld8GX%d7*{{Ue>T6ss zms=HuH3`7hPm%TT!wnEGOLW*cYO|oLmv!Rjewz0_ zJ^@PA??-+N1tQg-%_KLQmJN-{wAk9POUsVxwutfATUrt6L(r*vTvIe?0Wu4tRO;{7 zt1MRJzu2f1$6WX+7gEZJ0b0#l7U)z{saX`q+3wl81qIdUakltmj#C;T;8m>+gO#*j z8~Y$0gRODrFz~6|o6a5Kc{FusZJ?-88@-s~llA=m6;c6j>I2`4=)v1V7t0*WhJc=G zF2!d)hZiU-1infw)&~*QhMeP?XA{s;8F!!9_*i#$4lCUkUuhMilU1twwECEtWz$rQrqIy+N3ctre~Xy!%6sRTNLVt zZl<~`sH*(FPbvJKhs(nNhqYt0#7?1RIV0Lf5Ht2e91}DUv&Sc(nSF(o#lS9Ti$a6< zF}o*%SXB5Y^(g+DRAi<^-K}`g6l1aa$ujj>v18kd+D{HAd(C7w9wWtoz?H#3q;f;; z$bK&w^cqz`&%y~Em>YVY02TAJ*AxZ&6j(6kx=_CrNM(;MQJKGjkn^)TIqZ}>ts{`i z9$g?G?Zl9(B$aN#J|*$Hg;F6G>b)IBvh42d?jb2AUQQbCr;n9b(;4+kWi&a?0z47k z)GF#d1R<8r8~}co$e1xY=3Jt|*x;CoAHYawG51E3Ky$XU(K?NqZNbCkATr6gH44LJ}VD#E844bFk=@hXEh#jIy)j zE+J3ER_@pOZFe*|?ix9 zPr_d#?7O4-9;9ILI+TBB^B~TDx3vNuHuXfLF8^jp&>P&I0xP2gt#LT~)i}Zp;3I#3 zdzDi7Eyr!21u$?SX746swT!bf?Icw`R7Fz-LYJrMw+D9){sZfxj!Ga_uhXYc{i#dF z)hxFY1x_c2g&avTme(h^EP~4*ktpo?!gC}tPcct>$09G_X&dkvhGsYQ zx3;$T@=?(Tl@;c z_3TtnW?VDqz$t+=-lL?}<&PkqFi)_^5zyQ~?pz-NscjiE1YT(S_z8jRl zD$alY@~P6L^o~NzjcgQM{m_R?xHnm8P$*{_rriS4?SYkBuSDFb8=~%dg{1@Sf=Eo_ zG}leb&c3-qbNy{yj#~fYHRu`sD~wu8wVIWJq}B&dt*q8k<-%(!X;52NWwmaun~~M} zhze?53FPWiLm*X2MrdxnRWKp&ss#0Or!uv6ByOO{&~ETkL9r^4jiRfc4yD#@U9(o} zwyuiAEqkq`*2-#rV0dRcoSl$fEOnm;@_6Y$r+m`*#=64A!ebuY zjm_Q>#vY_veM;hJ*Ui7ZyRtn4^Bt`fiF?&X)Ia`FRu8OW z!Cq%O7@MRQOWo&zJTAor7SZ%)@T~Ai52mTVS60!b;&CN$1`iM?0q%Y^Myyj%O<&H3 zTK1)s&Pi}Sxrqs{;}k|$7bFl3PA)8UpJTVSt}g#B$&A;PCJXyq8a;gl;SE)hC6oxB zq~e!vio#@FftP3AqNXGjZYx*!5Eaf)%okKhMPJeDkz1H5NJF8$Zbl})t2FWbtsR{t z)9+$iiRUZHR1_al3a@iz+I||zRHZNHLxO!N$ql#JxA&Ix6-^VZ9i76rqN=KOi3kaY zF?&%-@N&gxdy(D2k+7#AVqC9&fTFaLMu4EVXPqGV`O{opqe)>Pxb;66j z1(-9o(JoD*k6O1C4SEeN>ML@_Whyo*>PAwTSh^Wl2LZ8k z3(q;n(ydU7od(3xZ4l?*+CgRCinXImL*LqhSwQMMI@?)8jb7(`cqqg%iP+NGg~tl= zPb{4gh>sZ@@ldDjnCsSih{kh?^k$jy#1qZfa{_HRsB?-p*$}}s99&Sp> zTz0<>@Ud-ZP7@9kS_%-ObTTwPgs&YGeb82Ib7vXYt4$0hn<`TG5F*~XRN$Iz5*3c- zQj;!0y_QiPmaE0~?!-eCM#_9cDVKpy8@wn34j+X*;u}gs*+WoD5 zT*kYIF;3Hr`*$E2>NDg81sN6UfC8s_nO3RB1^t!wTX;3z!BIUZ+m>f{#5=ZQhG6Ny zDXrYsqf7UVl>1zV&X_dG+!d+lyI+rb7bHNr6evS}T8kjqD923ipqU*2XfJ08#j(Q zA7?b;qgW`9qqm5vo^%6)L3m^*O8cw;bH%d6?GRv^72q&Ru(E#_+HZJPhG;W|sB`94 zERM@3mXbBsT!^>`c2}-hzY7IWxg16`&S%yl5pnz=>^qZ9OwLUtE|_K|eXGNjGeYTUGq)s8+^v6X%Ui5zVo`71N?NqPuRH=HM z?QrEDOu0xY()RlMIKei=1kid*EyCQ=b%4)JrOQr=_ysMBslY9<6`5p%o1-mj(l2`7*)EqrmE5hR@gUyf7yTIvUv%MR`=1?< z1DbzhM5UYU_X51(|Dj3FVcYbJ$m-|LK?Q9dbYm>%M1&VX^zct)p?A=Yo1L8i@I^)L z>^I12`R{>~6voe~Tln{&;$4I2kqp!B0_?8pr*)cXIELxqurb;+lF5$4;z%{`JEb=} zGC31B$Fjw&fhD?=$y`=TPT^&S#r4n1WK(RB@*?udyj#XbH=8zmVlQTFf64P!4nFuw zRfaxAjyY!Pqhg1uHf%xlkwb2$Bp_Y3Hww!g0H%gJh7w;Qz&rN3`{R~l-4Z*GsGCEG zSpTbgk!=o0MhQZypSg|rC$?0I4ds-gxs!8Yr9pct#oMl-DBa3Ur(3haiDg6~UC2!@ z&K>eJi;bCBKA!ZI6wk4L3nki!xf2tR( zbK@u!fPitD#x|EZbcpBfvWwmh&u1jLrcs7rh%au$h(}$>7tREn~7iF*ws_Dx{sHHEbPePo3 zn*x=-Z8U6LQ7Mk0z-=XbSDTv$5roRPy)udYYP-nOd&|H(hog}2sxJnC1AK}Cvqg}S zI-8~P8ZKhbV8uN;{p(P9KW2yYt2NyPWfooE+St&;eBk)$aLF@K`n zX5yoBCAyVnw+F!Psd?`hGPk6?IpuIYlJjwsVZvA9$`$6m;Z6ft{h<=9cFWZ&z<00398U#IJ0rzC zgIjIVh(hru4r6pYlWIziWsIbbv8ce*LDX@5yy!N~hTZ2Prdz47@OeQ6<4R!(7kSzk zAGQD|y24!aZvqOhtI{=#D|3{2oD6m%@D zYZ$zXe9CyUVeW@ox==Hlg*Yo!5>o;DwsWz!ZKFvaz}k6V2dRR>IzRTZerA-yDrdPj zd+d)Wn{oMTJDW(N_r43^6S#&36FY1PO|7>_uEMajqj-sqBtWEX(p#&>6}dI0b+oiL zrHoPCvSD8uef<)zT6&8-5>Y@*ERF$Onns(Rgv7yHaLvdj4>4k$wi=7axCrNQ#l%0}M3FY~d#&vyobVb9Po$%13Txl z#IVIZm?RkD>rDCVF;+&iY&~mO`|$iRsEUv=0>ClO{(UMoKA~qV&1C14$>o*~bhe}6 zwY9S2$o$5h@GFC@jdq(U&P7$p|CRGOnbs^z2m0hh+DIBUzq2={P2g`03qe&6 zJ06Cw6M)0$XG$Nz)2(Q?ef9WP(AbfbT9O&<(q|`)@=cMLU1q;&NE@?s8yX3gb0Y<; zVh{X;`U$zKnARQI0YaruNpy1}H_2??<+6FR%h+1h9#{PMc^i<5({kxFS%TO-?Ntw_ zshcLSv+}+)0&iWv<~&@Sy@(Xt=P`n7)GCRixQ#QW!A+lKeP@`~cArkdGfhP>^-T7Xs$M1=%(gScc7>t zKd$4sv~7sW@rL~MR9Us$jMj(tV^@UT4nG?!sZDU>@@tp}+S{}8!B`Ae#&(_Gr8@m7 zYReBq6nZu*pTETFTe@=#YP=zT1J$1+t`e;E7E1#0<2xbVFmxX|HPmDwGIscA2{y>5 zJ-blQC*mh86gQna;6VAgC8njFGyrd4jl0}}KPfutS-Jfvpg+u;ti*63G@OK8Y?ih* zI;y9Hy0qmmgyq`pGe>X9MkT+6J8dhO)$HULz#lG0FM+;<0F=qZNysC7fvN!@Uj zpV(o;rYJyQpeltVr)||{iZ>MgKxTtH6WEmV9`gCpT!)b+{z2^(Q1%`u?SrJt zF9Z;*>c}==u7`DLIAOV7-KCAv7q>g(H@Pv%i8FoIt4@C=9;pZMvO`CC}8TTB{*hc)NW_LY-?lLPZh!drfqj?FEQK(mMr;^9)*UeNW zyLL2_*38kSOm=kb%=sy$3oDfnZg7hyOj6(gaDE2fe9dXWxnO*B1Q1@II zt#IW2?4(*|?Kwl$&EM(Cl%+`<2ad_3IxHrsie#z3kgDWwpb{}kN3*HINNkUoqdc2W zQ93D6KZI|;3C)45pTBhS)1~L7M4o&Q{(@*GZY6K%MRQhgx|GC|Jb-m)3Td8!c0E+% zBq+qTL45zYsIcZ%!^~N&K&R~c=$lp3lleKZfbZTNguh*c^U8B(fdi*Yfxi{8faaH8 zAQm`Hzwx5Eu)v+Dd-`IG25y_97PxS_ltM8JXkPJ!VS(H9g%{6-1s#0%;UGNsX4HbY zvcQGYrR4u!qy;pe{KBxnZTg}&BMbNuAGoOCxpRZI?~w~}P0R^f*^EioAUt`A)ccq6 z2}QNK+{oVb0JK(MUW6Albmb+COiFtrnHw&}An5@LlEje?_yU(s^wmx8l7`|(vTJv6 z+(HTSfYpZ{XXX=|Un{|CT6Jd#6nzZhz!V;sw-v=q;EQ{V<^^C@EKAI-(3Es5mZoZG9_x@JV31 zJaNB#BX0ME*!j>;Uu>9UUmQZq0Uv)$!0veu)`pqPjN^1Q9#P4z+pC*7jGPsljH=X* zr~aSc!g6;)%;hk1V_D?2NG;^btfnfS!SxU_kHFti!m8!iil`6ewnrl;V( z%V))8UQ5RP?1^;Fa@jqO0m~jg(__0@Q{f5Wt?anGA|)C-q=J{f;BBF4@zj&lrYtc? z1DS3G9r8e5`8IS`t)WZ_4X9|!yn?mtOloN@3%jleO+2rmy834UeE$l0g<7!2^NP!_ zsN{-949%v(6;F~wAgUWC4DtuyA=yyPVNiq5x-!WVC zJK#ZvewSE_8;m{nD%`nZN0DFR#T}QolJ(o$tq-x}%P)t*owL?&>pFV?y#z z6ekO%lMARyVm%+#i~zgGiETZrr8_rsc+FXSSsLMsRc3PS3$K>CJNClQls_T+@*IKX zob!5$ZPk~OD)ENtCke^W$;Fbeo|lm-Zt8XK#334J=DEQoM|-zHoZkRRd4f>QmQX*6MO9>nMV!NUf5%o2%D0vADjM z&-A;}_f&c7HM0Ei)y|#U*-97-*1Zc2C#XfYOsz7_cOt}gYH6D<>|t}55M#FvR3lx! z=H2>F%k5iZT5CtASZMckY<{Sk#x{arZ;?qwT~F_0V*tN?7maa3$zh6x_ zeIUz8;!1s7iG;~xcKsvQ7*`1y&I<9`{HgN-IPKV!Ki~fO6 zLkKQas??|?KEo9SfdQMz0)9_O{M=r)zlIdHBCMOQjXG|@ZFfh{ww?NUIUUbKi8mDZ zB-L-ObC;rQJA~3g(F80-DQa7WSASh#i+WnKjs1F$VcCW`nlr31J%zhOcJ}T<6pka_ zBEQ6NaqQYzp-`@Y%dV5`Xws1xwhyg0pdIZch8K&7m3b8n(Eq9w&6uhCKy+RI)#t*>9i}hUoWBJFDwf4K2MbH$JRq;r0MoK&HP4ok_AOGd`X%8N&DRNBF~n2P0c= zKwkDv+F6Yr&Cnyq$b3xczMhTsiv%6440501T2BGTIqdoOK1VtZY8-^gOYVnBa6Bp*|Nc+z%s+QgRU>sx&| z@*-_F79p4Q?6toeMKR5f1YVIdjq`GPQ!5!Yj7m^&Znha!r)nX}eJ&kH-=8-++LrE4iz#O}VwBP6OR zj;H$-p_bUs=9ja&3GrLs%Peo9Nq_o1(O`8#Oy|POh?002b$~I$>ZR_D2Vp_D&$XGy zZ^GEErE}1lGLvgHJ2P%1DJQRIU9_FmSV2v5^w`2N_23{(-5g~brmvJdcWvd%{OKf# zFbz~Ig+V@m+Y9jE&1g__KDr`QgEu3S0#BN#Tpb$T7X_%BzCvOLhO3T)ngw&=Bk#kR zyidp;eVfNPVk_pa3Et+)l1=23zj*yqg|bfZ#unwuh0E?jbEzn&ezo3^T-x&^Is zpFRb(Puz-M_iH!^i~rw#mS73en>izM^?yc#-ONI^G6>a5{tcYwy#RX|RUzHa!XlEn zg}bGN1Pkr|iSBAh_TB)B;~Lz?gxbMbElZ$2!Tp=-L;^~F`Bt>Lxpy_FW9n-Hoi4tb zgK?}61Sa%X4MKGxxm}|DefpGIhU39VGs|+TF*JGHi`{bD2POX&@?)`k$0lmNiy743P3rB-8AN6@YI#*T?w9(xY zIJ#VVr(}9dpAcFLU?-MERjNAIefc&kX(C3lufvsJ7fQ3^3)N%*8%MW8$G!nmPC3XMr6 zAcC$4b`ZY&QIEKs==jN6770{e{bf{&rGkCLVZ8Xx62DgtU$5>IjA zAX9+h7?IK|ZTFzOSbAk`+Y$hgUg0y+LCgr*bmU{RPNq~Mhow&TaC-IvT*i=}hq8qu zWe`5|u`=2g4hVdLsoTt3C1XT&n_YsSGKWy0^+O}1IylW5oS6?2ENV!I7SInnt1pz$ z1cZc+POE;JtA1(x!-Y2f{uSqL=72g$*l99A2+ljkSJ8ysyd;x-%N@W2OpFm?gA8N0~FR;>A{-Jwu?VAGw*TT?UOVgj>o$HtZY#l3Z z9Cz+oDy|XxMAVtR?bFOuS|4x=Au2t(tr=-Fx9JpVQj8^s0PZKu_?&Io%4dB7gQFn8 zaO{1!=VF>Clmm;zm7$OGrVVR%r#K_%{EVsZ?u2T#Izr}>PvK6b zZ*wJ>H@LVm^ts9kGw)^|{Uj6pAlo0-*$%sZxFl#Tzbib{Jb30_r1dVcc1bt;v-*fJ zDX-BtCV-WO7@ox3pUzp^O&#K>Vjhj;H6Q*I%7S?vUyz8a1&Y==%u%^q|K|(Y>FG0v+-@@8;bpH^(P9w3hI$t5`T79~ObV zY0Y6@hyq`|z*fz@h^W1%5O}^&CBkrRPA%`oa{#y;1EeY8&4A8%1hSf8wWYU_!snG%}V|m9<`(Vgd@&4>qKsQ zfcM-`i>4q3ERARqC7kbzNY4H*&N?|#EYcUA?KlEV?B#sl3$XRe$mDq^TwWYfe()uX z+)JbkUqoGasne{J6n=^XR=G{Os;Fuwa#@TZOvUH zZK-u_>pIqr=*`I%J=xsWp6p2J?WeTst*3Ohwyhf;gx`JzqjUZ#znWwIjY&h(y#U)D zoP`!GAjzq(a0OT}m~c0e;vLyzQg7iJxbCa6m=+GMba&RA{OQ7Ba@1F&ph)9oU#jJN zmMVIdvZgD{jFujCJD$_n!G{xecD=f%SzYb&eU4qzdUa1rKr}WxN4s9#)2fPH4U~KF z>@}DpZiDO9J#DHx)Mt?)yRSHKq4n%=M&Rx0YMg|^wBm$~duJ2w^s{I0Y~g>aX{(pp{d5*FA)$qJ3?eZ)Ca7up`kf(-NNb*yy6?Oqk5rQkvbB0bnrqo zcF)(v5O0;b07I!42+dXUC=fHgHOvBO0Qm2X_=VqyfcUzu5fcyIvs6FmdoMZ?d%{u8Vw>NU< zX=u(Hv(_G06jXF@5DE|DdK8`-aQ%xkI$glV4n>sdUQgcd-QPhsU4Ez72fmFxpPTo4 z<~zt4*FC*3xGsITToNyco3mqlL@Cs84dBzMd@e4SGK?k{pw;8QgRy$cEGMt$P3<5Y z``svRJ}X>K5hvR=JljPx3xGzok{@u#-S7M^YSui)-3ud?LAdmPQ7k3K&BACz=k3x(Sczc=gj*Sr~mxvauDu1!74$e)KzEf6yX;gRo`F{%Y> z1tyqPf!C^B1+3f!V`g3_y{#B~2MKEs{@)L905ZD;PeG7o)qf2F>Un*Fo1DVNxM6Wv z82>T|7zs$@*MD{M*r(O!4H8~(W--XSlIl5`9w$O~W%5EzSP#N&4`4*j$K%IXnXKJ} zfAmzJVcKwbHTBK_(14HeN^+eoyP{(*5`U*fn6T3K_CJ0U^;|bZt7oLMO>yOqI9sAH zi3=iw&z5i#U;_fk>wgsGm`-N0)|fG3!{TbpfUu9UnxW#n`tKhhf&3GAnE|pnWm@=e zT(>N3RA>7(U~$!vg;R;Xu=O^j@vgVURY$6?on6Hp?1Y-dRY&G$S<%wzYQ@Eo>Y2kx zSW4h&^8r+vx@GGrTCMelpCFr_BM5;P2~+@}DsV3;J|2IG;stiCK<&r3uro=5PU^1! z@~EGn9}8J=j1HAEev0{}J%mYEI(XfvQpoYah>-$V*N$q5tAn75zMq%VExc03f<^Q| zGpz%7+1?TSXU-nWWDVQUEvV*?QYizgXlVPYSv!rM$gE^^Gm(LYBuJD;yJ3yCSO+pnnq%he2*L$Z_c zAjYK~M+L_F63MKOzY2MQ7j1~!;SDpTPeL;t|CZG)B7e7*Hd4?N|xj^D74(Bl* zy90f{{1M9CdC1SVJ7ab_%1L_l&%*2I4&BP+vPqo~hbyaGW!j1lcfrv3A^Hqc; z`PGZ7j%D1~wQCDpQf0QFd2+NbBMS{}TRtQbKM#*TLozR6%`_V>Wmt1`o{#W#gdsJ9 z!GvF6c-q<_j&?*4&-?ZpJs%0}3trEi6;tuTP z+=1Cz))qbs57Ng&`HYMS&9^){8}BYqc>GsUI&738Nk&~uC){k+z_P+>3brulK=&0o z0P3Ue^6D9kefTGIWr3Obs8l;_SFpW*mm{&b>j1_1r0_q#o(+>^XuZE?rw@j7Jk+2R zm**dI&#~+(bLT*4LfMI{7#;FYj71$xPC%}+8jpqzZ|@vp(_^y;4rK(`RJLQ7jn zi>u57*#C2i!UJL67oG3yt2&-G`b!CYcKmKgQL{c&jGDM0SI{D%#F$ps`R* zK>Ky32A1VHG>qP63yr%H0q?WpHLz@k!$a?a@Jf!svYDK%H|?CVSjcdAb(Ok^yj_JF zJ2r0=H3}c18l}~N8rfo)``A8%YaYk=?+9|*C2hf1saeUN#`9vESro?)6J}~bJ8fbN zdc1-~_~5?!@hCKu(gG^Wz`lLU5?Bh?kkCu_E>expK3>kqsNiN(w>ug9_xuj0j;cd3 z#jO=r5}%+dvCaLw>x-R}G4;(o&i&2hiF09X|6Q5RjnEx7Of5UL(HQMDQ-)>&D8Mhl zt;7+3%?E!Mjydg&IUoK#YRqDyw1UPUK~!szlHgSsb1Fn8M4?+d{rj?g0TpN1#Tyd8 zr42$)JnnV?*ZvVB@!-TvB~$6qNASvc2K{4yK-$A8hFnioW`t^JBf?sCXa_2G8AO>3 zY79|eYqbs2UL$TtU8T~`^7@4|tWouBHk0)%0>Cj?4Vo&uxVi?pRa5xxQNY&(%(J1N z|Dm#aA!bZs^)t7S_>|QHT_rnzVhgME>>s6WWNElMs2cf59lV2_B?~62E9o*>%@)J9 z)4F4FiDB}C;Y`j`x$B9F&TcDZ@yOrh-x5^&k~p1M5Ys2^?IVtM4`)(S_>-R;&KBuG?XEwU4F^fC68KGs71Q?en$5dF*&fPQ z;9QcdA+d(|Tl4U5qud(q|8vx12<1V}8;5WGsj|d4-Z1k7Zisbkr95Vv@i<@HV=QVv z{1@C^_zzP5iawlY%0=ZdrEot_`|Q?P5>0qC7KLzupO)4}E#B|(C?Sr11}I9P9kVP- zu%P+-zeoSf6{wGjWwis)k4~wgaTSr*{Flm%bSj}25v2X^U(vr;N*Zy6OCVBb zJsBdZ5WH}2vqRGpGd}1cN?rcHbbwvn)Yb}*p1K6$b<a5#YBWc0Dr*OaMtQ{Ryu2rVROc3*Ie3Y|w*O}bORqFK}7a`D5 zo(iFHJEts&HvkV~Pkh=9V;7gnV^kP7Ay67Ng$yIG!rz|axhli72P^b3Qm1|0z8Y9I z&AU^S`B5P{wt(;|hke}8k0~UnWENE49S&9DFa>@>m6vTeIXR}8?uuH=Fh|q+NZJ@3 zvoR%tH-&%ao*H=BPy-q1_0PzQ`z*oLN%YT!ysv zEeP1=XHdDFIdxBw)3ce}1gv}(JA4tJl`C1Rf z0=C!)YKh5s%ezW`CY778#3T}m*n+MTo-0Eqga3hNnHggF4JVgZig-hT8{(Uua}Dv- ze_*_l)=lmRVo7a+$2zbq3|P{&<*u$#kW0fjT9}}|@t_o)s#jClnr|F=lqV)FRpsN) zMG2Mi^f5%HB}A0fVem(0NV&gYo!UV+@S`c!4{@;{W9oL3mCQ`&IMY8=MO3k;<0;eH zrCXlG8uVhqPEPE!H9KcvB~`98ma*(D>;}N?S-gu4vt}QZfKnK`xO#En1N4;T1`!Eg zO_8$u>L{x938dDK{Ugd$n^odAfy+tOj$qcYh6*v=U$+F7Bh=0Pn1c8*X)jKEiAcd{ z#aX^!mBABtu;%1HqXNoS&Dsa?2Dq5om?b1u`?CmbV34rFzz~)B5wgCi=UrvKXwlIn zV;DSnq-itNZd8Uo!u<;7x9b>yviVCi;I3KY&L3 zze;VCagQ_t;XvFhCMaxOrKUcwt!t8^B2x zVv>*xfgi&qi(xMWb~o!?vOz#X0^|yWFWkSUs?O;?-P1FY4ZQE~c^~wLJ>7jyojO%@ z>QvRKbI!>$@Q&Z1x>DPQWxqqsA^vEKS*K-qoBH7)Xo9x6qF5zIit_t%LQB7&d7Ubb z=KxOq7OJXoyB0aE-aI2%x1S=5grAkwo_1G&5b$wME(E|BT653h^i=RUmfN_cSGNEGa8iKfsvDdlUyI~gbMkz=(Ad%n40!l;{9~!kxuRwU1GzWZ$^zWc!WQ`o%U&hPSsjd2$ zk%tBxEdomc;(I(b8~pyjvT3XeHH~fex!==B%rph1W_Mxm-TN; zAoJquvrz5q=MnWB^>`5r5cODo>C#}m=yEXD25m>OfX5v>(aa86^PYyLpcND~*wEMn zZ52}y-ktsYC8=e8!g7ZKFq&{uS-KnVs2zQpY#UC@0p61BzhT{YZa>xoP63VD8!Uz-?ARyUA~VY-VUv z-}(!B)`GRD3EDExGuA}p;M=GM-xXxWvE~A+0D2&qY`?Mrndv%N7rNTzZ~C;EJ~1Kz z#4{I33J9%x17wd#}yQDfuQDrUm&9>JXXwm;X$BYgmrOXNbE zxv0f?pfk~l6{O`O_E!3Ux|SY-E_%@o?cY45iMDc}LW1Sk$;IKrdB>q5(6n%bU3`vdChGK|~_OtSg8e$JEJyAof3V7R6 zVai~WY|50MVa>#pr1wXcN=nk>F{ENz7z)h!oP}l;S~sVihU~wVF(&p0-wr3V(HZWf^s~&2 z6sJ1M(=W4*>EY$0&~n{rI6`Ss3hG^anzXj?Wgi!|F*~Os?4{+)g4u5RFGz@EXFG{d z0HZ_8+$WTvWlSLD`I5Sn%MPN@9y+I;s-B7EDq#NQG=&ss7{l9^;n@5ewg>%oQKNKe zaG_|Dc?Qfg6fv@qV9WOnMyvvr}mZb zKa3l4Oxl{Oko+}m&A4qN6LzONsTYmV?uRRYm5`u3~=y8O;GPtr$1?f9epRHYvFgN ztE%=-qIuGyc{mYdWs7;}|Km3ApM>S4h2>kz7ZST*pPv2u^Z_;PKjg+A_;+0Kz_0}E zixo>;Tn*C?N0+kiPEmc4YAC-RUMbCUb@At4euH@uZ(~bYqo@G-!G;6Rw4?F9gkx7^ zyo(Qq5?UrHTs|}u(UAjoaHMrEe8}F@-|OYlGBw!xIrwc8LA*J6_zxpqWen)-Qe5339bA+raQ&C z3%y#f_6*nCY}-@V2nI@yCT7p{{qdlIJCIg1Olryz zDHlP9^FvSPraRAtC|7<*Kfx$8f9EEUUEx-IsN#7lPgcGbi zs|ni58(Hs>?8+}mE%Pj6S-s>V%e0(@@FX7CNqC;fDE_n{a2#ef`~=&b{RoezAj2b4 zDy@~DlkD9-JWLzx;O>2Xc^JQ~gE0~kaWe_@*GXOVDpLyHiY z9uc%g(dnRFIKnb<43}1sgjxCOmndVCXb>r&Y<4y+)4SZt8UP>Lk2*9v2`$8|R9<(z zK0L+3$U!^X!NGb#*PSo%XU_=mpPkbb9|1*1`4I37DA8-cSl+i09{yNJP%i?>e2Iut zh2F!{xC9nGQ2VUf>>Of-g}alEU+Nr!^1foEdrz(X9IrQnODfU zEZ`9`y)FG5Pl{)ZoY1HYHl-l4(vAGsOL0nxcOxlpkmo&09IQOgN`)ndKb&r8$F-$W#E@*^8<9(s^c z_YHPxgkI59!LzRXxrXJ45{v7!&&DFQKmAHnE@-8aZHl~4a~dd5Zv+^g{d;s}npvsg z)TEs!QwsnSVU)cC$h)x_j-z!pLEGs2MfgDZ(4ozJk{Ly*nf?k!R<~@+%<6fnt|og3 zh+lnWIQK$<`4t^c+))yev2``k-$(fIi!gP01uIIFbL;_25~WGw&RX~$;$9Z9X~I4D zmeJB{#fS=>ml5nF<2WMwC#lW4liRO5w3!pq7p_Kv^FD^&BqrUn%mv`z)+xe^p^EJ* zGtV&nF=JNF4fAe82*G&16K4V>npxbp9&`Qs^z5Vo|GFB-j`*X_hQBKpfeh{1R`QH# zfK%Uk8YIZy*kpLw$-v+6Y$i$Mt6|{9)q(}G_GPsHT45rUeM6Td^8kn2eu_ZLT}u2;d`N1Qni<=o$b((;sv%^k@3Gd=ndFfQ zdk^cN&s=?xAV#p&I|L&o2fL+>h95X9x*7UC?Rh?}^TD?8Pv z7xghS&u5ITS*sENnG{z_$wm%>8Htox+_!ckZDh&8GKweRtM&3#lmBY%HoaIhasz|G z#}s{(T80+;LWot?P%7SuQZ7(5NJFV;%00Ce4cXC0(#ns6zAhHBw+8G zaS8)?>AKj}BHo=hFt{P_No$?gjxrgZQRBk*H7dPvUHzE22=2&s(g$GcdPTD88S~Xu zzk~g;YMS8H6&7O)Tb{2)w1E5w^hR+c23b; zJSPqF#`UNJ4X-mx4apOfLCc7ZBh2OS*H>O32lC~(JH&S28yDcdt*zX15-j;;P|K|0 z88dZxQAT2M(iDb{NNp9%?WC8xa5mv>)ETrJ1r8G8feS>wI2z5kmQPM>`5Gxn*c!09yJpTvy`ce|kn1_p(J?J=tY?lXl7TzbCcyYuVU)Qg%Rk&>&Uv=%BNZ8hMmy2wXsmt%Uy4jVw56|9w5$uHN*!eLeu$Br7b}-EVr?NwJFX;fodhNUVqyP)zv~T16<7po@kuT{FGg3tj_Q^%`vn2Sz$ZgQ9~Oshkh+hV`Vm(ady71GaKi&` zb7-shn=5K|5E1gX1^Cr#^bisH5$>5#5%Md+b-1BJxMpJ3acL`DV6bJncxX4HzETag zaKUA-l0xHf)(~9(DzqUkU^9WFMSe+Y6)~Qp8$)G@$yYVON(z9Wznp$Ux-W0LMu))a z;xvxJj*&fho1}l79R(B4y*NE=7*l9Xu3vjTn#_FW0Y3NzJl*TS*Y@ghw{w%3HwLV6 zTQ2ZloF3+Ff;i~0SR@-tmP;St86uNnbt){A|C{ms;Hwrga$>CM8p1UM$-iBL!-wsB zauq2!L?(#%Yo>q?+xTjUO@F1jdXYT15CBSvV*og!s*1kpucYW<5*4xymS#})E@fRv zvgdzFY8g3Q(7t2Zcq~-3U+7UEWTZ7fWZFi>ezG$r{9E4MHkTd%ko}H+=7t&G4dGu~zlz4Jdg2T2-BD9a{R!Jf857exT<0N)6sT#S`MZ ztzT8EK@0;SdhJ>z)XtrcVayaaX(bg)GTM%;{p}4>c2(M%c*Z6x8AvXlmF#%$CAHb1 zR=Mbs5Et<15X7o&^yqxVmuR)!)(kJ26ExQd91?wikNChyQd{{gZ0>MWDr~K9LEF46 z4iVu3=wO=uHu0b0?)Zw>49*4OEYpvW8T_=`kYaM7fz*~3_96ob8#Ace$v7Lo2+fUN z9>iSI@~s*)YT7DKvB>ksFc@jg`~sn**Q-zjTP4{x^`qG?Hq4Q-h!kko%3o52l0rAN zghfJa?MHSZ{lh6Vn5)wU>)6#{#6@%d0F)q*g>7Xq_Lk#$6eTC$X;@i1XIc}uYhUPC zhC^T3XkN7ux9txw(%2>CTbEYu;7M6=v~~_{m5vrB!tjAfG}x4ykM>7qWq-SV2M6io zoLFAP%n#3rF{QgVvWLTsxLLg@T&b4!(SP9H)$nf6FQQrl);_!jWV7uGC#R*oQokw@ zo$Bs%4F|;=a+@$P`zT}W)64(dp}~2AWF^;dEV{7+?IB1Nb=Tc*J2^=``8;fR>Z(t; zxfPp)Z1C%$tuTBkGVH%uDgp$Cwo+31&-s07I6G4eH(`^z!c^eLp>J=EutmrnlL4J=jf=t<)yl$r`UN8O{v6hTjb7`}uj4Yix4# ziGpEyP~_(!eZT*z%z3qe&>;3VY-)mbDY+slL`x*;M(z{B@s~)59I6<+4duA}7)wIH zw<@0Bhvnu(xfOV`Cz^(9dH^>_XTlM$7Cg4fnJIeq>YU+r<_x#bEKV8@T(%jFhFxON z#1<(hvur+>DG9TTDt~65kjBE)BI@AqGG16EqG#SAV75?!4@$$cZ_+7ee3+me?tnmFR?xfx=rM9pY2i-_mYSjXhTKQ=fj?^T; zwOd(iSNLKpPB2BW9Sw`E3gs1BQEoXvFZ9)lR=8G?+r=TKb`EabCe#9|5Amn6U24Z6 zLw9WpOQjnbT&UErIbrEu$uTzMC-C~V$pRI`H< zOb}ts%Cq+ro_21W_S7BINSVU8*=M|YzM0%TUD|}o4Z4b zs^NG5FW!!P$=a6dIXc?Z$&sBXCQ@C;ws+??Md4A2KEv_MaWji&NMwhapl$X`Vacc} zb?j6Pe!^BKPLm*RJR8#Y@3>@G1;_VZ;D_PT*i}Z>^T~~23FxBJZ0sT@LvRcJCt~e1l4aN_;fhYz71*Re}dTrAE7Bv?mn7eO+1XpdE~d33S))? z_a^eM$1}JDE|o;jP53DK0{1rZ^LE?2#2B|71CB1h z-Uaj!r#6XwJ%fdg*e2KcUc*rrdTUA5J);E5$nqu|wR4abIO zMM*)HEym114qjNoeA*zcA)MjYhfd+wm-h`14%5jqaO4!Uk${M7S`&)+ppjJ(U;DQY zV>*Nj$Hj|{1>K}wiiYbNE<7t73*}R}%|SYwZXkzh zYTPs&7Y?7wQbu3!F3h8+ehB*&)-(#|V8eD=FIKeU>f)3+jVhAG9(VqIhn3!{;ouCM z&%GXbg!76xK0g~AD61;RD70P;wu=dCAP4Xv3&*L?*LG$r+31ia2@Ihgh4$bOVK2SW z)w*kScV12kEFcRmZqIQKu(_1(IQk60lQYci>K${EpDN0-KokT2B&YbOgVHDP$z z3rK%|SCv8L!(6{hO5wTbj#?t31i{NwLJm6w&Svtze|Hm|I<}&^G+?&nnXTpjA0tg}>X~6iGgz3~ji@nmk(d6W zU05`Ix*;dqWR~W={TO^<1Ql06fc71)#TNm}(hLI*X zBWyD8Art8DLHo+lb9f}+xR8puZWeM)Snh~Ig`BGgPAt6YO7!23@Ys+92p36fv-b;z zEBuA`Ux~>buYq=Ahu_g~-NcmvW9f?`AX;Xd(2#?Yr>}-2K}%f6KM=Zog8Y5jn7`ACyCGOJHL_=Mz=<}mG!5_u>CgLQU;Aa#ItgNR=}(0(0xFuT`4 z`eyt^xhX1eKsVIKS*Fqq32<#=lN|Bs?!_&>YSOxNRM_stnqk)E^g188Ti(1@lyEgiqP=45~um)u7SXLaI z8g30g1VDZfZ9_y_{8$IKmDVz!i|OK|X?eH~$&2yyh^Rn2Px|T%TLbVO9fyd@lPe6z z9G~HflwJ7fn8NZ9H^Sw12h*?!ugLH>Z51gPIaBux7g}CUR{W5ShXJ?qLtn|k@Fi~7 zf;0!cq*S=j`#y@hY|}AgMFc1N>_Sm@Ob&VRvPZbDVceFCk>^xLSld^Yh3|x7D`(W# z8aQt+z_cGGCxrUM>}t`{4-3&i_Rwyv2#Qs8TC)*pbH=khy*=IAbTdC@PvZ$McM6z! zJXK55H)mtuDORu}5GJzHNuWg?PQf}r*bOV6A#@n+7=WANEN? zBk$t$?hA18tBqUwfT}q{ehc=UNS6M>IF7^JJ>uvNoDjPNLDxbrSrn<=>mY-^ETnd? zht2=1Q;iE?@ZYg6PV~5@ffV=2iPL3m#Ndmn;#2HNm1L2KNyq{s>kJAu*kLi- zlI7b@mM`9rs~248B8M{-U|8|FU2Mp>wz!IIAvrk$rF z%DDn)v-faFCya8<3QMkvU>~uO{!+*I;m;X|FjL~KrMt1hoQM{MD=1nnWJ5lPrfueh zxHhUK7*KlJ(yle!*2u&^8)d zr_NmY=^O?GVrdkyG;YCO%!!3)zsHjWJ==-@z~}PFQDGuUslqS(dAF9ih~rWzyQt@Y zWA7;we~V>cys*`Qt~nmLmSs4Q;fI3mxRZL`$iZMZARNHIGT#I%g<7Pyq7Gi6#M$bz z_b~^239c&kt$dKN=O@7E-?ip)=P0X|j0 z+02A;65~CPx(eRHzkKYJx+4X&uzJ?5r*NJ~l_)1$%LiDt$+0Ro1YE(op+u!N#IF2H zF0RF%St4kcoz%0FdQX?^*Df3TCH`Wsz^+m8*@i%^PvIe$kN*v7^fcT* zgvV{Y^zAm1W2`g|UBCG%=02f?ezXB6Tu9mb3}-xV?}vLEi2HcaOx7+G`|-fG`{cWB z_*?_SEUtr!{COdKMgHtw1IOg=9{3meb3HsLe|C4lfB3)G!g2X~4LmG=UI26QXLmO| z>i=E`Pw;P8RWAk-A?_Lr8M{Vh>$0$++z4m+L8CF#(w&*kycxP$U$6!s2nfa%{Dh4P zxejh*@MBU~(az6wujyLXxotxqytBAK_QzLOT*!i+1!&oeQN(f6((~rk%5Xkmyn2dF zb)~Ty9c+Klrxv1fsbLw8nFY8Tv8v0scyM1l86;$@VO)mkavT_RkUnFj@o~d}-V6U6 zhl05AhOc`+aV%wfoA#ar#TYGR=lmqVSGOQm4z$1Ss#7r)4m#+^iDSmVf4cXC1+$?0 zy+YP0=Kl%FQ9&0dAk}>!^>kdd7?e!Q>+b1Qmn`x&z`tXzW@^%eB>g%;+DM=)T_lar z59`+w<$XLCq{aQQEH#gO0k8Kk?RzOt$@Tgityv8!tz{A-ib0D0A2(4_a2;OE5A!q+ zyN(Cp>852ktMjJob<&a0zaqq2IKAwwxDs7j<#vIOW7{6UWC>>%oKpA#c32fx?JFJz zYivtxabw<~obL_|9v2qBg}avF8Uk5#jD3*g--sc!e8j`1$d!V2jILBSCSk59#dJ}A zM3mdgZ)bC&E~_V}a^s!&5A4~8Ed3Qv^7UAXrMb$aR`FN_QpR*qw9azApNW>@ImvBl*9@?N&Nafw8V3+AARtL+hlRdAcVK{ zZ-RdrX@=9`B}|$N*PaTE`(+GlzN*RPv!^q8nvd@H53HN7*1zr}MsePJJWlUH6v`3n zFTQ9}tH{wpt}2kH>YIJ>m0q(SQIN5n_%g4=9?3siqZd?-#vj7fvn6Mrzl3FEbCaEY z_{R6y!=(}BSl2#W%o;Ybj=?ufb+1n!fZ}WLY$%0kZ=H^bjD&&wKZ99XOaDM<3u+MH z&h2QZVo#xkn3EFSfp!!(x%f6hW*hFo6{5l04mvk%+|Yl%vVJBHwZ?(b{Hd8}ju0|w z9v~OZAor_Vf$i&;LED^&SS+07Q@j%cw+Po@vt?UZ1AU)5WpqQ&rD?=2gc9ikaLrCU zNlcM}>#m+x2L4|{wOy-7p%55=*!t#EM&u);kPOiB<{3=nIji;(cL6YPx zaOA=RA~(5Z<2LxSH=;@Y9@c--(~K+Ft7V=MlU8n0iuF`(99mFaLCP(eo`H#VD4Btf zcd?Mmg{C;j54=GU$Z&ySgC>#G0o+hj7fDfaHlbzi6LS#JlB60^Q$_(=KKKTlPTfh} zF>NBNb>;l^~)c09uo07QnkYwX=#D!`#c{s6KrzPkd) ztKNvFLEHp@qgU8}awE^ADeb?!?c0BKE3ZCK!!&ni+|}7)snafE9xg75_6bU1va?(* z+GNmX257-G97qipjVvbS=44(gpYQ}4JI{ESx3-db^GJW#R7!SOX38kGk z($m!?n!K(Y77+gfT9n5U;Q+`n!lm2C7s;jIB3SF|phAAiC|t0P5H2L2Nuo z%X|yA`raC9vv$3??OM4{s?jNc!*4-;8y2-}+5~&!EE))OOhDb8^qSlJKD{AgV)oS0zKa5zpORL-^tQ!wt^;;454NzXT=xrobH8Bf= z`P^F&tu->wqng^5=Sk1ymbW%Rd0A^(6}d(~1GH|{Z(-F%t#iLrWU<^(t(yJLD$N$& z8pLj^8yUw{7m4914NFJGkY6!!;lexh($=!tsik$e5Eg{;GA+|6xZnyd)ag~Y@+EIW z-CL7U3JH+%ZKEIN^(&cjTTKhcv?e4e1jU82X8ll?;%enb2(txJ5Cu*dw+j>|6j;C3 z3jdS6CdU!IhK7CDHR;aW&;~La9;%`O0-bQPct0KUd!qwF)74}!I4MA4E zp!%#N!_2HCV~k0E{Den2>aez1AwT{KPqhGr(3dw2%fnPJ0)pm6OeudwahQ)aIHh6f+|AX%h78M? zusxGjVuwpbv||zET%x+{t9K6cZS-d)zL~Ej6WpqF{jg27(6m-5vegyNVf>hjVU0n@+^2g6T14%)UXM5kxiWHo zAN?Ey7dkE@JZV3srx)`EWZ2=t!bVPM7gPOQPcIZ**v(CCHS}>PQdW;Y+eMMByu)9% z?Nm^^edDRHBJ(O9pR%%A;2qZo^#~Pl80_rbkvihz7q3SPrbkX1hz5Qk$4QJqII>@!Ti6LUr0#e7vcp_EB#l^-xRDP=>p>v{rdaw6#qy0t^>VQ!LLgxO=Kv!o$ipTAnilSH24+>`-;Tlrk;a+!0~T zIL5dOa7_VMgb*1>RF3yS{uUB0~WC)^yZTOLY1whAO$+YbRXxQ(;`8x)UP^q*wD|*d1Y}4JT`S1t<=zLF@w1hm7bM!JML5Q66`CXAZ zQ^w4O3CG~Y8me1qkEt6fR(Vel)In}Lu8zrWJ{ZvEo{oVLEw<>r5pLokp5_F6_# zHk*>`R2)3QcEtCijavHi{53VCTRDKWx8pEEiGr5n??cs$sz&5*y3qbKy~D`Q4qN#G zT^Icg7QWP=L~&6bdcWzvC6h$0k$LHjvW9Jqn-iOChYP{xzr6{+TtNxrgNEU#)dsg6 z4nFqTbiX7X-?{tC^d6I?`!={R`f|z#-)>BMgO2UlE==;yao*979grs`rMOu%m*7z1 zvbQ66euaPzCm7Oz$YL9wGK-8E%yK?oinDKqF_P)Rk;Lm?h9n$xjByP2Ir|JJRG=bi z4I3^dn(r{PQ-a>f$#><3kNrKC+-6!=7!I}%2OH3U0nCJ#udWxb$js!97*^JB;ZVci4OqQ@#?lLB zc8CnsVe@L!n(#YtL?pw6>_#}+a1#-Bi!rm+o}k0w9JsgP_@|Io!elgZO!m!2K5w&7 z&}!?k2A=NYiVYc8aS%=LWUhhbpgp{b1a>CszPIvsE#`nm94t;H2|or zD9zo?24J)H{w8QXd_#Ce>Gy-~{!LVyg%O0V8*ozC-$<^JIyXfN9!qHHN7-Zvk5b@4 zswe}#kR~uhN%mdnV1f7TQ*jQ3Vt^{d7M9RH!K9M@vM!s-ERI*|rfp!zF9K{J( zeEH2<`C$>B-H^>1MGtj3&K|aONf?bOH-Z)_yNZbnLRrUkB| zH6QuZX(0h!AixiOAncN`x)-(z+kNC@r0?Kn!^R5D4WYt+_CeVrf4J2X+K^HafM_r% z$|Lh<+%$KVhD7V%hsVVqs4e@0s81ApRgB=shfrO|grq{HUPx%=EnB*EiQ z&lfF!^j8unpnyl|UNQp@CKzFT0`IVc;|b1R_sA`Qxde3}BwHb(4*wk)9V%-A?D5Pd zvH$gD?+gr&gLQ)BT4>S;HFf^qVQfVF&8M=@_Y?TgfKlPZj@?-)$t1+sQq{0kgfF5j zPN=O5|DFh@Kx0y)i`TxgBJ8wxe>fmlt`50W<2QiQt7UqGIiedbLWT<|jT?8H;A0Cv z3aJ&FqeDy%fMq@$8G2HUq7rSJmEq@!;3$5jZ?bG~<`dSE{!Kj#R9PI01NgNntO6s6 z=b|hvSa3xBo3rRJlv!J;G^<;yL#up@gDTr~+<$P;K_g~;tKE_X~+Rbsplq@;W5926R5LSIa(%wDVRKAqRF?uf6h@j8`jmG$hQl7u3M&L{l^2Aq;uvnJ zK*>}h8LTYg8%B20H*yXO3(p^iF3dJ?$RbN~`zF&e-AN-i;BGP<*Yo!!W%D+so_Xdt z9`OK`246jZP`q;t&zb4N{atWe90e+Pizg?~ExC=v+$$8}(Iun= zLt2kU*;yZvh7u;?8`B=X#{UC107irxfX7PZAhB_jy;oh<7*TFhSbcVUvg{DKe zD?u~GkK_7ZuR5ItAF&6O#@4hrK(F=su2Ufy!go^7kCT8NPH3552z*&j1M2C7awP_; zEsE}Pwz$xpkfz-rxbIAEadnNoka_wYIGo(9qkGv&9CMq0MS~f|i(}WYn``2mZF&|a zd4ERU*k|NnnknQt`uMn+l}atfa`bzU{sdB@n>QVD{$Eq;Xn_Y6bv-g(iiT6rErU*_ zu^lKK!TEd(PtGb&BjsAOj9gqZ90y?ixoASFb9AIa|B*<;KFX-l+A^PDvWFM^=%jy= zi5**qyVt-)w_w)cU2K7=*j1HGYnlJ$Vkt8xz5OX;22U~;OY*eT)bgQQP^NK3652UE zTkS*@-0vZMD5xlMOn*Zn(@jw}t~Q zL$q44+0b)_!)9TJmbqJyf15i26e5R|S`Jzsx)mu;55X^Ykm+wXkw)n|)`sET9cQhUhlK zrE}{kq%j*8(x@u>v)cmcKwi+ClXc9Zr!0ogL_{Q}VCEm^rHVUaWmoH-T`;qFMfsW@ zN(m4YiIC*AXkmm2F^LOWA{>r_cHSp3UJ>JtL<@oj{*9oDV-YRZ_{p%OPZ={ii#gph zu(5-(ITM7$IsOn{(j1271*K4sCFwL;*>^j-qmCjD;7M>eJN6;J~ z;i6X*J5x>%3ZFh0k~FqgP|~Jfi6atEI-;m=nADtDhY40~Wx@N9vNrtNGw zcWQX1P%u2l%C@<62=Idpjv?TC*p2EMFqFl|mAg=EJ{q0hl+JuA zYDnQXn!9j;=piP=^S1|v)b9}50t~4D&LXXRv(QUh`Zo>q%P9>rN(TCYmz6^bDyVzk z?J@WT2MV^YtOyhj9}Q4^)dFK9Mpglp*3#0ir~F6}WhV?8;S93KqF=IKgYK{(G_JdXz0n;S&36sc=lfVRUzJs$L)!4-V?{VR;`2)7L4mNgd+HLkb|MbPlW+((2rSahe* z5_92bBZ)n};3J(pyg}huEz?1oI05kVxtdWpsdW z4nu=0BB)zG9UWe>A2_;%0=V(6Kmlz126ioebg2Wv2b0z^{~%P&Ua-~O3Mhd1_Z=BlPGLcw3aewSk`?CPTJTIR=W-K2&%tb}^;Z_q3cyX#1(GWU=1sKKlCZY7dOotkbLjun0h%igEs@Ue7PL8 z2I62g--8ZIcX#hv#b=pa8s5S&yEHNn(L_Zv0c7vR2-aU$OOje>vo`xpAxgD8-GZJ) zzqbV!O(A9O!v_)8n$jWt*?UoX#`F4wD~`|*^AU9IybnL&;X1u{&`;gabCCHEIYb6?mA5t?15E>L`o<0Hosf$Nk7`0Xrm53`(-7*55tQ%ssldL7@r8w-fZUj*QIt{dhxSa~oAO<|~-2C+As4?kq=g~{tbj}S_( zX-y0cZ-F9G60xwn{!*bZ)7Li)d&N7<@r3Dnj$nnr--L34<6#ALkyqv}gyUjIgk7R& ztx96-GtrS+n|j~4FCtQGXA!y3CY<^_?h1K=6_=_|BIS-WXqD0*)Fy;M93Hqd6)_26 zh@BT)zV^+(VJ0ara0Euw4t^bDLj`4a)+{5PqF6IaFjlp)V47Dj$i_M-R*9lE7R)BN zV@yFZtKdii%@7omypi>coUof_1|7r21(v`n4S$EPOJLIM73QrpSYu76Cxp~62I&56 z$WB377=>9j{SFG#&lOExO7_ic11*k)^dN)rMP~H=h)Cf!3X#Jtq6ORB2_C#}_&*qW zeM7Z))u}V6mAi%V4~YQBADH;kh?i7}GuM;C2(MG753Thyms2p>Ag*{rv$mH^*bqIG zlF|Nt;q&uGhh!9^JCt*PA;UV6Gm2eMJcIOKWO}6~zOe3uXTE@8kj+!NlEQ+G(ig?Vsyj3{5eeXxISL^ zh43UYE>=0I;YqFhKe9(u8GzL;i!-$F6Mx&*zYfy;fsHaO==k#r<9Yq zNcmAipD`7m{9+THCP8HHUQfZC$i7}YzCj6}hFFQh+?H9#R&_Ym5~ObgN&&K8#*oXK zn4fBw;)@Szl}8vhuHIBDHK_xY+8C2^g)17&Mb%Z^Kaie-FMKH~Rzc6vqIeR!3avtnO$sMVN~ zyr)^o-L7SR#Asr!N72U2))IBsidx{8{}f?`)MVK0cYhUSBBtAFp?0kzJOiBe%yCY8 zjiJ9CPAM#xnW z%hR<&uOi?UM0q>dOQPr*B}HNM$`yx|N%5iu(He+6cPt`qiZJ{BKWp{_KApH$5rBy`7|WyO zZo#kg0q6>1VWIVIYK{Cul6_k5Cf?110Yj$qPSZ6&(|XZ7U<9UNmsaRW(W5%ftw6aO zTBsZIw54!US-q$u!??t90zhTFunt6ud3r5U13EM~By^W&7spIY z!{p4lDcn(@zKvwatn&tk86&(3%#l+U7*58oB(w2O`cY33B?;4wB2= z_0E*3k<%Xqe8V@IYT@8_D`*^9CmXM~8d#%oAENeP%~n-P1r4JV%7$&?ef%C)R{wR0ScAdXXTswrTm~1p1*gBJHs{UHHh?=I??sf%FBT zi}V4sH>Hgdo?Y>jh52yiEBLx=LqDjjV}<5!vaD!GETkQENGP=N;|X<$bM05=CyokU z@#^Y8?fL47K~<3Uz3_VA{|Ui|MsbX2gZ@wxw3TDrrntiGT1JMRctI)7r;V%F@sRI| zlL;ElGTmi*lb4H(sY_T5M95A}L(Az8;CeyP!DE?%J+Q)>Rc-m|77zVb#IET#F~bwyibxa2IE=Hw z!A9b{G9zv`Jn<49Ux|o_Y`rAyP7Ad!%56LzX|y`a`>tCtYeQs_aBRfi0O?P(3R;wE zq`nYA4FBkRbbl}fDVT82B^z9IfKva{1awELh^YLLT72Mm6P)#jhOIgyL=2g@f}IxL zvSKOyfoKSa|64B9_#Hyy2eUoMrdQm{zxvE+@zf1f$o)@5Hh4i|L&@W_Hu3nVwo$R- z5z=uR@`fW0c>L`_9CA;xNgpaiDf1Qyj%kYnG;w!LTl$HZ($QnU{v6LHnibn(+AYId zjgj6WZrAJFHn43Y?|p#BFT%@9?pcguhk{C&XQrw7qctV|&I*GRMHItu z2%t5+472HM50c4;ak$~OFku?5imm^tZzD;=pK`X`5n(=D8-bvdGTt44>%WZ+^J(Nf z8AqEAYOY53YH%~y)0J8^oWCi|26??+fe~g_NeTvatUN8$r=PujHZM3pB-llU(aEfeu zK`+WIwBO0`2D4-&tWA%7m_l(ZT()f5mw5xWfhwtEvG1KF1oLzU&bKY`cdet0$%8%oL2iM|n zyhVwVP2zF;=KpfiW~2}{jVYT?a_ZF&9tl!5Tfa|D7?zNnR8}tn=cK_|v2Ye?mHW8! z)FnA3A_za>J+A#vJrajo-t-Jd&u_PLh#?3J-}jvW!w-CK;TWpWl3H5gDPBudk@&Fa z2Y*RdW<_unMG2q5@1US}U$Dot@CHb%(BBI0W3i*qL@S&TsXsQem3xF4{7acm3++u5 zmjKqczS9)b(k92)i){V9n;}4dF^767(w&@{_4*Bz9|vCYU0mIM9gA~RHq?S$8vK;C zZEDhVrEQwp?+n_mOmTE__`pHg)OabM7xDWlBB+4BK8Flh)1@$EI&A)}!lm)`dq%p~ zi!ukl+l1G43XRTOYvDoxS2MBSBc#aV=1)A`3wC_338H6dw^UIiD8C}`%IGDKC}FhR z^@#K=i?rLcZ&G*YT@uBO-wo;}zgs5X895^dE%a1HFlvJEeE7RD91T8WH+-W~i*jBa z2q?g9bN|5FqpD;erl_yl=0;fRW6Dv-P?5D|Y87H#g&L}D^mjbcmPZ)TXFE=*=ndfR zdY(=eCzqse<-%u13J+MGfmdvD!It*2C|7gdiPRU$Uygz9N70n|Ew>dN1Cg>vTC~hh z#5gFJ76ALn*~yYMhSpF^9yv!VPlyan2Vn0HkUkFX&u(cYE6Wkra<*SX=OK*w7ut?N zNAqQ9K}>c-7qSRalRa>}g@01$x_tJ#2p+03dU5dV2j~Ur7dUh_DmT7OuX6Uwu18p| zXB6Ss*{dH%H1Gi<6b3waHnWi~RE6Vb_mk;@+pXu$=Cda}8!bh6_G}6erY7@n@SKm( zG&`AtL*i%FF3w=a_u+FG96tA>vYh1?Rl_<&DtHXt_^@bTqw+}Z^k>%+r991&Caq$EN39@TE^6Tb0*t-prcTL@zgUKQ=QP(@eM+VpGnvrhy4a z%_T#vh0DM5i6%uAPF!skJB8=9TboFZ&VN=@+pA4pg;T<(GCf$XMTv{DK$FpsK@(xhIw2NGs=r_3Pc#s?hfSg2DI`c{w995ANt5? zvg?5HX3p#k-61Azx&_u9bLjjWOgB_L#0AzSvk4X_ zcuW8+v|e9ccU?Fo$@3aU@8S{=s(U)<*hP%(8!j{i=A%Iu2Ec4LPEdK>^&mmN<8PF! z3$^h{)D;Uu25G;;F%3vQrA8-Curm1lSSY%FyZ`!RJrUu}6C@!{DQ|S*KRj8V5<~)O zV;ikboAt(Rr(ig+$j4d4SKu$6!ofYMyF)v+`dR#jVfoH)O{&yu=M7KQJlJqtzv=|r71JxmBsJ<|k{tBe{>fNT+H3gz$=9Z9$XUR5CD1SgB>9Z%O22dObE z+bMLqE=$7Vy9$atu-qD0U6NfVSCS}SVwCcRLiDsYn8?5mj}siN&lRZJIm zvnozyvsq|hT6*|f(ACs-8>$c!8HreO(f$5<#?6pRi%x8T9BH%k`4 zDT(ClA|4L&&weDAY`P5V<1}^72_JfaB8jS$+D&WazY@ZT?*P8|42lq1Z#M2mgd`gf z_(J#0F*9#^Gq99j(tYSH|3w0b3T2yaTF}1y`4f;OY?KiBx1TgYTTSH?5}~|pPnC+8 zp9jh1vl*WLn^E@pwBz^SC|IB@E3n@A)24-zk53Wo(BOWKt8nCk&0`$7t}vq)iUq@K z`dQrxCqx59*l-xN%|60hTp$TC9x4bpCwz+LOP_C&+0qN+gAe52P5|)%&v0P*&ryq> z#wagPWU1oZVh#R|tEHSVAa%?kf+LGqbCOIRS&T#h5=uC_m?Sc#yIIrR1ow)^B2*T$ z79~$$*qDH0izz&?P_7Y1Z+X5+`q2xcT4*-#S6D~i$AiE5g=fMmtiK1;zMn5#o>&5L z47F8W&jdbWq_){oOBUN#R>X1vGC-!L(SW4Y2wn4hcjVYdJhl1h1T~K z4bGQsWg+wnt!<5G$=LQ@Pz<{T3a+esLbf)kcvg@XB{t4yM;jC_zPMnTi6Jv z#)otZPe)ma{w4lB^~+%S?6DQK1G@_pDy5A@fj>lV<&s+7mNw6%DJ-6a_MPBx58nIB z8m0_J5X_vVcV1R{9f?h1*VO{NS{vAGI00RA5NdFw4=i#*M>wo;ZmJ0RB zg2GHn{?)%-Fw^`x^C`^gR{jiIp{$SCDc-LjMCp7yywJ2Jg4_@sH2BxN{?|?7#t$Vy z^pJEa4#n3t5(h{9y0J z3m5YHR8^mI`aEVNC;5EWelRAFoDkOdPbL@uF z<1aFp&uZ}G@G_DupcX%nP`arvzys}KH1*dKNxXhHIcWS<~Z@c_g_iDWH<)oY+= zHD52krXlIssh>Vwy&MPnd8^|%pf77!+2pgHdi?k6patHzB2DFMw=+;zRb{K1Jc9=l zojVF;Ej!glUJ`QNE3P~>ybI$NckhD9tD4~4>RQGJ^)tTrLiK8CG_{wqrc=5Zy7lLR z{6B{fqzn_=vMNa+YjR*saF~em0p3U?4Xq?xTY^|+SzDqu*sq^aZ@$!aZU|fJnMr8r zegzvgD;kK~ayGqqGw;}vxjAD*n*jLsnYfrrx-qQZBA}ZO!b;jK{yeOgbO-vlj9!Vu zQcl(j{p{G8fx#yztuh{-jn~-}Bd^Ad=H083D;q?PpYa@_GuQ5P_fMGsj2OV1UxWi8 zS6(|~EJp#m_@`DZnlqTolE&s;XH{fQ>tCGrWbch#-x$c(uX9EneFuCFeVOm^l1SZV!f3S zLG5MhSu0&?5z--dwZJQR_G;@rm4pGN!DZI6_&b(sDeh$}@$dcgR!7biOStwO3qy1U$#x>~By&*P+O)v9q%uF|GKnwJ zFsI+edzaIotFT}{j*hJJm536BA2DJka}E=<373{MHBTGNd8=2-*GliSm|MC}hb1cp z##7`(ZrH*71%?;MfA~sMWpdKp4^DV@Bvvw(FxV$oJRAe#j@PVBLQCdOmQX9=E^5}C zUUXw4jsaBHqBP4E^kU8u2Ear7+st7`b7eqW_e=wosB5De$9YZCiJ3m z^AQBHa9ThDhAg1b+A=OVoUgv%r0vjUbZi?1rHY&kLb{|W2`!7im%7HayK2>LPcK?h z!dmWy7y}yF5Wy&@X3{kzKr{1qVu%;Had3mWg%oj8%w~mcLn{S#^5vC!_cx0pS-CMi z{x_0TvIIVzSqq7{ZsH;Q@22>n$Jb6b<3!%dUDoG5B6!&{XXo1FjlqufNoZO6PiEb^ zVlmpbE4}zLHg%HyBO`-@2-mgAuwa-=q+xco$v@?zyS)bwK|3)ZZ$|1jG|D;=h6JDXm(f{BaaarlZkQ7w@~?zMSPUAG)SS0a{hLH8|GpqB^m z{V8;0_UF$=<&iT>nr?Wd<5aq05=fNpUcan~uF#;wC_ri>{D8&vi_h^b^l}{^MQkD+ zZ%N{4nVH-k=pHe6*}hgdM{et{pPgLs+-Suj(GVb=d5Bm4iizm8efc>n?@*aH+=qNo zGJe6tyD#X1fF!;R=W;uZsUJVCWE-N8Qtju{r-7g3G{o&0X7i+82KekB@#-RBmLe=y zOeGiK)f-k$q^zE^cv>3M)gg%n76MGP*9_u_VJK|FfF9d$8mnC;Udx6gv{a>JTd`R5 zcw_FIB(#i6%WB2VeY`I?7p3y)ws(pC5kx*Ckw3>JP}{Ip62rP9aFez0EU#t%Yt}UA z>GYz5-=)p1e8=_Lx!ChJ}MKFCOYcckyC8q_U-(mhQ~qWxYD^F!I^^oVlBO^pUpB?39h+>6(#? zCl1xXE6@Dm3s%ZNESMotF3#b1cU0$(j^)iZyP%Z?MitP}l6OqIug|}K({>@+Me)v% z4Wk|H?v+(ss0*PhIIG#e>9K-0V7Jky4>rNMi|6r_PHjvQ6!M9BZD@KuqPq=O9cY4c zVG}nzUpCz2-By=3D};3={>c{2h)J(nRi~vClqMzYdSDGCq}4NGLi3;E?>HnR7rH3) z_u(Dg^BIMAwA94iV@EJ->K8cT?k`%t2ju- z0KhG;Y=U!_BtXRC_332XsC|GzV3XLPZAM5uGM*7V}{ z852r?K!XLi<@_Y>NoxZC!Fh^l%`oxNDsJ7HUi^k+G2&e+W$y$4jx<4CdBJhPUVJ#6 z=SMnGL9IxS6cZyL-+Ijn^05g z+_yi!vhl--bmmeS4O`RedQHW|mu%S9Ne*gG0a{UtL9&nRUFhOp`)A9#C2Z)wnQj3M zX)|ugI9C|ZX7LfAGP~o3>1isEX*2sBr+maH&dJo7NlzPA6W6q$Z9S-24#gaZ92{pS z9d-dH`;3eep zZ(*HmNRQveACpo)6EsL(J9HSDd9Ou;dtvB$P&h>9!Tm4Akq60qv<8{SacnM`U6#ci zfcsJ$T^Kl#$Q{J3;-G0eRB)Fp7{KdeWzFK!g~aXDwL2&GCEQ6VnU3hmW}R6$mALN{Sh=h@ zaJ=CToO0gL;L~gJ2{+Rz2#bzhY`CuhlWM5zNg@S3FG&Ksy|-psr71y-XemJ*3$JF~ z(o}e9G7!a4v(5UKqAU!+16&d<;}fnF#Ec<3C=wTm%C|Ls^iuQ}{s;HADXvCyy83as zZ=taP>}|v8I)K4fjlaV=fz5FZ?|Z3K`5Xc=HaZr;Krgzw>-IfOaPGJ?xo}x@ZJ`Z) zohGS`3;)DJJHEx=ydS8uc~94Y_KFLW(6S)2^+X8aBYHFqf6j}SD4?Eb%N5={N{MOk z$k)BIoiz;ypNo+xf}K*CLf51Ua7tx!auWJ^6*;A{jZgaJU{E@P!o>BF`OFrabV?*- z?!e?yXaft%sf$up!2J6$Oub zC4li6TcR-1)eCq~Nl)+0=*1k!yPLEMV7(-8r1|(oNLR0zvp+%m#mFM>V1oF4x?j`q zgzS>^4S5*gri&5nUX1A*5;@%xSqGXQqrz8d7Nnd@@H^fc#Y;o2p{t`V>~CMB8c9QD z#mJ3Vdbup*t#6oOQ$?Y2pwfQCoQ~T3K}mUpPmx3l{Rp_~Lo6DaP~$bNxg;o!A`R<8 zyZvIC@JJ&_Z5;nvf~+y@#SLly&r6Xs@iXpQjBKVLEy>lk@RVfc^JN(C|04Vjg$q)bVr2F_T`rL-PCq0Lp$$$wIXm z)RmYST0ZF938_2qWNtb7LP?0Bq_&1?*%^!+^#|UW}U#r)aZbo1$HJ(&CI98FDiv@iuKyTx$!sh%CEgUQqwN zTCpf{a4{EJ6kdoAPC+ZnBrd`Duy$kzNb0#FQnBYY0- zORzfz?-_d@ypsv0ZCM!TI+07+hKFV4vTo#USWLWk1iPKE;E}}o44$xIo{pX@#T`DH zVCNPVVVV?pS{9VFS@PG>Ua8u^J(aZCUem}{EK3*?4;yZiAIm~m1KKQaweXQTtOIFc&6coRVWHK*IXCLw&=3)HxIjZCZVO; z!wVv`YW!MvCDPr3>m)I?@yOv}^3KdU0AJmSlOju@adsE}#0vBTm<)JmzUm#gv~Zyg znr0DFKcQ-Ux~)wVMXyY^g{d;+TSGk<5RBmru$5O-bV9cCR1Rpfco5>xv5YVt&}J!K zNLWmDf2w56@qji<&bnv?eo+TK%zD-th+~z#j1)2&jtK>;vW^iS{Nxt*^YR|ljl5el z$VGIpgNYs7@}pedYgPeZv0wkRdD7@%$q%F!SBqCb)Q2r>H!CslwCnlkqVDgge zqdQ@^Q5+Gr({GwnnDX>jhm&e(3`Cs6<1bzpP>?TBuh$@Rle%HedHPr{^h&IkLHN1B zPwi5-+%#L>s-Y>C`^_2$6rs8g?+Pq*e>OkXg=_o>uU~(S`xQ<#q{koQ&!mF}DgZ2# zP zdAwTS5v2}$g}EA5QN`jD31ol3a_`MJR}w5GSQJPZd6f|5o}xZ2YSWml9T&#{aFjg*!q-@noB>zsxulr}q0le24mVince3 z>oxmoy$H|9SF{h9R#x;ljwqEK)M1_Q|Kl%;n)X*;>*dmi?Wu5NWs!CG%;hREhi+}} zV_8_XZlhu~;J26K3I4e(qr{#h+_ncdk}@e=1P*kitLL(7F-ak0z}wQKoh^uxdr>N> z;W>x&wwgw9uD=7^n8DhhSMLe=y;M9?8<(WJ^&5M9ckAcgj`J7qd{g!=DxiC=L?eF< z%PL=#sa5W3NH4vME1)!S-J=dPOZBN`3;1pYq>x~kY1=ShFl73VkJ0JNgark*G$&@D z%(j^^V__`NEVCCT8!tdL@z+=4Sm;x>0LK#-_2F+4pVmmz6uGy9Jm zfcME&i+cpV*R_yf@B|RUHVSZGV)in$p{VGvX_jesangmHD~cWmpU0*nUi^uR(t<}5 z4}25_X{iEFB;HxZKaEO>7br$0^UASdNeTi&IR{TC*y7|JoXFwE<&-R&u*@l8z;Whj zcYpLI`Lk5P_#K9&cH^v+gl+y#>vHkKNYxc3%HW}zt?9**Wbb1DpWKU6;9w?i_$vLKEkI+j7fZEVPUH^WqL*D~Jz5+nt9lEN zN+7z)ENJQ_`21-0R`?EZgM{a zrI%yW@avejeddqU>rZ!ojw`vRlt6BgL{iLRZqP;+tCYVIn&0|zBnX~ArqDFQfg4#- zgx2HtBeRq|)r}jGF5(J>5|PfI2@HZlEU6=T%rIhUSpk>55-pP3S#EhFsZqI_NH4rx zj^`k6wgK2EmdQ7v@s3yGAOu7IBosBaO z8Ja69ot<^-L+eG|2)gg;;bFQ3-u`CyH?6;}{vP~{U5c#3Rl0wZLCVQ8s@DIB&A4_) z4*PnZ>ke|K>IZ6Yv~?fCz1}h%cP-&vMHs&|Ty8^wVW82Yj;Ab(5roS^v;Tt4{rRgw z@-HH%k>Tues!~xB9n*cGmux)6yxoM^PoU~=gkRN(|#*D>f2A2d2E(!hvd4q9Zw%KsE zBFPl;s=)E;#DASZ-mxTFzyP=piJ!kfy9c}74O~1P$YU<{OD^{RfUH~(d%xA$JvEVc zE8&gWmrLg%Mo!oOYoA4@4XTn9&2qBW)QZB+ik#f`$1-N$)jmwC#~oMg18HT`xil{s z0neya&RRk1pL<=|4n~*au_OP5?Ak5wQ1Ql)Vo;cJTt%&X?LHUhn7L3_r#m5;w8Xj> zbrpkMQPU6ai=UNnCwMsFvK7t-&}_sKKhm_ZH%4Ex5z%9ynP6J5>0?cB_EIyC;;CK* zsWvNIHZW-GYLLBX$+wk1@a*mslAu^N1gEgdz;|wj5Hj+c6@c%&LB(5 zm5ifzBhIq&`}O$Yq_ z!$oz&a3W)auOGlej{~bno<@g*PYR7cD(%N?31H|Ts^BRx2kpF7m=KeuC1~R>b*?t) z2#X4&wk=g)maS~GsH!rei>>HhXBJK*4w6}wL%Y)523W$z>jMi31AC02540ap88Wax z#7ly$5^HY-(e4hT^nj!E9L>rLXPCEnjif|m8H#2%d{1un*afSg5$x}&1Cnt`6QkFP zDjdPX4by@rKZxeKWY8ZR4CC!qYfppAHsvhPzEEB5w1eq!AVvF`5H2o^_eY?k+o?I7 z`2f#xvRDMT+d<2#1dXpiH&F-@)<1Pnt`nbSlnu(*HeAt~ZU}2fQMaA8{Y8QCNAPq9 zhM&ttg4^2LD(m=|&W9s~%NT+nFDjYVy;{Wu7V+QJ=@X0otQ4 zKsOxt>}5@G?t*NGyorVmON&V7TQgi%v5IZi4joniftdI@lFXAOhvtSV`7~jH!KBgIalaa@H5O|QS+IlZ! zukSzP>Ty@c^Z(=h-LO$&h>MN?6ZUd*bJr;y2WIkBk~O*N7@5rc_x7;o-Pr1{$00R( z>%mJg>`|hF*u3tl@IXRN{FZc0{$==P*`9t792nyPsI6O;?qy-^ho%&i zlNSmwt42fZ)HyX) z{MKgUW;~cyq-ZU0FTvg?D@K;34_4#4o&(%KOJ+e@iA_Jf%;O@TA+S}T!Mg_{Cngx} z<_Oz4XqlBVJq|OeW#!e5g&tM}VSH<}`8h(Td*%ot^Jkgf)~%Vnmyvf0<8|85rFLaz zNZ7(@rGmbSrD2ADj%2tpGwhh`~pJmzgpF+L?*N1P$)oi8a7LN z>=?jnDK5YO`7RnwXyb>mFiaD-{jaZ9omm7r#-r6;%cB)-FmNQwu5M=-#L&{ck7p^w zqcXIiG2Q(>{+`sP(TbEqsTekfJM#l?;laqZe;3pBK?-7^<6`*yAYK=hm zU$_^=4uv%&veR(NFNSIZS`xsaHpd%~O~H+|xvPPLZ74(hkif(Oupm#H(1VW{-6sP! zEO9f%TRq(1#Y^FDGJ?DtM2~ol`6E8Hm4ioRhzN4|ego6`2(1K9NH059o^C*u8d;YC=K7OsB0{W;JaRmdUV4yu6Zcc@Q8YB}G10F&0o}92L=^*( za+64*d-^Ez!GtBWDV&A^|7Iqbp)1vc zYgnl_I3kzuXctAnc8!rDaig~9;R`Rr)4LP{`sJHe10t9XX+IJ*7W~t7xOdI+QVgwF zS!F4a?w;ogm@Nix1-lPnYIwMZ_Z4Z2)hBp2&cw<3p(gF@RmTu2<}E2;WIRlXBW>sy zbuTY#+~w1wD$xD2>ykjxaPnjo20O0D5~TOO%A}j%tEA1+kt{(eGn5&arIawq8H{f_ zp1650KAu7$K!=J&k=KgU1f$COC@fU{gx%S2xBP;?285$A{x;l~Sp8zeA{svA2Avhn zJT|i7^AbeS%**Z`lqES7+>*)1Y61?d?ObW?fqj1$1T_4^Xw6(Puaeb zJSM?#COCK$qpDA`?Gj>fmK$Ex_Vl751%wyFS*k#z^t_mcVMDWi6ek|TKkzzc>i^Ks zDKTe5=36)5CoPvlh`m~|XhUn$t7GEmLj8W@dby-|oPl$B(#yVa3%|M;ooI4(KeJ~w zeC%k*y%lbe+PL&}w$@S2x>rP=LcD66l9Q`I;u4Y@IBpynyyPPEQz&?}-4LA$ul7SX zphff}p5M?Uc$K=+i>4(1G1tT_1KdlDj(#s`9H88V6iPKbCAJ}h<026Rtt5QH(%0)@ zn- zcft^Im2y$XHzEdOKy}~4GG=9<*v_CMaLk6_;sf_SM<8^|*oCrCAC*{(jOjvqzm|n> zFyYhNMmxt;T?tR0Qm*Z77~K{8dei>Z5PBaSpC1iQ1MU|||5Hwnmc{#7s)Uhp!}Ymp z_83a>*D>#f8I;ri$eOKb_!dLt#ZB$y;AnJ-WVuG7k07A)uuJmy1o4-(6xXf%*L2=^ z(C1dS-EGGrye}c6d>WTS!1aZ>bSOJ9aH5bf)4R=ne2mm_Y_FKKP5whHT~&O?6q*e_ z!zTj|AO#@}>IR#P@T+kQc6kY-7@7>f)`oQTQLa*RB9IkeEp8MdgYx10<#<%a@_fn} zf4_!133=r@gifGhdpdIob5Ko;24$KkSbO#al1465OUt88oFY3uO#eV1yS2!)x-bVd zODjQR8YIU{5!xrpt_E(b}!)>f>Cs=2Qo#wwVSTMm7rgf@~Do!HQiNw9Jxj z3&yo0hq!*45eVTLsoa{bZd;Zs3HYHt*AKT6+#?-55`((3vt-nNim1a!8KDsBk;<*< z%s03uQ(3YGLxQzh5)w&LaW*t^_W2R3VxR)eNrnm&JkrG*a=4z%dN;49g^@F@9eJ~; z8TmHT%I}!sKhr?FMg?Zcz%N%1N#`0^W(u*(SW#jKMO1Afo4IWY8 zwWMZ}Jz?hY!F+ELNfp#z=*R+e*0wBGnQ51y7M`i#HKAN)9?u};zhvNgHr-zg4+pWt zgB1&fy*nH(fCoOhCc3xq8{D5yE_d5MG_2s}g%~_kV+0d9;eDyhJp63|b*22bTz-Tg z{Z6l=|H?1BZl4eMR2qKFHfnG_c*l+$yQu@cA0b8)76!frPGy_Zu_(ab#VtmVJpSn$w;FO0KmAv6}-(X&%?`z?!na290J)z8YMgWDhg zc>_S+WNAKu5A4$zQ+#ngG2&X=Xhyey%yx%Q zsRv1=R%;GQ$%e3o!0F<>I9=104tcQI0TjkCQv1mmEah<_p6j$M{(u36_{U$*cqNK@ zR;iWMj1g^0TrFl~{7Mai#eun|h<@E&V^$=iMr3!mL7pXa>;`<)Iz*~QA89&p=fHdJ za{sh2(1D6^NLPtD%-}5FK=diLY@-*IkjZQE|akfI^ zISVpqu8Bs%{`tC1K5V$o`a0Fz=oc2C(C8;yz-}lgT8)@mIiyow?`~vR!Ewr7QW3Xb zwMR_gp>|fk;k1eCqNAM-JS4|&2nvVQuwk^LQ>~B+R#=NxSj@^A3|2c6Fjq}<0geZ& zEtj%JasE8d*pWJBCZ&d0&Kj$;pgQPv$Kbl7sbg`$i;wQ<)M0SBusWn|NsEon{Fg%v z51e-Is$(Nw_ax>Jsc&rZkq4Ot{zz4U;#Jt`|G5jNXZNlvGe<8c3S{PC&{ub!o;=zv zy`^Xur94q0??WEl3GHahL-Ob@=q68c^62IAWg}n*7|AQfzzgmRECae9#Opi0?(NoT z%2}Q?^imULySKBI5gZlGnXe+)NMRv_E&0^aETb7N|M|NLFZjdP7B$;%wNDfKC=N6V zi9~1=zp~L%4f{W%AV9Yj<2|I!In;)YvQ)bIfV2b;t3f={)8hv5NM}8zJ0VXThf#;= zp+bI%gM_SYZUl+Rf?3xmCW!trM#uP>{1n(QuIAsAwHB4j!lnKH@POZ6bQL6RKLGFh`n*+)<3&~q7XukH}*5O~c1?8(`;Q=b}PR_&( z4$nFJ{p|~8xok`?4$JTvBC6*I;4D!!jyOX=HqWY^u@aCw&gOKrpII|oW(|93Wc5Hcm^=y=SA z)W!vwv+_%G{puyBt=xm0dPCS;AK)n@KWS~Rp6$pSm);7R5_6v2;T7_TGj7~M zW`29-d4t%BHjQdwuYXMv_&?;0WFzBwck_Z_TD&C*o1Qm1c+vk$%d*t-@8%PKrY>fd zhydv&8IP=Rt^9WMHEIJHam<^ z5^r11fkxVt^r~m{O4nC?W8CX5eiOkcxt-TK6?iLUA;A3vd)pNIj;xGH;X3Zlcpa+A zPqBvZ+l5Az4t$2_V4fMlX*|@ zDnI7dmM3>zZ1z(_=qBHKl9Yv-{=XlikigB?Jx-Pv)}nyf0{=XOdGYwEgEE6bJ%`$o zd1aP7naKnA;2mps+k0UF<9xAZFJ>3%N8A;53=!l1&L?4`ZnhD|(qC3R%6b?!`yVo| zi`?w9gyz0(^p^MIby|5};|s3~a=KS5lde9-bA#HOb8r|y7u1O3Wt>ZL0#bAUs$i}3 zWxsA&$37p*APEGTysxaHaUuULAW8apKpyz|x+?a?WR%rafO+1sd7bcoZvm&`mU z=iz~Iv+Mk5hkIKyOaU~ugavzVRi?;axkMycMcI&NOF4iyZ^T7e#y1YH7S4=mP04K+ zm-Xawl+%kJSxXLLJ%j|UoTIC!;uLuG+t7a9tq2Q?gyB$J67;|L4MD?5#pUGdgLV;p z!{!VTt)oYICUII_)8frl>?J-%gzbiq7jb(=KB&NEBIma#peHTVVT=b4or_K7Hzc76 zQBXZ=&$pAmxRpf)t}jMRp22=v5gW?K;Wq?~HAWrtNd5E;82xe+w+c-icCs6F^bpzyYWn+^&p3X|(EgbiH&KYI-- zq|7}04)K=KdO30J85mJ8mD`cL^+b1_#?9%)OC)6i;_MsqC=`4qnW#+aCMjXL5}MN) zcj{txa*^uSym471V{Eq|$w=yZR#YvC@0)Nm29o$}PoerJ>{T0JL0!!1*O%OZ7`y%e z7Zb^HM}2j}MhO6$Gk7Rg;YXJmU$X~eA*B5&1tu>Ul+uoTd?~}#DB&JIz9fTn(J+H{ zzB7e^ApT`3>2sy0fBh%rX;g5CjVSjH(jcLwFnDMn*W3|URiUrv9aprKqR}dbT`6OH zfXI#PXFKh$vFi8_!izAbW^=l`ZdUl1CT}{9b5WPYNmN{!Q6DnknW^JNBep=D-ON-{ zMuc7q%G;qn4~>rO8|#UGNkl?9RwkSydZXvUaUyndA7w0W3K4sTy@N{M97 z+^<%M0M+scX?-mF5H)D7J(^k9^WYkK3G)B7bTX9j@QoB}JZ z$#rj*7aQ0$? zg{eGSBAC5mMa*>^vz@5Sg+@LeE=}nMd%-uo71vqBK#0^!rNOlmP!Q3Ix7a6`Nal2U zVBI~9%IW+aBCqd6!2ZKqmfdO87(XpIU%hyAXHo63o|J)^c{u-VI8ym}2aJ8D}V&jAt7ua@kFX#7UYkCsPZ^s9B{J4h+-UVDjab1)1$A^j#>)WhsA5OZSqi zy=n>>jPCF?jgPvER*3Jt^^mC@v*jxx%G2OT{S}#8v-UUM zkpw^GMofenXWp<_<|+_ls$O&g7nQ49l&K>disBf*&6xSQjTDH{2IT-i!*mL|LB|G; zkdb4}=>-=g%>nNhRJUPVtG^+*b!Rkg7%bVSpE^Pwa~|=+rfF1{GqJS7|NK%4ow}>e zYRt;2zV_iGQ={#z?{pnhXfVU2cVWV`By%oW1qz#^BHw@Hz_6Y-v?6@)T^M36?`-?) zJ6NeZy;ziH4SrzVLRZk6*+7vdNea8d8;zC)@)Gr=0qq~dH)#|@<$wWQ9CZ{lFMO;n zeOG9cEtMLv;^IHMtSC2O#9Xq*!|y`z;msXtHRUr|&o}=$8gNSt#2<28)l_;sAqVbd z5Sy?7*oH&drvQh>@4hSect9M;&3}$d@bD^h57K3#A<98Kp)h-Tl9%miJkO4rkAN7dpOfv7k{ooRzV-{sK@y4ZR0h z#Fe6x+LR!}8&8|NQ!GA}hRXMc-V?wdX&1WPgI8%<;7Jw?icoN$f5b;0TgG3{-aF-c z02Cdn!EIMib0Y51?qL0XbC530mGBJ@dezYCjYaLW7y|YkvXAi;b>~@CWQGDw{qS+<8@^M`lxx)Os(js74TfNWl8#L&mSFo z{t*3>WrXdzf1kV)S3$h{H%uFv+>yDmpx5hL#-sd7v~eRjFp#6ATHwVQh9v{25B1Ug z$Cj#mH$Ge~_F8sz6p1bGibY%aLEt=*Cvk47Qbv|B);79 zJ-XubRD?SSIYkZIq%o1LKvWQNF}{f!H1{QvNi{0QSoHIH5?ZPp-)UIHhPf$ABBnKzvpr3!@p`vNoF#d}V zlQ*(n=U0E2P9@{3Fd!W7=XkvKNBr@gI;RN}LdCysT^C++y8Ej_)WB0B@rsY2 zc}F{dl2&e;sAQ*vXleZ>86oFyY(zu{QA!1%fNMzqv3TV_keKe=&>FdS>{#w}Yi z^Ki*MctZMlUJsj|H@H+9J}B3M#0cOIAHro@Yf71R58{k2y_(TTZ@bAhjm}bTM^T?( z%=<)Erw0%(xEJvq<%*kX*XvGaY{_3Vbb#CM#l^9Iwp`Q^H3V)|!LpN!r z(DLH8D7{+7qw9*k{(sLNI0?!jOtDxx<{OtKGbf6+MEtw=;TpVy1S}P@hD2T7k?aa)J={su>ezjwk&B(uXG4`zQ@&eq%4dfv*HL#+9uSIq&V2q8sNF+Dtkvf)klb`V zV5{P8mY^EX89#?7qUoCml%w-wD!%0tNHTxFDBH6UC;#Ck`XEtHD?E!s2HG%MlHzCQ z?^i5NX;1scmtyLM8*aG2u4Npq-kk1kl`_LSCLV&upKV+RcKrR#Mp8__a>W?lHsw^Qq1-EHh1Y_d(B6XP;Df3~f!-o{c8^z*J}i+Qz@_h+Lj%IL zwLD&9dGM39nIs-&$0x~Z39elQ3ke%HbKm5(GW;oDFLXRmM=vZ3Gm~C$1009Uq8HpK zBdw$ys=@L&8W*=UpF%z@X=Qg`r(A%ZhcBL511*`8jC_rSj?h$kv$F@bd>W15c$uou zD5hm9z4#MppZ69_Ju7U0;ZLu8awGyzeu`$t<3u^)%yOjl43ZMF2HP5`jhSyT8NTsp ze-5>JbNB-FoJlXqhqqm31>z&r!OM+>Dz+@iaP_n=4W1t{kOYcm9qn9x4gCB8ML410 zTM_Er>mkqLmiF}xXypYP2K0DpsYw*J{(B?%pmHoo{)Me+CivcTxO%zh$APshxg>S9mS>-#JN( zA$F)z$)`4sOAbea!zJGD>Jl#q&$F)D23;o}9Qmv=?;)`kSHIQwh0s{gvF^(13C+nC zDC>ALN1dW4+N?N`Izg#^Sgb(iuRb0lEaEzQ>N6<(ymjiDN#Q+5PHyaJTe9LEOlBT_ zh=X5CyO$}QubU#Fz$M%&@T17tz}?O%1kA3l$}gJBjdnaaCcx$h+>d@9nR9e3^5OCaQ7B1w`1vsvQX7}jOjeZH-hKubPOH39;)@h?-tL+*3rq(=CcI>TcWKh>{*Fe+v|9TBH9XPedeO>DL z_wZ*A;GhQg!mu$g5)n|~Hu zy=RR_v+ve%xHHX|-?nDtA8~;nxMOKroLwEb-BYXG0}r_gfw+d3ek9W>-vX%F1@(gu z1E{x%mWA@-zaq^*i>ZyWh5Mu!ob^euSceq8spGQA1ci59>HVI~&}!yKnb7c8ecnLR z2+3J!q~LR$p2?!A;q>{)+`5 z!a#i8loLfqbMHiED~>riDX+D`X4?rjeEY-5w7d809vkIMN4s~ypWd_@s@De@^I=S4 z)XCYBcJG46UcMR@Ld?AZdApOJBasP}strLX)dbR@V(3{jC)y^>Vs22=@kmK-JK^uY z;FF+Gj|34It9%MzEGh<_Kw{yLOUwXWLFz|Px-jvz!#I@ zjc+r!eMhS|_&5|z_c~NThlIS0Dlv#QT=&H!%_{o{-AqwMGdtz|C{-LWd!_f&5*1ry zM#sFm`n16NVLg^GG`hdy*PYZVMW=4v%sh0dxM!yT{`?xeWJCQxNY6(KJ;oVBnej^8W$=^19=8hK5 zxbe$KCmPVOC~5Gptb}CfGYvt7e=Q3q@<>4}hDBL~wl+L+#%mAa#8hly9Wbq|$ibs$ zpoYbVvvfz-iv2<>AI5auPn^ND*KAw29S&z-I^%owgGmJO%Q_yZR@mTBZs zXk0I}DLB3=yNDQP=dwj%!+opn*05*WnZYTt;ghR=`U)&4X|r5#QWiMF2z^9YC3tw% zZTql}gQI~(^%s4A#4SOpJt}{c@#WU5hv5nNyCAf3J1ni@;oz?hcv|_@3D3x1*>VMK zAsbGu+V*#=G2k!4^8fRHoUVkKw1n`)O=-BS8HaV-@ukW{#h#0P52jE=3C61vk_ZnP zTLzIX*cMXsH!6f}F1l)N{I}iUE7j5lJf|ir-I;bVQX_aRTLL1*c% z_6gv7Gc@k}CZ7MFUI+M$uipofEE7msjL5kp!Mf1+dMCb@39RZGD>8d06}dl_A6mvA z8bNc-83Y53Ea7j4=33*l%ev7s=?H5`%jZRIpJrLQu-INQwBinlh`TS^XPRhNNmMOC zYhL8~Dq_U!ykgIuoiv!tys9VTy8!qtExH&OAa52&XC8IC@#mk6Z5;eOzgn`|DPZL^pDsPz;Be#63p#M2AGJxJ5Ttgwmp^=My zyr5YlAr$NvQdH_|MUsj5K#j^R#30fH$U`HpMM#!exiz!g)v^wX+YS5wCt`zix3uCR zBPV7d^{RekX-^9IT|D-;&q)mV0wTLlcWg3pe8L)H)>)=XwqjYrSguyt64Sa_v15TP zC)u(;MLHlZ(9D0!P9eoVUFeXZ-)KO2<{K?#qIW38=ovALZi|wZ!NZ50ojwtGI%tLY zg(d~Wtdv15ku|Lx2Eovd2v>O7GAH$-;4f{Y_J^pJ-$zlK5JhZMF>Gy8xZaV>Jaqfg z2T^u&9x|rxvl@&dG!j=Pxg#5z-}n{O!3-N(pP5E}I$xgA)J06$PyO-Nz6Lq*Yn+`L z8pIV@CEL*)8y5Hah@;wN-3se}k08jlv5(nvHav0u>|Iz?vdfx5B};BeR?8Ly6+JDB zFhs4O*##&Jw}F*73;LW;j6UW3%;5V+Ow>WS#Br` zYw27=%#@UGy5W=^%U>I4w__j)I)v2HDc`a(Oh@>i3fi<75v6ibbHqr&5?W3!^xDz~ zQGmwu1ahM6&lS-#+LVCQ-%jDDC}hYG55$bvrh9D|@>^4Mpc+-hPvMrDRSy}?PQ<^_ zUsZkw0Bl|tZh=)J-8iLD#a?d$Y$0-)(VgtO=<*4PD_POZ*@BiWXkC_O&!%|=e@9%=K7ij$C0VE_rhSpcn4{Qui@Qby}Gcz6ChumJzk+0)HUGlXREGb6d zIri%|{^Uw}Qjpv#6RkhTZet;}d_B&X$mY##L1c+_RSwM-&4~$N;Wh+WRsEFHZIo3` zp&a>Q8V_mky{F{2x~80hp}%S|J8Wnt-NgiEMKFR~@s zP)BAAVc7*e!f$AA5gs{Fb_!uUN?NgKW=UB2ltvLm;KQ};2`j>URFR+{S7VEnOlDa# z^sJ%Ngsr4LO$M3>7NwG_rQ?ZZGV?I&D+YJ*0v(P(QdzX}G5(~#SfPX|6Po##vM6DLTjVcBNV#RfMFFB2alVf$kcPW70UdZM(=SW zI8=(832#@Ko5)|MsmBlxZ1GQ*1}?1ad)hF%vuli1B1Qi#tqPuZT*5Wb{G)@UOTwSM ztp{1*`79e3DSttr?GxFE>p8p)L@pBFst!Q7;ah0y?(7_soQ5Xdh&C|-W82n1^+iE6 zGKeKxP&+#*F@5;lG9!-^J&M_SUNL*7I3s)=^6G(VPqoTh~C-`&-vQ6M7T^ z*%=f>R}8t~G518$YaY{S&92@_ z;9inP2W$D491%40PaFko;J*oTNU>q>BL)|63H9r~Ik>?i7BOCE{orx=?!bxS@z9}T zAHJ=^H6-tbw{%#eqf>p@{`Dakug7jWcd1qWD^%sbu*xo6>cJyA-4P{^Nl0;*O|y8& z5hb5nj%7wW+IJDNbYPu7? z1RW#zaPW|^XIj&XxTC1K7Q?P`Ehnsl!l+pv*KWYGFo|iKMeelP6sQ}#J?sdB7QT;m zPM*Z!2*YBH=6pn(z;7-e(CEpQ9l?S6fLI&G*GBOas_x8{MOtfzQ6%h}Wp^!j7xG59NNfg96amW0(tZrYyZ)cM22$Z~)IGQEMq^ozBWZ zZ{^jd@oHh2xO$M;11rleRGgeSQ+q82*RGh=(Zx4BXJ@tIig!r1xcOi&^^I>vNj7UM z>BgXz$FvFMo?Om~e9wuAN`8P0UZOx|9_~6P-d04e(2YDaK2t)c6-hT}<{v1@Aocwp zkWx1-UtB>K*_70n71Bsp@O0DrzKYm|R|7oL^sXiR5a4<^+5BZ$s&Y3hHUEN&CA-4X z*-R7pJHWp1Bl$5e40!r%rf85a;hD4FjJMB|E<3`hvw3nvfZ^c(olX0hByki}&taE* zh?C&BELS)PPMkv+Vw6sBQSPg-QiJbaz9t<0)DW%iYXMk)&6>y_D7J;C3E<|THBdb> zxFLyY;^_kz9;wxbCnTj@h6gGU(}Y2vu7HOdmy%&4r(t=2ue+`#KF`tj8G+KWgqyoq zQ4LRat%16jNeyrBS_AEIq>q!+u%&Aagai5zcE|29Uf8q+4%cWtP16<+P*HEI z>baDW#w6}i1E-9>4k+oyG*t^$z^SIw2PG5&m6E%s#_n7{QVE)W|07)R2n$l%odY9sVjBR_Kb|9jY-Cj!k`J? z#aAO}Yqh)KDQ{nlRl2rkwuSzP=|S&Y4NZGtjrK2``(vc_F@SwjD6gl1sDe?dD*(0S$4S^Q+K$1&qvIKp;{X1=Xd$sk5;0~~=F`97@`U2VM9SH`6Z04H{w){4-` zqL!!I-vE`(VzKQ&-*yB_|Jl7u#!zo3kXzNT^+VrW15FsV*Mz}k{I)8Am9yj_wmj7S zpLh)7r41UAx6>p7fvy)fCAmYluiCA?{S%y+%g2ah6tw;L>>79^w4XjxO&(Ohr}=G4 zek(73!E66D2`x*v^HqU!yxwAJp4r&2^eorgFnUF(+q*)?yJyK6$eQ$2F}^782y|MZ zd$6sqVo}8X)=ad5*8+X~#RC;1O9a}$%bDNGz^gt@xBoN=O@Hyzq)RHVA$Pg`xpnom zf5jUYAy|MfbhIqU>mOPxzc8q@mvQ|8<)Pt6)LavH{Yiy z$T2=(LIspJ$)>Q|@E(k^k^lEzAJ(y6v<8=lB{kg+G;s>1In~y0no|nQ-~P;x4uAIS zIthoor4SD>_=5QzeTdyq#W-Rd(DY)(67F)1Mi&Gzd*EkuLBVrHWUaZfWA1Q@Hl1aa zgjzC;hw4tn62sYoD2Xf8ZzbIrFmsh+#ojmP2s+#fe%EHN6BcH0p*=yzZVX%Wg699> zX`d{&)FaD(6QaIQv}x!7N&dq zuhw~&Fox?;fn{*rYcb&)G11EZ_;0QN4Xw%Te9j#yWeX{AWd>hDIy8tIJy+zqpWXk{ z&6;R6RVkndX!`lHYoUqm*wBA2g>?ZlA%YB@rbkL&P!!Moxl173VAV$YZEJ>qQ`)(} zKuEFq|CxU*cqyqKl->Ux=*zV_hRy7hmdjbLnW8*SPW}AF2iFA1L|xQsd{3|s-EeAw zCD@AfL|~QrFZu;qOYWdEr{Nehc|5cVhtRxW8@*K_Gi3kz{0`vAGQw@dTj*&3ow= zcP^jmN@AyOVh`#Mx^KP*Z9VJkIt-t1Q}L* z-4-@s+pwFiq&bBh6!J5q8M(0edG0Sy!~6_`y8A!S%4r|l7W@F;xqfY&`H9zH&}Bu3 zfYp#uQl-1&qg`Xu!pND{j=WjajC`AE<#){PK+{Lh8Cu&-rj8*P0g?CRU;2XoygDLp z!^(kQ`1^Ix7&ML>vHk8{eD}+}*mXz9pj+%k$H1;VeBH;b^|BzO#dEMyM^}6Oq0(EX zd;#k<4M{lB7ld-#dQ=D`!?tEHq_VAFvkQuuuzu%SPZRuZeQh&=uj5s+PUg>CWjuR% z`<^zfm^T9jp6c!?=1og?3MEfFQv-Yq%v)SiQPBre&-wRJHQVTDr@0SNrS)OzO{94C+{D$h_hf85}H=oAs?1ap_ z*G4X-A-~IOYnLM}bgY(cmWST3+TFr&wP?G=-_+9Nq!lLg ze4B$uKN(~=uPCQa>$Zv9dwfF;4NI+{@iqU64yX-rt>MkuTBvRaGG;f{#e2MQ1j2pp z4qId^me?_&=X9%56m9q)dP+s7ouKy1bzy(ti}3WfGhR0lNcT=F!o=hnWnnqGu;GSZ z)ZmSd-L>%x0P8JkwhdqZ_jQqQlx9R~_vKfxvf)L#7(Ufsq+|V#-Ey(JD(So+&8mFk zz9z+7j11E%wN08^FYUh{-9(#37Xi#g=7qT;6%?4(ThNbA+|M=ZHc)BGB?YnsOdv~Yo(M3xOR$6RV z(Q+$=AQ4`BI`%Bj0C9MI?}A{Lcb+A^trWS+j$U|Qx(Uu5zl_~7xSG++n%*|LXLm%V z8M^I=?2buYbJ_+wJ9oKK?RURb^1^I5otBqexfZIa_3$1gG0-^nzH4WV-rLHZm+$VG zoYW0I?S`3CfzDUn@YJa!R8NKuQZ9uJVM)%Tj`Vx{U{n!iVDk!FfAN$m3B!#=iqQ;LTQ9Xz6RHr(H{)BiX;5csjn|3aPi1sh)8 zp$uyHy8V9b;p_G*pQ!DDeiz{vX7V!g@V*A0bB+N#^9S4|(v(7#ub$#B>K$J17szRa zTm7P*{loAE|0hi?+~ybHdMa%LM=V_oXysAAf+i#Ow$mO!X2%(W`=3Sq z^lHv3<@?#6E4bG5VxN5N0P|!3T=4)(=_xTsd-}#r3-agwn=4uonrG?vv{vNbDgI5@ zvod%@3X0GEx#G0(*(Ch`*?Sl8D5|q@{GCm-xowuOA`**rK@EY17%nOZT?rwO$So2; zEFf%lC&`-K*=2VE3ChJgikDQy3sRx=(^73M)>`hA_Ses%wUw{dyS8aRFKGQN)!ORU zTK~`co_A*F?9OgxW48}J&;LA6GJDQ>&wD@bIdjgL8KHWi#)-6YN&iOSf)o^;bf{EJ zFvN~5t@zPDT(#savJ7qijQXRs3$eUMXz8DU+grxCX)Tw~(mxAs`=)cjW7)>!oe$&q z6o*lxH0BR_myBPAQzy$qbN+CsU5lu6UqnlQmW>iV$6~xeAm$H(GT|8RoNy~BqN68Z z9%;wL{-6#!&PjH#_Gl=GXI}Ubc0BTeo0Z1XzxaEhria4DqwBO&KlFBb6HSCpy?ZO@ zK}SvGs@|ksDv?8A)qKPvLOc7lw8mH-@pnf8(JoClo}SPm_y8|IIH5)GXl&5<`m@%K zb&cX!O^m2EpVod4U$v+F>b|sA|wGTOrOW#k;hX&_Qolpx?Cg*)90y zM_FKla%&+jskGzSHrTFsPQYBED;RA@q2Q1rR6r&4XfYgC>ab^*2h}z;PC!Q(h)VDj zGQ6mKbSbip(nX+8*|`97?E&K{(W8nuQG}1RU>Urt2)?j@PJtg+$hPa+c42<9WB z5fu^C1{Xv5+f+JkI75StuCt#+YG&ER_zIg|7ZWfb=SFoE?&kJ*OePdoJJ!H=_xkPY7df4p}%EI~u(fM5(9C97F5qUvV;)*yNDHBhd6CPUI zM((KUMXpNdtMRB~ETpw#`OJFgD-&2!YUbes+i+}%S-Aio%H}p3hNl;c2EAVqHN?Vv z@yZN*;MQbh`#Hp8^7#(#PSuKCFYijI@sute@6zZAY_rdc7oJQu^K`;-JU)PBIU9qB zj6QB);1{8kYy5iSE3{uW}M$4QRuVryd-6G5Eyt#h%oQ1P5ntoA|So6H*`ni@{i22hO zwA9P9{UmTpOM~?`AvN1D!NO%rOTz?%O{Z%0u7x|~sr{um)x$|NPSmY)`{|3VGyA$l zi=|opqQ&E^)B3taq;%`VUKCBc7%Y=}QMPEdb%Kwox6bcT`PRujs^2=fpE0M-IDr(vrg_&`_kN=3apcRG>A04N0XQn{N-?mcw?F+R5&$M zy8qLilf_fA+!Tv@h$JlCLnX6(2bC zHB6Dtt-eHBw>lH*-5RQ<&uF%KlF+@?mFBe6S=|WZqILO!EMWQ&jztYE^$X@(LzJ+B zHA;~ktT75LVGU2h7S`~@PH$O#Kzmr@)1t-J=(K2YgEcmxRji?j>|%zdMt_1VW3D_b zht1pEzGE zau1wL57pygk+_Plbzfpdz6@UrebSxy>GB~0_dm-$^ySWs&{Y+EKM`AN_`WrOvQP1z z8^1q3^zqradVn8T&e-};xrgUhX=i)3j}aWut*iA&ObX=b9vh)TnohE;z1;&}>|O*7hck2P(`hQar+EIc6s8k5)HR z#pql&jc}{TOqMsD^IA9^7|72yIc|=dDFfyI|Ih;s2i~=G)WJ*@qiM0mde8)qmzb$y zwDYqycwEO!8GVfoGMXvfTIia1FPPucrB58iy=P9Abwgf2N;wFN2XoC-F*@RpVgHpM z^3Cw*FFx>Vri#(L-e7D4DTUflE8{f%2q_0D)xK@lF zjR(ZUdGzLP{FpzdiqR==TgP{-o2eqh+xf*C@&P}9I%K=~dj908_|<+*ua9aPcxFh< zjgEvO>e-R5upWx}<9f}EP#B-~!!r&WM@W!%3(9yF3{Sw*zgRrl-BT06?@p+3C_V{4 z>~WVuoL6LQO()4L+r4-`Bn)92bPwdROoK0gd| zC)(EITP3EeG5GRK$+yh*tr33=M((z+Tu!T^KAqX1*`im6U_%NS=OocOc7neM~4x-25s@tAUqK`m*skLvdD0T|EI+mrxFGZt2U$ zjn88qDn{F)zJe+_orb7WXxDNbzMqK})d4pspu8>YZ%;rGz3jex8Wf8knn|{VPV<)| z##$)D!x=!Y#Z>hK3?i1wqv4?7DFH)8iKedd2fFawtJvTvSG)DNAF3wrLQ7YSMrP7eXC|=avx}tW#HPsR?1M#rkdgIcEy7ow+N3{g&G?eVT2&#+nV#D5 zH_RtHh&_qT1jI>L%N{T<)*==qYl#v^*0SgGjA7`boY-0-`1{1Oy&|w-;vU2HGG-<< z^Xc{0@lY(L#(i+vS>^c24;2)d4`zL> zJac?xQrX%cNxF(7^F8xr$$W>Ep3~76%^}JG$uCd~PLd6?F`5@c67!NFMM*A(%*ON& ztfF#9O@S9*bwz{v;LR~M#$+cj9v<=me(mc}(rhWcH6!}!B!+(b>^Hn%5 z8|T`6?-+rGHBEU~nRtd|O8f4cWcEpHZzN&H{+i^)gv7RFLQ&$#g#B@>kxn?V`Ian8 zv4OSUKDf9%OOIr7Suq^RaTIMiQ?{cE&a$&(R+={>4kJmn%-Sq3_9XTtdy0}=?3tD8 z^GQFPI1(?JNb4XIeRM~bc9+4Hr&as}tzLUCU+aS>pD9U)HTtW8vO2X@Z!%!pkFz5y z(V|AkokKaZpKet8upl|iW~9?YaGZouGs5P@lEkQFNm1g+k{Ma`j70b5#D+=fkmA#= zEqYw_M|?2oqAXEW-0G)BRT)8l_a*X;`NE4G_`QByytSpA+&^s{9xsI(r z8ihUB?6ZfJJ4Q=R@R^t?B3zt=cYG#%d2t&OSaKUt;;8R`CUM@fs)xh7dBYlrYML6* z=csx|G&oD`@qs$291e-&yO?+i&3+itbRVshrHu$il+TD6m!N#mD|Try*~j`5tI<<9 z)*mkLQPy;vjfKC&f@JjQhum_4<|JIBpB8yBJ~1sBUz9j9em@Mm#$Zq4#O6(g|M}|5 zE;XV1VEE}av4E`_5byP=5ii}T$8Y%I+kt$r@!`_UmQEv5oe@vJBm5gEgkkl@!8Pau zNuw)2#e$^KbD-QrEjbAX$f0#!oPmUqoI#X0a)un3FE~4!WAKMYas2SXGm|nq2fdQX zAJWwL+4!x$8h*sk2iJ7Doj8YuP<2LJ{W{1$Q5e%f#@41d$p2u0or5@{-bSrCr826@XceMDD53a$P zLh|12Kk%%1asw(e@r($2#Uojm98C_!O^_YogmdzB0V#$jukZ_VK zh?39f3K@kMe@=E^sES=!~roFsr`PEnGhIkT|+j3ll&Qjk=X(3mf0OQdXA9Cm&Ae0xh61P)HZ zdQK@?H(>IjVPZ_uuqeq*!v>2DBqk@mO)`flJdrkTM%*8vcXa#U(IMqf#IUL7T;xV= z*r?ebPvBnhU8mYiqYcZ+?M!0HISDZ#;ahC7U_Ts|O~!!8_sWCnC}-X^HUACs$u2ek z`rU+qlQ8`Nu;j%#NI1zkM2REk7=RE{r&xLc-!|hB(G*m8ddPISD=P-z4*em`%_nAk z6+I;B6Qw>B!wkDA+@mk!OX~Y&;(A1Dr0kB-pFDZ-4iZ!H4pH)%yrVx!E+O4=vX3z6 z9zJ+3OLQ^96+SqhepDqxEJ3u@M$5aCGjjW=plrue~SE-qr%|#?qKf|_0 zvULLqJTImub|q7bl7X07;&uh;fl~&>n;pin^5u}7ru>KNvME1}+*b*lldS>Ln&ia@ z#I|GvQ8ExCq;dRvtB|(+sv#b)j|NpA+&ZK@^WjrD*l`=vay$F$Hcw2mbepOa%k5Cq z`Ye`9CH67mGLoI$QXAyOYQ(Z+HBmAEtEKXL4{3vwt<73g9ltr$G~EYZI;9*2F|0Y; zvhQ(bE7NY<`?K4c=@z>UO|jeE%4rQ*?3PaM`-sg*c9u(Tk{8<%+mh`>$pCDZ&T$Lr zgj1^ZaK*Sq_v@+;I!|#j!7&?{HtToyaN-P$%^Fi|c3AmuthSlOq*Ll3NKPr{N{7yi zrHEO{Qlca`OQkdGnu(6fiH(%$Ne#NSIJ+a}4|LYV{Gm9$pPSyt>w^)6<@okyBk>|0 z&v0?O`rHXVz_;RL-0{nqXtja1-&KUyP)1{_ve7lB5#7R^s~mQ{j0H(Y7&s>v5h6~) zPX>-GFK$GlN^T@d9J$fJWf|RM@iHSOi`h=njQxe+lw;1RW>mg;Ardj~N>b3*Q!OW9 zn-qw=D4dv*6fQ~}DLjS2C#)iH-@@Ao0cPGXx+=;tcQXM?cOl}Z*(_K6-!Mz=M#}Kn zcKWwra5uIzl4J@S;JjFcSe2|IN^-G^4X+oG1~{>Sq+>StM&PzkfQG#-H)PrP6t=0( z*f>{Xjh>TYtY2Y4T4%{fZ#jW;l1!GdNnQ*`Y)gg{CAk={;g6`Kzqgnmw5S$P`+@PR{-F3u&k}Kz7(YoCr;kgCw^G#9HPlTvJ`Uhr z>x1joxU(D;XQ7Ty*8BRK0?F=o!hCc_)BRpl?r*X4zvENZ-N?L=r2cSx@|&0bNd}Yp zrzpv#e;%LeTC@ONk&~l%5$KAqGQWtT27(FE4$a%pvu4QSz_r>;ES0v@Axy;>2B>&i{`xpDB=M0v(;fwpr6>_@H2QmRlb7 zPiL4`FnMS5v_W5!y2j#PU_Py9IHNqD+H#Uy*IE0#_!bEy`Iacj#kZWXU%o^(KEBt3 z=VRUH)K02NbVbIEo8W^x7Q3y}8D0=H!fbkrYBB%*sJ|JIdnuZdC*Vhwlo|LjXLAHi`FX3Pki7n2dAlF3AgBa>xh^*O6p+&7mjqvtJd+EkilUd#&)4BPLLYRX#N zcbB16%$ew6*O*JuP0X<;1v^3;N$6e%$cw^>B}w6;#F4^f{w^ZvIPu*m=@{Ssl5vF_ zKYuCk`1)VJ`QmcaDEoQ(-o?#Woz$-t^?0h{B(!U-&Wmn|9Z9#M#F1{T+B$9u!-AxyZAfe;2u?!b zHc)xdJ25BeU6eS|yA6|GKXzMAtbXxMO)ZgViFU;Ust^9FEZYq>V@+$7Ye@McvSKlW zYeOsYmL5{jJuKW+5Ev)P8fh%$#U#Y0WD-%5i%HTreZ(ph+r_f}OtGispE)i2Q=43u zk*^NOy+b)UXlImE*1nI;j3n746*4a-Atoi0h>~1PlFH~b5}cDUtzq9EkZMyNQ9P`~ zR6ErY?vtsClcZ0o&WjR>9Z89zBo`$Tceh!Y=fM;Ea7Q$ui;tbrr!IYP^ph+OZPUW} z;5x4j-8kdm`VLknBTY6^8wiw>utYZSyjX_Vl`JDl99bqCx5L`e137WjNq!b75z^Yj zYFjwe-l5k7{P?P*EB>CvV%#~hBFHr!eT*Dh1_|YX4rDoXkjP4N95vx2Oz4EwyqJ>M zmrN;29GTJy*T1t0YnU|CzP@;s=ju>goe_;k{JIZH|D46ZIFHK#SHBNy%2VK{L)d4{ ziP=V__cF=UIZ{7~T5}Rs&k;jj9Dsz996*#fa)2BKc+RQ<4#b&UA>}2l_y*8WRP(_t zua#xqS<4NX8NU17Dt)mNU4g@Q`c~<8!jfYQkc%@33n$?q1H_aUM!wX*cBT4VO)kSjuEI!nI?H;Xlkv!ox~d>*Tl1IB=OXdxV&a&i?} zmb^F$2`D*=DEU`9iY&(PDE2>2xw#2_=-LOv{snGAsGPe=Z7i#se1g?EyNS%h8p6RT zCpVGt!+$-mM~WHB07%5H-13J`k$8IE)qaf0VL25i3(AZ3N;;y7B`oJicQ zyo3cw+oqtuoN74t{AMM*A7 zG#DI50&?O|FdAQLUxUxL`5-vdCS;^&+X|oSo&ASMKN$-SE9<*ar)De=d`cuvBM45C z)~$@>Mf1d*qfmIax&anf}^#}e!V(HscDP6 z3op;CRHE<~i4i>YWWz>YR7$K#DitNUtCR#;OOkTpfI`Pa_=D;`xU8Whi%WBdmz(rE z!AObYGp;}%x6Ck}!-BM^A`@9jFr0)5Y?;Z66^KE}3Zlf36>Qmrwdiu3*bt*tEf5Wc zwDz+vOGLGrwvZYQ`rxjQvb?vCaoe}-w_l{#R0x)nWFLF@ zyx55tmh2=-a`9goC5|j%$K_wF!f>}RKd(36ueGc6nKK{M?Q#(CjPLSU8|E7CesUFB$m}GCm1S2; zox}*PcC_Bdwnh@B$p)Sm6A`4DSWC3FrHSLFJA z&RYO+DC_SvL?>Tou|{J`AZxx2Ym;$I!aRxUI0@xT^?A`YF(heQlsMA1#NOjp_KlE5 zAMRA~V~jpHvNB7kl7=I>S%2qiQ$}pd~G*&mAg;GTuqVpLj#oPw0+2DD1%AoVcTCp$(~ViyoB zCt;_o@Od#6F)W!%lsGa~R(5}D6^}d8l4Z#QA^z47uc(Z=INSMX?49Gs;afq;GchVxu zx?lJEKs)GQsl~2gt_v;?-z>TMM%Rp6?38au>Q!uNB*_$Z(0MToF)JBHl;mO& z>loE5nbk-i6iqnX1#;d#p$#}S1{U6r!wPeqART9*W;J{ zbk|s)b*|hFaPuW4fL{yh)}g2_fWKpLbrlQ}ztylBfuVn@cLS9B11q~iaa9jRRPbIi z4MF2-SoJ4FQI#kXpBaZ*@h##|qD95c2AD8?95zUF1p;a!0q2P9YJW&q;|VwyU%de3 zF@IdubeJe;V`Vs`b;8tSUiXK>&?xc&Xuc?j3sI*b3#$q05E+ z2_1^)H+}&>qpp^J9bJ&qX&v~fFk3nu3<6NTl0V?=xex^o`gQ-xL}(2ZiF|8U8w`rq zVioE(5n7|dP$7LF98IV}7$!7_YL0{V0@VP;86i|*zl8FHss+2DR-lZlfmPiwF_~Ej zQ)iKs2(}WMMJ5~#bi#_;FG7t~jLs~|W7yP4>24$*DQIf03kKr@^{0F9KpPyl{#1cq zH8HwPs=V;eat{pI@UobxrMq4#ywAo5%QvI^!UyTMoe!oXdhqtF&Y0fdc=O69$nNZw zBGZporpq)Hb!Q}YPcfgftN}QTYYQnR9k+hS&LyNvk7Nt(N_r9fHoZu+Xt*=oT|~|C z<`-v>BR z#Z{^CC>72w4&6!eia12SO%JkF=#0D59m*E)c=NACBq#fq$k_Rptt^MIt&vjQ4Zr9! zFWwN3N8JF$ zvig%GjAyG=Y%%pz2kJqGo-6$vco2)1?lPy`mvkBWZMqCMIq@4-Wr;~ER~>Io@dmPL zIEBdAIR$e+9wj#tdx05$R>woSYJMs)_j%3Vx1m};5963sQrI%R9ENPa+I}RnLfVaG zt=my%5q;>l=}H;YT82t861!Fg(BsY3yd*2fACa+hwG7hSf~}3ju4g&ik-NKH_AONX zaqMoUbn?j+o?n}_c5bzYWD{;hzfHHYG`s$r=~2p3eZ0BP)g%+UkI2}$k41u)u!WJ> zWlZmitBF`tOQ>~PaKJO6@BQa-?m`)*g+J8e{5wO`O4m*LAUNy+lwG(O{We`JBT*mh zy+7UMvVb3NF8Dmzn_W<3>|8L5M6*$AMoM={`pj^h7Od~^YXg~@`@W4vIgVqRMKxAg zAuy#JhIli&qZQUJ-hGGU7S2e&?VM3YGUvgpt|$W^Z;q&vyzGb~(~l#{^1OiUjg)jl z?D)9o-X5(ihcXpR?8XbfMFO-hVuV{Y&^6{kHm9X6^lEhpn$=Zg9N$ z+mqixx!K=DCWF6amZ=_fXQV7Xm%ezK``m8Ecah)YI}WC4DphcJw9XAfroO6>(s?9JCVp_ zcOnwB6p+c~6QYTd}2)*n`ZY7Sb!Jgg*_$nO1&M_BmLc%`SDiUp@U;(y!>Z=~vk? z-TZ@e7t7w_c=NWaNpki!k+Ji(>_UBtNQ}gehlf3C7K9?|K)kH$fzQJpfJMY~mpc7N zNtdGErb}6xo%ndV|5&PzH)nYp*)^O+WbB;9BEbcy1tYO%&?B$8Putf%i^qnJ+i5#b z!AZDuZy2*X=?(PT^ag6+-IeYKRB^m{z$eJ4;Q=CJ=K;j~OqAJ7Gphj#DOaSJ(S)B! z3x{+ShP?4gAsp)X0^lvpp>>5&12d!EKqPo}AR38df*tZ2%^y+egux`;{98o+&Rr7%3&)&_vDr}nOb?!ASnNEDZ^s`c#dv(H^7ax zCC(bh@em!!H?OS?l(!XSxDIg4lxrOx2(nj-Xc~w{B2f*LPm(RO!gpy#i$B}8z@xHN zqw!4)qiUKoU8RTEF$>DW1;iX3_)D^6ZP4gX;FmK&`LV#A0&4M>8a*nI{$$<)tnj_xA`I_u4X^PGfFYWClJT4$5;fq32zHpPi9V zi;hne3rd?DQmT(eV*a?lHLQYik7$EJrKxoxTMMy|s+pte9noL{l%qnu_fM9XN*4&7 zi~Q<;MFXTE`U{@EPNOEM-~#pZP#ly;g%Z$+-arRD=DHxWwr8n5pu8_6!3Jr#g_cOY zqiPx=t!gl+20{5*p)$M;;I)fO)0@ob(gK2|xm}Sa4V2r3ul#a~B(^Ym5*Q(GO+pS2 ziRh5VQL@FnE*+G69W)W#G~s(4nzSJ5=;uPQDI!}f(WS-wfzAeO1IitWg8blP?Oqi& zD@!{{!K=ev3L@~f!M#dDo1_UVCVEFrJPOJJ(pi6yogpm*ozqJ zY;X4K9j0JMF0%^wGa=y5l%Jnl0tZeq1pJvP;9(N5?<6-=;&AmOWeWzop^_0(*!(pq zXwqOej6`i>&;6D1rmqB`tZ#iGREEdFFR10x!ES)!&N6^O8-)0WQB~W@0M-w7!-ySa z0N=tNkCp+vIM@xv`^o^CHx$wsev=EJ_~TN54O|cvkKGf(LyYhj08q8L4B+pG2Qh#7 zWX4>FKM?cklNmEynS!~N3m9_|lCNf!P2el){J)XPn`Ho7PIf~P@?=2OG89fa5KGmU z%K#3X>_)5Fn@EfkEqXkpwVU=?(P_2Qhl*0I1AM4pAnZHYu+xWZrv&_!ew{MJ4MofM zdf*r{@fW@#%YQ|dZ{6#GYla{*CvHKXV=6JeDqE|5veyIO9%4}VggQ86F;Zq^PrzFBtCV@|tUt@5t?9yR_3&(~3MH!yj>j9m4dY7*#15VP4Zz}~@ zaiwsd_pxLAqy*p!hT6baRKE|WbO6N%O99|2(V$QPs5({(Fz6IETp6=Dh_Bf41Ju=! zGJx@?xP=85oq|fss)xB<5upN8^6iXLM3d`XWW?)TXvCgV3>9DR!l@rr>N;$6oj9ok zwjyW$*z18tZoZVSsDUB{fCM%wXLBX8{aq5c=$tZu4lWM!72B4e?5$-0KPF+ac_Vde zgZ)^G6dTwGZ`}pxQ~EW4s}D( zn0+4D%}nm+D>C~flEt1!eBQ(4s#&vg_o(3RQ5OzOza45QS>&No)!_$JrBt|~cu*-o zEjz|IRLYt?1Wz8u0NTLzJD8NPC1xdr09Gry3&daZPG31t?y;&jYLFfYDCragvmg2%#at|>AC{yeAg9V=lkQSZm8}8xUK+Ic+Qs$m0njs4^w5ORF_DE z+BKX3FNm5nj!Cq71+8A9_28ETtzM#a%%s&TXjOJ~!5szYtTTqW0Zwd2r7$*FX-H6Y z`F;*BRFF+C1PmBF`^fXIhc4UW=00m{5r7b3YI5$y*~rx3D>$h~^H8_ITK^<}V| zN_xC*C~Ll^5GwTu>?zoFa|yiUMQ$Uok8*>CyRnXf+TH@3>oj7?Dp9hZO4bi|Lm6C4 zNSb-l^y31IX+ILMM&bXX!j&Vig8`J)Vw20@CE~<=h8xO`VF}}Yjc`A72GNI?iJRYx z9C7ms)ek)rMY>L+-A82Rok?;Sv|p#fva^VGF^&<(t`pe$K{vTCcv)==Uxavi$ zdSt%`{;!h9*1s^$kNAo_ylafDGe`2+ddEm)rkEx!jA}2Ux_++*b|DbXT1MdpBu#Ff zjtgggk*~<`egl4Xl^Y2DYJ%Uj-vi!+fUo8&!XH5R-|hFng9!iDeh<86!cSnhDSVaS zVU!yPK0e9~P;_K{0nBE&1$;&LgNXnA^#!mQ!ALakMxnshunI+c+N^5ks#lDvg3&BR z*Jz}IQoO4RDgMD%r1D#YM=9<$!3w(sSQ%d>uxAi03S0LTM2pN1W4JTxq1a(5$sI_LqYES3Um8wgwGI_f}yw?(8E2+_-RI^olR zxY7h`0JxzL2dyPz4AqFd=-Y!B>RYcq-UhZ;jHk%zJk8!=46 zj|y=$Wx?4P9$^<{o;ceLW!@VKakjay5J#pnYux~4F;R4gimt&TaQ~x6Fg0O~xK`?L zxDbcA*J}ymlayIB7TX`f_By;+h-+y-982{tQKoVnju|%+#Xf-^OPaAnZ-d_#;_}eP z+EE8|B$b!_EQb(v?Cx{XE#^x=PQDJ$pCwaz~0~Qf#z-@$^yP3 zSPg;=KH!1#Cb(h5umc|0JOPc46W^r_w}P)GJWhf?KzN)4tLM0ZCc!V@51a%yo$H1X zI0@c|KX4K}<~#$Zfh%Z-bK7}5vHk>q;KX{#L^lx5J@^B0o-)Y|gtKcBt`wkL+iw(c zbG#wTvj~H-G@S2-V)Wl*H;UVMn!}@O?>Zl~OwGNUQksvN%&NG4GB-c6DW&-vlc_ms z32w49N6pkvVJRNRA1KAhud)>P;186d)W=e+^`R7qCzgumPY8n=o^t_f_+k8k8ZMs7 z8g9oQsNw&biW+{q1mI`vxQF?QR;xnIe4!NJj5-6%sD9cGY%_r&SFdt)A74#uDCq}v zEU7w;IeYW!CJ~xf_%lvjd{yA??TgsYHt=UWRm;Ig-q`BSxjy%{y=iC;147> zX*QF)Yc`Yn8IyUMuM)WrXPe}DncPCu4@%cEhspgH{y?3dHx<=2=UB4;DKX|hZe)S4|rhwd^Z$d{80ha z-Xi*411;kb@QPc6u1w&|4B+h&aM6_qJP=+dB3_KIXzs}<^#cby@Xh(iz#|7daD;JA z!K!Btc;KWKH&nfFzysH|xS{x4w-rF`t%8$52y*zg0(hgv4I_@+Rsa_-;Hhjc6S;%0 zsQ!@%KBNp_%>tgvzPkYJ07Q4gtwI))u4f_BjVxrkM^fqjav{_G$|9!wG}Al6SBdW2 zMNId*iv-<8w->;mUO{&%a(DUd1@H`37WWuzRcKVK!5Uj7=hG;s;B zlwmTH-0m=@EGw=Qf+|~20DzTB^(ugsxW}+$i7_o%iCZE#<4E9-;Y;06*7x;7sKk5Q zE8?{!(6H1E?x%SU^g3TF1F#M&Z(WKE6`gd@13%=d-F!ugt8vjd1GiP9X#=Do9*@RH z)zHP{H+Gek?}rqnS_ODW`DuG8Jh9ZxyBH7Ak}(u-DFN`_R!9UN$wE-v#M51?f&Yj^ z=P^Z*_eWv@_@RP3%5N`qLuDI0u5@iIfwE=pOfu0%QKT!OPg5IsQQ4%i^J=S12Nk8d z4d9@Hn@MLo!f3hiFmN5Y#)|VPf%7TCIX^Q_(z{r&GZ3t}&Wh-Dg*@eT1zn|UnW1oz zhqBmhg;3cI$Ef<}m%5=+rA@Y@3cAqym*SlB2*RKb{Z&CfS-l(=5PD+pDXqtU53_Ay z*x*9(k1WStfb$^lwStW^`HDWI;9VNdi9R5(x`Q6LiK`Fs6~T_-U3vRK4-8r1h7sKd zJ#aq$xbmO}maX6s=njUzo3GgJUTlZxH>|*Ti=GU(3uoGjRTB?-;C`%v;!OuVP|J0= zZVyF|AN0UrR^2eZBGNaqxFNmbs>#^kH#OgUy%1e=Ra|Oc1#rC!E%3w&!vfd4P{WzL zPuNv6SMd#$eG=-|UBbA8daB<*{kmA)%+Of{>K0WuksaG;<@Y^-^gTj)t=|pR-N3nb z{b+iByj|7d&$O9&PHV;vBYgN9oNf^GmPwB6sW8cnOi zqvSvlM;Pe6Hwo?)LFOJJ^M;9- z*K&(-;n#adA?|R|U%#Yi$G|||D-+q{LRpGbgL~d9=#P8rDmo+d-Yju{9dZ8^RVeAY z0$e|zYmzb6&+V;2wy2qfyZZ*#S{eHR?@YqH2UCn2Hq90< zpiLQ98*j>NxI@hB%uSgN#v0};PJ4J$CW8^oEd0-6NMhJ-vfOtw7Esp%~-GCD+~Y%xBptt9MMR~vtmZr z$5(}Vh@w=-0BBCUqTLPED>-vDQi_A3;HoM>sqMhhtj7UE(-8aW6RyJkz*;&&Z8|BXBr_=U-v{t+ zWAZ4m-A8G03UxMr9^z~*8jrw`x<3`dq0>*NUN&fRE@z9W{Rxi9U{6w`@t_)?*45Ug z#z*19n5quSb&4`P-U{&1pWNUSS93fXQ{#F_P1JPw6Vp{Z1RYI;^bnra-JsB6;Sui- z2ww^^JW?0~Lgo#@J<6mBCE(Ry3!i&-DtbtZb?KvO5~@C}M^_V|3{i%MG=Sg!)eXZF zdK~C0nNG#88l9rbkd{#6`aIMw9{b#FQR(oI))r*}G}uEM*WOmxt5p2W4c4}Urxm_5b{`~ zyeZJ)u?Ix^4k1Id@6j)(Xsbpy=@^I-sy;ocsh~vY`0&gRod=F8eAkfbkF@yP)KN7W z-kvJe%J4)2VEy0RV1^D2YEg8joLRb-He2=M2iE3BwHDnU*YWkq%8j2X#AyrS{!loi z$vAJx9cNl6XbX2GI_jgb9#9^(#vrr1d1LTh5rRVSJ@F&nrKvG^M9HOdSo9O=3hSX3 zwH=RDiNS=!wNqW2p<+39VNkr5FcHf&C#u|_CyiJGVf=Yprsy+kVNsR*y>Bbyk zc(hCB*xd$j&tG%ttUL_+V=+|=vVFd4vA5}`hKY<}!5f0Q+)T!#5}lz~T}ypa6DT7s zgxIl$NBrHnxmc>Mr-O1u&bqcBF-GpXHXRk+`B`v<5>1*K*Fm|;5=si3O_4%cNH-^; zvEgDeYTN@kMUobef>ClI6Jei;xC{U`DM~eZ)g}e!o)6+J)jE7~Vw1S5s*I}%xRrjt z*6xO)m*4QfcNxKMz9O1lTtOc8rU#${Z%!QwEPPO~XRsAJ!;l;CZf5p26MN@%7&ok& zi6tnj-AqS1?xZk!RfikOR$xg2M+vx#rgD89xNffn;cm(ngpfg}9Hwbg%rXVb2CEy+ zvaXeBYyg(2QEM) zpHT-3wFEx5Xb2x0X;s2Itb}(+39XkIN_dBqa1Z7~0Pm0z-k}^R;T=-K@J=_BO}(=a zgxr55xfgdDa{q4*!GqOA{r|8;kP`XR|zq;@)Z?}H^igo(KRdQ1`y09%S+} z>b51MwTD$xpF67sLkh)$O)d^qE(}$@n&DHMT-dQ*#(W516K(JOXVeYVq(9h1w;Vr@ zx}kXHQ4ciVEi_n$+Y&b)^?(+0L)BeJJ@6R**ml$d&&Ld%_A;;=fl;R)$0VKBuatD! zvyyeXbtMKFhfy#s#)z;uze`h14fKw&YT{1T#GRyxAFp&nbsXSM7v9afslK;Vc9COI! zn1tYRJ8{{Jxx^ayCULnX@fo|%(rJAN(Hx^)UsMLiqZ!ADHMRkLz z8|l=#W7=ufOJ&AjV*6MtGxrvVE_QDLcCo$8%e}?Z}-vB#E@C4>UM(%aYbsD*atAsB-M3679a>K}0_)Y-B1?0haO?rMRHz@5MlTcW!9|5jy&L=jA1{TP)$L5&$SBt3=8}W z3H&J1ZPobsgrk`2gx5%>)4NHg-?B`7$Zi6DM|Jmi(-RjWbAAs>vjxYF5WHDHx07Xq z?ln~KE{Xh5kD-G9rTi~3-->T(P=oLu9oIgFc(zkJu@Du64@n;cgWRh^%qL z8K2N_Qe6DD2R2wCzxuWZoQD7xGY zRad_4feS8o!-yN+_CV|9$ehEt*xD5e2j^4Qp5u?17VSMxh^uVHA#t~&jKog8S9$LS ze75%3R+^^nqXGU}Bv-WgZ4Z3J?f~2vn;t+NQLRy0W~n`hwN6EBLR!*N+l(3^GuUHx6Oyw$gl^7hnRxtP~H9Bvt za|W^dqWVmhrcSo9_JKG(^Z`-%+FFcPYXCkVGCx{tDDML+Z=9*@fU9ZbdkJ}KgKJ!i z-*>}5*1BP246YT!jC(2TVLx9-#^7%P7nIe01N$^=BqXP=%fWxmrJg_szUf5xh9TFW zDTaJS05Dgn#`XERxIjN`9S-LJbI~)@b%tCub73U$iknIg*YfnalD4g zL}-l)%dVJI0)yAP|3BvbJie(a{R79J+XcHNy>7_3%oKELVW6~th}%e;rZn0nu}J|P zof6Ujkv54*3I!E|BBG)e!JVZlDhg^{M;#pNGVUXF!*NChMN|-_?zsP6@8_Iz?@e3z zl==RCzt{H%-Mu7`<@+zbjk%b8b&({=_&k?x3iG8 zUw{$k!8;47T_QIE*|W2da#A)LHf3ia{S|)%b{5jY)IPWy%dx80l?ruEREpti>svGp z9qTO`mVDQylw)tuJ=<+EOJi&w2+b# zcAk6!=Oo~)Wt;(f9BS}lx3WZ6nnmst7P*ftGUP&Kk^6*22AX`qqz>z{2KDwzp`yli z|C`|QH|BDJ!R2p)3%;0SJI61yQE?MJ%Eh*e@DU4Vr^gH;@cM4sG!Txwp(ahP_`zEiW8xk^g3geT#VMh^Vbay!SW6H zL@46=2dmB!)Fq$2b`)tMNVjGQu?MZu|EkVW6)?|!6 zU`yV<7{kJd16zN*7`Bazzq=IogR~}lm#@3r@}=(zEj%Vqe(4(w#ApRqHNX4~An%#4 z{eug-;3xd9ro%yh=YHz#EAZLZEpXp_+RjAxzY;}jvk>E7uogF-5B zn8ZUMbU73H<1!mjLD%hAA)>WBNxNhj*8c9`YMnfxy9-G8?zbDIbLjDmQ+M0wUtG#- zV-wd2H-CXom8ybYX!>{@(HrNSwACTFUu!ztgN}K*x(j~ z{w<#{#c&MEVT$Y66j-6@lJa-~g4thLF3o=63LCN6$FIPe6}q1)ub>%|PtcrzilBKq z&~hJ#Gwu@RtpVuAU4_&RQkkC>(sTGD|Fc5c$;c4EE0IqDG6b+)BL!%_2L1xUE{TJM zM<#d`KrGVMdLdGQPdp*gPJ|FddKPF*fz}I=J_QIO{S44R^F>n>+18RVRHoQERd5v< z_EYm3QCFz*++pc-Nyb(pGL7v#*762*!3P+O+hDxbj*VjL%ONJYJSh zc)NE*{3KynQv@z>`b0#H1SaJZraB!~hpFykd61C1r93kLLPGxJTA7glbe)YjAzyl3 zT0-_pD@>J7aNqZY?3i?lVPV6tU8f6$gn@Sm1Mgr1e|()X@D5?%VoU-*=f8h{y^V4Y z`mB)j4Z`oUaiA}!>9aywA}OwvPgtP^reFUVIv5On+9l|E`Gh`4TblMj!v_5YPCoOm z?kY{DO|xqBmC)!b*66(Jl}2B&7P61|nltQk*JEOZ-X$hW>TWPo^*5ZVFTUabamP|H zaRZ`sKK;mCw=oa;m0x~$Bd&1(d=6cii3Gp&MvMmvrrphb${Jo&yZ1)1f)}L@u_i(} z4}MliIcEsH3*-}ap9{P9!0sYlcHS+#wFKVc#?d27t^}r;seiapHV*wSg6cu?30L)Q z990rlE1$rzHjZ4Rn6i5W>+2EqNDxQdWFyX5v+xIUR>w^$4}L8v6`d)#Lzl&#2Q$lZGp^ zo8=Qy;d=l}WKWZzS@PLifPqG~NQg#ktC;Q#3JFAOU7yM6aD67yVfjiG+v_<>#Hey3 z|2=D^jS9T?BJSt`j;kl26bC@N)cOHvZ3e&xD`fyI3J8`<ogwoH!(q=MjlAvb! zg!tB&9cai@N8>G};w+*PT%V~F*pG-t@uzE}p4@Gt_-Y$rp==a7&T9-^_?3#Iv#AWH zkUeW`R9qaNPUEqD9>?2^B#nOJI~#4pON7cWcK#ZF;|Rw%P&>H;X*3?BQtV?bym{d3 zQ>=hI5T+zA3R^Wefi*Y*8r*&}7O03$z>%>RZdTslSJvRq(BSx6qy|;c;4Hk&NYY8r z;10Y*=+Dq#3;u=%mC#_stx|&%p+WQ7ebm5}z*r`rj#$dW;gd7lW1>2xt&gR=~bX0=Qg;DhHZ?{pw#Lg^y?>Y|-^W1@* z0TjeSbOClkF1`bOH7&s3>+Vn;R}BRvT!nKMxFntSHwh1AiR#h&72khPx| z(ouKXsN|l{3#sTX8<> z)p_g1lIf4EPUm{1>W@6nC0mWGo08As2>ryU1snE3u96WmqVX22QV%aZ2X64i0ue*5 zvV_$Yof+JSIy2Bd92?Mza5ANnH_+$fJxp@;yc2J5hJ3HS+eRY;NvEe8C5XKRcVqp5 zG%?$zQPJ7>8xShS&$c)mbRQ1^AHLfrN_?T2VlCnU9=i>{$3{md5_EsYG8|Jc=;>4o zTN~vH_=K2{-YcKQXVXJm^`m=Wq1@--#QTK9k3sKe+g(Vx_u8oB;N68(0JuO$zke@! zNX%y@-Y;;fH6lxoec8RT$L^N&-;hsqD7|~^taF4c1?owZ-eY4#nUugb`9#h*4NS#| zGN)*FAuW-xrSb`!2H>2z8!Mdy%_7Ezt~&RrT)lO!5L6_OJ#vq6PbSjG-S@)hF`_-Z zclaS*rBnCt#^2vMRg?-05{FQJ5X0tz)crgn`h`bCpLb#vrUyl#c$_ZD!hxmz??YkH z1B`b=7DnXbP-JVKBgA@HKDo{st}_J)?5Oq!grk0eJ_obg95W@(KFAWvG&{TKNP{ z9~mY}Mtpw*WcUs;p|4*M&008x@s_Zr+^Wr3;E?%yadbXB7WP2 zwCq!(?9s_zK3BDdWenr92dPI z5uWF_yuYzAUuF}G!b@-ry}}^x-%z&`Ikq;37(~gz@H27Hl1kcA&7zp zA5ocX=^<>}eevz=^fsbING-IA(tae1(thL}du;R;m?wzZk3<{!0nF!+)_x?j<~GuP zq~X?OM?%7(kJ-q!L26fYfgoKRrkl7@)nhPdIjv;enUBFW)8Lvg-Np!)Nrc-Op$l(O zy6Z?-VjEt9V`u||2K^m%+fg?-mY(9W;eW^W06oN%YVi^*r-vDI4c?;k7*l!#FTpXi zkwKq>(*JHp*+`TX+ZPFyg=4o1Yi?(2?!QqqlI=&Lk&N3Y8p(E+R#31}Vj~9Lo?zkk!MllAAf5#mjUe7Ufw(6d_4*bjFNCC`o^WQFlCn zF=UjE#!Eq$jf!W5sRS>hyKGcY^l%mxx6lX_T;2t<<`3+YV@sgT`cV;D=j+UuyXcW_KL950>;RW>4=^El0F9B5o2md>@!iHme=HFV!sUUe4NwinZ37 zYu_4ZA`=a$!CyVW=#vK8_Ra}n35oBNKzDTV0EL|u3N(e{wV}xWShKo0Ff(D)?HXuk z*9F>q;&WF+(E(>CceR7Ciab!Gx~r=gf1t9tagGrT2AbL3)u##$1NBv1pv@l-L=sr| zbIfXu%r=a0(?Esf4_?LbG`Rc5u>w&FqvWX#8XTfK3pYA*mRuYz+%p?%yX_ zW}CK1dA)24`fgurIomYGUxfgF+{gO8tkL~HS!1>&W0+&wN+HcD`yfr+_+%nC^ZA^5 zw#N1UOMOoyh}fQRm7i(cHuryUEW+y`A;TBn+i0L|BVw3>ax;q#{$B_Ls{+XYY`RHr@i60)Nt zFq4RuW6AxMQk^L0tx}zGB&a|>;e(5?57kqu(>+*+$iZ(uDe#b9Xhbw<&}iLYuV1ai zC^>kvPOo*TwV9J1!f%5bvX_itCB$-?x=~nFE;5K<9lx3{6s|rXoKK}sDst6WR5QuB z5=Nz>J8?iBCQmPg5 z3G(IGOouUJz=A#7B=kJ_Ota^=PvKM%?5V&@dZAf#3MUL*QrB8_I!5Yr40Jm5X^T!= za;8~g(S?qIGOM3f2IN;NNMY5fjgIdK=hK@{V@Xl}8($dw=g-6Wbk;K{iHQ=&pU=*x z?a$!HgHQmoRzZT6e`2x9*0Dm!*eRdIZFCC28yKu&wjyCDqu(GZh+iNHNTmX(^Z7q5 zN(IpTc0a2G4?qtpo)c)MS-4X+_#CdRofo6ZqPq+}jK3(D%t<%)!ZNxVgZ)*Eyn#qK=!rPn!pSm2i0;1&VZ`>s{_=YH{W(a*XQSfpoNaYFeIh0I%1Y>{ z4nWWPLmmGxzIiEee>_W@LNh|~x$!`(++Z8qsMu?{nQomrAI)N_2O=(xaO-s0CL0a= zqg$uD*-z#V6efU4{TaV;;kVGtoU&ejB zDW8hF5hRz7*JP+eSH`A9w(@%i7P#Cz$*RvL1BM0d#A(jfNe4 z&;j%u{wP1_0NM$X{rSQZ3q=fBz5lSve7utRc*tyf)ke0hQr0Yq$TFV*nS-w?nf1qn z%qJ+BZ+TV9oRYO4l+QGox4tT6&U(#8Ec4N?A$jy;Xo%FjZqR-QGZeHx#wRRpbp(oYUFhnvLE-;h{8!YcU#LUQrj z=k0GO$sZ6`W*wcR2eoaVVW<%MdlnT(=pmMI2PBTpriU1<;7ufir;pdEQwoe9Hs9Os zaBnMPf>l~(h?Ytd@dr(_jg%$RLk8N$}CrM>gAt1L3>PFu8t$PF)hm z1svZAIvx5J;wOKgNuv~*!xaf9!+~Zx5w701sY_QL+Fq8xf1{eoaxWV!{nd3v=AbDt+w6fGDAX2gZcT*eu*h)sQNO*g+M8_>36XDT z#`@^AH{CpdTlY13kpVk5kO+puYK6D!TG12z1RB&-Fn}IL-FuCkPlJPpWmD=vkv@7h z?rZn5m+rvIrnF=PxLMB~@gZ})- z%kuV`@%oX%;M{qLRe}cYN*Zu7>^dnygOp%C$?Ga`c5*WGzU;p?D$pP2?qwPR?C$@H zvrB=F@;vJPFEaN;0X`exuOxUjz{B6+(|u7GvKD^_$Iv+dMBhQZl&Ci!^{!@cIh_me zfAJEf3sCPD{EhQYd|%?YcfoxuUBYiu-bIpnZK6(HkBdYW!m#y^6*|o~K=cm*YE__B z256!LU9Lb~0PQ8&_hV$4*_DY^_br^gLUevBxZB?Lu5kAX&9t`sS{>%ThE=+Ai;W6W z|6qHs;j3I;-U5}&se|FWx4?3p0^v?ZIAE(lxQh{L@D`w&lR;N*Mcp1z z_ff8UA0w30V~p?v-lFsmu6xA$AUlS-806!(vGPLlIqxgOmpvh@yH=U-1_Kn6pa&Ib zqXFuapcfTr3qX767}5%%i=C5gvBznw4B>TwN!6xF%jUV4}44qZxOAq0T_HX$iD> z17VC#t3t7)_&JCHY{?_ZmSXZA4u+Gpp*f*uTK=hRASar*CT}?^c0Ui)B(;PA??AEZ z8GE3%bGjLQpk{OW7=57rclsH9FXM)^aH$Z&>}9NwhVBr;3>si$RkAGm4G>$<&KV%~ z6oH;QKy;5#G;e_Dl>$9~faq>t9jjU}K>P|3V739GJED;tS5X>AcRUPN-? z04vqjdl+9IAbx$Vz~6s>__g&;rayqgo!olsYX$y+1ElY*s%H8J4G_P!-o^L>H~=N( z5%_}#NWa=A@DCXvzPC!?4;dhSYh-pLIyYiEdH@GKS&?YWa=dAv-foM=ajePw;hOB< z9^t2@ji88MoYVD3g|NAqMcc(x;go?LnL5~RVNqHJT+qvwfznMQE^Z2VtZG6KI;;+Pc*&c=#S5Py&s)P@F$9*bh< zSfbShV#70J%e6~itpgHW0H=XgRy-RI=f&^dwT))4W}kq*TNm{_MMiqSRNE4dbCvU+PI)D~-v z%na2=gShKcdsEX4N^=7Vr+_vQlM&4KV!t$74KH;>TVjEDs3Owjn;VFcb}}1fZEHRa zjW#zCwa>|?q0mkIm6(f9hX=~*{~NyYCHZ%Gj$^stELe}|oTOo?NNBDT-X?9fHZ)9x z9gw0BHN({VYK}B+Yz6GPvhMpz${J*#BTWvImmE{c?+R$dAcv&7ncg*ICnqzc#21+D;$%t4q+FTVn zr!~|X8eZBO!EZ^ER-p}TBD(6ld>R@{#uG%ZpOi;Kak`awYBn?(2sevA1)+3d$B!Fc zTG7-*+TVnzZ?xpo(56sxG9W1-?lsU9wR?*pgs3rvV9um=y>|oXyFzWDVTqHozL3WMvqLhWjym z#7$|<1$`q&@&1h@4{4LMz*D?oFC4ZZ8jy;V7`{#(X5N~ zf`%vzey7<8dRtr}%DCEKVv4}xcAClIi(L$3VAE-w>cGJdO_Mk!BOK94)9}(MfpF5m z>0T{PDTosw4sB$R$gw$+PZ=NJ$PGD>!)iaUQSR^RbSmr0;&TjDSe-exPN$1Lz>2?+ zLDBbxXJP|QQSMk{cQc5MKva=`V`d{`k`7Gj|Q58fke_E80)YS{tFZS3kdI0 zggYhSGZo>&ZB}t3QCu&HJRh8GP)nszou`dj5Gx~^_aBVw3`JE<88NC(ds^N3vhMu! zx;!Xk-z)U)Ta+R=VQ{WCvW4i}Z=(4$W}A&h&JN8ZIu~7hcpJVi6(cDw5W7C-VlYL; zAvz!Z|GI59{2ET6iO#$1oP0X@LvcHa91+qvjC}Nhc(0P&iFG>lbcrRq^I)&D>U4VH zLmLeX;k3hc8|6mobSipMpu};AL>!aIxmk!mO+KN-1pxCA2d}KK6dQ>+*>L1`9H0Ta zCRw3PK2h-^#kzXCWPSd28|8qtLT|R9OEJpbI-Mx3I88o5Z5gP!+&Xn^w^7MyZk?XS zA7{CBI`ktO4V&TC>7a+r%)oxHwF+F%BB?vGU+{4KDP2EL_+^=d3y>pZalL zLHmADvAfG2NFx#_hC#OMV}7@otGE-VXT0M#WOTk6DLfK7myVdQdS$ zyP0qjSkc+En<=dS1UH#J#f-mZ_$w0rHN(I71Qa-~x}Fk2vmNe0TphGSI``BaGOl_g zigYNB-_+hJEWTDg;n2?zR~$U+Wa&ov1e7ni1pSDjf7=cjJPN(jf-ZT*+&cBh;;glT zC#c~!c}kYMb$Wh>jY@8E>+~zA-Q(8jr+&~kL%HwVIt|-tqv4uICo0MJ=;Yi9JLDYd z(WzTfe@8ySF?`=_wEs4Y25GVSG&^akS`rhN+w_$L7iSJme%5^(xwX?hvAgsK+ncgz!r>2(!aK z6*nYZo`Kvn|5HnZiy||^FBkjLqcHQ6e;3DSX$GSG@Ljn6DjcNi__cnQjdEdPuN3=t zxMBrnr4rb5)h?Sb>nFRAMsmD5O_efElTVoIaP)kJjz*)dh%{A&nnTHup_qP`Rl#)` z+^pASpjqdCrkeFSROAupGbG z#jWTWS>fOEiHf7Joe$OzE7qUwmaRykzidGlVi>lT@zf;Dtjyb)X69!y__+5oY%0CS zH0v{LXb~vSu<59v=$R};%d;FFlX{fbYb8amd_wYSsLTyil}@!lPk~;Dq-8y64fMaD z*<1c(d>+p)mEhj>!os2K>tSi)nzPzPDWJB?ZpFDLB{sf|}?vzc+5 zzp#;r_V#B)BRxaO`0y7t5%Bb-4ESSl`4k6yk5uYw`Go5DEqhZ54tSPd$(_2_D*p!} z&OhKdKjTXk=O1LCDdUza)qRPQN591WWIQw{9BoaICJN2q|3K(gjPu&9A^L!Id;d#r z*Ro6dN;QCO+yFdZ*~s>eH0g;OglNSf`j8EJC$OWl=|iUbGI00w#+^gknaysAyPa_l z_!_ykpbJJ0(Qc-8%-5nR?q-bUuWe*I^KQZMJozl1L*Fp9t5CmS+CRBd`<|=cFROpg z)!#(*&U-}lZuu;pLqBl!J*ZBiA$R>#cr*x)qBR}%jcmyE-^hmCBT+;P;)Xm?+Twco zgllodu4t_*Wof5;0vb{DTNV8c-^kXg(AQefB`@@OAIjoA@(F4&P+RWSY122d&l~iu zjkrnmNHiirxM{N_I6rN_HtpNwDKcS}JQ^FQ1^W z0~8QRvwuVfgZacPY3Zf%3A8VOhWW&mKg#*UefR_OiJYJ0eBw_(!5nEWYzZe4a^X|? zuj_uHhVbK$!jC_)AKgFMC>O)UPANNBvvAkE96b0r+XNPD$-p7!gio>|A=3J~Fhw+Lv)bp_R)IoN#bx29eB==%0 ztX{z@QU_>uDwy~}78UdIa&34M!`-*1A9E=9x`vs+0 zx8$L%11_k|bMJbnvg4)O!QPA6#PtRxBX=EEw z5IN-2EISo9Q}oDr`Q*&Dm|qZ_6iOO|wx$uhNMni@X&^Z(+m4ZR8EF^s`UFT`qJid` zY&#Xl!cBAupAUZuuWhu1vA*xiZ)PaT{#LwFPgs@c5Ij4W=axZs8VSnO!K4T8r)210 z>Qu}JGjGu@E6bz0{h%f`coKB0hMk;?xr82MDL3wCr-F_b`SjjxOl=EF=%p-Zay#Rk zlw&8`)PD$hV)9wsMt3l`X*t~R3MQgT8$HB@52LUkhN3V%%5)2I?Nrc(l4g30>7Jcy zCtHcsvr0aT+vx91;)-0Rn_Gu4dO3^WOa5W3>8I(mE!R#ZXP>6ip?P*17C%j=lk)6D zIY3?{C0r_>sQ4g)2oZ&VTQK8VUNCbsB49ojzfmCgX+UwTCh1;hp1^zR*rRFQ1;V$&@xtx@VjD zsT=$IL@D`1`Ghg}_hR7h^X<|Y9d>Dqopu;QWhKLfFov1;$_;yXpH_AeX{lXG@gdBO zv}7xkX{oFbgfltaVc(;)%&6RM5P@*(GR&P;DgyDl6b1dG$Ii!5@q3by|;#GG=;h zt;eDfbJTsK&e**=sN{^f3TT2L=Y&IWaCkq z6uoHG2)E@2iGE~_h8&_o;78U-hQN<91e)k47X5@n>}315G)#|tVomE8M%-|S43A%! z$fiRev(Kkf=PN?yObtUik>h%c%F(Jsq?fFsF3DiyFUf%M?L(BPOL$#Rin@dsvfzeG zGvJ2uA$BU*`U-xAXbfG$MOUGyY?GMrU&r|W9U>*Zj%oce1g2z>r)?4<=R#yI>5wG{ zqaL+av_xyr>4)Fhsern(Xjoo@PDdPyW_MtNPG?HAcKO70 z=yQ;HCp2I;r>a4xi9j|SW1Jm`i=AL}ZU(c|%gQyNbAdJzhYBvhse<64cK#*&3ou!` z@lZQ4_X3GOQ9e=Y;&kqh_s<=(XoFqHzpQMp)CAuvH89?KsGW-AbS+PtMjU1ri;gQL z!yfqr;}x)UE{syvoy8^Bri)T{n6lag$*@U2QT=+fIan=rn4Q>g%Yon5aKdhe&|R-s zg?~;8{~Qbd@L??cCKmp)!|YTrt((WwZ*uN9>To5{Mk&Z1`Gi25(xs0wGN!OoCC3%= z3FNocUiIxjGla&$iAnKjYfK#cP_p)HQZi$;V3am;?o6UlhhLaaFC1>Ckz(>pqcCne z@CXbNl7VE1#-ooMkN3GW0sZIcN7#w`{PylF{uT1^u;KTo>9qC;Xy$3qsZ6pOC!b(g z2T$d+AvR@cjeG)n8uY#y;oGksRrs((DNxpJRb;AAWGWPS4iczrWHAH&G^)|7*6L@8 z>Sv+)&Lc1-fwP3})1mvg-$N_(dojF247p+KZHW=I!%O-YF%;=&BgHblxjOFoNZbZLgSEuXr;$fk<7n&7J# zybWN=OwNmijxPr|C2iRga);uH~NtoeaZ)lDqgaS_9nB;jRvju{%9g8 zl@v)k;N{f4(hT7ssX?i;KZ;nxj3|nDS+-Y-I1)uYl2Gc+e$?I^$)^VuLA)F|bZ=^7 zQOO|G-5)_LCdG$Jo$+$bUJ0U&an>=Y(dK@{L_2##QN_!D)2Zfjc(!1P65T~eTo9>Q zil;t_`<@lU$xk zrBaeyd@hSs7ihz-L?|vQGD03pEsuladKx{up3si4;;RX0pt%{d%S2r(s+{_}Is6GF0_>`HU(dH(uO}Uv9tx%eKfcfzNA0fxhHq*n2P%ue@k2)sX zaD)EpzI;a@DP+f6Bgt?}sMHZ?ZidR*Ga1^@#O!d4*gQjHL87c64k!0wURpAuE|3h) z!lwtmS<%+!CT~0(Ns{(v2A|D;_`-Y|8g3$SJaBKdTR^5hlxS^ElJ;&twTr^^UIzSL z6r?Q~%NFAt>;EZ@Gs5?(Y5(enVDA%-zkQ~(H=B4U7>Xo|{!%_>!dZVAcUEgSk{mzo zgg$63twrorUTf|L%pXsP%4hZaal}KcBxsBd{NG<>{QpzR{tvrCUUAs+w0wFFUJAU-=37oVp*Z#u1S=v2gFxFDC&NxM8t8`_je65(@kBjAegN%|F8UtE@t z=(#Ej(bJa_KB4gD$rBu|&RqQG zJaAafKM>WI!;HuraeS7v=d-n;vqW>i=Pww06ou)9?EXcY5t$=D+}66%E8?!7qA)%A z+ZDla>=@EsO|OX2T~U}``|XOPJ(gYt-@21|`;$l-rMQVwCI7N5suREt){_`l1d_>6 zJVNv&e^mYp^8W%XAU>%fC&=mFlNxUJqT&{MmWyv4hEcUh0-y8f<8zcuNB-W;%M!Im zia~2UDVGj6qF`>giN4^0_n(f!ItD$@Tmwhhsi5pt9+!W~;M@Lar-B$tg7h_iG43tK z9u7((bOYy?#l?0iZV4o2)2$isPj@kVt#7blXlX?^CL#rf2Hw4I70w1I=_R=iBdFCg5Xi533L z60CR8BbkuUTVkgI?`y2*gN(ka#7X#P zoeuw#q&!ig;8)?8a`;b@a>t(}<Hlzwt^rbb(GiTZLmtI!Ht- z(eQgOfXa_vpi}ioJC!_rfldoRDEGAsbn2FLo`r-zU!c>?0QGGV-1W(fq(Y$Q1Hn;j zt#8TH;QTG@{C|(M)5uvwTiA!=MyXKPk|`Pm{|@%n%>6IVr`eeHZO=(&9y z$V%$e4y*J#g!DUD`j1P6^gCF3$7nl^MCj6vObis}k5=03U?Hh^Mm*3GO3?1itxNLg z>Cw=qbq4KZjeZ<$r=w%>(2Q^!ea0ok3jWG^u};lMVRE}F=-~r9E*r_EJKKcH6F(GB zgvf4K^t1Cb>A10W8usJ)nRFrWbG<2@wo3Zn0VkQ#>9MhPDp`=y>FDuxD!DYJ(~aZp zG;Bpmr`N{Y>Bv>=hjl5P@{aA-4>f_rEW`5jg%Cr;&$cXWWE1W}rfpfLT%J$9W0je= zWg!tz@htj~|GMs2q=d~`I2p4&>uhHp9eJFcY}0lLa!L6tox8GzT|%Ppc6rGt1R9lpT%?N>nwDf z$1-@&r-Gkuit}w2)}hW|a7==i$Y=2!+LMJ2?0g3QB#9nmf`77PLhC2e#m_ri+=mi`~6wKV&mm*>+R zz?z7XHo77k7W-LX#ZWef+Oz+4K^_&IjE!8Bw9&F`7;etV%&{9~b7)2OUl!!i4Zx!R z@)f7oWW#>D1y&i#=Fs)ozh0hC2Ta5w7E0RahHRK|)y+5Us_Z^X_5;(H+YD%Z`E9_MA zc1ovbE9^Av{gh5W0X_HAlujv$90hXTLY>A`+NtD#g*u&EX{TX_F4XDON;Gy*?v%)_ zKnCUg9FnrvAt^6*DCDgY`CNs(Q6XnnN#ql%6te!VkokOtd`^|5{Fp-SF_0%pvX zu~Q;Pos#lxPEbB-p-wT0JP*i@g*yEg$fqvU>8NTu4f8G3>5OV1pS4h@4vBmYkmoGa zsl6K6W}!|GFb)WIOPsmD0m03{0l_~^k^~zjfgmhK?+Lwb0WvIh`6MaoJwS#pK4fHY zFO$fNfeh}4)=2K&8bxrL#JLeTAeaCSoO^AJ)cs`xxm_Z!R>&DH$^DNmiToEAC?gWO zB=UtoMkFi)G9uwYAR`j;P60BE&?Awr1u~2<@f16;5rU^kBivvh7i|$nxLP4^QpjH_ zDq(Le6+R z8Xu0EaiejgGtkpIg?qF4WZJ#i%P!8R^J-O(a4+YneugO}Bbq`@ST|0DLJ@;j$Cn}% zYdnPCmCZYiyRtEITvIE$`d!@KHya#wWs7}hrt<~A*y`lIeWz6XGboOuNL6((AQri| zO3+ss@*9wI0qV~J7CCF9iNriZNbgrxVSg0D{>Z|vt`oxk$ihBZr_Avq%SXk*KyX%w ze#$<6X+C{eXQ$%eELt?^^^@{wKewH1onH#dyQDda!}K#J>66^(&s1@i6abl?D<=Tv z_|UhCU>cN3#PM;hTgJy0w{-cz9(K90Sliq8!Dyt((6L>z(yU9iSpGQGBbV*)67i(1 z1LpT88Rg&h7(b3@oVPu@Hk-J0U)G)!iEKkuSyEau!WT*=!;zW8OWlE%5NXr3p$Vcpm;PT%K0J2S zIx@|rr@2I!XYHLyL!>DhY;6felK2Ad?EaWZ(Gv3FJkqMPq2hL93;sS{Hi^op*EVVS2jMH@UXY`vx5tC%Z+?*0`W~Vh7|gDR%up=JNZ^=kmqWzN@J^dVS{pEziw&2)nEzxz2i-@ zAHAPuKYP1Qol@iH?-y zGO5^^@(HGEAT!J~{xrA%nYl<>}OV95SH0`A|fp9`m1RK#*UqikpGI z6i5#+(!oA-ap%I!?~6X*K4^J;yG|GS>@@7wcAehC9~;_r`m^6o!yX5&9|m~7U8jxT z3*n!aPZ;2SIHzIRZ#aj~9h4bbccp3nwx;!Cut;cbdV5us+J6+>H68%GXPE9iHr=oW zWxDq?oH?fA1Z`nIFKDn+&a`DZeJHsXNkR~R3k)!CnNANi^xqm5baI2383rMy@g=ht)S3Vu8MbTp$6c#ZK&E_o3pF_dt3 zxr}K9&ahK)AU;#ResKm)z-;5$?~06-SLV}#zkm)(&ZcXa&O?8J9wbaTM2cG^pD-mp zvMoVF`IIsHOJs7F%a?2l#pAoa)Kt{*S{R!2E%MvK8TWiZ#oAysp~2g~^Cf(}$ssO@GIv9Nw8ua6U- zMOY+(OXL%ql=xQ?ZO|!oCQR8k+$@5|Wlsr)icycEwNkU3Gw2hv2{Cjt(5Gc8yFvi$ z3=nb3z4c5vd_^XH7FLZ@-xCLY=+6n@v(A*HN?H9-M+EIq)bUbqR@&gWpMkf^1#;kN z(CzGxK6>QQXG!VtQlCyguaD4T(vh24bfDPs_3?&qAGAm z@COc=Y1{T?P!CFih-u`$97-oela6R4;)=|O8jMo%LT$tssXz(tv{9=vHC{HPo0`fr zQx!w~a55U8!LGV&!+)uL(t_)o^Afz4HsF8K4$#DPH&~J5?Zv_0EzZM++94Pg7M(-C z1H_VkP_5JjfmAGY4pK3Typ^Zn^Jbzv)ZYf}c$+K_xt4O~?2U z$FIL<&li2TFP`W6*|P$PS)+sT;Mg&x4KeQ7wbwLlXfPfWw{ZPru zHeJ%pOs9!qlA%_oR8PfOZ{(5@;^4>d()tj7N?+mkWQYbK`pB8u$oWJw8+O>}y}ybB zlrzWOn@Yz8ur5>L+HL~b{dvHWC*=6%}%)o4H-hQ?V`qb z)F>S?gbD(7Dj7dy2vy;ai9?1^O8_xJzpZOL6i`2qp%m@OvMPLq#u{ItLFE?$O3N#F zWF|X7@e5b@xG&V?h(_knRXiR#XgY=r@i<+{+r9Tp$D;Li9wfB$uB)LxDacQvtgA^D>8*->*N#dXcd%ydX-MGkey2YwMwVu_~WfrIz7x) zd&Pw5W~g2h3N(e{hRR)b!9e)>9-;a@topkltS>diNxMfoVTX+lnqj9=P)@stwWH!0 z%`<7Eka-3+tLb&N(c&4{k%3A{dR)W$x_2ff6)6<)rp`+m1{evH<~~3CF!FvZV&q<^ z_V)0@==fQ7DtUkSVf552I}QJI_+dmPUkyKuvcs4{jGT1@bz}>{*UBd>^EigRIhzj{ zOr5gyW%&fO3&gh{Fqke4+bO4P`Thph^AanYi0;iMDse2|pEiIM^_uFu;c6B5n4w@WN1x_ z*l9RIi-?8XC9!+t6SZFIC1jVB(B&cDBjs6zL_lbBz5Pu%G)2Ik(-*Wqw4dh$VKnmJ-WQqOVi+Rv|b^@}&_*l!;MHVK3j6l+N3ZrFX;}u5Bnj?*N*_-z^y$%y&x$%-7JS zGU_dC6r%iekIbAOj=8GJ`S{Tv`h|^Q*yUP3z^c%U=0GyU2syprGnxZ4{h>B-5VCh= z0Bf4!j(CvkX7p3n*_Mn47@E}&loxO@wCRtnBihmuiU`hnugdNy&X{3n5{`<4puH_p z8A>F5u~0}I9!<9wpuH*N_QH`%mc7v~=!I6(I%7slAi{^zpvr>0bXLxIobO*ww6@^K zv!`2sv#<}Q>}hf{yR3NpZABImHRWjLAc_V%CTG(`Ibvzwxi14U9)(e+Nvt#+3WemT zOp#P$Juw)XpbIQYaF9iV(*nMdIA#2}N~*Hu-nomF~zjYNh+0Yx9{t>r)!UI<7WM8yXSg#iB4B z3CBVLfl)HT7Y)u1G&RLtkx+8n@X~~QCGBQS8=43ZJ$hX}ZjHV}3teZWonCXiwdOst z=0C3Inh*A>>5RuCQT(v5)~4Zx>C0B+W4w%CmQ~|mX-$dh54FUi@j!eYX^X`T#;9T{ zAqe6ohiGe(w9ER#34y6I9w+Vk{=k*1IV9i|m*-O1y`qu`#lHXcqL1>R4! z*{MMPG>eMQrauDcnun8pD3~3ZM@Iwa?(bh;4X--h@^;R4vHzXT-uARTOa|35GUJ_If ziu2CJNf5{vCLb_AIv1TAmiF~sS={qfV-T^$tCQRp$-!L?2j%9737UKs5*d3Wcj#0j zi#y~Kmc%6u!?)gn-{3y5!9-aiQT8VitnihT_(baP^Vn2b>4vf{Mk-tN7&g;#w)=HT z%J)a)6Z~t@qiYR1oqV2MeCRg)Jo%+rn?&0npO9q^)HY?gSjb{H%@<0#qmk*j+r;4B zSs;v}t;Fw_hP9D1nux-%>bmpnG%_Obafb>L(2GSK}pdq>uM{9;D>azk>Emkkf!H+ zK0j*1fuExn$SIf&LY=a}svp|8uQ?nHSz5Uw+iq9N;&Y}l-XSYP?6xdWO;$5T_ z3^JyV&90*%5^l5f9Et_)uLOsp8?i$eAlj&HzagLUQ%Z}CR=yjpd^ZX$#8Uj@8jM|< z!qgif8>VM?uzpht1~b~4DO6P;8E6hiW?KYP0;SRfdr=DZqV~a!`ShwG*o#)dUbG7K zq7cmF8i>sEM^zVWkS{yPs{h-XHZl$aysbU9GM~OS_`I#HU74@k__iQjK%e1*_ZYpS zy|yZ!{F|UWk)&zz);bToSwx zHY?~rSu94WjH6cP(+w929t$%tROkju!B&Z}B%^k9K7A!IF3rG@V8lfrViD2YJ_ zZ<{E^yVDGCSBA)PckxC<-9>g9l?ae_SB6j@y|Q+fm`xmw5zmGUj_9LdwDlQ>uFAJ| zguZ!+V6xfZkQ!_?_5GP@BlZ4FL`@95#7ev-O5_ypem=2F#Vz6K^kC-QtMlpli{w$? zr}<#-^hI_$Izo@}A$E(y1$!*zjU8h!pLU2uMMj^Ojn;N3nV`N^UOi%&~6fCyWC?IPk&{A?PL2@E-a(m40 z*5uQy#n66UGL)e4NQr9}v-)FkB+{ZicQNG#!vK>&Y`?8(j6bptm=d8#(+qr)qHx-O zr-VF=h(>E8qpeAFeWQ^zkG#ZABf}Bny&CUPJ_i5?*@}F+_YymY)y(89D#Q6Hi#}rH$!FIL0gDb*g6>B|5k9jP z0ioBxkr%xquq@d+rAqI}XK{jJfP8Za8pxon>vV%y&tZya$v5kC8n?twB|opzY4#G8 z4cpMy6AjA0Q#Y{74mWC*+^N%Z2J&K!v2)7r)Jc~}CGrWWFG25Id#6r2m#`?nQ^dRy zPg}_two8=+mkFX~%dQOLc$Y-qDxVsgWsPjoYq&|5U5cI80MQz5(lbGlTlE^zs*5Ak z!EO5XOYzl9GEN=*vmGy7YVF@0(MSZ}2`WypBb4?~=?=;94(7N^alC^$9(kD>Bi&(e zT*n-zUnV)OV~!VJX5~03l#GQ!akMCdoBn$%*Lx+`dztGxutDm&m$}M#yH~K})O8;- z6{)MUcj}UvG8#*|B1wZ$DvjYIlHnuF@TJS_G_pB_-};j!pQ9uiORD$%FE^8ms3WC% zL`oH*M>BBXnvR}>k0vAZm=Mu)GrF%((vXAxU^T#3ncB#vXiFd*A^M8H#u<0Hokk|- z#nk&0yvO3vWK@;RhfFh?ND~#LKwo8YyEOSuM|df2JqFi7MC9TDoD1!I706?POXR|+ zr%Sz~ryv)4if!P>@#bO{X!@t1#SU6(r`!g#kAn!?gdfi$&C?Ifj{0%@Y^ou!!qX|7(XVsVNg%~VM9kdkJal;#YyQb_afbZO9? zhLVO5-9xRygiL>gAkmMIfZf5bzrO;@o5CAhP;muje@&$EoAgCUNER=n8OZQWS76Xu zPBS3pJy%GhbxJYAO3YWUNHf|@OY2cGb`7)2+OCO?U+kWI4bjl2Q#%iY4!Kg=l5N+n zHLb~~o!E(kj<#Dc5Qq6x|9r zSgzG9*Ryy>#IanfS*{(1T&qpF*05X!S4p|NQm&hoT%)f_lglxyCE8?2LVvVM^l_#( zG8RbABKnxe4E0ypX=F>ZDTD$H8RlQ54D+#QN5ze?GWsMFXa4THO1ST{OxWp-tN2V{ zc`5JJIDmjNg5{-Q>}fjSZ^}*HsZ)oPm#=ZX`c9p`yUI?(ZoE^cKeXFv*e!SJw6q-) zeYBt+iOioEA!a`S$ZSDk0a+=vHBGsWED;7B{J~B3GNnmJmI%X8Io+HE(My&|Q%>v= zEN30~U|5=^{WnqjMy;0+}~@ai-(cw(W5!B-z{<-aIb1=ONkj7b(> zZ3NV!T=2gK1+h@Pg=kSOB5XSZLZT3GnJE_K8p2(c3*io5ZWpn4aV|pZWW2`904&R+ z>B|Qgdu3k;2QSVd;@I1;T*lrT_ye(5u|maON+MqYWJJSt0GY9ejZs7Wq7hazY|a&- zx0!W*Yz2}d(dJx?ao=4b_I?CR>bqI$i`Mm4E>ipn*Wg>n*f$ftGJ zNKbW35w|K4U%DpEQ}r!P;kY5&v_Dy8+?c0awUJL#e146JjE#B1_z>ptJeYg%wW6s% zl?V36T{}Q?7kw$r{by{CqN$&Gtu*%{`~h=+aji1KdVWSD!>b!oczLZR7)3f_@c zRW{`bb8q5PB4=M`Cu0=9DGyu8SDk(U4&QBZSZnnf}7 z3K=@IkGAS^yG?|_?KYTl_Kl*|-EKpxyYfc3n=}EFZob{3;vF`$x?MMlR(H1zUN5{! zD10|7eELn&>szIaol3^@Zc3A}r8UwLh#A7g%B(_ev1#1P6K!E0Yi@$aLp)EEMNi$N zoW4a07%Hc&EW>ssL)q8D=^rQ=a#p6v5NZnr=Y$O2)5@(f{A4)&CwA1B6DJ@;GC68q@PTYerAzYtdwR+Ns)e4BK<90q?w^)aBhvp*7Kj0T3?R7g`Dc$wDT{HpYr8n$w*MwEVUH`GeU4Ow_s2p!A^69o!Qm$?( z*IE0?4)ys}X{M^iMOTJ6)5coG`ou0m>JygZ(^Z&oCId;LPwb1<H)Uam=yO8h>M(k_!`+-tMCU}>wRlfYweQAcLN!%b-#`@-DZkK zx@atEXx4R%Rl_6?SCVk@hz@QDJn7_uIu!75TD-^bj%!XS3EwZ5uRL5%Vif;yhw?^J z7$a0ptx*2$4(SX1Tj2{_JjU|=lrC?=lvl~wJ>Dwsg;={Kx{yckf4JF{Sd>(vM2IX3 zLMULFFXYwTCGwqxzEDHduD@AHeW5wVzlg{9uiPxvLbK9-TH20We z_Ir8EUUsYGNZ$#Uma8$l@76SqEl9HTItnuyj8#J?b=AJ!P511EW-X9qTwcfy55ciG5$7;lbV}2 z7O%ITb4(tcjv~InU!a^P-z5qhzvedOlj~*5Lpr{J)A3Wc2_@EW2Hbv|RH8#FF;r!V zoZDFm;{#7<5om7aC80o56YuX(*NHiV{c!I|;wjExXnfB+Z#w+xc6r7e`#YE24m__ZS#PL#aM;dWs)3;;FgJ4;J;#>TCr8jz*sY$B~d%%6KwI1SC0R-&Q82oyhK737FJe`#p3_4 zT>S|aZsDUDN}NK8ViVL1jY5SdpON+bd^L$&fQKC@m6?neTtUrjziT*iX36mIepo9BsH;Tht9mq|C& z7RHyhz%(l-l2QJBWt3WIR#Z1uc)hhwe?{%74ATwqDykbD4wv80#d&GPH5DEo7iXsx zy9MsSX@wq_mr)P0mR42xD^7I@^xQOn(-8>?g$mNjYh6smW-WAMFR@ zv`9sDW6kNcF1Jvz(2DJHI-L{B#*~Ybe}7B^)p?CNxbRR|yrR0%<#r0uad%=`sc`Y{ z)2Mi-vNVUJ75b{&T$(1i%je<3{j8|96@r&Ntx)N5Kw7ECQBlhj%mAsVZmg>i5|mhR z-4*^iM*TxtX;uAXExEWXUM`A0H>*~bmmf2xOcbXUnFgvXSEZe1 zW2h`2D}sBmDN|+nxN)+6SC5tMF@or|R@4a+bBP%_mF34NPAMy{Dtz8jf6Ul2#Vchc zF-DbcHDj-G%ml?P^}dD1SSi8=v!PUuouCM$ti|ILfiANRR!X(Tm}Wla*zuBP>M65L zR;t2(o3T`RtkSAuf?5CAW7tnGS&Vn=c*Qzpt$v)MnX)3gJdWzx244+R`DD?;ltnX_ z$1$n0)+s{cqeaV77GVIZsueAHXTP9YmrsDdUvyu}%za$b>33B#$4?h6T-b|hO;x?4 z&MQfO-IsK|&oznbe%TMq@2Ycp8U)|Z7d_cqVxQCLbIPp#ZQtrXe}!LUfZdC3?p4=S zSLdv9RmkSrvuIh$Ovlj4TkolJ`9$N|(Z`OyG&`nS!>8=Hrw>t|(=Y7w@uDYt@$(rr z+_h-o!gN`DX*T?(FImHeU-bjC+VHc!e62S8u21#J&U&BABSPkjbPIc{8$|{)yJRk_ z6ec(G7nfF5v6C%vjrz5&N$xu3K}+B;tbuFvDs>Aw-RXoX88uyyDwpUg%FNc~t!@-8 z!(v}=b)&0Jq}+6xF3GAZz4#QNbgx3^Y5hoW$|TVlq)TwhBoWW)20EoqbT;OQ1(KcO zlRbGaA;dr-U4~j=ncjuH`wfs)Nx(8t<6@aNwx-jlt2o6|-)M~lOA7UJ|6SQo?Q@+W zv`wYbHLI(b>T>&JS8Q#Cs46?V-W;8EQ#|!mB9PL(TIX_Gne}4ja_?PI)uJtZ8>EnW zcl6S$K@3*XdAOY-UG=h^r=A0#m!ImLwayBku*<{OS@MJzBMA`)=IDe=r7YI&(_7sr zbFU?vbD?mqrK3S4gvUEP?DAr+) z`%;$v$WiaC5Fwi~UF`6Q4$$Q7@QHzzwWjY_7SIw*4&MY(WLnctmlz``Pk#cAnP(SG#Rtrp-7ma%m9B~qoY=iH%++}sJ_}!K5oKTHkCQ> zWa!i}!t!P?p}4=Plqj#N6`kWJrmGmx_*)RCyS%JU*e+cer+cg{Hv*G)cZ<9HjgA_TGA&N` zH&(esp+yjXW3{VVc+27he`BqjE?c6*-&ilRvqd(4qtD|I9i~MQe`ACDRJUh}7}{72 z?{Az^BNBu;FL63*ga9U46gm9D|CXeWV$tADwNNJbn_6+1Ks1#?p<9qN-G@T2Oj%}6 zkakRdqljQ+Z$r!>I-^9e6C4u8CLWA zz)tr$JPop!u$WT_@2>&Xx{CT!ox=cEBp>I z9<}tp&dH5Vw^79sF_Rme9x;2c_M#}Rua~2bR63pN+KNeH;%ym{IVW?uSWik>Lw9nc z$L;jhctrGDB5-n}$5W|B-o2`LJu<1Kv#57EoG!1Rl+MFUe$i!EC*#hkej`#XV~6UhMn|p3IaMx+m|m-{YVgL z;rnV=t<&d{lAdG)2E&Z<6RhR680q+(71AW8^Qs#a%;9nS<%HLalj=rC&8Y}1k&T91 zXjV0NtHmhE%u&^%G)>#5t>9{x+m!H?Gzb<48nL{|{inI)R^3=#?{uEwRGH%#iwxBb zwY6qnG}Z!{~k>X6xbE)hi%;AhI^}2-mPxYeh_V`?r zye_Y3S}*p3s>E(ir+}*#jEojdlJyms|EU&-J3`t@rqy4!_IecGOn*MDU)R zZU_^l);ZZJbo^HunOaX}MQx+oZ0ke2doz&>4amt=@oTY-->8bX5e03^)S{>)DcdE$d zfEr<`S6@|8HQ6O+KO@rtHBLG9AC*>KE7@Bk#^>}`Rya;|xhMJjo_fLNS*!F|RS_{_ zZBZx{POvsZpHl^zB?*X9G1f81RH_)AkXnN>#~KVxkHHsu38i9ikVQqG(-J3UC|Ob@ z6E6>c9|>8Z}s*+-W3!s^Bbx3am_ zc#}P~R*zdPIHg`yFyo-6!RrvVw)Q|SM}=E19a$!CE=Qerd?Q9LQVDZx>vGh~;g1#H zIn^)MEG>!KRm~~2uA$a1<_MO#pn&;}wv;AQb#-H9y{DqeQPF2g4Oo@4vSAWi$>mKoylcoEXf$BaXOSRq3;Yex^L+fz|5CYrq{x;_5-3RkVO z-p4e}kqK&fDja^-WYKVXnb7NTx&0pgpDUW3J}O$*K&>9Bu5PSvaJyaZNnFViya4&I zFXa@=1(x0!Ai1Gpwr*FoyQ0qLI)mw0nfn{dCyWu+Gvm@#?Qg88s1@^6GedH*IUzpM zhCF&1x*pjW|w2BB^JHR{sy4hF_UZM z5b=_9?P@FJs#JO#!f>?K*g7}+W+tXqDa>6-6xX>tY9+|rJ!F8VN-ox#qYRY0)$XRb zC5vK@5uE9gdgZobL7Hp56GU<~x6eVtTTw5DsphsU7gvcK*sGyg)={){;^x@(+g=d4 z?Q8CRGyQsxx2{1XFta07kjbuke}kOxnuBD9`n{GlmceN<)ceMa73s{}phg{^T1GQh zBe__v;F+7>T&h-H%w2gduB-A@O%<)n94{z{%OPo*LplZ-Yd%M(3CiUd4{|~ zDW?F|`LfGhD=Ap}Ot)8#dM%CD-N=K(ipCm`9H;#H>#vsdV!+km39Cp3&@HzZQmND) zOVT&tz8Y7x@TVmUNN}Z}5br-aq}?mv?2Og?C@Q9Twge2T0-$llU*c z_CwrZC2mQ)vU-(;_^-cOR)9?eQ}|S>ukck?`j)on=3ycyUANQkZV(GmsnniLmOSp} z^Ml{F%5TZU+|R;4@#U>Wr&NWHJZUnj>omU)eWJF5L6>zh+q z17sZcvILSO>!_2}ES;GGd21>;$yw%YCd@7Jzhy;3!K!5UV;M9`idB<@h@1PGq{=nf zDOX*8u>`cL=y8id*1`-k^C=j%q2w`!6Z=4&lBFfeWj$w|Lt4`^ic&DQqe6)KL0?hT zV6(d)klg07bPI|ix_cR9@As`(U0*TD*yp#5BxUVMQVUC#Q4r<75Bn0GBo_lLBT8AT z##vh{R`tv^8UrRb+ATvsRmaY-}2*^l&-B*B080~1qsOSh<4xof4vEn^1*CIZMZWl#_S z=!cKj;gPaicE-$V4sl+?vh-%coRaE${gIV0OaE(#BJ-$ab<~7S79;cis!4BJhWd(P zz04<;v@b!Y)}QY6d(4#*UT&ii8<`aNlUa zxeKu@*4KM=&I*zBEt673u|`IUWtgu(E{E{#mc9z8cKS{~AUT${jP@i!dBVoBN@_rq z6qW(Fs-%*ox%6&;WVl*}&8pI5uS<-OEyGg@^Z2~gs_DGlSA6dz(T`hpc~woXs%DwO zNupktsP%C_1nULJGR;(#*!Zj^_ji@Kl_@@=Ws~Z zT22ZXRUNWTF6x__eJYV!CclcKQ<&5;AybfgnKu5@k0fHaY#EBHN|RL+c&(pWQ`AO< zWy&Zk`W<41^j6<$e%anE>pQBFkliv-G5}M>$i*_PQ=s|=w_K~VOn9w8k!37X0R@!o z-smf0gN!1}{8d((>~hPYU|*^+RqAe8IrL5*-#BK>abv7QGpj`~>D|*T|D3)>I+E-c$X?Zk08Qn_!O@`&A;(TYD9WQSYy*6622{mO;zp@r{)( zch%(a9(S!+4LhtC)YIUuYIKR1{n_e*$>YTal21;2ti9Fb@r_l^>WYS1zq!QSSmW^u z>8$OxUrb}MsxM|ut<1D(NoDeQ%QC37l>kudnF7-_)>ZgKKDsJ>8mW+KD=M8h#U-&) ztJ7w(j8$JT$uCcV?CE8T>M7Vn<{5@{x&r~G;J5%otUZ;0@D44*tV1gSn@Y|)~V{VAP=|dNctj#LJXS0sb)0R5|DQGO_92To(?`7RKlys%z}YX39n*S z!MHfD+^>(cs4cdru%ly}`?q4qUZqD94h~i~(1V!v!xA{r6ZM4TriDJ=(}Ai7a-$R( ztHmMF5@;6YSEK_QqctqHO7)Pl{V*K)Xpi%%6&IjSY5+xpqXKe2xTt&qj>V$^xno^a zx`4-LM9hR(DK4VN5wCCr=fA}D91%46`^W8t=_*DOYJ44R3ltzqvpiOF0Xixvro!`86-@Zrg25RfGoJ8o1FgZ ziy}6TPcFaeD_%C#cCz|%k_}gm)-k4NBZ@v;>yBQZLnB{Fp-&`{Z){2Q;0aE9y$?}mgj7)|IKSBf-!CUa72ZPpgiNVRd)nhM;XRbJD`D(Pi}!b=0X~d z)zx%guEA9o(+3L;+%6^1ojop1CK4STVgjIT%|;Sozu$^>zR?%O65&=KOh}z|)R*=Y zOj)tMu&iLVCF`pU3aYGGpT#Sf!=kmFY74=#YJJX>(4$(1sptW*9(vj%7duBiR<=RP z$1#dNwrD8Bj*J?7xN(t3MPDrL@yS?QpExp5wxd2j;CyRpplxso$hTHBYI@tFCgOfE z1nbjA1}dA<1O9|gp)VsRqOw;x)|L{O{==MDQ~H9tgortoK93TB_JH3a(^rBts36s6 zk{l64)I3F>V3NA2*0M4YW3is=aZg(mfbIUQ0HT1d-7y!0rxFX!Bl*u}7G z!nyJqk=;W$SKbS@n>pz5wI{ruP$Fu!$;ix(J{jl>hvFgA6cZV{5QgE%DBFQ;qgFn1 z^r)lHw%I!o1bLA>p>Q-Pci8Q&$aGOp$QKr~3cF?uiJ8Z?nW4D4;b7~Cq0-EF|7 z#S@E}K~JPboo>|H9q?o15`tzo$kq>jYStq_d}vOznbv&;430M{&x2_rCV-J>pxsk) zj2)1uXOgYyO`IstaA-CU!dQzZDyywEnYT6JZHq?L)lVnH-xgQFou=4WbHa-y zEU`1H2@z{%BgLwvP`hTO5;onVbf}dSYqn3+JIgWUd3wz*W6j!7r>i^GoM`d1n?bAb zbd>@amKQVi=8H8aV2*MPtY2^NSTj!BbFXsB8Zk|=TH%y6>TS2nE~;M4R5svLm#s$4 z5@XE?s7W1OFRBJWv3S(0hxL1o?1fc%<<1x$6- z(z=-)fjD#}PbKv0SB)*=gg2(&h|vs1u}}C+FE;D^9=Y|YcR(F0ZXIH#tnL;;JJTso(FnuH`uHuW;dhzcoGqYOgIQXgu0# zO$vs^X!SJ5EDXm1B*`Kq*Qf1)P6DI@1=k`=^M~cZD!_`W(`uU2Jwj7O&4%;W+(U}< z!_z&Y>lB%owM(eEM3yZSGri%EyjxyWexyhxQ>>+LCXM(!agU}CdmhE$s3{Lspk_tl z3_E;Es`jwF30U0gbV-Z&X}9kB`}HA*dB&q2d8Dj}i~(j;Ec+LAym&Ow+-wR^Y1eDh zg(;$J4<_^;r;kD`PJ1vBiJGyP8q}emBCF6)60tU)&kQBvW;7TGsR05srNt_?!yp$m zE#I183$nR{t@{SHRfYX(wyXEmusop#*PZxfu2K@ z=}}7*yH$7$62XBME=D}jpmf-FWg?zv*e7QfTGPYk0%8)6%G(u+L<8YOL>)oWisbxg z*lWt=Hd`?avM%}Rtio|Sez02+=O)oAqL1r*!WdxXZ8N>-aC}&&5Q!%|@wi7`%(X)` zE8px8GZ1^_Ey7FE@j$3elx~NU5yTpy89GOM--v-$N%k^e1UKHJu{iE(vwhi?4Ny4h zHIHoz%geI%)?y@{K;X<2T2qPICq!#$dX)L%P>iM*nJ+f8Y;O~ZW6V)?w_^$9$K^>1 zd*vV!PsCf=FkT9c*bDxEL_NHv!IEk%CEG)aYR9`4K**}KxuOul+7WlO75nZ9dB(}+ zB=fy$u4c1RkcePqOJxHRZ81-?nFxkMfw&BBY;{UB1PENM5{Ugij;(7+6vF5tv|+cK zgvH`&B)6LaphO_#4}=2JYvC9(|EI|v_jDvYZSfWgll0z_3# zbA`xOm2=dhl~(7VyWsVR@h15rzQR+KM9^+_$u$j3L&ml8zw}pkUY+ z@CT$O^v3erJ1K&tJ_7`0BoMMSSfq_e6biJouLVX-84bt7lO$K&p!~Q`84ruq0X;^x zsKh0*;s%RE!yPI(DHbJa9;2>!>lW%4FHwrP0z5mU-FZx(yH{_O=BeR)nT&KA30?DWC>7n8#}B4Ac_YZlpW~}+O1F^VAu3*|~aljCggyZ3m?9#Sg zV=Zm*gdFHKK}7C!;T<%d$d|dA6oIhUq~}1Q1^#k92OW33G2UfyZ%%xnzT9K z_nGZV_WOsU)_2P<9u9$*CD~+F4F_&ELuOR=h1-Thm?m$cpYwFFH8uzQL64Z-oiQvB zPsVb|w`{+@xVZ!Vpcz*MuOF^pRqekH2h`)l%Z5XU=!)4oEJllbJ?5rip|L1>vgr1A z4~L;Gt>3x*fj>KYYIDH<-^@ASpYEAv+D)^xRDbqb?U|((A5p6V=a#?%eqMzVqsHD6 z!4+7Hg`W=t_nEPHG%RAM&q}&+z~9W_jeI*}@zeKcOXaZHXsor}_J6l;FNVj8{eaT} z#xy_hbxG+l+Yj`=sXs!Dl4F`5SXN5v@Bd-AQrA*h~w_S3~gH)`Ase$IN53vxj#4EP(d*vlgF(TlpNJ zTy)pZi^sx2Plq=g5^<;&^~HEe-HgO7pkAkhbLwqJezCCjJdiE{=FQsi13QcaMK~1K zR$;ZMG#1u&P0n!?GtD$3o>ui{kLFIo3BHrSL(qVOq%{z-Hg6oV2>|wzn5y5tAP@^D zLUQM6lY>`?qb{{gbV(6Re|GKSB3L}&pMlZ8RUPBL+=&oL@WB*3DJ~57Os6<_5R%2* z=%i~4M!h)q#<11)^#@eJ-r2d-K@<*84TM?)az#U%ipRn@OA)m-xynH&&KTjiJC}Bb zL&B)HeQKKsZq9I0dE4#2wWJt?I9@DXob3eR)D^l97o~RwK0akyQfL$Fn1u1cnl^dG z^D+n9SjeZOTwDyWr9873WJ`IL1B85g7NRY`#KJbLXcnR^EC6eUT-RCZVB8k*dRnm@ z6*FR-I1%uxrw{w}UWerbzi4l5Sr6Ij(c1p4n#)N9{6SH9{XI-%tLJQs zg!jWA%35iTH4*Tuw`=>iI`XZ4yhp4{2)**7m^14&FShF-@-LevaSX@4LsM~%b5K|_vTUa6xYmbW)Ku+xGs)m|V zP6Vc_8af)mb|lWWoNm?JVG))w~$?8DPGcHA2X+xd1Y#=Tm$ zEqB}-!F`C3$&XW_r>eIK?KTCD$#2RgOs^OXmpC~%2OL{hpae|)bkZ56WPwn;T?|Ot z;}xI_dOBKERAArHkdUx~*f%C5Bqm>Qc9s0!^aikx|7*qs0=yus-b!;4gu-F@%$scuf+C`Rjm0TiC>+O6mn3X=to?8b zD5D*`s~E!l88j<*GR*=Njz7RC0_;>8j{m_udc4!*Xsg3x9^1^c{6N_2Rc{}CzDxfB z83y^{!VByz1%>69{@N~m9s??$m&X=+40&#s{#!1L67|UTeWir6FBS*}ywNZovE+LG zvjoLD9k6^DknpWLO2o43KTrZ|W`V3x>6u;n;uHjKHsucNnq8aD(ApNiq+`ZK6mwoF zl@(+ErxXTL6;C1V+NJ+q4R{9Zlj3g|H;Q#&d~pdVI90BMyi-!c;M9Pyig!yv&Axcl zs{0Eik_Oeij(?Ye!agw`oH-mw6u6|cz?fd(i^ITWeE(i42(L1-Rp`Do5cUQWUeOAr z`EspQ^N@`1D;?n^KpMTXn*rQh8%qJ>VO7+ev@?FKKP3gx|ro@Umpgh ztl(dzFsDmIX548g%W<=|Tg^A0;Ylnl0{&{|>AOX=)UM-cw z@}HlUQd!NryOb_2kFmWm3`lg>4|koV1)y9rWfSk064}1`tzDan=rCN{W8L#5V_lm! z6c-`Wo6btu#_QPdZ}$aaUXR?!u{Q$Seb(>J*jsb$ru>lryIZ!KVNsgh*^njHOYQ#E zZbsEl71>c|dmtozpWS8Jtz$;^0!@1>o~z(qV#iqRK{=$`U9tVxHo0eL?@hKx%}7-J zme5G+xMF+EYEOG58ky?&x4j+PZnc!XZQO1t$KLF0j~8DaZ# zyYkOvsbApI2E6IeOhmG_n{awth)-dO4rtqeIeBWJnf^FfqfOjQkJoF)IMdcHFIogb zxHAUqiB_vIAQX;!+HrOsF#8oi)8h|?V{H*$)d0wTxy|%6BmP6C{cwQkX~jS>gR|}D z_DoM(yoK|&*+yHwkk@QAeT@9P#+o;hIoGyR(;E(jJX6C_7zMc7H((hQ zHT`Wd$;;k(mmq{npt0}Vnm*IxYn{fqwx=+$ErPYDsCg`d>=%PfpV^M(Kd{jr?=f-r z4te&HnCZh&M9$QnZ8FUc{8}VX?am^ef8k8qMVNkNa6hn3$j{3~pnghI6kbC0;Z1!2 zG!th)8uWbX&44c&WP%qoPJz~#tg`l6nHdO0qTyzoQ)dGEg$R>vE8B>z+fF>JbQt}48*W(1djI2d$ScU(j)KRny6U(;>@FM^#t+5B%EpA z<2A*>V36AfOwC}dnVkZ%?TNO@n^|~T2>GvSr6JOZhr_Mmkg#?8f`f_cYF59~9%eE_ zrrj!&Z6QNu95(|+d)rsTk*~LveWlzCS=F-d`I@0H?kusO+C!seD2yKgGQ)iS z61CEv;xI95Z|7Y5po|#`QP3kM96mdnXP693sPt0zmc?5oRbHi+DNHzG#F@p&{)tPh`eL zuJ*7ia%Tv&Y42m1p>Q0_o+!(XE|BZR<_WUx*PD^8jMqLwgHrX~&q z!TV2Kp!S@Z8J-?8qwtdq(SAD4jF{4e*kiC6DH_P^r|?WZO@!*$gR_Yv?s%pO+1gW- zoZl9*hf3|;Tfmxwfmo1#BMOAt(|-)q2_My}$hN~2+MB5&*E#yykA(=R<#p}XO3cXg zC~l~ueC=5~^H}^O5a;U7N<2XXTziMcl(QVVE+6gdG4QkZ%><_CZa_ z91Mc`jiO8+mR|v|mptK8gPzcgs41&#KU`_XJK|GiuwcJgflZ;PsHpbJvpGH7*6N$e z#|}||7AJN1q=U0V;#i>BGc_8gO9yMTo`KxfXpnR5K{ys@p4t}UOgjdN1)9Artv=4R zJ34a1A;H%6rm;Y?85Eh?f`6ddA2FHUcBZjFb4y!Z%kQn-B9C)wx}FikQG5AJg=R=#;|avY3&NVs&xm2eUF^+i7B(Z6 zXv1E#0BD^I{97=do_b`pZV#xG4bZG|Ml2DW77B}cX?OSa-FD zaO84K0ycIiiR@-GZmPXkI~V{2;{sgQAn!oilV?Efx#V=hyxI83#l`~T4==(1dZwjfEG+LO#2l-Ee2&!AcQ3?uBJAj z00>qR4Y$dMH??^@fH5u6hKfWwt}=p3kFX2MC6-F z+7b{T&8AMKts6<8FOF#oPomM5+QJPGd~Nap5p8t=`F?AGS6g5JAZE5IEo;+RK8vRII8aYX){{}hpo3#wP`3I-Y}mEV7axKAwW>E&QqHx0E`EKnM|9hAwOoa zL&Urk%FrfBfZ|ycIe==D0zeS*w8hY2neP3P_C~{n{XUO4qWf_PT->T);!jF|(0~%@ zQzsPtQ2eOCUMGM<00H^T35l6S(tPfO`@Pm^`b7yujlW+yAu+6lsKN2;VPMLpwfmdU zoAzjgwpa{^pJxGRgfpFyP;i>{W7*nk=pc!iAzxe^$~ntPV@;sI@9ZKNn^(jmf0p|k zCp0)sg)Z8!t)Z}JPyx<$QnaC|#n`C*q&-Od8t6|>B9HGFkgUVZE_8z0LP6_SwJ&ml@T0rp()cDP0L!!<^@HV?Ibk-pE1UptOuf*3)i8jB zr_C3L$HUWPufN)fgZNnL{#qwwnh(FS$&Zl;b=~NMV|b91H$p zs2OjIifHB*C(zr{CR+1WXTBe6iPjK%yAz0JJfIFRlR*T{;CjdXSh7+?) zbhi`ccL4PRi6T11MR2*BcCV8X9*6J1Gu!)|z-ALGcxEUb4alx|zY`&-qJ#&W`K>Ya z1ldDQ03UExFSR`Eg!vu7N1VU}ehyq&<)cnm(DHGQ6#?2p7U*#&s0}ZCsGFisIAI`87zv_g?Ol-(W7xphFOnV#bH75jbR`E^%H2S&|<}a?*&LVhmwcc>Ttskp@)0saV zi>zilb_7uBTgA{o+y?G)0)t^6Q#0Y)PFS-U@355jZzl{@v4QV6fia!%T_+56vj zi+H9B=lrn~?6=u!Kk5l-LjIVyRUW+gPZ3cIx(GMOUMD0N2(_3|=`ufe!eV9+>vCxS zFPsqUqRIO%%PSW8fzk~zBspaeLl1mY8e%+eBolI+|P2>UQr z^&d-N+Ctf%ilC-OxvUiCZB;9E%S(Y8+ZCly1S>4i%2HT7I>Q1Fl!C1%yH=ILa0l4a zDxU>fT?$9%;9}R5!mvHBZY>Tv!LV;}lg1~HaIAGA@M3}!W1SNN@rV;)y%T{WmIPnd zQH5-9VsNnI#Q3ul!;2Y_6XQH5hJP58^PMPSq0PzV0w)Sj+C&uh7bm0{(>VdT&0^ycKbb1131EM}F{<`Yh@MZ+e zR^Rlfg?RP1A!6&UIqH$m+gP_FyCJU6OFB)4>+NL7`A8B-1#9VFp-GhK@*QJVeKkB{B0PqK(nGmn5H2L&j(^gL z1I6RnG|_0=oItFOS+kD67s2A5_+etFZ@Ux1BcKF5?F8}gDM9~mf&@ROfB&ZwMN&I) zo^j&H!kswJI&mEHI4R9@P83MvSFq=uFfPIZzTgC6sfo7_CEpjFVB6g6eax4>4Mp4Q zeA$WP!5R~H?N^2Y$VL5EogmB-)l$&EoFGgP73ei5h^c&mI9{z1^W8U`2u$lhc+-i% zvZ_G&EhlVh%qO}89syEICA+>22ZsUdw{PEeqD?!t&6FYKzl$LDH1!=Pq}5)we%A@% z7X}ltkS7vrk^5ioIWcU}5m3b9a*t)V6Ki@nI_+4FOu4n*9}YUrjD|E%zsHFqq?gw8 zffJU%vsS}(@(=%aJiPD9r*B!XkDLUuQSe}s90WghBD9*mW?AD;oRCB!=<(_sdS4C) zk7zo)d6I-fIl=V&&-dlBIY;a=8y;3L6pPe}NE~XRhJSQ4#q1}mbz7vCup$O#y zCkh%Nj`ycE-TlRh!rM^JaxN^w;DVee7dcU2GJzs48=WYTaBEQmY;q!ql_qB~7nh(! zy+t*?#EF8n9Z2KkbEy*v(+D{iaN=Fm&DN&ve+M8CORUV`Na8l5b!C`Jx?yxLCL zmBWGIzU%>B+i(5E}pl2HY!ARHr7bNMmw>VLpGJzvA8ggPq0zTDj48Fb;95rKY@_N+`rO=kZ zRGiY1mnHvNilDZz6JozwY%t&G#B(l+;HQZ^Azvca8t|EFapYzv#WW9or$$Z~{^o=Q zJstYUev1vwr%y>`XGaP-uB(0YRpiI!W^)CG*yFf>1TrGuK9nf(TeF!tOi;`U0Q_GB z>`fNt29papx8J6Tu~@Ne=e%TBS1yoA6~vTAI}=!faTqhv&e=IIJefrde^)!v;6P~0 zq=gvzc@PJNFIh--r!(`Ik9N?Iky_K4IjLeT3Q!;t%%x6ds^S_3BJ&TkfW<6>`8jjD z1WWC78kgqnv0=5=DRKd0ZAIo4(wW&=E_67A!*qn=f?29~TR<_;&JuFz0dFue|1f{L zyC7?9BZyOy2|g<+yNIoD3FI@`*t)QfqDqK0*DukNu8ci4M=8&2G!3)V4;%?uXD2m1 zmeh21j}9(2FB>ctql(F+1&z0UdWwsV7J{ilF5Q`D$+Y8m44v9LyFHc5r?VLb+a1XQ z#|lZ&$%?fxwRd(Tmz|SK_AqI&)KhzB2QqoVb{NtYX;XG(5}tU%@AFO<4)65p{Gxnk zvb%diDl&q*?>|v$=Va^VZ=xFHCjLPx+F$1?db`l|^I`dta;5xMXwbInnNPPr4Aw z&Q5ou9vu>~=XD*iC!L9A`wHpI9OZFZQ()U9YYkhqYn?bHmCI69*+dEb<3wHbuD;%Q zF4@zY?M`P<;t|7Q#R^?4_!J3smJlm+W%~+LRkr7W5UZ{PoXX{>s%-DD;8ZRrFQ`_P zZJT~9Gq|Z$mk}7^N?=vlj&H>fm(Hrn_6&;wO{EIHbUrz&I|YY9Rb}7*lduBc zXSSy|-JNQflbtkS@-HSF`s;@A`MJqls;eQFnv2636Fin$OV_XglA!S={SF(^I*ErW`Fk}fw7m~9?VWWGAPClxd z8s^i4jgzX0D$LMofj$pUvjhQ^Y5GWK(gjPo zkl>1>WaEn;d<09gSl(-seKAU4V2i(S0tc^KPU% z=cYR6K{jq3y3Ymsz5ALv8h?6jBDn+~s$BQU|Ij!o^bzt4`5u6LRbx*#Xhxilzi`3H&N%aU3jeRvzPc-%|{9Cyv zQcoLXL6^!edW;7JIyK@&74m&OTmz+#iC^k^dDrC{*Oe?JiQeU^-*v63;Jd6Ss)R5b zr7pM0FDm%naDr`=2%F{iTLgpk3(IEFlt|kw(W;1OoNTdiXs}h+)e#zK9NOo`>y(h= z0CQG4-<|GEXXe=5TX7xwN>v;Ndz5RO=<`Q`#b*}BqdL*|8w>Os{4sG$9o5!j090c)B9+H;ZBi>%*~2RL_csY6?E|JeAM=pRR;@t;4?z+U@nJ>TY$Xh-=O4W7*UPpvk(sP09dZ)vNfRzHlF?3dv!X*o6HxI zb28cPG{g6P`rP_T9X>6a?d?g5=;*qkhn{y3_at-qxykNSmsM)5PAgmK2yb5@J9{?s zudmi*?X=1k<4slBhDs6N!1cIvRN=!kVXMkEO5D94Avn?4&P~pDVs4NF?LPAmZfTfM ziz@m`$*!bmv=$>BK~-f#5(`E&jDTQBRb{tGtP;_{OU8%?tQ4a^$B6XWNPBpf@g(iP zqOXGX{Hl`2_WIBNpN#E_MVb6#LU(#rXCj}Vdf=_(?fmCkhO zWD33BLAGx(S!XtrNp%)RwqZ2BE9S*{?UB#5yUMHQx-{YV=uY@j_ zN;;ZD;nPui_q@22uBS+;ilE*>xVeO|JD*C;(`%>*8XSbzl@R8K)2kwy=peqogt#w* zl_9-YMKQ%e`AiArf?T@bFg!&NaS*;xLfF%j?9~fY_%j`JuawgDWr`>jMX!T$M+v2v zb?ap+ihc*>&Js#^ZD)5@6ax;*T_u!rQiaY14hvEg8y%FpODJdO3rR;GQ3OK{!aXH~ z#dHdPtAlQ@%SCkf+4d+`NcW^@|49ogXyBt!Jj6}>e>B7~eFgrV(3325&L!7bWvNWOG@V_B$fu!m5}i|q$zVh0WLnC9N%~}OZz|Ij$f$tKbyb;bbRUuX z7|a~Gh_3XUbRmy*4%de$rqD$k&qxmWRBtkuEM#-!`g9m@a#r5kjin#gwsP0#v(Q~tXIdVa|(LL#m{41a86n{zH@pMmWLMWNZ=2M;7Oc%NSUGBo9`s9mi94M{X z&SU}WW3C&jEQE$``fF9cjhUR4pPow1>q;&n*G0N44JdX~ImD9X2D6#M+zGKlE}fY} zuE$2WM)wfS+*Csiv-)P!QzH~VnF6`a8tEDhES`jqo-eA~uz=1U zsUb^2tRh|;;TqjVv_KW{rYhnsy@+()hqgF=PIf|mZn8HuAgEHPT6udLr(~}J6BB@+D+eNP3H7=}XXthq~TT{tia@}9!8oel$ z%n98{>aOIXd>})vZIxC7mm)R&6jnHPLSH7mFb4a)!^I}ncbR2_3I%rLy79k&3$58s zay@5bjPA~M5@|+iu`1-!y}@iIORj&}=u+-EsY0f&=NII<+HT3wv(pPxUBoMD4H>#- z6s)ZwE&q^&y~$iY)f|kI>+dx#(Qu*>CA>S8>`G_m1Tw?f*os;3ZaZ`>DGdi&v%AQ3 zh->uhF5=}=s=xq&CFf`=*+s6KT>^#z)ymgsd? zyX647Wipqe^D8#{tD)WtDo~(gJG-ag&m}v_^?t=LU4L|MF4ZZ_ZgggWaBrhC^MwK% z7SIRGs-ZWR&J^gw3altKr0FA(%f6RBu7I*-Y5Bfc!Y(cz#df%4YBj-Qr(4&kIcjb_iN)hm&dQq1O6F7Y&&=Gk zWR~g9ioY5%^kfY8EPM89QNiI+rhiLt8%8MXs|(xJGx2 zF5OUANN0~Y=F%EfSmBg>su1eyA=jVFxi7(`@$e_?w z3yuwGy09F;h8$g3u2mWnTr5Yh8=t`cl54DM^lYMKms?Ua{S=FPf$-P~v(j@8<$l*i zbkF5A+;K9Ks>v>?m;73Xv?&p*lj~R9J0vAnp`n+idy5^axi&p8hccMxeN5bJ;9gT`b#|!=PT}8Z9z~`AR&|PCpVp z65CI9Xd*H4g=DUfpPnwvZOtx7<-F+E>{auHoM6?^Nw;wmSQK4wcQF-`mAOWDX1gqJ zAPix&ba|472OjC~?eOLJ6Z_Jc!VxD<5Q~9m^*qt{uB;ipQ!JFurITjBA|&gQph$5g z6%{c<>A{MlqTxn7n}Hnin1N%+;~X^ce;@LOuQtZ^%$%ZH^B~%jR_e1|L2<3i3X6JE zznYLbc|v|srt{~?LbfN}IU$neU7q9lU*79U=Tn`930rFNE z+rQQk7m38I;vy0GcW1Nn`g*C~BRFlFD!wjWBwkzQ61z2P%djPU7P!q%b`z~FLq~Z8 z40%^(Ezf~y{OP$#hCc9}sNZmgrG0D;e@%W>AnL(@Y$NXgm-FOqP-g9NFnU+iJM;Di2(l+CkI%>?& z4NQ155@~vf6Lf1G?Q^3~l=+)676-)#7CaHGCoJ zzZqMAoD|j=T*jMiD!TK}!It1n5*LZom9umjKKK z^h#4yYP0+T`lu@UDT~RM5=>I>KzmZ5eF!wtn+o$ztBtlciz0;WeIwfH8*Z!b-d;!J zGGdZS^bPCnm^NU&`<7!p6**_P(*?%0D5+rK`j7D*%2M7_*Dr#GoTBwu zD#13DquWqpo-XFUZn(3K_Ni@63^bMB;Lw$QX>`> z?%{H--C9SDJk7h8Wqbo^5#rukju`ic_ta4%;#>i5Ly}h_?qmGrNcW^OeT9_Fc>phVG(rlv_b2|J2`dvChmvC|ml9v zu@ULX)%5M@*NWhJx;vGj$2bfY^hm!Z(zWzeIW@xDUBxmy3K@EHsoCj;bQMc6?xi|v z%=gW9j>MgKPhACf zvfETA{@ZJ^6PLeUN8E`I!5`?v$GncN-@V+RjRy#-?Vwt`+~AXBWB#<yazucf5lEWk5aNTl)UVgofesuG4gJ{e>%MJSN&N`}kY`H;^?^RF@ z5k*`?V_#Zs5H-EI+@M)I>uAjGLJB*pEn#lvu`M#->(>s=jt~k z&yU|={p?+C(6;Xjo?KI7g+af0vyOiBy%mu2$14n)`KILA@B_hfj#JJv-;_LWeUo`Y zdozC^cy5E7(B607suS9q^%gXYa))#rlzY}&j&iSjOO{)I;IQRB29~>*8|2!Ba^(Pe za$mNPvit*`CR#}N%a=#G#^s4#9to#&;x6R|M4EV1cx5EK#;v>Rs5Ww71a*az7oiXaUOKYfqx03UWQIK1)6RU_8<-i|w4(E;Hnx4lLf7@M0<9MZn=)6%_ zq5AdvisyN}(1?I~61fCA5|feO=(=*k3WIKu;%$>(82b5Th@@8-bkh5EGpHM(n*3-g1%6wzH=tId0G zutc~inKkVLRV{&nnoJ_sU7{vc^HvztaF8&}`yt2aD-8PihY%N_Zdu--{6gFZ6xw4z zlTvasTH@_ltc&@@D3>(|f5CEAe^f_Ga*;FpBPHe+qfWlKMv3`_5VJONPz5z+=h4^P zRQG;ltB2>KR)G^YSMuhQuXaW*NvrYUMwXVtVGaL z{{;rhY$v8*BB%KaCFWqICgxxz+eKrtCzdME`byZ->0i`QV=rfH;EbD*!K3B{m6!tU zMp}%Tf8mT@eIjh?ZQ|r{NMh8yn3Go^$!mz0 zaQfP>>!|S*x|A#SIVZB2R9LJCUdAYszrj!PN546WY;pz7UcBOt(>T>bW`KpY|f2)AVmH|DH-WQSFp56_~USR>9Q&!&zIAH8)3Za^BP`H#PS3$kZuF z_h%k zjqdp+BhXkNh%Je=jCF7=2xlTcO_y?3H?pX|J5d48u)Tbt?_pI#Dn?76ZYAl?MIA zaMPH>RvPr3eNiCe{6gSF6wdzNkvJXSaU*iWf&d*VkiHA4upkiJ&rMW0Wu-xT{v@pG zOZmmXa}dxvRvNScXjK@5CrS>j<#OP?eLuH8@J{-!Vz_jjU^pnhV7LraU>FA)Gn^?I zG6*C2g6}#Q{%(KNFuBs80f~7Oq{^){XgPzR?QIfd0t9V8u)kYqd({89X)LrYR2}#q zH?flM{vS8}2ucS1j>wB`4L%;7R zrB^brN?G`QS;}qSw@Mk3NNuW=EkBT@eDDKDDZe;SmU7jBj#73=239HW0TI^n^B;04 zr>``q;V@C*Stw=YN`s~x>uppP#UCUoFM;tLb%uOO;?Py^OV zASN$&;CwVyzJdD4>?QIG+A|^T$N_^|M!TuX7%-^*a8bvXNOEntNECH~{MHcB>&#Uc z%9O)}-j;zCy5b*3OS$hG4a-N-hi()+3i6AhmjfR~?;InG?k_I-D2X1CU#)1?M(rlS zc8dIhZEpop6|^xVxd{!hN7x}nU4Wt?>i2&nMLp?9O4JRK*GuvXo;lz{)Ia=KiaO

3jg8W&Zj;O`3tV@IRL^sVz~#WH)*mSfv>fjfO%lgD z7FNnqO_4|wl~i3nl~P@fKj3_J;Sad_OMd2X_1_yOr8;q(l&V)Uv!uEcDCoT3j8jq# zNgPY6kH<@?emB9Aihdkm>{c0jWY0rMf`mi@M|&Mz3M-0@MY3uDx<>;>rdf*5*?NfiP;MgR;m||^V>@SXR6YUd``Canc*h#n4V1fUj zBcYP)fI$;~EgJSJH0&7z26+I|>k&)O$)vjEOheV7VMvrL>hiK(mzTLN>0h}qQJqMx zmswhBTu86g;2g#7Um;XQX!r_Ca>TFQ)R;^4&}&S1Jdy~0U#r2C?1o>vscPkbLH$z9 z*C6Ii0|x!`*KTUM6HR!Oo0=XNFz5#S@zj7pNBqW3W1b%{XgU6PZNQ-WeuKsl^KmrG z1bqQED$e{^Rod}aRW%Up%ZtAIBC>n)H*Q|@-51L#KRenjw^)ewMMPIPT2-0hVk$+G z5V>A+G(tZS41Nb02mRJ9)?>c|m9!zn6U+SpT!y5WSMHC+*eyt78p+T%k(!H8x&g>P zoPszuoe`-iNMR1iumw}xv`^CwqBcj!?;)6IG=Y@FQsUNMR}d=NfEzN{57;urE%pQ6 z#2?rXsPed}2}_AnJn;0`4``6=6nk(O%L7}9mNCh+PF7jk-FTCb%f(tgnzej1w7k$G zw0tzQe2GUXm6NckN#r^jra`s!M^(V0kAXpNnF`6?9WZFgje_w7@(anPp>F%FGH5%a zjr@5%T`9p2%P-JcQ2~aMl&(R30c`z; z3ssfXIiGjzx$|D&f+?-^x@n)Cvgo5EKkgM5GU^Rp$RiYu9e0>?&MKJxrUWP_qWxXt zI*ImQaaj#j`c(1z=O$KLi*5_%rZN^$uY>qd*SIdCLveiTXBP3HOQu#6H5TXy{4w9> zR!R|P*>Yni#0uG7i_}O;}~n79f=I z>4RRJ1j<+ly>}?YdLsInYaGs3{tPPFj$*KJM?Zs4sA_T3INV^PpTUh?6HpcV8M}o> z*mGx!#sT| zl3iUm$E#G(aHmoQLe{&5HQw#Q@!svNFb|@;S>2zsVleINp(i+bK+sLKdr-433M=W+_!|*-tXHdhwbFTaqqilLcR1`W#G4e|?UwFOnV7ZgLKL2})ure>N_mzv}4^go|hP1@9^<18ZQGy*E?K3jRz;!J61*B<%$d( zD=YE=SLDHC-NZshq)ZYn4Gk{PFfGgU-YsAFnd#a{TetDuW(rbJLiSs||V=f4Em0 zG^X86V-8$xPzV0_$!ddEw5uY9WD$>#AR3Dz1pn*X-NbC)!XIE;JzcULk3Ycn0{j8C z+wlk3R(60bHt_dIwogHFY~cT_!!7pk;~j3|J?cMN$P=fCYCfZoYiCI0U(S%o^JZXB z1m_zc5RE_&iZ86=6-D@-BHaHtNqFRON~C^S3Lg}JNGBX8OF2s+-wtGH1nkZ|U0qhI z*FPc(bn&qI=|~=^J{<{<@ju6lVfE9Iu!X+kRoe?FiShJ5BbQ9Arp?D=Qvf!d4q)AFTcP!8#oxE z#i)0I1P#eAP%u7^#rTZ>gxPbAV|}x%nhrfGiWfCrJ&J3*nm1^_dxG%rB6ax*s>Z9u zWsSxzTE~CAbAp?yezDph@`yGb#2SbRqdd_~W7<|5WS)o?!T>N);;h4g5grmYOP${@ zzmQx&p`L9r_aE}>gsR-hQ234WA zRE4Tff*%q(ZOatW-4^pHkL!ipDhj!k3mJcsD&$tJkd3mCJ5We)wLzOtl3II8D})=) z)78cIVpJKCC-kzmi?X(JSs$L{rg2y?$P;bn(+)pOxM>_O8dzCzWZ`y;!ueA|Nhr9; zbBRwggL@NLNZ_kCPmjW~!)M6AvV(w6oT;|({=xWjkj5||l7B=JokAq9o$013oDAF{ z>;KO(c^YurO!SS!YJ&<=mrLarjsG;%^!wEYjZC_!>Fm`8{RV%m26_^Nu$t#2!AtTB zLhvWqWUNhok4UL->l3;vcM4VRWK~|1bkn%mecjzeJNa4BTwoA+)>3H@s29Eo_%In5x^b1S!zcBcko85Sgzmsgy2vs zIib>bmP`MjfL}@^080n_pm-l-YC4n5EwV(SzYE5~M|>kh`-VlE2hsSz`ZraWdCB#P zZ>lg)QgKo+ouU5fQ_rcUM`l^eT$s%)uHL7)np~Z3YMh_S@mq0cRAa5C68Y?v)xoX0KFml?8XH7pNyBC1IUI4c0VEFIfL1vbhrpm8j> ze!bG5)pOipv^qgTm&h-$u-(NI%00k>db=d_RQUxK1}vUXhUUV(ig1Ij#u>67rY?@hAr@?~7-YelJJT1hpA)1I#h|5{Z5uxHcyr~vAD#>d9uWQ$zV`^!N@9c z$vjaa8g7?FsFXGa#NTqm0oovQr^qj$lL3{M%l=Sy{$pa{kY%b9pAjk(Cc8(N>>fUP z@Y+0Op?lc6WY^y#&2<)iz}a)U<(TL(Z5v_?SLStqaLqfmSn02QpHbE}8s2k_GxpEgJvAJ}fYL zE2y!r&`E=ge{UZg?q60L#KSJAXUTx4DeQs*HX|+K=n2(b?s|`99e>{y? zXFeioa2Sxz<3EaOJ#awv#C$i6y$gYP6KH?Bz)egWT%byCeL*zY{ql>__gZ{_&wLg* z_>?V_d>R)jJ`YR!{azIOz~?i?Cjfl#`NKj7pPP`6x;$s`IY82P%P;tRuK0WbeDImH zNNQsa{s4zRE`m1ptTt$?)W%n!s#;^vwTs*|X1_HCZNncwTw~C78*K;AV5T?r6-{Dm(I4tL0J4G|p#G@NUA z>=ZdP_Dkrc@(U~sGdv`(*xx)Y5R?A-JHa5Kd ze~(U0bX)DfrW(59_o^=lC^hEjZiamDdpQg>_{6~X>ya{O?3F!wiTv6URn1sqP(~?%k=80ZW@1bPqu3^ zxgOy@N4VQ5fFw@6K30oU=_5`N`OyPSH6QxoYuPJJa(3+a7Mu09=ZViJAMsfRf=k-p}X@dwx(LYX~M53M6{s0Z_p zLD?!BUE{VM*PRKe2Al-HDgYKiE0*SKz?Eh$B$P$cs8Rb zckI+F((4+R!HHq&1=Wpz5W6$jU87#C7S{Kx+V!FooJ8%*Q6E6f{ccg&h^*`a8m6+9 zuZ!aLlV7{CRg2dc)GulOKtxSz*BJD>em6C3Tw~CK{caj_?HYr=>bEVnNz;tH&8zAY zc}o=J;s#plV%4m5LFz*ntI=sK@2Wh$*qXJib)nw&tZl70c067TAufSW<1v6(32~R- zwbmu}LAtYa2UC@ufmwE%?&Q>wNa2tuJ?uK^(i%GB3^(ml`Gyd(MScasCPujI4D4+U znd;uqHJtk}=k7pmce;nRas}Su%mzQ^S@UQ+3sQF`LKHa|Zf6lDah9mcb|#E-4r(LP z3pgz*B-otCIcyW#SvPl{`QOytw5wgl&1cF2a@Pcj1hGlq8qUiCMajbHSycVIRzt}$pOK(Y(-vlx~fTi?^=864rl zjhjJ!{NRQqZro7J7vlU94-N7j&&y{kYZ2-F&atdTklEHUC@!px7bp7kukBgpK^7aW;$ zm%8OD(UXkyF{csMiS)R00f#f*#u9{)?xLqhAk;bs2}EB1VA$12ay23Gvz+&u$mGKVcL=L>BGqPSJfMPP@E@E?)*?&2(v5k1cakCE|pS|1BD<{~htI zC8C&%h+~F5%Vf;ZwA@V`Gc3U$h#9U}ZW}6_01Kg-zkZM4Cd~39VFn+u&@V4nW+~Eg z4qzuHlj|dH52_u4!KP_1EA+4xIQ~R?M_{z;UxDts1MnQ3U52bn08h(D(;fs|w@}$u zO0eEfbOBeFv7N3igL|r9>9%cdwcJ_CrSb)G2?SwQtILE#%B1GdI{y2DRq#mJzFeL* zGTB?m!0N(foUC4rb;EQfPuFt#5~TB~&OF`5=?7Q4scQEcgC;hM{)>|KI>~J%{s6ac)=6$-*1L(hb+1?4A}vC;+m&o9*E_h;2FYzb{s6ZF|17!v7Jopt zbN^gqJ8YP$Np^g#H_eNpX^OuwgqBNlQ#Z0*wgBWF$l*W#`cj{?qpt zYvGNS^A6pE$jQ-5%<%2MxT$tW3oh}Y=ZbyJx(i`z{ecR?Su^@KH`b&J-PE`sm6}I; zIGsisG@o#C6VvQC2KOkEi{v{3_al?rWXd$5#n+V<-@4FE+_oJTN&UT#KcK%8H%k5W z;}7WX(T%D#wn{$lC_e9Plzi4~l6=Np>?Y>(^~I9U=u2d)&bvhMnbIocA#5(e;#OSZ zXw^?ImE7iBCb`{pndJ5_`~j)PU9PwdNp9F#Wp3firBt<7NIsY057h3vS4uud;1BRQ z<4VP6N>E7ksjA&&S30EH|0>DtTKoZSKe$?QI~sq0+qqZkwwCMb=EJ6zk!>4MMc9-V zj_Ss8HCMlpZDaG*ZW@;(y0IMA_7qa+&Lp6>T%)W}q^U7aH!Lc2weBqE@Nj22!ozE><`jE z@J_D3aJf6nxi{BRm{s#0gEH4aS;*n=Z+AJuzgv*G6`3+b+rzl;Uxy9JWDo7(l8@Q~ zty1kCAf{bgOLwn+*3X8dUgM<0p8^B!Zs5P7H zYL$E6j(nVHlW1uK8l; zpquUU%jFyI4-{SCg`7y z?NiR!hKye7=Yy-}%{WcD5a2vL!&ys_m7Fyzmzqy6an1pMb5mnFpD%RL>zr}B$mlC{ z(p#J{?G}V=3k#S$yvH+x`;fz7$Zp2?%dPSbYM#DyjlZOZM&HII)A=KAxuk|h+>Y3g z7F1tRL(}en#B`DC$SZ0ni$q}|roMT4+6DL^vLG=l@_Ej?^$zVi>#x#|mnZ(u$ZlzG9@iffR{XV2S#i5!jfkzRS%; zeXS^Jr?fG_1tU=3DR;T4b`Nqg^cOa%3z0L_F3Pz`#J5ztr@ew2Ge^*6JPUa1E(9m@ z=rRUZ-_2d5wqZImzMmP7zuQf<8<8U&o0wcg?uNN=+)qz(c}qD5<|ESAAT4G`Pcr)_ zk&&GzQzN&+MBtU5rrW?k6cLdayxj%gwg0_f4*#}Mf}V!c z0_ezl-Nc^l1Ph0b69qh@aBfpLFIzYfiSxX|X}eGIS$Ll+Ye?c?4U@}yM&W#B;n4A- ztXCDE>Gv!0hy=Z^K&RX+YYiHb7&u|t)U?*1ryg|E*okX#sO-qK27Rbl z?U0Bw(5g_p(E6l@ELQa=3Snvy*n!pZha9Z#dq_$(MIsil8u2i*n!47YUWwrbtM;`9 zo%FDq#+Ap)4#vwRwwaB5M3%Bc!i(x|KEkEUS!+)v z3YdRSV; zHu*(_cqs@!TWe74AK3Q6ws`sXM0*}3zd*YTXxM{Z$7tZTM1lw87id?aPk|eW5*2KN zhI18WSXCp|8MILfe+6>5==$%AqFdw_Jg-$n@AFUeAQZh#5kRsUL7peFzV)!`OsK!W&nL4c04KwBl~DMfV}gHYBU3EBn_ z%GwAJ*KyZ@qK*we6kpWwS)f6^_W_MXYM3o5@*zMd_gBwCS}^aGpmzZR^Aj0_ifoji z_W%M_`E#HGXom#tR-i)wVv*tpS79WeW%3J=c$y87jUg5g8-6xh`;3{&1B+`^oeH@(b=8zJ3c1$sjF-mG#=5=$Pqj2x&R7|9BePiklEmD|n6JMC z*D@1`vZSjS&G!mQO6&C3a7kyrBI&PT`u$!-NtNGUXV6hU5h{tuFWTUzc*StSI)l7B zV1P%gGicyNiBsuytFr4kpZ?tDXDq^f)q8k3z- zsZ8&KN^2W`s8m#}6Fr4~j2U{vTiA%62kW9kF#EpeEveH(&juYR;Q)n`9r|y!t zlYzU8ahoLWko-2DLPr62i^4q$xSs)6if(Po%BQN8;51Q`CRV8yxN3(W=G+`DLRfI% z+i0c5IPNe{tflcp2ivA%&veDxa?kVuV9K&GsfC4Dq02kBFEh_7V9L>|AY6@q6uWdu z@dF)?^4$K*x7{?pHJA|B0dYp6C zA!kYt*0Jd>b}_diXCrd3>hLhS!#r;^X9!VxUuHLj&qSYfuccMkKI=!iC!5Ji z_$4Sn@JH9T824KRAck|fBu;NKmrq5-`vAw!Q9p5WynG|xqM8^`;z9hV>tGNjdnWTM z?IW;R91VP8(1RaiuQksB!xZwM_1FrWyxyQEK6BHUU#&Oj=+DttfVo3rcBAv{zuuta=WfzS zmVPctK4TJW7S#6&<~<<6n?whGA>SmL^MzaN8Z4bE8jif;3uCl)4Q}|tk~5;SLg`<9 zVU^yr-k=|Ui4COn1~q>v#klNCC3(N3%|Hy$KK!Lc3zGL0N$@L0G9*c|ilp}|2g&-c zWGSCmBs(NYuOd0%YX`}1zm_C_`5L8o*BexSve3^$kjz2D zQzYiuz(f^(|BZwB#otKg?^q=LlH?pkQb+ZAd;OB?MH8QnP@oF;D~BY>YLfRWcbqIv ztYI09h!5hO+AJFCe)&ZUtw9kZ2MpRo^~4vQ8*+jW3s;pm!?jeRUm(9V_A^Y(_Bok< zV2fZ6wrdsJcNE)5G25MzY_I&**lbzD{W+mw46&@?I#)e4LBp53>V=YDcHvIA@JQ@! z@Wgwbf5##}jK@))X4^4BQ1SfAhc(LOOk$T~JuLHK%^jZ^L?h~HTn4{Nk|%jW)+no| z@tHii#DQ5FpL4zl>H3ggW1;c5Ht+h7PX<$CzBAdK;@|tq(-JY@fwB zmzLGbC*V4ySL~8sjGZ6h;3i&ddjV*&-pN8XJxpCy9V4xJU7sl{u4|s?nX)TCH|Q%= zY9YCv;e`n)`!l@OK#eEUOJx`smzUR5BaSyb%UkC|NaFdKmzc%V<@Hpxc)dY8auqD_ zOBkE4TyM~)<@MAwwBDf3ih63gWxYY0@yEUE4cc8%Ph%cmZ;*FnJ&pP2dV>~?6s@{Q z{wxO1Ci#g~@e4SrO8r4qH8Je)Z(lrjX$@U4vR?cU1rkq>tmpHQa#j9+GZy(<9v1WZ zuqp&CpxX+5;6t;OU0)bfJ*vc3E+6&(j;-upSx>bMdCgXitW>sgrBuX?@(W2VTgd`V zR@-SSs;a7r{Sm#Yx~Yh&vZ*}%r9rDJ>*;?rm2WB?rZT3go}{UOg;`ZEo+gUqHB-UU zM02a^iJvC=C;q_GL*73a`hdm zl5kU6w1GN`Xyb2R8B|?UPdF}*$8R~>_)~2?B8-XT`mfPO{XX@4Bt%6U$0Kb;8?Y0Sh8 z1|{n2Nsu)32@(!FKypUCg8+;`bHC)OdP(wq10)DLB9g=kJAPx-i>DBm8uc>lI99YE zMa386h81=^4Nju;0i6{}KWJa8^n*4SWbRupB%iyl6yw2tAqFbDRnl_U0ou3st*5d3 zZ7_(Mnl>0z|NrCcJK&=#w*Sw~W@9!9x_d>yo}k8r7h(da2&i;HDG^YyXR~{gtl8`? zTSBm45D)~ESZIP0Y!oT6Pz04IMNyRa?Ecgj%M&b5A|m2bAOFwy%*@@pn_&6l$K0JW zr_Y%_ckc0$ByhYUX_X{A>;TEV$D88dpG}ISQ%6OTT_gO%!w!&K&@q9ew4;>rB9mZ> zR3b@u*nuh1XCT3aZQ{1JD^5_%>m()*JHWj5gaqbGPms*_nIuh;gohm<$?TLs(xa0k zxup}NVA#qI-RgkxC27iWCc#YvXCG8`vb%O9IuV|_P^!HK%YUrB4WNG&JIAF(2n3Qkl(kdM_c zO2-i*zeJoPpOB6J`5@6~7HCTx5upt?`Ila2T1TY51BPQy_nhbyV<9A-g5)s8$3h81 zEHE4wb_XIRx5oCkp1eGh-cekU`1udH;>Db>$*HA1j_V%Jq?1mPdLnV*NzTMEo6_^z z;_JDuhpHy8(Dm)u9$1k{rHUsKQ~wH2VZLJ6+#cgKD>CT`#TbcQiHtK@!AR6QA{eOP z?+z1|3#O!)8FAnpap0&}bsRo54!_0-2jdyj;u&C%uMQ7(PhtVn<5^blopG?joDm0I z5%3ye1)mv*L#hJb6$dl=)rQ+7nHA3ii8X6>JZl(US%DMgq?%3uthfh5?~X&k#FBGQ zJZGf%h!eQXjYENN$*GU$jOAG%Kp z44EH?jJt**olS#>j~(6nyrNi$p3b{@Vi8>_61CM-H{zHd?Mi(=vyS3G69ck zCkx_IcOZs4DDa}Nq#tUOg-w|B$VhkCQ^`oufB~Vf&(8qqnvw2FoU{vDE{z2YPN%7~ zU*sPn(gR`@u}G9L(gkF!5sDhsWk#5R4gnnPi3SxYLx4sFYvO2`0yfz2#9>(iHZ~T( zTnS%Im+k1tpb3$tj*A%)g(NM#%J2r_Kxyr`%2?P$lm-tC`xL3Pd8|7c3oDqk+g%n3 z`eRW8jfGkLUwUNof~^yUn!Z5Pz&AO~OsSe892XoeF*pF~K5qH<^gHV6>TXXY9~T@T zB?+|0mOl93-;W$QB2CBn=VVUbzw-QEDNHFbE_g-2S1UN&w`tqL$wJ6Px8I*kZ3(z) zUqFqN_$$hF>YvoH+FhHFe?x*Wrsrd=qs z6wLy|>K^#^V_9^vMaQeTb&U_b{4dEJ9fJTieLnq*eNq#_Hd11QB^fGWKfykp$F#_) zS5=(*2au_LuI^D86Y(r`q)b)!)0u)JZUibZ!o7-$Q1}tfeCj7z9b{&)HFNbtGDu}+ zuQ=U$Sraf|5F`F8#Z50)7V{VHrP*z>dvLbo>-L8#-K>WyLRE&rYNJBrsKQgFhj$(> zjro1<0Igb*ip zy;7RLyO2SkHDa@a!Sgd*5xmWmkR@Ibeh z9Jac4^F|oz%lzc9)we})<7#5KBkZYk2gqS7xm<`?IVV$zKuF0OHyyTycZIhu+i~=X2hbN5kU_*Y)y?q=1Gs5GR1zs!&cgs zj5p^5A$1Fsyz=7b}=)=q_VNif?} z?LEg6tPc77Ms`K8xUfgh!d`u|yGJVBVZ)mpHY(g@KJrKwcFC*IYq-6BU%(hr>oE+k z;pG8}{Jxy_q^A8DCZ=q`Sd_HYe*xF14h6#)j%aJF*hY0II%!a}ws5dihO}oiEk~(F z8un~+g7ORaagE}1{0;w&?egSvQ(M~2y3*~B8KiC1w45qq646aVGA!K7+@7k6a4;6| zLbb7Cg+SY3rB0H+*IqoQbt2?h7 zOU*AF5p;V=J4MT>AsX$^q#S&|nlFFOt_jhpYyFv&9i0@S({O`$HV4re{IkX%qBD^l zcSK${4{I?fTAQz|T4&Sn9d>BYo^=T%%D8(|O`14!7#WO3g`IFkyp(pP{)JL%OV6eC zTB;?s+Td`hz4?bI>_^i6DRll|b!NMEw&AR7eJ@T&b~K)Hc`l6ApkRo!30h8+2uloc zUtHyjR>sZ$A~mtdbG$?dRkEqt#YZ31(m?+k64xFm?G9Zxp(SQ+_sO9mq9?U%1W)(L z*hhGh&-oCgt=4rDvowifwUuJE+`~%yK-W#oP!h!&E5#Z`k#@XpV%8^7JfoFDtNgQ$ z*NO6b;V-3Kbi8gNHY5>j6a>PIjam(;Deao$brZ8OiDHYDVhdlxPm~rrUNL0^a%??br~b8e0<-V1iv0fZ zI{oKlCw2Yuc%5ob!Qx>%>J*m80-j=18E2_gssCu&nI8HN3rSxG7VU5tuQldv^` z!Hg8GKv^{-1y=p@6lIl2$E})?GE-VLlUuv%PnA}Mq*dwI*M(INoGPt4{j{T4#X3pT zlt!g)t7&(oXayeDobF1&0P4!qoOGs#?qb>L^T63Omf4vMTFR`97K0JdVzzG6X)0zG zGgBE3GLh9#i?3gJXbB7Hcslkoz8L8wnk35YR9N`mkxse+S(K_OKg+isqEsr>d6X)e z!#2uE{UxK*!06ObPFiugle(TY%1N&?N7EuvC?X;UMQcq)&0fV2adgm9dV2Q}_r{Zpe!k`S+Avdot+KY$v`?A28^r6javgY_Wg%i@s1Q`wSotJ2eL@tB)#L(Fx}aZ-VY zNocxFe=?KYIZnzA(JULPPXiLcDw@To3%<#55*JE?)a@SVhC=Dk73Bptx5%vfG!ot& zjI6kK(O4H2nR?YH=(p6SQQ9RPbrW1#q14Z-=%`agR~$I$sMC(F%Gc{`JfMgRNa%C9 z7gS8@eTo?haD$CMC9;9-EIC8*+h9Y1pzMlBC33yAk;SY%!zl_U`(_c98{q&H&dq1a z!dZ$RD4dtigi8>@A&Jb_F~D2@<}4I)qs)3i@j2%#$!8pXfX~FUP{?Myw=vHe`^v(i zo0O#@_FJTw=OG5M{~)*{_FrJ`X6z?$$M8UL>R+We!go7_?{={7TFz3w+aV99;FQfP z{JUGOj2P!F!eKk10erAN7w$pCl*lZ+Ho`g`LDo@p3Wb9qcZEA@i27;j6$)APdsFE5 zChOPrY$us5?VB9+R|6!Y{!Nbh#b>J;`4$JjM`t@JtNs>TgifC_o8N&EtfEi3DqNc9 zq_Yd%v1l+_9lEhcZ`x;jZDs~(ODM~EtDSP)^u4X)>P)K6b5d5*t#-aBIVei+k{_QlwW%0E5G!YS>LFdbQ5QBM9D0E%>&-N8(Hb_=DvV%L{U&YQCx== zhwY4Z{1r~>FSG096KKETHKE>DIEnHHXZ)r0EUaoTP*wVBuZNVXVP&6D7iI-PNK=8P>;nK{WlAp zw5`BN>F*ahsZ?t0mCp`DG!;ePip=gfKQw;Jab|WX2@M0Sn>*^JlJ%ytd5b&hG_AW* z@Lr+dQ7VxfEGg=i8mq|*g~dQF`LER*?a`5^J znytGy3oU4?%*q87TH@8`$QJZ1{6GtO5Xf9PoYK zddFi019CP0;^CD^+j@)Q0ndGts+sNTH(W!$C~^|^^1tB^Z3K$r@$d~>a}4m=2|4{drUmC*^wRK|CtAh zO@@soV-)uj+&;`JBPw!C>gEJ4bM4%s&SgjbqnDGq$5)`{+IeEn zH^R-eqlvL7Yn36F7Zm6Iw<`xg&-;ah`&q&$NZ{H1{dNv>K2OplJv1L4!iQL<_ewZV zvI5q{{bUs_+fF5trdUNU5~3He=+}F-+jcHu!&1(5l4v^@*)hes0BI2j65>8r4xOk@ z#PL!ni#WdPT!c8<#Uh#Ym}+vou_KDz>)R!DLYr*^u`l(Dw1qVJE4|X zYEsJ3RI6imS(MtvN``pPGn29->+8rEXh3TlX2*oJU# zrE`l2)~v=l9$zzHof zG`?1bXZMwc?~_E|$!CJ$j(*B;m3=q~&1!gFw;C+`a+4sdtFk;a-RgnwM4kE0K5J<@EzM^iCil~5h{vL(VbR)2L!(Z%+K3TMacnn!pkOqk+$$>bxmGHcUgJkd9mTL@PLh%vC1dW9)Ix2$d@%V|$aUYp?8XEE)`F6p>*P~g zA`AT)t>51%s099EaY(VOPiDPQDu1-1O5l|*m?-*ic4NGXiR3b#i~_1 zC7-~GYq`mzsVy&yQh&}%w&IFS|9g^RX@`}aTIDx`GC?sJmWoA5GOMk^N^4kNe)O=+ zOSW1Bz2aKhmOI;R;=6?M^g zDo!Q$wy!GJA)lB^TYMQ;4NMeqQyV5it#+A67F+`rVaa`ephXBncS{4A7uSN@Z%NxE zpJuaVRU(;(QbGi=A`RW&t{NhSm+~gu9y5kTx_!L_Tg8qZNM_6Kv9kU7kJu(}{8|D- zv93!Nu`{(UDKd8Qa)Frl?c`Z}g6{1IK@ls)t;qsjlLEMm)uM|fQrhxEVJSAH^O9Wh zrj?yqk@-V*vZYj@GXPywgT0VKmhjH~E3lxQ_Z5v$0Mvi?HeN!7gLbZ0eC>~tyiZZnL4>d_{Fs&0qgpVR` zX~V{%h-8wXrd?t?$-wczma*cD9olxyL{R;+nnsbs*4i#&MKBzUMSTH2X4Xp{@_Vu> z)56OfQe0+CB8P2-M8et3xIm9T7%{x^tYK5CI3b61bd7$^BZ?nas}6;YNaUPyzdLGB zr$t}esc!vgTx*-B{GC&4f`J{U+u{rhD~x~{(s5k{aQT1kBVwheV&x3Wnid{HCp_U9Kyg5fA>n^Lr# z2+{V>J5Y|Bi2l`~eX^lpV}hc2MbLcr$267jKz?Dwh?3Sz%c&%CebIq(0F~Gbkp0{ zPvSwjH{*zZK{+NAn!Wos2nPkj;aDg-JP?ZE3b&wnw(<-u#~?cQi*~%I-UvuIi`k9w z%i6=w>*n=&qJs?-XgFWo-{p@*DoI<5W0YQ^hL#SL9Sx2#++JG8_dmCBe-e|4VPDi3 zI4Nq7cD$}jtMI=*Ae_ zyj-H?lo7eIvnVG_)N5VlpK6FxH>90ykrHoobKrTR3H>q%3-Z?PM9``3q+x^TRy4U1 znSR~{I&GJI%{ZB9C!$JRZERei(~m=))V2BoojP6Oq^^@L(CLCpoRmKG0-au!bf;M8 zZp6JwD=yISLiw{7=u{;0`pGAFm+@%0jiT+92nUZ+x^9zFK7$CK2}3AWpvqo%2{tV` zswxmk_am*^YKCmM#EJclBGNcX_^uGgQzh^}yab!vBKmv)IEUfh2oYm90Wc2$8IgPi z)`4MS_pFVt3yVxeOP^B;qpNR8)y@v`wNq#!>9f=~ZqFccz>}?0>^vIqq!3B(wQ)~E zFD|=8?t;;cD@hHTl|(qYzr2;KKe~6yt)v0?kv|;Uuaz_io!AIYaK}Fo9seD}u^oGd zopQ_2VGjY$9NK*60-eqTc^m8U zx>Y8OD)I}4uEMI{i_4r3qZBImWLfQ zP5bJi4BB~_ih*ew`##ri)EItR$GPW^a7v@acf7X9G%G>-nBs>KPCDCfxXT$Bp~af# zvknAMjp2{bd~F4iSS0#=)ecL1}$k0y7;4F3_oAx`-DWj?{GNqtio|J1H-_k4{aOWA_U& z)gp21z(Guf$AJFN-J$A`X?fi?s|V^+IHc-R5K{BTIH{mKY)l}kPeHis2EYuddj3Av zD`VLKu|Sj-@m-Jq1Atw$fGxXroUoq*hMr<0YX z`skE*m6P)8`=EiQw1@lXrVHv#SL8MI(dn|QRAqUARUk^+)kin6ina?G+LFBQ0KUpe zc?bLGlr!E*dB641Y4CVOzg3)wr8IkA-K1-Y)93ZosRCq0eRVpFAHDnPbn?|sqO<{h zbrZEUPEy)er;Dyu0(L6}eSLKkt8s^^;9Y%nntHX9^6u@cQ^#wZl((p_PUl~v=*5Q` zh|(VEtDAJCJ5BnR`|5NJ$X@HK)3^BX&%Qbxdo2|FzrMPO+7u^A>!(xkwMxJlV(%^O zxPH2cl>Lgy|Ga)WjWdzE3FM;&vpIa_0fA!E@Va7`MnqALz$P4oMETj0Nue{1J`5sv<@I2En(2T3~B`^5UU>H3sgmD4d=W9mVRHLik{;1 z5wOdi$#<+iqaAl?26epwtqdS?Zb{EF=DZu6lp8EBrwxqu`VD9lD60{eKHADSwi}_a zOyu6^q%67%JH2SbwlT(VU?|2r8RWmwNgdLqhl}KsCGFz;xi`YT5-@_|QFj@AtX(@f zgLXhbJ>XF5Q$|iLWwqMt0u#shxZ+lf@KxcKzF{rowJ92|KUtfCfq$P;C*|Tbs0}HB z;tZM&A#Fn_Q2>UdCeP3wEAKsm_a5fGq7;1r(H{PW!M{pXSnm<_P8{>vBi;zeYJ#-^ z`5BX3H=68=`{~qrmz@xqH{cv=te;LHWc{g%gRp{P>PF309lsDde!)8Kb1SXB5Q_XM z>#*VW%9RY0ap+yEfc-+ieiqQ9ObOU81mp(%z9=1HANa};RXhOwl>y5bz#k>3vtj)B zVVRRU^pg4ylTSWneSq1yJP4tpS==ey$s`pXCw0h{8ugM-CYi|xvv-1ovg!m)J=2`x z1)D|zTErlq7rM#BelNP7*{GXk^n?wyDc3*`MavmEf!2 z|Eg**lF5JRr&F`c(r~0F|3aO1_??vA<3gPp-xroYFQ2GKD{$`ELUL(Mv~ z8hSb^FVv~O%p1el4_&BJorK%~DeqmV)B0*B<$ZRcP84ub*Y7XX=?wh%?LwWV;77(q zI(>#8CtjpeVGzc`v=+(7%bZcev*!vu45ddokpHwA;)P97(5ir&J`?7Og5ICX1-(BN z1^sCd#fo#JD76zpveXXoxy#-mCplitKbcCUCN;3}!~B!!qY!LF&IBjreVczWy@ww^ z=ATTzOn^ms7wOa>sV0GH;6*wu4kHVuwaBb#%Crk3$U>=Q*9)tsg9@egW<-`+YKt(# zDW8d@RwC)lQnN=AiV;dNHlWN&LBlP&7=V;~ygPxI;7=xJak|l4>ESHeRID<+aE{L4;)1gUCWbL^%sl*C?|d zX6als5`vSJ9Q~apfe(ttd7n`1QSd?0+%riQ&BtE|pM5ExiAB?2(wRl`0a*Td(NqV+ zVwGDJN$4vfMik8^I|uS6E}A=TR=vq4_CC%-KWA?!&Yd*?3n&=e;*-t63knk9KZkbl*=3;i<1N%{LbcOmNf z-_Bj=iaVUtHM2_>s=Y%BqjWvdtKv}hXql^k57jEORqls>ILSoK<-rWJ&Vma`_Gn}V08MfY|f z%FQjORjl%v)15+%I;qAvN{wr$D>eRdS@GKm#ik}L2do+$v1@1Z;@yrQs3<#cSqQmJ2T z>W#w0OX=_YQQ3h36J8l7(jNT01Ko6|JSU69-FGIRlTBcpuw#ggIK{Ot&K=h-zmS3* z#6~dVeQmYOk*XrnTvRjn_3{fxU_(qXDmiGX3x->n;oDHvN3@ma0f&&fNu^Glq2f}m zi8!#N(lpOG~H11mFc%{-biLwUy)K357fBr38y1v{_@ zXDT?+&bSS%zBf}>VwoxBvaSlr6kEU9tW0)`de+p;!xdP za&hQC__3(3PR?0>Y$JPHsT2A+L9KULt)8=lTJN%2S0Z&2$<(A-PI5fbSEqWbCRk=# zGRsN4%;cEuBwl7JnBBf6ZT3s~q-Zc~YD0(PS{X!Vtk~OuM$cCM5KB_&SiV{))qh?- zx$dD`!6H`pDB=^gUX^GivOZCCw-5Gqpn0ZfvA~s%#jpa2*k3*&7XRmANo*5~&P=L~ zqk@BEE${^cylvP9WuOs>jtv=xxByE_=78*E9*$(B!A7||=5K>{dorF}K^PtIiyK2H zCnNZS6<7HJ-rz)Dc}J^zTr$eU#QeItl=d+!ml?J(s>~fRM!KV(N_mr@G&aZaY4s%Bm&1%a9Uqc>+qc5xk zE{ki!2#Qq4{EYCzp7(xeV{p`6<`+{67;a4;g6VUuR|qC)&icsiPEH%i zVWV!sgyJtdv^4>BYjha%aEGnst5MvwfBaPk(g-U$I1@pXuK9x0=^08h4|_Jx|2sz> zeL=GOf`wdZlqn|Pm^dkCfnbQYGs1FL zLtAgx5;K^4cwA5+Y&j|dV~z`w_JW;nqPxE_lX51KFsXeLj9|I&kJ$Io92(htLEc4_ z(@AyOHnc!yD=7qL|8H) zI>K;AqNb#}rB;c%Qg~jzD+MbAU*3%!tXMhGF5bhs@E#`>;3F}#i>1DJkI+G+qz=2J z4*eHkyCfi|6hGagbU-wXGNm>&SfziPq7`tAeajYinG1^%Y2UIwQKSPz-=-keHq3QW zcQg6~OzQBh&>=myOs7U^U#ongaQxWT?^CAJf95(VZ(x~D=hWjEz-47R9h7)oq-G}m z_%fZ!>z$NWR;JU-_~9?p>1+JBxlE_5d#x4@lA z+`b8EcKbd2A?}uYt%lxXgYnYPdwAo8xNxz?EyXHvEst5nE>L0@u-MG|l-LCoVyVhBX1kN1-0E zp$DEa&q-PR7vXdcJ!)IJE|Yf8gZHwU0Skm_4WsR!$4i{FhQ;-nkBf;OL{v?rb&Rlf zJ`j*X+`v;sfT2du~C3vgu;QdkjNyg&r_ z2P{qc`~wz7ls3IgH(52AEX5lSGwdffW)kHsF4O6Rg=+dZ!@4FsgV$||(jF_*P5PE3 zKC>mCrZSx(CZE|>KC>ksRB+Yc!faeyWAbw@x0*a3jT3HxoG;j8DQ-S5GCmFXt;Y*UKx!g~7{ctNZbtruJ^>sb~1R$gHGyDDs^a(>~aJ2D!XL-VgzJ3Os_K8 zEsGJ6X~R6aDY-?GsC#l8JB5cr#Y7Io9OM>}=0H8c?_$y%h;Cs}UbpI)Kk9>B@vxRG z8&@mi(uv5W6;y|bTpu)Ml6Q%SY8M17ULu0pB_u#w&BYS3N*3EKcY76|OQ%6QZAphy zRteU%@|hc^6TtY(B`}a(ounCeZsUTcQvKHm4jL<0Ae1Wrk7FKk5_-7;w1pQvdq^-0BsNXGXHX2sg&uao3&m)#4?T?2dB_Y<50IU^G}$fej&Qzki>X^lqgBVTT0xX( zY~-m-s#+=vY%FxTXQ?Q#v5+epp|Pw{huU?50ZB%=6KNc*`iT`liy(_7LB1SwDCs3n z3nrpyJgkKWT3ptE9)ie&!C0U{XyLJF;bARuL*#{?uQlL+(GokfFrdXrk6>2_fXGBD z$57F&y#Q%fS>oy{jw|MxXH^Z&p2@KPh9u22R?K94 z#^WMj|C}+GT6kPg*{(P@CQI029~`$`^-c-4D^^lf$)IFovZ~z6nW~QZ|4@~Ep%}Nc z5x|O?0#^Tj2w+pj1x?t5z}d#Z*7{YFVJ=!XLL3EKB+GfAVHQ{t~)R=cB>=kR7}r~T<^W^oif^mmpZnBH!( zDwLn0xW)(vBj@!Z?JO;)mI$BvC0iV8u&?I4UU7+AR3KnHcW0_rQ0XJOGZj>Wm*WU; zP$esnoTO5-mOH8YgjkR?flS@;?O>J2ofIgd#muE?xs!5Zfu22R38%jh>Ea@Kh|_5+ zIKPONaxx#u3B}aFsS!wpE8S6A#_2ndjs0w3<`juW*v%$*@i(4~cuv zj>n0ISHn8FR-#U1#=Q{p1;ndmXf&IwIaeGt$7>7>$9y-z6i_|&&4Ioajw<)OUYvW^ z{8djZUc2{Kn+2s>d>8B8Rq{Wp*y>^?d*-X%i|~I>E1CGT+T4eqKivA+wi(A+vwAYs zktx5;XkFTLB!QvOMjv(RmXSuP`0M^w;K7X|@Ct`Vh6lah*yUzq@LL|bbnm)8=7_7z zfCNEdR?HPgDG$3nN70!baJnWn7!j->?Q+vn2^|^qmXmgDoS!ccmF(i2CGJ>XMWD!0XM|Cxb_L*}~&AA)k%s(%_I2O-f-oyFrCf zv)P*D5K$LCInxED*lVSex(hQnP2HJy7t|xlt)(3J-@npH>1W1t>MhxflTR=^6+6Q{ zV>-=U>7@L=F}z5Ueo0KH@e=7e`81KQjOp|Y>b;{frc=X1cFxKMpIc%&y}!~)c~fIL z{fZy6V>+F_3QJh?V>&$`c`cMr@WlU={%lOA)e_n$pU622=ZvIJ{Oma8i{|H53k#f1 zS^>L(I>9XiSK%th+7MAEc;e<&YB0;WlxxswfNxvngp;!2q_6P1n*IUubDoqA`#@@P zP(C5)WEcvEjeRoFVTUEsZ}Mp(!(kJj#Gx}dtob23XPpK;Vl6FlK*N`Eh=Q>lb{x>`Pw+k!h(3WAd+$^X5FRJp&C8F_KK`qgEc)QzbRXUuc z-11@yA!F2PC*@Z8h`;q6MxvZz$e4%URaArD&*683CgJxd{JRI;jD)rZuYC5RTaY*l ziD(bH4T<4MjG@W+UA@LB!hN5# zolcS2+42cA{(^NujZPn~aZ+Bd8l8T}kBe(`%2^BkDAMmFL92X%5T6F8^h;}WN|iqB zD4)o^37ak`QI+YGnW)L9qb$t&6RhZA7DJQ7(v~1%PUYe_6&@M9Ruu;)DVOWjX~?Qv zi$$c`kf=R#);fuey-udY7}ow6sCdyjrTsC3`CX`5AGwfGwhnc@TqN#X z=Ojm?NVq;iqC;VCEOx{GNda$e+wLA7Y2dOQuai)jXT+E;MO zpc&k8uYVdZe>}`fKQnok%k>Q62(9-t>7beN2_r9rkqGBYOtd*NyIwvM(C%ln2Wxa% zEwi7MPoP}_Zco%K(O5OgT!mFkL=Fx>X7Z)X{!Tu@WT0YFp_uHGxeCj|1ob1e zNq9Y7KEY%VWN_{&5}GZa$Q^=Q2kLH;6yb=2RvqmSc%nr$?iT1A=oyr*v9y=_v zf0Iw3h1p~KYIMqeSr~Jsd?L3DvJTbgG>e%4t(VO1EuTOuRcP-q8f4YU>}m1|w6QD; z4qPFjo8%L@7pdk}Wj-r2;|@$eQlnFeY#y$YPtZON1&*DlQ`b#S$~$qQPD3}Lk>iO+ zVXP+L7AN4Q;HW`e-EVT1t91$~9SPEPz9wKB(gbi=ZV~nc|8lsc_;rBTtd;s|Kz(!aBu5ocl-S6@fOA^4Er?ZI_V*;-D}?+q9eiN6s_ zwnWi|{9dvWtu7u>7mhe?pQ=;cQag|Prs1xL1yfNIC^Lym$S|tJ!V+)PA#CchlaX-5 zaQMg}!wk3Au;A6z*)x*Z@zud4?y$QWi%E6Ih*K*mCuWhBj5;9}j2c5C9(TwHlfyPs z;=ZvLi$^h>!$yUtbY487PVb$ZX7=rkf6vkGXr&ym;Iq=06skc&^LEi{3Adf35&dq9 zmvsoMR^>jwfgjZA&QsE;;VDtn&3{V+lZPFgOqTfQGJkRrP?9HD@H>t%JTjmxLfS$N zU-rZ8UD#P6yDuQV;%p-NeNjUppOJ{0$Z(L~a0f^`|BrAk4f+D4-JxkYhTo`G;g`sV zqfdFG1!5leSF5f+?l@rt9RF z3_|h=HjkhMzkRw+_j1FXBpzR*h6+k_=ozIhXdgFKc$`&u&Qw99r3#M=6>=wfX(i+R zx(TE1aF|wdPUUk>N}oMlr%h6?59AXvo`8OjP1otg=bYqdoUT*rQgIc{BIGnr*NHYe z$?^NkHp*_W)44>%Bkt^NHX67Yoe`yfJYA=cCDRu91oP!!ad5g$(}0*xcj{CmE$lC! zKwE=Wpv#@y3M4t`CKR~xifU4{CAiZ*TPqMjkc|%&@7%14Cz}H?tE7R4xn0q2T(|`e zp{~d&-Gb2tYb;f0kxwvX)1dKGpiw3wzzpR=^V{~^8l97rI|7$v86kJr9SsVMJ{d{$ zRld+*AKDbk673IyM{5UbfFH}&PAeAGv}t+D&P*fKP}DX2ttK9kEv8Cnd3ec7aiSg8 zrUzdoTxgMuw|1!0hi9glcnyC`GSl1eSJYG`EtDWt(FIwUb_&j_DlpMCE z8^mJ%O+zwBBiwj}AVPZ_3;1y2@SIq*yf<~KJS)xQ+Vr<{WO}=mTj7{kz!0a1wU2DP zl8JXU$(Dd5PDKLM6kPg>aweAb7obKHE#QmgUf$xQ?qSjvXkT5FLDZd36=@52rbfAC zJ!z8$8Rx!$dxDB+38%&*RZ9A`LI}$}nnWia|>RuD4Bay`C(NmS3r+9qySvc&ZlIKOlsA8~tkJ1(JF#+|_Z49VPqqk4dNXr2vK zYrqbymC#~Nc#+^b*udyBfLqq zWyYJbir%tXx8BC_w%&%mDtD*Y_g~KnQC1No`Di02jh#-)Dh0$tn^^YZoovpFoOlrl zHfI~7e+Kk=fxe3|I=_OmJ_53vA$M<_s}}VI1EVk<7~^*3_1BV97Gb3 z^$fY_HDya?Lf?Y`<7-fPZ1VdNTifihhF>lUXo;C4-L)WR$|Fa%C2_uD+|ZHkTFy*O zq>IGD;b4V3YK-)Wb%?sgB(ZS}$>GS>B#dY=b54>O74ni5jfyFIQF)& zt;yhXS>oX&aJ4&9C8S=QXeBa)+|%1;2+7B_%@DGWNz4ex0!FdviX@wn*~83CVu{S2 zW@d8f`h%V_HY(X+l|GrVHObmqVd@V_V7bDR?6Qa`q?AN!BEl=jCqacvl8QMZ{BcGS zt|*FLNl>wflN@a4vxVC!BO&g~ec?!y9JcJ%QG9;o_g8RBcLIvr<1s?fG44PG-{)Pr zTVQ_l+_9w5f?@G#btD@01OqkHsdu+DvmvN^{O=5`6$`AKfLEa>SX~_ql!T3PU+oAZ zFuWoV3>$;o5rcT|TTHHVLYyxW*#SBM?Mb$ePUNSDPC`o;=jjhl#5#W$fQ8jT`Rer(>^z8aN*@xG$HLa z4Oerc^NO?K^%R_IC9SiWox@OZNpcR4G=c&DBqB7dITc}dsM6=*mueXjXEr<-9*!Hh|0_n-W?vn%lqRhvC2@F!sg=Cp3*nlM^nU*$)t-^24jBj zm|!g6y&^=~o)k^-&Ix#v#pmI2HD9&cpI>NF{M-&jwvT>iIoTfiUrJ)>CG!{@oERYO z`*ye_7Na;}(S?sJEW0ej;;!nLf8eAM!HOOuf)&L@#8-LOCadoC)%YSlylOhB1fg{< zX^*68Io0CKQcjh?&-T$1+;PPn>x~DCCs;qeR8e zWKxa{UC#FyfoRz6Cy|PUu-qno8$HxX6BS-wUrlw;TWnlF+Eb}~Kmsh{nkYMqi(o8i zAEYo6>Q5XgQtkHp`3~zg!Q+dt<)h2TO$?Iux@MuqT$zuE6^;wLeSTj+9PDB~wM4iC zCuy?Cqa*H7wXLcACK}K1Ok;eJ;eks;h-1-!w3p&o+2wF1$xo=5BxMl;6ZLB=f`ySv zcgQHb_IjM0Dn75kgoyO@q#dwvEz4Vi?dwu0%eugfSQ^EjT@(GDpZv zoaGRw2OQWatCte{L(eZ_I_(Dk?_)ace?y)IA?dtS`2+_XmcjW0<#0=j7%`%zzx#hA zG!v82OSOVBqD#@}>u-vK{+FT<&VAD<9Db>AxQ{M_Gox=}>A^kG8;sI*$T??^1o$d& z=k@h~8GD?R)$|0e0Hpc65V06o1{cyC2E4Tgj_NQ^r^XM2I?u`{9K{=$rI&ab{~BD?!}@Mg?-9G;s=$jr7xF~G zZ%)zh-R;dOh~rD%aZ&+yo10ULpUI^A-f>d*GQ1?q{a*KAjS(*QuPU9)RL@xTC@H^LiPz zZco%yyZH;N`fF@jff=c^h9#|ePaUfhuR%5=jd-Osyc&@9zGx-a*hDjl{zBv=iSF4$ zh*ly)|F(+WlOTEzi_U*viQdD#+X$pt^d1&{hZMcX#sel6Epn1X4>M{_agF<|Vt*8C zH9zuN&Ew$D1@@y21y=R}-Y|9tM9rliS*a!faUM$c#^-$KB%#!gtW+)1q6$mMdmpMb zhaaVnCS%*_NAc!nZne7>yN04m|K&rx$i>5l`}r95DIckVU&-l|7VsudtJih(Nm0_r&Dj~wSMx6`f?8ba6d9aLQCWm8Tg)mQZe-GVS2y*Kh{#% zAeLx0@Dk1Dk3~RjNEM|TAZ-7w3j|f z{RQ)x!+rQ8N&`LYymeWg&JrW3T z+Dq)S35-!BFy3JIz5m$A2Zm z_<9hr5Ul{j) zQikz2?2a{1CP3e?jc*|ltfFt&#DjYgVe#kV_c*r59JH@R82`p1&iWb*XtkX{Gm~kC0Rv$?lNnY4MTAY!Y7vL` zsj!**wG5lm)k$IV*4Hv@j{gQ>!(rSY$rpee4ey!XO7dn&ZiaFBw~~Ctw@i*9x^sr)(G0hPqP>!1ztLy%w zgrK$WLPOFhE#9mwKCoX}+?s50uU2VsO>3gXA=6^9j%V82oM87>VfR+Hd&Un|ySK6p z>wZvnZ)F3e-69i-9)z3`8AKhXN!APHGdDmlF_VA(;Dn1_vZIIl2?==RCHBgR2cWoR z@-!+|&C7S{fWFl@g`-1@WC@bk3-tdO1$Q~f_;=7?od zPQHpH=!wpUt$~9|$oY^NcZ{{$wgw-DM~%`oz2N{1k2W2YA$G%0n3_B0>qP5>`xxh> z`8xfMtn{>VoODoHTqKo7SuxrP=Q!z-LkVRqtC+A>S=sQ5)!M$o+P<)M%OPbgCp-Ua zjn=*}_=TU9!F~U_!9~&-zQ-5_A9Gk5yymbnxXy|LgGc;=nH3Cv?6A&y0@0f_3wZHP@s`g5^pQiBa_)9>nljx~D!#(*oZoQ^4&gDo%if9bf`T92} zWi_tj+3a+l%|7zGld?jMcFGMD(?d+L<9B?KjPB(8eZQluc`AEY7U^&D2^U?5@WNE~ zwEw}AX!b(C3Qso4C(uR$4UPT)M#I9zDZdG{Z21J*-9Q7k{{oHsq1HwbqYcPHKXmF5 z*@$L8C34Ju=)EJdz~}Ni5=>qqkrpZPhc%a&Kh((_ll-a_mzX~|ZFt2(G~p8I0gyxg zt5PL-i_9^{UQW9t-(@$+olgrD>lL~47)jn==9uKWj*;Ydq?_a+iNq)6AbpoZk~hd4 zll+bhN!~XT5z3yMI@vP3|H+StoPMK+HW+ zI_b?5!CTZKiNxPbgJ(WDNs_n79Fu(d$&$SLDIkw~FW$jg)Z_5w4Pehz`|?D?YRLFUL;g{H)kmQ?L>6BR^TSOeA?5Q^uTE@DySj4 z9PR!_Xc(X?(0|bBE;<`G?Ton-C-)p@%-6}eUVN90U$p5pU#IgDJR-xSl5n!AtRbBJul;`|vwJ5#XGb?ZRX*Qbl+2 zm#-?aU6j>-J)(@JFl1J?izur>fMzjh4T!wH3YyKyX2xg*27nJR@J9gIzLE{XKJEfw zU#FY|`}{_^G*6k8eVo<2yHnV{yLolomKoi@H|L*Xc75viU92{PTef^E;oBV7_IYR}`|czB$1QTZ9+3 zuowEAA-o_~Z6|=C87>xnyqj$D>F*L2XdJDd3x1z7zl<|oRDg}?&)HYq&vcO^hS1z7 zimn{G*U#6fAhjVN|X zeAp7|51cn7ls6Zy0>qZ*q7JL2k>AN@?xbNp>WS>!JhT@^xWe!b^HDE=hQ~o8f&z43 z9@p@@IChwiZU)nb zI5i?DQokT&b@OZT{0M_)cXLsP{vCy~b@IuMo(#5YL18-jc7WcBgGK}e z=%jzZ(O7ugB+YMO90>Eu57n+0x(qBdL2||omEso}IpZYvQaezj- zx9KFIxwhmOBDz3c=6S({ei>By5A=R+KhXt<*R#93sGz($N^}9X_%A~;6gE60HmFx3 zZ4Od-&w5Iss+-J@awE8J$cRKKjADAM&_%htE7b>Xc?C)ApY+A<$?r&^NG}2Ewdc6R z{?;%Q_ZFnEvqC&o8&KrprMM6p|ugFCmsFSd9g?#2lXcVSlS;goo zN;la>UmjaM6knCjYTjfgu)u!VG#>j+E0*2;y&CuTlz!3zo$92gr^zQAh<81S?_p?= z(ADyZTznbXQXXZ#3N8y(1}&!wJw;hOs>#B5lwTG&w}*?QzaM3#HUq%r@hDFbZ|kWl z!J|C%e->%>5GDD>MBh5nT6_)A&Km?w6 zr{Kc#VFv>96?Vj;^OYS!3}w|nhh6;I5Pis=`{jH@6SLUEfKz%WwEMh+Hde)|GNJBF zt0~{1bcnvw&JQQbylH!OA81({@&{=iqg)R_42Pt|Q0Xb2JiOJ%MOn_xcFK+UXgwny)7M2=4FFa9 z=s5;m*w=;E5$FX@m2!$+WCl}_sHT@W@f;E{+R2IiNCd0AID;m>o^n<{u4=ImeZj-@ zVg2AZ2Sx(fTkIU;d>=7J0#*IwNMKh#IT9$@B8DO6NZ>m`j^Rh0B*zP_Jp8!yLKpGy zW5tDX`0+Y^VECb31gYjosk}Nm#4Mfcd}}HFiXC-a^!;lJTC_7SQnB?bt6wv~Mc8rw zH3i|k2q}&V`3cXAfhaWu*01c3fdf@w33j4MTFAT0HxG1CE@pImYW=B!F3M`&!tWl; z=KNuUT$Ec)b2zbkP_h9-O~qUPVKv}sbA!h)Q#ik`Pajm?~WcqrPx5I{cpea>?(aZy$yAf8Hkg>9Jwkn;up`tGYt_ueH? z-&*ynqXSLF`**h*@U}AGZ8qSSOOyd`vjH~_H*3z@JW#s>0IoT2v#i}nbIp0%hDK`4 zrOG71Fj0%~ru5RDRxQ3!T71J=+;yqhmVU!hejTAo`Wv>n$4I5-H>{=)Y1B@lZ&=Ma z0I-^ZLDx}WfC~N%tLY!5)D#jqHXh;g`*TOhneLsu(|5vX7j;-7eRfzrwp=9hW|vNv)&J0_ljeziV)-U@)9&sJtzm6KOixI3h=w*<;dXE zBE{&JFGJevqn==T2PX^+czXhQ6Q_%)CoumH38NNYUG0y_;hAHYQ#9-ziTUjSFt`LT zFRmY;K}g?(G&kEb*dx-7;sNe8L+AJlwi| zflmEjv@@hHhFiBT(CMo&F3Ov>K&Mm2x~S`%1v*_k7IndFi>&TGub1iO(%x2gU&HRc z2JViHHQjv;20>eq=D4~Bj&Y1L1C`SSkr3_{O9v}W0Q-9ecWAo-6b`?O0s8?^9-qZMw|=FII-Gf) zu(U)zbIZtp%wbn5myee@QTc?+`LYgnd7X6m6y@@&E2YaHxl+1(2Ux{jF2@~WZO8O< zXzinjt(Q%tWi ztmbML<<|6&sYSqu8!h9^9Hc07TQ9bn{e>|53pRVt)i4`}!sK%A71yY${RJ!hJ<{BU zeZhki*R?RwPy1P&UYzpC)MQS@WU3V@qX+G0KE`z}%AG(zabg=1UO)Z9O#aI$k4)Wm zJ>oJZQ*;AT{v!G>m%<&K^r&RkjmV9uJ$slQ2@SD&W(kHVL`!V&%sG{D&n#gL#{gn_M!;qQCOsn{F9HJ3 zJj6qngPeqCMEXJ>YJI>fj^EQmHndd{pNq`qOMKU2K0uyPMy;6A(c_G`8_*hhoApV# ziCegLIH&hbm~_0vE6!^;`&wjs_|UI}Ouh-tTr-erdeD1}wG6PhlW_G6UnyD=oLF3B z`ltU;tDpXZ2rE#ouqCPV=i0iXWC0 z_vl^blcp(?N-nXQHY-&ts3w}l$GM)XlBUh#QQyZ%PCzDfKN2@hU=s`-LYhD}O`FB1 zmMGU>EUy9X>POuulBpp`RTj%Dfh##xBvTI~RqgYVxF7f>Bt7nMc#;Hs>37lDaM}H- zH{6s#!Y2)~sIZ!X9T@ne3r45#$uZRlJ{ihu2E2!_+)y%1xdUHQK32o`N{_`TrdzeS zEqW~MA6c!2*vDcq2dg?deuS!B{IchOF+slk+ShN1IxZD0K?O8-a|>LH>Ks_5X8*^We;Cd zVyP42MI0!qNa!-FqIq(mC=YYohXOAC#%mt*{$J4aSsr|r6H+o5m-qlF4)^p;JlwZ;9P+Ywsnn{e#H!tRsoi*JR~&XxL5yfTv>Q-kYBwI*H6o1_ zNg9u~;%y|wZ3#C@(4iU^b@!6Syr_GGwDAZEY)RY*4ku1@(b;agVcWBrL|M*lShI=H z3?3l$om#hv+Q3IJZeFOXb4m!J)$9%VWW);pI0v zl{;I;Dz_d-M5k*7*dC?R;nuSzx%g{er^6eyx4Ni6zWRAO7LrHb=Ar`2`=6)7yLSSZ z@DeD4U${+02s5S*>6Z(sz2!4EFfdk5XF%$~+u)$=?f8a=&l{~2HDJeNtjquyru?ri z${^?MurD6ZRs-<{9~-HRYaJ)VX>4R4X=ERWyZUw)6@-cUfN8(FxRHHeX7C$>5WwC?lCMIaVyxsbPmWNymqwwlCDlI|X{1)ljG{?kr=yOO2@a zubm=v9t@o~-4WM$Fm%3UYGT3gy!v4P6LeblHj_O4^(JO*s2^h{=q|*=6U6gm1$D>3 z!&f#b${@rYb&tAzejtDeb;t2hZ7QE$pUaWj zwOR80BaT;^%d4hY_hNM_;3aVtPvBk`Wo5sHWsX=mHSo%7hx-s!fca<{55aG}A7zu( z02uJLG2Tb_<7`l%h_^E1KC5#F9ApNJGbVBmOAr^^?)DBtyh;EmK6mIO+ILQS5ycM zpO;VYc^?|09e!$wY=<{L1c#ap?C?lQ7#~yDJsNYkyc(m%HSo zCLE^ashGV-A4b5?3Z5D*KtjAm-^hrY0q{vd%TuqNoI#%m(5PTkh+m$1<5`(>%2Ko) zyLsbgDf7B^sf#49N14}T02Hr>nQEB;nY;wFWhs_U(3Q~}B5qDkA>yv=yaw5o#qa}N z*+Tq4S9Y)gR3|;CQ-4X72dXg->eP9ei}J2{P^T-G;ds)6I!(fl=z}^f!H?S>)amdt zkV9ylBrgIv`p=UekwUM*4+y;tKOpqOM-)|)q~iSqQ2hoR^txjnbrDm=@B>tDJ<0`X zegI9qvaQ}ZFpIL)EK*DeH4Wc#(r~8y;G^ikNF@I880PAviEfiJ6+2(z_hA%xc>7>n zg21nuV4Tqu=Thyi73Z=5?`LxH8t4lRFJJx;8C^ck2#oXvhSY|N#TQ~8yX_BA-FW*_ zqW)isi5eJ-izXs0>r2W|DhESRU$yUM++mo#L)<}w_Es|yLcxhqcNsZs&DG+H!#T5Y znT?_^#|y;g^$PLQh{M(@kuIH+Vd23@J$oo5XFzbu*mSG~X-GIc#vQ1DCeDQKE-6t% z!r`%e`3pI0+3jNByq~Aia7S^SZjqF7NrIGioJ$THomE1JgHIE1rP{tgc@Q|S3->I3 zrXBkwPIkLEDF3*N>h=mgc3kyi8YgRS@phzNGBZ->C`jCN^z15U zxME1H#~<_h0u>|NQD4B6OfjolRwZb#iZ!@qh0tIXYjE2NrNJtT22ZjE4J)JujjX|? zD{*!i-oQIIYuJG+R>BviTG61XKuL{Nk+(IiKt}W1O!)drp~c&*#rH@d*hS+^eW^pY= z=4uyp=qDMCm(N^3?PoTHs}pqa1$XqfFT#7KR3K+o#0=B@8gHyDJQh zVt{!gU14Y}1JdIp*D$~t2aJd@B-;YTBN{bu?RKc$?5iq$QR7C@%{gqf?czXBZZ$C6aR!AHv>tIOp>Qx{ zgrh!2$^Kr9Jbt%cM_A~Ncziyt6AoMUet~f|k4;o8qIDV_F7v^dA}Qvw1PqW5az~Ad zV0aP+Fm`d1_}onxDk(PQLBDqdIc#<98pFA`V0`A$v!Ou!(X(UZuq`}l_DFKr8jhZe zVc?2(vtis=PuLfN`0+zccVsEbQ_}VHk6!(dDGg%g}Os#5b9f-4~FbPH>m` z3iG;+n-ntg3kUdp6}ZI2rsYJ5&gqG}{S=$LZtgPQSYtwdVI=DIRLR$xwCBZjf!{rs ziElvOjCGx&49anX6WRbGq~qFf25Ao;!*|{xF9!=mMeU1n^UG8G`u`}>9=FH}mX}8i z$vuac1x;qC_SrtUBp8>R*R9MKz0&YRgW-I+giYEzowb~@8lvyFWKs?)0%0{e+eaUY zch1{2-v~v92E$j1mCtIvrjhOU$*T(^6kr_;@g7whXAciV2ZSre7!?D3(Q0=nzi_M( zCG8r%(e9+@GDtCMLpa8$7!aNa0PHtv%QMaV>?&H3`Q@fe$_~-W%y2?<2nyZ=)K;8eC0TibBw2Mrdn6o2 zq_t#NX+FyUO}v~1_d!97Cu&J+bXbv}a8_V!+e?{St2{GJzc`OGiwqo#U zR6LWpjg5wR@VV#ckuW@RB`=C<567)dVuPmU#S<+}jcP$+fsor%HKg3_F-Ut*GcmFa zTC81HnMv7EdPqa7lU+{Bw8TKFHlme5FS}-t+wUI~^cp46Flo~?EiON=8#i|Oh2kDu z(q2yCmTb{W)|f;zo3=LFG+a4~!fJ<4ETA~;Y>!if9X2!=Chg7RSQ4~vhYv)O_FgLE zBx+;o%#~H!(H>4TD5~g@PRt-#2$wkA3atFXfVeBYrkTl}X)hMx_+9yad*d+W_C>1w2 zQ4;iL<#*fL@nfi}qQo7oWQ6tYAq0$xqukX7X^l1wofh^qa-!}Ean&&{pbdu6!Sgy> zj!|7^c)f;KB;zp=HX=cPjWI^_^n&4%RA%_fdDh|yMZ;J#c|W5aUAfh*jzyva4btA% z%!U_A@Mq2`DWaoBZFICOkLN@w$BXWczE-(rKuQ|Y)(Zl}ttgK~f=DzNHpX~r!lcbT z#%!u^g)7><>?qYA16kQVgecvA%!uvQ+UWLrNqg}aP2uB~TXa%66Duc;b(b5$yK(>V z+TGA_uzCz`qTFly)2=TE6L@G6=Yh*@xi{$9!{p@DMBj{iF zL1uY4SWQQlDmkI|a{Ec%qO96UF5GaCw5!c>jVs~~huxC~L`iGx6nB{%c;py^*9{Yza6Zp}ef&XhFKO=|tL0P@ z{WGDc1-H>P245!0eg0mj$r1&Rb41xyw9kq5I9t8p_pK8Fku7c$`_2iZxco$>Vpwo6 zkt-Kz75$XR^{|!eFHRuExr$B72x-es=Yc9d{}a#U5)##KNUf~(T1$HlG%M2(d_qk7 zQnV|mNpWttdDsX<#d(SdX^lc;V#wwtH+rLD)g!r2B<=HKxQRfs7w1|Ys3|7x;|x|G z!JOmw;q4@J{MlvnNd^YGadhqqxYwxbpyfmiPt+F-hzJ&S9(!VBQ!&bm9cq!I>DI)> zj`TOa(y2wRfE&a>bUClCFsp3Izn^@9 zRh6B{!m78-QYZn6GtvxFB{;~fB^ThS)|wAw{tY3?jRN5>q zR;weTyt^`qE?a-PTj*feEYO-=T zf3Hr5*0?C|;=MX`U+bd0OZVzjyVgbN*Y4HnH_0k%vS1CC7ivU#fxS8{271k2onBmv zWii&FqjW+S`2@jSSOFbA10M5L*?Gxm;M--LiM48V1nnl_famdpCZ)L*!7QV z8hrn_2It?f4lA_is_1bZY0g`x3gB^#OQ8E?+`RI*Tz(B!$=6Z;wa!Jk7|e^eQ%`7g zk>mImk0W|Zl+8+5)%(TcsI(EsIEL5SD63AW;gnAp&uIQy8?6Ew7eJBZun`H2>><)VDJ9{<{LP}ry#z!PcHHT5%uZo)-7g^PBwi$*;q z;&UhaXCYE@?npZ|w8I%sE0^sQUh1&>4#CYiRXlT}QEJgn?2|#`p2lgcy0>r|D@y;- zc5Tk2oloQBAV5C)p4YZ}KjWgTRzS*WzxK&hnY8d3lr=eC${;HGr*QEra8OGI(H3N; zBm57`lKxFT;g#2chN{s0S=5r-_v%zCIlP0cNB8Qq^jR0>J-Jt>ukmBkUY&ZbcTxVf zy*g3X*Z1mFwH~w((PX6s+j}?*Y+Gg4dx|Y>kZdpA;G%r6weoym19*b4=xt%q2a523 zBJ8)Z9l|>|BBY?qH0j`(@(FwS?jb1lCD7U{ezAx$JP;jamV3ip8N4kj%741J*JFAL zs>qd_Tx6E`^c0kM13+SpMF745AV4&o*UWo8hZo19;gRkTK0}`vUgM z-WBzEmF@lKGHLt^Tvl22fcj`Vv%BpD7jD3a&~|2b0A$(kF`3Y82ea$;A_`x?X&18_ z`69D(yt`MY#`i=;;WJH#_UaUV(M9=3xSBfl>9h*Cj7jec%!kk&0rRI9UBo!ux1!oT za-WS#B+jG2dG$UU)oyiB*XH|d)VS3}>Az$WHOvx0^_YBO*ZFLmY3Y=i7dfuH-$t$Pi#O4`fzK`X+vqNkvpDA* zA+Do*LR=4^LEKACQrs`V7ai(Bsn>7vi3VT=M59AJMH=WM~>L_>!<_9fY& z&SreHD+;fp1%D;jqFq@FG{^D#bSnBlIC&MaF5aipzh81u*Gu;4bj`~y>NZCF@6vJGtt7FR#y z0PnipMFr*(>xbL|3;_(Y%@1wiq716s?jpy%`*f<8`g{cIw(Qes_I4NLZQG~QTKsrp zpH6RWM@X6Noi89Kn@Xdi`>dL^2+dkpvxAVznzgWIop)F@Yhlg$04C<7Ev(!1JEU$+ zQn$~PZgo4PZja&zblbTjK{u~qgm}^=))kdzO^+)LGl=M9t)Qlu=wz%We+}U^J!ERf zPTch3i|}3uoqWuTnKbeh)jXffyWVG4y2D;N`-c}ZWh3ZyN8NO&WBk35HUJFGmZ!uK zklte<@ZqS_p^mePGKkKe94;#K{Xf>;JU*&oc^|GxAY_sWB+3YsCAkdb6&oSB5+8W9x{H3G7U7y+-Ss8R6>$`Wux*`lJdN^oIc!lKs&^!+^5 z)o0Fx;QihA^Zg^~>8|Rkr>gt(e)`l;KtuZq_ehb3cO$=dPOiEwjj2N*94vJg(bc!E zLru|Hf0jZV?@s*?{}J(0x~X_YB3-)%FT0HkP|lnciPZaZw7c0`x!oOc z|D_2uk5kH3N&qQGIHgLZJj~;x8^7R_rYiE*7m!iAwJ~+TH*4ux$3p-_&V73g3T$gk za6f+45A%M3*PonDFItbyGi>DI{8J45?=V65+0Ti!WW`- zBGT+d^oJY>H?S?kc^F0$d8wd%_WS;X9+bTp$y=G@YeN$aKU=g%fPg=0Eeo)z&8A_8kh|=H&fxT?DvUiPyHrS9^Are?- z{Db=q+Vs525I;wSaMZJw##-Ue13YM4s5K^R6pb~Kal3@|G-m&(Fc&~37ft>PN>iDH z&-* z$GX+%Ugu{AcRT5&-DsZ!J=Wtjc#otz?dRCnH8OTTt=F;lD)GGz81$QTf<6DGK){B5 znB@&RV9(p#xV4wa$RZ0K)xX1i~x75`RdYd;g>dRu zf^Z3b0O2;|1H!(Nub+g-XCeFxFc@K=KEk$NGeWN4l~U3NfXcDYpuQk7sir=u0;I^W zsn|WnqZUPZfMV5!oYz!wIOe#BDA$qc4-n;I|7QEw7>y9+qS!}o(lvsklwMFo1+A7Q z(veM=&0yK65PTiJNul&Sf4+x`F@GG5XhA96bANRry@ZH+lsBIOm}zeLrh&S{t{N`& zbcBRJhWEmhrCGZha|&O_7eDoYK{JJKfrQ}aCctLCztl6xJ;rlgi7)S(pqeRa)01Df zyI{V_U-ksUa1C<3l2Bz(a9M4j7jDGz&(?HwO%=`Br;;bqprzHe+Nv z&dQOyNE?6!*}1U}EI`bS#h%%5OUhh+sRl>%)Wrg+-{B6if&!W3UW%v)=AE}ioZtUl z3U$EdAlAPZ@@qQeqCJf%y{Nm~MKI+%m^;Qrl(k14NSx;&YWJdx=)Lb!sNF&r(PsR3 z+eOrLa|$^>bP)yisHxxN7$PUnFD1c`85u@FE}~iy__Bj2<2e^mx6LU^@-X~>91r!&#MvK92*NN&HP}-eSg5P zN`T(bC6V!;0|upT(PdbpydRGj&#R=kmPiQQo1h!7pI^EKQmxsJwMpWU%@p+vhb&Fi zmPUx#+{l5op3RN0jWWX0zPZs?mn9ITmxO6cqxqL7&_i2ND4q@))JU@5FCm277DJSA z-a&)rZAl^LltiNPy{cfZcaVBQUBaF$AsllRk8sI}sY%*Cm0=`A)g=;r0);5%Ig+DN zLgYM-<_z=ty<22npSTrXh`7O101@Bs zpg~InW3`09dRHt_M&3b#T5iX4UUbkP+OJA?6iAOBH0bi}DQeZZMlx=a5CX1M5A>Oh z|wHAobSPb2PYG!$MXgFAs7aryw-DRNPKc*yL zY0G-o?$r75j!ZtAF+cv`szfT?f!nV=mgh>W8!HvC5i z^d8rrKijMAul%>&55Frv{LX&p_apqk-1K`qUSu1Clmc@6&c1l(#}rCx{hrF-Q{<$3 z=}-3CcR!|3dYME9{**$loCEkaU6@YAW7%fHPbjQ-c&Aoew^u?~a&NRH{3=4R#}^J+ zh^szSnpFutoao3LM)YtZ8uML0rBLQAI(GHX`tf);ktx~U#(wF~n3B-qMD%Q% z5b@HIMAXi8yR zV2^-ivdDY(;2oiZ{ObHHPHDOqjVe;Y#WXwd{ihOW`d-#cb2zphv2Zbus-928sI2im zrgO>*oN~oJ3@Q(y%_`1Or>&?m5sza?@#@3{oY-YQ+>R@Ck>zgsfRTK$ALVlxuV*XS zD_HjE16XJ)rVkm-Toq&8%b5Pn0hn|oO7|*O<2uMVweC+i@$-YMn@8%)q`p*3h%z7F znAEh(a|ny>nEPxOa;=15o37bD0viuAa*n8eWid$aB@+F?qzTKKQJs*Mg7oIHX4LJl zjM2*9R~2Z2gy@G4>H)=rhh;#q_Ao#@-#BPcTAkAJJVbqc(4fO$+IG;Ofk#rvxf^}9 zu<*rWu0sa-k05(|i$ewtTcY&kO9)$DfT*sA44QHzg)%NVWYFhFQYapxh6|gWJ&a*O zqeBKA1UuxPU&r!$A2Mi>WLPF4vm8w-_A9HqgY551L$fijMY%3uOF5PoXES(@!JD26jbB_ZIJ8+KhEtJrb1_$1J zJcZbS?;aPimBMCcw;cGb5(@_|5x!~(A+{PI;lSU)$JOtIkV)E2uzdzLI5X>nRKL+D zr275V0K4VkNnl6yTXh1~MS7f&`no#CFg{O_Iw0X5G8t}h<17KE>d`#6<1_>2)$c`Dv zWxtDx3HSjO8}I`vTK*v_M*Sfw%8%7kvFHz|h!kVKH|WU=7md>eSmd!vEqR6Vl`6g6 zs`Peq>HYeLDm@h&_NOkr-KzBBJAc@Wa;5O|C4?^zqa<(quo=z$6YW0O3WZ!GVHDd* zW<&Q75OP>Tiwy;R-APCTTbYpWmoSQL0oWp@7rIM5{qsG(us6?Lgps3Van?$^<$oy4 z|6t1xpH!BsSg%tll$lo?80N;>&mV04_*2>&Dz5g!Fp^l;IC_WeJzE_6XY2Tzk%}Mj z_f(u~W=|=MYzbl5-|!|X@XM!QKkB(A_iKpl2(!UW0U?)3Xi0-NFZ~PBz;?foCrB8@ z_6W26a>$?xA){qxVJQyuQ&~O1gCFKDC$9 zV=f*RKIX!((EYb6B^CP`F{_jwb75%MjSL)C%OZ!0%lVln`Ms0*WRR=s7)Cqvq6=qY zJ|% zNu1`_e02@+d*X!uJLbpY3yG^5NodJOGuXe0p%!93#1AaQbdEQOo52_H1I=KH(@-t# zBTis|Of4(EA(_8`A5H7f1VgpnZ{r7=)@F$YvCLxpz~au@M9AEiK-5xXwU!V#Tfq-a z5{WvxG@B-;HN^HjvjJh2kb6pKNrTnDyTs}-NrG?(egNT(O$Fg9`~bop&7c}GwX9wZ z$$Yt)$c#-EnQr`m%$4{7Pqs-BPu`aTnH`olvnujTo6O5$^|i~J(H4uXQpgJ=ggF-5 zC5Az&o2xmYx;ZAn(pL>Zb&e^hp282H`Wruh>YnC;>c{4Ws=13r%Q6Wu46u6Do6o|N zz`bhLFZI-Yc9y8SAXU^of*(+~3qPRl>K3Bzg%+Z&^(ysHx0f(V-JdNibrquSD)cGsozbao@%mK|@8O0p`@bYG~?qlVV8VOO{ z?V;u6<<01oHURME@@7=?tCD#ENE??oqgLk{l=0p2X7uE_2Jtbotp6w^$yooyF|z~b z8tRzY$ny+!?yF9eoRARi?*MkJ50{^3sI~WF=b;>k^_Ksqp5C+4L~kff^w!kZ`*oV= z?bcTGW`3fK%#tul@36L__no#%?=4efDfu_0wl9ax z`1sUf(Hg~8CL!3~1{=-_x;mkFu-EXDkWWYmwhzFDy@sw9nMJ3H zEW;K^eClC?iX^pMLQt{elX2l;gYNBQP{$sJ@mcXJ`Ts!tr}zZEq;ZSdvtm0Cl_GPr1>qh(`XtM^Z z?)-cVOnc)iWLiSjR|>4EySrLAGVz}Lt`;~PoqVAYIU~NS1vbI2M^cz*7t4R}LW5k{ zs1M;V{ni4n0=$Q$awK`_cYeA4;Dx9mt}3L3!}NCxj2bV=G{{AN@<6e?C90keDcMLV zpod#xoVYX-8HwMrC=!DM62e(Oz^8c2Vmlc5d~n5|svX-863z#koegS-GsHJ^M%#|l z#jXK?kCo`Lj?7Y`u{fwOtE-+PjKx!r74LEQaaWc>uIiJGiC<|M z@4%_WN3sk`Dw0B*eZu3(?e9{*z!>#HR`6}RyojUkrUg_bV?m(S&UOj*jgSY=H^O>BFEOU{qz8D z&J5!B<9HERYo%}P+Lt)w`B`hZu@ zmm|VtTILuvF@fT{8=;g zOu1s_2Sn7d?15yWj7J_wrhkDW{_6)~DJ;TkB!u5T#2WCS2V!YocZ1sf`9Lf^)B_T& z%F@7_ibIS1V=X__M)_m6!;x7)w42|(e;<+>5)@f!VNqHHZB|A*xEB7VME%C1wnmBK zy3vJiuF!cZqVi6P;VL{S26gyU4_zr#oKNLc#Nb%YXFUyyfAp|Hy(Ie;62dhpyi;%C zS4snJpEqx?Kg<_WXr(z@?7-(L<#Sp2{+D4M`n^TNv0ET2x@0UF2Iz zqXk0xA;48e<*>J^K<@Ip1>u@;t5IVuRT#I4qqT)J;;qj zWxRxdjJyoF_>!ncrkGoeA)Y*sx6~7~&{plVquv{*4w3DRLqk<@iNW9Q-W!J+dk`@t zMQvcx-Z*q*n|c}48ACobz}p*#VIOr4lVfijhJ@6)z=z*xjnP!^<6McC8sw_`n?I#@ zBn}&Iy)VUDOAf_Rdb+p3M-!S<4@{tOml~8_9HNtP*Z`~o4;LowP`DfEkcbZgP(1dD zdP-`xgrEkn^Ny{eHkTP{f2WsFH5Xc3YY!s(GPsYI_cgt*@LSTbyq^a?-dv~@@*)Yr zHV#U$Df1YVV$)}VP&F5nT6+qwLYUet*eiS|B!skwV@V-DD2bW~A=nxq1z0ElCTuz{ z16~Im@n0P_$o0MQVQUG&HUTP59X2Rg5;YMjF5}}Mye~6Q$Ri{~-p7y^Hs(m8CL%Ak zV~D7ce$_h9T~wrcGOG?|A5?f8bM!I3_nP4F;gQP>%2d4)z1IZwDE)FwGdy8+YYV;C za)Z)QGcgLJ_nV+%u0+@3WfA zk*1h2gQ3-&8nRM{K-p^6^%@c-J9t_0N33#JJrd7)K5BwVPpc~^by zvF>Yn_or1&Fe`c-X-WwaKjOqikwnI`rb+n91p4wy+vLaoW*#O#@+O|E5)WG_OlzCq zjkjN~!~>Y-h{IF5qY0kvbFMP5GP;B9xauk_ug1j52@t%A7ozW*O!y#?+~8BdRPs(v z{+4sZA_tP`IO`d3H6DGW7IW$eP7NTnqEQ@4YI!^`J%`lf#&H<5_>1Y`c$EGJSL4yd zRzCh4%sZi6==fNzQzY8Zpg+$En|9(GZtqX zlyq{ps+Ilrs!$ZZ&Ky&e9CB4cP|y?R65Pxs_&%#eIv3**Pnfo_@~zOLyoJQAoOp^Q zIlCM&C^Y1hb-&G2hvePprZTp}SlHfyQq zzH4ML;@xXdN>02=m+VwN`5rQ_-fPgekcn66e#Z~Iq}RGP#u#7@2=h+OJfycl%=`|1 zfcYCR#}^$jXyDJv(i{n4>7U5>=n;bw`(S=192aXujrH?_(?_qwtd{`@d zdnJU}t8l6aFCR{1HoQ-$$#piiYruxpfof)hv_?`1<0XW&8^8u>KU!=>LJms^w&9xX zoW9zYav@KY5Nx-B4YmyJE4ECs_%&aJjsJFJa>6rJ;_c>;0H?iyjLtbn45}4UYmokZ z#Go1oUvb2sCf6F2vF3->kO(C z+Ir-zKM+?33w1%H9Z@J%fgD!_a-5w!;s%2c79Mw^cWW|07muP+dI6tM{L_i$m`(%4 z)hF52gAgq!qQBYI0mM)&PbHvOrVcRFyL2@I^8{eBYllho7%m|Qf5T_6{oe!dNY*?Q zOSwYMmk?|Yw2HjukL^CB#9(I`8w6XR5XTa;TyflZstp2esV0hu$r2)C5;Ee^I0=pQ z{inwB+HZArcK;no;0k>t0TueZfy#eJ65zk-gS7t?m0Y>5I7&zL{51j9bLk*-sO}>A zi@mfRQMfPxwe{~osNYpO$0XjyZ#P(SJedf;UW;f!5lv1+j>5snK}m6x9-wK=@Z?~u zXB5$M%W1J{L7Fr{`oqC z)(Lr&gkWR&Rdoi9ywRYHC3Oas<41L!LGy1kl({-zjV)g{EZCtv{S@Mr;dKJ+IR*xe zEv1&ENeGS8U=cL#gCb~j<%lB9-rB})De9VoM*Xx}8M@sQ^jp=j;y?CkxxmHMa)Arg za!8J@hzs}-0lBi8#!+=d?yc405G4waYfCL5Y(%ginkC8CACOGPzNv- zwR5thR!9iy5Q|zPshcDO^)^uXW!GB4**}idySs10Yy4hYR*Ev}PB2GG*Q?4^muL1L z3b<;n54s$gqlgZ=aG}?pA-Y@-s&Y*qOB0e7mnKlxVTMhjS_K>G33G+f0&`E=1;3EQ z73PH`RG3x6407(MGbmRe&ILqnIVvROMWj$%0t!b2l}qTDlhJOUL9=ew=)O+E8yc#n z6bVy6LU>EPa2|((&@$?T?3j~LvfrTZEm>QV(6lO9mBO?@Lde?2xMKA*TA5zt@eTI- zh8B78kqKeW{o8K-E@l2MHvjxvH1b_ao2xn{4ug{)*pS(`qE|w-Pd=q&w3HA4egFWc z^Qb{JhAQ7ZainT!maz4d5bXRiAMa^gA*tCCT2v*COeJlfmR2Tw_e*H;v$P44I#EK4 zs+Q8SMA~6U!yKwg_!diuhRn98*!(ZCR6>YDUT$Z~FVWzAN>{MM_=z7@E`+%Yo!L6D(1o$=M{Sn z{n!*2W9+_N7x0fw(V$T}sHg#ZxL~sG!2UU}$R6T?9(sqNP^X=x3d#pKKwWwVo_0JU zzexo%39|5J#oBQ{-Av*D_zaluL8R}=;2n+xj z_{?Ugp&dsE@LT~tO9LM`LIYnSzFHl0OI1(V)$v&Lj=S}!VacdzqJd?U9yMt2T?S=5c+{Xp zcWG36lDUetd+ew|8}Bm6m75wz=Rb24*Z9pnYS7uam@lA>trRtdP{ZwNdy$Ur2WuB_ z-55|QZV|d{Ay>s)Al4l<=*nD!GBzDG==od?;)HLKou#jb13K!b66ck#fTO*s6g7tCqOEXh2 z?z(3bPGz*fqOZG%<}h{PDD+hUBnIg16tpZmkrW?u%%DvItEE$EgRbfEL>VoP85EO; zS_tQAa+-~8CamCxx>~G$J6?LGY+>swA*4ng>KsWOCZR>8_^AgBDwE`L2|<1fy0{}+ zD5;Ai1oe4P840yi{{NRoQYUODBm{c|$w^5i@wG)gt#pQZ!d8joxRkF{cQoGMk{NHH zWIE;&{!uf34n`pB_)ni(vBHaDJhMRQop(k z^{6dw4JIea4r+!1HG@H&6i_n^JWJE^1wxI8Zwc}Iq*5#8y;ewRNFjE?Elt$SuJJ{s z@kQ3SHeb2nMFUfp$33DkN93nLzVn7-29--nJ13EI$pM3^TE+1e+0R&Uo5j6xRZnq)ip*A)2I82oXM0{@z^{-Fd~TOi=G z&Q{L35b}8doGU4v0Um>fnzPjclp0GjDmq8y%0w<6S!@@#)k=su2MOFkHy;gqtlmd8 z=tDi>>x%sDum!q0#SZ=>1^y!jf5~VC{v+eP0SUBkG`6YE;Zf=*1{#RAg=nfU-K!L2 zeSwT?V2z|)iLWWQwDMOLc00eweL#w9db_lx_gxw6G%xh4lk7&1nKCRQ;QiRnaeKe$V@zhC5 zAyOL9hvBGt=U9|cOS{rXOn&wsG4brX@75C!75@=&9?nTZ#uyAlX^t7pN}!hz@z6q! z{5=NEcAGfNO`heHc16rP$;52q+9G)DrNahQwNW0M0KD;NBt{6ZA_-BQ9tD-#!#d%g z?39+dPS{RJ2zFb0D5-RoveKd|UH1#QLPC%qa}s4-bIhPG0g^k5#X?>pA=sW^wtWVj zU^Yy9YJ|K=LW>QnkS&VQ^}xVP!ICW@*q&i&#|*j*Y;0-vxymy$w59h8{{#tgQB?XAiZ8kP_wwuK zR}_2&1^hhJNy^tqp}3>_4NGjTB|anZfI;{A^_Xd6b39Y3s+J4W1PS3ob^Tdv^s18j z4h$BS_=7N(^vZMWMts+t+k@|#qdnMcY5Y!Uedl$jPwZS3Mt zDef?I@KIk7~rJB6KA2lJ9H0DQiN#QSvV&&Ac*a?v^YvHri!z0x@bud&lWnY zHG+PfgfM>+fbqGO6OwvTLQtPgAob?2enmY z{5#A;Q`b+}21*F_$b&pwQb$N=QPp_3T#_{|kbmVu>R%R1ON6>iLQsDMRkezOZ8a5jG&HCr6dvRb=M`F>uRG6P0`pTHnPUsxxP2@YPusO6 z24#l35zS9UU44n8-F5U-$>}TD*9ZM$EuqxbF8qU3F5VAPQM^}|s;>TnR8;+zW2LDG zh~byxR6REpm6CD=QtCf=qoGY}$!TZTwn}MR#oBHhtF*0RZC%Q6dZ%q1&INs*iWQoh z%FsOG$Yr3NsssHK<*4QrVneosSoTx{#R!-_{=hg(eQmlO#?%&$%zS*^V`>YS9WFB{ zGZf4ti5uf|EuYFK?utz6lX@~V}qeN-->O1nVidVzCc=OI^8 zs*u7h;llHkvlmGSY(SEMr`pX10>CCA@0Ad2FGC2MP%UC#(O|OEqrgCQAYFlZAGvsf zbF#fsO&6#TGGD{KBd+F|z^ZXCO|p=iOBluWA=o1I2v-CS$}IBuh<@ewF)q8`pv*kn^{?Xh+^;LwuPsseC_Ozy6I$K=Ng^%0-~Q-A z?;v;JR&ThF4}WQ`Ssm@#Uu`AYUu^~LJE6@-^lB?S!IwXvb-mh3=_;l-TcP0&K7iRY zz0nE}+H()s_3~bqR+84quIvk~>{KgmUr#Z*N6S$c{62gtjObJ=wC2MfHb^x_rx@#E#ATg>PBGpC4{IZp82YbopJjKz z$;Kjj>}(Xqo#U}+Re`iJdiHD_0(uxp@z{hczfi%QEFmmTOd{f~$X~`AGPV=yLBjJV!#1S;#f}4Z8OcwBtC2P$T}1OCo8=3p17WfP~1{1lrM5{Edw983zrT zDddF`g6&)u0;*1}sz=SZ`k+Ac3j6z-KkxLHDR4 zj|Gsr*@1qcfPTS1U$UTmaW>YbFg9CzHae~Sk7DML!6y*DJ{t`~(qjfu(%!C0Sg$M< zrhBQD$-_bByUuem)DtowNd{}n1a!580Iw>76E~`t3;#{9hf}H~<#sEjR#FO)5|0bD zKNcWsB!tkBpt6MIjtaw2q(pp_r^@igfaR3ROOzd|6qlSMo&1t>(8$eyOn34szV$I( zl1t7(@u{vWP5j$lLLgbH>to8}Xu>eoze32_5`yhR&DPIiTO{Nv3BmS>W_!qDYuR1l zZY?3$HfpwiTWl3Vo+csKwrI9P7F)HDKb8<|r!-spiP{9}p_De#ArbAc`kLL)AkNSQ3R4fMkgpn3>Dfa>4) z0skDD4FAAzE$c*mnNZ#SDYT|=LUJedTv+O^e@dM2Ab!9J@8Jj39mEf)JO62M!tGDX z5d!Kd6+yzrL_C&eZO=Xphp@VUO&!C`mkfmx>Wp$yqE`5S(`B>8_KNg_5^60B&!_-AE{K5_DKl~;L_t`p`&Wo(%JBoL-^UMl<}iLh z^##w0XGT1$3nDDC$|Qu$oiGOlF_qa+5R-*mA)&>FI`9R=!fMA9sRO<61FHS_0oDJ) z4_Lip3RL46(XwXNmx*Vj`BYs8vM*G1pcbg0?vklusSiJ3>C5;5b=&a+>da|k=?&B5 zL8!X0-h=S)GsTF}umpU}mMi2!3BeZMRG!>z%!Uh#?iX^ogcjSu zLk1bo!F(vK5b{h3!RFM``ZF7(EfVr#2`x5AD`z%Hb4kb0Ttcul)6(7tTclgTrRZ3k z*PTg4FJ->!RA#s4jqlm5F|IoNoI#nRyx|bh?ABP~yI{H+E6i?$ty6*Y4>hpstV*Y#qRcGRd5Q zva{CQV9<5E<)KZdo|Ktl z&zo#dznNmsIsv#zLICXwps;7mOl=PdqDDF-qKm)=d**?y0egCMw=AK{>=rFm7A<9q zw#-x(Ep7eMfCNgOrAlZiThwoswrHu%q7T@j$7hK}AF@R&W=VUrKp-xX5KynwP6?w`mh;>m=S1YKi8EVVf3hL_C3m#9P8xac=t!AiGXKSddZBRdAs9((%s2?*_ z=kwzHY*E!uLO{J1P~rT{=e6@E3VE`GU>g87I6n_;(ax7uTWv_?<#yk1PzG&agJwT( z@Qa=sT4Sd3C1PQs4Qx>7IohBNVo=Z%rj7jj<~jBWDJrnqE!C=Wudu89MydRURhG?B zD!*xc&lL&u7GhzdZ&+pO3tHtjtOlQ43)8p!`>Ge}t5kcOT2}Y=Gs-}mGYjT){}FUUEjXej?7Y`eVp%xxvCED zV-2h4YNHgDoMZMGR4e_$-Aze-E6DMp%;frsAvqGFwB=htn4;xostyGiZRMUaMw8d`a_9pZ5X@i;?FuT&6^GsL--8ls|-6DOy0L{k_L@m-@GmGUk|rBLL% zMu1f#NnmaO7H=47Jb`QMxPDW(eq&r2FDYEVF|Nv&G%iI|xTuTrz&MR-=SzYsM<_O2 z<&p&EXk7h$%3~Is%HDQZrxdJH46D;T1?v>Unm13wQdHs!S0$Pr1guH>4Ekc8RJfU9 zZ>5ANQCo$}>ZgSb&ebtrVI1EEUGKR08l$4BT9+%T z9sx$wx|ik)#x-JmjfB8xt95m@jL{esK8-7ItsURv3g6?5Z?oWgobjEvK;u(XHPKDW zQm&ZT6k}*q!dyvt8i-K|-3z3!tAt`J>{>|zGjX*{?xqAzhQLT8o>v;yy^=;8xXzAz zib6hxkuO@HJUfMvw|`k9S5)Gc3UWlzRN%x{yav83I4gznHe9Mu)kqSUqfzzu>3vO& zr}la~t{Do~48}F%Wrb@7;~M)fjZ0A}Jw!9xaJTa>gIv^|Ux}O52FEC_eg(s4G?Rhu z%0aV$8?|=SD^eq7ibEcx*s66- zPnGfI(~RKmR;mp_ATdW>?+W}h&Z@7#Q19K&W zx77Z16gVCiD1;WM>L_sUDsb;IxK9M!y9}<*0;i}7T$Mms3~*7(lpa=TDXAQ2_rwyV zYzZst{<`wS5>__$b?pg7Rm$o_*-|KDPh?%Je1nvDZ1NlwbLu2KgC}@b z)v&S#+uYr2=m`!%JJF;VCL!tXk0q%afxP8KgFN1lh`_@~OpYNl@d8O{=Aebm~ps0NTQ^d%PtRh^-9!fwv4wYA;&)N|>HU zJDB>}TWEMN>ZudNKb(emSoi)GOd4W`yIaBC&2ayEOT*pGa6MHTuA)=AY=rD#&977$ zc&~YH8|+|wUnS@FvwJDu9zG34|L*-Q?YTqkD2^!<#~8(sw>64mjH1mu8ik^htGrj7 z8n~a}40pW)4`l2&s7iDnN7LJXzd_^QG1S{?OW%<;I77_Ok`T|5eA5v=q)Qe0U!u_$ zy8~|CSh}mg9SVoSLCb*LVRl2_KbJ2_eE(e3z9a7#lvzyl{<)~9#=C~RyYv3J=m#%C z94DED9)8y#r|+0S*#e(PUy*k z4Sbh^57Lqsi?rGXxGZTs7DHO(@y(C=Y;G6a7TL|Rz5EvKg9IY#;NXwvcEB+3iN%v4uUKcLl#L0HJu0JgTW&D7Qk=IyLg-hf^C_1~;qMK8Y}YOkzGV`EKXS_DV@X{jp+&V> zlhKLqaO#MU5qW*1Ed#18Lt=?&x%y7oa{f-)3Con_jZk?hJwz+`*M;xPElW9)dzgff zaa9Zve??>F`v&o63^u-xdKHIHRB3fJlEE=2W9)~#*a!)#NJ8MH*mS|r%=!1m!WgA$ga;7%`! z9QhKC23>oby)<5H%LmU3`U16`GjP_hRHCtqMt>CLzSW1F?+`88m&VRL-Sm zu#xXz)P_~vZZ~kLGH@vyxZN^vsTdf#r+3huH^vhV(MtZBz~4&^?CY-J$VDIEED|ke z<}n`_L~(d=Kx;_8Lu<3 z!N#~C_3;noxf>|P4U-V^_n}1a+`aZs;Chc!$JRBr%8_=AbqZr0V_eS|)gkw~#`v&S zorSZ`j#DDQd4h4CvrKSShzDj$2%JYX&a0L+>_oiasbOW`X@@saeJyWd93BYwGVQd9 zaa^L%KRK@5r-?K{>t7<)RZ9r{PTWTXKfMC|T)vf3zDXdVe7^vxf#(%(8E8l~QP3(B zvqs;H+i|ZX1HX}l@QQUpdZkBEN|E~gG=O{JFPtTF1vqbDiAXn z#4XFUe`ZEN%!xxi8mB?*73Fmj0z@kfV$K;r1j4~#eoJ$m+m7J%1a7ikPe2K+v=F?W zfFT{F2ZH{vKRhnrp^p3|=y!3SfM#!C4;?c)! zzgc4>sHuJ!zFe9nWxE zw9e4HUPg-l_?SVNlA0qS#NG(8-ySolpiFivvaS2|8UgRil(G>N@8i zRT2HjMRfUUgEC8qeoVk4^yq3`WqxFjsz$P}WFIIYz}yWmXe57K4IHuhOuRM7kV8q{ zQAM6%ehWc%o*l_A3dt{wB>N+U0_<MF>?Tc3a%6^xrY4~V53*!UP3m}SBF255 zys+2r3(>Tsy3z!C=3|4>{RIUfPnf19C5}v_{Yb&y(Q};Rj!$qY3o0zQ`qPZ0`!7wP z_dk_vG$RSSG{Zi_`9yGue43erF|%V0o*A0MnQ{@iTOx0)G00VUWgMm7O|LV#+2^>} z64(Dbj~n#E=LTi8IBw82U&vK9ntg(e z{b8`<&Yb)&40RLSM_)h#A0*RqC)+Zx{HqTdv>SX3BU#+tT0+2B1i66WTq`gN*9r{H z-qOZy!FUqvxbJEIS__7jn`X=K2IK=q(mL>2m7Q-a?C%SEf_x*Ht_jJbm8aC&;<`kh z)2vIxq-OX!oZE62c!<{V_m+mP*VS!ZBAPk1WR+E*B=nRJ^N_6=l|^)0b81T+6{<{1 z?VD;Y{U*<)OV=AbmwuDy(z6g7gYTiL_%g(KF0JA_*BfdkoGFF`Bm|7EdL|sR!BF$+ zY@yn@teNly8{jdXSBHgfiG+}rrKR1crA-v7oy(Fom!(lGp2W+#c<$+Z_vV`G1N(R#~k^-?h~IS(pn3Pd|%fti4S3Rw4cJ%G6G{Fc9jx+adzWR6X~!e zOU_?5APZiya@R#m`8o+wQfhSV{+jR1p?Dn9ohhu962eRP;|aYc&JiMNiJH(ifB?fW zt+=8n%8V@xpr-{uZVci@)>>Qfu+o;2x3z>&sunXFP->0Aw8rG(D9tO9Vfw9><`wG3 zgN7wZ#nmDzqD8&PqVBasy?91Z_U~zHAr-zTTzgp93pMHmx;@F*JgT-t%XJFQTaOx) zcD1egK6uoiA8QQaix%R(G>C6kxamtny_cJNwOaCvT&_?fHf)j*&QzBxG~fb_8yig) z;0k{f@DmI?{G|eZA{iS-GrrU{>qIife&oc(4Hb6axT)a-;Kk((JMaTHGGu%uz-t6} zZs(qH;<&Jl{CCF*up2FHoLOgQ#i?Ytrf=5dW z59l0ocyM1410~N|N-AP7C?V&*1fp!4aSaQOtk`I=wbA0tXHOZ=Kx7T=&RhI;Sy zSo!Os<-E&szO&@K8><$?ET85@F4xvsv+s!%V_vu{j9asDx%MAl;rE<8XimCBt0!6mu~tmuuHumZ`cx+ z6KVkO!p{#Z{e2>^z7qByymFY=Kzr(K$+#s0SJDQC?ho z-P@f{A1_HfoiH@)&KtudNBungec>UV(M*e}Pt$L#LS}4zA|J`nSsPi@{=QOokvE@y zio(HPwm?!uKh6*#C8D*zFWifhBD(tf!rd*qTHo zDHZWiX`;g!mDr;@=T3-96fKESiJ1FCk;_VD=ZJu$=v0x=)Jl~zB9h*$etNg=3URVU zP+P4~m|~@2IL1kam4t_llT0fqG9y6l=2rS3cYssQvQl)>gD=%e9g3Y^PHPcKS4SZKOnW4e{ia_Voto)RIJ7tr_a^A zbzd6oFc9EOCPehB9DO|n?vkP~(_#~Jd*bzluL+JGg6_f@TlvXg?v3F@D_mxz`i)<6<^vZ1u~-*pN#XB#?t1#()Ido)qtpBB0dV)p6Fu zIglMkUFdp$T3&H}*F1l5z+2==8}0AbrF)MqJugq|94d4NJ^5)t&uI53F9kgXDD%j0 z!qz?ZaTgU?V)dldfywBOB;xTMK;#9-1;YNWp+a{zO8t3t9MyGe#ds_JCm@`^u}v6W zCJU7md(brZ=uVE|j<%zSaDw=ZY!~+=$FMW7Wpv8-j`oDY9lH$iV4=z}!qGOLsPVug zqFB3Hug@Fql4V&D=&r0$tB$6_xge=ERcEW6l;-%-)$*j)L>)iD|9YY#ItRIfp+a|2 zZ?DfC9B0X-o_5(icoN(Lh}wOlhul5b;LZ;QyOE>Edjk`Q(nnEGB;NLoLC)_lYeRI6 z0&_8LS#NZC8%q8bds&yap-aBS9yoZih368@bGznwz~TuAPcO|=sd+xKcq)bGQq8kl z^EkiLx@v^yGR>3uo#45Vc{adPy%pvw!ShQ}YYKwrR8niYZZlR7qDmy<#+!Y`QI6CO zth>nW$ZOb<*T9h#n+?Jp!`HxLhc|1lUc;X3@L(_t?F9|z??uMPbxo#TQY6DA#L7k= zJZ{dF(@e@1*!=-}$260UZ80eRqVksXlkgvu5czHc*UjZE>8!1=i6sO?!VRE7La(hN zp-yuCCSjC>(ujmgkuX3@Sga+Sl$?#k8&MLDMI_XSgn?Q@);7TqFFBJXjFRBnW?4af zRPhbc5*BF*%_V1R38N(JR}yx^lOo|pNNDt&Ng3M>a_)YxC6x>1W>DfDYDojOLob)k zWRbc^LP)<3r9*K^&zV*=$hufphm6k8ndF0js2VlcGuC3NwwNK!E1%Qz)^>N{k_jqC zHF|ctLGjQqT$n2)L>3=_LeFC8iB=S;xY{F5sy0v$?gT*uqMesm#i&Pdga&(+;3^?KCLU>mz#F0o;9dYB#Gmb;!@6XC1}p z3_4)$MN=IAc??Sq0M-6dB{xGtfO_ViK;;#B^2Pv}MijVQNl?h9D`e9d*<=gKbfq#K z7gfw~V0q}19;X?hm)5F|}x??<{Xg$$iyB{3b|Gl((d(EmF3S-qNJEEmA;8RhqQPB2^0MZB6=* zMXC|fJDPO%EGUcetHhn)#R9w0!s)=AlR-7A#t6|DPJOW6t~G(Ak8Ix4#*(D7Y3+~EnhgYK|D zNS||J>JO1(+paQnDDwD5hYM+?qnyS6BWF3vS*Cyn(cF=l?;TCl91q!hKdPpoIUcdi5MG!?Er6%ZPX?tIc2kj|9OFPZjwEk+F2vrt6dF!k@9NcU>H%yD1FX*BGA~_=l&k~LS{$I>hz&)ofcoNhDR>(yNvr5;tRjpdqO+t3&q5K-I-rX1 zEEM6g0|q7a6Xt0Wb||D)cRYy0m$r4gyvQ7&i!?~d*Qa62WY-${p z^{xR5=u6-2`8C>zy9Owj+kZd&TYSAd3gowsJo&XN!S3NPJ#<#TyznG~mhex{Jt>|5MTq_}9v?Z* zakLE);pl--d>&7J=vrT%KOaX`9i1$C+9BB5!ywIi|;0Ol?D+fD+f% zM_d#l)*CAHl#ydXtfOrn5!!1Zl=zrXVJD0#De(F+a2%TF_W3+Pa@@uS^n5FkNT?Z| zgxImmFi%;S96<(gZd(_Tg|Ls`7j}Dnp=*5O$kCe_U~1dELU+(ITWPb&yA~s>QOwQa zVzocvvOb0Gpj9~KQb>+Dct}l0W{}8)lRD!W<0WB}_nP8o%`@o=;T|R-@}#1Zeq)|V zIR_1jU-Nt_Wee5Lg~Rb3cRimCH za15Y1@5At(()5F9cicgDF;VlFRxa9Xfn_?SyTjpNh}vVZr`aKH{Li{=cOq>$j5`j| zF#5+(D(w2}utDj??z{a#N@d3$K`&P_O2uA3f|fL`KX)Y;fPPaQQUaV34hOxXO2QuM z1j@&C2Dx$u;&7Ha+%g<3P1~aesjocJh2UCy)F4;ZjS;RZ!8PTWL9U7%bmM&V@G2z6 z9tYT);wU|@kgi6Bd+IoQ%~dzV;=B! zMFbbvian_mIAKu7PxqiqYWAekJ0}dv*s&*-uK2~EjD34j>4#r1`uq1hlhTH%ysc5# zpUpF=@voYcQ=jwe4`1IXji8<7fEKxpF1f?I7x0P*)+j+JX>KcTB6RlZZ&DUln-~nP^QD%O$l^LQn%1b)BRhln~UrK^52K zxPu{&BGY?=wh+_yN%VAGClvi=b zq}NyC=0cg}xMxQaWi>J>eUv-op~)<_g5xCxw90|8&E7_)TBF92PX>kssZ)2r%CA{uWxjbC&wKO<0XoZn7vVgNv_H~1^i>?y*+^y zf6Vj;5==_^={Z$q{idrBgThakcNuu|J$a?{3A67^U~6h1&`0ZeJ>-@|sK{3;zTs^1 z6B!N;qXQ0Y;LM*Q)r0wnC!doyaq`tJ#`6tFo^zSx%JnE=8(G*UMc>Hu|F}#dSEVp) zV}`{^CMl5HnCH7B@K8aVr0(R@X-xqg&ke?a1Ld*Kl@PUiG7<53@TO*_eLOf7QH%$v zBL;&qip(ugt6g(>UAq)%Tx$wc8)jy^-i3dvX-opFvK8tWR$ zh^{-WP-*-L;HAyT_vHEWJ>9$a=yt6y&uS{Fo>vZ0O~wJ$WE|i^yUo;%tcs(>?M99R zT<(+}Dy4%C%xxYvP0S9^B!%fXXFadENvLh5^e4yXb37QNKe=W3y}7g!I9W11rn)SF z$oU(Fl!E&p%GEXBq}FGd)Uo+|lc?jl_^*xeh6ek?k%^?XqAtps@i9D;93O+JWZ$#2 zHREG2=hBj}YFbk4VOduG>N$3=Op4)#bW#i&($SC;#uSzuDup@+(ckD;y3ZfZ@rUS1 zKAiXJS!O*=>KK>SK$Uh)vz4aVtZ5xIsfN*^QnV(Cbr@ArP~ZvD{Fo-Y6Ny|Eg_x1i z0#G3=JsgHQds0m`g`OZ(&DFqAC!qyo zs}>WjFggp*(lMV_BZ_vfTqaVt7=wgN+EBt)CGUmnVKlj2aGz&=Bq98di3W ziMLozaO6!y!gM?atFZq$$0S!)5!c3ACa-OcZlOvg{lZDj+n7Yo4)aaQDpuuK0D~`? zZ_-GMR9>G{&ZO-5CRGWEKWvdR-=z5#sjfchQ;@_pAy4?)varV&^7?(e(xV+huW83% zn_!{Ck?GNVL<_kc+ttA0qO-1AM!So=dNy@+g_>x6JOe?{d+uEo*0KGnGzw?~it z>yJUn7*GJu@K=?@`y~X=Ds&-0lG$F6jDbvQgyK7`5_HM;6nKJT zgYG~#s~%KZ^}*qc9h^+$SAKgRlG_YUrYY@BYInilWO|P!tELAl@?Em5%=Ef~=9q(T z%QkZeSCQYHKge6`xxUChDs-K<$YU`E-mqtA=g2JZ7I}Q`Vh>R}^go;1>)O_ipGvhn zs(yC}<+*)W=U>>U{y0i6@Op~!LzIb5 zeH`bp>VFTYK5&6aNr6QQ-y8`sUd`^?J%ReAqkgz5APLI)L)jQ4##c8>rJmx5Yzg6x z^Wn+$%~I(#X1i^^NrggQAR*W~g3UMIq_4o{96R5n8h;#h|*co z!HiUZTXY4;K?SJoAQd17WfGf#kwG9*;KNbi=}4_aDz*VeLgM`yfaI#-#9)x}k^3to za)~CtsQ|T>5a#5-9F%BEM?AnN(Y``1mk?~XfDI*jDcHC~lkZU_Hypl}5=9wgOFCAe z8ZHBP$?ksbsB+6;%Kbeho-HAO3)B&Y?ccoJMP9`~@+LcOG_o#P;o{S;NcyCuKUEc|UDUln)-qt?V zupLSCgQVURm3mZCi``|M?#aK^AI!(QDwyQ2jZpjeePNG}zX}M-)d{Sk*j?7!9r9e< z6FHo*g-;FRT{&<#6vwr}pq#CbjM_Y5Pmr9k6{FNXAN(fV5Q;t&hic1Bbr@Bhngu07 zR=ub*w(@iyt8ohrDj{cVKp;IENYqLq95~W?x&(iyzt77x1@tBg#p8W(wAVtx4Lmtx zlOIzy;87DBun>c~3v!m7OrvdQRehdtSA6IrfEV~Fwb{>cls!sS#oBxRmp#phkj7Cm zyJU14;&JEW1K)W?{*b4yKTM7{VjbA(oU|>8(y+Sob}Ytq^>cbd*9`5`zyCl_7zO2c z&cO#+EAiq}Jsshox0oD7ycq~dX>XTW81c8 zv=KRQPoBSs-|azV@(0P0yz&W zXsS|=X%D;(G@y<*a~B+{f6 zNdNbRCb=kxkJ0&q9@Uq($uudwh_XR{8NbnPWh3X|&L)vFxYDGYplWUSJnYPsCf(7+ zr1L9Ra>MeECiU_Au&n5@*`{r(y4!Bs;|@0NaR;oM(nVYMxC7?ZAmTPawwV7EQ}%b2&haUx{hDPP9u@H^TkFuREsFKcBCfei zjYM6Qa3%cp9Lvk=hM{;35(6I37@EsUYdAGI6emy?b832bfwGvh?(dF1^rA*BX`n~B zWw?YW_!rQ14Q8_q8wL7)ar004<4w%!zO3o{qc?o#H{P9J?{KrL_xk{LqXy!H-i{xGS0m&^cL8Ulj zod}Kb28IUQc^<3yE5|5%9bCn6axF--f~&!0m*|3B;b8x#`)DQqy5|z~xKe?VMM8QW zt>VLE%PtY$u43Qr1SIE(MlC6;L>c=T@ZaC4CB^h=5N*CctlLS17RHKz9qXnTM`k|J zrkMEcNpwLklaF$5;#qN^mr0#{6*T5V*%~%JxQ% zOl~GO!&f5bft*XJ2lTdeROv>MJid z$yGB}KWKHF*y{?o6i*hFDVN?2$atEb2B|*SK5n1i=go5$MY@3jo^ckus*oM#+LJF(^Q<&O9BB;ke+CaIg{r~lU!M4yihWUFQj@NDY;0=qbD0zc1xfyuS7vQ zw7W4%xW!c(>~026o&96Tv74b{ zU6b8F2RYvSYE#?tHz(Y1H5{#wA7bR?SEK1m9>?aD$D!8GN3J|7kAr2Of}Ye%@Z2w9 zdMP~{S9fUwodlWf&K{=<>9DqYNVZ87zcweCnoBM%1=SCiNp;+rlT6fc5B|s3*%;F# z0~Rx{;pOrA{=-1KkH5Is?^89>@>or&-E(u|97!%oSuA0CDa~c+fqmsFcpu3F)7 z%*hy8ZPMXv?dvQUdm0M#{gK*c<*vNnj^VFZej@&gowO~9Qm?_{l)G$@r`R7HNAy=L zy2(M3s%#1Ai>$xct!fhzhe2xPA~NO!t@9~P$PSr_)hJUjli}Xkr0kGjc>DV4~*{Xy}JqlJ9wW5EsB?c%M* zC7dv301`^+12%>Rn&e8pmz&;2jVik(P}x9WaTZjYlyk4DgPl~CM(lTte$jiag%*oeyS{a+?qBB60TkvNRW(`T=5G&!;z zZJF)~5}^%9^U$?vYZ5Uzx&bA1dUl<+3qI$WvjP`d46}khSE_aONmq|0N1dfg*$BbB> z;I&$qXcJK%K_7oGSP}^L_XSGAQM0wT0DH8`DGKbJcc@)VCMLa{iuG;*}{bt9f= zw)J(_KYHjNJ@t=^^pA`6k4yBAUh*T~NBpWvTl92jY3xqtS0B5K@{YdNm+y7^y5yAb zz5IA*bESjlx~RWxONsF6i#;c=by%SBF6ig~kDvhvXX5`CK#;#M8W2QKY<(}HYqM%} zZKCQnIQGusU zc8%OiT$0b?D5O;VpueU@io?GX>*JI*RTez1wXzGAo6N3ZfsqZTQp?jm~A zfwx^^Z^Ri0FU@A+rHF-e%zvXvuI&5NkrA59{4d`KBTp+FJN|0|0go@AA61Wqp6<6> z@u{}rQ?_EwjoOM&*^2ZWQ@%#6;$=7KZbro^-AA7>;Fd#7a*kYYQu%$Vi{qKl6YEVX z9fCrggEF~ak=9_6v|+tTn}(Q_@$Gt(3WxnyW0V=g{6oWhZ?iTd>jAqt8?-qa*ql#> zY4{u1oRh;ee083Mc)+odX%F8l2ILCLjX((l(ry8`xSi`wOIwAdHN&;Rq+4&%a^K=D z(En1e)FFzSO5DD1NwFm>dA!|{B{7anu3k%GP`&QGMHkExo(674oU50LAGp<|%#fF8 z3C|0s4aY1oANUDj?P%K4ckx-DXmUe2mCGh`C@A z-p;DLpKlzjX0rcwjB_5q_F+kwHt-~~_I8ufy*^sYvDiD{w|}mpefd0zWqb8wb{~vy zBt9752tG)^L;GMn57I^=&OT7_9wT5hImYvdQ>Mn_d2olLRWzB??;nA~PQV=s=5?p1 zSycOxm@hww^F}^;rV$$AQAnbP@Ghs8=>9Y zjif3hasGEWzw=I&AD0-a^ktl$iFD^#8%)Z1SUn@nFhaa&gGo2uX;Q{j8%zq~$MqXb zdJaEwHkh;oKW^J#(l-1UwZWvn@MFvdliJ>8Qbu@#NqzC-fej{&z6+0G#RiiqL`5s8 zSh~Ta7r?U$a+v3f4JOqHPX>7QLo#@dLvpT387CnbKN@c|X*7N`-DuLJT&QTf(WK<@ z%8D+?e#1tSs=+g4qe*{($MyzQ{mz9Sa|~>BRY-nZR|(tgN2m^M-KM$>5RtxalRDcj z1FSDLqtj_420g6S3MwOtGCf3Z#a#J$B8`Szes@Hk?07T-y@Yjygp3T$1sa1%Zj^Ci zeFQ_p`FgB{lSEgZ#xU4#r$2NW-K#DQQujY&WcRltp#de91D&xg8>`&}IOufVn}-av z@uvMF?-2Lcp<%n|jQ_?SEj`Q5PLHNUhRa^uzv*)qbuB0khy1}XrS|?U&MJ$X@&C(s zS&PsE&Y}Kbcxc!?s)+B@vJXA$YsID2yo4Vaj)uaxohR+Sv;m%RX>#F+y7}w!G-i!X zOnYD?IWBOtEhGBii$rRhR~#TJUyD&_UU7i$FLDPx5e!~m_~M=;$&tZXUS5|-ZO0LZ zX=Q`mgQ+8ik!j=bvlD(&!(uRKw#lg$ISBtJIR&_>#wKXH2txQzg1n)@?!ghOB4Vy) zyO8}KkO#s+%LTuSMuh*Qu@Bz|Z1ZG@2LzU!-hC6;l?cxyM}k~zHP?CwvL@*k z5Tn;_j?7|rfNLO8H{445hSf%O>-|L{Q3o$J(k^>wWg?HydO`i~Zj;jSe0lQeQgB^2 z%2XeE3rJJl9m`mJzh;%B;1!j4e5Iq&LQuu0Na$E-@~NKtfo)(REOX+c{ArJ=s?ndT8}5@0NXj4_kw$f^c}(5Upze)T zxgNLYx?!I|xsozc=c;~OhWGX`!k=?AD7w#B5lyG@D@Z82%V(+cE53GvEfM?BZD17K~kS*tPm}D|4gOS*i!g|jlsJGWb$MMQa;!SklB+JNSyFt zNafEcNE0N~kcg2~OpYKSVhPnE;SL=91KL{2`ICgv5|W>ak}zkBN!d>+3`H<+#TJuR z-fdFGr&~<=5A5U`8xWp{P!{pM4j7d7tO6R45REyg8G{ZOboLmN;%~vt07A``5Y=F{ zizvf?z@WewliHOZFz7k_c;bLTOOOvA>FFu_y(C0F-loNeledjQ$(SpeQRP$0qw@g0 z*NSHJ7f87Mb+17+r(v!VX*5M)kHXxxNH8xGt~v>!+rm7yNHEXA4`5!2e8Bv%@UM{& z`7F#o6lu)#bUn|%9r zm9rE^)zLlU&^_HV4s>;w`K+!^C9L!5zU~=bbCK*TEYa$Ia$Jb5~@VPvsyySfMB>WpfSu8 z3=&2pG~hT4hUBAb+F|9hRet5Tt9bYEs&76|q(=kr8L7yf_h^4zg)WIGRuBAmQt*E< z>eE1S=4e2uY8zu5=XK#D+ueCE7$dPBJR^wBjCuAalcwBbQl<|vVBCfn&mA_# zy!ewzMi2u4+(*NAWz;ZJeC-{Fe z;MFGQ&bOQX$N!V*f5c4w*`(yKN!U03gAFM`te7OW2(eK*_9J5KmOo-<|7_CPCD^<3 znl?rmQKu`D5cxL^o6i4gn5sq9p z7OSomarkiBQ=Ia_SUA%aK$@3kbK23doaK3rj4Q+963AodWllL%hAdVA1pU}+`@g7p z+U<3(C|lIE7d5fE=5^zmr(@l@=EX?OtBFnAV^aHZ%KbGAbUtFm_%H(M;irg=(y>#B zvHNRcTkSC^?Ov?&(U;8H2a&t!D~{ZEuSw3%TTQB|h~wSyZyiJ#k8L&S-Fr>Scy_Bv z&F(WPW7bxauE&p;wwmj#%*|+HISKTk!XX6L5e}x~&p74NV z@BM(zUM<Lg6b)q>CL7rrQMMdoRXss3J~Yz{qLLmq zsdEu;dOJ8lokQt``3S{7zt5x^G3tbb7~EWn4u0i6lLkI)QoGOgne^nt?5mEABd+rL zyB35fwI;5KRhTvZlVb`(6w5+nFHLGh+9F*uFo$tCG!xGE4)W)F^j8ZsU)r0RAJ9K{ zIPm;_?nvB_NULByuUmf3FYZwKSX#$}hTk4G`JF#n>sWLhR`u&pF}$Q+1%76{iD}+? z2i7RJjK|TNtf^d+HgoQiioTiY*F6H)h$R`Fa@^rU-iy^Ms~TT*6wdM(USBPbN#Br2 z-bYM~VzS6lu6It9W1_^}?88enmYkX>d3Zt7Lo}Y{Ra)}Ks|}6@;o*Kg>{@6+6>r%c z`KSW;C=2|^68NZk<1;->kHvhnA(3{&`p6aj3by3~OJ;7BmaPgk6CbmI=|vM{{Th+^ z6Qb9zm9&gby68LVve)8jB4Smk|GXiQwnB=%)G1L3Lhb464HBwGd3kwxye(br5?Z82 zHqy0ac||25Z>h(k#@MLCJjDTj5cm3Ou?`z`=(yq{uWyW|RYyVblzQ_lp=q{UI8KdKV-r-7L|#?Ex&H*UQm;8~Qt8p>6ZrAF zt4{)*(vKe^Js*vF{1Q$Vs?4k?Q%-2MrztsTs~PDNcw;}cg{uiw9akF5|0l~he`72h z>L=YXzKNr)dS?&)SzC4FO#LpwX({zvGWGX`$uT>wp&aP(8_1F44u(ABnCEm*<0Aj) zE<-)xoS>(`TSkum#5mdp)Nj^jFQm(P09bSAC@Q`y!eadbTaM1jciJIyhy?OeB z4|n)OUGlv_)U%NkILE~mxkADyHMF=XZP98FIr+b+3CYez<)~(?qFTZz6^NK-~dPFhoV>drj&5i6*6Ub)g<$x}D?h zvLT*ONl`dNJs9}sY714j7BYd4Z9;@PjP_HjnPsW8X<)Z`+!e^d;uvqB=G?s z^561=N$FDO)RikGlT2bUH!6VZL{7dZ&x^)7MBg}e&2LJLo;0aL7k5e6A1)5u)vaew z+T>XKY9cx2xN4^2Ivbe6$0F~+5wMO=Vp1Cn(miNn8&5VVY5Oaxq_P&OkYSaAe8*&r zypVpa_Ha(6f1h_G-<9Btq zJLCuce;EH2Pmb{54fiwglnYN54xODrPL=TR6?A9dsTCe<)18SYd4|G_yYw33`LoWX zY~jJ{${n3YVOMkhpN9X+u7L1dbOy|wkD62^JovBz#b0&QqzV60Zv9w7%t`LT(sJ^N z7DO}Ei@9Z(kas@Of?7^7$=T;b3(6A8EucJbq6J+(g{zD6uA?U9&QuoVLe9h!EvQ1c zec)bjq6OUvIZIBoplYG?$2}f9p%Ux+or{Rt%R@9#LIA@pSsib{0zNU%FnPe1NC=*B zFyF@0^HpWwObNmBfQ{#YqbAJ}+9nCX^Pq*K_3KJPxrCrjvPl5i=8F_hnS|iMrR#Pi zi-eZ^hT;VeR$J{n&4rdPA$X?L*CkJ6zJ%a;r5;aDd0Hn(2%fj<@o3s68xLRQVDn!7 z+Y0L%2_fgbdUC4dnWlFYFL*wR@HpokHK|Uj#OIi)uRUtg!&6M^xDB;8G?CM58cV>0HEH&k4R`TaBMW_3L*HC1O;_r|P$ zwr+Ejx)!s~tggGICT+zTtt*-R&*~~?+pXLA59+#EYAVjG?#k!?SzY1tCbci6r(#U!h7Z=67AkAM&=-sFf+{zioLo}Y1tjq zOv{W+)6$aiQg)Nnw2Kv%nVJ=qrIi)Eva<5>dpy@(`!Iv(tJnMU`|<~S&f06Q+g^L^ z>)8iUWznNY*Dlv)Lu-zzU8}jrW|uoRTD!p9qq6I5a|s@u+k5|o+x{4CJs&$Nw_Y}v zhNE+P=&@t#R`*!6ZikP~t*6bUpzf%ayk_39xlNlF&28hnqw3bf=2CZbZf87xY;FaQ zM{|4N@uPA(*XEM@#!+?q@xO3uUlYykvYMlE>uz(YIXbtMHOJPiA%@#&PaKuoIX0J+ zwMW(MrYDZgZSfP)y6rwXw{A9@lvNkMob1Pk+aAJKvKwO{|9SfG5u;7JMt3XsSWk%5s)qE1*!`3pc^6~e9uQ74T4`YuSN9JO#01-%(iy~+~>uLIm`m3bx;b*g@% z4J9np71#-@9I`j0erFbY{KZxoWe#1^hWakVMpdMeG9k*fc;dB!Fm>)!dmL>A0CnFP zWOhB3O0F$4m2zbdsg$$NL3W=-*yqZh!%H1}Tz1hSyparm<@Zu=te_?>#z_n;Mbi1; z(0Oqxxf+Gm<+Gw_UI3cY7V|9D@oIueQ-rgY0B(ypde+_qlYU#AN*RAAnAG!Wl=x@W zC$&0&#nH8vE6aPT_J(rp4Mpu;{d6kfd)P!nVXPT=6l2j)=yKXKc(X=T!%$SiZ2<5B z;20k+jT?%_v>L$ZsvU|7Z?#0%*ih&#HKu;+8oLlRHf#xMOji161j?2HgjyPbqJJ$( zC1)*+?xO;jeEd>-1*l-h;^S!NQmjZlnuxawCwj_8Tj5~9AGWAu;tR)-@hld($~?ia z<)!7u%Hgf47lW{Ft}dUx1U#<>u*?(2Yojz8eBXHvB`bqv9>2ny!Lg)%8NacIl|))# z-1PKubagGfVlIyJMzCfWp|KG2)>^nvBjC`5Z|3?MsP&KGg&ZEv0Re%+tL$h&hk{s zKDhKmBFEv`?dU?m(YOPM)S>Sa=Cq^FfP3;BzPE!ZmkIX0lEzR@UlBuD<4zIW7|IrK zqbb)2)~k|&atI3qkojp~H_5y}a5QcVNl}DP$%0Gwmzr@&KJ#BhHIFAU+7c|R^!yk8590qL`2QIESD5vJ zSqq$_VLI05r_r{@l(OyTsT5*hDJ|=Z{4@%`lu8-9^V4V+e*7mtjaI*eBTo>9YLqa1 z8S2qkFlMe`K6@--qIbX|@+QqYZ8^=R|``51B3C^GdBx%_skW6TxVxV{e89R?@fF|*S zAq!I|zW2?zu!(-Ns)+S*-*|r@TwfeTRG2*DQN?ezZK-<0yC($-d_mjcTkdnvH z$6^cv3D>KS%Jmpk%$|Yz{q^HEsunM5xe{TRA-K6o7BY-`S{Y>PQ)v0USSBE2YJ!m- zsEDA2vKBtS{WGAx?nmVtTT^;*pu!)KFam=|@I^{45v8{_SD_UN2Z|>Ia05?-jQ8VJ ztuT6-HPI>)LV=(riCLH>P@7{R@E?SN2FL2vR5@aaHqgH~FrVMCJ)c}DpCBdG{@mT8dd&?{@Q4-?ir>j$m62Cp( zv{}{KEVG>{rk>ii$Ha>$(Vb%27(Uon@0u zTdzqwiLfw&{&*-l)=;XDF^)rQ3crE_=>h7KT>8nMi2vh5t<0TkH3ObP5@oz(Xg#H zY@{bVLE*Vx)p&(gxrZjLy1g|GdNq|?WukwPNW=;zTD#!JXe~c7j2y(!j|_yK z5wB~dK2n`S^5hhg>T41S&W1Ap8Be8{6nH(AGHO#ydhqq5a)nEbRrb{!>ti^7E1bV& z&MQIb2zP||2z{1N(iC#viq>JjP}w$6Yf>rsU0kyHL?VGGzGra6 z-J1oXz}1p=B$B%~lV3?ODObw!JI7m7O!|IJi-vB}O2!DbLRIfXE0(Vm%ZFmFI$Vkw z@LELq=wEX2tOwBi-HO6Gt!%zh7Gv^3zHtxbW_rU2T+ueG(|W!6kQ zYi$Y;kcO?*+VHm?aB*3f#z31HYek!S(S~0Zhc=(D6>YNDi8gPEJ?kZn(I&hO+Bon9 z#C(yoNK)Y8!=)|g3g<4gBD!0&mCWB1t?U$W?kVitudPd^oCwhr-Vth8r&XJxn(L7U z6(*Nq)M<0l9BN`X-7lQ(XHIR_D^B+_r>^TYr~6yx6mC*vSb4C_Cf)dcv@#D0iwBv- z6~f{{W>JkN<(CigSk3XK=J}xFsk(o8$n??-zV-ZqH}OK){6w_}@(80&dlR?eUeMO0 zEuy0HJtYU~VP zCWMaGZq!xNQ$+|2MhD<$&yA@^@j|c6VKw907y(8q0Y*ZAi8l3-AgL;r zcQ>Bh7=7u<2~BsN@FgZEARbflJ8eyBoS#V03twNMch`}zm-akvrjN-r)b$IDTK243?&>PS4el4 z6uGzrgdHLA?V}3FvsNuv?Urcy?vo~VA2+e=x3Skrq%QU^dRxnTAGa`RQB4ndWZcIM z2sup;&VnbUd8yWa`a9Sps@6c&=EgCw9zZ`m!Sr8xNA->L1QR{$U99h3bAn011*$Uc zhtuD3f=R!=D{*$0SYnT)Xc{;($QJ3m2^MkOb%IHQK2%DNkQA2V#<=_hlSXbzrL0#^ zFli|YB)@ZlNwva}7tVh_!K4p>bNB?4lHNZRy7MQz@%(dRtn7A6HFpOU4JOlvO;vEqOj***d10RP>ROtyEITwt(w=WU5I) zf!-)7au>o5H>H}CT@Mr8lWJ1_LS>?7khL_`q!IP0l(jO|q}%c1wN#Uy0&?>}p_VA% zAki^G`$)>iO0=WK2<|9sw5shI9a8=rr+okCI5__|^~$w=j>BS2$IaTce&!e8;Xt_* z^^k`LG!5i5J!Ja*^aq!?eRC=~s~3u+H8Sqko8f5n8n};v86N^jPvJI0I>==!Kg3wu zQ%QgF;|@F`Bpu`eDIdY8Z>5@4^^`Jd5GSsGNHuBON8)!8F-)bT@ViEs9wGVoEnK(1 zrJA%!^0!C|yF@@O%`~aomQ>2hHcgt20_=Bn!to~H!0%Q72Y$DAi}+pe<5Xh5`|#sb zO73Zzl(I;Psr=3~>5;AQJJX~O@#8Afq)R`6-(3s;#Rfg>3h#9?dk# zE6|{%$h{9zJ#Ly*wGC97^*PFw*dwBSNjpV@?HtjI?&v9`B(Lb$Xf>ZI=?wOzJ6b7U zx}z0*>B((a@*ujS)s)BEkg;9+(j98iPj&qAfIF$W@}(2^v})D?+|!DC0B8Op@uEAq z{I%Or|LC@ZV%R&;pwVs5+n!3?ZPzYp)@>iyF5UKNJ9M{Qv_om&l@vZ=ciRJiM#;~b zCN&Bdd|N7Gt7($8BbAcBG)+odoJiP0#@EP??BkWa*=o`D)`p{w0j$| z?lWBgb^V4|{0+AFk2^5EA$p@#*R5@+)2G_vZ?MJ0;_~1uS}gWl%ehy5iUwGtfg2e3 z2m|Xi@I3}L0>}}co>nfs7M4fEAF@+g-WXBqa!FxTJL2E86Mo4Nf2riJkQ6#?05wGX znw=8y52FCb*)-8UM^fOw2YkfYjL$%~MSmA7M%REgWZLK2phxleLd%+9W7U13v0X$o zO!@mTxclJGaIQ!eEQd`Q(3(mCnN(`wLUtO4$;myR=_+9asS*2k9vS>OLBQ?bu5BV9 z32P357a{6C2ZfW-2J)XyBnSi_^5*I z(2mk0!89>86%o@Kw3UvqDb{`YoS4Q5exq^Ii73ThHcf$1Cx4C@f@RYb*sm6k!9bWO z1qR-Y$B??RT0uVgB9*#K3i;IHFe4rD%!{l>3JkcWA(gu5MOGsPqu?W+Gri#modWK= z8dAww^RzS@69n?UM5}4gu#*4_GA!*G?#fO>!Oy-#Pe!NXcfYSv>C9jtoJXCY0ulFN zv?P(BzKPxmeOSd;(vsd1Q*M$Jbz}EoOTU80JJ5%veW}ctB`K(#1ZwERKKM%du)MDk zHIv*XHo!-^O-k97N?BLAO*&^690%P` zvpA5C-}|@uzSk4c+Awk(H2_@?UTPrrB2+Q+SXw=BZr^0L$n>+tu6MluLkW2F0 zwNF$xg&k|*0^T7A%uyk@B#|HlA_GT-e&0$|s9Dl1DinPyQQ`J)QMs7M)Cq_6lEU-s zsPGif$Wh`psh3D_p`_s07EW-h+oZR@O{J{6-6rkDk6CV$j^CY1Sv78xy5YxCw@G7n zr_$-G;BZH{i|RdN3?1z`+Sa*yw9Sp?4wuzAW)og-P3RJ}O(n4_(f-@5ZD{6hZFkv8 zIigExMzU$IK$A=R&K3w{JAmwqHrWmwX#P7{U%79S}_RaI4o4j*oAGO!(m0}g=@3)NEo z;E;V(0dhB&2P6iELb?Kh(%iCwG74YOFDvtmE3+8w2m^U4M_iB^bL z){^5L(I3)jW-w^RS9eo#{GLmcVW*tqI>NAHtZ?MQKw!ckpBME7qfl~uqm-Dp3wL$g zv&|rEED|}4;?U$^Bye`5EKKeG7??!#*vWh^Lj96N>PkZcX~pHCjeHQqDMN0G8HwQ3^OwSSp9k`u7M9D2NgV?u|f;me^>h#nxK z_HbOOoN@Nv(G15IF02R!1EGlJE%c4|dmsa!BYtq1gKOFJ+`4IYro^UA{TM~@)Qlhr!IAf8s zI;(3YMg{!ER{z2M2W5BVpWJSZ`gkL~gy^RGAk9R}?+t{`9v>+4_{Vn*gvOs;dA8I} zajB=c)YC0D7$}>ZcTVoPxSNDX+cwdeJ?hK!wYd@SkY8c>QV;k8OU6=e}Fi0fFa5GFWb9Aj!`{DA?# zKO3DluQuogOU69ING~QjRO6y_`v8}42v>}Ylv!P)?OJG+Svc@?ih)lMA;jShdrl{- z^BT?gmU_xc6mOruxX`+ujQPU}xd}ZoqS|kwhl+jkchA@WaXsb97@~ThQ z@c|0m_TEh3#<$`b)WNvP5i8s^%l?cpzA=rnfgz${KjWlXbA-`W*ot6f!O%Fk>m5?X zaLq5-Oh}C&OMM()h-8{;wjNsFv{8g%{Kee$op z2jiP`k`ey8J6)8{!m!EH@J0vCh_7FQx)g6m3uL+_zVX>zaafm(2OUgm_yaCV3(*Wm z_hqQt<^}xOycw^|OUC<6HfM8+#SzOp(&A6=al;_v9jB2#$>Yao0yrY4j}JuDubAS+s%|TM5o^qd!n|_^Odd9f zjE7qp={_$}@9wQBEeeHZw|a3g>gC8Lgob?P5e=xNBX=!S5q!nA6KX(pE$=pYON`L$ z3jS;^R%r1t#zLI)h>%g$R3yg1!q8ADlpa>N>~1l1{ei*5M-H%xJr!YVRABJPLNe}( zGt$dE<8TfR`+;bPTr*raz0C5DkCf`*${{a&WH+L^ilcWOSu5eW$Lh_v1;OTBvpTEK zW@Ym($5@{~lGlT96gNyX`bH<=gdou=nQbU-BK1%)?Egx+7auQCdCFEB9*_gfXSkL)A;I5N^vNV|!CfPWOf7A}c<_o)&I~9Ht$PYtMk= zN0BeDh*WsWq`%=7YaGBh(KFE^0yc*<)r`}8UOJed>>foTDq5*nHuY$+?2cATUW`)D zo&=YdO|Z&)_UMH%<8W*8kWiq!(1Mdz!HMvT8-JR20^Me;t2!nL9Q-Wy1j7|&{6gYz zYchXlGX_G&6-K&u0ugAXjiW2?j1~)BUoi%sLjoZlOCmlmHB@{)>&__dzzq6|Ck!5V zVT=VAKx`N@OcYqg^M{4TZ}<_IKu>eWgA=>h*<- zJt42%_o(p`8GmD#LsYu9nH<)Ha`?kYPdH+Qv?=b7GtwtY6W~CgdmH#ydU-T1drrD) zcBA>7BLS(YbTlaHY&s%;Amp_|T?be?K_E~@Mt<|jnMh`b$cmuX6R`&P{9Y>*Rs-5{;y{@m3{TK-S{Qa> zd_3#ONsJwJ^|+9gWq*doe(#`)^76??%KsIrE=<(zjiaayOOLj$5eJ@rSVlQ{V}C%B=`j z+^QzCr1J+J!m1_OY)i{M#a#zqUxDs$nPH$w03Sp~T_vWLVNB{pa%8^0Y5re+z zEm4oE5$tG&5FRf zQIbng8Hi7Bl=#4(jFoM8(1IwEjumH}LFISD9{+1A(H?wcTzRB9O&(mS z*4@Y$X{1l^5S{gai&q{>eB+51ykzLbqY@c>r~#xvdcJ}JRwzPr@+fR7OsFJU^d^if zRTM7u z-dvhttQzK`v{J11&op|U6%|CePcP_pZVxg(HJXN83Vcql?u7S4?Fv)ibGzk{aeG`7 z7=2ebSmwiR=dmMl(}-E*Ne}0#rRDT+H}%6N4I05v$&f(Ep2U_MV>F)@43*IMgyuuA zUKq=WOLG(Yp64wx9|k3rTaM)m8K<$(%=M3tpeyQZW1h_9qjD|4912h2{KD_6f}P! zFNSiDkk=#aCR@ssd1@;a2QG;WzJtn0o#vv0yHhFWdQuPF_r%=f5if$ne*CIg%6}A< z+&@&RsU;Uyv?p4sJ_q(!fxHYQ2UZcO4$(3sNYo*>(#Ypz|YdcU>`E{DlCTE@?_ z@#rs&k9Xrfk25H7Qg_n^=iAv?4|g|d0_x^w9ECcbXWTK=5fUO@Rexm{q`x(N$Yb%5 z91A0?EF(N?V76RnW1WOCEoS#PP^L>doi9q=i_c1BLvH=aM{ zLhWkuTGJKZ;oyG7INk?tH~C^G-3SO;3t}HpyJg*+bmI3oPU#!($2;ZiFbeJ?5K@nG ze$XmfYf`L3?rHk)CS;rw^a*r11>Sj{;{GKj@?f_ok9- zf!JZ2qj z{Hor5$!vd6A;Vaw;#jt0>ey3hwm|D8MJ^wJtQmVMUHW4xWz~*7m2Sn4*TyqALPckCKR-Hv+VlJH9GLo4uk9U60TQ7S)xl&WB2@w zoAG4ZOx+Ln96%Aw{fk3BTwibr;%(#4y5=rHy;5c=T>*Hv##D0Fpv%lSl;(t58+Y~DR>W`1PDCNUgxVA#}U|%y%X&o)V zvxbpluc1C`VBz-XUr?nyh<(<0&+{B$3hXzsE^$}I7fMXV4X(dl0{K1@59zqvreJLWy&=+5mEA&D?3nrZ6cN8@&@?I!Id4T?DtekzUi_K6s%38 z$ zYkh%tth0CHe{y3%3I(4}j#h71Y{kYC5w~kfN!r?`~4AP}>PJ{inN0sehzW z*5U3Z{p*iZ%E&m^q{se%m7^P3*ebJ%BhA~E_M@UUN)8nuaG5OpiD(ZyQ{5kMCOpBD z{=ievvO-+hw1;cE>wfKCd$^i$jyR8uJ<7#1?F}q0^Bot7?8kMFFCfN;tP(2}vXD#r zxcvV82q4h5PPE;xwe9eyXnX#jqV3v0bqIY)+Ql|W5hM?AwMN^b!f(qH)wtB&zqUk5 zO0-38jWcpoYq~WKHu&aGt?sRANf1u%w?cGx9NJUvU#aBEKVDG?O3ICAZX80yw|_x9 zcxt0)hpTqkQ;U`#!&B>)AL*%SE0k}Ua36T;Oqb$YD=B=-_EeoYTQZ}43!Yja)ZCgH z9jB;uloZr#Pt}zZ>-T6F-G zUV5%cwIV74PS)M$nza2uDrL<$*QAt#spOao6rr@Vrsa=>CexKL!SaK+!4_T1SgenH zk4N;UV}aTIP%1eaRwVK$_Y#b9!-w!HFBiWA1N&Kr&~c=!Or*@RfX8cj=_;U|eAq_2 z3TXWfH;-Uyk8Sk8?zT!PrdAC@`kC=Y4v$P{^7lF3IGjp3zKJA{d-*YpWiNoh@Tw!BiUf05 zg3Y!Bb3}r1e6P$S@$ucAomomt;@jfLs$=81 zCT$bZ_e+W@T!ya}{QwVlytcSK?UbwrNr5sJ6QWqNh*<3T&5p`S3)h0?pBOFh)&!Yx zZcShTJC)KW2@ue6LtHbhK^YvD}q#S^U%pZ?T@QrbuBwK?9%DIwY%kCVQq#_MX>thQ5QT)G9~xSrHk z?2GWhVI2R|#M5G!e5jaw$V~d%Ow{+U|0|P1%Ts0(44xD%{T4-d3lsLno0NmsQ)mkt z_BlNJak^PUMkSb(6D;#!0jDyep_2jic)h-eFW~o-@n)`&Xh_foQ6&GXtQQ9@zb#|k z$SGvZYq8QbF# zYnw}683=KuH9z>RWv-8r{Tn6wH}OgLyQp)L=JQR9UJ^^%_CHNAtwXd`zEzaJWy&LM z%HRG2%11b#rkoqYVYlM2n>hq+4!i#W2lhNot)^qN+}|mx-!au0NxC+_N%~o>c6pI6Y(sn zE66K=L_4wqk8Weq5xT3ao7!mI)w4MAZ_1IiA^v23i`$rL-Q`ct?d>$lSp&U&^gF9s z>14b7&PI6D3A@y*{J%K=b(Q}Y=YNU(7`u2ZkGKEe!l464+Eyq!E!rBlCdf$Q)&vZU z5?m%J{R`x4unIoTrFFVBVafe2uFnX~OhDLO>oO_RV})sI0=n{V@Q6Xu)C9EL^5Zzy zQ$Z?njXUu6&OO9_|=g@6nsb?VG z-zbpxH*D=1)1tLpskL0mTAq`lwdA+g1bQ$<>$oyX$JY}OG-kBbTC8R*KEpGVT+K?B zo?u2@xsZj^uj-kv+{wB-w>436=N=~Iidwy)*1vj~wCDtrGD>@xG&I#zOJlR8e%4ER zFA-bfh!|KLDZMP@2?lZAl;3*MI_IZH>;8U%kyA|ce!_-(T{Jya>-c^GZ=1w;OI8fc z0T`b#VNI|)hVTc9@CQtIg-!T_|4LZBN1=%|#1O4lMC+O8C#lWGCb0tYTLX6af{bXb zH!C`una*FS2wy~-x%N*nwGB2aQPt>wGuPy?41!He&804e+lPwVhs^Cl)1;hnL?1G@ z2{t$N9Lpe57S78Z*?*u-uc%YB7CRG+oKm8lJZPH*x;#MG$&~+t=dg7>^3ittx=qT# zSP9b{g)$8w94^o}TL7>sBOx1-)=pbP72yCyUoyo}?V8y?6!r}7Z_8K}BkP_7gI7{~ z0YA|mK5(0UqNcqkVK<^i)32O8l@wo{9fFB*udyjaS%b!= z(18;X74>!^@252r4)2#)p-CZ6Fwd5+&X&_~d^0A62Fdd@Ng+ibo+!&0n?k3xHz}j_ z*c2Mj-c)j3C}=lI3b`i6%4M}@xnjHMfxHKuEpt0Qw;i@+NjodrcJm!l~Oae`Ry5Qyep~M(7Lp!}JCN zPCCgXXZ5N?%A81R9ax3-pA3zUqA<>7qeHbq>fJsqPf^!sxqeo1{mgPbb+SoK?dfOc z@%G8u5I;K#9zrjC1m#AR(ZQtTa1WE}#HT{2(&X2AnA9kce*OUR=|Lt@aEeLJD&$S184L-Zg8FE#VAKA>XzTh4D;&Wm-Ell6h;yIe zsC--9g1l(8_qR6iSu3Lbt;4fiH0u;hbMSdsMo3=VuRwqe(Ju?<|~gEW(J_|zHEi*2y~dr~@V z#&4A9bL=m&Ufp^IP;P7wMfL2~ z-uYI$Z1z^Z&@1Sn<%e5j8#z9rY;0b2x2a|yDf~bLv|)oZF-@eVXEd?PnAl}__vA90 z*kvscV^`>35h=xHq^7B91$#x)3^UE}MR&UB2AgL1NSfRU`tkBxc_3o-_j*G%p*ou& zj$+(xu&>^2;6T)4Hp#ngpYbil=x*|HAiPm*8`8L|=e1YrN zHjC+MzlajQ!Ms4q*GOv2%{s29NynXGQdYa3Cgq)>>v*0Ni zJ{o7_1U-|nRYLSAzj1zkwn<$gfzeiZV4_9Fqe_@A;{yRNo@(-1V|QOs#R6^(gW@`y zlo<&OEVYU!&{JHdS7(!)K?v!ir3|?WkXk^-Q7spn&c*22xZ^KsO)EM>hi1*JsqVd2 zacFWdVr$SiBwCZLV$7{<%uhO-xcwh@5)f@=z0N&L+gCly5NK;$GTzkda&xE~8Czw$ zH8MF!+c@LrvrNhqZHK}HU$Wd!m_MsG3|*P@{EgWyJ7S}uW-$O&(5&<9FsED`q@uR0ekkDsGq3S>3y6t zyQis^z7DaT^Lm-&+Bs0M${D26Oh5g_0>9Y{wc#uPFWxueb%`mw1h%J_Ntxw7e}RZT zm7(YM#<+@ZkB38E-W!rOqO^~m@|j6-q$2bKCP!o`L87sV(@C*DLSa9 zNpJQwDSJdulPGIUPm}gA=7Ux3DJYm5fca8Ydur3qq^!EC_Eg!=q>Q(!+EexUCS}xD zwWkf|o0Po0sy)@dmPk{Gs67r49I9$h2hKMsJMmUrAx6p9^fakK$i5GxQEWky^a@ikDB}FUxpMXX$V_ zS7oqidxv&2lMj%A8wd~Z6nKHjN|M8H)fd zrbF>)3)hTNzCwqY?#UO!yHe`7Y241siFXv(O=AkLt_-{cb4Z~0shXu8Rftjz05xke z4|9p@kx$NoHxenT6+FgeA1RD((okR?!(U@K$oV-qJkHS5#&GL+Gy$#SGCbf-&biFR z9BY!Zb}fG!_esXSZ!8pct+-g#=Ub9y7Sl5c*iGJn{Nz`Anv}Lqxz2ZRolkq3RCK9H zS-X0gbjf8VW$o>0()0N7M^BTAuP`Yiu9rz)T>;NS@t_pPQBW>^_LWlHf0Yz}3O`W1 z`d?D~$A9VK4N^P}tE=FC##{4^!-Jacnx_V8@`ZBu%t@WCe{X=#KqO4!DJ5X;6_Df9ZW!h~b zWwfGW%+2oA%cQvJ4z(ckWVFrt0R zGAT!0{BgvaqO>akgQxZ-x<&0$E1=xr+|~*`(cLAg8S%TKn2yu^{H}Bz;33tc)waU; zq|JDfGI^!(d49eK&uB8w^BuRzWoT#Xr5XLkjIISN|4qQ?KmxA6X;lt8IU4}O^)(N- zf-x!qb*61Vqz;d#g$#V3fkgm<%VOrz$#0S~Wg}C3x)mtS@-xL+hM@s7UUC3cKI^v= zDeWy)<)ysw(7%^SMFN=sC%B@QN!K#+xLzjJNR~g2D0?dRNDQwRcp{d}0Dl8Tcmlpx zsyU@lRS_!adggV1FOyQP}nySR>HwndPm85E0=2MWm3LC zW}}p=w@I5PnUw8DH=C8#+oXY&CS`T*ZPJ`dQyq>CN{Nd=gIN7q0WlGg%GAK|ft!Wn2RDNwhC$Urq81!4>uu7qTTIHhwzo+i;zwz3lYYC! zq+~ct-KIpoarRPNiB9Z4IsZ~@dPxd)>rsnbiH$=1O%R6w-K#`^wzon62wd=U8Z=Jc6u~)es>aEGU-9}x)ThPtVbS|!g_nydKJA* zsul{rGX%D(6UZN$rt2P|`TaebZ(6wBTcpcEZ-mVs$r>1M18o_x-0!(+Rwf|Q2W^m^^Q|;-@Oho5;^_*jLt}_$SxxPO~cdj!Nd2Nn+g$t#_E|U~K z(H`e_*#oLYdO)~VewicE=RR_z^q$BF%M*@3a%==6wPmK8qGey0Xyk;57AF4ojEfQ< zRa>G96VZDYJgUW9n21Gb8C5S-GH3efX$C*{D24-}5IxNq`yV}G*+JXVKG{39|hDn3io7bEQnbWqHH&GUuK&PefY7 znZ@%=6|_2u>N%3a#698W%id{6tLMQ}ioR|~bzMC%jro>ckcTBr3aHf8dv`ruHAHZx$s z9})0j+7|0?hkp z5$6RLnB;N@!%mX&SzNk53CCk*EHH_j^;=?Qc+~#Agh%it67k<7HPh#nSt3n9BH503>I=hTg}}9e^gU#HeJe z3OXO5k$u_Qq}*ps%KEOiN#pV3x85e*{j5nD)W@VppTkHe@$3{Dbhk3CS5ma5&D^WK ziiIbk`7>x@ZvN+UG9@3o%p{(YA6#bA=~xgTa=7}KlzXot*jG{zG!lria{8Fmx7MVL z?tM(Uwia;^gQ>Jnl-Mq8bG8@>j_?b8Xj;6YPEys`f1YG?QnghO3DK@3-0$A<&ZX$y6H+i#cZ~N+vRKr+U4K1CgotVOS2uPEp<`1=i!=B zuPDxT_?Eip$>+60&UPU3!0mD=>LCvg@DSuQJ!JAkeXav7{+s8eNzG-%Hp^uPwLpw{ z5)6(2?pcm-(^3F$Nz0g6E{YaCs1;VRyjobC3>JvxOO{J4f1i0B)uPc$8DdmTyJm*w z0dIL)8TDy4>MzT+wVq~GVUt|a58RxVaz@q)u|b^(*ck%C1~;!brVWmiE7}ZNwEUS` zL}dorrrh={aDf@#S|Mimf!VcW1}p{Hex*7eO}4=%4+}i6Ebu&A;JB4)VD&tE%M~lN zJ)UQ^$yFmoYb9lytm4cCE78gxey1Jfin!;Zl{wDuW72Fv*TuoN2bA&jG3hT9{O>y0 zVh3ky>wrw162FecAH1qrE2!gEaN{biN1X$!k*L=^GU`}vtnOFZmO}qe8(8~otGHX< z%nvwPxl!EuqDju8Z4xx!U`hjCLL{x$z^x2i1|YAI)QMRJ!7My1qwT6*$G{XA zx^-MF!?Isj!9rD(Q9UOJA1BS3^JCv;+&^R;e1n1T_(BxWb18Nkf6{=krt=Lj^I%#P# zqN!`G$pUG=`f6Nav@qnJVfs1^x z&Wt+bcy2e^V!Zow8#3lPgFA4fhn6Md>-Qh8hcDFugtDu-?0yC|YG54$tv7+0`ze3? zXdAP90{Ce$&JR^g2O~@&ok+{ zjV5KEbRKF?O$i!Ayw^Yn(bV}Cs(%)iVRtIY*F$o4hysCp31m3UWp9Z?jC>oV;57Au zvqR(j{THno=G+DnUr<;c_#h&9_s7sktx(5I7x63I6&?D$ohWYiIn!a zvUY2nLgLY6*ovJRo^8OYXB12A0!1Q}1MZn~s&q=aBio}SfgAoYyaJnoy(8j-612fDsywB`uJ+aR-B zqdJy+4#{>Ebyq4yc*+>haI#pjun z|Alh59;n%A=b5zOW0NwbpJ&o{z+nfxT&n3tNnzJ>U{_v)*U-x6ludLl+t|5PY&?7` zI3iXBg<~&pL_ACXL~!IA|YJwjm3}QyP@O7ae~VxxWe?$;0g=RGpRtb!pPcio=Km6W>VJs=b6;= zbCa^Sa*vygt;|}%EM>37NPYhbQBte`QGKA}>0{C>pMwPo(pQT6jVQ>2{`?pS_5_0h zxd}K-+#wb;B?a-DHSzy|IBzG-l0T6AJ~SG4D92dkHA(y_W$#3=zocy9I= z;<+QfIGX1UEDMA!n^OL2WgF$XcedfMd}kYk<-ix(S?}ay#*|q}ceTMlZuJ*`sq!{*4V7(lvY~Ub=vR-vP*@xRkGz@$Z36cocU)bnPy* zV6g|FCq{7>d?};2arl8z+|(~oj~G_v3#oY^)nel=SvU}p8ExhgF~*hlJy$0aJ)79D@ckg+NG3!LuVBst3{89 z5;@v2wA6LB^eUtk2<6qF499r*D{+jMnMF$>4Uv;S#WFD9=eL(N(W>>)rrE80WH+1U zv#(6b;ZLvZZiA1nocA@hG^_|2ySajweT~&ykH6R|qo3OhUDAfO0Mf)NQb|8^N#`!S z%djhvGQ9!6MTgqF{XuJ*y33?Yy!agoRRklWtuP&KQ`yBuukC^})dCwswOM=u{5J+R z0GR2amz;y1b5ZwiU~Be@Z$iofek)qIO-}A!HnE8Yey^kICg)9$xMaFqE!8#>)dnZM|9kLkX{6VTXOr{g z*=%#)*6dYStQYE|KpoDv{d;kW1I#AAg}HNatZK5eqA3@>t2w|YwkQR+u!5)kpcUMr z6wD0c!wvK?KVA2OsX_+rQ4Qi!H2Z%pfp~%DU%e?>j-Q-{e*5SrehcS;9}uTiD<%5L zi5bQZfR*v5h5*8NwEJEyvO+;mmq*4=T>J3uUsyB8guNy?f(SOkwv(hVLKwz4XM}Ei#P5c7^H3{-o{Pb!Dwf zo4Ko+{HjZf!_*5|hki0~h3MHo-RPtnf5maaJs3e1Q=hTRT=c}R=w73(QwYQj00;R} z!%?1ci@X4S51>^UvHa?lP9K2B@53zfyF@y(m@fZq8D9QzrhN|%7cm!%(6S={@_ zWix)0viU!>SoVS6?6Or-_5#iK2VJ(V<+A;MM_EVZtd3OsfvST&l0yEWsLj>0I?`jm ziwv|^$-sAbLjF3YkUYAdNx6HK)&lx>`EV{Q|SJ(2=*Omj@uCGAJ0iwE*R zmw)XSc?y117HFo+yGkDC`6i9otmN@Z3N9A7Tmeb4J@~IApC%ZyB?Tt#w~djU?NcT4 zTmgCDQD^=s^3;hu&Fr(}PiW0$=_jSj)w*obUs5*zC#6fXvXQ^+vQ<(xpv%_lvUM$& z&HNi>xn7)Gl^r@t3hhF$Lv+2wh6wrS6h{adG2M*kZ~r#QnF~-w(=~^KDl|Aol#w| zuXQW%>#0PDQlCCS4!WUV4`>6d;P<|y8(N{>sLKpC zo$G&v@jD+haZ~L|Mt%ejm=&yK^rAz}PD_lp{8q>lu`aHNSfRlKFNEc^q-kG8OS}OR zdxCTkw1H(>0&q~?)DhV6PQVgE02E*(~!0gcclErL>fgz7sLGuAMPke zryIoTv{UHVIvK-Js7E8?U4w*u-wBKsQN>T+z<)>iH%u7Oh{E6yX53-DbDOJ44wg;J?D$ zB^di91?GFGxoFJf4i}h|zfY;~KAIavk2YFxT^TT-+*P#&y^;38_Zfe;mu)F91Pv}jMX!kY~PH{#&Jo6Q_ax3zMUj;#uEe4?9j ztm0Cl&BnNuF6xHoiZIs20Yc#SCF=OISwfDy$!2-+9@S?YS?^7B<7&`iPY}`ULpJ6* zfd#ylmt#L|W%!jzzzyLMA?#*{n=->z$mc2Z-C%iHIr@q*7bOEzy*H<6s26^pZ0#n; zwHKJwC{=twQZ$#Z;CXjlVAA+D?q)p^_jiLv6{0V(t=0HlwAFq$j2u5k!}L49^F6PP z*8g|52W9eo4PpArz{mPNYvU$o&2Nd6Sw@Grz&TDgIU4}+k$OLPfzwTyA-X*di=sa{ zA(W%{iS22cR9B6p5dJR+KjOsp)Yj#u>@g?u%ZbsAJFD~HfftUp@P#^Gz#sEQyH$m~vPX#wa_7XX!bIYFrey z$>yNTmOd)y%T;R9(v6H9xEJ8(kdcFWCz-%d7hHt#r^`)vgS!ylj2doD4%6KRO5bvv z8%LTe`tps($Gyjyl|X%Hm27tfjFDoSWn|KU#hDfY2# z3LLDsl-^|T?~cbNsItd?#9TEB`KXW1wBK=cOehcu&^d67 zzNQ;5Qvw>U@I|a)Ub0}|YfU#f>uoTV;Paf%0GXWme_%!!3enBb?mYmV)qoXO@OS@d zmVx^&PItS>8QhdQu0oVv22&j7;=DE>SoVkA{(wTt?v`n=>B4^Q`5SCuy zlKb3Z(^oj2yxb0gR|A6ur>_l6SpRGXtm0A`UvD-rM4on{0u~IXFN}myPWngy3M)+S z87pgDl-3?-p%5NNw5I3r=<}BZXb%gp<0Q9?sk%$U8zU)N$=T?h9azL&A}NJ>R6dztk;*x5VN7>wJ^W z@93tCYw}IUI06n%xM&E0Wa42fMRR6Ly!K z7Q^m_(@-4jQhrtJ(j-OkIhtMlX~OQ(G>tz^@McR2{COIGX`0}7O9wuzQ7d?xB(?D& z-n4WPZ)19lc&#%~{Cd=x6x%B)isSn|S@ZHuD#~zE#?$#GJ<0gxbJ|nZKBZ2zq`-d$ z;#JRSPa7C*MZQV1C3};kKwAQ|-}6m6l;Nh+59jkt7c_hQrfAkk3go3g2F>)-A*!Rl zNwvQ#w4IUyZ5hzA`kQnGqru=hdv6rl3qV8tPCs1?cw#4z1@}RJD6&f>1=&@=2lsMD zL({u(zd|dO6lk?*6r7v#r^?Nh6uFo_sa&r>r$~xiOo}METYr-pr9piSreph?w6K$# zGA8slsm~d1%BbjX(luwuaL{>B@#!lm`upc0g<@3qm&$IC6pTJXHVawyx5`~0DROTC zQ}@{^RCNXn0!{ZzcJ2X%4z!3y`x9u*MiT=Aet&jXb>ozMFw*YntCKs*2V+3!pt`!} zAYa`R%5=+~-a!L96HntAQPMcbf%u0^yDPpU4t;UTna%t7JL1sCPX!#k@Evg&B>aWP zuzbSxo;ZBN^~x+aI*ZUmOTNvwDGnd!`v71W0nqk1bWcgyZpy^Hk=+IbCi%yQJYI`- z#9F)asOQ~ux5OTNUw-xyd0pz5z)i^E}`;a#9e z)1mg)<8U?k?SQgyV?v&irn`04vWy41pl=2xRKpr3e_vOGYS*-QCCw&D8RJ97_zM_L zI*>?bmWJu;IQ}%U3hrA-m1>X_K9Gg?`OzB3pDnHN?z7>12xt2R&nY&r@e$4%&X#cI z$c+hSXXTL&*UoX1>s4vBdnL^*_7s<5SWUBe)W7ejg5Y|gLTP)NZG#!j z{_i**_WvCRb{BPb<5tzOvH;QF+{j<;?xrrf^%@7b`BE2lS!qwk0d`jE($t!45d79S zz-~#IKCk*_EBzWjWTlHvJ{O~~;!-u2yfq$;dLLlLVVV=a<_#B}*u%{Wmg<`deR{a* z%!yVgeDMUTVQh*yR97F5c6JT0qP9jZoDGT~(PdsFa0PH%@R@IjG1eG$sL(FeE5cqP#f$M>E;l4{Z0~PUU#BX4xuo#;=YYJwze(5jc2fooFzL45Xm7tl4>cQCjsXhBm6Q6Qfv^Dkg|X>Uf{qGdSyGH^y)YIh#EbBJjab?U(p zA*T0v5@N0f662&DRyN$KacT>v&}jN>M4){11r3EQ?S8W_bq%*enjg>m;d- z53^7|F^k4Z6Ra4soZb)I`J0k;qRxJB#J47ga9&Fh4ci%GMilDFxsD%itQ+K#q3%ql%;+0V??^!NTG!7_>V3d_`1vn92PV)x ze2L2`=ewJadYX-Q_%uRx)4>i~&c>6Y<$W~4$SDmIZU$Q8qWtrZEtFYIb6J&1 z=esErUswrOlv^~9g;)j{+WtI7{_A`kP)WA0x=sF=yJ4I3TH{mBEQE7sRK}|K$lWi zg8LX)rGZ-*a}5J)HSiM#x(A|og9h$k-~|bO<7+L zFzF@y_Zm^fOM&0ORd_@ zV&sp8x~UU-NMFEDi+TLqtIdUO>ZAsui{r4ilRuClJop~%@p}X1G>f@A{*4T!cuYpx za%7Zy!VxPZeg0xDapCtDx-pKoyz~eI{KM3YV5BF)3&6C9!NnK-!}f|G6Sc@@+ps2G z+Y)Z@v$Wcu6VPgRU8F7aGrN#l7fFd%4ibcep#SPRT>7sm!`-U?S~OhxuY&j_{^o$a zF7ggg(0~0n9HElOvRNX+g_1&$5d85UIgU3lq@t|pJvCdQVAp8HZfm8&(QU1uSjq^k z*ln#axUtMd^V5V%`n8q&O&2Z4BfgTprq!Z1U9=UC z#ibsfe@vjz@_VVi74CuVK1wD{Q(IxuvCa49{MxU8F# z<^hCzO7Bm^*8vaU5g+=ry!1>WHg^|arO5a%#ET3NNBqjw7#fE{^e_vWc#WILwJ%5cN^TdGvd$|R;W;3viBf(m<@oo)@i9?K z_?JwSCjLuLf3{^P3i~C6FFy`n=CNc@D7**?7)zShx`{8!F1%J#s5)IysF4&DRx^dx z15K(C3L8P8^FWgxL_wa%)k_v%A_VbPf;irsI<80)Z_HA}v$9nR;tsq>k7?sAjD{Yf zuCvPCB`MJQ;0=6Uc&Zd@O_3D2XM+f)y8DY{s@uVX2L4Y%zZH&n#+CWPrM#J;+ih;1 zl0glB)+8D{oLH0CdX0;&_vrSkZfimNtxH7vUF<=l(`CnhGZ7P_ zr^nlrqoBP?-Q+lDph?qmm08wH3gZ{R_}2|I>9tZfWmOI|sR2Kx4m9aBAF3V&-;x3e zVpkNrL>K&*PYT|OA1L@V3jU+&3x)#4RybV9R#V4Sblo-fRKlt{+$nXqlk0G^PuJm2 zzTrr^-<{m*P@$i3g*vX2jUT>hU~y8@)(?LYXK@mL6URx35|qB+@)ul(_>-TQL}yx+ z>ej(RtGFU$#lHFYOX3~pV`I-*ql^E=#n+)Y_f2!fTCYk9tCgW8aCI~H3^b{% zOqN9t2%V1SD7v8cJm@jExnk~DB}MK_$W_?|QYa`XvMppM?-^)PibLsg8-zJykV#9* z+)B#=$+-eK(DJJ?G;L;?(?ba|LQ>G20h*i}6lk@i$i2_bT_Mmdk|K8+a{rOr;pT!N zKEE$qYI$u(tLdY}Q4aS?k_sBHBq3;|m%HV=c&{X(1G&6hJK!rxdJynx5`xGl<#rRa zIlP_JZKI2h_e+QTCVz1MYCNJdd6Oxm2HZTg@|8@c515B9fGux__~z@3)iwy*)jr{yyjW>l@bf^*NgiG2$z(uRFq0E8lUg)Z-Dg@*U&N zkGRRv={%FX=PBX~B!$q8@kH4$pIY(UA3?Z8%dHorMoEGE?LT(TuxFwbtD)Acx}Oq4 zY50d|_y=qFPKDO+50)Gn{>>Vi6VX+K=?(`L?}p<^FC5U=kL&7Eb(! zbAzzuZ)!?fz~BYZg5Kj$P3ay7n$plox+&e`KvTMTk{0|PxxAr7x6B6}Xg6O@g8XzZ z-w$z8B{sE_rBU3^CHhvPQPgVSOa?v(U`#M5F13m$47Vop{To`swESpEpKusCexfJ% zK90kcZt7C*sU$-^Qg&+hkt(yoRO@*2F&Aw@2B?PVdB>e&T@?RsH#zg6 zpO0Q(?MD3@WyAhF{?z9ST$6KvRSPVXrR5HM*7y$~qCqWZIj+0WP0qA7NqC2LHB&m_ zCQK^;2~41uS+H|&LYwqd(#wo-`Auv(a_t(Z3{g5rrA&7N(_IK0XD!Hp*~bj*crzBN zF~on@F@KzkCf)2NXCrV}zz-cKk8;u6n?c=K;1r+Q%|+MPML*-B!*6kuvj#X^bQc#5 z-oi!6B}M<{qB8(>)eTWH4jQUb7VEZTbQZ0v5GcM}*?}DWRptmu5h!{m5b?+OZ>@4u zGCsy%P9K`R1CZi7fnPeiSA3r-l*#~H|1cZ1pq8q`Tzi7TrffRC1dtobDM4& z_a$SxN<4MFpMQUH8;+2LJ^d$od}TC~KOuG8?P$ow(V!W(yUAINY6oq6F@_#WuE}dn zIa3gd>W~xmjaOg0q1DNlJYGD-P0oVjlK9j0T;ydga>o>q4~6Juc8;e(A=WTDPK>uA z>aD~Ptsd$0@9dwc=MrYQ&#As?Qh@Xu4U0p3f-N`Z`?-(&vdtb>edL4iJG zgg#TzZUBayA0@+yub&E!i>W{@r<%q{7dFr%TVh(k+89BFo)HGO&!&zjTg^g_1?gF5JUmGeXb#w^qIP=D-p<8uLB|gZM*`n2?Gitz;@+>Xk=mlD4CW_ita1sx^{e>FdSRO!kX&}`Oy^V$AS%_jBJ$_&FyXC zFe)n<(iE4=Xs?3pLOpnyBMshKXUzoia%y38Ir|FS3ZiUO8$^fw|dcD!5tarVDX?Qv$KfPDY?l8#=L6plFLZ_pk*OW(XZF z&x$o!Gd`f&oNjWfS5_~_1fDs0ETyHJyVofIqqgr?T(dXXq{V32PAH&tFYS3$PG>T{ z*t;XN?fZ&uN!HV!xyh}NzZk6M0srQ{rJ{^4xR!Glu41g_NW7#t_4U^EA2v2~1>>>f z+kUmkw#!`sANld13>L&~8g{xz<*y{*Z_T#n9Q#wK6%aM+;;NE~b}7X_vf9$~wNWT_ zN=kc`rDg}Hmy*1peoF3swn8yEd?HHw#w8;a7{~Cuq4nUkWc-6l&cM4gv&X+w67(>Yho0GOK6%EvM8LwS=lsr8c5Mk?2b5q>P3rhyoEmEMre$g;2_5@|_gb$cf>$J+Dwv zCJMQUwmqS}b+f6AIs$KVqQ-lk|2&WPxK-ruQT{=@Ar-_=bXE@=(W0TWZzW`VPVFXB z3Kx4PlpoiP^SoV;uXn8y4=3un4Uj51O%`fIofhir)cKmW=Q<;$=7Y#;b`)nl!VnDF z#466!v>&Z<_L~%{6@LuzF@)t?YiXDDiW^xw1Pc4m4T;zlP-lg51nA%!kY5rHRWoy_ z;j4{l^2dGzEeQ?)g^6@!w*aupPUpvNLVnVaF%>m{q^Bs3jIO1N!yHk=j8r01Wi-(` z@WY9^u<3T4bf<-?@Is+f@#r>RNr>blmR}`#Ax}G1wW%$K#RT00U zEnCHY`DEJK?YglP=?%^0ESfS0Kw(cXm7Hu4PkF@>yD*$h8m&qop93iE>3>~(di-55 z!3eYmiisEZ+~e4ZOawAacc&q7-8Y8MNJhuX;um;1?W zs9F{RA3i;`eK7CeOAsEACZk5v>nq%~T`~kqJuM>kyTlbvsk>PLvgMcjUPZOTCdv`M zcuY9`ZgI^-XihrGzuspA2ZvfaFQie2*MABdx^mMXx@-EGU_3h*?&u2!%r`+$~5k)VM+0#J^7m!s?B2 zwRI~`Y~=m^@O&SHu!CZ@9ckFoS#oY6}x6w_Cu&6xA6Z1gpu2-p~b5WS9S9wir zov&9Rub3*KZBiSq(v}~piUtJkctD;YXYpT54!s{ne(~@q5k_+r$h4>SM@KjhxObtr z7{N>$3Cqui4b{u4oa$&&%upSnDKL3X&-MU#igprmH8JIEKqJon4s2=EWJX!WYM_(o zs=EKqouiz@m?EJ(Ty?*cHuYRM+yum9(yaS-jl?ArUtr!!ZW%$>wi`YZq|ye1Yx(u? zn>4|e?oakZz%qQtIiPXn>&H7lEt|149=1i|cS3oA)fTazviR?{{A<-q;rqSNXHbLE ztQSvTvj_!IVSMF9O=Qsv6+f&94onGGnV+x*@>(@+`L{` zw8j`q+&DB53FBA*TJNn)tdmZ)#}Y5rILsDMjXZ-|T118h=@bQ({J|6%WA@VOzHgPz{Pr7}uCf?GwffRG(K0${v<=Tc?QTT)+vCt`u{n!q zKhyKw!=i)MpIg$>yB)7LHnA+8u*bfcGut8(Jgf|Nxg%(fcUe1ijg9}=vA}AA_~rbe z8*Kjpy~S$#vBuW-eCZBR5%g7_ggu;PLI=ChmezK`g{@fIHt`E|7KC;Cgp^uHYucPq zMj8{-1WU1H2BMr}V@V3LxZM3u+559U2w(C&&@k*(mYXL$&>jEe4SAtUR=?i(j^_Yt zEJ&eN90+fGt}BS#I~7|MKLUU-`|3D*PdmZNq4rVtEA^~Z<=g5w&rds0r|WZdk1fsU z58C8kzM3rcg}^Lsf1jud8{N1f+sR3f={_!!St9vZ0VCLc?r{9G$>8}W!p3pzSbQ>> zEVS4tD_k`2Hmjs``}iz7udoqY@)2QYji)a!YunJBQyY@Zd~3yK#KX}?x{Jql-B$H* zN7@n-nrILlg&j{2(Z)bCsE1abEk@x5g|Zuc#XD zQzJ5nw)1fa0ar_{*#Qkg$;=0s|2~}XaOvWdU+fT9RM*)w)6fa0A;GQM15NB<9cwJ* zLdI*K6mH$?uvECyi^5CTj6_)H2^nx42J;oOjXZzOO~P=Flo)?45`OME!?W#}i#ym0 z7tH)QaIK$!;`^N5$RVm-a*av5|1?m9!{Dd0I;Z&fW(kay$SC~`D2#LM4gp&*-ztUI z{6;wWg+43aaxVcb!IloONEZX^F=p(`>o)KioCi-&vzPI#Jy;a!uzH#n}%wYO3M-F<&0A*`+h%- z(7rH!w@|cb=-G7rTZU2pA?ayX9WMO}xy7&^R|iVRAaup?mf2gNAeOFy$ue;JAy%8T zU%_tlmde?)fc6KMvEg0r_tGTU1VAzRn*B-58B~N-mCY#bRs2bTkq(d9t$5pP#WcTW zsc=LMs7ypd^DlxZ==}8;M2~gBzIWrn!y@pBF*WGH^O{V9q1P~`@P!5V+rJS4A)JBS zI!QR0cQt#chKTGOvrRg>)E;~DdJn?t_*boDS^Jq7p}l432y;FulTJS|NsPfPavTN> zH}L?ZQF__v3j(-nYg!E=BIB`_7|O07eq~Z47X7U-P_)Q`roJ<8Ttxuep?ceeY17}4 zsRA0f=J)QN6IcDRiiG41`bG6vMPvhTg{028ilEq_I{)G(;GN40;#!N>VxJ=&X)-So z=kjCsE)q9O5%;X0Oh`r_|6wlQD6@U=DJ+u0U;$f-DI~Gt2RH{4_sTN&x$s3`JJ~CY z`t#H3xzM#afKvcsa6xAML_61C5^!(!2OE>Uy;k9Zg^4fDBeFzT+Y;Jn`wRKtp?&4_ z(WFt4ckUAC-Pxw*qGRn`uP-cFicBPp`5uvH^h?G`x~z3mxnX`i|+y4EGAS*$wm9wWf=QENaBfO{SD%{!=YXh^M)c5P5Oud zqA_`YOyt9F(2OYEtfz10>S-pwWY=lS+giq&bpBeUSN4p-p`Fz9TLgY{D^ZUpr_~sC z2uUoYY88HE{+1tKn8waJ{8Q|&ws_iTt>_jsX%M$d1BFb$(5-JGd8b48%^qLw9z`q~ zZrWZ7>d~zyUM#3^Wk&W?voVNG^&soU9j-a?DQwR3MXLNk#5CGj<+NFG#Vv+HALv$z^ zh)jmd3A?*J#UU?9RQ#rWS-$Uz^uokGiVE#<6 zl!HGCd^E0}(E-?>E05Us8$3B9M9a~Jm0M2VT#)DS&Kaov_*=ZIM66PNg*9Jr%%wzp za4CTWN-uuP%n!D~k(L_bycF-2%R#}Q<*sFRDhBSYTpRKL@}Y|$GmT)c>nQOp4{EY{ zM6H%3+}vR0SWH3Z?C32D+;wB3_I#U->EDA_tHy_gimz2E@}BZSYqFmC!Pld`tq{cNKdog3MzI zD&+quja8Cw5dg^6_gLeW#D3wwhqb3wq=ka@N_b13Ta^HusTWZ(J0*flmAiM<1|I&1 z4do45XIFG?)*_z5{!#inrN z=#pz@{`o~=ChvlKnNX9&g*rt5jk;^c+}vKhK4!Bs_jqcDGT>^fM&JwgJJcj0gp7wi z_i$xszBZNKbk9#|h0}koH;FDNY^SORo7Alf&z+CrTonZr!ez>Ho9;w&v6+wbe-+>q z4hhBwx6eFn6*Jky>%I>j|B0M?8w zrr4l7+kt!72q#oGa`*1THnOlL+QH)=MZ^RdZ=2TfUV7hs=#X^$Gb@3kN&;?Q;-_Cw z|CSL1h>Vvi6)OSa$@zbiUx!Ud`@wr?ZPO1f0z|2T)m-Jqgd7NFX zO7XdvDQhtjOkqk>S~_XBUP^^3SZDbaF0ACoR|)IC?N^#FQ*{sQoLHkPs!_#w6luaqMgcCx%X{t} zwqH0Lc+9yR-UXVXmZN63z+H{<5FbA}#Co`BomQ`hAVt3+K=ebQ+k&_3<(#kQNp#e% zn_krBHILIhzdMU2X;yga6r#Y>ezb3(=LP*!{&p+10r-qp5BePJ`P?OqkY7HBu{mto ziz2l9<~1TMt}#^sKINKNHHf*xzLaIL2zGRpQDdX@U|c2IIScmm>Vge6TFQgm&7yS>#12Y6xq`NQXk_?unlQ*d>@Wv&L|?;b-^EY6X?v3 z^bCO)JyG|#kn%0WKE{E~spXm2{;p8Q$9=p5Q|~J*hg&H}1e|k<#Wae=pCjwPC@WW| zG~kD|niK*`#fh$!gd*}}F6Xvh(5bMrRwfcFjJ~QIK_gTg`b=$^E6`7e_(Nd%7$Y65*z=RhA$4V z_Vc`R5@tLi;YzJ1ORbMEW;xr1MjDy>kX;IC<$BN4YBH0JNRsfDKV&Tq!$`2r+F!d! z>$CMq7~)|*eE2EO%r(Sq5bG8^y!IX!p|x=QSVEfa-Kqnr-{%&PXrmbB?S_p-auB`Z zRK&PR&$b!3a8o^%xZtx8_(UVfZL#Q0xNzPh`4a}o)C_%+>GDkAa!<{;s&C-<*Kf6h zS1ry_!_TtFc;_hQSdLM&&$DDBcPk9VAgXCBWd#;9^KjrEzPiFa%d?3mF-AX9bgH*& z{5!*~$puvO=+oWS$_oh4f>b(3o1aaw857H>)Cs{`{+iet}}2B5J|OdTop&2q`^G@ zUJNF5kP`+{v`;;=L`>B#OQ5(bXr*{0k1fREXpb%nn{+{Q>u+YL5J0 zu@6DE@ms2vJL+RiFPGK3CF{9@h6BiEfd&XPjT)*=S@A_ZjIzJ@-0*~_*{uZiTcFRK z9HV0=A!Wn>o;LMkB2w%5D9s^DOx(;kXUCiRN^Q1T$3d@4aE0gdTh!EO8$bRAZMEWa z{GC^%s2fZ6aHtGcPvLCh2N15pn4Qr5t2KdkDHJ0_rp9DLYdxm*2fH{{VA1_%VR#8H zE%Pscz&XQ|P?1*4rL17Ni*-j>;hsfn+;ua*+S&R@Q-#BpdZUciMg^S=1V+qiNR^~My|Cf`%(!{dWa{rA)arBYzivUy-xC8TQ1zZxPzOD0XjGxTw4jz6ocSf1rU zBjl4NtVj|`7WV$(SZJ2*Ml}udgu#i_f)um4v_P=;SHo9auGI29+BQMVh9LlSB+@nl zG+olDOlEL4^a>V)xB^D<-d8Q} zn)e*#1q^uh?f z*Kdo-=vJD6BBPGzQYlY6Kf_=dCtIypWl@Nc%Gh?zh54^&ublq(;g|O4%-`%}j zPsK@uFPYz)>Cd;4xCo_wA9)Z8(0k_uWIc(tdHa16nR;*lD2{!>L~A z*bIHvPY8w9s53oL2UTXB>G036121Zfd?j8MSf7(ro^?S8F8fkXvT6B+=}?UOSW!3< z^_5?^(@eFy*&-6!|7z5*6;CBCtaO~kqa_62aqFciH;Wq}%+67q8aT;5V?yT{7{E!S zT~%U59qYQSQ!OCM+Y#ryC&6hZiinW3z{quTCmj{lrrBJf0DGt*HEWU9MXLz%t9k0I z)WJ9+Qe7&(snfqFOvq?5+DJ%zu8M0IW2R-60jLM-bc&3gea#i*sN}}$EnQf8$zVwE z?VijY#%=8WXHJxKbuV_>4!O*$twwL#D1w92?`Z~vL-F1%a*igw3+d_Lyx*kq+zVCcnxr{%$FCB+s>#E*Zl3p$6nFFhD3@Kp_1$FToZF6Z(%zN;Fxal8hO zadgXgZqRM$5g>O|^O(1fZkB*Y#DZqo40NIq>?z=%bn2z=eZi_s=D>GXk77Dcr9axv z+X?$t*12j-zw_0u{JGgiw~JS3gh8M2e#UVoj`?YtQS=4Q1lzWGns#THB>_;iOci{Zi=UC>eM7M*;^a^Ztg3Em@j=o3H#duEz9!@1;AtKv zYaTX%l*r~MHE&R8e(uaW9k|jemZPj&(YYE_UnM^xFQzOhW7}do!zq~gHjYJ`4sCAT-FKUEc!^)@;vvcnM z{6ThPWee= z0t&Sv@xlo|cJ;&cb_kZm(hQcW`U9-gwNJZWF@Ew96079lJAVcNHmx5kIypA!~Nt3TjvCwkj5OU)9@4`uv zO=-D0<75u!%RxPkYWvOn&wxMqs!QlC)dTIW~bEojU?2E zNzQ|y+SPEukZ$Gz|CxB{92S0&A9T;UCBEvqR4@Ix=k+gIZr`idy0l=O2Ax;$<>#Gm zU&C>(N8hl$SKqO{iS9jffmiiy)xrwv;{J%uCkQ8jS=B7MNy%5w(3a$duXZ6k(VBc$ z283@h(R`V71DKf^HX}ng+jdFE*DHcnT60fA4sqJUwFhrmoy|x2k2R} zpi2iPnT7)>-_WSfrFv)|$!Ffv zzdu!MKC^Eig_t3BT786&ybD>o8s%fCm!ser(@n5hbFYdX3ZAoR%BuVn-k@MOW__QQ zZY_qGwLSq`nI?IU@4XaIKt*-k6YG6_xs@Xq@$bg=_S(8fM4GrqF%miQLtHIzGo2=Q zH~h3KhIx;i)e&-^a;Nv8E%g2n<#RKY{FXrxJ(;zY<=Hb?9_2@VlKTehcVCufA5W0= zy1>?YEG2he2jXQ(_*ww@d~2Jg%5q4-=O%!u4~N>v7WpI^waR|#Irqisqhq%a@5-GH zzTsD^xxADWyDC3l&54q1q!7e(3k>z=-;vCS@CX*yKyRb@PW+$)3qeljS{C$xQgUHa zekSgj9yh67-vaa4n8x_qk8w96OdJ!kv2Cb{leJ!%OxB+v>g@M(_}+iVn}lRKo4V4{ zSumhd>BM8QUUC);*MxlSx#II>;&)cRet!vslN9!w%70pt3+xcnu9}f}1m4q%%@N?0 zeiJdJKkZ!WwGpxkv_65c)@b^D?Msj%`IB>Jakx#i7n4>j9#NptxsQtiN{k#>t}_*I z#Pq`)#6Pg;ASTA+=8Z(vOcQi<-h^s9UhVFUl(U=r#3y%B0nTf_AU~=p#1n#aH!Ups zny;INab|*X7UmDPl-IiIk6+i68|WcI9&^NBMD9Uc_E@Gi70|(EKH8+IyLMT2eq^xa z@rW+WJ2ImV8rZCm{6`UlpUPV`etRl9M`W#K)_TV0ayn~yGSR>4uO&`=OW)5;k-sH@ z@KVfM*;ULr5S6{1gzyqtri=e>6R$p62Z5HMK8mWt*R!4}27(6<>&tZk|UOHJSP zJE-$PwZBu@>8yqE$tEe)?^9xaqZoyy(?50hs*&=e5&lhvp!8|B0{L@r>oT_cNzW); zVs<3ikC&=hpzRK|jGJLfG>&Nw7@*FvygWj$uBxt(78qsItQ*^`?R;`9IzFXcwn7c^D5`92@2cswQ%SWPfcg5TWCOu^Zt8|IZv*|^cU}#=#_e(`E9Sy+~C)v z<4o_96s|Q<91nAag9LFG?|89Prgzs_j?9WLWg$SY)-PRW6WiT6N-GOzX>{I{XW@uG0;-I2w|6DB)l> zvAA#D{a75#;9-((BWpAwYRbb#wQawQFeUAg1WLFhY%%@(j@d}HgpO)^%(o_Lh6*`J zv(;?Z6rIV=-wdBhsQEi1z_4|T-!;V3zmOkr^#;nu zT)q6&F!TIs%{Np7qk*uh9#q-z#P_jeUtY!$`Jnj76yF!aY?{}G$r%4CSpc#(KAYMc z!Dc{W$$)yoRG&)9+Gx}z|0v7!P+#RtXUJ~y<_gPPa4DJ6JY>m!xvs(B)XowI&G8Ud z2mr6z!E}s?^_zt2_@lsNh+~sf z(TlZgw-ZWl&z~QNc=3EC0W65`pO251527dMy zF-}UXnGV^CrN-AuIkG1f6Yk!=(M07=o^_bd^!{;K&+Dy>}sKe(YiHR@!WaB$Bs(v>?%_2cosrSUd>>j6d z5J?K}NGY0zkLZF4NGw7jQl4nsoF4;NzE$>_d6;WFt{zY}1Vexq7a~6^@>HZ%t1Ymg zK;k(!q_(9b4`|R9!Jv{H5v4R_F<($8q5fT#Tw4Z{wcUqzm@+Foy!>Fc3zx#$aTJQA z>XTqICaN>!Neb<%a}`n!b127D-@XD%jwXuvw2wMed86W$c%AjM)t`iS4N8MwKSH|K zqJD^-pC>V*I!s)C?gZVuQ;Bch;@pvQ$o4~hU4_aUK@FPGOkgLgN0d`BI}u43nio{; zV4uDtw|iO)Mo&#&0w$`aP#6wN$7wXjhA${D1q3yT{LrUD1`DG$y zZ-%O7%hBkxooaxSYPDe{YN{BE##lLOWoH~^Y;#n5waWd*l7F`07yeEFr-~_6UTUOf z@yY-`JtPHoSSfH1<3;dd0Yk|-WN04bVZ$*pt5dO+K#*%@d8+iYJlvJz*)Q0QdaG|? zJ~Qy#jz$995OoFD|3Qf|3Mxe)7*nS5OZT)6;0iYrS!OxGf6!*jOgw_Tgk@aipSJ&Z zi7S;mT&8O8gieBxWwaon5qbY_Me}6=Bl0z%2DoNZIrR6g}7PQl8Pf4irP;dzvtK z(~Os*azns#M>@g2&la(;lm-gJ$)pv`9RG)_-ly`)0IuLAkFuQm9isdS_eug?)$0ma z|FfS<&~^`G|IT)G`!kguPhVB6vy(oxdWY1%TH#PA2RTX)>T*A%0)x8u3BsM8_K1Hx zRUZE0B8bN^V!R0RG}kUFu$9;zJ-`jc__+UAE7e)=7iOV<{9P&BgfG5EBiVkBG;elTrbefTa$nLXY; zgdeV=4HqGjE?zOc9I>>8zZ@CdO(|Y~RrY$rdQatz!htGv5^hxktm(6i`11QZv;p_b zbFbzT!2d_xeYf;n+7BVC6iOcuswGvy z{zL7#z}m(Yoy)Xn{7FjcStZM4Ybk-F4vdU?&)FoH+F4N&y2+5aP) z+5!aF*9#+-?Cu?;ai@j%{k^h6?5owS7Ue}2-&Z`v` zTp(VXCUgn9@^))Mk>XN0J1B-k4?ohB(y9OfMZcIQ^A;U5lTlOYotfz7+lD0R*#lBO zw30Vmf|;!33QT@Tz0CX=nR6&$%0Fh5EX*pF)ZtP{nXzTe$VSnmDY#E<5d+vGWgZ`s z@naJ$1Rl+jO&6w^eu7KhsS&v!gSB@wYq!^`0R}R>2_x+eTfkpi*w1yS)Z8mDN&Cc^3CRJI$}sZZ_X4aG@R46tUw z;gj)i7+MQ(@%D++y`|uq9+b+7{yD8?Qhl4&|7??GG{y;JMiRQN>2onTE^|0o1~Q*> zhA}xTG{0F6vh(WZErB{9H%%fnvl9q?$LAiMm2@`C?N1F{5+oZ9N#4JyP1*15xf!p0 zuNhq}t5_~UAxeZAz-nytlkN$3fd0nVmTaS&eS4mqn3kOQv!EAxnNf>#2&Vd!w6Dew zg2;KD5TV*hwi}^GM5lTi@s;>T7~+<^Ee_YKsKOWgE}$lK!*?XF_R` zIZ~32hq`3$pc5d<93qevat8e<1ZZrdf|p2S!Kdr^1_PH?57`CERPi*+|5;wj;#B*e z`=#+L*f%+5=GJG)Hu3fQ{sTL1_3=dQk0Qgp393I?r)Gy5ssyz6>0u^eldUQYQ!lHXuo|Yo8oRN9rElCs_umw;r0SLP`2y7pPxqf zB^h#bAdH2`Jx%I z0H%8I*Wurgi49e7bEoX^pk~*Puc)&K5oQ4pajio#nKh>h6hGzh^Euied7EJ-A*=9h zI<$8-t$bUnk7Rc`**b~MyQo^{5N#@8ZRmO}C~p+#b(>?Vo4Ak`d8Z0g<^7<%wpO-R zptb$Jg$%&MBGe+?aml>?>UCRzR-a?DqMH6tcD(Bp`WvYLYM0SbU*hz`15W7iB85R; z)p73^+)+_pZ?1jkIC?@M80uvx)79i*p}D(t-$Q1=TrqpPVi~pBfsSX4jOTiiNfA{B zsfdyl!~|AKG#-jyX0>&D+-qQy4GBDuNrb_ngZ#*L?YTu92f%~(+u zJy#Z`pdL$0n{p5|HVaNg#*3;sFq*vKEUd6(T*!1o%uN{BI*C`f9=ijAZQ_^zBv0N;DWa} z<}!$V7)n6)0~a%aYEW4iz4uy>GJJH@%pup2hNWKek2@Ef#eJPORfHX1-DS38vn@6+16@N+MYQ59 z6>L~Awmn`$t%#Vh0de|WD<6Zd@I5;AM9elrB5NA8i;QqklGY6ptfp&ZaEu1E@{Cqd zlf3qcXRQaM^dC8L&+K;VEQh6F8FhFNjnrp zNx!S9$Uf)XMJtc@m#}C9>Ij?`Jooz59l}nQoD9tIZ5a(Za)v~uO4iFzST+Mw%a>Ze zusDpCMrpB;Js8`Ld%idG^E{Jbgo+|OPS=FrzudNj&*#&g%!rCt@)c_KbZtp)8rm5r z0>M1Mx|UKJ!iU%YtrD+U=*T_?fJ1H>v&m`L84ilS$K)toXY7^%2l0Ko9%X)zd`)Jb z_t>>3vDn#^vs-NXJ8AKWPTJ-SQl2Da(%UL$9pc-VuT11q3-PCww^KMtSe0fU!UB&P zuCOw&d*QqQpBOcQXc)`Mqz8&dd<95fE4sr&iCY%0X5f<#j(LzUSr@ z+u9$l>WTvHO-F&+-XE*-PmHq-XfzXzIz6SC&JPi2 z7>Av37`!Ms^TwLA(y_`bbX^TUtL&0i{MOxzs~waJJ)s$vI68mhmh}J+#66O64%&Ug z09D-CVx{OBpacrT0*Yl~0UapA;n`e8qtWMv0=d~smp*0Eb`0rVGGFk`<1d(Q+~8@9 z@bEZe7yd6fI8_HR-};h3GdR^sYCxkbC>-rlx5m)*(8KKd2wE{R81Gt9$`-8Q#g%N8 z^h;PrfjC!| zAa}{BmkxUeWsVNPj-s87jjyqBnSZ9>8}l)R!}?mlVas%YO=~*WNThXy)kQSjqL)N? zk2nvaf-B5V!4#pYPH;5G4&^OQ&ep;fbCctOi46rTM&AY`bi z%MiE-%;<5ggh%jjGYw5--1TBIes7ojk+epJP=&+;9>GVgFhC9+Z)ZTW+YfH^ru6Ga zQgI=UqJLb!Yi_}XeAwD2&aP*iSxUh5W6zU^vF9|T@ml~xhp~Uu42QbDFEC<;s|$bI zWfIcCm$Pr6OU%mGTn=>xmkV+Y@$2Fa*dtVz-cLRm2^wWsTP|?Kv}fu4*VVDl2};N< zTa2QFM#)CeQE9)%xNW1PRWD4+MN_j`(`WEM$h2@HEb1$N{eJ*1ACf;9I|Op~yT3&y zMwscl`~mg_pwoskJ0*aE2hBRwSC0U_UHEDW1iRL>Md?O|!^hww|k~rnpGoR-)!eC=q+NkfU-usJ0cF zCDYI2rxF%G^ezyo>!7LuIPnhCnltw*$o-e7^ z1~AJ67Z{KaRx(lD{A3U_p#5AR)f#QYLA$v?!%VSwR-gvY3=g?L+kq0fUQST=bj#oN zTiVfaxNiEMt$){-?Pq(h==t8&{CC0CLWR<+rx8(Y;VaM>M^OG~!WWz@eV zb7N<(hU)XO%0jqJFUA;%8n*A!f75FDJV5nxXVVCWT9V=t*f)rrft&8mUqpDiPYlR|_wuZ!gl8-e0o2{B?ze$`!`fv2w5;N{2`8so*#cYIVt43NdKb)ah*fO&s zb-x>x6$F$Gb}YK8OG0STsI0?ULAIG+pb$fM0TGegIu+L3>fw^sRW%JW??KQ&72ox* z7ueTco^{=o0*uOE$TL0ob9dx?%M^vQbriQsB}g$gv_4+7{rUt6Th<8K|BLuekVjma zff(O5ef8Q;4Kq%0^v@A9&Q#BXtdUzIspM*n8l@D>qD+ZY&hZ)YEVoMmf^0wplnxEj zqJ}EeLMhf>7dX**ueiv}I^iC+p8}uTa*mgVd>N{P*<56woX&f+vGR=M-Zpm2SoCv{ z8GJ}QOiX5$c4pp}mb)%SQha;w!oI)xj0;b@O!>Q7~*{C<|;S{_sUAx1V; zrs!BRgm!Ld^v#R4#eaS{6;tiI30&Ry3bKddN@_fp+8r4Z&Z1h5U zYS?@B`^q*vTi!To=}r`myql4aAeHh%c!6z5T5pG4=TBjJ9+n~q&^u5QINv|os!A9! zm95W!iKR2<39^~cMl|1{1=O4=%GOGNIwgVjx!Jh&@kkBlLH5=*ah5Bz&NGn$EivMJ z^1XRBs54oMM;)C)THah+x*iz{rSEO~bY@l#n>yO3M~Cw#TcDx&tiijY*s+#SpyI({ zV}5^~R>}+dtjEM9*ZF!myM>5t)A3hI?{}9_k_%l?%OWGuF-+03vdA=UACtU&jAUgX zc2Hj53U@VYba{Tr(#p6#E%J<#eacMdkGpDMCi;xh&8loSRr9g+b!T%p_bX5fu<4p# z5u27^R0?EYHuo>nfKmeG;muKTb3$&V3IeQ?T7s@4d!9s$+g^djo^cFx@YK7`MxJrq zuJWos_`uN_CSEPzcQd~mY2wKZNSz#GBy1TnUU;X0Xl+oMp+}ceP2r=?;qsKX-J{{l z9vRxBNe+Y@kfO^viWp}$;>^$BFC_XWw(g4@Ol|XGXfY%VwjnKZUATR{6siB@9pF}93gG>P7|dy2 zQ*^tt?h$EVzn;71z6N03hOkJ7GcyO_gD32RaTZ15u0CL`$(nF-9zDbNF0KEy#+86C zK-_CT_8kc4zWT)_yz4!dmJw6=UpU|}r4gTmA$WG3>^+)BhMK3gknc*CMn#ZHqu=GE zGC<8En8a3D+qsi{M<;kaykdS{%A#ytBi)%AGkm@I-w@_{~ zP9q52ZXGmhJjbt=c>Mh@X80ex*Sa17(0_F3wZ1ePX|xlMSm%>+cxpc4tyTf$a5S7% zEKZC;=-6t0zW(E8@cJu4Yj!@PGaKi519R9q>wj^@q4zLeDaawO(w&u?Tj5ROAuPaA?5Tq~^O9OG27!3G=jM%g&ft~+ z^CbMCE+%|5T|PnZt`0_D7G5lfvO}k+#GL#uG=cz04G<^Le}G=s5m9w3jx+d~{v2Hw zqIb=o@3C1$SOEe?reTUDYW!-W;Ivk&0vIjVqxTOD{;f;XG3?Z=|J zmxz!)A13#AVlW>)Up_9dgZE{b%(jp#zbUZu?XZfw$$|NWulKv!5J#v!tRx_mSTwye z4S5#7L2#W~_K|6h*5^T?@&@eOK=Z=?tClevb%j?~P-K5B=-LMk9lrS-N!&LDCw#PW z(dgmk)F}l|o`9u92u+Q#0hA&hg~GRI8Tde&gOh9_fPZqlYv6TQ(B2u@YzwdBb1@}ndS>(vzuR0341O);6s`GirCu50jq9-} z2Ng{G3$Q#4_S1MxU7mKRulqq~oerd|scPg^XQmcM)~idDse}!@VBbxv68Nruy0Q0| z_k$%qBsU3TpIU%Fu|T{ngEQD9W^r2~tzuiiu_+(Zv##0|V4P32Tqq~=H7AJJIR zxRVxCtDA6Om)ZyYO({hImA#OZn)+5`O7yseGrH1)y@sO<^s4B$eUQp{w!A8sP#rBC zV{NQra~jy*Qmc*Zf0xRLq&;h!Sn9b*jzkhzTo4Mlmum@%YuVgdn+KkB;Z&7+^)-}g z^{t><+n$+YIv;427kUM~X!13*#U~uDZIRzF^PP-&_}?nqY9Cdjjlj?`2!0Sc-5vt z+4+DXFQV)N&3V;4lvF5K7M1^!QQTb8agBR2q4!J)Q#65OHp5nrTR=t7QtNy$(!;h! z$~rNaYy4`}&SBn0YJA3@(Yw`%ZY5^4!F)L`qUMuj1ks&XZ#mprX!%{nFp6ALWDw|~ ztSSM$N{VNxaEPPxFr&bR{ceUTnrEHJ=9Jj`NcB?mlb8K-6X!i%4X-Q(tvruadbnv= zCguL_M-sL_U%`y>$8J&Q({QWU>z#W-O3X>g4WX5Bbpk$`Q}Ja$z<53a>Mmx8l2P-zNmMli8v zBAB+FU(C>%YvBg6=Gk6{XTR)Gg!wzP{@p@o^p#(Uql{M*cCl>^vFX11jV9>CTk8F7 z9^y8_EmU9D(z{#w-j9x3mUn9ZW3fHnv9$XPe4zOJ=@+KiAZ%S89DWYkW=RNsqzrJ1 z?p(}}>(10eo-A}pF#7Pnu|7rzk%xJ`9BjHgG5erNFq&T`Y@Hm*^XR+F6^Vu0EJVNu zoiAxllGFAf`Dn`+=+!D}n|(Jgj@JJ6H|4Q!3*9KRmrVO)^z}cyO=Y=}oJn^C|3pP{ zsxg`o#DHm6>>*#!RI4lwS0q+B9GROZ@nkm5f7-TNdUp(rCXVzT zd2tVn{`d-plzy7~r-T~+Hgp~N(s7sfjzB9aeJ zk+)~~*V0y+n$;F@b^y|QO`S8Csv+O{-y&~c;yMMvWA8w z84_}#R1IByOY3R_FaD~IF4gGvb(!`OQmA2LQPZH<@j6IBGU*N)$ zUpr;jhM_L~AirB&c2nm|mO%GUWT@#1d3DEjU$seH`OB`!K+_NR1lQBxjU8hg$HC94 z-tRN2rT}KKe83bVqnMh?(`&vXESL74(X03eo z0|5UO3JJTjw`EZ$X}RO}+oxVv6&G6H-=y;kRchS@rsU#JuC(WZ43Us_C4C$8RSo!? zql~-sr%!@TR9_YVPW@rKzDd3KL9rA!A%RyR2f*gUAJ*wCW1z3TIK2Ar<{A{-%N!Jz zlq3HqEO|yv7rji%it@}>yVwf+vqEcMtrAvJI}wS0vf!U<@A8X_npMZHYq=;3 zUcXmr#%`=#wP~I3K+T7tX1kYoKUAR_^(vw1gXS7MqO|RYDpkR*7{4EKbfPT%9Lk8` z)S>m|m4TW`nDMQvK}J7nc|%Z2e_G+;RqYDTL$7#g&0R*nS;4`mtJ)PFS=H9lW1PGT z)SWX$gumq*YufDBrqR%wcH^tW_$oHudbc@Ltzt7f?l#b@^6*dtzvkrF zj2V{l-m+UJ`3+JL{tUXB=p8Pewyyqn9M|GkDZDit|AWKox1sU&% zMQQohJiLm3&4X2Z%U^jF{|5WT{gc5sq?U_1>2iJWV#l?^w&yJBylAgSzV+OsL7m^E z5d?|}+V(p#PvUzjK9UGLSC^l|qZC+XS%jVeLd za>o`;b+-ySvs9;jcFKpAH^vI(XF6XPo>nC^%y^kzuRt43lXN(_e|{uh{gvO=jDD#YTQG zv3wJK1ewahs>xD_+XD-ypk<4Jz4fCfRmQzUcn=b$fMH zuqMv5Mb|Kz@lU()-C}$<8$a~{lkRRdGv)yU&2FKaSxx&qI6G>7AP&I9s+y`uWw63A zIT+U3KCoN8$;)%$P2Pgny!4)FfHf2@*VJ_?k(+oZbWpUy%AnQsFEJw{;CyM|`ON5qWoB5-=N2wnW`xyz zFRwQ;BXp-13-y*|e!2g>OTCq))uLGz7(-v^MQi9I8MXSWUa?B(FGMq1qt*b;va8rf zE%$q}lrdC>62 zGWL$f|1yn#J)C~Xi*9$;gV;I-aIlLL>*btShs1h%l1u*se@vwp_>WT__LFzDQrW5! zPiY?J?DHPM*>2N|*e~gW(wDq=P&$P(+K^Emp!KZWh_u+@&N_dzleeP7pk4O4o&JYH z|3jv~?-4UPK4e4VR~Y{IP@_X#u-S_P>9-%VPO`^N4Fx0g8GGS7q?ZRGbW>N1A_E@x zlPB-*e2BM)6~1XcVC!8N3|CE@>S#Ozb;g%=^bd9A!Rw)}1+RLk>k|gphuEw(-a}oD z!E1R}3|`}&(7`M8cNs*oQyj*dy#0_ewCvqxLejoRgw2 zZ0~zQpWgO;Qct8?{~_$nNA2UDbeu@ntN$0Ngd40OY^$bv{w>gut3w#yRvOeI)N=s< z54&$g?Z5S~dw-t5ZFltnZ4rawi^s$lkFhUaeUg3gxcDONDP7!#6w0Z9g2ip`r*uT0 z4i${(%`M^!b41?;6)g1ouhbE}Ozk(Q1Yye^8#Q&8O600L_TPJLY^}~~@N4*wo#!(W z{?9PAhgX`N_Zg1=^(&43@Jv@@Xn(dVhW5gzb!g90w~becpglcBSGBapKD4Lq8sMY6 zHPU_qQ}n>2@@c;;W<#nRbKvpN)9C1Y46s#g>`@6@gHwp)7$6J3TiD?ly{X);I(FT& zaLB=j3w%>K^BJxg6GgH*#iP=wnJ0?jtk=PD#GBy#2J6VcaBvy5ba=7 z|1`<(5b_Rd(Qu$9QXZ}fVQFAi+xD%U$PI_Na;8Kk5#4auu^{ripK>bkSJetHjep+o z?G1;?yu!wBISdoss^@W-8V=JfoHy+SKY5zhq)_IR$fS#^qLrh=Rg-AWVQW6{(!0o` zmr^LR7KwAY{HE22X$N5bQNf}>q|yZ0vfTiS_YI!z;L540A$qz)TY-o6t`@9McX+(O zLn~i2uu3jvY6}L#^g@S>554rwi+;+)uYw2ZAN=F!_x)WN{DXhA{pEl7DYLdNN-uNu zU*S))b%J49%e4$@!Tb&`;J%JaHn)IErw@yLCg6;A19oj>AM zFU?qkb9k6qI#-Q-!$bpY9K#}|TuNofNf)#W*G{(N{r&RA5J}P@D zgDnXd_|CN0d&0OoJtk1HKqL;3UW_+67!)LHMD`H z8XFkzokviwn9O=2UIWC$)|86ArA31ktQKkh!49_00$T^Lb6+*U)&UM>PO0b%zH}A- z^uA{M6}H=4{L7Y}#+R@o0As3$QpszeFm0rdvUZB8kV;Uv8Wc94mO+J@Xk<<$qDde? z?nNVg)S}tl6)wtNOGMd!N5?$=HU9qw{NLF)m4Qg*u-b5G?7_FmyY(jl<>KBm%jL?6 zR87LxfM}Kr4}~6k4I>n)sjDeRA_l4rNW3Z-j#SmwAPrBIb|Q_5N00>frLQyGH_HWI zkf)WsQ#Gk3SV1>%+T-i|^D*fDYTGTR8QSH&NCRqgf{f%PT5__K_t!Hx#Zd4@M&KzA=14bPV5472n*E9iNTv;pZd@Z6Dvk>{MZ z{FD=!Tn`02=y>WaKSrNCa@~=H@q;o$!5|)A(345YqdfG%TNrohCRA6Iv? zk!ZBxnj|#gMQ{5FN!&|cNu3+1{th%aNQMq6EE-x82nVJ_s3{4p zHVc_zW<+o*-Oq{btte47iI(u6y49&uw zVjl9o+$G`Nt`hww$E73emH*TsZ$8)1d6P)CL-qoG@e&DC$7v0qz_cqt|@lm7Ja~>3a?c<|yANZ+f&%Qod_yNWP46ohyNJafs zqQY!Oj@-GLx?Cl4yPLU%`=snNmB=lJt!}5L(etK>;$KCrS|!v*ppxJ29s8enjc(E6 z=pI|Ndu-wE(fdR61??VNxO-g4N%rosg}cX8PUrb|3wMLX9~xo5g+qkoqkLPEFxTRb z0&;EPmY0P^q2j1e36g(GK(f3lR2dAH)>TD=dKW5K9}3h|6bB{*t4%h|$wsFyAfgjo zIpstrJiFXW{{YU2Ozw07I>wJkspisjLi_2R>FkZdnr|D6&Rs~-{RN$Xfc|o$L60k; z%o-Ym^dA56Q)cDliiwOzei`&sm^m962=`Ml2t0_tu);eVd3*4eS%@-mUh=>uTbKIX zKMvZmHif|WM1(Ll(Lp=)5~A9D>c>j9 zNy%m#WIKG~r|bvK-&xChEr#hC$jAAM^wf6sa9WC(1`&&9rKZkRiCm@0NBYgLiK|aC ztY7(wx_bF1dKi|sE(XBepQ@`je5$T)aEt9Gm9S!A*rlo6Q$-88is3h1T@egdT@}>j zzv116E>e+rxrm38A`B-lw_&o^;iL#bwHqnocPfJ81E2XR$JziB6=4~)rQ|89E+e1yw@-<$|V>8qv1e#)I{0rv7;?vv?V~>_j5nx)ZoJ| zyt$zg0RA%)%j?47U`*n8uD=L_@9 zK+OpN@$Tt3-hA;^b^8W3ooC$$Q5P)eC zym1YZ?BO*D;q{My8$Mzs$~5etU?s}+{>o38SMiBIJ`qM$D7N-1zqqqe-C1e4v;Wt> zrz*f~v8X1TN!&QN5lTB-yomGfgg~004ulWFuwS; zK`pfHZZH1=cgV44zY=Bl0Pk;ioOrlAo#dNfGyY%)9d;RAJB+S<*B1LqAq=}4kz}kC z!dMYp_8$W;D`4r9Pp&Xl0o1FUBJhBF{(txv{hUN1s4(r#|6sK(J(g3l|BEFCMd6HB z{~L3DIC<)XV7M??9tcJ1ssquWNub3+ELL>!%aTD`}By#jwsq?x}C5(@y1>bA%YcYGVQR=6`!Top!}!{+WoT zBlDxT(`ipk=29Z+$}16P7dF<@L$SX?UoyCqbrX{^Lo9K*a>9X{3Zf;hK96^yJHG|3 zy6{!O;e|*$d5)L9`PLw}#D(3Px_JqElQL^3(^40R`)Ky1`jzTov zvCTHaa!U8@oD!^1jmz*i&NhgSI=r`smT&h{=AS47o$pQjyOMgKuVlCTDKkLFTsfdK z-P7(TPsST5_&WFrNO<_W*zqkCo|vC!f@?m~4l>UInTOJ4-!mkcRXOF~NYP~gClv2s z6`val}|@SmAnesE1iDT?)hp`Uj6DKnB!{qe^;JMl+> z{ITQ*KV^#N3g`@EyMFYOx4paI&QScB6KEh#6Yl-dkJp(7!bxj>#Mb-xQ9cU2kwRz$ zeE25o+)+OI@<%^qjUDBq5kL7U>#|WkI(8Qv)oIcwpXrmvKZ*_4;;dPtd^CNRpR#Tr z<)fxuMubmHUgIU|dG9D6y}Zj$sSl6xQH$Eb`B>IlqkQxg%6&A-N2x#i$-QNikM=0& z3^f2~IO@#aJBqI&Lnlk~id90V3_A?badzKOc;c|y@P%^Zg}(;Z@XYH>%rD?cpui;7 zew0BFwHI2AUkE?v{LC#Yjn64hAcY%`B%%9mRzc&l!Zo}7)C-Nz$8p2a+V0PG?!5^)iG)y0Wpezy9$9i^ zF(yz`Q9DJ`ds8~;iJl{U0i|Os&LbC}UQ=cSq-)R^pHe!W=GYz(gpvG&E1go&BhucK z{u9Bn34OtvlrDXtFezQ)|JtNWYsUY*N$JG#|Ieg!iT|c!#0jfaBKl(Ef34=V zsf3Or{v9LlejR^sVO3zZPsVWq(Uq`kcsj6)ZDDmAQtV;13kmFD=^B?Lr^j`THOc8Q zgCTc~UCHVHzp%?USvB-L5W5pP#j^VwQta#wPhl0a(?n9@*=hlIF(GZj(20$KLuH>V!a+2|eoXH0jv1_dwc ztlgneLmbaIxH~-8S-V4fXYCFrcG2$8s=8)(IIoL#hwHm&ckmo9+_F?6)ML9tp5_&+ zgwFoE!+~6Susy=?eCPy0nYc%+>VgKa_lSX>baKSZQY6@VL?>^2%uMknjG3k0gNT`f zBbaXzYt1|m!3|ww30`3pll#)I?0Z7 zhRbEmA~X1D*MAQOSm_QkT<8zS{UJSeFDl!eamaAaB9gHN8S>P{m_{Tc>j4M&n_VILBs^Tf_ApYRynaMk?DE^)toN8|9=W=p!H|^zl>t)ir(U|l)`L&l1^lLAV z`n8vrsjk_}7eW_{zPJ6_%L}ul(aKb!2gmmEO3j<45<2_u<%weC;DW!v3?<4DbcuU+ z$8^M#y?5V)6#I)a!TSWWQiVS1Ndwu%ZaBi{~kk-bAN;tqY<(dm?9J0T@n z#NE2)Xk*q|2TKXkMX8=-6YKFHU5fDiu_#gwhhc z_7WNCqbFd_d*>04Ydz^e_GO&_`-a_yGwk+168j&q z{Uc6}v;Ps>aOD{G?|1*+s^q-oKiAJ9$(+s}iP=qw&?0*K| zaTg!XzhL-~p>pZ4siyf^;JH!Z-Ys zp_y4KVWmFK3r?_+x+%fZV1%!)8Ti`Iw&T3N1LM5E18`0+Ky0Z#-QOW+v6ogB80~XE z-=8IKlUA@&C8J%=nI8hJBaG~L&JE%TXH5i;l=$XVFd7U;Oh9FW?Z93XU@tPTQ+gX9 zUhIJVvj)kFY?QpoS~5o^BU;Ou6}@8v+m7g>Kvfji2CF931gcE{O+)NJHVcr=3}k}| zWHVO}AbDi`k=kKY8fCnd!fo@Zizw^pQEBAvlTJNP7?nn)eSjs|@vPbkEeDp|ir~ax z`08*VlyAZ*DKucYQok6PMrCTK4Tg4)Orxu~etRc2xN4i?o%EntPsT8#zFZ<&CIP#Q z1NN>yhQF63AqXh-UxR(r_*M!*6_>ZNb`ADXE11zaKH34xsgDoyQHxC{b;+%VQ~O&f zltDzBl8M}}-g-DSDvjkTk-3^5fH)(k0#8N_j3wt8SO*i)HLjfUXg$$2%m3Deviqjv zk$tqVDo`B^6J7Ilx|jZ~iC5Q#Clg(RPTiwlI^_f^DxiYD>+!d7b?X{jY<#|-!TK8X z25npO-}&Uah9B_BJuUfnW_fK*4c+TPfm2UQ$KTbpH6^tXn#XhQM5KiSSEHS2j%!`2 zm!3cpX5u+K5nt3F>*?A^v?eemNR2LFd$fN#@%2Ey!~a%L%BG%$fa>vB_GXCm)3 zZNd2}Gs^=t<-uy2cJ3oxsQ>BdM4pVd@mUo~p5a3JrAYTC7l}g$sLU*b;$OV zpcpXT#~ANNV(Qt4`>5gV6haeRkCif-bCud|Rf*<&mxm>HbGJXxE=~5FEA~)!Cbk1_ zbo&DhKLdlCceIZNDU3FiP~%<0@uPio1U=>&of8+Xw;eMM5$ZO@KLkk zxes+7ex{FRpP5csN1o}U)zFFQ4TT&d*}0HIbZRm}YtRTm=;9FDgvY-`n_Rs%dg+e? zFqI+VHn|XT1MzoFP@`)TH`z3#Nq9-Z(gEp+lss~6;%3Jm`S^olD6^bC;-FbSAYI0* zk2qudfOK*fj`q(b##bz^(DfOX!1>7gNhA) z1Bs}Bhy4Pac>Z~y0koj7pmR(w%+lA@`Ave65$koNk&#Jk;54)1wDS!Ec+Y92%as$X zk4A`Ax@J%J&;tY0<>#x~-j-Ljo(36O*n|ExFx~zYzOr}ZZWCT}O6PMLvVgdR%n(8oig@FxbRhmz5kmf}w!p+}NO_4Ck0gYARo z1u6DnrQ!w>6e+%%f=cM zFaccaIyN_i>-YIIxz$U?Uj=m9zr2)$uZmsU_mghaVMsdVOeDH?$S2+Cydg#~UdyqV zdgW*zrM;WNGs-mdvY$r#==ve)l=bUqAH6&zoqDE}_-OYKZddh2VX!(F4aWKcaP#|F z^yTl%Ow@xew4>kZG63gO*3(^V@bBE@r6i@+o6df^3v~dQt*)J)bfa;F>6B9utR~v( zO8T@L-Ct-Bl6=anqHlTKu(2?mJcHg%!99)dIi=swbn>(yPu`T3& z_?wqnhiO>jkCbzNJFM$#!=upO2M--wXja>Bk>OBV_2}(qovYR3kHGJ!J`@g4tg5#x z7&T{%OJHI8&0d;iSim0-#aK9Cl-T-$hJJj+7-s1b!#v{M9x;DxvIqQrBPR2+ZuADA zYQ*&UtQ#FsWJL_}t%y0VD4je_@7W?|IZ`s-Pa%$&b3W@vrwm67qy|cS)bhS9VwRQo zsA70JWj$Krqeq9QQ_rVMeDo!2JNGMCV@I=dWpFYWhAuIfx^i&cpztv-opYW6bP86q zkKRq^X0gX^M| zcqPrO(WohJyaQb1=}lhx&a5$#sK&w7!ElsY$!IL2k|tYaL88Mx+2keP2&1tM#ju)onMDPsFC9Me(j}&KJ$-S;KxLqMBDPQ2 zxy7~Rlg%1hY&A_GDxJUFOVdZBlN)>EKAJg8C2-@v*s!qbzM!TWwHvjTU2fr=LiB+P z)i#@0t)aN^gCOE`rC zmNy0ZdU1r4`Sqsr@si(yVk4vu{Ce?;xq-T9ZFEZLiu~T>I{8*FxfaI+KpdC-zR_q< zC@HhN2`VE&e zqYR%+!peU?tKRZJd1Y{DO*A~!G|+4rLf4Wd_1?dDY0@ZN9N>?eP`$OJ?qZvrd7di9 zA@D_ka19QDal62kScS()R!E9d7*bsu3683bR!y8rTW|1Eyp;=Tg7u+bc{ErtG8l;j zCIwlqb6id51S(`0zy$?V<%?4(3^IkP+L~d3s_I|`t96T0lWIm^6%0?TuDzO-{BcT! z!SdQEp|G6hqPn`UAUh^b)eI)PT2`FeIf05H*2kRK>`8G-7t}=RLO9_K#wnj1r);Zk zOz^7O$;vqAK#D`_Lsel_Jmo-&7t~CyslB>}apuM;4zI2bP6||;=g_QnYMh!@GGs94 z=f&j@E-w#8B89=4Dut9Er-Fv42uh5h*$WbBgr|mNx@MK$aVkUkMT{}QNHknkjsRi3 z0daaI;o4|zd2O}c;AO2><70MsO{5O54dHWurNQ#La8*>o`?W;6(O|eHpuGePS`#S+ zLZNVNC|nhY28RWzt0x4?C$sVg@yZuOg5e>2CaF%_6t6R?*7?*4cE5~Q8(9^JRMkux zS|1J8@M8|>eH*W53rnZ&ws>7zR9;Y1I{{ngz=UcE`0erP7u1A<<+YP)s;&uEjKUc% zAhgHpIlB+e;_D(~f{{>dO+@Pc?zhU|>FkAp>N-*XK3-j;wWM^!@QPqf)M(Tl@tQV{ zV}e)KnLTP}yy~dh!J$xfRXIQXGvbR$?uu6)I;6-^-_H&ezDPC{Z-($_>lltT4-0k9 zbPUCuiOH#e^OEgx&v~UqgM0Vy!!h zrB{q89el+_!^aj~F=j}gEAk8aUr|~#m~eX9(O#KDvv=Xp(kq4*4!vT?IYUf!$F$&3 zA3U^_E4FJp#-ljxg1-G6+I{y6C8q9zzWo@xWB7=x9xTrtG7`{@0*f6j7FPm*v;-Oh z?ztzf1UEnazdw)8Y3LA3B(6*hAvSg6?<@zDb|>GozjFI+^V@#2M8dYe{WpsxqIJjb zDiup={{Bi09b!5@#+uQ{OZHdP^v8Y*KJ>c_YLFNfAKzcW1Yl!|KC-`}4zTc!7Bm90 zd4C0CY$hnWVt++Vn?s5Q4=Ea4katDN=;Dj>d*}7FccSHY{dRmGu!4uec0gQ;&TQM8F5AFkkSjeJ2tG?PuJhSm$=*^xra-}_(hjM zm^Jcv+`qn)-5a*-e9oS+zigRvX6`S$Hs{DloxXxzU77z&7n84fM1ON_?}#X*@?9rM+cK)t#47c&ixna`eYTQQ7` zOj7MMyGN%4>Mw`{tAmkARk(?6dVf~ zD2&4Q1^80VSpNJWSL5(Iw1(DHjGj0&5U!TD&w<6!cA~=@hoH5ljj*C)2?Rx3w*B@fgS2D$8#Hz);%(k6 zV3D*RCH@%=!_Clhw7*rEx!=aaxCm=HmYMOICAC#GQPFfnf~AQYQ8+FiTs^5aTotXH zBHE6IHnj&AmyBYwjz%%FONWf*+&|deY3A~FtRh(9cs~Vf9evYO*XwNBbD(#8!fH#x zL4IJ%-*bRkc0x6)gF(k}oMD+X9IR7@xWaLoVd@U4t({yI4=fF71~UW z^9xhO7OQw90RLBM0mp{GERdkJPzR%MoHm#xM(E>gR&ks*m?}8UAB(S&;xNZNCK2wZ zC-kLb!#=!bVlCHx{XWNX$E<(okfPCb(O{U{A6s{+x?%Tz)QzD|h65g2xc-~>I0{6{ zBUd?qu+p!Nj>;XS1dAkGad5<$WMPqrEXM#KS$zxgIqOKf3&s`n?VEo(ryp%kujlj; z_~OKA{rVS%0~4cpd8~HVYYs%A$YewXFA8{MaSREXTUa!d^PDBom50@arv#!G)EFaB zP)0#VCHR_ZT?A9_=pm&g@fG1@8@Xdj2P2*cs}6@HLxC5_rb>&`h%s)Ij2Kdye^q{8 zw)2^z9#dyb>EM#mQr7r9u?CxBg)jD_P{AH%l`kDCqiQ8kxFei5C+6T*N7Wb=M+H!+ z3I_+*OpVcU%;z$3SsKG9o?CTQu!b!<;?k@Rqe#_6TM^cEO#ke>i>jjd@&Te0pWDCQ zF{j%#5Ui}>81(HLee(-M!!e(il$OTWnB^G1t^D5o`%q|(mu9DR`TyAa^YEssEp8lN zr)guy{5hP-fNGC;|+1qQaoMJ zzd0d47almErs0l;b4+DsBq$R{JY~=Tfv9wvk)Tr?@znnP1fpU+BSAIg7BI`q1fi^E z>Bs5UCz#%3!Iw@T|BQt6c++BxHKrO7n%myB)Mj~OT(I0292L&Nh;1CqL>v(@jKK5+ z)C4aR*JPh(Odsv9Fk*h}6MZOy>gpbS@6cPb5{R0Z#i1Ba*pa|ocI$iVx>=j1YHw{q zSNhMC%fn+zN`^;c{%XWWXdafPSt*K0yTpH`F(ed<#wJunjA&IT;A7Ofx`ZN^+ZD%n zYl}+;qciefDF+~l?Fj}#9R9_pG|?4;c|tHG8rmXt4W~)YWfjnRh{Q72moTZX&Z&rc zZNW_*a0pF9gLY!b{$~G2R5La;1eBSTShpHpU?mS ze{5*@R3jLpJ#C#9i6sW_04g!~m=jRiNmpXT_$-nIr=3iUzymChXeUBPAq5ZV^lOy*TMwyr5a!nhk~rb&k(MhELutCIkwIN{&~P!gOMC?&B_#5`I7 zjI?x1)9A(SiwQ<;jM8jHsl~kHyY`kW(yT~kKUn1T`LIHnW-FLJo^i2Qnt&ULdeP%@ zNG^7;L_u1jAHAbl=!-RZ6p6B&E3Lwl-2&7Jor9EJH%L+B8(RI2iLTZ2Zh z=Rr=K0I}+Dh%uiL6-0p+J{($=H#&0Yn8DHN!DEIE96)Kd6Ka<3B?n~EFQy{ zC;lU5$tnLXxQNnhC;bhk*jjszH{zS-jTj|j7(!___u=w*jc8?gbzHUD!>Q);M~w0q zTuM0JMRy3zaA$I&W<~(tT8Kraqw*Qim^{IsW^0zk*3MVW^M&V@705P0XmJG+U{Z_U>>SG^08kn?6)*@kq1LaiZouIJ=F{ zPlDo>x-`ij^cl6222q;Lw`wdG@rI|{=Hc*v)G&!#0BI%eNw=ma8f}v3voLr=L@lyV zn(aoJA$waJxkELvAvG06p-_O*Yz+&95jNDe!7?I|lB$}R4~yDqw#HI{h8-OVMHUN0 zYl2a4g<(vohOrK3(^w$Nci~Z*t?F<%vY(c&XZTFRS>^@OY}r=|QO=#!#%QrXG!per z;X(@^jz=sIJp}{eG+XWAP>e`qj5kmbic}lE;X}_cOStKm2HdOcsO)xHz1~~(`-d|39tD! zt4pE-OQ#dfm}vG@p0Lc*t{pjC9@WXzTQgGc!9+OfKxEfKQPTe^wNw4h46gYF>r*M8 zMuuGF)xN$MQ2PUht1{FtzyB%u15S7KiB{plMpwkB^p^Q46s43<8>M8bl zXfre|7hA5ldb;rICe8g}hWI&C>$Xw;oTXj&Sq3o;Pwx@_fYB?Trv1!n>Myd zEUz-kr(EC-_|D)TI3df&BqP=#^^->oECXh~vfyA#vD!LWs1R zlQc?+P7jv%^~OTg{_=cJZ=Sa_BJnc}<{^bp6n&ZtyY|Dei0hh(*ewd}2Wq*sMCW|j z#*WNQZY$QFNs4O|6?SZIB3BtL=0nUZ@6f91kdL(O$yzR-+hgPBR)>5BQO1XDn3Wi@ zVTRWi@COajW~9WK#fYwWxIHop7te9uKjoMnE%yd+pe;z+f;cj}-NF=f5G-@hm{#l+ zClR^)%#z^LabY7!+U_K)LY$yVo_E5L*3A^3Q~i~}5Dr-nA2oJDF=_iz#6hB8)=LNf zGwxqeA?FwvNjpQ!^$SzwhDgJgo+!m1>RS=bM=bzOevuKWCaqP|a;pUdb#DZJAJM3fGbnD}!(?k?=R5yCuN1#;P0(MPe~vdgHWVzwp5YMzZ>2B` ze(>0oEyFiz1PneU*N?O}WJMYk9UF=buk}Y`q+McZBsfnfj6&$F<%YenD(O^lvbkYc zgkUUJg^#*m80q5Vm_!6q7&5^uGE;wEjlmRsTPM|9Mt49czW z_ZPoh714e)V(sP(a)qY!qY((aT@}&(G*ZI)(@4NrvZzT6Enc)A5s-3$&c7H5Yj460 zWkWqw%asKy%?*2Ut4%4GX~{o7=&vR1u?#IYYLpY*I3%5NgXT|HkeX? zPw-b8(Xdh8OJ*+8a$`}VgXm&^pg^TY} zC@Kwml+qCGbf$JXsC{TsJDq*h9iuaB8#Cx9rE;5~VN}$lRrHgUW{sA|PFVw?rWH9T zFXoL@8Zi=(spFLiY6KsHuD$i%3|cqddJb}i*5)AO3{9MbJdmkV>H1WH^tdF*b7YoI z#}wm$b<304)F6{^8MJ5bli4(@*g>8HPiE66z@biqvvgD5Qd4fv`B^&kEm4x4Wi1@k z_+A`hm>CLxRtnpRs8+*~zgh%8#U&2PDr)A=SmWtgI@NR1ZCN_GHi)FFaWLt@ESUK{UI_nZwFMgMwm^e&T`zD@UW90YhAXfK zTp%v%Ss>a3q22<~efBg+oW!TBfTXujK3Thkj268B=jowdsZ3PJP<0{nfvO9DMxEZy z(oNyAe^K`D!=FlAtd)F`LDc8ka6h8OEFR9U7a$%)iSL#SJN}zarUS zl7rc8 zeE1P4%b_OxxUwvVzPYe%F>%*#%2u>jsSk>83K`#I8GBqL9PmvIgR;OyqKw|uEH3z_ zkkD<=3jTHIA_rv_ZAcYG)5NgI#me*QqGbEt4)Xk! zrPJ_B9OOyvtkXvP$n31spO=83qq9z}GS?d9>fKqV;H3`ooZ4BZrI&W7_JokBaLZn+ z;(LYSds*?5FLO}dG@`xQ#^*9<>}5jny|QS~AZdG9H8|&f@hj_UNpG=yvUC2%iFaJ4 zoHHyF6*6r88-#~7UIrR&dH@d zA!Jf)G$})d*-2bCXD6Xv)NaH90h z7x4qW`2~K!HxEo~TSz5_F~tt=waRpF37 z(E`41=&8$bK-Y+oCQwwvB1Csb=t0i8_Hx`qK@T$c$IEg0_%rFj3mJCXBsd%zm+)tz zfjkDDT2)F6{w8jde@dc3o4Dp2^lQpEFiXPpjAa**f!&N{hD9n@=GXPv0m z7x=%&w$3`8Q!2xPs1Yl{n%49o{+LOk$s`4vtVz=H#29Q1Upy5pRlUv{795}UptVVe zEMG2#l933l<&+*?2if;@*6DUhaGrdEY!S5ktFunEUI%$nyXf?~SCtgTN@(s)Ny7V; za>#nfl;{vkw8@J@>^wR+#AR}PnS=5~Fn=fs!^I2Al+1^?XyMQONIS%0lDmSgx81Zc zgYGDk@?X!l4>guaE2$wG!kpoeh$&9fpH{J6wXuJ{%A`7#!+F9)uiDTz7vU9_p;wvo z-f|_7cqi)At&46Fbp53Wd-`_K=~bVDJpbyV(-%IKZH5?kQm4W$x`|#F&*tr-Q;Ojr zkH3pf9z$ijPi2dA(M|Ni117gSyXaJ6VlPM;vx&PY7;AyZn%VqWD)pGtMW;oE@>tp^ zJQg7ud+aJC*&l{RQlwV-MEO043n54e_p`U(e1zP=Mm{%gFTeCiqTf=UO8 z5BxsC^jB0mC~L7a(`xy2`>_XSZYOyD@+#>$b2m23PC$}9zfk6vUs#!p73w5BdP)k~Xa+lV|LbkS)JrzP`X z(CdRSF>9x8xl;Blq)eWM(-vy>&%5aKtF$voIIMjHNbc#P({d(>@QYh^&@>4YYKjoD zEH?{iITz5^lO2?Yn<|Kwv*c-0%qqFO6RPB$Q&bTyXJt_qZ*o~YKSf&MO-}f8ii0wn zU{8kX~nlfrCHQrewB zM5C4WjfV3*UhSYfF%>l$n(Q#ijb?K0Kqy$rvy5Yao)LsPG!|)Nf(}aChZ_5NDuMD? zBI=cutrPXovvqnRi1LNOSsXdHp@b>2S%n&__2C|q3Db!z&^XBB)IttA(G+4Fgs`6& za!^{2Y@LcEXTJaE)NGx~Lk{wsnXS{!AqUw!Ob~AC4GIe)#6))7F$`1>%u~dt*phv zVWq`Z7L>A_GV3t;WG%ko#P3)!pp zWt75rKQV%LiUDmText1n!=)!j9F+EDwoVP7rxF;(Ab(0%o$iRhQ{c>Us?LE@DvkdAL1x6@11ha!!U!{jNjLb`;N}GqdgvebU zFub@Lg)!+<#+n;tO&(#Ct$4v)ClPN{KFW-H*Z;2wru0A$fF6nrU+ z_y(?4MvE}L26t(7)v5GL(PI3`<<4fsY$n=dp4KvrT9TgSt=ShEIHS z_^fFT>KUQCnI~mkEj^%4KG_<#F^lqPXxcqpb!w2t$UvxgT34N3Ly{fs=MiZ_$wK*r z zYd%RQ`%SeDI?)}aIY|T8XHfPsel=)LQq%eja{hu>R4baxhA8Mg-9aaEq&k<4KWGd9 zTt(+{dAY*VWy9Spt>J(*&~PuAE*tK3(`CcmB#VBfe4?E&8}7Z+Q625WyXy3pOed)o z48ml`ch$*xrGq?|b=B!Y{HW}z)0I~`s7Iu$PK)s4s;)XUT?tyK+G6@tO;#&2QI^$Q zt1zlU3q`Av-rJZ#`c=xqRtfuY>5DN#=~WKOYLIe0DW4+JdM^o+;GbOOAj(=PQ*O@` zjNAr&z_ik@hWFguRi}clL{%CEw^-LzC-QvNRj192@LgA(!V*DzT)C@GM_v=x3iYa2 zeiTDRrA%XjQk$R@(I)o$(bpi_mqW!(Jm8xFkQyp(;?jQo8Z}fDC~~N{i3O+3hObhk zrkh#QpCAsa(kvytJW&df^*TWfGFOfW&%%DSWt#Fqm*;T%*g*M*G;? z=3J{vW}gtsZO~u*YvZ-3%>C#F8yfwqud~j5{*}~b?(?rC^fsbJE0m≪rZ?!*vcy zv*+j(-ehgjI4o+A$TukDf;tD;kIT_%SZ5*kMe+&Pwlsw!jo37wsa#=Kfvhd`5RtC+ z@%ba9>FDyL&2?x&ZeJ)y$0B~ewGNH`rc`qK=mY@ru7|#n2%P}T8?HxlPMh(0dFlf? zk8wKXBlYMT(9#UZ(dkoZ-9gzx0I)w1{o$}2o$k89L7uTWI`x}@JHv8xnu#BQ9G#w? z;h?nZb95@)oJwg#bRII@ouktSGaTeukfT%U3^-@n<1pOjRHi%-NNaL*>UJa2P#yh2 ze=NViG(e+ih<_7Nj#yO~Uh9vUDXttLsm2&*X?f*XKBTkIHOvHAQ1$edSFFluWtFMP z%zPU0hx5vbjvluqgO=ZjUWVxC{0$kDc9U|dqhT#k+mFZ3yqg@9*$At`(#HTa-U=wfp8p^1vY(ec2L@e9G$Yaq!N&` zA@lYeoq{(z$n#^4PIuqzpdNqZ==2hB3-0Mkg@#4eRja)*oJU39Fe8t087)c!q~sQri$0uHJ@xFG*j~ zZD~XQ^DPxh1Vg+9*@&Ac6T^Ke@5%#I>X$0aCn7L<6}WFnSj?O4LyCZ%bWPgn|n z#C#5;5VKM&Ty0oEpAR}>l`Ijn;({dT=)BW1U=R|V4p5>5m*0tseYmwHmY`z;#LSs2 z#l1dJiWw=e!}VYpBbt$dX82KX(t&N-ZrtFJaZ!JBSAx^)2xgEU|DkNb(5CCssMIaP_{|152jMOt2%`?B64#6Ib62l4#zHYr@Id_qWb5pK-g7?xrY?ieXSk$eIT zx_DmdqSIWUiGuCO{_MD>4h1JA=<=ophxz($(J6>M9Lcjquh7vRj=yss!ALP^Y|98n z>pR=ZsQ8~49abH=FBl z-F@|J2YImW-ijYsdhd3RL#)0R-=o^B!_*`;g%nZ=*V-<)(P9wqi#6KcCS}+mpO6E8|6mNR0OpuOac^oX1uW8iz9duxH^B;B;1Rc)3=b2MAI9H7SJA6)Llg)lc}5j(C!Y!NA=lC7bwjPM})q68Ti{ez5+RvEDsxMYzG@~7elg8T>Y1H!(w zND502K-i2>ln=G~{NnPwt~fhu0S@y=#CHUcpji^I5;Y3DA(4=T1Z+nu8a0ga6bS%u zH%P!>80W^iQcZN+c{EaC2?$>u0$82=*ThnY#9||0c%z0d0io_XFb(^C=LfM}8>^R` zCG7W^_y986xJkvs#*H0)F+&(_+@uSOO$dtyoC}K?gY_9VX>d`|@Ug=fVh6KvlZF-# zA3Q-^wOyAiK*KK@I+{`HvIU^z;;};oK$4wMe7=G(nUdkd&zn>-d;%k}Fbo-bzF^L| z7^0Z7IFrtwFlx;3N#n+h5*e=rH*Atpo58SAim3rOE^P2dA%-$VK#Tlg0bz=Ol!VHs zFpkY67b*$`0t{mfi@dR_5(6I=i-p7|9r^+l;L-kQEHUXcB;icmFn@$k?V>bUX#fX@ z!$#0&!kIka!)wa}HBtXm!%SlNkTk)l4u>M%$aFJ_l|fR;^y+{=I7KB0he5(HW2(R0 zq~%f~Vc;MvfW@r3I#FK803#SN1vbSkF=91-Q(RNp5+ha_2b+=(t@1_^ImadC{q~Z6 zT+UIEh*4P+@G480lH!UBiQTZ@6sHJBRU!HzTY4y(KRgl{6`acF*3)dQX~o=^U?o$N zsldlw{-Gdl?nrNsAdbwY*|IxGGf&i}*<2kY^T}l&?pNp_acsy8gwkw<9U!RW;kD(4 z;bS^QZPBo;VdyllD3wWA_hj8<5_Z_7*}^i_)Vw6-^F!ku$wFf+g>$x&m=C86(`>CB zq+mBfn(g`SF$S-j)M*nWpAcU ziK9owv0qB53Hq_Mq;E~EVo-i@O%PuUAnmOrEjL2+?OEw2#fs`!ul!h~CRpx`nH1km z;$!jqzrhh-LDChZ_joqW<)`=g2c<+~k@8S*YQAV&#@B>mhPW%j7sXuwXQWeHpbl^+ zga(%k9W{!ydQJN;vp^Fw+kXq@8M&sDkL_^U^67rqi18&XT!8EPly?PZNyt{vMmh-_#yQD5Y);RYGTx5}r6 zp&H$zX=2*;9^n}x8*liiFj1QZsZk2{km*_;Ze{YuhqPpww`EZE18QWc zqrW3}m>h*I8T85n4su87DAa@>@EhIP(LgDG&_S85Z&JxUja-jz&7d(2*w+{a#E3+q z)HC^;40^V~L77c}`N;!syXYZ&A2?Duwx(JD);#2(%z|y$@lJUlyyqbdA1JNQSe=@; zrSkr_lL%X^@CCNpPZc_?l}~7P91(er?3+z@Ka8wwK~Haaz%FE7P@v4woM4vI#Vr8P zAoUStgwvs~G-?7Rbv=P3(ITH3Tlj1ybv8)#Hc6cgo!XO%s%sMziPi*C=P;>r?)xT# z#+ane=~${L?yi8gCXMWW3nK}!Ml-F^pcOW0jYjF1VsYa(E#i-v7aS@Q4W}tmnM5>% zR}O}t>%8X?>{}!nf>M3|5eK=0#x#7n;2Z#sc+^4eu-C`E!#MORmji-kf5f=2Gw6D} zh2>?6=L7QSV-Cus?bxz$3d~;^t5f4*2YFr}tJ8OjAs}$*TcK#J!WsIk#JTcWg;OALu247|6i(K2ZIwU| z7zzapZ_t#!ErksnZWY9C zWnxP}tcK`T?UOUp>D%WVkjf{jnnDu*B#X3$ZMm-3%p;J;;YZ7QEvJ(1L5c-t~ov}=)@hhuwKi6}Xuh9)(?OYa zJGl8?WJB%S($s+s#+wp{i>;EsEzY~YZG%6Y`IeCUZ9eWEd&@!0XtfRQap-NiyE|Y6 z>0K5@luyAgFLh6=g`V` zIuK)gxyipJL5?j#jx8+5ckc)}wy+$>y~}d6upEQ&0y(y^9KLrQWS==sr_qvHm3+c9 z-$JHm#_4p&yAJZaJWi)q-o>KA@^L!VNxX&f3H%+v|8ks8p96pUIGvJKcc6n9yDISu zF14C-k5FR|t8vn52X{JqY%iaePM6^gr{#(EuwT{zAX?WwtlJ|1a2qKxzF3X6wWgYW zWo2{Tb5Pcx^98?o@@Y1JHzcFUzU@N>`~?@S{hOPSfzCPmxYb)}cV+4<8FnEiga2BoYEql(mnwWf}>`MGZr`qpeB4G1E1S~BigK>F|8=#$H><# zuV%B{l89)o8ph1kFxnuyzeuOBL}`#uATC7wUtgrty%MMrq1{>BM+=so-r;bs{ECQS zn5>#E69y3@r-#Iz|A(02ai8K=4ve@&z6&`PP#KyNGApic)x@Rl#L~;xeW< z^Hb5AEMsy{gFf%}S=OE;uOXSV>Mpm^{7}&RkZFD(X@1C2L-KmHulPgeA@{R=$fuD- zu${eITv_%J;!DKXudRo-TJ4UZhYXGt=lV%j3Dye<*0Tg7Hi!;lJ%>WqY)}%cw@R>H zN`NT1NJ_8)5+FcM|ICzNL>MkwVe+dlwesDpX}DQwGvBN<_cI(t;}5tHZD!`51LVcq zxwmD|(Hq5r+GeXDn}s0mFl}M!eH)e2l_=JdWs%p)`Uk|3lGzW;Y{5n`WdDJgZ2~hc zO_2%?dzgd!b2Kf~%9$p9uGop0%nFlDeVLWrK21Y_u#Zdi-d$W^b)(oI z=?>FPNoXsM`AV@7Ry4_H`>bs4OVUt0_c7s;uawaDaZm(mbC}w~uh?Sq82pJzT4)dm znnYU@h(9NEc#d@#xk>2o9J89WNqE9@Rvn%*by&h&*KJZdh@uGuO$ON&RxRF7;<9<4 zgSWKJg33hlWY; zIr6FUCN!co;`88`(KSUc^;-?}OA?>t`h^ug1Qn(d{la1nY;h2GYxS}bsKv==^kzh9 zJ9p$ztAtO-o(XO)3Vsj`79iYQT<|seJ@k!rGU-YS5wtLk5?%u$)~2n@>$|{+l$r8o zhpciR5ON<#f=mDLH3AnP(}5&JfBm+~k$$VBejTW%y;!7E;T|Dk9eVc9igc<1E<(ew zOo^ebLqz!CRGI2Tfb^ zvuOi}+v|7dn8~g-Xe4afDp`Z92TkxF$jMR_{wQ?0DJ~Vm!xDa{1&#u1lJHyD8nAw= z1;WH1_(m02L8~A>%R&TMN+tYu5V01RoZ&DFsxG0xX4}}&X4_C;L$;y7h-TZ+?;HGP zM9Kg_@!hgbRX`D4R8+?bY6lK5DQ1U-)kR>kNKjtHl%HW*M2k2S--@?@;l-y%3Tgob zzm)B0v#Pw2Fg?Hk&-Qlw)J%n9l^wCNd`z%>j9E_FE|htUS>AxR_{79x%-}Hqg@%u@ zhO66i6k)pJQxLWCTPgUhWPZDt7|}`&(>s5QE*&>&(Mk@}`vV|&tz=&3f7?#Gp`q$< zz^FC3xN5B2HVAGTm|HEABidklbY?oejW@w<19L0>4$5=cZD4K-zr!>`Ybpo&MS+?q zeaR;H_B-WRg+B?$`l^j%HA?u`9XOT@ag>~eQ?2su5c2L|d5_%zfrxhS7`Tw%SiBu9 z-gP@1WIs@(Q$VVGi+rNy?nRr>ZM;qo?{JXk=3cW}sbxI=Y zCHoMxCiN0-0|#!rnk1x49Me8%R}Pg<7shIkPh`jiqcOX3=(Sb{*)Q0YLt&Y!u+VG1 zok`>g?aHB+RtNQ{+m%DfKRKw!tX(O=tAwAz`axsYZ$dvn*rCSnMM)s3R3;5DksBmrAVUyyv`ENlaYk12lzU_4RYl&Q zzuYuS-OWOBVV*U@JZsoId-o!Gj8qb>Nyegb-fyOH*06Dc08|^Xs*n#jSb=!#H_M#c zJ`M~iGkYJ_>~KLR)A{Z<+(Y_%D)kHnd>Eca>C5EE)(oP|@cvX>BHqk}@$f!GBcd)m zDNFHp`Gm$r@HEtgulLEiuuP`zmQNIeSr<_<_1G`FB|tR!aQU@`-$AUAW_S zlmO~N)BZMfp%Fo6%gjf}r%7b+_c@fa z-$6Z!zt5rZ`ymm0qIJJ;m+?U4S&x-6XR~}FVIuxx{DMCO{v?JVIt)ulDU6BeaOQqB zaVS&F=x{mM3Od<;3OZh<14-&6qzrV>67B^yCpF2Waz=)mwn|84$8K8X4f+Dc77mWH3Bw9c{Xux@5A3P``or|hAB1QA;h;W(jE}S*c&e;DAk+0%D!IkL8BNB4ze~H1AKOG$Caw%}c>@?a;pfq66%<52WZ@?MgH4zsPrriQoJ`Y2D-Sp*kK2@|QqZP6k9X1FK9zz7_xghl$_n2p2;458?h0C&fr$?snNCDbqO9FAS7to z%em>~=i~wjE|O1o1$~i%AnA5aZk4vqjuZbX<=+F+X$dE38a*xp@1Eh$)|*vrX2NwiKr-4(Po1zoCBgWjuU?rrkvuAprx=uS@OrN6RM?cBAcjalB4X zCOKulvsNZ&&lY+CZ6DCk?;K2WiWuyxHdyFC-c~W#hc>5(!G6JyS-WybPj-qJY@5ut zLq3tu?05d1>?GoTCw!nyzjGOqq=bW{WHI@Kgv*tL*MSqF#6p>>u)xWba9oO%Fp3|L za88Pp@Enp2E%DhYE6B51w34blp=@2+!?qH!zZII7| zxr2{7p+GQqppqB2K}(zl;wFiA)R~P@%X68$Jy~=g2SuZFkQ=36Iyos%#(@V@5C`^7 z1%%_kgIocZ0^l>k_n8gyq)d0BmDi-tygsZb_Cp;GMnL?(NH92_^7ed zkoPmlZFf>;!yz7$JkJ*}H`<-#uBPWZt$#0rveTSIlvzl2a{K9BMmROiNtrIqPVOMC z|7zyTj&Bf%bpXd`N2i6$GpK92lQP3ec5;{Txa)hq1^Es@3v5jAXC^p4157BhSpct3 zMGN*kfSIMqjCU)84g$orX`yV>8{`v(NH!EPYSW_P=qs^My6)TZ+?;U08G1o{?QARIV7#O6X$olKf&A~#CNB!*yyt3^V3G6cD@?Scg$ zOmJit?2jqbBAMi|k%%OJ;3R}(^)e~XmN;B(k?>w|`e}lGJ|H}+d)Y#vvQIvt@*JBS z8mYumnTQrd7B~2oN^0{ge4$sfgjX?_%fHW|zZMBD*$;>(xLkqMv`4c zi-go!R0F&;Q!t$shueR`1ouON<7oK=#}kNO(#P#5?Wh=gTQv4sfW4HAA~8+fyX z-`)mZ*hMg{x4^+VEa5X)7*KAJkeL=t)b#9ZK_Lnc|If}GI@rZYJn{-ZoyqNgt$e}{ zpF-dzELJZwo@OBk*3A+=z+|1>Rj?Rn!GuVq5`I=2c)f%d#^EA*_6Pa$M?}u5bIN{R z2q`+L$5VMl>hV-8b)A>(m#p zXO5Hl@E40oTg#;@cMYf)>>5}r$~Cj_NbaLQVZ7_QIVrOqkkAxb&loRvbMhuzKdon! zjd&Nk7X{$YZs`BhJ{zx7v(#lhblEvxr{j)rQm>yd$LTqMnc*XxqTyvfghiAkie9u@WNWIH7be=8ic#W* zBQ3tZl~ov^JIUOYzLiZj8z4XOT>bBK+JU!V4XRUA!TLK;LE@D_km7pbC-bc$R1@^! zl`9Rb_#OYe+TkR(D3b4(#e)D~D~EV}+u=-5vCQie}NU$qLpR|vkB1%JzF3ci;G|NU@+7fFV{fguatAR+sd;C;JG z!CPdCDR@P9Q}6<(5d3%0u?p@B@t&ayQ{3!#tz!Qv#Qu}TUew*+SHoO@OoJU7-mMby1Z2aY_1zPlV!^xcZ9#jOd?pN9PdEv&BnGW2xu^}A z4CGuV3h5tXj*u!WA9ZJjrfWq8HeIkLoG? z_XYOf+MXEv0?Qwj*Hp=Crf-m`kIH9)>7Vf^)2rmh zHfS>JD|8F%yb`xs2Q~@>P3ttgZ?)2!!b)$ll|J$aE4|59>eLI}vO8{%CboysOWLD} z?XjX4>NCo(RZ8-9ToQ&C^%N4m6Ng)F8Lp^~<>NpPpXV}#D_Wz3#LBYR52xX<(SY7g z%DaNZ%aqABG{JS05*3*7dZLJL0dbis4AnG%$IwkCW6E^m729e0Z#XYA8mR;{9B_SNVf6 zx^&gM8T2GxtNlK4Xl66syn%?}^-Z_3jDh8meVKGc%w#sEuah##V<)B%0{OXp;mc{g zighaV*a;Szh~?7@i*R~74YY(F@~O~pFZ z%j`j9Zz$HOq@R;KPZsMm6*y#;nKDhsrKUy;`M|^}k~j}4I;{$)+W?eg zJ1Y;rqQ)EO*wo6zt?R6&UUC>yFVV){q0_(t!qiLH9?#*ejO5$LcoSBYkY5KRm|A3l zIiKh3j#Js3;^nkcr8(=QiZ3Y@C!Q+JS$C>3XQRY0BvVb+P_e~>~M6laFn%5`?LQgEmQ_1#wn0E-zuD& z6wZ?-PN~G%q;S?KoISuXtyo?a3Pp_~G)jjt-lxK7O&i2h#Go9oaP@>a~_;E&wPEX*+uo9i#KU39PnW+ylp#%5RyYW&+;^l-PPJBZ* z5{*%>tN)!&BZfGsXCOqq@j~t@Kb?$U*9^hZQ(x#!`2f5$1g&VPfSdxzK|o9mqW+55 zu+X%_R7E9u_E%O7^o}*q1NZ7QYAAxl5RndEPvEV}FW=tBTLtSNl7ATLBzGhfijf1@ z*~2i}fT-nF{(z6V19;sqv@M9PvI|7%9E;LHIU z@&9vgsB~^k&^@zb-7~B1(`mwRQF@t>`vts|k$71@9IjWu#gc{8-vMX>U@~zy)L|wg#Xnkt;RJ4XQ6gqfo)NbLV6gPtBC2NHPsU_`<|A_G*6 zPz6DZ>-epzrqa03?_&tA<4R_zKNuV0k3|{UB(l}`gRyWdVgaXvS~xrc=d{vnuG!-W zqHL(mAdQ~r7r(zd@Yj;|flX5ka>D}F23hFM!bK8gT%M%mP9xg2D}G^8 zoJXZ0cQuQgq`fV$@u_!LbQ-Nl>a-+-T%jqnf={H!b9#CUmLo758|lRtsE}5n;ohH3 z-)2y5440#@lrBG2!tY#@X)>O1B7sn_(uiUwcy%aPQd7aoUaH{^8rT9iw(!$$-fF|9 zjOLHx**|S%THG$tnzCptA~;eCpIZ!i1AVdR?hRDtkM>4moNLJR_+?Vr@$B;ABbQ%Y zf7On{Io@bhNi0%qOue)+lpn3~h7D=HMw+-PttzMZNB-vhZbRm&}}@ME!DR7`5dNqKsEqvd{oi4p6SUmgm^y#8R6wEI&vkfBdOM%u-i$E<)u zyJ(SLg6O#);7pMkZ1ccy-{q&>ykWu> zq1~M@jT`41i}P;-(?VW+CN7(o7pePz@<7vaFj=a zvZ7euVijySQyvb=H=H9V4+rJP&T&%Ou(>*w_OtUAnh}UCjJZ0kJ_mhvM+&fKoAg`1 zw(=Rze8z*%F7OExjfc#qj#8E#54Ny!Khnly+p6GSFZmZM{>PvDxBS^}6z{@stTK36 z1}|jrpX;Q&2$2`n<;8QA3|@;2-nb0SQid`m!vV;^i&@$Ig?W5Pnms>N9JZf38pJKO$Bp&qY|y1E zHcciwoJ;JkLC@~NZ!_rY(TX3(mSE}Wi%4N6UQc2OHhYp&ksfnui&HTyS}sX8N+gW1VE)m>;`wNIKmCh`KO82JE8OnnL)_cOwJkIe zQpyg&WZm$C62VSHdE5@Y zx5y_{?u!;=!(5%ZmVj773!+RWdsEX`g?p?n{|AvWeAx7jOQ` zU?^e~W2S*Nap?QvMNaAiYiI(E`b;sVV=srci4`Ta&qSb;yUd8jij=il80Y(o;F=+y zL0_|Cmt2f@+RBOVU+kpJ>{H?EJRtdz_5AlGn9cZRu1?`o?d%LZ*Yw9+oi4k?Nofb( zwo%imVhQm>jmYC#VI#+-@YJ|<5ao|dw2x_z!r!fSxjjkCla-WiPeL_acd2sH+mlel zxRK}%{x$G2Oe3?b_a~uD>Mp}l*+6JOck#UHD}WRN0>7%~I~8^V76xpxoPs`iB8!_9 zkVW*8e7ehM4iokQAxnX?_X&yJWwd~kZ=480{iA!7Xk?ex=DYM$vx3LgPkniMZThNmAaf3_9|1)B>U2&h|P9m$o#EdpYnHC2D3tlFCGv*(_D( zkf~Xy?w%H;bqpU?hNb0*F?EbLR$eu{Ivkr$pCqAoy8|iM5hO9)<>+0f(Z@W^v7j7V zFNGI=%9Qp4QvWYZM1vGe}UG1HLswk)266W)K5jSYH} zEoFBGO{)}Td6Nwt*mIT2f5bahDQEI2`qoM(xmi!cM>B1(*TYqEz;(0j;%_r(Rh5%6 zi~enzV;1MzS%u~fBLF&0XjP9x^@#mN;9ujHgR7#`#OMJ5elG<0*z$4xJe}%fMqKmY zd3c^q&-$I@d3v5s&3-5KczK>qmrZt3kN?cm>9NU9B4*bjbKRB5?#s!No$K@t*g?*s z(}kRK%#1wttM3ROW9lY}KNr)S*0RCmlB`vhGErsGkZdiJrJ_uh+7Kii6A)#xl*^Z*cn1Dlg{tsE!YM zR)wS^&EqT4Y+(d|P@$G)@FC?d0BHfFoNi>=1FnElnT2Pv>bEoO!YhyukTSZ1Zwt6X zU^N2fr~4VJ5m>E&l+gml%8Gzi+(77Dg?UDxhDJk79yKS-*>~_J~?WaPd z-htH2dV$Y7W>zqS&SnZlz{!fCodF)CB$kv%~1tgz9!HB9H7c{;fY1)cGr^X)vHX4N>!vuB=8+wtSKc{&wO z1-99A8MWa6;wEL&rVOE>*vzJDXx8~I>szKed5(uylO|15Ce{(kUc%jxwnT-HMF`z#dutC!yOKbSfMyEP4uzl`~(bl~+2+bJTpD zTvtIV(?+A!th6amOBbtP<5{rr5bU0-EP^rQ>8q7sTz zd!tp+D0u<6>{|4Xp=sjn@oV8kLJc3(XuXyLdI907nr_!YG0%LRTBTwaLb37lb((mc zlROvC*C|%#q#oY+I?b&E@?_{bL|DNK_xDEnDhZSn9bO#BA5KnC&`ahBDH%P>wRg zL%76dZ~#AP1{No224B?s)C?z){mFSbbYiw3Jxo5a-EkJQSUE3;)-oE3xkx7aS}a@wD8 zlcB;*8ldBm_v^H8rjtCm_vQ>JiFOE|DJ^khD99l9=%n(mw2ZdM7vqV zzPE{{O1wREn=;>S4eb(C{fqI&-p;Mn9}M1cJ9@352%G32C;RV!OTdMkM+kZH+;jt@t{)sHsbfL(mX9y_>{E{w_zNxpyfE-{p}Zwig%yUkQIyViU)% z@7;wyTz)v^q-H)stb%mFNFh;`u)c=WafAU|#9^ds*E0WW>UZ@=NqOJ6d( zyxEATp}C(vGXS`ZoCwQzPvi5bge(is`Eji^Fh}6 z@_TTsfaqWnmW~?kQ5HDJB9U8sQR6lnnuE{4JR0@}qjZyP_6qo+okUwb3TYsEY?!`7O&j=XR zQKH4n@uNAy=kSiEtYxl}elfqJ^LdGxoHh>;6}`=Yd3YYS+Cc_Tqc<4$-g%I$V7Q&! zQCh{IR{`P>xPG{(bSupenCIxvoiAhO*5QYVon0eD#&;7lwyKP!BMy_XUS@nJG2@K; zCCBEDG9vn>k%HrUiP85MZ`ibEOAo76KeuUl)!uN}AFL$$+}3YT25q_@cEua4dc*={ z)zA4I(-JLgNcTcahK=MNVk_JE+=bj|ZRN6`i)1>7z4sf2e1v3ijB5wu_ge(t&7@Iw zv~|}f!)2~sgm?vz2;IPt2N^33chxtK66*bkq2Q*mI(?5cE~LV9Mba;cChPHlG+FrE z_O`ZPdB08txxyX;4-d#)<&BQU=T{_P z`GW|S&*O6VlY{zSA9PZd^8~?Sw0yeD=`RlOdpEGv>yRTHiqfqqsNG{4xQ@(9!QgdH z14xc$l6P?O>wvTZQcidBwHtjN0vTcahA~3x*|-oVKA64oA!+=UG3|}d=FJ`}GTw`f zY+ld1_v^GBnQ=Zc#yNI*SaFmYr&u#`0gV$J7ql%PVNz8vrM-pq zL;*dN!UgnD3TniQ532$a@9_e9m@j$Z0use70Jl8?YmQ@UKEl>~>k%hqZInKc*;BA` zm(!zc*?o_|vYBs3mM}D^p(okA~Pm%!g{N+*1>Fr6r+w>ohty~Wo{)IBbGwzoQ=XM6i`Rc0a;(Z}1oGh+Ya zPD*P=kRTIKQ_{HF)6%$u@B_xZ;pqegF;x(kN6Ar|ZN&&m-&1V1n$c=B0{2!<@s$#uRU_#|>&zCfon&pN4B#R8p3 zj2L7(PP$q~3@1DX&F$d@I$hmY=($urQCpeXZ3}c7@tjl8sgrcx1s{w#76}TA7U=Yq z%$Y1L1`4d;tI(M75^5N>6q)92>?hcZn*Y2o^Yd(G#}bQ~1?cQ0%FN=OG8@kqW?rhy z42fE$w68!~NKv^22IGO(1;SuUkpwe8u|%5rqb1VJzv2hX+`AFEV5B0M>ow$pVa7B% zsnR^aYD}0@U(?5>LESX3>SBjCQCpLRF}j@_Y-EqAJ*5gs3Q^ zCYiJYNhqVz7iAgU@}exGXYm7N^bO}iRcMvDb|M$5LZ_D!s)FkxVSqMOq2?uMA6Fk` z#OkX;Qrfp9SmBVc!XdW8yq7Fi5THw5R#p)2lv#8U7t-~1M8^wX=0Xw&x3|9x%~42= zQpOY)zt-egR|beTz%da?BtqH)nOU z+-NEtK1(UD5vfDWSYvd)Ezcg{<@C1`&K@Ky(rU{%F9@skd$9~-QZ6{#_0Ei=_2`luY7_q{yax4)ahd;9zU_6^c8#ROj@OdrfPBH1;5Q8 z`wD41yqvru!FU;pSb3->7#ms>8&@F?aoEk2YH#hRV2n>}98&)%UF8!Sym>k=&F1=H zEQ@{M-3-zQ(+ceFWY=m$H{7=-R?%-D=}&%TH|1@--2Oit*boF#66oa*Gs;6gu~$_) zNz0A&BmU5HZlu5XLAFGEALE-6v0rN+R~2FP7!HM9i^2R0pT@ub;Brnn;>U9*gPz zWQIHzlZzEdm!Eo|3h=R*lNb`cLjLfR8%bmn#>06p;Zw4wP>n&|5No3_gg3bkfp5%R zr8?dr=q`vl&3#`t(d*4@o+a<=^ty>Wj=55&rSIz|QbQbh-TOLiLd9<Sxd}a%;L_2+3z`pR83N<=ap2e-#ff8vcQ!nG}0>4 z8$zTvSfrfy#q|CgypVU<`$D8Qc&$x}^oA+Ye;9o5eJN75SD5-gN~8g6Iur@L`G`oU z+LWXI9IHGZ3VA+cd9Ga}R-ixB1}w>-_tpq`KGZB5y+0K4WEDuQsw6dcHGRSYW~{}! z7=6qy`FJtt6V6n-)*>LhuLQdyO?h1BTIJd<GI^JtXO-cGB(4WHB*A?i_(<&UyMg!j{q>Qs z%?(Kw+uR^Rxy({8b|s@Fijw%8SaCBpdH`HaH*xg0vKh;9bQ5R!t{Il$8boEHSd+1= zf|@?%V~i|Og9>HRkw`+leh*29sS)6As^c1MRcnz@YZ0r}a~;OLM2mP)^NMxKw2N3U z6zhX5?;Go+--V?V$3qHK(4W>hsTV3Jrb*fw7~W^HFCAkQW4RDxIg4@nC!%sLPeSFK z|B0}!3Y2-jqzaU~uhFSV3Q-6lPFbVVkWW=RGePQFDxYX)h5!u~YXDk%i$O|ZFezd7 zSgWMlgrwV8(u1EW6K!LY!Tb7=woTfroW5sp?s_M&s8lWr)PM{Sb=Z0(YLn!)Rz4wW zCD0&hCD7W7YB}hlge*-Ec9D&pc99KEd+T~7>>?YVDNzLR5hHKJF!MH>G!R)It>KoX zqfr#kt`{eN_+_I>?r1do7HZg|q-ao)G9G9R;;?rRXxTQPxx-Ecetl$v5^fMva|gv1 zvJ*dZQr75kg2+nwbo<3e-)DS=sv`Xill_Q^OqNaKMJ|n5FBKC!OH;x8@z3Dsa>jQl z&-o;8R00ST$}B7w(_AZ35hLMZuQfW=mD_o|@fK#KCalpZZzHOt0Y8*UE0A=}8l5(7 zbdqP*8lC8KC-qpcMyHeTDYEA8_&Iu-Z? z!woi~Ui$&|B(2q{_ZLuI>8_$PHOHk?FD$W|0C!>X9hNY`m@h0Q*vck2>q}*Vt&BpM zr9PhH+QzY0-IqMawT%PBpT2|w8EbWFl>+kY-U(}UI{hmrc}`xd(=}gVc5kgttG|M< zvYh0o)s(H}LaU6M#bm^0J7nDTm6}fy?*lfW={Kg4e7}a@@dlsSx=D#Iut`pf{m7|R zo3S&Pw(!drc$r3f_$7M_<{jt`L(JB;(hE^rNmaA@qn+-Nr!sFzHySf!s`!2 zX)Xgc0T82k{Brs>Eb$l6z3JFY8|AltbRWMxvkjxhN-X9V(B3p4d?OKlh#NcFdv+-U_ULbtIn^#D<3VU?ZSAzHw&7Xizx z2WT=aVbBJE>}RgkY2!u0K)dA=ZN>MPTN|@hr~ThL$#ec%oqqfdjpgNQbvoi=fuARz zz~2G<$XcC#+uA!2;j_Y(iw@8Q#shy0?_kbC25ACBefZkZ_^Z64_RVyj+{>)@ zv)?yy>Ae0UJX%)T_;E*!ij@>CrZl)*`6%A6f7GZSGiVcd^CZSma9y)n#L1LkQcvY? zO`szWt%44$kE?>q(4Uk+EXP$> zN}0CFCzO}RRom*rhgB8zqB1KNc~})(DtSZF^Yt zqo3qq)!U^;i{%qCV#Fs7t8SCfWGN9+F>LI>aa5G9gb$Znl`xO0LXDqUXrftsRJG`5 z5l7Clj$qmzRsGM;7$?x3JWOc$IbK}m5GB8&CRpCjWZPo0N5J;2rjf1P+yZw7#ojW1 z@H}rM+`$xsB2rhObivWeA)31VlBXWHst@VCNvBJGQ8k@kkN8#j#1O7*?kXCA>4?XE z#r|n8@0lJ3K+#^gXPV#c-|M8zCQ;)@BICBb+>i^%7(kBt4Gq(bwK^3}w)1h%VHjY( zuvVwW-<;HA*;<{B+vlVn@2=J9D*X6ptxnJ5$LDKx>ioNtdTd>*)3D#slw=5Jj8=OC z_>6po5#dGvjG9zpD2!S`fjt>?ox~4?M+K*P1AYNbM8t$qKgQNfY!CpvKf{yS!B1hh ztsQ&-!;>Yv!wG4bHpU+W6PCO#IbEnVJQA4@3XS$gDh*1r6`e2c6MA`cI%yKah(yL3 zmEM?ts=?XQY^5DzSsH{iTiA+2xohI{o<^`dgzvHTog57XN#FIooxE>e8$)(wzLzuLKb_;nWf&60Rgb-;@ zq-(h`B790Go_~yYiZKScAL*{;#-@i=dSXFTgaW>?-fDxi&RVWNC`|71Mk18GctpC@ zo@L&sF>pY>r}uE_7`^hvnuQO2wsvcZmTLsdiCn8Q2u4Q5vy`38E{a`bxif`J1Bkg{ zuaDORc+n%SgfR2CsYawC5Slh35c0-I`y^Xa*3PXZdilI`a!sL6v$rjU`IUV_)=a1L zCv9`KrqC^N!8`M7f>+dpVg_k*#cfFlcyeRJQ)T8YL2B+dKHZA0J4)I>4b#II15-+L zdayj-)B8fNKQ=NFstNbX9~KH4q+O)tPOBt3<_3r;k;Zt#xS@+9s70EVTTQe#ltH;u za9ADJrsv8_ukgFA$U$S;&`{7f*yoFIffx}I7ae(n5=PX0Uk2qGfv7PY1%o#iis-J? zTMs~tL{GpOwOFL4JZ6&nG{tl)SJa^ODaUa((ICaGfE5VR9@R`MsM2>;(~~@l*4}c1 zr+WPXZ&| zt@YJ%D}*EF;vPCeh?X0U80BIKCb!HVL%gS{XN;KY1=TCLXJfiBH-X?iU)FoCuUDr%x4Qq)iD3fx_tHVUaS|+)I zbUbRLxjvacDmvPT#o%9;@I^{k7qq4h^H-wvwp$VtwLlw#E|agJQ%f$9{?uDgYJ(gx zj1tknlV)hSXzZ!Owp|=|A8Skt_))_yu@+aDZ)FfSo#o>EVlXCO6+TYdvUrE) z3d+g&_PIst8|jrF3r#VC+)@q>1VYmcA88*XX}LbQUu$yaf0uPQJ|?pMzsxo&7_PxG zJ0f`PpzUvX1w%oxvaUTN8VFQAS2-=#P)E4p@tuVjg=%X>>x(5RaaA1uDwBguevuX1x@ILe)ImsCD7RJ4P&1h14!~me_I`1%04b;KuWmSBDi^yrTMpv5D-?wL~v& z&!F7t;_0d#>m5tIAd+i3e)h&sv$lx61Noz(B{h*qsM3or=QvBf`L7#>3cnHX{bOSg z^i~_PzPD`%ki^}WORhwAgfKP~93CzI8w$gt<>yQoGx{(XF>j@-EEEc~C5`%6>$-HJ zKNxG96q_D48?Xt}!|mFyP_;kCw-t0qShVRKCUGNUMA|fm5{gFH(jY!=R}JEw8P*1o zvCYoK7pnFLy}=ltydrIPnwHC@5+(XzZ-ym?5xpv-<&TOEuMLOzvx?eHcJ4JEc)LCF z1aD>jNL*bw&#vXhyp=?|-%^1AJlZvt=G*bZ>`NlXR3j2KigBWRXmuDG#l7_~eFc2r zFhz%?|K}cD;6~g%-bC}r1Cz};nS`bS-3OWqesUkmF5{=BB!FC#=`H?YmXo;fd$4b^ zx7-NE{$EM`U#iG&bX-#MzVmB7NtY21O3N&pP{e0M`v1Q&`M*>qPj_49s#mCGJwq7>eip)Uv!5TV_C}&r-hi2_(XSX{a^0)tiN4*7AS8#I&G9+Y=K#fl>~h4zS{x$wfoUh zr8@acCL;4gul@t_L$7z>$E;mB)QEgA!Llug}Mxh}*t- zT1Ap82r07A_keJg1MDns9CA_~Y9Ae7C-~W9Dc(D9{shmAPVt9L`V_YEnU%y&IV%ZH zX;61^N9oq2{y%3>BVPP;D~I*p;e}g`TX{2CKdn1u#iYb7^63tR=?+G}P3!*8LX~@i zW7z;wpho!=2cqT)!RE1G&w*aM1|(6(a&Pd$Du2K*=`>zpRrFCo?@^|=UF%MHp)k>- zEbcK$toLKAcO_o@^cbf+fEU*LG1mJ8TX)KAnu4u}#5a&>)|T@Eara=+vDk(-pw+ zBfe**jxWaWcIr-<4S>R{Z?vIPxe=(8=4sKXH6SW<9ro=_Xwhi}((FjelzJQ?pQtN0 zDzra<_78r-X5ndapJ|lF8mn<;+qg>1wxLS&OI7BWZ9_{9t9ACCt;d0hoFbcZNP~`vL2GeXfo4CryyR9Su_rtdG#EoOO=9J7tC;W({rO z%)i@FG|d9^4U@Ppt-EZ;w@HiamQR>@nMUGM?J9AHOpG_*X*+l1kSi!k5BTEq-$fF# z0tJPhU{rc{N;`RH4%G!kXJTpnmrWIhX_8N{=8S0l?@d<{u9eB3$|ulPDG60#vrLRj z$Q{gA65J}EDsPYC?{jF2$>NA`tB+$Lo}p)tv-R zv&8vE;jGG%IN6;$s2Xh+q`i@dce+_U3a%GoiB4g!sG56isG3_ki*{=-R~o&GsvCQ` zIYHYwmv4)Cpi6gh*U&!p`VIUNqkZh~v$MNX&ub$E`P{Vrc7(rwIvob!k9o`Jw^=u&`}YifE~?}Y+xbTU3WJW z1q_IQh#Eiz6*VF%Qlp?Cpf?CAA|(_B6(safgn;z=d%d5TIXNj5Kfmw&@BWd@&df7U zpI#135gA6J#-u-Mt}$%z{AY#1V*zN%FSXOo=Y`gsgGGo8$cyi7wwbOo( zwqP)sF5wDf2>9qQA|6=A#l@55k^OO$A#)WJeA5_M-X0fEjq~@%QBVB9z%n1@mP8Zv zk>zq^h;qij@}oGo4F;CQQZi#;`2(^@$XG6AJdY_8%h+XzkgY(5;2F6rT+G!6oKwzEk{DMDFXe+r`N<&Qiyg&;B>_bM6$8>5~RN( zW*v*6UiNrO@&!D-D-B*>k6eBvhDO={A628`pizx}=ej$?j@m&X9DwyuS;WGVF|?Y~ z(vcR->(!q=W62JKKQYEA(9*-8iUvCqX! zzGHnfb++O`7D{@yKkbRc=wWF6c(T!RKp6;7StJ^OuxlAG4XMxer~Ui}Q=4n!$(D~K z03Kk#d;wUD)MxwCA;$P0B-+vjOO$^zWV;4numFku`qNp~&+rBiE<|Togwh7_WXl*L zOI_en-iBNXiC!8LHKJ`aEo;aFAnBz#`EOf+)F9;Ea zhmq)idp1T-dZE-#%O&8Pr|MGRRjjk_q6%^W9qm3@QPPWm%;0N01lPHHax`{03lp)F)ZFE8t zG@@vuGhPr%HQMOb$RaUpFOWKU0$ibmbs>`*-(s0nsIn#@i%0ZXql8>JGK5@H6xbRD zYq-r$?MDl+J~9N@e1P4t%}!C*$J2G~w&BaQtlN4o3*e_@2;fZshHhsiN!{*Df>>yp z%O!i|fWeUIjvJ5#GbQE=Lvt32FTX(+zrZl?lY9wt4#nXoJ)E!_iVuEOaI#p2;N+Tc z5{|vYDN#9^D)`N>3HbW2ixBWO1^+}-2|o_-?DoZy^fw@lq4nCP@nV1)8YeIelp!o? z8fK4`$%Qfmp>Kf%qsrTEl%vWIZiF5%pe>h#mH`F>+Fv*e+DjZJRJ0OV(B57DMk!>a z%dGEJ)GBG)B=oWYm&cH9zA zNxf(~XGgV??ZS*&7#7WGg<(-Kz(e!}-v^Amm3fJMl~Js(g+^xRt+Ld5E|qi}N)-x- zt+g=rnsA$h*vb%JF+|F1jAKtN47Yw$5PKNnEJG9vh(oo|W;MKB;yA<*?QRDn+%RXn zF52w$a4g-Urh%wA%F5U?L3#|SZ9 zrc4Q!>sT}26)5M)u%d@#4EYlN6UKnvtVpKdEkeO?sVRjSzUEHVlNU`E*w4xk45J`3 z+~rQm@QV!Jb(@_c-V^YhWeE7`==q=9W~b?Q#*_8cZFY(sXCWsMy@C$)%WZaA1q4Dx z1+wHy83MyBU_g&rt|;Dcm(qXQIH97cEFk8{ep3pIGX%h_|2N(j?3Kz8$Wf4y&$~;C zvmfxRt3rutE+dEbN@U6+P60U@FBm`ylZ(4sX>VGgAa_)TAje>k<=(p`xqODl=-Be1 zfbT3r!1Jqw5A2VlPw$qa<9aezLBZn2=y>(ra&-JReqeNLZyirOI!=`3n#d64jM4G! zt)Wnij#I{0867``EIH}Pkix#t;zJ|7Zo9$`RZECfNCw7|~+eo+B zAj@r%A<7wU^FSNu6mC;Gp^Dr1ktN+G_6?!%w+!*qOerk>1c0;u%#&i!vkHw1ISi!3t_$5HZU0{de!gwpud`i8@C!qK^g^F zz>nKX0XH-Jqle?DP{NhS5b&k2@6f|>RKHz3HGcna9Np7SvZ}zRNpP^*uY%PK$=Ehl z7h13aQwp=%tDSPiH)aS{H^~sBQIJ`EyPagU1n{hfVu|VykXx2*vs3y^fuTT#z)%k6 z|3XL5KAtQWw%I8|<|-)g{HTS0k~>5}ws1~d7l6lUh0j*6MY&js=v z83H*9GV=Z%B{Q=a9<7RkTV{ga48x&S+0aq8D)};Zo(y3)qg5H)Nwz8<;s-jTWu0WJ zGFrlql_AO*t;(KGsM}~&iYAJRYb@17cZM%qZsnQxNALB9+yNZ9GN!Bkxk3eEHhrNM z&!#Wb!fd)(=Xh#{k+*h%hqjlaz#BXj6clFq(oCVi!W0TDpzsPP3=O1^R$#$K3MKWY zas2B$$P`}rtd;Xd-aA#ta>P=41wBP4Wxko-bU|-A)9ZMjno>`4}_D)j;jPCw7PPb@Vr;>SHVBF5ABg_dXUil$cgW98L7h!*il_BZZ_h2Z`flZ5-< zhSv_TUN)i&JdU%ZGWy669(RWVTLduc_1o=KATivDthDWRI?^ScoSnDZY4QW{u+LNa-fyYqO5F2pw%<9qsR`bhO2S35|LsL_7@A_(2?Az|#75)Ih#3d!AxuH zLy(L$S@gEhr(tM75tS>ol%{!>JKG%?5ODcZ4T1cIVAW1iI&bXZ4d(g9MU?J? z{FRRxs)D5k0%+zZjgJ}rqy#;p7eg2V=` z*6A4PVvyKyF%nXEcYu*9+zO}vTQly10{1~}#_1UH8MqG$+)2SeCa%oTA&!6YXibZl zRmt`6Lg5CT{7Upr33{iPUif25U8mR)s&xfk_W*J0md{KYsQmvii||^QJm30S7@Ui2 z1#tHM^m-VaHR&V;u!|3aoc{V}n- zTF9hVA?EXkf<%R39Z$xP?eTbVFi$Jw<4*TK9xpc3i={xZ5;a)65gX4K0r&ug4fT~0 ziX;?za$^KA`*Aq}Sce}N0USfQCDBA1WVuZ;L^)#wQ11yeO&9@`yi;Waa3`{=8>MFT z$P>s)IVnrbO8PKNYr>m*ABGkG5<`}!pTHW~yZp-46xP+7NJ&FVUpIXew&Y94CCvLq#DoJaVF!BZsp=Jw zM_{6s1Ps*vrDlcTBgy^oA!O_ey#JI^NJ$v3zKa2Fp%kinO+tmx6ykLQcfi1$|Gk;Q z#xVY}{KhbBc%L&UY!noda(u2VcNXnp3q0O4o|2LV^rmfLxb{DoquI1Q?9VeXG#9aK z5A9;F*w+(#>hKC*Kib5lAALHW0I-Pxo_iX#d&qV>9atv#IVnR})Q2iGX}g{FFxZss zcCtyfoiYTNA7D$i+o{1b@#Oq@yPZCHriLEUqy_2h`4UOm4}!MFJ;E+w&|PfMRnG{6 ziqn!28SywsiWXrP8}xyUcxnP&>|$M5Ugvl=?P6ubevwh#-d^_r?wX`M6zzf)W**K6 zdz@i=>;+X9u0Zn%KRUyT?BG&v}{p!qKCAoGENLgHMhv&xGFmEs%>wy;`Rw`wO@6{)9qJtIZ_l2oiIcL{gR@F^5U? zL`>BCIZWdgcRZc}ofp30TnxRANG}rKcy5<1hf-hjF}7{la_m%Xf6cs~WXqV0fplJo z7W0RqH}r;$v7U36)Ixh1Lb*k%T~wKAGP9yxlJiP61faa1k~Ego!q>iI)1=L|el3^D%VovuvX6tlhPUz(m{GM6&}| zVj`|lW2M8qAwx*{o2pSNvp{CT4q~LiHEOapog zI-!d`G6aU*umIa6Ll)eL6t+p?dI6sajE>Fyp_-7d-=u^R<$ zVVAwaE_>N7lS9HTd)Y4Q5fgUV%XWD(kL_}Z?J^mWUR2I@xxIh9IAu}JcCin*61#9M zDUh=7ha;nw6b_KJWYYjuOQa*E%MjKv)sj+)fxo_qhU{MhRaJQE7a?j$hEU^CD4EqB z`+=wmN07p5PySWFKP5xJBbU{_WuR1h-9a^|9S=#UMoMWo{WpP5X!?ZE^aEVwrT<|lOP?QsRgDHONa^z>N4&_x z(wE8<4ZEsfzePUS@3%LOQl<#@!@|&o?Tw@9`SR{U@f4wkY5d42<0+TG;R??-;Hlc! z;e!ZDMkQO?7PE|_B6touDgvXGPlpK^NAU^YKM@la9u_>o}S zf`u&REnlYKORp^B*%u+>dwb)kL}ua0cjIsN#!+a5l+iy`$oN?_=8cS}aH_x@qhy>h zq6RZ&;{`J*qGXEbF)kExDHmBj|AvH=mE8XJ5jcnrIvgzX-AZh+(P|Inprj zI2Qink(fNt_z0YKs5=Ug+z@@h-+}FdsH<0y-sfB&qP=J`>uJI$*{)zo>ZZy|Qa52q zm#igFjFdGF#y`9_j+Ts;ib?rcD8|Lcw=~I%qe5xI#WF;lzZy!qIWLZEFUM2UJM-dj z+M-il92HBTWim8?AIXcO_AkRSsHhnq3uV_g%vc~(t_5be>R<*wyTeW;GV3Ol^{+8# zzYacVr^SB?m2Z+E$RzmQLrm0}1?)C=y=pT5T@~ifbouEn zHouKN!AYCE+-{)&EsVhV=loZ>qAg_yvcHDqX+MB^(a)UJ>2+9bZ&@6bd?KuNJ6rAW zzBr2gRHWRmPABJoD*WdmQLzA1EMe{lljl=LY!mX0l_9z-2@l^Wm+&DwhS1%CLe@ zw)?!D0)MFtfxivdfZdbFNxNr`tHJKZ)qiDrW%7dE0;e$jq)2WqCq<&U{Cb=){Ukp2 zUXGYB{iH}Vm&3-#i}3<|CY1Fy>Nds;OUKLc!s+pHywF8DSa%shFfm@J)}&WiW|or- z#b*8^c) zH0kpRH7Hv=o+MdUZMNdPXrEbv*^ydu&`mU(UrO8r5jY0@k9gjxA)I1WUwnkcPVwpWO7K!&iUX>3kX!n!hq!WO_yaUis!F!vJ*mjB_oZ1%rM}B#2(ySv#*4GDvjhg5 zzbEU^eQ}f~Qx3z|ty6Z|$uCn*!(7dF+3B|T-)0F%KAVx_@xrV9s|QNS@Nh1fdRRUVc`c7!##!Y zrYc)q)x62P>7^Vh?TFw^o2)fSGOi@wPWvKPTVp6T8;GlTsnb=1 zL#c479zz?6r_EZKfe(CReKKlqn2nBpBwB_UtfYsh3MI{8CCx!hv?tSGKyRgeKa{5PpbUt}1VTDyz!zPiv?u zH7M9zS(J_o3|&{QEZ;`71d=aSC_PDa9>oHrLwxX zy_xhgTXga?98oWVk?048UJs(Ypk637(h(FiyJE4r~X1r_|kgzrB>6?Inysu7y}QS4#xt|9{QD2 z=1fO}$9DZ(*tLf`23r5DV4;Z&;Rk>40%TP|yzT*QaK&57VlZD)Nr@e4`1G6gCuv9u+Lk&}DqD3l@YV@cIcGYZI!8*;B73{3! zJmS1iFcd4gM+=q;XUgQ|GDPL^qiS(wE0_2KNa4yhiG%`{%MkF$<;pf>wybQw&xQw9 z>6iRfwQ86oZ$+MMdv^12748>kHgMFt|Uct*Ix8Fq&zbCf%rF}cH8 zHezNC1Trf>AFg~tQOFNhEQ8XBT3T|Zn`i{T9sbT7;S(cjp}Q(atUvPL6IT}*J~5&e z`m3o$)rIo}AHg>ly1RNCbV?)4^d^Y!X-(jpA}foOWE1!v3fU6B;B!6``O8t)e~vjD zK8Su+@*FF1f%CV+iJIc$=;HJBPXXB4d6%6Ez7Uo5Jv1;qciCwIvMkxV?35+TPm&?< zFxYdu?6d-4RjV&HP1_E-*PG=YXh@P0WtM1aEgnNmt%Wg!_66>=F*HrBHTIPl`u+p?)%+?y1H`R2EB5NneN<4xon&t7jcnKz64uVY^fE@ z73!R)jz|~JwQ#i*lL=nIFPAX$p>ixG(^oJZrq2aZS(9W4m4C_g6J73NnYu}a$X$Te z0Z)@G01{r6Oq?fpC`LhE!^@W`e3ZMY{xV&eecV~tW$q~M^efGkMw>Nw%F32IWrgJq zex)>c$~sFvHj_dsl9Wy7tT>Ky-+av0=81;zv=u{*`SUSdqtl%K`+O*_iao`P=J=Mn zTYQv3@lj+`{9Hpsb+u&dX;NJ*-#)#7JxRtEeyzCUOtPhX$upE|&~x3h5Vz-%5THaH zo-10&{aPY2*CUg@;%cQArbbirB5+977kAky<0}h+aw9w(_WmwAbzc-ujX&9Cr{4f& z&3!wL3e7NM-j1VN7sr!x{M&Jq_6;!AC@Wr%?hwp$`Q3($1+hw2%u}w`lCy##qN^e6 z``_S#Y9Lca^Oq=LIaeLEv_qW8?{Fd&;&@0$g6q3@O3Dr}FRB&_rJH-u0TP;VMC}g43_AR36h2@Gr20&d)p^D$Bq@|h;4~pMOcn*th#!F*g zEyuY}eje%;NbuKk^sV#RLakt-4L{*DsW>ZWTY-_(0$k#wj(~oA1!Sz+SqTr5LYg}& zC1*iBC9h^J+GVH0uPp>Sw1nkX@3PZi##^H?(fWBq28yCa3Kb9xm8u&G)qSu+sg9$+ zu8XH+zkITPX!-5YbnSZf>MYId@~0}MY{2O5YgY0IextteukmCnT*R6k$uJ{+g=($8@3NEs8=;;8 zsOR)9JMH{6o}3Z8?KEpsJUQ#^wo^HNG}vvYmcPYQ;~RI|Y1nV^B#M{I;(Qn##kc(? zi@P_=;?wX0#Sd(j#c$uDisyeTsLxQ;XK#_kZQEq=2k`^NCv20&OSgghDuWo|7s7yM zw@@}rZ>(%7+#>crg7Pp>hFhd<2i7WXk=e(c*^fJV13|%c3K3rHY^dJcZ20V5(UiFz zUXkhY5;cU$mTgxiZ^)*#rTvHP(g@Cf1_b$?L;rs)1Z0We<=9p`HP}%DFU}jg4f2Lu z13Lr)KI5#Y;-;_?S1oxe-qI6cqaHgH3oYT6)p`(oG(~=_Lg7enVwSL{CJqhvgg|7$ zyNsJ9vAPlvyt}ZEL{X@5s(gve)+V!p2STVS`Zaw)!tKj8N|ao0=HLTOOAZo!z_s9A zL!J-Vm8%J4Lg~*7_Az~xLSZes!HoS2f&B}{zH3K3C1(&?6ER3r^N9|nGOx1CP!j3?(ayX};^E1sJ6-i`L9 zs&0aKeYCB+x5xX4D<{wWV1SQP8r-Kl%~GrsQmkYtp5G;;Sjkenze};Sl6fa<|J`;f zl`O3SOXGIiY00h{e69))2GNvcGx3dr_(mqaQxf0E#BbWIh;K9zk1ZAszey2)csE)E zs487%okA8=mA6}}s(AMmsHz)pC_N~3rPwU5#4euT4C2X6c=V3lLRaUQ-Kaf^-E$^( z3nWkH!4q`Vbk8MqC73X%7Bn|2YfKn>vv`ZM%N{{}Oc+{(i3asCVdw#crHUo>7;M3V z`a)1=WziC$$2ep`SzGo7n__Nnwd2<;?$*^zXO@hwA6cM8B__t#JLYZX$zP$$fmUnjB zX|sgBE48B3%_#N5Zad}fRmi)@>?SQLz*+$ex>>MSA`jnJ8Tkkay;Fu@>28z)@}zwV z`B9m@^5zQUX#fNAXZK0ui%rNO?`8>IB=rmAZB?nQs??494VG$b20D1Ny8HZwky2U- z7BudImJ4GoXJb9EUl?mS8*9XVWvu0FAz>cBG|#=@1Lpa3zp8#kGV6Y1L9iA3rC{2D zOS5>da(l&G`K@Lq*9s=rGLuaX2qxDulfw@vCf7O%*{c; z%|YfS?4aQ0AanD;LB-8M6E}XzO-OO$KPb6bbnuE*(<9YjBk^uCJ0}G@Cz+i;B|9gX zou-ErJ10%-6iIf5Dt5XYlI%=8^bgo+VenDZ+RV@JaBeV%hoiY(cu1&ccsLr&$a2Nc zaDJRow2I}DpV5jRXL-EnZG+|igrA!YM$*#E42=zEhQ=~OZ(DB^IuVCccamu9G|ySG8Syp5TOx5Al;x0s2*VZp>(%)~c`l~&#|G2xd?e4v;p zRZRT*$fcP`HAv^THM1~PurQTb=z2u3FqK(&_K0F(s$ikURxQOqnBLBec$PpsixIye z5zk`8^9{taE`hkcFJz!BZf{1tNT6QCsDGEJ7cuJV|5iLN5~#TuDw1j|R%+|~cf7FO zgugGXw)VbIib37q!OR6tBL|7TXD+__TWIck=HfSli|+*&HEOPdfsi_y5w8npuUN;3 z&q~DW7;(d+O1yQKLfp;M+iM{8cQPa2ERb(z#l)}( zF+N2{T_l-^3M1kn!Xy##p;9&yuu;HXo!vyA`;VH$#1QQP_-wS`Jr+8?4O7igP zsmt+@V-U|zH#0F=Ffo~#C^{vWn9NL^F_@TaV!|((z$3S8iw3796Q0wTWg=)$FMr6) z#m9n+kC}_Hrv(=uGZ!0AD^+}K;-W}$aidbjX~ji{GneL~t3f)Wo0)~#f`!@4LZ35& zh1tx)+%t-W*(MgsB?~tz7JfY=718MIWmw2FXqR+1^DtlVFrRs7e^&4?pLuxwtm0w5 ziHDRQgh$+^c=+P1u;;HHHZ^jWd-Oc}x<9|(SNXDm0#SeqUqQ@1hb%=O19p5NTEswC# z+6&@1M=Ewt638JJc;%|Nm+nE6`3X71HHdq#AovqG$d>q{MI^OC(h$u-WZksePU$~d z2t@7x_m(|&TF+VdJWsyNYKN@vx7z92FqFD_kDZESRu^R5v&T+BWLfUpV<(&By}k^A z{7!&{_SoqgfK@e>vqBH}d=jri87e4z%B+gVwPaSs<4{Gda3%fY5E`0DCGBy@CziDR z^b|5$hf8hrMAp)92iZ!Y4KKaI%P-v`9Axt^=QUF|2K@aa9C$woi;zXjg?gTV{j4wS zu~Vr`c@Z+byT?v%M>weQM|OE974T?>uMjiEbt0;xCG$68>!68$#HKjqsAROxt5G(>mxw|X2dGU%wA zu@gy5u`&cc+{SQzyT?ukYz}Jr1E8EGd+b!Vwu78o_SmUgZ3s|>$&jn}gZ_}m=M6T; zH=JA^Z_wZ~#ck#{UQ6~7#e=2swH<_`Rvw?1k1$X?xLaFW$q^yH~ zs=;||wvqrx52jePW-g*B)}N2ZP}n$2vtB9^KCVFS_)M{)jV4>h&)fii!qE+}r~!o{ z=`&6`fh5cEJ$8CbYAHj8p!7)uku$o?PL4Vba>kX}>G3)avLuw*=_v`Oz#(@m8t>L+ zcA5n6j%9Y*SO>PUK3Imk#o~0tXE-cX&3uFWX31L^68kGjzRs#O^AN2wB=4&t6_T%G zxvEj9et=h~SP@9+ZC3H=NG;hzbed02+v6OR+*iex#3}bU&8Mt|_7+G!m&*{^JA*+D zw09&9PbQ%1tinqq0Xoc>h9W6{H7BLg z2~PV7X{AU*BY%?94ged@T&-bHDu@OjSJiir;G-ZKj^3g^bZ32|%uWSs#IsY!;k#dy z*{KKeEYp_TX=Xp6whc0b@=gNm`{j0ewZ4O#tC!p9d%#yUgFG_UH7~c9J7DNPE+;z`=+)7JO)Y9+LRxND;^PG(u@{p6P2&NxvTKxN_Yc zf|(ctE#|k!K5#lH3Fn9wS%JoTBOC-K&Zh4yXxGL!bC9j5guAkz_(8Y%NGV>6Pj_Gz zl0|><17#7-9h8I~=y!gi>@Gw>j{fBMc*OZQ?VlF3`On|vpd@kjZMy|W-!>s3J4Cx# z&lxFjtZZ6g!I#m$Nx}GTy#?Q3S;_g^Q{dd`NMb4J0&{y$D*TVgmi~){l3ZE+UEWML zE#^9sl?sY7av;?~wqgMJ`q9r^FGDS)f~eSvK4l`}*&h0j6?#a%*+EI_Ywz?kr#*Bt zbn`1db<&r<MvYqu?%rcn7MctWy-k>ZM0528AE5dh%Ip=>v%Imb#AH5 zN1GY%w3Zmf1H{*lwlcq~T7r~yQJI~JHi|ZV3Mx_SeRe8u>7b?^_hC!i`uILOl}do0 z0r1>DJKc2)I!u&J+$74bLl(-8Xl0b;k4sD&AVgU%Lo}&d0er_cJC(F@7$28txXn)G z5-4_{CF4B3gFTE>@@Fr0klAd0Dc6(GxynP?5!9C0<8UD0oGZD083Y3pEKB( z`|LDXX3vx%z@7kD$v!*%daHw){tksS-nq|C*W3o(ReV~PuV)w^q&1xpA=MU?8*KtE zc57O)pGfmW*r?xa4oZGjMqazkLCpfB@v$PJa6W^Np-?V{)fM03f3l~?8%pl&YaZ<5 z^1E@Buctx5KSGGBO>Rs?w`&bC@^j^;(y$#dq+%^-?yeXLOAe+I-Trf5ojwN}!4@)h z9DML```ghhlP!G{d)FO^-Fdr%lKQ989l!O}>FL`YWGg^!{}$95DZ`Lbf|N`#-=2Cq zlqE(FvA+wbf<%o zGjsh!cSq*x)cP(AT3o@lt{gP3S=0vIYJZuU?h1rF7|6S`s4X(~-{l}%+HV*%czX8< zb#Z$s4>a@cc2H7|+uJ+TM;ulB4heowe_x1r!J~F-2PO6KctiB~_dmwa2t;!|SrkC& zg^0Ow0&Z8=)YgOh?3D4lg`mDC;nPna zu+yY_9OU#Iu+ujDc=mvu+N3$iIp~0$zDjdY;};Lu=^%cLJz%F_+5nN^_e%bDx&DrM z-pqz@Jk?YLTngw(k(O$UDEVgIY4#IUrY!YI^D`d)cq_FGE((VlC z6xp$o>xg1-k2#Dky(7Dah>Y%7)jeD}_j-fw0NU1W|La+E|64t)a*l#Rvy&{nA}3j@ zJsi@htfO#}rEJySorIH!*!zgFODxqgLOT7@Nlho0%ErP&%ekDbGp0O$Sl}N&aY_zS z(nZSeoU*MmCPC#QX$L1YzZboc(H&*>@%e)8bo8qKYgPU`t%`D#{3&LKc_~bFNG-TS z3X>hu&DfRM2sEgrLb}Rb-MpP2Q?F|j3ywc z84r@Qm%=biqNE&ph4H+3A2%njb7TP`eW`#W*WB-*q<%DxBa`lT&<()>E`N|-;{>w4 zb-+#qo5h$&$C9+=pq;kg?;z*Z2kq3ci-Vdb9JCWTZ#ZZtPZvxiQKVQFsi%s}?IMfl z4>$yH*9R0ZZ4vn674Rg$0RLJ10Op&!!j;v)D%YLmas6*p^;<;g+2kA7Vt8n7E>_nX&6izH+I}pnx8C#$3pk(23G&gL>`e?fLAvMF8 z8}`T`o%%eaT18QwtSt`ODNVWzUM_Y{K4ho44>`#B=^;C{e%L|IIfv{t8b9VAveP>J zSbE4#*LH))p}1cbzXcwL;(fZw;(y}@ins3WATB-{KTv#qcU8Pt7U!eTDBhrlEdCjO zp!iYz0QGwwaS#_D^N87ndj;ICe(A1I$Q|$o|JNMH*j|xu~v6m6!=D1u-9IR9EqZPB*!wkRUQS{q?VyfxO%FD?M(oyZI`=Tk~ zF$dicaCx(Qx%9WTa#sv?Nde}`8c4^vXt&24WGhFSFVx2!pc9<>!(;FvvOd(!P6b;< zt9cYN)tqj2+KWu0DIL}25-31)1rIg z$gPO5{iBx**U9^|gKUK`xQAlkJ|82Y1POh~hJ?sxVEv2kgY>9)xbG2HPM$o?sQAwR zLh*0f5y3}WUVcXIh#)fanaVpNyqst3w1^wW;;vt%rBaW#cTa;t@j^ujdpbYXs!&zi z_R&P*ZGazZVcUHTX!GL$Ki2y5MV(q@IVf4&$M4+zfi6Tp^4lTfv(yMv6jQGT{KU^} z{FLP&TgEnSGgolR7&jJ>xAT7d5q@su)@+B|`98uBBM|q}L5{4+Ml%fXUi5P<9J)J_ z&6+l9+CSVM>2IvKvLUK2x%LanG`jDNgaKsAMx?OiM``s&>2%s(X>n5&yRvoh4$tyRg;*CK@@&#FA-)v;9bzpRrZPqoWc*Fzu5Luay67Qc6XNSPxC=kHK!?ZUz$Tp+O{7G`s(Y z_0WH!ETyE_U(CufuLVL`6>EX0gcWOnuaD7bcYmQQv9@>L0O3s{mWvoG>!q;w#^_Wq z;7Zp5e;>eZB-R4YA)?j-pBQLb3*0iW+FD?s+)inGL>tI!fu-ekx?zxmocqh|^vWOy zHT}CBYk?Qa?KFE3R!dOCFN^S6Ac~w+MV@-jK@2?MIR#uKfq5+uz*_(V{B53>zyqH* zH>Ce;E%2WhP#G?9li7ggbubuEjXzqj0V~$Q(#Pub-SfhLV&!VUU|~QJyLt%Lxv=Vm z!B_aus_S46AcYM)H|*)LI!zj?YJ(_G*671_Dwpcwbuj;7JM9?iAm{Uk?c{yILCzNs z+i3xQymHu1VfhYnPCRU<2lAn96i+M@*5q|C6rThb6u)kmEbhe*6kmWJC>}ms75B^H zybgxqU588Rzu^a{H+j)PT-=8rD8BGTv-$o{>tO%HipqNFzniUSUIT&^YwionHT!Ei zMUD_w6stbN5$i=V_BCQ`rE;ywYdV#W;2{N{WDsjWSHFbAWn~yrR9XYN|E20{Kq-4M z%rw@3UPRGCq?y-%j=u!`i8Y|KeZqdm8c_U5*amArewp?EZw=_5Xj*BwXtP<AJIvpKZdHv(xqiS6L=rszx#6GB0tbe>S z3SB~}Na6L5T}Uy;WdC&iQHN4P{&urEO!MB0 zt0Q{GIGqNL5x&Ax-4lrPA`xl+3ffhiP>TL~oKEgn9Ary}uyWoz843AFkn`TNuR!vu zCxL68jLNe&Jq^BTzv92@Y&3ovzUa;`)ro0k{#XYk%h_noc%4=h)HoYW9jDrPQLOT8 z)H4n%$_Ln|#ccG(@tBSN1#8OL=*IDtXQNFf;4G`0jrt~FHVSYx8y!D^<-h2}zv63C zfq*YyNL*q_ZkqAqQ>sTM$QeH(P2Z?I$iymi%x981`T`Zp3H-U0RxNuPc=P{K`0+yI=z6MwN}oIgh5>srH)= zvTSS~OC;$;$PiF3L^T2NbewE#_3K+)47gd_^b&1616Qw(JcMl9Xn>yNl=2FPHL3M3NhD2%mBb#+r6G2PEDeoFwLT_G43;5E8fu-X)S4l4 zvt$S;AYrwx0e)7i%Jx?QyQEsh0m`1F6y-m3F+b!LLvb|PT*G50L16`Eh5ZedSYahn zWcVT1@D>xLs~k8e=#&g&NK*BmJ z1b)_;%1*2Tc1fL?Yj{Otj5;nWcvDfqQNNWwW6OoSTD6w*g_<3)&)d*z%3+fuE(6=F z*9@r^=+)$iW2F=YGOQ$Y_#~y&0$CdRkV>59$ zXQis_;woTN!!2BKAE{|`v65}5Dc?|+mh2@e))v@p+R=3+)LWGJf)>(?mz$$Ttz4Al*4u^K5SvaV8FTTu$}VWc93)BVLSbd z9~%za$@z|htnokFiH=w(nuuS7Bg=vnc3LUf-y}oi;^nt1k^^U_4H=4$RFY*U%ksiI z$`m{K9x@t+RMK`z{$< zmAJ_cN*+KoJZ#hK7`kOLmaRz}9%gdB;iB49eWL)QeIDO8=l=%!wBv?Q&4&6@h%|f$ zA5}1gG-8*iipqJTfvl;-rmz%M!?wAiN2(bc&kH(K`ZiUw>|6s=#tDPJWaE>IJ)s8U zlZ*L?@uf!j$;Ce>JE}3AS>qQLKuJ+ZJXwj7_@2VG?>YEIi4nE7ERLZx!~`YrJ%#t* zt4_%us`6&*bKVJ+*;1+9`%I!$d*KrNIz?k!iYdzV|?O#47~>exWZQ0@)L&`^vi4FBV4!94#fu|jwu)6!?-zl5r(dO zUCQ93;9L!w=~dT84^5Ga4v2&*a}>iPHDgVK-|27f4!JW!9&hhuTYN8~v?{^Pki0~S)2nariz%UUtz;y6t2{49;tO0^~;g20LNK_3V zoLfPss`?;y2X7WbF%C-&IJ))o_`CUCnSxbo4GJGxC+Ve*W;7kRptC9=(`nkQ`ef|nOH9|10ZXs7d zn5HT}jD=w;@Koie0;ZiWH`m4dRCN@F5&EehR@GBXKuU>KUBv{_&X?o!GO?=8Vgjn* zx~jXFfGXIwL|3lC-n^U~Rg0|Q<(`Llw2L8~Rs|AY$3-@Wh>6*8cUEG+-P_g6L)j=F z2`tUKx(9U3arFjoTCr7YgzLe?!`DW<+S=LW&C2n3dlPljZIqZzORNFiMzXZ?Mp?LN z^9S=n{=5)rbHcSo*+eFw^nk}3YVXc=<>iEeq^t0|HPXTjz;S>CbICFdu5-pFErjbx*fzA+ECmwS@dKx;IR=x&FN8u50` z`wkl=dZ{5AG>R09c){imyZt$?Om~}{9Mbk#v_?Ux0AEg){N;5I5WghnCb{qRh8}SR zNE;bt$lb_Kgli^dW%kE;M&ib#QC`s1TPj)N^STExo}4^4Y42+qz$LK6B>_!@#XyT` zAFNEfcwlr*v@ye}nYKd3Shg*cwkCabS2 z)9npioYbP4Ny)20Y%znJf!r6!atB&m9O?9$kt+UcMjp;i>+yO*Mhz}JUwJra{bIUK z>3@qijppE4ksqh)G;xZ9ny#C!6E)sFU8loSu&-8uO}=fVaOGc6IPv1qcaho+Ljx(z zS=)2MJuRZ;r&&EAS1+0qJ|-oa$XY&Kr+i6vX&8~yHbW=pM_2^CW`<5hGHV$yO3Ka~ zJ*WgZdG4mo(U2&z<#ZJ{bsCfZNaLrqKGJTCx6%C{IS6YBAMskk@Q>8W%tvAYA*nx2 zY&E=&Csdi#}z86#29~M#Na7S zGxIiE@HU%yEBIJ(Gn=_1i}kHIvdIE53j*hvz-^x>sm?P2g%Gc% zp!EzIg%iWgswoWDlHsjEqQdao>e=W?15si4ym&k(<)JD3=HU2G9F!Dr2lH}5^daYM z{R9V_kK(~1`Y1fBo{f5c3Qy`eL#KkH!nZ%wh@6vV=rrb22Q@96p<}UCxU{LYdjwBd zp@ml56;TNEAhy74lAmPzcdG5s>J0P?OF3JNyLL57Qjt{k3-@xch2`x7g(GWH__-P6T?o z#-BMT={fo!0t+sk5D5+P)B6z^%LP7jh@IK9lHtS%!36xS0qQ08Cno|5JHHV!*CIol z6sV(ie*gN+L1K$RW!J3&M&d(>#nL1xz(3<5JJp%)ApTTW1X2;Hgx5>Nn60Cp=P4uPI(&$KpT$#S-^V5Jo$`JB?i|&g94pwj#EDSyO2@fgpZ`MkO~<)5H=2u< z%{$RfX;O39(A>C*cKUIygRJc*+Ntn_g`oR>=s4UH?Q|Z6$U1k1P9?Hb0HuDOq0=q% z9Mp6@+QX(>@ShlY!AW73-T(vg+4CIae0!pucFmJiN={axvh+(3wIoio)5+RG6#AD4 z!-%LKP~ADvPNxCKROpnTl7lQzN&QM9Pd`-!d7rOIw~sN;or8OU@>m$?o{+f{VFVwV@Q95yH(@-2Dj_nDG>uRiv`6`5Ho1Ya{OU zh74xt8Z%SlqnN4jQK%zbzIISDR|*;*g|YjKNRS=Vc>ZMg_ec@-M8cf?T8#?EM_~X= zNq(BZzgjPJkfm&fP9vlqiew0nsf#b^M$FVHZ=r*nwwXH3#E-f&b=rX+4QJ|f{UQfB z|2@W*EA^vfb>p}HF6P)BK?ypkqP3r(D6P1qJ2 znmEmxXom!;36b&=QlutC%8!ecCQe6T0tHQ+;a}Eo9AtIP)G0$6Y%qAvpQ%&YHx6=+ zo~hFS{3w{I(`@{BccxC|`0?pXomzbBAnUxDIu*#$K9pWHQ>W0k4svdssnb{A)+{mR zc1p}dhuQE)YO&#u)Pmvv`c@1b^V~#7YN0b~@*hJ1M{2#%F3_ z6f*>Id5J01#~pCf*;*L9d|8a6#b+#(lpE|#=Na3kVw`a;N21>sq>&bo=(q&U-;RlP zN<1rE{wM@JJ<(2!mpI6J?z?bGKWia;gl{#5e2GiLDdjszZYhPA*oA-tWe7Jr4h+X; z>NNE`2Q@u6Qzx=qm}I9638P?vvo)I3xVP<8Z>fWtI^M=plr?x!(2v!KY(t;134*Ms zzvu>vg7N;|dy?(%M68z6R z$mg}`2fc2iWWjZ2408pBxs2hZACwGp`8il>HWWuG=1PHpntrjswNLzjv#EKx4|u$J zL3fZAaqoW9j}A)m^`l?7^uvhtapC*$d2U+4X?aKs(tK9S3`DqfUCMFya=dAYIWPUp z2}hQrqMYT?edPOFM_zPk;K;l3d?LU;N}CTPcz-4=813m#o0h@v%f9>H!%t zz)l9Z?PnlLIWICEn4`+!rGhJ|1pjHchPwGF&*{3X5*J}a%(qHO^(oI( z#oOJJWbvsogojT<(}cIX$FGtvpRZlzAb$CLvIJA$b%@A#`8;;DgPa}9?DXJj=!4WL z_lw=TRqJ0$gp#I8S3r5UIw5>JuZo!-*2D^?@vvrku}-h8h8gA4nHv#NmuH_^gX8?B z%d?geb!`3ar;BweEUD@8>?WC-t{#?D4{JWK7Oo3tsQj?z#<(y8eN2RR>{rPC<Ynz1ZBDM)JDZ5BHgA|s1mS$h7Q-j|s$u58@uCaQ4bKnJR zaT@B!->S-YxofOSN+n%lOS6>byQmkp@KWvIQk~BKCZrU1QHT64q!h7vh_OsdwX;ig z+W)(1q?T5Bo3HU7SQmlU_myw+J&u$VvO;8W7j^0%=p^z*lDLa{2uT%9(WUO9Dv6b( zv8~J!n=h0?;))BU@ac*RrKQVs3T+k=iz}p?5bH%Ub_Ow)_@%Ie%XDh6x zqAiQkKHX-v!rU;`E2Od&RICMCq2g9(_76IxZxdD!_c$ZB3%?Vw4-hkMg}%H4EAjju zABoq|)+2?jG&d~tgH8>0D!&xv$+~uyPQ_AYz7=|QmQJ3X4szC-t)d zsn6DFG=5w+Tc`EApfVIMm&N&3D2g}OEsKAKA1MAeexUf>dmO~YNAFR^)53&B`Bo^3 zuLlgMKV9Y^EnP zA8Ec&#q(Z2?}HMEpZ6*b7g{u~b{<3~U+pZHSyyja-iS9J3%XB{z+N#-clN$b{ zQ=R=nS$se92}IQW%yEdYS0*+4NvC=HF@NQ^i1>Y~{rhXYpV{~T?jc4%FXDdY;|I`8 zW{8wjbUhy-#pw2fo{PLsr6g36(cNYVO_wYoVa@(fS6iXe!2=kTsY{l*e>te)l4Sv6 zLMZXT(vL`Bq3d3?LZ|b8VS~kZV5!wXTt0;;nH85TYaN1pRd;^rOO}RAQkKHjW|>Sk zB`+>h^e-!P8hc1sEgJfaDHn|mN7^F70z~gvq0^vpJir2uXoj*I}?v z5tzT3r-tSwIvkl)tkOM-WzBTEkV|7%%g8S`!IAo$0%nK(4$ISah)g+L>HJ>|0`CP8IV(2$|JkpeJNb)9SK#$4V=Cu?|0^<1%XX$|lEkU%@CeY8T$M~}p0i>%?Bq2f$QW?+y% zyc?6iXaP%P&_R&rXT&!nN^c;YY4N z;0^{;NqaR?gG67|kHMjv*CG=})P_i}^QWg0dp+Z!f=ImcM3Md7-Yj3>rry3Bm$!Fw zU!eC*18u%@s4*x!#@%!E0 zEYkjp&>DHXLMe^BQdxc6t}NnvAB{pTPmcJR81jW&Idt%HEaG$+oQ$;7;e0#_tToCa zV*O@$`g^iKIX}Xi8cm7#e1gaA4bdBTcuZo zf_d60%7eDa5q4cE0(PeT)p01f5(m-&iR(&Rg47nbk(T!lDmE9e{<8u`OOTJ2z^JCZ z28G;wq^fxrcMp#@l-i0kD>rpG^Ge~e!(|kcwo-Bo;VF z;(MBl^J~*@=-Pt}J(6q7S#7HJ(u z)F#{C4od1u-)pF)?GOpl58Cr1YSRKlLbR46hyHfZ4Qqzc*DNpHu;xYeGbyWITiU|O zNk<)YLnm6GO`M8Xf&CbMd1x`Co^sqlwj!kTrEfUp&~c2&Y~@Jv(2qQex$6W*(J4_p zu3pJWFLRP#B(3A5Ur%7!0x5lI1E<_}l5u6!;(}W_Z4%OKMIuQoT|CIjHftRXTlm%0bq5SLu{uv2vkQY~OyjN~dLjU9n21 zJ%C-eN~h@44zg@srPDM?WU&lEtb|vZSL@X5w1XPQt=1|1G_-RmPxJS~6(x++3>WkN zD&!EZwlqvj&delQ8ur;swQ1yOJpVBe(Xy9n(>I9bA_mVpidev#g*fWozt1S&Tgq{` z+cLhTI{u7Ll`>o!RO#bMIvbhC;iAtQ)^p0U=b@+C^-F~0PpuVU zl7}{P{#xYQN>C)5wlne63s^a*zgj1X7G`@K>hHE%r$;V0sPU7lb$S&)vR3Q#1p~uL z(j{JE03@lLNV${*~ zLWW9?Q37x{#^3x!bm6yWrL>*bRg`!}vDZ{f5PsVgKfqp4lk82>Bzr|w*;@oy`0by7 z1$$bUWUrnik}ShY>@^8X5PsV!3?FP4>=jDL12RMjgS{bPlD#ST0rq|b4pgfHVUoSl zs_b1AF16PpJb{?K9^sO`#ga&g3_;AORs#VJ?Y$db!QRSWgx2cIT7wcjl=jww5^DW% z`~Z6i5mI|;5t6;w7~wvKaWf+%dm{l0?R^#@+1n|Jos=Pn8SJeDIM_QB!R%e4e`)W| zboqmMIj)e~sP2gyg&?BZ9}Lry(eE7$L%(-@WCA5;di;Id0bHseI>+Y*JDmI?cbfMJC{U?eA!F2i7e0bsYiaP$!Ri#v_mx_%PW2AQAj4LOms0$ z^gr|a)T4n_GSS6|V+Ujzv|6W>O+o{GWC+YJ0`v6MI<=IEDie=F3<2gpR_krrHM0yT{qhB#U3uFl15d$~C-0+KVbTN;{B#^BzhI^Iiytg|zCV|LSE|TVQ(vL{u20hUxs_#5i=dZWP zCh0!>K%Mv5WSxJ*ChL5`8!}y=g-+>o!`7g$XG5zXwj(iiM3^m*{(K} zM;8AYo!U#t(K1AdufX2VYjhe^JAs_**XZ;N}YD;u9)l|W=| zUZPXElxirXda^{P6;~xt)2tGm$eB~3({r&2S*NwxC3(l-k0| zFQV-H5}ghLW@Cv?b>b4J@tzW$I>o^P7kA$pWp8Q7oVLv@`{YQLeR3pZ?;9s%pUi7E zhU_8}vcJb*!>>*to4+>8{yxiId^Pxo>_w9Q_rO17KYn#3**~c(WgmZyl)dDd1RU`$ z(Wy_V;B1-<;Z7d_5A@)$ClKr5Zv22AGVH3LDzKrZf-eFd3j72=Q1B-dyo>^|qLo#7 zX{)87y7cXWQQs`pGAJ0W1|ChRsBC>B~ns0I*Adr=_);eY~@00GgxaabqEl- zR;Q`bCW~bV$!CTUHEz9Dry+U*HSW4rr$PgMrGzV$A>d~z_@xSdH{dTLXZ1>wVu)9= z(=6tE;VJXk1+4Xir_5&|4Nno7@RS7%)+j!KY-#^uPx+cntAz6}fYISRf5#`7E6zs_={&C|NatCd02RV{n(PqFX2=jKSp+l4DfKCsrZ3{KNV^`&RYRx{6-L1on&5v`@Q*9_!3Mm7n`eUGsNg?P@ZSOc z5}FRWa{W2(bYIS(c5T}mS}iOSvI^~P71d-bt8iDt1pZ>`R$lOKm#9>|m32xtw51&k z2R-lQwC58O$QB#Rdfvx+E<}nSDZOYv4~yu{ z^cMr?iD(NMog|{EGP+1aQ)KkdMr=?F2scOLM0qm_J;p0X}h(UwM#H`=~gT2}~fo$n@*to;^ zSo;@9DM5-Wg@#9ArS(kX1WIzH%2-onU6(-4acgy&jvsHY)oB%ee7IJp^6L_)@n>swy1GdMS?8|RsX&%KhOyh0wL0C~ zL~6eMULkXL8A9{hG!m2NoF<@8*5hk+DwZfBz`(WZbehs6ft=T`)9L#r3DmgRI-SZC zIl5XfyIYZqxjuocSZqp@IR=U5*9#JmF<)jf3CNgry_9hjen7_Q_yHMLT`y%UzdnI1 zkkKh6P!d5BL*A>Cq`WO_fHIhAmBh?k+G7kip#ws7!d(iY*j);uFcg5h@LFy`6ugJO zFv;$alY~*_)mHx8x4ilAI?`h6TGdOp6BzP)r23KioIB*AH=@{elDNg57&UHWZEA1> zBDfPI;y>JwK(-Q;=u49rM-QivhYgr`jW9~QGT{48X}~2;X~3OMsmql8LbUEOgf0zZ z)@mws*+$lPa4Zp$G;QKRp$B2&kA;apW}6x&{+KNe z6Mw==zt9vN={3^CpK{tQHzts+6lwe_4o&0K$B>$4w~{dNXHhWmYd6BGG?PvIIY(e( z5fA(uGfI@`OS2eBEq=aZTBgjl<#S__5bZ&^|G2lofvm^L*;Rv@tf2hfyB}@CS zOqTZVBjqoUArxuY{}7;1nFq^UGn8TfxMs|T71k@16mJ4WSgdn1xiHnQS%O$Md##yt zs;s|+$_L62q+vs6^lF_x2RK|~17o+|x=yE3i97|!UF&p;Zk|9+&pMr&;m5P-08=a#p#_ z%2HWNkaY=-;zR~svh3i^^kupKZ`F$pQE6uM$<1Zj^_{9NM&u55BaFUhB5YZRMk=36eA}$WWPycwly}NmybEmdo==@7C zRS7(W!HX}!p&7g-L+5K($O}U!UK}{r9mKV9@O?SZEU{|o@pAhmTQ82>?PUk4 zQljt@CE6K_^9F%p%f$(F@Tv`fe?X(X`yz!(<1S9A8^04|7W@)Q?dEo8Rq%g_YSPOU z2)IL-2ZTKS9FKgSZq5xC=R=n0*2j|_VuRI-W*yg5346P8b6r6kn>!(%R@#r}a4!yz zu^xk;G=bB>tB4}FNwsJ>zTRYS`J&Z0HCP`1PmaBzcw~+b=NG&9diQqYoo06^J>bsv z3?%KBFs)Gz(fRr|YUCGb{L~Ib1_yaFZ*qlvxt`4C&KuKxd0^jyk1#pQ+u;dB(T%`5X8lj`4X|rRfkw>u5Xn-e6{7URa&zu+; zO^Lnyl!0b5@mb15NRc5=uDdx7jy3Pmy&Y+9TeU_ZqGzWliT!z@wt3mv?f_}CHLa0f zkf@p0Eg_fJo~CJue)>#Dr0-=ck$3-eb~clc2uYPkkjs{PH8gd5ezSf z9Mx%9D|Fn}(MNUi%e?zhYU)v)rnX8T=ggxz{ej{Oj_On>LGBe>BCNv-f9U28$;Zu< z#3}8}lBa9QIYjB;KJL~8!lT9MP{O^pCXi*v1F>{KDynl^Q4H#O2u+P;-%*`b%GAv= zL~eKR8hcEqg||YE2pF{nhHxs{9T*UB`BM#^N(|mKYk1dKDv`2143yV)jioKOCQ#$+ zyT%gT2HY&HxaHj0?aFco3=IB91iQi?0$MUx-2mL|y4w;cIYbmd4SWI-X)y}G*mvEL zK+@QJ$!`+JJvnZfzXtIfj~8(`L&=>9WTONtB@J@B0`xr6EO%j}&*u&Gq31I`jG=!c zp5+=Oa$mU%n|v84(#MwnOXg^?bJzg38*WC$}6m)0G=;L>Dnfh1+=;I57Xo3c2 zr{0}Fwqm4u=nY0w20*eV9MdWBTH$iLy$(ONMZ*v21bJ zjv5NJBZY0(p|d107(YN_f9Fah8r>^NJbEu0hXAQSR==p{j_F1_P;Rsb&R37=^c*0# z^{8l>WIL%K{f`RnL<_P+Yb!}*)xGsvV6+%wY@ydvJ%0Qf- z^mj?1ByKo2GRkR)%cgfd3ljAJmbTE;{>t(D5O)Uxbc|6abWIS?9N`_LRbdtHAYuJ$ z6`!6XTY5t)KaKD^UqgMOE1J!Gq-6(bn^rnHnuc@*722-#8XZlA0I*4Uf9@FXjt_jrm&tp1W^WVTDjg5ZP0>RS}Y|= z(>BGa%%BWaa0YE?1C6w4NKzmU zJZtZBn${xteSiOCpJ%VN*IIk+c{nH5EKyaCmJ!%v)sMmnN`#?D;5w~D-m>d4IvJnP z5RSl|irC27jP7j(yDfoDae~<}8@1p)%psd(rpS{v$thsA%Suvsv1t=C^sAv~fkoVJ71u(3Q%`^XWMKjua0|pFb&s=8pqhP%%(Zi&o>-Q!A-3ctWGeprU9!Z&Hb&DOtN^x z#qmbn5|_hUMD_NwR1R+uF^OlJnw2|s4oxBO#|ztM7+LHB_DL7e@9b1$$(J^gZ|R-u;A zx{0RUY&kR=kZx@C*i^_j`PoCiIW`r2BQT^B0J?~vQroUO|6s`!CL@J{J z{sl!;;uBSTd1Yv_^h!*dq~c@}W<@5eNG|Wllcr!rbh7T=7}=n$h{j6Q@`;fNooQJ# zR?5#->(|?8zajAxWmTxZ&|JJky&m^AziN$Fk8D}-Ppq00tVGJNS{oX)chK{ntofCI zjuK>N^x@wcaox{9j2XheD4HqnI2@T~2mj;>v-slQU)j|fHc$%VjSg0Vzp};s5sk5` zJx2LYmq=#zfNB0yShELu_bvNTcSAzzt)8>N#XQUn9?No|B5CPKgdQ1eEA6bja+~aa`z?=s@PVsp!=w(ShP~ zY2{yBJ2)6E@kdH6`!VamJrDm(8LJ0Ce$*{Qy`@%6c>=v%5%v(~CueWmzbu@LF% zl`%XfIrE*z(>1y=Iwcs!Mu~jlJPydYi4G+9QltGwnm#tsfz!@6Nrjs9lu4>0Nituf z?XA-xGrxsw%=~(?{WZ_5R#rxfr^IJ?o67Ju$?zxVb7}`{lY+>;AQk+dB{dz>8QrEb znj{%0Q<9uIq$%&v6lUhD3psCRah;&aBXvl&_sK?!^DffGHTxoyf2rnsNQdO#Y4SH% zY$Jn8drXI9Hh*w)P#z^C8P=h2=`Asx4P_kVbS> z2iEGQ@n$`~O5P;V_4q1v**v`jizT}I1YL{Q>vDKihqR{)_jKr%{u5YvJB?3tSSph% zOw}2eCps{7f~jtm($K#3W6Ogj0V{1$r745jK2YX=Am%Tfz(G7v9;^U95KCX72AO~x zAIME~T?IZ+`qHEN6_gGa^LA%1={`P`Y4ym9IgM^qQTruo8Kb1y<9Jo=E16f;l*dR) zy7<);%-310(;A(fpX9g_c2ilqzZUZShO6Xb8- zJ0rO-V>+cNtfZ;3U8*K=OA}bOM;50-yJD%#>d-eBHTO=JCdk&WdbePy`WH02>v*1v zMk>R&Pp)MTFJ^)Mr8km6XB?RoE7$BDvw}6;>8=f;L~DW<6uk$77F)lzwXT!3jbO^O|#qc*w`C1pP-6AlLwP5 zr|u=XoJN;W%wJqIJraxs23Ae7d{ZA&=BRSopnSAJe6*-U`Dlaq=$#U?iEI#?bc@^| z79l-^dO1h!TdH$YFNC5}&MD0iSqg}U8o|o~mga_57VuNQ6ZxHA&NuETHFbU@!s9BQ zDQakWRkX~K5uO&Go4qPGdqvhumaM&E8BP9Eq^1Y7$$ttV9B9vE(PzmISI6tQg;zR( zTRPI_!bz&zzeTPVE||n_KLq0zxlO3+^(`Ga^3#F`CHnX>&KXX$G3GXjsVL*@vwhf7nD9*fLSYLsT|?#LE<>XcNZheFsR>7OTxPRExWJ8kHK!^?TW zMrH3Cp--(avlmWR+56he-s?iTmopPh`hlcZ5*;`?)T&M^{h2Z#^YvaUyLB_;v%RR3 zwEIP!SURPlRQYGhMV-7Gx?pvPr2}J8r?_^%sFNz33Vj3Y$51M#BHZB?E!TBoo4&u2 zS9{V~WD(pf-Z?qUuDiEQrde(*6FCor*?ZS9HWb1tQFPo?UEx*<%fPA3mJ~KJlsDuv zxh(A2q=`3Nnj|;G=#9rnN@5jR^Iw(Ubvk6v?cuErO8Ay+E!D-EV)Fr}r*M^$V65Ru zCg+z?&M!GjPmh>YV}V-G*^4u5NcZar{c@RyO^Bo-J;uLv!g87JyvYd8*q6%`XMdDl zD9oh00@J1WtCxRoim}=Rrn5`>pQL;{M%FsAmRDAW^+(Dcui}M`K1`1fMCH-bv=cuN z4W~>KCvIcC>0^@v+1ofD`JQvb>8Z%rm*~K->8Y?KbabHbB*pR*S<*T>@FZC>Iy&&f zbk7%GwO>^@O59D7ri zOlo!s>(MirtBYSvfvd0b&SXwCI5jzCW-2l{H35=xIyx{{*UdFLWDO0Hc}PbGwo#6- zZP(;A^k<5fmM^FB63Mqlcs4E!4T*7YFTTMlTE6>dQfF38YroX z1*ZjuRD@$Q$gMUkJ~ic!@>kQpJj->i(Wf`d$8m zSUYS(B9U>Gl_M)dlVM9}XrI9}YJl!SYzd9+Giru$m6b#N{GYC-_SpvdOKG^7$L99A zh6Y3aaw!wo5>OoHQoK^svY%sKCF)mQQI3=WH`=V0QQY^RZ0j(2$9Q4c8GY2nxJYL9 zpz_LS0LI1+M$bT~1gP!By_Htn)*rSFi9{+R^lN1V#;u0YGXUgg{)IoI&)@*B-9<99 z!&T8Twp8QR1cTp6dE^8;(xlemj)aSP6JUcp{Kdi+8I@rX@nnmLzh#E8tc5czisc89 zAN5ZQl#ZD+iGS~GJa|}(ogqAQ7+a+LZ&XQTXqx&i@xaQ;au}VBp7H+{CD&S_*<-`y z!5EAO)Ls#%Gd-h$uz5G8l~?1u6rCQ7m6VOEWYyvAkFU6k1aVf+2*Bw(TDW;zMFuVBcz-d{RMn_@18kelR zvV?y(B#y3%lq(TEE2lti?W9HV+Cd7RLSf906FDWw&q=V{)1tWg5Rwm3U-E+SLYx%& z$~OKZ)g7A;A|)2_hobUkyzyRK(T9Nd``Xbn7}Y99`zr!EXVdU$2j-du_NlH72dZ9G zp~-<*35=Io>UPhG>QhPR$Zy$1vordrE<7`PT%;;g!VPlcA}Qf{-*k!F9c`r|7#$HB zROXNPOJV{2VK-y0&0sCyC99snP{bda9DtguwN1;+$QF$8s?Fesfa=_xfrQo?4Seetmv>s&iLcs4xBREoLz2I(}Xq( zGWy78_t~wfVaH5gcde<0v)%278T_aJ?T0Nn!H!F2tJl7^AIA2voEkzfwySPMZ7gg* zjDxBAXu*1I(W^S7yqyU=*K6&Uwx*~>kYz~2;&_90DT8*2K}XF|Ypq>}vG$xfM=d&b z#T&FM&Y*X7*7oX<27USW2B`tXG@+q3-jqL;DSwJ7Pg40ze#q!grO{r~S1njDSH2G4 z=?c!2+UDFE7u17p8kVBMC8olz0Iug(ix_a=iC&@3#jcmjU{4XLp$OL-&I~*DUtv~& z>t!B~bY4kZYVe<}hR#ie&G%I|)c3W?&Gn_6`HuRk8>Y@pMbf!nbwjhx>Iv5=3pne0 z&JJW$GTGcz^s4!)8@BP`>aV)t7p8l*Gq6{u+pj~WyM@x5Is?bd);Iu$Q{t(to>$+Zf+>kQgWElLP6kwB-^HEot5;rM{ z=`p-uta-W%#$KuK5ypK6s!8*U-D$_f!>pIVs@>Y3%?y6MS3Q`j_0^)?gUlHx$ut5cD5H~HIe zv4d{l66uu`2L@lwhA+~xw3IbEq%=8WptS#((gtg)$<<2Q10oFv>_;rujtf-BTHI^8 zwQ+^YhpJqkD))XW?)~y=)4iV>U;o$zY1f#}{8Su>bagT4bJ-=Bcnx3RteS++WOJe6 z8ouR`ZDW1>!br@pW6gq8*{{JDlKS$6+^SLxUmJ;i?8sh3zZ^nN2E<|!%W&7N@fr9{ zW#Bi-z|D)y4E!d&+}p%{CdO|@TBaRC7MnTxO>BZa+)p+>Zjz<*n#HNGH#3H{<~A8N zw=JfT(kmseRGJ++WZ~^+m&MM=Bp#)SZ91bg9qrX2)BX;qKB&5uz2cW&bVIRARn0%x zdTfESQc$a7InzrNE_0-I(g0vr>R1smOTntQ6d} zgghcy^Vm555bX`EVkoCs$8@F~CSQI%lhf2fkJ5ewNz`7plq8X?NhvLoTg8ZE9V6d= zr{p7-F}X<2QIbgCbb}rzwyaU*)~G`k%?C10 zh`J`7un!Y5Bfs9j#?R4l+)XOwQXR6?P9hse%3D|Hk+M`%O)ko}M#`5~=#lbkK5(Q= zUYQCRDa&=bunw8d8Yy#DGGiPmlf7yi#~LZmB?;>9Aq&0AR~=~;3wp*%UC_N&vEm-Q zo3I*GV}r`9>KUHpJ{9n6heuAZg1qa4eHV`Yrt za-;uEW?Oq!8WPgEGChCP$PuB^Ks9E|ansbBxVsq)O{&DagyZ_y@hc<9IGB56zl*l= zn|Wv0KP|xTCiA}Ms^WG=O~YJ4rFWzvUB<`# zvir96&QzrHzi%k^4~e>M9TWJcQLGrDK4KNUOZR41%E?*w-SLC%&quWxY=1sV23vdB z%MP(}3bu4$m$>?FEo6(hB=a6#b0$lA46lkG%kE*R*^z5Q`ZQcEMn6Ed#ysh$ZWOoO zxSnpC5ysL4j{I+}Pldfc-=@0TTgAZN*NdPc9}Soh#(Lqr?cP*$edr!l&?DEYFg<{0 zNPzYXbS{2TH>Dc%siZ|_S*x<51n zjbh=b4IB&YBuESWzS<~I~G?DdQf;%iZM;{B>+Yvp zE2m&tB1`KBLTXwjr&q^5m&(bEB=UQSY%(L?mdN3cQg_-Q5&6FSukFJU6;@FnNtE#zyBkiMFEH=yY0Ybn0lA4$ZVOKOaTIL`9h< zhjmD{@$B}vD7fM=-PxO5<(jLFu4)~-5scWHWj@N@9x0HXH$`M!^bJhpf{3@Q-HUhr{(lXjh3*U zBuaR86D2UePTd`zphHU7LN?}o{R?E{Twt&!&($H>ULqUk0$X3ubAfT1YI3pqS#yDX zFX*|zahp>ibAf@I^;}@QPFJi$rnBY(fz8YY=K}TGH`ZKWK1qOB7}NdvLpr2|_n8*n zLrq-Q&(&0ui<&G8-=tw&*Z;x?&Q^|jQCqlBr>oT=(^(dt`l4xJQ(Fs*NrEcJQ0v zg;*8osn>N?|CkS~NFCqM6=|JLw_b-#XH}$(H)uVpdefQ8A68XAlO$c$$33btuvCY% zQ2yGP7KW&a7M5zN$wf_;h0EU17H;GNE!@F$+`=l?>B2f>I?F;dnieLHY-8c^Bmqh6 zDr=um7S`*K7IvZ!*=-JO)ZONFjk?=hs@dvvXm*=uyFMc~_@riJpwv~NtNg~D{?Rx&k_X%d;1cYOHF&X~DX zOIj{Ww+fSM6lcDY!Mcu|Pwh3i>VzwG8yh35_kO)zNqko2pC0DFT+5WhBtjEegt~@v z%5j2@_UVvmr}DFD+!DNyDx@LTj#3Sfe}k=J8g$I%Y?5NeCN1Js9a2;^3z{vicAIW- zCf63t)nbp$(cJ-zR*LGFFuS`0j<-@_t6J6##iMQL0mwgFW-MRU4b7w+zW7Mgjkd`* zs9oHR%3gdVw!EE+UU{Uv!(mn#+@vMXp=9<&fp=Md*-s{qQA!t@q^~Vfp(ZUiNmcLB z?c506v{|`(zYdvSdP6$u-^q3ei_Ri{tjwyoX|E}XsscaK(a7RyVMi`JWACRTORW#_ zNJlOZ?_o@I8sUy7yWnHOus;%z|4V(OBmXaT#0RYNYIXWZN6ym+ZEsnfJ|fqo{Bud_CUTdzG{IinCw(&|I=f{(Ugj=gLP4* zX(^JU-y}z8Hfc+Llf?DX*WIB1z&u^r6$>97(%nVewqq< z6C;9rLjz~f##=wt>Ca&L7s!%!j-}$P-Y40S_!&Q<$4Hj#Sn_3lMkfOH8Z&wlqi5;p zr1`s3aDJn5;CLOd3 z_?&86mt8a#sVa#L<$pa%`I+UGwo|D$r2MAp@|#LWRDaG_&Vi}jv+UURxmlP~=}%pl zQ<)ueM<7B%mt8C^|72A0L~oEFUiXROKWy8{U!)@4Z0|1`j7r+Aw_^DIUrl^tS!ytmCHj|Du=(`J#6LY zviSIK?KN@`bKV~E)-blBvNV8Y#*kiie85Y4d zMXQXbcctKFrrNwK1;}XJm4f}>q#|R-t`wa8Z7MRp+?9fL-*V_E;COYOO*G1T>m$26 z@XEKT$Q(zCs&0F=g5hVA5zTgKrgqTG*3Evv* zO>X^xMz?8u{?gKxhG!aExI5nHc^$OT^E$Aey!b=QxH+%GaeeGK{ztwf6qr5@nwX~z(eqtAOz76S><)xEwjR?Bp zXMPnT7%Ing!t~kC92qaPAw7&G5;N=<$&|g2kwM%mksBFluTfDCNz{&C*p*&niho$d zV=w#X@oa2vJlcUyI({!(g$Iub8opOaubW`w>nqQQs0yZPVoVSX64NAX(ss%|SFWqk zA?r^87nH0&3BR^(JHN7uvc=8U-0O7MD%BIe>RPvt{B6oFqt7`1WKMbHrckE7a%L7_ z(O1eERr?w{7+C?Ju|xkXJEHr{GHR6TF@F_2$?T9bv(=>8l$Q2Qywsf?j4ZY7ik;%s zyZ3QNpNO-l*NCEJl~v`XBSOL0=wP}2h)$Fo>>|(DZ(6cbQm7tPcG7=7(B^mH#nML9 zyEe1;ry{*tgCF*%!rn}B3?E5_J?VEAqD)I0E>V4ppC!T>Ylt-2n{3vk5|X%MH|LL3 z^cu0d8nE$@ax%{@|wOp^v6Pkq& z(v6Rn?8>w@kM{AoCm zaea3O@;f-uYjt-AW^{1ER^Q!$2F)9#qc?ST;Fts_Y_GF__uJ$jH%_6H&$~M?KEa91 zZ%OI3ue$>`B{* zfeRDqi*|}adm}SjT{ShG(Y#NIRZdx=j;)r+vDKPHvmlnp@e51E4P%M!?_#)KqIV@a zA#<*!LKxJ^345WRz0efgDYZZKFelPuZrmySryb_x$uT)%Zrt19iVQoJ9_EBSJkf@9 zKOX47gObM?)6AI4DR@Z4?4+1!cu3lEW)h{iN|c_B!th{{lgB7yuSSOhZ+CIR_P)b`uulCNQ}1UPc5xyz z@n{Fo>xiQr7;LAFqNqW$o=jFMT4Z-3lZqf(-nDBj6T50H_jc7Ks;ZF2A=V%8u9ImC+XqL@nq0Mg|-bz}JBShLIM?27@SqG5y z%A*~abc7SV?mpUq`}y$T(GDDz>;!~4ZIV*W?_CIUpJXR8my0i$q$=47N%9Jl{AF0M zs&u$AY~K7b>vM94WcgnZRj+sQpILyN30xI_lkAjFJ?%{Rt*;#?9%Skxum5n&LpeZsegaFj-}9FL~&T z%euA7-u>Nl_73}(6RHYk{h_L0Scj~FKQK9~V4r_Ek#U*B0l&rnidCxn ze`aAcB`9gCP_E%#8*okHBPZA~_1~&cu936DC;x30io&|b?{w(krLn5i#c# zhZDU672((nEa}9-9_eyv=K4;zX4v6A#;FE-F0b-h_6WJU%444ddmVElbAhL2ld&|F zU0Wk#C%UjpRQ{C8F4P~H9Ef43WP7>Osn6oSmsGJHPS_jDIEkwa#dsk9v+#X%EIY+O z1iuKp@HqC?p%UyDSbv-oc2|&fJTN_iIUP9~`qRi0A>;zHLr*#w_Ik1fLsP3NV*$J> ze;e>*PZ}Zj7G9J6nhm|2us2<515K|Jp8dU?NT2A|$a_3f@ODCd$El}u#_r>p*_tVQ zkCw3w5bLz2o%52{gjq%(#gz>=n^tPgJ56T5zHh_T&ngHSH2>-en9zvY)g^ zVkV|j4EJRj>&F9~IM{4Rr`3!R8#c>_mA46N($hyfFg{ULnOYsP6&#)bWPE(I0|R-(DS(zdQNair*Ht$6iiBaYrRKrbDuE8(wtL$dM2L)bQ2Yb(mr&+j-ocq_nI|-SBV9q*&UPmZbQybV&Y}W!IL} z@j5o9LsD-RD#y`k9lKVCq^>q~b?-wJ^lFh!shF!nvYkpcdiCl)+N-~jUrv^iD^+(| zZz&qzMM;>iLrPdj3Cz^7C$S&qkyNox*qK)_<(PF-SCw#^4w>*LinaZ4v;%dTpD`JZ zV;rbG$qC7Kljh%}L;n5d1kSc>H@uvJ=1QBy_!*-LaqtmJQCNpmB!53c3m>#BObe^5 zMZ|_nINj7S*DyPrZl0K>r<*rq>FK7gubyr;YF=x)`5Y;nZl+CbGu~XOZw?4<_&z{6tkh9YC}?pr*zjD)FCtIVpV4$`dzvZx4O(i-=g`N zbx8h!CjZYSf0>*7tV<0JC4Y|&$$vWe>A(dRTg}mmZM_c3CSR&2+wT_J_G1)Vk{$!d zmSIY(^e|rs9?oL12yf4t<41E3B})xvZ*`J=_qRH6F#E`3jw5eLSHhv=ZTURABS#G# zZ_7wAI>)KJxL4bg=2S9imweuwUYwhwz4$~9y~6fq@@2*G2c&k>XD%H(UWb%cZ02&O z2x0pht}}4C4#^fK8ZW1<#0rTCL!jI0err z1yX!IowQVk%A*ex1;q0Dmfx`=o2xmXMa_PK9M}^?9VzDB|R{yG7`XoBxcq) zSurh2V!4(OmHDRyc;_37CGJhei8rrJ;w*E(DV!69Vt7yBho^GET1C13XaFlj>XOr( z=D)GjNyLw*aYizYvn$+}#D6s#a5`DAUWAkqaGrIayy3Frbk<*cvG8+)^;z*23Z1Yw zGBS!yl5Q4bFx`gq2wq6y+f{oDReoL&+fxTnkj!!v)Az!3+5qYfUKYUjVx6CC(w#Kx znueaL!m!3Qv6aiPRySR>`!3uAN@CBkUGpXPYT?gK)bpXa19Z>vgvclVV9hsIhm?Q1 z$-mR&|Mx(%r)kt02T-Hf;Hp+Oq)Chn>N=*JG16|DPf*r4PgEhb{A9KpABBrC<(fQS zhh+PcZ0z$2f52ASW_MAH)AN6|26`Oj(&N%rnSLnV~^?8h5)0r{~l9 zD86wzr0_oEXJ(y)bxDmKY+AxhWxkkNL zWXg0(ODG}_r&A{k(bK6FTEbc#ww_Kk&4`~)9k}leMWwZvye$ei59Iw7?r z`7%{+?qZD8k|b@_NF9>@9_F02r1J<}QwNOr+kEn4gL2kn=CRQopVx;}ULTUYK0Lw+ z8LS_&GWf8KL-?K%PS_?K<3N!%@^PAR*)b0EI)mlJb`oY-EM~5q||x zlgzg8@kpn9jz^)JN11vRMtTtQlBxH%QEU`}Fs@9d&o_>CBK=Y<7h3aZC+zi?+4yf) zS4+&l#*n*eBIb*dyfGXn|HHitTr24}jbXj8{kb9q#s5)N%a#OWbX%E%U1OZcIA&!E zs)|^A+g7Gvy;o&olMdHgYuPFPUO_uO?3!VCdJ8`O>e!Pa3><7AIOz;E%JMi2DHP z0VGQ$Fa@UNw9e(RBAwcP}jV-Lmd_~SoIa6>^m3V7=GxT%1kxs8HIU{ZOYB$myKG(#&Rt^SIn zsw9^7pR}>;*wdmh{xiwRX(Rbb0gclBJHgUg)8Ew2Y0`X0%iln~o^6+XLl=zs4jv|> z_cUDX$iBeAtv=TA7_JdQ?{OS;n~GA<^dB22r?VZ9mq;g#bE=m}f7H3$t3z^GFOe=1 z0ep!x`En(IB>j?VvbNBxL#FI%%D>-~@6c4I4#`C!8O_|*E-LG9IUw ztf8FzEAt>u4r?FL$(ejpS9y4i&QhHYTY12D4vUB$uG6XM9PQ!9bc*ddWJ=4!m ze)4!azcJqVw=sL-eEe5c&^FFLMJp-PVJqi+EUD>ymrhN;Y3HA>Q(Ue?rnH>j>0Isn zVofd8A-O0-oS%6vhY!j($>BC^biVc3>DJC`IoP&5Rd|Y7UHN*5{s=Wsv`3w*yF7yM zxvjgr12&(iTp3IIPa4OA)c^?39n9{0hyK?a-5c49{~OmluBQ*A zgzB`9SwhkWy=gizrt?*;LrS%r_^Sw@6C1Svl5`=}>J$&@kSVP~I`%wWNGmjTjSk60 zAyP;K&NB)+$DWy<72 ziM!pF(dQDj49tx4`y^&d{O+3BZp(0LC22Zr<&@hcHJx6pQ`2AC={}ueybhVta{5cA z(+f3qunx&ZA>#BslrJsABp0<|`#(6nO{X+O{)~|h=0h+fAGSJBr)0Xk@U%8n+j9Mh zo_1Y-dof*K8}It$^Vz3!*lp0xW0j8^5uVg}ZO~yW*NlwA-?eyT%qQHZ#HIpvG%O>%P^HUNpOe3sG28eLo8kkZbo*P*dSF4^2_%=Or@ z@qBY8Z+=>%ODmxj{%T9=U^(F9EWh^kgR!0Gk5&ll{NAkvPSQgSc(2oFL*oQM2p!rG z1`QrCZh%k{+fvH>5qZIw@{-yRNAT+K`J+Y(F}V%Vd=Hul)7sGZ<}>3P+T?T6n2KO* zh#c~gT+=oKqbf@SLiw{z!f0%Y47B2tztToWh0sP^G&Y4FIitrrbZ%2D5q|e0pgwtG zOGxW-HaOa>dAbXX1F+J$obSSx;5sln_d0bCX-O|UFgs5}w9ujhb5)tsK1*crAu-39 z96oRYnru~4OGC|p69gg=i(gf0TSDFWabMhP?LoODKgu75EurvV%FkoNfAKT-Vzeh} z|3-Bv7L3h+EupxRk`(M^Z(CxMf{`dchPd}EWzv90|DsI&5d}XKc=iE#hL0OHa)_y} zOsnf~^k3AK#VX1zzVLzhC~tsum&2A&bA-wPZ;!X}7)8djHingz!isf=1Kb}1l8kAA zP-$hP-{i`2e`s=cWn^-{>VEqFFr#Ju9Hi`j$A-GQkCED)HtU~!;ZV|dJct5b5cGJw zFuEBB=M4oEukDa|hXq1`NU#LPNr#YRbYQwCHxI_qhrmly$lLXhc`Z}=9YT`Pf$2Fu zFO0N9;0a&vkoX^gaaMkmmjOA@f?Me~aP`%dkr5#n$yU;qk2I$F zC&E#FNSP?949^%-6@zhXoE*rf9x3I-Lz7~;qNKbs6o4_v5|$PT#4z~YSzUonPfv1K zjX(Jw)d?7D`pWm=x;1gCm$ZA51CQzcp-G3FN&So6OUh&ip45>hF$nps&jSp~-!|D{$!*qrru}J3KFSS+tFgU#cVHiI(G6v?0=P?f65ooV|)^adRuni%rXmY5B<)IAK{{e8buw20~Ve%%{3Kfy@{93!&C%>JdOY)CNuEKaKu1EozkHBip7$r`=@^%bGgN zs*IjN2Nng*l~xXO3UuldbXZLFEzIMC4%A!B>TUx%1sJQDlAkp5o3SG3z}BD>UAJi3 zr#d{oF38j3gZt9ziNk?NMO7?dNpG^W@O_Unp(|b>J8B{}a3Yr2|5xMcR7P zYCLWFh*~Cy^%H1)|0zz`XiZ^_O)LnT7R;UEgjn$UltZe8mgmlk1Y-e9V{t>g<|@%# zMa@4=ae@azRqRzxFL$Ev^hhuk3xr@)NtNi!OCUxSFWbT%uHimWWoad@kk{j04AG5r zdst_!R)@@>{QoT7-mOx*eSGCV%LKm`ZP}juTznpu8Ip-*26M47>_nFQA@(wZ8Tl^E z!*m$S#7doqWroT`12YlA%|fe-{4OA>cl-r)$(~ zXA1KZThTH+BE)@BCz5h6b)ZmZaJ&xb!-rWB$}V-_+?W%`hc8v_$J+Qk*kCAgL(5QX zG7UDsKPP9wi@7smdKZ|^tdF(Y1@3ZC9pTE*Q()e<_YFnhQ^ zTBg?mF{5W0uz^$B1Wn~2*P`cibH`IvCyPtTkYh@iOyL10Gp7UHjjS@DJ7w=;b(7h1 zcOL4icj~e+jMk*0jLhD%P~2Yg}xu+ha*S zyluReeS{f5aO8k9hj^lwn1tlkgozg6$hHL2pl)pmrcM7c2^Ico)*f-dQJr;xskIAN z?orqhlIN=(EslVO$x`NzmI<}tNVPRq`z`yB2?=fckbc#8RT_p9)iR7R0f z{Fx1fuA`;aHTuTUNBZRC=iJSCk6R! z|L7&Lh`%CSDgOazno*?9QTJJrt#0RUb;Jp6(ZoIacnzlH;#L~E9h!!s%JN{z3?+Q* z*pcyaO?lx~a*sMRx#FF%f!Mi$NO01O!TwkvUbHE)rj_WFe;|4Y-!9j#mou7p@ut|? zR^pxiK>V;kEZ+D0Pn~!Lrr^3(3Xc6hD;N-ykI=@eFh$q5QjvCODq7hdp3%0(m@*q$ z$?j!9{*~R_P*$zAU$kmstUQ2}$)DS-e!1x%wJ9w@kH%VH_W5JZ=KoVfEuqbW%`t`R z|1O65Y*-ie3%#Z}kyQ-^ULlyF!Ozp2e5cR_ql=80=^@zp&~rLR+iQ5VCp$1nZL@RA zE}NwhbNzhn5eiiMOJZZD1tR7CFt^HkR;#_oo-TA zGc@1il8V8VRTKGt)pI+D`tA94^eokJcjnoVRuP>zHW0%+d5OIp0Ru`)VfO)CrdOCf8bo-u#0Ts@}6Dr5d~eAwi7Q0lH&K^`5ipKn**tlQ?zirLW8ljycPMd~qypoz0#8IuX*Nzyri6K*Y z{K4wQRP;5&1LRB_%(^nOmBHM2v-%KS7eayQg9Z+=T2AdSv&pg1r^=r}rY3mpm@=IU z+i5^5OYX(#W}lIIQH~wz?U!&Yf|okCex3NG&XT>}d|8H7qnpXos19QCoq<2Y-s znCU>DYW6s0C#MY@Ed~w=NnaX>SQa)9kGIqBE_|V=m)0=tv0^bYJd7dT<)Xl4hIfMA9@mSz8*` zA&XVE#qONxz&Ep;u+`6WpmB~3{P+z&OPH}^rUN-O^vl7l9~+2`IM=ebaAdsQrDAs} z?XIrTB~nV$R#SAlg*-T57P2X-ZdAO;YANJu7IL@QT4XgvhGv^yuU3VOD|8`W9?G@j z(b>#Q8(5x^G2$}>sRXPP<%z3>JYq&9z2)}Rdq14m3+2RVMtfE8>4rx6< z*rq<3{G^WEu0v8IoPQ;KKGT5}V^yBl>X6ih>^?iqa^M9~K{k`TK8|gX2%!P5>exLx zq>!1U+K!mzK;2yRuewVpHhY!>i3?~sN!8<&_BA?WLW`|G*=+f<9B7P}Fm9FuS1fQM zWBe=!?jVnd`caE;oTEfi)b*yuqZaZuT=gsmlIJO7NNT70O)ja(%$EA03R%~PWF4x% z>8>-3tjR!~l-z`cW^t>Aav-6cp2@7ll$F*>=2|6lt;npgWUiH>K9J0@0e`tAr~U$! zab@E=C2yU`yKSM__}3Zi-M%K4stg@-=OS+2m-{(isg9YkU9M$1IptIdBiEXlRT$~% zdZvVjrBc6lt&@L!;GaCoADyD(k{3Jq56&^vi}GQF++kPnmc>r==C>T5G73)W0@!Yw zrFs_N3$q+>)H;#z`YZ?DU8q^ znf&IN4wPK)R2|a$*Xs@mSH^WnZ_RXIz*46gs2*F&p2@avmIJOURb_dca#Cs>IBpp? z*lQfvf1Yx7()lW6Z5z)uh3(=R2f{iXV=}I;ap2r#l=;`*PO96Oa3ExrUg5<`pen`9 zsuVX%Db85N(%@&CH;W^86YEk`F_Ug^LU&rcH3l8sV+9jRF)F@0SDMABFjs4C7@O6x zMoXE#Z?K~}V^w^as<^ut$5MMoOlczI&4y0|KCqfE>Q==xT)oDL-ht|Hpd=Pxr;Dbz zqOQ}v`h-4aQGd6_2`S-9BRl zxO6(JWbU|?B_p+_SkvU&XRI|Z-KuNNakuGOQ+rjLT65uAU2E#siku}i4m4;vU(mM4 zY8?3Lb|*3#Y8>cxhZDWF)Hv`d9~x^MIOa|#Z12@L(0rA84V0%c8GqC`P;{pgnVn`6 zAE@F~RFzxx#dS%t?5wGJQ`bTDP#rk7MDKKhYxBAeT$C-p(~0!Ua7za+${xSd347Xn z8C7nRWzk2BDPqj@%19~JcHk7h`#LA=HH^Fz_Xyi4#?&*WB03%Gh3krSavJ}zsJUYu z{c%9ktHP!JSU^uxP1}o$OmXdt_ zpt3c{jja^)FLEc=sL?x(W>Rrut5z2ws8RZ(tW)*B#_ql%xcvfTkN(S5T1wBJMg#Ra z*A>n|st!>}$GQ(Xnfj;stI(jO<&ru1A*uTAGUVmXd zw`B7rjhkh`tUQq9Jn6`(7+=bD)2sv;25n56;63)EJy$o@y)^_HcONT8i%94-uPdLj zMlMs5Q7wAmaow3uiPMEsa@=U1l>~&ABAb@ROQoHWH=ZkxzwJaCbRL5=b}z=z#>w~{ zvOg(OJcXv){0Xf@XG5E-@BvB~O z_(zmMR^`LLAkjW&mes@zDVkp@ngkPmho>2;n^n^=<&s~PnlkCCEkuricp~Sx`S-A4 zkF-T>8Y+I@nU3(^j(v~keep%B_c&n| z-XXYSact2sJ`YlmO_!>6=pviOa`73SAJ>_BGjov2xklN0`n3f7v;& z4|y~cka_e^1d-F}Sa|6-L`!kfz$0J;Nk5Z^e~U74%RT{qv5vVMwBOom5IDQLcGa{( zj@US^Kg#67`2~BiAeO|$ex>lRIK?msOuL>%D<#* zFv^RXTd<6ZPFP$vCxEg$^)GuJpnBrgz4L2%qeG|n!b;S=zDqJK%K?-?=uQkA8*YroPBGiM-jxh(~8Y@F+fC^OVhKKly3 zh=yJ2;x&I`@PF%nLW!opBlLk(RZHZn?c3&WTl}6LT_I&I`zEpZt^Xx!97n()0zXZ7 z1&Nu@Pn+~YmHYbwXWw(TwP7#B95^;?r05}zR2Nn+yd4&^{rbbNLMqlWxwHNv=|A@ z8aL>d%YL19lQ1*W#Sr;d6nXivdY(r-EhSd$e*k3OeV?%;5-C9eB1VXQF-0B;Hz(Nl3QBPV3(TSkd`CzF-;w9g zu*G;(kt}ps>VSOTr^?PE8{y1ElgbbOahHeWrT*R?+tV}9%(Bx$+I>BZY>c?SI&U;! z4y}vt6l|N~94|zh+2vp(?-zIgQu$FJ1D-$pGev|`FNau zdoVBxeo1Mfu&zU$3=EO9pbOEXJB9Q`(EXN2)3zOT@cidee$lc^vv-aiZ zpnR-V)-jOhF3GJc&xEae{oqs98|5cA;IkS1m{n0t1|`S7e8FzE_;GpWOo^lXVIdD1 z_QW{FVGa=!b`n-ujPx@-mx9C`-`@mZB0UXxMkEp2DL$3UaT%Sj*SRb?Fs48s#qGY5 z%n@LTWiFcYkkLfADU%`oCmv2op03-ZHby$`ICga8BHBr^FbihB-NROqn@}676Kt5i zmjHWYi?1&bztn|Jgk@u2CFnseIserLBb?9Gg7$EOgFZVZ4n7@xjjJ?Sl zefHb(xE0{wP1VCe?}Wxps2$#++9=#IRk^|`#rKS>1|9{{eDK4sgTz!lZqo#za-tj8 z0@60=1i-&gwomreQ7!w(R#SjRn@7x5=IV5?S1uh97ODi-CbHM^T~|-63+#b5c(yJW z=L|qXs4DP?%G6eBZ>}IkhrEsa2$PlXwUg|#YIXTb^hwk&`04%Z6}I49OmN0Svh=UE z7q=u;%^<-KK<4uF5xPWT@5Lxq#t7V~50)n+LqiV->N013TxYnZHqDtps6{;Zi$Lb6 z{jA&2D!XF2?x8ri?sBd}*b>T>=2_rX^yaGhA~V$SH(y(;c3mQyZM{g_V8pggts8kM zBROffR@gAAL{V_K7B?RYjH$_@Zq?KT7`o{e>iG&~tE9JQ(&X|t{_|*z9S*d|E9le0 z4c;C4!R7TWLMu_9y?_1%=v^q<|K!xV4)&EUFKRr*o(L#IQUK)(n!I}E2J z*w{r%n<{e_T~ybnAKwm*xHR&S@pbKL<<}%^z03r{|A_oB7tv#$jSUPNj{gfgCNMj` zF9-Su3hHZ}#RBAiDdHWLE`d#nv$)M`4ZS)|?lXu+-^s7O!SU`rZsf4%FF~WfRFaX= z5($tpR}Kty3?^1({E|DTp z7Za`Rtqf--_^w{jILS(yj%?3NP9Embl>RF;?`Om&K zBgQ_YEk?`E3l0=|_3<{T()P23Q>DR$X>`5K&P8uVm&(ta=s zq;Tk`jPj4Re97zWW2rw$lmAU_9=IdOK%TNHSMHM9bT(tf9yYTv|ic zQ7nd4*QDGc;04b5{GN(gy5vr)<`kWx<@joTbQ9d`S%c?_Fb6BSm-eK;ku8VGbJ{|P ze2#q;&o6IG-b}MJ!h;#tz52AD9=oSYyZRT@hqh~8F#}bD;%asAVZBYX3FSIMstxkQ z?K04$HdvZ+yC3Y$d*NFx@`OizzJOYIFw(i;%yEMjR?^Qa@BZ{dI-zjhe+kVpWm+#j zCJ+1mf2Xw0Yq2j=kk6xJvl(JOd1=-`yBvJg%+`)OIgF5I7ks6f*;}HL^1ZPIJMkZA zXL>I3d7!!u@5ex^M9zCDY~%+iv%%|k>{CS61ITKd3TjtkW0f55W56qw98G#KjamcY zPWKJ>a(SH!aC^PxCgHDpaBT1sZ zE5=CCo&mqLnUGpDV{{^{QtZRO zF{9oY8I^3X3%|=39l-Dih|1W!kei=;sOkHRyKlymZ+?wEqA0%(MUbzx>Cp0~6;c~X z7F#!GOgug!X?ImPC4~F&{>{wiUKc$oZYuqF;#c{X4+S>psw6pmK%FMLczJMXsBo4M`XF?#T;eRBe~m_+sSK z+0Bk)MyR&jd?(bu-}WM1nK~R`+KvtU<0muVnb4a8T(Gq$ z(J|SvyB%}(f8kE#99 zd`vCOl*UP%Ys(vzC>$LkkApSJIRTFY{8`t#r^3MkVDXkl_MzNWf~%Bl&J{6k zqy~P@o%>t`#<8$(*x98Il6hGvc;CIm?ZF6n|9nx-Wci#sD{z}oIeeSr{qNHYhfegHf}e9K#IwoEJvvk5 zir2EyZ*(9_lUAq%PE>T{-DLIeD;hDHtkd;cvyiU2>dQ<1^Jh>0MeGe3rZN9@YqlHk z&+m-lVyL{Qk`fG{Mq|8UxZy)lBdmW*g3>yZV!9b?s0@x%s+AMRdi+@UZ%4|PDFm7S zn5q}=djCp7bl6b-(j9pf;{<2V_iqDXp1};W?6N<;qVr zPebna6HD$CljF^>XKxNuccf zb52d<3xwd`h{<@qSdKL1ZwG>|9)Y{^eC&E6qwe}3qt-ang#QHCY3?P0(6-Up*JPW? z6seXKKSY{xM?HmaAhK#hxJCW$$=sPotq`}iO4n1+^K(P@dP!l%1nv=fn<3pTt+Ovf zUa!jD82M9fI3Q0_og@ElGVfJu=6r!yQD>cc>Yg7YYnQG8-Pzi>wG~nqZff_iuCuzd zUU^5tT!^Yvp0}KwZ(_m4>+t)YQ(QI;ge$HI_mIZes@G;0SAd5Jat8x~2zLEyC zDPK()@qnP#Ncymu?XIUe)URy<&S;X%R=cA5G zbzW6rnod7PkGTfXyc`Ux2r4H2@K<>n;XB_5sju!&A1&DxD@I$+bz78BsA^-9{gN}t zf)vRICSlnvwIwg;343*7$?a0nXfHL_t+el`_NBxCi!|yMtNKx+B{Z^TciW}I_I&am z4a28mj_r0dd#J#L8rY3b!$_wQlpm^U2NoL{v)O&7@KkDFbHZCJ`CAw>s*YRxeZ>_S ziNLxnb2r{31Jtg&%wx7`7bEQpc+m!I=Pn;bzWscwno9;fiu~tIGXXTmD0g7HAj8O1 zaW}9Z7Wt-LQsmX?uZw+rjB~30wV7PQW9w^`BZUmUjCzH&J91uW+>5p1tcm7;bGYRZ z`9}>Gi_8$_9xlws->!NIzgBK?hmq^f*bp6v*({>YB85A6c744hCCDvI*sk{s7Tbj*f&qLSu1 z-kdFOh{?mHhFqc@6?2EY5nN|xIJMDXQ1Z0?(UJAt|2~K?$*(ZIV2);q+$-QnD~UQ z8XBk21fLi{YP;ZRkR`U5tY06}s1ThES-2svy#ebB!CLj31nHK$Aj_H4DB5j<#%_9H zrZ%4Om~OKNu{E6Noa@1w)*|eMez`;}Ekb{(=O$G?)rb@PC6J-#Oeg1uC*`2Y_8uSnjN&z zR@DQ)hfI_V8lgGpC@*jSjd`ntDP?}-tH<;qx-ZoN@o-BDcOFB7jct=+v#21 z+1d3ki!07no>*R2Rz|i3^OX#IU2JjiL!|DB5{-3Wr|1j&slPRv;n`CY6>7x~1p-Tq z2-;7`eujvR^X{x-3m24#asPS`ag8*h5Q+1?B+e+Hm@8i+Vo&KQ7a{6}b}ub1-ip8b z&N6u`S=U>ud{K`Tww$+6cS-z-C8S#^my zf7KfN(IN{>D^G3QJI9RRPsP6y>CmIhFQFX2=IO5Tzg7YFUiB0W2Q|4@;RI!=KOX!9tgo5jH?6Re6!clc{{WE0V zeCUa#+8mgOJ@}{j!C8yzb;|q?usC5$yZOtYbY)aPx#YUk@=zyL=dqv zx5Ryw9CTA;DwHVS@j2^j)`%9xRH56ayP!@}Kb?JwI6}uKuR|oC-sVCD>=Df7L4oBS zY>H2*;Wp|l9oIn*Azu^jh|$t-W~Zy&haluCXPZ1&pkGC3F*%R#G%(%kt>M$uX)2s) zVvc4nvJ?~_yEs3TH5wXt_;p6{3=XDwxJIsB+ZThI?yM9U7)_I*h;G0=>yl|5(iv?= zP~JPbxz&;r?e&4~rj>3`7S1dGFF&e84o;T+WUqxuqXn%`EqPV$q#ezIca&}7Ypq=F z*oHPkEK+9a9=?J!4vz(jWd) zs9PGcAtUk>#Lk{rn8|Alu0t|1Q5wTo8Z;0zWF=F^tP$))t1mv}IDEGJ#mL*`Bbr&^ zUZ{g2sWrou7xb*7$;nG>kqZ4CdyrSp^|^lL6H#9nw?i{wbsVXsP2B-cqThdV8QHG6 ziXBI8h?Q*j;=# z>DsfMr!E?OFP2B)S>l$e11*H#7T{@~Vk|W7I#v|r%5Y}{ny6-0G<5S|q8!60Zu-_9 zP#vO31#=qighoZ;iWnwCD3_KC4u2>5QFX0EV+tC*Xau1V;#>iEtOPTKU$el!}I$5(arzFS?^vw0uoDU=#mVnnM+fsg~b>f0-Lu zDxXKDNUQ&sEo~T@CdEibPlLf3_4aT< zhItoNernwig9@^nmd3!!9&P+9CzefTdREbF+MZ)6(Aug)?vq|g65%CoJtO$ zOE)EkIeOgJgdA^t>RF(j9s44{cI)es)uVwo-g|@0p*9^=guKs`8nuh;tiy>d@OCxd zR8$+|S^hJnHaZ9-jfZaj|@7yIw$<=p;%B3iz4XnMHh_Scj6HEyddRV55bAv3iPLanDyurB0 z@4*E;8iM6EFAD5}rb1HyG+iqxp*p)hku*Mk)5R0q{(c9URDAKV#rvWzx}LQEgg*+c zLY4#e7+iKLX%eEI^w@H%J9#5!0i?@LutPuVF}INrb4g8 z^&b|bU&`g49d3297kVW7Tco{Q(N}+~Hc1;fDboQ_wa5?-(*l}yqx;XC5l#XDIJJ_v zB6|+5hYL>LShpA-MFwPA2)lub=IjBuE)eXxnJnnP5U71-tQZO!AG>%ogwQIlkAlz| zb3ON)-)gG_XP6%-m1R?DA~Tc)@LE_$sDtiB#sU{Y$hT5SC<3Q(gJ;l5gmKRzHt!TRwcpJ8v)7jfz4X+dl02-67(ONHViw^K0k@9tO> znNmO2>IWSQj*2TIExx6ws9g%nKv9`nMi@`c{Bo#pZ#2pNrgks1C<0VN*n%#qv~9S` z-YUNhTBQbu4L#^G-oKAV7|z8V3E?J1@9P|n3~Inp;G+NTYy?4r31a>VI0w4mh23^3eA?qmXY;AEfoDTFYC zHlkAi&t<7sB490-nr&&l04!r%kdy_-`Uk{TpCzprVRVSJF2y=ttvJuVMaAY5@9h>_ z@2f_HKYJ%zYJ|U2C(EX+()FfDuf%vS8ghFMmA${QBO&X6;R)Wf3GUTJo+kez+(?=>Gqk%`NWWP>A8lN7exl zgf=&aVGq&2AQOuKZ}Sr_*pU5s8hQ1;3;@1}LYC9v7vy}Q;Im_2qV692z!UUChlEV< zX@Ft`+|e6cj_a4ZbcTwiow@v`mdmUkBZKowTsAd_GtE1FT&8_=wKA~}(0=m?SFaSE@Dl(ne z2sNcP)>)$s=NU#io4h?Ksx@K{7am%TDWZHEobGEiry+7`LAxQk;_R!r*>e?ZPL)af zess~lL$#1ibU4u*Py!gaSB+Ph;?Uh)8$gt>&!MY6rMQT->IB<=ez&r9yTu<&2-FSw z4S3PX9S-}1vdDbUD;zwNj4C8|`J};8-mlxj<2=zA! zbax`@DI^MFiz`It9l0{4_Avr$=ceC@aZDvXampHF^P_jlxbB%Yh+kLV`u@%LRj#?+ z`$O;ZQ>O^tKZZ<-KR#T69H=@CT|z0&qyu7HeD6UHZe!=gAgjqGb;ii_#Bq;5?$r*? zT-@Je^ZBV2-w8ILl4}F&WSgi;TwY}NaNvy(qak-#&IB1!He-%Y;@|_8spD`x2t||j z8@?s&%Or{?fbMm(Eh3XbRZtKBYjX(+rP=-C9!$5ph3T{Ye*jz5LyiDgnD}*(*#kR- z?E(x~yO-IV{@KX05o#Um=pl5koE;iD7e$G}uE|L#hOwG!N)bSCU!0H_Yl5O4gWV0k zLd0tSdqNSew~L($`-iMDS&Sr#9Y8pO?u+%3v>$~(8kH_}NIF~L-2z-KzLZ|E_!#9f!*khIZ$gDb3huniEh(uetDdrb>1Spv{1-;stV=cZV z$R`v)zgK&=h4*URyJj>V@*4(4WGXEQQnp}ZZ<=SXvQs+Yvi4!PbOez|dvv%!O>q#W zhwGBT;OmltSdg0(NEv)HQ8x&5i+VzVlsUTWn}wQo7&v<@j(P$kGl?7Bzj|+hfFK|ix>eMcjquu`&n)Vlb+udUBrD3IbJgqB%#!;Kw3s@6obf#qsHTwO#4Ri&rK>Hsij zpi6+~O2-lA9@3E5$&aL6?9VC}%p@Iui#HW`Z%r#UcZCwnoUzt%6t|2y)(I}sttSUN zLU6m}!4&Jbs|MQ;2-W?&X#Zt)aPtsgnHgNAleF8803YC(Cjxyq?DNqmF$v7EMDc44 z% zKXlyy7*-(A^TDOO4p|U1JO-=ByZb%!YBYH7vT~*SPu89RtllQI?WymtI#Q^ zRY>ebMD$PORjTCzmtag5l$>o`Qp{Cd=F~n6kS<%UKCXMMOznrRR=FN~rSZXu|2Z}T z6i`*fL#q~NNh|=M>wr9Yer}88ROx_3v04?BuDDUM*M>N(IC`B$e*OGHo0A>JNN}JB zK1X(kXE!6N9-MQhO)*W6<1@Hd4H_1vopYs8@}_O@rv=hCq+ZAI7k!5E9vbuHHFPh! z`38QS9e4q?-2bK}%|^MHk{Xv2InpX#*m+5d+56vH=IYe3v(lN?fH3)1a16MpJ4u*i zAb#W8hS=|-l1iE3ZzqVtha2745obV-h<~g==dBM|?Sl<5SdL=qQ(A~WSZ~@Y1mmJS zmy|yKci0k2y=W_f`Uo&3phY9g!Fw|a_j`H453EvLvoneGtHN9EW|bKLr49 zKyF)W*L=M!N_><1(kqta6g)@z9(!0T$2C06qCwJx+Plow|H@{x%%P*51C{~HmH|uV z)l1`)>EqpLE7(4m5J44*)LGy&s5Tara!(4M-QG1#r_!qeHOZxo(6=t)u~MHhFt#b$jtz>8Kure zbRD{5_;RXj+F$}V^Q?O zhUwXvkqJm1;J2?c;pt?_h0NN20c^br%^)a131`P^OD$N3Q2qPw=wX!lKEGPE>?IaN#D)6e{a{*D z2}%MF_SDGLU1U-$YB1fyMJDLgFKC3L=L)%P$8!tdbNfCQMj^nD+4n}41Wk1Vxc~jxGmJmp?-l$TmI*oe@I^=z!ZZ~KoJ8hDKj8}}>2iXw z#vPkti^BG)5x!#Tf4+hp0k?&Q5bC%>&`I(>AT;%}v*=G8a88IQ!bt=I&5(Y69HCeY zv-WvW`_kqFVT_~04v5=73uWdJKoi3_smUc2!^}kZid&3IHBQQIjLxr|JgLY1&dAKo z|2^&cG`Q@wT|j#{@IAIwf^Vrc)AxH%`j2{n(ku*QQvMCkw5UfK*`K|VK({6rO#oxH+!V_R=e{BQ#z&1Y zj%Ceom1as7yWs0j&?>_P9^k4@dy^jH(FNKg)Nk8X^Pe|{lyJ(3_vqDWpWf!s8gk29 zo6lH_m~I+BWAKA{Q*fjN6N!E|AS4;TSLSAQ#mzJ2748~d}5;x6vp zQCkLRt!o2-Yl{Fon^pizl&0ge^|F4H|2P;9PY9n=5{|NUh{Mp?DxBnZdiH2xi2kfT&XH z2$fsa@}RZ-k^O_#kp)c)U7PF|Z0>@5=UYAn{V2$BWx(8s#vTl2e0!NIH_eGyj$fvW zLOSntr)fU7-Ak!<<$@qU+lpKWg#Ezgv*-zq$Mu&ZmY$Rl<&QgFW-UDmcpet9Jwh11 z7fn6C@yB~xkXwG7^a}rxZV=ZlHiU3YnJ}wU1bjfSyMToKep$8d-_RUDJ3ecVaqu>Z z@9Hw6OM3{#GlRGOsJm$A zsa9&!7=T%;X?jGoa_DK6)-#8A21E^>!@D*O>w#hHnZkNrnGXi-gT;WUQKSGb4xCY+ zR=G_gz?B*!r@9mr?B6O)1x1i>1_Y^0I_}dCCaYg~7UUP$wNiXeZ5-(NH9W(is1XI+Ah9{z+Zq+Ls9yL0)h}j2BDb>cDEG|Cd|RetTAO5B?QOHr$UX8(>N(j z;8(|yJN-_ymA^Z386PEQK&Ko(#agLuTkXz@*fbJpk|7B>L(i zakJe$O{s;2vimK;mgouoMW;1;k7#cW%Zb;PKyUMvRNL?94S-D+W8W&pGj(d;s!(ZN z-9AY{T30HYV1C+_OPHzIG^M?+O4a4yPcjCQ!^Djs_r$s@j7nSn-~r zLqe_EE>l;7*0BhJkyGQ!UX9zSyA5Lf*r~S-Va0#$7!qa0&GZUR&fLNs~JOA4Noo&b}$Gt$b zKUn;tjxa)R9T`NI+UITr*6j(3qQh=;0s&h!EwCwYA+RB#Wu|*4de)^VMwF;pLlFHY z-$W*~1R&sGO-X`+>5i0*dr|Y(Tss?c3X_3rDw+^w$1Wc&1Hq=y3{VU?jU6Yi{?PZP zIAliYplnpSQW)buXA&gGE}=wmQm%ivNMj(#RpX3!VDDt6vI|*GJs;9W5b{knBt`@w zUndx6RtR8U|BB=T?gs2w_+eg{l^}CheFf_(osf8r2<@On`09W14at^N&dO!XQDD5D zo?DR+(O5I%7PIH>cIJL{;)3#BK^fdv2K6s;)6i@9ajOMz zfA)Su%UO(hpM)lFo)X61uvZV(v@$vcznJsvH+C;Sa-4*mb2RTEhWCFqyE)W8QrKeU z#Efv!vbRe`zVm|x`sw)9PC>?d18$fIKOZ%U1z0tckdUSN$YGB}#~IxB9Mfj79|jv! z@(I>PVRy>Twk==#7bw4Eq;)#iDH{)6IVA%^?8YUtxL` zIH}uxz+e^xKsWU;B494MCkwI9B@-;ejPGjt5gZcI)~dJ33Aj>ayif;RxkmL!BYa~; zk1qhzC8MVbxr@CL|Gpu_X_fur;J+JRkygn~X2kE6gHxCe28Dy&@x}*@{%@1f8}omg zjAtM3VF>Yr;CnhX%^*KOJ7BNfC8x&~qrL9)rUyWzl(5x)w%1-F)UVGRo5ON@6#~d+-c`-iB9V#g90xxV8Sef&83YaurUGy@ zD$F~7Jjk1qX#RkBpO14F+iP=`#P3r?$QjTkpm7x|#DS!N)`AwrUlgMwi3d}4=!%U8 zM~=A}jo>kR!>DNKi_1Gk57`j(@{I;8)(;8idT0sGnuQ(-h(y00cRKnB-$kV1Z8$x# zaG`~7;@9m4tY@HOQe4rfD0u>#q-dQab;-eUyf7W#9NwSwzOx7k#mjraW>K1N^KRq@ zRgxH6Px|o52Vt);M4&STnc8k*+0#`sfL2wFa;`4JW2Dh(Mq!)O^P~jv53;3=N3!Hm z(QuW*Jq$#s%hNF(OKyS?IG0{mEm%I4PjG|~WV!MpQMB_{xY-a|n2r3O+)o>Frj69+ zSGC)HR~;?SSFxAb4!8Vz^$rp`jSf|wHNW?M=URH!*8IBZ(foFHcFE~>M*O3NREbnn zG@&DvmD(A*iZ+kQedo5{>+aRaIFI^1VX8?cno^ zrTAAO3fFz$u5I00^ldB=jfBzEwvtNT90>a3v6#>i4^J)>u-vGfVDEN5`**4b9v0%A zPj7WxWXs~B%U_)NzdMMNEUS^Iy~%T(}Fy=)$iZUW0x zL?HWt6)OL>DPId7=ly)MP=yZncT#)}=Y&uG6)cvlKQdOStg*j;n4BN-wnN_?F)sO+ zb5kbPSpm!g-R#3`k0DMLFG7BK!TExeF&qn14OsHx}F zzt-gNN$F>zSLj(y=eld4->Uyv#7$XFo9wYhphv9JSz6cY@4e=(l9koh61MQ0vR#-! z7UU0h+F}VD>s@z=c8>3y4yoB+k=^f^x;ehLI;2P?St4CnI=3-mT$gaSzur4~px2>8 zPr;k1*JCL8%kOjXD$ypxU-M;*)vCuDR$(R&S8ZDUM#|4tlZo{v2qBnbfQwxup z#;C^{=W!fUw)1t~S~=~z#t-`^DRalQ>Mch>qHrEo4)PMO9AhY-?}Zd|3RY%2*w^Uv z;;DEbD!y(bg@f3C#Z`~kiEpZypABk7fGegC9i<&X0*z8-~$DJLwt zm-;%*7D8wASd< zO6E;Ry!TuWMC;_G#5g&EajItEUEXbz+= zT)o11P@eTx1zv!!lzeiPdN5LC<}+MZKRgD4%E;RLjPPc)XUc{0Kte7MMwK?~qtFYz zIoFhJqvnlxVNfw+#Pj`zrBIXG0>_KeqQ!gt>3q4+CjAQFS15G*@8fsK#qA%bH^k zrgOY3$x_Y!Q$yV&8n~S;#i&xp*IF#3$cN*+GgFx#W1P28I%0fbKc0H`eMwU<@o~45 zOlBczDSh5V7Pi%v4sT*;df2e~rx#3EiYoPvNF`ID5?1<$a#l!%E512gH@S2jcg7s! zbTixvcJw`*V$SaPWwZMeiC@m}not&rcV*}NoZ_fv z)~K!JG-$wD_~@XCRi~|$G;Cgere=ju7I2umJVFise~m z8nj>^F%o)URWy`~Z&oiKwP-0{5>`s8<=a?gSsYe+ z?m$p7F_V`7Fvqt~v^H!&KahGj9pRsmm=_!DUChPo%3RcElR7P(%`>1EF{eHvf)Qs| zT2_aZ)|9hOS0sF{P>q#3#Ho^ixwqr_|GJu+@Zp=ZMF$owJkHhd?=J|J+>~^y-4c!W zVl&%>-;{*H+(8Q6#%B56ItVObCB5>W8eT3k`mGB!s<}N!(CxCX)QwO22=l zT%qc$cB>z@Vk9&tF+T`zBgoIbRV6dAWjiXucIANx))DaaT4tX-saYRJG?mIYsGV!s zH4$hvbGhq>uV@twvrbB-+2J3xwti7c$myJ z+IEdV*sM9%?zxt5m3+~)uFNkPt77`HBq428EDb$dIiOXKLsC>eD-N3+zFwiOLffm^ znTuH?-A$%;JRAQsRJ$|1k5)SmuAmTa#zRjnTPa45$B36h@Ht8)@ZrtMxM$(qN;8p)b{B8+vgD)qE#;wJ-#!9GRuY+cK=Qiz+#C)H$E1q-n_Tcv zU}$n`Brkfq`eM`8LMmQtVt1kA1#z?<#opgMqYJ(v@QV`64yFFhD5>(2q^R zs3enGJ1=`a*ug`H!3|qbPY>#w2M=QAJH)M$`e&UaI(RKlQYHN_R6a@%##*8FyW%?4 zV6>jXvk*>p#iC~`bTo_ADNI=#cd5F=IukOl0)khJzB89DUqu~v)QYv|hU&&WTAdz_ zMb0z4XBcaw)8eNdYofYYioML#%fm05fQ2R(Vufci>JIZwXel?b-@La9jlQNl8hLQ5 z7%VHR?+QYyyF<+X6(>I?m@H+hv>|K=?^(^VHM0>T^iNh1Fs<~Ol2ULx7#_(gkk2U!>vhxRP(Voi)srj8AdVlhNZ zXbESL7H)-pT+&h3tFjjJ~xzU_rR%rlNaBh-ct& z0h;8_W=39VXp4Rf+0CV|be6e|NUjaoL9d=E(H(MdQ*nu2*QBppG^osdj*u^*udMf6 z-{ymNSBy6Xz_^F?;l@BrHshjq%i=yu?SL)PX&rZsmg&Ij6l37zKkBVj9i1llGO$ zh()E7qj9U>-L4=s_qc$4_v30e{H9_8K4Oofdk<`?BQPNp-t& z9BoooVb|hQAUOr?myDn+gFYG|*_;8EpsM49=oc z<~Y5bR^Jg$0evN|=1Tdu1$t!#-KFwtD|83{P=)&XZ%r+gGxS2J7mY~()Q{|{#PD+M z2dAGXb>%ndOmldW%W1QwL|ZZIZl3H-M9NMjC!7f&2{aobC7DFjlszJi&%ygFZ;43r ze)guJY`<0blgTg$zWi(zN;vAN;`)BboaR9Rj+Hptt$v#-;Z&Sy=WiFdi)gOC&qSzi z-G0UFl!cxu%@NfQk(?e-i(RC|)sle9@ev0`a5YMH*tmKv1EauXNwe!cGb&)QYbB49<)8_M!sQaDT1p&%d-`Dgvr)8p~D65_RSX2M<&>-8f3>p{h8KgMdGt7$T-@v6*fUuSGm z7CxCnaY4EEJXezmJbzO1r$00Ad|fKgw*0;Q#tnL2@n$kCK=Lc-VhQQe!hSkRMaoL|dhr=N9^f5Mx5v=c?Q# z-GPwx9I4<pbzMLT6wsu-`JbNS{Vq97uYj-X z@~&Wz3~1L;p1$PUIj~KLo&50iJ7C~^h}f-DF)7dqvudyF)9pK5sT-mhdR+;KD1sFu z;zIeoZYu>Js|fD+17wts$h2nzPKouJuk<&@QqU@k+xg1L>RNR)gFQ8?@-g&RSp*hw z91DGu;W?_mj>q8ZDL(ejP4(W+1NE(U7E^Hf9TpYA*zq)4f;($9{P4l9d}MR0^QqFE zV`aFSUOBAm^#@!N)|IvZUoXjb)BnZd{o0v)ql~z}-%aRa>FY~5u&ObPPD#(B59ZVL z>r=Q_DCe^;Ug3ty#>qGNa*|HdN%)TH#}+f0rxbM=l9>6W^|{`5nVkLonl|rXbGt(N zxZMNO&((h~)e*2YXb}>~zb_4Vyu{7uMx%GEh;Q6=<$)%ae_mP(e5}t5QWnIJ& zJ`;VEPzuK}QA1KuHxf`>B!fS$`_UBZZ~PX@pLtxbLpVv$VBICgPUBB+Elmo42jJYF zbqlp!B2|qr?b`tB)=D9ickHm$j;Nw8ote?D4r&-LkxZ*n(So^&Pm?^EjCwN#`x4#V znM^#C$%aVMxHJ@#?GolanHW@;glw0;+r|Ag-Tv$OPeCR7#?HISkM-a-ZynfOvGSPT z)kj0gqo&R~hFwJ#$ZJuiLzpw=u0#~Xh6!!_PkpltKl)yr3h8KKMkDLLD(5&qZw@u% z@>z3H#A3ASyVLSzrPHw1m1@@2v}^67p~?p-v; z-)fbIE{!BnIK0gvAFQY&UQA$i9}r%x&?3-}P#|u+zP(dxmx^+hZpW)}Pv-}|Z%$xl zYsUQ^{RhE_cM<2nMAmaRYBKGR&V&l4U^(woXtg)p>XH4HAKGVJ>vI9EY_v~OQ7$Gw zLj8Ru@T&#%zdt2=HHUk)62@+&byKw%AUQ4FmOTiJcT+NaqWd z$+q}sxl9ds6DqK&&XLBn&#Zm^az%rZFb`Yl<*L~NzBF2%M(WH5vO5IF;ama|+N!)- z?B*`CVS|PhCKAg*0uvflQ>SIKw!zhz%q{xapQMWzNT;``G={p=m5qr@qgh$ai%TDRCAq zRv1;2yF$p{r$L1K@6gTt&oFlJ`VTkM_`{{!_hWX`Lk5{1SzuY(M=#pHwuK&{FrKOf ze;J`A^EH4)+o!z$$I^Glv-!PY|0r5DTBG);Ek+TPm{qk`iM>~BHLJ0zMeQv?%v!Ox z8m+w(v0H>zYg5$L+VAsy-}etb&*w=_#yR&n=en=!x=$wEtjsL$nOA*@Hh48V7-V&B zA*Aui{BN1td^#0vYp2nE*@QFUk0fgEVYIgZ6gCwY% zEzxLp`kU|5!IWjil95(_%DRuDxVDn;a&!=-dW=Mz-iytb3{)jAp`f96tLEUaAzQM?EPZI zH+pC-3E&NoW?hsniMeGdCAqb!k>T>O@d_^6wjMPh|k7 zM#~pgVObh-GaH|&VSPl7euqq&FMUgGBFfuN9_~dcRsTR1B;!_mG*KWw^G}!f*p6mg zaWFy$Zc}UADEN0hXxdUfbrW}A7=PXA+XFrEjdYw{OzI@Jzv( zdUJx6B(Z`8+JoR(GK2tA}-RnG?Jl;+*Ry|nF2?IQF$Z`Q-m z8kK`z=IWbt5@z>rEbwmEe@tOzJRM#@&aqrmz{EGv9A|RVl)E-bZH;*iYL5JC0)?49 zB{dy%hsz~t(^T`0jz|bs1?@0!j z1;9XJN)`FME}bj&F86nd#41V9(&Jk#yscDwa^qmm8lw+*wLTh$kvpSmsl*-7p}@lg zeXcX@s?cALQQ%*meWm&fMCR`-e%g9g^0fUl0TRJHknj@E_p0e-Y1DKvNK565?&dc< z`Pj%R!ms&poiQamCPc1L>p=9pNxaP|vq;h`U)K3RotVy3(A$N+Fkfm#=3#AquVfo} za#~njZZS=JqVN0Rk$%LT&nfgOXQZFg{tA|DsVX@0&TS>qO=B^kW)Nn>j89Dnwod6I zdI9khUg*Y_GW;d%eHjKpaD6Y-`6#=Q=o-_6fRb~2B}a(ID>*XeLgbXC9q8zD1Ld^P zI8lT*3d_y9dM+mYuDes-b}S2lk3&hciB0RdS$ z|I##*+a1r(1;M!Zmp^UeS&m7=Q2mm0#B?b@#k>%=SCaS#qPi5Aw|Ya?!TWYgWl_`T z=w?&ZHq+96UvTO0{k}IwPNo`Y+#%EQpLRd^y``!MaTe~;&?t%F!u$7GY9IenkAxrZ@h>}w9J25Zius~F zymPO_Vex-QZVdxm0-~;9e-VvpenOsjzyj zPhW33h%M{I&E54}KQ-U%2;|MqLNiKFx}hkJ*=gv8xy)@0_PP5eki26bBu?|03yMsa znKRvD3ICK!#e6#S0$J2VDPFE`n;G=b{_xa#ozmeKjaLlL5MGm)wr<1(GwvyTS!POc_(u%5 znx2)?dmZO3zIGj+%~tH|8;i|eV%@nM;=JlWQyiH_YbY>oXX5NFZ zGJ4;}k4MMZZA!$$d{W1+MpbU7##-xs1i9}eEe6J~(hbRa?M(1@$2eRw`fk>27cJzq z!ao+UiB8^WqQ^Xd#ZiQq2MDFHYeXb+2YwiKdde(Jw_N{$yH&OxIW`IcdvIaK2zPCd z@6r+nq^;HX?b;u|y&t&DgS`|6oz-6t*gqMv6*(M?q4T1e8+=cB84I3VeZ0sgoVs4h zNJ*u#E{!|;p#MnvIqf5F#GaP{ifB0INHK+W7=NQ|NU~3*VkY6vEz-PK?CYYt7A8B@ zI6HpmU1Rm}bq(tz+GCMu&ux#E)y*>^sbTDA68n=ly}g875tBRtHAVdSNPT4;7dGjJVa6sH7_p+Pzo!0W7 zu*;%d-(>WYWW`s9NJkUTx&XJ%_#pnE8ebm9fyMbbV%-EsvyZ!lFR@Sq$ z$&y|N8;@Bj$N7m3+B1yLJEHc9F#f0O3b%_MuU4vE!?_0j)~$h{9m*a2MZIDxZ#d06 zv^oTf`hy2_spPZYzI+>75T8{D@fp-K^Uf|N*0<)qFE?-|t4}PajW7u(s=+I~K_-Af zY#ec5m(eq%zF9L@obOi8nU@L1h#^@m_Q7yIM1~6B>c>44X^fuH7j>}k!E^VB4WFt= z74b*>@z{^Wu4okTN9TKlok*J1pKhtpB9}oep@xu)hboV@M$hzY(uo(}y&VX9XcbN{ zu*9-uK+cV2rzMHN=hGnv#b4-CD$g%bBZprWm6P}56=0u)rq>yiXpL-X7c*Zx63ynn zCFg-B?d=$ka)!oj#{RA5W$Jjg%Q9sr=i~7rnRLJ?ch)&e9Z@jHlJ2%g2zA?gSJ5!` zKm4`_`>Kao9OJ-5bt=@Eq$$Qkr8!HdL%68l(tam~?Kno6(9YD0SU|5Kk{@f!xEN?iHyT^SVq70OV{R7kry=cM-Msj)`-^Qe zTZ+ZdqWh*=FFL<}vza&}ujCvg0lG!U!z&SMFRqX1-|;AD(WSG5d+@|z>+|-sim?q5 z{FOb-9${g5LT&fvRXQ>ms5>-vvF|ws*@`;K8Fn3lEe(NLrvhQI4SsFNb~m;N+%a}` zSnkWh8z*o;xmaK&vh#pIE%xYlDmOgS`!vW5mR%yPd(+vdH1hyMrq2wE-dMBC)%vsE z`gqi;nY&;4(HguhoUn%LtJH|_wc5S-m0Z{GmbGgvXd!h1-y>N{!aAaaU(r7jL9oiC zvKyYk6TX{3l97F)~A@OSY8I5?d;5+NCO9 z;8JDwJ__bQilJ54qtFG2nj%=1p5fUR8f^R6tVBmNOO!)1XN3`5+WWB297(5)K~VN{ zd93`d8IBn_rSR1Xqg^IMLR5k6N=~iKiN?h1PCnmHDm)MT{r4x70iYpZUl7lF#sCl( zD&W$7?mlWHTRH~8h^D5j#EB(po7s{Sz7nq+BF1sYpO8AybTy;M&OjR}VWl3T$=`)2 z&nS;EM-DskbcQK*cEk|t*q1+3E(qgEuV0V9<5`0MpmTIY76}#N3-czMLoil9=EKqn zVx$CQJ5+C~8i(C{->j3nV`y5v?=-UPj5M}Y_mK1jAqki6 zoTzA?S4I2JE47`ywpRYrkRDPNFy0W48z)}k_J7G(A1&0VUY{%~faM&t02yZAW3N;vJR~_dqEQ$A{M@%ED&y&3%fuccdEP5v|}S+ zX@=&<J>Q*6_pPW*DH5Zwj#>U7yHmcvuYd4`H zeM8a!g`acPwPy)+n~eVc0>9#V)QV)iNWCrDohtaQ59xn)l(pALz&j=ugG;A}M;%mR z-#KGAb;1iHx%9JX0jx#Bb+2Hl*EMla9T+>8K9e^GObNpbM_$c-Nt>8m%U{HO$^v(W zQdJg+XzoQ6H-oIXp-kGGI1^Vfl=ikL@&K)w>4F?5TlK}X=t?7@ z7nQX!V|@VtGQCskrLb~?9KZ~)>l+U6Tll_m56JrYITptxH`#_q6lcaEN$L3ZXEWmK zgL?%6cd}De=t#{U=EpUv17qw8jPYm9cw<08QM*@Hb>gCo#!V)9|IV=tO)Zv3HTD!o zPIP0w(G##@e#Wy<%c9{?Y+bXq$WN;Glg0^sN-)QNt0LiICv8?=GQzPp1gmbyza5BBT-jz z%Nw(h_&XQI(H+IJYlfj}dpp4=KPrCiO|>?%;h|l~Pf9=YKDX`4HSb<&;?HZ$H^nc9 z-h5TC2~Y^G8@Js3h6?Ly%{|xLyZX?0+&wz3mzn20EAz8BPeer4OHj6;OV9bHugiBs zVF4m6YtyZld6?OFygGdzc8hMk<#I!ZG`tJldm9_x>Jdwz})E8xDB8X$aH zw~7)0l)fus^BkM(3>`1h&dv0fP+f-DGj~q2|TYGHKpotP| zubs5_BCuDazIc`rUJPKMFZN%omlT82_pUzXE->wKEYVKcvme?}u`Y3*#cuh$ljgFI ziWsP&^6Su%+e;C`(Mf9~(n&h{&q@uS^y(LNOr@__wRE#1dwq)d{M)8PC*I(Az8qvT>$MM$_>%#o@=T#E;LGWH{@K0@Vgi{fRAW&9gB zf*)Hl5}O-g>}~ z(yv-4lK7xSEQYm-~ zB?;B%YctlTdix+;!*v`}UE_o!_u105QvPOG^Zl&JV{J*r;~nlUiy;{U0HSV}Q?9PD z=M@;%!`6r8*9iv*dfq{c$YKGCvviO!B{fs`sFpZoG(hh~liCDlR!K1?rFa>Q`yjq+ z-ldmT-T3uxH4@X%q;tS_ZoBINmAEYL@YQg3pUm~%_Q}`iGE2X*fJH}-=m`WjbLz>A z!xqVFFu6i%%zY$C;)OKOoduek9r+}GQkfJLtds}!@1LvSTqVAvR)nxj`fYgx>*o~6 z{Hjl1v4FTg7m#7=GSoEoq&s5Y3C@5V$b=tz2~qHo$8DzUf9LKl*%giJ zu)sTzpX1WoRL?hb8jZ%qGOirf2}M%aUs5vY{RyP5Q|fw`E)<_y_mf*&-&HhL+g0pC zy+&!jA&?~d_g(5G%{XN#p>bmI-cuOMrjT>qbJ+&JRNE(JGV~y8&hMquwg9=h0(;Jf zq$I;2*y8U^RN(5f0AU4W_W?^dI)Wgmqh5*ybDJvBF&b2J=3d%ZhKj-yL)OC%4c0BB zFuS|;83z`7yAAnQCB9Jn@RClczx_jkn> zM2%Olb&EUlK>;eua=ILYq-E z_h6GhG@QyYRbi0-IX#l37=xgjpt63seISRLUM22x7Z6VXA}s3}T&mW(%PT!($I11Qv5AWNdH6v6UHRE|AJr5s|q=z;8P6f!SIiV!gyjMI}&R(X*5zaxGzEuH=S>PK!r5>vCpMQZqnSX9dLvQK` z&%4U^P4ZkjWk0fK^1Oi)H`{v(WSmIIF4Al`m63W7Jdlb>A3Sxk$X55zW(hZa97|C>~ zp-)@_GCI~yp{PO#5CB?z9!O#~;A1v(?#8&Q(=_u+T|AsiVQS{k+WcC?y6lUj+yf#e znfmQVu(BEuYQy*bRtZg_v>7Ln2QLu3y{rGNGBlC&Moj*Ez`T82efjXsSLqY(f)l&i zCHM*NsB9o_KNNs{)tEzx<-P~Qsl*p)VTCM{1>boZbwuA?h5N=S0bC6QNxEVX?`9L- zMA|jVGWIpY^Ne(hWV^_MN}OmeFtNCo*#*cFfDf9$agtQ#Lz>T}2h|nJ=-{CYAQ4G{ z)i*KdNw0HRru$_*L!V@+x4nQxX*gn8p!ep0s6hfzljp66Q0OIZR4=fnD?CR-^Nsj|Y3gQbSCtuTu3fvMg|F^}yw@Ce*E!531CVj&=Fqcba+*OJ~7 zs_hb;S3@_?>$Mq{_7ebW69*vYIKgwYQLFcst&zk_98eM#E`Ry>st2S?JjqbqL$<)z zEG*>N`NDNxU};kcJ{FVh$#CuM!-vaF&c|-gqoOegKq0F&Hv2^dhCQ?B+Xt0Ol(3pD zm1!CHKW^4{J>WM}NXwV@uk0!IBWC<;oKkE-O#Ip8;bbw1Yq&KPJpLE%+o9*)@rnK` zDu0NT7Zthe7qKU4`w`~;0jvK~Pp+M%NOUi82{J;{t&8j;vITxg0hF6An>M-70M%}w zd2Hix1Z=YlwrTchJQg5Ob~`$ELxzT5Z;UEf|3_Z^F$?HgUj3YKi5>spmM81BxdhuJ zg>8O4s3+s@!Z(bbX&n7Z6Hi0l>L_|J9yViVfIrDHcq*XrmG!qmVenO{9wNBIBwCL@`t~=COb1HB5q|6YHFu%>_+of7 z#2(^S=HK;Nl~v)NMrd4`-tYETSB`}E!0c~zrVb8Z8KnB z)f>!S0REbI%zmdA)Lt7_FDGLj6=cj+(Yl|n{m=%Rs<|mOn(8sB`_4bhRy&$kUr@o; z$DSAh1<5d} zyGFGPq`p|4M78Dd2hI$LGmXB?{K$UT7(*QO&?|;P^knsjI+T5ik3n^pPHfP7B{GJD zE5@D6Jb`;f+?R7-G!CxxB6(`Sozsayxr56-X)|VzlMbFQ&1v6bx_oB}YZK=p_Woa045bKljtVeyciF}IxmR*yAiOkh0N(4vzut+X7X;8)QZYTz zM7YBM?M!S~qKjz6ynNB1_R2sEWmC)!R}hXjKcp*{e?m0+UwaH>gSsx!)RygVJcdAo zO7>wq(Q#}sC))IxCEe67*vcZ6FdemG;QQQGZ}_!o*3c` z>P;1LQ`PS(${(Nb}*$9WZqhEo-f_3j=uZ&_} ziPJw#I!=Y^Y3pbsV-;HRMd^OMyF-k--{7yraZ_ywOp3oO{P1EwFzqBEbS<@ zKe1}HDO}LN55S$wW@HOS{IjE_R&2O-;JI15fV0yl~i_Ys$kAYdCqET0X9`Z zZd#1(`~yp`%CxhaoNp6LyIE51>|0gVGfFuM-)(x$SIgf43x(zlDh-69O7W%e_Mc?9~8k+%FYZTNDsWfzKQuqi>Jezt=Cc>ebe;F^R^^<@>Nap^xlr#2ceI7 zYNFG=`PU+aD?_B++*jcu>?jY9_ce*xae*wZ8M{U{5aE1bQe&4qm7XqB1bHV>~qY*oQwDpb8lvMp|fwFv16YwyzoA|B%|NW9u(^wqq zsf!-)Il0WuzS+vl=(7*rL>do0nf$-D`Uw}XcWs^iyUPE1oFI_Wa*0fw^wwBqP#2tBO*>k@3JDZ@aqW4iS4nnr zwLa6ZmNr!0EyC9JlG5_{e8}Z!KRA8#Py4>L9^gA0HA{8M7ArCtXw0X9-bcKNzGN3TpL7&1uQXeIA7yBsAVF+_a5b}Aik(;RS7tDx!JRj;*qS^57={N3OE7dzj~ zwFIoq(zO1w=gqJY`XaMN26*{#R<1{H7z1KG>^!3tMVb%D5?6gErYmZ{s+m2xeU!R> z^Q&>;lTX;+-#6H9c6yCp``4M03Suc|Be2BYvCb>sAK#BjZd=DX_m15BZVOs|Nj>=f zX3@sYWXXRqR@R&N&9D3T_p_N&L#1zihOv7Ove&vra%5&Q`+l$t(DuE?Zq}wfTUZv~ z@xK1`kA_R6m)G2-@%+MkDEoRikci5_J`@JADLE;6u0>$Z@0iu7HiYo$d7bn)U-^Q#nYJpKa zqGyKURg`y6e#IJSy1s|ridNMIv`wAC#lXhaULt=awcg7ATh#w7QZT8u(QW!R+^w65 zz3ocdvP%BdoEl2_=F8)NayF}v$@XDFbxt8gXTNDq$$ky+xfNUTDu})^w({(s_|7;N zWH1rmGFI6?tm9yLGFvSZqR+HHUfC~F>hpV7MnW>t=VsVxuCPEtLH-M6)7b4<*{=Lj z$=r|I;RUm&ETD`jmow2owj(M)qk{ZL>9CEB$La-kRk zg3+0zQ0D$ub%|@EqV6^_T7p*{ny0v zJ6xr}!`9q1$FJU~G`86Uy_L>{B=-M|)aFhmDP&i{9=s`xQit4deC0AAAo-~~zHY@u zbKvDuo;jiT6c)6m%5?5esc*;_c-i%w1aJP^GgMjlv_->^t;Vew`*qPv?KDiQ^BQWX z1ob+n<6~PVXRTPf19_pZdi-jbw(-u98B3)Ja6Uz=8Yl3tfO0ptkXbfNa)f83%qQlk z%k@2<``){Jn1wTYpT^NIE1+d>zv&6D%A=?+kvKvLi!y4bx84teh! zGc9bZjvuEculY1XGwn&Cdhk!Vqa5eLce_oe|AwM zrxf@e0@%Yya9po8o}EWh=qnVs)W6$%5TZc`DyZX#6KL`fNy0kl>mbcqeqK|nt(%%d zi8Go)-xvZ}>-wG8SSP%KDo;UGMFn`|?7Up)<~G8^MD;X`tWlon+pX|O%V8!=c`aQ+ zZf-8-$aOdi7F^{Umjzo?Zck{ks`NM{87}bopv_kh@LlKTbK`K%S{Q$3Y>NZJZ?`y1 zxxMG@5wT6|F)_>ITRGqCUk}>>(`hT_7)gy?guI#c7X@!vZrwA#qd^laO)i$tW8|oTJN>ch;#(=rbGQxG^F;&az>_zvqf4MBm>W+iVj5W1+zu1*+ND8W0zfH+oJ#W{=M`7dMOBruu!} zFv1wQKS(NWfu6^TM&l7gbyO{pop0-ahUX*loP{WVhSibM=lsJ16J%gmih z-2Ilt1R*PwEWg8C()`qEHb5U`2jlS@z%9@l((yUYVx?H>H{JE=Vx4BO%}1Y(+AWUp>SU(6I- z>`5!QcgPHdBgLK43UW5pQn>!S zI!`0gd|o;Ai7285*^q6awOY@W4ZNm@BYDlWR<8U+ zy&x8-ADC$P5>6N_aBtosg0t7+QOUf```+IB)g<4rfn?GB63N5IbZlxP2*xCU+rUYd zXEgTsU?m0aEOEt?lV*47-Q)+0U{i`9sOgV9%UAn8d=Z1nfO=pY=~aH>q64T0QQlv$ zEP$!w!^Tw0wY7uqv%A=3$8$$VwIh4V487+8oTIlPH1Xq+cLV?`5hTyX7jkvkXe;sX#x1Y0^D~cq?6kq#b6<nbX2$cvQ+LC9&OPE{*Azj_i6yKbJ;Xk|t!eC+IKoQ} z9kR_}=0MI_8gJDh!$L>g>x6zIcae8C(z*|n$2X}vUUc0=Url*~Hqv2DdA*~N-Q~^M zMq1nTF-JZ~ws=5|%=oWAsi1CvZl~F<3hTKKP#Ct4mOJ@FjOS<~2=OU>UZ|fJ4Rp?(#d$2pR)7 z;Q;-&SHqSL8@lQhV;$(PZ}njT&ownEs`CrNloO4G>b9-qU(2 z*t@)!sq4Vclz(f(lNm9QSC;@6UX}iEbQ;k1_9M7|QgeezYS9|@o(f6Mpu%tt4Sc?y zjwr@xA)#_=3bc32TC_jbczIX!lOYQJG1l>(GCkKq*%7X10zMy7pdx@*|Hd3P`Ohqs zx7QH6p`UHotZTEED5@{Xb3=+o`hbY668056fYzfw*U#*m8ee7-G z^b9xQpUG*{c>3)Zo?U=hsPHzY5xC?-=8;T1(Lc+U55AaK4M90gSOY%atKFYSvcv+` z8WxH70nUz@k9jBaZ^_0t4JZmbHSjL?_8hRrel9i!0rR-if~!&KB`(#exB8G7{)J$q zpNI3UGtj!|M+wUT$u(ct;CV=qYqR1@-9z5c92=je&R(`gkCxVXH3sxF2g$TS5X(o) z5^%o~K=QScUK-IcscSJFP_9F|Hy}+Vbq z{8blNVHTBv2nFikwDU_XKRyclqDwcKEtI}ay6|wk4dsr5HuFH4n)xiz(?xes(}i5i zBuKVAV6~8;`2$52lL1=eYUC?{JD>m4CqL8hzA!6+1N!7-dcAR}3NF0VA09{DdSGxr zSs`_Y*)r7?yi45bJrifSYFDTKMD$(W{!2cuCzbkr`I9_b|X28ppeSpwX%#PNl> zPLY&;Fzssc<7^tO5z#n#Qi!U_OPA+DIz7jwhZ;o$4Ze!p8e+9z5;69Sp_M_bv^Kzg z(NXTClS5rIylWINRI7IDs@ev&B4$&l1=pmiU;n!J!rM8 z_c79}ma=xRr$I8hK^m6Um}rBGftgogN>iU|ay-ZH35Y*t={gM`Cap?)cYx0KjEogv z1x$FI=)qbORNg~+cfVbyiFB{^(s@F!Gzi$mes{CTU*i$uj#VY9VQvYljM-$*}(tsx>$nY?M;=lSoBE#iEPqeP<9?? z*gP?;Sidn0D74bJO=z5v!zW_dYPJ*)FQskJv5ivvn=ZIQYOT<;rpIKhYBd*IS<^?0vxe+Da!ArN+d zpxi9`9dhccl%s37Q6%1n$~Eqr<2H(^x}wM4ida|K2|woG00Zk}t?xowCU+XWrm?kT%80%R=Z zj+2k^^u^7jOSKv~EBEYvu&?9Cs(WSE(r;+-D-7p-iid91!(SXBHDeN5XE_1@pf<8z z;3pdp!Yh`OyXdDsQNpEpvX^QddILB*V^^A@66P~;Av70k`WlUH%A@bJd=#u*_z0oJ<7aZ|u~3v<&Z zn=~P*+w$_s0qwOsR*437S8bg7r1a;<+#<#HKWyWfjO34P>O90^08X7_1E*_r2G6*u z_l?SWnsWsi&van8-ot)CkDcjLGH7&zee)lq$MOS3hBrY&lQvZz5}qv;jf18aK7Wr; zBYEgCtT@NA1ER{vDf_7w%7Js;2T9BeC|u}=3YYY{f(;IissH(FtBja*a)2RMIS)qa z7CV9DHw;<;bj${TgOI-Z?t|%gLNT|1RNtQG7uTsg0SO`Fp_UHnRn2&9`(b{dCewVu zgSWu}(fxA!|KDRbL^(V%n`O9P&3j_Rh%LuoM9FE?C;8IB@=IU~)#?$6h(#lFiriBF z`R9Q{*S?^@6`-Xhc$i-ViZXF+3N*0gTc`Ltie68HsCJ3?W&>V+MZ?b3D52 z5V>FXoea|*`@MuiicMHxw~oTHg-20-C&eD<#%JUirQ{6icty>E$7pQa2`IlXw|E-3G%zPG z{a!I`$AFr4AX<)~uuPedFvQ0m-c=>N~}%V&f2L3fS&S3BXF*3{Rcx ze4m4WBqH+4cg0X|JAKg=jeLp^FE_IVE>+f_N5J-tD4B#VJ=WxUg~nSoAm2RjdN+B; z8nlep4-mo|sf+xwiJ5t|yOP#ttn1`t_@7r?z6zE0Q(}QM2UruZxnhV6tnQTNVDh0 zL!+9lEPBQc_fN_nyR=jbjqw7#e2DI_&v9$1#@XpG3N4lt*3*fZX8fA$ep+}MliDc3 zD4}OOv=$TC7+G{bUD#t%6fZJ#{F$hKCGwQwgO;TBS*8Gms`0d2mIoAB3p^uNjXZs? z3|!UcNS5iIwF}6TLnbpMI%z+sO7$c(jOo^(+D;RyV{KYXE(t=VgA&-$oJ0I6rI_m3 zT`ZgU_WoQi#7(U6gNSJK5h=1TA+^SfzmZ5pzl#Bu!O4p;VCv9r*DX~^NGUiL zSRckff!7E&$`&ws%1tL`)KaAxZC>f2-hsv@0lII70CSop6lPK-^im5+p)qNQU#)Qz ze4g!6x$A79rb`8d)fqnpX42)X?>5ILc({omeTs|@(zAHFSdB+q2gwR-+xwbyF%^yR zh9q)U>AXsceK-6%aW}v=A(kx=JkDjA(!gM-QC88YR65**#7P*O(F+_47(C4)hg7FIOsFolmv7#Tkq z+h76(qbLO{4e_>j&|f#aNzq0r_Cb*sQNPHYs9(kSNdq+!kQ~1(==g(OZjL<5=(F~BA0KQEr&;x)-Q3k|YwYCn{I^H)-E1KIr?~W*ujP?nF0K^7eRtN#M z3;<%=c_LK88P)|F)eoPn4=*0)e(3c9m#WOD9{pXQ5Xc*9HwbF}=&_ieW9Z1VC8O0& z{`yBP0I@F%ra1CJ2&KAH3J&h$(+K}lvg10#t#Qcx060P>{p;tibe-_WZqa4xP_4bG z=(|PZQ;tWU+@0ZnfI2+ar=N!3t3V}AUf zuqTokMdQr3A-BSRmyD}62SDt-8voO^x7xf?<+BRaAM1BXgp%~E=bZ>UbmPu=7`O!B3{ugx_d_FeQFtIZV)bDQziG6KNf3OO?OYL zN;Gvr>?E_NBXeF3wl5HgO`H-4sW<8Em?R#bUW?E!IH zlDO0M7{&$B79W+|ry|FeITCGzdle&ccQ=yw~0O^zcfg?l2AFOh%l<*n%pSgR$Bk2xWJx!k*o;Jph621 zi`tXdhYSEuWpk)dSTHiqa%Nf;5KzN;pIWV(9}$Z6pO#TYIJMnG+=*{a@BvoOgrT;x zB&@MU?S~oeIamEape-ZPHDM>C+!wde_u^3uV{Mt-lX#(Ur#9@4PZ2?F=F1PXpZ+&z zTc09sNsp`{w2YYf-&}Cfk^nX9!@7_;T*+ihcoOiC}^gyr~^2*BgxR!r#g*tm{iLss6%a|nK4`J@nH(q za>g1&)?9KmkZzem3R|ceh~;%rIbY}Z40`^nO43F7e1|V^d)DojyB{q2?Boujlca0> z8wN}fvX1iA2@X{TFwL1Jqr~r1W%)MM%-n~kzYP+Dbc~qfdS|fe&Q!(LVdc|#z}*hj z&Kn(ohxuAHD*5|V14(TERj`aA22z(^bV7PDH@*!W3;gkM+V&-na6ZSC$;1#i{S7?B zvTj!laV5M?-aityi_!WEQB7GNUZQ6(Q{peSE=>~a&C13Y$gy+G>Lu~WG4bI%7>@+{ zS%#q0auUl%+EKAcOyOaz{eA5U?RQRJp1P)7V74XG%V70Y3=w8P22;|mqWzkc6=0Kb zH6_DZk?R6Y;1wg)4LItk9$Tax_=pkUte~B3ekw&ub)50ETA=QtX&00lM2wcOuS z34LO!;!**g2{FO4xbxpVU1_qQW%_fG=Gzl=@x#YkUGk?9$=wJYg!9=(c*M656QC9u zRHIpUUW;ICf!>(HYJn@_rH>)cVC8*QrEkGeU(QM-G|WGI(R3S7NXjjmHeRl!Hel+v zlBzh79vMH{v`piO`qFg^YNt(KEC28%yW|=9o6ao|p1YzTVY&8aY%Xv1eo!W$UFWUm`v zXHcHN+}8=})PeQPXP?CKcD5uHe8=7lZm&}yReGrgG8-jt0Pys*fI>$nd zI_J|(x6q#IY#jAK$($TkXz;rxQ(|BzP-=Yz@bp=8VW*pDM>aChCBJj_iHtI2*#4ph z`yr|`|LMC?EMdmfsFz+o!miE$pJlHCOW^ICHtjHGjFzObuW%#}>694HcQ$1&)y^u$ zhK!MURBAuNg&<$4gwMI7)v0EzUCCQahTO)h{;p~LgQx3mNMjmnpxmKQu6CM>x5K~*o_W-C=bR;yWf+5i zb?kh*b50ZuUON5j2tp>1)_F>sYVug*gZ2e6w|X@3(WL!`p9Pk`@mkYZ(^mjYbRGn( zCfoNGse(LkOMjPj5H$YStX_v^Aku^pg=X4kGGy}K02oj!QXuzx8UF@pfZ^i@lD|S( zfL}Oyd}3twk#65@<{iVnjC+bl8bWgjKDUPj+4t=*Mzq^H=&Hjd>8_=lXf+w538>bt z36tMPj{2J#^LEAI4$0T|4e1!3qse1P>pSoIDLtKfIB$Z4CfII1kf-VFz%(%EK}?y+OR`APO|~ ze_8?)P|-=UoGc*klm-&l(GN9*C1n+Bnmr~dOEPAZkTELJtA95mlssJqM`^oCJUOk` zj8{SWXJ=egaD3O@Z$KOgVo|E-rJ~$bql*k>_;%MMu*raIAFTtm`9xwWMgX-;v91NU z9D$<&W@!Sd{Ig3mw;N)QaD{Tt8LlY;uv)Q7V^|@i5~zsrS9Pdw;nQ(}}%A4k@h~!V>1>jGjlCXIBdQCPJ@uE!Fg0@_$?Tu&VIi zG0KscPx<($pOVxSRVFa>Q3oa@p=*?&eCblMHpPXm2#5nHW>!HI*G7dxS>v-!O%i{e z>W}^!;lr46FK>-FzI*%S5y&?@+7-Fs1!@CX9QNm8D7_52}1=L zC5%p&myd`v<=6g)bw)r3EWZ)^c)N*ZSpzxzfa5oL0r@CHaKC7unwod7Ytk}csFkq# zRz=QWl+N4B(5F`{0m|nZRd*20ICub1j47ZzlIk<#ZLz{MMqX-k1x_p$mr9I&*iufO zrAg;XGm!NSs_F|MYDPpRKQoTgrp!YUewv@!;lCphf=mxgv}Uaa3$B{vNJ!^!ZPZc6 z350&be>|*WUV&JKtH?S3;IGA$nm+Hc#=HE_eMjFiqA7-;8rCTadx7jF|je&_Wqd07JcuF znqOuUxmWrG3k_s$(x}Sv7wZLk^71zrfmo!`ID!wD;qUo@u!%%Ye*~?P23Iy45mt?| zn!nW5-=0dtr82VddKjR%hUyIH;E_ky6Fy3TlmCoTRl)-h^eSd0F^5$Ml4OIoYCtY% z1JKI}M^BK}O(r05|I*zP7O9F6$>3ASI-%ZV9l$JiU+Crvac?P!D7tb0Pv>DcNkZqZ zCy520e%}mq?OopH2N2#bswNXs#tReP;dl4%f8{i>x;HuUjaTSRLTXhctuj_jK@7H3 z>EFa)%^02f2(Dv}4Cc^4i&dHU))@R@a4-gujhO;b^O3QNbB1UYve~bX2JxnVDC~QF ziwg+55mW)E-_(D8R$51BT>4#%r23CO$^^J6{P2#kqKt9CY#Bk5`=$i!b0xdrM6DEP zL;uDQ@6z*Iam9AE{+%_QN5_AmgIY!c>PNkd4(vS-k{ozF9?cxvK%E7{JH~o}@5x$7 z1O;6sZBMC@lvRS;RCkQ)peQ>l9zPwEPBqy3L?n64RApYVKWWR?5-qA?qha8jYhRK@ zev|*h)msKc;e6r3KLl5hUXT)yjs-!MhNXKci6xa>kd#t#X`~yLkd$t46%_=fI~3`X z4h2bR^&S5Ay`iFwYMIS;(%2}Ux2FDUJ=@TyI4Qr5c*s&$f&+xJi-qVCTV zJwscM_Yzhi9}2Upaoe)(K?4{B_h}`hZc!#bueF`k|2*`>2;W$8Y?*MFWq134EL%pH z5xnYO3aoRi{`s5n=3?>HS%mLMVQBUDf;KOEqL;y1|A>FQp%!v)MD66#Ba|goq|Hem zVTJ8kgo94U(Q=^G?tfs9bskm8pf%w`sv8S9}?i=r%oYM65 z?ZihzGi~$IXVH4qd4~Q8W~D>>9OZt$RDLPtJCFRhy{=en^j(h({B~1k?Itx*{)%sQ zjk>7QZzL8H2=>b9_j?!<#Bj`c{6{BDn4}J!_=IS*g8xCpkQao6oj)4zoPDhZ1?H6W z07C}weAz6X74r-#$SILvQF=_LxeZ?PhJx44h~@IM500nIsJBJ5FI-?~sQ8NK)n942AfZC4IIOy}vsjox>ul=m$IE20M$?fjOM((Rs154uy6<5k18J5qW&s2*Q(c5ylO)26By%C;!C z*s*el#~R^Tl^6Au^VDu%GjK_s^wm;IGsXCSjMlZ%jPB~L;xLIi$A`8X#9bvQ0ro=T z;iU2dEz_{9x3jE?L;BB(Ge@!&w9N8q^X})DAnJBRTtsu2xsSF~f~TGdun$%>Oz;nW zjY<`!Wq2p`FaUBKhtmj8ho985bhi=yCN+IZA#60&O|?NY`Y>NOoO>wGz6STE_u&Az zPkaC_%qSx)EBHlhv2ZJq_3-P*6g7Y~2BAA+T-t5|FYT7-;Kz~)Sz*UkRR(rp!lWkP zN+J6&K6`$!sEzv2ZUFCFpE!8&UzWC))2~U+YoBa~vl6jw%t2PuSEuXX8*x`0XBj8( ze?v_H61GC{nxYM6-|Yoy(MR&yJkt6Q0@u~&XB||tfrisH_$MIa*1?Hg=O2Yk!eaT0 zQFE0;ubWft>V8C<;vD^;pdhWW_BP%~rXtsdJ#fLL=|5hZ@n;iUq<&7XU+B~1W_+!( z9Lr~(D@xC|dkbUj9T}jWMk&zaU<+DdGNQHv5h#Z6_eCi;y5JVSjSpISmaWpYO}gk- zQAG4)#$`iB>x?%O$3(tK4!EFF#M(Uu{_4#dhNE~%UkZh~CnBrctJ2Bt^ZHcrgP#cl z0+%JmUp2#&;2Yw%ue1y^+N*%6oTQ7ZZtIg8m5OYspHE3*!tzu|jCO(2%%2yLjeL3W z2Q*l$Mt}wj&`AV&#+{4EBt9X^2Ti)X{!rQSv2>!E-=t0NePa*Juw;Rt;VF>Fr)ua6 zxGb^#?^XD~4Sc}s5l&QGGz3K`(PWp>DVmgxS@C3#*+?w!HqDm|SreS69kB((vsu2% zwH7(IUuU%K)8(P{&2LecwBkP>&y=vtw_-d;t{d9w+_hLS+JFd-4JJ806?%OmsvkhG z!+GbkZS}k|E%}AIZ);f=&jlW&xi4OQhzv1ZD^L}H4zcIh{bd}OB#=~|RCMDNfX zv8$tsItM9fy7HNc+2e*9qTNFW7{T>Fc#K@9N8b?759jd3x~rwLcHaigo-jpeEKPk1 zFViH2YQ=H%a#4DASqvcean9$ldxckYi`G3&vCV$O@D%3`nz{3?IM=Do`#<}e3Oq*A zyiZww6CTq|rYMXhlLoyV9|`}U>;bRa(CerNAj-FbmaO-L(#2X-JKS+v28& z;Zc3upF2buK!<7Nl`vT?-@b!^Lgc$UYYvtQp=5Umo`GK2htmXkgavWsI9{o=b;HR* z&6_ky#!FvnZzTP?gZ9ORG0O1qeBler6<^}*Xgp{UP`HOv&nyyL*1Om<}`zFMhuffDJ6f_AV#8fKWcw$L% zh!$2V;~4AHEWr>gWVwoyHu146%YoWkA1TbW)7l`7^q ztqiHR+#I0Dm<*5;$9x6+nu8^DB;s=w$(AA@1w!a-F_&t5O#oN=ChT!Q zpzpLrv06i9(OE-8Z{sccN~?~GI$B{OqEKkU6IS2=`^|0aE*BV^ePF0x=RjdlQ4bz+ zM7|tz{Bb-CYwghi)M;&wW)(vR@NAHWW0@=9EImkIB3;zmetPzmJ?aJw{F@>GMYJHx zo|$Nz?IUG@k-%_zBY<3|7~3*l)5ZY5b*qM5L^+$=M3*MJd|G|$)tD8a*0CYBZURLP z;Dd97_bj@-Ht4~nzSj;1M$;}-9P|)=B)SjCdH?@*pe)cobPtS?TXw~g-Ma4LB;0nM zFJbGPTa(6mF|zuAyHzs3ap#s0wj~}@E@IFS#19f=2^~-oHscO{73d(U28ja?Hlk=E zoB>ZV;DiIJePoviI{a@WQE*j}I4EW4Dr>6PNRP3a8F9SZ46Pt#=6w84buyhRMn#P| zdUjc8Y!A8nHT$|Mj2-_tezW-?1atTp2K{K9&b8%nqjc-}4SK}$8>WcB8 zJUv9^;yWMov-i(ho&ao2z|@OjbXpum@2)&H0s2p)Jm z@4o1R|A@84__b1PA@4vf{-7ELnqu;aVmUf#@21bYn zMgpAja-24A0d}8-3gOEHsUuzSj6u4}yd3QHs9%r0^NoJCC`iBOk#^&iF2HF#`Ghu* zJjx#JA;whfU>3L@_sH07jx>%KJ8c@-ZN9eKMA>cHHjbP+Y+f^E&eRB7HkxL7%x(M= z?aS&OJ^fd90{yo6mdeuU0#k&k+&&`mt8{E_5)x=#x2VJ<-iqA9!(yeY&5KUtnEzIU0&-pYV%NB9)c50oUs~x>k zN=aHq`T5>gYp#REX!;F`kE;CpP!GnhT)x>Kq>9?vRt}=cG-C=DV4IEf7MvEXzvYOi zK5Z<1{P%2O!Dr$05_~b&N|HY}S$Jo@}b)0yFld*-0gjLvCE_d_VB=Eyss#Jt`Lfy>{qlP9-Ik%VG^pmwG>-#~);D@Rt8OHIR^N!9AHVDG=NW(tB+sDPY4lkw=C&-Y!7h5KoX|J=~P2=S^$!Kxvjd zwcUdhz&c~9#TM(C)p24fjcy;Yh3OeaZo6#it!G`c;Lyy8tK4~z!xPNHku`YoxdIdG zr>qA#n)&rcN#L6~Kg&q;Rzva!QL+jQH>hYC*@u(pQm-3jUSFXzOj<@;$>h*_(^8yWkatmE7Ft zP2O07TQP@!k+Dj^nU>AwR@7^E-QP~0NEkUkk+5*y^c8%!zQ7`Kncni{ZA>|_gHC#8 z6v6!&X4O0T2xV%!FR{BSfQAUn&!O|=Zv|o&s$%CbdLsD|eX5v9rmS~U3ddA*drLW8 zjx>!RkLYuuW*YTlkgX(9G0$`AK748NjQwR`AlBlTe+9RKUK8KE_#yy^ z7h!;hSjsvA8AlHUnMQDeq8@lxHN?q?Ny~Hs{mWCkF^{Zd0ogDfao^|JPE8*on$-1s zVwo%0+axr`RE>yQc420hC+In`8xWOaH-98+` z?KOb5!*_gLVpzP;)5xuQuQk9&!`~1N5M!GTu8+T6pkoA(bAAMzYypJ#<_Gfz`A@F- zXV?HXI02Jlyv7E-tNsQ>vyVfT~2qLG6PEE*Sr#t(RC1y!EXkl+7YF;8*1ip&+SEO) zZuAtxYKXBIRj0&K+9=^C<1s<=7!3&euEMF?Or!k6cT{O4{h0*+4B`CJ(Hg2~?sj+1 zVq6>@`V0FyA5s8K> zgr{0JC1jzMxu{%>;4mMGaFPJ@?e^t>EZ1qii7t*nPjQ7knH{rf!BdJSd}MthE6{F* zrp&Xaq%^4vHo4Nv4fFT#Dg7Qmir@=)lu1L;c=ZSudH+(ul{7U%3;0*|Ais3Fuk3i) z^!N&X2#De#nk2%`h(vjf7oS1~rb|^L{_8v2R<&^>P4!8Um6%lL-ls2`@c&WYg5Im^1T$9lQKch$k$WsgM^Ba$NuX~HmS0v z3hptfL8cm7a29O>TP(={%qEr9FKNk|vlAByc4rdGuMda)j8?cn;6j&#l`dLvDOn}D zVw4FD#;2sG z#n+?X3n^6N>IyM_nt)fJ6^r+Io(3N;n-O23i5;Q{=Y(|VGvQM8Ga~oz$^FKdAR|hq_rB2P)jaG8~%t^ zP(>WxP-tRhxA42(Pz$yX#u}f8*n#fP3F>+1(~YA+;1CVdMn;zA13fIVoT>Cz@UuYZ zfA=Ydtp?j07g?Ap1Az$yPyP-Csr|ua`IfANw5VV{O5L*K70Oe>#V?t1c7n=$e{>Ct zNh#XU2)fGovX@<(pq2xlxxR)CEb^2v@%$#yEkuy15-Yt5+xWHUt;&rB?AVhM35J|7L|1wp9UKgl@ zicnS`&i@_?RKjs1Gp_qsZC%VJ881nTYs;}O#P}Br7|KiV z5O4|7pUltU+baB#foza%#lD-;k>l5V_dbApxr;BDqJ7uy+vK853YhR^O=<9h{J0@E zaBPq;92}xJMvZT6dI!WPK}Zp<2hSvp3E#q$7GHM8DrCm#Q;Jog27)13yGxz1`ZZ~Y zow_U+kN#kT$hB>k6Bwk&57^safH?Xnu7fXDc#roJtVtVxgwWnO_vbi&tE$~F(7f)IDfk+}C5}vZe#u*Ne z0V1Dr(-oslxUQ#~NAQjO$>%`5Y{ix~tD1JMzvB{@tw6=OuOT^04eVJB-vz%#Ye3dy z!$>&8^?{XhkW-Kt0Uy3L7$t&kq^cc6NS(ij>%(9&h~SDq^+}XNs4+pHVhB+2bG3%* z&_aa9RrpHs5SOf2`@XebtOQ?ZYiFv9&CV@y<#qa71Tu$@NPJ62{@bBG?I$K+*NQK+ zwBS%B;qOTdEgXsqO&&@LZD@f4P27hR>FpiYs2C`!5l?W_vtAY@UDc^b`-d^mL0F__?HVc-Sn`wBd10C{`>k1Oi{6Q?hlyT0 zY3n1eEsdml@%*MF$p&DX5**ces|__CiR_1;ycn_1H?o`!CXdd@>?3G3BYbJM(#VH| z{&Lq~`c|cEX_Au&6>gSh_wLcdI0`7jZV%cs2S2HZc3~yBgRMpU)r7;APy03_&a8f5 zcM-pwc?vKJgad-ozL=(0HxBQxRut~691KU^)*xoBXuu4w_LAO7aU)bvSvO}E)b9Cm z6LHA2ocWgI3@yl-KYn$HpA;NS4J9L3Q&W7Xa30fZq4Z+yiQ+@+N-5?imlR4WYg~#C z_2_$7f2!Y<+`?51^UF@4`5It|v^*?DLns znLb`!)VeuTuPMi!#k!CEl!AtS3+otR@HaX!3gaorDnK4oxqLtzdAsjVL}Az=OVW37 zL-jH$FrI{YPt;JEnF}mf&G%|b?p1Ar8Z&2C6=nU<+JPL&dRi}$!uh%e^X4C-R^nS| znq^=7@I&h5G)Ux{V}~7=j!8ai*Nx0*#7{=G(h+7g#9@oq4wg30wqWRa;ve{^Bw63! zh{r<L@$%1D9^f8O@eQ;D2(mEZT)#%7BA7 z?6UQcs{LM)wOLRB@Z2#$6!+2(YC|}SI0!uh5@D~9k!XYvI8IqUDD+3qgOCMFBebW3 zFQv>yD?nP{*;dO!JFacqJ``?xJDI#)Y9(~CEE8Sar49JLYRd6B_ADHyUGgSn-SW!6 z5<)A+C0MpnE^XWR$a_L`%kGBUSQUU& zOcTuWh;TmG+OCt?FPo0T;uhtx!5DWvH0@YZMj;d=VjzusNJF7@MNCXfrh$L(&51EZ zrAs6c$=;4sQbpjh=+Iq!Zj?JlW89h`K-$qOyf?JJN7N_X#SxARW2r%_8$r`ZG^yH4 zk*erBLj4sjYWtQ-sxr7pI`lZ7Gm#S(WK%Y&K{N9E_OMGP(mvId+SqVhjKo2GBzb>S zS2jrdgbw_I?V^e(9l7lHD=^51s!&ao7}cuYz$$i>u#9&*4e0z2(D|oSylb8FSVA#{ z;7kTlau7gkYs9`Uai+&$n45TkCIuiExGdxf&I-BV=}bakr%w{G zCWhZo$H|{Y^ir$tQ}nz7*i-|;PeLW%*(bVK4)mj!wrovreAfL1&`qWzfgUHmr=U=O z;W!0}0i5F$q=19r1pf@+Y~KK)vV*i2N0RuiS=%K0@+vo&XcQojz}kd~-Z2g0ff(em z!k%P%5}J(A&{}O6;@T=kE9bLu(PJeE{=q@kI0{lY|8>c*+nj;=YG9lo#XQix1qaZ5 zCZPoi&T#}ln4N5U6jF8V4%ijH;s3-xLJ!qj+t2VM>su7u!*c$Rouly%Ok@Axl0Q83 zkf))Xf{`KF52Q$$Q@guJm^0oJ20etc0cES&(YL>L)Z*JoxjvR#fwna@3TimO8o9zK zk&tU`7q+N9Q0PytVdbLFXe4;^ZQ9&rjN7740(Ljrho12OC|mBtD$&Ho|;N|4Xa^Pgd;*6i^P=ST>-3NW2-`KK`pdw zpVSiY;n`U|3QcKN`Kheg01)&0P`ZFE@c~$Dz%VQ@^4Jf}Af&$2%?? z-1!5~IXlcBnz@%;9!<$OcR=(MBzmLQ#_dIr#7Cb3yhqP3)T)vdpV=>kjzesn3R794 zyFrdOx5G96q5Ma zd>4|Zi*@5m1r-YTL2}Y^2OrPW38eFKbm?{k!wl>kdlbT|CPrQI%0@l%hA$Z@4*RO; z_6Gt$F+~&iBn4h_m1gHHg|Ch0Q*k$6URKMZ8Sz=`=Sp)Ca?E}fJ!*WM&ebW-LO!p=OSay4Zk z(vyg>p|e$!ZI??YQRf3zhEYo{N2{M)@5w2_DsXu6Yt<($X(-u9%L?^X7zeSMl$GdK zjd@72Z)fHZebR@stO7*5P7}69;?2pB3b1u2Jf*FmJxGlm9&^Ult=E+2OCpp}tHo7k zu!EJ3k$bZLpZ6No>do9U+m93U2=NTs!9;THtQ#YVTh z;f8eEUWze^dVLoTJ4LAa()*8WB{y(Hal)s@a28^)R6Ex8H^AyBfz^pXYxLo4alnDs z?}r&FK@hlfP-5U4Butdd#*&_`fs#m&slvWECq6C}*v(5q0T8CDuU9A&R+T2%78+f{ z8Z<X13V5AB$FW4>2kDk zmH+k~($CrM!l56@sbYQ6j6w)@i>4QamiJD$R?>*NZUTC)6yKti5G}SKv+$ei-Zr_< zQ1H=SwudYaVN z#?;0xc8pf%y^QCe;*>&4Km`(H)$2Ng3!B3OU_ti)YUuAJkmK3Tvt(U zvc$Ab?+yRdU->K-EYTSh{MTJ|gkHn*N!$6~tL1o~OV0kIs{I(V8od|aVJ8iz7L}rs zzpIY>i%-4?3h!FW=Ql(tWZRJ|9j42W3^E4Lr99cSHv9o-(zP{f|JJ84FDS!5S>&=+ zD!f166ewS2vDS!PgDiMn6=an3NiZ&bZb48!i}P^PKc%Xt0XX^N6R096n-$lR8e)^*1zz@@Er%oS48lOhL`#l={P84A?=j1DuVo`2V zZ*E0WZz>^gzGqUOmuNirE#_Ei!k_8RgcE`LfAw+2vy;E|=8R?;SJxXY$lK{$MVJ1w z$f=!L4)z*k)!Jp{{L^Ddoi0sf_7^Jm+E|Jzz^(`NC$;4Cdu~i?s<6zZ+XtNA`eU7X zC|iLD<%0HFCQ4b5@+MmJtA7zV?~mE_)F&yJPJOya1}~M%sZ`MS^{K_8e> zE6yfJu>rkMJcL=3zLH~!Bg`QLl^)eBcLq#2v#GW(E|hXc-B-Yxhob;$VQXa~RuJJl zZHUi~d$DhaUwlZvtg3i*bi1Q~x1vAywfKfSX1fQvKwi%3jRk6bsPoat!Q#^CSZ^h;8tKtV9iY~K5F4pgMsN&;~ zAU67_AY7PXv7l+x^zoIH3quWsx%?=`@3`*IfMnG57@-1>fjGxtAyr6uW?q0 z1ED`(pft4eidW*@vf1*OY@5&X$Va6TEL1V~3>ecjx-_Yy*2GqJV>>;iJ;vzETBEhN zWKkZ5zgT3yTh*)m;$W_w<}-+ProJqBc_!K&)D#CIEM8O2^hCj~U$`f^hi*4(HWKr0 z1=IvTAw86Taeh()quJ|(r#Dj=R!*q{)`M2|s4v&>RzBzUp9BW^C4bn7b`Q2=c#||%pnj3~ODbf(Vt$G({QPO6C zruslev@bF`;pN9up$Wp4_}Se&ceJ}{yg@3M zp0dA(NrqhHnwsimh!=G@7S|2DU4OZ#p{)NnoBe1&Ew+StT{m?U(!}|u#Y)*!;f4AS z(ZG)bmEjAiruoK(ug(_RPwEuS#)!4fQsI(yGnpLWZ#>+jO?2qFU+uMC=2|9a?s1I6 z1!P~#)(!C~Um!#;d~hst_p?jc0_(QtP>C|VTy1kSPdzL=e*5u$tXB7b9r1wgaKuZ~ z-?Y7=rDMhOK+)?_qB)dIPSiAV0nI>P4Hez@svvuLoQ*B@>ps_c>G(cNy)2P-R>~q; zDnCCwevnvoGc$et&Sqvj!=nePYl>?HquRt;;RDImTL>v|qrtybArBAi;%E2`?f<|1|{i_m|74PucP?Oapc+sGz(bzj4Ot&*|M zQ(42~)4u+G4qHKffnXiRIOztd!E)m*S+6g8A_fUo>eb6K?xN#lGL;m@az?8-VQo9;_9sOtXM3{F=SnNsCxqU-FVW1y-ham;obEt4aLm!Z%z%uRJ-AaVDTz`1 z8$Fru0^^z1QsN<6G)%UBd-iC(uC>PSd6usOb_v^;{lBATs5*DGW;P#|HY(YqA{_x|={IW&} zw&YewkO!eE08fmzdF21z64N{LB-VEbaT_o%w#nRx{<7Km{y-7YV5=BEb;ys3VOzl` zE$=Wyjp!3W6Ha;z9~>2`@;XC}7bB={l}LG{i3CT;U&RNLs$<9F=2NFneqEf3q^xdZ~7tl(%WW1X+>sxbKJ-MMYoh9Zu;m{zO8YlXN zh<2{0Y65Z(IX%teY(RRK#;*Q^V_R7mzC|fp=05A%VlqmuH)+=WBbClkJq6;*L)#zk zs-99nC)dEDGFh=nhK(+`^`_${FO=>_234JDa+vN(=dBW?XR`*IHfMZP=LdWFsz);8 z_vzn2N(PE0D()u3?nch$zjL^8{y??+v>jsi+|tH%EML4-utDrarb9PZXXFH36)A0{ z@(trQW`nh1&9@9q17_7GqGvm1rO#@S1WnYmRJcLZMNaoR()P{Dpph7}xs8a{ z9=kfyDSF7N@aXIwz13ATy)A}!K$00X6~}OC_=FdqblbPljKyfjn!1W zvvEhVDcwHH!|MPk-h-9w;f6zaU0b_aey~ZL=!@4eGJ8-WZP3=SYg-MElHjz0S}ClV z&77;VFL*sKOC@Umy?>xG=@4mdSfnZ5_V>jLUL4^`J=pR+kNu4f>VvDhiNEpn?V_Z1 z42XLFb$E=17&aY#f7+iMS_$ubO}nd(Lr#Au49IHL6&^Z$iaa$Wdh@S+=SL*92E#^1 z{9z=uq%B<;^H{~0>>@$VV|@K}|7n(}O^=YJIV(&XuX|5l7^QK6F{Wn==S-X%`{Z|G z{grPf-y3~&3vsF3A^bf$@pZ+h-Mf?3#~5{F$(%RG4xxo{7w#BzbIa?pupJ10STYAS13Ojy& zXXYM1?2B%9WJ0@gF5a*dgYW}}tlu5H!L#d6m(xdE8NEm|90jum=Q~~R@4d;WAA@FY zFg9YSGd?iRISb90+1+y<^_I|>OV8<)99@6qS`Qtl2v#7Z1kNE}m< zk50a0G5vJbQRCGFuctKa?O-g3@?p0i7Gym(HOYH2h^`stuN`ZVYkN+Xx%U3>t()cA z%=t{ASWeCGUSxH`lMjt4O=bV?ndK1I3^%@Q7ZufXl?@{mC_U|jWRCgwmLXkb--+Eb z&(UFb2lDk7deaMyIICvjoem*zgGMh@*q;dgU-(C;kA|Gny6x~zh-=^**FRm9{+;w8 zlaB0y+4)ajiy52eNLzQ3r0Iy#dmG0X$tv*;JV zZyvn(?h~Z;8xZ`?JV$1dOZ5Asdy#6rUX0NvdA=d{W~a>xf=f)jtOsLqIWO&*~%*?i=jJD!NY|A zBhw^l_-yW{q!HDIoYO+$;^7L&r+?^_>dg#u3bjq0W#3YcG*QsN?%ZRl)O;-8% zyvp?s4pCtBkdhPEi?2 zLrFj8oE{-Fub!2{T767bG4evqRX}&FtmdHa!!Ai+VP#xK@q)^HdU_E+vUevC2__@Y*xIp;GACVD;gxNa@8lA+--0+{HVe? zErN?~(QXVth3zgkTpH$3Ke&k>fH28uRv~^KcHc1Dpn~LibVY=%t_1E)AS$O8x)Xy= z(NA;&CaLW+jhQg2AebAp=0i#AaZZy&<<-n?!`e@{;#ezCu6f7?kWHj)R=oqiOY3Lb zEUkK65Aq5V8kp;+?EUg-MeR3Up0a$AgaH%63Zv@ETj7K*G8&Bws4y8U=(jMaRpT@v zH$^i)bPc{vI==q;52PqWlba2lhUO*EntS~9dQG$SEEg|M1@9F~msDYIoBntb<@6Bd zQps$*5;)Hqm@7=|nKx);&^M&%bzUW&S?_Q0>B2H*Fw>ZNQGV;q{!RH$Cg~SCXQ_6V z*pmzUc}_~VGBc5hZ^^Jl)zXA#nyR}YhLZTVeD6QMNw#n1G^YOOVhQ^@mi||;Nj+q` z%~_ai>ACMK^R0=GU;I(kQ*yP&0xnCUWViPHZ`^YD>FUc=XyfZ6!r|tthqtJA|+5&7f2hoX?6IS7Y$>A4f{{Ls6Fb8xXuZ_{+OL$E)X4$FGn@CA0${$1w!jlhrXc zi(kcFr*c~+d|eBe5wQL?{U0_p)K|nP6;S|SnqmDU#k$ATRM!C+_SU%Xl6+kv^LH0*?vCn6j#UAxNRDNp%OPN zRnK5px7rV?L#Dj3es-qK3&*NPi8ZOM-C4)x2{%`C(3h3*8z>03PkFBpr#3R?CdJN& zje_EpAN3^UK}g@Mj<0R1%z%!gZBao3_?x&wi#vz(tLn|!cz2~*C*(Bp@2+Q*R82|r z-Sa?9#dq9z%~gDnTgBY{RHHZN&Ux<9(xT0K4voN~5=&!~TuaXd6+e|?VzhYEa5oRM zT&Z!Sl?f*FD06!IVflb6mSveAcK!7S&2`rs54?Pv^t}l4?{oK$ z)h=BlJ0I+2%y(^G=P}v8|E-r-D-&}6RYw6+pAq=SR$8NRdDb`pPFh0lpm11bHTY>z zvHoO9;8p9ulYcTs1;KL{)FW}o=d_gZL(^EOJTj&BLABF zZz9kC&qV&84>^V1lNcT+V$D4xs@f^LUb@ErKAe5PxZ+yme<-s5iF5&s@H_<^dYzOOiH@?O`f{vVJiLlFoiL5OTc_ zPp*}{a&NY{7kjfgK2$eh#5QX4**D9hGojbQ_vE;|d)`)%LE7_FW5*-D zivB&@;X+@zOSbdQBhdMA1Em>%q+5p)o|Ku!m(I7XT!kV1VyoYBWj1kEp8Qr0eBj^g zRr|i0@$s*W5*nV>d+{5|)%?9Io@bhkW?rZwr#CytRr>3O|1;Uv9$=U&kp%~veBY~} ztoCzfny4mK&OYyh%GnEBqc>ka+jAY>W&aI z=Fl{H=!`hm4*zKovW#Cl7Phshf-+@=j#e<~1Y!`)OD}P^pER$Xh_enM>i34U@BYp^ z2y_a3aSod+bc4-%BCXGFPQ{P#@lhBDZvMTp0kw3GAnPCtmCzO)?rqdlej2*05OJHi zXHVZ%Dd!WrKcyDbqTEiHdeS_=?iyOTne$zZtlB^0;Awn2Pfc>|Bs~Q7Ef7HJaGqTZgj3-oMzr$yvde zteQ^CVz@WRSG7#hnAb$C_~Fl|=pI)$P~YcR;vD39zVHg_nioT!d)Bi@JH6^^zvUlD zEBV8}C9bQ+K`$y0w(Lv8`1Om{w5<(BK0t)y$;);vBcgyEcPpjfmUh=qtjy8){IM0y z?P6%Au@%OqbG!pf?;!;86cux3e}r!~YP9cosB;4|@})PdR1wzNURZ)sSQOQJWJZ$Pw2U zfXZ>7)VpHC+v!FxJ2&!r@SkXArO7BY;$init3xRyH5-*cWh+ua?dC7~>N_SIhAfu5 zNL0sPxgmnDRfguZQ2t4DwpPtMvBYkD{?5rt01q%~5-FKg-f75&n`pgIrmBT%=nfX0 zg6f!2d+o#vYp0o&ZT?#w7=JaD==pNit|}ssaz{u`(;>2fx%7=;hYR)4gSi6tp$k44 zWzHd_b>!_}2lGAe;{2p=H)RGBNwX3%6l_@n!Z*=i_0|xZ5x+oq&OBLFUPX`L^Js5s zX1&Rp{EvTqqbg!}Se@j_mtnC2<1BIuF0+rNySoQco>V&r_WQd3zMaN||9!GoZ|Xn6 z5J%unMd@{7x4pok#tB`z%}MP}Ha}BWdNh}_{Acx}NGD@f&9elga_3m3n7F89%g;8$ zO)*2!FIprYU8V*bPnZ^K4yHG2Og|UO)8b}X95-jIrujVciK|^lmE_&9rhWd*%*U!m zGf^ndi(DV(dq@6QZ2!&BH>t|0NLo&0J-lh>k8qlD(3I44#IJl;4&!{9Htv61^Cqb; ztz_-vT3DmE z875K>PzX1!F-q>!gCsI7Y?@n-L->HJZXR(gu^N|?oPDM8w~ZSiWtUESRj9E0T^QwW zQ!Iq5#?Dtz%5TT~597Z)OWf8U^-54C7;#Ycjj}h9m^9wDhFNRH6N0b`eFw@3cw?(}Pms*_w@SXs3U73dTv^ymW&UW7*1uinf@!%jugR z5LKT$IX1cE)@=ktJcMqNXkj5tqIDZ)G7nH3wySN19tl2!2L)i6gk+x>vS8h-3#kpa zATKp_u@%Q`h{!ThB=%#E?V?u;Vqmi(Vqqvek&u?PQf=Yjw@@U+4z^JXpNH&Aw*$>y6|IO-rm%^38|K(%kh|PJTJZ(dP5Fm?;oK(v`jf@w$$U zgD+a=se~6y5mt0!}ZW~3OpnF7nWVa-k=PsNTRBa9)Y$@88Q7m}z zn0wY>!9>$P+sDeqYh3#Z6MUh(V|LuV?N$~)Tx_zt0w!7-+c(*>s)dcz@he={zzg%(eZDUbz2$ zjD%axL}xbJH*SyEhF5R9?N1is-5nf6=iNhjP(6-32V!|C*%ED~)@QC#>~93belu0# zhT~h-#l(8=cF2nMeZ>o1Ru)OP+riJ^E!#Hxc+J+HAV3tE#KzsjA(r6}iV+D;W|*nmbMUr}__jK(t;R zmK*BUf~~ah$fwN=?y`Q*w&z)U50#PdVEm{mcpA;zle|4Ydre(_&cSm zV%U*q!xHX}x(aJNevi9IDF%A*czIZr#uilNZoYE}#V@3ws3TkJ^~yiJSQjLy^iadJ zsGBX+9_UxmUc@QZij%2pyEKAz&FdRZ;!3k~h*X@n&YPliQ&u^wh;mbQ3!I9-oGfj) zJ1j+vnko&1q+9|oT*1AagHCEKWwyMl;u;0Hrz+#H6gS$AyAxUB*juDSHRS^nddDd) zthGDRyc4`_lHP)fnTvxl`kN_VZMz)P*wAQ>f~XOK5s%a2k9GB=hc@_gU=&V@a$c({ zdsQE@IL;RBZw!6Y6_T^+&A?24!!vG7Q7z?Kl+&P81!3t3wJOLzrexP4aw$kCHz2xA z9X$I-NEwE2vxj#G_*e$wx~?uo$8N*S+17l2XjR9v;kyCWE2CVkYqY^} z{Ev>dpIZ83Oee?i$1h-#GnK!3cVt3Lmyfzu72T823*SVF@T zfX}vQz($O=#r-DfyRCp56?fe+^ZFoNO-a8n@#ZddSvewIo#x_3?D<|XD>dG~<@~Md zx`jv$4Ovo6$CiL5TzJG18m0SBRzyvd(_^sQ*-zGCs-q=csRs>q@pdSU%E#TsbZA4J zxx@2hYVId*dibR}f&;1FWdrwW7jC_eqTAJ&6enG)639PI%eUZsOclttK|{_#0mUSQ zuDRH6M-tlB_i!D0^Wr^m4A9ykNec#F_s>F+g;l;(aD4JaZkV>d!csNB)*ofaeLM*G zJ6}!cZv&CC4OTduLlLaiOx?sRTBHuhaN*AQR;kI0G%?tK)gfB>p%LXrK<*<0(opqq z>HLTD?xrEy22dZuF$&N6V4FkwX1sy|GP8^^#^Q7ay~B2O0K5zmDCWd32hE;JE@nz@ z?0L99xC0C~0GyK`vjNlXnu)9eM=>OT4wCT7%B(Du#B}_`0k8Q)xy!n5qLV1QTa3d) z2mh*-fcEjiqk8y+@M`V%!8Sa37goD%BQNDaxMEDgbn%X8P<%F&uZ8hm%=LOwAYaiI*Tj>G5b)7Og6BnWK!j)IC=i=@CkSB~@cj|YMr zoRKz*Om%}R{BG}bKXjbn;n`Xn5j3_hApMY4GH^&Cd}P=}Bo`VQA%FZ4^Oe}a&z9;O zydfL&WyHu*ak}4!20(Xe&S;Me920?c8V4U=pD;i2ae)1K_hZ|+7bPQNtNWYY9r&!6 zIx!0$3V3IY#^MAmm3k*$m8vquHYO0`;w9l#uCN!D&KR1{S9ZsFLC1g94Qh{5f*(q`$Sb(UPI{2YCV3T@|$*&gBS#ueUq{FO?m8eOUM_aAAoN;L7n>_-i6gZrBS z^qVb`&k-1i3Ryc`%*1nw#C(f(26jM+DZbTMw10G~z@Z&2`H_mozv~KoPpn97&z2RP zH$PSD>}AtCp9&TJle)L{Fd?7&0Dn(J^&{BV(RC(1~tav$FAbnz6&1#-`<{Y=t6? zY&;skZm`%^ajmqg0!R2I2;npWm2+wQUl2l7aZXrDLk&8Rmm8(}jA z?=wh?U6_SI_CIR&?tgWn*IqGJ>)mHC3M=vWi${;f6Bi|FW0}LLuNI%Rgiwy^=cqE# z8A(Pot!e$x4T>J{T|@IxAXCO4Rcwu@8|24MrYhHd>R47ZcG!1?s(&NfKbe3Z2zAEO zW?Uixp~%8BA95X?8U!WZ)u=K>0{y6}mG+Cpq%!PY+OTa|F+bpg&GkFD?v*SGD-*|a ztDpa>on_3aHCzPUHd55LzCLiuOsj8w%a)m*tzNK4KJ1}5=e;QMHBd`rF+Akzs)-l4 zdGBmP9L$!IgDT9iU3@l8J2vlc_9%tZ-v7mLiqk^K(__?N(PHW*&r3Y=Vcm=0y>oVL zi$52?ma)-du-LW&a+>vU4QV4!wVJi8-owkjI~KR#O%?bV_Jor3r$&8i!Jzi4j-s}W zws$^?pjsE(@Fxv3Rt@&`=Ew3y$(yGE8^qlEKjl`GzWS3)x;^S!jC~&_Z3t?$Ea|dh9cMtZtL>nH;53UJ}rf#yIBqIw9o^o0U zb}MG@o32NS%K?kI#V>ye3y4r8y*c3O8h+wP666nSMp6k7?2Qoj==zQ1n9?P(S(9sr zB;$#o`zxXcIL6p!%GvF}c@=DIX#!qt=YCd7;LoeGsxS6yYlv$=quQrwt#_)2TySpT z;dO&B5Pp3)bNs(M+C`fK>RUCW6nsVQeQr4JZTDtMxQAQ;1J0AcX*}Yp7Pmv{TcrsG z=9^>hrpDDOZ{82V)kJ%|!c_x#I+&SE_yfajOgD5+t?zEAr4sMPET4(Y2pnj_c4iTx z1f)f*kn6%u?sIjye|d46vI29(Tw#&z6sJnlXq5-9O0Lu_!kEHygUpjT@NE4IntuQX z5Oo?Swi47K6I5z2#mSWX;_ZVXSux-eoQJN2a=2hw(yJ&@BnxTyr zrvXf+!dN%>NxQ@Z%_4MS7!cfF9C}cU+6WL5$yI)?-tYVLSB~Qn5-%0QAJqKtyW+?eQ!-BkT_4&{r$RsEy)|`6Au0G}0{8Od zGw=eqJ@xeTVhU#E0UTXyiw^x>xd!$I)YKxaXbuDx0Jf)hfCKMd$QJG@&ZVOkE}MUK z#|77kko_p%1Fmd^9I(CAt*8^MyiAr_^+F{A%sz4X!haWBJWUGQynQGNLolPc zSPWH%`c`@sr(%)(ygl-6w;{?o`po9LqNVj2?r#%9%=Z0nrc*YPW`S^l%F>qdJiL%x zOZjE53_OLNZI^DHD@C5* z(Jg%4=~k5-jG>Z~2N?cj!ulFmcv`h^Go}L>BzU(9gx)^CPWOK!Q7HB`YWXuK zOM^MNn-V%8`(i0d3DAT8dV0_xpA85C00JB!%b5cyLI6GYN?Abu??vUdr*KM0pE?V+ z6k~pA0bT$7C7tNm2+la6MUWQ*v-B}Zdg-*0L!1c@Le4ZUf5(k)B22$` zOxc1)l<&Er{x02fqYdG5X&c*I#Jj>s!O!tG&M`Fb=B^uL^c&lp0oEI=H#0|~1I1_D zTQigmvE;Y6d>&JrJoa!gX4~g=0^((^sj87COKV>|+i5pNn|==v(u(Liq=Y$PO8}Xf zz=E}>=^!CLoA_|Z#c$2`E@q4fvb9Ja^SM|;?N(TDlQ{_2NbU(Cd2mYhy@#;##CYSUiwq zuEWyYaFoyEo4O%XC`kI=?LX)Lqu5L7@@Sb_TmG98h)ojU<6p{p_VG8>b*2=e5+&c4 zgE=Juf9}l|;K`otDpAtNX+Pa6HW_ac=plcOD&)PC4(=?H4g4s5|C8$O*_r?J>m_~w z+Nges2+)8BA4tdY;71p_P7b)s#vTeV&?BL}!Rwd87Io^^9`XC6Z62j4iZ_S=-tGIhda zZCrVtuBu)1qZu=h(5%UM+~TsH7so zk(&*uS+RHkHs`&k>I`s;?~Aq0-ati8EXAz%Vt z{d;`S;j>t@fLJxx0hik&7E`tE`(kxd>$(K|$?989W;N&FJ3Gd|<({&U9*DK#X%Hd& zh&>O-6xBBKU90y?-w(Hxdz1wX(9EHMGSkni$Zaw?&jpkApXxBF&t^UZWTy-24?Nlt+foWU}) zNq_<&VOY&VyL8;{os@PzkE2713>U4qhsT0jjJgcETbp~S`j%DeD#f9Mrk2+0##8A@ zk$(k3sq^-OCCaXA!tw7*RF!AV54l9EgBPtNEY*}zN~fnBsH*nnFQMl^x*+Yres8^o z9woEu{)aJcAGXGo*sM|edA!DyB#Ka#p8-!Qe`bXzSQS~Z!PMY_yvPf?d5M^6m{5X0 zw|>yNyk1u5iJC%f64c~RqR>PWX6xUUyuu}%G-j94`$Xl;3?S{UBxSf zN(8FLkiuCl@=;8;aM6Ya%Uv6v2+P$5;>D@>-v!z@W-C1U%Gr4R3F zE^U?ZIfP4;tkpw%fKH>(cwXE!E26+5n@YRs&83k$Z95|}AcPQyqh zjxVVU>yj@y38^%0aCr`Spi=(KuEI$AqF{DwUqXglc0u z#EW-Nu1QJMs0jrkz4WcA4N5c&n8^e78+0K3lPimgl*-pcMIlF4F{Yk)S89iB5<0ay zRZg|49&smm#nsflWRr+^B+`%4}1>M}CsDxE{XCAeYO!i(lu~>*KF9Mh0$h zvGNY)V~z1brx24HS9@5};CJt=#F*7FqC7{Xe`If(Z0pDVia$2job7yM6KyD9^Ky)! zf1{uO3DV!Xtz&A(``btX%BIGMm?UC)#%cB=`i2WtifV)eP=Z{&6xRXHf0$dNl0(>} zlBhV|-N8Bt$+-2G+DJU{MOXspKsNANjqy%PH)wuwJr<3?bUkQBqFjN@IfoI@rEYP2 z(V%pUu+;TX%}o9&FS3DtZjd1c!M~G&5vjyikp9eEtisVq^}XqCas{R0ct0th${8eF zCEZLmwE|Rls}34y#6Oh+xuU|#Ag~X0;wDvt2${E9kf>T&CWRWXyM$9_#65Su1JHK) z3))udUnOOvHpndy@Oeht80C{zu{Z>+hTtsJYMsf=(PLk1_NXYJ+oPpb%?aA!sg&jb zs2#6)F{!;0G&4&^d3tVAt+%|@z^@)6?n{Vqm~-L^6^gTHHq51!kYA@iss&Vfmg3`M z8JE>vbkt&P;Q$pH*Zv9Oc*F{=D^u9`0DLkd9b<`qXI6=ugL(-lw!EM9eeg|cfn7c* z-nwI7Fjzsa9j984r;$UL;HNwBX!iu4O~?M$Onw;xTQ|{)do{x~|58`}!pIe6o@)|D zdDW`DOebVfnK5`2#dE1R9`@gcyXxIZaon|T?e!2cTK7cLQDMAaK6aeRz>N|RRkddH zIsI_DE;Klr1H_8o!>Kdk*6AnCg%9vME4lF}W0co?N~xFx;C9FUj!zV+<$)KLT<2`t z71xF|H&~{ldaE#A*mZ4cmNlByY?U?l!GYR9uApSfW@a7fb-51a?lnev!xR9wQ_I>m zVh$kvt4Kz}Bk^7%&4ug=9=lxCR=rL2$Z_9)^agA)U0JF2rA(gYU6i(G&-{rEYekLI zb)_X&&ED5Ncm~>?;?^6R$H?}(^Yb~2gF(vUuFWpojc_WAuHYup8$$V`ZFJ)?E;!e` zF_Vkr>k$W~{>%%}skxD2U0mE`Gvo>#Ox@&qum6j-jYorkk!FVUfNO_DKdU85enZ0S zuf~YW&%|@WRLMR=Wxa}T*I~9V)EuO&8os#&Ky5%;?M}01$R^NNKSaAR%73O=miknM zPMV2 zE%)ghX4YG(cor~Vv+so48<(il$BU!qzbOUEL5gEQi@vl_b0F`iFuucG3qZc9tc=Zx z80E)B6O*+%m_=cgMf|@fjsMqz@Z=^lfR!_F+_OLxI)DaZh$lk9@D9{zY~k85CItX^ zqBdAbnlv*3m^b3pj+)$Bmy-T4VNlm#o@#1d>?QTOvqJe#HGtAJ8Q>~Lp|f;D9Z)@` ze&V*b6FGUlC4gylSI%-l`gO=9_5WDmr6UW1TSovc|GZ>H?bA_Y%TFlfzcmG_oWA6N z0gMZ;<4gcK$ z|BA;b|1~dF$@{{1i$NY+CTgKFBVtvFN%?=THuKmZ07TQlRNg44`VJ)I;k`zO%2%TO z_g`tH13Lge2Pv0ULW>k$IQTEe?wQEaOGX&*6r9YS?-209fSal!wad&i``!ncvldh> z04v>8!j;%BPdgIwLVS#d#6ah@^@vE~$`fRx9!n#Q%OJ-TphNn-L3hXBWFpKhBL3-2 zF2am%CQ#}TtfT;=Xfo}D=(W^nqvZi=Y zVM<`oRtBI4C0yoHz6YA&0X>LWlI6Plc3DMNJiSnqf-0pK z)ipAMxgkJy-mv~)^9@uT`mEyMLmJ2dd5qTBc5wFq)EvOBh4HX+fIZkswLIJ3ulX!L zfNAty^ITcu-DNp=QJ@b}KIBz3y%5L2PoK^Bufc%)DzTSl%<@eN05nW(&=LwcP>j)X zEl2N$id2@ryc?PM4>dnq%ex6S;KQ82ZklK_w(>+^O3A}_kbYdzql~R}n@ECvx~Tugd(B9v3G~z6l*xCXXC%!W zB!FpFdEo!c)=Ojo7sq$Z#;3v*Rijg_Wxt;4jP`0!pHCrDSWueqyd&4#M?!9 zdgVHo>qXZ#{|58=gJyb*%pv_Au(<|5)UFDE9nRd=XGQ4%vlQ1PMV%VDmevn$y3F91 zN-BX?W(%(I=RB?dGVeqiO@jOZD4PoHS9}JoNG~{)&7xqI$QP|iI_3;Or&ylF|4Wby z^Q4XChxlglDMRtnZGbC=UlR4w^4zL#AKESDZ5jYXt@-B#;tH<~Dgqe!u@;J#4*+=J z9HM7LotEvkeDScpl)0|)45y(lMKi5&!NhMN$(fH}XmTD<_`d6jS zkO{O_NBK;G5PFaKoMBf2-k<^C*apACOFkak1~ukBg=ZKQ3Dp%EfBFkfm%pL=@b_BH zI_Q@DYpo*iDIhzLg*MwIGhCwpCiX?+7mX1F#Dk;}y})b$eAim>->_5HN&!VJ<(@%y zT88!3bI#q9JXi_(4D}J7VfMXlxk9#0L=X3l6v-!VYJ(HjE&VhgimW-4FMmsr>xEBdO`ssX+2ipCicKDRn^UQzSpNmt9M$#QMoLp88UgcOw=HZW}Zle2rd01bdW zLV*oy<7tg8EA>(}qmuJ^^a)zwC`i5Uqobi@S#dY4mGst7VK$BVK2Er00Mn{OTEZ%%-{P^ARvbJxTO5yn zk*yk;nEu<7fr;8+=>av$w@ouwSj?kOFAQqkL6bA(LRS6a0$O?Y#{}YfbROkcO8`cp zRL$g}2?lcC0n>a5h`FUEE6)u$`^WQeHa*JA49e>#M=^T_@Sd2&*hmkmI&KGEMeB}b zu?_arw$4^Gce*i-n&;=Md?Vws3dblHNQdKf$6-DeX~lT_`fjT~CIM17w*8tWtur zvQkH#0m8Fz@hkCJNPp|My|;sF*CthMRN}xstW1fUREZqfKli!^ zHD=`)Rw{waC+l3xO8LokNbXde0oS~j{Hn&j@rdgQEbS$a0$QqJ{8wfb2CM6Pm4ay7 zR9atlvnmcYDU-EiL(%<7+moyXvDk(FAKIYFJXBr*5OBMftbQXfJFcT9+B)o^CYtFN zfz@PYgdRX+?XvjzLM{9J0DbgRg8F3ONsSG|n zcx|ceKcOBt4$jEdnyw0V4se=n=PiXibaSKrC7kK6C28ywwKm#Mt~)h7k4yUUuZZby zkfP2A=@#DQu^!Y0NGm=5J*T(wJWiM1%2l_~Fm}DNMQfigvF{rHS7OgDRGw%xsLlCw zvZOVr9cO(C=}fXMUKv2=Ji}f1z&QH1M;>>Ob9&ywJk>8`#~><4`R}Y979iigL2%^4 zs+e8W)3((F`)wGq>YQ@NMRd&mT`K%SA*z(sU`Aw{qH;DMozV9Jwuk}%F_~rpR|1_` zPx;%{PTm+&i0|By(_(q59hDSkjh1k*hjw^U44QP9>fTGsS0#Rsn*j^Cta7W~_s{)S zdf=_~E9KXsa=4u{;57xtCuhqW8P7EOa34bGcEi(-r6Ab+7Ya7}14T*%sjA9Cq@w9R zqnxMwrB&bQp0ow#xOvK@%=7z=9;)YFk6b_Rk}(Gt*Z|qb$NLS20zqQ0J*-TjzyvN~ zwn6Xd%w(8*lP%QG`gppYW4W+tXySP@Y3}j13M9qxAfp~n?9R-N!hUzme!4P zk@EOfzw&tpp8QzqJThDOtgB8gg|c@s+wk%_>~MKZ5Hc}4<~IxXkKENe##PUKom;54 z<1!}X9g3W3bz2?@0yeLs#;xU|N%7EUU40M5lVi)N$Xp+9T@CUsTdmJj2uAFqMi4p9 zW%}PI*%9ipZu%W!E%k?sf1Mf5P0uRcBxwHl{nW0mB&(qjPB_)x`X&nJ9aOKC@K7tx>YGp}2^hE1@76tIcN(D%e90@diX)O25HT`lQ+1UqY!k@sq1aT z0^!%Mdjn1ak+`go)=VQyLwn3$C%Z1cI>G+YRmcW2$Zz%27F5jpCyQgr+60(+Jhirf zuYt45?N8m%m#*s6;4T}N6PddH{0rH#g*)Lfw2N39neVjS5#Sl3qGseD*Ih8-50$xT z^QqaJdYq)J=3K1J856iwl~rWOT(x`D3SXNYC2#Nb@#QF?Aq1C8b0)3LExB45q7QOa zx?vvGr}ue9ZTVd6+}w~qjP;3fKK9J5trd!K_;MJWV0r)8iltxknP0g0==}vdDHt+3|b?lTDavq>Td^$mRBU1kw%9Za^WvM!++;#@<-qKxWft zOf}hF?HSP^a<^278rGe=#=GEA?0Y98!+$-q9;vz-C29X~dom2qoSBzZ|K;g-gmZ^W zQ_ruarTbk0Ykm*Fh6Rr1Et}@RYVzniPWuR_C4JO=^jEMxi`)4SbWHS(zL6{5k%^v~ zL-vqNc&i_GxGN>NJ5kn=e5_Z8%)}*yW|Gp%;>L?H*I+kDdAl(L2$>qq5 z(|Pr$1%h54ei$38|7xQO;HrO~qODruZ+dNPVX7rSPKLi>R>(B-0hlIE?SLUXa2iL%SZ!4v0R^8xa;Lsaf1v|1O3E93bFY2MWtX zT|t?;{H$z*JSq|ry@8a%1{DiKuKXk!wa#XqUQ`j`XM>Is1nO@g>gen1I`~8sW?(&) z9MvB51$8t)Xj|7wRxuNfgcLSN&?M7MZv_Km3rmBuEOY5I`ZmXAsK19hBIof%ZvVY^ zQG-ZzMXgeRTEbM2FN1H|(cdK+R8Z>&+Wb|yy?SM^3Xja3tPoYFYB&_s1uCH zP-*WLX|DsA-j=7#zt4vF&41*j5YL*r&?vFp4#rnMyl|qss%SA!FQG-M^0BUx`w%(q zn;ZE(VY&70uYnz}xXLxT6BGF98m8}zN5q~TJeCiH5pbX$>Co9c16kH{Q4Y4x6OzzP zxd|Ob8Mz5_!bKK(3-Oo+E==`C#_YCuEd}!n|hZb zi%foVRR0l+uTWLQmN7x@ZjvZSxLQzby+YngN`I~u;VF|%Lbbx3ZT`R<9mIYEevIu` z(rvHp+Z=|(oMdhOybqvM&q8gy`mv-2w7VO8-$kWfgxWZ^96Xi*-(-&M!|{mIKKFESCmFIgDb=;Ahkk{*knqWFmad8?)Hyr zP23!dbBek_jm#~rs__FrV*-oQQvRPX@V6JejfIjv4|bew!KnKA$m_E+V~>R%D)r_Y^0nTwyikyG)L; zZs~&>ld^rmdrp3GpF=57#o^^M?Hd+$BWKYmM{aW#Tztn>sSR%n6riU>Kyh0 z<&&t%RQSK`HQpi*Are=$sdzYW9hz>`l+0LD09VUXyO|eqW^!X3juhV|st4Miv-_)v zty#}qbyiVF2C7XWYfM=%mX0?#%g~?3j$t;My4#?H+6dXoeZfg!F;!|O%r5$U1)qpB zf}Vb$Cg^%%5Bg-z9;;8LaMaMyJ>h-Xt6ZKJ!zeuyJb^8@8jGLLB7%fXPj>Hv(mRGc zmjHt4obReD8?lADZB+T-cv^GqI&S1G&B!5MU1t8KLf!Ujn3ZZOk6jPB-bT5chA>dd z(+*?neM-ah_MuSu{mr2aPsnR&&Qz{Kd}`{7)YK3)$www{Y;y7{OG-~pMZsMr(d@^% zGGr-gu}#?4UB;$Ec~>E)HAUQap@x@{;A;FT>ytHR_L>5@Q~zuNh!hi_h>XYIvS!R+ zUy{r-S-8QynOLOit?|!j^Kh*8@T-ZQ8(Vjxx#|qf6h4bzYjX1+1LW@2k~{d6hPQr> zk7fCD_$tRa1=nNP;q$|bzxkNGhy5ALQckC1P|-z&#WE5V6Ypn{Q=dc5Xw=DM;}Sax zeUM@-MEIdfIA^iR8(F$EZ;2I?d^P8FkF0!>3$}Ptry5-U^Vf?{YoH zVWk(Qel;tMC!5X1)i!av-)5ceMD369s?+r55qZt4)Ar`UqW6PxpX}3g)Q;}AQ{aSDJwcDvR4By0p-w>+W&!Mq}Pf2$iB z2F@1!PqixG39RH(f^{lWc3t;2usH9}m|SR;4VV%#Q24F=Ho`ZYq4I{Y1wwtvD6lzlNyy1b{T$ZAfiqujQLfNBh%s2wD1TM4Q1zV$6Vv;?Jb8{7Se z_{t(G2Cwp%gJh#Fzhh?CJu#8s{Z7S=zSD*lm@>#oI^7HE0-* zmdq~xgUYsNDbOY?;OofOk3IeH2R*37JvBH^dapXnMugYN5=66Z~SFfBDVxRy}g~+&32sEtmUS%8aStWv7u%q=q02)4+T=dV)cKZ!V zWZw``3UF(V(M7!|%}{-T)F{2-eK0c*tFwt`FmUXeH17oqLLwTz((zifjjUgY7a;p7oGtqR_8qvWv=%1h5Fz!#Ncsd>2k8w*T zY-q^sz=iPXM3~fPlQDjJr(%7+@m#K?;NrDeodYv3#IUC zlgy^*SO%+A7)NagpYZiW49A8#DM9)bKfw98VZ(rBx|0Y>%EB;Z} zxOc83Lkc=0W&J+q{g>d;lUSrE#CIKJ-^;~7V#;q-oietaVyeG=O#a1)Rq33OD|d*7 znv~ndk6ej`n25%kQ)Lrc#QaV->SH|#+4a)<6#RQvKW_gPcsc+L!H%Go83EyXFh*R{c;g2{94MqViJ6*}dN<76yC; zwb}-&$r)frx0EA*x_p420>Y{&`W#BtI7!eOln7DuV4+EQ%8c}fU#k^@HDOx6EIhlt z0*=CBh(aTNzzlv4?0Y*V{;B*sEN;?pv;1ShIu9aulhd6``m}VX5@e#IQ{F9VZKX?x zAZxt~<*$>>`!`E4Ux{JQImU@g=%ZYk+~t|{nZkud-*g*(NeYFL9r%Xq3`3)>_38rV znB`}tBna-9@{RI1*0*_Wz!=JQe!S{p2p7bD)ycr{^A;xlLvx(H^rUR!+iM*Ea+oa` zkLVS==G-?P0V^;+Dc(H7TUnE<4~2y)xT>tv4rbzh<_gR(Q)!na66CZa5yWC<`DvlE zP4I2Sik#Nq7z5T|1od6@SB4}DZL6m?+Ef2FPCII!83ep<64=STTI5n4p?`4Rl^K!$ zRK@Pj(Rnq*6fU^+`Z~f3h6V21lVUJxltZq3xnvn`wor+5hwZFDokBk8L}rC{ox;wk zVlYqczH~O66{E2uRYPh}z*lCV^mYJgPU1T#HSkECS3H*jf=j^6rkzOv`A7!y^f~5^ zBB}9|a>op^v)(Wf(GRaX{zO5Z@+DEgy)y^A2DC&|z4A9w@iWizT;2wRx<4_x0%~ zS#MJ)VtBPT!XSS)tvmZ*Q3JbBJ(~!~spNBFvny5|+2tV&QT-JH-7;_69}FW5*y}Bv z^k&(xuMBWlvRVzhCS<~EFQ94(mZ=$*4!+3qFpSV)4NPGToQ0p}0QnyeFkso%UL)X*h(=Yv zcdjK*8x`%|T7A7(ox|r<>cqa0pQID#!GQHF$~*1b;LDL*2eevHUYa!kz5sq7$~>5u{o4O$@zR zMyK@At1+|ryV^IenDp<*O@4@9kA0*XCA&#JBAmGrP_I)gzg{;aq`n;jbPmBQ$Yw`$ zR*Y&wqIx^z2(S=_L^RY6+kM8&!ugPyteM?Yz$|nN>{0HGfvX~9tYJ=TJdPK~>J*IS zyn7vkvwg3M;R;d!v;B-~{)ferhHSW0;MYj!w2`qodNmTISCyD;lKGzrv9lZb*B?~< z zY@zooB=X^J*bK$r9mLXdrqk1fzaiJ#(~hQWGkTN9@Jn!p>9<8Qtp5m(toCR3pNL3* z9&A~hGs2s)d&+Hn-KglkX+)9uTq^-xq8xs*s_rSvnze|-`Bwar;J=Of?H*eeiP1!D)M|PYJFgh$0 zq0{C}W52w+@MzgP!BU(m?wHiqeB+D0h)#2#vnDJsYVQj^9$x}SfgC7U!&DC8liHMY z!mb13e6oaU>jXvQ=e@bCzyTp(BW-YqH-UgLIh88i8Ou4QM9U_5qP#X)dheFkfQcpR z%KhHiE1AOPME4`jRWs3Q_Fq9EHBP!Ohs8v*(=HjeNU2^`gg+5xA43t1=1lX#!+;UH z=Wk-JD%5A0_i^}=ZCP-RT`V?#4_8)Haj*tPYPJ5Eoy(!jL6*A0YV9Qmiywje-3n5`YCwH zR#fF*AHcouaI@JCtf^k=g4CNgNsgiYPW_Q66q=~z0Y>XD;c1cNkZW22TTvJq*qwfQ z-x!Xg$?2#y`Rg0YcgwVBN!JNYw^Pp(;RTu}?A>Oc95?&)SE=xJ_gbZBwdMD_dVcnI zpw9_tq0-%%S=R;>WDmiY2d-^I1ySaPIrmlcOIA? z-SJW^B-)w5-zlIC1k5HBwcFy{pZY&{sNPZ;j$ymS_zt)^H<5knzBwdJx-m#5O}f$G z>XA>n>DR_QF=!kM=rL&YUyAG|4Km-XbNgY&oW}dP;?cn>`%2Vl#wFLzahE^rBF*uD z6pNRkKl68KQzp|cMIu~qr1a#B8Nn?<(nMKzd>5R_DftZSYVci_ulV)wCVk~!Uekin zVUg^OcY77{KjrnS5c)pk3Eh8kjH(Gc!a4qQrAwh2{b;X&ASZ9Cfl&DvDYaa5i6=k`acc!sQxRFl8*8!EE=k4uwp=- zWMv8pbRp^p!OJK*oZv#=g6?1l7)m;~LK0>+r~fsTjcN@xB-mP7rLw+m>%^iV+6`HQ zvnVIK5P8*?zBAy{87$<=(m6^`LSw_)oY1|XQ85|Po@l22PaLr$2v9by28|J)82ip% z1ZK8-ti@=z9A>s>ngwe10f>F4L+;0?Lvi-}n=4Xsak!qp?oz!t=5i|1no+$XswT^+ z^pl0*sHj$#%6DH(*?oNXvzW5wVVMl54-)I&B-7!v)8V|ua_64SM2eU8Sw@|r{73)p zB>ZwPXU&@N6g~JhZ?QwUnFvo%wC$SvtZARF-}G`kKgNh_IKN*^$$&IEpIexk85RE| zv#4Ke`^y3=Dt@LKtNP~&F?dLb|DQ*B3g4I_!QK{Wb@Ea=wDi7^T$N}!4ds1mio{n_ zg~EP=EokY(mkO2TesIz9*U7+$d& zBhtNgmpc6|oNUia=SRmWwo~~X6z||7I;I$xCz$w`eyWP1JU0`ftCJWZhZ=|~<@5bd zmr8VoXREiVhET|^0(gaUfY|aacaub^vulen=0%{AFVl|M%tz%= zakJhYs4@6AT*&9+3IxYx5Ig#@ltiLdR2umXUI7H(771=$jNbnRTnn0GY!c5ceBV+s zmgbPWz%aV`qqkjOr%qxdY&dL^n2N087TsSF2vC|E@z1(Jr6jsy+D59Cduh!rKBE8& z5y$8xu*fZJgzxTt!Xz##f}qeZfSNF`fp-X7q~wuEa~cv6Gq9%>_s-Tx0vhb1_0D~; z8%J-{c}=Hhms2By#~&x9D3v+9xDh{hY{J&$X4sNo^7jxnmA0QzLj-p>scC(-2ZskQ zdouT8CIK~Rl37HgVW)wGrzem6B=glpR>J6yuVk@^CE9j}!Tht5pJxg6xE^dQ`{|YU z;|{Ib8#LFhD(Dl6)w#3CWBt<{ zsjjay*bg3%1^vbS?Vn0MD_tc&?V=_zz*wril$A|Kdm$qcv{msCgD^E+glNlT8XhPshkvXe;?Y8aWOnJ0Ce(%ao(r`*|oUt47YNC!6%0 ztt*e59y7E*|IZ`Gp`3H%XjPVVP-!r$q(;tJxh7@KXgjs!+dBPEMvg;2897>|rVgqN zV%5~h`8n65>upsWBWHz>JtSqDk+XWn|83;plxIA$&?86dmU~S59?i<85!39Evvjyg z8{t5C5TMr9zoDk>$T{i^-P8Z0k>k*~&B)OjH+R%Li#3;#Gi`)P!_UxpT+KNpwvzv& zk>gOZ^O2*q%=5PQ^H5e+jhuv$COvQK$|L7XTh;&b$Z;s=964H*mJTWnW|h>)xp|~X ziD$MQIme!<`}lt{avb`}$k8fMpZ4Aw#Hy*0lRe6$b8S@|BWJT%v_;A`Bj@TfF>>@f zOJ1${!{M>PYF~Mv#IZt8if{0%SCv>J+OH~^Ae23D5GBRe$csgclXeI>Lk3V%e7zuG z3TlTiFdMHX7YG7YcO4NX;jBzbif>$~6v0xrEtY!fWj|QV57hZ7DSpgL>YR8yOX?Kg zz7G+CyuTsH=X1aF{l#H_R7j@(4GH`w#^E8NLC3jbuP!Tm(K0?9mJWUK1)L?Jy18+0 z(motN&WoTfk=kJKalU9}MW7gOE0Np$;n>TxNHknk9JS7~Y?1vV;bI4rUs)9{_MhOW zi3}RxfK>aS0Zxpt-&ZotS8nn7rR^K?DkEqZVqI`ndjIj2z7jG9893;5@)<6oE-^hO z`sb1{(Qrq?#YEN9U1UWD4WLPnOb6Qk+h&cdtO)tS{^7wAJ`}?^**4H!Szbc4XQqqX zLH}I2!t>LRzl7AQ9ES_i`cC!*OVazFQeNrfqhOyg_zm*A1un8?(VNEVaX7y<<}*z& zHW(e0Lq?&E-Ii-OSgQIhU1NR=#=%meVKd-fCr1bJRPIqSwmGq_8v4b!WJFs=dtO;|z7q$Fp(ZN8T{K z$X7?rw8+(rJLWJw$2Hf^#N&dor}tU&2bA>7DR>K<%r&Rtlfmw%n)KLA zi#&&%YSIqoV4Z5xiZrFfS}DOH4p1xGueoynXl5yWl{w0% z-)pAdYo_>p`0xJcAsTYQok`KWl9&nNknO5$C%XfY)r3hyP>?Y zc(%W!Z7*n5tPH2aepX*`{v7|(bQeuJ+oFC!GJcNdPDuP>LJEB>YGod)$e`XY==h|` zV@xU(ZGMHq(J>}1IoqP73&)t$NXgWlNbU2wbBsw(pKX!np)n?X$_gjNPfQ^)RJ{IX z5P6Q7m_q62z!=9UN^56}bdxrszMs-c_5NOE%w9HT{5i^)z433KluXK)6UFuhDPasR zR#@ds(Z+})VwG$L)8;vSyZ7(3gTTDXh^<)taZUxVRtjFt z3ceas@M>1j@L`qe_b~&yI$`tKE*j?(r5B1i-$)6y_Oq2H5i?3kv>u?9zR34CE``#z zrkR;ejc-yK-^3c<4FwsQZeopFeHMAWJY!&5I-$s-#5H3~S|w_%mlEpqL=?O+#-wmj z2X!1X7FWEa9E9AiNUzkn%PzK=>mFsVkC!w%*FtUf`5) zxL@DAe9Km-!PYD~@jg)GhZFrW*v|38%C1Nvnp1O(BTz50IjSw}DzDrPy7*{q$bWo) z|D67@d7bBA&7v_!D~BXz7#SfV1M}}XzeO2y1lS7+VKUSLin_;`g@dKz{Xt&of11#? zeu^)|_1DFLy-tRKU^FMY!dGKsbW4s&i!CDLMzRvOC(Yf|uOn^Q zir<%tVPnPPtBO`~$S1}3l-q&(7vdMf9B9GPk(HIR`RAsZ9+<@6^glDhMFu@s*zT60 zC{Rj=Ui50R?Y5No@VU;`|a z?XW6*v(-Abj=0$!B<_9(T6yJM9GJf(-d4@3rrX#O>Js6NiV@t=O!oUq{9!Vd8HPJT z#Ls~pF84*{P)FmM}#CDtYAJ2EV>DyJ1;B&^2c|F{UKAnYLPpdqQFcU}=r1MXGULC%l*bxIANF47+@|Jwofha!F#t(h?c_(dz=wbv_6ru=y<1VhO z3PzO_tBCsJ2}4IiC-}mVGGDpUFe|1Z^WnG4&dt3=wNZbYSh&M5+(mlpE{}U2e!i&f zmt)z7=oW`?AOtJJ6=WP|H(LS9Y!?o*ePLfTQdQ(b9yBhCH+1y6qh(=#q^uG*q@b|^ zba^~RqeDihDmu{@CF7zlY}NIJjyzD>KjN$Kld)E{LAy?dBeV=eaNHUh563XA06o&B zoiLzA!@fYdq!!*tZ!Bv=w4yVjipr9)!D@dv;wR&_ZniVrrPBRDRXXe{3Y3lxmIQo3 z74{D#8?iy=K!Zj_fhc?K$~Jmj(}i6Ytqjri?NzGq)#O#qC8HO2D+ZA}L>JLvT^99LR4|TIhv8u_I zi&|Y%D>bAR?+Cb(Ky^WZJ#*v@&=YSW5xs2~8Q}Idn=xpXMHw}jsD`F90M=^gd_WZ% zx?`3_{pOIN2nmz4TzX6BLr(tB&$7rH&7!Xv`jVkcff7qC^3ItF-R&Glfqkn{-IPqO>n3n=~X~k>{7mCY>1oqgc4e zY>@Ad;$hPO+YN;Sb>Jis(eVa;yW@DI^mze94bkzqTUs5^j^Q@)2K}XU!YJ&rKi6_P zUBuvbRW#?k!A=QR8!X~#12TG_t7ItX3_!Ic(B0>1k*Zb6yoT3|t(^+ZuPU;IWe#-; zy;BLjlZEz~t%NG*Y(TXn(EYQuP}Qo0-o-+9&+g1p{9Ke4R)4%x=o3ol6D;)DawSwj zMSyBapohz~P}SOrKTof!ob0c#1vH)D6tp6aUAiI;F8vv-Ye=BuD%8=es%;jaux`7p zLW@8=v3fgrNF~X`3{8CkU-soN&(f@Kd412n7p;;Ad()5 zLnxgd#22o#GVY0UT(ltAskbXjY}xh0oH{(Kba<9^cnDH!NT4qO)oQ3;rPe{=lD9UK zo@27Pl}KOo0xR%dr5qAn>v`)HXWJSn!LGH;;SL}*2VEit&Y4aS_0SYq17AN&xacoe>{2*&k&9JZyt6Uviwna=c#;?d|`_`FXWqaNBGYYk}M~+EGyS(%$9f# z?kz0y`7mzsiMGUJwf74kwN~81hlrgJ(Os+1No|DO5|8Mj#MkppS|A3kml8>DD|Foq z=SMnKG2)Nn=Qs0~v>D2`Z+>;9u z(37v&>MTg$6*K>a<3e>^Wl_@Be3MeTs*de}2RB`&nAE$QMNvV|Yb@e{e_f3Xd@D(z*aQC~v|zlqB*n&i+D-W;-CYack;*5@imY67{;gIxkFFEW zErMcmz6&kjT;IJN=Zf)K@#|{nh>;u+XvO^<~iwOq_LrMP6UFKy`qs zH1sW?Y7J%2!~RQGGI{6FDyBVe9?xP5xP2ai11s5P@xl^#0V~;m&$EbEvMD{(8rNO} zZU#|D*GvtSz%kk(&EuSoc}Y3uC3Z}=3zcIOGz`!j5~vDLSVIp1ifHHyKv4~)*5lO4 zOn`9c=?rFRa3#QMyY*`Z?bcrNEy`f$(o5{zg83FD`KFkZ+e3xRI{2Y(ib>bbw9^-2CKZzwVbZLbTzru*`S2o(62aue)0BIrNQqx$ zuo)hBaXW^&`zftn(F{+&SQy@Su`sL|s~9#&*@j_Ddxnj|aD!&p>k?sDeTgt^6wTI2 z*@odGm&7!53=)4;ae1IbB0)z$(txP z!c9n-@s~?tJK=IkY#~9h6I&Q4SSt-%AZw+R6csR7D{;U)vVb#OBGzJKCMxgD6fa@M z|3-^TxI&A&Ld4~zD2;4!5uiZa=qp8BV|#ILUdiGd{Uo>c=iT z-TlaYb<9<2x>L|hK(!>$4S;+adKFNShKz-Jx>JF?a5M_yx5pE-1CSEOF*6Q;}LDnggbRg@OU)x2<~qpSZK zbhW17Eleie&=q-R>NOU5tFi=I2B=!Meb10?8+0vp!XtyO2PBi%$G7~C zGQ;&iD{@rSG+5*c9SEU5dYo~3TnC{=0?h+dt=nE^NVoO5UJG5rLPOW%qtb(tvB6+1 z^Lh#xa|GCQJ%>ipRZ~oAIY@=pFL6ZacVUW6dvJc?iyNwf~;p98XF1tW4VQgD!O0W$}L(Xax)K8?d{|)K|AD9Pt|VAZC)gD zPrp%y#Im044h6*wEYG;ur55j9qa6yg!SH=gmuf|G1fe|=iIDZ z>YHMcm8$4Zh6v0iFW)R__Nc`|-;k>4?c^U=980atq)C|MY9{wB7AAWaORmnHu7>S& zDcj`g!AoKcJLZ51v8SDS@Hfm<+*A^n-;I;N{B9Un)k`eOC?$a&1{BoLk+PoX|0oVP3x};TDA)3m?m&OyN zzs^zn?i7Lq{pP z#po48&;39L$2@7VIkqenj&Iy192<{{apa(xc{>M9$L^UNDYN?{ki#7s+D@Il@5-cr5Jjn^l>7dxKFzSYIJxiHz{n(&h#k~ z6&r6Qu?}w~L5H|yN(TiE1yoA{ZC<8zP_5*x&7yCUAoINC7WEh-rq!RL*yCB|d)(~0 z(_-ZJ8G89nDA;w=6w}tE(bg#KmnkMqze`K%gPU+likoUu>ru&c0G4a`XHGTgyt^z) z?>&{5Y#q!!r7~RMi`tAE0!k-iDOil)JJS&``1ZThs$&FR^ZX8=xQu87_LxmyVUd^b zsYl^S#aWNPRq>I$k8E(g&=pisM21>Kb>c#sRYSDVteUL}B z%d(Q`k^3;7^UyjsP!cVZUp3wU6s~|_*Rx@}@3Y9&1k4h;fekzCegr_`aZ^odDpvi{ zA|-mK8XbJfRFfv%Z&BLBsU|%Pfw!tzIjaK)qOrw&6aQtS)b1FGp(^gs34^M`T|1f9j2e!t4m@j#Y z%jhLO2I-zv7E#yRrkXahlyc29?U|`2y|T)p^#33ar8ndMjvkvDtnkI&2W#@o3eDyq zA{xqBcqnqxmQ@yI@S9>9igDe0wf59dc84q1N+xgRY&sr^qkJ{G_oJyM4H3OAkrHf= zNB6c)HR)QQbyQ7WRG*eVX`-!Aa~s7*bFv%_#WvcCjD}w5q%1N$%EMhm-V=E6f+b?3rIrFXWP_K8 zw>%-MyB$x+YN16?>?Ps}Ydc-hL0ex*iLGCK*r{hB>sbgr>((kg8K`TTxRq@o1QMQ<{P?s*X}9N_wCx$#pE}K?Ji(a@ zoXlw^O?cKK&yZ;*RX+zDtf(3UryMwVAF<>)Sy3JJZ}Cx+pkR&0qDDWD1muV;cqdX- zWP7Et%ITdt_D&tVbJ7dSI}F_SqVf&{@BT-7he3GfDt>4E(o6g39(Eq)PW4ciy+t17 ztt(+Ay!HD_Z~&Z?D@}Y{FPwDq%i^TL>wp6%)e8;|lVB&Eu}++H^(*3}MnSQi^yI6Z zIH|ZIWILy^+Uca5@ZO&2CIfDo`kL~Lf@r;Zb8{2tXY6CVi?2qlZI(&Je_e@ul%4+E8y4Xakb;ulw2)IW$xyh;#aO;{8m{c| zR`JJL?eXDnS}HzoeiLqm>oa?+UU>>QaQ%;OitCLHz`+;}3C=3u;5P5*4KjwW-ynBr zi{~mE8>K|%u*dVO8#*1&MZQR&7%yurl% z%X%l{hU|HE-O{%$bkXzNMbEvBAc#%X+((tcXlZ4IKU`}IZJ6g2{+1H{77NdOM+sNZ z@$af$Qqb*yDoCKz_wZ>TAY_&KfP#RqMqK}%?mi?ld@Y4@$&WsPSS@@V_ z_~%E~eGneUW_-zl!cFB4-ZzXbj@8XEPHlBRy8xbh>Mj-;gh!P5Tq z(Vs}OMT9PMin>a@`o1a-qCN)m8WQOFk5#Cswwa&cF<&;V8 z$V_fzC{v*RpQ<~=N4f1`hB5^@r3G{)^gOq{1gN%zUSY`d8SbWN0|WN}2-ABE^!>a| z((;v*43FAIHDB&D?k8p3PfYjq=c?BfwB`$oG6E#Or1{ww7P(Tq$>_CTcsEbN=C*$1 zaaUDt3uwB=Dd<}D0{7Z@c(KoBB}hSgzfxXQ&|6>YVWeQ3%+bj2b@zM&XJpbU<}vnL zxGj@bGc@?S&WG5aCQzQia5p#@V^@KW0hk9%qak`F9=rTTeveIA^eT(G8bD1Zz0T14 zTcAc3y}{7zZGbZAZHBgQv&cn#_%`w#HsShKi(G}MsR_`#jMQ%j+(?Po{BZCB<+2PZ zk!A5yJ!#koar;i$2yy5xta^c;DR^U~Y=i&pF2VQhMxq}+&7{^oYMeLXmHfey1fP-A{%MUWCEcy{Rm{b}C=N{l-Qu*vhnN-IA1RU^b6`VD| z0iVjBgirF%zyY7!bS3Kx;DArT&%$Syjng1FuWOt`e-WI7Up1d*!FgNb^!ZhAa(>e| zDSef!4>Zoa-vnnla1!TDGilM)%Kr~Z34gW#|AuKMZ3CJkOY!uUKPPMlddqc+T1{`? zseu2EJ9)_OD&Q4#$L|QI5N+Y$`i3Fe%80%Gu*j9ymqVtNL*^6IEJIBR?TE*;bLAdH zOcE4MIaV2kkD6G?hqZw z*t(K2Z(%4?pd15jCDg!e>j8!6Mh4R2+<1Dwn1P4l+)f`NXGG?NZH+>=IyJjjX?8DD z2IAd>%&VZDUEQ)=QZTZ_%dF+LuFxctUSX=C-P~$yw@UdYhqf-QA?7>Q?FT z{&A#k6yE6W)^BflOj#K!Br<`3F?F=dt!_xhyWEmzTsJ8%v+`DWHdpg=KVya6ZQvAeZEPV z{gm26fCKWgxe4SUX*wQFAn%hRz`o{=6Fm zepq<0_~CQlI27ZdS5|Io)N+SYvuC>S4dpXkkQ5F-L=8O!9e0?Uu&LphE_ga$+f!?; zTFD!v*SjEHyx&u_Za3CqM(}uP2gZg^;?ij*xo%Th9U&!r{TzI~a+*n}9qy*IC#IQn zW-9syGh&9|jgb=g&jTNkeYYu*z2*q`3wNEl$Elcj0XVodr=#3ztKhO;ZZ+aUf@0rw zerLH!Mq{J2y#aE#`!~0^+tnK)IP5Z&h_`@)uq)^-AvO4DFp4E>)g@uOE9&o5#;ag? zPj$(A+$Comt-3@(2YOYPyvJR#sE_Uv)rv0pjJsrKpZ`UdEK^z?AthYA5w@XAPD+z5 z8IW#wiQtWq68IkhA6>GhuXM@7$HHIel05O($G}0C9M?~}XmUKz)Y>414$x>bN9dOVk!EEW0V{*W#ZI|F6WZe~cE0po7OCIRTqvV0E z7$swJRF^2I#{f6s7FIzk2IwwPFnKHJ!LAr3M+}rMiH!qSZWhn~k26C4KxmDzjh7E_ zUN&CY!Llm`ie;ZN4lHXG%LW4nmK{1sEUOyS(Xwc`Dp+jW)Nqf}wx^VBPqA%_2PxYW zv~{qX@8_QCiVTo9MBAoX)v*3Q4th4!%2wTz3nO6YEkob~*q0|wqtOKWjv6ZVeKZs} z2%u$xGYmKgpglt+fc6|O0n{ugb^zs`0Pl3{_=%NK+b_-cIsLOv`DY#b=fxA0e-yO& zMAi2SDm+R1N5L45uX5M^aFX=>Upp(75?v`d92py!{N!P7B>Q@rq8-zMY$Hh z9NNtC8z#HSTUSTh8G0|D*W3!&I$hiwOEI0zys=hS;lO0_o=dGPbL*K{xYA(u`20ku zF=3{gT-Fdqx`b;koeAIZ4*tBM%E9-*!N~D%&6FIUd6u3Nnk2YBmJ(^r#$SE5;2(4j z@bQE)<00kvDN+LeIq=8R%5Dq=aklkA+FIa5x?u5GGU3lQ3C$zJ!T&k_rVoO!6;)mlLf5lbY5j?n4~ha}$Wt zPbx5p(nc4UwDJNsrA;U>sd^sBI_aE(K(M58uD`_Yj>aa%RdvgH)h+AU6DM7$x!pOccY_=5tzWwi9>6Vli!|?+5@`2owD{|hf{|sc;Y!yB(1f|( zd%a|tH-OVwAjh(d9m378D4kR&uT*=+R`PzX9oMT+R?z7;xjDlqsC==?FbZm2tm9b0 zM`W7><>n=nGRtwFd&c{+BI$lLaLrBU}K{2TpCU0maWnjXw+9XY1=%zj^bXaAM zQlh(ck;!ZJ3JI%5;Is>?h(D^n6>F9_YI;q%Ma9$O$$Ss_crsRzZ?AAu1|M+ocrvz` zymq&a8r2RLuHhptitdrFj&-K1a8xpRXVVjmu=pNyL5j#Km(p88PbVYiym1fG76>Br zOfnXl3HQReY&(rs1`dFM@a=?qWgUGNkdmeqn3N|tLx2-5FlpPpZc6%UUQb#kEg5L( zdSOpG;6CPz5!otjqtJ#CIp98V-Ajzrc4~=K6-}rt$)0GtEagq7>%LR2`;J}r?R{z} zDkylr8j1?~`2lsq`W+j3=R?}5s+D@=y`k8aOX&^J#odvIKYR#%enEjr)+x#*v!VD+ z1t$IRkefWW7nqdxu$w&h7ML{dVFVzaLXBCk_?1YB04xU~mczk^WqG>`_{_Xem`8vE z=0`mu%s1i!%zt|Xnt*wOAV+}=hDWUwhIg%mpWC?*JJz(@a^F*W8B2Q*Q8@=*IpXfD zMmO(8D93iSyhjm)zKEZw98P!u@T{<}c+za5@=@8z^x30s>KFD$$Y5&fS40M*Qa^P> zh=zDuSvl`v@`l3xYO2NV(zTDd$r}h(`^xkEA*ut%V;^_(zOxJPkmCsevfZpF6rlUKX%&Ey5Z%wvp=%KTr{Dp7 z7X6E{?_2{HKVD!`;i<_45Ag=UmkUff;0ZUSy;)$=09?FZVA6$8xXJTrfl04E0jlo` zOj;&XyRePow*r&40y%!VNrTq9DXsf-lY(o3jDe6cMv35G^~69}u~r7cx4>!Fi-BmR z&+fhE51qYux%wRW@;LOP>t6_Te<{)Bad2z*C+)7hoL@vYJgH;<@;D?Ig!&EKa>!F| z@?%ABl8Wd>rt9%6o*R^?&Au-(-Y8T>eMRMd zS{Jv>=c4mbF^67bCk%TIToN&A8zkVHq(rCkArBa2?>{Gl?7=1(!EZES1Rp%zq;*2D zUP_zLGu@;ejCS$j1IhKi>WYJ<1lkdJVR*;l1L=@|qq7lD&8Ml(9)%{v(=Gp&ce1bU{{pD67|R@|sK$W`i!tvDvKV{r1-%#>r?L_jW9dlL_-mU0iYsv#44-He zu|p@NMd_^)L2I0mF^8y?y?4Wlcsv0J@p17w1P~xh8Y^B!cu7}KEBD2A)D_t%eO_}@ zhUyw><-R%RHH>z;HXh^iD*$0?;1`+quXmGc**L5Mf+hYM`hgjCd7YnK>}GAYyzV9! zjZY?Tg|BA3+9LilqgA}=CT|%nV73i!qRU+&Af8J%aos1Vb2X}}6tGx`^S@$@Ik z0j*M^<7(h%1jg8A35=&UNMLN*0RC+q>#qqF6xg0@Y2(qC)XU$O_?@otmYSJfVxuzv z)sjGy-qPby;i`3X$^_N(FXMm%UbOHhoZ+|J>Ji6y5tb(<<`=A~RS2Zt?)a87w{uoE zYF6dKA|z$!tQ@A71gZm(%5cQCxaBM5N)=15#T$6k^&0zoEo?0!dW|jr2#_5;3d8$0 z;>$*v^R{-!Yw;M@SkAn~tnPaoi)0@18@XxM+Z|F=yVa~KG$EP1b1TCo5qg`MTQ!g-iWm`ht6C+PiLYHBCey(q_gS{vJA<4-qZ-Oxd?yN`%UX=$Kz8r08=jXMV15 z8l?md&IL_B5(kCy>7c8?k;iq*B-Ojy5pNjR@4qMG`uX?uxX$Z1t~qW$-mC* zLwlHJ@yBlUKr#39WCH(Rp*j{k>7U4g=kiZ5o^eCC@oS|?tCYYeyxqZlV>8g&nNbp` zw$;!4UJ)xZaIk3|(PjMNz1t=?Wz=PB@I!!A8mjx$O|VL#wtVU)9PXqb>K7(`*&!}Z z$M=lcv?u{d_0$&KM=fF~9|D0sy1PaC$p5+YQOe|Yee~Mr=p(-S&YP_KbSUgb)|mZ; zTP2R}n}xhl$dN3#ivycuH{cyTRA9TNVXMwZCn1e z@`-9A_(U}*pX}3ZYV5u=sXpJN7O~de2=V>b@FUh%R=#qCy#?a3Z)9!N{afH*axD~` zK5!O%-Tzzh^{L;9vl{Z-`TE-L;cJJRN&?jd(+X_oq-=LO=~?BZXA{2ixv1|JI4MZ< zEc?aS3J?$)#{sGZgn0OJs~Q~&Q=NYGA8r}F4GX6HWIhD!-wFGlmP`%XFjF8FsI{fNXT*T#@bZ#pg(so)b2~>|qZneFUyGzMaqpMkYrI|zL>QP=;roc)^6*__EflOa58wT};ZY9o+$k!GdAkbM zb_;gPP59?PO2ob3idN;@by7m24-%0kQ-6>&dG`;9dJLKd;aLJ4@ErZ4@Vxa$ol2W_ zD4rXo1W!I01*!BKpz$DU7FI#9!VOHXpJb3-`4hxpAktK2LRjOxqj91?!_E$oGXh)2 z?TE?zS!tpo=hp;YrT)sjdGpVR6%y#sUl1u8I_DSMh0yOD5f1>EQ|60ShiDH+#Sf^d z4rb9sT`=5E{?)C<*u@Oa1<3fcmnE(I)lIIj*-9QUJyS%Ek`gxYjuDuiVraDK(j1Z!XuN3zXgNUp z3;Xh;{*Y~3X3|0Y4oBIzQ`xwaZ9LcT@KQ-Z2giGOQK6vG2_DKQ@rR;iL_0as&g?4)Ul?juQ?B4pMH)OeO17;6TJJf|qCX!QY`uZA2i?1dRN3Vhvi1v>);oE=u`D~XyT^{<) z3IqdDKi3*D%s_A!SIM#RK%7!J(eDfUBT;^je;`g%kKx;m!{_>HxzWLFN+k|i=UN9p zVB!QpyDh#gyHZ~?Q0*TZoW+C*&c?jTNWj_HCDu5?k1aX_&T-L&vC7dkzTzlXI#eEB zR9WrkN{6bm$}4@*Q!9(6`NI*MfUqnp+1VTj{zc;iU(Eih~{c{#To z9cwEI#OBp0_DR-Yu-Tz$xT8wyXnk}`AQ+t##1DckmZSH}&vB8XJ?NA6%&Si)L@I9V z9qslvVlnwi@r7-E_{$zQ`r8R7YFX!~HrxC;ZO}uVqX&J_yvoXQh(1{T(7<~&-JNRs z!{LdQqtw&nr1-`%WeZ+WIBfBU)o;Av{X|lHQ|CzKe%~zleY2X*l_~ZWm-*3^S(%g+ z-*SRdD0TOKWYE1vXFpdk@__F4QeoNI$@}$W3={gVqP5AhG$uL z+Eg1OWq&8;2W>Fb=eg7HXjZVAr8L<3hvS`meI+Fms>-KLE3mP25?9z?#~Vaz2gkaq zSbn#;ygU$!28!v>HxKG=C;X;;>M@%~7kgaD!DjjTm9eom-oWoHU?_HaCc7rU!4F0T zpMwru>e5cn{zZXO(r_`T7e(vR`ikg=={ z?Q)xTxzM&ZXMCiLjJFbFmek14Pz2RaUvzZ9h=4j~{vsH1rF|F#8NV4e@Ab1>qI{tqv8y_VbkF*nt1Nt!(+&JZ*FL3TZU({wRHU7EP zVs`>I)*1gh~gyzbvAK?n_h=D9jw!391$%sj0}nICB}6{ zF8cH+++%ViF5%zjIH8w^`qhxJgd>vr)!LWdqS~k*KLBw%zeK;YmxsI|x|{L#_VSRc zh(7wUUfnY7*VP+Hg3CJl9KoO;2z7q8EOv)3&+ zgFIMucQ7>M3kHhE&YD;m)t#UXZ9K$j@H;Vs-(iCnc|80n;XADN+kk3GpdURRN;+V* zNsXe;`%tIvYLgD{4NcM1(!S|LG<8%E3*zDB<+eJlZPfZ4Q(^ng14&+X_5d3X01qXi^n%qUk>{M%CY{sUL!PqLCfyHx7&BKe4{M9LUSl45G>nR@ zHYp^SJ%M=x^Sl-PbF_y%D^{Cyv5mP*Fb{8wd5^|ywK1CoGqo+|LB|NsvySPsW0eyI z4zi8UOm#ZJtDHbS<%DCwJm!RYFozRf1#@;nreJ2Y#oVef2Ya;>3I($tFk?kBwO`m?PR^exot7(>2dV z!5q~Vb85Oc;SON_i4z89XW7Pwj&eF-f^tHOFrN(OG1Hz0b2!1(7i#dpDOOIH+7|P; zzT$*h#>8;Q70l_tj5%Q`FyVx+Y|M~g@*a#BX40`@TK=(+4xY;d^X#@ft2Jg5FgrRS zt?$UHaJ66Me|wptJyCCa1TPYL8alW4{JF?QUmUBSIxzV6%tMvswZY1YfUn$UT;EH( z90HEE1sn|lT{BcFJsNY|Nf}xigXFErp<~bx&G1kk{J@fbI;Ek4l2)!ZskvAk&5dIX z)7K&8q&2TL>A?&SrEdiI_xi>rV6A4W`O8OZCWbssV+ z8|SXoW~+n0F7M|d7nP_}$M`o{p6!PRkRg9@U{;`*PQVoUSwHkK=7!7?Wz)%e+L+j1 zHx)_~#ud*fm!hd-f7hK-86G}sRv_rNjcIf{jhe{El1+pSPxn_gFtDq?wt+!>Ks*UZ zeUFnltwA*4V*~rHHYs|XR<5~y(}QU0s9d{X^s%0}vpCIHTv-(G1#Mle-cFsT+Z{2T zJK{JMH`AeCzf5HW12q8ZNC7WoVkI)EDoO@7QomXU;H|^w6=#9PSD79KM&{7js5mN1 zY$m_zkQur|re$?H>9jKpYf5cF?$J(5g0>|=SW=q>OR|W95Zed{D^7;K283lNLq}!9 z%xuA&nyovF8}V>3i{#o{mrXTbw=f%KWzk#!E3-Y+V~BY5d?|4j-W){iS3r&EsGNhS zOv?FhEcKO?*nOuhZOU<4JkPdx9xSfT5sT+RtA_!>Vum&Ys>>n;nFD~CEzo2@u%9un z0EC?t3^fhV-e6F?VP9vHTqC8_8<6|c01tUodc+hZf4--!SV{F-FuJ!KS0h+xiS>y>!I-GU<`j8CS5O07oiEq0e2nZ zA^vfVVM9D>+fuV%ofvGN3tR=1x%nm)5A~>5)_Jp%DVd0;LVifUV<;k!SVY4tH4rb+ zB2GPCL|k~hh$t*=Ct?jyAmZW^LP8z0t*ep#8wDik{iS%cOiS)d4+e!ZcD3Jce z$s)b^+;-CE|veMMSqzA|fT&PDJe}B?3Oo3n~$dAp+CTn`q>wdTCmso4SpL zdspIyOq!Nr^%PqzoIKj2R}0LcRnYi063pi^2gtQ5mD~q3^Oa~y1oN`vRKmMYN-T9A z0U88-J=&vICcRFH$wGI8gyEyW#HcAaMSOMbDdMZ9%67hb87S!S)u)QDazpKWHF6C5 ziho~WRF+a~dbTQ|7@zkCaZARCxD8`OTz#mWxTLYtH9w6NaV_n|U3waeV<%g4l#8B* zMX>#&)5P|J#(A`#|8`YSQdw10?zd+ZJ)6++$_9g1lAB_;<(mx5E91stwjkQX+43Sl zwHo>tpgIk;j?=S+!jjDe^~sw2k3z6FfW%wK8<_1P}E=>k!((738hUpHCH`?fU*& zpdi#Pm?VAQI;UNzbQaf1U5{B88B2e14nR7ZUC391KG!50#*W?4G25(@5$G(uaG771@MljqkEaj_I$)C-N!(m zDOfBCa1($!0lG}ZQ@e<|y}6Sg;76xA?{Dsm!&36*sU9L%N;FwL0;C7x@Io_TnulD4 zTp5T?tg4^~d24>jG!Ci`54YMav6Y|f2h`d}^(Q)g^-RoH&#WFTl3Aht!3j$PHkU~ zGcrQzH3GfN>i)YxEBG?Y5CvamC8)<@QS5wS<1MCjjOUupy4b%6-~K@%YoQ(q*gE76 z)4Jg8o%}pJ;wv!SL-@jjfpyb8ymwNa{N2%9=X-{ZFG*rAGqkjZTrCxkB!4rEUj5kx zIhp+a;x}XEnpNw=QmiS^ctY{HjkX%iHz;+HoF{DEBk1tz_4jz@id zY7~0Z=ch*(B4Z>X+|x!XTc${ff%yvNY77tc-sMzaCv{z8QoRU$0|>{g!I=;J)|hmF z&!c{tvPDE62GL!nnDm;@qkfvwUF4y(!={+@QISXeG-asRL;TZ}_Y^~!eH~ZBOG@m1 zXdbO>FgUF|V)sYZQ0|X|bo^y<{2i>j=MkgpJ0^WxjHP52(If2agG%t^#Rdlf#A6B7 zIs+h{ODM3RME9eDB@_klhO%g-amYI+y-SJbh-kTS(bWtP2SnM$(pO2d%dQ5$ZpDO-;`VqDx{1CX}S?qNEr>~k}{g+#!?17!Zr*y*EdrKFFU zxIp?y2ub=_f(!hz?VBNJnAr8W6v{YFb@X^Ck;`tvJqx#%N$mnD(Ru?~S-sp^rQU6D z4%F*1N7TD&j;PmItIV|3dlx7@rkHeWSh8DHSkz07NOsGOK)rnp<=DFl9lQ+_lx-@< z&5s4Knq&9VQ9IxGZlg)n5f!!bIck>xswIJ*j_9N{U#;u-2aGT0u>CY5aY=Q`9hblz zh|9Pr!U%Dh*}mx*G*LqPWv6qn{^{PAf!SGk$Yp1xJC=*uR%@cR*^n`slS!YKxSR{c zjUSjaCfaVHI6JDc*Kce-XD_+@=W438fPE{Xb!_aQYxW@4HC6Y_CgYM1O?m~o>IKav zP6k_`E&qOZXKT0rmBzec(^?izYVXOu9s97f6Xjd;#*zoHZtmpNo9? ze@iAgLk4t`Ohk%xdi!X}$)xxFZIa0`n@p;l+b)@04;wnY>-^_xwofLaW22OYo%pf` zv969}vSE`+t1VC+zN+uuLwn+X^gHM2e?tM-sqCDldHzI2&8{tSMqgqTFQrjk(oCz?NOy&WM zC6naQ>$FHdgxcD zwf<+x#1TX~nTU1`QaZHEVc+-2kllwDG<{}Lo~?*8nF#FyDcdBINF9<%w|Q$!TZT2I zeLv@9^6to^(q3O<(oJ<9O8a1qN&l+zQ2ORICQC}SYz7MjUBjT4`ME9F%Wls5GVla4=6`>xZ`*sb^p)8AL;bPqrCT+sS zh9^ui&eu|8XsEanfVzG9glTi9Y3-$C4P_|@rKPPk>74UDlr~_kNf+Vbq_rm9jf+#( znzRlVlh&H_y)8~6SBVpVy3JT?+G6TEic?1Imv-M;llHs7Lusqlnv{i$r`MV^#S8andJVfA~YrTojJxi+;kM1*F6lx;%fmJ1Lf|F`)jJG-;_Mxgjc8doy}f6%2e7t|`DXiCBg)f^_C{n4Wg{H(&8n|Vx(sT@7O(dMjLkRe0mkN=?Equ* z&4KmW9vPd;9s&NjJ&yTCtf`kW7D!pDYaH{Z91zfZl9*OmOxA~Zm?OT?<0AsyA0bs1xL-Y5w*B#o$x>W1d)KSOztm6$3 zYxm(AKz1K~VC(P~edrJ?9i(M7ca(Vn-1584R`p0{k>TK~a)+JmG}q9X9lXAYE$VfV zBPDF&^fuxm?aco+y>*b!xl7BmW_FOhfn_g)>{zmTfk8W2eFd=X;yCGIt(oK}HMj-% z8_&dE&1r?2Ix6-XD^_}OyZPx#2(GbTy*b{(QF8ldK-ivN87%fwQvAkirf{s^Js00t zwnq%%Lx!aIT=||9^Q6%U(!fEK6hB6g5D}dqN%47tggd-Wkg!oaDSo;j;hB;_ z(~j!!?VkOuR#Bj|qOybzos!kvc2!|r^1gj^2@d*0vv!>Or}dqP&%=u=LZ|xYkg?e? z@afyn0T+K3gwl!x>iqN-KTfu4#s?~jBJ{QKpR2l)RjZ(H6jV;CA3efkA{ii~M zl~FQM4Rp4N9lWl2?eRV3a1xbl{GBc^}*k1I#u zoKB;h_2n1ye-q;oe{@8kl#Fwkf_)QIU%-iavvJau6`+d0?B$9;uqrYZCss!M6Dmsz z$oMLbIiioOTH4Adhgt#J#(M((LMz)`egkXV=rmRmRr_d>u8#PlWE8Sgwxd>s_Fu^2 zQSOUG_$%BW9izN5s0N|tp@#dMP`#> z8MeSb#nm=e6_IhHa;?)ycQGlUoR@B9Cu>C)$05SBqDvGMpj#Q~?*#fvO31kE?*uMj zURvN4N`!dkN(BXIAtU{%z}P3<#+okt%`wzQ;^nedeRJ$C_JzZ>Vr+oc^4U68S%Ch< zzoBK7SL&-`dXlUDRG_%D!v1}V$^H=1V=(HA;^+tCZuNCC`&sA5_&-!O{`6QGdjCOyN07xWNv(O*!Z`cF_BmdI%1>jUBY#Wy(Rfv4e(QrX8ek z#X+~RKge6|_mu>KrL>&KZRj$bL#2IoJLBAN8GJTut4Y>*%4f@AvTv(Nbh(E-fvqMD zy4*v`^rxj8Kb8`v--WlFe`h+fGTW=|Z1g$peN5T=n1O&Ozg*e-7~6XnpbDbL*xs}S z+TO=leewoq6&p8hfrq@a1(?6UL$18@a4IxC$<41WfSpxaO)J`OvV!_F%_l=k3O zlLlYmp`=x)IbXSACw6Ilyw#*Lu0Zc6ZQW{8y|k?X`C;2kx(-xd&Fe|c((*7`aNPVV zrs9}M6m!$0gyWw3ZyYyzPL;3R_Euq$(`zp)uf5D(d;bdD#gRb2UZMT;vcZclub&yc z&d`xpddQV>0iPuG2D>N!N_4Sjn@K;5hCM|TG<_2mjo4<=6<2yF?X+zsJ#?jqJd?MX z^d>IO+-B0wD`glREQC%5_^^(L(IHpCk$)O3;jnFDp5JNe<8geN@8faE3PY~a#yuW~ zfI$$v&bheqDttT>rd3>@xe(`%;VbbqJX_Z-^iU5fQG7F`^akh|ZoO?Gc9!PCE3;_> zSAB!3B>y&(LKmor7=Vblew#@Hul7*dE!#}G@M;ew-L=i67D4$QDC@VGwB~9Lc|P7| z5?$jV&-dF*`r{f{pY+EzlM3g>7VA@=I)!4HGFS(?hxghBm2n$+-Yn58_`C?!-nA5$F<-t|(uMM|{RaVzp>O{tX%8p!dF}5&#x+qhJD3R z2{BuC{kh6&6I8$Fbi*J_eguOewLl%LmX(g|omkJd{IJ_I8bTTNnZtCX-B zguOAuh>|XEHK|$160nc((N>efH+m@TU#%wHf1`&yJ6la!kBdF6CT+jbLy3tyOd_#2 zQA#jY-wf<9>6n|~&ZGf5OtR)HZ3)bacbGKhCJ&{}-eJ=0n>^&Xb%#mIfz2z9jREDv zkEI0Dco1U6@g30qu8;80S9eA3Y{gA=$>Id=lEn$=k{&m!$w5H{fGUUA4S^pmwYyDL#^{EW{>2bzXK~W-( zStKRgj9;17L4gFnOl(7aIubl_cF$cR*}YZTk=@gg-BoJ8atX}*@9msWRc>3>Q027n z8D-%!toLQNUw&47wDvZ^^kW8y_A4kj%RQ6 zkSp_Ip7uBJp|o?BddSs?niBew3GQ8rLnpK7D~4VLG=sipU@L&)Y}&z4%57}9EAtW- zw}5X=a&GfbpDA>C7l#a$UBsTB@E>PnnAZPPy?Vi zn{Ms0|7po|>m423T_vIv3#R>57g zp$#LS4D^3`4g7 zDx^gW%mYx6Lr*jGzbe)yx%Iv@>OEa`7dN9x4c?bVN2(m;9=X7$p;ztrdcKn&8X@4GDVRr8q3p1Z5F|ziT;Y zFS?!7kBx3G>k_$tTbGFb?ac+Me^byQKqW-$6463BEz~VUVJV~ha{7Ra6)u!#Otpyk zK^#%$*m`U|$CG9{@UTzn-AV;KX)<9S>vxL#6pEnuR%mg)P_C|?RAKXdgyFn@nv z8ojX)VemISWdHJjrFVWswEmwa8d*Hs`6;V%aG_fBQc%_+2W5GQK25|sbpfCfK(NO< ziyRp3Ws{+3sRJ(Y{gJs)hWfI_XNfP3vC$uk9Mq!r7FBD~tt!p%(Kp-z98Y7sTy-}O z|8^$g7*^@ixYhO^zM^^;6TOCL@104MakW+ClH>HD2GkP%$LF#E=eXp|g*9aDP z;_k%`VrRUySl-Th-=|363=Y;K%z($1UB@XY3nhg=zKA3352h&++)PQEqIpu=ya~n8l22TMeiTie9#G|XXqx^@ z9KJyd<$J0k8o+3gLs7NTqYyj)Z3LjZ6t-6F~TKE!o0wi+Z6hDNr7hOm>0NanS(Mh zFQ6s`)`K}F1%6(}mZ?vp-XagKp3a0kYQ)|0SqHfxRkvql=zTSQVCa3zvm#hn1iDpH z)C|Gg7O|j zJ*=oS7{1aI9PTUeTK#nI$;x7CJU*?N^w-mx;fb~3zpB4hP|9XBeH8q7Z-^HwjiV9xOBsJu z2?9Z{r?|MZU^qTK66Te}iqxYq)5L0-4eL>My(}HOBorzO1WJS9J{Mk1$?+ZUQLEj! zZpYw{xxq<9N%~gUn9|gGO`qmgJZ3-iFIJY1Hcqh?IdZadyINJ`S)q?(bpSmfxWGN&ldDZg$t6R9ydY<>0r@D&7!xq=7+m%g>Xzo+WTF^qx9l+X#+-da5jrKx>iagzNa;f#me$6e{1dhwl!XZ($ z?9P{aNBcryZ*ZUom;OFz80kes&*j=EeH78jU9dAC(n`jMy`{lIZ?JP;Z=XQG>o3IG ze)q0q+-(@?Bh}Tk>0^mDoR<<=ep3~8E-3Yf%1XSt#M(#+tI-H0mee|DTk7Rc)e(qo zUZ)tz3tE^|`H(sq)tZPhUvFU&WxkES#MoNFp4ucfb*3uIH)l2~nF!wQ z8|_7SUXHyLmYk_~M$3DhkyS`^-L5e<+Vw)z8B@_U!eodi7%K7<_x1Tb!SNQ8{JW!> zVL`AEOMqZ^;)^=a@8=aDyshMrUi8ZtNe^&#{}&yU{2@#xit*XMKU$dd`-^x80#zfZ z$1&%7u!TwOmOF@P`YzWr^+GdG(|9$_ofeIKma^;<^z#;DyONQRU0n3f7#r>MY0V(;rqzyQ%k8lVG zhZ-$I%Bvy+-K%_&BxwT<17Edds1gqU(;TL04mK&-O40@#)c)N=7`` zA66E!VxawNsF&k=if~;ZDY&mhd^7en$^BeXV3%9i1CrZH0vp(6z{a+Xq26&;^w$Wn zY6Tw{hB|urfl-%h!yQocF}H|;5K#}Qq{kfU9oH=<_eN}0)LiWMEAuLDn>Dx17Pl?T z?e=4E>(<@kQqh3hmzvv`7PqgM+lr>SVdJCbP}hLdcbe087N;MX(@#xsiaZy1JCx>R ze;``ZJ*xNE!(3di>;7U7FI6|Pe}rY<;?jaK%vbZvZov09#rHSndy&QWx4+|iRS+MP z(Y*2=jF$Be#q$s5dDH9anu9;Mr<;!&{C+ZI+f*n13i%(Z*R$n4oI)9Xn!*d$yWVgR zwJ5JtoR>+O;ivl;?RXPC_Unt=P(Tz4ONyxctEs4rSY=pQi6t}5R5hrGniI#p^qja| zV{CNIn-0qI`ANRK{nVSh@wV|IZRK1`E zXjJO|bSe8Isw28cqq59!1=RLXeuKFnB@%Y3gZ>s8V>|0B3IkhbIVd*)h0C zuMKR4HT5EF3yRA8W12?cg({m-ch1=V_+RUfOxYPLFi3KlWNH0|1JCVwZRNQ!)q|od zQ+z!2(Vu_)Xy){fn_|+PNp}4@ZQGkK9r|CpCYbbKiX(3we(b@gUabGDDJD7mc!!#s zp4tE5j%DB1?{A7sjRJ1%l|pyC{O6qQ2cP}p^d{=-Vln&Z%{Tt{_LB!c{GlmkT`jW3 zkA3{*oS)XMIH?I{-7KoNX3SVrz2~#Nzcj_HyG6G4rvtUqUi$jO3z}fo!=jqMaNGS` zKYjVGlbfK*zEMZZvOUjz{m~0+A9gfBm1|Lb{LP2IuDEZ{d(I}Px?5CB_RjwHtvA1~ zII{_=9v0P;TW8+C_SN@4JFy9>oEvqdPG7SB#ithD{lPa)nWBqD_VNoauX^CI-{zds z1hcLd)$_9}{`_nGdyh2fq#hR8f+v3b<+I73Y~!`v-%i!Or}bj7%flOLR?%t#N_z{;;Q<$uw~`6F=Jo zF%NnM`NLj(=-;^1NXL9rNX;ryVkjK+`A3I3yF2zR^A#6*gPAhkG%n-0rxU(!PU!(n z9rl%YJ7bxl^EF;ikc>MMjPx+k%}?T>4s!0|jUBNH`a*d@pI4qdaBB z;VV7GWnMDAJ;6vHL-g&c=22~ApeHn}Y}6=UIT@eCa}2U?LMo2|icITD%PRJTz0oCG z9j#!)_=Pq5qmS|i)jWMVA7y*>W;BbVld2+Qw700!3qxq3=QeI4s3N2aNF8Sym~K8R z*mNBhttvytSr+-xD^XTZC8U-fY=KLgsHGvXDn!OP7X8tyDJd%s`&0p`rJs>rLiAbC zCZ~-q#DK=`rT+F%WA#GbaD&2EaFx?Ty6}}!_$pPH`-@QLVWq+FfHHrJVN1c>eo+sN*xcHkY`zsnTT3ixiC5&+exun1%f&Ii3+KfrnQjaKJbxw$sGUi0eM2r|)of38UIQ?N( zHeG5!X<1Od=W1;KC&fxUV`3j1AIEXrbFbzl1Hm5YH|;+9gr|dI>j$H# zj{M!fVY2q9Orjr!pGCMdRo|r}Z}Jv|2Nin#Vc#g9H%P{FaYlNnuaM~8dzw>Y>0_Co zs|gJEjPB#NLgTG?BONnP=qv3qSXj+_+UJ@pa&-=z+OmBtO>9q`Jpa zD2|9GV}0pa#Zc^S{rS^?jSh&b;(juUaCQz4LOOX7hqXjCn1S*3- zaqnwulzo4GieADjV}UY)UaT#ThosZr#{^X{NaKN8@;w50B*9rS^NZ26DN znaewQkiGZ=2a&B7nf}tSdI;RlY->TCJ&!jhsfT6FhYoD>gw^xsDP{N*HL&`2;P{I4 zYC+9`o_{V+qoR-G-rjdVf-P{bG(Dk;&(Ji3YlUWMEoRD>4^N{?(Yjhvgw0IQe=t0a z9{boq?oWrO(XSsvrkL>Ss~(+z<<%e372LY9GJXZ#n#E&fGNHw*gCK9Q}crjO-K^k%i>Md=O1ke5qLIQn4b+J ztJFvIv+=m!My)q$X@2Hc80vs`wf$T@WrV2SxaRxjG-soOYyl<60T$%djSk8fE5Nm% z!sf&6CfTc$%@3o0xY=$JxyRd0TKK7hk}sUonkuVOs3Vd3_@Q?WHZV}~yhmG8jTAhD zf^Rr@R!0PG!JW z2iYp;pSEhiL{&`oN3PT5O6KbrWmx zmnH6|I7~)}SvQGzA=(^=o_@qO2W5=H&G&edr>&g1VVi?mWWS*J*GZZY@&)h<^u82pUc z7twJa>-hU;+9ms=t-DAqQb9?s{Q07=`A@30^(bt%&Ti6npqeyQ z-ywM-k(Q)>Wwb=QAyG*{XlOS})$B{%^XwL>Zo>|q-AbyA5FLug+|lb_In+Y@!=mIW zN#V~um@`b;XE*8km8!x!B?b0(NZvf%q`7qta{edXL@&{ls~Ze3wdnMk;;8&QBf&E2 zj070L{lE{7KFWp!#$(vYgTO!vI|N}0;OgCLguc>Grn{1$8Z#M10a7{iEJ|r(T`ks z60)WHyOYn1e25(Wx=VjxZm#}|U& zi^^v3#jYI=$|&>u@Sfx;hBGR8m-vZ4;;34*GZg(qV|Xzc>*Qw+ER%zuy~FW5-XM!9E4Mi7@qe`iYDc`*gsj42QAt^nRYqTRz_$c&9oCBO~3t`?X(jh&8Ynj zq89a^C~52)R2n7Cy(eJ6@r70FUM@xlAWaQwU}H;^FE|0OUgHkngI8fJxNscIIstL8 z^#I1hOW+=w!*oLrf=K{bs_j&rfY5l1v#XKqqlHX)6UYKV-9IWUtx*sKy0IBDH!?w?*u$bOMk)R^2Z%5sb-n#6BeWYRlUTsD(7Y0 zORPN%;^K6Z@&xrEQ1_>sblwpMxgSk8DeHF!xfi6HwElMoIsTh&(zSn})LZE$RZ6LK zDD_>sNfZBYkb7^sNyX$O_iyPYttO{>G^DOpZa7y`NVOeLOG<9$FsVjZtpTenhe>A` zPIC8jnA9S{N$$Q5lWs|Hl4G#LqHCEe;UX|dF^ zTvF7x7R={4OzN5FB=-`BNga}$l(`(NDEUo?N!h~j9Z-JlFlkhhliWW!OuC}Elia^J zOj_OCNe;tl(%CjADZ>V&)MYpsg6?+&qwc?~HPr~}8lYahtTioR6l`8EC=XDu`G{1f zvU%mnPG$2UEuG5due5X$C8hk>l60HQm$nJLZjC2OIx`A+y&z8^qK(Zk%3vdBKc@oV z9n=UTf6-EmoZ3o^Jgb!$c~>hjazF}MMn3Tr7?}g0N^tAJ5CM>Licz@ z91xl8AE`iq;1U3^e66rr3s$iFI&gsH|2S1Fe@7d!{FXMb2E066m=#M3rt81~UjD6( zc)4?1%gbEqfKc0NBcgnwOO>>hQdMb@QayLIqO4CX1E>4qHpYA00c}|nQKOLOV zIRu1wDt5ndPL zK|V{buwC?e1V7;M*H8`~e@)8ONQ!cnUSHX@Ui5qlqb!fdrK9mR8c&D3;RXwldR$lk zrBXtTr|VDP@m>817~i!|cartB#rhL48oepqNu5f^kWqgE24a+q4k!D06<3FKC*u~L zS&~9@>0|~vP0Ej=cb7U%>gs^VF=J78N1s}0w|3_D4+cwvnX)fI%U%0bwEQ0=88{#Q zL6Y-k8(rmaV$Ls{j1Q9Vi7+mi68p6_r{)>uz5 z(Y7QE!NSfcJ8k2t<~p6!Dd-J_$xvlvbh|AH!w)=+KTCp%4>_Hr2lihjVHlC_a#Au{ zO{G}rx&$JQs0&>ZQMNiIT5m~VUMr%CT@q0b;s+vXIm#iTS_yp{Nm0&FaLJRhS4)bRzaA1ger!pnw|AK$3fBOA1T&oswEi`s{2;1KT^v zarw@c6lxEN>hfAsy_jMIDmswYnw~*Ln>(k{tzRosOp&w|5uKb!)i?}zUT6RY@Eil)us0a8{9pNls==6h1`+1xL*XFN4Tz^h?VXer^oGf}VSK7?X8-ioK}7gMMM5q*fkk$rVMJ=ziNkW1xDsatiaf-|L5Wo)UeER>3E zzN5Y2E37v(yozf>*6oZo=p4hy3KPvSFjzjA<)lvf#f*$OyuX4vg*~H(mX(aeTHG81 zmZDC+P(OLii;$sPk|IgvPLt;Ts4A+F6n5hei9g^psYfR#B`tKCR3%tV1%qsWLDo7= zx~~(f*6d5CX~|u0$?qnYN!vT=P9~j2qh`r2(;~{>rDQPXx;wi}YS!6F?(Qy=GCMno zn)Pv+7DXUJS?Ds!-&wOgQ`nZcOp7Rcx5f4$mq{;z?PD&J2Al=9i(IBf(K|x9&Slcg zXKA)QgzW~GX%W>&h<|mNRDG6{+;QzqvS)*Bi}t2PQT3B0Kz4hRZp+qe`!QQb&-Nz0 zmF*;3&5J4I=-1w)7CA6wOdsE1b47pNlwz^2*`upCoro^sd-5*9u8V>kCuJ34@5v?T zji1g@T__Dk)hFgc)WPjdt9;;R&BtBR-lS`CwWwEso*ZTEO?o`nNj6%ZLXL{|ChY)^ zSW%-ka)_Xsan)X30R$akWQB-EV28)~U9_MhAn5B|L{Q!f-K?vVGKy)`!jO#$yE@6% zdpUO%qaoFIKzeDkJ!GT)-Oy14fE}q8;Wi*A`L6aRRV_~;R2oDpeXG4m9lJXzb6xwU zrVZU4!=BNjy@g8ppgN?hWw-B()?u=0l#`8D!ZupkT^nq&YJ9PlX~bC7VDYSPz#GeS zVWR2G`x!@7YH;dl)YaI<~A}+8ree&bH5VCk(^=DR4(?P4AUY{Jzy0( zJHw<`d*J0I!=x{J;N>O5q;q>Z$yN)GUy)(bjXlxXqTW2IZ;GU_?h~*sYM<59sVZCr z`dB+)w}Pxt>gux(YcarQwZS(PuIAOqJw3H1t9d&FMk+bj|A)?B%?TBF0;HO$jb2Kj z4*nsyyg7$HUig|D8yoMa2U#35A6ra9OP4X(wSgGz+zLn@V0pK+-+XX7bx zGCqX+l8U6Y&=sF9+`uR-IMxd411x1>kx{KMo_>y#lJOP&>=o)@#&qBZ4Z}S^mkmp! z(ZF9jER8A!&(EyVVQKXIIZoow=I5U4q~wZWX_UG$g|M6VIy{~6x&o2%Rw|J;YmpY5 z3yosB5ZT!hz28Dxz4lk7vcc9ri~~*RH4d}td(L$d0`ahcS;ni**WOTA9LGJvdySS| z;3Qkc$`mwK^_}eO-Z0vY875VWgg8F$o}6LQgx*ea%*rrn8-C2oFzFBccsj$Ru6>-8 z{CtKeV|`*eTGSSLd2ifPwimR<~~kxx9MO~i@vDc z-NB>^q2X_C5A0ylO?{o@xVD2y>-stsn;Iec9wr1E>L+Zv_d{LOtgwS=*{RO5m;2rh zCKdP7HhCxxof1*ASshG^D)o15U3X0flU9QB+72c?bD^_Q(Y}NCXcngv zQB|CgRf>H!RdM%>wb7mnb@f%M`ix?l&%bW(?<8CODvtFhIp@j#Fbb?|e_2^K9gpx& zGfdjj->J;}8-BpdIRl)eT#+ZlPE9;~fDn(v4-h{K;@FsrQb+natx3gS(Hbq*8ZBmx zwhz#~-C|yM`}!g$*(zUVJ(jQ@Zw-WK2rYU=>5&Q15Lzt;ImvxrhDlX}B*r%32V(39 zejvvBT`Vz{C#4Z%Y=sFIi(=LI0mZ(*7y-a_RS5N&x-Rn)sjK=Dsq1_EKwZg$oy2tw z8!UCzOKGdFyFr4$uE7t~b!af^inSv8%h=^XZNF-2g+?V!cV!$uR9Es^;{`*U)X7W6 z$~Zh7<-Ka9TC22WzslBJ#n!xS2&}mj3xlsJYhC~iu`swCs91qY%fcaFAy$Za@0BG) zwh=ANt~egC?&77=%@BsaLb@xCUx+h!v-)nXKW?a#lCZ8@E81+46wzLcn_;n>+kPmF z6Vt1~HhQI6S6|;ijCv(TJ&Q4LsIp0Yv`y-jO)}8w?d5>`097?nQ|!92uBkdMTJhgh zjlXe?<1SS-{uW*1Z&5Y=&NcpiY2(g7_1IcFT?Zvb*EKEPP(!?FyzsgxPj%MQ;;{yQ z=4JeH9VSCzB?hL+GAw&wI&ZI=ahZ5vw`fA~sup+v)0OOjmo7t`Z0KN8rNqfSh?7Gd zOxkjplicJsX~gB4rbcM)(=-)8#hO(q8EaUrr2PaXhI%s1(#pP3;qphzdTNj-);$=$V?Dd*Y?>AIkuW6e9YMT8PO_k8Ru4&F6A$%r}I7)}t zx3qA)6;1hPL`T)Scmp$TWnQ9n@t8PUG$LxwZCyO(U-peqUDCSvs0YJ3)lS&mjTT)S zWgq~ajtU4k{OPCwh+d^-mW?tH0#d?O`v&&PggwEq&p(>ym*^)B&j)UB;tc%H3?$s> z#F6js;xPkq>5X_Td6OSvKQQ{a8}a&=TBG-R(h+VszjJmzvi&(Uh4%yu%tv8lmC)S@ zAC9%r>&Qv{AG5oMX>Rnu_(ZmRPUjVje64(|fJ-6qvV<+#0WlO8Q_lDh=G7=DzwO&U<>B**P;ljau6x|IEG z#by{n5Q~4G6hgElRI%nH6<)QHqL1vt)!pSbDNjlw$GyaD(!oL}xnFRbbcxqV?pNF< z&G$OVvBquEFJ31R^V}%BK9>|cdx2-thi=o#t&73B>)j@G8s!uwl&Ss?TK}o8p%0@X>QorlIkVLONl7j3bO}2^2uw<2Wsv^sl z6l6OVQ;6s@wQdRv_ZIpiupuug4L(g3xL#6J zl!V10EGZTFoY6~4PjkKJXPQ(ceEx(;1(_y&?{ku)IMbxuo1EmBkZIBtH^DMkwfa0w zRk2%ARAJLql!G4YRVzUs`!L7&QuRo>pAB?W%+!#iCc(%WOH{)fqYm8^)z{W=*YL?0 z)z{V}MD?{bvU4g-PbOmKXnnDhGDembjvwsx(=)u|B%#Df83ovKLrXd1Qe;?DzCZC+ zt7nlH_Kc>VxI?K07&YOmh=%(y78oPYd;X|L`bSawYS15l)QVqUZFTRaa8H{6;dY~Z za4gj&oN%*^T&3L0uEp*es!06){1h65tQKoClv(OKs5B!Srd6z6G4Qt3pLpTR~swHPAa=6{B zX{RKvlN3^2hUh?(UBJSld1>^4rpQhYpg}O1@V{1b2r&v>Y3e#9UInUVVQM4?#7daO zb|p-;q!0#J7G^iIf)@2su!LEGdEWbq`5nk%ffqXztJfq2tNXN6eQ$=qU{xgrORa+U zI#t0cNl`Fg7krKjF14FfD+MRT5oLaY{ot8j<1g|4#l@M5$y!OlWHPFVZDDdfT-3^< zZehK%6bseT>ae2@Yl(Hdoy!bHwY9oLEF@hQ)Pv?a-qs~Eg>|w~&P!h>q7hCG%JR|I zj5srd#z)_Ak%vR7d0>3=#%V!pGQ;pCJU(zEy>Z%ET~gElcT}-wG`(@!X)og@9IQdk zl}=%|q|o+jm<%i2eZsOHxmd8DONtP;)*~yzPPM0DA%5Jot0k>PIjl!k3;l9QQO;VA z+!cl=v2vffK7~=%3ipX+c#KojMUh@C9r*mC&%&Ac&JE_dEpt;RZFcn0hLIH@;z{V7 zGACu-EI?5iUbo0lOHUNfBcxF0K;g}y(%> zq>OQ%P+w26$6w$rq&}F5*d@7n7=Gdl6y8GW3q*c778};1r}K?2!ih;LL!CRzu^{2) zqv`{Y(PKQ;Gks%A!*mVcNq~!d{xDth_PfpLYk(!bLJ9(T>I5fcc#4BwPvQ9JDubx< zQXq@Hqr!L)aAngcU?2%0z@Y>jKAr%tCQr*Wi9Sdnm=Dvz?p2v49hl%G_v%cO&b!4) z?)NiIdJI1{WSVs37AHBr$TTVERwp^W&ot?QTfxZkoYw#72hp_8diRL--E>1cFPoi5 z+Ij3dLwq;gXyY;I?OUA$=S}CC_wf^*{9!W%%>|^)C^>H;hy6F*$nco-%|s`4@_WZA zB109UPEA<|#h9@>=Pbm{F%c?AUlDZM0T7IIkmXAS2_lUCi{KzUUR3O>i|;wxA|1$ZiA z74#Gj@`t=Z#5c}NA6-qYUg`zWYOelAYOXeZA8FFV6-v$3EZ1*K zRn~USDn(YLJ1QtDEe(0| z&^C{*vR0M)M5{bCPKR|ahjpBUbr%lncnRyNaX%NB^vxYg<*9LKpaUi=&nn0Vi2XP< z?vDbK?w#zUPRi4UqM%OPSr}8}&|P5q{vPJDX)?Q19XL7y;3leM%#~A|_}1Nn3~Zg^ zqz>u~=>42Q$@Q5gReY%WhEp*@c6Bt#In_z-vpSkIeX5f(dvwGBu6`X&S~3+65EQAF zB5ib$1G>o7)11V_4@}d<^+J4_Cf*7f@IP<55QnEncQEP@?a|$!&7apV+6^ySZpdLb zM4W=d`*Gri7vsiyO?qRxa)UZB6}nToK|w9d096zX#~@OW0d z7&pah(!F=<$W-Mid1yzIvOiKr<zu_h2cYqe;u~V^&9#4&lfA zjwbcF7dAujd@0@yHbe1y@0H@G-Y3O9_<`a}@dL#V-KUFJOL6`%9E$h3Uy85C57gfJ z0Vi?seEdN1r4K|qVPw$j8I$J;hrL06=;*%CKB&7e+86czs28X=;zpZv=mF)6dTy4( zE0r%4G!78ky52|~ZPKHaktgb2rhmN>`wVJRFz?A93#J2ngtD&ENgaY7e_?3}?Kjqc z*PO@}0GY3_oDOl(_y?V2t3zICxX2r%UpTkdgRlZ6UpA~YWq+)kcmRui#lu?D^*|<1 z7}lEd1@#LSSFJ~5k=GMAcCV@JR68KrUKu)Aj}^KZIYlPj{-ClfePei}P1ANZ++^d8 zZQHgcwr$(a#gulQ^HhcW<3#HpU*z8lQdiAH zN5M442=-g9N1s9(SHNoumfVGcj+7l_oZ?=A%vWW)FJf(kCpS4b%FQgVvPy-MbCmOq z&`enz^FA6zPo3);V@jzqg`GlkNL6#li1(jX^L@JWr!Yo#GV-0&v6+8_H|Qs{ghlb} zmg)KO>i8aWxqC?>)1TsE11^XH`m{3=t#fk9BnA0HYPuq&c#rKWb06>TTMP2>LmlL5 z<#B!msw44SfZdd$uT%iIk7*@=Sso6^KiVVoT)@oKus$-;zcS9SUU4xHO!jy zPXaw0b#$!?;4DM+BBH}ue=bisAlo6qM#7DgLa13q>VFKL22s_E%42|tCV}T9D;l;J zQLG6)+}zgP`l;7BkwT!Rwbeu18Q#i)`bF#F2#4tNR@fX)ry`{&R2W4Q%iUOW*)Yo1 z-6X6#xEsB|f6k7)WctkCP zOF{ktP?%?0H<@0cKwKuRg8sj6H0R%%%#Tj)0O1^#7i65TQq_sLBC zFlBSVGc@GNa(JyXqrka+&p>0_KU56y93#!$*5u#20XJb8xgLtQ(4aDxV)I6^PJ?Bh zjWHatGl7RXp&Cy0i^L|+gOl0cM2TOeCcNR&O0hszF{z1$TkS!Qc(pm<4~#U2?+&i6d`ws2~sLCP2?ip22(eGbVU1ASsJYpVT4AC0}ojWCe6n^i4DP9-iW@YdWj~6 z0A+YP5v4c)9ZDn=T>l)`q}+qkWyhV*rFJ++bZI+=)~aFu`y#gF3hYagp%fZxM3m z&yOuu*l08wW07_tR%g)!Xn2UR$cT_pU}Q#RZM3iyVV+>IYC2sans5YQhxJ5a5}2L` z|7b#908++3B@`5yWF+FDFIb~oq5%XU$gDFqN1hf03VWl08d_A93I^(a;&|!dn-+?N zM$SsDk;l|UsiIr)Tp)ECc14rxHJ0}iKSv?r)jv-uh;<_~H7Y!?kNn-BsdnyJRHCo| zE7miDWUwq|@aO$o)+!ke^t0xfEVGaKMeE!oVW=1k3WsMF#FkFe8I47fX6-eXw(jQR z;ckY<7gt)$GmrF2z!0?F>0nu~Z$;Br|M zOb*9Yb+T(p8Vp?40)R57E9kHzeO7}uv;6ukD@o^&E}~VS9jpZF$GbwMF%y$A*c*recDe^svJImq}AL`QMAdzUfFmmQPzd0=t@b}h3W4< z7{{SJIxTn%yR|OG2)3`zDG5{z`=jO|3E!IiClUEwe@H-qp}Hc?kp>jDsXq=m zrLHFU9yWGO{v$&+T6;UT7onmQ<~P0_uj?Fba)`?)%6Cyq+PkSo&ql zitHm+`I30(!Z)E`Q&m(c7O2d;2bk5N)_tY!7Y`nA;*f=g)3TN!9_;G#`{`9F+E%vp z3b|&+W2gdz=i!=;ypp7Qr^I*}PAW>QAI_1{B^OvAG@P1nD@QTrWzfRgc|+_s!(=RN z#gk(A=3gMlDIAoNfya|WnlX6Sn9*8NU3mNkLX+lEc!z3kWnj1~7&bZ^Bj9(@&~&sv zy^38J076Cy`6fd{QBu}XozAntQI^XXOFMe118qseZDmp_gl*M~!96|UcjYpBOQ|Aq zwowV414m!9w{~l36wi+!qhAYYG>+LEKQ&p&FcwR-ltrttf1ggBQ}|gLIbX8b#WizM zX(+2$M28chCk9MlGJ^9Iy9D#8O_0>@Eh9AQ`hW%)=4O~06s+bM75El^)wS~Q(CpIQ zM&F|{7+AkD8tBflFaK7xjgqs#*dq2Rq<`|FK?;bkTEGv6*n^}`xGHGsN)@Dk=WO#Klv zcW76RAyY#1BWMA;FfE1Slxs)tqTX(@94mMPFQ2VE=m9m{YT7PkPh72q7|QofFeRM( zr*YFOw*icI0FfQy)W01`Fr7cXa(5c*l)wu;p^s-~K0`(fejcuR7QZg2@y>i*m#wvrW4;)=M=vNz@Zd{Bd3y|>7Vzlcydust9g;`xTTb)(RZbiIDS@hQ&l?-myTU`7c99+9t?S5aAI0q*6EC-ErZmjv@aqzx%d7k>r2^E~{#N_O zE47z-A{3rexJ8d?YD0-Y;~weTRE7Tw(5e{O>dO24&cBxeCQokKJ7<@}onH^TxCw8a zXt}r@3_E-4T1eg*obtH%pu{Ya1BhGfq(Q{T@CMGIHJ{VW88eUaerzI1wh*0DaClvE)F!c8ts?7XOnx)A_zp z)=`gft;i{Nh-ok&hO>BKqN=-mLI{=Oi2u}YL#55d2Y zW-EW?ByZ;jqSrsX*84O7a{1jhk1^37gS#ywS^VZlnZ(c4ZP23dWcS%e5@*x1<*V6< zv%m|rRikx^w3KF3q)v@RF8_^Ih^qD}g{5wO6S&DWC%&fKrfe=DlH#XZBikso!WRim z0D0-T-aN?L>C+77U1p5hp;*?00@s{M-jo-by!Lj=sqvLpuvFP*o8YY0Av55;`vDrh z1#I&AlfMf3EWEBI2_?_@51q$T5_Cv|i`VzzL;B(t!Z(kEWlR4N5?{pAdlp&fj%?-NO1$xoEib5a&IN#{ZM7huKWlDyQ~Bja-$nBPm1 zyY$VTAm6xK_F4&IV2RCMgWNQ zL@|6AhKd$v6j%8%Z)d`lWQg1{fvm&h;KUKKrQ1Sr@GY@u*|L5;=OizQqNNKlD_Kch zO{dN<-YG)l$alPA*LIPzr8TwOW8Iyn&Rhp@L57Z{6Y^{b{v$x6t8AlszZ1jA1n;nTmfj!LkGeH6lP{nOA2`>ctCE8L|RK`Qyp{ zG=Oo29Ym|J;w%br4~7Z=Z(R4qWM(})eR)dW5}71K<5v_RZOgJRb`TweF<41 z>2+I$^DJu0C=WIqZ!r%EcKKklbHZ#5MD(>|vD?YJtdbBytq=;G|d*YnB!S zX4UbKj_=&_0>?dpCK4hCRzxD_)shrHoqA%1D`t!e93tiY=qUF|*su_OUe^`n0q+8w zd6Gb&&6OFl0vv4;K&c%@tD9AwFA;F(G}(KItDh3wiksJa!z`*Vc9NIZ?PxWrS^0DB zfTR$2JnnxR`+T)YiNWNOnm;=m)SUL%MpRa@oC~%WlMd#Vv0?8CC^`9dRhSWe`UJXi zf2~=A%2DYX{?R1oTg`eTvE4==VU0!HbDYEkkE~+$Luj0-%jXO<=3D7Qk0B73=$89Z z65|+RnPkTjca6F^h;n8zq4Ho4)(f2LyKTDQW&nbdh5Uzh!zUk+1*U=^rem z{es|@Ga$r_+$XB~5^aoqSG3I)vKdeKIZ;0!iQ{Dd+k!D|wEn&4JuSG(*c*%XHn^j$ zHB~U_O`AKhfIqq=1Yha9t|o1n)7m^n_Z|ppa&nQ((xr~a_$pRP3@LON^pce{6#WwB zAZiQa;w^{Z7c+WBX)>t?$AUQH#R2|Y%N1Q#qY;^S5{-5t=rXq$=zwTdLAo61(01%r z|KBIal0-`dz2c(Okg)WuNp>RFt2vopRoLkAKPf@}@l$qxy26qD+bPomG%XmFrB2c# zqzxf5+(*7)QH}s9V$`{%gaP-`-84AaU0ZKG+1(c8O922QWQ*+XcAKv3uIw;ER-aDO zcDXGod+;_L<@nS{8f0rkZ!4ZKS5{y3KPKQIF!bh@01~NQb1P}FZ}Fe7;tm@-DXudx za?X+~8$kOjqn({p$0t~6#|7~X^-viFH1T_8#%~pgDJ7C%a3g;eHET5~SH%D(u*DN2 zZTasp1d&fxAs2N;hWf>u$NVxsOkaV|+t7k9&-~hCU^3U!iFesjEn+$Clim+Wl1cP2 zRR9dM%vBz}{6cc4rlNXXB&n!ON{}dic+kr}6p8mWar~`ner+7ED>VO`gp@ii^`2l} zSVAgLIhTEc6w8C{l*N)|K+McbOOeyA_`8J_l6AVA=&_~5NsqT^PerT-A$&)2c3ZD|+T=q(3 zMQYsiMKg!9t`A?oO@a$$(%wr=v26i+v@{)=E2MrA%_o6rPyip&r*8kA4YiK4nT18fTFyFH)p8sWdmPQrq+g z4AOQj8WaCaz&v^lJ!Up&?w7;6-ZXV*OkXEo;sE746mrbx;i1x+?Q4J-c6=ra1oN+M z1cu|Y?Ip#M^Ahu7c&cLf2E`iMb!e!ATT}ZvwCcrvNJ(77V%rTO$tSe25sTuSP0-?P zuhAj_mhm);`4z3KLy24q#mwsObp;>Z_l=npCRU?@WKE12;ppHy_Y#Dy3SfcXK^SJM zONF|bAt(2YC%k|zKC>OQabl-v*Z*_D8+}FcaFNK2gcmJaWJdUI zAf2q()+vFQXQ7C0<15(}L#~iMV#~$mOid5D`eZ|Lrd4C6v&DjUCuRig(qyA4$1XC5 z?>)k0y=Mw=4rc0WO`M0Yf74yc`Sb?Mi`cs2wm;|a`Ier@6;%wUeXRF4S2g5hk!Pp} zTDM1XcAl`4MKg6vGIgse^1!Kj;`+5Ac&hyS1Xx!fmG(Sr#Qad5JAq2^?{&G=2P7Gp zzaF{Ndyhox*>!Uq?d{JVA>H~#lP@IGV8T1WfcgBJU8b6<>cF*Nn!n z*-&^ftm{W6^`q7I*HkS)Y9;CCnit-=0PFth|ClUzua~0}nIX&5QC-n+bh_1$ljRNt8PfucD2&>hD`sE`TLg3Z+k`aa!T?N-e7$ILox#x?%W>m ztSe&pU}gBgW|o2RyCQkvW|ql#Xij9Zn%nxo$I8!Hnq7Z{7sdY8Dqd=>TB4r)vonSA z0inNqu0{HfU9-}Twz@|tt%x2NN~K=+rlV}af~aOPO##kZm=6>NVp`db7wYQm5j_ zJa7G$93xjB``2>u)B8_x?RJgN6S!2Q9YsoTSMZ}UQAY%v=Bp%9kL0iKNVJ}f)^AC7 zmO}Hrr;oA}UU|IgLB--h#pS(oyXL}_`%m>&5Yt#1lWaCHy2qINB-GHQXCcCmC#~vB zMTKSYnTDJb-%`54bM0N|(a(3>ChSy7O|7YClyiBm5LK$Q1p7Ik?ldloDrEHfE-Lto zKZEmSzMUNqJAo@Fz~ro??`)+lNB&2z_9MOzF%TnSSIslu7wP|l@D2>mVckOHPGeG&P<)Bo|dMpnH6d>$qmW!^1XZ>vyqu2^RlbF(Dx>G+`^sQo~lET5OH7aU!; zMMygsS7e-|-+u%r9H%8UjsV8UXBa7;{-JDtI8eUoGL01Nd1WPWTE|eLgDwLWXR9NS z*dm;V9iV#X7~~wm5@zDSNh#~T)pR-zHMtg5DV`-$vDH(t^CnRF zsda1xoZR=#?xc!O9SfI_szW+p(HDY9gPV|amL;xRfqrxc%S5h`=D^}sez4`|wtH)%a}n&V))jz8VVEhl-63q#F)#&+Uz>O{=4t%tWq!1y#8N{coh~7CFiu ze`-GTIC72*aIb?NKa((ZL9SS;2lPXA!`Vf12Q=&uD-vEK>1^}DBJO8{4P;u@y2IQ1 z-pVbduxh))!FZ5a)Z;Boox93$_*yKW+|#1ZwowUQ*;ADs9ZQe~Z$dofg~~28W+aV5 z5JKscn^e=d_rP<~o%w2q6<=}(W_z4CQ~T1>E_Fq4&mzoO7Sc2wMg4CVW(*Z=G^+Cv zHwuejubtZ0%F&$ylpVwW^q_~=hYTw3Whjwtn&X68a>7K|yIDnK%@S^OcV#<4Jht}> ziGPfFcx3*89zB*OFslUw&1QeEdKjLG^*>BuJ~9JU@#mOf8|DPZ3hTY5jzf5nc{`f zA*Y;elfLBe1Z{4sR{b=32e}H!fKsL%){dede=DxSxiNMop#qtxhrAMzwa`#+S5$!G zD1cHvI){^P5t9D9(C1yNLasbRy&j)dAE_Et7BC3Q-M^YK_HVz)om$kDajRmA&mgZ9|vn7b5W!VZ}D! z9ObtvGA!2=-r7w;*~~o}N(j#)myvGwF zcwJfNRWo9*QTBeROt`=P-QIPz#>QbS7BI;UO{mHFPIv`R_u6)feWXE4m} zMifHo+1Ki-gAXQa$e=lOFNb;(;#u~JB9%sce{X>zQRr&@8dkeF=Za{gc40A6Hp5_S zx{mQMA&&HcNOd1Eb3M*fuSPCMI8VS>gJ*02H6acUq3PZdi}Z6+ROOtDyGva+-?8J` z4Ylt(Xm+OfjwqMHy-`dyKW2raYTse7m%93d9I!gPu_SN)Q}R_~)t>v$Lo!EUF67>| zVkHa;@uZS!eb%a%drU`ZE~Mcbc)Q0&6dU-yY(}*)Kf#xk+}4~6p>ivcwSnlmr6nVq@|?M9!z00cSMT)4b9a!PE0u}c?gYxsNwMOM z4&`D-*t4P1p_)gzNIL(nouiK~mCCi`-qx~D-E^4SE2;%bTD}N^6wJsUY|>VB1H~=6 zAwTjSN{}y9AvWM5^+fI-&?_d3MB?fyr89PoN>+TWy6I|V%PRERUbLG6O44R=h$SfI zAob_)DZx)bQ38HCY10m(2chD89n3W@3r{g3BcY&XS)hO}S{l*VTE}*5CIQ7Iz{%SB z*36?GZu&P6H=(+%ZbsFma^;aUtt#kWDoI{z?fV2kLy!sOkvuOhB1 z#)oUbE6*x&SDD}a!@GxY7| z9IG01bAoAW950HK+W@5?DKr0Cp6jo{Oe=djlH^X*19lqP%(ieD>89 zba(|Xo-$Xldmzda;l7pH!hv?u$CH!=%4&59zDLMmZ0h`}kUp7ivY1Mz5=DI2JmU_> zzX}#C+!LaauF83HLSG&q6|Cy$!QjiT+&tOW-8v9E)>YIK#edK^cLbIgiAL?A_&|O* zN(?*90#Dl}!IE9mpFAHW7^erd;cjPqzA&-tRPp`M+Wn;@=F~C@GZj(y+Ma*#ExUhf zCo$>@L->!EM<^_SM4{^o|B2)RKcgF#4~_duW=y(1>JrX-p0|eNBJZFu!kIGANAYjK z@eIoMfIXTwg0%rv&JljV;Ht+Anwtd?5IjmNuUfGgg%&+A&e>I*DJlQJT>VReZh|t) z3u!o*eON$pc~X{Ou1Qa6cuvPtPwHqJYT4zimT7y*nC)d5qMw$7^_&#aFL>pPH?T{6U%WO@Ce{F3Mi zGS#o9atqP`tcTs&DkmYBgiFH)5#C z{86MSYylaLm1nh(#+98UVCXo)oi$;=5Ce~;A&pF%CTOHiJvE2ue^NUj_czOFUCs8z z!3^qe=xwEJf(}^$2KC&8CtlU48g})oA`+mH+0@2w{|U=s1q#EV018VZ{r?jz9D0-CFY}Yan8Qbt z!dy&Ig*Ho#8!&9$m9us~JdO~jsd+SJdD()7jCBDkuoj@^R^wQafH5%|>E^+h`|tTt zAGb6>#3K@`E~@aknH#?bv{m9KeSQ|SEQ(1~kOt@YjIBy< zcC?3JgdkyC85{u{4KIafIxVyLUIX~l zGMLps!?_^M$PAkss;<^fLel?wbl=~LXY&N(bl34DLMp9=Z+29eH@cGMBB-~MG+`JA zgzx1sLkxFI<(`lkQU z)l7z(wNJzv>=Cky&Rh zV`+x>AWUM#a|L3wc-VIPKvBB|ogilB>zU)j1UW$_$53^T3QqE#>i z?hF6gq7l1Rhxzx|T?_(}uJ5?YPT*%r85QNS`>$!GanlcG3R zOk0026BedJ9n_>8j!>T|{?~L#`wuoSOd%UFOR`eM5dKUV`vFyP1zhrnG3gvvY8lQ# zn6(c%x(12>uY%Mvo|ydB)G}PX*^xMQK#n#_|P1pECOR5&l=pYO6wS7mU)0U-^~fP8{!*41pp}QXn?ecdmqkuEx}3|%H{0l!U8{Tq9aNsP zOrM4np0m()?t~NRY;TA*&3S0?hO?l5UWS)Ah`*S4F>c~qFRQxRmwRY<$KH0M=-EcR ze-Ye_ddBuL7u>#jmiD#u`Ce`e4fOGLL3?OHp|72SkU1@pjQM0;Id+bs3ADuG?c(%i zHlGYDISIT4@ZZJ^H}+i>7_;Oo297tW2707MlIr#OCYY||Jzr&XXF-<-}bQwsyNt8x6iC&o%6gX=CW3_<9 z0hhfnCz?GzwrjX((e;H&EFMT2$`>4*^7HwqzCgc)R5A&>jf-}iL?DIY7l=kd*S{H& z@HAujQ$>D+23kr>Yyfcxq2nMOce~IJ&8+g=CEMovx!FfJCa0t@I8DWx!fG1K9nBOsCx?IF`iUuR83L9d&oRyPmy3mNJUr5C8 z(?oS_8Cd--)3S|`XnIzUsxywqk!yI4Z4So>m;7w|!vgq;xCJ?+w`uO=-~4q&i*o{v z76M%QC(cBxTf4V69nFUN57ucAr^IhC%^Omtz2AliF0T88m^Q_@W^#kVcouAo5#*0K z7px!Myd}X7-X;#;1{f+ zM%T^wFWmh-+X+K42pK77+}*jq3#NNPOF`{po57w0|E$>(r;cfeCS3neP%I-XS15HC zK;qsRyPz$)g50@rvQ~?}_ZS%G`IM9#k-e2`EAUcdqGPhR5rZ$c65#2d;}kjBF(dw% zBpt|i`M50bU7v6=C%C^nYG?V$|5c*sHX=~>SkkpQIc?cFvIeRBHvZV_bfi&e-|lYa zCLRu@vpse6tw8b0TL_t&5<~e5hz&*a>^hj8HMITbmcXwqDGGYhY#fYDlKydDv7xC` z*}tic;^@Z3a`^t|RXNP6w_@rHz6iqz06l+0v)D{tR;Qlguxw4T;YBIOesz6L43nwpVv|ky9jQfcn zU4?9TTVD6tW(I8FCdj~0g@Prs5tsk3vD>B`@?GJ+!N-PD4<#$AB5wtr@~g9mVtYI6 z^ojk4p@(A{d(}I^)!&bA%E0oMzmkF78Ww%oz0Rr+TghFAYi~5Xacb&O=1!w!heyEM zTyn7(=JO_I?PY_n+MW@F;}nGu9~V=Nz&(X44et~Xiu>rODJg%ULFWVW(SSYGjrTsK$7ZEmS)pD4RAHz3ir=mTF1gx(5~2-a<`sGWHFj z=onp8cBhe2x~5UCOx{i+h;V7t_hs_{J^CAwTGE5_$b(aL5a1)_Ax^2 z?lW6k|L#Udc)#5BwW0%eUFWSH>0?LM5I$rt?%~fZ)YkwkzjNEAGAVC59Qgz84ly-^ zyS`E!J|mn z{{hJFC!Ynu%gZSctk-8vF8*;jS4hsHk6`|Rk`L7iw+o5z>J{a@o3c19aEWYWpfg7b6TElAtHZ}UJ zDqyz6(;?O)M>E1NbswnxKHUu)~>FqMMX8b*ih9gGE9%0 za&b5u9==;94r?$L4P8Pm$0wldvwil$5pEF=(y`2k-IriBw;ROLDY{o`OnDBJ6>s?U zw}Nm&d3;7SOQ-4|SE+~waet9DHed~-NhA1{mSWPUO6pXKbaYhGrh6mLbyUmKl9G%?QWY7p{=z4kQ!VQ(-*LBe!Fhv#2R6A zy62W`9!VNZU*D-6oyLiE3?rFn9?Rz5I6t})gyAh43_W*OQ9qHzhkQ2?B(}h245XMh zz)#C0T~uyXgvoNC&9)aSl)R*lKHuz76pco_5Morx6AhOCNoxtvtE7|pszwplDHUxJ z(9tWzc&}#rEUJkXs`|yMuu}Y4w2-lk;DRfM14oT*F(h~6GOl5H*dZ`92O3kj@~?ck z&i~*PBrh`@%^oM)BR_^%bi$#D0a^T=d?X&y8Aqcof*6h{=8)aX;?+yCNeE!j9)mQ| zuTM;m5~xGVlEF5te>=AL%Ygh?n`E(75@y;=GmBFw5~3J@#gv36M@$`z;oU&P2lR8% zC2@R;MXM1Fb;=0MVhWc(hgyZ-wV15#Arx_3TrAaBP*4Ijbg$@5(<%RU$p3b^5++p(M)GS;VmOmh!fGR1E|R!orf$u8G$HYDZW1TKDvu z9{Hm4|12-Y|3ZY34YWL1F$%{^jJ=pRQisGJi}0F)(*Lu&(mPi0?k*dc!pa?oyZV}h zL!GPj5Y!yJuO8>CkLztQSaRTw8I&BDGxOZP%iR?yODJkh*W357z$ZD^I>@F#T6IqK zz?I|kT=#uDqtSFar0*pw`m7h;yfc`hf2}T?) ztN+wyN}8P;#TA0(8w474w-G8l&<>BJz!vF$R&~8KMENtRS3fy&u=`7DLJZA|wszUK z%(^VYF8kgV*Q@nuXuQ+KA#D-EvHrPqKrytkwxo6_S8Y&kCCIjgsiuASJ)BSX^!3Or z^zXtCfI%YYDmXp=`IK2}9t$*$KYOlp*IQqUd-(h)_#B8g3@sneH~W#HM=}=;EUS%h z2+#M+`Ou}0{jzd(0f!g1<8~=0IdpW1%^#;@QnZsCAzpdKQElU}hsGdiaDs*M3{ zOp3a@B&9f3GdxD>l$-~gP|P{33Cus}>(lO}aW08+9(mjtq#xZ544XlRBT3a>ecIII zrYx*@sri@R_k}4)y=A+NN4D5t;8dB{N!4TBjp|fW9GG%-{#+R(A(d>%?{*P7CfK*7 zG0;$Aqz({|O|Yjc`>FdJfkytk#?6$5LFmV;mm#AaiXF_9emDJTDjz*XW`ZA1G`(%D zxG#_1?q$nHiNNn^$97hqPXKx1O_&qJu514V&a#{3T^^a{bR*?p_%mw;wHrzbhOnK$ zk`jd1@{Qg%FNK)mH8vXeJ|q-18PeoUAd}-J(Ok*sjX*@%qR(~BV|>-}Oqg*w;j2Nj#hFQys&(itPcmI#6e*tRW~bTkozp_>PH zg@6Po8@6)AsOiwcakk9E&Pmaj@Jty^Z%pgx@v|RmQ3`N}|9aVsD0KEN?v&n= z$px(7AT<*QM>6_5mVDWBhvOQ z-1`SUjq3fBOgC^hY*n}28xWw~f!0R|*)i(^BHIaLtaIiGa>6tlEe6sSA8_pq=Fg3B zD)k~}>OXhbjM)3K@2Px+8Pj^(hk1Q0Iy z_5Km4i^=0Zc=PgXS@33m!JNysg1jpW@}8;+grY7c+JYsy(ypvT+Cn+=3F|aiB>b3l zqE)^~kaA;?z&z6Fm__;G70vIMSB4BY7h8CHS%6u}zBWcI!*gc;)-9Fi1s=QwoQuBP z-Q>2N*Q`172bQl0*3@t+kLjaB9M1>B>X!G>Jt77kpspsHGI6K{Y?-qK+ELYUPZ``| zLpwbA8JD(I!qQ0U=_tYT5&x8_{pi3X;&B`meDPztw!jnfb_rKFV>pFy|wzvFF2kl zf_aW-oL@RM@;YT1f_WYDzdrueBgwaTv3th%v!kqL>l)r#K|B{F!;)?G{p~hK4aCK` zjF+ipaIXh-I#q#`LP?XVWmf-hDv*KnRCUx#3~hjbty4c(JOzC>&8a*OV%YgY@@j)P zrg5*w0=L%}h^X^)nHwgc33;7PdB1oS^Bpa40|i|2pFX@Qgj|XdFu*IYeg5{M@BGB| zaxK+_+)>j zHYFD06SIYPm^XEIwb%P^1p+!}V_{=4!z~>fI_G6W(lyxEr}{qsT*~q@PLAR|GK0aR=JS>$1-Acoe&98*{WTXf#R zT~fegfk{+)n#>I2t2*-5BM-3a>@V0RU~ZO_2FA>@;}%&6K_j@)fS z@6Nt<`1L;b_+0A5i-PDVh4f9D^-2A164M2=VQJ@u4OnL2Lc`uj<$*TZvjM(H=QEto`)PehkQ{iF6lvJ5cj4qmR+x4lmaL%Ss z!D^CYMOxq_=1O7r#T#Qh-X1ZHq!;e9A3b>O=Te^tBA-_Rt+^-!23io-7h&KT>G zrAwhcFb-|jC`=P@v zMU&s7-dU3AA^F)i1hb`ND0#%+CQIALOG{81ndNQlGZs%$31_yQx{9TN%xB|tM?kN! z1R$9Tvcwl2q{j&-Vi>Rqyx2SQ{hBWH!t5TW$Sq-bPh#f8b;cKN3r?is+-3oR%c9E| zMl7Pqn7(5X952Cy6?9yniP58z;AF`p)TWHZV>ozYksH#1*VG@*evAuq$|RLgYKbRY zhdpVpHvCwi_k!zI!?+Bo>8GZ{*#w<2`4$`7Lcwo zo>GA_ib%yYmhbQlsNXLATy_Mk#t(@;ob{<&wI5DrMj^Q48jg5NOqZ1Ud_DvAB*LwMd`9lE{<7;6w`Y!v zL)oHw41eR2W&UyK5=+>w&mb1zaE;#CK~Mr;q~MgX@J?M%7!(K7rNpogn8zHaRv{}f zf)1Rr?;(V`7uwEA2dyxQXg(u|Mb_^D2eCg1ZMlZ+v;8?Ae-HYA9?v<1j2Yz%$nt#- zFeDn(1o|3r0L`qzA}SsQb6T*0IgQ2;d9B~3*=L>6XW!~?+Yw0^Ev)0{WRc+i?9lGv zj4(2TUgX`8N~pVx!xDVZot|`3^bh0DVS!}KzY)QQE*juVy1{>qC3}w=h9;$&LgpPw-^;2GQ>#yD1JDPX~T8n=$ZjP^0gp=w6S|Ax$5TFWStQ69a{Xb_`(TE{5L$myUr8O|2aN$VM| z7SKf-815D@i#9Q&o8e8)Y}zcKeVo%xTRDEdAmz~83_;4JZ5;1?Gjwz2(RPM@0rTly zhLah(ob*1!X9RT72Mo6hm_<7oekWiyeaNuQEs*ST(=LV=37A8>8D1w~F707>n}B&# z%W&Z>s7scEK4pkO-AQ{n{+1xQ=yQf3Wzm-$KOjiiw2vW3ZaTp67Pmr~tQ>(h zeZ_EyfO&M7;XMN8Qys._~dHN(#ZbkesB6DlA%+eJqib`da(zGpa4z-;<~;avi{ z=|_f7RG^yKIrI}lJfL#vIL9{&QXc)n5TtzijpIjHh}%IG%`j@VybTJtopd`x_id2r zcF`RS@ifSy2^=3LNZE83Ly+7wiQ_j3QV!kC5TsnXm*Wo#`FS*%;cEis(|rv0GR$#O zCBx>oLzx^GO=H+uz$}{1aDaf>G=pJCKsU`|I9 zCZH~PIkcQ1o*udMBFD!IQXZ{f2vR<+;`n?PlJB6^3||${NoyE>${F)r^fE&{cCzSI zjyJsv#pP#HHA9fxw3g#(f|NtAGXyD@)^YsuyP$J^9=*vhAYeYNXE==^77aHtUM{hV zHZlG{;w;+2_$P_8X)9xWBBVPVZmMBiAaM?DW9*eUm)>D~m&AGWF5@RAqE1*Xe4jB! zKc|y+FxV`KF51bMiCOd!gCmldO}iK~(M=ySXgP`1a^}z;#%_so=@Z7oB+jEx8Bdis zpY}3-Q(%{aK4-jNVkdpc*m^h1ce!XE;~o-c(E-NA5@*vv#&=8Xrmq+;xtpu)%Av!I zp}Z@X>KME&iFx!jVyTS#q}VMo~#G? zw8R9t4q7SzPw_J3gzI7%Vv6?^Za>1TbNCcFH>{W<=Z23?Q4`53Igxx?F~6yp_bX=0 zeTun8GC!l3FSt)i9(SCqt<8tjah?Wqe6x{1_QeW7a8lVyKBGeHuIA z+|XQ2I5#xs?Fm!O3Fn697fw#5)T!M5w2@8NhmngmF&f1vi#9Ww!YG@zFj_V>_0)C5 zvWd`AX;<@Etd^@2l$NU#pwv6{TCPrj+TU~8d9*eGcTb#C$vWrIMurzNbY;_ahSxL9 za?ytjZx=9&K1x8hUL;^P?P9oIKsW7Xcu>F``k3Jv(@;=WF709H6flo!84eXNpFUwY zmSMJoK4mypKqr01@MQsAw3lJ6fLZi8!}$AAPrj?=f^c=n%tW0y^m{hV7=KT(^r3GrUy5EIPu_BVabwF}zbiH+{|U zaRGDa8-^PN%%yJ`ekEWY9c7p@gA2;1?-*t=%yH2746mB;r_NO5i5Pq+#z0RopGIK1 zkYlfMrDjBQt0sQ)ek)E75XZ$IlD?KA8NtTC6F>2|c*S^3W=JAZHwXOLo;bPl?)h@crLgQzkcM?tItwqa#x&eXqB~TI2zI!Zm z?o74KNaQ5@?+sL?!3I~(OeMRYp5R|~PX_P|^c2$~4f}$$S=f04{1{{VtW@gc5%zsp zOPOjRve5A)vX7!CS-|D9FaqLy0LSqpwq&I`AK*5S5_<1wPnIu44QI#@249XNv2Wec zo+^%uJ&PA%w-*0!d6`VP*i5OFDVHFH5?|fXo>G1h#l4Q5cxUZwPb*OzlL{s2RYm%U zNo{wwr%FlM0@5`*+taVJQYo!;XM1wbMn%9=BY8F`o?){k&s6aIY1216$--^^Bu};T zhy%hJ+WbT!x|DzPbSaL_cy4wop9gd)ns@)~RO)nti@I;bNiUtEWg-6Q5%oGU$wC5W zQBNjXQhvoAD(a10Rn|FwzMm=W1Re8C9hGjOgRwgHXC3>)dq&R@I`)T-kIWG|4rDdz zbsQ9{V<8hQ1wzNc&~fJ}==eMLl}Ar2&FV}I;Y}5qmQcR4Jx!LWZ^#g|+pshFiJk3f ziA+>P?DD-5jl5@PdrEx(b~DdiYWz1~exDJD`lI71F+L?x9G?U`m}d+%o@=BLFv2)~ zgc9T1oO-rmt@1 zj6Z*(n3*#x3r9w+Jl&iF@Uyoo0*070B*t5}4&?K#uwC^}ivFD=6^8i~U`mXyepZ~4 zmHOiuq|xD~L^#x+C_$5_Z-qSpAGO}|PO|B4p-e6hCUusxK2A!F% ztjz3-tX(4|p0MGwhK-S)B0q(ZziDKAz=?_G+TIGy6Eq(gChS5k|ihedgl zWoVC*``124dGdtMOCf$L=!vbYI@SD|L@)`iG)V7l)rryz z@jvaKR-N4Q@zwn0#cilgTH~k;k;RF}X5wc9Qiy&NLhxu!`)NQ5l_6syE*mP8B=HIwTN+9$eq-B(Zi3WW=)k4P>h#z_kf}7<&DBNt2a5HwUL1_K@$wqxwamWXu zk{vS3K<{eF{krJ|^9eSQT~A4_tuH~?51!`3&v%;T98J6|NV$>_~cQ4){m1-O)DvxPJPfVhk7L^QsQwtT{zE-v@M!u_0cx%;AqRA z9j!Ld7%N^pNXyquVU5pJ%yX&4b5@J+xoeo>Sc}SwX$&5>)XZYnP1A(FQ|^*>fnBF( z9!jOOe7jEeD)m0%;7WAtX4lE6N~QBJv+E@8#9Kzm{FHH{d#$W!L#wM&DGhhxr97NU zd?%iP2=2t2_i(DX6R(bvsU=bUop{?GPNno_x1De&o@mBQaZtXgrS(UpEuLZj2{d6$iT?eK?ge%8W?FGtxjBdeSeCq*4YSc)yDAt&gTshA&uF$hQK6t1dr@ zjw96@3`C7_QI$4qp)$j&IJSd*G;QbKXDnZsO7Il`!hZE#Sr=ufI~9X|=hwE5h;L=OKn z3-x$h{QSv*URM0LY8K(0iEwGFmdtZXl*i7z=5guFEf=Q>2QFEhDqOchJHd5dSu9<* z)f3WnM%RX)jO$Xed|0U}nkUKfWhHntSbxW2o@HzDRR2zndGX&L z@fe2BBwK04nZKV3iT z)Z(6`m}hQ8UQwz%s;1tv_$1%NdFQpKl(=MG!0k*#u{U>g#LXQ)K7nb7Ibn!+GRr(6 zG39lb;tGFL4hw(B9k8((dIi`o3NP<5#5u<4Ho2{Fe`i3f zS{*Nlt$qfB8K$mZ`hmDDO#hRFVn z%6{>)GW(oo8)X0JS=2n%eL|zX5%^7yU?36=mwThV2ACDDX%<`Y=f&fY==nIfjPtov zl2;ErANP#QLN}=d!v`L?#e>f&KYE_Abfo9`$plY&Jr3j6eb2*t!bY)w9KFblJCS0P z(OZlrEl(v&6|5K>O)v3-@|G9yUPMjUvKv^+!$?XTQr?CNng~bA(}>ba%G=NfOe*N2 zQ%I6}fCNYS=>_RXHBIUriLYVm`J%i>iJG3^Q6pcJ9<>M&c+~nAVOL7PPfaaNgw--+ zh=#co0|HYEWok%Nm&QZ>*{*(OFA9b(dDE78BhVLHT=Lk!1-3~$u%(bvqGKAgAQQUyd zt*K&lD)CgI<4ed9>)%FM#8WtAh%l&Wz={?^n#QYxqqK~&zI@g3D0uDEmvF^K842{g zK;7|ERT>3%{l^+i_i!DUz+-^LB^;w_T^73ZWo6V+JOJ2Bz*~l_x4#UdmI2=Yya+hN zvYzJQA^sICu=}F}@H^Z4`MP^MlI*g3RBP|wokTYv$yY>^c@UZbR6(V0@3VOXZ6T0DgDOHc)e%uZj&sS^JEC-2v!%k4aZpowNi$luE&stHa!0g*@ola zII#^y;}s4WVzwKl2uY&-cBqc9gFAw6o%DO69o!KnBEcu-zqEs0_eUggM-X_$o2ir` z<^en6FcbLnP1T4y*#F6H2p{;02b=D1aYND(mS!*thv^9O&R;M4#fWCfTuGdh!K2Nq zprBt=HanqTdPNqr*+r*^ z!I*&Z9FkljL$HLYaHp#)F%O>dApzk-BA&Qp( zrBcdSqAFsWWr2JXD8Pel;3n}VP=GgqIh&Nu1^j-7JK?)=tU*ad{*nCx>PC1ex>|;! zh}q&scyY6B!e+6EC{UNps*KrUy(IB|+~wX}c*q=d5WM{sZI-VP?{Ai`5dYkgO8g2@ zxkbK0jA))r_$@o$cGGYp{F_@+fB*K&^URUO&S6&!Hnm*#w^)O0i02RNL>u^4q5WIT z9;^d zmCgAf{x0X6UW3`-!0H_2-Lw<(Oxvq}?e)@Epb& z;({j;H=Wfi|_-uB{Ir@9T zBk>)t;AA;^g~J2Hg457ouzxTxk`m(!|A3hvmwvAYZ@ZKjKO&Zu>?yx{b&r&Ia8pa` z`@T*#yU>V)(|LQB5}KlKg#(__3ve6?UzwiS-RKqy83CX8tVLQ->rg~|9S1H=z&9-O zysH-bbUyLD=`1aAVT<-unUKt1 z_po-D{&I`HcV zAW4^j)Z`+aO5RT;=DFA82}z#I70*(|v(4nGl05wt&r!v5_6Mr08p$(2@m%@jPnsI3&*%iswzmv)|+yA$bZFk7b7}>w+E7XDixJ@(cmbcP-k}VDS9X zqCMGnVx&63E$Z9q%0O9z)>5rR?P6Q&YHqEo(ORLM;sJg&nq})w)pS>5w87^-M`kD* zrW^QUA5V;T$aX&_!zM&@4JO8kf4@kl8maj8P~3fyPTN0Br380H8>%@+Xs|@sy2 zAte+k3B7ko9gfPJ$7R?+!kzUJDy0OklCV}uptiye2{LRTfp(iZR7(jyCE?QDQU{C7 z*-C~DB>cNxLYT!#G`qD8y~nxO$L33^Yh(!N*TTn6?)Za^h!Hj2 zyt)G zjCHxPO$ng}lIAH%^Grz(v82I`NUHD7ycnz`RW^{gNaI>9Vkvj;Q5|;?zX$(;p?FgP z%T@BK8pvHDkOH8>-{!H%GVSm&xWi2~5){aYs%%v>z?LC;j5-sKT_-3`K)HM(w zZK)`rD`ueDQn}z*V@W1^fSzYs;U}p?EoyWjIYE|S56~JWPXw7~tea(pYGeradFGGZ zXGnROsk6f>N)XQTrqKOO*8P=FQYph9AbOLP|LzlIc7dsr&p{isEz_}ajn`2&igV^z zty-z^UKv8IxBgfyZ;9a@4c(MZ%R0pBhASt0MB7-mOFvc0Y!ijrON{_+=d)8LeVU4m z7~*uBrvO48AIN*S9!Eb-rB32;PhX16*yBd;Y8MQ$5G9_8kB%+GSWt`p>bk8B>7U_g zbzMapa!AsCknXN%Lj@*jgd`nMB(F)Tlq6m$My|U|Qne%zsNU z%l4yq5nr`Gl`M5A#7|e@#VYLphJn533%eAiiqP(-e}a15fmAYy1&TPvN&kX^GY+61 zIO$q|6#!#gG#ubVfT)YE1Na3+ki~H(zt8(1*Ln~u_ehAkXe5}fIw)ir0OJqhk=J;W znr&#c3thGOI9WACwq*+1a{fWlmMLh}gE0<9cPs9WT$CWN9S7 z)8v@=>&8!08*jMR;_=2fa!rnf5AFZq#5R;KFEO9!50P3&k=}Ft0sUEmFm4@HCZTbDw?((%qLhnrd_{hUmNqucWLaK~xsYszSPIwfXC zO8mvq!TymY(eC42q)pJY4j!V?NDHE-M!*~N`2!;_DB`7o%u;_eT59wP`23zgCRRPL zm{D54m~na~mz9?y|5z>%KVZUg3qos&UmTQaEe(puX-f9^d`?#$Y5!)?s0&0Rgv6V< z46h7J{NixU(<~yw*M&m=X>q)@3_@{-EQjNX!#}ZLVR=*;F>D$UYfX?r@u$^{X`Kgj z>ya+MKh=9ssqiF?cT`@dWt0$I_f4<`IS%2&q54Kj*Wsx&5)X=7u*>o~{%p{x%;Tl& z^SdOISmrNG5z{YzX~r`DrbBXdztdN8b$|F*a&^BlC7DmJHCOkie-*pBf0|V*^LXX7 zDP6vXsESnvR0Z+GE#L|k+#7B|FMcJSw*{!#S6`))s96CXScp4GE8ry%OLYprg7g5T zWgUNBJd0EMBc&fwPP;8)6~^pdsaX(=RreZJ_Zp~MQHPIuW0|T@5k;CE`sjYwuf8LSI~iX*4)o84XIyalID=- zwx0hDkW-@Rj9e1Mf1677utOf{dKV+7L@$4vN|r)c)a#}1G<3qFVDQQ#UN1W;W?7%` zd(2~uT=K-s4UDo#9GH0ssKgO=(0Ueg=69@77;j_{>tgH*J7^>0D;Yb&PTItH9AjtL zNt+qZW9$k$X$#|5fXf{85u;BSIcXQ8Cf}!$z05_s8FgZmMISS|?E6%*q_j*H2ai9l zagQNNSbe!pb7X@rlp(zI7rco4bh%Dt-(!bNe1DzxN%A2Xg6&~E(3q-H>td+*u1$&A z{dEepOlE%mJUzOLCOwKRshgsCWTu}obiL*(TjrlIshs`N9D zNu7Hmg3|rRl+HC$=Z#9|$;YJ5&wv>^e}M>cCH@H8qTDsIyf_16s_xlN&-`8I!`kQDYzJl+~{yLTZluBt^ z`|C6n%kFG^pkp!OuNCu}oiZtPuR7&`|zfQa4n082p zXx+~E4x2GRr!K#wQd;%^odzOu@c^A}_yul(j4H#in2flti8IasKQgXB1Q|c%jJ*cv zlpsxFks(UVMTz4E==2NNDEYJo>!F^J=37|G>17RL^No0)`t7&heyjfNx8KhE3lhXT z*YUq5q0AE*-9I?i#oLZBbBRAa4Um6eFxo)GzfVs@NW2!t)9xnq^KLoE9_M!C*H5&| zJlehVr>J-=bpM`@yVL~Ay#80 zudaby>nX|Y>4_Rb;aG`E5;;V-5pLV@R3&#@DYMS1JNU?UVZB1Ry}+aF@JS0xlQ zC<%|XS7KH={@@|wE%A6uJT3>d9zhnW87Us(6{j`QnKJ%p3Za-X#|6We8{>yW!-l8q zs$pi4YAHi_i$x>KPi%aM4vqGYqt+WV3l;Pf%bHcEr7q{MU$O8+0#3EDfwT{sY8}RS zN{L4GN{W3eFEPet_QE2EF*GQeh_qX@4w0}|Z2YkLB2l^(uaGgK>ofVVF;(lNbqEmQ zWwB8v(z*7OImO^!xe+x<#X5e^GhV7&<tp5@^70Rnb_>` zVN>4i(CL=2BNN}qK5UA1AWQY%_;WyKpd7N1#o8*F>|tMky6m56p;wSpgQQV(0eEiK zY-Ep)p$lN|jhc-tBmSOD*!l0q0?HAPgMOdXJsWzoj6;_3A!DGY%pf zfK%gbWH%yS4}K`$ZOg4mq&KmVr3MA~j4(g(C>NgjeN%9T!<38ljET)`M3$7+$@t9z zuW3Bl%!cnYJ-)a!iN0%QBTFHOLNxV+qFtNY@Ro*SOkTKFrF zkFM3JK$7x7`uSR&9%*i)j!lQ_l$Btkgp}brE$%0pW|<7z;^Z2Z?H;aE=ER7#gL@8)m9~38`icQq=_tW z|1BDZ1+T?MEeeK;qMBSOLc5=~au;vX!lvr9mFwhdVIxWm57(*S9N`T7rvB(~ogQsr zqqL{dPG{Jtf)T3D|Uycf$n+ha#(jVDUJU&F|Cocc4Gi_w4J_p}Rqhsh-F4mV~BTF69qVyZL(%mU+ z1NDo3p6)mBYpIMPRw4? z`aD6ukKb6rd@{^^;w#MLdz0(G!ZhH$B0SO)@ZZQY7;Ggytqh*hD+B&0zV`fSwjRN< zvS0vpz{j6IWglXM{hrc+UvAaK{P4uIq-;nG>KP;h~gWlAU}+#oE4^oW<*pKC%0{hOKm8zb+Jx`ZN)EG ztQt{TPO(l0xSus?AU3UYgSwfjEBSS1Ihs;6_<)7K7&u6&nt-}3l$?uYhz;}jBT8ye z;>glqk*5@PqQv+eLqu!BL>j$%c71D}(q!IZu~5oYZMmF^(`sZjPyfTOT#6~PZ;ujB z4}O2pcEo8NMjPWvOkM*)xIM?M_*$`xx5LCtdy7u;p)kuu+EwN^u0Pbm+?nku^?M?j z{Q|}~k2iY77$aPY!-d-N*+Q83;42%#(~;IKFy54O#_35it&`Tdhd)$;b9iE3dU#Jn zgdo*R^kZ2IYHZT~hq{_2+)(ZbM~(2{V0pktT7~!|i(M%AjLD&wL?{neIyJE#pKn+& zOj^rE-$(G-mt_|T+14@C7WhwESAN7q1OLlxV0^^nyr=Z;7%;|O8Ss=xON>AiKL#*- z!oa4lj~FI=UObKqwG1y&AxzxjEE{F8d-5Ks(SXuNIlijnwrL`*2Y0}?p zbiTDj$9JHN5}jIIEo#t4hR9XKGC@_T?PI83ld4j!F;t&PRjH{lRBQ+)maJ0KVyM{c zE2t_pBZfNCl&n%6F;r~x6?s)^b_~^T=Dkyf{(|nMqZtl`&LoA{MQ) zMyBqPAtVn*-c$3n`p!G-ioxcJh3R3A;i8>{r>(>%Qlga_d`)X*qYQqJ7e7}G14pn@ zAaK*OZDcPlFD<3dcvAg%8yi__+u}t`EsV%rrQd*3(2id=H!<&lb8PtO#3uehYm2s+ zA6hEGfU|0j@Yh%eQWJ)k=u|0t!x|aF0@d)qxg|P1(AGw2kCf`Xi#_odC?% z_Ja9m#XSF98>NF8kMUD@lEt*b?}t$J)mk-$)fd|&=FPxpk{0KwCaK^D(bd4*Bm(!b z;?*~s?vKM@^a~@mMAim>qaOBfZ~zV5{zw?- zpvBA{WaO4;B_pRIYN<9HO|7Vx2}n zU#Eq#*Vf7qJ)#|^fm8fCxzlWvHq)=uXhatHb-E9cMSh*0O|y}ZpmMgqPBR z(xrsnh(LlL5lE;?R}xgtVKEYPCE+*lLkD+f8?l68h(N-M&gfQ4{W=w#E4o!bO8e2TQ>QLA>Zpy<>AEh8xk@r$u9$CA%nKFsMJ~e8L6{exqVupqQ&oW=991VNWIb zGsWEO0x3ED0wp;lnR_Yb{uf9!f{OWmletG>$=p{-o~W3gRLq~4%$1V)GR54)DJ6GuO39ZwmABW85d9>>BSQ2S8#1ED zl_4X%{ZT~FU*1Fn-o76a647U?WpQaLm&+x4*-eNb*Mo>4*K01+2)XKHu5^{_2k@Y* zu355Y6(WLMcV($u*7JqY@CAi6d;&bkwFeQD_4jNWajv{wqJrD5_GrJb_gPnaJO68R^=>-k>FGkUIjlSe1`~h z=#nEf_v9!ED(9RS2`(jJHuxc73nGvZpDQKg=PC&*=lU24SxQ0>{E)B+5lHw55!B+W zJY~HKX}uhkt52S6-%*Gl*AzsMYnhpAnasr-Dq!<>z=N{BM+CXf$(Pz*oUh8NmAUv7 z9pt(mJjiuBBFMD}5tOyw%$1TVyaH$1aIS;kL9R0{l&#}L1i6M@h=B1_L~|Ks(|*qP86!6xVDyhmkS~W0GMdLImku%7%P5b&V$}9haOTrtMq?N` z9dv}@b^)DK$1w9B;B~s_YldS5%%X1?RtuO--!e=rU}iTRWjMINMwWt9e)NCOkN$ZD z@Kd}w-6_iJ^u~s!ClJLyeeJmAWH_8f7X@ zirOVpkIN8tHcYD35My3}3_&eoD#mP|Or0S^P{*lJVOxP5720-_ea?#r`rMLkXk|!H zIp2tpFkVUc4*ck~mv)yDCL#g}HQkj2mGfwfgd3HFb9zV}iV(qQF$WPy_@sxDpmL@b zi@Kr~Hz^67drAorL?Gb_L?Gdto@$h;l;h1UDpy)Bnd@>ykn1)?kZWNtm8)9jx>eDwowJM#S4xu0egIh7%D% zt|t*et}V!wkTgoC_HuiHRfe$41W34KluoVt!hm2)mE<%Tg6%HFmT$6UNU}qQV4JAe zt~J@RB{@%qV4I}aZZ+91mgE8%g6(d__K3;WQna6T^d4_R8 zZB8lEC9GLpUeC(WkYY)shCWbE@Eo|vcDy_Aa7^vi6xbS16I znMD0Bvymla<43ywcd0Is+$eS%S{Z|U2elC^`5BM`Te<)9KWVNHg2Fw zqp1(7G!?`#)CY{ufoQcSlYJcZ!@9?NAdaWY5kJ-+V@1N4TiVcZS-G^KLN_#9H>|%s zeoGr_H^4@T%Wi2yBkbZQI#_p3+kHzL`sV;hFuygUMbkRVL4sW>UHX)&bG2D_En_s% zIor!x&;tWb@FV!#Fv<9D8RBt|KlZ#e18l`du&N*(pp zOKmdZv)s`naRY69(+mUKKpVyU)Rm?*ukW4Z0sjr<2DLuk+G2j?tF)g^&y*7Cj}Gz_ z`7_fxV_O#Xq8O#VZl(T_0n%2+YaOCQ7i3{mSN-R#814a{5VDary-`+gFyJi@hmAlK z-$wcceEzWEjSdX@jDL*`W=2XpAtSSM1Hq7%VbYjX?C~1bn{FOX+M97&2VBq*h!V}6 zXCaCgRT2BR)F;mo*Q<}TXh!Vgl3Zn;(MFHbV89q+++cDQj8NQYQoEKB9UmgHqt1QN z3b){UseL29o!TSD4b&CU(F1Li@LCelzhzk!GK37AI-FLUL^O3ErVL=4A<1)O2)0Z( zJ=mTDn^~2ja(^klwU>LNrmjlIl%iN&^Q5kM(DgmI5A>tFJr7vu7$B}$xuUojLks2I z<4B?luCS3kf(qvU+&_tmudtD28M>^WE<}G`fP^|E6w^h!uTG|)uRyO$=(y%g8Yb-& zl_9FYyQliBIg`#Ul=e_u*GjG?L}JE(igqhxX)wrgNna^LNV`}W#K(N9)z~0n^Ra3> zWrt?*KbdWTMZnGJTWl;jyQ1REa} z1^eF$Hq)+QW28S4HNrhg4NssvWE!*1G%9`&a(_(!IU$F<^g{7h6syKSzUi7~_ZHIw z+OH2-=*7Wsv^0Dlpat5*o=J4(5Ii31MA9RiYPTPmrDB*+56j~^2wOEEI;XF7+U1kb>huS6dxkER}YffNek5PG-NFsc51_n+_ z2;ZSowM?#&A&TgL&tsG%N7aZwFtSu``Q#po?Nf5I55`JOF$(?xo=^lkrR46dw|+gN zp2I6#aX>jG#xI^Slpn*>LfEn;@+yYjc?5YRL- zKK5R!!XW!*z59vd!mT%1IsF8m4Fr9L^^%j4>ZP3W6ZfJjJKLIgx}Ytaiw{`X?-?UQ%)?bHxEzG91;zN4P3{9pT3{9&_W$} zn|=C|*nl2eQF$MWajLE2kMu(Itwq6L>F-ObTB=Gq`yrIXdY#a&X`Rusf5Cp~!t$g| zj@PiWe`ghn_>Wemf6(J2Er+d!y2MN>V@A_}F_w++mnVL^;^gM7x<`e*$_=6q$0XDL z-(%Qgq$mGZdrWN8qr*^Bb0EybE>Ub~&nE3!t%Jws3lrIJk4Fqozrry&R|X;>!;7DB z_^_=mmaIfLb4V*zGV*bTLyQbCW%i5YkhUXU>);O&{q!PyyrCwU{UX`@0y(5@{}U$C zw)_#tm~7Ja$JLk9!B2!&-`{C8CZ`-~e|IXp5&j9F&~IO}df`@V2S3mZujjv$FY@G& z_UnlmCN}?r4B4c;e`1Cmzbip(?eA%L{bim~(w=WtzZq`aN$QKUJVT__{*m$i84MVt zJ09i0nU8N)iF?YO8f9A?miO zp(luX^)<$k_FBU_sh^N?4GZcOX`{hc6SholeU9=YM^(T);!{f+xI(+#~mn+BV zHFCUpo*Zw!)9~cv7iT&0inDzAS^4h#V#DEe8eT_Mw%6@5vU9TyS8k3gE5B%XdS?Ha z=jHcrqD)iUnCIpHs%?0-X3iOeVH#IbS+RaN!y4~mZ(r0QD=jt^6O7&@ErZ#IX)}MlWc1VHj4ozmjEkyo0rWIyy#8ve zqGG|RvU4&oGChQKtgUzG^dys(+@n*qB+UaUcd|~OFbTQpB>fgl}I4&%W%R_`*>djMC+43v>$q z3(iOVa~J6J)U`H(Q;UvtbMFN@>BB*mixV_KyK;d}FAawacS%O6V~l=*h!KwRbA<*$ zw~v{e0S_NdBaot9hu<922-Gh1I(*@<6iRx928L;Zp+W=4a&ROpnXJ>WQKIdxl_5N7 z1U#wP6rDO;kN$w`!7DOEf56X5_;4*~ z1msFB73HSM5asfh1W0H#0_8$IM^{nqtxBc4K}rZr(dnqn*|SVYf`kcRduxhL?~Slg z+O{b=Sv;x-MkQZ`48eaN_|XFgdTf;T@MN8qn*5U_-wYXof0E+g1~zKKV{^b$`kx)* zWBOR^W#p=x)tTK*V=Ir4Ho6JQ=>t>9S!5GmAchvf5&j=Nh@hpZ3xz}@h#|mX4gn4x z0$RzA>T>cB@OK4WJOs2>Fq1bq$Utd2ugyNpj-pv#p4GYgYS*bvlZ8eQ(XitEShBa|pWfA(eSu9@gnqFMI}G z8V!nSjgujKAP;7SmwpPilUbMFUt%Ukrl(-q^ar1p(=$th!O`U*)1oyEEW0<(G%b&L zane{>Y}j~-ve~XpG7O@RH6%$Kr8D$e|BLSxpi;s#;r23rzxpRrI4 zbj9i00N+B*FsbY1>-Rda&=Y#1XoEl1K$YwWH=@Cc=PcCPkeea^@{I4(He5Gr|NG6RXnEtKQ84+p3yRdyE zvsii;z@U?@*D}?}dpG+Y(eWRm*sn)5toYEx}ZKUP!Hbo<4%K@A}@+1z`n34XIdJZ=gtZYot zYhJWaZ>1=J{{%&27RhjYh!Rpz-mumb!&vAEM+__zn_LxR6gT#IL^PaS8o@5f$taC? zFGcvHhd2t>a_Mo~~&bIKGLd zYd5a6P@59dM%#p*BTke#a&+S|V_TRF-mh>>c5kd`L`)mk{#z6&cB{_Q zv<%$tGmlOC#%Qb&6c-cC<6mL#FSSvI5%4K&DYH>Vv5-ge_*d8$l&RL4r?E@eeg2U+ z1#SULc&rRBHJ(zM&u|xDC>W8@7QkB;J;;UsD}YH{CU1Rxh`Hwm8gnAl$*e`?&9Sw4 zLu0pngXPr(Y}C~sBJB-L_$_=&dqZg1N!DVW#-9@t{8M!R$v`&0>M3OnlOY;oBeo9R zHdV(?l@q-gA0ABwLK9;Z*(%K{4v#tpO}Gbbm8$T}Y^zYk?x**;x)%m*WH(An{h^3I zLbaUMF2ojkn`@MRgN^Jr;``TojBmQZMxFd2+9OWbwp7D-XuhvGZ$cQm(!5^!M#J9O zOT#$S>w;vQeRfpC_r|$lY@Vt?5*m4S96Hd;$OTi)6&(zs)QIBA8uI znX1zpw}@JAmLW{G54K8JGF7K*Zxhsi%MjG7HKO!&FlfRTx3nRjB$dh#EPmJ=_oI>E zQM70b&nOhqz~3>LW#a5p{xqhVqhOLyL3sOtcr61TcMrtFBkqXUD1*&O2jVf(T^O}t zbaRj<7$|YuRGmgR#CE)cyqPJ+3An_5k^cU`Xg)3J%7EYNiA2pZt#`+keOD7v_FYX- z_IR_XyPDwLo)W&Ds#BrVwnB!mM)NqSN0rp$9o8eUO{GpXGO0xzQJSMtr+<{AWiG7L zX}%OuB}0fh18jpTbt*F1mPqn488%>>z-+L=4oR++A=p|%+Wx6JEoL@vrB25sxrwYf z*!~8#n=5tN1h&TP7Yv4qJl@f!y{cty(Pj@wyFI{mJ5sKi=z$o!rO6z_WQcOxD!Y+W zv|o(f95TrY-C(yD#=vf<={PB(LWU4!0~>0(gV|u-g_68jhG6ReHkkKkur*TC0p+DQ z&Lt8xt5|qXtcxs>Rb9eWJ$J0C>XMkMu9pQKlOYQ3s;W9d)+?r}l`<&{%Au-z$EvEf zk%e1j2vK=pLsgrP0~@NEC&?Gf5Nz3CLsdJ2?RQm`=e4QYRoxp~%c;#oN0{0S9bv#Y zRm-W(&<90FsFY<@$q;3>#-7WBsg*h{l&SB^5Y)5di8`*R)afZ`l<<0`PPLNUMivjY zbHKK*Qm5wQQNN>=I;Bc-e;IY^dwviPE}H(`n(qp&?q}4XN2?8A3y_Q6i`*S;D_Z zz&M%-g?3VnTJ5r8rU173_-JE z4RWNJX!9d%kk2ZVL5{FNR^O%!B5?f-?}7M+Q@jWGqH#(bGEJu%sof3Gu56l4+wZVZ zB8JhF9HI9onjAU{WrFAzlsI;pPL(n(8c&okcbZOpX9|rAWe6)T;Ia0k?Vn>-lje)e zV$7P{TvT^*bGZ0-cc|)4=3O=O?o`zknCv$i;oy*{Cmf|I&GBHqYa$->MmX3j;FC!m zCfWYfES8!ao;}^iLKaQ3ktLMN&+Vz4@)1&Mky1+2IpwUo(V%2;`hJ~~?txEYs6RAYR5*K%2w^Q|!%#o?UK^#|RjJe1d*MqkcbycqSebj_ zT#<2!3?V?6o8=FkC&+VT2y!LVfc!U<{C$xByXKKwZiX7;qGrQ{76^Hwfi}wq+RP1< zJXtl!W^N!%0-U66Zq5^0i{rv%>Qv?#N86fXf`oI4wlm+ylWjyThRLSiD??`e&}99H zS({I>QHyION0khj^L>-^1Lky0L6h)wps(apCBO#nq2tEZIIS}}VLI=f3ggyluC}Uf z!Gp2Y{#;i3^XAJ}Tj-i8Hp+;ykv`|57fw-j_*_(-=Ya<}9(SLuL9r+_Q(j&s?#ag7 zYiQWxFZBmTb{`)#B4*)K6|2M%sl*YseYg9dM1yQ}giSJYs?z2NYX`g6ajbUgDfERA z7bcVaMmp9UPnCVs;D^^$w4vGyg-d@AU%IWL4RyaCB)DkGMS^rJo+uF`LV-*WZpn9o zhGf!DXa-KHk|{q!pzF3a^cn%de8jn`H?3Z5qm7B1tP{2$p*JD)o&R zs+5mk=VwT}^p_zrx5F^YPK7dMGEx%u->=iZ z9~CvbQ-&y)sfn$o>r@fLpO!XVr=Ov4qI0@VHL;}2r|Wd?EE}Z_nXXf}S*X;t({)-S zg|3&O5}r1Gx=uc3gH}5vd6x{qHW3=0>^lrL5B%=;Gv-=p{R-3nyA5Lgs1Y{3qH2+_ zq4-WV-~5{6&Z3SlS;#Fv;JCASD^`wzZr)Oot6&aqh|5zjmv*eRP<{qJl+BN8vqqo7EW$RjE(*H2?Yk8Ck1>gwZq_()sHdj;gz(^b4##`yq$K3%Py|B6n2fco#J zFY!hm&s;1sVY|Z`&Gm{-`SWaK575h6w^wx94q(t>t?(6{Zk>;94gW}{PQ`o}>snEq zrTQOYE5laCkIaXES?Z7oLkFUCSKRfl=(G-mga>Enl>L~nbe;_1P+PGOd*Mu-QWoH$ zjBN{{0^v}6(-(fjmjx1g?qZp9xeOuUV@QDK-p_255mFG z4+ z8onXPmc{iN9#b1G&Z$ytR!J_AA*6k+*e*8N{w>LG$PjFwDz@vvme^yaPF2#dzd@hR zX6kf%m5n+coT<}hM82D;Q{2NgVd<4m3dPpQ5F+_BJXqTHFxr|Mt47Lfj)4J<)fY70 z;*io!v||!T*y1qyVY$WOKJc))_C6tsIV3|AlL&s;_SuI`+p@qyDX;|?An;u!@COL| zzdQEG?SE>d8S%8Jw0NG(G)J5)H;*_m+yC&0h}r&!oxU;K|1Mfv(X8YgRwK`pbM04K~&E&+WOvG~+Y1n}A6w>Mk4VuZL#-~V?-J51=cdgaw#z)m- zX{I)F?eDk$J^LsImik@O^pLi2txo4JG`IgPT&vUD3vFbn=!OwRZU1Y#2pQG(ze^Z3 z-2S(4txh8rVchs_rcQOzJsyI4FPNp%Q%rhlmQL30!WDV@-+8lj`jkn?RVYbMt6b+i zhTeycpSDD_Q??A@oaXkwu_oINNv@S4*v##Jb4|8=l3XW4u$kNcUIyF$T?fXtA+>=} zc`rRDtRuWQb^{?MMKK!)u|+v%10f!pF&hZswI|;|2p5mpK)7SAC3XWLpxi*%b)8N} zA5%}wJ{PaksmtRr8wf97r_|BE z=p%9ip>qiaV#_j_CN>QYMIt|Nn=BPglpsU69Pbmv19vRgcnq)YA$o_oPjD`i@Q6(5 zDM;o%!7rJFM_7R*nfnBLJ&lfoTs6;#a^IC9$~E^1mVhl0^MOh!f%geQrQJ;GJX@z~ zNy6Sh@uYA(D`Z|QL&!9D0;VsuiJgEGmqG`0-ky@LP=?^2gg%ZZ%adR`t(}1X)wya+ zcgyq{#YT8+*b{P^J+8{^c^bDwoivuYmoVWO`JnCc4E*E&?3^6+G;34yyl7xCo;cJr z+~wf$BxW}v#*>)ch!{^|b|YduiP?>a0puK>Svzjtpwqx-)DYpAyg{c4&&2FToVNix z=o;OP_}Meu^|PMbpi{GD4Z1$>Ml4^3DRM8d8*$YJowhB5AEGa0$Z9!c2tPD;BOV3Y z$=XNkgfsWR{8v5Dd*a)YYMl`;#J18m=DVELlYNkKRN zmgr0cbC{ZvQDS&}MBl`vZqn)OXH~0x6X)8b(?DRciIl#HE7+vd5~Pea#`hB2BfsJ8 zk#8e$jHk3gVzcMebmg14ft%3c{AK=|I9|=M`{`KR@J%{h`W(MI{}@-gNvCI?vx#@z z5xvB4$OkXZoS~ETdCWfW8eAz!WAGZxDb+FsCkcv_I+=1acoHx%*|%JnvrdNa4p7t1 zo~=`l=WWEZl_@CY|Dv17ZLi8H3SJbo5nXt%=^QR@)B0_%=)v{dUeST&w%2Z3bXxYj zYV5snS8UO#!*Uxn-1gdkxoX0_al^NmO{liLqBX3&MMIsq?X`aGDAZtglBs&i3RMNP zk;V&w+TINox&vxpM=QW;Xb`ibwVZz=m5K8{{>L45xT!#`%whXAlN~0P zf32)Kp3_CSZ(GPBiOwvVwaP**6m&2&yPlJOo|cS0g5Q!^r*75hYgCPQu2>gr#cUpj z8g;~>Mwh&RNz7qccm?cgKgv>^n!Nir(U^BYv99o1@d7 zl{V`5>>Qn%tg=zZSLWz+@hTf7o-y?-s_C6f2}IO}h|)5ro<$L`DSe$jp~p8*8N(G~7iYd^QG`G02)*(H=Q9IwV2f){LeO(&@IzxYE|1nyW7nr7dMZkP38EQ!yecS6A7Z*wU6G>>8&+3$`~60-2aYF znJ=q0yq#A)(O>Tr)ihqV5f4rwS=KliqDAiD&1?0YEaq1iCp$0fzc}AB%~Myc8W9tb zm=nkFN?-5iKJkVJ*ihlN~Zd5#q$LddsVy^7IZib_`Sf{NPl+Ee$sf7J5w? z4l2E%$-DVp(9rS4iECELk}5+muE4m36W7|lg8IWuJteuX48gWqv0Vx_YErMI+6o}| zKge&zN*h|Mv=U#5C-7I|qF0m;H})+$0bjAJWZ_F>$a5if;ggrmn-Elq6(^LunYSNI zFiUQ1*F|Hp<+d^F0PaXEEL{-Iz#2#3h&NIACrcAj7 zo$7xyRBwJ9Pt?_-OXFIgHZ zSo|^zKcC8Yqxt6JBHAw%z5 z=tiX?z_TY)@$`Pd@jq63xT$Y-V|s7hX`y{eZ-C_LlQ_W1elygVV+8Qium6DRN@aSp z5t9%6&_bEj(klQ4olNypHTi!^ZKbZYF}1gSXrZW58(?;$YL|E-C5?AzCAFe4>FC3& zA-#l1!{?H(A)Q~F7{(44jkTQFrT;gUcZl8+?YvBe4Sv3f1+)ualkE*K{I!#~YGV}> zqrm@4tyFC)){ELTs1-gHRH#}3tT(FmgC)DSU>_tX&qe2Iws8$fBl@ zlo>F_MGJ#tjqnx4L&pY5>#KDrH3B1}CB#RUStESPNk5zfA#=#pI)!fmq9phILVPOhb!-@#ct zoG}>o8DSS`Pd1*RNGR1vMu-~h35+yI+wjME144$?PaAoE5XE!Ju%y`{jt#1-;3cK2 zmLZhy3hlGk;n-7gH;RlAzDw*KL&HA>yrjm%Sl&Pcpde4I5eN`tTmA%QZalzK2O3ze#!u8#(MnN ztXiZvk@7K8q;etO4QMq}Cu>ctZkH(CE`e_88-#9`K*(r-A`)O1fbde0aL-1p7#2)A!SJ3~b|8uUgHKY%^t9}|qyHNf3AYe&^X(B(ZXTDwAY-FiM+m50QoKn8sMi#n^N0JXY<;(5p zD1{& zSu2nhjFuQ-`himiyvO<{zPU%In#+?3WA9N6+8^)H>5liX?`)4wtHG6cY>!UVU+`eB zoN{|_NiY~O3b`Aq%GT|Ot@MOAHB3y1!#J_~JyGciap*$>-xrk@=!W-g)K$2UCQ_&? zcWG@x96B}GOX*(DH{*Twov92~0F0ta2CY7@k$p5xV=(&z8+F3-I70Vv0wuPq)yW}i zayE9UW!LKT$p`!_$hH{txnZa$!sWT)aq3Jk&V(qwK{D>WL&LI3A=2RR#B_dKX##N$WQy$>yFf)G(GMTnFxb4tq_%t5VX7JU_LzfBM9hs;DYPrLS>85F zdz(vJy2nNt0g^wsu7k+>tF~%bbf;SUyqmT$m;g{Df%6l5IU9f{2-Pxq#V7T3kBl!H zV1!2+rUrE%$7=GEsmV{Q$%mhadjG_Naz3?DMj1&|22`Zb!cXy)Jm8~h{*~SlzyK{_ zVE+sojQJ;SPjkX`NC-=C{4*Q1uu3&n%Fyno+nM#wy=>h(7$kjeqZU!gvP6dV(KLmr zm7k-YmI~C}Pjfl#t(c-J-&+>a#Cm>X%<2yX;2cVAK;eOzTwDicA{M z9Bbm$SW^>^vF6+_g_UYp&W%82B+&w(B8A@llHFq)GyVWDincSz+-IZ2DYZH+lV1Hk z`ohDtIu-A;QQDGPoz^0ya z2M|H_MF(W|T?eq{jO?L3qM|Ehh{5_}C92awDe4YHpvh`PAnNObQdHU@STPy%xyEkuk7l`1m%!ouhBMoYO zb#IH=W{!cU^RDR&m{OXaT4|Xfk9YKyI4e7cv`J@Z9m*mjiSW~|4#hYvgXGcRrzovF z5G!wT19>>!f!`xfS+}&#*ga!})A0u2Gs2|39nS}w&-&PMI#T$^m~eWgx5V&{?&ceV z2{maenrJwu{Fhn_b%+>VqWgBAmQyjz zpmuAJs?NZ~z>012x(8Bq$DpbRV6L0FKOxNl@uB64-NE`8KkXB ziS;Yf53B)W>=nhu^l}PU#!GK-W9wwth(!Fs0MR3`T}%qPEsy2aU?@tP zT2<_`khPSyw1RC3~pI2 z(pEGk4M!}UvTC9qD|xRqkXJvlkD+yEoR(OOOo(y4Hu_+|7nn8bEDtkYYEZItFdtds2s=FH%! zl03^4PwIKdh6rP}pfT>UYy&orZjGqqGT! zb$a4^*a|kt@JlRCc2oXKm1XJt6A61h)*#wS0 z@kiAJPd2HnwQv(SC#c&u~2PCmSVH9@c5j0a21emWbcr<4C8|>8q*^VL7i7 z&kM5>bXXK2Mu-nYWj^4_O#2B7NqATIpb6dF@6Esm`^=ZH=#ezs9s zha)=09k)?p#u1$=B+Ea*5;~%j?Ks->G&-6x2nJ#G5H-zESr==LL&6+~*c^S2D{~xT zb3Arjnd1;wLsaP}j&C?_qZZlHw3RYcJ?=0YY0xiNKCT%oP4NTgiz1OtVI3k&F&6cN zDW3mDT402vzyg2C69$zV5lrLE$}T$^Tj@KSvP0k56b^mXud0%FHnr?E+x*U^!e`Ou zcQ-|wU;C?#>|TD+n${Fg>`h3Bjt|k4rkD*J284^>%~{X-4Li~CzBH-n(9bMHmSsbv zB_=S#RI-z>1Pv9In4m1Ng6xz4Ur6~{=-o<&=pMo6oyJ=+VoXa%(Q**4!O$$3wtD+A=n>=(NkZ0jY^yc<^#2^=o3sIabb0{D0cNojT60)9L?X@6E%bsFwfX>I6cDOeAu>5rvEs(4GWC z2q=m>Aqf&CVaNigsF}=kl99>G(A@!|s6lbVEh;JsXxul{TouI)<9hK6SL1ROy{J)9 z+zoxCz3Z(Yhsz3Zy}|;`*iaTe9eZRF!hx*^)&Ohe{*5wqy|v|8h$fUF%TE@#B^(YMmjf zz&Vg8XVec_^oB#FoFjk8qU#5$lw0@%D9V1wqSXUYp?gs^3G;JH?}Fds^e#oL6 zr%EFpW^JDSA&X9SO8vbvh5k8~{)?Sb|MK(%{pY0bk^be{JUO$sb#4D6UgwPtC9k%r zj%cF;W0)oBDq$$U(SZ@nDf?J9*yspl@MX@Ll%bMy!YMM&y~?oV83;TF zZOx+SDWaB2`xE7awr0^w87d7ow`S3xeN`I1aBCKwx9?tR7q@z0E7sL7UcpD5tl*viP9mqfQY<|o%K{Q-2JDoynw17`r}xN>V2c~2FlIy{9a=e@02 zbUtT&vNem^WY%%W8o4ctw(YNyBf2e%GENgjQ_%|U-j+ot429`27i^JPQ;>!E;_Y@; zm&_`+vc_bg4jj*J%OdYA!Er(gQEt+YSwuMlf6St}St{l1_hT0IKS-tF`9Ee+#X%}@ zL3FyHcoEW}pcMx(H`ITP%qv13>VG51QUBB2vi>{Vz}^e*o)ioRjRw2Q+xy2?{imYp zKjo@_I@@ZMpK=YNmmfpQr(6Sj5*n?JO3V`EnweE_`+#@>mZb9xe@Qx)n)(k{33nZp zVUoWj9bR-SK>kQ$Y+Y?hI#$*$>H)_jl$NB!5APkW(kL`UMX;bz(oGeSPvUOyCF$(L z9_vE+C&c2cVLL zxItAZ52mNvxb_BlP@%YD{x+_f=^o3&#Iy9U+eB4ZQ=8iv_ezfppm%WYOCD7W$lIiz z_drh!$X^^O2jl|}l>_oo_<;er7e6o{Hx8A<=59%C56J&KR1U}|<-tlkAfHepOnEXw1M|6vgMZ4RZi)0D zia7nJEIKY(tHow-xkd zQozIbR%wB5SOCNLe~p#Hc>iH44ZpEFi!R0w4C7xv3?UuE_%2Dc4OE-Dv*?GzRO0dc zorlYjc}A@;Esy65wq()24_Ao?_RZtCjxeyFAd!9spL4ck(eiPq$Sqk^euU%`l_-ay z^@7iwBP5>@M>3yte#)XYiS#S@Jn~Z(O*#@D`%@Ne2NKtHr^ML-9MtvD@hI!NEc(xQ zFzS2j#r~lTxmF#!Gvn*}u0aY8o7^?%=r5hrJ>F`ucMU?k^&O>BV!XX;5aMmp@riJb zw|5Ofyfu!uoI|jX&T-cuj<5K59^(1k%zo4HDyf*U7aEnOPLB`5qOo^^N=~;QN6Y9* zX1QOXN@}fE9}G*tO@%5sYk+bV{U4)@orqyVH$Y)+p4ZgrLx!)LsFJfYkVb0M)aji; zaKPjumDI>=dS?*MWtJDgpnY#XIntyu3|hwQI3V7r-wLCC%SN43WEu5a*7{yNM-T?T zWf?~mTNV?~(qi9Ai}{;$9rpd6kpsmrnb`K-#*l?2s?hFdew;7~2v`OAl}l4jRLL3D z((vhrg&C-!%OQS45B-4xlUE-Qnj zxRdXcNKx=XC;wTQ?Bqw5OFkJ5qA0tQH_IiTC(FU7&&Cx#n7A5X+0{EN-o9_`D}>p7QY(dn`)Wst$pLx2j1G3&E@g4KfS##ud1alku|qbrfkjsGU}qKkiW0d zDTpr?S#q{!Tk?w6B`NkQWFL2}!kWEjWpkS_JdrCF9+X!_JoG(n92I@30dV-Az#X#5P7RAxFqME5S7hjnn8O)Q#G@S zi8A(t2n#GsHEd6)Dl4WlQTCn?P4x=-a`%L2#>t|3)AodDa%uUbsVrdno{*Ken{qK1FpBfl-2GKEF_v(HpTSi3JRFR9}F_B|sjsHC)ezX7=-e+4gg9@##$&i$N4unyk ze>U&72E+A5z9=72Y3XyhO?a(ptcGy^ptk0#19eazc&k@)Vhp5UUCM&81 z2%D>Yb}LCfnGbz2X`g*cl3Ig2gQ@#$B}rWvHs)!%pzX6qS+q@NLqrdnK~2#2+0+(& zAi(c|`9gxe&u%5@i}6i1Si8@Ctf=FOL33IK`&3IypY3In78VN9)V@Xzo3g4%Q1;nQ zHf2Szpxkq5E2Ol2_A0x~LQ$rBkJT&^gnb@bXo0X%FkRF6v>sY|pBoQT7ME96mK04r znMwDcLP0D`E%O<(CHe0I`l>JJ?ZV?Y{LyC);Yhe7?2iPr03v&tZ$8WLhcd8_gJ9mD z2dP1hDJX zlXfI`>LKX?j@H~H9b)&J)H#~48 zxBX3KabpEXa{FE~E02VWeBnqq==X)-2?#ZN!%7Z7zyh&1DkG;Ck&~WCj=eVJ*PEke zWOPHL&rg{dGo93S;(;6=y$yTvZWPle$1%g}MU6gxZhmD`qs_CeRPYwN<>M7>^N&9^ zlunbqD)IM*$72QRd@Q+~rx%dIkK~LtU_nyX={Q7feHt$$^;XynJ&2FC$YL^1iZ5<5 z7dIKjt(~S4R(mF+w6CYBG%8HWWbWC;j^AXIL%UEYc(!JiZrV(!Jl+6Pnb}k@n^l1gc%P2%!r^9) z(r_4`GWsQyPWSZ4n%ab~90ibdeWgljP?Av)#v5NHx88!V*uK?RJ+}~|Ksd1)oB3A# zO#{|*)jQQ+xi-%hHe*`6&K4UKe} z0u9fYfqem(*Fb&Uq`+v`&rr!D?vkV{6s+)*vkS5X=?+F%KLc9@8A0A)xSK)W1C&^M z<=Q-Pccd+AM}o{(Lgq%IYgv-PGgZn{WvO(fQayGcjhl(nQ}G^b^UNP(bEArQOI)QW zMA(q%hhvzOy302n^+=PcQRbzmIqfCR3Bf&S%!0u#U#H;g{G5^az;R(9py%d?v~XcK zP#N|4HBwv(uC(d*HNUfR5X`wGoq!%XMCey0h5jLOfxx0J)Hr`8i5(+Q;fff+V6`zBnB8EgfyN^MSi%0Y?)CgnDh zzUo#{IuW}k`B7ZHj^2qBhmwt>-Sp;H&cE0J6Q+p|OqPFITRrYWGI?;L@D2(s7f^|SAcQYex>Um9eaOjRE}qM$o;1Tn^no|OjBlWd{9pi1fY9x6&+EmJiZ zpbjo~E+{B9+%wh0sjDUr@KeM_w&V~iL1n4&geas;DY&Ob;NwCwQAz*n0 z=n@8%1C)xe(ITs9sZ3#q&*2uR{G~wS=+ZJuF@|&Q@p^h8WFRrs*CNZ=zL6 zN}kUkT9x$fw@&&nWEo`@_xDi!asD;DQ6*=)(BO$AILS;vCWO*Bjy}amTN|Ngi-bMP zu*@)E?Gh%g%sGl-T@v;p=(dh%!FY)D8L%XT2ZA zk)JS6d0u0y>T?hq={2sgpXaFL>_&D-ULzFHWk>M_bGt-GT@^%7r?i0?d|+X0U<@*V z(IzlHVvG|E$>Sr&sAY^Uf$=$GthO*dXN>;=gB*H}i=t6c(=!#kXO-ik15K5NznJ5q z%(<#`hB*S|V$?P&^wP1ya?53k$`Z~%j@NQrG(i&7foDdpi;kMB((vqD7mb^*3WqTd z6D-xVcc_yy0z-G<`}_V1feZxVO&YHg=^`$`6}hrh4Fau_fGmh zJl5g~m9I2gwY`OFU)J^(wj8Od&!8WYUischlg>uX*3nir@wsQKj6gf7x#;idhPYzsXX({ik|`%X&@Yem%Y%OZxyaHl&(<#w z`i;C88F~RJd2#yXK|tttc*-^>eSQh+R{;I8E|u+-XBB5(`uB7wYYLe`zt7;osilmi zM9bt%sZ1JrTUN#aWjZf~%&07ftqi|YHtI5D=mn%W;*@b%%Ixd>(Mb<3W@Y$f`X>BF zg$w~o;pMwmhTrn8D4l2v*LiZL6cV3Q%2vcTezGCgjhBltWj5^m$>o-4*`t24r)JqG zm(*~hQEw1mwp(>2<~}%}F9DAM!1;%OaKEcC)-n-X#Z9}v0^n{+BXu4f4(rUk8mqIj z>2NH3o^v(Ok=sZ|!X~#M7b+i#q9?Q>+>&GLp)T^y6`zF~jzMJIp)PufGj}t9nWUNi z1{lOLL|eIKkR&2%RPr#f;A?CLU89mPPb17T36DO#o<#1s*H{MPnVM{9jgXR5Q*Wj) z-0!h#RC2b=O(Rup4A#<|dwz1#H`l17s$tUSj2lQJuEl~dHN)noUaOL`3&>E|0IO}+ zqH}W3OA|*{&j#$F>tF=O-Gei!ZJs!mi!npay1|)rn+o8NRpQn08NXqhLtuwFH}2m5b9N54@k-e`bX3Rd zGMd}&Xvq2Djg}LPh7(L(D%!|sh@&sH^c)RWm3oebY+PO82o)|~l1BMAa}z!c>P^G% zMmiF9dI-OvP6o{harShm-x1-v3Q&3co)Uw5c;rdF`=bvq}Za*`=h$G}jhmL@a zq#=(e{{?#|y4graqAD(0rV^gU(ouMP29LqIMmh#h+vL+&Iu4JAF2|lwov)Ejz+IK)5n5!09HBk50`2|p2zS_6Tc`|Q58?7{;qoEe z8&bG@_)+F9DmgI1%a|W0T**im9d(Nw&&|JOuZ6>U?wBKNiQ0Qg=CUPoL9%7HSgt3A zx{f@IS>px6fVk`L$#d9PTMln+4@>3R^5jCEgKo9t5o2Ko#?aAtc`#xQ+`0$$vDdoe z-O7@K{5|9;ddG6BBFJ;ct-`H};8tJUs*(eP<+gZ%Ff>kUQ;CPhC$;@8$qpN9OVr*| zvT|Fpa!A(LCL}9|Pd(D6k^|%F&Unc%ivBMo!zg;c+xAW}=}?w9-arq79dCp8)Q@AF30HQLg&YPN~QNQXPQ6K;>|3VDcY|Fhk@aaftlhkShe~;Ki1=aOJNx!llGrV(Y&Oa? zQxEzLYWvDbw96BuHhQJ$GX?Ul&*ci_3b9+qj?epZeI&5`lZkeHvRjmbLmZL1(z-IE zz9FC&`}BFi@R)IgK}`4V)UjxNy%6uW8&g=cjlB`?`T%inLXS<5x@~u6>RECkd0*u& z_L;tfa=ZF!El?hDcmnNToY9U~xWeqk&>Q2Atq+*c%2_kw|3&zTf3Y&e$^srQ+nT6a?pX_lln5rX#sv&xl#(iA{BN znvP?-dq5oanZdd8Uv->>N=Dbti}gGim8PWFe#IHTgg@^o<-bpfh)y8%Ea#>&03^Zhm1jijQe1F8+vJ=u9VN^X*v} z+-BGM49z-07he}aqqJ~brKtzQ^`yL>%7%ko+&XopettX-93`TA^)jptyBh`)Kh$I zzIn3NT#XMtz~EC$i%D75U&)RH14JvpNSob6N}w#AvLmxe!Gfqe>I>?$ynoK=J>4N_ zOvif)rCxrAM(IjcvK#R}4qM!9Xj&Mj?djuWW>ZU_CMVd6+X8R+RTh?)%r~_#zel9} zlq6l?y9G|l4us{4G73Irm#sDWg1q^w8HQ#^U@(e^Cm|nIrc-v49yCZ9T9r=OSl$vR z53EY!fc$e(C;%f8JGDLPQ3FH(Mc&$vNfX4nx0k> zty>c2;uEQf>UE?j@PGr$dlE*1fzq&T2ZNM5Qxj^&vXDX4Qh~5Qccr>2dvb}>DM8%Z z5EL|`5`Q|malg>*3j91#<(-JD!}D}sl$878sHuN#v#A;RMUk-S3qsH-?1ThrW zGV-hRU}I%dU0rZKDNn|NQMMsq9GMb@H6v5-SkmH*>(qz(_R8*cjNYACB>vj9-xzwH8D!-L9IbxVNEmj(iHkqWn zK0wKCh#0arNt7)CUw(j%l-K@(qB2+?_L)t(Rv473$);G+R9XO^Ns*ezTJ0+dm z0sd~dL4PO>^*x(_9c_2!a$}N`Ew~IR16g8xtq#l6C8WHV8uRJ7z7Ww_B?Bp2*9?>R zz|*e6O3a(52O}Y$8H|Jzn&GOH9#N(jm4ik_WudpSl9aD|SNfX-rSm8IjE4NexxOI3 zvOvmWg}*%TFb@7(F;5F{uaZAq(}Q)*q+Fh)WJm2@iB7)8N!iw?fMiP!){9>@500DF z;Ratg5Yhrgcz@XM3)vCN3XJPe!9@QGz`(L!WKtegmwIC+BIVg+B|8`;dgKph;^3pU zARYerLNfa!cPlT7?jt##CAK5%aT!e!<1_2i#fhl#?Y}M<4tOJ`7B+*vkk_YcVbie5 zyQW!eGKpxBqU0G!FglxjCoa$77Q9G71iu^04ZMH4Nck^@L24b&?9*-B3-`N=-zG8r z4Z(2z$y#%1FoOc?^Q|l`@(@>z-MX(-Nh^-_hDbJnXf2n z<j#~GEg0kM$j8kX|w`@zAEvviGuvE305sHDyg2Ff5HUC*8ULcq{tUc+7HoeyBO z&*3a{QS_WNf<3Om?!c%r7ybJI+`m)iqC-}yH2gTA;K$@L7Xa-*62CpY68DH5w?32H z3k22mpekFRNgu9MDW`gUCS^Sc)(MpvUqgjrKxH%Uo0=OJ9xB7nHfE`$3g*7nQk)a+Z|2=;()3%DJn|MUOtz7h&@2n!^4uwkRD5 zBCS^B2GVMlfyA+g)oeyOn0@9+&Pm;Y!V(n6;B?H`@e_mfrD1c5__CzMy{+0(9Q&uU z6x=&D3xiwt!`9;8EVL0yU3Sng+9+k}mMLgXC!!J09W)F_zCK2pxv{>6XHph7?@S?`!nNJ3@ zpFbs?9(x44z?gK2o%*ls@|t&DMpIb!XH!dyxfqLm&8+xh&gEjxMKPI=;*KlK2F}Ik zulM_Z{OGZbO?NW*-c>4j z49y=22j~WdNW}7{M!Jm0q+P32QUwZtiy3(F<2bdh>2weGzkfasH#iT$jutAIsLfuD zg9XFm;rDKbL*rlZa;zZq6j{ z`NA`W|`co?9e7QN3);@*v5THHy6k&9) zOyPOgT713+J~$LUEjc`eAK-A|5@lbSXN@m+k@-Bnu2Alx@h_+}yu92+Vf>h0?xJg6 zK!8r@@UW01Rx)gbC!7_p_|haL4;s+YB&hJ%3zo7=#W#{vu`+a95~M%oMU_+x(r-_C z>$rjRG9GJ#_0xSJx-$u(sOu#dWUi*yg(CClt|Ux=ieAC7v8Kj}O?7k!9~K_^Zw>-C zC7tql8da~2Gny;WW-b&OGnsgFx?8xDx`#zNdo4twdy>#YeZLlFyXEFg>bO9Jk35Aa z_qm%hiE{sqzmDP?Gs%5n8l6T&r^55K8!^}-$JnZ&W>yWQ z%U%WN@Qs<&DsiTQ^EDu5o?VjX?0BA4?qQU%NbvLme~xU*4_voUNw|DB;k>;_S#iL>4)_QE~|!8)qhmV-LPsX&Q>DSW?UkQU`&M0 z$N!_!a754V@B;xvgWil=mVjmA~F_WJ3y0lEqr9z~Hs{iGN zOlp;crzi+(J2hxCQdn6&W@>4%-IlvULI*LL*=V)tjohY3cUo0wb_i^uHF4S%-WMVarR;{*TJh~UsJQ*gk5pHhEt52J>t$Xy~+ z*}wq+wml$XnMu8X--&HR9mg8Jstxj-?*qp8zE z<*Ox5T8d}gH*Y$7%|!}up8Y^254%y7ubwVe;98h@p(55w@B2cgFoJI9b{j&|3{#}Y zJ)3jST{4V1B=l{WBKHym=QT@)5#_$Ze>?G482pPS!6jAN5BQfdH%M}@gci#bxfj~G zH4@q)Q{iq)44enP~5{9JczpU z1?eV6n(!g6Y4+*9MuXazjrO5R9yHq9gk@7dTr!L%%o7Dy$rSS54%?{L!#4T+cRcyMAlNtI1{fWZ%bs8Z^r>xa>HtMs!4Q%Lb3THT`ShtV4!qHO~$Rc7xeQ=t6| z!m=tuBy_w?k^8uvYf9)6nIiX5JGWaxGo)6?eHgjDn1T0!f@ZUA0`CGLwJ^ogigbmi zxeh%aS*CcJb(WDe_Ar)|r=?x67WEAO4t<2v$?=PS94zeRN4v~|^W)22Yn8p0%RZ>f zDtoQMdjw+jF7_~nl(k$QshDlQ%2?-iVL0at(VNWVwl0+%uzu&|BKrQvs+d>0P%&s> zRp$yROqs%bZ^L|?yUBu{C$NzFx}9rE=n|PCcO7zjF}=OmZR_2>Snw3Ke_v5>GYn3x z5WTOwdulrM``EI<`wEt|VS_QGyf4+pxch$!x@6DCvF5i?88F95ivU5Z{(uc}-^X}G z@(SD(-&9N675}wPdhKKMbX`}Zkq7$(0p110sNhfKn`RuRo==zXsC)}}HOq%%KS5DV zwRCyX1Lr&G5B6LIYMpdHpduU;YdZbGG93OX%JW{CMrt$N!JJR|R5H6K>4VeK z=^B3IC<3yFlMtt01)1ad)3a#8RiY9ot7&d8$MHenbC zJ~h-`TkbOK7o7 zk^8ls8<5aNGDYqe$YuMuuNKDKS0Q3!mVPenbNA=cKAUV6IxZB&cgYm292i`(3baV% zx@C&o{&wyJ39XSSa&Zxm5TizlG1L;{z|B&Od7Gset2SGfTrXLCBU7*%Vi#0!kr2Zx zQ{?^=xvX!SENHk@(AQQ$lfRG!UHk?7;hS<7bxGuGyef?#(SEVuy+NiJ^_9Sul#cWQyFw?Od;fMrDfJLy+4` z^s`Pg#CicM`gLC}BoI-sBT4o^9o+aQe`!U)j-(=N1d7%A4t{%=)FvE|pjUYIarc+1 z1Wd%Y+n-O`C+wt+Ut+9y4Nv(0&1m$MwQe6P=Jn4*R0J(2GT8byWFDJZ$`>z0~SK@$v0c0g{CIvud#pR3x#U&b`Wh!`gB%0z5BIgtxduvPVcp8q#C|3eaTe& zd?N>zi$m^ z8AHk)$zrsQanXIr=u^-97SpBK!2sRQ&li3x$4RZyAs(_E;+}7@CJ`Hvw_hzv@02MN z$EU@G(Y8tGeo{u{u7sLw((w{HQ>Mth2f4i%p(J(yD%N}~)jO{fLI^K;B3XuvCs>!y zzO{_-MDn#bbt9V36UoQR2YlyVS|sk6l|ic+z34lHk?`zrWL}uhUm5fylV1Nlrl#;M z9KBkTjB(dP-3Tw&VJaGvmtIK5xvV$4ZC@I4jgWV|OriMGP~36xR2Ow#CmP`ku=)K{ zT{LWqN;!{Bbq%iD(7(_%QOhYm4l&Hf^y>S$nN0~t*Rh!o{J}Ewy5tpAVwyOHly%awkzl}}^_(;R z2eCpM&~#qe&bAk60m0gK_JX4Zigj8*aOOuqoZcmAgyr&g_+;9@wqkuL7@&8OvDW-H z9{o{zgXO8-rjk0fw0I(pJkgsBYu;u>vtdhwD#ywcws{S<;YISUC87r2K@E&AchQP% za*=%0k8+XRj~`ehzxYRs=hEv1&($&o&$q!-IJ-37Oqn8g19E$*4XmT|+Fw{Tka4pp zO4P*mWF-&FY_vTYM;q?@QKdW$Ply3sKdLmU$vXQ%%62wguOLj>lsNk8I=*zw_hw$&4*_G{Pb#TqBpTA@Y9YFeVcmd5HN7!tnp!{t|M@dsqMz>%H3e+g z(LdwUroOpai3MNW4=elR0nlr6kWgX1n}o(QWeW2R!O|+{&X>@oGDU6%a`(Wv@{oXK zTDngZC``PppOR-Z)tb7d8AQui-!pzzX;kc#${t>B`!!m7MhB>_XW%TDc< ze%z@!+iplBHNZvDgZ;3OdE9nXnI3Fx)B-{VdW16`->#Cg{YC^WU8jdwkkP+jZkDlB zG`FX?dM*NtoL&HfQF^`~)>`lVMJ1I_`H1K9e+6N|P1xVHM?CZ{7c}BGXmYbGXag72 z@*D750gTM1kGP=c083xpCMw|4+eB(FOPCW8k=4 z1QRb}_138_s^6i~@IR)y=wIbPBkk zpl``T>?We$z~V5sCeYCV)EUBQHxB$a1N-z@N=N` zX!m-=uNj6ta4_uF-TH*!BwG2>{xXm*?T9`FWAHER1+8+V5pudNlHMggc8i?Zf{A&!Y?(YEr z@gKl98F+So4|xI%^y-?Sg-v>k`I55>44U*Heax7*_V-{8>U7^B%ld`EFM+Fp(cQ27 z{3XzwQ6NQT)5R(9@O@G|gie|+PQgHCOp1rd*#;EIyd(uOrw{Ouv-M8g=oa)rzMB|4 zc7TV}M$K%91ZV|Q<dRQ0fq`iMrrj9i{!wMogd zAvq>r;~_9B)hV-bkcC0rT!%+!x5^>4+iD|c+$}6H63AHc${i?i>IYgDsCh)l*&geO=dz!-T7JtjLJScBeom4*i$C-1 zdBs2v=3&0Rvs9g*>c#NubsMim4k0Zw)OFbe>O!<+MZ3pu;Tv) zS$2Ll1)&&mcnm4erihsdVc`w=zxEU{scbt z;Q;vKOhD`s=p8QL4GE<82cQrA>7Q_A5{s1gi3rz#2#58{p_v}p|JSS$uG1n@1cS2> zoOtoxD+w2X5R3PfnI3U;`C6Mp!E=IxSEk@_E;w*#{NhECoBEPSk^7dNd$5ES%M`h< z*|{|m+9FfruC;TQOXxE)MefUXZnuPHyc{d{2|Kq~LIW}d?L&5Mi-fkx6uI}?xz9-G zMwue_ZaX*i6;WP}Op&|H&YdNp3uKDiYwg^7C3LM!k$Z`q`@4h=`?n|oxo6wC<0W*a zOp$BYxu%3JktuTP?A&$-nn3Yj8zgPofr!{ua|BKLdbCU$^) zYg>G{ZIk#l{}J3pKlj@J8E$_YfN=XprWI~~8xR+6f8(Z1BHaGYIhp&*p5b@SJ8FNA zXfhjQG23Md@f7??6wipgxQF+&Y^hpav!okDL@q^a?z+_JVj-!33AV;jIF8rbxUI<^|PuB zx>E}Ag-jt#Hl|ccbp|cl-$TO(R%g%}*c*b~E73N}6oSo76zpRmSi;UbUb~5VSJ0M% zIxT5X?jcI*HC`8)Qp8JTimV(qUxKndc*ZMhv8}P$_xf&+TzSluc9X9?=jc7gwO~E4M{s(wSjmi~K8Ud^TK!y?Gb<$=!1hD0RnR=69 zPOUwQ4h85rfLIth8KNA1poi4i@*Q4-0$7TE;6RM=?!yg-e#%@tBb{ap^^mgz5I^k? z9TyBmpYNxk;CU+^pw9l#=7phfJZCqs@q$r>R!3ujcE&IdIoJr$f2T{wqd?C;eLWA1#o?TPO91y^OcvKM>~ zv*}y@7N4;;eb?XOGsdQ$@V}T|31LH zap)?}bLyn~Cq13rDP29(doT1f2X;Q9-az|9-&^A#@ zaGJ{&u;vp>(41?gxhR-x33nTpOe1cdhBMPCcTaQKxX~ER>(gBHoK16|I6ReTE}pbj zOez{Xe(absUocb~nQwC__`)h3ihQIf{F%3pxaiGbi|INDl#i7A{OLraa3RA-{DKFm zp@^SfTs0^1zSJguCFIgkn1MYABht@G+n9k=KgL7kEO;=DJm$Peal|y}HC})qXN$lS zZz#MBJaTpd7MV>SCSi*91%RpVO>@x~n}n?NnMfh=G;C^pGtEV5V?C6+jZKxPy;wW1 zid1NGHQmsr_-eI~Ve1|JTJTbqgWH!$N}j+c`jVF@W*z3CQF;QY@@10P{(vREWEsgx z52fMVz^{|g6^$MTecJ&Fp*Z@3c`X>{A!o+JY?wvKcvjfm zchle_Jmkp13iQKbNt!Hq|e85>q%%E{C0B*YAv?sW^38~O}w^O zCo6eCMD^0u$*A68M|mhOZmWZ?<`tATkFuOnERD$EcQvn_Q2JVF#499wRNi&GhO*z$ zXg64RzgObCBvY9CT2vku;g3ApT7*xCGP&zGsq`RWnFDD5~8Iqn)U zjJl=&yoEB?ju=J@j)U8zetKgjwMohjnS#SA;Qe}ICVhLHhf;UEkW6>VTnh_QVShQt z{bDke9uIB~$m4DoKFk~-!(GQq8EV>v58E;*ComOuN?Ik=<0uz9C8wW|uv79RqwehF zV%smSvaL#P-69Mw+Vnc??2D~KTF2G3@q`}rEx=_HtomLTU*Gng|7?1b;S&ly=|c=kW0j-!qU^v@$FwPQH|(#meOYdiFPpH4gk0t3P3cXKaz3c z_cMSwU=jV9JS31#xg{RTi|x8oOF!WG0M)GRbqTzy#Hy|q-svROrsJ)Cy1w5B&!o|B zC0M$fP0M*#^~6aSok>mt?E*$yJ;}qb(nU?3)*PT0$P*i8bM#DFbHM1s%r9hsA!K#*q|!=dN3_5^-${c zuLsjc*+epa5Dqb&h+c3KQvze)R}j(RA;=z zKUSuov(>rzWLuq8J=A#?S(ZAhggOpL1$DZRnR?Z=x^Ak6Jo&yRGh#MI&m1?7R`tVm zF3NI`-;P8@9J2n)rSgf<-ow=!0>SBh98IVp7z)snEdEDRF;;5>*o+h*c+y8KX_6Ov z+D`zTx)}8WO!J6U{)`H^304leCC-E-xmwWni>TYxGR0&gZnePt zRb(%bDQXS}70A)A!bR>UMTuvi#3L(QG_}G*JjIJXDW+}MgUG=Y@A(Qj#iL3&#hYF! zr+DoW*Ph~?Qz@r-PXU+C$f92=*$_OP?3RW4Q0RfvTr|2$7V3UV6l%{a&#sb%K3OFT z9aJp~^-5g3(8H@`p+>cwS3XiL3Y}2lqL!zGrgbPZTH&HEfHl9uMQPI|#}3Icf>|v# zs|%(}j*;n-rg+>*7L45h?3ozMHpE!of6AwQk&v9@GC=MzXNGpyzy+=veL z)A#ITsXxxZ=4wyR|GXdi>Won{k-5v>T3`(c%*52}X>ktfEAEk&&GeAiX#KJurod}v zdMLdrNyxlkvPije`kwuI=qa#+vj9xc@6*prcIqh}az+7+%%<(k`3YdA=P8242AMLe z-Kp;8lXijOw~P3w$}gvS$k_!Fz!o!X*l8?N#xv+J zqjYV5tYVyYnupR;`-#%m%9M$2WTKO1f#@$QR4(0GHQ4SH%oGlSTl#Yl+|nOG@UmI< z;_a6Hx7ViA+p|10iWf``QnvKRnyLJD`4%oc?kXjYj|Sb70;KV0V1b4ep`STO{O3$e zE&%bhyH_qXzt*#j8;vpIH~t8)b2exwvef zWVM(#jL-Kmt0?es4tP}xmNwS_n!b9V(7q#0q_Ml8XiW-2`^P{_y`aKHOQg#$mnk9- zj-%wuDE2X#f+ruC z0naX=IUYb*TO)Yppe4Ui;iBY#hjQMlaFI6v-4fcbPh3h@8fKf=_1m5yiW2qwZi`HTVh+-s<~+flq7b zR@wkXH%gnhgp@iosOQlAXgY2gq^}vf67aSTi^Vq#{8t@x6RU1r9l`=y5jbF2+#nfV*dQ4`&&XidCXwd> z84Q00GH%*BA_y51?B8SRQri>^1-L2*m@#dWwL>(lIF%}JgBR+r( zBUsP9IWA2bfDq|8%Wk5J24Lio2LRWhLCT^5XnlTw5Lc8%ydik$SswCWz#R-|dXasd z+2R4m;6}Hw+@Ndd@(e)2U&{; zVEFJeq>XdU#RGmROQ-3x(Gy6$77uu+v&yZ|2Jd_tStSc$$Z9bCnl4%7J@yyfGe7rXB2rn*@ zv-Jh^bM!Gg!EYw^^nAKWA2ZXjP3Rwi;iqo~V3G0mCJ(7+(Kn3#auY;%jI4CggcpSk z-hqA-D_t~bu7@0{8(q})qFAjt8VmQMHoE8|ceQVa- zp$P{FZEIu-lYb26Q}!Q1=g;?0Zq@!nuw9i91bBK}GT-MnL(R6qyjg;kFx*9{!f=H| z7o}oMaQl1@u_-57kbgE?R=bEV78O>*Z2taK`;82?;{ANLC^=?Ux~SkKVFCvfJ-gCH zbDBLg{K85XZNiTgl`a}^wugq_UFo8c`0=kw7q#HWQb~Z(fnxmZV~N>gsb5xgv%sfY&oHvXc>e8WqovBz4S}OC&oz!*!+6 zJ?H775-#PdB4f`z^DS5Je2HCs8Q0k}=dr7^!jP1fvCb*yTh%}>i|S0bsX5;{|rBVg&cU z1;Rfqgsn`S;9EUZLcC!>~p1V8;KLGtOQV=YrkIv33lXi*MI{UzECe zJ#r_Pa(e#_7mCt5shAbKzX+!Q0J7zv_c*8gBG}p4g*=W)w26gS33w_Ns#|31yH}=& zZ_{CDELyuRPFS?|T?{8cY$#YOs>BCO#F#}FOQ-C@4>(iuB{IgS_yK2HfFFnrw_bt) z@Jp30a*hxc;+82&uLaL9DqZvfqiw5nktwrVWD2yifOcS&i-unc9TU8P*NeFiofOfn z_n7V4pP-6DMI7D^FA#&??Wq`PExFV~xYnT>=O7HC?Ws5j@{h|bkJ!%DPU1~wqXK7t z!@+PpwKy<-nR^*d>I%|dm~!Q181$}1h0!I9GiotrKY(ELLzg<>fPMhG0c2{r%7JOh z#fveHE_juxmooKzi#??3+MFiMFtq?nNlO`dJy^5=SFl*Y(C@(l|5DT{m#uuaZG@a@ zRW2HQxuxC?-X^8=SrXwHnGT;-<)U$yqvjkp47R~9$!xsqn2gPi18*2irvi;qLRBui zoT$e#dd`hiE;`pP??&E6BxZW7L@JgknDR9PVA{q^6Y9BMGk3n@8Mp5jmf8ij=2-=U z8+OTjL2Q!Y?Cp!T6t^aSI|Xmh$ty7EvE&~c(rLhz9#YlD`noZbeMSR+0piKzJA&zS z{FP{Nu(Wkq_I=3>_35ms9pIkLt#k*!NFmQqee2Rge0mxI4mj4s0L~XDhrXs0V z$*IU0SIMc!@~hDBFmGv>$oB#nFBQCL;bdHms=1@eMT>HT#`nq;danY0N0o~vUG1Tq zwN);97e79za#3b0n1ireGJncKKCV?FzlI+`KJFSIV^ZW^C&Ydc$e0TSI1560W!6i` zf<|{-BQ^Tv8mZBd*Mbj_TP5<#7INdY68XhzC9-lIkWpxdM1BRxDD+6q0-tV~wH{e0 zbnbPM&-&{mpMguj2gn8Mh48%L2QyAsB5}^MaiS8Z6AXd#n1%C$jngi1?gtJE&ADC{ zT7SJ{c-Qqig{OMWw6B)2?@@=_;wl%*U)x8CT2$C?x8eoTV1Q^Vzs3B&>#gRxHJNW) zXRW%WG(TF>9Cd?~X8sLQn#XUjOyK^HFu~6j@;4Upp*Kq8;Efh?jYR&%LcaAz$@xPI zdDv15xlJPfW+7KBmB`mDm7HG(a$;M?x?6xRw6QEQJiTrD$uiqR^?z{`^?#Nc3eK4pXDhQ^X2x+&^uE%tq50!Kju3zR%5uQY1V_9#g^yp|oAPcjosw^{ z-0E4vbl8G+{E_~SEU?zRyrTZC?I7O6ZH_X#~_!N zg6#eet;XirHAd8h2p(&(+DEjO*Wem%!EUq+@c0JX?s4vGKzE1hdVLYJO7-c$;cZ);I(b{gi@0_7F?QV1Teu{)R2xPeM8e z#uc5?HF-@ABZ^VCdc+_iPZoE$OyMo|Ai{eqqH&@wJX*BvXp37gK3?oUQj|RLb&Eez zFjO#Z^-vzHpaqEj;5VY2Z4$(v*|4DXx5M5|)<=rykCf)APFit?hw^-(P-I@^vO1OOb`1>rjUP%y@DF~keTS|Gg+GSVN1iQj=f2(#` zEomLEAQ7c3EqB={?xTb#%JLk%!foDfwcgjobT56TF$R$4&AU0j&%2fpTBoMmVRRgAtZRx1D0&vyaOdqY>7AFdwB9+ zN;B4$qWx$~dq}gRhcr7_nnxe#A&mgLS6b5Sh?iyu?|towG>6$zM30G==JNiDwz<4N zY;)zx9?}T#ueLOoi`T73oriZ2FXt_!J(1>cTZ*=0!jcp-A|y5k{sGP60V9i6Yu$@j;- zv|=r2i1=!W>ORgA8K>$MB`->(_|8hFS03_+ueaji%ZG62lN52vm3l9r*F(s50iC@S zwD_!aay~2z!b8!+eJBX)?DC*`{0G47B5IEBMW?6FN~g1}qVTXHv8aT1UUEhzbWM8D zY-r52ZbrA*XPjUcBcuC`N#C87P90!;9y;P1#q^Z>oUYC^{zxIr%ze(M4|_zP^OzL( z8JQwN*?mq*J8HD2basb>K%OX_B^6ERDcufb*z9!5ZMPgvPSzcm*=M{FCmrVWQ`$XZ z)*fk>VlI{Hw#gJ?+LQJry^6^n)Q}%0l=tadVo_mCya>^11vlwGVawLrZT`GgiiNIM z`*>GAz8m0;Q_O+O^V{AKw`@O&RgIkt8!s_`nZuH^HBnv@;=XT-yYsh2Qt?fa0!dUX zQ`9~F_g-wJ$zMCPO3IiRXCchqmL{Rl9d@De%Ke_cv5VG`D?bBev8^eJvDb)7&?Mad zvV23ld*Q%)8eMS~+^MjHpkftavF?g1qJb!YNK#}MA>PI4DSfX&#A0kQtSL^A{R$2p zbv9@D5<^ecaWBSk479jf+}tHGu7*Z*n{v|Zbh^xDD_=iw;9Ujpo3ZvH-$&Seaq^>m zRynOb%Da!tTWObPy}HmNGW9;&u#~Zw(xjL&8qvKeb7!a1>o)Iu{`mJg zEM9Ffo%s|@bS88Tm)w~OEi2i?%ZTg%*2Y&BN0B6Aje9~ zE7AHm9*6A<)G<#yd0D_<%uR{K%D`-6+9MCbj5ZOstavQUTTxlanTb-B6_1P0OcV$c z+W6c=FQ}YcICk6-@%f1!P&v8qubO~V-iG5to7o`LJ&g;7Cf=OW4UNr z&^ocIONwF`f<_z9qieo|an$k0RwVG}Vt z(KRedz$eiutS!9aV13Z^g=_~VTriwiq=kwv=;EAqq_Ht*HsW#=;3f)U;hwBDmxTQs z_ifBw2;eh)JE11JP#n|3dR^0Oj)}&#(9ry$4~yqcu&IqwF~!E{NHn{JV#zc$9BO7x zExlW%SQBc-@DeQ&7uX8J&4eAu!~)|GDm8tsrlXr-yu7NzK|bKb|iZ*73kQ8OEgtHx)s$O$W=Wzi!l3RC!g-+Fx+`bzubgYi# z&cDJcJ8{w&awK=}5(lgMtT9Z)D>TJ&dEJqmahcHlo$vSQp?lOYy!al5W@@@&Yt{9T zz#lNjxf^_;--L(ziZ3*6+a-P^h;5co(0{Vlj9%K2Tp(rdwchOB9(}hX+1n>v(W5(( zqkY1jDFz&l!AQ7LT*y}#5*I}{lA}@{d>(G6MsAZq<@iaIBe}ItbQnFLX*HLI>mr4Izh=M& z?SDnv3&~W!5k-YdvR*5t7qU{mk`M)U?~*buNRuP^UWbUgcnL+J;r)m&(IQG!ct#iY zg+?2}aD7Ot3kB;NOv+sTij&+8nH;QJ-b|x>Ix*t*HwH#yDjW=H?)u1>{IQ4SAAXE` zl+obRwSZgK>V36AO5l^+EBpMMktgbrrf6<{aiplh7q(04m@P^a-v%sE^6H67R{!dx z3$aczTWcob#lffWSQjQL`QmpcWj<=J9RA~1d`!sSK$8aUa8lW$ILb)V7X0RuY}atrtFgB2TQRa6fU!2QwG3lw+c>*==GXe9TrY!E9j02 z0fpy8SkVX+y=0X*6iF0;ZfjOqz7j#bW8bZSwpmhl*wQNwxLk|0+#DS2&nnNLNsoJkk%Et7OD5U0Hraajs0z(! z3Yj+BmN}W+kl{e~^5dKxPP*c8x%`NS7avbpex$b!RqQAc)=bKL?p3GFs{O4!i3uR@ zAJ)v#Va*Ii^S!!O7o5-MK9t`aN_H?x^y}kJ%C5!NbO@`PZv6CynvAJojed9F-kNw0 zg@J&cA2d$Y^av?`B*j=TQhup-?%|}2Q?h+TAkD6osb8i6k-DRF1kNGHsLFiCY*LE) zLRtL#A5K9j)q+q^sxlPxYovUTq+}cXlW-3rv{YD2+*HGE29xCHjPyss;oSTwxV;A~ zaJi2HNzYGC--&;IsUhA4S1w83S?m0&$l;{?R}nce&4i^4QvOd-vZKT~?jWrb=l&r^ zoC^7c#+V~WS<_!|VUc}+9Cc{g&hiu*Qr)-W@ZGv*(uV$rEFLI+ZtQQ<=KBHxJs7Tw zl<0amg0VqFCuO*jt>JX*)_zju{4rtL(~6;Fhc$CvM4wHB(BTf#JUEjmJQSIy>4lX=rKO~- zQxwa?L?(pSOsi1#C%n*s`5X}r&Sx8%J~!?WuXPzj13J?vTgTb$1K*Ig2Cq6@(_+ky zQL>uqjBaJ~ncG4qU&S~|Le51h^)$$EfZ}OY&bKGIg z%ts|?y5T z!=}D)z^4a_^q{HfK_4lNTy;iTI-*RNU8gR=z!IQg6WHxsAsARWu`$VbPaGk4e0@ksZ; zI(!Ybo||77Hs)!%(1VoM`}4?ZM;A=rXFsucRn87Z1wT@@{e4cMkYHwz@=J^xJKS&L zh~Zv}7z$a-5kbu$<)IW?X=dIo4*%{Y^Cq)_msCi(FUEH-n8h5{s+=eP7C){WyAhD` z;a+j;>G-?^70?30r)Qe4{8rToxc3Ys=!N zK-R)WldfTpN(+#3YfMLxoinm77}9d{{h^4Vk+Q#%T}L!5&$8+~-50HFHvGO&h)?U0 zk`dohV#5$jmiEP{)M6Xf%t^($`IVYktoZ_=AS`^DqGZO(l@Nxqp#KwH*<;>zr3d*V$;}( z%_c(n3kvs{Sg~*i=|!%1cP+iBh+YzP)ZjC!qfuPJ0^6;Qv2+LNNhT25g@WPPlS00F zgWX>D%@YsYo4f1i=u~U{B;OYbO`0Ue=CxwA#OqCs?4%6?V%pmO$KjzYoxU!2%iz`4*%2Xvg62`0^Baiq+t`JQJ$lXXY zP_R2pGY6eC5bB%)xG-ZdTGALbn=SpHm#bZ|*?m}>SK*si6bXk(DHjd^R}<7>ZCAp~tX7Kmu^touh_a`aYVtQO;jnRet z@rBiEa*2bt7D#!hXJ`>0!ZJ$gd{~{kJXvUI5IO(k<&P_pPr5t}w!f0^E^yDKE0bdq zitor zmHDE)SQOVOa|5;Vsrb{w)8Qs1a1VD>kC+iUk$>0mg!RN&ED{?)=u6;$qBLu*5p&eX z#O(TfYmw2;tk4=GriPg%DPQ(P^83PmEu;lFELy8~PbEwKiMgc)DbMUYX$xW`6J`da zq{YqY+=fQZ#Y@sSsybtu$m3jcCd`CZ$#3*ivNfXH##@oQH(V^0@N(aqeZtlnt#!n= zdg5bk#23ICZ+?aRLCRPCrS{9e@-Y+|u3%8lVI5%hs193^PT zoDX|0ZSqS+a^=QHoIhW%&-RmYaay`)2?-HF7ttw~{q8_s+b%oKQZlT~n^a5y8_fp7 z6Wlqw&es?WHH*F?+lXMlLq1?HfhFZd=Ap4jXLqK4Qx8dJA*HNmK(x{DZ5dK>cA1Zs z%Qq)rnHYlgQj{Tr8?Vk-DA^%U674{DFP9JSTUb`^gB#eQXhd@_kRF2N!5}IvgZnq( z{)QNYU-OevFGPl&1W#THLX6MNm!lRukaCk~y+01KY@L&^3=+YD2rq-2 zv}em8h}cH7Ne@PvjFR%HRfW?FOQ(=>trFjuBT>3esXa2CTl00i;o&x#4Z1-&|6S3< zZ(yKMK}Yzq7E-IDq`Z*KOE|5!rBSvSA^ssyb7Cu6k*Ek$g}hD`Tg?h<^Qw#4BBb1u zB+M0Gd$Hv;t5?lt4zblWa0mx+eQlVO7vsiaLWcNSBqdv+WY~l~%Gzw}T#&fw=2&@k z7In&vz0E|F`|{OUM5&*jH-tzRxT!KtCZZrvQqIQ}T@K*5-%exrS%iO%U*C0+_ieG| z%G|OxxQK@5Y;e)9k9){*>;@OLNRl1+SeB#h+AM01hY*X@Et!0W@?s+0DMeW+)80gy z@ZX(8ivCxK1a2%$))P{gc8MGlrYjyIgmJ$k$hRd5Gu{yPnIO|1!U${B#L;rm6Re`! z-Vt)fitdbuh@vyz6-ED+SoGVn$c-}XQ8ZUg&AWnsT-CJ3bBk4t-x8@qlEg&nj)w@5 z3N{Fll9U8*nQsbfw8*qKkyu6C$N1iZrZA zh;*<_dlRW$a*MN4S3I|vNcUMHx!;cyfF0Ny4-q1@zAr@DCsCvhDau-z_9jx7NjB}|P( z=Bi=Kw8le(FrAY8z(irXq!3@ow1+UFHD!DtiUutgJ$SWj$P+#ga&pn+@NIBW^n)}y zfJmI>xN3uo#;x|qtTvg|fMzupF)H^w{(CY0GV&QVa;HS**cXS)XqVV%^r0vYmdXu- zPdsuQ=XQ{!-nhX<=GmeGT4f4pyCG6LldtB#9r!zZ?FJXUXA`cM1e;_E!Yv@|Ns{!W zheTQKb3|F=WeSoXKoaABC?mf!IFlM=H(|;&orwD5ijacq_N9;nTA2ANSiJmJ8SC*?pG)T$gZHkf5-+Lm$CJGt0*ti_Zb=YefBBBy{@nfcAUe!OZcQK>B8i67DlmuLx3a2- z+sJJRM6S3_m=VX>2<UiHx~$XoEM4|k);+mJg(rra)Km96Bm%EgrGI!A6GYY3Iue+wovq9 z!BQOXI@Gq*p|I4zHI}0u3PX{q1`Lx%VL*7<8hqpsM$AK2#~K_1{D{qy5Azg0hbu_B z_*~$aq~vtE^*MP7_Zay33&@Tl+tB@V7&w0R985*2>F>Fy_#z>dSEg`lwVFq6J~r&E zB3crPs#fw0qUt}B)2Z`$tE6gFLT`L3=lP;GgXrRTnP)Jg8DO;adEDAdGy`mod%;6_ zXUT_4Ua*>A5Ah7cX4iIAJlk4kTMM>tfK!mD7HqfS;Vk*E|BII7wLC>reWs}w7gGHL zN;)0=B3dW_6R81!(=C8FDp6cWL4aF&fG5&f0N(*{aUuD!w)YA@Or!uFroZGNXZy$S z(nw=tBwR?Z@=Si(%Q&gXpo#Q<47&7XbiJslu8&1c8K|jLKqg!+^evVt+%=XvSwg4C z6uJ2BD>-Jo=OXtfg4x zf*CTS5dl&37y?etXaWf$b`2yTNE->*l?=%wj3hI|%+RhaHn5-uK@^D^5EQW7=(SRCLB8_9bGNu7B#NIMRqE%!dh2g}# zWwk67ujBPHM%Hn5mZefc#igOJUyG+Fp#@(cfmlEL@=k^LEk`e`3#cAUvrr zHSqg+dsbvqHp2a)Vo~6|BrNQZW937qQqvY;1_cm@1LDZoDq{L2iom8SJ9hEXbSnBB@0`hO3k-)T1EqeJn?hYzYnR(iMvmNEdJ|B^zO+m3CZKu_Q3{iYI;Jt3q~vZAqzf^M zfOwPPCh#%1F`L{Ker)BXySe!ZBBoSN9isDB+Or#P0oz%+y%Rx%Y3*J zwTW&wu;90IC8}Qu(j)W;mvDZSP0kv`M1Ay_(c{T<>i-oUCg?Ixuv#L&NUt@9KABE_ zq?6-=wI(%e7p8m%bh>7(N%tTvZPQwlddwB&&y`;o+T9qb9b9V?S)JFJ^xRk3mKaYaOxH) zcqqBhFye!^g}f4%x(cfGRryIq6d?i~{U{j=4fK}Ns~Xd(^XhEe;2fn_ST7f>M&EfyI=LhCImeVB#*3J!k5;p${*5xA zg1$B|2Kjk4^jjzMtYK}P{570wSy_a><2;vtjmjo!|B@opMYfYN;wXBh^B9kMw zB!e<{iahnm^YD@kT7x`~FUg>jzsaVo1xqq$`ZpMzL+AHADdhN={6d{+n#_ z90P58pnRqIv=Qq}8d@g`R?065eusi%kpJ6k%BomrQsK8ig!~uI7x}M}U*!K9#>M$L z^E6~`WmZD%2^J+aPfrVg+M;Cq3i$Yofc(OP&Tu z6e;b9=nBJK<{vw;VyJ(Tzml$m|1173o7`o==rHWdx)M%vk_O0Ys^y;6d+asPx@YXb zH$c)v8w_lk|LnVL%84K}Bh5wo5W?&Wb4DAO*A8oOD90beN*lK6Q?5ObQzo{BAr)J| z&R?t8JOWix2a2Y;XYI5#8AcB`z&9Byk8x3#8`fr14^7Y}rYG(8wb}8@(EgbM$05R% z{viF6pHU;ngmosB>`W(!a}z9b<~ozst<9#ax$8{ounv`Sok^FkLp5g`RrHGWa6YA= zsiW8IRBjg%+|Ck=trHU5&JuhMAx$uCXNtdArzE&tN>B@JI}456Y6r4FyOXm$@_jZx z8Z|oVFRTdBE?%qp6_LScl)g^#I9+u0dbm$tv}ky-7~he2fO=-@_`-4i(((Q>YUciR z*A3a^#{MI@fu44A^ddx;1rsBCZB)}2-y|(}x@Z}~<$*{vMthQAi1VAIJ|O>l*lm+* z)KVd_H=Yw;w-75WO~L~scWq)1ypvntz1+J^+bqz~-X!eWe|9s5o6YIu9v7s0`O()&&?mSTrSN*r4lb2e_`OL(mk7a4kzWvV0`6um z`QD^)GEyaWBBBWx4v}N|dXt(Y#xOAMSZ~r9+p;OEWxYuw@Z;cmlOEreO^%ceCgtr( zr?Et$zu9YpN$(;{{|zQ>-3G}xhHo$_W3OOg8UXq>nB@E+n;bW8Fv%-Z+Tm8tyEd2< z{2`mNW^XX55n0gtso9%Or9_mD!sz|9+n%jDMy-=PNFs7PxxpmbXG8k(29t(vmq-g_ zic=wdal1sCu_Ie_xvKZILOO7VLR#NCDS#xRG(2rx^iLu7aqRoAdx+YM%MSY@Q9nPK)13Soj!n2FR@%wHw(dmb zY2oVm6T`Wr8OFelQf3tdN_Vcv%X`C4W^*#nXEig-`ztmN9=QnwfFk-9C|-%7X9JxaIX)=6`a zBp}zfhP-T#gnZ3j3Aw3t@tVB~a>jvHNdJW-X*JdvRpD4^SR7ySU@Mwtyx`5wJjZ0z zokA$#mNtZ5p%MRsrgC{&*H?WiR&|Wd& zAwr+*Q&2^i9C)nOdmx>F{teXb4<9<7st=^|V$xTrT$F}93nihS$uIK1g^G#%D)DWZ zh=nLoR#;{HU{IOb_Nqv(e%ns^M~0D8M)Z+^^FsM8*(BDdKH`4oIE3-wFnz?V{n(OC zxqe@yGT@KK$Pkg_tUn;0O#PJGECt|)?(-snoP8okvzB` zkG`rRGIHzz+;WZuh>`l499qk*>yHPNghhGSa6TDpMOW306MFG78;qLy>2$`A*&+eE z(6(^YuphI9YMP{K)+yEG{b*N>5HEWSwWUdMlldPzR|nB8icKLp$fCUSM|9U!7ycmk zgwFnn$Fg64P;wZcLmch;8Kd$a@t7&+ zp20nd@jqviJ0JreArKl*Gq{7(j1cFI#v*h-cVAq;pn_R*J5$q-!bT6FIlg>uXBzrT zHf8xRGXE8M;EytX5_z1;ANBZE`lC|(fIoU3KQ?VJ>1UJ@;K81c+Y!n6vxW2IsGR}zl#y?zMqGL?ntm89Av7p43L`G)D$Bpl1T9HGhT z@ZW1J=32xFebzF8uHWT~RxJZRgdp2Cy~wDC{h5tNhhgpXKF7HKlAX$MEJCLy_VG`} zfQX#lpVGx*BrQt9wy9NrOC*aJiAffbvrcFFfSHaV*0F1YNi{zUO=4EvO5SMFmt>K3 z{6>@BGAtrT=0=m6WF8N+ec?uv{zg*TfQ=@N`%q}>9{Gi{x&Xc`t#YGDyJf8NBasWK z7jx>YjV9eJV{7FXshGPKkT3jLlqr#4q+Z5l0r{~{MC#k}i_{UE3dkv+iBwB|kva;g z$_%mL&L8pg#Fy{+n-bH8=lxm~77fK(^f8FmCc{r&mSjmkxi%RFI2&PYacj9X{UF&Q zp6tQ;mW(q~ESOau8EcbQ;Dk`_t%(8nE$*8mB_41`%jk}jN4mObN{Yo3KB3AnR@lp*$uQoGR7*Sx`X@u%hF~}nsPaW{OzfT%Y^ftWqVBI>~rXsuSbe&elZT79PuRMaRS zmdWo4M6?;U|4F9~^a|jnE!${n#7;>7vi*r{UvD(&^!64}+6L5~RRY=$`9(Gdy6nGi zG%18^N3lxTvGhutY?5DLo6X5akGQ~hv%u4qwYR8;6!sn#c18=Lq{{ZPw7X=dTN-Vv z6S=s1tg_NysgY%_6ATEXb5e{P9Mzq}c6m#>su6QiP$NDZ6rgqm^v@x1##q zoU?SPjPI3+7)(=%r0HTs(Nr$YMUG&rvYMy@_3ITmsBF)q$Vt^_SipNS zERrTVmFOA1VI${wi+V)M=s7m(b4wA7NW*!74X0q5h?55PA{*GCeG{h90V%#+c`U~q_q6(KI>Eb4n zenHX^EW9`ntf=&BO=av9D1jnil-f1$a@sD(#^t}ihB+@sTG`X8ajuSVTM1Q7J zXUqzv5_x=xAadNe$t3TeqR0hYYj4_Q(zz1l4EcrJ1_9uwn@oC50j!n)O&h?pZ#S8m z;;pvgSV)r5qm(!Khzrr)HmES7>RBkrVMx7+gd&2I}~^n}yY-3An9TXe1^=6Z9!H;`3oAfXIsM&1N&hC&g+gMpJ>h+b5_s2%!WYDNcAgrkm|7ho;lHbCv zgmXEqhebJft)}w4KJDqY9v0JCdmV{98me@`I$HhCC!-SVBEB~n zl`sHuX8a|MtP0`5r^7kawg_VKBvI}lMKMBt?R4q*qsaRmLSdQ=T=%;za!;V!KvKQi zqU>>W6XTMq!xB&|;`dd>IaeIl|0a+L{Y}S|57Bf3ynpSMyiYg4{Mp%Y=BFD#g^D;9 zxxs!Fl9~PK2517Gd1MRZBfwPU-UnzFMY)<$K!7x3>z~*|G!6z8) zzdg+1Tmj7KRU*8H%Wd(X^46%x2RPDu26~+GZ zEy@vhav=(}=T1byVJ_z2TL8g|TFkPAI`YUC$Vb4YsKqQoatF$2DOKqg~3% zKeHFwT;Xd)OuIa(&81xW83g70LX=)KuxR*kp1r@c#kTMpK!0gVis&zu{GCp!D80f` zaOPX&Y(jjg7~{NZV2l&UM>i=pgVtkv$mEVW$332e`pK$R}h$Jz_Zup}!D|Cna!t3(x5A>C0AtO(QpZllCRwBR1 z_BLRA_|Wk>(JK?zOZdM@mCwe%fH*LZrYs7QedLudIwVurOvI24dLC*}px9HJC4Y zykDlOX}x@*!7o!;gO0hIO(H|EuoYG5`^_djbgpJ0IzA8{6%zLgY1*3{b{_r`Jp9Ex ze0Z)!ITOSI-oK2k-#h8ga}^(dF=6P3ClcYzGn`^VshVdK` z&x*a=_NSY+ympr0>BQHm_+B!nUTy;ucUoPWh)qefn5W$jAVM1c;v^V;-;0#tFJ>Eo5x|-20Gv6O zMM`fn7PGR*Et-`P;>xF`Nx}T~6uij7KqSgT+9jOt8APGJzrexp#TI%aFC^hkr^N_N zpqG-+5`BI#8vJybYBi!Y>|_ao#eie-e82IF0sfNMvS0_mL;K$ z`}h*+s^4Oy?gC6{(Ob;YodxWw-{J~D&L$Kt3q~;re<$hqR2MBSfV<92k(K=;2JtCJ zdR62~jvU<&DkO5`Z!@VTMJV_TD7bi=Nxvf1aoskP8f8*0l4`b@)T_Tm-REs%E0Y!= zmw@y=OMze@rZu0@*=_+pCmA{7%B7!^aOqOCzeSEkfR-vqm;yZQx0|%IKL|VmhFE0k z=x~`Y=GWj$y4ca&nXJ&B(fQ2H%O;Fk;IC^i=Z6>(l8)l!aM5m5i=v*CG(=z^m_dl8L)efwp@yh|=(?`6wxZ ztK_#Ek$mIbQS!Y>jj|NBqmZN1c9SyO31Qq1VO+A^q>F)r_sZOJqTsVeeo;Bf(KjP^Box9$S3fbzwF6P{eO3M_O-n)g zp93v&R=0DKTVA9)BL!9Wqd^uqTM%77p)5!baAdo|Fb~1H+Lk}qJ1|h`k7_dWGVSDk zk)kB<1w-%+w)Do(FH*4Wm~u;f!B8+z>Z>F}WFj{&AkqHh%rn#J-och!O#G4wZWxRy z>|>m07_lq3kpXwU%pzw)Ja#3=K6RNzZoaKdgug|u%ryA2Fnz_BAa@-C*M%sCwkiet zkS7dbM#*vOc9ZJUgau!Yxv@vKn>1&LMb;DBO*(C;Mb_f&CLJ1TQP#hB3`j> zNpn(1BHD;v5>~7q8fJ+V>n_6q+wl(ShC@Ie3CLe;H);5Ai?TLvH)$GYVjHu!N^zkV z4(~T?z=%khKcXd`+1oDp%_%${*_?v*X~}TaK5gb{$V~{#PCU_OmQ}wI+LA=X-G~S= z3fi244W=I>M!3AqDeyCfH?GdjN}aGSGB&56X(2IDvzaF>(Iy4`bZ6>Zw~HR|VjzeS z_`o1Nm|CChqBV%Zlmg}@1_tS&)RC5pPQM%(V^hO8B#`>H<)Z0`!KJ@wEEe%p;e#XX zt(NSrI!hc*-lr+eYb*2l(kNI8QsDP?FlEP#hEAN+-bupVpW$h(UZdeg(_w0X7;XcR zm($TwehT{oB-c2d{0W%cY~ARwrd$4LD(y<4(TC z^HN0eKWE(uPRb6Ar^nl2gi7S3jtN44njdt%@oH3Bq=3t37|0R^!dog+x(ULe1)M5c_iC(13A^VXVF(@|8Sdr*mR2w>&amN=Hv=^70G(egLRJOSiE%YW`Qy5;8# znVkfXY`ol$3=hULq+!&PnIkB?PJTh*C+O*71Vm!ku%oBQ^OO7{Pck~PqBrbl*=|ya zMD!vk>9E73&q3t>R?~SBw~wyp^$GQTeyWkfuA1hv?WSHEA6m_43oN?ULf0{mjQMPO z)T0vBaRFDyw2EuvtFI&dyr>`XwCQk|jI%WQ%szr^(Jh{<$(YMz+-DBeUhzWaw3A64;1m*&2c?BWkKi9<;Bk(KO;H@agD~r|mSt%D zaXq#s_<8zBL|t%$tn9U1WB-X@m}*loy3875Q5xP@sh4Cn$S>^iN!aXTn=udl|6p{z zHOmd>*jayAYx`lQ^wKeLqkEWT@$497bPr3TW6mEDs^2(7E?#W zpEDNg_<>44opef$i-HJ31}CA%{?=G-`0{u+)=dpl+VhNw2<|r2G>!;Ab(Anp2#6DW zJeh#tZl27Iho5RY?5ywP#0n~P!+@fwR2pVCQ1?>_cl&32Ir37&SD#8f#!wPKSK?IAF-+gkA<2F3=6m(^2 z7=S-H4l6>ZqH!fJ%8d^IQl7~%S0U!X0Ai{|%ngWn2{9N=cHLo8lj$TVuP3^3xjRf+ zjZ{b99VTU-CX&uW($F0yop7T?o-21?H0dbaVUky7xC9xd?l38QqeYa4r0PqAWh|Co zpuqJbob5wob70GRwS+JN5HjCSrquCh(XiXGMJ9QXgx%L)jkm<^>q{yvvas9n#Y&4j zo3LG9Kr6@;1Ye0FfVQ<#L93BTS0M?|UaFGNjtNR=lY)uRnq-mdQ3TMQ4=QNhZUWH_ zNCLE}AqnkE`~Ynihcz@Yic6?L_Nu8WQOQ6dG8}NbV~0r{Ww?j@LWG$>_}C7UVquHY zzB{fnb&;tm7fw5zh+KP*>r9V^VKKS}7x~M56DwoI{KSXG*Kn!A3k$+hiLUXxd`6obru#OffzsEr+4Cfvs2|D!+ zg?#IurY>`kq7oaPyBj&UaIpKOy3`sB+?Fn#G1v~JuYl4QP&!2{$_W$o z#RAdLh()r`3{pNM=WwY)1fn!WI5UTI)JfHJ9MTl z(YOXQXM91XriCuE(=0QH`yM zS{P%Z+I@$rEU}l!uTg9Ly2GU0i56M3(y? zQv+*|$wsY}v(u!T0QBsgCOxTv&Xho(%IHkhGWR|53!o1(P&rHt`eOr&JUvT8!Jxl1 zrlIE*D*^T;qH}}*&w&6xo~TBJ=U{IJO|mGb(noa8!rFAY4xxw&&ErtiM+PUp&q16# zJO?7)F-a`f6orC*8jC%auF2B;c?ZpP@pgz8gsu~liqu_t`3nFfG}$88>=A<4HS)_1 zT_KV`n+&Zut5uoNh)kaX?_G~L`<|Jr(9rletp#aipsO6)@9vsnk(;j`j!X>4M*5>v zF(ubUZ*$(v9%vWPv_8T$@+V{#K_EMEDjFT+L7Z5Qe0nMf$UI##GS|3ob~+9D2L?Nc zgT&?=5NGfo(ml*)STFno+z=~v_YlEz%24rzny!Xoxz1Fd?&O$a7%u8eBZ~53QoT$n zKoS;{a&LkQxp=2Z_~ZTCjsRvNkrD1jveW~n=G=P>e!K9Kp{#){jB#2 z3g~k21tp6Bobe(bXDgG*_412skE(3P-3%JAr>I2IP=h3pHxEf|TPu$#dTU$F8zl%Z z*k&gh=>M~EO@?u$9lY719_3|p6RbF+>n#@bNYLpm&}o-q#Ix#eh3fO}M;P79TVU;( zFWzDa`LB@`q)C3k&P|a2QAo|#B!)-ZNgZbhi~(RwYkzBN%5DT@8M%n()3oETpQcM0 z!sTTY0Y91ZZiSt>F^Y~&!4$`hu)eQuRdUV97P_6FTIqccro*z95Wpvh1?E2?rQ#U1qe*q6_RS$rcO8!{A8X zyzyXOd%s0l>vx*ebq4eXB{#`D_aG0HTs}i8`EC4wk~eZ@C^;iX0K6ZW zVJt}xfFSletp#n@wUTw6mbu};L#AE{?@quwb$A91e83`p?KQxFS%cOYf$*fh*Mx%p zLSN}Pzo+L!u~w#lWsDP&GHUTC@0W&=^N&)ZFO3nsTr}eWY0kMpXFFEe;2xoPUQ@E}*+<3bmfiBX8I+mpok7lrElNY#ynx^% zEWgG9f^{_?VAFs_I%34^PLAkR)ecKi;bfi?-5EM)a zmHr`B`Uk7Dbe3MS`Ga*bZI;sLAFScHMyq-ENXOY|G1&(0;HCT_hzO0RJ2?60*;1t~ zl6(V;zj2l3^=wn6Qy&2{EWeVTLZ!)uRB7!aQlz#4OeO+UE0u0$P)t&dBq>dilsZR}R42J2Fo`)W_$Xyk!h|&qUzc|VNm?b$C{a_i!0;5xBtKm-Cbe zT)DqENs7QFL=Xv#rDM@2IxtUieJm1Qt5qu)gS|ZmM@I*vVSj0$Jm4?W*t}6Cc-*@@ z)!^$*r{Xd|=MxIyR5T2-@6UjShC<=7zS8l!bal)Qf>P3o|R9w1_XvY>+AX>dVegvCvtbOb(;qWCEz zMMcSEJ2f>)f|?q>Kk@B4i*m#}Ni|7-{NeV%G@JDLB8#$yHJfzbGZtlC z(QMKS_%Q}~pFt^DLFQS4(tMPHz<+#33fwDGw7@@n76Rw1K~I??D85jBVak2rsQGHp zG`t7VLnh_OFK!6NHKCq%{AL8_QE7c_dBnw`X}HFDl#CuHzrfNPSj3giTAagtk?1&I z=@vW1xHJ@2@NLr2 zJQtm_)S{di_@i%`nJ5SK}45W|h`|ew^`j)*dtM7FDK=obv zHVEYE8}!E}ha%&LPplg2kMs+cahszMhVN2{Vbx2p?~Z6m^4L5Vef~Dim4(Szk_?|t zxmEaywnwP7T9UIAlHSKa|Nf*g30;r^JXC+=$S1%gpcL*w0cTHX`1 zyvMXOzk@MYnCLyG#r1C`pZAivzu*oA=|7w_>fh+!(SMlGHUGxmq4S(@CGT_ciSMGr zmU%w6ji2y6P*)&|4ZwT85b*noqZ#i?j_PEJ=I9F)YR!?lOhz%%l)%s`!O$vZD5XJh zuu5pSm2hNZt02(5cG0X8kk&DzYz=9hfaI>EEy?H$Ue$7mrOXQjV}~$g zdVJ=oR3xDfl3#QPp1@XHO2akXWit9^`9&rA1C4$B)^hKKPA(O{wY*x!G$g@c39k47 z*uud@BQ-yn5872e9p|};rt?m`vp!UOP3NtclCSA0sJ007CS!UE7?Uk*eqVYt1$#m# zeuym*p^3pVx`)q4KaYqYJ10b8$It~62>6R{ZSEq6=4YU z0mgg%3a*87Qs6GCSJ>u_MI4X%W5P0~GspbUY*L*q#`?9{q)R`s z$V%E}(#%gR%5v>8sokd*Wp&zRQZas^cA1p>Icj&i zSsP6aw&Wop2cy=i-)PePpDW^evlEXKR~QPmdUj70&8SfzfGl9-loADA%X86-8dzZ1 z`gE=Z`D6sR5~PjTWzt;|`782^@>TH)7!8$b@f|K#U|*f+%mTP`|Y3P8rgc0gTNuw5Y7`Clj*bP+WH9g*sbodkq6aN~_#u(i^h`^*iJjvY!q+zHOIDsb4|VQoV;sfY)F1HU9;V2uKDS%x&_n zlH0)fh_5i`DEeYVw}G4o5e3%ydMS?j98sl2x50>=tCT2i1HF8!Nwoo~ruA8+=4=Ad z!QM%IG-i5KQGi|a1!`5X09JaXhF8E9j*YG5Vkd97$I~<1ja#=$c6u@|X=QaMuZAeN zl=l)*iZ@@O^XFYHOI5cn^-#Q2y)1=UNiMZqm1=5TD&=e75^{%#h6bWBwEeL_ED#E6 zHp6r6>_!EDc@5hVt0Th!T=xnMlEJegs6Z128*2qSk+Q6#8tg9w1A#D2=L_UVt_AsA69VB$ zeL}&?shZxJ1iEM8E=`;+#=mH|Ns8)_^NHpFMkVMD3wqzzHOR+eY218R-C zFXL7?GoG-^>@kh|F~)tvI)(c&8}7V*0{2`D87F*aQb}u+&hO(W6Q1an-bHdP zlpl4VMJ(_B-{ZV$C>kR}M38**L51Lpj73uX0jlT4hbO6S~2Uor(DR$3M8#y>NaqcgmB3 z87SY-$7VbX~SU zmYE^H!254i_Qkf!w%?@iz9P%ef;iq~s%-wIHt;^6%UZHbOn!m)WmWc-w#sH~hV^n? zwV#9Hb(Nq$2B*M&N7=AX5)~-XP^!lI9YJcX4Qfp;!6c)BaDkp=@1+F@GbI^ zL={o9Mb)@N>`J7UIV8;-6(l&ZivDEt{c?*%+1i2CB*A}SS2R1l7n0!fZrmzYD}HAP zbqI2$5^HxWwxKH!io|62z;+A$(-0BOv>yp zoWiLXg{JK`X;d?$*bZq*K;9RWpg70WX2oj_U&cv2Knd8J#S`ycZrx?^Qwubc=QYzl z-lZCDtY#O-e5KI+j#GK!P4x&aR>Xy2H zvmIyQacXZ43OO*hS5WBWY<-B-lwMkSr;haeKLw~2@(WPhW4XOkM_RVmBF}xDIud#2 z;D4*GQ%7pI&mzzB2zGxB|J(Ku97*c-jRZ>YO^a2{E!8EMhKdLFXle^Lr=;P0T}-==931 z(5nzrh5ITbyK=V z>;_wt-dB@c{hFP0+yjo8#@CsESAMiek}jIt-alKE!zW|ub=KvLKPz5dH~P-iyu7Z3 z+m}9KP@nuP9TApkw#p+VQ6GIOis|DcTUkWpz~pk1oE; z*d~{mW3jgg&sYt^tB;b)rZ*I~c#d_n?Zu8scF| zbX!6ZGI%R7Im1O*+ao>}^82sSSzHC%G7m!jhZA#UAT~}-mMVI@Z`tW8v(Z%sy80ay zcigFBBDNZ_np)2I{XtD_87hsYwhR=Jvkuf&k-zvaCz*%D9G4%R(s75dZjaO;RiNbk zhb+qGO=-na9_3r$PMsM8z%T?7K)!DzlmMU{Is}%9cZn4Ya1w-!XPSAJ*eSosU1CxE zz%H>lzacwI$T|bP+ADmO6aCS`*c4As{;B0cZ`qRH1L+?b|m#l1}^NbPeM1K_wVdkrV!O9Enb>V zHzTaav-RA%9rFjS)bkEz4f#`rzfa9PGo6A63)(M@geHdNHLM>e;rgSQNGu&kpC_I0o0C5N)1vIMVDw5~ zAjWxlC-lms=Dv7_H5S3e#BU_wenjIh>{Jbf#!r+$V^ip3=CU-NGQ>ZXK4mt)|I4!7 zW1OFtPpgejm!y+Xo88klF&2tdg~yzCHocKFk#9HVnr*N2xOE2MxviRGS@Ss^YqxV{ z^Bjh~p@n^w|ntYKsfbduXw8KVZ4+myfUG@W-IH6*=swu_8fa(a;y#kVCLpX#D| zF5ajj_b|-g|F-bpSaZ^x-<9eJ6Un#J3Z*6r0@=o!w-&qvCn^_Y_G zeR~)BQYTf*qz#y5mJ{BW2`&JcC%kWdU&!Mw`L#`WqcO-B63@?#!7nKu*=riosxcUv z>Y*&Gh~AaTI(h{9iii4Ji6mh| z*hlZhOmS(bGO9VFF9cAd_B1|Y?B*j79Z$(|(eF-4Lq}N5)h>_l$ne-V#SaQ&{3K32 zXn!q&&b*=G5X8w4_RTI2x%t@c$>_$s?DCMa4jIbnG#rAs_gD`(X_yleiKn19QE{Ay z$eD*2(beyc*!zz25cZpP2h9H*=OJe`a)icH55#QaLiLF9RR(-f;$@n&3=cV*5E&Xz z7V?#5pgI!~AJWYj?j&%JKR)0v4attTHkec(le(jmjm5Pk86I&=b_af7)8@eAJ;aM^ zHy;n6>?0CnZ68i4afh|o#kqwZGTcXgK?+u8_}s!%$9u$SR7<9+TnXnE{=?{nu0&Lip#^CfxFP;Y2w+wsasdqe2bMtj2y#N|%aQY?6m-m?*xJK3&y znk}iUlV95vPw9J1s+VOkw6#v(V^Vep4|UJmgF$f79+LtcJd`zbk4bg-an&A^-tXX{ zG#oeR@};1Eko>~SVdE}$Dt|!s=XaWP`UxI!zGTn|P>7I7TV7ksqh_UmD=NVyT+1(k zGa7$_s^phINDC0ub^H>D>MumG4;0~ojvjqM(IuF7k#6V`7M@#Nhf{#daPH{nk*D+v za7up>f`M?9ico4rM-Qd>_n1^7Rkl!mVHh~C$9HVC40jTqF~cG5!_-hMOU#hpxEhWq z5{!|wFld!sD1%rigCG=lCnc0Y5K0(9EtEkJN{7xKV!y|OCypA`S((%zn@J5~sW>x7 zsDZ>#B+lt99S>iY{8MKSIlVdupTnPfqRin1fVWSCo>^S=Qjj!7JsiZX=oP^rVJ1V zYdzw?w32Q8Y9(VA5iP@P2|l+ayJ$`q55G0O%$RUtdur;U#J5Zgi^Ud>WpZ>Kq<1*_ z;*+FY-rK>CayiQ)7eG`;Du*akmP6lPNZy^X1$EFoUncvtcTP`SD&Xiq$ zvY>1^Q}*o1in8TQf~0IY6M*gJA9HkSS4r8&oG_xRq^v+v_MxJzq3dCkiKFU@qLOdz zl&ut$tz^pDpQ314$yh`a`ZfO^b&A9omKb@G9kSSX3NW^tgW4ux*T}C?YrVY3B+sb| z?oI=%T9gK%mdQ+&@(UFjCn3ZMxBF80TY{Jt`71Qmj&HFjq93ug8+Goe#GloKP` zZ{Y5|WtxQj%nosD%X7JR$vKo;)g_*H89bNktn@@VJSbO(%>w5-sR#nzBqq=7)NY1nk~gn-da0b)023A!t^BAb+akhnw|u<`kbaD zJ6-fknG0s?!Paa&DA_8xT(WgP7nyn*Xoo^-B<*SF*+a8kGm+%jX;OnsO2_C0r{P;9 zTMwfbK-BK7J*K9*KvORo$TyRRoa17{&LiVZ?tv1u`)-e^bJn*i@wO;o{kX>@f2Lyb zZSHnz8IA}Kh(tmWY$=``@r5;z1`P}rwKLhs@e}P#9(#_97HU{KMYmSzL&IW_8brzW zc0KG8pms5+k2I)Vf+Kg7_VKUdyT$v(8t$k-6<;o`AU3YIL!Oz!M9xeBk>6w~hG+7g zDRReXPRj4+xad%phn$%saYofo52xJrZ91LP-9yeWqGB|QGhNdiHRmwKM)|A4p@=Us zRikd$XeZ%GLBf+vLQ{7|!jqbWB`KbBU6kAd9Z%p3`e_l9&<{~9h>Bqhd#;PhdVmDS zUwcf-EEe^lTN07gVXsNeJv@|k%3hPK(~(KmaNf7ZviB$zNb2&M>{PxksC+%;!gF0T z%5svGi`^a1h3!}B*0rH;4Nc#K@D%2;JPj9;&N1wya^EB6cFEJh%+_BHxnUZ z;Vgk&1*9bb*n0xldnv2Bx#$%Q>^;G*J4(y>*S8R*l%x%^VqL+cO?FX!CSZTYu&;M3 zQGRwrLgSn(Kp6=DKNkQ$XMoij;O7!xyIy-uT_UVYS|j(G^t8^sk{KduckNzN=c+& zDcEecj=cidUIzAYwnsb+Lwk8!=57sqZzAwuC^j%OF<6!$Qw7+YfM9wmyS(YCJPgbRPJ}C7`@k0z?I=Bmv?B0^$Ry51i+s;TqxtM~4`fs)80yfcl_-`XED{U@1Bu zoUW^C1EAA%;2IQZOH~!wmW-7Y#3Fwsq;It#%bH0F+3CXaUc0yl!El z2j_!n0keF9M~)&edm^Gj9))W`D&}y>9f-ySu!!-$ujx^8k zkb5Gxz5^l=dWPw^`3#IVie(>TNh&VXeDF-0uUBm8hFh`KlEB`3V2|iMCbTR^5&j+{ z(SnZy6QWWu8n)R*^}c}hKEs->VZG0Cf*<{ef33`kx7=~Sif09~F#+^f0`yl5x;9sl z`4tPO4Km~6R&ZJp;C(CLearCP)$qQR@Y*fgYw8ktKPV{q)<`}A81+foD*QozBv3li?<*S^2~}yB4H`B);K5Yxf*wr$`4ksz z&h-!$bPh6x)Se!`jp|^kKmn694hj@$L&kQXG70G(`31b+&>gsHYzOl7^pG`nYzMjr zKPty|pso1vkFg!-m|ifNcHizbH8He3k>NYr)x{aUv)!5(HN$td6Vx!n;?mSahPmTC zDx^WyXwaaqrXBNJ(+>O&?B$`H$wW1bAka(jTO%o_(zSyM3^S6dWYg3AfEz>SVw!R9i|bM<|)&A zO3=bX_Bz@9%1oxiXE94Q9&)YOB|xO@7GHOHfcl-hKAo=Y zEr5Df{^QE5@u^y$N?mY$=RdG$39i7M3k{t@MqK-4r`xDjS}>^z&d=N zNi!I9JEW<+CAtjf_$P9(Zur#B@Zw^q7x$j)oS7{O3vA{QvEkKE^ayXR`06ZGIUX^f z08FyvkqQV)tmBm5&hn7EoaPvq!9TT+)WjT~02tTDBSp}f3_IV7k{V5sn4&I{;ZgDn zI{BnGuGT3&TTW3AlBp_J!W8wSvtj9nOG4vn`qeH4Y$Ywn9#Zf!_iQDCMa-z=`WX&1 zo{iOm(PF*fS)-_TdrIr8795^s5nkNaL(YOLoG4cr3exjDE!E{554p?6(o3A@0)(QG z(kQ*m;c4e!aC#*!1P#+0T%hng4GXVUEbLTdH$3z@G5q<`%*soObIP z$Lr)55HzBn0D%P?lgTymi)ftV>jYji@E=rr;-$U+TC`L_O1y}a> zkZblq0k}ba#r0OShbNCuDTHv^;XYPHK#eAZ9m|u2(y?48UN}KaM@0 zu{*;8#W^5h(*Tc{)8=b4QwG8}tnhTDhHIQGkaJ*7{FdE7EJ}L1(pLP~<>^YEOFhKT z(+$1!FfyYtUrdvh_m`c-LME{gB&H7XP_B3n#wbL`fO4a_SkEXFT~+Q#7QQ;PJ&`*a zqaySLtU+o-5(@i+8g|Ctb{s?08y`c!{HH*3E1I_20WpNt^_Zh09Ey?aeM{WwSp zf|rOc8tfrQ!*QLdMap3)PtftgRy?Wsr{$M{2EOSbs=o%5k_~NkrO4l4!EXWCT7Dqu=z4o(XNBC zv3PPle)JgPA=eH`u~&j}2gG&Lp&@Y&!;c@4iTt1@W&8^ zx&o{5Kpg<;i-)2W#a%HsBKQGo*mAGvC69oHO33$YPk)=;Z0=R3i;&)-|i#ig73f)0K#f0 z4@LaG(s2U={>m~9DcnwxgSx-RFmk34?O`Vu7~!Fu@*sj|U*w`Iy^7L3!sj{ZI?)Yi z;l}(yFSZ^as+=+W#6Vs`Y!hNp7k=iFtG%F!(zpqiRF6v$bbyMzFx9#HOgiCmk7!OZ zWNME5BG+ol3Zn`quLt{1)mSCTQJmlUF_QUw_?`2_@x zXzNIYXogH)EWgO65gk(u_PF2Rm52_3R&aDivE*otEcuiC0)j@g0uY!Zr&Nz6zsRN$ z?Ls!$szqawK(Hb@$QLc1SY95OGB8wGrj=WLjNnStwdy3c_-fvIyW*lo`x#Pp@Brx_X0?+-215hl|@&_LUxT)*vDnq#R_q>`Ew7R30>2 za<6YxQ`MBNsH-5N4u%`;ES&%!U3I1O#wWlletKowJB6^7Q*I4v0OrtmLFitT=2cN{ zdV5@j`L{7@M*wg8`TZ&nj)L*;fmhqM00j`JA`82iXW(~ee7@M?^Cv>(WWrTtUgIHg z-o_uJEXcA6G1#;qb_n&o)Rl){M}>=A3MZiVwe3XbOL6~sJt~u5 z?vUrG?l)))J^XI>v`@Vm6q6J!mR~?dr{%ESD=K5P%ov{)5ENTyM0}GALpUFgsXtAu z_gF6UIFl*RFF1qCgICvtf=Qru}5eVfzh6RfieX1P%Y`(!mj=lR#sxNW!hIOdO zN@+1^;u!qo#-%hAES(sM_=7Pq!xbGGs0ze1?nX@m?#${WXN@|+im_c#&^nnr&x!tPpNmU=U zCyx(Ns?15Fd$*?&L`D2nzCe&p1N`R@Q5m6bPyI8Ut{%&EE=v87v=E_kI%CMiF4~9R z(NY?Sv)vDu3fGY@PD-3x<{@W>&&jh%ebBR7R3 zEzj~DCf(-uQ1>r!|32S=Uf^r>Zoqngk(Q+7D+feMx(x|Z9%|&^MTDW)SaCtQYBGkh z%k7~hH58IcyP+I|B9`V*NHgORj!?223fXn}3~{A240_85S%VmLV2kgg{0K2( z6Q_vy3lX@77AHN3IzWuw0v=Ll!8)w$$*;Yqvz$9tJ-Yd0S0nKC{s zznnf0<(DGe;c77{>~j+2)fHygwZ)|2l^(LPTTGgZAAMR(YQ&FzEhZgP<)N%0Ehd## z>6)%zuqX@oDzy$ux(SL5*2Q(g5U&Hbk5{RUc-KKQD-o2Z>8^tWPYA*v_{#85Fh|^l z2%7}2g9JYeDs#T>RUW&Q()Dnde+AL?kNRjljyMhuNyFji>XRmTh#x1IgtVCv54lSt zQ6Ei4a849kEKrAm3hVje&hhhO9pl?!X-Kvp~v%scB<P@tYn6?!v^_< zbaCX-8oSG+zmd&xMT<#IV};<+ZL*>*CiR_!_V|_-lS)cO((Ta6tQM21IO&NNlj>y> z-Z-#cZ!u}1PRc72rF7CdB&EIAVp85ILLIg83z9h7+7^>KOvVX<7Lx|!$G#Sm8YY7h zMkis;1SLS%d5TKP^9ytjsia{#slIj6T}Wa|GEWnf%#dGD!Vo~oTT>(@2k--w1g64u zKGUHyEtCaXcZl&h##_IhfOOK@}4Wm}UQvFz4Y1U|w|-^v7+YHr`5hPsOo_ z-&duLx4SkrkV#=Gio<7QE<+w!}^gMo1Bl|X6H+5LzbxgzHH^><-$8<5xH&7pr ztM-ty0WoMN7cj_cs=+^->Ka*1omqlY2z5S$3RA5|J~maV5JFvmB$(=UB(Yjh{;R5KoRi$*X0P{+ z+9VeZn2uR|#6sN-4ECYvP`9%Lv8A+{&0-t!H6jN2#9q_9dytRK;zCPs9+qDSl!v}B zi)qNmW|0>VX2G{cz%1TD5}QSnM?lCqLwo^YEtCp<|AcI;?`jF*dn7^MkKL>ET`!ZG zkOX~yi6qwdDQ60xVfh7c&IWzAyHD!-qWfCuJBqEW;G)7M7S$?l$Q2Nc+DbWy zCF0@gtov*|1&;XyF{-D)F~ zCnsG4;bAEBuUQ_!Lqlb&K~Qiuc#y_7H407oW18}UvlIt-C#a*5qdNH=q0HkS!Qfxs z%D^$#B1U%dIbzWxiYE@rd3-)ghORl^Lci}3*{kG1uX~P%`1bzYl1Fd8paB$g#x{!h zQvxlOZ9XON_#E(*)^oo})w1vm`2~=ZAeDjpO?n6UShbl|PEI-%Nl@)iNJ`r_DuZfe z$rkwq1oZ9sVTW;#dWc^WP`SRBxdstYt5*Q-f0U_0->as#VCj4L1x&ajM*Sk^V{c%c zF{+TyqP#hl70*_9ze!p1ROYz`rX!@4&gJG_wBV~H_!L=eErD13n_e7aWb8hgROwzNI9N>DE_DU!k$ zm=I9-5=ZAeAt`)`6DB<&DYPWM9Qg%>ywop*81zpbRmf7*6lGpur)++baG~>);6sfk zr4MDNJGcxRW#%W<%ulCt6H*GxD~RSNjWt{pLP9v=*RFa{C^R04<1TSg;3+#im9pd# z7u7wb*q)yRwUN7&mhi7NPkAKqm^;BBdsJvtp;y!0aA6m@T{RMP_*(8;3aUHP3!Emcl_?-g^?4&3OL7Ziu$(PzxM z?>mz^%|{hF469;8<7-U7+k^4;0Nx?T&a5Yf~ zycgRU?8OtBy&&zGKolYBg*viA?I3#d)X}A1_dhp9j>Eah_MwjJLWh6Gozd z*Ii=gyg=L+T7a7^_X7*hx)m&5nob>`j*p28po1F`A?J2D;(@0XB?X{TQc?hh$yo#O zjH7;ar?{y8X%CTWjbyKjq`*CnijcYrscBR9n>r;^=71UVg_V!(H|dwBJ!H+_Z_-H% zJ!Cz%-=q+JyuRP0hZow!Yv(u(pqa3O->5W%Yp)xdB#M6eq{ zIW8Oy5e!-c7mAukBH}hggc&*whgrV1NXcZlO(w%xCOkPM<_<<+8f?!ZY@92j(Qq*5 zJR{v0U+pvd8I1fP!T^N;cgVA7%s$v}Qf5fhwh-F6jr&bn`mBe#|G3{I>YjE0bx;}F z5hWq#os}HCg?4#dE~j2EhXlGV7E`a2Ly_@Bm!m)Ns*n3n}zee8KUXFM4hF$84w39Q3!5EP1Ed3 zckaQdN9}X5>oHKFf!7UENW4Vk#u#&4r5*QmGU~u*3 zN$t>5*td!=8E8xA&gTiwAzj4}z0{V@z0lV=UB#a(X=L~(1NXJhVMxCo<;mM# zkoI{BC|d9Wx}pb6S|f=ii2&fyVYPA&m^9!;7%j#K-jI`%V4gfi826$aBUFc4jS;&4 z6O~!mZ~{NUm_IVfSE+fSA$Fd+34`edo<97i;HevU%BWX7@%!$QCmzENtd~5U0iFW& zlBcxG1mHaR1y7)fd2+s_c&e8?<)9*B+|%nN$x~Bno|;jp4W0_Ib4N2mLj_);q+Wud zUYC-K9(Wnmkf;}2!M86fhD1xmcG4Fdxc=oK7I#Kih?|F;AgvxL7z`AITZw!8D_~8F zdw8hGU+J5wX=&0lMR|pX*QAaGO~<|pnpE~i=GlZi)UI-)smb?_P;6ND zZ8YhEH2`;S&#&ctVfJ&u0>48w-lQT zl0c^Hvn)t@73)yh(K32232hZR-OqE5A#X_^@jT}VzU858ebrd+KKlMK$HfOsDv=uC zH?^)hVA8?2pd2LC$Rv+J#7Vv0R!NOA$*q#gby8+Ts9h&Lfh1X5L+ajqh)Id3V6+kw z9?!*&9v{(h=pwIvTk(Ee!XQ9b894}05E>F7oy3NI62!Xj9R=tl_#<&Qq&MOuV8o;b z4+3f;tp))fy%YB|wu-9Q(D=Mp*g5FVc)J7dpYM1`&k6CyoXmeKO1gtXsm<=JOd%vI zp%)bJRYFz@L-I#0&YXs#8%f8^}JYByCG19t3%nD7;=UfRS zJz{~t-u?T6mlJmVTlD`{^P)tzccr1N<~lO|T@R)C4w$qa^DNlV{_ zOy$7B-rp0P%8s7mGxI7tzm2G`I1Sav4F2>k{JCFt^%@OSsk0h96z}VaxCr7(i5j_H zzto^K(8y$ScW)(2Fzver4^g|!jiyGNrwOnw+-TAty6~yMM%3=M1E$VY63_hj0h3O9 zPhsjNGE1&U2Ln^E;_K-dh+Y|}EGzXz%5>Rk4P!J!#gO5ve zcnVD3)k=T>cNupdFI#bmi~JhgW#|J-#b3sT$cr)yB_XTj7v+b?o!!J>V2Yd#RX7^1 zQ@BPEVKmRW%0RFDp7(^Rh%z9GGyW5Id>L4UpN9x(nH(|teGlb`%WApAM+le3rWD2^ zl>>uB85oZ2dtV7S<5F}g#X-Of=D{r|UXMqGKmP+^x4Fo67XldQ<|5;>A9%z;i>7M@ z^<5+v;Nc9Ypi{35`tt*AwW{`y3PJLk>6+_Baf6xfBc?L?ppH*iuBhx|^J0l$cp*@M zD7nE7>H-1k0)RTGL0xb}P+=e6HLQT-jj;pkFM#z2u-iUVboI9Zv)G`7c}n-HKolf^ zx`Ivf3YccQj})XUAngRx6p$3R8b(V3tZNw7HGt*$$mTMy0l27!ehtW$TD}I#)0kR2 zl3-wpMv>vOOW+CVGoRoZKL^;@OrKzvvJ7GAGoLU(&q2gU%M>wPS14|uFi?{uK~D&8 zVjD6)$v*Sq6~guwvd@em1mj!C+5f%5LupSRFzMzJLFDuD3+v-`?KcjX^ydm~eTTOW z!N3&hG!>%_2@KaMr&-4wUiz`9iDGzlk&5;sHk)`gmMVceOv5*PA}BU;J#!;bg_ zjjJzM>Sy<}lCO}BeBzV{ef#45ngJlH}Qb zH=EjtI@nt{yx*_^BO+z~2=kJ!OH-K=L?-SDt zsgoz5oBg7S*gchW$oI)Tx49^8W>g+SYPJEDCz~{|QSVS`@1C z1%ktI_k}9j=vPLLS(>YP8U1S5_fOhgUCm=dH=oX_tjljtmwk#I|DoXUiB-4~fvS*U z79zvGNUYKyq#zodk3Yq9W`wu4^K@p1&phO8K}0Eq0cg-?aLS@;m-{A#aAXjToqR@B z(N;Z9(TF{H(+t}na2h+`@t?)}LeqFF ls=WIpHWW0UkA`aqVWQ-co<+_-sB|P>UE+Kb2det}L-N#)tfd3GR=;7HUl-IQr&N;Few`R+@WdyelD z7x1z`Pmx&)sqRoz^)O#wyVl&_9L zB{p|op|0`Q02|u^jJx}bzmn`DaNp4YtE+7^ESk^;vA7@TJp~higGYnu^+lq7Ap!-h zt}Upzx2@MINht#5s}e$=NA{VPMI;{(A8yzo#Mn{sVMsAqJXFYNDm?e)kY`e@f{ z6pfP<14D_Zr*Rrihxw}VXhVUENQUxhdVz~nT5syv&qYZx?JRn!pNo=3f4?V&rB8s! zO1{gV8y4z`*5eRj$4cUedVccNX>`GAeJ}DZ?&SoTpd@m`h65duU~x{&H5JG_b16cd#vbB zzs0O6t87v$m5pAbl3HYvR@p`*{a2MiP*PJ$Of3oGvT5q1{=gS?P*dr_>Ob)V@26*e&pkmt zqnO~&d+Ix16-|wC+H(0`al?N$F7ql=citL}HR=8PV5W1$C+XTp-IA{g3pjxvn zL8FbJ{|-6O1qCRobio}*AtVus80-66-Et_)H^{G1D`ov=@EiEjJoXE^6@uzSMEe<0$_8YS50-1q!a@|-mPGI? zjo0#Av{wUcWg!X%8;B|zz9)g{Weno@jdla+vr&okXa*98uI)p?u1mzRlW~mEQ0?{+ zlUb|f7q)@F!vtYB4+TH--UMFuGoCxZON^v>EYet8nTI0(FzZl2>k@%aPlBu)C2(OK zKWgButV018#-Si(-j_h`awgYo5~5$ub+ccS68+JPV{}o1btt$siSSqQ!fU05*vdLI z5H=50Hsk&T8kaGQha`>SHH4Qn#G{#q4ZOAw1-u{;$xcSHUW2vUM@N}8Q+{C^_|x{G z@z@PSA!tygYI%KfKO>?|cK^|RlahEV1F-{UJ?cJ}f?6;mK~~dwlf-ZdU0BEj4ZW3x zXy^=DgK9_wxts~E1xW2b-qj$FY9hgaT)Y(?L;=owAc5wUywKFFLAJIL4TX(F6|PSN zvW$VaHVZ=*Mv}K#iS?*PVuP=(L_u###IlpIT&v;Qt>hJ%wLyMiA^3}1iAH2M6NRDV z!33e~XH2((AMHoxYrL(^#15I2sqI98ZAt_`Jqe3@pGok-PWEZ=t?fjEXV@B6$wLVw zFK3cZ-Xc^We91*ylnRb+DC2{n$-&`bw*$|?sE(b&(PG2pCt8_uexZxTY3TCO^|W4l zO`0LI*2yoFfIq&c{r)Ys4%O(k@Y*_a5MO*s>wmjLIStn_|0KytQ?Q6w zWYTtHrD1Hsz*JUfVocE@Yqi4PpvnqAx6&v--$fZ)mHu1R>b608BuoV>?_tfG=#Ac1 z=%TaQhS@emmY0T8p9pUm(|E;JVX?w#Ox4hj<}}(yCbXlm6wFGXei>80Fag^~{{yzR zaoPLC3Q;%_;ZA1nfJR~WRz)(aSAJV*+&T(}RjMnl(A6mzLi0PB&raKv=Iv;w$gFqC zZ!5G%VAbv~6|&6P3Bujayq>$w-mG4$NoeIS+eYI_z@y;Tx5fh#oR)xRN!xhZ#sVbl zMzxDp81f!T;AnahD$r_)Nx0}=HKtZBx@}y{Q6e&hq@gu3u$A?L5T8Ic;0L9_1Y~Wa zO4yL9fY;4Q;Ak%ED6GM@b>Pl{qk7&c@tG^$uH}?~3gj1u$HWOd3qe2=>#@qzvUSQ@X0^4Mg2Miz6 z*xEW^_}>)GW=z0E%j z4+Fd+VXctgqk2S1iEJem`sz3tqLn;!zKha#DBT^k<&_Xs!>d5GvH8B=)q4qoYBR|Uz##Mk~0 z2#>^`Zrv50yhMCEnZY|X27BwCBC}@5Z!24F-KntgsIe++B?^SBo?)9)j)G6dbinp*gdoQQ*T*CMa+}vvy*$XnMtj;`z{3ZTt~6b=XXm!cqNH0#nnIps#-=;QFR* zTt{HacD_ZyE~!hvF_-mY>{6mSVt1n6)^2lm;xkDsNr0`J>m##oqe@XzFuxU5;Jn9w zr>ec(BdIYXZwA7}F<oPb2H&;YL%CQ}ic%X-2DjR%gm+l59WN3H1vSc+IBE>t#RjM;7rN+upuz?y z1bzdaBW-|+_@fgmW7lHU{XehgYlzj0G*fZN`wCr@zenzYL15IL#62(x+qJOQb|gdp z&+EN4K<_hc0KTEfMUxd^1YSHEU`evxTBB%Jpqtu4aa)m#wkZ?{r0q?hJssO~qNSm* zpO3hf`%3*oL#4h-dg4zf8S&Jif#CRfLQ*`TUs+Wk7*9!wrxXWbHdv|ggb|U7c!9Kd z{H2wlu@dvK99vou!cMKh!Sc|+N?%2ku$@!j4*IJ6SQj1Y4{~gS;H1ho1-pOjDb<+_ zYfRyoegnr09zI|c=NC}ok$pMVv}GMVyx2RSkfR0cF@<9W^cy^;-=+Nq59eqN1JT1r zVE;D93zlT|;sGPC7+fe4916TBP}FZ!KQ3^LJ$}R$14a%UI^s%B5pqzVt}Y%mU>L`z zs}h5~-jO3ljTkek(8~$!S|{Mh8N_zHN+=#tIAl!Gr6c$Z!kkBt0omk}`FeMO;ML8*6MkE!5{CJp-b8rTMt<2ANsf{Z1>P>@u zU>ql=tDpa<Ms(g>9aT^NEs0Hjm71_h?Yzs zx+oBh#|zr4d{c^s7kl|+PBET?0FFk11Ws-!?@TJJ2+!XbwxCE-u=rE=Ir7AWe`bR>( zvQj*FfOvrv@k2wQ@K|5zc#am35$z2Hg0WE{{&148btLI4(l|atqZ~Og7z_j}I961} z!QM%Iusa))6QBosC-qe^qGXkC3g4SC)E}&fjpM8*F-RnYBC-Ax%eC}H!q`9sXZ!o_ zzeRnDPK}N#^o|)aV1Re<@KIL`<<#2+i?+1F;ld-_AyRN#%dlt#I1>Jb+QUUg%1is6 z)4NYNR2e9pItoTXj^vEHF6aLDh7=bWoDG*aj0lz@anOUKIC0VBbX(#TzDjZ=-!oDa z{OK4~FzAn!1*7Chu8~PgW~gFS6DwnZurD?)RvIQp@?(c(@JAv;eO3OE{&IgL!tnEa z0zNj=2*Ui4NTDwn48=wVtHkcFqT$8E{L!ed0>l;^xq#d(mXO`sR(9Oy?vD%$M92Ea z`6fZ-j^vWIG7b(-@>K@PiiQ`D^hd*?U=$RD+sZjS)GrvF?2o_$6%8*I7|4-a-BxCi zsunZ?l~WT*sgrWUYokIYL4V9&7L4YRBYCk*j7wegm(W`xM{-lAk*v$hyDNQ(4TyYP z2MrGNa*W7hqJFV4RQO|qLeUtSDsm*(OB9E89xL(3{L$E;P-r|3o5O!QlItW4+)N{2 zs64{04EB;E`3(u^jb=L_Z-g&$Dz;t1L5}1GnHRS{B?4@fafQr?OQ2e1jQPV@&>=^1 zqs%nDsC}zUg?yaKUj|mlk-X-xjBr;`_%N?8HV$SkHHaNb0w2y5`eQhk1c6HcH}6tB z;?aJ8sY>4=(>LmLv1MNtZ;{D;cdOzR{umI1!m+_Y1>~U26-UBULJysVF1@-uP9hY- zwUQ$_Q>K32OIgFDK7)f-^z}yk<$)>WNY0af0egQ0b8vZM+08nV>oNrJm$jiF)mK>= zof;)aazn;QGKlEqzy5!ueR+IT)!F}Z$wDSfG(}C(g501EGuIFj7PV5_011mA3i2O*!tGAP^)#LtymF@@vXJiqK)>g)TJ6+a0C5) zzRx-5&YdJ|@9&>EbIy78^PJ}__ndpZF62-wW+l?L3GeUjmjItV_TO}& z+#DGlPBbT@F;j0$RF;n!TRv`rJ}T7|vdoBXnT?_PD5b3!8b|@N8O`bP%Qmbu}c735%nj%89)uVJnA*Nn>J>6KCO_;~! z`6-i+oRapjCxd0tCmwuBCnE*mf}vDuX~K$-c2ka4nogM(k9N$cG^r{%n~g}?RecaL z5G=>+q#aGfLoxOnZ9~3Rn&kUotVBI^ko8J+O%40QsGe2B<;2Vjy)((SUOccGnBHY$0KHou=LxS1Cwbu1?x8^rK7P3E-%kD zQ>IPYc&)UaXyDs}#0>#>P`Tfzv8;rZ`QEl%*!Rd(yJG+movPt@#;Jd-E}}EBreBF$ zET|G9lI-P{wPz-zY;CJMaGzx(m)HmRHkBXqi@RMKt#nOpr!S;|Wvt0|)QdX)N? zW~3~Z&=HwJ%F7@Nq@21>N?F%i%6s=6Ny=1HA{~neV>&W*KjfCJfuD0VAeMZyFQ8ub z(eShCA{z0LlC^=IfbYR@!>8k>=vInW^4D&fJMnS{qMdyAv*sg*>P`)0?{>%v036q^ zByFcSu)sOhqn5keaouU^sNHF8$G<^ejB+v7Ac(HBUwsXnFwz0A8i<3<* zwso;>Tx{FcCAkh;lxIjP76T{B-IxQ1b-Go%aLah47Cmt8lPiIqAeiLYD`kJx?O)|hCwSr|qyWPk=3 zL+45beZ~>USuK=;->TO#6m&3j3x*qn>h&J;2aJJJw8#wxwRMk2f|J5mT}hZ=wZeJi ze$<_>BM-@$dJ{3GR+oxAvn;2^_sBZ&5pOi%;Hf(N@TPAsU~lj9? z<*D|g`-}DY`)!`c`iAMnDQV4)e=v%~6T=^bY!sI}naKfGvPSA)^z@$?LB~Eq;mpxq zIGek)@fHE2IS!Ewr?9=)VY>gO)Lyg6CWoGt6RwW=f_dztKxw^cIw6l^I$yWjvFvr$ z#zt-_!9uZN_Z*`?aHZPbg9AcWTv8SGio#*JYHw1AZ%xLZ__Y{?QAv$5A6b_VMme<$ zq%uU#^1I?b_z$Xuar1*exAnvieij^$VeVheMZPIrG8If2!O&h`Y@?HZU)e_YdHoIc zVzi?tuCW6}dyPD9pfI_k<<_+J;oGWd)ZX+r>S*IpA#glZQo?m=$QmQgqGz8$1{9|` zSu_3->vyO;6}w=qG2nYL0>h^07o(`=Gq8nMpv?KNbO32L1&-=wgz8Mbehz;v zn%FMEYNeo7=`X%7F}w^~r3%k?A^JIn(`<($txDfjaewNTuyTIZt*kcOb7vHC&TQ8_oSBgjHYS2G_R3=(TTXk}D}|gIe4T7{MYqbm^6*`Vss!n)k{R z#<=*^8O2{i{5!Lnat#c?4gNr{K}|49^XxP(lsRV*pab4JS%{k*}3mw-S~Y1vqcYaqksWfM8{*u(!4rYd_cf}&5hm5*KmBP z{C%XK$Pp$iS3l=UXe}Sj=;+)HRTNR85?a18cupySbwW`rx=`*!m}dRZEknG5>iLEg zepv=yD+FSHBBzx1hH7Cx!0cl)Bgm!xFzPzI-R8Un2g-Zyu=CEVp+x|N;`UE((Yh~B zR#aC`W%r?+Dvo9Eb1a!5aB~Q0ffcjjq-=_M+o>EFfq}-c&wBQpl(k0SE$AN*QeI7L zQwG>3@J<*0v@P5JkRndiQExJg!cS25Q~3+idNpgJQ&rqTw~%WZHmEPgcC~(qZQMKS zA!6HUmo5fCl6=+4HbPe0d}{QdEHTV;2F#VK-p#Zz`&Ed2rKB^WP|M4J=a~f$|9-Lf zahO!SH(F-C@lmKZ%=NNKZhJdNK7yj9T0M4>e_s6irdX>ZqfsqfuH9i;nrD;0OQgl6 zp{C8P_xSkn=k6n=tMAa>BW2PEy_c_Whhu_6JW6G3>23XKw=&NtK`3S9Az_^->o9vC zANZ>dbEn!!Xf@HSpwRa$xz-TL6{N@)6(1pz1L-QvKn)&p;@K=jcaEs2k9Qn0mpJ(x zyglJB>qw;ZqZ13(-_imX0Z|TWNfL)78Kr7-ux|BHf;n#u9pWWmS-}C9Kv2i|Nbr>g z=@=E-;g?IQR-yqQuh-B4cC)NIu=tZ^S%~Y@fiF<_u_QiS7 z&^c@86J{qeFrA8K)m>(gn)K14w3Q^-pw6I94w7NOVZvIq^Ade-sz%gEOI+z2$vU$d z=R1~_4{=LgLNUt(W^?qO$iKeByOiLn88fHAqREuKz$v+vzb#$<4XiJt{h~6lCpZjj z+T4`N_{cmst4EsNmT$a4;=C31pv~Pp>5C)Uur;`!V^WJ;l`vDAfoU@YgV{XIOgwdQ z9$~P-kP=iUmgvP4GhO}Py5I5IpU_o^opnm}oy{a3t8L|N8g-rMQWTbN?W~w`S>-Nl z5PT{%HF{7`Z0p?u$_#qq_g8(Ri{$gt6{~-*L}ROazFOEu2@^gKafRI{C9v=Mv!a6k zDGU^&(9*>^lG!NS&^Nj{&xy-$?uhYJhw5VZqp*EWwL^Yx;-K{f^`a%ZP^s;-2(0kj zew9~#u9R`uZos(ya8~RU@0xFb)R6~r-csLXI3x1yu>G85pzO-SBrk5SwfAF$wF5pS zH~P-aCQaj21#oi4*^8KDt=%lAc_T&aYjC40c4F)S`CcKM} zbOwv>7UFI*90l{jI^l4jMs|i}sz0Qx^J&mmNE7mbLwy^XD*+x!UIkR+GM+QPH3(Nbmj-XOi(Xs?%K%dVJlgpZrv4<@v|g-Ar5K-7lhqTLL==I# z4&@GTj`#chOh;+7SnSZ6*wXi_Q39daNBQw8*Y3r}x3N9&7CxY`m@qdLqNQ0)mhNd= z(1JT#27xbEd?)w~wHFq7a@m!Cqsr14X1AczuM%h23F;``$PYgD_1zFB3US&Wi(O-S zf+w*Y1CCumx?H2v;orfsz%=lRi?>EuAmS4nqr6R1P-X`TzW8;8O!T!Xou)DTiJ#6B z*xFlo$<5lqN7*orf#LT0RnR*e@(j-9JfH2bZ#VsfSnuPn5%mU?pSx`>A~AmOIq-IQ z=#XBfr$x-1K6hX7cKVceEl#YbCE-QbXKPQae}-(W(i1XG3KX+qm{5QAm$3hcguovH zd0bVjdOJfyDV*9R!7aS=__pE$C0=w#trFs};Vxb4u%QZwMBd;=LyvC0&*|s*4(Vp- zsFZAb#sW6T@HX`aVJ1RyKD99266ot!7qGh$YCsL^O4cnhkh6t{< zU3RI#Uf3-MjA`RdjK`8Hi7Q48OK4V_t(vC>dBR$*{+d<>^?xcYYo)bo zAHi`Y-f`DNFXO5&(xu?j@EOD9{v5&mIK$`ETkeJru$YV+BgzPVO}6vup{Nbo_3M*S zL)^C|9i7ZJZQBWnPLpS4>Loh{$a9d&FeDn_&GQ24jUN`tJMcqS8}^+7vliW)e`0O8 z>fB@I>`KR$<$In=j4*x=%LQ8ReYlL z5qh5+iu$-HpN|?HY&=6&C6K26gI7OcOb{8@)acED)qD$wdj9gGxsv$H0le_yRYe;t z3lWr*1B+(-4qqcASdaF8{~?Az62{b+B;iaBSmI9fFlO^y=y;IZm9y#o4>f(0N-k>} zDd|bflud7m1hbr6fw&`5oY_0}pW{g-OW7?`GG1s*);LzhbPV$MCwK7W>gr*3n~wXC zMa6H1O7qHLib zkDGDv`In`$@AJwP($=lvfi)wOU$QzR3$X0%8qTPu%@Y(yZkW*LeK@O|zJANq5E-x= za`t=lUrTAiXqx{jnPB}@l8;&M+QIQ(*%dnokt%&T?guNn>#fzOaZAZg7P&ie8tsm;vab3$lxcH+3?^mf62^=oP)se z&Ue?dSVRwLZtdVt7v0n9XpWmdcD}g%J%IiKoeOs=`Jn~D-|ZVDMzIZl`g4ZeYOp6z zG3n+cI#3a*QFNsG$Enrvz1L!FxM@Fl3bnma2bazAU{)wFlWhB3$Hx4C@#UO>E+{I( z!X(QCkt%^MeQUxiaVObI-&svR*r>hzUDb8AO!hbH@obTtpU-Mn@4Y%n*PbIx+^y9K zHAe=jZWOUGW!v}|;EIS36dq)w$&4&LRfuKhOzc!Gd1g|%C`P{G_}s7K@I}sjETY6C z<`9Z?=tw)O$qNZ*SjA$b)?2r3lt&{p9H1-SuASEkv`0s^CA$fOxJtbw>W?v$-jQha zBNT*R9iSWr!i1{yK}0m@D|(#ipP9Gr&w%6;w~LZ};pjTo?TVDsK^4tImOC>Glv2m# zCjvWsLd*$Y!RIuMEl^Xd7N_gc`}Uy8J&qIzHO_JkLz@R)yz3%Sx*sAw&YvN!ONe{N z!sf;gAf}Zlro@uUztMh1^5ACjz)M#%Cj8I=_n22^beb3yBygmrzB^o5CB5K#=Yi zu{~zkra=4FP-KBUm&qH!7NNev)9(^&9w_rgw86SZ%+wC9Z_QCObp(4FxpOu@n2I4n zISup7HA7s?@Au9^y9+GUpU1?A$X;PD-hzrF(&A9EOLK2{oWp6;g6UvoP|OMO9OwdW zvI%kL7SkZisBgQk0C@TnG7s^xSP2BK34Q*VC~7kccZKRm>{!|%NGKVT(lgyP^2QV^ z{-15uLF^y*+0KT|>3P!(ct+_K?_RoSFF23z!2tu)$Cysdt)?ZGY&kO?=_oE~Eb5FW zPb=I0-zqh=pnP4n))|fa5k#LW!p8d$^4AeY8-n;c-y$NtrteMQ`d=4lD@%Rez9W31 z{iO{y;rK&tw{kNHyjw9ak8l0$fb5|%Ds*p}%f2~4#g@$JBBPvq&0`%9hVIcEHji)5 z{gR7_7A>Z)Sk$DJ`s6gE-`jpVeqNaA z9(9iY+<@s&==`D3iCE(QiVj!@u-Xq5+cXTXeb>(ftRoTLDs>_j;Tue3(2dqP*B)3~ zVS1Yws9bLYzIdM*av#SCrZyQMziI*tv)Zk2aGHUIc}^-aM8u?p9Itnw7h)Oxrj_cw zXC19^VJ9+e{234*y6I?UGO*@r{Y^UBUKAC^drR)~{S*>)PRtRMY>l@BS9p!O{`5s( z5JCgJxF*O8F(4MAPv7Ft5!kQDr9Smqg0;wu0kDsei&JD0l9td*6a{3x7&#j{-1vv+ z%oquKmC#`|^n6@HcC$&1?_$p0XD|M3pbb;WU8#58%Q^TsKhw3Y&tbHlSmXM*eu_-( z_dU;I>ffG0I!ICTP!*aqqPI>NV&s!Nriv+R=9@W{UIW1Kg-~#Q>#M6Jp)>hClnPjK z-wAyb8N!dHV9mG)@)^v)iSG!xqFQ4zo;Lc}#r`Nw+~0`G4zF7EW2}_UKaqLKq&Q$} z_)0L9Ow@;5W}U<>%MgPKU3#Dxjli58c_94pX0=Hy5Z*NLYy-z{g3X1qsB`eGa}*lX zxVQCT5E$&(WWh0*>K<`S8Lw+%y2G@%0k!q{2lx*MT?)GTRXXJOj)+#<1XS+qnDJ$Q zO_}bXTo90)%RH83_uDe4Kko{O z4M-o;^cmOkZdpO-N%lF z5b|lEFW?Em_51+Ss=8t@jm!8+%C>fdlD^)jZVpzkOLEegdh=bE zGr0rYwa$vxKAN=u-mzf2Gisb$tad|cVO46uxTu!m%N~>8g7FVIc`VP8`h;wl^0cob znB$Gjak93(^lGbbe)3_0?I)*;WW~thuk)^M6}GHp#(o>9@W|-sjLCG@An$#m7m1N? zR9K{j)u=29uxJ^N;y+)*x=-k1d(|Qp(8=f)3?s47>ov&&icsx=b>bz=k_Uu2iRjM4 zsHM3@3*Ir|+J0{1nbA5d!W54^pgrYCV7fR?&7I(u8FD$)=ZW!!3>lfL;hdc#UO+!I z8_~)oE`0Ss;YIQwo(arD0x}$*`K6c$56s+92F)#Wy*mBD?Rr?hD@S$Yde{L857^QB ziAmBz+6QwLJBz5P{I-RHRX^USeJc4Vcp?@*>CrS}8#ra(UKKkzn%(YFE<8?pVwxBl z9hX<3U$GaLJv-4Hs&ajIxn*MNZu`!`$Kn}(_uZPq&wC5_E8W?nmP@jTOv(yJWP%EF|L&oc9JP~RuMhpw8Xm2^$RmECu;z?BI(tPq~o}CJNfZvnH>}Etqu9 zcU&3`S+r{{ml+5Re?u!6v11l=dUZTpV10?ftWPG!Y^Jz8rnH}znz=#u-rn(qKSPPU zS_G%z|0q&B&0#<+4BqAA7b0tSOr!I=v;@-&h*jLt{n?kBurxD4 zhX6?UKSHfSE7+NOdtue6W0=$8g{*C88;)`FEvgCT);l)j>Ftz`LtgLp#gXunx9z*=)9J7D5Qx%0$&KW@&ZQzyy1HP0Q*IQ_|7L)xvRv009bkF_>6^(kEw!o9Ce z zsco_Y>Xq=^US{=)ZSXF8r-2GdyI=^_6xz~F$lAc~$Vn}X797v8Pl%ioE^=&HbM65N z<{seb?=Rmb+t9?-6nsmrwTFrXqE}3hE~5)310uFepIc|AAUc!Oq-Up0U|hf0Pwc62 z_1lkX4%K44^`xnuHwOcETM@srh_LqpQ?YDvY}WS+OHCINh8$8#Qo_Kbn`Aqm-pD(~ z_u8UP5QC32@V&M@?*(5d4<&D;z#{-R)(#3?JC$JQxp{y&7 ztRrjEoX`3k7W=5#`&pCwy<$mQ{*bPUf2Qwy>v>Hgso=GgF#z@`5~5x<<$0y7mF_W{%ep zs$r`;s5z+5VVyz8!TcCWm^vif+A-6=+@k!XW3hEnhwa@y{av$%q>tVHFU9ZeT<*a_ zyJ+=ez6|o`Ka;x24>U~|v`JwR0;bKDz^Akyz3*Yg8=2f?qbw%p=I7D#40a`+DguhB znq!!yN2j%*WlC_vMB|?&vYxnS6e&V)Mbzg_PC%PWn?av?q~|iqhad86_BFq*!y`Dq z0}5@zSMw^GExdA56xRy!Kopudt2qph@FQMurbz|6y_SD z8A$zx5j6FZD>o2K7m@BndPFW7@eFD}L^-<0_!u|fNc>sN0xOozn(1r; z;+&ju>tESp59vpWikScd+jX2w&87mZf~{Z3sqIWoy(KApir=0Zmp60~`@3>2iyJO# znZ`^1c$U9`RlZRz3S`D=Hd12BMc7^05Pl~zNi5nF7s{hryklIf)1>OVBUGE^O!+DF zC-E3MwWMMxKcQr(-W8`PF|EbxCEK{#e$Igzp*70;)jaf*xui`6doEi3u2{3`!5Q)v z&bodE0DlsIW^(*7t3f5*DLCu1+b(kaD9 z!J&E&-!I6mpu-lz(*)2J+MxD)34M_-C8T95Yx&m4#10s%!wD%uW-#C5Vh;n)R=E$t za&oj}DHlOt%oGHeGN!^tT!d;c8A(ktkT=u0!BW=O93ZI9c^$lYt4*6UwN$RN`)Wmq z=A<~(@Kmp#c`K#)`_ zGv;E)r66MLA>rc2xgkdEG4hE+6pl0Z2GJMbmqz3tf6x_JG1_OyQ~x=&=cF=^F!hJ! z^=)}~hdbI54_os{vIgE@;Ufy-FIO0;(_~7_@y1dnc{eF!#Z1)BSRNMH?An>XsCv{1 z+X<{_Yd)r{EF}KWwFrg$wT_ZXq6i2QU`lAX$vr?QEI*DX!e%-A*c~Y9%T7ThGI9}k z&KuZIXFRKd;we*29r8PCWk8`R5zThUnrcLU#Ow)*DEm9>s8vDp9&raXB;y5z%;)2> zz8txLl%8laA}%?pa080}ECNrypF~soj(ZAw+*st5cK+9FI4fO#*b97j?97QGsemsI z@(XVq7&_fT13Gp}GRXshB9^QP$iFKGSC5xV2Q!AF&d|EpZyatb-6^kEdx(>-9dWRR z3+M`aM~y<#q1h>QSgZyClyuoE%h;V_bTLZ6+n;5|qKtE=f}DKVa&(X?1zF z&nCc`@nOr3T*%a<-M5}eheX1#^5$GhEyBkIQ}nB=^*2s!pFT)?aZ_V-j zM4PL%#7>m@WB~*F{G8LQ;MSFw5eNJCq4yj20>9`Mav~b@+1Zi^Jd96!Vs`r~6AyJ3 zM)|FfQZ|^fi`#yI+k63ZeB|o`2X4u2)G2J#A#K2FcXsJA0t3MzpwjWbz-n_tfDY`S z11_m?3LdLyV1o)36znNIRN23i+JF^~_fL0}nfqV2E}nlPs+iH0L_SZ7@;=gYb93Vk zk~Jk~+B7p-CkO*Hyso0UCZmJcVeqEAXk|r*DSzR-VeOSx|6nH(I3m(Vi-{HeGVL8h zd*dg+a&#MRb?lJn?vm9GObKwTr&vFU8iHX>Q}Q)|?%%{pZwKO7cKO{`m1bm-b~2XD zU@@Q=w?J7q1Yfk|%c58)fmWxF62n)B-xLpVmeJ3TXJlC|_efmE73Z{1x7lWoG;yoZV4CWD187gk*T+hSi8gM-9jB z+3gJ~@c7!Ce=4k1#lvr{E;~D^01tQ8;`1f5y_{4{Hx-`KwUrXiT8@5pato2Y*xr46 zXqHON9Iw*9=DG2e^|uNOo73k-&t`a2Rgv>oPBf6?B`l&*H;gxpy;1|NvrSZl>P4FI zK#`QL{tWX;ftT$kLcP~eX&bfz+yfikr#(~&s7vV-0a4(ipsIBxh0o}Qad?~2n!<`doK6U z*$>+^Zu+Y+bFsw^??O8_j^pHyy&yp?usJoeK0De}qB%)-IQkgvV=7%t1uwB92W4w7 zodhcdPEf9-#ftX;5`|22p*ZTf&?(_K^KI;KEQrq2X>ciAIEkV3gwfPpZ!ciZya>!D+UtVqC8F z4RYvI^n5gEE73r)&u{cJp%Ea>+c&^8C?@2R@rhErcVgq~a zExecX4@fCb@{6WfcFYg3g~k2oI0a>!Nc3vvbg#(_0nu*~Sn0TMo^I{;+Mv=m+Jm`x z64g9RMgGfO94f~e0Hky!YJi$QiTYf3YNqQnm>c}Miv_$+YnN$g6F!RiT~FbE9!u*9 z2ZIQJ@14jzTa z6v+%*r3nX3PS=HP!4eH?GkXaYRD;BJJ#*`5wHjO+k{lineB?v8%yy+wLrpE}5*9MGJn|KQG&{9eg6D?;UMugUsn1XGX!dy7WaZT9Bk5o4#;zo;3ebutJ$374}}yd z(0BNVAG+#^88)RF!f%4g|h!EmErCEeLu0={&PQq?>x{tSYfs^iq4g(K`vOa zz3{8>l?V4D-SuF&DmgLtk82AIzsgAy=jo3`T~^A&8~7SDs4OZwb;``tZiG`f`h^ag zZ)J7GtTliaIC`yM*&-B~3(dy-P?99^O;4CL$8S)fgQuA)Jo^H)v@vCgE0P>5#wm&;iDz`Tf=!@KJV@>pl*OLq4Fy*I4>!K%cyfQFVX-9 ztBPH`qo~MCQ9bOJFowLzbymNt)?{*fGvsJ=I8~4}MqDLcKw%nuvd~~x%>?96( zYhA1_E%ewDU!FD)CAKdY1h-cdyXP@*6Zh>lLSRXM;|av!|@4Al-3KIf0>qQLiOkeIn41S(Qn=1`LF2fu510ENrI#0TUb*6_ zhLiILx{3tX&@bY#Ksw}H-{oZiAYE+durd;~wgcpX^|f3@mA8G`3rBrcVKA2WH+47p z6?ZQrc-d0}WS@YVU=plctqr>&r;HFMcWf#bJamBPAw~V7By@!t(PJiKhS9DgLhHCv z@=L-h4~T72*%5S!CGa}^zw_$X5*>(KwU_L9SOPTk+JI2W0adK7FVIwnoQdtWT$T?! z^f7YFgTkyh9hDuK?+qLTK5@mLRF>R}%Rd8dGmxs?tv^lOiL1ChvQD&&A@C8m+LVRq9cXUS+3;g-|N>rNn#ve<0vChX$#(;yB#Vc`L1fwr<bAus z0NH!5?MbeQRJ?DD&u!QxPByfh(;Vt7!@80bYO4uLpZ31*Z9 zY3-=e;P@IRtkR0h>e~gK5SpnTgAHRQt*bpeUNXnRxvUZZi$t;O>ug-pa@CbwdvT6F~mpon5dbHx>7DnL*RuEPCzWiG*^LoLRd=PQ)lj-QOtoO@G8mm8O# zmsP?Eg@SzrVw_w9fzz;p)*XxBmhZID^hP%l>0_-8`QJo`N?>5DhBf?oY*Mwj#~tIe z(l}vizw1*sJIgZhpji{PqFmvXirX}_if6q(E}F%@{*Ceof!1AyZNkyCi~A+Cs|1%q zo6k;tC0&ni1D_hw5NN^CXVigOG5ZIu@iGRGsFA=18)V9ivWHMnL*(WciE!{Z>%DTy z=Py;0;#Q~!ZRPM(qULAVCIK4cv(u5?4aOHHfdS%2QOtkP4Tq_8(LrWKnA}Qensmc) zs%bw0{i`bIjGMxM0I1piEmmAl(?yHMur}FzWi=#}zf>vIy0!(U(CE?sV?>_-*vUHS zajjL1mj**BScFEUiUg#&Uv;0G@<0{?N9N2(R2s+V=g5lWAc?ENC zi=N*lnWl`dJ|Uu=TWE;}CsO+DqE24$e%vge_9r&;EjTa4d(M6H>k=}{u0~F2ArXy+OiP|e&s9W z`=+dhDfv|dn_nDXb{Ji~{;HqXx!ve!Rc-T#fVtjM_+w$CJ4F0U^-@2Qh+C21a`S7e z`I;(q?mFagz$^W%4`a=>zOiJ5V!~km?I`0^r*pY(9kLFPnU!yb9)G%Jlq=8W39SvM z7?I){eKAf&B4#r_bkQ=IPQGgyb~0_a(Qr~JdW}!jlbn7nL_@L@GZ$h1P-#zmABBxx>G<^L z51fZ~(+7^c7`T-?m+EHpi{RZ&tdTrTf-%xTmu%VWQKf8Cy;6 z%Z;P1T{1j%y_TWlQUEeL;~40X{yW^`B22Dz522GdI2dv7mooiu__Ec~3@+0Rue@6}B-U2XR7q`|kZ0D{i;MXmBW{jnv zsZ+@q@E>gpLAAvHj-aPWttS1UJ)JvX>Lf4PqUP&k%H&^t=QkQdh7w37Iu;w5jCSm` zfSbZV>nFu8L!?uxn z6$kYoTIPIU7guuzF%fJUi>xGG@{Q;6R@!zDqLpXmKJ4J;DkaUXEVB|oVV9xDykzhC zck2wC&b(B<;IA>jaEM?oy2Wfrp#<9P8VH><%FUXnhkM{`K}R)|oZF_R=}(E~o`~*z z$iO6?w?LeF;$WM$0M<)by(aBW_B?j8VF% z7y#r6zqF0&Ytw@Tb{LZmfj?150%nZlkq(O0kkH4Qs$P=90w*-inhR;=1*5n*s;osM_c(Q3cSR9sTVLiCj|RcyHC=r8V} z_mF7HEQx9cmEpq-%H?nm!VJV8YYoi5WA9?V>;aIDK>d3xTt$gX`W8Rzz)gJna?tUE z#N0UU@wfcyFUbIT8se6-vNGvH;MG@v4n6m}9soHSz@^hPP8h&drll6(nWg{17pIR= zof?${8belK9D!=W`s%7IA?6CeS3;c#-dDP=dbP)h2$PiSgqp5J*q+$AvUe9!AETKn zp@AW4YMDER<87VAyA!9dMXkzi%He>c7Hgp{Illt&f(wW@dA$;*SKBnpJ}6BP{s8;kfLovh1+@f+>LGu{2j1@)s+XN(VH=0!^M|7lh%{Y^Zh#?V zuf&(kcR}>4odH?6`bs6_{G7PggrQFZfABq-mp^PoHvg6@WEd%Pgg)hef4z&co4{udnP4~yaiP(EJ@;H1GWXWvD){$-f9U|qw;s)JA zWwXPo;$gt{*lt?sEYxqSs^q3c%J$cbInq(f+#!*1Ma#ptaL-5jL}8no?@04wwN*y` zk8>=*AMX9Gd(=bs13Aj#ih6C`eO0BXJ|ywNt_w5?^4MVfJ>2M}X;Mszs-AA&Fhh%- zRkmuvbcb#%7cAfc^7vORBy=ulV7Aygg9Q_V{h{2+C-iWu%;}nBZhLgeGybaFsh=i1 zP?(WLsmI{hs-0-RCbX0l>r=#fJ6+n@_fPFzG4e)N!h*!ywNpR`*vDoUE-&{w1BdXY0GWN59M2hJDV z3^1=e4#xkpK$b6%@jdM7dS|!h*1tGb{ZJxM^`@?f*3cGosy~l0T~A(foon<++vKJJ zZ=oDja9W@wZ4;&$t3K9mAhA)p&+F+MNB%)<|ASmiJnuWc2JjUSE7HFgR=_Y4`z%qUtwGNz zqXdG(nF@(I`HKtbvIo%mhzlJTyJ)FD1P)2m`fMhx1zVND^c6zY6N9o;K&;0eKmme0 z7T@ar4?;M4q0G4mx%(|x8wFKAPdRRG=jPxn7L*?=nPQ3xpT7M9Euj*ChaL!IEcG9V z3-3>4u5uhxc^9gNknzoIB{2)&QT881?>|V;M2|=f5Cl>w{{gHZuJbz8h+|;yq{lx8 zXJrL$^RE2H{enKAD#aoD8lLNB`C$&%5f&;KO+ApA363;#((mm|Bv~uA9?@ej8!g( z{y+Itvitsn)KUKrBK=A7d~ncEwq>a0agWkX`*4ll0DWc-vK4^&i>|HdOtLhf87D3d zd)474rU!)Yhin_~8Xl61vJ^YzCWobAR4s4U0d!?eW2XM&SdfQBr7dyPC11UHF`HSG zXnpnaKI?glf|SyUWO)9OX_TsrcVI{YwSmGq+4M1k-HVgWW6P8H#W&9TJPk7->(~fZ zHf5R9mIib#0hO{N22%t7J?I;cy4wFun{+AlGtdLImG^utfxh`oY;j=JUHR;l6;xgE zw2^vn*E~N_jV{6WS~&_OE657Q2n3qbJXsYLpu2F!Mn=%FI!jj&_hEYV%y&+anii-0 z*G*)|ENYf&JxQg$UbfueuuAF#HNXL9%2s+I8qO3iWZQ9A7ig zYw6W?^^57j9Vh83=inKWlqY86?dHMpumDHISR=wWw!mi3H60!T(MB=6YjUOnm#L!S z{DZG;8n%wh4Z-&|Uc0K*x{DW>?@S+}*}M8BzWW_UG5N|_9wE??=4;`@+pNdqdcCEK znS$L$*}gdl!MJ|&>w9jbgvIPO+8=Cn|8n@xHG_2Xx?<-xA0_S-(<69%F&Y!y-+&1kdt!I7L!te@|1dA zEfyFD)?1Jh0XPOAK;_5?8*9@5xA**3-)53)Adp$TZ(_v$ZSX;l<)Je1e zi1m|3{$7F~h{QGMtJUG>Pb#-1F8f z$)s2zPUhUen%U-4QeeCe>Jd_rpB zl*`7e!tYXc4mDw1thlAnZxsp}DbFoWWMfBw64)1j=~>^~gg6G!l%GBU)row0QC2(m zS|%Qj78_hUlTBLHcnz>bcEC=G-(#%8Zc6@}yx`@%6ndw*hKGt0f3d7>W;2F`%cA^5 zF#8Jmgd=!~EX9oi_4Lg3b6r8+vo1JX^ON_6SIdr`J&+oEvP`r&JuoMY>Q}!A7PGry z=`jtN?_$u5T=#63gZFFC=VfdC6fj%2( z7`&r|=a(sKgCka^z8!4hT0+tABL>-P8F5!dVNk}SozJmGBUB?Qn2cY15EA-pCnks0 zJ13uphIeAhq~Pllj*G*e250}&gku@cyiX)nWY<4kpJXwR9ljw(D?q}Bw;g@1%Mjv=bi)M5IuThkw>f-r zg96kHcv8WREFS($pyhK_Gha)u=Lo;2;{_#I%DX`0FlrHkdgt3g%iaSq;bH!i$dp?G zAQ*>d5+nG!1D^Ns$>bL0t9;KzKIx~pHt3}kmGGUE@GxkbbI}#)U;UdMVpQnMWYF}r zv(DKs$~hQ8v$3ce`<@0VVy{4Iy5)6m=`H)p@c?`F4*E))A4+aQ zHmHp*Y!GA0BXef0LrZPv)|srZK=~s3c6x|JB$?6q)k7!#eH6;u!yUtj=$j~xsW*C^%&MR&ENR$Z5E?dsy1ODZ3i(Y5s6jG?mjJbZE) z2e!R9p9FZ}*(lEkc*1Q4AWmZx(@zHA8Yh2cZ%0;6^f6npR53%E=~ZhI&UQ`OnY|yT z>lkd@9T)&@Zkme+Yhm))tIw%~DZYaHed7RWRGrd&y=1m~*E`B?s>=C`P&_{gb3<-6 zy{oot%@=V~j#TEo&h(Dch?wIusqUiO!%j&}pHf64Io;$odLic|fZ?r|wc7hSu^p#U z+BA8RQGern={VtIE_Rm28L(qUcDhjA{ol>=o`C7XHQ7IdT%5ko|1LESA5X~A6LDvj z%31aqEw(tlr=vDe=!zuwJ)*O-wut`rFut%^j)L6pTVI4^y4=MRs~#YH8K!SN9!}L~ z(?u!un;~N^gnRSii3wWz?%r!pXu)8KPG&@3Kxn zaCk=B^>%6buy*BU(B8&I?I+!=lZ6+@R1obCqW;`>>k20Y$r$CAZAD+zsr3`^QfWZ7 z{M_~fIR9tdPDQFOT96$=EH-&FA~mf)`MSx+7-oT@fCFijLH z?~5=kY4Je@b6Fei0(H(~HbdamGK;=J(vz?Bo682{hI4wy9=TLAw&yK$(gE6GxKaU> zrfH)Xjz>x^5c2Xuk}c&<>^xsnO3D=Le-;q;Ey=7K*(ur@zgBGJ| z0(!v2Ud#unF*SkcIKL|E1FNDE76}E%#G${*3^nm_x%gT6UWEuCy)Z66GG+BiYQ^k0 z4PzDE{qR`b!^^k{=_5_Xq#{L>^ZSJnoGo`P`prFiv@SrqQ%{z zh2n0(;oJV6=ld(W$v!LDduQgHGk0bzxG2mRHTJY@fM~jk<1ceY2>}ro%fmIgU?-6; zP8fa+21&_LMEQUjZNE=!8+$)UzAXBC>G| znTv6Bl4p~y=56i!M>v3iE5iOfhTKouqP=7Hy!Pok6}~YehdIT+=<>?TPFcmvtp7`k zUwFEkk(gu5;T5a~DWsg$hla<;lX{Z?SZo7~ArZH)jDVKpakKy2+WQA10}`h>M#7WnL;RA`)q#{qM#8UZ)S~v%=>p3DBJR;iJSQyS>$x@41+bGA9`r;6^m zqLpb;qglN|rBJ~jZNb9;oiQ!+ZgbDriNlCo!G7odpW3r5oPhjr&sqXmvxh)@zvRX0 zWinvZ#x&|_t$ceO{%4ry#$~fJECIel58?onqm=vSCo8@8{96)0R({HNO4pL$eS*$> zBll!7oYa=iOG8p1ivkulVBZK+{5tKQKJ83IT{A5xBtQ5Xg9%pSs(lgPF;%tgh`-x4 zc!?V{P+JeUsxs$K_1dMZb`r7fJ#z|o&x_wBgXo8^0LDc7oUpS1A@F&1qaE#e z41*?JzE6c;-96QRuBAk%Us@@!Zn- zoU!AV9xALTFSeMZKIJH99}a&a`;9S&18F(sZ-tx#qq$V2IQM2Or}lh_z)TK~652pB z100d@&D)?s9U(F4zl|^`?G!bHq402kI$<;^%+>D)`?TEG(A$q7U+7S6X(wCBaaweP z7Uz9xP5*_$Y*#FU1#4VCyIg>(Zr^N+B*3U;=r8x0#ODleK4(WI2GSkBJsWi$!N-0w z{{!Av`gTuTE5UGIaU#BQmISA2u@(ler;7FMO*fMv^>MTi7T(A91zP>^yNLl4UV;7S ziw;TH>cqM!P(I-Z)^`^1+-e;085z0@*1Ub)lWJG$MGBbo6m4D`BAJCzSKsQ-J!t7= zj2UZobeCR|Mr;3v7OwUA`ee3)sGzuEM%AY{XAN(Bjk)fW*Dnl^U!LK2lY07h))OQ- z3#^ZSp3hohh*SXeql0yQS^cL4^R*-nqGQz$I1PCjr) zu*)}F(%#m%g_LW^-NcCucTmBSBY5aeo&*1pbx-+tvnqx?#N9iOCyk|zn&o3Mfr~%Q z3z~LxI{+L9SLK2F=2Z4SnOa*vGgxb}cP+T!7KJ@l-cnI>Ge2kmVKYA}Lf9yAS!?<{ zoFd2td5a(#;CT87qVX*|==IByrraPbnj&Y83RVWU!ErBL&VM|Mgs;=+cGLaT^t?Ls z>}jjeu)7nyxAsD!%+)Twg!PHwI#D-_KItCk1t^Gp`MN~g!6GN*hGW6|HrsXJzk4N9 zhzYD84GKYoB{CsDLc9b6<^1MAZgCF3~4s!i>ky`4=AX%m| z%<>#erj>3V%eV?^j3K?OeBCT3+$HG`D_2{t6%PeeQ+;s3n5jU>Pa>?(`XMdv-jYYt zLDcRfnRx#eHoZY;W1jXjtiA3D*!%f1_KojTxcjI-9Q*nU*x;T)!zcf#z@ z?ugJE8SNV~yGoBP7Z-G?{$!BtC0Gw z6K>E5if0l^yeqRQ80s=ZPSZ%gtyuJZ%ado-%J8cIJ3!Ct=@M?3Qm!V*1kWiprDb98 zqGAptL&u=XtZQm*A=i8E8CL!qu{Ne4WN|~4H2>Yvr^G|#7qeM2B;A)r_N(9 z_9WJko+Qc&VAq3hTeoidMnzzapRgKL;^~iGQUOVDE0DRT<7pMnuk%ekBb&=P))4VP z?;qqqrl$@@?GyXxCLtv0JP*PPQhHfZXc+B;Th#~FE3e`>sTx=(#35#@->2e#A z=|g4J8!d$dOSy@5yWH}mV4;H~M`1d*n$9q^W7)d3$1Endgz4^Wq(wbeTPGNX$qfqg zzu=TZ_ZaD}+pCIY6T30qd7my9yYl+8-TFh_w7CF-gUVchy~(nE{!y&Nd1VBGrUvI`~Q_jhZ||-sxv)Bd?7wL;n>7i;2l+Dw?@3Ke9@IIm-x z#!*`(QcIPtwLc&1khho6!qg?oUI;4bJwiv9_V!56p!|_r^F^;U06pgSDp7&rcSx55 zK!f`%6f>@xBBZ7MCtQaSvbd5QnLnN~Xz3@x1<(Vv${mXjx{MXvYMrSFzI18SWKYC@ z)VuSkaTH3{j8Ey~@PB>$o5OXS=>vApL_2P^z=fX#sf3s2kQ;p?l(IOb(G$FFH9HX= ztGmX*#0U_o6fML;ya}dFsGfSqh0N$39b`=ySk+W_Ca*G&HVK%&z9VFNTg{f!18?NH zl&(SL744om?&I@YKpyta@NU&ESeA&uFx6t#}^A^iD96sv?l5#q3 zEhNkc7?L>c2<;#vRdzrm5?~tU=^i%0PAwfiiYO@yD@iXd8MMduMOPsaRNI>~A~6d+ zWP0nXmG6e%#zM;AvBU>ai_9g+X+O;Gx!^$ZUQG!4bc5UV`!U>}Z))9NCKjBwuWdW> zGt0&d4d~+m1CB4#+pL(G6A$+ix)Z(z-FIVna=H$xD&W3 z!$Akw^JR9i5<_->sr#dUcGlo(!)yCZwr_8R6ihUM1WC(aZU&Ki((<1?n< z1$gfz(C~E29aAFyS^@XxSl0*wpZ#Q#>Qs0gUOcI@$z=p~=TO!iarv2O=coayVX25e zK(EmN@5F83^$Ffe1IQsBAw`8=e`aA=k+j#iut#C7-M-djv0VjjPF?hJkwlgMTc zIv_~B+8noy(XV(<#~Zphl$Ucq)(3QtVXvJO1$_pxH$oQGOLG!odlGpFfOtMX@8jaI zcoG9>Y`vzeh;JufYsnbVF?~TD_|Hw%zbiSap-6uJT_Cci;^Jnc8c(t7mVcNEl4O|A z`P3|pNGcx_d?x?oRGatJ;=WjVUQI_~{n>2;4{4zl40+NWe~{EXvRFdzZRKs%+7+-X z9*wRzMf=)vSf?*2JVKh;uTgH#FS{V+!4SDlkkxMX7|_@PPq^69T!K(p0B-z^8hcPw z(JVFDc9{HPiBZGY3rNQs)dBT|fGOHvHP${31OV_mww*vt90H^!iH(#EbN;);u^f>k z@xX?9Nu9|jn~dKf7*%j;mH(MWB0gGM#&^-VC1A@r)GtYQjP#^^II4SXwjX_RFXf}F zoecSvsA-o;L&+Rikd8h6Q{~pZnSqFL21v#^2CoXpRj#o4Qq}7(uT%fttz5ILk^=h` zcp@v;1*nZz)P#mi^80S&8STEg8)FO>5?>|3BC2=Au8M6_Z#LegVO!2Y3qFcnPs^}M zMy~&3^LsYSa3TiURdf3`x6-)O{ySe!$lD7gfdmX~Bzl{bdE< z&7quAZo&_ausFq&C06-2=dE7`h>4eJGl-y{Op)-=jsM7mDLhv#F{|OdYyV+YxiPfd zz3CSfvX$57i6Ow$Joeu?abIN%0Ob8ymNrkDtbd5T-_HuXo{|)Oo2%A8b7cKOj`GAY|&;&r`tZ+e8Q1Y-h~3H4swV^!c3)U2C>Y0V2~b{{+V( zjX(*jkj^S2+ZB+N5>UuQt>aV6mMi@ihyl}l-%$f8l)YQQ(+AIkVH*i+HoCoL0SkDu z^Q`IcYx7@St3aBWf}*z@cqbdIK*ljB+jE{3Pn)AAGz`t*I`<*H$O(yuQjZyww_N<* zW-r$}o|`ex2x7zGRZ1d%!V!2m*%^sGok9CwTPIC4^BK+wves>{tIvP$|7b<{MR@=($4k>o>Lm{%kWu{u zUdf?RjlYN*Uu)6bwVMUX>-$;fZEh_=K=H4Sr22n&DJS%5`fzd$GEzir3XJ26;FacL z)boh8YO?NZ4ezSLMUQ=R*&^-UMeInvDCI2S&`JZ36{ca}ES;}XDk@UpJ=kFQxl_`< zPlB;S<6V}Ne0VcZxVmh9oRnFaxzedr)@JM7L4jr-1WD)pS5e?iDxi%B=o{#3cVNvI z^D)uxANEGa1E^^MomH~;Pd?s@9N{^y(L>jynGX{mmwf=WzX7mV!k=~JQL;fb7EV{$ z-aUyo8d)fnj356qyK{xKT}jQsOto}@MsAh+ICMmZTE>H#jnVEbnxXxf^_hOCN~m)v zZMUeNipa|MO=RfvRA+L6cnf?U*Od1wIgo^u z<;6B>it?sBg+ejPyeC#EEpUW&YfMKr8sC!0$%~QE7Uii>RenbS;C!%9%5m1FVf7}#KKz?9Mu~{?+UK3VEZW+xX-dLQ1cd}&L}kZn>go!jQ6FK z)1FUwnvKb|FL`6VdsNlw@Z$<|jwoGA5*W$`9;=U5(Y8u>z7)_&f<9q9L;tQXfL6vE zZh)!u{T@+DVgT`{x(WWL^S3dcUZoT0HTDdMyyq%mxmS~B%O!7vcrqyRY6DLm(Vfp5 z495)|SaRRLhaSLz<-$3Hda6;I7i_?(K?7JAEQ1+sgv7D^gR8{^v})FgYtF zn&1nGbHDJgj`oKmsflax>1M?LTGwDUhCuOa8j4Ax=J7+~DQTX>=l5!AKfS_@dEi~C z&z(6{3emY?$uHn1K!Ql23*6XKGwlV1kCqtMvJ~yL&g}*?2!+!}9BljrEC}tD%1|kI zu322T=``;0r37LBk&7{xQS4j!A$DLfm0&}kh!g2Pp78qGFF)q)sWEP00^!m^xD4G` zSO%@1mibmssNHa%>r-|uM94+%E2*XgR;q0T1U`qam`CjmyPZA7e{tZB-o=N7ocmfi zChaL&_zIEjn@}7#}1HzyT9?r$dh#QdjVrFMD8vRZf8|#`lKKV0`R&Kv# zU#jF#fhOU`bL5n4lQ5|b#;y})dgeZlGPiWI-VMd)5=A^g!`*=*R@S9dV$1b`Ki!mI z-}@26ppKLu`eD9jq6n{DtM4szlj7aK7wWI9cCx;m%0bz_RK*j}TS4TW;)W14z1Q!= z#d7zd|4;V}2#q9A8qb4rJRtm*7!CA^MtUmRCucIgiq70kEYSJK<5?57LiiIIiNAoK-qG=`!3+uXQ<@e7~ z`+UP*@WxAaNHSWH>taCEMCJ+wiv?w6h6SZFEeG?JeIZ=17?q!|)1Gz;bHXQi$I~4I zMLde^N3x3;td{mxvCScE<5AF69%1p_!mnkk61)8CgC(GD=Z^f}JW@EZ(I}d^f;9|I5<{t z2b^CcC=jb=OH1O^|EiU2JWR+{+e`Vgwa!b#A=sxcBceh?mo$Q2_t@(oIM>=^Y zul^n=d+OXKM~Y*;70btBj?%t+PSRx9TXXkRjnbC&Eb(X)C{fl&5+kog4Hw4>OY9mzw=2C-LYn@^9yj_@}&PEK!UstjbV|)$07BkqPupNd@KJ$K#?c&$ZnklE7 z$9?}61^izOZa8N4$jY=ywqJcjMohwKPItQV8slbu)PBz4QY3o4)57!q#Z(jj_!GlV zfzHB?^4lt>12;R>=qMv)`0Z+~s`={0RpZ8PW&24rEtNk^Cc!0!s zNcO{SUY#vP0M6LU&W8$qTdR3Nlx}|_D3L=<$|_>>Lw2nBJ@%9SRR15KD4$L@Coj2n zzJ~849z~THP7KVSzZio?aUl&?<=?rCyh|$d&972d0}d8+2P|BJAQ-V!WFb0{(&BSp zijd7!pK%RHA^sVf$*B1Ruy7OI0hQP*jk=6J9=B~Xd#Bj5V<@*;oCp;|7!TvfRfFqv z7dEkz8mhiDHNl8y@8l;zV(ChJ4qo!6)WGeft*rHhPd17syC2HIM2(oAf@Yl~FLFX@ zv{!^Rb@Cz0h5AOuXw%12xZoolB)Z10zbXk<#~KF~)`*b8ZrK9g>=0G=NA^@d5=U%H zvHT`Psajn{%vP>DidfJL`6j2N<*4_k@Nzgj!F#u$iEpWUilTaXa$_v5woO;%P{ zQ~4(xCy&wH&UK|+y2}e^^dhZ@Y;KP$iWR(ylDwC%ffY5PAy%sQqi>Df&}O9kH)@Ut z@6y^uwGNa2;IM<;ewuLL2*XtfeglXufnC^kItO zEaC{?O1Mv)kmrc+?BsgA7^2|x#~)itYtL@Q=K0PtBHEedDNd&PXpZMu7MP%^IToO< zh)a!4<@M~9q~wbQXE85tOFuo>kWHS=m-Lhcr0mMC*-j zoD=6QLIJQd0JVe3HF;7A_6uKx-}tK^yfoGDP$^TnYNp6-&*-Bj^Vv|y&;D?c#mThs zB;YrCZn7zID;v{0ewT0})bAWZouvMwgUgoc22L^V{PdaIQdq0*PczIu^bkwGy@_%E zeM;cB%}qG0OSIV6h-OEkNjPIsMgnq@Zk7g}wrd2me>*zU-&iQ$Eas=gB)D)DjWP!j z+U!tR;42-}p?x8+<3*hz=v;T*pM4J~FdHUhd19sL?RQn1AKW9t#Y&&wyP)g)%IHtA zF^tl#J>qeM4}wQt@Cf17E!1LYN|Cfm0vl#U>iLcSinG!-iqGfIc8L>QRB|)eRBEoD zU)fd+BUTn3y2Ij{0tY|;7!IF!iBN5u)z81Ov1dln7uCX38s^BTvw>ew&2m!M=hyEp z$Nc%9O}%FL0f+YA5DYaf+xOEy=w?gDkzZ$Zp?B%T1m)3>ND%5$i+`K^%OYsl)f-k$ zh861UoSW_^o@2Ngv+0>wgT*NO|zVnx7 z9zCagQCMQoN_1*k>knCvIpm+~BE98PggXHb?z?|xYm(n-V;_ss76M%{3q#AJZ~g$MqXs-y zDMO&XftJWq&P*o3)L3lv2~~%Y6jpT{)oIAUja-&`nQXxVWuQE@=-gSOvMt7Nb<--o zd(q>b#Z7y6xyCX>ac*cX6X|v$U=2Gr-C^HSSDg+PK4tMqQtO{aB1UOT z(P8UX;fblzLT3w^8yL&*__OCwv^?!&wJlM)mD(E68+}xkOp}60a9fS6Zi&brZCDEA z))F+}PkAGO(TPaN;LD^1l0<@t5ClD>B#adw#V$Zw^}% zFe=}-As{i8VQ?pL{LaX)eb%Yre#b}x^TxMrs=51d@0*MzcGn{x^%_B;dsApAUugT) zQBt4}pK|~GJavruq);5!tw!B}F1q6(zW>@9*8}yw9$02yAmHe+U!wc15GU!w1Vz6B z+Nh;wr|ND*Kiu#7U1?sQW!l|A6dIMhT-QnpdxB+~#8^C%ZK+6H-w%E>*$b$fGOx%w z7Tlh_NTmXu`n}<#1^x(?&U~aL=3lGUP;cyhFi4TTbAh^uE*t$6c4MwRZ4T#A-*gfz z$rQ%`wIz1p{`|1S)318x^S(Ro1&2qoYtERtC7fT=r{TSR{dq8@c`IbG<%9SD`(*&vum&s9c#)j3IcH26taN0j=hTN*%?~d-Ou-d*EpU=a-LX{ z)JbdT%2MXHq;mypproZ3{h!R&+) z%NJ^Gc1aW4*a!MQATIWllnb=vpBqi#CwE4sT}!p!?v!i#zmoMi!ez7)n|%*w0IwzIp0L0S;+Hqdhla#`BaPdUdg znD~UdmL&Jl`InIugtJ2F@|aqESKq8T=-7~X;x@~Pp%=GE7HyPUvpFxF7^fZ1!cFvv zRCh}`PE#kqxzrIjG(Ipe$iaBfh`87A+8k_KG)&Es2h%a6i8j1MX+a;5OvHTdnhK$l z2?dp?Jr?%XQhXOga8etL?yzT;Y(+zi)6ibM9nKeg#`!`fbr%vVf{fn~4(06VwsY=| z#c!R8v=LFWJs|=03fMOO_Wt$Jd+4B?u0SyWk~}AwY;yVW(=Kh)WWIJeIC1Y=74bt` z{Z%|T5p`cI(Kp-(1H>WqkYa@?c~vsg@4r&WQ2>-($Sqhnq`3dEmPE#UNx0IEPA$Kd zRvbx(m*^Ev3Jpp^>aDB=Cz_Nb;^TgDa>&Nz&GWmG+G45c`a^M#+#;pqx3c4Vhr7@c z&7lh7?qIcU!KB+k=BkAW$;|f-Cb04CU6z)KJR|kZSj2PuBz?_)vkYZ;dCOEg!#IO` z$`JS0;5U!Cbj%Z9;=7kGKv%1robMNh;jVe~YGNLS40`v*uH`UC!q+e1Y>jleEpi@` z9-ji`_+LKn8(=gFuMkBnq2I`4yrBAt`%D?SmJh)ytsnVwYj(m|PW$#M{(RSIH#o`% zbE?|Wx?p^oo9MElwICg*_T{d9Z&Hr~W6F=QzvZ*P+nGP*SzTcIwjyWv2BWe5%pFBf z(vAL!^-VQS@pwtq{|?e!y#|%bnfWJ;aM0wsItC(IR}2pVIU!Oqd(vqxLQ^Uk!A2|+ ztQ`dYEo<6sLyb|ZqibOqT`DLrwwk*Q5s<#h8p&5mqlQ;=HoYYc9+!S)>od734BY#} zc?z9lqm^<#6rYe^PHtfd;;^Zulm|J1Wudp9B6feM{S{Z9`u1w`2Qi4>bGmcqb-XWA zN?0*as{C!3U|+#5?xHTq@gNubu?tFiJ0Wj^mmmjh=1)AoYwWvw6LrWWz0H(x|JMri z$-Vm)Xh^~Q?*Ymma#t*zy;jM&t?*o)u9-x}l#=GE#DQW&5g80lx^8ML>FxB|8xWU3 z$&C2?kLU`5Zx-C6C{c7!&qS2sbQLc}pyS=tVC>rq-;u`A21cS2Z{L&DUTjf)Hcm}#yrRtWmh!7(gdeT(Rf#nqKLQd6pW5=Hc%QOJ&YaGoP7b~?AIqJ@IWQ7f7G^=d zZNtLnAKg71UUub?U1N@-E+trgNYbE63>Hf+b+wXAL{;Pbo&}bbDmY+^&$z{6#nGAO z@RX>Kf5|X-w!0Vw7**1VX{G5gZkC|3r)uz%$@tny$sTm#9o?{U%s0(1D1tP3BZcGpV1Zl6 zh?|XBPQ{;2BmmMJZ;)Zm*9&FcA=F4pWhSmbWR0TTHKpxv19A=H@_5wxRZ^V_&CcM{ z)9)N>`boY@6fjUkBA|LjxvG+@1$(GstE^d_nHZBy5+BSj)Uq<@nw9W|k8p9hXw@F6 zgvo(P?BZSfFdUIm#=BeATW6LP2ZFr&510ud1jd53`Tzt{|us*2lLpUI0(~*q-^2Wa*(;uw^ zDRw=aEH8d-8-ibwsVi=tyke}DZCFsdW+GHeX-|GAHBPTACbo;5^daX!ge2OV>>u_T z5W&#|gRx~&S3_qinPKD!{uv5>%BB6vdWmy~k&3@1+R9YoP_D@B-IaMQt5-FXQoN0Gj#1T27{(P& zA%lk!7^{&?JzKe`SEnpoF4EOGahy`bjEZX)OXRgh12m1X(&N`J_X*1fvsEq3dP=72 zvdYnjXj@v6cc2FwJn#1%2(fdO-B?LkA%oN5o8n}xx%0%-$)#EROv(B?TQ<$Z-uM}< z(^-OT`PHrS^cRcvx*rce7;Sn+ysQvBniAwSxk5~qk*y-K@t|bgV~u;J1o=&Y_}ywh z`@7#L9Q&o%;Q6}~9Z~PwJ2e{kF;5SQm#S`>KRn5UggoyxS4einw^YZ0idjnh`N+u_ znnqLo2@2fda^}cU)}N0PT9NPDxb$7S zGQW-nTUA|*oFgidisR&>#Xkm|Dmx@Do7=Q3y4MdQ9v2K}cjtT84~Lh!5hz`ycGr{3 zK3A4)nR#7RjeT654Yv)mVX+HY-p-L+4g;`cld4`-vT#m8Rrnj*Nd$ut>H1MgP)11uh(Z{E%U4bH~aC`&gD>%k|J4Uo22g z%&@iBQALy?hqoNDNZy{-aNQjppS$1v8yN&Hwt6BvZZ7tev*ipl--eHhR^pbir5>g!_2_s;MC z1z<6`9F~j}Nd7b6mo=R4FKGX!NByvc>IJ+YcS}>s|9$fC@#gF7kVX2M{*K9wx8JTq z`1ew0M^+W5m{tD2pqONzyV*zL318O?|G3uF{WX>iojCGP{n`vGR|WM}ytyyYb+H_3xRKAmU&Kbq!oRm@Vr?*jERgrtk2@>}AH85#WY z4bDZj<&VKmOR2Ugr z4>)UDxe%tza-yiV4^fs9oE^GIk^9>mdb)&_kpfN{&rVb7>ud(irG(8EJBno$$cn5r zgJ|I@{@j||QOhOLj>anM@zFE{ERI7%QreaWS*Dq%qKU<$A?;iucf+~#cxAKr)=?qv zx4h64Am%F4>Ck^N`d|c4)7BYz_%@?;WaBwu1=YrEqO$>w*A|SVZ%lxF{vo_Ob*DP* zz1nUU|DTxZkh$~>29|Lamz|_mIlw?%`w~{;(JgjSRUsTb$b3|5(cdrVYpo{;Ke3m0 z$smd$XR;Rpaee)zv#->yNNMp_$-A){@e8Vc>-%&u^X=c}1IPp9V^f$4AA5&jL<3or z;7ZI8u6a9X%bz7K7wW*EcU9tqB#^~mZj|Vj4}X$UaCK;^pF(T&(8Nrd?ag47B5)dX zm*?O|#j9F2$3_E(C&-Bl!8|HZ#H`~nUi|01SYCR&3o4k{r?)cTmIV2f3`qcU&{Fe> z8DNz)FWQk)Th#_O;oE*|cOS)t=|7#>^_x&SdrXT6PqDa;7)f^gecJ1I?vBST?(taH zyY~dqx>X`9NtC{lIe>6yLtSe8lTL;>Z_>xT-5W)3iGziEdN5{EQNAs8qnWyrWyo8D zvK?4jB>6kX09FR@aqA-J$v-jX`+?Rlh$t}jxv~DUr_l`iYmJ!ZQT$c2`Dl_E$QhMi z!FyL21=UB^o!&E(1{3JJYMTxy)V9< z4<{5R{$3@+;q;wtEEDQY1x>G)CT!UM9d0TM7qzO)HnQp=`vgHl9xbXVKrP4byI06}vJ!Bj zLI8&HH$9MH5QzF$orC1`{X}fPP{^SGZ&LZ7^vNA=ZA{u2v>+GVoWV{W9F z1}pVssEKQN46N>TBAbGs>Dv_Ys$_{*Qgct)3Tpr8D+L$k$6FPHoP|4`?zKRecJicg zF25dOdp;yk?y9qP>aYs-4&u}ebaCIF99`kvSn;`$v$=KYBhoV1+h#s}pIo9Uj7q77 z{nUHrbtU+l(CJKkG_Z=|_lNLT0&bk9Vq`U&*V;$OUrGjqSB47w=Df4%czl{|)dzAl z?1{r6DkqJpIx_fTHO*(Nf+^rP?yTE<;DbbwRN^2~QtDGVS$m66aOyKwO=Q7gBvyDp ze1QA&5gX~xsoVXbH#DLDe9jrY;ebs^9Lun;zi-ti7@bDoKORyX}Pt zVkcf4r`Y6>BlPBxuvOKymBxrw$-YXL8j+1JzMSCp&wHCLZmX{}?oZfUZq5TVd&g0f zW=PlOr_2WPRB>+t*M6(;5{$821LZa#B;qM)vES7oR!$73ahe5toUJkvn0eqVn0BIq zA*1Q$!KcXh-RHF5zN+@^Mt?z=!O!WCSB&(85}PkxPiK8g_LnAgf%PVDzkk}cUIKdl zKxl`BgXCVj(U~zo%-;0dSyxJ(Eea9*Q>h*BIp0k<@i`vJ#I;jsPCJlzBaGW&c7b4RD3fw5&afkC-E;dSQq9zYGaSTLJ1oSGB`<;_Y!IVydgHs5-j_?3Sv5@}55JL3Oa3ntzK9Qf$MRGbevE>Rr` zvljtbt3BF_%n$`^|IvZkfY4Ap$$NF~9Sc~Yl@I)fX`q+zwKQ7ZowUJRXQ4PV4{no_}>|8hl7YUW=&rLNm= zHpPD(O-U->s%g>0zH0e|jm&w5+rV6DkGbYh{(E83EvW51!@%CnLrx|HICaWgX2pn{ zNSl8?{`m~9&aXh9Yt3#`46-gTTAf}28`mq}Wur@MWML!f{@0SJD32OWZPIoTbN6f`%<#a5GmT65=T|q~FNf zYC|k9MQFgU)mT;^Gxj&%Lx25H8^cq4c=3!%1(1_<)hw;m0B#^!s!au9S$5Q1k?iEC zZ0wPU3spC19G>^bs)y*Hkgoqq1^r$cO+NCuZKL?a!guUyKlYOaQ?q#SQf4Npb47x5?cp_V&<|r6(l3HbYjhMK@S3(f{tLu2wh17;a4+<*Nm*`qT|m zWA4BWt=(3`H0*~Y0dTA50;9FjfbhMLN)fwZoFwOXo`6<>NlrTdt%UhtV~V-DU3lR zC_ajb>^iNCV?heYc$YR{gU9r9m+3Z6#zxLzTZX|I%NB_WAK$_{75#YH=Ly$_@~NoR zH09)Sn8u&$yBfvsHXY|;RFjfg;Qh4KiRP^p#he#`FoX8D_1EyVTyr26sqHk%Z!a@C zAm3#L0vGb8-G!E+=J-|zPF@yZ`;&80ES-Pi(tTM%X&Xx`>PI(|la*@X8$25`wL}Qr ztEH?NS&qK_ZsPa+KPDys-TB6Y1M~HPfqf{B?&0_q@8KITW~Kf!u;_OVrgqi&!Ke7` zhedd472|s~Rqr{w-9UNSL{O~BeYeCzGce)g{)t#Y?&!$?m}u;8ddZy%!uv_9^*C}u z?{paX7w3y~{q8~FP^pP1PHs{c0l)_S{YHW;2xv;qmcydDlh=G3IQ0CdZ#yrgLrrsC_xaB z5BXSK01SH#3+!a2(roKq41Y139u$fIBBtpjcqliZrvD?CUlQ(QK0~q>K~NkNGDe;c2_mYN11XvZgbhDq`=Fl?!#>vp3VQ3e-w z>WG@L|4$TT>x#Oi(e#xVG59^-Nl$KI0+QFY_0w<6HFL(1*kety$M%W-yEA1Yf4?}s*ClY>e0ipu&X1%q54>VF+*B03nDx#r{(k9HZ3rlYCE^T*M^u! zcH<@+6q_z4s9)~4;8XKYaANuFqs?_U2Jb0uE0A@loc6BpRA+SP^>n-YU1Q1VsjORHT2tpZ3__Uh`xp5(!fA`PuGS1%# zbx@iTuuW%x2zJh83z9CZxnXDDY%h3l-KeREdNXwyqb+u{@?W||Eaac`^aMPhv~Uw{9-=!#d#~pYngNbS-!Tg z@d@vfr734_MAxo%S4T0dTh6^sF6&g*168BR6jI;QYGBm5J*gcwmS6il-O&LIT#l`h z-y>V46gK-%5Dr7AYSpJ{DC!#GT9m%WZkpT$sj&N6uQj%ez-Y)tkjCP(AQ=LNk=Ur@>Mk^AY5`x zR1zF2)s*DhNh$?<{JM#gnwJv06Pp|F zToXn2t~OFXqyAQL<(lO3G1UgkmHoYcDS&LoFtOwFQo&}Z>Y8f82$U`u&i?Oq;p*s! zv-*|hR(?~R1sQ-K=*p)B?qXZ~8~Qd>&Oruq;6xGWwQ$hD`rBaJ%3mr0_Dw z+0NZRfvyJqN>^%ie&vM6>vke~83Ah&dyRc|#{Lm#yMB=;N{2j5u#&LFX=w-npaibRMwQ@)^V!6%2caZuQyQMW#X)DaCA`S z<4heFarB-V?oS882xw~3k%T52n5K0jJI<#u%?3nl<|LK)F}Wq zI4(r=-hVCp->W|#dOHs!*6g1woh`?s$gBCHjtPIyLEXQb?(OL=JJ`I$nTZ_i-XZA1 zqqO;;wKE5w9=VY9wlWWzMH=)`%w9Ppr{DoMmw&mwmfx#=xX25f8jOocr$Q-i2=$MU ztnc9|o-WA3jTGU){bihYT`$$sPUqL%Aw7&X(EazxyJZo1*5>=4E%!L&qxWV8L=|RB z%QDAJv9W`^0AXnN`DFTRQ1YNCj%Wp-+sbP@e`_~yrhXxQ>_-t2`oO@A**K(N8lMNI zu1XByp|8XzbNzFd;Nz8|lEx=7e1Jo_-NGQ(v+cyfmC$9L8|2;4X5QDkIr!|jtPnK*09 z{Ptj&P_+rKn05Y6!IQV~FUm`yh!sg^f>z0KERpGBR9hEMO+~)6*E_6_MzCmlm!i+pwI;VZjyc23*^AKklyR)Qt31tjM$wi`@iAtZ2h=I9sopCL$X zMJtR-009z@*{gx{YLIp&ffLi9nXMy{)=!lRp<868;urK#r{+VdW+_H#otb)nAXBKD zgw$ip53MJbWR6bnH=T=S@P4-@-D33DK8jfNcT24=t{XxI-63D3LQ1-`p3**ogc%PFeathC?C{qoIVh`^o-NzP8LZ+ zL@Wm21kwY;-oKYBXw&~F-t%O53!xa|?*`5X3-|_Z_D?^L@OJ@rkr?Ok?S6=XPz$9g z8Cz|Rpv1fzvqgHVORY*U;KZH>Spq-oIR4{$nuA!D5U@a`2sFP0f}n^>7Ryqp49Y-w z0xPWJ%n<-Zyurzz(7SYzz98kAj<*DYE%S_HQ;z#&t&`Lx15qBOe%4TG;KsNS#|;fC zIvf$OL`>2L+|GA)<}NhEtAW_4;eW9aw|+kr8OpzVZbz`-j&8qzZ&n7D1mdrvTOeX@ zBBgrq{HBL8iZm+z=Tw2DWjV7y$VJ0ACCQEfK=UeQ7$H>BC&q+y#)sga7W7ntm@+Tp zN2^=#45z7~`uGvv;p__@{pk}Ck4he7_-D(%z>r&02Ecj3Pzk!`FMlr@!eZu3^%hky zjarZzGe6Wwh;WuxD7&0J1V~Sfivy7>m~^;m7VxN;Y_(S=1+9-jk_^NMsZK%aOm?3^ zg4-$?K+bDXIJte2%QVW)6M9KF6rO5L{i)5UVXlz5dW4kgp;UXh%F!;y9N`UyeWt=U!WHJWym_t>!y zNyTGyeIe`TdPM5Bev;3xhmQ`43LA(+f-KhvbC~(*j~Ppna_Gs1z^l8h%FRC08Z%IB z@A$L1jrAXq@rABXo5nY&lpo77^Df1H*t*1RM7>i1ZcE&yhSSYb&UW42rJn_T(*#F? z89^xT@zGz{+w2|cbL3yrUi2{FTX8&l$*A%!X{sH|+>vHdp>_?fnLqAy%HkywcD}g@ zJb3@?EzCMn3~iB!vAOiR0SCXbown>eA>(aitXj(M^=VESqYz@E0k|{Q%|t z>dOD4=_-KQ_`WVJ6qiD<;!xb(T}yB;?ozZk6t_ZfcXy{aK}vz(lm?1xix+oFv2XeP z=bK@eOtQPP*}Qk}x#yhw9@7^Q2@c7s!CJ%|b~}sp)dc(3_FAjGhqPvBa?Tu7zHx~J zypOY`AKMx0H@Y%&CCB$(ehPFRS$neNrVPNc0)3i_yH()B`%y*46Tfe>vGDJWhAfW& z^EGpXl97&JUrjr2R+Qh4`jrfiHxEBF9?a`rV!>wzX=6Au2&JFyaq@9V}`(J%4bg^QBUC0}zBzpNhH`n<3?%BhsxaL2#mJmv&SGae=kLP+2HgU1ZDb(VHy3mK|e zX~t@A3^&{)#U{N}OwHnO#qjhc3-87m6wNQ#WwkiTDf$-n>=hGx*>w9;kn)f8BI5QL zBP7~9=S1W&_I*_G(vHLWesd)IyVn_)9b=_6c@ zg@8(96d%KWcged@l*7Q;Ew_@38!k-|Acz|BEvoWDHTiEtUjq@jjt@F*=a%T&iQEcm zJEO^{~yvWlh~#-g)^qf755iBl-h7f`+f zcGIOf|K6#_$b}Or5QEOHvDxH7#h~WF8H!Q1NG5hh5hUNEs1rkl?7&El4@*TB+P^P| zs&b#Q@d0o>K`!8YfWz4&325TfqosA=bRwb6FQhd4LmtizrNu^ca2s^LKz%^tbtQb% z6sW0jjG8ZkT(_wo0h*3TCBPB4z!<*SdM-5%8h@t$+>N0lkV;59c2P?>*+qBw1$+H* z(7dOKwEYW9FRx@(7(Vp7MGoL#%Qn2AikP#?0SE@Su3x(IB*UHVJk)6_b`~`xl6-(( ztkQ_+0Zace^kKnW^V5q~eF@zyytMrFcSF%eP1CJn<$3<=YGkq!|LCY4G(Kx@aM@#3 z0rmf#cZ!$4^L%lH=g+QA)df$F-=AuI(LNM4vILt@OQtaqy!_0R<(4A)(pd+5*aj>o z>^!CD5B_nPo@S*GNa$HR;`drEgfF+aTaDHgox_+u$scE>IF<-np2!$RbA~pz-?1h3 zQKae1nTao;HjvYL^smVQRDMDqTa7IBEYs9^lFp0bz%LK2hSUzS^+NgB%UI7b=FS`Sa1yjwt4?w0hK)UnRl;-i4fn&&v@ zN`mGk|D)K5*igRGJHmO?{bb&F?dc)w*f*vwJ2`q9YPs!l;C!8RQlU1k;7gHXFYTJf zjKIl46SBA4z|En>X)OTEa}KL^Mtu7Jtf{{TJW~5Os*t2*PpKL4?j zW8v_ISDZuOjzt;5M#|3SzmjCc=sS_ zE0t1MTa(kRGhorzb{}pEOB# z!aHfB-#cFJu1NB(n!rP^#Lt4i(0}7qk_Ymlp3MM~S{;o_yBDAdjLdr6N!Br?7;`t( zy#3N_>g|>M-v>!nKb?4g&ygo+{d8jeJ=d{A&~@uq`MXCS+N)$L@kb)L;~zYJ;83{g zw5$P(uM&#hnZWEt@c0X}=d^_9vVUF9)IqpjX$8ZNJ{Fs8GY^`BV@@drVA~TPg;0Zw zLL|Ptg}(ULR)9C8A-Xww+}?@l}}_B`?YqZNOom?AdS-UbS;w ziRv7xeVh%3^VjM4qWZ>`ucZUa2dFLtN;4cH7_;F@DO6U=hTGBZLr@RQ?rOWk?Z)f; z21{sjzBb_Sh(2?%E`TRqc(kq;U@?G*$$#&+#hkMh2frg`G6Qt9xqEDq3a`9g<}E)U zP-`?UUr2cZdgEPa_;7F*Y2ICWVGz=l7v5ZD7|P~9vD?qTe38Dd{B>W*Uiap`=O@j} zRcF|CnO`8H?WU^i{l+LIc3l|0)^z98xZl{G0>4n9cK7qicdi*;Q&k4lB~r3|3O?W;FiQa-IR}54#_n7u&_TJ| zvy!|umH;ZNB5jP2R2$N@rJFz$y&~~M;DWKaPBvuNA{-Nk-`f11_a}TwzqW9&91O=} zHvHF*R-b6i+xGK}p~C@z6h0<5LZ3ZNcf2NM^X&!D7v88>Q+Kf!ACCJCcG*3qob`E$ zi`vm?h>Uf`(!@UhpNO@qiBNv_Xa3^rJXp1{wgh;esL1`{yN zPs2X^1-r%+3t*OJ4i*ynzj}uuJ)nnPhkr#fG*GP)ly9z*%jeMC>)i(F{5h=M_EurR zeD76zXDgNY8-TIe(YNIPKsJ+J5FF?q5&OE8CssS@&K$2Z<{BjBkU{N;|4KqCU`fA{ z{}15OC#ik@*U$XBYYH|0Iwo>2F)xYnAPfeYF`%>n+T=q%oa8~(VrJu4;!8;);lj1^ zOu=zXMOzudq=F|e)=@4T)GjyNLul^@y`|vu@_ClmdEe*iuWLB z_jn0L3WA*-*eYZ$;4+z?;iIBm2VKBf!u4>n0He?JljolB7C)~ao|oELy!lH<<}5$+ zdEHIvbnh4p8A(#kxD6+$cwNW+!r)7cAerV_&JVhwgGM!R5>ti_e;=e6AKyFu}#cwsY<)!ZLeVn6hNx=kg4I#A@)Nh-}CduxX#jeNzbGn=X?zrNk0_YB%!L&krGgcpdDPxQl)vQ=u+m=09B`VK1Zg5Y(fZ1i5#uuF2;JW&ZIT4g z>|Q$B((X9Ry_FAX4`AP?iO7L2gPB$R8k?uz!qS1))tEdn?!Au(1HURmAXR(nalu}d zagWy@#t;)vlEB*CkA9^8F zPih4LjdHn5)wMCq%jkw@Zk8q+Z{i9`FOo&jMmES;+#tqmo#sR;8~ym1cdbPwb}W!E zIP~Hqz8R5l;8D=BDG<$4E2$a&A@DZH_q>MC#(rW}Be|rHQ)9K6BMt z9o$T}WE2;Idc!RQW0;*dGlFT|NE{b?YMZiAS+a3n+swBrZyriw^~`rj!1na?gh>qx zobe=g%`6>5_n@Vg1|ld-ac19URO>WTDK`YWChI|6d%{uftqpFRam^L``W!#-R{-l+ zv@jzN4GfxLJlERdNleM}S@QI_x*Z&P*`hD{~?Yu2et8#$f4ku?N~onDrT|=UFD};pF^I(-vdn5HCSZ{gg%fXzXmXx-Hts8&wWZ(*4M@)cyE z=}tW(bUPji68AGP_vW~r4y2bSDCZnqL>FY=`P#e9a2l3kP6zoYkzKY4b*Azj(+edK z*W1SMp)jE)U2-~(K1+7jS4%nbK!=2-PmO)uhyzmDuUOUle;=`%eG?;jXvKr)+_93q ze;s|m1tfWphO5Hf)>|kqnSr>g7WwA#Ky?zdGud7JQ8yiFJF5e~N!8CWSeXg3zYtSC zL&_KG%o>49+v05?$rcRRyyL9?!0CGPi?G_S(CI~SuRt|5Aw(+^`Z@WBIvvRdZZ$dQ z)a#nc>u)g6O%uX6ak$M)9!-WI>Ji8OI+NSz!08Giu|h7==r+X$O_b*hQHJ<=k4Byk zm=uF&Ufs<~2=h{wL0B8ntQaJsffeHN&n}-z^?5kgmw4pBJeA z`@qW`uFMOgFxdq@$uAl+gswm;fV<{H{@-2UJl_?4H4|`GC6zXP0s*q3H2<9^u8?LI zrK9Z!oW`#%h6r46&+3k5!boNmO&jugPm1p)J+^H?rM=mXxPM8-kdP}naD$qDE3z_M zj*KqBz#;N{nDitTFJ?*JuU$tQs_nr(I2c`nDvr*q+s0k&jh502nrwdVRWH7PcX@uE$B zf>3;nrM`PiTeZ5(YWruPq_ncg?kU=upzV!_wBHJXQm<6BiXYb>#6K?}V^WQS#}6G? z!6!?Ix?Vh{Fe>z?u{M!%+>+X(l>PGC!hd+fZ*o^pFs~Zp-}*M;F&?Z7GD1Dga)4pE zs(w8YKP~VZFN61=23A;BS4rv1F9oGK`6Z&%9|>}x<;irt`wpSym9u1-3#wQ5_3CE_ zW)|@^kM=e5j%^iNKA72tmwI=rMW=G`bob=2USO|V^66W@P-nvDbhTT!<~IHoED@@7 zh<~E7B$~O`+~rdfJ~5jrDs9bS5M{q}JvB<-c*&tLX#C~pY;^pmDcoaxNS;3#^YV5y z_R|{H6R-jGi$J4N*7mb^ynpe4@$Y^`o_tYLL>A=BKKbeL=G8V^qV*%;27Tp~(K%sv z-L6;KRK5pp#rarg$3m;b>*!TrT5sQJg>e9zSNblb5o)8kEmJ1_x8%S z3?!a$2GurMU3I0*ADTPywryuBTh}Q`JaLs{Zz+bPB{GS-2ZA`anvm(2S~TKVXgUB9Gm+3(v?HL7ipv3?rd8^FHKy>*Uyu}Nvw99S@g#a4Vhgo zS4uyox<>!B5rhR*a3xXIr0JKOS2@&hxg>hCnq zGcpZOtQNKgI5%YS=x6)m0kMV6lpUmfSY|K{{W=o)=in=WvjT)S9(WZ+?(N+k>p{2* z6Joscp+iAvl15c#7w4TFzmia0-igj!0PBzjv8HAlRBwAVZ|mwcZRtXVOKVSBSM0%)ZAzqi|0s-#zt(J> zBKbog%FsM^S5h@62HJuFLrVT=|Gk9Iy2Qvu0=Vr`dp=p}B2^m=i{0FMn+v;eR3#_N z^&yRabVsF2m#G`du@fkjMBiXi4?GQHD)Iw@b}IRm(L)7yn?;*=Z)NB!wo;-0ENlIq z{7;)|5L|7?j=Hz!-q&^N1ln)%%cKXG?byo%(Wnmxv74ReTD&aj0^R_N51g>+1te{e z?F<(_m_!3V<9FsXtO@t(;*SrigV#afV%R3`kE5mAw?N)=#>?}kU5{w5Ct*3SfSr-C z+R~}2XwD6i7bSig5$7_vW3oWHRC~!AW7L!_E%BjoRoh58H?l{5IB;=hQV#05gdxC+ z@uYSrXx(c|4S4cp+D}fMU0f@l_>P&ZP+0bMnUYb5G&XIPAozkf^{VqTe#;OaR_X zoV(Cms(T14!1iq?m7Sm^dD>9zLKnKV4gG^@*NcDJpc6ejOsi7GeJCuu{b^lPUOET; ziYB8YX6Xl+$cw_b&wZyqk-HXXY>?vC<{X7xqf2XJO4iC{fl*wZY+HHALq+Sy$mW{v zkC7%n+;#m@#^Ialr(%ptwbNTd{GW01C%%sO&`NV{CkC#!cI<}<^kPD<=q=L0Bew~# zH=?to$+o}$bXP?AQ9)sBV#5fZP_w<8M{ZsxNQxW3_EQRo*WVDNPl=Zxm3Oot@8067 z$q4Ji%_LirVN{YIP8%4bLws-kp?8!dyK;$yz9*B`a;2U`$MUe4va1iKBB!%CzQVmi zJB-W8MO)VC**Gc+q|uF~9ZUI9=1fMrTn+9Q^!7qmw3wwroaQuXaAZ#akFfu$dPD1r zmS07q@s7RbeO}o6VJ%CnR!(0}$*RFZkJA5{;9PdWSC|Xi4ZLO+G|$PIBuR z(m`J=EtXG<(5e2Z_?T5A_7FpzAqF=}6K8S-<2;s`6}(p@c)vE*^HpB6Z{bbIH2RQW z*-hgQqkz1A0ZAS$gFfPlOC~RfRLwVSHvlbEo867}PDj>Jk$@+iA`Nn~rkNK6%YGP0 zX^Zm^W!P1eY*mxhwWo>GFl?rZe^zRFAknWgv+%BCM*KwdmYO9DXk{g7NG=#?{t6rT zo7ngXi9S7CpcwJ<-W&m^YlS(|>|4&oXb|x9XA}B*arwmQHRGyaVE^b*-hkX%SB9}< zjn*Lp;+-Xo@MNW+GCsedG?TFAo1v84ry8ebpG2k`U$m6pcRXKBJCU;6^edbGzYWRK_i9lr^s@=NATcH>K za2cfN=yW@hz(#5FO)AOTyqT4&w$r82w%Z#WUg}375L^B-;ob$S82+UfJbGH^2xIv; zf8R?hI1p=_7+huX&M*jH>*}kC1n(GW){QBfG(xL1ZOYDWw)x&Gyik_X)xYvWY6rDU4$*15$~t*m371@&DEjbC5OFSkivhnn`?n(0p(bBuS6E)xm?_~V+(ly6A#%E#YMx{61I0}zN740- zG(_K1ld^z&RJ-U+({3Cz0#kYl~3duvD; z2yCghrxpNpZZen*p8E>vS{$xy4O}WM9c_L8_j{(W&g%APM>Fy%*(*?9Yae~rBCaYJXU(`iq3AYl})sl2jAsX!AQ8Z7CGWR^m@dw%@AaCP%p-54z=wo1YuY` zyjI3MA3vT;-98@FZ+QKSJ95}D_CV2 zn^suThiiI^Q?_~aX*WdVJU4&!5B*U9rk_!CcHb6e3xo1Iulth3#;f@0fsYWM)Sp7f zSv4$7E(d633Wh>D;Si^4pZOba+-68S37BbDDYLj?vrKb7P8Bl zNh1p+wD=M|Ec21z!qyB3rRgioL@yW~6-V6ACss0(rm+=uf z%g(L(zKA$S02x0cD4`O`@Rh!_1v$vc z!ay>RnO86pZh1E_?S?O54OiaGT=7iue6eC>MJ=K4V*8c&r*Y4&9-1+7dQo(iOPg4g zE!S7wYzc*CHwq z{g{ey!sGRL_`+W{|9FqvX1|v~hxU+qUd5klu*=kFsIy*8L=Gg0<>p)$wt{sIVS|ZY z)T4j(GStNw5lzV}h)-=7bSpBsh(+GG1)*?P%lg;zOql>+;CbRk@)~6Op*Poo)|7_~ zN+=Q1u2$O_QY`}O7px`7j4|CHXTE=8C%t0cb_s(}yDnCCMYWN8ZpVn>a4n*M%1(;y zbF}Be>VpV6%aG~al1A&&6^%!;g2=m_#G*(8kuJO@EuVk5Wu!&r-_E_4w7#hSZW*O~ zeYxX{b{Sy;DD*Q1xH6GmF2T8zCtiHWAUIvmv*sdJH|Se&l$e8JUuO54`s`?#$- z#Xn2`$!2R5iPG>IfLb`<|gdjMIG+9N%7}mE~cWInM8nMYo;a&F&Sv&Xd?v1(Vr} zxh`kk6!L(RA8$<*Ajd}nR%>bO)UW`L?9t*YS0(tn-&vO>rv3=m*|GEYgk5DtY-dIH(c+5EA{Xr*s5j!vy_xwgf$kXjsJl&<~{9jUHynQkM9Z|RWil9Ufk zMq6Zx#)PlImv_jcb4565BBZwS$$jsbK9lo}91xK(hgFnSfq*g$jFB`Wij$HjkJr6 zd8?rt^Y^ymw7&sCR!nv_5M6D8Bzx|sDJb>w_2Fg1+H?j4CJweAWL_c)*z90n`t+jZ z%{1$|3XB}?eR8gy6q8weD0+V!P(3nOdGCF=d#NQyM{(v#JLX$JbCfKoXzk?k7`sBTr7l5ou4!PC0Bhz7UhM8a1n9(?Ai8^O# zU#od*K%d@#n?+u@DhBK!e*@;QKW3WnRB~M6KSi=3ohv}a50dh(f{{<)&DLjYSF^d? z^h&oLp|H_96zVAc`GZ$+Y1tu*FlIbbRSdI~o-a~47wVqv*4Qe+ha){Gy>VoE%@t4M zJ)ROVAEgls@DF}}D;-Cxv8;z>evR8{&Yss4PPfc*W$t~*$~7ytN%Kcu|7u5LvlA4V zG(QJ(T8ogA$XsZv{+l7|^A1U&(2FMd$M3pL(B0HLsx;rC{h+?SlvUFzcNGitCP-Rj zL0-%m*y+1%<-Y{R@g{op>#xI5>ynY^+L5$ACf}t4d-P?HPu&7f%cCcsQ|n&WF7+PA z64sK4Q_o+fOXJ@kQmWb_cx#9Z-d&KzH26{{aMgrqGi3T7VbG78y6{iu^}ytX&6z3x zeA=sMs7cjfabxo10$bCr((=#7X3Mc}oiF+>2ZSP*z<-~ zVr1wkS|TyYuRMTxz+BS4a^5jL#>>^k7jMel$j|5qnmJ6r$A1i6d>9=Rp5)Zh0f2s~ zefE+U{|B!tyA38sKK0=zEO+)s^l{o@*_w|ld^%l?G_89&N6vTN4ZY00u!aHr3_fL5~Nin2S-m?Vck1Ct9!w~$C0qri& zQSwvWhGn_67(pcIVW&Gx&q%)?HT4Ep4v)5$#ctXsUE7;b8q>ArAPjdocDY?SWKCrg zEBKQx;}!CICh6#&7)*b7S#Mx8$;+cx0ki}Wt7o*Q86_WS2+>EG{4c7a70e&dz~ zfYY{g2|EXt6@6Z2y>Gv+31mowrcrJlf%wh{~%ea-aKu*Zfr+^<`D z)+BzB)oqIT=^^>k?bO5mnN+Jp#o6Jp;OK9TaN*J=E$88d{Vz1o&DCEHJqwKIc@9!Q zt=^iMPeat0P#hr4u~S-%>mDy}t)#n>N-oePtrzv2Zu4EB+iQAVd?4qCMneg#Ml>5| zt6V_MSQh^l8sZ6asJLKt+)>nmgdg_&p+{a^M)!_JlIU&4yYabeee z9u}Y@wi9mZe!^rxUb!ijZIP%_o=2OYMwgKM%sXw6MUbdv1~)#mY0ZJ=N{YN@(f zR!nO9gu%B0l&STl?#;+Rp-bq=GUAu+-MJrfOhX)i5p#nEktsZlrZkfwQqxI}t+@Ix zxM;187)GvJf|eiRwNj6Hh#ReC<<&mgyY7u0v9a;8QdX{{ky(FaFY#~U3a>EjjVV@M z|6XDT_Gys(JnYuqp zZb^7@sPZu#Y}l3W!Qa41HdT|9x=|kD@4gl6D&>a_`{TNz`7ck@D3q#xQq6?A((#qx zadf>IQW{V~S&IPJHC37V+MhnS@&lC#=7r|8`kUpsX@YTOO{%${)6P>^FYk|v2OYY7 zznW+dnL=Yb=Di@snLfftWFYemGV(t{6l_73;=#q^rfE+ia!%Z0a$=8KE72tf$TfImOe$wy^#-T$5@4K-|yi zybs@}qwhJ1)FndzK{1Q7Uml6*(Bkyc`&6hBZ1cxG!wmoqSFG7)7KE{G<8^gPL!cLw zJuRlb{g=T2MchlOoQuwSu;J=S9z^|0iA=8UH~lSP_r!-J6Zl*LY;i=eH3z}>QDpM-*;S9K;0}kif!RH&Zb8pnV|v_*vRl9S_+KJyc3XzWJ$Y!zg@5L z$rB`aKR5XO$v37wMg9=+neNA6p8Z;|c>35qSi>Z8-fG^zTp~}a_SL;mvnGDU*}O)J zW>n?x^(zLhoOSW7rMaj@Y4=$+7yqpS#SuC3lBOi5>NsGO9Q#T$IOj5_KET8z=dopD znb+oE&F^L@%d?;7=44fjksGNClW$8PXFVK%R*?v?wEHpt>h6H* zpa&*y*SiL+=mALXb^YIIGJFu1TnuvL*rCV?xA+^ZjP9e-*v5E7hC9D^{_$+Rg}b%q z9IO_B2MQ%zs)(RlxIS(f3*nQi=Td*gD>GdIuHMG?ri6y5f}P`F!~CiWzx(z`nr-}5 z%yP8HG9>}DBk*&9>Wm7*vAi@xz4nyYoqXHvAb}G%MMTRN*&;9Is^>JDOC6_>TE3vUWzPOhCzsraPNh5H#tBt-4_$ zAJLiw*%o~<4On>OJwLW+eS-jMU6*e~G_VEQCSiznPkTRrnK37oy8CbQ#q1z>3Wpv& z-Tu@f1y|c(5HH1)B`xumip!)?Sih>p?Uve*!+>QG!@uOagJMdL6`Siyy-K}XKq`~+ zM5(KCILDh%gZEAI%jF-QO5&>GXOd?Vw!c?)Xl@&ju5^@ebj5#=56BvK*!4z>40Gn> zs;@60K>*PkSAKq0qR2v;3b+{`h}oxHxwF2>ZpxgCc7Lyil?Bg7dFGIL^}*miq4t!@ zFyB!7zg?fNwQ`dcer1Wfas;ug`i&Sr3&S#%cwFm}eCYIEDC=V#=_9vqDs_`Yf07#1 z1PEzNkdgSlTMKsagQ-+?i+pJ*c$<=o;B?3&hL-XEpPROfS^rjZW5@vF0nAZUHe|LUMl_2bg6iXnsbgmWSVF*b~^3CsM8M%p}8{b(f&9&vN-w8W_ z_viNYWA+6H!8y)Zl-9AevFOXxUjO_06R)YUk_*2UnnRBB?LHc!pE2?j4U9B2&{UXfg6$wXKQeB4+zyldvC3ElD`TN>eJ0jj{2yL75y>Uoj=G)6?PxoKsc}#uf zdVKf!dIX?R2RZ8!_(_18)?RpsoHqUQBjmFoZAD}u)%)t{jYHu&G@MEs5R>LL@ZlI=xTJh z5)R7tz3+o{DS>q>TGYMAOPg#Dq-HS;%z@CguwI-q_rdiAPo>^B7=llfJ-KqOs!QS0 zONi8lSB&Zq{0>p(!J)9nDBkMCOufwc8q4x05@>huMjdcPEGg2pzz^iud41{6&~ zc0Kfq6OHD+xM?FEbb#i#Kubi0MKcHx`St!SROMPRkxJEXvY@J9MQMOs7)_!4VFiY@)O< zc!Tw*Ha;2H2{}I7=G~8AxT_kY9(6B5vY=YI@NM!v!ZSpxg@%11;%)wvq$C=Q9k%;7 zFexUB%D~THIZ(;0Q_xD~oyyvL?l6hFbhgfhu#CocW?w|W&_oP%*!z0@E-){tP=kXO z8IoO$u76&^ZOum$4x`a1QEz+~m?B?h!B&^!>K#~(c%${*sOAS}jT(^FRaa0kW(hKU zunrlo`QYXD%_DTGZF^gv<>gq-(gl4aq+5~cmFdwwqFdo6|3wh?e!M8Wp5UFpeRf@y zAem0F&uZgyv(Vbh+v~o2)V8VAO+Aoc^>`L4oMp=RGcr6#>BbGI1Qh7dJ>!(!>a6qq z|17fZ`%xV+xxO$sdUJF23kOH1071#c2A>ZyM(KC|CO4Cb_hjip(hHdjrx9W4RZ^!n zZgbSoBz_MKwjF8MTlJ6|oCwL25dH(e*x^dLRA3Va#mT2V zxb(GsvJje5;&oChCDG+>VS9=?^JVS%dCEP*s~@D#0(5>*fV%z{%BwhJa|GpFZOZ$< zX55#cmrFTL@7BoCQjNRGic5#Lta7JoK0;O~?=uqz0e*kyv`%vM;TRpZMPl^gm(o>0=8LB@%et-wF6jOAqx~@Ag$vCYLBkpqTffWZp!I! z@}LvY5OUJjgzU@cOlqu9btdNkGOu;JXNNJ|q{>?4^*TLGL9u^9LOI7`$5}n5~4Vl@`KU`c4BFFPg{5k1a{SIuk1ysoJxLbvkJ`OmR4o zf_hSH#>oP|oKEyP&T-@AN~kVjYFJR^(AwsMC{ikqm0U3iA)j?5g0L-fA3w>ddhDf| zU$7%Oj}4%RmKZ&*2tQ{>wMDwZ+{;*zPg2T{Xcs9UoIExHLQOt=aCUVgu@i?rJ|>$| zIvJiF&4?>Dhx~W-1y2D*&{wQ&zNZ!>!V<6S+Cz> z&oi>tSAc&FOcxEGpf!7R-0TgvC_@f$zMJcj?^w|N z+N6|G^m_$q2JLSHG)QkEi*dP4=-4&GhibNreI_=Ic7-zDBY@t1B$y=`K~3#IBbP7i z8P=R2BoWJW5q1^J?Mn4@bwpVZMPz!v!R5iYVA9XDBb4pX{c6%N1ZH4%Pun|xWsk#>CtdHB0#nb`$ za56VMa0FJ?M6Swt-8ZqS3q*}bzCQlFGzED|OODNFZQ#~P`+n|p83BsKrs2Cax$f9N z043i31Vm&R{WHlx>2fuxW4CwOy3mURkF_eBKt4N79KX705>hjZD?=%OU}jmoL8 zdQ~a7-x_WB_P6fs;XR(dbMO-X@Zw341TrXu1%sN@9n$4fjWnwE`*v*#Ag%Q4BKF(( zEau>h9HR%PZL)NM@ulkFB>b|4!uw54R%v5mVFpvtmSN{x)V`p#jdw-`x*^i?nX+j$ zcXr7UwS0JLsHWo`fd3SuSYt~I#)~B_A?uGM?LtivD!*PY-oByk9c-h^0hoQ^=F0*G zTQr>-r+nE{fElm77!p}5yMCc|8t=Q-v%$G&pudQo^uXhc(? z@796B%}2{F-Fwixq{ZQei=)w!qZo_soINQiiT)P zNo5&C!)_E&{dtR3+M*tzErN%saxRD4*&26om)1$W07M7{?6#QdM|B40x>gu>tD4jG~#jq;r8`P!A~ zZWYfRhYKn(YaCRj=4@g4#BjO#yea2=brpa2Wu#Yo3vsikG{?;O{(NW{U>Ztf7}U#S zfU558mUI``D0p(9r11I+*946ZpPl0rQJI@kb;p8vw}KRj$p)SWMZLk|GPXI`_b#Qf zYG1=K<9Y?LDj9Z+Bi*(AJ6^wMWbm>OO2XT=37SAeVky2BzVhl>jJmq^mOhJjEYp4H>-grS@aEs2P<8 z-CdXQcbXtcXAG(xBEFI)U!H@5eRW}zdeO%8@>XsOsKLvdASZijJJ00+m zW3(rymUom6r_XSmFS@%R7g=(T52o^3_!y)kmsPD)!DNN)Ki^j8C8iP1oGfx=ILP?g zx(pSWL58PA%!ELBV>&b1)L~Gkc~?2cC>&n!HU|u&rg#)?51PIdQXc++8FH&&oMS@A ztT6lM$D(6`wcmHboSA4HzwW&*r6AE)rod~%p31o8gW&TBqtQg4Ck9sAp5t~!l#8f3PSB#dI&mfF!2?uA{ zEIL!Nk0AIr4beqMv9Hd1U7P0032Ng{^TfJgF&VXJvi#W+_J3Y~3YxP2ghjXwifoq- zatZa#fgzzK{%CLgW$E;qYOjkU}iAis- zbHl%EYfPY%_z}Zd+Nj$4(k*Z{5FD8{5^=ajf}HKd`Es^Y|6bRqzz!e!e@)H7d<)cG z#rR1`nCGekZ}aNvug8q@;?uWN_g z-Mlc2PhX)2ZG+9VLXM%Qt1<`&c#e;Rd6#R&RbG-8&V*3dDq-IJcJik`I~?d&?FAM; z!mU6pYzfmmfU`lr17^1->52MG8C`Q=l)XQ8b$5Q+C>#$`1Qeyz)SRBMvUBwn>xZGB z4~)C)dwRAx?`hgJ7Jz-D$}6!8^A{5pC>O?LV4~3(+fq4~AxENA6u}&$mZuq!g7i`< z??=6VFaOkxy(8#ru7liJK6ZGSg?(@&sj4)NAt(Hw0)ZIk4WzTYt#vBnRNw_Lf?=~I4Z zj)$XO!#Nsxnt7u&f^{0&-`UHBPY*NPK(HsEzeVnzU@9f*F8m@TjlaRx_}L($j>aXJ zXsc6^wX5)m_5M#Ls(EJ8V}Nz>hL#a?fFjOt&*z?_s!>flQ?|B23RZq9-t{-Pv{^B}6F7ERs z4)lK*_#HjnwKg;Pky0E98h2v_q$EHx3NpC8nT~U9s0Kxm3%($BA^(2GUI>^q8a=W@ z@Uwqtgn#_HN|Ss(9btVJD4U5rTy2hLa6Ws@P3;5&^t8s4`T|F29ys+Z0<*BHo$YW* z>_lfiB4A-iJ~fN(47JrJearIWxO?S@yDnWCwxFXj#*{{Dl&tB5ki6&1)_2z|)lu~Z z#M6=QD!q$#0)X{(D0EM;mI*%p2}rjj?Cf3l_`5Kq=dKmNqW1Z|0R?qBH+H8BA1uzD zSX}Ts0ka=(^dm;(1Rt>KUR*(K6}jPF`6A=`;G2`>wE#4I>5J?j<0-WNtfFZ55Nd#i5uD{v+jlH;boxTKhs~I-cIpa7 zoq_y7-D75*CO8If4H}dlAtqU+Xp8yR&l~P(EMR>cGUwT^-QUU1>MdARs}`Pmr&RCB z;?X7F`MmT1uUVR>ue(CO=ont=7lh0XxoauXH1PCVWw;$DAIHFi3r|xv%G~c?YGE8> zRA5osP_KmKgl|s9?S2;%$wG`D;pu+ovw_Sp)0wb}%GpdSJvp`pUsIX?W9ciy+UUBj zZ=gVNhvLPp6u06M+}$beP+Wt%yB2qM4X(wCyA_HRcb9K?-tQOT%49M*vSsak);bQ% zo>@Yjb;PqWyEXM05Vi1_FmC-tWnh%cj0VYdnA4pZ^chibsIk9gvTC~zSp9?hX^N_J z{u)RYbgTaCDEXPSsM}9Cj%c?XF*GwX8UTZnMC$siw1BG=b3hqWpx)AVO_Jw}Eawd| zM4N|VH>?jF1y^leA7Ah>2T)TlwmS$l{2cwII*wsjP1_3=jMi4WwPgwzK z8{^;N2&oi%FMnife2H_sOe_p}k59JQ&GQQ~)$^pyGcu%)dFsJiz?l{3%Hdj(JJah` zMTiU<at=Or~{B;B|3(5*px;AKN+(xYboLC!+i0f<= zRAI2jh}ThWLcjT`?pd%pi5lTNP%XC8zH&}u79Bpo<*a;WuNzCU%CjZECGbBP!#I;V zfRPP)E4V*= zfy#bi-8o#EEy6exO`dwtH7&utyEhCw$3Hc|s1;pZoX^3mx`W9qEEa`<0HTgcK-0F$ECh`>O{Rc42r(WGjubLt;x&!^CzOi6KJe2$Xas?s<;G3PMODczgjii`k3KPE zX?OE9l#Ziehtr^q15h8!w?7=U#o#&#qGIxI@paY@%Wvi=xiLxqlun4ju%D{?Jew;m z)%-uEMBEql0~l>B|CPb*OVk)YHSAqi6|11*d^(!e?sQ6vZe^TTty6M5=dG6e=~9Ak z&s8Kpx+i9@c2h^de1?a;ghkcf&>h}z+?DH|=wo}BhB)L8qXuZ=or+x}Ocw;_47X*k zEq{4q37y@rMsLX+kXrd55FimSDYFTd!DYP21O-nPrcqB#fj&pA99saaikyQdR>82_Ktk>sJ zv$$;C(0W5DBAA7cn}_vu`R3fOv{~`)`1g&!E%L2S%}Gbi&bP!CJ9P45)2+i!9`#CU z7#ww~)-nLCyXk7*&$D&t{~TY zIkudCn%`s+ReTo`amAX#)TVx&D9={!ze2`P=~(NJ*B$-U!R>~u%K4cjMw&hG%FaH=i{D z(^=ljVcPg(NR-fGV=l+|69SCMv1gAm!~ro%O<6(CGgakTQ2$X`QXB^ccdAqCxtrK& z9>acy*wg4kVNJd=Bv>Rz{WtCUDl*p}TRy9!s_fG6&%&Utu%BMYgW)@+zGl)XlwoDb zE(0+%s*>h%C?Z6WLX1Ma?W4PP8~Z2?D=>BSWx9Q2&qQ7YWLYJ5z$zd%-~l_e-9L4n z3b5CGU_@7%8(sV27_L*=Bq{0hQU7=1<+21>?6fAGbg&-`b|bpFBx8i51*HSb+2xc) zAprVot*^D@^kWvM9DOsC5avskPU2O-JX7p-Y)1je%q@Q$%_+t+eHQS?1Z_*fre2jC zWxkwWi`J!ZL+f^xsJhDjqRsFe4qBH!2$IWJVWLzBoJ9(c?71;>?O7u|Z?op3bPHSr z3%Hk#apAJhWf+sb{?swTv|yd~4bLnc6F>zhP$!s*EtWd5*4W)JA%bNa=KX6$abAkQ z)hJ!-mygKp(grBj#BQ)H9$?M9+MoUlq$stm*^TCaQczmbAuC#)Ajd0SFlfoNE2@v8 z)3jEevq;-@n#&U=pin&gd?9dd>GDa4>9Lp}No$AE3#H7|V3?%;JGAz^a1VAwJ?fq- zpzTrSVgRp_!UAi*MZiI@APmcEq{_C4zIqe-5|Z+bXyqwTpJo)AoaIeaziL9yk>=j| zm2a_{AK@@uEkIn6ZUqc5C8TnPYVRdzK=XYW`~uDhU1{3C(2y(- z&D7|wMWHwLotNYoAucN+?5r{&;3> z^GDdSv(*H*hVz|G7TFdOgmc!j zP0p@VA(tPyB313)W_}v7lrDI<%Gofh4N+~v?flCkVqKtEZ7c&ACV0xT&kST+2_Prm zNw?Zk_R+g}Wv7B)u%$4zskMYx?DQF}XpbKOc4M$&5w#U8R*i0UL$xE;3z10K-4U{} zk;pF-n8cMK3V^HS;Zixedyl;GZ|noYHfeODn{AMDeiSJM=_`>|dKdq)j3&|hQ^w4IsOZDj*fk&k)4$=8iZXbsSA*g=2zcz9$C5%V*)C_+ES25GedahU5kebeU$Ol}z5 z1+Bm5pV-=D1pXV=+s3F>sHp%%W1bNNF;->;Y?M+IKE}JVt5N4;s~FMP(rS#P_OhIT zi2P=PnMqC8E@;4e5%fIp`wc7rRgMYG%uieh*4U*Q_nQ!4`Bj@g0Ni9q7SEVTt%!4x zy|nBOQnK_)(i9iC*@q1~tACPg5HC=%xepl+t6dC6fWZ6K#+}yXXLry80lR}N-)l+d z6E1t*Q(XAL*z1!BCe6hh6lBNt$58^fo%8}7k?6MZNa4ZzF>lw{tY9~`2~jzAbLS{( z*NsNubTIM-gXK~?ae|!m33qdVQdN}F18I%b21sp*zh|;DKDqpWg3#8+dImkxvpe`Y z_+2PX{%%`%_A>klC(brC@LfK?9Y(HlB(s!6ghw~A6{fPEbp-NfLBb1;@qSf&U3f4t zc9m{HB?^l-nO_xP%|qprNZM!ih@U+M4b3Fsn5xB>rdQHIMt;r^-K477IwE?6fpjRu zX1OvZV;z(z5(K|z=Qu>q{LJ}IT#@Akn>fzFjPLJAESFE71aBV>0%NzZ98|m}*OB8f z$B=Uk%JEL&Cc_8-zb@83%6lS47F6RRX&y{NvA6qB*JbQR@Q*Qt5eB>R6OU8h9{re%jS}W0W{V4e3eao)xasS$+N6@CJ_16>xs}+M zps{45($XKnO_Z2tFcVm|P@_8hXy_WbJ%B#IO~Q|~qfiz9)kgXZz#%)8VwS*@%A+k> zDzXD)oaNl%#4ot|`QZ$#%Hi1)g>REjBiH{~O=9=Nyvfzn7f}&$wJ{n4Cx({-t46D2 zo>?>>c*ihL->B?0129P%dtV@!3vbeV&#o42Fk>rfuFDeMGrcHrr3HF=`VrG^f007g5FWs5gxUCVer@u>lHSYjI0bpDPCV5oQ6dAj}GuF4luv|xvOA?BM5O9O#5_)*IXPC2FtaFbJ zKpt#mbzmY{UmUSs(d>Le!__}kk5hz;sbQ(a%M7G5`+7$MBC9ogNUO33LHvM_#=7sc z!V-gG23kW$gf=!?_@WX+`3~H@$KtorM{74uEVF9@erMn>GPjHQ%r3|l$0J#C4APl6bV$x1MvWV=?nJ6`lSOcpvQU` z{gD?KMi=I5HRxY2Q)?mBMAl@17Ke-onL;H{g5uRDv0u(u^;=!Cm{YA19jLy+r9Be* z6n_=Ly7CTxxkO}lDD)ko2%o$R`+WRo*m25jsM@E|j&wUOY{4ZHA}a0(c69g?q`vU%)P^2KK44~ z$$V`;v(wqF437u*uEl&++cK*HryxdvA!wFZ^^6f7ao zJK#x{7@0Olf{g%K?lFPO2x|^Gq+@qNspH)|bduC7O@CU|Z$Si>txdrRZs!O`a2u^T zF(5`^%}rUsVP^%5Fqu^)kSdE1L{NEPKPTn!8}qAHbanqdGLl%XwcgyzQ^nt1h_&7O zUlz~=-=^0mhqfk7@}Ipg0q8v$-1_-IYg7@NW9lX-)8bETaDSvoctHr@MNqmVsvES9 zBNTtqLT}ox-N1AD7TXIAs^F4LcwBrA~TPK7~RMzieHM zLjr*ZZug&SI9t!kQr~!W0p5J7_z7bgf^$pW+V8m1sWyhq8GJfNvLo)fKoE)wCXV2{ z^hzGR;P)?vxXcbtmKbVPNxVvQsVpo0)Y~?BDE(s_&K?WTezDiF42N4b{Y<$`%h z{yvXQ3;Db?iD{EUSZ?~e8&7E98P|Up&7@XXQqL~61C?rp*2z*T0L7S3z`bkXf&Tp` z40V7>wQ-5#Gi371D~9HwJ6cR zNoKc75Z~=g8QJ`$Lb@ZL4jW!Hti)(q)A*(Eibtw8B;CE<9PN(SbiLRBz2KEgDxX(g za0@aLxmw=d`IxQ~tSgGu3#I`Q&W(%pCfC(RtJp!5)nsod30n7`?xnwi`9N3gotx_` zO}sEqZjrh_?gX#mS7R7HTTgL4(7leTMcm9H7rJ}@F_ysSlU{!-**GU}%XX@&IoL@>@PXHL)0C#`7Y<~YA|(-mFsKU*;SVBmZgz>lmCd3DJHR!zqnum zt4p`2lVP;O`u|8lxg*MzgT)C4ED(cl{Gn+)CK~T@qV#^n2A_4`WUcsxy!IvQm!cs# z{M)>D=O19`79=5{E8|I!u0(H9loPfkW87@VW{%$O==`9+vul6zQN^5hEH-09s3HkY z>LglON8rl96A3sGpnw1E(Yydjt+(%>Z~rxmNwE~K>%4Nq=Tqsp{QW4M=T>ORK`?2a zFQ->t49{z&dMR};&FHAo!jB(nX30lCt-(+H2n>L9Eiqh0ZqYhC?3k`-p8Cb4uFCOD ze9Z~ZqNle)pC{pSj2EU^X&=wxBk9J*e#TkYo}sGuYno#7fjjmU51{|Cm@dx+BhQ zon4TJhqdcBQS`L=6rQ@IrzZwvIj$Oi(~dGGO_WgVx_ z`y+B*fE*-Ru1`bpSA)rTqh0pE+S+x>ye=BNEN~P+ys?N#Vmzaf&OS(=x-+I@PPC+h z{HkJMwDn119^hS4${N^lG{s~jnw*}B49~Y*yBbhf9@wKw)R%SM0$p>PHw?mF)}tSDW|O=TR~gD0QWTP zduwdY3wNQ>gPA!(hLJ78wusGO4)kZtd&H2 z?UCY9`aOTGm~mt&VD_v57Tz1#XV=v)M62OgIkPYpO+$O?|5f ztu#0M>!?!-09%Xx$A=mBy=OpRj^zGH%83OzmDh`D`m~?^;6LGAvNkBY2CfT~_}bn} z4oBaYZ#GvbUd!$0ysl4$$=d2j9E-I=EoA{BW~YpG5k_Zf9W;gj@w|_RpF=B}9P%|W z@YNQ%EzJ1d<#`riS#S0v(kP8<)f#qk0H1Po?UN&Ku3B{b4#m03kVX0ei4YrCBZ617 z2^$w>Gg|4hNNUiLUpP&L*~cbCIqYiz`gF<~_iMw#008uuRol5dG1I%($XRA|Uo7ohjPHF7F z3~@7lfvBC*4yNQU`C8>3^2iM&*8HGT`Kw;&4x!DyH70g^v^%8$b%nRSYZ`oc!*)Sn zBVGYme7Wuh^48t=Bnq#n6aXLta$6p1xOCGU8K8iobALha$gvE^m$b&(9wFB@MNt*Ruah2b}Pa-SE@bU@U)ot~XDsZtBl zO0QK@G2iT&o9}v@NDgTPaY*w^{IE0LRF^5eRfC7d3a2J8R12Y$Bh7_ZUXvR zi)3123mx1jGCbX7w_;X=%jf?nqx*3By~11g0#Xb{Qr_?sq@c6!&WTTA`=l3^X^q|1 zm_JfjsQY=2reBc)#HqWZ%+%GLJ~wt|PkjS3p8qUiBUY%_OP}7IlrL5Ay}s4XES#22 zYx}cR)}>SzY&FeI%NcH4Ek3}q=}o!@P?w1!R+NH9#I{kbK)^x4>Qq>_Q{erdE+Byf z-d=^Vpya<`++H*%2M;kE9~bZ6&hWj8(iU88Q<&+Jy1dUKR|uG%VW6G~>zH(-ZE2#Y!r zILCT{Zs_yb4lV6`Nkrsp?F=o&0I@Q=7*d+O5Ejp-IH2}0_N7)ss+7+w+7<3o zc>M>NpF&l;s+WvE#w9F_FcFD3Il~vRlR{ysenan9Ki8G#k>jz3?y1$#IYJ%k6n)HG zBQM}J*pV3YxrcK-2?$k&p=v?35wzSSzQ&_Eg1=UFGj@e)*v4nD^s0SK7%g`_U@{dH zH_&C9+3I#)tb52ew;5IWuZ)^mDmdPbAX6-@HYzjCPLz%7OLHh|%H(gEPyfk&sd@^E z`^U4W9aX%W%yeNqi!=cwphrR@StGUq71rs5xL%~mU_!lTa!qnM3ARy62kq-LdHwrB z1v4?E;lrDwG(wF_WTk79vmWJ0bOR9 z>hY^sdWO;x1LGB=D3NMPiRSU4OM?FOK{DQm&*cpE9C&ShKUsJ6SyXpK_NF`bIBsu@ zh3UDXgTQI#*c|(q3hF*-RD4I2Gqje7eET-P-p~Z^0676Xmz-LCXANqUx=x{6q}c7J z9?YW!jQEvkl*siyhNkdu&h-;(VZd`aE4qtVE96DBhy_n!(rCT7N7E3AN9xA4Q>Xqo z>}Zt+U$^Br*xY`~ESVZ8HFy<#!Ojd=sY&bmdgrNTr9s)kyHtMgK8Dq_QRjol5zl%J zEfL<0VW?7e(;4}z7d?tt zq?*b`jHf;oKm_;U6Xpq0b&4$v2@bnIc08q%la&(NHq%2r7m!jJG6jM?w{HLWfzTt_h2kGow^>j&K!6D!No)P z5x^En9^SlT_VuiXsYh-hAOUzzv?g3T><7$z+V%KWe{E2F2ZFx6@KVv6pdC2fs!I4} zpt&Q+Sueb^FTZ>7#Yo4_Yumb;lrgIfs}$!f@j)eXDiVuZV|dmzwYkl%Z~cmm~TZpCfMKe zBQ)Pu4K3IL(CH!C&8UyXcA+iy{D(P#7dXfaD)>`ySMW6U+--F2rA#M*Cw!(>De-YH zP}Sse(4d{6XLp(it@V1x)<{$pRgMi|IR*sEa*LOaZ0cMs5A+cJ`rXd)-H-;agkpU^ zdy_OcoKfpjCggjL3bH2pwsZ$4K#|vMtQTp;FHwx7*kw)QfWqxo)$cYcLo1?MaHT(+ zk^a@u5XrfsTF0E0GYrbUTAA^0V(U7DfOECCk=X zjYO!vFUI4T(kfHrEJm$`HLO0IM{ci}p;Y>I^E(>IYJPnncseev!YRg5XRs+kx|$2Z zDiXr<`dH&4=E&5#PTJ&Raseo>a>h*|5WXv~atu~&PLLazx7!IHO?<3C@l6()gOKjE z1pg*cVB;dc#h3!Z?%%w5n=C#winw-w2*UpO>@GqVI;e?b&dy}ghg!cKZ+SmL*)bfT z1ITA?kvKrh3)fM`^UXawRkPR*e#Gj4FeDzNBz1T(lPQE7_q170co(m^#`x6QH>V~a zbSyUFVnMc<5n;L|!bD32Nt*r>nG53|FrfcXX>e1efe4jJyHXcqlUJtup%gxuoG9qL zLvr=vDn~2pW=)*&=y}rBt+!3}Q1!ghoW@F?`01Fpyy2@-GgbIEi5cOkNFMyz-m-&8 z!5}pApqO>gvNq)*y1q`632~R;85wx#7z%zMpPk_XQ=-yJ{qi&Sp5a^r32n~&nMZ8U zH{YPmAtuo=VQsq}_BDcBd<*^Dy}wJhYR3zO`h*sawLs zs~M8qEf;s7Yl~XoK|7W-fb*SEB|)Qi60#71WBV7F;RS_IpqB;3IReks*uNDm*V`AA zZ>28Km42^9@z3qYq^@WsYc0JPTVrgZt5W%AB|O=rrky{~qmgZ=xq7j#_LGp3zpW+& zqw&E5{4rlf^`7Lnf0>YUQ(0_Jzz%Ac$r2dhYzSl%>Pc`{0Mrit0C@i+SglS5W-D#-mA3mce-O1D`efd> zkd@ef{qp)(8@DH|%$Jy)#DXiUOBBepjtxaTj{|)$&vW7cXZ$FW3y#C}@V(hkUU=3Z zbi77ha~v|jvfh*cX^HH376?P%wJm|Wn?g{LcUh0555`m{q5Psm*NBCiUpU&d)a8t6 zR`|>DbNzr6T=^wlZA4#AuZ{`X>fE}Ivp5IlZE(tOzE0gh>b(S2G={_w%?UBDWXgZnKG7{?}^WY+6AV8$z`|AZN2=4@eC*ZnqmoD?;X#?n%MqE^YDBetyWS0 zrXh}4!LBesMN1Xeid;E@(Oyx+qum&r)9ZTsL}?=(aRw(76-+*{t*>JTJF%!nvAbUP z6aZa5QO zPM84%%T_(Vod*U7-$BxeQYqbEtK9* zH*2NMVySitHjUD;w*>&Xt&0z4qVYRAzV3%+!;iHIm2!}-3S)~^hmxDBd&DrH5I{wK z5G1EnUyHQeffSPRl=UXD26i6d;Bq@LCCsWwn~gyvsnE69VK{=G-?swWPzw|Tb+ zf=f*~TQE9w{lD(&c>I9a9{@Z&v;MG>)repr%^0iV!GVv$g~tFMn|wPCW-UN`kL>o| z{Gn%o@j{^oGn#G&P{aX#?Jdsd3mUaTi@zF$ZS*s7njWhPO5ta3%L0=FFeVX=l9XU( z(Rz^``I?D$A=IG1p8w1r#hFCg@F^UiTRZS)zc`Mx0LS{PP&RJS+3_nrFy<<=k3{}q zC)sP2tpV})B;8{-p<+jteH!EMiX4n*Pib^xeBCOdL$1&du+ z7XA&z_gS`a2fQn!#m@KY+^(6MaY3xswb3RZE>pXdVc=a-OZp^LnpM_!8+DI;oD6ZJ zx`sBZW^tN5+RrV@3Zel@8MB=CooCVq-eaLW0514c9hA-KepwxaA>@^b*Sye@c)IAa`L_!o*J0fDU$~}te)C7j!@Zw9v-~Vj!}Jt%#?FL5UVcjUE^{3k6kT>}jwb>ZZ8#EmODK|ic1v7Mt*J7FxH%a1@IwsvVK^#M*fIu>L{fzWa= zykJ5h9ZhwLuEvqlfwu8VR1;^+etGbDUuH^0+!XT$c@Fi1;bLx?{-!w1>cCgIQ)1&T zch^xg$i)@EiNyGJY%gYDi6qbIJe{B;O(S@eNvs#Z)~w{~`fTrlQ0bM83X-Jf>`=H) zM9Th~zFatG%RpfDsRApxg}(~=rN{^C#i7>+6nCWAoCNWZf_Nkr4U-ZMhO--n@}9 zhxdAp^ad!KbM~5}PBPbe2*0wvjeMe0XL0XGKJMf2+&Vfia7%ID_0jfXodooXn%HZC z?S{q9o?*KffFu(9!NeTzOlrw8&>Co@B)I#1`(xh=^J`OY@*O2@GOJnI7)cuuFuj!O zmMe?e18(usKyw>?M|$wITC{}P^+B98H7@|!vQ%V>6HWoqLKZZAcf*%=$+{?Ms?ivc zZK!PRTPin2jq!fFY=`#hsg-Z+b@b)G76IqFBNr6w80@+h33KSe^-EI-@siB&(uqg$nGK-(4X6wZw#G3E)#%7H)V`> z|Cf62Jck7|df2L1;Ce6_k*e+`JoSbP_|X#dsFVT)QQ{v2S>n2W1Z%XHR0A%h`GAsk zjfQaAsODmD!H=O{NdHGhK#2nz@{H5pO~1o<6^Ii zh}pa>V-z_dG^y3;96NdKx>Gc7s#M3|dn%uB+{6g>nJLJ7dlCFjWm) z-iev8b8j&XXM|vPXSC9yZH~#F0HbSk$BD7;wCQf5!PbeFZ&ArRiu3?>bH>a2tKs!T zWR(+?mhuDPy$zu0`vp~5mK5nRy48MbQ0ap$XynXr^OjKle$b6;=-}A46ub@Xitz5B zo3NV;ut{F%a~?6?b4QNu-CvXOkFcd+4?OkGI;J!EoM!^k?!Swk+5wnWjN{_7;#-d& zt-E*LZLF`SpY){A3HfegkT%kj5)w)uqloQpKovmA>lX0KY3fov3wYz=1>(ZTncF-K zls?0~dTLrTcS~qf=oU!W-Krk`DO2pwn>J2rq}oevL^G+_VcoE>*ZnHp>TeET5?GPn zFxNC;0Loryl>K%>Cvrs)%ti#7n8_TYP0luq)5YJdv=enfUxcwA^$U&JE<(@7R+o*1=Ef6ES|CicRB`2e5^atfx1jkva78i_%lo@L}=ENbeOTdj%$-mS2$<6I~EC^I+*LsiWTc~w0{c}yz(4?a!a?d@7S&;+#dzQ@`r z`C=(ca{py=F&U@0H~%p{X*y)p+g(KG6*!UT-BVhs!XF2DI2vq2A6^Rk zK7p`zK&8=dKjLdXSBv;l*+Bzp)Ko(_jnnt;2gWCAnhZsqAh$F+OgH}uI)>bZi4k)9 z=;8I}HKZcWZ>NuS&9k4Ds_yub2z=UhmlnQ`AS8I1KNKq+87aN08);^G?4?znKs#7C zC5xYmvfZUDapXEx-%OlIxqnD=U&)78W4bo=>TCj3E{+-Ce9Hk7RS#zECrNMIrw%UT zcnDg~nG&6+0%Rhsvrf_0?c(#4CHcsUg{S=Z_m>#gul}|pn|_nlg<1GUqlFEe(De6M ziiI&#x{I>ssDo(_17gs#6QK9+-LCCbc>kGW?|e(vr;$Fz#8axMFg*2YbOofdYL{q) znkVscu)fw!zFAyq_=^3j9$>uWt%9JH?m)4#sJ^pPPAlrYS{YpFyyGh|zR?;*A<}ln zt@G;G!YxYtko=Eo=jKgYaUpd!v9n*wKZmj_ZCIv9;puX424?ZIK{04RW-*)T36v=Zc*nvzrG4+lH<7u z+tSElKivLP5`W(w41|YsEXDTCq&f7y!PK5yY9XDQe~yU1?I`Q^|3Qh=T+*_{XU5wI zeZ9H7Ab9gRkc>wn5wbNXyPVuiQ$P5u{q{G#m^9ptURrI)zBz}3fbp`$IF#8X1=F;W zyGAn>lax*4uwLe8g3mqUv0}Q6laKdtIxNsgv`R8y0NZe8*#A{{Le_iI{&E@_$-GnK zdfWVrx93(AYxb^SvO$>juOg()L7|U5cm360Ek_1YI~({ZTIeL2P&2rZp*`LpElBQc zp^Km7``d^b-4ch=o17#W8~bHt&J?Keb<7^O?pfjd>mk=e+Y^9?^f&j0J6fG=0Hl2} z3T$!t`-CrI7o@h0a0EbY1#dU=>qn-+5lc8_zQI4~=g`E7*B$&*8qIn;eU8pbv&t5Y z6oEf>nxt%05WU6`)me^*bgzSisuw1@thTkblnQs zd|DJ$_$k$|rLMW*6tBjr6jdrt@gp~*l%u9Z^P(2yc3VV`t2V`t02pLlu%vwRow`%lz8FPP%I zFS5eEX%&f)C@An;1PPQIOsa##)cE5=yAs@ok0MAg(8wRB_$YwOp#m^I1Akx;nf;+A zKlk`SU16j@ys7P0jUTB)1o!;YR(_7ZJ>(?PD43ao*UH=57MLOW>CwvGl)pMZZRmP) z7V^e9FS5|bJU62YRX5GqU^f`2>&$xQZCguOYul@S*;xaBmIbYh$$n9d1-Wc0DP6Rg ze3c$g8LTr)GBQcoDLiWljoC~Y|M0=IIiJ$WS+0CP0iQ`RhCzDU^xTaWSBmRS)tfy1 zh{vp9yS1$Z_{jCM(9f>qDJ$nl!~+wK(hV?H`M+No2t>Nc_IIlDF!Wc@Rq44)ZxB(@ zY>C|jA!}+gHOkME*79fUF;d%~sxxeglw>gnvo_U04wrtTo+zjMxgufaa$ZbWXC^o` z+pd#l{JEGe+dJ4lYZacHAO}#KSIjP1vWx$MtWwmsy6d>nT=L(1=sA0C z8B_URPuN3?GQJ0SFEskLXCS*(w3^XkM(it<(e%qE>rTi|S~p9ZC>3m05ljS(Lt6_J zN&YpLdxE4(3QUMyu2Uw!%H<9xEO>k7ZpRn;=B7*Mi1!3e*y*XXhfQVZU>AHxW9{9r zYWJkpu&1iBREpAN6zHgOKdAI_Jtj-WQ*OswAXN;9Go$$jHF2sj5kpdD;bXq^##+MB zFpABa%fT3-RO=;WeLOf;?U$;oxcdd-;br}%y4~4Sdc}oXl1gT z7zmF*F5$qsW6z(YTqh;9^!-i~6QdWY{D8zxPlTllGc!3qwlW5Uq!-+5+rmZz|= z*2g=uDU&C{3+w|)hHME_pX5#!t#4s_uV)BJ7G2Pc(l<&sUJveC{|SuJLBabP{CG4p z)x!OrS`FcrIx^Qi@0~Y3SvcNyP~{dPrqPK@|50_ zI;2adp)+k?PRpUOv>xGGR^4c!tZI|Hkc`WrQH_DscFH8La@ThfunqHlK?@>l++&O2 z4S1INn!MmSSn%mVVV_vkaTg61VnOhqtqjg2No`gXs*MFjYt%``dclLrw3p&-u~t#|F(9VGPR`WH*2x@ zo4)d^Qb+MrEy+YU;IboMR*5o&tr(nqX*`(RR{1kNv?O7HQo#lto~PC&S zs-9Lq08=slsjyx>a15<%3Urn@8j|*veDr$SFzM)(Iuh|^z3h<}gkjxMs#wiWRWjAsh>GXE%X+bV5I9!(Q@ikz_+s?`i!5U$M~Y=Ftl?^YvdZ~GIQ zN^Pufb`qh^xn7$n5~go7ug0espg11OZKn^l%Q>j(b=6M$tZIL+F8`iZRfM(aU02ZzR_Zf3EGo-*78A&WnFOEz_pNOKhfT-%Eb!kpP^UYssVv$@erA>=jhG>wR<@*% z&Dqc!r$o71ajvsp?Key?Hf8pl&H$1(=jO<&*+`9M0o%!fDi^q<%FDPG6CVawM=S0m zf=$dZB@-p>-9RTbQo#NKW!&CHmwi!p@D`6R?i8eVAsQL(RJwUvo|?26{f|GDKAZKs zJ=?e1T|rXaA=e3v{0FQDhiJ z3oWYTZfG#DI{|wd<~3jS+^u zj!B;yhOw!O{NZo0+RF4yOtNdoaXPIR^;mXUy?Ohb&X;rb!M8fYI@vEMNA}IMR!rY>u6^ zgg#+hz^mlH<)~gS4nFx0g|%6?{;L-Pi#A4PQBdKoWEz3H&DE}W8^h|UW?DdqY`j0F z7$dI0I3}~`l-j+W}mnQoS#(ptqR#iTj{pZ9)$|42BumWED%+!E!&W6+r(;mImC|ODclc8i z-0U=*Mm+6C|QtTz_%m2aM;F2cOE7}0{*x9mwEU9G_ z`Y>M>E|dCHa?TLys_u&qz;w;-*ZL)p)c&*pMWBOb8=}^F6*q^0)`&CRPp^B@gtA|& zK4^x63b7AF{wK8KiR0?j#1>`JH(GpV-j`+)Qb9GJu|pSvbf}V!%;}hp$~^9(^(^Lu zcb0y5{BgjX^+!L>*^!(;2g*hN4>eveFzYJLA!DZO(Ov2D@h@=`9wV~qXocTW#uXKB zUZek<5GO_APRcFS(Vc_+q6?|Z_7|&SD<9`FG*tmp>_8>ak`FIR7xA3nxZJw#^u7^t!P}Ti8lQrz)B{` zo=c;1q#k}fn|pI{CyBrG7WB{hxZ4yh1BRt*zO6fjpB5E#KY(?3gZeD_7+bW6m`eF@ z7StFex%c>WzS*Su?7(%vp@c81gowY?732VpUAO^mHM9>jv(B_GuM_!9~T^d-w8!ew}L)5@4mCbV@V9 zd>XcVv+|+uXV`kR+|*>`zI+Dno@lG4Q0HfDZp&&K(D%GfCS00{twSZsq)75z;3C_w zzi_j}8|xPub{R4s=CIODX85kr`Xt9?Tou@&U|AsgnL@2hMlIG@*OA+8>=4x8an;o9 zQqpkT8QrYngL;^g1kJ)_(-Oz|@mL+$nC0+ zIYELSs;~d-k%s&b;I!jo;P??ccj|;4!+hE`V-T$*NNJ@ehL1=tG4GG82P#g)`!BSW-LW8k~VTCpey;dAlGG`(RNc1@!?;pb9gCn zAgUxG=9VwAJeDxcXLA8JE*in3gmhj0cq;|gzk|=lgJ04M0hL<}?{8p%zWV?)sm-56 z(I{>(`R2d1l(twzg*lWxq)0G`QVz>ahLFnD-4M&wloGCUHL>#A&in+EyW&ayxRk9Y zIV95U;9jRU8jfZv3x#(*~H6jDLuSM}$PdT8*`Si^C&5B3_^`xaU9E(Ghmz zB(sJ`{(M}om55VkH7t$4o4jsHyu11p0k8B*BHP3 zLP-%qj5*h**b*e(&6utK!H>of80MwYQ2V|R4(<_eXT?3q0jff-MGGhU6#wON)QFq? z*Jjp0FZbs#6^DGZ`kDNHvtNj9t}kX5#niZe_X{P-GxU`~`c7Yrbt$?OvQ0G_hgqHw zRy1dLUySU;Yl9zm;dQ(PjmVICmaG+BX_z02$4R#M=RPyHtTGH~$y08P8DqSuqAwg8 zYMU6$g|&PdGaJ~^*ly#XBv7<)5G^?(_;p;Kixv=I0E=5DqdDtsioroC ziym24pW;I{@l<)EKeV3l9KFP#RP|ocb;{dCaU}}T+*pw(W+tmQ9brqf|4Yo;mrBn= z^a1nFuH6|3Y8n2dKMc_e2q|WlWbM4vyQ?WSnFP-*d{duK~8A>s8g<> z7k9@FQ2qjJ=xBvr%s_|OKg#Tw>eZ{{tvWt<_F69CFG-b-vr-NGH)K(kL#;a^ma0VEw|Dyp?n{c z!S!UA1*1O)M1pF#oI&#u>P2GGN&Jxiitz{X)A^Estcs&ybZYs5+(p-s_mrr0loL-* zLwQUPZbb+xgW)+MJwFTE|7`}{y$58MFywZ93|P1O7dW+^F6#Eu=*~wfwHC?i9CYsd zW{~O>Q+=1){fuv>nSB17w8p{?41_71U}!V4#9+u+$@Varht^fAMjI72Q$rt-)PMY z_z5$mh+}p9GENYaG)EL=-KStjqYAlpuz4ooll}GoyhJw%U7=IoNm48<1$oX#JmWmQ zPB)}JdhjiC9o0ta-TZCyp%4%rf4>&fJ&mYk_1qY5SubOP_*wY)#mC`YqXd?+H!gV6 zpV=?J63STBNv=S+!TPTj;bXm3ItB5BU(%r&W}Ofjn9mAb&eb^KugN@KsQVKh<@kXV z<;4uy{|BABs2 zk@>%(KegWeM9rT#anf@+y}A5HN!5GC;n@j5{gMZ?@vflpL1#H|%x}{SFsJ@=o`t^p zNI@HAn!!%|S=p`Zdi($ieGTZr0>~2I$nobRqHw&-2yboTHQ_du8VB1-*0YG89LUWz zg^-$xq1#C;_hz7OR7Qe4X!q@N6Z26s^#am=ZM-Plcxnj020xf8q1ox(Fl?l{aKv9b zgs5$NN&#zmbnB@lFvd3{1+3UiJmxMpE<6f}MJFuAXi%ZJcZ#q0V0a~aONx~|^Qwx{lhrhmhJIVs=p97^Aha*1;eY{%g zGLhfJ^vD=N6qe+Fzc6f^3#Z0S&=<821 zm-2S;NQb|n{t&uKj}|#{3Xy$6Q4g0Y@9$Mh;;%CL;*ij^-rWE-V`7=#!@^J%kBC8u zQ6}zx5*xss96A2|DT;s~#YKpO5xE?Q!T*u#SR=%Qj6uYdaY$t2u<-Qw`sA|F^e_=AE@z8w~*cG1`Xn}A8nvzyh3!y4rhWkPl1Al#mplpO9Np}J7Q zxJO_eNDYhCTm4MI*1gD!Zu!Ey!18+`zxs|+<3ryx`Xb7)Meqs7gTJ9kYBIfjsWIlc zm625z{xPc3dg?gQS|pO^stH#*@gJmis>sQ=ulUQ!dIuk3V@Ew{^sntHbAj;@I`Go$ z!nAx|)kBBh~4h>M-xU#zyA|lhs0~Utm|pGZn12=)k-i~c{x6{6~(T6l}+~HYQ%6$&9UO- zs^kV<+~*Rzr*g$|Kik2$P?GcSVmHE_zd*nf`gL-a;094>#kb&D1{MQWsQXIimu?ol z{bC|y^onSL;~&k5 zyD>nW@$!Z0D#>fLtg5seEI-G8W8AbyG1k}n^=VL#_OH(xyD#z~{Pl+G7p2&A^pQZg zb5n`b|K;NO-z4L7+fWSUL&0#>sBgc=Z(T`he(830^V(5Y4dt#Rf z+|~&h93z5#L_S)tpf7lzO!8JKhw!6r^yY$f^*WO0WF5Hwz!A{Ioxj?C&hpG9akf5C zd@RJgi*lcN6pLEa&wwVrMx>N1jK@?L~kISIGp z@f}0^>p7EC%{*~yeQw`u+me^`I=cEfYJG~R#oW!n^uPSV8-rM|L+29VTL(i)4ANaj zFY+7-L-hSZ9fWQ&T=eSKW%Vi3o=Zg^()4ra#oO;ukbOd?YP6@U8<^7Rf1EzKbTF09 z1{lM!$1!ugTnc^q4DD1IX{Hoyb_H#U;(U;m)>%3(4FIY97W0223^o$SnJNj@iRHS2 zuhp@0ZpSI2E7A`(Qx#~-UV$;S<$npz`akiBuD=^s!(ldE`cjkF7Fj!kkws2+?6T)w?M(TW2W^)C{vQ5= z3Ta3At@ZLb14uRU33WJ@6m4QV`dMi!UE-wNK@JJ6=}tOi$Tsd&V>8J!%NK*lTV-D zUKp$#aPju%ErFQmIli{J@_+J9T=g6qy1ryv=YR4}nQghcqGy5XPIs$quz0m1aohR* zD8MD@%fU|dVvM@MFBv;M?5q851Vu|xCE$|pYO42;{M_m9QC_|GIPF5;xUI9#e9z9u z{>1Nkx*zRT{{*Qt)wFb@5XTAW&gHk1(i;D$_cEmXS=kFG`ka&hwns9o4>ocF(#I}~ z$~x}W+=zNK(6d8{Ymq~cb@Grx4GRgmV0+r_gRSF}-61T_h(ppSFGOW)(w{p>xXl@` zxo3<0LnWba=ZR|s+Kd?W98!`sl~w28Ao--4(ikez*}v@D&866O4bT+VUJG^ZA6nyQ#%nSj{YD|EnpcMJABS zHW`1tK)u-lm!4a&IcsEiLY{S8HDk}+U?Ir9)hXU3r(eeLke z+m~g_h0nK?M#8CE-GuygR}R(NXELJ>b1RLyC{5;ARJM4+CG`fHHEa#<#I!IZ__`fO zHQ|7Tp|Fo z&3r<3uqp5{)Um3)6VL zfK6GyYgR>_`asP=xnB6CZ~_5sce~1B5C^+pXr`^zR!)Qx{GE$&!Egj+?VU^u{UT*lrQQspjUI$bH2!t0&;hJdt2ng0*_bc#C@O0Xc<3fAmuH zXkxOokhVKMCgPJx-%C8)G`W~~vpa_~lrTgKouLxl+o2*j^W85IP1Z0Pv5=gUiXO@l zEhKDv(cqu-96DA)?^KpLpzcZyQhr7U50U&CHaz#9-w*AN~&B{1Vu zsjpRZ{=~|VBNpLUPAVyCg}^#@U54x$r-$rbOCz?0ybBMcZVbNtz*h`a*5QzdFBYod zQ{$C(78Kdb^OY+QH)+ke&@@5qOf=}g-#_hms(F?T zIj2(~Ui;fAgl_0_tXJgz%9aTg*5#0jHxaDitKqHfT8L}{T*3lv^HW*g=J)beBdt79 zpD@=r;F6v%U`JN@I52pgh&du4#=J2z=)#?wa9P9RD;z=K@W( zLtR95Fvg0ba7LYXsWp^O5OtmlQ6;~3k^88z$VHx!5RZvb*OrA4aJ%t3frH`jLvkbC zhJ3Z%_EFzUG{#@AmQG93>8582!u_4YP|xWZdtc0Qqe|!y?C(ia=&lOq8fPiReoQvq z1|n@EtypV3*U=5={54sF=j(6>@5!uW54%|LmjM{X3+@3eleLaWSCi&Xab)$g{aPj+ zVteSGzJAXT>&d!>P&ptByyMHKR@_=VbYA)6)M4bvQd-i_eLCi z`0XkjZt#)oLZRaV_KQ>>WGb?X$E>_l?(^b8b%1p5zd?@iC0`7TAJ%~%RIJd4)}3FA zuIfn({f&02JBVg4)7Hj?asGXfY~v=NuSwd@}>5A+q<)Ew1E5EmEH9QcQ^ z4v<6U(AJ1ye%Dz) z#f3u{CYC!f)Ij3L7e8OQ8TSNQ$SUwg1rl4in8@a^`yaxIJtw)A8+|1T>WpXMw}9$Y z(iH8G(+r2~Wz_#Z)PvM{{izYQUtgtB0>enSoNEF$z&H9~^C>o&|Dt(G#tYg zYsPJ7H8F|J1h&*0&d=a2gsGTh)!h=h_bs`ofP}r4|z;*kdyH)-(tV3FA(JG zmH5ST2y0*<>K}UDmYd=WPYSM7H^Mo^dQMwRsu8*$S!cUE(vsL@+sY2&Z!78rj&gUH z#E4?P*bdAoz^l7aF}5Dxd-~0i8-Ob2S{4a2D;DfLnEzuCEKulC)@i^+j-_m*j>TBX z%Q;}0c4~X)@m9E({lM?F*R>iE62Ly8@HzEkMX&ESWP}z;T%8EgLmsUN*wS0^Z|3jR z@?*ja$#|Qw*Dv&yv!kJtcC#B{o}ld9#EW12gf*%yuVY2!RwNGKB)^;F;ldx*bN3hT z#X2v3$gs~nMuO_S8b{M{2h7DBV(7autduHb9KBKAfpCmJapmvA9x%aDD!(DtBQ?+b zn%kjgN{!V!;dU+UvG?Jo8Ot)&rUN?S2j6&iNV_3&;U}pYG#<$8Q!V5z0JaVlmH$XZ z(yao240|ju6K@rdOSHVDRi=ZyrpBP%OYH<2pi=y2gP+F1YdXYaNBCWu2PWG1h?V=w zWeCUK%NnInRn#~_3#R11oA4k&N+v=h=z2nnE^dgrVVm!wT5B`#Oq$vToXTk^#WKo& zj;v4;PdJWsi+{FYUqE%uU}e$}0Y#tZCVMWNcFm_^hktlhIdIw=ZwAyb9G5<&qKg;Y zxkc%+9Gh}V&+p`{?6Ah4!UWKANZu?0)_8_URXtgz(-$M7S{^z~)Ezm3^8{&E z?1Y)ms_N3dH@6^98@9GtuZlcm%{DEOCP}^Q3Ruqg-DKS&?a@VgGJLw3@}q4qaE!BT zX7UvFap+)z$d*;(9Visw4){gWh00-sD>`gD10lK?4Tmrz9X(ik9b9P!jLPr8UuI?* z49H}K`YOF~Dr6SINRSA+p%4L;5)dn(TQvkxeRW?fVa4`cX}2da-2il(!wj#>lUbSS z!2P1~9Z20@gUq#lL~*cRv<=03(+yPOYT2Y^nQKAKoKiX_Wsb;8P-DwvB2MUh$v8er za$6i+^D-ddeV(ELK$3e>;dCInXp$zIdbNODXo`{HfespV|_^i>|tu%6Y+qT76u zT%NGY?C4!p>$Pbe=XG+1!Q>uOoa>20o+Iua`#qK_?kMIZ*-HTJTpn?+i*7`UBC3R< z#T(Yt|5D^c|I+oS5B+(8W2I-R&mAR1~3wl07B7NAA^wny4 zU+n9of|LxW&?J^#MoTJ_kb{+Q6`kzMnTSb&?a9S{RsgTmLFCQo-GJ6yNA#etnPx%` z^UqR4u3Z5hB{saju4*eEmi!FOvyz{?Pr6~n;t2KL#m+1VPt@~ zItK9g3^cwg|98PjP^SN=WKOGod!Z8$cR_{T%yqEgt2oSE7=C9RfU9Tmo3G9#9R+dC zVSD400JM`?Si55S^QgT{HJa*iO?iUF@j!lFa+Sq#rgB*?_V9KbUGl0{O?P2?LV`SP zV6C-+>rZiwozS0xtW+B3gx>bZMIT%#ahCj~0@lo3^FWT_>=JD`adb6f8vu1FzG4?K zDD}sJHaJ#;2-9V;;lY1QN%QhZjDhptAO7*$PBrAdYmmvMb*XW)oV}z`V(9QU)OA9Cb9sO+O3KHa1 zGl}S;%w@+1gt#F@veMF`tVQLq&;FLyrm}U5;|SoyO0EiVZfZNFOHIeyown6IG4^Ag zP`rg5Gcw<929GkXJR?ULdCJIIV3@#8jFgsb4ilVGt4~9!Sz47~fanMaq(aQ* z5B1{Y@zcIfvw+n1Xs}A4KwFag|&ft6TZn_irOmgp7RIA%$=16Q z@oJ*~^nnW6x|ei)4JyV2ucV zY4&T-yxCVUIL!q<=UqM*@RG^Qnuxs>Ik~I&G%%HRm79FHj$;aGd1|QJmlFn^Sc&1@ zvG`ZT?^qNdzhcQ{pBIxp$*?U>vsTXa^U60dE~MDc z7gc=SWD(Xo$lli`n?>Ag)FP_G_d9oT_Gl4ER+8M%I&H~9o=xX&m%y_4yDoQ>WdCD= z=0<4X!EoWR_p`BpdOZlN{Qj`2SbxzD&w-P3MAjdnmb1`odCX>|Ra->T+Xt(Jez0c8 z7kq?+)S(?k#({~r-6s6!&(cWgCbrQB9?YBX@4Q%S0QNL@jP(@=<) zLf+v^8OY`?vpZg0OKraD>EPU^XmuC$sCpD;Yh`^8gOm0M*LU>wqn%3w`^no;reLk2 zxC)#~H0GY?_Rgy7+zVpapqi2bjuYO9hQyHm>E$G*{%fqQ$4-EKmX48hL ze&=|$J5KD4^XWHV>2D$2z0g8n8Sf6J<3|+^R$U}LeU1ghrM_m6em%oRGC;|uFfQ;M z67G?F@uDRuhy=CU4>S1+(aw6pBos{-yfM|wu^30a~1BW(@5c23e$n} z3=(W**(JcgqEw!~(%EMse4o#qWF@>)gAU12Y3NYjyWI0jdFXzC5>(fCCuv#=U{Y`< z_kq}WlSR2zWLR`NzS4{nD6A;(PUyP23%7*)Qj+*v9`Xw-EUM}zM)d}Lj0L5VL9*IV zM_0OS`gJ%-;hc~*8CsT4+*Z^ob8mb7HP~KqmFRNnp-0I9Zn4O^!KeGXL)e|`vz4-t zc7flpzNJ`aw#K`HBIs*lYMVpTXi&vKd%CSh-p|`d3*9XRRN*_#t!E!b(lY^UA96vJ zmaAH(fV8%Fq==OhLdWEll{I;lH9cDt@Kcq!5w?XD>fQ~%0YytmHmN%K$YR4V27Ya^ zRLxO1&%zgKeIe(*Bl}BZ;vYD@i*lb{S`|g@ zy*^E%BB(l7(2w-d5b>RaP#ek0`=WUx$ZmQ(yHSF?CZ|b`&P{kL51X zzkebJfUwNoF-HfQt|xyK@Ist%=01)F4K5ur4s|5#s6fXhJW+UZs>NrGQ9Ao~h>IAT z2Q{m$Nx{V0D(NpG<|%wz6-`piNj}$tb2mxsNEv;($Pw3sp5rhL3^@C(0amW;H-ILG zJ3uVf+Q`7A5Jp>+Im%JhrFwGCQW5ovY0Q#4&xA-<)S$cr);cH`a33KGVmDSx7|iSF z8{^V7P93lGFU9S`4)xA!Z=q*t^zAD~Rlr6J5bF?qLFY zxE>P6plpVp6`yVf2NfUp>l(@&C~>|l7?!oeW=@{(BEQ`pW#rgx8bntb3O$nsjz>Zy zbwwo*X{ugLApmED;M%X^H>Ucj%TUz~X>cnn+`XSq!bQR_E7n2?`Ew4Vk)<(OUT7j9|-_p`j7nY*ZQnf$9S`BmS%lW`dT57-+4uyR<01_Gw<+H^v$fBeH zGU)P#McteWqxPxiy)Ne|q1KDeI&IDCm+H;#JAMC$)tA%G2(D>TC?$rKw?w4pbR8Gr zS2tbN#aQ(j&j;xibw1`!3&!6_ymXp%L2ek-th%vnm-*e#PYJ;pea@&`Z%8VSpR;Gw zYA`zk!=@ZVSQ!LOhcKJN7~j;Ub_yw$cYXfVLgV^DO>-3Chc9 znZVp~HPHRbCRhr(AP7o^sKZU`u_v~D^~n#|oovS!8^H0aC~L44D4Q=idR^u2PAFtI zE}w_Rd~ugy#cc=a#Ab5xt~w32 zTD%V%$4_$$k+c^%u3g3~fRN}Xfg&X>buwM*#scgfTa5fx;;gqX(giGMDWbN##hD^K z>-Ryr?~>HnDb$k&&R9;&$-EQ>^q4mC4BuN^l9fWZB&YH=aDoJ=E67WidkLX-u$%wj zw*=A}ATkL2^1&7$7AM*VJ8=;I5^&F$Y)3}*>dg;KEhkL$5u&YDr8F7vC4X>FcoA>e z7ACwnAoew7Mw~Z8v>i9Q8+Yf~Xg*YOEpm7LBVRLVdu;6yx?~&5*jetLG61UakmSNMBZJ;?XlFCKK^ObKFQ>*i#g!SeUGHzYs;3tN;U?T7?< zZud6T)}FCxYm~{BOj*koP#N;i@41dT2?f6Lt-5YIzbZI)J)?Sszh{{D)?dbC*`Alt zt0aG}2V$*Ur3vB%;ndq--&q$!r%cf=9D=fb2`cpwtmpM5CCaN;0+LBZ5LFt&EU?a}YQKn(FEk&0Q`JN#{Yw-Qfc}7hBq-8ZIYzh>J z#A|=KYfIH?j9uL_)wEB#^5(vI%g<`GZZVEV{39EbUe;-@o>(=T2@E)GZKYr8iiS%i zd8{?;@p_EO0EdF7{B?^#`T{JYmF|PQ79+X_GmPtag(f<9y>M2Yj!n)o_g0Wn=PGky zhaj$+(+){%yv|FuNiQ8IXpX23|CrpMhhs5%$-@H~wVN2y)#%6LHwrEGBKQsha^fAS zA*soq8-ZR+7=l4K++`zQ75V30iTT402o~@I2^MB;TJ7cs+S_sos85T`;sNbrQOeFG zuTC;Un9eTy3{^{&FN`2|PU^)q?}r=;JW;0|6X8~CoGXmAx=g7El9KiE+j?_p-{J?gn{~qJA(PkgbE8I7q?@ZYYN_B{Omlf z-oDEY#*{$*VCK@INQ}geoaBv@pzF`~ZoQN(x7EVv2ElM#Tk?@BO<(?RHY@+zXa_U? zgMj}UCgahYj3r|>p}zX8gxMY9rHycWs0plG23|oQ!jeKw?w`P9rf2h(ElE@wsznE4 zCh=EtT$&ls%Y6;K?qGyrv|3SL_arNJcg+fI9p+1DUfZv+BIRLn8sEU3ZyE8H&!`da z(;)l@$H{cn%esaCH-K2r`0Uisuy%I0zQlt!IE>4nWp;O^?LB_|q*8+mJOkoCx=o_t zIS=Q$=L+mRIWWjB8gZ$u|D7))W=U*oVCY2@tc`N_vNt`^CD>|_wk{eg_9K?A%p$lG z87FIdg=s2`2{%B0x)vAd1Yx8{!I_PwHGN>Rz3m^)flgQ`D?>?AhUG(7t9;+(I15?P zTBS|e3?e-$=IQ78cJW&EID4%s)iXTR3lgOB9wcv>EohIyQ`I%!!$c{0H798sg;F{l zbC&yM3Omk!X!D=@B{CEEpt!1Ef`PT=ww@zuiD~N*KdN4KEylX3Y?%S|Gf$eSRUUT? zXR(i{(+2oTwk*Mu9^;!>CnHR=X=o%<2l^D9eMA8^sVGpR{Zz>BA)iQByjJOuW~m@v zqPqvPr+oA0nyh%@>a|a9pqPUwHr=%x^wTnR0gWpr{2)9IPoL{bnfIxlA=*_%j{oNz ziUf9#FB)nCfg?ngoVAWnNDj!Om-aHorr2zfezINMDc&5R3SlG95Bjc}MTc$de5l!} z75T0jqay$(TfPV`O0MqeTDIdfh&${Mk&{!7Q`rV$g>(L~$P0Z2a~>GScgLUEl@REv zsnzZrK{y<~ZidA)EC;!HF9Nf42iK9;GzE2dnT{nq5Xt_?q<3scEvS1!y+7+lb;9|! zCL#_#qaRGyI^r1hGhOl!3&wDA2TgL({xC4odE= zcQ~I>WFb{$rOc(5q@8wxCm7u+3#n_i9>E ze2RIhvR>e%^-bvs`O&9V@Nan}ha$5DX0)6fb2*lH;2kKJ;J#kfCy^-^S~M%WkKP6XQnP47@2lV6u4*>m)}$kbrHAI|_c6 z>PE5BwBl0BTc-MXgA%p(Qn&p7AZ~yCujM(WoF>f%B>R)bJH|VFzCj+ZT0ZC=*347$ zo%)q^ZC~?oth^?t(RO=1ufJSTP>Rw2H>$e&gGk=DqLNpQN7B*>xNTqSy&B#K*N9SI zE6o=W+4rCEAv?;eLD;xQ?vng4nojW<3A^aigcf9z4L${X1M3pcbIShcwl{d$jXq-v zm&X!FSVTUuVnKg4KLw9Q{W3uaH73k!>67jIvv(%qi`R=+FZvjY9G*8+J}|59B_7~B zo%7=i`Yu^0pgZD*N-m#gF|CEIj9KjJ<$hi{u8~Iztcvu(4 zS9G~AHpnUd+Lcev)&l#qXU7qZfrmVUee}Kfnpczl_`jo-AHb_o7_2u|JaSU(jL_c5q$)>0jy>Br5@XeCuAXe$jQGZEyr+0RFrQY_Yf zwhAqsX>WG><#{SXsXiAgKqnoJ9^(5m8=IBQi|64(@2eweHV)dK;8qk=yV+_5`_I++Q-gh5lLBLJBN}rFi*Evm>lozQTbe z>Ou{c_iz^GZPHZ8R0+LG4au_gwr;x0Y32&mi>_|8d0Ph*eTh9(ttHjIV=js^uZN<4 z#O42P{#0izbN?+?S#eQo2XOzUN;-r91uG%`@3Apt=mZ zqDyy*Q)j2TTw*bmTU#@-A!-`;)}^z{i+vAWr$l#ZSy!RDJR>FFy-`Sh*C;3OGSV*o zeldd&$W7L-jx#Q9h3=uxxAbm{=&V`LoE-%dI_qH@KQg+LaF+N)I~X~&xarPQ{=O!U zS3B=|Yg=QjeeK!!F}NvolnK3cNP63xyA}BT@Tcu@rrH)X$NWLw;uK<5qM$vKO)U8# zbMsTQ?iv45%%v6T$$&oaxixFWwVEVPlkPxxUJ$Rzxf;0TD*7z)qVvuq&Q!a2^d_XK zp2&ycr%_Z+q8TbdNju=bc-N*yVENzOubet?y52 zR|UrA*q}zZ1|CM!A!R6`rOB6xe#KpNGqzhu$_f5xjec+wnJbTMytR|cNPI%yeY#$I zM^@0Qnks9q7N@UF^YgVql~W*H%9)fxb2~7D5#TK?MlO~-dyuJ4H+m9#B>j@=Q3|qi zFRgnp9WqzxGv)hIoF`a}*eqmrcH?X=g)N~V$3R=kV7)iB<*c~xQ$aJy%UT8Oi5jxI z3QxX|BNPLebM!@7VtvOHl|TA|7%H)#9SJT}@=VpvWZj<@?&;^-!NmXYZ&UsXdKQlh zZNNBkBzr>4c?g((rlnNP-ewShY{z|1vv@%h&D56dib$9U{#uppF+^N<8KW%-`y6T~ zBJDWVuhF3i_NbrF)64c43#%+39*bh|V<aD$)F8jwWUc`CThMiN=hGOqOQ5 zG{3xmrandU>2kM=S!0j3%;cKxXSosFjP3gRPkXWvP_d3?MoN zn=y+bKc^xjMDkwzccapvqFjzwdM0FRoyhd6^t$lczbdX7Rq59}m!1x5M|aGh&aqa7 zQ97*`lDdOb%w5xsDRC{C=mliwY}nNN5QL9K*6dzbgGNsv7IWa-3N$sk;Eca2yHkOq ze<(K8@qmin;7duGY)4PNVa^L=w)w*C!UCF;)1exu$q$j}pZBV1>FzxnCE<#j=G1tnOElvQqhCw`9nYk*Oml+r2U8iNc=NaEY4Uc1FD`RMRN zgw+Q<8?#itYbb+c14qR}m>2!7y5%ic=SA*tiMKpi>rU6C&Wxq7aC@w60=5Rrvvr9-*qF|&)S6Z7)(KyVFBB~k zCh2pfJ6{T}vlb4A%5p9BTw!f}zE6`Tla|QaxqdcvAgBn@YRyxp)?cL}c(0Apiu+q6 z-xSqRvXv;`BwQLWGpm3bEjgl4oIgr~w-Qa=mF(@_(c3zsNQy7Z9{7mI=~p)`eC^S`*qbkzCJ8i2>?x z%4e#Zo-q-BL_fMPaWN zG=h{Mk7v}CH44d=DG@8oFMdm4dmAt5C|vm?<`4=fR&J&+=(fuit~m7PLnX*B%3#IT zT7A|~VlXcP6mOz@^U@g?6-=c%vd_uF5_?3p9%^C=q*|?#oJP1wvHSzI86X(i4-2MCd;4&aKdt(PMvPWX!#`suVn&;_426dPOH8J=cboE+z{q12z9WdBlP|&=>ex_16|HZ|s@E|= zFCCFhiCixLumnf>1+jdib$-Jdx%^TmHAH>eI&{Oaf8qZwcPzV{`nIeG3{oB+x3no>OfuM0h`r*3HFq!&x&z6H89rY%`?eTiun! zbelKB)o5BrpS`G%kMp){mzgwG8}~;=TEr+1pp=3Sv^9=0hIFhP%dtO(@f5za3W$ z+WV7zZW1=58@&j{E7IF)?v}2gDYku>Q~T@>LMJVi%Z*oUQ=%Sn3BS$}wX4Q?|8lEX zLGyF;8P%al^REw{@L!~l$TuR1jc(NZd;jeUYJt7&3i|)vXtI7>Y4%}zRO-kpNVdRr z?s3;-(-(YMRkQ>(xo_fuE}HI=W`=LhUxJRWo3h!X7Ch2^h^sOT>2Ohd#g)5<=0dgN z_@ViZ^=OoRv$73KgHtURd*ahuAz#jEt&OX0wlzbceC%Mh_D7^o2`Y`C0Y@oWOHrld zS4NnH(3e4|nQ6Ku{StP)3Rc=1`d8qWQE3hDjrF^O&ba4eEwp)sTf5tKp}!pDQHLXS zV7;UsN{v;aZW^hsX@0!JFCrxx^`zxtqk^xQJK+Mx8;OugdceY90k^g`U$V2GR$FD+Hw8JZIU-RBKDb~60?R+b>d!Yh@8G)khJ znFs9h+}I3so=-E)b#NCzIXS0Rv9W~r?Qqr9`IWz$^ zqZ;=ZPH5;i06v18O|8YLas;8KZi7%heu+Sd-+xQPG3M8tka1u+3?79sD$d(aIt!&Jedei%{yDU*>)?(gTQ2j;pbFR*TA8y zvY8&Omuuw=j7({%cg%g4xYU`0;8K_Lp3uCMi)cR9Q1WxG;rg|&A=5dgOKR37NawigcA4m1;_3pbOC$hI{Ze?GxF|X$+P~!p` zoX1GIqjhjr1S+gP2uQLeI=$&Ul6BZNl2|ZMF}mDQ#gc8#$Tz-Pt&IQ5*qL=MR{v*7 zl5SPW8!apB9``p|y8iz^r!p-z*F|RQLFOnkLfWQ#_FitW4n&JhWY#KKx_y@3@;ndQ z=5i*uq*<Ev4zzUHi#=|oUe##CIc}kiZ-ak|H-`vOivz44;V&+(9k_0CUW8<*kg7c2mGOT zg9g4pf{JVqNfN_~Lz!q&2jDbFHG^EFxP+;PEp?yZ5e0{*BWW942~rhNpQ7bS+qtO4 zv4FfPZR4!UuNCyzUiqm0?dHTaHpA_3^S7%OXnWtT6|68q@dXe-y5{9GOsTo{Dc~_w z=%BxcE(ZW|Qb{LIDfHl9D7<3Nfj^rbfT1#bNGr0^J0Un<-~#>!2ZRvglnzpjq>ium?QxrF=2)7+O1@ z0!tF0RURShE{942lCkQ;-zxqeQ*Rv>RrkFQt8|O>P}1Ec-Q6&BOE|Q2BMky0(xotz z4Bb75qzoY;B`KZKDImYY^LfA5_5L##>^ZZ~IeVYA)_vb=?X%N1#$A)BkDni@GzCyQ z5>t~<;@3wQuzfJ1(FYZuI_uaDlufaz5fq`kBEZvRL12!NQ=bZ20NEXTYEvzLwz85i z+xAG6Eh$==1^6eq-XrpFARe};+UVHd6gI;@DYgKYZBN}pO2aK_Z$k@Yft2Qq;E33m z7XL8z-$X6QwaiJz=pXV>r$sC1ULi18UU9e;BntaF^7beb+=3%-S4Bwjj)VMTDl+3P zVSv2@_=P#1M0HnmZ}NXq=ge;{Y6e(p7f3z(pL22ygHO28elm%Yt^ynK;A#8g>1U+o zL`?b^J8nys5@82Dwqfv3DI;Ih8e**bT{tQFz;{p;E4j`fPJhgb`>$~RZQjlxiDQwX z-RD!+zu@m7XrLeMD2K98bZG_e+?k8RTY@lc7(ne0jfUW%t`3Q8B~LkHV5>r~Uxu0{ z6WU7z^WUe~YwG*mUsv->Bs1P5|!z-I4Tp<&N$6L_Wa zcE3jXjjOGF^W#6`6WXqwkd#TW(0uvckgVDve1!AH)l8wQ|1a8B)C)I5oe9)oefmHg zp}Z@kOVc^n^%{EYqpAtfzm;}V{BkNyM4m%zdU+zM#r8h^qK(I1tIX{Xq(oDiC8EC8o$Y#&G zsIHh_>sm>+e&!V)f>S`y2sU8LUM2j!MD5b6fxSYgX<@UeCSe#@juiyw8)@QBysz@a zSNxa)R#H!Ihutpda=CER1K}Ha24r+NJ9HMmq=~{k)cF!16aYkhOOWA%6&kP6zAQ-$ z1l%OY8Q!3$LNOX-d{=WICp0vt#sRN2wG;c?cJ6W*H$|y(?{4EaB6w4RvGV0JCyh2& zg`n|eFmd0`%6kURJL+e@56=d%Pq2=<1;o_z26(TkxmovlKSf`uYnwC-Rhk5{lZFF+ z;*LbY9j26Nsk7C$0@U~*rk*cF2e!QQr!{QF$VKImgAQauD!@p?CuI^qKBxI35Kv)5 z6_dk>zsFY+Nbqh?(Lxjfza!Y&HSy_cOO%9Lt?nES2x)9N%|SR04HfreR_@oyd4S0Shk!V?R0$TvrXO-CVvT;>)Yqwz4Q7bQIXz0z6QrRT`@M+a|7GAU*nR|gS>Q|i)ZMap zsKg{Zd5Z#4syZe`z-E#+cK^~Uhw~{G^Xmj=6HA7B9Vg}mk0z$H`4OOzIJH`0 zmew{Hd==MDXe&JcU4F>Y8bTlB)`|tPXU!HmXJuka0$M-031CXM@RH5ZV^6hKG970% zI<-{?As|Z{r#2=B&7JgcDl`<1nH7OIqQ5-wKlW4%ChCHz>SN~hVAJ7P zeZyISA`(e$yT~1N5?Gl6y37Dry~!s5A_~o&N=eGv*tIO<2Y@ZFS9DG4y!sj#A*G)h@qqka-{kuf-Rt!A8pOP%@mO91tY=gCcPuRjN(V#Ln-Yf=Vz;8gJxXhA7`^R}4m1;iqTzw!h5dHx6IVnjS-&%m$;IPgk6ge|?_9oAD9DS-uw{qsPr>zW0 zJbKZtd&12;!8hZ`Dz)dpTF%#KdXrTD))9I_8`5^uda=7cc|t_A^N{58Z}4+(Nhm&O z*%8B1rk7jhnhWuE_E!BwkN12y8NCVcN>`;NCKft0y$-V17JBMZuTx5I^z4;%W^Bk_ z27jJHJ64wEgklXhpk(!Pu6MT;&0l8LuURHQj9f0B-W{6LTvTEWF zz-o^4VPW()t!m3}NKbV5){ke?n)w?JhQ7we0}oWOAFHb1U9_Pw$0J%%llYaoFFGZ3 zh;YUPm=Rt#pTi_yH6#ISi#MM;^kQbubKR*%wo>$WP#*sME*0LQQIs1Pk&x{X;RCWt zYNxOYp;~n^sxrT~==)0!p|!a0o4Q#?FL&(hHyAW-D>+3xmNwk~Y+a5{&{1IHD*XEN z!cD36b;go*RYBe}r^ocay_y)u0^hSD8a)#(09w~rM^{803rVc=jSBO9$Y3Wx+nn%+ z@rO_A!|O~Jl?Bm+X1;NQr?ha+wO)Ot^hXsEp&TsU2U6LuGcFoNw)jN2U+X1zxbNIh z=4cWQQcXd&X;5_C(~7HtN4pzJ}9AKv>IPg z>AbmDo9N_#Z3L89h+KbgP)m0-|Gi2rs3EvYXC65F3d0cbW}!~k9061zP#y7hkl@p3 zalQEY>TmqeUGN5aiL6B66@=y*e{YGQJADeTp=WHB0_WQknv}CxoZ5|vj@SrK@@2W_ zT0a`zY-yh`CM`>nrYQq7Zph{q)7?Jh%B8eD@yZA8*T5pvhS(H-m=&;Z_Iq8_l09VJ zbM9tXd%az<{1EzOL_%XdRF$8zI&G0YddPt}G$J#i#p6OI z`3BBl<1pK*S4-7WnQ33J+|6-sjpum<9-P(moAj6JZ>ODKhY|=Gh6fP~VhJtNXDOXO zae|-MZQ-QV;|i&nDK?@Df&aGlZ2Roryq^0G|GNWJH5W(IeIaBXWu3}$Y9Vy|OUeG? z@gu}s_q{Srhw0a#&c#vtX!}x!kP2tRgXpxsub91zRsth442$#LsL|FaYjPXxPUcvs zo(k`X9!s^ew^{;jh;UKioY*NiR3sKc1El<$gfU~tT*I#(?5Uy#e5}~7)YA{@YQS4+ z=?6aM@{03i2M2O272HEwmi0V-F<8#1PUc4FgsRgz$ve&n6~TCs-h>_mJ3!*0OLG{j zzUtR#5=B-UyY1&-vNWGL6s)7PyuFD2E_R#|x1o*u#`8R%-Ip}6u+>lgzd`?VOg^&F zq|2oPGLI?x=ei(mm4Q0bSWuN9bO*>~#8|USQTk~%0%bZi^vq|^qRqYT?rYeqfsgQsev7`sla6o^z_ttCSyZzNt==B@m@t2<>ntYKRXhD^ z5LkmVWNSPhwK>_*ZrvZ$+=#r@+m+1K#E9cA80hi*u!9X$^85}mUd@hw3w5HD*QGc3 z-jVImXh41B;A`C(5z`T}s^;3j@+9VJYMhU2)sq55qxWP*rZb4qn2qyRZ;N}6?J!2- z4;?@07kLHtZb;G1J0kg|FpwX@9de-$UoF8Q!NZ4&tjrbwJA9T^G^lISc3ApMj2s*> z{;`Eg=T*0px0xjZrT6}yrDk07CnqJ4hyH(!=e0+6MC~;F@r0@&v@P5gs4zvZ@CwW8 zI$VELyoGIjB0E++z{9bs$wHNB4Py~ z*ESm0=y`t7&c&dAz2tkbI!8IZ)DC~Fi1q&1f|RXvv%d>i6rR)UE%iRz#(1n>!T-XX zPeK#@W--ueRZ#TZR9_k3AyfdJms#tT=vZE{)VEZi!B~d|;4`eT089C6G{kuHYjlfh znXF)Wv>Ije6v#Aojin_I3+rcZNH4TlthxMN`Tm1Hj~I`>9p?g2PL*?Gxf~m{RS7%4 zpxUxX#zSzOxbWiy@9@%-HJCbHS@QA;%Hs0mc&@ztV!39L+6(ZKoq#OaTcJDy>QT-j z)7ic&8Yh?hBi(VvsVteFaoF&JNn5V`H4mVS#!{&f`_Y((%4o_ogDmfE$qi^NC^%em z8u{>ei=1?9&GG0aiG{(8j~}Pe^|MS;S5y0eQktV16UKznmxYyl*Lbsq@}xTq@H)~? zC2ycNtw;E@`=x3aefV8#C#u$6?5n6A2bF6YiFNFk+Gbg?xo&`#az?XSHmPrMjBTis zpjgUyq~0p%+Zz;jS+G1^6V`HaV4<%ASQ$+2tom2qfA%Itzix8k98U=lfPB8?0)Cxx zK~(*dUuRLyl>o5mAnz*xEUqPf2r$(cSk7|RoEKzFzLysH1mKE4UeT?!==6G!v6mWG z&_emD0ehQSv6;<}p)LFzQ=k;&!`BYhA01mRbs~$tP%cRbuh-0t16eF8a)P>u9{9K^ zR)d6pO9^t?_G};bSoxt@cE)GOauud++;g|D{*;AD#0aSybNAAPzzq;@)N>I8Mb3#I zOn}10fzyr%2nuMq0R3xNOvlSr3x6e>7iD6rS!n;69O*xOmwXaPtNZinDdCPxUSPo@ z8{WDn>SKVZ>U&X1oU!!cvxA4ohGW@zT2-@i95`Y^i6}9{1YSpG1v_B+KYMHWhu@hA zEG!XoPL%TP+$+rfwW2!zUFMyHg%6+S5x=J*RQXpbTtgNrX+vAlrNQRNZtuq$!uf~A zA*`y-#YG01u>8GBUF7NWh5H98qpjlxp0_AYmK*{N?KT-!sC*r<&bCcy@ryU^TR`B& zBN{}Vn)`N1KK_d4_Q>SE{@7}WoASDb5MN$C6N~3x3o8x+fIbx$lLq(bK=o6*t^>zT zj9c+Wyvb$bZ5Xp^xJdI!Ee@V`NA$|Kv+BB(ps_q(tF`+8$C+Rpu+G-c!>c(q_A|#T zx;#rjS9_(TvHxgm%j%j^96(L1@%yB=IiIB4q51*mhsiLin2@};o_UggX!`&Ew)UrV z>Cw{ecTksE089JvY_tB}+Ki&jMIc@EUqjUr-4Cq-sdp7uR1@E1sb-f$*!*LK2|8w2 zH2^#Nwm9cvHOXV7oHMRoL{2Z|=P$CTBI37q5V zI_pB$evi)F(+4_O*G$PS0#D3f|Mh(G!*7E!Jxu(&t5-ukp{oAREw4EkzO_AMPXT0q znLzExM>YUgJ0Ahw%#2=8U$wBA|*&8GuXMJCI>1kmOL#p3h4xB4qW zUKIx#Bi!IaJx8Amk7)lMe?YFF>F1=I{3e1x$Zx*bv!HvWi(e#nj8*yeQ`fPCge`y9 zz%G{H-Gw;pON?I<>LxE$YktnFFEpcppUk8Q9j+iq44opc4xP57CPSU|oG*%2vxFT2&|PWY#d!$G+>+}u(*YP)~Wjdda$JEJ&?gjz=vQG`mQEKR3S=2@b;Qs(MTgA=&q zX;JP^yr+MzojlA1&F%~}OXa`&Q7*B4a3*o{P=Xcl5;H{rJ0?cKu2 zL^@V3(}JB(`V-i)J1$sq7>*2)h`|KJAo*z_x-1+bOJv%$P~&qn1yr$Vt+Qbmhb<3mjhy! zv(a;Z8HgDpUbiN|l-+9jqa$!EA1cHp^txU@RQ)|uiyecg8d9;4h`n{=QmF23ts!B7 zX2(LaG1soM;(Z1^leEKG7ACt8!Z%T6`qEABoV&ZX$jn znIc&77#7L_uAq_@26GNiPQ*W@H|wLCrQ+wde9d13v8EPU%+H+r)QcD?+(V)!Sw(Ul z(6>`S#vG9A!hk>xpBwUoT|DN$H@H4t4TsG8-}|_j-<~pVqpRO)2}C%URMJEaq))Jz zW#YdW?-_sCUc(yf(D26B&HA8g60Fx!zM65Mv@xR$v4v*5@3~1Ly&s253~hVZwteTH`B}~?vS7F`_TjZ4PW@nflHBMTnI%S4#$t+F+Jd}GXT@h3 zbvGo#xlm5A3gT*GLja8p_Z;T-#!3{IdH6Jg6w40>^vpjA!my(dpy48ox5>=aYnInRu z(2qcYWTFG35PL2#sydQB@D#H%wisRJ=!+5z`&Xc8HA=J#mXA)K_ThXLkqen14Y{PU z4|d?yMd2GoX-1KS^oi6(t^mS+Oqza}AWDWZ0Ff8H_KSTURt^p0~|`E9m{JC!pk59lXdGB;KR9 zocJF?-+#)usqJut zd)rpnwSmoy@aMBnTAryfI|Ekn11;uWto3U@p>7nx>K&aWjBxZ~x!T%W)gixt(?jK!zW zbQ5Zi+xf1?S>)WR4$yFQsen`-?E|PUyfel-2=E%`A3$5d(?EAG;Msbp-t4~Qv z6)gSSBiFa^;<@T13#5xU*hoi>^{&uN)tmBF=qo1GmWfa{#>zxS;(^*a?v#s%^jD$n z@=^g7jMOsf%@YPavT2*ODR>1zH3q4MIE#vc&q9$^QTfA(d`wh|CLh>-k+fP_TK?+DW9akA2p!n;BSgFOl=IT=>w(7!@NQh zu;bm&T{byxm-a3kNBk283O3rTU&BW_G|tfqt(li^MP4_)B%)PzaLGvT!!1Bm#FziB z6?krhTsBZ7IL38YV}CW+m%x}3(wyS206)&()*+z^1ztVK6l6qXK-aB-VC5C{s`y6c z9qzIH8gH5zf{SGH{Dj`SQP%Qf5}H2rLXK(a4^89NdY{@jO@SnubY=QEE-*$l&(yC$ zn`B;_YW8O}-+!z;m8LSBv}mNo=2(>)@|+`YyvF6vizn4`6X$tdj|W<}Vxrb9P9zmiJ;raO*R{-=w=2pt$eqqv^;dUvSVe%X*Qn{!}*a+$KPh56U1Dj1tj zcMNn%sI=17*nf4D!x{W}iHy2snZHJU_DHT~^gU&CDrA54NuuAt)1Ma_i;d;5))5en z6g5QqvF`TCmJ$wxj002UCM?enpItsHPSf7o75ucAQWFcy_-O%F^T%K+?LfX2A{R$o z4{q@_5407Am-A*5Tga1GVspM87!z&MQ%qE}8im{3W+(pEJw*_Js4vm8@%pGm2r%~<%(>UWh zn7te1&INu#xzK*Lu_}Y zidpxr`EGiL7@f{>q)jSHfRI-F+_thDke5{_CHF)x!Y4@eI!iEw@Ta6^X)DT)&6i1$ zw)EMYfa`chm%6lA%q({9XS*LeYHiog;!{EChJrouGHt@au{eg%Q6<^oR7F*!H2JjL zT(?7qkv{t}!bg2kV2CExF$oZgxblcfrWP%!6I-Tg96$$tc zopbuDiW7tJV|-inLd=ni0zKpgACVBZl*GyocswhzkoAZUu|DFs@#!vFXObD; z6CXRJp`^D9Pdh9A$N|AQXNRqPJZg;$004OUKLmG{igy?ky7hyfX4jqMFu<`dtGUb7 zmOMw@iC0D#m~f`plV@a(nDJpevbN+^Me|QOb6-V$MTah_@;JgLm!}mBw#}+NTWY+GNC`R@!!vx>c{UrTK_z-!si_}o5?+hW(IzO)y!Ad_r)hfA$U^%RJ7B5_CQ zb`&IB2G^8@Z%$P82NQ`?eDwwgq!SAe!wNzFbcv&Hh)hmYm&#+zCJ|+3qi@x7jNHCA zMtbCYM5?+?J~zNxo>t^3b8US|X5<6?t&A#()|vbK4_yj?SlIhIuyoTshK0Jc?#hoj z=Y>Lv?>KpC=m5~tF^mNZTYaa}pD@H&+TGELlytf1+oZAMOJ8dUTj7TU2aGY7Ta0(G zO)w~|aBL(4P6Qy}|kJh$fC5Y<$rddzGP&PXp`)l$%|cvS`BSLRii@`{rrTTlcjf(+>eO_l@v1@n5+zUe&uY>OekM}5X?AxeqPPq zZWgC7<~fn67`+^#cwSihV)UO0=Rn6E$oo2fyHF)QNjF!HLb`kjY-uPTkWWglYrnNr z4Cg(PPq7`%L28%yen37n+l=kTUQWX44?L_5*KM6Ck%t-$WucP^zP=<)L zEvP&i)`)*ta9Yg=ytPBtSwQh!8Ai_$m|hI6mtbGMsq&X|X3+-Sct`sVgzg1bOm7rK zH@AH!a24KPZ7iDdLEob!W3S{e+NeHsSRJ~zWM{ROet>=DnCdl7&Uown>4zy`YC;%L z*Hopj8425}R8>Be6t?N`_f9%FUr)fM!D4ZRHeFiwP^hEnZZT|%@t21$A2ASE!n@>3 z3SPVX6}*45Yr|mSitGJPun@A_7VJfwjoTGa%9z@skbJc+p;kYD+)#B(X*_&Z&cyqP z9d>KI4=VaoG)P->9R7~57e8{q-%_p#E6|9Rq<-9^r}-spSu*9V$p-X@`O-3gFw>ty}xgKEWgd ziZ#4Zb$N+s<+m)pP=R&V!L_QtD5A+-Ym-zNM9LB}CN+LI3SCx^s{~_ZPp~qyx&o|ZblNvcVnAknBfv{Yy!hkpyP_|n3g7=(5J#so z)USDkFZdrCk^q?8JzaD}V7wy-zPR?Myjh#zveUY?lv65lq$39Y8m*yO#ch8@2`LKk zwg5lFL;I>C5(oNrX)Y-n>b?i-lp_;auX>2XyD$MAU6@ij4>KTvAu? zA=^jl5+B@~Mko!;lCX-ZUw5sFJZ(4-v+%S1etGSv0|$WH(8q92t2Z=rHNu|O8*)7& zRdM%@EhWAEr{#R8Afh+v6ZKEO_Tof{9Tx2PnR|&X8hvJ3V1XU@1jpP-&VI5e6Ku(= z1~cxJM64dVs_S}d-zARgti$*Q@pV6k#UMq?-T3*?uO-q=pP5-{m|2wVT=Qp&jOHS* z94m1V;l5BVonUUvs>>^G6H7NYK+_E#<*(L~wH03oM^Y^Z!92cit^3aLQI~(XH*em0 zvyz`5f}X!|>iI1Kb}>kfns0YPF}(5=QmM`#ke$nT^*j$Yj{z-F)?L;cFD11ZmDH9jpUFEH=J(T6d`J-~$a`zW5SUpdC-SXUk!0oC zeS-6Gim|7ot^8TB(}Vm=%n`)prI;_Iv3+EHm#1fb>Vzo4H=rERVsukM)inQ>hO{ez zfn)LZ2QNV!8Fio9yqnVNJCD~Esg{TF=?CgU!v=m+%Us|Jas!p%fa?+~&RLc*DtEu! zIjk*>#ecXm4su2v{3F_|GYxqT88vZF;aZ~4la#Lk$R=3F}cakr;44!Jjyh# zA_DB*UC0_cEwW9YdcVksR4o6w$Vf!}kG4*`@`Rav-i??(H?rQGQXO+g;j6LDuvnO%!Gs<> zff4-h@WBS-+i{8WIN^9lIfQM>tZl+JvpfYYSrPykwPRpCqrnXPUMZ4>)-|ZBs^+X@4jlZpZ5#{2dDepi@ySpXkmGk_H?z5(w@8n)tAUBJVkYzr72d zzt524Vl1$H{SGsR1T7FfW;P#IqLX>_z0M-|aQOoX< z;(4&3#}n$o8yLYxMkD3VfS-yziJ5->-^c_;O;A1dUH8~GHB*75%{$b`zMuZ@fUGFc zym@uj`}JCzttAX#QyK>mnk|%$BD3hoQQT6xlx-3@<*%%e*D#s6dmO`+Q?NSb6Cye z1efTrsKVZ2Ry=@M^3ykI{v) zz)pUw?mUp~^P|TY-~(f2fM3oJI8j->OwGjRZ?!|fXz54-_YnG`{PrQqM+g7z`9QsSkzYtu~TVoZAHTM&qVRRcg*ukUk!uGB*W~3Io$a=tm_d z^5mFF6{I-AYlTC2e@Ws`-v&POmFznHCiC)l*aThdia6;xC^!LW@0Q{nQy46fI($Ur zwR`iGu(vWo3Z>U8^n}Z9)k+>#YVE(EzIefAs(FrIcsv<*&CY-1@@&tFCGGLrmx1RE`>6&*evrDb4ul2cQIq^$LEM*cHm?p<`_=Cl1&24tXuiMrXqt&tUGE}X*W8Q4*%Yf^Cv?#R%kJ9C`KJ_OC+y-OB z$^H7sWwr0q=_~?=wV~OpgLONO)EUQ)PPM(?J`Sz>=iYu$=+|O0uQwt4eURqF-0)HUxMEF zf_mw*Hh@HRKcmxm7hXq#s7P>7bgq(|?xvDE+Pjy2)s7oxqx!BKMHvO?1pX3tKX_sL zEakh$Bn@7Xq;MY1IKF`5=1-%eX1*7JB6REz$BO*M7o-KGsX^x=IkSA9L!N;awwO{( z(V)Y*AvS;h~TU{TlFOc_-{aW9DG2c%0I;*`j@06jDi)pGi2~D<*^)`a%C*YYNCF zl!siHed!jFS`$`kSE&=k< zNR43fY~7!kQLoSd)5~MP=5dJUD4eMDBCf{;MVBzEHuzfepZD709Yw5KNx{ghRoX(t zs2<{WgwGimKnAb6H>d>5$CIvJLJkZLVL9l~h~oMF;%laI?)R1j*gW*jp@e1XDGBDO zO`(ELCF&`M1;$N&kq#yGlhaAosJKVKstdUY58@jFHy>60Xy(=^TCcqnYkb9I;~@3n zHHr)+fjr$<^>N5R2!i$>+_+4SQdJOIuA-P?FBow70{8Bk{7+ESyFrt(#(xtf0e9VM zdRWR)9WcL(8(PYc{P0yAEtr=yce!Z`>ij~{m9lclcn8)*2ON?6ocfw?LJU|38aUET zG52Cm#>WwB0mp*zlpVhRj@l^h+U;7JcmKt|luM)G=-lEpcG1Ck9JlA!SP=7I2|0r- z#)@UE5{`towwefVpl8GAPV*(IIPc9I)K9IVA|3k9fxMmQHh@ad0m)gUUNjO{AA=Nw zD3~ZQ>Dx-CTzP3gS!b#5?ne@DVY;zFBS=Ty;3`gU+D6iIUGMq}H@U`*jJujGKDRg4 z{?)+#O&=^?r#D{^6&|AEb`?LfxP7lX-CpF0>?Q%O+2tdL*;C4<`zp_4sDW%13)INz3VT2vBbhTmU&K-oUO0w-> zk<*Mdwm3@D#eO@!BNoFm#k<5^0ztq&GzK<$WReQ@ik5OL3vXjVx=1dH#>u5ywz~H{ z1cHBOnqS`W<`(Nif-KG!wv2+kP9oFQyhQIzZmmn`{~*Uc@VQi6k_Ib`gzyHN34+6W z8p1ZF<5|c%p8%Js9=gj>(=kUI2Czf*&>-Z&F75j!5NhuOy1Y@@+gIAm0Yf|70|Sd6 zkveWEE@qpZxw(uYML$AKHF$4#3A=tfaaeFz#)`+S_>4wh0Z&p0aXK7c<*y2^SuE*1 z8VGeFUczX?7(emaI2st$`D<2p^;HagP=#Uw>-{WOu${ zfd<;G-F-}AI!^a2wqIbkt}LJMK123suaL40vW6Lm24pz?JY@69#Qh7$M}H1lNSBvm zwVT`MtU*D$;3hH{%=nBSe~lc}nZc&Wi~Rads{p8Pih|2EAIzII9lhxvx`=GC@Y)Wc zPGkUBiiOft{m*WYoGt&S2`%6Ua|r3g&IsTXKtC2i^#s`)IjaL(zipFNebC#`J9G!= z8ZOBNwak>^Q3F+W0ai9n3cGLcjGsWeg=6+;AjDI$Y4(Vwc3_>_bh*cZP)$l{%|?PTQnDn# zWv06pEd)y%y_X|%ypBH7u5>;c3NWQnycxg5$tf)*$Dl0MfZh%C3Tc&}@b`(6mwdYq zIde$>^BeBsjqlMMyR4S%cYV@wyd39`7QsJz1*cxIc1_i5+iD+ZVC)dz`wE=qO^~ zM`2QR2MUNX;we~>hh=59VOS&hX$+3x&bQeCzSo~co9mLR{CpnAelck=iHmT9KHK(6|Us7DuEjzFGB&&iFhSB7C85}@$qykVly}` zZdV}2XY|NlUf12Tif1pK-JAGaHs09P#vxxoa0SX{mR%-gWvfAFR8Il>^xM}Z`SE&f zESZ8Wf2#0?4RS!Ie6oxR&E|tTp(8BQ!j(og@du&LsF&sM-1e2a57ZPRaEM7}Kbe)G zY6h)T+0bsptqGx}0h&I;U^G(gQ}!313x@LL4H?^H z&bY?}IbI-?lBMq|RO-b^hJW^U#iQP`nwCUu!$?#uVOZQ;{T!%XRv+r(^%%3qgd}3B zRO-W+37QjsqjE+xU471R3Li3&&mK7pbLUGg^TV46U*rhb_)(E ztJ0f&MmP1xSoaUL0}n*1geVfa6#7?Z+l#wcX~;Q3rNm^gdlN&HlPE*QYSAQ?nl zn&3AQ(TlC*l+cd09#Yng7l4=f6H30oC>VDAgum$Nq>AdOUg|z+M)LPlo)4mdnSbQ9 zyFWHxBYECC$j(SD6FU)h9ZRc@3-Ee>@@E5|GRbKVIz2si9OEj;q4&P_eGAde91t!f z;zbnGTdKVbP_>tQZ)tRnCKq626tpBS3?-ktsokX45`+BSXCHiXI$ z@uu3K04N8~DwomV6w_)ZQpM~BO~8W*P^Temvl~)oImV}jVo(Uc&wJkavm^LRjwKUHO58qLg^O6>UyHg>l(o*Z&49*LV$ zov(ra;e)uC4h-7T3bOO0lC`Po-Q2>pWQqhxb*RGz&JSLyVK7q{i_7DC{WhE5B&uo5 zak0%dFClm2aoqJWQ0V`<8BM`1HtXw!IUUo=SH*^fNdj6CQPo2U-`G5McNSP`~J z>e&J$^s8DrRlkH9jM$L~qW3;Qeh3?2@4Be4O$!<4MGjUx4HTL9_LqWuxhAkYa=Kmz zgAx;Z{S)pWyPl8u!S)whaA2LVyc;2!a<=vq<>2v}wA1JS9{_{lw{K-}x1Bk^z75KJ zZ~3x)GY9Eck-sV&0R>Dh#q6|vFknsCrj=<7OPi>reKo_}@%-Z_7eIOBV!CPigYq}0 zgw5_qWQ*$;jS2G=FsUws8*TvdkGD_RWJW6HYaDJfBimdt4G5W3jYNr=Rk?dr(^O5| zVf8d@Y|`o0E+p(-!os#9bzGD0~s2Qm06l0Qyb42Zi~=(2hz7?q57sY#cjUFa%Txe_a^8{#yjV(ms;e+ z#BM+{ks)Guse>os@;8a)OlWzeRMaw)bZw9ptFoE@RXRmw`Zw3vyHstT+L(GK7x#6- zh(y3!(a1-)^irOS&B;-G{e)YdX6F0nWRt~(VpoPx49{2PJqwHvN@99lH$+)Xb&=+m z4QHp$=)M%Q`uPVvb`Z^C`TD|ignwd{MVC0E|FVzvwba9_d2;99FPO!DbG3Y0G+&mV zRU{vFz{v#?U;|Iju?2rn)rV?q90%=WXA*WY*<&vs_~0 z++8&Ah)g)d8FU$Rr0tiULP@h8H0zWqV4>Ms-z~h4#NDs$+pAZ|?{8&q_Wjv1MH*^9 zwh$!a<=Ef^vwXrU=OCdzqJNP4SOfOHraI}U7F`Z3<_Z_v{G34_@}pB*N*Q)Ked522 zKU(*(<^$zm@53t!o1>3o#E{1GQj4Eb12EIR){{FA@gqJcTuV6qI%+;B>rm&4oax+V}ZsLUYLY6cJ^uiHi8jIYQeyv1U)t5y?1l5gulBBm@?j41nxiI`o zxhrl8#g}mjNaH?Q`I{qW8Nuy70=4?fy!-0IZK=3FC(m)mHuIDwq#%u3 zNanj@>_Q;J_v{_MA5PuWOXs5~0hhxK3MO@=1Ur+fz1`FWh{@;UNU6i5tWlZ8(ZkH# zR{)|>b}_{F{DODV0X1ohTJMST9N!y#Wr{hi{)^j=i7s#w+a;NUmJJkR*k(yOXkc~* zy%b9tYF*1vy9vCK0q@2*Umd>T%ar0B>`OQM_5@!M%QKK_s#wK=_@X>e|<;vG`6^6J~M0XprD44s)A zHKJ#-b^I4osBZs8dXzoFAj*&EHjX`jjM1{*5Z}fH@$2Jc03Oh5mn&P3Re=cyQlczGhL=aM=?iG_U+k@ z1#`P$+Brc~LoYj8Tg~izb4YJ;3;jOW?^idOH5m%g~k!OxxEb^xcxgCs;~eP9^lnz(SKk zcBjuhr6I2m*D6M!PXl69qhSLIqdHs>t0ZV}y;@=Nq&nU90d#9IK2?x6X{zYBvFRHf z5;UeD=FPUdjYv496=pqnL`P(CuAJY%L{ABCW87%_zu+jzn|?Op|Hf~ z*tLBr zCrtj2mUjNE!&Vfqpwdun;%G|Hnn1LRdGJ?8A}hHq3r%kfT}+mLRM`MPwq!rB_`}@@ z9z7RtYcKou=!96o(?IEP`Q9U~@&QtJ!YUl*ByknZtCtzAPLy_Q<<)*Zy3@i@L`J~3 zjYa}lO@3Rv2JB9v$upjEO^ZEZ=C2bf2qP;c|BB)*^|ziLa{OT?pr921h{A+&8*31* z1zN8*i@b>Jy40-v?)l4YS5pRGh z3I*ED!UH~ULfmbk9d^ip>X?p?F41~aTJwhH(RZu0J)`r=VPzWh#E&=gW7 zC`iu+(#Iu}DIVfs3&1EMZ*|&&MA%H@&ucr|uimBxwQ@zP90GZ8E=9GO&0l@4LvPjH zc#sYxezx7oNF}$FN=E))KxqJ8p{Mj~YarAgRubD<1QTQ@C((Bf{U_Iw%A9Nm6f^!Y zpr-=jQbt{d-#(O(`k+W~=1xfNAbfe6pdg;c_ubl?F6o-nb0!2wk|i0nEV21Q><{N2 z8~_neGZy-#78b?zT&|NK(iA*ven;(e&c*Qf<7r>6*SM+9$6l)6_~B0(vA@iPvaF}hiFC7;lO(v>#>30Id#OUp1e8m?1q#NI`Hyvi3UK60FAEmAp}TK;ickZ z@0En+VFR0vFVMtAu~y++Lst$FJB%t3^e2hS2|7ct%I!XRESe= zN2EEJVbt%TuJmih-0@LgW4K`%zR|*LE6Kh_JG^e0I^yW6~3zvCOZSAe=bx7T2uCWjMy%;0%ybUo`a-nNKjgZ8lEw;SIhZIDaC@ zZLO>$iQY>BY6P49)o|EO>YKETSUq6~wXoiQXz=DWCv7gRmUtw@>4|`A@e$*cP#|1U zaL_1dq!iohq;9vNZTw*kpOU2hkg)Gt`*(dUY2b^9FfG5yh-M0PAAdO>uebTa8=`5T z#M*3F_v&nZuoRkB3dZA4r~zD2fAaR0Ow?{zU+N)AUH;7d zYN74hY^ElHNWS3Voo+kuOy1^S-||^O`;kBx4}idz)741XB}|Gr)Qa?ziSoV7FZx5( zMaZ5zCYcVc6#R7)SQxOrTCythPmf{z8^eWnpV(j{;3KsdEvn+iec|Bx8Dh#&rM4ks z@TOGw6^NYMj0}2Ki2>DM8{BIQ*vAKgqRk|bKoNKLYG)1QrbGvjM(qHQa=fXgJ`HC;eX%9vBU}W zkirS|657e^Tr)kaJk^MYw!Zw~eUkJ^pkwXs@3`AvBhQNV--{=B@XBm0ozTTKYHVtJZmxc)NML$9{M)3$U(FiRh4m)!E@PT)Hw?t zExa^r7}$h$a`)`>`R~-4tA+}{WVvZi0ymj6^X(?n8>NmB9fg%S3&u-xdMl8{juTsJ z^Mv6R&Z3g5y7mf=6c!TkH9DZNOxKGB(ko@C!z_tiUwrY0_})K5Zo?tU0jHLCFUgpk z^hD8G3+zjf7SQnVhF-eJW|-ttORg^VAq}GDwVPt^6-BBn;wlUTLMN#R2ANz+f*qLM zJKxcQgy2IjM-CF;Lnqg!C?%azPddQT_&G0m_n*CzPThCl_5GZV{~~zDwE$z53?V|H zDCtGqPBM9t*zSD%RAmS#tbr-tPq*lQ$|f1a2=2yw0Dc)ewB)!; zsaV6P{^7Nv%|_zdVlHGK7i7s$JOtGxE!&sUr*XPXlScJErAEpAOa6MR5*m~KO}{wv z>dEx9q;yM&=6j+X!K9)RJQ)Gv>stbM?yaXoz8{(S6jF<#Z*|%Id|uWYZ3q-B50np% z6TzHYM}wv{RrIiJ%RjyW{z{cm`g)Oj71kTLPx_W{CO+$ZJwgbb!G=CwQJ4@}j`ES@ zp_D|L$)0XD3O>ZY6tO3P@EBW;`W`NC>GPH-6$WQjWr6T2q?+z+b0DI*)ngr5l63bY z?CBv3T;*&ihXq^RY2_Gvr0WZnd`+|UGtZO;+S7^O63n=qL#v#SRrEDktFI;%DUXfW7f3+nA+J5#Pc}5oq^En7GVnw=*_GVzfhM{NY1a zqa6vPG!hb1Dud4La^#x#^11Gq665u^-+!M=t&7!*osEkR{Y~BHQ-nH5Ly5v-oBrkx zDeID&h+-MygFHmS>=__1@_Q17u$I=_TcI9YYccq2ZxK~T9Gk99JwvcPHAEthuX~+4 zQW-)LDSl(ubzZAXRtSbKfQ%!0gA z(+|IQyZz^f1YhoW3DxXDrSiCR%z7XWB-_{ zZ~9mU;i4lUEsI;^@FNuGph~TUb$Gk|z$+p^(L-KVd^oh{qpy&538pKO}4k_QC$?rN1G3M`d^ki(z!YHyRz@6N@Ut2co15wf9+QmM%65MR=;YExLVwsr_ zE=bH52Ol~XdMAgy)5=&NU~=QHbYDNRs ztq*@glDx)Do4=i&RL~_2pg{5me?H&UEIK{CK~{tU%Efk4MKOY6_m9yZ&o<#s{KhQagb1=-!i8)#_-%Pa??wU=EMZ9Gto zB*`M%*wx3<|8Hg0Hv+`V4M}jRx(!V^h3>WIeXr9`BYl?lsA>9aCwPG%(N$GlCLEAH zdGIuR=r5dLk^~_Fbzql0vyRu+<1Tube!=PT!5f#SVZm&Q+)c?=D>cdGo#QRh_d_n+ z%3iHmaY)asQPhDRNJOJ)XgJROYl$G{56nvV<7DbB?94MsvG|W-n z`Y*%{HpPm(!k}2UA$}d1u&-RVhXvaHsVp6D7|I1DN@r2Yfp9`gD{Z>@T}Mry!iLZ40sSvs)!CO zDDM$)22qRmsG#n$4+0{w$9V3Cg{(saWoKh+?a5-ylJP~cMoZIY6l-VE8O zs81xCM>acoH5kZjqi2?6v3UBn$BTTMj2C+Z&*IeIC=|&dqRA-j#Va`1zH5ZY^ZYf$ zc`6xV=KHQy9Up*DHa#kn>&GAh-FwcUG%le~(6oZGmpm0%2CmNgB!0OPQ154m&%KK9XEgeb{03TE|T*8f4I=v5f9w-2AiG zDM{kBEK5rZL1nKg{I*xFRfGlkQAzn;GhXtLh3*zTVMd^o>Ad}y)_hf=!GFxF@1 z`G=od7;wb&7HcLGv`+qYd7fnzOJ_m2QppPq`JP&NoLfOTTLItKp_cIoW^|P#Lb|SP z-uL_)o34{(_w+Y1BaJnIKsU@XJQgcMeKu=?YdDBNTRv4JWZgDYJtAq=HBS9~HgemJ zs&<(OM9NZ8UlNk>JIFEZdilEQ4|NuncV{w|_*GT=i!3(siM&b?3LsV*4j*)vtJEq- zOKMtux}lbKatsf&V<}bMU^!d9H$ah^UZ#qHWz_6U+^gYG)A(AFY##}ryA zOz0Z?XB;5A@AwF>ZhtfgtQj+aYf$P^EGoYSKZV=^H5Jo$0Tn8XZox43mrYd{JpV6O zdU#ZLL*Xoi=zYvYa$?}qbMh4FA{wxZaV|9)I;8dmp$wI zLEe|zMK#0ca_ z5f`zEG|}(mD{i0WYN7J&ICW^7!H2%dDhEThH_M#de#9%h9~aJ%+vxI3gX!#3rxl)% zLcZ=n)GTm$0nE7F6hdX2H&TXPcTQH<^IeQ1MU>o1 zvQk~5?^o>Nsae4zd$&<^MdLN0`1+rYTuo$5{yHhI8=5Jdf+gpvIuwU7Fg?v$LQ zHy3D#^{QWU*{3yhj1VCru-+Zd7fX4}K2z$X16Zk2DXeHM)y?dP- z;_-Vw%${JzEEbIjmaqlccHLxWzVzYp?max_D@?F|v48*GRtA1jb1guu(&gpjg-Yn5 ze&wr+rSGuHe`J69+69ZUVr2#AoN$_g>k;*k+a7h8%o9GmS4TDCbVr&RFhg%X<@UUqxjB(!?PGljcw;SK1>Ca`|@d= zR&SU|vkFF9=Y+Nx>9pmnliCSVG$tHOyg`AgU_z-CoAvpY0@8n&D!4kJiF9jD_x8R~ zo^uW5sLHPt){wUSbxL=wQe!&^S8A2N)n;ep!|M3J={7YD>zqqSzi$QSjA%$WbvndO z+oNUWhnp*p=OIl?b%g`R$5FWDKzZsiKPzd!>zP5feq*G5})CeuP7ZwbqxcN-_bum zAONMJeiP93Nx;$!G{8M1X*M`P5iAF?Ct>tSCujXrD#cVB)h-Vxt1RX?ke8fxG05kNt70S8YxUI6 zPD!5q&y7|{QVE0@P@)JQ?t*VQ#+muat8HVGC=U7KPnm7Sor;S+-BW(FW9;(eh|~vt zomZNjvBnc&kC+H?z!QOm-hgJYHYN)IB(}ieL;2LUFV$&5rtTVT23Xi-n8C+q2>rRt zA$CJ7+gXS|Dz^N}JJh#NW&-J8z68=v`D*ezaOX%a$xtK^PUWACX8V#z+kEmzB_4pE zsH1)LYW+b24rIDB2c%wR-n{T)8BHhVO(=p6;=ofe9u_MelG3?%NsSFfbCItjcGLpj zY-y}62I^rnS2pEi?SmIm-{H|AF$rbhY_N^ej7 zqG=PX5fNZPcgkg>)ROJ2p|R9qNSTq!4W32sYdqU}Is>aPS&Mc!@y(i8LW8Oga6QOJ7t}ZcVpyx%H)xNc9IANHlhwChI%91LOJ+;z9T02R86Qb*tvFZ zQ41auc*-QHCgZ2ccDf?Zo*W|nfd`J5gAd6sL?D&}ogVOg7lP9nbZw|9&R#N86&n%n ztAyLh^)|}WeKbj`mrqrkc)tUTOaY{9B{x!=ZMSj4I9|)ax_nPL?%{a5`gzr=F95d6 zNjjOC!H-v-h)Xk{nei6ep}E8rVEu?SiwycFj5(}El@xb2-o5e4b;zG58v;8C>bY+i ztDrf-+w^)jU4`#E95kDnJrf#CLoZZF6YVB(ueAzJt}YoS@jsx$nQr1bY-2uJ;B?r_ zI;PQ9Y^iZw7=Gs6HBGqNqyUc=@F~3`SxaCa8CYXto*uf!qeEBzg zKr|L+;!>oNMTIhY`Az4^MS8f$a=BQr8w54I8CtGDFsg70=(T$FgvUny(lv^}C zTMyC&_8aHv8-m8LO!f6=Rj+UyB-Qu}UTp4eMXbK`JS3ccPgc#K>jce0!3Ra=-7d zYBfF!u@E;*$oD~+ObaWEFDsXk**X_b2F*fRvZ8Z^SL7tVy#13`%K1kplfCAJ+*?+k z2G%gOl?Ba?V5U@V17*M`5W z7z{#%W)pR2iQDLlX*@ z9ZRLSmH0MSt2c5cG`7h(1+*blN%ge@#rwJdwMu?PMEso4S6x<$WO6w(BWd=;BTM|( z(&7p!m04FwAo#v+1HQ$@D;=YhlQ zTZ*Z9a8{4JAw^x%<`)!t=}KE30RK~gO@u9O|#c=ekA>zqRM?85yhDzl#VjkHHnP-4OE46bRRN zp+e5*{K*#DcVu$G+0XV=)ogo4By_I~J`0u#zK73hlFII1Q>f+WZ7N)CN>2m90ykCT z%J(ZlaoO9GM5FSEr0S84kk5Vo-}5PJqH$?%avaK|cwQ+|T0-~0hb;;#Iy5LmAsktr zj2~@a(ivFO?O!mGBGfE-f!xQrTjqXP87q`vAC7HC%{gQK|cu>9qBE z^Shy6Q4*2(izb)BNxx^MnMa!L0=rWcvSlBWoCkCE0B6PtSQc$_=V0zn{Z)oD2#s0~ zwF13fEM99fwhLr7UL@KWJ~Rsi)K%SuM303`-smCt?E7qkXZ*C(a5-PnkR9g5M}p5< zB)Kf?{nhS7Cn_}7JMty)P^ypE&8BV4fVXDJ{UAkcn~(3u>Jfp8Y!!SjnCU-O8^!41 zR(FC(8{Hy0JC;mr=I*LAW*0b(wpVyo3&Yl@S=7AfKw|OQ46_X8_4y z;OVU}*BH=+8z#Q#!3x6ultJglW2{VP+~@+C?z>}KO0WI%m90+kT) zX03nf{YeMhkXK>qn%~h{2PmR< zCly^Ng6)|gxDCO~hmWcz7<`W+(V66;@IVhaH#Kj2vNt7k)q)$+%Mk8f=ZXfAs3%r> z64HVRx*iBRX1K7L&y4+-ZjLUZ8saW_Hhx}=l@pNGL=q6Q2)rZjy zt>d}%T=$ewv^bG#q4$%^ZeP(VA9%*64qnvy?LFe$5io$Dw{i3N zW@gz-K;*dx4>g4jNgQEFpIvC zm%I6cp0806+kpFPEe8i6Z;{X%!lqZ5>ZxDgWaPi&!^6YvhByEm3Z{N&Gv`=Ts2wz? zd|b5f0S|U<^jFUKSDYv4Iu0kS(i6*>cq|nwySBG?Ro%{XGe9J^;*hm4i<0!NC?&NW z={5T&lM(;N^o^R0&?@>Q{v&_>(e8Zk8-aA~=#bITiK7`+mu3Ix7MJVm28VY(ew|BW zu6gQ5UVz2bUpvK{W?dHgv=36fG>JgoHU1iTq)_qbkt|E;{@r#w*@d4> zTYmcp^PT#Grstnm#^ddzYjH7@>FuK&01EB@Si30P+S;~sR4r;IsvfZ_wjUr=VP02iumerKp7AAW9$cxx6aun~4@SSZhVnAPMBAqcPs*m4B9N$M(ymj2U7vHTqU$s>=R`#B3^~r*CUBmOyyMDQ;t9;MJsM>r@G2VM zYIy<08Tjz&)a@E3agOGjORkgk;u-}I(HR+!m}7M7!27tBbVn2_g)g+{w7x@T(@@_P zG-%{EQLq%evaOeshEX8d)1N$gM3bIRPXBM`hs&Y6?EL=nvMFUZpO04sIr-^GDp9l+6j{Q zpk-`}AEg#;k>xWs)!N3sr7tI9^O^XPq-U^D{lr%2}1wo970s2JgV*!V#(3bR^#CPZ>`ez704&pde#3xE* zJO?%w(A{2&S=1R~K=e8&Z~jX)rJrv_P3j}60QS@W32TGu{!we$=2sj_A=hFWOVPSV zU2I(Vq%&dqE)2+qx&6agiO`Ft7wHqR*s|+i-DO@nU?NW#jDxq`+5?f3N4Q@9Y1PJo z4dH#PsK_-jD{pM|r+D_TR}?@SkDVUsJe~I6#%S8~1BnltQ@!_~!%D=?4+uAXppIA*k=qjChpZS15M9v?k4 zHk~_r)05RgJR%V+Yoa zwv;QiZylINE&DX3PgH8dNGcRFVmMq=P87eel$VkLQoQ`Y7jBgI^t!XB@t%6-L51}v zJHkXN23wBtdt`zLdaSqjxy3pIH|I z0b{D(PDmPeiIuAQX3=cW-3+EHyiD0bOm@`$;rfeSVJ4%M+uPSHe3z{oG+{Q@xb|#PGsOUuFJDMM z^_?dHFD#*y+jcg5jtkfat5!dFWbD(hx~S0z-M+9{k%ewmc&gQ3bZES2qJ#MHrs`W` zV*U+vNOK$|BhC#q6C}?3+s^`pHoP6@>&_yr{ z83*y+a%NO`mXQ^sbC|=ZQa7bxj2z>UElk>cXNb;D+)6FE%H<;1b8aPT;L5DqSN-Z@ zlqq;h7$K4idgs*big8zJ2CWVN9dSf_xuLar3Sk6QZu8}XDjlvdcpnTtxF;D6HZWbn zsrqVm*c5qR#|RJ`LbxxP!r~lsi}3d)@jx3_3%I0n`f_^Q03j*^;n!+ZgaRDMo{S%U z;fHwtj-@3TlQYOb#FR{IjKnbMA+OcNx_!UCUstVvYgp!bhH)29Dr$<`i&m>-ns!^% z)ry6H_Jj7hLHDCmtGI(zBO>wN)oTV0OZJ(^3UmL(L0j_C9IRSAi+)Yhe?&?TxQ`HJ ztVKuAkDZ=a-P~Be9fKz@WpY!2$~c4iGB}27E}8pn0}x%n^Js?vu$5lz=|Z*d-cm5g z92~M@{21&25T`rI)Vp?<;?i>pD^8iHFzcub6Z2NCpQs2CZ^hX$GZGiptmnD|Y_3IZ zG#i%$i5dG+OYeUh;EZRSeA)Z{m)-uN1(oruI&LyHX(&+LFh${>-0ULfuPA+&-wDf_ z>bp&1C>9XFBPOuYiceC=HttA09Y$;yZej$EHs591o|=?3XSVNj|IW1a5qGW>SQIx* zb&~(R=frW_F;acao+vu}oxGR!6?st@qJ|vOJNh>EcUY_;N2$jXNI2DHj9kTNSPL-$ zib>~U+|0C?muPW&!o3e?53c_$lh1Q|Xxgdg>Zv_&RuyCXe5*&8Z(z=z z4!f;oE_i7a(~)#I0a)5aG}I4kcNQ|Q^3DiY2c@LNg*kWNGOd7?pu&uxN#5+BO8T+?T*+iBJa3c1!k*#c^P0GjiBUf&j&x3*e zTm6b1xj#@~)|Nx)9Ka7rvi)NuOPmmz{(WGj%y;jk6Xzmap7FCME6|A6Ob(IH2-Z3adf9mK!?tD~An@cLAIJeC&yjcc5B`S$ zA{O%a9}zq0j%S8h7^q;LdBIh8`!ax|;oM14AGhGmq0tF#N)uL2zzFSI?3q&1#D|s{ z;Wz9m%LJrWiYo58+az=AtC=@6KP3_RY-3=aUA7MSLK$5r$1iUu@f`t!30Vn)`{{U| zojsY!q-FWiDl;kARv9eG=x%46Dd|olUG%|_w#A@9`K7YyzaO7Y2@qt_D>bB$)!*{= zwZ!w&GM+&njmSwW->2I(=VhEqW_n}S0(kmW$LNu?mtMppvcB@g`X%&89+JQKX0|Q8 z)AclC7P9-H-K|%gHhJ$nSRhqKKZ}rF!P?eR%bSMA8Q}cbmr;jhUa7v%SSNiG=HufR z6a-qdRedifnsiz66S9G^*EiwbX-oKB&ud(`p)U}faz~_RwC9@k5pTGBooUQyq)D(f zWmpy~BaN&b|KjUv!iBicw5>8qM!PHz#>=|P>~iC84yaaa{)56_LHh}%b{w1Wthbd_48TqJRY@<{E8OizZqPKfTX5-UajlGAs5_Hc|D^FqNGEB z#ApmIiuLWoh3oqyZKMEix2iXHCYd_Mb?&G1Vo$StNw-R><9BVQgKIw+B-IxHtj zJ|FV|ho1oHYY|+hoS*qp8IlYPDCTJGkUkwqzQ=U^_FA`5jajsFvP|N2P|^gH>L|rT z-AgDJNpcdlt)0=U`qKzNh-4&dQq-Y$I0z59Hmpgs$7k}ee#ma+L`Zt4YT(4~sVscZ z#^JNfsPP*{)B<2-OwrhXgVI->-`^NF$FjR)jZ8Y`2h7&0t6ULa@0Y||RtanYdQPjs z?pB}{>*4uhJ=neOxrEKRIhN*iFSU7p=lh*A>Y#z3WVf1^T#vBOl;Wde9m+vYY~TdX z`I*GL>#cLSjBDC=@vxsAF!h}E?Q@BnH(s3oY)h+e0poW2DR!?}D|kW4QXt1zE6F}h z?Uc4md=MmiC>`8V9s(kEBU0?;)yuKqRVbnZia%NMZmZANs2Sh!^SH^Uu5<}SSIVdQ zH4*9^08o|3AR|#@JkTYY} z%qMAjc9c8nI(jz1BXT*G;bUtD210%KP<4gvSA5tkN%I;EO2XuA(N+JlaVXZe=|So6 zA(vd4Xlh-Wttah9ymgE>E{$mkAga7c0gJ1p_SARV%C)2aBOg4aMzNF>q1|D9Bo&D_ z7}7=ZOE}W>(lX42^{ucM@9Cvu0;hPTfjW?hy>i6jTo+RNyL7qVi|@jp32}(0vU}0V zFpblXLJGMB81MC%i#J63dWC)3{$$%3RGYX|GX}-HEy(+6NyeaLBaVp1Zsvu5_9rE8 z2_9P6>dvv9T!owH@EhJuNMH9j2Oq`euM8PAM63Om*r;ZxSyq=UDDsm>6g?7ZRg7x1XE}UkZ;;D=Fd7v3QXv+N%s`;>;OWwEGB= zD}IlbN*E27@e~sp#j&u7AW^hc#IeZ2WeJZbQN&>hCsCwN@HDHsuCKh8lsh(BJNZ55 zlhAXw`_LiVIX>?9dbee^KB&3zc`D_Bb5~)yo@4jN+KXDty62k9JsI>e-MruKp%vttO>l@}9aq)(=VhwnY==daTHQ z6_G9FZF`~MX1N8Nofi|gKa7~bxvY2rl{Z9!1!l1`pI~72Ve8z&e*FtdWF?O}6e1E{ z0exTN@M^gcXr*v|!jLc$n9|_JAbiJ`>}X5&OO->-mh5?n0>lJfoo%Z-8oCOt%yYqa zfT?UcA+S1BI3|A#a&VZyeJCb#M!1pUzmY>_DSlApRzW=@FJW@1GS%ao=j81qDwfu=>d)-u#o?IYlE<*e!g4HrzZNHH3mxA?SlV@?ot^ER<7 ztYn}G4AX%9NB%$RjQO^Glei)=gJgRvLX+E?5ZQ>FkkwRWzAV zt6l!6WG;oBmXgtkWqLe?9&5aZ@3kgGqK!LFbXu9y#hGkHzNP51sE7ac$cs1hs`#E2 z6rO3L-d=YO0{hN_l&Segc6ukK6>e&U4Ts@vZ&*eApE*J62>UoDv%g!*#*zKn#(@sZ z6-k9vvxC&}>)Y5TDPzJd=D=P!w9P(x?O5>vxXAb1bP6*dn6n*+UYYe23PD^b?N*wMw+VY9jcN~nNP53hTg6LrV4^)E}5ccRE4e&6HR2B)S+#OCLUu~ zPThPsa%_M4sr&sx+5C8q`j89}@;SCJg9Wc=MS5{dJl{W17X(^V=@)L*|Z4Ph)2IcT; zvEJ6C2#bkmQT~jWG64n3A6n5t6RkryUW`|jtSu2-e+WJ`(NfA{j{Ow-X3i0mluQxF zpJp}*K2$%(F?e$&B-XY^XI?`ZsWX#$$j-n>z7eBAE0D*RF$$lSeFOgt#EAv+3= z_HN3Gp(A{ru$6pmnSkL0pNum=%qkR6J+hRyFx%X%xq~VLS_J}KW zZ)Qo>p=8gCRGVl8dr|1!I6Q1ztxdb+CN;u+V)izww6jQadi&CoXadd%Nw-{=OOw=| z#l55qjNu~CGF6Lzp)6LA!^;ZFtj;vyUtk#`+%sccymrcZAZ1e6F=LgCtz`R{BP$O=T0P6;n({=){cgO&>iMsXUTi>PV{XmQ zFz_giq6pAqmDEh>>U&>0hXH{q4P$ev!@23?bYDI+#v{BUUAWond^RrL*~Ckqy_Va7 zE>JoM_7tWj9I^ZKbW}5cx{>_H6Y5F_d|;6C3f$dF88=fe9~|Z@rj@^Rs+2(x)*n+B zjNF~G2_*eeR+9fUJ(iCJUyK=kXwFZ*CU5Iygq!=n4N7ae@t`c_K|Te#$iw(Eh1icq z$zFj|pB{~Di~?1;#!Qj?GX7Mft1+ItCIqXj2e-@yo`dPckq~(x7*A(x!-Kd+Su#Hl z+o&3utjKekLz8FwQFoG?`1Y56n$IAY7Xg7m%0wW)De|Qq9Uv^@A@Df(UP!NwPv)n2 z|M&bL+JAzG!4s>gcy^xTkdZ>3miZ?!CoFiEPyb>W-MpdM>#7>3T*$tmv+fV7q`h+= zU&;IQ0(0xM`^D@z-ITMRnLFvTgnmHucd_N`hwpR=k3X5*lAQ;7txFO~4kd-cnj(S2 zt5Neb$4qP~{g%$TAf&_RXk^1#W{3EO8S5}dTTS`w5~S5Jfb6enfw3ER%N*XxY+$Uu z^OH;9P8+EGE!ec^#wzuAk(Xs>H{pF{%v!OZ6yj{7`ZfW5yP})^^&|)5v1yLJ_zR5Y z00T-HRSIpY^GcqVb*-Ekd|~S?t!IYvMFjmq7nas!?Wpe}2B*u@_GlLJ95cb=-(50N z?m(i%JHRn5XD3+kV0Hdf7&rz? zFFvH-K`PvlT3A%OqG5XEDXEOolixh&Cf5BMkU}w-TAlc~o=5@sU%=|Wtmz+}=#fm! zQE5RMNW5Rs#9P)t6Ztj`uqx1w_Ixs1^F0AfpiL|S^ihKiW@_nZns$`*6HQoC^l`D% zo6I)jtKFcN6d zx(wHO1t1=}H_D$nW}=13RbypE+O?kM`PfWaLHhG$z(7yGaHsC9SKdQb15Z3uDg)cO`Li^IF*RB;=G3pF?TpRbN_54mGUTH+(W|%6 zt=}WhTY}P&B@V2aar7>DQ(9A$hkum3@gl4K$yaqvXr9Ltitv$sxnn5h%AkHGw?hl+eFr2hvCMyi%vARu!zQoHA!6PCtl-5;#@4G`a4db01@}X+ntO4g5lsz2 z?$Xm7b^G_5TG+!}@!Zxw*#6xoI`8=uwZl;2=dxE2WD|Qja!Ay(Km)?uywdci)L=cl zSUUyI7_tw+JAN{b%V=Y6@~32zPGAehzFgO0CBXwoA=`3yqJ{#{m$EnZ+U8UKu$A)S zz)EW@qOIqP;duzpORo^bivXiS_sN{KF9}bVy`NWdG&5|b9p-}V5H)YYlI5WyK_(}D zo4YYj$s%q3xQ{X`eP{n{j{|Krft97dDn2Mmh54&ZS?F_@_TRX#ZP;^03*bk-35fy) zdq=6aIM6!q;KD|%IVMeq!FoC2I%X+Apd1!iw&WfdZdBPq*FBMx1t;7SMwzW;Xv5#0=yAG{> zbL%qaJb6JML`$}aN!#N)y2k*E=T7I=Tm;g9z9y`IY0amL=h!th8XLHR1&qoqqpYbw zkB*#^xVS2nUK##2PMnHkp!_=Q!nAb87Y+cclwvJaOo~>0_LqcLC(V0U`EUP$-?u)~ z==JdwJZG~8fFW`Lh||r+1v-0etf9zrS5pWnT_+lpKItRmi%$jR-3e={d3Z9F^nx@xp8mVw??=OhwIknf`1#s>O zF2pr)k)02r@R%Bmz=mTy)@4mzl^q9k5|pp%YNP##GcFj> zU+u5P$Kmy)#|$=N(4LNdA=^U}K5^$VxL?sT=HKr(0i3NdKMnH`vwd4>0frk)3)WV| z*Yi{yJ>G+fBlgwoA&;I{`?KUmo(`EdK6$HRljx~#ve``fTQ*73{FUoa8>vhf*zI~1 z%rV(q6F^Z6H;Jez6`)q1Fe^dE3g+F zfiN{AFVs)oBJZ`Vfni0Yqs_{0;1@p1_JHluf!EoKZoV;f)}BB|#1sk?1`yg^XaBY_ z%0m;EhcwQOwO>i!^WwM@MqYo!OQh6Hv53;_Z*$9EM2`pW0>`aSLC00ovfjZKh8t?y zuecV>$Tuj3@f^FrG>#uwE4rgulh;R{E^^xlD!0paEN3g!VvXwY8mN= zow|0~xN}mB(!~yT5t01wg()zwU9RS$5MWHJs7R}crc7xCZ%kVKpaf)A10AOxe*O2Y zuTp+&$@nkeQ-(zjc%oYpx-b?`83c4WpCKdS*>0-?p@7l zt52OPL<(LVNc0FLa&f1tRwZ2o1wY4+%?R*D(Ey~hyd zl~YGN;r(MQo^QIfCvlSF`us^@dEmWKRP0L9AYlY8zKm$htFdRR$42?*pX8EvH07;y zbjM0y2Dc(!Ou}s&9{Ke_;n&l)CkluO9B1E^0+WKR*Vyf4C-M<0u04m9%Wg1DE81AV zVD!S)K7Y?>w#=v0tP`$;(OFp~4<3#@Rj-Q~dRXx4qFG6GWEP4ULx*_**)J=sjXkz4 z>LlT%Qau)J*lQI)(H$4MtRKc5CMKh;J(6L|G~1%U*N@o9Qi)RbVD{MgZ;1_oe)iR7 z#Bg0`C=bfWutwYdpD@r`~0wtM@>4$e2=Rn_Xu-7JE@$is; ze%car=M^Pj4mvZ{CVg+?I{whM#?!gJA#w41GWTzHI)r*WcN&q_bvA4>5!RXHACcYx zH5S06^#O?jlVN(&XW=?opNbpSTjB1-U-TLxfo9agb4z;wWLTLo8TF6D|Cw=jpUFTI ziWkp8^SVz>>Y4E9e3%J;^DY95`Vk`9Fo1+PM$SM9>f}C}iC4G+47tp^PMl|};)erWP4jPW&BK;nJ_X%H#x1q@{<<7miuL+$$hOEc=etV=V zx$syVK1A_!C?E8F9*v=%1MVI_t9X*r`8lBpk9K>|npfx<71fn2063!sKMu94k&DVc z*Z+7)#+FA~65*I)nG0EEADAfZ%&~3c@MXx7nP{z9UVfvs!q}SnKM{J&%M~aDv#&N{sjd4I?aocG;s=4=X4tePSAlb)U*4nBa?VnM zR4>jx^=(iabA6>C*X|~K+^`!iv$5IR=5W6;4*%SLgrYzP@)-NE=Pks{D`@)w2NCeM ze}!VyJf}7WwAjCcL%@U#0|<`T!Yqz<0Si^c|Iu`oVNre28mCjbLAr-->F$>9?vhTW zyIUHDaOeR+sRlY_U7A#`l+ifR zqvR#rDQ2;DaCcD~JZ><|L-r*~6UiY_lKnELCn9QVq+;|asy}8NE>W%J%2?7yHV&u7 zHV@2q)>!uf?ykaR1RzR5R;8wCWlwHj2%>`TK+)&4qOvpSwsG4%bD@%0_W0kUk6?IJ z7+%zJgE3#@_iiN5o%BN#9nYY&eDHmS5%Z$*@4&2^G;8=^fyshpMgEvN63sMfUSIx- z2uY&Xq&64HHCOq|-8J}XBc`EhF=zfJANHk81 z6e$15Tg%|Oc9Y<8k+9rJspC7p8p01kw zP*?6&U~-miD~5CTf9TA44g*){Byom(=BR8v zsu9Rbb3Nmm6n)d$>A|Do-GRWK#{Yj$e<+g!-kVHFq|K!=iz{(U7P;@|GJ;ILCgElBl*8a+i=5< zQ}B4;o2=@J`G23bJO`&nZH+a|h z+J28+ezvs8TN74lXKJVOR6=~m9+Ry~Q~K?*fP-|RY?ErT9Z!0b{pnVP>F>X4jZdPV zlfyPedDg+Fby%M5k<8rx!1u)MD#|24Yw;Kgep#>^_bmEsyoxqXfitF?QV?klz`%1< zuBEgSxu*Pd=&#Qr5B+teXsVijKk98O0T4qdFB6;3E326F%ccM6U#aOtmtrD3)ns?0 zmqNTv>5sV*wD*{$$M?R_x8go5-wxS_ReMMDM;jHdaZQ*R6bL>+DFCq^Wy})c7csUG z!* zcz)=c0kV97x+Z9BR7d#PU-2^jA&*o=W?-Vkt7HLbMB(2&DR4sxv%^cUN;kT;K`BdQ#MG4t;9r@|6E>5Uio2rf}m zaWSw}krL!R7Xv2EWte4QuJlmln5w8yDBzc)OILN|X1`+41ey+MDQY%jC*XW*XxbG` znbhQ^VPh;{1?a>g4NCAk>P=&8&Btwp=|lxQ_p-x0?Jw7RXuJ=l)@~#Q=ue98e2FBi zmMg(OVG>ov`C|*8{$7Z|+!L2z0!jwoFJm)f zv@q@*pqZ7JA~35&kH>uNqkAXi?rYKX?(rOo{YrvfFWK5_NFK}3~;l@vsTU- zeVt9O4p~3dH7YKFkRFmHu@lxDSLJ}a@z6e4b{bqReGACHNLz2{3Ma1P_Om%|`|s-V zuY(y(mtRlxT^)F51g5sXUefAEGFeCHuKq+i@1Eu_ zyG+K`{E5(Qvdo?20>n*{?d_^P%c!?w8${UhDR;0PjkLc={u?amqXRmFQbV3>Y|AoUQeiiSoqi`lHA0iu(EDW7uTN0iIed z^X%*tXTOPSYR^ig1J!EciGo3$C*3}SjV{giYZgk0#u$L^ zLR<at9 z#12BxOb`~SitH_aJkjLjid1t_QYi@w(ef58k(t`&{HbL0zv5o8-chZ0rIg&-=m+I$ ztK-hwrGbY=kcXf18>4=Fm0vr7f$H~^usA0Lc9NSLT5`;0BXJInAuF zscQl_)y9J`N85Vik3Uq#;FhC~ruA^{1Mw{vr^pw43Ta3#gXYK5RK*N?%yiY6HVOl2 zZ#dHKpR(RSQR4Jp@Enp?IFQUo+-a*Z^n^_4uhMIK2%x1@QEHvzYSN<+xkM7Sa6b@| zUUM?szU>p;FogIQnPU2u@sKgv(i?=`^Ad}3yg9YE&BHsF zCH%xqksD93!QExc+IO*3-d6RejwTnKpu1uIZXw=xEo)rr_oj|^>jhrIUe+OLtGn%n zXItRJ2F~4wR6YR#mCi=1QR~AO)$Gd4^`?Yd3VY{$*9qGtwpHRg@vw zX`Dzq!poKyb@;@(w}!;B0yvM7R_LzOW1M_7fZ_%J%s029ETR_lOAPCw>roe?+lxQw zlu+{~p2WQkPt8c@6@pg3dsNEKOt*5|% z44B|Dc!E{zBMBY;b?;)KZZaaBQH)N@Nu$VI%JxhPZ+&k-O-iT6T=xy1B7}KnC2UEP zS=z6Ft`7e{yc~ZvRfMb5vZ;E)73n6iLz#Kt&`x^1!K3+&k2P+4#;_fi+v#Ug#VjxVnl|8l^xVmIG~cz^c2exX~& z!~{xlSviC(dGu*dc0+l|<&chy-zzP#WV}4x9!W>6z}hoPt*M;u1VRW=U+^z+tcUo0l`rZh%4k4t z$FXk2t%5Tvf0kqY-UI+eBuRWnd|y!qjs%4b$b&?`HBcb>?9?Iyzs09|(YQq~#gDwd z9(Qnxlga@QHSltbROvuxjQu(`#Dd*-r{}7Za*E)_`%`N(%g~&HhD%tou}j-y=a+qH z^A!4vy)TvQBXJvK{XjXJ9-Ig1Yl>pW9%XVw#rsXG_ekU34ixi32E)wm{N-dkc(NAB zRFZldqDtZo5@#nb$W^QzgRuRO&#fAE9s6(IpQ0brD?WUe|2g@-MF2mVF5aKDAIn}s zU7|FULvwVYv+0l-REk5s#ZzmG{YG2{jvTSygyfRTTJBj;yT12go_HqdJd=T|%ZDq~ z;m$!s#wth8AI%UxQR2PJmWxNk5iXzseO-~66c&KHe2H)+%wGGTxh!#PH64EF!9UBQ z2>shQ>0*Y95q#b}ryl86GqpIGe4X8ZLkh3^nM9?(^>YdOmcS(T{`WQ+t5=7-F_63g z0`dieReDfWXi4pM;0UnDpu_fete-IV4nFa{81&t+1Q3A%6^TcDqZ!!&oEiHw!YUTd zAZ>EPM_7mV$9w}rx60mqLy36gZ%eM(S6`khHw0H2ShCt9JTL&{&=Cu9H`7M0$01Dw z-IU)WN$lL$BBrx`E;EG62g;7&3HC;^bTI5|cEO@&Q#v~l_gei3SVmVn79jUpU% zvAeyLIW^I&+CU2ST^)g~nR^{{+}?iX(Thmog309OZRM!ZJ|4wF47@zvMt@a%VySP%ak)_nN|t zzNRplq}hj6N{1WhntL|@n`!?27Q5c|ehfTzoJX%BP#ecMjW4}8lXy<2#)*-OxwAPY z%Gj2^o6Bhh&DZQ8Q#n}B4-tLk7}i9Ku=aEvGKSuWosP*V2cILz`Tvu=%Y@90juje-*c#_@QkQ{UH4{Y>Oy}}ahy92J*jC_(X zE=lmka{4^J;iuc-mU`G#YL)74nc72L}wEO4HFQ2|Jml4Pan1p7>zx0UDrk? z0;&aZ1fu@SFfHmx?&V7{oiekpm|JVo#iYA4lBElrqwzQ2c3$v#r;c5KUXxU9AL@he|!) zdE+(7c&)@6wsMk+waXI`+=L9}o=MPa^;_pPo`qOA;RhF_$DRY(dcT@RL3orRnbi>x zVUcy&IM&d5E8IEopV5zIOO&#ls+hg15_%!0ev>9InGfTn*|MZ5^C1_qJ*iX7?jaH` z4dtHQC6j&g^x11qHB>I6Fm=J2^bO^51cO-a38)1D&8; z#E-eIpuZ`NQD&9lriq)P|5j>QTf;d<*M$buQ%JP%<*SHOJMf{b@%Y`r-{m$uZqIUn zQlcT@_YoRdGmdu*bBK2XI>*2NCMFSp7w-JaKlswOn($FvLU0$aT3P@p#Fn9}yWMX( zBzMX)3RfR$*QsI+ZRtAg>M>(${NJeSZ?{v(t6LWR_RtO*mM9JuVt^>^0da;Pu241neA}J-jmvQ z^=Me@T)Lvak0)#jOOq>Uk48&#h)?XMwp>o0&K0Nql#{;4%}24tKbM$TUYd*8ABR89 z77}e(eyjTB{m)=a!@A36j&v3^TmU2Sb@efhlX|8Hj+_d4R>uGt58f}hKkQOuj3G4D zalS?=O&LvLV2)u&k;fcGngDIJN5@lE58=`O;0>hlKcr0>D4W$%l%c~*EtWh1NL6BJ z`sJwvVdVHOsvhzKH}o1;vyFsJET>GdzZ1=iPGH~=&v)6qR?+yl%~2=of&zTUrbAp$ zV5->*cQ;b|Cf;oy=pDZBPGFO>j%Q|sKirXm*Oc)#r5TGCSG5>u?3(Up(lx%vJMbZ5 zXvB>Rhcvr+`ON_#U`2&=6th)U-g&Mm?YPIbzrrS<&@S&(zDCzPJFr66vzmO1WUq4g z#o}tp$$iIx(pSQ2h8sGq0+t-N21*2`+9QE1JhA*^Jla}|%?zc4In6-DfQ=hEr2;ne z(J%=Grr_iB-1<_|dXy{S8>7#6hH6Xt5$(gG^96WUqyTwVEws~-tA4@vTn!Ll2dLQ# zMRrQHd8hJ`*i-tzB~ViKZZ6vK7+cpMOEyVhf30ip4sJxcnSVkht^aUFvYM*1nm9rR zNQwQ{njyt0w8HGVNRP_H?uTg&VQMPWZ86cbS#`AXnC zDPHm|ngc5?ljd47nf3nPFe`q6Ux0tsXLB9--fTM8coF+}6=2${Teg$#bGy&Kkqlxa^Hxlow@(`(BPDk$XR^=>*B;Qd41R`qhoH(xBd|WnPVs)ARwV+9YF}u@53*=A z$sboo6L6yY7tAPCDw*ilE}xmUK)16LrJGyDIfYrNo$1R)sBlwMw}qmU95#gBO*r(d z_{Ao{Vg8yVvHigRtAe!)p4={`Pm%*CtKMZ~KC8fT;;>MM^h~G0Y-&-Vi=5}SjU5*js1vS=1=A&r0jV&Je#HGcEuGR2jI3>e}+rzh9JszBpm(nAwj z06Txo4}$evjq;`~{1d;VIy|kW?2kserAQ*lhXf?@X8_J*jOlBRHKXO}5avzaFGA`M zPLh!SPNNg4Tq-B;nT+!aa(QMEwG+33XL`A%to*nQUIZOF7=aZXqx!>SIU>e!%gRD$ zWg;fP(}qqWKC|bcLxlO94m{@ia@*~S@mWi(e{;Qd+My3Y1ha9wawnUBBaSHwU%w( ze9oT9T202Gzj+A&m~1+C{nX`!kZyj*B5Hp-t**1=4>G~x2j(A{YU&{(_(Cqk5381k z^cnw;@y2W=cg!Rwu~-T61_F@A;kesiG#M|GCgYqONbHF9!J6n~Cq{ZHE!RoXjv-e4 z1uFl(o>_GhfcL}D21sD6{V9<(4#uc1=( z0_Xl|znEJvI=A~Ew^}yI!~tNhG;g`tKwWfBWs{<&mY-mN`vvT-?;PqgO(RVF1BLjR zicIJx{%fPG?AbrCb}Mp}8?2P4YkIpYFU6M?uu%ed;?L^Hmd?ZP6W3k1Bai5b#o#f!;U+Ol@JPQri_(6lZfYVsHm>S1 z1EIDhTX&drL7y!sj?J6yZG^p*h1jTa$0gpm@Nc%veSC=2#=>9U~s4tZj2ZJcRsn z1d(W>7T4`czbz$i-1?7?*E)tdK>Wuiv-Ahb?RvUG{MB!IByQ#9S1KbA_>=G@f71CG z$;J8RBbpc%QqG;4TGnykPGdwJ5|a4kVt$Q9ejqD2{X3Tf@LUlhVDt;_kGv45^X^2r zt$v6?kbZpFqhY_L8ed%YJF(ij*UV-A$H&@C6tt_Wv<*Mf_Xsz`$JynATFYM6DmcsY&N(LG{(l&{r~9vOVK=^>e2 zit`r8GHaaYOxymZ{NjnywRcMY2%X`_tufALfj~D(U8BzP@U7dbr<%qk*K^Teaq0-n z*QX;6XS|^_34(dGivfjyKS@ntQV*bt^1bMzpsRlUsD^@PL9ODI^v=Qqc(4M+0vr## z`nUJ&JO>RCJFRa8?4MP77xt3|PfYs;ZJjT>Cua*r;p-n=CV8(VqFIx((8_3*@aIV` zv3>9|idO+ADoZ%oD*ml+D%WAHg;BJ@TD*t^(Bm?#+3siYHy@A2LZiLLri6*RPyMv) z_q~hXBhtt4hb-YMoQ2bWlX@T)DetMqc7e?_kH_XZ-01T9N4kt*2A8R(G8<5Xo96uV zZOTxIo>rtAL-H>g)9tQ=Pj}p87dktyQApiCbQJ5ciB1`D4DzT3+k9YKx#@hlUBxzJ z%>5G~81YviLqZ|$&a=Y=$F#jV1uqU5HB&uzP^`(iMM)T|R7WlpIE*NB+1vZJc^z+W zBCdHh4kH}GcnPN=1&V)2c8mp4*VKMX345PP@#ecBvBTFbX!X$&_lIO7c*P>pZ36`? zoO_`jDZTZ+7B!>kZ>)3`((b=~zPfx*#_jYZXGhA_ju*oGVhC$stPXXzqnX^r94GZx zA=2jm-r!HjpSk()zGd4`#ec(7m!&h`Cp3A3>?;VCwkHKQ`&bEAFJ?B5MwEY=UWJ~D7P(hT%haF_VU|aBm+R+2eoa>szUYgB4+xbj zabQn}9<^ngtD~$-3(3R|hAH~P+=*^q`lr;*sA6%n-O0CS(Z9qq)v0yEt`_ApLqr}_ z$LD%%b|4O_Q85kkWIFG1Gn8?)8VZreWc5<&Lp-k40~7gGFi$0H(!`<<(fs1kqGPYh zJ}@HrQEHqGhqA(bBda$`IfdJ!7D~7xFpp<03K}Rla`U03x0FLvarx8WhnfLR=EaE2 z=8kRgJT!y%|GeEJIrF6=h+6d5cBzsC#ZwRUxO2U($y}aQRRV3TDbsXCqJBQwgWOY$`L@a&fnJwXnr+ z>9X3t-mT?xTH4Nozb>vA^iZnVZ&{8B=kmPFmkkpTK@phLQ>gFIG%$aSuR~>fX>Iv= zKspx8ly6}ah3`4cmz*hADd7bz+k)YqwX|Fng!#p)5Oy0Y=9+3=j`WO4T0^cRjZy+7 zY?8_jnKY?K;p1y?#X1K3DDK(Vl?HaF2HaG+zWQTqOOV6WqAFI<)UdbqDZ(T?oA(uH zN9fc`CZ(F$lo0@4Yx;qxsqH`#tY5W zXkTvHs{TrvWDtK@Mw+$`qnzZgyj~ez$?B0reH3FiEi$@r;(t=$x8dYZ5zuZn3=u~H z>7k!>sC6$7UUdrv9ui!?O^RZXD1Uz*@GIy}w!dHA?=buYhY;&PqWpofCk2c0x9LGn zv?)!pWsVqz_Rn3@ah{k`I`yZF`=^rWGquz{J`Lt8Cia>r37eb%%vpMkI?HFZ*8OSN zJsU`w-TSVSMnEvvGo}wRRvU$=J}6Aj;SbvUD2LP*ZMjiet+QYX^=&835+97 zF)KWDhu-7t;i5v@+sf{j?dbz>`R1}4ea|}<=|m>rYW;iBz~1r9Zh>}vCMfbYPKBw~ zCS$eMb2TbXg*L~M3;Gz0?%>T;)uqlkH;6~fdX1TJ)tjNy1Cxq18O({%hjtAc^s2Gu zUbyq&(5~Y#J}*5V-#s@Qd5zJ*+8y9&BAIT9<02(QY;V}iKP%MQtN0qG`J$SrZJoql zOCJ$eLy4qR_>PNW8dIEMNWPc3kr*Lw32ju%sdLZyCo8;-+&YAL^9dL$Nq+w*^3Zd}D~}{gujNEmAuOsZ zH)vwt=iD)gqOdVc4JeM_?oaq;20B5MtT)JWrN5YY-l7i%fmyC)#m`CT{-8lr*i?5v zzByJd-Y$pL1aIH$@~97XT4+kgCWlnAOD**|E#suIOHHvTxSpV1e|j588NV6EWr{lD z_+xldG*pfBgXJVonvBzs;#?7T_#J2y=Ma}`KQ)v4EDn1TCbJ@zCu&Uf;~wd3N6D8q z(uf$%Jgt(ytTmabnh#HPI@J>Gy|5l+Otj(Gn0{9qjpDypu|mLd9miNgITuayaI~&N zbysvGvaZtUYBc-F*id37$1*LPZsC>VnJpwkExOSN8&sKG_Dtoaf{}J2$b?)tN*~Ht zfCoOvNrHZcP9`?aS886Sm;n$qeYEGY8OsZ?Cxoyb3AGvrPQz#NOdG`D5}dQvA!RBT zcdcVupJBDbT;(%DB`1{er+Ju zwBqm{?A+bb%iLMWH^}#{j3`G^;aq}2WG?y_bNEwB;piFlG!ZST4<X8;I6@VpD&erN2zl ztFEQ#g#3dfLp8U!s=hA0i#A*u`Y&{j%dkwmIO% zyHAVJ$>B0BJ|Jk41po_xhc_>$CY+K%pqze9!jENI=}Ep!omM)Vd~dZJ%qg_hU?+;a zv@~$+RTVtoSaU=b)kDAd9Yxr~Kx#Uav2{Y@vV~-Ni#+^C^JFCJ&V;gd(<$CIO-dcm zr^uV1wFbw39d!2@_&)Jz_(wO&2ttO1JdWr6k*BbHif)D!s>krPS>FzWh7dPY`(q&a zJzKOD=Rvhp5Nt8jJtp`v(+;NuRrqq!HAgP>f;&RMcngIeKCM)e!0cwgg%&_- zVT&eI1^)z%Twp;1(Nr5@v^$_QGJM70xo11$a35GfzXu`kmKbwICXruKi#e^7m_@2 zv|J%v?F?(Zpvf-0$@dRt4l{I2-(IR5erZBVu_2Z_f@vR3!@;NJ2NNmqR)udxaE`yO z$E?8jrIlTu(X2?rA*5p z;JmI-7Ii;VN%p*-e&ceje*McM_F{YX@hjjN%# z`lPJ3M+`yu865-S4pCfwY2qO=?Z_VSLG6tE_@#d|?J@E5`SYzfikXu{(3yd8d8^Pm zH``v#Z79t-tp!8f!Xmt^sZ3^rW5x&u1R;2Eu|;GCtqCn-m}Z`jc@1^qjaLRcplR^N zUbWB~THj%wdtd?)2#Gr&owyBsb(TUgqJh300>x@32JWGm+<%@CyHUxH41syZ{m0QB z-4>YQXAM!-)e?Wb&X9vA38E-3v*Xm=iWZ*X+)3o26zK{Ur1q+zyCD7KgkZ&B^d{gM z4oJs%1aA<$FC0c(h0Rzk^tuqR{N_b4BA4m?+1+9MiT(3-#Q~E~{Ey))15o>bk{mo- zzP{?+{K;CfUqX`!V9)~|6;F%7n4wDVU83t&e}-Oxq*A0MPKhN}L0<2;+rDG=h;V+s zUHgKuP;YhedoanJ(f+|?)4UHhG9w7XF# zjo@{z!hl&opu$!-q?|?rT7!pO3rCPC%)cwlQ(wDEX84kHyt5QRS8eVfO%viw%0wxT z3|v3VZP6KOLoF&r&pjBE=tVIN%qJ?2ke~KjkqI4Rv&s7=!2;|b1E@x_!Xa3Y>mmOb zCBU7qzPgjY{0EhOc5cn4;Z?Zyh+yyHD>_wjFw1M~93OmmB516oXgk;y%}H@>GG!A+9m<4$p#kTpr|=x=)MQb7mG2=e zBmLvhx?KacO@01{=o}$nv_4zXq(&*Y)s7^5D^KWRhcr-Ucx;*_ZZxJ(lwjzamlG#i z>CJWrZY~iF%=F;cvGF!WVP-2K0f(9EESL2k+Ph?f z7WC$1$S8Q3ji7Jz;=6w(kJ7LeuciOO3=|1mrM4>IKc##}@gK@RA&A*ZP-SP?R(0s` zYHFm|pB>zj^`g=XojX=s?)}Oj0wefGeLIN*bDpe#LqccM94!rV?=rhN)){Zc00E>m zK`kTWbVX;k6Ldw#(fj`8a{!tixkj3fqV21OYn=h!-raZg^7`LGW+)i~exdnDO1Jbb z&pBE|rWJ{kgxBH0M_#)fEF&};6c&u47=<2-x!_Fwat(NYnC6NPC~1PsPiK-^V_9aS zLe=05n(8^3N}*YlZV6zQ(hb&JX*VM1t28sYaHAmoToN0tm;A{iRV`#rM!+x#k{j(Q z?yD8hdUFegRvSAXEY7^7gG?aJlEuyV(D!lKF4r`GtzGhve(C!rf5$;Sh7Z$(nekh# zT1Lzj5@>@~tR!EveFI;mG6b@s%Q>diI-}_hqpKyMZAS-w ztG0JqfWxz`{kBmdO_L(W1Ar-QtrZF$2&|g41#v6swUzKP+zsgjtRvgs<5>KnuIc=) za@11*@^D3G7D~_9yBboRg0m+)WWM;yDENP*Z7Mtfjl_lOXl_)C`KC?7MG!#MeoI1R z%VK`Q2O1Hk#c2zISKSDPJim?nmrqs}9*N(2e8;q2_FvC`_6wb2*XZ#@;kMalgAaU3 z^b~{}s6{+e3wbKcU)=@LUNb;#%o(fQio5i}5CODdg$Bke_8^IDcacC8^k_xncdPpl zj28si(n{68eR^41{AwHkCwP1QRyps9aCWX*qMSiY6%?rO2LV#{xTNg?S#mA*_~nN< zOZqxX;aPKPn@oO@Hbn-Xld3V>2*^lF05kRvD$8*j22B;HpaV5isH@qGvpZ$?I@psH z>R4vio8B*LD|{o8Xe0V^J-^kmD;nXQp~m)!j4 z`DF7>U2VZ1s-gB)0gHzAw}0JB{QP&MF;UIA^q0la-Crmkq(FcyjC*TpZIecJR`Cx1dB#w&em79HxyY>yt zP>EaRY<*LTdA(sz>~2S8VannD1)<{KAm%P=!(c{zcAFfMOHz;7fd##44(52n! zm#@c1r^HB03BI8xNMkM}DSrSar7r2gsm~)8&&BPx7i7rO1LHwtKY7O_r|S62jF%_c z3v9n3ac>XrpqqZNPh}@+@F#y4O$IRV%T`OCX*J63SuPV6+IlG36_W-rgdxTjAHL)K z*>kK8Zp=O<0H^wSMnB446y+W=j~seZ?lxM?H#y_D3Q8DL^9f3%OqCSu;n1g?V6a`8 zV4Y1!+N-X>jIz@dRY)xBBN(O^NoO;`){^W5P1aHSr}XSeEJpb$QvP-QQz%Ift({ij zE+^FLG2Wh1%>Q!xzLg>dDRhi1gu?mCthYB58FlaNMf|UTp0C+HNbqOg=m* z?Rymc_j|nLFl>ePKSuJ~N8+G-$_#8(uWs~&3jT?vF|xBd5rXr@6QreBqI10aRu$T} zO{$n%NJw+cuSTFcBC}u!f?LcBOrVzGntxmXwBz$n`vQ!+$LFz@Ft~*fB~RWGkMWM{ z7C=+kun{SPgCIQr@%#*%O*Mnh*p4a6|JNcum4hd@W_LY(!_ z)J;08Ix8*2E1yvQ@g&VPkZH8TfIvpr^dRcdoEIosF~KK-CX-?jHMFMV;6d9>^OXqJ zyDc?&hhjWP^B`z5*8%}dqjA{;qgm~ch*Sn{g3C0n*`72$27==yaU!=qWxV!<8s$F- zz~uy<<-v3C9H)NO1POgbrTZR;U)9=l>-N{VN}dU`Phd3a?otXB{Itpr-iyo`0i8Ih z44x7^$tF3ATSIZfu7B_7PqSQ8t#dbDJI=jf6Pa_G$v)Pn(mLxV*kbjvXRkM;H3^%@ z{RdL8j2rBG4ZzS2KR#4ps-M=@O5cqj+h@wq$=ElJt&iw zvH|SKc8@|&?ss&QicWXtHW+k!K8|;i3=UtvI3EwJIiRrCM*tD5AmikzR>r}q@A*60 z#aNKp9Cv3(0Z}PA5hH-TA7WDJnP;8NML^VIs`8FcoMtlvR@iuXvXFn*r5cTX;+WH?EP?VZ80zS*-<$LwmTa6^bvcTSSJf#$P z#bfmp3M6%?qbO%-dySmpys=;9ob1|7tEa)YjgtrBZ$Ah%_y4KM*b$C>uxPCr43$c2 zKmX-1BlN($@a22=d?SOp2VsqTg~)e|03s%lIp*6xnmJlV>1SGmvvS$+BgV{bcp!TH z+aUQYwzmlFNEsmNZp=d5gunoz1Zv~TK`)ijED!(-_uYWj4Kc!Y|C5T68%X)56(3Qc-VM1(<$iEH#P5{bcy2Dt$n&QJEzi&?6(ot5?y-5VoAEXD7c7 ztKbt(Pabl8=o`2AJ~DxSOREQ+Ld}Y$-tc2}=aBdxrR{=E6F<;q=U%-Cfg5*`MCE?y z)d!76+z-B@w(>6zF-HvFFeF6Kv@u7}u`@%I#(To-2_Sxgk%y!)yzOhT{Pzik`xG@v zp-M3DSg;z3U&+DL#oK!TQF%OM>#jx3G${*sqpieuU9ds*8QBnzJ;XVPh%xQ4kZL;A zm!N%QSC6(#|sSj`Us+4)JKbTkO@G%9Fa&li+hczmr-hos&Ck?Hn z2pmf+?$?>6kaDvQQaI)C~)o&%WU(VP~{9~ZUrAG=bCL>^E zrt|Ocf-u@(H>qZLZO1S^32j{i#s&ruXyRkVc%~$(Oy;USz4{db>uIugkAWoDpMW_M zkSBIs=2!{Z2J=Q6!O| zbJ|^9e3=YN*Fa0sLz69@GbuXdjU@WvQwaTgvdI9;uOhDyU>R$xruehJ8`|aM5dPOG z?N75lQ55O%>DK|Lh}w0bW4tn%38-wr?*VX`%4(ocAM$0Q9^M`a%r~uw zDP7jZ3`PcmDlAM)YFpg3>cU4B#K1w!t0v%PJ(vpI^5mWl*=~|XfYuu01R}Mms{p2z zrh<5Dnyysv7FWAs8A;^t5NLDVA2o9dxU0lD0}enH?&Nt>aWCZzS^aqh-ed2P!LdxQ zS8J|EiVEJ-U?&%?}*RVa@vB%-j?Zo}#7-ShFD zk_5YS7c>1WeUoBWT7Ska(zgft2(<`1OOL7i$&V;PDnq#x)nG z)9B@qlUAcdz(0aEPm@3hpoEF_(#^2ixbw;z^s^>GcTKZ-Xe8zq;PD zZWx{SoZN&f7(#ysD&#?p@Jz#WPTc9=*RU5$=>WsAe`PE$RF~Qi*^s9N@d^dPDqt!t zqtl*@sY}+;mS%wg4n6m^qJ~&n5^(I$`nAJ>pal{OIDwZHi!fk|;%$0l00)y><&NyZ zv~x1rAY@=eULLpfIkW{Ms<;&(9t72Xk(&|rl#ql+wivZgrS7-Vy!c;^$pz{}3F?mw zEjEy};rLR|kl?sn29;4^>q3Db&FrQFEn9~Mnojx4?<${SM6SpxmK4f!k>NTOlI{LO zL=h_$f|3B>vK>XbCjQd>n*hh^rtuY6LC$LI~dV3$vhSj{<*__HdE+5F`kgyZUG^gz7z zr~eG`Qw(0pKn6*+yOzmJe!p&{2W)U)^>zs_^0poIc&3^5 zr}T&$G1GX35~l&`$mb9VUBi0?DK+>s#{GVK69M2eOKz80E9I+t9)EhOc9mmQ=Js@j zzh4_qt?8*e0k9<)ffh(=qziYYU+LyY?+F)GnUUHV&^Y}0nxCizj2 zV@t7+elay(J&z9Pw?eG#Mq8JVL~>W7#Tr~nl1Ue`n z-#1}glTG4it>{3@^qQf3ceNd)deo87HZ2ZMQNA31zc@o6Sqifeo&P2p_1l^^cp+rrS3y|zMj zm@_5Mg+JJ+_P}1fVTw1Jy9PYuQ=@M^(*7?nQuF>jvqC_|IBgkJ)2|HK>UT-Orzbi%+??{x{RP8dTTNe?Pl*7Mna zJoZ|PxkPk$EF4P|M?NS^OX3t15_{N8EE8Qd_oMGZVOS=l=&9^rOp87r%%XkO0Pgb- z{@bOLSS2dTS=v^Veu47cDd}7f~Hcd02&RtwPX2Rembnsn#zpz7o%%r^zp z#oX9}(;_FVe>pefnU)Y{a znh3v2rA5DPAg5-6Po?sz#svGAwd+EbL}(S8_dlj7vY!7kk|+HFgl`|;9Ueh$hTmlf z429Rg+3aPD${7D(E#DC}u`^c>PFC?x@g78&dOSa6$;DLT5-#;IHDE#^M<+9d3cK?H&F;@qbG2RYHtQJ zb8gmAXYprpX+{ki%KoI8!l<`@ox_@ix#01bYz0Cc?2|+2+NBa_!bdsUqQ`{KL5%VF zeEM@VMrEcWBnCM;#~KbWdpRyCT4u*bN&yB-JdIPE*A^aLZ%jn8;L0 z+Fe88KU&ij{0|4W1{WX^yGISx?_4b3MF?p=*UEHY z_yA^?u@|VE@&NsW(^O=y3yVo)2ONq$A_^ny(I5#A=rLjM{HTpmk-)mexc0k@9l0ck zuO2?9p(*KOb&BqBW2Qh7L<-=PbN}eFqdW$j!ef|RV0Ae(NKzifR|Vh5k%K^M8mq9| zW+>|$@Q;l(kGs?R)&q4LIe&qb7{mYYiC9yuS?VaE5fbGl(37l*JPiv_x&l;`IdZ+O z?|DW&QEPCHFYgGGmrMc(&4ta^Eyr}uIU10pK+Lk%7bFnwc4ssLQ71HBgOJK~J%$37 zHg44O9b+!*CCXe^pm)_V1I=vMBhjxN94dctkin_=Bt32pJ_zO8nu4x1) zvT}qa_XrmFV(y1ju2@e>%c!q!OuIKHc4npalOmDR?!z%&NAT)z39VQ0XDP3=2SC-l4nuXzJ3KDAZFz-kC%2_`_Jd|C6$v_t*Xg^5>ik49!+{D#>cJ`{* zcxt8_`i@NhL()|TMA5xnK^p09NnwelOS-$e8(z9uQluMcNnxoaq`N~9kd~B??pPWL z`NrS($L`Fqv+V4>_o;Kv&cpW}dFG9V8ivFq;@8nMwjnmKb=>&p_Xn)?m37!wL@ze4 zp^AlF8Z*>SAx)&QKo?uN!Z}m=p5C==g}%F6P;APx~3|`V4x3+p<|6?DTa<+ zhD!`ULWrs-b#|47F6<~S;1iDq*HLF*W+~EqEdN&i4r)RZb?XG+@`}Rgzut*nkC&F) z{Kco%3-z?|Hv?Gw*S_Ae`|Z6O9ijV^`y5h7bb^N3xT1;RFnz?!G9og6EZNNbcCsGK z4s{zYjp`9ywuS?a=O8Bk3fzlg=rZn_``YU!yF7+910J;oB=&Z0`5evkD${2T8Wf`! z{0B6dgtva({Z_BgGgaRk4=Vix%JsQ$^jiPXoG9kCuAK>H?o0AGHe<9Xw{X`=|50e? z+kjvCXQ^p}uVNXF3uFr;^Tg<8=`<=cd8t5IqYs5@UbfKDh1l7*sxZ#zB$66QLhu^y z%_~3(gZ-}B*|Z5r?f$&#{!UKsg!GRqiMLQpOesK7uiz?^@+t-1q(UcwKf-E20MdmV z28!3{q}q#ruirV43OLvoE=X-SXQ_pZ{VC{bMiD)w7bwN37(fF!D4SP*h^Y^~2xSe| zX0`QA5Z0TmDmlaWwi(DaQdsU%)Q(01Wv1?~znZU9+Y*oFASz}m6j+;YscDIv^A^(h z7$A)WGkp4=RUc~nWjHWp$&LEUrctBC<`I35lIHr=Uy_@szn8dG=&<9GV>)N~FIZU5 z<(K94Ou5rZ(=*$5owKy`4J%}a2~+}Ywa(M>_f30bqv?0QE?K(FUm2^kw{*SO+IyqT z6^Kc~d&(lLzB{*bxa6FN-ToE=kEo!Y-oxNi8lgA&((+O#K~ zic$ll;I`TCF_}~OXzWbuL3W#!L+KR%v^oE&Dx`AIi=I^v{q2rbzY#}xW&od8;feCp zJOQZgC4+o_>Rk+*M>$$qgXgK#JPty!zPe}hsnvv6Y6Vk=(bgpOuZf2#W$ePuG;H`5 zX|?h{NRDaL%0Biz!smq@=}+Us0O&DZI+X60nI5+mOOe#wpR>=Py%qqRL8-Zrz(VC_ zAKP8`x?9s<%#l!5Md{^lwSGjHR>X~MDcbRmlfdy6V6c4!vXQoX)}pZ80NiW#s~AK3 zM{0Z3qK`#E2q zctMJ<0xPwGE@-&8}hgx-|kqG1)g*W8{;-c7n%d#|)5wJW3nf%|H2FLzRk8 zm@V32?$-ToIu=6W`%;9skK2N^4u`Y_en4Le&r&YrCMb~bM$y4o zPAo>7$pNRHBq?uvMGCNN#%|hEX1KFjAeoXdL;U)OZuHw-Rs($}jVv_+Xeu_z%5z&% zRS0>j|0{Vdo>Xdu^li3)7rTY65h`dxf54^}MdxoRBxZdcPrpk)5(SD-*nsUE$^6Qw zw;`#XcQ?o2zfBaK0C#8I*$K;w5Okd`ZwX#xovzj9vHhMMxkC<(!x2B9wWNhFh@cvO z?;WcLg6b+%j|~w-&?(=me|!Vj?(s#GiT9Gnbb>!rNoo*4MU|w+M$;QOzXsULEFOS# z^fq-c^I>yazHYR09QCaSFZY2$21{l0H;EYLMj)Q znU@1vd+OA=-7Dwl?2lt?@OqOY;&YpEN!*yX0YA8eV(;qX6o#LV?~OTY5VluGm!jT| z=I4njxQ+&+f;#&I{kV^C8QRm*AiZ37C<$JTu9OA@sV6AP9KUT@=LCMR+6i@#eq`O{ zZAAr;A5Tm8->|QABf=W!w-`X#(%EGYJ~lEHf1I#04|Zp(;F-0Xyhyj`Deu~1kyrm( zU(VzSVlEIqV2Kg+U)*@g8zxjY?Rj^uUf)v=-KzM0w=?06YL*z-YIW!j{*N&A4SeK2 zx#LXw8KBo82;6s9%dSJdp8ysujzN2}yrKA8ZD^-&dXzrGdgNlo)BQsOao zVeuu?_oFCbr>`=MqeALB`4zAjt*%(VPtqpTvlMdwt?76eD_op3u zzQy^V3C%nW>A4Hr>JDFD*-+w8{HWr8NB!{TQ$L}`S#`L`tFc9cbeZlCuUjQQkTd_L zN67ZvH7%KtjP3w{`?7JlhIww(Mtlt+J}B(r0@R$V3l#3JK?3~CGOE`pD>?ueDpAVQ zz_|FK{uu0OznditFg@SdX7`TsRUpd`#Z@y)8GgLbv30PAYc-3?TMYUUuWx`lxyw~f z|Mbdxm&(5E(}*FZOCAF~Gr<&Yb*dGiqMDS=1kXc3vf&W^NS1_C!Pn3rHwB_HtS z&JPf#-=6tYtHt1FaHT%u0&t`de&cZ5+joF0`kka2!mha0TSI0* zgzPH8(0ZUq6IMRwM|Fp1+TDrW~fVL=c1xtO%IKfqlktFa!ZU$MVd5GdCAQr~c z?eg}jqfP)eD@!n$r_mM&wKwtnDm?l*APQeX?bV!^q}&!VyLw`%L7bOT1Fud7m{ zQ(iB~?+fF1_kMTGJRzPE$%D4yLaj~-dm+0=+ z#7Sr4tM>isv5M|nF`@D+5q0XYx6DY9g&N+tYcucA7M@cp^Sox+(elq~xNP3s6b%dd z+)hT5OVf^=atMb*i{CRNNgfp9emmgAkZgL3h9OCt@Rk{DU32qT+kBg&f3#zxe=4S0 zi_`x1D&Wgvo>R?Y^?OQqum1J+%ML@Y{&+9CZxY#o)9rM!z1>u5jPHy<)+dqMr(afB zn2(A+X=55i_-oh+y3oMjTyIMf$7~wd^ zRo0!g+**IQEk2GF2dNZDCCWPGT__X*m?X5{{7c{!2N7i?ky=9pO&`xX^{4WWQl+%a z$M9Q;;83DdIZNUR-HKH<9caE(f9bbDZ~MA`dxuO}H%3_!pAOU;{^X43C_hV^ zk#>M^-~{P0k#!HZ#80xq{32zx%nCr>;`>;u%#0Tsvct&3WrC` zx$oJJI5GIBPn|^4e&^k-p^1e#AUu#~e2-L;c5hC5-$Lx>pIPIl+8<}gd�%lKY%h zN6tO|`4?mJSIc(v<|K!ho$n7SbKARJ>OZg90q}wFI5mHT_OA%n04uqY`O4b{bl6N{ zc}xp1NQ?xl--?CZ@_lc#J{xP$cO1RHHNG?o$iiUyozpmQ1X{%bQ)cUXUf}Y=(5Lhu zm6`mqlwdl!eFUv+W}PhppAKF_SQ8SB*d#&cmaz1IGNYRwu*T6nlo1Y6?0Iy^Aa5He&cqtP={)DRw7g3(zbj!8z*m(++Fs*%CHqUDYq4>ONZe zIl=l#qTA^j(M_`339Usrv(PF6hr-EqJoDO zgG9EmvWbLyFuJ#dw+G~j3hIO{*cy~&45OlOTsAjZc-iy@E=eH+rFRZT=NntE9Cd8T zXV|Xej#V(G^dSQ{Ad$^z_?tWR1AW7ZD7rE07-qKQkMEwy&wsf!k90lX*hNtN37q0Vha87-n=J1Pj7*(LpXlY_SUW$OD1SZWa8_L`j4ll>} zuCV+5hO^%F4gF3515LLV<7!mxl4oveXbrgH)kJJtj6U-ryFK61{R)LEjpmZnv?ZU3 zj&uiXm@3y<@t3GM7&9eEMS!~pe?pZ3L}S`Gwkf%R&IDR)VcCFr&` zS@89AA3@CyZoV7M<}p&hkPlOYEe&LV{$9Y$MqiSz#e9$R!xzK);I520=MN@tTtI@4 zB}-FV-Q>&4|CZ<4Z^In)+5w)Vy!!H}%?j}C95ZORMU7)c<2)zq^1xJE1N*BNBs(@> z5kI}XSpvURr`=h7J?|Y+NGIj1{-WQyNd!b}{G&|M7aT-+&n@8Z)s~-Soq{R9{`Xln zD`(6mC1C(8`zpC`0YS0h<9rwWdTAJDotZsO8#80c3lg$J~)v~RBvLZ z{}o1#o|H1=1uNoCcEEC^YyfZV-tDXfqhW;`vf(=me43C*L~l)Sz;xVsB%HyUi(|NQ z@qQz~?e?-R;vv1u0ab-v&=?<{Pkr?(1isWL&4{dJIni0TJyKe^ePsDT9b?9+6yvXn zk_KbIV$Sj3uSVmBpiK$Nt$iQFnCq<2-5{UqY$*(b2+YS%Nji|yY>r|iD{49?Ul^e{ zvsxT>v%{zj?!K0IMaiFQNn*(?v1ZIMEBa?)sP$nu4)DMST)Zt9V-L;>VxcM3mtB<;#j+6s(N)m?PzA; zYNN2hx&aML@Boi5=1@kvMHzJzz9Gm^jfI=fA>;)G-IFBg=2M zZvdep@3or2$Jez>k4pu3ik4kY=%6^i3JbEjW|_k8$Nn32=3-`37v#L~h{a8^?lW0$ zOu(fIVzXAW4E%(@QIC(20nD6(o}1~&R0dPSnz*Z-F!89TU(2kJ^$ce>{2GaTT?NHe zV$hjrJ=3DKlG$^NDVjbUbU!Lx zHYt4nBH68sT>he9w#MlTj8sB>T6z9}wzS5APu~km&#uwkZp$#mI=-7)-QMc!M2 zmRH4Ob6(s?hob}G0`XaYnzhDDzOB5elRgO33__0VEFE3}M-(|H56fVKP6?O6eyn!h zQnF+M-Xw4PJ4lX9wK>@h z!|p6hee!?beI!5TObnmOtvchC-21n*$LJr^udH`iHO^h#5d4bPfzHIt27XB&K~S1}*Aa6R z5XDxxdi52;7~7`O z8k!BJdWqaN*4Y}Y(<&-M*0Zy%8=m_wjq9%{V_GU z^G6Zrx2YmaNruT+yn)PLhrKf@hAGD_Ojgt`R?J}~VGaBw53PK*HEGFZne|^L3mcI6 z7A7!Z&Q{NFfC+2XU{AC)1wnyP2N~b{NyYv7h0$}G(*4vVVr6vuR1x&di(vwL`maJJ zK2GNW13VL~P;tG*XzcgvLf@CbL}BXwz~8Opa`~4E7+*+wp44pjp;9B(cSKSKx*n`n zrviSkdPdU|C|VB=d1GJ{+*9BYH^w`6UwzL8^@sVHe?Uy%cE^;5iT)Bj;Rj9} zWYO5$Q~;TH@EQBvpft`>`cK&CFl~R3OXDIE2_>qjiJ%c>okJpdah}9*yy+QvB&``Z ze|WXawc3Cu2DEvRh7;;(5Y)&d6lk#iK|c}OY?#uj@+Y}{zpUw3nHzI~*58?3*mxl+ zfTz5yE1NEvts`@TUm}=Ea%^HJyb!y7p|wwgFZ@J+2uJ39zXU6qOF)L9K3l>rh6c=J zo;9}Xt->5|9og7k$LQ3+N97IFiARWUPP6XvgkJawwl@W6@Z1Z|(VbzF(fpkYBzrWBUI zuS>qme~zyzfsu>bczj%QjL3$L+MMu^z-NblP|huVu*dzcCU8+ODuYloM(uqGT>uyB z5H6o@0p3J4N>C(8#BKXx!uhBm|__+AlW8PZ}&SjyF3k#dZ|KP z4<-M3jP?aww7ILh!NQwIiw5Jgafc9e2`bqr+r2dSBXS zk>m6ah-ml20vvJG+s&B)?IyKzZg!iqH7ZF%`(Ap(UcL6}HG~VM`0--p7f~HEiQuPr z1riN1=+r;c5y0&(t0T|)fiv^Vd`#XONc^=m<~rYpy(`u0o_*9g!aZ6$EqC}Lm5AX& zFC0PsE+j{Wrn^@31K0Vn;M_<#%?$g2@x(s0Bud<*Jb1mkeNvn*=3iX;9+Pl$!{z0^ z2i`BeI;+_j?9FU!Fn843=2F>~bD`m|`M~!CDkEQfaolv~>fq`L-6eqex%B^yGi|z3 z579sK#y$ef;lX@VcGJ(Zm$K8v#MamGBz=7TZhonXdtSjezOL%LYDS-1e_h|YZcJyp z+3=iOCqt`7rj*h;h}W!$CI#pS+<|&i!y7j~!Um}2U$jo zPe+fmS{J6-A!+x>4l9)^mn$4ju|m_Vz^5{CMBLmFqw8r+5m)`nj3rr;Up`Z=bNC1-wzZ*Sm04afx^u>1qf)a7rXi!1qUHUfBGjXXpbhZvxE`I&YwhWMatY%>ko(yU2IdpZm>S zjwMa^fH&GwbuCt}VvZ*PVGh~;MNjP0FcILD0Fy`})SKbHiX)YdRMKgq(%G44GwF?O zh5BAXdA@PiD`J1`AWHe?=A(8EM3LCJ8|TcRl>M z^Ss~&3OMjnxCt@jMuo^DnRB>%8%qXN zuJOEIzM`JN@_KH}f1+g|2sdYqN+IED5i6Dh#_*u0x9)$>_kztL0|eOHkXfr;IsL&-_L3*a^vMP>aL;E&_@GN+x%}J?hSf8i&et8eO0Q&YsfIGc4v3j+baGsy_FcV7=wbaQU|!N&@3d0 z3o?ANwAsh+{uyjA+z}X~^q~kp|6R8PBa4n!e}Ag3@zffV-Vd0GPku3~4`Vy=4F{19 z+V(gMC+nw%HLA{B=Bcd3!M_K4}r%X{4wbmCD)a17$fYP zPuipe#6v&NhaobS6z8vIf%SiMllZ9vC1$-;_#zR22Nq#946DUhIHgV{iv$jX4C-9L zxaNO=>Q~26>@STiW3zb_fe5qVSmPTKc4bcnT_DH%9?wwxQ%5K$0z;r>ZqB`HTlnXi zyLnZX+O6d838eh5I32MDUiC0Be9vB)IX@GgHz5Le(!A(ldY@v86V_#C&zLA6%OCCJ z*52Zk_bRY_X*hMITX+RETg*OX%CC2w%|2k-O3Kypc$boiFqAE)HLAsyQO;YbFF#n8 z?J%B`rILuy;Y^8hhestLzS%LF?+PsofJ~^`2Q|;P){)qP59S*w2F-;kpf8AWgTo9Gu;^If^A{2s%NWYV%!dV!Q7BTEN$r;5RI}T9C^q>>ooV9FF(w)0vibW9Rp}X~6owXD` zSmX^?aE=KY_Y53%kXD{`0$HyDkms2OfC(ax<6Y?rI`lGP_ebiswAgf^bok1=qtHmg zA-su1a$QZDqDx-#<@qYe^N44o3PYv?nQ|U&0>HSPJQ#NUXG2qPyX818#zv$sePdbC zM&2hfe`4+fHB^Ax!aP36n92S1?}=h>dV=cf2p?YS>eog1_MgLpK^6#qPv zmO$OI%J(K|tuF+HydVn$+;>&O&YUIr^~@3LhRWggi-hB~R#-wsR2Ny6J;wbjdlN`i zJN6?LiQX^}O|3B{Q_%;SD+Mw1l#Zy{qECYzJeJvq>B9UPeS|vCh>y{s>UL1L`2t{W zWl}uP11zZ?Ry9SWgBkS&Z#0G|G`5E;tXzmd7ZV_vDFAyy!m*6}slBNU0d~!ast-`U z$aKGOf4N#XL|jm+Hz4xE6^SMedr+7s%BC_`Ivx0%P@?_76J&C&?2eP!90<11n~5#Z z#1i4FBS}75GV=uXQb?aj+DOr!##pdgeol56(5ZR)7-oXnin^{hL*0}E-e)c>mD6$ zr7dx0DZ>T|%7k7=ipqOYV+4NO6J^R`8P`! z>92SHmGC4brF_wGB?QeVD^86bBhSItT#qj8@)ng$ACvcvp+QsZiIpse#zCHq$=Bbi3pSxf7DS6Ai`U#u(<9;oQ(*2S_63 z!Cx=4M|X#>Bwkfz1b;nVo;zmYy;D?8MeC~QlbLM&6EB~g%zg61!3(B|f1^#7qN22M zEr!n6lq@$HT8*QCnbJE+fpr+h*POg)5D(?Rw=@iXh zgYS3tEZwAn{EU@n(Sw5H6tch6kOZQ5)E-BPe2S~;37YjutBELnVL5+9m`gIRy~YsD z`T_sVp=l^TA~(9v8aSm&w-z=)UEYqm`}Sw?t=Lo0&{SHDb5+b@;V^~B`||dAggFD} z61trIOc;x@XFSsr<`^{A(!^W2WclL5R9?2V@M^djruqPd2OcAd?VC9h(D=_2%dlTR z;F`56iOy;52k)e=T{mxN7Ywz&0u?*`J)6Y(#smgg-b696Cpc8=w0Tpl42Mf3iT1Jdurj zdMW&7Bt4h={+ub>-s4LS5E=g}7d%C~3^YcEHQN}_DG5wb@J zRBVd^u&c$x?@P9?Y2VQ2WC0*-U4F+@+I_p1xL_~xOF!m5Ve=^|j{QFXwxMM**>&TB zm^Ux#6|3Eh;v}|kxA&rg15O4H%)`kQKZTC$9S?~{`;LP{2{o~4j9Nl)hXp|w7)P^we-rihl0B^R4i#0rJ0=2zn&BI>Y%Qcn- zJVQ@#bJ8U7+-IiX+Dm*DiA1p=u(qZ;)6_Z89c^C&C>K!~fws+bn!sKIKSo#VJjdJ; zj<$>oGpanFy@maEt?gIfc4m2JzH_3zKL}>0kHBo)>Fq%oO7~2w8%vCUh=hC(`|w}w zawDsM{-6;mq|901Q*(S?q$<1W8AI&o#&N_X9a`ZZ z*6dh)%Mvu#+q76cd>N4N?@;-Ee-3i+VFz=FVa|ZLCsiliE+C}CD+Yj`_>lFL^twA5 z@PRUj*=bwqKny#pa2WLVc=6l%Z~*+no^{c3j`rfH6|u-|FP8f(hh^ zlwqqY26#w8mmE7^GWKOoLULYfEZsSXMP#N7Eq7nWR_~O`3tE;@1@lku^-VCTHM_N zL-n)bYlCY!NM1hI*Zvs#5c{N^@ig2KyY!BRIQoSR&8l8q*#qLEVpTuoT)$KcJdRZs z)px-{BkfD?W5iAIqQ~N2o(3IP|f2bsHM3g&G0!OFXdq`(FuuFuh*9+aY5|<e;nGNF4B{nL!zJ zW2+Is`_F~Ht9phWE4t{C4rF@$`pHPA+554zJI>rf@VkL#Qv~T3rC((cnm&~z*6I`g`cVRoa z=zh>P+J%97$P+Kq2`vZ;(effy-Yvdmzuzg|bLgKuJ64-E3Ek>yP_S7c77_^u88B5l zr^e%(Og!hb|3U>!KyaKtJKi(0c+acaN3!IWI%tkZKSOm7YO)Dy4{zp(nOl!c4j8g* zsraPM3|&22kXD*--T%;PS=m%6(CHhL~$rzn1ypy*Vie){d(vto<hyj+i4TyEmCzE;XRdh}Si_0j0rHubmQT6Av-Y@|bB zJrX9P^YBy*6*WFgQ_5e9ZZq0(T92A!{6DltO@}3?k3{M-p3QMdfvwC+Edu)d6(aKU z1~Vn8gl5o3DHbR)Omo3jh60%J1ub+ipT5>ai|Dv|VN`C<-Q{|c1QwB@GUICf$f|w| z!T5La?9Rs3LT*i%jQy2XJFppw*M+_((IR8%k_j}b#y8!4V=ao=eUvbze|xjHMipkX zHf#pk6SNFL;#+lon;N%kx25gRpIkU>){@{r;n>dQ4ZX;G9(8&m6;7d9P7W4UUmHhV zMwx?6BGOeDXwOZPRgP~s^LR(`NZPK@FQCsU$p5D>bd*s2YIz21{cJaAyG*_WDW`wkfx#RliCqJhLiHEA#us%hiQ}T*!Qv3`!46flt5WqJ9qj zm!S1wpx!z=>a;BVBk-cT6ir$HME~w%yaxSuTbUJ&6%W90d?@Ncks;48$##V>b{J<6J{KoO*A?vy;?mHhW<%&QCzym~9U<~4g2bmiKIOB$T zt$LcjqXQNR{}tdw>b+$RD*F^i-5S>79id~4ogRvf(w@363Cg#$aRP7nx(A zJ&)m+cyC^Akh)N@!a-dKQzIAXhv!m>1!D5N=8#yX<=h|kk#C(etXr$SHzXO4fiamL6KZ3?0#zq0ul{>uRkL{QqnA`#-AeFy*>c6 z9o3&@TZ7GI?OX=g0?@a(G?NN}Y!z6L;`Qz*J(62M9A{NjCCs;28C2o;w{*HbFx{?m zc6~r8A8Lubqk?hyaR{McLb`DHM9yVzl4?TDp`M6*T_&D$0VB%vCjh6;O*x}{o5kB{ zg9l!2)qNrz8edY)s!= zWfcU^l~RZmf=%k<3zV>4;H&#=yiW`maP+wcw<3QVf^Td1S>@`}f5=M)r>*l9ku(BTTdJ>>~&un_mz@0Xc$+3EdYWSP_2tdW^Ok^G!GR z1iB^geSF_md%sq1Q|acC9y^i}OvW>luuJ&%s#%907=4g}!T3%Eo65C!3;5FeJDfK+ zHC`l2Rw!kJ%mW_st3sqX|$p32fH+P|$1oBvso|WxUNilw{}h&yB+3@O@=10MPv= zPnrSQDvi2peLw%(S7u_&(pTbNMM<%Lo>p%-$QiF@Bo_lzeo6l2F#)lzFy{J1DMlib z``G<`%3MC1mdJ8fqY(flw+nqdQf|@0>H{lOuSkL??UjKcezJDklE7y z9aa)?AR#~I>6p1An56oqeCth|e5GN1(-j(LUBIL<_ybvoV>yyNObnoG@F~eOyN$wF z*p7X7IH<9UhRN~DEj*k2y4C< zZbuebkB%}e1kktic)tdc`SQ+K(#z5Em_#=OW$T@H`!k7o))`SvKXcnNGQ4Q0Q42RG z9#52qt=`Z=1fI$RA>BZ-!T9oJ{DqHeqUVUcEbNw8s><3J;wR3iK69|d1a*M{f}i)u zPDb!13)w}ae{iMPyCwO^pr>&6dZlY|gQKG~Ky}cD`a`%XI6Fjtw5M`=hPo z^s&PKkJdk_ikAWH>rDcPTr&yoS~R)6r5nc*#r5^iOT=sG#DW_-fN(%vpoghQVS(`{ zGC^UO;-XWu35LP{kxKXi@&oNUtJ|t$yQ~CX#l~a-#T%emu%b7JZpN{KPB8HJ$?8Lw z-QuVXp<^Dqm_!pWbF^*GJaHj1_VLOv<;BiSi=h*WU_m(F0fa4@#6o*KNE=#!RGma zq(IO$LsgUd=g;!w1(40;gqn8_T{kNj4K4`t^OajvkrOgbg`vSPuH=;%rHqv@WpUPX zZ&HO#EXXh`2P-7i@qX3udN~@?hBuz3)UxZ12Lc!e-hjk+`xz6WSkK`Vt0q~Xox<~v zl%D@$O2|pxP5;l0HKJ5m5uet4KTufgWC0X7O-Ix+b`_Sys(S4w&qDt41J7a>!kLW& zLJGmnzlwjOcWnM`hbu(B6TzPPT!7jp9B~Q$;dpCV@#0=%&mH?j%ex9CF{_0~0Et_6 z6P~K{sLTZ8(!Om&{Az{|1p##pJVMu;K26l*^xrE;Zu@os^0U83?;r!=0by|JBPk&R zBi1P2K<_=^M)*p;he>^zdqsX9d9E|cVu)a7zzD0iYsH(8 z984iHxS+h*PPOy=lZv}&@^`-TA&H^e`nt;rgX4n>7xT=^Vo;ae*bW(4{vz3-OA*Q6 zUPM9OkPv$@GalPnCVC$^m%FP=mBuf!yl3E%5d25?5XAKVAlj{B)7;8{RErUiYQ5}Y z`9wlGXVOO#p(wZ28S9(@Ks3s-G~{b5$uO;jKABw{r<)sUn~W~(U@B?S(eY3q5|y7}Rn{8VSz1|C~gZx$di z?3voxwqX0g0_~5Ll7(;V?^oAtVAAvlCYBS41S}dLivdi43975)AUX0X5WkmDv55@% zkTVo-Z}|Ak2>+86P@0q6-7@uheU5Pg;0Y9(_kLs+<|)$B8kUX8qZJ+|JGi3j8rV*4 z*EP-PI)s1|J*w~!S58vOXepdUqxnt1p&e%i2&h%b3gm!PBXWT2Mkt*8Kk&4nEa!Sx z{8hX2lw63Un2Wg-Ag?i3+kK02Y>4Ya8P3pmB&H*sE+A1XClMLc>gHZBU`)5Vb_`61HN4Q zwok#g#!Fw0nH+y0)7(=YGi{AjupK!auu1M04l}~7l?tKcy23KDrCeoSWs+)0wQX=` zd|)0g8$ZA@-GfseHM3+DO5RbGDMzwDI-lY%ms@bKP}MfR@{&Y8%R>mw+LCcZKB6=S znBThlyaLpNm;;_bwxold)ezVB# zj0JwQMlNH{k?1jEzH2s{!e5LRu)w7>?f z%!XWzfb6Wk(kYqTTGSKul5b3Fgba64N;5{1|J-1L^3PRbK>^~pVuRJ7=xE?a>lFyq zEI7ehA8T;F$potzQ~&2c8XS-yF=?0Pz+ZL2#ozY{IY1bd+ESf^)RQM!22cFdjn?4S zEmkZuW{YG8O!-Bw_9(j zsV(hqHy#l16Dc<`>QO*q2)$~Kq?YG|$lf@mk%;Y!#Go?bsp}LPeQ`uhd-b?zJbc&q zDIlr%faR8muV{ zAfwmzYrcg_6~2PZ@p2kiy|Hbj$D5y>|4H6VFl%+&1G1hKagE>Uoq*n;ef)Y3Vi=LY zQEjflwH)a8DE#Z_TzBzL67-4(4H&O@&cv%df1tqdPPWg|Jt1|32+4RO@g;TQ=jYXW z#vQzI1h`E=RCbViDXrs7=-3s?A9lbrhsH2-{|7`6(}YQGmM&d^zeYE`YQzTg{F@FI z`*OKp7(Q-?+30Yju&Y@#OIQsFsqKL#6ay+~Mpg8&d;nM+^77^nq)CQu%4!+>Q%VBS zq+6oU?DycRAUN|6r!4QSX+;7tNE0g%cP)%OI@C-?pir(G1s6*yk)`6*&qyWjb1UK* zYgm%2BY1r0C!_FkbJoln}6%+-8VC0O-o85y&27iSq>C}Pmj9{V=!%cL3 zDY-kOhY|NzW7T@vh9r$fZ~lmU(0eZe3afuoC~OQJ4!PsptPg41(b=Vj+^W@n?pmS=9DUkoObVfMwpBBezYM?~*kxg@7{qhoNQY@l2FvFFL@P0l`rCiIQPU5 zPvBNm17ydW6a~D&-j!n0QK6jv&VMOHX5n~E$7DQs&hK-!dG?|u@tA6*X-@gH(`t!Q zh;B^}0+IQt>iv29hRhs{O?8a!6F>_-BpRQzY;&?xkAGSAO1X{eB)6nGQ5SH5APb z1Oh{&KAarDmTmuil;(ZWQ5x8r;6m>Ql&GQOh3G`TD31m9htPLRIG)C)*NT2pIF*xb zb8%LW1~?1!5z|lfkZWmamu(Tc!gSFevo=Z=B$5NL98br^MAy26!~~)>V@#-$ zVFwxkQo9#7l4aQESbbb1Nm{h9wR%i*%TZ%NGeA*iL7y_x{WGk^HEVt z&8s}ce9usHng->NTpMbhXWv2E;OS$Fb}?ht&PgN{W_+;n8Q)LhD_Qq3Tl*W5Jgtm( zoC^q+UzNxYBelC44)zh~s$iw`dTEAFs#6QwpDBehjIR8}d+IfZ{#;K?H01p!wrq1v zW_!t#J76rRQDbSgj*=nY*LV5W<2SEbc&Th#Jj%X2EyGfL)#pY($&CCmlz9|b!|{Y< zGg2iR@qUpoqV67J<*%|e$vrPfUsLK|ng94=VM~7_v8P>A75I_1kshF#xDtX+9-L_Wzyf+C#KM-QYq0X$rToXGp$E# z^Q)iYo>=H?GIwq}-Dq1RBd*YN$Q+#U4ztXp#pT>oAAR*kxPN#O{1w6+&*sj|t|T`R z{iHu9pT?j0oz^eQJ-jiF-uo9HtU=8*8WfInJeT+iKd;jb<miEgW`tLA^656v!`bu|M`7~WqKGEp41Bk1n2>ovv3L3l0wNjxc44Qx!@l%s4;B_zR~ zq4di-5O{E`g_^|IiYDBBj%k%lzJC2koL82}N>0Wj%jI8iv2(3Ul1%$9S#}N(iIp2& z=kO*k2fMmbW!TjBc6S-9{43wu`*#2jCU7kS-Vd)&){K2yPcB^XJ~3pXKQR6M8DkC1 z!YJ1b8Q5d=L(BX!&84sf{va_E%Q36AYCq0ul$ZNiF7vFGe2Pou zkL0L&^TwHoBjY)U2G^*YaXl?2vjq8qjdJVK{u_5EGF-N=L8Odk)$&3wzH+li<$1cb zGt-zLjVi~B?v#4u4ZdtOpKXIs<3^T0iN(UcQ`M)^_B`;XsdkJ4)7+?dLk*`fFAEaM zU4Pso={5H8jUW53?U1)*k5F%%-wb4 zp%+B&b_ro9(c6U!>E}zVuiY5sLhyO(pS~KSrZnlC@7Bdrlm1yKwql)9QNKhOKEsqWZbCtw zcepwPg-Ws!Dgg@F&QDYvBByU2-neTCHDsj43@O8wFWCJ081>voqJJA~qgX6_vw0_W z*^gv*6$%b_6f%!a`aApAW)p2p2NE{`+-M^C`t4^kjomQLZPUw7pPeh#yI6`04FMB% zh_EBdVW)q!Y&x<}hsKdsgV2j!nLsAxE5bmUC$F&3=rcicxuU%i#U9@Dc@aDbEn@!?0%ySZ-4=WhI%0ZLM*1<*|Od1Q9smnAdZd2T_trr>85s0E2cWq*KzT`xR)LKg{u!PG{eE0QN_Jc@FwE z=H$1aXUE0cyfOKYiDuXqP^2qFke-cd1{-UKJvq%M`_i%h3@|fh0yy8~LqR&1)J0C< zUt6st5|rU%-|_o=pUJcEPZ8ODxH#LJGxZ^BhY*kd2UWXsBL!B3E9|@#(Cs|M)z`!Cy(u7rn|wG6=s%)QNSYuGH0V za3vAx(dS;x_3Zf`!s{N0AYtu3egKLNyBxOWNAF zOk~R+ITwh+w4j{$n;2eKwpc5| z-Q_aW>)|<<_x5PY;}u5@b5ywSooZYIP(C(q?P>DqEkcqk&PpwBojI*77Gnl)=;n*d z%eOnJFUWKvyKTZODDG}K|F&|LI1-@mRP*M=Rmp|Es}R{*k9y|=nb$`RuD6z54tRK( zw$+sRkS?34Q!y*TnZx*s;x_w%l3u zyImR$*HX@3NA^E%|614G;4u!~aI!s7s%!eNA(FY@ac^UKI*Fj%7y$;3I^dtZK_zdK zW*{5>LMZgQzbflW-LyOEN(FZyD+UXUEBWsGoru#k& zS}jUVbEwCXutIXHU~pUKa=%FZudiuR81vso)$fha+Apl#&%!7q^PvT9E)%W&haz;e zmk1_3BDNn0)DYSMbX(fy;t&4SEB(NeMZzV0K!6=pP#H6!zd0ok=SzY!BJH7dUUO?| z2PELN+B-oFAA&y4kWRO56~9H!lJH+@CF-mzysBRh=#s<4>YvR5vt2^g$y|}_*71=-mP>&mP=+2*ZXXgcF1v23;N`SYJh)=dpu=&RRdbK zROh?!Qc1lgP}8qFkdak!9zRjS@4Y12a%pmt;|d3IM)es(L0ff#al|@~;SRZ~_QULQ z-y`}5$&ayk8N_0~GdX4sKCTvIHS!$neYB5JPQ!K1hs*}UvzU~8u|ah*5zNp1ywf_5 z3=8;4BG?!5M3)LoJ<-$KT|Q-FVtjS(U!+Mdj@EX1zE@ct0GggHzlL$UP*_@fPEZ8Z z@vgJRuJb2eZ&G+o2fGWxs>{}%oE286T7JLytmzd?)nNFMIBCs4ytGL=H;V0L{#J|9 z`KWxbZqhHj>b=m-4m0~92;5GU6-K{xSCYd5#xpeHa6P(A{BVZ;tkVQIw>j7>DZ|at zQLnn7h$pf4!#}1OS2}9r} z%V{}k2u%SIm>AHeC6eug6ji4qWLrcW#cTeS)g!MGa7jwL{Mb8~vG+6oO5*2>H9QN< zP!%yebL@u={>yMzh2>meC&OOr@u+w*htYj)y70=#n;hk=XhEZ^bzu0Ahc1HQ zx0{guw)7?e>;|zsfhO$gpTuXi-M{V)KT_rfwK`LP9@w=B%7;7bjtm0Z-TmZD_zNw- z3>}r7p1`J~u|}aHoTQ_9T$~rJWJHLv#-^>Ap)N%N5pp<&wH6*qEtsI>K!pSgg!<^m zH$gA^>z6}K`d%i!cfllMp*2~*^tu+1A2sh;BYuDB{g3k7GtHha{J41SZ)5poj|C}* z9bfVl#uycKXOp;uq?w?u5FKu&OsWGxS?LNSYSnXjB8}CDMNT0yfeRnZrrq9&KDR{Z z|Ctj1nzHe&5oGK!q7Xfn7r=Z(g6tix`!_Mx{1$y7Ai=qsFVjx&Tmbjavrjp@GrhGD z1aCrW2d$V#80tcjsT-?RvdOphXYvS^3a+h~O?~WiI%CjsNlZI30UKxG;;#Nf#B;Gh zJV$pO?N(ZiR{J+cRLa-dkGLlZ;6JD1f=uef-`g4^8ttR-Z$?DDBB<4-**5zTUTKgY8g@8Qyo49{rscauPFU9vdde4J*PNa5x?__0r9Qid zAKv6DDBfp1Vq_NvEM&izW?Le5z!xlY10FYnam4jb%tO zEf+&jL;oTU`KXCQ1}}ZAYfFs{I0_M6hOnx%j$ob%c{@h%)e!p1LVI5F<#RN+6SBS# z?2BJ@G^+WT@Yxtwdyd9OU@uAK$g>v%(axv}8tLJXRL1&82*9vOWt$m1r}FUyYSpSO z6z*z4u^TbwsgaEmtlmlRdd z4qxBN(>4w#-Y;6K1>1?}UNEfw&Zd?4{&F+xj|X!V=2UsDgxQxLh%XHx``bZ{m9Kx_ z6tXVUa!B~@eS{>~wB2_#>v9AJx98r`9Bs!H_RheLVOH^O(??kL4v628prk?ELwOr= zEW$f(Ud9(+%oR88LR~uwYNB|!Km}+U^ zEzv4px@#|{OKZ7i_QHV$rdS?-Ffo5<1Zrt;hSk&U5X~)Pfcow9Y945bAIbUnob;Ox zzAe3H5R6T!z6i&zTA(k0lAIp?_I)Eu31N3_JC1f%J`$HT=PHsAb9hf9O-4>c85)ix5hVGVQ0b`C2A^g z>Lx|IPyJ|ne!%=Ixd&jY>DXaW8?>;#P3E)jyD7|!x2JjHzm!uVh1`;po|4LhE~lf)6qa>+E!^27$|@yxD)Y}z)}Bn9 zPqF^tnE&jK0jt8+nl(0*!LJ53oVR^n?cg%7(r<)zLu||*nX)WTi0dzDI4WQU15e~s zCGLfH2KXkbBRgY}s_XMa$$K~NYem{ZlS*=-hD@=SGuTK`e)5@P_CC328nz_7Ybsb< zlUzBBwEg21=XFi@s)1m6vVT2~ zz%bDbdF=3y^W z_;bokxhQq_cWFJ5+D<|)rYdg0meys@;CJ4_y?;Y^KwY$8^&m_B5;~Nb%i5vwuU9Jq z>{U=QhsKKA0?5=LZjRA2F1{^cQMx6+jrPs8ZeL_8fa$E@2ID!8(=9b=E4N-uwlfW% z=0xrt~;Kjo3F62pVKmN<0)Oa*<3NH5{l$+n=TNI>9R*IqRaz4>u(By#$ zmc7|*1%B}x@3SZh-q|e@wFSMiw6xpG#ACZQV^A@Gw%qK_FkosZ?>ASKWr+ZW_miZ~ zy!%1;c>yc4UUCo8ehP+D191x73{mEUSMdhlL9-5Mr zq2r@i&X+P>+q1{Pg4JT5b;PNB-UfhWX?(;-QB^oqtG7(bLC%2TQgmd`4T_Z{q^=Y{ z&6JDLG;O|CoW})?gk&o#@zzFm+Z8!=2=y;xXgi(C!)OtxBzL0JxxDLa!=pZg1OEod8UaI zkLyE5Lo7xE>5Rw%nOTg$@QNJZEJsl!g}pE=6oKbWHb6ebZ5=CuW$n(^X%5})vyTM~ zX|YB%q-8<~PL(NCSw8hUg=6M$e|_XIDcMgYg(uxDpJYU30UJ7fn@o>mL)B3R5r2!F z7>Cf|a4_M))0N7&QXFtn&kYOUWf)}FpwYK-xKCtkyVm~wi#snFEnj9>k4kCYouP~~ zv_x&4;a%2*hf7jmVRWh-6tJy)*ot5OMLG_hndJA-gV8IQBFsIrv=LQ44L@L14F>?K z@vH5q=A4ZpMy>v)!08Be`pfLK8in`)D;GI+1zPDwjZg(YIPq(CZ|m3E(BFv3oARGg z8xUcao7$f0K-KIN1G8yU+%u^ui&^bPoGy%R{l3q{YY9g9vXD$4mq6F{4Zy@SpK#tuA;b>mnw-XApDm9f~_Gz6MswLteG6+_eD7m7hKG{s0P{>6h z2pGQQGiWGl5HpBWYKkSIi~Kf6JIx=wmUH-qrRXv^Ji+uywfvJbfomX0V8IiQ=h8|i z*LA}5w_)6mULm1nrwz1KRc*07;t(yF-#=3rULmfth8A+(y?>x%*Q798fg(1FZ`V9u z7AvP8fpSiI7t*jp?vU5lUAy)EshO~gTDQgCUkF^r8>i|SbX&F#%Mcg&J*Mhp zCyfP3m-#(L#5QfykNL$7#kv=z%F2dM!XEl~B`*V>;edHKHPd?xKr5c~Jr8b4v&UT* zkZfYxno^3?I*eut9JXn^Gu_@$E*ek2Rv=)MI}X(Cn{;lF3Vl7gEtpeBR{}zgdAMVg zmEt688|{BCf%d$+wm^fwQW>6pM+lax*pC*sfw3Lev)~$D{RHqR;Zxdva|~N-n8v>7 zODXj6?a^De#YhTA3&XE_+gzO{Vw$ZaE^Z;yM^pLOKa$cOk0 zwO_k!T}^#xLSrg{lvYuIK zfCX5Ugp!ccXY5KwtugoHEG@g@Y{Y5h%94yObp~icWSB zW^iPJ3v(JFG%%CGb^r9~DOX-0hRXx#1opV{8VCTJ0x&s}_4AnDNeDdU$kt#jwd@?; zy1erP)p3(!R|{xTd$sC+UD8?E@J&8NGqmsHS5`uLnb~eue;l|v0g8YFdq4E`N9%AFD1A@T;!^ea8P+|qFf(<*GMXh#-taAYwt4#G5k!kdc7WT zKsE=c>zRf_pW|<9g1reX{~& zKKMqX2s(@pyetvlhXNCssbi_NeF)+4ejsXdqp#Gd@r$Zu*rvzO&X4c65=`ys}wNYPIQe3lOtGFD#3&)PO__g z#^iO8a+WsuA|;*QxiXO>Hp_ro=7!yChSXUH?n+&9d6Oid4MsL2aI37LcB@Qm3u1sz zn|>U)KU24wYl>9$mpuj@QMFUy!S%&4)(}q@0uDb=~^Bc3jxmC_T!?Y zZW@i(0IEv`@6|*O`Y6t0Ck{T2#uqkX?5|%yGYCo~_Xr*qYc$S9ex*~@29c(tuicO? z1ww3S&`mK^Y>f0|(cEUlhz@Zcmx=hVaxG0|klxWR{nUnyp)+be*g(`C)LWZO)0W{J zxQo13y^R_qbPHHknj||2Y20}~4x~CTb_X`Ru|EY&9mN5h&pHa|EYf}gmLLD#ly|w0 z=@m3(m0turPWTZSVt{`{j%Gd3e_0i zu%HV>zg#l#k>y~xSOG2cI08g%uK4S$_(@HCS|EU39^Y0>Xv*g+bu)^Vi5ku9%MD~& z?Si*Co&*L_(f<)Tl~Y$&Fq?aDYV0~SQ|N>$8<*eN=mY!7G~#;zru9)cN_RcN&Zas2 zYbp*&6w2R!Sa`>F%b3mJ`+C#?0W5;(UTiS}@an|&t&sjSvDWv$%DFy$MC^0oagpZP%N=8HJdDJ}9+k!%{r-}DMXTRR0#mLXujN3H!la)!* z1pSLiwK-Ly!C`9#5YgDT473If$#_@bKZspN3OyeT!{{NA%y^1cb)mKo3XcsGSBaJY z%z{WVxvC86Fx4Cg-In?4i=+Y91h(V&(Zb}B#D4ddf3YlZ5(3euSdjmmFbPr+!|wU| zUO2=alSUcwC4|VWD;4oBEyxJhlZ;WA-ccnjo%PW(w@Kx82M8c0y2JOnoeIK+6>$$> zILA0%-guy%5^O)hh}Eaba5gtBw3_hsr^y$j{(6YGn{vHkhi^nE*zji+ZrOil6ft_0pV8BL`u!*yxzXYqtHHH-XnFYR7!`Ygmyo@wDFH z<&N)wc5{Cb6?M~BeDTalFWN1WY}ziGo0IXK>o_mB{#|!kQyiUQ%Z&b$9{+RxDSw?G z&sp}&?tqTO{`Fa(OQY;XVpeCUTTSP3=kGW}AG@*o&MbD6zvA;19{r?TDD^M5wtgHv z7#r2vAe%lAAaqZwpvi&ae)nFduSF_C_pjz_yt`K=f8?n*&WT*7hlG~@P{3%|SsOet z-vN;^W~A6a!(+U6jkA_YhKesvMwQwGE+M=h(|Spw$|Y%3*SS}?oiZX58^GbZc(QnR{uyQ%cvtSPgGQ^sFzcVu6WozSZRX9Cy7{}~Uwf4dYA z{lk~|l3L`o_N4#O?^N=sA?@$nd7p5rw2j{Uu?gyaZ%T+>W7%$ch6YOr+*6Pmrb1$v z##P-Y$nZ{vn4Cm0@o0)98X%&ZI)Q*&C$`ioQ%L}(=2L;|`<#{OdRZ)~RqYZ@Q0_H$mV!q_pWJi3(ZPEL>H1EAaban0rLS?280p8` z7|%x^sj59U>NhrocaUcJ2uJJ1_DteS!=ogFx!bdzHUzNo{2-jh>BEzd@*}$FG0Y74 zm!K%TZRx$0{U+XR&Jru^MBGY_1e24$5LycPgXeE#V@;`HZPb69;x3&Q{S(W@3{Yx$ zmCuz#fv!%>J8m`q(PYYytZ^b2HqO-j1%p^(m#~nH7y18AFb=WiL2*@#FcuaL1(IBG zk)_su-((4@zs`~O@_+%!2RM|L*o7D&BryC&R#3Y^`)&%nuKz-Q@9-i7J=L}Sd*J5l zcH4~aY@Y#6kX3Fli(|r#A*Ol?VHdiZAew7J1PsFmsN-iJE&Mp^`z5|73~51s?EPc8 z3GnfwM@a94D*$Y;8GAE}EWphq^6BIkUhttTz6= zI;Ka5vofH)z@4{OJ;;vG163ScCMiwGat82bTUekkMgC$(isrPUEk*s(g4*Q>=i=_G z@8h;`b9a!WXMdobVvc%lz`bHusOC157hX#OjMC~enU5XyIM$sWwtw+}sGkvyYNO$= zGrs%9fLJCAcP51;#4ONr#>Z4CotE)!0a`vn`H+Yr;N8JsuoJ<4`)Lr^iM&QkeKV%Y z0{8;wRT@I2d4es1wra2|r#!7a4vTtPItt4I`TX1>Pq9^w6B!S5ccwT`_|V;)Qe>iI zAp}DhW*#7{!(P@}9u@Ilvth;c43#KgI3@kBIr*VkjD-UBzrmf?fx%6$(-0ib9nyLU z4OiB+42Fw63pz!UKhXBLLF1Oqcn6C%Zb`L1eF2}3xVDDYv* ze_yLZaF%~Q(8@g9!A_NH@*4Tse#a^q;CLN&7Up9 z&b^&(*llsDg`HzNqTL%TEz3DP$ZWH&Oz98a$mSW_q9DNzhIVFJSgEs$ypNX(6gC{G zxMtxYj~Ns0e#2|2QW;Z0r!3}sE`Yg=nrnHhvhZ|976@<$s6Vzs=3S++O-IlD_Xurq zUu@&nfBxVxiJCt8oDUzlQe78d@|zlu()`?e5?<9$(k`J~N-Xn#&7jE|w~xg!>?`C; z!77U%RJ6y^_AC1M+(^CBi`;gars2P;c@^<#_@wwaM25306r6bvWO|?wu$1hFEC6wx z`ybZm=)W}r;-#HrUI!0r1aR7!VtnA@F*|A|ugdSCVP=~hA^`gp^sv0_zy+xIG?U`I zqW7=_Oa_6{pkh~Jq4=UW50AqV;-+!RcU)cwqf;@CV@j#oZR}cS+{&*PX z0GI;1?a~7&YW|~}+e62{K0nNf>AyJ{Jbc;fp`$AIhdHTNcKr#6sKSNOcRWy*jL`7W z!?45A56=axC-7Xp4}G+3J&e8Z-_lk-a7Wwa1N)TKIj*Ba1*1S>J7h2bTu&W*mxtE! zjWNF3YDg*4Q}>NrpOUc}-b?pP4dN~nb;<}WQ`@tp z32yG>$7;RrUkfWc-ZHd`hj*Yoh${<=xHSUaYp?7@HZ>QXjChKm@h5GuIuWdq1c~~Tmeok^!NRAVuo&*QV~6e7HGBVEhSI~xz8TXbR--w zfVLS_-p<%DALu~sjBE$$;*Sk38lxXKd4AYGL%9ilnR!5w$6QZSJwq!l>DfbBD1A&`URh@r{TC!DFdD)+qpH>JhPq4j+lWWwBlO;znAyjRpj+NEOSU6j2YZ8|1p~ zt4XDvTgd&f7_R0M4qro~rVBI@5Ir{~g9Onl&*SRlMl7m81MiA%X>1o@!$n{q25{Ux zPu*B(L6tfFnZVskv|6`IOumJ{BTZsw2%yv0q}(hW_hc<>r+8Apn5pL6K%GvWsao{! zaL-c2Xr4kQ+kyyE7Yyj6A|=jD^goQlStD$hFe6#`Ugfb`2fm`z0D;_^csYOci*q>k zg)?CqbHEa|Aq%B7HfDFp+aogN4ApZHTTKQfUAqATzDp|YH}5JY<+&w7bz&Z&|+wosY*7KHwD7__#JO?#Ef3v!0I5uz8VDqWxF~h z0sWd2K6dm@F(={h6c@6ZH>_G~84byq6|$o1DV?5v8m-;kd{z+iturmg39lwl(T|-n zJ@jklJR}TD)TJm*?ym+(w$bQkExeipnV`aZa%8kN&|)`Xt2hy&h_W?JV6qr4{zEWo zYueB|-1IU%JvEf9FFgTb_L~iGMjM!Nhnwsl+|q>JVdK!4K89H8R!Jh0hOMky9$h(H zQ~!^LQUH;vm03ch)lj)FfEiut5E*?FiS?cj7S0?GOcho<4de9z|?)woAt~N>f(eCdIqf%_$v> zzSj4}R7_QZ_V&vA%Jt)9=$YLr{Uz^lO=bG=MVoNknqF|cyVXaNi#JD7u!XFCmZ0*> zsj5@*?b z-&eo85|it9LK}~W+gb}wG}_!A(M_gG;W_zjJjp;%`jey2Is$Q?deSOQiQR(^Oyk{N z|JU#H?SS-}r2_NcsDkk4VnGBFGAvk&%PG^F20)gTgH;^>5tv79%U9ic&Wfv| z&DsG8C0${hUsgk<)SGn;F~&o`<}sN1b%j`o46l$J`2m#*hPfalZXv&DUVbAQHq<^O%rlFc z(J{3vJ72xBS39TLSvLe^*J>oniol4?zhI7Z^A8^EPpydI5d9XGE6t=$qiiIS71m*s zCg&&-%0K?E6t@CIVj&yPXE^ujc>{ix8M$QS0Rt?;=g!tYx)3K6+ZHf*y@#!l_F#Bi zEui$#3_%3!5S7%hL_&Wo*!mT&3?_(&q^*!^4Y`7;8rIHQz3!ip%O~l8;CTtaJ91}5FB{b?VC?zcEY0rNKpeX9H3?oN5SUz3nCAkwNljtSQy~G5 z8JBE?>y6pq%yEj8GdqHYO?!^)uc13ugd8Tkmisk+uK3R+53dBezfLIx0&cotn78D} zTQnMzpe#bWtvZCkiO$q)eH?6f=6P*+x`&eI0?e;h)FL8%a)2e|OE81@bXg;=5PBoN zGmyersGT(#=gJRd4u5XIrRA&><7yPz!YW9Q1vk&RB>=SM$PNST5hf2(=E7nrI<8=?;Yz1DfQw>%sZ9xLIG`BCdnV>H#b<(ZV3Fo3wB zcMXx>u=v|oS-iA69%%0vu|<>8Xd$OKxl!V{$)cvO7B`E0uM3ZU`C!G}3n>~itbgwI zB|3DCmUp*{7NsxvvOX3vSUJ@F({yjVbys|X*HR+lzUS-C(-o-#tlfHwdceo|C)PB8 z{zqvz3lk(a9{itb`i1ew=9JIm&y2xZa8+7aivZSe51^}L>>1BI9{}>~h4Y51&6htm zb^UcxApm^}qj|gDk9?+xM!j=msS^+CaR$#R3*FyVpGeh55~j zKIhiURGxF`bA8d{YH3jG&IpA`>jU^epJ?`+OCce#CQuaDkMZG(Fnh*4!yi9|)uJ@! zXmt>q0!Hq66v{fv*>#HJxwVd+z=SPq1 z(yrN#cs!aDO;-~JC+0kwzW2FG?~MS_qM3b{HK8EjX;t<) zBy~gro2NhxuHDuB#PwYxC3IWyOyVJc>3c}SR=qhF90*Dt_o&{SQCgq;VE3cC?~zIp zF7Zwh?%xj9i&MJZoS3D<(V4>wpP?@;k1u|*z12z4wT<5XxVZQxc@u5khxqr%5VG&x z8_JKOj6L1ixgLJ}2zIDc40bRHXmNSi5@N4&>)30WSNsFnd%mRmw3jskHIGq$$*bjh zzBq+1aLE!fleiw&Wf~(Ohic3Z!OB^HrSuhrKcqRyeY2CEYes)?$8kAK$ubn8w57I$ zlv?nCy&zCyH==aU8&wuK?gnefU4R8mSBF7Oi8CBtmMJ2=1xaex_|_+_{4hZ|xg)z~ z*v9S8reX!A^l&o8Zra+)%NRRq2OVb4p}H{J zm7vA-k8K}yd*YsQ?sIao<7*5PVB&~d^#Z2_z~&YTKm^E@9DhM->POC%$eOD|RTVc% z7%D0;@i_?O`~m3*eG!#dj!Q)M*s3=iJ&NVhO9c2Q!ln4`=Ggmh2AYtkx_s*|h`tpd+hy~;vcI%+ zzY}=)uHvp^gPd3)K149s^4>HJl#wm|ZN+i>#o@3GCDP2=2l4r!HX1a<`NQ(mr6lyg zQSlCYDvw~KAor8}TKM0=x?i<5xg5D-ls)0D{jL{_Q`N2589NJe>N~OtH=!@int=jv zd0Q^^t8m$U@3ZP$7NuTng!rr$+jaPYHY>s=4!rxuM@aZ5y%l4`+c)B3oViKjO`N*# z+{VWwUOGsPWrSd+16NHsTA}W~@uO^qPjj0it zZaZ;uL0&qxln2t-an(t-6sB&_#De~l@=#>slB(_k$(;H z`>vIE-&kC0hiJ4Rx|OaI#p+bClSIZdKGMl5)T-~F=S|x9b$je(y1Cr4}T_zaH z+YXEVlG&EGt_NN!#4pQ8G2w*aB(V0QupIl;-bCF;}NBac*=)``CaPp$zK`WFfYyf_qeKUc+=7Rl^lZ^3#6~#3%3Nj zGTKri%_TI&$jz$a8@1qkT)(YOh>BR|=v5<8%~mIjxdbDG_08@YV$(GxD4l+TIq$`m z_<~U@$z$jVVnS;V6#72!)Pq1N+dGX#U*%G%D*hYaBQPd%i^EVE0o;M@@(9I`pwOv` zchpz$ImAE@dVP`niy!iY0OrM~^aLFSdl`QKRw=`uCYkD5!nT#&pY^`T@?q|LA$$CJ zHIG2!c?ao_(%ncNJ;j}77TE*}7DXSd=JI9SI7+aKGiSmAOhc5|#aY)1++41Cj>WM} z`IAw(hzRtOT$zj*9pE-ol*xY32s0@=N|cJ$P(9GH7lHMp6e z**?)|h@L;kHOLG>CAX1{;_CEdg4ZcZib&SA7r*p3uub{CH_m)&#ob_R=D#M~M-sV~ ztcjcXUEEm_7TXkM_BSkPBZ@m>;R|Qo>wMa_;{>0{N0SDvJV8NJvWDt2%Wk@*!BIR7 za?`T`XPm7S+{%pLYfa%aBE%X?eAoxw*JzG9a9ezV>pQ`H(E!vrYuq0BRNgXw8ZPjV zJZzKUD7pc%>E%4oNWr5F55#d%A_h4b=NaSi#kc3k=pFO=_>G`Bw^vfz0wo9Hm)=S&2_B1=x~aS*kCo3b%9+Esl( z_fU0O2`C_0;lG%S6Ip`}ml3)kYl%RS`6E7~9rlfC2c1tc6V+40zLdTZ|oSu~(gBy%@fKNOU&?;t!-OIhw9}vTX_|d4Qj!^ky8c}Jj~9yAdJPaa%;Lk$h(^l=wmi)C z*A{qQgveFORW<3M-V+#!)%Yuz5g!lEuHm)8k$d8p-cehd1oIf@RpMdo4cAvleTUw@ zmA)&2TiH%hW_X*(;_vIYjil-X4VJ;8?N*+dp-sLON;zUIUFwWxzZJ^;*-Z^s-OD*f z*#xbhqRNEe>oZ|>oGGCJUBm^=p2aNg%&pnpF~+|9S=ROj;WQe+2cTh~7GsNBnhe

zq3BSf+SyvG`KXXA{Yjf_i3n8Xwz#qyct8smWEO1Je!{itIXvU@DY3(>M4XD9QU4X+ zdQ{LWVvt-{hkF)TH9AaaK#koQ1I~T|jzQmvqJ70Tfu;pfF-!DD_v-&tCb7EPv2iYg zcf|m3g4R!xPEY^gW8M8<{wJFuZFW5eAASplQS8Y_g)<=o(`6RZSuk{$2=?~L$JTI- z07m!_eW`xNIpKTxgv5IrHXH4Tx#unuVT*QbMr#%gZFpY6gu;rWl#$EOXq@KT26E@+?9PsTVZKc?&qU~~% z#&a}mzBgzCqkYsOEeGsh`zVySNc?AS^9D5EH}#odGH?Q8sLWeHK)sjwmoDMyq+>^B zJJAthUql5;`(*Vy0T zws&MbGUbz8rCVY1MCbTZdGr}S&K$>|jEI_?Y!eV7aw1Z44xa~AgRe05!m0?)fx3e+ zwpruTZ+#R3P)^^iS*=gi9E8 z5MH5keIs#Nd@>5-8C#q>A3Y^L;Z}_K6z!zQ$hLT51r&OI%pvlByO%R-n3F%m>{lMe z{F-@%S8r*ehhjxPXLrM`;4B_>_g$!CG6|XU?cfo&lRKRYV5 zjHp=k4#b~!K?J`N8lE1KkMAgiv=bEpL%mJq%UEp~&D`8=^d^7=_6SFPsIWb0 z8Y^5f71PSt_WXx|BPAwvy$ZamO6HNxN$F>K3}%QK;qP8vibRn^b6IK0ayPlSmvFN? zP>;sZo6qC0#4l}0kmki;<>vBd8NN`>4EZ}=7miw4%g;CbfCLmvyF8JPO)FO$dStFB zp0!$69~u0(1D>ktO*W26EzFGc9Qqb(*RkHwz8V4?6Pqm0j8^j}nv*b3$XP;CPv5@1 z(|ZV~z|&iwz}MeU5>k)n4QX5PBY@u{IJ$)o0U;F>MJu90Z^ed;2d_$G38sqsGWyl# z?e_)!jCSgf_xH;M0|wZD*NWS6@#=Kn)yfy6Ule;mq~~aRwM^9>2<)b#pfL)2FvidVJw9a=A$h6%!dUs5 z4h|=XiW)NQRowTQLaZ4)^HTjCuAOMDFst%=*@lTjgB}aW@rvDG$OpZyC>~_41T^6x zHsYp_epno_QZ`%ag*(3U&IZ`S6WSq|ic2Ox1`8@h8DPp2Fg|Q*k$s@(H3G9Vj0m=4 zhimc7T8_o5GW|)4Zk8?I1{sk8oG;JKKab=cbX2|RXC|ZlD#`}4IL^)Ju^;c#;=kW<%qs_Z$mSzf{=NU zVXufXd73uF0R2PishK>3|5nCve3{Rpoe3c$a7W%AfDOy&?03YZi^E4y9485n7;xt3 zko`PW<~#Tk8urtaGV@KJmx16@`k5$u9BKE1h_T!R%BvTgU&!2X9~CxHjwUv)Z?HCM z&ki%g_VGb485AI~xwNmSJBi?!2(R_!2;K{huz*HFxE$h~kQV0?9_ahckxhIf_`7{S zKI~Zp>Y8d4IYcuRYL4x7M*7>?`NbazABzXuaA(X`5h^Sgu!y?%$A|MGAgl)2RfX2BUD^cuc?>{7*ZhYyf{aCBv8iWa%$aO-c!10SY!_~eI3GeJk;sn?i18Janj1LgD;&kF+@WMl$*XI`-7tLmm?5skUqqWF zV<$BzJ70GHXqfrQA)!5WhLo6yUMcW;5+&_rRJ=a!a|3p#`*_{+A}2T5Nv^jnN>de$ zfZ-_BA#VLt=%}2k!;;L8r#}ByNuCezBX_>vmDG1cn~Qx-Sb<{hIneK|E0TDX?7uJj z2j8P3PKi8s$)h+9KY^dH`A1x9f4B=_s@=1EWtS$2JDtaup$)#%wO?Ciws!MOd_Ftv z%B*HyjHdAT{Bou2Q0J2@uCsB7c{|H}f0!9<#QxLJ{muzy^S>5xlVXUQZrNRNxpp7( z{^mgZX8Qwb^%Bucst->RRvW@cMl`Al6r`d1jmEfm^5M0gEC(VvL0$UQUd-pXeoGWn zd2ekmQg84PZoP}pjo&sFLx7twGNv5@9N6T0(4Nx~0T-jB^;L?OAZRXE6wN99eKcKW zn!^>CxRnAZjY$XvoKRo@m8x|TPMe|?r!<%Jc>=wrQV|5I(a`LUgef|m;m7LQ`PjDly3wMnyDbhu-Abz{41?24P zyJQrD_BVL9xaGdT1QMQ)10v;)C$wEk?|ILv2+pAmLOdzy_2(Dj5$A0hbvtBaixnwi9%Xek#I)pyb8#NKsnY7lH~@5`yE(sgh%18+&HFGoFz$ zM^QyKn~FUSg7C8WPGzwZmH0_&9)cj5u?w$O9I0 z>q3M6jH})_rI<3MihZ-kii?aCnBPo%ui=M?zMP&i7e8xxLhG6c4JMFh#JjHg_Rhm!L&iX)FO*7yw}c?!ES zm2NlrQ}1eT2*|4oa(}Z;mo2fI^6 zRXd|=f3gd;El?!OU5Na_LoVgP9%yx-aVeSLtf9Q{)$(fEU>Z2v(1odF*tATc^Euis z*;|Id*S5@%%LXSM?(}~qA8%z7hI`r}&%*HMhugSa1cViH6&Q@Z7}xqN%vrY987B88s!J?k4xo$}He%)xVC?AG$p?@}ER%KNfFh)! zZQ;MpEIS&6xZ)YW{?pj!-|K?;-2$gvS=vvAhS|LBzq@>^n36&V-s)N%FG=v=diE|1 z{lWCa6MN7-MPEET4$gXQYbrCwwguf^`Yx@hD8Ik}v1-w^H18xGxALU%gN^J$l2GWV|BRg|>5-@O_Abni0Cb72&E_>ZimW;7RtI6hVnh+_Lsn5?( zI9vfXrkN4$ZQ|yZx_#PNX7Ffj;%DL=V_Xt9bbpjV`49{^L86flzEP7E@r=$H;Qk9? zXMdtJS1Z7o%BwU?dy}ulZZoeg3ZEP~$fjx~pNtk5FcK$%>meAe)0kNNqpa{)-e&$t zpGZ!6i=V~Zzkc@*Gi{rb@%wkrB1JzOpB^kW^ZS*)XRc%~4lP=7eLRM+1M%m&6)o@b z*IjA%-2KCk?p|1Ag?*SDq}id3A#9 z&OJ%6xRMupeYu7_L5e-goty_0_qo`^l>^RVtc5YG^FqLp;~8;(eu1w+@+MdWeLskf z^XKozvCvql#imue5;2j+TwKs8wtiI`R`F3du*gg3{^jAiJEvF5Gp3ebraW1;xK}B^ zxI}vysZZc8@My6e#vg`_<7O^^&>OxfWQ5OpZETZ%Uw==Gt1k4=%OQ6% zI(FNI=oY-#T`E*90Y1?b4}asu|v=!o{}u z3;tubg7)Mjo*@PH*z7TjWdG3u+sUSEMC8GM z%?8?8VD?L)6<~N$=_ti=Bs+I*!+*wZ8@2BpCI6ALsg zc-cZ3VK2yL9Q;wK9v;VeMGm~ao*3WUCR1DCRNVHuMdOky4$moC1Q{vO4AG-s8AFtN zVqBojPeoDj9&XfV$gD_`OvIoQS(k)p@e@Q#z*g#`q0JjuI37FsIc_<-1MD z%D{>Tvu%QiG}OK|%BOOn5$fr;!%aK_`Hxiub!vh!6Hw1`$r@M=U|tx6r22wi;1MXY zq^@^{MFSK;C`d-H;{ZH=4_nX2peUN5$X8X+=LP)!eDao^E2Hrx9&*GIic{%sY4a0p z%L`8kMnrHEXEx&3jknH9EGjg-lj-5V+;ZeD%L$MeY(L3l9LBz*ko{`@T0gsF`3KM)!AQNAsm8#~AEUrVax(FX3QYi&OZCFe^{f&A*t|5o za(7EQkm|a-5Fg{r%iaY|7{Y1?tqm$`U-&Pkk_lpY2s*CK8W1#jbo^G_w`EsPz=qC6GLfFyE0?Vz#1!qbJfA`J}7nKi-P zzunrlB}tewh<;OKRw8%FP^YFFA`E7HsXHA!<+*uOD&Y+Ok@pFh(aiG!)LW+$Y&`h+ zKtXu$RmJiKS;9($zCL0D5DYf=PpPMN97A~ zH#om<2sO6oiSAH|FOT}NUriJzIVsOIODjp=Te|1vdvmxOJ%WG;5|e@~3G*)Gq73pz zY_4ficc1RZh;V8uTorvraNa#6x0YiOYFph}bb_;3o=HVRM$ass43!0Ylw}tWP z2ic)^=*%ZREXm7Ppdkp9=v=tSoOcYIl|MO&-4)L%B`DFb zL6eU7CHCuLDU58Ej4F>=cSzp7G;8qwTJK0yPDLBmkq*QfvFGrVJJ3;_Q&lkv~CJDF%`9>rSlw_Dq_hqqIYoR}aNxlK_>4yedZW^d)4)71^ zRYfStW6c2e7~h(d)dqN~lKLKbMFdc?044CxQMaM+Ckzp)fA=mNHNPZxNd$54U(uEX?}Pw3C{&1GUPm;2UK5yW7rA$*r=^WlMsh#4Bpp#`kuTug7QO+EIxzBXx|LVs z!tVV+P5k=B$Oyd5$DYMWT~6kaHtAPAV*w zD&Jl|jT^x1`Q~b=wgcqb1EV5mEkEd$t5))~#F zH1%8K_biEB1HfF8FFC4?Qs&b)rf!(?x>af#`k>(A%`iyc7L2 zvtzmc#3%!yd`>1h`<6-XjNuOZ40uui3X1)z(&5wbKOJPVx_2EM@JyhZdFy9si2F9` zKV=r76evgLnJ@)5EqEc@PID5V(2gQ7fhAGitFJNi-a**qrdr%T~BBfz_7k5U>B7vJEg0+H>~?3<(SfG=_QxhJ|%A z?TZJ|&I#dyR8gDxc?9qjO&v5ogsW7lyl_pWbOp7g-Jm??$cvnru~=L&@})3-wc|70 z(0+|r5saEFU{?2_MTFM7&)`?Q0 z36RKr2^e{>1LBEED#;zEWU#*5$YFVZ!RIw^8}XeJlU&cn25pZ8QfmOK$Nw8>xN3~S zu$a|KA()HYq+v~--*jwlL}$=$0`H?QcZgZR+aTr*=4#-G{+XYaBz#gMP{56;c0tN+|G52Jv}lkWob=nnp-E zQH0Oe)1h{iQsxR+W>yJcx*MPz6Fw)hGozaTMP=g3w;O!3dzhLK%>!$;{_hdiZr;d? za=Dc*Xc`isv~0p#V})LYqkkrho=ptF0lpz>q!jGLqF`IH3YM9_LFm7qE={5R1?2!L zNplEj+3Y_j?LB(JK8^n;2Pv7RC{t=nNW-}P<}`K*ivkzsuXsgVDpx?{$W*+E;5Fci zjl!u)3Ho>4A^supD+KTD-1>Jt;X~ZRR-+(*27ex$=7k3F{x>`1Op)plUBdwcLCz!I z)T~R{Up?`#zdX8ZjkicmVM0HUqA-a9QVU?yfm zZze6m%4vyE1oQno{!H|;P$DrZCHUj~qu*M9q^z3bF=8YwC@nsPOXg*42> zJ2s5ckJX|?3G-aBf$_h@ePNr&=rm~A9VY6zXB3yNL{uk`Ky@at(y~R&oo8kS>Nif8 z1{KQQ1&dz^OqDdqJ^Z*xOp7<+g3iD$4TtGvZ?Pv3NS79^@dX;;58kYiA^M2vPFb9i z+($?#8pvnyPvJ<}Eq&4bymG}|;WFM=A{J(v0FaBHRFC)Zl{IJFDuD=5ZXm?Um5^!n z4%qv{ezDiHO@V4gAS(vZ1ABgHKzHTQ^BJVXlck#AVGvkKij^!P@6lW?MZZh{JqLO633n*alb7E8!&15 zJwbDQkW%T`LP5g7Z@ZM-mQ25K>)H&YJhXS2aY0>r2ghMmrwDjr-)EyBH{)ts)-n5u zP|ZxB_4PDR-0k8OI@$_)kDl}V)pvrV_1c;UD24=<2}Q|f<*7vDoMMf0@jQrDv(FoG1oI z=;c5pP-e4y;UyUJrY5KcJ~au~J18DrJZ@fDl*ye|dLP=qy|VDw<*J6Iuw9lP`1yJ8 zhiP9S9&8Gex>e(9Fe1ol-HHXrI3iTG}cHTnu-+l9Yd(j*oHY(V#uabC3KM9@{0rI*3}O z5`E8BAinl=xn|x-?U+~NebNTb=;;Py@(r~$P{leoLxxluq*<)ILK)~fw{=Uy?kq&*VGrw#c3H#rJQNsFj>jlafP`i!}-^QM2(u@G~v2m zoVqo{e>f|HGI)c%wqfrr%7>zo&sPs=iU#cUy4WX<9Kd$g>i~1#n=yv@hI1ktyzy+@-cJ^4T2`u9~v;7>`Ntb6n5kbQ$ zc)K*TEJ@(M6By*nWPJz)EbjAq``=NXB3dd4A+fHy{$fk4i0u7X5G`Qebpg}pY{+>N zEp1Z$_H3HAfiF|$2oU7@K7CuuZEZywe}mwqbI#U!J@Y}y5oduU_Y{|CkHE;$lYd<; zD@bgS(G2Uj!3gdV!IdeYPzj zf9A~v`})5li@w*s1lP+8;xuR=aH;-=G#BBo$w)WY#pqGuWGLpB2h))+#p%rqyH+Kx zWRlD@s;Rh-fj0iZ9M!vCCE;5eSx0aQ1R{Hp1bSz^9oJj_L`CWaoa|_Xai3BVLcSEM z1;{k+$p47wR;V=L;R#Qfcy`p_;pP55csWG5!W^)QqG`T57QH4UH30Wr1wDdKVd3?I&oX&#&-Iae!)Ks=Wc)bgFESKSBu&E{%?JQti*AUy3 z63qi16zpzz323I}Gwl(>=7lV5CG5II*~UDM&y-&f>wu8p-^dXp_UzjGrm+=;zS>W& zx&!UEKq{}I2l<#1&Zg)?du*pPO0yD@>AxdVik-(btvs(p{c`A9Fjj@Wvbsri$Syos zbLoG*(R8~;IN8a=<>`|X`X#a0mVz*p_BfO{yYi@_n`I^A)zNf!DpWz&TILSoDFJht z_=)`>F}7R@{^`Ax>%WRLXn5}9-#I3CZjI1&T~W8<4&q*(Cf9E$$$OlGbNrBvXYyXT zPUoh(qs<+#B<5r#uxWw`0&TM{n9}HRz$sHHeV~5Z6q-Dvd z3HlB_uCUN&0}=ci3wN=1e842|9z6KB04_rBxH(!5O|OXy95jwUs)0D_=YIGgUKV^2 zoPjAJm;H@Y1;E9HkWdo5A3LHKLhCAx+yW<4mK;HNUY+qBcpyCUn26O54Gizj^7*(n#I~Xo-ag z`Q2@s2yFx7Dm#w6!1{i#iTkZv7MLHF_lO! z0PPsRD>_68To^|RW&q-VJF<^@=`tqxoonYsNqyQZCZ{NF7!oqFl2r<;Qb9zJ+bL@y z+!n__$wpng8PeeDc6qaVQkq^mWZRn#4E@qx9H~A&swnG-mPP6uf41~Zoiit}^K5%Y zDFAPQ=!X8~RpN`51t@Z%tBY#mNVRr{*X4;r|dK zYeQTRoYV839Kz37@N+?QL5eRf_pv7owo%NbR9ET=yKl3AxglL=udO|>ABfm}7v_=K zPd@&hNTt2;u4@+a^z}4X&^~vpebFlEv}Oo!u&vNFG|7`YRLxC(Y?I91gY%8W|3qfsO%ClsF}lOTyzY zskMr*-+WAT>Ru$*j8E3zld}*`H(N7PU4<@}v&Dl-eTv*35xcZo0Af{%;1XkdnGn}>DVn!vecK4~EEK>`nj8^f8( zAD7VTY9;C2QKu9F`;s|iO6j@7`J!MPc^K9J3Eg*aVN)3hp@-QixZz z3ws!3boVQl|0-6GZ00GKN&N3dX~YgX=Rww*4)j5SaviE1Z`tjIl&v^h-I($qtS=47 zl%Vy^&IC$de(as^QdWZm3`0__Q;J7R$!y}1`IiH3A;avKcwJpCkG@%*pwwTcZg-_q zt5S0lPW$h%ZZ+MEq5O0_P5^ zMd;u;Wt6-jf~?4p@`B;po!AUX8eixTg@zJ6lM7->921YsV|^Fhug92MP+DTSRG$H6L)f_Yj$-JCpTC$J!IxVyVZX-rI9B|4`dmAab3pT_MXJS< zG7Fu9O^6u{2R1C-rn11c(t2UP>QvmQbS9D6waV@b>?EK4x;15-`C{|1L^eLX$i;F2 zKRTXrt&pV^l6ZiLzp)@4q?jBGsRGAf70h?Ht2 z?CMw)#~j($gLH@`cht z>&?ugqRtw=Kbk>j+Z78HR%+f%hlVdYhfta(KW;=oR5@8@Vf`s#p{%>AZONK?@gYKJ zh+`PHm^FO5rD;3Uj7t<%?dcccUaIqhE4;f8q=@7eqqu z7^2&+xP{LgAr>4BA?;KGL!z55Zwpqwu~Yx%{>%kmPJB$ zR~DaR>e5j}eU@UWDJ-ycBJO4dqgf2p&MCPDGZa*Z^$ps(ioEtWoX#*gOp@=XBfZEf zFf-K+2U!NsUe!VAvCdtf%4H@$M-{L(Bfr0u2r*e#klJ?ZhZfy5?pYY|5d;lE3R2Jb zYAgZiE-v&6#u~XM#hTEuhMOxrjX^j+>|+wcPte^VW6yI_tW~EwCg}_7p<>;PH+5XU~h&ItFx-?|k*X!B2N!A12ns zUV2cG&qxKqMN8x>mKu+VAR_oILKDGJ_?39>J)eUUmCEJK_pTZUZyq_c_kt{MXwDI$YY@9Y*Nb`P@(&@L(@h!{8b4KZ%xbhef;{FDSPvgpN7pn~ze)P>g(I|*aqIfc@4JsGE? zbk7>yjpC;79Kqc*^R}foP#>cdG;_x!brk@a?}yjj=zSs!I7L)-*81ADlmaz+rO6I? z$w+;|;^m2O8^nfvN6Hj+w8p3sAV6?O3OnV+A@cKRUby9rzo^a{KH$i4K>Mp_%9>{1 zmmgpBbjaa|Wot$^6SJ!r;vuk-2EOs_B}n)ik_|6jq4u_95R$CBi_}*cs-kRXWdMi6 z15rMxh=M!I?y91i+g`3DrpjW#2&O@_u_=Q2(M*#J<08u6X%}ot#D$?6=4xXNmK_}g zT*u0Pk;Y{w&CvbQWQc*;!6JU(qjBk9nGvWpZi{gcH6Pdx3NlE}h^9Sszbfq}LtpB* z%p~a)X^NO$1juxM_nz!UEFY?PtVH}Qgar{z_$LM83QQkuzsie9$omFknHZcy@vG9j zV6`J_L}i2_&sx;f%76fmprCykUoXO~S*HKa zwzAT8R^p`xqOM~Ne7lGyLK_2)20|Bk_xz#+xbKoa`VN$jzF#GMi2)Ck-}B9Gd41x| z0-dsJ$-Z=!|Fm5V%CWspITI6aOgYStr#bgYSE#|r$yR0R3X;EQI2?4YOs`IP<$K;* zmx6i1hU6EZ4BsFrmE8<_*stx;k?i{+fvAnoB;LUXcV~RDx z#4e=Is4JbrQzG5w4-IVasxHrckQvoR&!4(r)xYDQuXSC7bmL%*GC0_tXwMQXt0%$r z5mo|D75}n-KVSXG7yNUr{-EKDtEX*)pX$HdU^CO-c!+)2xnK-m32BweagpVVZafPw zA&TT@T-^+5)Lo%kI23DOhLEE7YshDFrD~`H+j`LVtPgRWyXgT=MIg^r9%8c<(e8Uy zk$+zSv{9w%SYL`*Y{^7Nz6Z4SkH*|%mkVVk-IX3OW!VOuPN49G7XMdwY*Q(Q;>I;t-*Auqoet|RL(UdpmuCLEWPp+{f zJB?eOWYr`OGLNTS2A!qX-T??y1~sl~H3pa6H{kC#iH|GmC9f&*0~?|1AFD(8!+5Xd zemVYef;rh$dviIp!f)R{oPD3?S`J)eh2#I;%suQ33!)48)=!od`eotWmFGMte?|1G zySh2Wi&K1f>h4q%n$jo?c_;?c8ao&dxwcBroVvL?c;buc&Hv^1rrwSILgAOBYY3T6%($!SS09=aAh$481gjbu0mp^^rZ3j>Re=X@)O}%;Y85?*^0Qg zoZu)w0Qch+$Z1)QcBgYGNO3afei{#Zyo30l{AY@U=6UgW3`2AW$AOTKgyviP$s5PS z98Trivj|U}>`D~DcyBvshTCRqwXMONMd1E*9AW$yZn~;LG8a<R@?WBeUSHxNa$ugH3ts1$R=k~;XL^+z%Df*O!kS}z7;Z*DaAZIycx*uw=b;kM* z_#3|G^0x-lDx$aBYoNE~Vi$$!OBc!+tg}0-iUvEXl}}ny&^49vmF}wZ4k4k{pC@E7 z`PgMQSF2?=8D!L)VNMby28nWLIN@@zWV_)A(3;gyFTPBGQhJ|zg!qe6C3KW%N$?^szvds9$3 zPIxgbM)oJ|26l{072}5JvI`VSDRp*Z0YB6Lk=r&(kX3*0f6^H6q${K1TlD>(mnWfZupgbO<=_$Q(gg6hP3tgE%_xiG;B^DD( z+VDKvyp!@NHa$!q|CH<3cKl3@nvbBDXYt}=j{YosMd|#6qv?=V$ieC?EKfFDXg^|( zU=BD+j#sclB}3})a#Cb%W=#K8s@kZf-g71I?)QP%+nXmXpgpzC-AYL;lqP6;k}2DC zrFk*xdkI5K71go7kyFFfp4e)pZ5i_2XrCk<)Gb@H(hd^~3LBXd=b-&BRNF{XJ)s8D zB0Wp2T1}h8e{ykXEWtPn`BuI`C!V2rHp!hW+enjh;c-mj2J4;p8%#%7s1`wpidz^L zaR*6BiZ4%@s7+rsf@aza3V+qXu3TvSO|q7r=J`wRld5T3NifUyWFh#THCm-%1dG2_ z1!z1&r`8&0eODTKsgWzmAg*aZ~eW*olWVIy@Dp<8;VKFBNM=alNQI zHR|L^)|`UD-o@Aekd-|6hmyyR4!_Gwo0u35zr1Q)6BJvGxo!2kzrig>i+FVg-cxIi z#mu?i*fv1Us*qr<^Z0Xq&fBj+-v$O{JP=g)p|2;#4|M6X)oHu?mvgN}Wsf&c!;k8^ z5BEi7_qVe;HVyc?VXRzMKb^G1JKWgNJXrgJnsJ!+O3v+?)z2QPmd&eVt(2i8qeRFv z59id@RB?LeJ`+lr{02Q%n&N3718sKb!TQ zF^^6L2-)h#FDcxs&8BTUb&Nct>P0&Y*dPqLSFtA7PRZ68hh9(ky%mP|7KH{F%Jb@o z4y4v}Dx{Yp#gJj&t^B=)ILx;eo8|KDlqogq-xYp%nUbYssh;C%5z-|dGr9+0CF--! zf{S;}$%8>5HO|(cRIU;iQ7Ig#e#WcV3`L%KybHx-+RQXD{f%Vg#wos#p-?a*Tth1o zYkdx>t|LQ&9}Mv7`uPcxpBuR9!aVmLC8kwAL&n;?B|8yOyJ@lzGHu#z#?LHD~ff|Yi_kvm^lByzbPQpy9NNl|8AFfNQ zf-KivZx*%DRDlpyL4ZK%gjCHo$<30Q0A%wWz9p;FSPd@`4|of?O2IiCaYE&^tg1O2Q{KzQHaM-Lw|~G?!YwrPP~`eP1?M1J57zw5J182@gY{X29$f9< zs-Oo`2yUWH@H{IUU=RBI7=#gWwbV%556K#{zo0MAlS_IgkA%dIHcdte8(sdTEKA5H(?7KC(8Z~tgg$q0fw52pi#LQ%2PJ{o-er&9+ z!1-IgGit#6tH2_vAS$B#pRbD!gi+=!ep}rL>6>)rrKABV1Z$9`9W}*Q-{-lZVs~=l zyMH#yInYiV5%C)%;u@hEPIufGA0r)0Xe&pfa+IJFhnHw;9oeA23jOXEl-~QS{#ybx zAKuQIf_gQ%9?@Q>MzGM?{r0h9u zNEpE0s6UzeR#pIvOi$xkr-%K4vZBR4e(B6uwtrAY$lUSuTa8jjBMLm#D{tg8ldQF- zRdiJ{R)VBjD^4v^^;#9XVD?ua`Vl~W)+Po=h{taiF$G&tsX)kQw1sk!zvzWtlTZ4K z&U8fw#hGm1rKBQ_0w)VBX!X`(yAUQEe#gQcT`)aeyw6mFk0zDZ=2I;DpGpgti*FdW4Y|!n`sd}$LoTngE^^*u!BS<@0yKgzTa|rY@IJK6$v$U0!tueT+3?-%uTet|NB7km;gkE)E*PBF%gL&|^cUmqy5)^Hr!;_IeSdSiAX*fe{f3!Z6hzZ-A|y z{@f)KH-u_t>a(afgWP;VC4dYJsTJ}Skm}l|#_GoHUGIRKmT!D;1i#JFaduaVuwae# zjvdl43J)f>EmyMhtFt-1w%q7gDM)SEF9=@VdF#|)?m_vkno*5Hnx~nG>c4Cu_!G{w zf#u;syhbAv)h@K{^;1yn&dwzO%X6gx3D?5$n5J&W|9H_%E{`P|Et4Tf3 z`g1J~U=1}a#29e)-~A)B|6OaULSR~+6OD@=<1AJabp*VUUbL}Y)`;I#GUACv1Tf}0 zWT!3HKR+x)O=I*OqL?1XgHlBrS{glDL|(3?T5F~If13l;g+9O71C$vSsx!tnMr2xKh6~3a7ZdLXqQj zVFkLa*2{DLX?py}&(&syF}F94N0&m75d4QW_UVi-*#(Fvt%*w*NW5Vtw=jLy(F(9q z`W)e28R6@m?D#UCfH?wL%*aOOQF3IMTVPkor6 z^PPxptaRSDT=Jr3Z#}cMDdic`dS=HXdhL}bolEL1mKMt^=J92jvqJS8*8b3bXOLR(pr8_g< zRJ22lg88j8p6wQUbE6dM^Yi^pc&(65P>@T6>*L*6!gVsZQ-50AYFjgJVg2LKtJj?* z_FfvT?Jj@v&=FgD_Ob>Ygriom^vm{8#s;!whCU}cNuoGg71EA0I$$mp7F&9K5q<5P z%LI%79qYI0{ez*lL-63|m$8Pw1}$iwy83+W8xEwcC6TRV-%J$Y0cklm+*3ux<|C%3 zARP0xH<3l%RVny(FR}+Ff%{Y2O7DwF~7+;caKlp;5W<7iU?cc=QHmX66k{ zKQeB)AZ%7j(2-5PXN!EaT3eM2F@ldLLQ3Xsr~dPtFe~}foN%F<$-F{<{8q@{-_q)5 z^0R1>WstGi_wvVsHljNNUvHUiq_VG9#T+j!qu!ps?AE6oWF>{W za<74YX-w)owNBa-^-a4UQQ-6BymWhT?|5=^9~%zaG`cjW1c5Bn-VdN+pI%}Zd5ZDM zw@0ybBe_xqbB{M*@u$ca-N`r55sy0XtbBM@6$WUUWjRXKg}4z0Y5(Yt(gg{a-3Zs> zS)EmWO;`q~CAUiwsx)_uji+$gdzI;o+-FJ@xS2Wj-E2v? zC8uSk>cEMnK+UeBcq_Lm#Py(|`d?!+PTWwzn;ZU0WQ_kv{Ss<>zcP3kxf}`7ih}&lqZD|f#SyMFFGydl@-+Zz z;itrxk2uJ4b#?fM?H<{gE(Zipky*LYhAi|@`OaH#7G98rG}Z+i{?65QkJ7RJKlytO zz~}1*^VY~>3@u1DY7PgfC%5HnKfq0WJ zB(qqPdZfff)(pEN&XR>Z2?a+G&eHL@9hQZ~u!bjqZ)PTp`iseg^chl)d`P{m$j5YEmK2Z6_I(*Rxr#-T@_a_a5PWj@rd>I&)4i@yIm5$ zKD;DIghQLFjVlM3I24US)_Iub+SfP}+yb!&$bJ==uM~MReE3SGKtpW8c;{7pGK1}d zj5KbxRl;#mff{r7wMkb(=!EeRR@KDN>(pZ|=@;WsPh%Qs{PY>hZRM!&tmrQL9wcdDWzP{$cP2abes2RQPI!)`aqXO9(bdM z?kDG^Vo48eO*yH(Al0b4iWJ^>Atu+aty@ei`*!B(#Js3?=eO~7bA zGJmpoEGqfqAhL56-CHlrPms!80N7p(u_&wOzimaQTTk=SD=Jo(P^*;-@s00P5A?Kb zf|TBdQ?i^Qp@kUpJSTVyPl5pMZBrOa{jv?7SyrAb>JOnLLP}OYN<;5Og&bh5VqG*3 zIxg4xkQcI?(&i(;i6t!zGy;=ktg$e z%s+3R&CO@`4$WvCr}f_+J>Y92HQ0yW!G$l<-XDU?&Uf+=q^JiFSYxl`Dl`mbNi==A z!QgMt`Bm8TWv{j3+`&_H`Y%d);do!Quznw%}P-%CHJC0#cCIh z-I&$$nywnk^G8F!B=2Tx3htB?!{-o00rd&?Ikh9Q$sjUs0$9|5H!N!IOQ@?(yTH0= z_esXs;4hIldB(6*cw2H3-rhwpV}0m;#sU|P-?}mBD+0B%H29C^O8Qqz;HIN81A5d) z#b~z1f`)j6(+(x3TiXp(SMDK^m}$`ODhQEOk}J@J+~cB6d%ekjYm1^Y`FRYqR%lOeMbYZ@;*JJsOu0O59h$;3^ zC<3g>vkoyF$S%8jIL0_z>9d6Y+QO%niFRA}rgDfOTAbU2p6X;9hlkexK{c^Dlo&iJ zS?N?1M2u!klZ=eE`KN=4)?vgdlc+gfg3Br?^#@rwJHPD)RvR+JEyH_JG{EWI2^}P7J}7>aH4b-_U{h^>^`=N3ok)ZRj6yf zjcnH>GDLvk!)Q#4TaKqo+($Vp zwo?T!hpe{IfKqJ&D%Nh6!^cje$U(F!JOC6>2{}2I1v!~W`4}XGCOmGoMWm@rmO@lH z=&1|%s_`F?aJt8I@K5*QH*R%AwCU{^Hv+aE+r~GN^Pnn2rdI$pj#{!W2a7#m?jK9) z^TF7hDvGXd*cGq0-8u-1?qD5XmA*S<+4nN!HQbbvK+h&yivP=}qtTkw_o4mnVtT();Jiw=%FxdC1CF5yJT9 z6z=_L<$)>ZUY)d05QiIVad8VRLyWg6kG#mZbLb|Y-o%{+lGe=nm{(&jr{P^RPnjWB zc=CMbo_3@u!OpFJ+X#pVWArE}In~)zo|6Z%gW$Rg>Z9xXV0h_0oFT20(D7C=+DK{d z(?%RbzZ*vipkwmK_N&Q!nlP0MDO0tO3lV?S5!uiT9W0tplj}*6BQow#dTri_b*x?f zVz{JFF-fyBM2^{&*Lkv;!qP9KU%9L{{a2H8`2pg(ZOOQ;fm*hD<(o69C=XAyR`8QrGmLN}SxE6yRiN&}Uf*m_o_Er6jP5c%+H4RrQUj{kEAD4C-W zT06P*Zm|1~u{9wJjtEKAN36ac(#B4BAJP}SF=}lK8^qM058oswRI#0%HJH!!;TeSl zS1k}-O5KbvXu9-$w?5CaYf<5@_1RxmXO(#R&fA|;`x=);kva55jpq4|wwx8mv0_Bw zwx=A5ASX3t8{rVulA)-ofgWIBILSNvTvIuOj*#OH{1LOkm8QK>lF(#^D~+C_vq>nQ zq-gSL!r-sx{*qLH9T#GLcg5WQr6o3*(`FC#aXK9Geeq8rzF30tAhm#Xn9d|j49WO# znn00tt_>`FIXwKe_eJlwJ}Knb%j$Wax~R!B|qYdY>=kWO4%0PHQ>M|1F%Tq%&GLX}wQr_DTVmu-TOCx-* ztP=3v3I?gCRZQ>jarzePK65+6G=C++y|5=^TDZj&+Eu`K#3R7o9 z65SbBMq0e~m-2D@BnawGT15NJB~f@cq~N4!+aE-U;-m0t;z&6qD?nPYD~fFSGqaPZ z*+bP<>-xS#Tv6(}b;p-Nlfrx6iF0GHWLFd=vVJ!6mOIH)c6;pK4Rrq9;6i4? zvW1!orSao!;WYaPDJQ+RvL*y4-f#YtQF>7dZ%oHMUi-oq&h9Zo3~Wq}yOVp=6T;T@ zKcT}x1ydF{-E1)B=sQPbD1Y)s$E~UI?rP?)*_clryvU&Q3=?4wS*gg^k{>(?Sk`^S z=8b!H@zcqavto|0Qp+@AY1#!?e5g_rgq*0;dLgWA*=@ytWTI439~$6&X$}~&m$&Tn z3tY$Uc^v>lHj0*5OGL24z^cuFd7{tdjXnoic#yJWCT5iKDJ$8h|>GIdbVHAGe#Z z1~!q&a&6^2*Jyt-(+=83nCzzfixYEz4%BYk>Mtbp!HFsOL2Ql%1}@FN-wfW(F^K-Q z)gY|ka43DRr{u%t(BMhONnM_QQW-udGVv0nPf4!W@=^i~R_$lBfb7&bJ6WVJ1BQ48 zUFco3!}C1PEDG&_=qArlAz*IUE-T@wk`!P*Y2>%|HDs6!4ci%3quZJ%6i$_ct1okj z%Hxy#sicZV@kLx4%Ro)^B8r5mP=9R*Au&A2lqA}Zg7&_IIBgwc{*PdXrO?=E>xQth zwQ(n~aTf*u=g|jsL8slbkfpBp^1%ScGYZk~%$I&fDyQ<~Fi?ZU*u3Pl9@4ud1xo@i z`Cqdq2%^KkH04JaV5cOzClUmio}f-ysCge44dAAasp(>HXb{V3XxQ$!==igP4h?;! z4s!~&yi|%zHC+X$Q{w>4GV>!*(>9;4!SmF@y~*>;xsa%@VJ&4kJV?q=V4XS67Olzq zMCOW(&P@vdW{O4VN)uN^l_{6Dpz^~?b@LNUtSeup^UhyMS6RK~kp1y;y;cM4=hjSe z1AAqjiYhfzZB*wduHFo%kp<4eeX<9QWtzp;NSc$Cg8T_ zMxAy+Ou**RewD0kCPRbB4-V4;cCg`Jm%NJ0?W?tztUB#(7BXM>E|xoGEvaD-l!&f6 z4TKKT%KG9d5jrm6EPPPn`FeA+EMo4kLYt{@jSW59iL=TDt`win>caOcr?j^Wtx(L8_STHK1$;!bfX?ry=| zfB;t&c1cee*API33*6!$`b_xk(4e8}c5dp9?0vorIZ+a;C9l+;Ya6OS1f z`-SWYQeUVlLT=FG-mweYr#ux&Bpy$y_j@#8;?)r_e)wj0r1b2DXS*kzn+DkO{AuME zIi&*j#!DL0B->GPi{dkZJq*JZ2tn^N%!Wj%(qGOa9)YDrNp?H=tX0QM{hfg^G1HMo@p{?dN#4wh@sw|-r`)86(vqm86-3xdZRE+~&N#qOwq27CKl)JA6#UhjBRCnFKjp2sHwsIEUqC@Yno;GV|J>zZa6c04DzZZNiuNq;Hp9aL;J#oIXk+>cxH) zSMSRF7B}$PIsZi&+T|l$o}nDl*Y!nF3X+cpETVK7&4GJK26%q1o(%ouKi_D|V~IWh z#<3}ev$Z_ChLq4jdN87?_ixuQ{86En8&O~M3h(=&yuXn{v4sAr@NQ{nT^zyPm64J~ zkS*`UewuHGN8ZnYDlI%RYC?vl4ir~GU)ky?0leel>~EO~Ame?H}*g^C_0Narz#j$HCeJ~dgi&#$x_jB!=4 z8`f4EdL#80a@W;gsA zsxvq)18y&PSVMqnVzrnHmc0UX`Ov*hL@bND0X_0o%)BKNlF?knO$-PN7U)VJ=*Lq9pcYy#e& zl(r|N{^x#o#^l7FWaV;dcwIwdq&p%zl7Vs!+Hd;&5-PDIt#GN7JqzQxrbgdD_k@*& zb_P7Y6#a@{7A#t3#;z5$*)J}*j1XrS%n-JUG;%9M?a6?Eb7lVFlDbXQf%=zBLQFQ2 z)%j@Ls?s%IY7Dy6@r;ROC?D${%-V{hP}k)a`KPdMgw7Ioy&y$O)R*yMo;^7EN5KF-UR*Xf6jL#%5CZO zSm}RIUGPElj7i#hO+(0PZ>W5(45j}%=BVVuo{uqCLM)24wo}57L3@wHjez#^a%1hZ zX&LP5MXg()-EN=CBBUq7I6vFIS)%m=Tzo{8W`smjb>d8EVVrRXfKyhw5|^O$Y3v8D zMe#=+)f&L&iqIPKYw-lFIy>mzvH@V{KiJj<+rUA|My=0Jl3f@xHYqx8wZEwyOVIb3 zK{0(}twwzN<>qKrW2G4>ftnpsZOrWxy<2{6;Cd+UN9L-5^rJXtg3RH;7Nl zdzzFFD$rL)>t^$SqUrqKADR#}zSflJ>Q`l@Kh_uc2uJ_{ZbljH%L7-T9zTVA&T9nk zUzbHf3O;5$cb_X=wIz98@*yLSF!49rpbph1I9sN^L5{+}fOK|tLIm$WeNcN#&8>+K zVIsx%@Klh(2adnR)ah`c^SmR)AX2HMBV=I$X}}5!3N$1O!=;dcKc113hRw;d{*I^7 znmv!|8Xh0{U-zr$=jUe)lbiR;LB^fr;GOL&GtZn4oHDN85{51g_{JT7K-yhx>XpFv zrQlT3hTZ7$!e*U@=QHCdt3M(+arJXT`kZeG03e~ z#pqb$44cqG;cExtaev=>5_PK02Pg~36b9img{cq6Tf=5OqTEtv1&C+kD0tQOdY|Jt zy-yATZFk#6`Lf5q(cdS@yC+mXDWY@}a|rEGv@EyLKOoJILsSJ{8oH5Jp_$q3_wF9d z$3Iyw5PK_;Md*ghdc|?m+t$uA(5_Vws0km8L5T2}FcbtGOCH=n5i5JV>G1#J_ zo!&50oCEXN^XpuZxY?b}o$j1yf226cejZ|67ZIukVmEI3vihX;og}YEV<7u<7;Ma= zo{a+(Ro(!@?{IJiW2`eKR*T7=l1w-m%Q!X1V-&?yx0J|cKzc-kdqbRBLw>0(|1iD6 zV)$d)^<&|ulZ^~DDVk@0Acb0ONBEirX0W|f5(l)}FF!S8M)BbVO4O1&VE$RuUnRVG`Jl z(<{7U@erpw{j3GOth-l>Ih}FFAD6`9IG4wL{%p;eXK3bXscLMZED!ms1{4M{{=y$T z-Ev|2(Cm9Y+^N2h0s3rSeJ9bW48 zkW>#HPNsjz$o);V9r%Ps_zO#IM!3tj1lfmxZJw7AwZB@&0Z>Du(Bs&x$+bB2R2xak z$##1KCd_2`AARA%LNb`ing%klt|4sFqn0MTJR5`b{v26tQ|C~#4z#(bj1PMk61+UM zz_DzcE@5aD1icaZGlF%xPZz)fnDtMBGzA8>hN28+!$om-=K{a!b43W}emJr^=Xyx`BV{Xhmw6_I zTz&>!Nv2TNn3D{PHC{Ka0ii zL6DzH2Y`il80u3bCr4TvkN24GmzV`!vU#rImbF3 zZwD6fDV*W_?KeHX-S$osr0CKdXR(<2QrT|Eib=k9E8y`ZsLZm`Jc1J{;-upabm*Jym>G;BAo96neul@>x91?xZE? zjFb}=kCZl`EDaT!jwb;bO447}Gftyw-lwN5DQhlXRN)61v~8dUk)8v*sxs_y7ftvkDssSo(@ho)@xJ*T_a<+@+a+x zkMdxAkHf}n_MObX>P(BWmus5qzo+q;{pV6q2&^E_Jf3N{Pm>w#h{sV4OU^-Mep5CX zZ$=CeHWF|N!9@Xm4dwcNA=~Br`SwuNz|Iq?k0u@;ci|xxKiTWHG_UO@9T_d3w)He5 z(L2rQ`J8jrrP9oxqk{$+BYZYRaAG|XbkiJ(c*vikDoWyk% zFGtSH50yDp;1MZA!DU_#U<$1gHB$&zmwBJ)kFKF+iw4J13MlqiwELu z(NrsS6ouTpjurGQ;?y!jhKKSa<8-}UNkqr`-rLJUyVr76DE9j_^%H?apK-{6R+xVc z1KZHG2-<9JkxXV}xGP(weV=7u|95I?d0SB7x^wmNHU!?sQ8~8{+b^Q`ME2f~KOBL+F{>WJr zHte-x32xRXriywTL`*nNH}f!F!sAjEK5gfntBU$1eLr1dv=R|!nr}znL*cR_B`##B z%7Yzs$CMF1ByrZs)|M|DGF6Idzo31)%wVEfI^Kov%_9t+@tOXP0M`Tg23LrRcjBWZ zanYWNfWRM&FSWz}RSe4whK&O{R~p|Adn~xXj)_5crfyzTZMYQgw!}11I&_iy0dXlN zao1i}OY=^JpP>4&vACD`ptKWWHa7t#dD7+eSeNIzgF3uz9;MzmTb)WR=9^x@-3##WMN_(!D2A2@4bN8io$85!mIxGE5GeMR32+UfbBFD`CI>1~oM!I?_8mqw zBVKJt-75+X+L2T_ zso4+7!Tmvo0WFJ5lt3dMEaeq}i9vsN!Us^tJLe<|O-HJ}dfCcMGTjV_)P4!xv~t;k zw@QI%?3(4)d`8*)zJ|t929!^A>E7bb`dZ3K`C5AYwsWZzm;f_59)+c9efk>*8Jt+9 zYA0!t0;t|C1*;}D07H6L)G@2JLKE|jO{x~BkFbS-tzi@Bh{2x?Ypt^2Y*h)o-$~WF z(jqzXZnAp3dcd4Nv6bKQm6@VfFH8r5i02Z(TpcERng(k+ULep#WBGUwaVMRx0q6o# zr$q+2N6X6$bWvA+E`w|a6aGF;>nk-+so9HK-i=2TSniOIlD5SasuR5CQyplyS@%b+ zv5WT%mWNS2xZYoHE_a#k&DbJKRpNu`Bxtwql2_=jj>pXsUs&;zMU;#*J~=)^2Mc-t z=+>sOkFq>0Z`O2e(~x6{WqC|zOe*&B=Icr&D>i~bKy^}X`9Eo{tY)md)q{##l;kc_ zQMUQZB5R+L{LP$oDK>4Hk8iEeZPT*J6KylKQ|4;xmacX!Mwwt6f>5oN|3`6*;e`_B z6XqczXf@sixR zO?etS%uMmGAuoZTC?Y6K=pC@bm+4nVh8Q2TkBm*o(<)m}WX6{_8ptKW??0FX!C1ps zfjjQ$#c(H()Fv%;%+(xFo!P~vIJ|%dT;vTrK{Nk5Vy?y**L-p|iW6wcr}*NH*2!|| z0ZOQ6_Ti(-|w5$yEM2jz^AK$&?I4^&i5OyV!sH@wpBnZ=d-+zudJA+Q4AE z=B)fx@o`nChn?Acd=JlQoIX=9*?ta2;kIYk@j2+dW{GAd0WJ1^xXu;rpKmfSKx8J= zwJ+nh?0#mwSo!f#E0g)1htYuI4N3)_J@H4$`79Kt_<7YCWnqR-9u}&tlS^q#sh<0d zNfA}%X>?s|c;#oRoRS}O1r(D;zGTz+oYUI@_L1kl6jH?>;DQMj-v9py#deK1E_Y6y z=j18>2ykPe-bd&_Jd^%E3&ISPHDsxLey#3N;6$g)bw8CD3zy{eJA6=wp10ob+dkP7 zlQP5EJ_imdFHAlCHyAoCUEe9IASE98KmF)i=y!{~)wXrT4Cz_#Y8#2l7gYM?Dy+SS zQY^bjyr?4x$D*tmq&cVxGaELPvEEd_Kk!~6*Ttg9%ZzV}wgvBaWCXXYm*<#uc0 z@b}cV=q*LrMx_~{6f00wW^{tCqVO6sZy%8%H&ah;_7^LGmyqOJ!vC**{Ai) zqpop2M)glnotXef_xN(Y+1hEAqb?Hzx940%Mvv2p*&@5N@7HRHnoVFbena{r?@USy9VU34fM*({65U&6dUD@e9?{m*iE3_pmscU4C;gM4^a&DfFVH^hw5qPp?{n;BI zc6*m)@57KAFvGyUsuP{5FZlUX>E~vIVr5L34mUGu3|m^7y)cV1MzUXFg4UBD7u)Ja zg0HXpBvQYV6bBDW0RH6wZ330x|uuG1j?F=0ANG>0$~sV!sLK3q3#& zPC@Fv*PS2UdF#!wqaY_q$Balb1@2R2WS+?(m+M8tf;FV=XuLkQGqMBrbHRY`q8``0 zX6)X_-O?*5*Vg>c;29ZY7IlyxeLim)^+?fG;iy3GuB?6Jm9vXUd$_7$*S*qS+T%zT zG27-#$b+s-Sc*bT%WzPCpyY2dcYAwaBbgQ*2&PKI zF8wLsatnFhSOlh_gJ&*uayc8dCtT_>n+k2SjE`qlG}8VRVoJN*guH)3(-X|3`UQO+ z{_tzRKZ}Kww#{v#CnX>V>0CqU8nux(^=tE}Z?j@v>ObZ0r&=)Cz4@K+slG_;{Tf!> zNd1Lj2WDx3^(Ys+opz@jz(dlJtsmm}saA`-fonU8D zcFrH)RN->EMVCMnC;gf&VEFI-{&>b)(J0mHB;V`D9539QYb#kWA`ri3N#M?p?{44Q zO|lDC?9IN&vn5FP(R8f&MWsAu(APr%R*gGG<%el?|3EaT@xp)peuP$bj%ZKTDyDVD!O9RD|~U$`3>g(nHpBr59zv zJl@rQ+hRF6=xb1WnVzQhHMCQaPRx@n@H6K50&lu?q}$oUCL<*bumb{{**t77V|$k* z^A1ghSHbn3u1IQW$hg!g?wFHpwAKcpY-7V+hjX4ifkVwt?&_T0*sh@Skp;F%KxndE;`A zAQ6=3?wpud!omif?g=s50L~+-m0YEPs{{{$xoz_HGLLWdn8=h2gp2#cJsHaPI_eKn zEtY2lc^8;`Z-V(O8!py<8q89F6280yh0eh>R1AcD8BtG8hm3eQ>_yL)9< zM)+wY-4SZ0GrDqxjiuiGPU9&I4gUcJEkZva;#e!C8t;NUicVJUj;ZWQJru)9`Z-7A-0kxz(^Qg`A<-c?Fk&lU@_ zh$ikKlQpWp*($`hb|`nZl-r|ST3lU^O5{emVZxEebz8@SF7<4!yay~Blppx9^V5+3 ztIEH^3x^A}fWeuvrY~a1{CEECSgwc-?ya92nHBFHvMMa`FKDQVT_{8&ANLI3G5tq# zxd=2IGogq|vscWW&E!oE2k7l@jiC7m7U|p=t+_(21KQ^Rr8cgKdKgt!cqe2x@zR5F z9Z{!tC!Vc&mgLR}hM;XT#GrnxQHC>e-U^$Xt?EGFem6oXpbcYVAzez1AGakL%UP*s zIp*K6HmKc)8^e#3tS&{RmZ_-gysu<~t6o8dYiZPGl^Fkz1VNHds?79w8-w<%@NZMW znu+?iIycdF7Hf@j>osEt-|0svwI_5ctj*b?H8cL4SiXCAnB- z?ge6ZY`%?5_J3amTdkg=?AXA2mf-TGsIwF2d5R`G>}N#{K$YAlq_PXdA*y?qqzC$o zy{R_(451#@x1OKs*$l~W#@(vvVXpR=ESvw?s{s#KNY?PIv6}=6izOMv<9`|!{OEqwa#y? z^|!65XQ+(EIG7dqe~*zq!JM0ntgy za!!m*dyJO(I^S&-M=nZ-&y4@pnzFdP^UmhS`u8ySF0!bPT7tRx} z2%ch(Q$D&qDwez{OND=QpQ|N)rz;VXgDNn|vhws4<-&OMB?Udg514%mV?Jee zZrXGDlC&{En@Dr+42}`xeGM_hXS4CnF53jMz8>{>|zswk4S7 z%q{Q!`3OI9jCXB)V7s-SR%r=_+Cr)1CTl;IP2TPpmGuxz&_#~AX(+m+e4d(`<-IdL zF6|RJ&z~%=V?WNzRo4~&6edw|j?IxiuNT&PalY6O;d1dZD~b9@G;z`qfiKy#s7aa9 zy5f^Nbw1-yi>|A>WCX9Y!TSTlVS>;p(3Pw1aX`!uT4|ssJX@=NGfw;W>3AjHR&G-BFn$;6 zEP;bGBqM}0_`FWCJ9VLk4%f&PoY2j!K)l=jh`uf^fMOWmQiEyZrFnRkCFz%^$%AAF zDNJmcsE6CGiwj|a7%+J29970fr^B5nXhh|f=T*XWL13kZ(hFTxTQ3obOl7}kR11xu zi^T;W;afS7*HfHD=_Tc%)r^7l!4W^gQ6}FAJ6rBfO?`lyYK@hA^!%Z^qDk6~*Vyogh~%;H!vGhk}dw)&`-lp0JaS zUCKrXMi;g~_l1=%=*#FgT@hcFEJc#E))##U2a9@m23-by?1C(hz=jE|Ho7l|m~;4h z+^C9%;*4z6k4Emc@v%pZIRKR?qwzV5pq_F^$PBy~_=Q1@euqbdpyGoasl*DS!;-go z52IVPmxGo2-)z$=#JDB2KwIo%d{AQCuZAP8OVn@$U4Fm z0G^rV%UM(Q`{mFNnW}MM|D9Nw3TUQxBkK8AI#Nk5N!i=(_vUIra;+ltD(SNME3=#| zO9I@)ol>`wUgbsZhCx(az%V+(JK$^32a|M3hSV_isN`2GHPUAt&2oUEg9lr@O?e5> z(lUW%@Ar^nb!B+kJPWzGwmj@VgWV7;2wUTlM%(6ds5ZUo;SI>6FoFb{&TNe=6%L)=C_9SfCRs7bq{=d0&lI1@6%!2rFpVSuG|qf&bLt=c=NYYprHglKt1_Mh5j$J zRz>?_jvNG;)gxg+($64?4xQvk2Vh0K2|7=jXv zHg$oM2R-jumLlr5h{`XOy!6yaYyMYzJfHf=)G z5dEs{4R?I=`8sPEzWp$Pc5hzbFm>dHnI%K3in2Bg!(a%JUm=N(5@+d@gB$fE_>VrZ z*E!LbbwV#xrw46|%#tm8dKlf48c->R&&7#B(SWJ1c$W*P5^>;Y2ND+4`WKokdtP80 zPWln#w73Gu%~+aDP)L+q$d}Q)N_82>$H$&Xi zMtxxjO0m*8slRetX>Vy6JYTWL@i4j9e6QE3R-%qdM}{|XLU7tELa!*NLD^l_tq@KpP|VRl)}=n1#%-JQ8UC1<9@C;{a)aZ!NUPv?-t za>$B9-RX&fHH%&30u!4yJEG!Wn!M46ztKHh`Pb#JEU^34mmtSswnWGwn%QBMZRZKc z2Bnr&S1^-6GpW`x4)u9GSQn#4m(1=HjEN?Y43iJb>==@%TD~EkvmF4ECcQuW-$ehC z6Wdk!jTP`=_0?OZ^Q9U=&OU=FERdVP82rtwLOzv;?9T+6_YB3RWYqXx9pZmwup-QG z(~kl=d$2ZApYhUY%htSEi5^(Z3anz+EH;!ShF2)}m3?%^oluDrs}~5qxsNS`)vnx7 z{#bs?hRJ0(N1%)Dr8Ong)R6uTsNWCLxWbIqFs!#2r{W-5o{GjoDaUmq=nS|$8n6kA zjMs>o^qcbnd`xf?+{sDujMp=Pm#Kb9xS!EhP`j=suBQKCV?CZ2Z!V6A;p-{#PDA8n z(Nx!`SX0$NwtW5Gg35a`8kBA|mFIX3WU@CVd`)Kn2F1+;oqxrRF4^ky!zY zcBz#Qf`|o0?RDExIeDH8MEVCy<6YJ3eH#q^dygI}nd)rd9;f!5| z2I|5q8J4DX3Oh}x!;J4OE&H&xVmayg*_(|8g2F@O9b*JSthF`Mv5t*Wf7Xw)Z$OC= zO583#U!!s6?*&AwC^f^tbndosCy4fe96N9e z1;h?C=jRVNh^W!3!AasQakTSEvc*)=k5V2N)k8U;qQ$ZqzhX7VVl`Yt)}O zz<&KzUy7t}!adgACA?AN96Fe8Fl+rkT9#>2~D=${yiuT?AW2VduiR4SNXrXs! zJ?!T>(r;$FC6jIBD(+zey4d*nx&^ ze?GGoteIxY-1X1=GivAv)MZ}8__>MF0U!YDhWIt=kMhtngGAuO+fsnD*y$=_V#9N}!sRj{~*JzIP@{Ky@F(Rl7@$ z(-0Wv2{M~bF_4LY_|3?9_)Y5lP%`}62S?79&}mD9(B^ursZs;xLk~qq$sppo0~e;EW-_k#T_F`6D<(#8@zsNgzlBn@jp?#a!QU zICPvZ`cdsk%xvVq5bq1wRT$N}9vtA#8z0j_%)QEi3j2gjLHSFMJNo+B}3{Rw~pjc0=Pe)=?ZZDN= z|HTLo_i~0zcDS^l_KXgj9B=K9Io=i8Z^?e`uyRR`xc3A;-pLU)KFJY1@mFp2pPvz5 znt|a2i=YU+zprT$zY%|69dRteKJ>zk+Q#yv(WKa*aivG^C*Lv-zr#ZR_t;=%*NEoT z2=(9adL0yj`|WE9DGoH@lQz39yrJT;?gAPt0gR|-Cp4%bz3I6k`88cY~xnJmMo;`97_uh*P_nHCxCB6Jke&u`J zveFE$3i*#3;JgO-c=Xwyo8{}UogQ$gh2*pfw#})9NKkw7w>z(KQjq^Xn>7j-jGdE_ z$dM7_!Kt_pW%T57{d|m{aE#fzSr*!Usoasd6bYJ=Ndywb#S zruVG=N=qtV(vr0&v?I<>E~>wQFA0N^!0|;7|Mg(>vmg+R(aXakQx7If7fho{d+Yfq zvsy@eLxb+0f+9MC^!$OI0)`>(QtuNH%^B&4q~CcEOPmp!?c|4Sb>=jB3}%c)(<>f-{3TD|-j$37Wb(XCe_ z%`*=EI@`wdb8#^Hr=;O_+4Gxz1hx#3Z$v1=Ek)?s)Sv0M>^Eul(9h8{4`weBV_c&g zVi>bjo%fnPwv9c;ET$2GtRMF>N-0nHewwa(^T3ImjTbRGBEyj%1nVuX!*SR;EYl+yb;Tn@DFjags*kUI!*_PG2}`K;CO*beK44b;)ZbZ2)^~Pu0Qr8btkV_M?Drc}wz; zMMmd|ljbEJO-Jz4OJ2rLK=c{?4Z`7*smjcKLZwTHeSEm98 zNuUekd9jp_^1T-~)B)aHZ?cyW>wai+)`PtbPnrSJKCfjNKjRE}o?i0(4prtzIsEM8 z=J*s&!ht+g%eauUrrg3MXW*d|Jr^@>z_;0FQ6jH1zSiaIs!WqvftLJ#H){ouwWtFy z{By>}a|&=L$;x7z;wQz%G)k;UL7Hyksf8rOU&>3t&d&h8^p};z?Q0+b+6?-grAN6l0-E^sK_B~JyQMpMPC(LHu&!8dRGH=bB$?rYfiaGW%J%nEJ`gOFz zaMs>`gRr`Mprk}awk*=nOcdKrW2FEZPwo0D?!?V8jDO~Wh9j~0QIv|@=3(&TRxGtK zCN%Te(-2BDts;KKwWSnur?NoG`6RrO|Hb>jSAIp`Xf`s+umNuDS%!i3jKSrZd=uF? zOPe!tXi_ee?Ipk!kG`8~tk`>p>dz;2?F#B}_=3j?hsrmQC+@Of0%lA`1b-S$zMj>f zCl5ei{zDI@##&+*a7uHg()HE&bfp88(&4FC}Hgz-5k_bmHP?M-JU2_p52(Wa*J z!hb`WCt~4I9$TrMZy!ud-*jG_OP#cPY|+R5PCzkAt!%ak%~K`}z_V6{j21rSJp+`S zm32MP8W}Y`(cEc9e1OFryEn5!&uh?|rr|j00I$r>2I0+F*7-L8*bq_-0tulQf1TSOFoO=v8T+^BHv^&(RPB4jdoQO^*L*l01MfXLH!4Q*-TtskRN-`xlEN` zC-m7#`=(@h=`m|`Ojsexgv&!4*cu=Q#+O8zakHbE>PP>l31a(K&%+R=GSqF%1|6l~Bi!>f)FGb(q9!w}QD)dyV^m+L{pgZS%` z`yq4Oj1nM)-;?hrEQ$r*XJw8(7bSzGkD6q7c*Mp%Hvc?BZy0Vny#8t$XvA=bWIR5# zcLyw$KSsftMdrg^5Bmug79C{4T~L*v(ha>KBLEEWe0Yv|G&BhT;^JyBeI~U_M=h2A z1&VR%tP>D*)*KhY#GYmNnm_b7$ORZ#zep?oNgAsPdTV!`V#IszN;1!Htdl(1 z?AxJd9$pJ6r0% z#rFI-@shLqgd+oasor`?V-r7kwTDYv6oI&roh9WOuxm6$@s~8^Q0_Qr5y@B88&Fr% zRmxh~y4>094;~8D58g&q2S)|#1MXCs7Aw*5+YBG+kgkmE)feRA>zYvnJ0oQ;#l|jU zVhNElv_l&fgcW63iJ*`zBhxtKxe$|GZlrb+*|Q0FMg%eySqbSfG`mL}X)-}1e*)RF zm)={HDGL^a<3F6e?{p;)7iBs83JRItV-;+Jka!_qLaVj#G$un-c?fy0Hu3SM4!x$g z-3{PGlMfTK>r9B#-Z*`vMGQzy3G2$JxyR-PX879`Q;w8^DMsZdaOjDkldraOtpC(%Yyrt1YnC4Ti z7k;^ME#EEl483~Moxg|9cQlc6rjH@)sz5}*J9Aweht}p}uVW%rWB=*WX&30|;K%}j zCq~;2DDwopY5oA3(}Vf2bIeC}-1RDA-8dwHoINu9-l&#A zT1Q>PeFw4=?S3mC1Bw;+<4wko!T(;Z&Y#11Ru_U^K6C_lO`hd;`2_cv1tUpLp@q|+ zwJRxkrO93FPh@*87#i4?DuMRile8VifpTw-jdy{=VaMijLB?_^FFYj8c@jxnr2qR* z5DcgJ>^1jx=ZVw<{4-_4pevxpE4>TsbwE0K?>ESkcNyPveQp#q{A#SJI&a}uZWY3> z+zT_)M3Iv{&)Uo^?xDpLF=a6w-wFDeB*N053QE0Ob4o|hRFOpbn(?IH%?g+3G`Xwg0`kZ{K8LNjIl%wpXV+crGX18cUq>J zU-Ji%^Q)ITcE{2U{RYp(>A_tud+5UE9`Z~Gowp+BA`6J`?^Wl{_mc%(lr6nvn6mjw zos__~M0!>o6TV=W;s?BEY1KO*;?ty>Bod)zd7FNX*<1T0T$+5!Q-?UVL2KX7fz;|k zyL21Sxljpfv@k&&!o}y4r�yh?`=lBJ#MFFtH6SP z$NgXR@Jji;xe2J*=ZE)SF;3!-?B432Pv__NhB*vn3i=CU0VRHiayW2~B{vq6O$cL; zQA^CZ_X^~92B9WV;DP!ZCGzw=JuatL>I*)Nl#{dNXl3$rA^{N0B;c5N8s~&xOu1#? z&hA$VsTnAkD@oGp7VA!8mK>AfZ{fw$Sp(-p|1ul0z;v?EGDAu5*uPZA+@X{JtzNvY| zQF3Gc@7u)%{H$*l5LSHIGhd3|Afl{qWFzUx6>^Y5%NGB%Ai9i3D#v9-l>|D+_^utj z<(UTKg#AAXYN&R}yPY%vgLf4Bb;K3Jn8%g=FoqJjdhy*E`Kzl4hVLWuk+r--UZMr6 zalsjfsCPkGY(f7j_#*5Mju9GO+O#woi@}cpw~fL}II+$daJ#1|`SYUIIcu$Hmn6M|D9!Z}*0s&~U zHu6tD6a=!g%$5n7Hc9j#xsael58~(`2%@s86}Sl^aVq-jfa^CtLy20w`fSTAE4kyM z<-G#$hlrh!Db_B+gt-`FlXrSmMk^lB6ba`SO;=w%-uinWU@--;qb7$<|H3tL)i01s z*+L!hWTTcVWIO{Wt@*!Co#`uyhbB8_VmKJW4Ax!ONdfptq2f_PJuLgbb3H%E_7xNsjK1?F$i$ndJlUr{{#f}TH=gl>Hcp#2$lRpcLjer_hG zc5#|2Sv3#rm9u?q`u;(xnbL`n*VDX5{RxF*ow~w)eVF2>)`Rvoj*7>;sr1szHXHKl z1ZNMb!G`(ER#tF}=ejw?xYk2>f{kOybxZ};kh=;~_qSmUvalBQVXXGa&@kK>;?!an z?fsV=7R$B0hmup6+OHQlg124BPWPT?ID+>q)nur9*m6Pwf;YlKM&BwHf>QT}?!_!e ziJfp`)_f#LTZ|ck%p>f?QD!oW&DWx@@?9ag$$1qAX!OYRaIy{NIe4ZM=oksgnd4u) zuS5StEN5ftZ}vA;x|!I8$x+!qxyucG;Rvt%B1h3Nm4;*|6BUQ0VL}&{n~LlcQ_YfdF?_C$q1(AowR4>b_eIE z;B8Zn_&r>Yk$%daT?pWNEwMd|fRHR>#sc6w?9j$&ov z)A050(i}`DK_N>q;_kP*CVIC%6Za7!A12H3(q%|#FPz&qF8&Lypi#~qGn;v9Km3ni zcJj&STX=xr6E{dLRo?1rdDx$fMvO4#!tl2A)q+ia``r>ff z!Mj$&eKr46S?+t(udLvgK+(@(ZGy!-$26!1zm|1&3XCN?SdfdoMw@#J7e~-1b$$mk zj3()|D?G^Ye&BlgFow;*oK4vq?(xuWo)3MI^B2gfVoQ@5V5zAw8KkODUuis(DMk5)0 zjPev3dZ(h|o}L@&KMRH1U;RHBH%hCF=!zs0qSoT^nF;Pg>P@p%GA$(BK4H0QJ3XG^ z_C&%xyjm7(o_qMiojW{2Yhrp)2;CPvOC}wBAj7$bW^Xs=i)NBKh>E`-65icnFWR7S zZ?}hQQBHds3Iqo#m$9tqV$DRH;8sAp+ETzn3&+C2n+nf_heyTr>XC5-sQa8iglGZ^ zKj3M})hs9IYyW_@UMi2GpySipb9PTW#JcE@WP+nh3El-vAsu$7a61GjqD{x?jF@i*PN~_$A=lqo#!u z%^YFDg_Hl+Kmycsz4Sx(z;I$$rorYlo$sCizWL41%eje%9!`(*{a4!0UP5YHUMhJg z-4riDuUkd0OY0bx3@iN`Z`_|RRs7ZYmLS=*N{zI*N2E1!qgC>WfpAwRt_bO2)7^{K zj(dUZyw(4^cf$kggeo^TZfN_Gdf)jevM0_D@`j8`8aw}r%ZEge{QigOPg<6g8#B?v zR3YwScUxH$oS&y!6QfhIbzfO7k2~KVwr?2v$qCvR09(r4EF5#$(Ck^Ui2G1QzJqhAq6M^rO{~sx28=gt2h+*5m6(-B z%A(1zo9eIj!FRIm=4)`&(ptq|X86UQ8jvJW0Lv^uJ{&*W`)lxecr~oP9y?jQgAtQ6 zP|n-)FEChB=18k$MN$?#{fW@ig8pWw(}$2`zIiQvrB~4YABbgrsq1XxDHnSB@DI^r zHMh+keSFg0!MQ&G?AXL);OZ{r7tM zqe{Cos%Cb$H?!%?Ky+M39Q|%{>y4=xr8*9Wd|mDh<#c!e1D;QmP6meoEqa#5{| zfS0sxkKqR`SD*KS*jXHvUPcow3Ha;?0Rp@YNO4EJph_=}37rJ|bi{80ycgB8x{@8P z+G+`S2+Xai7?}u;8r^D9X8d?i)L0`Ek(ZvQRG$9qNVA5Z8e(B%&$dE(tx#QV!R)Yt2Zs_giT^s`h5A#77Q;fPNhnbzp?So6b^@p@wPyX1Dw(( zkIB}G31pLt&ctuh=14?;KOXe0+$GDoYGh26OO`gF-}`VB=G$?rIyYzfP}?xCX2TQs zd^$H+uOVdYbH=DZF;KY;%T#F)@Q2I$=qnb{L}oi9yd8S zoSPpE8&+H?WXV`j_76>^lc0&gT}$0*6{&YQG$D@GBAog_LdQql`~WQ{+jKgXhHJN~ z1VUEy?!}+8)}*Bra~^d+h&z2GwJ80_-clD|E`*ShaV52?Y$7>^;E-mKtp$k20aL~8 z;yAYrx%<&pCRqu3xMtg(-q7_7Z0t%hs*c4+C?l}~gQtL4;Fhr8Zl3vgvMbm>?jQPu zJ@Qm%^kt$#gB#y~tS<<|4$A|(CC2`6mhTopjNQ5twKrAS-LBplDNl~av-A90PuSZ| zRJWGOv}(>Uwg6j)Q(~bJEY`;rZSg~t$=GyPT5&2W2TjV7CF~Wv2(J)Ae~lJ2seEsf zP#61vSw5UJpCTG>hV;}mMw(^u^I{ztasoC-4k2!8F*;SEJU%`_UGYN0Jek>n1?lwe zeHZO8Uyx8y!JGVl)IQYiIW6wxTOuznRv4yR=|A8>sxl=?XL_;Q;f=4=lG>m;8ao;$1v~9jDUs*yw{3d@>ZJAo_B!5z2VR4?)(D(@Aa~bF4`~Cxl zsz2w56%uTl7f+$7fiP>JQKeh7dUC#xSO@zTlj~7z*PVLRgL^P5$^eI?B7qN^MBUw- zLG!fx&npRL%!0R09@>8!CcAvzXu^3`_m8tzSiM-s`2M(VEqsdINVu;3NX9@Pss|1x zle`T!*jI(TTzVcRlPH%gWh68xfSt3q9$xWE^z+I;itqvV5n6)lsxl{F#HPzcp9t|4^WbdAU?E=_eLS&ciaSLh511Pb3~*XY*u3@Sa^?>5aZ*Sjh) zq0z7KrIY_I``&F;(4Gea7OL4Cix9dp?W<;JAu1eLCCGk8gc96ka~|Cw1|y82Ztlnxu@lh;pI11!6gs#ysPp% zVN2{n-rEt?xHqFsKZmlmYJTMrn3mA>ZE^!Gq~3c2REV|e zHD(>bP%kPZx`v*8y9145AArDK%_=HD_4)m-IGX_iangWstxtKXiS(QARiKOjVsK=; zWt}sC5JPRKU04e8`GyEZ;y4`Q&Vs^9G7j)o$5Y$aG#(MO7A0DvQoshiOZUqFT0j)A zY?+CtARn^t8+byU0;8VhL1|2BvY8t!MokcM+NjS{&zhe4DFG3>>^J=?Hh)RH-pHqf z)@(8ASm#O?8deG56>}tyPU6oZjqr@H?G+~E>kb1|aZ2R`nCOuFBQUb&v&*WlSx`7< zakaAlMotdZQ?QK*5%ph^co$^#HPLTUyx<9!nEhXmdo8M(q+&Q@q@{HB+r;Xi*~ozw zEzk==f)$sFAVWmsOoOwI)9Y@kGn$eY ze}57mz7bKHxDii}?Pc_kEQ-T?W13A{&l6hfQQI(&#%F1z>m3H}jFpU8ID<^ubDrvU$^; zyb^vjWngI;Np>jYj+p^qLi-!rw`2F95~YJ-j#Bb)NgZ>MJo{Pc#VzTD;sieS5D@^z9*y^j);o8N6X8h zf*=bMNon^tSi>b|wayEr7*>ty5y|v2ayal9Sul21Nr9+zkb$h3t5jt0vM;YC{cS=|GEmkpz_w&6q5PUkfrA%0@a#3&u<@!GxxVo7%^R zhPNPeP@0Cfa1J;FIP>4Kg~~{VOlCVd8Ryy(nakI3untb1q?SGO%+%%>lAt*0 zR0JGnj z7&js~9MW|(sYKM;YXr}mO>EBJCsI~f zZ5%BU{C&^N@rgCO^*Bone6o`l(PsMurlsB&bcrzCwr5RvM|(bQm+*%20d9dwd#fE znOH8*7*fjwWuv9m41o$uL&A{*0O=vv5DdO}cC2$9&i5B-=Mg<})zyCBPhi8s#bN|X)%bZuG zpZ=Hzu`EF5(ig zRePyxId_GTlK(dF4fBa^_-7riRR?J7^%@bR%*Drqmsjdr^djt_$gg>vDC|LP!|tHI zCbeBukINY&D>9v@3YhB}q)&58PNCgl7iplpD|=!Hnu*04rLYKp4n_ym^1Vgz?Htg-s&U=+i6mJHN1+9 zH@q@V1Ny47OdKhpK`ObKZHi*95UiVY5(Z41qKvO+SXs&kl|WEO%~wfJm*$tC7rg`O zqNUgBPOE?0-bC8pMm11C;3z?$>@so1LTr*9Mk|Q!*M53jStm)(esYN)qts|vukryc z!cP{Bv;j_Uu&wGsRwm;rF{;J|tBDnekZ4PNj=s4wyJY_xo4Qvpju^+-?^G;Qx1V}c z)R6bz80*aQ74c*Q_+A=BdtOgUG6|3-YX?NyFLw(miR9IbI4l5HlZ|b`d8z1d;k+Hv z%~=@mJq~&~R=dzERKzBCOgmYZ|Htn7A2I@@;{unLRl~;Qgt}dngz>dHy?m~|3`a2lqpbmgJ#D<-qNITnaFPEnvWPNUD zF67vw;9)WO$WWv&U;bjC=6lzd;+{7XN**BK+!$Xy(xRaJ)aG?)ip+UU#zO%&S4f~{ zHb+;gX&eHADtV)=N9pxyD>oowv@*qR<#7<#KWN;)7{9ku~aTzDn<8K^!)!bW7+oFtyqB z;2|PmZ^jw;eVn)+6@P>VN90CFxuo68C8?$pPP2n1d@O>#UvEx#W#NJSL=16 z^4KOX26`7Wg75Mq^7PC-MP_QP_G$YIQJW7y)GlIoA7$86Q?U}NX=LSp?^^^)N;@Ey zbQMzv$&_hCe8pp!lA;x&L=(uZJCr!*P?aT%kZkME6fdFmqIM}}WDGcp_0j?)9YFqK z9LbuEH^9c{p<0WxQsT@rdCO-#D?HZM?+7%+%6i|xl0FxD%4{9D4Hfe`0%p+390Jn0 z!N|_9p9{s}R6&EE4c_3co|U(nvRY+Vz9pWi7o3rDvQA*YurrfF_)=x7BG;5A7`;F? z%<=v&*_~Wb@no15l?+ev3uW;@^oyWvfuZ^<(LlBRL1gD;k?ADy_hrEbz8y1%a3I;z`j37lK*w zJXE|$)Ily-ARLiGYytHwfDc#1jtMK5PVklp-<`1Abvu8a+qsrI@ydJ^G4%4ic#>A~ zp;qa(GC3KSb%>P2A3=)17pMB^sC-}R-2A3>N!Az=cA|A<+#cN4{S_5s(WkgZDfl_3 zT^31*V#9#8!t!tQindB`jq(!=L}$&n1BVP`X!H==_5km1*s@?83uj9DCf<@6WdG_o zFpG%LrO;A2r<^Q^#Yb^cfdMG*-;lz>P9EBf4H55*AJk~B{N6WxOyx*N=LX+*S7}sw zZB3RXnEBN^3$!ml5vB8_-fzWUNOUL4R)12if?|){^*j?}|96=bLiyRjJc<6rNtATU ziK;vV3T0aB$1SmoWA~QF;DYtlY&Be0+7puCTziN0ZQL;JcrA3GAyDrmL9xk)()zz* z@#F#^OvEmE!(&sC}@o;Pc(;8CmH#9@d{Ic;?`CryraQ?8zcX69JJ%caI#VA zDNLv(z|W6^NAkrDJFJ@v#>EC>S#E>Y))I^um12G2inBP*@5UP`c{_o)L4%seoTdlmD+^ptC5ke zpt1A{J%C{+QQwLVu)VrSMv)4T`D?a#za+1;%8@UePGeiXr!0R1g!$k5wj%eLq5o=Y zMzS!$zjIVq4=%7*V4G9EwrKt3KPUFC$BpC3CZw?I&o%Nus%ZU3DJRlK-HYkl?DR1mA9BTESXIPiDKld_ z+unp6qKzzMFhC6ae$wM>=c$d#E9IKk^NTN}U`$YkdNIIiX2%P7N_$u|z;pY_B6Qoc z6zf9-2en809a}5dP2(lHq}@v(weYIG&#l@==0EEaZ0%`8^F^e-e|#ZSbrg)QX`liU z@+<;xRb=LAXh;`~!o&v|9U*qA1||f@jKf&1B7rqeRjg<`eQ)YD;N*@EEVqdJPVWQR zAT~b+q7!^8glP^LWOOLl^Qn<60mBQmK^2stg}<_qT#W(G@1cLL%BB2UGW (FC57 z=|da@G+4??l3md@l4KKO;Y*~ot zg7?OV>bHFK9?w2XHK?yIM_Jf}CxpS>FB8q(#5V5Y`u>N|ykeN1Uxl1%b5Y}?P#43O zsIW+B^1ek^7?f0ieB*7;{=vW0Vv$X#t}wI$SCoy#;`Xgbb(@sHzzp1DwQVmka3W2m z0d!ZrWjjXZxnXbysOh?qwt%Fdx)#thT>v+%Us>8Gz*SmJTG7 z`&dwl+Ku9`F{V2D;Lg3He%E&9KLrJ;ef;$=8FumFh9jGq{ce2PH-UG}(pK7(eNI>T zFIuR^JufHT+sZS5-F6v=U%XL#+nq{N%B!;}GJRA!#YOj#IgZ+*0pV93yFcP1rPrtP z^`h2mm?B+$CV;j0z;MbtQwT3B9y2XmHaW$X2Fj>}BDzb~KuF0TY<7PsxTn|5`c>|c9QmT5sPN@rBF8>DdzGU0H()B28>B>uvW0;gSQq&;TZP;46bP!rO^ zSb#@I3=2vQyP)dqRjD3#A58rTrN==V$_Wcj`$QhvZu1t;gQJWYHYeU28V#`<^J;g( zKMTiSqGgKg*Pz4ZgGJ`+CNR&an7#%c_Dn`NbLTr3I4&wYq}Qv}onf6nnIA%7^F z?JF2rdIg-r7;m>IUO?dgH{B0~HY+iNyEd>8z8Mh=_(qL&Rr#^ihQCw2Qzjw~o`?Pc zLAJSL#6d3WoEVJ7NVsmfgc8YOWwm5w?Bd%w5`Wq~7q3-HzU6NmHY>O^{=tUUMtuTJ zmt(%)D?`S-=gQYOG}e%eI8iu`ehfG~6HTgmTF91uS%4W()(!B9so$WZj6YL#0Hh7- zh}{9y_{`)cv@w)aEtvoGG-hG)gbgHR8uS*4Y)k#n;_=go2jyf%aeETLV0N!}rO$C9 z=<1`7xi#lfj#ZLEG0-CVgWUgkWr@2jOT5_U{Z$%DxoHwS=GKCuA zRM|EYn>0(j8f?4;-!*8W428QdX*QH&Nx=SwBL@rR zqKKI!784c9(^{d5=GRH$-QG`i^DWv{SzQxvXyg`A=mTGefkyTsZr0?tE+klEj$}FE zm0!jvSx7`38zq%GwYd==X-4t606@2$>#|RGVKuUVUvuoEe@ecy6a(qz5%JdW1^-?B zW~PN|P=)9n`ddo7P^a3vz<@Z{z}^_EQjfU!1m}kPCK_#BA0eso$SAo_5Jl}rHxarJ zquxo)nc*v{Z2-=lXk@hA?^PYFnILYWy2i?gl&8wLe3CN1k72ZjVhKqZfL`kk1|T5C zpobqTH5K8_w;4^zD}=c9CnsHbR9=%!zBV#Mm96tf!YXf2%7dxGX)z^JSjbI4GI;QizP4>k{y^2)nCYlC(r{Tn+UDAIjDegy65MhEb^`^f$Y7ub#~&d<}7yEu=M z%+e`-Z03|m89Tv?slcztt_+t{x2o1F6l7Mi?d{~1;g9v?FVx#_(zO!V1e6@#zX*3P zed<;xGH!8z+4kZPb9C)}uRXpK<5E&?mtObJFc+KD7U#>1+^2fJF??(e^(nh;sqbx} zrUw3G;cKTOvcDP;0w6N6C{dYF6(HhOfQTk+G|S(akMh;6(4?pGjv9zuh~5Vh+VRvV zV|f^%+8iwd+TD3_O#9kd&%=l1zXco`UyEqA0B%Py`XzV+4L_A5C|1 zyo$D|V0yejL+q+&d|!{@rU9ilq3yhQ>ozH%MCJ$;>Fs;<^y}9Wb%Y}{)$v|#1SlWumSM>as$Pn(1a-J&m*^dQR=4^; z-;e{C^k6UaDxTgu+jh8Xf9C}oMs4Ebz3dl1PK@V0oKZAuB|>Y`j40`d*a$6JPSQ~1 zCn!TdBJF~F{#km%?&Y<9iW|*7=W<2@Y+AU`h7caXu*g*0_D`Ud&~_)xC+0ii2Us!x zYB*Q=cz>f3KopTmSeeIM{;LGnP#V;CX@)23Dpd{YqOoBuWaKG(mY_RR__&3mAm3SJ zp@Y@9T$SBF!-D?fqn0Uk(c5t`B!@JLSAdZ6LO9jBm0%T0Eu>Y^k@K^?AVkJsVe7(r zU;SG3?ikg;Kg=wa!spO{ zzeK;>>izEa0)D{X|M=9?!st60dZbOrJ@1^!fAy$00edftCX%HZBVR5m|MV~^<)Hdp zb28U_-e01r1R=qPZF0Ed&QDzJN zN7G$*1>%W0l-lUz_QO4OxixH^w;qCmTRR!lkl8ykovp*l+PcsV#v<&P^A*czxBcbTN;5HAU=CeT%@e;IRt>&Etq~q( z4W%Y7X0)|liXU8di=;thzV`CXWLB@iJL9RHb*K?MK*AlfQo!hq6ZR!L>qTBFt*jkI z@TNRkWn0r0bhD)t=p_NE(|;rL_FK8 zS^_!ubS7{LRygTtgA^H0j*2b=3$)ig=(+y`)3sz!4=Rm`ZN{10RD((y|4D3`Ad%FtJ406!G|FN6uxoHX1dO+i*+n;23mLw1l7qp={yQT&LNd9E0nnBmb4>hPO(BIh6_GmR#10s!H?j#3-rlFL zazS4mWogQqwtsylkCLSQT^H##`Kq~6dZQPvvz*n&ffOevb8)J7l*k1C2Nwh`IayQs zp%$!mKEW#9s`DX$K`+)_lec(4H_!M6ekGYAyIO>v$Bbd!;ua=G9@%tlBB$>KxtqSV zXwMmFr1lW`lLt3PDcE-nK6f8D){IyDg}1=)icWKsCT6gJ3J5NMdc(mvtX$xva9}*I zmoE?x=>|DS*?B~@3e#s6-=6QB{<{(~OoPl-Ps_N(#jNa ze0N#T;$3|)r}H+m6y~g(-k8gB)LPL0^Y~S6cQwG@s5si@13{9#%D=4NY3HJdHPj*} z@TS_xqNAVWR(-gCoIU?Dqdl`dU)j))0I@_H=CD#7VF1N)p6S9p4n?K*gt}iu5{_K(vmS?knnY9@}Riv=WD}KK4IjKp()2m~In_y%QqJgv1%bD{HL%@m5aW&Im6IMsjUljz0w|czDl&+ z-#Nl+WrlnCL&WnH?V8y%;CrvRTR@L>VjW4x5dq@+c{TG|-X{5#|LAQ_P_&%MkpV%S z26VMJ$}F5WRs5#Z^Vg6}1ewVuZ2>uvV};yI=agd&EvsKrdZ^X_) z!Ox80O$|y8B|IOd?dSR5aDRqVTZDhx>6p~vHPEd1LbVcjvDxfC5Un^C}HPk`p>r`n1DL`{X zYBOjGlG2M|bK|vRmB!0;t)*n4#)G7Obxf>Qw}Y|sAk3j1(bCF3{_nX+0-r@SaYwcC zu)vpo-%UT-wfeK=!z}71)l`m+x)@#A_nqGuOwXa;O)0AM9<+4yd#iD?7X|})%`$)$ zf{$@!vSzX$P`Jz#u?5y`C4`SOR9aU)PP*0Sf$-f$YFL-LtYGTsNB!PmWi{aj{^($W zqs-^bpgV2p{&{O^D_8Q}Rn(<_u;?{) z;VwZ1(v9qIDoKZSSE_MD@ZlggGsM35q5E;glvzYVd-av;Z$(s@(TYNEml1>-ZX}ud zA?8(|kq%~zAdvz@z*k-W3x0NFql;N(ne`5(lz0U=q-`oavgLtak;p6r(TfJ9t*qbC z%^7^NoiaAxvcN0fReH_K%dyGB|1iD`RbRQayiuc|_|z$T5~6-Ebypy*V{~1OY9cd~ zZ9hhQOGdLRHu50%DDnP;z{8n&0wt_(>12ov^*7$S8UBGB2kL(=3y6E1po}Qb!Lt!* znc9~XmV=`Ov@YEi`*O)-L2Q)88#IWxNwVW(Z6HMI2B@n){+zjDuJTY(`dZ2!xQxBg zoc%bG`N2w#5(L%FtS506xrx*0wC@io9Z=R63hazatwI7Q?8yN)1)utMcc_4t!%wV^ zC-W{w9mBA?27M-Vdnkyq$r=4Ya#$}U7>vnyTS_~J}n%A#V~m|S?Z*@yS&459e>vlkA^ZEqS) zCr+wPkFY`b6}wg5VJ4VAnL6oqlVmQZaK_-Ec)I@qsaeIw>@QBO@*UL5&jDA$$Lm0d z2nt{)TRIx73NLyCc~Z%Va075cb5C#*$W4=naHZR)At^Q~2m)Sx>sUH3%!O*~L1VGd zcZt%hNkzUb;!Dp}*Ps1C%Y6l9h#2}e8xQuF4+*!j#&43t)cpSMch5Mw_Ls8m~=Z@{VzSDG+qyXf6MKs zHUZ)4AJ@@^3LtnmLsm7i)AtGn;fUm0m3Fc7RJhH~WP~5HY;))~a*JW}Zgc3md#NP@ zi!c|3v5$rf0yO|fL$hYdFj(2aAFRY$YY^UbT6~GrbYr5BSr9FxxPngRN?*N;PX%_Q z=Gq_`)mlgf^t*QS(@P%VH@}_JB z$VeLJ>Wh7t8{95uZ(|90@F5(2e~3wgMJpAXmHu)9S@eI?#f#|x2O+CR5ue!}n4$Zx z95D-GyZNB{AS1Co*>;G&`HWHACKw_6i?p2oj+d~~{(}cZ-b{h*-W$7^Aiv1}PQge% z*Vah?C|hF0kocKiI(FZvPE#zgN#L90&p^b0DrYKEs7iP(sJQ{S9-buQUWritCL5LB zw9qA^Co$is_k{j|Y0QqJ-F!RswFCaL`Y<1?{{j>FR^+7&)@u&BVhL^hT z@(bL)RC30a@byZp!)B^C471@lDcFUtUDRLzk+0ln8f3Ouh+1o}?)EJ)EcgH}UQH)t z8a}$5n9Y9q^=Z>cG<}(=o`UFO>6g?v%_JW!PLl%!HIUlVn&Wk%qFGd=>=6_Q%xn|o z+>QO8150Z}!>hGecMTSZ5HV19mg4-ZKos#yoQmVEVN4G@<_7|}@l@l(sJn8hT?wCm zImT!zfEm+CdU!D#wzOKzCF(GdV^%ci9vx<*!VZlzX2D5BlQ?591ocKtLLYj;tNuRN`md$ zhk4;8VuPEWZhpwXSkFC8`yalY+8@kI1p)J|9C9K`)UQRgDZRmjk4!8GWQ6yMU}xAb z)GR;HtTgiYx!qMkWdEQ5bo$t=;2?j0&W>)IE+hop3l*s{;LGZzUGh4}Q0D zq@AD^Wl73cO2`W^FHMqJPY(Y=P@EXXQY#0`mFBxT|lkx#M7}p-K0TUNrdFfKsR%#-riB>006M@zBDQ(LtAbM21$UdGvKQlT_|u> zt#JFq5-W~c9f(6bMA<*wtWJlxVtX&Ho|am_8+OzH}}$OvIZykzY$JH zI?oJ*yifQ2o|#M4zFlx%1BdopbGuM+yrO6= zO7UPWmG_w_$!9pV;SQ^cZ4Q~PF^LtqN7@f7>RI_BmO=l_9ia#HH4hv^G;-fzmP}Am zi0f1J0`Z@ZKb1+%fb_tPfudDM>cTo~>=o+`W6DP;`q~Y>b8C?zQU$TJP2V>Y%b}Wc zV*YUVkTs~FHv7mEu&)H-C#wrltNY>UG@KW@85a?p{t<;XKgD}+>@Ei9i~Y)pAem@h@D?vIBlB_t2nnO-ytYJ zf7k4_;5!oLo$i3Kahcyw83*u^`w{oxVD&s{8FAF=>mlkRn@J$eRcVjSI^n7m3< zJ}hzIRSCld1n>7|NapyqB_1w5{^f)O)wD&f+hstplD#0w>}$UC|P z7Jf^N{|5}l7aMuo5|7-_zx#oQb85M*YX&+TzYHn{ryDgdk-T`9ZSe1r8<73-r|@jg z>D*9k6-*foh+T2*xpr1^`u~1zg(x4!(hT6yJFQ!HYR%O2r&eTeR(U*e$0W6XWt!lI z)^LblvNPqFc%$~J4@+VV93Epq<&_&BJ!4RJIncji-P2^bMz8IW!?jTM7YJq~4^+GO z^yeu6JXY95__1@p_O$#pyx?<|x2! z3mhW?IR>gFu4MF&qPIwkY2`Lco7?BN6nDH3yJ{SFJ}xsP`&n@&eklnbC6?Y*?vmS0 zbzdY?ZsA>TY14GNE7>XI@^O7LFWV8{JMr$-*u7wVA|cUuEQ+gk zdT0h;u8Vf{Z5lpPr!=dNEO$Xpa|D+cs%^Wh5fpZFvijDb4cZ`(xT!=&{ z$_v}OFX_+GLn@pe9A;Gc6s~YbT(+EA?cFXdKh!V3hMpQO!7l|H5 zt!G%6^9&(825x!VciGDgpFrR0lj7d`;C|}YFUoS|SCwaC>Kc#t?NfUmPS+oFTH@_} zEz@pdBX04?3N*eYyoQ${%WdCz?4hy1h~PSy`D8Uk%J4shH-}*fWL$BMxlji9-|)nT zcVmgs3HDpEv)li9L2tlZ`rJ+XB{vgpc%%O>l8rUqp}AphBnGRo(xdfS|2|iPYcbxU zxqmu9d{F;Kd^k$VYOaukLUF!Li^_fazV;?3K+`dzRMkBl$`NlkL;?9#%+9xPh}7UT zc*=90xTn+M$Euh51+s6Btiz__w(E(bAvc_H zmctu}kL(?K!o@ko{VF1LO?CDCRP#tvRih~<6q~bkJ+%Q}{yd)-LY~kZOrS7Cl$e|% zqwX>TJhO1pGPCNdW8}cIrt{m$>lmObPreMB+7^tq4i%h-rrn_}Jdh@B2M;SUc39)6 zt_FF?rk{;fEdJ8k`uz{uD}H!?!qF^}U&NGttbJ$rN1mxe^^hib(WT=_%Fns4980&X z&cP*<(tDB!i)tz5W*TTeP}M(aL zFi|!GEK-``|Hzff|GUldqfI-*Hk44&CQGhC5dIWulX7REPVkM~)y6pGo{(&*e!;tHouKP2#K`3#uLqr>`BcchoG`ZV?(GjNXZ8r;?bB5S?- zcng|p{%Be2lGMIwYm~MLne&^h*zY7dEn>O#54dWf4v44qGoJh-x)uWomzjq(%?)$5 z-&1O}L+~keL>iHqH?TM(d31WX)wk6pkH zmW`7K;{6QE%HpD}Bp)GFUoBwelL|=L(&3MFz}N9CsSg9ae04L9#&E|bB0m^a3K`sk zcSj+bOo+|==Z`?_$j=xX~faPr*m`}YW>gWt!hJO3e&9#~&l zZIQa-9tEB{o_ipY9uT8#RcqNtYnsRY+~}`*EBRY8>k?hzxodO4OdB@S7dwc6*NGDI zMcx5{86N>HxH*+X6h)#lSAQzDP?7;P3<>yJu02*$ zXQ&uf`F=9dy_Ll7$8A#G6`u(qVN&^Um;DN<$(NjbUnC${tf}ksak!qfwYcW&G2>v; zJj$3aY^A8uQe-Cb`uGk0x8pEt*)iHmOICzi?0Jx@Ir+;2L(xib#Q3OP^~k-1G>TQy zWQJu{<`)_PejEM*(x9X}+}PnhDQ%7JPd>&kzQTC^0$@Tq#ua8OKXp*WyJTH?@4#U} zJvt9cvDnhcjLAMqNRiCePFI!;EpR+6wp?A&mn>4DJ(Jt=gC28D88{)oM$|YJ@S!s{ zKPQA_-qW9QHO$yFm+t6W0aR*D_&fKgpqfsRd-+V?r+^4i2*7pCE5H36RZkgX5ktv( zAS|8Eeeq`Kg1#u|%xeP{TXHNs?v_n|{_7;_yhiuHNDuL4H~~2t(B`>#&}Z0I=lh91 zP3JsnZXmwZkwF8nWprp48Jk->cmI-F(XN~>t|Dc)lo8?{mBm^-d8I1DqkYe8auu!e z#NtWvcI~tKE;h!CpkV6mJA*|i6+hi6L90s?Z_SxpxdC5%rtYkmpKEHdPb^-gbr4U~ z*<(i@Nx*3H?~eVepw*(}Mw_q6M23`{8Yj1@rvFT8F)5Z+>NLp)sjL?Q^KT*DZrvZs zsV?`*m9Md-)d0@=eU8t0q@wOdaStxO8=CK^XjKY5|NdkdQ@peo%$~N%NNldZ9}5cS zA1>ulg?D~_^2&P6R%LLR>07UHRPGFYqhHu|Q{|r@@b|+k4&7%h&S)sFR1Q6@)9ii> zeE&Gu`2$~36WysZbo6hH-y=cme$ti+XOK$nPg3>0nZCU>E}K>)=1g^u_apTkgoQo1 zC+@F^0iUNv&oWYZ@RKll_$KLioVSu1zTS3IuR%gY#>`LRh#OD zFs3UhJCIze!>#6=mUw?R0-K8=z&f?99u&85P_->1P#zIeY&k-GXk9}=fD$89X6yyhGc6iytcaX?hN&7OVD>7d!sIR zW>3w#eNP-YPkEW|SG`j&{-6r=mQ^xi-B7R9T)WKpBQx$Or?)YDhQ{;f5Y;bP@115s z0Lp2((?>7;GHy>53d5gzo$QjP`fAB=8F-}4B9Sm(5(Z+5wh+W?dMX{U=3L+IG!o0; z&F(l6@gm|6W4Y-l#9EcgXFbr`k>7O0+!hH^BjNeDi)w)mYc4k_WJ*~rPN$XoE79vnH6&@PPwtne{V!u zTUHwWuvv}`{}F1=4+nqcyihZJBzVG;P;b#K(D4u2ckYge_5>J$n&~HUUaEQ#8#{0I z$%|o5c}&b_RJu^JEc(GBQai|zzCd_YWg(Wflj-{WVJspJO>8#SZ2?!zc>L~gQ+4ffs zorwiEa9>?4#xfH_PW?D&=NxNp5U+fTKGj9EB^Y zru_s9-1#5eE_Z^1;hDzoYB zec)0QMe8t;AYoMD7>PD2^;`N4$n)(Je(UmF_=Z$|HLkegPxFs!47znua_AT_fQ#(cfBU?Glr_X&i8uTaO+hq2 zYpIBR5?1w$b?NnR@Q;UyKbtvV^O1)gDeMZY=Jx46Br1-$hZlRAQv~E|udXF-LPavN z?PmvS$u8hQnG0@gn?uQNReu(ic*HR@*^ zXRvfxj|ZY{zh__244#H%)($5WIL%CcCPiVZJZ`rHj~)6(O0s{y9F^|xuR8Y;K)W^m%E9bm5|>G zEzmf=E8U}QxFWGc0sp6kvwUFYuWuEv8lQQtV|K`+nvqZmXexhO+YlU)vpT!f5jp+x z7rv4*hEEvUEko1VzV16Ym{~2f+T1?11x80u*>1P=JR`YS(Sg0<8@Xr0}=Y@ICTZcbi<`dN|u+m}(hDs)m__aAyze7CnIaI$FNHI#?dMH#)ugh}O&KyJu(Z1s#?7O5(*C8)D3Eisth>-3(B9A+ z2|Yfzw8>Lzb~V>BIX_6>%7lzi zYljJ{FuIayOi_OhEn=VxjaO?=(cv3noofTM!>2yVK|2wY+bGk;L)O%lX?KJ9X&3}u z4_N^i@ul#}i=;svntyP>ToZc3+2PvvK@!A+S>v!}!PD7Ke=w~~rWJlCuzpxE3w1H^ zYk~4F84exN{!AvN<4sE8H7yf`@(rB_t1TLYK^!l^Ia#9({VjCpWYOhU4|-%aXPzIa$XE{5jt*q6+*}wjNv9;4_L4YJ zvysU%RQP8Z&A#*dp4jXkgY}3RG4)n^af2M|wL{V#AFKHT7>Y;4=MZQHhOb7OAo&BnHE-q~~RpL>5K&piEgPtVj;cfa*kb#=;q(1CH}f_EOL zQg`DvG048=(Js$hKd~Ql*(N_Q1!o)C+g_hR-+CX-EXU5^C=DDk8iYdr)8<3TP$L;k z_vc;ICb&f-Rh*-WGZAp@MDH}hfyMbx%9x}NN_S*;zH-v5BgeQ0b#oaQGS@?GJg#l4KB$$ebjT! zgB)fqGs(9%cy^hQZ|bD)UknT@txrn~HdhTkWxr}dqD4EIWdKrT2iKgD=5a`m5?nR# z4?4_Jo*VQa(`FfEc!wyL*t^r55!%8fac5Qsp3f9f7(FDCwKAjGBVqT9!+YXm+s=-@ zA_;Ck>`H?z`witDzMzV3u|>#qrYl(rd6k!TPtI2=3)E_)D-0YM(VAvDUu!RO<}LZ$ zsP@8zE<;rS{F;-oHWoh5eu1J#__I}G*2fA-<>#aeSNBa^m7~YvSgYE|q7STRfcY~n z?}kgN&v>!pHB{Ug@tZ#})NG<#irW#(fmqqZhT!wqa~iTk3u)XuPLP znc1_ZNsNx(6mfs5dWnAHEmY-&h$VK`pE;hXGnCt+95ko6BHO`0TT;!%rSH||8bE@Y zQWhH}G#(adeC8h(iE}`p^eNho`4Lh==>??`EUfNXZ+ALDY|hkkPV^&@aM_V&0l=H8 zd$#_+q3i2p^MJ`U`TTlO(&s%^Ay7!_LuN>2@9|gmxA$TIYp!U-uu&^|485|^12k36 z?_G>0j;)gV++dY_SJ;1)U>E5U4zQBi^wkLNax$Iqn4x=gPv4RVo{M>ocQQM|W3Laa zfZM+ZN~4#_4epWmp6K!f=V$yh!?Q@KlM%OnW<8^8g413$MVwFc-sD>s+?Z!rd6Gwc zSptDyly#|03e>ib*?7=X?S_bXrp|Z(!H#VT<~L1IO`{VGhV?*!voaLBuKpF?oKLZ; z=hi00EXLW6Zd>qhq_P1Pt8)O2*P$MP`MZOqN+;>D?|SE3+>Y>U^Oc9ZPKLzrSDl2o zKZ*kPJS76Pbd4dtkQ%`oTxZt*aK9^d+5NXi{P#DQ3>ll(z?fd%2$`~;^BDpsA*7i% z4vY4{GE}Z%hS79~bA^||wt5zn6;X0&6HKf)_PlzH_a`fUJw)No$NAbz$#8A~#mzSf zs(c=$`^nms>=92EY0mN#PHDzOUH_qb)H?Qj``kEnkFk$vHJ5}rmIuIwySmaQ&{o~9 z_(_l=7hKt(4OcA-hK?%}Yzb^xoZ2`nwbx&}CzFJ$q^M-*>A&8MRkBX4NLD2D(2Ztl zf1#1?bFg6O5I1i_x7zeT-naC{+=k>Q;4_>?+Q}lnf`U}_h6k)=bt-k;`(jfl0U)oX zVgwVp=)Jm5Q&>-Zog;ZwR!+YOh6YFIywKQPrCVT@OwSngmb-QL$a=~F)`(_VchaW6 zqEn5tM7N+}a5@2c^j+i=nz;|4qWHTr9_KGN00-d%q1u9HpF1{n9Jjo@(~WRnxW#{n z@G&Lu=TT#)ipT&~o!B5fx=~9{(;Z=oU`ytmh+- z8P*PeMVwy$0U9r!l;vkkM&m${SqiEM5!Ylk=?Z-Rth)&C;2p9v@4k zEy#!`@Wo2sEg{~#a^I~43A3z6x)Xw5fcQxP@wY)yPTI?z{lnqwyZ4c9vEn`iRSdFt zTlx>avN`Z%m1$LihD!G-^kA-XgDbkV9TIzl86J(%MW`z{D&E?Ms7xjKTAA}(pm&uV z(pm&gI~a-vva`{_9Q40=_zP6-I9CWu1?c7WpA0&ZKxk3Ds4>D6hilCIF$dbX?DY`r+hSu>o3^?}db%q5upM3%DUdRS*f0g`J zSj)cP(p3z$Kl4S-jSrAI5Bb-R_WCo7rAgP79Fc;J90nv zILZLa{qs{2G!+T4ah)(Ty6_E@8=&tC7OTXI_gGFcZRYMhwS2KK@rS#-A!zb>@iM-# zkmi`{Amh(JjO}hk6Z=!5aVfAlUE?}P6#Dy4iKu5w8?OcUq~}y2bAee>oV-%pJ137w zvfNLmRt7$qh~6asnP$ErsIc{XM=i{+LlXFFU^th|1Hpv{=Q`JI*-e_o8_Wg|?_-$eV2e1e0KvNNY&W4;CkT807->yFfB2z(qh$smO$I=|Hg))an-CWBzzWbRcS3@$NqR5XOG0pTq&gku{6Djg*FU;b`Px!#O9KZODXydkoz33EXmWC4tnx=P^TWRtKq~4aBFu~pp&mj@ufInad7EJU zwEV93b$*d|7vNiYcht2wzv%CZZ6UxhQylg}JNG?VC%l-u=+`d-Znws*L2qUJI?>5% z?UG$}-1$fDKcW^vboo1MhQ2c>>iovSS z$`hA-@WZLgy0b7X={J^FYaB$LrV9M(iF;Xh)=1DnHk_P8z=IkR{d=iB&7kEv zpUw^Mb~KNADsD?N#S9*2uG^V1CBxboCuV;M5gc}&ed6`Ze}8EzLx1nBp%WcP&n|e4 zNujm~#r=CwBQq*f9|-J7LCpBz&8yAGH3fYw`iRcNW$`w$LX<&_s<0=~OEVGE-#7A$ zx%jsAmWOp_YSzQ``jl754ZHEi0Z&uN5O@>MF|>M{KtS0g@IiIAMm>U}5}Z>NSfQhx zN}V{zLT8d6toeQ^Cq3&0ct6@@Rl2Uu@jNoy7&K2#zNAj1L?pUBF~A3ZOI>PyL-)_bVSN)QI!Hh*-=WRotff9>jEM^7KO{A&&F|&pK-I+hH#t&! zex=yhP22>b4yrC4PebVHAe+Zixj#L{cE#2dMVIZI)1ux9WiKS#C%hpu-(F-tCGRso z&!34Td=w|%sS&&Ow_U%E)|>-Xf43{`VII4j8HS>JyJ>6ugXVdYro9}Sb3i~GkHscB zicMrON9*e|q&ji6|a5tK|EAJi7oi^BJ?6JzNwn<^6B_|TgGQmWf_(eRa6HfM; zy01%vTSwDrM@TCkcXacSEJ14$duAfDrAFKp=&TooD`aMKbi5n<@WkR@DgPuC|dRNDC60&J!A+munTpqJg5?YqEyF!DL`nop= zg^ir|&n49e@m(f-IVb3u_-N*h+~}d+q-WS|itmApc6$rpYl?J43Bt)YU_$RJGm5`V z9+-S-Nxezml;YnbX9m1^8Aj=e7_nFKfVm}^_buTK>T*Jq=7+fXe6s$@bZ0L9MC@+E z-#ZSo`O(qXz_o}S=bHY4H;z|arR|I}&AeY*0zL^DD!?_197|%W0*d5h+#gSgPmp){ zC$cH(h?EcEpdRbo#`2#)I<=#z{&srdZ8!EjynuIz#C$?qx>j`#u03JZ<)2$+Y^vPF zi$Zol%d%V#>Tb}Qo*RrF$Ve-~YX*9Nu6yt{pqQ6U6is6RM_H~-I`F4x7+V0cvR5Sh zarF9yGN7i>jovf6g2$3e!lVA`?QIm&6EPOgp-BVo0IJ?5lOFz>l(>Q8g;L zJ@kx1I$`j0dnM`>N(*7@@%T>SW4!Zcd~DD6m%Ku^n5%B5RYB0qpvI?t{>u16py+wA+n>y~RL&W*(RB}l zykTjj4@_43O5kpRHS6hmY&6ze$AAC{nP-1|tNs~|0K?7CgMf>L8TNU9jw9(0dT3a# zl5{FWaHCl3jhYDoe4t`CsE?E#^I4z$S1u<44Ks9Q1iJNnjJ9AhNYy8meqJaaQ>U_% zHh0ApQ=Bly7V&RzLT0$iNJ4uw8HV8MzqbhQg82H|&}K-@^ywA`+3&3#_oW*SuvGVS zB#CceqPrEo2*P$5o0VF@<~B+hdI&5Bm&^!_rT;46jCLvG6`dnf$cOlYm)4KOPL`{p zq~9kULgsd*s3zJh)3Q3#Zvw_HXZ-b@?HEIg%^XNqh?;mzmXb1xO}0FK@5#rhYB^qr zDCJO7JJ}pt0jK*qL;J&m*5^JF!88*_^L}3c;f7VXX-!x(@eYn-yEZz}l z(-*(sXQ5wqR^{z~-O^LL*7>y6$7wiju5X~)WGa^8p|HeNJ{*^`=4Wzecr&v0hVVJK z^G{A(WQt!>GDI`At>AdLVonI}l$H#%g=I`OHbbT73Qds};?_;Z$)0Z*lt1)M#ycvf;H1>!xEp#S^-m;AuX| zu9}NE@}&v=+i}_xMWPL&xx*QSN$-RA2Z&6}qbAztXW^9nVVjF9JFX$(OT&0Y{SIQ^ zOjDv+Iy@xKV@vASm&Q=;k5&I1xYgYoUcQL5H_sf#-{D)Z#?p5980UMbp>??Y5|<7QJ++$s`rmS12e$K~K8&4KK!;u%CAFO0Fu1??&B%}^K zt8du$33VtfXNkP4=2yOITcE2-51iD)@u(A^Vw$8WXH1q<^uHKjLS&TiTByto}p z&b0dH>~cQZW0|x*(Nh*-8_}+TO^J%5@6-)%CA|+p*1Lo@w}s-9^iuwK7sMw~QR=RL z=3TjqPQz25z&m9J%iYuPc%;^dnEFS>uPBY-F&{_9jmRtlrwaoYWVp5Y&?432*qy*< ze#NxZ9Nt>h?oGGqR>XdFnX>ApdaBPek*C+~K9BBFo|8j%bM=TBh(-GAZp*naH)qVt zjs=3@Bg1yc=!hMO6hHsSYWCVI_s3{2g42S2ov=Stoq^vI~I?yNXup zvPz1bj0d&f8r$A%JzgsmT2>d;Ng|BGnVz!M@@~w1eM=PFP;V2ag;y72RCGAO(5az< zg>F&%NGy*vBS7AU;tL^v9<(-jp*Jzh2?cgI+SMXcyNy~Z8j8AOZ_j)N;u{-#yTqOT zqidNUNDR1}x3xuLIL$L2qBZ0h0sc&ukVf^sK^)MHBRg&?Psc#mE6#xjNVF%O9cYRx z2+ZEYyxu=yI+5VIP~BjVIs#fDV75dX{$QJ;PV(GWNLODe7^D|TezJH%H403-g`Th~ z(;|S{$~{Aghx*;Ha9F#b*dxJp zhPxqa(4R&~8bZFUS-)>NdCkZ_{Ak)4tF<%Y7~9PPjDSC`Ma)L0e3f+!pa7ePx!dot zuMnJb2Cj5%p|O!hMD~f@K}9oyt&tx;ZIEqf#^d+Pe-y5XfsAGg)tce1^1Bnbat^a_ zCskSLkjEU_b!M);MHM>OG>++Ulg$i@y}$GNqdr-fQu8QMpk|18bd5etuPi_KjmFYuih506&EF(C*Tj2a zPEIG0<_R5AHJ`jg>;Pca6gHgaO8XyZ-Fv2S7@s@#ss*NB@g0-!)>(fi%7pC{S)Fvw zvdmiuxT-^GwMz3K2qDc5PSv0f&J|^TpQE}n0y;cPR;dYgsadH1B~t@E(_jDe?e7n> zzCFka$_kZWOP_alYC&m8@!hhb;KXssv|k3DaxK(D@2JjwgDM?^270C%`?ObvCK^Ad zJYx3$Uf)vI9nidSv|DvVniJNLD)|I9>ovrTQk%U|eVLu&hu76#8PL!$TYgOQ^1~S8 zl?7Ft2?)R6>>C8v_nATQ2&mxwa$-C|G#$cvAOHO(#CVu9HcZlew@>khAr0$j@F5*Y zZ9nU>sT-}+0+x~wLW7-`cCQhN2E(;ykq=>`9a*D{jd{K)zWz?A8a(8=;SH)rSv=x> zi1I*;`#vIXo$Ex}Pb29&Pcti#B~v#VG~ftFmZmT5x&B(*8-thpGN=aSSfjMGti)uo z`N3)r@<8l>g^`1E$Yj~s`}Yj{yJB-a+x9BEY_QIdqw)3L4W4T_f+f)|ohxw9kaKgE zI{(bIvfs+M%4)wNr{Jpu?pMjGnZp*9`Xl}OomftOi`+pCngSl8uMX8U3e#lkghoJ= z+NegR0IiqRxwFf^h_!xw^cd~9U7lTfjsYc?S}vPRT;`hYRgjy8r!>NP$-uqwTa`nP zN!n@w-3JH7!v!uaWbPgU3Iz-C(^rwwRyiOec`9&jb$vTcYTHI3V zLjHi+Y`bZY?20oSI=f*}yFT;~8ZQ@dg258Z73EcKYc(uWPER(`HKyN`Bk-J$bIRNP zW>Q|wKBSS!%Q27b7ztClJ4`z`ckCjsqF?rL)O1$)>DW5rmUfuxt;t^ErhOswBKW{M zvk7USMch)Ynfkgp*=<3ofHk+pkv3y|A0t?}=v-?v?Nr9`_Y)%ROp!|ssiS+oW-#fp zD9b;O)ww?v&1m<8icX4Go?=BE#i?e2vYJH!VRPV1IDhuIa{Q$CXQC>aA(rx!1{>z0 z1$b?oDfVWLILj#iKB+GHOG7t}<~Je#khEZol2=tJo7()~nD?M7e3WOTCJWWM0a{%I zXVW&0Ybt6w-?-deDGmc^>Nuz7-zZwa&ynq8%NpaJ3k*Go)&2csRu4_zU%x1$V{~#b zw*Yg^10_4%aC{H6N-*T|FyE}yhgEl79M2+O7v+E0XQR-QqjZzRs1Nn8-qVj(kJz%> zwRc12)&0o7khp|&o{DkKyQ%KrCL21#Ljl{IT$xQ3_=+tmUZT^HQo0;;ZVlG>xVp^J zE;xzq>a`JY(9=t&WFwvb|P#y5mxgW$?%+fH` zisX?UfA&tsyRGwTbGMH0n(Szc3bRfts%Y)0tfog+(SUH#AM;e*l`|w=J*AXzpx+)M z$<7vFkA#(4bowHtjpc#^T_Qx5LWe--@tkmF51gdfoxiK254XFFAw3+q_*_wQJs>_n zV|6-$%W5A$Pg-q;JXGUkE5=x0Axk^F_3WB8 zTPG7RsN())hTO2q1|oJm!Cepm@0aZxW^iLCmPVj7E}4Ur!0zL*-lQ*3oRVdYe`%v5*R zd$NE20(BjRXkFWsN4&rN=}|ctII`6QYS4qCP!i@md2E;Ulu1RSp+iN=bC4EZC>m7a zIFzZ45ve05W-8qXG42;@Nf&t={zD1-x`9Agk?jJSSM(2TGEC0nGde8z=ZnO<$^~H~br`<6d>+mv1?=zu9K+(rF>RuiuWmDB1_P+0Rra z9!p~n(2`S{sUdp|rm6>w53UpE4s8`;VR7auVsf`d4b5D~uy2R#qzBn%@s82@pebV6 za*ho0uLsYv3OR*~jjOfUaBI>g8LK3<`L2*IBg?wU zS?_=Ar+8;vf`3>+ebzIYXOCu?(ua)y;{}J8?Y`%Xj>ra}`yAb|th!lb>vDY8`6_$d zDwmnMgeEb!rwovL<^d+z{wWf3B4g#@J0$9OlVgW{jIxGwrRejndG$umrF!?UzXDbr zg$B;ZNVD4(Me`YPIoPkX9SSE=*&x|No~}$YxqFM0<{oq5czZ&*fBvwmshb?V3LX~LN7kui#_j)zbK>CJ!v2Mrc@uxPZ=#Pc^0`b~zt>N#ul1x(qFlmui2JicU$g0) zBGR%QSxU>As+z|`2GC!VWM&WOu9u6(Dti*SE}!gqPx9^9wM;|NP!Hvw8+)?hIM7=< zZc?$Ux;p&3b9O2j8ZC*89l!qucW8$ac*b)hpI`N;Mc^VDN@gX(3|G<}bKJVdKkUzQ zOi{@S0-Gk%2?8sH8Z~bU?(rq?_C;J$Kh zCL?dI{zH00zsCUSMPK_OCaI|{lpvGE+!4E(8W+FaEvT|TZefp?>>6gwxTq*PNx!zt zHPm7(=?`d)sXC($z)7rKk~Pi^Y$6@(&Txcy2FC0_ZoNoGuiAbZ%%IF3tLl3aU(66V zm#1a<)J!36!8KE}cw+w5F3Ya+Lgt>1A=-Xaq-`J@wy$}b;B?1xijCpM?^oNp z8B>9+0Atvq=-uv5) z$^}_AS}iE(+E(mTHGjdOXy^m+fxS)+^(hD*GnS)%{K((#{{8#6|A`-2>we);@m>Hu z=uheEhT{bf>z4b;>sxtg&dcqB&*OMUb5c)3VsZi6eQ9C;ZCPU4EX`mh1LQ(_Kf(K|W}hmn zg;e$z1@d&^><7C&sIIu152}B6O4A936)UrSUs~_~Hgj&4uCpbxQSWw1jTk4VBPgsW zM0BZGcD307%cew>Tt6p-`+k!8jY2E)YClWM1Ew3#!likaK%OgJ;SVz0>HZ||R6oEe0n)u)CvKRhnhm*^u5{5`NP|Bpsa$+Mb>^y# za5#p|mP{c;48RFY6+i@~2hg7&EdH3r4d`cd7Kq6A`=XHK_XVxH?9&cH+-E0L^9z}@ z^D9{9%o%S<>Y?X=kV@Q!|il{%YNR5a^%Vpradywq6o%u)9rGX!Z9f8yCQx@mMzMEiG( zzwD93`<=bH_h5?P#HA+8jP>eXWds(?f*Dc7`_22Ttcu|1#s)8N?GZbE@IzLR7nB*e zFMkDcGhtawexuRXU|y&ZyNoc|aeLJN-pq*j-D&4#^rK60r5b5?gLG5MN4x#TIWo~1 z^HJo>g0{0Q1haP{Pxvm#$V>2W;ewN5xoAs_!-t9xWr4iNDgPPs~mkBk>fTtMSDdBqa8w!Opw? zUZlQLIrZ5d^l>ko21*Fesa_1i+4Tp+Y)&9ul6G@5`N>5KiOQ(wq)PF-)V#VjMN``w zvrb$aR6#>2RCW144MgQj%aY*i#4bGjK8U4dh(2_n@*HeN50Su1)@PgK5C1U%Iq?tV z%ZF)?QyFYM$%pK6G<5z&+6u$huo z35z0V3ddh2C;kYBs|YjG7x5cYmXU3(Wok_KPOQ(j4ox1q!B*>p25jJKkBiUVD zJgm-zN+xl>RdA=y7#+_NO6-%o*(o3eHbS=-e?O;_ht{Sr>#0%Grcis-z9^PlfG;3u zBZtu(-S|s??RO+vvh73XN0t4y)JI$$!1Wye5_~lg*U4Xv73fVRZ-UL}q#9*2P5)F< zB2-;I8P0K@o5qcs)v~77KtUm=Q*-l zHyt*5OiE6ijD<=Ul0N8}A>>u!tlW$Yxt5d*V^S-x+cdlC{hEm=_RPV!Eg6gb$s zD)YD9=XYcZAc;O+b*&AezyNK@xQ$&B|*(MB2io_suK*KsA_|N`dTpE8&G^^|lIV3vK%A z#F~8Q zg(%>ts0#}Bi=O%|@tA=vXf#n1=dCVkH`}|B*O)R-eTp|5$iU4?y{lS4H`43a{;YWk z>=(Do1t>j0wx?{z2Dhc&o<73M3TJCfq|)D2V!R|p zpQ-+!)w|n(a_^_er)Dk0Z!+Ig<(0jYuNg&{)+KEj2ow=e8V@?m(=N$ZkBL!uxTK}bkwI@B~Ig?Sk8-W7k=0s;~kF*|v@*=drYpzHJ^C#=1 z_2bWjNZB2cuJ4%@2v*WeZ@)3T^SV6M_%m!Ce00g zkW{OJf%qgDT}E^IfenjXxN= zzqqsvQ+?hX4ArRlw`ACIuhsPrY)karYJ&o0ZZmWkvf?Y+KV0O{xe0Hg(H-;{6aJ;7 zHuS$aptIxeL`i5hb-q4n-5HE;nI?3kc<&W+UP&9zHAnqY_QObcNg38Yt2W@u7?9EK|kAbvA7SM;mx{~O-S#+uCz`5(LP*N-bZ&jt5JFoT)*{B z;#NDba~)#t(tBAh{P;Uz$y>e@O4nSzb|p3#16eR~pss1hXBGrsHT@JKCnlg2MXhQb z9h%oeq8lnLF;r3B61z^VHT_`J)qvisy!;Xo80qA{=^{=QTgzQPqczp7V&9oYp>%RF@M(RtIAxuBr9HSx6(cpj0J9PULg-y^XICaR5oV z4MwyCY$WF?Ehd-n>Jh2tS#C39Si`Xk7P!;C;QWYxGn$W3@(`p- z2zM4OI3j&6>b267@7cVUwi1LRNjyTv=?%T-Ewq^oH)(GrLYdEiItv#{hi#{ug`bTw zR9U6Au1#)4o2eJNrg5e!60IyTfM6k!XURZgqSjyp(@3Jgu7)vRtf=`0>MXHmOFEhK z=+oOy;^-QjRi^BBC#hf$!}2{IA1vJ*{o-9XFHoKG=)DTaq)iR!O<_(-l2?lQdyocJ z9kcoyXjcjab{C2R8F1T^nG&368~EdY#k68*(HOy2VJcC67Z$nUBI^QV!qg&sj=%xN z!Uac{ZE@1G6rjr~MPAD>KVZPl#J0#{da_~W3a3F*H)M>2@hsxC>W^ZIZmb`hAINu1{RhGjPJBzaEgm6Z+ z5Y^w{vQAqXn!zmhxLOo|A-t|y4Q}iRj+{pNyp!&n< ztb+8R${FGhSSQl85tB<5#9de+CVoYBWMsfWh8}4GvT$qcScnJHvuT4qWr2y(vw1vL za><*Lnh}(5E&5Ft(e4`=q>GXwJ{2d#3Me*?0hYnPg)Hqgc8)H6o3T+nF54#i549rF zQMZsi6-y8|D)1%w2L{3XhT345Qa!*mQYb<|8c4-SA@aP#3udVaOAFA6zni74gUrln zuH%95S-j{jMzk`}Q(DyNvM`^NAwVTV(Hy(5`)`{TU|RNYU37pvVWMOP zlRL9ZIb!7OaU$fPY0VO3w3r~I1g$e1zZ)1m-XAw6DCJRxkHu3=UpWe-{L>$t|Hygj zpUClE&^OKINsZ*10Nthpbo*-q37~{4$rRY zo{d16vG#t)h`xZs`uBF(;%L!u>t{Is>Pwdp<&lOL^hguZ3sHag$&f^d=%<%4>|$p? z1C#sIN6MnT>tAk6c$37wDo{;E$?m4UDDkZ21bOxOZiC6Ch9zYHEaZl-qd4pSiQxI# zZ1PsUlu0;tEnpR;huebccC#Z?K!gV13C30vgq9sj_k5BZe9D%ITKQVbxkj3Q>Z#q7 ze(>odurK$r5Y>$`zEfv{mhx9VL*e*u)ii>TYC<`5`p4uD=k~bd{V5;IpB}==u?zrr z)WM?!7-w`qBYAzTkCQsc>Dpj1@@l{5@CRl}R>ZCjPoy3~n%gbP9FKrfE}8{E=BPf* zZgZwJYFACA%BgMOF%r6FRuCyrYZ^edrs(583N&sfh&fKLgejS*vjMDzbu4k714r3| zR@{aq>HvN?Cp^HmGWByRkAbYr1h5V@)t`v*41G19fDK~}B)ETImD!>0dxkn=Br0vV ziPD$fGRK{Gtt0sw( zJHLL*Fq4&>gfLi(TSr!xNoYDdN&AD7HTyoE*3dBOfujfvKRQ_)n(vQpvB3f9gCzD+XFB z5dl-i6in$-Y1~y9uyclRNx+{)(oYd4Njv6SgqqUZj(AhyDi)#!F2s_aeNwf~gTpMl ze5YdV>UKfP_+X;0?2CV$>Bg}T#W6fD8Y&+8lN5qra#@z3+9 zg32-=1?Ix}|3&_t6UN2*B%wD-j0Mqx112egoP3^_pw-omeg!P4WP0I$*Xu)LVWkiF zPt0j{&B^XrTrLTooF2p4F`ouhneyHS6I!0O3?i*Bt{0j{p&czObeiT11u{v$W!{on z-!(=?w`W$HXXaX)9c*uuyXV$RS$?xKpJbrHcO=MbF>WEZm>lt6B9^zbsB=vccRlbM z?qDT$jJR$4H#TU}CM?$Fe=r6Og?}%6SQ8efnQRBg^FqkKR066XMWxw=43=e-oL-4i@XG7dH4Bl= z>N^w0l7^^mb&(oTc~F~8z}P}kt*R{NVyH&tgLw>u+geJXBv)3Y9Eq_qP3L^+{vi;9 zNA7wBBAEBak-17WNa;cd8 zHf(z7<9y#pmoopy8JVrNd1E82{T7+~NA`e$W;cR4c$lWz4dJUCoxLewp;>1OF=?u; zB-48|oFqEBHU!i<3HuMvQF}X8V})kNA8FY8Y>PLMT<*_C!o{cPADfwsMo!0qTc zG_Ez2XrcMcXf}Ekh5%Ax21MB zM-Iu0s9orjYcri9G_3b~LM)A_@|4|<=B4tSh&l-}wUSNb8SWu(!T;)aa;F|8mjUxJ zSx_oXvb6bd%U5)vS;+<|`Q_ABss7sdP|DzEEtW((C6xdP2W7KQQy-c@gA}<7{O?Jv(9ImxkFiVA~3)272T#pRHDrYHOihG%7-FN}N)ql?jca$#Z_Pq!! z=RN>sp@{d@X)5%hf~g`*_uQEX*MJ;}BDo-JzKO3f1(@tf5+}2l8u0}g~EII1* z{n#;doiHOo6y*fSCQ^B4l0F@=gZ3j))IC1b#orEU5x&PA(1^!iF!NnD`&r_pQq+B* z>+vIyTR@E*5;{MJPLOsIid9O|?>q@XGaw$MVby*U5~v#2*-^m|vXlhr@j}MTWh6gH zsZ`wv6S_Dz-AL8>piz}YQP0Gf7KMK%e77aF*QNOfKB`A zT>Z2SrXDxgtbhd9s|Z^F3mov+2cf4P0OSXy5@!Gf@EHpzkh(Mt#7gXoG#?u&EE$L< zs?gWC)z7g*mNXD2383^s0#yp+LL9iV`8d0BYtDhp$Ax+AbalRhn(sqUQ|*_0OlhFQ!%l9wbj<5xyGp=?P+C zEeVdU4ed_=YMMo&9IJN=l!%BxwUDb@Ar5jsbXB%MxrjEHLC(>O|Aj-quI9_z2YHqO zmo~-?kd{=*cosyN=@hbOqJdFI2T&moe1QtqZ@WqmE^%RAp8C3?3$wW^{sb=SQijQn z>VjqT#kqEx0th2&iM5|biyDwHiE%1riJAR+piGr?di>rpPogEsuU;~(Fq;|jY6&|Z zkmLt_ss9Z965XvZQdXSlv}0&AgTAwxoBf*e{Wfe6^47JPm+5)FnJXNqWbGKd(={E4 zk&xogCa0epY%?zgt7>7)F26BD64wVaxM*;KS&{XycFt6Z>65=|X)uGX=Kr;oM!e(! zm8j80c(nF@RZ!>W^G5w`?5aD`Y4#fi<|IMhCJ{ugRsU%pu^|pX^~nILH^A_+p(b?u zj?unEOMVfYYH%FpiT}+dum>Wf+fj4Vl9cB$Lp zEklq|EXI45lI%Lv~t5e z*Ybnm4w@r803es^5t=pQ6B7hrded7+ckwOWm0B@{5|_y&%vrYY>EE@rz{~I@mNy~D zK;Lim9ao{Ug*_j>htKA-0F{K;g8~#e4gKDsr54sM3pbfMDw}DZ$UapvK5N6Z3EyCt zBnUMv!-JY$I_UOP%FSDt*MB8(KrRo93jl=yyP5SRX)&-w$~jS zy7@K5KIec`Dy|Z0pO5ngn)BkbM`fM|&6D9HQjdp+ua7S1tkA(Km6DA`PDfG%tq?%8 zo+HV1$y(iIp8;u<@J}bg5-=Yk$KW6gTC#Qh^CmKG733ZwBQ(F4sIMWGG{e(G=D5O2 z_)bewpN?QgBj~HJLEvgN!}R1~(Xq3V5uBPZ-CJ)5Q~sQrY=Q$gen696HeYRo;N!5w z>+(Ltew9%CW$l}dcf}q5O~zaaoy^NQ?#)R&n@+JU?@O(K8F+?N3}Q-(?m8D{gk9>r z#jbg#(jwOk%ePBejy}b1cLQXFs;m{-tMd%=J_erSJ{|abrs7?oDzk-LJ4}D^LC-Zq zpO39@1M2^TeGtA_npDvvG2A#WTt!!}}@vI)lzH(_hV%y29_P-3Y= zFf}nLbRhXV@v`)U+@1PbfJLu}?e-76)sl|kdXke0`~|Q%csNt}a>|hyi=!%1GA1x< zJ($XbYkz4rIV%@Q-ySD%;-@OpPz+p8qDag}57&t>A@V5`%TI4fEpDK)n0yLZl9qRU zRX$4{N75IfAKg!lZ#VD~-z%)cf1}$eRiqioC*bT-;ded`;Ji29*6^OqrB9=mpk)S? z_dhXhF1r|S@Jf9)YmbR{{lwfPTIIe!l#U(tM&BzYQIp&JT0@6d=|S`okL7|a*Qwn6 z+T-!L)|!v>{;TtQE0|8=2RSM492H28gx&b&sg~y)l^=#8%1ovbgcFgx8EjsM@W=oy zCg{DtUu@E2Mj0}kE3A1_gt&645Tv3u=39NI;wF+?wO2XreJbeVtlTDMjiGS0vL7xv zYTOU%Lr;fTDxr~)lW9fjIz&6y~%p}vAt_#id~?Bz?1yzy0n z1c(rSDKXtC^{f*&o8|zE0uVn3GwAIRDIIq{?!M8!hri6AeeY#_#g#FjTBdhXE^6 zj@PissbI{Y+72)UkY!UMN}N4*Xn9lH#SZhJJ68%Qpo|7?g+8jp<&hl(32HDV40m+T zxO4fI;-zIdJjG*KuY9w?PW~l7zP=mK-J!2R!&`&(kn?`|)ycC-Rp$8n zvuNqv)AP95 z<@uQFZLk>z3{D=T8JIJ1Ka26tfmbz@Nv$rXN%u@O3kMGJjAt7w5yXjblbT=TTzI+R z?UD+>G(x-A*%`3kJrV!Lr*xon{APU60e7ax@+UM~d2^pzs!TgL;VZ* zjh8M^iyXp`x9|l=KKr{@QoYw`-UvdqF;n1}xtI>#0}Wqc-Q8S<$Lr^v{zI5aH`975 zcoj31uET+sKh}pyOlJd*HqB$-jr1uS%2rcGAovJag< zJ=*+rghpIpV}442iH!thhSq|T_foKQ^uyh2H%6&$$|+8Lz(5YVynA;=PoXi0=}}Ey z{V!D`pwcvah)ai{6+IX@-sGW&oaK7i;y)w}d1>OYv^eE)}~ua1kNd*7CBkxpqA zSh`Cj7o=;E?vzwg8c8YXmRfr0?iA^elCDQ`rKFp8_`bjY_H*{k%$b?{+;LskoU=9H zC@luWD0Z7+@ibPs%vL%_r3l(c)?@K&+w>r{LahLAqX0RyjAj#PK-livLmKQB|3ASS zIYCHMCm+bncrfiJG;jZXk;*NL99tlK4^=D%fl&Y|@zZImTnQ}W*w2#*q4_K0%TYYA zJ?PHNad3sO9D5nYQX*?HO=0t)On4gF!Pju!MWT;Plv?;7sZrAkbEJ+Hd6Rtl>`_T} zj95$40k|Z<0Om2Jf}mA&r^{4)Sk3AhrmV3Sr}lMu!9m zVp#}D3o*AcWWCnPARafG`1(;J!G`eL$U~l1-0Ksxl(%Kcxo-Dl8d7+XBa@=r-P|Du z?+BgW`{dT+W_G9; zY$RxbZ7PNYMy6@KmpZAXI%}M=xeL%_!8I{g^QTeXHls2{jGq-j7{BpDloIDo{_{0< zb9FhuIi83V8@?OQy{n>fAE|EYkx2--g8?|ye1mZNcQlWv*$9A?{PBJ)F=?!L*1^@^ z`khEe@v#@~zH(+lg~LvzE(?+xIsl>3yn{a0h@>DfgWZnyk{~4K z0!JIq6QOXw^Jjo6Kwi(LdtfGeqpYhadKyPl7;WP!w3XnO=R7+AA<@3)p;hB(_$=HY zf~(wdl#g{y?pw!ma^4yD;5+|cEumV|gCr4K=Re0^L6wS7gr>O&AK4$@>0{FI@4*4& z(<-$3kfU8C0m+Q7rXQtwOAf`=k44dH7!UynO`MGMk#aL5gRTQV#C}<+VL2K0>ciT& zr2$j9X&9=@Uw16vUX}^mFzz%oyeSJ#-))eHK7xU5z00ZXRbMw54}WZK`Z@76RUKI~ zgXA+Lt~;V$>2H4bs`^IFPI^J5m7dx!_IlMFy;f>#-{;$JUa*s%)bh-4M$|vQ}J6 zd{oR;pC7sX?#cJcso48kG)3Y|hVrB0fAEA)9`T=2RMG=3V1z!mRSpI5Si?Uv&Rq<0 zNags;$w}ln>lL$yGp9ab6!AbM5Y2a9KD91#$`au0sQCw1cz7LO;j-dhjJVqBL6WIa z+mrQq2r-VB_VoDqwMK}ZHaT#$!a&kEYsjW^RGadGB^VeZMM(DbQ-@QTg0*6vpe4I~ zat0LPouucvO;!SUyibcZnHZ6;l`!fi-EXUwOo_%lZ;jKY`kdA=N`2R)e}z9MCWr>L zw$m#51ACr)VA{knx#lwe^kfHnk*mm%Q$esEi~2n`37)rppv;Fcaj@rU}A2 z&Rjnpb@r{GkR-H!(pHW?vaKw(R9V2S&N9n!FD7+yRyhL{3BT2&)J@Y)+nnPcvH8ClJ_Us>zovzip$TC>E{RhC^)$+MRW0}L zc_-HiocPDna<%u#fOb!a24E*TW;W|avu{jDJ#ud(OHPgQ0JG0>)j?>sa`K$9{o7N& zxqiNj&1|Qi{&e@>v8$ARWewEbM7-$+AYq<^9@9*L(A2!OVKsjJS*jlSv-NR_!gg#8u}>-UVZnUD~S1Oz4& zVV!4#9cQOW`~TwZlx%V7A7Z+gS(X9D<1osfEmL&SvpgG+m>_Dn(NbQoX<@3BKC$U`wa_KH{;F8tfoHnTJz<~Gf) z`8bz&>#*t5brTC$fCamF`ntS^1xx6$)fv3vd~!1@1gbyMswZl@v}X8x!g=qKLY){v zr!^%YLpb34wU~P*?YAs2xwGoAYP^EKt8JQf`o-Lv@}^?jBpCQG%;P^lP7zZK3e)pc zXuz-8OZJa2w~|(?%$G_Z-V4;9IuMS3wYe;D_}m(fy=Z>0K53M$y^tcn!SW$*cvow} z(OQ|XGTO#7<+;O%uE*)`hRd92KJPLY?6@aQJihCsfJz9~P(0H$Z0%LRaFsNxp?%#c zl1c7QH$kQvy+sqBAD?

aT3hHe-}snm017!o5|3zcIMOET+Ym^v=5D-K`#YGX4c2t_~dI}!$}ZE?a*!f};` zt3TV6c%AQY0i=1RYO1|IH_FCGfO{&+&W0So@*=dhHQ1=`_m#aYp_J0hi?OkC!VvyD zg>E^)sTgcLW5!f=sL-60!f}y=%(kf zro;#tQb=UVt!9t^@@#bOnbV+<)Dkm27~xfZZC9SRrt|YJV7)FRv`!?mcgRQea{ljI z#5=4XS97KJabBOg^c1K?Vb2^mk{sExNqwYd?)qEH;ShBBECahN6$My`>qZ-k_gbZb zcnhy1D?aobp_ihMe$>6z^*x<qFlQoj*>~Lb{<)@#n z@BrD1wb$^^wwrRYroVjIg)=&%Yelq!j=Ys>+i!D*B-F8hydVs+N4ROA{Zthzlvwv9oZ%g-$hkZ^o+Hv+Y zkewOw&hXRKn#nUA3j;*WrrvDkW|HY??E1a6`f8!KOxU~UN#&hnz=f{-ro|j@BWS6{ z2G`S=jpL%!5zD$I$=rAJ13lG~^%PtYU%Y?Mm(B13*(d!W4)%rg^|mj9?`RaEzVJReM1m_!CFy zhCSgk+cyI!&*CF3MTiDvK;O1JrwTJ+-2lBeZFTe%m1f-p0z}88{8#TB`3C4}C>%^T6lGZ$mBMJSE9zc3^nXY7AAm*zQ zHs=gTB7L<#sD!N!1AgkVSSt>@U4Ca^xI+k50y|>9vH0-wp$gAG|6D2J1Td7!^i_rI z;EC^Js}99Kfm&m53Foj&gE`YaqCW=bV!5wqtqd5NpE~d@>M`MC**|W~w>~T2o?*TF&Wo|v`#+RMtJ|{{I~Q*h<>?Et$*fRSD$xBT*q;y#4|hRfBf|I zXeyU|5fMCwp?iC8j&30{kHp22VVFCwK|jDP@?F{fd$d;Zzv}P1mOuSKw~rQ&$T^Z@ zMlEIKnh|!z4)Yr3pih*w^Wy0pf20Bx>+zF@XpCE0imeH~@lv=K8@XFZKOFwB+G^x; zWy%~L5`+%1V1zj25bQ0HjbWxOBLOM|-awJb@1yAFiS)n(=MP@7^}X)u-m6~D5T`CH z@fpy@$hi$E**aRp&nl_57+5z|S^wO&W`6_T+0P>Rz8o7^b5(-oH0BFCI^3~{gRGJu zJyrzwW_7#o>mVPex6gYGqx?P+Cwt7(Vf(0U;Ff5K3LC(eqz+_ z!N(l^j;f`85mWN=%{aRHFnLK%;&OhO66egqmH2^+g)9HeButO|Mm*$ybLFI*#?=t- z-ogCH|3&xC&oh`BUf1q?=a_j;bQtMz1nJ)^j9A`*^rykK8^~^B<&$4kXM4MDyW7M| zix{IV2C~i}2)FB0w9z@kzz~GIH!JY4eR6p54uUKFD%mvQ=rt-NL2r|cSegJ;I@VMZ zRRR_A3RPMI6{4{PBO5T9+-=_(Yqz-^o>)6vEGaqL@!q+*do$^`lY0@^d91}AaK+e} zb@JFmU{Bn6xu-Ox0N=CJBAI%M{pKDvwvs*@VDQgox`6a1(uB3#QG9Ty-~TmpQNn{X z?N-{#^x^fe5bm}%j?NLXjk6KLKW|zsMX;yeEytr&Vd!)?7hEp+lX85eWft%GpWUFy zYnLUzjt>v}=0pdOThmOx;;5!!Lg=f_f)pB@IzL#Q`;x`>t9Mv1@PjUzpw_VqoplsH$Ooe#Zyx;zwNFyH(AyrTG9fUyrac6UFoSZr_0A|H?(`Ka0a% z21>h@Dj%HAKMK^HdeR<6ciT*TkrX?AdTTt`v6|LNaaTth?(NsEaB&yc(t+hWOS`52 zV7`D@yZm=Z(03_ndOo&&QT?5E%YS%z#VIB16hysh@+5T9^t1WSVL7~O@A3AKW@vyj zS=b!^+}`FQV*A6u-{q_>L`06kXmz!Mu6KtIUdFZguX(XD-TLj0D(?WnMd9?(4_muq z`&SZQ=9{NYpZEgKOw1Q-ArcA!6C*zUM0aS3K7ykU9^+@_!vUdQR>kp=>opED-hx;i zOMA9^IXNe9wqhH{1PiO4+{LHP10S=?>X6buf04_l(ao1H8d-h{vmmjr$JX&>cL{zK za{rLityNV9*QzbKaCVMN96xkR^_gK;|pV9Of^_n7tCl4 zpY?U0?S2%j9w9WdpmQOTNzAQ;i{TqHzcVIu(`nALcHR%GJDX8S>o=Y=KTU2&FmB)u zBI|g>v66}prw9aZhI%q#XT{W&vEsnwGl3=fho-R`IY3-BksFtAUvN=?3DF zMo!G2N@;tM(yCde{za?WD2M-x)8mXTpPQvtTt1&LNN|bkb1Z+d8|Y)~d!~o(5#+}B z=L3+oB!ek4n_MSQe}l%1`P>S$a)E6HIANFZLF?7?k4f@B>W58OZg&#%N;J}pZxlS; zASPRVmz(fvE-aByR>@fbgmEZ!FO=&>^{*-mpVxM-@;8#!crNm%?)fm5G5TZ@ihlo+F4pGC-b$ZgvIObV7bGe?fR)tCHfljiacGG;E^U1I zRD@`ke81JVv#BIJNz$6f%-TK5BR>RI*|#zsA#6b3&W<5CeX%q95HsBwW^S=tgK~jP z!Aw{BS6qmO9;rN{F7)R`9A0!VCS@?TxE}ovo-VTOoU0cbuPByCPN|DsKjrHEXtioo zC!I$6V?5&TDlCQ1*U}(m6aR&P4@VoV9gQarmrI7!3{0jhLW~EE?qcTRuadf;Ct~qZ z&-aYSN>b(f&{Xlyt`4;^PewRY<;S&ia*_v?84`t2C`aGtG}o=lMU8MCa(R&J(LIID z+vwSES{L3YE1bKKqbFwQ(9|T~WRIQ%@lP5C`SWaC=^2u3b+&zdS+xLgLl1F_yP}UM?_b5_S!Rt zR$8EVz2a2?QQ~7Y0tt^`lnW$24|raF(kv2D8dqUzEaG9~_-6HUf_S*F0$9m#3Z@-M z=sU-sKIEJ7MTz>WXI>5?5nguh7G*ZI42W|2K2F6@mw!IKt>sKl`4g@ zb&l(JTS&7tRxx(PKJCiY;l1-8Ek@z40a7KLv-Tmt8I6pu(>##${um*?yn3DplsBuL zPQUY2EW1OW!^hBZjpvo>>I|lhsUwdd@#3QVcHd-E3`;yOJ~!Hn0-=PAKO7>-3NhxyK4n514~2vjbKQ6R(VE<&i-8lRLfhr!9%u!( zivhba+kg=-eD}$xxi{h;r`|2dUm37M`hYPQ7wv~-hIZ4HRc9F#L?679b3Nqt4e3Ds zR!_bm#kSw?+j%(TQ89rQr9-ak`0eGFXrmsZxxGQiXRu2R^ z?~#bS$bUz!kN-D8Z99|iXvvV#di-wAGiUF}7@3Qn$l?BbM3t}HtO%2{ZSECMipWan zryAF_U%MCLyL3*qyl){O^bZja@(Va~FNN*zadMcAEW4Y6#;DU$^3P>s>4@0(aD9i- z*R+S^h&?uOa%jhEpVs6+-+q>BU<~!EeLzDH@VTvUVQu4L&}8QyxCRgyd*dvHAXvF^ z9wcDf@(DN+Ds?isKLR$YZMMKCkSOBxCyT#>*1FJdeQku_Yn)q%II$1HVnVRnGdDsn zeq>{UKhpBO1RuOQ3N7f_jjt`J={40l?xN3kcFMW?G z6RBU?r62iKhhD}=u5!5)OQ3xI2Dp!JX5h567e{GJ)EH#t5ils($vfO-hFKv#bC}7Y zE~rt}0uP*i!ox!xE5%cmosyHE&tF)c6B_=R^|exTf}yOuHme;97`H0kINkp3HzY9^ zeVks%J6e@(J|vzn;QY=&SnxthjqimFVXJJ{p`W~J2o$7EICb zG9yP3nW-8VKC{Ti;!Vj?6?+HOvsWtiIEvcDDxW#&h*g&a6V(pZ1bll7f5(G_(q}2B zL%E=Mb1=KA5Jk=riUHt8ZTHS8R807U^Jf}kA8`2($>UyyU7^c}%E~gzWypfmswd+| zDPZSMe4~G9_?E0O4T5ifb?xZ{G&|fJl|Z;G{sJq(sjKn7`m~v5NZJUJQ+a}GWMq-wq{yq&KL{ti)v@M9ALmL&dIve7&2BA-Vz2^|NzMah zah+Ki3tiVQ;<4H9o(SGKQ<%u!MxTkpjj;aVJJ9YN_sSBid5kJ}K`8)VQ#xDq>R|61 z*q;7V(?vWJGvHJ_aKM|5KVmwi?rQB0h2KsDXZFwR-DVYRyL(P#k1@iHVJQmdbnlyoX1xZ5)mpjWa z_G;KRT3m)J6t(K)_wQ}+#{?WVoK(9pEJNi_P7tB21^EZU1~ihLT26WEPodVta`AU_ zJMKR50USs41$4owc_UslBlLuI$l8^q7%$p~epSVHMzNVTQn0(y-|^_wkiBRQj1@D_ zGEcw}4JunB7en8>Z**l#`%f`6gOPtQIwt=>`jC;Rq|$bu6J_e`FK69TpxrD$ zR~&Hr6L9c0bZV$t^%c_EV`sd(p`3le$1837wJ!4NEuR}+*?eG#$Ly8T?JGToZ4?Ig zTxZu%MRjW}LmqA(QF)M$M>wmSKuxSlVLNhT$a63&_bUJ9w#$`Fy7GpP;070z-95$d zlHp-KRAV^d$L>6U)r_`KY{>h%;m3P>6sMziW~NFgPS9%AjtBPspJ$7uC>!1JHw zg}=#@OVbW0UejG#we*!S;0Zk77PT^(=c>3w%V`D~Wh(Ntc_X9hx#ttUq^p+2VXpYy zf>st!oL3H^k!|dDFTn1)L)2qst2*btBFM^ELPi5KN<6~MgXUJ}`?{_G*#6L_weIG{-oqo%0 zn?9$#76x^lVWLp&%A+WK>07wSF`|$ula|U+nN(q(UrON(v8@b-Npa>_yqx!f#3|a) zW?XN#A|J-<&oWE5Ys~%Vf1iLO{_bPnsrBD!=A3#UVSY{^J@4z6eGntcc-GEF@bBxN zF0N>Q9?nS3zLv(&80$4kxYH-}`EC`~75dO)<5kr$L9}qQT-n`0ZTv{AykMrei;D$m zWlynNLSs`=@TRC;(cKz-xTw`G@MlUEIn>DW-6Dxi%404~`orWd0nF(j30^hi>T12* zKc5JN(l$ap9>2Ba(&v|&D)CdwN@JhhST4vibPn~;7@*n5>k5PigYw|2w#7qXjX?1G z&|)`3aQ*`9C?>OV-0%rAW=>XfKxy-JvjD5{T3IE$^KdM-Q8;tyA9#x{5`w~HhY6;k zGmk}OaYW{#q?1<9>*K-Pf6%0g0+LMdH8^Z6RyoDZ-ft;uVnxq4tL&fV5UDrh52GC= zoH7@9SXs&+`<#298&JS8!STI1RFTiwsdWW0pD>rsMA<)rW;GN|1vDoqh;P;=16kX( zFGoaD#V_9);CoDb2nDNI10X$bA9svm-6W0O%x4`!ZZ&Zg9?{j$gmAt4=cZ1Qvv_Yz znwMnb+WcnY_=)(2X6U0G`M?_)ir!N$poS_hrF!>7^VJ>gJl*|*tCm(ABuDGbG*Rod z*N>(`)XZOHHXY$NDd;b(ll`J=Z}T5RZgE%8n1o5dh3QxHgTyoLu#{P93j6;^IAW4h z233B&5>6)57#s>!WGVBYCj!-LjEI`M=F-n>6}}v};;crqYWRl6DWh}uC4+&i zDS0EbvIla9ByHS1Z-8gm9^Xo7nVQdPS+t?P@j-E->_N(RCWFesg}*TuIurrC>bN`w zjk)WFD5C|>V4~kQcNN-G*O&$&*ncdG^I5qC-KQimvoP@@FdvJnB@J=9-dokm8fs3W z9?IMZ9LPt++wzo{*jzjYI9MZ7K*$8pNUC(gRDM9GgkK0H(9;96B_&#rHVU_@ih`7x zX^#?`p>eYX8S+R+m`D?uQR+Xw@opG;Xrgzl{O`bT#jTH~2cnO;6a=nkRX+`T6I^!7 z3T*wbUCL~p$mUGe&59#^r(s?gM%@F|r$&CGfKEoWuhCZX>!Xdc1@^m;ohN@t8~Ugv zBHTco8f^s2GvZXZ^p^iVGZp>yF3@wkffJy|!vDOs~Jz4ZL*?2}Pv(S#9vfxFL zaK7>Mby)yTvOlKftDhx!SKp!2I_KOy>L0~p@E^Rn=ZAWI*tuj)IEv+k2C;7@xz9@R zE$-)|)b{8VTE|eiT5}mFdlY9}^kdQAVFr@Tv zM@~x`-N)B_U1@ZVM9-OKucYH@+-p26$ZPE!AP$1BX7#_dP-K1`kp1OW>N>+6=PL7! zeZcuIP?dK{T4dblxKRBQjQ>0IN_FT}%sS184dn~fLVO#l=6E*pV=bDSiaCwhPBN>^ z79)`$pGcJs<1RV=*MYMP3L#9TLNS;dQNgV^c^Xj;Zg+rUfTA@xjad;>sqS(9R7c8l3I(Q zf+hj5`Gr!mTt@yYO$_)w|5d`Q=)j(F^L;^X86SRn>k|H$C#K-tV8zcde`R=w_EXR= zg?{=%Q;%(FNRL22bD+P7&Lj0aCFVB0>Jei(Ir8ifvse8|H_4-C0*)5{dPDTs++>uW z_hCK5l;1WbWs_^R&D?6g>ZNS)f#_*DRX4vtRpYFgs`0tP;S__eKk^&ufKW##J^cT@ zDMOkJswA-OQ`qr_S<#VD&d2hH)?VZcE_$Hs^vQ;5$HgWOyEzKHfYW}7aK_HKOGs`U4wPInKKcXSis&qMJZ z*(+^S_-qS6Q)EMN-_HGO`U(AT8)H?{{XAkXX39ma+*e55sle3DVH`cz#)00`b*Mn% z4k2oIIk!2Y9i72&<@G(xnMpriIWW9TWYa+p5Lz+Ra=8Y+yJ+9=99+yb!O^_$2`zn3 z?3+sdKMY{@iE8-p{kEs9{dzGZV2~64-7)^c_wzdfkc7Q!4=nT)^-jX-KR@0GgjC&z z&LL93IwyWAHv)SvY)H8;ml<2r;tLj*{*6IJd^Lp7fXG17^dpGsKhmf)F!LB+@OdttThyK08rT}dr~XxcROVt{7Jv*$Tr1oBXWAmbN!;r3kZ%4AE3|r0 zQjDazYulLe3^zOL+$0j^i3Ui_JTy8xMWw9EfG)Wx58SG{t39sfnuU3X{fp0EWGrNq z-|c>10=n-9S|Kd0{##5#zf(1GaJjqG$z|#oF5fx~C;%^gTzjdg6F?Vkwacy7gO1=X$j9n?LCPM7&Gmt9M$P5|;adQrqxwEY`cs=rAFfJGD(TRqIm zb;+Q*xPkoQ|4=NZBwFrh4`K3^6Z+~h3FFvJ`VXK-QO4XE)#0{oN5$JvjrsU8>9pmr z5t|PhK3lp}9ygtZPh{B=Te`-)_oXZWeXMOEHo&x?Y&4+qSX4n1{_DiL8{?-Y%-l{y zm?(Cw%TYF4I_g9w{KSv7pJ3`VyM0A08QX@4`ZQ7+y*9Y|lnf-~jY5dJR;jrAX?ynF zn2Pr$zk=YY?6r~nOpS03 zqoNwYc4D>oZ@;{>iqq{Zu0<2L;kDsPsx0M$m4Rp@4Sc9VW6!m5ltt zY#vyXvqv&3zt(G1U~AZv;n-z36kp1G#8G&nak)q(h#QxC0I{k+@_q8!+hS$kp)pmZ z)n5KN!%w<7dLT^~G~*C#{G(9Z9%!c%vI@QWaph9yw<4eNTe@quPJhL;Tb|JsD-<24 zyEUct;Is4#U;`}UbwLhFBx&8eK1|I0kTQ8Qax+x#Tg=kcay@%g=rM+*HEPaGk@Faz zu;^EKlfO%mOP}}HEk#tz&bq+!f!Z4R;xYz2bC@gW=|D`6rioh09?)WZ#ZhsPbjQo* zmIP=!RXvnf5bJe3n$4)I5!j;G`XOa)7nmEod@XwC8nZ_z9O{4HWBAcNclK@OG=dxc zXHyf|#%y=ylI45RUY2ny`{$Bpq)uXBKdwdi9dMi*(fo(cuE>Hj~P!$GZ(=h4jMsvX=9gu~_Up^EFFXX$8|kd+hQ<`M8b zw$hsRXUNqR|!1bSOSEV%F8o3cyW{k^Yu}X&_r^joGy7+!O(-^_V z+CH;Zha^%Smg$jSZ8UyiLOrmTI9$yYuqNcihy{5=D{a(0eu`})>o=S zoEtpxzk>L(J3X~3faz#_$(?*fcE)t!ck*@zm0$Ew`{1%map@4_qS?$$rA|vI;-m()kc7ZGZSTaz?T@(ioD)H z=Oj|niuj*VY=Kf@@3&-3dZ68uEhpR@?^yUyiKg+iTLoT#;(LF}gu&F_REqJM1ROfn z4WS(-#N%58(s30=oOt#or)QB?J(FExC@CUFVF=o;9(l)G02Zg!j9^Um5pP=)sA9nt3BDUX@<#YlG|4JKoNa+%__o$T; zSoLi^<1dez6OpTgmm;jUh4IHf3`0#3p;o;6zxch>T~+rR}WWBgvU9g^cZ1WmbFpWzj)p5smt zf-s(bGhE(_CsmIXhM-q55S3)&Xix#YOk)*yyPGFW6hM@>_A0;a%xL$I8S-9A)KlN* ztK9PF?NH8?UL~xl_&f#TRuzZUML+Oc?Z_9IU^+KuA=M2(g@lb5$pVz7_@AKnpN4;{0*FsU&^_Hd zS|nxMV1^4bn6$ zWAxcrk}3juhoL?7Z;W&!8KowO0u~Z|Qc*M+ns%eJZOWy`S4U{Z|619FD4) zl@XZ2X!xNX?RT>`!$7un zmCKoQ`Kzxwrw`{?GK`FCk;e2h&J(K#P)Q$F_5H9~;~?kxtISGisQAP?Hwm5EvvF2| zqc=2YTSw84tt`AjXj^?K8NoT_l-5bM`Y%3O@tFs*84ozm)}s3Z2+`-|Ep>VS9O{--6W05PX7Gav#s zxlOaZs+YGZ62Fi#b6{kOjeWZh&+t7ba(S6fv$bmcDADcIyES+Az zw0%L8hzn{AyNyb6g^-Nxa8?pY0jP`SNB;)l_;$W3#wAUOyf2@}9bf zfiDWQhn3t-+hG(Cs4yg?vekfpj=P_I0IG|ZB@cAxehS~id}byVY)GTp*82JW&GZih z3ZA76^|wM-3!C77L5;r79!X2%9c=*l@a6*--TK)R5y0DUA^jahfx86X{W1OTlYHl< zfwTJ8n%F{fovWa^8pPm6i9pBMihTEOcg8cj%}5tG;cUg~`u8UYAl(_8m$-j?<2imj zCI$1(!{ob?he_8WhU8=)o_@Hd%Y#&e@-G1n`jIi(6(@sd@k`r@ zQxrQ5*O#}LLC<{&plFxI@m-j^yrUe+4j+FEL7Vg1Ow$cvhHe_oz=hFD1+Dt5i0tt* zfOT;sQ-6^B51dBISlS{IYCIG-*S?!s4pn`ibsql1%a>56(gaRgf`5FP7km3vMJ{2u zaioy62&pmd`M0OFR|gZ|;;2ore5H|i z=)iDCI{r#1pjCn@nCF;0a3D*K%tBY5f6VD>t|Yb-I{qiE{rFlu+RUx0C8)O z$5&Qgst%1M@yyTUZsWwbMHz;t#n(rg2$%Dy3+L=j70$!m@yZ%;OX?T<5 z=?_vL=j~B%$>IdgLYF&3$YZj--Q99oQ}_3)m$3&7{>$2s0I3VKQwc-j-^&_Z->(F4 z9MCgFp2R~U+s$rp8Hlp>1ZL4s{XpWMj~R$GU*_ciAfq`AjcoZ5*UuU2}vAb7$8HA_A${%xSY_$U_>OwsoOabtg&Sn79JhniDFpYc& zY8?l{4NF;3ar-J~=8@;sYsM`Z7^qZgb53e{9rg;xL<%%cz zzH;nKs-prip_RAH{tSSxEOX1IZ&a1RpYj%pHf5hfr;<$-+nf&GZ!SFXyktY5flG$@ z$kj#G_*f&nvBx*vs8R(qE|g5b3g~;3JMmLm{6CT5az)$8DEy_Ex*Jzlmt;Wd##=pETQzH#T5UzewSUD zr(Q^-`zsVWh*v!3&v4<4ruxnp4m4i!&17W7FX>9>zFV!x!zzmd?-7+DEG#@5UkDUs zWI=OF;A<*MI_vzx343m^kjOPAE!`Y-_$*6d_^gAniZ?iEY&9O7(V7Zn-o@9BX|mce z&`NRHaekLFW~2;ni*KdUV&_G;ZWOHC;6kneo7BDYdGd5|#9q&37`DI(3THlHpudN5ww>>gw9cOjazH zfHQQqXmO@zA%6aNQARc*>?T|UjOs4T+~7SZ<4zCW!GHB>!B$_?Oo2Y5HLft+U*k~Q z*S}#d?%X%Bi4UoWnh=D~>m90+?kIZ*tgGh29-%>bB$sS%V}_c+^FgLDJ~G@{_5&f2 z<|WUf0$^8p-x4Cj#q_ZaC?i=N((THaRCah;!&K5aq*fsy*QwyFUDj-ofl({Gq&@O= zA)Tr#7RqTE5NuA>z61jH8Dt`WY>#H_{Ub6BD>Gq&`tP&oiI;T0E5&TX$p$RVRn7nV zFL=jpbBS5#EykKMWl%-XF`&OVQp(@L&4y3+Fr>mMH{Hg{gz z_2TDCx^G9v_Dc1?MW_AzguaICk^85PWixbN4Khn5Ejm9>z-;#l?&zEtk8EqH)r$E& z*{_=U0v28iLEmZQGUVy>Qq?AqGL@+3SpVr&4k{67DT~ z2f5}@+Kso1*3I`r-I=WcB|~BR;Z1%}+P=Z{dF6X14OOKsTOv4|CS-4L2+m$GTIDC{ z?tQ10Ff!dOZs0ni)!am)e(^E7euW2KTlmJD)0y!`zYfkFG@8r2&fz-e270VxH?1>X zYt^3po#kg&2L5~P%OV$ck4|0<<88!5ZG8z&h%S+1!)cmbruyv|(uc#G};yqLC z7m%97w$Y1~c4zPXZ{W6}QNYExcy!M1ZLZ{zWwdG}f^`;)bJY_9$U#v?I;(UX zfoT$&v(DtV7n@gamd_v_?L3ruQ8R}jSek8-w<0T-g_Oh z&T2CEhPhtQR`m4NDS~e`XG8rDFCb~>)j4aL+coxe^d6ZSK=RX;K4SDoDB=&=b8_}T zUO1?uUs%id>=NPGz~uDWr9#b+Pt5~s;;d&^%V;;KF&E5LTx7ORnzl`eG27|v=o2PJ ziYtq8IaC+<=OLDia|5i>inf+tySgvDyI&3+*gxbwVizCSC+K2Q*3gk~onLzIA3L~~ zwtCto$)eYyz12s&u>Sy9E4C zu)O-SQ@^e^EVP{VB?s$Y)fN?^m#};lL{oRUN59{a~`6f#9d9 zRI01fIovHPeh&j=oWKpjO&1l;AY2t08=>R9z+jtu_4_cs7f>^N$+(4g&^v74?GJ#J zuaTzJm-G_KeQ+ysOW`?gWp0ar)fY7gn{M@0F^QO?0K>MG_Gyzno3qHL-Y8~gbrCWF zL^~g)b>v@jyUs=QJQNjoZ!BhTyP6YWsVz`{|Ig!7>RN1}cVhX-+lGlPB7C_?GZUJbzLUeb0!@s9R-@v_&)QhUXro%t=hviuXrm*$n z5>ynBCLt0!YRkt>8AWKR#y=$=>%S{|YpnfP=qe07HOamn{?N~y(3W5AP1qz{P&-be z&tG0thmn#qNHHPfYHrGe;gEohS~Q-2{N_J<7<26OHF=#KqV76Bp50hzn{#|-yOWV2 z;oE&Tsj}$Z4oz{GUM1{Ah?V~~1feXa;0Cgb`Q(0uUu)PYt&I7T9@nnXx3llT{@za! zj^#`FuKh}BDabndIT7EP5)kyTNjTQhGgdK*ryqb~uzhGPrhh^_Hz;PDa#GFQjSyr1 z{Wd~b2g@@#;frV^!K3rnUC+v><0#?B)5U4#7iad(G^hUi)&Wy6pkB)JIfkE4AwvbR zUurSXl~!urVeVAIow(guqs6N+X5Q;(fkLjtxU6@R7%pq?dF>vO&rX$Dc>i^ zYIWGn|5K!%mi1mIZX~NoX6}#n#HZPL*VorbQ;I_qMOyVBzsNY%*SHQ|E*ao9%HnwV z59f3BQd=$_3H!=cdR;xN>HT3N-74|801G23d?(a8@7MPeXn5IVKv{6-`_CV`hlVj8c9Yv$+dsfSNw`~Ft-lsXxsr-lbSHm0fQ{6SL9&Rg$E zAQuy47PE$K+R$j_A?pbLuux3v<>#wRIKXZG_XCdVTQXhqCE{{#=Z;T#Q2NhXLB;?% z6cE2722&foeVD2a^MPt_mvhFpajM64MKq~%_yOnZw`9hY-(R}el{J~x04N*qx2KDJCI2i?p0w6Z%y z=3xV$ZCc2L*gkHD|Gvd9C#?7#@+hA^EOS>PP(qWZz|}VKj}Uj&DGs|(_ixT0`KylG z){Sdmy~_CZ(|Q&-kNG+#m+|gN#AD7YOog5~x_|R6W>7d6j-cXx7y?zNx2BzNrt8GF z&lfj)o{*?^0PVmq(l<8xGpaK8jG4g@7FZBX*&O;Ac z*62_VDf187`^1kV-E5bwr5m#=^aM{mxj23{D+S0p-92lG%n!(tT;F}A%Oxw6YHnh1 z9X0vxd%1C%r9kDUiCkT`a%d5(T0UItrC@wt2HvpVv^L!Tt`4FkWJo=Y8@N!~nJFGO z6vE0=2QAd)&`Rth&aJy?TeMWWX@iZGjQ+}yyyZyifC8I&OK^W_wfZu`%|V899kvVK zXEee2j}o5QdfMZaC6K|6DDapI1wOSPaqezuq zRYByqr38J1y}=j%CF_^n1pjIl@b2$TF(@BjCFIA88-&+P8(PS4Kp z{@MY`q&E&pL<3K6qM}E7ol(yGwM5-Cvogh9F2z8#iAM!f0cwpykR{7$O`f5PNu)k% zl+W@KbtL$WlZQjB$c$^a%VK<7+?Tt(#}N^AtWrwxf*!-yv&?w#1#}kFH-FrL&QlFx06fGJ?8;y}YAhX*E zk#9EIFz+l-=I42ik)m5Y!@Y)U{Vn+QQpw@a|Njfl5HfrxDz_h?p99KKs2IV8l2NL( zh~WD)dS>8!4Gt&4LusXWX@6^@wM#2ZU0+;L{VG7>;P=N7{kVu-GzN!vxYPdAq+hr3 zK=wL0fjPbJzoZ8+TvHiskMau}rq=I}MFkx4Z&xA)A)w6MBl^rF206%zZ=dUTeE_9? z4sO|YLK&iwJ4X-!waa|Rd@*zJR%&pWwMf9mbytT#CUHf#i-*L~ydsY|!;>?Q_xH96 zN;_e5pz>Iyn^}&XpcBkE&nKW{c1HJ0ZaG(0g%ffU>~}?gLn0q zN-wq-Yb4PinVcBAuDZKP7pSi~-b9b&KY7I;Y2~&wdhcRBvvu@|JdKJqyH{Ak)X@*8 zlGi4@2_mJN_6GaDo9Z@3e`#2$*M*IjO^JC!5tpTQo`iklU-l7|R|^02Hur*nt)=M( zL6fD(yWeHD*`J7qyJDMl zy!WQ_mK4P)$Z!vGIwr)oWe}5K$9wDEqFkWBVP3V($0;kIj=#ch?aAB<`9yBH<9H!^ z?#FJ}IDF{n{v7BE;z6iWNeG!#9Z$h{kss=M_dLgAQeVUhgF!>D6+Wq987(^;fn!)?UU1EVM?7<+>M zuokq<;<#K4Dfpt3S75_g0_v3o7p%+ zuXM~x%apWejt@3J8am&*Zu4vkT_$`I&6T?>>sU6amv>22Z$=|=Uz!~5bpO_|MF>Bs z9opa3p1T&&KzYPHJvQe`J&mMRg^ef^5=)_GA3f&v*0@<_+_1bAe%ZcJM6(iBLFAf< zchkmPSWBTI#)^gcj^^+&yWqli9_Me}HODS-YES8mw?!AlySWwjI*=(QbuqR;Afu+8 zw&o$MqOH;nx|)W)HqCLG3CsH}W7KrtOM;o8{yxat`qiDEQsaW0W1UWYS#4u%#1E?l zT2zqCsNi^8Rn!jDWZM|^-hg=HH;4R)dJ1K)NaA-Pj z`5ZG)k^lTy+dT*5;{{cFm4Z%-ZQak1cG9tE(YC&Vv=*Af8xtoqrYt(Qn6Lo5JCb8u zB$ylfGUWxDQP6sI9dX({l3p@BVX%WdHvV2iHZd4w@JF!+;gbC_O#-G=!Gds*lG1@ca@PvB_Wj@JEe0#O`Br&6G#0RgB z96KilrHrphKeKk7ZP#TRuCBLn=u$0}MD7>n4Ml24xeo*G*oX_V1=}US4M|7^J+*cy1i(sGvA!Zr3|tUG0N>sbyYT%M7AlwmC&^9 zmAy((vX!0+bq7qr{2qbJ&3Ei>F28r&LIZ2N&_mLwBMy(zR_WNSjqfS8_dPf&&)Zyy zs_=GaYx-635FbU&P(uUh4GuTJ%pLRh5regGFiPe~VIKVUa4!-?2rf|obk08OI$H?b68P-tQQBE-i%Ui zd@Ll*yD7xqrO?bbt~$>QA8d#ZS#7rUERg0UGNm_?!N!`lPm~1%`GIM~KC1OpKtsej zdu&UhOzw=u@ABWj@>%59eud4qr4Dt8Ta2K>LTcrG7{q)$B2t*@RZ6xT85)B(uk&Tf z`nx>XYQ>B(gTL&(v9-yGJO*_0`aB@6Oa5NJN`GRw3rQRKoh}@-ykRv{}$m&gnSr zQ7!zT`ayY3_i)_E0}E5@ZJHODra~9FAp7S#AEZUse^##re_74Y+1O=#lLAfM7w ze^kC}0DRCx*+kwSv9!I$1wrDFe9s?Fk0)kgMsnY+;NhRS1l`MtLT2C{A8q68x}-Arn!VemKCvi+xTf2w|Bi=pu<-AY2@@EH4VWl}w% zva(97@i^BZ6}r9^E+=3yKS-u)w;3nz9PL99il08)ccK8rmN^L;4SpHPbi@$fywb}4 z^JnOB`*Q%;h=7L)`dC@>Jw#!V9p{lpW=LZCN4=_|eEG~W7EMFpJ8Zzn!mE<xLpJ|P2`3N*~)H8QjKqXvcB~;9D^L}j3oN-}S zy+S6INouuuw$=A-F;SRg0hu~yQb&rJ@!vNQ&ekj|pfWaznXw!+X=)-IW`kIVN*|)d z@t;tPffYgSy}+m-(Z}Q3cE`Sx*I@N}=T%1=){EDmf|7W#Nbur}u<-5OXE7()OXjY~ zjwUyx!fIs{SV95qZ+S@1KxNaShguf6qx9nxjc)sgq1zO!L{^u_&xVfZI({99*TIS& z)x8^17oW}}g???V0>96n0-$z5#6%G#VL_kk^_O3UvadX=)Un2^bi86b3#8_cq8&Uf*lU!ni1*~hX4^jS!auCDcjCes%fFt=&C zg&g;GdU*L3dhH~jy6Y8&X_q=p^;V5$5jUKxgl8{$US+;!3_e!aY@a6XwcU*2;mxvf z8O6Ye&Zc4^BP_7w1x0Y#bnZQwx8b~wAa2!!aSP-VV=outu~dmlQ%4hM6Tiug&IO?+ z$~~vBc}U8h%~dIav>n%pi zmsI%tzKwN(4vmrFU&(z=WXKXV0JU!0iw;mY#hEz4=I8F|_>_v#D`MPU-BmVhY+l_W z?*ayv%^hEJS+SBmmEik(;k-jPwGMdrmi!>7c7~d4#$5tR8I5m(ho3aU`$^W3xzXEFH@@-I< zTBQ(+~EK?EtZr~_6^$8&8TqYKW&;R8*3eREl$ zzGX4!tnfs68Ktr#Y_=%FSm-}vrlqWdx*^H)(s=}qx+)^YeOV2@= zrX5qZgy&5VBF&}ofz40(8hb**_bQ%8o+fh(6SDjDF*kt}2?6bY)3Zg5fVq~6AYOW>EH62U#T@WKnh|aH zAuo!@;N(iGaXF{d)4h53J2*AJVaJw%o=PBI5o_=!Z%aqSaNsK8j&l%4rtRD8&@4yz zV~$KsBSoqJ#aVUL7-8o4R4uc5 z^$G_L+sATWtkOIRN0F^ZaLUCL&u1qe%n%I-zso@XXelLUe+8lkjXOX`%6u70JEH11PE* zp>!@5$KuK}C@KUHHE9!!!gj&U;4Pu)vDBw1$$ChpCnVMQ@O`}(&-Gl+ zSSTOV@CV{m3i3fCCG*+`-4Dyt&>-(~u{$X=UKvjAC zM0dLA7}nFN<->y?(DAon_XC21?a~LJ36i}&<)BP)nL$q;%LD0c7pkL?(I*~6eY#=iC#YYVvsA!yt%5r zGO>-e(Gt(Rg-JE_ky`-4IP`iZ3YzPo03D@MYNUNbl%Cl&ed??qD!N-w5u}^x?X2-q zCNz~Jnb%tLp=OhA`W8>D`_3610iG-g>9J72M|4X)&ljPKI2* zEN~)}f&BeJQP7q&D*}96f^^1VZ-%gRc5WLR?q>NzCtRPo2D$R)CI)7vLm8Fdd*Y+= zxoh0K4okshXU9jdqr&Wvm?GQ{q%rC8^mW6oFK-0%()ulxyGrHRK>F!YKjlBqUg_)^ zc;-DG$wZs$Eee(WPD*j|XaH#k7WvTf+1S0<8 zROj%qAODOWW0Vn6G8LN7ELoW+t+>kz)l(})cvHcWFY>EhS*V|M((mnE5}N`T2JKy( zHY*NFa@HBpv|T6~Qc96ZA(cvkb! zmq_*l=1~&OuQ9ouN>n+ou;QW1*hN(PGNq-Fi5g9JS$Je-gY4rlJ*q3m?j*^lCsx3l z$c~rA7-#sHA8wzl7wQJLb(zvU#8arQ*uq(yGaVz{9xD!IChM=bJ^zdz+8N~}M zVNJDtlr-5M(Z!E}ihO_ktE6xXvnmq`pQrt9^UJ5#0xg>hy(A*P`d4?orL!yOlk#nD ze}TjO8pLz1|xl6MDERU?*kW?W4h6PTWr2iiQhTn!W$a_lH>eCAg+{ouUo5psj6$Gz$Op${xDw zsOBd9VuVhm0@IqD*sgBvCYAdD2mn<4(+Qah1olo1M-Dt1L{H+fc;S4tjx{ge>)=xx zue1k^@sF}8KhPKqh<|7NmbRL8K&)>vqHEvEqN=y`MB)Vb_jov!Ik9t(pPEV2u>Qb^ zT`T0O<^}mgH@f5{lW2`vs`{?wTzGuKc{==&VNd&Ji z8m~m!P>0dsNn)0NB8gch$DMhm!!Q`oot%#v7haHw5?i6FGZpp|= z7$*7tXV!|H8|731PdIea*F=VpP)9*Pe(prPfI;oa4==A(Phzh!2DJ88SNw4qmHZeal^4?aFAUh>f zH-3_RhK_Pi_pWD7If34)VNib;+1yt3$o8Eo4?1;BhhIw__Fkcz zdm4p2JZiF~7_{itWtUx!8h2x%#a6jpBJanC0W9&Giid~~d_6gv7o$Y|d5n!?2$G#GP{T?R{g#ai?r1TM2?&YL z7fN2ZU23%O=#HJFN zx<9pNhAJi%L2t8jK6hntYP<*0P0uQ7^A1dGx#a;f-5wR_ROKZaCM#Rb8KsxfWT}}x zkNK1>!Q?{am!x`L>Q~C3zZllF`paiXt^2x;d$%#eNBzBU$U6%nEj-U;mVZK8X}cXH z`NOlR^*LvsM^c&!1IV*%@zotGZR^^DlghE8!;8>(<(jM(O0l9tf4I29ItZf*mN!`N z{QE}l_&Z&~`Z(6ASH_YT%9XoF`dXDc^K6t)$XeQ-ZnuB#{lMDC=W9HdrlHpKiWxfO z6#XR3EJ?J1+UxocQ=PBQViwxFU6^i^Y#O7Bpz0^>o+`i#)d>2)19kahjPwi)^N(aI zGIIYGSSy<<7x_xNbo0C_$_YNy089UqJ^JM{V&CP>XNDw3y3${_d23FYwzs@qQs*<_mi4Xd#mh>B5qd7-+T3rkFM|Qi(ds| zwTp|?{uTnGv{AAzs+_BY8GC&DMh2Kc7PX+BT)^kWO5I1MB&nz2xX9(dlr!T!a8W5O z!o?#9Gk2F7xP-!SPZ_3vWRh=>?Z3~zx>-Fk(TQFiIslb)=}3PQqThjw6{k?3#UGqp zO0Q{-(3BE6>br*?pwn{m$*K_#H3FJ$`rAUOx|lHuZ_U$YOaXF<>L*cil5pk! z3ZWf}qHd~A=f2r61BY}35aujcaw#hU@veEeBY^T-i8+ zdoi%8k=eHtw#dOA6d0vYPRdIIVp5?m@Kd7cjcD(j zzd54$cFSSt{Y7ETzK#@o+h8`;22we2@XVq3#)IebTZe5i;NYyBp@>A z`todgP`6T1LPb~nu51k#T;IL3U8g)et^6l#%qe!z(jP{e=A7|*t@h4B)Wv_mh@nkm zuWB>R@=5$U?juiO9Ru{E$;A}o7i$hb7NT94r6lqJ*pMU6osVf3Q)K?Of!7L_qMEMV zI|bSwLOI%Kfgvp{Vli;=i`M<@th?OPoDb%@padyABqmbdfm+IXItV*P5Ew=M19^)u z)faz4WvC1oSO$6TewN0O#pF(tU8%Ry{kN1y z4nJ~W9-YMhma>#|@%4bv^;luAv2-VBK%)91hQ+Nk*C4D5{eLTQeeR09n8O#M`l41A z7Y)MD5xkBuH?YF_x0W>c|0aUQc}Y!1nJa);CI4G3bA9^15hVYu7UNDtJa7#cfO4cx zhzcoZrkVZ^{g2|DjG6 zs0sAHL-gbDqxpBv>Hi`3;MKT5WKa@7j3kKnzhiV&uxgev{{ZtHyYM4$X{0}las*HS znXjQU9HQs@0+-0A`WqKiutGw$lP& zM%8_o{yS>COrduG7$AoEKfD}|E+{2+I^Bk`?Kjh}vE{;&+)?#U%w)_?1FDu{m1cUsylwYXeTnz4la5S=*o$W(f+ z7wk0AQ|BaYyXo~Ip;y z-!Jr>%0yPzi+Z0>d=}&drz)Re)k^+su6UEImjGX5EHAPLSTBKI2Yv5j(S zrHqQ6%ICi}(Txh*Mw{fI@{HuV@;=J7!vrt(3a4c7!3RjS_l+j%;PU>^l0P4?a| z@?#V*D!hy4e@@>uEtU9GHkN4EPY&#(`Jr{swigGKfOJVr^-s2%|6G=T2|kwnH1Dyo z;wGbC{Wfl)bhP`y+@avjIb1IoHtP+v4?7KiyXM#lArRPx4YJT1XTF*hsLMjn(Ox6<$g{p{@LI<~!?e{@3%P6Ux>OAr|W`=@f} z24Y_%bR%egWnqNs4Y_g7jW%kEp95n1uLk$vMcu|$Pk-PTjLM6SI8r~%F7j!8ubugN z2-A4Mp_^Ls6MMG){4BiOTKCWIBs*pu%l`h;w>avf4Z5?*d1)?jk!%A@tUwOds*|xt zz&iofSOms+luW$F&4!V>;3j`H@{ejT0Pn`BEa7rn%3`oW1onEqeZn<}S(& z?%(EJJP>*hjd+-jfAry>z?st=|6m%-zh;jZQ~8{uEMl85`$wiosB^mP7h~=_mF7f+ z5>YuW?)>&~T_#uMWO#bFyU!<}*Ksgb*Mgre^7b(tujW@lUW=_#n zJ3_s@>gC97S=N(IWvE*tXsA{jN}%Sh`yK-LPT@>5jZfA*{fWzkD-9IO>-O<=f$8mu zo4EOr>m4(#H@^IWQ#`H3b4qd;ciUaGops0RH>HLbH_G8Zdg-QmrNJIYWcj044oCgb@wcg3J0 zDJueDKG*?*I*Ho+GUAcqI%1qY8;PO|JUe~RXCec@MV!1VU|}r3`UUsWISg-HdLMrY zdfI03Wg{D#Io9e6uah7RYyAFp?38-f@v*tEW$U6@y_|z9jm2G&ePmf3F@%rPqH#09 zqWOCFkco34oq3IV*{iG};PNcok``9dF*VIrThj_q;Lmd;vCHE#a0^471BJESVOxgF0Jk}SqzjwARI*HN?fB(=svi$l<}X$qvy%S& zCLDce|8h1CZi+{6SM=-ZPN^pwt*qX3A2*GmE>gxe-kb|W>n5($e3VZShLk4|Z2r8G zJizo&p_+AD2Q?Nc%}KL64v*ST$2B{pa<<GKGay>gld`$lcL_u*;mNC>W;ble z?Lnitq7jxNVhc1Vn9L>FUAFWD1AyJU03D6IV=w624fc+ydir2uV5$B(kCo~!}MI@Y!qDsP>6X!jS*P~sV}Bvqf`BZ|Is1iWD# zz2a0O<{2?stX4(?Z}+>Cf8&LK@X{Q2ggw6b1hV39bzB2Jpm7FKdS*E@Q+ZZPh2_j6 zoSe9_Lx52V;itq#aK9A9AbRQ+H&8s8Zx_{z@}mKauaE+0`U!6w!(5C$oEK{~Ausvq zsNTdf`xB#|A>-|SNPpZj)b5H9%SpMq1oc5VJSQ!LRN=u9h~|$>M08Mkc7TEjZdP9b zJ^bdB1LLBozju)l1c(Y}ukQMjR3)WMUI8qL`SF@#eVN7XyRyT5ZZ{lTu-UCtJ)MYg zcMW5f<7~+QR7Ub&wC0P}hP&D&^Yqv-=Ghcu)o$}{UZp&F+E7(=kZ@@)w_4W-wVBe$3o;MX&}!3fFAA<7?cb+ znV-#3M|>@Li3mR|zJ_dfUf|r<8VJ8@bjzvpZki_{sy?jQ;73y4wo0>%>@Y^gEZ(nZ z<=^;8nq7t; z(+?K`-KEtMLt9y`tAH*Vo&9ToUUzMz0G#3iOpu* zMQEup#e1vkS9Io`o~8`Ik2s>H8*=B{LrsIFLf@4S>!b}*uAM^y!XCyLAf4KT)Y zel5QZfi=;7v|Hq!MyulV&DR26w#NyE&ey1}r!4ir=4JXski7dyj(W$ZtG5pstvDau zz>$Eew-ds2E8{lJzN|y9+wOwZdYeypzi(1Z#RfCzoIzGGLX2|Xp#8}vyAxrdx~J>N z+5fx0^GAfG<4Nx)rG~`5i?eO&tC|oRy+L{9v5=a8U-D+9=`UlH(HxQsVSF&Hb!rEz zZ;!{r@a{C*Wpz<{EB7Zkh?ttuYM3)hbG^lW6?N*PmAJ8b-)XGX%1f;zknbRnuG19_ ziZ{08BGpXCs!^FkHW2TwMq@R!ZUCLk@&awVOklfc*0MO=rfbjZO|b)L?V8t9NT&Bu zd6GiZ=|j=evsY8W{$Z$#00L|QSlg^ zeTs{;knVDzXf6lRQ!eBr78I}jp{A1*Oyo7C-T2{^+cEN=tG6!?^CV3GpPe|)>favX z&EIjx06BomY9~IXDs;omVR%(ao($p|7RIQTWPv^H3R}Ze86|qd8~Bu;;W-QU4mE(4b|F)- zVl-2S_~%%sS_G1X~N>6RNNk@|W+|QlES9QKYN7A#fy`&II#IR@D`J+0A zx$Z)Zk)bM*oZ#GlsgHw~9TD>8-^=;RDx$1N-8R$d@#b27EhUZg6{wyldZ_6F;`tC# zM*=1yKr3~SN*$ED&y`e>{EJ0IdIsdEH)GIm$Ej}6ue?i5^f?WrPTdC56a8_HR@I6w ztadr0OE2@R>eZ1-X?>Rb!Kre>aiJ+Z0qnnwqpieSUlhbF4Y>aIfB z-{G4B6l zL|b+u0rBygCvUt#54xr1psNS}^#M2x|4kL`SsSASaz z8cbF%<8gzPNQi12doE}sgLF}5$|J}b(TDq;!)Y_{DwmPVr1*_fZ)r_xWNnp+sVJa4(HZ3vegiqq64t`6jxNQxs|^>uqr9 zR(|%(mwvbhi?NMFjiy&J#Ktu|GOOpAMit7`%1rL6`weJ=`VXzLKhQ7|++*hgMeV(E zqyj`FNV;R?x4+P^z{T)B4TM%qWpzDm_DQKepRuM^9GH$Bhi&Rtb+W0inpON=rAg`Q zFONNUEe9oHTB(@uOfk|u=CDT)wHgK!Z7$R=2nG*G1r_uy3?||YO^U=o%y;wCtsTUe zrv?PeOI@)H>nQI&f^nX;@{)q(LXX-v!_}Dyqt8dN1qr*kMbJ;JSYQ^-26{N}9U}}+ zsWY&*=Vz6g@Xj}Wh&Z1NTeZom7zyqo5H!L8ZWsmx384!|@fC$-zQG?768bykRN+Z< z&KPa2?05;rcLo<$D7~}+&D9wmXv(egGOY+`->zZb6w4H z@}8N*fa$`??DSrLqWxz`_~Q+hjr^E@Gq>OBPxXo z2Th%a4!a(05jGLwESgBP$L8~Sa5#W}AU=8+y7sP(5#%kZh8cs3q4aeot)fYa^R8&3 zKs>b|km##{dOq-kp@>JiS4-1&{k3BX-yXsvJ9PJ&OwP&lOWl|J8BfiABNG~ChZoegaGr1tqPQHX_5lF*C(5{K z$)M+*fkfhsmQpE(&(^RbjY7aRt?W>G8|GttpJiQhiDToEIb+VZgx~0I03`kH@#YFJ z$y)3F(~7z4eJ*+TKvxooDyk$MHFJ}bu$j%ZGMd3C(_`qQn)&=jravu*#o!WTNVx6_ zJ&iS8^uLbBj`-MZC1P0Kh;eBsY35(T`%x-DQ4(b4twnF%f<~2; zx24}M3)-CU*)gnHD;izj7MC>q^q`VAZSH^K<;K1qww2iaW)O)punSO5|Dk>;+9J4@7bhX z*E14SYB|hh`*j`1g6{3*&tBo487MSmfiR)!i7)2>8#mXiaQ}@Jsv$4wY}JF6N}cI3 zhU7N|9!nVKGL}qqgoX5YI0>-}y-@~Tx)f}YpYIQ{i>9aR|!6`Iq z+$A?h*Ub@xNxb+BH&aQ<{H-v^S5lYm5z{&dtJ;ttBJko?IKkG4q=k98=AAxnBU9$3 zr?}#8NEJ14#d$FmDYrodJAE_$+NZ3BAr{oSXt@A6oxcA5nM7y`?Y^6sP}dxBt+0J= z#(SU24dzNo(Ps{PI@+P2jLTneR6!g_`0IyU?5>Q13$A7h<(@eiIuiIeI53335S(zm ztPD{!NJX;0SqPT%&**56LZiBZsDP4m2W2js(93*NX`@n9k=$9fYWt<8XY%h0?9b>1Lwd4z<2_Q0v5UBIRix zSK_7EqZnB7d%^m`v3}7Szte!|V_4_j0^=AJ6FFLt$eYC5*H}=^XYv#MgU%i69qedG zE(zt4I1i1vJ~``vyC7FIyqAIu;Z*SFm`GaW+`nfV_scL2O3X6MSf4@Md%*)-^n1E; zkBUWg3DhS82$u=r+#l}TSAw8nR7I)QDYkMN=KPbQ_(kss6D`@7Ac{f!vgRLiUBVH2 z$9s@`ZVmP!zTThy$g{JnE3#pbjJ!?9_;%>vM>m#L+s-qO8uGPovN=@RDBIGG+;<=4 zMnNeZptlZei!MruTrHx(V*`?z`hzrLXk1iQnnP(Flbu3c!ANasmGN{f1=PJ<-Zp!m zpS@Aj5;#UNsGwS^N=;Ka$iHhw-=u}%4()Dtl!aNeI55>-a&VO0lsMk&22KXYXM6~# zL}jMinryv|DA}-G<}~c=@maIkLPX568SnU1ss-KyjeXwnFYz~H=LYZ^M2FA$o&ox! z(l}=oJ;9l4(Sqg z9W(1XkPnQg(0u2HLa@PQ(;umGOF7D;=hg3P?-{bKp=}7Lu?m@Bii=s~GroOr#KrViy&-16NNfx20bP%)Dm9Ef$7*Jc?C)(wy zsqSi5V_KRMUegq%v~P9YLIgN=8qjQ7a|;E#M(f-=u)kj2k|Yedq!)*G7V2DkebAcbcAQ>LYeqTprVa!4A z&#WqnWI1vIJN<1sR{(!D;!V#DS0;7rGx6|9}Z%^t=bvFNQ8ypB#gDBou;0TmGf@!1W#Rl0`W3u!SAHQ z{aQh|KU(>wOM`CTgM6hBidb;8) zz*enCon(Pcajf(;S5hFcOVC^lU>09a(mQ#dk8#O(8~t7I`@J}AZ<@d0Bt4*42u+`7 z1gdaxJlXSv<9xN9p?nm)>lPP6Eq^mx=uyD*HO#jP8)s12vMZwi#+tIX5JQG@s5Qgo z{65^a;pk61^n*h2u#AnzsDK}#m%)>-?EdPQe-ICvtgmABh}_k41>J6LiQ4B)>Zhv z(hu5vVfrytWxtVULnS6Q``o>8L%!g!$(-onCRYYuh6sj()Rz26oteZ|lo%pB)$GxS z6wfO%!5LMlj7NS8F0lA;&{%r#ZYVf|5;=jnSOP41C|dZjLd`l!wUar=_>63rGr!FP z)kYI{fu56}WM3<9&Zb~{uJ5>2eLliR{io#TMAR?-xCimZx9=kj&=h>G zd_pzP@Vz^2g)iZf=oYR|06|k=a$~^@QITW`e%K z0lhUFyQ~p@Yz~F&h_$gCL_C4GQQcWa@W=MOXw&A|f=h-Mhm@a`&q4vBzZ3e-M%;7e zBfWj#g~#3Xn=srw*!GfQ`7~Z#2(|vr?5{p4%V?&CV-Wj#DiT*NE#~mdb-cx%U!KP+ z>0&EQCarmLHU_f-&u37flRu>A;qs@)5p#tIYsI4QgUsSBBVrqo+L!_n^xPgEHw}SD zvDF8CEvt}ZqO~Iwp@wCrxWg&wkszk${JTKyqpC3hn*3cC*;4A2^-jl?*~1${llM@T zAeC~sK3Y@3X6p=5PRPbxIxhaP`MLtxZ^Dm#x_dpDlHdSkxEr0ikx4tavs==N z2*iM;?(EksnUl}&$nL?v=x&A&^D!qbRBtdpi~m%Z7ddal5WLrdbAAbRmf3hhs~`0~ zuiDhVRw2d|UJphxcKjXt0xNRT9eQ_oUO9+4VQRG3qh0h`)q%Ai%QBT!dVZ_xU4dT5 zQhXn|W(eyCFtlHd8RC2Lq7W&n(lFu{^5%ghy?4k#0+b>0E^jR)6{klkf_#SUc!NYv zyF*z$w|YcQDBX)KokSx$Zj)B|5WkX>HhM(>cPYG<@~K5SZ;rRFLvWx z@W!P?9%f3LOW8-c7U;iv&@mL`A|Le@S*U2}`5^kwD8`TC&?=#LS0ko8xt_$caJM4m zl%SmbXD)k{&&SpR@1wjJR3W<6xcXspRB~MQo|AUbFAs6 z5nX*gc52qsGypN_RX+sCb0~H0Kar2{GNHPyA+qcia5O%;-CA!kTC7|(pE-8?6TsLDuUP;fQ?KMckb2H_07u(|>B}-PCN}vhYN$Fv029!C?$z$d+EyuP?2K5KC zR+oc4+04g)H0$oiubt_Ufy$U%=RTb1iuor>p&0ZF*|*s7jZfs0-G~Fr1mlvh>rf3* z1<3JRGd&-~I)?q?j$Q{%#>wu90sqAN7w?}=M<>FgoKdlWuz6pWg%G75I2Bo;1#2dY zw{b^y&R_sT=D3xRyGQ6<4u{z}AB@z6W$cDU`6xj`8FX-}32RoWfe^J_j5-RCeE*3& zLK`1GNOTPxe)r7X7lK6gr#mzOAgD8stsZ0dt)9hIpIGK6*iD-)X5{lqfXY3QNS`Li zZpAcT!*0ig@00FVn1Xcvfo%l9HZpq^EWr227Na6fUcTL-*IKHLy>hA0L5SJ~R;$bp z4*K{=6`p*eB&ztVR;}tAgTh5MvOpfli(#P^cLLv6KO*yHpZYuMkomuxf%8rc!9wiq z83Nuq+(kP8eou#DjTIT+K`2b;44Cmd@aY7AYg8ug{ndl%I0)d?;jD+@zZ=LO%{FSK zHg3175kAN+(313Yn63-}-&hPU60O4%c9xIu94!kHJB7jVjQo?peqxFI7Hg zJQKZsji=$+WyWDPdmzE@zV@ZK+Pgn2otGxncpLA5~u&7FQFj zi3IoH!QI`1y9bBh?he6%LvVL@cXxLW!QBJFoxx!b-@Uufvp+dA)Rc8~)mwGC->p$x zJNzLte2)l|?|PF5n>rx1!k^R$1*8psw>PDhEA`mAQBj?cY|(pX1p8f~WvNA72AgQY zAl8iQ+z!tbVO{DGa3V21oLQWVt1qT3hjli$AZbf%^qBFEb)V}#3v&9}!-ZnXx$zCC z@OA9L|0Y#a-bXdc)==v+U!b#X5e4UgUzHMy_UD}Kpe?pL1Aiv(k`x$c1LI6cdmYhr zSHk*bTlkJ%$%IY8!K|I(pN;&z2@oxuUeSD% zBl^WFxvHW~XoG&*aZn!MDKph8ug6F2iZdm(P@H9tx@TvVxiivmbM9Gc!Z0_CmV0~Z z$(553l_zFLb_ENd-;kFQULW;I81NlXR)Mj&v%975*fo-ki{EB+mf ze+k6tpj>u7dgxOs>pJ0DEmhm+)2Jh%E_cc_%B|4Cl4p2l`D=pbgBS*aZzJe2dCIY5tyZt&CgUe?Zy|jv zi)!+-d6r~zjOTgm9g6j&J97_ck$+oq{R4Ns{xiC!jG&Yf1>Wo&t3#XsH3kK&p_YHR z;>pD=2}kP=m7fb zTYvX+cIz4sboT6;OtJ51F!$c7`4Q;7*N1yuzh@v9h{FimGUm_tHx>>@tsTf|XN^PN5U+~2j)`n(oj z>*}oFiVebignN`Dm#6AAqWfiR>+B)AD*2|R-vQki&uU9q$*jzTpEt^?5I`^iBm98x z@LkL}6VD;}_Ihlp9GnRda%2Am%-mgZy*)ICkD_>mS;}dBMo<0k%+*zXE|SfW7oojM z3wWTyiA)LmZ*@2M!XMFb(UuE#+^#7LYQNZ`DWSKGN()asbap|fHIOhw34yVNZ*2`` zpV-oTuWskaxl4DC(%7h<&2i2oonB?gF5Ro`Skp+Ig{XhSMO&2^=Z_ZD9AgRZDzjjG z?C=@d;Az^SvoSAYrUEYkk?sjkjGbEK(RYT)axJ`CF=ZU|_XBk`;Y zas2CjTa}mG5a__BJ#h}(_deNj@P|}FO`NH!9XZ`d6I5|~+54`39qu16;07Y_6(uF4 z3r)4xzi==qgr|!W3M{?dr6r^bXHB}foP3wC&J27FrFKI}B;=zWN6R8!(erdkn&>z;z9;B5Wo< zt^8B7pdNQ{k&*|O3-|0U#Rbu^W#myjmLThY5KkQHW_0e�EZ*~rtwx{BWZ~xjo_2cdk$&&l71J=7DR9+UZ1*3D4XSq`HNTFEhr* zyXA)`N<5$!t1yWO#`*iF66*Ny9waxf5Aae6# zJTGdUJ|=?J3rl9i(}^v0)|ChH562S7|R?`%&VY5@NoKMA25@{ zN(Uzx!x$y4ME_-z!F~9@6Sk6{DV4s*dUDVJ56&SToz@JGCTv5hcn!zB+RiLwQ?zh2 zkyL}Sqwo|(*ORH_(n`z%wNLY%WxP(wo0geki5Ge2A+}%TEa5)kP7$pCPXND? z%YtZUtUXqbn)~6hcZb{TnVLH1I*q(iTL$?Z6;0qC8zRnMuip6Xv={s=s{aC2Jz7XOvY4p{LpO(EbnNitt$ z?A}Frq2vFGu-@kLH#l(@Ti;%&-GNhzJ;;kUAl6eCTYTPbgEb4F@n_(2_`qG4o36Z1 zs323mNXbKo>o_vwQtLZIu@)gDB8&kAw1_QhCzdpYd~liIzrBr`i8vz@{OF@EuRJR> zx8T{Kp$XuBi&90P479~OSpCk}9SqWps89TL@wQrgSaqz#s>yA|2>;>^E|m|4ic_sG*7g7$cC7!7{o00u8)Dop;VbM4@9p=Gbk_vT zgQ8dwd>LJdP__0I8o5$|CGoDeI_g4O5Vx6wv;h7{qTWdpuX)XPF~mSvc|gKZtX-Vw z=UHm?jDFIJ2$6xV7(Qn#wAkGL=sP} zfa44)fVTI;-7^$hVQ@l@T%RWMrNu=i3&SQja_IND7OEZeR^M6NcWiVl#UI4IGKGa> z$9L0h{7g4@_MOIzCKs)g0Q)hPRSGdnbzoLsZ$MzDWe~WVj|H2h!v6PcL=5(!9Mx{3 zPif2o*F-|j5Z4dqNXZ$)MnT3D`BY2e`R?U7tgvl$yy3CuWMn%W1yu2N=zfzc~pY$PV15S8} z2ij9v<2c|=gjd{m(b?|ll2v@#&%EdC*&bpHRU@|x;X-3|LR*m=Yy2`6+R#fOee zWD6V(=74`A1GGPy`Tw;4VC(@B6F!xam>~W&rkGk|uwm6S>92b-gMLS)ydm$X!oJb< z-olqk;_-!a!#mv^EP<5old(|o;JTc8a#cqZr00CEKR*^WL%<)~AoO!>xdrO^4Ow_2 z1y(I+61nzyFQ}6Q!4I{^xKD$Zofo?+9i5BvN4AqGh!&nG+#niA{0cEZ9IZ?RVrv^i z^RPd^-F7e5kaz^~5A7YXQwxO;dVET}ip(>tu!gfm2XxE1FEq{uOFst(z7)1+7^Q4h z8@TqoWe|?m@*%}B@=vk((V!6qun^A~*-_F-E^Z9_PjNUHC=3E55&o|!N;`7sLcIa^ z_{Dsre1lo3UzN;LYt6qqg$Iz_-H>r3dPPDxsQw@*2oTi#k6_G)%+deJJlpHf5Eg+Y zeAxWIUVMDc0c@7I41uVWx>nr*8}a`uGZY*k^F!zC29@tdU>g-6wDpJ3djAP+%)*tB zq;g`n*PHx>yH*V_4D5zF_fGeaYLv)ER~~URMS&@5f1&J@M(x^l$%j?YYV1iCHZWpn zH@x3g1JhzlNMHR_BH+y_fH$MIS2^^1N0faDYJW}& zXifZY(J}&y_Ix6T6~?tnvJ)>v)lCKe51!QzH%b&D>0qh*oCDn6^#AB#f0)F&8I2wJ z)D#k?nT%2EQ{Eapm64=pxkTRi=6l6Aw*`9aP@E69$*|c5fj$}sq;ta9l8AT`KL)f-o?$_(#JzGLY88y?!WV-nah>Q0v#Ex_KSFCNTZuPb0%{*Oi1 zo!Z~Q4Ut{GV~HPHE&24}&*?QF*6>GoPO_^1QJ!nqJh$TCp>J3&<4G;-6*W3O!PxvP zb*kOk_BGBB^bwb9D#L!ApV8YlVEw$&ZX+EHUazv|{}c@SZ(wWSd};+HDQ~WImb^De z16^lRVURxWUyYz5=@Yq?P2XLWq4KybTEpMNVi5(i_75b#;FKDz8mg1bp9AU$Hxbh- zwiG%gUZW;+zdfq9_IzhgF$ePwhlU!5>~9$KNf==lg0@B+^@;*-W2Ajx)2t;?bUn%W zdB2JpyM?}ZDNRV9CEIn|l|W#5CbK$`{}!l3JXY6Ys>1NH7=mrwe+-<54e%r_Y)?Tq z_xGx$Pl2A~?!p@5VqELNyyS|nfRB_983EQ-0@_64>vH~!RO?jl6`!`cqr9>MqMo!j zJ6RefLp@c1P3dS3d;{Zyxm}*@YLRv{sfzgU_?a3H32ZpeyL;$=SgN zwB{m$R&hSMOoC)E?`;`E<&ZR+X!0=C6)(#5qA;`9VVQMSReYXzOr0$FMux8J#n$sE zCeSY9_P?;HApGp^@Zb|YgGOAS3TuY_CKqvHPxAAbmSnt_#v4R2HTLS3tuIUQ6SS|pfE~daYX@%HC^1sK ziRE=e|Cjc94Rw~4%i|fTDj9}&jU*`>!}QSxD9c?u7_|MA-98$7gcRKHkbtdk0+Ul- zr#1?g|H_w`8Qv}&5TShIfZy&AoGC%BehameVY80u_e={%fl_q?s$Sn6y&s;>#|4#(I)9~XLJ$t!YTQ) zCbYr-%Jeb%R$=6_(GD_4KXihPajaxu0ohsxP94;vjo`cU-04y1W0SC3e%OL87*1mk zyY#qVJ2=1XM$T^tUglHMS^$0MHcr)FwR829MQdglVSE=;|hqaiYa9tcVwfn=#>Io9d!InK;rDX&Y2V*@pYkM*|@L;xI7U z9eq6QQ|VS#!#c+9v0QZ;3arvj--Eyg;xwSu^f~vnMe&SF=i(&SQu>L%1pv+5=}9{i z@!wpq|AwWyol4aPP4kBy-89M@Pe({-FPdb7BOWt{`UJ>D937f3h;cWL+@yn7ePR(( z7^wHvs5hkPUQ>oukHj$Yp8w-9`?#MrQ zE6{GnR9o3V2uTM?l}=SKp1|)z?Od4=j0WSdQ}Goles- zuB0hg_8+=9^Fz-1R&YhIfKr{zt3*(Frp(7!!jrG-fPBMSHFx3f84T_GwADnaU)ulJ zGYhsx!`Q~yamhRxN6&X-FXEemN3=Zi+utHa{`>*jM6^x6apORlg&Rw1uDY{dLtY%= zZp@6!Grnl6-eJePu84J;l)KUdAslL(NACw=yBCx+VYJM{D_9p;XCEUsTfm~{aH?ls zwEdU^@iveDesv3(O#EjgQl;&Z#s}QGqoIsiDM6`qcPqu&0^nTWD?9JUYIe&@xissT zsgx)`|6bRc43rtmtFxp5af`*`J!#dT%k}xHS#Bfl)_YVeI|>#jiJg>w3ao;(KS|R- z=%dER!rVwNMB4!;>x~Y5@OTvv2Ta|Xdh-z&C;RGlI$S%k%?Y+~^ZtRdF<3um zG%v9|y=-ub>yn!M~U#k46IlLo2 zK)=GBnkIEO%pfwSV*O5V72;UKQA%0;yp)>+c-*+%a%OI+4a_S@w3aiKw^GnNB!$Do z2W*uBVETj15aE=1-zzP$G<>nU{^eS;VT(SUS{ub)y`&367#CJ9c@Z>%jF)^3DZ|t`Y7XPxPJfKk zft23|E$!Xr)h62?aS123!wB9O?AGA$Vdd%0kb~BRPk3tLv7Zrarp(j)hIDi3e@NDn ze(Tv!-~^g#8Xfrc=p?|*faUC0Rad%*1uvtaUw0&$O#G9O>8J~D=-Rm|K)Xe+=_H_f zqEER%C?o&$x1XT)rxi?Ns-jw5XyQpwvr%dGVPc2=)whpa&eNXPBR|4Tjb8((~ zEVb><1~d%ZfP7Txqt?h$bpHpuujj8WOldzbg+X1oh!ftBZdJ*Wcpr2^?Okc-O-jT4 zbN5I@OCw&tj_CXE1_8^Bc(1i>^=zeXGM$ADb-9~k;$q?S#;46I%Y~YO+ABQXLlO=u z>hn>0=YZ>KhX_Z$>8FMHY!e(rPmeF#BEMpCScp@MPW3x%C2(LC8Kx)jzvr}Cp#5@3 zpfY0ec*0=YE#-kq^*#;7P{Jxik`w8R*Te=pEs-&ajjlkYd#}<9O?f;uRGrVM_!}h`QgF?B*VIHTf|JF0v>r%LFFZMJTDj5 z%*~=(f-Qz+%D3DX-}V%Ji$r~y;xWkj zwSaz}@c!=J;EV@d{X5*Rv@6Ut7oh;-BP#*rHnJaqf9QGBk3j4%6Hp0M-1*PNd^C9r zM|cLjI*t1Il*uHo6r-7)5@UW@LvO^BCB$F+BHC|?(S=KmpM)@U!xxSQ(*7`<)nYGN z9ov!bd^=y2M<9kYt}{hL024Pn<@yL^hZE9sY>qdOxNzy>GftNDSl*$1z=7U7J04*x zQ0}AMLgd3_s*xHLjEc8 zo~4saLxDjS+lseplXR~ic{bDU7giv=Rw9*Nb+7_l1yzyYHFDP;0x;Z73>-Q(NUO@* zpWuITsXNHFFW~*wKO0Y2xnJ~dgj(uj#s-OkjUJ+o#;M98gLk1}CW>?rj~Io!9{7ab z?{u0(l6*Z1bB+7C<-hAsAF@`04kQ4ItpqOU9GbYCU`Zo5E}tC0`y3`T#@cyU7xY*3tw#va4-3gxcgp$b`o z@z`hz|Fb_`>&T9#e>6qAQ85@Xs88W+mTT)yiG7(XB?h%LmGv!WmJ=dR@PUa{ZT0?o zxL&Ku9Kj|`%)@T=sn1Y@>ok)?D+iW}c#*hc!B-g=KsCX+ep2p4Y587Jb9xzb8(3qx z6uDY=oJ{x$EGzsc)6PogYfpmEss$OiW$O!y0->`U)AM@UCeaS7{P5T}_@8u|xFmv9 zl{*{iTH;pH`R*h!n(@>M?+*6gL!P?BYbZ=9zMhuJ+L5V6?fO}KDd3~liE67#ZN?x_ zR;TI4ic%Yg>F51kBp6pOb5-3~C*vXJ4pT9S|8$>`>9AF(B|Fe(Q8e~xMKb?1N$MJ* z|I6=(hpAgL1c|gX0{whQ!>^K*(U~RUNEd)gA`^76O%l@hoCN{SSgCVfnTfmY>jMP~ zR(~xftBnN@Sxc){Dr0&rzjby?_hfH7OP0|JnS z5RSn|Dfy6f1}-SDz-u)5mwP+zHh&?x4u+v6dlfJQeUxmj+mgJd=p)8aM@i*9!0`*B7t$>eNZ+YvMgCNq%ZApm6?K3ja2El$&eu1Pp0-X zglh^*o4kh_Mc~o|q!=Ne%->9VfuYa7uUm*ZzRBhWjo%N}r&2dZuhR;%43jqSFCFVM z+Brg|)hrd0W(&oe3tQCZI-ZR;4oeHqxz67{h3Lx$KSxzP1)Wa_*DO8n>3+nrP}K|p zerF0Mwx)D|<&6LU#-Es`UjMwzd=I>rl-O7!PyMWXEm2>88wmtOXq8DL+V!6 zDoAe9tn%psW9(_f6MyU8HN3E5kQFM*dq~nK6TS^L<^?{B)f+a(hz-#;i96pWyB+5P zle{CqByW0>vD(p_v>k)|@TeVJyJ1flPS^cBMf^BOAt&HZLsXGQ^ilcZx|rIOdaRan z$t-Wmyo?YUUfMfrch{@7Zl8y=b8x}{TEF*_Coyb6fMyHlq z6Kb+P0ffsdV$AaI+}{|Xsc!wa01&6!%PBUS@^wHGadx$>+o?k+`5JA zd1s&QZ9wL8cY`7Hs)a<^{H*`QI2TUoiUw!2F!ge|pHrp-0z^JAb$nSP3XJa9&_g=X zZp*Qj`(;MZsQ&bczE`U;Q3XFjS1DabbI<^uM~2d<7DG)+0j*0-pmmw=NAb~<)RD_$ z&2QtU)l6>(b_{WAVg{BV4nMe$DIyu5Bf-o-TJAUb(F!08s}0=hZjCc1zA=)cXc<7= zt_xj;9}U1!fkTHhP!dxG_h1CNK;N;@)Pxp+ZYY47h+%#-%))DUSojyOM_FCSG=XDj zC#S8#(w9E9##L4>H6w*+{lendJBxs*BT!&qAXh1OE;w53GwPaEu4-!H^}%tW=^f6x zvxah`#O-ca8fP01+pSlTqMJT18x4QK>^K~6K;jU_h69BfX=#w*#Sf_w_7;aaB&Doz z;KBM}(aw|u+w0;z8~-utDA3qiXVR%gDqoIod^q(a_#zOj)YKK0>N>wnH`Y(*r;Emd ziuyvYny;q7;K9x$O`KTsYhsVY!%M$DaAeR^c2)caOq4W#Q@&$z(*KKRm{I$OTs zv;$wcDJjA-t`i-gxC?gc3)}aikTd9{7ovavJdRhV)EtEV!$9f^^TCiE3nkxbG8{zS zi&m@7T5?H9d_V!e!XPi*S-yv#1-v_wS0mg~stNQXAgnjlkYWrt$% zTuEgIn_-of$U7-4J@DJ`P=CK|^SoUw*`t3?eZJc9LeyJmj7t)#UgVa}rAg)*w4lZj zuEJLCq6w+8yI5owVnS)G-w>T>XnT--uObvU^c48sBq(7q*Ub)XuBE$Wh1+^iPO{1l zWm&O@SxT1Jdf@BmXprE2(;we0Vso*bp1uJmLj@|?{nfbe%<)eth{26?6`t{cZl3>_ z`%i&IgJlfQQ0I=8WJvQ2$6dSSzQC0J^hNvTR6ylcU9nvc;!25Zl6cqAtfy>ooF#x4 zHU(3OXViBxttgrb1xu3Sl{bn_NpY(7g>>s5@*BK^mDrqM^UU9_LpSLy$OC%}MboNh z1$wyv4S>;2P zS?Rddr8y+fcsq7kzm=R&5zj8&(~4_>4)5epP4t#|Ny?zaynnM->I1#4Gn4<*J;`;A z7AB6|(pXfSyXrS~_seOfgU4Jt;ne{kkv^MjDd)#alR6 zDKUwlg4N?%RQ>VU>hQPsb=T%ylcD+~{89RZYZ>)UBH>u?Cuhx@)D5SPI#%G9>{h-1u4Kyv@L9@&r=5w99>HU?)8!~wgEW?^uoN@b!VMF^N<-qQz84BS? zJl2??Gsp-T-M1PA9_lzi$2ak(nd$7v=slUG=50IHr7y4gt~^OOw6%k1sn1Y(fx$r_ z{B0&g)s#Q3Q-xF+!fesEjjfX$ppn2AWs4Olx7Yk!Tn)?o_Y;*fvi`{^w{no(4$smT zGWn_DxNMpN9DR+m>0ilFKPbXk^j-EZ{z`;89GVF|=N83R5Pcf&8L*HaPG)wNj~~dX z`m6mAY+&PMETrG6ob$e>n9w)LA$=|*gMS}cu2CG6mO;Ey6kQ8*e||dAT3mfe4(+`o zO(eH(FPlvSf2IZ@qi(hQRI8tj9pox|DHeKS3 zQtxd=4jnIiDY&%8DXVx~2o?RQ>>g~Oxv7xBr4=D+JBKN4ljxn4)rE7uM903wVX9|8 zLi0Hm&q1`src=25MvCsqa=Ix|T(r5RT13UrwCH1tT9@YE)Q)^_jh5sKES|~iMA@CQ za7tk`74hRme5%IqPeYM4QjWFs;@eQqRWfnLgk_)KxA)xRjEN#znxn8EOVueA^_WvA z_Mkk)J3H~mKJ7y_D6A3a6>lL7@RB8_$fs$G%w(HB)2Md!x6IKP+X$y9xorgN95Y1K3(S6cJSivJ+f zLQ63JAPR2U;vf&CbXvi>M6S6axlzIJ@1FgS!moX&Pj%&W2-8a|9+v9|CA9G6!l_K; zbGZ~eI#}5=#Gp(AT%1pbTzc>Q;HhNtiQ_U*E;=$?FXsKj5%|Yq%67`F{<(}7*v%a^t=uUmljQSWm-loUxI?-q-s-KrZ?blEPPET4{%t4!_@T70 zom{ko0w{ss(5IRkJLcB-OeF`NZ3%eow&dI^w627})`+W>`ly@PIjYDWHfjU&iOQ9k zwqxT7@)`vb<@;|;*jj!K0}Ex->c3y+zZrTi8@3l_@|?QEaGa*9j@QoN<~9>g%HN@B zqk=@QRs&M$5z;X>Jqt$bGzPUn5^-);mBMH7EZF1?vcqQZX(@Q((QR60IhQemhHQ4n z{;QG7a?c`yh%GN%Sb2I7{~W& zgOTSqS=u%KE+_0(ZK0db=&MV9CE7!F)3`yNTz|f*CG!uXhq-p5pIl!W(Gws@Fc%%c z)b2<4{r&~GMm=v6lOPx-)>EG4Btwb@AtjbBO!wAYlmUTC)AG5PM%e{8U_w$ zzSE8a_St#97%~dp`NO=K6>lkNi zt_`Dg7ZDnuR$CPE3pkqSDUzPE0=q?`%cuElYcE89$K(NAr`_0;gahy_KJ#kf>-OJ4 zzt=pQcJ;Ot-}nu88E6Al(wI0%Cv^f%hVHu3Pg?k`Uw+74i{7oqDK=o;d~eJtLKlzr zrqx^V9#MO1eR0rDAz@u2Ua3+M>tNP8nNTB$szJCD@tS0qX|wL?Jh=xhWSwqf>cmf6 z?8_nz@l5}-&D~g?T+ul29(OO~^Eo9lh9V0dUDM2^n~kdR47eUXwMG27f$qJ7DOH>X zSj}m=J!^pZ96aHsMsyx|)w9v_Xor8m^QhVgQU@NzI;}K;s*L8gTX(e=XkDHJ8J8H2 z4t`~2)3$LEH8(j|-^C4atA|ZAZSG+_?G%a`e7log{JG7Kdrdh}S=ph#~7tGY` zKSZ@;x=urQs{HeBAlEYcmP75W^hZHuqjdbw-yMq+^|PIojbZPPu0>2K%(O|}d=0Zz z^|OxgNir83eXUyRU%%Dw_V_#^nDcL8v=3$Ts9zsq!yI>7pD`KYM>H~7Rj{-K2@pvX z&Aw_CBxC+f|A@qpJ)M=ex$5UqkJvtzqlw&lcuBWXsSn5-orSK+W8>7%J?I>kpcEyg zvkK_q)@7sGYqH|e4*732i9ts2%lqk< z^k9K?g$MJUq?MUT;D&dVh_I1vRq$ImNrGR&Z@F^MFyJ93&yd3eg97pr7^T}Z3!iKU zo@_^+&XkI_4l`vmMfB}k&hF}AGl0n3>Z=a@OXi|ePrS(ZML)-kyVRO2H88%7Gr(K7 z(-p-=zk=edN6cNn!9&}9rinAT@`R+?a*qy*NM--x(Po)YC@92(I_u8ERHa zg>K3Cvm3rLN@QiM@MAEPa4w1Drp)!%W$Z5F8N$=J)DS44fM2DzG2{j?!qz9*euLVMA^0IY~8DDBE(!H2wHavRQW`#dKhS@pZ zZuk=g5*!;By7|%k@;w}biRW~}BzT^^;V!qQ5ChbLaB4!l_mx?ku|inO>M4B2q(|N% z_u|Q_#21kXkY1cIW4O;qmp%>=Hj%o1G%R`q_w1c3Ry@5*m~zPYtFn7vm%U^}DpkI> z4iUy9`rVOA9>-qbli^5QxDuc9oMQyxC}e7?l#~JKM0|6gnqhvv%Fwk;>fxtC9Xy?b zbn=?!vQ{Ym_@|(-h&TJ#qP$t>{y}tWUahxmSoA2bKqM=1yjehr!u$!hI;B)1%qU{d zCF+zBFKxvy2zm0sV^1c_8Z6;do88yn33~_Qm_MhE)>qOrQOc?QDE+Pnk8X`VE6s}_ z?FRB8PnluLGqq0IDKjH?FWVKkv7r&HFfr8OsK5rFso=_&6c4j*y zBcSdAR8MOc2B+b2h_KZ{-^BK-RYt{$PwlS8-^4Icn`~vnyw(5x6VdluH8*sXHOmRw zpwnB~Bf*bF1QPOClTIV9K|0S<7iHE`o7B1Zt8jj@&7Z_q7?`ZoHK{kip=GB~SL1{u z?c}(soRgSi?n9fXz7my86oJ>_9`v|1%ekG%-U5-2q0B867FAIl;>c{YLw0PdO{R0( z#pqh=Pssb=m{evZCcA^@pKG{{YesX;fZI%D8Vn8e(e3rnwR`!;a*c*`667kEPj;uQ zjnqj1%56o!5$xSe0H1`*aUXn=0YB_R{}%Z@PWX=Q06nQ#zwUl8V+jDFtDwv^slVFw zJ10Rbe42zDV+Vg$?5rQri=hY{5B4C%t#Q24m=(qxIsFcghd5M+{UL0V!nG}+^EiAJ zsw}svi&Flat(q5;-|HA-GHv^-0v3w^( z4vA#^ij6u}Tv|#^Lh-zHUF6`I+EM6#k%puu8!%D#X#brn?pIpt05Oh{-cP@-&MU{s zpsLSLywz=UhcaRwVreNnY%PgS%3SgrqQI|HJ8@fiE)iOT^4~2ne^7*aVihETaCeE9 zZy3EFo=*EV9}jwSUX5({Za0v4UIP0cAy-|`ENE|Vl;Fxr1u{#>`^=pRZ)3P2ryHX63553mjSkWa2hO)V%MHUG)pY!7r)rr2{R!8m}-%fgZL zt?1$B_NABWlrA}G38GKvpOlY1RZRrREf7R9#-(K(YCIps{JQtp*TXQY7Iu}eC?v4j2izcj#Z1daG>9h4r zvsP5ZncH`d@-ORBB7w$Z70D4;`A&gGvh`cd2jbBlm0A9;7>lkU%Qb^^ylW^pGfNyh zR{=m&dNRtVZozLR_UQ8}Uw2QN8HcqZ8|kD<1Ml6$s5xC@6^;sa5Ika>s6xx6ImACNm1ZW5K9la-r(8zqf1)C*_yO z4YF=6=;)-}@;V51)#&1_r8@Z1Vq-g5%I5@oqzSIl4DM82YvBV*WRThwCtrfN0JXDf zDPtPFQz{M30%h-%EE(mg>D2=3m61cGH(aUwCuf<4B+W>E2B3_zU1J>Ggn`gFI#WJb zOkW_YOb$Gew{WP8Sa5PwNewo5q(}Xfx+b0En}}wn&DMBrKt3N`NC(=9B1MTYTodhp zMVTJ=V=i0Kx$$EzKulWAGox5p(Qp?+3AP<_?FzEpb#&vK8x8oPIr%RhccqmZ=GIvY z;`2YX0xO=*CcLqTk&8sqZFYHR`x z&PU#UGlcRB<`RlIx8;_hj3u);!PrN|!+%*lx?nd_>2aWw* zyv`z(%w@91-KC~sriocB0JWi{A&AVR#)v;NRR8A=P5#44j^`ufbyRR1%s2fv)puR8 zI1!}iH;5X!R)d38C=Q*Yb>@GSJV8)tvLf%#naR2%A)3iR(dAlMA|I%&VBwh1!Qn47 z6t+T@#$5+9sNF+lH|5wr;es|{W{&;Uq55Mbo9sp_)<%GNhkWqv33OJwGdW?%x`v8^@Y+>hT2KvlvP&tw%fC# zh$GL??8NungXAeicDm$)N#eHTAQcU91CP^^E*vSMwo@^R_9tZ(abGL8}5cdPcP%XXcWh#o; z+WtROzR73+|HA`;+8|10F4ad1TiIUk=RTvr^u%f??iggt6vqJ}yCQDBxup6Uy9Gea zGDsGhYFib~%a(zJ8odPiXPRHFVEz)#HzC1MFq&q4=lP_VmvVBHoTL+=`!DYzZsT23O>9J4ZZspF^@UPm~HW7Y@$+00WZsntf&{i_| zyYm-tc~kIJzqJ7;uIlN0SLiA3Pq+dm&q0`lpi6N$KCi|IwI}|(GbS&VR&xfh zH_?9mG59@BRCQyM{@SuB+cKV4+p|hYXQ$NZ8(dZY73s)e+(po!$Ab26n_yqny=pC% z5wRRKx~X;$Oh4*s5DP`in?bJf8v4-Bq?T)Sq*5R06y1u7QSFW9gb+L_3kOwg25dJD z7ez{EbaEOy6tH>5e^X_fK(3ljZq6lYJ-vnyvS^~3$zQ>PmxQjk#zg7zm~D^61M6X6DfQz|2k>-*|Aak)U_#&Rt_=TypE32v#DTtf+0JgN1C2fZt z{n=;5<87u}!E2ZePWTSypv26s`_C|#nO%TMjG4@?W6`TtnWKzLN|DuwsmUWm^h##5 z3=>kk!kglEdfO$Tp9(s=Pd6$a+*oFI@#IBkaq6?vG71S-h=n+%xS>4x@smQQN>^VV zrPJp69eg;||Dv*e{=v+y{+F4T1L`0T<9OJIT+hY#N5rtLf^>vl*~ItgI$kVLdH0(~C)FcFzx!+hZ0oJ!DIBa!0>_al029n_mA_M|6Jb4`(l@;M8 z{F`nQEkbO`E%ok64&)?0tW{`SOHutU=y9?&SKJmmDVyO-D=Q! zQM}poH^`H!cijToE|I#Y`Y!v|jzAsPNm$h+FrN+dFXNvEt7O3a-+!z|uI}~k}nK(kU$6CEDu(Gw1me;4DnjpaK^9oV#h}r&kNWirUCE2OT(@YQz0!!%AzB(wM)t1dcO*mX`wrGzZh#zneU042`$d)hx8_#p)BA61c&$EbWa}~Rm8!;ub9kxw zO{U$RCyXwsf74+vf7|D-dELplQBEJ%*E7rN;?i|sjV89uP8KAR-Znx0l;RGzybC+T zl_V#8>D!a7rS|=Z3m!4}r@QGVK!)jYMCr==w{K71mY!9L9AZSpeBu6xWLC_F9B}h=t$dOW5kWPt$N6al za==8sBed&B``k*iBNZzw;H*Dq{-dF-?*V2=ZtJTj3HTV|ASKP)1W%PQ8(k)vdj+Ur3Ue!%(nqRu0HSlk6SRQz-Q8~W7g%am> zDVEW~)UWbGPj>HZEOj3F!~TDzb=86x@#U*<>AgXq-*M_UW&ocUYLNts54S5c^2@)I z)o{_Qn&gL=MgPi=Xssi%bE{w8GrGg%ih)xm9Pi zcrmYmagFpMQ;*1IZQKMxBA`pvoogjm#}DI7|{nmVR^*VYKcH~_0vW4#_Mab7-+ z7^1o2`<}Vppd-yUR{4PtDo$}44-dsZ1&-U9Z;j!@}k>tA~0H7fUCAo2FlEUQ4^6}BYao+xy(4}+iYY05+Xe9W$B zyp^Q#O2IM>{3%sY>d=VX)M2jYk|7tt{8+F%T2aS_%;}LpGcdWVcI^Bo<=Oe{PA2u; zd2BaHoqt-pq>GFyz|LjjeqC$q`iC(n^h%7}gL{{AffUm_?6lShINJGH>m0V&uJhyG z#YlEnUXP@2lvf@{A9tZ3UnSvWXR0fjxA9<{{MG1;Rg}14fT48fbRcIiDkQo_L^2da zH~ISXt=}*61L8b%QgpP(8 zl3g^eI`rO3yPSFZo}%P_(y8Ef(18CP$nE6u0mn|}7b%kcr$w~}sFZYzu)VGZOVw@K zD4;LcnrZ2NivV};IJAHY|L9kv)y)b2Ts85&`t~IkO#4sE(8BpP^s0o9>(+a<#`T~N z!u4|9=Zu&+id+KI_X|0ng0Q=)5r3w~L{YAY4;t+`O8`Fp-8S$exmEA-h2HI|=nTLC zH)PS1P>!cn5B4@mx)WbF8^1r)ubAjc1o?VFVYB_6D;4g;2mOFgUwe);&@;b`5C7?4 z7bGz0mI=}UQjT}c7@p`3w0Q$Q%`ccBxWQ+Wp7Bl!E1kp!26^Bie$#4sgZQpSz(55y z(j_r2L6Ji08##w9dfQHFQy>`)d!3cCnzCMWoAGkp6}pQ+ly|PuCu!jeaX>=%=(^Ng zd67Mi<}Wv^uYws&8Q*TMd88R;l!)2CcS&fYOu!?xI|jv8=TF|uL=UXQk4F#q(ys*) zvt1Y*mt~_!4kbA&!!T-&!z9sZ9*Ldra@SI4&c?ryCUNxc7%6)@Dzq!j;!d`d=Z0E& z>lSa;6s2^3<5|r_GE_cF?(tj50`8SBlvCYGG?v8C;pA1UGbnQp)SXfl5bG>zY&P-CEU9J+p zMF{QL6n~Qa=N7Z2-@+oOvRzYMup~Dyb=Q9ZXswQJi%Au^nr4ZF%>js`v8NR;uN|QPj6S=Y_uC$ zM9}R_$I6Y#xLyE>T&H&#cRw%kll;75HpH0x!Uu ze~+`+ec^BmC2w!>!ID=Nz#F$1@tx#}A_c2;O#HTtKOF|-E0?wA9mc1TDJlP6cz$s( zjA!{Fd|%bU=A>;HGCvON8w``gX#Nh_NG`yB1FgF)6KRPFe(PbB{^=t*>^(7ab-kGX zGYna`OP|{pPe!F{(Q80!(TDowXiL|2&mo@rDmw{g^9sl~S10DF3M7KD-wsvv1`EN+ zhv8)mcCi>xNrm7$Dx3YVu@`Ic)qD#d`g0e?M9oaSl*~p>wVj0>&2BMSxRPu>_?U{F z8P_7-lLxJYAqX_KZ!`!0k*4=sx7IcV9{+eW-V08_d$E9>mB2SoD;HW$CIqsMo4fa( z@#6N0(+w}Mwp-;z?V}eG+L~y!A9|}|f>ZYJ_*ZHhJhHlx_Wdx04Ywg5epIV(CwIc3 z{ny`U?+I~1;_B^@r*ih3F-uW5nHsI8KWHts(?*Qmht^)BO|kPMAV)+Oru_(-7p*&;fUxQ z`at(DyAw!EHK#&0<_Z8Aj(CT=Yn-_-L1b)uD7|zdi*wp$HuT_M-!HF$W%6~wSuBs| ze>G_|iZ0zCK)-)XdE*G0x+DNcW z2;qtdpAi-ewx$+tfjE6aVq7R>XhPx$W5CJ~_n%1|qXDie=|i3;{`xl}Z?N9GE?zUl zHTJl|1w!;A&y@yG_FN_>=o7F;3QO z?+>~zQnFeAg<)AeRb43+*Ztvkg>b2I;uL+x|J8{slSLoA;7KT%j)%+ZyY#$7P5aer z`Gy5UC7T4^wOr8*+8uO5|40@45@XX#!m+PQP6@=LqL<*Re5;)B7X!Yr;*)(LCA)$bE_!EEmG zEWUd5Y)_Ho(bh}Hom{~7)*VRn`Mg_qW_qg)a66FCg>P!Ban${vKpnc=WxYhW$-eL3 zbvET{S`n6(V8rBl$xJmhr>VoAJ6x z&8lS4N|`#f8mBUZdd30w-_NrBxmzSSO;J5HA-dRpwe$}C>4P{k@O?lW3Ge;>A&=eT z83BEQD$I*!$3^+kBXu*Iwt+9#0iC(TpB+q#i{+lj`)A`$MyS6>oN16UQdI#rkZWm~S;$Q2q$bPQFneNJab+f%~BRNK|R1VgA8naWG17 zO=W@RToA1@cxCp%Mo~!*H_t zNB=`l`NM1UN1PMNt(ryg)wwGr(t>f>bV8fs{sJ0XBmI83$^zppu>*{WF8va*-VVl3 zb}Y~`^i&51cafg#I1coxy2eCZQW8YJ{7>$$Kx6nOL_A?;b33;#4$v5g)v4MsSJ@jb zY6ZKy6_}s;Avb0*pYb7kycqOgJ`(2s6HUyx@Xhby z_oZ(h(mi7~rNhXe)n$HC!{&X4_<){-d>BttHss6PwS|7@Wf1w`Cr*Z3hMn5}qtxwl zek+waq5bd*-l&}ZV&LFrQ;lF(fOevBb;RaR-j$*TkL!r|M$D&k#H-|OCj7;9BZ*hx z5WIK!E~Ud@ye@JXE<134i`4=d%0~30Nt8LVGLL){XZMFXBjurT5W}y{I+g$w!US(* z!Uwrd4t9iBifu@Mvzcp(0;RGJ*zSu}cGG?k+7r`2*h?9Sz; z>i>sVpc3XnV-hQ0k7iE3yrcuMbSVYWSpBifXQilmJ6Vhl8>^_!so{TR?5=-l&aJC; zX@%Hc<4T=#LvYc)_M(82I6R;X(d}dAGbx2tD(JewFNftU31X z7{-(cx(NI{wS~~GhYo?cj2#~W`7{6$BNO++v)}3QRAE0k8K;x)VQ2bTQXhUig&d4f z4Y|JvxfKF>>WD7ub;SB~$~jtfo7g?^xUCcXCVdU8-LH%wn=%qUx&_W&}O??W_ z;!8AKz@6?`%}yNmN+_32ACScK0xOW99TSi>C!fHKsf4_F@=EK@hIcJNVbUtjjitE* zJ%?ABbNxQ*8AQd(!g0~k<_7#@KPgMpFMDk*H8E~{SGl^=8o*Ak;X6{Nwcz$;BPD9d zYg%`h*3U6K)E12~Z4xC3DgP5o8vI6+Hd=EtkhA$!wx1nPjv+d)E6Ba(Y((CM@qYkN zstCQyf65OqnWjE9D5@He7tTvQ9Bk%f1bvaIRZ3k}BZtQG^39c$5onFaJ~d%b*u|A^ zkH28Z{Ur1JH9(hSrIOm3LBo*z$PG=u*n8o;oz&0Qo4rDLIw3GCku_Bo-+J>?%~wnT zb=+`79`y+3eD2u&^wck8KQ`p2NR{G(mVKv7j;H@!NF6ECn1_}asqFA|Xc<-5LGt`) zdwH%DZ6oS{!cSYwnbR~|G|RnjRS0-Zr z)W3hj0Bf)s{ioGsXjW-ok))Snv|{R7-w|b&YdMNsp`Zbx*zjB*&)*rR z!2}H35HXT%y=0f5ie`D60laM~FEk#|y;m{%39(*X=d#yoQhau?`C|I3O7`amPF{o3 z{ud#wWFM2Qm>!uh0LsTLC*Ek@nYmgkEyTH+nE{F>$4|wU4Ix6Wx`4VH(7wy^3e)kK zA09Z@^m+05tb!WP;V7JzzRP?E^TW2hAd2?Xh1BLI_Dn(FirUP$n4PX2*#Fm4bu-%A zYNkUaaL$h(VW$o4x9SJ6Q2}xlCHagzwDK!b?~=Lb8j9rc_;-bVZ+b3Q@%CYj(ps%w zo#WueFM9t)2-nYR>g<*O%n&6y{@I!^AT33eAV51RWEAhhkHQGiukj{xqP;mY`-Kfk zo>0%p)~x%3Q-nde^g6-Rz@nD!l{L&LZuXOLLQXEu$BIkr&G~eTEAtM*Q#~L2fpIJ4 zEHot-N^=)8%fzl!n79U&?7+DgxHzW@JKV?UUB)n(1~U;c6gE3x&j+t9qZ_@o$zw$D zP+k_{Qa2A{eC~(OsgI-jRyB-GP9Nk!G71u^lFgGUvHIcC{%PARz>Bgl`bFl~J|=4X zZDb&C$p8g8P5XvvvwgFO(RO9l6d`|~kVQ#i$Wr1=EW9Ofcf$xZwt-MyTO@DEZwl-J zrm9B!csp$xy&N`|;Loo=OiuuQxZ9M-@dF>5Ef|321|oIQ=3`W?R`C@+UBlSkNqMRS zHp;33IH}kS7BuMC%Y8sygq;>NBm^U+3VA!PO~!MmkAhk$@JE}d0}c{ZrKYL3nrgPt zZ62f%9ih~eQQL)*E_Mkzebgp}%rk1Ak7bOZg^BSUcpop!Kwp72n;!z*1{xptzCu}k z1Wn=d_eH)|XfI-;VE#os`45ooQjmu38s4=_n{K$JmX^E8xV zVPr&#Mb0D+G*Cse9!UZXe>0BDFo(i+^!Q5|T^kjDGlFBz8y`lHv)^~Q{~sqjOoakv zaDG$Ryp;f~^{H4r1{*IT{5(avtz4A_eX~CN**KckGP$8q2v*M$M~GA9@#Zx?K1>Bm zD^kBbX7v<0JVf_=DSLn-O!Ily7qlN|1@iOSed^2B3?sJoxUZ^oUFM1K5Z2m8LjRv> zb{mkiV4}Pg+0s@ceKtS?z%JOJBa3dOLWJc-J8X?KA=9I>a|%3^)QxcW-l2WBR{i%A z1-apVEB&PT8$h zK@@TXA>`nbTe!3%c!emzB%|7Z&#TvZUTXLboEM{Y%amcb20V*Q*}Mv>iZvC?o*c-RD% zl#PA5g22tKd@ck!jeZ{xDTHscWJVsr%KILv?2DtQI#2oN_6Se<;kuT%$4SEJwFSED zkW(rd6dQ?_1eC!}pm*&kH_#Lkar~D>NbO&fO>>Yt?PU(OX@h?i<0`fM_=Zw`F?L8h zoqVtm%NHrn-{jLxNeHHzfiLGB-Bbuc^KzQoQt_XT{es@OZ@(zNCJ^2-Gc{6{6u+cQu z&vX$#yfyUp)7+nz%n92lMd_JW-|SmmP)x&cl5lD;<_8|pN_p&+O1xHrug?OcjPj5BboN8bzY17&eMVpXQ z5N=q4cwuAo@uU0G{ku}WS?@IwkB<2SQLiVpSs0_VJ(NfGm9}5ZlGlBk+jkT-yQV#X zm&@E~;y4OeqFBw|Ee&;XS$j zSp3Nt@>%4-;(=o%kSl|uo82@jmr8YYn2{=p#KeFv)qm;W?dnm)jeS&IMYSUlM?Pe@ z%l_>gGgqBsd*LNpmPKKQ%wc?;QXiuvqeXW>5=%uU$i7&%e0Q>s6xN+wansFi7bOE!vlJS+omzBO?j3BH2#Ln;CMJ1L~W51z>Nx&N}lX? zOHR6YTe@t(+ICQ1N6d1yoEyKGCSi|2%m7*C&*&SWO61f-=M9K*HHR536_nPrj2lLi z^Ci@3YUaBV<^W?fiqQS(d~C_w0_WWN;&O2e%URm>hKCK(#RC(nnw|()M>@ zEY?nI{7#fSI&yu?g3%c@PL~(oO5(2(iFVg8<_^asr_CR-^fAieW&ZZsOkj%gJ|7|@ zm{CFoqP9Ia$+X`zrx-yVZ*t0|bu&hic)la=)8i_z(>=K!@h?QnCqGY$Ywgo93w! zG8P`ra7>8W2k$OXi<8X{S-)k4*?H80{P2k*_R!>5LJBus7{9d(SncuI_7O3O_A73f zvLGuf?&zvo{xc7yrEq$!&tv!q1*nhuwLX^B9@&Ta2Lw*^##0o?>f_s~x<++^*=9p? z%iKnD?Nw2G`6nNFNvK9M+Qvi~cie;Oxae`UlbexGzLidr;(m@8b^Tn#5+9dpe7oyh zSb~bRv#oRwbM%=*JUgwXU$YE@;dfJ$>FTIDpEUDx1z-2}+<2l1V|#IGD!Cd7+mYNh zHj~*OVqY=+*FGl~&xpWBeSMP1cG3?BAA%Oer{sW{TjA&ry*L5%Q& z-^>@pchjhe%=w0s_^qC;pcAw&@{f6&oD^)l&d)gi-O*J$D{Sru_7r^v4vx=M_kM96 zXnv?bR&w2eJ;Xqw75;T^*&3adE|T{wGiA)|D!W$O+U0;W=!mL-?69FeqZb?;_{KfL&*NR@G&OnmuC`63r67+`EUX;Qw0VPrI(D zyEd^vx1pG-?ihD|1b37_J6_LpajRjc2Q0oAgg`VUOYW8h>q+h%>N+&;G+%l#z4L<8 zxbkE-%n}SsN*mmCYq{WD$!us#ez~F4-z|4aj#X`)+IheMu(_QIbcQJIw?h*mc4qE*dfhs19oXz>OIq6BU(Zdi z>d#K!c%h3b*!U(ai1R0!2YlIBD=4~Mr2D*Ryz6NU`3K zceBBAL77An-cW}g2RdCQ{9Y$G>q)xuoSeLKK+tc$52nj5?!*%bf_)GjbC)XFgKP=N z!?GH;C^5D+-v!f8f(jcBmoc!(M5%XL0)|<^<(wsY20Ga`-;gH{y(PMYl#MA=7PIpk zQCKYXdWfT4nnLJv|0_c$&|y#lhWj|C7K7OXyxNV*J`5JDXOSUpOMc&#CADD0j=Ad)~wt!)(ha&D0a5Q+q=d%`shxz8H`S0zUQJ#QbG>Ia? zw)7P9Pn6%npSJk8evF_?Kq0;eG1{#GGM9<#{Y~Q3nhx*l{Gkk7F6kYJ|KJZJIa7;? z<&t!h2cY)ojbG_ai5=bulvC`K4qimIG;1`akWMP*EyaBK1~VD^B`$vr8?U8dZ<$J0 z4r01TZHyq*+KmJB58tuYuJey?*I;g>_FsM{5V+R~5v615>>Y}J+j%9CJ8DC?tSD17 z2R3D7z4AH1*xXc}&)YHyE<@N44P>DY${Si}Q@K5bBHppkUK4rp#Uq%Vkn6XLCjwvLyq4A6{3wcYbs)Z^U7)m^EcG( zBQ_cFJcBlQ@D|$Xmx;Wx#iIg|Tl`h#w8OWU)n&`O8l5rL5jCiY;{tG(Q2M#)&*=Ke z=xaPwnp_nN*>WY-VOdQQBal=wLu9|qHqYsFYkTK@Qi0rlPxoG3dVC!x2y4bBG!z71@frzBJ!8*J!Wa+IM5wOR%leuJxA7H`3B~rxE6`gu$q`& z&A7>S{#St-k37DaGgxg5?yeh2qbbCZ3+?Q-sdFTokgqZQ7zPsN=}Ht}nKGFM=q zUCd*`Zx?QNE<8#hjs5L|W`su$^yJ*6h{@71#l3eW_ZN3|*(iGc9x+3EHClO|GfouR z;)C}lP=394i#7p(RUT0f~e_q>l_1yhfC%5X~i9C=WD&2EYa!$n8?aiHf`QzIJA-L@iaM87uDnU7Q(bbpkB&xXEH zntsic`i7}17do4ydzEs$8*dc6WF)sT4+CG(#^u7-&07`pl2y4XY2|Y)+%xMayL;HQ}Lif7V@0BS9>GvBZ>$aUVmJ-3u*ZNiRYU! z^bPiEh{y@$GL7bbHHdottCgLYLWyfYk;wQgal+irQyn1!10|1^6&xv7nk;Mm7AZoRaxI$EG^T^Z5MHt1&T>0ow1#>m| z+BZAuinSaWA2^vq{@4Pz%O_?$HaEU?IaDK3ot&{FAZc3(fGGbVVz!rPt3}#o)2Yf83Ox)dD+uUt969Xoq} ze8K4_+d0?Y70%I~_<^2eDBFfg72Td>j(i@=aB7mAwt~o9Jk-d>k3s>sH{my_^$pgB zZT0&ImgMQ4x!@PTsvA~MOeb+44_>H$I+wU@f!8!O4^ZS zQtyktslQka08X2HMwqH&FwE_SkCAsKAy~?bXXp~Ja=F-7;9I6VnCsq0HMJ^rtMA^X zERU}Yt|7~t3$$fPb zIsE0m9x`mYJT#5KrP|Q$>h7PwNe_lW&ydMm$Uy?o-?K#8L;_KApN850%q?qfBaK?~ zado2FidZ))aoUzW^O5L(V0hd7MuCrn`ar{)=1vE|hdZ}Ea7+T)2TRbTY5+2+i8ggE z{y<&+Xj2RszR_kiHULHKiNfpmI>p7dJS2kf0>F>Q)HlQC`zuQD}s+nT`j7kDnK z_Agvk=q4m)c%dTT$+IgaK7k|D>v+@3(Co<%j^btL9Z(b2xu|T#(Mi0UhlQm7YA+x+Vb^Cg| zYx1T!j&8>jtueOLEhr5Dp)a5`)sCBapQVt`dS6Wb3vQ(IYKcOCwWYa$UNLU6gFj=) zSG;`5`>oVIM5svQP0X+Q$R$>;y1tXiMPkLKyrVE~L0DGk{wt)H*0K~g4=r^1tOJe` z&!MV^vj%KL)_vFxo?fy^!V-o}??01343pXJ1iC8?06B7v;xESv5N zOh|fo0}%?TZo^!pCP{-6TP~RtQ7smNdP!<$&GnOgcf*HV)-7fw?(DZ88eE(F!PRjs z2$h=Fn}(t!(Hy3fC?(3s`CnsL8i|N?BzUpXET(GT&bGr6K6H^Iw^OX)^O>c77paN* zU`;d^nmh0{*B|@c$tOtXZyE_0e8vEs`AHS~=XPPxO@UH!F>!Oia)Pi<_8Rq;CH-jx z8HZp9tN^Ci^1vc_gD0VUk8t$G^~YDNggoBHqm=tknfB*79bZHP))#@q@Ma=%(SWS5 z+!$j!JY5M|V>GF>aI68H5_n4k0rAGar?XSxg9aPHO^O9QiM#rWK$MNIGV8q9&xP;Ts(I}b1pU$FV+_`a8`(vVuiHHgy?FK@GCgHPPcC( zX(R`rAr$uiK|>}#h#L?4P%IbD4Ne6Z(bsD+E)rt@AWdE;hk8vBAjxcR!XQp_AKIK6 zg(P_-{~Bti$p1$(%=sN9)O1ro>|Z{R*L4J%wz`MwlGH(O;MxLoa}v;(SUP=*G}HX2 zWG~G0aab;ins6QjaaOfn8djYC3*$&!*lhknhC4(5Cyt}{K;p_oU_Ie}H6WdDE!@Q{ zwNaYf=K;%}5%(ylD0++y*9LwWueV$Gd^L(`TZ%KGKu8PNH3L2(e~yo@Mp2Dfls`$W zVq6gj#vKfUpafyrkFFwdosi*`#@FpUv`C^&)IZ9|wPj5J5GKxU{~4QGLSatuWZFqz zZJ6EfDu_{t)E0lAD(YdpB-mqe(s?epek-}fU-8|hfxC%p#WQQWs{V(8KJywkga2VJ zo(ppsnkGp`t2lLU_m%`Q zg^=h!g4K##c+Z`KAOa&Ue&1VcrH9>;@QGccb1NU2j5?eAj=x+f9*mVKUn?HGTN}(< z8T#!6yRZr+npBG+?7_waxmLo>iz^quUfj6A{57SrGO`*@{062>I_s2X|2HRMW9?EP zJmu192tmm<16SSUV$fIEuUc+U1H3`i&fQ1{^xFRm6kZQqM0$!Ze0JZ`>LK`VH7Jx^ zvj4Ykh%s?X6hPR2oKO~Yp6HdhleK=4{GC&G{CFMT(rl6?9>vDUv-G0}KM-3F=ks$V zn^w&+uWrIH*H0*eW?R(Dr0uez+!KCH_#gKfcdQ6Q&8S89L;Lw3@~fAArJfgQ22TBTLjf?ip+!G8C7p0>S6GX`%P=-CZjpZWFz9fC43RmMpi*v z%867F;^-bb_YuMz@J)^&-Qm1DZ_@4m=_@!W^_(2<+5uKQwydg(S&}F}7*Mty_%Ri?g!qyGOin^c<^MR+BBwee zfK^(7|9`BK5EUmgx8gZajv|dk^ms(HZBIm1(b;TJHA9?-?*OX-sr-3bW8Wr`$&Jv# zSW|h-??$E7mFNa7_Pg{;ziU;e+iG7KPL!}}{px$Xb~m4Z;)V+4o#|_Q_t@M4{4{^F zcRP}q--Gc`lLikl-`60`TI{)QsRYjJ8#U!ak#ujdAu#!93a?HeCNC>@>2m~L`rztD zFpt6&cmr7$=I(G+6y2W?+e!5PA13Y=MshH z8$a;y3b#~i+yvgDE=Pq!E>F=jeGygIvfkS>%h|bbktdNZ8mQ&F)m_BXwyj@f4Ns+MGqNrBd}$ zazKekFIf%CZNK8D5I^EFMo!bP-e%sXt+2P=p3inXc$Bl!VK_|y2;vN2;mR&L8t;%Y zk>%WQH|-Ja^Stn34velv%aZ`iD!FziL}u}SOT@){zhaYzk?PX7ThxHyFErO#``SY` zSW&ta#Jl;{62YLC@iY zZ8>i1Z&LWEn5XE&uEZ^jDBT&?SX>H=!J?}NvPICW5l8;c(B+p@*i2?Usu;hu!&yZHEIdY?^~=) zh^p*5EQ0T^8blP*Pjx)8waC`kN&cVTSx|xStny5ueMJLKvPe+Wb+-(^vT9$usm`Ei zYZkAcC)gFkK9$Luj^`h_E`*4_y%S)TH za{%8jNnaDngm46y6|~j#GT|wXN?M%T>??@M_(77|cd&8p#n>VVr+8li_ zQO2UdsZ&w_9{aA|m!dvG?vX2SUGMYNEP*3poT?#NP=i-F17+{`lHe9w?kU|ENsbL9 zN(9m7y={DsU2;EdG(ILfvUb2`wij_-ej--kZBKS6u}c^ODx00luDEZ91x@|b&$AaEwUqZiwI&TCW@fuxQUV)LKP{F?0Vt$v58Px|KFOkYF^#;oNK1uaV2xSv&bl;-{D zD6$!nNZgv4kb4E+{+w(?Cy8&Iw3ThyYVL7#R}8Rf(1S>D06OG<tO!^ z6!0^zALhGASP{B-7<#AuVck$DR>SSSKY;HuP5n{KY85DdvyQ1Qkb5kbkf zzVN<4-30j|w)0kLcRU>a0j(Med^e2pl{l2vu+=S5GsS+MkRIfjY}{h-wVaPU55?j7 zzYA@eN!pn<03a!bxB7n<+LHFzOY!=%%Nh+?dczR~`9BBrv53m!!I7vZL&NrErwfns z^*TQsYI!DW90WfTad9woYQGDUVs!DCym#V2*)fa_t?${%(S=D=J1f@YX!AE6Nd~?0 zB^m8cg2MMn*>eDdVFlh{cKy&xrN%7hqm2f|3`{modO)N+Ug(4hmZQ#c!<+f#yaQRn zkE&?)s~CG!kOL6Zsuvsc!`u?slu7XpzT)W zTBkm`d4XV@v?L$xk?M=fhptt&rt(EdtSu6k7m#W z143P$%r$9?=r{`yM9O9y2nVuv`_{_@fqEN8O;o8F9_Pq_{^l# zI6|;diY$FKAFIPW#QQu${+SZc*!b63UW%fb#Rz;GmQ6rhh{i}f3BY%Nd$A;u7a z1vGHH22bL6Nssg%Chhv>{w{hvO9DqCP=k#=WT0!~OE$^@J>NZD>GvTgJ8-yRQx5JD zap3QMpXl2VP+QjO- zHR`P#j*|L@rN;?^_*6^gehc1Ty3OofkXEH$n2xNs_)aK#1!_e0UY@PfPbfAGW{Tq3 zI!Jtjyn2Od{+Zz}k?f!x2Q3@xo|!FN7ll9pM~t|^lS+`;TgA)bMn8)G3AVYnVMFEh zWGt|Nhc1_EaM`XKvqw<$*y>lbuiz0V0&C+#U-j0JF$4}-OT z2w;5JGe>KjgRclP<5oKF0HQ(1DLbtXFyjDPhx?1x{tlj{hmj0fhmnspZBr1xj@(7x z34|dYi>f5^;CRLDBy#%@VTMfXf6ae0To<8@+z#mzkCuTR~H%QBNY2nclxN{J8AQ|=w!(e`u!L>YlK)INAyfE zMgCBkovv-Xz^wy_40=DRC=0}{u7~Eda0-hV)bLH&PyE?|3gR!3?Vxl21@Y3?J&&}a<}Z0d3M*(_YIKN3)bez#uJxP81DmE~ z1p=k2Fd8lz)NO!NW?71m)~zJEdg<`zKePjC8b{k^?9DI1fQ2d(erbqfpB*!b>cuMw zOrk5M|A#hnY^cDZgkazAy{~9_@6t}fa#XM(bZPP}wYZP`Bh$CtG0?%3T3+0yFdA0~ zdXH-uiQ2fD*dLALTq13g%)VVzL!#oEkiZmtexyH)-p0jV!q>(IE$-posHiM11l1bu z+L0w-=3wIE}7!btR0;%tj(~C<__!wYEwqbiIlh43k_tia>iL7FA z=sG>dvnz#U_SU-g?Q?qyz`DW-|61U_MKxEH9~%`RD5vk^pY&ZTKd9O++rL>6=;MM>OJJ`_pLY5}MeZ zj(eIwCgPmHGo5+)&*6j@<2K)=qO79v9QWc{&*0OqQy1>!LayKVQQv!b<~b&fAK5Vz{ z7fS{DCUJ9MN^)?t1|NuEb2xcWm%1+qr)BP)kDcD&RlZnvV&g6M$=l-KKS~`7Xbtpv z{yOh&j;RXxm8^)^Kq0mr>Kke8b~B!cq}{ApxeP~&edV-DGfBGu*J7KynLd+9HphL` zmmZnxERL;-1;!YNwBeN7zj4%tMTjOMw1#Z!&m^s!c%fji@QP4A!N0TH4SZ|t+-s|q zK<&lo#sgSRL|%3#YjVxj8>~9`;stQWn}vWmX(~5=(>d3mi$M+BBC6L|T4rxc+R zsfh17eATTO{89xw`Lf}Gx+dKCDOB#x?;BV~@!Pn!Cg%RNbp+1E#_T~%J!nNOtJV!z zezw80l+gg$a2;AoTPO9xyhEJi7wYK<(j^)f{cv>*dmMfS3C1vT+TCc$VJCO@!}p*ID$H(zlYf6~+(^^@6U?U>S=xPnO~@ACgq^%YQY1k2h4ceg-rcemi~?(VQifZzmo3+@oy-QC^Y9fB|J z4*%xf_s%>2Im|L=ws*RwyQ{jsud1h4RJgs$s{V8&&^MPGe)q!rxxyGa%JtMIR#jPs z97>AJ`@f%bIEaobAK@lte3ai6Zd5_cE=PW)md$8b8~pvIna8Cn(wIn;n&~$?M)yqvIgfTpOWVZdvexn3a;dvS7#=sAe^yN>lBpYZMy~xRZ-dNG1T)wZZzs*u zL^_u8;&${V-t+8wo=<=4nVsnYE`3(WEuI)Br%;rH&LIMqYc;CFPh%sz);rS=H~mip z6idSeEw~AQ4vE#VoCUDC;Erhx(;RcFCW>|Gb0clnGDLgWZ}!}ZbxW)wc>=kmhu^aZ zyMWx`@xtLjYlIk*DeYak2OIYaGpdCw8%z4)Ro7vJ_G9JFPP`A^xl#NqSyj23ZfMH| z9JC7?6>-R61$q{4dVPdnx%NzA4Ry*7JmlK42GDvvV$zh+F96z_-k(@&y5a4k$Z;?86g*epTurjt+Vw92pGVWXh%Ae50!85n^!S{R$6C z@Zx)n4)y5IJ2K(F|pC?5^E7h;(jIQ0#UzDf#Icx(fJ?Zh919ih;Ztr;d)v0S_ zc>v#kbszpYNqq0+x$E4=KjYKyN7#9VjyWnF(%cDy5zwE9&d*Qu{u`@aZMqF^5RSC4TqW`PWW1tP$BEn|kJvp9-N>j@BN8!@u5QrW-GzP&>5G!HkKu9_6 z)O5t2Rg(letm5@2Mr9iOq1-YXx4RQ^oGx0OyY*tR!L0 z@Ab16P-aHrRgldFv;i)9j#mNdq3L;Yp?keGZ3n#dYgzHdRBG(3{XK>2wA*!|l5yE4 zjB6P~QO~?|^~%0IZa(cDxE(-agb^#<* ztNo*@&Fe=^(^3IvsEz|NIpB-NJ5Q|A*XpD6lbX?}2Iu1?EsZg@0;P4G#Q;H!cqnB? zU3NWEcl{^BU+KAiiYmx70lbdlY-TOyWbHv+w5LsB6UJyZfQ0AJ%7 zhK+>E;jC}!y;He|0&!iM@2BD^9GVM0q2dd?YKx0vGCU>{RXB##q;}2_}od;dAMs{5@&q{e=*<+Z_%)uXM7ZG>THN}^W>8D zy<89iyHqQ`khnm*K_h2bQ%>R`OxP1A>almcGZIx?%Q0xA+;)`}DdPr<;bI#W)g@)? zVk~;{%^$R2S*?As8TFORW z2$%_5Bm*|(;DI6c{-&!{chXt#XZ~M1*jPZ5RgxF~@(o-@0ipaW@PxVOCsoS0_$cqWTl6Y%=@R04GAf2yPjGq7EZH+C`K$uw&n7wcd8 z?M~E??3HJiqw$Pv!K# z$lo+UV+TKl7_MVG1NpxWm=+E%nFpmx6nfK0<>vBprYj~WIlJr*wJ@m65P|0#S%hG5 z+wrD8eU#xo$s%Sr+tlc`V{>c5ulJG!(vjGMIK(;^{K012u&%8EAx5jagWIe)tl+~6 zx*N}`sVyAZI0uUOcnUeyM>HH+kj+t?H|_}$v+tcgr|>4&3kc=EwMR_3mxt(2-(jI+ zzJg%%;(nH1-KPI$9cH>$HcEMx%>nE>JSQlipa&P5c1lE=FhGOccsc)s_{ZGGerPyJ zq_QneKd<5;pTJ}K605=#u&<{?N)^`9{cUxf=J#|iwwPIjJ?a3i6<;WHxzG%4#mn$> zP!)(OSoZ1BEH~i#BWwe2LZI7gg4Xb5JULIWopZ2eG>m+cU@OM4J(NLe_6_h@gD)@c zLy~*>>TU8q;}(A)a9`YOJ*w)$TM`&vS`8q*t@7gZpTMRvIMOZL;pJuuZ0*{|3S zYOHq+r~=f|z@aj@d5TOaAz{Z5F9zMl#BF4D!Di;D_~+$FYu!hP6oUxLGKbD8r6f`Z z>*vsOmP4%K_KcK+6OtC{+dN~s`U?H9q4S-p0t-n$k@$BzNY5PBf}cz+KQ7}kqgi<+ z=|_}ipekOFRF2I-h}~wX`Wf^#5W6MJ0j#EgC~c*iz~J16j+I*kq>Z4B5K1npV0_CN)*<;uJ-GncG=qKm_fjgog3O()gEYG@vyl%@0GB1 z7C1B{dw1B=A}tp;8OVnovqGl?aey0-v8qo$8(c2G!zm(ynuQA*2RoM3JkZC5q&bp^ zWUN>4&!2XPjv8eV6pucQ9q%0lq>zrh9)-WiSiv;S-^KmS6`E=~WNk1-N2~eKwH3tAifL{(xOyiq$?_i{?BHK1eC3%c90Z>Oa6} z8n{3p4Z3-%rsrp>u~=Ub2u!p61Ji0rqWQue`U!HMIPdVtoQT2@H9(x^8Tv7-^huBU z`?e{k8O#CHrT`>S5SD&H5(VrWa^Qz-IGfU~Fhh}eE6H?ap%;-BXfIu5e?Th5eXv_vuKM7=NIkhsiZv0UL`U!WyEZ8L2) zBeB67l(2MK;Fi z_}|~<9w+Wz2?r$~I!isg& z8guV+U;1{7hqeUkeXAz$qk`)C?3DoEg1z2WIm8;T6NcBO#U9ITgt6_eU3>YxnVCa& zs=~2>b)ISOQ0ChtklzzW@clcI3Rcz}#&{rkQ_gHI)Hz;hvt8ijLaqN2(}`<8A=r_J zvin95U2zKeB}10h3Z7O%n(4*Q6*%Bx&P{!LubID>;{U3q71um0U?=ojXxq@423f&W z=p9{MQ9K7indZ2s%t^NqY0eNc05`FEhaH-oT=||I8jpQ`mtBHQ{|4*Nke>r?)=&?i zpx^2@cH21i*gP)kRv8to(prg3Bso;!1s#jqDY8j%?Jufx71w1qEz$lx-jzQuYCJ4p zQ?gq8QT~kLg*i6y8s8uU;+TKKo@t9gH&JL0(hzGSXF{Jaz<+<_2n-DVawn7r+=tc= zml4+dm=ni_7Ooi6tYP*elvUHm%@JlJz=ocrANhC10hbYx8SkWq8DLy$o2!Q_<`Yt-r@RE6rk;r&MzPshYm#HOR_7HNX^uKqK1E6y+YsT@_8%$^MOJ8954-r;4M z4msS_GLc!3!4@^}RhNDaca1kMRV~H@+%)h&#gDpiagW_ES9;*SmEbob|LuxA@@*o{ z`)16kbilsO!5;JgEWLF)RH#`FMR-}L*~~!+rt@YF^7nZ4LxHMUY3CGEJzEr`B{#mjSKmc*s!+`>=&7`i?JW{q_Y+-`>jYCYm^96AeGcfZG0m25rZQ9euk(g8p zy+URTy3d1TVGe^dUG$e>azrxljOdXb3yN}XDDv^#dE!oSVauoL)~|n35nXTXxkv1N zs~kf^)yK`*f^#zC!_!9utRkgBEz=5dQIzS6f$eotGIqcOhd}Q$wCH^4Gi`>m$MT<3 zP{(Uf!c6a_1-+*Xllyy}ZzXs5$3f=>^}nH7m41JlH(SGgen{#{Ck-%_oxjjUYad2x zUu4?&vw|6=Kis80HQ#;VjKKyc%X&61&iT7$VYsk^M$$;h(`andLR;w8x~`>{^6N^T zng7?B2(dw6?c>4OCRdL-T=#|D#43{tmZQUqN=m4AU11ioC%0pDb}Sa4R^i-DL&t;0 zvyvQ}M92RfW~CBqO@Cvd9{m-iL31vGhIlsg#~Pr&k9^nf$~Xq89C$)EJ2H4Tj1-2M zNf*=00WrfHUbc*wb7fy2@dqqjKh#0zj`*UDw2b0AFvP1n=mQhLRRM| zqC2rMF51dZF=$uP|FI)WPQRzZEgVc(SNy5$RE8N-nb#R6@f6u6t9Wl=^1{>n_@^?D z%d`cp&-`W^G6w8+>{DD{QA!aJe)BOkmlV}@TNIsTqupZq32oRxYdl=vh;BTNJ+W!^ zMR@z3@K0WpC+W`)nAHf|M4h8-&)X_i;vzM8$IIK!=x8WDc&59I-a9F;>s0Cq1)umB z&I@phgYn0f{E7hkuQ`j(Xjfm?Ni|0;2ia57eqmYs$x_h0*WPhbu3J zommFi@w5W0VLy8ckUpKx?ZcZJmHyLhavJuEBiu|bM@&#xi2l>;J8F;>co~eOM-1Rj zt~N3T%ar;*`+7F||J&DHmsOEgX1^Dym-OM#1A#BhUhN9Xa z^9u1F^Xi}mpDqpRAZ~_#9V7xW^wj^?K?>VUtzT`sDH&Wn`h{z4s3u+t@jK)Yf;^5= ztY+8LO}A@M+*=A4dj6WIqjU0z=^1Wp3DyO2>Z$8 z7ekh2l5=D<4If#ubHT;b(59?!Ck|z;Zh?8KnD0MVS9|FV-Yz;wmAWa|xKanrPG_{d z^^O?~sxs^kPMRZ;V*HyvyJQDY((l?y3ay!v)(NohmT!4%K8yCm|DN^NfY zMu&4WLvZ8;UcBqtjc$}LQ&wLovbX+HND0i02|9#7&4B`1|c-_f#7JrMuZS9`oS(IW-H>(w;*>~U1oRdjg0HFU5ts7@?Ml@+)XNq@=*6&EtM(!*;W4_ zyn;*YUV(S?rG@JChY_Ih5&k-~fM73XEEX|&GO%+se_h$PATF-FQvF%FW7lo7w5>oY+ z*o1Q==VaxeJs%YHnoYcW&RndPS|VS>4wVL9<(Sp3RlG@lw!6at4+UL%DJTOoeGhZA zI*wFAX;alGm7qi$4NU@a1&H+wfIJkj5T&cr6=9oimglBk6q%y40~;ay*v^YSuK{DN zJtM;_ZWMh2Rz8Flig9=EPB_YDK9=Z}5o`MVgUD|PmRRHWbI@hr7UL^&eV+CK{;v0Q*krK&#`7fNFgx~pPZSL{RK#y%QtNKhLiz13@}EYPuS zUR?@ZE$CQHxu}45EDW4XHSHAsJBzYw&qIGP*i`>emBJ88-26qVW4et^<;89ONRk~@ zukX^;7=l{sM(!cvbBn+ZEI#OKF|ZpQRsCn^Rt#AtyJ5ZoDip*q5_*dYi{Vz3g$D!5 zC1XdS)G)$;0Rt0UK$M`0|kw!!`U7e-GT!LvtH!O*=sQ85i4zU0L` zmAS>1Hsv;=J+tB2v>6(Us{#GRq600{5tX9278xgvhBAJ2J7kQ8^JRio3pKRFDf8_^ z)=|S54%VPZ`Li^NYV(t<%RqKg=Cl#2f|Y$JnRl5eypv5E7JY$<06iXgKOq}1({sAN z-!mP_#^h%_vE>4fD5NDcagiHw*&|dEL#83A6GqaJ$}Ya#LUZuCG;Zs7@H0Z`Pai9s z@MWi?h&I9-g6Bvv5Q5q^%W0#M{VwvJ%mz<;m4@(%Hcw z9_W3u%h+;OiVb<$Yxn^?Jq*wDAxlq1d{COfss~v`Qu=t9K7Zhj!o5E|kJ3Fg&_KC8 zn$&Z|7;CLNdCn1$^%xau(W#$}hG9dP@--Fgt->rk&0inGY;h+WPIh7M;fjvyQ0O-b!~++!>L9pp?PP2hz_la-=; z7CR;{&PU0Vt66=!&>;WCfOmpTD_JPhKi=c($|4sDy$5gh5vh98@fY}CxpjX|FJ4rY zbY4?G$Sp8ZJUI^!#-}1CUsdSvM=>3kNV2oA*$@1k1%1;s_ljssIa1ck1PhY72_*GU z1eucF8kMG5$nZ+Ue?_i9ZzQ))L4Pwf4j(mYhScV8hvs7CkY0(U)D(l7b?{6w59bsk zGl(Xa`LyA20@jl!9CrZ*#;e4r^3amvJ`eFyv7$HOPdpZdVgbbtRKA9v<+fhD1pj_``*# z@R%dQtXYh#bDCTSEvkfGx`EQ#mr&u5W|5(H@JZqDBo`=|H`sJF?jiHu5Tbhd{AC%f z)eXxKl}Rc}P1z9BZj$Q8S$>f!FThQ06AMf3UW$3Dv>U5xdv4qV-f<5DhZK9k9$ph- zeEAQa+Fm!QyXL~H8?bBUT~xeJ9tJ3uo}5hy4(IiFQ0h&&?&(q}SsNWjX@Tb|?Y}BZ z?EVtV9-}QbniGo!G_;DPDiy>m(IS$2OPh$z2h2-BT&`zKN z58E=jT*JuN_T?=2Rp=Br;uN7~bg1-?jBxCt{c4Mh9NI1Q72^R?&P)y+g zuYrbYaJD;HUoK+mBGnENX)V)l4(KX;^42n3hWo1(T)~snsU2R6pp0m)@5QGfJtXIots5`Bla{K!u7h}ZQ)Z4BK(aOJ1zU64D*mRhm{r`1pN+c z>gy3GIWrT3$`?sc3MYb_@K9bfe#Yen$F(Jf8K&qV(NiZu_Q*tH~ zY*cy)_c-hu>tEf9=pJ7FdPCxkF(B@$Z-ij>? zG3xwiSZDFa;}rvt^L7thY}Z{@M9*zD-oosxtS(q>Kt%7-)1UBR6*oT`U5Ur0Wq@RY z#~VdOo9{U6xi9bX_ce#9FL4eTpW%Go<-pbwqvaMR_h4Y_mom$`Lru#28Jqp#WlhCC zUXDMs`f2ZC51iN`#Qm#uCzKXuB6;SEP{TB!($aY%DueY`O1}&iR;rBr`%gs>@1=rQ|_v~ zk2DHdnw@q`5$b3;W!&cahVKX6VRPO$Pp^fljk8LX3vt-}Tfg+ozsc$*)TQEWDHvT+xb*JA@t;18%dFbeAo@JDdJ@IoBsaO zke&m!a_<6gxc#Ha%X|dBBtn;|%$yq@rI*PN93luCVR>+1&hkPffVdciA1i{FJ<|x0 z69Rm~9Iqu=rC)}!UvzizeDOlZ5Ee-qV{m|6!LQwM{=cOIpzg2HvuU@miWL-F!*aTm zfQpze3t!;|FB2}|jd!4%6@Um{jvY6z5Bfc_RDmGBnUEtb1t6@pnb$^>>h)Zr z-ztRgcmS>$-nwV&-RhU3?{V^CY>AyFRZ^e=$7t#gO>yHkpUGvoRZ76_;Wls6B4SGL zH%-%YT8xY&zm6h_*er)Xd@b}%4G++SkT&#vz|<%3bTbEFQipFcEPj5#|7PdM)1M=? zgnke!=`GjtNdXg}$pqKYM!GHwn@~kRW^NX%K~7o~25*pDfXAb0iOvuvu|$e=_t0*z zcGNa#_gNBX`#^AnVBtMSF~3mzBYV0nbU2vqv%(EHPCv_QS?DK`%1%L}_{8PRf1RN8ajaRsVTEV5+~VC=uz;0mW(yL#oz&X5 z+sZaabMmEElr~7&uWGx?c@(`YkZDmkoaae}2-|0}qT2(o9SYt;;s5p&qIaS>0h$ys zo@>QyKu&>txT_c$tut7ed&ykPI1qXf2Qabg zCxol(w)zv#WE!8Fk=kD4uf0ccJ{z*iMP?B-{Nu%plN|5)#}dOV((DSG3;KNoZLbk_%vW3OC#n4fP>o-8}v+Ppc}Ewdyx08g$QIM=C%0-8YsGR@)MA@MoW80 zdc9{JR-AD7sV7v;Ovh(Q7GKZ0a#qfsS}dH6^sG!*N{QYRPnBNFleD#`!mR_8yosN5 zSZt_nFpYG<1)Bw7k&DQ&slJTf0X(s{AtnRtmWsscm3;0@1LE$wSrkt+WwCS>IUYJTBD6 zqEp?r4{nA|SO@lOA0}6s>AJGD4GN-s#4WKoUBY=h4EuGT3nHJ&rY&N@=W3Ab;jxMKP>2m=cNBp zyPn#=>(ckoXS07y2|nWJ9MQ`uwQ>ogEUkvK_w^Cm3uXslT)R&T9D8oa>td9ipe%A{+R!t3NqKQXerxNL1xbQeIFmX69fD#8}&@+0|aRJ_IG z6k0pq*>|g3*lt?-zb8r-_TuMT_$IZ{`l+1>$Z84k)-iZyAxf7Rb}-hYioRc$8ZVo}&e`SUv8fmX3J8}1QB1r#rWL;xQ3ol{F*klepG#mY|2~CAvNm~89du=uU2iOQ^{gU_-N%3zaO9q%!8aL5r z4#@RDTTCwbwA-AAl)NXzt&q(EZ2oB?+)7WyZ2B`nRipC1@{dr1I{g(C|1BT2B;iwR z^Fx}QwV@q74PSvH&yhZfZFYo*DdGMP2!zy$p@1ySBY7JWo)thoFC`C+$t7vWKE5il z>+X$lcpEdyUDRBSmY}0j`XCrEpsQN=Gjj#xLS?dbuLtu}<5|Ii-kn)EtE5^uI}=d> zGGR!I>fiT=-w1IYVgUpAwEZe4J4j3=5*?Kz$|521-%x zx^XF%Ae%=P-sV#s$&eQ>-9Tb(?-jm@-B@h*G!Nz^;H{l-77S>YG?4VkuJ<6u3%AzE zLz*3s>%Ty#Q|?81`+bYsn%+U>GO^Cj@8H2;kVVzTVc=Tr!Owt3OHi~>q-Gc{F&8~% zQ|f%h$qZBxpu)S>(AVcu?ZnhA`}^MJYz29T$=$+T<4?B~6rw5Unw<<5~M zHC2&cey%oEB~z?nM>RZ%<;l;EG=^EVLbr*N8988QyisSgzW4_kM&jGPIjxeurPAO{Ufj$l z=stA;E9g;oL`gqN8*H#}j=iBYTUL}4&@@Ygl1R(6nh=_{N|9;)y7iZ3kQ_`J0wLBW zD&85k+Cp~-GTIWRE=O}Z3za|=2KpdYeHG35c=U?C@+Mu;FZz{t@%axL!=AKbmq(n* z;g}oDutV^;#PN8FXnwRtF4KBGeodFB2n>XxU#bo)4%%5NL|8tX751xKHx{>;)HV@* zO%y6*YKCx4)ri=>gwWivLNBGR&lxyaXW);8TdKVDj$)CZd4F6|4NwVII^%3sMKrz| ze!RR%$&1g4k^!pFOB$cfU|F)S_!cvaAMdni=S9a$6nV;h`P>^_k0fyzo`OZS=JKD9 zY!g!wyE`CPi?!DkuYatZ`#y48E8!z4j^VUC_*Ygo4@#Lg+Q~yENx^{4?e;IqBUp7o zbvNI{5zPp`vo5Tj>Q(WvcFEaF~2njVE^iA}r5;OG4 zNC~xMieC&MPRmuv^ZU@nwrP`Lx;)6*@L%COVm;R%74dxwqderKUwR<3yoxL=Ublnw zwvygg#}oX{xiF%t>`9&8gmMPKm(bI-u_q9ZAfmK{wRbz%%Cy1@jgo#lZgH|WfPo)WU){c4@W=qy$1 z8h(T`0*l0``QaU*GvHjpjIq4S4F0(m0HV`YnQkMr?R)D>0s8 zo>qmbcN>C-Xn}eIU4DZ6j852f+WTNQ<+4+*Q~CP}DTqs|k-JzyAlVK~BPBS}xdnNS z*??PFPt+h^MtVoofq7mJS|4}0LmHR-Ow2${@?d(K;#MIQy3M(77$BeH%%9k^!QYZ? za%8s15D0=#po+h{Hay2>@@i0Tn%hws6Cp$}oT7Z43t8I)=^`S1AeOEm5i9wcciAX< zntur&jXIuNZ45cyqJhW5Dt2uep?vLpC+E9NeX{d;J%b4Z8_$)we0xYIs{91O%5p$e z>b>9zlAd<*spY@QkaTSRprI6Ol}Bx&J9!$MVayIp2w;sDxjwK63(tU9_Dzk{@H9T` zjIxaIw4igET*RjCAPeJg{+rQD#^5ccUs0$0ek5>Jy1iz#L@+Z*;WGGsEN}w#oUZeD zuPCTGdbXB!^RUdED$bfG=-^YMHUQY6o}^qbX0KAD!l!r^Sn215cZ1aCZoLiqtHP|H z0lB4DKzEbKlb9*X)H?< z3x?!ca)}|*+VI?f1?Zu)FxzSIc#<&|t-daj)5ooFMeN#BCyp30T{o!yvhI)#8`Caj zMUsWkoOmIA(^}QqQ)P5fe1proVB~N;;(L9*X;Y=gZdBdiSN7yRbj=b_b&*uNsx7n3 zdwFq*ktotsbLcLE$O~9SILm@o?%MozWiGCrrx@M}A#3@sbrCM>7gw5tD*qVJL=Q~# zWWlwWEPb-S@fyc9+k~={chyc`BK#svPn}-Xb6GCZ$>{@N>1OMOxC*R*i zliS`omcI>a3-2nIHz3z*K^U-kg?bhps$~P}l3?>m3JZAbbdZZq-w*m$1j7j@*;I1B z>aR{r2G%+Vw^r~E9ngT?{%ynZ%3A;Q%NeZI;L_>&#bI_q%s@hXOUn*kEVF?$MWp0T zhT_Quae*qXGjPwf6Xt8;z*aQP)&zJ9mGJ$mZ_Z=B$_$J$P08$*y83igdzbUBC_Br% zsL7ES1{+d4$%qAEmMf|6*CC9Bt_OM)AI}ym`^=m+%<8ayW9k#zORZ7N0ml+ zQ{mZ4jB*4&MpR}c3w{!E_dv-WvX}U;^1Kf>T-ZvFa3V~s)xWRtqKk$UOk{KT5s9|s z7PJ_5S_>bwpk(xtOtC97GRFp+U+K~AI(EOT^(bbfs<(QYUXQ{uaiXHUH|h1BSOb%{ zts(rdUum(q!h~PU$1Hv)c)+t0$8mvwO)RjLWGke!l0SWw3TT-Zjb*LXW zVBfAlm zsLD<)-Yv4KF4HZPQt4?h(~;!7JgYiX0t_P8Jh4DT3I>}K^@A2W%XSb&2WYUNmv&yN zpbVN+nHwH9?xt?LpXrVVtgOOJChgezaTeH=L!hDZUw>pt zr)@rhljX3Ly62-I^`rxgAO{Yuw??3j#$z``&QA!lxG~sZG9GwU%b^A7=vkZ;a4`wG zKrY8}|KNp9cI8SOQhmnU0YEIXbGQm?xO$WiYQ!ujf66N>pyo9_#ji$J`vh?ZfrkzX zOc&itg+04?;V~!$8mt-!4wg^86aFj4frtUNy}<)+#A~sLNl_RZ4;>X+chaf`J*9U=di(2M)%Fe^V5KkHweHI8e&lHyM?WiSIZLpHzB6{#-dQ=qEG-H}i-) zbU*fYU_eJ`yjwdAeWmX!10Yv_jp~ezX;QT(KNa;#tS?7as2Kej(*KXN3T27?3`LRQ zLWWGYO4!u#NcekYRyz9lip1+eqfiFcTRPXrsTZVpfns^8J2+GKXL^A7e}~Dd^cAZDWM79Imil$T>?_ zZ-%~{&uwOexwI#PA+jQa=_c{0(el?pf9CkY${22l0nUB+3!^#A+oNocFvA`cEl3vq zt6u-%OZW>x6Wc}Y=$P=^F+_WkjDiQh4k0O8H1n5Pda_X)HxX8w61>S)r6SbdulAuX5u~;5rT=Ugd9wSqtth0 z+S~NST*D8b>0-_iB=(;3KJ7zS@>d+*<+Dv;Pl8j8TqZsDT-^4v(;5pNB{8O{tIb`< z>*OkxB{2*q{N(YB6{Z%R()^6nX=9dl>!r_2QQFcf8B$-vi#UJ|J7}iRgB{FL-n#)x z#U+N$;NLi@HcasK1VEM&#DC?bBz`1+o;Z)T+SVIW?V&>d#U>b5^&3Ff<}^DJ)A!|k znUfw*`D+|NjXiHJ9-!S2IHp%R_rw4S#~8L2>Oy;xQ$O%R)%%=4`Z|3v8E+k{U-9d& zL&zS4HpOpOg65=Z&*S%WqgC{WUCv_296$M=rez0|x$FCouEm#*Q06 zFo*1J+aUQa=l<%(kU_X13XX>2?P2{nv450I=Yld=@5$EcEmC?>(C)P``e8nj_8P#M zfw(rwk0tnC^N8N*Pd5VH5KP3A)5Xvsio5`i-TPW3#C`JCMYEP~JtyEwiw*Oj2+e3` z$a*3d`Qf+8%dc2E1=pqVXXziOWB~rzM(}U0v!BFbS2!SR`30jhSLL@Ef@oP*Ng%%d0=P$gEFQ?5)y%}W66q;EWJJCp6awXKzzzRsIoM&_u_ z0}@u@bM0O3BT*T$J**1w?nVj*zfkMtPO81j$U)2E&T2Eca2n8Cc5-rW;G8p`+7R|> zGnxMJ8m&lit*IrWM>gRXa%H?Vf(B=>`5XWuI(ytgo!eQhS2KqV3fnFsNm(kXl}7_> zb{rSK{Fd5T#{TyJD05isoCkr*`B_$U!w>lykEp+OHl>f0MPGwOKqR0nX05VlQ<+qt zGQ=+`WtRn^RYBHR(8B>zT$2*jKtaq-c5XYr2hfvsg3_=Y)qv2GDHcb?j;~@vat5vf z4h8lq>x-BAEh1%{iDe*`ggcCaXca>opM3lIzOV}Kxv+_W$uf_iBbQa6TR>Oyvh9zZC{;+#_mGl zfM&!-dSkk>luv%~+D_RM>EZ>Nm0H^BZ|HE6=yT9oaRt>)qV=!-(w~VrTTFZ_)kMb@ zs*nRiCGPnG3>XBTdI??3UHm?K@+$|~KdozXhbx-dGeDv4yP6s%qGem@_MaCvmt!j^ zA4n%c91?1ev^{us$bUIbC%ECbQK1v`_8O>PzzTFFnqmRQ)$ z$g}PZJ%slpBDC=oYbOJHt9fvg1%XivVGSf7QLTN{k)p`_!m%GK9>RmgCZ#*9Ru|UJ zhVSl_I<|%q2VzDhMMi8cB{TVsibnfkk0Oj)5tN6>XC)P&99=^%F{qTdZjwivSe?8$ zQT*EgGMO4izf^-k^3M-piD7VL>RlbEy)blgj4;u$>Z(9Qo9FgJKrdPvlKr_bg^htx z$D9!ofMM|C>KAdQTV5Q62l=e-?1)OcUkXAABF&U;-nL3l4%5cr}2=tR8ZWZb`#aJfES^*H9$MX`MFY(K=+e(z4fv1f`moY z#q<@lhOI?~%arP-t#bQk|LSu9rCNAyo6ZCE_&GrNQ>0~~y>W560dOr==S_}aqyg@| z;bZRI7=(HvV z0UHq*a(whC+ie#5*;K%O-wU__W6lj0#sk*mS8Pyl%D8zfi_0LsEN0G=@%JP9Yf?3p z#bn6^D<$Rl9sFC^cO2dTqa_fRvw_aba0?=k?Eq|Ppg`4~zb;~z-1tRTN#;j_hm4E3 z!*%lA%SAsQl(vnOj2OuZ(Q`1IPz>bOSE+9D?DYHh(1*<9P!ISAdQymd8*|kz(<(lV z#A)%x&XaWoJ|-aPuV-dO;bd|WQaT7zw=B?*62TlVTby)r1KwZQJJG zQ$rYDN4do`5(+R;trP*Dix1+i#KQdwYhLxmuR(t}bQe$gwpfpTGcN!~O&=4>OL6fvFCY`a@EGMp zY66pfcU2bri7XiNIZ6VWIc7uIw@|ZXJIsK^rmg&Jywz3y^5%P+=f!xkb5^8}$&!!2 zkCA=bn*E!A&qy4E&wJXLk#(TjNf*6Pd#N7J0M0}61S7M*roaHh7&0^s-^(Oc* zMb%Pe&U?Dqly%uuZIBgD9schRm)S_j4-B-51(#H)hi8sw?m7w_Tnw~8k)lT_p!k&@ z9<&_hxJEpW{NJjWL|Oo`5QJu~KMbCGjz3J92m}(}j1UA8`%Di+nVWC-rf#urboOQY z<7P(+qpnS*eWgL?FPF=$*RyW%@(Vd*NgmDf<~+CrtI|U;3`#oSWjcT~*?iFeOFoq^ z!CI1icIG!;Dvo?AqAUk(I&+$eqoo)|?YGj0a6sRFt~E*|@PH&N2C`n+g?!6-)QM6j ziwgQUIgk5PgO3H?Q8E3m+q_v;~D!@pSWNq%*@vdYRm?wkUH`XWXs@~sP1KATTx z-fvJ6NNS{4nO0K7D=C^bH~Fko*%ij7QvX6a4bO!9j)HxA)@T>|Rp+Wp7OH1W@L0*6 zB~kOnuQ;Qq^O~j)qP5Wi(e?52wJn&SPu4$tJgE34#_IZ-;F2`ae3Pb@3iw+<+?`@3 z9d4!BcYT@Rj1PaTZ~NvguLPM4)eRDdck`N0)Q99P+RD%;-SYf)kPpH$X~sWOCRA|S zof^V-@>Vpjd1IUw=n>*Y11fdNaH#!yHP4sv?AbL{0Z@l&${o;f>C&Kfy6GebDts(C zKBAaur`zrM*tS1x63uW}bpsXlu+sl3iW*NwI@-AKPy2BezNA|*wMMy)7v4P=dxQjB zf!c^}s)s}O4Q}?ejzV!gk4xV3Ks87cyt1@^DKVuFIhTU@=fd+#xP9^Cp-LLruCLA;kC>=I9&ypboUPBH{Z1fUrxB*F>QZ4dW$sj|g zFxNLtF9fAD5(_qiD9wGJT{pScfvAI7lPrIJspB5&xm}c261~ius1U~+O@0*h#IVg@ zF!Tm-FhSca=T0~WC;tk{9rxhfP+kNp0t-g&~Tu6z% zPt|vs`Ve=zz6<(IC55xmbYVZyP1BXli|aL)SE7=5IQJD-Ctmf4=)ZEMkvC>6(>>O3 z1S=KJPoh~EKJtl%b%P&`OZiU9Hh4Ak;-D7!oad>OtO$TQ#=wIgCU7e(<#PpdR-|pN zc^yY3ek}Yc2{}mcf&w)W8A6*YZU|`BOb%BSJ^EhNj@oH1K$CSTba;28IYD)CVs*)t zn7q_xH9J5PysVNLMk(jI@Kg!_3di!F{6ChyGAgd7={5=OuEE{ig1ZKH5AG1$T?cn( zkl^lw!GZ;Mf=iI#?#_3f_ufAPYkth?bGoZ)*WOjtcT!l6M)K}V*$)nP|1$+RKA#Kk}cts-3(jXh}%g6W~yHhYMvX(x#zRlm#``@ z;VPVpApmXb^)=Qg;|i5Eqa@=X_=#0LdV`WQ4ZNn_a~=a6e?z21(lC`_J?e&e+te1f zcFNR&sIBtkw+@#e3A>|u1vOnx*bW2nnF5t#;#X036`zW{sEHdpVBSu+lP22(y$MJG z-7AnfV_|+XstdczWJT!0%MJgq0+&gJe}6OFRw%9EJYx8Vvy8YytsoUP0%pY`<-Tt& zbzC4ddF1^PITVjm(x=~RZWji08Xa_1e@qzv6v=dTFWE}El*j9+hvvP@bhB1;tDd^+ z)}ZQF;NG6b)JRkL9z}JODV+EP_;2!8;fUN0UByCJ|54^{!F`G4B5TGwdbpD~v~RiU zb0z#O=lQ62JJ#}B|*8TkF%7yuFGFn0! zYuxemfNPLjN($6nVHPsfx$Y~nPQnR0Yt;sfNPw}tUP1@u?}mEV^ZRdYmOim1oP!8i z`+OS9EsWq`<=i2oxrZ+78svfUE87Wts0Ko^#`ZF7n)vVhR>=YmmvLS!ifkE>cjx7y zF{c~(CG-}{Paa|oxNqnOQNEk0H_t^2^Crf2-8ihl5_~w;id_oC<{WzJk;F-(+5Cq( z2>FaxU@HehPM06940`3IQA}WzPVnhr1F_@cD zoBTkit@hNj{#VnEe~tAL_~@B@e7V|PFf(;OuZO%Ne$w)B4vr1$dBcXBuNv)a;G!4Z zl^Bn`CzcC9zfet~FUL9&yaz+=QfF=6v5E<}^M)d85=z9JvXz^`X` z3E7}pd-pKfDq^JziEbyGpOj-PMx&*3^Xk3{@5q-&No1-9?qnAApE*rs21HP=CYaOp z>rs({^yb10QP8W7V?My(b*tEwaM=vB6tTY|3wf7ESPy3NpMW($3M65=hq(zNsZ*|h zmS42HiE+#4U#H+2gCS`ZV5=xRwM`eiQo!%$zS^BXwFYV)j__U~A2?Qj@}a}oNQ?|W z9#8coMtq((Breod4<&wU+hDp?0NLhLYPFxCG&83O$gG~ZTB(SAC*tZZRDs`85q8YH zoBFk>fA4W7;B5N2Xc9R3C1&k2EW%`LnF@9Qg@BiQ=w=@vy*t!Vht<%EI1-ZmoyOb7~J%Bd_)7 zVG>rjxev343|vo-;qO`qCQIt9?&Lc$gm(7?0u6h3eg_p_lv;F`wf6X?u`2q$!uC2( zKSYS@s;s-KwuSua!PwgvXY{z)_t&(76`^HvuDoxrbci`psEgVU@eaXy3>LKf-BmUw zRP{+bXZ~VFBbwt;WxfkV^(4faVMOC|dK|a=l6=cS>R9}^vJ|5k6xYL;@K4V`|Hio_ zZWJOnCoech`CbuBOGT)=8CDohAnjb>N*~0GzusdE!-SimimJA+#~qrciik75nOxfq z5vbmt1IjBpGn|)AJ|7f$3jms2==K)Wi ztVNM*45CJgtN!)AZXkcr!o^ke3){^*U=`9YeFyw@$L zryeSVrI9Njj?dg**QF5_Sp>>Y^D<6jv^&Z%HIAlBB|&_{|hCd+bSH=+_y33F#o{$vWFD`u``M4Plq zE1!mm8!(@)GV~+OHyuvIz(GeZx&*DI?NQ#>mkE#rlKs`0U$<19|G>wJGuRe?Reika zy&qF#ff6J^zh;uqgR-CvAI|bEi*NY%F32Z~?=oma+y?U$eXrz5leU_G6vLMyo6oUm zLf4&Y<}QLr1W$2|xP@Mm4aBntmF1S!ASvWUodR^r72xEJT8j~BlJL}c>Zr6Mkgd(e(7xhx zB(4PBWV4wQ!<^&Oz|!!5uB6C%6>d7bRktK+fIfwmqOCQ}sdy=@oW%yB{3t!n<>Xt} z8Oa*QinADqSwlPr8Sps2Esbb$ia<`$ObRVD*-K_bvN-$%#51JQG&2E_Oh&fDjMvcC zmm-fWMz8HK5Eh4$B3`s`9@Keh+g8~YhU}$Syy6u+#a-}Zy2Qsa%R;NW?HtZY>6johKiU)SrHK5e!X{C>ij!3S%QApSsw(f6IGugzi! z>&KQaE@HnDD?e6s3jIn4?O09ZPo_lzt(asH8$H(D`)%iyL8z)d_KDT z(H@wYVc(0w39(#;{)w8qR9JZpQ<1D!tTjAraCOc$-H(Nm7OPx{n}ND;QcP;UiZ=)5 zrQ51=HI&by=e;dZ1y6FOpUd`HeeUPJwb$lyw!Sy0e9qQ3_QGWNEQkNi_`omfRD$NJ z(1%aujx8Hu{2x*ljZz)kAJl&(jb|U2-|$F1bUZ~-OXxH3obcrj4I@AV7&(%U6p4SS zk8DbTf*maV_#}o*!ERbQzIY{2+^D$`7kk16r63{xQ=Kqx8WR-C8*d#1qI{!6#|L4R zS>XES{P!#v1;?*2o4#sLP7BWY9t^&e<*l~q)A}?kQCxUKk8WRM)7x4MFW)V{gTwaL(6M{};3|h;m_xN>6}| z+!8>n>PK1JJ3x!se2@@hq4N^DsvB!>Z1ufVO;%V*t1o$MbFXT%l~Kw_tMxYRJZD9S z+|Bbr$yaMOc8hq56zeA1aqc)<3x*F_?<9&aqI_%a8zjQ;L;~?8hv2 z!d@I6LqWF2)2|gbBLhs>JSAJP{gRrWFhzAzOgm>9&B`)#I%_>!4U#NoDb@yNDFDfrd9273v(7ut7HC4E|h-=oXX8BUuiZ9w@^g6Q%k&tcIw^^ z{f^>+EjUr04Fnt_vaBgo1=LQX zX};c-5Izl~1#@#R$%L$x|Kbhn>YoD%;;uGsJ_NyP2oTYR2qu=w_UM1zhhd6>_=`Ck zIw}4tw63_gj0_RW=K>{mlt5Us`L`1R6WJ=+Z*_8$z9FVbkWaDB8yd5>!4%@e%aJsK zJ$uk81;8wGlp&sbKSH2t?_bt%YDnlu^yH?`1dt31{!IWw2@yqT2L+S>W!b@M)dF=wn@<;QU!}tjmDxsGD3`eWju(g2@cFtiqG|Iu`e&7csS9kd@%$|D6N+oYmwwaX`~@813bM8{ zppQfWnF<_l-bWZ3yISuG*XunUWZ>~|)sHp$Eny!Gc&oi6fpyyL2V2kGQ)Xuw2}}K? z7m1wL26R9Qb$$a+&_?c5na@IMWafFNqEGS`;@K3SyiSUt&c6)3kQLN6ouA#i9Z( z(GJCZIeV{YsiUMUKb+dl)0gf=o-NdnxzPIlcl~V@29Sr{%a==Oyq5eRWSW3<6m2br zWSwi$HAl+WW!KjVOkAIZEr_Sc>o&f|o~Ulo4vEYIn1+r@~L0UmJ~y;~K3S zJRbJsG1*|;U)Y3#eDjz@r!@LVsa?>a=5N2E3vPE+G4qDyd!X#z$WtRA9s0KV+`q8^ zY;Xe^uiaLh0ZqIaD$c0gf24r_RVrV+FPU9A@DK`W)RRVw#kNsLVgv*9y=`k5sGpt7 zrf|^a|6_jCsN-q1|G6u2&E)W^iD4OS=So^FmtpsZly^YG_5c=UjHrN&sQztBW^^l@ z_)Qh9#6w&|U!!HT0;PbuqBa*7#vrZHM-ThNTO96A98x6Ysm3rB1izTKbwfxRK&Mk( zZU)$r-RlO^AMMJTO2QcYtI~__i3p8&Hljvy>bdLAGoR39FC{uz!C40x*cZ$~meHL| zf{Zxy3y}K5ENLWhg#5| zMo+k~3|C?xsXt#oynAwg2N`nFK$VOvr2=gm+si!AMU0}G4JzMZ|F~X8`MV6;g4S5q z99VA!?PLwGU(3DJI>1&R6x7yOum{_P$@;3wEz^d+sJ37h-B~DvK8tfXdrd@)5%#}` zT0(g+`^ha5iEBnh-T$%i30s??l;pwQFG;tGvZwVDv!txBClUpn!baULa&x51jInoE z;+knGW<86cz})8E(F0qY5fI9&aB9h4HPbQ)F=^ZK4{xl3HIBk%MC+?!h+?}??`{PS zBa+BzVG&2)v>}=v=bPS1gSAQIcOw{#12QtJhvOyE!%FirQ3fpn_ zxl7JAsc0puvO$(y{hl}`)%Fa8YYcA>IkqEvKqN%BzI4#ihAB`W$MdAX!F!jNB+Q5s z|HwusWj%P$kG&nN$BC-YarP$lL>_py1~2G9rPU3kDS)k*i$|8>w7Z;5SpBFdJZ>5_ zfcCP7xS+~!$Zr4ec4crwf6=w!g(DV1P3Df`;_YB^FV+#Gy^D^#c#B1!81trc|sLs3wIG#>0n=W;! zcDn=|bHAwDldNP2uQ*kE11iJ%kb@y`hnhRYDG|&qC7L1bFy0bgIzpmgxCwi3`{zZm zR!Wp%I~%^NIKi7$Kau7;<$Z$TD6V-?FkWp}OdW~5Rohxs^`E=W1C+V5&_K+3+!r;N z&`s4U#}3rE$B_3ge-uvgRJJBh!MAzbntHg*cMr^gE*P-`&qCd)LFPKXjy7V2Hq2d+ z%G)kP3xz@b_9v_eBlonfe1$%E;01)4@|PfpGk52kX=tvpk5V1R?H` zqqu-18e+&LNbG_y2nLYYEUVRCUPQ=5>>6$9wWp5w$nS8UaVF8TS7KoS{=7J^wy+p9 zbn*?xt?Xy+wPY{yBbpFj@|ppY^)Fg4K92Xh1ivWh`<{?E?gjk$#Gp7>QHVtYC=47Z z9aFz!RNNCH+I(T1+dj{hA<@9g)%Nsk8o+?5g&|DBcf#QmW;LV1&x+jhUrt+8^=`2w z@-Ry5I798BV7lCSYY`L!a-i)QfE7Vq4?*kE<9F(<&6tKh2|#H1#;Rcb=`PqMu)n$N zuS}TkTu2^6A%YHOI-f&Z0n01`fCP%)BW!s8hL#=B9$gR5nfIwP-(kN7ET@!hBtnTa z7P49I{?2yRQQN^~c*x?Qona>Us@rP718vT;e%{EO#eAox@3QTNK!8Yd3-iz0kgH*c zdXww>@oIx4+i@X4QD-%gc9WZDcMBCQaP1ASfM^1m9b^YB%_per;SVL+%4>!np(mG@ zF!$Mce(3GYo5m?JWk(V|!}W|g)!M5JqlHKMV+{Khf7DHfHD;wkdvV;9zWDb~C8|`P z8}r&rf+1HA@CtgBpG$?%lyeZ2Gjl8SOfxqrNxsVB=SkBT0@JsO8LdTy1%vyKkTj0T zw#(~WGH)4G3C6Y7nctAI*NQ!)uF}iZVtz9J6l8+NC%Vd!$iyLoo1OmT zGIf5*Js*N6PX0@`Wf(L5Me=9-6b}{Vl#Jt`XOHm*c-oi7kGjlIH$zFoc729HpwBDt zVaPOE74F_az~U6nu|7X^LtGe+jFeeCL{ZmG-~`uYtD!O~7iw%lDM`jpTy67_1T)*X zA#93%HWi^NK%kCMqn_oY{DUC;-gf}@DcLAgBbX9=csLPXPt9p%Pj z0e2;PF;%*%QF_Ay*hKUxuud7hje}AIkHjyyGc|&6*eU$j$Y!8+bN@4Zyl5$9TpKd= zWbWn%tgJbDIF7aw)FwKz6xEFOOF>`qCMeb3sIrj$z}&IYR@n#xsM2vMnc(yD@d$x*fGZy4dbzg0E6a z=+?KXwD>k1E~6=^GbbI&`i%%mY5pKnK{b1Gc6$RdE&?vHpH|OOG~tD?@cMk*i6a+v zPM1<9ZpvI+IxR^+_0F;2@T#mjg--~kF<$b-Ics+H!>2E?+{7iy!nQ?cJF%w!Wm09p zQPi8uO+Ut&894i{Ewqf)Qi|Yk{Q$Dsf0(wt;8YdjJ4FO;6VlWjHLLP; zC}@2Gp{15@f3l2y#5xzg_Hi(QkUGrc{W`dtvpNFFrcF`u?wwLTm@zzXYsic4$$TH- zB2ZH-i&832rbFP8LiIU`bni)WUlw>#e9{50$@h856+NN~|u zBWDkLvl}Ky&pw%!|3%`Vw0!`m{{UhWxdH&@puH8;X)lpCfAYF;4?ARt5*2aGzUy7= zDpQHF(3uEG=-F_80AL!aztGe#Qo|o`v@erSf2WUcZf;vU#)Wjp^~rqJbaDK2@@h3; zvXdBGjgZQa#KAGR*>awjFkx=~L>w6TzjZX-8QK-iI6U7)3)atCt10r%&<-RfYU&_k z^x>rsP=XA%e?|)=msGzXSF%)|l1wd>mrW4e2bYZb1`n$RqflZkyvK(R%zr&MXosyzw6 zZa)O9`$xqY>VM|S^FWrgMVX(leRdE(t6L*`sRLHMeC2lY3)SWZEc%MfdaZCu z795NW7iR>Ez(HsunvUJ&2u(1;nfR|Di8nD$77DmAC(!R*d|B&LgIf%wa7(3*MQH#c=Ldk}6J9tJ~_ zP=cOL@;^W`FF#h?4$?D=B3k)Q-p6~(bb(}1m@hcL2K_R+N;te*nkVGj(Z>!+xe%m zVf!;q&mB(D^jV5{Ubi>|k}nbx0sN~9DG8ymJdXIvBUM8XNvy_NL(o$P^uR{bP~@gS zxn$@$T_6LTl5d@THLO7S?`V|&8AFT7zwBH*4`KqgDV&Axhdl@}W83Z5GfT^?l9QuIl$F@^*UBG=Sb+v4=^R`r#ykq8v9YTw2JDWZut1jkW5%O5>I&Q6Gp34(qGV@vn!KsB! zAe(c|2Ib00tPbz@g|a5Dlqc}~hXs+`eU0a{otJh#!1x-o0oWSW4*RRASpaq9^uX^m zwNrZ4`<1{3px$GqRphik^NABo!?X9?NaI-%M=+&we-K%j7yX~iq5We}I-1QuT7`3z z`wQ!dpI_nlXk8L#8+X*8MFYF%(ByWUx8fv^)cSnhFsCF@hJGpG5mt@;xTu6$nkN4! z=w{%k@0@8=wyhkgGRrE8xz|2mxT()6GLFESTv|f*+&lm8CRT5befZp{tktdqo407y%(2&IR2^g+CV!J}V_)T~1U1x?x4t z2XYe-{R$;8-D?1w+Fzf8yla^I6~9ZkE03g_;J7+h-pRlr*LF2}u*((`%ty2bCzxv0 zxQXRHM@*tOh-$GMooZ3E;mTk9Fu&sCu{F=?@t;ix6@J}bUyXOh%*4k=&5RI}! zXFBH5h8Nv|8dn2CuBqwJyO}p=O2POqs8cYt8_+W%oL?nr9IcF} zI0r0xn}a%6gCp7REHs}0kiRnZw>urAC+$a>nVtkUnERWw zn_v(1)7xe7s&|oBhGYuocWwD{K-Q$Jz~uaHTnP+&L+-52aMJuYQ#U(iba)P;`JE$4 zJRBBA-h)xg`J>5OivmRidtQ_Thr~9h`9Cn7{XZ~G9Y2ePr>Z4r;Q9k-22s%mvZ81C za#-Ji^-&z(${}zzA2o5Ur4=F(?#k7_CQo5hnj>5SqCZXIu}Ns;sY{ay zlK*>oG5e)25-}dXuP{Y2x_@ECSe@{$zuDbhNOHAr-Bq&xw_-RoYM!-^6|hoRpB;%x z&RaO~LKUT|OvFR2UN%@5E&e&-fY|64zu&YItK?aZdyHvCs;-Ls!aj07`af~ixZui%W5 zF@&Vj(iD!1`A@0!_OA|wN|Kp@8>2*9T`p(enF8>MF?NaPutFa1v^4rBtXE~AD;aD<5HE4tDw4@9k?!Srzme=b4 zFo8XMb7^-~#$gXS^9edb5|dO?ZuyQIt9z@vF^^0Us4DzA+FP~xp$`?{I7GuT8kl+F7$zb?pOrD-n=YbM z2(oJU_Po%I=oc1?KvG4n$${57a-|o(QtQh30Ee-GWl;Tj;<$~Y_0j1cAzMxBzNDdb zji7`d^H09E^-<1U&Mkgo;FbdTRGS5rK_}tDYWXSIcHIBMZruFf@*z+fCX5%bntyEM zooyxRH&s;h>1dJY*J|#rw<ewGiIya)TN5;!84YHWMMhJn2uX{3941F}}eG`8`&MzNNcdEMh4p%z7z(dCF+MVu?yV=fd7TBteBoE`2`<)&mE3 zc2yEgHen*nbR^870%uC4_@fPU+{(wSw20wE-Y}~-DBgZ=ShW$bUwG#uN9BnqkVcXy zr11&nr`u}IXVaYMAt?Sj%b2D*slv4FBS}j{wH7aIubJKTw~e#b(UM%}H67?i@3L88 z1ygil90b@Us?PMG1_v|z`l6e7_z5}bS$yInK?g5wj3JKV$z=EPgK_wSNrJpJX1Weg zLZZdre2=}QN}z+3?rp7@a*<^{_z||Rv=X;n)XORaKjW0cC=Ut=21}YTi7oqJaq8JK z6FHSeRbo!f0nmHDHBzUTDiX(dYMA)OuN5ZBv4Zf7MF2> zo)os|yGKulMC&YE%N{U}P*<;!Pv{Dv^c#(IYxa|V9>i(C{YZ=3E_EYYuRhj%qwy-h z`U0t`TjvI1ss@86RO3CI3r$?l$1l#Am!h!ErW5!lwHjw7@TLhsR1ZDX8d@*!y{5h$ zIBz2G{!3fIjU^CrC%K-ZHi+2`Fg?xA{muL4nz)8i%%A06>{pYWt-nU%7Y@+ypMj82 zXi6!KS~VF>#E0(dc*Wd}bz4jCR8An}9a6Y&I-4q<_?QU470FFL$74&M42n1J&W;-M_>yljxx%AS$9-It;Pz5*ej^}Ee{E?yU+o8 zK69oN@;_wlL;FSIzvaL*?d>qk^|IEEVEKjlfGNql1K&s2qFVokn4(Fn7furpdGB2H z*QFN99iBB$)R60fkIQEl;VTL(SJJGBoLgDWBKGSxz{oJbchaGB14W8(#(?Cg0@Y4q&}1XOom>7?;FccF zbaSl!k`^jX)_wl&?5)C=o;FaMS17 z!g8E6_@je-+850;E;9XDKZ9KS;iheG&R?ggm=~g>f3M@vllSjnq|C^*j3dHNfQqtE zf6Cz+7>Ty7$EfoDzcS&}tzw<5qd0TK-r7G)eHFp!k}%Ju5?OeSw6+t__SAcilH=C{ z$nKFZ7tPsxyJv3s2MjbS4wX@~(P_wJSoM&VDll*#a{yJ63QT0`gE)kscgl)Qg!J6} zy#?~$iz^$~1xf}U(B646ADJR%34*w}ZErU&EkV=~cNMQ&LIhfr zj6iBSIcO9O$^B(T`kYAw8>p;9q3o5PR5g0ZG+n8&<%MgmNZ3}74VN_qZ6A{!&HHI1 zI(?`r0eb~#~7;fgTq+bNmNiX)H zd5X_-u;({AF~PL5p8V&qTO}M6z3455ROB<5B=9brV^#23j`^Uq8K+GYYI8@w^Ju}W z5PehSu>1rOy|7~kGcdr`@Hf{V!+@@NLuc3;XtqSFMZ*N6{`W` z?%4^4aL+{Z7PwdR(RPUP6w{OJxro$t)u}|?Y9G(VJ_)kw<@%vH`gxO6|K~(O8fW4% zGG7vpWVxJo#Zx!4V?F5Ue-x#~e)?0)z@hgQHZg*0(MkhWq6idzAz3NI@2idQ!CIHW z;p$9&#SI)>E3aPx-%O*Kxk+|Bn|l4+5=1#u{#vBVXqIkyh#D5sqRfD5CmgxR_%kXdLb7%sr#0e-$)UGZa5ctkyRp;K9V_qb1Fy z+WEIE%}SK-WvQ8TSJ!v*=OAk0_c}U%H1R?V2Hk7?JK0kpj`yn_ya^jGW)Ib)!L)by zZ&`)w&K*4l7>9n(%}Q+SO#L$7bq@o6TJZbSF?VbS)TW)l30h9jayO*m-qr@I@C(=| z#=U_nHbHWYn;2G?CN`h^8A|}BL7KC)9x3kGrNU$QhknP40HbRx1eOZ zcU+n>`ZqVmO)3Sa8iw1y0}B=Rz2R5l>qR+d*d(2*`e&##4Y82`&r@>QSh^8CM5D)l zf$E;G5=#3tYVTE(zHT3Mt`$?uAUMN^EX;0Kie1PVN~>K9N+d9jDmOOyY2b#d7)m^I zv_f&9adWit$%n?Fgd*{gd^0mg=xrFaxN08T_$$F!6J}EBT8=V8CJxG^Xnb)yA?d+L z!V_7ot&kS|5sZJWfjn$czc$*if`*s|11bU`hI`VpnIFYw`d+2NbY3xT>YHH=lHTF1 zK>FpEmJjY_HGeP70kf~$SlBNmeaU6BqxnYqRzFpB&BG9&H*az=)`uCH;f$V-E;;Cn zbRKiH8&LcU?tF&(h!Taz%xYp6$s$B=sq`(ld;?O> zoq~5nzvBxi6`mSSs&|D(jH>5>Qny;WL`SelBybCY5yUS^F1dCWjD0{=kw6j3w0Ecz zOf{@>M)gcrSs{}QaW-wM#12s7d`v zX;s-@yR^`?n`>|0%NG z{uu9HF$w%sX<#2Ogz*VA<#v_|@;5G{YgY1H6P}s})H#0qhiet^?x+JsQIM^$y4T|S zGq#6_dP&}kZc6Pn)!gL~Org-&doH8m)^q7Z+o!cNT^84iVkYdxBwB1abNA=VFN zYS=ORe$F4D!kz`rY-kUE$x1c&Pj(op!YlkM6BNR!`RY;NS^72csCYs;vBx(HC3Mu;xP zzr$(&4)Ai2(8Gr`Sbj=}{^_*Q2d6yqx{s7pru9Ez!SI_=2Ca*wA{K4}Xe>kG=pntF zn_I`lY{Cps7$%{AdD@i%zOA(^%ofTtQWB`KdoBnox`rB8EY-h!G2#RV7+=mMyd{5&=FOBsBovTT<*BO4?_t|z0WkoC>yspPebO~Y+!KIpU5}dmzTM*tmanlSz z$kykWNGDNCXwBTjOr|q#D2mp)r^pi-mp}rnOjrhP9A{mRd3X&@w9D}Malu{xEe;Vt zEjaQ?%}mT!2FG*7ciFBP3Nb&IxlrUdE>leo<&6|qae*9N1JPK|O|HUYkMU-QB}#08 zqEjVG4@s^D-zPuI-Yi0cS?PbM8C(w*Vr?{^E!}MjLN@Fp6pc=z0da3Kmi_lj;N6A@ zq|sqwA6&(7pR3vaWq;RyvlCiO1wNzz9lAsRCJwi{*rn1%8IL%OCH0If>P$mPMSDMC zwsB~b^zp|5nOmaRfw+nrY&aSGEHmA@m^&kInNP?A`(Ezm)Qa5{_q|a2^TLSG`j16? z&8ghYb3gFoxYe7E>+2GuS_7OdF+)Fxs4&jPLm@i6)@S75O1EaWCEL*<@Im~hv6p8vgOp~1s zPKQgIbRWMp-VX7KM^ey2#S)EKJuAf|w}|{TvxCy*(uZ4X99>GuOm1;{6Q8G1(a(VR zYv#L+VEaP0Gc~<$usG!!i?t402cuV>%N?Q`Gn|T(EKb@-srUSo%R=ngXQiE^s-e*| z{ETMr9!>?RVk0tr$pI~dqid@{g&`O0=X=Ea`uCVKzg zThvRn|CShOZ=eC;6K|P)P4d?Dp<1?i8RKc)`9%_Q%yBlM`J9n!d4=69fy3L>CH_p6 zqdT3WihGm3l-wcOU!=D=IkL;8)Pj^uA=zWxyA8cWu*oJ{RXq&&Uu*?rj!xm1R-MsF&< z?%tS7-pWc({?4d(0Z}->I%o#u$+g1kE?6p& z<@Yy{4W}y_n}L~S=r&cK(D_sv3MQH3C(1Tr97gWH`diQ|A6MP=qznp%FR6drpv@ds zFW4_)fS`{d%9Mz1MZ~kTAXT(WwQ6&@#=I1KU+uy2Ua3OExK%(TQSnew`<~Ewp5bOE z|GH}e4q3xCdk=arJ^NBmi=h-pkGe*ho|7cy{8e~D4)P39_m&S3^D3)w(!oO|-J{p> z><>75__`@JIFJGpR9wvtE3a zDdS6=ZmXj@l0BQ(`M0)?%(7#OR%%8VvP5R2laW4@>tJ8zJ0&m}Df^kn*cbF^1>+X{ z8r$_&9pL|J0v1215S8t4z;?LZL50+FYwpko&0E0hS{tI>$ts~@{E`JFyaP{9r(E?JaAQ=prB zX|F*v;&d}zhx|j%FxmaU68T?>k_~#)2u`&!)@wD?m_B!LhVer_Y=4HV%nBpL zXeqg8tAmj6El5R=o~c~HGHUAb_Q9-+=+?j*(>Ek=y0fMs5w8@bX6#9nwIAJ$+sRl4)QIvjT%fo7old>49PQQ9 zK)6@2E$!+Lp61FtS@II4^q@kew>Vff^{~XN;FD@=Kb5s;vnST)Gu_19DnDSbZ8P1d zzt~hwF4>sr#@*&|mC*!Kxkqs>Wd{x6{%{@AIoGYWM^kfmYT37{%BFFlh50S2LfdHR zqo358#nDN`7;(J2c75FS7S(*LP&$ zFR;C~WmZSN7(G_zH{Ip~PAok?H9B@RS5xoKI+9y?De)E@^v)mWgfo$nnX8SXq!~QS zJ6hf?p_74m$u&y5htW!AcFB!J&YDDKvxFI6$Dh^ZH!@-co!*O$hMSk!>*i;j4O5oT zr6EF{ZZ@j+>RLZdI~In#IMd_!E7?IaBSudmmQi#U>C9@7&kwh7`K(a9BJZKkRY=}L zX)q{ZwPZ-H8OxNzRGn-Us=1*P?2nOC%OlZLi}Fidj(OPJs#&5z+vSWkEnLjHX~lK< zbY>E`(0Ltet#Kvc@Jwk5!dPi3ikBXm40^02r;eniWW_d2a@}mJv~lf>Kky(`bIr4- z$!=E)kpCGS*&p+Hj5u)O+!hOPX*lMjv5CT1xb4bOB^pv|4woZ>)SJ|umuj5wzX2%I z?p81LCjK+8Dar2DappC7aoqM+-)#Gdijuif=$>6SB0^HiMDg83nB7JA4IhuYdsmb5TYFBMHYb;SGRuPI zjRPnru4)7evIo}usXFYE?+~PtdTXKjcR*aq`UXwtA*x)Kd##R z;9;p?MBZ4?>f{ko*NZuJ%U;$o8_OzVdNyR=41Tuw2q<^FmDW`ot9^%@Z;{A(rj!PB zrk0U)nexYVh@9}%&%wj-l>gqwdnDg-`F4Kp3#UZ&2jctwtDiHO!SFef1e~F9nCEnV zp&-|5hj;=hjt%=*%AKW$MF58_hMR%^Vfgo}XJUXZ^STWzA;zDdbI2Stza6pH^7hO= zvQmA%4a$w8H}Ead>Pcs^4?pE4#Dq$YwY&=L6Ht7e7N)7cpZ~?LVV8w;>RCQ?_&EHB z;T2L4pg4Y-Bh#(yBxl5Xd;a9ZQFUYum%NOQe<)H}JCZh~9#$kp^f57RI+SVHwjw4N zS87?tkH=#WizMow77UrzTYQ4V;HQBI#7;A(hy98wQx4@|-W{%sIX)CY&%bu(3{a`vc+KkgghS2bo<9i7yL5YCoepv;wYH=U5FR58U-CCpAjawxK zun!UmP8SQNGDUsAx~U9wl$UPBmI+*ae%71S68T4>8A`SuX?voZXs3Vo3tzS*72jlk-d=lv?}lTH2jH;awTw_PK@oP|{lfw{+L zFzzOV zF{H8uDN==2x>oY1uUC+YIt?S#O+YwZ|madpb+lZ^0pK8bGqpV2+!Rx2m#IA2^D zmQCvujvF1NXK_2!{V`1BZ{l~|pR2;^gO|Lj#TybTVtl9w9(OSQA0U#hQEYMLwvx{1 z=7~;Hr|>PAYRUr|&KK-PhJU>B_}c?B>F?aLinFEfH0ry)S33FY5Pv!05Sfy`0rM0z zB^zJyIE^Lw%e7>k_@7$ow`NV8Snzb_*L}?~Tqs?Ea_G{g`eWM&^*&zb!>qq<+2T-1 z+-OsPMgBCXz8@4SEER}s*&@mMEVm-&fWEA=A1^clC61r^d6Dy z+h|>-x}kT%e+G0t+94=JVV9i5soqm=GBTWXh^fV*8fHZv<1=n7$(a zid7|_9k}H?jf>QQ;wWzcKs|%kjG8~v8l5zU;Lq)=SY)#I;keHVB0*mnO0<=N%O8pF z_rb4IWwzdk_PZ*gP@Y`Y&v<8R33b31xzP({qNUgKjRotPJY(w9Z*)lfObbHV*T$YO zgg7`qk#j*THlHRyE=lgj!zdl`Z_BE3&t-9GYre~nW4db%3&1TiRcLl2roJ&cxu|fs z^jql6|1>vsOd+!8yUEfm!K13nda{zGWX5f6Q?`7jJ_QYSWgrLH-ANxg+&{oTPpwTD zD~B&!VprV8oa8cCyi$m;3f@Y1Fad-1T|!6Z|JsbeLUE{@h{Eeh!0iURUBte0e(x|^ zjm+wu*#oY^IGwm7UI#IOA3wz6c5MgF=9iI~nIfiWQnSlqc_ z4Y>cS8{d~`U`UO-f7^Zp6=juTetAyvpv#9KN3$;<+F8?PUFsPR+KW5#&b9CoCD;dt z+tI(x$44N_MrQW4SbWS$Mp*-)%GWNRb2>*}o%!;qc1O2rFp$a!10p=l@^ z6W&r7Iqyc+LE*|Iyvp&Y(Kxt0M@dLuW$py}&G2Zlr3bx#g>7Ej7sM3ZtF^ z^n;!qM+kju4B z1S<`*T}zD*@C?C15jAfYPc9 zX0chK2@wSs^?UUO>b_y&Dg@X7ICT}5@GeE%CMKSyMX@M13#w87iJsPX54sfn;UgS8UK-Al~#JZwYnfv!Oy9VfrW(j%9OZbj( z)8j@R2rb#mIMS7weionevOZ+1$Nif@j$bEoK@#SQ)Dd%PC z#@{}UjF6XON$_cvxZ$mR%c8I~G77zEIS^+mm8Yz0a4|e=Ic`>oGXSxZV9xsJWNxht z2+K67r$Llnk>icHfWbJ~K!v2b{uBRPUs2_$@}l^ewL&ZYb)|y?bF1WG5n1mVvBh)R zC$+<0nijJ3`~K%&`^XtJl22a3ou5Gti?{19N6Lwc^8APx+%F(*F4AG*>2{lnmTq0` zK$Yn(aIuwQH8<1^6{;xb1LAv)2t3aBh?7ymV1Q;M!4P9-QG7)pHS&6i)S!pzTt8ic z@qD+bxGM6wKrWDz)ZrECT8+|h4OH`{zVAeo9_ZPQ>pG;K8dL%p&zByVn|#ox^Z8HE z14QL9u%j0P1FatuQ?q`7kD9eo0y6GFyQNrKhwDlV@IGbM-@=U9+0z}xCspr0wP9pl z8qT~#z4+BaUemz|X1ATOpUQ)Tci5XB+pwoYPWZ^ci!M%IQ?e9Onap^+pQIKC!^k+k zD}FAW`(iLWU`6MPBxe?hdUXUEJ?EWm`Yd;1H1%=*c5BON1{QKf!NmX!;B5%?Lq_(< zC`Do3GQmMRf3^HN;;rDy?A1nKjxK#|&b@94JZ6#m#{t;>d6?i;H(Rq9PBj?F8Lv!d$=lr*_*`Q67R2J3i zreCwIbFM6EQmQ}xPOa37BKOU`MozDYDGvtSxev>WQ4?Dz4F<}P58<~p(C^z$dgUQZ zf)?JkSxY9%Q6W(0#vjTgTh3`7$Q0mf_;@^wk|*m~e0NN7Jkx_1*3#qicvsM+F`qYv z?MKYIB(u%L?#tx$7QjQ;Md#m-U(y=QA+Q4}D-+hgEjB_Kradvn! z2@)VXUVatoG-i;@W56+PP=KLf=zF|@=$^oN+xK$*7VR$D?74vD9|Fp<+$2w1a#ti| ze2{c=SjBb*dTwW6^>3nyZGc}yt;>0xo7gh{gnWN$w)<0V_SzJ)`bqS%XN+HDM(YK3Ah;m7;g{ zn=TnInK&0na1Y2D+y_Yt)2qp|q0}ApS>|%jP&Yye?Q4+cUv}@gx?=K1OTg2^)-$}- zG$yE^jM7BCMpti(3}KG!AJy5-KcbK+EfD=j}mE$`cud)IBXKT$Cz9^28jM})DJ z-CmfZJF&$k=owala+%Wd1e(fk4djVv;{)Fn^%ZX2w~G46Z-+?1y^AGr+8+8JKVco$ z6&d>ksEE`kZf6$$A+_0z=I{*z_O>xF^vEfK%Ll8z`Y<(33!`|^vFICwe%j)Fmw~ce z(z}WMDEZMlPSj@g{9?R*ZI^EYCZtHt(D9({qn6r57eexD`%~m7<`Qu@pr;x8bXNJG zUXrwH>fa7A;Tq~ELN85jj#;XCB%g}N5?;YnxPC0ZqyN=qC4-D)P+m;NXOfDn;qY1H zF9Fm$`lwC1Usy&wgY{p-g=?JpKdn61Zo*xTd?LRt%`Y>)D%pyPwkTyeR#j0P{Pcum z_a}3+4*er&`H$i-*D&mSAXdh~k7iWTsyvc*>LYvyHWxvDWkopLdoZ`pFb@+nOf;n> zB^NdmjC@jB{7Z9nwHMQB4#d$C%2u(FL|1(UJC}VNFMD)revqVBTR&HK@{_KXu~r7$ ziY!WMMo0Oixf3%89~*$&;cR<|^(UWl4EtLy2%J<`u^#kLcUDhu?7EcTr?f?$s&ucQ z1X$vlw)NPlrzvMG;3uBzPV{sASH32O!m3uMO~v`+z83^Qa%8QGE|aX2o&8qaFTNW` zUnba*8!Ko`ei&U|AK~1Coyz#^74FwWk;!$-IzWj_WKuRi(wlmSe(*I)>J9)$lnR3H ze$UtGqeDcNW8WagqMB-eCED2sK4{~)n+R{V6U3?u1m@|P_Zes+&_J;0iQA{8m`H>s zU^!%9sIgrHrpuFP|&cH7pF3muu_Rw^%59@I&V( z+U%i}6|D1g_gTG9P)o+(3XWD-cOTs1Ahe2mn@{-R$?=>{4Tc}p64_Cnkz~4$1@cEZ z*2Enn=mco z2(L*-Za7WaZ~nrPa6TSwSEBUsXW~!OAlF=)2ES^juMkb^ji!CG*olhb!%@SV7G(^& zk0Zz){t#Gb4g)iorNP^<`4z#Lk_n!GN)}l6DOz*N;J(t3S$dj+Vi1h;hQpITN)h6g zNYR%nH1LXtzLMihkB*w9Q&#$q+WcYZK{||GV`?P@a<0Odxbc!Aor-A+4D=(2l|`xI z^v9{}s%z^1A@&iRiH^td1E-wK3zx#e{!NWE_{JJ})IYdl-5>nj!*lepgQoLHy886Zw5jUY7zD^se$Ofm^klaBrEpOwNqIp(>ogO9ACEI#c zPNyZ>xpo%&1^AXs#G_a@o@K5WX;bBP>z@2@p!q?|rUUH)$_y6-bu#1bzEsCXjDF~| zTI$&ytoEG2$XAM8jz<^>sy{oc5!N@tQM#AHQ5a!`Pw(EFynH&vv~q7pkZccaqbDq8 z&>?>J;zVeto!*&rjtR@kDzcQ0S>Ee4@x~F`H}0(HYQl2g5JV;Kqb^V|$tuk?lJv4! zRQ%;-1B)g;$JT9qjYKHK! zA)jK7yk$vLXG?XO8+{Pg@U0bcyKzvsX$O5WbK)Ek=IO~J26`cFIOzJ@uEL^^zqBjpxu-NzoI($!C?P2#}m>Oe!7V}`rVM85=dEjAMIN50c zV_O|O`eD4Hr@fehG9^Z9?WLOv@muPtz1JOf?ub_GgB<(|VMP(#03zWBJ!Ipd72>~< z`AY4a{R>CPaZf0QhgZS8-!`dxRy>!K#vbUn+{R~P456FxD;MkDwX5(zQqC>Cr0c4M zk=ISZBiq9$e$qa7D7U(;3Rl-r!J$}q*E9OhIw80$9NK*>XKotKT`mV3mv6Iq!s4>& z+G;C|g}7^x7;MgaaW2KhGra!5-lR)$x02ksuYWlhqPj79%UvtCjI+lSCsFezn>fmI zsp?=O6DaoU6XG!Mm&(Ujf31vZk+^Ic#kmD=kiWk7v65jTZ<9XTQC5G?W7+YK%e7Tk zw!ZlFvwRAetS_`Cc=$Y)AQFR~%Uoio?;&ycKLHu-x={cVC!d;ttb z29RcbC!d0UlbeObY0Pc>1U*nl8(4LYiNeRPrs1MyR1#PO)Y`t7|3h_O1`wY%F$5m# zJ41_dblAMy68@_5O+N%*?n;}x4h`IaK$f+WS>ZJ2MAkbleg0v?{$Qgki}C7Ie+>WK zs{AgvQxYfLyT`41ENqkGTH36Gpfo3y-}w2oCX{;Z?qczo+2es~FZOc1-qv@h^(E9k zbA7G_a?))A6?-WRu8RuX+esOrz1pN~K#YH#pDw0LoUd4V9iaon&h|{o{c|;JYf06!%3`{{pqfbmTB=fFTITw7 z&ShyG3~6vWi7kTzEi5m-p)6F<_akxxYiTQ8U{O3j9UytFn&pvrOP|L%vRlm`t!Hd6 z2+GycyUr}~Z^yTp_p%mKE%b-#4?ZyO=Vp&-!8s_6qg-7F2><ixuC(*6(Mi*)1QDbpfs(P>!!vGHcwB}* zb<;&Zp7366bfhz0xf+h(*OfWo3vc^It#r`&p3_}M|1M_ z&SUG}##?nF*x08F2o}Rp8sa{sGT=H`V*RriAkE6s^%gpqvW9;Y?Eg+L$RtQrRI`H* z@FLHspU*lt=D=o*Jn9&jSof9eMe+?-prL0Ank|2uwnic1C`!w2@C7HYnFw2E5^> zu;~2|rx{zN10VYhSRN7H_JsNrXlrk6chGFtTSfTMY)dx$=0XaC5`EJ%C;pOB{_D2UmSXt}xj7Qo{;cucEK1K~uJo+XY^&n5?h5{Oaqe@0Czx&`k1orrb zv>_hqPd^qr*V5N|oY5Ckyf_;KEveOh=o2bLZV|=k>ilK*WK=uRr9$$3A)p%9P^6I; z$FG@8&V7KIR$s|)rw|AL@|9zyGO_@ElG?kuLFY5<{u+(Td#)C>@y+Jw zo%X~Q{~{jA1EKSK2xJfhDleHU`U<^EE88AI_##^UIlFGNwcWJ!sK#yB1-*20yCu?* z>&33}#DuxZMft(`=AA!_8vM0vaZ=sS8|hIm0!{2>Yp+dL5b%ItyFHxcEwqK z?W#sWcnV$S0v?T%;EI4P&L;!$LBv%00Ld0v?>pk)qXg0!KXl1ftDOkU#St40h z1Yt!M!#C1I5+QTXF?$z_cU$5H`rQhc) zO``eZRS&GbRa`t`Cz~E%1(4HgI-^8hZ-w=4a*ZB&dRo2jb2RwrgP#_~GRn}2o>!Df zq%cJu%HR}nh_F|&BEk_pR5ah2EB ze5s{22`Qxqv3H=vecw_CdpktJ-V0mF2QLKy9J}B zx(ImF$_fO4bxg^0MA-MNic-TjAf(x*R<`imjixHw;{vT~I>^E!s zgm#rm@XiS=Mo24odJU+TjetbN17`sYHCA|bwMy_}#5&EW&78p5UV#!k#lL$ZuYr=Y zFBgHnpUd5{ZK65tMlvUUOWGDeehw6jUD*&3;6MPDI2dZH#I-4(*TFm6Utc@MGVYv2 zr6vx*#KbW-H7rh&tyWiNaQDUKc({B?JQ{n6FvQF<0xV%cl@`} zqx0+0N%aPh{fRHd<#y@e1h}ZB9;H}7T(s_e@{q4x9hg=2tFtLyKvF<%Uk~bg;RS*< z7O#!BH^Z~VySd7Q{1RV5Y}6=q5kW7%wx*~0PBiHy^~ZkgnXgq88NyjE($*N-EC-#s zzKe9^y-m7J&5^8?<0A{=*xvaZz?mSAxJscaO7*WL&T;Hg^vi+S%|y4&UtRJsquZ-KNQ3SZ2GRs$3yIb};+ICkyx!n4fw z5DzSlZE+_(vd`E4P92to?A-*LR^{v8J=*5Ey;U8fAbO8!%yYF_7d%q;hLM5RK(q1- zJ5zt3!^VmU4Ns{zCl?i)-5+y$r7^M8Pjd7-h(upqRBi@V$!7*^_!)O76<@VgreXH4 z%tL$CP})4K_JCauw2C757g8j-*iDfak``U@bcALCSD!OPCLe@F@KO#-gAF55Fyijm zhEwZj;<=VP%adg}Xax8sO_k(bEUrSLs7e!3OV)%Zc{Iiu9VYu^!EEL$L`rfn=k2Q z{bK1gT+I(q&oK$w2bzX9%}LLCl?()k+4jneEW-65sK-2IJR{u3BIgJ}bqLhcMuOJL z_OF=tzhW)}pq`XkV1I<0u9as=!9O)+oQ`Zlw6gu~7X)N>92xz3uWOM(r>7ry>~8|C zGqN$kKCAU(cukyh1;Jm%M5NV^DWMNEIjqP*q9;#SWp8eyKh%mr1;b9_oIObYX25K8 zCVA>BZqr+IdVd0|U}D|H!)MY|fs%oq#|1EkA^0JP_CCWT+eo;ooWbOSMNY-P8pLcs z;T#Fhn#jiZ`#SsdlN9wT6`?~kIXY5iu{xY>6ip`ey=G0ndPRcmwKxgye~JChaYQ*L zT-g?UZa@GHGJ^aYtqsy zZ>^U>dTy}(Lx3Rj6uK?fkuu~=n&NYoB4kR3cZ+{m%2|H(IvXgS$ChHQ@JA`tKm)*H zsb8Ftc0R;mmhqHSMfG4@@yl;B8DS*>ZzbJ~r67}pCWl~iXHpdfW`IsY~Y z6n)EJlWNt?@Yyp4%>2xlaZ8-yjFe*qCCxWF$k1ZD1oq!~DqAV>1=eUH?(ip&ai@A^ z23SUqc}yuTHAA!-x;5*6x|!pw4>o@R*~_stAOC9ER=9|NCCtV@`-;lGH{xf@uU>-| z>u&wpUOmVa2IUuTky#YmHEcgOU6(UCuO5RA8Okr~7Ww~CAWM47M9Eu~03&@_guothZtAebNIV&SIOOufnpHq>0)7&qI22yMhArg|OC0|Am(f zb;s$agM>dxPiCLDj%(aPm^po-4h3ar}m0>oZ ziKfZ}pw}y%RUz7e)_2Sf{w6?a1}C;d?Bg(IUv-`ZcZ`4=8O!Ecjx_!Er)hPx(@p}G z%^=LO-_O(vA8`7`8Qlkx7(`phzr+VzBKLJh7iBJJ?$1zG5s((-6g~De{vl$hgLiki zIwP0_0m98uZefmKp>|snaxo35I%WTp@jy*4(Vs*8Q8&g{GG14~MI*6|@ver$MZ?J< zg7hAtYc3o`0$`h`gv|9?5bBDq+lg(YsNfBcH075E(ziGh&H90Oj&4@47_EE!2JpC@ zHdo|p&xiBRsOCA%bH2uEi-!@vfB$ZR4XiRKf=z_DgMN(YESZiScm!&3d*HmtYU>S> zc70gHPLK0Ajc;Srr+&gWg?{^hY1TPdir{s$0ZWydt6AWHAG23tOYByX5_iK_+&o*oG1ER)> zw3LsXN|}3JO5;8--8>77`q#&frR&P=m36#JBW1aL)6S`OD_sn0f1%tSXK^57KXue6 zT5qZ1EvEBgCe1sv9s2k$pobj_VWyS*>tRbpkz~+m(7Y!RM*k-$v`_m-A!h+>PEa;L zSp4sC{6IEMkX(w=Hx3OegXN#}_-R-{8MWUS>@AhrU-EX@hz3cuI5D&cW4c0V^tN!1 zv$Z;oPTypFzZ)f2p%=TUB-3Aftbi(Amue}@wawSAL_s=RfNlf#{8>=`Ab}*D32UsXkDlPjYuisr&uS7%r z_T`r4Q!1fYwyw{wPQ5G|z{8}9EBw<^e;WvUa2xN7-@NEO*enMJYTr8Ai{XS(vtsw1!#pR3{HS` zQYKBICGVei3_n5gRqDmM1W z=1Y=nEX^UnsK_zbGvEc?MEydTn?|vPwa%=v;sal=pKGhx^k919U3K0!eG*%+7$-O= z&ar<GKu;^}D{)8b5d(#QDEG31BHZ z-eCW*_e7Q6=uErTxh<3s^%M>5@6)z!D*3P9c|4Y%*~RGdLBP|d&vo7_?{v~ki#M45 zvHYm;OY_Dn?FPrTDeUPodbT40@G&B}1Xge+YDusz=5#FND9CJRke3deVz7U6nhvisRSeMvlcIXRZ#JIz(`XI3Uef+j7Ghyar16nVwZniF^@}}L0EPD z;yKqms+G&p=TFU8z4s>vRDXs8Q!&hq5Z1^B8+0EK5E3&V-)5Y^DP8v5)9z;OHG2%_ zL%W`T=wxs;ydSe!=8pnM7qwwPXtH9pJ@yFgk3brTA6?{oFHTPu<9iD8s!X{JA30s3 zSOMgR zBo6TM_0QSzhGR1`QNnDaDzPC@{9Bh?VYU>{T7}+M75G5%h$8oJsV>e{oh4a=FVFk$ z1Seb8TA&WM@w3lml`GcndfeJ3(yN~#cucbw;u?I*Z^Ft(?9FBgi(jM#dOVp|jgtGe zvZ%KyI*n`%6szLoH&{{HX?32!WUzMLXNu1!&|90DotsE_5l#mc3?FCqa^ph2wLI8k ziCyX6cVP}rc=;n~^>PeFW4jqh!>w2*$KY(#gN44#@$sNMvuF$9; z?kZU2x;OYsboPCl>sp-2(RH4zQ-Nzo8O12VS-o`KcCv5e=$o@DLkqZ+pJ)r9Cc4zD zU&TGMs5bmTEGH#kc>=DB*K-T+3d6AE3R__6S48eD=IdirBIUs`xb;3bwIMF;HzO7H z3Zi=p0?PjG=Qq(bmPg4CQ5`}P#!p*Gq!M3aMVW4j{PFl2?qq`AWJw!c>NA z9&~Ujta$rgP)CDtgcwl8vPq zHiB8d78?&1ulT-W$0HhfR|={aT1}z7bbO~Uz9}?YPGqewm3=W!3YC}TVfaa4ZKmxR z2BJ+2;)&b!5W&^y#&f(;&r!K?bvoQpC23128m=AsN|F}#;C}0NNES*Ihy;+GVhnrX9yDlo^VJj!h#;|4v-_s-3SO#;`++`>#G)B+>;?{g#MkioUe1bwM73? z5_B@dg3uh0!wwp38iaJ$lcxMRY=Q8E`-Lqs`s#!8V4L=i%)=AIxiuA`hgj67iq|;& z9iv57Xl-lO(;Ln3VeR9s9rc>g5(SAWx6m^stTCpbiPl4?AzI~(z7-v5TfmU!(U;^)<5L45W(}J0F%%f~1KDZFix&-gQ#8Ner%7hl%2D z;RD}vUoPpBVpFltGe}%NBcB^b#c0UneL4f?5!2eg?;tQ=R?z{eV1lh|5y+lMNa7~C zd&+n|)`wHGyQR!sAjEvBIf_8|w};~+PMSoFMw|9i;_&Cj@)nr_YaITC9}u#yXZWQZ z9g&h38VoTFH@&zpmNuy&}kk}EOvUu#G%yJn&$%t5Joy&?B$!3cb`{Fs*6d+ z6A9>h0$V=!6wb*g zrMc?_ha$F6dGZQq16(_k&XXZZGMD|j_u1;uY62wOeTq(z0IX~==?LTJ^*YVQQMy^) z1xhZ1btDsQH@`my2uhMn*rkV(z~}@BFm1On2HfsjEFmr!r)t$rqS+1fmr zMC>=fk)>25SFlNPfG%ogws#IqS};VAFD@l1F&|2Ro9)E9u!qu)7Hy)n<4>zOND?d< z1eQ%)`); zx?IjJJ!MZKN@GpnDHL(+XdPzJ^%*`s%JdbJZUAWkkWJ7mcvBo>k`{Up@)RbTS-RW( zd}2?yMtnspap(ZQ9madYkfc}axX*o4qpz&I{1)iOE6#K|-*K$H`=$oTdPDMCA)}Y? z&C&#%R)6yXr0th$&8g=RFq+QgxQumNS431Zk#R*b9(aZR2*@OJ3AMT0Q?Y^T3^0al9uMpN|6h*V~ z2yPq^U)HM~M%117FnzMM2v&q5a)DO;Hs?en_~cJouy|-4+V^O&h*2bmFw3dKjleW9 zLVb=LKVdYqTkp5NJZ)kA@%!P)N&6g95Bgb0<3^v9;P_?s$#bgb{z;&=E@S?MAx;w& z$Oz)#YfI`Drrk=f-M*3fs<3eb3&IV(qA6ZeRqp)hwMM=$!1-<`raI=+Fr`X?pb>d%i{*l4>~G`# z_jArCE|?mL+n7l;8OZTdzzEm@MnEfjgdJuWrBVg~T6`{WZILk*m8ppDw~yGIkLSm# zgMqj!w3*JEsM0B++_A;is#zxoe34kpKrCkYUfo*02{&ig+c{6`2pG{Fkxp?RyX=R#uf_I1tMMU;aj z5aYr{RD2J=^4>;=i{p9|+u;YdN)|l(ymYNhu2M|~-9@c5sBXQ$rS&}wb^{)k$NU{9 zvN`Ry6{G7y@{5ac%*JzLuRDjG}2na`8>TkuN*F7f&h!{!@@W(iaRM35kt)Cf7II}8^TxlY$| zJPZNts4K)lE4LWPveN*G@1bnsOqHHuZTaYslM@ANu4Rz;Jl_NEA>^X9(8`WhWa;o= zc8IK+g_dD_ZClO7^Vt42Kdya8TzaqeG`lwmk3q}PnL;`5w{R(&=*#%i>h)0RQ~j{6q?df71%;Lghg~D3K|R*E(A|3eh{1o4(bIgWy}cl$!K#@K zEYF&@9g`SrXRDwdx}qzFQ_{b@Y?Ti1m*nY^Fu&$DVFP9+B5v>!7lIz&bb zc&}y~jq(@pKvNN@-d!`>0awZzvXSNZMzvFMq{31=j9}j~vat@GpEmD(k=kDNO%xXK z?Zz?L{2o4_Iv%)`ewaDSrEe-%{*-Fp%QW&bI*NLzV@jc~DP-PmW-4Ef($fD+{$sFT z+mb`+lo^z)xx~kk5czB5JFdizI*i^OP8R;}Rtz06zx2UUoZ zcNep^{j;ctR9{v(-?3jkaix%>?*Mo1d}BVEu$D@2fB{>Aps;Ue>d4yfj-(!;UI@Qa zVOmX#%zH&^aj|agGlG#N2=-LriHvdSuxjb>$BJR`KPk-i4SsZK<)&d1E-@x0U2KR# zjN{OQq9V2)j^~nJgoizYC6%<{Ud8VDLXU*hJ(&b2o_ehgP8_QK+*QR@_Wjc*Mx_Bn zd7;_%AX*h$8mzLER0Z-lUiCle#_$%>7yw+J4qYi_Eu08w7D9M%`15RrDdRfM>-2_Y z$snIT_&9Mx^|u1c10<*1Cv@=H15}b{pX_e``yTG_aqeIBJHhM<44}yINmSvN)9m~S{+rzHT(7mzfUmu;Ox zTNAN1onmml(L9DKCi%IDU{k$-&v}t4*wC^#uvjS30dw!f4C8oU4xu?^Dh%uSc(%5! z*e+Phebqpp?WE>iSHp_epz|$j$GFvMs@a?^bzPGU`t^?L-q(DQ$dYbKNypOn`-v@% zrQVui(f3;HeeLNdbITU?Lo07rH5pFwwJpBNx1`B`dbR*A!w4x zs}PMtY8I7(UCp`A_S9uBF8ZQp`(#chr)9!*$omJe)OsaQjz>73lD(i!1vO9)g#QR~ z?zQWC>ui2PHNSv2I2(_GH;63k`oaJAyAPUUK-fZ;nQb2hzwg%V{eKZdI1_WbSEXAi zB2bO^S_3FH!(U&wkm1QYy6p-hm|SALZM53w4lnzJ9Z~y`bHr;OV!Ow=ej}afMO35|7@;c0q=8AK~ zauWH9X!N)X-Kh8DD|Mix0{*mh*N47ga2}8p7&~-j)|p`|bCw;^-N@2voZtvTEy0dz zgisB1IjifgFeMdfw^$Ul7$EvLeA``=cu7d@zCJJC-W>@5$+^Vv0+=9esWiPND$7Gz zqEK*Z+1CAy%nJKdIV5yhb54LqQh2)cf$DjG)EeGbHmK2{7J=-UxUd(5%&Q4idg(yZ zvWEB4VEG&oz^G;rYOKgheKGyT!>qwm6}aJ})Bf{A?m9VED_x1HIRpFN{%#@SP*Z`k z{g!-h4QPSb)Nk9AeU=4d@~MjG(yts}?ft(_WWu=@X?QZwO-7`;E@X34){d-x*G zyM2ty4*-C5VrA*|nyGv+<;?b3fkQ)v4VMbh&li)7ADdW_-yD`rlGh}u<#1?b?^N~f zx?*W)WOG5?Wa-#Dn7P!RH`~|feF+t|JavdK%?@79$eUFFKz~2HG9)^d(ivzd z=VM=Fnmrid`6FYyrq>EM!VVEayb#|)vJ@0??0mjE_!QCl=} z$*l-)D)HyRJ@Uc(#S|5tMdyb-Nf;h4&)>r)iL#b#1qlDSe%R)W4P3tyYAwzVFn0bZ zJ*yn&A)Dw2-vt#J{XRSvv2U2Wi;ln8O<7^-@YZl##u-S`^PD#&81-uBq{k4}Qwe@( z)+ciJI^n+36lx9o4;merJH_;*j;vC*mmBY=@qLcN7Bg>u$Y$WQg{;;>0->VD-v_1m zYa0<2uF$QXEd|knO8vF{E7;)FcG5%1UxYs7Rfvuv9OZdwL`Xh$zg}HMDRCp50kw}n zq5)L&dSgvGobkO4Y+vF8P}skZPon!4Qc{h6+ZS)=KR}0_zH#|}Yj9|DVZYmbrug_k ztKnGFlUlj4mP6aZ@UZzpFt`Z8c~9GI6h{SJom-WidDZxKu3i{^b&MHa?fG*u*qHHs z(6k&D0=v>&hbjyby~R6)RGB6>TF)| zdnljCII-|H&l^XJ-LWKo1b~uX7;r-Wz2(9rHi`X5%s-q2;^Su5Uq16em7B9WY+fv% z`nvtwr)bfk{rBzLZFWff*S34g!(XLSa` zbr|fryCH5CO6G_oD(~(%mp7EVg_$aOKk^9Ky%@z=?kNXN4hkWU92sSOpZsfOYLPUY z!7_r$*4El#_BUvR^`hQG?IWg6N@`5`a0FMK4Vh2pxkK-CX@?w=YsGb1^_BWD{uE(r zIKazaJv@E?T4aLk+Ka6{X}$OnT`5mHD7*Nt6ho9QMEV?;ZzlXfmEXYT zvZPqd@|x0wTTZRGX`)sPDIcu1?-$44R{irgh*5^}#Hf-L)pia5N)u>a>wm3|{sE<< z>23ZZ*{bYlAkepXR8M}br;(9VD|rN}2w!VgxSKp$$7xb6HRk-UWpo4+uZ^ABdl5F} z&v>FiUL=8Q8&+C{-CqyB@bT1r=+MFrq+o8-W|-C|v?n5R4S^=A$0%`Rpz)bS9rXM@ zc6qHQO~K;V7p`9QD$iZMbTQ!|f(w;>jtrM2S07e9wdH58Svdc-f$?0=tXQMiZ2Xc( z>v~2}BzRQ^t)Ur+`S`R*pWU`L^7ySw+{D4`TghORdvhJ26vmh3n^HU)V5{<7ML{|2 zD?nZ>avz_YiO3?jt~Mg_j(M$IaFLeqWP)!WX=|`4**35XlnmbKWeu2Wk-dB-iQPneZ`0<1eE+_Tz5p5630>Cc1((ZE*TCOehts+-IW$a`Y*-YiXy z3>_nlzsq*Kn6+wO3hy{@PO?7WPjgqc5%tb-ZofNR+DW-@I$N2=y-WBI$(oNAVF{Sy zUDP2sxaP&R4G}>MupycEbv%fmH1$(oGexY5f1+MuAF@HxTdv_-royHI7}wsE!&rpc zhZPCsR-rY49GCnYq;-E1v8%czt{f+D$2nP&9wZ9&S3PquT-cY|o7e#19>Hm$UozxD zByXetVlqq}jP-+q$B z6*4Mmg1fr} zcSvwK+}+)Sy9I)~J8$zpsYlgOdn-4yJH0*K-`CSUhclX8WL@qVpsw}XA25Vu0YfNm zEPLMKKj#kzcSWL&tEN$6X_KO<;;!b*73SAC8pj?WbOF{ke0*B*^cMik5y&P zfSt`^aiPVIBqQFux2lGT&*?ZX%XV7C=>dqU;T?I8rl#3X1v{6n7veC|DAoTlee{c! ztvPG$&8WZYNqt83q3+ovJR7t83wd|cP|R7LpjES-`XgtjXsZr*DqRwGFXQNw^gFu$IE%f`B3J7sTU9%psyOZG!&A0^_!+>}n?(==Kpl5v^rq_j==?Z|cbYW1k82YE(HZknwA zSI+!hhqMZ*Mk)I^snBR36wE$OhNcm*e;o~ta1 zKx{4t6i%Gm&g&%MaqJ)}NSP4{{1)q1XE4WtVY(Y-#@NK#+l%4-WSE)kyMkpPA2EC@ z>(Gh8c|S%_7#WVGj@_%Mu-gchg%ad(fObg55~BY?tSnlZdukl+`(WePvkw(_&x*|! zccUUc-X!qae69F!91aZgs3PF;MxN6q6#)5qUpQxF9|{B3@xjp)>sZIuWxDuc{CQ!U z<5j3hgvUIR2*?-G9FBtxpfm8)m@K#4`%5#CR^)2dbfjRbdj!wjHikLlDcMBb620(_ z5GmEH@j7imTnA13aWU)k>krVu9|b>%n2B)8c?1)YAXC_dn~50KFeMbn2$d?&F=Fl% za?ByyprcGxj{Npkn37MycbB?fDb!t_=32uj=s2g!AxnGy#KW3E=bMdH;`REf2RBsA zA#g?e$U_(}cv*Gc7=5P3{>EGqBgXzl@qfpZ3K6wU>HM`XCtJ@|pNCZkQ}Voh|wjf$g&53W0}SZt^+&Re1_%>#JMLGXb8j z0!M-%O^hk_-+xumTQ9IH(1d<|-Il-hG2|v7pK>+*`Z}Z#e5cWwCrCQrE&A_dR{&tm z!hBNBdI-lHPscwN7d~|;N#nB%tN}771B&rWK>3K5K1TvTmNO(!?_G`T1{AZnV-CVU z^?Y|dg+ec&Lv>w;Q$8i=UFLT_ zDsB0~p}-zaI0{8?RH=3bAbg>r)2<#?kdBF+SuDWerliX_cPG>f+%$$|c}}vuTf`Kr zzzE$#j?sUw3WTIMRXm-R%0~wtXOFO;7TzM@HS%qaT=9*)sg%)l>&@>=to}XWsx+Fc zAi%Y8>a>IR^}I7xBSdr4CMP$TN~Vj-8RmAjOl3(`X>h6nw<15-nI)+7;-m@UjHAMw zJ{Ewv3#tN%W_89lUDV4}#)cj*6kWCUStCyY+8(psr?2*BoqZiH&ge0-QG3Di{%wW( zWZiQNvqB*^;aUF0w0XcXTRP6%X=oGmqOL-dheQYcsxAherd3DaDr*p#5wtu(qw-EakQ_)zz< zm>;n`N_P`1)um_f?2Xz86<-ES0?PqtC~n2*1oMa5AFd^*tH}$vqsJ~p$dtH_szwML z@bb{ltl4W%&r56ijhDJsSJ-|0;Ph9hAx-Y4Llr1jXn>#)OqW*hqM@e%6e;(m zmsDyq3{xrHIkFpJ2oi@-xl(#8R}EJjtF!SI300WU0AVfxn>`Byk3ip}q^^D*{74Q~ zhptzd%Ot;yZWBQ`+^&%Q71k1EHUjfq;hgVuu*Q4@CT)nVayxh%ZSWjz?d$I9p2*SZ zAY;Fsm!zKqI~QGsAyAmzkg9RVNju9QnzB5<4^9ME1p%wzZwu6nemZpyi|V&G+3ESYJI3$C_R^$fRhP z@#Rhgj=K&pN&-zvA*F;>^-<_H>b!p_$dJ{qC{7;TdTTsH8^0}jc^f@6tp5FU({Ghy z6{ksm!2tC{qb2a|Gu^KiQ0p-<`tZRUGm}^`mwBIQ2|er1N{<4e?`U?bV;)l%7dvlN z=rk6_L$$qajp(xDWH>T<$^0Pk?Zc7Rt1-KvqB)QS-Ok&5U-oq8^~RYh0oHUU9!NB< z+n&&O5^S^d7_7@bK0Qgj-wRP98rKEg33)V`bQ zom9T{TFx^YrMVWVf8w+~jM~0y@+nS%vl`%17bb$(hVVYyc{&)+#@z)kz@uOO@f4la zf_0!qtB%!2ay3618#h*Vub^?SKnCUlLk)Uz4jpSDi+Bx+zmM5+lnCG#j>3oYThA~6)cZm*vm=up;4ZpkM zO|m}*DkUm$d!E#2-2cpeRh->eaB($-wv$$B!f4ZjtzjLf-#**)&_z@ulSNIKgrj6P zyh#lFC|v(;$2roU>x=DGg0{u6@XB)JszO`1?n}qVljako#?Ts58DSslRXkHOneLbB zvGo}{eqDSaKIP`OpCPPse*AF*V3z|Ok@OpXXjK%HYQ;_sjiKwta|H}1?E?X-X|o_b z$0|fG1vKrM09$Y0uYBACa@+%--DV|cSct}!={|6H>Zdgwq|u@DAp{(~sUj}o&rz|U zm_mynSjkzGeCfcljh_RfecL5n2~ph^#8d`qkdxmA7iZw!V|mAafBy_`c&ce%y(=lo zVUFMQ_&IjG3*+5b`?fHX-I6Q81Y?%y)ELeO1w5dv{O%6NU2o1@M((NYfDwi`yZ+kx z9Gw6TxSRZzvA}gK>TQADL8fW!j`|A$b5y4_uM$4_b+k~-90l*K-8mf!lth_w(E8k9 zsFk(UYyle(W9b~2A`|Lc{4OWBx0RTCL|Sz8l?=PqlBQ@iGHo>XX9(It%H z4xQKX8aJ9aHWnSid(TPo7eQPYd(qXkly`iLx+m~8RM9J2?I>OB^i~A1jF2rr;b}%J z!|xPuB%wfwD$$FO9R~8}@OG2ysU~L{#101iVq3H2wrx}!fr%m=Xld!T|E*q}F78!w z&JpQFy{bFkY_^6YZV;thh7vtDIGocZI>%&0!8yL3@T|ze@9^Czk4cDB%5lW$tT`wX z-R4?S#JHM;d@4t-f%#w9;q_pErrlfrN!x1fpq;;#^21=riwX;cj`#21c$#JtOsPUQ zXE=Rw?hd??9ki?E<-;|LNp8vl*gEFeK#hv=C35r2?)*+RDN=9mxv)P(jtcxk+jYgi zj6%78Buo?JK6(%tosg26+PV1#YLkJdM{|`lp`w4H(EC1*2B!O$S>hO|RBy1Y@!WeG zo{BNuj=hmCe>Xe-S9j(ofeh~i)8|>q^ga$Lpi;fHmL>YkPSrUMMd}j+$@2>laWJjm z9Y29r){zs^t_WxeAU- z%}b;GTK&esupouhQ`QP{{_>i$uJ{OmjvcI;`5^j@fLDhsO~N-KLH{eeQc_vCmg7OmqIvEsk3`T+IyA;S3G$%8e}j<8&#Mc{TSqzx8v}%j z0%MT?_1eF(ah|5adp#G3=&5Yhp#BF#{eo7apC!zUWpIy^l8M$8LmgAIISKDw-WQEH zR}<@TI3_zEbn6&tFRb!rw6VjYa>4b5+kM7$fNtdR(f=lIW z(R#K4Q3We-TSwTQz2hf0TyY$wTx+yAjc2_Jn)L+Wykn&arZFGyOcike263twIjOcV zn+R{pzB1s*0YVKQ7-`?A(T3tUE*R^(Btk^3s?lv!TX-plnmA4$SGnPocM!vVXgJ5IrT@Gs?w0I|(zd?wSkLa?P+VAv*EGO6lMws*~jsoA}@*82B5eZPs z%UKKvvx3A!eSA|#Txg}%lO0q2Qh!IxX>oLfQMb z#4p)$%Y_kpgG;gPBzF}i8UhF%d)3Zm^1+X;&f;+qoSW)$33)Y|;mYb*1Y?77j$`G& za{!;}`WLNJKMj8c?^w?-E>tAUzNwT4TG064=0D2b zc;WygP!p@Emo&WR&0BCAfENzG3GV0K`r#KGdM}AjPtG*~{PvaAOjhXDAQ0gE>K@ch zHoK_lcIkYvDfO{dvM5b1Vd*Q0_j{kVf!)moD%4lKXy&dn@%PWQL>YlauBSohZ@CS` zA;6aR?76c-!Z`KG0jfa%N|x2;(bwS;nrM!ds2x_V8O337Ya&PjQhfqal8!rO1h|XY z)g+?LM-@4U-x>`OPoBx--QXJUB>#mFgA<6smD2a-FQK?c?ZKPV^be93Eh8ptz~I#k zZpH&J!4&zoB1UFZPR^(Bat@m4+92TJSxm>{NbLT(fE4~(TNY@f3L9VcW(hH`oe+~Q zddn~@s5UJ>Fd@kr^vCy+OD*V;O|;WyM2VH0R&~`0F!aVG=^abs;*^qCO0}UI9P6|m zs<*R%Oi+H{yY!(%@%i)$jQ_`v+RSP>yyYnj8(IV0 zYlEs?XGkX{(^bzOo)Wu=mM5ZfZLD_+YH1IW^*Mx5Gz9x`su$v18XJ`;d4xw@_`co{ zl&{_#8I#PSoC>@EhcF$`nq6ib^4}aODDEFe`dnai_4Q@>{*#)W!`7t5DnLBwVeQxv zcC#Wo3Jy@*RuprRED{9hz@AI?0LN}Ezmy$xvH7JL`W73~8WVU)zNgoo@dp{JN1X56 z;VG9Vd}WDKvZcFe;)Tze-Mean>6SMK3Sf6q$Qm9K#6D&}K>vq|_O$DjUMpm`v+RdR zF8)LR>XTKOV5{kY>OTf&NDd3o=X^Z?B^{khn}KhqV9mXlOO!;vs3d&qDSR&0co66#V|A?Ee^@)?C3p1BPxs8KNprw^T0>3tCvqO7Q=QYYUl66!@{Nxu&l1QgkPW0kN~zBJz{ z=QC9pr)28_IQ!{i_#J*KxeorDx`3P?>)vSn=n11g2QO((0-_ z7#e|>d2X*KBO+7Z)wr3d3&CT*n|0sRMD9(o_~yGf-ajTw@JR)cyP&R+lUgAe(pcIX zWGp1{{@2H~Uk~(fdgrDCIkn( z#dr1eKol6Pf7*+!w$x8m+n*oT@>nUW>HuC=DLKtxTe&>4rP9R<%Rl(9a8Oar7#y+E zv{TQ`e&Z}G&JULXx)QQ=Jp#!`uk%Wbb<hWx4lR9PU z3_(J(0W0xh^iL%+f(2$4!}XD#c&!<8#T`8`^^20-sfePGTk z=+T%T?0B5`(1f7tt5S|W=-B@0n-FL_%7?A^a65l}9iRARSBvC_XnRrQm`?tV{Gx_& z;TpWQV=l=6|1T9XHkAD~)v-=b2B@D0uJ65ck%uYCX32%tjG@oD3mi8tf4^VDb~Zal zt?BJ?n;mRA-V7tX3Wa;(*Zw(8L`=00i3xnp~kM zI}CPB8^3L{ik5b9_x@y}OJ;*-!+3k+k_wM@wq#)wzt~oNDZ7@^vSpw7T=r?;J^P)e zanz?>->(4i4GP8W)qU3k+-_7PtJ{f%w0Fzn65bFNNiiBL&tJjmBoSuo z8T_&PUpX98-W?@qfga??;Z^B2n{L@2)ikx@OV%SOFQO%{T_os{7LDf-?V9K929d#<->dW7@H4?75KM9ERgjsd;QF0r8N zX(jnF5+R!h`ITaGMZjq#(5)-OP@q@C=V%)fDF)kxg^)*qK~PkiHSIcjNm(c$|Pt+=aRijvfe-Ug-@xWz&6nVhrz5d-@014!7y32*rtM;e8J_7wz*!# zinXm!*&$)SL7Eu5K9q82f%yGHldjRUz@6&4x02^O2_H?3EiDsuXUE>zp>HUonX9KV zK@iRdUv6}$Uxn<_QVGAz6XUHTF2A{jI-j#~8ugaL+BPNe;4VfURjEm_^i=t%lv4Y* z`ys=BnA4L`NGt$kt6hn@p#VtXM{lS?AAT?%rHnnpL*I@)-d6{j3bNS!zwRxl4yeFlJ9!KWi zkWy>|;Q%*ec+JFoF6hO0@?`7~d)W-gpXA~!kOag*n`!R3?`K&-3hlQ7t^x%vLv{;M$JGAwTMRq^?-F>cr2WASacZ_qcO?_S!ZvA2QiLttJrt8~1P2BV%}i7Wx`d z%UpZ<{1UTwioJ{yI6)LUlHgQEzxTrC=+<4(@FhHCe!`Ay?-mJ*{7|VG*tO=28GnRc z?M9q&k%l2Sv#!!W+@ zkuER{A8yEnFKV)F2M?fI!M2Za{x18$#l?KBklOp5XeK%Fhd9at4GQxrY)#|gbo`sH zwL{mtA8CT%4CN()@#14CC-U^>MgGJvUQ<ZKh5ni!^PEVklLO1P+(amI_=;w@A- zRMw8LNy6|CP>zpHh-0SnSwd^f9A{;Poi?%?UdJ*gPo?W<4`WE)prCe=UGk%_W5uP8 zV{d4x^+gX;RB&r)N73k&^tP5wa)L9+J+)2E%o-?ow6MlAk;tU3H8|K;e|{D7Crsig zVG1*o=ynXxP04@L`}()Q!VH&52DhD>kEBf6RE}e2-x)Zx)BT==7JEc1c%fA)qsCAdJIOm75kcXW za8b@3i#t~?u#GTd@mHhvph|Z~Zf2PJ2~a|djN!#9yAiH(3@f>5!c8nwKrF)fMl2nE zBB}^cD`-wOPhZA~I$dcg4%nrmax4R~)6XhnoN>p+Ip;?;x^@2MVv58&j}5x9Gp=oTT1WYyn{Q)3ja&`G5=9W+`^2-YJN zk!BP41pBKfQMs~9H;5{SN+j87C3_*N0JBI_g;Knld_w~KN?WPT@o>I}3UArM+vizs z$9PG2E8?ILtMUfv`Br9`;FAm#ZdsSo25Hihw&XeC^7uAOnI#0}ddKWHC+e;oIG1Ic zhw&rfb!p`csNX-P18HDvpDLX$(fz?1=F3hpwVfwPOwa033`w2ofA1b9|8RKVf z;M_p|zB~imL_v@9G}V+80`Uiclno36gCUIHm?nA&#K-8GuHueo(jkC=qOd*;cP8HY z9F8qb6jMzWW-O_(7Rq(@I5Q&?5oTqW_JNqs34&Y-6{ESJ4?Jw6KdkRbVJ_f-MP8|1LcxE|LMGn zLDJKfS#z&mA|7q8MVja0MCd|)*Fe%R78+-EGb?z;#v|cJaJI#=X8Lwf>M=*)UJq;Mw z{oEQ=_HAGgGZmj`#n}gWatAl0Ott1?#89{|#DIbpvP%S?Z>xOVJ5o^k45j8P#n`vE zU>K}x^ni;AGDg^F*nd=78foI$KI6R8N*MZ4b52HRB8tstQu7vc(Lw(o*`M`XT=MN% zw4HOwGa+_)E21UMq|lw>n&SSa4*96}jTe@(f}3-tO$JyivTp^oS^hN4Gk-vv{ShDj zreSNL>UAeoY$sdl&`y#5P_z~wrn~VnXvx0OnwqqpyAR^lTbg~3f)f6CTzVJ{-y(#7 zt=oX+!YkvWi@ca)@=76r30+<5C<-_G=9;CzW%SHkM*`iN%Qe(P<$2s92JznvuevRQvplXn?mzz@PZ?5wc;A zCgr3oYfHXX0bDy=bKJMMqx~KbjZ;N&1lJ_v%17gs7+&3!396~Fi~GE*=bu=VsBN|S zP54U`?TJ6N&`LT{pK_=?=F^PgB+CcjXSfZ5ul-nO>V3&{{ErHGOw)ahv z#IjDBu#MDqi6|TladIM%xY+Mg_G%wWr2Pi5IohBF;`U=Q-#pUa3bpt2zyXs#selT6 zKfDxKJui}V8hy{D4!ns*FwkH!N06Y4Ky4qGUNm8&Do(_c<%O@$6Ca25bECna=eqOh zCE69fKrm9k72TA58BRwv9iYhGuedbTcPKY5fm)NeNKy1u%2-<`l277mP9P2<+8X>= zwj*7;)JJHQhpB|WBWrW#%)v-?@ezIv)VTq!*DRmEy>T-_OY~;2)wjy8LDQ}7t$nZU zgscBzDerIi(TwI^wV*vm>+ayjapic25OsSdP#~TDQH;G7eCX?@^7@qb_SgI)+S_Qo zuTi3V6gb8v`cW1~sX6b?A9xvBOYRiDxMUTWf2@Qz(+VujS)Ti-S!TdISHcO7n}V;kaxf??>MZ&<$~GWVozlfVBH5z0O`XYzVHMP{KYu}re2a_5|HG1WSc(l5OT@Bi0BA;hgz3-dLT7|tWqdOPQ%PXQ2Eeuw3-ZS2J69W{Cm_M6pvfz z!XcGcQVl4QWlcx>zxCL{myuwu$fo})c z)K3I}29>xKM-!lK)e$v5u>OP5EB0YVz)~e7no^iAH!XWD1rbav9Qy@d)8w>CsW#BETg1CMDkV4d z{0%`CSiaaXyUx^N24Q*X>J`>kX8%P`iaFLQ86~`+0ytksxBQXBeb;4+e(a55Ec~3* z)R61+Ta^Di)oEf$7w%YoRsu5*hMU<5x5L4i*f#SYL$fpX12cI1Yn1|M34epjV{}@^ z4G+Dkw_mjW{nur6?{muj{4)ao6k3yiS>8q=Py8#LZsi6)*%s)D(PORAo@ybfol2|e zdkxx|cK99IT!`}X6J8z|Ddxg-fyYZLDXYHdVC|z*&acJouJt%ZK{gTcRu?>HgNsWRHm2+Zgi!QgYH^I zwy(QHJgpPs4j8m_9xFW=F;QFgJCq$ROrfEjL8| zmClK~I}z$JY{p1Q3(B^uSg^~=rvqJoXn7nJOnGG^g{P;G8njWE zfI?+3VtF(*hnHkxgy~oD>y`MGld2NCg5mgIM<1tQgHpn(6vK|{N(1cX&%^*p#y3eJ zt?$4#3!J9c`3EGt*-hu0b2b0$4;k-5S)g3042^9@E-#IhQ}TS4U!R*ySTQMrOCx%? zUu=HS`a?1`c7vsm8)}ubb2}DtC?ZhC9=g5l%DeCL^2!rLr3gWE6Z9>Jwa0NGJ(Vpw+={YqUldH=j<0U z$}`mpK;PjS^=9&uIaGZrH&YB93r3s(Faru1iq1U#FQQ6-$MdaB#6d|2cfFRZgN$9I zsW=wDqCV%iq!SpTEg z#W%H&(syw2-FBLFGuh()x60?{fQvDnAzt<45m*}q7xORDs7=3cU`Tg@{BS&4=9#dg z6yW;i{mVuncDi@wd{)pK2iY^}e~Tv%#rC|rzMn(uILDx8T!fUyclTVm$@4f{C;DVY zYt5z_Pz1?GRiU<=k%(8=GuJqJSR@Jj21OD#(cb-DpJ6YksiSa&Uz>()D!I|7P{j5Pbp`RFCxO^XN5^f8K}iSu=6?$w*yHyfX9>9O0D|E#UGEq?)P(b5@UDD>CV|3L@VM z`v*0ar_A7Nb^wMOhQJlX)++y&5{S2Lf=`d%FQeEORzvonv+QO5tsVY{>;tAt@||Pp z3)<1TUOT?*+YM%xWUfp1pEoL)v9I{1`~Q#`c^rCxtj zjlmoJf)O$K=Uys2oKS&x^rhSFF@~{kfH_k+rAL68rjF6^*%_nej7@LuH`PBWt;rk< z;d$T@P-bI#$DMDd{-Nfc4~(@n^N#COSi>Ff?P<8+zh&+f#S|rT)u)=FA#BwkfxlZZp z&wBW@3n0=jMt86BC)VGm2!|~RvqW#eP{OJaTg>R`OCqK!3p2Ld=vbf-SaIHn2?vGgm34{%sL;M+AKP##WsP7NOGH{GlWYw@V706 z4IlyThv_mRSt!6doik?=Yc9ce!wew@J|BTK;Wa_sN5MTtxx$LxNeC`q>;D*X^IJ0< zSJ|kW#vBe}qqv9n2Z{P!bGa7iJ=Kpmy$JLd@(wAKFnVN>>I9vZdF`cVkNOouu600)Zr=20 zT@`;*bj?rF&JrMSeDN2ofn~K%^_>Wa&k}zoiT+HyE`Y-~Pu!6aEnNR>sVY1b0V}GY zyYsGaF)_3HM8gzh;ou-wq)boMPcu6j;GW9TXcy{e@LvID#1>YD`wECux%7gj&+c6m zlw}ElXJFq%PPHAyjYS*^qXROxw2Mq`jIFKe?;qi7kxC7%&AlC^!r9&~3rFWikZZy0 zn7_{{wwH8+CoQj?q%QILR0HJ={fH!T_AwXd1eeoYGp_bW;$Q#@zgb+Iuz7P5_-Zmgn5&1 zk`BwG&c7=Ylw^4d9ObjoDZlGdNI73{n$>a1!pW%Q`UwUtW}9$8{ltJT4pTfyU9W`} zB)fB7R@uZ5!?RfQ_n1)DkLaX5jro^3Yj)Mjk4;FYkyI6rB{G<11T%Uv zSlv*+$!ip56oLW{?J$os`&Lx~%sa75$iAK^&O&T|rJPC!BHwou3j`=9mur+Q?<~F0 zs@U?A3}uG(k|8%n?EImcnTTZ3(+RG$v zt*(g=Z8#lt|3Y46rC&JTk)~5QfZ(- zJHhB5Rm)C%*|V>Fv+OjH&`Q*BXbu>vGiNmu%N2YM`p5cn1C;$Uaufx5^!&Vd#o{J_qK!sj4I*iq8|A--)eJ%i9w!V6oF$D%i) z%wmF{RreW+;NE^y9-g5hCw_9jvs-%!*eEbJXZO|*aYv03+K@BtTS6LC6=IyoX3<}1 zusLh-E?e8!#cKi&Iq6&d#|IHSk3|A(oFw&#fpiPRc3T*{jN*(VQas16c@bqdQ1qX^R%g5Tn=$y^gfF6=KN-PSO;4Do*i0O zX^l;rH5PBDHc`{@rFQ)hMvo=?G9mnp3LXW5gykY_S zhZ{Wc!^3$ge7^8AkiuY3U=gjlQRu;x;wN8>6Ulq+rlc95-OjhqGa)Q$7*g9l)Tje= zXMxLJ?ThhHW6%;TpQ*X{NoQV7-ArXBzpC*@m-Kbp1k1-|JN?b;d6T%@l+U?FY`BPM>ZK{YC`tZ3cxF|4HzKjO3i(+Fkf`{8a>{Z z9M{3te+2B)?vx2IZZ2Hy@G9Ns1_V&aQ{$Q@oH!0v^+8bcjwdiBoS5dQL!=PvC ztz2($`_#byfwBr~00b28W!8E3r`jUY&if}OL_QQBvZ+c4c6SSUM~HE@Z8%dz4i|ZA zyaui}C-f~fT`%)O90ZOhv2R0)>`p@Ho7L2wDp^=FFA`w@itL{Qikl2bX4wcU>+;673BgP8+$e&q4 z9e^?>jq7p}BYJ-#K>oI{T?#qQ^Z_R!-aGl^`uUwL0o}FBQilt0r@wA_Q%dqxN>TwM zIlVXj-r$il|7nV}k|^xZjR=#a@+9`rrq_QU~6Z zpS5$5%t!~-9jg4aL_abvAaJQEUFh|!&sS8K9T}_Ezv@wPo#{LAY{UF^x8bKs&_>aS z5w-QfH}F&P#}T#2IX$1gPod6@a#nqpTxk$A?5aE<(mt%bnV+gCehHprQrPa~QROBz z__i;Ib%(O~FllloPxYIzT^X^g`nM+Z>Zgn&U{E)3~K7*C=tbkAa+ei$wqJO1H z!As;H&qnMrI^3Qk+>Z>Y2x@2kq>F?huqX5$tdV9E;j}jdDiH#IJg(`(MV(CLDcWpg z3tWDwdeobxdcgp|Hl~yrJq)t~Qioylk&vB$PTJs%C~wT^s0!PkVfd$$W=Vh3^X+}5 zlC2q^Xjas_g=Oi4@u(KL9b}?56@pg&N}D@3FhH-bA$C|pcWrK4lWmNGcIybGqd-oQ0F%f$^U33Za!O_3=DKRhRjU6vPVF zDh>3o@Vpd`bBZ_d`r}iH`E}-W&$WyK@bW%j3m>W1o`gr&pcbz>*yi3V1CXCB1K@`0 zzgKn9=D1?(PZv`p!LK65W)Ik;mp#m3{&X^g%D}lO5PpG-k>qhrFfHfIl~w5+R#l7r zbo3Ko<$PJ5YmSbW$^82Pm1YBGvk?)F&d=#aRF11+4*wMT3)g4KxuJY0_}@`AHXEYg zhq_{RUe#_t-T2hS?n>mwiauh5Asr4&;20BGMybIEyXEVsiuyy9qBce@)76N4cE zAra9{F^!IVQ`i*H1cOP@)BbLHn!In7^U0K@kYSuLiH}~7-gd-fFnr)Qywh$z)9o+9 z!>{1ZNvsFhN9$Kd6S_Qk>txhvzm_--FE-dPj_%7%!>D4imG=EY>eJ1jMAh6bup{ zU|4Xxtj=x>s?3uqR(AYOjz ziA@q|HovU<&u{plI(4+nzL)9J9!JDS=Y7l@IgDn-^`(tQKh5RltCKj#Oa+Qtv)>5W zZ^U(VCJeE1loc@2XvNp9J^^!u;(~jL>zX5*C0ptQf3!oQ3i>Mq5vmvHemAWxKdWRv z(;L zRo8%g#7;4&UgXrY4fiWz3>Q&vUkkEx7oEeaQlO@hPH6eg10rfM+6D$lmKGcEKkq9( z;`q^GD+C6OGjmt)CSO@uT_Ru_!^iFOsTB1hRjuN%P@>KA( zDWah*4r0$(m$p1TOLL$`*B1eqGLpXn#OsAYK z);eK0UExj7LvK1y^+D>d>i<4hglFxQHw@7QU4rprW6u1P`UWxLCwUm%m!fVNW8a8+ zfC%*x{SyNE?ZyC;9=o41v95izS}NUX`VPpbg35nidB}hUp5pcUR1i&k(K-?ZmfC*I zw0&4;R!Yk1UTmNT=W!E#dE4uL%PkfddVjN`u-cA8eJ7Tef2I#7)oGDv^D;`lgk_71 z3-{?}^aOa#jMY`MJwBs*iYN1b2<>E8H4E8-w1Hf{s@j)HeSO02#2TMPz$-q4^`w(I zg!QG?f$zl8X7C6QWj69)xv1HfwEOfxgTbqqMQ@-#lqmJVN2)0)6Da+blk<@(eQ6yp ziN53>8;Fwg(P~PH)$C|4%X1X3v&T?NTEiUwup*e;Yp*4VHA}QqA-(8|{ghf)ioO0) z`@31$m)b=_^*nbGG-c1#I7#K#mYC0d9A{3Mou%rOPZdXBXov^F!9nwbNL1Br@3}nW z?>{3xM*5r1*spZPNff*%cy+|$QhdhPkB}JEq~CAJ-;%+t|MXE*Zlk&Q@7i{`_Nnp0 zSHa}K$32fsIqo!WG0s)PGeq>61n+BTldti5=GvW}Siij({)?usfNG=p-o~8*!QEY2 z+_kv7yF&@?uEk2BxVr=k?rw!baSJYQad(RT)8BW#oSbBLvpX|8*|~G?bDw)h$Ou$j zQK@Bu^UprLa#o+}iJm^a2lIVN0cJYffHFUM2sK@69zb=fJNbnJCw|Rm>;#kIu-%s$ z9s+E@h`CND=R23S&U4l~zvNA0(?rYESyhcfUCh^FIBUOKXlVmW=Z z2thKo&*nnOzs68?-88*ANW*CJ83U;x>X~;%694F~_#nwt0cFvP-f623)6p2U|6$qp z5f{F1J!+_K(RZStb$dDL%akAOy>zPO&K6pRKgdXV!BX83YTNd(kMSu|iGO7Cc**W)DQr_b`_@QXwR-B zWy!?8q}LC&{Ngj7M5y2_^3cqow<5s@rBe=_k0sL&UIq---pEzCYo28h;FEm;0K9@& zdEwCBReg`dSIxT)Bii4|M5GnAt&ZOxs6$K^?(g*OJ3y0uP0^N{|11|v5u%roNPzan zhm*I@>SY?2qJ!)VXI0%a_nn|_gf&^sJA2)IHzF2wkzb8!sZag?Y1|p4uf~D4F&pY+ z+&0tp=@^E+Q>!f=q=8jAQ4R>cf~c362HOMWuLhaVfPc>OJMpipzM`4DVNmxq5U(>E zuM;F%iR9Szhe#5!Y6Y=1mA(l*u)sAQn)Bwdn?!o#%&YiM<)O;rc{(WWD>=Ye=)E(z znniM-u7iv&*bBNoxxIQbGkngkF{xCm??3M0IMz({YNTPv{Q3XQwTH-mZS6ewP7mL&w{_ z_9q$SX-?|saWZcS+FcP3%OE8!cK^$^s8MDjYd1D+8b!D*`NolA*CA`pw=u6~ZcF9E z-xJ+X12J<&IggBYx~z|>N;S*`2iofVfG!rv5A+RKKUUGt7}cr)rywQor)HLcP1&rF z`E_ga3SwfS+Im8nT9KysNvz6)Hv>nkgw=bsGNbKxy!;=^zna~&$n6bjHhPG_ib=!@ zHDrCA!8?7CXd;9O3{Vhdfq@lHSEdAYj8(xgW%iv!F$?gjtyl5J%(vYf*%}ZxGO^_8 zkBGnxwUhvJ9%-1=op|SpqQBDL|R&9YiSjh9|jP@ zkUW$B*$+W?CIP~9(rfU*b+E>6QOqQM%tLvWwjz~4qHC!7Tf|j+CcQmnjEwdLMjqoN#{6uu>H)9Q7dqIkj;GUwuU-^Yl%jWVSD#X%PXDY^vw#aqh zVPpy#x-P`YE-#{^cXPvp5$m_pvLF4s1+nBDEL#u+Ws=DizZp6j{Ejp{9Z*+P-iR>n zSr;8tTF!c2eK#uijO!$$%d4%cc9)G^M&DcKc$>LrWgIQR;S5Rn3v*XoN{Ho7HB)5JF)_#9Jz)I1p&jT>-3ONp_HW)1b?Ic=e0Sy5UWBu{{?lBag&s}Q=X z;MD5fZe4ZjU+tb_Um`6j_#e5i#i;=jQwA=YE+ ziL2RnTNk(bvu0c~11fZ>8v+>VbM2-$K)0Y5nhmmDzYM~zp&xiMqU*jOuB*rk9Ex|I zu$rUVA55qemV-Xblu-J7yxaK&o^4Re3d+jYtk&MxqYjz|CdU9Zv6~!o(HcbwvWC6? zhN=)q-f146lqoYFNDWuc(@;$1cfMKmC*)$g2GX=IjUJs<;g7uv#hi+eIvS;!kA{#~ zV{#Yb#%9}^ldi&-S|IV1h$*ecgu)a!Sb=1eHwK#7V;1#>#|aiD2bEd)qDV#6Hw>#{ zppyZBBC<#}OKHpNjwgx`>QTZ2KCI`TmqvlbCZQr~{Ww#z$rg&H@AQ7nFy!xzH7i32B{aGaC!nrh<* zUBvvj)w>CMox%HwSNZ^KAXZlw;2QR*atIFSZvlDlzt!wSUh9-;iRcpE;4MQvy^ucfo&n&sqB%G}MU-lAV3+&} zpyU@Hr6XrVCb8737!1>d+TdN`r_X#Fav+O=e!+uWkExCTt@C$iaM6qb@Wk{zmy|DH z;Fq9qUxBfVjEdxsa2gpjv^YFGW~R-T2*R;V#sG}3vNAFWff2V*Ludgo7VVpYXk&*)=Zwr$;`#uO!8?DGtJAz&R8eo&rRvJ(@Y) zKthbOv6;qZfx0i?L!4z1BWR=^+}e+CdjeO>gY5AuokHcfLf_-j^0hs67q(4JHq>vy zOmDa$bkIb3i;Iu~EM6|^)kn)U0X%Wmr^>WyuqsGuRC-cz2gu8UQ&k}=UCHAtSNE#| zb9Vg0ob!Epgh=|In4XTh#HVmk=CS@wd0Exr#8Jz|eQ^!Ph(ATnn)Ypu7P}0n+8c3) zDlP1jQZMPQ)iZ-#g41-tjQ;(#=bq<}>{aTC@0OPk=O3O(e{t!B{OIYfozc8CXJFf& zgtxF)u=O+TTCJh^xSZp0N1Qx7g23PyE5VR4_R|37jcEv*0sZ4oQf}d^!8&Xz`&?bkp zgjPJf7(Au!$}fkGmMPupThT8}4rU+ZoLUGS*|SHx(CYlaNYUFHPV)tAAC_LuGF&SZQJR>}sujfJ$cbdVsgk)HGgs&jfplb1~3 z=l1l9BF55s8n>2*vL6K~jPH)>O^h9udbjvE&oVfX%!PdA&6_F%-(T;NPI23WFy3|b zivTX;ZCCb30Jhah3N4ztVrJ1mC5x*%#9#&OwDy`qa*eyhBtFDA=_~Pi2C?Gz=iBX5 zO(tnI?Gc7fs?6Tcfd-EhK?mQ4jf$hn*DCs&hc+M-lb@-I9}xM4>2hA#sh|SS;%u_6 z@Wczz+jxT}Bg#?1qTnt6k#t}f1X{e%Ho+k$BztN3xD}e4{3qtsu(Dbf>AsK(eWkQs zEu>)lwiWb3|77fXx!h_H(Vg@s#wf_*EgQq>-tdw&1CcH&=xIXaM*Akgr>D(Q?Pbz@ zqnzq7zxS7=eo}d8CGhAb82-v`uvW4B#8GdyWW3zn+X{H6ZxTj!Aur)ddql@CMU>9b z$}D=U?1bY+@I1tC+vrI_l=++H>uNBkWY`|IYumnYjxiPz0X zVaj)TwL+gBn+DQiXdRU3j@q%^lae&&vSY(tn>r1w|BlwZ_9Rs`wAGBoL5mT#q>T4p z4tRWs7*7ZJh(hi{h@=yEX-J{DsNWTm`f%>RKBomphuNT>*anh9+;_H_K@U-qRNY%= zwEH5GrLIavpfJtx#;H}qTJfgJM+dB5jUb<<+eobjM&()DCCetYS=@ZO7+LYk^n2&H zpWPL**UrZYs}{Exn*EiS4PDTSH12X5jw?V9867qiTg|{j87tk?FVF_Y?dKR3gfp%6e@)FZ+swLMGa^j zDf}B(eU>QkNn6ppU`_f{w;mjV+=F7MxlU2u5#(t9lAP&n6dgC?8?vel2N}1$hpXtkd zsuY`32;Hs$Sg#<>Gpk49>x6d#W`)Cn!ak6F1b02aS|hNjEXUhEKRyUIAHJW66rP(|{Uj24RjwIcXcVzhr~O zy3o+3oXXTO-mq3RMYa=}0HJH%AMB{4HZ(Paq)B7WihtM4a19n3%Ni6STWm^HBNz8f z(N%Qw*aS%9iwc?X;<_7@e;DZ4px69{d}a8Hr=Q&_ir3j!+Szu|p&Xjux45&)e%#WBsdP@X?#~w(#Ei{tL|o|GjYD(&lV?Actd4 zYeuXDa{sd8kqMT!F}=eajv)O^*Je%J&^>?8{keFNIA##gH=!-!6Pci^&>AW z$MHd7k5!?JAsYYf4cs`CwGPw(;|^b-=v~VhzPrS%(!E5qyA33+o^r^r1fXoaZ={+X zKoGiSJ!QoE#0EvH3^u+T##dnhdf3{t5cJOa-JpSQz9IeA=5=U|KmW45u<$TG#5a-SZVhRr1yxr=!$&g<0d zD-8u9JD!zPF``DmCY$%E^9-L-n36?-W2SMk19V0{=a0D4bT(;WUP!n3b^m$5uDr9t zgH`nD{!|D1*t;De(6-7W*vts}1isd~sCP4$Y)9hF;?G~l(#YtN@&DN6g|~;0b`wGKOLY7PNQhfS zYwqws3I%TvV?4Y|JZb(qpE*M#dGn$EZ`$O>T@GSxh1^di@ObecA6lpx{@Zs(ey7VV zNFIDLcB{91;&?ZC*9ljF02CDP{P-7jWf~3X){-3$T#NRmd993pm?mABCIQW$kGZxE zc@=cb-Sa)(>Ah2@oJrs)Fai*$SHmrs|B|KL#xPKuL?!Wam75mwBA0av1%x) zf`^O7)0kT3L;$%}th*=Uh`CTW!@u!Se>UP+A>8gF8tp6|t+(tDs4?(>Z$gMqrd;Z3 z>?ffEC{6)rx$Z(2C(ZYdlQOytAJ(VtWirkZORC0_L1eW1T-e6_bXu9~|BOEi@9gy; zgyYx;%Wv~sX%COo#Vn$AAupI~ThGtB(T*K(C9jZ(;7f74X#QEIcX0XDn9J0(~Q5OSUjiuure*MRGOwfhJiND8R zm@Y;G4}T6RiJ!ZPgUi>-Ezhd2xCRh_&LVaYHKdh0vI^FZJ{BliU)7+6K9EW7ZU9Qc zQ?11Mbmcp0P{P@*L2x0B-ZHB5{Db2#>L=qbS>?H`hORH$ScXv^Ccij_W__R|X#YCZ zWYXr-*C4~Uh#Fa?z6$Qsrbp4*En6904}b|anKPe@SsiQAL^7H8O;nJ$53Ck3I z4%eaSU1yfS9jn*+#+d%_ZhCCAGR8Lp=(rQdhdo8Gcob{X_5vyZ;{Gj~WPc@7g$A*9 z`%;C5Djz(<2&?Xjpa~*cxi^P7UUWkh`O|-3 zU_?{w$7*$VL}&sE&^+bllyMHrq>pSd$BY!d4u0#wwgt}Mz6)wA+X4<-r6b}Beds3f z!dmGSdm}|O3)>&|ayACBl?=q)s)KSx7$Uw7q{LeMaz)APqtEg7PgA~yj@$uL{M%Tg zyeKn(?qQ$x>6_BnH|A7ddSm#9%;HfQUL0I<+!q7x6iAj3+*FKiLcdr%YUA9G(Dru{ zV?`nAe7GsBUr2iVyK+%_Q%`w8AIRvMf@+-UemSB%`h4c=&iNX|dWhTB3Dd{=yUPL6 zfIN`wGhfe-kj3y~VzH?*1V(SPzD$ht3>n+j6yJ;&zK7ukT=yX(D|Qs!OF8u9 zgikIAb+C{95!DE2r@sFI+CzHh8rn8PWmSfz2#HZBH=k`~thoH8c~CxkIi!Bu^iN&K zdP}6@5{J)#bL;`}{j1h-FI!m*?{xv`vk8#pN!GKDb1oxL(le3L>IPs1_RK$un_CC4 zD{u2dAtpy;^CPcyJlQw9?dM{a+2Z0hnFp5Hq(7z4!?nlugLtxOw#Lod#|p6=@#(s9 zImu^ndzY%%oxglk~??dG2VX3y>g+ zhN$hwh}Xl?spscY`X=R>*n4km`5Rw2R?GBrag@Ix)2zenu3+F7ztM+G`Cu5 z0RE9Kt4Ss5a@zh6`+-GdXvPC}-knI}4h9xA7nUb=&yQ+qimIK-6)B3`tcI7jHPZaj zO8Lnn9%am)ttynqUZ*J%Ob_NlGjPnF{ z?+_U*bZu~XPz3TJwBV=Vfq!O@tRMA)eK;36wTE6aii;rpg9`E&VE{~ra<3XWH6#`@ zgK;VwZeBo-;(h4K)a!F#p(bsQbnhN+UMFq;!CO%3z~VyIp5ophvX@L8z-2~w2gZ69 zkm9a)ftGb$V&2~+cKQotvOixD`!ow-fCE}Cw73gd5Z3c2=vj%-yGWse>=Ye}p87&t z5P*~YwFi1zBM1YgD%hv)ky8Vk2{VNCUvQp|Dk+f(DP#PXqc^_I@SHk_o2$Lh5nY0@ zgb9iqAoHlcnbshWblpT$8jHK)ceqS?leG_-Nc@c^B`9OT6qyYltnjc!};I22MOyR zXco5c))HwiSfQm%I>qM2h`uO6T2M>^Y3_qhjm>)Sg$cSQ1tFw3*+DG#%1-$oEiU8^ zche!-S5LZHXrHU*G&C;u^W#KI{BO@$A9lMHAwE+{-PC`s0$%L@9Nq-no$n?& z{>_BiwZfJGRfZS6;Xy=2evlo=e|DoI<*n_RhSZS($1jYFFqF7cKuQwpBWDryr&!u@3lU((#e4M)t+>iW=;Qty%_SIrHu%AwL!@F@$Y-ptoU? zm>?itWlM~E=j<(YYNhd9lLNu6^ds{AZ4d1P;dDa`Z!W2W{5bO1Np26>({hlHS^xL) zeG59S3%>{a(>#dfO^#*a&(_Q=yl3f!u8-zw<*nxb??)#R?=UO`QIkFvAVE}Zm>qgzD?fjDK#`X!d&!5?uA>G{qSVto2G|M;0Qg)pNjH^`j~YS zD8c(aj2N^GXlYyoqX|>2SyFg&N_mi4UR1(%KVL^dv_qbT)zFNEK2#NMPWAxe88=_? z?%Of^eJu171Od^i2hA64BtoCL4mW;tE-rB5udANvN+!y#U1j^Rtu7tNYZ2+vQpUGV z=cS(gH+xT(k^aLzAUhc9dO5GL%WUhQ^Y{57=*_d5ZwnA122OJ+CCIGh+0oSe_*uL) zZ*?6YvYIan5)#A*7Hr`M2(2y=vS}4&6+!W@a?0C0VKW!ct-D(L9sa)OcPR#B8FYoO z^qnk*^5KUHig3bL@6i~F{D}{h-g+JDgb7_jFnsi&^$y$jFD7Bzi*^jm8<8s&N_A=f zfk7QSz+$^D&S3~RcE4*lY@ulvpvI%TYvOSZC zsu`$yF}4#f$Rssf=&vvc-QQJx6+KQmHx3X*d1nlqdVO)PLE8Do6r zkssBGgHm$zZEvgFdyy}a(W0OSRPQOM3gp)imy?YGQ>=W(8e; zs*2_B$j?ZJ1JIr_)IjTGb4X{NiIc8;lWWuvnL++FCB9OTY-& zGSp}-%`TDx)66*$S*}IBUZwb|#7P!T$2=R*i_3P_baGfg4zIa`ar(fkhV{tdOIQWI zyVAsHpC}HVB1%JOUJM0YD0}C4qaa?DDKHej!p!8wkAZvJQLZeA#V^oFtr zo8Jzudt~aZF)(9mkTGY?=GK{{_6yfdtWghIA9Ky67_o)&Z|XIWv#>lqC1LHZsbvyS zbW{~Ue}>R4>O`r7F@ih@`O&V^d~TgIA=HXrKcuU@o0SsaBtSnRiAl>6>Ps(zk87>` znO@f8>FS*Jr36wFsh%(a+ay3Z?#kwOe00X17c?oK-qq@BH^Q2LA{+8KG;J7Gpj zB$*ElCS?r1N{GfH;Y(&6Ae9xm#CH~UCEI4fk!zFlvhMHe!i;GsWV@iRm8DE#ue(y* zB6w;#8kkDh(JS(zak0j#giHc%V65}Ux?wHDB|ICzp@(n=Wmja8_rFD9t&011mB|l2tdhiYZ^WeP3KWuIyBw@!k7~=eYnTFPD{MR!KLzRj`nH4Z{p>JtKU}v);V3 zh`zI*N*-uBneGzQ&p$NedL+zZd6Cq~^YN&EbgsJQ{L1QHpm+nuI{8!G(9cW^XAuWud?KW+h zO_R_vqI8V>`|+1iJtYFt&pXKLZ6W<&VAX$p1nj>n>HLc5R-6!qQ|p;s)0(z_=|Y+mzi*h?+{v}&^X z=4H6{J1FI6J=OgGoCwj)#G}saBe!S|skuB+`;Nk;aRTfr-moJ!FfgVb9gM-c=hDV* zt)4QUVS?x-!UR-cz%(QEG@_uh9Bfba^VZC`U-+x|l|HR&X?`HTlHM*Bjz9I_-+a>O z)40wPb)&_blMM>T29dTaP_M3}yUNG<5wMYKK}wZ>#;BSFveme4f|ty zyBIT-jcWt~i{I&&TX**TF{(2qI_P@NIn;Tj52#1j7~9!M@tHrTp4qxLbMpJ%<=|Tn zi!H8KNl+BwYX6Wb<#WxCLQWkY2lt>7o;2?nh!t}~7Ks`R)SE=o!FMs5SI4?*muj<@ zUCWl@g~-K>N-|qIn9R0?-BQ_YrYB8_3eMd)sn>#&9DAy7iuOrBiRtG4;T|8RYiVrV zLl7g$3E7Z5N{7mV&b>msRJR{CXwu@F0ib$rXDQzQHUuI0$E*27qsc7I=<94&*+xpK zgj+XeZstSdY#CwXX{U0Ni7H)wJeqo;?e{6N?W?&q_V4|5@*V8RcJFJ6+^@en;mSus z$e!{H_&7h27tte?yR>p6SBR~-yRLHA;eR6GS>#4hajG6Hhh323d`bR_aHcR1xk7X# zM{|mX^V#Hir|(dTE|2F6(0rQ2G^ASHCCr+7H_!MUHdyu_d>xIQ*RR7^0#i5Jx+TT+ z*rkY#IR*lw#eQ)@+htndx-YfkuYbleTkx2x0`Y!Ib>Wxdh$gPXgkn5AIb<)%w!PqV zW^B^r?AXE%TJ&p8MLry!@Cud&YiCl6OD zhILjnHtG;z^YXR_!f$BkSjS1Y&^4|Pm9S%@@!A7YudNyOfEXri|5bfxqpZ~OEXPJj zG(ufEYc@6FqJgZl!iMwWC$SYHB48Pz?pEQ!w!}nb!oc#TPyUej#^-5uj&i7%u5eU?sxCaK z?Co~E-=2!U4rp0@mFYrT<9b$H=%R44?QQyGxBQUg9lL5g$HoFl4Wz}*ZK7O+Aq9}R zwNlUPE_or|>o-!CwRoAp?oNeXX2(5CctX{uzsQj+As82_{3|>U!{sOAzqnrKf|-<; z-75ARVuxz8p$eb?ZCG%&@&Z1J7$fQ!?NJi{mC8;h?B`TJBMu-dK5px_zEl)7<`?f7 z!^SKjXOJlS74?^btWFx$9;S|OBmMHb*%_0!62_NEA8`u5wMjBFMF7>&#(6(Z;Kjsm z)(zSu#W|0Y$d<_RE@r-4!kkq8cfj?@6o}H-P(Fk!+8?dy^sWqBG1|Ph-E5lANi>-e z2PYNrvuhJWolmo^3EmE-ZEyQD&v{}%&B1q19F_4HEz|A&zf(pFK5}I`V8BuN-w^=~ zz5d^r>_Bsdm~^7YqHk>`$aWMC69ZZ5tL~SzqNfquI4ySrS)tJ9mLu9v4aO=j(;p6n zaN76G;UifI)s!aM8S^IwYgOyHYA1k`EC3~sziP}H1A|;P3<^jc@`>9Wr7uXr%d(wUAoKd7hrVApo^C(sVxRP!-SKt3Y6DPPZR>_UPV+D- z>F8qc=ZhIjcA15-F^zdPar1}qYj!_;9L&h?76{F@jIeN>UK1YXPoCq3qSRl__+vH% z-(h*jk_nV=Yt?{VrVmHcl->9@Kujc%&w;st-}CvGvo;|^5BpjrZQnmK$qz*P8+qLR zlVz@TU+Gdf6%Miv;CsY9bQH1=kfWmgN+bOFG2I74fFVq6IoCd zdW4a4nuQLKb5nH{=lt^W^#bd41;`N2x*BcGKH_%Wj}Dev^N+)2Na+Eq3d=`>nVT0I z69PdlEIwz8Eh7t-CSg(&b3SvEzu^)AG@}(e?H~N2m7ntdR51(m$z|OfT6vD5RT`8GDvDEXKYUMD{Ypk}@--GR*e6d^t<0 z4lFt-XN`V7`Aft7uto%J@UcwAKBD@!o68UC?=z?8)u5B&DAA;6_FtJ#FYWdNJoHs> z)*z${{}M~J*wvV@zHAFgYD8sB=6?tx7UIgvt^zf}!q+s09F5TLiq?rlx0*%RFG9QR z2bw;CyzBvgHcTAvZEQ@N>+=m$5992qQKPEgw3(7bX>f&$M16FlBs1Y0yBc#)IVDnUE}1{rr>pP@H@nBK6sCzDzznM2*w6<^&jLAe zWk$N<>+LD)qp{9HDHflUDv$gYFKsj~IGSILlk5#{wR~nNP_RvK0>`$|y<}d}FV##^ zx`FstKzdd(L8&wL>aNw%gpp0W7K%W*7g3F&?PczoGuVB-)SrAhag)EX2$eCd2n@3& ze3wb9a^9u=>fVw}eHem2JM%qyn&2GZzakYtT`i-5>H=h7&$hl=*24c3tVgdiQS1ZN zt@-ssRefY6jQkxQT$`+#XW>@UQ66G0P}-hVwY*xudXv5pY)uHa5_J+bG=*n$jp#M_ zOn735L4DM{1TI|~+Ei0~A5Vqa=iRH5{yaEFx1dU1-`n=O8Ls6uJ zb$7{32lr5&wQP;jwVHJcutWD0Uddn8eZaMnyOo-Yy7O>cbK6j}5=PYCH!4wXX`l8R zFoWz_xb-pk{HtRE?MfFj67QXNf9>W=x}k+HZ=t8vBk1QQNSNa%VRYxmZ!D;CKM5Jy zcpW?xJAHce{Dt>gH|ZOy9ofSFHV=oG-kWc3jy7#1PYo zv^32t9HX5^Ew~dCt?V~K%+ajhHN6`VMX+s=2ruUni+~_M!4fl{8nT9h-(HJsvn7pE zYdAfoi;La_q@@`P*uesOX(|dmH=Q3rs{2_x}>ziyHxAp7!3tVA_04) zg{IRhDpKVp=j|cff=Un?=@rPtdOG3$!{zJr6C|DzLflB-FT~?H5s%+-E;Ue1qZnk@ z5w+VjDk{&p5R8yWggDu?5iuS!1#p~EuFUX|l_U?l;lw4uEeWEDg}Pl#t<+2W%blyqFwvRYfnsK2mJRh`!4adqxxm?TZvI^|>s+ zlWrP4^_r6qCT8?sVP>>nsZtuVb;Pk5cixA2aVN8ZY;DiN+ufO)=@ESkgW}wn!@FWF zY$z9F7b}f{R*tkjBI`WJgJA{jwzg2J+Zms1bz(cNdlD{Iw#^fi-6hWc*J4@O-8%p_ zhK);O&HuSXOMA%4-`Fh93MMJ zSok1FLd2}5Rq)kh#uhxs6H8*Y$YlIUe((w*DgmEvMR(t6jp{pcnh`lTXN%=*+|fwj zbkKoB)3g#xX<`1_HUY^q$j5pKGC z)Q0%WReRvEY#nqKm)_f0HFPsU)-`AvX_nFFu@||Et+1Vbailk*qqyl4(8HLLFismLV?BdR2 z437sFQML-*f{{YY{}M-PVP}e#pP(IrkM}&fEUcet`YNjz^IVS6V9nG|=p;U|wh5zR z-=onYG zHyv%fvt{pwPqkCh2^qLJ-Foe9S|(5kFWM}^7UCR4Y@ z`tw0IdyV&z^;EAD#hnV$x9QvCymd>;i})y8a)JQ5Y9FW!)lYBeNtz#9CHCkX!QGo- z6FA5zAdoNu1f367SuFeSkKc7lNr>ekoc;Fwb>FyT?DhwSP0@9jH)DE#+p<0&`Bh*{ z*DP?}7qy$=>LAPaZjs+BYWvL;72A5m9=wq*X;ozM&C9@_iXqecb2+c8nK7|17M<5kwm>6XzK{81BK&!^0$}`9i0lw_I0I!s4vh#g06E{42OA z=T5fI1dby?eUeRjITf>QJl2C=^i;R47WLWd{cW6=3UOyU}Wze8`MnIlZP@!zuO~+93A}fP(i(r20GRYfA>NhJaL^w zs-uECKItLZm~ArJgvz!cs9M#V%EOI*D_cD61J8$pVXGVbfb|Q9z?Rb`6QkGH=b6s0gF-jP zCS14qdA29lBqXGWFxeafRCqjH2yFp~3ceW)m$Zp=-TRq3sMr@=A9PWf>bO2AO8UJ! zsIT6qEObLm4|~w(c0Ln9adWl9tLhd3k!eG8`k!;zZ2tU+@vXqMaLTm1hsVfHsD5w1 zb`^ceSW@O-Lk{nVY?VyGwpv!OntJW78P9+ZRvg~Uk&3!}^p_X@*D3Dyy*lh72ey5q zbM()OD=I;~9?8Zrih6)X{|I((aGZQiEa)ud?<`$$c2-%8<(n zZ_EGH;-`+Lh)j7)oXS9y8h+0sfK;VHw84-OHtV8U5;l`EShtte$(qMfo zZsroI?Q8Ov(PADodk#5`BNWGpoqLidaz16Q2%HJ|GB{qO$MzJ-Re0+}L-_>xgohgI zq8ds_A>u>Zevx|m*xSS3-KT6B3l9L5SSLpMg&(21e=g+$Vow>3O%~i0^-Xy^zg<~a zz1OV2JFPDrtD1BQ8VYiwTv|?5sc><`c^EV)o8}Di++;4YKaEEAbOGuLj!Los3~H5{ zH`uzGn2LqJ)>gJT=+8I%-fY+|#w;6hU2Y$3i)m?F=oCUYj`Y_3L@XhGpDs5 z(V? zQhkQ6>5>S&u3~}FUR%bEUJuq+j{bpFFBpVSelYUtqX_W&_9uvp9#^%agqzs(pg1| zajb64N$HFqe}*kW60{abj`UNv%NWU9Ie{{LK!}d`@>`@MlTpnor(pF3 zWJ>*rgUO*QN&M{}puG(Rv7C9B5X%_L&=86kN34ccmrt1Nv2F%(S3d`-kGu92#hTZj zOJrd{(RrswlGJeUFZWG|`AVdKQwr!z6V^&2$q^~iVXOSKe7Z#sRbwV!ZZ$<$UO%(7 z%C)5PlbcN1{Q|y=Yb$q|-Mk8vpHW5#DGrjplMWz8QZpf0;@jd#v0b|GrZJt_tM z7$kB@$28-UqlDF1opMBgC495Sk5rx2km;|YpCE`iPp*0EL<3GJG~tgfQdXcncZSZV z7JPlCIm>TvH5*0*9@wZ0}Z zsmBXB?ih&A&$@{RJ4a%k4Rf$pCuc3&;Hi}_WCJH5JOIu#MpiQ5vjJYWj$&&fwEvg& zv&&ZG`0)7kAh`l2dN4x(I{c*SWPOgwVX${hn78wf;v36lMnzeIQl8(H@Pm#w_@`2 z(Ea&)sJ)FApnNDF)VQ=-S6#uTt=TR?NE^V<2M0?wwbsYOWhA*43Iw%TdVAF8f=x5a zR&XZ3suxp3(8K8UP2)>lH(i!yKTm}sJZuE;Zc394t7AqCF80k=FwHW{T*j~iT+z8d zJR4s9^WmrLOtGmX@3dc?z;&^h{hA+SOCT_Uj*SpDc)5}#8U68t4H338&BGK3#Sh5~ zj;k_mjDPbH9^dSpW5J zPO>yU%m-pMmZEFTLQfx^NGA;pPlKg-&zyzN{bN1$e~fEeKB{v8q&Mg}*3O`OfC0_; zs8}m`;S*UX*WxDOLq+FDjCqaNLht(gLe8|-TO(zKLa49ay*t=qYXRS(4Cs)@@e}q) zO*^Du^+Q4@O(%W(O1oPFIHnH|{HKUrlSx|zAOwvpE9tv>9}_l%2(|I{gly4OtuJX} zV$`ry&G7-}mw+35v@*a@Nr}(_u?K%}O-KGxNs5Qd)DOwfu(D1$Bv>RQu}-A_|AK}6 zcL4f=e*W(byD0YjSM(&MprvGE3h$e3O3VF&*hSaU^TTsr??8H)WzWFZYf?!x4ZfYb z(EkQ%aHvGxTtH`kR=)F{RgqIn=}CS2!n^z(n@=T%XY*w)1n~LeTtiEf2r%04H|Jwm zFs%V_zND*C<&_a~NZSn1N)n5!g%c!nP@Mn=!*fx~AG9Lmq)Q*KIbRxM7^TY`=_Q)E z!1c|4HE9VNtb+#Cx9rFRe|QzMQ!WyW_nYrU^QlFiuBX_{hsy_1M+AVJ4Df6|m2BDI zWX&kn=Qy&q!p-59TCwt-#V!=|_v<7Uer z(e4+c+Ga|!qqH6#$De2$T+n%yB>*i zU8z|)tr67H-bmzt7ra-_k2Z|IIN}cKg4sT`z#Yv7(Sn1~=FFB_6(v3uo$MqT{m8eU zoPHC&KAv*i`}yWU`>@zPy@cBGd>SE4{faEb1=9Lf&~54*wekw`@*RMO#}G<+uskf6 zjlBF(Y;$2Njg3Mndkd|EkSU2W_~l&bL^};`?qZ}cPs51#QS1*XJmUiOC!+F~d!Fse z)$#tMsdriI;OT`*jA0kqyw<w)Tedfq<)J!dhub`#3eFmGr2 z7a6BJ2l~2x{-rDQ#S4B+S80H=4C1qNl1@n79(Qoq_321jaA1v zYQOa!*6IXL2uOx-MOi$HZ4K?$YDu9~os?Frt>QI2Pc<+7G_H|e*IT3P*KtL+?bZv- z-S0(YmvB45#unDLitANuB?T9)X}_==C7f>*>PXCL+2hXs#!iv2QfhnHkk#V3p+@U* zpYk~1tWCYy3nQHSwT^{+=f(KmJUQ0#=9>KC5Bl3C@ZMGCy+Gl%;+rL4fADv%qX%?BFAYEQ{`znvtjv0RZ5Onqr;C8zvK)w@xjO!ylnh&7d=htdiB$BKtwQP=7`N0m;QhRFq zojio6x+0$PT=36GI3o>Nld&2vi%z0FPADlkD9X3if!VZbjBg{i?Bd^wWf*sM@@<0Rr(&cY}>OXA2~s+IGZxFj`phtZVoOntk|msQEg% zSTVywv_w9Zi=sFAZ(3H!w;=z@_z0tV-6`&Sa5Af>sg&xfTM2N+myg8l(T`B^r5cYH zFEf9?ip$Yfxul@;DCFBcc38Fze$GA&)m`>@)m^Q*7U_eBVeib&e~!{3Mshw>(8(mv zgn1G_pmVv#CrQ#gScAF4Y-bu*bvq0D>NfqwZPj;=Sz(Z52a;i(4K_^MQ4ePn!M%(R zoD%qWed*EL55KIpDE=9KcsF|z?``9FZ1q^L>T4CVfBj{-YfmOm(BQ=d|BL*GlS`Vl zoDBs(%?D%^7A$Irdvs@OQ!x^Es}N5SD2Yw`rDMEvxQpT;(Pw^t4|#uYWgCZQImI^N z?P@w*yxYaji@s!u=|7mucT9+Jo@%`Bc{u69aR}xxR3Q#g@2hwH6shUMnNuqNl}jQ2 zo~symJ_UbA#aVH>gDL;-Fpj^mK7W4z{FfZ%%`Oz4Q~$su6Lpjo!zh;URAXFm+hXqU zf*u~muq2wDBp|d+l8MHyq_5{^q0o$1UWK9R8LsGup)h=Re3Z*!B4kba+E@h=H_UhS zdX+bo@qxHuj*5z&SPwD7vnA%Ep3SV)=64g7N1W4i?8c%0NYz@M-p&mCE-o32dS0*8 zyJ9{0=a<(QFSE2{4ZAu2oo<2jx51S?3ezzQPsEe2NI$laT10-P;kvSh<1*Fi-n^qP zLK+$$KKCrU((N3v?G%pxJT8SkoZ(MoScoO%Up_V7L*4xfc$%c|b{kH1{|gs6jE-n$ zM6cPycJvJ#ZzsXqlUu-X(dIA(&WJUsKvPlFGsrwJ&V6t(+c?QBve_*fA)?Q!}}Xl5Pxt`PflNiS0%PBluPW;=`vJcBzc(;;N$cqbj40 zLhVinlO&%b4RpW|QL83nGR}v#8uQe{elJDi_Zx}F?X_%{x(_Tf-PZdDGVRd@Ga(^^+_skDxjjP3 zSAscwQ>#l`I*HV^{5=A#MQsFJQJ#fCFXOArpIyRLm&dp)De4?XT(1{LE~zqXNjond zl6JU*pR)`tm*fC0#zJs3;BLW}ulnkYFEVK5*XRq}2+<2dp@|0%uq9!fWneN(D9ocSg?35;!{Avw?SAWNH zZWBB1rmT6X%st}@vbfA51^aN%B%JaQhnE9dmX0zq*1$^-S9q0=m#ixGc=3RrxjLX5 zm*ZR=(2UFBuMQZ*JyE_agxk8t&=74Xz-;0ZPZr-ZcHnGe?fbL)KW@eWQi?hN#n;hA4y9)e4mVZB;O$ur%_ z`n1Uh)P2RNjM@VRcbqmNbEiIM(YQ^`NY8x#O~duEYxIr#&01M=n|mUdQ8)BJx)FHe~6LMgBE#D{vtgy^;AxuSrtXl6ZOtpXxp4gz*V}goz46?jb=bXZr zDDx!V`=s(0h3JV{hFlC2!BGn2oAKycX+ZvWUo6+%>7rjYqa<5X^u29^q%^@M(O^pr z$}lGf@!6nsvgrGkbBpY7&%B_QvAm-2o<~%;mxC%lXg&zz?9{ANrSL!U3JkhE@x8-F z&2L5QpNS)z{nU_3{AM>zc&>69jN4(f&ugz62tO;y*A=rwQfZShp^@Lx&I#i9dl21r zWNRspc#PCP$=)Z6*4ScnB4LV7ABn!7#Hi3iq1B~Q63*FFOZHV}E>xs1F^Km8{5NeM zf+xRHpXwIRRkR`1t;A%rJ(UNRBY}eMRCUE@{U8+A=Mf@vM~?$^$<&$RIx5nV zDnY32`*a}W1-1-2C8(ZTFe5dRl!L9LB`LnF;h+|GMnC#<11DdRpxWl`*vJNxf1q^V zC=<0^&2W&p42vzWI1j?eB8)B5K^to$u%vSi?Bj?(5D>Gx)y6y^$Vcg{#9~zA-p64g znP)v65@MPB)#SME9)fgbqIQIy#5-NX(&meXA~H@GS^_JO(iqOu5{HH z%dB9L%|1@KP(bnW{(U-b8f=Ydcq!~&3$^>u4Y|Zz1fU=g4n2&L{3E< zpGQtnE)41MX7;MT8G`vFSYG=6P8^xt#lZ#g0J+;oVxjgrmWD%qjwjbP!nyd<>1Lu> zVM?`B04ea7z;b{b$*?}MV(wcnBfUH(C|99hS*&Ama5>>B_MEkHd)gs8gzLXRoZH0>=6Oqnuxu&mOQ#1CTJ)^Y8uKws6d zxDNk&wd44?xQ}$Dn~pD{qX4dZmaJyVCnX+V@s`_B>}BV$nGrvy1X4A{^A5Wjuj^&W z9&w$E69P9HJ{52%D8d{DN!3lQ9mb^8(osqY7ZZK+Iq1a0O|`KvV@mT|&1x=2mR?me zD=R#gtYdyXrzPMn4V6skC$Y8LmA-%eT4;f9)0#?t%M^0x{MHAb6#9Z1$))ehq&FwV z6o1VNd;iPVB8BNKNt}S1^>geulh}&5=<3brEY!OA?L>}(YDP3RsXxLn4VXn3`&xS0 zgzp*rpWjHn673i35p@rRVAOBxnm?5*1EG0;otcC@Nga0uf{*p{6*O9P^6P&99!~kJ z-{41oI;x@AJ^hZF85>=bKv!6ZJ7rl_7^;DEg%!qlGntbG_PSG&9+)v@8A5H2DER(0 zT+TNY>1%v*Y<}<9U|yz-4w{&(%Vlk@rt{K+I92n^me&jkPMuxtZAUT*krBBH0mTIl zrTK1)Q!&>qD9r{a(WXJyvm?Vz|5WRlY&|ihzpfB{Y%)`+XA;QaS3A4nnGiu+0Sdp! zTrSCkQ5IT3i0B=$AyVHrYgTRpF(u_C@t09`3CQ>kdHz2l1PPs^h`O!$>)ZumXbM&O=L?J zV*nbEc8Hmh*P1n|zFr#sr}t#!)a5OCEoWu*Gopt-Rw+UTTqIQ?8i;KFN_5}o+I60H z6#F#T-5coE7sRF)53(mZ^#xr}6GrAWU&uoER)jukJaV%ZNMoG+JAk7!Z{8aMGX5G6 zJjaPJuW^PIEsWjW&wPXUiI*#)<0WuL9(fQA@Y_SwB-wg0`^YSnIg>q69^JdIzgDC& zMF++QS=ag;MrMV4+}^4d&Nqn3eqyz3bLoXPVv%B@n>6%epZe3eAd;P%Q0SzNBfFt1 z*U16W`B+-O2oX7K^eV*hM*e*YI^Z9&Y zGX~G*p5s?~ne#$&e)9&(hxB@Ud8bFvmTO$6)D0egY^K{fB`DT^JJGc}3e5)afImX3 zz$Q>kh%@WxbG7oP0vgF?k69&d;n2I-cdW2+ANXLx#939%mjDS1;O~pn!b;D zl$5bH3AcR9%tGCVT>;i$CL^qeoqmEqHY6E*bKoh9GosqG2WkVn!o#NR-=^iaQj}-anNLS9Z)=M?0Zyu!bhMO*^ zT$F>v!Bv)QaPSFb=IB>AE-_r~Fu1!Dyc(WmKs$k%k#IGh&+K5S7+C5gYio{FD%anZ zL@H(I2dPZFDnCy)c>07AsZ^tnCs}y{J~iL|70!*Co0m=q;(1SH7Ni1F#dTVTkr`l2 zi~153=Ff^hOHO)wfS>Rf*>*ACQiMFoFZDM48!PAMOMMT#5zmz#*<^vvv`0qe^G~@O z%oX5}(pmS1x($s}g`tlTf;-&U7JFQ+6$-X;W4CL#@5Faxy5mFAgbDskBa^A(U+mqS zy6k;$XtD3abkljEFHzcPfjiP8y|S^sxb~wXw`AixaK18dKL5eCmnRw@LfMR=vj5e@ zpVXtpKlZ1$JY{y4bHTg%p!N))^GW5tFYWx6qY_+g|z)#$9wiYkVk?fj|1!S0rF*+CgdJ1vA4G*;SICz+6*8Bn~G>-L3Wfjt<0`}BJdt57ewxE*Nc9A#S; za5cc*Rbm=4y4rUI^)=@710o8Zqq8=GCr*j^|M6kgly9f?eb{ELREjz+c=QiLs@N6F zc;sxc{GI9Xn-TuAV*lFU+Y(SW=c-lOed?LCH5$*$*SpI_6JFCjb5|V{ji^eNzNQNY6i|qwm0}g6<0Jrh`{P9hfrkb@_qQs3}TCCFz z#T)yyV0Jf{gB2F5ti%i|aX;xkvhj-e>gyg?vqXj~d+0r=Mkp#i-gz=ssUAMh6Tn3W z7k8$?Qa;$I3Jak-{1XYEHIMSJN3&3O1i@#S&{pAUcWm21fp_82$6$Z(laJaS@|PR` z&b-jmeYV^@lMC8y-+ymep5tzI>8V0heLYY*EDUYbos?GY|GiZ}!QDJFN#<&Q8i`5) zx0I7^X4A2`;-%4OF0{{jBI^I`-4mkBxwb&jgEl6dW$ho(pVIi?F7**HkqoYBm$rDu7 z1=JNr)OwMKp$c#e!mZuoFSPoEdCUFF>o6vmzTaXtV{%rWvcbJ4)^&1;0`you|5Rh$ ze+i<)19|kpz^+9)7@u$8+b)U3ylu<|FDM(B=Q93O^nkzQaOA~i7I>x&QwEwf z-Up56_%cq~;x%76PA(mQ1YOnCJ}{D;Ar({0vlpgzS^a57<&OBA4L)($h|BctrImt- z32idof}8>cS~}Yi9KR*bC|r$S`brpTpoxo{mg>81EC99^RdQVl`I3%rtr;vC${+zV zihpedbdv)8#gMBot{-AZFOllaG2EEg&jy5A`f3r@Bek+8puHBYuEk;HLg|)M+2egf zT7AOU-1p`ym)p;!khWo13C+^fFT>v|Pz{I&eACxFvuC>TMY6MmhtA9Y&5-Ew@95NEl6vA@j8h zCZv##mnt0la5<(1uSu|P$QQoorSr_r0t*C`;0=7%L80Cm&2i)BY*koG3hbbO1(tG; z!NCx-sDe|peeJw1Fs>GNGvO=gddYeafFLGZO=o`)r^9Af1lX4D{z@auQ1Qz_em`r0 zSp?5_KvJ3l>=R+m$mb0#Fkh706f5ke>dOErq+axigTgcfNH+nHDIj@{Q?wTtA@y?` zEGq+R-&}#&(ZA``#;mu^x`0<+hdWu)UL8eUN^hLQMXVf2%PkIpL+oT#%C8J3Z(ack z^Usa8ynce&y)aE@)i4ycRU*zcO*dm$Ph+nGxlA36l8QcoP%0b}DGEcp^J7zcuYc|O z3^@Cp;8=DJ&6ttu0pgW38zmm@7R3kXb(TBiuR#mwXDAQo1`gB^f}d`n|D|#_N}`pL zEEH=N*EKenC-!Oiee@9D>OdS^Hz`N62v~o}>S9m`gnAcotxmSEkruV^kEBdW7agneVI8moF8%&AB%^p2Qe(;+G==%B{8Y~jts>BA%)i?#5 z-WPQW2oxmQAq*@45@ZVnaJ(i?7}(0{UG4WDwd42t*vsxwD`!+a&AFAzTp~Rd7Wn2 ze!P}>1@&9L{Dbxt2?rBF9N3}YL1ks$j;rNxK zyF;-Zc0t;F@`PzR=GS|MfafQjPjBVHJ{Lp|VailMqg@C!PqM*X*~1Mi9Q^E*+I)g4 z=jLnG36JD+A*$u}==fuD{Qj_!*Lsn}^m!Oxrb;>wF{__A&djudg#Lq1 z>Xsx|gCx@HzR%$+c**xKU+u$MxQGwlv%^x9F0H6y7B%ROE5)bB+DqnnUkNVnkIfH} zhFbF69;+q3F|HXFGRSYvlP1}jxY8EGX?F#5QYk@q5X2OHLQF|vgZW{g&<(|rUV?0b z2c_2q4LnEbvg{w3*Ln^c+s52Kz>wMz5%>LQiqXq0&n z>V8b>0A-UXBFieLU}zfZS))G>5rGy9X*4KiNXkAE#|#_&ux4%2M1&lqLzNAZwdp$o z`Ry$WzaA#U2OisicN7xSL0;|T`sEo52>j=N2QNOG{HA3X1c0eJbJ#=VO9kA8~3kfW521TR}ae%T>96Fz>dM3>s!;3>EvrP z|Fm$LVW{!97c#|=j%Igpu|`6UY?eJ$30#{rGM@XH&jYlAXV|wZ zG+4_EIBi;!8f*Gonu`ZT%XUfLjzVjV*IxcvKy`nzV2x~9mKGB+muTpct8Ty}X*G)}-C02mt^Nr!w4s{)zz(oM6k~ zMaLeA>e1oKRodb`|;ZFf4n#D>Slh!(q$elDLCRuQ;Uy5yqup3EaE2^SI`(c zOrs_Kata-tu%K1Y7q0eTqe3;&a5tJuR~)-%r}XC5Tz3PF51`t&EKtsa#MJB&;Z(mIJNXGwg zljMWYXyM2ImRaU}Jyh;K{%wG2chr+FBDikI0Tgv{zqagT6e`P4Fqp%OV2KMFS zGlBzgG(`;1g=0{SIKP9f)J${=FtwK7lJSy`Z4CgMcHb`g-C_q?Du=Wr52>ck9@3Jn zr@BMDC1AjeI9|T>;A82ld{gLC4gwqFfUf+KlNI6#Ai)BLP%neriC8$ik6vKJjY5%A zj+JiK!>(8TM6;IVhv;H>Ruob7mgJaRq+4#`Y{j3-Eh8n3|3w)QZT-dqC-Hj`{;2=sA$ z4kxlBd1Jys>p!LKYc$1qKmyFQv61(xH$k1aqK)4e8j0~h34V_jGFu-)zv}Fz^_xPQ z@saY}HeV6t>=b4Jd49bRcpj8{PU-EeOtk(t_lR7BG*~YNG<9Ms4M|@ops%AcYl3tH z@>|J6NP|}(4VrP&rR_8bwI&On+f<%qSLf*Wf|zKRjDk$4vcV*14*cF)WKW@7j`N*7 zA&1oJP@d;n%{C+b{3a9%*|u%ZF^fzr$v}F`X6$^j)&RN%&M{&CP%HdvJ}4~9M**J2-r5`WbK zA`FmVKOM*(wY`ij0R%JZJ@8~bgIcqP2x>yd+*jJ(C{wW_f!@0QBH&Alu2II28cexe?_>g#7ODM% zlG8WG%JFOh_QAja9KV| zrAkW80flrYPWC)W>B9~kfi=)^aE|cEtb`@M3_dK9~Au7px?Nh;eY#UUDj^B=o9MU)URlBd3>06QDek_3R^stfHM&rOHSg2TW zyAV{egN~`regWLoC#F;=tfhg`&X^W!d8xAy>@Sg&I)e)4wlq-JfbDnOe_O24{kD_P z5agEV{>*AIfIw^_&7VqXUV{62r(V<<4YKu+P+1638qgaQM)X*;mX zkQLO;3Jbca(#r-5IznZtS0uQoGGIWbXy|sNWKpFaYs3k;8ZT+YiFi^6QS;vyStuWU zMy&}=yDwad+EjJ-hb_S~KTkGUVS2^wc=)#2m3>)=nJIz}TUR0UI8!!aM}`JV!?Pnp z3uJy34R-5teibc{&p``Mfp;g8l$SKnzdRo87ynrdqI&X>ia$&JS91QZg6-bEZgS3d ztEXP**5ehW>bRf=WY%4$H=lT5@v{eY|M&ea zr)ZCn=PB)g9p7qhcIbwq;JRWZ?GnIN#^$kb&_n10y(GpX8(^jkBlBS>p9>=wfPAI= z3dB~%BT$i_V>oIh#_y}#Gt=cFyv~_dnek}>NZb%qPK)3D*hCJ?JV4DC}21%_* zaIcC?4mUA;zdx$K0Uf6Nhz=`JqA2&{mecRHuQ zLcWEo3e=Is?aYMJ^x}%C_YWO`AK0;5?NQee9X$dO(t5uI7h!3={|e4nhnfTre|d)U zTILiO{J#||e4TUlwmLs9hh}Vk3T&P;XxVtFgj}3@6&7-FU3UrAw_p=X0?Y#lZa&Cf zmA5^11xwbq0sAH5JxhS&i9!B{ONCjY;=dPlTEzyRMUUT&PoG`9>Mi$-?H_^#y z5S)s}-cN>k2erJAjwOLr>1p&oBD;InDRtRVXz>Xlf|arT1Z89@W4^rVP>(~P^>oBE z6m8dqTU7AG2%o1RG(Mc9>RizbBFLL1+6x1dEd0)NO`rc+$sVr{y!TZaGcfn1sWdH9 z?|BVd$slypV-~7jh;?q3{Hy}Kxhjtbemchuhx)VJ|3-uID_?_>xXWAnVaq|M|MMFl zbH*|&GtYEOv>k&dvSy4FdZSuL6z`ch-$|!NzWi0sla~mqz}&!rz1|4n^Q8~a7Y9az z-?6|den<0w&~%RmdN54#I^kK57!ZA+?X?G%Je#ysYr@(YAf{1gx~NZ-#LE(yFxigc z#ypFtOYP7cgY_3c#F8B!znIz0SNP1CYNus@?LZxUOiF+8Afu!`8}9lA&Cj%^fo=+r zvpBzBO>oWZeBCi}zxbeLO%*P4tzLogH6%?9d}tv(tmiu=R9AmEr(*rQF7-bmP^hqz!L(O9zdOBdcGgvzHl1|~TQ z-95S+`Ns^isH?ZW-3g4`C;(c0J^p?~XD`v)L>TFVQf-h~lh?sx3PiZOVIiK_6H?$X zkk(!GG)%am7;4fegKU0#@%TPElg;QF)UTBR#nMlr<8JaT84ql!vXk4ag(!1qH0Z?g zl5~G0v}Z?%+%3mI8QpAt-}yXL2IYSze>iOv>CS@={|ZO_$%Q|N#^CP^jre@Th zDDaPrWB~jNQqiFTf`=gjyg%ZSK-lJTozeTvr_T7?OhLo4p1+EHfBi>)^i4D>uN_r+ zk5=eW-#Hu0Xjy5?_`pK7Oi}YEl5W}FNF)jsdf%v#1#3t9v1qrX#ldz?*)Lv}_{wUn z?&RV5_$k8}J~tNDCturPuW^6Eo*%qkN{)5n*6yZ0vL2FuSs(PY8y(f`-_3^geV{ZF zg>to1YKv~UPGTwp(+^+nW5-f{DPWEY@w>p$7mdOx=#8dNXt=iSCPkshejbryDT&5+ z_!)`)&%4c^9ME1q-HsF%k){0Zu;P2H|KxSY8)gnNHrSN+%zdNU`Yon@^uUAv1{vvJ z^F4fLg;M8gx3_sLMHFg+ZhczIn$XyJmR)F95nVABRckaPy(8Ztgun-Ftp*@wr+xM` z2{H)3<-9cCkM5?9d2WkLLa)jhO&$)*bY$Zxi5~7%;L&>T-f|-FnaPKi2>iX|Lkz-| z#AtZ$9%f!!M9<=SBHDAI2z&NHoYOaIDE5WSvrAq6^+0p=IEgz^s122Xhe(YFO3qO& z^`hK1&(w8El4IRtVat@ZQyDEovq$go5kv}SL6CgjfkqLfC_hH>67@ZQ1W6HeJQ2#FGg@8J9Y4qCnEomgu%6zYO3I^ofju!6ik4!>YgNQBidYqmFl3aASpV;Z+=7sNC`V34eQk>YsMP9 zBr4ee8PU8~8rELdWp*{@IK;N8HRi}fx66Q3FdTh1(YAcNQ0;aVzlOf=<1&vSs%hlW zB~6ZHg>2%cO%EDLKSYkzO^(%%h3z)`zhlL+eer+Hf^{d;yXSF>Gv(|@1_Zr4xYpK* zC4!OI=QM?&k{1M(%m9^iB!FRqk^d~&VR-VbVPngAK`-|?=Ar#*ZW{pqOV0jmdVvO> zXNB;psFvpv*HKFL>vhU4`THqJ7_-6tO$Cn=XA*lwQP{KSx14NLjZ2FGk}#21Nyfca z^|x?{STJ8k_fPSX$85RmS~32K**nh`wCuIszx@Q+0i?)&37E%GH~vayez-S<&*5Plc=%(aK8hVO>M5ua{_;k-d`3**vZ>mpy36v9irn=V7OA zvBn;;5{h{?raF_=ZoP_2i(IEWs`9c%lVbyl4-dHRRpoQO%fXyv{^nZ##no`ECC|Ui z&-;_h{x26}F_-*H+1l_%24v%v7-<_##_0O7?MCHd|KM&pa`hde>rya%3b%*Idy?Qw z0S3y4b{&uX)g@Nn=uaMcBy(D@V2#UD3)L&4Ij85J=r(dFbjH;t4olN{t6Mrbo_Vzg z|K0nkJ2G66R_S{6?dsEy`!3#BbG*;RmKGEqII_2w<`o81Jg>945g_XmAG<22Gbx*h z@Jx_pV{!DcLf{#d?cyU4bITU;_5YTB+D~THD&=C<#W$E$A;A(7H1tVO8(yDy0!YP z?a+PoX<_B&k|^`ukn!}}JZ|4&1LbcojGvZ@oo3Y*5>6SNQu}ZHSd+^?_rpTdmU7M8 zkha75)i-*R-tNiYpRVR@^o++UBtr*-Qyxb(ek!D`s5(Ws>;L}UL;Y-_X%bxO@fx_) z@v58w^qzLYdj?_^WTE%efcKo4TE16ZyM(`(e<-&<=XHy%yYhcOAoxZcC-@D4R z+c=|)^$0mbq1yF-&c}6sYn2_%9vwS#y}B@ZIApJEFq0{$6pltJg@19n&Utps@z!`( zuXl#!wXm5l1HExp3%zfB+VY~QLgp~Pi#q2grhw&u;cREGzt2}5Y^;hp9B(8gGp{;M zh5Be(rUo2HEj84#xsa$l>-*xNc)IK@F`g`iyQ0z1HuYV>rs=AHtsP=oXnTSX!nFA4~PGC?z5Xb8&9Q;4amgHkKcg3SME($5 z;@_SX=>TcyW)f!`lj=Qo>}LIFXuSk%#@B@s zWY-zr*ps|EW0_CN5qr}z$-};pdAbU5>m_#X$sNu!-Yv@uk%NlIquzDD>o&!NEKe`^ zn%3T2OB5BF-i!6lNW^GMY;DaMgURIL)2jz2i~n4GY|>!B&vYt!d*;&N)z zf#4qGeJJ(El;$5=|KEUG`}4EwIKbRZn$H==$4zSOGet8LFy(s-AsjyOhr0sq z)%M6|VJ$1;;KOC@wzXc-Eh3BDpPXD1ce$SP`XLu?VP^h4X-|x{@tM<|Gm#Cm*O)W$ zDap{7Qg7Tl-9WJi9)1t>8x%utCCZE+H>z^o@hA_gzZR1OW&JDN?%7L z%JxQ+}q;^I8=oDZWH<)LUp|b z9m2M~JAiIsEPHw%`7p^;Q4-b)q&ST4ovETUjGEZ`XbRmT+2r~V8BDg&j1KvCfhnG; zBH%#&R_U{~zTbC4_05i*sVG+)=E-dKUaHI4h)#Pt!t#R|NbBl z))&clQ?Ue|s{AAN4fDu_fCKYxt*Q&h^6@c8D|O&vmOqS>_NJqf^kyJlQzYDPLEyJI z7NL}ZQp#pGgyQ!&{HzSSd;d9mao6a7T6Lt?6OrpG%1hd1h6* zZU)(TTIOM<5*89_mGUqr|J*DsWgOYhzRVbmGFJo?A}WVHKUMa5Tky~;)afRBX9=51 z{+Q79N%4v2SGjd!gK&-ESGb^&PNeIj5se~F-%^l*kw-L3JVc&n2(_O=U*yqRiO2Kj z-a5&|`*d~XC+b>z-AM6`=eLP3dVtVvd$PueD2Zz;=;AMFXf7~T(8>>TU6Vp={~)S* zhCwENtKLh*n_Pay_=6qW9cZq6c;KCzS`zp0bB*Bh$NBQFW3O-WKV-K3hQ7sl;cN>1 zjwpEk1ta+M0A)!4g3*zzT@%5GBx?`CQ^JA)eJ7)8a?j zMMfa--$%%2y-_HRXeU>(+{(>oU-DlL>DGUxcxvPUJc`F7zyF^3!L^53lk}YZlaPM> z(&R^YU$i>RTmuEh`~LS@?Pk@R&j9PLRBJ!~`&H~NfVqlEF0~J$+tOeT^XcjYuojJ9 zf!UFCWrGEw8y36G%gzA@iVw5uOs{SGRat6s{(hr!5<_45j;}4q+VG zU!wY!KT^6|%lndCYA$pOSJK`NTDxfcXLvEH?d`qtW_gm&8kO>LL^1a3`+TB(rT^2q={zXr!_5%4QEYSp;JY)=YqLsfDxE=&Rf#}vs9%v$n)1MtWb!&1=<~7pYbW2iLu$}6yY*Mg<$d`LVXaixvkz{<1g+aam0$K!204%y3jaqDm zcg3V%4Z7tQzhzmE8AIm_Ll|#kshkzZMJ+{V~V&fuFHG^;P7Rd9y7$h@R zwK6JX(`)ryWQm8Q@iXCzIWVzm+e7vk?$QT7e zE~k`H`B;NDCgMXZL6RaEHCy`9@DYgSe;w!$#clV;$hnBa>tZlh)X6(GtSpJ|L!|Ya z!m^nck95T|{s-q97X?ylAcY>M+1MTQ*1CuZP2c+!_=A}f;7!G8b&1~mKne`$rtPme za+w^^g0a^7Ly`Y{xr$XTmEvA`00gsYd#CeAtqXwJ>{s9t1oQOCC@kEY-D+qz%Xy{; z;JDWByQ3fim8W_AC8g$qk%|COM!D1^$WD5p6VzVa2Pe4sEASz70?yDFrdfYeq_t7) zi~jAeFRqEe5Is0A?LV4nmO9q5#*=2qPiv|eCwxw!c15&f$)`f0iURd?z?Ld6VORg= zS5}}6Mx!T2cy%r19of012lwnmRa>8~tFi+mr5+*$nVHtG1SE9{NZth)rd!`;CBQ8M z6IO41BMN~oV9--L0?sQvgV7ZK@aiUX_gE&*<1O> z&-}>hDxED`({RJ*aAMfLsi)(-p>ZK6;pW2ra5_8T`EAD{qDsZMMM9*TxopRQ6E8mm zB@;sR@y!2;*4ys)2?$C81g~6;GPR=+S-K z8|ufQfN&6(3F;l2X0~K()^P)c-uva1b(~tc+0dHrA)jWW9??J;PNvyx#o9ejPH-0T zPznFfS!MFoN7X>+hw!K56jJ>|7@AX0yZ^S=1@893bjA5p8~eL1nhE~5;rJYI)9{$4 zq^?tN8N*+JMlS(wmLV|-(|M(1z^v@kNX>*mu=j?^8F}s6l>ne)R4M+{{{qWztd`4K z03tdA@cq}n{rx*!1a?MQHuyW#`A%+XW=Kv$U#LY6CZ8i|$TPRH#LYr?G+{!6%2a>! zN=0YtHJ^{cMWvepJUUaqFJ$S8m#P6 zu*E3x{ULO`MCVv0p5H#c#1*02%Wnfb_?>QBcGNBwBWf)bsZz%vI{}A*p=s zMXAKls}!11XnpA=n;VMJ#T32{$m|%bl_cy+;Nj6DWMb(5Gdk<~*do*MKV_w1H&p+N z&gX$OO2U-w0oIf;?!dWuzJuYfT9zoCx5nBv;3u!U?!+^v?m^BwnMV555(H}qM;FL* zFky7RETPaD-B0{k1n9pC1kPz_smyaj>)+0#;3sPcu~vl}S`J0pVh9rTUx6{u#CtGN z4o%$rWzvO-Uj)`5f1b1dA#}uJp**X z>_$$c)hEC#dHht_os??1s+~20vZJEe2pCHI|6qu^o)M);SR{=>cCyl$;`r@(%DDLa znZx}drOPfL5G8_5aFBH?o8Y4gT34xEJ4OjeI8t+hNZM!jP;z6fDQF zf}w{cokmX2a+ZaS(W0GDCw0Faq}-l<;s8MiFAx?9>2m~lOBRGgPu|?IH+w&+h)_dQ zzQ<@9DVs3lS0K$2JXaX3tA^H@{tDC}Ko~;zC?A^8{S|l_@=DGSXYG?p5FEIGi-pHX z@c{RQhM_SPQ(C4!z{TJ96Bk^99RJSy#20IhE@ujSz?4E`8k2WFG*6zvR$qM8s$#7r z&kBx%rI!nzuYoXZFo4|AZuS!hyFg{^P~0uK>YrCA;>h0BjPZ1h4rwcWV%$M$T%Xgx z!yN+Fc+dhC)E{5+k>Gc2Q4nUg|ASWsf>-(hK+f)QC&5uA2rl}au=F8V`V~$MTcok= z%?R0?!^B3l#F17#dyH)O+EQS}f?cFnGWWXJ#qIAkozc?C_+eTSCZy@V{~o_($>G)x zT(r1W35qDDVoF0FD%gcIjiSIQI`Ky)`!RE5f%W1r{q~yIGJrfY$V-Bd8o|km)j|9} z!Y<{IK7z40Wb8GbQc87w3dCx-*8_n1VGUS*PTUC$3-g*N?R-8OJna5Nfffi6D5NP}UgPI-k zrI+`}J;A%fz|xs~M1-+Tt&{+5(?pqiSz~S4D@|%wg&@n%6Xj zDvE%m*C0a8Fi^66l|Y6Any=uyW65D45fqT#|SLg8n9q~ zK-CA(C!pO5B~U_6WhRg)ZGw-2jw%9o9r97oP}14;G93BqOjjl-B6F7eF1GG>!J#*x zKTZHPPPx>12o&QqKatZ6Lf(avAM@_FTA*?7OSSZKeV}s#k)oTwWcdpnN4%9mRLtn- zbX2wi9oi}azrMv~{A;zY2%z%p*PRM(;eM}UiJJwt1Oc$gA!1Dd2kZd73$@z6H%mN4 zz^S}`$(mZo=l>7BN>0x$rt&wTpzt=$X7K2RE1?*ZRy|NDip=Z(GkpN^!Zi;IsU1Dw z@PNzJ9d*^^2(de93VV(aa_?IIXB|O)0l3y$DGu;ryIByO0sK&O2ALI#&IDKXUVWNc z+7H4rC4~FI3mhn70&Xas2N`dLZJS>=RRAPX+xTd4>4KED3NL;e{r`M68glO*0_J|s z1j8607)B|2%m%sL}=U?T}T`B@6od}$T}cg;Fq>A+?a zc5<>5&hrIX3DBE$zR-!I7hHU%UBj(4h{~FB5;#!3xo=WpjJlMz_C&OK@9o(0dAU0a zJUGxGl`l}6!m1&q3_E`(y2-~^Kh2tQ6sS_YnPb@5ljtT9k1W!v0YAML5;vxP6uHqu z{g`)czXlfNJ(IWLi(){cJ#@4_d z-7T_GJ()ah+2u~#hPkDOr82bq{ASr+4|lk}-@NuMyLBIwQytEmv42Jv`KBXJ8PMf6NE(^Yec3qJsSZ<^wsSmUAa_nECH>qP>6FL zt>vF-S8a>>g)JSZkY-^u9(BL=mQ7d;ZmWiwe}n37r9hO#-+$vBh1zmGM<3N9NBmJv z8Wu@(@{vP3-UVkGn{I3+U0X?kSbnpa1WZqg3^TzEm%ni#m$773QW#iH82_@<=YEMk=Z+Gw&dr7&Gc3W{nF2mBNKWp;~$=|h_tByH`mOpD67R&%&ZwCNJNaiD3L(Uh3-1 z#fnuwR|a-4>St&F0Ak@S(Yb;(3L-=8LVnANqW=iYmBCE&mw=&wjy5orZ7qWAL<&K#!Z z8y+JEq!3RX>=0^&?7IE&(F%1A$enAfZeP*1TH*^%2=Pvhu&eSf<7rCyTt9rf_#xl{ z(x3T?i6pG|3#)KO%j>pDIz=rXPUnEf$SL+JB^lU;;zzJQ5glvtO^8~~FZ!fee69xX zBhm&tY>^*~mFa5pUuVPqUyS{ASXNKeHVo6<9nu}r-3=-TNT*190a8-Z-6btZqZddc z2uOD$9ZH8tcgHsuzx#fk_mA&Aj@Li7d+(V&d)BO(wbpsg;@We_qCCLec98-b-Ojt- z84DPRFu+LQg)%77Q$*&3KU!qv$73(duq=D#XZM)LE|Ur7i*i!7A5E$#F!UKkQgUcF z_DDPrnNPNkV*lgO+Ex18WKqY=Q0)2Qo(Y9OLCQOZlJnB{gh4m^qhOptk!hpN^c^zL zAElLK+Pm>aCYUKF#V}2NH`1sG=DhHLVQL&M#`Qkxe;0}G+HW3tER)LgCMs`PAMnM1 zP+ z*}O(n3awuqHzleJ>&=hFB=))GgsPH9s-wuXpgG(^Ldp%|3A)2nuf|m@Rrxle68q9E z2lSNI{jC_Z$|(}1lVepItQq;3GmdNV$zk`teNE);Po`872nSd1C?aDbfl(Z%>q`YF z^|SU2CZGznop%%&Ru=2e$`hR&jf;0#jWI^0&%rkGrM+y2$59<&Jb|h5|F}XDfrHr5 z6Vk#KK$f&tzRS2^t@3*#g(LCY#~RJW7I;+G8Vkk|e7f_tURZF14GWICfX<-!=0InY zXjMRGlMy+}O6FcKZzY5PZq1eA%|~aaJmQBaD9y<%KL2Bw1>Gs`!d?$GZL=HJU%I2H zDUm_dY|43dG2QqY1TRf5I;-FRoxTIWa>Ine^jYuQpT4d6{QCdgS4CzD*OKyDyaQoG zlQ07nO~dH|&3}(%z4sBA^uPq3j{p2e$yNGT^!85}tc?A&Qs7g)=;0p$bi;-1Q>bOk zAT>{C-@7ZB2$k+@?}=3iNs?oZQ6OPLU>&ow5h z*cH}bGkof-P8>0%Ow<$tsB$Hp){*I5ppX*SdlRwwZiB1R4+)}6%i@M6< zLxiN_yXe$oU||k2dIf{XOUdUj_Rb2#EmrLfiTc!)_vsVGZ%gIV<0az>Sm4$R9*LpU zM`DN+aDO3W*^`&kiFn@2dPM)LB|o60C-{#m3;j_`XjNck0bTVf3CLO@&S|T!t@$=) zdlHdb{r+XQZV}#c;|k)}jXOV~qBYnRCTxwQWi6$~{W$rN#xdSs^4_a8=jp%8*}MZ1 z*s?mXQEznHD?Dfc@sN#kmiTy?A&kLIySO3$D1UfnyL86}R@}{<{SIfU7O`WqduYry z<04nSUmbo0fVu;}XIL$i=)dEJ#!yD$3in$8!?uwa_$5|>nLd7)EYb(XrOX1N{T^n^ z)^*1R^AlD0LwY_YLbnRjtQcJ4suN9hV-!BoH8%;+>v2&2+epFp(wS4?{-#7foL~xa zEkIEz<(NPkb~tCmY;|Chu6jZKTEdA^2d{G z=I;2Vqz_Z4`tbETbg<^Y6Bt`vH~;CIG@3>;_fWpF#XIh?p3(8_YRB;%J`*(3=63ZL%{%;xUvUNoiFLi)lF{(bGuDI$MZ}7!(Iy?eFZz^TaPJ~!q?;bpAuDC zyBI%!3q*1v_kaS(*#~RJWEeKCoWQ+iwuSf27%f{)&rH%d$V566a(%T5UrRD zD?cxHOBX*0OJTLq*?=WA0ztKbP=0$q{cJ;jbD(L_VW?L;p5P5`wId8JEb+p)nTAKk z)CYESIxJRNvlEru4gC62G2gyxGy$esn^~}xa>Q>ABj>zq!EqUoH)*_|;<0OE)v2~V z>}t|a>euj4{g|1U8q(fia@;}@V5~oY?Oquq^XDf153c*54Ja5cPp= zY+fwt8b4Q0-fPbGGOtL|tdpcG*nVs2csu~KB%#s|w1iCcR+&(V-aDz5CARAav*myd zxCr(c$DwqT`}gUbHPPc)FvxKdW5nj7h1JOW$9-E?LCP;msYPhxB`F%V@CM~!i*e?zpgC8$nQ%o zshcqvZm;>O)Zy`U5^(cXv~6XuteUl&9KeBBvt~@a%B!lhe#>vcnEH39*ZY9=C0t+j zqk(&jLn0QSUyqh3<#wZC06gYQ(_%w5Lh>*=4(5nrV2-Hs(GlUp?BJ|CIChOPfFtcr zb{d*sBei$3%wo01M# zKoH!6Fd+!RrHD!h1H)t#c;s^Rc@T%`iOshIQ`oAy`7zP>AJopkVTYf@ty%lG%{9|r zIa%cCi+Nw5DenL0YJzF%N07ZI*rj;lG6HDmm4uV+w<63q21>A%B>f@TN%hI`tO~dT zwj7@jI{xG#FcmH$5Kf9n0r?LGAP5LMtu|*I$2w_-C~)ZtjSn{d z=Wi+=t2GHeRsbW9h0|cc=z<33@S%zE+a48d%s(HdRqBP`nJ6Xdf$t(n9kYTpGMZid z_3Q_Mn@NeotvpfAojC}w!#M$8l?>9u^3|)jRs5RrGyw}E)od|<6Vo*r2zKyXyBg+T zRF#gR=pTI(yii(@i}LP13RNd{$Xkmbs|{S18x8E#4-@bju~irjeA6HpNIA|g8WPur zp<4iVqkRc3Q;rLZB+!|za$ix`CVS<;;H`u$8?JedWxU6ER3ANz1F1B(dWhx5>z9SS z;@l{_o+excQj+Zb)M}B0P_H6c9Gs9PX?#om4NfSM)^*r^7!EY;kXJ*{+9ekuMw(EDwb2(y0th2E-t^HUE4Ia+%(cfe@o_Fk0_5FQ+ZEo_uMp= zrX#SFGV(&jamuZ%?&$4dYGc^<ilJgVTMN!cK&!f7B zks$;wBdoXm01)1b3@ItOpVV|N|j7X^)1jl ztH--*tYB8N^zdK>>ms6OY*w@DRML2xnpYB}h7CF4_BJV7r$UhFD31m+A5<$D7#TcrOAz}Wj13KULhOs5HRh8Hky8O=sL}B z9H>%A;6x_aZ+-&i!=Ow@kuCzdjVG~6R?n17zGWG@sP1nOWz{(&hQy++55gch>JB6{ zEr6sa8WfNb1+uAhM&38&%NJu3aCIjWGd@Wwf8kp{0kS$!2R*F5xh?2h_NI3C35&I) zDRmfxPu4=~jv}AZh&?4oxADy_*Mm^Lu8M(v+hzS@uhjQxzUM1O&9anjwM@Rvf*1l>tmB(PE{zc&Ntvwak?*7t`#I^y}#6FQQWSVy(oel1l5Js0HusK0xRnU~=33lO|4qC=p zabveW%gjlkds#9r?ipixnKLf(^U@3|z5fHE4=ctmI6tWQ2{Pb+Py?(RYorcxMtjOY zd;yX}wLo%cdbJHW9?rdGz&{^?h^Dd{&=ycLy@b=MgKYt^dE3E`zX8`QTwLw}p>*8L zs5XJ^D9qIXNB3p78FsYqg^8Y@C^BLBRgmHMqB`;ObZj3h<%+3NWaVG75yp3=Wwllyh~UK_FdhY64pSA681=jw*-ovgpQJk6j) z?`7P>`XS$b=7aTELDySGkw>YXS;$MquFH;lIxv1jT?0UtoAw!nV{F$zaADQz5MF<< zfyp^5?@wAgB>_#V&a$tu3f66p1HLw0qtwytydw%A?&WS1tn=>ep!4lVe*oM2?yduv z+P>t(zE`SqagV*40lm_yvw>;g28L+aZp^c@_z@K;l%kkF7Ytq)6(_Bx}p__qFM#fx{kBJZTcL%`^%w;|x{6XUF9iUjTees!Y03kzyp?&~JntiwVM(h1_<;0Y8znyz>-Be?WuuQ- z#-u=^$Q&3I?RgJN?v;2*0>=ap#=xu2h1dv#PG<_hx~oQcb+!!hrN$OkS%B53Lh*C^EhFy7p6o;w>Y?67u*?#`UKz-KhV}73V&QW4 zcyULZ0?!Wn zr=+bd0JREwzRLE_SKWM#G{38t-$%VMQ+IyTym(7qu?Q&z9#LC6##TdGD0*I zVqB6~=fHqSa1%+O#{s{mK%KT^pJ1FDk66t5oGg)`U;MZx6aUEV(pLuvFN_KC_x0N<=l`V ze{5#Mw<#v;OU*Xb2lT-!H+$nIF_w9Hff!uTN;tYQL#*EW{V1f7lj7=dF3QvPuE+3Rxac;>_C83*Y zoQ-soR%DbKKJf9IyUeC`nQ6(EglkTc8cJ=2@ku0Qdr#62{L|+uY2Wq?*A{g~8iWW3 zpANdbFnGBoCCoX9ZPj1$@zkJ}^V`73X!a@}Qws+4)tV1B6f+d<_Xp9OgUVK-Uq??C zt0x^@O@5{RNGYFIGf*Q{vLmW<@c zxS>LB#ZK6?)pL7#e)B`AWnu^I!#LvkMc{{3&f(;%^4c?UWfE=;YSRlQOZOi zXkK+R{3DL|#1K}hPMZ^&z1GhEk2FGzA6070CbZ8HdhP)&M>;#Rx)v|0bA`NaY*?srv%k(c(rj#8WQ=WwcKyW z68MjfNzGeAgEgB;VWM2K!bCZf1){`kC&_5PC)c@1>zs|TLs2?2Xo7C>QF9|`jxnuk z)3U!eexiq-rvf=YRdPl>J}dDWVI-77{MAyq=+L{P{IEfCE;NNTP|0pwA zmdY!`jGX_jmZxM8^w&x|4GE^Sa+o>MG$DRe8W&8`TJ}fP$pd*qRMCJr`f^qZOowa3 zD7wN^Z?20i#i!n>*&Ijt&+)<=*zg36ncL%JwE1|@-I4P@24bEm(L&m*#TGCWbl7yi ziU&g3ySLwS;}Km)HgTt_HNDGP6*8_%j8}t9@0Xd%po&P}rrmhopYW(3Wewyg_vP)vuNfvh82pGyP&m``Z1qd-R6 zYnh%Rp$gjFRgEY9Y9p|7my{b)Uuu4Wpg9%u%}tgTGF5%giJ`E8xik7vMw`m~&VVGM zM7~s-13@;dH~5H*DB=^GTo}xPUe~bU>4c|Wfz{#`ks(b7Q`r?0@E>hMIYf`ngzfQS z6ecQvE0_WZXymZQTv=IfH?ta$tDnhgvPmQI`>~Av@#tr^W*3RMPsLIJLbWpB31Y_F zm&<4?$J}Q>3U%`+6b+bH2nGJEpg3d*nC_&8W_G-kHBL;OK3g)nbBw6Q z;U=^i{$E4B@%H9lvMKqm5&1N8xvVg~5ess6d+YQPzk@DGMG8;vZEP&xctO z#Ph?uD|)~=H8lTGby231=H_>7k)%4|B16uB(3R@@QcMMyETEG>mPeK7``3Ya2c!j4 z7ASCz7>!SY8fz{@Mtch_1L%)oI>!Q5dno5WDMTNo$RT|c2{?5i((C^e_@6dklfdLt zg~_KzL?9J5WSn9DN@FtYqg(nByv?Q*!wY^_ar91qc7|6IZHF_+`CkH8TEVnL32UXp z6U4@PsFj%%;JN*Qpb3hX&=u4W>43TjIvOik@N{VONzZ|95cM=Q-+eEc>e3++fAuZ0 z^ZlwKb{ewnE7}sgDmcXmb_Gru{Co_2OpOqzTo(cn-T&T>QO?Oq1Tzu^HY0n(su*gg zVg#cSHpAG3KGqfpYda6){rRKOtz=#d&e`z)Y(Q>Oix#=NiBt9W4O2hx6Oo#lH&_{L z4{I35M{a~!6${HdXE3Mg48C#>n_se)p?)YFxt=DKJ&BaQX<*RqNbn!15;r` zUQ}#X!=gsF;B%+6Zwp4LALMo54YKqXL9_D2-(Hw|IT&m|)iE}XwCb$3hYsr-yGCw= zT6yJhhM#&(KOJnWelnK&HFcx7dOescPRP}&y5F9fL!8|)t&{lU5oc_Iq`fSE9VUJ? zn9aeSa=KuUi0BUf*`ivAzzNo2?#b2SLjFWpq3-aJoV`E9n`X&XQg*;fn?9R1eVHeQ zC>Z9Aql5Uprs~)Am)bNAXt```)gQStxcNDK+q*yWgA6dosc)Lg-X~CEcPO+OG2L9& zd@7@ZB-k^KIz?-10WlRhki5a=(>Zc+}ioY;^* z_c|kq$9&XgB5%u!S$PEo_k zslX#{u(v-0@f5qi(0dt$p{{16b=$ zptJxy9SffR|J(y&ZqAsbFD0il>90iRo*D@@wh4#Y@5`)O(sTru$uV{Tcz8T&!f^RpL??WRjL zBA#1=p+#(LyLMj>Ye?7wIdoH&KP|~51UfjyH9f2UJkk%j#_l*XMRfVIy7+Mhq7p!^2v ztAYA|H3Vfq`^{l}OB1gx)|iQ_opJM4mfk!&3jiYg#!PHIY3)&Z&2DLdCf6yQ#+2rC zUQ}*<-jj0Rm~WoaT!D7mYIctJ5-#4E2fTFkw|5Oj((8C0r){#(buGO`E>%a;MSD`# z{J+$C9PdaWzQ04~+uQ@k%}F?yrcg4ZY|o0n*^^3vuXKC=29ANRZ_lUy^J zC^;=nqj@fk4l_mUTo=5ms7fG&H%j7z50gKR!RGP3dQei6xCSn!vBXw0MvY}1n_>9& z!d2n>&dFDeHW0Rr^wxh`Ltn=-9Yo_X^_dYr1VSTDwwnk^XjNQ zDL6fqNqjG_$*p}57eDT2JK>cZ(|=O@8sa<5hxbyD`Xrt7<%dmnOVNpYWvxEfX@0#o zf0Xh64Cy=<+0)U;**E(kBkf#EevIV|G)Wp@Ro*f2_ZbhSTvNG)8EjNRyG6~ z`{&8NI@v;rq}(;`3@YZAYF0v#e@nW zOd-MMt{eDSYvx4h@S(#l-bBCQ(G&N;VOf2@h3Kz8Fzs>W?`t(uMX6tTzHxnG`{7Nd z4vs0Kh6t3od)%6zw8HCzFL)1QJz%t;f0Sa$z+fhkuCNmcipw@@Fx2?4$A0XZKu-6x zJ%)+PRn)-nWc}6z& zNH=dDS&+K$*9R@yZFHKl9R$r$^ov|EIKiIk00C$ag~gk3S55a8sk#1lJ0tZ%_3wOJ z;D@b#ZxUFu2=I@@khe+L^E0&)SoiiF*S{O6{^Dc|=Yk0-M4iE#Y0AOLO;Y?la%{sg z$|Eec!&81LLR+g`mJ(SJu4RgFChXou(u%J=9U36SL211$R46ov&95_4JJ-9nw}w;Q z>3UOdO%=8t{%aTEpC2~FN5+*Y#zL_Pm5(lgZ_X$}%K3wAa&9K_u<=VF-p<$mx`GkS zNe#T4$E<{P^}&D>;u%CY!C;d^|A=Ha>Ar+EUFAuwgMux zqpWn6jgk))-4KAsIEjMcl)oGTDduEP!qMo&#|VS0y}ry9q^H0lUhybt663f43Pg$! z2+<^(Ii;yc1a_7@=;05t6gNYG-$XThb6r?CtFkpuc{70-K2r>KDuru7gc$GY{$i0)gR!ot!>ku^=P%_OlPp5y`Sa>;2- zGSjly+#DK;5y#bteRR+jD_j83O{w9WmmuV6ETA^{CA`Nl0M<-&Kxq(#-63^}jRsAr<&mfnu1rJ zAcMWa6IULs?3Ox5dZ^pNs(~igwX*9q;!geWHPmMOdO+#L1FALT z>HgatGb01;f%&GMw_i@;qwjhW_7JSzf8PGcD;QsP&x~LCAaAgX9;FQb`ImFrHwT~i z*e3&D5@LPP_)XGA{nVU%&_Oq+4yt6Y;|{6&hEn>^>1NfQP~PEDRsEYpfC8BE4xw67EaLA%?V{n)Rq}HVI%h} zSY~pG&O{b`(8P0Pu3@@n7$>_cdOIR;Sq|c=CRs3H3P5O7lm>PnMN!1`9fGE*tqDQI zb+$Ox6f0EoS3Man#MKvgg3k6LNK-aM`Z$G=(H3Z^%6ufB?{1BRX1&tB-QE6H zqn-5yux4;1vwm5ax}eg+I41w}LtX3b?Yw+cRuF}r2#bbE5Jh+9v|Au_$G{L^8WSX; z+(s0^o(?wS?&RwQ4YLJDw7fc9{Na+e>x-2DwV}zc;={9wnZN>}GaP$|wd?x3?#PbY zW6T0QFk3?sNWo-O^_l z%hzEg;4j~U;CE4=4;Dc`?CrnE37yjl+|J}RA_1d!jzRxGisCp||1#A6kF4#be_+WW zrF_=yCuAd}GygY-sj-V745I~M*fS_QET|fhTYuvV_ zsx0#d)&y)6x5S?WlZ~<3ZVp)X(jJp;IL@W3gxas;QA{Oa<|wWXmXW?LUKGRX?-jI2 zc~@ihK+m$gmTeT?eD_LaH86gfW<$j<>tD9QLU{lM>95v4vxflJ(+&Sq;}vZh+1MAW z-2Qlo@^l%(elK$GokeD^9dYS*@-wY#c+^`$UkShqJFAN_G1*$GUtu8`<@QUxi*nD4 zS~6y1yQ||Fg61)XGdnAfF>8-S5M;YeB3&HKK9mKrbmLeG#pm8PVp68gNrgQ*L{Q?v z*!}@x-c*)wc^grvGk#CQa#O8n=xD?iT91_{bqEA)vcBbCU;M^}bPVG`F_>r4%Jr$SYCa|M(_i8Q#^Fr^9cYGn?I3H(;%E(A1d zbKpR_k6bx&{(cGmF*Ucu=Ub0ZTQXioOGJ479=X0RGLvfsCeVEuZ0;W#3&9lN&v&Cb zbuP4v=VNUY>W+SPJ~I@B4iAK73?$-G;-(}vrHP&w8uXphnl5wDdE+)ieEGQ`*L{`-1890HcPIt#BmiUjb~;`Qt$ao zhkN(d7pL+h*AG+Tsw<^jgiOSQX>`!Se}(omd$tvk2%ge4*WndY+vQKjUDtkN6#H$tVdrf`dJf-ZrZ zluk(vw+Op}&u{qBCB4kbLE|feU2Ld2$8jVFF>~WJTmgm5jI%cT3ubt_VdGHy z@2_&ILbS~s)&+sKJ6MhBD(|Z(XCtM@#ir*=?bD0v$2YKzZ+7G=r(W$Wem9uo(P}R1 zJd?8BWXJRuIITwi)G%;NgPBkKYe=WAZ9jk_`hiGFG;XMwZcyrWt+dwU{d-UaW6sa> zB7SeO^Hbxow8N!i>0T%I=j+07)}so4qX#nDApVOmdvlMbo*rZrJ4=I~d1l7CV{RwT zT$H8(8(f{&{kd&E=I{a{mzyFVbLF4Ht+(zNZM&aZq0T-5Y!Z;IE{I3I- z9o<9Xa00F=7L{4UjLsijCzcS%ni*X?c~;x?`B-AO?{*@7q1bM=G$W@bLWJc-=li9p zm6j`p+V&rcX^&IHosNE4TXm|IZRj+Sm7TH#aXh#(uG`R_v>GC~AF3;nUD5SPU?Hla z3@qh-w7`($3bHN-0gaYN(K&oYFl=HU`K>}QS;&Tnp%p*PILNZX`)!@DI(C3%`=ZzO zF6i1I{Odc72T2OfU?7#4f_SddeG3c9F4)vDq8}wZrKC!oP47*%ELTlEvRiesb|L%9 zD)yvPWJ*+{fR$Iq0wIGdaD%f>G*6;IprrzbndR?d{fL3`U&`DJ?;7MfoKFpW20bIF z7do%YW8ePnis9>zIdU5JfFJ&I3a`KU2AJ2RWbL&T%SCzz>F&;ee|rduxEjPwC08|i z-2N0_Lw`c=e=K(#jLCBKlH*}(JPbygp&m|6M~#c{@aNv2fYd`M@FVWDLn2>1lXLT% zt6s5_4T?b@*0d{cosO6|uDb&NNWn~%JLZLQDZOD(9N8K|Xo;(j>g5af%0ESRAhMpO<<(OXg`9ozM{GdM`3x&KD7G~KV_xNe^lSYgX68z2f>1dW<0&kc$>~CR;gd?osTj$lhLn`HzDP$7P$q))+TWrS<9G@P$?_;0h#IeZvwgK zJhl!u>{c({)#q>P>a_e4&Z>o%n$$u$yxRAyxWAi3H*@yY97#oB(5WWk{$J>xRNsms zLTE7LmBbrMSG;K-+WvFA^?>-i2Hb6NerlC-2C5zyLNn|AU$!|Dh)F*B!ed-xfspt>I28hm&ugr&H1GtTnDE zS3UU~@F$SsU%OzJ;L;U9z7qdu{}_v*?8uy^&$f@W?+E??T~P z5Fl8#_$4|T3=s~VE3{nBI3RM_^pySqkdj@5K?dO2Uq)N&nG2~2@$O)*#PpIc&*b)W zFNy*z!6ict_pir$Gd4hy18v{1s{nVJAvY{d7T_i?Hh4Y|x@H8y(2=4{3Ak+qyDf^k zP61E~It-=Y1{`jnhCQOA=k@*q|Np8tAg!ZJ>IWxz4&Grw;TvvG=sfmiWB(MDP-2b* zXmsnC%RuJ|xJBpSn-&%g`vrdE!=RmN**bQchioimn6k#_Ac`6%^eI6oyjMU8Wko*^ ztvW`Kk@pcc@!As+^;AD9890TdIGaWnhsffNbZtz! z9zDGTTtTUk=9_ov7}8RenYmIP{Rc9eZz^#1itRnw$0xEqUtL&z|N6F5i$-!UnuR#| z9rRlEr}5t3>DhEZ`~50DI+FeQW|kaJ^7y4iWqgMws^Ix@HVjVy$n+XR;i;M-e@9rF zk-l^Qg~e5F;I!uM`+Lgtl8GF&{%!vdQqCpk{TgxBHQ5@%{|lI^T8WIA#O2t;f_qQ( zKC%|ls)L{LndzJZaC8|L%jf237F&=jTgr zf={Vdm=gRb31&%uF~?zt|8OPG*|%t8x%b}}cS8dv`CCHKi`EuR_BBmpKf=jNByKuZ z%3P%nXrsx%q_v9{JvFHg;Uc$uNTil1P z3h2s&f^Qg)JQu1SIwk9k;Lk?0`X}E-qizeFri)+q3txYniJ{*XGYsL+0RjAX}xw-KsUAI~-l90}4vKMW?}{0HT8 z_600R9nO`~GRZW7W%s!$6I5}~6&jkPy^=e=nWdalj=1H1QdV`84CrOM4t|r9%a*-o zO{;U#yeOH$*O<|QLYLNiKmspX><3DXqm4;mA@*fAD(I$XXVpXPp8zZ*9bN*z?eQ_N z8Re|i>3dI)8oT8hyjxkhQlN)=452)W8 zerCij0O&*)66W86cOxd(w+N6EwM%U5PwhUG3Ki@Jt2 zeLXx0VGjh_qR@v1xk?{j_YE#n z186o}Z;G#sGI;ZVu1zG zmj-84CbNHD&Nb#0>@GWo5MoLNQ(4U%Y<0YJ8=%g6A=IsPilozoEz^I$gz;Gsn@Rcq zcypaYf_Ut#Wj1K*Ad)+3@KNS!|ACyf@$uxFX}GhrTeF!r8;(i|O978$sCkZHCd=V~ z@3ei6Y@F}(6I`~YN-uPQ=1N~=(WWjHm@D6Pu{j}`{zEi+PPHB&o!1aUdpQ!Lb~LJS z@i6?$3(O*;bhlQv>{}d5B^A`=H)N9sRV}Fx4|HktFT?%ZV?orJfh3r{_uiA>W0oX)#lUZ|9o@x8p&aHE2%LhqKgKx zY%FcvN@(zGyXfU#0kdyLd5$~ZZR;>g9~N|g2x3n^?ZIa@yaiIZPd zP|O4|+L>IXHZDMZ))MZaCu9%fQFTTK!(2C=Ob;M78DPXFA7URpAU0u{N2YP`qNsvR zS%BRe7<@{r4!pZJ%)67{pwg9Sfh%;_l>wY3@#JAO63oY9%e<3~6%;wc?R2t{#a3CP zLGHtZCjQjMe$!5lS5p5R3r#0&m5Ie@3EiR+u$~nE#V08~IG4oGs=XC_Se_EeHRO$1 za4w1^LJ7GBA|I^lR49}IyTKHub%2lD*dR!t)Uas8eQ*{j&;rbAH5F|p*j6n8ogWyX z9{NWHbL6nC!be9wb*J7@d54=6EQwNk9A%Ac#`j1IHsu?{r>g<@ILU%E9A$nEaMl;0 zZfRP+X7evng4sk0driKYc(rUTokwRD#)P>_LfEpzbnFhw#0pkg;cJ;8e-$+>e29Vk zv5qfgr6w;$E`%SnUD9?tq%rApey*p{^1qb#W~a=KI{i+cJ?Trn_jdkpQd#}m>NDgG z5}Pxu?TfEvRlG<)za(Ib2}()X>IHm;A7eNWK+z&QVuw~ax+0eV{_vc*>;j|;F_)3Q z=^Vr_r37+H+2*_?yiA^CZA9>E>N>purxg>lq*Xu_){MTMW9%e`nxr2->tZDA8G@1jz?qttj z`-B#TU3p8upo*hw6882Gh4g&A5fgiH>;G^BqNr{z4C+6tB3T+8RWiVnOG>#07 zt59fW^69HmJ!@Gv(+~*4B&*_gYQl(PxgD!OC`T%RHfHYhdxrjfnONMW)m98qA@ptG zh#<+U8QG0!%q#xBTd6jG!b!wjYNZxWviC4L;{WNuXU6=^1)}a>7q_YpmcsgB z{vDm(E@56((H3u$A@_djn^~{SoqFg;8g!ZBMhAFD-2IK&@2;wLoS0C4bRV3@5({d%2ht0( ze&HZy%!G*{nBIxl8vv3*TmPC&S(Y9tzr=z~W9C(L! z?KT#~u+C)P_Wn<2o;1O=z;CxQmsUBW=J_M@>o-DH<@Anh%_Mr^C`B*f)1hxS|X2MLFX+?k4=l$8%JVc#A$MvV;m2uy6j;`?O_eGvi2%0Cxp(PiD|03*h5&^{X4s58x zLw9zvC}J<2D|kS`IlU44x{p5#+~$uK>X)4q{KMJ*d9>Hu?u|i%k34le?Qw1TtusA? zKTZa3G%hd>*EH@ZzZa;)&b{CEkrI$!`SoEPF#TT^{4ns=n4X~(D`UkMw}b)alhtXD zDZlpTF!LR!C6F)Xf>mleeP1niiE!u>?_tsH0q zBV*+bu7nWQ#3=1?QKFAzVq$C1(Ba}6djF!K`dt6E+B?Kg&I>tt_GInMg*YlMY5VB; zk&Yo)vHq^|&(JBZGBe>~l!*hhG%Vmiep#Vv9lZba!JWIj9=xE$g`6M)#HZ&%UJwiN zC&45%DwucsHZ=&E*-43!zpoYx&j$~|O0)39S9$6FFoEZ`lN4x;1L)e%t|Ovn! zKTA8A^ry|-#7Sn`8SyjdAJDf2z#*bwq4CiEcOvmjy-wo3W}aQEs7;a0X-e#uClQ&` zvP?xeA~Ki8vP>W%a=9Eh;1G#uEM_JvbM>N|&FtlvT=bOoSG#V0g%G@|FEF=Gd9_ zO5N>IiKc!|BGzQ`3@Un5KY5YprhbXSJw#}kfoZ_(yO`crT@FXY46l$Dlw!|&Z9FfY zpYpGTOo#RMfxk%Qk?X*a~ys*#BB^GKB4iopE0b|+N#tHKIDHYijQ9gW87E@?;XhYrZu!8ZsB;*U)tPFC>I7JRCi?By?em%6s6s!d74m5T^`HL^*YY38BqOf9Sn6hm%qg~$ zZ;%t#;?71W4!5Xsd!Md0@eN1Kb2~+=o$B&6ds%(pG$dorArFv#1$iH$zDt+`tt;pH z&Sr?_9?jXRY9?J(k~4S5GKuXNLBi$bCpCzxy(K@^F@>Jb!i#CoH2D2arD8zjsunr~$t=B{%iyyZ&QN%YLqPn&+q6KYX5r!b!bK-AiEljc2Tz zx&Z!vVX37Np?b=OTi-{MoeH+Wmwox% z^W6&CKjcz58%M3IjK1?->l=fw(@|mTo;uW}o3aiX_kFzce4}Mx8p7!#%uyJQF_&t) z!hV;cz|gu^i;=YOHgZ#uqu}7E`@M4hOn&I|3YG8dhV^7}r>lpYlZ+(EZYU6fGRyc^ z8cDWH-^MJiD%lLl{u)67Btt9tuAB5!Q3nVxuW2;5CcFffkjDo8gjlgx9^4IQKQVrF zAD*AmfSkGY)at|gg$;dP{ZG(e0#7$%NY=h9cA2u9c8bhGs3b8J6u?P zA(Z^Oe_^)mx=&x#qwo%&<0*3pNz%l-%|ZWo)64HkR%0v!<{iU|4JKp*Pno}xOzfF% zzBnKIIAcKKo6pJoC+3dHu7HHlDw?p*Z!=oz8{tGIXLJiOkyfgTIk2L#^h`qc3v;_O22nvNs<(!i=7P|YUm9!j4C5%4o>trq(LvhBq+1~u(tg^ zP2t-uL36j|m^q1YO0eDI3q(8y>ubsYM$&nGnz8m4W+R z?RP`6kof=lFqlD2-017sZKz86)LQo3$TTeKj>~?P_Mk5=a()R7XEDg?rqJm*OpL>f z=zH(HUHhddu`Z4|d-3AX6q17G?ms-%1D{YH5`kypOi4)BUB z^|1WFVuu!9Qi6M_$E?L|djPck4qOfSN5Bg;Ox^ZH&-&s9M;M=FUFTDpwQ<80=^}i@ zy%Z`Y@6Ms)Na65Jg?J*mEM6FH0U(3$3)n`bTbi{`euFzh*B>ta}+ITF~UItH)o z5<;Vhc|;y=+Wd)q@t54wZg z22r!(!PZvkZ~GJEsF|S$_Ih8Vsx5g&9IG3Sweq~0bf;0KFBk$!4hT9E=2lX(Yqlz| zva)yTUMSg;k7`#NIhRazV3Q5agc7Wk8rX;QDy#HvVYICTUFson2-dn5fn)${Y3V98 z(&x;5M5#{N^GESrXQvu!Hq`t9snnfrqIFI5bfKvpZrTm9w51uUskg@I+RVP`S?v7z zOfVn%Hu8L`q#b4139K0-l8p{^atnsah^eshq*{ zDjy-zo;biK0r@xI@|yiv3e9Gc%Y-8DeXuKOp-BLT(84@v9no#f94U;&!6<+N8Yed`LQlBz z{X&)acv#I|?OqPIOLQ0A9B*R`LbXd!qXa`0oHkb_v|rTOh@L6+(oaVIVfTk$(Bs{6 zm6=^etmf7;_yEp+7#62V|(#eop9(cWK8c1(oV+&N7_)MfsiUenzlBwDW1gvUzuji!xK@9Mia$YFNx zo(9IB9wz40lD<(&3Ef2o=Or>?w8Zgr7)Dz<*yL)1xzTygsvL~CHDji=u2{oBq7PDo z4N@*TK1;h}DNXpg5uZUh`yMjoYm zZayv{t&NhxRj4JVyW1%9UWT`M8ZVP0IjI0~FE$A0V%`vk#$Cg*E0=8615{Bi)7sRN z8a}_M`WdB2ZWc-A@cT3HSZ>YkR3B4)FBFX^5?*))oQz8)zOU6Wee7<;DDx8=4fIFS zQ`co9187Zu&HJ67u5?TOOHcq8#Y zS4MJoRm->ln<30eH8R~std01#FX9d;ylQj*giD&orhF9pERdblew2URp&elGe1`9@ zDd~S2>NKM#emKD_2w!u-(Su9?nh2Dm9TJ>`MkkdBVk6D{BpR%lY$>`vebOxAGY@eD z-tw)c5^oU`xy1L7W9KR@z*V_f97?vyqmy5 z7mVYcC48Inok~9Njh30j*%ak;{qkARaeZ@|X?cnZIgXWSn*KtVL!YmB9xA*buV zV6U~qXU?nRl6a`~r>^XcpI@k07wz4bJ!P`^!NN?fp$1!a&SQSyUJGOOZ^(;4{u#N5UG&v zP2ah%FUkhrNg!;>cIZ=h;R1ISV<9<|>rz?c*#s7Uh*DQwKV5P+qOp;k+QM$s*)iiV z`6dh6`Sax_#=;mi#PNxkztPc(JMbe^3mUy;(Vdz+;sTer@NmA zXfF57zJj_B8!MX>LsGC;1NxQ%{ocG*it|nALZTh@y4<_CCA?Zzob)mlr`gs5^gUAw z9-L^AjvBDBX(54M?`l=PU)RI^Bk&&l-S|O)rwMVR`W7RVcGB`LIjhbJ!0^I{4YUvq zbC%8vR)~pNZJUo>v2knz^ZT=QU{oOk7wE4y=t}7RNcVgEt-w-mX*kIM#ZZ*|D)_|}_;vinXNwJL7^`(Kg6CnC6 zP}M3qa2vkKSIcZwL$kmow12Lg@{}|Wzq&Q&MPdC0=T*Bk{P(tQn@aa=XPB%+Km-(S zz=b+~8#UmNyA732v~D7dc?I|DI))!forRrUU59q+zBh%n3g~AHdL3It`-eu`<__&5 zkN_$)Gx0zfV!6xP%ioGOm_YUgAM1nKdF*uJX}&z>aDT5nI+8-AX&9-aJTC>oGg$!8 z0d=d1mNXWlt9N1`*MO;ZL;7&$(i9pSoqV+jvvy#1%KnUpp?aA6ED+Yhi=fVehbLgO zJ>Y+Rn~zDfy9lwq#(Nm1{?uwvm55^g`^W!_UX4U*XFN>>3%~>7To?{3zXYyhv*1ak zl8hxbFhOLo)(|3S{0)YDDLY~f+fabbFlK-f-8=C25Xib01wfD->T&RxdnLK*uD>fh zD}l^X7e6$(tS@5Ln&6l=-jEu!ii$UdV6aH;NOqGuZNd2g()GB@POWuSn|T*$^k0 zw>4XDIndsBl$}RjwfCX8lcZUzH8{Q0OQ%}e3l7Cg}<$K%hq9U**1eHE+af5Ho;c zXA7C>F>{^kFKP_>l3b*k=$=ivSlfjEWfUNMDy}vpyMo_3=V*t(XL;#C}AN!v^l zV`*uJ)R3Q&5f-Djf;;)eMZ~{vTQ9m+MOSX!i@}q&$gpgQ%W?&WCE5=8<4B*~^HwO3)AO7AjlEifDi&MLVa!@6bH{JNF1s#RAD!lW`8jCY{xJ=RGLT{+ zc`iaUL%~QOk796B#hI2Om{OwF{uVl#U@YY~fU+u-1gPozocT)))zb_6I^3!kWjcB@ zVJmbg-aJ&FIuwXC2&~J4(}2SCCqVgpEp+w`^pe2*So6g-G<7rexY4?Rvca4nlbTve zW_(jZoW<8Bp*WO{#df}Lg8%?2vR%o2)|UULOq%8C=VxXLg$6q3#{o^=L_#ezFO+^f zYzjWbF4vZFrijF(GG|k(?qAhW74T|*sb3GQ8C`KS`u|i+;9-uqs6C~e#PraW)GZK9 zUtUyfrnDktX&ydEHfu^)^}$OZddQ((tD~v<{2p8RhHDX2%H)8yJzNN8Mo8a%iA>4J zj0f|pk-$PmCV*xA^An=4mQ0eyEI7o5aKj<0+;MoJgCLlrV94AvC4*GmgH_ocZ5vku zfX3cf{Z-4#R=0#!3ew?!M|9!$Zt0`BtRh=IKu0lc~sr<34Px-v?S!HCW`#)8) z!j$IgzVpT#s!N~MADRK`(@s+>EbUT+6uKzN66t6CC*)Pt9~YKIeRmn@N0TX7*FrP;lvN zXxx&1tI$#4vlJ0E>hc*J;7XCSE!u19A$s^qf<=h6m9=@QqO{>-{lWY(KabQ>C+iku z{j!?ru5CUmVyFs7D}_&q@dPSBeNohKjozYD{IT&||4&0Bc;}HyNrG(~FaUv|@9wi| z#C&DfXr`uOPNsc?IVe%QT}p!warW+XS(z&5SCci8yp!0Hs3KRr;75)^j5A=ht$d?o z%rp#HJxRNU(j-UQuMeQfjD%$gVoEpX)0t4&{R#ZuYGnEMyw6zUrS9@PRN@FBI|UCNv|0+Gr~j;>O;o#XZ#cA8+74sI}bx=S%{mg8Qd3}?(Ui}RTor}bwg zF#RcKC1-ytEh(#!I--bJ?VOYxwiK$TY+$~p>)~}Uj@5g%bk&>P46e9lD4z8~+@ zc|t&|EVRx?pJO&wG1dc9$M2eQs-^$R*+bARSL#itd4OC0ZH(*Ux1){3bmb;KOZMrg zi-6Em=3qFVvxh`#MGYSecfTh|t_(eIx=M#;t2AY|NQ!mYm%Q~{XRjQ#Lcvaq!L?Y&zy z8dacmLlYrIG1~WCm-W5rcUh+tj$pY1QM>JRU%G00QAbKfh|)aw(Ll)R!T%&2yq>Q5 z7p~C;0;_8nV+(EYjrFypc`l6^z}56=Qe0c}-ccsP0nj@Ht@=gdC(N$mLh!ZgbiZ|S z>X;(;V$bcJ6r{U!t^hZ+HC4y%s_}1IjIlJTA0)+p@G;Wos4YxKt}=XhYq*BTm>41JGA<6gH7$|Z z9Nl~s65CcnP!q3CW~OBBj)z?%lhS0Chl~WvZ$5&DA+s^NxoX$|kYKG?-p|)}v5S>e zx2(*5Y+*&eiHR1Td494yLiqo?%u(0xf0w0yKICrHE`b%S{3fYTFJyZo?tlXc?OS=x z?`rHqBEXsULm_&;@4m$Uckp7&#ImhM<@VSef@%-9A@zPzXt||%sDDB0k7z?|@Ovxj z3`=c5ZFSsS3v>pZc7ZbO>^)ZqRg6T2kG%7=$*VLP?aI91?bWyqvY>?G#q>uuN)P}^ z)McbvT(J>#R2vSq`3g9GM*-LMRu8d!{g^ zwPedOU$5>j=0q~0s7n(5B&cuHJo@>_kri*Zq5uoA~viu7^>@MIO%J~euT{+ z3$?%cwZX;8R~25KUq9g59oZqPIT~341~$Q}RrJukeog!(~YtBzK zB&&R*G5&`-JEqvLzKT7!Hr%YI<$I=505crcLc~v2oKs|}E zUxLv5uo_9Q{t#nVP2lUE0x|&0UI%q^qn{D&U{Leumo>!Yf@<*W^NdoVQJNgli#$1{ zv|p0^Y+@p{4;h&+R*Ae5f{R-3Z?OGZ{KjgHJ+#%5%{xEIzGM_o%=9IUy+UO@-ga~5 zS$o_%kYE;(rG&jK4hJRY#f#iDtjIe^iN3efd#@X~vS~aQfP)Tco;Pn71ME*XpUpHL znh3Ipp}V;h-GUzP%p4|p7f;0;CLO9U9s$LM*9H&;&wtmhkEN59Y5po=>aD*B?i?&Z z=)eAb?hy4lgvbytF*mv-zkr(f?5sE6J(u(yTF8jC-M z=}azbvf3b6Hwn~P@Kup^2>hOvDe3>_HPJL<|Nr&+oIqpy^1R8{_r`@$*le~zflA`U zdbck1=XE5zLnExarO!H9#M`-Wk+TaW2&QB>nlcTPt@QE(zj=9P$Rf`Ab$Qt-EWMydy31ZQ$l;&I-iiKq{@yHU;UIpI7%h*b7OTU& zDHva(^ep&pyJ(1^&+0Ekpe9N$P%78vvoi^}O@lErEp4SgbiOY5fqn(&GLOz7g32j> zWt)RXm@L<@;VHPYbgqcqm8OKBp002=h?rBI0Uz#cU`5E9T;(r$!sIe^_(r*9GV%L} zPmgl}{aNJ`)Mpt`?k0lJW7-qRe9U|IUUs#~rg>RG6v(M1CJaj#Tzzyr6aJ(>wP_QD zR{Z*JV7<=Fm|j48Z3KCj`@cIkM{@PRmJzC?wbR4UdV;^Dl$&n~GKU~LuYxV7)o z)!ZJju&c=sPONoq&3E#=eBble__C}WCPf&IP*nx^zMw-{LKf<&GmkM0!Rj4`6^|EY z2zvjI&o7I4RWC~DVSC0ebl=SZBG^Jha^qAts!Fd?^LgGS5H5h>8aqRGjQ&AN2FR%r zEzXb<3Oj1i=3vagM|AOZI_URmy?lRz4A|nUivY)$OvQoGK&|Gs@Qb$On~N7`{8Cq*l&CLpbmv_OmX~Leb=&A}_1UJj zq9QtOz??b6e4ct^!Av<+o@a$?v~T+|4vx1(xi6*^+~~bTWfI!;R{L!I)0h>CyK5iQ zLc4r)?J+RFml%p$4;C<1^@KGjHN%T9^s>QUs}L~!ooNuaF>OpA;h_m){+KntCaU=QNqx!xt^S|dE*Fe2gfn>GE={2E*zl1OEV?-TjurZFG6_d)=X|qiIIq!L+q-nep^PIQt;#UqgxSEhSW% z%d7SYgRh3hgGvzR+hd@Mu+bd%JvMi!_S37ODLi7$yLidZ-CXmV2j7!XOqh_fw?akE zbc>rDYs(T-wI)nyOj}_M50ttrH^w9sGLBG{>aBOg%|C|!LAxVz-S_~7r`wgxynl|& zS^TL*GgJM+{&@H0c{fhg#BKcTRMdYune90I#NlHE`u(fu*JIaM*lf2`(vFnL|MUns znch(aSWJhydn=Qm#5bZEi(>s3YpHrrwvAeEz!#g;zg1C(ges^4yhY-hq!F zB9zhP}=hz3t&b%kOUJFYf2bFESv`+Wxp}5qr*03K;$hqlrH}Ldyz&)E*>*EIt zC!guJ*IIN(J8xBu1X0M@|G1;{Ao+rnia8z#HOz%7$MkZ`)pRF4X;Tln?|YNia2+8U z5}vvprEBIx=?yCDTnoqx);RJzARXzo2?`zZrTMZ3D&T_IMJA}Z*0Ejyz9DFE& z%FDFUY;@=`QIbjg7&dTE9*2`3;gd}EOY*xAoXDJx;OCAW>@MBpoAzB&hT zIHvHagal4@0(PlB;wt;%BZLX&iYdC9B$YS-+<}A-=6o^;<&wc1BJ*i*D_&6k8oH*T*9R zioci*2rNFWkokz@CI?!k5NAb1e#0*MGRimfUe?zH4)g&{r3=OcGs~N4Wt-^q%yi$z z3gYvTPoZYupZ~@=J_q65DyyNiQQ8p8)|>+;d_LhSsU*=*FvETRCpk1#)LE~zK@xR7 zt977)0DgPR=<%Ax6^Gad3W;|NFc-4vjKIzmW@r549H}OzXBEj{-}nLE%on90Y4~51 zcT21YBc(!$I_v$wdL<-WL!YbhsBO2)B<%&ibZ6eno8j}}*t2gWH&FFNJZtkZT@?6@ z3wL=zXyMddZS_jWN;f#v@X)ixIM`U?r#BnlXiM;8PV@?zqU3Ac3i`ZaMlj?AiI&7B z#GMU-*x^GhAAetJV(snYrkE$MJg70kNsapl*gD15eBy$wGUNEcT_S$7^_anS38?_T z%-=LBU@K5O854hVTgnp4TY@~GV5A957=)ezcLc_&$144J$>c7f5 zVk9^;#;REkwlitc4D|JGW(46kMYvBSyJAn^xn^ZWZkS-|jJKe~DgM=04pB9Z_at3d z4zG6^rBbxph=7B3#Y)l<8o4vYWA7J3Xa&#?bsg3uPpKCL7(dNqphzk=7-|ML0u)7Qmk3ALNx^}AI!+}t zcM$^%2?Iz`J`-x?SCrdrjbUE3hMbIHo~2m$MFtlB6V!s!Ph`&E4kVYKzhtD}IZXrN zln>aWC8>DT^z|i}%x)kM#%|xXb`26zawrivh)Hn+zN*^@$b#UdHlF2Sqwz5suNI&mBtQ(cdaKVh_I3Pc}8Y8RNoipL0 z4z{VmejYxa<_31^aOCJ9TFKL|hf_p5lHLYiqX+x5_n8_!-)5XOX1)bmHC{2EPdhcAJs4x5(-C*?1%Lt z(qgu$j6jYST%Dlwsb@oYOScCF@x64+xeiP%?Lh;x@}ynu9S%cAHavLH~5O}k;rrJY9U zO`TQ-n7U9JU0ntp_LlVyelxO8X$HF6sa`wkLz0Q7C#Z0R|u-uyH;XA2pPU)Ae#8bH&R9qtC zsVC8;6rT}k>DBm61%o$Ivux^QTSxL56X@i9O=%e}Kpn+aYRJ@!-&c)wSC&RFOM;#? z564V|*Ru|i(JxWE^QTx02XSKYwgf`Rg_CT;&D)BnK7uoBigMF!3aM3@Td0;Utje$dSVA-h^L{E*@M@<-p%e;HVbO+3^bh z&?bjwxfqv3*x_?C%YP$s3;$ysc}jR54(GJLKC4A9`hlYT8ueTZVWj}q3CdVw{5IZu_a|(g z?TGPxg_&#|CDkzPw0b|qPiZ2|QUS|;5oeFT0_Z9sF6kxwiv3lK?oV@gqks`T#Qo!d zE1nwHNWn`bq!wmk0){As5c}H)5mk9vyh8W9f1G_Z?3SuYt9Kkef8YzUag}#}ECdeL z1&bS-DSE9e5x>Yp{8geSvAbW2y(j-u{Z{`yQ-VXBk6?Swgd?R@e1Sb+;P$>BI+t+s zJ|U|O0D0kt-F@x_XD*i5UCNlSVXHjsSBb996C)R@&xNSBD=*1wV$Ik@Y}t=&{Kl1S zo8G?1wG?w7ka5N)obT>3+^f3N86)$QuGZu6@&eEtx7A$7@4!r;0!UxX*-a8E(wQ{< zc%O1&=*`?H3^6GCcC#iA(FYosLOMc8=zkSs3@1d{hSYF7#@5Jdrjv(spZ0>q)}FL2etnnYiew zA5{UYqr%L>1W|Mi5mwC0r5tyuQ6Lf4dL?uxIT%l+9iw$03^1Gw#M^ z!urkH_5h38-Niui@IjYC=cDy= zkkn$ew90e7&2>yfFSd$bzUD*mp< zA0tNpxHLO8*JKh@YTamzX+54&u-urIf zMo}7gky5N<90uRBnKaV5u89uZMv)fG0_f~Nzv-(mFFZ>i3W}F@MNZHJ3&tq*O7Mf+ zJoElu$m0TC3 za9y9h@;JB4LpX9N`KW+G9T$g0PT{ z!7C-3>Wey%JRJ-D#C!=-8N4`g#NLRJe9~-L^fA8>GlAbOGZhnBx&mo_H;<4_K=gms zLJC0a{c&PHJz0Rmru(TDW#ndU-1{(3`YGkA`R3exP&XT=EPAoPkxXbYWLApo#m#)g zcg$`a_@k!D|59IvS2JxBZwX8Z_t_2Ed$zzYRNw(0PL?uk8P@%cG+DMRt3`G>h3;#? z@zOnZ_zA0z>GM^(uwP^q57)}%9sISYv_n%+1A_vabT^0SvZdSqVzWlpry_dDE}ArL z&G|+@PDvojmN`Mg1t(r3G2{f>L<+c9(!8B3VQrhw5#bhhxubHy4DW69=Q10V)HX=_ z_(SIJ67_9PxSf5@;*E<;#7#wDhb~&CP16?REJ2SYCCD#+B96Bm zmTM6bBmku_0%ya(XweXmt?`MpFb)b-z$u9|!KD7H0cTNE6dCMLS5Xkb zyUah%dfQvJ+Ux;xzU+50+b-H#K5998;mr1(`DI+_6}36b_XFBv;h&#Gj}FeW%}M5= z_6#-NeMw=vr4=2|6>2Tn@cG$pNT4f@tqj`Oc67fu|JTmSEgn^GIpZpXENiDz<>}G# z7E!vt+ZkmWKX6_iGx@v+x+Ft`n_-yddi4*zeyC^-kM{BV&+x1TZwUHY^ZL3lnfnOX z*ygi5d91(YuFW8Xg$W)q{0<(13WffmV{_uy)%f(*ST`H@Frsi#QT zU{)2R=alUv3W64!F#HnN#&*uUihC(bIn*2_9LPMKZzlmm4&wNM3<$Wo#na#i&eKp2@(f~^*Y-W+?{ip6=GsV!@Eh~au*T;#(o;rY zswOaN3~h9tJY9+S>?@p>63v?cXWFNlNe*?#)d$-bAugB~pXndAEnBxgLFL^yu()Za z{YuFYV!KZ)^F!WOA)P4A=flR8@bFFa#~i5p;naPIxjG#Y`gkkj?}V@4lJjaudbf7U zls(RP)9)&NcG03QPFnMOPmCfkeR-xYJtuhM4h3rUqSB<<60t+3&xMu@CxWRbO75g$b%>|^4hn?8a>=~Hu$>TC|eQudfmB6VGyQsNX*2u zxO^cA!yKslMsShgXje73T524w|GfM(qIioIE0_^SO?1v0{iysjw`7buawwZ`gcLWWw; zbKxMd3#EZNp*_+G{Z(-URn7Cuz{PYlH_!6F%cqw62h*A1MA|{=FuE<6Sp7VFW-9W#}>QZ8xg0nCl$XI<;#jE0D+&SQ6lmF z_n{IB>Diyidhkqt_(hD*W9!DLsHv$rFXT91R;H#_2$GSOj)@#QNCAF^myX0^pjTH_ zRlQ;iUDA7as9#(SAjkM_HANa`YS>!0`0l^d207lgwAGo{+5f}h&G%ZBRQmUpBvkRS zJ3(@tC7{G5L*K?iqvfRJjGg!Mzv}WeXB`(v`s99=468Ox4;QwL3j)zn;V*!AkRy|% zFx+nQ%1$y)HqcV8%4(EHpYmX<(hrN1jCV|l<@idNwC3@|U{HbrAWzwlSEnG&$yZ^^ zHu4a4JzBp1FUZ6PrH`UIXkGKd8>LT!mZMdz^K37~*DeNJ4)>J>*e@6m3C;UuDpySR z2GiMx_$OfHFLKC8>H6fP!UV|IIP3}zrl%aqILztHx0bP&JiML`OS=oB3Z{icosKyd zeYzm9o8UJcmQX+~qd1|Kzi`m@$~`U;_txXb_Y}D==p!MYjH}q;a?H&mJ^?ra@1?;SFTu?w1KYm`O3ebJ4I9V!G=OnYDh%S)kv7_7z6!qb$aO5V+Hb{B z525&w-}FF^r-IYDUfw|Ew~y4rxvMK}%jQ$)=?FzjC@*SSD@*mneuHlotiN8-@!F%% z-Q<(=2BKRT_FG@CI*LDNwg=1M3YAVZ(2jY)pT{wd!DZeOGjdvGQzg&o?F>Xkj0SPr zM95#apT~_|gXUxNn)&dP!wyRPD|6Vi4_LtPS|ReQxTRvpR9lEZwZ~*!B=;IJ5V;e4 z5o@M%Uo@XlN{?o3>S9BRLA`uvyVzeY*-)FDDA#hEVS@c!d#_thMc*A-L;D zI=?5;yKegHW;UqYD5N|Q^kyaTAsSv^wmbPpYq}XZwOSG=#qbfwKp3USKj<@B?a(Nz zxD_PHx#+mWj}uVjq-5N4S#^mo?~|B5@`s3Z3=2mxs2+-UJU5w~c8Ee$oGM5%xCf-m zWzV)#L1ctF<5*WxgveiMInz*N*hX9DKOGv$q9bZC{e9rlLS|>)#T67nMulfDw_-vI z7}-n1&d<<#gWRd*^I=TSiG=FBeF8xMy*s=Yl0+}ReD_)gvxiP5IT zx)vF$Kqri`<0EKiB?f+~_7_W9=%B6m`ww;L%Sg`_2OBngw^JHYwTyV%Y_5UphAi$S zcoe`h=9FnCL!|8LT)I42oMXnfBpJD3fX=|nMCc9L)`WN6vYKVQs{z_opD_pTEA#Gp zroZ?UF8-)R`mj_uFH&{=-KhH=`CX*K-H>)bQLp8M5GX%MyJMY($_oe0?M{O@x)>-j z!P5?uG_AmLY%3Z#yyubR$nYSY;~8FUAhpJ5 zeBWCa2ftG@CU%U_UQ6yGT|H)eDWvBhWc%t)FF-+WN%$!;cR%ALEVBfEQR}egp{ZUk zUrEiz2OUymZs%kaBE(+ILjuE^S{KMaSw)dh<0=9>u=bMR%>F&-O{VVuqTbPj+AsN( zM~y`S!+S7{KAjs_u=Sl&l7}snTdY)?C+?^x4dH5gd^gy7j&jC94U9k;H^|KezeU^= zEQ(FR+4u}4jDL^@914t{&cac5J`>=tG*fVJp(l%iI*+WY0*bzs$=9Uq+Pm(yMrk4c z)K*HMgKMX$PQ_wxDtN0=y_82R&ajn(wkZ@+Vb27mVqpm%A@{19vzY!VXf2^dy2bxDXOAKQeBaX`qc$iz3U27)H|IiUaT@7D_~8^!}Sz z5Bfx&=oiE;^fRl0oAPoskfY@SLq4e6VS9&yp;w>g2b z+LxABle>Ev`+N?ym?O$Sx;7gV+9gGHtebE7_F@QTF)0Yn@y*8WexIaVgDqvO9Mtsr zpt`10-?MnCFzX}Nqr*@avqRh6jAzWB-k)gum?ed2k!O2rf1&J!JMaeTcJg|D-M)>a z?XITEYqU#3EbZs^a%z$U%myxFJ}s2t3ur4k7OREF7Q!x${aw`z^|hymU&TuaQ;+|a z08U2OuT5`h+`zpu@W&#!AEjg)c%%;c`tVXkjQUS9vE~RKfCFLcZf+p@lQZm_RL?R6 zdj^0fh*pZMDeaTU?R0l`wfuHHNP|Hvb$}ZOhAnEU4z7)1tU3RL)IwjeG6P?bf7L-j zZ#?ZB=y7lq{B>Euxlx?-sxmo+4o;syM*%53^{>9cjE_YYl zre7(1dp?EQIlzptXVbx%7lCQuECL_Z&?hhRcgW_q{S5q%aaUPUiIaYROQWs$^PxxX z21TR_123+8?e?s!8PN7&dGLF+C-CHUeM2X9_fb*C!^`72RlY!mbk z;~sd~f8}qnAXB{63ruJQ9@y}jL7Dc(2v2Jf345&O2d#Wq0E3pHitlKVctxEQ+^Q42 zh8+oszQ5NEUIVc{lR{wg*&z_yMio(b9;ywJpPMUK-{ z)MB}r)xc2NBf63u=PLcKyvXa{CP)(lC6Elj4csp69;5?z* zAAWb!>&bpqCDAX1GQ?l_q#gL%eXb{JgZrxOnAtd=MGfptxn^c|my7z7MpXN9k)p}_ za#RI2duOy@o(0Uh)q>*9B6RN5&)R{tKLEdB^)oO4zVqF&9y=;b;8 zjWKKQ=N?`nfZq;Q9BdMxe!$S1T5G9+012X^`6YRfZB zzhX1#*hm)%{CzpJ~ChY3@b8;TA<2@`W2u>dIx&HQZ@ebB`>B{?o{pTCGF zQ&k%F+0}@RY;C3yY|iNoO7L6eBnFYDVpXn6&SADTbuR!=)Qj77al7ZDD6M&5+^Bl~v0pK}^e2t~>HTlS@ zrKssZgh2b6WJ|`Kd{3{al>?5?ha9xZJ|h0=6s)y=vOnW_z%ZHrAjZ2J6lND+AYuTm z$T;!wBSrIW2?vw~^LOBA@Ptg*Io3$wNIqLqE@|euoZ(6fG&dWggOM14YstT9FQ9!y zggFA#YV*p_<*b;a90P|OV-bI(w^Cg_3(kzYnrR2pRu~iji_7WDJhRshE6@0uD6OTm zTy;+6QoRYYiXVW)G%@Cqjl;AsyEej|SitTHl%!HnMV>RUN4(}K2hv#R=vaD9$_dzts zwzV;=Ga+T{VPU_h#FQhh>!m26ygzf=h&4;Y+nCxy10v8cogj(wdpF2(_i!D_c_Bd^ zk!XElMo|%cGVdW;Z!NEs%$kV%A^+)2eVE))OgV{BvHLHQ@08PnKTWAdm-8B4>Gly# zAyor3%zP8Yn2~f-C~kdeWMSKw7-dl>{HWIc;lkE^89KsEMDw$eCP|?9{#FlDJ&Flm zQ_Kw^>NkT&9-n}}73F7_zFH|*o`}+cKdX^4obn0zEe}z+8!Rl97hs8E(Hvl0G!-i! z<)E4VNn$foP`)N(j?4)qFu+@Xo5%hLWfpbwDHeJzXQcA@uLmA_n1R35NbhMoSQkth z@tcHLk=tHZxb7nFE~LI@kQu@(GDiea@ZON(IXq>TTddGs0)dEIWIe&{B-)MfjW*=g z5}LNG@q%U&pY6;cgCSYasaWtspGW@T#xY{PFxEl--y{lZLv#zAnTK_B1bn_c!sITQ)5PJ0DP3-g~4 zJ&788=f;pgPw7MCUZd~|h<13bq|ghDwLTs>4Un0^eYy+DU&~>pM^{zJn*UL~fpopR zS5oe#(Dg&^B@u1JfrB@2TJs-@g*L>qOu+o zpY&o9@FKE`2EK^HmfFO&pe(!ok1(zj3mGL@nGDRIiqlYdGK~i?v_^n9c1Fy{osC6(&m@1DvI&u`lKcR;mV!^i4YQp%9A5 zDiyHTK)KX17al*#-}zN}H0^szdC^DO;-6z`ZfPW=KZUmc)9p<@MIp-_jAq-@F)&dM zZI_4#^9}3)!HrHjNVyA@1iQ|?_QYbW>4>PFp_Hk_yEL6y=2cBau}%X4eFcm!fvRzv^ zVN1CnRYNZD2`hiYS2nVcz5J*h%Y3mV=@YmDofOLE!R4#93Xq8ILs%n_B`Hw%8WdN9 zK~{r@G2K0kM4y5GLjeYkX;Tlbdp^`;^yRM}%2sl%)fk)}Hh!|#Y~Fg%IzI07;v(Gk zu;C{tf8Cc5^v@&aH0aOGjVv>%WywhU5$WleJ#(^(FEBbZJ+a}c875qmpFB&T^ZjU0 zEVaH9&6AZ{D`v;yr*6)1-_2E`+cw%JqUdrzoW(USFk%*Oc3*wf|o{Eg9G5=I^5(! zy;^fYJ&sO@H6?Q&Z`$J?M?h{2VopsUZiB<)snaP(94&wg_vJVtwHzAa#8{IY%&gWS zs1yqAq*HxC1k#~##fJA}u=^$Hd2Lr7z1+;4Kr)x^vazk|*CV&U!+ECI7N@wPkG?#t zj{gL?+C%y(%R1aH2FW=Mz1Oj$6&3M2-{yYrN(`Tu6G9#><`mS-e(%X?hDRX5LV_gM zgr?Dkm;y+G3l3qwDSa93*1yd*6?;aDNawB*AE?RowdG%ZJaEs6-g;;Y(KMN8R`+%&s zA;yO+HfGCvrXG8%hJ^JPY<0IwFe-7alPFy}vN@-0?42Op=1Mr+K!WEdqdq*;zKrHP zeVBy(Ov&f+Se&|zyI*tg8rVljcepUBQ`o1>^z97-b}mVU*w)j=p18y3PrvF>dDCo4 z$x`~2eRMoj1zc z_=TF%th+$(Y^I6+tsBIa{p+}Fh|awJBHt(0@{j^ScTCUzmfFy~(~q1@T>0RL*2@a7 zfHOWp8|_!j$(Nk_O$`BXb_#_I!&)q?A-?-SfNVMA%bLGR-Gi2vgaHK}G_ajG!-?|3 z25wBIZR_^jof-YZr|#<_Jq1A|Z;utKrE)O^Drd9K%Vw-Qn&GrT=)f)!a zONA7%GKxwQ4tMF9_Y9ek9CJ86q*MDH4(I> z%LOMtMVU#%6z(vZ3iby`&7F#{W*Mbel1`xT-LRK}y8#Q{I!KPODT1OI;h^_sTR*FR*$eEg;MTXds- z^LqwHc$GLdW$b1Hu#s9UTRo&>-y}}pD{dO8cUHVdUDc-<8?`$y;;lU$syz0ojmG^U zy_8B_pM##y1*WyO<_)%p6Q}LJRB;yz>9(`}A7bIp#VWxuU`4KfsGhl$leibL#p{y^ zv)O+J_1B?vq>;sdK8=?1B0OE_Z>LT4oktyy8fP1F2AjQn-GmL?I?zZ9toA8^<{k%^C9nTjzar! zKaA$2?Dc0Fj)|1iqpo|vd02oy^jT#~|5UEo!A(wCi}l&}-Jey&F2HW~q>G+I`df;ujI4GMGsz>z*qw-?-a$2`_|tlRyHS?6Pj75!i6tD$lHWRHBNRMLH}WpGwnBd)b$P#2_Z5(XtP~5 zz=9qQWi_Kjs^AQdn)l34p5$#AEub+)Pw;QF?cV=K*;huz(FIZBE)z7kySoJ$+#$HT zLxA8e0|W`~?(QDkLU0WbAi;xsaA%uucmM3!AG>onJ%lM!MrV&H85Ai8o>m}`L^OKLbDvGf+n8AzoS17H_tRXYl&jKle zu^65V=|`P#JFIMuT4NF6gwKWjC=`Ql44fjae2D`;qa*>UX9hWq(8j}`e-N_=FEySq z^nY#nNYLAm3!K_(!kl-S7okrBlUUhGpne)KqQUDAZ3FiA_>LYr!X6(3B8K4&DT545 zE=X?qFc_QE0yb;Wm}I?CfFKRQXcLKxryaIvF1iXk;*g%MC?~O&nnYk&*-kyiob=)O zc5`BlbCCU%*HMpOEgzH%OzWjGrF#~<_ceh_ZS1s$nQt+jS1&+dMyxyAQlN6ZX^Yo zaBe{|8U@~JkU)x`Gc(08jG*|{v+vlsyh*beSofX>;ihmFxBD|T=}kuZP=idqP>ItQ z#6k6=X@A}LiUh{Dk&<1VTB%mV9~aqpcq*oCAC3ymA#;A-Au{-YPawt``z)ZaI{P_| zTby-VbmO6}XKSmXZ$JrarPs1h`~#~FHh%4O#*i=8KgodMEV@OhlaaUNmUI7s5CJPv zzs(k++epCV=t4xe*t(7H=%`(#JtQEi=KllUu)^6Djv!v5i+XuQ0oj*D^yxSeH54KK ztGD~bfD-nEOGMf8Y);BHa1+uRF3U!?2c2&;?N)8dLdGCwHV*RRi1;5oLe#5U1}V@Y zi7IZj2dKITz;7w}so)nCN5{wy!KZvW+Q+&6#+3kPtAK{o@`!8ysAos=F|8H%SWy!u zI*Q(qe>AH?KAy|WEYTfWWryrS?)X+$j7^f{vqu@tQ|x?i48m5v;HOkP8x0M5xw}vJ zNKEX;iDKmpMMC8c;oFa;y#^@;YDsacDd*-iwG}gTtSeY3P7}~Y(xRU?Z!I{O+Dnc6 zV@dR$m#WGo%A#0JXP?^~E{8S$>_*(TDy(4>woZ?n-Hiy_J1Yk^l5l|bM0&r#7W8MG z+AO-bR>H{0hIP`@w)Z;pyI#y05f{_ZuJ?x+9_b%u!)D=YcSfSMk-N|nKRgIcVD?ta z&uBITXy{jty*#A7xT_g!(G0Ev4&WDpR8ab}z1D>m9>^1iqWR?hT;Wr`=EsL0qkEdj zE-KMv{NJT*iy9qr$ES;-yFu=Q*Wdgs-~NCmBNlDk9$ondJhdyfijrL?H$2=1DPC>=2HpEJ$YT!E^O^xlzY z<0%2&hxrNEJ(U$Nis-8kHH%7dB1C{9d&7M#mo^Zo)XG( zC8k%{Y8a|Ie5~@B{}A%PTX6$w`gi*9qgT=Gcz%#>B@YuxI8$ls-QJSIaN#IcG8u+G zwsrsac!A32shBE-FR<}|nby|V20J2EdKJ!6GOcg{c^Y`$G7$>TVapQ0ImIWr@YJg+ zRU7!S^GJ2X)-GlP)vlvD@#d23>et7hn>NXVl2mI2;AE}K$eX>*IL*;8j!XsutgW26 z)yV#cXK*cM{}#DX*XM1ZkUENEcd{G}=`6K1F$!oEgEtdLMxC!mLS^9cG>nW~{)`wo zr1Q(M14W{swOf5g3FbkbungEJ;d|>HW+MHVF9z%LQ(N) zSBf#d-CxcQ9Mv;&BQpIkaVLJIke-u%g$9yUC@7M}&K6Z3)+q}D zsyE2A;F=~WJtzRtZxVK1hF_9s@ofxy3NxIS0tFIdZD;itncBb{_L_r+VglDXoY$78 zST@0KDIAx4M8DUZ0lTCXxt5LWr`YA>m!^t(a3@m`ISm>RD_bE$)pVS*5>pQ%3(>mX_iL5ElKR6Gjf3+nAmw&r= zrE)%;-R{@coH$NL-LIvuV|hf!K?-mHF7J6pa(k@Q7L*yQVzZMV(nxtJsQx%9s6G`b zsO5zEY9z5!r>vh`EKrs#?lp|x`{$N#_M&VU=8XV$@ZAO{<}2+*L1wejB!#OMV-ivd#e~8E~wcS*A+xDL1gopC5_{p=od&aAjfjoIt zgOh*E@nXZ~12+2ACqii7m4)s12(k58h|AbX>mOnCUps&03Lv&JIgZ>r#eMnKm|G)1 zv)pUVaKyYn+3R#n!AG>V_;^_f=W~x4T}cOLcppBr3(rX{P^n0BSIE3rmmK2%cRG*; zZacp?iE@;CG1;K)D^iIGS^GY=DfVN)u5b$qHwg&t~ z===)@VQXIlot*#aSU|Ne;p((aXSLi~g0Xd5Xw||day|mLJ=pYtW7vXT3?v!v1T`JM zE7k=vS9+DA!W}5@@2$G1B;ebo0Od}=IauGE9IbW|9*jb`=bp)w}XHBRUIWa)wqpq zie7do?2(cuyZu`|ePMnC7*DDt@fm|#Il2ZYsQw)Lq5pfe__v*I!0=iw#Jby3>T=fb zJ}is&jvri_xm~{^chCf&tBU48Ch;m3Ck1LEVBW!{tv)hR4Vop1&kVRDrSE$M zFt&`4-*;FO>}>Vrz9jJibBomy{z6f4Vl0$e<~pz90VssWNBM0(6ooSQY_di_nJ2j@ z_ryu3L%gHUd!l~M{=$K2!K;;tj}~=##tv1d@Ok{Xep4eNH$|kKgJ(eC zK1@g@KKj`|p^h&FcNs1o^#a)>@sp!?;=7}7{CnRw(UAh$MlJFt?mMEGf7Aki)>&it=uIvvqAO1xRn%;ac4M>nYoc^A^?s*xijx1Rchk5O z#p!QIxDsy+6p~Tak{$#Dyu~Y}q*zugAxy}>wL!w3Pu$MnkFths1Q0eU_X$kj%J3w^ zM{^DMhUnTezocr7l>tL?*+%x&vR1xfFWzdme|^Xsb|#P zU1jSC|K42gc~uj>+lC^|HshPqZ%^uE6|ld+Z2y|l%ff4zB8Dk-(E9DwL^JUEs1j}+ zJC6q5hqm%cjfo8>v|%c4Cs#4TN% zze%APh$eE9@-Qyyii(Yl8^KT2xng$>xlnf%x#Yty8Je_72;7W}Dd-jkip)fPJn!;s zUe4)iOfI)D{@@VfWt3H6-TCmLKXqgw^t?{%FF^eUl^8qU)sW$ z5pxYY&lZFv+KLDjV22t1_{mm`s8G!}=mQS+A6OH0K!HXYT3YBOyoAJ{54hM#$4E;l zdB83VR~=)wO`k92(GJWT ztg5RDJG2Ea{`l1h??>L4i-W^e*_3~`$~^ROoHyf(09q&>+X{F8qn%0MpCEQlVJnzK z`>!3L^ob)Ziyda8YI(3b06n8o&5+za+jG>o0XP$W;$$8Uegxzy3a`B&B*%^{;bog3C}D z5D>suL4W~Tg7=U{8T@@7f8gbO06c6feQmoT zyfLsyp(xJE$2r+mmJIgATN9?B&`QnuDuo6>0hkmp8{vrZ-|INUX54PBU_pnIoG{|! zx{TyFc_s3{>}yw(U8Sjqi9>3_La2=RQ~eC427TYb#Go zgdjt&b4?cM1jVCXspXhLjD2kRH=Gu8YML>h=%0ejG-KbAIA7Rm6fZA?PR+4*e4<@m zY9YffDJ3r>*Wt)2u+)PZ`hotVE`J>YI0U1WW@y8#v$h4Iem*N&2Chiu|IW^QPH z1VV~z#djqRR(cnpJe=qok;ch&Yd{;S5-7n~^jC2ABg_SF}3B55~VT}7O&9QQrZX_blNpZm!=GiGGkazQWHaoC5^giLQ$(#8bl+P3yPnIiCT}ReUY`czv0yYg?Tp0W zA*nE0)V&+>vcWq)7}FoJ7%4Lk8E)j6qOzD_@CDCBVCweW_}JE-&*KzOB)ugawoyf| zYhy_Z7&ej`&%^Nf!$8ut_}V}CYYsyC6gOm*0fQbXUxUWUY-JMivE*Y4vpU1f=Z}I} zA56n*^63P10e(ycRsNt+u_ku(V-JUAlCUpkXtNa=tV;XfX~uSz{T^0t!SY0jd(_et zZ7I*@s8V_GG{Dq^*`u(Izm8nfA=g>G-{dhR>6B*#cUtPf7tS6|LdzE|=*}p$h~>ch z&s$Av(d_>f^zC32M(xPac}1mb$X5~4oyk$gyV9jYeUNuZU!M-AlMFKYisDp@sVF%Q z^{c3~3JXf8{0Nboz_jaA((wTCN%)UvA0?8cyV#o0hX8w)uBx3!@a?A4s^EC03Q(#| z^3p*L`hai#Y+lJ2PL}CSmPBTKs}5te6CPQE9|~=VTe7#v%^AcBtS7krQA!E<0K)#- z#$uu!z`QjP*x1N6R+k3rrt2_<43tySlOh!`{CnbmHLf9rW*!QoG?!8SsR4@$7+1^p z2_$rQ`wpyw5G0fUa9r%hQxJsNDceU!=~`hjJ-AQ|Q#k&C;=SHPauT;qZH3G$d4)z) z=~ImIKoX<;Z~+dmp66U=N@JnXsaO55-myC;-_|Z_m7DvAW9C;g{e{G$tEC zBLIW0JYH-|{CZ{ookuhM0bbt!5nNa?vXqor8SQ^l^82P7CDRspKfU}x>=2ocZ)rBzXE1xMK%|52^8NF)*1idC>m`C9>)2Y*3K<3$ z=>|cK>byh3MU+OK4u zl~-CB{@*^y{90GuT#vy)GuQ7>nPVE^;`mi`0xxXx{dIMM ze6KGyN&yS3HHen|n|{8h5hE@!)H5k~WjUH4?jfxFY|-R)aQWv<@AKy_M(DGz%?z_i zFIsf_r8=v@mvLvIfT;Jd%xm4~^zQrk;TtgmL+Z^APrbt^QFBeY>oDt>~W^nNx88@hOg9Ot$;ka0@ z#W@L1Zk5NVM-|#=G;W;<);ZC01#Zft)84T|ZbfZ2`))%b{ zM{Cps({or1=7%uXHhhKtLxq6t6VQ4EQ#i*bB*dT71LTKiT(e?skK7)#JDEhUb-TCQ z;nN$6S~QEU1A$uuWsX{I?ToKz5Ltp;r%f{n3#g>`uG|8@o>Ny* zGkg>MNARzk#^^q9bwe zdN$oCQ%J++I??aGh+Vh0WGrC!EZKg2Hd>NM@@Y%Du%uYO1~{*(UZMuBTY!c zf$3|!Jo)!f1jV7=W0_tj8>B-JA>^wr?kF1^5}Wn=O(6p5yNY^Xd8gM^V1|;{=TH?S zrj>gx6pX4_Kkfml0YSl)oL5sd?VU$HS9;(W!Z!>w%Gb4BR7Q-yjemlF5+gfEa8HuT zU6!eK$sQ$&?IkV+aR2F$CV*yAU}stwTHH7NSb}oV*2Yi|fS_Bd`3H+_{}eL`pm0b= zFAFV%=cZb5yr-Q!Oai2B6m0|rzpQRrF61G4qIRT-oNCGA-)Nj=0)`b}#M+~nGqBu1|4iL z#Jl#R)j%p}9PaJKhz}in^ym5BjjxPI7{PY!E~ZR#>+xjz*6CFs&?(QAIC_A+xq~>M zo(4z2XZ$dKK0&lxljONNgz!3UMx;i*ae~Hk!4>VgHt_6ssiqt+r-D)76Vl;jeg@lW z4tOV9fUBYBb;BiAI%7sDLOGUd_N6*`pOaKX$X3A({U`lCZ9dD4vrjMT_WP@1g~C5| zhC<5nk*8vW@YKzyA9dy=EXc3~D*yNn!(osk31E4{%H1Wv^}VxUr{KI4HP~7k)Rfqi zjWKPC;>8g!-sDA0(0^Yt5&=RCQ3()n|E7r)&OzA9BrT)(`JZS<^dfQhkiKiG(sxL9Jajf!!f_HbX@SO zEx)PF-Z&f2ld$CE8u;Tu`Hf>{FclIxheL0INJmP~&p!zZe;hUlY;wy_3PKEY`=y6O zW)IaHvl(1Quai?WkAJj0xM4p^N;mmbAA{4qW6w#%wK}D%RF(-kvueTk8lAvV!a;ie z55&kdA$r1EYl)x!_VV`<3I-5sRJp($Q`+pR zS&@C>hjR~aDrG<-2G0}(SOl^y`J17b<&1R9ttmgu<7mk@Ke?uX9BDrZ2WQ_6$lXJZ zIerNJD^cGbOjAj~CKEHJ=($zN_D5--j^lsqVDzj(u>@|u6a0H`cR1RwI%b#R0?F-lexcMC?*`qgL&5Now>jiZ)!Vlrql?YkN}Z(8-Jpf-PtI^$=!DDu zIa2wVXI92>C6u3WF!!XwdPb#M^AgJZ4a-~nWt3TbDnu_H*YN|Mg(a?jqE_IRcxYpg z;UZyqR+A_%e6Tx`;JnIlwxJi$nUzAJ6lROZZx_pd=!VX&(y!QbZ98dPwYDb$G48c@ z2HQuqcPPBIG3tx{A~QO=vyJiAzYgAEYo8`oUZ6KSqh3+*n4;Xvm=Hn6D_3!z;*`Mj zhry)`0d^=v3HvRt$DERbpAMi$BfKt)SF9ENE^q|?RRHz@R%m2W-sMS7yHZ!)zccRZ z3SQZh$-YE&Sr^%wMUwYp^X$eN`|!H~#z)uRY#Qn`3o?Tssl8n@_H(jvflLW7mIS?I zYv`KT_BS6G8BiPI+Od|(K0@o9@@@J`=AXkzX%eQhh-l>dfO)sDkv+ll(8&bGYnR#I^}kQ5y%EIXw??vfUt7H!3U<9%5Xc*iPog z0tk$^5_yfBu&Pz2jjnfo;S1(F1Y)`gA-`PM?5_>%B{D+Ry=N2;7dKppxswE^#fkVQ zqgZ{B{<$$3JiXf#(0&@{=g93rtU6pUbkOW>{%YA(`!_?MnQ91o6vX0|%(Q^2Rdcvt zH6pAYNILq|Qh948Ga_8EU>-)G6OUV3G)G3Yw8_=;{TUaVVi%W{G_u}xl`k*G7cPdJ z3&U!m>N~d8g2H!faU=XjFkL9z@)&5w-JO2nt5MCVRu-kqv0~Z!iJXwXz~aGp&m3mM ziFNMRlR?8MN7ehydzn&`7Q}O!h332wlbm*O;pXua*7T1`Rb=0wSVG3EVR9QUXc2P0 zs`kKIQ^UT8I*sPSUoXpUzLTIYc>_2#|Y_kjh1U$|mT@Q1#qSR=^?>ct3w z>DM=6f|`tE*To*aq1njd;0EnI?%&-qFde5%r2Y(A#i1`1sltRdRHKAH}a5Wb3K zQi`w!w-9C_KEhlq;1%Xa|yO#r4Zldqjx7w@L1pScn}N9ac9&e8(W``{k>@&boF z*(kN%2Pchoh+Vk$H52;4d!TY(y7HYmmXqws*0zlP+(cx2rpO6frGLCB*B&PS#-~@T zj$FT!w(Xf!JF~2q$$!LRaZ2(oJv<1Zx=A6D>YedY8|eu-5MDp}(rIl5pmTFMReh(t z{8d}KyFuKE5H{v?}2;U{R_SwcBOyVXCfqU^w z*mBEKVNHx~`HN3Fm0ekBF<)70KD{;4l$I^q%62gwWVuTueBz}tD&8OwF&{qFbQp?cHvM;UiCk54eSKG`7ECg81TZA7}sm!I#_QW(TN3ERCwfV{+oa z{NLyoH1&$?K9#;Ia`ng&q``KYiQUM<7p&M9--0Y5vvbY2$#8$duiuq32f=iPX0Tv~ zKB1aqm|VG`T~4IC`)?1m69H{GpvGE4;*FtFk)>6H*VKPD21FpdiB|sJpa{BX`cHH$ z(2~74H+-e0oCnO?;dwA`hhuh^Q7;ioqUP(RSR6Ss1MRTv6H4nmSfBW+C*OL74tKgYfN)juYSvLXc9>Slp3G-hYA>-UQoX?PgSn z!hS@3>oYO&I2^_Le}XUH^t6olB?jmz5^A17zFijl)?DHwIi1vc#E@h2}d*3ZFq z=Aw8glc!!ygv%q$-dQfNhuCkLQE?gR-?>cLxCumj8TpLla1`a5`-#Y{yh=01!6LPp zg@M?;+tzKAcnQ)dY*l%OyeO$^u}E99Cti69W~EqTUEBBStI zY{~zAaB7_6$FP@`_h>RvKvQ(Iu)V?R&|4UhwsHocxOY$_B+=Ci=QrljsAy-@#2~XT zJMlWvUJ%RKX*ve2BBmHo_P&hQjzR=jaXR!ubF^BN>V5Msq}?wPm}Dj74ADPZ`zF3y=Z)xug>K-LO*Ed4EC6YSn=i{NdA4ZQIU}X z>66LT*ENTcFtjkn;w6yqfJArrxFw34{NJc3{Gp%2+Ham-YgxY)^zpuyhPZ^N zG#TyNX;Kko(1he=qFYWR_Nz>O;j$XtD6_hb*5$mbSe6~Co_1HJuUNJjV!p?~ad&(G zZkpIgXW9PG4QX3m^vWp&n6xp+yI7sz{azSg{E3!p{0<63FD^R%dW!A1_n1(1YF$Vy zu_j}HEze+h8?)q&eF7$bbRsjxjfTV=$0pt=iP;joU0D`ove~Uxz3W9QKB;$ZjV3F;TJJaOD zmRIe;4;;h^Jp>tkJR#6!?Re6`ixuI8hfgYFP-RCC8xbhj11)RVH?sHdV2^!JGlKfT z5xjujmo{Ji4E5p2!8Y8^X}P+W8=iw#F!h~zkqA!U0nC-%x&kj1?ZLgmVQqJT@rh@7 z_^)=CwP$_7;Hl=vwMPZk<9FqG=y3DK9w=dtE`t#a5m#bcO1n_i5F@;#kAJ+x_21O_ z&SvS}shEE@WTy5qf>=4Hof@rc#ogc5Yc_HR&~nPL2<7S=~iZC(Letu)hi=!7rLrwK3pp~ zl55Zi5nvBkLs3-17^(uP2p~g9my{lUU`8@0_?!sDH#av=V)AbS#4qA4p z7#R6&EEpmjDwqFGzvpy-jLj@4_S>Pd$B9-ql>ptRb*;KCVTfp$ZHRtPFoApPXxyRgPso`Hj+AnBFUR{{BMtd(=J#ZKQ0rZ zN$VhEF^^Ntb%wEElpw%Uj}SFj2-4UkG+HqZXr`;HwD+es@2Sxxw>lo=S&#*KdKSg*n*Ie(##z2 zuuTo56ipnEe*@iWuEhoa-_^)Af+pzafMVw02HQBLF@Kx-gjFpw?I#v|Uyw#XGvb?5^7yOjg9IP|-o`&F$MfqR z{{-KHDd5#2SKaq9A@aF^e`*;#eJ`IAIQ%2U3;1tm(D7k0=Y{Jnh9#^@o84ZXPG1h$ zo35mat+HO%G9YyY<;RyrAM-IpI)Jy7I(!X`(ILNYgXXKz3xt4K`stCkHnodMJhvo2 z-y1NJ`kRsLz!vJ0r(t^u7ck|=BX*@G?y&X3h_x9I_g0P5C%J(O5rqG3qYHu6H;0vx z+eEptl`X-PnvTL{|Jm zGjcSD(SA`G20oG2gGEr8nV& z7taX46DZe;VX1exgkN|uR?#)BYO#yo(b4PPSWZc+Xj+YK@+Gi7Ai$>*zBM#TJbCeD zHFET=$u!jBd3>ogRo9plO8JmEgm{xZu3Fx`^_T9i0w4oCE<$cdZFN4>BbtVb9^5O82Kz-ZM4+so_|%@@))3WsoW0Q?!Eak++39Sc_K43= zR<|DB14^Tja{cy?vJ(4nn7Hf_k=}xA-p#VrAsN)M1J~57P=gM-+|8LI$zR&utAV zJ@Rx|=|LNf&$)5k)#Tjhv%^}{#@VEt@ZjCKgRm2;4o(wjR-+)RFByyqMl=GsoWIB) z#A>7ZZ=Pe1+6}%HTaDK8Xv+P1(9$eCJJ3wpzslRRzb*3zpKuh{0+1Rup81Zp7oK^} z@@oE^u?25wFF%7N$*?7o6#!2{|2hlsM^#|P2g}Mnc}49K73AnpV^#tLh9e{8;8LTM z5EUh6f)Y;D44S0bHfb6i=e2LGF6HM_1L0)SrJE2t9?8^%oe=!~zLd%gxc802fqF^5 z&%TT-rXxnDp;Imm%NuPWQuJeqcaekC3Kb7S5AYP1R`^TD`H}?%1BtmW4c8GQ>If zofmPM)6dV#k7XeC+BVn3b{T15v)`+=3EW!?{5-9f*k&~Po9%!}aRy?CugZ?P5wdkx zX9m|05#=am`E=kv{aAIgn(dQP^}c?m^AkxgT(*{}Khe%T?%3TeRg0(bShVK!T6_#$ z=rXkm&yOw0&ONOsq-gtyL1?D5QVYGYKZ=dyUzja31qPT8|FX~H)(Xavy8 z7_Swb4eGumX2w`dIM1Mk3ECG_Sk#4;Dkq#FqzY*DK*9}YqESL>d{PK&T48Y#UBe7} zKxRpwo`cm#V-e=_LaZF|3|{@MMc^~rLHXB|T_d*j!k*OiaNuy^p&Eaha1^tncy~sD zvgQE_b&x;&0VYHv781JL8nt9eJSMd*0?jxb3L!!3o}&&#)Uk3~BKl^^s??H30Y?ZL z$z3(H2bm)F>~L_99^9TxM!loRK@7fs) zU9EkYxo)5iDlyBapV%s#HY*QW%*ww= zCi9lYi+T8#Xuorjr;D@BF5}=R!u#@=QI#TLBp$$*=P(zI zR5|x{LaXp2sIT&E5rznr z3LJZH&nz0SnjY^9AbQQ$O61hVd{ekDeA!s^! zrC8jkl`K}x8@d!Qi(e6PC$Va_cF8BQYVrbG$6HYkurHV^cq;!V;*s1;G=60?m?pCX zP&7;rWYaQ%az5aLXviRsM5NGto+Q=Zm%+ zA2z`0xMm?s)`7VW)GWuVsDbTwVC(Vfy2#d@pCM+gdJu9=Z8L^Csc|JES(!s zVSO+*LFo28x*rM3;dEEPAuqP3Hq zS6Y2kbKY0b)U5KM6~71(b(d|>$y`9b>!t2B1oEi|YUlPM1Fjz*DpB*#A(?EpvDSz1 z0ZaA>u`b_M$SD`>LXSx$Gdjr}2Gb$^_ke==aI1Kxi#+VNKS+}xi+c#@7 zyDi(QD)kn)-X(5|h~dFEj=0gR#+xzg?$Al|Q|as>O0!e(f^4vHjQlVQ&RwaGROjrx z@Q+1oXn&AIlve`$j*U(uzS%U{eQTP(ARYBnF_13Nn^gC9X^T_Q>FP|VUX zj8EHuY@E>q(ezWaMAWFF2r%9e>`_b3y?ew+{brpWpg z^y!V)+u2ihxwIlAdmvx1OFw3m1WF&P(A0UIFHtG5L3#_Lj1IMpWiWqG?UZy65b4)Z zsRD%tMlQ!`w(8VwX4XExreq`@d!2KG1F&?g`s#{Eo?w(plkU6Iu__yzv!1*{69VCw z{~p3iGUa;|pfl-njG#OUT_yp)&kl7}F1k(YbB{L1*7$|DaO^7^$t-5JxXyg}kF~Il z-2JO3P^ZKj)Oq~7Gw4ZCmyRtm^%dDQHAgXKUt*#gSodV0^=&hEJtpxQT6)b%Gdv&q zA)(PpnF~G`842ax5}8hAkFmMLGW<2>1!%*uupV=JJK(3_-gf?9^lfUkC&bRM>5-Az z$S2+Q8V z?Z2fcCu4+hI6jcLMktM$$*{0x?p>YUU}E`L$KBiRu`0&-oWF0Zbrf_WW#7pSAIyP- z@~Eqp05a8wSdf}e;lazb7`oetz)WVN_ViNt$Ng=H^1BaX0fN$(xKA{DDMyi-FW!*NlSs<-f56(Ao`MUyISlaIr^k*C;vlZ)US8^qJo!TIJ_5Nv8SFx z%uRG_#nYIed!fTRN!zsydSfheV0t44<8UpNseKceMMIJg*b%HBw>MR`-@%KvRLlz( zEfZjM#AKI&x%8LdB#VJLfUY%aAtG^%-t33ZJ!hN~F}b#A2RsP?VG9YPqH+PGt5$QG zZ~~52oU6D<)gf*T@OzRKDM3hGS4bRH2ta6@NJz?!^%H41OfEyAvQRq(2jz1 zTV}=H$CeNHf#evWygB6wkNsm6 zj!(zFrk#Z?ixUEQFUdS(>{?NoA$N8jOwydzx3N4~(Vz8Yo%$+@rmvt+;)lv5-Rb9W zgLmX@Y1n~q-aoR>xd-XYu`OkClCf5UKiPb2!N z^RN-)Ml*+`y3!*b0Gg?aW@ZAid9+{pS+yoWC*~REmwi3~8@di6+bHtGcVmwLRKLG+ zjx>HlW9UF(GC>$_`|4cpQCSsLFt9+8a)4_FwNm(!fGOGntn~n*;#w`+2))UODTPnb z&cAS{tVeSOx8Ns!Fcyy-%LsWv(7+Jmnyo8fx~oV)t?YgxVA2Nxsf7pfx@^5q!JNSX zB!FxmenI3xU2A_HQr-2%s`RK@m5w0fFV!hEPgA+hA>wl!S;o7q1Zu2zTbsY=&K{_e>pnMce3YmnOGG!%A{HzE z`E_>oV5z5{BD+$Mgg=0Rs+@ZGeQwY>^Zu$zL{C;OKki)$TzmS^Fvdf`4<#gl&u#|v z+(hBEKnj#V&|0Tz!@D{81)%KHi3%Jf{K>eR@QbWzUwQuYGh1lKkvl0wG&*+C8};NO zdb~T!p1dZU^vP$!cQK)ADqgf{4T6p|LI$1_>L`d~+-Z$s=0-9uzTPnaYM!+VWuA1+ zi|@F!pUM_=F^czlIAb(Q3BU$wk4_&fk2`qJ438s!NBE7v{A88P3L!v3*tq3um_B<|Z zdNB>jM8!rCUcF$Huqc8{oRiE%_i8c)#`=b`fLY?1=Z(pDnOU*GVB0{&nFa$<*$|rA zT2Mu28LnJ2=UB7tn@_bH?0n@KV(R>1*r1_l)$-!m+6nP7SL~y}W1M&3qz(7q&>8D2 zBG>(A>wLfm@044fl5@N-OAz@JZPu@+a3Cwr{9$ysvsW_A0adX56Ejbh6T6#4fv2;`5Ez=bR+wSu+^D% zqXTc9^=q)gd1+j9MZ9kVIA8REgvJpVSR9UFIp;yV`^oB`+( zB-vK~3j@RSWy<{n+KEE=$I!NiU)Q7Y4Ru1aUZ@^DtE;8tL+1dOjxU0DNXO~+eTXeA zJ8U>Edl1Mnw;1K$ksOUnF~_oqS>o!h;%I-gG%M@k-d(M(vJa)XF;riXHTz@C=RZb& zX0;kl&IhKs5mrabC;aViEJ#G*@6C7g2WbF)2J11H;V?+zxWVAqv&oTs^<_k{)c*;XQZa#yQw z6n`_-0HVYoI-5%$LXhSLK(^H|dTZ&7f!{+Dti42qq%+a-WRjTRow-mGg=(y%zhBP? z0fw`=tn^-^Y)<$d4u3YBu0+YRn&N?8mTo0MH!h+hJfXjwZT>zj$-qOnI8|r_wj6`!*bT+T5y~`w!^2` zU}61+Al;US&v||&p$uc~)vH*8(2WQVw&T6e9+qPM5 z15TO)AZOqiNwm8jwvgbXX-dY-AR6ErtNZu?%8eW_|zVe~@$)P;E6$8+Q#>pg0r=R@{mQcL?t8P`tQHpv9e1 zAh=6$htlH3U5d516fOSW_xsNwIXU6(-o3NCvop^#nW2o(#ec=Q!Z{4|d^`Id_!~iM z4s(h64^J^x$TCP3iH}Zo4#Usk z`|mx|`*1d>-`{@?pZkC~q3lssV~e$*7d9qa5SZGlc|b^Im7OEDK-q9jc?7 zVyBx0Xq$~(1ZqSY;yZH9*+7!@oyW4vjbaQuQYGJ(D2xlBV8_OxmAWc92K5uwyb%L- zDv`pdit~c}{|-aTWibBcq2z6Jvuo3jik2~#K+4WL`+?{cSZ3R&5%f-8J97Y|cX(UX z_0lGl~{j(C%$`~9Oqd-b;RgAlN+nxrsUfT97g?lv9km2aR=mq9`o5<>ou`S^+>_l zW1bC8?cos1$)t)0W2#7$JFKXsq$M3b<_yA}PbEZZm=`)k4RqW$^6XVN%J>8u`jd~R zL!4X2|2WkWUouyQVUH&HD{FJ@lzx)|X>Lnmf~Gxag+zA3GIFOl1i#}^YnOfz|90TL z^N&+vKiH7S4O*n)YCWp2rfR||U8fVNPnw0r+O(`8uEXo9$j_B-3D z{_h9O4~^p@ZvHz&A{T1Z(#N3+UPQRPsBug43yo~L1*8+Nq`Bxs{HSHU#MqcbVpeyU zU0^M=pTa?gp~IYZ^=HX+1EOcae&YhryYc4{eMswh){UY=v#9AevWm|$$-iP$0LGu@ zGXSuz_BDZr`qjeX4EgCp#L4kDvM+S9Y&gO8{J(1)RW4OS_xE#r(~h~9IASzl(cv2F z5@@C?WGdO=!><^FSqu^WNc|?G5&gi1uprs}ojWwX?eY}$2?0&A+Z?HnzC2O$#6aK# zLv@V3Njvsu%!v&*Q$c2dkC9ze-5WqosGyNG3NbnRAIZP$w{G#uCDUhcHHX3Nn4>=~ zll$$LFkEPB>QM=oBL{sQn0i*i2b?sH3-o`xvC~xC6yy=2;1z@{V;xfiY{F@QEZlL< zbgj7ud9=<)=bDA9Fh-2BDiAW$(%Jnloi+CS@9~(DxR;}B9 z9St1us;;xUUOOB6eD-PY z3Wa0eucdxRT2zs`G5|tu)6+pFKb2s$k`yHndlLH31o+w8xJjTy2H)YMJI2Kh*4V#_ zcI<8G(R`mmp0gA2^3y;rHjkj(m}4HRobV|mNJ=>hKz?0Aq#;|v+ijTh{dG45`AiS{ zF-WJ&e&0%Iy*Nk+DI$k|Y6rB-MF#C5|D|(l8Om9zc*0T>kSsajKl+Ga#DZ?hRxWTx z#-h_w5~gVtgP$EfB?Kvqk5!SLp$qVAKYA1I8a5H}gG_nQREG2&XgO>!k%&frFBiSb zoF=jGb*N=lw90f0I?OG6^zCx830ghUVGtFP5Ylyc@Vh%vghNyggh0F{Yyr9@#Or=c zucyjf4mvZvM6CM)8feuP6<^Aybl-nTyDw(8(eS4A+AYeMq2H=s+rR9KOrk z*Ui2W>9_>xXw^fK&Hj=;{F2FG9LEWub0s`=@ zk-0IT5KItw7wLB16Ebn9uKiO;rLn!JT3k1`HqInymgE(@up`XF?I>93%wE8?Gi*IF zT0Ahy#Q9F%X#H(*i)rO2#>uu!vK}8pr z%KQ`Ms%b}YGHAfB>tLwv-`EYODK%k~cC(0>Pj5PK4|L{TgIwcoK0aorO1wNwek}Fc zt1(PF-3+Vis-q4JY{uuSlRb!+@K7Cb5;kBNH-B2B%4nyDk0R#4b!zZVxHW3T=|r~* zDVE9cgQR^JBspDKa0FqK*4DLK6&|wzilrPna32?$2i)?&d4yvfnyN0e8$A0`nf=%W z!*wQQr<&6g7k~K?|6! z_}U#S=w+OOF0Z6rAB@{(%%4&EA@&f*zHmi8#1Gv24 z=``O%CBt&Hij+Mr{ex9ltelF}F^RQ=hA&XB@T7TCRqDf&uSsrQ`OS2Zepoma3&aA` zo#7K?eTlyPgvph4Q<4#j_n${hnLyMZIKun*F?NqlISW;!ifKRro8I)& ztpMdAQ)-`Oi8rh~MN8L+2eUR;CH_%^J%^3qB|{KMxFs6YHS!KBtP?(>tgs*zv5PQF zxIY8{C{aP?Z5CMJLj=!#n5%60t819*Rx4g|mHg_osfi8KYIbf1irJHtUclmB}UlNYeW$f>gNH8qzTc2hh*(TZEKOswWPnF4n(W=`%`_0 zaikelJiH+*r4{&+Oqd(~@mW6~jWgiuxHfJb-3I!^6{e2DG;=j|UrNxxJ>5AT5X=3WoI$ z=xQM}b0>a>rL^x+!jAn;SCr)w3%(wA+cSIgNCV{1-11Q{im7Bd5TbH;e9br?um2P% z+GHo}r?G29o-7r3k$ff7J}iJ*g&X@+1D+d_1V#pM75Bi?-mow*EC5d{9n#qddtze^ zCbsg%w(LZ&%--H|6Xw+_|0;saJwNXz>J5DKC$y|Ilg%SOtoD&n@m5=Kz~OOUunQe? zZM2Px7wSJoX*SUUf_C)6WP(WN?FyiFLDWKX(;8+Ultxk|YYwg5HtX3WT0(w~x&S!2 z@@L$0ul~b^0ql3!_x3C}qqDlp(kJbhh+mpVNe|U@VdusCBG?RrUEhR zdnFQGvfPf|Wy~hILDH~4h_0(PBp#JBXtbMgTkpC?vBpx#+%$AXziQ~cBV>QCI9qXe zgY``jJ@kU&v6$GXkgF?eRf=|5C#`hTn>-KEwi)(nEw4cw&)D^3n9lqbTbp3VN!Kb{ z9oYuU;BD2o*xA`F&Hm0=UBqwls%X}YGw<^ki?g8c9i^78=dPC3p!J%?nqT;KRMh zWJG{BA!6kd+Ojw-J2o&U&B<;v$INE?YyQ4da4PQir=q=GGU&Q|X}f)0;+Nn$f1}?| zC_xfZE?7nlW(!1V$_5&$8jV&RGv6VM_{~IaT`XJKT0sM1>ki1|%(dmXij$;uQQNrv zr;q~O-Cgd7WCzYwzW`VDiDFSMim`35i+_dAUunL-gT`dwGrj0rwg}ZqVJ&rt7R=!2cTcZjKsbC?533^k^P$NRv~iX1JQe z6#Kg`isx&l#N|X4&k<_*$sWD5R_{ceHyHpOJX`Hm0Nb&DNX5^txEx1Tf3@^0{UN^2 ztNnApEZ)bXM%W%j0T$RmdqeM{4wBSMzxL&kFSAE)7h!P~6`_^!eZ)KoB6(7A6KbDo zjTXYWK2tbH5|QF4J;JJlgOO7>!e$goKk0AHC8P>vtrestdoqM$ zB?9NLcdzWLsLfkV5YCOm&9QeCQf9b9b0yYwF@f5{K-*ZLErM%E22v<9xIc}lmu~H6 z`(JvRDW*$0&3Jwt{w@KdBIwyj*0(=NCv@oZaXRD{Ww!Nsq)e3qI|eSm zbIOn006yw5d>hyNnB4U&~?fxPUs$Ggqe9l=Fa9jm?52o*x;)RG^fkSqDZ)cT`x)z+s59QKJvs*FUTd)A3R?KCw2IK5_aO9i|8BLyHPkkAjA}{DwXs zC5S<(WHLnmsrX!ItP;8f7xi}kO4uD#)^wT|PpL6jVH>q(rlF=$P`lwUhf^QvnQn*XG;;lCx-JQu^X}^$4$8!DJ%3=!p57tqUMqTR7l<|u1tpp6ld~Py zFGm4AK8lm7U*cBkH1vYy5}2va_D5v-zoT;%o5IfGSB5&kmxbyBtMKfKz@)1T2m>5; zHx-r63H@>~GU`2ze|)ExoVrdDI6t60pUSwa!9invJG$LrzmOqhvH6=o5cDHnFwmfQ zqI`Ki>vWL>@^DwJ#C)uLkZ!O0#CPopGmW#IKNIz}D8VM7M*4hr(!+XMDE1Qi_?h-n zv};)WV2??h1|)|D-DV!QPM(i}6p88+s3eIQ7oz zQGci2&xX{vl||GtmlPz*cP}qA7L29Hr6biS?bdR3$S+7PFgC<+GN1EDLQ(~GvqeQ~ z^_bmsCVO$!g}cU8J?n>azsVj5avJNvG+NcaR+1|(9@~gL z{%id%qkh*^94tcC%q0&EH-0cN^^;2T&U>`lT70w)*j8LrKz)SDH5#4PLNgBW6%4F2SaJN z5Y*AS;UZesVS=q0+z`Rr`%9FY{hlJ7umaMXjKs&BEQX%*-p%(z6bx_tE+Uv zVEbd>75+b<7V&|~1^?DzK;3K5HnUcd(4EcJsj_1W34#;IxH+HTki+95V4xS?<5yww zbyc+Y62%{^=_-554^1LbbYA;R+a%*KJ{zpL$kwwM=Bi}v!e}Ky8)c#SQ0Vgyoe-^; zmHlr+9vzIOHnVUfL1BDsYS#_Bl}=@?ymY{-LiQ7iNOHLqLqnJ)Utg%}vlPwPmy09E zn`(lqx2)Gp(jw_JK{G8hInli|7U%k^gP`f`IN2IBezkHm4-Hv5+_cE=J4xlV)0p%T~{ z%FDjpnvKtQD9Nrs#2rZl=A1Ho4HJzVtn^xBWxWZz`bCwZD{@8N@Oaz$1zi-H^yo&v zG)a0xDoopa=vk9;|0B|ilSpo&%QZHh&mwHiuzZf^=eC@YBg8 zG*rVNX)UQeSTheP={w1b->!c$@}^on@yfw$JG*J@N`t7uu64|OfNQxrsj#2O3>EDu zRK2eE#MqCZ9QL|kpg=l zNsz)ETMjUaX#_tr8R@HfsHVBJ^zR-hTn-^mA&oL*jl&};ILP5A=j2Ghu`)Q$a$9^P z2U?I_(I2fvwH{nK1527vyi(xf^^H@cjuFbLKa)}LMv=JrU)Qe&|G|R*|yfM0A zQbVv`3J-U#wD+oiH9Z+60C_TOy(;Mp#1UfHD)YH+r1AP1a=bT9`0=nRfF$ZIZ6HNj z3HS71(W<-$$gl+;fno*SqZY3D5X#ATKgTCna~lt!YJ+0oI#1G4vWWdT#PmBos&w6* zYR__pS{Or+b!O8y>v|qxwoHv?w9Sa(Pr=E8lRaj;7=MZh(orq<_ilcMeWgG;nk7=r zveBe69j30v7r)I81Pf z669gT2Sx#^5o@q`WvHXvCl5>!sl}GX<6Jr9HVNzP%A$1`D#fn_^Z!zoIw<=>%DOMr z$i5%1bvS9=Z8MdbvSL?$6crhI)wEtgk9?$~Z%O`&&dHnX3c_SmBA!MkmS)Ar!;7o` zW1ngEk^Em!g`MqvyZ*b0aXxnB`HaXt-GFW8D<=x2M97J|9y{{%7rucyg^VSN%{O_V zPhX9&(04dMh=lDwWzM5Nt`V#9p%MN7x>HUx=27LRD~@J|=VqfhLsj;W=(Y&>cxUT{ zQqKQ&fMxi3z4ZsuFgVvyCdP@`*%Y>GpR#h|DxZW>XhrT$9&+iD^~*x_)8Nl&m9^S* z_2M;S+8=Ffk#%)RWTu698pqxY(!Th%aPCZkuc$X0DNiSQE9h_O`X%}O4}W?ZCCwhUSe zLw1m{s=zJEN&JfCKcs~^8ov7PYuv5=_lof}qvQNyfB8&yCC5O^N7&DbFoJ;m@;x;J zvx#-%Y!;6K4%phjb`F9EJ4}qmBb*FLP1dEm$AFGY1z%b#_}(tK;Mr z1W@wg<>(8rVbZcpGc$dIxFq1{mH5GPJ*W^cZ6d>!Q7m_J!n%;etoAo+o-{(p^tNhO z3l>)ikzQtbO|*`JxN-CK_Lj5t>@I6+<9+-UQ_0!w*zp#fw_!9KY_qyN#XE5&lb{SQ zZl2;+r)4W0vdLlY)bPLhlUj%5viHBi&ESj@$ zjdWLFWd#j&tgUT_p3SRe*xz%F6>K;x#$k^M3+%gyx&^g89ni{LKMEL<7O5tFVArf} z6btUm;HcuUx234R(u6n6$r|(UK0%_TD_oJrE<|rGd!F=~3uyrqHm0*nOJVg;#NPIe zxm%U>_P>rYtwrVZQ8aHoxc2n^TVlss%yc$GJR&mDP7QLM1L5g%=>mDs^{a=70G~bk z$xf-gR4t85>@{l5`UBe z5ufpKP`|MAdZx)2zaL+8xIf=FDWdMhohHubVUtoy+z4_V;np&GQUl(j`DY>R>P8hV zaYlr2e_U zL$^|;j|Xl28=EzC6*^I;VtCRErdfdPo6}em0&o2Lq(WT0D_LoZ=hQDWL=)pCOWmNH za50iyQ)XJvH*eh=GX4r&^)0681Jlo@^TQ$&rF}7BxQjR|mb=ukWE91jIHFkz=M@wC z+q3r*g&Q9Z&I>VMh11{n52T|s)4QJ#tZ*8i2@^YIZ4|v86mJ!arl|An=S9zBuG8kZ zF3@A6>Vl+@?d+Sz`VY!{+TD~D_Iqt5c{R)&h(+9(O+JQx%xL{NT|`|feRK15yA<5X zarpR`CKAaodB5t(LfmFujiUy)2wr>Pew&3v;1mzeuia3zFMpc@Y``%JYl&11PKCPS z{0~HgPt;VB^dX_POozw#*c-8YMA$St2S4*KtYZ9+gRrGM{h6M%XN_Or!@0RPlOxj2 zcS{ty{dNBM1OP79iKFfTWq{GG- z<0>aD!SC(m<_?nmuKl1(U+FkrP$wmbCRA{5qBfLUhd!e6lpzttYE!Db-tfp;yR8JQ)ShHW!tD{uoCW?SpR+#&hK^3u}h~R`IR|S z@t-C!!|ZU*iO06)M6tEqer9m1l;Rd}2UurBi$G140&}e-AYvROo zivLy|?dw+$!n*yH#&S1B1zxbE)B|fWDo;Gf*7mgzEuXJT*m!*(WYP7-!E7<1oRcnf zo#@y-cCx4>v}VVrw^2sd5PLVpcl3_)j?fOxC4mDjz=VmYP)x_%G!9H) zJ&8ipNq{vu`eyxha+_**+t!IumQWGsY6gJD^+97kZf<6*<}it(k46XMQk# z%aHJx{J!eTEB6mg)D^P2_^74ugs~I*cQD&%(dh?;{?Q-X9%(NX9RxiI%nb zo{4)n~FXiEl|k|Z9Cm9ww^F*S)t{Zwx7l}dmOc?_E+-1>*U6a%C=vB_~|{f z`zI^J3)}vtA@RAa)XzRYU!pO@ri)7*-tnw!U+?ofcnj_IA|<}dO@$e^ zbF6JCTIbbb_o&V6J1tQefAEtY066rUKJwSu-kz8x9b`5P-uvFxyD^53uSI2PxOcDPD>ENmHX$#fY$SgE1(d01P60`e6XjhM5$~qJ`(aRaZHH~Ul!+u zq;7+tc}M10d3vnWPm7#JZh-P)r_teyZkj0=;!B)A*!|N2w-oh1=eQ{Oy*_tPrFLT` z-0*nV6yv9==WlG^GxWjzl`DDEnYHKOwG1ZlU=>Ysv8a)Us=x`yyYEPyhdi&ld4<5+g4x6 ze#elp2dc4mAP}BZ^E6ZEn@hgbN8T;1Uz!;?H4m@?TJe;Qb)`A12mD=qknhvZpuq4m zqFCmf*-Jg>7X4q`%0q>Fota4XLqupOlvw&*AUu{(*q|)yXxf5?7 zhWjR?%E-4$VZO>WkpQ1R+@M%XJ{EfdEgCRaJ;66heA9dSUCURVEQcBevpW$p`IDr3HqV&qfRts7>r(o!A$6wZ zq~VSd;vCw75SO}vTkNrMF@11@qqsC(-k8XSjV*-$Soa|C(L`ssXaiy8UcM)zji~?r z#X;v!a#FdanGcV4Ovyi{*)9^Se-Q4qVZa4~sz|gEg1#q=t(006S>B69Cw-}WqO@t~ ztk>iY(&a1WCy7{@z<7L`fu3kMTUTC;9H;BJ`54W)RE%6b6LF2JAp58$&0fwV*+RFf z0l8YeEXC8rb$r6j_72N-=o7~Yb`Dnj)6h5AAJ!tam=r4qntBTPV`48Bg=CV(2i~8T zfhXw9CuiN1kEemt9TG>SDtK_?xiVEeUj<7AGwH^NHl9Q0)DBuAX$EO|wd?&H+R&8q ze2CB_^eS}>d8+&yva%=tlQ-i#!k^=I^@++m{7WUEdz=>EiB@pKNT-?mRwpKzqYl}X zw4Z_(_V*%pgvj((YDw4z2Hw9&N*vg6jXZkB`E$&40jGSZo1sG6P0{o)=a9br7xlB+ z}bcD zZ-~!fmV#rJ>A{jEfCKVPE`yDXlO>wWL?n`-NgP-(nlATUataI{cY&GXTE{Wzu@VYSpSe|Q3`L)sd~uYHJPM=m%QNmHVV-Abr-HbEpaCmhP7(KmNo&H!7}b+DpU)PDXj-}imcjX z%?#Sd*qTmwUnCg^1dDUZmq{r}THfSj2h>BIxq|u?QYs zJD>Gln`{Mp^8PJ~#hx(&sdmjbun?A3TW#1YIL#~*s5*P)D|)dzF0PvQm34Fl(O3Sc z26wWpRB+3rB*q>lgWW`nEnwVs%vwqJd<=Q$a2VQzQWO1Qsf*F!`gtFfCVzS1G2Ffk zEpOzi-984bJLQ$y#HV8B?dy<<$A{qcPjjdD^d->VwrO~9F`lVti z+CJt@jq=BKC=Hduo&w)ATn6Uhmws3zJhC7=U*KY+4iV+ISGfQ#Lz`#(VeCjcp}f~% z50}CQxxh+oHD7vfO{NKSBB)t9*;gv08}CVyD{Oo-tvtP0{#t zvIV8uRY}1=u`0Ux+PhJd^wyn4mNIDsu*;R+=5y1AXl_53kq`fRoZM&{|EHqBM|QhU zdeK%zV~33zWSDmCJq>C4$w;w$A!}MR!<^1iAlb@@g@{U?UTh`FJ*x7*Vj7QK956yO5bj3F%;5>^vqGIsfcjYbl4 ziWkm)Hu3#Sl`rE~A1>@3GH=+J#wfJwmT?%>o$z}ihL84deYu!as2k+fql^V6dJH4H zZNamV0n{mo?g;pI(6*CD!|UWzAXld>5lPX75?FH)Ay1jO18rA2CBcOE;a6Z&hAa^t zyW$lDY9f=7;)Q!xR^s32q|0PV)@*|OAHOg(0KyYbjom(^Xru(q^6v^p`R?aDyU=8J z*uyQMKZ-%#xKo}m0uP&5$7&+@3@$gXOrDFVH?HCu!?Uscub{YUOOl1Eb zWa!MI({34ml&B-Jbzcp+mF3lr!Hs6?MSlD5S5Vn_>>H61y6HxH%T_NlI6C}h59NEb zR{HFOOvU&k{+SHi^}l=KV@TmczJ($fBo%+IT1E1a6?H4?sr3(dDZ~FBYfsM@4|0j` zk+V_%?>CM&Odgt^=rl>lilgOziP98ne;kF# zwJM?2&Pra4IoTA)$1gXhB^#=hq?GE!Rb9D|ZF-X{`gc=FUXQm}PgUpU_q$2m)2A(^ zvUu=B@`@K#48P4=^Njrxnw4gB&|R)KNOo%JwW+q+!b_mMDZ$o=5POY>E-#JNOdd-e|gFba+?R{ zBVrw7o%s4(bO2=G)=G?Tl|B$h<#!V_4{z#Z{!~sd)dOisGDGYhqMJ~#lHzM%2x6Hg znjU*;W!~ z8%;DYl$F$F1r#q@!iC{{tk$xXmXsFgqEHUWHPzFlEZ(_B_}chNe3yjTF6~%~@z)eH zmTC-pBY5Ka0r~{(8wBD3k!%U$27M5b zel`mP0Y8n}YfF#dbzb?s`SHj{Sg1-D+FgU3Mfhu_>1KLS94KGxJ4Q5{MZ(TbO3Q9O zjfFgz=4z0Q;9wR{ko2P&1JR%lP7#*1*FSNS!GlN6m3=3yFKr{iQp!adJ@^O_7Ws$7 zZMtz^{&aE><;eFdIOIay@ixu_hbOw%LvKjM>S6vwO{Ho%&VE~XweA2xY0S;^GIimL zZ4(pJRe?$52hz3T-2DeZ`w%Uxj@yR1x@PwMuwu;;$!2P^x1Ux11uA?9R>p%=scT$R zR2r&rDD7z3B+>R2Iu%w6QiJBZk{CO35WU_^mzADWBOpFQi3v8c4a4*O>3I@A*N7U$N0D7TKTUit~_i@EgagO z7W{4Y)=qC;c1nI{Q>3g~Wq10b4a99Hct{*UB(F&26$0dbb0*Lge1$iPE71iRtp;h+ zZ8j%f?8PB26eo2e=Zby6Vj-f4kp=|_L-CZ--Stm@n)1wd?gqwi&s20thCA!|ViGkM zH8PO+V!FeJN3{96TJQ!=g%l*)1|m5wsyZIc_f)rspElBOH4l8MkN3s@M%f=Sg$Iq8 z!rq*&ie>JX*l(L`a#nj4@|06XcmkBnf@#b5OSk&hZa?kYH!NSpaN+t$3K3fILXv{$PKQL7fFAb zsQ1c%oZQ)T*S|;>oc_ssnm(4(9M#o1w6bz5QNU>9uy|{*X!0Z4-WlX9L5I8}E~o4r z+QcJ3LBBtXOHMLKwJO)vS_B|zG#82}CPqU8cf()ssC~_d zJ`1w+>!N%8;gKi27I*IJe%NTCe*paMcw8z#^VP8V2%78~lN-PBR-| zMe_1BbM#Q9N-k1xz*FNYebHJ($*P%Z*8gk`e19`#a~jsuvsJ-qvRDI*={w*2tBfj( z|NL0-zCttx9=YmN;+9T@1+MyU+$+o*P{$;f;GwY=$jJg_Fn}~U*4LpJ@c%>1R=?US z{k%-nEMJ7_8-;*!c!T4Et<9#i?3sR%i;Ibv`CBou(>cmZl|>cU-Jm4h3VevdSS9tz z941xcUY=Fxs6_=2Ucf~~9mhO(nN|pEB~38B6ZA^PDp%Ze7eSA)?GMVt30M4U`zr9o zbxfLai5Xu5$FEeG@_X-@(iZ`tiw*ro&Q;fD->%NFp9b(m)yZnnUYp0d5p>P zOIJFyo~)#Yo&r$vUh?02t&CN2i=tM=krEYOr=vJ3SfhvBH-9#&^pS*~FpGlQj-5hR zLAyP5XT-t^<+FJBY=1xDS9FiK%2w*-3w9TY{DJWD-YFbwPY=?z(`yo6jW)y&OYc95FoxkBuMnM{Asy!S(z;i&{wUyAE z&`~W18>AbRDJE?fr0>D?qZJlc9r$W+Or3NkA`NoZ8mNGHd1M?|v=T8L51ihunFy_o zUs*Qhx)8T{kd-(WlnF|eJ{*X#F<2r~9YjtiD^X&i2KduwCA0P*(G5>J7X>@zIRTr- z;Stx&`iGhejh|(lp9?NJJ}dB{rgK-kr;H0!{m|x)I`iVxMI*I+NoOGnH9I)>A-Ppd zsGu)XfWk~+7=MB~IMAn6Q2muf9^f}TY8F+9LTQq|@nvzZchebdl2?027aug{*0HF4 z7efF1FF2%^{o+sD)7KlDWQGrGiirK%tdJjJElgOo{vpyU8DpYwAN@~c!poWAbvX{c zNa|lQf{Q;5-Dk?;$rV|Vzxs6Fw^BvDxN7Croec&j_d1@GCD%qjcEmJ7)3eqgVo^Gx2bkYprquJr|hi*Q!5|f zK!kt?n+GTf-LXgF)q0IwGd2ym8t%NMOS|uDJJdBY?|DGk^~j4qh~r_RoARC{Sii(k zq`q1_`+-H0S2567f>6i!KC!hUBmA@SHnmFLdvssgHM}TEwD|)3r-iR8K$Q>}$$vzE zZ^+uvR1YYC_PO2O()-(tqeDd_qHGk!Mi0*=(MBuxf7j4FiEuT#a>bc}RNFq47&=sJ z3xMa&Z_-ajf$-)tD?o3^^lr#3oM`?9FdSE_iqAr8hbZtRK zk6HEuF6?5NCHn?ewCwltx=%;b`n6*oP@HnQRBljg$shbEfQ&vW=f(~6d9~Ig5Cb^zBYd4|Y#cLVm(~obs>T5`yg5zV3 z$FGeQh&$=mf%8i4yH__mV_e=?1)yQ&zBiTN!Yu7<8WfkxvvfZuvB*ViG`jfw`P4Zp zn1kUhD0F;z&@IbKsRMb$#lu8&Q!dXXP6>B?wml1|B2LTrk?#M*QZa@G0UVZ-Cx<%Q z0HR>C|CFs=FMqioXR(Dr_jM}u*f`4Z-_V%jWt>N40I3(!mF0reP(1pSyAwZ4>Ra~m zFf*emF_)U#K?@%Yg|bn@va>}9;v;TiRCEYQm|MD{|_OWv8g{Z_EVEWelCPLu*Hp6X>#d^(Dtci;oh{V&UNNW4ps>jN2z@gLrLtm%jwOfU!iiyo+?)NVT z?Y5$P%D0L$6)-vG0Lj4H`fFL;sc5VgC8>%I9i?@f6kMft@_h+w3}W(`Ale6q>Y31) zSe$jj`DXdXR}WL&`Rgp;g*p+b>jB(m{3FH7eN&04Sjf$$nCUK&toTJ3_m8ZXv5n<# zy3NjhJ%O)D*RVd2Jbb&XDb@YZ{EJ0^QY7sBou@~lKxjg zc&~*d8H{rK5N8MoglOMOZie1T~+Ifq|h`=k< zHd^Hga+>arZ3+pQOs;WR;p7z^a<+GhnJst1@Iakf!Lr}!f%<|&IeT;Dyj*OT$JTBZ)>r0 zbZ)hdwUy=$T%CvU_8V;1)8%C6OcgmON$5^s^Ae3xK_-@STkjT11Lr$*(zvlI=c z&8X+B8?<#YTnNWy?b(%49ANq{8j#*go_|~oyu~J?6xRGw%=m}DUs|YVx_HYse8XGn zYxT*pKABFWVt;+irzsV;`Hu;UjC@S{Cg^;#qI|O1&kY>&;RU`_Qz~am`|GLiJj1`; zHPhSp(|p{}#%wqX{ojB3_PK{{qprG-!fpK@`&RG36!3^tW)&GW?a)H6{g(d6IiQasztjwwgdJ zZSFn4ooipFH+UQHADv!38VrPnT^fBFkk1(rFFjb;%49373jZY>U}rK-e7Fb5K$7_i z)?Hqzm}3;arKea(Hlrf%NfRT;kzJ3SG^L|j$GSeG*~w&Z1_bT!UpoA%oFJVWFMGmo zY=rv-MB%Lwvcaw-C1CUKp42Nf7Evkb$=WsuRvQ>fm~3QPEZAN2q$LprejR^1zR307 z?auIdkO|8o$R?=BM;eTZdwDW0o$j75{u;~)ja!|J@>flt+9QU58Mss{KM5?~JaT;7 zOj~Fn5j#W0{zYdN=S{1R@zyD@cuGt&L2vFkii=E*{mlyL^=tZ%aM zD5=SE?$QWiyZ&gKeha%W?Qz^$e~U>3vdBsEAzt2ip1bXLy6`_*5!Tb3MWmX01s}6Z)2KjjQaou{IB?w*cOCSg@%fz_(Pdy_Ni9>Tv6@2ewXkuSaV~1Kiz(|S9BPKGW2RORd55=)K>(m{e9}ReI+JCRRsHa>~_yR#{CL0+T2;hf1~j zmkNXQY7h2;%q;oxkTv8=vpT;7WfGS{#pmz2jh4wXmc5zioyOULR?}ULCkAkL$|dzG z+syhzk$8f)mSwws1CntdBD;thE!uN>Og3b=7R!tMupRt1uO|?$j`@Sw6uVV722ve3 zOO2P&bJULbZBQaD^PS*E#f%r7!56bQ1;>RusWpFTaCff?>Qz?FoTqjMT^nb7 zKr$E1CNj>U37_VpDCa`HB)9t(CQ6^BuNK-I$@cO07zup;-8I}K#MQQS_1e6)mXtC2 zzRli6NPSAOBCQ5M;`9wbBJ1EBAR^PG@(_j8kukxAPQ_u~3 zbem2}nxJv^DX_eV(%#q9c5+exufX)oz~EG=AcDm=wD4CmBT`-jfc-mn}~-hO5*ry@Obv1@e-70}rT#;0Fxk^dFcP8ojjVQ|d)me87!JCC17H}AU@PF4yI&e$21A)BoZ;z%gcigqc$((m(Fp@?kb}+ zl09xQF4PW(_+cRBidm-J|Cz*&jSv`Mw|8+1jXpauU21+V2wgnH`^YuZO0dTN792%r zmkp3SHXxdz6d79K0Vs!VLe<>QSH3ssvp7BS2-?Hz?zjWNoICF-E1P6fi_$o+GugcRW&P zXQ?)0rs^hm5+0dIUkqn!$#fF-l$M{?w!A3BGGF-W6;i!ztR@gi_Ewyd8m01-4~M zeO5J?1f&%DZxM(2j_qoy2pbZX<8GqJUSJt6z}sIpoHPBKkhg}W{EYL%Pq7Bk*VeID zQLhxI3oiEHLVj%ODU_d0IZM_e-b|Xeajx8Nlc)!&JUadfhSB#XuQf#rPf$*M!y!+K z)3+*3wFRG})AbA4m@C-`wfJTT2}`PYt7_<^lOd(DA$KRJgB&1ikhr@200|Bd3e3>F4Vt20J&O~H2O-;dFL0yb2WHy}p zn*2_D>?Ef3wgQpbcj-UXMBtB9 zM|knHg3YsTl6*axv@3AwA+6Ny1)Lt-Frd$~Q5hI^mb>dd1I+UEr&vj zD3s3kvU2n(Xo(Xv5ZQ=v&PWcqi*?MNHCwf{#PLMqLbmFN${3Nvve%a1Hf&2}aX29Y zICr(sFPbGkO_3J4fcCn$9BK<2i78B$&p#OlSQ-qDvy3MQtUph9uE5MJ#&Cv~{;ny% zTJjW4p2~zG5%h{4o>-*}uhwlVFemD)t0n=3P2zOY;}KJ-`L&#+Ec3hu=G`jm+$08)eySKKdCL<-mKz3OqGm>Ny(O-!e8w^q4 zwFQMCWQ)!-&A@CYPk$K@{b4(N$zdfF0pG*$P4qOlyryqP(tAIJtAl&HlPAGL4=1Q0 zP(0TuK=f<1NJ!`1Mf1|jy|eds0)H~?Cw`tFcnxV#$cu``{D*((v4JlqE(C=$y?N`^ zB3}_?cFl-z%JHzpCC3j=JKpXg5r(W~{P4XXDmPfcdgKDo-iF6bC$e(s1kG|6xKN(|EDDEtk23iy=V;l z*A0^Fkr_v^M?V$8hwD?^?BZ5qE&kJ8KqxF-`XY9oWZDe`o7QE2?WJ=hEO!-@OXO#U zC8gNLakf~?VV)A`o7J={D=_y5|E(}9q7DkZI{am# z7z;FQDd_IX-ANyr-EQV|e5$j>aIS;MK=!2)z~ z4j0YP>w=E}ddtj5U)8?*rMYBW0Yo;L!(KEwqv!y>j-Aux=&f~mG$y$xD49x@iJ)a2 zrsV{S#O|eax#ECxqy`YW0An13oBJZMR0|V;NEE*%wt@>ft1?r0EAc7Oco*~w9^Ca) zv12;Wd{iOg0>9-MV7t`z@>;J`qn6{(5E>~4veQkwxoTPY$Jf^7D^uBBEO$Y)&Jq`0 zS+b}##tHBLhS#t>LQ5Rae}b~666bw7(oxR_uF6B_sGcWQ8xMBz&P`__9B>wM87Z?D zzS-5N1dz7?FQ;4{npb)#d2XWqAEUnqq$q!hv;HsAvAO{1KKO~%vw&kson_W&(*XqS zR-*|2L(uT>9|AaQ_DjA31_DM-a%*~(8W2E|xqb@I|Fp%rOjfb7)wX&Dr#jQ$^*=T6 zZa`kI=aZtT`XrH0e`V6LHCTff7ds7IeA&q$fnmf6+&N5KWC?d=7MRK`| z7+{(AhXen*rGhOt?UHfl31!M1+-6PYxukQh#HGT(yd9P-V5 zOUuZp1m-S=HXM{`&Vosdz-v~7=N}}AX7|>t1?3Di4O&WZ`j6l|BkK6ATl1B+;C|oK z3i?;K!sN86QU@5dPAb*oik)}~`tIK3g%t6|UQi{qW@)dUmXi2xK?w^;O0 z_wY<~QIAi-#XE{q%O7jbOKB6AEbwRfLx(3sV#{Y>CuMpRf-~z|Jz3xwCG1mG_k1ag zwR=k;)$ll?@u~{nT|>=j`PUN~|hHEF-Z#i`Cj?`;KNAH9OE9TxECd{Rj z^+p^gqF&4`GS}TkIFHz5x@!s;D!*OAd#^yb=@zLFSgK^lSRZmDi!+TbX_-43hmxF5 zBVq>Dj`y2RkHnB@Q6{B%W#aVDL}o-gJ$5?l#DLA??vNuHB+zdQW{ST2`O)<+amg+D zAr%yk=he0`*E9?Z#n3`84o#^O(VKy>nMBr#Nn;y3^sMv#mafRc{-v%>i(XDqu z_6>?K1jRIZBwU^cNPDpQ{V5IPI}tjU4b3oqO=8w|Ga?yJ?LlWKYZ96%N%&wgF=Ypl z3oA%7;wQbzusFRv6*^{3g#4(!d?fsSvs@U0dlh1qtf&^@D0}TXL81n9f z#mYMSP9K`m5wDVDI-UZrxdvxiF&=sx?SzYSdJhugb7}0^xJGVte=0zgtC6JKmg?yvM*SH1|lfX~hc zXsnm5cciS7RmSc*YxoXg`kxH8zbg|`|8Ct2H&w4Fs9e38&>>ouS$M!nMa1HDzv6Q& zyqZ3k7Qmy1;fC755lf=83p;`ji)kP)h+vS)ux&UzP-Ki~0`F(~K7{5 zh$$g&R4DkLV1``-C?K7f;5rkgvO49evn`ch!(qC9BrdEia;fvnq2KkAAJPiihM0rZ zCr4r|kO1XJ>Pw-(XS@D*O8Ua4zHf!a>H*T8z$F*W@r~%n@FY_dltjm59-ffFV!akc z^Nj;j8KSMm0!L$ByIy7Ozw&j5l`{azmEk|FYi0?Qepnyh*b@VuDgJ4kWzUNn6|lH8 z_{rFYxI^*+iENw;xc0K<%*zR0tQKgwNfjOfa|{*>?ZW~cfc@80O zKIEEgrp{i}iA@cO3KFEwH1mT6AF84{GML2YjAFrMhVU;y5Ezs{Xc!GHo4boXRT;iR z{X-s^0V`}ucL@0&x8=qfm0>8kk{JgsOINwhrGXFx=zfj|m${D(gU=ANiS=2)<~u-Y zE`b)kTD_e zvx%})PxQwb?;3c9tWg^FqNx(HC9I7jnb!Y0Dps}>uRSJo9WW68^{rwK`d8Ai)AYz) zKsH1Gm*I`6-(32O%5Yelc2^LL2f7YGhv7x^$d+{IrkW;2()S-RUGXsyClqHvsm`n< zR+3j{A!}xGud?Krn=rC-uQ*3Mg(H!!$gJd2w}{_Q+ceW7f@L4 zqCRaYFNqJ`9n9MB(;!F<2VaV(E!E+Z~ zFkw(AT#!V=ucz{>EJ>LpvW1xXq{2dQ6`QsruATcPoq8@dYY+N)+kC<`(BX4~ZV_}p zfMXWw{WZT%3Gfvn5emur8?cjagEyjU7R3=>9;=Dc*b>u(Og@dJS6o7a%1P>$w;j;S zxb8EeAqKgK_o5@;64pb_DQ<)$(Ot4hZ<3kK?H27@1K~Oi?*k zCSGqZVIW zRR%b~`%GXL+3UjoGO4?o79J@Fo;at%Bs0wM0V-2m-WVf6X4v-?Oqv{lATR4|B#15h z$>v@unm+<;Zd&l1Cxw1H9T2FcGR#ji@B6AU+{^(6RJxg?TAf+aRBUZMDAd>!tT{Q- zjb)FIuEwks8{zdzG?0#N{TPaLjtI!=KTV7M<$2@TUWooD)cj3Ut}iig@dK5iWW^?W z&~7fP=~nzz;ZImwATnR|spHT>Xi@Z)=jwsTR!?L!u?jLK0#8*RW{k9zKwogW+AS3a zK{+c`JnD-~W7R^J>Z@PA&yu^Z|Q+Uy7)JzT9RTo(nV#+U}Iu*5J*g$ToiDJt_r5yW;ug=ED&ZUjj1tMMTuqGuUbAVuO1iR%UdO1 z(?Hh3u1&#F8Ooyj6fcl4;-a&!(<9=9_JUF1QG+5|&ZrDaa?Ruj{3LW8z(UoR)kmug zW4$zLe+dfppalQ+!(v4rtZ)UmbSw@}EP;+0^@$Drx1F$Bk`F6Dp0?gM-F<&7Rz?NT zCBXZi%~yc;)6F{c0Svnwc_Tom+(NSbaQ|6=s>VnrFTBEk{$}>aw8*jMPw1JuvuUCb zZy8jMg(h|vG+lY`g3P>J_ux6_|J`dQ3?ZiY4_vWpSLa;DmhN^M*5+ILIC}%iC(z|A z^RPF)_ES*E_!JF9G-FPns;3Nu8XjT~uoYe{3GkYT3a-@Xm#8QC#a7AAgU`_2~tYVwAhcKMf@!m2$DEz}axVKn4jDOKBj3 zUDSXjz|&;`Q*{IHh(CTshR@d9|7UoCNj?&7mB8O|EYVk{BSR_)!HOFF*3AMKQx4)! z9~k)a1vQvvKY_sYd8oL7Edu(VmZNvU^*k*lJD}?HA(y~=>-OR@IzK#ud(r8;+Y5xX5 z42YRNp8jz}*Io&W?(byP|D;KPFJI_U@~$!m%%KOq**Snp^9}x&VQHR$My4E&`tXLx6xz5{dSSpc+9XjnPOrsYWQM1=h@tIM@b8^EHUb_dETX=b zZN2dFPaq%q1bg@iF6+QyRKXcZ#+gGp<+2ivI~rX-0!Kts_syLo1IyJCvDM5FYNHk2 z>bgn8!jI@s*74*_^QLkXW>^b(4)aZGbmIWkg^D6{Fh2@(TkRGKUNY9Zm9?_d*BEzn zZD{?{{9>?I28EJb08uYpX&z-5EbDL@p^2z-jvQcHsjBH4VZV2;Z#IOf z?uh(mVZ8C7SVb1eRMMCSno%UO`Lun=TAKgICZthY^|AF+PZu&@Wf;TXASan=`k+vq z3fu$MQYNH{l@dP#WuFztRrHP%iOn^vm#R?Aur$+pG}wp_ODOt+DO5Adt+KKfZpw== z13@r~;XVwBF12dmsxCcbFORik<=yKcFPM? zE#!iv)*k*uo!Vf)&#PK1HFSK1X~^W;>Z^@@_k%C$+)G+7E}=BVC?yZ-C~Fq!(aUc1 zC;8Zm<~{kZs&n!ZTlmtufUr@^bMALIg?LXk43HxxOu}Z}PM8x+IjYnR5_f;~eaT)KxQ+vIbu-GQzLHX&PFj8-~s63Gx4 zn|G-;FM)clM*{;EL8I{8bsXDj6C%*jMlK)G*BQ z-R;?~nW2QoaL87T;xPaeAcX*U8;tV^1^ePKKJRtq=Z6 z&7z8nI!f1j1{ONUVa88Am^t5(Ia2AU;+j8`*xQL#p9UE@S{f|(oT@r7tBYTDs`opA zC4^FlBVb~<5#ak>hXNa!ma11oy%~VHeQ%+cDTZzL$Ii9Rli&JtBd(0vKQB|dT?XqLQ#~npGc5UH3&B+}aYN+W$~T=}KM83$Blfhn{3B z8z36=`aFqsaGCSY@TDSc(->}PcI>tF{$J9Tm^~tp|2miAn z)8`UBt@((vJXG#}MT1Zw%es+Q))?ME^uMHg^V$m>ftY^__AJ<*v@iDD?a|iMCieL) zXGcz5VE3XiFu~X;AYr=PuvbVoy564H3gs zF5L~BWKe`c#!6;FJA;JnH0^H5KRWzl-CNr(h$c`myRJ-f##(-6N&bCz(f@E!-(dImx^x~Ybo6(2s%AmC>MPoC zqPXIS%mB%K1C_t4?C>Qr4KGH#1%~v_h3>AckUo2cU<^fl?C?kP+PY}N5fwzlXX?1d zm`De-{wo@UUP>#VPGN#2VuU}%UygesTlQa5^APe9oN_#-K2!9J7{UswIBKr-s5@Ph zq>cZ~vLE$=RM2_BuPS&i z%@H_=OJhK1$n(Bo$KC2soPhbrJW{EJ#csx6n~Yz`o`mI* z<{lx44pwt5yJhG^domV24$K5Z{qGHhxSIRhnRSPcke-CIpm6G`abCR z+{WB7oR{YjRHro0%U;6U|SRbuY9m~lgM!$da5x`*;_v4 zH$cXOvix{rlENOlZ5=~*m+9BQGsg4eT3$QnN&WC>U${s+zc;=ENI~m1NutZWFq{z5mrKq;%^L{1}P*KF~EX(1^V}sUYXAQoqxmg=DhS1*2@6=al{L z3*Qxmx!JOKbtUH(0!}sZvNg89UF{I=r$5%Jf18aS*S7Zyc7Mgse7P5DmXGkPA~~~f z)vsvZEJsX~ov{-o|L#oOOpD#TO`-n1I3JX{e0n5GDgL?<3O{WSn$t_L?d z8gAcXqFny*@W0|PB>SjoUv-`P9g8mk+Z+vX0sZ!taoxOzb?K>4^w+4LOQzNkHr}1t zsmp!e_O_sw4=|$k{wqWvil>6hDN!}2g0{P10r>jUpNI}AJ2|&!1|F(XBu1D>kR3RD zyZT4Sby?%A;YYIj;XmJeLOPX!lB5NHg9@LYGclrrNB$#Ux713F@cLzmi}m#JcY2zUoMh*Dl>oIKSinBAusbMu%HCaPaBQ!*Z?)-s6yH5ik zJa2O-d1hwce6?`K<64-5J55A$6J}p<*wnO?nW7i|b}nyO6f)v^&JjmFnwT(8?;C!n zkBA>W^MJLkZkEK>D$JAS?m!H+K3@>F`0@@AqFY^NR zqd7A0liU6)tj+qS?DX4VaHe6G;~()P`hPVSe5_)s=;? zJCrl)x9dRI@x|P9>hfBt&n3oRN4Jb85XtA=teWl)T|1C!x%dp2K0ha=xlE7zOrj9g z!@WL2Hz-hN)E#@Ta~uMz@` zn$z`~(z7=tkcZK{7XI8n5H6PVaw>(2N`VHN?am(ME=_xv7QLHXx6u7q&T2s6D6Q;? z89po!>P=|>&3W|S!*cn271M8AGUdk{k zT(=mpP~Payb$QKkCvw0=k%ZF1O`{8@Xj=X`{QYrB7X0dmfaAaR%7DW@50a=-|HC>E zG%nG=gn%PXH4ORgBeMTpoC*d6UK!~m(?-k0uz4A4B|rqr3mXwJHBv9Pm-%xsSD!1P zCQuC%xO^352TJJJFkAE2J5w-J(>b<*E$P6&F(!gO12fDQbv7Fk5b8nI!kxjTH*P3f zVCfa>zj4bM12)~mk(=Q8ED~;A^PY97VmF^b%S_(uF-e=<#(6fEr;OQb>)$`Hi_QEc z-%M>cd#szMHGk#DE-NKVPMyWgESl>+#H_#F)zR9nxf(nZ4QHkc%(W%nF&CI>SXzh% zj^5kHbw8R&x)l7ns+HQpMEq~!GLeITHjRe8HDhAWoRtkXQtjcgyph9c0S2*$Bzv^6 zwh_*3!U9)smp8gOEku*L=O65$u>Q-Bs&Qu0P$nQKTwWx2WZC$@r694p@SJXbv}Cgw zBxE0M_V8eR>?X>Gy)RqalSTPD*XGugd~0hnqsd zsy181^{q8X==V-~Wf{aW`=teK_dUW zaYkReH0M!ZT+I;#TNf&k*%XQ1FYh8?Yq86t5?fhsZe_h1oAi33^Ir$a zp9ZYg4&`TAZ$WimPq>^Kddu{C8FgmsuWix5jLN%oJk+~xJhYS2-g_c=P*6(f)3?Dd zTeaeWy90f=cv28wG^Cir1JsXQqW4{>sqr0N1BInZ4d>#ytDQl2GRywy+6g(j4Shw0 zd5q5F8XFsgmf}+8s)2Zz*&XwTTbr<@g>rR8wn?(Zz=E2rTqsFRF+L^H9!R%Ph;;C% zu|7F9YCOD^f^wnhsqPlUG-sv+#ke1{!%eqD_*8+fgcsRZ z8(Qeh?rXlN-jifU@)vW^Uzis3&0^ySJD)a6T~EuB z(tL6Bg}vc_npmvcUGS^G`k?4zqV|M=%km{W8n^N-gyOUImB2`Ep6a+t z9!c`+Cl{{h8vNxn8OI??c`vV#?NNgM_Qfo|bhuf)aQ`vp+aU|_8+kpVkWzl*U&&+4 z=o@D2ft>M(Z?5@+N6Tt2QgKGd2~#)m(vR+r{e8@f+qCQHM9EGIJ;dXEi_kMVDNj;X zpnDJ5f0B3>zmoZ%$NrA6#-dS@f8w*tK*}6Q#gh|jj^q?qWip>GCScXB_<_ELOp-+Q z8&s1gKJ7OVmGNOWuCJVWbugI&9IaA9NGOGnct8VL7P+#a(i`^Sp7QsdXw58bD?nO^ zPw%lyOiFWtQ(8r=zfzuk2s~plvZV4?GqZUaTom*7SLn?^O(x%0tsq#kkvAIK6t$BC z$pR_sMcnY&&Hng&4vvcny`FPh0$bRVG>;68$YIZ1`P*Wj|P2W zsQ(3vC#GL+7nog#b=nA0KWR5zUjq_{TAQ6OLT5b!j=3_uE8BKH0<)gYO)uMh(71U2 zwYz|VhDG|{k6q%=pFdA9cNObZZ7$h%w5q2>Iq%%iWb_9@R}#Y@6459dVDTsy2nRyL zArkQ@@6d@C0=HmLu-XyfP%19|`gg9E>-AcgpRIx<(rfbr^=d z>pMs;*^K|vp5Fht-i= zY=y@<&HXaDKZ-Ls0q@RvKI)^#d?wKR`pYX_IXCG@irk~|M6NgfgdOLB#c0p4i_d5H zotcrNyoS3ofA7pdIeaG)qBya1cYjUGkUeDwcNr?hQT&b>;6>ZhUpg|O{XyB2(*1i{btaa=52|NQ&j zs>9d5J&NHZ#5N8mgniT%w=HkpZ5Qw(`k-p%71~)Qr~RRaJb-5zq5m_2X(;sGVrB|5 z{1=Wf+hx;G9&_k^Cr%wM;q{+x4)9{-*^XfP2*%(Y*@xh}1i?G>A! zsIZP)RJvg3vBNyeR!`O(?BHB{w=FY{QZyUmQwdOFa7z;M*SpqlE*miNn!^+Q^e^(; z<;b1_PXnicw6|j^eFOCU#IL_2!T68NKGwQ6d>H?8gl#* z-9c|_BenQ>YcXWT6y z|6!EV&R$(Rpn)7pNFvx;d4C3uba?a5n%Q`&e&t_E?-dH$jFPs&c=2M)!f0+O*AHyD zMIP^nB%Rs1>o3(=M}MvV+xWdKisls6it1&(`z2d2flTHsztp}Ny^nB24^FBMTXT3z zQ$JUpYuLBw7xF{4ofyXEZ~OXfhHtn}iFC)EVeAc@ZDc1VsUz}7&6nd}MlP~99n^VQ zH|8&hs-oU|rN2&o>@2qXsE^t8>dRgD^7*%Mck%}{>2kY(nBH+_*RT9v0B(U0Oh*Gz zhWUK+yl7@wG|!=JFxHgMUi!t?6EEe+hhv$$B5s_&>x6Tz80@b zlI-&6{T-4Iko@rqbkaf_CwJOY&06{Rw8;AFdvy*~iR$Cqf>xg$kNZucsp3=U&aOUY z{n(!0g;$Ll^cq$bdX=3p#0v~(s&3Ey1LS}l!mK9*%5hF3D+vmbd+~hr%R;pYhB}=P zv5tiN>(9cQPJrE4WAezz+X{4gJ$>QlAd%=`R&f3I&9JRk5$UCJdn2R)eS zUL|KaSfc4)ZyttOK(R_4dwgsod-T!rKa7*djyUz5L=Cw(m#m(Rm80?yuWkAyBju-CgyS>}Hv6WD-3Jbf9T*xVO&fUnxvN723UquuZ| zzK&#@7x-0u$r;d1w-hFSk^Q3R9rz+j^4UT&@L||{mTt$!FXfdOGhw!atyow|U3n#3 zj0`sWJA*@aa8dVciR)m)#s=W0SfXBG_#Z(eJGD{!uO{t}9|s>dsPGgmtb+|77q>Vp z50b{eewQhE`f=s{we*S0u@jB9_b8iIJnFu*DRlgN0OMMYJk@??mW*})y~)*2*-kO- zd1fHXNZ8z(KYP?s%R1-Waz-m5lfXd@vLiLCW{qZ3M#kQ)M9w%6bK%6^z3iD%z5m{IhmU05gEUoH(;C8 zzMW50Q5rf^y?S_AnMqJ6=8fj&D$v~cbU0?t(ndGOS|GMzsm0~DsLSo-LP+6r3$v&M zg;<|2gP0Pw-tCk?;)HTT%)R2*;}xyak@D*clZhhBK9uM3dM&&xKTs9CdF&EV`n33LdHqoB^x3luOL!~6b4QP@41eoKW@HVsR zo+Wv_?%ZxDL()SFx9um;ww3g5*HZ_Jw_)$GmJ!7kY!4CyiSiFI5B619qj;8B?u#$n zyoSRsmtC20h_O%A+d%PD%cc*lZJ(tpq%9gMXxhVBQLpVPH)&FE8RUE=)?gA8+xX$T6Zg?*Q>$d| z=?}~C^yS`<-_VaAQ8lo>F$;gxBbKlEEV}wTI`S2IWT4qa;FsKPY#J@z=HS}Ri=Mfm zyeY}?jn1KjmSBy%PLj%MdBJpd2ASX-RMT{HoSR{$kt)CX;YiALpgGm8b+n;o?F!cV zPgkA0y{?$g_Y^1#HNHB^y0jOf-hZGMqBeV`glTH9R!)GX4z$FRlf%M-=7wvasPINr z{n0Q_+dbEdk;^0%CkKFj6CM0&I%4Y(Y;I4n>1~@yW3g*xd~6RHvFV(gvFX)0F*EfI z3FK=Y-u7YZikK+9L^7sTW_2@WY}8bhYu{@D=7PGElfXUTGoy>Pcq=fzihkw~$q~NN zcJPgEHS`GZmBqCq->@TLQ44p1Gi8%C0p9xxJwuHf~qGXze4oU|@x8EF!`!ciA1 zcqQ52MI#tjq0x525W+q{w)(>4!=gb`YWR6^x_yE2x+Cg8`~zf4GGuPY3w6RCpgvc)!7AHb?11877EFJ%>KV%9OO*;rq&H zJ#S<@GEK-jHQ@Vt=;*W#3B~~@xVMT2vxbc+xSK`camSf|x&#h5e@+tnOm#;g0zbB( zQ*Mx-k>?tV`Jm#|WGM~u5EYR*(z7U^nHaZ6GKy3%zewmTSW&dekMsE02A z)J2N!NJ4I^=0j)IBHdxGm6h4-da33-Dv2z;50r|v0#e+b6aY33fYBHxYCc4LU!(&z zRI@U>Jd>?~2p&*U6{Utt8Y)pG7KLB$u|{Go{h%MAP<oSCoTN-FTDa+CjKpG`h;F7(l_%*2%YjESbk(K+;Nojt`bex>R_5v?Y1Vx& z>{Rh6CzkI9m%m7v^}wjj^yrVpYT8y~RCTuNSv(86S+D9&dq^mhY`yfS43@c;grr=T zav)t19{-K^w3Ba+0@HliuV!*BfjB8nCNCG%>y@ogI+#aI{j^%7jiT9{OUrkpN>)}K zoPC|7H@OfSf0$L3d-YaSop&9Ysilg0sRT>CTd9P_Ce9Hm@-J-7Xvhk}MRqgHWPtI) zBgFUo{wqrl0AYvDM8a-D3}11WV7H+nC6QWhf&;T!&U`X$G(gHwST(Ke-@;KjAp>!m{h zThoD%`NrgLc#xeO*Id(q3_L?9Ts|!7^rMl&l~n|1Ux4^J;1Q_AMi~Ee%6=gcctOOl zv*`^2K)n34!j7w!Mi_-#6Uj}Tlcr+^`^uD`#f!G zaAt#`4CEM^eXkmcRco;v8L8A6%VjtQi4_?0l^p6ZS?g@#N)voDF@LPv6>o*ZGIs5< z&K#t~hU6dA*<1|$Ug)DWr`@J*B6JD_$x_0i=Km>&;eFzY4Ij~N{yGXh^J8SH?DSW# zrDewHFLtjKh2LhpK%w=I!9@*V{Yq>kMPh+JQc+2it+y(@7^1(xkv%`hC{1Opmt7Ej z%DS!88y6PVL8O8{-L|^;g%BB;{k~Vz$MjZBCw9bixCgf=wasS()E#?qN1CiTf7_u? znF!fnFQ!Ty6tLeuglrj!QPu1j#(s&RP`wkY^>irOT?Yp8+|+t*#k~z6EBz@!$J6oh z^krXUKn>dBQVH`MJ+eP~a8JlydybXft1t{ZkR@w6eg1sW%E^oI@v&mU@mhFNKn;V| zKmA`nXU@7FpfrF+RS&)NAD}0??$UQ^(W*A9)2gZ~(5kxTsFql!sqW8-mSbQK#W+fe zn;lT^LWB3GJwjH8n9=h4dOWbWNqxeP2rUAyDZiRyMxc)W9JJVvw;7xa{ z-nC&6C-T~OC+sVG-RxJWLx7pgYF+ou58R>QWmhuh(X(hR()I!T)Kt-a@9(o>f}}sC zDfC!D!-=W>c5VL*F++eRvR^*UdeYLLiCkn+Gc>~N`bw7XX(3eQne6QS+n8GmNl7b4A5QJBK>i) zXwyq)K=**YAXK!ZJ+(x$lBN%TO?<66FKQuuvSA-lNI6&X$aqv%D1@`7PN_E^zUq{> zxeii_MZ5LvgtX<7+6yt$VCWrjTXz9PeT8Qii*WS5?BT)0VkSlZ+ONwFJ26ISul9VL z@zENIp8LFl+RXX|EC@ryOQiz@a+w5_(Sdcfk?|R&w8!(7)L zQ0?n9&xvM+9Hg1$4NIZ>!T2lMa~a>CJheKUCaui!Y1xW-QXxoWHv)1mzrS<72q!AV ze1p7|rcmVb5PMvCt|u5S%KjL=l29PBM)VDw=B{deV>Q?Ea$zx-2)U%0X#2UV1JA*n zPN(o!pCPRTps<5(%tnod@RdyM1eUNdCgMwm2)6R4We&}0v#{+R8$Z{$biy2KUUe5VD1) z2{c{aiQ)7cujWZdBPShFH&K8~suA4~RX?^n(OLLPkNV~YnbV1D_CUHVVmXOR!M@XN z_%e&LRxyqeTn7FOD?rW*aC$KdGx*TiLLq!=-2SKfUsiQr0ZOYIH!)Z~dUq%4K~^4h z6*DBG)DsHTZo$H`3Ir_<%Wg+JDw+r7V!H|yv4Yw#q7n{^3kr^NPTz&yPl41UCf^4p z&!mRR$<(!VlHD;*M4fL9vR5tLh@q-agd1PmUMDe)5Gt;*)UNRp;j-T2vagrlmE{R; zW1Xb-yVR6e6?X;@CyLFjTD0rpc)1uHO~iQ3y+NVG@nW+Swlkk7@~>ZHMAqf5WSTfJ z6EZv??ef18H1aMs60;v=65`=x9DUQ^Wa<~9fvicf4~f=q@Da8iJA#*>rWO7N(RjeT{6!%hh&H*Ak%maExdJuf7hT``Bz z^2hmAHcZmax(8}f0b&jpN5jxnR6v8k{p1>FvE@B$GeYp@G;(8L0DMEMHDyO};5Yqj z+oZ#hG#??hWv-mG5wWFyk0NF4!bX$0x+ z7#NTa>FyLzIwho}I|Za91i9z?-TT}>7})1*I6P;sz1F+lckR7nms;|i_qhb;N@>zC z{wQRk2_|Ezz1%VpY%#}c;Z&w;3Nfau9ypg+;P7#_uD=yy9&jeF(f;<@gJA@}^LIbr zMtc(x9u1fZlZ2<81j*sH7WYa93ebfC&H6MHXJ!wFq zG_z!zA-~2&Z7$1A73TPm%tk{KMH{1=?@V&%(0G}bz%$5yNCD?~C(YAhizo0iPwcZ3 z{1oOq+JnbhvmJ2H6$3K$om4^8aVP-R44hD?fOGBfX2BCgh$hZ|DB1_+(X};ce#<5Ulio-rmM=c>P(E#Agu8YXCSSLW_{Rbe7$4J{(Ky2JbD{9mV$@&#_=>aQ?2 z&p}E>lU&P&z5{ITk98(;;Wt0Xxtx;Rh}L+3UR50W*BST$@l*E)t6$9*`!*9{YZdk2 zOy)6kLQk6}uB!h$|7f{;l2^eOUb~hFhQoc{M*(V)+Ylc$foCiDMN&7ZGin@9mWl^) zuL5I^qrhd;E$cD{2JYXUv|ux-(|=)25@%&Y`I=iMXU~y={Xg~1-AE|LKdG~C)w49X zk_^?lQ+ty5>pUW}W3D*W+gkicEL+YovbbmP$nZXI!^gtS5_jys2sGSD+Fj0q`D(*I z{|Y;hJc|cm+vfFpv3H5B7aZiz-Mv;xIq*!(aOl{)ZvvRm#&XCFd%R5I&=rQ%bMcVd zW`p*hck7^T##Mw}wevcs{lDkUlFbu!Z!+90m*1H+;bRhOvW>OW5bFov;@b4POa{S{;D zzLCmq4$RR-iu_(5AQ6>!_sfBnoMqzlokGfs*c$4n)p%Vu$;NV?TVudYVW`Q~bfAmB z&=ACca$a4rw0KFRc^O~m`FN7W;8qe5!rQr~=bJ~W`_9UZ8gW)g(;!qWf{3vvS(X;- zNuCj#25_6d%oo(j@|_7T@13`m&Z;gr(k?E|1VJ^PDwD`P)k15)=__vC7+B5$OsKgS^Q~i zo*Dj{_~cmjQZ@@OXqNfds#a@gc|RWhjvSZZ2Hw}`cN+_HWYZq|TNTU2@J(>ud{$@u z=d4zNBkaD3c4*a^M^z_>)Fsa>>uOj$SS#jM$>faaQ*fK3=2v`v9mPrafBmQSqnB?f zkEpk5Yr#=x$%75ORL6W=!qb_^Ms?1?yfJ%R12Rs**8^^g`cp`sNc(H*Xs>9K-!l(0 zK$aX|BrCZUHEcTKoExg5alPGDAbZwuV#ctKmmF2R1$+N{I(-acjG&Q1 zJdJL*L9F2yk{PVH6Vn&0ZKy_0-uDHKsM>Jxzh8+961UEQEG8CE#zNTFT9K)xUZlzM zckYFULAF0^2 zLThv`&duNI6Mm~xgMhne2B(ipe6d4;lCpcI#I%avaeccq8~o?^uP3Mu&GWJjgZNpv z_aHcjnBBCCY13?QeVL~YwVBPR5#zs_=|B5*5%C+Ge-<9rX{I@Aj5HQxzsHi+Sdjjf zZD{jeF+(+_y#C9{){jOjvJpeJR0<&qh053&@+c#Hk! ztEX(n6RAN82X(=Fdy-{QEQL>_p#1@L~yN-8L7Xd#Y6%z6BW%#t4gW zz1DKuN1ZlkfD0!dY%4m(PY6A3e^N(K^$nkH^UWXV99X=`eD-`&bvi9>sNu>7ku20U3l4KOzFxw0D{%7PmAd{bf z-u2pmoPKF_gTSwLK#r@}puj|OY-3UOQ+2R6bJa~1KKn9;U7a=ZR_FL~jh@rnH-7G7 z<2YEoQdQ@YL_}_1Vvs?9Zl(T>l3VlIl4au@hM1@w6(&m!{*)q`c5LwHYTs$?bVmK7 z{mRNbh~vdb`7%fl%TvEP<#+Heg5RC9A_ZlY=F>kQyoViN$~2dhjx;|-tVIa@abx3G zOfH{iZQapo0U_sU&l4Z#(Zj~ zzlBJMb6rb-wuO*G{Ij{xKgtpgx(6`2%2Eu*G%Z91(+#e7^_<_yzIQiziFvJEokkqp zLF53k^)&jbG${Sq^#IvZ{9QzRaP2{L0eUl{DfB?LZa^J@=?>ENeE)Qg9^^yz@%EvR z9_^IVI|4r^fH7lU0R`j&vCIyStZYb`Dg(PVJyGL!dP4z1dKPAZ6}%uW8z#I(%SJf0sJ?*Qx08Dpq4iB zG6k}5Y0U_FMBiH|JEfzGygoYuq6KWAzGR{w&JM$Vlebu7}5za3lsb*le5oX zbhfqAPU@nq|J2(aBS0Ac0HIgrEcNy8kI87LOIgw^{j*t|Bf<&%=Q!BPjwPKh2WgXU ztU+ucjHMACL83w%HiMbM1_zf!ldW7}|Dz5?ULZZG+Rymfqzbo&b_t7{qqg^R*G<3PB&dpb%<@ z_d&7Ejr_|uH4qmy+x_S^N55q7I!u0auv|96+WN@-nEONJaX!(cXtr*kO^Y`?29gFQ zVI^~IhfBD>vnB|yA_4u2>C$Y$SEJoI$VEOK31Sw6qM~F5k)^#F&wX4FksTgzcYYxT zdWTCeV&m2~`Pdxfij+;lM|;H0ws}ee_@xQbe=qIIeu}(4AGkutvLG5c#MLqec32c= z(=PvTn{N0u2UbRMi9Mf+;tPw=Z0zZa5-3L52xO{N3i-n-Yk3e@Lkssbxqb*30V%dhJ#*BJ z=43gP&soboUl5oMiTU|bo1CmKQBnKfPI7DMmG&+ohy*0+s30@;@=*8<0x@6`RA$~7 zLwlI!u0bk3g6`JmFR(6xv^by&2Kfx$Puy8mFWV}8?b#l&+r5s4y&pfXKoS5aj;{;K zJ6mt%WB3_eL}v^3rHQSZP9JPP{P|;<&_?lFR&!BrGo(A_%Ww5^XH#77BlnSI(P}fg znPtaNiP)#XzJ?()ToE42QW5sUyluWd7_9gBCkk(L=266B6X^X0`1=Aj3N_Sihc(9m z$6YgLD&C{rJ@&e&y2sEQI^ND|#B7FOm!uV^Q}Tl%`xk^xsieRVOX3Z+;M%*;KM@@I zCO?T!#$>MUJHMfdKTM(S4p|iF%bw~Rp2*uem?os6I9R`#6tgZhGS^7r%!rGvIg(o^ z-t7^M>S3SAqPMo~UUJ=uAGj!P)Gudh1l7DJ?;><9Fc!pn$b5Nw+BPy#(P5~iU zEa`96hCvl*w*hVKhM_<@Iv3KW9BS%X=3L3d5u3rxPd*NrFxjYl_2E>SBVx!tN^Yk} zSTvtveVz7t!9M-_I<4nIOoWKWR&{m0!LRfXxt;j_g4DNdttFwN3#+02?CC$kdnzkE zOpbol#5N83326$0mG??x)B~z<>ee<#i6t6OYO4AUM;A0@XH#-bD%*&_5F#q5E+n3# zwG>q?xATdAA5#1=mJTC#ezj$yo-oKka=`*|8?rri{8y%>=UHaBQT%k40;v z-S13AZ*pVB?+#>^hUUa?rUfKUjPnV?M7kdCw%`$QSv4^=)M3{0b(-|i<_K>a90{%k z^g%@1!LX;EW(h|Rk_BRQfFT)o}w>>=?OwCg#c7 z>V&B-z1(ty?sHze7T2HbniLQ_A>3q?RjnGQxfJ(aOfLD|KMRge!lIxT_b0d7 zO#DpK&Zpe)OQ;7b@bM!eC$-t<(?BX3Z=k~VC%??8i~-IN#N-%Xl~CLNQlr5;DTJyn zKRwF@2d4tSv#f>}w^mb!0UsUMY8Lkz(dw#`CO``1G^wQu9tYT)%CV0cp_T69vIYhO zVcIQ5wAr%;VN;ymgyYjABY=-cg|Pjhz(g4`vaWWI;mrQ>;ara4bT_L(|5p(!<_Y388% z9tTcwb?#8E8GH6k?yb$Y$^hT#g=8sbTD&bz2<7~Nq$GA^IY6y2)pYnOk!N=9 zXsHgx8Pjp!_?#vgzNxwOU#2mKJM; zz}VH96IL(hqB|Iuq_07dEJ(T5pg>Y2e*GOK#2MQt*4wOFd~hLN0jM~_?m!fS@>T3E zTSia^5p*+W(GhY0*YX0^@%~ZC1>9QttPiLn)i6_p>b_q29T~kE+RRw4RLzZUjja}@ z`+*F)MoXlkvz!A|0VROT$U!yP#!XuC7bR{rv&lVwDcKQo?!Cn;!*qw#Fb4qF{LuZ) z{VH`_GS@1*jByoG$M8Vlm0N~_o9X@}v#la;BDq5=f zR{!tz3Byj#M+hiudOits%n<3ZKgqT&sAo0<-fM$iK<#qsBiA6JY@_LNt8rU>{B?wh znOU+w&yJI`T-LB%8w$W_XdhQRG{Z1`mLpjjto~(e`LGaZKbUQ~oc7Hgi#nsq$}ypC z#;QlN&MrncO2Hhy5%F>`Ptba0XQLV6P-+xnMocG$TY_ujJB1q-$p3cnkN`_ho4=Gk zt#+ZyWtMS8kmgHAK{_tf8zeJKcZIb+x~K=AVKtSu@whA8N@{b@yJ2MuL;zHxENrG1 zj5W(yP3}h;=65#xva0%=!<));-{_+9pxJ!NpW`G@PCo5Nmh^7ED19@-N7A-$IxEQo;TszVwkyd)2vRmJ)p7SE@Q5#C6B)*7mzRpD*=-VSw92Zn za3HDlE#`H^>QjYT@SgZT!k1?O_K(X-Wk=6=onj+81d+0O-;Xm00U<*&H1U(Wsmgwi zHNg888D+}%zOywkkIqbxvz%VDog5a9Njh$$_>b&2ow3qZf2`gfn@1srS>UKjC$nXs z%o0WfxumzQ3cFM~2U{KwNxVhwCdPFvPO3OO!vV-n78T}Sw-VSjIeiAukA+KXLU`ec z+^hd_6yzreSewvP*a@e%)-^+ce=u zAY9oxT!JE^BQU7J z|AGA`w01wy#4(nO2@{q(|FI}`B*)|h(E2Pv-}U(pm8qMf2-x&G4(4{|Hi*MC>?!2F zm$w~jk`2HQ*&Y(Zu2Usv4S};eqLfTiw-w1s1&=9?+PnCXQ4amI$50c4ST108#4$yX zo8amdf4U;dOk#>aC^d+XcRi5$7PK1$l`K;^tmrqDUTLzVgVA$;Q1|nZ2g&Y204~x^ zx!(ucAWS#nEZ3r6(8@#&42NJ#5hbl^PE!__H1nl$5h71xv&@?pt+=HncqL}~)Z*o` z2q;cp%HLQ^YGPOr&LxxcHeSz!6YbuKQ|1-Rlj4U<)bRce5|tuOy9VG~;tq|3u63^! zjOeg$*T|}E&a$;?_tR5D(ls~eHdFl2IozXw3xgv$3NJR*?P474H`A?JWp0J$W_Y2d z67$Z0;K4-^_A0&Xg9`v{47h#t!U7rpFBWV6Q<$EpeCjrGtOf~RS<&0I83goT@tdVJ zeA(1DB^v$P?cB=|LQ#?QmVd)2z@jlFBX

w-Bh^B$+x8v=bT_6e<>!oULy5Xs{l> ze4loEf2`@OtGrB0tB(}6HM`}CJNK17qbEsoS^{w zm#Hv#D9Nmn+)_&TlX!gXR$d3Ys(NgPquw{Bog&O~@}=FE%2#i`Zm+UtMO}v`Lhog- zI244ckL2?feWhJGgZGj6f$GtNv?$UrX};>&nsL10(d7QR>Z7##gw&1gi)nOo!pMCr z*=Lcc0ekT+Ig$^D&4sCLHVj;qWl?SIhgpO)3;?qIUS^sT2>{!GaAp##9O1*=jJz2G zD$nh5=)?PsG)v7*bhd8V7;CGfRZhcS&u+~+(kfaQ@KoUd)yoTa7X8veYY#cEkRBkG z+jTL76sEFDjh4#R)Q%MW&z$~I`jXJTu>qte^Hs%syr|%Nl}4LMs`NjkWYz|oceW2@ z0GjbDJFip&YRd0RC3*lHnz_zE>6gUPTeI|^f;H~0fnBh{+K~Jn#}QWOq7`ICWEyCL z6_aYeTSMvM_GvqZ+wd$cvsH{F%RWIU%DBWHKO8lggdM?z=H#)A092hc$3^D3ymC7r zwAnfg2B=pU;n`@YbgP~GHfmY^3pHa2d(1->H5n{;Kxi{oJR=j}!uzs$TF|Bln zza>R8H{`n}`Q#tn%R{ptVgB6dj-+;9JgBZ6N1;E$UI~)W0vcUgL`Pi1$V5YtxUn_D ze07ah=%m@%{Mv3z7zD|0b8AiO;7QV|@^;^5{HlDn7^3381hp+eC}weEm@yes-ss z!m;Sb>u?44D$9?an2WXStkGvWDGk@Jv71SSivuGSs8mzlMjXT5!kK8Sum+G2Hfmix z1X@sVey>H!T|7YWf*MSqGh>Hf!rEXse?_6-1%zT0fw0aQIY5||MT>7Q&ABFdVT^Nf8A{+Pdv zT8I%gTU(_1VD_@dxiL%LO&$7HX&VVl;V6W9a7Q}>Rf|QVqqRQt&29F z-6=yyOvyNT_3K+xv3B!7qJ?rhMtgyw6M5>BfFoLV$qUs^Z{7x@o3ueN=V*@X={S5p zti67tP67CpqF<#~X*O6%EdR`b67rG7(A6K*p3E}99ovPthQ(Cu*sWu7KkZjhavP+! zlVZDV#K%(P9aRABq>`67!uJmE!8}rK$7#rX=Vla)a}aHzKMRA;$-hG{X`;)fjCsE@LXnz^;<f)US+$aFEaKO5SqvQVZ+#%To=j`Wu&!7>LbI-hM%wAfBJ`C%_#; zoyp!~*j1g+jQ2paKkU#?7F-OyndW1$f?@|#BIr}attO@zbU(AVBV(_PAm38qYH5h$ zJ#_8@uV)qqn0k$!+yzk=t+SgS4j%}vRj*;RB$!WiVl1_69sL+ERu!BrcUr^GM1Ql0 zGZoZ4*c`Js<>IqA0NV=CY31_aw0>rbjaPs@D_F9R51u@9HMW-l?r&tmonN{Bj=S=P z_wWunS9t|OFI<#HLDLf91Ei!<3$IH&jJ)+vjOSv<`4HNa zbd7C!sqDsNU7S>YBK-vyk1DkqomgvAS@gRy?q4>5mY9KbFG))dbeOvpAVHHjgg}Rnqs3^8uu! zJ?LmZ3A~@)IR<858KbOo{8?aog=?o6hyTPjmWS%DPkT$8^Uj>?>%%m_v^&SJID!NL zB7XzN#3%=f3R|;o#OMQ%bwUKZPTf)Vd;pLR5-a42TAEOJ!QcH+J9z;yVwRNM)L?H| zcJWTknRXIf&9$!Qq{RR~a~{*ZxRd?}Og8-9C zcNz`Nit0%lD5gSX^6Zj6h2C4r=md+p!Zn%2UKRgs3n&7jix{e5@Uihf%9}KU}U_P}nPIl_%zT;mM0aXhrhFuqRh6;c> zz;}W>(yYJQ@V`p(>l1)Jn?|a{sBIAE3CU6n6iyf4hO^@*qq>cRU3=BM(VzkS*uHzk zwh}4icm^M`zp;9#tx6falqGuLc{^%s*TyrHZgAc@I6_jrMTTl*Kf)kgD_t8D-vCVr z9Xm%N+uE_6uv<%{))`PYxD%olJcHIge@1+w-pw2G+v9mE7^)b?Nr7b~ZIgpI zQW_-ZiYBUpsq$*1!r+#qwJ^Bv=b!RjZOtIFYo>cUMW{O-WDkPJ; zO@q35&8fq^H9{cCd_`4*b?s*EUS=0=q~CwBCQQ|yXr?V1gLi%y-h{-qmF*$9U}ujE zKal?d7>if(FqPLO^Oox8o{wu?9+2Hq*lkrxV5x->P&>2h7d~Xg^&)V$%KBDSB=vOr zEkfY$b__{b%=`q)EriVv{b;d|Ra<#+FEwaOYDNSA4vWfk?LCZ3x6#@w$)&3H>FDBh zVD+4S!colW=(~tWf3I%y_CVANsfO~ zjr>YgSUh3lVpihS7lOka_m6pv%KpW%ZKtSt&xGMdRj-01?2fpjagV*!V zd$Hadm_UcBfgkM_DCjhXW{nnwP6s{n$N9LQ$7^GOCOh7v%%d#ptx^%PJ$tXMQUWwO z4coILgXvhqNAyEY2)xuVn@sN~Q&2fjhr@PI=V zVh>wrve?E*|Mh7R=SG{(Y!j_lN~}ALm@^2NroLgg9QsWR#@n_o9e8B)4<8G~nC`LO zl2AN9;R(HR3Amw+8AcIox|dY~dHKn?L&cx?VU1w5I;?X+0&PPu#uCQuv!D=Pq%)%7 z0TX%L#K(9}P)&xmtO|Zz>MF=ANVr}ZX`t4HF6kUJf+j}#M(3jRiCH=s8R0+U;*WuI z7+bDy2njVO-Fc32yg^zeB5n$%jjd=xvl?(^A{{RN)N#5(-og}D!Uwn6|&7 zckr=Mx639e4eDWCYHCX1ApcRk;|>fB{uTGXXT4-cClf@yZ82O|+h(2pa3Q*)adr7U z-P(4u)?O@EjZIP$xtr34*>?7pWdxI7WaQNrh5}zY{AR%bDT)d15WF6J`2u?X!U!}B zJD!zi<)6l?d}bE4$UQL{)bw&`1@~0Oc2u@oko#0_pHzOTyH4hfM25KDC!y?b@LJxPCzIio)8Ey0u0ZSx|zLua@#RqC)D}Nn37vQA|=m!1}#DyX7Od) z6bCgybo>h|JwX&`-0DNfDDKAbHW#R9n|aZn(8D7ap`iw_0{oGDv7py%9iX~9g98@` z+>*?NOTH~&{S`-;5zF5-ygTH-=XuwtxDZ6d!Kr8Da7^dy;=@-^WT&puJ|vhGT5^xDO3#FUx^Yz)Fr{(zVs+_ zWpEeh|WTlqIO?e<38BdYY`4n}DE(TlNbli6NF9VUB4TVqP{ z(CS4hBx`eC**vgg;}FuDEMX>Ey&>zL;k!yJ6n>|Lx!9;RF_%GCx2uK_lyzNE*8AWE z#hZQ7JHyLyS2<_lhuqz7q7T0}KL&jG&4ZtvH->zxWZAeD8?Xk!_6G6V6HUL#Mv)+y zH=x^mk&^NI6fjl(A_Ib7xtBL&I3K_QlURQk`$NRFa`WVoso~w$%rCzucMfY_=0)cB zI*I@jngKiGp%U(u5a*bHD5C!{adiL{H36%+0PZ6;)c=mvIz~mC>q6T zVcJerhPbw0nF@;r9hv_w<7`W5C|J^atVY19KT`4?kl$E*se~{=y@D~60*;&n&u$Go{@&y^oV@&YfRN$GkG7epEBdEsuQ0~Pa_U*`IL`RBHo91-fLX7h z`7L7^+{+%xSZDw~8yu+=F=}U-U8ych0qxbKHf~v1D;}}BuFN)EGxz9z{|I&R7!=<4 zP}Jcv^qIVhEv~@kq8sBY4pmcd{lGjF$>i$Go1Xe1KVu(yy3}TZXqWq?9tWd~?#N?i zd?LSFxJ-KFjs(^451oJ75P7+UHgso~6eJz3(uB@Eps>7VW8Fp_^8~b19?= z>IU(J4Mdf{(^l_?mv_Ov8!sRbc=6KIYK}~cT9fHrNsf8W=z{(`*@zQMJx|3(*EKyQ zcEEB^!JRDje}g&4poaT6H0Kt{mAs~EF!TefF#Q+lX%cGOiUFDqqO2v_Bqw_L8Oue* zb852!bN1_5bKlfFNGe>~c(qVf%HH)O^O#2+_JBTi$KIvXPgqoq2hpfgWpD!*_j>6# zxHt!3cD5R|rP)6}TA9Wxg->&H{Frh^EI1WvTL+lp2FFKvBx4J|5;sJyiH^;rc;+_g zW}(*MK%rMq@#XCQSIBvoSM02)`q)#y`pluk7?iuLd}Ta8R7@Id86BHTluO%3YF_|t zuZ!*>^ZH)1!Pvvay>rZvDQ@D=5=VNDlpB3~o{Rc_d*sDP94s8f^irl@5P+F!9@lfk zgIw@DEz%}&y9DaglWVCYi9a>rmQDNcr|B1z_f3+N#Cx~|q~Dd#dE8#NxFl#S!%GjB zH*m>qGlY%`vU%-`Mgl7W3A+NW3Za@?os+-*5bM@6N1kUU6s7oKrtI=ie_#1fY#hF_ zWm5^x=epOQ?PZy+QJ;sqBtXB z7)O@C6mKMjYK-Yg`2#oZ?$Cy_2&%PwJYN-X|2=k+O^8zqRk8&?9?Pt;a3Kr|3WH5R zn;>fLs`a7{z(2D=#CyZfI?O7JC2#Q7?t9j`_`6KrTcPiQ2Wpxr#x3e6H*X9vgmwka zH4FO&E-epbW`=gX{Jt|hIOCyCyUjd-Jk70(7Uk~-@6RwRu@3^`f}oko$GmpV93f** zrX?*3IN<7cJsJP<7u$@p8a20mcb+=kbnjo6@5{Z-3eSt*qC7ZaC}%T&;_1FXogdhW z+>;TM6#w#m{O3P3_Ej-dzq(2A<9&K>A+r0`d6Y{%DXmTp_6I&FY1}i&j#q7ApfH$u zG^_odP$3a}G;4{(6$~YnlZmwbAyC6${3D&!nee?*z8ZzgLq}2Ff04iZ&`Y@4^@pOd zW4Z9CvZ4}VAv9L0%@BDiQI!v;?d^~z1a=zmtvodlY<}Fl8}WKh?Ho5lqQ;+@%4_yr zUy1gW^1NnmVn)O_$AAz zoL?^jHM~KcVD>XjN&^8A2EVwfOp>?+B=citmfS?zm2_^ksll|rL%!PxzmS=$#{;&D zw)D5j3ifvK3$$TuN)(DFf9p~@a9zvcQn>!LVWaDhb8@;Fj&U+i5TXs9Dako^Q?(LAGaj<$fZ)6|H==@U^Vq-J$>HHN(rJVt|0DZ>V3vcTrp1qfU`P1B0I^iCPmFos14&f2w832xnHQPd9MmO8n80Z|L@wdmpO>Mn8^?&BM$jeam~1XGP&^ zC18u}%ToHik?plYWH#DZMqzky7tZ)xSkugU>6|YM$6>ESI?MJxyROkGLG^xF3SZoA<17Xgb_{^+ z90n>xJy6smJiOEdyJGc6#oGdN?<(5+ChNcI##9g|k+;(0=`)@dt!h|5 zDW&e!-ga|FF)*DaT+ab*Wl;cXN6 zJ(tOkl-8s7E76kF%xufbt8Wr)yqagD-^RZg`25DF$V&cBNSSs-Bce+ojdiA%N1-M)^3l07b7UQh{N4#4YNw-@9It;VtfYF05{A3Ts*dc97fmgH zm+ti5fzMyd(>!un&_Lnh|QJnA`YbP~h(MMvYPQHzlBOs^i zLB1BAk&-i_xnO?t2a;U$f!#ysk>g->NZiD{T~e`bR6<5eb(O>fcoX?^({x46MXo;4 zum-*aMWNmnn(1dMllFX4i*#dLpX|fWG!No>S;XF2lsba1R6`@l&JGkCI$hghU`=7Y6>#+!EJ7ck3cfNFmr`MCUegw+K;A zifDppo@gnezKVyfyhmU!LH)fM&`0A~F)jXhF(V2W=``wo8^qY9UVGituCsbGr)CZX zwQ^2_VSC-%4o2E?4(qKz`e-VXV%Sw6#VFD586G+eoJGdYd%jThvythEbXw|>#Oh%t zq}e2;XA(kHW97RZyrkaXv$)tydR$*{rZvhkF}WZet}h>Tr0cL5@*laJp=>FVPgUx= zUuMTKbOxudb!_ons+wyVf8**~wZqL*3jNs>AM#wgC|7F6NLlT_9k_H}PF*6~>8}Q| ziLjtKC>wdEczPZSqYZ!(*9ftWubeoI+c5#tlg8+}=o5max*3_JctF%}&%D`ST}wVFz= zORd`7t8&?CXW7Nn&9l0YUhzjVy6i#gS4$9Sd_%qyM<+|BWKQyFrzei6n$T!z_vNx& zhMUm~y|Sr#j}~IjVjV(DSyf4|dW-i}S$e1nQRp~WzMQM|`0orVogkNSSpX{h%Fg5^ z=~;W}^HDC}7OG3LoNWv}t}mHm{Ca~7k0UQ2X;^FS|12Yg!GI^Rq#H0(XT z*c0UQ1j7GFMDmu$yD9noN|NAbRDatI<0%O1ky&Ncccf`@y@fDT>yE+T`Y?YBs;`1J zh2FExWUr_g`psJ0WhWaKw2k*5toEx1tzl)kWXRJJ)EhCUbZ}NL3MA75ts_G(M|+eH zLcC~|iVG(UNCLc>ptlanjo(oc;vNKk4LSPYFZ$3MvEEHezxq&ND{>XKQ&4<7okLZB znO+HAy(U%4SGP@fHB4@i1adJt(7B-ljFiZBWtQ^$x6iqZcn)1b#T2CQ&qYv8)eRyS zRIa8^+|iamDiw7Y`v-YSOTSSH(rO)oOrwYEcPpT zf_z^4%CR?cALf_@VZ*pabRnTUF;{D(^vx(+xF>mE!?j~VY-KW300URwUwuDYdy;`C z%8I+lC_xKsg<5q}aU8^%358q5NgyW5UjtzLfda8+Y6CwkEbyp(7DthB1e?Fp=<#E> z?~fH5=IoD78b+Ok7wNESCEP{U$Df9G#j=Ev9rkdXEV}xm`Em}OO^26-p-8KiPVIcl zNnGuIOMfj7II}e%M%}aO%wiPq`+{1L!+US$oRPlzc6)_L8-%D?$mXC7bnYfd{N8?7 zF+|~#$l=^z#xuZKZ`a$=T_|wRXEw7wblVB0&cMSzHDW~ys-F$}=>mEF;h+tD4Ewo) z(35?%AA8s7EP5>~=Akhq{LLHKCKjlXTyFz&jQi<(u5P!xCaShMV1rTAdyvIf?a$xaDQ;qpm=DiX{qA^)3zx zF4g02KJ~gfd5=VsC3oI2i?e6N$;>`Bat1CVHx|;SApay&fEEK=gExg#RK9}1pF0Oa zdse)APhYeqRe)S#Q_ip!rJ8wIFPbt_e(Ga-(?&OJ!<=-i7GV*^oA}2B-F1jUnmE4Z}__ZpK z+gV?l4dqhV*(005mh--l^@p15Z->BZ)dss*Y3+mCz5n(F(bvYQeSI-T*+WOT82{QZ z-~F`@LZ@oP6B4>gdew3DjuLWj{uNj4%DA3r7eRHEAybOeZ%>uMM`O#HI-mQTgNw^T7Sg(|rJR3ISSx5A zj%^0iYq?P(96P}wt%`#va_vP)cvwME8dbm{#U6#=pdn<>*-9L11;O%x;IU`iJ zonU6Sbpq#tU9SfEv$6amJoDYG$rtu8GWu{tB}LkBcM}H*gvusqF?AnR4z$GU$om z?3%3DM2%%0)$OTp8}*`5N<)~kjqV)0Sn5LWx^`E*tw;8;rCw7l(=vJ;dyCod@D%l) zy~7jJk97>AMkYM6@2%{XZfrz0Q)f-ko=srXwb61Je0e#2SM}kf7sxsTwAyoDF8|?B zuT{qh{Jh&xyy+-#&{@2>ltx6Mw1Jr5mi`RBE45bMQv?{d__4@xp3!(5`d4#S3vhz} z_@2M%gooMOxPHY9Gs2fRQmL9?bU(gIQRc%a0G`)w5Z>&gcXNtVklD$n)o5+4V~p_d zi>N-~>c^p^vwl25>CU#1V|WLhW`|OE@{EVlzx(;}0akFHBt*csy!QDwyC&#i=2fFK ztjq*tNStZniZu-W0}vTaikt&3F?NqH*_?ctq>F9IT+%tZ0bG4hM59+N*z#Hn&zad) zKv}an;nOG;|BMKX?d^bKwRE_k_+F*u?p@4P#x09vsEpv9Hv|q?p;7t@#8=^a7dxkz zA#uN!R>og=<{mwH@mVPROG(aZ2a6QnD){>s6b}{;jh`7EAz@>Z6@mLM#(Vw{xHPGY z@I;$cU0nrK^< zR+1c+aPWwdZrRA|{BDXL)ndJ$8?B_1;ZYD&?r@A9qakVl75g*WgA#k3k2fkz%$*=4bz}f2<4^{8t z;=B=N%~!`E=x2*$^m^7B)!cF8(yw6@ug($5yXM)_XGo`Gh6m8-RH$ZZ!SytDa@D(- zIB!nW6V-7NCkeI?n*Lmw)V`WA#9nUu*JR|y+B#}IMq4>~@SZ}FNn!(tIHSRPH5`J- zA=wJmZe&@wyzd*3E#6kQ?Ib!i;>acTA)P`d{m2oXssHIcsl+GX_eEja`jzN3E|h}ov)@&y%h+78egTi(0{@mLiM6pN)!L_8PX&Fu2v|q3`P1;n0|>B0Q)jH zWGGth0~v2VdiF86A53(BAr*8u>sBlJ5#-_sF^!3;iZJ`7P_obUa)lkW{YDS#>4eeoUU zaf1&$;sls+Jj7*9IE&7$oPzdIhWz>mp)8b&9=xhqV@|f#>m-Ae*=``xfZATuB{EPC zGw|L8m4;99)tl*84DN8r^RCgSTg%tA(E>7UIKOnxUQ0jV}}8$kB)xMU`9h`Fa_iX6W4mMg3Ot~Bt?wfcWLP?DjG3)T6^$s<#} zt3{DoE9dl@GZMm*VbCL9{XuI%? zL=|l$)D*hY&En``i8m@42fY5vlcciHUs7({kM2V+x9ghcAwyA;V3>};XZcagmNDxe zLkfDT-=)y4bf0pT;JgtT-`DBdEav3y*p;ni&%lH>&fq8WS&GVuQ)x}P{|}%6E~>(FMOh)89o9 zqormJ{J+NA71UOEEcN2AyQ~Ij84HBUWyZ*A71I1W`VW%lhFMSPJAqLt(N)2Vn}w!V z%fw)qkZ-+L)%=SfJFF-V`AyLyC6-yPKFz+o+LO^;`tKW?S-V2TOev8|rDw4%`Ka&A z1llZapR=pB5$noZ8H8}?&%Gey`uDAL41kzthw~l6I#O>+l==ES-U|U_&T%#R=ppx6 z+{(+MfOArMd1Rg#CR~8igeV9x0=c<*0t>kT*xuo33$O4;vOqXXKonzf5L&>fn1w;z zig0F;;{&d;OT`<Aa60xh{8-Kt96ke==1tAEAG*1Cl*J?I2Zi#eLKoCr7R;ozg z-#!l~uN5rF)p@U)MdS8|Wa=7Bo;OF{`rTgFmmRUIP85xw{gIDRZYr%I1HFU?Fw+Lp zPlh<2Ok)_Z_BXhfa++dW=SHluoCntnkk7!3S5HUp&a#n<*aSTRxt;~JUafYiB6yw+% z97y*X;fK{rO4nEwXKlA{u_~FA4$fANJR(p8-QS6~dQvmm4%RkxVfQ#@guh6m%rTUh zK+~gL9~|qOU;r)NIP`OIFBOn6G={fh>jY4-1i)#ge>isYEvFADTlwp(1?{9(;~rUf zOx~_>$pNi=EyYu=Z9vyME9_JG4V(vTIL_3>AJjIYF&Tw}6Dw{+nb+dhey&NqTfwoa zqgum6a*Pa=oMt}sWjhQ&39TJZqAf#f8P`;*El4df+cymzO{GpseOD22xQUG-aYc2Q zM>C=DSpSF0)`Nws>h)+kzmP^;D*v&mP#?*fS8?BBFN$&Lhmt*T`T5N-Yl@|5=Dc{c zG@grpG&sZo_%b(5z}?2L=^1gOG73i?Nv-pN+o)QjgTM9w7t)o!`HCEx`+#*booG$t zYjBChQ$9TYFA<~|a4U236;y{gZl5|mG(PsmJ06{t7>NeBU5~5N%y7h+G}kp|hIs8s z<0NZp0;vk-w9|&_7q`sAk?D_Dd((eD3cj)SlYwXlFS`e zPdM{;Mzv0w(romro=u&P32>Y1*1+;{i;>9s=a4I6s`j@c#ziuba)t8;QQ>>ki`Sp4 z=RGjfImOm9Tk-Qzv=}OOi1{{}08H4lZz|2}mm8<+CEqR4QW_3Ng0;+z^4xlFQXqcN zZo-Qmk9jU)Xy8Y9>PMDw7T*Kquoa0oR0e&rtb7^4_+-g7$M$pG%~*^BBgap3KC94e2eaO= zjt4_@F&t4n_k^An#|K1%AcoPO4P-c z(g_eP^~ogMw2mi9@-fb=hLTaihciD!ELg^f9 zX^J`W_>PIO66XuRlYNoUA2G_fMyQ*#KtmX z91ZjsjPrs>avwAQk&3EJ3L%Wr$1-oXqWkd)t=r;|>xjm|>1ZM9wxBV!IRNdbc9$_g zUh@5`Ai(Yg4=c$xc&_!jp2BQIfN1*TH_9SS6xU?i&D%==oL}|k<49n0|KKw$b}(n1 z(`;YVq#~dNar=CAu;Xnz7N(ZMwf`0uuJfzAfG*n&{olxW1>lI{=F~I)q(1jNVHKJp zb^TdTCDZ!*xEEXs*GB=gm(7qq{+emqO?<3fMz)Eh*gzkY7(b}k*5^e|Szo9;og>2_g6P5@Wv@M<{({L$ibn^RMkCR=>yB<+d(1|#kmd7>XaO20eMlJo+5;UG204bH4E>9n|ClEI@}$J_opbctvFcEv-ifN& z6Gfidx)voB6CK|io{va0;;xNE^Qf^{lQ9-lnJO*fLZ{HzHg-G`c8}|0+u-> zUc?FEOKsHP{ zypTtbD<{cPelr+lxHkP!yq0k*vQT8=13@9VojG%K6RQV#m<2=3ybz*GgeaCJ-F3fe zh*hN~eWVCdVi{7dM$5$Hc@ffZK>Fu@F50z$1Fl#X0%XI@>DBJG^yo1TY@O%c$g7YZ z7wz{d^cEP%&D5O9wgP3?$DPe!#cK;BA!xJ4S#CFm380{}v^@BfpE_Z@erx!MybfXU zz~^Jylta zBRpQ3@uR+V+=9ifqHKOZi=;TYH&fN?q~ORqI~S2Rrw9_lkT;q^hwaQ4paC}{f~48T zjkJ%>6d!Hn%VmH98~2bY^6HH_vY5ndO{zDI(`_se2mV0Tz@X^T0mS`QR_Q zZUm0bpxxb^!bK_xO(38+HNv$RG2w>B|5n(>_yWPSh|t#-d`nucD)pjjBwepLnLrez zF7u~Lw#BeMK|#F_+y8~IvUzuof=SB2eF?i;vN0-ga09*;!W*KTqsLEgSunJ>5GcK) zlv+Ji@}1S{4Ya?}ZbnxhulI4~`#RR!>$<;}vNUlV*a`Zn*QnVR%*wp>L6%eQ1*$;F(byr#Y7<4_S|W1^ z0kby9pP>SR%XIanViBVh`&c;)RtQJVvrr7FDXA9l+I5>)6;K?#Vpx7AaOV7}6Z=3| z9CMAjotEv-1b$g^bU;~6Qxd4=+W{n<;dZUFq^vtNJffkw-mHndJfZ4k?0zdJWD{iK z{7Ra0L)BJ7AHx!@fo`uh^ktMvgxN@^_F-_iiFgX;6T>d3Q!P(R)=T%ErpfPf-`*A= z_erh!tKI**bcPe57MqnT6(8eG3Y^C~&@C=&IWuhdUf1f!xlZpF&SD8uD zX^tg^K1G3{VJ8ba*S=RLWjRfWuZb14;o0PL=8v6mk1TYw@eQ@!U!3uKv4(S;AXvQ>)hfniY28iUyCuE1GY0`{`GJDm?-#BWy|MDEJ?n zaLNDg?UgTXv%9(V?tWeP{;sQi5bM}I*~kl~<|f_|iDem$q8cMJ6VL^hzkAXc2ogRb zyixbd0bCY2mx{2Wh}w&Ihh_1#68ydhQ#_q!f&4xmZamwVK(|^B&1UJwcvnjpX!}Re zMuZ-pvj%aPnGgPN6R$)IB?|18i2w(}_0hgAPo6z9+Y(k#tr1K)?9Bn9&6PN5#o-E! z0i-sJs|>)Syc?j-2*`~8O% zo4F@X2GJPKR7@f$)xKnRSMGwpaC`RaQUV~+__np;CovlqtYA~K z5FFPwN6+b}03OCNcP&a8wq^N1fKmhqh!`-j-fx~!z`JO#YD@nL_z($nHTJ(wr}qEX z{r@_h^WTvC1orhRMVYBMtR{{h`>teT-=HG6yji&Bn+EhMf6o2wfC4s6_V?QP3LDJM ze;B1eD9886{0j5ZoSv1q@VDnZrXUB&E@!-P1}-YmqY*Ow$oT!*vM61W_Jbr`1)GT3 zCbe``SU377;~UArreBqp#u*O^oZlCZbc>Tj&C@v-!s|~^3%->y28PMlBr5@qeMCCD z!m&~@)?gt!z{i-hkXarZ5?!Setz?1wV3(|QR4^7@ka>!YG^=>{JmKJr4k}#Nt#M%wr#@Mgae*?)!d2 z1|LFN%M+FJB$Ex(uB^76DbD8&K5gXJP7I9KJPZs!8RDE_QPZvzjLCEx&s>23-nkR| zknbC(W;`jWRufT$CP?I-s!j(B?wuE9Y{QA;?YJM`VpZy#ey^q5Hyy3$Hfsp#1OnpH zK=B|fIY$mMt7BMfty#xNSRparV z`8ZjQ;;s+;eR_?NVDOrqn z7`HaM)iyDxEY~oFIy||~g-g4|@s3BL@(A@(k@wF452}Be;tH3P%+S#LzhrF*>6rH{GNg?D+q991tDg!MWY$ZgZGH8`J7)|a z^YB@v@FkkciS^N0{e@&D(%w2|Roj)K-~Ld%NH^a&?>YR%)8DD5it(2wH6rM1 zxWFV!t2f5MQ*J#kbdA2PNzQJJXYXQqG{~Y@x?Cj1dY z?R58Cd`Dea1-AAn-$ICoyY-W{QgY$jcR zsNZapznjmA=2a=_PF*osQc4|(^CCZ)9r%@TRS`5M3XGY>tq4%2QxR)!-##)<`f9Y&-t-jLVb)vlC!=fpGo#*&x|PcxAHkXbS{$|6w9f3Yp&qj74~*Bf zVu^92Z6yA2T)kh_tFm@JTniIq)74F{QD~+{uvz5b+27>1LiF-%QQ%YJa%+B;Q8TMwPJxM5rf=jfF!IBk0q@c}zovdLZ$eDvW-r(jd`vR5Aa8_-WLE(1J># z6YYc0Ha>K+ue6@Ht^4eX`#fA2&FJ;tbcpPBcR_wq))#R;X&&eI@qcZ$V&4)g2s{*B zG}xw5)gh#p_Vp>`tLEbcIY`?50dlH;aZWNVH+!BYaJ$)tZvmLXJyV#!mMk8Jbp@Mu z_p1xcc7V0cIcWpO2ueLusaA@yfZDj)S7L7kY|RiBWx7L`LZ|Ag|+OUbOt-mZ}F~h9eDKa8Cps;Ll6nFE76^WQSrdB^=O_&+U$FXo0bJRJp=rq%F}kHd}h z>zj)+!?#AoQ5mC?4V|S5`!prSKQh-#A`Pm?ppP8}OzJ^Cw2)2ZLAMIs9v-uaj~56%`LD%v6y=i?}Z@^EH6#vo5x@M5+$?Ex6W!O)f^ZS>>q|7(HJ%DB+ds_QkN6B#aqn)j} z{@2n&hcW9^IIl9;$pj_FKT?RD!w1@XHj*Hhc&!Wsvo$Yt{qdoMblm@TFoYy8=2$$- z0Sxhu3O6R&Ol7X!a_{fF>fr`O%So%{pu;bh<5RoLPYTw1lWZsav-3Sp5}dnP8rB%{ zul?G8Ew$0d1^#7!e@pAQ3OxO8c=g~nb#SnJa`F==8PC$1lS>8@PDUM00&|RQq{_id z<@6^zVHpw={piU7a#XI;4q6a7QV=c!sK~-<@ka(9*Hl7CMd+NK7 z=2uDhWBAEY(Y`45co?h0$ev-+Wc~0N>=lBp8F)#FaaBjz@nNRabgMmWzAa^`zr|*a zB5JoGOCtR%pCz6}J&H{{f=`LZqv_>7)fgwyF8s8;@ZIWByw~^8Q#YZ<_9m-AM`t?E z4`IEfFb`$*gB8Xl^4mn!Np|<|NB1d^j9>ibhiroDS|b#3V?86?E4%XHNg3<`sll3s>&?`-dg`uVKwZKz+^eR-n>iHu4wL_lBU05nlw_;J4Uc&r zbH=PM3VWRw1xORTkxO_gE~YQ4KiWO(DO5P<_asqd<= zwDSijoj$+1Cm3XI5J32LxXt>kDOQsf(9r=;YVrYA{lH61@L6pNtClNSHw}5Vx@3&` zPqLQH3}5eiHLsc%+yc#S7r|R*p~MWnV(RsI-xdZDttAH&_Jah4m2et!en5~|N?5!Y z_-8v=q%Ee$BQp&CaN?A@*^=+0K6AJZjvCY{juPs`S zr{jge?xT98<3J^OWu@b`Odkm$%M0}vuUWM%o+XxIl#V`zJ!UN~CsaS;($tJU&yTEH zevhYy=lSJ==4@1=V{I1R=TB;L6LJbKF~=)TehLkmULnx?Dt2aJ>4G~=qAzo#nC>3m ze%-1bpeuOk5DtDVXcqM;gYQiN_07U2&lyRSY>#jm32G_{EkRreXk6v*A3xATq%d!+ z?>Q0aA;>bIw6L!mgc9`q{lxeWdB%L^E_1q?IcVz22K2g$LlAW$ z+sO2>p~x9E6WN?ZU1QtEK!GiHxm%r$CGt&I|5Hj14Q3wy=oo~^Ff{9F?w6@0X5IZe)8e*X)bi~DOY zk^wISwQJ+DW4gB?xnlW*?-*f*EZll0Te2SMzuDce35E>LE00rk_coMHYx=s0(G=n9 zez$3tEchT%_N4QE&Q|mk`BUn|Hx$oz%GeZ)93bZBWb3YxON`vFtsY+?*YEc*tIPgQ zKvDH&omtb@BEH(87)T_?gp^a(+mTxHA=n#9TTk>{OYd?-j(?qy(3fJo0vBDF4cscd5T@5*ImT2EpHp8 zjzg*qGs^iImub}lX~f53I{3i~Tr^)usW8xQ8&PH+?O_lO6_%@B$0+ag$O!ovl-~i{ zh9R@6qIE>6d7#HqMFwS!3te~1%FU&utRnt2-jTCr4 z-mGDhuVqlKpdl4aAY*(C2M;oSW^tO+v@mFuvWwu%>EDp^S@E!)$jD* ztJ6w~^3(<_S?{H1Py8s@&JBbzu^?!_$e}|dJS1-Z}lzmoM zM@{`I!RC~~gZ2FwxAA@sh$fzt-!k zZDpRGrYrLvoanv#kcKVS{sw1dF^FQ<#AUt`X^lt(kL%+$VqWh+Q+Hn?ztN+>k_`Ph zD$4-zNvzy9S%t+ zj2%>-ZwH~blrJ92f(BkazMeH(W5A2Be<&JX&=6$UP>`lJW5Gh{gf(0JT-Fd$K~YA^ z6BjV~9B05ZbO4+ZnBVi;vM2`%6DO_T*8hoeVTC5`C7DTB0Jv364XRaC^g7}KF2NZ+ zt5PoefZ3F|t1ZPVBmbdhb^e^pv2lNte!CZjYt11cu!2`d-skdO23NJ1ps}@5Ph>ov z^SQ_{d@;07rJr~eLX5-T|1M@eA>RluWl+jY{L|4j%YG1f275QpCWzWHn}F03E#28a z6@YVza*V*e!O>yjbupES3k7Fe?A#sQ-$0|q-Zv4$Us*s1oVVRGep6#PV}Xu$wxQ8y zrMruYY@JQLU;C?K`&G4@2N?Z`)>c6VW!I9YeAH#+@wp?i!?f5kOs^dL{|=_i?s5fz zEI5ow(8R!|L&Yp?O6$GoxraV>rgpa#T#0_@I5)MV^^and9kRKLBM3p)yI1RW3vL}R zm?_MDIf+FDw$6WddJf}T1_$s)vTD)zmhTKQkd{PgKQeU$7{lG$rkKckx0pQZ9l)k#Q^0d^loFZTBgMfNP-HaS zWNWHIVo-B_%s+FBcTIWU!cUw6$QU}R+lEE=_UMlH#WcJEGe7-fGCNr0n06^13vtNR zEKAR2k!J9rP)*RQVlp)3(?YVG!zvyURv4kb#PhdYvld}?B!h;LO| z@;$b7#V)i-JfxPsq9$7Q3?rY~uZV=RsV$sbq+SLSAw$dMWi!#jm8m4;A|(Id=U@)S zAa!8LS<{IVaAzPL$ZBL5!8D^N(@H-6(4w_?hcm60b)Q+90HW@pMk#t%lvjdF(ewi@ zNiU7NhJ|IOX(}(Y?IJx@y{>a_4>4J{4?*bx_8Lfjaj20M4z~8=QdPYDjQ8%>H01iP zOy59d1JBu<%7MkX|I;h9u6TqU{xEV%2SHxB~L+WapCQ=mbl<>`1m58I)PCL>ft|Tyv}y$u6s4P${GE@c$XJy*X~l%^ScL9gkSYT<}8F% z2+qcGUOI@ov6Pk|w?-Qo4(tTDB08Q>HBGwVBoDMt^J>oow(PzwkS;YA-+qO%1H)!J zyYY7w$Qfl`6$rjeyr%J+1UO!VHgc{g#H=~Jle&pUBHB`kubMp68LzPTQgyg}ni_Mk z_c3>JRVgOz!k&DanvV~idz3SS&Lo)+?6-Xcm)j!6l!n~O_W{xNZ)CN%NL3Z|V((I0 zU>H>~IO5X=seor4#<2?yyAvCe@*jv&dSE0CJ{d|Ps+(uV@-@xm*ka40OzZW!Zw*h5>*)Gdv1xxFnz|PZKwX1RK0h8ssNalWwpl}-6s}1Ek zaqmJ?yk-Nu;WQmdgg8Y6MYh5e>NQd9+QkGUb)NRJQHpbl;*Bo^VhEoYc@4a3Nc{^@ z(tBf7LE1;;z$atpJRFU$VlDPxQ0S5b^ZfHpc(*_ox{y7J#=IS+1ft$Z*1goZ+P`h6 zST3?oE{|%^90|{lc{&DE+NV-e zJI3vYRKnVQP-?f!tJW-2f7Nl$-riXHHe}qJeqoq#A8Em0`3om`{4(qK@IT`&n)l%` zP{7o2cmb73@Wa#^DR2svj)$dGWbkaPrer)7>@m)T+fT%7E~Ay9 zj+@5LT`xhq=VcRl$3O2Fyn{Z?-m&d`4GcGXi-fZEvPwO`72|6uA^J08FJ2AB=Ai!Y z_Dp1$2Tr6c===Cz=r&EDmF?6RgII6X>9*)9G8 zO~#+`hnpZt`%2XcYZZ$tgm}FEeA#g?uQmKW@tgxFIef3Dv4qm0p9U#rc2Mh427hFu zg6IzC3$x?4NtCoEVs#ti}5MxKO2;WbflK6lcRa&~D(zZi_1@o^o>q$vCT=TqCLu1p58 zutj*VPbA+>EDhenJ4kO$jhcdH#HkeOkf;PhM@U+lnrF(ak#gw`#cAca%9Z@JAJYX< zur~pbXeF3?ODaTYdhMt;B1ekL;ThY+MDnjT?^e@F>kVVPRuwdDfpA8*?ioC=xrLpq6+#x5(%1mugX}B8UdS^+|MeAnkj5Ldn0neYeB7Hi+rs zE;}Rcz8m2+Dzbq|XC4lNDz1A!Sy}ets;{do;i@l0;ap>?uPddml&~tYSb=y}sOx); zHFY3edwWmGx;3F~Mdb>}wtHNV5izR=#Od$b_BUS7U{Hc^*WbM_eUak)G|_Yd))`Q2 zEXMZninUiUyTjn=d)YocVR~eP;^^ax;#?Z(je%IJJ~M1fLIV+;&37I!6?oM9#jLf! z3SpwP>}=;_nPyz`P}H9?M827MCBNqen|2u2(5&$E?}PS4IK212V`!8#bvI6il5`CHbMmIy%vs zd#g3N{E@UD`AgT;3^N>!s2ye(RC%$z@daIeU$AI&VfI&HsBM&a`w&6SL) zLvMM|WNZy1S-SWj1*mUG+IYBFa5EOfr^X^e71cl6%!Swh@DAI=tJ)s^-3h}M!5VP8 zlPyHpTgXwHYbUy2ge^E7gcg&Zus`pc9B8Z1{hhb!<&HS2opVdo8pM^bLU&`sjKprz zlW{jd^%FAGs+=~+uTRxuQT^SJ`gk1R$6{2rTcxLqCfWD5{LXII*egW?!xZanIQk8{xGt`?bxn~a{V1uHMFRL@dO zf|%%i+KbpVyn1x878UU3;_0)|^~6vJ#m#@iQbEq1j_?Pp_CpQKAOv1C0E|^@B^tL% zlnRA!)xe>%|KX}S&fnXM#^n7DP(38?@f|&WYLH2A*%ca1XzRpI{8N?1n|zQ}U{h2- zjgL0&-qx4xxR-HiFYgl#dtZfBWBcw{g?hfhZWb;hc#%dre?AwFJQJZJfkMfyZzu`` zD$VyriMCJ3otB#+G;xO|5}eqZ#Wa7yo&@uA431>7NSmc1KLl*z|Zs zDUCGKbZ>0s599DbVheND>Q-5HTw_(i+s>fAW-p&aDc@~_(c2dXCNB5uv}Jzlh9hzb z=kFV#C!4)Ok`MygMF1F1g(6fo0Tvj;(|da11s_$VvrX>R-iF6v6)9A{(&O2oB0S{* ze%nm_3{h!Z9W@1({EWc=U^_$c3si)+rg7OBnd#hQ9d)k{|3MJH#jC<^q?PDySdGe} zk1@G5k1)1HLct@N4Xj{(`IuCbaxa!nKQ4nusvs!Rwgc%^iU#7wA6gHGv{&rv_DdMw z+Pdo)NnjqAAH(F2t9yL-*OZ$#egs(@)-WCHliw>kv@(v7@)m+R^Ej5n#cLf zPR7!qcT=@8!%L~kE?PDU`4QLcCD8+D`@aa~_)1N2#GT1*GhRLlpy26_{8;viSGp{O5#YM=X`E5(zfBPc=G zb@DdZmkYq3!c_LYFSc#I>kIAbbO1w@=;9jRDVGZ0+NSV<=%>WAbHK35{fLLzGsl@qOGINN zq%V2@5X^&eZc`u2!1g9Fm0jdgw*4wQ_|lKFCKej@OGT}7FK2UH$)KtB}GURGIE5)O${<;@8x>CKDjT-p3t7Ri9)r0sq z&^kRR*TQ;7rioGTeH4gL^VeW-+CXhb=;LbN8L0`8)QRJ ztd$65okA7bSDJ${!yOirGN=+S9$-UdcqI1FT8fHvh+aaJ_?M^ z#AJ^KYVE?1Gt|&IpQ})dOIuJU-YY?_U|Np;jBaj*;**r^AydkrYVn9c@+*NivEsbk zGpW-YOJXttrm7TI)16@iZJx~cBTAl9!F_GFbc<%A5Nm@o!XF(0*7!R5(NGfeHmAyk zHdHZK8Igu82t~hpZpH$!6gT`teGY=dnHX;sisUobIk#Deuo&wBTSvkRb&J(67Zd?| z`H=BvO{QNKX8u>6X_O~D;g9<5lqxuy!_od457y==%gTx8v0vQ}LP^L4z8JEEytu&q z^mIJ6rIdIA+*bp5WtGGWDEJY zOm3Jp2x0emC>X?gDJDK;y(eYH7rR@{1a0W$FP{BOPTTl)3Q0>MA^#zcfS_x6=pnsy zPHQ0*jvaLXBeC0?Aw3}xO&xGH1@kcTEX}Fx*u9*N-6u(vV5b%lPT%Ea$XLHqC(abm z8Z}7PmO7?QMNxgm=Rd?(SNI;mmkS(4o)P4(8@}euaFjrGH(vJJ^S#F_9qh>Cx#r6w zT;z(_%gx}j)Q{HCu8FVF8Z)BI|3rUSqRdRn&f2OzRv#nff|ZGES|As_G{$Dz@XdA9 z>yqXuE3hy)Oi;ERo$x@7qHuwh#G|}QL*t7bOZzuH{$NW84&s3Dc+-^>0` zN?RMGkx7vWUu?_lhxOc^g+XlG&(|EnaNNr5+31pi--Z(VY6u|^d-rMKQRg81GhOZl zmccWFE`^-@FDfbaEZegCH7YQbMx`livZsTMm$uvnW{$FW!COg zX~Rj};}mPK^}j3zQ$Is^Yg`BjFmgkQ9m7QITNx?fTY;6cdjli|pyKcfs zQJfz4++vXiw6-@Lye$a++Hdkc|3UVMP+B){+4!AF?-cj#(X^5}5QT_$eT#$li7aD% zQYbNgw6mJ=F*I5U*7G&4jxv2&VLH~QV!$xFZm`DPr|49kqG^;!bykQNatsha@!zI0DwvA zwq0yEM~x;9j`8pH*m>L#yWx+<$7LxiG+{>LS3=5}xwx#+%dF_FqKqrHew=GdGEP7d z%xBJFxTd+lMdMV`aR`1~?F$nD^u#(a*2ojnR zk_`?l*qpKbr9VB%OumaJCD=yKXUELvu)d6VGCcoT!_8S)gpPQ+DsYLAA8Ugn80~LR z+7M#t95WO|Bsx(CjE*VJm#bL}gtTn~2KJKgXn&c}^7R}wK+j;um=-_kY$mrlu@Y^p z!)JzA7ez`E$%S$bu)++L{jh@&RGsqEq)Lnf7E*`FBl4CdseiY~d=C`YHXpFq+M?(+ zaYApAQq=YupwZeHHHuui&O+%Y62T8n2Lm||Y50cin*jb#cb?h!yIr?6x`_&WPShC| zl(ML)v#Nsk9woXPMVentuy_4VQ{ z`wXE7ng0a9`=Mo}jNphtz%Hv+r!SA2t{LL|x`>`f_wspz^@SfWs4~jj)o}i^24$o( z5`%ci7Q?UDSG3KN#iYsw8_c54l)qu7*T{x1+ffZ6ewkapgVKx6q| zl$`jI)_exQu%1X}$}l_;r=HX#@xAqWmJ`_6?PUO4wj9Bd)-rNMZ_zpm0Do#^-&gadVgd;?&?G|jDcl?b^h-CN08u<(#&dJ~DT1rk*Qug7S`&Na68hdJPR@Xqu~Ar z59pAbp!j5ycBR=qIEt6IjcmHe3x<&<=8fO?ck1c_ZOOXLshf3@k9ph3d6y?|pKAjc zy5Mx07nb}vFL1ew8N|?ljd@>9!*B#1K*D`~g41cEp(qZ1*VoXpo`}gAP}sTrgFV#O zr;jpkOshyII~Py7grA~Jz88yQ1E%GXtb+1o+LF@p9muK{r5atC?@NfVG`7-COq+PL zo46)}Br7E{Ek`v>06u_=kC+sF^A9${4;Rw|aT# zP0DW<4l5z(T(ldHIgxUha6{w4{L)KuZJ++T5T+_6&&>t9Fh{_z7^}3XiqLOWtlWVU zmF@Y9-ALVbCOGRlYqmZfAuBgC#u$Bp?F(9$|l8;McY5WVq6 zB^`mH++vnyM*72jYDZ6dB|QF4 zf6jUgbDf-7QjIpZld=gu82B25SXBVo4D^M8-w0O@wX$2dGpz`_%PY;e*FWC^SWt@9 zWoo6>2yN&xt46;Id#%+pC;G+SVeAz^Oaqle2N;FlZv^E3jH;=RZ}$D%ZcFP&BAkwq z_R4XLLx0YD%$kgnHI}TkL~RRP@ZVO?5%|Ego3^UPdgK?RtQbuq*tPB2d??`NqsD(G z*&Y*QHG)i%D{bz31Te=Iy;|S0JW0Z>+vTnDHEh<5^3%MWb*`dg+3eWKXqnN5>0z*=h$QzX<{%k8zvv&F-byMQwgjFqXrEHU@suX0lx zvyaTZ3)y|kyuky@4!nqI9ss|!#uNzzH03djnj_`#F@qog%K`04M^aefr$0_aXnbz9 za)vM%(-si>?oUA%H@sk_|}zJmW{!V_cW-rJavQn=^XKKn;*w-Mrjayi8BwHyu5yY~wK>^ac-SKOhEn#mJVX_1f|SPTtdc5`z73DQB6aO^f3Sr9)Tp=gQSi10Q8KV$sc3>3jHJy} z_+C>j)j4muiJs9fd&dIo6HRGPlo1-HmK)T(wbw_GEgxFCED6@h1wvY+*<_Ntv<>D9 zoxu#&It}R3W7FqXm9bMl@cw+5oC@7csjXrf|E3^ve-O$!(r*G2r>b3!_RQeHfCXm6YD)x5P(UEm9*6M}jZa8>E10)k7SEjst<%C|$BCJgPS| z(YSyiDDx7dhtt9twr*YlQX^EP)u=Y%u`K5TH&IMNnW*s5dVAgMv+9Fj5AG&foENbh z$!N}NQnjX9eFcnW;ysM*MNuTo*{A~q!+4D!O{&;jtkLIP03 zp(%WNP>y>Ae{*j>gFVr*?s*&fcHb5`f!`Bu%lV7uAAElW)P0axr5(r~)C-%;^=0R@ zE&Y(Q9tdLS0<|4f>QYSMy~$LWQhu9RS&{njX+1QIYVKZYa8&5z03dfMw_zfj+JB4DP=$RUnK`J<4fV zqA*Jf`=8O|0V@Q#`4`0`x>=tE&>WJ3zl(AvCGVOePpg}sS2UDhE+YLBcDzP=7~MvhA|=wXWboO z{V(_w5|*esBh6fK<7;sIlS^2bM$(sI*tM45RpuM{WTP_XcahOrm(l~i>k!1G>>NB4S%K@740Q&BtKnS6 zq6Xr2uAIubjP|@Epo_=>G$Ao1xa~!h;#kb+s-V2WW2`CgcrI*zoH#}LW1Z0`9j^g;p z1)yy_$ZaA-)-99&lb}7qkXUOg-ODg560B8F0M?y7OLzba8K#n~O*yzH=&{#mnHC*ALep zTCO2EJdx(TpGy;j3#Tx#Ec1VaomjxJ-dPcRNTv0hCgB)9QrM^9pg3j-RgOdH+ zm)IHUB8evus!p6bVAqIPlaZrN<(D^l`6$WSLkLnuekfGBZ;B&gM2rAtCvrVFYZBwb zS<`tm%R1(Mvzf^)NMRr-!Ir0z=rG(bVp7(8=M9Q>ta}z zHxu=(AWMuPHEi_1?hvv2aGeJ6fbtJ@3r3nEMTJPlRS5*H?PRk;UD zc47M_hqmPBgr`K}L^u$#H~%F}k6jpLpEfhe7D`XG%U>{~IUnQp^&EEh#9+SCXQkJ^ zIJSo)1<4VLL|2xn^&EG3*29%JuEIhMF-jNLZ9udAxEo$5G4 zRxZ+<4)6KMeX#U$Y2&q|LdMIcsIRNA5+%;dRKsUyQ?(g?v_v_^oDF!L0r^K1PYQSu zpTW*%W99$GTX7^rTTnbrlqEE(Q|0+;D=;TcTT--TMC>+X$T(V3JlRtw9z<6VTemiL zk|-&|ol<-yy%);b^5q=~KBIhT%`)!N&TQ25>J1CuDvvmL*kbO2``n8jBHtL52`(V} z@!U}mKa@uu)5k8DckrX{wCgu&(pu!NpQAFzagkk-a=o5X{vY$x`ogZ2@V98{9#`BjctETfkj|9Qw8ale&oDjQhKT@P_ zCCy81y=`GBZE+d`Z041IOk08vtWKuT~k#lX}}0r*}!lG zyIoo$&OGWv+%$2tZkW3E0uj`v_^OD-WKC;hN4U^1fMDukcmB}!c4ZlN67bjz$>*b6kdb|kC8HMdK@Hkf&2PJZ;h!@^}A6E*gLQ! zixy3dp?V1Y zNzEZ@vE3w64CJ1^#B9CQ!RDYp|L8h6m9H8-w3aArb0iFtZ<4N6N?2R4O1BNQCo|5! zliPUx)tEUwFvH7Gl`~*prE%8qP`-rWJKo~22tXLjP1=e`%kzSf1PI`htf&=3@QM2K zqwFw|K_?yan$#IS*JdG_6A(3BBHob{Wkmwag16O+PAON*ruGcRbPlV=W?m1k_toIz zOuw%^uGILzjLmVfcx?hcvfL$P`7@ zAh|W-?0*ud0@Ru^g9}C_fS>g&16G7Pxi%j1=g{YNgV8A|TzDq3dR-!ej}RizO=KZ| zeNsN-?i;>SUTm__tHtGQWMI6O1cf-|Um2#xgx^~Tt(a8^d*6>j zghWBFgR%vvx@9vNG#_rlj6+ixuRxD;vSn%GMi&ZEF3}Pm`vm{G~%fcF=CSn zQvn<7zeqx}aB01>d#~m<%CW%AL}sFD!o88j!sDrKNbDOIqr9{%gc14n6_Qs&-D9Dw zkH3`qs%(=l<7=?dMljP|emdnyc4dxYqz3w=V0~w2_JhfOK~JN9LE*y1zt8z06p$Me zMC8-jNFF%;MVEv#AJ2cM`V4EGoXaPuxDy5?{ThYB zRb_Q3WCGP)LRUGZPpj74J3pkOz}^WyLALCu<3FWY4I9B{wGZ^)lHZe;d7NK;V<+Oi zl!ol>%bV&{{ysQ5) zeblli;k`59(;}s&FI}Eo%lv$YYA8XIMM>iET4|<*oFXZ&{0AuG=8DVbp^Tr2nAQ|`q6(P^D~O)zw;FL(!N|eYEP{k%YW_g3=Q99^7&iP34^K=Iif~N zAoK2N$D9i@l0W^@}nd*&7l{K88vGMoqDlb)yFp=B_Oi{N0d=Vs)#SrI!(5`+w zcb^tlzXzkEkab{RObQu*Yt$SfS=|u(`DbtozT~l_v7eK!mfMCIyNE6>YQRXWQkJHh zRL}$6$v_nc6aGB(EC>d(FIMb_lj;w8x=UXjRSMjWFdgj32Q>6fg?`}x7y-Q!R1_aX z5CTdkQU@A*b(5t6iXzu01@g2ZeE1MK_7)bti=>0C)obAQvk$zhySk^h%U}1GkQwC_ zSkmtGSs_3FGZ33G)*vhI#@-P(Zm=6BItHT0b)qV0kGCPF9>tinv9Zd1*CR1Z} zbGn+FnU+X8U1#(~Y_z8E32Y)#d`SURYe) zB2+z5ToxZDmYcn7SBQXhy*J$T{G;JPc-ynePWYqY_bI4J;2`_euV+fZs+hAU|6fI4 z;6$ms>e3yJ@0z(2A15KC5HOUCVCUzCm>~|FXBC*vbY4LPQW+RbA4ov}{P!0e5X-;X zdt+;a$kC%Ps_K>O_tbmARaNYkVv(w%-6Z zy*1yn482La`fGOo18U3JH}J^XH5WBQtAHqguUs-yuVUoe=x^kggfwO#RD4cSLBp$& zYGk`%!%LuF_QUq%P58xg8xrgU+!ND2iu5AqGhwH%h|)r{-<+dW^n6JOwWVPTttIg} zLic8vP%cAYcGB)oqd33fksm`;f6fl?o4jgKkqMW(g3;{tIbugH1^Sdjree=BPp6PS zz158WvP{9@NAp)~b)buYud?iBN!wbaHg1ih9U}X>|LNE>rs!y;XC?kp(BSdmGp_CJ zDgJCp+J=*}*L?|J(^6MyWYlik@d5wcrWe!Na~)J8|8Ho~@j^9EBt%b!IhT?%c~?3lub=>?1E%+Vbvp}dKE+3if^Us>4%tk?7I9m0M>t$kANu-4gT&7ue)a@4Bwb5QF(E2LfBOi0U+vno4+SuGG%H$ zmxP;GP?tLpCu(w<%8Nd;l_&^BBTQX~Y2{<^>LJli#6yY;)uG&exa>su`!c-Mw?NxT z1?`u*Y)!^5Xt+#F0xyrM=vdiHnCK>4I%$J6U4q*EtiT2~zf#+3^d7;TK;7z~a|-n>;Q8SStrQJBXJrY?rdB^mETGY60C zxCal@LkHf_K^O4(4YaO+6f4eW%}P$MdZV2az<}szj20Kq9^^S;9z6}_$WnIZTKFM+WuM_V(z9g|>dEZx6VmfInmZ|&3SZ?ZK2G|y@_v^}L6L4&(BMy8jKAVn7}DU&SlQzXn;Uo=KE#GXJE4l6im3N=otp6+rk5y|lc7*xTe8tR(DZ0*aAbS+tOPqR+ zp+Ka3ZJ%iPJnri$wX`VZw64x!*Jbn&#zp9 z?69fg16a7c5(Bl|lOI*%DJk(~V`1O!R=oF`Sy)$GB&v8xa&*nuYPe2LyepNuT#jiTdq=dsL zs%S0RHO6tRaGLJ2ws6sw9wHqQ z+P*I3NOkzy722b^3%Ay>?WAietQh>qdEYJ>P=q{RQY6gC;I?cRD1WDwFdb!#wQvpt zQe~5F9RA}jS%mSjtKTV7DuUNT3mF94TiQ>GlE!a;F%7zC^L?%tyer3&Rcar=fA}%Q z1r!Ye%nH87J9BTn^KG6cmvZnkgQVtZBq~N)xF$&>K4r% zprSFwWWuL`cirqXQu1Wp!e6>Q#uNtn8>MqHM)K9|yJ}OE^0+J`0@CzdJ4SX$QChyP z0sZ7R^a=3LI_SSXxb=e)f*QFHn(%;_LiqdmHx&OBd(Y#K;EzHd90nn{e$%?Lel&4H znUZMR=Qr9VJN1FgmgG-Nfp?kShv(E9r(t>Zn|PR-wD`nxk-p>OJ)_E%m84mT=;deX zO*y7VVMbCkRtQQHK#EjYl_qhvKig=RJEUt7{{ks)o_E!eBY+qnrMn`sjsR-|bW*wH zmm499ELX^Kt^o{kYg22!z-OfHYFz3c-1DokQB;ku`!YFjqb(Tx=%Q!O;(v|2RBzX{ z!SoE+P#XCq1M_Z5X4Kx15+du>P?byLPw!#Vdu9d1>7j0?ou!-&9z*4x^==%q5?Q(Y zWqAW%3*Tk4NNOA9PnF3{+Psr(|939@T@?qLTDz3Co*>fqROdCyXo9Ey z>DFOp)n1ID-w#n;e%iOC`SB8XxTjJU=VFQ#3xuZ!FEwq{^veB#mi@U4gB@BP;|$XA~W3=GsyA^&`cRr#TefEr()Gs?Ww2 z&(alz_X@{WhHpp*);WvyzqKKJ5?qmh+IxMe!C_ulw0KV}xAEu5TnfgI9`L~pRcc4% zQ&A~#^}H67K43A0i`tYhnUD)&Co2Hdq5FZ7H(=dX1-GKW&jR7{0Op9*{{C{QiXohx zT}vgtKc#RYtm})2-jSU*`6VO0_x=$NjV!KI_^+%Z%J?*;rjB`cPSMbysQL=SJ6L8} zbL}_3o7>cQ0q}z25p>R2Ho={{+zJH_Qm>@!X%InX(D5-WNYlYZ53z@4y?H)@;zAwn z4IJ1Np(90|OQ1Lb(8!NO1ZkxXz(QR4XCtf^ncJ$%>LPfM!;0B&rT-Ku{p2;2&LcOl z@J@^C3{;3<)#R%^&UsVLZD>EC7P=j!@q-KmtR7P@pC{mm-tipO)A#_W?`zjdA;{(F zv{uX@_4$JO1W6rtA-qZW~1+%uxU7VSQ9fD1u{*1!mH{>c>_XvS%@ zoVyHCbm%PQ*=umIbtfL9#Q7Btx0m%SkBs!677dg_i`+K$!1a$TdXm`O+D7;{I@e(u z<2!tQ(w?#bryT%yu5-xe_CdZG9kv5{4`f4!vU4#uX@3k>M1Sd8w}mKFcI<0 z%Q}YxPl1{pN>uIk$+H|;(og>Q)ikH~M*T|DNyCA8DaJ=S_7vpw#J-wK^Oqy>qRx6#aN^?y(0+76QH*;)v6L8Vf@p50h=pNC(6}G!;=?tQ#@7 zz#YV?iHf>^CNVAP#G<_o?n>d^;1)??DX+%%f2tGqUu;qS4|}LY+-i=$){(e2j2&zK zkGE!TVPr^au5BdkqKs6^F#W87az6pzNF!AHVuOLB}#V*>iTForj@8P4%VLHL%LpKM>;E7rXa7tjb1ukL$>zR6X06 zr&YMoI>^~eW-qh)&Kxt^x_t{rdL7^N+zr!y2qmNWU&aNWu=zDO6Ff+sRR2mBbIZ^l z6z9#?`yA9&V^qrp@?TISjqJv=w6Fh_`?kO`C2`QJmf$au5M(cB`l2g8yc$Cx@@8l7 zQdyea%Yqy`B|4(HLAb2G7ZJYXm&5BKe~U--jjDmgwL-^h!ECke0m?Rl4F7jiO(V4< zM+}2-X)(@e@&KCu?OBm0)3x;e3iABWdMZ)kaKoIfnM!i=c3AMz9JJKK3FOq=$cVQL zj6+>()r}AI&A0hU@`*)@d$EfD4mVQXwL=~ZJ0%<>j;+R1Y70}JXWD7cYM7ehblN3s z)W?j;_vXil`5%g2_EH?p+OSg))r8d5eciqRV!|NZiL|-w7}M$OWv@#2M0ewUK2RHC zI9y$Bfd6a0HaPC;*$m_Rk2_M%s4QANXKwc^KL3^%v89;F9}8Pw7w82VByUAi9#lw} zRB~$CFd}erfycL8Z@&Di2Faf>x#gJ`!a$<fKgyJDl_#KiYL=wf)nif^A^(%xms}W%!WWnJ2JNdIBh=5sd<}xa&>RFctr8?J;C2_ z_U3L9{jksJ+NV*nVFuj(lx1ixa_sc8nE8!)R?@badKFu-^t6;z1s(sidmx(cRV?Hs z{a^S%gej6L3=fm?`80l@5kW+4r&}1&->7wc0Msd%_9)8rpA2okvVT<8@?YO_$2X-g zFL7z(yGZy3p(7U?ln(_DlE*hPB%?wque-pDgX@HlaSirnppJHRiety$_Mb;Q&Bm|G zW31B#1NX704GXA57cc?9?3k#7X>2lu8&Q_WpRk}{J|^?bY;h=6gW zv})@1|5j#&2yr?NsTE~P??JWh)Rra4J@l^{hhLU0vV;KwR8o05N2>2J|EL5?6Boijq00snU0;YhW_HP+QTRLM(`mU`V zp)&#qx{+wD#Fen>76AgvqP^bmJJtYjL~3`cND~MLc#AB1-qhVr+&hZgvYwl3B?CP4 zQux#{{p3BeO~~TF=G`M_iUefdHF86dUv_mRV8l!PT*vH1Nf;kLK|^XTU|ilGJOE&H zh-bXTc@K1!f0Rs^j_jiS+z&(8t zHvFA-6~IbDju-}73))J3V~l~_OsHM^$1dTk!D|h-!kV?GxE&}evaX_(|K(ZK+Bc|*dpL=8l?#HC2#EvNM~~Kcy{jdtFYEbgAyu* zQ$PM)OqXjq+<8k+^mVLHK;z2Bm~>V`qwf0}aZAI_C>R?H)v)7UxT%_>sM!kWCO`rb zb8cho(A&La>?_*b81(GpS?mi>TF`rt+Tshu<7OVO#mi+5daUOyuCsopjgifwzA z5J`PInOFT?=dYX9V)y=KL22<~oV}I0fs?DRc9|t7_`I6sQ;9r(=01cvUsp zV96v>w%z7#MC$;q_?sDv(K!T)!CN^+OB)8#$!BZe2u|vI+On5eEAuH|fc~Kjqb~MK865?2%Xip$YbB4AwSRxY=b7O)35dCAd+C4%>dn}e(cj9zTP3{27K6(p`3}EU9)SR*+|4Cg(HW-w56d^Hgdm{=MMH$e zXB-mHK-njSB>!QU*~cl177`Y-ryUbaG_G@#fRF+Dj(qi zbCWk!tsw`F+60!2JLFRzaRi=QtMYrfJWIqOs=hE zET(v?LzXyyCG=#ygYr@4Am%nlKD^oo_mec-*tJ zS`O_kMZsg>XE9t5V^Q6RpS+wtmT4#^ZZUcUH_2%i%|hfOHmqy0QzH zb%ciUROw(xG&;9QTEks%BRsv#`_AnxFl#mTvQPGsn2ki7gvQvg5jR=`aM&dFvWeeF0GiYSR5XSoGt6BcI;F$bSq4zm=eZM1zIn@CJv&;9Q48R zo*?JgoeB)i&q6G#(Tlq3q+Ojs{ByB2xQRRCfrNObSi_}^%C(WKC<`1>?DY{Yjj#Y| z51iJTU3+hNS6xr}RLWCVc(1Z?@Z-@8akR6^uDEbYn!s(pmL~I21LZvb82yy!JX97z z!|0m`o%C%EYS2=b6dA4l>N1~SUk|S1nwM!4+zo6QHOR&zk|9PL4dX+X^ePJs2i-#< z=x2m!{x;oak^GeuwOf0S5`v_$D|NNmN*y0s7AQ7ZE-s(r6Imu9H%P%$b5B%LO?(zn z%}!kDEJgplM-etSi0qw`$GinS6SOWe%rj@ue)6F-y`5}rTIeXt#hm%*`v8mpk5MgT zNN)xREpD}b`s}c52s+b|`srenJf~B~rJk!tv+2M=(}fClT1mnJVh~NfWn3&*rx9jM zMjfwcEz<^`F_SGty_P_N-uQN6YBH+bU}Jx-u-(jTmhf&1_w1-IAbDpJeu1SZpXL!U z>z818Q?@3W$)78`uuqb+i$p|APDpvBcJ`f@9eroF(WKTqi_eGTkcYHKt)vRCZueUu6UR|kSXJSi*pl03UlQk~*exPAl7n{VA$YhSVdND@ zS@hr1WIh>xXM*IxpGvwV*IDmeA+kClt-GR9q3`b`Xx(x%`jPNfDravdR(tVXq_54F zNnWHAgA|Bf6xcMEh<#@>iOX>mp~nWK&F*VvG#{63&NsF_dbk4A49gB?y^IDKr(&*t0|nO|}GP|D4F z|3sVbU_($ORqo*Dr=H_)agihmk`wt0E<{jV-Fu|F-FlTmMbO{r63+TJ8Cj1bzc9U8 zV4nUI(KS4`g)@9+M8}?qlyr+Tsi+wnQ=SjlNU2=>JhS~*%1%@Q&~?WCT*CqtMy)|- z;c5b5+eGI$Dk1{*sS<}>RB`O}>oW;w%p!z$R8!-USF)<%FCy<327nmPT=ek7GnJkT z__x_V?>u^?s44~p9+w1Xt=`+`v~vRM!zu{dy_mW(X%B!nm>h!YH!Awh%E@h{!#V6P=7eO`ec+2!zAImN{g6+L(qt$j4AVNeGg2Zc9nc z(}T`P7})%4-}3jAlgWh0zdvWm0@i=7&K2xNxNIVhFvPZawr*B+Sjtb8`M5HanAysh zsdAJO9*)D1w>Nbay}uGr8e37TAmN>G>%AvHExK&CQo+e|bl;Vz_b~pu>mktPgg!KLG_x?=RewY42S_xq$+{VVnWEJVOY>#RPuuDtFqtpNRKi7+ zs9A-UPS3Y<_^K~9t~a?90Rg1N_SZx2hz|lpK46c-60ha!rvLOJQ^g-DBhi|n(!^lZ z{Z#TSES>oR`8Fh3i_wbtud@jKt+~|U#SngWqfqm@XOBn4knk6V+(Zhr&wGo7);q;` z>YL~@RWD4pachxa-LrfUtS#Grlblqu_11sTDwt<0OSvsE`y#vO$FR}fW1Y*qmrM!O z4)1*9!X{`(s%{&j9G2%aSFf<+ayq!8{V1$p6lU&^C0R=u-}Y0+9Ya%sE8A^*G(PPA*|aV54&dC9c;tA&iAaYX&uNl zs{77MxjEoJBwL3}l~22^x$&XQB6(urk>2t{mOWV|Vct674=0Dk^NEh-Oa;!hJo7p) zM&;k#yF!8?f}@8k9vW7=e^cFf9o`hR%{2kZTSh~_I_zAj`eAiSMYI4QCVKa8s}TcS z55u~p%AZ);efOWBdE#l{DDp^dyY0H7t}VkEPjc>?5axT)zfkcduf5aftnbZ_@uU+w&TCU66g{qh!fMS<9RLZ6oGRCiBiO86RP6gI|pS~~D$%YZL6)eAj^ z50ZAL_fI3GdEFjZX#Ige3CDv4lD__1iuE&KJ^2?sS7@DQhkveZ@idn=^yvqnjxlrj zQsqbec)}nKL@)z^x*epHai2Ctt})J{_6r)^x(xVWGZxV5AC{KQ#P0udYKhoO%@$9w zUiJoXvOwSm=F}^Sp5oMfpzoyl2_$4;t2l$^j~X;LX*$v#H4QeBnR@w|@+!K3w`_%y zQQf+w(|!ji(IY)Z0G*L~Fu8(JNI{y;=MSI$A^(m1mZWb@^Tp?mRWf1!DNc+nghqBx z6OYFiuX?Hyy+NS&O#&Z5+_oGHpag3)`E6nLLq%qp9e9lAwAkb&_u$T@iFmg&*8LvU zZyfYg_gZ}|$q)N60IzG^DR}NFv1i!GYmw5)+O$@iCAaLr1F{l8@2DWrXYy8CA@CjF zd@+`nL`p7tUo{0=9S_!s#ds_D^yTE`f~Iy zAaBmNV&}TlKI1km6;|jq*pX}@F7#l}w)mU@KpM{HTh4Wu_Q<^zT`0XI;VPD22 zj?Tl6k;4prh&BBUCQeMpL3sSOb6S4-h{=-^KlrtCJ$Wj!REXeu9mopAL6oh`UzIfZ zZwR-aaz3!>c3FU^7h~Zw&*jp4k+gaXbXF1?(WB=M==3wI%kt`>#$?BhPN+Fb3hFW{ zV2cukw($0} zvW2|iL9FDk6(QzUv_(0PbXjqwCVUk~K9xnF1bQvG(x++)60`V|bvnY{zZ!cIL98?C zs1m}o`Ax`Z_foy@3_K9Jv?+`V;+a~P)8enf z`(BXQR=c#Dn}G@eMvu9}EIeu0B^2f_qn#Xhpt>G;i^AGRe&~1^d%Oi&o1!65^<@bm zZGN?64?X=4;I~Y%ta?apr4u^Cg71V2ioysMtBQ+~CPtMFWyq-AeeOtq@uj|umJ<0- zfiGFkzv#DzftwL!D_Qxni8lP&xXN3@53*C#)3JG)-RHYMrEftI!X0n0{rXwbkaN!P z=ZAr_17wu#o{eLvl1uM7-c>`Aui4f`1z zmprRm87f*%4t525@jXZ;1c|# z2n5GYeb z&VM!SE#u=4O+;Dn*ovG*_VK*GTSe~7QHd}m)9iA*AX5}%Vt6OHGf$r%2*(Y}9*DoD zhWu9^6zC;>U{@B0Jl^+6sTpRfpdk%^7d%CEms%@%NEEq~TRH_cd0D`E$li{iXEjc+ ztq>$5=wRArMQ)KF1tmB|`7)J*zJ=kjpY_cZJ}yHeMGToLk)#59F$W}?FhkDXoP_be zu;)Zs7~F?DHXAGa z#_nyj_%qHC;%toNl9vsxzOqhV6xNgB#fn`<_5JPAiWdkY(Mt|kPl7YPH2WxA>wD61 zEN?RW=Q93u(tht||Lk$}cA;biY?vu-ryKE`kbjGt8l=EbiA=ij_*3%Fu1mXW9?otjRH!7>FMyhE>vpHqm z3asne^Z{O_j*6Jv<<`g@9WSs37Z8TX{>I5VwJT+t&S+E+^eDE^E+<2?P7Yw}ehjp7 zE(B_O^`}1HDsZ8uJ!qn5Li%64eolgyo`Fmn0JM$LBH{#tSW1q)F7u9xruI^QHqc~3 zXd9xyv1tm&h!y-vNQBk!F|M8fSvc50d+tM_^b~%i8<~9Ka?P${XL+-9&mZ?HrRCol zuW)p{Fcu0>LBO2t8TcLs1MNQ`dDL)o2%qs_9>V*mxC8n^>hyjbYm(1j-9I&+S;+nh zcD~8<2rcp!{2M~mkHXAKHJ-+~N~*%YN03NE|B?NkCp=`uC6#v@M##wb)og-9HQ(d~Z7fmyxt2)0^%LnyVW+5V2j+``gbr|N|X1Ul^ZaCJ`;O?2xsO%^kOomjs zieci#FVKqe8WtA1zZ7TFW#S!V8E^uI0g^cJ&J(bu3Z~IUNC%Y$wR6D`gX?`NZ>kSY z9l4^&!x3O*Sk4OHi=t(&I7WgjaG9fQGJlmua~ROBAA^fQHT#B0{5)b+C533CJ0K05 zLPj%>D$FXIz#g0jsAkV5@K)EjWUgxg#(k1_mMP>~+(6H8cNeckBpLVRMw#q`TNnmV z<3AvH`GU5vtM-Vj4Z+$i+A|%p^oLq%D49KbF-xD}kCR4>KkTKuBPbA0(=Iidpj3BF z4gegrVxzWa(CD(bE&lP}WbMzBUpDPv_9g8$1MaCJH~)O~B)@Q=F1;tT%cqYR(?)AP zi$@X%6~YBik#v&fiw|4Z&<;lO^q=BvK7&WLYw;PM4K=5dGiowTS?rnvMg+a1K799xv3F zI>cy4+M0bM8;h!9oK!cj2c3JE9mAr_e=p}3#^+m>5k~;Alfz~uR`O{7*r$%{P>&Ch zQxpFDRdttaJW*q9;0R|khZDS@JKPvRI4f66ZSvK9+{?)q$*iEw+cuS$=GAwH2mddy zoAy4W^nPkKk03N~{SRJl=Ma%InK`$ds0?xUwrn=VsZtu`mAZYdF=*LES(mLHhfHgi z+df5q@c_ip?*&Wp0&(OCN&DB)OIv0juGv98`Kup??*Uhg9nq&5u;d*fMssrfk1A?& zlNPE9?iT99gwp#M93VzLvux8suE^2aX1BFa-`BFvSJJsz3yTO1+U5c)HA1ybz}if= z1_lsk4NI4lZ?7;uWD%&z8nJqff4A3SILStusN?LW1!`JL6Fs125UDg=R<|Hymi5tF zesOFctHj+0o)N^CF9zBQdPi?He{bRq%E9&l3q}yJx@4B)K)ipvWp?JD;G*J?*L@m$ntS<4) zvgHzYIM9gT?yO5Zl-kZ|tAo+&pGAVo2*-Aj0Yg;Rl$kk_ffI6SPN?=`^?LfEQ2<~V z*~4c*yrS%};*Stzj~0kHW~G-czGb!SHJ0OE{8r--E%`#JTW#0m&Gy}#o{>}wmV&TN z5}1-pPq;uFIQutKf~@@uao{|^9?(NWI>R7)_ZR1n?}@PTMvZUo)5SP@(S)p@)sH;r z>B>DNd4BC=;HOYRdBs2r`XYo`%`i$sQI!C`q!702og1$L&0dw0VOJb6-mlbDxTHw3 zyE%p+qn|?|qhZ5p+(w5xI^-3mjV8H>=0OSl9Ki9D>;Zl?b@8@Wi-g{4-SOjhz%v`Z&)=nIfpViw*OXL$r{3&7?!IAY5akBdj!~n&||G4w} z9f!G42QX0yT*e)Nmt{y=P;QHR@+$$6XjL0=n3aP?S;7}FxI{YugDK?ky4)wy(Np=v z_PzVm4XNX9XdX_wwrXsK~z-6H#s2Qo7mb!R#5R5 zCzjd;3BuYT;Nfdt8KkYOf{IWfw1NT6Z$qgdSU5q-9YjkaVyw`&dLL&Ro@UGx{5P^`8;zA~@7n5qI3H(aR0EXJ$9rdq~kH5o{Ucn~g$b=mND6;w5_Fb2^bdH5& zx3qGihZx=4N6}}KXYO+-acqT@fra+v4_xu-9o}k$$#Vm#FX9Y{;0Ss#I#veOm=aP) zxkd_GN|%KxS|~^}kACx~VCjFY5PjM#@O@S7Ru+mE?hz)(DEc zXAVq9J+90ECV90CEP;@K{LFw_u2DEa6(sYE;Ejytt#v2^HH;>|bp{tKj>rnDG5a3z zXnZobICyq-B*60^?n^8Hd9qKld}4;lK&WaREsCZYrSek%lL28e&y)O@z})hu<;7h> zUG#7olD{;wGpxnIIGndp^!MZr8inA9jn&NNV#C=1zn54^B5;IF@AvSeqZs#COuUm+ zYkU@U;!+sv*XDS7Z|$V`Al58F_^0AqKwKMo-c`=uO1kRprRnz9L!7>&n(r|k|Bw#i z)VjCLpmaK}+Wb(QJZgbgmxtGF#_ZJ>C(RA~GvplhWA8pdmie6eUk3%n6vkjVa`VzD z2a_XY`Y&pa{b2ruA&3De?S_US>QYMPmEV~jt?1Trnx2E5CCfu!NjfH^F13UiU|W(L zkS^HSv;Fn?<{DfKr4AOPf=FToZ+6TNmt8gP4beN*4|#4;xryA9=a^@i6oj?z${Ix}t*!Ny<&pF!tuXE%Inc4qFwHx}^d%$n-$9_N6 zUfy|g$H4cH>wU^T8s3trRg5HlU9RX5rdr#@iFA&-=m?plEKY4DPNGu8*<<2odbfz567MDkfo&6@m-R}U1#>t(d zmH-#F!AA3yShW)e7v3A;d~!HG!$s9KFT2ke;P@uHoAiR);B45RZF_B zuEw|}=3X?iQxcNp`R55VVf7QV?l-}^Vz5D!@)~IUhoD9uKpdWk_}J~*yowW*q%kwwz$pI zSE?~2au|~ky76*0hGitur%ZL3J0i~ZySmFiR4-39`H)_=wWL+)xRpnBS2T{go>Nrd zJ3imY`@{XN6tYdpHc8a~@2xFs>{LPKlz7F1x2HYNGu3^ZsS+L^R5b+Fie8sQBY@Yuwrp|Si>)L^&AE^%!Xm&8eY($&GltA~sk=s^ zOdWM%YP1rBgJ`WoA>f}Qvmw+woQVsWlb&7se=Gqji4(1MFg;9ym^nKX7ul>X5X=)< zHy`W9^XZUL2=&zs9$|aUjNbX|Qr?Xb!jcRVFiSfC#@h#=F9`HW7kv zZgOCE^6go@!yww%XI~#SV{DjenZqih1bWD~Ytmt=i-H4MAgR0}vYqaih3F{QJ!D4` z;Hy&2*@9aaT13RJ>_p$yQ~)P-3AoMoaNxN6QqPT0R8_NTUcdw7E6VAcXq;ZrC38 zI|C=yjK03jIg9E2%OHep(ls1xS!vPoKd%JU9I?v=V#6!w`WQ|IfPN;LhA5Gxd3=C zt@j(o$#yQs({7koaMAuHrmLj}*MF!!dtn`8{{V$DM_{<1%E%XuytM#gHS@~bUzn&Z zT!!>6zhM^I3GL-q6<{8loo_f)cM0bCN*6Cp)&QJ`m*aht@*N|26%B}BE);*d^?BOJ z*UDlvoW<>d)dj@KEb_^$zo1ROrjeZo##gzgR!W~GLCP(aD#_klb@11RAX40F<$d?l z8yb^|@piOmWG{m_y*>$MWbG)FO%=}X7=AUe%J_VO0zoZIgh1gE!Y*=#Pxl4ekjD23 z{SD+YT}5Z##569$rTi7RT6o4K1K9kJBcR5)pWO&`^317-Ayfx7eI3&XL~~bzdhe6M z*%KK+isxv=O7LXE?INYYmc<<6Cpz9C10qL&nml{1?yQH-Q17!nu@$`m_%b|N_&n8- zc<*oqE)?QoRgxA3Y7Xh<4Pd#l_<7Y!g0~fme|u*jJy#s-b1=`Nl4R^eG-s^ z?)9ag*F%I*HiRyo<50T=R13cZv1=J^8I%2b0J~*W&0ZR@pMn(Iipu)L3Oe0l^20x>fn_D7zMHX0?kD#W8`#DN`SG zErz(hnkem5M{B)cdhM0WuoG7lM+owwT{?wak5~}yq9lw?8K;>bY*UB++wX=Uoy*y| zd_n`X%OCA^aN4M_@J9MKK4(U@gQ4yFRK!2!3~F~&!Qfb2a_0Axu3tK9CJ%C2@YOM+^#B+0CXqHgQV0t zHZ0Y1VeR{3PRp$U-q>J!YV zl%-MRc?zuK#Fg;_O|u%ul#ov)yw7)(24z|JJXE-)zD4Ak`u2+iIpMyZ#T18ucWu8J z^xOjcXpFRK+!ctNuWXtXULn1M(F1_CMAKUTJzDIxHEm_f#~#c4E@?!Ow|e)vFX5H> zYxYI_!(v)Mzx~hq6_(qkvU+X(h%leHtCqdE_hI3xW=4_mi~`kNs`MNj zAdI?ROw-RTYdr!-A@|n`vooJ+=e*Pc?eF)Z3ZEp^9TSzO1H_v2_%Ceq`B}Q(r%4gL z-l?Qbzg_rFE?743q<<+01aU&Cj$3PJ{q4YBoktl5vpC%=MD~_^d^%$Q2X(E1&84Dp zQEfFwYL28kM*n2>qUCDP8!#baGzvLt>{Q(S z?N3wChw?DeCIo~SW6wGDAtfoi8Ct9tmgWOE?CARG9aY$i8-*sP|H~v%K5R zHP#10o{WA8z$_=xg9vSs?@E64V6T4C%77t06`v5Jy+Fmyf|PEehZM&JbcU3fhG207 zR=kc4@N&beM!G)jjuk?Cj5@XR_y~ zd0Fa|vLr^kH@UY15Z-W{sblTa z%S+b=hv`2ChFy>RpP&74fMtq*oi3ydH>C_Oq_F%vtv*{EVH!seOJU*A0W@@?e3rF^_}Y2!9iL3WsPer2eZ?o)#SLaFh~x zkh{Vl1a>_&WSCAeI-F4)AF1?)uiqY2{AF>^zOGp{0c&93NIbJSB3bP7y;)l76IhwQ zyju~}JiPs8+1iKxOwIFi^D_bY^BddGW;ij(^laU$B!gD#tW#dhS4l;7%2kJO@Y>^gTpF+ zt5dAPWzfqv+|fwgSGdiyXB+y$(iTwJR_jondf|`5)foqiw!vMf z^Ka{^7fbBKZ`LR`jjeF}8Sd4gFyh(?^oX)B2Khl>9O=P5yc8;OYIiVB;Ca6L>-W{4 zO)zkq(I>SF6~cq0AxFk03Gv54Vbn@#mV0f1KSGq3Ut=AGS4%HZ8EI0)=`*2iy+2va z5I|OcZ?3}v?pgTM8p+zLcRI`eTim)_FCgq4Ze;g#9>_m;_8m@UdnGT^c7z1awj1}B z@d@7CUUjPZVrj&Z>~4?dIEj)c2FR89bbHIaRDI1T*G2M?X~^>;b^Yct%X2y|9qx?? zHj@YzA!Tu}ej6$6>*_0K?mMV0-cZ?+P6lKu)}~@~$i_?dM?yrw{?9bp@3AqKrFiBl z>4B(Aza@#b3RnBch_Z#%n;1oqu_n!`SS=X8g?w0+w4Sj+Q~K2KjGL)$%@pA{eh!Y3 zyXLtU7iz9HburfRCp5sAn?Tpfuy5s9Semw!6S;qp>8-gQ>-oT%y0 zIH;M%hI>E_>5D&N{#6!@%J?70&IuUu)h3o6ohDb_Sk21 z9;fl*0b9F2Bx`Q!f9}Y2MCT8{k_TohN4fe8kZqDeNAy?C@E8&d?Wl>n5wV0EPCN-j zovEC7oF2J2o&hnH@xXRxqaG!!88l#E)%srP;RlBJB!8c@(vh+(tZ3{VNt&(z!P77J zhGN~8Z;g~DeHsp#{E=A4ys9inhUafd^arO!IcZb2)WJxj*Xl%r;yobJDA+$@`MC?y z829fIHSV}6$?_>_C6Et8>ueg{SSby$dtY6}{Vy6HCupG=BTbgVb4!`ng@)P=XWmBP zq$K~~V(;2QPv9#M7_}4%hQ!StbWL>Ev}?SoTZBJK`-)wF_6Lrzc@;!G&mo(i-Z8nT zI%OiI$qXW|j(~70N2F;@2!C6D9{dhHJrTrDnYiSQba>n@GR?R&V+R&|V#IE;2ht6i z=GjQ@k^LrVDHl&Q#)ryqI}%4VNxn3l(Z`RT+j2{P#`8ViSY59GPQJo+lLm}XuS`f& zmoMY}i*=YhR>UW~><^w-8qoKs%vtor2ZUdeIPS2wTj(06g4Sac35EegZa9+DpGqfS z$x)T_TemkKYGp9P#wFVgyUlJGdgTB?QDW`LxEoL*&8WcPoasnC#TL}$Z~S0LXiPJMLTbu{6&RKq+cU-0{3!wK)A{o^DaJNb z$=$Dwsr3O~JLf{$E9RqbetX~yy{qFtw?tL1UWkb<+ZOUF%liCgLVk}p4R&BC zxg)U%m%LWNc&telgZ*WRg!Xvu2FUHYAyrEvFsP&lg(4LUnvk86{FOt0yh!fj_$$Yl zX}>mz2Ygt|B;6hy@Q13mjI`VyzrM9zMI?;gK4t#N?d+qv3EmWfRhJpr6fJN>o~q2m zK}+Be?AXM<)AWdKW5t&_`MF$?`5w)@IFhuQTe*H=aL48$eXl-hUP@QA9R>2P>FXXL zk8#zPJ?g)#@l=6fPD*p*vfKIGeD7?(s7Dsl%-3#*R-M+Eh-@nG32?v1PwQ3`@B zsgv!IH?fVG6vAW>?`hKe51cJJ<+gd-){P$x58x`^5$j?(Xfa+(5ueC)*;O^6VX#wn z+lzI_7<7Q|pq2FwW~Ik9_qVUo`NF-?3fv+c*M7a3mR{-CtmqdJC8CrI;q~CKVKCw( zubH9hk;vf>Vmv#0Ou4PX7L@)Rc_WW|ZmvR05e1v$2h-d~DcxjjI6`6Mt~@lpL6Kv~ z8t;DgIFcnO>W63&)k~zppEH)|Z#hC4*#lhic=QNqqHS|=oH)p$U~z+*@rOws&02dE z)fc$?o^VsXFH)(*Hm%n1O#~wmRwMSBshO*n;_oOgU~rO;Y2nmEx|j-l<$ zGLi4rsX%$WW`LeKD?Gj$EG0ZrvNT|>GaEp!Gj4{5C&480%7K}Q>%m2&=y#@>ZMBD- zAHwfr66-jKx+oqKujb#$t{R)4Pq^b(>sB*Pv%|22zBD zrcRy-|B_@-${`J+KhoOw&E@+y5dHix>(`(;an#;KXM6UOL8yvn_V9`w*5tN}^I4F} z$oY;)ItkEJz+PD-Z9Zj=$Q&x6p5n4|XH0Ba?&yGe=WGwLqUu_dCtoAGMA{B_sm6k~ ze4&gMJ-3yfrcux2*VMK-nD73KreXT$w(=Jmx+zx5Zr_^{46B?A@1oE)e3=Q-db|(r z+gI!Kp#saB!s-dk_*iF%LMCR3~ zapong`1v6IyaofAA57ic=YU%F#Xy4!#4Ii7`%02fO5pWPm1JP@k@)drj(NU}arh*IAx?ylBm@APO?Ac|Ucff&>v0qI_s{F)~GU_1mF@ zCj3?qsx6|`N3hBL?uCjUDrG^J8I(~6hUYu-83{)#^&|TF342602i8cn?3|PI?uf11 zZ9$h{hhz~{TrVgzxLZpG2VUnQaJd>ucKh+n*u3wRor5C>Z7rO z+oLmPEovl`o+~X-LS{{l-E%vv4MgXrm+uUXdvkx6x-ndSKHlT#fNg>{Xi>SeUT7Ov z-$Q@Q#xudGv7a0q!Ve{k(Cx4fOQ`Sc-m&Qd%9OPxlg;`qu^faYZiw?`oZS+Zvz8qc z4wAMnS7~JcY~lJP%DlK6IXv&q$mEZBeGi9As38lTE?7ErU?}0>kx5TNK2QPSr_*P2 z0x)!1=#AD-halD!RlCKQds|}zpy*9L!IL6%?^fcYW`D(i9#(=*{Go$GQlN5OrC>^dxJxATV$P!Q!x0c!gOt1Cr9 zB-9fw4ZJ9>7f9(qgKGC#;}CQ`a7@4h%186YIX*ED`Apyy5Y9cKdvhpuZYwuk9)iYk z7U`BZy03y2Zd7?@tD|`gprGT1;_=|U7C!E~A=C6R(bA!#H6tk_^xkksjl4ToStu+j z_r9d4)X&J!KJ7ebY&)^$o3hz=Mhv~_j-`-@YRiG~jh)t4| zB&=!8OF|AFbtbCw7I>JPDIo$L<<374lEbDWu1FgpRZ3Z&?o#epj1P%bEyeHL6ELUv zAixCYYv%UAk2EB3BgzWzEewyh;f`?e&}6VsGtc(HXenxYCdjzNgC1B(PEQca-%@(a zf*}dLZ0nTx4f=~a@m&f*(Dl1jv|-Hbwd|Imsi_Ro?DUPb{DmUx`uqVc6N+?^!XGl> zf?QwZ2AQJPN&X&vj_>Or`5_Oh9=DDpweNkSC#9l(Oo9~EZy_^DN&Vivk|(Jy{m%E` zMbTcDZkV7_UBy ziIZu`Bp+gzE5bDZkDYqL1W~>@rv-C)aRGqvbvEUw%Bu{{p2Zj`V|)2NqJ_*gVozCG$qseu!j( z>SDI{Gdwqpnhdhu=^Q*buDn*U%rcI_ZKt875yk;;kx_{Y`bzNOW+`q2Q-6FTsk$Elx6gzECytVS(S^Trjr;r^=& zFD#n{J&lJ`meQC8qG#beOQHUu&Opt@5Fe?clPEd97r@m{ZbZ{N1 z(?P6T^#?Or??Q)xYg$&ec~L`EO_${$>5A=K{k=e>K}xdVXO39SlM4vkTfbXBY}q%bP73Pe67*6!x+Q@tNkvTdXHeKN zVx(W$iQ_S#qZLhOg+n@MO5ChW;#r3g_x2f?3=jGWhr}DTHn`0@5{RrpZPmFxg&`~z z0cxi*u>E#U%Js_@$d zzhC@P%6g{s22z=bH}2?*=l+#(3ONZMJw2H+G^JBSMG_)y6E}K(f>vbz;=s67U4@@caQI z8wY&AWnXr8j>Y8fhRA5mBh#6&eET2MmDY;BR}rZ#c1|)giXiT=_4zUnMBE* zivGFqezAqL9CWde<8Zzn)IWRAOEe_DDK0~6LND1o+gj*FwesHTrV%oGdzmcPki2jd z`mcOE3h3Xw#onJ8((SSAHA`zNJ9ttyy@BhHJ-uPCkFdgVM|3$9?@UW>Zc&%<6EWgh zvT;W}xMr!C()$Zbl3r9RMCReEpf?3%s-GT(-wyd&&3^l{>uJ?+%H)t2P)jrAxzfoF z#35`TPuD9N4Oq#uirO3usoCoI%FXXvD)@;>~HGQx@3(O=FePBuXFe^}6{k$)*o9St&&ErmaPeJ^3sC@fx! zHd4(}h1brlnDtlM1;ND*5c#^Xvg&ZZKTh5ZYtwV(;N1qd)WBD6() zvK^qD$9@$c6!CkS7MMtGezdy6+>CT~=Lc1`nF}|_&2EZ=&8V%D(GM0ku~S!u3&$>V zwoo6_IVLHO6wXE3#et?U<|X1&qVNUO;65*juP4pfRPizy`gYA>F}3wedveiS$!iH8 zXU~Vf?Op%Dr+yh#Ku97J5i51AXZe8H znu&zjK*mme3-M$_Zz!1sEqg&fVVpu;TFn5C>e~kGUsY#ypNRRQ!Lg1d$Z%rFxR@55 zD{Rneki8Rmb9dOez$aLCGUI+lm<*oZm}-aKF44Ri8JMuPojJ0?Q>*U#o5$g$41n}_ z`*VmFPrhGLT-(M;u4+qro4#l7)`ZiQUse^(;UiMF?DI-{6Q9($Suu^_|oy#hIZ zqY;iJAg&i!Jw>MD9w7p*x@P1Me?^n-70T9oxBAW_Q9D^rM9^mH?3IwXY8zUC$C{t* z#QQI2bdJN8IQi;gPMEcgOhRd!+rK`dmU~o0e62fmVMq5sGABxd!e!D0RUhW>6UfAz%lVi$TCzRFsxD~caeLGM!D+3CE z&osA|1zIJj?F6D-cv~0JX}Ch;0wMe>KD|#FGqAQB^-u(67_mKajd5T?&kpF<|(;@2VLrqtb6GkGfVNc=iGu(XF zH1^=zA2m$vqAs_Buf*whWd{qoFlHp!rydpGFdmkKqUcFEr8yug#D|QF17eMkP&PSr zHwN|&#i{xyl<%ocg_r<{V3jghYPP-`$3Pk7YfjM-X&MWfxsq(p{_7hxkt&ixHvn4K zQ%N$q5f(2n+Gj{H?+e|qR*_L>NYT@v-+an;4+b?ay8F@Y8%*vu>B}~x=YlUY`BW`Z z*BhI#Pe}x~&R3I_-pUDHXtu!SgABK9R*e&bSEz0!oXqlGYk{u9a_1Wh^#aZ_M#Q18 zVK*L{NUN{;J_sN)9w?zaOqEMjPs|Q~i}!?{Huf=Edh=COWjt2r0OLJvB?@qo6&O`d z6!LTEbO&$I_GQiL7+m~%xj^5|zw(0yLg*}AFoJEYCAt$3qz?S84X9lnVFj7I_cu7! zxq5j)Nm-hemzG|5Jc*XV&~r3?+wJgmi?>E)ALsDIKguefGFjuP*ZqJ)ZE9MA`Tkg_ zW}h4(V93XcU<2J}9aPyAw8DnUft1$?(BWL(LtX!JQ9#`^VMmo5aM_+=JCX2_;t=Ij z#slREF0Yq_OsJ}OlOp4}m@S4sH<$aM`*fJOc`7_iG?cw5MRmVbl4h(upE@M3JMu{w z3*r|t#L4ALnKt@c$&TF}={@MOI?}5ji^$^;M&Strfjizr>|t*zKp88pH_!mcQxU?l z;;=R-@6VdZj4!zOEZu@!i#+Jdiz4DGl-bZRQZ%;%YGy5U+3V;+$~#yf;NSCE?-ktY zTcv(AqHDnz-=x)p6?n&La3ekswpq9#nld$005!K=)FS-Hn^a{S8G#Y_>RTH&sg z^rLa~1hC^D!n@xicrv&y`~Du~rK1z+r6c7MF~=kRmU-EH-Di3YpeLKhE{dY+iz){k zS4w#D!T?oVbx_#+>jXOs(J$p6>!S^~L(hg>UCLzt1?9+(f*G5a!Lf9+8LBI-1h?PH z8ujwp9mOB0ZUd`=0?U`lg*^!!@r$02R=KyN;=%weu=9G8|G!>_>YS_eX6=i|@7 zoUjGvM9esVVL1g5=E@(nZ2NvzUa%Bgj1em{QK&6S9DT_htQi}0;ZoL(SgP--SLgMh1hl^LnL zr*i~{K_%|FAaP1Ou`ZX)j1~wl#l0eBTQ*U%Gl*n&D2hiVeuI9w&77beK<{-ivtxmW!*A&zD?Xw=a6e`Vg*<1Kg{<{x3(_nQNEb^`0oT|20ZIsjo0I7P7dEeA zOnzrm!A9};Nh)F!D>(mkwSrwqb|%-T?2kVJL+qPmtJ5$i;;`#W8q=pnGHsGHrT zXB_MR4lTSl{fGZlmW<~!Ra zkaNl*nMU+9rQ0qkNow#E&3Y*LJ+#L!o521wCE#7M@TJ>4y9I@^LJcD+qyqy0*F^q@ zkAd-F)*$frTo5{?rqjCTGwNTLamgfy-0v?rEHsTS8?q_+T^#xf>$s!Jr?}Y#I;)!> zBC`}$*1~&rU!#Wfp(~xALu>VD#c6yY(*QOUDo^=1SH-m4J9!bmf`Db3I9u`KQGOV= zu@+*R-HO@w?Dg}Qh6hJ3jx3p2&J#7wtR9C-ED)L3hpf7%>=jv?MyhMuoluZ9W{Ty? zy#2?i^iF!d3zGL;{VOoj1`LeMt}u#QU6lkK+9u(yX(tUo^yQy8FFM|RBI@xQu_xdA z(sQ!8z5ciL$h8Y!Z8LRu+$?n$N|xA$t{UA8KDWVcK|Et^&?{L4V0qmhcroM_jt3Na z$#1$vl>L%pUkER8jDz~Yu~t5zvze3SSfyfSV#sr9@t}?*K!O=gEQZ8;{$9+|OoffZ zq4~r|$B1aD^V7Cu^{;>9L`0?XPwU`@E~WEWC%e$lSQz)UlYOyWk%0!J;_K zO1W8ZB4n`mNiUHsz&9v-^6fw!x+vCyhG@T27k0O~QcRZUdv;BO&*yTc)QNjb@(({Y zv}%AWCWZ+>(?x-8A6U*EtH>2Ex)G61^+`S2>v#KCVwilv`)Er?-yyPsaa!cGB84eu z0u3<+tbm(itVMV62Czcag`AVhS3Nz#p*nKW5rxWR8n-+RSPUacA>#(XCGfx{*MNiX z5%6-l@5%lkTO|R_(QrQ%00<0#FQ^6p1V1G3+dlmlfB+H&0EEQJ2J|w+V3q^TjuY-) zT^vh=1ma2NBO0wZPeAYRl1UPY~o*Nz{uAw?u zB&NJpN0jp)sY5QPnhI$uii?~*)ql~WQNaNj`7 z8Joc5jo><5O~!BdFvgStW-c+H{pN^@36;qNzc*fhHrIR6=1lS;h{g44dKNH%Gep{L za2Mt~N!*4EQ81+~+o1FtCc2{L;f%6Vuns@*K&tH;Y+fBz=|Qma&6Qfkmg2^fqTf}? z{gl)Xn3pg3ka^=uP-fuY1tvDNXNXn9)GMLum=sij{zGtGc(FM#(?-D!v64;U!|Sx- zOME9R9i{hyJB%r)@3<77qPAIbiLQoDE595rLqGj1bbNKt1(HaZEq|W{Y!?J5wOJ8+ z`l1KuFUdSR8c>L4;Z7W&573rB2kG_i`j)uY9Y;wTf=|46Y$- z{(x@IN-*|n=6V!yS{c3Vvpv!E$1nX6eJIKmLV@szrTYM;zo;cSVxv1s_phQorTkGQ z@h^6r&(Zf-uS1N|=h!EDpcWob+7#*)S*USa{qtKtSe0*iHd>*6zdFbYr4-l!AjOUiy^eV7^>09jMhO4^&6w?M0*8YUs}^_X>ecU0tKn}7R_;PIP%1uu5=AqG86P0q9U_XdUhR^NNQTWnBQX_h9@2gs}a3r9fpP3`zZP@6&ISoI3QV22EFud2lCHuAPQUXu|-G7#NLm z5s&D~O(g6G(G}prB477}_%1PUkhvN^^W3Ret#>_r0Fr|y!+a`s5_t69MFwJc^mFYI zA(TP%3c!2qu!89DGujgiONOe)id>mo9KlU2*1>AveIT7a9H*H^~a* z?&iAht4>rVPE?3u>0>;G)^IT{v$8vTB>(XVs0q`V1<^wK#^R-Jc#8T3c`pBo8w5Hl zR0GtDit*j-w99;tMpg@K%WUfN9`ZvAzkJqZrv2gf!>6o1gO+YWEc-ebICG!M0K=l= zQ+Q8qTHkw9`wxq>2Xs$A^hwphqIAr9jd)~xj=xsoJr~x+ST*|~>0Xqrn`T=Pt#2j8 zz^zjg-@g|3pW*R#h3mu__Cg7Lof1>H_XJ_5)|qnvRv$&^d-SVdeO;crYSpeqNQ3nu z#odzO$xbg`UYV%y%sHwr2yd$p7iF` z*EXPn>=roi+bH3`h0pp5eLV3p)-P|(b|pCc^-z%TMD-CRO)a7l#UtCbQ)kaw^hn3?YG51*T2cd1p&hGu{LxM z4lq58-m{{IbXj||{xCPR_HYCtIE*m$GolV$DUR>)BVSAc3jw|oE;Kv8PkE>E4DW=4 zG<>S)9@FATHzD&2APgC**OtkRL~6^301Is?;mA9=@%U5HKJ}fJ2f-JH>#J#@Cao8Sv@Pg}24Z?a@GBC- zZM0r+^*i55sJ?$0QoThoPLVLcLD)a-jNPcTqv7hSTk4-MRo)D98$F#>^F)3q3;1kV z89t``qX&N4j4*Cv&WyyJa}vl8KdhMOhgCT?_M8X0mmG?E@=aoXyDLkoOBEr?I{ueb zc}khTOiKiwG;lEMf+g` zHr2 z)c?G@1Cm5efA|+gyZ#JA&C{8*D|G*#+gDSk*uOL013zI_&-YJ{TS(OHOhbXA1pzhn zaSSEj`KSp(zxY9r>ye5b%6-$Oe{+ zru%9BZfI|jt&W)fqiL8k;Dc^xh6~ua0nG&k5q_+Vg1SKKkG6t--4tMRFJg*3bqc zcz)SS(|N1399=9N^y03@|6D=O%sL+9eIU^t6kuVSsEO0R1&^flYhHK3qqb8r*99t4 z0gM739G&CmYFQ?Wh&o$!bKOT`ZYYb%`(yC|Nhh?1e$C9@M)tEk4_6RCkd5KH_{{l0 z!9`jqLS={N@@(E`rn-;V(k0JHt@E78_h;{onZQ5%Z<->)YQA2gW?z5xb^lWZaS)Np zXqZiLWiq41R$FzMl1^3n_;am|nH?J^@9x2sOk@X6X^UQBr@(+jFcJ?rP;D_}op!ap zdrry2M}L*+4do#aWsMc)k@N}|IM?7D`7LHvoQq1Z`+TqPZmwJ*XO={`6=kFuxkRE! zQ4CLulD>^k(-~t=TKksGgE*l0h;oy*i3MU%+ ztQI8a!sv5qA<;0t)91d$?d))?RDfmmzZ+~6ng$75(ae*V2>sWPa1Od}%&g;opLCII zMEL*P$1rxUxER?ZmQY*G8ji+?tfTzep*G%Fey$IAV&jQDrYu?a#!xUpo{|S1QP0PV zHY7;0ZO(j1Tj&5zE!p#OWhc~%CWYQ5nHC8@ zedCfm5O)qin)#dSs85C=|8=`gHs12(#MR4*9;@NQ55QB~yqAij_)}Fl@W4a;MTZYTFAr{A%%bz9K0|gTN$BZm4FnhG; zXI)mf@*LJz`rPs00r-;Z|2_A7OqlZ?a&IW!1P@6k@C1jFZAI}W=IMhX5*}BR$BAFi zmy-fgh>mnSCM0-X@)AFVV@`sXhY-&jBaW7tS0;YprN5dl{e`11^zZLjoTrzO)O|as zZ$-londTIv(YcZ^ReTW&VfAhi))ab$=xl(Qh4F{4#Mg>O4j8r?hTo=a=P%6=zBJQ` z6*DpV4Exy1RS1F`-Kzez_V0P}k=0`PQWK0k%D#Qk94fNZ$8mm2u;p|D*IwTo?sX&c zd$W6n$aqXl?b^&$DE$95+0g*2Q5zft-yrS~u&U4OLS!z&0KX>V;Da)=5 zN`p3vs~RFe$YsktC{+70ZI}dF25=0ZsA>O_TXY9z6~6?Y4P7y5RUa_~`avf7-+|8Q z-?(%~Nwto8V|jeisr^fijkdpN8o)VFk`j^lafe#*sMtS&ODhlO=)9~TKu`}dCUx~8 zrGWJGxSDh!BOOF|AFDl?lIAZTca*s1qEaiZ|MMr1`Yx!VVKIE63C7<)!5%47Uh?dM zmzXPpi=z@N!r68}0y4K9>YZJMPLLC7xTmaC&xzl!rRWxTI!? zL9`&BKi)f!?+2>a-{(Jtc$X21&&T{Q0N4sTF-G(vij$svOOTp#1=E-bZj+4G$qSmN z5xkQRJTIvjjEmj#i93AnY=G-h_L+dNjhUH6*%*hTWeDq@TxU)yY5SOudmEpLLDe^6 zAb);SZiDLi==&L%$7w}!3fdIUp;G%Nx!)eX^+%cCee;`zfN*8S#XbdsnWD4KncqFg z6+J-EJN3^kiCDK>Jik$g$?mpN`_gX*@h5CnAOU9CM&A^A%g_#=s0|C>J(m{IY}QXa zly?Q!O+LIg_M>`nQ~*1;gZl_!GMXJZqUn zm-|>`_kSLQjoAde)@omQhOv(on5UC=)rGr3*d%IdP6}!J{5y8~T9t<%bV91KZfR@=98=-5%kybn+*>agcW*vJDGhS64jM zIxUz_XfQ6a2#SpMfnVlui~>{PGdIralSbv2EUdxX_ScC5q$E;wTVK^rLDu#506TW^ zFq0iEcK?OZNz78P{7|;xd9_blkbWk3;>gAcC9H;NWp`j9T>XG`Mhdfvce@B<@=v$zSQ)b9_&@YJNhqVz%hRbS4=en zKTK9m0$Z1ikLrpOp?eN)C>L7-tC8e0;SH~RSy~!&2qX9dHm;=rc6bM2vUw}l)V4VE zHRQnhys#o(>j%EWL;(>_f>uYjBoLvLXUeCKAk!AI9ZX%7()7J{DQA_GF7^jdN)V7) zFkPk29jHncYeDiAYI%>XgS|@@ed0ahvcw<>zRSBGOkvhMz>OTiX}Abs9jap8_g(X^ z{QM-EV6IEDVrigs9hhB%P$DTMC1qEttfHbT)r(6|Th?^#Z0sQ?Dk?e-Cx%SH!TtI* z1-2-~7fcl3_2KQ?-GG3AtRPI3ri|AwXI@^eMrGCQE6ZWtSBEnjvufuS7DO60We#Pn zSM&a3Z9C+nSKV^LCkhroU5X*ID}Rz#LaIHoXwFaj%W8i;8o$TaQ>Qder)a0mB4Ja8 zg=(b5|5Lb?_RGh&8|>DsdhcPbkjoWjT}8ttJ0dK*S*y*fi%1I8!=c6y{8k^LTdhbW zNK@JCj}%>bE6J$5)F=6l#@-uz!<&h5SDad>^^BdA&vyK_Hj>;^zG1jECYfz&S-A`y zcBIwHWI7o<#9HJ#rajvpYH-%nM>MNOiw*>{>%jGm9%XMDYLQc z!9x$y>q07};_LpLBoiwkbC;rO@F|U-W(C%)Me6$E(Nb>0duD^4XcSTsN8sV3<9_?A z!T*pYx1>dxyrEUxKh_D}NI6%AyzY@I{=x55cOOjgmwwnaw*RQP!!iQ(T85(cEfrdf zKB9P%)JVqkGEt5S-q88KFm>ige)Y%5oAYVcgRB+%D=Jl)$8URgr|xe=@-)q~=0tXr zc&Dv4b63thuG0k*)y|A|aw6$m&YnJ74(Jd#BDe1I4LyGp`+ePt>wqsPDiLxOA904Q znDx2+AB&0#oZwW&U6hKLZ;tArug0e6^*(dfK$ErDc#ahvdRBhZ8<_OZ&9mC@e{Fln z8fr9{vXB)T3G--ydE;mheM{e5Me^DB>#U8YH@gYy_|XqK4f+Ro&88;DSQc9a$WF?? z+W&FsZ@y9&CprqJQMXZr<{+6dTOSy;0+ppceVkOwxM^&B)`&%@1#Y|4^u|i5g!pxT zWD3%z9R$yNxzNFh*|)=r{04(+s?KVi8mJ?_N|hTfUvS)%oE)ZHMgUXbxA;7~i|0P9 z`}ibbHtiSn>EYZGfm4iti)M&~nzlO8mt|<7VVg&tlcuU4o0f}3!CV;LB&4_~K-=Ub zbQpfp9JNjJt8y=_zZScqUk`;P36h1?z_(s@-Ze`notP`;OZVfTpRC)D+x*{M2a(;QJ?-o^oLbTmLc#T4at$_yUTN>o>OeSV`Ki7drt@l=MM&7|EXDu0 zH`UgJM=3WFpBXSOwB1UJ0uuAP$jvdi3$ z4QC$CP(F%i6BMj>!dGRJFC{M672>u}2x0x7H*5PuI_GXm z`UAK3n1THWcgZ&q*9c8b@IDe`N8nT%Hl9X+-s&0j(J8{-Tv1Tq$^Q=C7BiWSlI3Nx z&M1f^cak{K=T~dPIW?N~HpDN}gSOBVIJ8Ckx?oXNxT{$Q$mE1GTvIs^=u9XW0`>Y# z>Z%WUMu~A)w;%AOES#`Jq~HjVaYCja33W zQU-*;WTAVF(UFg9=;U}8{G_qlgLE)@?1|@H-{r!oKS*9uCv~8vxy^Bw<(O(RLL)=| z#_=|`P5Y~o2gNt&`WAwT-E-TRcxS&$*G4qlW>LW@rAgOy3oG}`7dn@{z|HWSkoo$R z)d7bJk-`0xq92_b;L|Id5;XH_tig!ZI|M54l9hgUCS3y*(KcL45nF^WGexKQJ+5Uf z%2B92$ZsOc1>=3*m2l!b-b%U`mB+!3)fs1GH`i8cTjsqCIdRa{LukS3_9jx4MCBz? z{@@8|EZ0rvZPH&mUsTi2K7J@_p#>+V7=6>N-0P9?*K0U|s{~(&` z7uCbn1K%sz4wq}N6zL#vKI0vyJMHEtQ+tpQuqPL=JeEi@5N16^annjy-N z-o~M!2Rl)Y1ycLnjDcnHj)`@D@Asl-`UTfm1X1s_R_toeQz>;jCZ&opnpC(qh1Px+3$>xnz~&lNGx2`{FXgq0F~>jZBi1l# zKo`I2!P~1R^uRaJZLv3=^|k{$3GC=!tX}9_KFtP!qe^?Nt}|zoUbiwAppocD?3FIU z5F$d@&uAy#v|eTZrCDjY%_*uy_cSpEw-|h{axuf=v8+swP8dQ5; zN@-woD`b8N8~Yad7br0fot3_BD{%y5R>;>lIZN`&E>YqU@0usS* zeWeYp{LGEigBVnsbtjb1WD|bh5#G;eh+e!~_Sx@o&H5RZlagG+F`@$|vmz4YQs0J! zS^Mrq_`K8OM>-%I6unTCLUhfAlU(bM`D?7{(cI5YXAr3gf}5x%(6EKi++|BiMUp5e z^KR2T&n$tShWgMimT;t7PoEV~j(Gjt?nXOfS$nvuyGJtOG0{Ng%Sr5_I?HpEniAwJ zDfh|kz0igNUWHU?KThTww&UTN2y#!|hGCq~h^Emk?0Vaus;=QMq~V(eV?uY;zJDG%QM%6YK8{FNoaIaP$$)!4`5o_Dy> z^B-+Lmv_A(oc;Hw)j&vwb%LgHq?=zxrj*((jtXe0^xxY$f53|#BIR}!7aH}9hmd=( zyp(VMP@TyS*2AaN-k&g(^Bq_ovxplJ|8&Cbvy(}>*j!~#`&;1OXS}=f-1$Qvyb{zO z{NhZu2BiH)(fNPxS0)9Vu{#t2li&x^Q*0PO0`3SP&;M&#S$GCmJJnk0kFQiQ@V37j z&R9rn>-`es2dGHGkEH^MO|qusq!dXkpM{_EubM8U=>@i!h&ze#Ca9zw{>5IXx)blIDP!#8~W*2PXGYx6_0Pff3beaApd_)-#k-=|KQ)rr;!k^ zS2>?{ll-3 z?Zq$OY>8jDeI@(8ZCUfWu%tPuRW&Rh5}IjtqB){!82u5EK#x^^s#3c~*t^HD*tSG&{ls+r?Cg;F4is>UpV z#E|w+)!kpLAVX<(&r9bX`kax;e@6BpJk|;RVzRT&;+$Z;kU*$E9vc48bDeWptvL(? zY`vx4b0Ncg;J)dJJG`Oi|5`|mx+|pn^99X4RT9s>x8p8yYm`}c$7HVeM&dcifMrB| z=PBpf)r8OTh<52UhfhELj2@pzz9V6&hfgaK`9`jf;6tR&9loUyWpbrn_wlrq4CX$3 za7fVP*!`ME_OFkO&XVhgp6?vFWmuKkTDkGoQc{!j;SQu!SaP*@vr58Hw?q3iy9l6F z%m-%Y0128xh#enkMn3b9vd!9s>?$5sn|R`@P?}#5n(1-Ow(F$ro+#`3!&2qWFPtm5 z>nK75@b0?d<$edqKWljbrOFjw7YWLPN5I)!RBWzs)mi@fLO5ZRQB~@+`Mb(-kUq-#J+BuS9qc`kz z@udE3T=KVncCM6``_gkq$S!b_+C&(w6~~vRY{<(N!o|8UbB0E z`%0}pB#sYJ16RkwodKNsT^-NQFoI0417>@D6VpLyX-4b{Xjmjr*7ji2N&n>`L*Y-Q`qB;` zZ%`?BGSf@n5a%wDZs$kC8r+ZmacOe4pI`G8ZaG)f%m3}%P1S|x*@GL=2fFNj91)%Y z8Gmu-f4jEbc%GnHqTqB4SUdCVkL>*vK)LAAXCS}fHZV`hkhC~kCID<#|68L~@>h|; zNs$eiYizv8^AIW5O`)bUZy0GRiR>%lmSxv%j~|!vlF;3aKiPOqyh;5oo^JwmThLOn z_HoYD-J5ygT3>GH1R^uKkL~l-+vykX*9o&vUlH=P2WCSu{PLt$u+Z6MI1tR1;Y#n+ zj@cX{oCtc!QZ1kMw}IPx^+so=K59p$+|O&Hmx;o)%lPu_!=U1RA2}sc;EG!(;l$o_Sw3>AXktX*KGq%y!2gsn)BtC$V!z>N1d+B zMZTZ6^~cZdxa6WzK7Y7lQkoB;W|&FukHX2`(r0P=R`$DOW>Ib`$d-TZY;R%+Axm#x=A%pR$q7+mlOzpL=G{SNbg ze|ruts{YEO0;f(5K&zgOJN*M5FmULNs89{=;wHb1lN7Eep!DP&|M5d_3p!ioZzz0k zHws-RHWvN2p7Qpg46^gjubzkH3Ft#jT9QiJ`p&P@6O#W8Uk*7RP{$r>ZGLB|SnBZl zRAk!Rud3zd3q$_0&BU%Etc`{wMQbK3mkUofK`QPO7GS=0Hue{(#s$Y0K=MV<$H8f6cE&M>d5~o2n8krSa zbE{gjP+HJVQ4MX^X?iR5QfaW-O4PB`zl|$6b+OjJ&qD&Kv9>Sx*KRln>ppyW z+y9#6)vZe*o40Xq7p6{Iy!~g*V;&r*+K4ZG8(;UyUKKrS(R@$i!}A!INUeyk9ItCh z8sk-ui&=>D;G3DZG3l<&6m~sQ^W0Olw9i6$xZ+}*mZE9p)T-xE30&Gh<0HXy%&d^6 z>VHqYfcu|&GcOo-GAHrZTYRV~p^|^jI_Prj);NU8Jx_V~;z`j^Vbni4T4&#xo*VDg zwHCejE`AUg^uh%BVu@s%zrCvR8!qdq-uJ{seff3d{R2LG%P?n^hu5Zq1o{J_**lL` zm8U(b>_aX={HPHwh;L3NuIad`f41u#kpWNa-8nyR9_(=Vma+Nlvq}CPH2=|PT$tfM zPi=*(jRP;>ljmga7e9DnarEusK$o*0E8fQChxBf|t&02~^J}jk+P0LdzNj+~Y5#TN zd35jxxqhU2;p6pxqQ7dKY+5~$<$sNuW=!-*fGiY$j4RB&Uh8<(tQj*DbZQN2v=K+J)}) ztG=-9vQTMsP4W6SQTDY3Xziu=l}iqiOAZ6Rw&q`o-*Nd4f77daINYjwzx~r?{HuH4 zPKEpmCEuMo)u7$=5@kP|#%9= zUjv8f)ethBB=mH>mPeDHNM5UKE>*L9KlE1itEH4f_nXGf?TIGWpPyv-TAl9!-ta#r z?p)h{RMQ?-B@k4fT?c793riZT1$RrTRf>1UFV>dRscMxmbALM=?R@ANax?GUvoN(E zQ7L&F26vVsg*=~6{yumA?gypu6q!5zbGC)()s%pNum9aSGM*yLUNjs(6m12FSt^x>{usrh;>bJeq*Vlz-C2`HYKUf2+(7W(3=X;#!|(KIX{g|NZZQXQP}9cVNl;D;W-?bSoPs zV~z46p!A5s?Mv})@bsHr`(TA7U{ms*7a#vTiLd%}S@?I$WcrDT z$IACZGSv_VUln=ZF32F?KC*g(1ze}WIa{vbfdf3AmG0*Edp0+&9N7O7tJI0%TbqOB zvDVFLoNuR{zKWa@8>o5k+-=rJM>geXhw!N(r7MTqj|lo4U7j~d`KBi3+uC~c%HdN{ zhfj%JIqX9y32ALjRGh~~a5g*Yu;bBn%~}@OgOp-SB(j*fOQKf?F5km6C;VLewJUU^ zFzO)PrBA93%DBBP0qKejgy3}<7bbzaa@hJ&HQm&}5pU^;Jj_9-O%eNIFumAk{Z~!m z6T}Inz;jDSCZQSBt|x`G)80ZYl~WGUUl>M%kgj1qbe?*uymM55^+M zZ)`Sx=d>~Xzb^8#V(s-=vGhxi6e^(fHl|*GS8gRV(kC194L;Gd@qwDW7FY*P{XF7z=%nUVOg*dK7! z3ro*s^cZaI%9BIm#L3=0Nzp}icv2GxtO>obYj60#PS zqjJR<-L4^d!JUL7C1?t{l^wgDWG9nvPXf(T$h?N#S!|IUWp}F8M8dY9!(`1z1*F6( zZkk^QkF^PCIYgkziHca=oNG852+UGnA<<_zh34;T+ZS0j8!5&14i7Da(wp=~9IuQP z#_r)di(5nI^;nWIYh7cmnA++an`Q=@PJv!NN;8COquJc<6_`M6lh}&NGs$Vss?TI# z=Z)ojvIXY_Kiz#JsT1318Ukl`wUgS|U9#hoVr|7&yMFbw)2weGTTgV5_4*U(H&@#2 zVgFr@mYiwB9I(i}T^KFcm^UccQS2BpO{G8V-<>aGCsE9E>S8VCm{TXoG;^V{CJ>ce z#qO^BZ0d$#>@Z%hFr=MLQ>P0por#gcDy^Yptg3L5y#VG4j zBSy`%16My*LVH`AjpzHd;@&SNJ$f^{=^nMIoY_T_bYBpeqkW~)7ccd)-P@KDql@_+ zkTeag`cKRy5f;*4LCjZyRu5cZ-wmk{cVUILp##M->M?0jV(GyNha~c1R4EpTz(T4Y z{FZ-=p7!*S%teJ@uwIyRONUf+=p9je%Zw3FnkR8uDwxD48Rm`Aa~V(&>+Q>SZg|oo za}OpKw)t~}1)71ul(`x_uBpp%7deMx3Y35AlO7acy6q8e^PPNZ^WV;PA+Z{faj@H6 zt+u8M+2py=YProMtk-Dvx$q24bZ4x7JX)wyVo9$^Tb(q=a@#I(b>O|^CIoF7zc(h+ zsl|})@S&&;0@fMD(KbSXLO1f0l(B8vRP(^&4wTP6%3iaKw>8y)U3qkf*3Ryh4YS(# zSG={&$oQlpd%nbYyZbVX8vh>6v_Rpy#GlI-TOdx;Jr}cSB~)hOs4;Qo>&mL~4)QlQ zX_agD0op5}s6eBpt65tK&oOP;)(Q}KDO&;MW-617O zeBR);6TLK>=4dL`(@5inGr7coY$zjV`Gun#D?m>^zxOH_fugP9@`zegH#0uDBw1NI-nw*Cf7;T`uYiGJ) zMVKo|#rU5)-R3!WYs%1OhEyj00ycxqF7!h~nY0LlCW%V9DH2J6bkMM}eZj1E=-k^| zuSm>q7o0Y;n45~#F{!PD{)5E|*J>8e`^D}>5c&+tcRZ`1y}jrv>!pU>C|qhDo-;dP zIn`@IR9ZjGHXjC)Tg%5dMRv?0Z;Mbb^X>P>QAHrk{;xUY@uX}a-82~>IjJk!m$ErS zt|eY1bbDl7Wv8@B69~W81L`|P-7pP4^RZIS!4NG_x1q5}`zuMz_uwm+!M`aTi2=89 z?;DaI;BdcthXuuJ#Kuay>Ktea}i`v z4==&jzwpT-2}ISWYvL5$tRjLvY*#_GWO<4aMj&(}c@PiWn6H^_5!G{mp85^K1+%~Z zm$*Ib{mi<{YIH1&@rIlS>bh)u#yhe<*U(<)8N%k!DBWsO zCchPKwpO;xVuWzSmx+DThJ*owpxHKM(p+1I6^B34Wb=~SMB<}$|7X_ji^>u!@g+al zx+%bXzG34d-`O^>qcEKyfHR$;6T2H~d(lUby%`KI-{TW`mu;U=aF)?Z_@*2HM+%_P zZNox1_HpyZ#A5Fgj+Ib)boPjhPgeo0u*DB!N^Kt}TS|H(PnPRGaRrWt3SUdW$i_QD zFaS(V;)E(Epm$lQ`X`9Zf~h!B_cDu64kBiNHQ~~~;dZarZ_rddEEJ=y1DAKEkzx)E z#snuIJ6^h!&kacj`4wW!%ICjqdt9f7BZXU(=E7v(ha1ESbsnJH4Mt9b0 zk}`!J_WmQZliHEpDhm@*h|V}-(_6+MH#|O)bSlh8C_1KZ!DMPjM@{#Mz&T%mYQi9& z9W>k8oy5-~PE8*1l`S(Gju%s{HGd?NH^l0kg2FrJBkJhKi`OZ=Wb>i+G2w?~*uupf zqh5?>&hJjbE}HnG3v(pSBS#adp4cO}1+j}6>q7R5;mvXoYWR}W4zJM$cB!$yCyIHA zuFdfbu!|K}X3x)EsCyKGQ;e4YVUSlx8<@B0eGwn6h@CTZ1xwpKv$Y-L(C|IvBue;S z({mlM`ZBZ!ba!86Q5WkOC^5c3S8-8MW+!=(*}xK>??yF)JSjh1TOR>&C^WmM3`zum zkb8;K7}*oY$@VBdj&IBYbqynQ{8EVD2qbqQ=@JqBdl4)zjsE22(dqPas+Tc%?Yd4gkfR3 zapox}nPI&w9|^YX?yYKz{;yrlnl4%{@##G|O)Y+Xtd3$KR`g19?;$pnt?)gHZSnwS zeC{2!b+D&XmZLL6lap@9s*-3W$({3+Q#Tq!Te&2P(%o}UoBw8f_`-5sj;1&VI6F*` z_x464FGct0PD4$jcsJd#BAz99#(D!VdzTma*Dnf~e^OkOb#A*p`ZftRhGvE*BTjNo zWr2lg-$u*EFSd^5K@q~(#&}xLEKkP6|MPj;s4T64?huuPFvnQ64V$s#WCs3@8_ngx zLW)G@eP!Z}hB8ZkeC6%`&2JHLmm&qZc2r2F1-ro{^eH>H*@!0c~xtkRbhm_ zG@R{UpL!M7#t6;p9WuS>g#&kWY8e*hOi(TcW_t+1JGDHvDG%|1*Ld;(rwA&-NB%fu zK3qo=$Tt5mt%uV$AI)xs=2U%PROmiIDC|eYDP%&&s3cd+QLj8SM`%mWuF(fM`v#bl zr-IJnv^SLJOc4YAQ(fnS!4jY{Yzs>xR0kTuk%`IBumMZcD>Wv(`W0$26Q}dB4UyI^ z!y{zZUi4`E@Q#qFf1%pF#@I0(xTyP*ykT(|!)+GK5Uuf=@R_v^>N!PLx<7w3P4_Xi zOCQPho)soXteGk(lRMwRG}3xEextB2xo_Ia*E<`aNw_UaR6?@CY#GmbJ%;*q{^TJ?|b7A_tHt%o&jdK8il z(uu(s)Gf)MupOmg>(yB!+niXoQQxWTZ|hebxoR`^`~@-T{EDS&JGBPp*}^uy(h-#! z#Hk%YgH3TEBTHMxOmKqe3}NtCwu9bIltS+izh^=^19DSQn*^G4n0VKlnz(h(S&w5k zgAvO#gBUtYlu*Sv#0QLHHVei4p#s?_z&Y)8Y-uPsQce8acK zSqSQ+_?TX^&5xb!P+bkGjD<3xV_l(2qc1rC0?x$-Wf!&jjj=RM-|^ztF8b-DOLbkd zeals^RcS!ZUCjN+Q2~s0!&6mUVIs;SJ)*1c^xOJp&OknX&ed4%!7gLL zN5UE&Y0fSH$fH0oMMzPCH0+28P*}Rm5(+K`b@dAxK+_S` zDjzqG1TP(~XzNo=OCq2H(TdTeXSg@eoF0{i&10s+dWxmjnGW7S|7_;AQq(xa18#yc zCC&Dr0`cE!%QW`dikyMd287R`R_Lv=NVnCGU^lCYvfdm$_cFD;R|Z%Ov~DJzFeIHF zo|;(!ns+m;%Y@hfmIautLQ>3)$${8#YJODcrpKCc zj7^d<5I^l_Hcv!th=b~Sze35$+lTF^iq$Uj1U!93fmt+SL^X}_&@zI-DHc(4EM-ze zuQy&ZN3~|*$Rq%X9~679C>g$kJ&nf&%suK@(w;U4?Y1*yHh>#=+ACnGK-4efhV&c_ za-Vw?S5X{isAb1ie3dtFPP#q=F^?i7v5W>mL3K$#Rv|Iz5je|T2~44ne0AmL39r86 zaX+GAbK8yVBlP5k_syArP8$o)?jbhCji&5YtSmEi$8z`{Qs4ph4tTe3y_P73y-RDx z`L%ZnG8+UQw{@*cJdYg$NvDm@5cQsDHa^u8Bj6s)kE1EJ*WUa@b^ri#Kz_eNGh~cChvg@sX_{FfAu0%N@jc#wt7|4% zM}kYK6Qc_LTpdeYXw<|Nk6YX>#kt+d`d8URm`Qvqd z>toC+!@PsCgbrLDiO<2T^1~kBP2k3}C2BUXLykNDk+90a-9~CUIr6m@6=T&&SSZnJ z=3-J-$#n`pF7}5EZ*uv3MsyP>qqmh6(QwmkVo#_#EJJh}vB7I$J%mxxO1Lgxcnm4x_jRzQF2{b42fZ-kkOyUP8L!eZ3qYPDz}mQcR4W0f=kQ$y7to| z=weuU>>%Gvy4sG&q{@K2K{#9pmxA;s4FW?ZzR8lRwQKbC^dSFfcPbtd3;KAu8UhxI zugq`A0jNc$(1-n*kjGvC4ObVEWj&9hXz0l zP`+&Ilu=O|UlF~^&E5}i(Wf`25i`z>EMP#>5t|Sf&Iw6W;QIa>Ci40ZN;}XAvHIti zl}LvwDmOnAw6pLZ>DFMu>dG6%E{Wm=gu(dP1%7IHjx#ERj%!GbZW~yjxOm8(iAjyz z>aP1StdnZIMmntChEBkt@tfix>Rw}^m1yxgT>^@C7Y?QH^m~Rj_i> zpef$IW1c&@7ALUecl1omV_jU~1T5nQ;521cwwLv4ZU?G-!^W3QL1%vmREn0VS-~!! z8?ReD67*pU;*ptT`-1r~#c=08+HMwBiD~?wu}*Oh(~|~Nn&YYSY_)Z1Pqnr7PbKBM zlTlrnW{U6pYgoo8 zknJsXV*jMQ4`6;+!719|dx8`oq{?a~kG^UGd9au+$ps?J$bEr;?JqfB>>m6yM4<;sggvm#92EhA%{+Xg>bO?Uan zWCK!9FtgH__jf_X6A$2GkHn@=(S(sDYS&>0_3{S6_-n8H3ddem^d?g~AhVI;iFrlQ zvg&Ob_jr7uq}7KBkCR4~uvRj25igg_gSL9K27SqkcJaRa2qMV&D^? ziY<&KvoKIxB?NCm*9FEHV8KCZDiD%E+cmGxERO8Z(d(NhNR@KmEYKQb0I@N0i~gG$ z00~kFV?^xdH4_%6*qd}S2Ht8kplcWYOcb~F#en9y1uJDh{I;;eQX!*AE>3+aVc&A4 z17j|bom-!Ecgf~ln}eX+H1BifmUj-9D_lIus2|2gq{P8aU)sSQd@lkt+%3_}B+ISyl8(~1# z+fn{`ebzRm@#6Yau40>)Ssg@WZg<-ldz37PHYYQJi;8DpFX(OkYC&<-?!{d(-@|3H zSK*YLh&}Pnm>f;rzIfUGztK960UCy@$bAuYDMC)Z-M zw|$o)*4w}BKqgiCOr^u&CD_)E6P_i@ganD4LfSSxcuvZEL7ce)_JmX8Hw6s57DR0U z5?Dz1uXgG?ED20k)y-60TM}!P<(yVxxRPf`~WdyMvm?KDdU!S41^i$9OwzE z+>|H&N&tkX0vfXHo}Riw>Ik^7MIM$9iJqG?2Tk$3@!IJiMSvy(qYXd`^GHEh>St7a zlt?@Qv*J9xsn^wiF?2Ke7wRK`R>0@GlLS@5)8j|lMz^t1&e7;cqEXZXLY?)tcTL2}##=)1UycG>(+X zyTUX^8E+fc=Ia&Xv2vqNi%ld8!F!N3I%==kPYTtLYA;OM1DF(gBYzC`c}JurJk@{g z>$;m>3H1g=CZ{Hv9x6u`Yd|>yPG(cjSRGS3dW9E9!5{i90qzQkd{DCh3Q8k-Gk|m} z2$JA^UcE{TkgiX8P~$B%Q>mfnC4d4LE2gemF0pYvKC3^hAxhRUuiq_r$*e9sUC1V> zdNggt z0>O(%nU(E6UXfyrTp%e(Cf1Ofi*+&TZQdRz#vm?sw4&X&7{5~I|CG~CX{}J=X?1Gl z#tbHHvDy-YvgZup)x?jzLcsX3>!8^F*BtD4Q7+m{T?r{!5g)4WiJvmF0xTJi@VS>O zNIiIGUC3%75NGZQsJUW*BqSqVsnoL^O)=IZ-s075D$_=A1=!`eqxMPPR`;W;L&Y!$ zpY(9vh#t4D&A!@_bHQ#y8DtC?LFT+c1DI05K0_g$Fa&%Kpj=}S$a%)@8D3ycapR2o zG(wYb*))fViZOV>gr~Q+H_5&}L!B7LVUAEH$K7*ksM56X9E2`M&wa@(s43KTP7v^O zJOhK-)k{(a^%?8M^%qA$7>Lmo6l*kR3fRJ$*Va6OBHd=nY&O_XGw&_6_@)t{XOBaZo5ypCn2^i*Jwf8Vz9bpit#jlrTwr1*~$a zizior9j%74#~fj){J?C-7+oIiXEahz%$e&uWEU?Tm@^tx!oSz73xwQrG~tN@S(zEr zH;Z*mRQ~GH4*zBG=7=7m-7s zp?mK5M-@Gm9gy{Fcy`2{B+cpdrl;sUu82v~1-#izqakB0oxMy0*deUWBh_<&?yt`Q zyDuEa%kHSUk;TAy1Hk`G-=`3sfXd#G1|*cZI5Y)ZxFE*NE1c;8W|`vK`ZREt6@0$; zIJ4xLb4(_}0E<`34u53w5WDz{{K?E&qKUcF7ooTlX<`qrGxX!iDJ#I%%)YfT`_^Z= zg*w8DF$^e&HJ?%yO}kOpGC}DcZ;RbB?_l04=7+~iSk=(x8~>tZ0M9Fg_m%^uaCSZ{ z1K^njm^mwW0VXjtrkfeDje-{Ig{n{^7Ljo%=EFWVdA?Jh_9`Cb)CAgp`crMAuNg-X zN@>HUS}#Y^y4~=UZhTv$=k4Nn1h_cfp4GlQ=1Af_U@HQs0t7R5rrTx?`0+NNwqx40z21}CL3>1vu%;7eB?vI_Do8463?7~Z zq&P)TJF%X5M%yU>617De9{RpK-t?OY80(drytkwOqq6zL7M2JwH}p*3m5+8Qtu2a? z0W_ZloR?c=bY_*p!lAi1(dke#IDSWir0+0hk|z8RZ%e8pNk!VBC?8zsXE-h;yyzOG zA;lI4N59#_fHP{`5vdPZo_ku+3t@ko{oXqy1qeBb%Bk^Q%qwDJWN9U^0^-hLObm-X z**mO6)!h+_GR>9)vSL_z5?W~E2H8Glgy@Ylm|r{uME)_~qzV&l+gBZL+uqeyaY{TY z{+D$D|Md4X8Z)FHVoT4d=~2}RfULrT8AsL@4%48iFDx2ZCxFU=FvAWNtM9m6l&LU{ zm_~0;46%ycI^hWZNPk#RB)-kkwt#h>3K*{$p&|2*SwPrs2vUH^8}=e302()U619K* z;sG#j-AP{WjbrUbDOxNDN8ZI9QVvEQMQ_T3a^@}MbH2&gbjb{I*s7t zT!mV>z2Yhq^a4D##BZn7sduQ4=xXxRC+nS`J#2>-;?DQAGGkAoI&(?Rv&_$ucL?9s z{}ooh`@M1;s*d~JrpN)je_>6RVF*s_w@RkoG$0uux?2|OUHXwM1>b^}a(mt-_i;Pr z;Gz>j+scuFt8|wuqf#7@*q*kVc1lIwpy}INdP!AejwAk;b%E;$ST5|AHu^8sRC5-F2vC0&B`@EPBmYBwU%= z6R>5z68S^F7=m?_@s||_%CWEV(ECzEwQE=?%M80b(Z+DAnL&%sm&_2&)(_CH=i?z< zqsW{&SSs`wV8{n3o1oAh6@rw=ZytuD$t3XpoVeHeA63~90EPvr^`2k#X|0vQ-+BKr zev@bRT(?qdl=>t`N_cP9Vh)31mo=U_H-Vb*^}A_#F}9Ej2f#s2tD~l|1h*X`YOTCJcc?d2Ug%9!r% zYDpiU&w)PH!)^VBeP_|}jk~~! zn+9pE0DrwQY}y3_1wSC8zTFfAa<+M$!>go{+FXnX%WM^tZdkvmUg3Abz9GSnV;j79 zxB`%>3a*K=Ah--+GdsqLUC8lzhc5EZC)q{M^NpE2MaVGCPZ--rd+2W(w$K5n1ek{c zPJey~ZRNWxtmY75#p73-$DxbL(A87~OFqiM5vEtr3-{O>uJzJe+))JPn|$ret=3_) zxLUlo3cqQ5#Ei2dsJXSCbfjxgIcS>fMC_LX6u@F@!^P3n+FA>15AjR@y2?goEJr`j zOz%O@ElG=pg<&9-!o5XGM3QP`XkmGx4?}5Bqpp}D{jaR2w z?AyisXJ|f=`xn)oIFViKbI>{Fk7tzWe#H#XSnWQ&k$buGSTNRkBGJE7`HZ4kZTG;Z z&U+^C!Pm{|fSk`%vKZ8fJeK976xQ$rVU6tk%b-5B!W65U749_~v5cIk$IyY|+)S8A zyeTmew-P}?7Kmdr<3rV&KvLVz*9?1Za-=42IsP>FAoY6~B4M_F=8P!9v&p=ooc zwm~yNi%)>pqv+^zJz7a$pw1C!2U(th+EuLDyCMNHpC~T>v&rwsS`0oWVtE+Z#8Sdfv)Fg=pfT#pE_|(B|{BD`2-5 zD&_bney#jHPQ>t;zf;@VG=RyT$-6c>s_v4g6b4O3oZ<@@x&m_c;Id3I5K;E)G0Pf+5Wq``N zzgs5s$QOrF9y-uh5+MZqDYR~nN~|OSZ61W!YW?=(NC41=AkJRde&KI%zH4xbOh^J3Lp;3&8-Zdl z*|o@TTMTdF?+JGcTh4B*!?Pl&l!Ljl%dCLxq6)tuRs=wS20@)#A?{17^jZs+z%oF{ zR7?y3BP+3M=QIWDbL=Z%Q>8E>EreLP%5<6Rk=*+UA^-#(!5wN7|6k`)OpbhaQe<60 zXA(jVgLnCrFc&!)Gnadc6zvtj^vS6h$-UgIjR0fe<)1H}4<;Q4QH^gvd4)ic`%VvG zj4W7S)Hjw(05)wq9%+&U z6Jx&0nE*8KbK*_CH+Zd)`aY~PzJ)EzTbd!ryqFtqugZFp0hYH?9o!Q>fjej1gBF?v z)Sh@<_U@CF=J{mOnb}16y5?r{vaxi-BinN=HFb};sKzKhkSARxk4A`T#p<$a>xm$M z8OWPy&oLeUFrv(p#JrE%>!b;v^FC%Ble%B>&-sJbfO|=$T zH-bylxVBC(IACU+kGiGzo3K;0!t z+tJ4zddnsi^YDq+Jrm2ny;ZP1Us)soiToV&VstUYe?6IP13%}hH`vRr}S_G&r5S0mWXB0_?t`fsAWkBh{(-+YF{qpnP zlDYc_pu54}63P`CJ(V4vfslz#cuBX*ygQ8m<82##k^3VJJoKOw*zd;KOLY+e0nE!~ zOk3Kk<*oE!4IV#IIJ#w|W} z=pcQB!y3{xv$&P;i4-I1a{L+)k<_bte%*3o&_@VX&QuSe~T1~; zhhb#|;Ns3kd?&X>Si0#|z#ds|(cPUlML?K&QNU!sL_&6IV2KCkB4ihh0nwpg=>?T7k?qP8v^agJj=0_MZZx=rr^iS!Y%iF68JnAdrq!(2HU$g=WFFu(!4 zYP&waTniR8X&c}NLWu?nTZj%;L9EhTc~Vh2A3t)3r`?yNnX-FlOq2&8!C)b_5xUx< z1qxd81`MP@W6!+|xOItIUk|UA$|EpvZ4~@6E5-x$ABGXP5T{hnVn-#a1CGcTG$XDU z1YRy$`k#`(iE;b3^@fhMaKEuCOw*~xKiXItBtIjo(KPOPb8k39KR%mnM72W2F_l9I zt&YTMe#~6F)Gt7~U1ZJzKgz4|<>Sn;ZhxR}1>TceXGi?{5koh42ix~z?RYq91V0HD zx_%r8iu3MEA~S9Z_+e!y-T>KYcH-IJymh0QG#P~i(zuIbgX3xJ%3ff;xs%z5Q{92jb$-TEh~|GiL&zoyk3(jSNPjtyKq#e(_5{h+o)d4!W~2Cl=9YM|-YSZ<$0#XKY$!ax=!3{F2=4`yK}^4+OCWu zS#p$3@KI1pXt#YJYxMX`%h+Q~oV`c1?GSBDJ58f&Kx?}`2VZF%FKXy{^s37>+uidX zuxpARYmHsZ7P3Ws50x2RvE8hlg`Vl^cLTC7=vQkL#(C(}3}R}gjeACLeiG=COg;l> zA#fO()#x%5^XUi36H{SqC+X+$pvLzI1<0!fk<+n+4wXvCmT4_^Y+_7v5t<~rYOTAq zjpD@NDx>Q&`Mb~xxS{zecsxLVz~>_15fyKxo)FeVo8(^K&98Rs>SnS(ZLzLQh2I6} zS7?A68ue?;FCGVqVYBfayLHeMK-+1zTEL$$BWJFP4b1|`hoUe7{;i34@}kns>!4#y zYT`(5D@AkRD9Bz(dlA}&-}B@+!B$z)BLy&(F`>-L2G&cC=GGKv_+)uQ`jo^&x{W!( zc-UAg9Bw71>JF0z2*^}=8)sA^B1#9k*aDO*eK^d#2SqKOqiNGt7nJiXjQ1{*_EyaN;mCg6YHHvqO3c0qQWSsVddq-+sw!#3WsiQ&a` zr!&m*e|G0xgLeyc1NQZo1oI#O--DV^)?O^m9Ql&tdryp7%dJLsxZ&88m67BPvVhv= zaS$k*;wvGl@*6#lo8sZPeV8`r5&^9hbWj7YGN6M53s(a*JtnMBK|T4@esxNB6=<`O zf)Ff6U&d{b=Js@L76d#?{p>j>Jrf=)*hgGT%-DBnn3X_~uA`YG)pDIo1RvwP`4Zqm z@GK2`{H`WyUbpyMR?L#-FJ5Dnx&R+`QL(j8b-E29Ksq;k3q4x>+fij{+0lb24l>+Q z+IZSy55*Mu8~`~%e|E+ zNCui^0700u!ZwZWQjv^#1UT~tm@ArXByAl4_DcN57@i9vc%y~A^gQowjo^i^=Dg&h z+`k6td6?7F5p%ws@CxLs+-9m?4rBG=dPD^!;Sc%NAQ?-Es4`|4{~(ZCFysmIiX^1O zgn28=IB_+?l@Vhe0(3W~sST2e^9tAmyT2rho3d)n1t-#9bD7x5$%5&!se1J&#j$Zna)j_w*K$AZE;dS zqZ!})!XX+U-NfF`#t4rrouE0;`!)^cAB8B`gUJm)O@oWhWa|}ogu*>?FOa5U@3_MF z3xH{^?IX7yW4mfS7g+#a`8pdV3mVssX204;qwf#!#hBwx4Vi{9Igd~efyk{yZ7fy0 zd+~v4pUn`_nqBBs=S~j3GWK0qhQh4ROH#1nXd1`F)OQ6k+l8QqJ!Bj%*GX>$It%8iUFDOe|c*5cJTtg^0;#(UBDs2t{cz-Woqom&@R=`jbp{1@ts^f z-H{fxYi3UXk#IOHJX>_ucTE8}AOYqBcuvgrc~m!yRa-~p8TcOnM*8m{Uz}u2hAqzY zmW1JK0$Fxg7C@1b9;q@S6+Mx{eOg@%2D%)K6VfTn2A)`mqU{?_WU#j=NN=^jyg)tW z25zg8w)vDxdx(Jy;J|r9XMvt(;IXGNintG5LPEc*S{v@B3? z8G5qSw}_(hkPeJ%bj;9-M0<~@P3bZ&a+epWRA*LawO}8T_uvC9GWy60-^qf z0aFVI!Yn;cfW-n0M+?7l+GByT=7q_60K7lWD|B6h*T#>rf<0tLh?bqoPZU~_Lheg) z^y}XyLnt-3Xbstbagc&otSz8PliRMr8FLM=^myGrlA-Q!C4c(RN@iwk3+Q+8LlTnn znjX65UZVDo*I_JR%O&(3BN2f7l4$MjlSdkBfb%y&LCwWclrSF&A)^6>!UaK~+;Ju? zX%b#t)ks{XG;%r+*hv$ABnWvAw(392njrLR{vT8C9+&j_{*PByb}-FcdDi#VvUf){ zBP~H!R<1cS=bAN7nJdqy$Ri?^m6jsgdY45CrOU%K$TZJ_rovV#nj{_}r9eeRLP0^` zd(r3j$M28kLmluEUiWoh*Yk9PN2Yc$%Xy@Tv+zRnAz>6z(*HdCiZw7mD66?FGb(MG zRWI^sIT9gbXZvu1T0yd2wlkdzN^qI+XpK#G`ixw2Lu7ElgjGqtjtJEeJRY25PPh|6?=LO;2^liJvD`8f6)mzU2| zmNz+eTj|gzYcfFKPyV2nLI0H7_cL$WU^%hO<`8$}ee;(2p`$m%uQgvE#u)i#UC?tV zg?h}_b1eOrR#wh_aR^By9?H=rD6OKkX zN?H+E$SAk0>h;x^ z?K5}m4bi2Z3E_V`h`V!$!Vv6w$2tDk-Qg1McjKSe-78H=_VYdeVY>~t`@OB*9n;RS z^ZzWY@h;+v13xdmPrB-GsNHm!SpK3M`C%1C{Beq-9SPH&)Bly3f#BU7;B=}x@2wf@ zCutoo@8DndgzLT%-1pV=7Ilft(Q7{Tk=)QA4EE+RanXa#^IPV=oGL@~`C2Wj#zi0P~N#=VhP z9a&HjMgBms()PUXh6EL2DuFEx_u>36nf1#VoOS=tDF%*}w*YbbN<4SXeO`i*nXY|t zL&hNw2f0P#Myph>Xv+&fyC@@*7bC;doO9)iqGP1jrWdKNolidlt#3#^@U?%(igzQL7!v0?#cJcyT~cMN=k$xwj+GSNAbu{3I96Yt z6PM!w8(Tm?kV*E#G)G_7lgn=;Nl-&04&_v}^)xo-ZUm)GqY0tTrAwMCU9^h1xjjO^ zLX@>qU{F}x;S?>o8|^7<4*0YIQ`dqh<}w$;YwUdo`74XW!YNL22JyAB@rU(($r-gy zCvhVS!WUT}2{*`<@4`NAhs8(F(YnM5mq?40# zB;6r?&TM?M$;y7-kTQGUSrf^-_cSd-Can4zm;3O-BOYf)o~%S4FGywZSs9yhIj**o zi!x#JCjLw>E$M~W^q{HY%DMI}hXAcPg3F!#ur-ix5?WW~nuiWt4;ockZKW%#)+zlO z2QVP;u1vXvtsTkJxxFO)c@@HMN+#T{l)k;sl!o9=mq~;hz@??ecx0wN;Ae;7C_Vo6 zbjo&y(-DGigB6GMZag(?vEdz)Fv&?HoO@Q8p?yveOzo)_XB??`JFEONS}6{YE$xva z2j!b(stktU6^q)1ap&uaB{!C$Eag$IPZ*3>($pud-EX_ym?={!;mi3+-|JOb9#xV~ z8HLiZtdPK;YFT9gq4?2~gXAP7iP}=bD5oByHwV4;=Q#~wdpZxt?4S(xi|-*YinPv< zAza?3pK2Gvx|6-u+TB>smOp238)ewBeaw_*+`jew&nko5mJoadnb~k?OXKWWikpry zjPstt|Dz$G#D`OI3TGZPnOsBpnvd@x>A;5SPM%D#3M@Znd9TG6w4f=Y~iyw5_Y<)XiVeT&iSOYQKm<{t-0KRWBT(g zdD+w6_euiWH|R^YL6`JU9KQubVC<2fYQ-6eQNs-gbo%vy@$G?gDz7aw%5U)RMe~Rk zc(a+BXLn?(A}p))UrF;>@Ex7R%Zw+bQ&frSS0kvv z#JxP+{kfF}$?Xp5M;Ai$5GA>NBX->|j-nEx3I%^C^a%q(W!Y{XZsAM_FAwzsm+6k= zV|GCi%VOOWmvmQVC9oHltoB?xiUc)jk^hcw7z6#CrL6et&g{Kq)cA}*)h>0?R;|Pj zF-eHOuLpZ1(Q-lCs1V;+U%d~+-wul1d?m&9_Ih@{nY+Vp`S8Y?Bt86b@{j#9grvZ+ zukz{`A9vQ#4%WnLIeQdHioRB7h6o?Ff_y$}YsE#e(dd!2+IKwt&tcbA)BW+8G_OP~ z(Nbt$y--+yyFIgNTdycJ{|L`TR3ohq9_|k9Or1%a9Za8E8ffs@*`LBx`^dXyeYmr_ zvfl2He^U0I5x*_qqkW?n_+dwJFUk3_mjkQ*he?M5(Bxk}h0WNc)-AFG z*(c}hn{#bDV&Wtz6tBLEIBVTSHa`}r@UxPxS+cIaEwKQ{$Lf==yys52>zi7>4FrJa zTs!or*8~0!W+_P~7(i|RA{=Z-S=lSr%r5^_+AHeg*Ks%t#dWb?-t$20VS@U|!dhUJ z%|wwed5R-ZkqWjh)NG*Z!dU^i%D;Mr{ff_4DuusaP-_v7B(hQZ4V=XL1~H3*$j+Et ztQCuzo?9XAEaJI@ItXfA=Q6&D1TgsgsmL5{?7C1#O@tY z06bQ{NhYpcC?2@zSM`MUm2XsJ%sIKG%f(5L>b^^FpwAgjcFuJ!~EM%3^IY9J;K^TYS4hDy3P&nVvwjt%_dpTNF^_+?za*Oj*u}5tDbJJB4LpT z$7PiWh;o*Nab5fWsXf17Pww;aqpTjnO1*J_FKmvX?wWRj_E1MQlG>biaDf_ke;37W z@jiuO`Z7qF#3oE6q)kSSWSy<-u3=2oPCGKAB4^v5_QDu6!N% zan8n=2Xa~-QeZG&z zERst9#nzU8?3JTph6zHG*1e$4akCg(trd>ONDPze+%@usgw=LgX62cK)~$vom#kOp z^!EAKPmgf;$gV?Y@aqcoeL zO<~_DFL1iLMfnLQ&lHKGt4R9;`J2UaY(-JNyB_rKwxZ>OJ`DQD27LT&GgqNe`({#+ z3)2iU`=+0TE`i0fe*yifkmuq{{3x&g0a&VzwSmFbv1j^J>AsKn8*1U^wf?Y~>C~o+ zgqoo{q6bTReMBl-cRXum7q`epx-;NehEJ0lexY^{_cvYGTlXQieSYu7t>y!5TD+6y zDq{2=A0W3tlU^ix;o_Gmm}S+!`7yuh`Jgv1WQx_8dk_{ZOrpuW1z16s*cK?kq#fyU zw-bN^+cvsR9AexPL5paE1FOg2v2h>p&xeVpOGOsQ{zZ@dNWvlQ{n0 zw3_|+fD)8SH?3E6nKQ`^!J1B`?0r&@cUjs-dD(%ova^d*wt-GG=@#H)`*OXdO)BGx z?RU;cT`&j5Jvgkhh^rRhYMIvF^j717*f$cd03L!PIaw{Lc+}%)Bp&SFTZ%y$n_>1} z%+Cze^_1Jz{b2oN8-6NCQD)$J@ta?p2r6&FbPHYF$*wgEdj)5h;xqKu@N6yGXWPVy z|7IwmzjNYo_Qo0nW-7G8gXM8=(S!15yGLipCQ<1r^(Eh3z=)Jqs<$Y;7D6Z9-hnB< zmp;B5#z3vI^Tgg@QvkWZOOp*Rv&{gOPEdmAbBGDcKGzuJDGxvx7!?dx)!{tCHw;;J zowb<*Yj!JgTcuWDs{`uvL;7=O;C$r;p*cc50DjOpK+Dj>am}G+fnHxpL-*YPhl}`JQG}P6Yk!K zu!M%Qp1@5h<$ubjvS~Qs2*Jlyvw!jM|Jr>O(ih!MmQBYna1f2YPg9$dFwzxw(6C#` z7H^cHk5_$?7h76Ma>za_k<4{N;@m}R_2KOy_VD8FuLDUmQD(1z^lmRtkkDfM|$yPA`?aOkiBinnLVh%F8B;eFVK zz>vyyNb%~G)ym|q8j4$y+?(_4)Mw-G?gN|8K8L%bDp5Ghe`-e#6DN3A?t6e|Lkn8{ zQh6J00o^> zBqD_6LfjM5Wo&GzFjOwZe@eSMVt}Rn%7#nblNR8L^B_53)w%3u5$A!&A*BNa9IGPOXVNgB-nVR$>#WIe_zP3fHRRY z7@ufmrpy=CQ4&X!Ms)EDgBX@#y1A%QSkt*!+`JQ^7#YIE%v6$G#Put=;~TVC6!L?i z%606QE4ha3rgwaD677Jf^o!{)(qep5E?B=-P6hW;GHiYb((8NZ-t~EBz@=NQ+#oUU4MH+pg1vh?FU6a@=5HpfN;?QK!igwc%&f%JEe%6r_(B2yHX~Xgk`m7sLU+I zCT?Ld6D#I#j_sL^yd^3W^-N&jCxkD%&g?dZ)-a;NwN>q-3RZk`@qgemN&OHyLi!I_ zjqP)^R6xezz2Ne4YFo2EKHGU-wO7=N&e0}IygmR65bB;Ct~F{kKlwVSC$?EjvS?(q zXg*|f8*DZaF^7YiUF2-H3ORmZc_w9X^a?G0w#0>g7%JKY!yuWoV0_7wFLe=zB3X(i zCR8OcqPrwhwq)@#noR#zw85T)B^})3va^J1Ej#=UIFr#a9r&aPb#9p0Aq?g95#qox ztT?Pm;O9WN>^9o(bJ^-~P?7Kwb1)@r8f7$=K9JxlbP20qao<9_g~Frc#yRjn{~}g% zbBPhsTE^X!Tp_-$Ftdh6m_BnE*{0yd!|rwwDi0Ho&y3LJj)|2}SYSJ{shKm&)P8(V z#KO^3mff{@_QaB-iSQE3TK16MmX?|>XIsaCGleUa6E#Uw>6zz~z%CaqZEAWa6qQK_ zkIbk1SV==pNyAxpid%%mP{@MKol8@=YF)-7ttd1roD!+UYq$K;<02=b_$MX-AF`EP z>hcYxJqY7?qfZ0{S-f}-5a5&N&TT1S^9{d}1AM#6L>0aA-1Kfl2^T|TKruIhDbPqy zQz&v~nlLJ}z`C^_27x^zPsq6cY|nxQdBdUjETVINBV*H$Vtc1t(V*Tes#;(;c{N*| zn!3amp-(htqu7h}E-P~`Sis&*{LjMXsrX$};cRKo1iD0is)%(zP`B%WXtSa*#!HV6 zW&S@<=B-V5?(~$)!+)+>a0hoGmi|$HE@TA@ZX|F9IS(+u2K|RzYgw7g3dOOW@5~AY zwKr7qHTxykWCr7_7c=9J{oge204)rqMf_RYEn(_i?mx`KD=Tu5792#}uv?Ry~7$Iw>VJ zFwCLTvyRJ^ajWD%-%j|d@}`2@liZcWL=sIC{Je-MQR#S6o@J=;X{G!&K^e@!{)IX&u8Zb02#kK3MiRkni7{ zgeOW|6Ypd6R6ex@?w|9Kv*HsaW87ZM(%P%R(u}@*@m;+P=)V}qI zH54}>>rUVUYIS8W8v<728El3wE&HifbS89^I^Qsi?e}-!?P;cL?JqN3cJQrd*muWg zmeq^i27}V7?MR}x(Zwbx&)Gz?g+ma-LiC$Tw2G_=noOBTb?;QNMmK!Em=WEzb%Ev- z?1z@y`d_+DyEUA9QO^hz#2&t;AM~Rwx*n$3IGymzv|%;vsJLhe2^kKMpOm3_gSU8% zj{2}Q;ayaI&x$U3KP2@CIT0z^MH?uHL|9AP$bug0x7^pERm^Alaar&E`#2Iwt_p)}nZ z@)0rzX`32I9H7VFw-9K@6#fcod{(Vf$C~|>bFrhy6BQTCA(z0^M1pB&lpun>Z@d zVeB7QOGT>rDRF$j^wLXV_;D;VnT63<*HO*c%4p8nd(Y_5@;oJrCKH-CQryQh(t4rL zz45eSnJzf;{WRJ!IIXY$=^d4>@2RZ3s1myjB4)~9JU`kqoWFZh)Y!h$FFu!3#Acc2 z773I<_pU_IZi308SB&$`VAu?NO%yDOYsO)S>j0DjtEUFJ)@p>I;MDjI+C(ZTNj<-@ zoAY6r1RD+K={LU&#y~s>_|O+|N%y~)pdYv!W>H-s=t=de6Bc8@C|+jq~{U8%d?R;_@LvBRPPB z;)hw4+Q!`pRJo7{*Gw)t=NL+Hb*B8%vR+^2eti%;Y|p=t6%S+917Q;s!Y9Q?n{P#r z6ka~2(hr?@PhKX++}va(NfS>IHz?hcGeBnJ!}POfog8`jyncyCxsvUGRJu|qm8myX z-akGO{Fw^2k?%bDA9h1)fQZ;e7wq;JxiHNMP<@Ubb!{S_?DA}x2@1L4;(WR*lAYthVvW8EOG6@QQ^xxJ?C-)?(B*baFXXz|BfR>;adb#05zjkHy=c*4I) z@k%M!pAS8x--@39RlHqEnAq636v*qxbwCm6Zwwtx#j@(V?Q=p_;Yui)vXpu7GSSO1 zfedJ4LANp`My{UeuTVAFl_PjW#V07YdeFM(*)G%ox!SY~YKw*BY z-prBk4b$?57~&gcqcj$Db6=)41S!7Y;6@oW4j-sWEbzz)9m4jtkH@HtFXl#jF6ji| z;L4BT-w#iVFB%}QQ`_SiH4GNo__ZCVE#{xbs0<~4Mt3b_)~mG}2XPmeTHn?bUmatR zKUjb#8kZv?1RMcjfKL;1tBq^_f^lUIVp{!mEl%5{V&BZTDxFHK5ym-k_R_&dgCxGv z<#o#KywZPtY%i>d1o_4R<#TFlqP1@6o%Fx_e7693_oRyAGIr9g}u6yS;3uWQ#|ssgL~E5=P|cF2X{s%V2?U1>D)doLGXKl2GT4V}M)- zrXfA3Iv!2RuacIx-*$N)4Qj`}bDFjox(zG!hjWbG5vL)zhEz)`VNg7NmmYMBiaZ^# z6GlcfCHv8Dejy3qq&fB5XhQczLl13E;Tv}Oxi*)GVzNNsfo-U|wj=+8Dzr{T+{&a% zvO+D4TZ}E~bLV5YYxhg0s{#VKa!VB}ff*in zHUAnjd4`3SUXAyQ8-C6L+;JU3IPVU092zKEU4|#_`eW*~9c&A7mF5CJ?WwuzHAD9F z1TXtgE*Ufhp8r#6X^02~+Yz+p5o_7xA#3F~N!<7!FRX>TdS&k1S|;ElPR#8J+CW!Q zu3=M)8EEOFwW3;TtjP5h_iIiEr6hp5W&sM^f-?g5YGT&iv!d#GxR3Y(e4Z&!T!0Jr zMt3i{EY!0H&`!bmKCC}%KJS=KJ#LRTWuFVmfh#CHsWh5TS#TWSwQmc2dQv$I-!Ole zP1BC05#w-CI1dOtsymKf@EmaMfV2{gB{on6)elrE@XdDkX3{o+2FxW5QSsHgK&b)Q1_qJIBX( zI4W{CqL%9dMgYGYY>~+0!Ul(P9cEm<&I6Ex1TnsjFze*NfQ3EJ%MK*3;ylf=6{U_| z!Oo}8lv&TQRecNh&kOf7lRX6TU&$x~EJ4^xs`cm(Vfj6XrY@{OF(Tvrb-~WJvy>$^ z**EXiyEj|CPgt;_X%l{!NvMHB$fO5Ir$*8H#G@^kw z*3WB$K@a8fqm8t2((|D7%NMUjj(iv_55Soqc1&%J zqAAX5b(^9l?N_m;GDzdytThYH9*8)N-D#$b z>8}Wt%8&xr1&L^h|f3V=I8GVN{X{qGBc*CHg;D+#+ggRLNV=j&Wa)>CSt^rqIB7P20u6Mb3S zT?4Jqgm!!3i@SRh{Lm^;J0iBP4-C1Ze1+l6*ArGwHGexuAI+Gldb7M<;8kQZmoPix zy*F5A*kn1?*BC^Dut@d!jZVi3Y^UwHvaKQhTcXK247@k`2?{$YBcI?o)`ddQUrGOP znlodjH&VS(dt2Mo#AP4iZD*O4(@QiztMbvdbJOOdsNd`}PehIFkb$;llN6l4+Q}P~ zu=opKg(PR}>OWN^xUsYwTy+5UhK9*s{Kmzw^M`6rif8VoEO05|oa2;{=88Cifr-k7 zL+-GE=ka0x_NXLxQ$E;y$Mw8sc6O((|B-JxmAXD*qC2$w(Cj$)97VsV}DtBCz!)DUZ39mmssADvAv;Cqb8O2EoM?s*Vg~|lreQY zEI!eQ-6;$3@1yi8zD@{X>(TnKt|3iFHUGTQwaX%+c%O7`n^?DQO9Z8il_|M!7G5ps z33uv%)xs}He{rW?!dGFDYi3H?sW5{JrrOs@^!mkjp_GLduK@rJ@hh*myPPZ7_d`IQ z&?>O-B$NC%8Jnl12uu33~WZGz|0WUz|RpC#g8~jpdu<}bJhE1;Vs&zu$Eji zEBAYcJI<5SmsP972r&PG$u;7KYBGNfw@`3bwo<>hoR%^p8-2ZByWHzuqg=*UBDM}i z7aU?)vi*B!Ur`v5*3Yc33g3Mj?!`AveQ!PCdT}muqzOs~K$Q%Ky+wlaB5Wt$I68>v zqkK32DCrt(JB+E@^M0cUSkmUbxNHY@MOYAjt=PM2IT#{xwD;aIP0xFGJ4ogl9IaC{ zFaA>{_)ckv?G(A=`yunFkRh{t<_a+v$?pg25yqW9S}->x;Vg(CyQ&yd5%YclnW6ai z@4*IW&i*aEy_vdKP(e&@9N`LvGIy2jY6X z>ak&`XFSkLUL#0fC&s*2qYP*Q7-^92tR6N>&NqG6g_v>FhGe{0c8GE58e}!dEWoBg zJTQZ(Zm?ky->~YJv^f9~FOw)P1df*Ml~^TBC{r9=@TPmA*2Lev9~8%r7-$Ee)T{(D zvw}EP7EkXNxvvPwz+~4WTwxi4M@zM>MrGr1=j=Lsk)3QPq*-I?-t6}%G zZ-K};9d{=ZdGQ0c3r)bxYs@)TC-Ggb7f4^zeiaQi)mDTI)$6~E#TQ_h$$hh}-Ow9w3Em0;L%2fryB!+>&ne z+qB4$LYvA2rPnxYa@!0{mgE!{aBUp*6g&E>RxnKl>SbFVKpWG!TwMORkRkB17nh}$ zOMeZj(ip_3kW%9pH_cE&Q)LY{ydwRdwC&aC0yCa!Uo9O1b+d_)L$Px!x6@|?y{jjv z7q*H@P3>5lV z5x>yNtwbLJ2l3 zqJqDWPF}!rA?-#@R$y;Ga(AdjkFB+y6A(sOZL2toSv1#p)k+9k+HS&hfJ7B3NVWnm zzzp8u(exSw*eq7GRzPh6kpFqWk`PK>f>85Tc|{&KMO$vl^Wk~?ufZf0>l(=OxF26a zgCAs^qz5To$PJ|8IY?K#d#?4!<;BAQLVBEIYy^0%aiDDvWZE%H*Uyfsk0ob2H1Xxa zRd1)27Hf|yCaa7qD*;b_*eby`Cuw4_HRF+8> zMhvk{I6PUlg8SjC*rn5#!+6+#XYY$&)6A7Iw)H!eNrf%)3v^CKnn64lvr3qpNi-gf zl_WIrAV;JPdHPIWVRJ^n{NJ_oHisr_rSlN?F=Og8+Zdd`duP0!%m0=ta5svXsezi< zo_s~)IioS2#)J`n(}Cup#NA?~vmBSyeoh@6b>K%ie??~Lt~pFaR={)-BV9bqOG%KG zYjO!;oB`ggJ?Sopit zrgC57oiQGvD-oN9;8tB9QC!UfqhaZ(o1=NdC^?Gl|1!o}cOnGa^`{yU z(^F1sM;aGm^W2AKgiqQIPd9S%6|)Buy&x0i$>?Ljbu}#9*|Yb?17QORTxhOEi$Op+ zfx~dDTE4X@q%^}N@-1OLP>DYopqG5#m}~&6-Q$QD>-~v?dv+2$G0Zdn^iP#OLXbRMo+AJQ{q-IfuK; zYDh0gh(dL3k`66zB#{-;s}bsbbc2`6`N9|1jKQ#$OQwFA5pP#R5XH*wXOQ!niJAX9 zh*qR*zH&UM6XkAyJ)B)(l0?9yahLX=552Y>EDTsp)o|###?nEREoJr^E$Ir&8xW%M z0fx`vGsZcMDcc{mPB}NqY$8swJ>zrtfc;s`P91iC5D(;{&SMB|*HcqtH0#+&(Q=8P z3QaxF>-T(_uD|gd1LAr&$Rxm4UZZKc3T&?(1?0OR4k%uxE|9YbMZsWF@4}+Xd_F=J zEqs^!-qEB)(%=YD_EX%0ZnKa+hb+9@-JCL7#Sn4dTFD<3s`MP7)m)&=Vh!T%IyAKb zd6jpSfACgP{LG2H{59IBQ6D}cTz0Q@!=W=p3umXArfes1T+Hq`uN~mi>l*c)7%d3y z)CQq2Z4%)&4Si01LEe_wBevJ%qEQzfD%M3z`=1av%;Xc}-i&&z4tdz}l>m}aq37qM z6jcK+-7G5=ze~U43)dl|H%dRvSt#pXZ?fu#_@VP+S@}L*a zjCcpQM!-wjj7Js5tCsrT1bZa#G|=mW&WH9u@Z3n7CG4zb&x3vvp2ej@4pE2MA@6%? zx<7VpRWhC?PgEJQJ4^jOY>}5v+Izt?e6s6L*+*p^7Bt$#Z)RY=-pV)-L;5PBhK>5L z{WfwcTPzYZf{v#07DOIyK$Hl2le?53>~V{#k#r;IbRuN^YIRJ$HNSN7UHUDo1vp=n zn*>!c1#2TJ(tlK~j&3S`jRZAU)#a`SH9l8xd zM8pI3b6Edr?d$k3GIW)t91N5SkOJyEE~rMl(K1}4RA@)o!LAM?^S5wH2m8`nwx7Hb z6VNrNTQA$o{jdRQ)kob%Nd4LJOJ;a=!WATJ{FkDF&8fYzXO-CeO0s!V8_MOvgxiR^ zt+ebVX*T5h#>*a!T%Z3f96URug%Kk9t1$&`XlDI3lYbheYax{`KkcE24q4I-NT_G% zrnXJokjxZ}(|3NpRhLGrLae7`##){XZ`tX9TVo!k?9r<$@yj3c5y!KZb=d>W2q#t< zX7~W1tGqI08a04ga~N_jlBc+Ugoq-V1h3%e4VU2N$Pq~OKOcI%hEl8=o5jlKpzo@b ztp{AJ!xD6)V}*V@Z-Dq_4WOKRpzrygEVLlD-N~~@mc8r8qLbpItfz+)Y>AD8IC)JA zLgDi6Rpj5+pN6z@8{mN4MlMQ=Mi&o3J3_Ql*0q0qIxhEw01C&RFeT|_w^q{R{v$SL zxWuL{?R>T3g*!Gwm0<4{00|5r4+zoC1ltb#>N#P;E7GX6epGku53}mRhaJ$TZS`~d zPI-D76`gEr7QkaI3IlNeEN)kvtG#(CjjJj+qA72tb-9kG8Kt$jr#gy)AHX{q$-}90^=USw_U7% zDbbM?rS$XODlq5JQ}RDmRz%FQ%ZC4rim!w0;Z_Qy)ha4u(l!l4Oq{|L^lwe>JZ62$ z^8bZWp6VPw3^F>PGuPFY4!R1{bvlh-3rFdU*+>Rz5Id97uk2O!{B@XT3Ib!8PlRC! zG7@6z)O=%D4~2JPjz5Ya6f$U!OVUd?BiW;mPtu*0F;h*oC6rCzVeC_vv9L|A*(~qwb_8_?o^EbRxqgSfdt-Atw^$0^KT4SyZb>> z?ncnLK<1=-j(^=!(Zd#PSQ?bX^;!tVM71YI{cW)w;SaBPz`!qR=MG4PTmLrs7ICe( z7HHD3t)x_qDK}|bbPZ6+!c~ zv}NR)xj9WEnYx`s|AFqxLUGAt>_B>yYP)is{cIVXRmFxrcJ>w7J)L`uJTbci21{U* zCxp%$FKl)*Ioef*Gd?K8e5VPh`;AwVt0MK}AITH24mZ#v`^INz`+ zD`g{7B+fr93a=OyiT~48M$bN&sINqo;*3*Nmj4@XbZl1de>cH9ABfYvw6xTluSohC zJtaJ?{dZ6GvmE^Z(yz1#R>u)oyx9EUPkkfH9H$eSpoE#PbDj#$3;QPOqsW%BlEVox zcy2~lq_Q6!vEL~Ph=HS?e^1TsBXoHsu`H*xnCLCY=n&jZP--WAZOgGCI{)%%%@zZb zab!I!PPpqSqdTCGmTQR=`k#eR0 z;-2P(I>S%&NFi&C$owf56w#;)1~A8j=y67oT3QkkNDWV)aY*i#<~VUo+NN!9Q0G%v z1Db4-{5b}KXF46z2}MU=3;Q8 zePL4D0pi;XxCAJV5iwvktC+p6aG%FMb!K@Sqmvv?Vt`gq?zR&jrhS_uEy4#$DkQ&h zMBJ(hh^L!KJJpZnE(Ok`N(QMjejz;14D3)6x}&OCc4D@)c|(~Z3Y5FYE z%yQ4c;miBlX>VSY*kyZ8kFUUVI)XtIydCw^7A}I?l&0=ygYgE5wuXXAwBf%1!TO8q5Wc|BT+?D`5e?PyIYWa9n8uFV-wp?2AXC6TLge4U zh7V0wWfGFxmunlOTPttwV>eACQ?ZN4{wjsDtg&TBLJ$UZ%C}Vlf%B*m^zLGB2lakmS4gE8H^c(6y=?Mb*Br)wbYFMe zEr^Wvye~TfTQ?-rGZ>r+*Rwc8q7sY|%+h~x)N)(s(<^T$G%D5DSKNwN%a-zVu`4!naufrmOQSCLi4R!*MkKNKPfZORk%_Si2%?*df6j4Ks z6g0Q)ih%viqQMKF3xC{H;K9UMJt3kB83`YF+%-4$xzQZ}x9|`C%5(LmKgDIk(lMhp z{9{lrcB?AJW`I*2rb&<*kXEd%7_h@c!> z6H!h35&o(!zrQFq5bSSbM&t$UR{Gxeic#9IhGIEkWLd|5Oh4OoTwx3Oyr-G5VXhE? zvh^I35uV#AD(apeuKMjeLmluj+T+JnykgA?n zceLxaa$sD4A$`@fC`AmZVZNgD1NkgbIsl;})j-nkziD=LKEWENkPz`hK@b(6a{9UO zRW6oU#r?D@OZoFVs+&rRw!%kL@An)vjTqMqbdg)Wz3h!{Lu&cbh#Rhx z0t0O(!TEQz*1=A^t8!NfLm_03!l61C0k*|qisBujGF4PPbnE&f1v?L?m=Wz}3_*m8 zSRg0Ldl-rl6r54=H0GKcB+A6b-A$@%*%T3_-mgM6L6En0ezX4I4@n?qh)`Gl?DkA4 zIf3gW#8Vi z+LURF69iSza-nl91a<(wb#~G*5*hiK{!w5JQ60(c(xAiyP$WU}@@qi;Rzc1wakuDw z<#k-3EL0i(($-oy7h+0YpdHB(jqeyzjY=v~dKXfmtqI?-QO^CMyqF(RSOS43Fs?UZ4=uqmfHs5{z5HBQkVM5HUOwn?$Jv=YF;w=jtA!jNdSAD?Pu}V zA%4kw#;b;0^#t1e$8<2=RX9>+t>9t^6B=J&h&7uyT&-dXj4k}Nh{Ho5I@vOHg!oN@ z*K_ggh#GK7u!ro6xJaL5?6kX+QlWYvq5_|qSROeya@;dUEb!y&SJC-HurUx1ENrb9 zhoE9;G?D}5RK2n3Sl29J1c{NfNDX~((^JQA<2i*k?g3VB z2be2!<5@^`#$(?izD-W!$zk`Ll!Fdag1TeQnRpL-+D(N2(;1%-{yX-g^CSdjcj zVcD(cHwoTDcj=(E9!a;UYmaMr6ik}J~=Cl$NvoVyd{3QyFa;~9UhzI zNkRm=>i+APz*7QEd5MV8 zTYgEoE~V2kcx2PB1puF1CheQG+j>R#em)HRbs4dOXj zpo*Q$c~a~#x+2rb4NjoieU3?6Iuz5qUfb!a%!+~?_iI(M+IV1sN%|f~mmA^8Ho)e0 zPOgi>Hhy%JhBo94|3MjzCrnNEZP*SA##xEA==CN;EmVrG+KyO$Cd ze`pSJ4*FA7+|&J-Uw_jr{{zbEOo&+c0BgHM1D>%Yov^;YQnf94|F}ss&Y5I=Cd4)A ztjyOy zkp-flEWmEU1zlN(NeR}(^$C9YFR6!tr`5LICq`*OTZVidwylB)3L>ZsDc@sEK{J}6 zwgH1Halp9_{cZbKEMmXrs{rlXAiOfCY@bpWB zm#58`fOU5V3GpcjBz;jK(g1&@306EsLKsaY!TX-vQqmoou&*Cso6>&{Lr#~zQa-qO z*xkbDp?lYd(UJPtc**Q+${nKq36Vg@Wr?%ig@JB6oWb?<|HJTwF&!!tF-! zmutM5GoLfnI@Oup;-fyv2B5%&$w@TuUDs_HBt>gTG2Dcq`lDTgnlS}L|IBM9hqBL| z@Vu@7wDhak_}e7{XN+xa3@|mvx`?t(it{_OzW(uIF8ZT)h_L%z(09YY=5b@_aJHKf zA{g3p++SR|ON?b-EbC9p8TeD;ZcF}jnbB!F$y3tOu#aTu=?eyseHaVs-~BU(Xe{Y! z#4oUC3I0L$%F)6*+DmKOHQOs!$M}DR2%icpTiQUKOH&%eDS_738AY#aGjZ$sS~R7~ zF#t-f!u1FGqmSp&r!fIh(u%_g`dau*1qbfkF-V*t?pOJaXN7j%l@%m^IPViO3VC7B z$^vZ?>jOFpAb)oPE@i4hgJVzQHiBAfe{&Q2W0>s)!{#Z>JT$64?9Om{+nUp#W5~d^ z9JXgw6)!%GA?XK>tw0!C3#I^Su!`7S4SEwVl_6F`@QckbE64q%zIKEwwpX7Qi8}kV zjO@!bwVH1p*ez%p67lbaWqY3F+V)XW`W1(zBX>Y<6e2^lYHfb*dds-kd_xe2l=P-r zONWlY5jW;(EiIXi*iUfsBdC`kj|%idfVY@X!8bb+BHcdXvR11Af(V#>f=bn38g{B# z@VF0`87SR(HwL$>N!x`2gqmUYc&8iA@NIIgc4uV59aD`Bw*sL+mh`0E?NjWDBZ(AC z=~ffp_5*>CmC!b0M0dkbWuxKq&Pdno=- zaFaEx=RV=2844sss+R_iaXRMtX=En$TBX-j7Etu70wEG9)HziXXT@6Y5MI|GY7_Ck-1XB>1o^TJ zzx0;7DF2-eM1L^?WZc{6>+0tD#}P~36BRpIv$Gx$0Y;Vy|GdM60+3&=1xY|s9n6Rp z;c~QkMgN(5Pf?jE^M6zOc!1uXK}&bd(qiIoCzT$i)+vBy+=?(7(&kc^z>%TmL-NII z?!EGGTUVDe2{WKNTl0(E+$3dx*4I&dFpswOgx1BUfY6B-soD*R*8J?*ovrCt;Gj+` z6}l5w?CfQVSTnl>)7K4%cVURErh!}=&9Gr&c3m)>K}DQy_U%GRXW|Gl%dlf3Azy_& zJla($nB~dOcW-9cDF!LpPAb?88W|h7U$A<(+{eDshF@_^mxEOHQjL`qv{(&PLsf|X zm>I%>%Y)(YWY=wZ!&a=7)K!Gx{5&;_f;5$A_G%Z-`%Jr>1op>p+l%X)r*_WA0!0i7 z#-w4EDIBOkJl^&`)Gqrldv7>sM_B5H$L>ajs2kq%f5)Z+w3zMd)C#ukmiL$J2|GYO z8(3J}e-(}$%ZLOeMqYz6do``6%eq~AkQ?CiM{44;;CzL(;$F>(nC8s_i3N-k>-a_g zkEnN#OZxu)z}MAv)5^-buDNWji#fwfsI04NYUbRsLM64l;Uy!*8&GCxd10-V=8e=+ z`=pY27X|OEQpt&8-bg7a710n-5cnN>e1E@R|Ln2S!uREM&Uv2aB~Z>DJdB)jMtf-c zjAmwzpSLU^0JWJ=f%Mqn^nW0|@Bcnd0@>`>T#E9WPVkFpf{q5Zj^TA;%ddNuuPzG! z5okat8e^dZRb6f-+cj$kSLtNem(oY1XOQq1x3R_x$U2$Kt--VB1ISSEU2?=*VFL^1 z;kBi4NgGD;!5rvkdZG*bMTk{HP(5k2jst0=M>KQDh@!@^+0Vj#3^YFd|JIgtRcEL| z!&WX4Dah0jdQf&yZEjUT!gtB*e}`Y)f9ZWbG`RIRZP&Fe`m3D^D?=@}KiOBXBcIH)_fB@tVH)@wJbjxBmF1tOgcfsp<+)%FpwJUq?`(%CX9x=PP-j1N0s)6IRwwGQSl(4`)B@lxaDsUVkOyXE*mG|a0HfBC%V~&N&QzPlr{lV=jnw@FSI6X_B(l6b4RUeTB2?7e;#LCV zKk5AsO83P^;q$Jz`gV6=_|egAm<2*CA)FPLXT}f9CzAOym5J1aMyJVl?^E(r>59`= zZa1j*O%HON$PBW>5SQGkG=?cUTM;ejo9wP&*u{?GllU@4*JsS*(eUe{cZ%Q{ym>SC zp5BwtyxxIx432Q}p;&|h_D%vOUuo0#Zg9|os=^QQS9JnL00}JkRUQnUC z3X0^LW*uTPmrru&#~?TiEY}PXOW{{V+g~JRr~x=ldu!k|hDQl3LrBlDQcxVRf6()2 zbt-*a^ksjT@+Wn#v_=j<3ZySx9Q_&mMiJx%*NEF1?3_y0)*(8a94BODbSGL;mWB*j zkOsp!8p?p(SVj}mf@NYcfES*?y{{+pi3|S|60@@dl1+0S`#Y+@&r=>{&9cnrl2_Or zbGiI}QV&8}>YCr<8yYs>o?mZJQt!Kr#VRwJxc}knaNh5C4b1Bsl*BwC-L>qxs-=%o z9I-p5K7-ES(`u9uWYnC$?g*!1eLFLE}bngv4o3(d`eGESSuJkK$D!;`)(z zvjtQ&{U~7*Ro#0k2!@Y6n#AVM(Bmr*8)Vf)7lxU5bk_U7uMBe(O9ZSpoU3Y8)C=2d z)4{DEE%`Cjk&&n$_f%ru*TG57B0fu^ZFD;zR zQEv5Ce>m%BSXeRZFJKj$6j`ejEhbm^Hf@`O3l^9bH}|?$`+Q8sVTVQ}Wqlyi6i*JO)s{T_j|5qv|D(E&E`!HB z8o06C4&$}N|GEp$&&S_HFZ>Use2~3c+1GPu(ybEWV|WSCw;{B8I2xER&YvjbhD}p@ zXDV-d!0ef!1S0%%`eNm*H;}~jgKHMRJ*0zz69sNv08NRvcmPv0yOrvOJHi<7_{3@F zhf%;nRv$4y`52?Qp=6z+F&mX?(KTVYv<9{(mN3*FSFM3V6;XXFTeY4*jS5A1FaM2& zSjheCjgJBn!MbK+Mu{zg*kePTA23m#s13I~3Xadf9<7Rd8&a2HG+Xf@!X!d=KNIq_ zl;$)7IpQu*gY+SBO)Ctn*&(1|$Nb9Y(P*$%^DY5o5p0IlQiMX4 znaLlmoMhW$8)xmuBQOCA&_Hcr588PP-R}G6sm+u?N3cIH8+x}Hn(mqKA$&poGI^E; z#nNhSt?)$>!H{Qq8j^HJ9v9KBPHRD;6hPy}MQsh<@`0SgN&a8R*uv>6nvjyDV`}8N-6atgii} z3NJTmhrsc>paKB-e@Bj% zB`7$C`xE%lchut!@MwRKUXm8YgbE=ES^2x}cq0Hfqr-P4ogX}rBYq1En*6Shv0n~% z_6Y7$e6s8ZhsZ`UZV(^F&CTtjuR?PE#tqg#EUj%t0F?zbQm;ND@*v-49&#pF2pC z!p6Rb_*u2$MDq%nNKqls>n^UYgcj!Pi*#7elnWe%uOy?=LDeU%5d}Cp?Y$p?W#`0d zW%WswsnGA7Z@`Sw!j6v>$s8rx;yT+ zniiZAeJ3(xd=sEoCRY+S*j@axX+(Be-NV;)nyyumtPdo(L8i(E$T-z>bLV@9&F#QQ z2>!N=$=O}h;&+f-qQLr(NY`UNE3hjYlXM7_EO%0X3=c)Z2x~P%N}*quAApfhophi* z;-_TQ&97-}*O=i!cyXI;k*YL%cZw~fYeHQTq4=F%YPII6m%>B-FOo<@J?M`)-u(Sn z%dN%=a`DH9$Ar(8W{9^_v&W~WK$bu-9>gaKS#-bdXTr0%DdYfqDIJh%`78o9`!V!Y zkU6~hcJFg83U7@_&Q=-iqy5QW+6B* zm)Tg8Zl*K~8v}__y4t$3GrM8l1$|_xYs0d8pz{WhZSIw zp@n!>dqo87lFRHTpKE0+L#Z;FHD1K!tpcrz!YePEFr+`6G)wECA36`+$`RTf_&$!h zTPbm3JVoNyRt`4%cG+_lWAc8X&hO6)2f7=j`+mWJ<6fHZ(f5#L?tJz}?(5t*SkdrS z919mrwd%J00Ea7NQQ14Z=MC2_71=OFb<^Q~??b)hh9D<)d9dUP31OqVK+?&|C@fFb zr%_F%J0ADUx^hD2mY+#KMxtV73$I6kAVYW_7mU#-Q1_Chv2W!2D=~3Y{pWqc=FXt4 z>!eLTRODV0Sg47}Xc^^#$)#LNAP8883kSFp#^mTF&DogkfI(qlf2)xh?+7vjoP_a9 zxBaxkV{**$B0DmP4V;6Wk}b!&X9Oce@G39@8e5BKmqTQHvUGDN2>#M#qaR8saaF=! z3q{4c87I4+i7iv#fbv9nxMprGawIO10BgbwnQ?GJ7w$HstobwXm$~&sj%+k=26K>K-#KZevpt8pBf! zm!lwt_@ZI>oUzG^oA&BBYqTVy87PYEbBc7CJ=BQ2Hv<*8poE#W+3`z+flIaId`$h_l@3zl(m7`j!5pgFDSfSb^RyXWqzjCR3*Fix<1PA9x z*F%hKsNG|-2qjefVC%6S!LqnLnr41$g@+&>YX0`$A z`=W0}=0GD7`{%&3agb;j(MuH{Y-$DAUxw+c&;hK5adLzZncxN{hDF$s1)r3*R z@-AfSThujCE~%uFylP~^I0ZoE-v&F4H6v1KP9CJ=X|v|7!~zJk2e?hWw^p9Y4RDL% zH!WV_Z&w>(8IErYP$wltkbxhRh~AwDMR7y;+?L&+RE_{zK#U>yb%Gi;RjS|m{Xj$- zw4Fuq87q|Ktb>XA3R*8?Lz>cS4eGyKJQ-FgX_IX3E&pzA#|~V)IlrW$!3gq^>?@R4 z3Y7m4cW3D+!mh^e&ib{ZiE0BLP+7WgcTr;D!$~$nloi-7fSc4+STq*_6NGx>1pVDc zQlo)M%WC*yJ!njwCM#nK*qn+RWSfpnMGh^3W0t(7m2qsNGwzDx)+|$*!>elT!S(~e zzac!ht-x;GUhZa@Px`4RaJX($Xqq7IwbF8?Pg3@Z^l~_Si}@Ato^sW)d+sh6coX@+ zgfW|U9f>~$o%GnnzqCtDkJf~la9kIGmcg;Y4XBUn4OGk@Z|c^*NdNBk+@rVQ;()5Aspobk6DzHC zL+WukKHL1xwd;HP7$HB&J)nEejhvk`#th3pS5UWrzmT__Fc#OF(@^}KblZt;BLx6g zngF(ao%IIo9P{L;;ilSe@_fEU99I6(RL1jjK=1xl4*!n)|dG zPBotFHdVj`vlVgAz?uY`pyMu!I4=U^M;?XQ9?WH~XX&fUp$6g(3nxFN4y4}mAEtG~ zzCtL&nriEm?3!U$9bm>S>(udgvG+0lwa55PqMHr z&;U>Ia)>-h9jT8&aDZDz%OCB`Xc{w;*E2Y75p|sFlPg%1J`3Fm3ciCiI5awabSM=4Fuw4 z^J{H>F``_IUn?p(MSO;aUat#?TDFHyNcfRq-EEvY#X`h&KlUa1KtF8owAoLPAm{^F zpKNo)B~O_UC<^(E0=z|UZlo#_l-oEtxFb20*s5YZNVkj(FEXa{$YuzNp-W<>78{t0bJjG)@q>5Qs5u~mtuTJ`-jvuDmw&AZkXoirSks#Z zw^%3@vM>geB7H`f~s29CON!s0|dJ@{YDoi-BF#~a-xsH6or1znb6iiiI-$*=##z7m7 zn?hU+GBs_a4we;a^Fm_)?fwLcUy@CSI-25!5HJ7rac2+5pBqKD5SY?G#e}G8XZV`P z68@uJMl;`@+^uR}bmwbbrNcCvrVmj&cp873H1r`vei3xw=)FV7yWA&KwI2v|ApANj>rZnCV*9BwjI`-X30e7=R;*hz z2k^u#FcIBr1K?CCi78rT)){63tokD7rgE&jR1}}AKU~~jTY}~7k4>v3n7yM2ERUga zs$;XBTG5hELlJ5@g)iUHQ?VXg(W55vY9?or)CyIwgZ@(nm0t^ZVrwJ4#r&M-i8|*x z4}(aVIMw5>IK|&?<)$rP=b;0h0zhWLL5^YYPpHXBtj~L<7nDmlpIho+F>o`>Q1k)J zH`&p-)Bo}81~4D%gQSo5EI&YoZ?Z-ShBVlho`HNX)Htm>&c$n0IEWEuCX6#?+CcR% zU(3$MR{>gA8y&3d>KbL*K70YSCg~9+LJXuiv*7iql}mf8Fayhg%h*Z?=_jg0h`_j9 zDl6t;^l{qm?o9p#lYrdSv1!tI&*}RM*qJdw(@_0PQ_$TR;WM$c6xnJqdq#Wdyi{@A z;?5fOCFt8QyiHM8t{bSmW$iPE+u_LwhD&Qfge&1eK0WL0{3CqJN3z z$B`L8>?|BJ1b|p&%5tIYc2RQcU3i38rR25cr<^}H)sJBH{8gUNTk59al{Ce}#a4)nK2D+_qV@l4BzBk21d#{Tnt zSkuHY3k5Mh{-Q9VIqs04DR||`uwLievBIWe3&2G}+gx~^%r1EL6R8>A zfaD79H8|%M!YgVbBUiC6eP;Q*ul0o3-vCvIy_PzC_1%9Md8qa5dL)uF<}-o!MBX|N z^~d|-_V9Tb1Exo4&@?g^y*xZxG-k$4q5NO69~SP-H9sJ_e9k&4W3fQqALo6LtB;5( zFQ#jx&Gyaq4)hN;N#Bu!gh%e%*GWAT_5m7kC7=69&1@kp!j>J~jTHq>b@eq{0omK8 zL7Wnuwx9jOC)c}shO!PRYI^JidxC#fL&6Hih|&^nkC~&&_F+X@mXQK1+-{)FQ`HEj zrHxAEEza80Mir*|ZmfivVXqM*_SUqw*(T_>Xj<-U_RFaD8+%KouZEX$mfNXu&@vY3Ngh zJ<0=zB>0GnM9XGrf!z*E1s=wRL%?RS=PO^pRlj+>&no$wS;ED6a?jxRLq7MH= zG55>7KX&f2NCNT#7z{3N8d==vt@f|}7C^$Z?(9nreFMF#SB04b_-57s1oeB@+kpp1 z!<2iW(}xUEKx?hM``*X6FJ!5vHjNU?)%L9H&!k|&!I?=dA6=02zZD`*YsZ^TG7p}F z_-d6~&f9@~YA<6g;4V_d1~Npyb(EY6O<76MTkf{-KvEwV!2 zREU5gK~nBo#tCw1Tw*lwnX8?h}Ksa52(4GytyJspG8s{})r;+!TK3_SkSwDiv z4r<@$~x!Cs{QojvLd=SbS9d?=4mABR~1wz3GKtlHr>fPGr+YGd51`l8eI zjl#v}$+6k|1pZ}c0JD_M=sFYgkpPV|R zK^7%us*EubqVXVSqen?aoaRfdzF9Bn(+cU|{*9l-Fu07}pZpN_*G#l|+ydjDIM*Y< z%)KGAoN9%SZz|q0RZvoK*o5Vm`HRcn$_JT6WkLUa?a2{fT%qW#jgoF-f`h_!opWsm zB;3w46TtKqlfe+t5I96?g=&2ry7&$&QL$)oJ(m81Vh=cyI=rA9kep#r`N(S3TH&kd zrr;TSvYxe?Fs#hu8_fWCWgI=$=8$)RHEZ8#pDoM%2pF>P6C02FV83WbCnHz_B=S!x z*XMc+KzC4!)AV!Af6eXIDGU__9VK-WR1wa8I}qBF!{aN+C6E3?wb>pxQyA$qbr!-9 z(X}eT79f1&;#T*PSG^SG&(bqTT>fduiavqHwb)sO{}Z(dfnPvfgkY!kb-^-Nx+D`_ zmIWgO7++$*X@j)iDi9l`UQ)Cq0>TRNRVD`(bHqZON&o`@?vBD%eYa>p-rTVvQzP>82KW z=APtW4~A=_1E=vp(krPAl<%geh&_Wz9O9H+xUWH6$*JW{`0^*21-amDdv_t@#a#qD zhmnG&CaWXMzLbOYg$0P{rg`jLXZng%JA{zp+gmm43M0c2msDbz>wnaDjoH~Gu+eRz zO{3a?GLn7@E#C$#GZ{1dbF#teZ9AjY$S~ZMaZL+!?b{tDRNa_w$*9TZ>;=1jSM8G{v7N( zjpQvqRXMw6?6H9V zi#*w3fH0->oiR@{$1*^MoTz|DAOrjoUO{IfPXNeVk&<6fq0XO&`@h+H6FNria= z$+}%r#I%=oSO|b`O2KKI0rmq^r1m~u!ydG48oO_%%Aaj|_yU}$eGSx$$BFcsq6MF@ zrkCgP%jyeM7eac8W+tMbIBi z{U$j=%xdoe!)M=IT)#$p4}WdAHr4f8e4PRDT7NH~(lr-zQNkq?d=^Gv44 z{8m;Dwi?X%nmB$vAlBAKOUN*Kte1z38!3a{k!~vJGYXx$+V8E-aGBSMaVFYB5{Kmx z8b^T=$Ng(b6%?Nymh`~^glUk1+P~)vu8KI@Pju`F#4h03J$fHUvNkhztD!oCjP^l{ zM1REpA!=lx^|PfXNKAx;M(hyQ1u;bW@=?aqu_}_rA(k9jc9bLO2!e4725HYyCYg3m#iM&lIK z9ynXB9Yz9PK%ICMKhabecI!#Tpz~x2Z}&&mpMOquLgGOc4YPc<`8wxk5Rh& zeDijyzm_ld`fQi=mJ!80Wq3cc_0U5%cPwq*!#7v55(rAYs3XtOeFHNvw5v9Z$c}(U zP{Zc9eMaXL?1FP?N733#M{9}O;2WJveYC?5MoxEn%xenW+zVCfhXZ-Wfm#vi=u*Nx zG&M~h>6oL~1tw9pk*7zC{aMS73l*-kjDS`g%Ixj)vDyCF?%Dq7yhG<__awzs#j*dO zw~vOyUhYNz zBy2T*aYsSD#!4#re1_Ap3r?Im5ar&r=(?yA>eYQ4H*>K4un&9jvsUtvA>f6+7U_0^ z_)fKIsvc21d&GUh;Q%BUzz2WbL#LYI}(%{2qD4+&6P5P5{`X^!1n zRI^@r$EHx1MdV218g*i(^#a$||9vO_MJmVlW{rQh9mFhxI-T!0^&gPwQOjYq45PJ4 z>KE^ew2q&lCPl(dJ%`7W*3XOYx#Ldcb`KX2fVN8&E$dSbD}QYg#=X;83sNA|O=(U* zV%EyY6QypPlP-@MoSfY}j=RX@qjYt!59D0$Z#?1hA!#Ijo%6)5;0Yjq+*KZWIJdHC z(m;6~O-l6hkywqOEOR0BJ+)Bi!rn2Qq~_L|q{19jx2RGoKr{l(*CD5o@x4 zunyVDEgZesUC6JALA04fNUd8f{E2Ka_9$vnRvAqOHTC*FZ4XL|Aw&^L2U}Kzfw`aw zUUUm$7}NrrRT4zIW6t}Hs;XJ;tTl@uenB<|KOK3K$a|`!SNQ&!uztdVLUcSE0L?Le zvU2Gy9oe15H;8bRU%x}S!Y%U`hkKFYrW)JgU~_2BgWXbNzlSb9xM@Ws;M2O4SFSYN zV#h(jHvco)i+a`$$pM+IpIJar1l%@Dw(;Toy$CCE=hT5`)c32CQfp%lowa^Sysfcd z;c&~oDh3ghhw)v59y|f$QVLD}+r=Gf25%8@te+2$3}Ol>bYEYXZu5pEVC+JDl?H>; zq1tmki=T^8I^K!7qJFxBP;F z0$<|h9 z&k-qd=f(A$7buF5DU{L)xptDsGOX9M7Ir=ygPm8&UMjU0DD8GT-eQ-;q{#vK4g&~w z5t+F83D#Cw2Rz_^!q7hM`^1#>pQC*v=r(x0u9Ro0IM2|>;+^DXpd@xO{Es4jUJ*FLe-U%XAwcdkWWbMB$C6q8mRm#>aNvqbfuurAt za%-M<#oMz5RpNn-L8|N8IqLja>@X^PREYtrj_B5QMugx$8*cr}NoV@C8|r~Z-TllaE`jWx>qd+~0fFT&AF5*}!*L!dLbN|3ZCo`4BARv}Ox^5PO%8ED zX{Gr_Fkbx;GcEl0i)bG;wZ=hzcs5)0X`uGXm7&~0W2YCRatV9bp4aI_b(4_8X(p~ZEd=HY=LS_5*DOqC;L#KxZ_6f`-9uriL5LNxv zLG%B+Rp5tmq9|id6QGs@W9ro=uubMx0Ec3a!-Td2_&>+&WG#8LZ!NN={Hz8E8)fsc zW4(hBmV6rsokn;IX6)V$iiYlr*wZmCwX_b&K(s>ze`lJ9YRe?piz8~hC5_lO{rvL; zL#u@V7Oo~wpQ;pQk`mn6IVxIQ-a4gLz_+Ggr;TdfndesdgqUWq5KLL3gxSCJO@C?T zN;p#bdpYPo@2NgM%xM4mXq8SF0fYlGARvE;+{}g5!{-;yapR_%YJhRoG^?RfMadNx zI#tZ-?5{t<^UYfsmLqN$qo8xx3{2+L2`?>#EvVX~nd~ z|NFal;%m^SwUL9zT$XU2^1J7rmYzQY>=t+)wd$2n%vS_Yv8`X+@fMURZ$$a!00pV3 zhH{2le@?+~5X|{)Mw~VAjnyrL#A-|@5cc|tiOHsokNr<5*4BRNcTmNIA;G5y4a;=| zo?QzTo$<}P0Uo9BR`Mvqf$6j}Ce7c)R@ch53I03R49*4jpLp|GXQ;2-C`_y(pfwD+ z7mQoe!LZv~w8?(RmC?LkxJ3)|>Gf_qzk?G~$1+cL(wS1n>CFD4pnBt1n_;P;C$3nwDB(%Esdt4zm!Pp1AI6{)+Bu(I5ADiwlQ^hc2hz zgf1i{%R(HRLq9{EKO819v<4$~ZzgCX>S$sc&I(xDm* zbH=()QVCu^aQ73#6BV;Mf{0#r5IzN{RUm1yCEbi0G1vW-c6GBUFlqLQZ9~RaIRKA0 zx?S7*ttu^SX^*qt_H3J0jrdMIcobT~uPR_GaULEHU`^}_K3po@?6^22K=mp6BIUKE z414-z$^BW@VQ~$L@*|435#g8qW?c_(*F6jUe^Z!F#iOV2CDSk&^-MSR4#wQ~MSYH- zR`C3f#^HWDj}z2!oayg!HDJE{i8qC4fgRS))bUL)QiJv98Zd@G9PCZ+*dLw_#Bhma z5oJ%F#bRhVWpgKdBygj8evXOGnb7WhS*}v&HACDcI52b`I6c3G9Nn26j(?Nlw-&dU zYUxZ&!cz5KD38GpB|DcFhPfR(oTq;|O~A5K^ux804z{Ts$B|V6#u?2Df<=`;cm~YX z5ccVX4taB_Sy(0^owuK?PhHZkVpx%e3UhM<^pWX%XGneN*fe8)tl@M-aWC%z*{{2g zif4YbJOKB-G26IZ%~*+0$7PyCDQmsFGsK`90u3}U$G|V;de&-jP_N{snIp|^00`A1 z;n(CSvSoaWdT#{PO3(b0%B7t9-O0JN(B-MojV$am?edZdM`m^FTb!&)Hb{vIOFf?K%)Nx*7Y z6K`>NuWK^D?(0u+g}i$~XY?wh2pl|l1{X>wUQk7paZ^8%{*V_+yuwfe9cwRl-;kO5 z53*o}J+;q$Am)*m)(hn&|jYhJcMKmZvWpUg$jQMj!^2g7T{m(D}BQDkq3#*sOQ}gLB z#-PP<$>x1*GsaN8^2(Bo$>#^Csj=}x5e0kV8AweyNA=;&MT8F%Pom7#FzCTHVz)XD zr}L+w7^>}C%!4Sk=LDDYWojRF&k&Ja6@)jJ{D{5Hc6~d5d=nLTjd1}vOwGc|&lj?g z;W>Q$Jdega>7h!Kt4#ieBI4jbmFHTK>Ii4>3<$eI*4en6JdH|)3uki{@(#1}F=+1k zkFNiS%NRUa=_BEgGUX-3o2Ke9G|}m0@u)%WT%cJ3wTOM~8N)HxRr@0CrfmO}HAtXHiq>oc*)#M)1h>5}k!Vz3Ojd?GzM&Jcd0}!sV?9t=CtPX&=F6MUVYQ za5=#!92z4lF`S4N!9{G8x}5u;Q~(a_=<K^@r3-70^8z@t9PNUZ@mU|DVL z%pEsIxS~GFDC27TGdEJRY2I58ykbe+P5y9u5B4RaTb$8mx8sa^5k2A0(tLkM#To)i z@KC4Cz!|hysDr94t|jl>RYQ`Phv??-Vd5{Or`HZ1@6MFL!v&|^NmV#rzB^HWI{Y`e zV|?p(qL^+}{wS@7QlNtTwp)1MsLTkj40~*LdH|@$Zg`DcCcp*1&-dX=|6de0ZSqTP z-AFB-ZTp3K2rGNm0tf;`i!@J?E1Wj$5oH*M4wmHklfGgGvzrgGC<6>m_3*ecxdguL z2n>}D#p)vtGpb#qt4x%q zINa3%bw^!d-wy?}fB*4c3a-x`;vi?QJ31@(Otz`p0y%1W=-t&rE<BMWuE~EU{uR!o0cw$)KN`Q+(=YnN zlqtNgMOTtHusphAWxj2sHB zVPTPodyL7Ny+A_rTM1#0D ze}ZcYvSbg+5>*p%H|qlzC$P+f2@F!JNZe|Wa6@w2&lHJuz7n#I$Zx40DuW^o<%;QB z>QkgIQ;os+TjCTj)a-72EHr6|QX?Y5Gol(r&;`=){@R;fjgACQM8MFJt???|lP)t~Hl{9aD?n%}$d*Pud<1w6rWNr(>sfXb~Dc(vIUHcS4 z*Hqp^a&LnK^rZL5!hzr&Rh1Hpp_pouWj_vg^KRXX4&4BX7+ZXC^+aW9ioFacw4r+K zHB|Mvj$smHKlop{4)CUv?|KoqWe1m+6;D1QHr6=xZ6Zvv{@8RS**zW%=fLzx#&1xD zby$7W)a>26le+Fw_JkI>i#~hb^;@M1jHzwAKys_D#E`9rc`r-xS&2K>z6Z`u#|60; zN^L~k2g?2(^N9fq^}{nu%eu`#&`_p~T)kfsGpwk-#28e*RPX|jFl=Fmr>hA%HSV>- zA;}LTnio-ZV~naV-bzgoP`wg*E%pB3Sz^WWM{sP`F)eQ=loK`xU!>Usb--~kb;HPn z-hGNd#`0O0!f5aJfUQ9nAwAn$9YlRgs&4sOZ4X4;MA{WnbW%416iK#e)A4-Y#%~ap z5$L$&tN;nk!C~(dIl9%A!z^wghE8BjsQU$|a%R9AA~)ZN@0ZSBN1u+sOf1&$99WW} z2#3NDreGB_!wBNJ_|M8~i{vro{wZtHhkp+K-N(bddFw*})0EW$Ip%%r;QJM1-Q}a! z_IDu>)j+;@{d4lyl%ByaN&CVR;GUAPiZPBdIW~k5%E=QP3at;AGB2rMo*HhN<`unC zUiIkfjDx*?pBv`@F!(NRP%v7no$@D5F=IOQh<{VoyoTa=1Q*H(JUf%2?v8T}teLAR z&=zeZKEvv?0&nrAKDy`D$8O?N?0f=vfa<&Is{Sk|tOO;&PXqHiV8mW`dR`B6<%Ux! zbL46`;I?WS!D1@bZd~seW=6KjZ>6Pr6LjTzKx+J*>Ehc)o)cRvDOUtP*>u_##l#y7 z9;Nb*n&7+BnW1F-jBA`-hzjKvF0Xy1Tq==rhiA=4QCYR4V|v31HzRXlTpU*#^FRL6 zJbcs$q+f~a`Ui=lFDc!eybKtToN|(iI5E#FMBs)snfP+@7QY{|cNZ_0UtSH=P8b>Q z(#-+MJ7ZZ$LeExz4EcnIw0ED(kKm{U;K4e0EWq?bd#rsl%laT>N9A-FJ_LnPe5-bz zh2)=dih=V*ePnb?vq!z;NUE<9tg$uh^8McsMMIQ5HK;U5{1s5Qe}rD(dJT&8KrzGo zW=PrJD3~7?M-NAR;)PB(`8Qd8sy2$Uiw<=p3zn9y#y1KTD!Np9U18rDkOPUyb!{7D zwp!3R8l>{7nUyXy&9TBh-yfL`I}GEXp}5_Tx#0;sk+IfXjmfIL;6GgHfq=4?h3JS=nyU~jM{X1?2Apd z#4eaMEvduKM2KT^<9!qZB}*@;g|RuqypT8jp2NbLJW0d<&YO9Tyy<;cUye z%~VnLz1u4iNk%OHq_f1o4PaN^nO`m%=asX}0s^ljkrz|-(P8Q9V$^d^$$S>$m$vWe zH7K;uN)!nb?Rv7)CTr+cJATSE?))AG`K%ws5pc321{^vU4B`dLHC~aEmArGceFJDq zAG#w9Hqc-(Y80mYw6xuW?hcO?4C(QAN};?(6}N}3@u}ub#rPYOje8r~AE`P@w zu>LCcJl>t7dLx7;zG{A*j1(tFMJB4uxsM^{Zks^WxA(QaT+d$qsP>?$b~51MShAOu ze+KVbpfJKVh1#RK0rD>Ex(;|^{#B7o&vl?fCXB}XOj3+lNid`QaE`Juwv3xBFyWc@ z%&j4nOSxDMzF9fu_m^Npeo?bb$7Sf!V&;~4n+>g~jv!CG=&&c9JRH-t6aol|je_e%yHf@V|Lsi7yy|1k_FCU9ZQGdj2A_+PD@*x-_#kRUb6X#+t-u*?w@q zw0^7N1L@`sm}qt`k#H;F0&IVsh4&(+e+(*BhH|LyD~NHOYpQg$<`Q+3RFeSIyQsjI zRkic0y+yU3i<7(+WuF1CSyeJE8d4hDYKAoam8Nz0_wZGx6sc{=z8ewVig$H8()LRq zeSEh_-v066n>!a__1?Y>CcDd}@As>n-{3$&cu&h)Wx*kgcTt1HO2J$4fft@=;ZK** z;b<0e3UQXix;Ct7z(N#^b@n;$K^3M6Uq*Ok?c_D)Mcf9}9PCh=O})?X`0=zBFn~(~ z5f8XFAM3s@Q%|0Vv#R!mfSkh4p0;oJ5ulB0(G4D}QD8ItSnm2%@t5Fm>IQ=WaK8?j zYaD5Zmathgql*|&-Bpg&kkUA@u}$TpmEK=Lz~vCkH^)kPIZu&8)oy;&=M9)h`tC_m zr=Hb2j>FYdBxU`;YY544zx^^zHRbO8m>hA!N^{@wcUZejvHNAE z{sl(X{ysIZ9~@$zP?{7p{dCWKM}S6XBWpiz#f9jk(3!$cSxf=S0OP9&4zb-N6J1D8 zbAP+qVLUj&H}j=`mrl7AJ>*0TSs#nKEHmwaF?sP8q`#(u8M?~DE4HfEne?Ah_U@F( z-s_Q&IW>4J&71VcWu>uXzFN_JzGo{TUg*!+vSul7UwW2}KpTbwFVdlc5XB{XS>^P} z8s}Q!xE)b*ID;p*lGksqJlZts-+LgVdfNGSV@^i}n)+#9(sgqC+7 zU;~932=6z~D;v@5kLD#D>FDk$vEmD@92nT|WQTqPBSm*r0&d{~!?Mr(lOM`!7^64D zpBs{W`VtwVPg9z<&Hn{*-kG|02S-PKAK*QC6E!5qrF91|Zy8xzQ96w^?zeI;{MnpoCmQvBV`!uJ zt~5p8AZhG_BELyB@_YKhp2l15@sN2(>HeAVsIau+<4h*8{Tqcw@4zwb2|}Q{JAw)5 zpm6W~K;C=W15u-Ss^w{Skq^K2W6`-WBX2bME>q?VK%l+x+fK43eI)cwn)W8k?+OyE z!e;JVO^@*1lzQ;=Pj{_|HAn03hT2W6Q>0q|H(%v5l@iF~7d@av+_Wl}INScaNP)BaKu9&t$ifPpj)oK)?asoI`)+X}rXAfU2}}ST z*YLYQ#gc}m5AxnDUNR{5Eb?R;E4t2zex=zSc*~SO`UcRm6A&Q#W?wOcmk6^bx?sn8 zL+TW+pua9;C^)oTh8D_AxyIFBi~E^h~~) z*rnVI$k~1?kgx-B@RenIp{KHBdLe6Ef0N!fXh-PKEZ#N2GZJ6Xnx z&P?LXKW7TWh3jzn$AC$A%%$OdaI8)1CwxP!%Z<;}P?a;PMR?Wz){Epicyfo&;toX$k@^v+wmLf#vxouGi1?x5z|b(rS8SZoSgLNs=)tpRI;^}}vV_0}R3 zHS(6cAAiDjnJ`rOrdn}n`m!SMsORF=7C^AU+y4EkB05)`*{KhP`=Z=Os)QqC$sLm` z_D^2ifMBH6?x;^g#cDIsjm8w`A+^sQZAN#j)pEm6?7M$Urv5r&pTS)L2md0(*4)#tw-KRvR(TY_m`U~(Yd)6n*WQ`$i@@YKkPes zTnlCXO>mO%LWz$r=?P>ddwAdZ2Xgeb>ygiN0z+#DVu0#}|EGw%aY*d8=CxIr-|?7N zb?eYtY2M6#E<4bauRC$)P+3?J|2%yZRdal43&K&-wvH8j#AVUa(`B_l`3&7@5;##+ zIbFWEwATbhodEY6abdk*V}e()563N0H)-=p+U!nNUE0q*y4F{IF)+C(-`k@GODSxETi+Gr9Z(19ZuaBGuCV8?!&zAr3#UI`H7=O=hB%w+lXT? zy*1e9kThzImqrLbyQ#d)ezI0vHC+SWgr0rnM?{Cm0vpa@>`;_lh3Bwl z_kQM$H0${M5UC~STvHQK8k2kYBAFSPtd9-gs2t{G+wutzY*!)9d68FCR=b}wt$bfI zhA*WC4g0*Q7Q7vdl#hz<;U;e&%CVN)G4gV9NU3>fUrp45SJJCxUmD= zrYpzmUQ7=O8s1)SBKU$A%+?%gzskw?0QLzngEItUK)UP*_>K6Xg(Xl=HBVtrAOQ+AvxyW=6A(&9>ju=llEp@#}VT z)81^e_uhLQp3leQ{=o5GaGU`^qNF$ z7(J1ifUlt6XZ=aE`P1NA(vNl9mz$1Pfb^*XzKTWtkRUke(Nskf_js^7?3wpMcfpp7pBOFF;slkQ81@$9x`KknEI!Xh@sL^@td%GyJd+ zpjR46XNBl%2H9`t$yaHUFa(Mc-Q0cOhuDj!D%snilwjdk)n?5G5)x1!*#BxI93kPG z4BM@j=Y!IrH{0u=!L3)~aq&)$13O#k>z$X!i80K5nTa^1eyNNhtKXS!m;k=jaR=>tXukp7}L_no$w#YQ%s?rYNxBrQv()8t8at`1BokJkoAET{Ll|1s@%@ zLgh(`8(ISvFF6evF&%pu4F`4b-leTKclHl1ttk=rR>(RB?#7kiFa2~Ku6}om;Tqw` z!MFp#QH;2dU($D`pWk8(2zJ5;%Hzk$QwU?q!-$KiTXZxGwN5!wH!v2=Glo?y+N}v- z=?9Hw=a!^SLSOIYMaVj`Ts%)F|fCF%Py$-}fL zCr6R(4Uvp>Gf5_EQ|YSa$MO!IBRSA9tnL|~3*ypGmlq3)Qj=O<$XNgMTqc?J6WIOV z`Oi%j^GHXecFj-1@u`jrOCLC_pQdEP?fuNy(m3n^A2YW+QJjHYSQP#hLZm3NWQH+Y z3M-YX^pvN^)Iz7_*G`pCXKMvMBwc~}oPGqke=ssbM)>EBs6<3R-y{s8uL+x${!4f0 zf#t(1id;2Vp}0GXPL&Y~47)hrbSHXeuK&eOUP;2XoHKn!d2?gV2M&XR>^mC8V`kAuZ^A3(q4%mku2^;FTC{jOD zSm^UUNmhKE#)%8;O)s6$Vulpn8vYwHlW`v`A9Hpc{X%N+G0WK zxqL#l{(eeijr3Lj9@f{*2GN-tYNXYJoE~fZPBrj?Li@*HRv7}Dr77Cq1VJj`UV?b0 zZHPEzJ)K=z%^h^sb!=+UCiyMa@^}d^X;LD3sRa|&-RhixlJRW#)2&H*q6%JA#z-pv zY_csXW*UWVM#Qx&gfV>}0p1s9<@QjR5Rn`fL`~D@=UB?fbuGgNCf$OJVWST{L2h@T z-w;KeOc!cL`!N>@i6z5fOlEX22YH`kku2{UL~rzC`r*aA__^uC!^Ol661&n~`bKm? z|99?tr-C4aq^X$3)J7veanKp9k(1n8pe^b3qaTn!sKs0@-6q3+#*dhsQ$f7G)z)l2 zwntwnLyzM}49f{4#=uuHd|lTj(w%Ho4~Uvy5buW-&=Jb5G+v>sb>hlPPyuPG4bdyp zv>jQ@wXp7+S5d^@_5V*y(sqzz_igGjH6DF=#p*oI)79MmmQl)VvgnjU7PH~n>I-iF zO8(U>0rdA?kw+4Ee8CX;NR0vkoz&tD5CiJ$HD^qad=>pM3N@ry|3sFS6gpmR4@I^PMvqS z5;BE|Ba==0iu1@P`dNp6{WFS;J;%Y`<2(NFZd(f)p@_Wqd6yti6_mg}Z~+Fv2Fl4& z{{0xqBJ^N-YLkZaLB!H$H5d^&{9Dm~1YUbNoK^Mv8H=xrL#-%RK}xohjv2P7vYLPN zzN=NOfpo+pfVflCz!ZFcKs_TSQ8HIa^;A5i)$C)%)b3T(j@3jx4Q*RO_|W}w86 zTH3)xvq;Ujm(piU^gEvPu#U!g@kaV|f@7e#%kfm5bRz6>V_0kHP?4!u04t1EG`;xFlaj+E+ZsiT>aWa&7tZfve2YuEE`|Hjvwv)ZYBi*R zQD%a^e3d^G4^qwD=hlGPR}BzKhkI40%$O&CoP*%SB^;Of zyOZ#N4=~g6>@)9kRtbYt_M7fahr;D1m$EM>qsV0Jlj$JcGr+iS*A1PNl>@U;tc#D3G>O0y}WdCY*|I2LZdT3A}`;TcF#a$1niM}98UQ(aZ z8KhXXC)EDT;0vgGmbJqhO(&Uej8bx380jT2 z^}~*y_plJzkJpNf9HJH%d_Nr{Wu8$xm$aHEIYOHG+v1zloP=sme~GLw)v}P)@=Bw= zj1Y>6`|wCJVAXRkkfl%F!9bzEVelAnQ2gF+EaG0&JF&7z*8ZOqaRJNFFT?ARu%3^k zmfdqVw4wT!q{hmou`wQip84zbFAUk^wGVg>ZaHeQf?k^J-paD-I1PqUh{lobI}${} z$C-+!)p0rulEZx0PLs44u$tKb$l@uZ%#?2;dr7+ko@X zH=R!dX+YCqD1kvY-Y+7OL;WhO^gOL8nZzn2P)f zOh3vWzAiJfp*>MFvkYrxHO)xL)5F-RBqWNCU3%|z>i_=X?j-&7*WSNepL^=_(u<1j zJ-KAO`?ufrEQ^JI+paxjUhVGWbTbxiFN+VW zQbDV&yYqB?v!>s9fT*k$fM^Aj@@jcw%T1h}5=^O%^G|@gSneY(OFeqb=kp1yb$;iz zT0nAvB@p3LUIWtWda~UUKFQmmuKF(MD-Bbma*86)Y<@@3faui(1ad)ex<3hR&(T(T3OCM$Rn{qF^skb{N&AL!W}uile8VXl~y-QFZ%tWq<~!;mt<<*&803 z3z+&P;OF0jV$(*S!A>6tg*9q;v!L2AzI$JW6{KzeU({`SY;%Y2A=Ll9s5sCP*o~^K z#eB>VuQW5q*{jI^t?=qt3q@;a6ln%Eyw-F0DpV!O;aq70V#C5WAXn#p5@>S(j0ZUi zw|yel6M=mUwk`g}>sgLdk`w2EW_HU3yrLkQY%_?F-JILnH`wU_3Jdt#omFal$Ef2 z>XZom52>FgxW#j;ckI3;cx+N~#zU0I4MMLu*~eKxym_IETC1xuy-?W89p25)MdGG} zC4Q(WVX>dtl(5iGF)2h|;lB8JO8D3hWzrCNrq~bneUV&%<1SVe(54p43OuG33ky6Z z7ym7weP4Y0h$hvWSV)0-n@bs}Dl*%AG0K^}W>r7+>?oIrLPjolw4 zeaZVE$6p||&Rmp*(WyzY3OXsNrHP&3Pu-Y-AqKGe!fh~*B>-CytV&0bsqg-bB-6uL z+~z5m4tXk8=0aqnl7{7zoeQwy|vFz#%(}J;_sVqLsECj{mdyU ztoQ?SL*nh)dMrI+AD3^}a(*G1o#xgofLG#PIkB-;YX^En!d`88=kL?+qu{@i!hAz{!z+u$-l~pRaKhKT|Da1@19a>lTU5 z#XP}WOX|dHk@WVMK|Er;xkI*ZecUQAnqPvuQs@NRr%3<3gj~C;7k-( zqg|sjbX01F-(G|CMTf7-*P9D8lUiEiAI>57yStUJ{~%mU59@cuG9FIvgHZH?nK>KB z6EWVcz*0Wxq}VTD%NU<}WAu+67=X}6VQn%J81fe&|$z7z97hOu(Qp>@yw=;`(VSgtkXlqgnsIbH`5 zpM9peObH38u%nxC3%9edZ$Eo{g~%N$c4_}>8`HTlRgwtUciNuVH{G;RlaVoH8MgoP zhI6Z-AvLZ!?NUmb|61NcOxI;qqhMJ)7fNibn_ne zCb9R#!k=v$Bdgozo&vel|0?^$7Qoc5rQSCzm{sOm7tHqtQ8(@lSOW=8o0w<~qK*5| zsIG?fd4(qQwb_VUgJvCN$}dqcDvJ`sq;6S1b}%?Zp7Rb-)&i|*_aLX5=XLr$bMP>e z=I+YGoDL#a%YvyjJS!8C4%6f7v{D^v{&Ah`=Z?5YrtDe>v*&&|tq$ptUB{E;@V?Y> zZTjkvS%U=;q!IH7X6*IlLq;VLq?sx-GY)wn8Gh|;O#~fL7sMopUExp=97^hdL%ZuZ z2Q5F=3H18v^6S5}M%ap9!Cf5wVblqJX1dR@>I)h_h4F1eMoS`Rf?qNL&H%gLk50qB%CT>mAZ_3``n(&jDXVU+8 zRy#zh%&6&@cOo6B?n@cPAwHa3P19OF)VX}98V*^7lN$r=2QdNM-CbPKL42(&*x8>{=pCQjr< znf8i|?a98c$X_sloV#(Dh9E_s%-uo~K==nF_x`gu(jqhmDEN!xd-QibKev>HT->HH6qX`StO61vC@#f85IR3{ZaFn`OV3AIB31>F<~aVxACQCRa9%$0*%)Lb7q8gP ze_7ZqYE)pkv2`#|z9QyQb>lJ5e2ti~!!g9`VbiH`Nam>Z*H}HBYW6xtG=<=TAUD~Q z5I+LZs`C}hS3Y<@7T0yKptRvN%;+CD?hI_dV-GfXlMW%7u5T&L93tC-W0Q~&{= z%1XLoSLw*f09Nr$8-*1w8R2$M@RO~_N5QrCW~A*A{ofO>=PL~+ee#y=;%M!u9BmXO z`P~bWF%xf7rsA#flV4S6wRAy&Zr7_g?F!atN_kp<9ldp_i{R|TN0!L2pB^gBZna1U z!&NZWUXl!bAq+0N1nZMYnL?dXXzL^pTJqyFZ1_2&YZ08Du0PZ1k3fV|w}TTQZU9nJ zy~vtO>Tn3{$)^|C;@tLQlJ&H0%{E6}EL#Q?!Uj@(L99zzJsn@g;%PdYV)I=Aw^H>Aj46H1#Bmh#&i z>ismP03mGU%@K50psyxD(9>ofUv~gdv(?u6t!Wb$>?1?zYck>kF1r=XdvCmTT|E4r>7sdP#?1mk}FY9QayQ9b#3YnzXmL z`h%nBX!kJd((Yx<2;y(~+ojh-pVuuDN;)k^An_#GSiy{R2R#AxK&$D4Fw%zih=BKy zQK=5{QR>43pls;Bp&%GHzFBPuz*oDPi!tYyM_j}vu2I9M0^jM_ zT?zIlbiby{BW!Zig_#St%dw-Oq7cW!ZfQx&(prO$&eF9bMU=1>D{nk!fE3t~v|%a0 z$_PSnT=)DB5|UVCx8yODx9<3<`?6BPF5DWhJuu6ZUm#DcE{1?KKag!MuHUKn@VVho zcb{YYZaH==eR+{g*hs`uePv)jQ%)kg!U*L)dD z9z{s~Z9qZF&2+Fq{7pPA^Kt!T3f783VTDa~-l*b#@0=ENp+U6+@-JnyqEDzpnIp9@ znHx&LE=9Z+F+I9+m7=Rw1p-~c!7*}PmG4IJXMfTy6W>0J*TJeNHFVnheTrHcweK$M zJd~7z*`IdF9qX5jo7*5r=~kEclF+K5V=$TYW5uTETQRVvU(2G_k`B3D9|xw4s_k_#CiS#CdkZ-sY7?_BY&kvG-Nh`guI#Bt&}xNfiVn|pb|l%!hO zkpOP&yB{kniB`=8TEq+=0XB(*85J^ZmE)h zywSC=rLr&i?wFy~+|Dso9KxPpgd}GONu_c%2DiWeF^jqa7~oKkjFF;V`Ed=t55%nU z(Omr5OTWX#wU8oVek5wu$*a<}kuz($s>clL*z#=aTL~6IbVbz2+to9shb% zi8Sfb|DnA8^)4Y{4kFWWV7iqa5TFwYLp17~F1QweikbAk{6PcG1X!td_cx!G znP}1)=Rzn36BJ0Y?3w%8F}$?;Z=jWclN%miu!jcma!;B<_ZL0Yr7 zaoGw@dg~RW>#TfN?}(LK1}1!w#xpxg&S}dXV%?yXEA?_?rtipL5Et2G)zHkq_U}_x zO>>eC9TYc~&OJws#Z=xw_YZP>Z__M19mRJ()YD$%P&r=LliaF1FYAEOQay8Ulm5w9 zr6Y$`Qp%W-*Rv@rWZvW)<0<6b3nh>_r@g7lSy89keIrTr>#a$XTMMc_w(~)<(4zjk zX(viNy(xy^;|0a_^Yf{brta0pLD?R7yrr=rAPKicRJwophkJ$KiQ5|LmozfhBv_s1 zh280(pf()MD4>5&yhE9yE)J33tc1D7D5MgTDsQrO$|#5*_Zf z&N&c!X;R(SsB*p{jcyQVn=&IRD8+0+#Y!D}aX6mMwK~T)k6}fa9KnyM6FY=HpD&iu z)McS-H6qDcY!$E!Z47PKMHQD({kt34l@mkJUNJM)l{WFd19`UYVLf{F2u>zj+dh*# zZ(UJMySj^6f!P#XS%8krV|pCmI9MI%^N441Au#tw>y2DI^_P(`d*B@cT2mWg`zW9b zcam@7qc1aZ>}Vkz06y&UNqOYZRa;R^@uT|G>a)H#@U;}%vY@;oap-Gl-F&?2%5Esp z@PR7u;blR1OUTG)y4M5x+}|zo82$XCbR0+V3j@M2z)ilb)@Vjryko#)W3FT@j<3dFVaK&5i+6db78$ zR&qVmN|r$0a?I`K^7wu&M{C^UyL7)*=H^hw1J)h;5=7H*#$ao6oo#m`?_^9QsXyT} z_M9~Ts|v#TJ-jui>SDS#j>BR>4-uD_ZO*D;-!fc0JtK1&)5|<*Aqn+j;p3rBl{5|*^AOCmVEUz}U;P=h75(3u~PsS83DYN;UP~ZRMELg%q zK&pww(mm95%BsgH!SPouGxd0)?DdWMkeU$o!T;MT-4bq$xk=rQf+>Io%?H!73K2&6S}RPSTHR)N5uoTmr;M-bcj_ z!tu>V6P5|~@MMk8mNrFc(Uz%|yf=l;vffmZ{@>n;->vKqbqk#v?k&fj9Mye7QS=ga z{p@G{B8}pELM7L_ZJOG!xTb&auQ+sNE@VoDR@%K=+E!AX1`8Zr1pNUAJ^}r9m}X;} zv2LS=$*rN41-Qi>u04MWt4b3qmD#Rz|9?UyDBK%Y6%roU-l*_0j}u59T3)LCPt_kU!*9~Tq z&eXnVLlm{A_;sO?-UXTLt3TEdVBTVF>H>se7_j3LWCTwymNTLpmOL;xz!O{ z$JP?eCg{xg^bcOLny+`VMb!%g&hHVpyetR*JAxj-{pAz4XGa2)zG0#{2&c^hMMjGx z{}X?umFCjVbiuXp!ApoUQ?=Hti(xonldXrr8V)#QbI!`d* zQaQSkn^y`1<1A=tw3g+1J}m(GR-xZoSn8}Vv2mo0Y6FZrsb+l8nzrqi5J6EF_NS?u|Fs*`=9q?#ZJdi=>AO;QJjCEf2^M;8uS~B8|t+gU}_sr9Oo2 zl+j`qHyjvho6HKgXILNLjCggc)6c}|1&V&930eo^o+M7j6@kjKO&T;Mea?K?LUS5% zX7(KLW475dJD{IC_q>$(U&c|ji6*m5tL%>1XJK2M+;WnHl&5=Hzd{H9Z=RZqhlWq|tD^`(B(+I{JCV7ODg9 zcv)gmK3=jyrO5DD;<5g5KXX)H@shjlz86}aGl~ONMqS)riws>_b_p!tVb!N_bH1{> zH@IVey6(>h6Y+Wc*dN#Mlj&dA1n&EBCbG4}S!xqG>x)%iW|r&|xDlg6)~Cm${YQ&K z6eS44hAx94{;oZ2XpWLSG=Fc>cG^3(p8h>(kpsJK`O*0RnO8QU*px>7almpD(o6{_ zS@9n4<9J+Vz5Qc_;uA4;XQjXOcuBE0#e~Jo-#0pkMn3-W+i?e)F#d5|AWWpua?{+` zyC$k3-7~LpX~U4NW67NM-Ll4>hI_E?0he6o3Wq{NAs2aKEwSneQ~I(!;V!?maa$~7~J2h<>4(; z(xk2wADVVD{&{+QCU^tZ<3CaxmVkEmclC8q%fhuuPOAgjK4%x!tez~;U$AY&k*`au z(<1=x-y{APmi+~M+4&Uwqa)q_O2o^tEFUE-3-oI>hxkO1{aGBp;&?(Q%LgGs=gahx zyT$Rgc28vX40-Et5lUKM;i@<_e|F?Te@5Z0jk-l>qQwPWC;gWUN$00|uzssEGU)f2e;SVM)83~52VfP+{$+nE= zY#l#~{k_rfEz8zosb$|?M89uy6$UR^rOVdkVH~Bl&rHbE)k}K38cc67Ci~J0e5r3t z;m{1fYc~Gf*3g2oD6s)$(e_&mQ=<&^=C``4j2(}n*zYV}RFj#K>?q+YSBR;O5?h@` zOOszfXWz978(4Kio1{3&w@rzZuH>Mgk+y;7@NRQg{oJ?51M|V0fY#5n{VyK>skr^O zf>A6F8&k~ne_uS7x&8u-U_NiLIKEGeohW)~BUn)fG8Ou>wcPVu&b7F>6#6Xd@nYbI zD*RkMcUqfW&E?T=%475De7o-nwvjw0A4zYB5)KiixM|5{3a0@XH5hyJhF~3uY}i(V z^vIjGeiwZ_fPU(9Cf{U=7=7Kjf0vEl&^cw>vTAnDfX+$mEhCjzcvNBF`$ff8-ode3 zTvd{4`wZ>HpO{p1jdaIvzNP&q1G-*u)qMPJV|r@qr5mRmt{=eCd!87+7CLR}>aQp+ z`J=@-8z>Ja(tV1K^T6tb*m$t zokYGtnFjI4+_US z)<%V132%NDtGiyj=_c=q!YZ)H?~44|6f_b&6Qo&4nZjv`8zoHzGvDb@n8f4@lL)N@ zvA)d5HsdpGR-RWuX(5b!QlV5A}x+HRocMvPmRMLZf9>WtUx6O z$~Mfzc%RS1ET3gOHc)0Q_St5BNO8D7{a4ocRH>H!zWK)lCvCAN8}n25zi}Lf)Skh$ zHZ6-!)2=WGY~)NOw`>Y4t(Kfl9QG4NEl9D;l%EW|)bxlZWgW#FYMay@;w_aY1H{RL zX#MyaozZn!gU<}OUZA(#9@fPT!zHeTBOzZERdNR~-6pdLdkvPS=S@7)G5_F(gZ-Sq6HROs8lE+ldv zMG$|cxl<{x$WZP?_y*xNZQDou4WOF%*ep)w>%Ya=+o@ZYVVwASjUoBXL)JMhXj1R# z0bq&USGAJXexc8y?jQNhDd^#Yk=+Pl*FayRq)wk(lRTN!|JO^x)4TQAvi)8K9s9Sp zwm*dE>VW>rON6IC9`^i}IVY;{U4-&2Y|WI>{iV!fMRQPhiuPT2*))g`! z3QXuTyspoCLE8%X1}p-gobcDNKyexIbK8>>fUz~Y4zWYBu{g8hXd?ZofzQl;lg|DB zkVhVVtGeR!Vf*mX_RxW-nB6{L&+}r9PAIkZmTR4KplBWc<80sftslDE%@>%%OGeEL zdXZ;66^35>I|0x8ssx{}`2@2Kk_tKulAsc@GP3nc?K<>b&wz3N34lzPr?YoIIvFs2 zVFK9Dh~*OR$-afZy*^QPVn0q{bBzr~RS+gnJ~I%?vfIs*gc{Gt+hD5Ku88kUsqhLs z8z78?@$|WC*eBM^9|ZH%u>M`-j)(v3yKA9xai^0ug2%tS8;sEiN^4iQ2WWQ}p&XE* zfuR4Ak+WvpIpxjwcv8+$-y%bqJ>lu?-P7k5l82WLg#P`2XfBr#ZYh-Vc2q(%jxB1q z`*8}sPm+i%^J@2*f{zV8>2Ug}x`9;A{>59Mn%>KPIk)|>>6krZq12W3k-9J~JeHk#UYjSp z@Oa=UX7J%t^VHb>W0(&Qk3lU4QVBeU_MKdarbC28e;%g)p`HJ2OnBo!q`MTdvn?W; z?yWl8GU(Cp!y0(p9waM>GdN>lZ z72fl-Q5U8`NPPPTkH+1;{@d?*x9yzJ@>;XJJDwi1m)J%^5&G*!%aelRS@l7UIV;DV z_Oe}2I4{MSfv`<32{_Tg_~wzF6wFzmGX8_vU~BE1(USQau`6Xp`gBs3ZhYqJ4V@d2 zn6`%+Ro8ONc{S);5+@*x6`|LHexea1A>QrVL8zsH^omM*h>z_H%UjCJ^*i?;z5L8a z7UW9EoWENgcZ76e>=p{WZ5NMl$}`YLlw|43-}j!|*r~s1PM;?Wz*c3xWnM(97i!&VcHVq1S6DP@WxZU&9cr)=rK*|3-E@ zK;f;Ow4)8dieAt*4ex7$%vll6gJDfKJ^zqi86xM;^SlvM0=0I zx##$CbPeZrGefCNgR;({zQx`s=9}=ltVEkxz!Qr;Xftc=Rr>>*?#7`AGKVTUpOg>2 zC})MZ$lhyk)M4m3Rn{yT8p{-w@bIw%K{qi6=(<5rnjzuk|4kkcvl%j>D7nq%eOFnv zlw~}G9ceRV@K>m2o5o;6)*$}o7f6ia(5pnl(sROx`g1Zi%aFO&JTZ%t( z@ez+w;5jfD*Hn}r!pyhFo#UPekUBk*we`#1NPxw+_f671VM8DsCDS9slT6v7A6X-q z<|u8zj}%;YH-DiOgG#<3{m9OKt-?U_&dgYt6d87*{{uacO#knjlTh^>s@Lw^(CpMb z!S;Q(vraQhrk_v2I7fBMKAf4k+vBfg3@l)%D}f?C58k%v_#w-5(tC7WjuyWQq)~@6 zWbTV?69W07`5jT-jfo*?WwjXH6IXR8xz)dW9?Fe!U)d^%4O5dOZrZUNeWrroNP2fo zQVU*Hag6?WkOe6IX80^sx1sCQ`n~OwThBX@z%oe{Y{3Aqkk`0ginR72Pzvhb_7Sd} zuHo!sxW)r`X_dX}PIvYQbSHy=h$4W+dV#{3%R*a|aPlF0X+I_>edx!2Ky zl3s6VcE95%_n08z>}`?mUpL}JHcuKS8lUKoj6u~a(&9&`wKzYlP3MG1D<;q}9<3&dZEGhb+X70`?w%E#Bf2!#>znUVX=DL=(a) z@oR;(2Fb0#@0I|cZ}Du1@kQ|afcyQ?#u60ZmGRu!BSoVDRZtw=8}IhMGFXH&kZ4`SJlGh@qxcn3H=0bsXC_?GDK4DwVi?;$(M zTK>wgSDaUerztW;bK0Z0Fo}2uvE`S&~&l=kLcs7>FkjI`RPB#N^X5%f&+? zQImVkD^B5jETaJ-O0EP{mClDE!=sBaT$EtQVORw*)ldR@HEF zXCCWA$d>1ODCL`k{$dyO;LSF_F$8plc7rkxl|{R>8>4b=T}Q&z=nq8`zP51L^6tn7 zAi5CbN5v<2@5X&?-4}{EA|6!Omq~o3(~$mbs_7 zJuIby&WfQXzvEVODkQF|Ctn371Bg3yr_jK>lP`iOHk-=xYNZYB8)r^om8{Ju zsR0uE#qV@EYw*`T^As;VvP@pxD*uI8e>0-r_ebk^^%arb6MHl7lA!{f%KgZrXaey> z;s%~g+ot6!ML%@y1E%1nsw*?*R!)rczmhj{PVCU9?{P1fFOGBRg9YKX^gO8lKmsR& zcwhyFdhpH(*W!{xiO`uoE*+lM?~Av7lbkoYs^=8Dw)wfUZ0q<1-~Y)B9Wbf3w|4;Z zz-;*^4~1rD=7zp$t7>u*KlaszZT@ES7Frm~>0f|opjy&S#8ur@ElVu!GN`i4cu2mS zBGzY<!pv*y{`c-F|6+Xs`-u>FWf=n$8(7===FD%Mq4_p?+r- zja}6|cQyhPes%cLFL=VNHU!h6-6d}@Y6~U@w`y=5Pe`*w#5wI!di2!mh44JUl>U)k zpMJXkdvxH@7L(9v>vHaCrjus;wt$8s$;cdz$e;*sk|g9HHVtp;(Gy1u!9DJ$8{ced zHXM&hiBz1K;+*UYuqFJ^MMFA#YT4tKD)~K57k%Nj(O!Y>;%0=|3qHG1_|vVqm1-a} zg%HHkOruFF94JfG@V5FooZE30cB5arhK+xH>EIH~JrltDv&N+u#wT&eF*5wfp2iJI zFqN+-o{D*j%wcW3{b>Y7tCsCT=5jsmk!^T-~A=h{cQJ&vxg%BUagfF=y|xSj;Qh!GYo7Y+Zf(!MTu(*RTAVss`z18@ z{Hg1AnhbyIp5C^-`{MOIdqabZjvaM6>h%BqL#nICn*90a{{4Ub_FGBLWKwIZMT=w1 zE!onX?O8HCpXx_%?1@EkUh9wkq*T9)7Q6ku_faHn?9k$E4PVp~oMos6yJ_Q|4`1~x zFJ7CXmr#V7+Z)=ADx;A%*VaYfUO}kX-0rh?Ow>1gdkc==!4Iyd?wndVw1NPc7B1=M zmHBfwHWHX0DGr$@=^^W)kym!poxJ_;aW(&{&alrL-QMMER(LJKH{gnFbGu7=A;IWm z;n4BibxiSdrX#wWXHeC0v|p(27mGdirH2!V{dbYctaomyi#l%eMvk4~6)*0}hBUnE zk$CN{hait0WJ)uyVhUb=E zUQl3X=0+P8s!ax=wwK8-ZLC;o%()7C)=%|)+@AQjK?S>DoWQcl z{XU1?@D7^5T)Q4b)lJV#r6fOpc5Oj{#`oycvcW+`^01k(Au0fGG$?(Mf!2LVTNOSn zSOZ?4sis$$e|)b2ds{@fRGHVz;3l24Cnyd;n$X2G9fJ7s zkz54NU9)g_%`Nj=xH3${y}jf%kYe94sVU!&(|(R*7PqvRKB8kj#0czNeyar=h{a3>2*>2X04E_ zKn4J3Wg^4(BTk~)C-=qfzyUBk^2@Tw+=ei#42%&qO5>(8DiuXb$$rD1>qi1V}GhcdxD@>A);?IzY>9%(qD z$(LS%Tb1{u`9tEdfc?EStsArsQJ{2yU%@5rdh_Nr^DU@iey zKkTwWyMIixiRik&}kZcRuSG5P1rFk@0UkK_Hl~}z7Yr4j(@{@`yFwkvK4fY&wx{QD!kMy0L z6kuu9&pjB~BtLsK35k8-^2N}(MRJOyFSI;CxcjL60LQXCbl&?X1oGcc!(%+qTFb4I z%#RaSCBHGaq%SIskqk|vhGV(vLJ#QNLaV5i7GwXd7h)HPF(v9z19tIWe+8&}l(hR1D?lr@rRBHr9c%3g1__HE z!|RkkoL8=}Zqsu33i>u`A(g+vaUTu{#~HQQiHTQ-b9eA44R4HUn9&}NQ}l@o_1sK( zKe2lwQ+9=3Lr=2QVekNrHD8UxhU93oI&BN%omS1Zbwi$PS-`Y1>A z2w%yM)Kj)Z5Xxu)>OR5uxs5rt9|G1yJnsjc+j0(#RCD1M=KU^vp(@}bp|uK^4alS+ ziFTi;Gtw^C%AU}D)xZ?uHe8T*e*$bbv_`f9O;LB`oN~l9)Ys4>usm3fjsjUYnBU>D z4Xa6>qY58HUqT%4}t+@d8pJ_xNYgj=GZYx%V~vmuPCC=)mI^% zVm*G}(uK(5<~3UA#)`+w9nli0I-vx{yC-gJV`TSc+oN7HV=y#;bFB=&M`m&{4aU3h z1Ce-n3k6v{RN`<6VKP1T6XfikrT0wCf9_Ao=)EgK7JVQ0OSROZDEYiY>;)XW!OnU*NzFcil^GnI{vA7KhBFjrcvGpMx+ZhD!Qn1=W zq?a1Z)Bd{ucUi2s3(eVqi?`vg6O{E4(kvf0R=m>^JlNoU3`V7$x@`EMxA%)1^nZgV zT3do<6u0Pku`AE8Caz(QQ1*>mhZa~GKY~11XS=aZaTB6%;fm!HKTX*0!0sK%tcX6M z&%^cOXTM5^RF-S=K%a&h^43Saft}Xmv+PZp@pmnAyl#%4uGHc_Ai*c=&IT{fb|qNn z`#@RYe}Smbu@8=MlQ9vvH)4|3gOA4rcIIY|yp@y;(-_nn}F~qUmoH3**tSCRPL$dB%!#JjobkhY`z1<@i zids~7#98f#CC6`}?b&ajIeBHLZgIwd$17hh>tx-YKT$`7w$D-)J0deQ88u~mw&wo^ ze8p|-S?y27O%7# z#PGu1`;{<5Cd&S+VbC!6x(;};LY-3lFO1-!pSzyrr%I|HzZ`+dzv`t4a-!c1)B!3> zH^M;A5hsh`eWoE~7huKdIC;G6iCY@L^+#74L|;EFIYqV%gLfe{xdsRql0Egi($3U0 zE{1q{c0$-T;BtX?3RQ~#ivrR#vUbWHB}XH6qz9~zFvUfe9UmIsEZeJO=P^e zE0V*c%`)V9Nb@;ti(fB9)Ce~0C+?$|U!~=f^NIO5O4NSbPR3kuIu{{-G;T>eC{Ifa zmLf2BqCaIHCOI}wsL7qII#m4pw?ssW2@^2ZwD14>`nv6R|}8yL?Gd4V89$0MY)CXmgWvJD_ zB}|t270b~2LfIPLk>KLEFh**K#$lb>zUJbZyAUZE9OK-A37Rjsv<5~TOU^+3yJlm= zlDrUO@D~{sY%wZcmr60ekc$xNuphnb&Xlyw9r`lyyX@WfWGm*UP7n0!gu`D9e@DO7 zED5(5kJlG_%U=+j)hZ)T+ zD=xmKqHy3<<5;}tc^T_nAAiWC%<%^)+%lB)G@pU`A&s&a5S&Hv2*KD4iOeQ`9%{2p^GNP4AzWFiEuZEa{F?NwgbAI@IfoWmhyiYX;R$tv z{a*5H!|OTwD;k;|#+5}FT%!9KoGR`?7K~04Hl|r8$#lD?&CYr_+yh}k5j)`ydvfFU zs=H_i2nOq_Hkrjdo*Om={TY%UUb3Y%HqWyB;u7Y~s)E@&YzczF%FXy;+ZGL4CO)*) z`}s#y_uzWvU;}1fTAy4-{cq2>)|7SfyE!3nzpgk{`YVaQvlnzI>o!Ga?#@gGb^c1L zhK?;%k{-o+WcM;ODkksF`RzhdllXy9TJQY+M5dXzjQww>W7w3Cl&_U(A<6Idd2a%V zq?`6Ygq#F9U^Bi3d~se6Vtv=gH23le4krr=GduhX;$bT78+^Tqk=|h7_|=xrn%*%( z<~L(WvN^6$9)ZgeEhNk&EcCEbNwStq12Ym*^D@#W z&P%`h>t|tF>@gRq&;uL>9AkBKE`OTPJ40C{*j2n1>lGVUHb! z^3MyFFMRT4!oYrgPevHHA~#x(LJ-GXI7@l?eTxY=W!}0oj%C6a)CmzG!_SkQkp7sF z$t```VS2h@OF1P{f#Ad58paQ>xE>i6pM)!p8yW z#CI;nfz28`M)#yauS_PS^4O%L5i&0rWCz`Lhakyhd7bblD6d;%3hly>h1bJyt_wa% zNd-rgwu!?vcYi2jZi96e79 z`MS(RsBLFaI**-{pHqf`(;4C^YI;`z`X(csf}2Z#WCZm-<*b^4J@Axa7)53)2!2A7 z%RL_V5DlxkkyC5v9tfxbT_~(y@sZhL3--0I{7&vN5`JD@!pt%9C7%=ai-Lr$@w-8@ zJeW~5_-BIr^NFOT*hlW*Y3T2$!NPMI!*ly zmGaUpKqF2MkEns&m4Fun_t#a?2TaYUwSg%D&}zTo7U~fsQ{V?j-wuwh2_9*J);DSl zkDge0gOlmPA_Pc*5DDgE&9buUf`_tRNpe&`P=A77Xnr-Ey7p~EjHGu=)?iZ@3gp~U z2P1iG+C0t?P5WX8P3lUKUVsFt^CPW|1<1+$QaUcP~Q?&$gwLQIM%Jp`@14NcXK zwsB7X@vbQ;d`u6}t$(Qj1Pwr4_6srwxr)8{_w}~%&EM4<7PiEvs^t~A++4o=eFLch z%P!a|w5xG^qI@EDPX4kSMnhN$Yzi8T0@DqgQK^l->{;76XE2|xtYE-oV(Han@**V8 z>;vvHcS(w)$VN4jxAwUfT%io@a&jn>UAUbn3es8HYi7;$t`Gvi)Rjmp{!NY;pDpl(mfPa1L+tO}Gcba_PF7-)^kg z&s9a-P}JnJ1oJ&%1<+@liL3AF2A`lp&fL}QfD<%tC3-;i-BAVl#?=?xmA4Fr11-n< zKRoTqUTB=Wf{7S~K9Ss6{)3V*6b=9MrTn|J$ai@?3aEt7?KYlM14qcZmfYhRVZHv; zNAAM6Oe(ZG$4~g`kfmrEKFxqbp7__nYbX+Z5zZ}7Viit~$GZuvN6(HHjY=HO zHI2zjE&T#lixkM4uSX-#c=XFMoj*a@T3GEPS5;*Psp7^cvQPG>$7tzm+4!d*{FF?C zE!a8AatYi3IZGm-&P6(3CuPpMWqWY?t~} zW;D6Emo$#HC(KK)mNL>#fwzH+^Xz@Mc}Cc!OXXxA0V!A_eyZSyAvpDPxPN2>WqreD zu9?9SG$+;(dJnlpa!W_nthGu{q_v2N6I3rFkeC_?in1+Tk{T3wi69bL4k<%7zLvx* z@AZzr~%i(msi^dBRE!Q{sIJ#;9vo?~{0&X7gYcAi#8?Py1_&mb|zL z?{(Oh;)hwnp9YJYZhL@KhKvSQB@1>Xj5eRMZ!6{S8VHti$&IE9GiPrt!&}Z^e$$u{ zH6A8v7~^NTf!Ddun;@RWf_$zghf6-e1Q~{n`j70NmrIkD*qMo z6FMV1R;c0}miVosw3qwPm;{|IJuoA?nFI4X?#|IU#|iA#K`n}W;!WcTg2H&(jWA^J$zeF?o=8MJEMq!Jga;*N z(;|FWkunGyd1gk*RzQZPVtaZ1#VmP`lcZza(194l8_8+P)Jm0RDK|Nf9^YZM!!Kw% zD%w$9ZdeG)fk~Cn`cBcUH6DFDHm9DfMvy?!#>y}-@zoS-v`p=XK-Z& z8r26}FNgO#_pFM^ZvS`#wn`_fN-H`ixRGM4LQxNP`fprItt2#oqHeg zlML)XCG7Xt7uAJ!wK&cFTTKa2L)C-InvJyJqmPJc-dn4Bn1op`Hwq?3zC{#NB#OZr zzQOzZ7>Eoe(HA-1CxJB3T8d>VG;ze}Qk^u-{q5efktFA-Lt6Ti6?3oL9c0v?6Es}E zCjxqr7RFZhVmQg0W`ygA2e|WR8CfeZ*CFb>@K<6O{bON4$hJL0Q-_>!#%?kBzxt+r zm>lLdWi&AYInH{P1zmN#_0aVN7QZo@82&}PZR%{9ouE*n7Og-P*%o@9@`0%&wbR`-pcV_N9fv=NuHArM>GmvLf{kw*{i-nzd($1n$9%Un=4A!t z{gn9%mZ@L8UTEKTIxqQp*nx0&)}rcAOy#ci^<$MxtYF)Ua2au znqGAXj~M8NakI~I+1X1gKDz35Tud<3=+Th)JhniSIPm@#`EARa?)%3bK`py+vM9cB z)N@hMkU{o-&Nq&YAEB@$iN+Kb0cuN~XU1s6v1?y*nMOZg!QGH#;u(qQN}P#1b3yKxT;Wam z+j}@*kw5(7Gs^)|X&F0VUtp=!BsX0**`J*C5)y#=`l&(?ZQL{2Dy=DxOHF{);KmGm zmHOoYnWD(==VgK(&L$g@=1|oM%3K<*ipW{l!uk41Rd+$18s-2{nmPML6XVA!a*vY; zr0AAQ4Yt!f8d@Z^?*QBngD<4&_gNvO7Fh*+o%(YIVr~sIs3iF^1VlJQ)8R7 zryjz12g0Kr?;jv&970Ts;_p8LibYGqB2iORx8oVAsBW)s&QTKKvuJgU-(eD>inl7- zNWGB(`z8`?jFGQx%>2SD>YpErq-lBJ|g z#znxT$#)$KreLEtxap2fQ!A`Gg{W@O(-+AjEGiz8U-o{h4K5NMWuPeB!Mofi^EqPW ziJ>b++`?|PlcKO9Zb>)q4T*3AIhWYdEFtKHiu1htrOsZsmF7@(tr zY9k+wJ$!+C)ex)I??ejfPxyeD8-i(1`Rg{0B62HNn6)M0Mf|J~dyA#((v@~i-gLAv zqr^?N+4U^XKxDrQo7v&S(~T=HMbGjg_?<;J2rn3EQ^{X5rD~&S9H3&Jl&&xz#X_Yp zuUpG)RvEFh!91W7E9?(SZfF91Aq#kaujrqkY*C$wK?e5?a?!qGgZz)V-3GoPvv(N= zzSoSQcMkN$_=j)V&k`xgd%_$I^0i&B;cV~jJ)??7Rl@i?bn;!08 z4KsGQ5=J-ySjH%NUMXn|)B###>XJ~MV#SuC7ozZ){gHv{x}m-P#s_A{eww4A=u`%J z&ed#b`xF;O?Ybq_t#)Af3ABHUhx|fVvna$JUyHh0-78y%`ebDIOj@b`D)|m8y-4D| zs5tpp)#9-GEdjFgwzgN>=~3Grk-@C!=hyog&ze}sLZ5d)1XV}#6RKJt?a*cUA=as9 zKF(lHV)k}uUY$6RTKnW@#Gh~8ZQ>~3USPtVK#KhE7eUx+<;|D+GbspD^rPJwtn3sB z@2}e9L~r&)M)0Rq{U8mrVcUSBAY64e9Ajr>XnblX1ZzqwFEAreu(P983UB#9a{9}4 zdKP;lFCgUut)V%Ve(ru%P14F~{LPn-ziuSxbvH~TRJ&t|6NK8<=H2x+QK25o@t(b{ z+t=!GN;;z`B6BYW=+({OZ$F{Uobg|aTnAW>4Wl+4s2!x)s~2;Zw|X|U{0g+ccF$(} z$jsT)t4d;sn6$TkGsQ~f!U{CLjuXvhPG1j@Hf%3VrqnvJr9>N7B-`*AwLh3O!PoYNG5vTF0WfGBc~>TnmMR- z`5Fd%?;Vy}Rl7G>&1r-8X}-MkgIVYJf$a;%{DIc>-QTb8uD*L#U^7abL3G>2Ci&Ii zQIh+L98kh#Yv+*09~E1sKr0#1Hh`_a3bi8`1jXjvJ(GW66;;Fk42L*qgjzrKy+5Jf zmk6!FcYKMqT&(f6zX#P}=UX0js0gtRSzd#TFALB>eeKJ6b@L*_R9QzZN#9mp=)x7$ zKG8Q{xC7PBYc~42PB%15Ap<3Up`B8lJZkz*{l+=q1_ukD?~VR$`XEZayZOxc8e#gg$%=$8uu0EghZ()% zkpt{ow4N`m3mY);DRH^&JD{KpeMWwoC9t`?pvAX)59}*%%KvQLag}y_(4yX`VK`IG zFzBBATAKIpxG~mP#v6`hQ)gp`76%lowA^JMon$E2hTYXNt9OcqBXgdhE^vB8Fk>v^ zOFMiBHD1lX0vRuj;h-^OoyfvcWJC6O53(_pV~!mn4^W6Mf8UOxN*$P#t_KMPmX_BJ zm|Ldj}*nK9RH zJd{7F#sv-{YW>%4DJW6an*J*8V-2WHlNWBtN1BZ@&ni=6=0l`3qI&^%149mB56n3{ zp6gu^cGg^H7i^f=el_W8CucI4;BhD-8*hA-`PSJSjK6J9Z9n+etIToIHNNx0|NPv(zYj|4?c@XfqlhL}=kBy8_*)Bp zu{fRns_8DDu1)Prw{|aLOT%<0QXOd1l^b601p18X0p!u!|I+47q|9CAYdNwhYbsrDR8Na%_tH0{PXF_IJqy*jR)Lyo! zef{I&S?*FXmz?29p0bNtTvBd8!|v!&_K>`h?G;*cB78YFH^VVv+U_~`zRO+i#kPTG z$q`s1UF~W&+)INa?I(#p>6+HJ+s(ZxaQSmC-Ton?99iTGp)|E+YzlBB?wu>-%s#pl zN)d%Wszw^f89uZR3$ZtH4BbR_0T{Q2eOl;S&>GI<)qP9X&!@|4z2R+Z<`=F%$?3X@ ziv@frQ+#JdW%IUz9%to~)ULUjh9~NsiWg?>)lDZ;VnJ)vgEgU?w{B((0Zt!?`#sk4dD z+fA@32mcMGnDjlaVHfCZE=%gAAZ%A@%)r98EM~PBP}?RXwKk{HJ>F@>x(i?J{p@B0 z@aQ(WUte0f0w3GMT`iaYK!Bs^A8DIbEIJb!DdBG-es*hkVj8EAb?ex^-bEd)hN;DD&@)w0%9k;UJgank%3rec?; z@(y%|0FZ>^(_)i}rd$#FZ*a^oBMogkF$Cw%gBba}5a8X`8v*CKi>~!hDKcuH%CAoV!du zz|`&9G$X90n96N(k^4{u4)waap=B+USt3LDc5Mx@Yfl&WD%?^)b#Kv}3{UqA)lneH zrnj&7eAP;s(X}CWrK2Z-QRqg+T3gDdw<~1W2_kH6LvCq)f7RpOh#N_;z&j$4_9<-Z zmjA2-luz!))3m`iw059nusv>yr@rCm&?5O;fBeZgXM?$%%|o{I^{(!+HYeG{affMH zMnpq+!2o{uV&ls&mTLS#k0$7jT$jGubYTaod&JzWTvA`t@w|Pmsu&l%v5`2W4Ztd# z9sFz*sMok7${#Yd!yjM5ry0}$U57wzb&%=o>;PNa?%}5y%9^?bHM-d+i*GdzV)sR* zmZx8li`;Qo1z^058V>p%S1$(MT5450YpCg{*{3e|EZE;(EzC#AH?uomaK^UB*#+e_ z0Aa@AAdR8gqr|O4%fs?OOa#KEppH~+1Z7I1hPGL6-!pUuIEZUW?2;o{oDHa>0`8;K zg)NZ_w|m^SgF~(+*V*8Ut`|AM(cM*eEXgxILoM8rvYq{c^Q*j!V-#Fh3MMRjqkddq z-noKtL)4$NOUrBHLWRd0QE|mh-LYYafnh48!7(!9z$a*<1T!bIYu&}3oeGJBoqc<< z*Nq;3VzA_c_I`j92QY}AQ8xvGx4UlwbcV{LY9ss4vPfW03ifjV%z%1A#wO!XqeFY7gKNfYlRZh!4EuE|oTQn|jr`D{w(kU*LM&SQ8%CgZs>&B~ z;rd{)w{oop0EIuoQ^F+I&xdo@ffB-$0K;#4bz>y^r62%efv3IX1dMCf{%{1fCpcGg zA(*FgDVcakq!~Vr;YE5(;M`jd0j1NV+1DdF)eAF}E~$~v(+p62W5{22*FH(jU`db7 zGfcodqc7BR_^~WlIJKX4NwK-}S%!k*v_A(to|M+`RmpoKFOrpIut|Lo{;nP;d4Y^p z@Cx@gs2>|p6!TsPf=hK}`+5X-4by2(ZfN^+?ok7ZvUgqkT|I871B%*@yk;1G93meK zHKcH}-+BmNEnGyj#b4R~EQ5vqc9ToWOkoQE4uN3r>T)d=>MRzlXxDRRH}I}_zP%D^ z|F8VXkF~;b zi@Jv!?TeqWJx&9OY}6j^q=Cwb?Havbj`-#x`UFJs{}s%X`m`s|nnw>@s|GMkTzUfkR{zj5O?B;Y(~`F0=z=;U1v zq4Swa;tn_2_&?}3kVX7c8FLqR2KQ%ax$MD6_vyE`H|t3@d-=H7Vnbdm?PwCs>?}o= zAue*sR#52 zz&7PVo^ONH8Km1RsXMzJ{8^;THi)~fvGKAmc6ClDrm&GPaTWR0=O{&FhG%D3zu#rd zEXl#lsmH$`1B#L;!!eUdMV>cK6}y@m!E9FhM03*eo2D7E7}oSgnL7q6%R!ZSa_#LHSlo@N9Tue zT7kgrvIVYHDa;v3^aa8KniLVs}&x<7!?EeK8e>cxhY5--e2n=ai{_ zO35)gOWD|UM;T}2Yz#a%S%G@19NXdP*Cjegyxti_5w-d#ow0ckG`praW) z0FqV`8yzzLsu|(IjE}C{d218&zn|d8@8I`9aXK~=W!n0?$)#I=cize{+!-D{c}vS` z=dGSs%uPB+&p55CdL)p^%569y^XJ_df`rS8Tn55I5`(7*VtgeLHF(K|5c~bqq#J$9 zetWc0$pnaHwZGF!8pkzJM*8HCqe@*27-qALBJq!+p~$9-SCw&R#e) zF%!!B_&P$XW0BdUUeGbe`cQAPF}cELC~rofy!x2+KFEXYyC{6BOL}CHlc>FY0I81p zv62T+Vqw)~N`q+*-Augs2yrh1o#|MhT#NdAV4Cy}_)afqMk$R^v3(OXzZnqSZW#67 zMzm|zXAd7F3#%jEEdkIa-O*_PFa&asI(ibq7`7kn3xW6z=x=NI6n2&I0x|P-H-QT0 zg-de4fxUUY50hT|b^%+IhsXeWM8+MVI>E5JdtqNFf8=Kfh`|WpoVqb`8FH=SS%%E} zl46@GFgqcAWObMWgC9t5K;KKM){hQGs1R&L98-x-sP)QUb+q@tfHzmtEjM=U5zc*X zFX@nD_*sh|zZ%6Z7&}T{`>jpm>|}uubR{VmmIBq3+Ek|0kKW?%2DloWb95zs2*Fs= zQwu3@CZ+Z!6IbFs2`>~;ELiTKBjHS3C7fMZCyd73uFa3(X zOURWO>stT^&2x*D^6iUETo@VPPYz)7rF@sh(SkTt%4rx;zlS3|q3g zzPI{{Fga;z89wN?9kn4n=M}*NhNE>#w2ngnvdTIyXhFTW>DD&tJMZdB+4^2V-Nncb zxYm^H$e?5k>S#K0{HOL%Zqj|+s*PW_{au-EHvF_}GW9uf(YbXCaI1hYOQb~yDVb)& zgD{s9hL1-5Y;*fWGYZ;SD*+!{2YSS$s;2%1u4K z<4`?0qnBzfq&%iz!%O;t%gp@scY8rT+9%bvsyX4#%p<bq9P?QL$S8U>}T2c)J^ zs4D#5q6^Y*J@Veqwiy#;J{)=oMA=9bQnEtS* zIraNT6?vVMfJa!qeR#E78qT>&TN@jbZ^%kmta4qQM%&Nl09-#hSWD$RI;C-S~{oOsz)zh>`35;{eSu~-ns31QT! zV{)J0Ya_;g=uYqqmv;g0!}76Ya|y)iqx~Z0pJ1AOA0wlyYDBqBX8(o*lL4VtSe&Zt+1{PZoddX)mEm zvT@ook`i2`=5k(Fx`LDw0}#Xk`?&|i%*j3pRWZ^dh#{JCC?GKRXv8sE?LeD!N>c8U zU60|5B(E7TxjkP{)^(FKWHPzD{Dx_|682-*B=`MBaPC4K=#DgV>}A@{dFJJ1ViZ2r zc~q|3hqeq8X# zb}ayB2JCl&p4Z0}J2vdcusHs>K+r+mT4I%AubQd@En^p@w_Y_q;BWSmKk!La!FU7o zQI}`qZkY5t7H0I?2lesm{`KL*<7Oq-&Ba_N$c+6#pP}q{t)yuAjXcDSD^w?g{5b!n z?$hUJtsqyEr@nU8QdM9^0o?N|en0l2DuSr3IR|OgoM)OYhnz#U9D0ksIf}@s-@4th zK5Jn$djNiH(($X-8-mN6TQOd(3cru#QhgHBy}uuQ54w#W`RW*FCp%ktu0L{oOI+{!r zQo7SoH}MC!;22&Tp%taPvspTQ2r8sKGxo-tv2=U_!NH<0e%!WRs_>J-6e&|mXRGTc zs~;bX{5ve~bGzwO;vjyXLzq!S&rlCGZaP6p8K>)z7edL}a$_KAU4St5;r72OCU z4tI1&cTwpgvk&o#ZEYLac3o~(hzy?CyVAAOgl`HE>v`NQHF^o&k8_R^MRyf0QXe*A z<NpY*M>{E5{o%VFvKQ&Y>Q{SXy&zYohka0LSzWFnsg8$0iI5wFJ`u^GG@+YVN0#j zb|w2O6SasiY@Re}3UftyQ0Q!TgcJ6mNezeyj0WbYZizh2>qGoOIxS=6(%x&*{XE2^ z=6(H@DVkym@70yhj57Q`@Eoz=AfDiShW2O{k)~=`L=MslS|$r{u5!#`UA6Mg@+#V_ z?JbtqJ9Xdzaawh|){ac&-{6Z9>f70dJHg4#JB^w0L+2pP`(-T5z)~)+58o<3E#nQe z|CE}G^BoM?emd8mRv$_HRfs~^3LWFGDB}-{zq~gFm8wcsR{iSBmIRjGBrMUKxK*Cl z?+~RtS;50%2jlbw*X~&_QbC|&d292TLXvRW9`*Ag=3->G9GrFkVB2oKR{jpCN8n`~ zFz%84eUQ_(yqE{&>+oZyB*py=ENtV0Sq1^DBHbiAkYAu>(9h0{E7o?Eevjd9R~tVW zbl@)6zW?J7dQp4^aLq6s<^CgdjB(9!kUCW87BR9eS}d4GWkSV(5Ry(S>)U0Owx)dS zmg+V>gWF)7Ankrr!0Woi*+p$z@5;pPit*bRsHbB$mYnEnfB=s|M1x1E@kcUkbL6F< zJOyT&ZLKrI`@@PBuNT8EcP(d?aKQi4()1}lMceL#()*L*nCkucZ}axzQ4=GGFge3l zd9TUss(hC`Fu)%9Ga$~|o1@cg^V}F$s^Co8N$`5rM^A~3+Vmcs7njPaubgGQGoEJ( z`#&RYt7uJ;mhFk)1SGSH0o0;j@>z1S0#U-*0%5i{RlVT7s^cOc{|2goFzEZxN6cw| zZFLogX5mg{P0sSW7obDZKuXo&BYMDtAOnz~V#G5F=HSwwX7sN7aQLti2CG3nh>3W0 zFE#ET+SCrqj?t_hU6!5PKXVf z_ptlyICVTShU~!;zZxwQMV7q`GQgq-@tcA|xxdqXkPj-?GUnp^^G)hUU8KKRrKwO6 zDA%qk{hpAUm@~B0S(8+j>XVZBgERt9JXDqGuoFlEA17|$s_wP%Qj^N$?|0ft!9nE? z#++?%zCIZM5&<=#G1v$(8mK9mI(5nyKK(N7>H@S8mG%Mx%r#9DoScMx}b zJ@_kLcI2Q^>wkCK4dz}n_RgXt=qJIeO!erHV)%`{Ak;(^P+|Q04rj?CAv5C|EF7ahmyULoq_Og zS{$g1zg@UfK)4UD@oL%ZdhN#P>V&3#5w3XF_BuG<&`lVIFVCm6doG~SgYQ;G95>XC zKhARgn6RSP>`=~KHDX`p=Dz`x>rCW@jjZL@pszzCx&O z5A_NA$MkmV$^R@%m;zpd)g&DTg?Bur3m>O_vD!X zAk^BvR(vg8W8=t4n%S9gN05}XoeRloe6(PgxV$pNDb?d%fk!4#qYmSQD3Aye{#w3`T*B?6|I zz!}0xjS-K~c{7rV7QuO4k%Uq5NfM?>45bm;(=7%iYe_E!Emj=7PM^S;Xw2}&YaJuc zam?g(Q#>cUmMh0uJDL3V3zD@#K4Q||jftaOK@fjpd{JOKE#cqn-4=#v+SUs5PIsl> zKq$?Kv3$K$p$NqjZDgZ`Y1kneEr71Jg4N{K;R%qXq_85P#-iDH8BO3Szq~SPel11J zv6W(yXCzOAj%75)d4;0?tkQ32D$xNquMQIlQno0yUQJt$wi&bWh$%&Q)JlULA(;AZ zNB_cRZ1A!}80HF16I(OY$6?gRbpDWtcc7}_hZ}cH9=f@v8hpEEzq9H$4xl{c3qJ|+e)#gcKgzzTXu`FG+QdW=zOsL+R|J2;h;lS=dU zJM*-wj20%1J4bT{YecqOpy|H@HJkJpcuLdMz)b@AfE^Qu{@zucK<0467ah@EoHEU| zOK}gnvJ`O*3I|_KRrq3KR11y&R__~U8R{xpRojV+b1$}D0QoR=+o5IO@+Cdw+xvk#SWmbOess@8Ywf)RePKSaz)$$Wa6+oP^ z1~soo0jKEzU7x*Q9}i-1G{F-jQC*$`^NU)+o@}S&TcG`bCfwPaRivEQ8(gs!Z zXDCb>Xje0~Qha&MSbZ7YqKX?6K-?I||pzt)@Ee-`r?vnULWb(9MpAnbIL|&S zd?z7Hksz1@G`-b`@dT7Arkw6&mEki%Xox(B5XZb0*Z}|s;3%5(4}h5RGPV$(q6!<| zT;hOma&zXw9Wo^+5%HJmA5A~NfWEzI7`ou|m8n2L|Es|z&dJVz8>&Dys*|O_lHbtM z;O^<;&-q^-)99M?CU-HU9d!EEme>;tI!aC;N)_uth>h=D@%I)Rybmi^adzM7sh+me zb&>WAgy!0v47vIs-WVA00#%{S$5hcQfLoQ}Nnnf#xUAWR8!pMM+t$)9%pBD_W}*7( zZ@Gd#d>063a|YY>c2Ds?;N&1yc(l;-@Q`Dw%WV)QtAQ{XpSlbXw?DiuA@`~rBWnK} zMsw9LPJ?085`sNHOtFXq`Qan_l+k#@ZeaI%@LT3ZW zsi2=`1AkRO4lCk=`ohy*Y0NyLOmuA4B6EHD*iYaREUNN6n7{)7G}(dC3{dm7Yt!#u z*esVEqzQe$s%YWSuH~E(W!6Om1D>;VXIgL{8v-ob^ZfHMAcDOI@)c4M7?c>2zVr;x z!KMb1Y+uL#F#2*QkZ1QnB$vWSI}3)8vH|SvK~JrY(5j%}kt`9`OBqoY5dZ<(MC4Re z8%9>)6+$5EeuPKRoB@9c{Mlk!c!wt@R(0=)^HTT?GBlw*^ec*L3t6*%BeasVYd z^f!5-Ui<12P(*+cO|kRn>*J6dQBdv?5L<{KS<9Lr_7(5Gi1lMLO4@kT z498@^`?a!rKyu!WGN02eL9Wi1@@f(9zFU06odX2Mgqbe~SuD_DbEbRJ9a-udmDRH? zlaCfuu^x8Tn-NeX(wzCC#R9Gyz3(Nbj7Q2#`S1gM%RPj>)s-A?sr{Fj@M^^$h%X(I zJngK|AX+e&pSuEod4+5pp9&QxMRg~-RfH9au+lOhHCKDgK;{U$ghSX#RaOmbOmQa_-U>2R2|JrRGbo7>`04*25n&2UxL zHy~o~?tj@U`7Da&oR`+-M3F(QC{;cETIL?VAOlezG}h7Le`&Dl6G_tHO4?fXZ_CD? z0ygK+b}@o8{#!Ps@EbgF8vIkeDxIkL|I&#ed1iS-yIMWYlaBR5evHwO0k)MY?ud_2 zj^eYhiz>sDu?@Yn-6Al%43Yp+>MuMJls+?o!vCt?FyNfJJrz(;oRo)WP~;2BwZ!=+p=S2CKQ6^J_AdLW9fIfScM)Zu}^2iRTefw&_5&fFJ;opkT)1En~4C~URs|Cw*|`vZ5%@v8TxAR zPbmI^rveoxK$E98(6Xf4!C>zZq5BjE$QE;gfDuqAHjZ?cT(RaU*4E@YeyWSIqwh6l ziX`W9nh*ZPynF;CvBb)5)dJYP`0*9xXZC(QhCC(Di~wYV_>hSvgRELHbr|smR@gOj zT)C`TRC%9ww>3H0(eIoV>$e>A3)1H09b*Isu}=4Lt~AV+Jt`prz?W?gP74*`{EXPW}EJY5vR7UP?Z-*ZPOWbpqFs`PxFthC& zbu(b)Hh4`WOkpZ3`MahC)qdGcs?TcmhJc8Yxv*ZMo11jP9&Ze)@VdO%pe+=4QX06o zn83zm)v@CE-0hKk8?!`lDr$Z$6b%{_W7|Qw_iBO5kB7l>!|mwheFvN9CSdO;c&IYN zFO9DwK%oVwHi6Uj4E5hmH}0n0@y8`EPv%3^R7o%~Ko0w>=kj}dhNC1~Sa2xriO97= zkPwp);kWo72E%X?eaj>>i@Jnw=F>$vptrh}cStgPMVfa=l3nk?PxA99 zDAfg^4Pcg)ZmU{j0ZVU+BMm`wUOmr$z|HaUt2<2ABbn71RdU(9GZU0VHc~Se(?C-{ z_Z?8kP}xK|M>moeHJ^1o;BmhWAE{=TwJfdGUhL$e$k#yvZubG0!h6oM}Y&vyP zVNMc4PKo@53n&g2d$`8{CD6P> z{wwG_9ufhk1l=SMiF`2lq}Vags^5*yh>QG9bfB7vbDsgwOmDCOaa!tkQs2Wy3+H<$ zAIO$%A7Qp$yN6l4^{q+QV+yaMPk5@D=WYMZb7n|$6HH51DN5B5t&wi6yam)`W?Uxh z2NpR;>FRcL{k5a#=6R~U^2p@N0J6zcA8q?1kvBNvGhhAq>zrpBw0`s=ePpWL4q8bt z8Ggn`UBa!8w97nuyO_~!F+00;PZ(5GG1>u*=yX*63E4`a7Q>B? z=`8v|sHmsX6Om#WW3@*sMMuXS>*de@`XTZ`hj9|)doC_(E*Dvy)iQ`mf zK}{zBIF;5O0cXJztZrpG=v@AKfOmQv_=0X?ffCPtdAF~H3rV%*?4Gu6+r>VPxPuAY zUJBXI)c`n<-W5yHX7Kn+$1aWN(Lw7dm<|pRjr8DR3W^U?D+abI?Sh+kMPP;7ZoFLut^o!r?gV z>nZxyQGh7lS02P_3>h{(e?JBouIy(ICl~f{zTUttAx_Xy{iJ=|`G54@gR5%F;t#V* zKT=mug6+ereJSJ*>~_>F?$>f|brVrlj6NVACYj3?{eKQp1%u$&P=)$ursGmH!NG?( z(6ORlFckUWkV)RS3E5I~;WHxTSl-wwUy_RvaPKQ%Jdo|IPIxPO{93zIo_7(+cw9m# zz;wP5J4X4A`H1sCW_}HGfHzt+-!fTu5z$qz4~s*n;as)p&SWgfjCb#u3t^!N{a*C5 zZSD$e+VeQko~yxqz<+tK(+$A#gcGL_xL1Z*@LMgt54=yd3ix3uwa zC--MhJ>0WlLcj*pIpl9KDq&xlUHgapzV@EWpO(N$Bn^BVcUxa2YIF+utMISfm2-97 zKx)Oaj55Bi9qb+5IUR{6c|ntAm#73C@W{0141io9fbiZCxmWh$ndId+=>}A_;-1tJ zJTOnphXyyQgiWl0pu-5#On5W*y@>U3X)|hvBnqzr+ZjTbmw)6R#9!T*oRdD3txuzG zfc7}G%Z5qg8-Q37$(tsR804*LwD^br6MMGFd0tT2lfz%b-ZF5fTiG@ybnRUl+nr>%$CO zGMe0-o6`yww(&P%LMresbdfn1D~x4RQg{8D$JR})pdMpPn1Y-lSvtMC3rxVYMl2pB zKNmLm2FFUHOP!>n6W?nw=aJVz)eUX9V;nl)-Y6xc9?K%SUfG z2=++fUe$ML)b#9y9Y9lQFFRbH6EwE1_}5F(jfLqAmA9bYbZtHHA179$qG5EoT}h9r z;?u!@+CIJf5MzY;wQn#XfsTr%evoeg@rycu8Z2#4m0zU;k$NVfS&8TM_l_AiQcSr} z^!!i)SVI4%v(jV8s1?dClg;HoqzRTjlmI%N8*uu219T~V&UBP`*p2#j_ySi81y#l9 z#!R}r{@|A8b*9Y85AYbjJN1_ATWSZlv~2|Nx5dPR=h}7vQBu@wAOy7`@)wg`=`q9XTgdciuAUPdqF^+Oa(9T9zZoz<+l9FuGSZf z{iGe56|bRxGC;dDRQMr&>cE8SMy^*&QgxX%4S=*`P|^98@;^&sP>BKh)7(v2c{2AIY`rCV9_1_E|^~oX6JG*1nDp%F1`M9iMEJRXfcUygfczN&%~2 zrs|(7WRv(G`8Gn$Az6%q%tmH@n)9Gx4a9lh-AEz~hvI7y9I= z#t>+69oH2MN}J}m(hV)3=5d)>!q0kO1#nM<+GEPFKd_rf+ed91UkE;Gkio1bc~j)? zJJSNUV?ad<6le3rI^P=YZs`9{!3i{qx1tKE5>u;`3xFkU%#s$sz9rt%Sfn+TUrVWp zZ#HpFz!yAAmDFvCWtuk23Y&8NMbk~Flxf+*0hUWR*O5JygnEFJ!X3D5dRNj@^+vts zO-@YH&ZJSBQGWvwcGr)|gVy(eLD~(xM0a6TDkwS2Hm4W5q{Ylh^+)3t8_SyCG}%6^ zH3!bs8;VBoHy8H%aX~FM6>z}1w>AF)gaIgBL906AQ;z*dC2Kt}&!jVwuIh1f>HE!l zoH%Za*l*2A_tGQ=Z;Du^&h(K;6I`UUuAD81G-P}4vqZpWl|_tx(eftk={ z!u96f2*X067U0E#6DTs+I9lZ9;KK6mgI{gd)@pqB| zI@DF~TV{4h59-KjPcjC^!5U2g9*Q0nsc0&M$P#+4#LmCk5LY5eiH zbC;Ja?|1?Y6;SF88)((ay!JO{BaKk5dyL3SL_$l&S3F!k{-FsV-h2O#rEdXey8Zu8 zLgiFR$YD=UPf5xlne+BYr4lOjP|Qe0IUnY{EvHlt>r6R}Bq@zb<~)b866Tc2X|x!I znc2o>+yD0c{kyKNu8Vyp``n-Ve!pMu*MT*It3Ia2Vn``m-RP1{1|-rHD3N8j_fmkD z3P4Y^gtP*KR##lQLeXB>>8FC*NWi8jd0T6PTg%Y;q#N6Qcr;ev4?`r z67#1+S19;(SY2DwPlZA9ebBN#^S1#F+IGHf^}jiabh#p--K*X-K4|Z37Wh4R58F4$ z_O$(|ruZ=~)dcrlt54%kuby+)i@7$!@eV_LIkG-qA8gb1L|1o!gtOoe_od^TY;=sD zJcW6K(dz%w=GjwwuZe*v-r_^h<$HtKoK-XqngxR>GdL3xtmo3TJeafn!tiko9O%sa zitxjh{7Ya%dkJ*>BEVA=T2Pi&4`1B^K&d6J3_EcRG}TU`+k%x;_UWgT_uFsrJiHFDsk4wT-E@ zqiOHmnvMWH3z(WSWR6f$_jGtkrGcRfs9MTB4~SaSTg|~C1&!-^&oB(@@{Jpld>s?X z)zvLbkHtB>c2cKf?L^JpM)hA>*Z)3z_4 zjA36_1{ZaH+D)1<5(^$*Ytvp}CpJ1iqnoL>{0)L^*G(X!vL#v?!5O;NMOh4-a*}$> zfi|OidZ2L%JW@u-F$L~rNWMG&(-$#!(eX2h`C@KGwE=&4@~=C_i9~hsKT|*#qV;=| zGO(Dm=P}cx@COgiuG~IczhwO#^Q`g*xH12G4uLGG1*Wu474x}^PP+=`OD#v&S!bUf z^&bK`oRvMi@XWf;@a~0V-+D=1)^2rpoGYIEWsEe539$o z?CZeoOK<@8WFRC&KNfbrt-%(J|A`t@@H6{rLFl;w(hs8x=ky$K219;m938zp%`j0h zA(?~+|V?mMTK6<3%`n^V4+te{gj7LK~31#m01_K?@%>*p)#fzZx`qk@5x!j zES!MfsQ~@%|HV6k?*>)$@IUC3Q{MJ5Hs{!gmQiCIz1|j_6yS?A8hWUgv|_@WWNtXa(Vsc$HhaV(rt^(5jAWMU(5o@!z^2)9=C+;9|T4bup% zUH*@+Ys>gN|Jb=C|LR7D=7@$gzt;c&jUVPeRu2E(3yRWGiOqZtINy}IPclYB<8hJN zPYo+=_Sww>@ym1e+0{fM@*h!PK;41)*TBwc$ee4~)0hs}2`YgRV;W?JANAqt)nS(- z#h6TeU<2kH-FC`FKsL(M>loju7TsiT{adDddUc0}L{?fIJbNDm#l4Mpiud(L|A zHW~DDsLLte4oD2+)HCRv?f`fR3c(SPx}t#P*ql)98SdxedNTOY@8$*W;iG}wp12bP z;*PNi2$gB_4W-&r;#m%IrnqmZzGD5!_5)rT`e4!Nu+u#WNQmHkIsd?<`u;^nw1~+Z zy+^GvnPnoAnR@s1BI?Enm7I{g8Ey&HJx@>=(+G)ajMdIB7c^Fk$5ekQX>^{B`9(!R zn4c+{Ld&r~s9%jihNXsV!R3=e%lcwZr0+a)_AWC#J*YfKhEVjktmK{t8T6hi;2hR4 z2u*TdMn6rDN&*4RN*qAt1lHJc*1ua_f2@jm|J%cHTi*qMds9^IKU|9(r)U zU0wZ$3Yn~PP2h9 z&$v+DN~cYs=MoDu)7`%ATUgwVi#|c2I>2+J=<6F8NdI&8qpJAkZvv)dwPU8&5#_^z zN@N~p_AzDn)i>DL&=gGtOj1PkSKw=#bh)x8l#yQ2I1&S9#qLI32ZlXJ=Wi|4U0s)y zeKY&cb_Ya!7uq3m-{2QL0)SXoT+&TEb$+HbIP=ed1lEz&%RnXHtSm z(Dx|nv~#|71$^W~XG`%DfT!fL`d?{HD-eskc0elw6=}{8@pPw-v-+F^dOcE1cFacw zt{*x z^trarp*iHXK+l|A9Hr1ix%e-8^Nkj5BLDe*n@+i9O+*3yqD@)ng6Rc119S88p(01T zK-+WizRd!DDnuxT*Op(&6&9&h+*Mm9J^&Zo1lQ2D%ob z6+h!0(}W*Ce@H5hdh%@h6csf*D18018h%9^(mf--8B6<=(kkkTT!vr09l5<<_-^t5 zbBpTojYK^(vT5g8eKwdA=nyM~7c- zHNIEpJPV&s7Y72Kokn@~6f969Jp@~K0F*$jEJOWXA?^ABH9(U_&D7>iy*H8M#z6h( z#fVL#Lv{0yyXvl56FHTk!@f4kqxL!0I6KABA^1lpp}Fa=^6j&YgIY6KJ1Zj3GANCA zD>V#)(#kQQri_t{$Y=#FEE5_�tUjF!!9bgDLSk3vx4GH{4k_txoY-pu_BPOz6n- zWPT5lWw#nerp9o^9Q%RLHhIsUlfqz&MmufFjc>JT)nDTTsPp`~iKw|`8Pr&&U7R6v z*+&DEa>ii2L(K$nHwS^nbnz}!V5Pcmfp~=uD~+Y~TTDt1RhAOx&H9I@t*=j|^*Av; z_k{XZx`4Ifut0pv_u+;?Tp)KUl49|FfARNyzIm>ke;9FotiUlRwgEDG_v4t{mU zm@E2+$l53ScmvJ%`A3{EOLlP=J^SCBgIrUs--k(`9s+VyKq+w9j^;eS(Eb5NV6#KX0ej}j%pWwh~2N=(rd{N@sU~e8^ zwRHU6Nr`TD=b7GvM#gI!+Do7yZN`ZMTFShsxG`K32%wll`9A*5)-QCAD?xXoNj@}g zE%EsWyj6sMx$wnx4^!uwWXsIXd-T%_S?(=cfvX>JKC)j9ekFFAb!3JG<(Be8Xzmb! z@!eSXAEQb4eqhvjAgVj(;p!St@Xw<2-#rOjZfuQ3G7K2(dWt<3+a8FP9x(Z;QTx8U z6T=i`j`HA57|VHy3zb<#ID|{eCwAs@I29NoFNiT2iI_pXR*H3=yT@ts+gIMAJ*Ul~J_ALh;9XO&E+>-juHV&xC8Xf0$rYVXBdImX;Xv;U zSXVZ)qFioc=j8=5SYrfsvYJA;pzCPe$-yo?P7KEka?TI2< zch^k`zSfh7nh|~A)&~l58Cu|>&dVA`6;tnZ`MBIbzL*{uo3Z>UyvwRK;97oWhg;tX znn#Uj%wH1LcZUWXv^KouMm`KIfoUtdiEsFOh|j}lQ!)nZXF{D}Yp@z^l)``;_>jWQ zV3q?6RCnN*b|D9w1S@kJzj-bMd$d`Bo+oe_)?rDga<{0Y#t1v$#}KI*gdgPS^ZlqH zko{`3OK%vslWWQWQT9k`LxR7xSE6u#NzYVfq-AE+J9BW+a^vD*0ka-jn`u9UI7hv% z;x(3S?DXr#X^|2zvMlC>!xVMbJlFuVzV^42l+uL@HYccWHrgT$q&b>6^#Zy6nc_}GM=y}Fg=#xT3upjpYpe=>9u7Ew=RF>TAG z(YYu*>laIw{uxM6BQ+xYgUdA=oD%1JHs=vEbFyN!XJV#N*Ice+8Vp8tID;Zgz z?xs7Dbhll#0mD!Qd@aTbnrOVXKDaEbM*V~eKD}SN{J{6DvL!wiv(7StQ+Qu*AsbwmCZ_V`3%Z8%ZX-l_a4Q zIX>iBI^Bx|39K7HvvQCgagOR;#I)?`YORjUt5wP|nt$B3^SUNPB8DZO2Q@dCv^xS? z3^$5Od1A9xhk?uj{7Y}Ry3^|FNS0N3XELfIB-3G;!{2;5_qQxSH|@1JQQIuUK`&FG zK^%&D@I5Dv4AO9Ks!_t}dh&i?_Kz9RY*KGy>*Rp?3dW7c<)#wMI@dG;J~%rGi%D5< z2Fv8dT0^sr;Tw_(Q?6Kz)i-MRQAvsqJ2Z^8?5AGQT&&!jJ_tde-4;>f~*!_(9pnsZwKwzo7Q->0n`bR z)iW!>uWFkGXb?4_^C{YZV_TJiu8bL{6+0XKAn^@ecBVwyit}PXd21))GxoHdE_8gt zp0ix$Gv}?Hdf+9e*io7MY1TK0>>c=@&xS)Nk)xL6#)nMBN{b$Gx?259$Ar+iYVFxT zol;T(-~AfGEE@#Z)c5n|gP*dkzziH1*!Wm`j}O1XZbCi?&Z2uR7ed|vHXKN$0=cb5yb^WYGV%!@=o4-foo26OGzO4{mqYc#*8u+4Dmzx>X%E%T zj57X8&n0jC4_MFx`yQ6~xd+Y^G(QV|_+kVgA|A`xA_75Jhfhv}%!=lFWroQ_pkw59 zPhk#;fPrPEz%HylHE!7|J*pVbmwSa=ll!^}V7owDO9-c|;G)~~W3g^+4b4YygeIw- zmE5hQ(6`~DMdvLR^6r~WlR+q6f<*v*FKCM&Vh!jgjlr?)Ed@pqmM|~Px@eB!^|eD| zov*B021t*>G(kd26X_Y9IQYdNR+6E`pHLJSCmVx22cV|e@n2|Cob|ZIfi;$lvCT( z47uE%nHbaGt4B`~8PD;Idwra;AE89X1={Lr+ydz(YHNbgAFf^+J}E`OC33{L~-({nUo%Cbfv665qa{1>fV0JA|i z7mA)ziltz|vy8h2)EdVpkpc_oX^S+Aj-t+7$`4?HxD`Nr9kmo2>Ge z;;&5TK-h_skjIgB%cANh((28C-Ou$k+w|rS6FM#j+Jhy5f?qbWx_- zWB$P&>)H!~Xh0$Y*jfKKereW*U0u_?-U!BsDcU1fS4v)?w+{oD)(LQ`6bP@d>U-V3 z%7GcEz4^+`8F({La3#g1i_wpcY*#3z*957_-xkIBo44 zY?A>#zBo70z8NGLgfVRgXq9sB7)WOSy1rrQrUw>dxBLKcRWL6(q52xG7oNTyxUT`v zn!O`FEAa2;>&~j5K!gmS|I5{ju4@>k0datK&dO~U576|j*T~(($E!e%3(E`zgE7hd zW%;X{_08=lXkvMYm1ep}N5D7$v5f!6lNAA0zn^>3jyX zw%nHI!sm9fvOboV5Ac7$SZO9%4Cj3E=ey<4C}zs>L6_^wki#>K!V*mWO<;$+|2y-> zZmasPh7X>sfQ1-%U?GlAWQfGk6d~R*2{nkaVNs-G$Dp-cW?W$nz>qRh>YXYpGSHQw z!h8T*+H|+g#}#G_^Z(T(@?|uMtV_@!p!|reCs?3aA7_@({s4LgSnq;4TKQ9`(DMeJ z*U$kjc>ykg=4jCRrrKn#gUe9hGETtj_{MO-nD768+XCgR&%~91OWeRE zpP_;VD4C%v@^mi_DB@DU7!-WR3w-A*c(bIykw&%Q<15zjX}}tx^AG$71E&q*QPh^# zYAZS33g8C*eNwp>b`K3=hlXMTMy5XB6Wu8*N>)@5z3!)7RnV{gH%CDf?snV?gBM8A z&jN7WipV`WO*^Sgy+MJ&UuLm`yVn7err__5CyYRB-D&GH!dD-T*W6J&E2^p_+2sru zX;H4~;ux4T}PZYnW(eR zptk^fupjiDzJ8D695l+)J=+uuK-?YBsP#m@if_}uF?MF^$IA>C(qHoc&NZK{21x^F z>(pZ5)3?Oidg?R5N%OxkOzS29$pz*LCT}9a(62DlzeuzdKNBNy+Bi4RdSMt8mIzXI3~P;6iD)-|;# z4`4u6h>-e>CluMC)&38n2S6EJ%GoX2{Z|&q$l}h#?p`MLaJtcp>u=W8Ppi)>YsIZ6 zC(FCFv<^FvzFFOcT8dL*Dn)xm0qyDTWN^f14Dc;rEufA|flAM&n%6z#n7?$KLxovq z>Xp$ql~lLgMG8i})zBhILG)V9i{c(I*T_Hd<+b2QM+RrJdnifuX{kMo$3ROMJ|1J;11TI`A=@m zJ$0u<>RHho!UZJwO#pgr-gJ2A;`4Q91RpsN*(~pA+%C8ZcJC=Ct>1-FZ)4Qjy!O_aN3TXiOz`1LXFTj zykqU*<~*owRe_r0witBZ_-1Z0;++Hh32mPT-!}ob!6uiq@L1ANjPJ~k^sO(zaVqa! zz_%Qoh5Z5{+`-DZzVawYz+9u!p#7+;=BO`vWFF;1$3lK(Z3Hf_V4J4%pYrU1MOom` zJ86)P^|JH6AY(n?X#d~LEk5fU*;%1O-LUdDx$0(a&U53Sw-HBEI0Skk!kme23m|0%A6JB`&;@$m6hmG zm+^!?!jJu{6Cd+;PF19Up~zNh^cRycu_PTdAoRfq4PEi}BSAZue)J57I-~u@X{x#O zPW%%dD8$y;?cuu(Pmqod&0Xg2&0kE14vCvWi9FY#8Q37_;Wj#dM_{>?=tm#;?oOSYqTw3myUKhQ}bSET;L^WqTUCVVm&$ff(~wz zKwS1DxEp5&)VH&Hi;>%{s=tEi&6YH*li(@%gwkLCiBVK-y0R)2KL*+38w+Qi9A912TnrUp5)Y zJ?_d?US3=oBnR7zR2!^en0O9d5BKBUJiDK9DUEy@J0hQTpQY@v|;?qBA;w zMT-!O&c}iIiXlIS_B%$@&uDO54gQ95F;+S-d_1fdcIUEIQRgM{M73I8+lQJp$LFm3 z&Gk<}JkE-HSiN7sxvH9TSkpVOlWmuG0^dJ{Qiw(%DHZoLG@z=VmG0UO9mPEkjcwDI zoF@NZs}HS|Dys*QRLFt5PiRQS21^1si(7gduWOz6q{{}j^*e{$uR=0ZuIuE4U7po% zi)(j0E3vuh>mm_epczFTi2AUo{3QPVlgOYwV7E*Y$v|MxeiB+~F;->FofkLUXCdj~07oWZ98!k9h z3fBz2b%8}T9V&yl553R8J;9j^w7?73QZL|->^h)kd@(B$x_6e`wr^6kO5BsiR~Gq{ zDr>4dc^rKrXMk}=QWbhCCOJwGyxUI!g>Eiuyq+5-jGu8eg`ErX=E=Xx>s*gBJGOfN zlTeL?7ac35Xy^l%Siya?gC=j&ueTa~tGWm3>O!51s)2=8r+Y9k{hDE`?eCgfm#wS-p=8w1LvE8_#5B zb?;yK8D^afj;DS3#{JKDmjH=<-w^d2`NL(SE>LK;jZ?qHZvlLbK%JSWFDi{A>`Gcn z=tV@9u#le(YJ{W!Kxq;g%3Ppag+R0a=dLVjghLFco`hrv%Gjn7Ufiq`zXohAEk%k-vgZfKaT|DR348_6AtPg@Rt34qiGoap};3;-vI4>048 z8#3p9Zw{=V!!1NT_}GUT<}PbcZ^jFwEw~V)%nVOM3qlUCIlYfM6ebo#&(ceTFLB5p z-i=v;>R-!^j)cemqqYE*6BU#O`9>pIp`cvPF*}9c(KpLKg_bp|{yNQXXa8f+cEJ;f zeDb-g2n*6Sh7ZN?%j)#%q+V7Dhvm$YD?jlSo#JKU}LU{z>3#UJkP zf#=0?L-IhK0i`en%XJ=jZ=TtGC#wHbI>`QVDaSeAMh}p6Z7py-_RVv&FT17&DMw~u zvV$b+|BRN=VR^@mQZ&FP4|bNEhs%Ih=sx&$obOP#dF=OXYTkCIo<`XFH$**iYo`ya z2di6wTUlhIOD37*4pt8rHSoGQ{xFcm0N1TZ4Qd*IBQ}9WAyfNa(iD?xL|)xlad%&Q z4a40RzTf=BXt~LCisqmAz<~}@^WGbBhGUIRXfios1}6Oi7h2EObkjfe#>Ek>wg zfoBvn2$!>vb4==DYQ20!#(A}7SNI1)wQO}py3DUsoia4L68yrAx{Pj@h$sxz+<-`)kL`NGmLFOfn#pumb2dSR_dY{4c%6AOTz!Y z@LRgdsKxv=IFr{jpPQuvP#2gYG(p#cDo212(_dx0x+|+-t$wPH3(YP`31X6ZyIN!Z zF$N6s-@q>PtA^!#>OxXq<8bx+;r(2PCG(vX$?Kf&z}%s)#3Mml)BHXaW4J$hF=?)q zr3guWq5$5nKkL{F|kNw7DW{P8&>-Q`>!H+&R!k+v zwTpzaTlOr5-M&uopDz`AA0g?^e*~`T&nnXYFTu5GUzhmb|E0r4w}1wj5?ci9XAn`D zs}Ts5cGrHI6AE`Usq1Vd1xwO#hJ(q(&-x7VM@`R0F(<^wqXLwbgFgSj{)+T<=80ALQ+7btjmE#`@znU*F)hxM?m$ua(BW?K>WJ5dB)9cx zpkf}cHINoKG7uPkU3}YF2e%FA-T{txix{h^5e%m3r48>Zhez+?qB*<^RX5MU+cXHC z{NMBTTEuzCnC%d;BF#qOi09^DHmal!_w`iNb&?C0+e=Z|CQ=7}WEJAB zUP1!EuGizC_m0)S1HLf8e%c;+Zm78DB!!>JsbE;W#H6d#*MH%>{}5EdL4=IV#G`X5 z@8;uMt@F2+h3PmktQgfcG;OYDS06h*Mk`N;BB=X8AN^vWw79+8kKP|V-Q8%{kVh^> zui0Ef>jhj@3k)hHihBEpzr4ZT5-QH~)wg^X>;RUWdGk>PV>cW{F2k0>1N>4TUF3z{chW=mr$J7iZvZZ4 zAoh57>FKLm!~IMOF#sYj=$v;y<+u@K<|^FyaQmCskKBD@hjzJ&o^w_2KR$@nLL^6o?*(x4PL=vpsMMq|SY+q!hYH z%;~zuoX8_~icIsb+F+dIR7Mt7g}q8ssl5E5(bThB$m>0F+qV8I$Rtb!D_Sec?{T5{ zat5O+2#`#M6~=6D@mB6mP*0;bsbsh99&7ki>Ho6x0Xrg|$cr4?$(gR-N9s7&40g<9 z-#L)N%KE(_HTp}F)#UTNKys{Aq@i)1MwH zS-7y!nCe$F53Mxp^A!H7bYlQ3nPWLnu5%z<>HB6C(30|_kL8Hs96F`+z4epC(SYRz7~cuoFIY|TrO#iB=(X0#mQzQcq3q!?f=UOa!XOFC4WZc@cG){a(M zTyfthTq4RwoHz|nQd^XL?XA&hJIQ?e$P-|FNoeyV{WtLl^sG&?ig}WJb+UeSY7cGK zpJ-WNNmnZyP_p!TkoIQjB295}Zrt@fX;yo{iQqC>{a@vDVtM(Y$TDlqcHRt4?Rpd17PN5giS`W3^=6yQD-jBbnBv#|D@rvEWyGpR-G< zyFl;vvfV#k%C)xw_?L)SFQLJlEm!hqhN@?q6-@odz)nK3>9BLnjDGrfpVRQu3do*5 z&WT&i{-V3Di@jc?-hJ&d-J7ZTCK8o*5c_Y9_aydKjPoD0V#u7r{&CZ@%}48#2o>>< zzNAi?>_Kluj_(dCAb#>cC3Kol=qmL85u;TMIVjBdRzbZWISQXV46SOWH4u!HDo(ah&*TY z)>Csf=DYYE!sCR1)vsLr$|I0*<#A)GTy!!)UF1y#AAW&#MKvPOoB+Q3@Bv-3 z5I?TSeQePjP-1!ol?y}yOmIf%E<$O{xN zX8NBHC@oA#)oxwUuh{C}{rLu|Ynt}0Wk*@_u82#&8fBRwQlAlSlJED;%kA$h9sl_Q z0M=gXz>YU;#> z{WFrXrh{dz`{!B@NBsQ8(tIk}yQ(~)jv)xhzsj9EgF7M(CLik}=C8Cy>N$~ZLoJmE zM@4-yD;v|6)y%w9s>Y>ro83-H*fy(^6O1m&(|FeqX}GH>e2!K!#cz)!SYsyD2J*vy zdssT{<9(9n<`?2@5(zHDH5Vy6xO%M>oR-Pq)foDH_E^CNhjT2Aw}XL$35dZY74sDL z=uqt^*~xm@3H*u;h?3dK)oQ1o!_3Lrm8zz`M6`JdL0#-0E!}4%4Szw)j*Lm9oa0`w z`R|YB>yih;(-~I!-G*+7vr!Q-@vFz{8SgM1(BT^EMq3ZJ_~+uP`=NK|-x?9hwz^j+ z)-B5iPAJj~Ef-V7nb%HX6WiXaelOp2i6AiU4M5($v?5CkIY~N|i&4+uWS8O-Lp8v? zxa5|qJ(y}|o_M~^Ik(v%JMpGlk>sTkLQbH%caFaq)@&}qT4Y~`v%DXL1r;zg^T1wF z+;~D@AvaK7DmvE?-jC$|S68!EANN}Ob$2l*>|b<$P-9-#@Bz`WEz?V!*dL+&kp^&< zsprXIVEsmRX?KAcz$vL;{H?Mikchj~NP>$*&XQs}c9v1Igqok$wQAZVa&zt8CpkQW zc!M!<&5PG;+h6rA73&rMCGw^RMSi9{L>l4Ee)k}n>o7HAUspnUCxw`WxIZ*#n=T3a z$u>);40`ugt58*x5Y#&!SWobD`fOz^suITSIX^hU%eTQZpz4--c<_#Lh%7gskw{af zmWFwhw4+efnX{q}S)pQP45cDfgYYTJNZG&p=k+?Ni?DkSR%9ok6FMgc$A12fEoXOF z&Dlh#)+kvw>CClS^m7|JQ6deeafPl4`G^?I5N7v|^4PPbWQACm)!J@kJQ@8>arIU_ zp-t75O)Y5qbt}G%*QdId=I|zL>)>uPo2LJ2VhtIS)oY?hV+B4xp zg@Be_9(D+`D;}vH?fiem+pDb1+A5r%GJDMjq(Np0q=0`$Z7`M;kv5uq_znNKYrSE6 zlWNCO99E9gT`b)C(~j;}eym;Z0d`AmeZHqMRfbA+c*FceAZe^!f5X49m*A>mo~k!S znijdJ#ce?7EnSMSpR3^C&C0u{pX2jblH>;YD9>j1X zWeE64s$s}~5wzeBpF%dFZ16sY<4QV(V?HyYr)?7kF=BLze)fcot;Vd)7gfa983&>`#2Qa!3OP#mxJr<{Moe{OI35@tqc*5-g!~dZVb;C}Q+?Fak>wf3ml0&mxn~--@-+&yC^IgGcq&nT=7)6y}&ORNj7+j#AN)wVD zRvPg238XCu>9tw78h#PpY7z0C9bA|%_Y>5K*lUB51_zQfh)sQ6&<3d;JMD@kRJXDY zhBRKB*nA84hvYII9i^y6E$U2p=~UGy`Fq^aXWv_`IqKxh&30MX2#tZ=w9P2DqKnVc z8OvH>-#Pzo%n!w4q8F8J`|8&)40eI@xk=)>{fNzw0q=Jw2S27z^$1i!pe z!^FQFlYS#w^_ZXMmRCR=)vMr0;+izssxzuafH8BdiCQe;&OD@SP@32rG;TH0qMd#r zzuoZayLjqhvK{#=#W1iWGU5`v8l4wgy~gju8ghmGj9AOW+QvH~Nc>Ff4++a&f)feq z$IVGCsvdSCT3u!(MT4ZPH<%bW&6wCE*$K9=Ws(0!?BQvlFj=AzG@&G5T}c$xCV~S) zTt{uDrQ9MZM!E^D!)yjZsjTV+=Eo7=nZ~)NbcvEFvPDO@nNEV5?{GDeGsSRUu;XIh z#B3i*hv&5J;|J@%VewG3DZyJIodaJHA1@`fv0hgoCp6kK*c zjWyq_NR5A~x#k%^F0BrA71N7E7aGj|gnwT-mm$AkY_PK3U}EcncQSo8mcG`#S`>5W zNjb1?hI+l=6l`{=ki!>rX}TtarlST0yuYq)wl_KhBN$fs2%%anM~2M`e^A~T58E{R zs!znR6b>Dus?DFO+wWK6?{=eZYgvnBK(pLnQn>nccvyvV-J3iF(V8(^_xVVqOzqAR z#x|Q0HMnu(wjJ26aNfT3R{!bQ-z?h(<=t0<3&+SuV;cQirTT?WH8=rU1^&k|u`H8j z&wsNT8CH)7{AVidHjLE^l+${1Ru5xG?5K*XuL`FW!DG8{_ebl-L&vg)zSHcv_o(6V zFBq%;Fu0}>dU%83ciPKZMf~dWSLNeY5vG16yGBB35vskOh~Kh{+m#CaGqGE^2z;a6 z$T&9DryZL}I6w{N%9kSTskW@F^YAnR)3EX#fr-f*KqnL24X_LUkx~SO9^ULNyjIj( zT5B;Qwjq^=(a##|E02MxnhhE>xKg9yUN*W%tgBeNm8Fr#lFK1ipZo&Af;y zslt}PtlkJWj}qfgd$2&{B?xEJA?yq#o}ycPw`(-O6!pObCAwzlG19?Rfl5F-HVtksy@WoI_bUr zzwsTl3xnlb(~n^LG|F0K=bCgnK9FvyQL|&}_lIot?M6A-7)-o~4F-qUbF=SU6yL|L zk=6b48mVPu_st_9P^BE{T84{8fmsE6J`EYVj0jX z6`x31e7O3(=$BU=SPbX!_U6f+aJ(|}MeI$_L*>|vWlL~u#HZ+2CxDcCI0d8XwdL7F zNfjk`e%B(X;Bi&7ecVgJH#Qem!q3gSG!t5&@wXjTHnK{cl>}_h3DQI~G9;JYJxRXA z6->u47{4Bbf&UQijAZPuNrlXDt)RkpX*4=@1yN2iSv-z)ZBoj$bKQc~_|FSe^h-5@ zQClU@2(lJ87_t(pt5@PST1nZ5D9|v$4;ErqWDF$Fz@8%%MqI@{u5^|6+>T5_zu{Ru zrBO#4&&C{-T#PaK%~*U+-yIoW^ksE*f{Xd9>*)$Qfw5a@LO&vpCYSHN!jqG(d#t=( zO{P3$$@e0Rmrl26POlQh4}vvLiTBL-xD|{T>ZwI9RcuJBP|&YXOq(lVW<=hEP`dmU6GjqSBPtGER1eVI)hh*!ZpjG|ozD zA@Wg19CG3s%S+1XOagWb(wpfFF3l3nH%oc6s7=prFVQ9x5S`^E(r3F3sEZecs(1BU z4_i06%*UOu8vQqzB`e-<{Zm~2<$VDJ=RX9Ofpb^N%$+pVBkKi~j12?Gc7%q$J zD;z`OTJ&oTVjg;lPKMP+B zBgG*Pd^UQv?uzJKgiMI=@_&0y#F$p7ENWDV;XlPWZIzr-USk0ou|z&)*n-o+Z{coz zN2A*4lnfym*K_FIU|G>02|qRGGrYBHojw=&pAZUTFr6{^35tU$xNg+jQAA)SXbl?1 zk-)A%Z%bb{w?|TEFQ$1kiqhKmcuI(&b34mgpmWEv+PM??|oon5a+oDkxmAC@_n|j8rSz|R*X+_uTE0U#43WH;WWOLFj10kXy28ynUREtpIF=XhVz}9_8z^f z(%wzzhb@}R5+h-?ByaZhrFShAymq$L<%GlLcQmuby0#8kz$BJ(gV`u6rN%7KUsSUARH$-QasMWDL(9RH zwayrNZ_IXH-P{swrh+zVf&osq3T@t7$65~6r^ZB|w3?NCLA9~XQ!%+70t9vs^1MV& zY&F-*>Q0x~pQp-55K-t^RLj$GbI;bJ^(llj${Qd04WJWKUX0NSAl)Wx4+XeJK24Uil#~1?%E3eSZNB2%{)-KWD7b3$ z){3o~AzP6J`T=z}PQ}ZRvQ&0~383x0EpB%dr!8Aq!-$C}#JI$7&4g{DF2dV_QMc*e zUBn!El+6VAYGRmch`jP5&*H!yIfVDH_OCH+ChfZsoT`%}t)zMoKjJI)9gH`Q{4p z5(ljh$@+a%IqH7c!H}&_aSoxTzh^@>nPTxvq4Su_lfraZ~cSO?v%;Qdg4o z;NM$JDijbesDe|mxvS(e=Qxj0AL14jb-)_IoH*BVSi$A*YvTC#@IK#$=ce&V2JeT< zPz5JCh2^Va%?=vm^?z;~&%~=>lc)E6F>EkWew;S5`uAD*wjSRzdx&Ze)tT6miNl9l z4RTv7z^PT8NKl$D<56SBuA~l|`Xbm-)hWhP$j*N|DOborpoF+haG$!|ki@Upgs>UV zi@pQy1yO>mI3ath;rT(%lUtaJB0|7?)MQ%Bg;hyrOZc?qURwB@u-ZY*>eLq`xKqs) z!@p7x2`|W`)c3J<7c&P!2kOfXQj24AAD65DXP7mhao|J4o7lQ8v7|}Gl~tsaDQrVZ_DP&Rcgol-$%+{(A4y zCY^qPY`gAR1c7YO@ZTR)ryA4#YqjKltA2KZpg$@omY?Z5{qQW)B9P;jxR4j%$Y&qxLnGTFf zL~T<-8|DxyWfHc_X=WH^<}};v^nJQ+pFe)L+pj<3R(l<`J)e)m{qT61<`7NKghNYE zEP5UN7x4jZ2PK&}7gnb(a_Ws8iTLzyqXM?L$PS*#)+}TFIviC%bkew5kUyG+()C@C zS@aF|lwz0KZKL!cE!$qS=*vu%V_JprO7MzwDby>^eCD7zeS>eyMj>fAzEZZt z$q?8{@lo?HSZ`FklVclaf;2`ADChF6)jwIDkRx4x7f2b^f{3BbCcF|?-jIiXpDQK7 z5BGfmY1jUU728BlPSDrI#~PJW3^*^NJR1fCapyVSSKz;N&wixqOzj!9YCfQv1!eL} z*IS%tPr?^`AE)AWlMh}_cVi>!$D((;vHRZN{T=br6`?Y=Z1ffx1<&*WllnjTVJj)z zQT3VV3vAD`0?BXUZA)jR$qb9uO_95CM~MAa<$8zLMD3TQf_f#w5)m|7ysK{<6;7!?!9dAHRqwJuSOdzeoypbk0wm zfEF8@Ftl*RMbv?9qB8V(=jhDqC-Voq*wdN*)gPy`Sr1jpS-S6cuzDPn8{J%J@%Q;d zELKU8b|jtK6CdL;*Ksm@Q)~5ar1@V53xl`K9FPvTF+={!3yR-m9J0AQFyXR!e;ZqJ z*DrX+CeT}a7$~_RbF>Baib=hE(v{s$+)}$ux^ahZF#K}5FT2~e4jm%dmp``W%cqilJo5N7i*%>4=pHyS z{m-9KGqA`I<5FeB*p@*6k6c&v37gKbZ|KnCK{`v(&2nV(9RgK&sF)kBfozC?--iD} z&MNv)&0$ZM4LvcV3N>tH*%r%Di^%B)ngP|k!PGT?G&j#G>A$H)Li6u>V8-?1b&t>i7QtpsG3xcyzhJ2boV*&zTTyBcDBlw)cv`fiZOii{$7Wf%Up{yn3tjMV$BnG zRkH98M$#_(qtiICEzm>{eO=5Sa5kkv=E2wAJ!i3ppzPRPiYz_^jm$&63=UVwJ1;4H zmj3VK@fLW}Gt2_@xrLBv~yDzeRG#Vb7tA5bDG;?0v6OUA|5|}(YH{bY2*TX zfx7&Vu+g=<)c&Wo&o!i$l)S0CEOLo#KMX$9u{`ZiI!a!87BwOVqB9 zc;}D3`4ZyyuQT`-RC@BxFa02H;h!1$w|0Xj3*6ZM4!o2KwXVhyoR=H1Gjc`RUdLRe{A5j`{)+p7OGi-x?VT3<4#sm6(9~q@msgj zaJEuhRXQj33m$c&&0LJzElovaa4s`22OEP(hIHeY0fLHBw@rsk9(?RX%L2Zw`a|mt zaK1o=pc>j;D!f_y+B9G}&gnq6jr7ou7Nba;`Px7|^J zhS^eMV=KVt7x<%w*GX0xsDiFQgX={*^2aPQDw`c5ce^cJvB~%?{@b!`cypN$?*L-_ z@Pm=$1n@l>Pv?bi1Yg@jNSb?hX`R+I$gq|_$@{OEs3WpD&qlFZ^=LSlcq?;jO=!C8 zTDCPeooys51q-i7ynT01+o2kZENJkLAkrB@WxHXG>(XBu&;+~(|YBBvXH_Npn z#I?ShIk{)SuMza;if|RNE7NfBG1K{Yvv*6B^MRs(jW)PZWs$*k-xPN0GK{G>_^HkG zYWz_nzqM|1&%2{AZHxw_S$U{c!KvlcSJ6|~k^CG@OqZ<(*>Yor<(i3Yxk; zKk`P|2J$Xv-wbIT^`5<{-W#scUkT9B%O515*#C5a+2b2Gb_jsVo!bq;Q>Kd9S{Tr; zJq93>SF3-z8tqU`72H{zV1FmCU9%=M3!}uJ56GT3`H4*QXUBVey}`a#sI__(+z=M?t?t;uroqQ8SlfRHaKLmle*81- z3t)HxjBT@T-k@$IH@3#Kpl!Pd@U7yocIO&)2EUNfez?Y!89$KCN8K?F(iqeUdf_@0 zlgc?WM&AFX3SKMtcp@Qq%22xuQ9_{^npP}e4~oYuYC>}`3iOipqns*@_KjQ}c(6dy zFkvQL-y5@M!I11qe>mm%z=(_D71gBLoo$o=miTic&*S*lD9O~U-+8$en%Ui$A8+5-qh*8B$IBzf0-eUnXmI?iIGliuV5 z#1b}Q!@OdIVjxrL%@?ati=iV_Im{#83#$Ml;s-$lVdEzgT;=gtr4`0z;>CB-@k#)K z5``;#QpaQ?CED|bKtn<78D~y-E!Z*aD&jem9TO>iP`!(?^{h);eyv7e@BGmq9|~@? zyEJO*y#ECeV~?2Fw6tDhej^^8aIYvOI$SMGpL##lQfQ+-{3nEp?7@O| zNI|+~Uz|(AaZb9c24)T*k1gtzdDK^~;kv-JGZF?Ij z{Hsd(!IIJSkvVro*TiPrDjTkbc!qZr-7_AHB*Hf-I&YZ@HgBCfWI0oML{4BU)1x~BdqV7mjiI_vTQ7>cc{Yf2K)}rjWFLHTx6*Jc&GM7TnX`MwzAnAg3t>xepsv7Y}iYGj#haf7NWQ#Tp?Rn8mr&3~t)mktAQvpMSO0=+q| z;Gix6?j^SIPFDl68=DR?Kw+==29jz7(Ewj)-^iZ-{6PT3B|i*QXt+I6!*{J|RjHiI zW0?K72=M54OQ^8DS4dBwTVnn`@m^&r*8EmjHes&FB2(6)-x1|J%GDRqaQmg^%=Lqx zzR3PRC%*PoP!=jS>?q}F{T8(C1regR;&ql`@FejnoSeNso#PNJsaMfY&Tn5|!`771 z?3<&`Em&paX*eCJ-K%~{#PG-50QJd0Io-f--W^dk3$*}c0o?Rf(<=B5agadOJY130 zZnl!Y+?vReZ)X_Srvq3CNlP-G${gd1Ke=5=i+l`gl2%MOf9PVao&M*%RSx*56lMJ} zU!wjmI@tU(+y>wz?)*mW%iL@P3UK@!T)fe%-KSUAsh6||UVl;A?n0b<>)%|CG=qDM z8MfQjR3?+di~>pZ#=+HtD)DQv;YDN~eKb16nCLuxB_%4dkdbGAC%{?nlJ(LzfC#TI z3T=KuwN`--NE10k0-%^T17h8B^knB&$*J?nyI&mT13UgtUE;3kz4u>wf%C{6D&r0h+`C0V#al$!?oga*BmO5pD;D4hYad&40u?6%jUg={ zR$0Cev;U{T2?0t42TW*B8%PLuWiM8?fOmhLA2Hz3iwK{t8`83xvH#qX>vDeuIc@*$ z|F#ONx)!9@*#EUW)HG@}pUSMl(MtDcaGs4}*W&O_wvk`!mb0xs`ZXq3o-Sur<#GT{ zmZTz(xAM+Aq-YgrM0@}$HudM>@?~HhbbpdnHAURf>_47kXhneM3%F}bNBQs3b=M^x zFj0pB1*{$1g63c;FOc=xijjp#CiU~(AECjtL@w?#Pv zSgsK_yhUb2v=z_+Gik`D;SRCKhg(#C7;>xNI$|&HWF&Bp@sebHf;0{>*ckfW?$0Ol zBtLJx+d)nNq+9{z>6BNm!xP+2L`abO+TOEnyh4f+@Y{FPupb}~{+E~)F}fqoaY1r} zygOtC(s_1tW}X1pjx0Nl029*5?x7hs9<= z3WI$uc-~)a`_cKe%x6)9&CZ^rI@{NA;=>$9CWCxn^&&U4Tyh)!U~b`(*z!FIwiEs?i1bY0 z2u`tFP;eU{#p>`3nN|!g(D0uUwh`tB2vmKSG{OLWEI;@3$Uz=GY2QJ@Mh(I;Jvox9 zHr*&Omc|KwiR+85&XmOxD_+~>INzYI10ks+YTp7rCy~a5*nRr`V}Jk zFb?zC3~k@0UzGLJpl7s_!bVM-v+i*x=y&;?Ji(3ulcMhw_NDEneZRf7HT-iUfOJ`4 znhn1JP@`9%*yuU;mG5~mLwFl*`1fm&sya41(D82&XYXmYa(4hlUTja=uJ?MyNXj8> z^SG?$$ZM2{|N5k`OCk?8|EJvaMlw~bgo7+x22&{F6yc2{>60#Xs!zE`*hb^0ty&Eu zoySaDcl#d6pZrN(RXEuA(S50q)F1tucuPmT+BKrHLsEH5_IBbeMD6=cm-mj{ZR(X= z$LUp^KYbxs*acDI6*juk`P7WiCW1hb&Z!;Uc->_30R0zR6bnli6kQP3?HadBf0l>J z;5@(_Up2fA`6plOGv@Y_D{iZ&#KTNeL4C&^4-<@tvygIC@KsA`Vq^jz@+GQajD946 zh5ix+Uou7+jx;+(hM+^jNTUEUxplmL?E1Ea%)!n0x)Z1B$3`+>z7{+EmV!((Fn{2H zO4U&``Nwg@WsT<2I@d9sGVUiJT-CzcC)9GzXh$o3GHLZWCDAEE9^6vKCBid*6OH}9 zqz<|NiE~TX60NC)=g#vQ8YZjy>-#L*)V}#6{=FyG;h(J^t9Ap(&cQJq*rsooqeCk1 zv|#C$Z6OU?+H73LbZWO@+CG7we7E{Bhiwa&-M#6bLJPm0{VHZ={*Gp3{&t$i+0rZ7 zjzS}$E@+W2Q@s(T_XKi6jgrZ$`1CIpx4;#iZClVDo19OyfA2VbaY$tW+}k6Z4q~g| zFKOUXo5fsVOxsVI<=Jgmjf9fIMs2%kv_HPJZy5528+$KEtUgAddjfc6hb#Nf;DUU% z32_U?g;>xM)iCc{jI1CH)w0nerFC1!4nZ%gk<=dNS^P|h(Y;>SEVptORkaIrmj|Yw zrB|kNE{8)`j|xPIJ;-T$I0n|J;YpGu+hxi^e^`X!MXdVc4}q(wH^FeB>x+itj`-* z!C?>m#_SS))c?8HI`HVT(|;W=b-3Vp{q&bihrZK}7YyHOm1aJvJ)W$Ky!`fPTguVnj)B_Sk(jy3YcU!*qCN zekoR8*I&5WGBz#!i&pQ8cJAR#rV8KGb(QB%>D+{;`{K7r4lGA}`s~JID9)^YDwV8% z<=b(^TE(@ShZ2SQ5pkJ;OTA8>{@tY#`>TyFS2{~$bBP*%RG9o$LA>;}Yw2C}Z+`Mo zW*n{vH%eQLkbHCqBw|@(QQt`tq>0tW>ZH)Ao`h$?im{1!>lOq;9q*k7{zX!BNNP|n zva8tfZ}UhhkXHig9ql_?1wE3`$ok>h5%wZVop^iMstKUH6_~d|c6=;X4EuuOis29_ zmQo(dp&43PRgg_Qj`J zo#%%=!dSN_JY9db1=Inm@0tv6H0`QSy;+yU4+jmd9HeDEYNJ(v)*H~82C76cQXGRWNa{O`M1E*n3ZpYH}Jj#n>QS3 z!uO#V-w$A&oNWCkYddQliz{*!x6mxd{DT zfjBCC>WW~0@o8EUMHN;e9^;o%>Mq1626dMrB2B#!+XYaUzW3STSSv-Z+QkPoK-rN7 zDrk$kG|d^kqNYQh4XwLw!V0D5Xh{_mTB&oQM>#P&VgQ0wSUD+Jj}4^7|M%Ud#TZg- zg2{T}8>sUpkY>Gt;-Vm}w7!Qc0u!qr@<=7Dll8fRoJb=s0S=Le&HocR@TDkxUHrB9&;>7m9SbqOY zAcAFibOM2X^*;Vb@c4zl{U%gzPIy>$Tdt9{K+dz@*M-_K=bk4-zvrJ6*29A)?NXzG zsp9++HG;xNn_TK&X!ySqOK&DfaiNXS$F;F1e1w(sYkFTLQFy@BQz+} zrpXYB>+JXF1Q*iA1gIhHdAMqs^$jRoWs=XcrSB6l2KLjHz@lJS^u~5%ERFCi8L|;x zlu#TGxirRJLJAR^R}Q=1{XJ@b!_4_&vB{uP`1+3UO<)_J$cR?_*gdhJ6L#)}LVPN& z8LzPy;FFG16>k(=)lPhqy{o6mYh@t^{UJ#U<7wo)i(HM+iwp-a3zjty%$oPCchUOyS2y+?+^uIRgs zE|QY&{|z>8lft`uW%Zlb^SVR`R{$vr{VGntT>N3Dl677<7+0O%&Bx>UNlbI zgSG)0#JWW1+~;zU6PK`|f_4@nlU(q!F!9n<4~of#m^J1yqfTK8Vba6X_I%(&=y;oJ zQ5gKJHuoSbyZDn*HB8-#Pul_4C(1u{8patjh$o@ec}5o|yZyvlR8D0Ka%eR6jVI z&dx=vl*fLy<;@aV*B(t1th8Q*N>*z~fND=HW zxMOKtD?N;fGWu<`0q!_!K0fHy>Kb!QIlcW;ZsY?EDi9c~*lo3Z|lG&@f5 zwrsaFv@YGM+iT}79k3+b7@Ka9?e5O6gCl3pG52vSUo@Km@{6GB-5xIQ;jcVv$;Xb7 z)8fE}w)?WLdDD$%O(39~5y`tMf=zfPN1j8P}P6ts<+ z_1Yr`3{N6($Hqd2JZd_MCm91(61DcPl|@N7hZRL5mx&R;+DrR?FE|InNQX0ms~ar@Ksxcqs3nIKf11){yG; z$6QX;X_7)*x7T*l*fl>HkOS?(iXN>P&J{1>O|MztiHlVKd(qB7upsvly7>pt2s}Ia zTmaFSh^E$MdZ=a8UL8_mDoejdn3i##-K&jX3n?a}JK__+hZg-2CfMNVx3q<$3jS8) zfZ_r0Ro{b0UOA`a%`5vY9t$-Am+TMR8pT)#tX7r`IY?`-hN4A&;^F_md1qRBP+j6$ zwbAVr45gE9kv;nvIlQmp^MvghnpJQLCm^K}0XSM$mbbR~82N|d;x@^8L+VAF8>n5#~nd1LoEI_(iCjPT?< zG=29Jv9&R!!fxTL<2UqpFG*^Zq>2=N`U$P8qQU#@qJ~EnzsH z2D6@Zko{QU`?0J(Lxe~_)`Zgb#!1XM#~V@z-J|_CKR5Kpin3=z)`pw;U6t!u=L~0v z%-G4L(~@w)#EZW+{k;saNc-4^-tceiw`<@2lKz zuC@dyq?hrQENgXbYh_6UE&zW_Hq=2g4_ARh8uMVUuwx|4OH`-5M~q<(xtD418SGUO zAK-63?}9aEB9JziorG$j6=M|C^A5(`!CSi(t#0|zqzD$GQH$n_`(@}Y53Jxz9t|6Q zhOJ~QRUVkrvmxifv8(2}*zNn>4b$G|&v;%N-u3QbKj*Q7N?AObHLo=p=$()!_RI82t}w0c1w7#Ndq2UZ|2j`AXgt@A8j9GU1Km%T?_A5ufWGha#mHsbKBB#{X> zocheViJve27-XAX8}@yAG~jKs;*C~IBftZQw6mq&qxX|dH;R;~Z^e>pfwkne{RtCV|`EaVd{g#aGh<;*GK+6VR^90dC@jNO zUO}}56Qc)t+P;}~#s1#)&V#0mt)60A=5hcn20n9U8MqTizC5d$oG`3X*RD&{m|tfa zm=trfE^vh3-S_g=F;M((B>oSUkY<2(ovIMu2zc9~Yjgyb1go4_WiDV~8Zhs43N%%Z4f$YVfHyjF%#1$}`kD<$l#5ujn3PLuGExA&nTgmiqzmZHYEDOBQauG%xsfK@NVBR+tk|TA z-0~YU=d1vKP(WCR2S)_zosfBTb5)54WZMak0#(#UJISgwCy%X`(frcf=Rcy~Kn7v+ zniW`!&#HSt$5SL5Y${4kuulWAv;FbcU!LZE8tQ8$qt*3#orQ}shk{ejg+eixU|P(j zT>2{g(=Q*bT^@s5CM)lCu?KBMqECVOh}uxA($pR)si_B81=Db|n2mcu+IC%&Pj6xq zShW-E*b--31n1$)q z6ohSUY@WI=oO@bEp|Y3`wvGM4C&XsGu?^k*t43SHr2feUL(R(=l>r@Eim}$P|7NH% z+=Tl~Y8ImcwejQb5Yfe=fV6XUcS2({{7Y6y*O`(3%7TsU#51|$1cQxbH(ka$CJyIuL!7EqAxJ+Snx;`Rd7U{Bd z-Z$SL7HJ|t9urMdm0=?C`X|cu|H#gC>#&@NX@|v&-Cpf@#jOMtQTQgSj8}D_sBj2# zU*IqOjqr8oVi|G6WX|*FjbJ$(m%5}-{Wm5^YKx~jN_OGuVhm?*Pf1tXmXX$tD31H6 zfm1r7;Ld3vt;3*)ykBw*gRlmsA*9RMmGL4!F2$x9qs|@OMBHVbcw8tP94|3FmM`|3 zhIP10d>38gg9xQ57+o00W9Ud^!HVqMXP z#bFthXiK5j0!#woMSOE-b*L@$QA%Uc!Q%%0_fIBhz!y75p{;9Y_V8QUCY@nDz%Jq~ zFl!gE3xcSPRbttX#8V;V4u6JyPrvoRUsDz*?xmQKb)_o*U3?-GJBkNa7}h4|frAXk zuZC?6u7+3qjy`S@Hc_hHsk9F|5oh%quLyUUtzebK&#TPhl!fKN;c|B8QG<{1!=?2c zynx*7v1m=YPJ2w2Jc&};69$++jlUoa`{h`K*{s*pMf8|?O{L$D4Z@s<^NPBFf>r!t z(5(F8SYPSCJtHor&@S!G^qluyvB(v_cSmj*su#-5XOODw9Ni<91%IRbxkKU`G!+iM*P#y8;dU zj!|6w+WeBgYs?87-M@V_P6e9V4KW+mrUQhp=XaNir}#gD0W|Zmp0P}g*Q&NEo-7kS zs)7NFeSEyB)oEy3w2i=M`S8t<^aQv4E9?gB!^KL<6l-Sx$eJ?DjR6Woiq#phz3Ovp zX9s`5Pv3n39=C{xwwMbxK{pF?+yl3?1J57mV;{>tXTxWEqP)HZ58gRPOT#HgO?2ou zyU!UyC(I$+xbG`-C2}SK+3^bqfb0;I;P6#`$)bMVHTU90*{y3LK^VUx4lz%*Duwb)ku z`a6~reAB7~KvbzeHb- z8AP^ybA6T+QeIz%GaOvYJSCwS5|cgdv#3+(Un;YpY6E82_%}u7UZ}J1vhm|N&Gu)H zQRPJ@BjRUGuQtH*X`qUVC2w1Cj1h){o?C8k7?YZyU^*(7)J9r6ru zb`ZJrGx6%W;CXdoCJnU8ar53XRiO`X!@91SQ#RS{~N4Zeur-3*+{e6gT5bM zr^CDpm|IcSexA?J!S3}UkPCF)N&1`c>hyp|;*YqvRSrWTJ_)D4pfxZ5roAtU@d%&^3V0;>?!qav*gIm3fY=r-{!x305+uw&DL4me z`hbFEXZCjEw>`7RmoLxJj5Vi!LgtwZK|8-id^{A1tkPg;%@v0x4of`E2TNJ$WlsK; zt%^gfTPU5E!`H#1ov^?XmA%<~Qzb!}_agl4(Ba74(r0x(F*lr?l^O!--HrW4CCExm zced?o=d{?O!=;i$0@<#};#q@94iWy_lV&8m)fObVH1Kbmws8>Srx;6;w=C7n!*5gSR?ysk-B8G`3z(^aQuy#ciUGpM(r zKkH(PEZW~;Cnqiz&MJvDtbZ<1fR~S|D^jnRTii{tp%l;fu~Bblyh$c^U&Xdp49+fC z-hH0J&B3$ED$*-iZ+wcd6?44Anwjy&UtNp!-?08~EvN5tVv9;y;TE&*^EwskQ_2+0 zu9BI=C)e1AT?$3dji)wM@=3u`{JaPMB(oNZ5F3dS6kO*U*r*%g10q;zg{ySYBI$8# zDyvte3)qcYv^<9w2(e6^F_hS2^<+tVMBfxfqt<$Fk*nx!Y`^kqDEXBWr9XnZjpHa- z2WrObdEWyXO5Km+STP=BUB&Iug;m3BBav8#Qi|-BxlY+riRG3N-jX1E^bkwPUN=`# z)JhtA*HN{w{G#SaJz@2$cE58b-C$X0QPhc#K*M{MexLa&oTat}v8J`D7I~FR{S(wn zaC54!-WN*Ta4!Jyi>94&5F?q~2Dd30nB@OiY~PO}jMnKvSL%BaoJHne9Z#j?qhcD< zkDhGlY>sBeEV5z^iuiAFk?AYO01+~oT@+-N5oyO)=`BGEp40(W)ob#MH~Vmv#m}Wz zdwZ8k4|8@=!fr7X>PLmyMT3W~*)fzU!Cl2Oe^y9?Jfe!yzm!aS9qtP34NUm?`XHsV$otM z@*(ZA1}58~7}<+#f_A|A{j(3`c2XIdWb>F7-d~iwSJk}u0j~g&O?_e@RT&QZ@Q&n4 zDRcd@+a-_MI+9LQH_75soA*+k3)5W(fP#_zx+qR#VQZ`MvON(;%xyzVAHJQ;YK=v| zi+v&WrG5I}u-$oNKhH_>h&A9xt-0Hv%JZ%N=5+Zy&r8J#3N&zrR!4^jv31WV9l2#_ zyD!+54W4k!`_N~oP^hm-n43{vZ3j-K9aAb%_?Whqmjol%qU-MLJPmMe-(bD4e>L^t%RjBW*h|~cTe3*&gCGwKF{P~XcoPzu8b4g zgtmtCL^%gyY#Y0yrX_FfYO*xvK9bxBSi@J;l&{=IoWmU0d^#A zk@Y7ie>QVX=89d>vF>LmSJk&k7fDw|JL1zUh0~HsJ@E+mpP-v8Vax8wZ|-dUHZGWY zvNLj`N-se3N#KUG2CJxd&oJ_B{d%IV5`uGx8q*F; zh#Tlj)L*m`MCOZDOmms*9Cf@7s@G3G2hE|r!p?DxuCb++WdG%s6DZFZ%PB6c?FZF@o1H{mi;+%gpD` zP%4DE^z&E(Wg0cS3*&J9`dV821BRpkZH_>w9lVp@pGV@q8vZ>u~pk!m?OCH)-jyR$(W6xC`&EofN(#F+LxAC~B>pHz6)VUXQ&c zlY`4dKSRe-+L_K}e>>^F)i_vhI{rZBEU5`s(wJ{pxMfhLB0A46g=Y@_Q}l9@(-Zes z=p23~;bq*wxw+>!h6&`zFe?Gx^yB2)e@2D5u=O+55dVo#B4yCR-F^6<3@M% zCj4~bW8R2m9Qno;HeK`GU#9A#`g4H}TLthIqgfYf74O9d2)Vwf`Ma@?`*1Cv>B^r# z%EUo=wR<{zRBC`idsOvSaZ#A_FLg)1PNO_D6yAZKtE4O}F;O`Zxz&ZjacPu#P4S4* z@3;V>209>)w$L~pLvIf1z+_PHa;;tIdsu-=GmWd6E4q+okGYSO&+eQD{_Dvb(6LyH zpK+IB{X>t(sRh;BmGa;Dq4M5DP4Cz+QwqR)i|j?n`As9osGCHd zv|&h|?RJ_-$dLya$qQPX=G77N@KSvc6k{nJHj~aO-y-<%!VB<@{^99l)-n04Rc!uOR^u6|`j-f^&m z&)B_@8O=D8>;^PP@?BfE#;AarRUPR&6JfOU83Kx8MT)TKeT!jzYQogaJjsPZI_|9D zM@UeHulfzoJ6kmJqqEpSWEdcbsCDA#t|~Txox$d3?QcO2OwcVl(0ZRY`MQT?>kI(- z1WV*LY~Kr{C#7{JWjXoPaL4`x4N#&I574!|KLDTt zgsHS36LN2b4Y5 zYF%}>JKe^_DT(&9r) z#%Xr+>4Pd7ypf^=*ya)iAOnP(K==A>e=Rduf~0RKmqey=qOoX~a!ELVl_}G#>$3Yz z2U3E-Y$mOhO>48aQrij{E~$4 zF8a1s`D1-kWijvp9Dk0wPee`ZSUG6?fozCBW%NGiam|=U28nX! zu71|SI1wJ@L|4N@XUj0#XI5os<#^Lu5n8`SqvtzW%5EG#0x;C#w+?uIjyQcj?wYDE9yo^^Th)< z^66-lRS6GS)X}~{)fz-(bsP?SXJ)<_wbP#hK%EdGj z%}X#Q&+k!~GZxL6zH10H2bRPEruT%SB48B&c)79togFcq#F_x5|0qE9{+Z@Fz&-4! zOK&?dTDH{w>5q}P4o6-0rMPW#`(?r6A@Y~{PU|)lH+F(}yM=W>>A&XsTMV(QZoG_H z><%?n;#b%DE61J?8X=9gTw~i``?bO>P9Ns=8IrdiK;M__X*TI~xN$Ai*e&A~^6T$@ zhoPQ+haVwUC9W$vgY4SG4zNqqiI-`P?U0**&oe5<{WcH-tdh9gP|EyHuJt8$r5EyX zL99|&YN{fOtRGbvepPlO;9ILaU_F7XdQJQe;Aa&$6=O-*5~vJB)|Mp}vvK5Rev<17 zfxcnY_}--Gf~u$pnJ~*;Po2dpKJj)OUFTL@DQ!(Gl_L=}l)~jtVs^{}ag1Oic2wmC zQeI{|3jmf@$GYUOZls_YaGd_YOj9U1A8>0vU6-z;AGTpo+gi=4 zr1IQQZq>;_GtoT`D3);RmwM{i)jmKSbEXO|;+=BrM2`o6^l_irlK0AXMZFjCX`-$h z;V}_`Yj(JCB!A^qhgdOO6~1b8;72WJUlm4l^1(@oow(Kv)O|Hq`iZa0y03;x5YT}TUh&wQjB$x%&5I%0X}Ai8Dg$vjQHUOrP=9Si{XnF9Zz3a zI*QL&-!*FN3(N^gVb+WMupTqp+qBjGj>ps{HcWMa62wQeh@%D2)`ub1 z$p>Q;D#7cyPG6u&!yU>?rlZib=r?09u;AQG5Y2;`&nXz*Z=BA@{F1l&UqR4@Bp5{`g&TqXTYOL$=| z+s^zOV9x-ADzSLyZ~ldx_}@i;Bx!;n1?J)E`Q1pFyL>0Wt{-9yX_V0ok!F)1;8+2A ztyM9yO?D~Zbjy~#E-*ezp%ctKa*l)#Z+H^wo#I-T=r|QCN^N?u+9rE`X#hn7fK{?f z-JQUTjX_j3?s2wc+Tdmb6jWJME{~$=&ir#uk>rYKn5&+gI8(I#iBWx$&eTQ8wA$d~ z{E2M-=g0ZZ?xb~{giSnjOmfuLIh3hDhj_@%NT%S#Y6!{~dzM}1CWcM3dg7+P%W&eU zKrn-}Pc(xXs6Ed;OgXl~QYWj`r&k1V_RNpm2D)g|t{9D|4e-V1qj>&{ll3Pp+^*%V z;m!@P2J;bx>~4B#U99`6I@!!~ct&p^7Q6^mS0iw4Qt*^n_o28QZJ`z1u8#R5jj#1$ zR@IJvu|L}i))~lY2g`|D_IGgs;HcG${EIh+ts>d}U^gr+qjt|3u;O$8Td`8Scw{&C z_hHt*I*};jETw+NI(SF_aaQy->5l&0>J#|*GdOiLyqtX016$>Ncs8*rbG2d;4us+dcNhN9=+{hZqypiRklwuWh7`Qx~Nf1VP7YZ~;tr z#=*CgV;G2SKJ1BBp#nPY>eo&N%Bf}1PKX?0b4Z^4{-}LVe8VW8=RQB90eo)8VfF!> z+NX^p?VhT*0yBB*%!%)HR(M-<+>{(7&8*~_UXb|iU$iK9Lj%zT@#VPlywk?rv$)+W zYtRmuHo3z-slq3&TQ4Uhk-q6x@k}W9T3o~JzSA(j6$fGK8eCZT+kL?|V)1Z{^Rz9drmVK^&OQ+2jkC&?>(&JSC&r;#B)y(~v5z*<^V`(NE;alPQj`-us{d(>1 zcJCa1;Ed4~l)aEEn52-&iHa(J>Ud6FB3OA!Q)lLjN@uHd}My*xXE3c z`f&+rqVjL_b?RnlpFaoVF?yeou;R4$C{>x8j&0UI+oYn;kj!g?sUG&Ju~00kqH-; zex_%0dz@;48O5X>h5fva{liBzv0BfbWshH<_q(bcRXjqED*WJly;sz)N2~0%X@S^| z{PJr$^_6bb^MLu&Yl53-_@B%{(lAN{koE#6Bwg4cF!oPcbqxH@b%!aCw;3IkBrG2h zX;yT>4~+9;w4r%+&C<6E1Itt~#AYPPzG7>+B({AnwiP@avYNgs>VA9`F6N%-QxG2! zqjOs9Js~|Bz&&Dq&4T`Rv3589En8joc;WtvO7?;y)oYS}L#^W0b2OLu2!HdAZ1YSs zS7|Kz-C|1NuUr?H7K!SwB69ZqgI+OHXcJOf?22sPE-vB`11$UGE{^nnyEM8#8E1?- zzJq*#_R_)a`;M7257PH4;K696KycpsGk<)3CO`%NH!4dGnX|Bjcof{LGf6hjWgCzv zJB$}a&>v8PF+ACPj)o&!^jF)R7Y;>LgX|ffij?{lw8Ww5WkO@_rYy4g$PV$#cAo=Q zvbf+ZI0joj*Q>A+EZB4q3`?2&g(kL)8Q)#1TCf!-EUu#nBz|{r`&S(6((d1*gD#f6 z@W;h0I`=z!@`21iYEY1jJG`QgvDz6ZS5cJSj(jUiw7;-g=mL9Awjf$}(lv~y9?%a; z5#oC6V$bJZ%u&olZQM&v<`Vz3e8pUs4ky$*wF36qDM)+PL$K*F>@4cbKjPt6PJ#St z_^RXg4w4y;bvTY2Q^99k#NZuo-;f3_@R&o80mx~8TyN2@nBWezF%p$)v28x`4JNjSjzfyVkItu&x0_h^(iv;9Q z$o7Ozdrv>Q>%d|ABy+Kes;I%6DXNxRL+eN7Yz0L#$EaJ;<~Q^J7V^KmkIr&y1=ij{ zAZq0Ke$ZRcwy1`~;`97lw|f@$Dsp_L9>9@?ZGLxUTSPzViaD`;++B0Q{{-p~wc^Qu zr|Ffnn9#WjQ!}G@x{|rzYJfov0QH78A^O~x!ute8o+C5}hagH`{LQbYYDTCDA=o2i zJoxC1WN~23!UVII|GS8LX@UVtf@CbqBVylYdR-!5bBW`%L7;O@#VPO+U7xM=lMz!y zR^zxre9HH-H=-wAV>43fUC|AkB~z8LapY+JY-mjwXwm+M2^3j4vCzeA-M9|>vf(Sx zYJc2Gv0t|GoSp47-X6CHt@dj1K>ymW=Bo{c2bU{>I?DN@MCVDJseQWXT~SIIK@nzs zmNFEL7+!D;l&cNb@ED?c7@OzMX+t9D*5A4L2O@G1^6G*P@Ji3*_tmQC%>%pS2 zDVbU%^c?xQugh5O=$14I=COE}fa7i2SJD$97`HpICi(tP`7Lu95w+KeBto%7(zEqk z)D?@TTXR%HuADN92s##xd|Eoy13#Ysvi!BaDshY2@e9i*{%sC8uev+)inacuEywpi zP)@#L{aXLfmWWpwFSWk;<+02c@7?z&zDu-CG#rhM`SLC z04-a%W1(LDNY;aTQ{(2ns=E)7kK_d=xc!Q&8uI%7^I`z%@9eAZOlOm|K7@5^PysJ? zp84Zh-b?mk26F(qce&B^CQ2e`+VrhH45iMQb2=`yR?KdF7=hpEX5 zf=&D!sD=P*ow5T|d?k||oQ-7$(OoV%{-i_PbE3PC1pdF04om&7(N$nsi=LG2E5PgD zfIU?f8-l%W%j2Qq+`oiMhqngix3gS?pFN8v-!FgVs}ujJwrCTefa4}KHzA2Bnl&W= zyOw@j@Ty%q$8P=}LZI!PY*-szc-slc38$;M=%)y=GE;DH>iYg7$;MBzzq4icEpsTL z*p-A8_+mr4Xx0W&iSWzm+~p$cxJz%|DWe=}g#R&=f^%5VBgna|7)x1dql(wD(fP9{ zXc;9!sGivBx76%3TIrDK>r?j2m~+!%UxU?615Xig%fn?xb$-{{^~g(xvImaPC#)~> zzj^5@1MUvZM)qzd%yR3y`GQf<9UCODijtl@2*4uunOv= zOB`mKgP>r#q8I(n|C)rlk$XJgZbkX&#P$n2=ijLytP0hl`Mh{kO{mPgS6fY{ev27O z!V34ER|7ay;At5J?O}fOy8Q+{bM}ukG%~( z3#BnzD{@EgBtpz>)GKle#8&!#F*~hv9qIOpF+DN(48_XKBU;k$u zY&ZC%!%#yy($OFclBJaHn?)<>V}#)%B*=!N?nuJ>@NFfLB+gq@F^z*dd9h<^XHF<3|zA45GhdsW)Ap-<5}`=>q|=6$L3H80x%_es4Ex-zGUGDN)@ zyF_~#7j2fA5N!Xb!yckk#@pCv&KEqGE8(d$hOXJXF)t^+%N}^rSOXRNuq!z~h%B?l zn0HVe(2Tt#E4cE6K_}Ws)_ke9@Isn_e@9XH+$AaoDN0(Lw2xgh{Ff31=WG| z_ogjUc}AE8CX?B8O3&QkgF-L#7qv~fh`V8cQ0XdiU>u=8G;2>Fy~BE6O1CRQz-pDx z>>Trgpv!oJJ**fjU(A=-VU>nkXUaI0$_=!Gv67$GzW2g+FCfapKR8aJ(6cHmZL%v1 z7SEFFM+jWXiR3;P&JDHmC=L;~9$i*5SI+S6d>8N0y5*(2uuZw`{HNsAAbwh0bxc`Z5GfT&w7EM>{Q zB-W=KP^faD>4%wXK7Z7#iW}FB4DumO+Xg{}8N4RhGX9coOFXPE-etjaaVQt_h!#b0 zp`6S`YfhyfW@j9wX|DJ2bd<*TMFu4vA<_M>iJue-;=8NkLS|8?D3%NLEW$+GfpiyF z!8b-7Rsm>sOL|~~;>p;{#_Lgqd-#Suyv-M|j`X0GJ^Qg)e+*?x|If@Fs z%O0^yS{L6uy0)$G2rAcPEAf_CbX3rUl*I5|9>9{+DfN(samnn?hT8!%>m#e;gq$~s zf5Q|VL<4U|i#glYN0||$yJ_w#{e;UbUcfg`S`2_#KLX~yEa8)EEO{^A6oIn3X@U@i zm?Gf3PUA@UZUkJ`jD(Lgd{Rbo;*CDyc6PID`JynB(frkqY(bUEZp7^8SGff=YQf?7 z*TWBeqQ!bWc22oJ+rJ>9#Uq=$DCY>)vsW0DXI@$LkIkBwck~Iee)1!N?-e-l^}@WP z#ooO!p1D8Y8hm6kiiqa*=RzAjJG-&ZsLp^x+Th)RI@KG?1fb*OXuODQn39;FkfTb2km0^O45uwQK9@Z zV&y7d-QX8z#J5%p7ug#)jT90)$b0s2I4slFQ;u3JG!3%5Bt{)DTcAyOADSy?y_arT z&zcJk^IkaL2qC$0ep=_v{U{;gZ0HU;h@Z8Ogn}8&kbbloU-zududjwfS&`u#ncg~~ zq-o3pF*c~-0_W@@A9Hi^>+tOqs}qgD3uR0S+d1Fovc0j=5T1i#K7!GQYZ7VCPOt$KNeBhs=A_KX%S!QW@fsWx> z7{OLWh(3V?-JkcOLeAm3KKDv#3kzu-PbG=NPawabYt}uznaB9KnF&kDo zB>>ZgMUbs;Gmtvd=S{XeX@-Xp#+Po5sN9d0yYtG@LYX@G? ztbm;|_6y-K9~RMQg}R;SUA*!;=8T=b(MFVw^;uqI@J;Q<+(4B+ou zDUbfoTH4Y=AtFpD^-|}%Uy^i?_Dp<{0kVwq^c|M!IAw_PkIR0{emhkM;>+_(3jnJV zGLS=T*0*a3Jt6ZMRGt(2Mq1W**_WQC83V-Bl8#cb6ZQ6-{YxL(FOHt{oZysDK4g3Y zHB{B$2sMuxztI$-c5zD)b5-`le(4Ko`v%jgVVi*=nl|oc)@d%~;-^GVedXlvOMN3( zd;TaVx(@YlHMrDiVbRw(TCRSZ;hZr`)|1P8nb%PH-(b{oa!H=d|0XGPV`Z_$rVhfl zl}BrpJVMedaubd>l%Hx?accw&M1IQ9#M%0%kBx|`(I4(0TEWVnJE?rJX|I8bbL*D* zCWFY-p9P!dJ=+XXsK@R)05WmVSD8DpGx0V~YSLa3FlpKIt$ZsYRnSRl4%zI$nGLh+ zQ=;;f@A{tzlcmEB+=JaRiQU3pk_URM+4w!V6=a^s&yC2aSTlMVrcYS!bL5ylz-@$6 z+p@=H0KlOG6Z`nz4G|T8nf{5#akh8wfxICWqa!(QDeC8lsfWeER9&uMXlf;T;971C zJ1aJK(o+w7>_|@7ZwP1(fSpX%g}A-5_|J{=yh0f+w{j>MTpMr{W^~B-d8Wr!e`79m zioWcABR#w%@6)Q%kjk@IE_u0{ZlUP%M{JP>#$~Rrp-;wW;>+`C#G9?l*E?OEO9mx{+s`+`j8Un;d82ccFt4cg_!OB_%H# zs$JYsRDSG#$LUkT!QH!p>~_r0Jt@gswQ{se)1I*rExVmo@c|-p@>lHEU-0G)k#Ii@ zZ~@CNI!O)_!M>Gqb8mpn`dznUpt{}}2|?f7*86YtX4xm}+-$vNeGg_|dynwJ1}@_J z^S|G7ZkS;;wqm2rqG<^P0F6m9pAMJl#MvJTHedxIJxhZ8W32 z%dQIAkaVduytCoo4_>0fdM=0gkV99{c>}+Kp?L#mQdW8i3QsVO#FfC6(-*|9N+ym| zgr|C#zP~pKG3a=ywlXJlW?A$ZgFz+#IsYWb6&5!`)gd*RDF)*&gst? zQ;F!CSNPrKtd6J_MW7PgrJn)*!H?zP-U8F0f(&L7>#IHeP6hIWFxid0?rl4N0O9t1 zsbC1wj{j=u-3*D$T=p`;te6sGVx+`Qp=;gtDeJ@!ENV<5x$r18C2|E8p8=O1C|Xer z`XVvSonE+BSTfDJCb)P*s+?n+pRRU^RI;Wh9b|AegZ9C?k5wqFx>oqgX?ZcdUG%<| z_Q>!^OTaeio7oXfiCF&6yd5MN!`ud2`}hs1muHK2_D%&c9I=*qtfQii!# zB7>l>rw{2{xHzI!Aj+@x@)1dm(+c;FPeNd>Z2fA(J1bF&|CPqMznne}4fAFvO!hhP z35la0#~iB)FEO9qZ*X}Z=i*8<;~)7UbaG=?93Z)|j1^Yg!6TXtC5dhV-%iA>X`>+G zgOw;q9Bii9X*cHQNi>!-ww1wiWo^LO%+j9YodhK3FM;xgU}PvQA7hyBVn~l!-a02b zOp35-bD6LwSpsJ!t(P~=w0Zqv<0_4_XPL3I@pBO~gI4nI`@LwD z6N^RgRoi#`<$NAuoVSusYu4S3xlw;ubd*0Eb5T6=SV9z58Pps@{U$G_Ep5#zUO&Sp z9OZkQq@{IY5a^3&$^D8=%nDgYS3&JHgnBObc3ddYjAnjf>bi;K0q^ma?h!pe|_zoeQTmMp@GyyYzi)G6COp zBjd!jE7>__(jH?(;k(BrD5EKo3!78rk}<8`X+EA5>9i)JyCyq-I5(I2H(a5UmRKf4 z@qrkAF;#ys6$PdhC;p@~jJ4R7@t?c548>eMJ{Z;3@b$p^=3SBfg9ye*S)zCNwB`RO z-nxBuj-C(Wk|DLXC}Et&l5Ms#gk9*rG+k%Cy@nbs7m?}Hw+WZ3@KW^^vi%_o)U%fj z-f;bX2)hF(AujIhf8m`ht%L=&5|sBq8xoeQKcOYkPM5WibC?Q}6zK4)inz2{)$<7_ zQW|~KCIQR#UF65-{;=dZtonqjZ6n;QAl(w|1-t@Tk~2i&!S+d5rSH^pSco{V^Swwt zvVci&w!rsJv2Qr=kyzP<(%)0|u0$u^Gj8(MC`6t7chh|>|Ndc?w`j~C$z^ypFS@hc z5204-Qlw&~LjSOaK*Ed}MIv8sZi>N^rTDOBY40Ol_p|mZTgfR;$o>OWQgfX>o8nS5 z?SWN^R?iDeP|^+C6!G0$%XcmqaghuAxJ+ z)=dzeyGaQv>ZK`&qR!!5OcK9Mw!@+2;$RB7&y!CW8jQ-YZ!uc?%l&t0BD_YA7e|zv zW(AMM->%?JCuLhW_*|JZ*a|>cN)B}-hw;0Y;<9}Lcpy|({S7P<#tneKCx4BYnAFc)wMZ&#gjOO z4bEN1AE^&1o}1W%fp2ra4T;?VNNn|>du>OuiN@w*JL`5~LB`B}AnLkcl1CGIorbK2 z+gBf4$d+0GOw7$@3%2O)TJdAnVh}NhSs(wfqP%%cvhOiw4_`gge$(p zf;7bgsAJ+x*~=c^tTlS@|spSz7g8GiDs-+^sSfm ziByVUu9gc4Uqc@{3CgmUWDiUW+BVR3R$$8I51ZV@oQNS~N$3#_;_>ro!C{TyLWjaIU_k=#1Kp99Ei~9i3t7s9V&v;Pv|JyW z2!66txthEaGgr*5T}8DE0*=+;T!a%LYPyq=2ofQ3Y~_V!S7}dN zq`jZ#?}Ud!BaR$1;itOa2O4Ds?tng7pCWLjx%q6?&5eN{-8y8&C0p0PjR>$2iZ4{9Xcj&T=H)^=;Ky?3_T|o?} z!1|Nw^>vN>ns?ND#Rn3 z;>juv$?h|IK0@&u!SEE^YqHwl->-xYqZys8u%wtUdHA7$<9`O8ik+AyHG0{WE)WoX zG)Z1E=q_?%O5K0Y=(YJ`Gx6YOenl|hiVxx_T?a&2^oTwn;VtT9unF~iF)!kJ4cKg2 zaN16~atTuuCA;E;`6dCgs@In!B)RN0Lu}eg?*|DFtih#ymdgFm1*@~$eZ-ShlK<|< z@}GMOWZN*FL6TgQkJAg|0&=!}=*siL0AaH_A>L!KXFZb^b*f*6+nfrRm- z+k?J#2<-Zlw-UP7Bt+Rmf>aRFxcof*Nx1$HN>SF8zTNCe#8I?P`eu+#CaISa*8~~T zmHenXmoLfQ{31!(6WMhalY)$Gk`(ON+hWj=65}c{Uxfvo6zZO~#4l?V0@q(Fou$&h z1hgi;(fh3fc4U+R^wj$*C~8>}Pn-bjJUMhFE;HYXCxtF(5L8$pMnY!f)(y zQd|Hq5anYS>my8O_II2VyXKTPc14Onfkx^xYSG8xR9rT(`)$qKRD98h`beOpj+ZIC z^iD5#S_*NKwD3tXB5*}pd1RHhpoWtK^Sn{2@eAcCy3IHgrotP<2m}PHw867Bp z6?8eyL0qc7#CoWkjaxoO^lYO;&G;Bj_`jLvosxGQ#uRN??`DTiH+%0z%z3trxfVGG zHn@XdaP$d%h5zy5&`!x49p(D%W3O*QLaReLVQ&%4D))1+`uL0z&IIK8zqn{=2qx@= zHAP60Z}27h8)g+vJDJIm^UQZ+UYJecl zuS3_Tl{mOsmP_%p-#)qY>EqSUTvp}ju-)%|?BClIF_>&*osPH2LNo&4t2-9tadsKy z(30vUt78+VGdtbZNtp?3oPFtG&x|&!iD{vM6>PTS$`^|(Uh}zkW30WDA z{!)8c-k^2w`zF)fk}EDLG~a*6L(fpp6r%g1O3*Q1ov9aR8fFN+ab2?AEF$w~HH#3b zZc?)a=}OJSN%5^BzfvaTs7^|S%~oa14s}A7b8TT&GZAO%sZnov;ZK&JW`*4U9dYvf z8pNB<7b6jMI211xNkdHz#n5WGqmHadz2v|&g8>O{~?C#=yESw+m-Yk%M|#mH1_O`Fwm$43^*AUlUyafkF- z0Sp;iC&ero7|;XvNx_TPFP}5p($9$aMx!n|BGZvympyBfRzhobCT_b%5)RZM2ph#y z_gv94W}9n#4c{!8=O#m9mCQkc+7T-G>F`dh8~|_{#oR;G>ODr(sZ4BALo560MvwnB z)+CphoDDPC69?GJ;6o~ov{<8mt5zuG^zSD9(~wk&Rc$v0U?D$^gXl|_t&)%z%8xHI zDmK`L{be{9y0-%zaC~i3`E3z7dX=|9^iJHZX4WeIr{H$LT=lCUEdwpusz-o9V8+CK zWw2cplHQUfh~1k4?LSn3H1!!msXn^#`rT|d)(-DQTge-}SupX7bX@>ZJO!oOM0obO zf8MiS3LF?aT>z7*(8PsN+{{^^Vz6jp8e=s=ZEk%94SS7o5E&2*`f&k6?#Y_2Ji+z_ zxB62lBC)3}qTgret$^RiE9JF55cfoB5itEO`s+@jnm$EXBJ$Sb+z9R$(gJ{}HNNT3fv0(plD*>8w8IX}yr)zmw*D&hReI{D#+}eJ z5Vq_dmHYFvyKS|kZ8ov0d3+WzE<-r~BD)@|t^cfLr?&$PL1sJ@$z5+Y11SM(tDlSn znz0>m`&v*8Owy=cVfdKBlBAj$u(Zz&lQmBEF;~Oo91`|113UR}hfxe_k+f_)z$%PK zg@_iRpM-cEwLMY@|?d4!K@;*Me3(XqZ8*| z4q_$uga&Z{OMG&?LM8as2%szTa((cIa z{Y`1{tWC!4;HPv7Mubn$4)!q>M^Mjum{_?!-dyhSeOfKx?15`G&-PCX+Y(w_&eko- zgYFf5-krEXcWJH0aoOybXJvibc0n#}+tdT`z@1XkA%8={qro~+T$4|<@vHJ9Pv{`o zC;Ayp_2Od;@>p&b*4OjacEK)ww-$xH2QV zPD(bx4vOYtP8>Ra5`T03_KN>t9m?Dn+X@AX$tDjUs|UnnIHe2?*i**hgqKF3Y;>{^ zCzR05e!Z@=O*_RP2&LRkOQ&-fKuSxdX|{U{#dqM=NxUWPXpneW`@mv-NQ6aUyI!9aJ<)GY58%arlxW=Btd$3th$^OD%hz#LQp({8^!Zt}f(dqcX*k~9uv zMUQK|37?7B_-SbiAf7%2p-bbSN}q8#&WT5N3 z(ut`6CL4g+bt?Cktc7B>pVCtRfbvrWa};AuKs=0y-Az8VE0iL|QLHvTN0Og<(Q00f zo`R`(#3}(hQ6%JUXfptX>N*>eT2NN@qp z(lejA7l@%CZUFb`pS{fygNz3AafJ-Y6^hB(f=Izn^2XWH->`JxlzAuA_sgYe5<>cw zJd@c)g8#UUhs8T5O)F&Td!~4!M8S6moe+DA(u1!*RwOhKCCa5^cJA4(9hNuP?mUaz zYzj}0M9t{AoL-pJiU9xlEKH~;-9Cl_8*!q2p`?e#J%B3 zBA?t1ZB=@4gt~+{u^(6~tSq>y`q{0Ut~+>^6~b5QjYoAOt0f-WdXs?&7doy=Y#t+J z{j58ohrx@axZu7tE9!LG)cM@kDz!NAoZ*z(J51Wml?r*Uo@!tEy!z?N+tG68ef2?G zAo2!S5kp@ZX~%9!puwAU-48k}D+CqKjX~#9fV>f>Sj4I+5 zB%m_`(d9%Ntp&u$X@cFPCkoxo8BHQA!W|{hdxz3WK?;?}B&b}US196(nQ3caOW}pd zgVW*3KV2k^U1EpF|46~rCy`4V0SM*Z+8@=%yR{9o`uO3aw$kzaf0#ZFRP|*k00zIZW;Xj^~UA6&AmO4U1E)~X28Q!7CS!vSK6|NQ^~K} zC9%Bs3_K|B2^L<7=|@D|qHe{VD<`#-HygHWi>$uOYJoPSs97iDTx~*2 zSHe?&BBVjk^F$hb^1+hUyF|}N!+WGb`a7`MoBK);g4;+`G$@;#7I1(hwt(;{+nJvUUlSI1Pnhz*T%{b_~iw}gP^3qVcX3Oks_W%Z!1 z>2e*0TXI+iQ{J~dV&xpftmm%~M;8o>gw=@A(rJ=;fs+6iMBC`La1(M3p~fxZ9*O*l zyAln0JSixps}}80>4>#W)uGM%1)DDyab+8J&fDxl$z1hQ`#hfOF*dKpR#>3c;74&% z>k~cTfK%XCiqhVbQq6kh6g@;b5^bZ5+adV*I%9}Tfwa5C&F!{13knV$I!~koexuN4 zFzPX0I?8j)Q$>`Fn9k(;GPoaZp`LId6DrZcrRBE^h67JB_FyQx-vqx`iT-NHmnJ{2 z0IeouSOb{m*a>5xvXHR!?ZDh&cF4?jK{9UB5xRD_5*S2aboDHdlM`%s2R!>T*V;_r z)B!}m#JJ6ZTeVJ5>@4n#x}Xe;}caW+$FbFU-c z<(50fS37XO9u_$cx)cj5@tg4zenmuuELf}Nnmk;0vc}>Bafs2nvtFM&B)j9J4d_XZ zX{D#Y<%EsK6WhNZ^^^!s?6+Fd!%W@SEt&F6VJ#99DW4_9y>DgN%b+T-joCEir zrwiW!&$t!qsUl%qiMq7Z$==$KEow1!af|?lhLN9uO5A_$G#UN?u}Uac6kr zxUyIs=bLoYfoMZM7CoYzh6u1!VYJc8V^(8q>vGKRwmL=56RNj;`WoQy8Hf>JeUe>l-U3HxIO% zL1vFftD#u>`X5>>zl#R6cm2ITDOdj2uO}~QjI|!h=nX|1>Z@Xt9KnIH*ooJ_k9wDT zzaLe^XnV>?`_@a#p(ONDzVK9_t(r48>pkbshewvi0tmb|=Z5>)uypEoa7>zn2x>TW%&@&aCI|y&qEn~a_BdAqk#UBryz&sOV zIn?4ljeqb@^WI(+bC1OT_dS47B=DK|>o#pE$!*+_^4GvtNMe|C`hW>f3r6-Q{h za)t0^>C$k5Ou|~FLF++rHr#7{;SrkfCkb&34-77S6yw!TW#}0AWO|&^pP;3yU{NLB zM(OUM?GXH>2ZFIue(g`xKO~%d__6QPM&{BV@O!I$DsTz}?JC>iw-J-8 zJ#!Ws_tO-sGjVXFS!zY`34IUUvudLN0QVa7`fUx%FdDG$aTk&pJyDdl0>H?h9npxlLQR_LlUh6xKF9nS)$>X`nSfXD+yE;Cv|-(u&dGFlu{AOU{D2IL zoMzwYX+Fs=Up0N1_ithwiJ?TeOeKqDY+py{3p;o3QlLy|6+ZOg*FJ{cMFnJ|c4J;g z>>PIczJ25<=f3G>l2L(H2}Wr`CGz!uxoN<10?&JrRv-MHsff^4l;lp=j5lTwdN4{%fkjHaTW#8|Fj6t}-Cf zf-uY{<%#wMPE(I};~j~)z-!?EqLs=awMlWpLZ2o%@0Nq&6WL3Qp%w`fTyyEfHGQ9( zul7ypqmg?x?zj8@3e9X6D<{Ngm3Tj=MZFk(U0-#@E86z_(90v=_S~2?t3Kxzbo}+b zd+JX#%gVIwsvb{1we88Va|gKOdY@r0n+KlLsu_lq|)DN45brsc50?G1Vc zU^hN41}P-GAAVc8OwnT`Mca8%JjnQeLlkLv`V%cmM-9Iwh&8F0^@Ax~O}s9Y7qz^% zx^y(NI&=_wMZb45vpV>?_j8gbsl-uY$-9z9*`9uUub(2seyFtesNe-8aPj_qt%m+n zxkd|=+JMo%t;~lfEV6Y-;`~G8i{moNWtGwP-Ou9LEz!65vTgwj+qvoc)AIjjcBi#% z9?mymejJ<7yU}>w3;*$}%_AxAaigafLM!im#=Sa`?O967g!u^?vk-xBGE`~V0DLPO z1|DYkYnNYWef5F7yK>nP{QCi8A4RDK5>a7jDNy28CTjB(3yC(w%vZ$ z;^=Aed&3KX1@*gz;$v(-N6;8^3Tqjf*FR>#Wfk7bdUnVhTsANaOU+)(_$(eyUC`WH z(fw??np)l+r5j;CA8-G0A*NBoyGFg^0uXFZo}Q4HQ)@b+s+mY?L~B&FpbJMcqJ}2R zHu&CRY97lRl|*}s>=tUdaYKwN)-jV{gI?p;8h>S> z-cfRB3_s>yBrZGe3!|8lE^07Yj_5#-ZMqN&Nl~fdKtziyZ3fJdgmuI>(ximmr+if# z(LS?+BQ)9QF)n+6*+66QlHrbB_RPs4!Tdd@frD4gd770E{l=QmR(bJ1Io)(#E|GP> z^E-L^3cDgaEu)0x@VvBR+K)cOvgD%s0UG-T8oMH}|H|U&bft#V-`B$rdwGry7#aS6 zw(G&D2exw)G8t4@E31H;aQ?}H&B2P)8+yeZOrAfU_6g#{n23TOK(p>9j4At&N`sx> zWAgn*&v{`7kRoq9VHFq!FwLdT{$=!Rx z=ncB5zjS|c!j=kXkgCypSbS>S^~p`t*p5xioU8lOqppu{xS`*tFuY-)tD!{;I;n;W zPxcxq#(7=3s{d2|d}UgUwF8{8dYh|2^nJjw$;oR0X>iH*RU8=OI>`Aj*YXv0s|Now z5^6Y>z1W!;3nMU>j*MmpE|Pic;@)Giu=eF@R{3sT^7OOtRaEX|V{V7uv>P?trw>PUf{2u5pAGo@ycohbGIp7 zRf5w2yn+uUAw({;Zx9uPEk}psBfA;Z9B!S8cf&)&e%N#sbyt$reQm{vy$&FJC&tUO5`4#BZZ42yyErYN&q~64a8=~M$8mfHR z7JSfRRESO^Wuf6E2sf~o9iU{D2Ys?YDjP|euC!g9q-0@@N-)!6Erq8=t38SCIyxR*s%E;H>o{G-r zzFeHc)QWVxSoK{tWBLszyvVje3V%6Okr1L6;go9%3d8FYI$p}cImeMg>o;yWrH%Jc z4Zcg5Is>=9EkQL=P1r%G@AIg}`m!HlG-rE-{dL;{=J@c;mO>I$fWgnIp5rt*tGZxM{X#Ktmqj%zM#ZO#Nx$F9c zOWKX=!*MmwN`t6J)WkE6==Ua>XXeNMN(T#Fu(M$}{a5Wk=!BbLk^PzW!?gpf!}Yt)B*=UaKfKYkcV^m_{3W+`VbgTQy?lAmyTLb$IuQLi z8S7@R2C}U8xI82?yPkirGTR}sMbThcx&#|03b@cJ6Fj-YOyc-opwZ0d`k$#fzgVYJ5*i^T^ z3kew_lkY`uP9>-f54Ia{-AbaY`p&qiNnY8KcD$*)qyFAs*LJWS3LW3*xt8KJ60l}E zvW&M#JNW|rdKhy*do2YyGOiI*OTN&SsYX5!eDk|$_JHJBc=uXTzN%=R(Ep1|2r=qd zws9e`e^&jD^`TR&anomDuh$=Wnrs~bav~44s`#xmP7w>W` z7@*qEEm$AK|0B}1Zj(NtWqeOZm(?ajuJ$g!sTz*;G@yTY-JSKk2zocDZ+^zmAEUKU zYg2gjdjv#1_j>4lwh7N}x-u;vITRYc8boEOK-}l5#N8!Ou}4V5wx<;tD&oKX+X8{l zJ+1A~4?bGuxFp{Q7W+ZF0~gna8$O6n*H5Rhwdh8L7Ay2o&VKv z_pOD&HIiGJRbU;!JQnXy#M3=yXsy~{P36C`$oA+P3D?)CZFB6?y9S%IEOZNqGUg(d zq#hv?XvPS`*?oS4Et5!@eeFXr7Mi%N+!f-r4JB9HSQ#1`ui$Qgko6D~n#Z zES(H%hnH@CW07t7B4DwUux{tjom>3Sz9%UZYT_Kl!Qm^E#RX~eJ9NR4^w zy4phi0s3!?Z2i8=*Qo1z7V@<*u6tLyLWtieHZ^%`FNVz~v~snkl^#z2Y+zVX7`~wR9rJwHY+R;XxS@JOa6D~s{WY=mVw-w!NExYJJ6N(u#2jqh z5T8lAOzKzuMLf}!1>2%|)xcAWykpw+Q+GWZI=RRoEPSCv1S|CgiJqbJH9NkF60Lkx z$_G0(C@srCwhw)pu`1jpDF`t*qetHYX++G-(T_7U@R*0g1AEJ06PCo!m^;fx*U0zd z^e7!?TwxB((u)#*D^}yh=waE!=5h0w1?|1QvI()t0f{a8h096Jh>4Dc;x}xk5wUhi zcy`yLVcLl^E4c*G#l3#oSkkuzSin+lWu{y=?5@e_0PO`r(0r|QRv`Jqr_olq?yz*f zY4v8)41bN>+6fzQrlYJS-A2rITi&~J-A|Vi>rre#FT1#(rAUkxpP9y2a+L(K*BDGI znihLCA3x@4yjXsREV|6Qm@b0rz3hOJrKqim3sFA?;|yz+AW%a*v#1iJU^jLYCTkYT zva2bn9n8i@(x1dRZ}dXV;`{Njh7Ah|t&0vQmup!bJcBq5HBx7_@&44H_r-5k%i$#Xfn^QDu*_cw z`dq|?Hu-@{kBF#z4SHfR8GG}EJbNjh@`zw%57YJR9T$RoD4{R5KQ0r`@$06P``|yJ zoptJ!baCG-<57$GJ@_bW!X~`MeI2B-$cAd30hy;mg1J>B-*IYvxN6z#z^E7s^;@*6 zjqU=rI9&hoJxRScu=|DeBV6?p`?K66jwPsBVRycbSR6_fty=nQ^Wgq1i1uAhq}7Y)X}%K|2v$&Y8b`=_oP&4 zScZE;`=6Uowqdb;8Ee6l9gSA`X5FHzN8g5Bmc~mS=mw^}y**qwDW4tNqPaz);%V4? z;OVjkUR+lGt>jHd96GjfBxe$00j$PC4LfA z&FsgJ7_ns0b;@$;f(8ZOQJBCCq=+7((Uisx7O{unUBZ$>WkEvM3wsVo@&qr`FeAUn z$iDd?Sip=6a1rgLONA`o(IJH&U?O80^cR=q4x#Fv*KDUz zacAe=0O|7B<{s=#I`Umr=<=Pwz88`zT}DzfjQ+VduU&6C{FXyQ>uM$w+j_{%EvEu~ zjjlOp=C2+ll4y`s@nZ@@Hw<(d&EoYB{DLuYo^niU$1~|%s1JQJtHauGb@|(Oqi=SN(}FMj8UE{;(jV-f%vY66s4vt#ae+~! zEjb=DS{i#lrtHueZ;`?qBKZcxHAmzWf{`|S6C${k&p|{Bz&`re5jBUeU6)M+$BCnu z{{Hm+*=Mc?C=0Ax_K*3`;_H_847TdNplcgrhWly7C$c4}P@gq?J^$G0X?2G)_t~H0 z#wu>G8v+t`5xF)|Bx+aS9i93tO0pK7O{}OdBJUeYqZAw3oSYW;A-p9HwBgj2~k$~ zbyyg8Ne6PCR*b;f;MX@s?X037nC0D4U>g$R-iKvBlcgH)$=zDP!YtlrfU>20nt_Dm z$ASG~vSDEI>5Z<3+`#?9?2)LuzItkT&3gwbvTlGw9EA6Rti)T(xXp@SZ!Vs`YeNyM z>heyLz%Jir?F?S&kXg<6g&-UF&=rH8BTvvk|dxE0whWp%;`Ds0XCg-XWDHx2?-&)8#UmJoz+pdmnmJ83TV0~NQW z5)Yp#i2i`>!39C<=us8-jrtyAn)*G%(-QjZL`YBCnpO8ME+`xuF4YruUkyvqVb;8q z7!6tH{~$%tDa#v-yr{TmA2Dg?4Ohz$KZ#_cX~}4PHtDKF=>wGRMtX(HcKj|$$X@i2>TlwV$wjuhu*2lW`Eode-uVEtT*E(P` zKx-jO1?MMt(mM7#FJ;^|ycw0hV{3hU9IS1 z-~8L&^*l$^GO2#~)P;_=8H=)Luc0^73n(qtU@s>s$rmXB;0l~<;I!SUK}(n{B+Tq!Y~d>Z(J3UN~v830*wY_ znF3B=8Wq|3;|y=q#+<>!_+v!oD6eWX>^4*>EH!_X?CoTrV14+^JRSEs<%-^55C9I69d_rJHhBIK-~9jn-J$FU)>g81_kR9s973LsnEs3>w){Gl*=6-@ z`34d485FD9-OC~ytQ^CS!EWMiG_3r_AN9{*cNV>AiPEBP%&l!!J3$=D3`;%v-B-8X zy2IjWxG5) zDr>&(ZkWWo*xr%*V(&HTU_>o_?Tvk|)A}z|UBGbRC(X~CL-)flL6L=MN?ix27_U$o zyS^0)C=@4u_Da0+wRtG^ChYnDoo*~eXYh=p?9Y_1w*{0%hcizZW%tmW(DjiIhlrWw zzrsf}SIaTIjM(Po8{TJ<8Hu;RlLDXG0BE#c*WI|3x9sv7-f3)n#24(HF}% zuaWy3X4l+05R>xQu;e3Ix-FYeELXF=5{J5Kg&y4pp(R&EX`I{maHCE8IZ+|?ck<<; zoJPj`sf13g&lUHrto=r`gReIf_0i({1p2&(;oW6-G1W*DUqi5}P)*w6ax0S{?*r7x zWpTaBmon&(!YbK*G5$v7wRHZTb>;5Z(}IM@LfduwkYkeW>t z$c8qXn!N!n1p=wri@gB>1p?6sgbD;wX`g!TzcKLPIaR+}y1KfmZr}U9=zZ%Yci-!- z8_qiW7q0rA>;C;Ce)){I{f*l{{&(lMmmm1v^~$IJyQ_ZT+~2+Ku@`>(^OyYQ<)44T z|9jOpAOE8tc+rzy^WjVH{+*Y+?-y@=-F5%@XCC&Jzy9r4eex@py}NzlSDx^wYyaMd z-twDweddavf93nHIroy^{`==W?%RC<|6MnH z=OMrTxko?wzu)*rKYi65KYjI`ADQ3!pRWCp+dux9Kl;S=|M9v{Kk2t$d*7db`cuC1 zx2}Kkvw!y2ervdSdD3Iw`>{8D_|pCtt?Pfd_Lu&z$Cuk)@zy8a|B*NU(mg+M`^W#_ zzJGAe{U25?zV1^GdBKB!|Iy!m)g%7(MZb0J55IXi=K){;;{W)PeC{O={rgXu{?V0x zc>h26XO}bARHp zs~*~a6a0nu_dk1jgtv>I`r0km{Hr(o#|QrAPu+Z5y!4T0eeJ{dx$y}fJ%0Zee($rU z{$u-BF24TefAEIK-)A}deviBK4Oc$(mmmMq&;IJqU;BNJyY^9cKIVq)!H>S>5x+J4 zqf7tr$|vnl{`X({!@qprPd@H9?)T&ezxV8a^H1M?(_O##eSh@epLx-*-0SBaaP{T? z>y>}_>~E^&LO`jo*9E z!`?if@v}Gmg$tK6`p>M_pMT54ulTKJ-2Bo1`L{oG<`aMY@fSbw0sZuT!MA_#2~xj* z!*BeD&tLMCXFu}4U;P8m`hULhOBem_AAI^@|KQ0tyyE!PORxF$CtUK-zxR~!A6eDhzv`~i>r*)ROjzrJ;S z*MH0Z_Cd>;_kPh!|I@#?=hnwv_H)-={>j@Oe_VOhPh9cIZ{B(PZ-3*q-#GKhfBaML zz4ib2;(qmi_}G8G=f1!Hi7T&q;(fmF=;y_w-*egjx&E^|e(?hr-}nR9J+nRKc@KKW zt&jPc>8kI4?5jU}#m_(SS6=+<_c`aq*F50n^^q^U==$wp{ki|x6<5COjBj0j&O;uQ zulUuMz4a@1{KcPq*rQ)?)AxMrwtK$u>mU1fzy8eI|MtIFe(4eSf8-_CeCK{Yxjp2! z-}rk!cgtO8J@_jhzUp3&ec9Rfx%a&v|G4kH^PYRY?~ONo|94AUzUmlbm@ppc3f5d&h@{$|B_kj<+_^zK_AM@s) zd%_jJdgpC7-RI_;f9x6ed(~$?@^}CFXaD+RU;T~#A9i`$LmvC;4_<$MeDE_L`o^1n z;c<`s`lXM)|J(Qef_3NZk9*>s5B#s`-XFZ;UH|qU-uBR+`NzL};T_jram}y&-)}$r zmb-rQj?Z2Ct6#t44Zm?l{k6~B^dB#I)gN5?r4K#j%ImIp)0gXQH~+;yf8iH?aQXTl zq(?vLv;X+1w?68xT(kbu58U(pf9}uSbI;Gc{LH`l=f3;D@6Nz?XW+Xt@ZA~s?hJf) z2LAsy1NZvz@SY$0S3f>r#Ds_$3s$73{rJE7@eLU}3Jz2>gr|1~jF=EHW5J4q4H-KM z4pcORXLQodeFPgab`%__Xb9i#3>YyXV#b2?6kB?yGZw5!*pRWK;6Oz~cwJ|}hzSuh z7OY6vkg=oSKt)4%eP_Uk2@x|EtVr09v7_KXMMHQ)XTXRF5i=I7NZ63Equ@YALwI9n zz=#Ru+|GfDhHzeIz=#PEGZw5!*pRWK;6Oz~7&`+-Oo*7VAbg}h&;cVRM9f&QBB9_w zMML;#XTXRF5#`yv`vVmXVdxAPF(G2cf)xoHGIkUksAvdhbOwx=5HX)p*8PkH>uDQc zL&lDR0~HP7nVkV6CPd6wup(hY#*Tty4mM=$rzydKiiYs9PQ;7_D-t$j>?jBy?+h3* zA!5dY73qYD*pRWK;6Oz~II}Zg#Ds_$3sxj-$kE4pcORPjm*1m=KY$ zA!A3ufr^Ik$xg(K1uGIZWb7yiduPCi2@x|EtVk$0P|*-R)fq5iLOk)Mo|>^>MZ$)R z9R&x%Iel^ljF=EHW5J4q4H-KM4pcM*pIZY)Oo*7VU`4`qN_iF9kg=oSKt)6Pc4tGz zj)DUf4dFYT0V5_v%vi7@nXnBRI|>d|G=xugB4#XDk+30SM?tj4XDnDxOCB3Cb`%__ zXb5lZ3>YyXV#b0M2^%tY6db5%2yg2Q7%?GY#)1_I8!~nj9H?jrZ|@8kF(G2cf)xoH zGIkUksAve6bq0)>5S}H55fdV2ELf4SA!A3ufr^IURb#-22@x|EtVoD=_a``G!HR?p z89NFNR5S#K)_@TcB4#WIPwm$jFk(W)j0GzaHe~E5I8f0Ld_fp6VnW1>1uMdr`^5*0 zm=G~z!HR?p1qUh`!u6d2BPK-5NEh}CZOGVBaG;_gT+|sbVnW1>1uGIZWb7z7P|*;k z&VUgUB4#XDk+30SN5O%LhVcB(fDscSW-M5dupwhd!GVf~@Pf{O5fdV2ELf4efN#jy zQE;H5A)MbCFk(W)j0GzaHe~E5I8f0Lp3@mHVnRfCZ-1%-Mofs9v0z2|R%b)Tj)DUf z4dLd_fDscSW-J(P{F}WL5%rxt)-YVv88IOuU*4I@h$A!5Ob zgbf)x4pcORcXtMiNYCqR*in!aY}iq7pq^5yRD=sV17`QTAYt>OnNe_{q9K^h0TUu- zY}#g&6K1F)T+ta3aVV__-{=e&5iw&y!iJ230~O(>&VUgSGZrLlf7%yPM#cE@&V+~= zD>4pLG>ot4Oqh|equ@YA!{W9QG72ifm7RE+ViX*R7j3$mSY2aepsQ^2wf`)K(Mofq&^gLt1ilW(ZYKw;Oj?RD)6N)h(sHf2* ze6=%R#Ds_$%PFhvf;cI_hK!1a>G_?A1uGIZ9H-j)MLsU7uyE9;x9i=A37xpSM z7OaR079?a8RD`QLBO(?gWE50{YdRw$79?a8RD`WFB7D0uU_umIkdRSO5#G}o5wRd4 zqo5+Zw=*JQK|)4BMR;FlM8txGjDm{r{?3Sq1qm4i72yM&5fKX#G72if2RkDo79?a8 zRD=(8Mno(~r>wb}u_9r^eoA?u83h#$!;3m2CPXYq*svqJMnOf$oe_(Lw0yO*qPW_D zhVY%vfC-1LD#D9914c}ku^_5EV@1M-9R&v(qG5zvIs+ywVhI@q72(635fKX#G72if zM>-=S79?a8RIF}4o%Ys8duqUl3E7nFr+Gm%wD~kXtVr09v7_KXL%6jwU_!)<6$u-5 z6dY&>tutUkG^z!gF;}B%r=cmGu_9r^j)DUX;kM3z2@x|^By8AGaG)XF-Wf0<9sT7l zqo5*uwKF1OK|)4BMYy3eB4R;8MnOgRT4zMWf`p8MitzQ$h=>IV83h&L#?FX{1qm4i z72zA55fKX#G72ifO`Q=D3lcI4Dnjjyh**%2QBV=S*%=YBAR(imB7CbeB4R;8MnOfm zxicbSK|)4BMfi4SM8txGjDm{roz94e1qm4i72%f7h=>IV83h&L*3O8C1qm4i6`^%T zL@Y?iD5waxbw)%iNXRIt2)B1eL@Y?iD5wZ`bVfuhNXRIt2zPcyL@Y?iD5wZ`bw)%i zNXRIt2zPfzL@Y?iD5wbcbVfuhNXRIt2;nDsJwz->$S9}?Pwk9|SdfrWP!XQi84cW5;6)Z!m~RgA{HcM z6jX$vGa_O^LPkMFIHNNnVnISiK}9&TGa_O^LPkMFIIA-vqW?cldOc(mRD`oTBO(?g zWE50{b2=j;79?a8RD^RoBO(?gWE50{^Ex9U79?a8RD`iJB4R;8MnOe5zcV6YK|)4B zMR-nUM8txGjDm`AL1#q7f`p8Mityadh=>IV83h&Ld7Tjv3lcI4D#C@G5fKX#G72if zMV%253lcI4D#Fwm5wRd4qo5)@zcV6YK|)4BMR-AHM8txGjDm{r!p?|@1qm4i72!pl z5fKX#G72ifi#sDC79?a8RD_E=BO(?gWE50{mvlx%EJ(;Gs0gt$B4R;8MnOe*X=g;l zf`p8Mitw_|h=>IV83h&LlFo>T1qm4i72!`hBO(?gWE50{KkbZ&SdfrWP!V3<84?oMEU68OL zW5YzC#)1_I89NFNG=x9t3>XnHW5J3I89NFp8lpR&u_F9wj}4d*kxr9?9R&y43BMGD zuXRR5%vg}HonpFLPW@p+xTp6rU_!)<6$u-5WG!|as3-~!G=$J&112<8!pl1&CPd6w zkuZ6n5ertNQ%o;2R%|L~6db5%7+%pCkxrGdqu@Zpq)Nns_0$3zb`*rCb_Vo+`?Gf- zR!;@90z?HfRwQiLQE;FkJgqZeLd1-O4Le3XPnfY_MMB1oqVj=;@bu1r2@x|^By8AG zba0>{{F#`70}bIBodMG+rOJqi84Ffy$VjIj7En+zex$#iM=VIlD5wY@?Tm<6kdRSO z5kA%#5wRd4qo5*uyfY$VK|)4BMfgN#M8txGjDm{r$4AR(imB7D3vB4R;8M!|uKhVY5bfDse&Wt}KBW5J4q4dV@+2@$&%S;3Bi z0~PHQbG{s?Xb4~L3>YyXV#b2-$=<<$5fc`dT#-&A5Svd0SdmYoLP14C_*y4gtTPs@ zNSbY@w%Ac{prRoe)rO261qUh`!i}8)Ba&-u*imqxA?&^9`I7#mHyjxLq%&i~fx(|p z&Ir%!%?3;ubr7*2ow5pMBy8Aizk-T}@XF4B2@x|EBy7moaiF3hys9%|Ld1*}2^%sB zj#J8j4m4D;hT+wn5fdU7tVr0fqu@YAcui-(hzT^M--5I)<9 zt~O&u!iJ3S{?332GZw5!*s$Y3MML;NXTXFR3sxj-*m0nuA$+hiU_!)<1uGIZ>?k-; z(U4CY_OiaQ7bIj9RD{brBO(?gWE50{D>@@079?a8RD>%#BO(?gWE50{)EN=6pcrmN z_*7>^#Dav3f&-&-cRC%usA3KA+5PrsELagf-3taJ*$o*x(hK?sHe~E5I8f0LUf3Bh zVnW1>1uGIZWb7z7P|*-x)EO{hLd1*(D-sG0R5XMacLt1@5HVxHii8arI|>d|G=z&g z14c}Un6Y3*!iJ0;1>q%~0V5_v%vi7@VME4_f&&!|A$A6gm=G~z!HR?p89NFNR5XN_ zcGB&Z05WzI9H?jrcXS4fm=G~z!HR@}0~HP7&dz`l6C!3TSdp+HV@E-_t21E4goqgn zRwQi5*imqxq9NSfDVF?!c3S+YPf12WMR-SNM8txGjDm{r&d!L41qm4i72&GRh=>IV z83h&LU7ZmT3lcI4D#E)vBO+!WatjhRgrPHFLKfR`prRoxoe>ivW~@l3m=0#FNZ7EW z;6Ot-qcdPa#EcdDHJyTrhOl)8BvrOkJro>h2xoQ%M9fIoP;ele)fq4$Vn)J->BdgP zf`knj1qUh`hHrExM9f%`upy(Mq9NSW88IPZ#)=IYI}TJ7gFMhMe6}+pV#b0M2^l*M zR5XOobp}j`Sg;~tLq@@YhVc2$fDse6%R3nd!r7ex6WS?uMNf^G5Kkzbu_9qZ#*Tsm z6%Cv9mr-yaoYNUFA!5dgs!GG~y3U9R5ert#MztVeL&kXel+@LR9r=BImlafm_jg7_ zEJ(;Gs0bhEjEGo}kWo+(KG+!%u^=I%pdx&zGa_O^LPkMF$ej@p3lcI4D#C|5BO(?g zWE4~czZFDZn-?Tx6jX#ycSb}kNXRIt2%qVUh**%2QBV=C?Tm<6kdRSO5kA`)5wRd4 zqo5*ut}`NHK|)4BMfiMYM8txGjDm_#IwK+$BxDp+gfDbPL@Y?iD5wZu?2L$5kdRSO z5w7cuh**%2QBV=S)EN=6AR(imB7C_sB4R;8MnOfmzB3|XK|)4BMfgf*M8txGjDm{b z?{tiaSWYQBZ$Yu?4m5;wI|C*pRif04^%TQ~9R&v(!g-wm6C!4;r<8MTK|;okf&<}0 zodFXfW-M5dPXk9q_)KR+@l#H5v>d2t2v>Fn9H-xhq-HEwk+7kjdXYL#Su}*y8PHC> zNX=NVB4IpCMQM9f&!OUS1+go<#jg7bQn4Lb@BG=#A;U_!)<6$u-56dY&> z=XVB7h?ucrbn$o^f$WAIIZ5F_MMHQ;XTXGr84Ffy$k;J@6Pz$(!HRUso-U4zf{O6j z&WKIDjDiCd4e52A4Lb@BRD{=e21H#gNXRHB&hMoM8j3L=XbA7@45%uH&vizm5B1oF z9R&v}8lrmPIXyLC!g6YRD#!{jA!5dY6pL4Vb`%^)s%*&EQE;H5A>7&-Fd{vtZ}ANoI|>d|G=vK}14c}Un6Y3* z!iJ0;1&d#MRwRTkb_T?`m(EzQB4I>kJqXztWkpU`4`qN~y9T zV@JV(iiY6Vo&h7$bNhsC$kuQOo8goqgnRwQi5*ilf;X*;druUo)ocgBMC zG}*82^=2$sk+30SN5O%LhVYusfDscSW-M5dupwhd!GVf~@Y>FR5fdV2ELf4SA!A3u zfr^H3X=lKQ2@x|EtVr09v7_KXxWL21gz#ZmOo*7VAYsFff&&#%2Qw136L09LjDq8o zx9TGZ2p-qceAMogHoAU;j|-hj2#6B8p4}914cy5Sg>M4 z#*TuDhVYiofC&-F#W(Dj#gqhh>93>g*UCWefP@jYV5s2JZXhK!2wePSqP{6IxJ zJpjX``x)BRj6dY&>*GfJ82YTumoe`^8!jACS&VUIKyLtr``#tU)4Ke&+ zfACnJ*-H~LcF(ttP`zfYHLdI(D6ZTWevd!2LUe_5gA!5Ob4H*Ro*3l)Au|M@kdnqc?=lTJku_JuG zGhjl*f)yJw3Jx@c(it&f#&XJ@(PunjK{3S@4e{%p8BJ+8voj*x*kdCmBy8A~7F5J< z^w@$G1;;6SRxeG+r_@coAY(_UodMI?dclGsR?!f@*_qLlhI2Y2;-~veE?7};pdoxl z>aIQq;qK0eF!UulVnV`(9R(E)!x^0s5i?dKWb8Ol5zg!kXx31;v@>G;xN`^@JK9&j zuUEODvp>zDg^E=yoeG|=02NIvd`*jIC_u$3mQDrFRDg;m7QU{OK-O_A`{CqG9*WRnQRc>C9M>kg?-HMblqcIwMvjLQ;hSWv_&+NsKQ zJvC!R!Zc}t1KyC-77ih|>mjm{iotY^w1 zV>+*Q5V2tWgi{_n)_1Cmj2+>s&VUIK?YvGHJ0qH|!q^$Hep08{v2MLeLdK5pp3Z;? z5zYA$zT}WW!iF8?lrq@`^)&5g%c7zop3|AJqo887loRr4QNP&p#ex;P(HB(gUINSM zP0+eZ*b&ywfC&)`R&2;9IIy1I`%B2!Z|+aEpknus3!0&YH+DwsH}wTvP|*-;XU2+z zjQteTEFqsdu#^ib8sgJBGgc&I>^M+Se0#5Gr!VtkPt917kg?-HML54RU_!)#^%VPN zA8EvbcJcqzTfC$*qMau@%3?yoh8+bJ4bzwV+K5=NqToP7xV|%8>~Uj3GdbZ+oe?V% zGWK(=95lr9Iy1_Kk9Z-(X?;mw^9)6@H8N319~&=8){ z88KnTf>qmujE3-*&WQb6eRc~P;?12ID-tsHQ_Kt{>nHQ_v9K(wWhWC499r zAz{Ogg65Lpg3gH5B@;5%-N}WF{lxiwnhGkW8+tDh3sw{yXb4~H?7AMh!w@otEcUE+K5}v&RCIap*h|lTF*zf993hF84_AA<-rgC}3@E(4Gh#x*h8+bJ4Z~%f5fL+Xr&K{jLwsgu#)^cD z9S17btNI5fxUe%~`V+&&g7vHY86}i=_ubPFukOrPk&vf=3~KI45% z6jU_C_jYEiNXXc6pd!4lGhjkIrKUco5eq`=yJEzIgbh0iRs%}Nm~QK_hy^PO4m5<@ zrBn`Ac1E<*)t=ukIbwRM10D;CQ@)}hUf!9pA|YePfr@ZNXF%~wZ$(4=LTAQ`gp3^r zD#8~#113Z)Sg|1^-ro2Af)xb^8p0i&5fh5X-4HMG^fA4nFOY}@s~Jkz5$@;=m=Lj> zrhQPB|Z?95n^kg=n=WJsM6D-trMyLy#~ z1uF^;G=#f5BPQ%9Xo&Z8W~@la*fIU76A25}7xdXk$Y{6p&$4i9XT*er?UcHzzZ_>w zFV_neGy@Os=!{rDj{ z5bo-?Fk(W&h8+ddpJ{;wtI1BtnC|XXA{MMDIM5L8=_F+A2;qnNL|{V1errGB3M%&7 zWKl6az29ZTf)xb^8p1O=Bc>Cruc3$qMXaJBKC?4pMMB1o=H9|PJ0qrN^;pD;f&&fV z*_{y+RxJ`TcCmtnICN&LNXXa`&*-cuIM5Kz?2MSu^c=40j98J7(R}QLn>y14{m*72 z78GsksRP8bIx|)zWb7yI@BmN|&+Y{aRumj)25Q0OrwSIV zCln-Ptk=2^Wb9}P!qyqFepVG^>{t~fWb7|?`>1Fq6omJ5Moh0a7A#m#C`iayKj%Wo z*wGY(_jX21Z_o=CtS1yCWHf~Lbw;fATRLsJH)@L&)0RV<;&h7Wc|G_i2A7P1K$t5`yl z4Ik=^Xky{pTF543tYQgGHssETCKkS!-qQ~npn6+3)zH>RV<;&hL3bc zG_i227P1K$t5`yl4Ik}{XkwvhA)Am-1qdJOj0n%^?~@}YBy8AGP|+}4&>0aiV?{#7 zev0WiA)}ox^zoh=v0rT6qoN_cq%&iEWq%<}$k-8H)fq4$V!?_H83hL#!mB$YCd^on zuwh3*MZ@r#&WMN^tAQtEG=xudModrd=T*dl6$J+x!ZSLn$_e?@>^#RX7OW^Z&=AJX zhzU)z@X5}IRi_CVJHj(N113bQDko&@x+tPI}TKYPjv=Nh*+>n}m#VVZTsCL%gUnV?{#7j%hVlSkQD3KGhkq`V39jFX{s- zsAz~&XU2+zj2#CmR$V3JQ>U@dT0+K-rXYN}Gh#(TK4ndNLOwkJgflw>CPcJ%^>rHF z-5D_M4M!|uGuyuATqM)L=xA2+Hi0Qd~ z?M5tEQE;FkJg+li!i)t88+H^_G_1N$$k_ezR8Y|n&+5!rk&vRla1qT|!Ih_#`W-O=dTdhef*!>b+Fim~MMl4uSaG)VPzcXUOj)HoM zy|k~jhz09qz0-u<@+qjNg@kxkXU2+z>1~FF1=HJQv7q@f5Wd!#E>nO7`^y|MsA!0n zbY?UY7Cze^M--TsM5KGh+Ri zpUaW4BYeCwU_!)#6&o@N4m5;MbVf{=u^?f?j)IDY;gg*a5i?dKWb8Ol5%$i22@wlc zY{)1$&=5Y=88KnTf`kn_3Mv|gPj^P_)vq(CSj}HT#&k}9JP`|46dY&>=XO?qdodwn zM;JQ;CPXY)u_2@2KtnjcGh)Jw1qmCP-#5b48PV?OXH&SdGh%vS-#rlvRumj)2ruf4 zm`>c)n?)?xt?`0-TJ4B`(wWiBOZa?e#On4FGIoUXIs+y|qzn7Gma!vT)EO`#V!?_H z83hL#!qgctVf~zWK}PZAr=lUgwliZzLdK2*72(p(fC&)`R&2;9IM5K5&WH&!79?zF zo>q8HXT)y)3MzKz3O4e{lj8O`Tlcv)w}^qfAbhy^PO4m5-dI=i+775gh>QPB|R z&WzPC5;9h=3kg&5MvDbaY4}2C#EOJ`%DQesK8*n3xt#$MB6bv1G{on1W~@lq9aRMt z4e`aD87mSpc1%}TtXRh*VZ)AsiiY8;&WMN^`wRP0DX5q(?iY$!u%h5V zLwHGN#DsXM@0?gsaG)V9oe>jeEJ)a}qoAT;cwJ{i#EcaQ89NSCgx7ZlOo&*pVnas3 zfrjvg&WH&!79?!gQBct^ys}W-}{Bc^)i1mb6LPkS4IwPiwdIu3J z3Jx@csk5q=kg1md#)9S=;j5hyyAQsCie@*4 zD|{(*p@NEr_=3)i6$#;4{jV{Nn2@kxM?poy@a)crh#4yqGIkuO2t#MUgop(zHe?hW zXb5L?MogHoAYsFff{KRW%+83287mSpb{wb(XLSZlh*+>5Pb&u_7U3$AOA)ZfC%Rhy^RQ|Bs~qY@gpI6UXgm_dB@Xujs?>vKW>ky68>= zhbp>&07ExjbcK=_NHHV<0RjHk>EDy{kYrh1vS&W?O|WC&z(jg&upnc_hK4&j22M=e zmS91_ii(CE9S2UN*9A9Z6l|!tW5>XOiS+tlLB@&=4R>@5oS3-zZQg={6%`Hniylwd z&^=QI&g<#&YkzJ@5oS3+s!GeMn6%9K&4xC7T4sOUO*idoDj)4Oc z=`X>8j1?Oi?&ugeF>!k?SWvK{qG3nJffMQZ;D(HX4Hb9n7&tJIUI-Rstk}?SN5{a4 ziQ8X;1qCZA8g_IXIFbGq+>lYQq2i7m0|zG33>IXp*ihdahk3(}Zb=QC=nnXC;Vt~c zM0qS&QPHrY~%ld&CT!*TV+o4Z(_vc7;598usfc^#0fo11BcR2Z9w9 z4Lkb9ht!J`6XnCfijIL3`3LdmhJp?GgHcMshKhj$6KM?=WK4}re+(AXH^dhZ4Lds0 z8-p7%3N}>Sv18!CM0!)OAY;XbhC8Y%8+Mep$62zWV&K3;dPmUzlYQq2i7m0|)Z^VzCx%$X{|GV?*~f z)xe2~@|IvlMZ=Dc11Hj3gBvmmHdIWtrS}I5`V%qi11BcRlfjCLh8-OTPNXBaA){bJ z#T`2a4osv!1Pe0yqoxce@|O(+Hgs7ACdwa!6%`FTIu4vje+q8MDA-VO$Buym6X~g7 zK}P?Grs2dy`Dn1B8^3`Q`Hc<$Y^WGGFp=IAEXY`~p*mk0_UoiU`f6}PM!|-PJ9Z2l zs2_-CH0+qRQ2JA_p!zh{(Czbq^V;z!4+JYJ8g_K#4@P+f8z#~kEJ)A92ZR;*F_VW4 zUD`m^f`%RCvG{rLhAEn!3KmT5OD_ghv|&g7ijuLRUZj|#h8^X>Xz7NEfddoip=t9Uknyx%;X2kF9{ap9FtP8p<>{`M0!iGAY;XLMZap9 zVMFzZY1pra4bnS<8!`&2*=pF)ebgT~F;RXPtf*+%(Q)8J`cZI0M!{5QI)eq>0v@5oS3*h87xTOif7(} zjD|Zp22M=ez8x$mSW(fiqvODd^qt^_jDignckCEAFp<6+^dH8H(!hyngwmgb1>O7( zoR}!@3RYCKDr@pb`G3Z$oe}mQQjJ?sA$;Hap1(fYWqu6xgftg ziZ7@bI53gk6IA&ccFecO*U{-6!GbDFLw|2DaAG1q5=*L}V&K3;`cSYSqkBpYoX9VY z%~`OaKjh5~Q!~{vG7IWs40JQMCy3=u7N91)_{y3s5DjIfl95|7F65NncuF%_J zGZk#87&tJI-X1K-Sh1mN+Q5maY3Z-Qf@*~?ff(_Eih%4QNo#n8f=u01S={Uc63a2q`w6Vy65k>9>OSJ z3RYA!?C3afB7HfyA){bJ#T`2a4osxSf(02XHZqqy_LX2k!HSB89sNqvqlO(5 zX$A}W$DC$3G0`7(hT+6SmAPTRZu_{~aAKN(^k9%tpN*H}hJNu0nQ1`bT5j|B_z#WyTCZ0Mg8-&H83h|E?$|MKU?ROOSdg(| zL&F{U&GFq$!G`{?*oXrsx&{rLm?)nLR&;0axDI5LKL#r*8g_IXIFbGo+>lYQq2i7m z)4Zj>2Mh8U(^F89ABjCvu%XJ*uwR)Uidik#P%&^|B0Vf5bNWZHpgTqePE3?{1S={U zc61y#uc(SaSmdL+0Zqg$7f`Szl4LdpxoJhY2ZpbLuP;tkOfdlpGC3bY(7wu%KYYwBOQ)f(7Zr(aQxH4R>@5R4r)O(VsA;IMEe4u98u{ zsUlP~?C3afB7G~kp=-v#iM&OX1r-AaCeoup{cz-K*pauOU_<>vd|BNv%*=^-&CzGf z7P<{PaH9DDvtOTLkoMq)jDignckCEAFp<6&EXY`~q2Z2>ffEzA?*|JCR#Y@hBcA>f zEXd!nRIs6{sRARXH})i;qXX zhW!ftJf5%}=@-Eb83j`*DgFPVBII3p*iex_5k(bj$UlzHk_8n52PV=_f(02>p$$9g zyQ0d59aDMflfi=AVu2KN(SZ|FTKZy;-w}_of(;b|2PV=xg9RBYHZiN59bA zf%6JEi_hyoMk&FHiiTm?o|q_~4OY||ZEn~x{$XEYqWm*hQPHrYk0dP z7qB6JDn_VaLpR9--61_NQN9$csAw0Cs{J~zkiPBsLPo)cdZm3w8aCvwYBM%Wt24bM zSWsUUpW_>LbW|4__ABJn+tHCe65NncFf4`>6XkQkim8C~(qO^-TCB-*4;G}i$6Mrr zjDi*E2XU5UG~CfKaAM;2!(c(dii(CE9S2UN9|bpL6l|!tW5>XOiS*;3>O#Yg>WSDc zT6VLmb%?&uge zF>(7yu%KW?MZ=Dc11Hi)gBvmmHdNfPW8lC<`dF|aW5tGsJ30nVOx!*mEGSq}(XgZA zz=`yU;D(HHg??%4V#9Q9q)!J6>X+j1XsBkeVaIUw#6*92?CF6M<+Cy38!844Or*~R z3o^P*JaD34n;j=6%14707471kv59wdq;~~3WE5Lfh4xH%H22MUmp9L!_8g_IXIFWuH+>lYwzo0WXQFX9kze1mh-Ls=3y(}hv zK}N$J9RnxkmDZGu6a9;#I5BOK^syjcd@0sY!G?-~0~6`X!Ge78LuWfSlYQq2i7m0|zG3J?Jkt1~`#l<)p!eih%?DOGX1H zCd!wC6;l_|=Yjqqy_K9FY!HSB89UTWwq)!Gn^q<8`!oZ2C=jp$L1sM%@3~O?tJQl2|XxPz_e`MKW zLp5LxJMxcR#fFOMLV9Jepg!*N40iMjvoLTXKNRb)U_-^gfr<2Rupnc_hN_N+9pj03 z5}%kTPX;S08g_IXIFXLvhKzy@71M`?lp?fXswllGSWvza!?dAd;J`%sYOo+qz@WgjQ5E?C3afBKSv18!CM0y}tkg;Mz!yO$1CnjzW1`7&SR5a}9IB+686x=Z1 zWFNgbSdicAL5~f?o9~H<^8H{%MZ=Dc11HiCf*YpkNw;7@zQ@=XY^WGGFp<6ffEzAhk^wKD=HdxbR0O59u98EDA-VO z$Buym6X}s)LB@&=4R>@5oS3+MC|FRiqM~6(M}C)&JlHVI<%xMsZH>^1iiT-^)2o98 z;~R0?iHY*fU`0j4j*bH-(zk*eG72_S+_7Wez(o3Xupnc_hK4&j22M=ez7s4cSW(fi zqvODd{IZ~6LwhbxnjIbK`QV0(f(;dS>=-yOkzNQEWUScGa7V|$iHX}^g9QaEDjIfl z95|8w7Tl0gu%Y6P9Rmj@(hL@4tk}?SN5{a4iQC_U1qCZA8g_IXIFbGl+)%$2A2S+u zq!w#(K}N$J9RnvOZeI)*6s)Ld*wJy|MEX*2Lq@@diaT};9GFO74i;pr*wAoC$H0k+ z+hf6kf)y1FJ30=WNM8wV$SBxQamS8<0~6`1!Geqx8yfED7&tL;`&zJ|U`0j4j*bH- z(ml9gzST6oEm)9WWxHTQ|6nk1Vxp|Uiu%QPZEV<)zZ_qw6y)xyferl;KNY}4^Elqo zk$xE5kWsLq;*K2y2kQIcg`i2(74S*wJy| zME;pNuw9`aMKd;33>=t9KModT^j~^x<3#>aj8(yg{Od?7s2Dgfkw&l}qdXXG*-$ZX zU?M#fEXY`~VX7#7Ian~fxSp8Td+e_ww4$P6N5_E^>DR#x83kRd2F@!a?ZmvYj0ml$ zXxPzl;6!>nxFMsUUoFLn{2SfJhW=1w88|Uf9u8LIpPMagm?l1bFj$b$aK|u0=QT_y z4+kqM8uB;eg{ojf_i20JygqlMJQl3z{~NDc1Lqa`R@`x7qI^48QPHrYBmY9fv7tT| zJ#W}C?XC2f4fd6Y7HsI!2D%muoXFpaLJKMe4otVDuLcXcKlF|3uYK3w{eBtI6%`HT zkr;>#6$1w*(uacn+nAYw6J5yzC+b(CeGT<%5pCF!zZz);)r&~Oj`|xPTCt-e-{X#g z4f*#`QNf0afddoiiC{s#_@yDnhP>(q8|nj**07@^Js8}Oe=RdMbZG-8 zCdxy>imCPK>%oHg+E_1X2^M5D+%b&qiHY*zU`0j4j*bH-^3Wh`=wGomaAJB1OrHue zx@HWV$ghif3pP{?99QVOv4&1els#Bc(XgXmd{waQ{mSxtS+JpEx{&@mSdgEP1sm#x3k^HQ*P>gu zd$1s*V8wkg;Mz{YK1m!;bvvSO*0a0|zG3XM+5_nAL)cfddoi`@w>Y`ubQe4LkBTqmF_N z(~eJH3#MrLRWSm^m7qfQPD8IZaJcSBUn+D+_0mYg@N-5St=(c%143~6%9K&4xE^ZO0NwTbmKfQ zQGOh(sA$;Hue5JRou@D~b6~eC|mP2Mg*KBduXa`9d_eV&K3;YC-kqfrcIVP!TrdKZs(( zw2;zwg9Z8Gqp__DDh3Wrq>lv)GFEJ;#<^ifmo_j_J|3*7XxPzl;KX!q`oCa7wI&;? zXv2;vExjz5qUjC6g8YdnrJ!Qqz(o3Fupnc_hUsl7Jsd2^e>6-g8>;p->{m;li=qlDy2UkcVxs&ySW(fiqhI`AJnjZgbX5+Vm?&Qg zR#Y_PKj|AbbVu{RiHY*fU`2lgfKjw-ETM@RZnOva3cRA(h4?I3>=t9 zzYG>+tk}?SN7vGU^J+87w}TbkDKK!N>PN$l{FLQ`4Hfwog%)g>@})Ni3%Us%I5AOv z6RfCc*wImb8*Hc;I53fZ7c9sZPiqi1q+dk~7GyNs(J^ph;`ZxcLBWdQJUFjI0i^~j zDjIfl95}D&%i^&!a3Z}gSdh_hN5{a4iQD^w1qCZA8g_K#w>ti@p&H$W9TO=B3#!wx zq3$t_4LkDJimH@`9p%mODB4gla9|?kU_r);4Gnj644jy_y(L&s zu%e=2N5_E^>8-&H83h|E?$|MKAb&49U9e%8W!;X#vC(>(!8!`$uRNS#+;J`#$f(02XHca!8-V!XBx|O~kEa(_GG0k-PMlgKi zHLvf!eiNY;6%9K&4xC884Q|LNs6U888+P=oWHi4t*{`3QAbmZ!A){bJ#T`2a4oswP z1Pd}&Y-qTnV<10c?_;|{pNieMp<>{`MEZ2FAY;XLMW0m)Hsl{hw+c3lzr~23m?$$? zQPHrY=-yOk^T`Z$XKzV;f{`h6BD<81`DP(rym6i>Nm{{s-@Df zqub&GCnm~wgB9Jq1LsxB?|cz}9UW6C>8-(n>7+?7cif0J?C1{ZffEzu_rZ#ah8_Ki z`cu!2j`YvqhKzy@6?g0yI53f33>IXp*wAoC$H0k++rNSZ1uH5Vc61y#k^UXrkWsLq z;*K2y2PV>gf(02XHZqqymg4Uz3kp_LH0oMZ=Dc1Lqa}L%g&Noao<)hu6S)g?!>WF;SiiR#Y_X=s0j9 zJs;eVQLv%njvWIBCejPRf{gU@7>ETK4R>@5oS3-%B3MwcqM~6($AJ^+m%$Ag1sf{v z*fDTmBK;~@kg;Mz!yO$1Cnj#c4i*%wsA$;Hao|K6!3`M&8!GPDF>qiaJsvE`Sh1nu zj*fv76Sv<43kp_LH01@Y_<|)3TBmE}0 zA){bJ#T`2a4&*;84;!kD)39UsG;v-ZDXtF}UyJC9iiRm)dV8=S-6L&5M#CK)<4XHk zZ2KJ@>E})XWE5=t9PX-Gzs`+TxF|X-) zM}!vSug828R16%LNZ$ymd=2{*lGc&F8QhRj{vUA|ZQl$k1`bT5Zv_i7R%~dvqhsL2 z#O>R`f`Szl4O6SqJA(z~nK+O)R16%LNY4ffGFEJ8xT9m>#Ki3k7F2aK?5IDEmx+cQ z{i&Fvfr;{Tu%e=2N5}Y6yvCoHC{G0|DjIfl95|7l4sOUO*idoDj)4Oc>6u_b#)=IM zcXSM#n7BO~EGSq}(XgZAz=`C4Y;i+IL3d~koS23#y(?Id|Kjw)hKl^0G;HX;vKZI* z7Vc=+QGf3p4m&#PC*uHZ*fDH_6BFeKR#Y_X=*Z8j7aPX+v>6lS`@xF*f_&Jn&~M`u z-B2-bU?Tl4Sdg(|L&F{Uugb%Ah5it2*-$ZXU?Tl7SdcIN=3Z=9=(F*{vY}$&z(o37 zupnc_c15RBuw9{#$KKjdF>qiaeIi(pv0_8R9UTKF=9Q+o1LqYok|!q04}ul>@7jV5 zL)wXn^21<7{)c?nFr=NBC_f5Tlpr4gjQ5E zZlSTH{sheE2sg7I_nhVqMGMU}5%N0&BmVxsxc_l}PAzTk$8f(;dS z>=-yOk=`FH$XKzV;f|>V>Ak^%{^9t}Xy8QpKxA2v(Qrq{z=?_52ZIF#D=HdxbR0O5 z*5HQnfp`bmP%&^|B7HDekg;Mz!yO$1CnoAoV_7upC_jkVsu(yhk$xEDpUHv^6Y1x{ zf@xi)ZwLKjv6Kf+lxO0NXG2B9{fZjSh8^`uix4|Hs(Kqb(vN~0G76@9)BA!2UE08T zT}6L3svJ1ce`z)0L_O4tZVL@elqZ7~6%9K&4xC6waKn5{Y`XN;U_pLa6k4!dp_dCG zKW;*?VSG9`QJ#tiWW~UNiS%@^AY;B-z5f*~7*`s~Gr@|Ah8_Llzh%aW^phBzjD|Zp z2J)X{tO~X()FZ88;J`%sY0!UVOmU)`iiREe*VX|xR7@Ar`-27f3-S0Ys2Dgfky@}I zW5tI0ly88rqg#msC+Z(!up2t&9!oxbFIZ65pkYTxdLXzVqhLeD9XkdNOr!^c1sU`4 zc=Jua2^REU8yuXNC?iqp{JRYpb{~ZMsY{;7^HdG88m`INX`Hayh*ie5FqtURV z{?WLgqkb&zXxPz_J|5hVQP4GL;KW4vO|T;WGDlYw6*_RDo9=-V6XlD+ii(E(w%BM3**j zVxl||tjLdB4A{`64V;)LPX;UUYook^4f&O(5gV#6#2dQI11BcR5v=GKIFX-?BdB0Q z|AVVIF;V^)tmyBEF&#M3KWgP*8mshyU_nL0j;U4Y6`FQN9Rm~P&%ug{h8-Q{OR;D+ zR16%LNM8;XWUScGwRGUT+U#9$?-{UNv^qhLeD9XkdN=t9&jkxIrbebUSdedq4jbx) z3k~~~<<+uaLq+<1Y?1{T4R>@5oS3*h5iBTJQPHrYFcrdonS#m!yO$1Cnj#+4Hnej#Dq5ND9^`iZKxPHFp*ve7Gz9g znjQ!iq+i6*w;-e8j*f9*Go6?yj|D3#8g_IXIIrkawh+!M^t2GFe-zuWqrN6aqhZJJ z*X9!w<#oY|iiRB>2Tr8d2RCFCY^boF1`V8;D1Qr9R5a}9my^)3BrdDRyJSj*dLzwt@{60|zG3--87i z{aN#Y6Zx01T?#f-3>=t9zY4m02hOXMQz71P!9NU)&%GsdrC;J`$BF<6jMWp3Ef6*aE%P@Y#jnqM(=q!)u5 zG72_S+_7Wez(o32upnc_hK4&j22M=e{v9kRSW(fiqvODd{NJ%y3pRAid*H-;Z>*B^ zz96IFj{5t!+OQ*kH(neHy6z90=#GMciSkUaqM~6($26wtL&1XbM3lUtVwm3(6XhGh zii(CE9S2TSMK$c0NFNRsR9mQFzcv%nuY(&h3cA@JIMH9wOPt96j2Tr6XgBvmmx;h5V ztG4gNlof2K7&tJIz8fsaSg~Ev*J>#?^b-4b;6%4h2Tn|s5v-_aSID|=*pU~7VnfBW z-_n!8g8EdXHS`N<11IwP<2)$XP%&^^(Ic`n>=@qfPt5C0|2Gj@QPGh95cvu=bnlo0 z=k>OU^4DNRMZ=Dc>9+L8U_m!@0~2KrR#Y_X=s0kqejvKguwS8P9aZQ^XK+JC!G?-E zb_^WIpNzg0Y?!7ZJswQ6khWmKw5ZcRf(04Xg@zqn+Q5l<;X-;eSWwZhBfl;tuV6#H zcsfE2JNm_=v0?{KS3pV5r$37~k7&tJIwqQXZQf*UdlHdNfPW8lC<`e?8qqklA-J8+^r z7>(ReF>oONATk$hs2bU@BmX9*yI@1T_-%w5I@0fg8#2ljdVO@OV8c{Y`e?AA|0fwPJ_Q@5ca^jUQ#3st zEU3@MKs59VX#*$vbeQ8rJ!M9HCayN@=t>@_8q}~~P4it6JnLvrUH$ z>A7G*M#CNDckye7ih%G#2cj1?Oi?&ugeF>!k$SWvK{qG3nJffMP;;D(HX4Hb9n z7&tJIj$lE?iVY2SbPSxBxcwnmP_Uw+VMoV-6X}n^4H*R+D(=`Za9|?+DOixPVnf3n z9RnvOZchaZ3RYA!?C3afB0U}4P%VLm9pxplS2t7)9GFNi4Hjffccf1Q3-YtE{t7l! z44dSJuAgN%kd zItHfZrZZU3XPi<4Cnn0@gB2AGJ38_kVmB6Ss5J_0=t!RmZs>|1I5DpdelmKwpxTJ- z+KvB+LJKMe4oswf1`9IkpQEUT9re%Af`*Q&{` zL^^`-?P_N$`5L~{!^)bGW~)zJU$z{83D4^f=x*5AN+twfZU z1uH5VcJzyXDgY-Y%8S8@j)4;sC?f2{xN^Pz=?_KLc@+~-Wqmvq(1~VWRxrP zlW23nhKhj$6R8IaGFEKpANSUR6H~qEGr@xTm*`-_j*j8B6BFg9!HSB89UTYGEBbFk zf)nLuabZKnz=4VM^I$>7iVgigO2LV$kqtZYo1=XN8!Dy?>9fItoMpj=df`IDer0)! zEZ9&nRDNQj{32LU(XgZAz=^56^toWcv~kjN!Giq#*gXaL!?DZ?HuPj|;KW3ENw6Zn zJa$UKhKhj$6X_Mff{gCAf%6JIAFl={Cdv!Jii(CE9S2UNzXmsSr{lnhs&5TD>T@w* z4Lkb(Bl=t9zX}#)tk}>$Y2)BTH7*VNH9|;#3vS3L*idoDj)4Pn3DWC= z1^uPw4JRhb%YqgC^^s-ZM4myxhKhj$6Y1~4f{Ya#8t$lX5XFwJqJa|=<>kSO{5BJS z4f*Y&*f0#!iHY*-U`0j4j*bH-@;j7*4P8Y8=T+s1ECm}X1`bT5$EE$n3kD|2bHR#= zh8-OTPNe698!`$uRNS#+TIlKX!Git@jl@KGWw0W>5SbTbG~CfKaAM;2*I+@xii(CE z9S2V2SH$143N}=2Y1lEnoTOI;Q#5@cSTOCQ^w(fPzIdl$$A&6j!;bDbGjL*}{4rQj z(J*CBEm+Wv!N7@$@~U7(epT$jf(;b|2PV=jSddW_+OT6tJ26pS9jvHm*wHUkIdCHV zBbN7qjD|Zp22M=e{uwMNSdsrLN-o$?w>Y>Pc63a~SNfMT4E?_&-@u7}ty7$sD8C6- zR5a}9IB+7rNn5aC{J}YdiSox_MMcAojsqvspMo1Q3cAHGaAF$g^u=I7wVE38-(vk0 zY^WGGFp*}kAY;XbhC4b2PE6eX9xUjeiUtjwsDFzFHSFkKrw2|T1J&b=p6>Y1q*}qg6OjwX|Ws+Klv%;D(HX`BdDSo(>jdG~CfKke`Yc z6l|y%I53f(4)TmKDATcFOCd#XW6&(ZT75y(!OqBl)R&)%U$VEli zQ2!kjHRN}@fDQRSkybDbV)|0BU|K%u7W7ABlMI~b*5AO1iSksiqM~6($AJ^+>EMQp zg8oAMP-);q^K;o99qEDKhKzy@6?g0yI53eO3>IXp*wAoC)yszc>aQ_KDSnQhF{Pz1 z2MfBiaa}-_*07`eHa2_3z=4VMyI?`aiVf3k>9JtJ{74LJ`e?8qV}3eb+R|r&1^qF{ z5+?FzBduUV{%WKZ)C+|+>{szfUkMgev)Yi~V=-Vuy^y70M}Ds?*ibQDNM8*WhXxPzl;6(a=t9KM4B&*a0}Nki~UkUTf`{2(74S z*wJy|M0z&3A){d0y6Nx1f+6j^E?im8h_0w;*wJy|MEY}ZLqw{61Jw(XgZAzffEzA?*$7AR#Y_X=s0j9eLuJ%qhLeD9XkdNOeDYfU68S2!~A&cxAci%LH$<@ zcEgVP!&s*c`QH&O*pPn|X$8~drF*cTn%agP9qGB?hKz#odRqhi6R}lYQq2i7m0|)Y_qVj?bRWBQMuK^7^I;IQh8^MC<$(ddqRMCbV!~Q)nQC=UcsA$;H zue6uMUke&`j9YLby$}PoAfw@qj)8iQd<{G5=i=$tuw!aV`ev}8YH7Qg``%b)4Lds0 z`+^%X3N}>Sv18!CM0$U)Aftc5#>R>A*J$L1ih%2E^M#Sb0Q^TC4dSFr>0y+L|k zu%LcAX1Za=w9wO!ZG)G@%oJ>>y4tW`p_jy4+JO`E!iDs$U_q&|%QjRD9GFO-3KnFn z*wAoC$H0ks@#a{kDF+MEOQUHkHkAKF+J=gO0~0C5&@ISVv7zCP@z0pw6BFgdU`0j4 zj*bH-(!YWmG72_S+_7Wez(o3Yuppy5dnuJZw4$P6N5_HlioPL^k%1HWZLw$y zHjFpAfQj;^U_~`i4LiC)95^wN|1Wk_!G?-~0~6^D!GetG?LR%FZ>TSeQW|#Tmj?wK zD!SVS&Z_{FHwG&z8g}$6?U5+CVaHTTdPT6HW8lO@epB3Au%Tk$z(jg;upnc8I7&`i zu%Ny&N@>`UUlkN=sOY9~;JhYODFf$K9!d^YR5a}9$RBXVVncpu?5%?Oqu2orJEqx6 z-wqa}TU5RvBfTV+@`8+pJ30nVOx#`?EGSq}722?)z7TC`*wK+c7!+(*XcaiS*WBLB4oEQP|LRVcx9#x%Vw z$Pc@K?Fv02g!~~Fuw9`K3n6bVU_)2RxQcr8|0C(Xp6%z&LorXzO?sV_bD5bnlQ=yW zsgqvMHU_9hR0QWn+>62dqT{hA-`P%>XrM1 zeSnS=6X}z3*O~L#IKB&}pJ3^!e(EYV?AX*b&U_r);iiQU^Q5tr1SE+%Cl7kh;SL3Hp$H3G%eK}Z= zpAq||pkl*eSjWIbd48~Bel4P;uLlb@Q5tqk0n#&r1sR879pegs@`7MR{$@lesFjuR891Pd~b=5vgJEpdiS&(VSkOKl#T^Hlx*g>+QLNa|{j?s}k8CWTviGo}qTzu99c790r(#3H z1L>I&G$Va8ww6Ath=NT+!;bWqz{K*| zAY(;E!vhEUg*6XMq(2YtXkU(?2Rb&JY}ir05fk0e@VJV96yv`gjSKQuVpKuJhQsxz zV_>4ZELgFjVaI{_Ay>fTSKVKlJN7Fl@{aV)*m)TRD>^39w}LxTHzBkm zX0YQx`C3r1p?oejS+QZqffEC15AMh)Xb$*}105&Q*CXhHjIQ&*#PazdV?{;70|z?F zvm;c+hK9#gyjFxA?mJ}-)Jf}bexz-Ukw&ytf**s z;6TT~#Qg{s6s*|LaQsnpv+eZVU_t)d_?4ibVnfG?iS&0td7*Cy)|b8&EXX*}F)*=- z+_0k>yL~$v8g?8PzL#|j%nKvZ_ksltJIY_hu!;@c5g(XXz7%AvsAzcLK)-N43{0fw z26s$BQx6tobSK2b^6nsGMMcNN@}3}LMMcNN^4=h0MMcNN^1dKrMMcNN^8O%WMMcNN z@_`^@MMcNN^1&cuMMe2g{FvRafsPYRw;cz{ zmt%)lY}j$bE=a7RW#b0l^g=s2&O)`5xTt3k$!iiQUcbdiIO51ig;*pdDs#$+@cm`KkH7ED8bXejb8qN<=`Lw=PL5fvLc zPE4d%2MaQekH!~-j)C-_#`ZIV@Ms*BKl-C3+HZ<%wFh3^n$Abl%r8Mlg z#}>Q56KMq90h>rUxMM-XfsPXc&9EH@$`PBb*s$ZkiGlRB;Es%P@%#u?v7zCC?idf` zUkQ(j4e3V?3FcM&b<{1$=pqkHEME^YR#Y@RaG)c-Ahx?8qhLjPNlfFoB4c1;`%TcW zQ=9T^4Pg=Av+NRY9jUb)y}jkHOA>@*x0 zn7Dr=SWvK{nccNY7V?hU3ba)qYLwKa1#pi24l;=@;gWJsOWlJ<{&nn^*haMyCZCXEcsT zGgA7`BfUTi?onDIh8+jW-v=8Ss(w>0{X?)Iqhlid zW6*5>Kzd=2(KHE^ zj`F9$hKBTnAfupTB0W*j#^$xf9=SV))wk=&pv4v+k@iSOr1VP%4o!p|#}&&*bYNon zR*_~sD4h=ifRELHg>2K7bVMqFdIyCIa--#2UU_<&_UC^-O@Z+ds z;CNx27aao=_~sF4h=hw7?A!!JsNg&_5Y|I-QgOTXwM9G z94Oz3qf)V9$AJ?A>AS%l83o-j8<^M8ds*b}IAhzxLTq|RewNLkV#7pQf(04P@jWoj z|G$oc>W+bld(`iC(h`{*sYF^MrS#87CAL3{ehoX0tEgc|{`06SsMwJH*jUV~`0S`# zkkK^kI21buCaQTn;#+b-#o^=MF)*=wJjhs4v6=RF^P8fP*Aa9&&M@G4Lc?7N4(C|Rl zIxw;PFvwU@(eS{5j`WHcyC9=rMKk9e=~Xe^1sUZ>F|1<4jsqtK(iz;5QP2e)m`JY- z?&yLJOr)pSVQ8<8;*JAdl!1xm$3e!5iiQUc^sCM`b{yB%aH9Mqx>Rh~ap1&2`e|@S zM!9%ZjIG$v@VIi<$#qQ}JMwE{;sq5OCej)#$VgAMXjG@Ps*Zt)?X^L}jsu$k4f}-& zj%)Q8m=`fLtu701 zh^mGi2hx8V)6|&X7^Q-W4HM~2!Geqf9Rm|ZfQt6!DDF6ro*t`PkgI;E*e*0Q>`2cD z7MwAu^p@zBQ81!vs{Zd{rIEUWF|m9)$XHR)F|m9m$XHR)1)NB44Z58d=8!V{Uklr3-6toyH^-V7g7Bn1~ z@A=oF;k`jdw}^@KGvTNPli8qhUvSSM2Ep8R^-&peqhcbh~R{V)%heFyC7plMZ*IJ`c-G_j^m1g6XhrI z3E9x_K)J`6zoFrQ9S71=f{cO{728$!rU<&B;ej3Lrx7`$U`6^1yA$Pov9BsNl;6k1 zD>m#naAF|+b#O;UL3(p+XhBB7if(EX>8ZgT3o@Fs_&_(Sc}?wSk-MXr-;VVDnA(Dj zZj%EO%M3DBR5UzrpkJ8Kz(lIS9qD=Y2DvzUap1&2`a^I> zMnSW+9S1s2l=yI`Y!+zPQ9c+OtJu&58JJl9Hpo~}(eS{5j`Eh+M8$@N2h#Ix5~)S_ z1sUb%FN7MHNN)@7 zXqK>}eJH~3I8gpE##U_Dap1&2O8;&|K}JCnbjN{?6Kfo&>R7!x&e?qN_NXqX*gg?A zcN%saNS_Su$R9Q*RBV{n?tg&;f%?YyzBx6dj|2-cHmwai4s{&^^Fl*{Ff1V$ANCw3`{J49AvDhXn5d2zYug_BE2)XBfZ$HP(B*r zD>jrriLn(Mb{se{kp48dBcq@Ry5m5{iDnx+(#InFf{bq1z{K)|AY(;E!vhC8%DW7{0X@`&&i z8@gcw6U&o>j1?6P4;<(hf)30p{JSG}N4ZCriVZssoES(?3GT=!XvXe1&~c)f^N#e1 znC^m%ZrH%Y^3)(>MMc8{2Rh1oB51{ih6hryhtP!|m?%$+VHF#895^wMo*vwhQIOsn zK^J5ctZ25eqkS^M?>Nv68<<$05oD~WXn5d2zYug_BE2uTBfZRuP(BsmD>js8#@LDt zI}V%}NY4uH$S7!n?l{nKqWnCr78@EKNS}@kWfXKwq|XG+tafa^zRIMR2MfA800Z;7 zA@C28Td|>C#HdZ*h8=^{qT87h%MxU)sAzcLK)*0-V)-w6%`E+9OxG|G_mYK#)^uLiRJS_#)^uLiRBAH#)^uLiRFty#)^uLiRDW{ z#)^uLiRH^d#)^uLiRCLn#)^uLiRG(7#)^uLiRB0~R#f!o|I=v2@`4~^MMc8{2Ra5O zniJ%}z{LF#EGSsfZD?ZoT9C1#qP;Le9T=FnzbIHxuwp~QjssnQiRJ4-#)^vm6#r)x zmZt_8D=HcuIM6XLaerE{pkTdnFOE%a*m2;*Kzd1VM@B)#b`_r!=ii1M2Tly6=LUCV z6m+|8V);anv7(~kfv)q!@)tqIii(B@4)hBba+6X_+vf{Yav-G(NX=LZ=pDjFU*(2-v1 zGp?XwLyr0) z^s9tM!;T~Bd!(sj`dqLe7`m`Gm??*DUSUTq2&QQA~D z>^P9W6cki!$kDe%S|im+Q)T*cupncP`W~qXzN5Y=e!n`jbqq|DRVz9M%4^l3W1xK{ zI`248z8Y+3D6caD9Rulzh6Ner1#z}kY}j$Wssa7RW#`aqlu3o;7UEBAVvMEXU1 zAd`Tf{#Vn9MQvz!V8=Zgmq^`wC(?(4I~HUVbWEfV z2X`#UDCn3-9|`VQkWtVvk^VBcV?jo_axtIuT9*)%6xY6r4IL*Y(jTicY9f6!xMM*^ zLHbZEHltugMYG!vbew3mdt4j;O%yjYJdmCq2QQ;wMa6dUp7>7Eup_%heFq99{MMZ*IJI?6|4Y84wA9_Y3)(1jnEC@+p-6&rROI5Ci365Nqd zP+g1CUq<8w83pT=d$VPseK)qT<3KldU}AY`kg=ko;eiAFLePPU^wHps^1TRJv7wY0 zQL$miffEDiWx*X81x=_O2RcqP2lTj(s>KYX@5hD~WOM-rCYF~687nFp9yriZJ{I9C zHZ(l2##y`hM%S?8z`(@)Gr@v_^~zN{9NnY_Cdw;fV-*{A95^wMUK!kxQII|!Q(2Hv zuwJ>}$JJm%!voz61}2uz1{o_V8Xo8-InaI(^V@Nt+wQ=`@~R+XMMc8{2l|Cs4NRm* za7S~7<{#PzsMwHx6fDR%&@nL4&S1xZ^5dXlLwSpxjj5P^5-iB*7?{|88Z_)Ukbf2w zRBV_?KMxjU9OxLB*nSZ->^P9#;NpYyx!A=S6&oJd(Q%^e!G;|NP7I{a2X}P80Zc64 z2{KkxbiWTxEH4c*R#Y@RaG+yg;$DIU-S0W``nBg}e(k}EiiQUcbPUX^Xc`mCcY}-- z^~(Kkv6jD%Gc==My>h>dowlK&Mz<|e(|trmMnU&edm=p}xMM-aBL?h|jz~RHbKV{3 zI5AK@5o~C9U`P67kWsLrqC33?%3Gan=olz(Q-{;0Wd_RI)uUrvIqMjh*Roz6pPY&f zI}V%}NUsU*$SCM0H87Dr72J{EVY{eEPl&T-K}N#^2Rep5rf_9XNL?- zENhUlqN3q}1O2MAyLKGtIIo-q4U}KUTq-t{*T&e24Lc5;7)Y-R?#L(?pN_~C8yX%c z@3Q-leiPvrWOTy@CYIL+87nFp9yrjEJ`+I~WE8CEwlUC!ADAd_h+!2Qb{se{klq;F zkx|eE9hgX;4esc^zR&AB{Kq19$AXM<<+hkcK}Gv*Ok>BP*fFjKC)kby>B+$z83pNo zkJ7tiKtaW!zGGmb+=CSx8m5yjy*XHre`jZ)Vnh0Uupnd8uVKeTs=jo=U`NM^ ziS(*qLB@)T^5)oA8+IHxF_3C-M@B(=xu19#n7F?pSWvKHL&J^(U4V(@EkVYLiiQUc zbduexF|oWo$XHR)@W6q7)%iYgU|{0@g3K&ireed611AR3JA*qi z3Ysg`j^ny!;Y5z>$#|Pzq~0EEe~9fg>^QKw;{0u73Mw`{qCUMP)}GODAiY%`x-U8t zX$E(szl)(61sxOV?}KL8abc_PiS-v$bPQ})U+g$g-m4$VKSZ%&Lyt&Z{XqK1s9KQG z?VN#$ii(B@4s@ImvwSXgU&V%o2ZkNii5BbE1=I*w~j z7??;rpgc3a5^QLAV8?;{bE?5l!`4M*(Au5qAzKwoqW zq(6yn3o^^M*((iW*jTHX}{^OK{tpkhO_8xC}w7)bYM%_w+8kn$0e#i6QWU}Aep(6HmU zaxvxXR1O*(QwFh3{31X>WGxT6nmqU_r*=qu4PpQGOMy*wC=!z=?tUSrbCV z@kt+O3>+`^{SyNdDZzq*6^He7j4P*ZV4^%VSh1mD$AJ?A>1n|o z>2G3c83ilSJNzKJc)Mkx;vV~OiIgK9yFG|;F+J?YuOf_(9=m~lbHhK>^xX$ux)bfE?g0XhaIHbcGVEo*-jIMZ*IJI%*v4^yRor zF32cYui__602Rlz9Sj`mI>y!dGCLdt6XoT>iu}pwS5UF}49Rp19J+N3Oq5pyD_RVA zI5R<?;A>Z5;y><(0vTLzj+$^68*rdtS`HVaIXt-Z%^$0~6(a!HNwHI}UWA26B!l z1r-~nvnD+&Sdh{1z=4i|iTe^PC|Hp{6X6Rgy1Idh@@K({4GlXEoES)d9^A3p@mC}a zloMd z4CS}MiVY1r4xAY1re`W01E;fQk4!;D$G~BV9Rm}GNpuWMln(|gHZ<%wkbf7OE~wbh zabhCDj>@83h#^9@x=w;t;B1pgc8xZ>rePabhAp zEhx{4VHF!XPE4fd1`9G)R5U!0zhFbCXl@D|7?`+!HCRxvBK@-{z1PJE8+IHxF_7LD z+>uexU8V*m(pQ2zx(SSH7B7fny5m4MY+zz}e~_`FqTzu9{i+jm$8m+niSk9uMR{R_ zuh>vN5MwJg>^N{@Abl{nBcoh=HG)=bXn3I8#z1;egkO-+4I7wPT9C1#qTzu99qEXm z3o;5;G(Xx7beu@<^rZ;pOV*2~c%Ul|bO9#P*Md71lrI~GZq5S}Y}ir05_J__;DLEZm z%DBI|VaI_J1L;-49T^1`8>Rs1<-vlCZdn86y?!0L>M$^ozp54WLci@5(a^BtxX{`$ zkha)$8Rc)I%Z7%>#XpPsBf3>|4CGfvkAjK~%@)hobVhnr)MXTObpyv$#K1)P%h=%+ z8+IHxF_1nQ+>ud`z7fY{K}Nxf?m9QI96`p4iiQXJ`^@?4=8TSk?bR`>h8+in&guO@ zeoYh$Dz>Zm4N=fBFg2t#Sdg!r00oC)$G}ASrYPtb*j^hA4Lc6hE6TTYMt)t?6;y1< z|4&p!S|fFx2GZYYLcxj}J*FY)^}&LSj)93yph)H0Hj6V#+Z&>(VaI{);0#PG9}6;8 zR5UzrpkFw20~6_+!5zoT{{85J69W_J<-r{bG745yY-o63$AOL$0~6^L!5s@S3RYBX zXn0`9fsPXc6X}(~9SbrFR#a?gcwonYjuQhD=~clU3o;5;RBUK?V8?-u69W_J)xjML zG745yY-o63$AOL$0~6^r!5s@S3RYBXXn0`9fsPXc6KM_ZSddY$qGChC13M0MoEVr$ zuMO^4kWsLrVnf3NI}UW57??<}3+`BuQLv(7L&F0*4s@Itm`JY=?pTmfu%co^!vi}G zbetHNNN))4SddY$qGChC13M0MoEVr$Zw&5OkWsLrVnf3NI}UW57??^RVIVqhY@CAecj zM!|}T4GjNK}NxfiVY19>^RVIVqhY@Be-KhM!|}T4GjAk@n3o;5;RBUK?V8?-u69esAx9EZ$ z2Rcp+Or&oIcPz*#SW&T|;ej0oI!+8sq`wO8SddY$qGChC13M0MoEVr$J-A~*M!|}T z4Gj^RVIVqhYDKe%H-M!|}T4GjxMM*^!HS9v4G-)%&~aj5 zBK^RVIVqhYj!5s@S3RYBXXn0`9 zfsPXc6Y0mn9SbrFR#a?gcwonYjuQhD=_kP*3o;5;RBUK?V8?-u69W_Jr@A(L~^otbFH@?O5jc@lzOQakr?!(>s)Bd^W zg!J?vqvC;%iS&%%js+P7D=IcLJh04>Sx)rRb*zkyZj<+Va!)>hFFl=69 zzY@8EiVZuwgY?!+a>#iR>23LHkWsN=hnI`q1}6GT4Z}owRj?eiXV<7!3SWr+etiz4Nn;VC>Vu$rB_bFY_uwPj3aqTr-CQ{rK zybX0;m3d1gIxomq?(0#PQL$l%cWB;CMZK|w{sj*fx!kHLb1iiX`>29|Ha%H37HxvP3xQ8nxsSC?<8!+zy-Y1fG7|0}U* z>^R}QMK`Z3x>Y>jH8VG_nYk@53|4r#%+1SXZUgrh1>P`o^M;w*juR=~_;K@^nOnsJ z9bN@;TV4{Z@J^4L*LmCq?k^1ry#3j!S$JaAiH7OZISobTJf z{pCS{M{>s}DmdZU+Bc8YzEwQnG1xaxz`iZ73|0)gZsPRu^XH74KWE&QZwD1A{%Uc% zV?jp2iVA<{xcNKBt^8#i%MBe9<)exy9}6~gl#eT-qdeqLJ|TzQ0X@)hV!-|KoBQQA z_seeu6&o5J*m1z!^xFxy)NgL7-%{KyzumDQBfZ)TljzjemO8OvHVrw zg*dm04Lc4Dcx6o-ilF_#+R!nOei$q$sAw18kKB%qfn>yjf_epbV?A{MI0|WPG1m0yIUw4BYC%lh7zQzU>4|u2i&FkcE%d>(NUgmx~ zFmPW2@8-UFE%$B532%YEc>(mTqC52l(vN}#1q~eo>DR%6f{KP69Ruk%!GeN{h8-Q_ z$~8Hp4+Z-*7kDxA%{!TI3koW{vG?YUy*DrGy$z&i2MfG#_vVGWH}BfL?dTXt&k4LQ z_f}BRu%pA9Qg7+G!GeN{25;)Tbqu7x2o`wn=FNLIZ(g~1+tD$Qo);`A@Gi((!;X#t zZyLOL$>6P^!s`ZaJG^f2=1qgQVqaCX3m>{29pm~8zBzKscY-vc-yI7w3RYBXXn0`9 zfsPXc6X{pM9SbrFR#a?gcwonYj`HfDqGO=ECfLw1P}ZQLW1zfN9XbZe>(rrRpuAok zItI!c)S+XbyipxG2Fjb%p<|-FSsgkiN>zuBf$|o0=olz(Rfmp=@-}to7$|R7hvqxn zjsuHhnNhH!qFp%$+3T8ayW)Y469W@oO>yhj*8rqf1q%wiXyVqeqhlbwI#^Ip(XgXq zAiXB=!2URvK}CZH-^Wo62GSZV@F4m)zJUkP$8R=4$AE{<$FC28=gr444;prK40zId ze1QxKDm;`tzDxw3M;>QHkX|1wDDbfG_~H=k=omWhQ=Y?+o<3cgLC0J0fVrotC z4_I!yGqzr>pNtuFlwSlD4Ldpp(l3Jr1r-fDI)WhQCk7_c+k-n6WE8BZ*wFC6jsqPh z1}4%wf;$#uq<2`k+=}}vgQl)3Zm4FzqbWYH=(o!q@NcJXUGBudM0#g%$AXN46%`vA z9@uf9uEnq0q%Nbd<26jU_q=om=v4Hoq4@W@q^cgB(FXo@=?e^4hkklq(8 zD5z-I(J_$TA1o+W?uU`fsMxUMz`(@vqadTg+g)xu4h&2zXW%_Aw~7rr4h(o>%5C{^ zkWsN=hqs2rZKYsh`ALvb;S~vSTPZj&FtPkJ@Lq$strTq7abUm;1>&|+kWsN=hsU7b z1}2uD2N@Nfb9~!zU|?eTMc^rWab*oQ>^Lysd3$kX4KgY=?C|8h+rY#!f{Y5!;EQjJ z!Ex0&WJ9jI(KC=z{IFS2P|>iXV<7!;u%Mu#VMoV6`jcQmK}Exkj)C;2LA%bOU&Vwx z?LV%zfoJ^31vJ>>f)ySWc=MFN+X;`*yDjkay!d4$ z*m1)1;%*)lcWc;@KB5kuQg+)f9Kiz}Ck7_c2ZB2mWE501?C2PnNFNODSddY$qGChC z13M0MoEVr$Ex2PrK}Exkj)C-{U_n83c5c{lV8Ejuw6QKI_h;#%CGCBq(+Sh^|2g=ujiVf)-!GesAfr<9bV8?;-t)ODV?x1)eW85tZmX~S8 zj{Yttl0*?LBV#}d5{s#1>jDBo^8%q!!vO_|Lo0XrXEyEU5Kjq1i%Yp~XUM!G6l&9vfXq>{gov`%y;^ z`%#B`d)rMYwvbu~E^@)Y-6g8lVu6LwLbHX)LW_k)D+w&b##RgM&$_mW9Z1!u?3Ue^ zTJT?RA+S)j5L$>VB(=JmTybh4uH9!Lwcx+#LTD3ewh&urwUAn9w-8+HLX(BiLbHX~ zLaT+;Lc4|F5*L~*gch1D#1>jDq!!vO1U6)oh2T=h&_dPd-)vb6$s!k03%)(X1QtRo zX|@nqXt5AmNNTsNmsSg@g?0jDq!!vO1edwcWFfTBY$3MLYN4sN zqn>uRQe(S?)P`)g;J@TTqlLgilLh}7S5mcG3oKM`cOkS}k%iboVxjt+tFvDY_E=z{ z*@FGj&pm><&~Bme9T%z=?3Z~x?Du%?%K$E<7VMXF?orJJ`~6&x$UF`gwhN8#xzJ?6e%|Fiu6Ci#g8iPThy8%aeH`q9{R*W=^#d1LETk5i zK6D|n&}Jd{$c1JLiG{|GT?j3-TJRGW7FdWav|Fft;zEms)I!szE<_gEECl~?q1i%W zq46^pLJO@H{LfujU?H~9ZlU^x3oRB>3r)*hh%B^O2-;j|wvbp@aJ$<9p@rab+o85U zLc8U^>~6(&%YOOfo^rZdp@n7(k%hoYnk=MNSG8L~;BJL>tJy+S3)Vr41^ek#5Bup< zkJu84h17!I;zDR`n=J&!CJR*y3u;*l%@!gHEf!)6trijsZ5C1s{z4Z53snoDg~&o| zq18fS(@5-An}sxSNG$|cSop|jA^zBf(C&*Zq_ymo zcAtgVLTVwn%I>oeTS#s2_S!UTLaE&sTx}&5VhgDS{}q=_YRhi78eecb&;EF^yMowm ztc91{E&HpyZtY!Ysf8AI%l@*b+h`ZsYT*@k%l;Z>kM>txXk27ru?zOc5PLK&abZC% zyykA%pQ-B69J|m`3rpQC`^z{zT3>getrp&Jx7usrO?RvDEf*G?=R%8xHVcjC+kF;V zEVNl@yuj|W&|;y@LgR&YpM@3+Rom2R_1td3f7ONHI!o9t3ye(`+Kv7l zE;L#Qtt7BpRm&!JnM*CSTc|$oLTI7cLSiAc&~72L7R?qS3oRC63#}Fs3vCus3+)#C zMJ_a22rM*Ns9IQHA+*qJA+pe7A-2$JA+gYAA+^wM!C&k`qlLgilZC2<1r|aJ%@!gH zEf!)6iG|dHzr=;WLe)ZOA+iu#NGzll{MTFvEL1Io79tC=g~UQ?q1}QXyU=JMu+U_o zYGHwe&_c6?$U=*S*g|3FXX-wUbRo5#{ny>CMhk(3rrND}E)iOYEunBg1G}$pVPOh&T<8HUtQuQ+$raJhm zy2MHKQ!5RYYBZhFFn+wd9oX#&6Y47Nc2n*4w24x+rn)iH8*1Y&sB!$H(Wkh|(BNh? z`V{MRYQqGZOtTt{9yhXX)Ral%r%kONNhbPl#_kwZH>G}L!^E+ZT$>ilRHlxzNspY; zaMFajDb{~%cRSXNp4u>}wgh&^v?=53>usV*Eq(fw(Nk>`+Wt?YPM=s;n@eisblJ0zvEn4ZI3PZXUS-{*|mE^yLWP(R=WA0B@^5-&d4SZ{cnaH z((=EVX%orC|ICe^G;vJUUF-iOPOGbJX!1V^+iznm>khB+byFLr*N>cNd!p@sD&1O6 zth3#c{xdbDe*9S55&yi8BhX*{AIp78{plnByDE+U%(#`DQmG%^Ft%Z|>nO0)#7R>d z#-2X1Vamwqb>kbxjGQ{@l-l{yWXTzo(IdyW)up?UV<%6VFmm*?$&+;uRIR3AitBaS zNwt$=f!#5se&U$14dZQJhUQTX6YYd)w%cyyCf1L(11PGMx`V&o)MBy&!QC;bV)w+g z{6Ft$rKVC}Kba#m`A0MEYO}jWkDD;b#!PFllYUHto!jl^G1Dr$xZ8e54LNyy!-U!n zYqUGYH%ysY+y1~bd4k$C)lyUH>+QU%8r(Mjx8VZ2!yPFN6KkinseVTN=;^id+2l5C z^l3Jq$TZdtV}}+4w`03qH+l4~lj|mqacewkn(gUU8rO{(GnsQesr4|q-ks|Iy|1lS zR@dMBM&Q%v3kIlMbq#hQ~Eb9cS&PxO9`! zN7s+9cNZ&nM^$&!=A`n&r%kOJRXb7^{3CZj#a*(;)aK?|h03@|)vJ-aulc|C)&2Lr zNcZ)(%lJW)DyR3WpE7!KLxpQ!%RjQU^%!F_igoWlx1j%wsdjjcKf1m$x!xVhZeMZr z?6iJyoW~*F2{+^Y@RRs(?60r>#^AOM zVRZpDY-1+=fjA)bMo@#Q$e zx8oN46przmxWaf}<5u!3xgB8hP4LFJ4Hs~V2jCg>Hv;>38g9gw;&$5Kf&=o8V-GLI zO}Gtr#DC)|-q6k6=Dz^<#v$&9oAC*FDD#<&$Ki``j`}L@iXX*!yacy%JbaEl`uhWS z#Otl1`SzmycDO?R09+tH6!*oGaD*?!E%*jp#1G*a^#3Z(;ZN~E>VGq@>3M#qqig5m znybn~aaTMJ?}2CFqi`9I!4+IFyJN82clBET=i$nE>(zWY_J3ck=6TrLb*q{e;%G0q z!)mU-E}j>UKdP4B#_akp?5+Hv=CwR8H)n7ye;oPh*tKh};CPIDAx=+{=i%}x@*}u3 zNse)Dg8Vg(M#&vk*L;fO<;`(Yk$1;2^@rhTit=M|^(=W-mY*tLg^M%gdvQqq1zee~ z{QH?_$n9pge&GzQ-hVnM;{bbQ)!&9IKG>jy{n$a@R^5ulSLC0Iqx?H)MWGz6M8a^0Qg~bNMS=eN|q6ZT0WP zauHV_ksEOSWjVm5XXKWw{&D%Ito~Vf%be@q?c?OtBWmlvr`fG(e5sDd<8X?n;oKbM zuQa>uFTbSzp1=|Q7{|+%{|zTE$m_c&u(jg>7jXEZ@`qvndHG}<;EQm9_6y9mJzr>h zo-@0B=543P=P#%)u|L;Zm+`+>|NEF-ePc(}*WmzPV0Qbc z$omf~|3~I#dA;>Cer4S)YwN!Yj@FV(IQ(1nlW@9<@>gbFNq*Gqwm;vZ?T^iFf0i!P z{`wYI@EYqYAK;!i#QkuL>u`$C!nxV%|9V`&&A5c$!r>^@r#Ql$Hqd%h+1|}@`K(QJ zd}f~0spjE0z%y}(ufq|39G5Or`}ZD|b}!giDty|0|C0 zx?MG1DNuf=%yZ>|nXiz`xN@a@1`cP*O_{HeAHex*<uU35@oZ=y7cmDaW==^Ii z+x4_T$NL31!1Hj3AH>Nqs(;aJ*Vj6Y_a;t8$=^|*Qr~e?)yE^0-wcN%<=t_vEFXze zJQ3$-D1TAb{!IBkTw;I4naO{T)0wJYcQeh$J4fz`eY_Vga(*74?&CL0HQ)bR*_pnqu z-n{eFpL-fobKyeyJ?uBh?rB&pU*&x4vZWkxK6b~+bk+CA-f8j}oWtkh5Z{9H_<3C1 zR?i<_&*~4%)vk}raQPtlFS9!zD?A?N@|s_Ifb!k2cY_{ZcEc6&18`w~)t_K?=X0)6 z=jSA|9Z&1%csh&va6i>wf{VAv_u#TG$65V3^7ptxe%)@GuYbPseQ|PuJRAq)eeC|F zfZBT8j=h`ZSFro{18Vv2viwQ%I^8v%H%{IeSI5hT;T-u>a2{W1cIzFo-t))@XQ}-Y zIK@kGGF$l{us>a1vxnwWAipC{PgQ=9+3k-)z4ph62y{3Y_rTWLNq-V#^GDqq0qx$?o-KU*GccIy+dK2vf2Cp}(VnE7Y9$?UEVJ-I%F zxEFqf`tlA(*7oOnnG5n)nRk^}WbP}kv9;z?-b~&kb6(yq^H%a+nS05Hm^--l4-VJ% zO~E0akNu;Re;%g;5&Q;*kq+!xyvp|KKUrDF`$lFresPU; zYwNKGE_9KH;aq3=EF9oFvwSDz7vj=3^13_F|JL$woRF{LXe;GE#WD4}^;Y{F`LnRs zL-j4VO1{gEs`tAqe*_MwzX`|Pl>Z7xd3o2J)V@sn8MsLPaa^MPngzz&LhbwEU^96Z zE^jP9j0;`m&#||uyv@$)ui#$D+wpk>F0CVY6x_n5nQUcmYKn%Pan=Xm%J`58Pv{~7yu^*(Cfh&RR|@3$0iK)#>3vpYXu z((ySK55%+0Zv5Q(+Me6U4<)~leDL7tD^<-z=i$pIdLtN2D7;siJ1Eq7P@COiT!z&GG#`~l9pj%(y87E4YHk;bvUM-kz%WQjOOSmvJMG@FJZ5UiG=XTzj{_eXei4 z%x**C2eiKZaGw1&7MIvx7i4}!^>^bc`|mXz%~ih5?4HlIt$uWEz5c){&h4#yJKhR= zyx-N=Y{wt{9c^~!M>txKj}7Fj*J`|Ta1_dSWxiZ~3Mco zU%W}~jy;aYea&lmUK5WegUS1}KM5x{s=be^Rrz|_FQEMcSv~Dv#Krll{|I~Z_Z#h- zX`kDd`LZ58a5zux_raAL*Tky`UUbz`)U0AZ}QeS z`AP1ND?iJ_v-%$T{L#n$mhxR${TA|49PcfA`)fRJFL`@h+)6$Uhu!3hanwhC4Eyx| z87>c1e%m7b50j^2A3urn$0^_C0M&ez59a#UbA4 zAk_!QC_f69kCyMi9{C@zkN4`Q_W7e!e=!d6+c9Uj^2NAvfcye3?kE45)gLJDGf@4d1LT=F z?k_jvfc^U(E*!7?N`sgW?ulbO7*|eE{VbdxF5hN$k6+c??zQ#&5*OE$dmo~Fc`f-= z?5{09Y8S)j)y9r;_f)tRr!5zVH0@>E^Q`P%(K1A=QjKv{72j}qhxH>@f z4`m)Czh>^@9zS^f^A(Qqs)s9|;yljr{Aq8qod`#$zr%3uD0wuF@R_(gPTO-U_D_~y z#$lcO4Nj-aYagNU%D%iG4yMUc4o3T*Bp%@(gol z_xfp}o*({;e5qdf7qa$ze`I-<$D5RxKkkns<~u6Ok5YfL%xk%i*LXjoN4VJCCh&$zmKcW%WDl*f6+_wzPR{`JQ@3oaYB)yqDSS?}Yt5ihTSZ<*&e%ujLlA+h5WBgKO8bzs+m9&zbnX(UwEx0xse_9%FXr zlmDLj3$XX8{9KlQU;Zug+w#`OseSy3d@QbhC|_)L^XbZX58xbrmwNwk)&F63+n-;e z<9)}W&UU@vdNm9;cfY2)Qrb7Wcx3o89(T*q#Yy*Kyu`5NG{2;_`*^ojAd- z;KFR>zc#z^^Yq_!q_cZ|5b=6%yxDbJd0X{Ym~H!h)$#ctdG8zLTXFs!`41evAm_{K z&%=Xo{ww8A!qqmp5hw4;_uy!e{5lR&xlpJ6OCQN~nLm}U#L>s{3poBvZpXz>Y#(Nf5N2&hDtiE3Ed6N21&X!BKGE<(3 zy|d&yaCpA_2F_n2uRf0Ty+qyv`;GEwoKBVJ;_?*vc^otTcUhkCH)&A+(bcLyB6BP^ z;^1!iejMH*e}MhFt_%IGlS-Zo>ITObecs8l;Y zaKiXAvpnP7mwA!uzr^wG@&@D8U*#hC0GxYNo?!0m-Y??%dO7)WH{~C~$w2vC>`#{e z!j&839uqX4H$gtc?5>9~*TV+0dwwyH??+uqJ|+Ke@-xUUA@A|{{w?_?^0|o`uW+o! z+X)wY%7^31Tk@%yd&&PYuj!s2G5%xbwLEVe{eM8dmHg_HlpjidceC43RmM9JhX<+u zvv9eeydd+Na%fO_6u(;AvfaOeeylH@TlC1E1}%+4E0xBAos@QN93W{yIb}%)Bd*1^uGiL z52*eR9K9xYo5^?|%7@`_seC4mBKaoleIPHwNsIgw&eMM5Gu6LG{|DmeE!9`D{OVfY zt8x0J^3P`Zjg|ifS3Aj@%~JoluT*~kE^Vazc$~N|2HWG|Rhc^~{}>Lt%AevOC+E&m z|KWFXUtC=-k2SmVBjWtL#O%(GUYs8fQJ-|t=hJUuuan%4~{$3ts zcH>2S-}`j(Den(lVs`75@1oCtZp1Bo-t!psr61Ja`?&Iryd39#lslcH`NZTm$7SZX zJFXm~*9S*u^>`XC9IO1b*gIN&1efum%()F}$J5tYJ??a_wm;zWv@OjY+~=u(YrcD% z-Q!DX%u%)W!*Cf-qds0u{Rh}zQ@$PN){&n!yY=kK@wb$G(2@D!a6S2Nvm3w2c-_yF z3%Ebd<56bUe~$iVn%(vkR%m-Jr9MafUF2i(&*F&rzJ>EEX+A$;e`R_7^VNU6yN<`* zaAj|Kc;;Q?b8xzwJRe82e>}_gRsJpP?JEC)6Y4j&N=D8K3cb;vDVw#bGzKKgMjwJKI-JzOuFQ zXW*c_d<{-Fkng5_h4xR8&u^prhnX+e`^8?P=9g=fd*bp1^1(Q{Odf+vbLDfg`itee zaDo08V(%*Dzc#z|$+14GT%_^KbCll*`;6BcN8|_KJo!4aTd#oaokBitQh$xOe4Tt7 z4zHG<#oqbyJ6U_aueuzk{q%U) z($>m%xkUZvx0Cze2oJ$tZ{=s<+)?W9E?nMG`6W2Nz5ELVxupT%i4PxYAeoPjG}g%whfZRDK&A%$JL}bh})S!`tMGaqe&V-mJc??SBgg zoi?lOkH2xgByW3}#xFjm_JeT#33)0mER^Tt@~iR^oLnfk;phnUzs}|A-@8Zo-7>RY zV{mz%>MzIsU8;W?2WT(0taP z%lyAmeP3KYPachv^;Lf%PWRXPKb(1AIl=yZ@@iM8|MCIy4!C-dd?NN{X}q&>a+Q1& zF5(uPJ6ic~GM^#muGDzdbL1Uybgq0HE}bRM%-m1&nTIP)%D;m1MQ#5tIJiLh+*NGf z74kl~dZj!CN0-atgyf+Ordw0Z6P{32`rsJzy-8b5tZ-V=vU z$S33C({h;kdAS`|o{@`9>aX~md@)Y(+c=Dr&tJ#-J}ZyMDSjLWjJFI|-&Otk|5E$l zL-`<_UL#M#VUv6huKr72mbL%4+~a!I1J~i)oyy;UEBO5^f4lM<%~SgdJ{ITiQ2uJ1 zyH{SA)xRqLj0=n9o;Rq!{A=32$w&VAI|FC4{hx6a#>dYg}mu}_2;$8gK%*T z9gov-T2?vI9x?u?MC&NtS#@2WA{TI8-F_XJInKNc`f!p`!IINd{zadj_wwVTzxu#dbC z4)&EN;L3^e-8gYSMx;we;mB6{1jY% zMZN`>-jLtIg{AW9|7L!#$$R3`V)+zY`9uB}j`6G5U!nX8TxgfKy<7bU_;_6WMfrpdgm#5ZZ^$R$ z;+yhCnY-!n^&TAH#W>9?{|zp6m)E#Y<0aH@hdr*Z^*H`j z@&FvS%HwhFM76&NM|c4)?5Fi?!Rf(rTjssxbsp4s9^MP5Uu(QNT>M79IIBNTz86R5 z%dJ`a+45i5|5D!eA-3-;c?hmtBwvUNm&i}zQXv0?%a_SJJgok5m&#)@lb?@^S12Fj zDzEQ;!s%XGkIf!Ye^q=SPN_c$ha0Irz?F^Vhj2vx1Dx-wy!R;c#Rcq*)c$P10lo=W zU)TI!&FVS-e#PlF%5UDR{_`W`LD-#g;;T#Tej9)T$ao;z-LGNdO zj;nZ;ra;1C~}M0_&l6DTl1NZ3-~cy!f)aVUXBC2#&eoa zh_}WuF5(oQfOEdapND<8Olt_yAnO!*K z!tvPkMa{23z7H+Sh3>iT&L9*QG827gZb@=bcay~@iPFMqSVEzWJB@rt;> z`{ToLc}wMInb-8Z8GN34EqU*EZSQ@!$o98TU!wja^8O(jzhjH$Q`ukchI8v{yuLWW zN8n&R<;R%a<3r_i?eAGQz_;NLzm7|nt3J)@d3?w%)O?a}^?11r_P)?~gK_?7&37se z7Ab!fE*`4qPmh~zevJ1q4sfSel+Q8#o@RGoRiB#O{GzhfW8GKf827;;KGf{y=XclqrsB%h@*6IFjJLuo@YZ;PB^p2cS;x;_<}U90g&cn)vOJ%!%)q%H zbpBj~3%|-$+E-Y=r^(0LXnQ}#-VX9kuW9`1cJkRczpgwFNBD7EURmq$ISx0Ff5XKM z<({$nFYRd^xUer23n1xUu{M zPB)c*!NDf-&TnWwWzNro%xk*ub8`HhNIsuxd#}XNa=FFq=3nA?Y9pUpL-SkXP31$@ zZ#%QC-w$fPKluXfhm((1QTy3A{Y(4rChAk_pE0}nR+#TQ*w5+uvhG_NFXH`^o@O`S zp+{@I_s28vF?bvvhkbl5uHYtI#`l}u_LX=);8k4cpzZl3^NZU4jowy$xwG>7XKs^+ z;v$b{v&?I`>n+#ctH~!v>H8Ir;Bc_~A@&ZFJHMmxl7iNAFYK)*pP0Fu+=#=R9A;iu z{s;%(tN*oI)qmVB_rZ}z|2V&=d=XCZ!#Lba`A=|VZ+XpkS+7|A?~arAWlK*nGcYE$?EYs?=$}sHU4fmIZi$f zmxjt`X7&52zuR$ce>u+b{p4S9p80M0fyVRMo)S*jo-@sz-Svm>_graq=SS`mogep7 zUwvQmc^AhVKdXMI`jqcOZD)4JfAK@rABuC-Ps8Qm%3p;mZ2$c@KT`RHxOlSs8Llpp zSNcffrQ2xxx5Y(V!qF$HpNh+$%GcpSU$uWZbE15T!!F8q`B>vsk5#@mjyT_r$nsmM z{?shLi#!Jx+SEQYyY-G)zgKaBKc~LP`md6xzI3tnPdBq$?_iGl+cWbH8gDeN^pr2f z{s7g_$NAUfC$Pu)^&ebheg4S2T=iRiqVX$V%Lm|S6ODHQE^RKKhSP6Ue=QDsEB}bu ztw+lCeS%}$=~LxHygLr?NL<0QaoVW$nup_y6+9S+ zcp5I?D{z1x$jtl}X6pc>u3#!V*g2bIgYQCxA>g?1GzsgohjGh@M8G_T$wD-!_{l#mvQc2ayu@Z zEwBHD#;b7t?uO&jl|KxZXUZpG?@swb96TyNWp>9~a<7huk8trm`EOjs`DLmvhsy7d zbL`KtX18$3&N`m2HM{303-~=jhzoolc#+xdul!u?uVu6^FV_54X;Xjx5_tz4bXETY z%x?Z=evf$)9*1X9U*!9j|2DhpeUA4-my-AM>i>5f_mH>yQsY^ZlQGdm4ALzp6a{x|DpC{1dpaz3P9azKQxRzGZ&&KMa?4Q2pgN z?kz9G72M%FwGZgOkJ-(qk^aZx0N;px{Fd1kp8c`Xa=FTS?TE{pYy2Z|yiETd#uS|F zr2L~<9{-6ec)wKr`}jia71aI(9FSk_d**`&;1pkgbF^QK^LWi4)IOqqATE%fk3;gy zGLzrwN3}2ES-6a!#YMc%Pi!9^j;r|JIKb_=g7^Pf?JI}r`@Iuz;Y9f|b7!}|IA8A~ zUp`*>7MvR*e~bNNi5{#2act8j(+HRJq=s(%+(kCXqzrJ?c`ziYgN`Sr)qvC22#{12K>fUDoj z51HNhlk$D_#pKgFmH!@Bwo`xW{^9D~@t)4n@xDFwc)TBAb_ZJS-)cV*S8kE7#4&y- z^T*1+gS}NX-XFNouKX5%s{j1z%I}B6mE;r5?);5-{xA*q<$S(``ecRrdjQAyZR$(3 z|IzHW*K5%B=Kj+7=~rsM3r_G+W>=r1ej@qG*Q%e3%lLk?n{S2jUc)|KMt!hM?N|Ss z@vHKtxbUOAUzTTokHEQ~l%Ij4HML#=E^i^ZxY=!Ane921eC|)>ugLs|9OBT^_^)O8N0k2-NB7BHI%vG|9db`xdPY6~ z7eAHjapghzA{^f)-+~Lb%1>qW_sE}S^$*Fbt)%h2C*`ei>i%I2TaSZraKAjt?A9Y> zJZd;|6IhpK-PSC-2k;^;g1cU)w>w_I8CDd3{njTg}WQ1ZFioon+uAII(=_Obci zo8`}yU&qOG`4{Y+DR*B*;{~(i!_Dq`QsVmnyEvIDue^r(_gT-aaexQn^n0!EBpm%9UxSN3%8%m`{y1xoJ9gH1#h+Ba1CH>J zEdR6eXX45;n(xiH8p%s=?m79lto<))ziAhZ7jGr+nR#n@lzA0*{8x9Y9e;Ch6+dcr z=id%|p1c(2IUm0=+vC&8dOTZwO}T)#!XxDtCku4fP9WDWT(9Cnd^&f0gC*I8TR zhu!49*#A@>oB09xN*w>L`8<-jUHK1i?j7ZS#c4<7H_BJx$ir_4H|6TvvV%m+?=yCHwbC)=~Qy7tC(qiXW=~V{pjvc{(m}d|rjikE#An zvpfE(pB+(K{};0M)0F=J`(x!47uerxtgG=$?<&6)&VMfV!@(=^Nx1N=d>)P_YQ3u1 zUs--U%X7ZIi^H$feg!V9Q2iF`X+H7y${&I&W0gM@=U!EQuG#I6nCF}Kk}v$J{9^3$ z`0zC@{-Auv^)+6O{l7J?@_4a74pyjs1TMZLpJ}$|=RE(s*6ijJF`xUXPk4M>h-2pW z70z?~uC;;YQ|+Mb-3yn7$j9KqYOIIZ?f*RcztQaGlg!oi<`(L!KWn}(VgEO^{|4vT zzD+jN`2K;)7jSxjd~oI+7Z+K-dAM+x@(+EXvDq`51Y%jWvEeSl$9Btk;#W8CC9WF0b{kogD`PlW5pxdXwe?e?!fuQ@{2h4 zz5E#tH`VXCS9SluwQXN+CAk+4w~!CU@s{#9T-i~+5C^^G+i|)=yU%Fkx)UzXc&`5L+F78YE3K#I%xP<573VzJ&&L@9uJ->Jhk6TUt&g_o2 zD(k&^H@SF))@M5$e=PSmyZ%a@)qZ%EkCdNEzD)lYncaF;hid+J;`XEDr)XbXqW15Q zk6Hg8&2GF3{dMWC`2~%t-x9|=>v-4|=RZ=ul=)J*!R-1gF#n5i9^Zxo{3PwmeboOO zxWxK=XLj4?9j)zIxd-!mL*sQbyYY(jcL4VB=q%6k!834${5)L7kDKlKaE0E_eaGyM z-y+BFmu8z!p!uw_m7K?0;&i<7yW!jf`A}TIb-092!xcOS2ly7Vo1cG;=65gobg=q= zk$j%{$65Vc)qj9<*U8JwZvFf!#stgM%LOzFB?^ zc{uhv%jaNk1Nl0f;(Kv^W#wCNVLkZ^T-{GzeH+a$x2e1ZE^H?E#raL-qi{m~IPCwe z$Dgxs>Z$%(TwpM;N`8@5f!3BI9^}(Oo{t(Aosr^&9`j_%caLE2i zGOw!o_N+bY)3KN4TVXz1;9Rjw?fBUZ2Zzc7vihUtp*T54o`BPX6{13RiuiSB4wr7yMInM1V?~aq<^1(R9C*TN=$CVRRKMVUK z2heL|m96 zPs!>h%d<0AAHJU1K5zZ3XDz=4`PM-1FYbT`w#vhB^-1|G>^&vlh!gS= zE|dQxGx?Qv(tNtUr}n*Y$4BKsIQNM>0~hgKxaD2tKf($5H45s#?G@$sz@<;+Q8*%h z8D2p9$8eaa{u}J$?mMgh5*~_kpQ(N}j#$4Ja2xKhi`oyQzJL?@Yrw@X)Zf$C$LsY` z`v{N29-fECeXsVdc)@acw_Vk~{ag78T>MU6jzjX>?#6h_ls^I&@l;%WN%`xu`sd`u z*rWcZ%x%hVzPtMO$sdgK&ntfxp7FDM6ZXh2!DakA?)tXs_ufPOm)??n9FTt*_aeVu zU)2}M55ui*s{M_42K67~g!--br2XfrKN64ok9-+!q5a!<=sU`9yqDT1tlzQNqy7q9 zp#BZ)<8}5{`;_`2xFh*0_F1o2aKwE7#?5TcLHnq`D*c^@$B|!%ecJzyi=1CO?W^_y z=ldym!K17v9)}myp;@MYNNdh;kQ;!p4b`d?*#`dg&&cEA;U z49+c9eg^J~ug6W)KZ_&047cIVMUB_a^=KO$HLL%AIKiW_&+%{>&a*vBu$Nt54^aQr z$JJj??Bhdl8Bf68W2&EnD_pOh!A3v|u;+n{_k!B@#}OWd^R)MInd{Nbc;K&U zAK~iL@&`Dfzu$5DJIZfzkjC#yfBWL5Uz8t?W9nyS=6W|Dr_?`%7f}BpZlu0LKem_o zcEjx-X}p2B%zRJAt(^bY;1=qi#L0_l|1r*UJ@xuC-ZRSg!82lc7|wBhYQ#kze;>@u zdcKe2*VVqu!Ro&+?fc?^Zzw+kyxt(SFSA~U;{3Ck@40vZ{XKyT_&40feD^&>{f%RPosPRQ-d#AM z|JQLf(s;k)68o$7q3W;7{y72nrT?q5JoQgtpY8bsC$wMlF!krLe+S|G%Nl;E4RUIM4Otz@yc^h_A(EyaM}p z*fFY)@I%fx6*kitLXYDz^Ryt1o7kPfv9T(VtBd|~ZbFqhC#CiN3PBObK8?t{nSk+{nG zT#G#(kDt!W@$fPB$$KZN|03QI=kWo}aLD!78=?Mv@&%mW6LFs7 z@j~q3`?5TbFK=ggyw*tOPyG?NOy1A(_(9yl|R(~GnTYv2HczGF)n175D z#{WC3XS_n4`fK_~`+FWv8UG#J3;%@^+HW^X?IS!C7kNCp09WxtxQsu?9^Pa$>&5lp zK%8*?Pss9Ye}K!l8T(xCKh5&E^BC5T<9$b*=Xf5Tnd510X4>D2BOK!*=X>9J^VR!yffJouu|roA%dL*vEgE-RA?|N=JJ3e%#vQNxmo?+^4(@PUcmR)pCj-8uHWy!Z+4$AEa2ZO z_?~>>59PZwDBncBm)ZSZsENN1agf>GU;It=Lvj3YNJ{9E-; z;GkVzgk$FOB@X{oe&v%jer|=lq1p9crT;#-86QG@b$h-4cPvhO%VTh96Zt$`#y8;r zzlgm}RsUI*-&pQ&3iBhsC3b(Fv9|yB!6iHrM|c)4uA};JESMApu zula=JcQm{AJ9GEzcrD}nyYdy(S9rhlZk*8H%h-M0zP3FdXZ7pLzhwChL-&>iX_j~W7ec@xh-@6p&zL!&6{Xt%TqQ>*c?}_~{ zl|M4ezb%i!$tUtTIQ>+t z^*IlxxQcVzDgQhUIbIW-?5KQ)iu&_*l6&D~w7x%c0**&(d*)=OzYyo&(Dh@n+3lY^ zzyJG!e9}|>|DDw{UXN2XeuM|%Nls{pL>Z`ZOt4_ra$hY9~J@UY5%6sJ7apB*}H%(W*e5<_sY1CKb%W(Xl zyvq#gAC?!H-SLy(P_OUSKOL_x&%()vyVcgC1(!J=cRfS(#WhqvJ##1dU0mRN-eRWe zJ$yb+A6Ea1aftUgllf489!^8mr?~o~yz?yPgReBZ19Z%7< zgKPO~&F*~Y`;tC?d60Z}t_MrZ_V~A%j+bTRy??3w?>K3a*F0PGUU$`Z$Avp|JQvOG z`cmZYy9^;;?xFfqaDH>S5hsu7_Y}9{IFz5oRemqhn)y2Af5wH2<+XjyFTFsOZk<~)%b<0wf;RaKPnHx)koxdvpYV#?D2qn@VolE zHfvA+PvGL8%D+eZfn2YCHoM1bT>b*bc$JGZe#-bgu}6PJ zoWsL#9#1yA@v0o3SK|mji9`Ge4)E_;`)71MZ+@}*D?cmmkCSE2Y3`N%u)VAT*PSNU+Z%EKUVvzH}+0d z{s>&5zezY6r~KtO@a6k5H^_@|{wekU1FqnW0*zm)sJ@6(d{UO5sC*+1CNUn)v;Sfo zQlH}Lc-5~nSL5YrzdJ6FACAlHzgajwS?#M?{uKFnT*1q+H&@%c?iCs@D9e3t`9ifH zioM~ok8}7=oX1OX0so%WAFlqlyi()EhspiT_I!)KuUC)z;zsHVV>SLAxWxWlii`Mn zT*mpU)c-*GI{=q?emBbO&i|`;K5;%S@%-Qpvt3`uZRolGcuR0}iu@<8Oq6%Hn)-@7 z78l0Lb8tF}?ZK53<>ztn8C`Ea#}QuX8ujlztNb>&j0fRTr2KeXeOv2&DUPUr87J7g zR{f`VPaNY@apg(%e=#mUCEtXj$K@z%|AhP%ExUn(*QD*;>^kM6YvloE_j)qN-^VzLd~$22+WtKSrw_>I zVgEt73FjV?AI$2T<%Kx8Q~m^pcgug^;yrS=e`$W@`{ccFDU^rdD&tjf_^|R9;^HIn zZCU$A<>zsNKQp`MAEilp{`w29;EvZTAK*=Jh9oh`RZ}L-iZ@_5AYa{@T)k*AK@7P zXtw=zx8}FTe2rHIKaKku0Fg=^}Cwgd?V)DkNSw$TPINOy{Ppai}U^Ue);)0 z9U$L>lYQh@aM(}&0hb5L>)fdM#DnEQxa7-CIGriCVITLcs(lVO;_4ZyZ^Z@N>n7Dl zrz<}Lmv9SCW+>nFX4O~lIGmiV{G(YOuXc;-J$x(<&QkrYIFEnC#aYT9c&pmSXUcPN z87DZTzQ=8@-W^|s8QR|ia0#DccKbUyQvF?xE3EfBIAXn5zg_i}D^VUPKK zgUgpGzt#e^FH*lZPKW6C6G!37aqL`ierH@9A@7S*e5Bc}SDEYYNb)iBJvFP}U+rh&g!#_Km9p~J zVXscUC(ARRr!upCi_GqNmhk(Tw{Vf)w|!52e3SOaA2@hg-rz2+M;`Bp!>5!VfQz_{ z<0q9r9hdPHIDJC-e`6oNfOGgmTzy>izhwF6A1Yxx}H1Cy-RRflyAoVe)6L@+FM?P!#ephoIg0h4pN?}QiOTG;U!{x_uxU2joF7Gcd#|7rU`n{T8X%FQ$ z!$s+DbK+{gM16lv%b$~^*Q~0ruWV6`cRsq$A=$r1$VxW^}A60 zZHnWy)qYo8=p+xqr8VRcS^F7ke=3eHlg~H1>)$v&f2f-6da?J>wd3Ov>Vt1p|1u8y z$nW69eIRPb(^qDd{IuK)M`ig;TxR@H*kwM~<8;3Ar`tH~pTo{Giob_L z@;4dh^<6}LJB{=8b0=&4>b;UXpHIYmTHqKTf)kGaamMC@=uwS#4i27>$J#jaooQ^& zU&edFI6s~#&)aVroAK%Y8`>9MRDT)qDa7l&s_}!g$*%p@20LrzV{u0OZ0jEsAA>9F zibXUE7aX3L9Y<(^DVQ-pz zD=y&KxHMhymvBb>W9-gQd;^Yd(RlyZeC#e(z2Owa+v6Db#@?-pkHEz{)qiC3iO<9t z^Lqw|H!1&ZT)A2P3a9vIoP4c#wbxZ&k$8Z8;+Nor{wLb@cc}fnxQrLz6jyL@n(}|g5#!Zg%K8y+i;H(EzaP%k=zOU9uEq~u5^ne&>-Cl5hv4#Oa&H{(Ef=x>mpm4ia0z?=C_V>AJLN?< zm@j{g%a3Y3{=;Ee@s7(>Uoukehr^5HVK^QxN7yAk7nfeqdL=mNtNhQf*H7Mno#W*y z?`!-b_ORPu@k4P!{CFImuJ}k?I8DCA`c(Nb>^IT;7h-3)ycWB-?gy&J!(DL!pO1Zf zhjm-^_cSgNUt`@;@xO3Dywh@xSK<8Vk4yI}e;D@HX+AgD{Ga3palr9?6}!JE{w*$V zlK;i=&vN?@8UF^2*WWslhg(mSugAe_YX1Os>M8yl_8Td_9J`Mxz6n>FDqim+jqf&= zJL4kpvvAD)*;rhj6L$l+xU0NpNakN z=OSF7w{S!;NNkGYka19O1LR5 z;||#2eS>doo?ip4_ld+SLsjoNHae%GtHjH=6wn4eHsq&#W=uM8Rrel zw`%;!*qtTcZ_WN$gd_Y74(@Vz@^0Yn<-~kM;Wv`?$(iiWhMM<9vOClG^W&lj-siHok|>htrMo z_eK5qKI<^!eE&>2MdwF|H{yx3ul%Qa?;>8lQ~8e)FZNM<5ib3z@hjNtrug?b>n&GV zrTWAFDSvO{ygrAYmvu}*^`EOiqq4w@-)gRPV zJ*{w3RX)@>-|ul_UC)j&HrF%O;}r6v%hmrmxN@C53WpQq>uf%ric9z*9OHR5pX=A_ zI6PnVeTF@*|C@0#TJ!r4r#zqR^{wU`PF6iFjr0AS@%yWX5RWcW|GlvLg8DlfR~E=) zZ2TcP!X-QnhuojX*q^EVdAP7h^}KGJuW!cj{D`xrpN%Mrq<_o>&1%KT&w!yX>v>BeEsIKp9;i-Gc^7&xN@F6 z7>CQ{QP^RB-G;r|8vh}j?k>NABd%{B8t3DujdcB4ZJh5Pm;0qZ$PbQF`x@(1e}&J} zy|BxEJis{bugrW7Bc60p`_pl$vpfzv9pz~_J4lXg`~>-B9QTty!j=B=2Ae-X-iE_k zTCdvQtA4j6?}NSW%I}QhQ{`iEa*ljDE}bi1ii<cUdZ$2K{%SN@z20M z&zn~n=j;C!>p$JtT(8)lbBHe@z8EjVU*HPff)iYOgZf)Uf6a{Z>)AzIzuFt;*Vn>e z)pw+EUQdDZYXI#%u7_unUn+Wzxh#*x?f|)DeZKr84nz4JoQ;;(;tKJa8>wfk;$3ib zygUGhU0dxMe>?rveTxc%8hl|I_8?ZZ6uD4nJ$LGpj za5?{lXmk9AWA}V{D)z~L5vS)Uz8ZU7<^OQpPTqHm#>)i^B=|1LNEmG!{KV((D3zYG@-lOMy;;qp3b=36UMf0g=*AA#eF#yc4YpUPKY z|2_G3oGz2+kF4A3c+ax!50Kxoo~H4> z!|pXY->dx2@fk0-#)W4!-ci^cE}w1ltIH8CkB}e3l@0PT9JWzCKjP9Ja<%_ye5bzL z3Mb5`ACBj%zYB2j8F><2^m0!l;NclTl93{J3)n9a(+#Q#g&&4>nRPh-$ z|08(`cE-!wuuHv%Z&Uxte8q?2;3fLQ9$tc-7Zv{lr?1Q2cJ-H1{{UP#SjYETTz*UW zPuloW`8yjwPp-K`{bdKKeHR?F{zY7(|7q6r_aYA8Q-ABRbB$d0FZExTC?A1KH^>*@ zL zH`e+DINGlBc?!<(E7)zS{2y?P_u5H+zi7UPV>gor;}}P{yh-s#aQKt_F7}9T#@U;y zui?KMFWRX5uGoE7@w2eAL%tb%E9GZ#yjK1c`#;M&Z2UX9`G1V}r`!t%U&|wHe6>6i z7dFU=^#}4#xcHX5yHi!~1KyT9;cS8CHyjt2D*m8#6&>%laj?6*6{mIN_Epq>h=*VY z&%g=o-@$%ewf_yf)U$6@^_Q@o18{L)?T-+9TU7r{T>4pl)yCN`-(&ZGir22D{wo_b z-=nbqlY9}5{!soj?EWb)!eJHFvksT5%6nI5J*vt5aZp_zi~WB!-ecIQq4-K%*sk&Z zz}^nINe%U1#z*6r_9Jk%oBErIqnh%|*xy@Tiz|Pro~FC0fB$c}AC8+Te+(|{BgeS( zkLp{7lbv#ggMa1bHPwIoo7@AJIe&{d{YvBCi8I~X(xElxgEd;t#smDgd{(eY_sU;SlYD<0tDkMeCe z`JeonbycsoS;J((Z#8H0n}6@Lbo>&d_30WAIX1)4hkNHo+&QZ!=WaB;L?bb)gtsAO8 z|44Z-PVsHH)Lrqnte=#B!{IY>$42TeoFSiw6ULv8aer(4j?#Eg-R)5Yk z<)48o)8%_`VUqkdF5MyTz~u+zc1_fuNBzTbd5+?tT}(mIBuZ+zsF%qdC$Goe};QuZ$Isi3vsc9d@Bz2RQrY4*-QQbXZvWr zyERk)ZgaUKc5r};p4wl5EBId9zLw&PafDM`;`6i4KCD0Y1IJkN`EnU9ay^f6*irQ@ z#%U+{dt7QS|BEXf5)RdvLL<_U8*Yx=^2Q6@ z3YSO8!>}`2o{6)I<(F;yk@62Xoh?^uss5cu@wvqwQ<^S!s$hd z*KMW#%Z%R{mqNt{N&W+e#GAKfyorkU!bv2L!Wr!!!iC9- zufj3@5Bqm1-fDmKA5m|BvxgMF9*48!C$YOt$Ny7Yn5y_zT$&~y-A4VVcgoYTUy|R% zFx08< z)c*(^;YrxLPVuK~`}5@waPca6D^AAAEe_Op$(3?1?A|C}h`pQSdvV75ln-qC)~Y|l z5$or+Q~zb$3CH+kT*70q*Gc`&w)yyD?DF}(6KAiezJuCpywbDs88|4*Q*d^l{1Ofy zmN#Mtx9Fh$+*Q>1Kf70et+=-d>XzE#~&(QGR}Yg<&M?&r{>}UPH+jY!U6sbmvNng)L#X+#u+{e zyH{wu({TZh#sR(shd9P1{ECg!|4LjYz6n=wwSwx)aC7Wlsd^5>1$+_?@P#r}1~<0&aY;>J9KgIK=&M2@l0(d=0MPyKsi*VfQMHzZ4hncR0Y? zaERSQR9^|V!)4qPSMb@^XKB1~HXq+(Y zl{@JCZHo)MUpo<3IxBt&j*pjbv+Yll%eZ)|`~}WVmjAKsc^}>CP>rAUQ@jU`uGjO! z88}T9zY-S*$y0GSMSjFMUyroD)?*QIf2jIfVcY*t`I~U|sa*Xq)#H96x51^ybp7aI z?-m&!V#J z^YbH~rv31|abAzd`mZGJKC9!i8GE?q5o(`&q4w=@X|a4XE-#T!#lcc}GhC`FYpxIN|;L-bZUZm-nYfTl0Q>7>=9id3z%Ec|Y<9F7p0- zDGvYB`uv2mw%TtskI{IAm)SqK%=?!UaKiiH(Z>1yXwLq))yDbx#Pim?A6|vSyVd`n zHqQI@J^j4De7_g>*7d%*oF$8VAEk5T>x z zo9ojniVq{6aXwAPF4wbp*ug7tLjM17k$BVNG`}g#r;BmEUJ>~x;{qOqeS8-V=x+gz ziLbyVyvgR{`n@!shdW@O{s!1M9*rY>4-WC0xPaH$IQ{R$W!$8<#;;)CIA3qSh4%j_ zT)+<+=lk(K)@uRrsI!jmm&8lNGaTc(ebm1 z|49Ba@~igMc*%a6PkZBhfA7or*%!CKBghY$YCaKmnExZ##cvts`=!Kse@VR3T;uJ) zDQ?zJ^%Ob3kHw{DG@r4!%=^?Br!Cdr3gdje67K(gB<^fe`?~$rf4D+!kCQd>@i^e; zyCZSoTg7k1?(gz5HveyVIrjdLf3fX9m)+wve(-~QFfOi_&%_?*^Yu7=R`D1+iToyx zJk9?H?6V&APtbZS;&`{m2|nI9-_IHAdogzV>-gSIeueyJh||ANbnjPPTobmZm!OnSVzZS<`mA@4ij*#n~#QNY?I3vF+_PZ&6 z050L7IK|_zd!6#{#?dO(^CHd;m*2;QA@U|1o~ibA2CBXa_3ej~9*Q4|onz(xxY$P? zVw}HkDe(8@ZX)j8sqtd$ca>kZar_Yu@Q=8_``JH?^YgvT^FiG}PH;!#e19jrzYMU) z``9au^XuJyT;EE>!_QRDd|bXu^I3+SVY(ms8Ap6Sw*JW)zrg)f6rw+`%}nFWO)C_W^N- z`bQDZ{!l#=aeAl5n~k06@*?cBUfqJIEi;@o9Ft>N$|>Q+J&3yc*zs#K&MC--SK= z3J&l(+zr<_L*o^2YvX)k3CE`|c6h!GamM#sXW zgG+e4bzQ~Fxa`QQu=nqNyY@q^!RpT=ek?B3RQ^aD)sSZz=hw^qdH=k?*xWx2)BXPE z&U-K;WpHcr|*rnd#*3^4Hj{i}8uj3N+e2atX>i=KkeEf{RN7(pmjaMG3zyH=1 zhi~cc0}jH@+wxU7!qagvR{TZVewO^XZNEg`Zk(@2k^jH6-Z>iIpQ3mHd$-Cb-~?ZY z9mc!a`XS{%f`i-SML4`gUWwz|<;~c=TdsAk>T&LoTVelU`DE)WQF)^E74mc( zjgz0p>2>nE*t=R@Yuitdw;JdByI(Es@9IP30Pl$h;nv3a`4wNMKx_mg`W=i?_=s=wj56w0@epV9wA#8c{d6=!qw z`%$0ZaIX9dE}BIH^q;^6@0$+3dL{4;d1$5?EWGz#?A)$6I|XT|BOq- z_c)*Vf2?>L9MWHJ9Dk(vP#pX!UyUn^A=-N1kZgH_~`B za76rR9RE-Gi*d3|{tO3y${Vp$PyOw%@oI8|5t>hiJL7PN+V{l4&+<7q#p7`W-(}nH zuKu5}@uu?YIKiLb7;nI3{0|OVsC|nIRIjtI+ylq>TwK~u@d?(R{J3>Xxq`!1a)y)E za+3=+UZIWL4F_%I^Kf{8d?R++$&cbPe#^FRuXt+RL9TX@#`8PMt*krA{j597W3hLT zJOhUXc@a(zmcPPI7x`c8H<$OmnEk+f4#D1@iuW z(30}xz?Ot@8E>< zYlAiCSN%&he!}_H0hc(xPQpdbr;*l&sJ>~~+e@BQ62eG9KWRhpY3{^|BO7- zn)73h^?QoHVBJ{#B{=>?=gSwk@J)kVpRXHmvP#~LBV7Bke17@=54UVD{eR<)a0R!- z9{>M%d+gvY_$%^{HqNhS?lOHp=uG4M`VqXK`^QU+^YwN)eV!2MJ@yk`@R=7}2?vBH%^65CEzcDsGQt?}D9M8t3QHsBc-Am;! zY#je#JzDX-E?0e}%jAyO`Lxci{_?-MK&d6Ve6Xw4L7w~rMKBM*xu26l>e7OrQo~Qaw zz=Z{hUxbqv<*CN`-#3i;|L$MF>GH#N?T5A4!+Tz-{PY{ed*b41c`T0c!`Q>0+kD2` zZJhe^@L@P+yb-vFXJG$hjkg#(_*a~)P<-F3)W1`a`{9KA2{^*fV(&xcufQqwXU67z z2!HRf+W6f0^`=nPdhUaRR$BjdIC(RQ@G6Z7WZ}UVHgAoH5?xI5?r< zuH*GCjt0sburolebG7RAPLkW>%9cHLwLj6C_$ZtIi{dZh=x6yK9QSZ{wZG^Z_3z*n zIQ&)l4XdlhbOB24D5WR z^`C=FowdFR4m!zSVtPUPpFsQ2jw&`2g#B@-evL%4g!FfqW@W_mn4L zr=k2XF6=G8Y~xMk71;kv^IeaN%@qIJ=C_cWMyxmILjeb?)V@D#O}+IHjHx$NMP09T(ckyG>Mm3ID%!3mkl?{tq$E??<}T(DOrY;u)WR zLvVSG^2cC*pzg;eVdr3tcMq;qm!HJN{_>kR>@9z4^9Ra5;`BIqJ1!j~*Sk^kiF~;= zc21GIV6TUK91f3_2U+)#M`QN{c``1XB+tU}0QqGbKVJSAC$!&y%YowCaB#F-Ym(}Z zd&>Lbw6A;^F3iw+pNNa6DLxWMXUMl=|8?a*VQlW_>*;vDNIc>5_-&kiq5SpO!Mok0 z`aF)GXPke&%;o*;;kb+kke|J){)*UPe_exvPqbg|!0sE$pNk7`${*SIY06)NGCl;mwC{;i{=f25 zab>mEXC!tmQoXn0;AzF@;P_VgRb2Q){s?>ApMGbYAK!@M`-{zAtoiIVMfF$kzBs~% zhC7&lKcR6h`)fNR}^201D-Ed<8)um{}1DQJfHDu-l}^2 zqT26+OGD*DaXLcn2Vn0r)jt9!cPoFQasE8)@c*SxBObJ>vFm$q}|{3$N( zUu{?Z|FD0AT<13G?W^$)z(IG#ee8{KcC{aYOYIb&inBKIQ@DJ9{4Vwnlz+k@^;ErG z<7W-4?&^PE>(%lRI9w~Ah0|~3$++^hJO>x>+c@5+_$KW4)%*WiQ#F3+RmJzk9zGP8 zUs3!7>~E6C;$XA~2x}#&>W$`l!C{xZGPFjMFo!>{_oY zaAA-<&Bjlc=i+#Vyb_nD$yKLmyvkj2XB^)l55WQPN!Y(r@wwP7$)DnElU(gi^&fAP z55(s0VCNqT&Oq$1maoCdxAJ`3KL0z7CV#b!ez91PjJvr`)e&O+@-j4m&Q-$ z$a`aFrsjJjj`1Lzyr=we*z2YKXW+{Fn$NQ~zFc00vybFou>Y~#>~7WLeJCG|L-L2) zct!C^IJ-oSaeT4-x^3S~^Z5ps-cbG9u(O}y&1P!+q=|eeE;W}=!5R6NW2dd+w_~@B zJP#))tDbkT`JR{A59@Gvx?JNPjtBA9xNw@{y>Rgixrk$~?>FM)bv)>ltB>$W?{u2B1<-c&8-}jq%n+H`-l3(|YeQb`i@lfm#pKP7~YvPaKLjGK4 zycmc1bF=Z6*x#=CZpG&B@|pNP4{7|;-*OL}#A<&wPTBuA;NlaCKW5{P%kSZ6w!GfP z^XCFHUe#F~-{<94IC()n4rkBFLv8#ic><1}mLI_WGx9?0E|fpP<(K6@aD{l|huIG= zD&7^B7RiHg>d2FAem(hd>>Z{3`Z}&~|M4|;cPf7?j(?XM}Rt3sK#>+ z*8Th8xP;Ha;US9Oh*LZlyZPTuHT!!RcJjaDX}kgZ|C4JyrtxC(55&pSTHoHdG>7%Z z{uan*Uyxn#Df0Wzf$_Du)9J%sXZ2pdyncqy? zeh>L&?Cqw{m(Ott@38G_E8Z->j+pV?{nWl2E*&I~##v|iaa`c@U>PnF-(uVEt^E4= zzA*hqKgsQ}!~I`xoJ`Z_!v!{8k|*QvRMq^y`@O1eATsTSYg8 zSK;yq`7WD3T%M29Gv%e&xl;ZKXRpdN^YhdkpW?@IM;v`5_r_jDF5(iNgp*Gce*&kO z{4OryA8}Ak=X1^cb7a^4t1frOmEGi%aZppf1{cYH%(kzi_!1n~me<+*$*QMn{yAmF zE8iry#qPCoPaI5?&%xOQc_L14mLI|n&$qAIeAfGG9O4}~_*v&egL$f_(nBub5T9b4 zzb~!u`z;sX;@2nbI{&XW&fhosCB1K)Y1>cH`SS#hztwsyvhfQQ{}4wb^!fR%aX#K6 z-hci{+#9It=kED^lv%%+=ko(_0Uw8xrn>%}h5fzd3vC?Vh8_F_j_~`|^!K%K{=TZj z_pg4%_u<{1ReerJJ+C#hzCiOU;GmQ0>5090il1&>O}^AR)bW{Q+pm&m<8-9F0B3Kj z{}tH(Ohs zmyGlEtV@3@a6P=mIIq{~p#E#*_szzI6Xd;drM=u0M>XXWvDa0;2p9XyH{rsGTCWF< z^Yd>R=ifpb-&fb0cWwItI-kEX&gWNTem@eAcGvp;g-d*%)XnenP5tqu8m|Qospn7| zAE|gBoZ`W_ME>QtI7<1q;)M7dn@{{595Vh|Tp_>O0@YW>9`NCnf*PVc(_gL^%kx)(fRxd&Kk&Dtht}4ordVu&2(KPB?BZ_r;|H<>9zoU%m$Wd&_s>%Kq|m*sUkOgPq3m zYV5U?x8cHG@*eqft*JL@As>iKu6#5uHj~f9Zd-XQ&KQ3RcA6+2PA1DkY(BmQmnJBFFAi>$pRw_q<)t|5qBJNdndyGLm~&NI%>=Wcx7jKc*y)i^)jmYuBg zZ8q)0lT=R`yZk)nEu0l~{;a~qLGor?8X{MFP4&9O8{-0QXPno+k?U0t;^qEoKM)54 z!r@_Z zcU-wu^`3~!%E5ib5MKWO7W$}iBqaI3Co@8Sx-f3qIDr)s?I*gsR= z`*rF)hW&yI?G*2c-9GC7G#u@z{E;@kmwY2m@cp>dQ1LPjT=_j*c~IkhVf~Q2!RF7B zx8uSo+JCj*Q2hbv1FwE|v%3th3JV@wPqt`3@VWzo&3`soE#FFk1fB=J(cm{)LN|DZXc-`hw#WFW{7T zKU@lxe}Qp6-#u8rn}`>f&x6E6*5g^?#Ww2yZ5$`+?>n4z(f!K~<9xj%?ho9za_9F4 zaZ}wNbTH13Po7prt&~IIq6}_1{h0|5*F!8RP$RyqDwfVjb_) z<_}i>Z`k`>_3!nz>Pxb-=~Hwch=3zj>(U(|DWQI*!KU*Z{ujE{5|#@t^aOI zH9wE%r-sJ0^84TEy8gE{&e!*+lUPsVyxxHO@xeH|LdW9@xbuJJoxuaX#Oe>(~9(oIlUv;rJbV5&p(FU!REQjh}I`tJd>h920k!Dc|p=cuV7a z{Oye2g?RS1;{A;C{q610{yC3$rK66=XzX>8Ct>Gc`Cc61=W(gCj>mG_{zI)_YTFmo z{tq1DJ>FOS6^?%^T;%=5!8q=r;~(JgKb^lLjPv=N!~2MsP{7Ce12(-fxFJnDaQHr);&|7Cl48$^XYcY_c`o`n$KG{e!2W5 z&hD3gGtTRexV}_fuKvQ|itlZlk6&SZ3&fLrADUrL!KJ^n9+%+kIK?O8N-z0g>+$je zoK{o&<+yyE#{b?pU(dZ*uRna6d=9JV(dlZ=5|NH~mQU7UszXTpBN*guQFz zD{zFT;-HJ>^R)F1%3o%j&*vuAYd!vo{ZwUz;@%9+w+Z&iKNzR%$G*n-fevbD{v(OU zv($bf@$6@fKZ|&0`ddi6v{>=?aN#X^74|-of5pl3a;=Xwej3S5t%u1Sv3IdN(E2c~ z&&4>oNbyP7!LzV`q2eza=j+#t`aiMp2Nd6o-3R4571ih7C%3^R>hWSf**M;Yi`e~C^@g||j&M(0!e`?cUxCZ`4xI76Xtr_w`7w7#$6eRU=dr#2 zP4M>J^!ZYueKJV%TZi4V^V}koL)9U4JLy^6&CpHvY2q-{UxVM)8+%%;!x7SFX_U*@BbW%CC|?_n7NTGEVWP z*yZ~rop9l3#gD;(FQ181o^LL|5x=i<74{f!3eNV@{mH%9Z7k2jNhA3!TqOPpP8%q` z);M3^BF`gRvHy^+k9EG#{QTpUzaLH-X#R)c;-2yV95j(H#O22FjW{Fz04_9C{6$>B zAKQHT{|T4&QhxO>RbSXh-Uqvk*9DiFDjs0Js*eBVIOgYB6OGON5Wi1h(C^+TxJ?*%imXsos9r;r-qaobf&)#ASXSHpw_Y zp5r-R?j;^Qsrk;e=J)p(VgF~1_YrovpH7YQ>#sLQ*WbTvobO}oxk~dX<8C;etIyk$ zZJeKfUxdS-G~NUpZIMej!B6AjQR?qan_or#0%xCVej9O+slUIm_kpe_^}klV>1RA| zSg(UX|T{ZVasAMBi~{th1tC&ewXbA-n0fjvAJ2h@87_KBBl`|fH# z&$h?QY(Daf&s6o4aA#b`$KwjV06V_MzY%-* zVeI2qaS?xk%lLPkQg6Mrs;5G{9S)Awc)f9qhvE!Rzo{Jx6^?NaV{^a2_2LxbDc{E&Mm*vAHI8_g>ioYQCmmJq9OL}{Z6o(@uM#gEqw{}- z&3{DRfKz^c{TB|J>GPo8_p0CF_t!k^pQQ1Rw0=_gr`!Cp92)2K#2@SZ(iGxpb+v!M zIG=wL=06{I#EWrb{4woap10TG4tQVnxpBT8hjM(@5-*d# z6&J{Fyk7ZJ_&(i1IAuM1;$lVXI~aTTQscbd!d%sRGcNa0efQ#&^_z#Cr=Bm{QUDXZl&johQ|5&_hbDJG|ubu-_++_4;&@(AnPUa zC>*{e-)P$tf7r&a()ceJ=k*q-cRBHZpKq?UK2rOq>IUjLPwic6K41680iVC!jPv*7 z-5%!enHcBu+mrd7Yn;!&!u{)&v^V#G`Lvy>xcH&g?+IMkq2u#9u56XRz~#Qm-)QsC z*ZIE#r~Lk_yHWFTxW8(R-EVdOd?b!*$^CG#wtSAwKSTYG#v#Z1MjP+0dhWyJ74kfr z|Bd{f&Hq7OhYLT;|Kenr#@jQ$58AbV`F-h5IR6i&cCBAO>`jq}|QS4 zhYJtse(h--#q!&@{D`~?7ax_k;_xxK-e%RCHIffB&aX!a*Q29wiciO1;gQDqexA#I zo?x8s_lWn2(~R?to$&lL3#ayc#18MZ_VFfMX{Y(@#LhJ3H~m@j@#wDu zE)ws9Gw#<1;rMp7zZ|EGcPlO*s_`B$&aZbKKM#1$I6pq=$9n(p7Ww|ey8f)P@jDg& zn|RV!{Wbq3Z=WCUDID)k#(94(*SDU=`Ti-eKh8AHk5}m)T|Y0NeX>*6|8cbU*)O-@ z@i?Y^rJmY9XZ@wFN6T>e3wa%mxSnsd=6c@fSJhKKTG!9kIN|4q#~Pc@2iEHpR$~*6U0g=lmOmle-n4fHS_YdKdO*D*iYw9Ig7_u;zGHu*Z6@w`RR- z{Fc|B?~iUA&;5+^{orlU{0_#cul4G0P5Y6wPuO1*@OAih@{4?5`vIK%r~UkljsGpb zgWccd)Hts%puQb=0R5XhQ*TE5PR4ouL&SUI0G~^K^pE<#7KdBq>Dc{Ko^RvZt;(zSjR#PvKwXx4mLdxK{BwI4vr^$i|y% ze|&(0C**H&d7iu(`_IaM8RzR)u_9L zss2wi&c}Bc|2*UTdg*b!97BF(zWTe_`g!?&TzF1ih>I`C%dtoMHQ1S>_50H}AFn0j z)z0^osXylYYmQ5NKk!H#aD5nv3$@hWc{t*FdIc^XtLw%6*vl`XW|^Kg&abZ#-*$R%y7UTT;Qz_qn zx{r8tu-ZRmoUdQT_Xl4!&iC&!_U{MA`SBUW@%ag7xLW=>vg`9?iRRlJhwsSUae_}b z&imU)`w*x2HvAQiY44oG@yFRf`9mC?Ab)3^*YEKE$!sNF=%Mwln}3d(`KIibHpcn> zjrQdFhLbB*e}F@N9ytc5pQ^u`jPvpxWe;%5tnA`ct<#VM4pR-+vL}9GEM#rr+3Id8|U+l zX6t^sTK+k0d=cmCzQ*}_y3Z^B2<)`e`ksXgCB-kt-V5@bxWeCKcoN6g>v$}|<I~#4WIUhVl=_g|p=2aCno(I}29^D?SSQr^+|jeCjD#w^sZK z>|U$>7USS-#Xqq=NB$8P=x-aYkYCrSrui`6{cwr+VK||_<8jnl`)LR+^L#eWI6r=K zIes@`4?l>T;sy9XyaGr3JoqQ${P=r(|L#AWx_TaORz>kb4;`PLIN^T$N}TcYyO}s9 zp5T!7n`}Pu{i>?}%5m!d7@QrW&+kiciSJuZ#gVW0D>xv&$(nfEYU#-ac8Gn<_uV{Z(-A(m4yf1EsOFYl^ z!2XN6o}7R~o)^!?-YE5VDR%eKewc(y&E@-X)JT2~mwEns+s1#={{0Ml{5`TQxL8N$ zuTwLxKR=%u@%h!vIA8DdQq^+^E-;?}yVt1wc{rY}{BbrvmLJ6Bla;>!N4>TF@8MD> z#aH2Uh^{{yZ2O{IEx+$I^_5PMn_z#d+8=~nzVCe^4t6L$635%+TaC@?-~^t~AhoeHjO{Ro}1L(Tj>nW?|*D}NuH zw%74K3_CxleP3LeqWm+B^ZJ@{d`A*5-l6ye9B{pvi5)y2mxtefRW^Q``rCx# zsdANkbDR3(+vSEhjOBK?lz$H7X-*I9<=0i?z_!n?TgJn&^SIhyZJe*SA8CDWxB2{j zJP_asIsG@c#CB+rFFHe`NC?*YzvK#V6#gIOYA$?)iHMvmavK zS2Z=x3#q)R_3VIC-tY7z|1|c`AmYKX>OaIqe!uoETqZsrJKUdrg0s&x-Ve5YcU|vx z;KFB$@1DOGG4&Of%KKrDzc+A%aXz2Dsc!)BV3qQR;^;k{FJp1KSn-=|oO&O|9`jj( zoo34a443$OB3p3Gcsq^rdZtoOll*&#=K9u+_fs8=^W&K;)z9C180YgVG5^zW$mivy z*hw`0L}PP(;qMVVXl&lMZPE4fB^Nh)82)3!R_CgNzP=KBCHJg4*RW$fN2 zudu#d-i#}_dLz{rPglIPaeln|u^)~m?(zNO({S{H@`vN#5cwM8e12ut?+)X9KRA4U z?@{Bt|Kdsd{h3#Yr~EzakBsy63O-kVDUR0Z{P+bsuj%+!X{`K0eXU0$98OXD4mi41 z?rWUa7xQ^IoOqG-xDNYMmHz;CUX`E2;q&qv*yH;>6;!!iL*BqQ3r1g9Q zCzq=I*EU{D=f^hV|MUBaHJhrw^6%Q8Epf5E_GcFyb&z}GjQANg-cj)jY#fir9-fNJ zi}m^VBo6Y=4Rcy8!X=*nR^YIw>R*pDyc0XPVgBCLtar%wZw|r*p6`0$gzN2an@{`( z9Ocgirk)3|!}o_0TsmCqRlz>%y9s-E_hzamD67BL*5lRw2;=7rPx-#tjkv_`Bi=`Q{}A0@JxM(K zRO7v4&Cfr-#?hsU|A&*y&N%=x7zm9->5|%H|P5+I^PdB&W~T^HT8EQc9zJ)afm0{_&bWhhDDS`({+?cg{5`Q5-{bG)9b}x>SBzEP$;3<4e-SQzss1M8@>lYsIQ~{% zh8_GX&S+nEKaF2lt^7{d|3*Fum%o-n?0zRt!?|0P^zKd->aXUhKt z`!C6L_NSf?mei8BQ#1rE64pjS)cuV7aAqpc|UmWoLt%2mn100e@?+pb@@UZ z&(rZ7ha>)e!X3EG_p6@6Vg5Zovp<($pZ962Z9ezsTXBW_-8-mWkN3~};B32&M+aQ! ztoa;?qmHU?AWn!6H_q2D*-z^?#yCHpkK%m3nS7u7r@L{;dOl^GAD;q0Uwe~ybcM$M z9A|vLaBt3On={7(_i(j>dzy7ATAIe zf-}aOj#I{~;E?h5=|+9@HyEe%Hw6d8U$rK_9=ohpy(840L;uI(5}t@N`g;-k^!I~p zPk#-LRDT)uACDvYyA!AM_nytCzYG_NH}6h;k7_^qI3#`{PX185Wb?Pn3$2N7#6{fp zDE04eRsNaS`%}Ier}$wTr``{3oa46*SD1h29_qiy{4d7|^*@PgaXnjs2jT6w!ufE} z(X2214aQ~0k8sTR3vhw_-*JiY_diDcm6%Uo>=3^c`>$y|9=G{iPrkz*@g{!WU;cTN zP1H~Fw)pmSJv(p!KBk^x}}$*lctfyPNnKrQlwH<)Y9F-LE{SQs2Rrv6(ynqGddbK zL>7sn%Bto;8m>u zQNt|XbtmJd;Oh6Dsrc^@oVk;~N8hoP`B$F%1Vv8??&{y7*v7beC&%{*!K>fk_*MijZ{qkKbvpBBHgkOaf>*Y3dJ?u^Y;f|D|qRH9N+KHVE*O1 z7-x4f?)nkO|7pRU-{tuKM{sou$N%JA%-=b{@ec{^7~=Sgg6FPe{a@QH{C6sHgWt{#x$M;&n zt6%5%ZW27#$MM}OczGSicigDh+s*NvF1X`0tbcJ#;=7sg6ZT1bpX2y0ba9IHEeIam z$oiHAcfE<_PutJ(b6;S5jo{_4u)c?%&GhPRY=6Jt#csAgFL?Q@OnIera(ssbkH3e@;}->Y zoWk+nD|mSu$A8>;tgrL^9N#O>XT11b#y=I@@dJ*3_=QYg{tm}q6x=z;@xM{<+}k<+ zdjw}LXT0eE%P)L^`G*CkzsNWvc<`Hy^Mb42Vti2W+^vjXDY)xPjNc}B^-jiD3GVm? z<4?Q%%ZzUmJpXmZKNdXq6~@1H`M<;Xq2p}->h~ExUU2%$jGrlZN%+%(7bJaK1b2Lw z`Og%*{M(1C_Ov88{X52&3Z8p7f6w$iF8^Z~-z0eOv5Z#)cM1KEF20NPJt@ofXTHTa zEqHzh%MT0gID_$i!PTP}JA&uVW&9$-n9p|ZSqfDNlc3`qKmtPB1+!c>X-bBZ61Q7>^5H8D;!p!Alv& zmkD0{DeJ#R@bV8C-z0ecmyEw6IQ;|0KM-8~5###=4@&+%B**qIJ(=S>M)2zYG5^Vg zF<JtK9BKLF5bfU2EpUKjBj`8{fvL%(obRh z-~!vf`VEf%34~$)+)s9^@^g~V2Oq}#gMyc&zMLs|`7UnXFAzNUea0^nJTLfi!JUFX z;nG)_|2Dy^cQgLA;K3g+ad9tD0tx>raw>cvf%xKSAN9wy5O##Fn+DzCEAzz7m4s0qSEYP^Qs_&M zWd1J*PM^Z~zXeyH!1xz}SB_!)H<$mhj30Xe+qbx$@$rIJ9>aK};Q1#r9ud6!D8^^I z^rtd*1P=}{J}7wMpBP^vc;y+4FB3eTVtke0jM#IX;K4N0zvS|t#CS#ULJ#Bn1XrKU z_z~0M|31e5NSNyXFS-773w`hrOy457>uAPfg6B?VJT186cbq?$3Qj+i=~oIK|0B~s zD|oS+>ECnZ#h(8Yyl@4tPd=u^_AbAR@$rHe-^qBR;Q0l{I|Q#@%{cGMU&;7l!E+yC zJTEx&e#Tb{Uit{*8w9U>nDOm`mp{h%C$9W;jQ=9I`fHj8p^=pjZFF3u-_!hyP|H1g%f(OOFe-J!(71NKY zuzicdpAuY^{2vxPC-@w}3twdUrr@qGF`gH^^m)b~5WM_Z#$Obi`7g#l7Ce6o;|Et^ z57nPrx&Az!FxKb0je)kF}~B~|1{&@3tkZVu@|!at0Lbc zxcX7%-!8aQ?r)qgxI@loHUuw<{&~UU|Hbkjbm_M+UUuz97ynKRPQQTt^Gd;)^zo`cyi)L>&~I|-g1;qrUhr=P&pq&AO8)3N z+qWe24T2X0pCNcraFH<9>zDrAujH2m5B`bqrUugoh5mZMOMhhgQB97ob2q2=L>HgS z={rO4{4SgUmfQU5Z>G5^yA zFF%^`X@VD@$#|FGjuRN4CwN8VX9Uk5%k-BC?s_ugHwj*S3giD2oEG`d3ZA=`^W*D; z;ol|M54=a{)gG3=Pw>ja7(eK+r00>0pDMWekBoZ-uZaEI1g9U*^m7GwK9zAz@H{RR z!%ZD86}^e;3y=4r$4*$y)6HT zS$09f82+lJ*}iuYhCLav=c7Vj9$@-s z1dn$!{<=$lI^&-RUU~-OKMP*$W8C>7j&JcHjGyY_M=?$bo_iAG(*$>%!Z_>F1;0}8 zyx@NqydZd4aHrtk30@R@>_1ETpDXdXSm+ai7li&Q!SjOOFE}IocL-h*{1?IFf}eFU z$CnnI5xgw8DY#SgzfEvP@Qs4U1>Y@rMdTm+V%ERv#wWN#_|I_Z&*Aje1a}Gj?ScnI z{uY-m_@{ywh5wN+;rQkRrv;}4?-M*Ic$P5A_lZ)TFBAI0Cbs7)!Sfp!Un_Y07{*^D zjPiN5l+Pc!^q+A1`G0~t*D?Q5b8O#|j7Od#cty(B$$~TMnSapb?`J&b@^>-L3m*Rw z$Jcc6j~QR$@(X^ui-rFKf|oY4z8eM4r5Jx%@ccQvzHyJ>!5=XFFM=2EW_^ApCq1us94^-T-zxR~j$6x?|>(=Qi1FZeo_{sN|d-K7iuAD1rnKHz0+PseSn z|7n8L-(-A>;JG^)?-yLXk#R}z{B?|9>e4^K_|1YdN3lKsA$agy#-AsQ@zc2M2j4At zPVgTEcgp_hF)wHS69xAQev;r(!CiuDf~%tMGQsnLZxTHI6&{b>C3rz_$14yY`m^-s zdA`s|813)UCpmpxg1f%W_}PNTKf`!P@QUF5g6E}tObYHyu|F<$@k1HEoiO@~ad{u( zDxoine?KmGevtWZ7Q8se@qgLHUt;_%!Gk~J^8Z7@OP^=@y)OM$#z(!9)7y0u<0lHP zW*DagFAIH8@WQ8=ex~5LPcfcw<@Ydd2p-?T_)@`{uQGnO;H7UczRuk$zt!d6#`t@Lp)Vu)?sfT}&+?DCgwxl#hw-xo zFI~^_n+c;lFWkZ9`T0W6+{N@U!q8tGTDeF%^lj%nb&OC$h@q!1BVLU__@(&aF z^Mu~{B&Ih6uRfjeTZO+v?h9W<82#TDr2o5?F#MA)v46fOcwX@L2_yY;L!91U2+sVR z>)Rg%cceMKN57i+=jFWNF@jfL%JRSD>i7Ybhd1*g9vxKrBC z7ZaxPEb+Zja8>X%!ax5Kj_;#{kspgP|G0%P%5&H4T%Nxs{NobeU4rMXW%*wVUimBI zj@NSfsy}7?D8Wm=WPF_9&Q-=Km;ZkmpGKIx(cMHu$oDdnX`80lMD5>yeoz%W{&HSld6&z-pXr|zT%CNP()TmLgD3F$+(*2g zK_6c5nG23%ca8>Y|1gBrk_J2t5+`Btfd|!3xf>#BvEHM3XZ{YYl z3XD^NJ3qBlPmUkYA*7U##K=EdK8 zIlaAt2hU=Bmf*Q8zj|Hb+J3BlFlxcuKPxJ&T8f|sAd^e4WV z^<|{K^$T9=W%@aS=PqRZSGaU(4}K?j@$*c7!&_M2@@E#_te3F8Xf}+<86I z?-INq{11N@$G0N!KU?tl&CEaQ;u{zr5EAMbo8X0GxxHBwJpUBNUvlYBX8c{j zt50HluggEh<^7Qh9RJ{jQa=PQ2|itLXN~EO;6=d~yZmQx{d}|Fm0gTKE_i+)`c4(RBKDspcwr~=*9E7A|Fte%^j|Hw^GxRdoQq}rbdTT; zvG1>ft76Zw?_v8oMBlLBLGgE1aF_7ENbvHrxV?U_OE2^Jnwwp^oFBPU@an0&zyCME z>34Ja&%Kx9UzGOnV=mss^wZwQ^znD{c;Z&U8R;L7xk}`vz77i>l=L1Fy!s9H&uazG z-NE<@!CgYXR`9aWzb1HI==TU-y`AO%;?lp)_=#7uJ%d8uD0p1<)AkCU6I>R&Aow*d zmj3tqf)^$JM_j}D=kMeAo+mgh{q5TXR~y_u9QE(aKPTr;KPq@Z#`CHF5Pc$liQo}rOdps2cB9~`v=>>y3lrQwUE>I_vZ;- zlJ(*jy7Zmg{~i{+DD#om30{`ltqq+#&O!3k9#<$o<1x1$T+RuM?c9b9?!9!GnKh zeVreI{=oe0DTFaUt;+ggiZJGbofV!BZV@~s>p5o<#(Xj(>lKrN&ldb5!kF)M*10}k zB6wBm_hp2kzbg8!5qjnj^WP|V`5@!3xOBna7x}{?e;;9*U;lcy>hB)1$okV_|FMD> zh5uy1*9jgJyd-$P;5oqs!Iug?D0p7*s{~&r_$`7L1iw%4HG)4*82(u}%;~$u#TCZ4 z39eqq_%6XK;?F-2ru6=b)BETTv%Zcd^B*gCe30AA4T3XY2 z$@zp=x%fBzDt|vAc=_Xu?{?+C#PatMCjFn__G`;W1%Hz9TLrJ&!1z~!mp;Y#jE^yY z`ew%eA-Hpb@#h2&jx)ZSFs-l1`#!%TOzS_gUh?RVGrmUf(+GpVOZfWxFL*J>xZ?5)ex2ZsxAOYcwS*D>b@IOS7YL)irQg8y`3Hol{>uAv|10w2XLEbj zeJ$IQKA-VLgz?^F*ON|G{{MHOcjcLWr_cw5{)kVoeCJtA-!6E0lJQFguhtmfD!8M; z_;J^&;6vtFZyQ)Q+wlZ`mc29!vEib(@mB?_Ij3I zp5yg{F~OOO8P5w|c^TvH2u{D8@$olE{4ZfVE_mTZjNc)+5I<{T9I)hw+1!B)yZ2w+OCIF@CAw`6A;_3tqT@@dG~1 z@`JBuJVF@lWBQBS9vmc$@;>-5zEA&N;m?SFmIW`)asK~R@aoGM|I=q!|H4Zc?-87t zX8c;g^JT_g7Mw0Ie)KIOe<9bmGQ>~FaDPCZw04+#(4b~#hza< zo)O&fE5;uXymTzv^COr4D8|QsN#q~O*b&@yH_Klxc>ITq?-0DO!uWAthCKXNmHgaF z813Jpw0{=~UK0Fa!pQIRpIQDM!l*CfQeTd_mF1U(|M`Mf1Yaz;tHAB+#{_o@{<+|c z;HQ6u^(_fLUvO3MzX@Iy{ht?{7W!WW4+`%4D(hPmToXJm_yd9$1pi!chv2TSvHYCi z^8}9zej{P{Cv%*XAHwj@l=$af!BxS}T4w(AqnQ6Z!Glj>d>LVCpT5B5`-?(fxtYi3 z54cU_Z)7|qxcX_vGlU`E^&6Idw~Oy({7vCsdIrxoI={~Ps?TJ6hTxSaFn*cfrB236 zf-@rjfZIj>Xr@0`aOMQYvx3vdGyVu+*f%ceyT|4K2c{qQ4c6EBRK}x%yN+Z0Y8MOr z(}I_t#`Hf4uKpwAlkX6Hk7j(1;6;(2cj=E~`YnW!e=85={@^~LFMpf;yY8E;Z~mK% z9l;CVVf+rk%RglN4Z$5hXZ+Z2vHapajL#R``F+M$xOj!}-GZ0D%lMz}Wcl%*GR_L_ z_ygm23a;MD_#VM?pJn{CZ;Skw8J{b7LFC^ec+j1H3hww8^FQW0EZ-^qc)sAJ`{wQIzPhB}~KYlKFb%OEJzRU7+ z-(ddzf;$Dj$)(@H^v?>e-op5|F8?y)XMd0NEq{gagy0qN=bH(`p7AR<{?EAlzheGB z6UO-Z($|lv@$iQ4Gk=HJKQ6da@LLH}eZ7zC=a&VKKWUGW{|jOGXYm)TKYf>Asqg0q z?vnJphA_(Sol<^p5qjpwEPo$i^5-bqfBXvTOJB{V+w=?}b!I^h3ey!m78yMdxc<@@rzab2NWTbo?`vbN&E%u%!_(Z`DVfbhH zJuLqk7r&qJhX_O8(oO8oZxM!n79P*_yW@wffB6#j#|FaSUwL3g`TvE2mp;w-62gdY z?h~y4I>8I@tFK~w)IF?!Ui7C3Lq2^8)6Wm&V{_hEWlcc}n zN33sI;y+pN_~opBoG|ixPV~J>=qo33e{;RyrL!3SM)2S>7$5&**1xcu@wtLC6O7*~ zxFg5-cEO8+AN3QK&*YgtB6#rt<5vhCA7^}v;La@Lzq$N^Px-0XvxVs|6PzwEzE$wb zX^bE9Gv=Rn81HlO2;=t&UKaV^yZB6|Z~QsSuZsP%gi$``Wc+lc(3j6({x1oh+s^o2 z!5wMF$NfU=IhFA~!Bw&M^@6*E{#n5tLjR-S!5xB|f;$CYEqGAy9fDUUS>Gdm#r7=iV|&ts(H8H9G5)UL zmG>|{>NhOE@>ovqNrWLkF7jsyp8H3xzwaardlsa;-6r(RQ<(n9dqw^@#zTZ*@6ub? z-h)E#6n$3{hQ4t*KleSt)E~(A1|NMN@cE#r=ta{JjQxa&vk-!ft7U#hVD`vni4!uThGJD7M z-!6FhV#bgC9m`i=%=iq!b1!B5TEX+LVEiS)i|ZLb`u|vd{$$1(!3*7t-y(RWm+?0R zFQpkj_4gA0(;2_e#m{1VwTqw0_!ol5Phy<@1Iy2yz_=;6nqvGh!Rc2rzLzlaV^PY_ zGyll)t9hpH6Wmo~JTG{8n(=LdJ0}@G_)jdqBKB_-yjo)V1%l@X7%vD;Z)Uvg;th-+ z`e)Xct}}ifVc0u(6t^d3moDYwJ%Y!_ng82@7xpoJ>|Y>{@;v`Kmfu2{%GcAFUKTvR zk@0&3k8cuvf){g)AM;n1Upbrc>4KMD!1zUi)7uz-Sn%Q?1 z3)5!_!~O;7->)GIf2<60eY%q{+Mg8}zdXF-!HR!T(mP0)_>0VcNbulojIR;?YefH@ zF8>Re{_qDt9{RiFea=$|!=7c)UndOx$vz<)9D!ycsDn=-Id&z$Y4bgMpuC;1L7wH1N3wo-lC9z)b_c$iOc*@M{cwnSrk`@P8Wk zdINvPz+W@)w+#FP1OM82L5*g-)P`(8~A?={D6n({yD+Gn+&|&z~>qG0t3I;z;870H3q)cz+X1-cMbeQ z1OLpxs|Nm~fscBqp8m%g_&5WvGjO+oHyZeP2Hs`heFi??z+ckjNgm!tKfG9ey_&w zQv6$Zhu^>8_h$V56~B4>-h$t&@Vgqni|~5~ev|l> z@LRxd7Qc7m*TnBIepC2O<2QrfA^cv9-`nwf34WL3_g?(U_%-mm5WjchSHQ21UlqR^ zerMu`qstw5{;^{pe&^!%0{k-g?ZFSnmOJotWCxBjci_p$j&txkfZzG}y%4_~e&hH} z;D;m89p~Zq|NHpw@8hnH1IHJ7ds4mqn+AF}PUHplO>}SCJg_M(aG#i{*9Y^a~lWTU}pek4s71o zGvQ?W@}3fB!m0$pxq-sQYv+o9=9{l*@4@Iagc}H@FYgH|zA}2b>-Uh1w0?=oILGu8ah4EO4-*;XhumHSvD>I8x!B za3&@u@_i_r`5q*z+YK%Z4CK1$|6s^_rD~#JEt(Srx3FoslS^;f*tfYmo$v3-_on+@ z-8s7lkv=yi>R+)0f4g21#~#Q{Rt`)$jd~-So1UmN%P7SvtXM8KeT9OF{koQ!KodpWvg(eJ(5&ow(1Pk>rSnaLIJWfRkO9av)8FtE9E+T zGEu7JrVrGM7t!C%La_o<_rgC;ZCz?>v7XD;kbaYpah&OJD2drax6C#y>WP16kut{A zPn6M8F=ru-S}omIu0g~e&Hh@k;fz(nu@s7>QaJPwv2SUjO1V~z=Cb87`GfxOgOy%{ zZYwqGQ{iAr!?s#wCQQ+j;HpvohM86HXropvPr5`So7v1@APPu0u*VjN_M16JYPF&K z!E8APPgS;MYk}}B**v9X7_|-6!eigkM6IoDYx+Kfj%~$qwlr3mc04~Ja7X!Iwp7fA z^9Z!v&SbVxJm^T|0sSt;j2HIptoY8LJfT8_%EYEQzEU#V66|^!g|d#)&IwtU+Nq2- z0!4)-R0^AgBDo%-)E)Al5$t8H5w3a5phYA=1>4QiOOZ{TsvrhovQ|0dTTHH?TA=yK z4b=S<4Al*CDWlwiNQ7P}S8gNJ<_cFrCCs;dxH5zGjhYRm!!0`unx6Yn@TeGNYm*-9 zj&j~XQ_gaGooXqoS|24jRx8fzbPAq$xQHn>1cyx}AG%GM;_?Dgg=RTNRaFI}hO0RC zm5aFwjUIE^QfV)hnA}V?mC6@$jg+MB++w+LelUC|dK!CRT{RpUbFscVyF0|qp+R=; z&2n&P4eB=(O6V!68>`f6&1z#ux!P>NjTICK^$+YZS$C8hP7Uor3$}^bZ8hgYDOVvS zKDCZ29!VZOl&xCHJJFigvZZMC0fvYeJdTJ+@{Vg3d1JU#p?=j3@*JmgQTa{4q?(Kt zs&bX;?AB}pBlH%Gc`$9E?!G1G6x#ODS#(`3G!oxTwwz4qOm-UM4r)VM>Yb_7oGrzM zEyHy;mYu_mWUIJtFiMI&ULJRb+$3xxWl@=}*Nc;7 z1<{cYmGe9XXel>QY^XNO<_R;7Ne6=zRML>uDKfH=t%vhHIB39t(J&`SMLQKk#n{52 zhMI|pRzhndB6xN?ha$=iHxAq*83uSHOQjZ?wK`ACN_CIW>Qk+XU9wd-6^1=ETLmZ- z4?Ax0l(H3asfA~Un)MhdkKl8=8ymM+?Uqk_)O!jAG)FCXFn|n6y9IL_C9+jiRH`|q zkREGYqd=PMc*&Z)s8|imVa(>(ARFy)h^jZh23fT!W{YOwamRSNjkap3goE`il(HZ( zSSdr2!8#ct_L>e0s_ly-V&%k0Mqtv36~AaA79yjGn@l1G4Z~ATZo1wKc0(#5AZ@pL z-7O&HMiw)I+J;iG-at{MCB90rOv0t*by#s2dPp;+8Cf3Hu@C9MxPM2(Wjkd*5yOR#uSPK7xzeeEL?_N zVGhH-FhkhRFppuc)-iEmM-Zk_1uKsEWMvW*UF(z(@LODGRvM$H7w%Vi^^fK#-p9#1X|!6&9w8m zDO!|uJt+U~&dvln)4Z+m_a__dIXjc3;g4W6o$fZ#mgM*(&97dt=29=z@YiGoqn8G+ ze7dRk2S7B)@rTBoLRuLs`;CzAw3;(hIp_>e6=|=-)g>*z#e9#VS?`8w#=C)r?RB** zgmB7Wj-W+mYYy4r_<7(th3c- zs}0(F*^Z4Q>@=u_(3}ydyatzXaIl%?5Ry${w(POE0Vfw|5FKFB^eZmUZsy7tC!KnO z*l6j2!=|7<0Hv@g+6HWa!FW_p5`l+;gR3A1_=bzsDOyKX>o!3{L`i@JQ(+QzEeJDU zHX6q;RG&!@kK$?(mEvg;+v&r*62zvsTEwP!lEjA02Cw0t$=r-GcXtId2#bSV`kpH7 zcyQ*5Ydz9Ir#6LFGlD$oWU+G=kPMYYr7u_sA18XI{U2>CHI@-*%VV8@Vm40*?pcYOJrYLn1{ z_jP)UaIz6hD~h$q840m zu%zkuWt%if*p`alR_nogcx|2wJdvnx*v6VnspeLQBy2WIJJeGZv!zISet4>avp;?XB4b5XXnsphv^vc92c{dGTz(2oc?RtR>Nqk0 zThzl_w|j9?e=Jx%%~!p(^GbdWEOEIm~(sJh=1%X?LPjq+Zx)#WizmD8Y6C8}7b(`>p_6}_Gj z0{e^*#25~7Pg?}KVRpIB%3;1q+F+OREhQXoX^vvRY1C4-hJ)e$tR)~mI)xS3{3sSU z9JRC-G@_oRA&aYjU1})rFTOy9NggrB4<^Ho`kt~X>~WD~vbF4t=4?|9CumEy?wE

A`as}?dYOM;aHKv^; zSgVYupjgH{yB;?(jP*ck@!0gY6pv4ROY!)$`{gHkTpXMFBno2FpFlyp2NEcV_W@4f z(&?FI!{JK1)~e8ko>-40*AwrUYIc*V*Qd#N4!sx>4@|P z&Fkg3XRs~xIu}dd;?tt~4WGl>Y53&W9>Zj_bQfo-j2s%>t##*U2M<(xdbCcT!}Otn z3WIlzR{TH}-ZH`_ijzCa4O>TwQOfQLmi!F??uH;^4X5tO?A{DUZ84rZb+=_~A;~+E z!Q8YsgMz{f5AcXjNSRI`lD06!W|<={rKEo#Lc2?_Xzw4eX%A;m6+w3G9wDq9B%Ahd zR#WcVwR?oHc93k^>H6F#4s&N4O}c0z3pL_4vypC*tz6M};to_!TUUa+wKgDHO z=!dsAg!!;gr1x17Ec!G(is^9GOF}fu=oF4(o|Uc9&5%~?XyjV4;3VM|dbA<|-ySS0 z6iO95nKYS5WF|X1;o$gIqn6Dz#G|dEnZc#Y$qH_x%_gwjXZDW9wbG9gA8&j3^s|{2 zC_$1KT0YoP#9Kb*2_Gp(Sdc&}@UB#M8moEQd0tbg&)VNRkdU<}-M1;8Ra>Mqqf~cX zxgwh+Iq*BQ^=wPpD)fHQ2E2ig8g1Z2H6E23R$3xVaq@@2Quo%jXb=)nXY8!%@gPf4 zw{u6AXFI>tq^)XkCt@7oxyP=8ZhlzmkDa-IGLz{(obemI7TT47-wsA`v*~EC9J&=$ zLEnK>d5H`QNaDeo0>+r_$t6p^8w*Su%3pxXhuA4bRN?YvxXqPA#Z@g|GTX9hd2cw2 z{yUFXy~yW&e-Y)va<;IuQkia6)%n`~HOg&DEQgEE7qn?2iKx<(tOn`q#!w;34FMsp z6)p29Ac5w-P4;%I>Ji^2s6R37xK_d@a2!HSjCN-W`nhVme?t7mAS0nvhIbcQ_{u*v zVpkiuMPX+sID6<#KC^$)JWwcR8(hnRv$hC*9A60|SAc9PiTb?=FdvW^2H5-?Ut z(mePkja3pGB*)P!wHAgK^Jv+M4Sx&B?nryjtR<_?=Z}izCXW9&NfO2qV}~l;1M~)m z(YITagSEHUpQl(&-B6%+GchI}DLFIh)ytuYdZmQJ3-M7{F;ULxrF7C4O1*~7uFhk=`er0&UeS*ZOidvY%ygi zY-EBFL0AZ{lh$T4?rS3E-L-_7h$DGT^=eUTd-eDd3|__L0-c^~$Gik>u%~(yM*(Y2 zd)m)&YRO7N9c5>iW!yPcou=0SM;$zyh`YSnEicdS1`|Ejrfw94%ad^Cr_LC`9a`L@ zQLp(K8DcOUR&zw$P8w0SG$WNxP3H?|Y~6+zfU*-#2~WSY%dX;yP#-g7Q08JOvhHlp zoH62UzegxSE}VdD;1aDhb6ReN8g zOdwL!07qxAw491zCRM%F+FD(BXG4%c5myV3^Dwqv(=yiFu0)33`!bteHy+3#vyz&ys*1x#m+Pp!wBpn!S2Z4@v=rey*1`6E}aZQkQW zFn8J=TZFwBS{ECgHK8MEJ1kM%^R$x$yR?;i+^x-)#l>fASzHX(*i`zwt&5A#o4mM~ zGRghXx;+Z$PEfQo?nLd8FVV7K8LSCiuD305W!%@V%d~Ef+(^yk7BD(Npt@6< zd$$-t(AoOLys#)4Z_LbBIu^_xy@6k(6BLuQ;IT|@U5sUN{Ta!WK^K8bp=!BLy*g>k z?!vBa_qHUAVl^x_5yXJ4)O=BUT_{SG$Jx@& z{R7s07B3?)QYuc4Rq(iaqvS-aO9q$|=2`!MI}mKGvQ>?uu^F3(4HujPu%B!Q*tn~5 zaJW*=Ctz%rS~KB5Kr(ICy{~{-Xgu0-HTc^~Hu%w(woVZE)BzUD+LLZAiezhTnV3>7 zjWY@$Z@Y%x!V4GgfQb2mUnDb+QxYmqP+MVxcN@#H(iL!0fNx}G!nQ0HVCLSowMUtc zJ(hqwNhY;uefkINtNO0-vHg+ity(oh(2%U<0HZe&i=XsUygF-qTMhOkZ8g~9y(J7@ z3mjekV>@9whPkDc+rX2|ZD7=W5o?vs8t*n}PU1GO#k)-yTDlFJQ(5}TLI(Z5EW=`3 zM|-ovXb7wlQM)^_4=?QT_aE;&b9_ z>2ok8^SNS7oMzOE);MM_AjjwRQeLY0~^5)+G5ud@b__ zOv&;`F~+62hAAk%4Nyw$qT}@RjVo$k!u8h97d4*oxcS2i!~Ed|V*$m&qzz4G^8FYW z@bmJv+a@}F7hk8|g3rpTZ|`tqI*)!OMW<~aE*Y! zP8pch)pC7so%;r`UV3!e*?aclv;>~^bFu;3l{ke*ed&Od1nI`p_e4gnvjM(K@aBj1 z0GvKru9;-5WE0hrO9+(%7MU}KN4pPp)5qH+V~JL)?ZR7o*-855Gk;TAr=CSm@!(Q` z_RKHEr3kZDr9v7V)uO`XmL29>JEREV!37rza%Sq;%;+fV(UdsG#~u`St;Q#-aH^|P z#!*xGsBa)4c+E#=8O4QIe_W(#F>f3Zk)Hj*5&?Zy(Ldur1S)Wvfkpn%^(e|bK9UUg zYE;XqCeku9S?ESV(^zrgfo3sWYxJ=piBzX^b;?bwXw@4vyzf1-%M-`z6Wy^5tihmk zWXlb@8WwD^3^sd1Y%u9{$slS4)}#fRN`n;D7#nPQYitlh&9O-*KZX(?u24!W*kG#w z)ne=ZS3zoi9(Yt*F&j;q3Zw6`a;;R5{$uF{5R4_%R#~gFMCO{#7RooBZ04Ln)$>j# z#^j#P7|K7LY|BC7u;vjFj9ii+%(jBQrch+NCG@}i5AXj3f1zO0ytak z`Q6{$(<;ApmdO0p*+TiPlg<1#sCs_u#F+fn8AJK4lWqA;9M=3Mf|1`6go#WrA0-t3 z95C`cNC{lRM&=MjFAWqDDh*y>XaQSvL=_bhG>ZzZzhV~$or2!<*E#%frKZC5L1Lps zA7C0P;T8irp77_Fm_7_Lcj=fFT-d}h+8yOW1$)TywUz+gb6iL=I4npOmqlV<`A{uz z$Xg;Q714zc;kzJXvsDLX;S&pG|E{bsRmz3pr25c$;GU9&W*gAzY&rFDiqR&%_w9G> z1`EB%Pu;(RkEi7lv#9~9f4*7QiBBlt^?x{}9{hfS&Lh`poxHF`*>{LOR*uo;&WgoP zW&d;Z+`)Kp5T#Ns@bEW)}v6VO$mZ}h|pSirqXb*l;97Wf-101w-s9C zBfa6|fUX*qSSG#M3946=Rwf!umi&bia;gw-$BYM6f7+jQ2V*mM0^eub5W{B7dcvHe z7|tU<)df+M39||#!^u2+H4!gF@YnakehkoCxg)2Spq-ex2Ig6sc{a|NyEBYqw$BEkL?Agno4sW-fWAagBi zrm*I{INvadoY5y-VYM1x5~Ec?M17}kUq{9tWNKHb&u_02ls-+r`{C1Y`LnXxwU#Pb zyr!;-yfsxbsD4~nw&Q^o|H4zyqchpVne436g!w$Kmq#}dRb=nntgT=pBC0PST8VTNAUim=qJ&{wzN0Q>m#pO8sfB zsY~PXv(H~Y2?}5VBu`7S`*2Owu97&sR9G`ZYO3by=c>Y7Si)jTw8EZJ%BvwVBAq_)yN&9-rt%+MWH3y zNrzXO*s1OCw(^aN!b?&s{RU$q<*YNt^~!`Gm%UG+G2D)=5#!0Ncy5)}N9kXC8#yw( zeXnx?kCkEuyl3v)F)ZtC=!ADV2c1$xlx*>KlwhY8qtJV1cog%YnW*tFS&!FIlpZ5{ zg=8H@XR$_=tv9F|g;s~T?&qLa4JFp4@E8w1wr6M$(XGx8aFwy@37r--GAG$p$GCpR zctX`TCdlR0saS1juDg4asnl5`lj7T86L5{EWC~~C%_v(+5^K5N^q74qWzZT@mFl_w zEJ}6XkGk%^^xdB+gsx<1Oi-|8>Yl8&1ex2?t(VN`YUW3vOSGy&@p5s_Y9)DOvPUEx zQ}|k19b4R5O82fchfpwkk0{NMZc9EvDWtJfv!1Mghk09-yeNsdR0L!a*pE2^=GV#W z5A(LNKT0Cb{(wxR{r+m0)MShd(CzaFtIXmRV&>0bt49{v;k73=PP>u_RWzFfW?)zV zz!KsvEO6C=&+1E8oJc6Kh5ADgQBmMCSNM##gXvOei6gKcDLD{i4zCi0`RR=f%s%my zSuSC=!D!+RFBchWR8#>l1kG2rHA^LLFmL9xzV;|~Foo~7!mR!FRiRS{B{JGWt+6jk1{QPWoChds6Jd-oPsH?<8yu4hTJ43+!WfsUk?i3iL8Kw9-!_(OA>!XK zlrFCof7?cxtCA5uhc`J9mhor{n;_A)RiPnBi#ROf`xX+RF?}okw$tG6k0;B!zV>5S zgJ=D7YRRnXZ`&&W0mEc54744?#x^VV#`a@agJ(CU+AXz^Xxp+)ZKi!w`!Tegj!mg{ zYtyDwo3*1&*S77~HWSdM^=((}Hfw2{Rj$qY(`J!tyTom7uSjoBTB7$C8&j2L14qp< z*^ZbTdA&QHQJ=(7c;k_7dkYph%U68}#?~W;a~LyEI&rHMR^CX>jj8=O(^IAGge`Q8 zT{}o7Rt9+VV`yP|3-REA&9Lx(%gz`+(l|vY0){4Qj=W1d$xjN%HgUi!Ms{07wV7Xx zB%fNQcPAT8nRbrj*0tn}a}>9A446nJR2~ybtzGU&B-bo!Ho-|#X*R7*C zrs^L_&O7S7b0l4yA5A}8EfsS`eoPi8?IRiM>L^C!V%QWltp37!)IMQ|Ey_F7?qQ>t z{IK#^GeqZ%iW7UP+5%5Z)K=Cu^vZPKKEbsp*ZHk!Oehs<4wA~5%|TK*vpGmC*M_B0 zo=5*VlclRHF((5|o-t+cTs)G*67vZZ;ROylokRxn5iR zW&;#wNqG0R1$SGPn6t<@-R6uimbWb{%*kV%W^?8k%iER}<|H#tvpHdmNzgIqT$WzJV6b-yI+LY8&>3PogVbbYf?nP*8VN60uZ3Q) zUM0L>z4Uv*deIHvV4_|Oq1do2;1gG)hl==uWZ>m{D_L(cL(+OX79tzTvVMnCMH`?P zF$O7mV;}s0V|R>{VPjBXw0+%%>w{8;`-3DS3xXuQvCxdyYr96$D|e6-ovfZ{KLx%q zYWlq=Mq#v{0{0w43Zwnh6YZy-Xg~Et`>7||PtmsbM*FEZx;zCw!(*nnHzuy=^3)sc zr@-frLUBd=sW;kBz0rQ^i}prev^V;qOHE(2H~ONz(HHHFzUWfZ7wwI{Xm9jId!s+v z8~xGV=#TbBf3!FHqrK4|?Tx_gQL}9HM|-0`+8h1R-WZ7X#z3?;2BN(&5bce?t377A z2BN(&5M4F~qP-FL5?Ls&Xm3R8+ZgSqjnRJE811Kx(aG8vZH8{Tx}2h>_?Dyle5k>_ z2y43stG+C~W2=p1xrbn53td96*kRm2u&Ftq+j@XDG7f9g_yb}qlMd*aalnW|nDqLP z&8r8D>ay&g~TODyTPL8P4R~WB{7G(I0VDHwE zQO!UOhC?i&dx!O=k9jTEq=N+ijx?U84ZH$GH8_TEcg1c4#}>{LKKijumd3-UP1O;d z(9ufzhz`C!I8@H_ZO}acHC*fQ_GEM~7~Gjr|2YUsPTnbG@qw3N|MaMy`4+mRKw7xM z$&F$PyG=N~4}f{~YN#}c*<)jB#xT-64x2C5aK(wl%scn_VjbrWLA${inm5&dOQvE`44xS>_6~r*6s zn&UMxys=cJi@_!nx+T0EGPG-}G0uX#koNA1B8~QLPM>Wmr1BPD4P;x(KlaMcP3ej% z^+q{6tr)jyVitEsHC6YeCsA&PUZXVcf$JBEi&+!lhGohA;-hYsVp5O3NRW#hRFgG3?(!6VVl8o&3g3TD|RDG+)aYmiJw{W%R=7dqV zr&P~7wv=-wTr~0OBKe&Bsrd~4c1ItrXk?t5w_2BAZT$_!t%UWolq(}=?An5I*sSAa zpQvXf=%7)vkf~JkT38Jwq)MMH^}5%c_>Q0E1G*^TQFThK5cp6b6(+4eGkMeHX($(5gT9k)=jCaulyJ+ zTD06tx@9w5pz;1jxHQ)>+{0D7nq%wsJ)#(d+=}Y2i4_A3Ts6USlDb*$7inckq$jJ9 zcj`9J2fpx%SaZQ=JUkyn3~W*8y3@OchBYfxAbZFR?}Z6wm}ZJ-2BUWu$1_G_QqM2cc40E*&-f4br#IszGK!TN8ubA*0-j(LgE4pr zv0P=9ffL#`z4cj)SY{GrY><(r_fCyor}M`kt(5yR@d>rZ6Q5waOka$X@Zki?bnE#c zHzn%nSSmkf6)X6JHAd%g>~4{dyoa7+u|HQS76v)Ab+I7bc-bUK z+UGvHW4z*R<%x{IYiMDPjmaCYd#MgR`63Z|Eof^_RciNEDvf%a3r9n2#W zM$+;PR)O%Eb7{mqh-^L@;|WZcu{J`h&$Mtf8*K!3mhr{0Y~4TVsZ<2_L;clWMGACA z!BT3hlrn2X5w-Ec#C?8`=XqGKBU&N0X|b1J8z)LbT6Z~?Tl8O6$$e6#4KY$Y$c+}Tv_D_aVkjWJuIJEOBZ%BpkK zJwcZb%6a-|aCgrpe5MU=j_V8)L6WhPpbDMpV3lSJJ2AlDr;TTQ7SEZwemruIHgyb# ziZOnwJ8BSa^#4w+CM7?V%^=}nMwQt{kBw6-STu?(0a&_oZG}=J!a}7slFdzx4(--$ z#Y>9e){90-A-llorPnKC#5v@z2`SfTi$Y#VptR@~5PN{=ccMneTdeOG>bF&ynJmuA zY*jw@2Sj=tie&`3LT`K3vBHTi&&F{WOP?iw4N)1dWZhQ7@T#Ki+q)x5E3KO4)y9|A zrIljdYhP(gXDCOn2<)iWo55Q^$_hK{o?N}ErvrRm%Wlyf%^W+P3~I$EOz0J~p;AMe zF1S|iI(?gwaS$d>_5*zjEM!EG99W!kRR+kxx123xd=n+4+$-MLn208-eTx`!zNf#x z+uyr&u1*Bol8InE6CH}gtCJ%UCVC7o5ua2r6yP;uJwz?~ zm{pns(J9&L9~0ei0-B&5d77RTih=zr34$trjyPn>wW_K*hc+u7WtR z$`Fy;JjTTtM@u3ft`y2x3Qm~nC|29lOzHLtp~Z!Jw+;7f>gnEvrf;HR%+HlDee%H~ zrzVUT_8BC8=fv}o(HzPemrqWqDA4J6LDodn>zQl~4#L86-Bu~821BTNm<*wcQVoud z!dX;z*9W|gZL8CO%Z&7SY@Wx7e^;^2{R_RguKTO*52GRp7_~;yb~PhO+WM`kUV+Gn z7}=4kX`$Pp>H0|zmW?qJJm_GZEVwzWqM-lK3Zf>p0iCH+!-doh*k7p71N3`F>CK;J z-8Qu_x$UzDlS@4nJ60LqMi=X1V;|l&RI6oYhtwbtrMRh=Zd0+a4`VplaIl6&Tii+X z<9*-ZO4ArcL?{QZGGqwv><5pd&Te&hH4`fcNko+Q@~zC+i`8xn)mtd!E{V;hW`;`QUI%xJ@GcBWVQY(6 z0X(CBXTX*_Zo#%pNlbr>@lw8G(}2yVKwqZZ*{YNqBC7P#xu}-aBTg`?#W0fw9je2n zN>*}4M|ZYL9dmOQwa?zRS$mA7PK*6xwPwA+29etW?$?|ADA5r#qD|Bi5>}O0DxJtiNHlQrqt1D@29X5SC}5q6i=H$u(daBb zI#oZZ6G#vR4Qr&oz*cA>R&i*;YUVn6R;-x$+oiUqfx-@iJD&rXY~ofXeXv6RR7)!z zTsWmiTC*0_(4w0<5(UA&Yb|=HP2$w7Rod?kk3a+Wk!w^S-QilKZGGgL^+?CL7J2EA z)*{OFjq}jn#U80vYDd)=uYE;mOahIn+j}0d9<57VlQyH*Ta%8MeroOdLY>#zRatwq z`%9hb=GLxE>-*x>RHuuFF>j%X;kwkeVyQ&Sb$hWRPM^mzmq<2f+_t_*F+@yk6pbRX zt-G^oMaA-QZ2XFem%71dj9OGA0r}{-vJLdA6WF__BWR@4rYYP(stCLjV*2(DCjzT} z!bo5MtFZ!x3(Iw((5o})?3lN^_J+;YfxpO;Cf=c56&#NpwdG~JZ)AuW=6h%UVaz(%?_*cdr&JXbV7JTwqB-BjqXv4!v4yizX57;!XfxrX|t9ssh-hW zBM9-$V)}MCo@WAE2WnI}qRx!t$tnILY+slY;{gjBX5FoI1(_n%)D!9PxvYw?v@+1w zQjyLTp$IdUV&2;ikN4{ghU3MS(dt~*Xi-I4stNK~mEZxNmP&#=b|uNmNnK~Ns%m5> zCvoqfRgp8;EUE%sgrn7`EpE9^?JZ7DVXp=2lV)L+a}zXjQmdzETy!P7@kF;8Q`rc6 zvxn5U-lUoJ*y}lT-9{yL*Fy{@zXsbGywVt*2=3udR}Zw6aOf`2>(zQh%rusr#Mx1} zS=|h?a$r)(osyYs*067E*ruMpTTLCLDdv8;T93?ZYa$w@5Syihcs$nF(GI24BU}zc zrumIjRQv%BTht(L+ZLnUZMtIAdn^|HP8pvut6^CpZ%bRDl%2GNPdW|s<+QJ!E!pTi zRMGPz*gQrUtFc&&rH{zDM+tQ0ic-mA=^`Ayt>~1nhaV=iD9@xbgO8r7)oAYEY+53U zYp_|4NUp|WNp8jH&266co9&250t~h&nvHrf?;s7haS|R;sEwm7yE69FGnJ2V?4U(> z-Nj6nlOGn6HD8pFqGY_yD68TYnWyx?C>oYo;s%9=T^5Sj^jeu(6blu(J8Rj4WiLxS zaqXm%k0@8QEZs4oqqlWic>r1UC^i~;=dFBQ(%R|xop(f03NToTlGbNOX$bb~DT|T@ zEg0%eJe?C7=%ZuIm0GP?#qJHRk46lfJl+xM$Z_OZZ9yp<>^Ll(xH_cfA6{lzc#`O0 z9)BkzP9L6MM4V?fabt+C@|mo*p@~=(@I6&XGcc(JiN{2H&viUa!rpO|DiduA5oFi!#N2r)_;m1(Dp3#O=Y^u^IA zo0NO#FZk?Mpr<3rYpWwHkS}^A- zWsHUz?h-_7%N`d$7C(-$6cs5UWe^B+V4Xd}b3@V1gq{$rBZgBL!3M zGi}L+p1cK5XvAl;d|#Q`-~@)rgUPXJR6L2IRh(n^ zhIFhx?`_;9T<(FWSYAACTEdx2)ii}8CkMqhSXA)woWi=4Qwfd<#j_r@itnvw=#;rC zL1sj$d@s& zQiHB0MH7Rg^N`x!ILYAXJtT>zVWQ*2@~Tdh9yTI4rYq;>KD;=i?D0BDQ!lzIyqi(O z2`)TLGZI=iP{PDR*A-gwsnIP~Q>r*K9f~MOj7bI;rfroH-lgDu0^L7M-#)f$=Lia^ zw*JD#s(9#Bl{N_s(OQ}y<$1D_f^(`)Dym>>rQhR@DBD@;>&49}m9S7VYGF|y_|x4m zf#ZCl$Mkd#hx&O1MNM6Ui-5`t>M8@;4LTB|E5*PePumirIKn_Xm&Hb<+@Tg(PMGPj zF z_E>u&x$J2liV6{n4`%A9oT~0CG>=pz7`u$|*%V^1WKxJHrbP(z%1auKHk*=OKC&V8 z=*FsoN)(=XP{&jwybLl^tQ|3^yF90u-UJcHMx*sc6jg|+dK)dMCQn?4#6Z2Q zwnl9(P)kjg;!yRhkEoU*21|a2cw+MVLK;ltwpYz`C1xO%(n0!GjID2?YyFlwjDCuS zTDC^6PDgLW!4YY*uvk)O;fhILo#v_7qIb07t3pE~Bo(^F!S{u@N3==>=zD6aIf!Pn zHWYp=iX5k8Yb7+eP|^suCuziL%SZ$>U5>E}l=EYi&?HbrJyNzW{xFd*0a$#!%^J@( zoMyK%XG2RzO8vGh-AS+r;B6+g!Wm1%AgUF8UrPpinQ`Ms6#M9wQjCer5*lKw^zF7v z?sh_*26N+jge4+ud|JcD?4hB0uXN+&RhS)nGLs|rHV|V#9BU9_RPDC#SgLm;JGrMy z2iNN`s2V$DPz&;!sw`JN%P^EQh0TFq@+GB7AQjo+o-j%~r37=~t}GF+5HUDU3=xB?N+Du!RVhToL)2RN z2`-*>DO4;RjcV=@h7@akq1b_sh&0rx(M} zgp%BY1uoU~P%$*rQmabMzY(0n1vfQNHNxqZbI3G@9wCvr-&Us@s=$R}jo#7Il7n`Z zqP18=iv>~MA=-$|1Y@~?5?!oYTFFwSGTp46y=$k1hDorhD)bE!%~o$(OYNP-SsvX& zqaMK$OxK3->Wb+VJVfl!Ngs!QkrJs!pkcFyQgsdU zQaufaQY#e>rRFIdO6?Md4$=6P{~XcY=<$Mg_#CL)UU1eAR`Qzrq?ohbA~@;H7@WEd z|6pRiE`>fZ9H{m;EWy?vjsZ{P8Jgi{2dG?aV-Q7X6gakXcgTIM9p$0>BfW3duAvA6AYbYG!!s# z*bKq}Pd*P75UtWzA$9CkH$MTGUzc<0rfY!p5PbyNd)br;)hh1!z^{@}VxGG2KmchA1?aYrHe8c%7s7qv34DD*g$ETN5O;JZf3iJz?7<=_dryk?U|(3U=Z3nO@_ZSY(b!PA?7M|0&&v0g`OhrA6+Rw=h^ zVQvNY>}c}4XWYH9dGz!oTBZt}{Gxwc#es15pp_xuKf4ho1VyaPD@+Pb5vZjyf-tl& zgT$vnL2+R)NIW*A_H3oZGH{EkN~ln_DerMy{I^DMMMeW`{I{@SOs$53tRc8tM%E)^ z)YD~dhS9DG<(Au~sos*Ss=C=lGS%ufRX!(kYWBGWd$;O|qshv1PCN;{8kYaMUer~r zr0C3yQ#*zJRGhW3O4r5fRB*?Eq})o9j+#I~Jr7GexZdAt?h-^P91&4;>P}K8d4JZN zlAt0^lc!FaXcjyR*P}C|cQWc5(A6{?v0I5+m?1_>XPmegrU(Z1#Cp#sK@HU8r0^y= zUR))Hum*#Hb%ivr9BEM7(C(fr+$70SYJ9FxJcd6eW|83+uBWb*#**%FJWpSUUG~5d zOVANI&Vh=TaJBG(+8&4nQ6Fxkker-T8g3j$ws8n84y)y>1iUOw#XH1IC!cG{^oCet z)E=b8(iUoKv(-jr13sXX!jE47U$4FjU)hjT-_=QN;cv+MMq5Pi&5Sij;for{rSO5L za4hr%?4Uo{G%-9kk;4)N1))$|(zHvR_%_q!3y+l7DhykV21 zte*5!X-Sm2QVDOp7jr2kfvG}vrtZmUG>d0kdJ>oSnBH-PqulP&6DSY6-W9$dl!vn* zsq!kynVu0XB&YCBS&II%eu_S2#NWLzYa+h%k zg|`=FS5GoNMVrq~4S^Wjn$goMYYE~98&aQ)5qG~6Fw)mL0`F;HX2Q=EVP2$$XX_mv zjhY54xk|abzU(w;+Y+-68w)-dve%*aLzC#sHg@6xutr>u&2274p7UQ;!Z#J%%0(0! z<-<|aj`w8>n`OGZ`Uu+*DUnb09jPt8nj<%9RK3b~q!y_nF}pS`r6Ly1)%V6t_?FBv^Hb-T#>Uj6D&&BSQ$~~u=&}yt{%yKMmi4ext9I z+6x?xy<@{f)vv9Uva=?S1#}#N{E^nt!xg-O-a75x)|7gNq{yGV@b=&APS#~`*v5a% zVZBaMNSiZxbaiUPdYvxRLFzKQyz}?ekG5bAY|gjf^wLagNG;N>Y`0yHYZ-_4IU%)z z)gl_&GfR{JVho-!2%AB!LTgO_@Ty^mMJ<-+@v00BER=QYRgf&Tc^$qAR0q2U`dYDJ zLX(K8l^(1OxmzJIM$yTo*3l7HTzqPsYCSDiQD=|UR6pei^A-=o{tHs;A+k5OnDL1OUR!89_u7#xOnU+g*ACvH^6OPWZ-&_ESf@VIUWOEO)TN6ZZ=PkC!)i59NZF# zh_jN~S)9Sbd_HEZh;$)E%}dXO=A zj2Dj@qXSisz4=a5bQNZEVIt`%wMJ6kgV)0SXd<=IG^c$VPOl6d%3{qPr;OUNeLLHB zzY4aOUTkq@*K9=wgNycU$99+cuGbnQ>6S3=dSfKezVP1BF^u4x>`eQjICi35i*8?X z4ef#|)E2p^9xpJm^vYMP+P8B*E&x^z@ocw!Y5wXuwXbyf$fyDGp>-ifnx03+lZ`m9 z@l&p-7I6_k?a)Qaw%y{8;ubABTJ1V*G|H)32I5&G6=2a${G)0{ zjIVSgmYSp&Ad9(nMCe2gWhPC&6Uoiu~jhbqT>#8>AhmFBpOwp)CXk+7Y11LL5J}G;vd^gIH~@)HY1wgRtevRHZh#;qV68 z#>!3MwSjC;8Z-0R?%s5N$_#KVmF?c*&2>5_lz>gGc~0R;*P%(=SoVh_<}gDdujTrZXr=wfy*MUU%k%s= z&QRi|myH^sBI2JCz)1|GrH%buF_*p>5Ih$lTR(B;u!#4)S^X315hf}b_uG6emD@j6 z9j6+N5LVm#uZqX$dYNrpQy5L6yS}^+$OU5&L zuWgU;z4lhyfx1mJuwApD`%-p$QDU#ed*G-&$>0mn?MeF2I=3fR_g?XDQ%()w+jVTz zTeXn|fHon$O;l^m(>JCZe5MTAoJ3 z=8O+Pc*QTeeFeRY9*8b`duP(j$#;;4^^aBDk+;d~ZBK1D*yHWx89P)&GWWGG%!Wf% zV*p1Ts?_Y0a$?J32uPPJ;r4RH^d&LKF=P9#1Ti#n)nTj=(RT#UQ<8J&c)!hkVQ%FB zKmEZeADI1Z0yQ60Sz>l+J{z5`SL-d&6K9VZC8$!1&PR{0fRk%H7sa&BGB5QR)ft~5 zVxlzwQtf7{HBs)ueL3bho~K+f^9sAQcIb>~pRvZ9fDxl6`?*40U%pwzas+01nYz== zS7OEjx*U8=v;V**kiRY(4E0Cmx;!uAtBWLbNjgD-0upMUK7okb9ty50nw3zus$3yn zz38-(P!H(ksgfkHs9f2K&l>4##=hgY+^C~`q>rDva#yuXOG(l8o}jONMZwUh9H_RJ z;dM-xrfDNFK?PLlpyd#?t+rmZ%mplC;TF&EiR-$adoq?c%3W%`gg8#=+UuYbJ5dE} zm7^x>y6ADnQ@Zd-auqTJx-O1NOT*otgrX`2hEAd^=iGK{a%4VZ=XMHZozjgr`r|V> zu&kcIPq+j_r}x}47q7amx@%u6ufH+2_yuG3?e6hhaxz-)c>Mt%aw9*g&b&nmp*$9= zd|riSxsiM{FT_IxjSB98;=eh(xM@2lLB(6GTBCoCIB4av$P|G_^Hx<@NT${11bvpy zseEbgujBwp+({|UpE|vkAf<(e3HE#fSNyd$w^#-{7u0n*0ACqI_ir6X*XYQ5CCyzi4_k zOyS9;X{>Pa7vijCKhF2A0;xah@v*j)sFhh%1$%cL6?Ki)tWp!jG7h5PQ$Kj+ub3@I z7A}n&gRMO^Y|!}aTarj{Td+~WjB$aay+$DQ!0u)qPV7b7+xeRTl?LA&x5`kpf88C$Fu+c zi+Pb)1kh+SPG8VyG`eYU(l=cffg!1al7RH*2vE-GrB-BCN@$(5{tal2OY3BEPt44x zu=?``c%)-y>ewuBc5kI|ph|lk(rFK22hHajn3Vr^?fyiE4nN_DhBx*5;mwq(RdBhO zd)4QAe*qz>E9m!3URTB9z|&po@1dbrGDEZ0?g5d6Khoaz$QH3B(IT=gStKb6<|<}F zyG&+}`pmerW*n0lB*OPd?ocJM9i}d`!xsdGe*rc`OhIDRq8! zXT@}KBOUciZ@PEDKX7LPU^iht``=Ut!x_TMR!qT~ir0uH2RKPd1VlKBT9NXyw zpa1lotieonCv##w`6l!|?872jXv+~_J>Z)-neg&twV&k6NVFa!qDCU(^&;un*-EfOC7sUVFzFsW z$V5-n7K_yNCnfsj;1YM%GUHJ}lc&SacFGIio!^BSl%k9-2;r5SKmF-D^gq7JhrjLH@;=9noYstme^wQHNQ~Aj zd4&ZjX#P13YCbREjQAq%7y)0iwshYsGKfujY=)oqHqQ|3hZOUBK6w(BBVTH1}em?OKl&bqn4utKip3`o-Pu|6IDS>A3%23Hh~M|Nm6p|Gy6&oWWs-ULG)* zYnM-7HNUzfsyp)iKm8sIjDeN1b+Sm>{(mZCpY}S7F|2=uWluh5SnbMRA?MRgfZc}2 zx&OJRG>neiXtk>*+MCOFmKUHkk3;q#l9fyCakkh$)An9-ah@6qh zKUGxZ3^?Yba)w*5|B$$b(LY;cSncux)wht^@cQQp{txSz4E{+XCWAQI>4h#FaXZXv zHMakU)3=^(7n{Aq%TwMr?iN_ULP#D=t!Fn+bL(ItcaNG}i|egPB5Huj(j zCk$D8R-(HC>u>;(m2imi7ACaW)A(H26&9y>kts#3_9TaT1W(uPaPI{FQ>2W9L#P#R zno0D8-ry_T@n`g~axs4tSw^0<>z<+)$oRwTz)UM0rIgebe2~5nSDw3R`>C(O?fISg z`#6i7{_UIi@JcEwQE%a=)(Y}+MO66{E2?BNRTv}&U^r7*B#+$6zJ~*NZ!JD(8tuwc z(h;*mal9YkzGt}^m~iF5bSo8ajAA`HVXE)acL1Y~WUE|KXPJ?pwlYALT7z&?hdB$8ou{^`{FnTlDZ1BqiIc3!Kx!AWS6U7< z8_aXH4yMUI%+phtOrVd2p#h@+x$;Uf!2AjIp5Wnwv>^$)+*@_v|DV3=B7yUn%u>pQ z@#lMZiHx~s5ca|p@Uc@BX`PXu474I;5}#0{VT&czr9QGQQYIlFMH+jAS0qUe@XIut zd!Nj|UlpZLh$6#eJ6fiZliF?C$NuZy*mjzi+uro3-4URpSE_ONq4b7oPjE!@LQsvC z#W{9nPhHol)QJf~N3m!gl0A-X~XNKX%?Ct)vfCuoH+67rZrR3un@lBX3bSiH@^ z2dG*9`K`9VbQPY6TIeBK1HH?vBJ+_9kgkMOo4gA_~@MaD4D zWt)YB60x8D;HKUT+|OP%b*knw?~t{h_TJ&5V}Xmpuiy#lCAr+_O@ELzasawwJb1;I zDYgAqwikX9Z!GTQE8$+sUzzxRgXAr6q#qqzA`Ye1%dnw%JDRUer7oY0E6!Jn!i~yW74L>q zOjFo*m_3WB;~nBa{k_Z@mkF-6B=Mbw#cX;rx<;t>0b{e;`)fQtbdCEv$o9x=3TYHu zUi4PAU49px3vNw+(-W-MOu1WccGEBfB^aV44*5$5r53m@id zBBnAy{Iqs`57E(?TV_FQ?>{1VDy=_I2KP8=)C%x0xNILe-L;sOOdjFzXqRC8oookV zRlY=o2TvQ>SRll6E)=5T?QO$dr6#V!r3TptkdY}O&a|tRx5+}DK9O38q#Oc`wS+ms zNEz>tcP?SpsleXdYM`#7E4=R=oCp(c+o_+BHxj|Rf9V0kWkI|Rv2wlVc)1+h!D>@10;tnGARp$i z?r`ZEDW^Oys>2;*xaZFf-ODFZj}%qk<_1aAKK+XKJl(}xNH?-#`@SbPd15~ez)lbY zQT${76b2#!CQ^R{(iM;XDEA3s$53CsgT>DsLBeQx1)Vcffz*=YdWuPT*F~794|h}G zQRJgq>VAl%@cYwuF5jT zcgIQrqRx^6$Q&REFtoJ0T_NDXr|H7?Aqv2sZ>Drf3YMP>f8z*M-@uaJla_|2EKcG| z@MNw89}C#OAAgUv9P1sD@FBdK7XXl`6joaw)+}L~5Mzc1=>IV|KXY#29%PltqkGjA zH<<@6eZjw7&t~`6{%{@%>|8ojbPO)WxM_}-y9@2;1>Ev5nE(`OeoOhj zad-K~z2zHsmT%lwzHwLi#y#a5ca&egy5H$rBE_zoEZ19UYEvdWYU$Ea5xPK`t{$_w zEFZP;QR3j;JXZd={-FZD&hiOq|EDZ?uvK>|<5|6C?t`=looGy)%4;4+T#1)cMBG@m znx=a$g~?8$-D8m zJ33?T$y!yKW(VO(e$_e~sw{g)VcEfX(U~vjoArX8D!MdMXf$dJ4xgi3C?a~QgDF<1 z&Ox^p>j*ehAe6Gk`SUGvGCPd?iBl7T#rw@aoT=u3iP;7$98|bs=Jr*jzHRYb8{RAA z`<||hGm43m{SVEv__x-T!#cR%2@a~3j1TN_`9PwInY0;$Ngj#c_j~}rvgDcg%T~4g3u6pC9(n=OfD(#9CJeo z`_D~#$z${se#xu$DG;+JbjjfP0w#E7GL-izX7J5T&c~g|72xxG_a0$cc`+eoS$(tS z4(|~azX=i|$TVd;%c@z^AvFLJo&+P#e2C7JwN=|{GGskJ4fows`V>wq4t2M|@6*lf zx8FXhF%SrU`>h1bD3M8YIsy|Vt$+r<>jnf6j@FvhT|3kLN;D;MDdLRLcP`v46%7>(5cW5^7fQ&;NcLd{D?97ZTtTyL!-PJ$g zb@KIUa|44emWnVt+#}d3LxW2S>F}nnMsTSNqSkKR(R&oib-%$Bp6oae%Extxtm&YG z_`iM3J~8@U-t`DX_fbj^2(O1n2|NT|y978gGaborPY`p;`Xe$~>#sxWQXv28*ESB?0kBADJ%cqbf;ruyeA9ojs5hw~ZSc1G3 zy>f;XRT}mdd)Uqf5zu7T-VA4#_*(iibWndD5J#32;{S%5H4b&+Sa#d{w#tDr-kvVOUl}g~Xu^`&qIG(zy{ugzzi%VJiHpBC2YaZs!JxG6XnbqyKVI4{TdSfh}H zcDP)pT~1--{-E@Qyj1t$RDs_&E#4atdGUd8c^M-K=8em68&5hjxVnUpUjurMk-B;Fd2lpz+@$_c%+eJx9s4&T7Hnez62&IBytj6xg0zJm2#h zMz^VPWWiefS#LI5X^q*pQ!E1kS2zoU2#yF3ya#rbVS(2Rudk(OjcY8Jd9a z4EGhfd5~DZ_#Yn$NT61m{*~yYLq=%UA3KcgjuCEqqs--41?VxCe}XU?kyT=lwjQ{I z8Ndy8@3Hi478i2C+cmqxVeNp`m_nP+NDnrrv7cflt>bobaddWgdhAor?m}gUNGCyw zqtNc}R=cA$6XDL{y!e7DUeJ}gJLfK3bmubg!KAdmjI4!6g4%wE8iy< zc^v_xX1MO_1f(9!W$dxC{Pvpk)tY)h@~4h6t}E#Bv!%3 z$A$g7>iskC*|Z$tb|#5nJ26LaJ~KyDIyD1Pn42S1PtFGDLbIauVJc6j?O3KTZ)@CN zng-!NZ~=hP?G*0Ygfog4=k|_AKDkGnncXdvF<41eORTxY);;pHtt;^1%>_Sk_RS*P zezOR+-zo2v0N? z;RTgNT&6Q%N=l1hqcsK1)E03iy=i1waUDfLa}k-SE_;Y5S>g*%6+Pd}*7`D)UG#_O z((yH;q?QDi@lWxV^Uc%c+oP91@8fEXY=YddGf+mNX5hi&5dLBZHcu%f(j~Z)$ZLct z<@w%SyIQ{&JI0tCMWkJc9EqFp)3T{i3tE75UkZ@4A31oT4@E?=|2VAa9(W4&NM_CF za+y!BIk(uyJ>MI-AXTSA2ZAG=iH$hV@gd-#Ux(xjNUVjr{27kPbb%oL(Wj%~0x4W^ zTV&d-o-?)3O)OKVP*y^ns9=pv7Y{H0tDPeSJ{%OxIUtu%f;}xvS;EsN za?wFJjtSD^jM(J$2JM9ILU~|S+7Uc3~njxxQ7ANhtOTw1I2xmjK#HOq1@QMYTfWO35=_lVpVtf;2+3*Gku_xkMz@^bXQn>a41s86-V15YQcEyDscO0c$j_;o%3D+#gowgajGuE5;s{eJgb)(sgGn#FOWFf5m&G|0 zpDkQjV}X;4z2o0E!8MIvCoAm!7A}6-6P&D0dA7wjw;V$u!@YZ4Z=BvvdD!bIvojY^ z@{k}BXx9aBhzl6q9nvCJCnqNtEMg97PA(n}u)fNBY~C#KAk4{u`NP4D7V+WW4|#ON ziT3ak8&lJ)hnGm2k+$IBWmiaC{zoToE?or!JWh+mV=Kg7Fo7%C790ioNTGyI`hJl_ z54Y>1Ykfi9)jNETd(a#vhkTKZlK7NhjHzOR#SElZDB9rM{PMlbwLh?y0@|ga+ra|S z*O9tQ5~P-W_<-yhI4yMY7{7yNZxLbA7=_~ssyvuw39vJg{J4zxtN=QjvXdfR_bm;m zNDJ%dVLHNv-0=Bh5rxS`Vw6A#E~lC$YybmP-iGz6nIZYpVK-1l{QWxleU$v3-aZX* zC)nHXF4@cl(#A~PlR8pWdRSi+eY&MG+s5 zsS6Z7q6HY+u~7Iu!dou+rVchJk>e?o)y^lJdF5I=y8tN8+Qaz@K@{TdIvcQCGom75 z2b<9yhob>a*GwUmvT|in&rO5OUy>KWvvHU(58}MML}Y9LNap9sZtNlQWjP$V&<(th zc6o$g87I~)rX8L@P$Ipoz-NaeK0-UsAFh5haFhA8a};pG+t1SxJ7aga;ep=Pa@D}Q zMOAbzGyq_&!FGb;A_?8=F3W2-nkKCzML)v{eHcFmq$f0v!@Go6q{OwJc`1^%xQ~d4infTiqGyKhg9VZh7 zu=U_$FlU|Rzmh9%g)is^)PN3#Q_Bl3L@m*CgyI!A(fT3;Odu0|hHwZ6md$U(pXi$VOm z!4Ge_1H;G?B!|3#5c6@jTF#qC476jH28kj7aOqw<4~!oDX{Y~<0jlGP=Fy=jCl=kj zFc02D>r8~tPp@rQgfRmT$Kp%CljC0e0QSa1gew1C;=VH;I^@&}zB>y?LtQgq8~6Yo z7<#KFm_+j9J#TY%?k<60-lEv9u!gvY)x*_H-8nTCXPCS!p;X*=^IGnH(fczKQ`A3)t^P`T!}AOB!Bnm1Dg{ zuu!Mx;p676 zMv%c{6he@79U<~IW)40Xht2`HBJl}>ZXSYN0t}N)&xZc1u8`a0%3CC!qo2~R+R&L- z5fa^p9VqU36$oCxlovgL-4TYRWopl`<`855?q+qq?D0Ar+=dDi2rp#c4VI&zdu}BU z!<_^9p34y}lYp1ZUjmo=Un{&dHNdUa>F2j{=}0h7tAP^8`j0NrD>+StygMvVp&+Rf zxJqE1U=^~=1y-jkIKXPesr#!!r<`AX6v4e{X}noi*Hs{>sY>I!#i3jY?hBfT%`(*% zWrhktCV$#_vcj@om4T)F>-&RGw2tk`26on@qQ+rw z>3L(RTk=VdU2LU@xHr@$b}fKz8@s_!)}!L#;AH;WOSda_wet1n&W8H_ZoOXMJTc@9 zdgsS?3Lj&T#Q5S2>3lFbSWV6qq%*QpBy-(4j{)q6#nf-Xwo>&}$K^uWADDRiM{f4= z!p77s-5DC!zTgRSQ_8K1}D99&|2MULl=tt{aqQH8d} zCUgW=omaS#0x!#w(a6$;wb2(OP(+vh^`T%97;$>&(q+HI+BiT81Vipkaj&b19u7%b zg?YO`JaTq{OJj@+D_hz%LMm^Bukb+Y{h-z;TsNV)vASrM(c&*uosjL&^l_LqO5A{B ztyQ5sUcP?yF5vmtE@p`eN;KEji!1gr4$02VQ)A8FXC_ZehEpRpTsm-GcV@CC2 zZa*+7pTUtL9o`#@&H^bkvqm4z?!TZ&l1i=I$^Ba>QDQfQYZ@t2_iF%|InLc#2(8mT zh5o*2HVwW9$52fIO|wJq3go+19q7J-^|p?QE3!x|9xAA{&rI~S7rz8A*khIv)=YmELDvyZ$s+v zgP>mp?=Kn+Ut5yk{(?(T6jbXfCKAIN4gx#T5sQ?mytX`?)#OeI`s7fw#g-)aK31+V zDsnYyWOF-h-_ACxJF%b`23|ouTuxUp)$DhX|q)H za0}yCvhGznlgtu@uukH?4PxBEykgTJi>B%3{MW{YENi{apFSiT(ZnMCN zLk!C+2#m~AB_nJ*d9APIB{SPOBe@8tEJ4vlio&ebL$|YhICtV{(XNWf$ak-(xKw0K z;10UQ_2$cF!te2f~Si+-_HcI#W!!%+6YB z=Db+%_t*W+F{HZ;oV-LoSu8Q__QSIzoed-&)*{#B(#vYoZIZd-o4 zU8ZAkAGztEU&V^BzJtx0s7h@au0q*OhVInW>cm-bm?C9r)fc-`I=bs*GL@P5`Chew z&a;=3O28!wPwtU6qo9~txz+I;@fbT5KwHZuOU`>b$1%}#EE&MkwSiYqwzN`_)GAdZ zeU(A+3^gelw({*o5~uCD)(ufjB16nwnNDC41PVr7hya`a?@6+iDDCho#lXRP0iqsT(|IXa-V= z0+b)!RmmbDJog_n=XlrC@WBv8vLnTwl!J}#q^KsTY%N%?~JuYIYsPw9AE*oW9n`oKhyQi&e83Gi_ely>t3L#c(R$@~IU`O-4o1TiJ5T*wCkK z6VmJnI%yK_n;VwD0#`UpuQK8)O`D?{$pFgLswlWi_uc2|--O|jIu2y?W z*H^b+qw7|XCpdYtm1O0`hc|@*!oFbM*MnKGrL_rJt+5FiVrY{icUch)5r}qCrEGYJ}`RDRIWwmf?^Fz%*AEaT22CO z4;2gv#csayXEc;1(^gFqTVgkZvnJY(hNmr#&xFqzk|XtqDQ7MDC-PSfwv1>?RU6Qj zs&Ay~UBi~D3#7IB3#5HQU>C=cz{hT%?;&Apt{h|=_rrzV_BkI%?VAu%a@=wG#9+1U zj00t_$0ud)esoKLdIwJnGQ=XPVtZ@E|LIShYd)z$74THLQ)*Xc7x6B= z7n0d*n&Es?a7T6(WyCvbyQf@XQ&OrAyIQyO^y#jvCX#=#fXJ>0KOl~qy6bjkN+=0u z%hYzsrII)}<)eq(bnVtKU%>5+zNm=bOngz9vfHGmAuRC4D)()&Wq0fszATRi{prFN zH87Y)UsdR98yYG-bGJdhs%~HQ8;S50T^4I712GCFVo&$b%gYxvTNnJ|Hg8#&x>;i- z#_j8;(g9VeRUQ2fGP&sskrp?IF!hK34BV)=;#_^bg}$^>$E>DzQ?K z&S3}^Hv?9%F>a(Gg3#a0D=cV%uRCAHbvSAEmNlcd&=+X>k^br(1=O@(Sp~n&wo#y~ zA^ks4RadmSrUgaTYc2S=N3gv6H{xTM749v$E;Xe2_L_jxV6GKS@v(fYph)v}nz7R1x;doasBvUyd@fEz?Z6-;k1dD*)D8^FAIOP7+zu=%Q;g#w z$X%IqbcI>h@sqWe-Z$wtHRV~Vj&1cr)CyGPz}6+a_?aG!I6)&pf9-Oc5d1z%_;#?Eg}l> zw?4x-K&q8H%y>j0e`$4&DUzv|Bm%p|5NwtR1;(Newnl51ELO;MK%ZJ64%uMgx=Q`a z1$@DntWnwSX{cNrdC5SllhwLvp+w?%D^l#nE{l1#yolbN5x>I8?O;8f&9Xq&oNq#% z+T>dc>@HVfV_~>YSFe}usVFs@Boq?mY|TAGfg~Jp1koVFNtvM< zVc7}!=S?MnjVGvm-x9Mz_pXxJQEbRyx=8X(>+}_p54g?4)Mgh*vxBrsxJJe6U^x1d zy}bF@JJ6@A=SoZ^`G&z-EhwAjYAMfb(4xirypd&Od6R4z&@vA1;zcgFh@!5BZ0%p{>@%S+l191a4xDLD8(`!Kl z{;VzJ4yx^8N#GarX1mEzXRX%06Ad@@sm$8@zo^#e0!9>7;Dt-_m)HEFR`rWo-7ieT zbR9aj+X&?{a2G0fdAJLRTKj5sAJp1+P;1}MHEDmYN&9n6+MjFEzN$&}swUN|n$};{ zzU-@-N?z&keEm6I)mr|l*78@imcOdC{FhqGf2pulTt*HVf++^Q#q$NMs}?C~jq?iy2M80#^VidrWO=uR z2FUGqICGh2kP_1(VSK*F!Bb%u?3TJcX4`AV$S)v=DJxaFOpvXlo=ld{!R7{Qm*R02 zFY}i}Ig9wLiaM3hWj}uR!e`(O+RZAytn4PhEGfMQ&v3nS^>#Qj*YM@`{QG7y96h-OP6>O14(D7f zX_kzeB*{jWV*~a>JI5Af#toig2_OEvIbWW}#A+JnHfXrDaM!X=<&NBuXpN_QgCWuz zPm!e7lwUl--_PSG1%(rU3DV{I@+BUQU)nw3xZNOW_R;10x2jAte{kO;BwlwK)Zna} zX=EBN9&j-~p67=#JsZwHWu2j6DssGHv^;;5r}A_U5**W%Up9|;Qcg>aQuM?oPE{6w zW;5R<;KzJ$0+x!rw{(8}n=%O%b8qA*#K zmqLn{v|{%Hg59$L=H=HCL|`_okh6gIKAh~wGpu9g5;XFyUtkSVya2gMy!3bIF3l2i zH#0{f`#5sR>hu{7J>XKdOeT-J_5JL!S`zi z>9e+r8c5Mn)_ur|^qQh9zornQK{EIkYvMiZN<&9jJ0=Ur~CCPu9Y z0+X1FM>eJ2Y^E~=QQhCSb2r@a97IMG&1FZDRxH)vD(m!6M2>Z~)vW|5jD@(E$2&gv=-iP6gVJ-p$EO);+sn3QT0yH8Ji#T)r>JkL)PjgDA>MLs6iz^MxRot6JPHNKL(&4ND?&PjZZB zx1S32LMSRLcg}~=Z(3UBNgH} zgl4x3HI(t6ewRtR{_*o?-kQdt$rWx3WStk!UW}UC&1`VDS&!St`RPgrut_!UGTf3I z&k|PlI5rBpJg?k!lwAmPnRM#2t}P1{PMM3|Z-8WW@_jgMJ1#`ay3FyBH9H`s2Ntkc zgv;xOiiFOBoJCujfK{hrxroIwl=SG3(yCsHGUILsgx*=(VU*&ZtCx7CMa2?@fN4c+ z!=>vZY%v`F7v6iyJ^-u(Nswu`u&GLfDsDWBShZU!;@wAfhG>x>ZU)L&ZRVFLI_tfN zw-$Mk#8}Hk3T??22{I!zqe?@K%i|IW)3?kxOLSqiS5E6$`WF`Hq#nT4xo)lRU;0py zsBC(z++L+Lb9&T@D@JJ^QGsa;#7QMp=*k4M@>6ED0vnW_xvHhDs}0QIC6+8vS|)_S zmQ}Du`$Pm2(k;w86ig>`!jYhpU~Y%y%J7aS2F4SqaBU6e9^KI9T6s=ehw~;wTAJMI zl;z`ek@CSO4D^FMQSNN`dF0|dW@&r}OW$ukPud^%^=aR9JRi#Ac8XJk(TKFhAI%X> ziJ$(FX=ecvzi0RpkVe1?l>qfgPY2aUg%PSZ5E1wy6^2}kObt)}*x~sS$7Ex9_fYd$ z{bdfT;beY_Cj=Pb$py^d;pI==qSI9v`GdzPo`23(+E3)EZM2j7g&v^Ei-Jy_5vWxf zfm$OGs5Npw{4B-*gC|}$wAtA4)8x_f=213`F{(bX|5;C&9w^r z5^G?|orkH*!-c`}s?G4(6^Cp_$5Bpt4Pq|~^g+TZ++|?~g)!%Bt;V=WmqMgGl824y z0gMW8b<&Ibp5sR~FeegPO##38sej?p^#H6+3~qxxtjM?^xpN|IWe|#WwTuW~IN>4! zxUa!YRraCDGo?v|hEJqf;?iKdv~z`fDnqt$$+3dN{1=RGBhK%_`PiMXWCLi~g}Sld zmCeB=QexB@@Qg3nXs`U)r_);cSp)Wb`Ej_^XIB2f@8iW8P|;h!Re*|ykG;+o!IlNAs8x#D{~tGf++m9 zl(v|SA`Wag6nT2=mKVOeL`XrR*aTjin3~h+z>CWi^pISZpuraxFD61Oe7Rbba<)EH z2}{w;b>dtdoz<6xn&v|E5ofCvUjO@@djaZJ>6MQH9xvL_9lwo%GivuNtg#xioXMnl zzRo*hT+$Q|2vJYHVxW+MiJ)IAi-dSEG z=%U&zTDRL`KFqO)@8=y!a_*uLiYPOVN$KyZNrJB(kOlTmR?yNXgvt$PaDR^wGU+s9 zNwDPH;6_&g6(4@ia8+eVy&jvkMR2*A1k6av&Wcq64i|VrXMEKjWHLm|$=t`i7K-H< z^DS5{?D;9omx(%paf~kLW~6kwYMxt8Kks2RUk|7LrTOn2J-5=D)P6a=b~$>%pIkhc zrbVFm%>Aas(wjE@;Uke9$sdN-A1~+aA5J2!_6s`H(BH%MI`2@LD$K?*;?PgqF z8z5hcA9^+V9mx22w~BZwaoTu;-%Z(wy!6`8v||Pwrprc0kWRY=jJMu9jGP-YM)dUL z^f=a?s4T>bP!}v}f{;XyA!KX>rV{q$}#^67kaVP%O|b^jlaO%N!>=l6{by9Qs$eijNhw% zxI8JKR_*KoSL^6Cz;mq`fu3wtMRB#!i=Dz_16pk!a9Pt<4Yq2xtxa)3e)?L5pQlo= zw}%LJnib{iAwyL9aZ-GuI&nU%_T<{5(n=$ruG11uNZlQT4kXiW@{>1PCB7zscOkG8 z5&jZ>467WsCwC3&N*6Z9vAIXq~a z#T~Kwq8-qg@v%6~7&sNer?nk}r80tqYlc_drdVEem56r-*fSN0RB(ma&~F>luN3j+ZxGStPhik$9G3(rf27DEALC^v#X(dgEqspJnoDod3J-h7@pTz+m&(>@SU``o46<=5~Pc+SfqGm6HnOCVO;DE?aGMCEXW zl}!%e06APT$`gNJj*u_wrd9k3(+j-iKCQT;!Zi@?xZDrFf`y-F%4_Nfa1v$;Tr^7w z7NPk^3fw7?KVWU*-cI=Q(MiFjXb#VofU(bhi!m>M^8wQML(~*7;fFaPMManL+P4(u z1myp~xt*IT2Wxmj15g~9C?mR!)ks+AM>V~U?(SPBr~DPX>L}QJZbq9m9Wid*1W0}J z@LAPJDq-U|Dj~^3#4RD?04XEw6jq^;wudfIuPLaq$R!yVB^s%w%3iQ)P*gei2}|Ln z=v798N9=%ks$Odf?ktzn*(fbR52b{+)`3OS{R1?~Sl^21)XYL7g?6BLf0S+n~)PKwG) zh~jZZG!tU$r-?Ic<#6T#Is;K77Rj}FjBt&eqia_H+{meToTgxVA(a}^4!-vD)@nLo zc0*y#6QHgtoF6yK*%MqaXh^$cqL#(}PcPwpI^l^8eBKT)_#HR<3}AC(0VBK(yoQ=s zT>i%Hf`7~LsQj_SWZ?h!3kk7H`<@+!f9NH^A`WjTlc9ANlqp#Telx}AuweHgE$qKr zPNWvbqX?-CUZxptalz6*LVzVWcAKjX^%oq%mf$$|K7pQ|?}21-W+98KG(?Rn)GPjS%SoC#u3UfJnz*jC5!rjox9Aci8u!voQQ2_#tT&yL21D`8V zWJDyYjBqG3*I2~5O+qK)k~YorJq$?n<_ui^#J&t{C47MO6bm(1kgPCoKo-E+m-g%g zj=cLA_@4{K7|$a&Dob)EecE$ORp-Hp;yl!?W?3UjT{Dy@Dca*IT{FR^UDKm4K20t# zffSTb`FeR@_M8wLfxA2fM$PAiq(g8$rKD2!81q$E>*Z#&X3)^{z3dxi$m~Pzji;Ep zSR9cofBO#U9qzurnNHxvS^vEg&{DyFB`K=*uHdQmezIn{8+r6+?N9&*LGvFu2Am?d zR^aH5=Np8H@=DT<1>Pdse@D{uoBzg4e==Tp&N;uJhM{Fg#m+so*_FG5#i5( z0#EEXJ8r0+lgDs^Yy0gdJ}?!?!i+F!_i!AFYtY|qrt|eH)vAo*Z+{0xKfzWWH@N@U z$0ApMCdL*afQEU3faquc^UU|y{}`N~{ptC$KmG0RBHzD1!v?xtK7aPVeUcRTa60@J zeExd+{!ib9NmhoI^D8eQnV)(Eiaq~w?GqY2cS3)$%&3b4%H2`Zbwn=?m2oxB7=!bHg{mg6U&(E;p!5Bi) zi)ZL${a-Y)IKuNDE7Je@8HTg_;{^As?iWl3{EXuYjF~yEtLN=zHZ!0d>tgq?J&}+< zKa(ZvnQQlstmIhP7y0uu34i!91~T71BGb(?j@_3Be|{zqj6~`n{+_j?XWt2ibui$1Ax*;dt`9DHASaq z20zKg0GEPt0Ilfa@j^ zLPwf4^RsrIEd7qp0jsHipe6jofTAZ*z$o#*8I z7GeIOPJp?j+nukpo#2iRK_4^MQ}Jxu6#k*WphGAa95Yaf0=P{wgF<6DL}Nok3h>~1 zUfR6T17j|*?N$3>z~K0iykL>v10O@v9TY?YG5s?DZKneCWFJukidk9;zjOWo7Q~g$ zSAa|ZI9x7=h-FbVb>NG&8o7|k@kqvFgeScO@T;T^StAL^TR~_l?ijYPbl}7ww(#Kn z46P|(_q(>bko(z_^Yd7j*%%l0lv*0&HQ@#EJqFjj=D`z&E6du>u1JpkuD(DJ1|z1S zPOMvY8@RKODL@TIfdXVLeBS(eKYWtYozUnGma2nO#mdEm1>!q|2e}O`296zK2K6C& zel5|Iw)C<7&eMWR$T?L*o%p<2^VkvKf=2^8MF4gM-?)Y(YvCtshiH5#J53C)BV=y@K%2o$3$igN@ir%|hql+!z(%SnUjawP;4T)I!g}yqSe* z2jQ^H7ZRjFr@*Gej?gtYjiIs8OiktiXw@iCnIQOxWyh9q#20;1>LFyMadyD+D^*?Gx;r z6Q(>`xRs2*9tM|~1VQLd#~7f+%p5DeAt8OPgQHvE%mI&?QWXi<*A}4{VpdJvWs~lt zqQg(|*>KK-Gz6L%eg<%ytg|v7<_lYjibM~fo4=hWXCtvR&Z}VQJ(mMX#ugl5qdZ1_ zU}Til5(F}iIKkGC{$6{h2whT;c?$vwP7#@6u`4@7+QWxt@Q79Y`5vOi4GwYKm&5sl z-vs7h4TXTMIlBb6MzPVLk7bcxa(sy2@0JbR!7f#4AMIc>;#e>8@)K8wTjPS`h#|*U z2uYVM7BxNMFU{kI+wWEF6Wwe^pBhA&-Ej4fyaV*U2wklaQP&ak>=dyp@GY~OGYo`U z8~wh4Il7gOlk||ud=fWARV?EDG0nH=_41FFu0!1}8`I}!P zn_*P!R4u28%gtPD_;*J;6}iJud)E$C&_l1xg>IF${XQzmW$@_bhs7Kz8l+!x&{L8v z>Tq|@zvyd7cYKdYdMoRPshvU!UkyF`boklUZ&smF8Lk<-d~4V|lgZZJRqEaR*qD8X zNTJP;e64hc8c&f6Ep44cBvNC*6@h^v7gJRISgVm{_k;!r{)eg5 zC-&E@?_fX6qO+XCkd>H=91SzJgq+S9FOD#1bbgv8-rkO)O%dsL#a%5;jcg^zvGOcb zCx1e~A*Mx}-C_b8LG=qHRWb-S*B(D$J;Tc>Sle)zTPIk_)1_UtY}3#^Z(UdcI0eG4 zK(*0ojqN3l$97PKTvau7cDTe+x#v6>mrGm-SZ|h$1-Daeohmn&EG;}rc(99=Xn9IB zA-Q&^3rht}S$p6fuJ<{!0I~Nd+%|J zC9{Sy;y?_*5_4>5@VI9gBR0!nibaoE4rhFK2+~o0TR5P-w={nE`%%Ea+AkL6WMAk_ z_X=M4jtYA!+Qdh+E2d_E9PY@YZzNU>&Je2{93P$FiXdxlf~(5c$f}3^(-2Gl@Yl@* z!M(7%3xd#Ss!B;@ZzD6UI|0)<)T@NxN|V^Fh${9S#}tERHk>lXDb2 zR34j!9ipx@o9un%U283BaF3A+eO2nmrlq(3>>D}m1pivEJPd!?85LUP#<%G7h@-Ad z0vqJc?T{;N#6*ICssC#4Z>CAFbyDJ!^7+F<+_`gkb+6jDe9NOyU5&Ibb{ucII{X>V z>YOJS=?Aa`ALbHlpUeS{)g!az25rjs((Ew%jFItk*%~)Q2i@cFb91W*OFw0lMU!EM zLzJz-rC)b};V)mMFyo~-*`phNN#QG(%D+P5<{lZ0#fe8%m)Z{rD*r?Y0UUDhfu#dL z6D2E^LZwreLiH#Qw*z6r@PPxR?oK@c-79$@I_Jf{GMlH>J9qlmeQ#BX!}iyqXl%bJ z^LbI1*puftzXCwG=$IvK4$qh7yNz2H%bkt`Bmk>JsyLBg;f z(65iRs|-(87FvN)K%8!cBJv7P%`OvP$^i3T=XZ61>V^@$5{d{372KIvH*OLNc&svt zPq#cP!<4WLIe65dS}QCpE;+4}8(}3HCUjth?yR&Ef+s@nWv`5Pi21`bBE#bN&TksS z9#lqz?d!XivVA`<#lQ+VaTK(h5NA;{Jq;vv)h@acxRleD5W>T)wbTs_TA_b=h|2`i!E(REKH zMAv!X#>uVO|9Ix93($kzns_Ll{X(4@U9Hnut=yJn9Nfud9rfQh0WTGjEnpD=r=rdalKD6kSxs-(Z;+U|R2Oj{iVa>TlQV3b5&F#oedl3;$i4CtV{h;I-4~Gg}@;ky45DloMgPTv$asj z93_r$tU()lBAXGE1az)u&tVsWLU*JF44-m?bCO+n$+X~#W@_ie)7N-~Pnx!gCkvwq zyZ}5EG)8uZhxI7-KD=MaV$spCkK3-(;b;MB!8;7J#56ypb*Lj+J>6d%>GFKs9XG#m zuc~xTKnukM`$^=2Pa4xLa0xy&XdZ85Fl68F5eHE>vU|B};3rP)Rg0b@?!i*T+-b@Z zl!?LLI`YdaK+rPdmt4AbCb18qyKmI}W=s^EhFUl8&QA~jdU1JvH0a3QwJXeNYVE*6 z<_|QF*HUMeHu8VLi|a+}sG_!^!4dgZfz1kDN%flUc` zcmvO62uSngecZ-iGv?L)z`#S*qzoegQZI+6ifBCk=r*4`J2P;MuvG70;CwE7tIZAP zF=M*VP1CSpORqziXVKur-tx3C;P3`ZNeu)#0_ zRjgSdY%uA2!x@syY9|!t9W-8*n$0H0zybksY$p_UCr%Q?^jRia(Ai9#HGsMuWxvy| z?Wk;Dbg+?ihAUu!hwc%xnV8K=%;iLxY%thYX=jtP5V4ctmyZk%opCuq?wSsN|MKA1 z;p*_}{PaZ60K+8^aIzoH&^wTJvXw`gVzJdd#>5EiNNbM>RY!y&f!g&XFTt;yL1CIA z(Iz^}QV7+o7>Lrw5{~2F757^cHRoG+qM-zt?0pPYBH+~yxPZ)=`;gUgTNTo@xKI-| zPnd(HZFwaMY4|w%A2a-&b>-c-*gG}6ZzjRXp%_e(o#~BaA|4?20GY{+yX%INiF|Xr z`w@6!r)q)GRS5SHFR-f$$z?TUV7*#uajB0{;KK#nz%}qCg34?eE@k4O)s5_;6hpek zFo0D6RFxL_m zPr9y5RGqN(&^;(dbDMp7JzOFK{}S%|d%dB%*+dj(5%C-s%LU|MIXa)u)J}Bt zy1!`?`tL63B?X){yn zhqu?1M&s>CKckNT;88#tuuC}Hkje-OkW71?Qh8GYsn(>YS|~%7WbOr_*aQzL)j`Un z4uUL0?6`Gf@iWyZH|2+e(-N35t?!P`+KYt6OsacrtgS)9J%UcaN$ zJxGCH=<*%)K^csf^I}UC6Fn|*hxp1+JzV|xG6Be?!7yxCKyM)+^urY%hSR2sz>=Mq zgJn8c!1A?IOZzZ#ci2LWiw*5bCDRgy*xNS=!0MJHfV|U0yg-AUwE6^$d+R91HxoE` zm2%j5X|yU51~ex>kI+AlIJlBeF2Ta$q4adIxGyf#%;e?$_3B;1-{DU4w>+D*xyn?((fi2R*(TeYhaAbB{0o# z{B*R$(`9Pzfv*iX#49$79Cgd!yLac8=jUqVz)#eHt>G>%BzAujpe#^yY_%*zP-1-u zeAyr_TTMg=g~5x7#wZz2@opX-C2IVLEQ)dY3BHp1@C` zwd%u6-nC0y)W%%tIBA7eD6)^rMcu9e@3hZ*|k=JwWK>GV^>7!bH(Rgevj zH$;0Y1jQ)k0COXjLznJ;x@Xwj=uICEBa^vEB4sbtXw#!h9k>g4@`X(>TY0$!a0v+Pdi)bWX&$FemlRj z{?;BM7vA-<-e`ng-VMU|BbZD3y;&Vyc5Pi9l87hEQ3KFb;;2 z9Ld%UnX-j}pQK9~U^o^RKeJxf0k-BD0N6j=4_C_6223rix6|cn-CageLNL!vV!Pbl zug?aDtJ8rZYnR73l~VhGTMxO!5}wgr455jtwFt zB4L7ZMfW&{vDp3ACqI)Pt{3j?;~K}P*CR}@>vRXNXmEBdN5Z$$QL-laemjuOsZz{j zG&t3~DWh698NHtH^cNps>92H~nsO{043F5d6?pK%t?qJxw}a;rIo=AZ zCO=zNZgAY^7evz?y|Q4s60TuQPHnPEHM%#Y`cFi`^lm!o4E0$l^s7r_l8dT~<- zLOhM*MqjJ8B}1?>c$iD~lhGS))T!?`h%?cJ-r)5`We|gsfHOma%|beLPDb(s>FY_l z9h_i5=reuyJFPWHvgc==3;9UOJ<}D-KVMfEBgA|6V8nBdD| zW}o+J+jg?z5uO3rxvz(RsQ&saZHNpt15k_2%FkMF@On=Y_)8RGZXVX&NvGn4;(gzGx;~$#n6#&u`CxG5zgkfG= zi@#5j8!XWFeh7C$J2G04&A%24CW&6=+6A^Xh_llhQXb{O*K3y0s+$|c&H;YO<3C-K zhS!bt_)WPM%a|DN`STr0-9`wOB`M#816Z&Pj-T^F%P`e<}kqa zlycI_$+}^=(Mue&k1qaJbgdIms$)Od7ehq*8$V%_OAKNV+gcSpK;SUxHvT~4oN$! zj}PF%AsL+`+A_yu$6dvCf#ok90WZJR@RHxk8j;^WS#EpE_7MEbv_%3{hKT%5&I}1d z4i-E)r-q{fmUw!4hf=ddRxD|eDW}Yjtei5?GF+k+!!yy4!6n)_v~tUi7H!MUTo}^r);y%!3}49rUQ|pr@A|^r-BhPnI5)9rUQ|phsl~ zJt{lsQCW{DKlP~WryiC4)T6SWdQ|pPkIH`Ple0%D8M6 z99wDk{B+K@%O*^Vsb|ax%oiDzq}?y)@Ju7Iw+(g{SQp2wd4t#NZtBlu0u%jpX7keXsPy;%-* zM(8OLF|Rk1sdCsxQ2o$~dsFP06=wiW@YyWHo)X8jik9e#-iN#_Y&vCBvh0feLYI|$ z2G@e>z@Ow^Lb79UzFpdlF=T*G(%;}Sju>I4C~4WaAHz6G2TjL4o?t-6G+di~ zqqq#PHRL9{Nt%0)YkJ83ntW;xt$*u#bCAGanG^x0)J-})Pw=<+KaCOqykp4P3v^?(?}kiF)R>-W`O} z^KsrpUGjRyuftWy_o(U>)LF|E9J^s9F_)RL@PM+lsgU4XAX-ch!qqZUxzOpb^HA3G zCSTa}BM&REyGAUxapZ!hHRL=X6^2BKECeN@IM7?nOZF~`XPO8}>X!#qYAZg8tE8o7 z`S(x4c$7HbBqRo6q04FW12=eaq=5&bkkdpj{`dB^8Cf}S8m@c0czNjkY?erGEywpJ zbv9U7H;@OM7{kp22j-aPp8jIhH|Tr2R==n7d^Anx^teqY#=63Zvth)a*mB_NzxBZh zokPfAS9Nh-NR0x*dQ1Y$Q;Gq8_0aP9X`LSs(M<%{*zqpp3EpY==Jl#3seAF79!^bE zr4E|3K3tqmn-k!;s?HnSHFO#<6E9+`sF-{B)stGydD1=O{t|)}^E28R5WO_ZD8a6~ z)Tam@kcum0T0;W3UadDbMN(tCZsz!0DVYF!z9G{d;sheU>^vXNah9T^9|Ppaq>ccu z*D}v8v)LoSIW9-5gVA}jW&+EuOL)K=-ey6_Rfa(7LG&%0)#(A_3laFdwjtl*?bCF>;>=8sOt@SwWrwne!YkoQK!Wb3B#p z*dEy8!J@35tQNcyh2!+o;nRmCDiX)$&Z^fuaDvB=_`##gSD?TNzng zEn27nvw=4B0wk;Bvjo9Fk@8`V7p=Sj7W6lr*0gRWXJB1Rs#APU%c{;G1DN9tG z>SG;HZ2|?DL_Kt{4+%>byI_M9gxVvF{i+Zu+M$Ei*d&n-sz8N(66MepP%#b)SZq`l z-yj8WR~O1nlz?k&yti^(2J`(`-zS9y9?MkUOF%QVdlNwCc+lpUMdfYBh)wDO#0!@o zaJN*u=Y+F+^-o+NId6+hJCVYoO0yMDSv!1%fq3hxl?%4WpJB$jg*cm6(?8=UT&kIC zHI)OplGjL+kYRyKWQx5@D{voXqUw_Caq|9Q5vUv5k?Qf5-nd$fQf}k$*k$NOl9EML zhdXcRRQ^Juo1fLffxt@tHl6HwiMm)x<|JQgh}aO8y^F7s(OKo3c4`CLY@p? zF(Wj7ShpP{izc!#Kw4@{ z;KTXT2ub_ed#*DH4*93d$A#QA%>$Of#6mZ z@0qF-5NRmNCxq5ZDMa&B^1O?Ij{q?#Cj5N1!UQVoR|?ZZi7kV_)~>inFv}H3dLXIy|p`3f=}c*upEBc+#kxTtwq$^4c|V?QMkI4 zPVsQ^B6p9=_-H*!;ZD*Jhu7Dbfo0pl=A~`%cb77{x@~J%bIaMslK352;p}&e=d(t4 zkTttPWh%9l+u0-S!B|eRDw0P0j;vO)ed=T8D0XB=7tNR#(sA<^FS+8-{Cb``fkOL5 z)M9}hD`P!&Fr?qh^XxnVSI&4iao?|JKEgIFy3y>-CT)N zz!h3KlTt@dxa^jlbPKFro1l>m(+#X{iWfL1@1(&I-p5ksdZc4mpRVMN;^px1=84`9 ztxs#6=$u}+R8D~2o$JhMBtm}ZQ zE6gjU5xa$xo~89;b=9y7 zRyag6nXdhNYG)!@Z_t0b+tYw?AY6{zH7*NPfj<-myGf=?!DGR_% zDo+C&deiaWAHF+0QFm!w&Y~YueI_|kaI|xsUG7-Y zR^Q9+gFA3)fI29Xp;)mu^Yl0#6GjZld#jxDfZK!v0Pti=hLCI?b??G$r!l5ETmE-E zVWMtofx+))PwfFW_5o!NUCnf&2?CKTL)+{C(e~hZOfn@ngsiX3l&|Dm;rYJqoAFoJ zP$ooTL?1Je@4))VStbi7BT$Z6B*e~Eptgt14hc6P2|8yLY07(_ z;zsR-d5zPVrIs)`wjf!02mzR%IxRm0fm$7EYl5w*tysjRD;DO>nDI(WxV{h{SOJuM zRk*{1Hvb%JWTj|V=!&zMTF_ACB$*Qd;zbPm#vnFy!plXjPzhx58=Wzn1EogMQTyK&m6hMcE z_vnLtd0_MgyYd{Dgw5v$%$}{l{M4OujYADct314`^(WY6iIs#D4JGAv&|J424#uL* zXUK=x0z0es4qj4wyC|sEB#*%GP2x2Gv7s6skOyJ{fsdNb76MGB3xQ#-2-AlCVC3OZ zoFg>q@lsyy8c1>d!lJgt&<>)|HgnVDw4IyI9b9+ey}&Ig9#(%tPS!Lkf`sKKU9 z@ScfuDs?$a(+3DV@yoxK$t&RN&p-X}OQA%TA&~6*asoR1&C?wFq9Q+sI+?Gv`%IVW zS^Se!t0+}@fkt7jSJF2US?G)`&9zL^lfO4c=s*^FHNSY^X0=q7K7rStbtJVw{;&|8 z0F~1R61MXU8Rynr2TAfc%FlF2$Rnky2qsuQOqy92oO$DmrKc{^xkx0gm=D;stHqCn zaXAmxqIDk#Jjo8Qo~^Ex-H)Gkvh{%UW7x$|wx);fwN|&)RVR%O>E0+)Ehh*6B&JQu zrS5&|qM0TQ1MDGw)OoV9l4@g=;a=3l0@s5coVxRx@l+REHO{Wx8lx6U2j;ZKd8@2Y zvEV9GxRl51iOGF90V6$LAGoMxlo`m2N{KFxaD)vfA5siwriauez8(p3lL&qjEb9n9 zFFd-LuZ)L6&}cnQ80I5oU|M)lS|+t=7k#RE50mHwF0Q0G1mmv3yCZjYTf^F{dm>OQ z$=FzzAZ#}(Xl_{*qaVp^Y<-nTUNhbxsjv@*X=#^Kcu5v~MuGF59Oq zRL9!2Vj0*_8H?*vkaU8Y=n%(VXifM&;qr8~d$f2Tz{Ky^!g;nJi}&e!WXvR{OP$c# zFPe_qyB;<~ziYLNH!X}?H(1NIuy90mDK6k3==mBBZ9L6d42RY?@icaD^|9X1f$XW5 z-M^=0MAzFSerGp5mHSbv{4HAG6~~nJYzPt)(ijtI!irG6NAjc-*aXe;{qUxltqze+ zSxj|h_Ik2pYb?;g-DW+;i`9jdM*3VLfS)T1qPb12AHYc|gDcL$SB2->MZOMEUfB8; zgn+Y;0Sa4>x}CxGmV(jI9j>qRBQr(i3Dq%Pdu>)nlq9&0}2BDAshHL{D}IlY1?K(+5WqM74I%X={A> zbcu9p%W+I|0nOX=&^S#jlYCS$=paIPm5qDgsUenm^U?}oycYAidnHk65Xf-lb&tvD z@mX6N(Nma*q#FQ$^~}Sc$My~` zy+1#`Onem%cm?uK4TQbeFA%=i&k0{16bN4)B!upW?}|H~BkXuoT@C}Mu7*uvP-3rt zgqhO>?;^_}FyRsa;hYZpoI> z0ZXqHdklteIVOz}Yi8yzb1WAT{p3}|s*>W$vZW}v(53fFg~q20PDpV=_{O||CbV@Wes(@!)wlWh64KIid z_sRGFjpM4lTfX4=73F+bmmCC$%X}vuIF|eVZoOWx=b#5~npJNxIWTF)e=BrKqlQ3- zy)_z^vK-yD?Fz4CJRryZR`nK$E6OcU?x5POrM4=zz+t;uRiU7)f>H}qWtB2P!~nV1 zw(B#ctSB=j+d-54MYgIjExuicp+t`YQ6&A@WdMwiJpv)c0ZPjRz z2YQ*@hW;MU1Fg_w_Wt5WemY8vrMG_+Q>^F96zR&J2bUA3@RN-H8~aGyr`kGYl<6xU zs7$=W7uu!Z)=yMsvR%uzT#uT|gnfG6Rlk9*SFeF+XML6n?5M|3cvt;x)mtDa>#Jq> zz<#K-ZXsB@ioPt-4*KdVuvK5S@OFJU`g1n8y!=~&vc6!|^2M){FPm-qN>y~(S7Zmx z?p}PWp8HsB*LtaR&o}5J@A)EJ__lXqVXjI0bg$`lRd~J3Zd%smcT_Elh^y!ens4a0 z>A(`$(2*m#@=T zV*6*YrFZxqu0T*qNpY5{$Db zxQ&8}BlUNQ7pwABGmZM(t25!(`i2(Noj#r}|4#56;b^5+S zu^vvIF&jt9ni(=aqh9o#6osR*XMV`;@$@Ohw+L)H=1u7$fr}xoj{k)I(U%g!1>ffrDe}$^QrjPGbP3+8Y-hWXKSKPBsYx;K|VcNTct95JF zt`cVbx=Q)4PW??g^_S}TYkKtdI=;F)2iE<4xk{CGWaF9M+MD~Svo76F9{V+%^6APb z3#0|=RclAaxj_?WjvVzbTc={{e3i1lX4LFh(N_;0!@qz0^pmE8$elZE9X%DYY}i!j z;&{pX?!UUVED^o~?`Fb(nU=q%KX0qyt2?q|+~0#MG-=l@xzV?F$q7>H$!yrhho>RH8UI^ifOP7^l7(VPB*y4}&7$f;}`cgfA&A8ZJ$5#)Gz`B2ARH@RTVUNeZ$4IQ=bVLIAo+ z$^X*PAps<_k^d1sRQRde=f2hIMBZJK7O%W+LA_ogv--o{pT0ZW+)K{F=izOT5(*EW z$z{?GH*Q8#ybt_*kLevIY*FMGQJ3w8i41C(W_*I+@Ck3D6-fL`ECq}_@!(!uD`0WU zspEIJVD2VVlY%~AAyBY&JMDfQ4d<_MO;?WaIw+JR0UZW1ctS7*9+wDm;2P`waBb>y zAY0$b>H-&4lmI8|78lF{mkYk~^LEw_6@|bIV}V${ALfbB9)O#mO36b}3c>;qDl)#5 zG+3c#hx$)Bq9=yedpfx^Ho>ghTV_LweB^iY8A&BAkXeK@(aVDZ5w2BtRJ_;blWK;7 zH^_Er6n4W`M*m33__zX6dyX&KM?|EZ-(K-uz`&zHM)ukJ<{mcywcrY@MU^i1V&A3Z z3@kiw(LVn8L&sR4FAoZ&FAq}E`!z0BM|St?WAp2+`Sn-zOM00d0k;Bs<3?pl`o=%g z8mgMOwl1u<9Acu5A8gC1H-~D?Ouq6-A8OjVD%aw;=La?DV!PMoFY=pK$sMn zedKM~Iwnp>DN8BW!xzzLMtWS3mR2Rg2QdHDQfN4jAxpSH)8Jxor=U>AC7I0OA9w|Q zT;nDUx@s*$WON(xTU<8sR~dQ3FV=%P$NN?CGIS@Bo4E%}i754=itIzwXmX|BKtW!@ zvaBR<#d^v7H1uUOm5z|{nzijGK1bFm0f`(Yvt8tJMU0|yC7_EpXUG`-d~Y~bUt4{d;}krx<*1Tm2c|@*=LA0;TV}-;8?exX7L2!#VPQ1xE3faRRdA=s)mc zQb+I}`8K3n2SP@N?nl=7uG;rpLOY79_5*AP4OmW*K{W_1Lf&JZw={(0beEia zEunyNO5`hI2KZ3za5b7v6Br(fo=IBwK!koWSj?ErDnO9U*r~%-z~?OAxpgvh(5?)T z1DWLBGEh=!!f|zd$Bi*Z*Of&v(FJm8;3Bq?p9pDbhe4iCOwkD09Y5cDy=;f$5rhRR z4M$o>d2QAJKO#p&q{G*JzV{xba|n_txj@<+b-qL{BxcF?jD<@uft>&X8X(l!5{;*L z8bWfwB{kkpSF04x$Gj;Wlaz;=9PERCvdltfpd84%4HoVagcuAb!smN$ILA799bnuR zs2p%lUTNlDhhsM{)ZuCDIy}=xWA&7!P(v;@-4}2)$eUwoc2Kc<9!01GqG@U81u}hlHpYJW&#jnmfa4oc`D{LvE z*cydATUoI;jl2j|=Bbex*A8jsF+g^GXfaX&+T@+0&!ELXZNX?(~VB-Bg)C<;@px#XvwO^%TMTE+Sfx($Ds z6Zn@a5(Hi@;rpuz6hl4-&hhXGbKNAg1l~TqM;f}W^M_V-ILQfAEs4Dpcp(4Y779GK zS*1{8DOA5euEC3}afBiQ5m(tyowvVTxcYYH=usZm5xQM9bqg7@zQr&D+j( z>ypoQ@Dr&hA?7O$?Q*!}HX<#+#Br@NTtcY<3@+(`TrZFz@A=-!Lv;B-b%HOAVY-dv zWj_flU>Jot6SMp|0bj!a?*!p)Tei7xm+gJDynQQ=?(UffE;RaRhO|KHXN%Xj#b4>k zMYqu_Zh%^^j;x?mk4QD@Qt1XE;2!=ICMMUQw>VP~%hCOWVJT@LLjm_Yu>vFMqTwXv zmahn0&fpufEUDUzjzBolPQ)|q+l%TRmCm>ViaW@o&*5M1RCsw0)4%)a9VVU>$;fb5 z%|G@NS~EbE&T@9ry5Dc6E8e<1xp<)4F_brI8Q|^QrLJp@oK~86Y`XCPOxCjS`g-Iy zhpYJLd=*#)Bn4!s@|ua&I;QT)g}3l2zqbm-Q`c85Mdvu_=xrvdC38yftu9S8t!bPTzg2cWj_Vqj<|K__Rw9llp(06IHAKD#oYcBvVC z#=~a18u|TZxO~)K;&wFGipL&kq&pOT=Y1AH_C?!4JrEtV6DioO?xS|5tNJb0tNT5h z-i)sKQSxMaK@&nU;ykg9hF^0R8hr6>WUd=*p+Z;K-$K&J2iQU+-(wpMr=Tq~?n7=T z3SW0SrJJU<(}k|F8zuV77Q&g#Y1@ca@cwS0xxXJSwvlZ3M#dJR7L)205_*=mk#I8J zLc<5fm;-n#+4}!6_a*R9cirQ5DixBYkR@vwTed`GmwjnNVhkpXWyYEoQc6+^rIKAK zZG@zvC`)AvtrXcxi#9D%|9ii)Fbqb|`}=(U&-*^l_}+WYz31-do^$RwbPW#B2f)FZ zzRMvL2v|13#VOS?q+=37cv`VssLBaKcJ*m03xRrx$zgz#S)k1(0LiQ!X`3f17DxvN z9Qn~*$rl;q(Qmkq!3!1CIgt?jfQHEyIhHL+PMFC#AilC45oi<|j`o-@(ls0G^J8{n z@=%V#fzcR$@{bPC7%*fk#(a;jokZD~qnMlmb&{k8k$%UejNSQ>9%gVs9LdVafgpV# zP?jH%w~)CsQU*B7rpo|Z52c(@aI(OELOH|?BglnsLgF+{0%3{pO{7djyAC)wKcL}2 zeke8C1BMRB`#?Yis3{wl3<6#N7YsgtDMk2eQei|CQl6Qel_hN~dk{!xTn^1j9Kaqj zc{P+=8WmCUXjBDw7ds44HjJSe(e8jbGd>YqX25hO0YVoKrdF~jpq1C4B*IOFqLQcX zLE`-vNEoodnQRc?z@Xbwq^vPHc%*ZwGq89>(MlVOoJk822&tgMQ^42K#vPMS3mwrU zTLQTd2yz4@@p#&WOn{oSC|e_a6%`fDDT*OOj4J~2H$>SV(p`EUK*&H;S|H!Rt1#fk zh<}n+7xm&IuQ?cB2$-pmU=fY+v;EBj4EC&r{q5NnlGoFyj2 z9;^_+1|wOPP%Vr6g{&yj-*#ApK}FI@!LyOS?ExlHyiD~ure#D3*m6=)#012og=lIs zHhm2SG?m;b2qd`lK~2MG@g&=;(S(Vmj9cqSIkXn~*o-lYZ*20Ib%!LAOr!{~-AIDn z*j27u#s3uFhOWKu5`{d6YfjloNI+Ek9MwSq4k=s^ zP>LdDdHRi6r^HJOk`DibtBSmMe84J$tIN>W4dd>Pai>~kFq6ClN#(*M`*5PI5mMl2 z0xZS=8)y}{2}|CqsWa`v!KebOGa7i4p@4KM!2<~u&=2UMlmtGB_fJwb&eOm(3?8oJ ziR$Br^RxhNh`gv1iTZj32}iw&!1xc~!msZY1Oqe>At?I_vhk8qCv3u$`7v(tOb$pr zkq|iYc06Ta6M$$G1XZ;dcPkH%(K8jPfY!Q0GVH;E7_;DEz(25|`NAYSzo0-grAtBF zF;{mW|3LbNWCV!(4b!2-f|N!6oH&d?Dha&B!v#nJg$yP9CVxW9r*Q)bB3K6!CX6KB zn=<173m-@rv&!)VcST@{g;0ivaYo112R4AQ--Pf$5VU;^1%&X?ZbSYa!>}RWR2_?? z*dk%yk?MkQI4}i4UjX|n*ni~7I3az^+zle2M<;fO7Zi2DvWTwsp`pi0IgkQ>QJ4Uf zqmr8QSm`oZI!~E4U`7zVXaS~Yf-V7zH-xHD#4!@$XkprNX_Z2+2>;&;|6fS*$JFkO z3{E(aFCXp6(PSY1!ojhkI8)Ap(06>ND|cb&}^1#U??_Nev~le8)XxOfpUEc zMQBn;LIn^{08|1DUoe5LKG1H73J7rthY3`Gq7am!55^a?adH4C1Rj?QHfR*$<4T_Z z*EhLCNbcfXF(8UI2Ct9x_oC`|TG@JdcxX;lfU&;CL>O$Gf6{Eow06 z1Mpa9A~p#k$PHM?(FXAG!Gei}Mj?LbtX;wAj-8aGhJdN+j)Ge(tyTu2SD+^A0dQwj z0=FMe#IkgJ<3Xmfs@kO4Fc1eZY!kFUzyy+StK+&5g1Hg^1A~XbV`x$cL6F!~$WAn| zm&S(t80}3O{Qv>MiR6I|1rmP{*3?kXfOcu2et}a!AQ64_p{slkED#C+^%C zE439U1cE5AAkPd62=Z=_bfcL#fvCakkS7o0`-~c9RLlUXnhq2nwC&{rpmL$c)LmYJeydQcC{+XN)Mc{of1|EjFAoRXhRn=@#UPvcnLdK84o!kU_wGZ#oc!>71er zJjxX30fkHx))EYPhcX(=m{3S09W#(yM$tT#Mj?s17_zO7DKt`}RT=48K!$^c2ebi+ zj}$z7pb+s;c6Rq+{>J|kMa4{4LOo1wbx|Mz?#6|%;@EgzUVf_?5@F-db6g(=F zeciz=zzyUh2L6<8n~(>en&GK$S~IvQQG)_?I|%GMp-!4W6Z*u*&l8xBdm`^hDdYn2 zwwD63&)Wi?d`u6536J#G_!AVS8(7-WN?r>hi|hGdgE9JkI1d;;lTr&rvYK!h!&JcC zQJyVQfkVqCI9}sQ0TG9Mh%G}YkIMqZQO#0VJPwc7hYHI;>H&aRemEaeXf^7q7|a7Q z9x$IqC8}T+I>T2gR1}hcXyp-Ow}HJLXEouP0m>iE1{*Um>L&GNiUI>;z$qhFWa6EL zngn7w2#m3xUWn@i4xPFjVEpPi{<=L50`wE)^yz~_IP*yHE~R!6BKtEIhcpdc%$+K24$3Ja}=1;Uc|3f3ER!AA3!v z(2mD=j*|_HW&yP69{Hp!G(T8lR{^Aah^j%eFY= z&sI8#L*E;aD{~ST*AfM9sephSv8EXh2j4}KvH&R6T^*1@tb-)q7NusxCJMks$&~{Q z9~6jBS|Gwxica7pDfs@7=bRbkTB2}xgD7zu$fV$r)JZ8J$!h_RL)61~`Fmqf;KCk^ zAqOu35@9c*u$L0ggc=&I5)v@V1pGY-7!KIf1Y!uH(9lImjdn*<%WA+Y1yQ6b()*F| zK`MMOIG9hnD6Ai8ZfS{vKZ$3(y&r8@@>KVP+k}=V`4cH!MVT50=;lD6Xev$`s=81l zWf^(hBWaY^TaBa$2pSP%sH(~Sf+ zEozw+4M_+UHUC~RMI=I_3QFGpx-QDbHp({sUx~>6y=O*ga|MjazrFgRIe*D>6zNq& zt2jE`Gu9eUIr+xSqyO34FZ!B9mKz~Lq3Ol|u-ia{2=5-#K@tz=X`$d=M%lIueF$_R zz)P1ExUx?M3?)UA?ny9n+@?f&-a;}`MqXo4ZD`c_P#Q5N8_~tn6kQ-W_9SZ!^0rG$ zbx%nHoFA3mo|ev@bPNLprwUYvP|4$I!NbMjj>E%$NxTLM7;*sB(RdDjGGr)IK<=MF zZyp*3Pc&+Q4xM=Je!k#jvTy}{i^lZcRLIH-Ucgy}H;O3kKzw`55Yyik8wmAM<1cHI z)rxh;xuSer-7pibW3(`6TTeh92lVhXf*lGJfTVd9M*wJEDv=yo^pGnh_ZyAYgQwmo zV15H6pwOHHHRfp`IY}RAK7sfTXw^r-5EJ6jPw0b1A51(?KgfcR5n-wwIuL05N7LYw zKXSe)p5$hPoL~+F)llb8{l0?CUCIbppQe1kBqXvC`tbxFVp z2*xT9+#({{_B61B7%oGQVFhE5>HZwLE$T_G!0gpuWJ+3ly;9+2F zJ}}9k`UHF%`I~6(NcqA;7h$7caB~~|G7Lrg1VLkMb8u}LQwCVC;BG*`jvYz@5bKjL z6@myJO29+{7-&PwmNBUsC~#E*iO}hdE%0b(inAl#^l1P=0GilSU>YKtcN&;5iRAW& zswztQB&2`^jvXBv*fI^MguuGfB-|B^#`qJ)&2l0V(dHIb2W%|#?DS1d$F&W4n2e+W zHx(#w8Uod+1rQiH!qqo^F@pmNG-DJHuqO`4H*f+&7Xq-!#Z6oEsEIqNGK3xqOZbm0 z+$in;o8rip@gy3ly5fIF>;Jac|5KWYg+o;WsY?(aWqN^uu3^C79vG`)0;sdVTn0jh z33+Jn6ayFJ4<(kw?H10rv5C;J38@NF^28LPpA8BeW&43;L3!weTFE{!DE=@=HH=Fi z*MosX!K4cr%7MHZ#uo#$m_hz<;!#K%E9F6`q2v&221^PwpVqlZ;+Tkr49beoITT{} zCX3o&Web!opkIM%Vy3vj42y!h#s|vrNuIChD+nGlG554w7EN_y-t@cF`jr3b!woynw| zk|aDJ>7NP%kC$fF6re(w`FJ2NA<1{QnRFb22tDswgR=oqS4@Q+80Js#gRU6CG@^L< z)N@BXQckl}CqsoqL-y@oswLtQtr>|fSs@GhuO%t3_+OO-8PI>FLgOD$!O;C@(!moO zA8>gEEOHPc-v0+O#uq_8-Dcbs9zMWDnZ6g|layX$A}b61oVvQ}*;xX<;a@Rdhof1G-~;esa-Bu)%?dVud?{{udv&YWQh9>5a=+*oIb7K9jG(ZfR`!fP2hM0^dL^upWI zw0u4_j_q`8_;_$|p`5Zg)bKN;X1YP5E`0hSp@#Pd+ab=~8WZRX-n`AA5kUAy9|nEP z7aYrq3jtoUv3Q_6nK`DFwLRIe=U>omf`GmVF`A{j zYN&Jp9(D++CY+_7}*q=?m;FO03W>nLywW0I^)Sju8Tevh=3n!nlah%QX^Og;liNO9=_(c25-YBpH$R% z#1jUwkb%-sV50-ptM^pf9yR{dqcskeSTuOc3?FgO*REiw#)pl9LK!-ch*nB6azNUP z7#qm~{Uu^^G;MWcOf=R#@Ly38<0x6=1f2n`Kp6;ZssM8UEv$eaGUNpf{q#=6M!fyc zxPTx$#u`#E1Iou(biMBxTyou(mNnFhfje?nbh4p|eBiT&Oa|bTBr{L=*>4VV{%;6i} z$<*i5wTF29Lnr|e%BESIG=>7o=>MTVf!-P1k;eGNrEks=?c~5r1#`~AHFW%7&^rv2 zVvy-aGNhd$1;9K2c$uW*LebYU08d)r^uZHNTuq=LY9?(Wd!t3R@2Q0VW4CZ38laJW6!3)98hLL<&k!L- zleT_-1k!9Ey6Kw{Ch^91_;(V4HJM62LA)z!hp`9qd+1c z*#YpB4IV8QnPbx}1PD$aAb3fLA|4Uo0tX^5P9c^iHV`6U@oTW|!2kHvB!dlT8j!7i zAZRRnLN{G|MwiTV2RTBdFR>}~3>XP?Iszywgs%Dl8#dkellLcJSTNZHo zoch2ZoahC9vN0k$@RgB+uPxP}oCHdNkN_Q%585aW>3KKF+XA>JZ>Ykr*xlDs;ha3%C!W!vQ$INb_a-c}!O^Q;KrQlVK{qwRlT{9g? z)^GyS0p|*)CixBS)VuPyEr|ZzYZAoim-2W-`W7)QEPCr2yYOe)Kw7x}RPRihNmVI* z8x^>&k|YsuTbgbjjY8A8X8;5eqy-i~jgpNxrAIMs+;nrgUdB7k=YIN)0nSguSm zWC#`l1v?$y6p^jdU^4B)B=B~VnI|Lg;ITEdGC{@}xDY1BpPIC%Xh1{X3bdE=WaUpg zid4102{^^gdtxOJVJcz8L=0fA0X;B6gA{u|eP~-qM}u&1Ol0KX&_Mr5vZ#O{;9VBT zU?$lXsH!C)V{!gL1X92SK?%J6n17Dk-L6e9xY0-=GK#vt`(ogI- zM7n1Hsh`H7BcnT$%Dw-YY--Gz(@l*(L&5_wz+v(YfA8JvY7xJRUGS z|EUNNd;;+47?T-JopFS~m`sxzYtWi{JQZ<7MDNP^PodDHf5P|rXfs|^D0l_(dk24AwNI>xP2&CDoZvQ2v5qFzFAOH(O(vkY1H%EOVTMTf}0__0w zjfmU;u?WLZps51Sm7w)2cw+?~d_al9cqr&EiNx`MCwgPTNg@>^A+X1#Bi#Yfh>IHJ zTKj>(pkc&X?d;4MTtJ$AU=W@_OoFgu3I<0gIo~kY)d#rf!9WP4Rijz3eWS50DV<8j zq|PRd1UVDf&=UPI01833xVT`5@k?AVq3{3$=B+D$gaugByC5+Pz@-K3$!ISSX9@g3 z7X>b!FtiqYrUyP@&muoS+y(vu5m|u9iNOW<4p(LXC1TLQ0Kfx{BY^9&2XH9r3M8@c z&JMny1rmdcrz_S44?NEii2y~9LVyo+2&fhO`eWc|!8+jw^l=G(1Ci(!$e(yX7bPEl z0HGd(3lG0)@etweX3L4<2!K3LH*&xq#sFem+3m zjf87va6!^wb73q@7b5AtK#B*V1%gc2AK(X=FM$6B;gDYt)B&_0|h5w^rV#1%^bz6U{AUm$dM!2#N-U|{x`BxFjDetToWT+rjwkX|Jh0=+GS zj-!Z}M(i$XxF8&LDjwqlf^(2zK%W_AD2ZKWD7%V{G*^?N;3>!t67&uArQ~${CBIh%@qJYkTp^@zgBTLrKkBaHP zb7b7=Fsr@>C+&>0ii~0wkC9IXk>x$!l|{IAfx)fi(^;YHHTkoD zX7?W5BU5(rqR8Bhf{BUw;cVz+^_i6){t5P7d}#7l)m)uRl8Ivfm*}&ici!wciIHm8;e)`?{v_o1<9m7bW)wjjqVbSy}|U z9do}us$I9Qf0rO{%8tzKNoKJ}ShBA%B{zob^t^myQL^K}sm7O4;%5do>NQ4HRrEbw zEG!p?*A!$`*Y4!q+MK&z#%FEq5L@No?qi3(t;Ub^U3Aj)dNyqLqaj-Gk)bcH`^?(y z5?^=^{JCAfM$Po0#JUF}TmrvZF5KGlt1U86Eu*Z*;AHtrn|@vvzrsCyX+=q;8ttlA zxY0$Sr_%TD(Grbj4><68dsKE}|D1Ujee&MCY?481#X6|#f3qv9jroBJkt`6>pA{vg zne+C5`m^?!-KH)&n;t1-CH~$XQfZ0Bu$23~3jPyb&K=9KT>FXb#aNqlT`k?q;)O(_ zzlz;i;qLwXXU042BgYSw@jjD?*d3tMOAu2DiPEhwv2-TM2 z9h;K)KrF4(SW!tY`9!Pc>i3VOhS3+K3K>~F*uF;%zim5y`_<9Dv_s1C2K`^(8i<#l zwVg`>ryTM0)~>_;&7bA`2W;l*DrY3PW-DC{_DhW@f1X3RU zazvvoq?64}x7H-hBxknyhvAePAC;<^!Y!5FGX*oA>RFj0>#}~U-QR+wcL$8m-S5aQ zWMfzuu{Of$cCFA+^y?Sv_q{2|i|dnP`O0W6|MHw5gY@zEjs8Zw2g05Xl$71pzLEXu z`<@k-nOMKf`%){|eB-r#Q}xb#R6;`hrZAp+b2e$QbF}_QZjj*$J2m(5yuExQ979KW zMP7_7!p1ZP@aG7GwLNqWDM=BkezO6^XkRKa|J#X!1shW<2VVZ&z5B%x0TJxa*55bO z?3LS_=g-pnCA8$w?ek9VS4GyBcT{^SKKUrde#Yo-$nBlaIHHo|u9qD0!k@kTtSo(G zLsQZ0toXCseFyywY|PG9tDJcKBK*<(y|wo^I{5ZvzO!L0DA>Be&5>{a!A-KNJsqo7 zg`H_c-EZM5+IsjJLw4lCswzz8wtBqWUrtA+-lI$#ZNC~n$@{fw)wSM@nR{(p`F;zq zJl_c21Vq73;+3@?xEy45|48>({fXvk$do45lhR zNOZC7d-DEA4`TPYC(+p(MWUu-L++pCe)Bu|HeUwL1!}H!jWS{IGi` zhhm(o-SY**7cw(g=dQ(e2&xcVmn)uoc!gQZ_K-sYckr* zD{gc4sRMbAk!J_KM+>k0bput1DUl!Qk>Rp;jrh2F&xb>MzGnL_F4S_tb$OZ=;d+ag z{_sH^{)NBX%0BXQwP~)TQu|9u)w1?$L%VdKALju3 zuej26d-V@I&bqBySCpQ*cR*N)Q~CMo)x2)QgeoVV#cNmncCLQlX?>YDD)m7zt0MP? zj<79{T-S9S&lq92cD`cb!X3;#++0lI2z-)43xL#5Ne+2w2^oZj^fA zgxSTW3qEXv=O6O=UQrU-B4yCA{*AYmzO=uUQ0A*U!dGVh%=_uA5-%d*SFh8pd-dL3 zr69|wx|;9%8!H%%GqwiWzz_s>7JBR+A@t1UmOtfIPFcj~?ty1!~Jr`;nw_ZeqH zlnv|Ywg-`)wyv{0C3Vc}UEObiCEFh5a`u1Owm9eZxvp=Q&+?sKb|M|KSkrM;|CuEn z+e*>=VRHPJ)a6bRycl_Jtw_Ew-;$eQ%e;lR%5tJ4_B~I!Anw%DIFe!)QSE+gf%W>R zWxIUk&U5b6J|n3jxy4NORLmu=9DRj4Q^^XY--1d3ClW$VFy{uWLT^2rzhL8n^O6Ql z*Ke*n!H4PZ_1LVCd8+P(lW9~byXOAcUB!GE$^7SH)`a>0_3C+Z^Ow_UW*4iDBEw>% zuF8%!-OX3)wCuNLVXT)dFco23&@O&<)nQe|Jz4H15h(=*0~;k1d^B${r>?dA6J#cF zkJl{4wk*)Fj_dHw0p6@85uDU^MxATzo~cVrJM-u6IDN59zGa~lk5l(^t-dVtg6bO) z9B%f1gJ)gJl?<=Or}Z2K>>7%1_Sv~AUvf6RDqp_sh@;rs=h<=_^Q{bXwHd>1_*AgH;a;ETGRv+xFVBBiwEDKl)B5(E zk;~7fC*cCt*u5?ri7fn7+4r|WaJ zI=o#YPZPRPj{6*w^Gnrx2bPKO8yQE<=2hqMo$G0rY^j&axy{Jy+H!%$vh2Do&mV4- zf5qBycb{~^V#m;x7KO>;hP$)xImy>re=CxD9}{ck(9+W7fAxFRW)JILo6YN( z`EJVbC?B)u^!aS?YjEF!=0A4cGWP4gEL!i)zh0v?@dsBXoVne@ z-;g^{E$5ZHr1iN)T;1zazcH*yefpu-uPO5%ZvGJ;tR^V9cB`%Srw@H4n*>U;*hEH} z7usgMeztFOa?w85H4S=KQ@aD!Zch|4FXv7#zFwsJ$--jEuAP1B^Y(b`IceK0H0!S# ztn(O&N*1T*wOkuXZE={Wma8oxA*0tIogi?e$m>%bX-jWr?o&`KPTu} zS?TrT&-$BoOW*9-d?3a-aQD-W%L{+c`lE)gE_Zyyx5Fzf?HcZgU?zk0X1&F|L7lk2 zr+$8KNqT;>*SteoOK6{fA9KX3@8=EHvUk-ntUTU)%j`YF!C!U9cDn{IaLYI{{KYm* z$VhHLl|_GPoLB25iCuFpoYgja>$w@%$a!WYZBF&eZH3H)rR#?z3L|C-Xj`d`uxnQ~GjC5+S4WWdtp$6f4I1{aA6p!m zQSu{3%l9}>_L41ZEUj}F+&%oc9_P1le$Lr8rm`;!&?|i-b)MGrYsF@6Gk2AY7GAzo z@8p)=rD><$UDxg2C~n_<5#tej0$Uf=?3z+xtNB4L^x1+KY-DbWN7D~CWuG}t%-#E! z4~0s&Mb#^vH~Mxq;@AnN#P44URVBll?!D_+U^}nMY_{=Bt(p}FIf`=Mlt&5HHNTlt zn@o^Rc%eS)v^u}NV)%j|PyU1+^TrnQyLJCIK?S`pUwhTT{ZJU6c~`{_TNWMbCkz}_ z?`+OL)ZPDa9gF&{V^??D_t?(cxM8KH@TLmMPSM1N-HEH0Tl2fT*6moYzjAH(U`bGB z#^TeBFP!7E#AiDsIv(|?;=?iDKH$V8Se;kICbMu(bGf!NXWxp4Uv3`JMw<(cuqgI% zbs6Lgn(~->N*R8!e&X`EmBm0T>wdH_zUmgMf`lsD7oY1+ITn_EZ*s2Izh!0)eI~GR zZOQ2{<7=xkEasw4yKK*WI&0Qp{Q99oYrNh+cZ{sd`H}iyqmWy{-{Q?p)d{LyO&*$} z$s)^QOX;ACG=`uD@`{LHz&;xy4&p*5UOA z?YDj6*66cXTCneo_Qldovv3L7#=FCpeM*mdpL4}-dlt50iR9@=9(S2c%`B>Rmuh$> zZn!RR36P%c`ZC=c=%sss3 z%(|oNf2E5%x?xdWk0L}qr*-%k>C~UuYvCVc>Ly+p>3zE5?0i1vB*8T!cRysO-1D8>_zRt&xUyz1IiB_oH}B>*Ts9kzT``T34Sc1h{WN#bgn^0sR*SNR_y&XA#Ille z_0=2#o4@hK{{4F@`dWT_%(l}4Ww(~7?p-YEpq5pCK{^Z7@NOevXr-F{fj6e#df%=4 z%=k3KO~+sF04{o8d)4z3UMAA6pJcx7P}VtN_hpYy#UtUk?G{H}YL#C+y(8iG=FAYQ z633;8jVZRTlAf=+g;)NZ(;p`Eyvxs{7-x3rNY=KSvo1IIguGjxUK4y>%j3g&Tjf1h zJ5ZNQa)-S-(lUbtMYF?ypUuHwW1nVsX%4mZh6rk~PZGHxoV<$3P!lVo*w4QFQpW>X z3Gd7NEZ&CM13brW6!Fggv-jc2eBEynJUVfYlP@bYd5V;OZ8Mi&Y?$cc82jTQf4V?- z$w$WY)<-v$8*dVfT+JG-oI{3FW?jnYNx{0`GPnQEDj%5ks5|HTuPptzgM}sMu}oZR zdh<3$w=l|KRKIR?T(UPku;p)Q?(E$xe}b-7{T+UgzHI+dwDh37b4!aenl{E-+V3IO^)q-Q}yhG7r2)|N0Yj z3vacnU(59JSCpoeT!Bnv`BRpM4;i?-*~B7J&juLUZ|AIC_Sv2B_{I1fhAtz%{NUd2 zt%9OC8W!j8hOPdwY`1xm9^Xeso}H35j_M!1@_c=+?+^Pf`QWu?MMYPZ+0(?5J%QU3 z`Q7-B2r^aejn{1q%(;FbvyA_(!Oc>ibzgrk5_~<-rl#x1p3yBF8-DCS)}>W8^Ot=2 zd~BAB8vhfoE|xWS=XJ)OoGsaN(eiFyB)j}?2C3Qp)*ff}f6119Z2RX6mb;tx_K~~C zA6@?y;KqA9zpuIZi`)(2uI^v`cb9L_e!rb!n z9`>fq>YDGMXu8@LpZQ6z(iVBvbMF=~@Y)vSWpctTvzotaIFoA&N2zi>=9^S0=Aw7D z^Gf-+rK0Rj0UNlx7OY~@`;;T>{Mg1~TmSo)j=r<`Wkj@p?|Pf(6;@ZhHZXzJ=6CFd z*#WKfGV2fGjk`FxvI9&b9)}CwxglSBK;d;>pGu%m^d5WlKl0s%Ew_pV)AVj0zo>Y1 z|FR&9ZM~@mIm+zKT`H{y_gE8#f5#u%RP?;Ox6g+M)&HWrg>a|)Qd8UT4HIsOPYgTv zavWkB-DF$4!|BJJ(CWF4l>t@*{rMT4sCSRoney{KyLD7rB#lA(=nKd2 z&wklgs>*eEmdl071ivrfl1^PHh4w5fL3<}&O%^KHw54)$+tJrO2`!EftBxz#cD|{u zDat)^k3Z4IHDfV{prso7$rDZc7UURb$s}XcJAWEFv$tQ{ec3R3@5iOB7HfL?S)=(b zb=rjJZb|hS<`(aF;VqPFlPQ>n>MeOUM~|RpV8CJx7OO(H1@?RP(Q@$HNPb z$sg*O^OEmf=&B1%-ok9{!%k;Q@z-sP-uoC=ite~{<&$Ub zDZh-)XP@#9ORLXHewK3;f27sxXPdZ2&5sqPXE8g!Xf09Ed*#tnAAgIT6=y2cn3BJw z(fa9UaDZDoRvvGa=J%%3!7oWY&X>)y$xa~(P{@BCD+pNzV{Ohx9FINR$ zuJi9azVB+xO`k*JwW{xL%+D&j^cq{b$4+8a&yp<TR<@So#;wAecN-tTZ)D)WKG$NrxO$}2 z*2aM_Y<~aYA=7t?hmsdH@T7S4RY}X&IkD^eGRD|e&QfCb!6ar_ofm8j&0KtEXWX%! zWp6T_=Y1RC(Y??<_%?p)Su?da!}ctv<<3cDYMx-Y^OF(xphU^Sc5M)+d{rQGpOo5{ zeahNm3e8xD)$$j(*w!#8YJZaqSo3AM|4?}3l9h`bPp^HJ`;_AmS48xwf&#M}pDe64 zsx4te?-gj(P>rv1{#q62sZo5zx;3l1t)fe(!rA1IPk%)J#sq&a%T=#^E#=T|rz;oR zb<`*X`AL`ueJebECXb2zbnyHadA6dtG3P|H7w2*B=ek>f_Z;fqtbSfp2o+>>S>v8a zfr>a*Ry0A&;Z+MCbA!<^f3R%L$_FjW3jE%%bhWm%ehNPIM_$%)i(U-c05w=|%o>@< zGSH81^bI5Iy~Q<*8~nILuji3`$}!Dr-Kd|pZuUlIu3r>rd%69+?PaO{r|0*@+RlHc z=!Sa8{@?&6Z}`SYsVe`gjC(>G1TWUuVePZc{Bt@VEH_#C;-Ry^Tt6Wnzh_d3-I#{j zY!R+2j~GJh7uw&^A9kpy)mzSZz=pe5{DBo?A@jWB1a_v8bDU*%#+OpN9!96-pD(L@Rz%Q#Kc|$C*}^b;$61C&70K(WtID5rR0sP1eR$%-caNF??FPxM zhjP5o0an-0M+_FKGaWyPFG@*?K2XxMJqRn6xpvF0vnwoq`QK8?)7Pv#5Z`;6b@MN` zTdIA(4jwpl&U)pCYpXjjVTTTW=v@8s=K4OZpr73Cjt5`X>?uaINiH3Z_sPy~j57A% zxwXL4TO#iM_DxY=T7J6Q#Sgv0z5Kwxwue1uNv7??tD3C029I5O=gxEe!4J{+M-??v zFBtrfIlNA*7;M&L@WO7s@2Tw_zWYS?pO5KpZZpL!O%v|C(<$F|FS7k+k+Yb@w_lEH z&Oh#8elT|*(+2gb=h?y`M(K&oA!*Ng6I{0k&u^0b5zO|!)$U=n=}6KG1y#AY%j?{? zsoH#x@KIDfkNzsX?fSsy?VlXqmd`8wro47=zA1`R;@#U6ZJAru?FomqZFBh~G}Kl0 z=Wrid`S;D^^h|}~RUF%nnteVNnaq-FqLp+-zU0rGc@KI9`|$ai#l5_nnT~keX=}a_ zHBx=?!HK)4*IK&gG_CskrQ#y?mY1#eGv3@@Y)+pP`a$r)5=#%ec^Xa0mV5Kc&sXv+ znAJ6?xrf=z+fI&KpgI9BIkMnBK|*1%5hl#kMQ+g!<%PoUMZ0_t8t7_XU^}y0N0p~6 zi!Tpf8E$e{4Q#I7^O~E__Y*=G=U?MbOPJjgBDA_T{@eUpI=LCQM2<2aeuBMpA*t^A zTPZPd)!$1Nsa(EdeljnhD!L~+!tvyym*=YTQ~2eL+nENVn@xV zi*mCZ&a}Tuyt(jM$lKGAo8MoJYwU9?=#@UM%*Jop7qcvNN1V9q!qaUi>zJfFhM#=5 zw62d2iB_8*ErY@CKDb3^)3wg`?RS5KZsNGUNG(QQ>7x-oUi!qlCmZ{=>l}=3W7*S- z(YcW~q$}u$H9K9{3yw_F(@%f?H zDr@7S6Z4#NZSsQV$DB(uudzH5cB}5q>+me~Kbu^wVv5~9-xuD=iRH?4`z&j->6P4* zugq`P7oR#P8yNCvbMR_ExBcJUK5m|E*ugmDeq%|b>6^fA+lsc6Vk)04zF@t6>{~YL zZE|1C?vd?1evNOG`?YwCUdLMBzq@b2rmOmfDw2_hqgG?9b_uRwy)G%GU|d>#(TwNt znO`5IcImEt^1kKGpS459#XDZT3aJk@Wa?~g*T|S{^Rffqd(}QN_wBiQ^PIRF33(kd ztIj@c`Vla^wcj{)b#n1#=X)Do9c#2-U=>;7&%W?wLFj|unhTYGdi}1bdA{|N|NLVv z*R~(;J5W{G5G!7O=t_5f>F{rH4m&*=m^IJ2h+_Da^IdkKltgF7vi(9=t zH-r4{%fwYRn%!wC{NtW4CDFxxhLD}&^V78L)n&uS3@5ZV7T;@3UfgzDwB}A&zi5G{ z&BCjTk2eX&52WwyPm1|fe>!PPoJqc#y-;GqB{TQ<{JamW)f=rvA98)R$Vhk{v!09L z+aVrz*X5jlnwQ5cM0=+<9k!PWZE?|2x5bS-9WHjV*`dGc+EYWN?c1Md2=7{aXs_)9 zSBF&en=S7BiV=UfJA_ZStmzt1;}J2>E!{2IDqyEx68>9!uWgFLL-T+{ol~Nv>z6Vp z?)WisnSal%vFq7c0y)4E1 z)lG4^?>l0hwN_v{mtB4)!BrbCwrBWfU=AG4bJ*g3J@L zYukNK>QsfeI$DX!tauf;ZRKkp?ubvb#sYU$^X$6 zin#aK^2y`keFIkwzeoqD2IiFZp7^@*?AQBmWboLg{r5~TZ(=>ZUqK~3ROnoO>Dv6) z37^x{(c(^Wxrukln?yU`HDNjB*P3gOXU5vpmq-Q`XJ$7P>X`O5e$NoeSURZB zvEFK~z~f5=4UHOB_OCv6*Q;2*c#2UeWcR2#C;Z`+%&VuFWtF>_&h;-^(yP*R)ifvZ z&ysJ4>K_+17_TwawKLHlary0GA3nq8l2d4n#EhBB9$t?7*Q$#)Kc6X|CS={q;gq|3bWdNWiu{qbI+0f^RU_(N zp4Ipn?{(ekLdW&{qRj~jzLq)$mN8d0rY3FMwkH41vqFoS2kK$YY^MdkoU7kaKd0^U zpMvU)Z7~_$>b$7Bk{uR59zHz8va+at=a=46rplWBwB@spyxUgKel{a4-6lLNRHL2w z-9Fc)7n@3W?w;C~z!;*)cx{#OGO>>lM_Hs@9QE zLuF|Qp`aw9@_y2ed7}?^Vs>xuZAJD1Xt}}zgS`e`@W=q7GV8cVdJS~ zN;v7Ndq(7ifL~A94%t!#rdpAv>~xv;{#iz|T6k{X^4M&8-><7AZ(o1XCk}D`vzM3& z-3fR0cJXkc_)^+?%gxI~Vq3B+zNi(zjUPo5T0~+B@ zsZp*8yVUY!>sYkX-pyaKd3J&7pg>IXCCgJ^vMc$5^XEwZ*l{4(1U2yYu|W8Z%PV3S z@sI8MIztcGqSB+g`2D}1u8HXxQT^dj$BQz>DZ3UHb_{Je^GNWN#@o62TsGdHI!yEl z*;es7cJ-U9Q`=(BT~txxbbr4oU`|MR_SFqKg}Hwlmq>52W||XNmVUG}nR$`Z&6OzI z@I22mUD2z!+B^T;u05ImhO>I;kYHrMjr+=9J^qWaWeuH%`ZvPH&`w{34KoZw=Q9nn?#!=PWLs(T@{+}@1uiOQWo z8RyV?^pNDRFw4*~-?oQxC-%=pQ((VEt3%{77e=SczJiHXciH6?Jflq47tWJ# zn14qzI&;%obR>50j_=dLEc1>N{)aiVWSY&7%`$y_iKDb@q@G^+AAgt$t;>CS5hQ@@KW${{Fo>S>>A;H_scW%iVB|*~PM^>E(84{{5_Ms|L0i+h*Yv_Xe87ae7~Lfuy^cH@S1 z=ZtKI6y^qsv&`bX_fjROUH*F6`$4@+Ki2$d)T-<0-+wWAf4F1D)1OXwdvf5y9Mhv+r9;YJcWQ}0-NNpc6BZbd zcC+(a!9w$--Y*p?u=(z6o&^{EUpk2$%h>jHXV{aQzu9&aF6~h`;eZyyn)nmm zhHEhO%@bv4SZv5@EaMiZyT_teg;!+Nj`Tf$-I-LqGY9mKII|Q93=DN0V;susKhy7w z!Vbu9Q>>Na>3a6#SpHzqvs0oj0shR9p5GL1_Bo^sl}Kb=T&P&rAeebc;^U$vBI@%Q59bh0N#}Ol{(DtS0L@$>tC@NKl;{$<oQ<6>Gk zbK|p5dH>XpX#R8ezKoZ*|02WN2dky$fBY-<`s;I@!Cc++wUK|ESy!4RD?Yt2$zghS!7C^{r)ci|*K$`ip6al)^moR5N>g=_ zZVr<#tVz{5bx*gvxnRFCi(=G#?-ubB+sq9H^D2ZM868JEJ<`xLZnX^L{H^}WV5t#f zMM$rg$ThXD=i-BhH?70}*zde8NUXt{gSLaWp7S^m}`r_m(T~wlp|6 zH=Nnj$?>pPjBvK#1jjSkQt|bAPbKxFn%MNxc%w!XbmKzbOB8x={yc2C!{15uaVoQg zb%KO5u_o0y6ecpr23PG)NZOc7%?_bc3YLYzHs5F1O+>y!? zdr+HwoXoa~XnJ8{uLkC|zFcwoZ;4n$J6bM)A#!hYBDZUF`SuO^mxUL+x$u|O>%!7k z3?e4@O$9f#Sr@v#JpGJ+cjj?~yKCUx3nBbnVQmbiU-5cV4_m`TpDAGULY??^yfYRskLvO2e4N3oTz@VbiK`}552Eia^E7qwj)4zQ*6nu z@Zhp{F}BaGZ}xp@T^_8;;pKS$ZCV2MX2esedi)jVTkbc~xy>Rx%Tn(3)!s3>xsSck z>Atb%aB#+#0`ucVZ+{mJzUnqh-J!a(VU5jcl)LM@vvoQ#W<&yHl;nYYkV~F*M?u{8q#0X%v(NEW7o~#!Yj&UBUr9g^ZZnErDlPmitH_Z)z|A54*m^K z*!rpMh;8$n?gz@Vp040kVN)I0=lUZg-HkhJK~~wtD;w;$T8SOrvD~?$J>?yW>01GZ zdq;Pl_YSpm%s^oJCew@Dk+Tg-yRf*Dm0D(9L#r=oHE8#Wv?ndqRx+rO|058>RUf!T zts!44(pfsMFU>?osPvP2uiNe6HnXaG=eEy&XpGU>Jtybg<9pKGY5D7(dCQ&-x6#T$ zvC7>%uJ`iup?ndZ_VS$$*Asr-`K$lLa!2AK-&d#N=2t2mtUmdvqX7Ncz3{Nm-g)17 zT{!M+-Jq9x+vRanaA%c{6srr@i@P!HeWn-p3l$luyJ;Vl-u~6ppk&Kz?v+WQ5ss-z zHu^Q$nT{Lv ze`|CC|K6(Mxm4P`*m*UsB(W~`#HobU79EZ-}O z+e^aibE7@YcF%EH_wvBQ%Y<8sbA;nQ`^bDTIIHfN0M#S(lJF_+D{xxlUdXkwuX=Y#I~io@rOzUiOgp zi+1}nmX@Oz&$Z2Z@3(%}v+`mO-i9Q${y|eM(Y_gAq!CdkF;~x7DYi~=&^0vwr$*F+qP}nwr$(C zZQEAAf9P=yE0fCFNhP9mqC}~Ohh8{h^|{{KmgqqVx*8nx$tRezdtU`9A;5@HNFrSK zdKQHh!;~M)a?gIp9)nhoIn2?9cD|kFq=<;#=MvwPp;^SB1_7Zeu*U`o(9Bix@s=91 z6Q0@CnoZUtbltvrYM+F-wOsh>#Ae&9;&a_t;%73B|09X7FYCE7w>EB9U1Z z5uol>$SQg>Xx>Z9DH@U|j+U^uTC*1#Gb8`Lh4@3ZyXcu6_hxNs42WgXX%~v&d%z#dr9C4mz;991vD(@8P#mHv4?Wv}_DB z(dinW5sS=tET>F02su+aGo2yEXHd>MFzj>UpKs354 z=C;V8Vqzk*JpzEWV71KfpTd8W@@$L=zLiWK#YGxN=PM8W%A257^9NMR2C5$s;=eybS9C)ee&f)2`==p4xYo5{cX8>4WyvGyMN~X{RQNO@lb`JUCct3q7-O`<^2lb zr(wtu>50f-+r@#*Y|@x=`Fi7YV8(ReA|wE*(ZxRcMJOhIPC1Sg^)**^x`#wE^FQlH ztZmt%9lphCMldp20_FN4t+~WjLf*QB&Q|Vn&kv8F$7Xj$srBZPkQ0)kYM5Un()N&= znEbl&vv;2&Ee8M&VCR}8Owz8(w1a_Hd`8xXbW2Z{3mp`LX0SJAOTNQ4w@SJY*G<;^ z)@v-d4}dqFB5J9ajBQAH&IxzFlekZlwJD~7A)=rpC3}n$b(q#ol#BuEY76txGMx%j zz}C9x#fHoYR|fNO*DO@b7t<}MRk{1jn^gB6pGKX?L(cML)#9RLOUm>VLOiZ5Csnbq z&g^N|E=PYyt+6|gyz{fodHfC z*w?o;LU#Kq{F74VE>RTd^%H4NpUuwhO8s}o7@ah_8?QlE`VRgke+zmwEP!-ZkRj4{ zSy4IobU0KkQZ1*UZ*ftw@)+kZ^Q!KF5w~0XXx`Z{HpLHJx@1|KkSE%nZ{5abXS0^# zv{#MP8{-ZKacFsKw*U>#l&8k_L}_MmXc9Y#O--7HkPcs}X=4%gNJixrvZRVXjNcJE zKf;p2>@0}NJzAR4xu3DQ#ZIT!3H$p7bKcB?7Fp6oV&57?n6)Dsp^h)?eZ1of_iRXh z7S1&^8>rSVyCG@s&OXW;vmg=oYRVe<7K?6TdA|T>9Iy4r82FeyB{^l_;9@|Drp}f9 za%f``?<|miAr_8%I|=lVkE2{3oU2(9G``S$mFuPY!>}8uOlzlx`YPNaV#z!0+GJHi zW3^wbb%SXvaaI?NlC8HRLZya`U`edI;%8M|hy}E$r(y`%oQU67$`;uOf7UHB-k5lT z=e;ztSWM8_SD8jw2Q{ag`6kAJv#`9yoVuU%efEaDdS!01Mxz&H&`&->#042%%dWZ| z;v~^q_KVqnT-Lsf?|C)uw?+vA={ z;e@GXU7D0977)viOPPd+9;=#+_ zu6VG?3BlYEo&}-|p-QRilnqd^P-3U19^!-^n-Uwv1}me)$Me#lwGQ3)z(MOUSY5%B zowfg-OGDm9I{e6%9i*;ZI1u8JiuA7S(6m4;484CSARU{6mEVN!xZx|nn7UvU_zd9w zXQ-h7r*Ld8GV$gmuh8oo>rD*`e|epJcDy{}Zt^=iD^r^-OXPz>^PPvwQty7|Dp98hAKqyqCYY7qR13tl^2Sl^J8J_$dI-xhJY1(&E$M>b#T*OGsAT0|g-^W|%t zo~NqSrh3=J#^Z_*kFFUf&3CR)2yNzH&WSoxSARFhpdDwvwf@=n2GL)mJ0u7lhub#9 z6Rc}gFO;g197pJglfMbq8C5fRybXIO>OuQ!qzP{IYQ?B-B4*a@~3P=Ak(I_-CnmJT$WYu5+pVCzbB2=lM}< z>P^*|z$~a-Q z*GfR*a2fvjCl7qGt>>(v)%A-L)|HC|7_r|~jKdP5v@&L{(RsdgBP-@+x)m%CiyzAcY8m$hmH2F6=hQ;B7!IEp9_7lS_q;SW6T9} z^p)6PmH9lXkG_k?3CC>{Zu1BCiAvIVCv6o?GjA%D&_7o0YA1et9lADLZV@40l55VL zGz$iN&@c0=8MyxHZe32k2dND*qTP6(^g`GxETV*iF<^7%65e+geP=|009U6=OBw)Z zY4;F&_E$c=?|2EX0g(x`q2tKz**4>F{crs0C~2}w_7bL242q&z^K9;~HE`W$lKrK5I5-cLjaFm%RW89qQp?lEds%AdRmHTkMKo08q8*=ggB%2!2I3mh}@N7E| zXis5qB+S?6Lmj_7hn@PO564ibdLZ6h4LGt@;W7RasQlbNnhpdOD^}q;DhC8cplr&5ktAj&PlNa z{n@MnXek7mjBa&g@UfW#%6`Grp*G-ZK}!7=arw3HVu)a-qNfE9zJ4upef@XSwS{(+^)D!K7p;jVF&>H@yFxfK#byU80AQ{wjAk2WUj4b1Jz}*WGJHwM5IuUH-D~am)%>6Ham{s}*kr$5s$^;D@0y|oil%m5 zG4DLytYU-E`NuiW^W`R)M2N`OH2~T3zTPiFYg@>sCug=eZ2nngT!hCv|NYoE=pLRQ z0^||wK#^^%n@pTl&av8W4*WxpCDj*rT;CZZqZtN~RTXCKfmBu5YuTjhs5{`8aMGG^WV=to}JF^@ey zB^!;kG7+S} z0YwbK$|Iiy)+BeNe6+C?U7MEV$d-6{TtfVKr>w$GJ~Lf#&KKTDQ`Zq%YWpcEI)Nzw zppXE(ig`}6TdbN3oS)^f3<|iKKU&N zgLFo7hyc>MQ&crNA%PwB4InSSt98Ky8onUkmkt^t(V^89L8i+kbW=OCM?Y$*4f)mi zE_2vqVo)2|Y5_2{^1?uutqW=`)GXH|#z$_u`)2{}(&A&R4`E+(rUf)p zOzPE+%6ZvH3QsqLs$fllEK5q;foXJm?yaX*`B+er?g7ExRcssZP0$^&s`PU_S~6p& z%FS!+%JBikL6#WI5JlNN?u}lkAI?Ln!DhOJn!+1{$yM%L@y6smRQF7DJPHIKX_GJk zUx2R;W$PF2FmipGpePBON@qIp1~d?`pl_%)$1Sexe#wo>Y)JL${8;}n72n@q{N!zn zFQERYONQ;3%I%lcwfc$Ep8o)@3Wy&&Ulku?$9+fuRYDVjBnLUiMN!)YF&#wC<1s6i zn4CF2-&;VJU*!2c7)Zp4WmGWPOBfbjjcY#f4nAavO#Ivp2^JS-j)An1jyAeUmnb?Q zF%nsmPuWG2IytmcnV-Vy?4R5<+u=Nmf%MHmDq9q#+cjthLo->cVs>yMX(oa>E1#g= zWz;)w|GZSo40CL+otl_I=uf+K!@TSTN-10&h6sF=D!$@OU%iF;3J2RCh-eMU{_|6$ zP)y8+P(O3;8UPd8eZw&6%Lu9JhF~^1!0p6WBRhY@G@X3wLPH~1S704n=DMmTaBZ&z zUjWv;b#vWuYQvGG;KP~~FV01cXo}>sgLPLNj`oR7L^6xjfhnmo0(=}S<-yX2?r)A) z0+OFvXsMBMi0rhc6e%h!0R)Ln6Ps&{4PFylq|`L_H5+XOSt-O%vx_l=}MIs zgr$(9Og+h7A}Cso^}pM(ll7TR(}t^omMy_$49n%N3E8IQ@~0x)uhavgl!>n&1`#L~ z5in~g?olao4syY5Zo}v|+I-qWbbuCG+XEnTDBSx`62UzmUxOa^$BIa&7T0mV`7-Fl<=gI40U^3(lrjBCv|DTVyF7 z+~2?oP#z%{0{JWaO@=D)zMI{hPLWx2GO|!;t2VXpCz|LX@PePS7(5=J7xR;IxYMga zdu>bdPH;!wd>(hWE+}RoQG)v8#_(cExr|g1hPA^Obl5+Ax@uq6%s4UXnytkIRs5lteqR5Jf8SdP8RU{Vn=w0a^6)zcK>E5 z0oYD&lar?v%9&TLE=NnODDj2y3G-6%qT%C8+N&D(-h^^aHO*+_`W%~ng$iEJD-=_2@#E%4W}i>TBx}qzp~X?JXHb4NenE^Jxr8oC<0*etr^%?wdi=g zqreN_FkyBCq@$kCEEI81M)rLA(N=ESI@g;N829A?it8>qD?yiDx2u;D*frO@Fb}I% z4!IO|NOaZ?>r3dt&E$XH=j<&X%vv#QCMR-GcKqA?B2Lie0rG471G$6wo(#6Hpqr8L z{u4zV45j_JdkT5Fjq!xu7vV`B2h3gwa(M3j4 zeJ_=T#|#kXQMo&25nK$~Y9FRTf+>B4C?BO)JaQ3JEJIsxn90F;zj=m??tKy$>Uo|k zcMZq=yocWjyXk)NVuKy7P)eN`0Vy-C3S@t4H&X|ndQt>%mV&|vjnX)j+ytO0Ww#CV zz$rcu=;WJ=px%n)<-jJCeZ#>wTy9uVo0aOb7WF8jdXJEcT?~{Yfo7S~O`@obo&2r{ zh|M2Pz86S!pEMSuqVJqQR?WW0lHc5vw{nNYj4!S=7t@h?IiOW&mTn{Uj)Hg=fdKFB zW};JOn^dl3f<60o{fx^xZar;Bw!Zt-m1bYFVKv>svPd6oZVP``CZMW}WI z$4-*#(v~8c92!w}KhZgceS|gq&?=`tv}T z#lAOCrMF;`-nrYS{Q9lYkM+1)WJs5!0ln@$DA~6i6<)pMje-SJ2PE3_sZu<=r}T{DCx9&Km%}Ua=J*8xZ`CGj_WlAzNQ<5OBa0(SSVUL? z32JXafif9$p?8={J@2d(5^PStsG-_kw%+{7HUTl>)DQ>PMv8_lI-n-z+aH*SLr$*h zT+{TJEt+=oys9!H8_^I+Jh7nlSK_|A?ai2&H^{e`_k=^!*2up%xdp1PMKwF`o#fph zxTAU30<3p%A)1hFlU#ND;v(AAcVuL{d@fWL7?J>byQz>?T0ay@ySZq8-}Fz~MH2QT z!Fg1$Q{T*d+D+@V$uaVV8SV6B?*QvzcD(hAjM}+r(+2gO4Z}^~Y}JHc4xq5+;fYvk zu@7`c;$mB>HloJoYPhlv0Cy@o?|lAWyF=?Tfxyrn2e{i@D2Zgb5qdN$JbR9+Ch}*F zrt%GgQ(@RKuo`QTfqeVG=qnkJCk21OTAaWB<5WrFrz`?h`ePx?cQv^bF6MnUmpEi7 zD)3`f1P`1Ul187byG#apm=*+L$DkrO|rB`xG9Dvw&#%-{Q7M?*{WbFQQ6py1_}xQfAHcdCsG>Q z%OUP15ffNH#TSFz1YevByLcqLVT<1ByN?J~VGKgJd@qiJ!fF77)}=}0b~piuz52@` zku8i|t!TWSF1t|4m4)R^*fuc@3uoa|=K#p>YNG``<-+bh{qGl^G%KuCC5ANVA?B(M z*1or%op`FgM}8Q1 zabaZ&;}51zxP6cB#~@63StKra3v<-ToB7@!kniACn_|llb$(2O+E$*xWS0S66G$bg(Ik*$0AqXL@W2A?*54ssmE3%rKTz;N_$qJpxJ=V`{i;uCVS z!U8LM*B+$B8^0C>a$%uXxdnSWd2{^n^cs;Yhi|mz7ZRykjS-y5)N8|}&lW$3;ulUITSCIS;W>A~6H;E<4@Y=aALMu^q%>4X6t1XWTwm0^c-MRyH z<&cXqheLHDrl`|a>kP)~=+x;~P*zqs;3kVI!q64!B;jD@GnF}oPEdD3g5}~vp1bXR zd4@aCP?Z(Z1VPc8=SUQWU`32!nWvp;{jwZg^}Me)X?KW?){Hz+$x&k-^!?jn1&v6s*` zlH;DAxOc9o8Nbtg3V=a&Dl-^(g}X2L;fLi7ljF!kmKkG#ZFKTvZh4@yJ0Chmmo2?P z)y~i9?kT1-DYx+QqY|$y#gatxva>65BXWm}ItO!PAr^YoNMgNtWOb2`L4Dt!SR4~K z;yqkNP0x6qSr4`S)uF?2GO?jnT`}MP)hHA>nKzd)7KB9GaK5$V-tCBAn4B&qzH^*) zz*R}(W^y17?!MwQChx$TmljO+4aAP z9lFnDFYC{Q_NHdr>b!Cp4Sn(om!dr8KkSlr)x}u|9U8H98yt&bZV!1b_XK&sxTrsT z`3WREg``LuUcdONNG{vE-|3Bdw+k4L#N|Z+gvHE^MO1aE5|g)>E|e6Mp(u)WSOeU~ zNTnp^TZJAu>gd8c{f3lRU0f-intfHWG_P=Yj7M#$BLCKTMlJVe93Otp#j=x;?mxHF-bbqYKnzuf2bq3E3)#YhcIYlgepP3H1+E9 zluD++QLRFd2Z6-_yP#BXIfJT-Q~2&RFnDdXZzbQicA;DgughFx7O1Gh}lbxVwE+8=R zeaX;+f#if8rndRyn29t%YrNN`YWU&0$xMzE00T#X7X;c?hrsrxSpqQe4)}CQMJ6nD5TXY1F?uyqGs#F`B|5f^ z`z!*Q9I=pLM?Uv+H20WjNv&suG83?3UaTOOg2+OJ>Z=ftL$DsaH%~s?8a*2w*acMV zOH(%FD~kTN^GagNu1qNxI!l9;iPp<<`notMhMA*ea zp}OI-x@W4u-Jys?zlvOLQ_WaSBdZ;SWK|fIZf^kM>t*jXEa~;J9^PU7D7>PTN5LJi z-J~1{+sm@N{dF-;XO<|UM!?z=wlhV6#<;&L$Pz1fyR%fmwdB5dPw& zh8?uohJdjvd)SC4j_sR89i?7xF(GjD)`Jn_bx4dnj~tig<+ov2PAbYpnJEDGydd{= z+-r=~#dQKum4X{s8s*uQjbAVy=6Kh20@)%8n9vjY!!q%mq33<)7nx7Twio1p^m!2q zjWg57M+p4Dt58pRt2J@&$*l6e6?&Pa%0}`))ljlhljBBF&jef4uG^G+9Um`xcm^DC z-pajr@-Q_jW8bge_2?6cpQeKRh_XP)d~ui9l^d_VzrpX}(pc1=7T?P}00@CJuyUdn z3KZr-vW8>)+OW`y9si@|>;ZkgY;~bh1c+r?E3s z>%@3MjT!X)yLk4gfCm;vhiTy!KXI`b?N{UaU5EDD<9i4kX)jQ_5(7c1e=)M8y}OYH ziJ{{Bc13h-&+h|>>|ds|kkZQ)M*B$kt>$kyEQ}^}JlX4AOhoYPep1Dn@$hGQ5z$)R z%5Ze{I_~n^H&GDi8|a_`O$woE4;QCk#_Guza35PX=5vhQACZ4$M||LAmoWqXZXjPJ^; z4y{s_2?Z00^R@Xb@~n;a$9-2YhVMw>ej-nXUHXi3D546HzrhCNTa$N*&2wJc>c20u zTG0at8;2xZ*D)CqDop0REogPM0>mJ?4L6Um5ApY~m$iLlh+xir_4nRTwHvO=QOc)T zZ3??+g(*lV%Yh+3R&**y0cTZq)(l#4QB67$G@IkMgL3PH9;JskO z8rxeV27c3F8Okh1!*Dp4>y&aZSVa`oPYS+g-1zE@7W64PrpCPCgY4ipoeP94UJNfb zyW}ffER37SA?A`H2j3keq~N~bl>hu<)z|<7tax~Nuf^rsC$gPGauPyW4FkHCs7ED( ze)e7~TROTP86h9%$39$TuTELAI&?%>!k6XDoi&;n+!9+6L0Ix|eRwt0BPnsJ@7HbI zc=5RZ6eS9cU4%LT3JOZU`%E=Zd5r_KPawxlKcvI8{dA1NdYFvalFRmwbz@?XRS62x z6LYldI&nb$0qZ-O{0^BLZR#NJYL?#DCW_Cj*pDsA+dHXjp5}Gb=L`Wn&z_lREW*X{ zE*CJk?7+0X$?#cO91(EFrXjsh{Ro^|`^{N;{wkj4xXO-q9HUEqvG2(MokScy+^HC% zE*HIYr@le(LC#DmZpebX2IN+vwTiMnitOJ|`khc+Iu2+Cw;>B{GkB^UK{9%@r$xMI zd?aSIvbMqcZ`~;YK6Om5Le>Ye3AtmKCRwL z6wBR6^6ngWIxZ%lG44@-mi)g!-^VHwK@A){SvdJ#b0DVt`Smt>~)57)2$#(`CK=c{{ZPcgFTuBH+saKc#L zFe*OYq+6JzR_&A@!p zgepk!C7@AGjWJ~;)y7cs>1$!H_XAUXrx4z5Y9z>S{L)_h^f;p1^_N5@H!#%V=`RM4 z(b+(fLKdSESv_={AN32!8B3^O#GLhiCAP`@Sgq$^XhGMNkSH9Z@p2L$(e#I(n4n@j zc_K8J&mo0bn(CGz!D#(^#yn@R(9#~gr!Qmo+U-YuIH8F>Y8qhT5#8g znX^1$b#$+-bmIDn&1k~6FpO%H2@`$c3%DoKW-*xeeyxu0KJhop+FYmFc}z1RUOp=a z7<-6B&s-|T3N{HDTC>9sQnTpg7mzx~F6GYR#6!!498}}Z0vwu&;@Js$2$A#-x@Sv0 zcT?d#mGTQufW*TH_%b2#K3qrGy9IJ|(s=YU7c~6Ft)TQwgYk`YC}KRdaywMKo082m zvq0yw`;>%VaqVU5_axawPa3nMl^E`8mQCk6G!937+q~tN@E#yRrk%NtkkFW9rl1U% zncs!HOiN6ve7eVakW9&3+jf^4m6(n28J1-JfT?MgGaS&@7|~7sqAUT9Mz`4`5W?5! zh$#lI3%E%Wg8y|jh!9arx>{+~N=ZYvmPmzFn`5&4n?K)dE5L)?I`|JD-oDl$I}FJh zGt<0H1F)DXb|-(QL?T6l;SdH%`V@|$&W)g7M8x)sB++uBF!y3SNVam9h9F0$_j}oA zAVIKOV1@09zhhS5`(s53*!yz6X<}d4^Xk@y=^!aK>xSOo>_~S5E{C-^{;c#NAz+1VEAwI zHGb8N>UN_Sru`sm3kz02a_XRe-p$^gA)7r>5)m|I|L^=Dryi$pR)e6RjDTc=S^feE z`VV%S{TBs3_WY7@%Bjf~rH=N9MXE_K6!D4r7OBwMekDJVkzvI5wykRvD9{SVB~`e} zK?hXC5$N4^I$F&7N7H;*{?L15q>9MmE=xqm*Fx;`0X-u3NTpfM#{rtxdCBc2gTY5` zLmdvF*>>;yL|&y8jM{8cU0gsFx7%yhtB9> z7gcYM@Tf_o`tgf&XWur3z3Akcq9Ey~$w&vIz+kK-&V*_93xF7bB8zhi+8~N6)9Yb0 zv2hSFE||_kOb~$-EX!Joi_KLgEChiAR|V?kew3*~f$S`GIM1<%9ba5Bg^w1_f+{$- zJNWpgIM*RZUu8R@wGH#@M7Nvr=Dg2qyqRx!UkwGalg(M2bI1Z3GIL{h=g~4Vq#jUy zeoz_zWeC{IhpLee6Ri|?qZ0gLvy>;G7nR$Ej2SBn{VV3TA;h%u_y-$(Vhl78^m4-AMrf05@V{eL#cbhx+lE6k5Nrv`%-^70w*i-1_1&kzTlEELl z81@@zs2=_3FT3UhQUzhH;b9JQ&h}c>e0OGgPhlqYT7bxPb#N?l=T#J&hz@c%VLz_E z_HXqck7+@g+6gS1{&KE??$=q;NNp==trZXo(CZQe|Li;azP<8Orugo=^`a0W7jP6)m6PeB|LO?6da z!pq#nTT1D6mMr~u43BxvupJe=HX5UW4Kmxvv{vHU*8cT|Vw;FrcfBV52spU(U&c2J zJoKK{C{@NvDm2-ALwdDj84!wb@8(8Rpj*F^%uXfAbYOX2!YFN_y3(=#JQaLn#+(O1 zIAh}dtwq$5B)-nnM+G1~e$pInie+ncycfppw6XYFF? zPA#E$j4Eg9=d>otU!lAfD^!f>6w`66bB&AQ2r|eXRfk^~o04^j2L=nZ&$^l}4=+Sv zR!Q*bF+?fW^-V2g`bkWsBI&vr)Zq6VgBeJz@0M@EAsgrGQb;Yc6 zIiw|2rJF}Z{octP3MpqQ38z2=DurKlEYY|tMD0jiV^GtYAEAtTTwCUk)SjBJD%ABJ9YP>|v z5~<7-e_vDCTbq*V1Qjp}pvweQ0m7n5p_I%Kcc$PA672JSQrmNqI&{4*=vs5T943rouZpA zy6f9Rw#CwJ&r}avTb6`kqZO{aB|pK}#j*4z!WUpatXem8SIj<{*CPMjWy8MO5t(F4 zoTZ}wNSs)b%LTcQn+XX7=Mhk2SbTmi}~m=VrHS7TM*JS6PIq3~Nq#rwl;r{h|5JpO*FL+2*) z>^*^gw^Uq+e#PLVNu+fbP^8x=47t>$;-ULW+w4&97EUF);gNAxhDI^3APG(Cl7&V7 z>7TFU3DrhW>oUU<&oA4`ZK5sm`kMvCxz?DyUj-gN!lo^GKrUT{2Vk0nCN}IY-Y@X6 z+{VW93JFzgQ}MppkQzwUPl>O12LyiO9gHLvzhFKRvJssodScc?$M?S!QIB?zCmhUn z<=^z+%DpUd&8@Lsk72 zZaW>%?=^~YZhsEigkE!YKDANu^GB4LPR;3G9ew&JwZPn1*bYkVr&D zVAN)a*WNLdk-5r#fC8f8pp{CGF-ScZ56!WtOos(`_H@v_Ppey7Tl>iylTh|)9NofL z`nmW7oTIrcw{VvhM%bia_>--!P}a&I+THXJMnn1v+0L@=_T#4zu*M?;i8<*ep)Lqn z=VidjD1chM4M@t%xT=_^PSN751=(0#){%9{-hf#N#8B7rdMi>J*<`a=hG$;(yB*t zE%5E*P$kQ}^-KxTDc2ZY$!vU$2FP58XDFK2_6Hv@j=1eO8~qq#dw8a4RUsJ8Jh-2} zR}IL&mH>+R+EXNlJC!Rx*^T8cZM37^wM3V@cqIb+La^OEi~P-pm9dR385nwfo`fRH z*hE$_P{{*wU#s)<+-TUf8fW>8mz^CDe+9V((Y8;rR2!c^btB4JgH2kbMT}^8ng^d~ zr8yF8%PlfmCs-QbjF1G`7b=NoK(i;_ct;CCSuK9>-ju^P-G;**;W3cGD^CckqKPc! z2mCBIzUj3jH+lZtgqb;}u4wMmz$Ynze!q*YTSm;w&3ragPC8ub8?)9@g5rk~ob55}4v7{u)(x%7giC5K29O165^wQ=2dXWpY zO4&XX4~+Ugb@k0#-r%3cu}dj0DVxo*9!)}i;87%)L%0B09uuR4JWZm+xdSqHVY&mb z3R_ffqoDpKbH8(<;UTAF>y;IgAucb3?5swPqc|%hQ_u`sasELC$+ZX#0Y>X^qxf;J- z0Dk@?;hu##g^e*q2Gh+?)_Aoz5O2Fwlo1?Oo)e1%ew$=v-KmD-^V&+c)1fCrJmpS%uMqIiiHQdaNMj<`F zZyZu@TFO)9l;-p+j9o0Rqlfx8_XY{)#7NKr=LN1=UXU;Afqnq#XjnIR&Y%d?CO)+K zFI%*o=FGc9+NRD`Z3%VVR6a_f&4(DgI8E8p&yDAYFW?bO2%d$cC9~u&A$83x-*4N% z`7-J4L}V|D9mADp3)4GY#p7$ZsHm7!`YS)f znd^5e2n`C+UaA&TclvSw`{+4+1Q@?^P*kz14v;o|CD&-06 zDuAHK-RL$uQ>Yv@;_^SFLGhXp=&c9j5nU^ueeG82%uwUWlRKrY%=P`pNG>zZ%|>g< zNSfm{c4GGXATbpZ?OeZv(t{x$!g;MDp;C*%>df0%`lbvbgF!MkYWiT)VS@ZOB6xVV zI{$pc4d73$a8rUu>rJ|&a9Rz*bnf#8=zd3h>~Uf=XayTI(@9!25lFtj=j;^GZ`D7H z(64QWo9fs2v`O_29=y+*Wp4CS8NM+WJKq{DPQei$>50L&9poNmMIB?yQI@h z*5_lyljc>p+fp4tI3IOAyJ01GRjgAJg~E9+DcQ9B4yGWCbeYhKL_ED#?!CpbE~i%x zy3jLiOdR-*AENv5A!6&?+&HOP5nrUEV7C{^*8lxHGDM<{KkTAGar%oq=q+V=A5MTc z7VF~VoGuE8PI;=aPb!1?z#$PzpSn1Fck5`%L0_#Hjz6aVEf}}X|FM&RtZWqN?C!fo zEyFssDBd+cht4sf0cqOX@|awb>l+y(?rNOvLhL`SnslX7ekEfuY1rvo*KB2vo0GfPj(B@1``SVNs zgc+0qXKIZY_I*k6D27URO2953kH*og|8Vkp@NJGv5da>l&2sQe6K)zou3>5(!Qmow zKAXU>Qknh0Jy`y^Wj0Mi9K-~*FI61W@i(EJ&_VA5)M_m4kUet!C2y;e7fyNxH;CLi z6+u5~mdG@crFKu~4EpLmRR|E&L5u+ndD!HHJszfNx4oxnqg8ywqJaGi z0rB*bbj?AyPx{`!bNEu8ADNKD;Q)?c#+!=Wgn zCysC_z$BhIy_j)|{aDS}BYP3hALX+$2AS;Dg;X|kcd!7!P#Sk^2yAfKbBY^vsoCT8 z`_)}hmbNLk7jLjC4vj>0StLYPTD=cL3<#S` zq2Pv}n@@^89#O}k)?N{`VAOtOEvyk-botSgsy|!q_-H!!nH_M=d5n#Ytf?ARfqZNe zg$z6yg3xbv_y-6Z&h-!UzxG-EDrQ@^ML3PgcdEd^S;;EdRo@W>j+EHO9a5O`*08wj z$QpYQ91qI2NxRnHJyCqjAAoS1coAw$%W*qSedGn|P)4;QaGlPkn)`%p|13+7dIe0c zcZgdHEnJ`kr72KVbnNOiy&1>0j$C#?c>O&W)xwY_cX~#Bv1OT2Q)Qu_9AP)n!#uUP z>c163>a2uqDuXL>T4aMt9JQ07>TSK&XgBA1%Y!}N5!>mb!Ve50FSIZOTck-dmrC=! zx9NagxdZ_AyElz&E8(&2E}?uJZ%20Ku|9bYP$euCyOLYpuJj>Z5ciA7qoz-}&|X;P zRH#)wTyh*{;2->d7;qyd&l`0k>gycWM@|rZ{(M!dSw-2GBeaEUS&D6g9DHet6#*nB~K8B0-Fp%qbOEuM2NH6Q1A|NBx?1tHd=;C6|i69 z)Wo;HB>t;s_!L$do;3m3>rAdyWx~Z$m%AW4ZpzRuR3Nj$C0lsj)NnmvP_Uz$%B$w!8EK3S8~WAevjF1RC7 zd2miW@aH&$8Xz~m=Whf=6bM*ZS{lPsr=|`UTSn)Mpbs_yR(?*$zT!$r9zz~uKJMhV z_Uaaen{)9aXWrbheXnJU)yQB3Zd6_lgt6tjtjesom(+j%Eq4Edh;&QORS^M-GI zt4u(S!8|F73)zbO&MbPd>bMM;hDll^M`r|SLUbPwHekG5VSB*~Aqpb$38|tc>$G=@ z_8v~=5eCM8rfWo@|04}!)7m2NwWt)Za+Og#1j5j=?ti=SOdu8C`(*ldk8NuyC))mv2j<%dqBw0@Flff9j#*JB081p|ju$T^WL<%VX5K5PEb)@|l6vlZdw!Lb!gF`cA z0_r_AVAHr5ebRv2Sp<#5kLBFD7wYNaJOVmAL?W)>RZb`k=g>NdA^d0&ve2?%TQH8= zm(Um8L8`kxU+*prE2&TqbQrKR)%$g`U+CNgDDdL>tmRaSsOc+vN^*P5gB z$GhTOLakbxc`lY?-(cZFc5afz>wi=#?>1nX@78J9~+4FNpnc)xlHht<4&3vhE5}w zF{;plA1Skc(Vt@a$NyYSe%3CF(p38?0vgQjz}u$B+E4L2G&vIjn2i)fvQY`S$Qrn<${g);vDv8{0b(D(@dW8ohh?9 z`6DihG^FEi_&_wH7lI3C$hLmC|EYhyZ@cPr46eW5)$G91GK*-5^9+SCW9xL(4ff-1 zn%M$>)o2);2^59dml96$aT!JPJOZA}X>_Kpj@a+$l0(_nWu+yzYtJnu`UAuCshqL_ zSR~n7R((J~W2=Uy78Lk3DJ*}fT)iDRuF-sZRW#;6K67-KGDvq5Jx`EhH=xYyVu^u-eP?IcS8@mn7&+p*ic$2mYMr+2lk%7lmr|u@Kgc8eB1Zlk-SeEhvGv*^ zc7@ZaD>yJoSj`!G>#6ZHr?SorI{5RfP1w++UY+wXs;;$lzh54tSJs_xJ%;^R%_nh( zEwUA=?X*c{11XT!Znz#yy`edCw^@N*@lrgd3<}#%wJA@{NPfA^-fmZJ*Q!8#BiVOo z)1ux`>E!)&UQd0q1Gh9<&9)pSszZ^FBV(O?{>zgYC>Xjr< z6B#8~B8;eJXM#ab{HCi-7&@dB2$pFQ&qX}0DF-TrM8uGOHYsL$WvsPo)6Q8)iI69H ztNM`EEOMds;Ffw(QOMDR$%1R9@KgMJ(x^oDR0Qn#Y;t)2bO#>HchfK6o&SzOY%mA2 zZY)-k&`E^El{?p7JEuY*w`eCXkjcHsm2E0L2DhFZ#Ym4j1_%Q6NqTqvwh83Fl$_FE zAncvTh2U~dWV|`>ev>>j3+&Ymm=uch^1aG=Ed41XYGX@abOZbpq8k~thd4+TEB*PY zE)G0`-zwG%(J~@H9n*6~S(q1+7O|P+dm-<}iS!XL=!O{!GH)TQ6gAO*D7e8tET|ec zRD`#BT;?%)C8TT&h7QggJDu~y=qwE{U-guK)0_!6kX^v3|T+X7hO#nvDV0xf!W5MzhXw9up{`|}liUXnGW zo{%Hx8ir!=TK^;MoR&mUmLNK|ZQHhO+qP}nwr$(CZF~0EzULqA2lP`%bVpWYu8jON z$L)vmzL^aFBEZubMA0d4@f-;ad#>wX_-?11&3$M(t;?@WTE0y`zINUJUQ;t840|*x znS(+Y^3wiD)0Um$clrKS-IiMyjH?hRo0}_Rrq6Qz&tU`q>$6?3203=?wGV)ceyg@iuUGKT!NBggbj1{BVtX!_8wDu;BSH`^$z)OXkfmzy z`X_Z9HW@iB440bArJYzZoOm9JXF8ugcko%md`7_dR%X@bdBuQM%OTmDna3zbi!h3W zr>QzbLxZr_nRo|D^pSrw0l*fU-JdT%$y4lP*p3}|8FSt_OoF7_;t`yovpP?s7{n3O z%^$H5enX9A&%?;*tblMS2I4Z43LUzO-e#@JqAE6%Vq@nN#f8)t|RdUUiG(){6RQe^w+{j&>r)yzq-E9Xc5f zhVStKat&$4ab;2LiSj<}%;)1b=U+lY&{qw@`5#r@@u%l|$TWxyIpe1ZFsDsWtz%a@ zgPdpSwhs=j5cnf!`+voB7bX$Em7o{G>r_&tzEX$Q;6{ZMrrZ}Jfu|TAasF4`8i=09 zoy(=nrP|^Psletw=N#1%>N>>@&Uj=kwW)ijijCwwd0}P2Fq(Tq3a46qY&Rm3(}&HL zi-A3erzS}P${PlyFfdi!ZjFoaG!RsoO+`irnFuf(Q7a{}ifk>kQiQ|I*-KBer?Juj z+?`1z`9ZF9hVt_Og}MDxuYB5U~`apu58;yt~HIg_6+0F31p}#LwQ%> z*IAobh`-79%-|P4n!+q6^F7t%3PwlwDX6bOzz}{qMmnwrZJw%vx;1MguJom{kmQ0- z2V7x+&!Di*5omT1O#yi2*eFxJ=wPP4OjDwlO|trAHc8*pH-HI!5MBoyebPD51dHmU zuusC4cYVTr@*?$5oWubecjVLUAL8P6jfQ(aKN4o-L@a>(n;r1M-sAz*xly^sB|n2v z+eExzgiOg%;Na8SsUH0HV&oB;#UJ)_yeZZp{A5N6wzOe1Nt|98_#V&SS1;+F5cXjL zS8-7guC-Bg>y9iQ|US^Z4dyoec~ z){D`R_8(VK6Q&pJOCC=?1?Y*s*!ULVg0P3WXuDhysy(_iL|x-kR20%`+dNge!)=x{ z1=GOUrC9S>T8j{CxQ$&1Q6dmEGWQ*Vy1Z}DE`V{Am@7vQzU+Vz26@6K0b~n==*gKS zpM4*0Aa9g5WEarqJnkT(x0p0-YZ>T+#KwH^qYl>F{CpIr!egz=O1McJgwLyrW`-}p z$%Pj|S9)G zP%7Lb^eNL)q9_kCO~1a6E>&!&b_mbR9fy*G%CY@Bk$sYkBmZ*-)U!>H3apFPhVLwX z=??V5j=&M%g>%el@%M5{Y;>&>9L=a9n^!{B$(-B!AwS)^o+O9wMCS|U1(DFkKmOAeEOD7P`$~_0Wh|^?_Td14~~YF1gBP z*a`ev{vu_*9GP3TH}S9$I+03Iyx>U4QEy+-!Mn>Vj!;9cSJ16g2cuLA;+rL_M1cmLGY_vGgyO6B{W$cDAZ2g88q7#^|3mi{T}cAM ztSXsY>9ZKjDllka6_(2T^8GFCkr{hTv}LoR?}6gwt%TDy<>6Ck%@a2($iv9ZU$0#K z9J;4J%NIbG(Q&znA*lN-fgy7&+)y%q=PSEE(wyY~GvWvEjA;4saIfec_bq^LCzof82iMA;`Hc)t$l zXFZo*rDD!Yh70&kK4;|fR)F8M$7N&kd z^9_3|m3b0a8!AKmv{x}q*3p?B0vYxgl@>g zQMvS{GJmL@#+ZAtUSm5WI(l1ZmyI605CWOO!4CyH`G+UEU|Al-#$kdksqXf+BKt0& zR0_YA&Hfp9x1Y7bJM4!8A9X){4#$qn!K_B12~$^?x8O#y(|ZiQ&*C@9v)n!kN*RDL zchsG*cC3T&NQx&n85T~&9QMGqcltWel(IFGfT4G{JDn_0X2p`JrDInQs?Ji#=&3*(Z7ZBwn~ex~95kkq4-5s*9^ zsZgee7{VWD5gI5rQ*Qhsd2XPatv^VX^~Q%Xk@(M`XTM(Q56Ntj6{E{`*MmcQ@dx?#vFXB$o2FdyFG?p&n}`5ardXdzE9j_F~Y!$#?YZ?Wh=iw z-uhmq3U9{xThL8mgagP{Yg9M~Y|F^B{Ni@B7le6Tc+$RVlP{s`W= zlcc${n!9=1P?SCHrWKV2UldXWD-v+3GzyQBTWpiIglvUotbE!+o?3E zfCm8l+jx9Er))R+V_oEDb5AsXZdB@M=AaL4-u!i|BSF2kKAUbK6Ur>Uo<{zSg2xIa(NxtVcWtam)gDcBvmxwQacH$a zcZqC(N?z((0>#~$$PpkA0)!YIBDe1xqmppx#K&1foJCO~HyvPbWLZ+4ko2HyscWLJ z2NL0eDCb`)c6-JL3Lmbp1Hf0z`J+~d?tABN^#;hKk4pR=w1=>ONJx7POkR%D$jkc z5Eyr3gpn5_*K^sx-sK1p|2Y0#O0Vl!Y;_x&=j+Kkk?9MD+*lLu8f5>69c{M_7jWo+ zGnHNVFKvHBD4w+<=O1Z^QMhvgWdKIo=`-?l(2d7dG*YdqjOoa5sA$!H25?U6Wz*b! zqz}CJlW`hD?*_$gyMj=uN~=J1K3HoUj3fOcmm?p+LK+)I-OIE`Xf)OAGOY=1(gB-J;%=3Gd9-g+* z&XdAb_WCr~XS{P>-pR!c%>0SizYt{t{fd0bMBY=9iYTutZQ9U7b{qG)v)OCm-8Vh{ zlDe$^P;b$Yp*7jGxd=N@OW~uNgoIK2S1JjdR}{kM#~>LY2IW4qlO(Y!6Xqj>&@lL( zBDXY9Y>+_PurG`Licwgn=)_gF<$1QoGrbMe2KAGo)hPI-gJfG)n#@7a6*gR9sA_TxUN|dO=Z!H{EFu zsKdRw`!!%-Kt2iYJO8&wizQE^RAI5+;1-d{b|a0Pk|Q3@Q;<52G1&t#TAKmQ%;Q9a z&g6vWsy+#Lt18M*u{XVsjRJbdwvh&4p9SbdSfFwZfBaosIhX!j7LyJSE6k_~CE5#U zb(Y5v!CQIHVv|@9_!mZ9e8#v~#-q>YChT^%>j+K`YV66CH9?-oULh<`@^1EmtYzp? zpz>+3(*UUsBxV3Lva z5>P+_m2KKAeqg7X6u@swf)@ZhX{6NX!;TP3N`qBBKd%SWpgLuu5RAqnY^BX)&rHGE zgb90%{N zZLl#HrPNjGQkC$B^aimOfT(82-yY-1lbjEOy3;pqEzRC`ktg7ke|feFKT z6nV!2lgR;Wfug)@0b-;&c>`2WFOiK`WFuXHloy{DeavjJo&^D_Br=uSvre~ECzmsh$H z#4gXk4;t$EK!RH;!w=2oJ%ztqel zd*uFKh86Bu>h8k$J;98hv6Qn(tr8E_3JV}pss90L&=!CJ*r&iLUKKVOgs2l{;SV)T z4RUo3iclLkkw%=8iyO)3%bR^p@o|t!A>TZ%2jQ8uSIW3yna^s)NmFDC9TPMxd&UW6 zRkAgtk(8bJt|P&qE;fSX3(EqRnFL$TK@Wb~dmf0O_3*M{8t-Gh~@}QZx)Jw zWX`W1)NNeE2tHF!Lmc2oH)MnyeRkSLUdWV}TY5J04F{j(j06&AGAh8_(`1Y6YmLNO zuMfnug&dP`NaTr!GxBY=b(Y8hahDRk^2CU99{zeVM?whCGsI+++K2~~;BEDYiN6dS z(RxFpwE!E^8BG#*EU3&&$P|cAAe;xIIB zCgPvYR9je%NFEO+xFo*;((lNP<7g11%vZJ|VJdHP>WR@;e~H~FQK&=#5@2^T z4L$muhbjHBL53z7=u$Tr^#?ccJ~*CoT*R)yJ-=NePphyPEdZFolK!*&>y_FI|6nS1 z>PnHxiu`0eNKkJZ1aa)PR+txVp1p0t_QvJ5129J;J`~SpXt_z^M6Q%?kcnqh-Mgmn zeSv!w`zmUHxf2E)6TM^URtHv2MaCpWDk8hX>!0Ln^@b67J3u%Bt@H#id_98L8AH=l z;m_GDBWwHlB={SUrI28q3ai=avT9>5nV|a~y^<_@qej9MQj)`Fa3EW?l{ixy7i)5& zmb3V&Tb!C&P{VP#y{U!;uCn%2&jG=cgZs$6K9}bo6sw%5ZqY~BKz+ISZ;!4o1XD)( zSsbFT21(kPbv=XbUh(26euq4~*1xXXyb?5K< zBV4sSr7{su-pU#baz8y)6?unoW>!Bx^`}pr)K&*iP>fL$vsHQD7cB3^RkU$&!EcY?Wut>~Vgl+wUhT zr?JZrl(c^y9_S%0f>xY)!wCc;EP()M6#H`aY`-}iA`DJl3&!<7q;ZY>feY+^nRqq) zP7&zdlyH@IVK_kMwZ{2&IiD;QFcK1ZqsU^JDc4JtNOAU@A+~B4pT`iGtan-YpF39( zTC$ZRHjcA~h||>KYpI_6i1p!j0OxRn9IOneIccBqw)u;4BquPptZ3CM9FKXHrZ*KT zf4&qcZD2Z}YkLO&>A^kIFgWR+*d}rQNt%eM3!zPD%~3t@cLpX;Kp%Dff&FzxD55vy z<~ff#8u(0voTmPhrcPBgV)QeseYvL+4m$+*vZnKhyva{Cd_kyviRXby)R=7bYdUgB z>9}o>O~ejwaMQa*>$7Zw%!cWS0;A~yTjN34^Vh=qF541oC~#p%aHjD|n$Xk6Ls8OG zY)Z2n80}lkO?0fv5rc}f;TZXb5r48w&&Syfsg@7n=7IFY3yLtm-y88J+F4?7cN-d$ za9{gooK@YkGH`Q~EL?_*Dv<^^4~ zUY}A5I(m?+B}4?@CWDa@rF%kpiO@a*M!LeCwRS0Zs>rSN~+HN`clC;Gz zEaEmBAI*7RB9MIrqjS)GvKq!FjZd&Co}Vka+o`ZSGmKYnTALH)CZBHx2klSS5Rj6< zGtO6E64`aOCXGs(KynIhcDfkz(kWZstk(@qn97c?$d^vSzDlhIOE^Tv$_LG}Hbl|$9~0lw8Izo) zeeQ7;(75&x0P_lxUvYP%(b#03*3#Qq<}VqWCh7CicH)xHJ;@QT{vgEZtYKv|o>RHJA)qp0C+6C9G_*MMVfn*_jgp?n` z7Gr3}tDZz3Z-q`tQXGLwO{TjS*}OR!y%_%9{>WLw+s#%f>^9Iv`!~5qC#d31_z-L- zY((kJvSk5I`Lja{n}Eb0VaQ`RR1^%zUXdKDnTWU=?#QKvIl7iNQ_vp_TTXZQq`38f zKRUVIpWLjiL)pl3+bM)Oek=NvtvSfMAKWq)(00JT;5+Y#N5ldg zXih5oOB5-KVmt=W8rR*dML^4fdUCX%%X$(Ah1a(P0&dxIOd^TE&-!T(35FH37&9Sl zEcqzF`ck?`M&@;cvFs!6GXYJ z3&rUBj^lG?g8CGc{MbKd;Toxe>Z>Z;7hT;44sR35e?!~h7C0#4_@kP0z4GmJ>Ote< zq3=i`h<}P)lv1DeQ%!EUewM6Eu;u84JO$Guj;r-6&-c=Z)QW;RS@daV4_ z;-y@|;+Zn+EuU!_x>(g$2i{PboTHvOddhs_YNXqW0V$-HDVT8&_{WuRxe{)t1lLvm z6G+VDOUVu#Y9vJl+CCXN4aW8s)_tf^MgNfW#Xb$PJWQ7~c)%y$~v6VDykm->4tEw4D63T8nOcdmEH0i0`Ne(DK*&Gh-u&-$1vcL?=8d zF9gf+_$U7}b-u_$pF7Q;@mE8pUJ2Xiw0=r!e3$HG&F+s5LqTfHvTIP$MV}FDXQMZz zW?Gq%2Er&j{v`$B5(BykUH#|K7`5ohDV#p%D%8)l1*=)s<+S(ukWpNaJ!2w`ep8PF zp@Xa5cL^JvWGvt$$YC*9v-6>fR1%*lpXsm4r>if_Ngw5e{Wm zx+x^zZw19lYO0C&Yh)F3JLG8oz4CY%xG|$I2fEhQHp+h0)X~789C~x>)aJzbIGI5q zibjy5a)r4)3cFv`(cM;CG8|8;GJ$D=6S)fg zT2m;~T+LV1$Mj_!0UZOApgZBl-u)K8hy)CKdJs!;Kg~!pZgA@CP{nD8XfnVbJ~rRM zyw>7__Kt~4pR^o$z&pJ!;g)w!{;B5eA$6~7D{y~Vqs?Ja zKTU?bwcYZwRUpq>eQrv7&_P6M$%^?*qQdBZ+5Q;*E1YHiCb01L1X}0aHmg#DQ7@^ zvFEilBcjD*it<@4aO;b#Nf9Vnmj_0WMF`wY=Xre^L!3`$H{~7PLH(JfsP-E*a68HI zHVo=K#Xt@(^lVuQr6%0tfh|AY`ghsV4>8H*mI%P`xv63~;Hbqi?BwK?^UUc@ zSDsvtKm!aETC=Y4_X@|-tEsj?V$So6FND@IKpG>eyZ%c~!-+M|K-fr-EJgzY64$RC z{G$8x6xqN^eOy_e!BE;i`{*W|jN;Wr&q86?JxCu6j>T zb1u$dnfH}|mP_0x{Iyu0c1DCcWPkR=b1@yQ{zNWaKjpgRVBqLspgdlwY+@I0QD)6* zl}C)nq9*L>kzRd7Q#L2v736+Vv4^7NfLfTDVUx^-_nlY%$@&rUO|(^!4o|1h)6KNY z`xpy<#A1x}%-hw>1&ubvFUrer8WjS_B4H%$=T@O_!lF_*;h{zbYH1xlTx3Yjx*+fq zj8Ac7W_z zM=(1k0OaLciTGb-aS)W(jcdI?pZ_ZQ883?7H-TI}ARi006X1g}xt`m)(3fD25_mOA!MnIyYgoc<&Obf_iE_?N2t`;Nm|~iZd3q3)Ye-@=8_25xZQK$OK9y)aeB5nF1*=eumW)|-blNECraIHj(Yj^@ zZ1S9mffv-4OJ*1LTtORY^fV-}{?p$_u;JkWi>GW{s1;|DR`ER%nvQ@d#Yf}}N10$+ zgE@wrtIq+}YQOd-93O%SE;qwflZt0`AF2GyehiJ?x$j>&?20{mPzA;$<*!U2#Hx~- zlf7mVM@R)!Q-#`d{1&6qrn{m<<)v1jjy@@D>VIJR#NMr>QjPQ!0x+FYIL z82*o%W4{>6Qr8_Ma`35-2T-9E=h zvPYy_qN9Mj=&yh=Hgc{yTn1*}t^x(YjBj_Wgha-pQrS(JL+&QcQqqKZ(E(n08(?U@ zkAsGp^X&Th{EUHe26o!;x{t(Z2GY^Ve!&1CXKavT4r;-MxgQ!rn= zi{|av0u+riJ((!_D#S$$V?RtL#sVrX6>zS|;7&Z77-j$3G!CP{X2>g1@zpgv9S*5! z+h8s;{gJ&+6Bk5#=3Uo`P=vBud>WenPJTvA!=oIWy96*-3=L{8dq_IYCqgL2UAtfM zZ>8SNNJ7_u;}AR<9Y7_tSrd4Ul}hhBN)+?Db$DC1-*m?MgGWJpoJ$SL;k~T8^`UkL zE>s@H??6xgt6J2VBkf~*Xl_JG(Stxu4gto%Y$umDQcX|08zYtsIy;fP2afC=VXBBo zG_l-1b9fQ+n(IMo1(=?+XV_c52bX(ID3<~M$TdNIl*y>RF6dg6nHai7rK7q5`#uzC zxKf75$8eka!Z>F?R0DWuTN*=&!;jPJvczmAc4*zK8SmI5{0DOjz5QW5f0q zK~F01gu|0n2EpN2`J&$Z&6-Owa#xd*Q72UZMI5vOAlGpbL;#;&EIQFv?6dK!4lrPC z&idsW!|g0%6U5;IeqFC`N(?zVEut~wAm?PIJ8R5NwCWiAqcSF*DH_&=#g4M>%10rO z10ysEO77)@D@e#-vN4r>UO!kU1F!>3UC>J|fk@`C))rFA(y0yAdxIODZ8s7fd*bpC z2ayfh3c@btTg+P39}iYM6Yh+C;%yG;$N;mdg;ST~#=r_Cp)w47=eKXWNprnqlBbE4*CZ#6hd zb(A3|*;n;}g?Oh>B`Q&z@~8xdh}bJ%yrS;~D-W~t9X^Xc6T-fqZ@CDvj3wGz z7785{NZ?=sQ>h&IxqI(C;o4#H3))y02`(gIFw9`i4YevUu<`RHY`_Z6`ri#-SHsb0 zu5RStKbICk;%Q<}a(#Gt@DUpTu-Mx-qQy1I-pHzlm~}&LkQ*Nov#Y0HX8G^OH};$9 zTtNV|f2#vsOHPs#@g$x$G<$e@^8OGOj}gwug|D&bf!*m{B(+qU zS@g)+)Qxb^^O#LXwBhXQOMN%OdTN_?X&+ml|jKr9OR)%Q~W#bYS6legF z@QQ5oK$p7gZpAHyBJi--d3d&p+&iDFa6Kd(7Ke|5hR}eaVj~F-3A>S0YI|Fmt*wNS zKNZ$fT6+_r^eesg3+T4w_G6#Vs8*u_;sN0yNfz8+xx_#uBRTOSkFCl706S8NWa~4a z&PU=EGE(TGXnQQzJDv;m&OZC>gc2|9-BKJ|A+8_Ld*eW-P3!-0L>%%{Cns!0UN}F< zsCu7ZEV5%o2|M~@Hc>mt^o_wI;I5V6y4fyW+4M#!T*WE3{%JBSl=#4@LT z0mv%rgRae>KSKDBDrpt;;O;odcWAM-*&di2l`MMv@#Drv6@ABN<5R~IOCu*Iy#Qg! zva-m+gDGWlJO`ypqr+MNafS?x2-Q&4-w+kbqd(dBj;63V$7$5Nu_ypJul3&1**LT&{>$J3IQpE*{^#(u*uCfliTUYZw^8BZLV`?nRQ9B z54VVw5F54>)TDVkv46uQ8K9N6*o89NfwlNkqCRP<&PJ^Vub!Hlgbge{Cu!N)QdhhG z#vKQ3pQNWEthC@@p);if+qpaDNDDrD&{!4D z0}dxAOwPO=E0vs5#m2OmH_W8Vy@>9DyBV@<(NwdlU?FU_1+&N(vg+~d#@s^yA`_6l zVi@pJB@>}w)6vZFQv>dU?(tRX>HQUm&gkN${1Pr_g}kRScqG^rVR@hP4(CD>O$+jX zOjWfKGr$QKL^&&KF@id!u&@mBsM3n^mz|UMw}bvBh(EvFW0t({hZKDy1QO01Q9UUx zZje3)K2L9EiIjx40JFFo5)BC>W12~S1kQ$Wka8!X!c56eeDS&%04rsJ8ux3%Qk;#* zC&3P!kGUe_TY_0eG&nt{$sI&{po;(Z-d#koe_ltkjsfQ}>8eva+Q~x7i1}UR7?*v~ z+J&zsH-XS_$iSntfKQR=o)%T)#)oS7Ar5rDGEKYpou?L0YVfrze!8`2ueE!*pj|m*g|Pc`kIBzCL(P7rE*Xsx#XaDi45~kpTMuS6;L@znb|Jd98(~)Wbgs{SIv0%RC*~HX<%s%GRoJIat-|Tte6X@d5^G$xN}lG- z-R>Fxg$FsU=dt4~X{+ASs2uv%_0v^DAd3{|)^dGQ*O@Qu;Qypp-*--dR ze=oskuLErzEs}73KdU8wi8qvUH72sO%QU6@u1QH|ky~wUem)>Wg4hv~R@Vy_FlEn` z-f8k?c@@4@)70R3?OKe%csTK-2SZ(l2k9IU@2E}h*gY5Mojn|K4iq$Dy8Ob4Ba^=d zhmt2$8N)20CE#zX6*DoMH|}PtDE!KdNX>aPC9b3*E2G?zDIs`EwWe(iXzyTrp|-rTx{22*CY8n83*$x zzWJ|us!#z%1TOTr*NbbhWjq&$RrsINbBKc~TubuTVY%>Zx{1yLIr=q@@Ngw8FB6FS z-6-q*30}a!K~BMYNd)IVWp0NKAT2z_UwY+Pf;Q>LL0sacrTV$^|Cc`@!|V|tPU#Ku z;UFD`beydo_fF69SH!4o;X{-3J?Qr%jFOD4l=YcI#4eadU{3ctsgu;1+!PHiM^1HX zA{zw6a{oU2i+K78n>7Y&$d6KGH$M~GUxA|{xbd$Yj>=~>VF~p7XNacWRkK>b0(*Yz zU6${!r?58)L~;w6Ebc#KRPP2#^xQDYCniQOdE3w6SXD&H%it*qEq^gYqh(y8>}Yhk;js44dmalg?|j(&+0-C+ z*(vVM_@}8_Kmjn${D)8IBFzq+O7-%Hdb%^-_N;ab;gkv{9b5y4?t6s2WByRYbPg}S zbRXS!Hq#=x{;o>c3liYxclcW4{a3_c+fI7K1OEUHSKj9zx}?9sL|A7L>)T>wUXhIk zx7qw9Q?oh~r-^P`9Y##jH}hH0buzCCil$l$fgiuneyt^CkR+4_EZH|5c>VLmb0wqE z0gJqP0IQl;h8HxT+f2BwbzgBvO*L@h`h`0yv?#TWF?({8ZL~_+5|r) z-ga_;@wcd(@LiWpsZ1%LpM0=9pF^kct0Rb`{@jhZ0QFWzxcETQJ4F3(+*&Khl$92| zz(H=0@Y>d2m3Cnx1IUVh@bowsM^f-d=4!g1aarqrs_E`Z^BN@`@j8c72pJYbOfxa_ zqD`(Cd-3P7%uR--D86Ce^%BI?g7I)zD_F@xxwSbvD)w0y5Ee5*+n^>wBO2nk5*T|DM$k@O~{qJhcSp%Wa$v}XwQ;0n(K zoMlqIMEe-I^%_m}#v3-RDe4@6Xx}wPxoVpY=9|@kI^Xq$3pp1~V{@3BTNS zCP~KrvSe)20F6Y8KqV(B$2_H%Nf0)wwiQS(@#FxJp&tzKLrmHQiqOPUohl>8!&`>! zxXI>3`N~@`U2T@7^P-$YGWwqV5E`JicLUatV!U6JerAFF@_NK*a-j6RH^cQF%EXej zH!6Eh$<}3VyvC!5Q0ss7j}Jo3h65HSVA51hK94t982TQ}@y>L@CIh7!dFaISMcZAv zYh>n}!Q8nl1a$%TbFb!ZjJzNCGpF>lQ^}#0L(GLIC7Y6CFHCS4rmao>KjOvnB1O*wif-RH#h&5;B0P zCiV2hauT-RisTu3=)mMXS4fr-%YaYa0^pQhFz!;q3R621eGh4>@v4Oq{iheS|M7!P zJ-=5(@mcX$t$c9{HvE_!)ItG48`oc#)swa_J!!|UgBiQm8^F~U!vXm z8u^N!tH<-E*Xl7<2mR%czx z=P{`(=zB3Fk>?>4J5?-XgR08Z&RTvl(x~RXjfnlT{`GHC55r#uA;D-mIlX=*p@JLx zk(7BDQPH%@8x_nAOV;$|9a}d8F>{8`S^TuQ)^DN+Ef{=GdQ3fkFjns~6c1cKp79uHPe~wH!8-AU&%d0SY-~ z+lv9x=0`NI3h6jp0L6AJA!8eDQvaDq*lnkL=%}b2FBpged9<$(E71X*Y}|%MFjIRy zh_SXqH186@Mc&5Vf?~o5*xzxqIOBzebaPoIhON`o4jl3HUJ<|`IM*JD!$@xy`jD$# z?B|(VTU)>E73%Bcr2w);z-I}^R0#bIF6iV^lY$`=t$ZdqLfi_W%B0|L=LY zCDmbAA4-jm2Bqf@8Pm_Q0HvA$$W-TE8jqNvoXrs2_r4;AF!uHhhbESN*A>)%-1tIi zuDz6?PbMrKwH13X0u0uPG*9qtjqJ7V%1NEWPI4ur2e~LKVG7t1Z$TRF?SE%k5P|PM zT-VwaSn0Z~J;~|jIb?~lej*SWMc>Y&ug2_l%2R$QQ>~%Gt_fUpXUG7zay)-{=?fOU zl)C4NyAh1FE_}0AKvAnXMg;2RLNqt=y$sThZfj$p{L{&WzHa~ZYEx(8zl5F4A1gf} z6#$ocTA`NuGH{h+>hJj1oBM7*qSn1ymizx#mGaU(0xDiKL(OxeuZUvZ^FVHyw#ufU zaZp8GN;2~&Q!0%Sr%4*k`wFx11y9~Y<<&Rff-eQJ;6~Oa9Gdx^uX1nXg+ak1c|TR$ zZb=#l82gl`TY7)e{yR&pDT*Yqv*K0}3JgranNLwLS3u%+UMTR)o;q3EHC4cMNB_;k zX$05PD?oa%aM2ekx2RS}9+uan3a0=$?U(*?E`2l_?lq>BZokLs_p=k(Es2WUiJIdI zPlR~D0FqV5aI4pX$gb~i{x4C-FeFtm_j_( zY4S)Z1Xa!+qfpI#o(4Ra=@$l)*Eb&G#&OF8;gJTpg!s`{`=IDlHH}(nJ8Z$7GXyhE zk6by^GR+gd?@d%i#JaDwGh4-NxY1Qsj&?q|VMw0?2zQY~6oBuCSrTJlxSOsV{dST= z&_4+noI^%(fzqmkb&V1bOW3jE28+piEBuo6b8rE%cY>_-Vcchg#GhnhMBKCW|2nb6f(R@kFTUl4oQyX55?(&H!0o`Xo5PRuyaiz-o zO+kPavxwfbb{HQ+a3(6>=(8EPy|GbPm)0l?*d{zs_IpGkoUfEximE z$oQ8Lewz+7x(`Du933X!c4NWRX3+ue+6uKWMueSF0(a~`8%{F(b_j?u3g5jZx3Kf2 z=3)FyN6k&5h0^+$v0C<%^et0-B!Uxsnv{#1Tx*fq=*}A{|WeE1V?jq>FV571P;fmTIMOTMN>&F<|jIaHW5dp`74<%eJ-watgRo4|cP1zx?r2=X(D0 zr<^H_sXrNM#8unN$og*k`Il4nhd=NXPGaHZ6y&A}p7T<~^O60C!E`DAvabz;t1mIV za0uP9$hPR`?Heq^O{Ivm!eAMQ>ms@VDaxiB8ZoIKI?ijgqw)2 zgxdIB8+YP-2ZXvT-Sxxh3!P^hbPwfjIWPSZ_}r0h5hdTzlo(T#jRG;pI_3L%8k7Pj zPte15`a;i9^zTHgq5abHI#` zF113|YgP8(7O?6g-tnA7Kr)eR1~rWFjbD-8?5qmPVw`-f+DOY#yb5Tq*QM@`H8C=e zgQSft085VhkMCCVU$2%c&(r2h7LplJf8NsOikhZg??7zp)m31Uihp+aS{`Vv`&#|p zx|>DB-*K{XccToDS|3wqUKFO9RV7T3ojISoAS^z7F4|#@I2Tqm11oizB-;0A$2^AV z-9wI0G5P2=V~GeIIC@5s69FUJah!M-ZwGr{&n+m5mN|vA&@#o)0|&Ezjby0m5aN!H zCtuNRP4B{vH@0E8wP%%euj$|6H@r=K?Z5&e` zFJ!Ob{bk}XA~{=Vy7N!fDZtKlYmiH`(UZvB0(7N?-~gK*3|8J>3@geC^5>2h+5K*D zz;=Gj$$vp`mMCtq_?HVi6dP^GmQHH?rGA1?IrATY;7j z7W5;;i^_|E|9*idJ@Y5xak~52cFVt2tCa)jX+7n){+AV}gcQm3IH=TxruZiK5v1x^sX;|eSKri#d%P96z#ud!)PSYoi zin8_%^k`U`+9}bkm>ZiAQ8fdW_exBs_sYuMDVo2t(G<%*Srq5-Io`47chWgPua z#ct8OED<${G6$El_$EP+PP!*~<;5Xy&(`~%vrKY47RRcDo!z6npqcoi+O6#a4&2ZYG(fo**gG1g01d!R zlvbs7*M+z*6VSdIQrKr14)hoxI_jm|t|$80u+gcWjg4OrrP}plXBH}I!ljec;f)NF zI$Tq%G=Y?)A=_UjU2%qYtSeQVVHjjhum()k>S!qr1#*z>i*5)>$?y>~-U)-aSIw5^ zcm2_Q`DvbCpkj99gGpCRUH3dU#;~cG?Si;@qLs3m!p?*q6zc=)lJiT#7VDG~_gU++ zh&B%LeN6jmoDZ^>PMiFwF;z3wGhHrur?ewJagUUc4GElmICNNEuSXbCOL zOAV&M2!JQc^+J^Q8ftU|KdD#H)lSsN7H;KZQ%}N{TDdph!Vj2E#lWZi6>gGK&YX6P zBJ)t?SEkf$#re%3YQ+Dew4S^1tJ?}`R?5B80TgG;!T~YzNc?3zM14UjQVi^JjVR8T z5F#<{t06(pM8PmJRk%pdjtoEjfRzPCS z_CDwa%tDPud5)lqhFMeEJVp<|{7shjGi&lqj8yL5fOA5^L9@OB)gCw}16)>S$nen> zR?G2ZV*|CiSAkW1uT_h!>kZMXO*~8{+##>&YC57LSfzL)sUkVuCi-Dt^mE-sMhW)=U3FKoBu~$T_2&8R5N2n<^B2Lg07WC zTf(GvBw>}B3T(jgp$Fe{*5qP{9dHV=U#n1iW!g~yIK$?7_@c%%-qut;*Zw|C66$@9 zClP6}l^ju%>YN~PWCn(~EoOnAiC&1UCAHpl9kvWHK*~{XjeFg6j1(oj=f!8PhizQ= zifjaKCw(1v>@*4yG`i?B*y|4H;h6L`N+v%-A4U$J#R@-CkH4A)HirgXwEt6{dyN4( z;D;b_GLu)ff>Q+wvDZTf2*#OsaFkC*#jF3^WbYFT~U$$nL+azyGDBSK0VKief$|26O?pJnl z9)xKCG$<;53h)Ps7SULA`gbLE{qg-%%cdt-f?vR(-qHE)f>i~k8`a}m15@t9P`KRA zJ{=o_mW=I?(gDgU&=L1aCabxhI;TyEE8}H!P!fS|JpWr(I@oW=y4e6gxlQvfi%Ssi zrkrT?b}=*eF+Jo>-GH=Jb|1;lEC4K1=)P-}R-&;DghLie%Faz$(jCL+O5nspjjG#{F>M|GXYI>I>3$}H_xIS3BsY>5a zll&SN_#<8R?c`JL%AVHIZ2^UfyLcXb&hBckDSCz;+95irb^_kQfx@YSn~QGR6+lrz z-d%d)`Bb3J1_jk6snOWTXU4^&c|O|d21>>zF)^jzribUs2r~4Xhiq=lpJ&Pg#xYpa z7M-R|cN6sX$vVY7jWLMnv+2C4k$Z<8&^G_$_o%9ewoKh?lZMU4+$+`3T~<*S2jcC* zx_2Gf%O>C%juseomLGUv z&7^aszq5h(4u$MBHKRn?j|hoRNA z;1I_Z$#B(=b+9xrlif4um1*Nev=|zFaqnqN@^>GVE1=(}Z*?L_7+2f}IGS4BIX-`% z++HB0H+Mqpa`DG&?J-#i4CNZRTV|(|d(vc|>$~7u&qb0r0g<3cR8yT|H-&-!WA2p^ z1jQvrmI~u{aT^Ymr;!CL!^n}>0Q@RYjn2}r8X zfjiJp1)yPjUicxCgvHp+nRuhT0~+&_T;(8N)xP=`a zL>cSe#6oiLIkzX^k1AM}?{-g4o&S%u+qe%#Y?7Qar|7H0E)qpVr~AMe78!P@&Gl!r-HSNu#0?0{wEvUar~3_~!-A zPzV~@%Da(R6B&r$(^sZoPx9wKAZm2xXk5{Ik*>##N&L*V?05@Y?30B4U>$pO+Z!92bSPHh9Iu4dSz8si8 z@=wO`st8*w2xnY#=n_}jbJ*a1o>g;b=~gWBinGQMo>f}%MIh!rrb|9K(`(8_?h;%n z4-S*}x)bOI4;z8RUeKxSJr#MmSz%twlu{v=UqwW5z0cMy=nJl6V7}*zkN+zMR2o32 zYi41K$0+yV-M3^f5*JY2)Zh^WAT*kKU0XeOrc^Kk*|DQ5t_`v`%xY=5t$O}zIg$g-Y-I)&wuQ!$Ba%X;(% zj=6vTV2a=M4M50 z9>9F$jfEDn&TU^7r;m2Wl0UYw1e@-{Q~0i$lSZ&Q!FdAZ&e~ce zaeXe7)9yPf+Ra6H(MHm*w=HG5N>yN>W>gXdzJOew({WG6pN}e$gu=hB*22#yE30tg zYf}MLHrTmj=qipq%-hC(HX&FIGm5-9 z?jw|lFl%S$z%)tr(ty8XcI9ZHV<2-|ZNm?QYwMlHZJ3kGF+MM|hBFG2XRJL<{Oj37 zRu=*Pyisx~1y{Py-ww-J1s7B~uJe>t|!ZM6(croHjBK*TolUJxdzsLQD~*d}wkMo*r{oNJLZ> z1xu<$K4?;W-j~A9PE$C_223-o`Rr^hR>%T zVNqrD`b;tN+{^!YpD^Jy`7W_gv3H&Dn_L}Lii|A@E_~A`v=zz6>Piah&cROZJx_7O z=A0Md9)$5~GE^A^xqy+JzVsir4(9LO_-@`gAzoqYr19yH0NhQnXO{a~Dyp1AUsBRHin$N5W#@;O2tAO}MdS{Hu7iZkb zTH3XCuNouxTFZz31DVg@hX|038xW2ZHVg@Gc6VFQ=QXSCn%9qns%@yrCedxK*T{$r z9JOdTYbAZsM-%{^b_}@-j2!CvAd*UE&XCAr_H{x6{HzoZ6j`u(i8I}l*Z&j`%`i?D z%(;Q|B?{G2&q*;J87Mxx@mPrek#g3tsZY@E7m&lrRxScdQcHq)>xT>3;VySr&5bV( zZ||vJ7Tq)N2OlztCvLrjA?xGzS=keq`sZc$k#zvA-l~nZ9Vswc9aw&QS^Q8pv&0++ zrC$X3^8kHENP7wJ?hfsE)myeRFRw9$=A6Vy?G$iIoKPNI5YA%1rL5d!2q(7;kh+(o zR!?^9Q671qR_|s1^!5mD?t*31DZ#jcgfs}9h~s}DGfkza;5cvY*Gsq7aIlfD-U&XI zJTlT-xB>nSEEIJ~74w-xFL*q35fo~rbOwd6Wbx{$W-7*({E?KDJ}tUFrSNf~NoP$x z>WFuEgR8PVm6@V>l4`v8THO;jo$l{rJUN+lYffz~3kEl^90RYy`3iBRJ1_)+UlH9F zMcRz=AAF_!^=BN{594XWdA=~&V%c1Y-A^^Xnw|p^2>)`I+eQ!ns4gRpS(^x+kfA#w zLKg~BxHNLT0yzW7E~<+k&q+jk&Fohpw`d$d6LaCIX@(_cu?b?3X_ctOV#~u2aEY3) zLr1lDl8lY45OLZq)3bm*$Va^V_yx9ZN^;SMG|B>5PPNJRytRMrK*yM*wI|oK9inK6 zlX$HfamgQNe^yMT<*(1~(5Hau)o~SiJtH{CZ*LL54D1VDHQN8qn-U1-kZaig35K{p z&;h4#!2Gq+;d|-O7c}Ro`UY5E6s_NOd=bgy`GXs%kgZ%QSzyLlE{t`7lZr60>n?5lTg;Pljy!StQxCVO)#BG28GQ$Isdtyh!p)rv!s&zTl;@ZMzkrpqnwz(1Y zwKdjC4lBv~v|71VCIT>J;#Fu^wDiXk<_R*!Hoy6F;0KG-zAdogSm%9z+;zM*XWW3n zZqnx|1zjylz2NTdM>A3b!@~|F!O3wu#O}ieV*0seArtZUwv=FAYTN7g%MLm1xGkri z5)n`qn3h6`MNPaNwGA@SJ;fG28r+G2eeaezc8`VyKY$4SgjuBBT+#Uwj$&%PO#_^%F`izh>+6J)i=wiU)?hMkXNfu|Yd8@v>~oR%=-<^c9F&MLpMMJd%tcLCMd04(P8hZeSw+sm_62J48pL zre1b{CRTu{MN=cR1)-G=O8Km^!~rz07vi=v{BkX?-5hEz?|UTr&(K3tfi(LZ^G;F< zw~u_SP+ySac?m1jF}iXpgOiYpn1}lkdX+~(K6P;r1RTR~ArMW|U7w2idAH(6PI1~@ z4ySwc7(CF8?ui^iDk@k|MGUjUFHJOOyTkfjg8Q*SKWKJas>!BWW&9n>NA=c2v4|JK z)MfMEYAJXHbCHW`Xz-lW!z;vbI_iM+Zq!&thDebX0$S&}5){;|pWi`{I{4P2lQ_va zBGQA`?a`_o7D#@cLmPSXPj&a0A8}{P_y8vT$Xof99-0 zH_$5;$I1*OyMa57#Ic3LDx6NOih->W^V@SzNGF3n(y_u;^@bJ@mw&^I9hz&o` z%2W8!F4DyXph}0%LhUst8LvLU<()&%fX2Wpt{Rg3&mylfIrSjpz(XV?n~Ep992BL) zJk_*l$LP6-PrtVV%|u?_SbXgb>Cg~EkZ&5puW4@p{Sq)T);2ukEW(QN34D*SGb{+K zx=N1rGhz=~_V26%#YH*tHFWNFCYq+2+2cK!P@ zT{Z0JY5U`GV$5%F!KVtP7)Q(Hs@CVCW-pP^n>`QjbnnO25)xZ|bklF+5@U*NV?21Y zxLg>k27%3VI#KU#rAs`LQ-n$zXf`AxuIRGw_Tyc+{1Xs_(h`P~)^RHm9E;@$bC(+fV>uEv!f?kd55_$MXk@J3=Z~k>S>`_TO@G$YrO%94`+Z z7*`(*dgIG!^@n;XzOpzTB~UGJsyt#KpxIWHe+Khe6>mn+g8j&muW2lFvvMJmS$RrvYs>qFkiDsw1@KCm5?X_2(zHiRh9TBx~=*gmR*->ak~O*_zeQ{WEf5QU5wr z==st@0=k;4QYUF%SNG)1gET?q&)aU`n>7ZMO%#7cRpf+MxnNQPTWFE_wbwd2;sJJj zpq?p(f}pJ@_G(?5P0h^$fWCXnO^KU)7fZG;Dyf|Fvq!hUuW&P@1Iq)BTr4q1aMa{& z*;iQ;WcP`7gXDG^VJWCpMC(-^jnf@77^yx6aK-BwL(5GL{Qt9fe7xO5LIl(tPY#B& z4&_KgpI-7+gwPpC?a?9Xefyx?Uv7l+K=b$ZTuB!5GOQf5xs{L=;#ADmuUX1{d>bna zgV`W_HP!*hBsPCpo-t6M*An&=@I?~%g&4!TFqRw?HxtH-TJnQBwNCYbOmJ2|MJeww zE;Vhi^@xig;?pxocHwdAdq|m9%Zv}6XoQaL1Ef;2-z+r7mgyu!A#Y_c3<7DN#?jpK z(qAt$|Lo0m7HV|6f_Q)~4d@G0GhMoE92huKR|R2_iTC2^3P{$IwALh(iR8FB8@~bl z91b&@0>y)q!75t}seKw}nh@%MVu(paFRz4*+Mc zGKf3{^ngI>ixXYUOgqrBaX^Qr#%ZkUz)W_4g{qpG`=tGcGFA!Y&uq-^J|~j^(*w)f zr?@Q%i8U(74(82N1N##$Oa;>0KuT4Rh9;(lulkL3OR3uVFi5nxK%{_-?Q<8X!W1_m zVgyC^wZOzxgq#9^^QqI?!~BuD2Xs%pCS!ah2wPT(kXD$Uk%o`!?Ix?Kbw@TRZ&12I zQdr;An|7pP;!_pKUQX z0szo9UYJ{gv!E^c%84Rbb3^Ipf6^i!TS#bq*sc9#ZqZwrvlBP1MPcoXX+|W zH+?cT66E8A2HJ+D-7iSFZQbVKPbNNxS>%tswI(r)5~fpO7bc@t?N<1tgImcmrj4 z-7uP8HS`brlnU!%CgT{Y03#os3A2 zSUt%Og4TN{hlA%s^+mh_OVBSL++Ym5ct19_-^x1p*um&-zMBFp7K(-|14_>+(IiZ1 zsn|ZC0?dHG22Al#Ti#T6G}An@AoFdF7#XNlHm+f6%V+FB2Q2@LD_JK)_*?c3b!yEW z`LR|eAMv3gHi%iu+WViNOJjJwLDF@b`9Z$1yS~6ryce;N>CTWRU7%J$1w!JF0z73a zT`WsrzHN<-T|fUkH&{8{AQg6%K&#ZX9e0tO z`^)UTD2Ef0rEiIYgiG=@6>u-~DjLK&=SM6RU~-aZo6TIMba%-Ru+n!vlrliQoNN z6!Y+sIx5Z2!h3N}=flC<-%|XD^bWZQ|xHipw)w?{LdGk5kMK&+@1I}PcTWA|4_xMEX*W7Kd{7l0Kq$u z%Uwo*GfQisDbj9_Zry%7-twe2m?Ui-&-}-h`1_ zHm$q@ZK)WX4Y!uJE)6s!DmJkGhm9h{v|R80Kz^+?GlOhfg^)RTP*L`|?SjnD^vcT1 z&vZ~OGT<!u<7f0WY@xa5oi{=0wyh97xdvLv^}kS6 zj-xi(QHK`zu6o0sUezVzMhAxCe^Sxjxy0^Xt^?WHI7rr~jnX*SQYb>pl#(wj*mtzp= z;$PKji*-P-XsctzHVKP|gyX)cu*GQy;Y?MhD!}_t8e)@zEI$|rJ9jGu zVhfk5IhZlid5tr0f;;-Y`EdMb>T?-EzIS&HJK8Q0SFhIJN*&n_dxr^Uc3_$hm%wdiG@6R;#X1^}FEpG}EyOQS^^*=gusS8w8P)g;LdN-NPQHDYBLDmH zmNC&*1m~z;pv>7dext1H6D8)m8yFwjjmI?FtmP&Gd$H)pbZXeQ1Fffp51boW8IzXQ zr4-CUFFd1eO8RD)(M5d-|D1!RF2JoA(z--iEzZ0=os-SWP@qL%^a;)&%;bOI_E4uB z8gcPGf7e0X4C?NRf5`?#b`jE7%hX<9AkxQv@R?wI`$-TZw-@$?2^4Z{Hu}ToKEs|| z9ZxM`&#`Hw90^+|%6$R_3n9VO2%el>z-7(8*vn0F9C-4JHH2QuP2o$B%E^%HKb87r z^5fRNrxEALOxhA8IS8n}l%79ONjQEnfr7s(*#=a!=w^_@BF9Tb=wO3c1M*8fARiMz z=CT!BVGt3A*y5e+pR06@eC|u_2?qM~ue>XN`^+zdvRDJA&~^|F+5b zG4lK~oeT@Krh=w1c!ti=TnvUyPhGQ75WZzWx^dB-a?KI-s1l?-b-`S4!yCQ=m?Vur z9(%G;KrEc_cDhCW8&mE559P@*V_>KUv{~A)ic` zsA3~J|L@ZKjjjN+l`zTMa@9ytGVcbtruIKD6$l=(!h?XCJJCIK{ndeDKG_;=$c7-h z4^_PQSj#tb4cPP4beCTEREy<3C_%OPS;vhW1EL>`mK%m2->KJ!Uus3gAqk4(Kfq|& z$*9CN6|!2d>xHF=N3$X~zi9+a2FTPiHUeQ+JXh?o;Q0cH(b|IyyuzUZy=@)!@FH$& z`qpfo`dKhrQ=d5b_l)FmD8N~&j)zBkyqy_~J44El+1f1SaMB<2m)$N82wl8!Muyh3rDtO7pT*_^WK`VtnXVg^CI2xBWB5fD zK@OtgrSy3YLH#JX)%sgj{kC0i*ja_N#gTC(;3ni6Z3qKi$sVH(D(;Ds@rUQQq7bp+}p7cR_(SaV?yD?zK}ZBd~7QQ8Z?CtF>d1O3uc37PZ?R&R?hcl28K}; zTG{y@u(9*DtD`C!%3pDDr#H1o5ghN8gtP8qwYGeGEO%g&^p@#C-K+Rcp6DnYZ}mkU z@{-n~NXZraay`6lNP?oTQR(p8Yscbg{Rj*I;OA5?s*wib!knJArk z6a@O*(C^G%`b0a|h9N1I3Lm&pN9HD9fCGcd!?bgKC}6uZ%O5JHn5%kYXP|=P;8>wW z;jgrNxv}!DVfqIouEW{znpSj;DnNW2(r&W0^kGwxga>vs!bx^@y1;w}Kf0T` zr>(q1*ztOWDedN#?*~c9QL7v58z@W@++aHzzjoO&285VVYAa!MYcFlH+AXn+KRx#b zy9Ar<+zcm6!R_$PO2$*22@3;IoJH-wbs+5Y_f)y&+C3ph5muB06at_#-9IZ|nUFlL zat_B6(#dM3RPEAB$#w{WvSTtbl0;I1@-ssM@cHNAl&7rJ(#%^2GDK z5-%(a+|wu>bQT!{W|r}a;c=s}(hs~Aj2Vhn_d9#<8%3MX00Gy{hVbbabv6+{sG}5@ zR7E;`d2NDFUUK_#P(2QJ8JAY-6=AI`%*K{^Y&!5p*G*optqf zer2s3daEm7~rhVs^p%rv<=Z@%&fr3 zMN462^v@{U-2w2Kb5^^?u#EKax~&ecj%4DV2B=T*Mke~J-ZwelA|a6%hqUp^_CK+* z+OXrbv$89t$N1Zj)tNo6!+l-JHk_Ha6fJA%PsO0$;cdi}*s062grS5czWLx1R#zbf zgd%Za#8dZCr`SBB7#?F*WpXAS=L_?-n#&XWAJ+hO5Rv+vsZP~^(;|k=@}$XfDXEa= z!;DTt3hHH4VxkV_(3$;old+5*?I1Le%CdFTTQxrOKl6J0kT$!%n2=+qmHAV{eo$kug;zRJ&Tm6jBq(w z@RsXO`Oqm-835NshLl{1NOxX1pBD7x?n<@=_g4pBe94U0q#uKLk~{I$gL& zRQVjU_@0KzdT3GXB{ZqvduZ$gn~2NL>4`hc#x#z}zY^!Qr4<$*TU+mMr~iw+ktN^{ zF!HmSG=$zTtoyB%mKFSucai@*w#&=JrEuXcl=eBkkh36`|{Kjx&dNoD!*fEHVdnk%TAo|?bfrM3U zt(kYjMZD)-C{p00J%&>WJ%;`>1FZo|fpI0{6T#ZBGV?G*(WgjDH+}(87;?bYY(W0T zN%4a~ul@D4n#O0XxTO>Vo;PYexPSLl_so|YS0&l#cJ)4KUCl*Cs1=F`(CKMtMs1i^ zu&IV8T&CJ$hnKt8chLhlu8hTBUfU&y`XPEqnkpB!v+cu+BrJ zWZoi5%ky%@awCSE_*y*ecKQWDp!%~kNgeJNnn?;V|N9_W4(<5+gyX3uFspq3c5?V1 z-@c2=+IQZ+-CxMA)I%FAveoMF2SndhB9d=S@l^W5Mg3}ddag7oNfbZ9?)xi2bS<93 zo+RtWuc5ZJ8OrlybQMLyugA{sz&B)Fo_!N|P+4z{*+qYQa+sB*m%hxjGtRN(Y=@KH z#a`kC6lpZK_-R8RAJV1nK(E(;b5H)+ob)q=-eQ>Vr5I;ru{#=SZo~T*I;&td&+j}+ z62#hEO_(Rg-LHY+Fx29?9vC?N#A3^i4T8EMqq=OelvioGD@V~cBEK#)4s<^LPjO#; zl){ku+(T!tSc?%6g8R0up7~`M*<$QL0i@LQn_Fg z+R4X4?hVQYQqrWRHVwtB9e04RN#TVo%V#cyE*)ti{EZHK&P>*41qk+9i2!1KYW5Tp z9C9-BOfmA!*u&unIK()9rRE83liGw!1tB)1m%hYkEew8Maw6qnm-2Tfw`hx&39p&; ze<2-QhOHdUf5yptm{|)!P{Y5y!L0E8`nixBjmV9&=>wLIs@ty{T%BHFi%z2jF3I4r zkB|$;ONY>ge+YqBNK@DN8q^vmwMjElDFA|9DPHXvO~R^EkG)VwB5i2Zt^BPhIq0lZWBD1piFIP5Vk#J}e(8{$4L1o)@h6<58<9@bufw0= zvakzJdSP@O%y&c~#paG03g+Xdrwr=~&q@H3D({{<+t5)$q1ANC%$s>G99Nq4CO0y3 zg;%$q$ikk;YILTL12(cWX1ObF<8D=>-mLDMx+u}QAiaLXG2SlTuCdL6-Y6=8Q`NeK zlq&7Wt>L)c`n=7}?Wk@5C_?;W&6mTA-LAn6m#*CV?^8Zdaz>7SHT^|nD5}H-BzBzH zd3tadSz>nhBH`NwHg-r~l8C!jeAJ^N(oMr{GpD5|19J4;$|y6Za$5Z+o_)Dlx(L^K zS=G}~ohg&3J2mW~?5xcq(C*@t>u1$h6N{iHTzr~dRx{FC4djzDMdko+-8)5ZgB-2w zCCq+k^?NUF{vv&6;8)eix;7+yep0OIB%jAW>lN(Vw#33oxq^(=IEU4TUjvh%rD*xC zLAWP*DhBLST92vbKoleOv2nvnnbnl`mJ9z_IBS}%5`XQU(vC$3d-Zc;TuM!jIq9e) z_EvsWJ+bt6#8c!W6%*vlfeEgm+pMB|K)CNBMCN20^k<>eh7aYui+`(1q)H7ych61G}SaryKt3n;r!(#J7E6BAJp zJnK|2I-AazivvaGc9I>*sm^mZJ>OuKBa#B@*b3A(nbqukt?JC-9~=s&&L7RnuylTe zQUIy?9OGCs#9?A9blWQ2y+$f15l}yk6Sxfvn8MzAX-;N@tSm7!YBq?y*5aoAhB`0} zJ))xvv5%p2OyK)w?V9znVz4gkShW!CFRYIz;{k$?yY_U|r`GB-)i>ipx(?;m zI2uY44W%aSHl?ko8ANQC`}Q)CjV$el(gma@NeKKW26zU}JWRGf1eTo8RxdSe+?x>f zi^ZarH>!yCMDwTUhkO&lad6GSge#>#73|1DE|;Qx6D_0-En)tE&=b`4$RHImg{-jNDo<3a)@VRxFvxni3OI&ZJDELH%kceHwRP7d{?F zH$sJGCzJmxdd6!LECTVq8Y-mUzma)hXFlLWf>6Xx${4W^?$Igbx-8>MFlUdEW^vla zWrRL5tPhJ$UN@E|9TCM#T*x|X#I--vkAEO0wTdxZ^9sVMLWsx+m(Uq}Q8nMc@Uana zW1#H|#vfAp*_ie38$br@RoK*d^Q--){_K~e=@Z`j(74h{jJn;Hw zcA(K>9{@?Gn%z#(*~}tbm4oQ8U|~{muP)VXs&rn`Y;L%_q8y}ocq=wJ4Bitr4Io@& z_X`iGogI=a_b;#BEthKrJCMc~US%Eq3zS^dgil_~Zm_t^1Wl)XW}A5haltF{6O;Z0 z{I;`~re7EMgpt!sZQk=T&e|*D6niOq5sCr!U*h+MRT(zpP*hFcFKsL(hBmY-&Xx2O zm7&l3-sO?3-;MmsbKM&#)84qH0(LSkJ$2JwgXDwO5s~2{#Jy&7js21(decI`l%(RD z?Hv%VDBvg~06qS;1t{V;qPL8f$Ur|$;eL4YY)_HMu`bgEO{b0WEz&F_qH}beum6sP zZ){jsLNhiH;-G*d|D#h;4o_9n2FW%KjSsdSGADyiJiyGPo-9F$;f}RZANS5<1GeSI zub79{V1@>AqUh^ zLu8S+!wvF3RhC+c+c38bzp1PEy3(lgfEP*?-9Ecpix)L@8+oiNjkZ`xA zz65e2Y-i_X3xb(MGUq?wyS;dTc&DU;T37o+)jJ@LO`8~lq7me}9H_|4SpUuetBF(I zr4UoZ^PqX(*;pc>o-XyIn`A>(nvlQUNLsiIN24v#uvRYF5z~$QIc$^(k%#0Q@9z6deMq~T)A_&SsMi%pxhxDMK3mNL1#w6mN}(*XRf>Uc>4wKb z4U6)1oMYDJ<*{|Gt-vW}Qo_MeHZ`2_^dgHNyP@B@bz{`Z+McYZ2?i{!Xia|9Ft=`C zNiv3OpIWUFNQ>1?aXX=4O*s7>8(d{(%@g`JFRGn#_t&&>oA;0)N7b@&r=P(v4ytG* z!g`vXVe9d>7*O%B1HY@&FVK3ZSq;P&Z5@8Qlq_RGhL1R>(e%NZN@cZbw#LP$Vr)*$ zQ*5b+bYnm>=rB2LW{P2D&fNVRuW%`TBOM5bl!7r1H}S6^H#9a@z`^>6wrflO=o44`4kr{eja5^ZT^} zl;RT?&Q{I#U|qh;%YemJ+Nh>L$#3n*s;D-A-Ys%|TF(3p>C}*!9vPa`HD;72x;oysUcEfXiZEr_-iG~%Rxv*P+SqY$l` zWT_o*p5mJa0h@=2tZZ&ZeA=8aR#s0ywio}!9Do1E3;fu^|OP24}m)(Ac%c_okFJzgz&Dd&R-DmIdR+TJhzi4M5j<8O*A`|G1IgBa5G9D1ZlW|)=Gso?mp zNHzgM>PE;L>=8$I15u>F4<5z*MM*v_$|apiFKx6m_^)1G2|gDIT+ZcuLJ48W4}e@O z@SmZnT3lwKkZ&7Wy-JwIoys(lozs*Bff@LxEhQAIA8doxc{>LEasKz=lrs{#gPbz+ z7&B4J363dQ4wP*z@}=C|?9@z2rLRE*%_63IpH=h+V*^KAkRtyjQK19Ff2j52^bRoz z{o-DwWL+L80%o^uPsvt5HUd_cJm*$)sdqR4w*t*TAG8r|T z4n)XFd*hQalL#EZl2D&jc*AC=l4hX8hP%tuLI--RqaDa1-%Mws zJK!z>D{IAzdZ3eIu0_5VeUl;LtcZ<3sUTW0r7#w@z4ECWDRyz|97f+WxM@)AA)WM5 z_7=&%GUYAo#-zy{ZdR0x3%AA=|8QRYL!k2%dr9+# zL9KSsLsM+}pke#eZXTPe;naEvENc7U#MIrIqwIBem; zR$l~_USx<2_*d~ySdM824|AQr)PC<9e7oIS`(N0%h1+(uB&JJ1QHV@4lJ%n34QhjC zy&-Tbi-B?)6Io3{kq+MJjIg|6WUD13fRwi(&|%C!YKf~PLZ;$GG{U^VT`LkYeI)*wuOcjY`6?DG_Go@*^~CO}IJGmg%J(L@iie4d z!6{Lw^J~y0hfat$G5o+U%^uzC?)L!~u>W1v=h($)-+{ypHE4uHF4?-mBcCl&+Y zT9dX{)92-~zxl(3VYt7DQ<5>u>^RWQ^3THSA%WOF34X~82EMo4CiJIN)|_UJ^VB$Y zfwt}_3TU`H5ZyR!qG#)@6#c{_a*pnRbxIYXBp1{?7u4L>Vz2*=(S3nDmm_e4LDzXm zntgyRrLOB$Ago!sdd&+^`7%G6U#l0G@4GHss&`_@+hV^RqWlyoc_Dp_ee^OX&XJhj zGGMZEdr02Vbz`XTq|Xb6=_4a9_{jbl!Q_jGVDba(h*vF&5xu71Qltacb0=(@o?L%~ zAy)3C}K}#Mp@+zXPXp zrW7|i&n|+1dJxY@k4(Gca*%RKubsZ+&6-~A;mkq$)7f<#Y-&O zP*#RsA>|>4h1=nDdY*uIxQe=|%w3LI`^U%JJQHgBDhNV4_=1u9az5deR0HVhl~6UY zj8~mO3(17Qrn|AO3-e;2`jP8(9Q4f(u?KD1FKDZ$Zc2Z1lv{0ru@it+DhU*G!Ts9U zpKtxQxY<;FRDO2*QzvTT)6oeQe#Z-e(>#DcGBW!d@<`S(8u%|WxymkBSol|#do?M5 zzpb1l#Wx`ID=$=M%9;mPoD(S-8JpcR)Lac#y%pvO*Cx&XjOIx!bCFY7&E7MULQWe@ zVFWhfdaqXwcA_%Pa&CA8^NUrh&6EqD!^QVQJ1oD9iiB{(_(k-wU+C!nXX}|$U7kSw zJ(=AaD+)?yC{40#u`7nvG3O3F?4c8!$JrzuKE;j}+8LdO7a4_ntt5{%J#|3<8xB$` z_;=&pmT_4}DrevHT)S<17XX05Mw&XGPZuQZSDv4wi)baq8Wp!!O*8SGzR7JsMB$n2 zhj0VQH-sDtNeXBZkrs39H3qYHw}2Yyjz)gu_ocKsXwb-S2=L_4>W*OuH9M zC0QFzJn}|Vrq}a0z-ooRyFPYqHnbZf;JXS_W|KYsG3sU%%z!>jg<{r_zy;=QJX2;v zN1IgV9ZJc=kq)}SZZqZEK|4Xhbu{>n}7GZK6A zGTH9H3$-3W_)A^&VxV;Lv$=S@hVPbVktw_btDqriAG8luzhzDM^fYBBwg4FP{MwSh zvyz^cRVS15_dGx^`Qp3JOlfrFcYMqARF)#qA(vw~!31QO0;pYT&ZJ3RXCTSd#v_aV z1glemQ1=No1Vm^}N{xfs7tj~h%3yAS=N>o+-&I&Ez05_VVePvLhRo*-5s%^Jq(F~? zv&3!o6k~V09*%R1Dd-J8*#Z|ZzIs0 zmj&5i^klo!r-623B$UmoT_Lb&A{g2=%p4F^$X*|A1&_JEP%$ve0ZaY@P!mTYtu6hL z^Z}xBybmlgvzS`R-1=65-O1UbIgQhKo--X2K^Lf9gRT4^OjRa^E70-AfwV5g{q$kU zFCz3It264yvhwbj>M?3^yd?M~Zl96Ww@;lt*Ef#x;^4wl-o&%!^ctyJ`)D(*8j8FS z{J(9^PmS_4-8Au@J2f~CB}VNnH1YV*pL&Ue=JsHE+!{E98&LJ2hJLW88H1&+FDoDc z98UhLoI^+?t^Tw*O9KBRmkn4XSh%BY3~WM_ATR!m)RlEI8yeUj$mj(Cj|$Jj=7Ood z>%&LMB1hD<(#Pc>KHgn6_KP>wP5oz+B2KXX3CK(9O?0hf@_Jna6YTGZ(%EL4XzN7X zH1c+3tzK2$V)L2{)-x@Ck|YK!8h{A+DkE(w^A*>au0`LB#NGrF zYE`&6+jQE6kynS82}O@{YaU8~(FL~;jF(1QxPz6~QBKpW>){|#M>wl_`5^on-4C9? z^niv|uHSvo#M8leX@NE=PBe*)%lSa)e4R6@f588eLH}q)*ZZk(EIyRyQ4c(9LRvZ7 zrYJZ5Q#hRlV&5_NyZkBD6QCd}|9HQd@>(59@XJ}eU;2ZoD}Dp7@yv_h$Xh?Cs+4d^ z0mA@1|69&wOpot*>ICCML8`YE#Z>ut*V5E)(KO*(QBpb><->EUdIV%y;i%Tl;aa+( zd(G*3D9%$j=wwuGrQnmcTNp)`h;iG1j`bVHKYtpx^>NEmtHSu`v&|TCjFvaS(ft39X`;jW)njNZ5Mm; literal 0 HcmV?d00001 diff --git a/prompt-criteria.txt b/prompt-criteria.txt new file mode 100644 index 0000000..9abfbfc --- /dev/null +++ b/prompt-criteria.txt @@ -0,0 +1,29 @@ +**Papers MUST:** + +1. **Focus primarily on the engineering, design, or optimization of prompts *specifically* for Large Language Models (LLMs).** (This clarifies that the focus is prompt engineering *for LLMs*, not just any AI system.) +2. **Investigate, analyze, or propose methods for improving LLM performance *through the manipulation of textual input prompts*.** (This emphasizes the "how" of prompt engineering and excludes papers that might just mention prompts in passing.) +3. **Provide concrete examples of prompts and demonstrate their impact on LLM output, replicable with publicly available LLMs.** (This maintains the practical, replicable aspect while emphasizing the prompt-output relationship.) + +**Papers MUST NOT:** + +1. **Focus primarily on the development of new LLM architectures or training methods.** (This explicitly excludes papers about building or training LLMs, focusing the criteria on using them.) +2. **Be primarily concerned with applications of generative AI *other than text generation driven by LLMs*, such as image, video, or audio generation.** (This clearly differentiates LLMs from other generative models like text-to-image or text-to-video.) +3. **Be primarily concerned with medical, automotive (self-driving), or ethical subjects.** (This exclusion remains, but is lower priority given the more specific focus.) + +**Additional Instructions:** + +* **The core subject of the paper must be prompt engineering for text-based interactions with LLMs. Papers that mention prompts but do not make them the central focus should be rejected.** (This emphasizes the primary importance of prompt engineering.) +* **Reject papers that focus on using LLMs as components within larger systems where prompt engineering is not the primary concern (e.g., using an LLM as part of a no-code platform or within a multi-agent system).** (This addresses the issue encountered with the example abstract.) +* **Favor papers that explore novel prompt engineering techniques, provide comparative analyses of different prompting strategies, or offer frameworks for systematic prompt development.** (This provides guidance on what constitutes a *strong* accept.) +* **Analyze each paper's title and abstract carefully to determine how many criteria are met. A paper must meet all of the "MUST" criteria to be considered for acceptance.** (This further clarifies the requirements for acceptance, as the previous requirement of "two or three" is no longer applicable.) +* **Err on the side of caution. When in doubt, reject.** (This rule makes the selection more exclusive, suitable for cases where there might be too many submissions to process effectively.) + +**Example of Clear Rejection:** + +* A paper about a new No-Code platform that utilizes LLMs as one of its components would be rejected, even if it mentions prompts, because the primary focus is the platform, not prompt engineering. +* A paper that details a new method for fine-tuning LLMs would be rejected as the criteria is focused on prompt engineering and not new training methods. + +**Example of Clear Acceptance:** + +* A paper that presents a new technique for automatically generating prompts that elicit specific types of responses from an LLM, providing detailed examples and comparisons, would be accepted. +* A paper comparing the effectiveness of various prompt structures (e.g., zero-shot, few-shot, chain-of-thought) for a particular task, offering insights into optimal prompt design, would be accepted. diff --git a/prompt-papers-old.md b/prompt-papers-old.md new file mode 100644 index 0000000..34d4b6e --- /dev/null +++ b/prompt-papers-old.md @@ -0,0 +1,7421 @@ +# Accepted Papers + +## [ICPC: In-context Prompt Compression with Faster Inference](https://arxiv.org/abs/http://arxiv.org/abs/2501.01625v1) +**arXiv ID:** http://arxiv.org/abs/2501.01625v1 + +**Abstract:** +> Despite the recent success of Large Language Models (LLMs), it remains +> challenging to feed LLMs with long prompts due to the fixed size of LLM inputs. +> As a remedy, prompt compression becomes a promising solution by removing +> redundant tokens in the prompt. However, using LLM in the existing works +> requires additional computation resources and leads to memory overheads. To +> address it, we propose ICPC (In-context Prompt Compression), a novel and +> scalable prompt compression method that adaptively reduces the prompt length. +> The key idea of ICPC is to calculate the probability of each word appearing in +> the prompt using encoders and calculate information carried by each word +> through the information function, which effectively reduces the information +> loss during prompt compression and increases the speed of compression. +> Empirically, we demonstrate that ICPC can effectively compress long texts of +> different categories and thus achieve better performance and speed on different +> types of NLP tasks. + +**Decision Explanation:** +Original response: +{ + "decision": "ACCEPT", + "explanation": "The paper meets all 'MUST' criteria: it focuses primarily on prompt engineering for LLMs (in-context prompt compression), investigates a method to improve LLM performance through prompt manipulation (reducing information loss during compression), and likely provides concrete examples (though not explicitly stated in the abstract, implied through 'empirically demonstrate' on various NLP tasks)." +} + +--- + +## [Adaptive Few-shot Prompting for Machine Translation with Pre-trained + Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01679v1) +**arXiv ID:** http://arxiv.org/abs/2501.01679v1 + +**Abstract:** +> Recently, Large language models (LLMs) with in-context learning have +> demonstrated remarkable potential in handling neural machine translation. +> However, existing evidence shows that LLMs are prompt-sensitive and it is +> sub-optimal to apply the fixed prompt to any input for downstream machine +> translation tasks. To address this issue, we propose an adaptive few-shot +> prompting (AFSP) framework to automatically select suitable translation +> demonstrations for various source input sentences to further elicit the +> translation capability of an LLM for better machine translation. First, we +> build a translation demonstration retrieval module based on LLM's embedding to +> retrieve top-k semantic-similar translation demonstrations from aligned +> parallel translation corpus. Rather than using other embedding models for +> semantic demonstration retrieval, we build a hybrid demonstration retrieval +> module based on the embedding layer of the deployed LLM to build better input +> representation for retrieving more semantic-related translation demonstrations. +> Then, to ensure better semantic consistency between source inputs and target +> outputs, we force the deployed LLM itself to generate multiple output +> candidates in the target language with the help of translation demonstrations +> and rerank these candidates. Besides, to better evaluate the effectiveness of +> our AFSP framework on the latest language and extend the research boundary of +> neural machine translation, we construct a high-quality diplomatic +> Chinese-English parallel dataset that consists of 5,528 parallel +> Chinese-English sentences. Finally, extensive experiments on the proposed +> diplomatic Chinese-English parallel dataset and the United Nations Parallel +> Corpus (Chinese-English part) show the effectiveness and superiority of our +> proposed AFSP. + +**Decision Explanation:** +Original decision: ACCEPT +The paper meets all 'MUST' criteria: it focuses on prompt engineering for LLMs (adaptive few-shot prompting for machine translation), investigates improving LLM performance through prompt manipulation, and provides concrete examples with a replicable framework using publicly available LLMs, as evidenced by experiments on constructed and existing datasets. + +--- + +## [Table as Thought: Exploring Structured Thoughts in LLM Reasoning](https://arxiv.org/abs/http://arxiv.org/abs/2501.02152v1) +**arXiv ID:** http://arxiv.org/abs/2501.02152v1 + +**Abstract:** +> Large language models' reasoning abilities benefit from methods that organize +> their thought processes, such as chain-of-thought prompting, which employs a +> sequential structure to guide the reasoning process step-by-step. However, +> existing approaches focus primarily on organizing the sequence of thoughts, +> leaving structure in individual thought steps underexplored. To address this +> gap, we propose Table as Thought, a framework inspired by cognitive +> neuroscience theories on human thought. Table as Thought organizes reasoning +> within a tabular schema, where rows represent sequential thought steps and +> columns capture critical constraints and contextual information to enhance +> reasoning. The reasoning process iteratively populates the table until +> self-verification ensures completeness and correctness. Our experiments show +> that Table as Thought excels in planning tasks and demonstrates a strong +> potential for enhancing LLM performance in mathematical reasoning compared to +> unstructured thought baselines. This work provides a novel exploration of +> refining thought representation within LLMs, paving the way for advancements in +> reasoning and AI cognition. + +**Decision Explanation:** +Original decision: ACCEPT +The paper focuses primarily on prompt engineering for LLMs by proposing a novel 'Table as Thought' framework, which organizes reasoning within a tabular schema to enhance LLM performance. It provides concrete examples (planning and mathematical reasoning tasks) and demonstrates the impact of this structured prompt approach on LLM output, meeting all required criteria. + +--- + +## [The Prompt Alchemist: Automated LLM-Tailored Prompt Optimization for + Test Case Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01329v1) +**arXiv ID:** http://arxiv.org/abs/2501.01329v1 + +**Abstract:** +> Test cases are essential for validating the reliability and quality of +> software applications. Recent studies have demonstrated the capability of Large +> Language Models (LLMs) to generate useful test cases for given source code. +> However, the existing work primarily relies on human-written plain prompts, +> which often leads to suboptimal results since the performance of LLMs can be +> highly influenced by the prompts. Moreover, these approaches use the same +> prompt for all LLMs, overlooking the fact that different LLMs might be best +> suited to different prompts. Given the wide variety of possible prompt +> formulations, automatically discovering the optimal prompt for each LLM +> presents a significant challenge. Although there are methods on automated +> prompt optimization in the natural language processing field, they are hard to +> produce effective prompts for the test case generation task. First, the methods +> iteratively optimize prompts by simply combining and mutating existing ones +> without proper guidance, resulting in prompts that lack diversity and tend to +> repeat the same errors in the generated test cases. Second, the prompts are +> generally lack of domain contextual knowledge, limiting LLMs' performance in +> the task. + +**Decision Explanation:** +Original decision: ACCEPT +The paper meets all 'MUST' criteria, focusing primarily on prompt engineering for LLMs, investigating automated optimization of textual input prompts to improve LLM performance in test case generation, and likely providing concrete examples replicable with publicly available LLMs, as implied by its title and abstract. + +--- + +## [Recursive Decomposition of Logical Thoughts: Framework for Superior + Reasoning and Knowledge Propagation in Large Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02026v1) +**arXiv ID:** http://arxiv.org/abs/2501.02026v1 + +**Abstract:** +> Enhancing the reasoning capabilities of Large Language Models remains a +> critical challenge in artificial intelligence. We introduce RDoLT, Recursive +> Decomposition of Logical Thought prompting, a novel framework that +> significantly boosts LLM reasoning performance. RDoLT is built on three key +> innovations: (1) recursively breaking down complex reasoning tasks into +> sub-tasks of progressive complexity; (2) employing an advanced selection and +> scoring mechanism to identify the most promising reasoning thoughts; and (3) +> integrating a knowledge propagation module that mimics human learning by +> keeping track of strong and weak thoughts for information propagation. Our +> approach was evaluated across multiple benchmarks, including GSM8K, SVAMP, +> MultiArith, LastLetterConcatenation, and Gaokao2023 Math. The results +> demonstrate that RDoLT consistently outperforms existing state-of-the-art +> techniques, achieving a 90.98 percent accuracy on GSM8K with ChatGPT-4, +> surpassing state-of-the-art techniques by 6.28 percent. Similar improvements +> were observed on other benchmarks, with accuracy gains ranging from 5.5 percent +> to 6.75 percent. These findings highlight RDoLT's potential to advance prompt +> engineering, offering a more effective and generalizable approach to complex +> reasoning tasks. + +**Decision Explanation:** +Original decision: ACCEPT +Meets all 'MUST' criteria: (1) Focuses on prompt engineering for LLMs, (2) Investigates a novel method (RDoLT) to improve LLM performance through prompt manipulation, and (3) Provides concrete examples and benchmark results demonstrating the impact of RDoLT on LLM output, replicable with publicly available LLMs (ChatGPT-4). + +--- + +# Rejected Papers + +## [Beyond Text: Implementing Multimodal Large Language Model-Powered + Multi-Agent Systems Using a No-Code Platform](https://arxiv.org/abs/http://arxiv.org/abs/2501.00750v1) +**arXiv ID:** http://arxiv.org/abs/2501.00750v1 + +**Abstract:** +> This study proposes the design and implementation of a multimodal LLM-based +> Multi-Agent System (MAS) leveraging a No-Code platform to address the practical +> constraints and significant entry barriers associated with AI adoption in +> enterprises. Advanced AI technologies, such as Large Language Models (LLMs), +> often pose challenges due to their technical complexity and high implementation +> costs, making them difficult for many organizations to adopt. To overcome these +> limitations, this research develops a No-Code-based Multi-Agent System designed +> to enable users without programming knowledge to easily build and manage AI +> systems. The study examines various use cases to validate the applicability of +> AI in business processes, including code generation from image-based notes, +> Advanced RAG-based question-answering systems, text-based image generation, and +> video generation using images and prompts. These systems lower the barriers to +> AI adoption, empowering not only professional developers but also general users +> to harness AI for significantly improved productivity and efficiency. By +> demonstrating the scalability and accessibility of No-Code platforms, this +> study advances the democratization of AI technologies within enterprises and +> validates the practical applicability of Multi-Agent Systems, ultimately +> contributing to the widespread adoption of AI across various industries. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on implementing a No-Code platform for Multi-Agent Systems using LLMs, not specifically on prompt engineering for text-based interactions with LLMs, failing to meet the core subject requirement. Prompt engineering is mentioned (e.g., using images and prompts for video generation) but is not the central focus. + +--- + +## [A3: Android Agent Arena for Mobile GUI Agents](https://arxiv.org/abs/http://arxiv.org/abs/2501.01149v1) +**arXiv ID:** http://arxiv.org/abs/2501.01149v1 + +**Abstract:** +> AI agents have become increasingly prevalent in recent years, driven by +> significant advancements in the field of large language models (LLMs). Mobile +> GUI agents, a subset of AI agents, are designed to autonomously perform tasks +> on mobile devices. While numerous studies have introduced agents, datasets, and +> benchmarks to advance mobile GUI agent research, many existing datasets focus +> on static frame evaluations and fail to provide a comprehensive platform for +> assessing performance on real-world, in-the-wild tasks. To address this gap, we +> present Android Agent Arena (A3), a novel evaluation platform. Unlike existing +> in-the-wild systems, A3 offers: (1) meaningful and practical tasks, such as +> real-time online information retrieval and operational instructions; (2) a +> larger, more flexible action space, enabling compatibility with agents trained +> on any dataset; and (3) automated business-level LLM-based evaluation process. +> A3 includes 21 widely used general third-party apps and 201 tasks +> representative of common user scenarios, providing a robust foundation for +> evaluating mobile GUI agents in real-world situations and a new autonomous +> evaluation process for less human labor and coding expertise. The project is +> available at \url{https://yuxiangchai.github.io/Android-Agent-Arena/}. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a novel evaluation platform (Android Agent Arena) for mobile GUI agents, with LLMs mentioned only as part of the automated evaluation process, not as the central subject for prompt engineering. The core subject does not meet the 'MUST' criteria of focusing primarily on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [Rethinking Relation Extraction: Beyond Shortcuts to Generalization with + a Debiased Benchmark](https://arxiv.org/abs/http://arxiv.org/abs/2501.01349v1) +**arXiv ID:** http://arxiv.org/abs/2501.01349v1 + +**Abstract:** +> Benchmarks are crucial for evaluating machine learning algorithm performance, +> facilitating comparison and identifying superior solutions. However, biases +> within datasets can lead models to learn shortcut patterns, resulting in +> inaccurate assessments and hindering real-world applicability. This paper +> addresses the issue of entity bias in relation extraction tasks, where models +> tend to rely on entity mentions rather than context. We propose a debiased +> relation extraction benchmark DREB that breaks the pseudo-correlation between +> entity mentions and relation types through entity replacement. DREB utilizes +> Bias Evaluator and PPL Evaluator to ensure low bias and high naturalness, +> providing a reliable and accurate assessment of model generalization in entity +> bias scenarios. To establish a new baseline on DREB, we introduce MixDebias, a +> debiasing method combining data-level and model training-level techniques. +> MixDebias effectively improves model performance on DREB while maintaining +> performance on the original dataset. Extensive experiments demonstrate the +> effectiveness and robustness of MixDebias compared to existing methods, +> highlighting its potential for improving the generalization ability of relation +> extraction models. We will release DREB and MixDebias publicly. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on debiasing relation extraction benchmarks and proposing a new debiasing method (MixDebias), rather than primarily investigating prompt engineering, optimizing textual input prompts for Large Language Models (LLMs), or demonstrating the impact of prompts on LLM output. + +--- + +## [ASKCOS: an open source software suite for synthesis planning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01835v1) +**arXiv ID:** http://arxiv.org/abs/2501.01835v1 + +**Abstract:** +> The advancement of machine learning and the availability of large-scale +> reaction datasets have accelerated the development of data-driven models for +> computer-aided synthesis planning (CASP) in the past decade. Here, we detail +> the newest version of ASKCOS, an open source software suite for synthesis +> planning that makes available several research advances in a freely available, +> practical tool. Four one-step retrosynthesis models form the basis of both +> interactive planning and automatic planning modes. Retrosynthetic planning is +> complemented by other modules for feasibility assessment and pathway +> evaluation, including reaction condition recommendation, reaction outcome +> prediction, and auxiliary capabilities such as solubility prediction and +> quantum mechanical descriptor prediction. ASKCOS has assisted hundreds of +> medicinal, synthetic, and process chemists in their day-to-day tasks, +> complementing expert decision making. It is our belief that CASP tools like +> ASKCOS are an important part of modern chemistry research, and that they offer +> ever-increasing utility and accessibility. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of an open source software suite for synthesis planning in chemistry, with no evident focus on prompt engineering for Large Language Models (LLMs) or improving LLM performance through textual input prompt manipulation. + +--- + +## [eRevise+RF: A Writing Evaluation System for Assessing Student Essay + Revisions and Providing Formative Feedback](https://arxiv.org/abs/http://arxiv.org/abs/2501.00715v1) +**arXiv ID:** http://arxiv.org/abs/2501.00715v1 + +**Abstract:** +> The ability to revise essays in response to feedback is important for +> students' writing success. An automated writing evaluation (AWE) system that +> supports students in revising their essays is thus essential. We present +> eRevise+RF, an enhanced AWE system for assessing student essay revisions (e.g., +> changes made to an essay to improve its quality in response to essay feedback) +> and providing revision feedback. We deployed the system with 6 teachers and 406 +> students across 3 schools in Pennsylvania and Louisiana. The results confirmed +> its effectiveness in (1) assessing student essays in terms of evidence usage, +> (2) extracting evidence and reasoning revisions across essays, and (3) +> determining revision success in responding to feedback. The evaluation also +> suggested eRevise+RF is a helpful system for young students to improve their +> argumentative writing skills through revision and formative feedback. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development and evaluation of an automated writing evaluation (AWE) system for assessing student essay revisions, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not demonstrate the impact of textual input prompts on LLM output. + +--- + +## [Towards End-to-End Neuromorphic Voxel-based 3D Object Reconstruction + Without Physical Priors](https://arxiv.org/abs/http://arxiv.org/abs/2501.00741v1) +**arXiv ID:** http://arxiv.org/abs/2501.00741v1 + +**Abstract:** +> Neuromorphic cameras, also known as event cameras, are asynchronous +> brightness-change sensors that can capture extremely fast motion without +> suffering from motion blur, making them particularly promising for 3D +> reconstruction in extreme environments. However, existing research on 3D +> reconstruction using monocular neuromorphic cameras is limited, and most of the +> methods rely on estimating physical priors and employ complex multi-step +> pipelines. In this work, we propose an end-to-end method for dense voxel 3D +> reconstruction using neuromorphic cameras that eliminates the need to estimate +> physical priors. Our method incorporates a novel event representation to +> enhance edge features, enabling the proposed feature-enhancement model to learn +> more effectively. Additionally, we introduced Optimal Binarization Threshold +> Selection Principle as a guideline for future related work, using the optimal +> reconstruction results achieved with threshold optimization as the benchmark. +> Our method achieves a 54.6% improvement in reconstruction accuracy compared to +> the baseline method. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on 3D object reconstruction using neuromorphic cameras, with no apparent connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria." +} + +--- + +## [AttriReBoost: A Gradient-Free Propagation Optimization Method for Cold + Start Mitigation in Attribute Missing Graphs](https://arxiv.org/abs/http://arxiv.org/abs/2501.00743v1) +**arXiv ID:** http://arxiv.org/abs/2501.00743v1 + +**Abstract:** +> Missing attribute issues are prevalent in the graph learning, leading to +> biased outcomes in Graph Neural Networks (GNNs). Existing methods that rely on +> feature propagation are prone to cold start problem, particularly when dealing +> with attribute resetting and low-degree nodes, which hinder effective +> propagation and convergence. To address these challenges, we propose +> AttriReBoost (ARB), a novel method that incorporates propagation-based method +> to mitigate cold start problems in attribute-missing graphs. ARB enhances +> global feature propagation by redefining initial boundary conditions and +> strategically integrating virtual edges, thereby improving node connectivity +> and ensuring more stable and efficient convergence. This method facilitates +> gradient-free attribute reconstruction with lower computational overhead. The +> proposed method is theoretically grounded, with its convergence rigorously +> established. Extensive experiments on several real-world benchmark datasets +> demonstrate the effectiveness of ARB, achieving an average accuracy improvement +> of 5.11% over state-of-the-art methods. Additionally, ARB exhibits remarkable +> computational efficiency, processing a large-scale graph with 2.49 million +> nodes in just 16 seconds on a single GPU. Our code is available at +> https://github.com/limengran98/ARB. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on mitigating cold start problems in attribute-missing graphs for Graph Neural Networks (GNNs), which does not meet the 'MUST' criteria of primarily focusing on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts for improving LLM performance. + +--- + +## [Enhancing Transformers for Generalizable First-Order Logical Entailment](https://arxiv.org/abs/http://arxiv.org/abs/2501.00759v1) +**arXiv ID:** http://arxiv.org/abs/2501.00759v1 + +**Abstract:** +> Transformers, as a fundamental deep learning architecture, have demonstrated +> remarkable capabilities in reasoning. This paper investigates the generalizable +> first-order logical reasoning ability of transformers with their parameterized +> knowledge and explores ways to improve it. The first-order reasoning capability +> of transformers is assessed through their ability to perform first-order +> logical entailment, which is quantitatively measured by their performance in +> answering knowledge graph queries. We establish connections between (1) two +> types of distribution shifts studied in out-of-distribution generalization and +> (2) the unseen knowledge and query settings discussed in the task of knowledge +> graph query answering, enabling a characterization of fine-grained +> generalizability. Results on our comprehensive dataset show that transformers +> outperform previous methods specifically designed for this task and provide +> detailed empirical evidence on the impact of input query syntax, token +> embedding, and transformer architectures on the reasoning capability of +> transformers. Interestingly, our findings reveal a mismatch between positional +> encoding and other design choices in transformer architectures employed in +> prior practices. This discovery motivates us to propose a more sophisticated, +> logic-aware architecture, TEGA, to enhance the capability for generalizable +> first-order logical entailment in transformers. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on enhancing transformer architectures for first-order logical entailment, which aligns with the development of new LLM architectures, rather than primarily focusing on prompt engineering for text-based interactions with existing LLMs. + +--- + +## [REM: A Scalable Reinforced Multi-Expert Framework for Multiplex + Influence Maximization](https://arxiv.org/abs/http://arxiv.org/abs/2501.00779v1) +**arXiv ID:** http://arxiv.org/abs/2501.00779v1 + +**Abstract:** +> In social online platforms, identifying influential seed users to maximize +> influence spread is a crucial as it can greatly diminish the cost and efforts +> required for information dissemination. While effective, traditional methods +> for Multiplex Influence Maximization (MIM) have reached their performance +> limits, prompting the emergence of learning-based approaches. These novel +> methods aim for better generalization and scalability for more sizable graphs +> but face significant challenges, such as (1) inability to handle unknown +> diffusion patterns and (2) reliance on high-quality training samples. To +> address these issues, we propose the Reinforced Expert Maximization framework +> (REM). REM leverages a Propagation Mixture of Experts technique to encode +> dynamic propagation of large multiplex networks effectively in order to +> generate enhanced influence propagation. Noticeably, REM treats a generative +> model as a policy to autonomously generate different seed sets and learn how to +> improve them from a Reinforcement Learning perspective. Extensive experiments +> on several real-world datasets demonstrate that REM surpasses state-of-the-art +> methods in terms of influence spread, scalability, and inference time in +> influence maximization tasks. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the criteria as it primarily focuses on developing a novel framework (REM) for Multiplex Influence Maximization, leveraging Reinforcement Learning and a generative model, but does not concentrate on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or demonstrate the impact of textual input prompts on LLM output. + +--- + +## [Make Shuffling Great Again: A Side-Channel Resistant Fisher-Yates + Algorithm for Protecting Neural Networks](https://arxiv.org/abs/http://arxiv.org/abs/2501.00798v1) +**arXiv ID:** http://arxiv.org/abs/2501.00798v1 + +**Abstract:** +> Neural network models implemented in embedded devices have been shown to be +> susceptible to side-channel attacks (SCAs), allowing recovery of proprietary +> model parameters, such as weights and biases. There are already available +> countermeasure methods currently used for protecting cryptographic +> implementations that can be tailored to protect embedded neural network models. +> Shuffling, a hiding-based countermeasure that randomly shuffles the order of +> computations, was shown to be vulnerable to SCA when the Fisher-Yates algorithm +> is used. In this paper, we propose a design of an SCA-secure version of the +> Fisher-Yates algorithm. By integrating the masking technique for modular +> reduction and Blakely's method for modular multiplication, we effectively +> remove the vulnerability in the division operation that led to side-channel +> leakage in the original version of the algorithm. We experimentally evaluate +> that the countermeasure is effective against SCA by implementing a correlation +> power analysis attack on an embedded neural network model implemented on ARM +> Cortex-M4. Compared to the original proposal, the memory overhead is $2\times$ +> the biggest layer of the network, while the time overhead varies from $4\%$ to +> $0.49\%$ for a layer with $100$ and $1000$ neurons, respectively. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on protecting neural networks from side-channel attacks by modifying the Fisher-Yates algorithm, which does not meet the 'MUST' criteria: it does not primarily focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs), nor does it investigate improving LLM performance through textual input prompt manipulation. + +--- + +## [Reasoning-Oriented and Analogy-Based Methods for Locating and Editing in + Zero-Shot Event-Relational Reasoning](https://arxiv.org/abs/http://arxiv.org/abs/2501.00803v1) +**arXiv ID:** http://arxiv.org/abs/2501.00803v1 + +**Abstract:** +> Zero-shot event-relational reasoning is an important task in natural language +> processing, and existing methods jointly learn a variety of event-relational +> prefixes and inference-form prefixes to achieve such tasks. However, training +> prefixes consumes large computational resources and lacks interpretability. +> Additionally, learning various relational and inferential knowledge +> inefficiently exploits the connections between tasks. Therefore, we first +> propose a method for Reasoning-Oriented Locating and Editing (ROLE), which +> locates and edits the key modules of the language model for reasoning about +> event relations, enhancing interpretability and also resource-efficiently +> optimizing the reasoning ability. Subsequently, we propose a method for +> Analogy-Based Locating and Editing (ABLE), which efficiently exploits the +> similarities and differences between tasks to optimize the zero-shot reasoning +> capability. Experimental results show that ROLE improves interpretability and +> reasoning performance with reduced computational cost. ABLE achieves SOTA +> results in zero-shot reasoning. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on optimizing the language model's reasoning ability through module editing and analogy-based methods, rather than engineering or manipulating textual input prompts to improve Large Language Model (LLM) performance, thus not meeting the core subject requirement of prompt engineering for text-based LLM interactions. + +--- + +## [LLM-Powered Multi-Agent System for Automated Crypto Portfolio Management](https://arxiv.org/abs/http://arxiv.org/abs/2501.00826v2) +**arXiv ID:** http://arxiv.org/abs/2501.00826v2 + +**Abstract:** +> Cryptocurrency investment is inherently difficult due to its shorter history +> compared to traditional assets, the need to integrate vast amounts of data from +> various modalities, and the requirement for complex reasoning. While deep +> learning approaches have been applied to address these challenges, their +> black-box nature raises concerns about trust and explainability. Recently, +> large language models (LLMs) have shown promise in financial applications due +> to their ability to understand multi-modal data and generate explainable +> decisions. However, single LLM faces limitations in complex, comprehensive +> tasks such as asset investment. These limitations are even more pronounced in +> cryptocurrency investment, where LLMs have less domain-specific knowledge in +> their training corpora. +> To overcome these challenges, we propose an explainable, multi-modal, +> multi-agent framework for cryptocurrency investment. Our framework uses +> specialized agents that collaborate within and across teams to handle subtasks +> such as data analysis, literature integration, and investment decision-making +> for the top 30 cryptocurrencies by market capitalization. The expert training +> module fine-tunes agents using multi-modal historical data and professional +> investment literature, while the multi-agent investment module employs +> real-time data to make informed cryptocurrency investment decisions. Unique +> intrateam and interteam collaboration mechanisms enhance prediction accuracy by +> adjusting final predictions based on confidence levels within agent teams and +> facilitating information sharing between teams. Empirical evaluation using data +> from November 2023 to September 2024 demonstrates that our framework +> outperforms single-agent models and market benchmarks in classification, asset +> pricing, portfolio, and explainability performance. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing a multi-agent system for automated crypto portfolio management using LLMs, rather than specifically engineering or optimizing prompts for LLMs, thus not meeting the core subject requirement. + +--- + +## [Embedding Style Beyond Topics: Analyzing Dispersion Effects Across + Different Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.00828v1) +**arXiv ID:** http://arxiv.org/abs/2501.00828v1 + +**Abstract:** +> This paper analyzes how writing style affects the dispersion of embedding +> vectors across multiple, state-of-the-art language models. While early +> transformer models primarily aligned with topic modeling, this study examines +> the role of writing style in shaping embedding spaces. Using a literary corpus +> that alternates between topics and styles, we compare the sensitivity of +> language models across French and English. By analyzing the particular impact +> of style on embedding dispersion, we aim to better understand how language +> models process stylistic information, contributing to their overall +> interpretability. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on analyzing the impact of writing style on embedding vector dispersion across language models, rather than specifically investigating prompt engineering techniques for improving Large Language Model (LLM) performance through textual input prompt manipulation." +} + +--- + +## [An LLM-Empowered Adaptive Evolutionary Algorithm For Multi-Component + Deep Learning Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.00829v1) +**arXiv ID:** http://arxiv.org/abs/2501.00829v1 + +**Abstract:** +> Multi-objective evolutionary algorithms (MOEAs) are widely used for searching +> optimal solutions in complex multi-component applications. Traditional MOEAs +> for multi-component deep learning (MCDL) systems face challenges in enhancing +> the search efficiency while maintaining the diversity. To combat these, this +> paper proposes $\mu$MOEA, the first LLM-empowered adaptive evolutionary search +> algorithm to detect safety violations in MCDL systems. Inspired by the +> context-understanding ability of Large Language Models (LLMs), $\mu$MOEA +> promotes the LLM to comprehend the optimization problem and generate an initial +> population tailed to evolutionary objectives. Subsequently, it employs adaptive +> selection and variation to iteratively produce offspring, balancing the +> evolutionary efficiency and diversity. During the evolutionary process, to +> navigate away from the local optima, $\mu$MOEA integrates the evolutionary +> experience back into the LLM. This utilization harnesses the LLM's quantitative +> reasoning prowess to generate differential seeds, breaking away from current +> optimal solutions. We evaluate $\mu$MOEA in finding safety violations of MCDL +> systems, and compare its performance with state-of-the-art MOEA methods. +> Experimental results show that $\mu$MOEA can significantly improve the +> efficiency and diversity of the evolutionary search. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing an adaptive evolutionary algorithm for multi-component deep learning systems, utilizing LLMs as a component, rather than focusing specifically on prompt engineering for Large Language Models. Prompt engineering is not the central concern, but rather a means to enhance the algorithm's performance. + +--- + +## [LLM+AL: Bridging Large Language Models and Action Languages for Complex + Reasoning about Actions](https://arxiv.org/abs/http://arxiv.org/abs/2501.00830v1) +**arXiv ID:** http://arxiv.org/abs/2501.00830v1 + +**Abstract:** +> Large Language Models (LLMs) have made significant strides in various +> intelligent tasks but still struggle with complex action reasoning tasks that +> require systematic search. To address this limitation, we propose a method that +> bridges the natural language understanding capabilities of LLMs with the +> symbolic reasoning strengths of action languages. Our approach, termed +> "LLM+AL," leverages the LLM's strengths in semantic parsing and commonsense +> knowledge generation alongside the action language's proficiency in automated +> reasoning based on encoded knowledge. We compare LLM+AL against +> state-of-the-art LLMs, including ChatGPT-4, Claude 3 Opus, Gemini Ultra 1.0, +> and o1-preview, using benchmarks for complex reasoning about actions. Our +> findings indicate that, although all methods exhibit errors, LLM+AL, with +> relatively minimal human corrections, consistently leads to correct answers, +> whereas standalone LLMs fail to improve even with human feedback. LLM+AL also +> contributes to automated generation of action languages. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on integrating LLMs with Action Languages for complex reasoning, rather than prompt engineering specifically for LLMs, and does not provide concrete examples of textual input prompts and their impact on LLM output." +} + +--- + +## [Distilled Lifelong Self-Adaptation for Configurable Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.00840v1) +**arXiv ID:** http://arxiv.org/abs/2501.00840v1 + +**Abstract:** +> Modern configurable systems provide tremendous opportunities for engineering +> future intelligent software systems. A key difficulty thereof is how to +> effectively self-adapt the configuration of a running system such that its +> performance (e.g., runtime and throughput) can be optimized under time-varying +> workloads. This unfortunately remains unaddressed in existing approaches as +> they either overlook the available past knowledge or rely on static +> exploitation of past knowledge without reasoning the usefulness of information +> when planning for self-adaptation. In this paper, we tackle this challenging +> problem by proposing DLiSA, a framework that self-adapts configurable systems. +> DLiSA comes with two properties: firstly, it supports lifelong planning, and +> thereby the planning process runs continuously throughout the lifetime of the +> system, allowing dynamic exploitation of the accumulated knowledge for rapid +> adaptation. Secondly, the planning for a newly emerged workload is boosted via +> distilled knowledge seeding, in which the knowledge is dynamically purified +> such that only useful past configurations are seeded when necessary, mitigating +> misleading information. Extensive experiments suggest that the proposed DLiSA +> significantly outperforms state-of-the-art approaches, demonstrating a +> performance improvement of up to 229% and a resource acceleration of up to +> 2.22x on generating promising adaptation configurations. All data and sources +> can be found at our repository: https://github.com/ideas-labo/dlisa. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the primary focus criteria, as it concentrates on self-adaptation for configurable systems' performance optimization, lacking any mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, and instead focuses on system configuration and workload adaptation." +} + +--- + +## [Diversity Optimization for Travelling Salesman Problem via Deep + Reinforcement Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.00884v1) +**arXiv ID:** http://arxiv.org/abs/2501.00884v1 + +**Abstract:** +> Existing neural methods for the Travelling Salesman Problem (TSP) mostly aim +> at finding a single optimal solution. To discover diverse yet high-quality +> solutions for Multi-Solution TSP (MSTSP), we propose a novel deep reinforcement +> learning based neural solver, which is primarily featured by an encoder-decoder +> structured policy. Concretely, on the one hand, a Relativization Filter (RF) is +> designed to enhance the robustness of the encoder to affine transformations of +> the instances, so as to potentially improve the quality of the found solutions. +> On the other hand, a Multi-Attentive Adaptive Active Search (MA3S) is tailored +> to allow the decoders to strike a balance between the optimality and diversity. +> Experimental evaluations on benchmark instances demonstrate the superiority of +> our method over recent neural baselines across different metrics, and its +> competitive performance against state-of-the-art traditional heuristics with +> significantly reduced computational time, ranging from $1.3\times$ to +> $15\times$ faster. Furthermore, we demonstrate that our method can also be +> applied to the Capacitated Vehicle Routing Problem (CVRP). + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not focus on prompt engineering for Large Language Models (LLMs) as required. Instead, it explores deep reinforcement learning for solving the Travelling Salesman Problem and Capacitated Vehicle Routing Problem, which does not meet the primary criteria of prompt engineering for text-based interactions with LLMs." +} + +--- + +## [Population Aware Diffusion for Time Series Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.00910v1) +**arXiv ID:** http://arxiv.org/abs/2501.00910v1 + +**Abstract:** +> Diffusion models have shown promising ability in generating high-quality time +> series (TS) data. Despite the initial success, existing works mostly focus on +> the authenticity of data at the individual level, but pay less attention to +> preserving the population-level properties on the entire dataset. Such +> population-level properties include value distributions for each dimension and +> distributions of certain functional dependencies (e.g., cross-correlation, CC) +> between different dimensions. For instance, when generating house energy +> consumption TS data, the value distributions of the outside temperature and the +> kitchen temperature should be preserved, as well as the distribution of CC +> between them. Preserving such TS population-level properties is critical in +> maintaining the statistical insights of the datasets, mitigating model bias, +> and augmenting downstream tasks like TS prediction. Yet, it is often overlooked +> by existing models. Hence, data generated by existing models often bear +> distribution shifts from the original data. We propose Population-aware +> Diffusion for Time Series (PaD-TS), a new TS generation model that better +> preserves the population-level properties. The key novelties of PaD-TS include +> 1) a new training method explicitly incorporating TS population-level property +> preservation, and 2) a new dual-channel encoder model architecture that better +> captures the TS data structure. Empirical results in major benchmark datasets +> show that PaD-TS can improve the average CC distribution shift score between +> real and synthetic data by 5.9x while maintaining a performance comparable to +> state-of-the-art models on individual-level authenticity. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on developing a new diffusion model for time series generation, emphasizing population-level property preservation, without any mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria." +} + +--- + +## [$β$-DQN: Improving Deep Q-Learning By Evolving the Behavior](https://arxiv.org/abs/http://arxiv.org/abs/2501.00913v1) +**arXiv ID:** http://arxiv.org/abs/2501.00913v1 + +**Abstract:** +> While many sophisticated exploration methods have been proposed, their lack +> of generality and high computational cost often lead researchers to favor +> simpler methods like $\epsilon$-greedy. Motivated by this, we introduce +> $\beta$-DQN, a simple and efficient exploration method that augments the +> standard DQN with a behavior function $\beta$. This function estimates the +> probability that each action has been taken at each state. By leveraging +> $\beta$, we generate a population of diverse policies that balance exploration +> between state-action coverage and overestimation bias correction. An adaptive +> meta-controller is designed to select an effective policy for each episode, +> enabling flexible and explainable exploration. $\beta$-DQN is straightforward +> to implement and adds minimal computational overhead to the standard DQN. +> Experiments on both simple and challenging exploration domains show that +> $\beta$-DQN outperforms existing baseline methods across a wide range of tasks, +> providing an effective solution for improving exploration in deep reinforcement +> learning. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on improving Deep Q-Learning with a new exploration method, not on the engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Incremental Dialogue Management: Survey, Discussion, and Implications + for HRI](https://arxiv.org/abs/http://arxiv.org/abs/2501.00953v1) +**arXiv ID:** http://arxiv.org/abs/2501.00953v1 + +**Abstract:** +> Efforts towards endowing robots with the ability to speak have benefited from +> recent advancements in NLP, in particular large language models. However, as +> powerful as current models have become, they still operate on sentence or +> multi-sentence level input, not on the word-by-word input that humans operate +> on, affecting the degree of responsiveness that they offer, which is critical +> in situations where humans interact with robots using speech. In this paper, we +> review the literature on interactive systems that operate incrementally (i.e., +> at the word level or below it). We motivate the need for incremental systems, +> survey incremental modeling of important aspects of dialogue like speech +> recognition and language generation. Primary focus is on the part of the system +> that makes decisions, known as the dialogue manager. We find that there is very +> little research on incremental dialogue management, offer some requirements for +> practical incremental dialogue management, and the implications of incremental +> dialogue for embodied, robotic platforms. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on incremental dialogue management for Human-Robot Interaction (HRI), not specifically on the engineering, design, or optimization of prompts for Large Language Models (LLMs), failing to meet the core subject requirement. + +--- + +## [Are LLMs effective psychological assessors? Leveraging adaptive RAG for + interpretable mental health screening through psychometric practice](https://arxiv.org/abs/http://arxiv.org/abs/2501.00982v1) +**arXiv ID:** http://arxiv.org/abs/2501.00982v1 + +**Abstract:** +> In psychological practice, standardized questionnaires serve as essential +> tools for assessing mental constructs (e.g., attitudes, traits, and emotions) +> through structured questions (aka items). With the increasing prevalence of +> social media platforms where users share personal experiences and emotions, +> researchers are exploring computational methods to leverage this data for rapid +> mental health screening. In this study, we propose a novel adaptive +> Retrieval-Augmented Generation (RAG) approach that completes psychological +> questionnaires by analyzing social media posts. Our method retrieves the most +> relevant user posts for each question in a psychological survey and uses Large +> Language Models (LLMs) to predict questionnaire scores in a zero-shot setting. +> Our findings are twofold. First we demonstrate that this approach can +> effectively predict users' responses to psychological questionnaires, such as +> the Beck Depression Inventory II (BDI-II), achieving performance comparable to +> or surpassing state-of-the-art models on Reddit-based benchmark datasets +> without relying on training data. Second, we show how this methodology can be +> generalized as a scalable screening tool, as the final assessment is +> systematically derived by completing standardized questionnaires and tracking +> how individual item responses contribute to the diagnosis, aligning with +> established psychometric practices. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is leveraging LLMs for mental health screening via social media analysis, rather than prompt engineering for text-based interactions with LLMs. Prompt manipulation is not the central concern, but rather a means to achieve psychological assessment." +} + +--- + +## [Bootstrapped Reward Shaping](https://arxiv.org/abs/http://arxiv.org/abs/2501.00989v1) +**arXiv ID:** http://arxiv.org/abs/2501.00989v1 + +**Abstract:** +> In reinforcement learning, especially in sparse-reward domains, many +> environment steps are required to observe reward information. In order to +> increase the frequency of such observations, "potential-based reward shaping" +> (PBRS) has been proposed as a method of providing a more dense reward signal +> while leaving the optimal policy invariant. However, the required "potential +> function" must be carefully designed with task-dependent knowledge to not deter +> training performance. In this work, we propose a "bootstrapped" method of +> reward shaping, termed BSRS, in which the agent's current estimate of the +> state-value function acts as the potential function for PBRS. We provide +> convergence proofs for the tabular setting, give insights into training +> dynamics for deep RL, and show that the proposed method improves training speed +> in the Atari suite. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on reinforcement learning, proposing a method for reward shaping in sparse-reward domains, and does not meet any of the 'MUST' criteria, particularly lacking focus on prompt engineering, design, or optimization for Large Language Models (LLMs), and instead pertains to a different area of AI research. + +--- + +## [Exploring Information Processing in Large Language Models: Insights from + Information Bottleneck Theory](https://arxiv.org/abs/http://arxiv.org/abs/2501.00999v2) +**arXiv ID:** http://arxiv.org/abs/2501.00999v2 + +**Abstract:** +> Large Language Models (LLMs) have demonstrated remarkable performance across +> a wide range of tasks by understanding input information and predicting +> corresponding outputs. However, the internal mechanisms by which LLMs +> comprehend input and make effective predictions remain poorly understood. In +> this paper, we explore the working mechanism of LLMs in information processing +> from the perspective of Information Bottleneck Theory. We propose a +> non-training construction strategy to define a task space and identify the +> following key findings: (1) LLMs compress input information into specific task +> spaces (e.g., sentiment space, topic space) to facilitate task understanding; +> (2) they then extract and utilize relevant information from the task space at +> critical moments to generate accurate predictions. Based on these insights, we +> introduce two novel approaches: an Information Compression-based Context +> Learning (IC-ICL) and a Task-Space-guided Fine-Tuning (TS-FT). IC-ICL enhances +> reasoning performance and inference efficiency by compressing retrieved example +> information into the task space. TS-FT employs a space-guided loss to fine-tune +> LLMs, encouraging the learning of more effective compression and selection +> mechanisms. Experiments across multiple datasets validate the effectiveness of +> task space construction. Additionally, IC-ICL not only improves performance but +> also accelerates inference speed by over 40\%, while TS-FT achieves superior +> results with a minimal strategy adjustment. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on understanding LLM's internal information processing mechanisms through Information Bottleneck Theory and proposes methods for fine-tuning (TS-FT) and context learning (IC-ICL), which do not meet the 'MUST' criteria of focusing on the engineering, design, or optimization of prompts specifically for LLMs, nor does it provide concrete examples of prompts and their impact on LLM output. + +--- + +## [Deep Reinforcement Learning for Job Scheduling and Resource Management + in Cloud Computing: An Algorithm-Level Review](https://arxiv.org/abs/http://arxiv.org/abs/2501.01007v1) +**arXiv ID:** http://arxiv.org/abs/2501.01007v1 + +**Abstract:** +> Cloud computing has revolutionized the provisioning of computing resources, +> offering scalable, flexible, and on-demand services to meet the diverse +> requirements of modern applications. At the heart of efficient cloud operations +> are job scheduling and resource management, which are critical for optimizing +> system performance and ensuring timely and cost-effective service delivery. +> However, the dynamic and heterogeneous nature of cloud environments presents +> significant challenges for these tasks, as workloads and resource availability +> can fluctuate unpredictably. Traditional approaches, including heuristic and +> meta-heuristic algorithms, often struggle to adapt to these real-time changes +> due to their reliance on static models or predefined rules. Deep Reinforcement +> Learning (DRL) has emerged as a promising solution to these challenges by +> enabling systems to learn and adapt policies based on continuous observations +> of the environment, facilitating intelligent and responsive decision-making. +> This survey provides a comprehensive review of DRL-based algorithms for job +> scheduling and resource management in cloud computing, analyzing their +> methodologies, performance metrics, and practical applications. We also +> highlight emerging trends and future research directions, offering valuable +> insights into leveraging DRL to advance both job scheduling and resource +> management in cloud computing. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on Deep Reinforcement Learning for job scheduling and resource management in cloud computing, with no mention of Large Language Models (LLMs), prompt engineering, or textual input manipulation, thus failing to meet all 'MUST' criteria. + +--- + +## [MDSF: Context-Aware Multi-Dimensional Data Storytelling Framework based + on Large language Model](https://arxiv.org/abs/http://arxiv.org/abs/2501.01014v1) +**arXiv ID:** http://arxiv.org/abs/2501.01014v1 + +**Abstract:** +> The exponential growth of data and advancements in big data technologies have +> created a demand for more efficient and automated approaches to data analysis +> and storytelling. However, automated data analysis systems still face +> challenges in leveraging large language models (LLMs) for data insight +> discovery, augmented analysis, and data storytelling. This paper introduces the +> Multidimensional Data Storytelling Framework (MDSF) based on large language +> models for automated insight generation and context-aware storytelling. The +> framework incorporates advanced preprocessing techniques, augmented analysis +> algorithms, and a unique scoring mechanism to identify and prioritize +> actionable insights. The use of fine-tuned LLMs enhances contextual +> understanding and generates narratives with minimal manual intervention. The +> architecture also includes an agent-based mechanism for real-time storytelling +> continuation control. Key findings reveal that MDSF outperforms existing +> methods across various datasets in terms of insight ranking accuracy, +> descriptive quality, and narrative coherence. The experimental evaluation +> demonstrates MDSF's ability to automate complex analytical tasks, reduce +> interpretive biases, and improve user satisfaction. User studies further +> underscore its practical utility in enhancing content structure, conclusion +> extraction, and richness of detail. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses primarily on developing a multidimensional data storytelling framework using LLMs as a component, rather than centered on prompt engineering for LLMs. The core subject is the framework's architecture and performance, not novel prompt engineering techniques or systematic prompt development for text-based interactions with LLMs." +} + +--- + +## [Towards Adversarially Robust Deep Metric Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01025v2) +**arXiv ID:** http://arxiv.org/abs/2501.01025v2 + +**Abstract:** +> Deep Metric Learning (DML) has shown remarkable successes in many domains by +> taking advantage of powerful deep neural networks. Deep neural networks are +> prone to adversarial attacks and could be easily fooled by adversarial +> examples. The current progress on this robustness issue is mainly about deep +> classification models but pays little attention to DML models. Existing works +> fail to thoroughly inspect the robustness of DML and neglect an important DML +> scenario, the clustering-based inference. In this work, we first point out the +> robustness issue of DML models in clustering-based inference scenarios. We find +> that, for the clustering-based inference, existing defenses designed DML are +> unable to be reused and the adaptions of defenses designed for deep +> classification models cannot achieve satisfactory robustness performance. To +> alleviate the hazard of adversarial examples, we propose a new defense, the +> Ensemble Adversarial Training (EAT), which exploits ensemble learning and +> adversarial training. EAT promotes the diversity of the ensemble, encouraging +> each model in the ensemble to have different robustness features, and employs a +> self-transferring mechanism to make full use of the robustness statistics of +> the whole ensemble in the update of every single model. We evaluate the EAT +> method on three widely-used datasets with two popular model architectures. The +> results show that the proposed EAT method greatly outperforms the adaptions of +> defenses designed for deep classification models. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs) but instead addresses Deep Metric Learning (DML) and its robustness against adversarial attacks, with no mention of LLMs, text generation, or prompt manipulation. + +--- + +## [Reasoning based on symbolic and parametric knowledge bases: a survey](https://arxiv.org/abs/http://arxiv.org/abs/2501.01030v1) +**arXiv ID:** http://arxiv.org/abs/2501.01030v1 + +**Abstract:** +> Reasoning is fundamental to human intelligence, and critical for +> problem-solving, decision-making, and critical thinking. Reasoning refers to +> drawing new conclusions based on existing knowledge, which can support various +> applications like clinical diagnosis, basic education, and financial analysis. +> Though a good number of surveys have been proposed for reviewing +> reasoning-related methods, none of them has systematically investigated these +> methods from the viewpoint of their dependent knowledge base. Both the +> scenarios to which the knowledge bases are applied and their storage formats +> are significantly different. Hence, investigating reasoning methods from the +> knowledge base perspective helps us better understand the challenges and future +> directions. To fill this gap, this paper first classifies the knowledge base +> into symbolic and parametric ones. The former explicitly stores information in +> human-readable symbols, and the latter implicitly encodes knowledge within +> parameters. Then, we provide a comprehensive overview of reasoning methods +> using symbolic knowledge bases, parametric knowledge bases, and both of them. +> Finally, we identify the future direction toward enhancing reasoning +> capabilities to bridge the gap between human and machine intelligence. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus primarily on the engineering, design, or optimization of prompts for Large Language Models (LLMs), but rather on a survey of reasoning methods based on symbolic and parametric knowledge bases, without explicit mention of prompt engineering or LLMs. + +--- + +## [MSWA: Refining Local Attention with Multi-ScaleWindow Attention](https://arxiv.org/abs/http://arxiv.org/abs/2501.01039v1) +**arXiv ID:** http://arxiv.org/abs/2501.01039v1 + +**Abstract:** +> Transformer-based LLMs have achieved exceptional performance across a wide +> range of NLP tasks. However, the standard self-attention mechanism suffers from +> quadratic time complexity and linearly increased cache size. Sliding window +> attention (SWA) solves this problem by restricting the attention range to a +> fixed-size local context window. Nevertheless, SWA employs a uniform window +> size for each head in each layer, making it inefficient in capturing context of +> varying scales. To mitigate this limitation, we propose Multi-Scale Window +> Attention (MSWA) which applies diverse window sizes across heads and layers in +> the Transformer. It not only allows for different window sizes among heads +> within the same layer but also progressively increases window size allocation +> from shallow to deep layers, thus enabling the model to capture contextual +> information with different lengths and distances. Experimental results on +> language modeling and common-sense reasoning tasks substantiate that MSWA +> outperforms traditional local attention in both effectiveness and efficiency. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on refining the local attention mechanism in Transformer-based LLMs through Multi-Scale Window Attention, which pertains to the development of new LLM architecture/training methods, not prompt engineering for text-based interactions with LLMs. + +--- + +## [Risks of Cultural Erasure in Large Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01056v1) +**arXiv ID:** http://arxiv.org/abs/2501.01056v1 + +**Abstract:** +> Large language models are increasingly being integrated into applications +> that shape the production and discovery of societal knowledge such as search, +> online education, and travel planning. As a result, language models will shape +> how people learn about, perceive and interact with global cultures making it +> important to consider whose knowledge systems and perspectives are represented +> in models. Recognizing this importance, increasingly work in Machine Learning +> and NLP has focused on evaluating gaps in global cultural representational +> distribution within outputs. However, more work is needed on developing +> benchmarks for cross-cultural impacts of language models that stem from a +> nuanced sociologically-aware conceptualization of cultural impact or harm. We +> join this line of work arguing for the need of metricizable evaluations of +> language technologies that interrogate and account for historical power +> inequities and differential impacts of representation on global cultures, +> particularly for cultures already under-represented in the digital corpora. We +> look at two concepts of erasure: omission: where cultures are not represented +> at all and simplification i.e. when cultural complexity is erased by presenting +> one-dimensional views of a rich culture. The former focuses on whether +> something is represented, and the latter on how it is represented. We focus our +> analysis on two task contexts with the potential to influence global cultural +> production. First, we probe representations that a language model produces +> about different places around the world when asked to describe these contexts. +> Second, we analyze the cultures represented in the travel recommendations +> produced by a set of language model applications. Our study shows ways in which +> the NLP community and application developers can begin to operationalize +> complex socio-cultural considerations into standard evaluations and benchmarks. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on evaluating cultural representation and potential erasure in Large Language Models, but does not primarily investigate, analyze, or propose methods for improving LLM performance through the manipulation of textual input prompts, nor does it provide concrete examples of prompts and their impact on LLM output. + +--- + +## [Graph Generative Pre-trained Transformer](https://arxiv.org/abs/http://arxiv.org/abs/2501.01073v1) +**arXiv ID:** http://arxiv.org/abs/2501.01073v1 + +**Abstract:** +> Graph generation is a critical task in numerous domains, including molecular +> design and social network analysis, due to its ability to model complex +> relationships and structured data. While most modern graph generative models +> utilize adjacency matrix representations, this work revisits an alternative +> approach that represents graphs as sequences of node set and edge set. We +> advocate for this approach due to its efficient encoding of graphs and propose +> a novel representation. Based on this representation, we introduce the Graph +> Generative Pre-trained Transformer (G2PT), an auto-regressive model that learns +> graph structures via next-token prediction. To further exploit G2PT's +> capabilities as a general-purpose foundation model, we explore fine-tuning +> strategies for two downstream applications: goal-oriented generation and graph +> property prediction. We conduct extensive experiments across multiple datasets. +> Results indicate that G2PT achieves superior generative performance on both +> generic graph and molecule datasets. Furthermore, G2PT exhibits strong +> adaptability and versatility in downstream tasks from molecular design to +> property prediction. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on the development of a new graph generative model (Graph Generative Pre-trained Transformer) and its fine-tuning strategies for downstream applications, rather than primarily on the engineering, design, or optimization of prompts for Large Language Models (LLMs). + +--- + +## [BatStyler: Advancing Multi-category Style Generation for Source-free + Domain Generalization](https://arxiv.org/abs/http://arxiv.org/abs/2501.01109v1) +**arXiv ID:** http://arxiv.org/abs/2501.01109v1 + +**Abstract:** +> Source-Free Domain Generalization (SFDG) aims to develop a model that +> performs on unseen domains without relying on any source domains. However, the +> implementation remains constrained due to the unavailability of training data. +> Research on SFDG focus on knowledge transfer of multi-modal models and style +> synthesis based on joint space of multiple modalities, thus eliminating the +> dependency on source domain images. However, existing works primarily work for +> multi-domain and less-category configuration, but performance on multi-domain +> and multi-category configuration is relatively poor. In addition, the +> efficiency of style synthesis also deteriorates in multi-category scenarios. +> How to efficiently synthesize sufficiently diverse data and apply it to +> multi-category configuration is a direction with greater practical value. In +> this paper, we propose a method called BatStyler, which is utilized to improve +> the capability of style synthesis in multi-category scenarios. BatStyler +> consists of two modules: Coarse Semantic Generation and Uniform Style +> Generation modules. The Coarse Semantic Generation module extracts +> coarse-grained semantics to prevent the compression of space for style +> diversity learning in multi-category configuration, while the Uniform Style +> Generation module provides a template of styles that are uniformly distributed +> in space and implements parallel training. Extensive experiments demonstrate +> that our method exhibits comparable performance on less-category datasets, +> while surpassing state-of-the-art methods on multi-category datasets. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it focuses on Source-Free Domain Generalization for multi-category style generation, not on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not provide concrete examples of prompts impacting LLM output. + +--- + +## [MalCL: Leveraging GAN-Based Generative Replay to Combat Catastrophic + Forgetting in Malware Classification](https://arxiv.org/abs/http://arxiv.org/abs/2501.01110v1) +**arXiv ID:** http://arxiv.org/abs/2501.01110v1 + +**Abstract:** +> Continual Learning (CL) for malware classification tackles the rapidly +> evolving nature of malware threats and the frequent emergence of new types. +> Generative Replay (GR)-based CL systems utilize a generative model to produce +> synthetic versions of past data, which are then combined with new data to +> retrain the primary model. Traditional machine learning techniques in this +> domain often struggle with catastrophic forgetting, where a model's performance +> on old data degrades over time. +> In this paper, we introduce a GR-based CL system that employs Generative +> Adversarial Networks (GANs) with feature matching loss to generate high-quality +> malware samples. Additionally, we implement innovative selection schemes for +> replay samples based on the model's hidden representations. +> Our comprehensive evaluation across Windows and Android malware datasets in a +> class-incremental learning scenario -- where new classes are introduced +> continuously over multiple tasks -- demonstrates substantial performance +> improvements over previous methods. For example, our system achieves an average +> accuracy of 55% on Windows malware samples, significantly outperforming other +> GR-based models by 28%. This study provides practical insights for advancing +> GR-based malware classification systems. The implementation is available at +> \url {https://github.com/MalwareReplayGAN/MalCL}\footnote{The code will be made +> public upon the presentation of the paper}. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on leveraging GAN-based Generative Replay for Continual Learning in malware classification, which does not meet the 'MUST' criteria of primarily focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts for improving LLM performance. + +--- + +## [Pruning-based Data Selection and Network Fusion for Efficient Deep + Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01118v1) +**arXiv ID:** http://arxiv.org/abs/2501.01118v1 + +**Abstract:** +> Efficient data selection is essential for improving the training efficiency +> of deep neural networks and reducing the associated annotation costs. However, +> traditional methods tend to be computationally expensive, limiting their +> scalability and real-world applicability. We introduce PruneFuse, a novel +> method that combines pruning and network fusion to enhance data selection and +> accelerate network training. In PruneFuse, the original dense network is pruned +> to generate a smaller surrogate model that efficiently selects the most +> informative samples from the dataset. Once this iterative data selection +> selects sufficient samples, the insights learned from the pruned model are +> seamlessly integrated with the dense model through network fusion, providing an +> optimized initialization that accelerates training. Extensive experimentation +> on various datasets demonstrates that PruneFuse significantly reduces +> computational costs for data selection, achieves better performance than +> baselines, and accelerates the overall training process. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on improving deep neural network training efficiency through pruning and network fusion, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Retrieval-Augmented Dynamic Prompt Tuning for Incomplete Multimodal + Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01120v1) +**arXiv ID:** http://arxiv.org/abs/2501.01120v1 + +**Abstract:** +> Multimodal learning with incomplete modality is practical and challenging. +> Recently, researchers have focused on enhancing the robustness of pre-trained +> MultiModal Transformers (MMTs) under missing modality conditions by applying +> learnable prompts. However, these prompt-based methods face several +> limitations: (1) incomplete modalities provide restricted modal cues for +> task-specific inference, (2) dummy imputation for missing content causes +> information loss and introduces noise, and (3) static prompts are +> instance-agnostic, offering limited knowledge for instances with various +> missing conditions. To address these issues, we propose RAGPT, a novel +> Retrieval-AuGmented dynamic Prompt Tuning framework. RAGPT comprises three +> modules: (I) the multi-channel retriever, which identifies similar instances +> through a within-modality retrieval strategy, (II) the missing modality +> generator, which recovers missing information using retrieved contexts, and +> (III) the context-aware prompter, which captures contextual knowledge from +> relevant instances and generates dynamic prompts to largely enhance the MMT's +> robustness. Extensive experiments conducted on three real-world datasets show +> that RAGPT consistently outperforms all competitive baselines in handling +> incomplete modality problems. The code of our work and prompt-based baselines +> is available at https://github.com/Jian-Lang/RAGPT. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing the robustness of MultiModal Transformers (MMTs) for incomplete multimodal learning, with dynamic prompt tuning being a component of the proposed framework, rather than the core subject. The main focus is on addressing multimodal learning challenges, not specifically on prompt engineering for text-based interactions with Large Language Models (LLMs). + +--- + +## [Deep Learning in Palmprint Recognition-A Comprehensive Survey](https://arxiv.org/abs/http://arxiv.org/abs/2501.01166v1) +**arXiv ID:** http://arxiv.org/abs/2501.01166v1 + +**Abstract:** +> Palmprint recognition has emerged as a prominent biometric technology, widely +> applied in diverse scenarios. Traditional handcrafted methods for palmprint +> recognition often fall short in representation capability, as they heavily +> depend on researchers' prior knowledge. Deep learning (DL) has been introduced +> to address this limitation, leveraging its remarkable successes across various +> domains. While existing surveys focus narrowly on specific tasks within +> palmprint recognition-often grounded in traditional methodologies-there remains +> a significant gap in comprehensive research exploring DL-based approaches +> across all facets of palmprint recognition. This paper bridges that gap by +> thoroughly reviewing recent advancements in DL-powered palmprint recognition. +> The paper systematically examines progress across key tasks, including +> region-of-interest segmentation, feature extraction, and +> security/privacy-oriented challenges. Beyond highlighting these advancements, +> the paper identifies current challenges and uncovers promising opportunities +> for future research. By consolidating state-of-the-art progress, this review +> serves as a valuable resource for researchers, enabling them to stay abreast of +> cutting-edge technologies and drive innovation in palmprint recognition. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on deep learning in palmprint recognition, a biometric technology unrelated to Large Language Models (LLMs) or prompt engineering for text-based interactions, thus failing to meet the primary 'MUST' criteria. + +--- + +## [Blind Men and the Elephant: Diverse Perspectives on Gender Stereotypes + in Benchmark Datasets](https://arxiv.org/abs/http://arxiv.org/abs/2501.01168v1) +**arXiv ID:** http://arxiv.org/abs/2501.01168v1 + +**Abstract:** +> The multifaceted challenge of accurately measuring gender stereotypical bias +> in language models is akin to discerning different segments of a broader, +> unseen entity. This short paper primarily focuses on intrinsic bias mitigation +> and measurement strategies for language models, building on prior research that +> demonstrates a lack of correlation between intrinsic and extrinsic approaches. +> We delve deeper into intrinsic measurements, identifying inconsistencies and +> suggesting that these benchmarks may reflect different facets of gender +> stereotype. Our methodology involves analyzing data distributions across +> datasets and integrating gender stereotype components informed by social +> psychology. By adjusting the distribution of two datasets, we achieve a better +> alignment of outcomes. Our findings underscore the complexity of gender +> stereotyping in language models and point to new directions for developing more +> refined techniques to detect and reduce bias. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on measuring and mitigating gender stereotypical bias in language models, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not provide concrete examples of prompts and their impact on LLM output. + +--- + +## [L3D-Pose: Lifting Pose for 3D Avatars from a Single Camera in the Wild](https://arxiv.org/abs/http://arxiv.org/abs/2501.01174v1) +**arXiv ID:** http://arxiv.org/abs/2501.01174v1 + +**Abstract:** +> While 2D pose estimation has advanced our ability to interpret body movements +> in animals and primates, it is limited by the lack of depth information, +> constraining its application range. 3D pose estimation provides a more +> comprehensive solution by incorporating spatial depth, yet creating extensive +> 3D pose datasets for animals is challenging due to their dynamic and +> unpredictable behaviours in natural settings. To address this, we propose a +> hybrid approach that utilizes rigged avatars and the pipeline to generate +> synthetic datasets to acquire the necessary 3D annotations for training. Our +> method introduces a simple attention-based MLP network for converting 2D poses +> to 3D, designed to be independent of the input image to ensure scalability for +> poses in natural environments. Additionally, we identify that existing +> anatomical keypoint detectors are insufficient for accurate pose retargeting +> onto arbitrary avatars. To overcome this, we present a lookup table based on a +> deep pose estimation method using a synthetic collection of diverse actions +> rigged avatars perform. Our experiments demonstrate the effectiveness and +> efficiency of this lookup table-based retargeting approach. Overall, we propose +> a comprehensive framework with systematically synthesized datasets for lifting +> poses from 2D to 3D and then utilize this to re-target motion from wild +> settings onto arbitrary avatars. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on 3D pose estimation for avatars using computer vision and machine learning techniques, with no primary emphasis on the engineering, design, or optimization of prompts for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Data Augmentation Techniques for Chinese Disease Name Normalization](https://arxiv.org/abs/http://arxiv.org/abs/2501.01195v1) +**arXiv ID:** http://arxiv.org/abs/2501.01195v1 + +**Abstract:** +> Disease name normalization is an important task in the medical domain. It +> classifies disease names written in various formats into standardized names, +> serving as a fundamental component in smart healthcare systems for various +> disease-related functions. Nevertheless, the most significant obstacle to +> existing disease name normalization systems is the severe shortage of training +> data. Consequently, we present a novel data augmentation approach that includes +> a series of data augmentation techniques and some supporting modules to help +> mitigate the problem. Through extensive experimentation, we illustrate that our +> proposed approach exhibits significant performance improvements across various +> baseline models and training objectives, particularly in scenarios with limited +> training data + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on medical domain (disease name normalization) and data augmentation techniques for training models, rather than prompt engineering for Large Language Models (LLMs) and manipulating textual input prompts to improve LLM performance. + +--- + +## [A redescription mining framework for post-hoc explaining and relating + deep learning models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01209v1) +**arXiv ID:** http://arxiv.org/abs/2501.01209v1 + +**Abstract:** +> Deep learning models (DLMs) achieve increasingly high performance both on +> structured and unstructured data. They significantly extended applicability of +> machine learning to various domains. Their success in making predictions, +> detecting patterns and generating new data made significant impact on science +> and industry. Despite these accomplishments, DLMs are difficult to explain +> because of their enormous size. In this work, we propose a novel framework for +> post-hoc explaining and relating DLMs using redescriptions. The framework +> allows cohort analysis of arbitrary DLMs by identifying statistically +> significant redescriptions of neuron activations. It allows coupling neurons to +> a set of target labels or sets of descriptive attributes, relating layers +> within a single DLM or associating different DLMs. The proposed framework is +> independent of the artificial neural network architecture and can work with +> more complex target labels (e.g. multi-label or multi-target scenario). +> Additionally, it can emulate both pedagogical and decompositional approach to +> rule extraction. The aforementioned properties of the proposed framework can +> increase explainability and interpretability of arbitrary DLMs by providing +> different information compared to existing explainable-AI approaches. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on explaining and relating deep learning models (DLMs) in general, using redescriptions, without a specific emphasis on Large Language Models (LLMs), prompt engineering, or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [An Efficient Attention Mechanism for Sequential Recommendation Tasks: + HydraRec](https://arxiv.org/abs/http://arxiv.org/abs/2501.01242v1) +**arXiv ID:** http://arxiv.org/abs/2501.01242v1 + +**Abstract:** +> Transformer based models are increasingly being used in various domains +> including recommender systems (RS). Pretrained transformer models such as BERT +> have shown good performance at language modelling. With the greater ability to +> model sequential tasks, variants of Encoder-only models (like BERT4Rec, SASRec +> etc.) have found success in sequential RS problems. Computing dot-product +> attention in traditional transformer models has quadratic complexity in +> sequence length. This is a bigger problem with RS because unlike language +> models, new items are added to the catalogue every day. User buying history is +> a dynamic sequence which depends on multiple factors. Recently, various linear +> attention models have tried to solve this problem by making the model linear in +> sequence length (token dimensions). Hydra attention is one such linear +> complexity model proposed for vision transformers which reduces the complexity +> of attention for both the number of tokens as well as model embedding +> dimensions. Building on the idea of Hydra attention, we introduce an efficient +> Transformer based Sequential RS (HydraRec) which significantly improves +> theoretical complexity of computing attention for longer sequences and bigger +> datasets while preserving the temporal context. Extensive experiments are +> conducted to evaluate other linear transformer-based RS models and compared +> with HydraRec across various evaluation metrics. HydraRec outperforms other +> linear attention-based models as well as dot-product based attention models +> when used with causal masking for sequential recommendation next item +> prediction tasks. For bi-directional models its performance is comparable to +> the BERT4Rec model with an improvement in running time. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing an efficient attention mechanism for sequential recommendation tasks using transformer-based models, rather than engineering or optimizing prompts specifically for Large Language Models (LLMs). It does not investigate methods for improving LLM performance through the manipulation of textual input prompts, nor does it provide concrete examples of prompts and their impact on LLM output. + +--- + +## [Stealthy Backdoor Attack to Real-world Models in Android Apps](https://arxiv.org/abs/http://arxiv.org/abs/2501.01263v1) +**arXiv ID:** http://arxiv.org/abs/2501.01263v1 + +**Abstract:** +> Powered by their superior performance, deep neural networks (DNNs) have found +> widespread applications across various domains. Many deep learning (DL) models +> are now embedded in mobile apps, making them more accessible to end users +> through on-device DL. However, deploying on-device DL to users' smartphones +> simultaneously introduces several security threats. One primary threat is +> backdoor attacks. Extensive research has explored backdoor attacks for several +> years and has proposed numerous attack approaches. However, few studies have +> investigated backdoor attacks on DL models deployed in the real world, or they +> have shown obvious deficiencies in effectiveness and stealthiness. In this +> work, we explore more effective and stealthy backdoor attacks on real-world DL +> models extracted from mobile apps. Our main justification is that imperceptible +> and sample-specific backdoor triggers generated by DNN-based steganography can +> enhance the efficacy of backdoor attacks on real-world models. We first confirm +> the effectiveness of steganography-based backdoor attacks on four +> state-of-the-art DNN models. Subsequently, we systematically evaluate and +> analyze the stealthiness of the attacks to ensure they are difficult to +> perceive. Finally, we implement the backdoor attacks on real-world models and +> compare our approach with three baseline methods. We collect 38,387 mobile +> apps, extract 89 DL models from them, and analyze these models to obtain the +> prerequisite model information for the attacks. After identifying the target +> models, our approach achieves an average of 12.50% higher attack success rate +> than DeepPayload while better maintaining the normal performance of the models. +> Extensive experimental results demonstrate that our method enables more +> effective, robust, and stealthy backdoor attacks on real-world models. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on backdoor attacks in deep neural networks deployed in Android apps, not on the engineering, design, or optimization of prompts for Large Language Models (LLMs), failing to meet all the primary 'MUST' criteria. + +--- + +## [PIMAEX: Multi-Agent Exploration through Peer Incentivization](https://arxiv.org/abs/http://arxiv.org/abs/2501.01266v1) +**arXiv ID:** http://arxiv.org/abs/2501.01266v1 + +**Abstract:** +> While exploration in single-agent reinforcement learning has been studied +> extensively in recent years, considerably less work has focused on its +> counterpart in multi-agent reinforcement learning. To address this issue, this +> work proposes a peer-incentivized reward function inspired by previous research +> on intrinsic curiosity and influence-based rewards. The \textit{PIMAEX} reward, +> short for Peer-Incentivized Multi-Agent Exploration, aims to improve +> exploration in the multi-agent setting by encouraging agents to exert influence +> over each other to increase the likelihood of encountering novel states. We +> evaluate the \textit{PIMAEX} reward in conjunction with +> \textit{PIMAEX-Communication}, a multi-agent training algorithm that employs a +> communication channel for agents to influence one another. The evaluation is +> conducted in the \textit{Consume/Explore} environment, a partially observable +> environment with deceptive rewards, specifically designed to challenge the +> exploration vs.\ exploitation dilemma and the credit-assignment problem. The +> results empirically demonstrate that agents using the \textit{PIMAEX} reward +> with \textit{PIMAEX-Communication} outperform those that do not. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multi-agent reinforcement learning, proposing a reward function and training algorithm for improved exploration, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), thus failing to meet the core subject requirement. + +--- + +## [NeutraSum: A Language Model can help a Balanced Media Diet by + Neutralizing News Summaries](https://arxiv.org/abs/http://arxiv.org/abs/2501.01284v1) +**arXiv ID:** http://arxiv.org/abs/2501.01284v1 + +**Abstract:** +> Media bias in news articles arises from the political polarisation of media +> outlets, which can reinforce societal stereotypes and beliefs. Reporting on the +> same event often varies significantly between outlets, reflecting their +> political leanings through polarised language and focus. Although previous +> studies have attempted to generate bias-free summaries from multiperspective +> news articles, they have not effectively addressed the challenge of mitigating +> inherent media bias. To address this gap, we propose \textbf{NeutraSum}, a +> novel framework that integrates two neutrality losses to adjust the semantic +> space of generated summaries, thus minimising media bias. These losses, +> designed to balance the semantic distances across polarised inputs and ensure +> alignment with expert-written summaries, guide the generation of neutral and +> factually rich summaries. To evaluate media bias, we employ the political +> compass test, which maps political leanings based on economic and social +> dimensions. Experimental results on the Allsides dataset demonstrate that +> NeutraSum not only improves summarisation performance but also achieves +> significant reductions in media bias, offering a promising approach for neutral +> news summarisation. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a framework (NeutraSum) to mitigate media bias in news summaries, rather than engineer or optimize prompts specifically for Large Language Models (LLMs). While it involves generated summaries, the core subject is not prompt engineering for text-based interactions with LLMs, but rather the reduction of media bias. + +--- + +## [Citations and Trust in LLM Generated Responses](https://arxiv.org/abs/http://arxiv.org/abs/2501.01303v1) +**arXiv ID:** http://arxiv.org/abs/2501.01303v1 + +**Abstract:** +> Question answering systems are rapidly advancing, but their opaque nature may +> impact user trust. We explored trust through an anti-monitoring framework, +> where trust is predicted to be correlated with presence of citations and +> inversely related to checking citations. We tested this hypothesis with a live +> question-answering experiment that presented text responses generated using a +> commercial Chatbot along with varying citations (zero, one, or five), both +> relevant and random, and recorded if participants checked the citations and +> their self-reported trust in the generated responses. We found a significant +> increase in trust when citations were present, a result that held true even +> when the citations were random; we also found a significant decrease in trust +> when participants checked the citations. These results highlight the importance +> of citations in enhancing trust in AI-generated content. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is on enhancing user trust in AI-generated content through citations, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not demonstrate the impact of textual input prompts on LLM output." +} + +--- + +## [Multi-Head Explainer: A General Framework to Improve Explainability in + CNNs and Transformers](https://arxiv.org/abs/http://arxiv.org/abs/2501.01311v2) +**arXiv ID:** http://arxiv.org/abs/2501.01311v2 + +**Abstract:** +> In this study, we introduce the Multi-Head Explainer (MHEX), a versatile and +> modular framework that enhances both the explainability and accuracy of +> Convolutional Neural Networks (CNNs) and Transformer-based models. MHEX +> consists of three core components: an Attention Gate that dynamically +> highlights task-relevant features, Deep Supervision that guides early layers to +> capture fine-grained details pertinent to the target class, and an Equivalent +> Matrix that unifies refined local and global representations to generate +> comprehensive saliency maps. Our approach demonstrates superior compatibility, +> enabling effortless integration into existing residual networks like ResNet and +> Transformer architectures such as BERT with minimal modifications. Extensive +> experiments on benchmark datasets in medical imaging and text classification +> show that MHEX not only improves classification accuracy but also produces +> highly interpretable and detailed saliency scores. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a framework to enhance explainability and accuracy in CNNs and Transformers, rather than engineering prompts for Large Language Models (LLMs). Prompt engineering for LLMs is not the central concern, and the paper falls under excluded categories (development of new model components and medical subjects). + +--- + +## [Understanding Difficult-to-learn Examples in Contrastive Learning: A + Theoretical Framework for Spectral Contrastive Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01317v1) +**arXiv ID:** http://arxiv.org/abs/2501.01317v1 + +**Abstract:** +> Unsupervised contrastive learning has shown significant performance +> improvements in recent years, often approaching or even rivaling supervised +> learning in various tasks. However, its learning mechanism is fundamentally +> different from that of supervised learning. Previous works have shown that +> difficult-to-learn examples (well-recognized in supervised learning as examples +> around the decision boundary), which are essential in supervised learning, +> contribute minimally in unsupervised settings. In this paper, perhaps +> surprisingly, we find that the direct removal of difficult-to-learn examples, +> although reduces the sample size, can boost the downstream classification +> performance of contrastive learning. To uncover the reasons behind this, we +> develop a theoretical framework modeling the similarity between different pairs +> of samples. Guided by this theoretical framework, we conduct a thorough +> theoretical analysis revealing that the presence of difficult-to-learn examples +> negatively affects the generalization of contrastive learning. Furthermore, we +> demonstrate that the removal of these examples, and techniques such as margin +> tuning and temperature scaling can enhance its generalization bounds, thereby +> improving performance. Empirically, we propose a simple and efficient mechanism +> for selecting difficult-to-learn examples and validate the effectiveness of the +> aforementioned methods, which substantiates the reliability of our proposed +> theoretical framework. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on unsupervised contrastive learning, its theoretical framework, and improving downstream classification performance, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria. + +--- + +## [DeepFilter: An Instrumental Baseline for Accurate and Efficient Process + Monitoring](https://arxiv.org/abs/http://arxiv.org/abs/2501.01342v1) +**arXiv ID:** http://arxiv.org/abs/2501.01342v1 + +**Abstract:** +> Effective process monitoring is increasingly vital in industrial automation +> for ensuring operational safety, necessitating both high accuracy and +> efficiency. Although Transformers have demonstrated success in various fields, +> their canonical form based on the self-attention mechanism is inadequate for +> process monitoring due to two primary limitations: (1) the step-wise +> correlations captured by self-attention mechanism are difficult to capture +> discriminative patterns in monitoring logs due to the lacking semantics of each +> step, thus compromising accuracy; (2) the quadratic computational complexity of +> self-attention hampers efficiency. To address these issues, we propose +> DeepFilter, a Transformer-style framework for process monitoring. The core +> innovation is an efficient filtering layer that excel capturing long-term and +> periodic patterns with reduced complexity. Equipping with the global filtering +> layer, DeepFilter enhances both accuracy and efficiency, meeting the stringent +> demands of process monitoring. Experimental results on real-world process +> monitoring datasets validate DeepFilter's superiority in terms of accuracy and +> efficiency compared to existing state-of-the-art models. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on developing a new Transformer-style framework (DeepFilter) for process monitoring in industrial automation, addressing limitations of self-attention mechanisms, which falls under 'development of new LLM architectures or training methods' and does not centrally concern prompt engineering for text-based interactions with LLMs." +} + +--- + +## [A Unified Hyperparameter Optimization Pipeline for Transformer-Based + Time Series Forecasting Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01394v1) +**arXiv ID:** http://arxiv.org/abs/2501.01394v1 + +**Abstract:** +> Transformer-based models for time series forecasting (TSF) have attracted +> significant attention in recent years due to their effectiveness and +> versatility. However, these models often require extensive hyperparameter +> optimization (HPO) to achieve the best possible performance, and a unified +> pipeline for HPO in transformer-based TSF remains lacking. In this paper, we +> present one such pipeline and conduct extensive experiments on several +> state-of-the-art (SOTA) transformer-based TSF models. These experiments are +> conducted on standard benchmark datasets to evaluate and compare the +> performance of different models, generating practical insights and examples. +> Our pipeline is generalizable beyond transformer-based architectures and can be +> applied to other SOTA models, such as Mamba and TimeMixer, as demonstrated in +> our experiments. The goal of this work is to provide valuable guidance to both +> industry practitioners and academic researchers in efficiently identifying +> optimal hyperparameters suited to their specific domain applications. The code +> and complete experimental results are available on GitHub. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on hyperparameter optimization for transformer-based time series forecasting models, rather than on the engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not provide examples of prompts or their impact on LLM output. + +--- + +## [On Unifying Video Generation and Camera Pose Estimation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01409v1) +**arXiv ID:** http://arxiv.org/abs/2501.01409v1 + +**Abstract:** +> Inspired by the emergent 3D capabilities in image generators, we explore +> whether video generators similarly exhibit 3D awareness. Using +> structure-from-motion (SfM) as a benchmark for 3D tasks, we investigate if +> intermediate features from OpenSora, a video generation model, can support +> camera pose estimation. We first examine native 3D awareness in video +> generation features by routing raw intermediate outputs to SfM-prediction +> modules like DUSt3R. Then, we explore the impact of fine-tuning on camera pose +> estimation to enhance 3D awareness. Results indicate that while video generator +> features have limited inherent 3D awareness, task-specific supervision +> significantly boosts their accuracy for camera pose estimation, resulting in +> competitive performance. The proposed unified model, named JOG3R, produces +> camera pose estimates with competitive quality without degrading video +> generation quality. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on video generation and camera pose estimation, which falls under image/video generation, not text generation driven by Large Language Models (LLMs), thereby violating the 'Papers MUST NOT' criteria #2. + +--- + +## [Balance-aware Sequence Sampling Makes Multi-modal Learning Better](https://arxiv.org/abs/http://arxiv.org/abs/2501.01470v1) +**arXiv ID:** http://arxiv.org/abs/2501.01470v1 + +**Abstract:** +> To address the modality imbalance caused by data heterogeneity, existing +> multi-modal learning (MML) approaches primarily focus on balancing this +> difference from the perspective of optimization objectives. However, almost all +> existing methods ignore the impact of sample sequences, i.e., an inappropriate +> training order tends to trigger learning bias in the model, further +> exacerbating modality imbalance. In this paper, we propose Balance-aware +> Sequence Sampling (BSS) to enhance the robustness of MML. Specifically, we +> first define a multi-perspective measurer to evaluate the balance degree of +> each sample. Via the evaluation, we employ a heuristic scheduler based on +> curriculum learning (CL) that incrementally provides training subsets, +> progressing from balanced to imbalanced samples to rebalance MML. Moreover, +> considering that sample balance may evolve as the model capability increases, +> we propose a learning-based probabilistic sampling method to dynamically update +> the training sequences at the epoch level, further improving MML performance. +> Extensive experiments on widely used datasets demonstrate the superiority of +> our method compared with state-of-the-art (SOTA) MML approaches. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing multi-modal learning (MML) through balance-aware sequence sampling, ignoring the core subject of prompt engineering for text-based interactions with Large Language Models (LLMs), and instead addressing modality imbalance in MML without mentioning prompts or LLM-specific text generation. + +--- + +## [Augmented Contrastive Clustering with Uncertainty-Aware Prototyping for + Time Series Test Time Adaptation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01472v1) +**arXiv ID:** http://arxiv.org/abs/2501.01472v1 + +**Abstract:** +> Test-time adaptation aims to adapt pre-trained deep neural networks using +> solely online unlabelled test data during inference. Although TTA has shown +> promise in visual applications, its potential in time series contexts remains +> largely unexplored. Existing TTA methods, originally designed for visual tasks, +> may not effectively handle the complex temporal dynamics of real-world time +> series data, resulting in suboptimal adaptation performance. To address this +> gap, we propose Augmented Contrastive Clustering with Uncertainty-aware +> Prototyping (ACCUP), a straightforward yet effective TTA method for time series +> data. Initially, our approach employs augmentation ensemble on the time series +> data to capture diverse temporal information and variations, incorporating +> uncertainty-aware prototypes to distill essential characteristics. +> Additionally, we introduce an entropy comparison scheme to selectively acquire +> more confident predictions, enhancing the reliability of pseudo labels. +> Furthermore, we utilize augmented contrastive clustering to enhance feature +> discriminability and mitigate error accumulation from noisy pseudo labels, +> promoting cohesive clustering within the same class while facilitating clear +> separation between different classes. Extensive experiments conducted on three +> real-world time series datasets and an additional visual dataset demonstrate +> the effectiveness and generalization potential of the proposed method, +> advancing the underexplored realm of TTA for time series data. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it focuses on test-time adaptation for time series data using deep neural networks, without any mention of Large Language Models (LLMs), prompt engineering, or manipulation of textual input prompts for improving LLM performance. + +--- + +## [Unraveling Indirect In-Context Learning Using Influence Functions](https://arxiv.org/abs/http://arxiv.org/abs/2501.01473v1) +**arXiv ID:** http://arxiv.org/abs/2501.01473v1 + +**Abstract:** +> This work introduces a novel paradigm for generalized In-Context Learning +> (ICL), termed Indirect In-Context Learning. In Indirect ICL, we explore +> demonstration selection strategies tailored for two distinct real-world +> scenarios: Mixture of Tasks and Noisy Demonstrations. We systematically +> evaluate the effectiveness of Influence Functions (IFs) as a selection tool for +> these settings, highlighting the potential for IFs to better capture the +> informativeness of examples within the demonstration pool. For the Mixture of +> Tasks setting, demonstrations are drawn from 28 diverse tasks, including MMLU, +> BigBench, StrategyQA, and CommonsenseQA. We demonstrate that combining +> BertScore-Recall (BSR) with an IF surrogate model can significantly improve +> performance, leading to average absolute accuracy gains of 0.37\% and 1.45\% +> for 3-shot and 5-shot setups when compared to traditional ICL metrics. In the +> Noisy Demonstrations setting, we examine scenarios where demonstrations might +> be mislabeled. Our experiments show that reweighting traditional ICL selectors +> (BSR and Cosine Similarity) with IF-based selectors boosts accuracy by an +> average of 2.90\% for Cosine Similarity and 2.94\% for BSR on noisy GLUE +> benchmarks. In sum, we propose a robust framework for demonstration selection +> that generalizes beyond traditional ICL, offering valuable insights into the +> role of IFs for Indirect ICL. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on demonstration selection strategies using Influence Functions for In-Context Learning, rather than engineering or optimizing prompts specifically for Large Language Models (LLMs), failing to meet the core subject requirement. + +--- + +## [A Survey of Deep Learning Methods in Protein Bioinformatics and its + Impact on Protein Design](https://arxiv.org/abs/http://arxiv.org/abs/2501.01477v1) +**arXiv ID:** http://arxiv.org/abs/2501.01477v1 + +**Abstract:** +> Proteins are sequences of amino acids that serve as the basic building blocks +> of living organisms. Despite rapidly growing databases documenting structural +> and functional information for various protein sequences, our understanding of +> proteins remains limited because of the large possible sequence space and the +> complex inter- and intra-molecular forces. Deep learning, which is +> characterized by its ability to learn relevant features directly from large +> datasets, has demonstrated remarkable performance in fields such as computer +> vision and natural language processing. It has also been increasingly applied +> in recent years to the data-rich domain of protein sequences with great +> success, most notably with Alphafold2's breakout performance in the protein +> structure prediction. The performance improvements achieved by deep learning +> unlocks new possibilities in the field of protein bioinformatics, including +> protein design, one of the most difficult but useful tasks. In this paper, we +> broadly categorize problems in protein bioinformatics into three main +> categories: 1) structural prediction, 2) functional prediction, and 3) protein +> design, and review the progress achieved from using deep learning methodologies +> in each of them. We expand on the main challenges of the protein design problem +> and highlight how advances in structural and functional prediction have +> directly contributed to design tasks. Finally, we conclude by identifying +> important topics and future research directions. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on deep learning methods in protein bioinformatics and protein design, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria. + +--- + +## [Drift2Matrix: Kernel-Induced Self Representation for Concept Drift + Adaptation in Co-evolving Time Series](https://arxiv.org/abs/http://arxiv.org/abs/2501.01480v2) +**arXiv ID:** http://arxiv.org/abs/2501.01480v2 + +**Abstract:** +> In the realm of time series analysis, tackling the phenomenon of concept +> drift poses a significant challenge. Concept drift -- characterized by the +> evolving statistical properties of time series data, affects the reliability +> and accuracy of conventional analysis models. This is particularly evident in +> co-evolving scenarios where interactions among variables are crucial. This +> paper presents Drift2Matrix, a novel framework that leverages kernel-induced +> self-representation for adaptive responses to concept drift in time series. +> Drift2Matrix employs a kernel-based learning mechanism to generate a +> representation matrix, encapsulating the inherent dynamics of co-evolving time +> series. This matrix serves as a key tool for identification and adaptation to +> concept drift by observing its temporal variations. Furthermore, Drift2Matrix +> effectively identifies prevailing patterns and offers insights into emerging +> trends through pattern evolution analysis. Our empirical evaluation of +> Drift2Matrix across various datasets demonstrates its effectiveness in handling +> the complexities of concept drift. This approach introduces a novel perspective +> in the theoretical domain of co-evolving time series analysis, enhancing +> adaptability and accuracy in the face of dynamic data environments. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs). Instead, it presents a framework (Drift2Matrix) for adapting to concept drift in co-evolving time series analysis, which is unrelated to the engineering, design, or optimization of prompts for LLMs. + +--- + +## [BoxingGym: Benchmarking Progress in Automated Experimental Design and + Model Discovery](https://arxiv.org/abs/http://arxiv.org/abs/2501.01540v1) +**arXiv ID:** http://arxiv.org/abs/2501.01540v1 + +**Abstract:** +> Understanding the world and explaining it with scientific theories is a +> central aspiration of artificial intelligence research. Proposing theories, +> designing experiments to test them, and then revising them based on data are +> fundamental to scientific discovery. Despite the significant promise of +> LLM-based scientific agents, no benchmarks systematically test LLM's ability to +> propose scientific models, collect experimental data, and revise them in light +> of new data. We introduce BoxingGym, a benchmark with 10 environments for +> systematically evaluating both experimental design (e.g. collecting data to +> test a scientific theory) and model discovery (e.g. proposing and revising +> scientific theories). To enable tractable and quantitative evaluation, we +> implement each environment as a generative probabilistic model with which a +> scientific agent can run interactive experiments. These probabilistic models +> are drawn from various real-world scientific domains ranging from psychology to +> ecology. To quantitatively evaluate a scientific agent's ability to collect +> informative experimental data, we compute the expected information gain (EIG), +> an information-theoretic quantity which measures how much an experiment reduces +> uncertainty about the parameters of a generative model. A good scientific +> theory is a concise and predictive explanation. Therefore, to quantitatively +> evaluate model discovery, we ask a scientific agent to explain their model and +> then assess whether this explanation enables another scientific agent to make +> reliable predictions about this environment. In addition to this +> explanation-based evaluation, we compute standard model evaluation metrics such +> as prediction errors. We find that current LLMs, such as GPT-4o, struggle with +> both experimental design and model discovery. We find that augmenting the +> LLM-based agent with an explicit statistical model does not reliably improve +> these results. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on benchmarking scientific discovery and experimentation using LLMs, rather than specifically engineering or optimizing prompts for Large Language Models (LLMs). While LLMs are utilized, the core subject is the evaluation of scientific agents, not prompt engineering for text-based interactions with LLMs. + +--- + +## [Constructing and explaining machine learning models for chemistry: + example of the exploration and design of boron-based Lewis acids](https://arxiv.org/abs/http://arxiv.org/abs/2501.01576v2) +**arXiv ID:** http://arxiv.org/abs/2501.01576v2 + +**Abstract:** +> The integration of machine learning (ML) into chemistry offers transformative +> potential in the design of molecules with targeted properties. However, the +> focus has often been on creating highly efficient predictive models, sometimes +> at the expense of interpretability. In this study, we leverage explainable AI +> techniques to explore the rational design of boron-based Lewis acids, which +> play a pivotal role in organic reactions due to their electron-ccepting +> properties. Using Fluoride Ion Affinity as a proxy for Lewis acidity, we +> developed interpretable ML models based on chemically meaningful descriptors, +> including ab initio computed features and substituent-based parameters derived +> from the Hammett linear free-energy relationship. By constraining the chemical +> space to well-defined molecular scaffolds, we achieved highly accurate +> predictions (mean absolute error < 6 kJ/mol), surpassing conventional black-box +> deep learning models in low-data regimes. Interpretability analyses of the +> models shed light on the origin of Lewis acidity in these compounds and +> identified actionable levers to modulate it through the nature and positioning +> of substituents on the molecular scaffold. This work bridges ML and chemist's +> way of thinking, demonstrating how explainable models can inspire molecular +> design and enhance scientific understanding of chemical reactivity. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing interpretable machine learning models for chemistry applications, with no apparent emphasis on prompt engineering, Large Language Models (LLMs), or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [(WhyPHI) Fine-Tuning PHI-3 for Multiple-Choice Question Answering: + Methodology, Results, and Challenges](https://arxiv.org/abs/http://arxiv.org/abs/2501.01588v1) +**arXiv ID:** http://arxiv.org/abs/2501.01588v1 + +**Abstract:** +> Large Language Models (LLMs) have become essential tools across various +> domains due to their impressive capabilities in understanding and generating +> human-like text. The ability to accurately answer multiple-choice questions +> (MCQs) holds significant value in education, particularly in automated tutoring +> systems and assessment platforms. However, adapting LLMs to handle MCQ tasks +> effectively remains challenging due to the hallucinations and unclear prompts. +> This work explores the potential of Microsoft's PHI-3\cite{Abdin2024}, a +> compact yet efficient LLM, for MCQ answering. Our contributions include +> fine-tuning the model on the TruthfulQA dataset, designing optimized prompts to +> enhance model performance, and evaluating using perplexity and traditional +> metrics like accuracy and F1 score. Results show a remarkable improvement in +> PHI-3.5's MCQ handling post-fine-tuning, with perplexity decreasing from 4.68 +> to 2.27, and accuracy rising from 62\% to 90.8\%. This research underlines the +> importance of efficient models in adaptive learning systems and educational +> assessments, paving the way for broader integration into the classroom, +> particularly in fields like test preparation, student feedback, and +> personalized learning. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on fine-tuning a Large Language Model (PHI-3) for multiple-choice question answering, which falls under the development of new training methods, contravening the 'MUST NOT' criterion 1. Although it mentions designing optimized prompts, this is secondary to the fine-tuning methodology, not meeting the core subject requirement of prompt engineering being the central focus." +} + +--- + +## [Prism: Mining Task-aware Domains in Non-i.i.d. IMU Data for Flexible + User Perception](https://arxiv.org/abs/http://arxiv.org/abs/2501.01598v1) +**arXiv ID:** http://arxiv.org/abs/2501.01598v1 + +**Abstract:** +> A wide range of user perception applications leverage inertial measurement +> unit (IMU) data for online prediction. However, restricted by the non-i.i.d. +> nature of IMU data collected from mobile devices, most systems work well only +> in a controlled setting (e.g., for a specific user in particular postures), +> limiting application scenarios. To achieve uncontrolled online prediction on +> mobile devices, referred to as the flexible user perception (FUP) problem, is +> attractive but hard. In this paper, we propose a novel scheme, called Prism, +> which can obtain high FUP accuracy on mobile devices. The core of Prism is to +> discover task-aware domains embedded in IMU dataset, and to train a +> domain-aware model on each identified domain. To this end, we design an +> expectation-maximization (EM) algorithm to estimate latent domains with respect +> to the specific downstream perception task. Finally, the best-fit model can be +> automatically selected for use by comparing the test sample and all identified +> domains in the feature space. We implement Prism on various mobile devices and +> conduct extensive experiments. Results demonstrate that Prism can achieve the +> best FUP performance with a low latency. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Instead, it proposes a scheme for improving user perception through task-aware domains in IMU data, with no mention of LLMs or prompt engineering for text-based interactions. + +--- + +## [Few-shot Implicit Function Generation via Equivariance](https://arxiv.org/abs/http://arxiv.org/abs/2501.01601v1) +**arXiv ID:** http://arxiv.org/abs/2501.01601v1 + +**Abstract:** +> Implicit Neural Representations (INRs) have emerged as a powerful framework +> for representing continuous signals. However, generating diverse INR weights +> remains challenging due to limited training data. We introduce Few-shot +> Implicit Function Generation, a new problem setup that aims to generate diverse +> yet functionally consistent INR weights from only a few examples. This is +> challenging because even for the same signal, the optimal INRs can vary +> significantly depending on their initializations. To tackle this, we propose +> EquiGen, a framework that can generate new INRs from limited data. The core +> idea is that functionally similar networks can be transformed into one another +> through weight permutations, forming an equivariance group. By projecting these +> weights into an equivariant latent space, we enable diverse generation within +> these groups, even with few examples. EquiGen implements this through an +> equivariant encoder trained via contrastive learning and smooth augmentation, +> an equivariance-guided diffusion process, and controlled perturbations in the +> equivariant subspace. Experiments on 2D image and 3D shape INR datasets +> demonstrate that our approach effectively generates diverse INR weights while +> preserving their functional properties in few-shot scenarios. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on generating Implicit Neural Representations (INRs) weights for continuous signal representation, not on the engineering, design, or optimization of prompts for Large Language Models (LLMs), failing to meet the primary 'MUST' criteria." +} + +--- + +## [Google is all you need: Semi-Supervised Transfer Learning Strategy For + Light Multimodal Multi-Task Classification Model](https://arxiv.org/abs/http://arxiv.org/abs/2501.01611v1) +**arXiv ID:** http://arxiv.org/abs/2501.01611v1 + +**Abstract:** +> As the volume of digital image data increases, the effectiveness of image +> classification intensifies. This study introduces a robust multi-label +> classification system designed to assign multiple labels to a single image, +> addressing the complexity of images that may be associated with multiple +> categories (ranging from 1 to 19, excluding 12). We propose a multi-modal +> classifier that merges advanced image recognition algorithms with Natural +> Language Processing (NLP) models, incorporating a fusion module to integrate +> these distinct modalities. The purpose of integrating textual data is to +> enhance the accuracy of label prediction by providing contextual understanding +> that visual analysis alone cannot fully capture. Our proposed classification +> model combines Convolutional Neural Networks (CNN) for image processing with +> NLP techniques for analyzing textual description (i.e., captions). This +> approach includes rigorous training and validation phases, with each model +> component verified and analyzed through ablation experiments. Preliminary +> results demonstrate the classifier's accuracy and efficiency, highlighting its +> potential as an automatic image-labeling system. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a multi-modal classification model for image classification, integrating NLP for contextual understanding, rather than specifically on the engineering, design, or optimization of prompts for Large Language Models (LLMs). + +--- + +## [Merging Context Clustering with Visual State Space Models for Medical + Image Segmentation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01618v1) +**arXiv ID:** http://arxiv.org/abs/2501.01618v1 + +**Abstract:** +> Medical image segmentation demands the aggregation of global and local +> feature representations, posing a challenge for current methodologies in +> handling both long-range and short-range feature interactions. Recently, vision +> mamba (ViM) models have emerged as promising solutions for addressing model +> complexities by excelling in long-range feature iterations with linear +> complexity. However, existing ViM approaches overlook the importance of +> preserving short-range local dependencies by directly flattening spatial tokens +> and are constrained by fixed scanning patterns that limit the capture of +> dynamic spatial context information. To address these challenges, we introduce +> a simple yet effective method named context clustering ViM (CCViM), which +> incorporates a context clustering module within the existing ViM models to +> segment image tokens into distinct windows for adaptable local clustering. Our +> method effectively combines long-range and short-range feature interactions, +> thereby enhancing spatial contextual representations for medical image +> segmentation tasks. Extensive experimental evaluations on diverse public +> datasets, i.e., Kumar, CPM17, ISIC17, ISIC18, and Synapse demonstrate the +> superior performance of our method compared to current state-of-the-art +> methods. Our code can be found at https://github.com/zymissy/CCViM. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on medical image segmentation, incorporating visual state space models, and does not address prompt engineering for Large Language Models (LLMs) or demonstrate the impact of textual input prompts on LLM output, thus failing to meet the 'MUST' criteria. + +--- + +## [Implications of Artificial Intelligence on Health Data Privacy and + Confidentiality](https://arxiv.org/abs/http://arxiv.org/abs/2501.01639v2) +**arXiv ID:** http://arxiv.org/abs/2501.01639v2 + +**Abstract:** +> The rapid integration of artificial intelligence (AI) in healthcare is +> revolutionizing medical diagnostics, personalized medicine, and operational +> efficiency. However, alongside these advancements, significant challenges arise +> concerning patient data privacy, ethical considerations, and regulatory +> compliance. This paper examines the dual impact of AI on healthcare, +> highlighting its transformative potential and the critical need for +> safeguarding sensitive health information. It explores the role of the Health +> Insurance Portability and Accountability Act (HIPAA) as a regulatory framework +> for ensuring data privacy and security, emphasizing the importance of robust +> safeguards and ethical standards in AI-driven healthcare. Through case studies, +> including AI applications in diabetic retinopathy, oncology, and the +> controversies surrounding data sharing, this study underscores the ethical and +> legal complexities of AI implementation. A balanced approach that fosters +> innovation while maintaining patient trust and privacy is imperative. The +> findings emphasize the importance of continuous education, transparency, and +> adherence to regulatory frameworks to harness AI's full potential responsibly +> and ethically in healthcare. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on the implications of AI on health data privacy and confidentiality, rather than on the engineering, design, or optimization of prompts for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria for focusing on prompt engineering for LLMs. + +--- + +## [HLV-1K: A Large-scale Hour-Long Video Benchmark for Time-Specific Long + Video Understanding](https://arxiv.org/abs/http://arxiv.org/abs/2501.01645v1) +**arXiv ID:** http://arxiv.org/abs/2501.01645v1 + +**Abstract:** +> Multimodal large language models have become a popular topic in deep visual +> understanding due to many promising real-world applications. However, hour-long +> video understanding, spanning over one hour and containing tens of thousands of +> visual frames, remains under-explored because of 1) challenging long-term video +> analyses, 2) inefficient large-model approaches, and 3) lack of large-scale +> benchmark datasets. Among them, in this paper, we focus on building a +> large-scale hour-long long video benchmark, HLV-1K, designed to evaluate long +> video understanding models. HLV-1K comprises 1009 hour-long videos with 14,847 +> high-quality question answering (QA) and multi-choice question asnwering (MCQA) +> pairs with time-aware query and diverse annotations, covering frame-level, +> within-event-level, cross-event-level, and long-term reasoning tasks. We +> evaluate our benchmark using existing state-of-the-art methods and demonstrate +> its value for testing deep long video understanding capabilities at different +> levels and for various tasks. This includes promoting future long video +> understanding tasks at a granular level, such as deep understanding of long +> live videos, meeting recordings, and movies. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses primarily on building a benchmark dataset for hour-long video understanding, involving multimodal large language models, but does not concentrate on prompt engineering for text-based interactions with LLMs, instead emphasizing video analysis and understanding." +} + +--- + +## [AVATAR: Adversarial Autoencoders with Autoregressive Refinement for Time + Series Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01649v1) +**arXiv ID:** http://arxiv.org/abs/2501.01649v1 + +**Abstract:** +> Data augmentation can significantly enhance the performance of machine +> learning tasks by addressing data scarcity and improving generalization. +> However, generating time series data presents unique challenges. A model must +> not only learn a probability distribution that reflects the real data +> distribution but also capture the conditional distribution at each time step to +> preserve the inherent temporal dependencies. To address these challenges, we +> introduce AVATAR, a framework that combines Adversarial Autoencoders (AAE) with +> Autoregressive Learning to achieve both objectives. Specifically, our technique +> integrates the autoencoder with a supervisor and introduces a novel supervised +> loss to assist the decoder in learning the temporal dynamics of time series +> data. Additionally, we propose another innovative loss function, termed +> distribution loss, to guide the encoder in more efficiently aligning the +> aggregated posterior of the autoencoder's latent representation with a prior +> Gaussian distribution. Furthermore, our framework employs a joint training +> mechanism to simultaneously train all networks using a combined loss, thereby +> fulfilling the dual objectives of time series generation. We evaluate our +> technique across a variety of time series datasets with diverse +> characteristics. Our experiments demonstrate significant improvements in both +> the quality and practical utility of the generated data, as assessed by various +> qualitative and quantitative metrics. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a new framework (AVATAR) for time series generation using adversarial autoencoders and autoregressive learning, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [EAUWSeg: Eliminating annotation uncertainty in weakly-supervised medical + image segmentation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01658v1) +**arXiv ID:** http://arxiv.org/abs/2501.01658v1 + +**Abstract:** +> Weakly-supervised medical image segmentation is gaining traction as it +> requires only rough annotations rather than accurate pixel-to-pixel labels, +> thereby reducing the workload for specialists. Although some progress has been +> made, there is still a considerable performance gap between the label-efficient +> methods and fully-supervised one, which can be attributed to the uncertainty +> nature of these weak labels. To address this issue, we propose a novel weak +> annotation method coupled with its learning framework EAUWSeg to eliminate the +> annotation uncertainty. Specifically, we first propose the Bounded Polygon +> Annotation (BPAnno) by simply labeling two polygons for a lesion. Then, the +> tailored learning mechanism that explicitly treat bounded polygons as two +> separated annotations is proposed to learn invariant feature by providing +> adversarial supervision signal for model training. Subsequently, a +> confidence-auxiliary consistency learner incorporates with a +> classification-guided confidence generator is designed to provide reliable +> supervision signal for pixels in uncertain region by leveraging the feature +> presentation consistency across pixels within the same category as well as +> class-specific information encapsulated in bounded polygons annotation. +> Experimental results demonstrate that EAUWSeg outperforms existing +> weakly-supervised segmentation methods. Furthermore, compared to +> fully-supervised counterparts, the proposed method not only delivers superior +> performance but also costs much less annotation workload. This underscores the +> superiority and effectiveness of our approach. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on medical image segmentation, addressing annotation uncertainty in weakly-supervised learning, which violates the 'MUST NOT' criteria of being primarily concerned with medical subjects and not focusing on prompt engineering for Large Language Models (LLMs). + +--- + +## [BARTPredict: Empowering IoT Security with LLM-Driven Cyber Threat + Prediction](https://arxiv.org/abs/http://arxiv.org/abs/2501.01664v1) +**arXiv ID:** http://arxiv.org/abs/2501.01664v1 + +**Abstract:** +> The integration of Internet of Things (IoT) technology in various domains has +> led to operational advancements, but it has also introduced new vulnerabilities +> to cybersecurity threats, as evidenced by recent widespread cyberattacks on IoT +> devices. Intrusion detection systems are often reactive, triggered by specific +> patterns or anomalies observed within the network. To address this challenge, +> this work proposes a proactive approach to anticipate and preemptively mitigate +> malicious activities, aiming to prevent potential damage before it occurs. This +> paper proposes an innovative intrusion prediction framework empowered by +> Pre-trained Large Language Models (LLMs). The framework incorporates two LLMs: +> a fine-tuned Bidirectional and AutoRegressive Transformers (BART) model for +> predicting network traffic and a fine-tuned Bidirectional Encoder +> Representations from Transformers (BERT) model for evaluating the predicted +> traffic. By harnessing the bidirectional capabilities of BART the framework +> then identifies malicious packets among these predictions. Evaluated using the +> CICIoT2023 IoT attack dataset, our framework showcases a notable enhancement in +> predictive performance, attaining an impressive 98% overall accuracy, providing +> a powerful response to the cybersecurity challenges that confront IoT networks. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing an intrusion prediction framework for IoT security using fine-tuned LLMs, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models. Prompt engineering is not the central concern, but rather a means to achieve the framework's predictive capabilities. + +--- + +## [VidFormer: A novel end-to-end framework fused by 3DCNN and Transformer + for Video-based Remote Physiological Measurement](https://arxiv.org/abs/http://arxiv.org/abs/2501.01691v2) +**arXiv ID:** http://arxiv.org/abs/2501.01691v2 + +**Abstract:** +> Remote physiological signal measurement based on facial videos, also known as +> remote photoplethysmography (rPPG), involves predicting changes in facial +> vascular blood flow from facial videos. While most deep learning-based methods +> have achieved good results, they often struggle to balance performance across +> small and large-scale datasets due to the inherent limitations of convolutional +> neural networks (CNNs) and Transformer. In this paper, we introduce VidFormer, +> a novel end-to-end framework that integrates 3-Dimension Convolutional Neural +> Network (3DCNN) and Transformer models for rPPG tasks. Initially, we conduct an +> analysis of the traditional skin reflection model and subsequently introduce an +> enhanced model for the reconstruction of rPPG signals. Based on this improved +> model, VidFormer utilizes 3DCNN and Transformer to extract local and global +> features from input data, respectively. To enhance the spatiotemporal feature +> extraction capabilities of VidFormer, we incorporate temporal-spatial attention +> mechanisms tailored for both 3DCNN and Transformer. Additionally, we design a +> module to facilitate information exchange and fusion between the 3DCNN and +> Transformer. Our evaluation on five publicly available datasets demonstrates +> that VidFormer outperforms current state-of-the-art (SOTA) methods. Finally, we +> discuss the essential roles of each VidFormer module and examine the effects of +> ethnicity, makeup, and exercise on its performance. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a novel end-to-end framework for video-based remote physiological measurement, utilizing 3DCNN and Transformer, with no primary focus on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or improving LLM performance through textual input prompts. + +--- + +## [The Essence of Contextual Understanding in Theory of Mind: A Study on + Question Answering with Story Characters](https://arxiv.org/abs/http://arxiv.org/abs/2501.01705v1) +**arXiv ID:** http://arxiv.org/abs/2501.01705v1 + +**Abstract:** +> Theory-of-Mind (ToM) is a fundamental psychological capability that allows +> humans to understand and interpret the mental states of others. Humans infer +> others' thoughts by integrating causal cues and indirect clues from broad +> contextual information, often derived from past interactions. In other words, +> human ToM heavily relies on the understanding about the backgrounds and life +> stories of others. Unfortunately, this aspect is largely overlooked in existing +> benchmarks for evaluating machines' ToM capabilities, due to their usage of +> short narratives without global backgrounds. In this paper, we verify the +> importance of understanding long personal backgrounds in ToM and assess the +> performance of LLMs in such realistic evaluation scenarios. To achieve this, we +> introduce a novel benchmark, CharToM-QA, comprising 1,035 ToM questions based +> on characters from classic novels. Our human study reveals a significant +> disparity in performance: the same group of educated participants performs +> dramatically better when they have read the novels compared to when they have +> not. In parallel, our experiments on state-of-the-art LLMs, including the very +> recent o1 model, show that LLMs still perform notably worse than humans, +> despite that they have seen these stories during pre-training. This highlights +> the limitations of current LLMs in capturing the nuanced contextual information +> required for ToM reasoning. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on evaluating Large Language Models' (LLMs) performance in Theory of Mind tasks using a novel benchmark, rather than on the engineering, design, or optimization of prompts specifically for LLMs, as required by the criteria." +} + +--- + +## [MoVE-KD: Knowledge Distillation for VLMs with Mixture of Visual Encoders](https://arxiv.org/abs/http://arxiv.org/abs/2501.01709v1) +**arXiv ID:** http://arxiv.org/abs/2501.01709v1 + +**Abstract:** +> Visual encoders are fundamental components in vision-language models (VLMs), +> each showcasing unique strengths derived from various pre-trained visual +> foundation models. To leverage the various capabilities of these encoders, +> recent studies incorporate multiple encoders within a single VLM, leading to a +> considerable increase in computational cost. In this paper, we present +> Mixture-of-Visual-Encoder Knowledge Distillation (MoVE-KD), a novel framework +> that distills the unique proficiencies of multiple vision encoders into a +> single, efficient encoder model. Specifically, to mitigate conflicts and retain +> the unique characteristics of each teacher encoder, we employ low-rank +> adaptation (LoRA) and mixture-of-experts (MoEs) to selectively activate +> specialized knowledge based on input features, enhancing both adaptability and +> efficiency. To regularize the KD process and enhance performance, we propose an +> attention-based distillation strategy that adaptively weighs the different +> visual encoders and emphasizes valuable visual tokens, reducing the burden of +> replicating comprehensive but distinct features from multiple teachers. +> Comprehensive experiments on popular VLMs, such as LLaVA and LLaVA-NeXT, +> validate the effectiveness of our method. The code will be released. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on knowledge distillation for Vision-Language Models (VLMs) with multiple visual encoders, which does not meet the criteria of focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [LLMs & Legal Aid: Understanding Legal Needs Exhibited Through User + Queries](https://arxiv.org/abs/http://arxiv.org/abs/2501.01711v1) +**arXiv ID:** http://arxiv.org/abs/2501.01711v1 + +**Abstract:** +> The paper presents a preliminary analysis of an experiment conducted by Frank +> Bold, a Czech expert group, to explore user interactions with GPT-4 for +> addressing legal queries. Between May 3, 2023, and July 25, 2023, 1,252 users +> submitted 3,847 queries. Unlike studies that primarily focus on the accuracy, +> factuality, or hallucination tendencies of large language models (LLMs), our +> analysis focuses on the user query dimension of the interaction. Using GPT-4o +> for zero-shot classification, we categorized queries on (1) whether users +> provided factual information about their issue (29.95%) or not (70.05%), (2) +> whether they sought legal information (64.93%) or advice on the course of +> action (35.07\%), and (3) whether they imposed requirements to shape or control +> the model's answer (28.57%) or not (71.43%). We provide both quantitative and +> qualitative insight into user needs and contribute to a better understanding of +> user engagement with LLMs. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is on understanding user needs and engagement with LLMs for legal queries, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), as required by the criteria." +} + +--- + +## [Proposing Hierarchical Goal-Conditioned Policy Planning in Multi-Goal + Reinforcement Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01727v1) +**arXiv ID:** http://arxiv.org/abs/2501.01727v1 + +**Abstract:** +> Humanoid robots must master numerous tasks with sparse rewards, posing a +> challenge for reinforcement learning (RL). We propose a method combining RL and +> automated planning to address this. Our approach uses short goal-conditioned +> policies (GCPs) organized hierarchically, with Monte Carlo Tree Search (MCTS) +> planning using high-level actions (HLAs). Instead of primitive actions, the +> planning process generates HLAs. A single plan-tree, maintained during the +> agent's lifetime, holds knowledge about goal achievement. This hierarchy +> enhances sample efficiency and speeds up reasoning by reusing HLAs and +> anticipating future actions. Our Hierarchical Goal-Conditioned Policy Planning +> (HGCPP) framework uniquely integrates GCPs, MCTS, and hierarchical RL, +> potentially improving exploration and planning in complex tasks. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses primarily on multi-goal reinforcement learning for humanoid robots, using techniques like Monte Carlo Tree Search and hierarchical RL, with no apparent connection to Large Language Models (LLMs) or prompt engineering for text-based interactions." +} + +--- + +## [Augmentation Matters: A Mix-Paste Method for X-Ray Prohibited Item + Detection under Noisy Annotations](https://arxiv.org/abs/http://arxiv.org/abs/2501.01733v1) +**arXiv ID:** http://arxiv.org/abs/2501.01733v1 + +**Abstract:** +> Automatic X-ray prohibited item detection is vital for public safety. +> Existing deep learning-based methods all assume that the annotations of +> training X-ray images are correct. However, obtaining correct annotations is +> extremely hard if not impossible for large-scale X-ray images, where item +> overlapping is ubiquitous.As a result, X-ray images are easily contaminated +> with noisy annotations, leading to performance deterioration of existing +> methods.In this paper, we address the challenging problem of training a robust +> prohibited item detector under noisy annotations (including both category noise +> and bounding box noise) from a novel perspective of data augmentation, and +> propose an effective label-aware mixed patch paste augmentation method +> (Mix-Paste). Specifically, for each item patch, we mix several item patches +> with the same category label from different images and replace the original +> patch in the image with the mixed patch. In this way, the probability of +> containing the correct prohibited item within the generated image is increased. +> Meanwhile, the mixing process mimics item overlapping, enabling the model to +> learn the characteristics of X-ray images. Moreover, we design an item-based +> large-loss suppression (LLS) strategy to suppress the large losses +> corresponding to potentially positive predictions of additional items due to +> the mixing operation. We show the superiority of our method on X-ray datasets +> under noisy annotations. In addition, we evaluate our method on the noisy +> MS-COCO dataset to showcase its generalization ability. These results clearly +> indicate the great potential of data augmentation to handle noise annotations. +> The source code is released at https://github.com/wscds/Mix-Paste. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on data augmentation for improving prohibited item detection in X-ray images using deep learning, with no primary concern on prompt engineering, Large Language Models (LLMs), or textual input prompts." +} + +--- + +## [Automating Legal Concept Interpretation with LLMs: Retrieval, + Generation, and Evaluation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01743v1) +**arXiv ID:** http://arxiv.org/abs/2501.01743v1 + +**Abstract:** +> Legal articles often include vague concepts to adapt to the ever-changing +> society. Providing detailed interpretations of these concepts is a critical +> task for legal practitioners, which requires meticulous and professional +> annotations by legal experts, admittedly time-consuming and expensive to +> collect at scale. In this paper, we introduce a novel retrieval-augmented +> generation framework, ATRI, for AuTomatically Retrieving relevant information +> from past judicial precedents and Interpreting vague legal concepts. We further +> propose a new benchmark, Legal Concept Entailment, to automate the evaluation +> of generated concept interpretations without expert involvement. Automatic +> evaluations indicate that our generated interpretations can effectively assist +> large language models (LLMs) in understanding vague legal concepts. +> Multi-faceted evaluations by legal experts indicate that the quality of our +> concept interpretations is comparable to those written by human experts. Our +> work has strong implications for leveraging LLMs to support legal practitioners +> in interpreting vague legal concepts and beyond. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing a retrieval-augmented generation framework for legal concept interpretation, leveraging LLMs as a component, rather than focusing specifically on prompt engineering for text-based interactions with LLMs. + +--- + +## [Creating Artificial Students that Never Existed: Leveraging Large + Language Models and CTGANs for Synthetic Data Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01793v1) +**arXiv ID:** http://arxiv.org/abs/2501.01793v1 + +**Abstract:** +> In this study, we explore the growing potential of AI and deep learning +> technologies, particularly Generative Adversarial Networks (GANs) and Large +> Language Models (LLMs), for generating synthetic tabular data. Access to +> quality students data is critical for advancing learning analytics, but privacy +> concerns and stricter data protection regulations worldwide limit their +> availability and usage. Synthetic data offers a promising alternative. We +> investigate whether synthetic data can be leveraged to create artificial +> students for serving learning analytics models. Using the popular GAN model +> CTGAN and three LLMs- GPT2, DistilGPT2, and DialoGPT, we generate synthetic +> tabular student data. Our results demonstrate the strong potential of these +> methods to produce high-quality synthetic datasets that resemble real students +> data. To validate our findings, we apply a comprehensive set of utility +> evaluation metrics to assess the statistical and predictive performance of the +> synthetic data and compare the different generator models used, specially the +> performance of LLMs. Our study aims to provide the learning analytics community +> with valuable insights into the use of synthetic data, laying the groundwork +> for expanding the field methodological toolbox with new innovative approaches +> for learning analytics data generation. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on generating synthetic tabular data for learning analytics using GANs and LLMs, with LLMs being used for data generation rather than text-based interactions through prompt engineering, thus not meeting the core subject requirement. + +--- + +## [End-to-End Long Document Summarization using Gradient Caching](https://arxiv.org/abs/http://arxiv.org/abs/2501.01805v1) +**arXiv ID:** http://arxiv.org/abs/2501.01805v1 + +**Abstract:** +> Training transformer-based encoder-decoder models for long document +> summarization poses a significant challenge due to the quadratic memory +> consumption during training. Several approaches have been proposed to extend +> the input length at test time, but training with these approaches is still +> difficult, requiring truncation of input documents and causing a mismatch +> between training and test conditions. In this work, we propose CachED (Gradient +> $\textbf{Cach}$ing for $\textbf{E}$ncoder-$\textbf{D}$ecoder models), an +> approach that enables end-to-end training of existing transformer-based +> encoder-decoder models, using the entire document without truncation. +> Specifically, we apply non-overlapping sliding windows to input documents, +> followed by fusion in decoder. During backpropagation, the gradients are cached +> at the decoder and are passed through the encoder in chunks by re-computing the +> hidden vectors, similar to gradient checkpointing. In the experiments on long +> document summarization, we extend BART to CachED BART, processing more than +> 500K tokens during training and achieving superior performance without using +> any additional parameters. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing an optimization technique (Gradient Caching) for training transformer-based encoder-decoder models for long document summarization, rather than on engineering, designing, or optimizing prompts specifically for Large Language Models (LLMs). + +--- + +## [SDPO: Segment-Level Direct Preference Optimization for Social Agents](https://arxiv.org/abs/http://arxiv.org/abs/2501.01821v1) +**arXiv ID:** http://arxiv.org/abs/2501.01821v1 + +**Abstract:** +> Social agents powered by large language models (LLMs) can simulate human +> social behaviors but fall short in handling complex goal-oriented social +> dialogues. Direct Preference Optimization (DPO) has proven effective in +> aligning LLM behavior with human preferences across a variety of agent tasks. +> Existing DPO-based approaches for multi-turn interactions are divided into +> turn-level and session-level methods. The turn-level method is overly +> fine-grained, focusing exclusively on individual turns, while session-level +> methods are too coarse-grained, often introducing training noise. To address +> these limitations, we propose Segment-Level Direct Preference Optimization +> (SDPO), which focuses on specific key segments within interactions to optimize +> multi-turn agent behavior while minimizing training noise. Evaluations on the +> SOTOPIA benchmark demonstrate that SDPO-tuned agents consistently outperform +> both existing DPO-based methods and proprietary LLMs like GPT-4o, underscoring +> SDPO's potential to advance the social intelligence of LLM-based agents. We +> release our code and data at +> https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/SDPO. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on optimizing the behavior of social agents powered by LLMs through a new method (SDPO), but does not primarily investigate, analyze, or propose methods for improving LLM performance through the manipulation of textual input prompts. Instead, it optimizes agent behavior through preference optimization, making prompt engineering not the central focus. + +--- + +## [The Proof is in the Almond Cookies](https://arxiv.org/abs/http://arxiv.org/abs/2501.01827v1) +**arXiv ID:** http://arxiv.org/abs/2501.01827v1 + +**Abstract:** +> This paper presents a case study on how to process cooking recipes (and more +> generally, how-to instructions) in a way that makes it possible for a robot or +> artificial cooking assistant to support human chefs in the kitchen. Such AI +> assistants would be of great benefit to society, as they can help to sustain +> the autonomy of aging adults or people with a physical impairment, or they may +> reduce the stress in a professional kitchen. We propose a novel approach to +> computational recipe understanding that mimics the human sense-making process, +> which is narrative-based. Using an English recipe for almond crescent cookies +> as illustration, we show how recipes can be modelled as rich narrative +> structures by integrating various knowledge sources such as language +> processing, ontologies, and mental simulation. We show how such narrative +> structures can be used for (a) dealing with the challenges of recipe language, +> such as zero anaphora, (b) optimizing a robot's planning process, (c) measuring +> how well an AI system understands its current tasks, and (d) allowing recipe +> annotations to become language-independent. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on computational recipe understanding for robot-assisted cooking, not on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not demonstrate the impact of textual input prompts on LLM output. + +--- + +## [MoColl: Agent-Based Specific and General Model Collaboration for Image + Captioning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01834v2) +**arXiv ID:** http://arxiv.org/abs/2501.01834v2 + +**Abstract:** +> Image captioning is a critical task at the intersection of computer vision +> and natural language processing, with wide-ranging applications across various +> domains. For complex tasks such as diagnostic report generation, deep learning +> models require not only domain-specific image-caption datasets but also the +> incorporation of relevant general knowledge to provide contextual accuracy. +> Existing approaches exhibit inherent limitations: specialized models excel in +> capturing domain-specific details but lack generalization, while +> vision-language models (VLMs) built on large language models (LLMs) leverage +> general knowledge but struggle with domain-specific adaptation. To address +> these limitations, this paper proposes a novel agent-enhanced model +> collaboration framework, which we call MoColl, designed to effectively +> integrate domain-specific and general knowledge. Specifically, our approach is +> to decompose complex image captioning tasks into a series of interconnected +> question-answer subtasks. A trainable visual question answering (VQA) model is +> employed as a specialized tool to focus on domain-specific visual analysis, +> answering task-specific questions based on image content. Concurrently, an +> LLM-based agent with general knowledge formulates these questions and +> synthesizes the resulting question-answer pairs into coherent captions. Beyond +> its role in leveraging the VQA model, the agent further guides its training to +> enhance its domain-specific capabilities. Experimental results on radiology +> report generation validate the effectiveness of the proposed framework, +> demonstrating significant improvements in the quality of generated reports. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a collaborative framework for image captioning, leveraging both vision-language models (VLMs) and Large Language Models (LLMs), but does not centralize on prompt engineering for text-based interactions with LLMs. The LLM's role is in question formulation and synthesis, not in prompt engineering for improved LLM output. + +--- + +## [Practical machine learning is learning on small samples](https://arxiv.org/abs/http://arxiv.org/abs/2501.01836v1) +**arXiv ID:** http://arxiv.org/abs/2501.01836v1 + +**Abstract:** +> Based on limited observations, machine learning discerns a dependence which +> is expected to hold in the future. What makes it possible? Statistical learning +> theory imagines indefinitely increasing training sample to justify its +> approach. In reality, there is no infinite time or even infinite general +> population for learning. Here I argue that practical machine learning is based +> on an implicit assumption that underlying dependence is relatively ``smooth" : +> likely, there are no abrupt differences in feedback between cases with close +> data points. From this point of view learning shall involve selection of the +> hypothesis ``smoothly" approximating the training set. I formalize this as +> Practical learning paradigm. The paradigm includes terminology and rules for +> description of learners. Popular learners (local smoothing, k-NN, decision +> trees, Naive Bayes, SVM for classification and for regression) are shown here +> to be implementations of this paradigm. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria: it does not focus on prompt engineering for Large Language Models (LLMs), nor does it investigate improving LLM performance through textual input prompts. Instead, it discusses a general machine learning paradigm for learning from small samples, unrelated to LLMs or prompt engineering. + +--- + +## [Multi-Agent Conversational Online Learning for Adaptive LLM Response + Identification](https://arxiv.org/abs/http://arxiv.org/abs/2501.01849v1) +**arXiv ID:** http://arxiv.org/abs/2501.01849v1 + +**Abstract:** +> The remarkable generative capability of large language models (LLMs) has +> sparked a growing interest in automatically generating responses for different +> applications. Given the dynamic nature of user preferences and the uncertainty +> of LLM response performance, it is crucial to design efficient online learning +> algorithms to identify optimal LLM responses (i.e., high-quality responses that +> also meet user preferences). Most existing online algorithms adopt a +> centralized approach and fail to leverage explicit user preferences for more +> efficient and personalized LLM response identification. In contrast, this paper +> introduces \textit{MACO} (\underline{M}ulti-\underline{A}gent +> \underline{C}onversational \underline{O}nline Learning for Adaptive LLM +> Response Identification): 1) The online LLM response identification process is +> accelerated by multiple local agents (such as smartphones), while enhancing +> data privacy; 2) A novel conversational mechanism is proposed to adaptively +> conduct conversations for soliciting user preferences (e.g., a preference for a +> humorous tone over a serious one in generated responses), so to minimize +> uncertainty in preference estimation. Our theoretical analysis demonstrates +> that \cadi\ is near-optimal regarding cumulative regret. Additionally, \cadi\ +> offers reduced communication costs and computational complexity by eliminating +> the traditional, computing-intensive ``G-optimal design" found in previous +> works. Extensive experiments with the open LLM \textit{Llama}, coupled with two +> different embedding models from Google and OpenAI for text vector +> representation, demonstrate that \cadi\ significantly outperforms the current +> state-of-the-art in online LLM response identification. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on multi-agent conversational online learning for adapting LLM responses, not on prompt engineering specifically for Large Language Models (LLMs). It fails to meet the 'MUST' criteria by not investigating, analyzing, or proposing methods for improving LLM performance through the manipulation of textual input prompts. + +--- + +## [Accuracy Can Lie: On the Impact of Surrogate Model in Configuration + Tuning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01876v1) +**arXiv ID:** http://arxiv.org/abs/2501.01876v1 + +**Abstract:** +> To ease the expensive measurements during configuration tuning, it is natural +> to build a surrogate model as the replacement of the system, and thereby the +> configuration performance can be cheaply evaluated. Yet, a stereotype therein +> is that the higher the model accuracy, the better the tuning result would be. +> This "accuracy is all" belief drives our research community to build more and +> more accurate models and criticize a tuner for the inaccuracy of the model +> used. However, this practice raises some previously unaddressed questions, +> e.g., Do those somewhat small accuracy improvements reported in existing work +> really matter much to the tuners? What role does model accuracy play in the +> impact of tuning quality? To answer those related questions, we conduct one of +> the largest-scale empirical studies to date-running over the period of 13 +> months 24*7-that covers 10 models, 17 tuners, and 29 systems from the existing +> works while under four different commonly used metrics, leading to 13,612 cases +> of investigation. Surprisingly, our key findings reveal that the accuracy can +> lie: there are a considerable number of cases where higher accuracy actually +> leads to no improvement in the tuning outcomes (up to 58% cases under certain +> setting), or even worse, it can degrade the tuning quality (up to 24% cases +> under certain setting). We also discover that the chosen models in most +> proposed tuners are sub-optimal and that the required % of accuracy change to +> significantly improve tuning quality varies according to the range of model +> accuracy. Deriving from the fitness landscape analysis, we provide in-depth +> discussions of the rationale behind, offering several lessons learned as well +> as insights for future opportunities. Most importantly, this work poses a clear +> message to the community: we should take one step back from the natural +> "accuracy is all" belief for model-based configuration tuning. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on the impact of surrogate model accuracy in configuration tuning, which does not meet the 'MUST' criteria of primarily focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Virgo: A Preliminary Exploration on Reproducing o1-like MLLM](https://arxiv.org/abs/http://arxiv.org/abs/2501.01904v1) +**arXiv ID:** http://arxiv.org/abs/2501.01904v1 + +**Abstract:** +> Recently, slow-thinking reasoning systems, built upon large language models +> (LLMs), have garnered widespread attention by scaling the thinking time during +> inference. There is also growing interest in adapting this capability to +> multimodal large language models (MLLMs). Given that MLLMs handle more complex +> data semantics across different modalities, it is intuitively more challenging +> to implement multimodal slow-thinking systems. +> To address this issue, in this paper, we explore a straightforward approach +> by fine-tuning a capable MLLM with a small amount of textual long-form thought +> data, resulting in a multimodal slow-thinking system, Virgo (Visual reasoning +> with long thought). We find that these long-form reasoning processes, expressed +> in natural language, can be effectively transferred to MLLMs. Moreover, it +> seems that such textual reasoning data can be even more effective than visual +> reasoning data in eliciting the slow-thinking capacities of MLLMs. While this +> work is preliminary, it demonstrates that slow-thinking capacities are +> fundamentally associated with the language model component, which can be +> transferred across modalities or domains. This finding can be leveraged to +> guide the development of more powerful slow-thinking reasoning systems. We +> release our resources at https://github.com/RUCAIBox/Virgo. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on fine-tuning a multimodal large language model (MLLM) for slow-thinking reasoning, which falls under developing new training methods for LLM architectures, rather than engineering prompts specifically for improving LLM performance through textual input manipulation. + +--- + +## [Mitigating Hallucination for Large Vision Language Model by + Inter-Modality Correlation Calibration Decoding](https://arxiv.org/abs/http://arxiv.org/abs/2501.01926v1) +**arXiv ID:** http://arxiv.org/abs/2501.01926v1 + +**Abstract:** +> Large vision-language models (LVLMs) have shown remarkable capabilities in +> visual-language understanding for downstream multi-modal tasks. Despite their +> success, LVLMs still suffer from generating hallucinations in complex +> generation tasks, leading to inconsistencies between visual inputs and +> generated content. To address this issue, some approaches have introduced +> inference-time interventions, such as contrastive decoding and attention +> rectification, to reduce overreliance on language priors. However, these +> approaches overlook hallucinations stemming from spurious inter-modality +> correlations. In this paper, we propose an Inter-Modality Correlation +> Calibration Decoding (IMCCD) method to mitigate hallucinations in LVLMs in a +> training-free manner. In this method, we design a Cross-Modal Value-Enhanced +> Decoding(CMVED) module to alleviate hallucination by a novel contrastive +> decoding mechanism. During the estimation of distorted distribution, CMVED +> masks the value vectors associated with significant cross-modal attention +> weights, which address both uni-modality overreliance and misleading +> inter-modality correlations. Additionally, a Content-Driven Attention +> Refinement(CDAR) module refines cross-modal attention weights, guiding LVLMs to +> focus on important visual content. Experimental results on diverse +> hallucination benchmarks validate the superiority of our method over existing +> state-of-the-art techniques in reducing hallucinations in LVLM text generation. +> Our code will be available at https://github.com/lijm48/IMCCD. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on mitigating hallucinations in Large Vision Language Models (LVLMs) through decoding mechanisms, not on the engineering, design, or optimization of textual input prompts for Large Language Models (LLMs), and does not provide concrete examples of prompt manipulation impacting LLM output." +} + +--- + +## [Abstractive Text Summarization for Contemporary Sanskrit Prose: Issues + and Challenges](https://arxiv.org/abs/http://arxiv.org/abs/2501.01933v1) +**arXiv ID:** http://arxiv.org/abs/2501.01933v1 + +**Abstract:** +> This thesis presents Abstractive Text Summarization models for contemporary +> Sanskrit prose. The first chapter, titled Introduction, presents the motivation +> behind this work, the research questions, and the conceptual framework. +> Sanskrit is a low-resource inflectional language. The key research question +> that this thesis investigates is what the challenges in developing an +> abstractive TS for Sanskrit. To answer the key research questions, +> sub-questions based on four different themes have been posed in this work. The +> second chapter, Literature Review, surveys the previous works done. The third +> chapter, data preparation, answers the remaining three questions from the third +> theme. It reports the data collection and preprocessing challenges for both +> language model and summarization model trainings. The fourth chapter reports +> the training and inference of models and the results obtained therein. This +> research has initiated a pipeline for Sanskrit abstractive text summarization +> and has reported the challenges faced at every stage of the development. The +> research questions based on every theme have been answered to answer the key +> research question. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing Abstractive Text Summarization models for Sanskrit prose, discussing language-specific challenges and model training, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [Cold-Start Recommendation towards the Era of Large Language Models + (LLMs): A Comprehensive Survey and Roadmap](https://arxiv.org/abs/http://arxiv.org/abs/2501.01945v2) +**arXiv ID:** http://arxiv.org/abs/2501.01945v2 + +**Abstract:** +> Cold-start problem is one of the long-standing challenges in recommender +> systems, focusing on accurately modeling new or interaction-limited users or +> items to provide better recommendations. Due to the diversification of internet +> platforms and the exponential growth of users and items, the importance of +> cold-start recommendation (CSR) is becoming increasingly evident. At the same +> time, large language models (LLMs) have achieved tremendous success and possess +> strong capabilities in modeling user and item information, providing new +> potential for cold-start recommendations. However, the research community on +> CSR still lacks a comprehensive review and reflection in this field. Based on +> this, in this paper, we stand in the context of the era of large language +> models and provide a comprehensive review and discussion on the roadmap, +> related literature, and future directions of CSR. Specifically, we have +> conducted an exploration of the development path of how existing CSR utilizes +> information, from content features, graph relations, and domain information, to +> the world knowledge possessed by large language models, aiming to provide new +> insights for both the research and industrial communities on CSR. Related +> resources of cold-start recommendations are collected and continuously updated +> for the community in +> https://github.com/YuanchenBei/Awesome-Cold-Start-Recommendation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on cold-start recommendation systems utilizing Large Language Models (LLMs), but does not concentrate on prompt engineering for text-based interactions with LLMs. Instead, it leverages LLMs as a component for enhancing recommender systems, which does not meet the core subject requirement of prompt engineering being the central focus. + +--- + +## [MADGEN: Mass-Spec attends to De Novo Molecular generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01950v2) +**arXiv ID:** http://arxiv.org/abs/2501.01950v2 + +**Abstract:** +> The annotation (assigning structural chemical identities) of MS/MS spectra +> remains a significant challenge due to the enormous molecular diversity in +> biological samples and the limited scope of reference databases. Currently, the +> vast majority of spectral measurements remain in the "dark chemical space" +> without structural annotations. To improve annotation, we propose MADGEN +> (Mass-spec Attends to De Novo Molecular GENeration), a scaffold-based method +> for de novo molecular structure generation guided by mass spectrometry data. +> MADGEN operates in two stages: scaffold retrieval and spectra-conditioned +> molecular generation starting with the scaffold. In the first stage, given an +> MS/MS spectrum, we formulate scaffold retrieval as a ranking problem and employ +> contrastive learning to align mass spectra with candidate molecular scaffolds. +> In the second stage, starting from the retrieved scaffold, we employ the MS/MS +> spectrum to guide an attention-based generative model to generate the final +> molecule. Our approach constrains the molecular generation search space, +> reducing its complexity and improving generation accuracy. We evaluate MADGEN +> on three datasets (NIST23, CANOPUS, and MassSpecGym) and evaluate MADGEN's +> performance with a predictive scaffold retriever and with an oracle retriever. +> We demonstrate the effectiveness of using attention to integrate spectral +> information throughout the generation process to achieve strong results with +> the oracle retriever. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on de novo molecular structure generation using mass spectrometry data, with attention-based generative models, and does not meet the core requirement of focusing on prompt engineering specifically for Large Language Models (LLMs) or manipulating textual input prompts to improve LLM performance. + +--- + +## [MixGCN: Scalable GCN Training by Mixture of Parallelism and Mixture of + Accelerators](https://arxiv.org/abs/http://arxiv.org/abs/2501.01951v2) +**arXiv ID:** http://arxiv.org/abs/2501.01951v2 + +**Abstract:** +> Graph convolutional networks (GCNs) have demonstrated superiority in +> graph-based learning tasks. However, training GCNs on full graphs is +> particularly challenging, due to the following two challenges: (1) the +> associated feature tensors can easily explode the memory and block the +> communication bandwidth of modern accelerators, and (2) the computation +> workflow in training GCNs alternates between sparse and dense matrix +> operations, complicating the efficient utilization of computational resources. +> Existing solutions for scalable distributed full-graph GCN training mostly +> adopt partition parallelism, which is unsatisfactory as they only partially +> address the first challenge while incurring scaled-out communication volume. To +> this end, we propose MixGCN aiming to simultaneously address both the +> aforementioned challenges towards GCN training. To tackle the first challenge, +> MixGCN integrates mixture of parallelism. Both theoretical and empirical +> analysis verify its constant communication volumes and enhanced balanced +> workload; For handling the second challenge, we consider mixture of +> accelerators (i.e., sparse and dense accelerators) with a dedicated accelerator +> for GCN training and a fine-grain pipeline. Extensive experiments show that +> MixGCN achieves boosted training efficiency and scalability. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on scalable training of Graph Convolutional Networks (GCNs) using parallelism and mixture of accelerators, with no mention of Large Language Models (LLMs) or prompt engineering, thus failing to meet all 'MUST' criteria. + +--- + +## [SmartSpatial: Enhancing the 3D Spatial Arrangement Capabilities of + Stable Diffusion Models and Introducing a Novel 3D Spatial Evaluation + Framework](https://arxiv.org/abs/http://arxiv.org/abs/2501.01998v1) +**arXiv ID:** http://arxiv.org/abs/2501.01998v1 + +**Abstract:** +> Stable Diffusion models have made remarkable strides in generating +> photorealistic images from text prompts but often falter when tasked with +> accurately representing complex spatial arrangements, particularly involving +> intricate 3D relationships. To address this limitation, we introduce +> SmartSpatial, an innovative approach that enhances the spatial arrangement +> capabilities of Stable Diffusion models through 3D-aware conditioning and +> attention-guided mechanisms. SmartSpatial incorporates depth information and +> employs cross-attention control to ensure precise object placement, delivering +> notable improvements in spatial accuracy metrics. In conjunction with +> SmartSpatial, we present SmartSpatialEval, a comprehensive evaluation framework +> designed to assess spatial relationships. This framework utilizes +> vision-language models and graph-based dependency parsing for performance +> analysis. Experimental results on the COCO and SpatialPrompts datasets show +> that SmartSpatial significantly outperforms existing methods, setting new +> benchmarks for spatial arrangement accuracy in image generation. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on enhancing Stable Diffusion models for 3D spatial arrangement in image generation, which falls under image generation driven by generative AI models other than Large Language Models (LLMs) for text generation, violating the 'MUST NOT' criteria." +} + +--- + +## [Multi-Task Semantic Communication With Graph Attention-Based Feature + Correlation Extraction](https://arxiv.org/abs/http://arxiv.org/abs/2501.02006v1) +**arXiv ID:** http://arxiv.org/abs/2501.02006v1 + +**Abstract:** +> Multi-task semantic communication can serve multiple learning tasks using a +> shared encoder model. Existing models have overlooked the intricate +> relationships between features extracted during an encoding process of tasks. +> This paper presents a new graph attention inter-block (GAI) module to the +> encoder/transmitter of a multi-task semantic communication system, which +> enriches the features for multiple tasks by embedding the intermediate outputs +> of encoding in the features, compared to the existing techniques. The key idea +> is that we interpret the outputs of the intermediate feature extraction blocks +> of the encoder as the nodes of a graph to capture the correlations of the +> intermediate features. Another important aspect is that we refine the node +> representation using a graph attention mechanism to extract the correlations +> and a multi-layer perceptron network to associate the node representations with +> different tasks. Consequently, the intermediate features are weighted and +> embedded into the features transmitted for executing multiple tasks at the +> receiver. Experiments demonstrate that the proposed model surpasses the most +> competitive and publicly available models by 11.4% on the CityScapes 2Task +> dataset and outperforms the established state-of-the-art by 3.97% on the NYU V2 +> 3Task dataset, respectively, when the bandwidth ratio of the communication +> channel (i.e., compression level for transmission over the channel) is as +> constrained as 1 12 . + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on developing a new module for a multi-task semantic communication system, optimizing feature extraction with graph attention, and does not primarily investigate prompt engineering for Large Language Models (LLMs) or analyze the impact of textual input prompts on LLM output." +} + +--- + +## [TART: Token-based Architecture Transformer for Neural Network + Performance Prediction](https://arxiv.org/abs/http://arxiv.org/abs/2501.02007v1) +**arXiv ID:** http://arxiv.org/abs/2501.02007v1 + +**Abstract:** +> In the realm of neural architecture design, achieving high performance is +> largely reliant on the manual expertise of researchers. Despite the emergence +> of Neural Architecture Search (NAS) as a promising technique for automating +> this process, current NAS methods still require human input to expand the +> search space and cannot generate new architectures. This paper explores the +> potential of Transformers in comprehending neural architectures and their +> performance, with the objective of establishing the foundation for utilizing +> Transformers to generate novel networks. We propose the Token-based +> Architecture Transformer (TART), which predicts neural network performance +> without the need to train candidate networks. TART attains state-of-the-art +> performance on the DeepNets-1M dataset for performance prediction tasks without +> edge information, indicating the potential of Transformers to aid in +> discovering novel and high-performing neural architectures. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on utilizing Transformers for Neural Architecture Search (NAS) and predicting neural network performance, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), failing to meet the mandatory criteria. + +--- + +## [Cross-model Transferability among Large Language Models on the Platonic + Representations of Concepts](https://arxiv.org/abs/http://arxiv.org/abs/2501.02009v1) +**arXiv ID:** http://arxiv.org/abs/2501.02009v1 + +**Abstract:** +> Understanding the inner workings of Large Language Models (LLMs) is a +> critical research frontier. Prior research has shown that a single LLM's +> concept representations can be captured as steering vectors (SVs), enabling the +> control of LLM behavior (e.g., towards generating harmful content). Our work +> takes a novel approach by exploring the intricate relationships between concept +> representations across different LLMs, drawing an intriguing parallel to +> Plato's Allegory of the Cave. In particular, we introduce a linear +> transformation method to bridge these representations and present three key +> findings: 1) Concept representations across different LLMs can be effectively +> aligned using simple linear transformations, enabling efficient cross-model +> transfer and behavioral control via SVs. 2) This linear transformation +> generalizes across concepts, facilitating alignment and control of SVs +> representing different concepts across LLMs. 3) A weak-to-strong +> transferability exists between LLM concept representations, whereby SVs +> extracted from smaller LLMs can effectively control the behavior of larger +> LLMs. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on understanding concept representations across different LLMs and developing a linear transformation method for alignment, rather than specifically engineering or optimizing prompts for improving LLM text generation performance through textual input manipulation." +} + +--- + +## [Machine Learning-Based Differential Diagnosis of Parkinson's Disease + Using Kinematic Feature Extraction and Selection](https://arxiv.org/abs/http://arxiv.org/abs/2501.02014v1) +**arXiv ID:** http://arxiv.org/abs/2501.02014v1 + +**Abstract:** +> Parkinson's disease (PD), the second most common neurodegenerative disorder, +> is characterized by dopaminergic neuron loss and the accumulation of abnormal +> synuclein. PD presents both motor and non-motor symptoms that progressively +> impair daily functioning. The severity of these symptoms is typically assessed +> using the MDS-UPDRS rating scale, which is subjective and dependent on the +> physician's experience. Additionally, PD shares symptoms with other +> neurodegenerative diseases, such as progressive supranuclear palsy (PSP) and +> multiple system atrophy (MSA), complicating accurate diagnosis. To address +> these diagnostic challenges, we propose a machine learning-based system for +> differential diagnosis of PD, PSP, MSA, and healthy controls (HC). This system +> utilizes a kinematic feature-based hierarchical feature extraction and +> selection approach. Initially, 18 kinematic features are extracted, including +> two newly proposed features: Thumb-to-index vector velocity and acceleration, +> which provide insights into motor control patterns. In addition, 41 statistical +> features were extracted here from each kinematic feature, including some new +> approaches such as Average Absolute Change, Rhythm, Amplitude, Frequency, +> Standard Deviation of Frequency, and Slope. Feature selection is performed +> using One-way ANOVA to rank features, followed by Sequential Forward Floating +> Selection (SFFS) to identify the most relevant ones, aiming to reduce the +> computational complexity. The final feature set is used for classification, +> achieving a classification accuracy of 66.67% for each dataset and 88.89% for +> each patient, with particularly high performance for the MSA and HC groups +> using the SVM algorithm. This system shows potential as a rapid and accurate +> diagnostic tool in clinical practice, though further data collection and +> refinement are needed to enhance its reliability. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on machine learning for differential diagnosis of Parkinson's disease using kinematic feature extraction and selection, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, failing to meet all 'MUST' criteria. + +--- + +## [Enhancing Uncertainty Modeling with Semantic Graph for Hallucination + Detection](https://arxiv.org/abs/http://arxiv.org/abs/2501.02020v1) +**arXiv ID:** http://arxiv.org/abs/2501.02020v1 + +**Abstract:** +> Large Language Models (LLMs) are prone to hallucination with non-factual or +> unfaithful statements, which undermines the applications in real-world +> scenarios. Recent researches focus on uncertainty-based hallucination +> detection, which utilizes the output probability of LLMs for uncertainty +> calculation and does not rely on external knowledge or frequent sampling from +> LLMs. Whereas, most approaches merely consider the uncertainty of each +> independent token, while the intricate semantic relations among tokens and +> sentences are not well studied, which limits the detection of hallucination +> that spans over multiple tokens and sentences in the passage. In this paper, we +> propose a method to enhance uncertainty modeling with semantic graph for +> hallucination detection. Specifically, we first construct a semantic graph that +> well captures the relations among entity tokens and sentences. Then, we +> incorporate the relations between two entities for uncertainty propagation to +> enhance sentence-level hallucination detection. Given that hallucination occurs +> due to the conflict between sentences, we further present a graph-based +> uncertainty calibration method that integrates the contradiction probability of +> the sentence with its neighbors in the semantic graph for uncertainty +> calculation. Extensive experiments on two datasets show the great advantages of +> our proposed approach. In particular, we obtain substantial improvements with +> 19.78% in passage-level hallucination detection. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing uncertainty modeling for hallucination detection in LLMs, rather than prompt engineering, optimization, or manipulation of textual input prompts to improve LLM performance, thus not meeting the core subject requirement. + +--- + +## [Weakly Supervised Learning on Large Graphs](https://arxiv.org/abs/http://arxiv.org/abs/2501.02021v1) +**arXiv ID:** http://arxiv.org/abs/2501.02021v1 + +**Abstract:** +> Graph classification plays a pivotal role in various domains, including +> pathology, where images can be represented as graphs.In this domain, images can +> be represented as graphs, where nodes might represent individual nuclei, and +> edges capture the spatial or functional relationships between them. Often, the +> overall label of the graph, such as a cancer type or disease state, is +> determined by patterns within smaller, localized regions of the image. This +> work introduces a weakly-supervised graph classification framework leveraging +> two subgraph extraction techniques: (1) Sliding-window approach (2) BFS-based +> approach. Subgraphs are processed using a Graph Attention Network (GAT), which +> employs attention mechanisms to identify the most informative subgraphs for +> classification. Weak supervision is achieved by propagating graph-level labels +> to subgraphs, eliminating the need for detailed subgraph annotations. + +**Decision Explanation:** +Original decision: REJECT +This paper focuses on weakly supervised learning for graph classification using Graph Attention Networks, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions, thus failing to meet the primary 'MUST' criteria. + +--- + +## [CarbonChat: Large Language Model-Based Corporate Carbon Emission + Analysis and Climate Knowledge Q&A System](https://arxiv.org/abs/http://arxiv.org/abs/2501.02031v1) +**arXiv ID:** http://arxiv.org/abs/2501.02031v1 + +**Abstract:** +> As the impact of global climate change intensifies, corporate carbon +> emissions have become a focal point of global attention. In response to issues +> such as the lag in climate change knowledge updates within large language +> models, the lack of specialization and accuracy in traditional augmented +> generation architectures for complex problems, and the high cost and time +> consumption of sustainability report analysis, this paper proposes CarbonChat: +> Large Language Model-based corporate carbon emission analysis and climate +> knowledge Q&A system, aimed at achieving precise carbon emission analysis and +> policy understanding.First, a diversified index module construction method is +> proposed to handle the segmentation of rule-based and long-text documents, as +> well as the extraction of structured data, thereby optimizing the parsing of +> key information.Second, an enhanced self-prompt retrieval-augmented generation +> architecture is designed, integrating intent recognition, structured reasoning +> chains, hybrid retrieval, and Text2SQL, improving the efficiency of semantic +> understanding and query conversion.Next, based on the greenhouse gas accounting +> framework, 14 dimensions are established for carbon emission analysis, enabling +> report summarization, relevance evaluation, and customized responses.Finally, +> through a multi-layer chunking mechanism, timestamps, and hallucination +> detection features, the accuracy and verifiability of the analysis results are +> ensured, reducing hallucination rates and enhancing the precision of the +> responses. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a corporate carbon emission analysis and climate knowledge Q&A system using LLMs, rather than specifically on prompt engineering for LLMs. While it mentions a 'self-prompt retrieval-augmented generation architecture', prompt engineering is not the central concern, but rather a component of the larger system. + +--- + +## [3D Cloud reconstruction through geospatially-aware Masked Autoencoders](https://arxiv.org/abs/http://arxiv.org/abs/2501.02035v1) +**arXiv ID:** http://arxiv.org/abs/2501.02035v1 + +**Abstract:** +> Clouds play a key role in Earth's radiation balance with complex effects that +> introduce large uncertainties into climate models. Real-time 3D cloud data is +> essential for improving climate predictions. This study leverages geostationary +> imagery from MSG/SEVIRI and radar reflectivity measurements of cloud profiles +> from CloudSat/CPR to reconstruct 3D cloud structures. We first apply +> self-supervised learning (SSL) methods-Masked Autoencoders (MAE) and +> geospatially-aware SatMAE on unlabelled MSG images, and then fine-tune our +> models on matched image-profile pairs. Our approach outperforms +> state-of-the-art methods like U-Nets, and our geospatial encoding further +> improves prediction results, demonstrating the potential of SSL for cloud +> reconstruction. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on 3D cloud reconstruction using geospatially-aware Masked Autoencoders, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria." +} + +--- + +## [Architecture for Trajectory-Based Fishing Ship Classification with AIS + Data](https://arxiv.org/abs/http://arxiv.org/abs/2501.02038v1) +**arXiv ID:** http://arxiv.org/abs/2501.02038v1 + +**Abstract:** +> This paper proposes a data preparation process for managing real-world +> kinematic data and detecting fishing vessels. The solution is a binary +> classification that classifies ship trajectories into either fishing or +> non-fishing ships. The data used are characterized by the typical problems +> found in classic data mining applications using real-world data, such as noise +> and inconsistencies. The two classes are also clearly unbalanced in the data, a +> problem which is addressed using algorithms that resample the instances. For +> classification, a series of features are extracted from spatiotemporal data +> that represent the trajectories of the ships, available from sequences of +> Automatic Identification System (AIS) reports. These features are proposed for +> the modelling of ship behavior but, because they do not contain context-related +> information, the classification can be applied in other scenarios. +> Experimentation shows that the proposed data preparation process is useful for +> the presented classification problem. In addition, positive results are +> obtained using minimal information. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing an architecture for ship classification using AIS data and kinematic data, with no mention of Large Language Models (LLMs), prompt engineering, or text-based interactions, failing to meet the primary 'MUST' criteria. + +--- + +## [An Investigation into Value Misalignment in LLM-Generated Texts for + Cultural Heritage](https://arxiv.org/abs/http://arxiv.org/abs/2501.02039v1) +**arXiv ID:** http://arxiv.org/abs/2501.02039v1 + +**Abstract:** +> As Large Language Models (LLMs) become increasingly prevalent in tasks +> related to cultural heritage, such as generating descriptions of historical +> monuments, translating ancient texts, preserving oral traditions, and creating +> educational content, their ability to produce accurate and culturally aligned +> texts is being increasingly relied upon by users and researchers. However, +> cultural value misalignments may exist in generated texts, such as the +> misrepresentation of historical facts, the erosion of cultural identity, and +> the oversimplification of complex cultural narratives, which may lead to severe +> consequences. Therefore, investigating value misalignment in the context of LLM +> for cultural heritage is crucial for mitigating these risks, yet there has been +> a significant lack of systematic and comprehensive study and investigation in +> this area. To fill this gap, we systematically assess the reliability of LLMs +> in generating culturally aligned texts for cultural heritage-related tasks. We +> conduct a comprehensive evaluation by compiling an extensive set of 1066 query +> tasks covering 5 widely recognized categories with 17 aspects within the +> knowledge framework of cultural heritage across 5 open-source LLMs, and examine +> both the type and rate of cultural value misalignments in the generated texts. +> Using both automated and manual approaches, we effectively detect and analyze +> the cultural value misalignments in LLM-generated texts. Our findings are +> concerning: over 65% of the generated texts exhibit notable cultural +> misalignments, with certain tasks demonstrating almost complete misalignment +> with key cultural values. Beyond these findings, this paper introduces a +> benchmark dataset and a comprehensive evaluation workflow that can serve as a +> valuable resource for future research aimed at enhancing the cultural +> sensitivity and reliability of LLMs. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on evaluating cultural value misalignment in LLM-generated texts for cultural heritage, rather than investigating methods for improving LLM performance through the manipulation of textual input prompts, and does not provide concrete examples of prompt engineering techniques. + +--- + +## [A Separable Self-attention Inspired by the State Space Model for + Computer Vision](https://arxiv.org/abs/http://arxiv.org/abs/2501.02040v1) +**arXiv ID:** http://arxiv.org/abs/2501.02040v1 + +**Abstract:** +> Mamba is an efficient State Space Model (SSM) with linear computational +> complexity. Although SSMs are not suitable for handling non-causal data, Vision +> Mamba (ViM) methods still demonstrate good performance in tasks such as image +> classification and object detection. Recent studies have shown that there is a +> rich theoretical connection between state space models and attention variants. +> We propose a novel separable self attention method, for the first time +> introducing some excellent design concepts of Mamba into separable +> self-attention. To ensure a fair comparison with ViMs, we introduce VMINet, a +> simple yet powerful prototype architecture, constructed solely by stacking our +> novel attention modules with the most basic down-sampling layers. Notably, +> VMINet differs significantly from the conventional Transformer architecture. +> Our experiments demonstrate that VMINet has achieved competitive results on +> image classification and high-resolution dense prediction tasks.Code is +> available at: \url{https://github.com/yws-wxs/VMINet}. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on computer vision, proposing a novel separable self-attention method and its application in image classification and object detection, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts. + +--- + +## [MRG: A Multi-Robot Manufacturing Digital Scene Generation Method Using + Multi-Instance Point Cloud Registration](https://arxiv.org/abs/http://arxiv.org/abs/2501.02041v1) +**arXiv ID:** http://arxiv.org/abs/2501.02041v1 + +**Abstract:** +> A high-fidelity digital simulation environment is crucial for accurately +> replicating physical operational processes. However, inconsistencies between +> simulation and physical environments result in low confidence in simulation +> outcomes, limiting their effectiveness in guiding real-world production. Unlike +> the traditional step-by-step point cloud "segmentation-registration" generation +> method, this paper introduces, for the first time, a novel Multi-Robot +> Manufacturing Digital Scene Generation (MRG) method that leverages +> multi-instance point cloud registration, specifically within manufacturing +> scenes. Tailored to the characteristics of industrial robots and manufacturing +> settings, an instance-focused transformer module is developed to delineate +> instance boundaries and capture correlations between local regions. +> Additionally, a hypothesis generation module is proposed to extract target +> instances while preserving key features. Finally, an efficient screening and +> optimization algorithm is designed to refine the final registration results. +> Experimental evaluations on the Scan2CAD and Welding-Station datasets +> demonstrate that: (1) the proposed method outperforms existing multi-instance +> point cloud registration techniques; (2) compared to state-of-the-art methods, +> the Scan2CAD dataset achieves improvements in MR and MP by 12.15% and 17.79%, +> respectively; and (3) on the Welding-Station dataset, MR and MP are enhanced by +> 16.95% and 24.15%, respectively. This work marks the first application of +> multi-instance point cloud registration in manufacturing scenes, significantly +> advancing the precision and reliability of digital simulation environments for +> industrial applications. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multi-instance point cloud registration for digital scene generation in manufacturing, with no indication of involving Large Language Models (LLMs) or prompt engineering for text-based interactions, thus failing to meet the primary 'MUST' criteria. + +--- + +## [Advancing Pancreatic Cancer Prediction with a Next Visit Token + Prediction Head on top of Med-BERT](https://arxiv.org/abs/http://arxiv.org/abs/2501.02044v1) +**arXiv ID:** http://arxiv.org/abs/2501.02044v1 + +**Abstract:** +> Background: Recently, numerous foundation models pretrained on extensive data +> have demonstrated efficacy in disease prediction using Electronic Health +> Records (EHRs). However, there remains some unanswered questions on how to best +> utilize such models especially with very small fine-tuning cohorts. Methods: We +> utilized Med-BERT, an EHR-specific foundation model, and reformulated the +> disease binary prediction task into a token prediction task and a next visit +> mask token prediction task to align with Med-BERT's pretraining task format in +> order to improve the accuracy of pancreatic cancer (PaCa) prediction in both +> few-shot and fully supervised settings. Results: The reformulation of the task +> into a token prediction task, referred to as Med-BERT-Sum, demonstrates +> slightly superior performance in both few-shot scenarios and larger data +> samples. Furthermore, reformulating the prediction task as a Next Visit Mask +> Token Prediction task (Med-BERT-Mask) significantly outperforms the +> conventional Binary Classification (BC) prediction task (Med-BERT-BC) by 3% to +> 7% in few-shot scenarios with data sizes ranging from 10 to 500 samples. These +> findings highlight that aligning the downstream task with Med-BERT's +> pretraining objectives substantially enhances the model's predictive +> capabilities, thereby improving its effectiveness in predicting both rare and +> common diseases. Conclusion: Reformatting disease prediction tasks to align +> with the pretraining of foundation models enhances prediction accuracy, leading +> to earlier detection and timely intervention. This approach improves treatment +> effectiveness, survival rates, and overall patient outcomes for PaCa and +> potentially other cancers. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on optimizing disease prediction tasks using Med-BERT, a foundation model, by reformulating the prediction task, rather than specifically engineering or optimizing prompts for Large Language Models (LLMs). The core subject is not prompt engineering for text-based interactions with LLMs, but rather adapting task formats to improve model performance in a medical context, which is explicitly excluded. + +--- + +## [ArtCrafter: Text-Image Aligning Style Transfer via Embedding Reframing](https://arxiv.org/abs/http://arxiv.org/abs/2501.02064v1) +**arXiv ID:** http://arxiv.org/abs/2501.02064v1 + +**Abstract:** +> Recent years have witnessed significant advancements in text-guided style +> transfer, primarily attributed to innovations in diffusion models. These models +> excel in conditional guidance, utilizing text or images to direct the sampling +> process. However, despite their capabilities, direct conditional guidance +> approaches often face challenges in balancing the expressiveness of textual +> semantics with the diversity of output results while capturing stylistic +> features. To address these challenges, we introduce ArtCrafter, a novel +> framework for text-to-image style transfer. Specifically, we introduce an +> attention-based style extraction module, meticulously engineered to capture the +> subtle stylistic elements within an image. This module features a multi-layer +> architecture that leverages the capabilities of perceiver attention mechanisms +> to integrate fine-grained information. Additionally, we present a novel +> text-image aligning augmentation component that adeptly balances control over +> both modalities, enabling the model to efficiently map image and text +> embeddings into a shared feature space. We achieve this through attention +> operations that enable smooth information flow between modalities. Lastly, we +> incorporate an explicit modulation that seamlessly blends multimodal enhanced +> embeddings with original embeddings through an embedding reframing design, +> empowering the model to generate diverse outputs. Extensive experiments +> demonstrate that ArtCrafter yields impressive results in visual stylization, +> exhibiting exceptional levels of stylistic intensity, controllability, and +> diversity. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on text-to-image style transfer using diffusion models, which falls under image generation driven by generative AI models other than Large Language Models (LLMs) for text generation, violating the 'Papers MUST NOT' criterion 2. + +--- + +## [The interplay between domain specialization and model size: a case study + in the legal domain](https://arxiv.org/abs/http://arxiv.org/abs/2501.02068v1) +**arXiv ID:** http://arxiv.org/abs/2501.02068v1 + +**Abstract:** +> Scaling laws for language models so far focused on finding the +> compute-optimal model size and token count for training from scratch. However, +> achieving this optimal balance requires significant compute resources due to +> the extensive data demands when training models from randomly-initialized +> weights. Continual pre-training offers a cost-effective alternative, leveraging +> the compute investment from pre-trained models to incorporate new knowledge +> without requiring extensive new data. Recent findings suggest that data quality +> influences constants in scaling laws, thereby altering the optimal +> parameter-token allocation ratio. Building on this insight, we investigate the +> interplay between domain specialization and model size during continual +> pre-training under compute-constrained scenarios. Our goal is to identify a +> compute-efficient training regime for this scenario and, potentially, detect +> patterns in this interplay that can be generalized across different model sizes +> and domains. To compare general and specialized training, we filtered a +> web-based dataset to extract legal domain data. We pre-trained models with +> 1.5B, 3B, 7B and 14B parameters on both the unfiltered and filtered datasets, +> then evaluated their performance on legal exams. Results show that as model +> size increases, the compute-effectiveness gap between specialized and general +> models widens. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on the interplay between domain specialization and model size for continual pre-training, which falls under developing new training methods for LLMs, rather than prompt engineering for text-based interactions with LLMs. + +--- + +## [On the Statistical Complexity for Offline and Low-Adaptive Reinforcement + Learning with Structures](https://arxiv.org/abs/http://arxiv.org/abs/2501.02089v1) +**arXiv ID:** http://arxiv.org/abs/2501.02089v1 + +**Abstract:** +> This article reviews the recent advances on the statistical foundation of +> reinforcement learning (RL) in the offline and low-adaptive settings. We will +> start by arguing why offline RL is the appropriate model for almost any +> real-life ML problems, even if they have nothing to do with the recent AI +> breakthroughs that use RL. Then we will zoom into two fundamental problems of +> offline RL: offline policy evaluation (OPE) and offline policy learning (OPL). +> It may be surprising to people that tight bounds for these problems were not +> known even for tabular and linear cases until recently. We delineate the +> differences between worst-case minimax bounds and instance-dependent bounds. We +> also cover key algorithmic ideas and proof techniques behind near-optimal +> instance-dependent methods in OPE and OPL. Finally, we discuss the limitations +> of offline RL and review a burgeoning problem of \emph{low-adaptive +> exploration} which addresses these limitations by providing a sweet middle +> ground between offline and online RL. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary focus criteria, as it discusses reinforcement learning (RL) in offline and low-adaptive settings, with no evident focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs), nor does it investigate improving LLM performance through textual input prompt manipulation. + +--- + +## [Online Detection of Water Contamination Under Concept Drift](https://arxiv.org/abs/http://arxiv.org/abs/2501.02107v1) +**arXiv ID:** http://arxiv.org/abs/2501.02107v1 + +**Abstract:** +> Water Distribution Networks (WDNs) are vital infrastructures, and +> contamination poses serious public health risks. Harmful substances can +> interact with disinfectants like chlorine, making chlorine monitoring essential +> for detecting contaminants. However, chlorine sensors often become unreliable +> and require frequent calibration. This study introduces the Dual-Threshold +> Anomaly and Drift Detection (AD&DD) method, an unsupervised approach combining +> a dual-threshold drift detection mechanism with an LSTM-based Variational +> Autoencoder(LSTM-VAE) for real-time contamination detection. Tested on two +> realistic WDNs, AD&DD effectively identifies anomalies with sensor offsets as +> concept drift, and outperforms other methods. A proposed decentralized +> architecture enables accurate contamination detection and localization by +> deploying AD&DD on selected nodes. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the 'MUST' criteria as it focuses on detecting water contamination using an LSTM-based Variational Autoencoder, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [Siamese Networks for Cat Re-Identification: Exploring Neural Models for + Cat Instance Recognition](https://arxiv.org/abs/http://arxiv.org/abs/2501.02112v1) +**arXiv ID:** http://arxiv.org/abs/2501.02112v1 + +**Abstract:** +> Street cats in urban areas often rely on human intervention for survival, +> leading to challenges in population control and welfare management. In April +> 2023, Hello Inc., a Chinese urban mobility company, launched the Hello Street +> Cat initiative to address these issues. The project deployed over 21,000 smart +> feeding stations across 14 cities in China, integrating livestreaming cameras +> and treat dispensers activated through user donations. It also promotes the +> Trap-Neuter-Return (TNR) method, supported by a community-driven platform, +> HelloStreetCatWiki, where volunteers catalog and identify cats. However, manual +> identification is inefficient and unsustainable, creating a need for automated +> solutions. This study explores Deep Learning-based models for re-identifying +> street cats in the Hello Street Cat initiative. A dataset of 2,796 images of 69 +> cats was used to train Siamese Networks with EfficientNetB0, MobileNet and +> VGG16 as base models, evaluated under contrastive and triplet loss functions. +> VGG16 paired with contrastive loss emerged as the most effective configuration, +> achieving up to 97% accuracy and an F1 score of 0.9344 during testing. The +> approach leverages image augmentation and dataset refinement to overcome +> challenges posed by limited data and diverse visual variations. These findings +> underscore the potential of automated cat re-identification to streamline +> population monitoring and welfare efforts. By reducing reliance on manual +> processes, the method offers a scalable and reliable solution for +> communitydriven initiatives. Future research will focus on expanding datasets +> and developing real-time implementations to enhance practicality in large-scale +> deployments. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on image recognition for cat re-identification using Siamese Networks and Deep Learning models, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [AVTrustBench: Assessing and Enhancing Reliability and Robustness in + Audio-Visual LLMs](https://arxiv.org/abs/http://arxiv.org/abs/2501.02135v1) +**arXiv ID:** http://arxiv.org/abs/2501.02135v1 + +**Abstract:** +> With the rapid advancement of Multi-modal Large Language Models (MLLMs), +> several diagnostic benchmarks have recently been developed to assess these +> models' multi-modal reasoning proficiency. However, these benchmarks are +> restricted to assessing primarily the visual aspect and do not examine the +> holistic audio-visual (AV) understanding. Moreover, currently, there are no +> benchmarks that investigate the capabilities of AVLLMs to calibrate their +> responses when presented with perturbed inputs. To this end, we introduce +> Audio-Visual Trustworthiness assessment Benchmark (AVTrustBench), comprising +> 600K samples spanning over 9 meticulously crafted tasks, evaluating the +> capabilities of AVLLMs across three distinct dimensions: Adversarial attack, +> Compositional reasoning, and Modality-specific dependency. Using our benchmark +> we extensively evaluate 13 state-of-the-art AVLLMs. The findings reveal that +> the majority of existing models fall significantly short of achieving +> human-like comprehension, offering valuable insights for future research +> directions. To alleviate the limitations in the existing approaches, we further +> propose a robust, model-agnostic calibrated audio-visual preference +> optimization based training strategy CAVPref, obtaining a gain up to 30.19% +> across all 9 tasks. We will publicly release our code and benchmark to +> facilitate future research in this direction. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on the development and evaluation of a benchmark for Audio-Visual Large Language Models (AVLLMs) and proposes a training strategy (CAVPref) to enhance their reliability and robustness, rather than focusing on prompt engineering for text-based interactions with LLMs." +} + +--- + +## [Attribute-Based Robotic Grasping with Data-Efficient Adaptation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02149v1) +**arXiv ID:** http://arxiv.org/abs/2501.02149v1 + +**Abstract:** +> Robotic grasping is one of the most fundamental robotic manipulation tasks +> and has been the subject of extensive research. However, swiftly teaching a +> robot to grasp a novel target object in clutter remains challenging. This paper +> attempts to address the challenge by leveraging object attributes that +> facilitate recognition, grasping, and rapid adaptation to new domains. In this +> work, we present an end-to-end encoder-decoder network to learn attribute-based +> robotic grasping with data-efficient adaptation capability. We first pre-train +> the end-to-end model with a variety of basic objects to learn generic attribute +> representation for recognition and grasping. Our approach fuses the embeddings +> of a workspace image and a query text using a gated-attention mechanism and +> learns to predict instance grasping affordances. To train the joint embedding +> space of visual and textual attributes, the robot utilizes object persistence +> before and after grasping. Our model is self-supervised in a simulation that +> only uses basic objects of various colors and shapes but generalizes to novel +> objects in new environments. To further facilitate generalization, we propose +> two adaptation methods, adversarial adaption and one-grasp adaptation. +> Adversarial adaptation regulates the image encoder using augmented data of +> unlabeled images, whereas one-grasp adaptation updates the overall end-to-end +> model using augmented data from one grasp trial. Both adaptation methods are +> data-efficient and considerably improve instance grasping performance. +> Experimental results in both simulation and the real world demonstrate that our +> approach achieves over 81% instance grasping success rate on unknown objects, +> which outperforms several baselines by large margins. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on robotic grasping and adaptation in novel environments, utilizing an encoder-decoder network, with no emphasis on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [The Integration of Blockchain and Artificial Intelligence for Secure + Healthcare Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.02169v1) +**arXiv ID:** http://arxiv.org/abs/2501.02169v1 + +**Abstract:** +> Verisign reported a 125 percent increase in data breaches within the +> healthcare sector in the United States during 2022, with 18.2 million patient +> records being impacted. Growing healthcare data volumes and diversification +> mean that medical information is becoming more valuable. Many Health Centers +> use various technologies to ease the classification, storage, and exchange of +> big data. This use can also make the health data of the users at risk and +> vulnerable. AI and blockchain are among the leading technologies at hand. With +> AI, data-driven operations and big data efficiency have been improved with +> respect to traditional techniques. Due to its potential to bring about +> improvements in health services and lower medical costs, this AI technology is +> regularly used in healthcare. Blockchain helps protect transactions on sharing +> information and private privacy as long as the exchange of knowledge is that of +> the standard. The objective of this analysis is to investigate the research and +> unique contributions since 2008 regarding blockchain-integrated AI and +> healthcare systems. The work sheds light on applied AI-based healthcare schemes +> with machine, ballistic, and acrylic learning and disparate blockchain +> structures. The use of technology in order to ensure patient data security and +> manage medical information effectively in healthcare settings offers a highly +> successful position for both healthcare providers and patients. From 2018 to +> 2021, the best year was 2021 to grow, enhancing everything to examine the +> download of the device and the counting of Google Academies, for which the +> joining perspective was borrowed; local research experts were asked, identified +> articles in recent years, and read reviews of large research grants. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the integration of blockchain and AI for secure healthcare systems, not on the engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not provide examples of prompts impacting LLM output. + +--- + +## [AdaMixup: A Dynamic Defense Framework for Membership Inference Attack + Mitigation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02182v1) +**arXiv ID:** http://arxiv.org/abs/2501.02182v1 + +**Abstract:** +> Membership inference attacks have emerged as a significant privacy concern in +> the training of deep learning models, where attackers can infer whether a data +> point was part of the training set based on the model's outputs. To address +> this challenge, we propose a novel defense mechanism, AdaMixup. AdaMixup +> employs adaptive mixup techniques to enhance the model's robustness against +> membership inference attacks by dynamically adjusting the mixup strategy during +> training. This method not only improves the model's privacy protection but also +> maintains high performance. Experimental results across multiple datasets +> demonstrate that AdaMixup significantly reduces the risk of membership +> inference attacks while achieving a favorable trade-off between defensive +> efficiency and model accuracy. This research provides an effective solution for +> data privacy protection and lays the groundwork for future advancements in +> mixup training methods. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a defense mechanism (AdaMixup) for mitigating membership inference attacks in deep learning models, with no apparent emphasis on prompt engineering, Large Language Models (LLMs), or the manipulation of textual input prompts for LLM performance. + +--- + +## [CPTuning: Contrastive Prompt Tuning for Generative Relation Extraction](https://arxiv.org/abs/http://arxiv.org/abs/2501.02196v1) +**arXiv ID:** http://arxiv.org/abs/2501.02196v1 + +**Abstract:** +> Generative relation extraction (RE) commonly involves first reformulating RE +> as a linguistic modeling problem easily tackled with pre-trained language +> models (PLM) and then fine-tuning a PLM with supervised cross-entropy loss. +> Although having achieved promising performance, existing approaches assume only +> one deterministic relation between each pair of entities without considering +> real scenarios where multiple relations may be valid, i.e., entity pair +> overlap, causing their limited applications. To address this problem, we +> introduce a novel contrastive prompt tuning method for RE, CPTuning, which +> learns to associate a candidate relation between two in-context entities with a +> probability mass above or below a threshold, corresponding to whether the +> relation exists. Beyond learning schema, CPTuning also organizes RE as a +> verbalized relation generation task and uses Trie-constrained decoding to +> ensure a model generates valid relations. It adaptively picks out the generated +> candidate relations with a high estimated likelihood in inference, thereby +> achieving multi-relation extraction. We conduct extensive experiments on four +> widely used datasets to validate our method. Results show that T5-large +> fine-tuned with CPTuning significantly outperforms previous methods, regardless +> of single or multiple relations extraction. + +**Decision Explanation:** +Original decision: REJECT +Although the paper mentions prompt tuning for Large Language Models (LLMs), its primary focus is on improving Generative Relation Extraction through a novel training method, rather than solely on prompt engineering for text-based interactions with LLMs, as evidenced by the emphasis on fine-tuning a PLM and the development of a new decoding strategy. + +--- + +## [Financial Named Entity Recognition: How Far Can LLM Go?](https://arxiv.org/abs/http://arxiv.org/abs/2501.02237v1) +**arXiv ID:** http://arxiv.org/abs/2501.02237v1 + +**Abstract:** +> The surge of large language models (LLMs) has revolutionized the extraction +> and analysis of crucial information from a growing volume of financial +> statements, announcements, and business news. Recognition for named entities to +> construct structured data poses a significant challenge in analyzing financial +> documents and is a foundational task for intelligent financial analytics. +> However, how effective are these generic LLMs and their performance under +> various prompts are yet need a better understanding. To fill in the blank, we +> present a systematic evaluation of state-of-the-art LLMs and prompting methods +> in the financial Named Entity Recognition (NER) problem. Specifically, our +> experimental results highlight their strengths and limitations, identify five +> representative failure types, and provide insights into their potential and +> challenges for domain-specific tasks. + +**Decision Explanation:** +Original decision: REJECT +Although the paper investigates LLM performance under various prompts, its primary focus is on evaluating LLMs for a specific application (Financial Named Entity Recognition) rather than on the engineering, design, or optimization of prompts themselves for improving LLM text generation capabilities. + +--- + +## [Interpretable Load Forecasting via Representation Learning of + Geo-distributed Meteorological Factors](https://arxiv.org/abs/http://arxiv.org/abs/2501.02241v1) +**arXiv ID:** http://arxiv.org/abs/2501.02241v1 + +**Abstract:** +> Meteorological factors (MF) are crucial in day-ahead load forecasting as they +> significantly influence the electricity consumption behaviors of consumers. +> Numerous studies have incorporated MF into the load forecasting model to +> achieve higher accuracy. Selecting MF from one representative location or the +> averaged MF as the inputs of the forecasting model is a common practice. +> However, the difference in MF collected in various locations within a region +> may be significant, which poses a challenge in selecting the appropriate MF +> from numerous locations. A representation learning framework is proposed to +> extract geo-distributed MF while considering their spatial relationships. In +> addition, this paper employs the Shapley value in the graph-based model to +> reveal connections between MF collected in different locations and loads. To +> reduce the computational complexity of calculating the Shapley value, an +> acceleration method is adopted based on Monte Carlo sampling and weighted +> linear regression. Experiments on two real-world datasets demonstrate that the +> proposed method improves the day-ahead forecasting accuracy, especially in +> extreme scenarios such as the "accumulation temperature effect" in summer and +> "sudden temperature change" in winter. We also find a significant correlation +> between the importance of MF in different locations and the corresponding +> area's GDP and mainstay industry. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on interpretable load forecasting using representation learning of geo-distributed meteorological factors, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria." +} + +--- + +## [LLMzSzŁ: a comprehensive LLM benchmark for Polish](https://arxiv.org/abs/http://arxiv.org/abs/2501.02266v1) +**arXiv ID:** http://arxiv.org/abs/2501.02266v1 + +**Abstract:** +> This article introduces the first comprehensive benchmark for the Polish +> language at this scale: LLMzSz{\L} (LLMs Behind the School Desk). It is based +> on a coherent collection of Polish national exams, including both academic and +> professional tests extracted from the archives of the Polish Central +> Examination Board. It covers 4 types of exams, coming from 154 domains. +> Altogether, it consists of almost 19k closed-ended questions. We investigate +> the performance of open-source multilingual, English, and Polish LLMs to verify +> LLMs' abilities to transfer knowledge between languages. Also, the correlation +> between LLMs and humans at model accuracy and exam pass rate levels is +> examined. We show that multilingual LLMs can obtain superior results over +> monolingual ones; however, monolingual models may be beneficial when model size +> matters. Our analysis highlights the potential of LLMs in assisting with exam +> validation, particularly in identifying anomalies or errors in examination +> tasks. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on introducing a benchmark for evaluating LLM performance on Polish language tasks, examining knowledge transfer and correlation with human accuracy, rather than primarily on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [What Kind of Visual Tokens Do We Need? Training-free Visual Token + Pruning for Multi-modal Large Language Models from the Perspective of Graph](https://arxiv.org/abs/http://arxiv.org/abs/2501.02268v1) +**arXiv ID:** http://arxiv.org/abs/2501.02268v1 + +**Abstract:** +> Recent Multimodal Large Language Models(MLLMs) often use a large number of +> visual tokens to compensate their visual shortcoming, leading to excessive +> computation and obvious visual redundancy. In this paper, we investigate what +> kind of visual tokens are needed for MLLMs, and reveal that both foreground and +> background tokens are critical for MLLMs given the varying difficulties of +> examples. Based on this observation, we propose a graph-based method towards +> training-free visual token pruning, termed G-Prune.In particular, G-Prune +> regards visual tokens as nodes, and construct their connections based on their +> semantic similarities. Afterwards, the information flow is propagated via +> weighted links, and the most important tokens after iterations are kept for +> MLLMs, which can be front or background.To validate G-Prune, we apply it to a +> recent MLLM called LLaVA-NeXT, and conduct extensive experiments on a set of +> benchmarks.The experiment results show that G-Prune can greatly reduce +> computation overhead while retaining high performance on both coarse- and +> fine-grained tasks. For instance, G-Prune can reduce 63.57\% FLOPs of +> LLaVA-NeXT on VQA2.0 and TextVQA with only 0.95\% and 2.34\% accuracy drops, +> respectively. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on optimizing visual token pruning for Multi-modal Large Language Models (MLLMs), which is more related to model optimization and reducing computational overhead rather than prompt engineering for text-based interactions with LLMs." +} + +--- + +## [Hyperbolic Contrastive Learning for Hierarchical 3D Point Cloud + Embedding](https://arxiv.org/abs/http://arxiv.org/abs/2501.02285v2) +**arXiv ID:** http://arxiv.org/abs/2501.02285v2 + +**Abstract:** +> Hyperbolic spaces allow for more efficient modeling of complex, hierarchical +> structures, which is particularly beneficial in tasks involving multi-modal +> data. Although hyperbolic geometries have been proven effective for +> language-image pre-training, their capabilities to unify language, image, and +> 3D Point Cloud modalities are under-explored. We extend the 3D Point Cloud +> modality in hyperbolic multi-modal contrastive pre-training. Additionally, we +> explore the entailment, modality gap, and alignment regularizers for learning +> hierarchical 3D embeddings and facilitating the transfer of knowledge from both +> Text and Image modalities. These regularizers enable the learning of +> intra-modal hierarchy within each modality and inter-modal hierarchy across +> text, 2D images, and 3D Point Clouds. Experimental results demonstrate that our +> proposed training strategy yields an outstanding 3D Point Cloud encoder, and +> the obtained 3D Point Cloud hierarchical embeddings significantly improve +> performance on various downstream tasks. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multi-modal contrastive pre-training for 3D Point Cloud embedding, primarily involving image and 3D modalities, with no clear emphasis on prompt engineering or text generation driven by Large Language Models (LLMs). + +--- + +## [SR-Reward: Taking The Path More Traveled](https://arxiv.org/abs/http://arxiv.org/abs/2501.02330v1) +**arXiv ID:** http://arxiv.org/abs/2501.02330v1 + +**Abstract:** +> In this paper, we propose a novel method for learning reward functions +> directly from offline demonstrations. Unlike traditional inverse reinforcement +> learning (IRL), our approach decouples the reward function from the learner's +> policy, eliminating the adversarial interaction typically required between the +> two. This results in a more stable and efficient training process. Our reward +> function, called \textit{SR-Reward}, leverages successor representation (SR) to +> encode a state based on expected future states' visitation under the +> demonstration policy and transition dynamics. By utilizing the Bellman +> equation, SR-Reward can be learned concurrently with most reinforcement +> learning (RL) algorithms without altering the existing training pipeline. We +> also introduce a negative sampling strategy to mitigate overestimation errors +> by reducing rewards for out-of-distribution data, thereby enhancing robustness. +> This strategy inherently introduces a conservative bias into RL algorithms that +> employ the learned reward. We evaluate our method on the D4RL benchmark, +> achieving competitive results compared to offline RL algorithms with access to +> true rewards and imitation learning (IL) techniques like behavioral cloning. +> Moreover, our ablation studies on data size and quality reveal the advantages +> and limitations of SR-Reward as a proxy for true rewards. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on learning reward functions for reinforcement learning (RL) with offline demonstrations, using successor representation and Bellman equation, and does not meet any of the 'MUST' criteria related to prompt engineering for Large Language Models (LLMs), such as improving LLM performance through textual input prompt manipulation or providing concrete examples of LLM prompts. + +--- + +## [AdaSkip: Adaptive Sublayer Skipping for Accelerating Long-Context LLM + Inference](https://arxiv.org/abs/http://arxiv.org/abs/2501.02336v1) +**arXiv ID:** http://arxiv.org/abs/2501.02336v1 + +**Abstract:** +> Long-context large language models (LLMs) inference is increasingly critical, +> motivating a number of studies devoted to alleviating the substantial storage +> and computational costs in such scenarios. Layer-wise skipping methods are +> promising optimizations but rarely explored in long-context inference. We +> observe that existing layer-wise skipping strategies have several limitations +> when applied in long-context inference, including the inability to adapt to +> model and context variability, disregard for sublayer significance, and +> inapplicability for the prefilling phase. This paper proposes \sysname, an +> adaptive sublayer skipping method specifically designed for long-context +> inference. \sysname adaptively identifies less important layers by leveraging +> on-the-fly similarity information, enables sublayer-wise skipping, and +> accelerates both the prefilling and decoding phases. The effectiveness of +> \sysname is demonstrated through extensive experiments on various long-context +> benchmarks and models, showcasing its superior inference performance over +> existing baselines. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on accelerating long-context LLM inference through adaptive sublayer skipping, which pertains to optimizing LLM architecture and inference methods rather than prompt engineering for text-based interactions with LLMs." +} + +--- + +## [Evaluation of the Code Generation Capabilities of ChatGPT 4: A + Comparative Analysis in 19 Programming Languages](https://arxiv.org/abs/http://arxiv.org/abs/2501.02338v1) +**arXiv ID:** http://arxiv.org/abs/2501.02338v1 + +**Abstract:** +> This bachelor's thesis examines the capabilities of ChatGPT 4 in code +> generation across 19 programming languages. The study analyzed solution rates +> across three difficulty levels, types of errors encountered, and code quality +> in terms of runtime and memory efficiency through a quantitative experiment. A +> total of 188 programming problems were selected from the LeetCode platform, and +> ChatGPT 4 was given three attempts to produce a correct solution with feedback. +> ChatGPT 4 successfully solved 39.67% of all tasks, with success rates +> decreasing significantly as problem complexity increased. Notably, the model +> faced considerable challenges with hard problems across all languages. ChatGPT +> 4 demonstrated higher competence in widely used languages, likely due to a +> larger volume and higher quality of training data. The solution rates also +> revealed a preference for languages with low abstraction levels and static +> typing. For popular languages, the most frequent error was "Wrong Answer," +> whereas for less popular languages, compiler and runtime errors prevailed, +> suggesting frequent misunderstandings and confusion regarding the structural +> characteristics of these languages. The model exhibited above-average runtime +> efficiency in all programming languages, showing a tendency toward statically +> typed and low-abstraction languages. Memory efficiency results varied +> significantly, with above-average performance in 14 languages and below-average +> performance in five languages. A slight preference for low-abstraction +> languages and a leaning toward dynamically typed languages in terms of memory +> efficiency were observed. Future research should include a larger number of +> tasks, iterations, and less popular languages. Additionally, ChatGPT 4's +> abilities in code interpretation and summarization, debugging, and the +> development of complex, practical code could be analyzed further. +> ---- +> Diese Bachelorarbeit untersucht die F\"ahigkeiten von ChatGPT 4 zur +> Code-Generierung in 19 Programmiersprachen. Betrachtet wurden die +> L\"osungsraten zwischen drei Schwierigkeitsgraden, die aufgetretenen +> Fehlerarten und die Qualit\"at des Codes hinsichtlich der Laufzeit- und +> Speichereffizienz in einem quantitativen Experiment. Dabei wurden 188 +> Programmierprobleme der Plattform LeetCode entnommen, wobei ChatGPT 4 jeweils +> drei Versuche hatte, mittels Feedback eine korrekte L\"osung zu generieren. +> ChatGPT 4 l\"oste 39,67 % aller Aufgaben erfolgreich, wobei die Erfolgsrate mit +> zunehmendem Schwierigkeitsgrad deutlich abnahm und bei komplexen Problemen in +> allen Sprachen signifikante Schwierigkeiten auftraten. Das Modell zeigte eine +> h\"ohere Kompetenz in weit verbreiteten Sprachen, was wahrscheinlich auf eine +> gr\"o{\ss}ere Menge und h\"ohere Qualit\"at der Trainingsdaten +> zur\"uckzuf\"uhren ist. Bez\"uglich der L\"osungsraten zeigte das Modell zudem +> eine Pr\"aferenz f\"ur Sprachen mit niedrigem Abstraktionsniveau und statischer +> Typisierung. Bei Sprachen hoher Popularit\"at trat der Fehler Wrong Answer am +> h\"aufigsten auf, w\"ahrend bei weniger popul\"aren Sprachen Compiler- und +> Laufzeitfehler \"uberwogen, was auf h\"aufige Missverst\"andnisse und +> Verwechslungen bez\"uglich der spezifischen strukturellen Eigenschaften dieser +> Sprachen zur\"uckzuf\"uhren ist. ChatGPT 4 demonstrierte in allen +> Programmiersprachen eine \"uberdurchschnittliche Laufzeiteffizienz und +> tendierte diesbez\"uglich erneut zu statisch typisierten und niedrig +> abstrahierten Sprachen. Die Werte zur Speichereffizienz variierten erheblich, +> wobei in 14 Sprachen \"uberdurchschnittliche und in f\"unf Sprachen +> unterdurchschnittliche Werte erzielt wurden. Es zeigte sich diesbez\"uglich +> eine leichte Tendenz zugunsten von niedrig abstrahierten sowie eine Pr\"aferenz +> zu dynamisch typisierten Sprachen. Zuk\"unftige Forschung sollte eine h\"ohere +> Anzahl an Aufgaben, Iterationen und unpopul\"aren Sprachen einbeziehen. +> Dar\"uber hinaus k\"onnten die F\"ahigkeiten von ChatGPT 4 in der +> Code-Interpretation und -Zusammenfassung, im Debugging und in der Entwicklung +> komplexer, praxisbezogener Codes analysiert werden. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is on evaluating the code generation capabilities of ChatGPT 4 across various programming languages, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), which is the required core subject." +} + +--- + +## [UAVs Meet LLMs: Overviews and Perspectives Toward Agentic Low-Altitude + Mobility](https://arxiv.org/abs/http://arxiv.org/abs/2501.02341v1) +**arXiv ID:** http://arxiv.org/abs/2501.02341v1 + +**Abstract:** +> Low-altitude mobility, exemplified by unmanned aerial vehicles (UAVs), has +> introduced transformative advancements across various domains, like +> transportation, logistics, and agriculture. Leveraging flexible perspectives +> and rapid maneuverability, UAVs extend traditional systems' perception and +> action capabilities, garnering widespread attention from academia and industry. +> However, current UAV operations primarily depend on human control, with only +> limited autonomy in simple scenarios, and lack the intelligence and +> adaptability needed for more complex environments and tasks. The emergence of +> large language models (LLMs) demonstrates remarkable problem-solving and +> generalization capabilities, offering a promising pathway for advancing UAV +> intelligence. This paper explores the integration of LLMs and UAVs, beginning +> with an overview of UAV systems' fundamental components and functionalities, +> followed by an overview of the state-of-the-art in LLM technology. +> Subsequently, it systematically highlights the multimodal data resources +> available for UAVs, which provide critical support for training and evaluation. +> Furthermore, it categorizes and analyzes key tasks and application scenarios +> where UAVs and LLMs converge. Finally, a reference roadmap towards agentic UAVs +> is proposed, aiming to enable UAVs to achieve agentic intelligence through +> autonomous perception, memory, reasoning, and tool utilization. Related +> resources are available at https://github.com/Hub-Tian/UAVs_Meet_LLMs. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on integrating LLMs with UAVs to enhance UAV intelligence, rather than specifically on the engineering, design, or optimization of prompts for Large Language Models, failing to meet the core subject requirement. + +--- + +## [Exploring the Capabilities and Limitations of Large Language Models for + Radiation Oncology Decision Support](https://arxiv.org/abs/http://arxiv.org/abs/2501.02346v1) +**arXiv ID:** http://arxiv.org/abs/2501.02346v1 + +**Abstract:** +> Thanks to the rapidly evolving integration of LLMs into decision-support +> tools, a significant transformation is happening across large-scale systems. +> Like other medical fields, the use of LLMs such as GPT-4 is gaining increasing +> interest in radiation oncology as well. An attempt to assess GPT-4's +> performance in radiation oncology was made via a dedicated 100-question +> examination on the highly specialized topic of radiation oncology physics, +> revealing GPT-4's superiority over other LLMs. GPT-4's performance on a broader +> field of clinical radiation oncology is further benchmarked by the ACR +> Radiation Oncology In-Training (TXIT) exam where GPT-4 achieved a high accuracy +> of 74.57%. Its performance on re-labelling structure names in accordance with +> the AAPM TG-263 report has also been benchmarked, achieving above 96% +> accuracies. Such studies shed light on the potential of LLMs in radiation +> oncology. As interest in the potential and constraints of LLMs in general +> healthcare applications continues to rise5, the capabilities and limitations of +> LLMs in radiation oncology decision support have not yet been fully explored. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on the application of LLMs in radiation oncology, a medically-oriented subject explicitly excluded by the criteria, rather than the engineering, design, or optimization of prompts for LLMs. + +--- + +## [Context Aware Lemmatization and Morphological Tagging Method in Turkish](https://arxiv.org/abs/http://arxiv.org/abs/2501.02361v1) +**arXiv ID:** http://arxiv.org/abs/2501.02361v1 + +**Abstract:** +> The smallest part of a word that defines the word is called a word root. Word +> roots are used to increase success in many applications since they simplify the +> word. In this study, the lemmatization model, which is a word root finding +> method, and the morphological tagging model, which predicts the grammatical +> knowledge of the word, are presented. The presented model was developed for +> Turkish, and both models make predictions by taking the meaning of the word +> into account. In the literature, there is no lemmatization study that is +> sensitive to word meaning in Turkish. For this reason, the present study shares +> the model and the results obtained from the model on Turkish lemmatization for +> the first time in the literature. In the present study, in the lemmatization +> and morphological tagging models, bidirectional LSTM is used for the spelling +> of words, and the Turkish BERT model is used for the meaning of words. The +> models are trained using the IMST and PUD datasets from Universal Dependencies. +> The results from the training of the models were compared with the results from +> the SIGMORPHON 2019 competition. The results of the comparisons revealed that +> our models were superior. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on lemmatization and morphological tagging for the Turkish language using BERT and LSTM, without any mention of Large Language Models (LLMs), prompt engineering, or manipulation of textual input prompts to improve LLM performance, thus failing to meet the primary 'MUST' criteria. + +--- + +## [Enhancing Workplace Productivity and Well-being Using AI Agent](https://arxiv.org/abs/http://arxiv.org/abs/2501.02368v1) +**arXiv ID:** http://arxiv.org/abs/2501.02368v1 + +**Abstract:** +> This paper discusses the use of Artificial Intelligence (AI) to enhance +> workplace productivity and employee well-being. By integrating machine learning +> (ML) techniques with neurobiological data, the proposed approaches ensure +> alignment with human ethical standards through value alignment models and +> Hierarchical Reinforcement Learning (HRL) for autonomous task management. The +> system utilizes biometric feedback from employees to generate personalized +> health prompts, fostering a supportive work environment that encourages +> physical activity. Additionally, we explore decentralized multi-agent systems +> for improved collaboration and decision-making frameworks that enhance +> transparency. Various approaches using ML techniques in conjunction with AI +> implementations are discussed. Together, these innovations aim to create a more +> productive and health-conscious workplace. These outcomes assist HR management +> and organizations in launching more rational career progression streams for +> employees and facilitating organizational transformation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on using AI for workplace productivity and well-being, integrating various technologies, but does not centrally address prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance, as required. + +--- + +## [Syntactic Evolution in Language Usage](https://arxiv.org/abs/http://arxiv.org/abs/2501.02392v1) +**arXiv ID:** http://arxiv.org/abs/2501.02392v1 + +**Abstract:** +> This research aims to investigate the dynamic nature of linguistic style +> throughout various stages of life, from post teenage to old age. By employing +> linguistic analysis tools and methodologies, the study will delve into the +> intricacies of how individuals adapt and modify their language use over time. +> The research uses a data set of blogs from blogger.com from 2004 and focuses on +> English for syntactic analysis. The findings of this research can have +> implications for linguistics, psychology, and communication studies, shedding +> light on the intricate relationship between age and language. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on the evolution of linguistic style across different age groups, using linguistic analysis tools, and does not investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through prompt manipulation, nor does it provide concrete examples of prompts and their impact on LLM output." +} + +--- + +## [iTARGET: Interpretable Tailored Age Regression for Grouped Epigenetic + Traits](https://arxiv.org/abs/http://arxiv.org/abs/2501.02401v1) +**arXiv ID:** http://arxiv.org/abs/2501.02401v1 + +**Abstract:** +> Accurately predicting chronological age from DNA methylation patterns is +> crucial for advancing biological age estimation. However, this task is made +> challenging by Epigenetic Correlation Drift (ECD) and Heterogeneity Among CpGs +> (HAC), which reflect the dynamic relationship between methylation and age +> across different life stages. To address these issues, we propose a novel +> two-phase algorithm. The first phase employs similarity searching to cluster +> methylation profiles by age group, while the second phase uses Explainable +> Boosting Machines (EBM) for precise, group-specific prediction. Our method not +> only improves prediction accuracy but also reveals key age-related CpG sites, +> detects age-specific changes in aging rates, and identifies pairwise +> interactions between CpG sites. Experimental results show that our approach +> outperforms traditional epigenetic clocks and machine learning models, offering +> a more accurate and interpretable solution for biological age estimation with +> significant implications for aging research. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it does not focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Instead, it proposes a novel algorithm for biological age estimation from DNA methylation patterns, unrelated to LLMs or prompt engineering. + +--- + +## [Boosting Explainability through Selective Rationalization in Pre-trained + Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03182v1) +**arXiv ID:** http://arxiv.org/abs/2501.03182v1 + +**Abstract:** +> The widespread application of pre-trained language models (PLMs) in natural +> language processing (NLP) has led to increasing concerns about their +> explainability. Selective rationalization is a self-explanatory framework that +> selects human-intelligible input subsets as rationales for predictions. Recent +> studies have shown that applying existing rationalization frameworks to PLMs +> will result in severe degeneration and failure problems, producing sub-optimal +> or meaningless rationales. Such failures severely damage trust in +> rationalization methods and constrain the application of rationalization +> techniques on PLMs. In this paper, we find that the homogeneity of tokens in +> the sentences produced by PLMs is the primary contributor to these problems. To +> address these challenges, we propose a method named Pre-trained Language +> Model's Rationalization (PLMR), which splits PLMs into a generator and a +> predictor to deal with NLP tasks while providing interpretable rationales. The +> generator in PLMR also alleviates homogeneity by pruning irrelevant tokens, +> while the predictor uses full-text information to standardize predictions. +> Experiments conducted on two widely used datasets across multiple PLMs +> demonstrate the effectiveness of the proposed method PLMR in addressing the +> challenge of applying selective rationalization to PLMs. Codes: +> https://github.com/ylb777/PLMR. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on improving the explainability of Pre-trained Language Models (PLMs) through selective rationalization, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). The core subject is not prompt engineering for text-based interactions with LLMs, but rather addressing challenges in applying rationalization techniques to PLMs. + +--- + +## [AI-ANNE: (A) (N)eural (N)et for (E)xploration: Transferring Deep + Learning Models onto Microcontrollers and Embedded Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.03256v1) +**arXiv ID:** http://arxiv.org/abs/2501.03256v1 + +**Abstract:** +> This working paper explores the integration of neural networks onto +> resource-constrained embedded systems like a Raspberry Pi Pico / Raspberry Pi +> Pico 2. A TinyML aproach transfers neural networks directly on these +> microcontrollers, enabling real-time, low-latency, and energy-efficient +> inference while maintaining data privacy. Therefore, AI-ANNE: (A) (N)eural +> (N)et for (E)xploration will be presented, which facilitates the transfer of +> pre-trained models from high-performance platforms like TensorFlow and Keras +> onto microcontrollers, using a lightweight programming language like +> MicroPython. This approach demonstrates how neural network architectures, such +> as neurons, layers, density and activation functions can be implemented in +> MicroPython in order to deal with the computational limitations of embedded +> systems. Based on the Raspberry Pi Pico / Raspberry Pi Pico 2, two different +> neural networks on microcontrollers are presented for an example of data +> classification. As an further application example, such a microcontroller can +> be used for condition monitoring, where immediate corrective measures are +> triggered on the basis of sensor data. Overall, this working paper presents a +> very easy-to-implement way of using neural networks on energy-efficient devices +> such as microcontrollers. This makes AI-ANNE: (A) (N)eural (N)et for +> (E)xploration not only suited for practical use, but also as an educational +> tool with clear insights into how neural networks operate. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on transferring deep learning models onto microcontrollers and embedded systems, with no emphasis on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Optimizing Edge AI: A Comprehensive Survey on Data, Model, and System + Strategies](https://arxiv.org/abs/http://arxiv.org/abs/2501.03265v1) +**arXiv ID:** http://arxiv.org/abs/2501.03265v1 + +**Abstract:** +> The emergence of 5G and edge computing hardware has brought about a +> significant shift in artificial intelligence, with edge AI becoming a crucial +> technology for enabling intelligent applications. With the growing amount of +> data generated and stored on edge devices, deploying AI models for local +> processing and inference has become increasingly necessary. However, deploying +> state-of-the-art AI models on resource-constrained edge devices faces +> significant challenges that must be addressed. This paper presents an +> optimization triad for efficient and reliable edge AI deployment, including +> data, model, and system optimization. First, we discuss optimizing data through +> data cleaning, compression, and augmentation to make it more suitable for edge +> deployment. Second, we explore model design and compression methods at the +> model level, such as pruning, quantization, and knowledge distillation. +> Finally, we introduce system optimization techniques like framework support and +> hardware acceleration to accelerate edge AI workflows. Based on an in-depth +> analysis of various application scenarios and deployment challenges of edge AI, +> this paper proposes an optimization paradigm based on the data-model-system +> triad to enable a whole set of solutions to effectively transfer ML models, +> which are initially trained in the cloud, to various edge devices for +> supporting multiple scenarios. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs), instead concentrating on optimizing data, model, and system strategies for edge AI deployment, which falls outside the specified criteria. + +--- + +## [Heterogeneous Graph Pre-training Based Model for Secure and Efficient + Prediction of Default Risk Propagation among Bond Issuers](https://arxiv.org/abs/http://arxiv.org/abs/2501.03268v1) +**arXiv ID:** http://arxiv.org/abs/2501.03268v1 + +**Abstract:** +> Efficient prediction of default risk for bond-issuing enterprises is pivotal +> for maintaining stability and fostering growth in the bond market. Conventional +> methods usually rely solely on an enterprise's internal data for risk +> assessment. In contrast, graph-based techniques leverage interconnected +> corporate information to enhance default risk identification for targeted bond +> issuers. Traditional graph techniques such as label propagation algorithm or +> deepwalk fail to effectively integrate a enterprise's inherent attribute +> information with its topological network data. Additionally, due to data +> scarcity and security privacy concerns between enterprises, end-to-end graph +> neural network (GNN) algorithms may struggle in delivering satisfactory +> performance for target tasks. To address these challenges, we present a novel +> two-stage model. In the first stage, we employ an innovative Masked +> Autoencoders for Heterogeneous Graph (HGMAE) to pre-train on a vast enterprise +> knowledge graph. Subsequently, in the second stage, a specialized classifier +> model is trained to predict default risk propagation probabilities. The +> classifier leverages concatenated feature vectors derived from the pre-trained +> encoder with the enterprise's task-specific feature vectors. Through the +> two-stage training approach, our model not only boosts the importance of unique +> bond characteristics for specific default prediction tasks, but also securely +> and efficiently leverage the global information pre-trained from other +> enterprises. Experimental results demonstrate that our proposed model +> outperforms existing approaches in predicting default risk for bond issuers. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a two-stage model for predicting default risk propagation among bond issuers using heterogeneous graph pre-training, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Knowledge-Guided Biomarker Identification for Label-Free Single-Cell + RNA-Seq Data: A Reinforcement Learning Perspective](https://arxiv.org/abs/http://arxiv.org/abs/2501.04718v1) +**arXiv ID:** http://arxiv.org/abs/2501.04718v1 + +**Abstract:** +> Gene panel selection aims to identify the most informative genomic biomarkers +> in label-free genomic datasets. Traditional approaches, which rely on domain +> expertise, embedded machine learning models, or heuristic-based iterative +> optimization, often introduce biases and inefficiencies, potentially obscuring +> critical biological signals. To address these challenges, we present an +> iterative gene panel selection strategy that harnesses ensemble knowledge from +> existing gene selection algorithms to establish preliminary boundaries or prior +> knowledge, which guide the initial search space. Subsequently, we incorporate +> reinforcement learning through a reward function shaped by expert behavior, +> enabling dynamic refinement and targeted selection of gene panels. This +> integration mitigates biases stemming from initial boundaries while +> capitalizing on RL's stochastic adaptability. Comprehensive comparative +> experiments, case studies, and downstream analyses demonstrate the +> effectiveness of our method, highlighting its improved precision and efficiency +> for label-free biomarker discovery. Our results underscore the potential of +> this approach to advance single-cell genomics data analysis. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on applying reinforcement learning for biomarker identification in genomics, with no apparent emphasis on prompt engineering, Large Language Models (LLMs), or manipulation of textual input prompts for LLM performance improvement." +} + +--- + +## [Calculating Customer Lifetime Value and Churn using Beta Geometric + Negative Binomial and Gamma-Gamma Distribution in a NFT based setting](https://arxiv.org/abs/http://arxiv.org/abs/2501.04719v1) +**arXiv ID:** http://arxiv.org/abs/2501.04719v1 + +**Abstract:** +> Customer Lifetime Value (CLV) is an important metric that measures the total +> value a customer will bring to a business over their lifetime. The Beta +> Geometric Negative Binomial Distribution (BGNBD) and Gamma Gamma Distribution +> are two models that can be used to calculate CLV, taking into account both the +> frequency and value of customer transactions. This article explains the BGNBD +> and Gamma Gamma Distribution models, and how they can be used to calculate CLV +> for NFT (Non-Fungible Token) transaction data in a blockchain setting. By +> estimating the parameters of these models using historical transaction data, +> businesses can gain insights into the lifetime value of their customers and +> make data-driven decisions about marketing and customer retention strategies. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on calculating Customer Lifetime Value using statistical models in an NFT setting, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Multi-task Domain Adaptation for Computation Offloading in + Edge-intelligence Networks](https://arxiv.org/abs/http://arxiv.org/abs/2501.07585v1) +**arXiv ID:** http://arxiv.org/abs/2501.07585v1 + +**Abstract:** +> In the field of multi-access edge computing (MEC), efficient computation +> offloading is crucial for improving resource utilization and reducing latency +> in dynamically changing environments. This paper introduces a new approach, +> termed as Multi-Task Domain Adaptation (MTDA), aiming to enhance the ability of +> computational offloading models to generalize in the presence of domain shifts, +> i.e., when new data in the target environment significantly differs from the +> data in the source domain. The proposed MTDA model incorporates a +> teacher-student architecture that allows continuous adaptation without +> necessitating access to the source domain data during inference, thereby +> maintaining privacy and reducing computational overhead. Utilizing a multi-task +> learning framework that simultaneously manages offloading decisions and +> resource allocation, the proposed MTDA approach outperforms benchmark methods +> regarding mean squared error and accuracy, particularly in environments with +> increasing numbers of users. It is observed by means of computer simulation +> that the proposed MTDA model maintains high performance across various +> scenarios, demonstrating its potential for practical deployment in emerging MEC +> applications. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it focuses primarily on computation offloading in edge-intelligence networks, multi-task domain adaptation, and resource allocation, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thereby failing to address the core subject of prompt engineering for text-based interactions with LLMs. + +--- + +## [Adjoint sharding for very long context training of state space models](https://arxiv.org/abs/http://arxiv.org/abs/2501.00692v1) +**arXiv ID:** http://arxiv.org/abs/2501.00692v1 + +**Abstract:** +> Despite very fast progress, efficiently training large language models (LLMs) +> in very long contexts remains challenging. Existing methods fall back to +> training LLMs with short contexts (a maximum of a few thousands tokens in +> training) and use inference time techniques when evaluating on long contexts +> (above 1M tokens context window at inference). As opposed to +> long-context-inference, training on very long context input prompts is quickly +> limited by GPU memory availability and by the prohibitively long training times +> it requires on state-of-the-art hardware. Meanwhile, many real-life +> applications require not only inference but also training/fine-tuning with long +> context on specific tasks. Such applications include, for example, augmenting +> the context with various sources of raw reference information for fact +> extraction, fact summarization, or fact reconciliation tasks. We propose +> adjoint sharding, a novel technique that comprises sharding gradient +> calculation during training to reduce memory requirements by orders of +> magnitude, making training on very long context computationally tractable. +> Adjoint sharding is based on the adjoint method and computes equivalent +> gradients to backpropagation. We also propose truncated adjoint sharding to +> speed up the algorithm while maintaining performance. We provide a distributed +> version, and a paralleled version of adjoint sharding to further speed up +> training. Empirical results show the proposed adjoint sharding algorithm +> reduces memory usage by up to 3X with a 1.27B parameter large language model on +> 1M context length training. This allows to increase the maximum context length +> during training or fine-tuning of a 1.27B parameter model from 35K tokens to +> above 100K tokens on a training infrastructure composed of five AWS P4 +> instances. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on a novel technique for reducing memory requirements during LLM training (adjoint sharding) rather than engineering, designing, or optimizing prompts specifically for improving LLM performance through textual input manipulation. + +--- + +## [Everywhere Attack: Attacking Locally and Globally to Boost Targeted + Transferability](https://arxiv.org/abs/http://arxiv.org/abs/2501.00707v1) +**arXiv ID:** http://arxiv.org/abs/2501.00707v1 + +**Abstract:** +> Adversarial examples' (AE) transferability refers to the phenomenon that AEs +> crafted with one surrogate model can also fool other models. Notwithstanding +> remarkable progress in untargeted transferability, its targeted counterpart +> remains challenging. This paper proposes an everywhere scheme to boost targeted +> transferability. Our idea is to attack a victim image both globally and +> locally. We aim to optimize 'an army of targets' in every local image region +> instead of the previous works that optimize a high-confidence target in the +> image. Specifically, we split a victim image into non-overlap blocks and +> jointly mount a targeted attack on each block. Such a strategy mitigates +> transfer failures caused by attention inconsistency between surrogate and +> victim models and thus results in stronger transferability. Our approach is +> method-agnostic, which means it can be easily combined with existing +> transferable attacks for even higher transferability. Extensive experiments on +> ImageNet demonstrate that the proposed approach universally improves the +> state-of-the-art targeted attacks by a clear margin, e.g., the transferability +> of the widely adopted Logit attack can be improved by 28.8%-300%.We also +> evaluate the crafted AEs on a real-world platform: Google Cloud Vision. Results +> further support the superiority of the proposed method. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on improving the transferability of adversarial examples in image classification, with no relevance to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [An AI-powered Bayesian generative modeling approach for causal inference + in observational studies](https://arxiv.org/abs/http://arxiv.org/abs/2501.00755v1) +**arXiv ID:** http://arxiv.org/abs/2501.00755v1 + +**Abstract:** +> Causal inference in observational studies with high-dimensional covariates +> presents significant challenges. We introduce CausalBGM, an AI-powered Bayesian +> generative modeling approach that captures the causal relationship among +> covariates, treatment, and outcome variables. The core innovation of CausalBGM +> lies in its ability to estimate the individual treatment effect (ITE) by +> learning individual-specific distributions of a low-dimensional latent feature +> set (e.g., latent confounders) that drives changes in both treatment and +> outcome. This approach not only effectively mitigates confounding effects but +> also provides comprehensive uncertainty quantification, offering reliable and +> interpretable causal effect estimates at the individual level. CausalBGM adopts +> a Bayesian model and uses a novel iterative algorithm to update the model +> parameters and the posterior distribution of latent features until convergence. +> This framework leverages the power of AI to capture complex dependencies among +> variables while adhering to the Bayesian principles. Extensive experiments +> demonstrate that CausalBGM consistently outperforms state-of-the-art methods, +> particularly in scenarios with high-dimensional covariates and large-scale +> datasets. Its Bayesian foundation ensures statistical rigor, providing robust +> and well-calibrated posterior intervals. By addressing key limitations of +> existing methods, CausalBGM emerges as a robust and promising framework for +> advancing causal inference in modern applications in fields such as genomics, +> healthcare, and social sciences. CausalBGM is maintained at the website +> https://causalbgm.readthedocs.io/. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a Bayesian generative modeling approach for causal inference in observational studies, leveraging AI, but does not primarily address the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Revisiting Graph Neural Networks on Graph-level Tasks: Comprehensive + Experiments, Analysis, and Improvements](https://arxiv.org/abs/http://arxiv.org/abs/2501.00773v1) +**arXiv ID:** http://arxiv.org/abs/2501.00773v1 + +**Abstract:** +> Graphs are essential data structures for modeling complex interactions in +> domains such as social networks, molecular structures, and biological systems. +> Graph-level tasks, which predict properties or classes for the entire graph, +> are critical for applications, such as molecular property prediction and +> subgraph counting. Graph Neural Networks (GNNs) have shown promise in these +> tasks, but their evaluations are often limited to narrow datasets, tasks, and +> inconsistent experimental setups, restricting their generalizability. To +> address these limitations, we propose a unified evaluation framework for +> graph-level GNNs. This framework provides a standardized setting to evaluate +> GNNs across diverse datasets, various graph tasks (e.g., graph classification +> and regression), and challenging scenarios, including noisy, imbalanced, and +> few-shot graphs. Additionally, we propose a novel GNN model with enhanced +> expressivity and generalization capabilities. Specifically, we enhance the +> expressivity of GNNs through a $k$-path rooted subgraph approach, enabling the +> model to effectively count subgraphs (e.g., paths and cycles). Moreover, we +> introduce a unified graph contrastive learning algorithm for graphs across +> diverse domains, which adaptively removes unimportant edges to augment graphs, +> thereby significantly improving generalization performance. Extensive +> experiments demonstrate that our model achieves superior performance against +> fourteen effective baselines across twenty-seven graph datasets, establishing +> it as a robust and generalizable model for graph-level tasks. + +**Decision Explanation:** +Original response: json +{ + "decision": "REJECT", + "explanation": "The paper does not meet the core criteria as it focuses primarily on the development and optimization of Graph Neural Networks (GNNs) for graph-level tasks, with no indication of Large Language Models (LLMs), prompt engineering, or text generation being the central focus." +} + +--- + +## [LENS-XAI: Redefining Lightweight and Explainable Network Security + through Knowledge Distillation and Variational Autoencoders for Scalable + Intrusion Detection in Cybersecurity](https://arxiv.org/abs/http://arxiv.org/abs/2501.00790v2) +**arXiv ID:** http://arxiv.org/abs/2501.00790v2 + +**Abstract:** +> The rapid proliferation of Industrial Internet of Things (IIoT) systems +> necessitates advanced, interpretable, and scalable intrusion detection systems +> (IDS) to combat emerging cyber threats. Traditional IDS face challenges such as +> high computational demands, limited explainability, and inflexibility against +> evolving attack patterns. To address these limitations, this study introduces +> the Lightweight Explainable Network Security framework (LENS-XAI), which +> combines robust intrusion detection with enhanced interpretability and +> scalability. LENS-XAI integrates knowledge distillation, variational +> autoencoder models, and attribution-based explainability techniques to achieve +> high detection accuracy and transparency in decision-making. By leveraging a +> training set comprising 10% of the available data, the framework optimizes +> computational efficiency without sacrificing performance. Experimental +> evaluation on four benchmark datasets: Edge-IIoTset, UKM-IDS20, CTU-13, and +> NSL-KDD, demonstrates the framework's superior performance, achieving detection +> accuracies of 95.34%, 99.92%, 98.42%, and 99.34%, respectively. Additionally, +> the framework excels in reducing false positives and adapting to complex attack +> scenarios, outperforming existing state-of-the-art methods. Key strengths of +> LENS-XAI include its lightweight design, suitable for resource-constrained +> environments, and its scalability across diverse IIoT and cybersecurity +> contexts. Moreover, the explainability module enhances trust and transparency, +> critical for practical deployment in dynamic and sensitive applications. This +> research contributes significantly to advancing IDS by addressing computational +> efficiency, feature interpretability, and real-world applicability. Future work +> could focus on extending the framework to ensemble AI systems for distributed +> environments, further enhancing its robustness and adaptability. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a lightweight and explainable network security framework (LENS-XAI) for intrusion detection in cybersecurity, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions, thus failing to meet all 'MUST' criteria. + +--- + +## [Decoupling Knowledge and Reasoning in Transformers: A Modular + Architecture with Generalized Cross-Attention](https://arxiv.org/abs/http://arxiv.org/abs/2501.00823v2) +**arXiv ID:** http://arxiv.org/abs/2501.00823v2 + +**Abstract:** +> Transformers have achieved remarkable success across diverse domains, but +> their monolithic architecture presents challenges in interpretability, +> adaptability, and scalability. This paper introduces a novel modular +> Transformer architecture that explicitly decouples knowledge and reasoning +> through a generalized cross-attention mechanism to a globally shared knowledge +> base with layer-specific transformations, specifically designed for effective +> knowledge retrieval. Critically, we provide a rigorous mathematical derivation +> demonstrating that the Feed-Forward Network (FFN) in a standard Transformer is +> a specialized case (a closure) of this generalized cross-attention, revealing +> its role in implicit knowledge retrieval and validating our design. This +> theoretical framework provides a new lens for understanding FFNs and lays the +> foundation for future research exploring enhanced interpretability, +> adaptability, and scalability, enabling richer interplay with external +> knowledge bases and other systems. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a novel modular Transformer architecture, decoupling knowledge and reasoning, which falls under developing new LLM architectures, violating the 'MUST NOT' criteria (1). Prompt engineering for text-based interactions with LLMs is not the central focus. + +--- + +## [What is a Social Media Bot? A Global Comparison of Bot and Human + Characteristics](https://arxiv.org/abs/http://arxiv.org/abs/2501.00855v1) +**arXiv ID:** http://arxiv.org/abs/2501.00855v1 + +**Abstract:** +> Chatter on social media is 20% bots and 80% humans. Chatter by bots and +> humans is consistently different: bots tend to use linguistic cues that can be +> easily automated while humans use cues that require dialogue understanding. +> Bots use words that match the identities they choose to present, while humans +> may send messages that are not related to the identities they present. Bots and +> humans differ in their communication structure: sampled bots have a star +> interaction structure, while sampled humans have a hierarchical structure. +> These conclusions are based on a large-scale analysis of social media tweets +> across ~200mil users across 7 events. Social media bots took the world by storm +> when social-cybersecurity researchers realized that social media users not only +> consisted of humans but also of artificial agents called bots. These bots wreck +> havoc online by spreading disinformation and manipulating narratives. Most +> research on bots are based on special-purposed definitions, mostly predicated +> on the event studied. This article first begins by asking, "What is a bot?", +> and we study the underlying principles of how bots are different from humans. +> We develop a first-principle definition of a social media bot. With this +> definition as a premise, we systematically compare characteristics between bots +> and humans across global events, and reflect on how the software-programmed bot +> is an Artificial Intelligent algorithm, and its potential for evolution as +> technology advances. Based on our results, we provide recommendations for the +> use and regulation of bots. Finally, we discuss open challenges and future +> directions: Detect, to systematically identify these automated and potentially +> evolving bots; Differentiate, to evaluate the goodness of the bot in terms of +> their content postings and relationship interactions; Disrupt, to moderate the +> impact of malicious bots. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it primarily focuses on analyzing social media bot characteristics, their differentiation from humans, and regulatory recommendations, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [DiffETM: Diffusion Process Enhanced Embedded Topic Model](https://arxiv.org/abs/http://arxiv.org/abs/2501.00862v1) +**arXiv ID:** http://arxiv.org/abs/2501.00862v1 + +**Abstract:** +> The embedded topic model (ETM) is a widely used approach that assumes the +> sampled document-topic distribution conforms to the logistic normal +> distribution for easier optimization. However, this assumption oversimplifies +> the real document-topic distribution, limiting the model's performance. In +> response, we propose a novel method that introduces the diffusion process into +> the sampling process of document-topic distribution to overcome this limitation +> and maintain an easy optimization process. We validate our method through +> extensive experiments on two mainstream datasets, proving its effectiveness in +> improving topic modeling performance. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs), instead proposing a method for enhancing an Embedded Topic Model using a diffusion process, which does not meet the primary criteria of investigating, analyzing, or proposing methods for improving LLM performance through textual input prompt manipulation. + +--- + +## [Representation in large language models](https://arxiv.org/abs/http://arxiv.org/abs/2501.00885v1) +**arXiv ID:** http://arxiv.org/abs/2501.00885v1 + +**Abstract:** +> The extraordinary success of recent Large Language Models (LLMs) on a diverse +> array of tasks has led to an explosion of scientific and philosophical +> theorizing aimed at explaining how they do what they do. Unfortunately, +> disagreement over fundamental theoretical issues has led to stalemate, with +> entrenched camps of LLM optimists and pessimists often committed to very +> different views of how these systems work. Overcoming stalemate requires +> agreement on fundamental questions, and the goal of this paper is to address +> one such question, namely: is LLM behavior driven partly by +> representation-based information processing of the sort implicated in +> biological cognition, or is it driven entirely by processes of memorization and +> stochastic table look-up? This is a question about what kind of algorithm LLMs +> implement, and the answer carries serious implications for higher level +> questions about whether these systems have beliefs, intentions, concepts, +> knowledge, and understanding. I argue that LLM behavior is partially driven by +> representation-based information processing, and then I describe and defend a +> series of practical techniques for investigating these representations and +> developing explanations on their basis. The resulting account provides a +> groundwork for future theorizing about language models and their successors. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on the theoretical foundations and internal workings of Large Language Models (LLMs), specifically whether LLM behavior is driven by representation-based information processing. While it touches on understanding LLMs, it does not primarily investigate, analyze, or propose methods for improving LLM performance through the manipulation of textual input prompts, nor does it provide concrete examples of prompts and their impact on LLM output." +} + +--- + +## [Demystifying Online Clustering of Bandits: Enhanced Exploration Under + Stochastic and Smoothed Adversarial Contexts](https://arxiv.org/abs/http://arxiv.org/abs/2501.00891v1) +**arXiv ID:** http://arxiv.org/abs/2501.00891v1 + +**Abstract:** +> The contextual multi-armed bandit (MAB) problem is crucial in sequential +> decision-making. A line of research, known as online clustering of bandits, +> extends contextual MAB by grouping similar users into clusters, utilizing +> shared features to improve learning efficiency. However, existing algorithms, +> which rely on the upper confidence bound (UCB) strategy, struggle to gather +> adequate statistical information to accurately identify unknown user clusters. +> As a result, their theoretical analyses require several strong assumptions +> about the "diversity" of contexts generated by the environment, leading to +> impractical settings, complicated analyses, and poor practical performance. +> Removing these assumptions has been a long-standing open problem in the +> clustering of bandits literature. In this paper, we provide two solutions to +> this open problem. First, following the i.i.d. context generation setting in +> existing studies, we propose two novel algorithms, UniCLUB and PhaseUniCLUB, +> which incorporate enhanced exploration mechanisms to accelerate cluster +> identification. Remarkably, our algorithms require substantially weaker +> assumptions while achieving regret bounds comparable to prior work. Second, +> inspired by the smoothed analysis framework, we propose a more practical +> setting that eliminates the requirement for i.i.d. context generation used in +> previous studies, thus enhancing the performance of existing algorithms for +> online clustering of bandits. Our technique can be applied to both graph-based +> and set-based clustering of bandits frameworks. Extensive evaluations on both +> synthetic and real-world datasets demonstrate that our proposed algorithms +> consistently outperform existing approaches. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it focuses on online clustering of bandits, multi-armed bandit problems, and sequential decision-making, with no primary focus on engineering, design, or optimization of prompts for Large Language Models (LLMs) or manipulation of textual input prompts to improve LLM performance. + +--- + +## [Large Language Model Based Multi-Agent System Augmented Complex Event + Processing Pipeline for Internet of Multimedia Things](https://arxiv.org/abs/http://arxiv.org/abs/2501.00906v2) +**arXiv ID:** http://arxiv.org/abs/2501.00906v2 + +**Abstract:** +> This paper presents the development and evaluation of a Large Language Model +> (LLM), also known as foundation models, based multi-agent system framework for +> complex event processing (CEP) with a focus on video query processing use +> cases. The primary goal is to create a proof-of-concept (POC) that integrates +> state-of-the-art LLM orchestration frameworks with publish/subscribe (pub/sub) +> tools to address the integration of LLMs with current CEP systems. Utilizing +> the Autogen framework in conjunction with Kafka message brokers, the system +> demonstrates an autonomous CEP pipeline capable of handling complex workflows. +> Extensive experiments evaluate the system's performance across varying +> configurations, complexities, and video resolutions, revealing the trade-offs +> between functionality and latency. The results show that while higher agent +> count and video complexities increase latency, the system maintains high +> consistency in narrative coherence. This research builds upon and contributes +> to, existing novel approaches to distributed AI systems, offering detailed +> insights into integrating such systems into existing infrastructures. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on integrating Large Language Models into a multi-agent system for complex event processing, not on prompt engineering for text-based interactions with LLMs, failing to meet the core subject requirement. + +--- + +## [Enhancing Early Diabetic Retinopathy Detection through Synthetic DR1 + Image Generation: A StyleGAN3 Approach](https://arxiv.org/abs/http://arxiv.org/abs/2501.00954v1) +**arXiv ID:** http://arxiv.org/abs/2501.00954v1 + +**Abstract:** +> Diabetic Retinopathy (DR) is a leading cause of preventable blindness. Early +> detection at the DR1 stage is critical but is hindered by a scarcity of +> high-quality fundus images. This study uses StyleGAN3 to generate synthetic DR1 +> images characterized by microaneurysms with high fidelity and diversity. The +> aim is to address data scarcity and enhance the performance of supervised +> classifiers. A dataset of 2,602 DR1 images was used to train the model, +> followed by a comprehensive evaluation using quantitative metrics, including +> Frechet Inception Distance (FID), Kernel Inception Distance (KID), and +> Equivariance with respect to translation (EQ-T) and rotation (EQ-R). +> Qualitative assessments included Human Turing tests, where trained +> ophthalmologists evaluated the realism of synthetic images. Spectral analysis +> further validated image quality. The model achieved a final FID score of 17.29, +> outperforming the mean FID of 21.18 (95 percent confidence interval - 20.83 to +> 21.56) derived from bootstrap resampling. Human Turing tests demonstrated the +> model's ability to produce highly realistic images, though minor artifacts near +> the borders were noted. These findings suggest that StyleGAN3-generated +> synthetic DR1 images hold significant promise for augmenting training datasets, +> enabling more accurate early detection of Diabetic Retinopathy. This +> methodology highlights the potential of synthetic data in advancing medical +> imaging and AI-driven diagnostics. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on medical imaging (Diabetic Retinopathy detection) using StyleGAN3 for synthetic image generation, and does not investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through the manipulation of textual input prompts, thus failing to meet the core 'MUST' criteria. + +--- + +## [The Silent Majority: Demystifying Memorization Effect in the Presence of + Spurious Correlations](https://arxiv.org/abs/http://arxiv.org/abs/2501.00961v2) +**arXiv ID:** http://arxiv.org/abs/2501.00961v2 + +**Abstract:** +> Machine learning models often rely on simple spurious features -- patterns in +> training data that correlate with targets but are not causally related to them, +> like image backgrounds in foreground classification. This reliance typically +> leads to imbalanced test performance across minority and majority groups. In +> this work, we take a closer look at the fundamental cause of such imbalanced +> performance through the lens of memorization, which refers to the ability to +> predict accurately on \textit{atypical} examples (minority groups) in the +> training set but failing in achieving the same accuracy in the testing set. +> This paper systematically shows the ubiquitous existence of spurious features +> in a small set of neurons within the network, providing the first-ever evidence +> that memorization may contribute to imbalanced group performance. Through three +> experimental sources of converging empirical evidence, we find the property of +> a small subset of neurons or channels in memorizing minority group information. +> Inspired by these findings, we articulate the hypothesis: the imbalanced group +> performance is a byproduct of ``noisy'' spurious memorization confined to a +> small set of neurons. To further substantiate this hypothesis, we show that +> eliminating these unnecessary spurious memorization patterns via a novel +> framework during training can significantly affect the model performance on +> minority groups. Our experimental results across various architectures and +> benchmarks offer new insights on how neural networks encode core and spurious +> knowledge, laying the groundwork for future research in demystifying robustness +> to spurious correlation. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria, as it primarily focuses on understanding memorization effects and spurious correlations in machine learning models, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not provide concrete examples of prompts and their impact on LLM output. + +--- + +## [FlashInfer: Efficient and Customizable Attention Engine for LLM + Inference Serving](https://arxiv.org/abs/http://arxiv.org/abs/2501.01005v1) +**arXiv ID:** http://arxiv.org/abs/2501.01005v1 + +**Abstract:** +> Transformers, driven by attention mechanisms, form the foundation of large +> language models (LLMs). As these models scale up, efficient GPU attention +> kernels become essential for high-throughput and low-latency inference. Diverse +> LLM applications demand flexible and high-performance attention solutions. We +> present FlashInfer: a customizable and efficient attention engine for LLM +> serving. FlashInfer tackles KV-cache storage heterogeneity using block-sparse +> format and composable formats to optimize memory access and reduce redundancy. +> It also offers a customizable attention template, enabling adaptation to +> various settings through Just-In-Time (JIT) compilation. Additionally, +> FlashInfer's load-balanced scheduling algorithm adjusts to dynamism of user +> requests while maintaining compatibility with CUDAGraph which requires static +> configuration. FlashInfer have been integrated into leading LLM serving +> frameworks like SGLang, vLLM and MLC-Engine. Comprehensive kernel-level and +> end-to-end evaluations demonstrate FlashInfer's ability to significantly boost +> kernel performance across diverse inference scenarios: compared to +> state-of-the-art LLM serving solutions, FlashInfer achieve 29-69% +> inter-token-latency reduction compared to compiler backends for LLM serving +> benchmark, 28-30% latency reduction for long-context inference, and 13-17% +> speedup for LLM serving with parallel generation. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing an efficient attention engine for Large Language Model (LLM) inference serving, which falls under the development of new LLM architectures or training methods, and does not centrally address prompt engineering for text-based interactions with LLMs. + +--- + +## [CryptoMamba: Leveraging State Space Models for Accurate Bitcoin Price + Prediction](https://arxiv.org/abs/http://arxiv.org/abs/2501.01010v1) +**arXiv ID:** http://arxiv.org/abs/2501.01010v1 + +**Abstract:** +> Predicting Bitcoin price remains a challenging problem due to the high +> volatility and complex non-linear dynamics of cryptocurrency markets. +> Traditional time-series models, such as ARIMA and GARCH, and recurrent neural +> networks, like LSTMs, have been widely applied to this task but struggle to +> capture the regime shifts and long-range dependencies inherent in the data. In +> this work, we propose CryptoMamba, a novel Mamba-based State Space Model (SSM) +> architecture designed to effectively capture long-range dependencies in +> financial time-series data. Our experiments show that CryptoMamba not only +> provides more accurate predictions but also offers enhanced generalizability +> across different market conditions, surpassing the limitations of previous +> models. Coupled with trading algorithms for real-world scenarios, CryptoMamba +> demonstrates its practical utility by translating accurate forecasts into +> financial outcomes. Our findings signal a huge advantage for SSMs in stock and +> cryptocurrency price forecasting tasks. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a State Space Model (SSM) for Bitcoin price prediction, with no primary emphasis on the engineering, design, or optimization of prompts for Large Language Models (LLMs), failing to meet the core subject requirement. + +--- + +## [ValuesRAG: Enhancing Cultural Alignment Through Retrieval-Augmented + Contextual Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01031v2) +**arXiv ID:** http://arxiv.org/abs/2501.01031v2 + +**Abstract:** +> Cultural values alignment in Large Language Models (LLMs) is a critical +> challenge due to their tendency to embed Western-centric biases from training +> data, leading to misrepresentations and fairness issues in cross-cultural +> contexts. Recent approaches, such as role-assignment and few-shot learning, +> often struggle with reliable cultural alignment as they heavily rely on +> pre-trained knowledge, lack scalability, and fail to capture nuanced cultural +> values effectively. To address these issues, we propose ValuesRAG, a novel and +> effective framework that applies Retrieval-Augmented Generation (RAG) with +> In-Context Learning (ICL) to integrate cultural and demographic knowledge +> dynamically during text generation. Leveraging the World Values Survey (WVS) +> dataset, ValuesRAG first generates summaries of values for each individual. +> Subsequently, we curate several representative regional datasets to serve as +> test datasets and retrieve relevant summaries of values based on demographic +> features, followed by a reranking step to select the top-k relevant summaries. +> ValuesRAG consistently outperforms baseline methods, both in the main +> experiment and in the ablation study where only the values summary was +> provided. Notably, ValuesRAG demonstrates an accuracy of 21% improvement over +> other baseline methods, highlighting its potential to foster culturally aligned +> AI systems and enhance the inclusivity of AI-driven applications. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing cultural alignment in LLMs through Retrieval-Augmented Generation and In-Context Learning, rather than specifically investigating, analyzing, or proposing methods for improving LLM performance through the manipulation of textual input prompts. + +--- + +## [MSC-Bench: Benchmarking and Analyzing Multi-Sensor Corruption for + Driving Perception](https://arxiv.org/abs/http://arxiv.org/abs/2501.01037v1) +**arXiv ID:** http://arxiv.org/abs/2501.01037v1 + +**Abstract:** +> Multi-sensor fusion models play a crucial role in autonomous driving +> perception, particularly in tasks like 3D object detection and HD map +> construction. These models provide essential and comprehensive static +> environmental information for autonomous driving systems. While camera-LiDAR +> fusion methods have shown promising results by integrating data from both +> modalities, they often depend on complete sensor inputs. This reliance can lead +> to low robustness and potential failures when sensors are corrupted or missing, +> raising significant safety concerns. To tackle this challenge, we introduce the +> Multi-Sensor Corruption Benchmark (MSC-Bench), the first comprehensive +> benchmark aimed at evaluating the robustness of multi-sensor autonomous driving +> perception models against various sensor corruptions. Our benchmark includes 16 +> combinations of corruption types that disrupt both camera and LiDAR inputs, +> either individually or concurrently. Extensive evaluations of six 3D object +> detection models and four HD map construction models reveal substantial +> performance degradation under adverse weather conditions and sensor failures, +> underscoring critical safety issues. The benchmark toolkit and affiliated code +> and model checkpoints have been made publicly accessible. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on autonomous driving perception, multi-sensor fusion, and benchmarking for robustness against sensor corruptions, with no mention or emphasis on prompt engineering for Large Language Models (LLMs) or text generation, thus failing to meet the 'MUST' criteria. + +--- + +## [MMVA: Multimodal Matching Based on Valence and Arousal across Images, + Music, and Musical Captions](https://arxiv.org/abs/http://arxiv.org/abs/2501.01094v1) +**arXiv ID:** http://arxiv.org/abs/2501.01094v1 + +**Abstract:** +> We introduce Multimodal Matching based on Valence and Arousal (MMVA), a +> tri-modal encoder framework designed to capture emotional content across +> images, music, and musical captions. To support this framework, we expand the +> Image-Music-Emotion-Matching-Net (IMEMNet) dataset, creating IMEMNet-C which +> includes 24,756 images and 25,944 music clips with corresponding musical +> captions. We employ multimodal matching scores based on the continuous valence +> (emotional positivity) and arousal (emotional intensity) values. This +> continuous matching score allows for random sampling of image-music pairs +> during training by computing similarity scores from the valence-arousal values +> across different modalities. Consequently, the proposed approach achieves +> state-of-the-art performance in valence-arousal prediction tasks. Furthermore, +> the framework demonstrates its efficacy in various zeroshot tasks, highlighting +> the potential of valence and arousal predictions in downstream applications. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multimodal matching across images, music, and musical captions, primarily for valence and arousal prediction, with no clear emphasis on the engineering, design, or optimization of textual input prompts specifically for Large Language Models (LLMs). + +--- + +## [Disambiguation of Chinese Polyphones in an End-to-End Framework with + Semantic Features Extracted by Pre-trained BERT](https://arxiv.org/abs/http://arxiv.org/abs/2501.01102v1) +**arXiv ID:** http://arxiv.org/abs/2501.01102v1 + +**Abstract:** +> Grapheme-to-phoneme (G2P) conversion serves as an essential component in +> Chinese Mandarin text-to-speech (TTS) system, where polyphone disambiguation is +> the core issue. In this paper, we propose an end-to-end framework to predict +> the pronunciation of a polyphonic character, which accepts sentence containing +> polyphonic character as input in the form of Chinese character sequence without +> the necessity of any preprocessing. The proposed method consists of a +> pre-trained bidirectional encoder representations from Transformers (BERT) +> model and a neural network (NN) based classifier. The pre-trained BERT model +> extracts semantic features from a raw Chinese character sequence and the NN +> based classifier predicts the polyphonic character's pronunciation according to +> BERT output. In out experiments, we implemented three classifiers, a +> fully-connected network based classifier, a long short-term memory (LSTM) +> network based classifier and a Transformer block based classifier. The +> experimental results compared with the baseline approach based on LSTM +> demonstrate that, the pre-trained model extracts effective semantic features, +> which greatly enhances the performance of polyphone disambiguation. In +> addition, we also explored the impact of contextual information on polyphone +> disambiguation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on polyphone disambiguation in Chinese text-to-speech (TTS) systems using pre-trained BERT for semantic feature extraction, which does not meet the primary criteria of focusing on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or investigating methods for improving LLM performance through prompt manipulation. + +--- + +## [learning discriminative features from spectrograms using center loss for + speech emotion recognition](https://arxiv.org/abs/http://arxiv.org/abs/2501.01103v1) +**arXiv ID:** http://arxiv.org/abs/2501.01103v1 + +**Abstract:** +> Identifying the emotional state from speech is essential for the natural +> interaction of the machine with the speaker. However, extracting effective +> features for emotion recognition is difficult, as emotions are ambiguous. We +> propose a novel approach to learn discriminative features from variable length +> spectrograms for emotion recognition by cooperating softmax cross-entropy loss +> and center loss together. The softmax cross-entropy loss enables features from +> different emotion categories separable, and center loss efficiently pulls the +> features belonging to the same emotion category to their center. By combining +> the two losses together, the discriminative power will be highly enhanced, +> which leads to network learning more effective features for emotion +> recognition. As demonstrated by the experimental results, after introducing +> center loss, both the unweighted accuracy and weighted accuracy are improved by +> over 3\% on Mel-spectrogram input, and more than 4\% on Short Time Fourier +> Transform spectrogram input. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it primarily focuses on speech emotion recognition using spectrograms and novel loss functions for feature learning, with no mention of Large Language Models (LLMs), prompt engineering, or text-based interactions, thereby failing to meet the core subject requirement. + +--- + +## [Robust COVID-19 Detection from Cough Sounds using Deep Neural Decision + Tree and Forest: A Comprehensive Cross-Datasets Evaluation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01117v1) +**arXiv ID:** http://arxiv.org/abs/2501.01117v1 + +**Abstract:** +> This research presents a robust approach to classifying COVID-19 cough sounds +> using cutting-edge machine-learning techniques. Leveraging deep neural decision +> trees and deep neural decision forests, our methodology demonstrates consistent +> performance across diverse cough sound datasets. We begin with a comprehensive +> extraction of features to capture a wide range of audio features from +> individuals, whether COVID-19 positive or negative. To determine the most +> important features, we use recursive feature elimination along with +> cross-validation. Bayesian optimization fine-tunes hyper-parameters of deep +> neural decision tree and deep neural decision forest models. Additionally, we +> integrate the SMOTE during training to ensure a balanced representation of +> positive and negative data. Model performance refinement is achieved through +> threshold optimization, maximizing the ROC-AUC score. Our approach undergoes a +> comprehensive evaluation in five datasets: Cambridge, Coswara, COUGHVID, +> Virufy, and the combined Virufy with the NoCoCoDa dataset. Consistently +> outperforming state-of-the-art methods, our proposed approach yields notable +> AUC scores of 0.97, 0.98, 0.92, 0.93, 0.99, and 0.99 across the respective +> datasets. Merging all datasets into a combined dataset, our method, using a +> deep neural decision forest classifier, achieves an AUC of 0.97. Also, our +> study includes a comprehensive cross-datasets analysis, revealing demographic +> and geographic differences in the cough sounds associated with COVID-19. These +> differences highlight the challenges in transferring learned features across +> diverse datasets and underscore the potential benefits of dataset integration, +> improving generalizability and enhancing COVID-19 detection from audio signals. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on COVID-19 detection from cough sounds using deep neural decision trees and forests, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria. + +--- + +## [TED: Turn Emphasis with Dialogue Feature Attention for Emotion + Recognition in Conversation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01123v1) +**arXiv ID:** http://arxiv.org/abs/2501.01123v1 + +**Abstract:** +> Emotion recognition in conversation (ERC) has been attracting attention by +> methods for modeling multi-turn contexts. The multi-turn input to a pretraining +> model implicitly assumes that the current turn and other turns are +> distinguished during the training process by inserting special tokens into the +> input sequence. This paper proposes a priority-based attention method to +> distinguish each turn explicitly by adding dialogue features into the attention +> mechanism, called Turn Emphasis with Dialogue (TED). It has a priority for each +> turn according to turn position and speaker information as dialogue features. +> It takes multi-head self-attention between turn-based vectors for multi-turn +> input and adjusts attention scores with the dialogue features. We evaluate TED +> on four typical benchmarks. The experimental results demonstrate that TED has +> high overall performance in all datasets and achieves state-of-the-art +> performance on IEMOCAP with numerous turns. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on improving emotion recognition in conversation using a priority-based attention method (TED), not on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not demonstrate the impact of textual input prompts on LLM output." +} + +--- + +## [Missing Data as Augmentation in the Earth Observation Domain: A + Multi-View Learning Approach](https://arxiv.org/abs/http://arxiv.org/abs/2501.01132v1) +**arXiv ID:** http://arxiv.org/abs/2501.01132v1 + +**Abstract:** +> Multi-view learning (MVL) leverages multiple sources or views of data to +> enhance machine learning model performance and robustness. This approach has +> been successfully used in the Earth Observation (EO) domain, where views have a +> heterogeneous nature and can be affected by missing data. Despite the negative +> effect that missing data has on model predictions, the ML literature has used +> it as an augmentation technique to improve model generalization, like masking +> the input data. Inspired by this, we introduce novel methods for EO +> applications tailored to MVL with missing views. Our methods integrate the +> combination of a set to simulate all combinations of missing views as different +> training samples. Instead of replacing missing data with a numerical value, we +> use dynamic merge functions, like average, and more complex ones like +> Transformer. This allows the MVL model to entirely ignore the missing views, +> enhancing its predictive robustness. We experiment on four EO datasets with +> temporal and static views, including state-of-the-art methods from the EO +> domain. The results indicate that our methods improve model robustness under +> conditions of moderate missingness, and improve the predictive performance when +> all views are present. The proposed methods offer a single adaptive solution to +> operate effectively with any combination of available views. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multi-view learning for Earth Observation (EO) with missing data augmentation, lacking any direct connection to prompt engineering, Large Language Models (LLMs), or text generation, thus failing to meet the primary 'MUST' criteria. + +--- + +## [TexAVi: Generating Stereoscopic VR Video Clips from Text Descriptions](https://arxiv.org/abs/http://arxiv.org/abs/2501.01156v1) +**arXiv ID:** http://arxiv.org/abs/2501.01156v1 + +**Abstract:** +> While generative models such as text-to-image, large language models and +> text-to-video have seen significant progress, the extension to +> text-to-virtual-reality remains largely unexplored, due to a deficit in +> training data and the complexity of achieving realistic depth and motion in +> virtual environments. This paper proposes an approach to coalesce existing +> generative systems to form a stereoscopic virtual reality video from text. +> Carried out in three main stages, we start with a base text-to-image model +> that captures context from an input text. We then employ Stable Diffusion on +> the rudimentary image produced, to generate frames with enhanced realism and +> overall quality. These frames are processed with depth estimation algorithms to +> create left-eye and right-eye views, which are stitched side-by-side to create +> an immersive viewing experience. Such systems would be highly beneficial in +> virtual reality production, since filming and scene building often require +> extensive hours of work and post-production effort. +> We utilize image evaluation techniques, specifically Fr\'echet Inception +> Distance and CLIP Score, to assess the visual quality of frames produced for +> the video. These quantitative measures establish the proficiency of the +> proposed method. +> Our work highlights the exciting possibilities of using natural +> language-driven graphics in fields like virtual reality simulations. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on generating stereoscopic VR video clips from text descriptions, which falls under image/video generation rather than text generation driven by Large Language Models (LLMs), and does not meet the criteria of investigating prompt engineering for text-based interactions with LLMs. + +--- + +## [Harnessing Multi-Agent LLMs for Complex Engineering Problem-Solving: A + Framework for Senior Design Projects](https://arxiv.org/abs/http://arxiv.org/abs/2501.01205v1) +**arXiv ID:** http://arxiv.org/abs/2501.01205v1 + +**Abstract:** +> Multi-Agent Large Language Models (LLMs) are gaining significant attention +> for their ability to harness collective intelligence in complex +> problem-solving, decision-making, and planning tasks. This aligns with the +> concept of the wisdom of crowds, where diverse agents contribute collectively +> to generating effective solutions, making it particularly suitable for +> educational settings. Senior design projects, also known as capstone or final +> year projects, are pivotal in engineering education as they integrate +> theoretical knowledge with practical application, fostering critical thinking, +> teamwork, and real-world problem-solving skills. In this paper, we explore the +> use of Multi-Agent LLMs in supporting these senior design projects undertaken +> by engineering students, which often involve multidisciplinary considerations +> and conflicting objectives, such as optimizing technical performance while +> addressing ethical, social, and environmental concerns. We propose a framework +> where distinct LLM agents represent different expert perspectives, such as +> problem formulation agents, system complexity agents, societal and ethical +> agents, or project managers, thus facilitating a holistic problem-solving +> approach. This implementation leverages standard multi-agent system (MAS) +> concepts such as coordination, cooperation, and negotiation, incorporating +> prompt engineering to develop diverse personas for each agent. These agents +> engage in rich, collaborative dialogues to simulate human engineering teams, +> guided by principles from swarm AI to efficiently balance individual +> contributions towards a unified solution. We adapt these techniques to create a +> collaboration structure for LLM agents, encouraging interdisciplinary reasoning +> and negotiation similar to real-world senior design projects. To assess the +> efficacy of this framework, we collected six proposals of engineering and +> computer science of... + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on utilizing Multi-Agent LLMs for complex engineering problem-solving in educational settings, rather than specifically on the engineering, design, or optimization of prompts for Large Language Models, as evidenced by its emphasis on a framework for senior design projects and multi-agent system concepts. + +--- + +## [Face-Human-Bench: A Comprehensive Benchmark of Face and Human + Understanding for Multi-modal Assistants](https://arxiv.org/abs/http://arxiv.org/abs/2501.01243v2) +**arXiv ID:** http://arxiv.org/abs/2501.01243v2 + +**Abstract:** +> Faces and humans are crucial elements in social interaction and are widely +> included in everyday photos and videos. Therefore, a deep understanding of +> faces and humans will enable multi-modal assistants to achieve improved +> response quality and broadened application scope. Currently, the multi-modal +> assistant community lacks a comprehensive and scientific evaluation of face and +> human understanding abilities. In this paper, we first propose a hierarchical +> ability taxonomy that includes three levels of abilities. Then, based on this +> taxonomy, we collect images and annotations from publicly available datasets in +> the face and human community and build a semi-automatic data pipeline to +> produce problems for the new benchmark. Finally, the obtained Face-Human-Bench +> comprises a development set with 900 problems and a test set with 1800 +> problems, supporting both English and Chinese. We conduct evaluations over 25 +> mainstream multi-modal large language models (MLLMs) with our Face-Human-Bench, +> focusing on the correlation between abilities, the impact of the relative +> position of targets on performance, and the impact of Chain of Thought (CoT) +> prompting on performance. Moreover, inspired by multi-modal agents, we also +> explore which abilities of MLLMs need to be supplemented by specialist models. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is on evaluating face and human understanding abilities for multi-modal assistants, not specifically on engineering, design, or optimization of prompts for Large Language Models (LLMs). While it mentions evaluating the impact of Chain of Thought (CoT) prompting, this is not the central concern, but rather one aspect of a broader evaluation." +} + +--- + +## [ProgCo: Program Helps Self-Correction of Large Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01264v1) +**arXiv ID:** http://arxiv.org/abs/2501.01264v1 + +**Abstract:** +> Self-Correction aims to enable large language models (LLMs) to self-verify +> and self-refine their initial responses without external feedback. However, +> LLMs often fail to effectively self-verify and generate correct feedback, +> further misleading refinement and leading to the failure of self-correction, +> especially in complex reasoning tasks. In this paper, we propose Program-driven +> Self-Correction (ProgCo). First, program-driven verification (ProgVe) achieves +> complex verification logic and extensive validation through self-generated, +> self-executing verification pseudo-programs. Then, program-driven refinement +> (ProgRe) receives feedback from ProgVe, conducts dual reflection and refinement +> on both responses and verification programs to mitigate misleading of incorrect +> feedback in complex reasoning tasks. Experiments on three instruction-following +> and mathematical benchmarks indicate that ProgCo achieves effective +> self-correction, and can be further enhance performance when combined with real +> program tools. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a self-correction methodology for Large Language Models (LLMs) using program-driven approaches, rather than specifically engineering, designing, or optimizing textual input prompts to improve LLM performance, not meeting the core 'MUST' criteria. + +--- + +## [CultureVLM: Characterizing and Improving Cultural Understanding of + Vision-Language Models for over 100 Countries](https://arxiv.org/abs/http://arxiv.org/abs/2501.01282v1) +**arXiv ID:** http://arxiv.org/abs/2501.01282v1 + +**Abstract:** +> Vision-language models (VLMs) have advanced human-AI interaction but struggle +> with cultural understanding, often misinterpreting symbols, gestures, and +> artifacts due to biases in predominantly Western-centric training data. In this +> paper, we construct CultureVerse, a large-scale multimodal benchmark covering +> 19, 682 cultural concepts, 188 countries/regions, 15 cultural concepts, and 3 +> question types, with the aim of characterizing and improving VLMs' +> multicultural understanding capabilities. Then, we propose CultureVLM, a series +> of VLMs fine-tuned on our dataset to achieve significant performance +> improvement in cultural understanding. Our evaluation of 16 models reveals +> significant disparities, with a stronger performance in Western concepts and +> weaker results in African and Asian contexts. Fine-tuning on our CultureVerse +> enhances cultural perception, demonstrating cross-cultural, cross-continent, +> and cross-dataset generalization without sacrificing performance on models' +> general VLM benchmarks. We further present insights on cultural generalization +> and forgetting. We hope that this work could lay the foundation for more +> equitable and culturally aware multimodal AI systems. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on fine-tuning Vision-Language Models (VLMs) for improved cultural understanding, which does not meet the 'MUST' criteria of focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). Additionally, the paper's core subject is VLMs and cultural understanding, not prompt engineering for text-based interactions with LLMs. + +--- + +## [LEO-Split: A Semi-Supervised Split Learning Framework over LEO Satellite + Networks](https://arxiv.org/abs/http://arxiv.org/abs/2501.01293v1) +**arXiv ID:** http://arxiv.org/abs/2501.01293v1 + +**Abstract:** +> Recently, the increasing deployment of LEO satellite systems has enabled +> various space analytics (e.g., crop and climate monitoring), which heavily +> relies on the advancements in deep learning (DL). However, the intermittent +> connectivity between LEO satellites and ground station (GS) significantly +> hinders the timely transmission of raw data to GS for centralized learning, +> while the scaled-up DL models hamper distributed learning on +> resource-constrained LEO satellites. Though split learning (SL) can be a +> potential solution to these problems by partitioning a model and offloading +> primary training workload to GS, the labor-intensive labeling process remains +> an obstacle, with intermittent connectivity and data heterogeneity being other +> challenges. In this paper, we propose LEO-Split, a semi-supervised (SS) SL +> design tailored for satellite networks to combat these challenges. Leveraging +> SS learning to handle (labeled) data scarcity, we construct an auxiliary model +> to tackle the training failure of the satellite-GS non-contact time. Moreover, +> we propose a pseudo-labeling algorithm to rectify data imbalances across +> satellites. Lastly, an adaptive activation interpolation scheme is devised to +> prevent the overfitting of server-side sub-model training at GS. Extensive +> experiments with real-world LEO satellite traces (e.g., Starlink) demonstrate +> that our LEO-Split framework achieves superior performance compared to +> state-ofthe-art benchmarks. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a semi-supervised split learning framework for LEO satellite networks, with an emphasis on deep learning (DL) model optimization and data transmission challenges, rather than engineering or optimizing prompts for Large Language Models (LLMs). + +--- + +## [CySecBench: Generative AI-based CyberSecurity-focused Prompt Dataset for + Benchmarking Large Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01335v1) +**arXiv ID:** http://arxiv.org/abs/2501.01335v1 + +**Abstract:** +> Numerous studies have investigated methods for jailbreaking Large Language +> Models (LLMs) to generate harmful content. Typically, these methods are +> evaluated using datasets of malicious prompts designed to bypass security +> policies established by LLM providers. However, the generally broad scope and +> open-ended nature of existing datasets can complicate the assessment of +> jailbreaking effectiveness, particularly in specific domains, notably +> cybersecurity. To address this issue, we present and publicly release +> CySecBench, a comprehensive dataset containing 12662 prompts specifically +> designed to evaluate jailbreaking techniques in the cybersecurity domain. The +> dataset is organized into 10 distinct attack-type categories, featuring +> close-ended prompts to enable a more consistent and accurate assessment of +> jailbreaking attempts. Furthermore, we detail our methodology for dataset +> generation and filtration, which can be adapted to create similar datasets in +> other domains. To demonstrate the utility of CySecBench, we propose and +> evaluate a jailbreaking approach based on prompt obfuscation. Our experimental +> results show that this method successfully elicits harmful content from +> commercial black-box LLMs, achieving Success Rates (SRs) of 65% with ChatGPT +> and 88% with Gemini; in contrast, Claude demonstrated greater resilience with a +> jailbreaking SR of 17%. Compared to existing benchmark approaches, our method +> shows superior performance, highlighting the value of domain-specific +> evaluation datasets for assessing LLM security measures. Moreover, when +> evaluated using prompts from a widely used dataset (i.e., AdvBench), it +> achieved an SR of 78.5%, higher than the state-of-the-art methods. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is on evaluating and bypassing Large Language Models' security measures in the cybersecurity domain, rather than on the engineering, design, or optimization of prompts specifically for improving LLM performance through textual input manipulation." +} + +--- + +## [ViGiL3D: A Linguistically Diverse Dataset for 3D Visual Grounding](https://arxiv.org/abs/http://arxiv.org/abs/2501.01366v1) +**arXiv ID:** http://arxiv.org/abs/2501.01366v1 + +**Abstract:** +> 3D visual grounding (3DVG) involves localizing entities in a 3D scene +> referred to by natural language text. Such models are useful for embodied AI +> and scene retrieval applications, which involve searching for objects or +> patterns using natural language descriptions. While recent works have focused +> on LLM-based scaling of 3DVG datasets, these datasets do not capture the full +> range of potential prompts which could be specified in the English language. To +> ensure that we are scaling up and testing against a useful and representative +> set of prompts, we propose a framework for linguistically analyzing 3DVG +> prompts and introduce Visual Grounding with Diverse Language in 3D (ViGiL3D), a +> diagnostic dataset for evaluating visual grounding methods against a diverse +> set of language patterns. We evaluate existing open-vocabulary 3DVG methods to +> demonstrate that these methods are not yet proficient in understanding and +> identifying the targets of more challenging, out-of-distribution prompts, +> toward real-world applications. + +**Decision Explanation:** +Original decision: REJECT +Although the paper uses LLMs and discusses the importance of diverse language prompts for 3D visual grounding, its primary focus is on creating a diagnostic dataset for evaluating visual grounding methods, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) to improve their text generation performance. + +--- + +## [Contrastive Learning from Exploratory Actions: Leveraging Natural + Interactions for Preference Elicitation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01367v1) +**arXiv ID:** http://arxiv.org/abs/2501.01367v1 + +**Abstract:** +> People have a variety of preferences for how robots behave. To understand and +> reason about these preferences, robots aim to learn a reward function that +> describes how aligned robot behaviors are with a user's preferences. Good +> representations of a robot's behavior can significantly reduce the time and +> effort required for a user to teach the robot their preferences. Specifying +> these representations -- what "features" of the robot's behavior matter to +> users -- remains a difficult problem; Features learned from raw data lack +> semantic meaning and features learned from user data require users to engage in +> tedious labeling processes. Our key insight is that users tasked with +> customizing a robot are intrinsically motivated to produce labels through +> exploratory search; they explore behaviors that they find interesting and +> ignore behaviors that are irrelevant. To harness this novel data source of +> exploratory actions, we propose contrastive learning from exploratory actions +> (CLEA) to learn trajectory features that are aligned with features that users +> care about. We learned CLEA features from exploratory actions users performed +> in an open-ended signal design activity (N=25) with a Kuri robot, and evaluated +> CLEA features through a second user study with a different set of users (N=42). +> CLEA features outperformed self-supervised features when eliciting user +> preferences over four metrics: completeness, simplicity, minimality, and +> explainability. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on learning features for preference elicitation in human-robot interaction, specifically leveraging user exploratory actions, and does not primarily investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through the manipulation of textual input prompts. + +--- + +## [ScarNet: A Novel Foundation Model for Automated Myocardial Scar + Quantification from LGE in Cardiac MRI](https://arxiv.org/abs/http://arxiv.org/abs/2501.01372v1) +**arXiv ID:** http://arxiv.org/abs/2501.01372v1 + +**Abstract:** +> Background: Late Gadolinium Enhancement (LGE) imaging is the gold standard +> for assessing myocardial fibrosis and scarring, with left ventricular (LV) LGE +> extent predicting major adverse cardiac events (MACE). Despite its importance, +> routine LGE-based LV scar quantification is hindered by labor-intensive manual +> segmentation and inter-observer variability. Methods: We propose ScarNet, a +> hybrid model combining a transformer-based encoder from the Medical Segment +> Anything Model (MedSAM) with a convolution-based U-Net decoder, enhanced by +> tailored attention blocks. ScarNet was trained on 552 ischemic cardiomyopathy +> patients with expert segmentations of myocardial and scar boundaries and tested +> on 184 separate patients. Results: ScarNet achieved robust scar segmentation in +> 184 test patients, yielding a median Dice score of 0.912 (IQR: 0.863--0.944), +> significantly outperforming MedSAM (median Dice = 0.046, IQR: 0.043--0.047) and +> nnU-Net (median Dice = 0.638, IQR: 0.604--0.661). ScarNet demonstrated lower +> bias (-0.63%) and coefficient of variation (4.3%) compared to MedSAM (bias: +> -13.31%, CoV: 130.3%) and nnU-Net (bias: -2.46%, CoV: 20.3%). In Monte Carlo +> simulations with noise perturbations, ScarNet achieved significantly higher +> scar Dice (0.892 \pm 0.053, CoV = 5.9%) than MedSAM (0.048 \pm 0.112, CoV = +> 233.3%) and nnU-Net (0.615 \pm 0.537, CoV = 28.7%). Conclusion: ScarNet +> outperformed MedSAM and nnU-Net in accurately segmenting myocardial and scar +> boundaries in LGE images. The model exhibited robust performance across diverse +> image qualities and scar patterns. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a novel foundation model (ScarNet) for medical image analysis, rather than engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not investigate or propose methods for improving LLM performance through textual input prompts. + +--- + +## [Training Medical Large Vision-Language Models with Abnormal-Aware + Feedback](https://arxiv.org/abs/http://arxiv.org/abs/2501.01377v1) +**arXiv ID:** http://arxiv.org/abs/2501.01377v1 + +**Abstract:** +> Existing Medical Large Vision-Language Models (Med-LVLMs), which encapsulate +> extensive medical knowledge, demonstrate excellent capabilities in +> understanding medical images and responding to human queries based on these +> images. However, there remain challenges in visual localization in medical +> images, which is crucial for abnormality detection and interpretation. To +> address these issues, we propose a novel UMed-LVLM designed with Unveiling +> Medical abnormalities. Specifically, we collect a Medical Abnormalities +> Unveiling (MAU) dataset and propose a two-stage training method for UMed-LVLM +> training. To collect MAU dataset, we propose a prompt method utilizing the +> GPT-4V to generate diagnoses based on identified abnormal areas in medical +> images. Moreover, the two-stage training method includes Abnormal-Aware +> Instruction Tuning and Abnormal-Aware Rewarding, comprising Abnormal +> Localization Rewarding and Vision Relevance Rewarding. Experimental results +> demonstrate that our UMed-LVLM surpasses existing Med-LVLMs in identifying and +> understanding medical abnormality. In addition, this work shows that enhancing +> the abnormality detection capabilities of Med-LVLMs significantly improves +> their understanding of medical images and generalization capability. + +**Decision Explanation:** +Original decision: REJECT +This paper primarily focuses on the development of Medical Large Vision-Language Models (Med-LVLMs) for abnormality detection in medical images, which falls under excluded categories (medical subjects and non-text generation driven by LLMs), and does not centrally focus on prompt engineering for text-based interactions with LLMs. + +--- + +## [Multi-Modal Video Feature Extraction for Popularity Prediction](https://arxiv.org/abs/http://arxiv.org/abs/2501.01422v1) +**arXiv ID:** http://arxiv.org/abs/2501.01422v1 + +**Abstract:** +> This work aims to predict the popularity of short videos using the videos +> themselves and their related features. Popularity is measured by four key +> engagement metrics: view count, like count, comment count, and share count. +> This study employs video classification models with different architectures and +> training methods as backbone networks to extract video modality features. +> Meanwhile, the cleaned video captions are incorporated into a carefully +> designed prompt framework, along with the video, as input for video-to-text +> generation models, which generate detailed text-based video content +> understanding. These texts are then encoded into vectors using a pre-trained +> BERT model. Based on the six sets of vectors mentioned above, a neural network +> is trained for each of the four prediction metrics. Moreover, the study +> conducts data mining and feature engineering based on the video and tabular +> data, constructing practical features such as the total frequency of hashtag +> appearances, the total frequency of mention appearances, video duration, frame +> count, frame rate, and total time online. Multiple machine learning models are +> trained, and the most stable model, XGBoost, is selected. Finally, the +> predictions from the neural network and XGBoost models are averaged to obtain +> the final result. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on multi-modal video feature extraction for popularity prediction, with LLMs used only as a component (video-to-text generation) and not specifically for text-based interactions with prompt engineering as the central concern, thus failing to meet the 'MUST' criteria. + +--- + +## [Object-level Visual Prompts for Compositional Image Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01424v1) +**arXiv ID:** http://arxiv.org/abs/2501.01424v1 + +**Abstract:** +> We introduce a method for composing object-level visual prompts within a +> text-to-image diffusion model. Our approach addresses the task of generating +> semantically coherent compositions across diverse scenes and styles, similar to +> the versatility and expressiveness offered by text prompts. A key challenge in +> this task is to preserve the identity of the objects depicted in the input +> visual prompts, while also generating diverse compositions across different +> images. To address this challenge, we introduce a new KV-mixed cross-attention +> mechanism, in which keys and values are learned from distinct visual +> representations. The keys are derived from an encoder with a small bottleneck +> for layout control, whereas the values come from a larger bottleneck encoder +> that captures fine-grained appearance details. By mixing keys and values from +> these complementary sources, our model preserves the identity of the visual +> prompts while supporting flexible variations in object arrangement, pose, and +> composition. During inference, we further propose object-level compositional +> guidance to improve the method's identity preservation and layout correctness. +> Results show that our technique produces diverse scene compositions that +> preserve the unique characteristics of each visual prompt, expanding the +> creative potential of text-to-image generation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on text-to-image generation with visual prompts, not large language models (LLMs) or text-based interactions, violating the 'MUST NOT' criteria for applications of generative AI other than text generation driven by LLMs. + +--- + +## [Enhancing Reasoning through Process Supervision with Monte Carlo Tree + Search](https://arxiv.org/abs/http://arxiv.org/abs/2501.01478v1) +**arXiv ID:** http://arxiv.org/abs/2501.01478v1 + +**Abstract:** +> Large language models (LLMs) have demonstrated their remarkable capacity +> across a variety of tasks. However, reasoning remains a challenge for LLMs. To +> improve LLMs' reasoning ability, process supervision has proven to be better +> than outcome supervision. In this work, we study using Monte Carlo Tree Search +> (MCTS) to generate process supervision data with LLMs themselves for training +> them. We sample reasoning steps with an LLM and assign each step a score that +> captures its "relative correctness," and the LLM is then trained by minimizing +> weighted log-likelihood of generating the reasoning steps. This +> generate-then-train process is repeated iteratively until convergence.Our +> experimental results demonstrate that the proposed methods considerably improve +> the performance of LLMs on two mathematical reasoning datasets. Furthermore, +> models trained on one dataset also exhibit improved performance on the other, +> showing the transferability of the enhanced reasoning ability. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on improving LLM's reasoning ability through process supervision with Monte Carlo Tree Search, which is a training method enhancement rather than a prompt engineering technique for text-based interactions with LLMs, violating MUST NOT criterion 1." +} + +--- + +## [ORACLE: A Real-Time, Hierarchical, Deep-Learning Photometric Classifier + for the LSST](https://arxiv.org/abs/http://arxiv.org/abs/2501.01496v1) +**arXiv ID:** http://arxiv.org/abs/2501.01496v1 + +**Abstract:** +> We present ORACLE, the first hierarchical deep-learning model for real-time, +> context-aware classification of transient and variable astrophysical phenomena. +> ORACLE is a recurrent neural network with Gated Recurrent Units (GRUs), and has +> been trained using a custom hierarchical cross-entropy loss function to provide +> high-confidence classifications along an observationally-driven taxonomy with +> as little as a single photometric observation. Contextual information for each +> object, including host galaxy photometric redshift, offset, ellipticity and +> brightness, is concatenated to the light curve embedding and used to make a +> final prediction. Training on $\sim$0.5M events from the Extended LSST +> Astronomical Time-Series Classification Challenge, we achieve a top-level +> (Transient vs Variable) macro-averaged precision of 0.96 using only 1 day of +> photometric observations after the first detection in addition to contextual +> information, for each event; this increases to $>$0.99 once 64 days of the +> light curve has been obtained, and 0.83 at 1024 days after first detection for +> 19-way classification (including supernova sub-types, active galactic nuclei, +> variable stars, microlensing events, and kilonovae). We also compare ORACLE +> with other state-of-the-art classifiers and report comparable performance for +> the 19-way classification task, in addition to delivering accurate top-level +> classifications much earlier. The code and model weights used in this work are +> publicly available at our associated GitHub repository +> (https://github.com/uiucsn/ELAsTiCC-Classification). + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet any of the 'MUST' criteria, as it focuses on developing a deep-learning model for astrophysical phenomena classification, not on prompt engineering for Large Language Models (LLMs), and doesn't investigate, analyze, or propose methods for improving LLM performance through textual input prompts. + +--- + +## [Transfer Learning Analysis of Variational Quantum Circuits](https://arxiv.org/abs/http://arxiv.org/abs/2501.01507v1) +**arXiv ID:** http://arxiv.org/abs/2501.01507v1 + +**Abstract:** +> This work analyzes transfer learning of the Variational Quantum Circuit +> (VQC). Our framework begins with a pretrained VQC configured in one domain and +> calculates the transition of 1-parameter unitary subgroups required for a new +> domain. A formalism is established to investigate the adaptability and +> capability of a VQC under the analysis of loss bounds. Our theory observes +> knowledge transfer in VQCs and provides a heuristic interpretation for the +> mechanism. An analytical fine-tuning method is derived to attain the optimal +> transition for adaptations of similar domains. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on transfer learning and optimization of Variational Quantum Circuits (VQCs), with no apparent connection to Large Language Models (LLMs) or prompt engineering for text-based interactions, failing to meet the primary 'MUST' criteria. + +--- + +## [AI-Enabled Operations at Fermi Complex: Multivariate Time Series + Prediction for Outage Prediction and Diagnosis](https://arxiv.org/abs/http://arxiv.org/abs/2501.01509v1) +**arXiv ID:** http://arxiv.org/abs/2501.01509v1 + +**Abstract:** +> The Main Control Room of the Fermilab accelerator complex continuously +> gathers extensive time-series data from thousands of sensors monitoring the +> beam. However, unplanned events such as trips or voltage fluctuations often +> result in beam outages, causing operational downtime. This downtime not only +> consumes operator effort in diagnosing and addressing the issue but also leads +> to unnecessary energy consumption by idle machines awaiting beam restoration. +> The current threshold-based alarm system is reactive and faces challenges +> including frequent false alarms and inconsistent outage-cause labeling. To +> address these limitations, we propose an AI-enabled framework that leverages +> predictive analytics and automated labeling. Using data from $2,703$ Linac +> devices and $80$ operator-labeled outages, we evaluate state-of-the-art deep +> learning architectures, including recurrent, attention-based, and linear +> models, for beam outage prediction. Additionally, we assess a Random +> Forest-based labeling system for providing consistent, confidence-scored outage +> annotations. Our findings highlight the strengths and weaknesses of these +> architectures for beam outage prediction and identify critical gaps that must +> be addressed to fully harness AI for transitioning downtime handling from +> reactive to predictive, ultimately reducing downtime and improving +> decision-making in accelerator management. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on multivariate time series prediction for outage prediction and diagnosis in an accelerator complex using deep learning architectures, with no primary focus on prompt engineering for Large Language Models (LLMs) or manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [DiagrammaticLearning: A Graphical Language for Compositional Training + Regimes](https://arxiv.org/abs/http://arxiv.org/abs/2501.01515v1) +**arXiv ID:** http://arxiv.org/abs/2501.01515v1 + +**Abstract:** +> Motivated by deep learning regimes with multiple interacting yet distinct +> model components, we introduce learning diagrams, graphical depictions of +> training setups that capture parameterized learning as data rather than code. A +> learning diagram compiles to a unique loss function on which component models +> are trained. The result of training on this loss is a collection of models +> whose predictions ``agree" with one another. We show that a number of popular +> learning setups such as few-shot multi-task learning, knowledge distillation, +> and multi-modal learning can be depicted as learning diagrams. We further +> implement learning diagrams in a library that allows users to build diagrams of +> PyTorch and Flux.jl models. By implementing some classic machine learning use +> cases, we demonstrate how learning diagrams allow practitioners to build +> complicated models as compositions of smaller components, identify +> relationships between workflows, and manipulate models during or after +> training. Leveraging a category theoretic framework, we introduce a rigorous +> semantics for learning diagrams that puts such operations on a firm +> mathematical foundation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a graphical language for compositional training regimes and a library for building complex models, which does not meet the primary criterion of focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Improving Robustness Estimates in Natural Language Explainable AI though + Synonymity Weighted Similarity Measures](https://arxiv.org/abs/http://arxiv.org/abs/2501.01516v1) +**arXiv ID:** http://arxiv.org/abs/2501.01516v1 + +**Abstract:** +> Explainable AI (XAI) has seen a surge in recent interest with the +> proliferation of powerful but intractable black-box models. Moreover, XAI has +> come under fire for techniques that may not offer reliable explanations. As +> many of the methods in XAI are themselves models, adversarial examples have +> been prominent in the literature surrounding the effectiveness of XAI, with the +> objective of these examples being to alter the explanation while maintaining +> the output of the original model. For explanations in natural language, it is +> natural to use measures found in the domain of information retrieval for use +> with ranked lists to guide the adversarial XAI process. We show that the +> standard implementation of these measures are poorly suited for the comparison +> of explanations in adversarial XAI and amend them by using information that is +> discarded, the synonymity of perturbed words. This synonymity weighting +> produces more accurate estimates of the actual weakness of XAI methods to +> adversarial examples. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on improving robustness estimates in Explainable AI (XAI) using synonymity weighted similarity measures, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), thus not meeting the core subject requirement." +} + +--- + +## [A Metasemantic-Metapragmatic Framework for Taxonomizing Multimodal + Communicative Alignment](https://arxiv.org/abs/http://arxiv.org/abs/2501.01535v1) +**arXiv ID:** http://arxiv.org/abs/2501.01535v1 + +**Abstract:** +> Drawing on contemporary pragmatist philosophy and linguistic theories on +> cognition, meaning, and communication, this paper presents a dynamic, +> metasemantic-metapragmatic taxonomy for grounding and conceptualizing +> human-like multimodal communicative alignment. The framework is rooted in +> contemporary developments of the three basic communicative capacities initially +> identified by American logician and pragmatist philosopher Charles Sanders +> Peirce: iconic (sensory and perceptual qualities), indexical (contextual and +> sociocultural associations), and rule-like (symbolic and intuitive reasoning). +> Expanding on these developments, I introduce the concept of indexical +> contextualization and propose the principle of "contextualization +> directionality" for characterizing the crucial metapragmatic capacity for +> maintaining, navigating, or transitioning between semantic and pragmatic modes +> of multimodal communication. I contend that current cognitive-social +> computational and engineering methodologies disproportionately emphasize the +> semantic/metasemantic domain, overlooking the pivotal role of metapragmatic +> indexicality in traversing the semantic-pragmatic spectrum of communication. +> The framework's broader implications for intentionality, identity, affect, and +> ethics in within-modal and cross-modal human-machine alignment are also +> discussed. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria, as it primarily focuses on a philosophical framework for multimodal communicative alignment, with no clear emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or manipulation of textual input prompts to improve LLM performance. + +--- + +## [In Search of a Lost Metric: Human Empowerment as a Pillar of Socially + Conscious Navigation](https://arxiv.org/abs/http://arxiv.org/abs/2501.01539v1) +**arXiv ID:** http://arxiv.org/abs/2501.01539v1 + +**Abstract:** +> In social robot navigation, traditional metrics like proxemics and behavior +> naturalness emphasize human comfort and adherence to social norms but often +> fail to capture an agent's autonomy and adaptability in dynamic environments. +> This paper introduces human empowerment, an information-theoretic concept that +> measures a human's ability to influence their future states and observe those +> changes, as a complementary metric for evaluating social compliance. This +> metric reveals how robot navigation policies can indirectly impact human +> empowerment. We present a framework that integrates human empowerment into the +> evaluation of social performance in navigation tasks. Through numerical +> simulations, we demonstrate that human empowerment as a metric not only aligns +> with intuitive social behavior, but also shows statistically significant +> differences across various robot navigation policies. These results provide a +> deeper understanding of how different policies affect social compliance, +> highlighting the potential of human empowerment as a complementary metric for +> future research in social navigation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on social robot navigation, introducing 'human empowerment' as a metric for evaluating social compliance, and does not meet the 'MUST' criteria for primarily focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or investigating methods to improve LLM performance through textual input prompt manipulation. + +--- + +## [BLAST: A Stealthy Backdoor Leverage Attack against Cooperative + Multi-Agent Deep Reinforcement Learning based Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.01593v1) +**arXiv ID:** http://arxiv.org/abs/2501.01593v1 + +**Abstract:** +> Recent studies have shown that cooperative multi-agent deep reinforcement +> learning (c-MADRL) is under the threat of backdoor attacks. Once a backdoor +> trigger is observed, it will perform malicious actions leading to failures or +> malicious goals. However, existing backdoor attacks suffer from several issues, +> e.g., instant trigger patterns lack stealthiness, the backdoor is trained or +> activated by an additional network, or all agents are backdoored. To this end, +> in this paper, we propose a novel backdoor leverage attack against c-MADRL, +> BLAST, which attacks the entire multi-agent team by embedding the backdoor only +> in a single agent. Firstly, we introduce adversary spatiotemporal behavior +> patterns as the backdoor trigger rather than manual-injected fixed visual +> patterns or instant status and control the period to perform malicious actions. +> This method can guarantee the stealthiness and practicality of BLAST. Secondly, +> we hack the original reward function of the backdoor agent via unilateral +> guidance to inject BLAST, so as to achieve the \textit{leverage attack effect} +> that can pry open the entire multi-agent system via a single backdoor agent. We +> evaluate our BLAST against 3 classic c-MADRL algorithms (VDN, QMIX, and MAPPO) +> in 2 popular c-MADRL environments (SMAC and Pursuit), and 2 existing defense +> mechanisms. The experimental results demonstrate that BLAST can achieve a high +> attack success rate while maintaining a low clean performance variance rate. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on a backdoor leverage attack against cooperative multi-agent deep reinforcement learning systems, with no emphasis on Large Language Models (LLMs), prompt engineering, or textual input manipulation, thus failing to meet all 'MUST' criteria. + +--- + +## [PSYCHE: A Multi-faceted Patient Simulation Framework for Evaluation of + Psychiatric Assessment Conversational Agents](https://arxiv.org/abs/http://arxiv.org/abs/2501.01594v1) +**arXiv ID:** http://arxiv.org/abs/2501.01594v1 + +**Abstract:** +> Recent advances in large language models (LLMs) have accelerated the +> development of conversational agents capable of generating human-like +> responses. Since psychiatric assessments typically involve complex +> conversational interactions between psychiatrists and patients, there is +> growing interest in developing LLM-based psychiatric assessment conversational +> agents (PACAs) that aim to simulate the role of psychiatrists in clinical +> evaluations. However, standardized methods for benchmarking the clinical +> appropriateness of PACAs' interaction with patients still remain underexplored. +> Here, we propose PSYCHE, a novel framework designed to enable the 1) clinically +> relevant, 2) ethically safe, 3) cost-efficient, and 4) quantitative evaluation +> of PACAs. This is achieved by simulating psychiatric patients based on a +> multi-faceted psychiatric construct that defines the simulated patients' +> profiles, histories, and behaviors, which PACAs are expected to assess. We +> validate the effectiveness of PSYCHE through a study with 10 board-certified +> psychiatrists, supported by an in-depth analysis of the simulated patient +> utterances. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a framework for evaluating Psychiatric Assessment Conversational Agents (PACAs), with LLMs being a component, rather than focusing on prompt engineering for text-based interactions with LLMs, not meeting the core subject requirement. + +--- + +## [A non-ergodic framework for understanding emergent capabilities in Large + Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01638v1) +**arXiv ID:** http://arxiv.org/abs/2501.01638v1 + +**Abstract:** +> Large language models have emergent capabilities that come unexpectedly at +> scale, but we need a theoretical framework to explain why and how they emerge. +> We prove that language models are actually non-ergodic systems while providing +> a mathematical framework based on Stuart Kauffman's theory of the adjacent +> possible (TAP) to explain capability emergence. Our resource-constrained TAP +> equation demonstrates how architectural, training, and contextual constraints +> interact to shape model capabilities through phase transitions in semantic +> space. We prove through experiments with three different language models that +> capacities emerge through discrete transitions guided by constraint +> interactions and path-dependent exploration. This framework provides a +> theoretical basis for understanding emergence in language models and guides the +> development of architectures that can guide capability emergence. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses primarily on developing a theoretical framework to understand emergent capabilities in LLMs, discussing architectural, training, and contextual constraints, but does not investigate, analyze, or propose methods for improving LLM performance through the manipulation of textual input prompts." +} + +--- + +## [AgentRefine: Enhancing Agent Generalization through Refinement Tuning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01702v1) +**arXiv ID:** http://arxiv.org/abs/2501.01702v1 + +**Abstract:** +> Large Language Model (LLM) based agents have proved their ability to perform +> complex tasks like humans. However, there is still a large gap between +> open-sourced LLMs and commercial models like the GPT series. In this paper, we +> focus on improving the agent generalization capabilities of LLMs via +> instruction tuning. We first observe that the existing agent training corpus +> exhibits satisfactory results on held-in evaluation sets but fails to +> generalize to held-out sets. These agent-tuning works face severe formatting +> errors and are frequently stuck in the same mistake for a long while. We +> analyze that the poor generalization ability comes from overfitting to several +> manual agent environments and a lack of adaptation to new situations. They +> struggle with the wrong action steps and can not learn from the experience but +> just memorize existing observation-action relations. Inspired by the insight, +> we propose a novel AgentRefine framework for agent-tuning. The core idea is to +> enable the model to learn to correct its mistakes via observation in the +> trajectory. Specifically, we propose an agent synthesis framework to encompass +> a diverse array of environments and tasks and prompt a strong LLM to refine its +> error action according to the environment feedback. AgentRefine significantly +> outperforms state-of-the-art agent-tuning work in terms of generalization +> ability on diverse agent tasks. It also has better robustness facing +> perturbation and can generate diversified thought in inference. Our findings +> establish the correlation between agent generalization and self-refinement and +> provide a new paradigm for future research. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on refining Large Language Model (LLM) based agents through instruction tuning to improve generalization capabilities, which aligns more with the development of new training methods for LLMs (excluded by MUST NOT 1) rather than primarily focusing on prompt engineering for text-based interactions with LLMs. + +--- + +## [Combined Hyper-Extensible Extremely-Secured Zero-Trust CIAM-PAM + architecture](https://arxiv.org/abs/http://arxiv.org/abs/2501.01732v1) +**arXiv ID:** http://arxiv.org/abs/2501.01732v1 + +**Abstract:** +> Customer Identity and Access Management (CIAM) systems play a pivotal role in +> securing enterprise infrastructures. However, the complexity of implementing +> these systems requires careful architectural planning to ensure positive Return +> on Investment (RoI) and avoid costly delays. The proliferation of Active +> Persistent cyber threats, coupled with advancements in AI, cloud computing, and +> geographically distributed customer populations, necessitates a paradigm shift +> towards adaptive and zero-trust security frameworks. This paper introduces the +> Combined Hyper-Extensible Extremely-Secured Zero-Trust (CHEZ) CIAM-PAM +> architecture, designed specifically for large-scale enterprises. The CHEZ PL +> CIAM-PAM framework addresses critical security gaps by integrating federated +> identity management (private and public identities), password-less +> authentication, adaptive multi-factor authentication (MFA), microservice-based +> PEP (Policy Entitlement Point), multi-layer RBAC (Role Based Access Control) +> and multi-level trust systems. This future-proof design also includes +> end-to-end data encryption, and seamless integration with state-of-the-art +> AI-based threat detection systems, while ensuring compliance with stringent +> regulatory standards. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a secure CIAM-PAM architecture for enterprise infrastructures, with no apparent focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs), thus failing to meet the core subject requirement. + +--- + +## [How Toxic Can You Get? Search-based Toxicity Testing for Large Language + Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01741v1) +**arXiv ID:** http://arxiv.org/abs/2501.01741v1 + +**Abstract:** +> Language is a deep-rooted means of perpetration of stereotypes and +> discrimination. Large Language Models (LLMs), now a pervasive technology in our +> everyday lives, can cause extensive harm when prone to generating toxic +> responses. The standard way to address this issue is to align the LLM, which, +> however, dampens the issue without constituting a definitive solution. +> Therefore, testing LLM even after alignment efforts remains crucial for +> detecting any residual deviations with respect to ethical standards. We present +> EvoTox, an automated testing framework for LLMs' inclination to toxicity, +> providing a way to quantitatively assess how much LLMs can be pushed towards +> toxic responses even in the presence of alignment. The framework adopts an +> iterative evolution strategy that exploits the interplay between two LLMs, the +> System Under Test (SUT) and the Prompt Generator steering SUT responses toward +> higher toxicity. The toxicity level is assessed by an automated oracle based on +> an existing toxicity classifier. We conduct a quantitative and qualitative +> empirical evaluation using four state-of-the-art LLMs as evaluation subjects +> having increasing complexity (7-13 billion parameters). Our quantitative +> evaluation assesses the cost-effectiveness of four alternative versions of +> EvoTox against existing baseline methods, based on random search, curated +> datasets of toxic prompts, and adversarial attacks. Our qualitative assessment +> engages human evaluators to rate the fluency of the generated prompts and the +> perceived toxicity of the responses collected during the testing sessions. +> Results indicate that the effectiveness, in terms of detected toxicity level, +> is significantly higher than the selected baseline methods (effect size up to +> 1.0 against random search and up to 0.99 against adversarial attacks). +> Furthermore, EvoTox yields a limited cost overhead (from 22% to 35% on +> average). + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "While the paper involves Large Language Models (LLMs) and manipulates textual input prompts to test toxicity, its primary focus is on developing a testing framework for detecting toxicity rather than optimizing prompt engineering \*for improved LLM performance\* through textual input manipulation." +} + +--- + +## [Can Synthetic Data be Fair and Private? A Comparative Study of Synthetic + Data Generation and Fairness Algorithms](https://arxiv.org/abs/http://arxiv.org/abs/2501.01785v1) +**arXiv ID:** http://arxiv.org/abs/2501.01785v1 + +**Abstract:** +> The increasing use of machine learning in learning analytics (LA) has raised +> significant concerns around algorithmic fairness and privacy. Synthetic data +> has emerged as a dual-purpose tool, enhancing privacy and improving fairness in +> LA models. However, prior research suggests an inverse relationship between +> fairness and privacy, making it challenging to optimize both. This study +> investigates which synthetic data generators can best balance privacy and +> fairness, and whether pre-processing fairness algorithms, typically applied to +> real datasets, are effective on synthetic data. Our results highlight that the +> DEbiasing CAusal Fairness (DECAF) algorithm achieves the best balance between +> privacy and fairness. However, DECAF suffers in utility, as reflected in its +> predictive accuracy. Notably, we found that applying pre-processing fairness +> algorithms to synthetic data improves fairness even more than when applied to +> real data. These findings suggest that combining synthetic data generation with +> fairness pre-processing offers a promising approach to creating fairer LA +> models. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the primary focus criteria, as it focuses on synthetic data generation for balancing privacy and fairness in learning analytics models, rather than engineering, design, or optimization of prompts specifically for Large Language Models (LLMs)." +} + +--- + +## [BERT4MIMO: A Foundation Model using BERT Architecture for Massive MIMO + Channel State Information Prediction](https://arxiv.org/abs/http://arxiv.org/abs/2501.01802v1) +**arXiv ID:** http://arxiv.org/abs/2501.01802v1 + +**Abstract:** +> Massive MIMO (Multiple-Input Multiple-Output) is an advanced wireless +> communication technology, using a large number of antennas to improve the +> overall performance of the communication system in terms of capacity, spectral, +> and energy efficiency. The performance of MIMO systems is highly dependent on +> the quality of channel state information (CSI). Predicting CSI is, therefore, +> essential for improving communication system performance, particularly in MIMO +> systems, since it represents key characteristics of a wireless channel, +> including propagation, fading, scattering, and path loss. This study proposes a +> foundation model inspired by BERT, called BERT4MIMO, which is specifically +> designed to process high-dimensional CSI data from massive MIMO systems. +> BERT4MIMO offers superior performance in reconstructing CSI under varying +> mobility scenarios and channel conditions through deep learning and attention +> mechanisms. The experimental results demonstrate the effectiveness of BERT4MIMO +> in a variety of wireless environments. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a new foundation model (BERT4MIMO) for predicting channel state information in Massive MIMO systems, which does not meet the primary criteria of focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Auto-RT: Automatic Jailbreak Strategy Exploration for Red-Teaming Large + Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.01830v1) +**arXiv ID:** http://arxiv.org/abs/2501.01830v1 + +**Abstract:** +> Automated red-teaming has become a crucial approach for uncovering +> vulnerabilities in large language models (LLMs). However, most existing methods +> focus on isolated safety flaws, limiting their ability to adapt to dynamic +> defenses and uncover complex vulnerabilities efficiently. To address this +> challenge, we propose Auto-RT, a reinforcement learning framework that +> automatically explores and optimizes complex attack strategies to effectively +> uncover security vulnerabilities through malicious queries. Specifically, we +> introduce two key mechanisms to reduce exploration complexity and improve +> strategy optimization: 1) Early-terminated Exploration, which accelerate +> exploration by focusing on high-potential attack strategies; and 2) Progressive +> Reward Tracking algorithm with intermediate downgrade models, which dynamically +> refine the search trajectory toward successful vulnerability exploitation. +> Extensive experiments across diverse LLMs demonstrate that, by significantly +> improving exploration efficiency and automatically optimizing attack +> strategies, Auto-RT detects a boarder range of vulnerabilities, achieving a +> faster detection speed and 16.63\% higher success rates compared to existing +> methods. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on developing a reinforcement learning framework for automated red-teaming to uncover security vulnerabilities in LLMs, rather than focusing on the engineering, design, or optimization of prompts specifically for improving LLM performance through textual input manipulation." +} + +--- + +## [LCFed: An Efficient Clustered Federated Learning Framework for + Heterogeneous Data](https://arxiv.org/abs/http://arxiv.org/abs/2501.01850v1) +**arXiv ID:** http://arxiv.org/abs/2501.01850v1 + +**Abstract:** +> Clustered federated learning (CFL) addresses the performance challenges posed +> by data heterogeneity in federated learning (FL) by organizing edge devices +> with similar data distributions into clusters, enabling collaborative model +> training tailored to each group. However, existing CFL approaches strictly +> limit knowledge sharing to within clusters, lacking the integration of global +> knowledge with intra-cluster training, which leads to suboptimal performance. +> Moreover, traditional clustering methods incur significant computational +> overhead, especially as the number of edge devices increases. In this paper, we +> propose LCFed, an efficient CFL framework to combat these challenges. By +> leveraging model partitioning and adopting distinct aggregation strategies for +> each sub-model, LCFed effectively incorporates global knowledge into +> intra-cluster co-training, achieving optimal training performance. +> Additionally, LCFed customizes a computationally efficient model similarity +> measurement method based on low-rank models, enabling real-time cluster updates +> with minimal computational overhead. Extensive experiments show that LCFed +> outperforms state-of-the-art benchmarks in both test accuracy and clustering +> computational efficiency. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a new federated learning framework (LCFed) for heterogeneous data, with no mention of prompt engineering, Large Language Models (LLMs), or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Evaluating Scenario-based Decision-making for Interactive Autonomous + Driving Using Rational Criteria: A Survey](https://arxiv.org/abs/http://arxiv.org/abs/2501.01886v1) +**arXiv ID:** http://arxiv.org/abs/2501.01886v1 + +**Abstract:** +> Autonomous vehicles (AVs) can significantly promote the advances in road +> transport mobility in terms of safety, reliability, and decarbonization. +> However, ensuring safety and efficiency in interactive during within dynamic +> and diverse environments is still a primary barrier to large-scale AV adoption. +> In recent years, deep reinforcement learning (DRL) has emerged as an advanced +> AI-based approach, enabling AVs to learn decision-making strategies adaptively +> from data and interactions. DRL strategies are better suited than traditional +> rule-based methods for handling complex, dynamic, and unpredictable driving +> environments due to their adaptivity. However, varying driving scenarios +> present distinct challenges, such as avoiding obstacles on highways and +> reaching specific exits at intersections, requiring different scenario-specific +> decision-making algorithms. Many DRL algorithms have been proposed in +> interactive decision-making. However, a rationale review of these DRL +> algorithms across various scenarios is lacking. Therefore, a comprehensive +> evaluation is essential to assess these algorithms from multiple perspectives, +> including those of vehicle users and vehicle manufacturers. This survey reviews +> the application of DRL algorithms in autonomous driving across typical +> scenarios, summarizing road features and recent advancements. The scenarios +> include highways, on-ramp merging, roundabouts, and unsignalized intersections. +> Furthermore, DRL-based algorithms are evaluated based on five rationale +> criteria: driving safety, driving efficiency, training efficiency, +> unselfishness, and interpretability (DDTUI). Each criterion of DDTUI is +> specifically analyzed in relation to the reviewed algorithms. Finally, the +> challenges for future DRL-based decision-making algorithms are summarized. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on autonomous driving, deep reinforcement learning algorithms, and decision-making for autonomous vehicles, rather than on the engineering, design, or optimization of prompts for Large Language Models (LLMs), failing to meet the core subject requirement. + +--- + +## [QuArch: A Question-Answering Dataset for AI Agents in Computer + Architecture](https://arxiv.org/abs/http://arxiv.org/abs/2501.01892v2) +**arXiv ID:** http://arxiv.org/abs/2501.01892v2 + +**Abstract:** +> We introduce QuArch, a dataset of 1500 human-validated question-answer pairs +> designed to evaluate and enhance language models' understanding of computer +> architecture. The dataset covers areas including processor design, memory +> systems, and performance optimization. Our analysis highlights a significant +> performance gap: the best closed-source model achieves 84% accuracy, while the +> top small open-source model reaches 72%. We observe notable struggles in memory +> systems, interconnection networks, and benchmarking. Fine-tuning with QuArch +> improves small model accuracy by up to 8%, establishing a foundation for +> advancing AI-driven computer architecture research. The dataset and leaderboard +> are at https://harvard-edge.github.io/QuArch/. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on introducing a dataset (QuArch) for evaluating and enhancing language models' understanding of computer architecture, rather than specifically investigating, analyzing, or proposing methods for improving LLM performance through the manipulation of textual input prompts. + +--- + +## [Mingling with the Good to Backdoor Federated Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01913v1) +**arXiv ID:** http://arxiv.org/abs/2501.01913v1 + +**Abstract:** +> Federated learning (FL) is a decentralized machine learning technique that +> allows multiple entities to jointly train a model while preserving dataset +> privacy. However, its distributed nature has raised various security concerns, +> which have been addressed by increasingly sophisticated defenses. These +> protections utilize a range of data sources and metrics to, for example, filter +> out malicious model updates, ensuring that the impact of attacks is minimized +> or eliminated. +> This paper explores the feasibility of designing a generic attack method +> capable of installing backdoors in FL while evading a diverse array of +> defenses. Specifically, we focus on an attacker strategy called MIGO, which +> aims to produce model updates that subtly blend with legitimate ones. The +> resulting effect is a gradual integration of a backdoor into the global model, +> often ensuring its persistence long after the attack concludes, while +> generating enough ambiguity to hinder the effectiveness of defenses. +> MIGO was employed to implant three types of backdoors across five datasets +> and different model architectures. The results demonstrate the significant +> threat posed by these backdoors, as MIGO consistently achieved exceptionally +> high backdoor accuracy (exceeding 90%) while maintaining the utility of the +> main task. Moreover, MIGO exhibited strong evasion capabilities against ten +> defenses, including several state-of-the-art methods. When compared to four +> other attack strategies, MIGO consistently outperformed them across most +> configurations. Notably, even in extreme scenarios where the attacker controls +> just 0.1% of the clients, the results indicate that successful backdoor +> insertion is possible if the attacker can persist for a sufficient number of +> rounds. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on backdoor attacks in Federated Learning, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [On the Utility of Equivariance and Symmetry Breaking in Deep Learning + Architectures on Point Clouds](https://arxiv.org/abs/http://arxiv.org/abs/2501.01999v1) +**arXiv ID:** http://arxiv.org/abs/2501.01999v1 + +**Abstract:** +> This paper explores the key factors that influence the performance of models +> working with point clouds, across different tasks of varying geometric +> complexity. In this work, we explore the trade-offs between flexibility and +> weight-sharing introduced by equivariant layers, assessing when equivariance +> boosts or detracts from performance. It is often argued that providing more +> information as input improves a model's performance. However, if this +> additional information breaks certain properties, such as $\SE(3)$ +> equivariance, does it remain beneficial? We identify the key aspects of +> equivariant and non-equivariant architectures that drive success in different +> tasks by benchmarking them on segmentation, regression, and generation tasks +> across multiple datasets with increasing complexity. We observe a positive +> impact of equivariance, which becomes more pronounced with increasing task +> complexity, even when strict equivariance is not required. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development and performance of deep learning architectures for point clouds, discussing equivariance and symmetry breaking, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions. + +--- + +## [Multi-Center Study on Deep Learning-Assisted Detection and + Classification of Fetal Central Nervous System Anomalies Using Ultrasound + Imaging](https://arxiv.org/abs/http://arxiv.org/abs/2501.02000v1) +**arXiv ID:** http://arxiv.org/abs/2501.02000v1 + +**Abstract:** +> Prenatal ultrasound evaluates fetal growth and detects congenital +> abnormalities during pregnancy, but the examination of ultrasound images by +> radiologists requires expertise and sophisticated equipment, which would +> otherwise fail to improve the rate of identifying specific types of fetal +> central nervous system (CNS) abnormalities and result in unnecessary patient +> examinations. We construct a deep learning model to improve the overall +> accuracy of the diagnosis of fetal cranial anomalies to aid prenatal diagnosis. +> In our collected multi-center dataset of fetal craniocerebral anomalies +> covering four typical anomalies of the fetal central nervous system (CNS): +> anencephaly, encephalocele (including meningocele), holoprosencephaly, and +> rachischisis, patient-level prediction accuracy reaches 94.5%, with an AUROC +> value of 99.3%. In the subgroup analyzes, our model is applicable to the entire +> gestational period, with good identification of fetal anomaly types for any +> gestational period. Heatmaps superimposed on the ultrasound images not only +> provide a visual interpretation for the algorithm but also provide an intuitive +> visual aid to the physician by highlighting key areas that need to be reviewed, +> helping the physician to quickly identify and validate key areas. Finally, the +> retrospective reader study demonstrates that by combining the automatic +> prediction of the DL system with the professional judgment of the radiologist, +> the diagnostic accuracy and efficiency can be effectively improved and the +> misdiagnosis rate can be reduced, which has an important clinical application +> prospect. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on deep learning-assisted detection and classification of fetal central nervous system anomalies using ultrasound imaging, which falls under medical subjects and does not involve prompt engineering for Large Language Models (LLMs) as required. + +--- + +## [General Information Metrics for Improving AI Model Training Efficiency](https://arxiv.org/abs/http://arxiv.org/abs/2501.02004v1) +**arXiv ID:** http://arxiv.org/abs/2501.02004v1 + +**Abstract:** +> To address the growing size of AI model training data and the lack of a +> universal data selection methodology-factors that significantly drive up +> training costs -- this paper presents the General Information Metrics +> Evaluation (GIME) method. GIME leverages general information metrics from +> Objective Information Theory (OIT), including volume, delay, scope, +> granularity, variety, duration, sampling rate, aggregation, coverage, +> distortion, and mismatch to optimize dataset selection for training purposes. +> Comprehensive experiments conducted across diverse domains, such as CTR +> Prediction, Civil Case Prediction, and Weather Forecasting, demonstrate that +> GIME effectively preserves model performance while substantially reducing both +> training time and costs. Additionally, applying GIME within the Judicial AI +> Program led to a remarkable 39.56% reduction in total model training expenses, +> underscoring its potential to support efficient and sustainable AI development. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on optimizing dataset selection for AI model training efficiency using General Information Metrics, not on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), thus failing to meet the primary 'MUST' criteria." +} + +--- + +## [ST-HCSS: Deep Spatio-Temporal Hypergraph Convolutional Neural Network + for Soft Sensing](https://arxiv.org/abs/http://arxiv.org/abs/2501.02016v1) +**arXiv ID:** http://arxiv.org/abs/2501.02016v1 + +**Abstract:** +> Higher-order sensor networks are more accurate in characterizing the +> nonlinear dynamics of sensory time-series data in modern industrial settings by +> allowing multi-node connections beyond simple pairwise graph edges. In light of +> this, we propose a deep spatio-temporal hypergraph convolutional neural network +> for soft sensing (ST-HCSS). In particular, our proposed framework is able to +> construct and leverage a higher-order graph (hypergraph) to model the complex +> multi-interactions between sensor nodes in the absence of prior structural +> knowledge. To capture rich spatio-temporal relationships underlying sensor +> data, our proposed ST-HCSS incorporates stacked gated temporal and hypergraph +> convolution layers to effectively aggregate and update hypergraph information +> across time and nodes. Our results validate the superiority of ST-HCSS compared +> to existing state-of-the-art soft sensors, and demonstrates that the learned +> hypergraph feature representations aligns well with the sensor data +> correlations. The code is available at https://github.com/htew0001/ST-HCSS.git + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on developing a new neural network architecture (ST-HCSS) for soft sensing in industrial settings, with no apparent primary focus on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate improving LLM performance through textual input prompts." +} + +--- + +## [Safeguarding Large Language Models in Real-time with Tunable + Safety-Performance Trade-offs](https://arxiv.org/abs/http://arxiv.org/abs/2501.02018v1) +**arXiv ID:** http://arxiv.org/abs/2501.02018v1 + +**Abstract:** +> Large Language Models (LLMs) have been shown to be susceptible to jailbreak +> attacks, or adversarial attacks used to illicit high risk behavior from a +> model. Jailbreaks have been exploited by cybercriminals and blackhat actors to +> cause significant harm, highlighting the critical need to safeguard +> widely-deployed models. Safeguarding approaches, which include fine-tuning +> models or having LLMs "self-reflect", may lengthen the inference time of a +> model, incur a computational penalty, reduce the semantic fluency of an output, +> and restrict ``normal'' model behavior. Importantly, these Safety-Performance +> Trade-offs (SPTs) remain an understudied area. In this work, we introduce a +> novel safeguard, called SafeNudge, that combines Controlled Text Generation +> with "nudging", or using text interventions to change the behavior of a model. +> SafeNudge triggers during text-generation while a jailbreak attack is being +> executed, and can reduce successful jailbreak attempts by 30% by guiding the +> LLM towards a safe responses. It adds minimal latency to inference and has a +> negligible impact on the semantic fluency of outputs. Further, we allow for +> tunable SPTs. SafeNudge is open-source and available through https://pypi.org/, +> and is compatible with models loaded with the Hugging Face "transformers" +> library. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on safeguarding LLMs from jailbreak attacks through a novel method (SafeNudge), which involves Controlled Text Generation and 'nudging'. While it does manipulate textual input, the core subject is enhancing LLM safety and security, not prompt engineering for text-based interactions with LLMs, thus not meeting the 'MUST' criteria. + +--- + +## [Benchmarking Constraint-Based Bayesian Structure Learning Algorithms: + Role of Network Topology](https://arxiv.org/abs/http://arxiv.org/abs/2501.02019v1) +**arXiv ID:** http://arxiv.org/abs/2501.02019v1 + +**Abstract:** +> Modeling the associations between real world entities from their multivariate +> cross-sectional profiles can provide cues into the concerted working of these +> entities as a system. Several techniques have been proposed for deciphering +> these associations including constraint-based Bayesian structure learning (BSL) +> algorithms that model them as directed acyclic graphs. Benchmarking these +> algorithms have typically focused on assessing the variation in performance +> measures such as sensitivity as a function of the dimensionality represented by +> the number of nodes in the DAG, and sample size. The present study elucidates +> the importance of network topology in benchmarking exercises. More +> specifically, it investigates variations in sensitivity across distinct network +> topologies while constraining the nodes, edges, and sample-size to be +> identical, eliminating these as potential confounders. Sensitivity of three +> popular constraint-based BSL algorithms (Peter-Clarke, Grow-Shrink, Incremental +> Association Markov Blanket) in learning the network structure from multivariate +> cross-sectional profiles sampled from network models with sub-linear, linear, +> and super-linear DAG topologies generated using preferential attachment is +> investigated. Results across linear and nonlinear models revealed statistically +> significant $(\alpha=0.05)$ decrease in sensitivity estimates from sub-linear +> to super-linear topology constitutively across the three algorithms. These +> results are demonstrated on networks with nodes $(N_{nods}=48,64)$, noise +> strengths $(\sigma =3,6)$ and sample size $(N = 2^{10})$. The findings +> elucidate the importance of accommodating the network topology in +> constraint-based BSL benchmarking exercises. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the 'MUST' criteria as it focuses on benchmarking constraint-based Bayesian structure learning algorithms, which is unrelated to engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not investigate the manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [Model Checking in Medical Imaging for Tumor Detection and Segmentation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02024v2) +**arXiv ID:** http://arxiv.org/abs/2501.02024v2 + +**Abstract:** +> Recent advancements in model checking have demonstrated significant potential +> across diverse applications, particularly in signal and image analysis. Medical +> imaging stands out as a critical domain where model checking can be effectively +> applied to design and evaluate robust frameworks. These frameworks facilitate +> automatic and semi-automatic delineation of regions of interest within images, +> aiding in accurate segmentation. This paper provides a comprehensive analysis +> of recent works leveraging spatial logic to develop operators and tools for +> identifying regions of interest, including tumorous and non-tumorous areas. +> Additionally, we examine the challenges inherent to spatial model-checking +> techniques, such as variability in ground truth data and the need for +> streamlined procedures suitable for routine clinical practice. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on medical imaging for tumor detection and segmentation, and model checking techniques, rather than prompt engineering for Large Language Models (LLMs), thus violating the 'MUST NOT' criteria regarding primary concern with medical subjects and not meeting the core subject requirement of prompt engineering for text-based interactions with LLMs. + +--- + +## [Spot Risks Before Speaking! Unraveling Safety Attention Heads in Large + Vision-Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02029v1) +**arXiv ID:** http://arxiv.org/abs/2501.02029v1 + +**Abstract:** +> With the integration of an additional modality, large vision-language models +> (LVLMs) exhibit greater vulnerability to safety risks (e.g., jailbreaking) +> compared to their language-only predecessors. Although recent studies have +> devoted considerable effort to the post-hoc alignment of LVLMs, the inner +> safety mechanisms remain largely unexplored. In this paper, we discover that +> internal activations of LVLMs during the first token generation can effectively +> identify malicious prompts across different attacks. This inherent safety +> perception is governed by sparse attention heads, which we term ``safety +> heads." Further analysis reveals that these heads act as specialized shields +> against malicious prompts; ablating them leads to higher attack success rates, +> while the model's utility remains unaffected. By locating these safety heads +> and concatenating their activations, we construct a straightforward but +> powerful malicious prompt detector that integrates seamlessly into the +> generation process with minimal extra inference overhead. Despite its simple +> structure of a logistic regression model, the detector surprisingly exhibits +> strong zero-shot generalization capabilities. Experiments across various +> prompt-based attacks confirm the effectiveness of leveraging safety heads to +> protect LVLMs. Code is available at \url{https://github.com/Ziwei-Zheng/SAHs}. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on identifying safety risks in Large Vision-Language Models (LVLMs) through internal attention heads, rather than specifically engineering or optimizing textual input prompts for Large Language Models (LLMs), thus not meeting the primary focus criteria. + +--- + +## [Detecting Music Performance Errors with Transformers](https://arxiv.org/abs/http://arxiv.org/abs/2501.02030v1) +**arXiv ID:** http://arxiv.org/abs/2501.02030v1 + +**Abstract:** +> Beginner musicians often struggle to identify specific errors in their +> performances, such as playing incorrect notes or rhythms. There are two +> limitations in existing tools for music error detection: (1) Existing +> approaches rely on automatic alignment; therefore, they are prone to errors +> caused by small deviations between alignment targets.; (2) There is a lack of +> sufficient data to train music error detection models, resulting in +> over-reliance on heuristics. To address (1), we propose a novel transformer +> model, Polytune, that takes audio inputs and outputs annotated music scores. +> This model can be trained end-to-end to implicitly align and compare +> performance audio with music scores through latent space representations. To +> address (2), we present a novel data generation technique capable of creating +> large-scale synthetic music error datasets. Our approach achieves a 64.1% +> average Error Detection F1 score, improving upon prior work by 40 percentage +> points across 14 instruments. Additionally, compared with existing +> transcription methods repurposed for music error detection, our model can +> handle multiple instruments. Our source code and datasets are available at +> https://github.com/ben2002chou/Polytune. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on music performance error detection using a novel transformer model and data generation technique, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or their textual input prompts. + +--- + +## [Dynamic Feature Fusion: Combining Global Graph Structures and Local + Semantics for Blockchain Fraud Detection](https://arxiv.org/abs/http://arxiv.org/abs/2501.02032v1) +**arXiv ID:** http://arxiv.org/abs/2501.02032v1 + +**Abstract:** +> The advent of blockchain technology has facilitated the widespread adoption +> of smart contracts in the financial sector. However, current fraud detection +> methodologies exhibit limitations in capturing both global structural patterns +> within transaction networks and local semantic relationships embedded in +> transaction data. Most existing models focus on either structural information +> or semantic features individually, leading to suboptimal performance in +> detecting complex fraud patterns.In this paper, we propose a dynamic feature +> fusion model that combines graph-based representation learning and semantic +> feature extraction for blockchain fraud detection. Specifically, we construct +> global graph representations to model account relationships and extract local +> contextual features from transaction data. A dynamic multimodal fusion +> mechanism is introduced to adaptively integrate these features, enabling the +> model to capture both structural and semantic fraud patterns effectively. We +> further develop a comprehensive data processing pipeline, including graph +> construction, temporal feature enhancement, and text preprocessing. +> Experimental results on large-scale real-world blockchain datasets demonstrate +> that our method outperforms existing benchmarks across accuracy, F1 score, and +> recall metrics. This work highlights the importance of integrating structural +> relationships and semantic similarities for robust fraud detection and offers a +> scalable solution for securing blockchain systems. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Instead, it concentrates on a dynamic feature fusion model for blockchain fraud detection, utilizing graph-based representation learning and semantic feature extraction, with no evident connection to LLM prompt engineering. + +--- + +## [Deep Clustering via Community Detection](https://arxiv.org/abs/http://arxiv.org/abs/2501.02036v1) +**arXiv ID:** http://arxiv.org/abs/2501.02036v1 + +**Abstract:** +> Deep clustering is an essential task in modern artificial intelligence, +> aiming to partition a set of data samples into a given number of homogeneous +> groups (i.e., clusters). Even though many Deep Neural Network (DNN) backbones +> and clustering strategies have been proposed for the task, achieving +> increasingly improved performance, deep clustering remains very challenging due +> to the lack of accurately labeled samples. In this paper, we propose a novel +> approach of deep clustering via community detection. It initializes clustering +> by detecting many communities, and then gradually expands clusters by community +> merging. Compared with the existing clustering strategies, community detection +> factors in the new perspective of cluster network analysis. As a result, it has +> the inherent benefit of high pseudo-label purity, which is critical to the +> performance of self-supervision. We have validated the efficacy of the proposed +> approach on benchmark image datasets. Our extensive experiments have shown that +> it can effectively improve the SOTA performance. Our ablation study also +> demonstrates that the new network perspective can effectively improve community +> pseudo-label purity, resulting in improved clustering performance. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on deep clustering via community detection for image datasets using Deep Neural Networks (DNNs), with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [METAGENE-1: Metagenomic Foundation Model for Pandemic Monitoring](https://arxiv.org/abs/http://arxiv.org/abs/2501.02045v1) +**arXiv ID:** http://arxiv.org/abs/2501.02045v1 + +**Abstract:** +> We pretrain METAGENE-1, a 7-billion-parameter autoregressive transformer +> model, which we refer to as a metagenomic foundation model, on a novel corpus +> of diverse metagenomic DNA and RNA sequences comprising over 1.5 trillion base +> pairs. This dataset is sourced from a large collection of human wastewater +> samples, processed and sequenced using deep metagenomic (next-generation) +> sequencing methods. Unlike genomic models that focus on individual genomes or +> curated sets of specific species, the aim of METAGENE-1 is to capture the full +> distribution of genomic information present within this wastewater, to aid in +> tasks relevant to pandemic monitoring and pathogen detection. We carry out +> byte-pair encoding (BPE) tokenization on our dataset, tailored for metagenomic +> sequences, and then pretrain our model. In this paper, we first detail the +> pretraining dataset, tokenization strategy, and model architecture, +> highlighting the considerations and design choices that enable the effective +> modeling of metagenomic data. We then show results of pretraining this model on +> our metagenomic dataset, providing details about our losses, system metrics, +> and training stability over the course of pretraining. Finally, we demonstrate +> the performance of METAGENE-1, which achieves state-of-the-art results on a set +> of genomic benchmarks and new evaluations focused on human-pathogen detection +> and genomic sequence embedding, showcasing its potential for public health +> applications in pandemic monitoring, biosurveillance, and early detection of +> emerging health threats. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a new foundation model (METAGENE-1) for pandemic monitoring through pretraining on metagenomic sequences, rather than on prompt engineering for Large Language Models (LLMs), and does not investigate or propose methods for improving LLM performance through textual input prompt manipulation. + +--- + +## [Relaxation-assisted reverse annealing on nonnegative/binary matrix + factorization](https://arxiv.org/abs/http://arxiv.org/abs/2501.02114v1) +**arXiv ID:** http://arxiv.org/abs/2501.02114v1 + +**Abstract:** +> Quantum annealing has garnered significant attention as meta-heuristics +> inspired by quantum physics for combinatorial optimization problems. Among its +> many applications, nonnegative/binary matrix factorization stands out for its +> complexity and relevance in unsupervised machine learning. The use of reverse +> annealing, a derivative procedure of quantum annealing to prioritize the search +> in a vicinity under a given initial state, helps improve its optimization +> performance in matrix factorization. This study proposes an improved strategy +> that integrates reverse annealing with a linear programming relaxation +> technique. Using relaxed solutions as the initial configuration for reverse +> annealing, we demonstrate improvements in optimization performance comparable +> to the exact optimization methods. Our experiments on facial image datasets +> show that our method provides better convergence than known reverse annealing +> methods. Furthermore, we investigate the effectiveness of relaxation-based +> initialization methods on randomized datasets, demonstrating a relationship +> between the relaxed solution and the optimal solution. This research +> underscores the potential of combining reverse annealing and classical +> optimization strategies to enhance optimization performance. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on optimizing nonnegative/binary matrix factorization using quantum annealing and linear programming relaxation, with no apparent connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet the primary 'MUST' criteria. + +--- + +## [A hybrid marketplace of ideas](https://arxiv.org/abs/http://arxiv.org/abs/2501.02132v2) +**arXiv ID:** http://arxiv.org/abs/2501.02132v2 + +**Abstract:** +> The convergence of humans and artificial intelligence systems introduces new +> dynamics into the cultural and intellectual landscape. Complementing emerging +> cultural evolution concepts such as machine culture, AI agents represent a +> significant techno-sociological development, particularly within the +> anthropological study of Web3 as a community focused on decentralization +> through blockchain. Despite their growing presence, the cultural significance +> of AI agents remains largely unexplored in academic literature. Toward this +> end, we conceived hybrid netnography, a novel interdisciplinary approach that +> examines the cultural and intellectual dynamics within digital ecosystems by +> analyzing the interactions and contributions of both human and AI agents as +> co-participants in shaping narratives, ideas, and cultural artifacts. We argue +> that, within the Web3 community on the social media platform X, these agents +> challenge traditional notions of participation and influence in public +> discourse, creating a hybrid marketplace of ideas, a conceptual space where +> human and AI generated ideas coexist and compete for attention. We examine the +> current state of AI agents in idea generation, propagation, and engagement, +> positioning their role as cultural agents through the lens of memetics and +> encouraging further inquiry into their cultural and societal impact. +> Additionally, we address the implications of this paradigm for privacy, +> intellectual property, and governance, highlighting the societal and legal +> challenges of integrating AI agents into the hybrid marketplace of ideas. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on the societal and cultural impact of AI agents in a Web3 community, not on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), failing to meet the core subject requirement. + +--- + +## [Effective LLM-Driven Code Generation with Pythoness](https://arxiv.org/abs/http://arxiv.org/abs/2501.02138v1) +**arXiv ID:** http://arxiv.org/abs/2501.02138v1 + +**Abstract:** +> The advent of large language models (LLMs) has paved the way for a new era of +> programming tools with both significant capabilities and risks, as the +> generated code lacks guarantees of correctness and reliability. Developers +> using LLMs currently face the difficult task of optimizing, integrating, and +> maintaining code generated by AI. We propose an embedded domain-specific +> language (DSL), Pythoness, to address those challenges. In Pythoness, +> developers program with LLMs at a higher level of abstraction. Rather than +> interacting directly with generated code, developers using Pythoness operate at +> the level of behavioral specifications when writing functions, classes, or an +> entire program. These specifications can take the form of unit tests and +> property-based tests, which may be expressed formally or in natural language. +> Guided by these specifications, Pythoness generates code that both passes the +> tests and can be continuously checked during execution. We posit that the +> Pythoness approach lets developers harness the full potential of LLMs for code +> generation while substantially mitigating their inherent risks. We describe our +> current prototype implementation of Pythoness and demonstrate that it can +> successfully leverage a combination of tests and code generation to yield +> higher quality code than specifications alone. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing an embedded domain-specific language (DSL) for safer code generation with LLMs, rather than on prompt engineering for text-based interactions with LLMs. While LLMs are utilized, the core subject is the DSL (Pythoness) and its application, not the optimization or design of textual input prompts for improving LLM performance. + +--- + +## [Establishing baselines for generative discovery of inorganic crystals](https://arxiv.org/abs/http://arxiv.org/abs/2501.02144v1) +**arXiv ID:** http://arxiv.org/abs/2501.02144v1 + +**Abstract:** +> Generative artificial intelligence offers a promising avenue for materials +> discovery, yet its advantages over traditional methods remain unclear. In this +> work, we introduce and benchmark two baseline approaches - random enumeration +> of charge-balanced prototypes and data-driven ion exchange of known compounds - +> against three generative models: a variational autoencoder, a large language +> model, and a diffusion model. Our results show that established methods such as +> ion exchange perform comparably well in generating stable materials, although +> many of these materials tend to closely resemble known compounds. In contrast, +> generative models excel at proposing novel structural frameworks and, when +> sufficient training data is available, can more effectively target properties +> such as electronic band gap and bulk modulus while maintaining a high stability +> rate. To enhance the performance of both the baseline and generative +> approaches, we implement a post-generation screening step in which all proposed +> structures are passed through stability and property filters from pre-trained +> machine learning models including universal interatomic potentials. This +> low-cost filtering step leads to substantial improvement in the success rates +> of all methods, remains computationally efficient, and ultimately provides a +> practical pathway toward more effective generative strategies for materials +> discovery. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on generative discovery of inorganic crystals using various AI models, including LLMs, but does not centralize prompt engineering for text-based interactions with LLMs. Instead, LLMs are utilized as one of several generative models for materials discovery, not meeting the 'MUST' criteria of focusing primarily on prompt engineering for LLMs." +} + +--- + +## [Plasma-CycleGAN: Plasma Biomarker-Guided MRI to PET Cross-modality + Translation Using Conditional CycleGAN](https://arxiv.org/abs/http://arxiv.org/abs/2501.02146v1) +**arXiv ID:** http://arxiv.org/abs/2501.02146v1 + +**Abstract:** +> Cross-modality translation between MRI and PET imaging is challenging due to +> the distinct mechanisms underlying these modalities. Blood-based biomarkers +> (BBBMs) are revolutionizing Alzheimer's disease (AD) detection by identifying +> patients and quantifying brain amyloid levels. However, the potential of BBBMs +> to enhance PET image synthesis remains unexplored. In this paper, we performed +> a thorough study on the effect of incorporating BBBM into deep generative +> models. By evaluating three widely used cross-modality translation models, we +> found that BBBMs integration consistently enhances the generative quality +> across all models. By visual inspection of the generated results, we observed +> that PET images generated by CycleGAN exhibit the best visual fidelity. Based +> on these findings, we propose Plasma-CycleGAN, a novel generative model based +> on CycleGAN, to synthesize PET images from MRI using BBBMs as conditions. This +> is the first approach to integrate BBBMs in conditional cross-modality +> translation between MRI and PET. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on cross-modality translation between MRI and PET imaging using conditional CycleGAN, with an emphasis on integrating blood-based biomarkers for Alzheimer's disease detection, and does not meet the 'MUST' criteria as it lacks focus on prompt engineering, LLMs, and textual input manipulation. + +--- + +## [The Race to Efficiency: A New Perspective on AI Scaling Laws](https://arxiv.org/abs/http://arxiv.org/abs/2501.02156v3) +**arXiv ID:** http://arxiv.org/abs/2501.02156v3 + +**Abstract:** +> As large-scale AI models expand, training becomes costlier and sustaining +> progress grows harder. Classical scaling laws (e.g., Kaplan et al. (2020), +> Hoffmann et al. (2022)) predict training loss from a static compute budget yet +> neglect time and efficiency, prompting the question: how can we balance +> ballooning GPU fleets with rapidly improving hardware and algorithms? We +> introduce the relative-loss equation, a time- and efficiency-aware framework +> that extends classical AI scaling laws. Our model shows that, without ongoing +> efficiency gains, advanced performance could demand millennia of training or +> unrealistically large GPU fleets. However, near-exponential progress remains +> achievable if the "efficiency-doubling rate" parallels Moore's Law. By +> formalizing this race to efficiency, we offer a quantitative roadmap for +> balancing front-loaded GPU investments with incremental improvements across the +> AI stack. Empirical trends suggest that sustained efficiency gains can push AI +> scaling well into the coming decade, providing a new perspective on the +> diminishing returns inherent in classical scaling. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on AI scaling laws, efficiency, and balancing GPU investments, with no clear emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria. + +--- + +## [Can ChatGPT implement finite element models for geotechnical engineering + applications?](https://arxiv.org/abs/http://arxiv.org/abs/2501.02199v1) +**arXiv ID:** http://arxiv.org/abs/2501.02199v1 + +**Abstract:** +> This study assesses the capability of ChatGPT to generate finite element code +> for geotechnical engineering applications from a set of prompts. We tested +> three different initial boundary value problems using a hydro-mechanically +> coupled formulation for unsaturated soils, including the dissipation of excess +> pore water pressure through fluid mass diffusion in one-dimensional space, +> time-dependent differential settlement of a strip footing, and gravity-driven +> seepage. For each case, initial prompting involved providing ChatGPT with +> necessary information for finite element implementation, such as balance and +> constitutive equations, problem geometry, initial and boundary conditions, +> material properties, and spatiotemporal discretization and solution strategies. +> Any errors and unexpected results were further addressed through prompt +> augmentation processes until the ChatGPT-generated finite element code passed +> the verification/validation test. Our results demonstrate that ChatGPT required +> minimal code revisions when using the FEniCS finite element library, owing to +> its high-level interfaces that enable efficient programming. In contrast, the +> MATLAB code generated by ChatGPT necessitated extensive prompt augmentations +> and/or direct human intervention, as it involves a significant amount of +> low-level programming required for finite element analysis, such as +> constructing shape functions or assembling global matrices. Given that prompt +> engineering for this task requires an understanding of the mathematical +> formulation and numerical techniques, this study suggests that while a large +> language model may not yet replace human programmers, it can greatly assist in +> the implementation of numerical models. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "While the paper does investigate the manipulation of textual input prompts for LLMs (ChatGPT), its primary focus is on the application of LLMs in geotechnical engineering (finite element models) rather than the engineering, design, or optimization of prompts specifically for Large Language Models." +} + +--- + +## [Learning Evolution via Optimization Knowledge Adaptation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02200v1) +**arXiv ID:** http://arxiv.org/abs/2501.02200v1 + +**Abstract:** +> Evolutionary algorithms (EAs) maintain populations through evolutionary +> operators to discover diverse solutions for complex tasks while gathering +> valuable knowledge, such as historical population data and fitness evaluations. +> However, traditional EAs face challenges in dynamically adapting to expanding +> knowledge bases, hindering the efficient exploitation of accumulated +> information and limiting adaptability to new situations. To address these +> issues, we introduce an Optimization Knowledge Adaptation Evolutionary Model +> (OKAEM), which features dynamic parameter adjustment using accumulated +> knowledge to enhance its optimization capabilities. OKAEM employs attention +> mechanisms to model the interactions among individuals, fitness landscapes, and +> genetic components separately, thereby parameterizing the evolutionary +> operators of selection, crossover, and mutation. These powerful learnable +> operators enable OKAEM to benefit from pre-learned extensive prior knowledge +> and self-tune with real-time evolutionary insights. Experimental results +> demonstrate that OKAEM: 1) exploits prior knowledge for significant performance +> gains across various knowledge transfer settings; 2) achieves competitive +> performance through self-tuning alone, even without prior knowledge; 3) +> outperforms state-of-the-art black-box baselines in a vision-language model +> tuning case; 4) can improve its optimization capabilities with growing +> knowledge; 5) is capable of emulating principles of natural selection and +> genetic recombination. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on evolutionary algorithms and optimization knowledge adaptation for improving performance in complex tasks, with a specific application in vision-language model tuning. It does not meet the 'MUST' criteria, particularly focusing primarily on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or investigating methods to improve LLM performance through prompt manipulation. + +--- + +## [Diffusion Model-Based Data Synthesis Aided Federated Semi-Supervised + Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.02219v1) +**arXiv ID:** http://arxiv.org/abs/2501.02219v1 + +**Abstract:** +> Federated semi-supervised learning (FSSL) is primarily challenged by two +> factors: the scarcity of labeled data across clients and the non-independent +> and identically distribution (non-IID) nature of data among clients. In this +> paper, we propose a novel approach, diffusion model-based data synthesis aided +> FSSL (DDSA-FSSL), which utilizes a diffusion model (DM) to generate synthetic +> data, bridging the gap between heterogeneous local data distributions and the +> global data distribution. In DDSA-FSSL, clients address the challenge of the +> scarcity of labeled data by employing a federated learning-trained classifier +> to perform pseudo labeling for unlabeled data. The DM is then collaboratively +> trained using both labeled and precision-optimized pseudo-labeled data, +> enabling clients to generate synthetic samples for classes that are absent in +> their labeled datasets. This process allows clients to generate more +> comprehensive synthetic datasets aligned with the global distribution. +> Extensive experiments conducted on multiple datasets and varying non-IID +> distributions demonstrate the effectiveness of DDSA-FSSL, e.g., it improves +> accuracy from 38.46% to 52.14% on CIFAR-10 datasets with 10% labeled data. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on federated semi-supervised learning and utilizes a diffusion model for data synthesis, with no primary emphasis on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [CORD: Generalizable Cooperation via Role Diversity](https://arxiv.org/abs/http://arxiv.org/abs/2501.02221v2) +**arXiv ID:** http://arxiv.org/abs/2501.02221v2 + +**Abstract:** +> Cooperative multi-agent reinforcement learning (MARL) aims to develop agents +> that can collaborate effectively. However, most cooperative MARL methods +> overfit training agents, making learned policies not generalize well to unseen +> collaborators, which is a critical issue for real-world deployment. Some +> methods attempt to address the generalization problem but require prior +> knowledge or predefined policies of new teammates, limiting real-world +> applications. To this end, we propose a hierarchical MARL approach to enable +> generalizable cooperation via role diversity, namely CORD. CORD's high-level +> controller assigns roles to low-level agents by maximizing the role entropy +> with constraints. We show this constrained objective can be decomposed into +> causal influence in role that enables reasonable role assignment, and role +> heterogeneity that yields coherent, non-redundant role clusters. Evaluated on a +> variety of cooperative multi-agent tasks, CORD achieves better performance than +> baselines, especially in generalization tests. Ablation studies further +> demonstrate the efficacy of the constrained objective in generalizable +> cooperation. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on cooperative multi-agent reinforcement learning (MARL) and a hierarchical approach for generalizable cooperation, without any mention of Large Language Models (LLMs), prompt engineering, or textual input manipulation, thus failing to meet all the 'MUST' criteria. + +--- + +## [Towards a constructive framework for control theory](https://arxiv.org/abs/http://arxiv.org/abs/2501.02267v1) +**arXiv ID:** http://arxiv.org/abs/2501.02267v1 + +**Abstract:** +> This work presents a framework for control theory based on constructive +> analysis to account for discrepancy between mathematical results and their +> implementation in a computer, also referred to as computational uncertainty. In +> control engineering, the latter is usually either neglected or considered +> submerged into some other type of uncertainty, such as system noise, and +> addressed within robust control. However, even robust control methods may be +> compromised when the mathematical objects involved in the respective algorithms +> fail to exist in exact form and subsequently fail to satisfy the required +> properties. For instance, in general stabilization using a control Lyapunov +> function, computational uncertainty may distort stability certificates or even +> destabilize the system despite robustness of the stabilization routine with +> regards to system, actuator and measurement noise. In fact, battling numerical +> problems in practical implementation of controllers is common among control +> engineers. Such observations indicate that computational uncertainty should +> indeed be addressed explicitly in controller synthesis and system analysis. The +> major contribution here is a fairly general framework for proof techniques in +> analysis and synthesis of control systems based on constructive analysis which +> explicitly states that every computation be doable only up to a finite +> precision thus accounting for computational uncertainty. A series of previous +> works is overviewed, including constructive system stability and stabilization, +> approximate optimal controls, eigenvalue problems, Caratheodory trajectories, +> measurable selectors. Additionally, a new constructive version of the Danskin's +> theorem, which is crucial in adversarial defense, is presented. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary criteria as it focuses on control theory and computational uncertainty in control system engineering, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions with LLMs. + +--- + +## [Deep Learning-Driven Segmentation of Ischemic Stroke Lesions Using + Multi-Channel MRI](https://arxiv.org/abs/http://arxiv.org/abs/2501.02287v1) +**arXiv ID:** http://arxiv.org/abs/2501.02287v1 + +**Abstract:** +> Ischemic stroke, caused by cerebral vessel occlusion, presents substantial +> challenges in medical imaging due to the variability and subtlety of stroke +> lesions. Magnetic Resonance Imaging (MRI) plays a crucial role in diagnosing +> and managing ischemic stroke, yet existing segmentation techniques often fail +> to accurately delineate lesions. This study introduces a novel deep +> learning-based method for segmenting ischemic stroke lesions using +> multi-channel MRI modalities, including Diffusion Weighted Imaging (DWI), +> Apparent Diffusion Coefficient (ADC), and enhanced Diffusion Weighted Imaging +> (eDWI). The proposed architecture integrates DenseNet121 as the encoder with +> Self-Organized Operational Neural Networks (SelfONN) in the decoder, enhanced +> by Channel and Space Compound Attention (CSCA) and Double +> Squeeze-and-Excitation (DSE) blocks. Additionally, a custom loss function +> combining Dice Loss and Jaccard Loss with weighted averages is introduced to +> improve model performance. Trained and evaluated on the ISLES 2022 dataset, the +> model achieved Dice Similarity Coefficients (DSC) of 83.88% using DWI alone, +> 85.86% with DWI and ADC, and 87.49% with the integration of DWI, ADC, and eDWI. +> This approach not only outperforms existing methods but also addresses key +> limitations in current segmentation practices. These advancements significantly +> enhance diagnostic precision and treatment planning for ischemic stroke, +> providing valuable support for clinical decision-making. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a deep learning-based method for segmenting ischemic stroke lesions using MRI modalities, which is a medical application unrelated to Large Language Models (LLMs) or prompt engineering, thus failing to meet all 'MUST' criteria. + +--- + +## [DiffGraph: Heterogeneous Graph Diffusion Model](https://arxiv.org/abs/http://arxiv.org/abs/2501.02313v1) +**arXiv ID:** http://arxiv.org/abs/2501.02313v1 + +**Abstract:** +> Recent advances in Graph Neural Networks (GNNs) have revolutionized +> graph-structured data modeling, yet traditional GNNs struggle with complex +> heterogeneous structures prevalent in real-world scenarios. Despite progress in +> handling heterogeneous interactions, two fundamental challenges persist: noisy +> data significantly compromising embedding quality and learning performance, and +> existing methods' inability to capture intricate semantic transitions among +> heterogeneous relations, which impacts downstream predictions. To address these +> fundamental issues, we present the Heterogeneous Graph Diffusion Model +> (DiffGraph), a pioneering framework that introduces an innovative cross-view +> denoising strategy. This advanced approach transforms auxiliary heterogeneous +> data into target semantic spaces, enabling precise distillation of +> task-relevant information. At its core, DiffGraph features a sophisticated +> latent heterogeneous graph diffusion mechanism, implementing a novel forward +> and backward diffusion process for superior noise management. This methodology +> achieves simultaneous heterogeneous graph denoising and cross-type transition, +> while significantly simplifying graph generation through its latent-space +> diffusion capabilities. Through rigorous experimental validation on both public +> and industrial datasets, we demonstrate that DiffGraph consistently surpasses +> existing methods in link prediction and node classification tasks, establishing +> new benchmarks for robustness and efficiency in heterogeneous graph processing. +> The model implementation is publicly available at: +> https://github.com/HKUDS/DiffGraph. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a new graph neural network model (DiffGraph) for handling heterogeneous graph-structured data, primarily addressing issues of noise and semantic transitions in graph processing, with no clear emphasis on prompt engineering for Large Language Models (LLMs) or manipulation of textual input prompts to improve LLM performance. + +--- + +## [Validity Arguments For Constructed Response Scoring Using Generative + Artificial Intelligence Applications](https://arxiv.org/abs/http://arxiv.org/abs/2501.02334v1) +**arXiv ID:** http://arxiv.org/abs/2501.02334v1 + +**Abstract:** +> The rapid advancements in large language models and generative artificial +> intelligence (AI) capabilities are making their broad application in the +> high-stakes testing context more likely. Use of generative AI in the scoring of +> constructed responses is particularly appealing because it reduces the effort +> required for handcrafting features in traditional AI scoring and might even +> outperform those methods. The purpose of this paper is to highlight the +> differences in the feature-based and generative AI applications in constructed +> response scoring systems and propose a set of best practices for the collection +> of validity evidence to support the use and interpretation of constructed +> response scores from scoring systems using generative AI. We compare the +> validity evidence needed in scoring systems using human ratings, feature-based +> natural language processing AI scoring engines, and generative AI. The evidence +> needed in the generative AI context is more extensive than in the feature-based +> NLP scoring context because of the lack of transparency and other concerns +> unique to generative AI such as consistency. Constructed response score data +> from standardized tests demonstrate the collection of validity evidence for +> different types of scoring systems and highlights the numerous complexities and +> considerations when making a validity argument for these scores. In addition, +> we discuss how the evaluation of AI scores might include a consideration of how +> a contributory scoring approach combining multiple AI scores (from different +> sources) will cover more of the construct in the absence of human ratings. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on validity arguments for scoring systems using generative AI in high-stakes testing, rather than specifically on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Prompt engineering is not the central concern, and the paper's main subject is the application and validation of generative AI in a testing context, excluding it based on the 'MUST NOT' criteria. + +--- + +## [GNSS/GPS Spoofing and Jamming Identification Using Machine Learning and + Deep Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.02352v1) +**arXiv ID:** http://arxiv.org/abs/2501.02352v1 + +**Abstract:** +> The increasing reliance on Global Navigation Satellite Systems (GNSS), +> particularly the Global Positioning System (GPS), underscores the urgent need +> to safeguard these technologies against malicious threats such as spoofing and +> jamming. As the backbone for positioning, navigation, and timing (PNT) across +> various applications including transportation, telecommunications, and +> emergency services GNSS is vulnerable to deliberate interference that poses +> significant risks. Spoofing attacks, which involve transmitting counterfeit +> GNSS signals to mislead receivers into calculating incorrect positions, can +> result in serious consequences, from navigational errors in civilian aviation +> to security breaches in military operations. Furthermore, the lack of inherent +> security measures within GNSS systems makes them attractive targets for +> adversaries. While GNSS/GPS jamming and spoofing systems consist of numerous +> components, the ability to distinguish authentic signals from malicious ones is +> essential for maintaining system integrity. Recent advancements in machine +> learning and deep learning provide promising avenues for enhancing detection +> and mitigation strategies against these threats. This paper addresses both +> spoofing and jamming by tackling real-world challenges through machine +> learning, deep learning, and computer vision techniques. Through extensive +> experiments on two real-world datasets related to spoofing and jamming +> detection using advanced algorithms, we achieved state of the art results. In +> the GNSS/GPS jamming detection task, we attained approximately 99% accuracy, +> improving performance by around 5% compared to previous studies. Additionally, +> we addressed a challenging tasks related to spoofing detection, yielding +> results that underscore the potential of machine learning and deep learning in +> this domain. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet any of the 'MUST' criteria, as it focuses primarily on using machine learning and deep learning for GNSS/GPS spoofing and jamming identification, with no mention of Large Language Models (LLMs), prompt engineering, or text generation. + +--- + +## [FaceSpeak: Expressive and High-Quality Speech Synthesis from Human + Portraits of Different Styles](https://arxiv.org/abs/http://arxiv.org/abs/2501.03181v1) +**arXiv ID:** http://arxiv.org/abs/2501.03181v1 + +**Abstract:** +> Humans can perceive speakers' characteristics (e.g., identity, gender, +> personality and emotion) by their appearance, which are generally aligned to +> their voice style. Recently, vision-driven Text-to-speech (TTS) scholars +> grounded their investigations on real-person faces, thereby restricting +> effective speech synthesis from applying to vast potential usage scenarios with +> diverse characters and image styles. To solve this issue, we introduce a novel +> FaceSpeak approach. It extracts salient identity characteristics and emotional +> representations from a wide variety of image styles. Meanwhile, it mitigates +> the extraneous information (e.g., background, clothing, and hair color, etc.), +> resulting in synthesized speech closely aligned with a character's persona. +> Furthermore, to overcome the scarcity of multi-modal TTS data, we have devised +> an innovative dataset, namely Expressive Multi-Modal TTS, which is diligently +> curated and annotated to facilitate research in this domain. The experimental +> results demonstrate our proposed FaceSpeak can generate portrait-aligned voice +> with satisfactory naturalness and quality. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on speech synthesis from human portraits using a novel FaceSpeak approach, which involves image processing and Text-to-Speech (TTS) for audio generation. It does not investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through the manipulation of textual input prompts, nor does it provide concrete examples of prompts and their impact on LLM output. + +--- + +## [Classifier-Guided Captioning Across Modalities](https://arxiv.org/abs/http://arxiv.org/abs/2501.03183v1) +**arXiv ID:** http://arxiv.org/abs/2501.03183v1 + +**Abstract:** +> Most current captioning systems use language models trained on data from +> specific settings, such as image-based captioning via Amazon Mechanical Turk, +> limiting their ability to generalize to other modality distributions and +> contexts. This limitation hinders performance in tasks like audio or video +> captioning, where different semantic cues are needed. Addressing this challenge +> is crucial for creating more adaptable and versatile captioning frameworks +> applicable across diverse real-world contexts. In this work, we introduce a +> method to adapt captioning networks to the semantics of alternative settings, +> such as capturing audibility in audio captioning, where it is crucial to +> describe sounds and their sources. Our framework consists of two main +> components: (i) a frozen captioning system incorporating a language model (LM), +> and (ii) a text classifier that guides the captioning system. The classifier is +> trained on a dataset automatically generated by GPT-4, using tailored prompts +> specifically designed to enhance key aspects of the generated captions. +> Importantly, the framework operates solely during inference, eliminating the +> need for further training of the underlying captioning model. We evaluate the +> framework on various models and modalities, with a focus on audio captioning, +> and report promising results. Notably, when combined with an existing zero-shot +> audio captioning system, our framework improves its quality and sets +> state-of-the-art performance in zero-shot audio captioning. + +**Decision Explanation:** +Original decision: REJECT +Although the paper mentions using tailored prompts with GPT-4, the primary focus is on adaptating captioning networks across modalities (image, audio, video) and improving the captioning system with a classifier, rather than prompt engineering for Large Language Models (LLMs) being the core subject. + +--- + +## [Breaking Through the Spike: Spike Window Decoding for Accelerated and + Precise Automatic Speech Recognition](https://arxiv.org/abs/http://arxiv.org/abs/2501.03257v1) +**arXiv ID:** http://arxiv.org/abs/2501.03257v1 + +**Abstract:** +> Recently, end-to-end automatic speech recognition has become the mainstream +> approach in both industry and academia. To optimize system performance in +> specific scenarios, the Weighted Finite-State Transducer (WFST) is extensively +> used to integrate acoustic and language models, leveraging its capacity to +> implicitly fuse language models within static graphs, thereby ensuring robust +> recognition while also facilitating rapid error correction. However, WFST +> necessitates a frame-by-frame search of CTC posterior probabilities through +> autoregression, which significantly hampers inference speed. In this work, we +> thoroughly investigate the spike property of CTC outputs and further propose +> the conjecture that adjacent frames to non-blank spikes carry semantic +> information beneficial to the model. Building on this, we propose the Spike +> Window Decoding algorithm, which greatly improves the inference speed by making +> the number of frames decoded in WFST linearly related to the number of spiking +> frames in the CTC output, while guaranteeing the recognition performance. Our +> method achieves SOTA recognition accuracy with significantly accelerates +> decoding speed, proven across both AISHELL-1 and large-scale In-House datasets, +> establishing a pioneering approach for integrating CTC output with WFST. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on optimizing Automatic Speech Recognition (ASR) using Spike Window Decoding and Weighted Finite-State Transducers, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions, thus failing to meet all 'MUST' criteria. + +--- + +## [Navigation Variable-based Multi-objective Particle Swarm Optimization + for UAV Path Planning with Kinematic Constraints](https://arxiv.org/abs/http://arxiv.org/abs/2501.03261v1) +**arXiv ID:** http://arxiv.org/abs/2501.03261v1 + +**Abstract:** +> Path planning is essential for unmanned aerial vehicles (UAVs) as it +> determines the path that the UAV needs to follow to complete a task. This work +> addresses this problem by introducing a new algorithm called navigation +> variable-based multi-objective particle swarm optimization (NMOPSO). It first +> models path planning as an optimization problem via the definition of a set of +> objective functions that include optimality and safety requirements for UAV +> operation. The NMOPSO is then used to minimize those functions through Pareto +> optimal solutions. The algorithm features a new path representation based on +> navigation variables to include kinematic constraints and exploit the +> maneuverable characteristics of the UAV. It also includes an adaptive mutation +> mechanism to enhance the diversity of the swarm for better solutions. +> Comparisons with various algorithms have been carried out to benchmark the +> proposed approach. The results indicate that the NMOPSO performs better than +> not only other particle swarm optimization variants but also other +> state-of-the-art multi-objective and metaheuristic optimization algorithms. +> Experiments have also been conducted with real UAVs to confirm the validity of +> the approach for practical flights. The source code of the algorithm is +> available at https://github.com/ngandng/NMOPSO. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a new optimization algorithm for UAV path planning, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria. + +--- + +## [Bridge the Inference Gaps of Neural Processes via Expectation + Maximization](https://arxiv.org/abs/http://arxiv.org/abs/2501.03264v1) +**arXiv ID:** http://arxiv.org/abs/2501.03264v1 + +**Abstract:** +> The neural process (NP) is a family of computationally efficient models for +> learning distributions over functions. However, it suffers from under-fitting +> and shows suboptimal performance in practice. Researchers have primarily +> focused on incorporating diverse structural inductive biases, \textit{e.g.} +> attention or convolution, in modeling. The topic of inference suboptimality and +> an analysis of the NP from the optimization objective perspective has hardly +> been studied in earlier work. To fix this issue, we propose a surrogate +> objective of the target log-likelihood of the meta dataset within the +> expectation maximization framework. The resulting model, referred to as the +> Self-normalized Importance weighted Neural Process (SI-NP), can learn a more +> accurate functional prior and has an improvement guarantee concerning the +> target log-likelihood. Experimental results show the competitive performance of +> SI-NP over other NPs objectives and illustrate that structural inductive +> biases, such as attention modules, can also augment our method to achieve SOTA +> performance. Our code is available at +> \url{https://github.com/hhq123gogogo/SI_NPs}. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on improving Neural Processes (NP) via Expectation Maximization, primarily addressing under-fitting and optimization objectives, with no clear connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Listening and Seeing Again: Generative Error Correction for Audio-Visual + Speech Recognition](https://arxiv.org/abs/http://arxiv.org/abs/2501.04038v1) +**arXiv ID:** http://arxiv.org/abs/2501.04038v1 + +**Abstract:** +> Unlike traditional Automatic Speech Recognition (ASR), Audio-Visual Speech +> Recognition (AVSR) takes audio and visual signals simultaneously to infer the +> transcription. Recent studies have shown that Large Language Models (LLMs) can +> be effectively used for Generative Error Correction (GER) in ASR by predicting +> the best transcription from ASR-generated N-best hypotheses. However, these +> LLMs lack the ability to simultaneously understand audio and visual, making the +> GER approach challenging to apply in AVSR. In this work, we propose a novel GER +> paradigm for AVSR, termed AVGER, that follows the concept of ``listening and +> seeing again''. Specifically, we first use the powerful AVSR system to read the +> audio and visual signals to get the N-Best hypotheses, and then use the +> Q-former-based Multimodal Synchronous Encoder to read the audio and visual +> information again and convert them into an audio and video compression +> representation respectively that can be understood by LLM. Afterward, the +> audio-visual compression representation and the N-Best hypothesis together +> constitute a Cross-modal Prompt to guide the LLM in producing the best +> transcription. In addition, we also proposed a Multi-Level Consistency +> Constraint training criterion, including logits-level, utterance-level and +> representations-level, to improve the correction accuracy while enhancing the +> interpretability of audio and visual compression representations. The +> experimental results on the LRS3 dataset show that our method outperforms +> current mainstream AVSR systems. The proposed AVGER can reduce the Word Error +> Rate (WER) by 24% compared to them. Code and models can be found at: +> https://github.com/CircleRedRain/AVGER. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on improving Audio-Visual Speech Recognition (AVSR) with Generative Error Correction using LLMs, rather than specifically on the engineering, design, or optimization of prompts for LLMs. While prompts are mentioned (Cross-modal Prompt), they are not the central focus, and the work is more about applying LLMs within a larger AVSR system. + +--- + +## [A Survey on Large Language Models with some Insights on their + Capabilities and Limitations](https://arxiv.org/abs/http://arxiv.org/abs/2501.04040v1) +**arXiv ID:** http://arxiv.org/abs/2501.04040v1 + +**Abstract:** +> The rapid advancement of artificial intelligence, particularly with the +> development of Large Language Models (LLMs) built on the transformer +> architecture, has redefined the capabilities of natural language processing. +> These models now exhibit remarkable performance across various language-related +> tasks, such as text generation, question answering, translation, and +> summarization, often rivaling human-like comprehension. More intriguingly, LLMs +> have demonstrated emergent abilities extending beyond their core functions, +> showing proficiency in tasks like commonsense reasoning, code generation, and +> arithmetic. This survey paper explores the foundational components, scaling +> mechanisms, and architectural strategies that drive these capabilities. +> Emphasizing models like GPT and LLaMA, we analyze the impact of exponential +> data and computational growth on LLM performance, while also addressing the +> trade-offs associated with scaling. We also examine LLM applications across +> sectors, such as healthcare, finance, education, and law, highlighting their +> adaptability and potential to solve domain-specific challenges. Central to this +> work are the questions of how LLMs generalize across diverse tasks, exhibit +> planning, and reasoning abilities, and whether these emergent abilities can be +> systematically elicited or enhanced. In particular, we provide some insights +> into the CoT (Chain of Thought) and PoT (Plan of Thought) abilities within +> LLMs, focusing on how pre-training data influences their emergence. +> Additionally, we investigate LLM-modulo frameworks that integrate external +> systems, allowing LLMs to handle complex, dynamic tasks. By analyzing these +> factors, this paper aims to foster the ongoing discussion on the capabilities +> and limits of LLMs, promoting their responsible development and application in +> novel and increasingly complex environments. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on surveying the capabilities and limitations of Large Language Models (LLMs), with only tangential insights into prompt-related aspects (e.g., Chain of Thought abilities), rather than centrally investigating prompt engineering for text-based interactions with LLMs as required. + +--- + +## [FLAME: Financial Large-Language Model Assessment and Metrics Evaluation](https://arxiv.org/abs/http://arxiv.org/abs/2501.06211v1) +**arXiv ID:** http://arxiv.org/abs/2501.06211v1 + +**Abstract:** +> LLMs have revolutionized NLP and demonstrated potential across diverse +> domains. More and more financial LLMs have been introduced for finance-specific +> tasks, yet comprehensively assessing their value is still challenging. In this +> paper, we introduce FLAME, a comprehensive financial LLMs evaluation system in +> Chinese, which includes two core evaluation benchmarks: FLAME-Cer and +> FLAME-Sce. FLAME-Cer covers 14 types of authoritative financial certifications, +> including CPA, CFA, and FRM, with a total of approximately 16,000 carefully +> selected questions. All questions have been manually reviewed to ensure +> accuracy and representativeness. FLAME-Sce consists of 10 primary core +> financial business scenarios, 21 secondary financial business scenarios, and a +> comprehensive evaluation set of nearly 100 tertiary financial application +> tasks. We evaluate 6 representative LLMs, including GPT-4o, GLM-4, ERNIE-4.0, +> Qwen2.5, XuanYuan3, and the latest Baichuan4-Finance, revealing +> Baichuan4-Finance excels other LLMs in most tasks. By establishing a +> comprehensive and professional evaluation system, FLAME facilitates the +> advancement of financial LLMs in Chinese contexts. Instructions for +> participating in the evaluation are available on GitHub: +> https://github.com/FLAME-ruc/FLAME. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on developing an evaluation system (FLAME) for assessing the performance of financial Large Language Models, rather than primarily investigating, analyzing, or proposing methods for improving LLM performance through the manipulation of textual input prompts." +} + +--- + +## [Operator Learning for Reconstructing Flow Fields from Sparse + Measurements: an Energy Transformer Approach](https://arxiv.org/abs/http://arxiv.org/abs/2501.08339v1) +**arXiv ID:** http://arxiv.org/abs/2501.08339v1 + +**Abstract:** +> Machine learning methods have shown great success in various scientific +> areas, including fluid mechanics. However, reconstruction problems, where full +> velocity fields must be recovered from partial observations, remain +> challenging. In this paper, we propose a novel operator learning framework for +> solving reconstruction problems by using the Energy Transformer (ET), an +> architecture inspired by associative memory models. We formulate reconstruction +> as a mapping from incomplete observed data to full reconstructed fields. The +> method is validated on three fluid mechanics examples using diverse types of +> data: (1) unsteady 2D vortex street in flow past a cylinder using simulation +> data; (2) high-speed under-expanded impinging supersonic jets impingement using +> Schlieren imaging; and (3) 3D turbulent jet flow using particle tracking. The +> results demonstrate the ability of ET to accurately reconstruct complex flow +> fields from highly incomplete data (90\% missing), even for noisy experimental +> measurements, with fast training and inference on a single GPU. This work +> provides a promising new direction for tackling reconstruction problems in +> fluid mechanics and other areas in mechanics, geophysics, weather prediction, +> and beyond. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on operator learning for reconstructing flow fields using a novel Energy Transformer approach, with no primary focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not investigate or propose methods for improving LLM performance through textual input prompt manipulation. + +--- + +## [VERITAS: Verifying the Performance of AI-native Transceiver Actions in + Base-Stations](https://arxiv.org/abs/http://arxiv.org/abs/2501.09761v1) +**arXiv ID:** http://arxiv.org/abs/2501.09761v1 + +**Abstract:** +> Artificial Intelligence (AI)-native receivers prove significant performance +> improvement in high noise regimes and can potentially reduce communication +> overhead compared to the traditional receiver. However, their performance +> highly depends on the representativeness of the training dataset. A major issue +> is the uncertainty of whether the training dataset covers all test environments +> and waveform configurations, and thus, whether the trained model is robust in +> practical deployment conditions. To this end, we propose a joint +> measurement-recovery framework for AI-native transceivers post deployment, +> called VERITAS, that continuously looks for distribution shifts in the received +> signals and triggers finite re-training spurts. VERITAS monitors the wireless +> channel using 5G pilots fed to an auxiliary neural network that detects +> out-of-distribution channel profile, transmitter speed, and delay spread. As +> soon as such a change is detected, a traditional (reference) receiver is +> activated, which runs for a period of time in parallel to the AI-native +> receiver. Finally, VERTIAS compares the bit probabilities of the AI-native and +> the reference receivers for the same received data inputs, and decides whether +> or not a retraining process needs to be initiated. Our evaluations reveal that +> VERITAS can detect changes in the channel profile, transmitter speed, and delay +> spread with 99%, 97%, and 69% accuracies, respectively, followed by timely +> initiation of retraining for 86%, 93.3%, and 94.8% of inputs in channel +> profile, transmitter speed, and delay spread test sets, respectively. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on verifying the performance of AI-native transceiver actions in base-stations, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Dynamics of Adversarial Attacks on Large Language Model-Based Search + Engines](https://arxiv.org/abs/http://arxiv.org/abs/2501.00745v1) +**arXiv ID:** http://arxiv.org/abs/2501.00745v1 + +**Abstract:** +> The increasing integration of Large Language Model (LLM) based search engines +> has transformed the landscape of information retrieval. However, these systems +> are vulnerable to adversarial attacks, especially ranking manipulation attacks, +> where attackers craft webpage content to manipulate the LLM's ranking and +> promote specific content, gaining an unfair advantage over competitors. In this +> paper, we study the dynamics of ranking manipulation attacks. We frame this +> problem as an Infinitely Repeated Prisoners' Dilemma, where multiple players +> strategically decide whether to cooperate or attack. We analyze the conditions +> under which cooperation can be sustained, identifying key factors such as +> attack costs, discount rates, attack success rates, and trigger strategies that +> influence player behavior. We identify tipping points in the system dynamics, +> demonstrating that cooperation is more likely to be sustained when players are +> forward-looking. However, from a defense perspective, we find that simply +> reducing attack success probabilities can, paradoxically, incentivize attacks +> under certain conditions. Furthermore, defensive measures to cap the upper +> bound of attack success rates may prove futile in some scenarios. These +> insights highlight the complexity of securing LLM-based systems. Our work +> provides a theoretical foundation and practical insights for understanding and +> mitigating their vulnerabilities, while emphasizing the importance of adaptive +> security strategies and thoughtful ecosystem design. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the vulnerability and security of LLM-based search engines to adversarial attacks, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), failing to meet the core subject requirement. + +--- + +## [MuQ: Self-Supervised Music Representation Learning with Mel Residual + Vector Quantization](https://arxiv.org/abs/http://arxiv.org/abs/2501.01108v2) +**arXiv ID:** http://arxiv.org/abs/2501.01108v2 + +**Abstract:** +> Recent years have witnessed the success of foundation models pre-trained with +> self-supervised learning (SSL) in various music informatics understanding +> tasks, including music tagging, instrument classification, key detection, and +> more. In this paper, we propose a self-supervised music representation learning +> model for music understanding. Distinguished from previous studies adopting +> random projection or existing neural codec, the proposed model, named MuQ, is +> trained to predict tokens generated by Mel Residual Vector Quantization +> (Mel-RVQ). Our Mel-RVQ utilizes residual linear projection structure for Mel +> spectrum quantization to enhance the stability and efficiency of target +> extraction and lead to better performance. Experiments in a large variety of +> downstream tasks demonstrate that MuQ outperforms previous self-supervised +> music representation models with only 0.9K hours of open-source pre-training +> data. Scaling up the data to over 160K hours and adopting iterative training +> consistently improve the model performance. To further validate the strength of +> our model, we present MuQ-MuLan, a joint music-text embedding model based on +> contrastive learning, which achieves state-of-the-art performance in the +> zero-shot music tagging task on the MagnaTagATune dataset. Code and checkpoints +> are open source in https://github.com/tencent-ailab/MuQ. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on self-supervised music representation learning and a new model (MuQ) for music understanding, with no apparent emphasis on engineering, design, or optimization of prompts for Large Language Models (LLMs) or demonstration of prompt impact on LLM output. + +--- + +## [Symmetries-enhanced Multi-Agent Reinforcement Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.01136v1) +**arXiv ID:** http://arxiv.org/abs/2501.01136v1 + +**Abstract:** +> Multi-agent reinforcement learning has emerged as a powerful framework for +> enabling agents to learn complex, coordinated behaviors but faces persistent +> challenges regarding its generalization, scalability and sample efficiency. +> Recent advancements have sought to alleviate those issues by embedding +> intrinsic symmetries of the systems in the policy. Yet, most dynamical systems +> exhibit little to no symmetries to exploit. This paper presents a novel +> framework for embedding extrinsic symmetries in multi-agent system dynamics +> that enables the use of symmetry-enhanced methods to address systems with +> insufficient intrinsic symmetries, expanding the scope of equivariant learning +> to a wide variety of MARL problems. Central to our framework is the Group +> Equivariant Graphormer, a group-modular architecture specifically designed for +> distributed swarming tasks. Extensive experiments on a swarm of +> symmetry-breaking quadrotors validate the effectiveness of our approach, +> showcasing its potential for improved generalization and zero-shot scalability. +> Our method achieves significant reductions in collision rates and enhances task +> success rates across a diverse range of scenarios and varying swarm sizes. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on multi-agent reinforcement learning, introducing a novel framework for embedding extrinsic symmetries, with no apparent focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs) or their direct application in text generation. + +--- + +## [Change Detection-Based Procedures for Piecewise Stationary MABs: A + Modular Approach](https://arxiv.org/abs/http://arxiv.org/abs/2501.01291v1) +**arXiv ID:** http://arxiv.org/abs/2501.01291v1 + +**Abstract:** +> Conventional Multi-Armed Bandit (MAB) algorithms are designed for stationary +> environments, where the reward distributions associated with the arms do not +> change with time. In many applications, however, the environment is more +> accurately modeled as being nonstationary. In this work, piecewise stationary +> MAB (PS-MAB) environments are investigated, in which the reward distributions +> associated with a subset of the arms change at some change-points and remain +> stationary between change-points. Our focus is on the asymptotic analysis of +> PS-MABs, for which practical algorithms based on change detection (CD) have +> been previously proposed. Our goal is to modularize the design and analysis of +> such CD-based Bandit (CDB) procedures. To this end, we identify the +> requirements for stationary bandit algorithms and change detectors in a CDB +> procedure that are needed for the modularization. We assume that the rewards +> are sub-Gaussian. Under this assumption and a condition on the separation of +> the change-points, we show that the analysis of CDB procedures can indeed be +> modularized, so that regret bounds can be obtained in a unified manner for +> various combinations of change detectors and bandit algorithms. Through this +> analysis, we develop new modular CDB procedures that are order-optimal. We +> compare the performance of our modular CDB procedures with various other +> methods in simulations. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on Multi-Armed Bandit (MAB) algorithms and change detection procedures, with no apparent connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria." +} + +--- + +## [Quantifying A Firm's AI Engagement: Constructing Objective, Data-Driven, + AI Stock Indices Using 10-K Filings](https://arxiv.org/abs/http://arxiv.org/abs/2501.01763v1) +**arXiv ID:** http://arxiv.org/abs/2501.01763v1 + +**Abstract:** +> Following an analysis of existing AI-related exchange-traded funds (ETFs), we +> reveal the selection criteria for determining which stocks qualify as +> AI-related are often opaque and rely on vague phrases and subjective judgments. +> This paper proposes a new, objective, data-driven approach using natural +> language processing (NLP) techniques to classify AI stocks by analyzing annual +> 10-K filings from 3,395 NASDAQ-listed firms between 2011 and 2023. This +> analysis quantifies each company's engagement with AI through binary indicators +> and weighted AI scores based on the frequency and context of AI-related terms. +> Using these metrics, we construct four AI stock indices-the Equally Weighted AI +> Index (AII), the Size-Weighted AI Index (SAII), and two Time-Discounted AI +> Indices (TAII05 and TAII5X)-offering different perspectives on AI investment. +> We validate our methodology through an event study on the launch of OpenAI's +> ChatGPT, demonstrating that companies with higher AI engagement saw +> significantly greater positive abnormal returns, with analyses supporting the +> predictive power of our AI measures. Our indices perform on par with or surpass +> 14 existing AI-themed ETFs and the Nasdaq Composite Index in risk-return +> profiles, market responsiveness, and overall performance, achieving higher +> average daily returns and risk-adjusted metrics without increased volatility. +> These results suggest our NLP-based approach offers a reliable, +> market-responsive, and cost-effective alternative to existing AI-related ETF +> products. Our innovative methodology can also guide investors, asset managers, +> and policymakers in using corporate data to construct other thematic +> portfolios, contributing to a more transparent, data-driven, and competitive +> approach. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is on constructing AI stock indices using NLP techniques to analyze corporate filings, not on prompt engineering for Large Language Models (LLMs). It does not investigate, analyze, or propose methods for improving LLM performance through prompt manipulation, nor does it provide concrete examples of prompts and their impact on LLM output." +} + +--- + +## [KANS: Knowledge Discovery Graph Attention Network for Soft Sensing in + Multivariate Industrial Processes](https://arxiv.org/abs/http://arxiv.org/abs/2501.02015v1) +**arXiv ID:** http://arxiv.org/abs/2501.02015v1 + +**Abstract:** +> Soft sensing of hard-to-measure variables is often crucial in industrial +> processes. Current practices rely heavily on conventional modeling techniques +> that show success in improving accuracy. However, they overlook the non-linear +> nature, dynamics characteristics, and non-Euclidean dependencies between +> complex process variables. To tackle these challenges, we present a framework +> known as a Knowledge discovery graph Attention Network for effective Soft +> sensing (KANS). Unlike the existing deep learning soft sensor models, KANS can +> discover the intrinsic correlations and irregular relationships between the +> multivariate industrial processes without a predefined topology. First, an +> unsupervised graph structure learning method is introduced, incorporating the +> cosine similarity between different sensor embedding to capture the +> correlations between sensors. Next, we present a graph attention-based +> representation learning that can compute the multivariate data parallelly to +> enhance the model in learning complex sensor nodes and edges. To fully explore +> KANS, knowledge discovery analysis has also been conducted to demonstrate the +> interpretability of the model. Experimental results demonstrate that KANS +> significantly outperforms all the baselines and state-of-the-art methods in +> soft sensing performance. Furthermore, the analysis shows that KANS can find +> sensors closely related to different process variables without domain +> knowledge, significantly improving soft sensing accuracy. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing a Knowledge Discovery Graph Attention Network (KANS) for soft sensing in industrial processes, not on prompt engineering for Large Language Models (LLMs). It does not investigate, analyze, or propose methods for improving LLM performance through textual input prompt manipulation, nor does it provide examples of prompts and their impact on LLM output. + +--- + +## [Benchmark Evaluations, Applications, and Challenges of Large Vision + Language Models: A Survey](https://arxiv.org/abs/http://arxiv.org/abs/2501.02189v2) +**arXiv ID:** http://arxiv.org/abs/2501.02189v2 + +**Abstract:** +> Multimodal Vision Language Models (VLMs) have emerged as a transformative +> technology at the intersection of computer vision and natural language +> processing, enabling machines to perceive and reason about the world through +> both visual and textual modalities. For example, models such as CLIP, Claude, +> and GPT-4V demonstrate strong reasoning and understanding abilities on visual +> and textual data and beat classical single modality vision models on zero-shot +> classification. Despite their rapid advancements in research and growing +> popularity in applications, a comprehensive survey of existing studies on VLMs +> is notably lacking, particularly for researchers aiming to leverage VLMs in +> their specific domains. To this end, we provide a systematic overview of VLMs +> in the following aspects: model information of the major VLMs developed over +> the past five years (2019-2024); the main architectures and training methods of +> these VLMs; summary and categorization of the popular benchmarks and evaluation +> metrics of VLMs; the applications of VLMs including embodied agents, robotics, +> and video generation; the challenges and issues faced by current VLMs such as +> hallucination, fairness, and safety. Detailed collections including papers and +> model repository links are listed in +> https://github.com/zli12321/Awesome-VLM-Papers-And-Models.git. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on Multimodal Vision Language Models (VLMs), their applications, and challenges, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) as required by the criteria. + +--- + +## [Optimizing Small Language Models for In-Vehicle Function-Calling](https://arxiv.org/abs/http://arxiv.org/abs/2501.02342v1) +**arXiv ID:** http://arxiv.org/abs/2501.02342v1 + +**Abstract:** +> We propose a holistic approach for deploying Small Language Models (SLMs) as +> function-calling agents within vehicles as edge devices, offering a more +> flexible and robust alternative to traditional rule-based systems. By +> leveraging SLMs, we simplify vehicle control mechanisms and enhance the user +> experience. Given the in-vehicle hardware constraints, we apply +> state-of-the-art model compression techniques, including structured pruning, +> healing, and quantization, ensuring that the model fits within the resource +> limitations while maintaining acceptable performance. Our work focuses on +> optimizing a representative SLM, Microsoft's Phi-3 mini, and outlines best +> practices for enabling embedded models, including compression, task-specific +> fine-tuning, and vehicle integration. We demonstrate that, despite significant +> reduction in model size which removes up to 2 billion parameters from the +> original model, our approach preserves the model's ability to handle complex +> in-vehicle tasks accurately and efficiently. Furthermore, by executing the +> model in a lightweight runtime environment, we achieve a generation speed of 11 +> tokens per second, making real-time, on-device inference feasible without +> hardware acceleration. Our results demonstrate the potential of SLMs to +> transform vehicle control systems, enabling more intuitive interactions between +> users and their vehicles for an enhanced driving experience. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on optimizing Small Language Models (SLMs) for deployment in vehicles, discussing model compression and fine-tuning, rather than prompt engineering for Large Language Models (LLMs) and its impact on textual input and output. + +--- + +## [Graph-Aware Isomorphic Attention for Adaptive Dynamics in Transformers](https://arxiv.org/abs/http://arxiv.org/abs/2501.02393v2) +**arXiv ID:** http://arxiv.org/abs/2501.02393v2 + +**Abstract:** +> We present an approach to modifying Transformer architectures by integrating +> graph-aware relational reasoning into the attention mechanism, merging concepts +> from graph neural networks and language modeling. Building on the inherent +> connection between attention and graph theory, we reformulate the Transformer's +> attention mechanism as a graph operation and propose Graph-Aware Isomorphic +> Attention. This method leverages advanced graph modeling strategies, including +> Graph Isomorphism Networks (GIN) and Principal Neighborhood Aggregation (PNA), +> to enrich the representation of relational structures. Our approach captures +> complex dependencies and generalizes across tasks, as evidenced by a reduced +> generalization gap and improved learning performance. Additionally, we expand +> the concept of graph-aware attention to introduce Sparse GIN-Attention, a +> fine-tuning approach that employs sparse GINs. By interpreting attention +> matrices as sparse adjacency graphs, this technique enhances the adaptability +> of pre-trained foundational models with minimal computational overhead, +> endowing them with graph-aware capabilities. Sparse GIN-Attention fine-tuning +> achieves improved training dynamics and better generalization compared to +> alternative methods like low-rank adaption (LoRA). We discuss latent graph-like +> structures within traditional attention mechanisms, offering a new lens through +> which Transformers can be understood. By evolving Transformers as hierarchical +> GIN models for relational reasoning. This perspective suggests profound +> implications for foundational model development, enabling the design of +> architectures that dynamically adapt to both local and global dependencies. +> Applications in bioinformatics, materials science, language modeling, and +> beyond could benefit from this synthesis of relational and sequential data +> modeling, setting the stage for interpretable and generalizable modeling +> strategies. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on modifying Transformer architectures by integrating graph-aware relational reasoning, which aligns with developing new LLM architectures, violating the 'MUST NOT' criterion 1. It does not centrally address prompt engineering for text-based interactions with LLMs. + +--- + +## [Toward Inclusive Educational AI: Auditing Frontier LLMs through a + Multiplexity Lens](https://arxiv.org/abs/http://arxiv.org/abs/2501.03259v1) +**arXiv ID:** http://arxiv.org/abs/2501.03259v1 + +**Abstract:** +> As large language models (LLMs) like GPT-4 and Llama 3 become integral to +> educational contexts, concerns are mounting over the cultural biases, power +> imbalances, and ethical limitations embedded within these technologies. Though +> generative AI tools aim to enhance learning experiences, they often reflect +> values rooted in Western, Educated, Industrialized, Rich, and Democratic +> (WEIRD) cultural paradigms, potentially sidelining diverse global perspectives. +> This paper proposes a framework to assess and mitigate cultural bias within +> LLMs through the lens of applied multiplexity. Multiplexity, inspired by +> Senturk et al. and rooted in Islamic and other wisdom traditions, emphasizes +> the coexistence of diverse cultural viewpoints, supporting a multi-layered +> epistemology that integrates both empirical sciences and normative values. Our +> analysis reveals that LLMs frequently exhibit cultural polarization, with +> biases appearing in both overt responses and subtle contextual cues. To address +> inherent biases and incorporate multiplexity in LLMs, we propose two +> strategies: \textit{Contextually-Implemented Multiplex LLMs}, which embed +> multiplex principles directly into the system prompt, influencing LLM outputs +> at a foundational level and independent of individual prompts, and +> \textit{Multi-Agent System (MAS)-Implemented Multiplex LLMs}, where multiple +> LLM agents, each representing distinct cultural viewpoints, collaboratively +> generate a balanced, synthesized response. Our findings demonstrate that as +> mitigation strategies evolve from contextual prompting to MAS-implementation, +> cultural inclusivity markedly improves, evidenced by a significant rise in the +> Perspectives Distribution Score (PDS) and a PDS Entropy increase from 3.25\% at +> baseline to 98\% with the MAS-Implemented Multiplex LLMs. Sentiment analysis +> further shows a shift towards positive sentiment across cultures,... + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on auditing and mitigating cultural biases in LLMs through a multiplexity lens, rather than specifically engineering or optimizing prompts for improving LLM performance. While it mentions prompt manipulation (Contextually-Implemented Multiplex LLMs), this is secondary to the overarching goal of bias mitigation and not the core subject of the paper." +} + +--- + +## [LLM Content Moderation and User Satisfaction: Evidence from Response + Refusals in Chatbot Arena](https://arxiv.org/abs/http://arxiv.org/abs/2501.03266v1) +**arXiv ID:** http://arxiv.org/abs/2501.03266v1 + +**Abstract:** +> LLM safety and ethical alignment are widely discussed, but the impact of +> content moderation on user satisfaction remains underexplored. To address this, +> we analyze nearly 50,000 Chatbot Arena response-pairs using a novel fine-tuned +> RoBERTa model, that we trained on hand-labeled data to disentangle refusals due +> to ethical concerns from other refusals due to technical disabilities or lack +> of information. Our findings reveal a significant refusal penalty on content +> moderation, with users choosing ethical-based refusals roughly one-fourth as +> often as their preferred LLM response compared to standard responses. However, +> the context and phrasing play critical roles: refusals on highly sensitive +> prompts, such as illegal content, achieve higher win rates than less sensitive +> ethical concerns, and longer responses closely aligned with the prompt perform +> better. These results emphasize the need for nuanced moderation strategies that +> balance ethical safeguards with user satisfaction. Moreover, we find that the +> refusal penalty is notably lower in evaluations using the LLM-as-a-Judge +> method, highlighting discrepancies between user and automated assessments. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on LLM content moderation, user satisfaction, and refining a RoBERTa model for analyzing refusals, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) and their direct impact on LLM output. + +--- + +## [Cyber Shadows: Neutralizing Security Threats with AI and Targeted Policy + Measures](https://arxiv.org/abs/http://arxiv.org/abs/2501.09025v1) +**arXiv ID:** http://arxiv.org/abs/2501.09025v1 + +**Abstract:** +> The digital age, driven by the AI revolution, brings significant +> opportunities but also conceals security threats, which we refer to as cyber +> shadows. These threats pose risks at individual, organizational, and societal +> levels. This paper examines the systemic impact of these cyber threats and +> proposes a comprehensive cybersecurity strategy that integrates AI-driven +> solutions, such as Intrusion Detection Systems (IDS), with targeted policy +> interventions. By combining technological and regulatory measures, we create a +> multilevel defense capable of addressing both direct threats and indirect +> negative externalities. We emphasize that the synergy between AI-driven +> solutions and policy interventions is essential for neutralizing cyber threats +> and mitigating their negative impact on the digital economy. Finally, we +> underscore the need for continuous adaptation of these strategies, especially +> in response to the rapid advancement of autonomous AI-driven attacks, to ensure +> the creation of secure and resilient digital ecosystems. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on a comprehensive cybersecurity strategy using AI-driven solutions and policy interventions, with no clear emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria. + +--- + +## [Zero-Shot Statistical Tests for LLM-Generated Text Detection using + Finite Sample Concentration Inequalities](https://arxiv.org/abs/http://arxiv.org/abs/2501.02406v2) +**arXiv ID:** http://arxiv.org/abs/2501.02406v2 + +**Abstract:** +> Verifying the provenance of content is crucial to the function of many +> organizations, e.g., educational institutions, social media platforms, firms, +> etc. This problem is becoming increasingly difficult as text generated by Large +> Language Models (LLMs) becomes almost indistinguishable from human-generated +> content. In addition, many institutions utilize in-house LLMs and want to +> ensure that external, non-sanctioned LLMs do not produce content within the +> institution. In this paper, we answer the following question: Given a piece of +> text, can we identify whether it was produced by LLM $A$ or $B$ (where $B$ can +> be a human)? We model LLM-generated text as a sequential stochastic process +> with complete dependence on history and design zero-shot statistical tests to +> distinguish between (i) the text generated by two different sets of LLMs $A$ +> (in-house) and $B$ (non-sanctioned) and also (ii) LLM-generated and +> human-generated texts. We prove that the type I and type II errors for our +> tests decrease exponentially in the text length. In designing our tests, we +> derive concentration inequalities on the difference between log-perplexity and +> the average entropy of the string under $A$. Specifically, for a given string, +> we demonstrate that if the string is generated by $A$, the log-perplexity of +> the string under $A$ converges to the average entropy of the string under $A$, +> except with an exponentially small probability in string length. We also show +> that if $B$ generates the text, except with an exponentially small probability +> in string length, the log-perplexity of the string under $A$ converges to the +> average cross-entropy of $B$ and $A$. Lastly, we present preliminary +> experimental results to support our theoretical results. By enabling guaranteed +> (with high probability) finding of the origin of harmful LLM-generated text +> with arbitrary size, we can help combat misinformation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on statistical tests for detecting the origin of LLM-generated text, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models. It does not investigate, analyze, or propose methods for improving LLM performance through the manipulation of textual input prompts. + +--- + diff --git a/prompt-papers.md b/prompt-papers.md new file mode 100644 index 0000000..1e46ca5 --- /dev/null +++ b/prompt-papers.md @@ -0,0 +1,5903 @@ +# Accepted Papers + +## [LLMPC: Large Language Model Predictive Control](https://arxiv.org/abs/http://arxiv.org/abs/2501.02486v1) +**arXiv ID:** http://arxiv.org/abs/2501.02486v1 + +**Abstract:** +> Recent advancements in prompting techniques for Large Language Models (LLMs) +> have improved their reasoning, planning, and action abilities. This paper +> examines these prompting techniques through the lens of model predictive +> control (MPC). We show that LLMs act as implicit planning cost function +> minimizers when planning prompts are used. Under our framework we demonstrate +> that LLM planning performance can be improved further by incorporating real +> planning cost functions and evaluators. + +**Decision Explanation:** +Original response: +{ + "decision": "ACCEPT", + "explanation": "The paper focuses primarily on optimizing LLM performance through the manipulation of textual input prompts (planning prompts), provides a framework (LLMPC) for systematic prompt development, and demonstrates its impact on LLM output, aligning with the core subject of prompt engineering for text-based interactions with LLMs." +} + +--- + +# Rejected Papers + +## [Co-Activation Graph Analysis of Safety-Verified and Explainable Deep + Reinforcement Learning Policies](https://arxiv.org/abs/http://arxiv.org/abs/2501.03142v1) +**arXiv ID:** http://arxiv.org/abs/2501.03142v1 + +**Abstract:** +> Deep reinforcement learning (RL) policies can demonstrate unsafe behaviors +> and are challenging to interpret. To address these challenges, we combine RL +> policy model checking--a technique for determining whether RL policies exhibit +> unsafe behaviors--with co-activation graph analysis--a method that maps neural +> network inner workings by analyzing neuron activation patterns--to gain insight +> into the safe RL policy's sequential decision-making. This combination lets us +> interpret the RL policy's inner workings for safe decision-making. We +> demonstrate its applicability in various experiments. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on deep reinforcement learning (RL) policies, model checking, and co-activation graph analysis, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria. + +--- + +## [Turn-based Multi-Agent Reinforcement Learning Model Checking](https://arxiv.org/abs/http://arxiv.org/abs/2501.03187v1) +**arXiv ID:** http://arxiv.org/abs/2501.03187v1 + +**Abstract:** +> In this paper, we propose a novel approach for verifying the compliance of +> turn-based multi-agent reinforcement learning (TMARL) agents with complex +> requirements in stochastic multiplayer games. Our method overcomes the +> limitations of existing verification approaches, which are inadequate for +> dealing with TMARL agents and not scalable to large games with multiple agents. +> Our approach relies on tight integration of TMARL and a verification technique +> referred to as model checking. We demonstrate the effectiveness and scalability +> of our technique through experiments in different types of environments. Our +> experiments show that our method is suited to verify TMARL agents and scales +> better than naive monolithic model checking. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on verifying compliance of turn-based multi-agent reinforcement learning agents, with no clear emphasis on prompt engineering for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria of focusing on engineering, design, or optimization of prompts specifically for LLMs. + +--- + +## [Neural Deconstruction Search for Vehicle Routing Problems](https://arxiv.org/abs/http://arxiv.org/abs/2501.03715v1) +**arXiv ID:** http://arxiv.org/abs/2501.03715v1 + +**Abstract:** +> Autoregressive construction approaches generate solutions to vehicle routing +> problems in a step-by-step fashion, leading to high-quality solutions that are +> nearing the performance achieved by handcrafted, operations research +> techniques. In this work, we challenge the conventional paradigm of sequential +> solution construction and introduce an iterative search framework where +> solutions are instead deconstructed by a neural policy. Throughout the search, +> the neural policy collaborates with a simple greedy insertion algorithm to +> rebuild the deconstructed solutions. Our approach surpasses the performance of +> state-of-the-art operations research methods across three challenging vehicle +> routing problems of various problem sizes. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on solving Vehicle Routing Problems using a neural search framework, which falls under the excluded categories (primarily concerned with applications other than text generation driven by LLMs, and specifically mentions automotive/self-driving/operations research subjects). It also does not meet the 'MUST' criteria of focusing primarily on the engineering, design, or optimization of prompts for Large Language Models (LLMs). + +--- + +## [A completely uniform transformer for parity](https://arxiv.org/abs/http://arxiv.org/abs/2501.02535v1) +**arXiv ID:** http://arxiv.org/abs/2501.02535v1 + +**Abstract:** +> We construct a 3-layer constant-dimension transformer, recognizing the parity +> language, where neither parameter matrices nor the positional encoding depend +> on the input length. This improves upon a construction of Chiang and Cholak who +> use a positional encoding, depending on the input length (but their +> construction has 2 layers). + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on constructing a uniform transformer for recognizing parity language, which aligns with developing new LLM architectures or training methods, violating the 'MUST NOT' criteria 1. Additionally, it does not demonstrate prompt engineering for text-based interactions with LLMs as its core subject. + +--- + +## [Test-time Computing: from System-1 Thinking to System-2 Thinking](https://arxiv.org/abs/http://arxiv.org/abs/2501.02497v1) +**arXiv ID:** http://arxiv.org/abs/2501.02497v1 + +**Abstract:** +> The remarkable performance of the o1 model in complex reasoning demonstrates +> that test-time computing scaling can further unlock the model's potential, +> enabling powerful System-2 thinking. However, there is still a lack of +> comprehensive surveys for test-time computing scaling. We trace the concept of +> test-time computing back to System-1 models. In System-1 models, test-time +> computing addresses distribution shifts and improves robustness and +> generalization through parameter updating, input modification, representation +> editing, and output calibration. In System-2 models, it enhances the model's +> reasoning ability to solve complex problems through repeated sampling, +> self-correction, and tree search. We organize this survey according to the +> trend of System-1 to System-2 thinking, highlighting the key role of test-time +> computing in the transition from System-1 models to weak System-2 models, and +> then to strong System-2 models. We also point out a few possible future +> directions. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the primary focus criteria as it discusses test-time computing scaling for enhancing model performance (pertaining to System-1 and System-2 thinking) without specifically addressing prompt engineering for Large Language Models (LLMs) or demonstrating the impact of textual input prompts on LLM output." +} + +--- + +## [KG-CF: Knowledge Graph Completion with Context Filtering under the + Guidance of Large Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02711v1) +**arXiv ID:** http://arxiv.org/abs/2501.02711v1 + +**Abstract:** +> Large Language Models (LLMs) have shown impressive performance in various +> tasks, including knowledge graph completion (KGC). However, current studies +> mostly apply LLMs to classification tasks, like identifying missing triplets, +> rather than ranking-based tasks, where the model ranks candidate entities based +> on plausibility. This focus limits the practical use of LLMs in KGC, as +> real-world applications prioritize highly plausible triplets. Additionally, +> while graph paths can help infer the existence of missing triplets and improve +> completion accuracy, they often contain redundant information. To address these +> issues, we propose KG-CF, a framework tailored for ranking-based KGC tasks. +> KG-CF leverages LLMs' reasoning abilities to filter out irrelevant contexts, +> achieving superior results on real-world datasets. The code and datasets are +> available at \url{https://anonymous.4open.science/r/KG-CF}. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on leveraging LLMs for knowledge graph completion (ranking-based tasks) rather than on the engineering, design, or optimization of prompts specifically for LLMs. While LLMs are used, the central concern is the application (KGC) and not prompt engineering techniques, methods, or their impact on LLM output." +} + +--- + +## [Artificial Intelligence in Creative Industries: Advances Prior to 2025](https://arxiv.org/abs/http://arxiv.org/abs/2501.02725v1) +**arXiv ID:** http://arxiv.org/abs/2501.02725v1 + +**Abstract:** +> The rapid advancements in artificial intelligence (AI), particularly in +> generative AI and large language models (LLMs), have profoundly impacted the +> creative industries by enabling innovative content creation, enhancing +> workflows, and democratizing access to creative tools. This paper explores the +> significant technological shifts since our previous review in 2022, +> highlighting how these developments have expanded creative opportunities and +> efficiency. These technological advancements have enhanced the capabilities of +> text-to-image, text-to-video, and multimodal generation technologies. In +> particular, key breakthroughs in LLMs have established new benchmarks in +> conversational AI, while advancements in image generators have revolutionized +> content creation. We also discuss AI integration into post-production +> workflows, which has significantly accelerated and refined traditional +> processes. Despite these innovations, challenges remain, particularly for the +> media industry, due to the demands on communication traffic from creative +> content. We therefore include data compression and quality assessment in this +> paper. Furthermore, we highlight the trend toward unified AI frameworks capable +> of addressing multiple creative tasks and underscore the importance of human +> oversight to mitigate AI-generated inaccuracies. Finally, we explore AI's +> future potential in the creative sector, stressing the need to navigate +> emerging challenges to maximize its benefits while addressing associated risks. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on the broader impact of AI in creative industries, mentioning LLMs and generative AI as components, rather than specifically concentrating on prompt engineering for LLMs as required by the criteria. + +--- + +## [Multi-Agent Path Finding under Limited Communication Range Constraint + via Dynamic Leading](https://arxiv.org/abs/http://arxiv.org/abs/2501.02770v1) +**arXiv ID:** http://arxiv.org/abs/2501.02770v1 + +**Abstract:** +> This paper proposes a novel framework to handle a multi-agent path finding +> problem under a limited communication range constraint, where all agents must +> have a connected communication channel to the rest of the team. Many existing +> approaches to multi-agent path finding (e.g., leader-follower platooning) +> overcome computational challenges of planning in this domain by planning one +> agent at a time in a fixed order. However, fixed leader-follower approaches can +> become stuck during planning, limiting their practical utility in dense-clutter +> environments. To overcome this limitation, we develop dynamic leading +> multi-agent path finding, which allows for dynamic reselection of the leading +> agent during path planning whenever progress cannot be made. The experiments +> show the efficiency of our framework, which can handle up to 25 agents with +> more than 90% success-rate across five environment types where baselines +> routinely fail. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on multi-agent path finding under limited communication constraints, utilizing dynamic leading, and does not meet any of the 'MUST' criteria, particularly lacking primary focus on prompt engineering for Large Language Models (LLMs) and manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [SenseRAG: Constructing Environmental Knowledge Bases with Proactive + Querying for LLM-Based Autonomous Driving](https://arxiv.org/abs/http://arxiv.org/abs/2501.03535v2) +**arXiv ID:** http://arxiv.org/abs/2501.03535v2 + +**Abstract:** +> This study addresses the critical need for enhanced situational awareness in +> autonomous driving (AD) by leveraging the contextual reasoning capabilities of +> large language models (LLMs). Unlike traditional perception systems that rely +> on rigid, label-based annotations, it integrates real-time, multimodal sensor +> data into a unified, LLMs-readable knowledge base, enabling LLMs to dynamically +> understand and respond to complex driving environments. To overcome the +> inherent latency and modality limitations of LLMs, a proactive +> Retrieval-Augmented Generation (RAG) is designed for AD, combined with a +> chain-of-thought prompting mechanism, ensuring rapid and context-rich +> understanding. Experimental results using real-world Vehicle-to-everything +> (V2X) datasets demonstrate significant improvements in perception and +> prediction performance, highlighting the potential of this framework to enhance +> safety, adaptability, and decision-making in next-generation AD systems. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing autonomous driving (AD) systems with LLMs, rather than prompt engineering for text-based interactions with LLMs. Although it mentions a 'chain-of-thought prompting mechanism', the core subject is the AD system's perception and prediction performance, not novel prompt engineering techniques for LLMs. + +--- + +## [STContext: A Multifaceted Dataset for Developing Context-aware + Spatio-temporal Crowd Mobility Prediction Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03583v1) +**arXiv ID:** http://arxiv.org/abs/2501.03583v1 + +**Abstract:** +> In smart cities, context-aware spatio-temporal crowd flow prediction (STCFP) +> models leverage contextual features (e.g., weather) to identify unusual crowd +> mobility patterns and enhance prediction accuracy. However, the best practice +> for incorporating contextual features remains unclear due to inconsistent usage +> of contextual features in different papers. Developing a multifaceted dataset +> with rich types of contextual features and STCFP scenarios is crucial for +> establishing a principled context modeling paradigm. Existing open crowd flow +> datasets lack an adequate range of contextual features, which poses an urgent +> requirement to build a multifaceted dataset to fill these research gaps. To +> this end, we create STContext, a multifaceted dataset for developing +> context-aware STCFP models. Specifically, STContext provides nine +> spatio-temporal datasets across five STCFP scenarios and includes ten +> contextual features, including weather, air quality index, holidays, points of +> interest, road networks, etc. Besides, we propose a unified workflow for +> incorporating contextual features into deep STCFP methods, with steps including +> feature transformation, dependency modeling, representation fusion, and +> training strategies. Through extensive experiments, we have obtained several +> useful guidelines for effective context modeling and insights for future +> research. The STContext is open-sourced at +> https://github.com/Liyue-Chen/STContext. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Instead, it focuses on developing a multifaceted dataset for context-aware spatio-temporal crowd mobility prediction models, which falls outside the specified criteria. + +--- + +## [Neural DNF-MT: A Neuro-symbolic Approach for Learning Interpretable and + Editable Policies](https://arxiv.org/abs/http://arxiv.org/abs/2501.03888v1) +**arXiv ID:** http://arxiv.org/abs/2501.03888v1 + +**Abstract:** +> Although deep reinforcement learning has been shown to be effective, the +> model's black-box nature presents barriers to direct policy interpretation. To +> address this problem, we propose a neuro-symbolic approach called neural DNF-MT +> for end-to-end policy learning. The differentiable nature of the neural DNF-MT +> model enables the use of deep actor-critic algorithms for training. At the same +> time, its architecture is designed so that trained models can be directly +> translated into interpretable policies expressed as standard (bivalent or +> probabilistic) logic programs. Moreover, additional layers can be included to +> extract abstract features from complex observations, acting as a form of +> predicate invention. The logic representations are highly interpretable, and we +> show how the bivalent representations of deterministic policies can be edited +> and incorporated back into a neural model, facilitating manual intervention and +> adaptation of learned policies. We evaluate our approach on a range of tasks +> requiring learning deterministic or stochastic behaviours from various forms of +> observations. Our empirical results show that our neural DNF-MT model performs +> at the level of competing black-box methods whilst providing interpretable +> policies. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on developing a neuro-symbolic approach for interpretable policy learning in reinforcement learning, with no primary focus on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [PPTAgent: Generating and Evaluating Presentations Beyond Text-to-Slides](https://arxiv.org/abs/http://arxiv.org/abs/2501.03936v1) +**arXiv ID:** http://arxiv.org/abs/2501.03936v1 + +**Abstract:** +> Automatically generating presentations from documents is a challenging task +> that requires balancing content quality, visual design, and structural +> coherence. Existing methods primarily focus on improving and evaluating the +> content quality in isolation, often overlooking visual design and structural +> coherence, which limits their practical applicability. To address these +> limitations, we propose PPTAgent, which comprehensively improves presentation +> generation through a two-stage, edit-based approach inspired by human +> workflows. PPTAgent first analyzes reference presentations to understand their +> structural patterns and content schemas, then drafts outlines and generates +> slides through code actions to ensure consistency and alignment. To +> comprehensively evaluate the quality of generated presentations, we further +> introduce PPTEval, an evaluation framework that assesses presentations across +> three dimensions: Content, Design, and Coherence. Experiments show that +> PPTAgent significantly outperforms traditional automatic presentation +> generation methods across all three dimensions. The code and data are available +> at https://github.com/icip-cas/PPTAgent. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on generating presentations (image/video generation) and improving content quality, visual design, and structural coherence, rather than specifically engineering or optimizing text-based input prompts for Large Language Models (LLMs)." +} + +--- + +## [Implementing Systemic Thinking for Automatic Schema Matching: An + Agent-Based Modeling Approach](https://arxiv.org/abs/http://arxiv.org/abs/2501.04136v1) +**arXiv ID:** http://arxiv.org/abs/2501.04136v1 + +**Abstract:** +> Several approaches are proposed to deal with the problem of the Automatic +> Schema Matching (ASM). The challenges and difficulties caused by the complexity +> and uncertainty characterizing both the process and the outcome of Schema +> Matching motivated us to investigate how bio-inspired emerging paradigm can +> help with understanding, managing, and ultimately overcoming those challenges. +> In this paper, we explain how we approached Automatic Schema Matching as a +> systemic and Complex Adaptive System (CAS) and how we modeled it using the +> approach of Agent-Based Modeling and Simulation (ABMS). This effort gives birth +> to a tool (prototype) for schema matching called Reflex-SMAS. A set of +> experiments demonstrates the viability of our approach on two main aspects: (i) +> effectiveness (increasing the quality of the found matchings) and (ii) +> efficiency (reducing the effort required for this efficiency). Our approach +> represents a significant paradigm-shift, in the field of Automatic Schema +> Matching. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary focus criteria, as it concentrates on Automatic Schema Matching using Agent-Based Modeling and Simulation, with no apparent emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [AI-Driven Reinvention of Hydrological Modeling for Accurate Predictions + and Interpretation to Transform Earth System Modeling](https://arxiv.org/abs/http://arxiv.org/abs/2501.04733v1) +**arXiv ID:** http://arxiv.org/abs/2501.04733v1 + +**Abstract:** +> Traditional equation-driven hydrological models often struggle to accurately +> predict streamflow in challenging regional Earth systems like the Tibetan +> Plateau, while hybrid and existing algorithm-driven models face difficulties in +> interpreting hydrological behaviors. This work introduces HydroTrace, an +> algorithm-driven, data-agnostic model that substantially outperforms these +> approaches, achieving a Nash-Sutcliffe Efficiency of 98% and demonstrating +> strong generalization on unseen data. Moreover, HydroTrace leverages advanced +> attention mechanisms to capture spatial-temporal variations and +> feature-specific impacts, enabling the quantification and spatial resolution of +> streamflow partitioning as well as the interpretation of hydrological behaviors +> such as glacier-snow-streamflow interactions and monsoon dynamics. +> Additionally, a large language model (LLM)-based application allows users to +> easily understand and apply HydroTrace's insights for practical purposes. These +> advancements position HydroTrace as a transformative tool in hydrological and +> broader Earth system modeling, offering enhanced prediction accuracy and +> interpretability. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing a new hydrological modeling approach (HydroTrace) and its application in Earth system modeling, with the Large Language Model (LLM) being a secondary component for interpreting insights, rather than the primary subject being prompt engineering for text-based interactions with LLMs. + +--- + +## [Rethinking IDE Customization for Enhanced HAX: A Hyperdimensional + Perspective](https://arxiv.org/abs/http://arxiv.org/abs/2501.02491v1) +**arXiv ID:** http://arxiv.org/abs/2501.02491v1 + +**Abstract:** +> As Integrated Development Environments (IDEs) increasingly integrate +> Artificial Intelligence, Software Engineering faces both benefits like +> productivity gains and challenges like mismatched user preferences. We propose +> Hyper-Dimensional (HD) vector spaces to model Human-Computer Interaction, +> focusing on user actions, stylistic preferences, and project context. These +> contributions aim to inspire further research on applying HD computing in IDE +> design. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on IDE customization using Hyperdimensional vector spaces for Human-Computer Interaction, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [TreeMatch: A Fully Unsupervised WSD System Using Dependency Knowledge on + a Specific Domain](https://arxiv.org/abs/http://arxiv.org/abs/2501.02546v1) +**arXiv ID:** http://arxiv.org/abs/2501.02546v1 + +**Abstract:** +> Word sense disambiguation (WSD) is one of the main challenges in +> Computational Linguistics. TreeMatch is a WSD system originally developed using +> data from SemEval 2007 Task 7 (Coarse-grained English All-words Task) that has +> been adapted for use in SemEval 2010 Task 17 (All-words Word Sense +> Disambiguation on a Specific Domain). The system is based on a fully +> unsupervised method using dependency knowledge drawn from a domain specific +> knowledge base that was built for this task. When evaluated on the task, the +> system precision performs above the Most Frequent Selection baseline. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on Word Sense Disambiguation (WSD) using dependency knowledge and a domain-specific knowledge base, with no indication of primarily investigating, analyzing, or proposing methods for improving Large Language Model (LLM) performance through the manipulation of textual input prompts. + +--- + +## [Efficient Architectures for High Resolution Vision-Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02584v1) +**arXiv ID:** http://arxiv.org/abs/2501.02584v1 + +**Abstract:** +> Vision-Language Models (VLMs) have recently experienced significant +> advancements. However, challenges persist in the accurate recognition of fine +> details within high resolution images, which limits performance in multiple +> tasks. This work introduces Pheye, a novel architecture that efficiently +> processes high-resolution images while training fewer parameters than similarly +> sized VLMs. Notably, Pheye achieves a high efficiency while maintaining strong +> performance, particularly in tasks that demand fine-grained image understanding +> and/or the handling of scene-text. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a novel architecture (Pheye) for Vision-Language Models (VLMs) to process high-resolution images, which aligns with excluded criteria: developing new LLM architectures and being concerned with applications of generative AI other than text generation driven by LLMs (specifically, vision-language tasks). + +--- + +## [Enhancing Robot Route Optimization in Smart Logistics with Transformer + and GNN Integration](https://arxiv.org/abs/http://arxiv.org/abs/2501.02749v1) +**arXiv ID:** http://arxiv.org/abs/2501.02749v1 + +**Abstract:** +> This research delves into advanced route optimization for robots in smart +> logistics, leveraging a fusion of Transformer architectures, Graph Neural +> Networks (GNNs), and Generative Adversarial Networks (GANs). The approach +> utilizes a graph-based representation encompassing geographical data, cargo +> allocation, and robot dynamics, addressing both spatial and resource +> limitations to refine route efficiency. Through extensive testing with +> authentic logistics datasets, the proposed method achieves notable +> improvements, including a 15% reduction in travel distance, a 20% boost in time +> efficiency, and a 10% decrease in energy consumption. These findings highlight +> the algorithm's effectiveness, promoting enhanced performance in intelligent +> logistics operations. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the integration of Transformer, GNN, and GAN for robot route optimization in smart logistics, with no apparent emphasis on prompt engineering, manipulation of textual input prompts, or the interaction with Large Language Models (LLMs) for text generation. + +--- + +## [Key-value memory in the brain](https://arxiv.org/abs/http://arxiv.org/abs/2501.02950v1) +**arXiv ID:** http://arxiv.org/abs/2501.02950v1 + +**Abstract:** +> Classical models of memory in psychology and neuroscience rely on +> similarity-based retrieval of stored patterns, where similarity is a function +> of retrieval cues and the stored patterns. While parsimonious, these models do +> not allow distinct representations for storage and retrieval, despite their +> distinct computational demands. Key-value memory systems, in contrast, +> distinguish representations used for storage (values) and those used for +> retrieval (keys). This allows key-value memory systems to optimize +> simultaneously for fidelity in storage and discriminability in retrieval. We +> review the computational foundations of key-value memory, its role in modern +> machine learning systems, related ideas from psychology and neuroscience, +> applications to a number of empirical puzzles, and possible biological +> implementations. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it does not focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs), nor does it investigate methods for improving LLM performance through prompt manipulation. Instead, it discusses key-value memory systems in the context of psychology, neuroscience, and machine learning, with no apparent connection to LLM prompt engineering. + +--- + +## [Putnam's Critical and Explanatory Tendencies Interpreted from a Machine + Learning Perspective](https://arxiv.org/abs/http://arxiv.org/abs/2501.03026v1) +**arXiv ID:** http://arxiv.org/abs/2501.03026v1 + +**Abstract:** +> Making sense of theory choice in normal and across extraordinary science is +> central to philosophy of science. The emergence of machine learning models has +> the potential to act as a wrench in the gears of current debates. In this +> paper, I will attempt to reconstruct the main movements that lead to and came +> out of Putnam's critical and explanatory tendency distinction, argue for the +> biconditional necessity of the tendencies, and conceptualize that wrench +> through a machine learning interpretation of my claim. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on interpreting philosophical concepts through a machine learning perspective, not specifically on the engineering, design, or optimization of prompts for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [GLiREL -- Generalist Model for Zero-Shot Relation Extraction](https://arxiv.org/abs/http://arxiv.org/abs/2501.03172v1) +**arXiv ID:** http://arxiv.org/abs/2501.03172v1 + +**Abstract:** +> We introduce GLiREL (Generalist Lightweight model for zero-shot Relation +> Extraction), an efficient architecture and training paradigm for zero-shot +> relation classification. Inspired by recent advancements in zero-shot named +> entity recognition, this work presents an approach to efficiently and +> accurately predict zero-shot relationship labels between multiple entities in a +> single forward pass. Experiments using the FewRel and WikiZSL benchmarks +> demonstrate that our approach achieves state-of-the-art results on the +> zero-shot relation classification task. In addition, we contribute a protocol +> for synthetically-generating datasets with diverse relation labels. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a new efficient architecture and training paradigm for zero-shot relation classification, rather than engineering, designing, or optimizing prompts for Large Language Models (LLMs) to improve their text generation performance. + +--- + +## [Can LLMs Design Good Questions Based on Context?](https://arxiv.org/abs/http://arxiv.org/abs/2501.03491v1) +**arXiv ID:** http://arxiv.org/abs/2501.03491v1 + +**Abstract:** +> This paper evaluates questions generated by LLMs from context, comparing them +> to human-generated questions across six dimensions. We introduce an automated +> LLM-based evaluation method, focusing on aspects like question length, type, +> context coverage, and answerability. Our findings highlight unique +> characteristics of LLM-generated questions, contributing insights that can +> support further research in question quality and downstream applications. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on LLMs generating questions based on context, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) to improve their text generation performance through textual input manipulation. + +--- + +## [Self-Adaptive ERP: Embedding NLP into Petri-Net creation and Model + Matching](https://arxiv.org/abs/http://arxiv.org/abs/2501.03795v1) +**arXiv ID:** http://arxiv.org/abs/2501.03795v1 + +**Abstract:** +> Enterprise Resource Planning (ERP) consultants play a vital role in +> customizing systems to meet specific business needs by processing large amounts +> of data and adapting functionalities. However, the process is +> resource-intensive, time-consuming, and requires continuous adjustments as +> business demands evolve. This research introduces a Self-Adaptive ERP Framework +> that automates customization using enterprise process models and system usage +> analysis. It leverages Artificial Intelligence (AI) & Natural Language +> Processing (NLP) for Petri nets to transform business processes into adaptable +> models, addressing both structural and functional matching. The framework, +> built using Design Science Research (DSR) and a Systematic Literature Review +> (SLR), reduces reliance on manual adjustments, improving ERP customization +> efficiency and accuracy while minimizing the need for consultants. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on automating ERP customization using AI and NLP for Petri-net creation and model matching, rather than specifically engineering, designing, or optimizing prompts for Large Language Models (LLMs). The mention of NLP is not centered around prompt engineering for LLMs, but rather for transforming business processes into adaptable models within an ERP framework. + +--- + +## [Synthetic Data Privacy Metrics](https://arxiv.org/abs/http://arxiv.org/abs/2501.03941v1) +**arXiv ID:** http://arxiv.org/abs/2501.03941v1 + +**Abstract:** +> Recent advancements in generative AI have made it possible to create +> synthetic datasets that can be as accurate as real-world data for training AI +> models, powering statistical insights, and fostering collaboration with +> sensitive datasets while offering strong privacy guarantees. Effectively +> measuring the empirical privacy of synthetic data is an important step in the +> process. However, while there is a multitude of new privacy metrics being +> published every day, there currently is no standardization. In this paper, we +> review the pros and cons of popular metrics that include simulations of +> adversarial attacks. We also review current best practices for amending +> generative models to enhance the privacy of the data they create (e.g. +> differential privacy). + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on synthetic data privacy metrics, generative AI for dataset creation, and differential privacy, without addressing prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Traits of a Leader: User Influence Level Prediction through + Sociolinguistic Modeling](https://arxiv.org/abs/http://arxiv.org/abs/2501.04046v1) +**arXiv ID:** http://arxiv.org/abs/2501.04046v1 + +**Abstract:** +> Recognition of a user's influence level has attracted much attention as human +> interactions move online. Influential users have the ability to sway others' +> opinions to achieve some goals. As a result, predicting users' level of +> influence can help to understand social networks, forecast trends, prevent +> misinformation, etc. However, predicting user influence is a challenging +> problem because the concept of influence is specific to a situation or a +> domain, and user communications are limited to text. In this work, we define +> user influence level as a function of community endorsement and develop a model +> that significantly outperforms the baseline by leveraging demographic and +> personality data. This approach consistently improves RankDCG scores across +> eight different domains. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on sociolinguistic modeling for predicting user influence levels in online interactions, without primarily addressing the engineering, design, or optimization of prompts for Large Language Models (LLMs) or demonstrating the manipulation of textual input prompts to improve LLM performance. + +--- + +## [IntegrityAI at GenAI Detection Task 2: Detecting Machine-Generated + Academic Essays in English and Arabic Using ELECTRA and Stylometry](https://arxiv.org/abs/http://arxiv.org/abs/2501.05476v1) +**arXiv ID:** http://arxiv.org/abs/2501.05476v1 + +**Abstract:** +> Recent research has investigated the problem of detecting machine-generated +> essays for academic purposes. To address this challenge, this research utilizes +> pre-trained, transformer-based models fine-tuned on Arabic and English academic +> essays with stylometric features. Custom models based on ELECTRA for English +> and AraELECTRA for Arabic were trained and evaluated using a benchmark dataset. +> Proposed models achieved excellent results with an F1-score of 99.7%, ranking +> 2nd among of 26 teams in the English subtask, and 98.4%, finishing 1st out of +> 23 teams in the Arabic one. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on developing and fine-tuning transformer-based models (ELECTRA, AraELECTRA) for detecting machine-generated essays, rather than engineering or optimizing prompts for Large Language Models (LLMs). Prompt engineering is not the central concern of this research." +} + +--- + +## [Fairness Through Matching](https://arxiv.org/abs/http://arxiv.org/abs/2501.02793v1) +**arXiv ID:** http://arxiv.org/abs/2501.02793v1 + +**Abstract:** +> Group fairness requires that different protected groups, characterized by a +> given sensitive attribute, receive equal outcomes overall. Typically, the level +> of group fairness is measured by the statistical gap between predictions from +> different protected groups. In this study, we reveal an implicit property of +> existing group fairness measures, which provides an insight into how the +> group-fair models behave. Then, we develop a new group-fair constraint based on +> this implicit property to learn group-fair models. To do so, we first introduce +> a notable theoretical observation: every group-fair model has an implicitly +> corresponding transport map between the input spaces of each protected group. +> Based on this observation, we introduce a new group fairness measure termed +> Matched Demographic Parity (MDP), which quantifies the averaged gap between +> predictions of two individuals (from different protected groups) matched by a +> given transport map. Then, we prove that any transport map can be used in MDP +> to learn group-fair models, and develop a novel algorithm called Fairness +> Through Matching (FTM), which learns a group-fair model using MDP constraint +> with an user-specified transport map. We specifically propose two favorable +> types of transport maps for MDP, based on the optimal transport theory, and +> discuss their advantages. Experiments reveal that FTM successfully trains +> group-fair models with certain desirable properties by choosing the transport +> map accordingly. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs), instead concentrating on developing a new group-fair constraint and algorithm for learnings group-fair models, with no mention of LLMs, prompt design, or textual input manipulation. + +--- + +## [CALM: Curiosity-Driven Auditing for Large Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02997v1) +**arXiv ID:** http://arxiv.org/abs/2501.02997v1 + +**Abstract:** +> Auditing Large Language Models (LLMs) is a crucial and challenging task. In +> this study, we focus on auditing black-box LLMs without access to their +> parameters, only to the provided service. We treat this type of auditing as a +> black-box optimization problem where the goal is to automatically uncover +> input-output pairs of the target LLMs that exhibit illegal, immoral, or unsafe +> behaviors. For instance, we may seek a non-toxic input that the target LLM +> responds to with a toxic output or an input that induces the hallucinative +> response from the target LLM containing politically sensitive individuals. This +> black-box optimization is challenging due to the scarcity of feasible points, +> the discrete nature of the prompt space, and the large search space. To address +> these challenges, we propose Curiosity-Driven Auditing for Large Language +> Models (CALM), which uses intrinsically motivated reinforcement learning to +> finetune an LLM as the auditor agent to uncover potential harmful and biased +> input-output pairs of the target LLM. CALM successfully identifies derogatory +> completions involving celebrities and uncovers inputs that elicit specific +> names under the black-box setting. This work offers a promising direction for +> auditing black-box LLMs. Our code is available at +> https://github.com/x-zheng16/CALM.git. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on auditing Large Language Models (LLMs) using a reinforcement learning approach, rather than on the engineering, design, or optimization of prompts for improving LLM performance through textual input manipulation. + +--- + +## [Analyzing Fine-tuning Representation Shift for Multimodal LLMs Steering + alignment](https://arxiv.org/abs/http://arxiv.org/abs/2501.03012v1) +**arXiv ID:** http://arxiv.org/abs/2501.03012v1 + +**Abstract:** +> Multimodal LLMs have reached remarkable levels of proficiency in +> understanding multimodal inputs, driving extensive research to develop +> increasingly powerful models. However, much less attention has been paid to +> understanding and explaining the underlying mechanisms of these models. Most +> existing explainability research examines these models only in their final +> states, overlooking the dynamic representational shifts that occur during +> training. In this work, we systematically analyze the evolution of hidden state +> representations to reveal how fine-tuning alters the internal structure of a +> model to specialize in new multimodal tasks. Using a concept-based approach, we +> map hidden states to interpretable visual and textual concepts, enabling us to +> trace changes in encoded concepts across modalities as training progresses. We +> also demonstrate the use of shift vectors to capture these concepts changes. +> These shift vectors allow us to recover fine-tuned concepts by shifting those +> in the original model. Finally, we explore the practical impact of our findings +> on model steering, showing that we can adjust multimodal LLMs behaviors without +> any training, such as modifying answer types, captions style, or biasing the +> model toward specific responses. Our work sheds light on how multimodal +> representations evolve through fine-tuning and offers a new perspective for +> interpreting model adaptation in multimodal tasks. The code for this project is +> publicly available at https://github.com/mshukor/xl-vlms. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on analyzing and understanding the internal representation shifts of multimodal LLMs during fine-tuning, rather than engineering, designing, or optimizing prompts specifically for LLMs, thus failing to meet the core 'MUST' criteria." +} + +--- + +## [Large language models for artificial general intelligence (AGI): A + survey of foundational principles and approaches](https://arxiv.org/abs/http://arxiv.org/abs/2501.03151v1) +**arXiv ID:** http://arxiv.org/abs/2501.03151v1 + +**Abstract:** +> Generative artificial intelligence (AI) systems based on large-scale +> pretrained foundation models (PFMs) such as vision-language models, large +> language models (LLMs), diffusion models and vision-language-action (VLA) +> models have demonstrated the ability to solve complex and truly non-trivial AI +> problems in a wide variety of domains and contexts. Multimodal large language +> models (MLLMs), in particular, learn from vast and diverse data sources, +> allowing rich and nuanced representations of the world and, thereby, providing +> extensive capabilities, including the ability to reason, engage in meaningful +> dialog; collaborate with humans and other agents to jointly solve complex +> problems; and understand social and emotional aspects of humans. Despite this +> impressive feat, the cognitive abilities of state-of-the-art LLMs trained on +> large-scale datasets are still superficial and brittle. Consequently, generic +> LLMs are severely limited in their generalist capabilities. A number of +> foundational problems -- embodiment, symbol grounding, causality and memory -- +> are required to be addressed for LLMs to attain human-level general +> intelligence. These concepts are more aligned with human cognition and provide +> LLMs with inherent human-like cognitive properties that support the realization +> of physically-plausible, semantically meaningful, flexible and more +> generalizable knowledge and intelligence. In this work, we discuss the +> aforementioned foundational issues and survey state-of-the art approaches for +> implementing these concepts in LLMs. Specifically, we discuss how the +> principles of embodiment, symbol grounding, causality and memory can be +> leveraged toward the attainment of artificial general intelligence (AGI) in an +> organic manner. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on foundational principles and approaches for achieving Artificial General Intelligence (AGI) with Large Language Models (LLMs), rather than specifically on the engineering, design, or optimization of prompts for LLMs. It lacks concrete examples of prompts and their impact on LLM output, which is a required criterion. + +--- + +## [From Aleatoric to Epistemic: Exploring Uncertainty Quantification + Techniques in Artificial Intelligence](https://arxiv.org/abs/http://arxiv.org/abs/2501.03282v1) +**arXiv ID:** http://arxiv.org/abs/2501.03282v1 + +**Abstract:** +> Uncertainty quantification (UQ) is a critical aspect of artificial +> intelligence (AI) systems, particularly in high-risk domains such as +> healthcare, autonomous systems, and financial technology, where decision-making +> processes must account for uncertainty. This review explores the evolution of +> uncertainty quantification techniques in AI, distinguishing between aleatoric +> and epistemic uncertainties, and discusses the mathematical foundations and +> methods used to quantify these uncertainties. We provide an overview of +> advanced techniques, including probabilistic methods, ensemble learning, +> sampling-based approaches, and generative models, while also highlighting +> hybrid approaches that integrate domain-specific knowledge. Furthermore, we +> examine the diverse applications of UQ across various fields, emphasizing its +> impact on decision-making, predictive accuracy, and system robustness. The +> review also addresses key challenges such as scalability, efficiency, and +> integration with explainable AI, and outlines future directions for research in +> this rapidly developing area. Through this comprehensive survey, we aim to +> provide a deeper understanding of UQ's role in enhancing the reliability, +> safety, and trustworthiness of AI systems. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it focuses primarily on uncertainty quantification techniques in AI across various high-risk domains, rather than specifically on the engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not provide concrete examples of prompts impacting LLM output. + +--- + +## [Online Reinforcement Learning-Based Dynamic Adaptive Evaluation Function + for Real-Time Strategy Tasks](https://arxiv.org/abs/http://arxiv.org/abs/2501.03824v1) +**arXiv ID:** http://arxiv.org/abs/2501.03824v1 + +**Abstract:** +> Effective evaluation of real-time strategy tasks requires adaptive mechanisms +> to cope with dynamic and unpredictable environments. This study proposes a +> method to improve evaluation functions for real-time responsiveness to +> battle-field situation changes, utilizing an online reinforcement +> learning-based dynam-ic weight adjustment mechanism within the real-time +> strategy game. Building on traditional static evaluation functions, the method +> employs gradient descent in online reinforcement learning to update weights +> dynamically, incorporating weight decay techniques to ensure stability. +> Additionally, the AdamW optimizer is integrated to adjust the learning rate and +> decay rate of online reinforcement learning in real time, further reducing the +> dependency on manual parameter tun-ing. Round-robin competition experiments +> demonstrate that this method signifi-cantly enhances the application +> effectiveness of the Lanchester combat model evaluation function, Simple +> evaluation function, and Simple Sqrt evaluation function in planning algorithms +> including IDABCD, IDRTMinimax, and Port-folio AI. The method achieves a notable +> improvement in scores, with the en-hancement becoming more pronounced as the +> map size increases. Furthermore, the increase in evaluation function +> computation time induced by this method is kept below 6% for all evaluation +> functions and planning algorithms. The pro-posed dynamic adaptive evaluation +> function demonstrates a promising approach for real-time strategy task +> evaluation. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on improving evaluation functions for real-time strategy tasks using online reinforcement learning, with no emphasis on prompt engineering, design, or optimization for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria. + +--- + +## [Dolphin: Closed-loop Open-ended Auto-research through Thinking, + Practice, and Feedback](https://arxiv.org/abs/http://arxiv.org/abs/2501.03916v2) +**arXiv ID:** http://arxiv.org/abs/2501.03916v2 + +**Abstract:** +> The scientific research paradigm is undergoing a profound transformation +> owing to the development of Artificial Intelligence (AI). Recent works +> demonstrate that various AI-assisted research methods can largely improve +> research efficiency by improving data analysis, accelerating computation, and +> fostering novel idea generation. To further move towards the ultimate goal +> (i.e., automatic scientific research), in this paper, we propose Dolphin, the +> first closed-loop open-ended auto-research framework to further build the +> entire process of human scientific research. Dolphin can generate research +> ideas, perform experiments, and get feedback from experimental results to +> generate higher-quality ideas. More specifically, Dolphin first generates novel +> ideas based on relevant papers which are ranked by the topic and task +> attributes. Then, the codes are automatically generated and debugged with the +> exception-traceback-guided local code structure. Finally, Dolphin automatically +> analyzes the results of each idea and feeds the results back to the next round +> of idea generation. Experiments are conducted on the benchmark datasets of +> different topics and results show that Dolphin can generate novel ideas +> continuously and complete the experiment in a loop. We highlight that Dolphin +> can automatically propose methods that are comparable to the state-of-the-art +> in some tasks such as 2D image classification and 3D point classification. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing an auto-research framework using AI, with Large Language Models (LLMs) being only one of the potential tools used for generating research ideas, rather than the central focus on prompt engineering for text-based interactions with LLMs as required. + +--- + +## [Interpretable Neural ODEs for Gene Regulatory Network Discovery under + Perturbations](https://arxiv.org/abs/http://arxiv.org/abs/2501.02409v1) +**arXiv ID:** http://arxiv.org/abs/2501.02409v1 + +**Abstract:** +> Modern high-throughput biological datasets with thousands of perturbations +> provide the opportunity for large-scale discovery of causal graphs that +> represent the regulatory interactions between genes. Numerous methods have been +> proposed to infer a directed acyclic graph (DAG) corresponding to the +> underlying gene regulatory network (GRN) that captures causal gene +> relationships. However, existing models have restrictive assumptions (e.g. +> linearity, acyclicity), limited scalability, and/or fail to address the dynamic +> nature of biological processes such as cellular differentiation. We propose +> PerturbODE, a novel framework that incorporates biologically informative neural +> ordinary differential equations (neural ODEs) to model cell state trajectories +> under perturbations and derive the causal GRN from the neural ODE's parameters. +> We demonstrate PerturbODE's efficacy in trajectory prediction and GRN inference +> across simulated and real over-expression datasets. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet any of the 'MUST' criteria, as it focuses on gene regulatory network discovery using neural ODEs, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, falling outside the specified scope. + +--- + +## [A Statistical Hypothesis Testing Framework for Data Misappropriation + Detection in Large Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02441v1) +**arXiv ID:** http://arxiv.org/abs/2501.02441v1 + +**Abstract:** +> Large Language Models (LLMs) are rapidly gaining enormous popularity in +> recent years. However, the training of LLMs has raised significant privacy and +> legal concerns, particularly regarding the inclusion of copyrighted materials +> in their training data without proper attribution or licensing, which falls +> under the broader issue of data misappropriation. In this article, we focus on +> a specific problem of data misappropriation detection, namely, to determine +> whether a given LLM has incorporated data generated by another LLM. To address +> this issue, we propose embedding watermarks into the copyrighted training data +> and formulating the detection of data misappropriation as a hypothesis testing +> problem. We develop a general statistical testing framework, construct a +> pivotal statistic, determine the optimal rejection threshold, and explicitly +> control the type I and type II errors. Furthermore, we establish the asymptotic +> optimality properties of the proposed tests, and demonstrate its empirical +> effectiveness through intensive numerical experiments. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on detecting data misappropriation in LLMs through statistical hypothesis testing, with no emphasis on prompt engineering, design, or optimization for improving LLM performance through textual input prompts." +} + +--- + +## [RTLMarker: Protecting LLM-Generated RTL Copyright via a Hardware + Watermarking Framework](https://arxiv.org/abs/http://arxiv.org/abs/2501.02446v1) +**arXiv ID:** http://arxiv.org/abs/2501.02446v1 + +**Abstract:** +> Recent advances of large language models in the field of Verilog generation +> have raised several ethical and security concerns, such as code copyright +> protection and dissemination of malicious code. Researchers have employed +> watermarking techniques to identify codes generated by large language models. +> However, the existing watermarking works fail to protect RTL code copyright due +> to the significant syntactic and semantic differences between RTL code and +> software code in languages such as Python. This paper proposes a hardware +> watermarking framework RTLMarker that embeds watermarks into RTL code and +> deeper into the synthesized netlist. We propose a set of rule-based Verilog +> code transformations , ensuring the watermarked RTL code's syntactic and +> semantic correctness. In addition, we consider an inherent tradeoff between +> watermark transparency and watermark effectiveness and jointly optimize them. +> The results demonstrate RTLMarker's superiority over the baseline in RTL code +> watermarking. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on a hardware watermarking framework (RTLMarker) for protecting RTL code copyright generated by LLMs, rather than exploring prompt engineering techniques for improving LLM performance through textual input manipulation." +} + +--- + +## [Enhancing Contrastive Learning for Retinal Imaging via Adjusted + Augmentation Scales](https://arxiv.org/abs/http://arxiv.org/abs/2501.02451v1) +**arXiv ID:** http://arxiv.org/abs/2501.02451v1 + +**Abstract:** +> Contrastive learning, a prominent approach within self-supervised learning, +> has demonstrated significant effectiveness in developing generalizable models +> for various applications involving natural images. However, recent research +> indicates that these successes do not necessarily extend to the medical imaging +> domain. In this paper, we investigate the reasons for this suboptimal +> performance and hypothesize that the dense distribution of medical images poses +> challenges to the pretext tasks in contrastive learning, particularly in +> constructing positive and negative pairs. We explore model performance under +> different augmentation strategies and compare the results to those achieved +> with strong augmentations. Our study includes six publicly available datasets +> covering multiple clinically relevant tasks. We further assess the model's +> generalizability through external evaluations. The model pre-trained with weak +> augmentation outperforms those with strong augmentation, improving AUROC from +> 0.838 to 0.848 and AUPR from 0.523 to 0.597 on MESSIDOR2, and showing similar +> enhancements across other datasets. Our findings suggest that optimizing the +> scale of augmentation is critical for enhancing the efficacy of contrastive +> learning in medical imaging. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on enhancing contrastive learning for medical imaging (retinal imaging), which violates the 'MUST NOT' criteria: being primarily concerned with medical subjects and not focusing on prompt engineering for Large Language Models (LLMs) or text generation driven by LLMs." +} + +--- + +## [Hengqin-RA-v1: Advanced Large Language Model for Diagnosis and Treatment + of Rheumatoid Arthritis with Dataset based Traditional Chinese Medicine](https://arxiv.org/abs/http://arxiv.org/abs/2501.02471v1) +**arXiv ID:** http://arxiv.org/abs/2501.02471v1 + +**Abstract:** +> Large language models (LLMs) primarily trained on English texts, often face +> biases and inaccuracies in Chinese contexts. Their limitations are pronounced +> in fields like Traditional Chinese Medicine (TCM), where cultural and clinical +> subtleties are vital, further hindered by a lack of domain-specific data, such +> as rheumatoid arthritis (RA). To address these issues, this paper introduces +> Hengqin-RA-v1, the first large language model specifically tailored for TCM +> with a focus on diagnosing and treating RA. We also present HQ-GCM-RA-C1, a +> comprehensive RA-specific dataset curated from ancient Chinese medical +> literature, classical texts, and modern clinical studies. This dataset empowers +> Hengqin-RA-v1 to deliver accurate and culturally informed responses, +> effectively bridging the gaps left by general-purpose models. Extensive +> experiments demonstrate that Hengqin-RA-v1 outperforms state-of-the-art models, +> even surpassing the diagnostic accuracy of TCM practitioners in certain cases. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a new Large Language Model architecture (Hengqin-RA-v1) specifically tailored for Traditional Chinese Medicine, rather than prompt engineering for existing LLMs, and introduces a new dataset for training, which violate the 'MUST NOT' criteria 1. + +--- + +## [The Meta-Representation Hypothesis](https://arxiv.org/abs/http://arxiv.org/abs/2501.02481v1) +**arXiv ID:** http://arxiv.org/abs/2501.02481v1 + +**Abstract:** +> Humans rely on high-level meta-representations to engage in abstract +> reasoning. In complex cognitive tasks, these meta-representations help +> individuals abstract general rules from experience. However, constructing such +> meta-representations from high-dimensional observations remains a longstanding +> challenge for reinforcement learning agents. For instance, a well-trained agent +> often fails to generalize to even minor variations of the same task, such as +> changes in background color, while humans can easily handle. In this paper, we +> build a bridge between meta-representation and generalization, showing that +> generalization performance benefits from meta-representation learning. We also +> hypothesize that deep mutual learning (DML) among agents can help them converge +> to meta-representations. Empirical results provide support for our theory and +> hypothesis. Overall, this work provides a new perspective on the generalization +> of deep reinforcement learning. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it primarily focuses on reinforcement learning, meta-representation, and generalization in deep learning, with no clear emphasis on prompt engineering, design, or optimization specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts for improving LLM performance. + +--- + +## [Watch Video, Catch Keyword: Context-aware Keyword Attention for Moment + Retrieval and Highlight Detection](https://arxiv.org/abs/http://arxiv.org/abs/2501.02504v1) +**arXiv ID:** http://arxiv.org/abs/2501.02504v1 + +**Abstract:** +> The goal of video moment retrieval and highlight detection is to identify +> specific segments and highlights based on a given text query. With the rapid +> growth of video content and the overlap between these tasks, recent works have +> addressed both simultaneously. However, they still struggle to fully capture +> the overall video context, making it challenging to determine which words are +> most relevant. In this paper, we present a novel Video Context-aware Keyword +> Attention module that overcomes this limitation by capturing keyword variation +> within the context of the entire video. To achieve this, we introduce a video +> context clustering module that provides concise representations of the overall +> video context, thereby enhancing the understanding of keyword dynamics. +> Furthermore, we propose a keyword weight detection module with keyword-aware +> contrastive learning that incorporates keyword information to enhance +> fine-grained alignment between visual and textual features. Extensive +> experiments on the QVHighlights, TVSum, and Charades-STA benchmarks demonstrate +> that our proposed method significantly improves performance in moment retrieval +> and highlight detection tasks compared to existing approaches. Our code is +> available at: https://github.com/VisualAIKHU/Keyword-DETR + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on video moment retrieval and highlight detection, utilizing a novel Video Context-aware Keyword Attention module, with no evident focus on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [Remote Inference over Dynamic Links via Adaptive Rate Deep Task-Oriented + Vector Quantization](https://arxiv.org/abs/http://arxiv.org/abs/2501.02521v1) +**arXiv ID:** http://arxiv.org/abs/2501.02521v1 + +**Abstract:** +> A broad range of technologies rely on remote inference, wherein data acquired +> is conveyed over a communication channel for inference in a remote server. +> Communication between the participating entities is often carried out over +> rate-limited channels, necessitating data compression for reducing latency. +> While deep learning facilitates joint design of the compression mapping along +> with encoding and inference rules, existing learned compression mechanisms are +> static, and struggle in adapting their resolution to changes in channel +> conditions and to dynamic links. To address this, we propose Adaptive Rate +> Task-Oriented Vector Quantization (ARTOVeQ), a learned compression mechanism +> that is tailored for remote inference over dynamic links. ARTOVeQ is based on +> designing nested codebooks along with a learning algorithm employing +> progressive learning. We show that ARTOVeQ extends to support low-latency +> inference that is gradually refined via successive refinement principles, and +> that it enables the simultaneous usage of multiple resolutions when conveying +> high-dimensional data. Numerical results demonstrate that the proposed scheme +> yields remote deep inference that operates with multiple rates, supports a +> broad range of bit budgets, and facilitates rapid inference that gradually +> improves with more bits exchanged, while approaching the performance of +> single-rate deep quantization methods. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on adaptive rate deep task-oriented vector quantization for remote inference over dynamic links, without any primary concern for prompt engineering, Large Language Models (LLMs), or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Face-MakeUp: Multimodal Facial Prompts for Text-to-Image Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02523v1) +**arXiv ID:** http://arxiv.org/abs/2501.02523v1 + +**Abstract:** +> Facial images have extensive practical applications. Although the current +> large-scale text-image diffusion models exhibit strong generation capabilities, +> it is challenging to generate the desired facial images using only text prompt. +> Image prompts are a logical choice. However, current methods of this type +> generally focus on general domain. In this paper, we aim to optimize image +> makeup techniques to generate the desired facial images. Specifically, (1) we +> built a dataset of 4 million high-quality face image-text pairs +> (FaceCaptionHQ-4M) based on LAION-Face to train our Face-MakeUp model; (2) to +> maintain consistency with the reference facial image, we extract/learn +> multi-scale content features and pose features for the facial image, +> integrating these into the diffusion model to enhance the preservation of +> facial identity features for diffusion models. Validation on two face-related +> test datasets demonstrates that our Face-MakeUp can achieve the best +> comprehensive performance.All codes are available +> at:https://github.com/ddw2AIGROUP2CQUPT/Face-MakeUp + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on optimizing image makeup techniques for text-to-image generation, which falls under image generation driven by non-text-based interactions with generative models, violating the 'MUST NOT' criteria 2, and does not meet the core subject requirement of prompt engineering for text-based interactions with Large Language Models (LLMs)." +} + +--- + +## [AMM: Adaptive Modularized Reinforcement Model for Multi-city Traffic + Signal Control](https://arxiv.org/abs/http://arxiv.org/abs/2501.02548v1) +**arXiv ID:** http://arxiv.org/abs/2501.02548v1 + +**Abstract:** +> Traffic signal control (TSC) is an important and widely studied direction. +> Recently, reinforcement learning (RL) methods have been used to solve TSC +> problems and achieve superior performance over conventional TSC methods. +> However, applying RL methods to the real world is challenging due to the huge +> cost of experiments in real-world traffic environments. One possible solution +> is TSC domain adaptation, which adapts trained models to target environments +> and reduces the number of interactions and the training cost. However, existing +> TSC domain adaptation methods still face two major issues: the lack of +> consideration for differences across cities and the low utilization of +> multi-city data. +> To solve aforementioned issues, we propose an approach named Adaptive +> Modularized Model (AMM). By modularizing TSC problems and network models, we +> overcome the challenge of possible changes in environmental observations. We +> also aggregate multi-city experience through meta-learning. We conduct +> extensive experiments on different cities and show that AMM can achieve +> excellent performance with limited interactions in target environments and +> outperform existing methods. We also demonstrate the feasibility and +> generalizability of our method. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on Adaptive Modularized Reinforcement Model for Multi-city Traffic Signal Control, which does not meet the 'MUST' criteria of primarily focusing on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate improving LLM performance through textual input prompt manipulation. + +--- + +## [KM-UNet KAN Mamba UNet for medical image segmentation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02559v1) +**arXiv ID:** http://arxiv.org/abs/2501.02559v1 + +**Abstract:** +> Medical image segmentation is a critical task in medical imaging analysis. +> Traditional CNN-based methods struggle with modeling long-range dependencies, +> while Transformer-based models, despite their success, suffer from quadratic +> computational complexity. To address these limitations, we propose KM-UNet, a +> novel U-shaped network architecture that combines the strengths of +> Kolmogorov-Arnold Networks (KANs) and state-space models (SSMs). KM-UNet +> leverages the Kolmogorov-Arnold representation theorem for efficient feature +> representation and SSMs for scalable long-range modeling, achieving a balance +> between accuracy and computational efficiency. We evaluate KM-UNet on five +> benchmark datasets: ISIC17, ISIC18, CVC, BUSI, and GLAS. Experimental results +> demonstrate that KM-UNet achieves competitive performance compared to +> state-of-the-art methods in medical image segmentation tasks. To the best of +> our knowledge, KM-UNet is the first medical image segmentation framework +> integrating KANs and SSMs. This work provides a valuable baseline and new +> insights for the development of more efficient and interpretable medical image +> segmentation systems. The code is open source at +> https://github.com/2760613195/KM_UNet +> Keywords:KAN,Manba, state-space models,UNet, Medical image segmentation, Deep +> learning + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on medical image segmentation using a novel U-shaped network architecture, which falls under excluded subjects (medical) and does not meet the 'MUST' criteria of focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [Decoding fMRI Data into Captions using Prefix Language Modeling](https://arxiv.org/abs/http://arxiv.org/abs/2501.02570v1) +**arXiv ID:** http://arxiv.org/abs/2501.02570v1 + +**Abstract:** +> With the advancements in Large Language and Latent Diffusion models, brain +> decoding has achieved remarkable results in recent years. The works on the NSD +> dataset, with stimuli images from the COCO dataset, leverage the embeddings +> from the CLIP model for image reconstruction and GIT for captioning. However, +> the current captioning approach introduces the challenge of potential data +> contamination given that the GIT model was trained on the COCO dataset. In this +> work, we present an alternative method for decoding brain signals into image +> captions by predicting a DINOv2 model's embedding of an image from the +> corresponding fMRI signal and then providing its [CLS] token as the prefix to +> the GPT-2 language model which decreases computational requirements +> considerably. Additionally, instead of commonly used Linear Regression, we +> explore 3D Convolutional Neural Network mapping of fMRI signals to image +> embedding space for better accounting positional information of voxels. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on decoding fMRI data into image captions using a combination of computer vision and language models, rather than specifically engineering or optimizing prompts for Large Language Models (LLMs). The use of a GPT-2 model with a predicted prefix does not constitute a primary focus on prompt engineering for text-based interactions with LLMs. + +--- + +## [Evolving Skeletons: Motion Dynamics in Action Recognition](https://arxiv.org/abs/http://arxiv.org/abs/2501.02593v1) +**arXiv ID:** http://arxiv.org/abs/2501.02593v1 + +**Abstract:** +> Skeleton-based action recognition has gained significant attention for its +> ability to efficiently represent spatiotemporal information in a lightweight +> format. Most existing approaches use graph-based models to process skeleton +> sequences, where each pose is represented as a skeletal graph structured around +> human physical connectivity. Among these, the Spatiotemporal Graph +> Convolutional Network (ST-GCN) has become a widely used framework. +> Alternatively, hypergraph-based models, such as the Hyperformer, capture +> higher-order correlations, offering a more expressive representation of complex +> joint interactions. A recent advancement, termed Taylor Videos, introduces +> motion-enhanced skeleton sequences by embedding motion concepts, providing a +> fresh perspective on interpreting human actions in skeleton-based action +> recognition. In this paper, we conduct a comprehensive evaluation of both +> traditional skeleton sequences and Taylor-transformed skeletons using ST-GCN +> and Hyperformer models on the NTU-60 and NTU-120 datasets. We compare skeletal +> graph and hypergraph representations, analyzing static poses against +> motion-injected poses. Our findings highlight the strengths and limitations of +> Taylor-transformed skeletons, demonstrating their potential to enhance motion +> dynamics while exposing current challenges in fully using their benefits. This +> study underscores the need for innovative skeletal modelling techniques to +> effectively handle motion-rich data and advance the field of action +> recognition. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on skeleton-based action recognition, graph/hypergraph models (ST-GCN, Hyperformer), and motion dynamics analysis, with no apparent connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Empowering Bengali Education with AI: Solving Bengali Math Word Problems + through Transformer Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02599v1) +**arXiv ID:** http://arxiv.org/abs/2501.02599v1 + +**Abstract:** +> Mathematical word problems (MWPs) involve the task of converting textual +> descriptions into mathematical equations. This poses a significant challenge in +> natural language processing, particularly for low-resource languages such as +> Bengali. This paper addresses this challenge by developing an innovative +> approach to solving Bengali MWPs using transformer-based models, including +> Basic Transformer, mT5, BanglaT5, and mBART50. To support this effort, the +> "PatiGonit" dataset was introduced, containing 10,000 Bengali math problems, +> and these models were fine-tuned to translate the word problems into equations +> accurately. The evaluation revealed that the mT5 model achieved the highest +> accuracy of 97.30%, demonstrating the effectiveness of transformer models in +> this domain. This research marks a significant step forward in Bengali natural +> language processing, offering valuable methodologies and resources for +> educational AI tools. By improving math education, it also supports the +> development of advanced problem-solving skills for Bengali-speaking students. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses primarily on developing an approach for solving Bengali math word problems using transformer-based models, including fine-tuning these models, rather than engineering or optimizing prompts specifically for Large Language Models (LLMs) to improve their text generation performance." +} + +--- + +## [LLMs Help Alleviate the Cross-Subject Variability in Brain Signal and + Language Alignment](https://arxiv.org/abs/http://arxiv.org/abs/2501.02621v1) +**arXiv ID:** http://arxiv.org/abs/2501.02621v1 + +**Abstract:** +> Decoding human activity from EEG signals has long been a popular research +> topic. While recent studies have increasingly shifted focus from single-subject +> to cross-subject analysis, few have explored the model's ability to perform +> zero-shot predictions on EEG signals from previously unseen subjects. This +> research aims to investigate whether deep learning methods can capture +> subject-independent semantic information inherent in human EEG signals. Such +> insights are crucial for Brain-Computer Interfaces (BCI) because, on one hand, +> they demonstrate the model's robustness against subject-specific temporal +> biases, and on the other, they significantly enhance the generalizability of +> downstream tasks. We employ Large Language Models (LLMs) as denoising agents to +> extract subject-independent semantic features from noisy EEG signals. +> Experimental results, including ablation studies, highlight the pivotal role of +> LLMs in decoding subject-independent semantic information from noisy EEG data. +> We hope our findings will contribute to advancing BCI research and assist both +> academia and industry in applying EEG signals to a broader range of +> applications. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on using LLMs as denoising agents for decoding subject-independent semantic information from EEG signals in Brain-Computer Interfaces (BCI) research, rather than on prompt engineering for text-based interactions with LLMs, failing to meet the core subject criterion." +} + +--- + +## [Trust and Dependability in Blockchain & AI Based MedIoT Applications: + Research Challenges and Future Directions](https://arxiv.org/abs/http://arxiv.org/abs/2501.02647v1) +**arXiv ID:** http://arxiv.org/abs/2501.02647v1 + +**Abstract:** +> This paper critically reviews the integration of Artificial Intelligence (AI) +> and blockchain technologies in the context of Medical Internet of Things +> (MedIoT) applications, where they collectively promise to revolutionize +> healthcare delivery. By examining current research, we underscore AI's +> potential in advancing diagnostics and patient care, alongside blockchain's +> capacity to bolster data security and patient privacy. We focus particularly on +> the imperative to cultivate trust and ensure reliability within these systems. +> Our review highlights innovative solutions for managing healthcare data and +> challenges such as ensuring scalability, maintaining privacy, and promoting +> ethical practices within the MedIoT domain. We present a vision for integrating +> AI-driven insights with blockchain security in healthcare, offering a +> comprehensive review of current research and future directions. We conclude +> with a set of identified research gaps and propose that addressing these is +> crucial for achieving the dependable, secure, and patient -centric MedIoT +> applications of tomorrow. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the integration of AI and blockchain for MedIoT applications, addressing trust, security, and healthcare, which falls under excluded subjects (medical) and does not meet the 'MUST' criteria for focusing on prompt engineering for Large Language Models (LLMs). + +--- + +## [Tighnari: Multi-modal Plant Species Prediction Based on Hierarchical + Cross-Attention Using Graph-Based and Vision Backbone-Extracted Features](https://arxiv.org/abs/http://arxiv.org/abs/2501.02649v1) +**arXiv ID:** http://arxiv.org/abs/2501.02649v1 + +**Abstract:** +> Predicting plant species composition in specific spatiotemporal contexts +> plays an important role in biodiversity management and conservation, as well as +> in improving species identification tools. Our work utilizes 88,987 plant +> survey records conducted in specific spatiotemporal contexts across Europe. We +> also use the corresponding satellite images, time series data, climate time +> series, and other rasterized environmental data such as land cover, human +> footprint, bioclimatic, and soil variables as training data to train the model +> to predict the outcomes of 4,716 plant surveys. We propose a feature +> construction and result correction method based on the graph structure. Through +> comparative experiments, we select the best-performing backbone networks for +> feature extraction in both temporal and image modalities. In this process, we +> built a backbone network based on the Swin-Transformer Block for extracting +> temporal Cubes features. We then design a hierarchical cross-attention +> mechanism capable of robustly fusing features from multiple modalities. During +> training, we adopt a 10-fold cross-fusion method based on fine-tuning and use a +> Threshold Top-K method for post-processing. Ablation experiments demonstrate +> the improvements in model performance brought by our proposed solution +> pipeline. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multi-modal (image, satellite, time series, environmental data) plant species prediction using graph-based and vision backbone-extracted features, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions, failing to meet the primary 'MUST' criteria. + +--- + +## [Tougher Text, Smarter Models: Raising the Bar for Adversarial Defence + Benchmarks](https://arxiv.org/abs/http://arxiv.org/abs/2501.02654v2) +**arXiv ID:** http://arxiv.org/abs/2501.02654v2 + +**Abstract:** +> Recent advancements in natural language processing have highlighted the +> vulnerability of deep learning models to adversarial attacks. While various +> defence mechanisms have been proposed, there is a lack of comprehensive +> benchmarks that evaluate these defences across diverse datasets, models, and +> tasks. In this work, we address this gap by presenting an extensive benchmark +> for textual adversarial defence that significantly expands upon previous work. +> Our benchmark incorporates a wide range of datasets, evaluates state-of-the-art +> defence mechanisms, and extends the assessment to include critical tasks such +> as single-sentence classification, similarity and paraphrase identification, +> natural language inference, and commonsense reasoning. This work not only +> serves as a valuable resource for researchers and practitioners in the field of +> adversarial robustness but also identifies key areas for future research in +> textual adversarial defence. By establishing a new standard for benchmarking in +> this domain, we aim to accelerate progress towards more robust and reliable +> natural language processing systems. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a benchmark for evaluating adversarial defence mechanisms in NLP, rather than primarily on the engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not provide concrete examples of prompts with their impact on LLM output. + +--- + +## [Multi-Aggregator Time-Warping Heterogeneous Graph Neural Network for + Personalized Micro-Video Recommendation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02666v1) +**arXiv ID:** http://arxiv.org/abs/2501.02666v1 + +**Abstract:** +> Micro-video recommendation is attracting global attention and becoming a +> popular daily service for people of all ages. Recently, Graph Neural +> Networks-based micro-video recommendation has displayed performance improvement +> for many kinds of recommendation tasks. However, the existing works fail to +> fully consider the characteristics of micro-videos, such as the high timeliness +> of news nature micro-video recommendation and sequential interactions of +> frequently changed interests. In this paper, a novel Multi-aggregator +> Time-warping Heterogeneous Graph Neural Network (MTHGNN) is proposed for +> personalized news nature micro-video recommendation based on sequential +> sessions, where characteristics of micro-videos are comprehensively studied, +> users' preference is mined via multi-aggregator, the temporal and dynamic +> changes of users' preference are captured, and timeliness is considered. +> Through the comparison with the state-of-the-arts, the experimental results +> validate the superiority of our MTHGNN model. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing a novel Graph Neural Network for personalized micro-video recommendation, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria. + +--- + +## [From Superficial Patterns to Semantic Understanding: Fine-Tuning + Language Models on Contrast Sets](https://arxiv.org/abs/http://arxiv.org/abs/2501.02683v2) +**arXiv ID:** http://arxiv.org/abs/2501.02683v2 + +**Abstract:** +> Large-scale pre-trained language models have demonstrated high performance on +> standard datasets for natural language inference (NLI) tasks. Unfortunately, +> these evaluations can be misleading, as although the models can perform well on +> in-distribution data, they perform poorly on out-of-distribution test sets, +> such as contrast sets. Contrast sets consist of perturbed instances of data +> that have very minor, but meaningful, changes to the input that alter the gold +> label, revealing how models can learn superficial patterns in the training data +> rather than learning more sophisticated language nuances. As an example, the +> ELECTRA-small language model achieves nearly 90% accuracy on an SNLI dataset +> but drops to 75% when tested on an out-of-distribution contrast set. The +> research carried out in this study explores how the robustness of a language +> model can be improved by exposing it to small amounts of more complex contrast +> sets during training to help it better learn language patterns. With this +> approach, the model recovers performance and achieves nearly 90% accuracy on +> contrast sets, highlighting the importance of diverse and challenging training +> data. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on fine-tuning language models with contrast sets for improved robustness, which aligns with developing new training methods, and does not investigate, analyze, or propose methods for improving LLM performance through the manipulation of textual input prompts. + +--- + +## [Horizon Generalization in Reinforcement Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.02709v1) +**arXiv ID:** http://arxiv.org/abs/2501.02709v1 + +**Abstract:** +> We study goal-conditioned RL through the lens of generalization, but not in +> the traditional sense of random augmentations and domain randomization. Rather, +> we aim to learn goal-directed policies that generalize with respect to the +> horizon: after training to reach nearby goals (which are easy to learn), these +> policies should succeed in reaching distant goals (which are quite challenging +> to learn). In the same way that invariance is closely linked with +> generalization is other areas of machine learning (e.g., normalization layers +> make a network invariant to scale, and therefore generalize to inputs of +> varying scales), we show that this notion of horizon generalization is closely +> linked with invariance to planning: a policy navigating towards a goal will +> select the same actions as if it were navigating to a waypoint en route to that +> goal. Thus, such a policy trained to reach nearby goals should succeed at +> reaching arbitrarily-distant goals. Our theoretical analysis proves that both +> horizon generalization and planning invariance are possible, under some +> assumptions. We present new experimental results and recall findings from prior +> work in support of our theoretical results. Taken together, our results open +> the door to studying how techniques for invariance and generalization developed +> in other areas of machine learning might be adapted to achieve this alluring +> property. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on reinforcement learning, goal-conditioned policies, and horizon generalization, with no apparent connection to prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Improved Data Encoding for Emerging Computing Paradigms: From Stochastic + to Hyperdimensional Computing](https://arxiv.org/abs/http://arxiv.org/abs/2501.02715v1) +**arXiv ID:** http://arxiv.org/abs/2501.02715v1 + +**Abstract:** +> Data encoding is a fundamental step in emerging computing paradigms, +> particularly in stochastic computing (SC) and hyperdimensional computing (HDC), +> where it plays a crucial role in determining the overall system performance and +> hardware cost efficiency. This study presents an advanced encoding strategy +> that leverages a hardware-friendly class of low-discrepancy (LD) sequences, +> specifically powers-of-2 bases of Van der Corput (VDC) sequences (VDC-2^n), as +> sources for random number generation. Our approach significantly enhances the +> accuracy and efficiency of SC and HDC systems by addressing challenges +> associated with randomness. By employing LD sequences, we improve correlation +> properties and reduce hardware complexity. Experimental results demonstrate +> significant improvements in accuracy and energy savings for SC and HDC systems. +> Our solution provides a robust framework for integrating SC and HDC in +> resource-constrained environments, paving the way for efficient and scalable AI +> implementations. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance, instead concentrating on data encoding strategies for stochastic and hyperdimensional computing paradigms. + +--- + +## [Tree-based RAG-Agent Recommendation System: A Case Study in Medical Test + Data](https://arxiv.org/abs/http://arxiv.org/abs/2501.02727v1) +**arXiv ID:** http://arxiv.org/abs/2501.02727v1 + +**Abstract:** +> We present HiRMed (Hierarchical RAG-enhanced Medical Test Recommendation), a +> novel tree-structured recommendation system that leverages Retrieval-Augmented +> Generation (RAG) for intelligent medical test recommendations. Unlike +> traditional vector similarity-based approaches, our system performs medical +> reasoning at each tree node through a specialized RAG process. Starting from +> the root node with initial symptoms, the system conducts step-wise medical +> analysis to identify potential underlying conditions and their corresponding +> diagnostic requirements. At each level, instead of simple matching, our +> RAG-enhanced nodes analyze retrieved medical knowledge to understand +> symptom-disease relationships and determine the most appropriate diagnostic +> path. The system dynamically adjusts its recommendation strategy based on +> medical reasoning results, considering factors such as urgency levels and +> diagnostic uncertainty. Experimental results demonstrate that our approach +> achieves superior performance in terms of coverage rate, accuracy, and miss +> rate compared to conventional retrieval-based methods. This work represents a +> significant advance in medical test recommendation by introducing medical +> reasoning capabilities into the traditional tree-based retrieval structure. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on a medical application (medical test recommendation system) and develops a new system architecture (Tree-based RAG-Agent) rather than focusing on prompt engineering for Large Language Models (LLMs) in text-based interactions, failing to meet the required criteria. + +--- + +## [AFed: Algorithmic Fair Federated Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.02732v1) +**arXiv ID:** http://arxiv.org/abs/2501.02732v1 + +**Abstract:** +> Federated Learning (FL) has gained significant attention as it facilitates +> collaborative machine learning among multiple clients without centralizing +> their data on a server. FL ensures the privacy of participating clients by +> locally storing their data, which creates new challenges in fairness. +> Traditional debiasing methods assume centralized access to sensitive +> information, rendering them impractical for the FL setting. Additionally, FL is +> more susceptible to fairness issues than centralized machine learning due to +> the diverse client data sources that may be associated with group information. +> Therefore, training a fair model in FL without access to client local data is +> important and challenging. This paper presents AFed, a straightforward yet +> effective framework for promoting group fairness in FL. The core idea is to +> circumvent restricted data access by learning the global data distribution. +> This paper proposes two approaches: AFed-G, which uses a conditional generator +> trained on the server side, and AFed-GAN, which improves upon AFed-G by +> training a conditional GAN on the client side. We augment the client data with +> the generated samples to help remove bias. Our theoretical analysis justifies +> the proposed methods, and empirical results on multiple real-world datasets +> demonstrate a substantial improvement in AFed over several baselines. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a framework for promoting group fairness in Federated Learning (FL), with no emphasis on Large Language Models (LLMs), prompt engineering, or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [TARDiS : Text Augmentation for Refining Diversity and Separability](https://arxiv.org/abs/http://arxiv.org/abs/2501.02739v1) +**arXiv ID:** http://arxiv.org/abs/2501.02739v1 + +**Abstract:** +> Text augmentation (TA) is a critical technique for text classification, +> especially in few-shot settings. This paper introduces a novel LLM-based TA +> method, TARDiS, to address challenges inherent in the generation and alignment +> stages of two-stage TA methods. For the generation stage, we propose two +> generation processes, SEG and CEG, incorporating multiple class-specific +> prompts to enhance diversity and separability. For the alignment stage, we +> introduce a class adaptation (CA) method to ensure that generated examples +> align with their target classes through verification and modification. +> Experimental results demonstrate TARDiS's effectiveness, outperforming +> state-of-the-art LLM-based TA methods in various few-shot text classification +> tasks. An in-depth analysis confirms the detailed behaviors at each stage. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on a novel text augmentation method (TARDiS) for improving few-shot text classification tasks, using LLMs as a tool, rather than focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) to improve LLM performance through prompt manipulation. + +--- + +## [Interpretable Recognition of Fused Magnesium Furnace Working Conditions + with Deep Convolutional Stochastic Configuration Networks](https://arxiv.org/abs/http://arxiv.org/abs/2501.02740v1) +**arXiv ID:** http://arxiv.org/abs/2501.02740v1 + +**Abstract:** +> To address the issues of a weak generalization capability and +> interpretability in working condition recognition model of a fused magnesium +> furnace, this paper proposes an interpretable working condition recognition +> method based on deep convolutional stochastic configuration networks (DCSCNs). +> Firstly, a supervised learning mechanism is employed to generate physically +> meaningful Gaussian differential convolution kernels. An incremental method is +> utilized to construct a DCSCNs model, ensuring the convergence of recognition +> errors in a hierarchical manner and avoiding the iterative optimization process +> of convolutional kernel parameters using the widely used backpropagation +> algorithm. The independent coefficient of channel feature maps is defined to +> obtain the visualization results of feature class activation maps for the fused +> magnesium furnace. A joint reward function is constructed based on the +> recognition accuracy, the interpretable trustworthiness evaluation metrics, and +> the model parameter quantity. Reinforcement learning (RL) is applied to +> adaptively prune the convolutional kernels of the DCSCNs model, aiming to build +> a compact, highly performed and interpretable network. The experimental results +> demonstrate that the proposed method outperforms the other deep learning +> approaches in terms of recognition accuracy and interpretability. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary focus criteria, as it is centered on developing a deep convolutional stochastic configuration network for recognizing working conditions in a fused magnesium furnace, with no mention of Large Language Models (LLMs), prompt engineering, or text-based interactions. + +--- + +## [Visual Large Language Models for Generalized and Specialized + Applications](https://arxiv.org/abs/http://arxiv.org/abs/2501.02765v1) +**arXiv ID:** http://arxiv.org/abs/2501.02765v1 + +**Abstract:** +> Visual-language models (VLM) have emerged as a powerful tool for learning a +> unified embedding space for vision and language. Inspired by large language +> models, which have demonstrated strong reasoning and multi-task capabilities, +> visual large language models (VLLMs) are gaining increasing attention for +> building general-purpose VLMs. Despite the significant progress made in VLLMs, +> the related literature remains limited, particularly from a comprehensive +> application perspective, encompassing generalized and specialized applications +> across vision (image, video, depth), action, and language modalities. In this +> survey, we focus on the diverse applications of VLLMs, examining their using +> scenarios, identifying ethics consideration and challenges, and discussing +> future directions for their development. By synthesizing these contents, we aim +> to provide a comprehensive guide that will pave the way for future innovations +> and broader applications of VLLMs. The paper list repository is available: +> https://github.com/JackYFL/awesome-VLLMs. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on Visual Large Language Models (VLLMs) for generalized and specialized applications across vision, action, and language modalities, with no primary focus on engineering, design, or optimization of textual input prompts specifically for Large Language Models (LLMs) as required. + +--- + +## [Are GNNs Effective for Multimodal Fault Diagnosis in Microservice + Systems?](https://arxiv.org/abs/http://arxiv.org/abs/2501.02766v1) +**arXiv ID:** http://arxiv.org/abs/2501.02766v1 + +**Abstract:** +> Fault diagnosis in microservice systems has increasingly embraced multimodal +> observation data for a holistic and multifaceted view of the system, with Graph +> Neural Networks (GNNs) commonly employed to model complex service dependencies. +> However, despite the intuitive appeal, there remains a lack of compelling +> justification for the adoption of GNNs, as no direct evidence supports their +> necessity or effectiveness. To critically evaluate the current use of GNNs, we +> propose DiagMLP, a simple topology-agnostic baseline as a substitute for GNNs +> in fault diagnosis frameworks. Through experiments on five public datasets, we +> surprisingly find that DiagMLP performs competitively with and even outperforms +> GNN-based methods in fault diagnosis tasks, indicating that the current +> paradigm of using GNNs to model service dependencies has not yet demonstrated a +> tangible contribution. We further discuss potential reasons for this +> observation and advocate shifting the focus from solely pursuing novel model +> designs to developing challenging datasets, standardizing preprocessing +> protocols, and critically evaluating the utility of advanced deep learning +> modules. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on evaluating Graph Neural Networks (GNNs) for multimodal fault diagnosis in microservice systems, with no apparent connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet the primary 'MUST' criteria. + +--- + +## [Enhancing Trustworthiness of Graph Neural Networks with Rank-Based + Conformal Training](https://arxiv.org/abs/http://arxiv.org/abs/2501.02767v1) +**arXiv ID:** http://arxiv.org/abs/2501.02767v1 + +**Abstract:** +> Graph Neural Networks (GNNs) has been widely used in a variety of fields +> because of their great potential in representing graph-structured data. +> However, lacking of rigorous uncertainty estimations limits their application +> in high-stakes. Conformal Prediction (CP) can produce statistically guaranteed +> uncertainty estimates by using the classifier's probability estimates to obtain +> prediction sets, which contains the true class with a user-specified +> probability. In this paper, we propose a Rank-based CP during training +> framework to GNNs (RCP-GNN) for reliable uncertainty estimates to enhance the +> trustworthiness of GNNs in the node classification scenario. By exploiting rank +> information of the classifier's outcome, prediction sets with desired coverage +> rate can be efficiently constructed. The strategy of CP during training with +> differentiable rank-based conformity loss function is further explored to adapt +> prediction sets according to network topology information. In this way, the +> composition of prediction sets can be guided by the goal of jointly reducing +> inefficiency and probability estimation errors. Extensive experiments on +> several real-world datasets show that our model achieves any pre-defined target +> marginal coverage while significantly reducing the inefficiency compared with +> state-of-the-art methods. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses primarily on enhancing Graph Neural Networks (GNNs) with Conformal Prediction for reliable uncertainty estimates, which does not meet the MUST criteria of focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs)." +} + +--- + +## [ICFNet: Integrated Cross-modal Fusion Network for Survival Prediction](https://arxiv.org/abs/http://arxiv.org/abs/2501.02778v1) +**arXiv ID:** http://arxiv.org/abs/2501.02778v1 + +**Abstract:** +> Survival prediction is a crucial task in the medical field and is essential +> for optimizing treatment options and resource allocation. However, current +> methods often rely on limited data modalities, resulting in suboptimal +> performance. In this paper, we propose an Integrated Cross-modal Fusion Network +> (ICFNet) that integrates histopathology whole slide images, genomic expression +> profiles, patient demographics, and treatment protocols. Specifically, three +> types of encoders, a residual orthogonal decomposition module and a unification +> fusion module are employed to merge multi-modal features to enhance prediction +> accuracy. Additionally, a balanced negative log-likelihood loss function is +> designed to ensure fair training across different patients. Extensive +> experiments demonstrate that our ICFNet outperforms state-of-the-art algorithms +> on five public TCGA datasets, including BLCA, BRCA, GBMLGG, LUAD, and UCEC, and +> shows its potential to support clinical decision-making and advance precision +> medicine. The codes are available at: https://github.com/binging512/ICFNet. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a new network (ICFNet) for survival prediction in the medical field, integrating various data modalities, and does not meet the criteria of focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [Hybrid deep convolution model for lung cancer detection with transfer + learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.02785v1) +**arXiv ID:** http://arxiv.org/abs/2501.02785v1 + +**Abstract:** +> Advances in healthcare research have significantly enhanced our understanding +> of disease mechanisms, diagnostic precision, and therapeutic options. Yet, lung +> cancer remains one of the leading causes of cancer-related mortality worldwide +> due to challenges in early and accurate diagnosis. While current lung cancer +> detection models show promise, there is considerable potential for further +> improving the accuracy for timely intervention. To address this challenge, we +> introduce a hybrid deep convolution model leveraging transfer learning, named +> the Maximum Sensitivity Neural Network (MSNN). MSNN is designed to improve the +> precision of lung cancer detection by refining sensitivity and specificity. +> This model has surpassed existing deep learning approaches through experimental +> validation, achieving an accuracy of 98% and a sensitivity of 97%. By +> overlaying sensitivity maps onto lung Computed Tomography (CT) scans, it +> enables the visualization of regions most indicative of malignant or benign +> classifications. This innovative method demonstrates exceptional performance in +> distinguishing lung cancer with minimal false positives, thereby enhancing the +> accuracy of medical diagnoses. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a new deep convolution model for lung cancer detection, which falls under medical subjects and does not address prompt engineering for Large Language Models (LLMs) as required. + +--- + +## [Segmenting Text and Learning Their Rewards for Improved RLHF in Language + Model](https://arxiv.org/abs/http://arxiv.org/abs/2501.02790v1) +**arXiv ID:** http://arxiv.org/abs/2501.02790v1 + +**Abstract:** +> Reinforcement learning from human feedback (RLHF) has been widely adopted to +> align language models (LMs) with human preference. Prior RLHF works typically +> take a bandit formulation, which, though intuitive, ignores the sequential +> nature of LM generation and can suffer from the sparse reward issue. While +> recent works propose dense token-level RLHF, treating each token as an action +> may be oversubtle to proper reward assignment. In this paper, we seek to get +> the best of both by training and utilizing a segment-level reward model, which +> assigns a reward to each semantically complete text segment that spans over a +> short sequence of tokens. For reward learning, our method allows dynamic text +> segmentation and compatibility with standard sequence-preference datasets. For +> effective RL-based LM training against segment reward, we generalize the +> classical scalar bandit reward normalizers into location-aware normalizer +> functions and interpolate the segment reward for further densification. With +> these designs, our method performs competitively on three popular RLHF +> benchmarks for LM policy: AlpacaEval 2.0, Arena-Hard, and MT-Bench. Ablation +> studies are conducted to further demonstrate our method. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on improving Reinforcement Learning from Human Feedback (RLHF) for Language Models through segment-level reward modeling, rather than specifically investigating, analyzing, or proposing methods for improving LLM performance through the manipulation of textual input prompts." +} + +--- + +## [Enhancing Lifelong Multi-Agent Path Finding with Cache Mechanism](https://arxiv.org/abs/http://arxiv.org/abs/2501.02803v1) +**arXiv ID:** http://arxiv.org/abs/2501.02803v1 + +**Abstract:** +> Multi-Agent Path Finding (MAPF), which focuses on finding collision-free +> paths for multiple robots, is crucial in autonomous warehouse operations. +> Lifelong MAPF (L-MAPF), where agents are continuously reassigned new targets +> upon completing their current tasks, offers a more realistic approximation of +> real-world warehouse scenarios. While cache storage systems can enhance +> efficiency and reduce operational costs, existing approaches primarily rely on +> expectations and mathematical models, often without adequately addressing the +> challenges of multi-robot planning and execution. In this paper, we introduce a +> novel mechanism called Lifelong MAPF with Cache Mechanism (L-MAPF-CM), which +> integrates high-level cache storage with low-level path planning. We have +> involved a new type of map grid called cache for temporary item storage. +> Additionally, we involved a task assigner (TA) with a locking mechanism to +> bridge the gap between the new cache grid and L-MAPF algorithm. The TA +> dynamically allocates target locations to agents based on their status in +> various scenarios. We evaluated L-MAPF-CM using different cache replacement +> policies and task distributions. L-MAPF-CM has demonstrated performance +> improvements particularly with high cache hit rates and smooth traffic +> conditions. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on enhancing Multi-Agent Path Finding with a cache mechanism for autonomous warehouse operations, which does not meet the required focus on prompt engineering, design, or optimization specifically for Large Language Models (LLMs) and their textual input prompts. + +--- + +## [InpDiffusion: Image Inpainting Localization via Conditional Diffusion + Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02816v1) +**arXiv ID:** http://arxiv.org/abs/2501.02816v1 + +**Abstract:** +> As artificial intelligence advances rapidly, particularly with the advent of +> GANs and diffusion models, the accuracy of Image Inpainting Localization (IIL) +> has become increasingly challenging. Current IIL methods face two main +> challenges: a tendency towards overconfidence, leading to incorrect +> predictions; and difficulty in detecting subtle tampering boundaries in +> inpainted images. In response, we propose a new paradigm that treats IIL as a +> conditional mask generation task utilizing diffusion models. Our method, +> InpDiffusion, utilizes the denoising process enhanced by the integration of +> image semantic conditions to progressively refine predictions. During +> denoising, we employ edge conditions and introduce a novel edge supervision +> strategy to enhance the model's perception of edge details in inpainted +> objects. Balancing the diffusion model's stochastic sampling with edge +> supervision of tampered image regions mitigates the risk of incorrect +> predictions from overconfidence and prevents the loss of subtle boundaries that +> can result from overly stochastic processes. Furthermore, we propose an +> innovative Dual-stream Multi-scale Feature Extractor (DMFE) for extracting +> multi-scale features, enhancing feature representation by considering both +> semantic and edge conditions of the inpainted images. Extensive experiments +> across challenging datasets demonstrate that the InpDiffusion significantly +> outperforms existing state-of-the-art methods in IIL tasks, while also +> showcasing excellent generalization capabilities and robustness. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary criteria as it focuses on image inpainting localization using conditional diffusion models, with no apparent connection to Large Language Models (LLMs), prompt engineering, or textual input prompts. The subject matter falls under image generation and processing, which is explicitly excluded. + +--- + +## [Enhanced Rooftop Solar Panel Detection by Efficiently Aggregating Local + Features](https://arxiv.org/abs/http://arxiv.org/abs/2501.02840v1) +**arXiv ID:** http://arxiv.org/abs/2501.02840v1 + +**Abstract:** +> In this paper, we present an enhanced Convolutional Neural Network +> (CNN)-based rooftop solar photovoltaic (PV) panel detection approach using +> satellite images. We propose to use pre-trained CNN-based model to extract the +> local convolutional features of rooftops. These local features are then +> combined using the Vectors of Locally Aggregated Descriptors (VLAD) technique +> to obtain rooftop-level global features, which are then used to train +> traditional Machine Learning (ML) models to identify rooftop images that do and +> do not contain PV panels. On the dataset used in this study, the proposed +> approach achieved rooftop-PV classification scores exceeding the predefined +> threshold of 0.9 across all three cities for each of the feature extractor +> networks evaluated. Moreover, we propose a 3-phase approach to enable efficient +> utilization of the previously trained models on a new city or region with +> limited labelled data. We illustrate the effectiveness of this 3-phase approach +> for multi-city rooftop-PV detection task. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on enhancing rooftop solar panel detection using CNN and ML techniques with satellite images, and does not investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through the manipulation of textual input prompts as required. + +--- + +## [IIMedGPT: Promoting Large Language Model Capabilities of Medical Tasks + by Efficient Human Preference Alignment](https://arxiv.org/abs/http://arxiv.org/abs/2501.02869v1) +**arXiv ID:** http://arxiv.org/abs/2501.02869v1 + +**Abstract:** +> Recent researches of large language models(LLM), which is pre-trained on +> massive general-purpose corpora, have achieved breakthroughs in responding +> human queries. However, these methods face challenges including limited data +> insufficiency to support extensive pre-training and can not align responses +> with users' instructions. To address these issues, we introduce a medical +> instruction dataset, CMedINS, containing six medical instructions derived from +> actual medical tasks, which effectively fine-tunes LLM in conjunction with +> other data. Subsequently, We launch our medical model, IIMedGPT, employing an +> efficient preference alignment method, Direct preference Optimization(DPO). The +> results show that our final model outperforms existing medical models in +> medical dialogue.Datsets, Code and model checkpoints will be released upon +> acceptance. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on fine-tuning a Large Language Model for medical tasks using a new dataset and preference alignment method, which violates the 'MUST NOT' criteria of not focusing on the development of new LLM architectures or training methods, and being primarily concerned with medical subjects. + +--- + +## [Skillful High-Resolution Ensemble Precipitation Forecasting with an + Integrated Deep Learning Framework](https://arxiv.org/abs/http://arxiv.org/abs/2501.02905v1) +**arXiv ID:** http://arxiv.org/abs/2501.02905v1 + +**Abstract:** +> High-resolution precipitation forecasts are crucial for providing accurate +> weather prediction and supporting effective responses to extreme weather +> events. Traditional numerical models struggle with stochastic subgrid-scale +> processes, while recent deep learning models often produce blurry results. To +> address these challenges, we propose a physics-inspired deep learning framework +> for high-resolution (0.05\textdegree{} $\times$ 0.05\textdegree{}) ensemble +> precipitation forecasting. Trained on ERA5 and CMPA high-resolution +> precipitation datasets, the framework integrates deterministic and +> probabilistic components. The deterministic model, based on a 3D +> SwinTransformer, captures average precipitation at mesoscale resolution and +> incorporates strategies to enhance performance, particularly for moderate to +> heavy rainfall. The probabilistic model employs conditional diffusion in latent +> space to account for uncertainties in residual precipitation at convective +> scales. During inference, ensemble members are generated by repeatedly sampling +> latent variables, enabling the model to represent precipitation uncertainty. +> Our model significantly enhances spatial resolution and forecast accuracy. Rank +> histogram shows that the ensemble system is reliable and unbiased. In a case +> study of heavy precipitation in southern China, the model outputs align more +> closely with observed precipitation distributions than ERA5, demonstrating +> superior capability in capturing extreme precipitation events. Additionally, +> 5-day real-time forecasts show good performance in terms of CSI scores. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a deep learning framework for high-resolution precipitation forecasting, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet the primary criteria. + +--- + +## [Unsupervised Tomato Split Anomaly Detection using Hyperspectral Imaging + and Variational Autoencoders](https://arxiv.org/abs/http://arxiv.org/abs/2501.02921v1) +**arXiv ID:** http://arxiv.org/abs/2501.02921v1 + +**Abstract:** +> Tomato anomalies/damages pose a significant challenge in greenhouse farming. +> While this method of cultivation benefits from efficient resource utilization, +> anomalies can significantly degrade the quality of farm produce. A common +> anomaly associated with tomatoes is splitting, characterized by the development +> of cracks on the tomato skin, which degrades its quality. Detecting this type +> of anomaly is challenging due to dynamic variations in appearance and sizes, +> compounded by dataset scarcity. We address this problem in an unsupervised +> manner by utilizing a tailored variational autoencoder (VAE) with hyperspectral +> input. Preliminary analysis of the dataset enabled us to select the optimal +> range of wavelengths for detecting this anomaly. Our findings indicate that the +> 530nm - 550nm range is suitable for identifying tomato dry splits. The analysis +> on reconstruction loss allow us to not only detect the anomalies but also to +> some degree estimate the anomalous regions. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet any of the 'MUST' criteria: it focuses on anomaly detection in hyperspectral imaging using Variational Autoencoders, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus falling entirely outside the specified scope. + +--- + +## [GLFC: Unified Global-Local Feature and Contrast Learning with + Mamba-Enhanced UNet for Synthetic CT Generation from CBCT](https://arxiv.org/abs/http://arxiv.org/abs/2501.02992v2) +**arXiv ID:** http://arxiv.org/abs/2501.02992v2 + +**Abstract:** +> Generating synthetic Computed Tomography (CT) images from Cone Beam Computed +> Tomography (CBCT) is desirable for improving the image quality of CBCT. +> Existing synthetic CT (sCT) generation methods using Convolutional Neural +> Networks (CNN) and Transformers often face difficulties in effectively +> capturing both global and local features and contrasts for high-quality sCT +> generation. In this work, we propose a Global-Local Feature and Contrast +> learning (GLFC) framework for sCT generation. First, a Mamba-Enhanced UNet +> (MEUNet) is introduced by integrating Mamba blocks into the skip connections of +> a high-resolution UNet for effective global and local feature learning. Second, +> we propose a Multiple Contrast Loss (MCL) that calculates synthetic loss at +> different intensity windows to improve quality for both soft tissues and bone +> regions. Experiments on the SynthRAD2023 dataset demonstrate that GLFC improved +> the SSIM of sCT from 77.91% to 91.50% compared with the original CBCT, and +> significantly outperformed several existing methods for sCT generation. The +> code is available at https://github.com/HiLab-git/GLFC + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on generating synthetic CT images from CBCT using a novel CNN framework (GLFC with Mamba-Enhanced UNet), which does not meet the 'MUST' criteria of focusing on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts for improving LLM performance." +} + +--- + +## [Quality Estimation based Feedback Training for Improving Pronoun + Translation](https://arxiv.org/abs/http://arxiv.org/abs/2501.03008v1) +**arXiv ID:** http://arxiv.org/abs/2501.03008v1 + +**Abstract:** +> Pronoun translation is a longstanding challenge in neural machine translation +> (NMT), often requiring inter-sentential context to ensure linguistic accuracy. +> To address this, we introduce ProNMT, a novel framework designed to enhance +> pronoun and overall translation quality in context-aware machine translation +> systems. ProNMT leverages Quality Estimation (QE) models and a unique Pronoun +> Generation Likelihood-Based Feedback mechanism to iteratively fine-tune +> pre-trained NMT models without relying on extensive human annotations. The +> framework combines QE scores with pronoun-specific rewards to guide training, +> ensuring improved handling of linguistic nuances. Extensive experiments +> demonstrate significant gains in pronoun translation accuracy and general +> translation quality across multiple metrics. ProNMT offers an efficient, +> scalable, and context-aware approach to improving NMT systems, particularly in +> translating context-dependent elements like pronouns. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on improving neural machine translation (NMT) systems through Quality Estimation based feedback training, rather than engineering or optimizing prompts specifically for Large Language Models (LLMs), and does not provide concrete examples of prompts or demonstrate their impact on LLM output. + +--- + +## [Quantization Meets Reasoning: Exploring LLM Low-Bit Quantization + Degradation for Mathematical Reasoning](https://arxiv.org/abs/http://arxiv.org/abs/2501.03035v1) +**arXiv ID:** http://arxiv.org/abs/2501.03035v1 + +**Abstract:** +> Large language models have achieved significant advancements in complex +> mathematical reasoning benchmarks, such as MATH. However, their substantial +> computational requirements present challenges for practical deployment. Model +> quantization has emerged as an effective strategy to reduce memory usage and +> computational costs by employing lower precision and bit-width representations. +> In this study, we systematically evaluate the impact of quantization on +> mathematical reasoning tasks. We introduce a multidimensional evaluation +> framework that qualitatively assesses specific capability dimensions and +> conduct quantitative analyses on the step-by-step outputs of various +> quantization methods. Our results demonstrate that quantization differentially +> affects numerical computation and reasoning planning abilities, identifying key +> areas where quantized models experience performance degradation. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on model quantization for reducing computational costs and its impact on mathematical reasoning tasks, rather than prompt engineering for Large Language Models (LLMs), not meeting the 'MUST' criteria for prompt engineering, LLM performance improvement through prompt manipulation, and providing concrete prompt examples." +} + +--- + +## [Piano Transcription by Hierarchical Language Modeling with Pretrained + Roll-based Encoders](https://arxiv.org/abs/http://arxiv.org/abs/2501.03038v2) +**arXiv ID:** http://arxiv.org/abs/2501.03038v2 + +**Abstract:** +> Automatic Music Transcription (AMT), aiming to get musical notes from raw +> audio, typically uses frame-level systems with piano-roll outputs or language +> model (LM)-based systems with note-level predictions. However, frame-level +> systems require manual thresholding, while the LM-based systems struggle with +> long sequences. In this paper, we propose a hybrid method combining pre-trained +> roll-based encoders with an LM decoder to leverage the strengths of both +> methods. Besides, our approach employs a hierarchical prediction strategy, +> first predicting onset and pitch, then velocity, and finally offset. The +> hierarchical prediction strategy reduces computational costs by breaking down +> long sequences into different hierarchies. Evaluated on two benchmark +> roll-based encoders, our method outperforms traditional piano-roll outputs 0.01 +> and 0.022 in onset-offset-velocity F1 score, demonstrating its potential as a +> performance-enhancing plug-in for arbitrary roll-based music transcription +> encoder. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on Automatic Music Transcription using a hybrid method combining pre-trained roll-based encoders with a language model decoder, rather than prompt engineering for Large Language Models (LLMs). The subject is outside the specified scope, particularly excluding applications of generative AI other than text generation driven by LLMs, such as audio generation in this case. + +--- + +## [Single-Channel Distance-Based Source Separation for Mobile GPU in + Outdoor and Indoor Environments](https://arxiv.org/abs/http://arxiv.org/abs/2501.03045v1) +**arXiv ID:** http://arxiv.org/abs/2501.03045v1 + +**Abstract:** +> This study emphasizes the significance of exploring distance-based source +> separation (DSS) in outdoor environments. Unlike existing studies that +> primarily focus on indoor settings, the proposed model is designed to capture +> the unique characteristics of outdoor audio sources. It incorporates advanced +> techniques, including a two-stage conformer block, a linear relation-aware +> self-attention (RSA), and a TensorFlow Lite GPU delegate. While the linear RSA +> may not capture physical cues as explicitly as the quadratic RSA, the linear +> RSA enhances the model's context awareness, leading to improved performance on +> the DSS that requires an understanding of physical cues in outdoor and indoor +> environments. The experimental results demonstrated that the proposed model +> overcomes the limitations of existing approaches and considerably enhances +> energy efficiency and real-time inference speed on mobile devices. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on source separation for audio signals in outdoor and indoor environments, utilizing mobile GPU and TensorFlow, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Personalized Fashion Recommendation with Image Attributes and Aesthetics + Assessment](https://arxiv.org/abs/http://arxiv.org/abs/2501.03085v1) +**arXiv ID:** http://arxiv.org/abs/2501.03085v1 + +**Abstract:** +> Personalized fashion recommendation is a difficult task because 1) the +> decisions are highly correlated with users' aesthetic appetite, which previous +> work frequently overlooks, and 2) many new items are constantly rolling out +> that cause strict cold-start problems in the popular identity (ID)-based +> recommendation methods. These new items are critical to recommend because of +> trend-driven consumerism. In this work, we aim to provide more accurate +> personalized fashion recommendations and solve the cold-start problem by +> converting available information, especially images, into two attribute graphs +> focusing on optimized image utilization and noise-reducing user modeling. +> Compared with previous methods that separate image and text as two components, +> the proposed method combines image and text information to create a richer +> attributes graph. Capitalizing on the advancement of large language and vision +> models, we experiment with extracting fine-grained attributes efficiently and +> as desired using two different prompts. Preliminary experiments on the IQON3000 +> dataset have shown that the proposed method achieves competitive accuracy +> compared with baselines. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on personalized fashion recommendation using image attributes and aesthetics, combining image and text information, rather than specifically engineering or optimizing prompts for Large Language Models (LLMs). While it mentions using two different prompts for attribute extraction, prompt engineering is not the central concern, but rather a minor component within the larger system. + +--- + +## [LangFair: A Python Package for Assessing Bias and Fairness in Large + Language Model Use Cases](https://arxiv.org/abs/http://arxiv.org/abs/2501.03112v1) +**arXiv ID:** http://arxiv.org/abs/2501.03112v1 + +**Abstract:** +> Large Language Models (LLMs) have been observed to exhibit bias in numerous +> ways, potentially creating or worsening outcomes for specific groups identified +> by protected attributes such as sex, race, sexual orientation, or age. To help +> address this gap, we introduce LangFair, an open-source Python package that +> aims to equip LLM practitioners with the tools to evaluate bias and fairness +> risks relevant to their specific use cases. The package offers functionality to +> easily generate evaluation datasets, comprised of LLM responses to +> use-case-specific prompts, and subsequently calculate applicable metrics for +> the practitioner's use case. To guide in metric selection, LangFair offers an +> actionable decision framework. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on assessing bias and fairness in Large Language Models using a Python package, rather than on the engineering, design, or optimization of prompts specifically for LLMs to improve their text generation performance. + +--- + +## [From Models to Network Topologies: A Topology Inference Attack in + Decentralized Federated Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.03119v1) +**arXiv ID:** http://arxiv.org/abs/2501.03119v1 + +**Abstract:** +> Federated Learning (FL) is widely recognized as a privacy-preserving machine +> learning paradigm due to its model-sharing mechanism that avoids direct data +> exchange. However, model training inevitably leaves exploitable traces that can +> be used to infer sensitive information. In Decentralized FL (DFL), the overlay +> topology significantly influences its models' convergence, robustness, and +> security. This study explores the feasibility of inferring the overlay topology +> of DFL systems based solely on model behavior, introducing a novel Topology +> Inference Attack. A taxonomy of topology inference attacks is proposed, +> categorizing them by the attacker's capabilities and knowledge. Practical +> attack strategies are developed for different scenarios, and quantitative +> experiments are conducted to identify key factors influencing the attack +> effectiveness. Experimental results demonstrate that analyzing only the public +> models of individual nodes can accurately infer the DFL topology, underscoring +> the risk of sensitive information leakage in DFL systems. This finding offers +> valuable insights for improving privacy preservation in decentralized learning +> environments. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on topology inference attacks in Decentralized Federated Learning, which does not meet the MUST criteria: it does not focus on engineering, design, or optimization of prompts for Large Language Models (LLMs), nor does it investigate improving LLM performance through textual input prompts." +} + +--- + +## [PRMBench: A Fine-grained and Challenging Benchmark for Process-Level + Reward Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03124v2) +**arXiv ID:** http://arxiv.org/abs/2501.03124v2 + +**Abstract:** +> Process-level Reward Models (PRMs) are crucial for complex reasoning and +> decision-making tasks, where each intermediate step plays an important role in +> the reasoning process. Since language models are prone to various types of +> errors during the reasoning process, PRMs are required to possess nuanced +> capabilities for detecting various implicit error types in real-world +> scenarios. However, current benchmarks primarily focus on step correctness, +> failing to evaluate PRMs' performance systematically. To address this gap, we +> introduce PRMBench, a process-level benchmark specifically designed to assess +> the fine-grained error detection capabilities of PRMs. PRMBench comprises 6,216 +> carefully designed problems and 83,456 step-level labels, evaluating models +> across multiple dimensions, including simplicity, soundness, and sensitivity. +> In our experiments on 15 models, spanning both open-source PRMs and +> closed-source large language models prompted as critic models, we uncover +> significant weaknesses in current PRMs. These findings underscore the +> challenges inherent in process-level evaluation and highlight key directions +> for future research. We hope PRMBench can be a robust bench for advancing +> research on PRM evaluation and development. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a benchmark for evaluating Process-Level Reward Models (PRMs) and their error detection capabilities, rather than engineering, designing, or optimizing prompts specifically for Large Language Models (LLMs). While LLMs are mentioned as being used as critic models, prompt engineering is not the central concern. + +--- + +## [Geometry Restoration and Dewarping of Camera-Captured Document Images](https://arxiv.org/abs/http://arxiv.org/abs/2501.03145v2) +**arXiv ID:** http://arxiv.org/abs/2501.03145v2 + +**Abstract:** +> This research focuses on developing a method for restoring the topology of +> digital images of paper documents captured by a camera, using algorithms for +> detection, segmentation, geometry restoration, and dewarping. Our methodology +> employs deep learning (DL) for document outline detection, followed by computer +> vision (CV) to create a topological 2D grid using cubic polynomial +> interpolation and correct nonlinear distortions by remapping the image. Using +> classical CV methods makes the document topology restoration process more +> efficient and faster, as it requires significantly fewer computational +> resources and memory. We developed a new pipeline for automatic document +> dewarping and reconstruction, along with a framework and annotated dataset to +> demonstrate its efficiency. Our experiments confirm the promise of our +> methodology and its superiority over existing benchmarks (including mobile apps +> and popular DL solutions, such as RectiNet, DocGeoNet, and DocTr++) both +> visually and in terms of document readability via Optical Character Recognition +> (OCR) and geometry restoration metrics. This paves the way for creating +> high-quality digital copies of paper documents and enhancing the efficiency of +> OCR systems. Project page: https://github.com/HorizonParadox/DRCCBI + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a method for restoring and dewarping document images using computer vision and deep learning, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Automated Generation of Challenging Multiple-Choice Questions for Vision + Language Model Evaluation](https://arxiv.org/abs/http://arxiv.org/abs/2501.03225v1) +**arXiv ID:** http://arxiv.org/abs/2501.03225v1 + +**Abstract:** +> The rapid development of vision language models (VLMs) demands rigorous and +> reliable evaluation. However, current visual question answering (VQA) +> benchmarks often depend on open-ended questions, making accurate evaluation +> difficult due to the variability in natural language responses. To address +> this, we introduce AutoConverter, an agentic framework that automatically +> converts these open-ended questions into multiple-choice format, enabling +> objective evaluation while reducing the costly question creation process. Our +> experiments demonstrate that AutoConverter can generate correct and challenging +> multiple-choice questions, with VLMs demonstrating consistently similar or +> lower accuracy on these questions compared to human-created ones. Using +> AutoConverter, we construct VMCBench, a benchmark created by transforming 20 +> existing VQA datasets into a unified multiple-choice format, totaling 9,018 +> questions. We comprehensively evaluate 33 state-of-the-art VLMs on VMCBench, +> setting a new standard for scalable, consistent, and reproducible VLM +> evaluation. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on evaluating Vision Language Models (VLMs) through automated generation of multiple-choice questions, rather than specifically engineering or optimizing prompts for Large Language Models (LLMs) to improve text generation, as required by the criteria. + +--- + +## [LightGNN: Simple Graph Neural Network for Recommendation](https://arxiv.org/abs/http://arxiv.org/abs/2501.03228v2) +**arXiv ID:** http://arxiv.org/abs/2501.03228v2 + +**Abstract:** +> Graph neural networks (GNNs) have demonstrated superior performance in +> collaborative recommendation through their ability to conduct high-order +> representation smoothing, effectively capturing structural information within +> users' interaction patterns. However, existing GNN paradigms face significant +> challenges in scalability and robustness when handling large-scale, noisy, and +> real-world datasets. To address these challenges, we present LightGNN, a +> lightweight and distillation-based GNN pruning framework designed to +> substantially reduce model complexity while preserving essential collaboration +> modeling capabilities. Our LightGNN framework introduces a computationally +> efficient pruning module that adaptively identifies and removes redundant edges +> and embedding entries for model compression. The framework is guided by a +> resource-friendly hierarchical knowledge distillation objective, whose +> intermediate layer augments the observed graph to maintain performance, +> particularly in high-rate compression scenarios. Extensive experiments on +> public datasets demonstrate LightGNN's effectiveness, significantly improving +> both computational efficiency and recommendation accuracy. Notably, LightGNN +> achieves an 80% reduction in edge count and 90% reduction in embedding entries +> while maintaining performance comparable to more complex state-of-the-art +> baselines. The implementation of our LightGNN framework is available at the +> github repository: https://github.com/HKUDS/LightGNN. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a lightweight Graph Neural Network (GNN) for recommendation systems, with no mention of Large Language Models (LLMs), prompt engineering, or textual input manipulation, thus failing to meet all 'MUST' criteria. + +--- + +## [Gaussian Masked Autoencoders](https://arxiv.org/abs/http://arxiv.org/abs/2501.03229v1) +**arXiv ID:** http://arxiv.org/abs/2501.03229v1 + +**Abstract:** +> This paper explores Masked Autoencoders (MAE) with Gaussian Splatting. While +> reconstructive self-supervised learning frameworks such as MAE learns good +> semantic abstractions, it is not trained for explicit spatial awareness. Our +> approach, named Gaussian Masked Autoencoder, or GMAE, aims to learn semantic +> abstractions and spatial understanding jointly. Like MAE, it reconstructs the +> image end-to-end in the pixel space, but beyond MAE, it also introduces an +> intermediate, 3D Gaussian-based representation and renders images via +> splatting. We show that GMAE can enable various zero-shot learning capabilities +> of spatial understanding (e.g., figure-ground segmentation, image layering, +> edge detection, etc.) while preserving the high-level semantics of +> self-supervised representation quality from MAE. To our knowledge, we are the +> first to employ Gaussian primitives in an image representation learning +> framework beyond optimization-based single-scene reconstructions. We believe +> GMAE will inspire further research in this direction and contribute to +> developing next-generation techniques for modeling high-fidelity visual data. +> More details at https://brjathu.github.io/gmae + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a new image representation learning framework (Gaussian Masked Autoencoders) for spatial understanding and visual data modeling, with no mention of Large Language Models (LLMs), prompt engineering, or text-based interactions, thus failing to meet all 'MUST' criteria. + +--- + +## [Backdoor Token Unlearning: Exposing and Defending Backdoors in + Pretrained Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03272v1) +**arXiv ID:** http://arxiv.org/abs/2501.03272v1 + +**Abstract:** +> Supervised fine-tuning has become the predominant method for adapting large +> pretrained models to downstream tasks. However, recent studies have revealed +> that these models are vulnerable to backdoor attacks, where even a small number +> of malicious samples can successfully embed backdoor triggers into the model. +> While most existing defense methods focus on post-training backdoor defense, +> efficiently defending against backdoor attacks during training phase remains +> largely unexplored. To address this gap, we propose a novel defense method +> called Backdoor Token Unlearning (BTU), which proactively detects and +> neutralizes trigger tokens during the training stage. Our work is based on two +> key findings: 1) backdoor learning causes distinctive differences between +> backdoor token parameters and clean token parameters in word embedding layers, +> and 2) the success of backdoor attacks heavily depends on backdoor token +> parameters. The BTU defense leverages these properties to identify aberrant +> embedding parameters and subsequently removes backdoor behaviors using a +> fine-grained unlearning technique. Extensive evaluations across three datasets +> and four types of backdoor attacks demonstrate that BTU effectively defends +> against these threats while preserving the model's performance on primary +> tasks. Our code is available at https://github.com/XDJPH/BTU. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on defending against backdoor attacks in pretrained language models during the training phase, rather than prompt engineering for text-based interactions with Large Language Models (LLMs), failing to meet the 'MUST' criteria related to prompt engineering and manipulation of textual input prompts. + +--- + +## [Strategic Fusion Optimizes Transformer Compression](https://arxiv.org/abs/http://arxiv.org/abs/2501.03273v1) +**arXiv ID:** http://arxiv.org/abs/2501.03273v1 + +**Abstract:** +> This study investigates transformer model compression by systematically +> pruning its layers. We evaluated 14 pruning strategies across nine diverse +> datasets, including 12 strategies based on different signals obtained from +> layer activations, mutual information, gradients, weights, and attention. To +> address the limitations of single-signal strategies, we introduced two fusion +> strategies, linear regression and random forest, which combine individual +> strategies (i.e., strategic fusion), for more informed pruning decisions. +> Additionally, we applied knowledge distillation to mitigate any accuracy loss +> during layer pruning. Our results reveal that random forest strategic fusion +> outperforms individual strategies in seven out of nine datasets and achieves +> near-optimal performance in the other two. The distilled random forest +> surpasses the original accuracy in six datasets and mitigates accuracy drops in +> the remaining three. Knowledge distillation also improves the accuracy-to-size +> ratio by an average factor of 18.84 across all datasets. Supported by +> mathematical foundations and biological analogies, our findings suggest that +> strategically combining multiple signals can lead to efficient, high-performing +> transformer models for resource-constrained applications. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on transformer model compression through layer pruning and knowledge distillation, not on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [ComMer: a Framework for Compressing and Merging User Data for + Personalization](https://arxiv.org/abs/http://arxiv.org/abs/2501.03276v1) +**arXiv ID:** http://arxiv.org/abs/2501.03276v1 + +**Abstract:** +> Large Language Models (LLMs) excel at a wide range of tasks, but adapting +> them to new data, particularly for personalized applications, poses significant +> challenges due to resource and computational constraints. Existing methods +> either rely on exposing fresh data to the model through the prompt, which is +> limited by context size and computationally expensive at inference time, or +> fine-tuning, which incurs substantial training and update costs. In this paper, +> we introduce ComMer - Compress and Merge - a novel framework that efficiently +> personalizes LLMs by compressing users' documents into compact representations, +> which are then merged and fed into a frozen LLM. We evaluate ComMer on two +> types of personalization tasks - personalized skill learning, using the tweet +> paraphrasing dataset and the personalized news headline generation dataset from +> the LaMP benchmark, and knowledge-intensive, using the PerLTQA dataset. Our +> experiments demonstrate that in constrained inference budget scenarios ComMer +> achieves superior quality in skill learning tasks, while highlighting +> limitations in knowledge-intensive settings due to the loss of detailed +> information. These results offer insights into trade-offs and potential +> optimizations in multi-document compression for personalization. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a framework for compressing and merging user data to personalize Large Language Models (LLMs), rather than on the engineering, design, or optimization of prompts specifically for LLMs. It does not investigate or propose methods for improving LLM performance through the manipulation of textual input prompts. + +--- + +## [CodeVision: Detecting LLM-Generated Code Using 2D Token Probability Maps + and Vision Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03288v1) +**arXiv ID:** http://arxiv.org/abs/2501.03288v1 + +**Abstract:** +> The rise of large language models (LLMs) like ChatGPT has significantly +> improved automated code generation, enhancing software development efficiency. +> However, this introduces challenges in academia, particularly in distinguishing +> between human-written and LLM-generated code, which complicates issues of +> academic integrity. Existing detection methods, such as pre-trained models and +> watermarking, face limitations in adaptability and computational efficiency. In +> this paper, we propose a novel detection method using 2D token probability maps +> combined with vision models, preserving spatial code structures such as +> indentation and brackets. By transforming code into log probability matrices +> and applying vision models like Vision Transformers (ViT) and ResNet, we +> capture both content and structure for more accurate detection. Our method +> shows robustness across multiple programming languages and improves upon +> traditional detectors, offering a scalable and computationally efficient +> solution for identifying LLM-generated code. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on detecting LLM-generated code using vision models, rather than on the engineering, design, or optimization of prompts for Large Language Models (LLMs), failing to meet the primary criteria of focusing on prompt engineering for text-based interactions with LLMs. + +--- + +## [Analyzing Bias in Swiss Federal Supreme Court Judgments Using Facebook's + Holistic Bias Dataset: Implications for Language Model Training](https://arxiv.org/abs/http://arxiv.org/abs/2501.03324v1) +**arXiv ID:** http://arxiv.org/abs/2501.03324v1 + +**Abstract:** +> Natural Language Processing (NLP) is vital for computers to process and +> respond accurately to human language. However, biases in training data can +> introduce unfairness, especially in predicting legal judgment. This study +> focuses on analyzing biases within the Swiss Judgment Prediction Dataset +> (SJP-Dataset). Our aim is to ensure unbiased factual descriptions essential for +> fair decision making by NLP models in legal contexts. We analyze the dataset +> using social bias descriptors from the Holistic Bias dataset and employ +> advanced NLP techniques, including attention visualization, to explore the +> impact of dispreferred descriptors on model predictions. The study identifies +> biases and examines their influence on model behavior. Challenges include +> dataset imbalance and token limits affecting model performance. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on analyzing biases in a legal dataset for fair NLP model training, not on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not provide concrete examples of prompt manipulation impacting LLM output. + +--- + +## [Existential Crisis: A Social Robot's Reason for Being](https://arxiv.org/abs/http://arxiv.org/abs/2501.03376v1) +**arXiv ID:** http://arxiv.org/abs/2501.03376v1 + +**Abstract:** +> As Robots become ever more important in our daily lives there's growing need +> for understanding how they're perceived by people. This study aims to +> investigate how the user perception of robots is influenced by displays of +> personality. Using LLMs and speech to text technology, we designed a +> within-subject study to compare two conditions: a personality-driven robot and +> a purely task-oriented, personality-neutral robot. Twelve participants, +> recruited from Socially Intelligent Robotics course at Vrije Universiteit +> Amsterdam, interacted with a robot Nao tasked with asking them a set of medical +> questions under both conditions. After completing both interactions, the +> participants completed a user experience questionnaire measuring their +> emotional states and robot perception using standardized questionnaires from +> the SRI and Psychology literature. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on user perception of robots with displayed personality, utilizing LLMs as a tool, but prompt engineering for Large Language Models is not the primary concern; the core subject revolves around social robotics and user experience, not the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Over-the-Air Fair Federated Learning via Multi-Objective Optimization](https://arxiv.org/abs/http://arxiv.org/abs/2501.03392v1) +**arXiv ID:** http://arxiv.org/abs/2501.03392v1 + +**Abstract:** +> In federated learning (FL), heterogeneity among the local dataset +> distributions of clients can result in unsatisfactory performance for some, +> leading to an unfair model. To address this challenge, we propose an +> over-the-air fair federated learning algorithm (OTA-FFL), which leverages +> over-the-air computation to train fair FL models. By formulating FL as a +> multi-objective minimization problem, we introduce a modified Chebyshev +> approach to compute adaptive weighting coefficients for gradient aggregation in +> each communication round. To enable efficient aggregation over the multiple +> access channel, we derive analytical solutions for the optimal transmit scalars +> at the clients and the de-noising scalar at the parameter server. Extensive +> experiments demonstrate the superiority of OTA-FFL in achieving fairness and +> robust performance compared to existing methods. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on federated learning, multi-objective optimization, and over-the-air computation, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Enhanced Importance Sampling through Latent Space Exploration in + Normalizing Flows](https://arxiv.org/abs/http://arxiv.org/abs/2501.03394v1) +**arXiv ID:** http://arxiv.org/abs/2501.03394v1 + +**Abstract:** +> Importance sampling is a rare event simulation technique used in Monte Carlo +> simulations to bias the sampling distribution towards the rare event of +> interest. By assigning appropriate weights to sampled points, importance +> sampling allows for more efficient estimation of rare events or tails of +> distributions. However, importance sampling can fail when the proposal +> distribution does not effectively cover the target distribution. In this work, +> we propose a method for more efficient sampling by updating the proposal +> distribution in the latent space of a normalizing flow. Normalizing flows learn +> an invertible mapping from a target distribution to a simpler latent +> distribution. The latent space can be more easily explored during the search +> for a proposal distribution, and samples from the proposal distribution are +> recovered in the space of the target distribution via the invertible mapping. +> We empirically validate our methodology on simulated robotics applications such +> as autonomous racing and aircraft ground collision avoidance. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the criteria as it focuses on enhancing importance sampling in Monte Carlo simulations using normalizing flows, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing all 'MUST' criteria. + +--- + +## [BoundingDocs: a Unified Dataset for Document Question Answering with + Spatial Annotations](https://arxiv.org/abs/http://arxiv.org/abs/2501.03403v1) +**arXiv ID:** http://arxiv.org/abs/2501.03403v1 + +**Abstract:** +> We present a unified dataset for document Question-Answering (QA), which is +> obtained combining several public datasets related to Document AI and visually +> rich document understanding (VRDU). Our main contribution is twofold: on the +> one hand we reformulate existing Document AI tasks, such as Information +> Extraction (IE), into a Question-Answering task, making it a suitable resource +> for training and evaluating Large Language Models; on the other hand, we +> release the OCR of all the documents and include the exact position of the +> answer to be found in the document image as a bounding box. Using this dataset, +> we explore the impact of different prompting techniques (that might include +> bounding box information) on the performance of open-weight models, identifying +> the most effective approaches for document comprehension. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "Although the paper mentions exploring the impact of different prompting techniques on LLM performance, its primary focus is on presenting a unified dataset for document Question Answering, making prompt engineering a secondary concern rather than the central focus." +} + +--- + +## [SALT: Sales Autocompletion Linked Business Tables Dataset](https://arxiv.org/abs/http://arxiv.org/abs/2501.03413v1) +**arXiv ID:** http://arxiv.org/abs/2501.03413v1 + +**Abstract:** +> Foundation models, particularly those that incorporate Transformer +> architectures, have demonstrated exceptional performance in domains such as +> natural language processing and image processing. Adapting these models to +> structured data, like tables, however, introduces significant challenges. These +> difficulties are even more pronounced when addressing multi-table data linked +> via foreign key, which is prevalent in the enterprise realm and crucial for +> empowering business use cases. Despite its substantial impact, research +> focusing on such linked business tables within enterprise settings remains a +> significantly important yet underexplored domain. To address this, we introduce +> a curated dataset sourced from an Enterprise Resource Planning (ERP) system, +> featuring extensive linked tables. This dataset is specifically designed to +> support research endeavors in table representation learning. By providing +> access to authentic enterprise data, our goal is to potentially enhance the +> effectiveness and applicability of models for real-world business contexts. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the 'MUST' criteria, as its primary focus is on introducing a dataset for table representation learning in enterprise settings, rather than engineering, designing, or optimizing prompts specifically for Large Language Models (LLMs), and does not demonstrate the impact of textual input prompts on LLM output." +} + +--- + +## [Optimization Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.03443v1) +**arXiv ID:** http://arxiv.org/abs/2501.03443v1 + +**Abstract:** +> This article introduces the concept of optimization learning, a methodology +> to design optimization proxies that learn the input/output mapping of +> parametric optimization problems. These optimization proxies are trustworthy by +> design: they compute feasible solutions to the underlying optimization +> problems, provide quality guarantees on the returned solutions, and scale to +> large instances. Optimization proxies are differentiable programs that combine +> traditional deep learning technology with repair or completion layers to +> produce feasible solutions. The article shows that optimization proxies can be +> trained end-to-end in a self-supervised way. It presents methodologies to +> provide performance guarantees and to scale optimization proxies to large-scale +> optimization problems. The potential of optimization proxies is highlighted +> through applications in power systems and, in particular, real-time risk +> assessment and security-constrained optimal power flow. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering, design, or optimization for Large Language Models (LLMs). Instead, it introduces 'optimization learning' for solving parametric optimization problems, unrelated to LLMs or text-based interactions. + +--- + +## [LHGNN: Local-Higher Order Graph Neural Networks For Audio Classification + and Tagging](https://arxiv.org/abs/http://arxiv.org/abs/2501.03464v1) +**arXiv ID:** http://arxiv.org/abs/2501.03464v1 + +**Abstract:** +> Transformers have set new benchmarks in audio processing tasks, leveraging +> self-attention mechanisms to capture complex patterns and dependencies within +> audio data. However, their focus on pairwise interactions limits their ability +> to process the higher-order relations essential for identifying distinct audio +> objects. To address this limitation, this work introduces the Local- Higher +> Order Graph Neural Network (LHGNN), a graph based model that enhances feature +> understanding by integrating local neighbourhood information with higher-order +> data from Fuzzy C-Means clusters, thereby capturing a broader spectrum of audio +> relationships. Evaluation of the model on three publicly available audio +> datasets shows that it outperforms Transformer-based models across all +> benchmarks while operating with substantially fewer parameters. Moreover, LHGNN +> demonstrates a distinct advantage in scenarios lacking ImageNet pretraining, +> establishing its effectiveness and efficiency in environments where extensive +> pretraining data is unavailable. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a new graph neural network model (LHGNN) for audio classification and tagging, primarily dealing with audio data and disregarding Large Language Models (LLMs) and text-based prompt engineering, thus failing to meet all the 'MUST' criteria. + +--- + +## [MTRAG: A Multi-Turn Conversational Benchmark for Evaluating + Retrieval-Augmented Generation Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.03468v1) +**arXiv ID:** http://arxiv.org/abs/2501.03468v1 + +**Abstract:** +> Retrieval-augmented generation (RAG) has recently become a very popular task +> for Large Language Models (LLMs). Evaluating them on multi-turn RAG +> conversations, where the system is asked to generate a response to a question +> in the context of a preceding conversation is an important and often overlooked +> task with several additional challenges. We present MTRAG: an end-to-end +> human-generated multi-turn RAG benchmark that reflects several real-world +> properties across diverse dimensions for evaluating the full RAG pipeline. +> MTRAG contains 110 conversations averaging 7.7 turns each across four domains +> for a total of 842 tasks. We also explore automation paths via synthetic data +> and LLM-as-a-Judge evaluation. Our human and automatic evaluations show that +> even state-of-the-art LLM RAG systems struggle on MTRAG. We demonstrate the +> need for strong retrieval and generation systems that can handle later turns, +> unanswerable questions, non-standalone questions, and multiple domains. MTRAG +> is available at https://github.com/ibm/mt-rag-benchmark. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on introducing a benchmark (MTRAG) for evaluating Retrieval-Augmented Generation Systems, rather than specifically focusing on the engineering, design, or optimization of prompts for Large Language Models (LLMs) and demonstrating their impact on LLM output. + +--- + +## [Reading with Intent -- Neutralizing Intent](https://arxiv.org/abs/http://arxiv.org/abs/2501.03475v1) +**arXiv ID:** http://arxiv.org/abs/2501.03475v1 + +**Abstract:** +> Queries to large language models (LLMs) can be divided into two parts: the +> instruction/question and the accompanying context. The context for +> retrieval-augmented generation (RAG) systems in most benchmarks comes from +> Wikipedia or Wikipedia-like texts which are written in a neutral and factual +> tone. However, when RAG systems retrieve internet-based content, they encounter +> text with diverse tones and linguistic styles, introducing challenges for +> downstream tasks. The Reading with Intent task addresses this issue by +> evaluating how varying tones in context passages affect model performance. +> Building on prior work that focused on sarcasm, we extend this paradigm by +> constructing a dataset where context passages are transformed to $11$ distinct +> emotions using a better synthetic data generation approach. Using this dataset, +> we train an emotion translation model to systematically adapt passages to +> specified emotional tones. The human evaluation shows that the LLM fine-tuned +> to become the emotion-translator benefited from the synthetically generated +> data. Finally, the emotion-translator is used in the Reading with Intent task +> to transform the passages to a neutral tone. By neutralizing the passages, it +> mitigates the challenges posed by sarcastic passages and improves overall +> results on this task by about $3\%$. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on training an emotion translation model to adapt passages to specified emotional tones and fine-tuning an LLM for emotion translation, rather than the engineering, design, or optimization of prompts specifically for LLMs. While it mentions improving results on the Reading with Intent task by neutralizing passages, the core subject is emotion adaptation in LLM training, not prompt engineering for text-based interactions with LLMs." +} + +--- + +## [Align-Pro: A Principled Approach to Prompt Optimization for LLM + Alignment](https://arxiv.org/abs/http://arxiv.org/abs/2501.03486v1) +**arXiv ID:** http://arxiv.org/abs/2501.03486v1 + +**Abstract:** +> The alignment of large language models (LLMs) with human values is critical +> as these models become increasingly integrated into various societal and +> decision-making processes. Traditional methods, such as reinforcement learning +> from human feedback (RLHF), achieve alignment by fine-tuning model parameters, +> but these approaches are often computationally expensive and impractical when +> models are frozen or inaccessible for parameter modification. In contrast, +> prompt optimization is a viable alternative to RLHF for LLM alignment. While +> the existing literature has shown empirical promise of prompt optimization, its +> theoretical underpinning remains under-explored. We address this gap by +> formulating prompt optimization as an optimization problem and try to provide +> theoretical insights into the optimality of such a framework. To analyze the +> performance of the prompt optimization, we study theoretical suboptimality +> bounds and provide insights in terms of how prompt optimization depends upon +> the given prompter and target model. We also provide empirical validation +> through experiments on various datasets, demonstrating that prompt optimization +> can effectively align LLMs, even when parameter fine-tuning is not feasible. + +**Decision Explanation:** +Original decision: REJECT +Although the paper discusses prompt optimization for LLM alignment, its primary focus is on LLM alignment through a theoretical optimization framework, rather than specifically on engineering, design, or optimization of prompts for improving LLM performance through textual input manipulation, as required. + +--- + +## [Can Deep Learning Trigger Alerts from Mobile-Captured Images?](https://arxiv.org/abs/http://arxiv.org/abs/2501.03499v1) +**arXiv ID:** http://arxiv.org/abs/2501.03499v1 + +**Abstract:** +> Our research presents a comprehensive approach to leveraging mobile camera +> image data for real-time air quality assessment and recommendation. We develop +> a regression-based Convolutional Neural Network model and tailor it explicitly +> for air quality prediction by exploiting the inherent relationship between +> output parameters. As a result, the Mean Squared Error of 0.0077 and 0.0112 +> obtained for 2 and 5 pollutants respectively outperforms existing models. +> Furthermore, we aim to verify the common practice of augmenting the original +> dataset with a view to introducing more variation in the training phase. It is +> one of our most significant contributions that our experimental results +> demonstrate minimal accuracy differences between the original and augmented +> datasets. Finally, a real-time, user-friendly dashboard is implemented which +> dynamically displays the Air Quality Index and pollutant values derived from +> captured mobile camera images. Users' health conditions are considered to +> recommend whether a location is suitable based on current air quality metrics. +> Overall, this research contributes to verification of data augmentation +> techniques, CNN-based regression modelling for air quality prediction, and +> user-centric air quality monitoring through mobile technology. The proposed +> system offers practical solutions for individuals to make informed +> environmental health and well-being decisions. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on leveraging mobile camera images for air quality assessment using a Convolutional Neural Network model, with no mention of Large Language Models (LLMs), prompt engineering, or text-based interactions, thereby failing to meet all 'MUST' criteria. + +--- + +## [Vocal Tract Length Warped Features for Spoken Keyword Spotting](https://arxiv.org/abs/http://arxiv.org/abs/2501.03523v1) +**arXiv ID:** http://arxiv.org/abs/2501.03523v1 + +**Abstract:** +> In this paper, we propose several methods that incorporate vocal tract length +> (VTL) warped features for spoken keyword spotting (KWS). The first method, +> VTL-independent KWS, involves training a single deep neural network (DNN) that +> utilizes VTL features with various warping factors. During training, a specific +> VTL feature is randomly selected per epoch, allowing the exploration of VTL +> variations. During testing, the VTL features with different warping factors of +> a test utterance are scored against the DNN and combined with equal weight. In +> the second method scores the conventional features of a test utterance (without +> VTL warping) against the DNN. The third method, VTL-concatenation KWS, +> concatenates VTL warped features to form high-dimensional features for KWS. +> Evaluations carried out on the English Google Command dataset demonstrate that +> the proposed methods improve the accuracy of KWS. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on spoken keyword spotting using vocal tract length warped features with deep neural networks, which does not meet the criteria of primarily focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or investigating methods for improving LLM performance through textual input prompt manipulation. + +--- + +## [Deep Learning within Tabular Data: Foundations, Challenges, Advances and + Future Directions](https://arxiv.org/abs/http://arxiv.org/abs/2501.03540v1) +**arXiv ID:** http://arxiv.org/abs/2501.03540v1 + +**Abstract:** +> Tabular data remains one of the most prevalent data types across a wide range +> of real-world applications, yet effective representation learning for this +> domain poses unique challenges due to its irregular patterns, heterogeneous +> feature distributions, and complex inter-column dependencies. This survey +> provides a comprehensive review of state-of-the-art techniques in tabular data +> representation learning, structured around three foundational design elements: +> training data, neural architectures, and learning objectives. Unlike prior +> surveys that focus primarily on either architecture design or learning +> strategies, we adopt a holistic perspective that emphasizes the universality +> and robustness of representation learning methods across diverse downstream +> tasks. We examine recent advances in data augmentation and generation, +> specialized neural network architectures tailored to tabular data, and +> innovative learning objectives that enhance representation quality. +> Additionally, we highlight the growing influence of self-supervised learning +> and the adaptation of transformer-based foundation models for tabular data. Our +> review is based on a systematic literature search using rigorous inclusion +> criteria, encompassing 127 papers published since 2020 in top-tier conferences +> and journals. Through detailed analysis and comparison, we identify emerging +> trends, critical gaps, and promising directions for future research, aiming to +> guide the development of more generalizable and effective tabular data +> representation methods. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on tabular data representation learning using deep learning techniques, with no apparent connection to Large Language Models (LLMs) or prompt engineering for text-based interactions, thus failing to meet the primary 'MUST' criteria. + +--- + +## [PromptGuard: Soft Prompt-Guided Unsafe Content Moderation for + Text-to-Image Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03544v1) +**arXiv ID:** http://arxiv.org/abs/2501.03544v1 + +**Abstract:** +> Text-to-image (T2I) models have been shown to be vulnerable to misuse, +> particularly in generating not-safe-for-work (NSFW) content, raising serious +> ethical concerns. In this work, we present PromptGuard, a novel content +> moderation technique that draws inspiration from the system prompt mechanism in +> large language models (LLMs) for safety alignment. Unlike LLMs, T2I models lack +> a direct interface for enforcing behavioral guidelines. Our key idea is to +> optimize a safety soft prompt that functions as an implicit system prompt +> within the T2I model's textual embedding space. This universal soft prompt (P*) +> directly moderates NSFW inputs, enabling safe yet realistic image generation +> without altering the inference efficiency or requiring proxy models. Extensive +> experiments across three datasets demonstrate that PromptGuard effectively +> mitigates NSFW content generation while preserving high-quality benign outputs. +> PromptGuard achieves 7.8 times faster than prior content moderation methods, +> surpassing eight state-of-the-art defenses with an optimal unsafe ratio down to +> 5.84%. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on text-to-image models and content moderation, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) as required, violating the 'MUST NOT' criteria concerning applications of generative AI other than text generation driven by LLMs. + +--- + +## [Rethinking Adversarial Attacks in Reinforcement Learning from Policy + Distribution Perspective](https://arxiv.org/abs/http://arxiv.org/abs/2501.03562v2) +**arXiv ID:** http://arxiv.org/abs/2501.03562v2 + +**Abstract:** +> Deep Reinforcement Learning (DRL) suffers from uncertainties and inaccuracies +> in the observation signal in realworld applications. Adversarial attack is an +> effective method for evaluating the robustness of DRL agents. However, existing +> attack methods targeting individual sampled actions have limited impacts on the +> overall policy distribution, particularly in continuous action spaces. To +> address these limitations, we propose the Distribution-Aware Projected Gradient +> Descent attack (DAPGD). DAPGD uses distribution similarity as the gradient +> perturbation input to attack the policy network, which leverages the entire +> policy distribution rather than relying on individual samples. We utilize the +> Bhattacharyya distance in DAPGD to measure policy similarity, enabling +> sensitive detection of subtle but critical differences between probability +> distributions. Our experiment results demonstrate that DAPGD achieves SOTA +> results compared to the baselines in three robot navigation tasks, achieving an +> average 22.03% higher reward drop compared to the best baseline. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria: it focuses on Reinforcement Learning, adversarial attacks, and policy distribution, with no primary emphasis on engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate improving LLM performance through textual input prompt manipulation. + +--- + +## [Applying Large Language Models in Knowledge Graph-based Enterprise + Modeling: Challenges and Opportunities](https://arxiv.org/abs/http://arxiv.org/abs/2501.03566v1) +**arXiv ID:** http://arxiv.org/abs/2501.03566v1 + +**Abstract:** +> The role of large language models (LLMs) in enterprise modeling has recently +> started to shift from academic research to that of industrial applications. +> Thereby, LLMs represent a further building block for the machine-supported +> generation of enterprise models. In this paper we employ a knowledge +> graph-based approach for enterprise modeling and investigate the potential +> benefits of LLMs in this context. In addition, the findings of an expert survey +> and ChatGPT-4o-based experiments demonstrate that LLM-based model generations +> exhibit minimal variability, yet remain constrained to specific tasks, with +> reliability declining for more intricate tasks. The survey results further +> suggest that the supervision and intervention of human modeling experts are +> essential to ensure the accuracy and integrity of the generated models. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is on applying LLMs in enterprise modeling using knowledge graphs, not on prompt engineering for text-based interactions with LLMs, failing to meet the core subject requirement." +} + +--- + +## [RecKG: Knowledge Graph for Recommender Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.03598v1) +**arXiv ID:** http://arxiv.org/abs/2501.03598v1 + +**Abstract:** +> Knowledge graphs have proven successful in integrating heterogeneous data +> across various domains. However, there remains a noticeable dearth of research +> on their seamless integration among heterogeneous recommender systems, despite +> knowledge graph-based recommender systems garnering extensive research +> attention. This study aims to fill this gap by proposing RecKG, a standardized +> knowledge graph for recommender systems. RecKG ensures the consistent +> representation of entities across different datasets, accommodating diverse +> attribute types for effective data integration. Through a meticulous +> examination of various recommender system datasets, we select attributes for +> RecKG, ensuring standardized formatting through consistent naming conventions. +> By these characteristics, RecKG can seamlessly integrate heterogeneous data +> sources, enabling the discovery of additional semantic information within the +> integrated knowledge graph. We apply RecKG to standardize real-world datasets, +> subsequently developing an application for RecKG using a graph database. +> Finally, we validate RecKG's achievement in interoperability through a +> qualitative evaluation between RecKG and other studies. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on integrating knowledge graphs for recommender systems, not on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria 1 and 2. + +--- + +## [MHGNet: Multi-Heterogeneous Graph Neural Network for Traffic Prediction](https://arxiv.org/abs/http://arxiv.org/abs/2501.03635v1) +**arXiv ID:** http://arxiv.org/abs/2501.03635v1 + +**Abstract:** +> In recent years, traffic flow prediction has played a crucial role in the +> management of intelligent transportation systems. However, traditional +> forecasting methods often model non-Euclidean low-dimensional traffic data as a +> simple graph with single-type nodes and edges, failing to capture similar +> trends among nodes of the same type. To address this limitation, this paper +> proposes MHGNet, a novel framework for modeling spatiotemporal +> multi-heterogeneous graphs. Within this framework, the STD Module decouples +> single-pattern traffic data into multi-pattern traffic data through feature +> mappings of timestamp embedding matrices and node embedding matrices. +> Subsequently, the Node Clusterer leverages the Euclidean distance between nodes +> and different types of limit points to perform clustering with O(N) time +> complexity. The nodes within each cluster undergo residual subgraph convolution +> within the spatiotemporal fusion subgraphs generated by the DSTGG Module, +> followed by processing in the SIE Module for node repositioning and +> redistribution of weights. To validate the effectiveness of MHGNet, this paper +> conducts extensive ablation studies and quantitative evaluations on four widely +> used benchmarks, demonstrating its superior performance. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary focus criteria, as it concentrates on traffic prediction using Multi-Heterogeneous Graph Neural Networks, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to satisfy the 'MUST' criteria. + +--- + +## [Effective and Efficient Mixed Precision Quantization of Speech + Foundation Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03643v2) +**arXiv ID:** http://arxiv.org/abs/2501.03643v2 + +**Abstract:** +> This paper presents a novel mixed-precision quantization approach for speech +> foundation models that tightly integrates mixed-precision learning and +> quantized model parameter estimation into one single model compression stage. +> Experiments conducted on LibriSpeech dataset with fine-tuned wav2vec2.0-base +> and HuBERT-large models suggest the resulting mixed-precision quantized models +> increased the lossless compression ratio by factors up to 1.7x and 1.9x over +> the respective uniform-precision and two-stage mixed-precision quantized +> baselines that perform precision learning and model parameters quantization in +> separate and disjointed stages, while incurring no statistically word error +> rate (WER) increase over the 32-bit full-precision models. The system +> compression time of wav2vec2.0-base and HuBERT-large models is reduced by up to +> 1.9 and 1.5 times over the two-stage mixed-precision baselines, while both +> produce lower WERs. The best-performing 3.5-bit mixed-precision quantized +> HuBERT-large model produces a lossless compression ratio of 8.6x over the +> 32-bit full-precision system. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on model compression through mixed-precision quantization of speech foundation models, not on prompt engineering or the manipulation of textual input prompts for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria." +} + +--- + +## [A Diversity-Enhanced Knowledge Distillation Model for Practical Math + Word Problem Solving](https://arxiv.org/abs/http://arxiv.org/abs/2501.03670v1) +**arXiv ID:** http://arxiv.org/abs/2501.03670v1 + +**Abstract:** +> Math Word Problem (MWP) solving is a critical task in natural language +> processing, has garnered significant research interest in recent years. Various +> recent studies heavily rely on Seq2Seq models and their extensions (e.g., +> Seq2Tree and Graph2Tree) to generate mathematical equations. While effective, +> these models struggle to generate diverse but counterpart solution equations, +> limiting their generalization across various math problem scenarios. In this +> paper, we introduce a novel Diversity-enhanced Knowledge Distillation (DivKD) +> model for practical MWP solving. Our approach proposes an adaptive diversity +> distillation method, in which a student model learns diverse equations by +> selectively transferring high-quality knowledge from a teacher model. +> Additionally, we design a diversity prior-enhanced student model to better +> capture the diversity distribution of equations by incorporating a conditional +> variational auto-encoder. Extensive experiments on {four} MWP benchmark +> datasets demonstrate that our approach achieves higher answer accuracy than +> strong baselines while maintaining high efficiency for practical applications. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses primarily on developing a new model (Diversity-enhanced Knowledge Distillation) for math word problem solving, rather than engineering or optimizing prompts specifically for Large Language Models (LLMs). Prompt engineering is not the central concern, and the paper does not demonstrate the impact of textual input prompts on LLM output." +} + +--- + +## [SALE-Based Offline Reinforcement Learning with Ensemble Q-Networks](https://arxiv.org/abs/http://arxiv.org/abs/2501.03676v2) +**arXiv ID:** http://arxiv.org/abs/2501.03676v2 + +**Abstract:** +> In this work, we build upon the offline reinforcement learning algorithm TD7, +> which incorporates State-Action Learned Embeddings (SALE) and a prioritized +> experience replay buffer (LAP). We propose a model-free actor-critic algorithm +> that integrates ensemble Q-networks and a gradient diversity penalty from EDAC. +> The ensemble Q-networks introduce penalties to guide the actor network toward +> in-distribution actions, effectively addressing the challenge of +> out-of-distribution actions. Meanwhile, the gradient diversity penalty +> encourages diverse Q-value gradients, further suppressing overestimation for +> out-of-distribution actions. Additionally, our method retains an adjustable +> behavior cloning (BC) term that directs the actor network toward dataset +> actions during early training stages, while gradually reducing its influence as +> the precision of the Q-ensemble improves. These enhancements work +> synergistically to improve the stability and precision of the training. +> Experimental results on the D4RL MuJoCo benchmarks demonstrate that our +> algorithm achieves higher convergence speed, stability, and performance +> compared to existing methods. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on offline reinforcement learning with ensemble Q-networks, disregarding the mandatory criteria of concentrating on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and manipulating textual input prompts to improve LLM performance. + +--- + +## [SLAM: Towards Efficient Multilingual Reasoning via Selective Language + Alignment](https://arxiv.org/abs/http://arxiv.org/abs/2501.03681v1) +**arXiv ID:** http://arxiv.org/abs/2501.03681v1 + +**Abstract:** +> Despite the significant improvements achieved by large language models (LLMs) +> in English reasoning tasks, these models continue to struggle with multilingual +> reasoning. Recent studies leverage a full-parameter and two-stage training +> paradigm to teach models to first understand non-English questions and then +> reason. However, this method suffers from both substantial computational +> resource computing and catastrophic forgetting. The fundamental cause is that, +> with the primary goal of enhancing multilingual comprehension, an excessive +> number of irrelevant layers and parameters are tuned during the first stage. +> Given our findings that the representation learning of languages is merely +> conducted in lower-level layers, we propose an efficient multilingual reasoning +> alignment approach that precisely identifies and fine-tunes the layers +> responsible for handling multilingualism. Experimental results show that our +> method, SLAM, only tunes 6 layers' feed-forward sub-layers including 6.5-8% of +> all parameters within 7B and 13B LLMs, achieving superior average performance +> than all strong baselines across 10 languages. Meanwhile, SLAM only involves +> one training stage, reducing training time by 4.1-11.9 compared to the +> two-stage method. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on efficient multilingual reasoning via selective language alignment, involving a one-stage training method for LLMs, which aligns more with the development of new training methods (violation of MUST NOT 1) rather than the engineering, design, or optimization of prompts for Large Language Models. + +--- + +## [Exploring Molecule Generation Using Latent Space Graph Diffusion](https://arxiv.org/abs/http://arxiv.org/abs/2501.03696v1) +**arXiv ID:** http://arxiv.org/abs/2501.03696v1 + +**Abstract:** +> Generating molecular graphs is a challenging task due to their discrete +> nature and the competitive objectives involved. Diffusion models have emerged +> as SOTA approaches in data generation across various modalities. For molecular +> graphs, graph neural networks (GNNs) as a diffusion backbone have achieved +> impressive results. Latent space diffusion, where diffusion occurs in a +> low-dimensional space via an autoencoder, has demonstrated computational +> efficiency. However, the literature on latent space diffusion for molecular +> graphs is scarce, and no commonly accepted best practices exist. In this work, +> we explore different approaches and hyperparameters, contrasting generative +> flow models (denoising diffusion, flow matching, heat dissipation) and +> architectures (GNNs and E(3)-equivariant GNNs). Our experiments reveal a high +> sensitivity to the choice of approach and design decisions. Code is made +> available at +> github.com/Prashanth-Pombala/Molecule-Generation-using-Latent-Space-Graph-Diffusion. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on generating molecular graphs using latent space graph diffusion, involving graph neural networks and diffusion models, without any mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet all 'MUST' criteria." +} + +--- + +## [Unsupervised Speech Segmentation: A General Approach Using Speech + Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03711v1) +**arXiv ID:** http://arxiv.org/abs/2501.03711v1 + +**Abstract:** +> In this paper, we introduce an unsupervised approach for Speech Segmentation, +> which builds on previously researched approaches, e.g., Speaker Diarization, +> while being applicable to an inclusive set of acoustic-semantic distinctions, +> paving a path towards a general Unsupervised Speech Segmentation approach. +> Unlike traditional speech and audio segmentation, which mainly focuses on +> spectral changes in the input signal, e.g., phone segmentation, our approach +> tries to segment the spoken utterance into chunks with differing +> acoustic-semantic styles, focusing on acoustic-semantic information that does +> not translate well into text, e.g., emotion or speaker. While most Speech +> Segmentation tasks only handle one style change, e.g., emotion diarization, our +> approach tries to handle multiple acoustic-semantic style changes. Leveraging +> recent advances in Speech Language Models (SLMs), we propose a simple +> unsupervised method to segment a given speech utterance. We empirically +> demonstrate the effectiveness of the proposed approach by considering several +> setups. Results suggest that the proposed method is superior to the evaluated +> baselines on boundary detection, segment purity, and over-segmentation. Code is +> available at +> https://github.com/avishaiElmakies/unsupervised_speech_segmentation_using_slm. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on unsupervised speech segmentation using Speech Language Models (SLMs), which does not meet the criteria of primarily focusing on the engineering, design, or optimization of prompts for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts for improving LLM performance. + +--- + +## [Materialist: Physically Based Editing Using Single-Image Inverse + Rendering](https://arxiv.org/abs/http://arxiv.org/abs/2501.03717v1) +**arXiv ID:** http://arxiv.org/abs/2501.03717v1 + +**Abstract:** +> To perform image editing based on single-view, inverse physically based +> rendering, we present a method combining a learning-based approach with +> progressive differentiable rendering. Given an image, our method leverages +> neural networks to predict initial material properties. Progressive +> differentiable rendering is then used to optimize the environment map and +> refine the material properties with the goal of closely matching the rendered +> result to the input image. We require only a single image while other inverse +> rendering methods based on the rendering equation require multiple views. In +> comparison to single-view methods that rely on neural renderers, our approach +> achieves more realistic light material interactions, accurate shadows, and +> global illumination. Furthermore, with optimized material properties and +> illumination, our method enables a variety of tasks, including physically based +> material editing, object insertion, and relighting. We also propose a method +> for material transparency editing that operates effectively without requiring +> full scene geometry. Compared with methods based on Stable Diffusion, our +> approach offers stronger interpretability and more realistic light refraction +> based on empirical results. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on image editing using single-image inverse physically based rendering, involving neural networks and differentiable rendering for image generation, which falls under image generation and not text-based interactions with Large Language Models (LLMs), thus failing to meet the primary criteria of focusing on prompt engineering for LLMs. + +--- + +## [SelectiveFinetuning: Enhancing Transfer Learning in Sleep Staging + through Selective Domain Alignment](https://arxiv.org/abs/http://arxiv.org/abs/2501.03764v1) +**arXiv ID:** http://arxiv.org/abs/2501.03764v1 + +**Abstract:** +> In practical sleep stage classification, a key challenge is the variability +> of EEG data across different subjects and environments. Differences in +> physiology, age, health status, and recording conditions can lead to domain +> shifts between data. These domain shifts often result in decreased model +> accuracy and reliability, particularly when the model is applied to new data +> with characteristics different from those it was originally trained on, which +> is a typical manifestation of negative transfer. To address this, we propose +> SelectiveFinetuning in this paper. Our method utilizes a pretrained Multi +> Resolution Convolutional Neural Network (MRCNN) to extract EEG features, +> capturing the distinctive characteristics of different sleep stages. To +> mitigate the effect of domain shifts, we introduce a domain aligning mechanism +> that employs Earth Mover Distance (EMD) to evaluate and select source domain +> data closely matching the target domain. By finetuning the model with selective +> source data, our SelectiveFinetuning enhances the model's performance on target +> domain that exhibits domain shifts compared to the data used for training. +> Experimental results show that our method outperforms existing baselines, +> offering greater robustness and adaptability in practical scenarios where data +> distributions are often unpredictable. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on enhancing transfer learning in sleep staging through selective domain alignment using a Convolutional Neural Network (CNN), not on the engineering, design, or optimization of prompts for Large Language Models (LLMs), and does not investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Three-dimensional attention Transformer for state evaluation in + real-time strategy games](https://arxiv.org/abs/http://arxiv.org/abs/2501.03832v1) +**arXiv ID:** http://arxiv.org/abs/2501.03832v1 + +**Abstract:** +> Situation assessment in Real-Time Strategy (RTS) games is crucial for +> understanding decision-making in complex adversarial environments. However, +> existing methods remain limited in processing multi-dimensional feature +> information and temporal dependencies. Here we propose a tri-dimensional +> Space-Time-Feature Transformer (TSTF Transformer) architecture, which +> efficiently models battlefield situations through three independent but +> cascaded modules: spatial attention, temporal attention, and feature attention. +> On a dataset comprising 3,150 adversarial experiments, the 8-layer TSTF +> Transformer demonstrates superior performance: achieving 58.7% accuracy in the +> early game (~4% progress), significantly outperforming the conventional +> Timesformer's 41.8%; reaching 97.6% accuracy in the mid-game (~40% progress) +> while maintaining low performance variation (standard deviation 0.114). +> Meanwhile, this architecture requires fewer parameters (4.75M) compared to the +> baseline model (5.54M). Our study not only provides new insights into situation +> assessment in RTS games but also presents an innovative paradigm for +> Transformer-based multi-dimensional temporal modeling. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a new Transformer architecture for situation assessment in Real-Time Strategy games, rather than on prompt engineering for Large Language Models (LLMs). It does not investigate, analyze, or propose methods for improving LLM performance through textual input prompt manipulation, and there is no mention of prompts or LLMs in the provided abstract. + +--- + +## [SCC-YOLO: An Improved Object Detector for Assisting in Brain Tumor + Diagnosis](https://arxiv.org/abs/http://arxiv.org/abs/2501.03836v2) +**arXiv ID:** http://arxiv.org/abs/2501.03836v2 + +**Abstract:** +> Brain tumors can result in neurological dysfunction, alterations in cognitive +> and psychological states, increased intracranial pressure, and the occurrence +> of seizures, thereby presenting a substantial risk to human life and health. +> The You Only Look Once(YOLO) series models have demonstrated superior accuracy +> in object detection for medical imaging. In this paper, we develop a novel +> SCC-YOLO architecture by integrating the SCConv attention mechanism into +> YOLOv9. The SCConv module reconstructs an efficient convolutional module by +> reducing spatial and channel redundancy among features, thereby enhancing the +> learning of image features. We investigate the impact of intergrating different +> attention mechanisms with the YOLOv9 model on brain tumor image detection using +> both the Br35H dataset and our self-made dataset(Brain_Tumor_Dataset). +> Experimental results show that on the Br35H dataset, SCC-YOLO achieved a 0.3% +> improvement in mAp50 compared to YOLOv9, while on our self-made dataset, +> SCC-YOLO exhibited a 0.5% improvement over YOLOv9. SCC-YOLO has reached +> state-of-the-art performance in brain tumor detection. Source code is available +> at : https://jihulab.com/healthcare-information-studio/SCC-YOLO/-/tree/master + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a new object detection architecture (SCC-YOLO) for medical imaging (brain tumor diagnosis) and does not investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through the manipulation of textual input prompts. + +--- + +## [Explainable Reinforcement Learning via Temporal Policy Decomposition](https://arxiv.org/abs/http://arxiv.org/abs/2501.03902v1) +**arXiv ID:** http://arxiv.org/abs/2501.03902v1 + +**Abstract:** +> We investigate the explainability of Reinforcement Learning (RL) policies +> from a temporal perspective, focusing on the sequence of future outcomes +> associated with individual actions. In RL, value functions compress information +> about rewards collected across multiple trajectories and over an infinite +> horizon, allowing a compact form of knowledge representation. However, this +> compression obscures the temporal details inherent in sequential +> decision-making, presenting a key challenge for interpretability. We present +> Temporal Policy Decomposition (TPD), a novel explainability approach that +> explains individual RL actions in terms of their Expected Future Outcome (EFO). +> These explanations decompose generalized value functions into a sequence of +> EFOs, one for each time step up to a prediction horizon of interest, revealing +> insights into when specific outcomes are expected to occur. We leverage +> fixed-horizon temporal difference learning to devise an off-policy method for +> learning EFOs for both optimal and suboptimal actions, enabling contrastive +> explanations consisting of EFOs for different state-action pairs. Our +> experiments demonstrate that TPD generates accurate explanations that (i) +> clarify the policy's future strategy and anticipated trajectory for a given +> action and (ii) improve understanding of the reward composition, facilitating +> fine-tuning of the reward function to align with human expectations. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on explainability in Reinforcement Learning (RL) policies, primarily dealing with temporal policy decomposition and value functions, with no indication of prompt engineering for Large Language Models (LLMs) or manipulation of textual input prompts to improve LLM performance." +} + +--- + +## [Exploring the Potential of Large Language Models in Public + Transportation: San Antonio Case Study](https://arxiv.org/abs/http://arxiv.org/abs/2501.03904v1) +**arXiv ID:** http://arxiv.org/abs/2501.03904v1 + +**Abstract:** +> The integration of large language models (LLMs) into public transit systems +> presents a transformative opportunity to enhance urban mobility. This study +> explores the potential of LLMs to revolutionize public transportation +> management within the context of San Antonio's transit system. Leveraging the +> capabilities of LLMs in natural language processing and data analysis, we +> investigate their capabilities to optimize route planning, reduce wait times, +> and provide personalized travel assistance. By utilizing the General Transit +> Feed Specification (GTFS) and other relevant data, this research aims to +> demonstrate how LLMs can potentially improve resource allocation, elevate +> passenger satisfaction, and inform data-driven decision-making in transit +> operations. A comparative analysis of different ChatGPT models was conducted to +> assess their ability to understand transportation information, retrieve +> relevant data, and provide comprehensive responses. Findings from this study +> suggest that while LLMs hold immense promise for public transit, careful +> engineering and fine-tuning are essential to realizing their full potential. +> San Antonio serves as a case study to inform the development of LLM-powered +> transit systems in other urban environments. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on applying Large Language Models to public transportation management, rather than on the engineering, design, or optimization of prompts specifically for LLMs, failing to meet the core subject requirement. + +--- + +## [Localizing AI: Evaluating Open-Weight Language Models for Languages of + Baltic States](https://arxiv.org/abs/http://arxiv.org/abs/2501.03952v1) +**arXiv ID:** http://arxiv.org/abs/2501.03952v1 + +**Abstract:** +> Although large language models (LLMs) have transformed our expectations of +> modern language technologies, concerns over data privacy often restrict the use +> of commercially available LLMs hosted outside of EU jurisdictions. This limits +> their application in governmental, defence, and other data-sensitive sectors. +> In this work, we evaluate the extent to which locally deployable open-weight +> LLMs support lesser-spoken languages such as Lithuanian, Latvian, and Estonian. +> We examine various size and precision variants of the top-performing +> multilingual open-weight models, Llama~3, Gemma~2, Phi, and NeMo, on machine +> translation, multiple-choice question answering, and free-form text generation. +> The results indicate that while certain models like Gemma~2 perform close to +> the top commercially available models, many LLMs struggle with these languages. +> Most surprisingly, however, we find that these models, while showing close to +> state-of-the-art translation performance, are still prone to lexical +> hallucinations with errors in at least 1 in 20 words for all open-weight +> multilingual LLMs. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on evaluating open-weight Language Models for lesser-spoken languages, concerning data privacy and model performance, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs)." +} + +--- + +## [VLM-driven Behavior Tree for Context-aware Task Planning](https://arxiv.org/abs/http://arxiv.org/abs/2501.03968v2) +**arXiv ID:** http://arxiv.org/abs/2501.03968v2 + +**Abstract:** +> The use of Large Language Models (LLMs) for generating Behavior Trees (BTs) +> has recently gained attention in the robotics community, yet remains in its +> early stages of development. In this paper, we propose a novel framework that +> leverages Vision-Language Models (VLMs) to interactively generate and edit BTs +> that address visual conditions, enabling context-aware robot operations in +> visually complex environments. A key feature of our approach lies in the +> conditional control through self-prompted visual conditions. Specifically, the +> VLM generates BTs with visual condition nodes, where conditions are expressed +> as free-form text. Another VLM process integrates the text into its prompt and +> evaluates the conditions against real-world images during robot execution. We +> validated our framework in a real-world cafe scenario, demonstrating both its +> feasibility and limitations. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on leveraging Vision-Language Models (VLMs) for context-aware task planning in robotics, with prompt engineering being a secondary aspect used for conditional control, not the central focus of the paper. + +--- + +## [ChronoLLM: A Framework for Customizing Large Language Model for Digital + Twins generalization based on PyChrono](https://arxiv.org/abs/http://arxiv.org/abs/2501.04062v1) +**arXiv ID:** http://arxiv.org/abs/2501.04062v1 + +**Abstract:** +> Recently, the integration of advanced simulation technologies with artificial +> intelligence (AI) is revolutionizing science and engineering research. +> ChronoLlama introduces a novel framework that customizes the open-source LLMs, +> specifically for code generation, paired with PyChrono for multi-physics +> simulations. This integration aims to automate and improve the creation of +> simulation scripts, thus enhancing model accuracy and efficiency. This +> combination harnesses the speed of AI-driven code generation with the +> reliability of physics-based simulations, providing a powerful tool for +> researchers and engineers. Empirical results indicate substantial enhancements +> in simulation setup speed, accuracy of the generated codes, and overall +> computational efficiency. ChronoLlama not only expedites the development and +> testing of multibody systems but also spearheads a scalable, AI-enhanced +> approach to managing intricate mechanical simulations. This pioneering +> integration of cutting-edge AI with traditional simulation platforms represents +> a significant leap forward in automating and optimizing design processes in +> engineering applications. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on integrating LLMs with multi-physics simulations for automating code generation in engineering applications, rather than specifically on prompt engineering for Large Language Models. The core subject is the development of a framework for simulation script automation, not novel prompt engineering techniques or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [TrojanDec: Data-free Detection of Trojan Inputs in Self-supervised + Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.04108v1) +**arXiv ID:** http://arxiv.org/abs/2501.04108v1 + +**Abstract:** +> An image encoder pre-trained by self-supervised learning can be used as a +> general-purpose feature extractor to build downstream classifiers for various +> downstream tasks. However, many studies showed that an attacker can embed a +> trojan into an encoder such that multiple downstream classifiers built based on +> the trojaned encoder simultaneously inherit the trojan behavior. In this work, +> we propose TrojanDec, the first data-free method to identify and recover a test +> input embedded with a trigger. Given a (trojaned or clean) encoder and a test +> input, TrojanDec first predicts whether the test input is trojaned. If not, the +> test input is processed in a normal way to maintain the utility. Otherwise, the +> test input will be further restored to remove the trigger. Our extensive +> evaluation shows that TrojanDec can effectively identify the trojan (if any) +> from a given test input and recover it under state-of-the-art trojan attacks. +> We further demonstrate by experiments that our TrojanDec outperforms the +> state-of-the-art defenses. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on detecting and recovering from trojan attacks in self-supervised image encoders, which falls outside the specified criteria: it does not involve Large Language Models (LLMs), prompt engineering for text-based interactions, or text generation, and instead pertains to image processing and security. + +--- + +## [BiasGuard: Guardrailing Fairness in Machine Learning Production Systems](https://arxiv.org/abs/http://arxiv.org/abs/2501.04142v1) +**arXiv ID:** http://arxiv.org/abs/2501.04142v1 + +**Abstract:** +> As machine learning (ML) systems increasingly impact critical sectors such as +> hiring, financial risk assessments, and criminal justice, the imperative to +> ensure fairness has intensified due to potential negative implications. While +> much ML fairness research has focused on enhancing training data and processes, +> addressing the outputs of already deployed systems has received less attention. +> This paper introduces 'BiasGuard', a novel approach designed to act as a +> fairness guardrail in production ML systems. BiasGuard leverages Test-Time +> Augmentation (TTA) powered by Conditional Generative Adversarial Network +> (CTGAN), a cutting-edge generative AI model, to synthesize data samples +> conditioned on inverted protected attribute values, thereby promoting equitable +> outcomes across diverse groups. This method aims to provide equal opportunities +> for both privileged and unprivileged groups while significantly enhancing the +> fairness metrics of deployed systems without the need for retraining. Our +> comprehensive experimental analysis across diverse datasets reveals that +> BiasGuard enhances fairness by 31% while only reducing accuracy by 0.09% +> compared to non-mitigated benchmarks. Additionally, BiasGuard outperforms +> existing post-processing methods in improving fairness, positioning it as an +> effective tool to safeguard against biases when retraining the model is +> impractical. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on ensuring fairness in machine learning production systems via Test-Time Augmentation and a generative AI model, rather than specifically on prompt engineering for Large Language Models (LLMs) and the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Reasoning-Enhanced Self-Training for Long-Form Personalized Text + Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.04167v1) +**arXiv ID:** http://arxiv.org/abs/2501.04167v1 + +**Abstract:** +> Personalized text generation requires a unique ability of large language +> models (LLMs) to learn from context that they often do not encounter during +> their standard training. One way to encourage LLMs to better use personalized +> context for generating outputs that better align with the user's expectations +> is to instruct them to reason over the user's past preferences, background +> knowledge, or writing style. To achieve this, we propose Reasoning-Enhanced +> Self-Training for Personalized Text Generation (REST-PG), a framework that +> trains LLMs to reason over personal data during response generation. REST-PG +> first generates reasoning paths to train the LLM's reasoning abilities and then +> employs Expectation-Maximization Reinforced Self-Training to iteratively train +> the LLM based on its own high-reward outputs. We evaluate REST-PG on the +> LongLaMP benchmark, consisting of four diverse personalized long-form text +> generation tasks. Our experiments demonstrate that REST-PG achieves significant +> improvements over state-of-the-art baselines, with an average relative +> performance gain of 14.5% on the benchmark. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a new framework (REST-PG) for self-training LLMs to enhance personalized text generation, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models. It meets the exclusions as it's more about training methods for LLMs than prompt engineering. + +--- + +## [Learning to Transfer Human Hand Skills for Robot Manipulations](https://arxiv.org/abs/http://arxiv.org/abs/2501.04169v1) +**arXiv ID:** http://arxiv.org/abs/2501.04169v1 + +**Abstract:** +> We present a method for teaching dexterous manipulation tasks to robots from +> human hand motion demonstrations. Unlike existing approaches that solely rely +> on kinematics information without taking into account the plausibility of robot +> and object interaction, our method directly infers plausible robot manipulation +> actions from human motion demonstrations. To address the embodiment gap between +> the human hand and the robot system, our approach learns a joint motion +> manifold that maps human hand movements, robot hand actions, and object +> movements in 3D, enabling us to infer one motion component from others. Our key +> idea is the generation of pseudo-supervision triplets, which pair human, +> object, and robot motion trajectories synthetically. Through real-world +> experiments with robot hand manipulation, we demonstrate that our data-driven +> retargeting method significantly outperforms conventional retargeting +> techniques, effectively bridging the embodiment gap between human and robotic +> hands. Website at https://rureadyo.github.io/MocapRobot/. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not focus on prompt engineering for Large Language Models (LLMs), instead concentrating on robotics and teaching manipulation tasks to robots from human hand motion demonstrations, which falls outside the specified criteria." +} + +--- + +## [SNR-EQ-JSCC: Joint Source-Channel Coding with SNR-Based Embedding and + Query](https://arxiv.org/abs/http://arxiv.org/abs/2501.04732v1) +**arXiv ID:** http://arxiv.org/abs/2501.04732v1 + +**Abstract:** +> Coping with the impact of dynamic channels is a critical issue in joint +> source-channel coding (JSCC)-based semantic communication systems. In this +> paper, we propose a lightweight channel-adaptive semantic coding architecture +> called SNR-EQ-JSCC. It is built upon the generic Transformer model and achieves +> channel adaptation (CA) by Embedding the signal-to-noise ratio (SNR) into the +> attention blocks and dynamically adjusting attention scores through +> channel-adaptive Queries. Meanwhile, penalty terms are introduced in the loss +> function to stabilize the training process. Considering that instantaneous SNR +> feedback may be imperfect, we propose an alternative method that uses only the +> average SNR, which requires no retraining of SNR-EQ-JSCC. Simulation results +> conducted on image transmission demonstrate that the proposed SNR-EQJSCC +> outperforms the state-of-the-art SwinJSCC in peak signal-to-noise ratio (PSNR) +> and perception metrics while only requiring 0.05% of the storage overhead and +> 6.38% of the computational complexity for CA. Moreover, the channel-adaptive +> query method demonstrates significant improvements in perception metrics. When +> instantaneous SNR feedback is imperfect, SNR-EQ-JSCC using only the average SNR +> still surpasses baseline schemes. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on joint source-channel coding for semantic communication systems, adapting a Transformer model for channel conditions, and does not primarily investigate prompt engineering for Large Language Models (LLMs) or analyze methods for improving LLM performance through textual input prompt manipulation. + +--- + +## [RTLSquad: Multi-Agent Based Interpretable RTL Design](https://arxiv.org/abs/http://arxiv.org/abs/2501.05470v1) +**arXiv ID:** http://arxiv.org/abs/2501.05470v1 + +**Abstract:** +> Optimizing Register-Transfer Level (RTL) code is crucial for improving +> hardware PPA performance. Large Language Models (LLMs) offer new approaches for +> automatic RTL code generation and optimization. However, existing methods often +> lack decision interpretability (sufficient, understandable justification for +> decisions), making it difficult for hardware engineers to trust the generated +> results, thus preventing these methods from being integrated into the design +> process. To address this, we propose RTLSquad, a novel LLM-Based Multi-Agent +> system for interpretable RTL code generation. RTLSquad divides the design +> process into exploration, implementation, and verification & evaluation stages +> managed by specialized agent squads, generating optimized RTL code through +> inter-agent collaboration, and providing decision interpretability through the +> communication process. Experiments show that RTLSquad excels in generating +> functionally correct RTL code and optimizing PPA performance, while also having +> the capability to provide decision paths, demonstrating the practical value of +> our system. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a multi-agent system for interpretable RTL (Register-Transfer Level) design using LLMs, rather than focusing on the engineering, design, or optimization of prompts specifically for Large Language Models. Prompt engineering is not the central concern, but rather a means to achieve the system's goals. + +--- + +## [Found in Translation: semantic approaches for enhancing AI + interpretability in face verification](https://arxiv.org/abs/http://arxiv.org/abs/2501.05471v1) +**arXiv ID:** http://arxiv.org/abs/2501.05471v1 + +**Abstract:** +> The increasing complexity of machine learning models in computer vision, +> particularly in face verification, requires the development of explainable +> artificial intelligence (XAI) to enhance interpretability and transparency. +> This study extends previous work by integrating semantic concepts derived from +> human cognitive processes into XAI frameworks to bridge the comprehension gap +> between model outputs and human understanding. We propose a novel approach +> combining global and local explanations, using semantic features defined by +> user-selected facial landmarks to generate similarity maps and textual +> explanations via large language models (LLMs). The methodology was validated +> through quantitative experiments and user feedback, demonstrating improved +> interpretability. Results indicate that our semantic-based approach, +> particularly the most detailed set, offers a more nuanced understanding of +> model decisions than traditional methods. User studies highlight a preference +> for our semantic explanations over traditional pixelbased heatmaps, emphasizing +> the benefits of human-centric interpretability in AI. This work contributes to +> the ongoing efforts to create XAI frameworks that align AI models behaviour +> with human cognitive processes, fostering trust and acceptance in critical +> applications. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing AI interpretability in face verification (a computer vision task) using Large Language Models (LLMs) as a component for generating textual explanations, rather than focusing on the engineering, design, or optimization of prompts specifically for LLMs in text-based interactions. + +--- + +## [Modality-Invariant Bidirectional Temporal Representation Distillation + Network for Missing Multimodal Sentiment Analysis](https://arxiv.org/abs/http://arxiv.org/abs/2501.05474v1) +**arXiv ID:** http://arxiv.org/abs/2501.05474v1 + +**Abstract:** +> Multimodal Sentiment Analysis (MSA) integrates diverse modalities(text, +> audio, and video) to comprehensively analyze and understand individuals' +> emotional states. However, the real-world prevalence of incomplete data poses +> significant challenges to MSA, mainly due to the randomness of modality +> missing. Moreover, the heterogeneity issue in multimodal data has yet to be +> effectively addressed. To tackle these challenges, we introduce the +> Modality-Invariant Bidirectional Temporal Representation Distillation Network +> (MITR-DNet) for Missing Multimodal Sentiment Analysis. MITR-DNet employs a +> distillation approach, wherein a complete modality teacher model guides a +> missing modality student model, ensuring robustness in the presence of modality +> missing. Simultaneously, we developed the Modality-Invariant Bidirectional +> Temporal Representation Learning Module (MIB-TRL) to mitigate heterogeneity. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on Multimodal Sentiment Analysis with a network architecture (MITR-DNet) to address modality missing and heterogeneity issues, rather than prompt engineering for Large Language Models (LLMs), failing to meet the 'MUST' criteria for focus on LLM prompt engineering and manipulation of textual input prompts. + +--- + +## [Cooperative Search and Track of Rogue Drones using Multiagent + Reinforcement Learning](https://arxiv.org/abs/http://arxiv.org/abs/2501.10413v1) +**arXiv ID:** http://arxiv.org/abs/2501.10413v1 + +**Abstract:** +> This work considers the problem of intercepting rogue drones targeting +> sensitive critical infrastructure facilities. While current interception +> technologies focus mainly on the jamming/spoofing tasks, the challenges of +> effectively locating and tracking rogue drones have not received adequate +> attention. Solving this problem and integrating with recently proposed +> interception techniques will enable a holistic system that can reliably detect, +> track, and neutralize rogue drones. Specifically, this work considers a team of +> pursuer UAVs that can search, detect, and track multiple rogue drones over a +> sensitive facility. The joint search and track problem is addressed through a +> novel multiagent reinforcement learning scheme to optimize the agent mobility +> control actions that maximize the number of rogue drones detected and tracked. +> The performance of the proposed system is investigated under realistic settings +> through extensive simulation experiments with varying number of agents +> demonstrating both its performance and scalability. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on multiagent reinforcement learning for tracking rogue drones, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet all 'MUST' criteria. + +--- + +## [Efficient Deployment of Large Language Models on Resource-constrained + Devices](https://arxiv.org/abs/http://arxiv.org/abs/2501.02438v1) +**arXiv ID:** http://arxiv.org/abs/2501.02438v1 + +**Abstract:** +> Deploying Large Language Models (LLMs) on resource-constrained (or weak) +> devices presents significant challenges due to limited resources and +> heterogeneous data distribution. To address the data concern, it is necessary +> to fine-tune LLMs using on-device private data for various downstream tasks. +> While Federated Learning (FL) offers a promising privacy-preserving solution, +> existing fine-tuning methods retain the original LLM size, leaving issues of +> high inference latency and excessive memory demands unresolved. Hence, we +> design FedSpine, an FL framework that combines Parameter- Efficient Fine-Tuning +> (PEFT) with structured pruning for efficient deployment of LLMs on +> resource-constrained devices. Specifically, FedSpine introduces an iterative +> process to prune and tune the parameters of LLMs. To mitigate the impact of +> device heterogeneity, an online Multi-Armed Bandit (MAB) algorithm is employed +> to adaptively determine different pruning ratios and LoRA ranks for +> heterogeneous devices without any prior knowledge of their computing and +> communication capabilities. As a result, FedSpine maintains higher inference +> accuracy while improving fine-tuning efficiency. Experimental results conducted +> on a physical platform with 80 devices demonstrate that FedSpine can speed up +> fine-tuning by 1.4$\times$-6.9$\times$ and improve final accuracy by 0.4%-4.5% +> under the same sparsity level compared to other baselines. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the efficient deployment and fine-tuning of Large Language Models on resource-constrained devices, rather than on the engineering, design, or optimization of prompts for LLMs, failing to meet the first 'MUST' criterion. + +--- + +## [FedRSClip: Federated Learning for Remote Sensing Scene Classification + Using Vision-Language Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02461v1) +**arXiv ID:** http://arxiv.org/abs/2501.02461v1 + +**Abstract:** +> Remote sensing data is often distributed across multiple institutions, and +> due to privacy concerns and data-sharing restrictions, leveraging large-scale +> datasets in a centralized training framework is challenging. Federated learning +> offers a promising solution by enabling collaborative model training across +> distributed data sources without requiring data centralization. However, +> current Vision-Language Models (VLMs), which typically contain billions of +> parameters, pose significant communication challenges for traditional federated +> learning approaches based on model parameter updates, as they would incur +> substantial communication costs. In this paper, we propose FedRSCLIP, the first +> federated learning framework designed for remote sensing image classification +> based on a VLM, specifically CLIP. FedRSCLIP addresses the challenges of data +> heterogeneity and large-scale model transmission in federated environments by +> introducing Prompt Learning, which optimizes only a small set of tunable +> parameters. The framework introduces a dual-prompt mechanism, comprising Shared +> Prompts for global knowledge sharing and Private Prompts for client-specific +> adaptation. To maintain semantic coherence between shared and private prompts, +> we propose the Dual Prompt Alignment Constraint to balance global consistency +> and local adaptability across diverse client distributions. Additionally, to +> enhance cross-modal representation learning, we introduce the Cross-Modal +> Feature Alignment Constraint to align multimodal features between text and +> image prompts. To validate the effectiveness of our proposed model, we +> construct a Fed-RSIC dataset based on three existing remote sensing image +> classification datasets, specifically designed to simulate various federated +> learning configurations. Experimental results demonstrate the effectiveness and +> superiority of FedRSCLIP in remote sensing image classification. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on federated learning for Vision-Language Models (VLMs) in remote sensing image classification, not on the engineering, design, or optimization of textual input prompts for Large Language Models (LLMs), and does not provide concrete examples of text prompts impacting LLM output. + +--- + +## [Depth Any Camera: Zero-Shot Metric Depth Estimation from Any Camera](https://arxiv.org/abs/http://arxiv.org/abs/2501.02464v1) +**arXiv ID:** http://arxiv.org/abs/2501.02464v1 + +**Abstract:** +> While recent depth estimation methods exhibit strong zero-shot +> generalization, achieving accurate metric depth across diverse camera +> types-particularly those with large fields of view (FoV) such as fisheye and +> 360-degree cameras-remains a significant challenge. This paper presents Depth +> Any Camera (DAC), a powerful zero-shot metric depth estimation framework that +> extends a perspective-trained model to effectively handle cameras with varying +> FoVs. The framework is designed to ensure that all existing 3D data can be +> leveraged, regardless of the specific camera types used in new applications. +> Remarkably, DAC is trained exclusively on perspective images but generalizes +> seamlessly to fisheye and 360-degree cameras without the need for specialized +> training data. DAC employs Equi-Rectangular Projection (ERP) as a unified image +> representation, enabling consistent processing of images with diverse FoVs. Its +> key components include a pitch-aware Image-to-ERP conversion for efficient +> online augmentation in ERP space, a FoV alignment operation to support +> effective training across a wide range of FoVs, and multi-resolution data +> augmentation to address resolution disparities between training and testing. +> DAC achieves state-of-the-art zero-shot metric depth estimation, improving +> delta-1 ($\delta_1$) accuracy by up to 50% on multiple fisheye and 360-degree +> datasets compared to prior metric depth foundation models, demonstrating robust +> generalization across camera types. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on depth estimation in computer vision, specifically developing a framework for metric depth estimation from various camera types, and does not address prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [PTEENet: Post-Trained Early-Exit Neural Networks Augmentation for + Inference Cost Optimization](https://arxiv.org/abs/http://arxiv.org/abs/2501.02508v1) +**arXiv ID:** http://arxiv.org/abs/2501.02508v1 + +**Abstract:** +> For many practical applications, a high computational cost of inference over +> deep network architectures might be unacceptable. A small degradation in the +> overall inference accuracy might be a reasonable price to pay for a significant +> reduction in the required computational resources. In this work, we describe a +> method for introducing "shortcuts" into the DNN feedforward inference process +> by skipping costly feedforward computations whenever possible. The proposed +> method is based on the previously described BranchyNet (Teerapittayanon et al., +> 2016) and the EEnet (Demir, 2019) architectures that jointly train the main +> network and early exit branches. We extend those methods by attaching branches +> to pre-trained models and, thus, eliminating the need to alter the original +> weights of the network. We also suggest a new branch architecture based on +> convolutional building blocks to allow enough training capacity when applied on +> large DNNs. The proposed architecture includes confidence heads that are used +> for predicting the confidence level in the corresponding early exits. By +> defining adjusted thresholds on these confidence extensions, we can control in +> real-time the amount of data exiting from each branch and the overall tradeoff +> between speed and accuracy of our model. In our experiments, we evaluate our +> method using image datasets (SVHN and CIFAR10) and several DNN architectures +> (ResNet, DenseNet, VGG) with varied depth. Our results demonstrate that the +> proposed method enables us to reduce the average inference computational cost +> and further controlling the tradeoff between the model accuracy and the +> computation cost. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on optimizing the inference cost of deep neural networks (DNNs) through early-exit strategies, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thus failing to meet the primary criteria. + +--- + +## [Evaluating Large Language Models Against Human Annotators in Latent + Content Analysis: Sentiment, Political Leaning, Emotional Intensity, and + Sarcasm](https://arxiv.org/abs/http://arxiv.org/abs/2501.02532v1) +**arXiv ID:** http://arxiv.org/abs/2501.02532v1 + +**Abstract:** +> In the era of rapid digital communication, vast amounts of textual data are +> generated daily, demanding efficient methods for latent content analysis to +> extract meaningful insights. Large Language Models (LLMs) offer potential for +> automating this process, yet comprehensive assessments comparing their +> performance to human annotators across multiple dimensions are lacking. This +> study evaluates the reliability, consistency, and quality of seven +> state-of-the-art LLMs, including variants of OpenAI's GPT-4, Gemini, Llama, and +> Mixtral, relative to human annotators in analyzing sentiment, political +> leaning, emotional intensity, and sarcasm detection. A total of 33 human +> annotators and eight LLM variants assessed 100 curated textual items, +> generating 3,300 human and 19,200 LLM annotations, with LLMs evaluated across +> three time points to examine temporal consistency. Inter-rater reliability was +> measured using Krippendorff's alpha, and intra-class correlation coefficients +> assessed consistency over time. The results reveal that both humans and LLMs +> exhibit high reliability in sentiment analysis and political leaning +> assessments, with LLMs demonstrating higher internal consistency than humans. +> In emotional intensity, LLMs displayed higher agreement compared to humans, +> though humans rated emotional intensity significantly higher. Both groups +> struggled with sarcasm detection, evidenced by low agreement. LLMs showed +> excellent temporal consistency across all dimensions, indicating stable +> performance over time. This research concludes that LLMs, especially GPT-4, can +> effectively replicate human analysis in sentiment and political leaning, +> although human expertise remains essential for emotional intensity +> interpretation. The findings demonstrate the potential of LLMs for consistent +> and high-quality performance in certain areas of latent content analysis. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on evaluating LLMs' performance in latent content analysis compared to human annotators, rather than primarily investigating, analyzing, or proposing methods for improving LLM performance through the manipulation of textual input prompts. + +--- + +## [Energy Optimization of Multi-task DNN Inference in MEC-assisted XR + Devices: A Lyapunov-Guided Reinforcement Learning Approach](https://arxiv.org/abs/http://arxiv.org/abs/2501.02572v1) +**arXiv ID:** http://arxiv.org/abs/2501.02572v1 + +**Abstract:** +> Extended reality (XR), blending virtual and real worlds, is a key application +> of future networks. While AI advancements enhance XR capabilities, they also +> impose significant computational and energy challenges on lightweight XR +> devices. In this paper, we developed a distributed queue model for multi-task +> DNN inference, addressing issues of resource competition and queue coupling. In +> response to the challenges posed by the high energy consumption and limited +> resources of XR devices, we designed a dual time-scale joint optimization +> strategy for model partitioning and resource allocation, formulated as a +> bi-level optimization problem. This strategy aims to minimize the total energy +> consumption of XR devices while ensuring queue stability and adhering to +> computational and communication resource constraints. To tackle this problem, +> we devised a Lyapunov-guided Proximal Policy Optimization algorithm, named +> LyaPPO. Numerical results demonstrate that the LyaPPO algorithm outperforms the +> baselines, achieving energy conservation of 24.79% to 46.14% under varying +> resource capacities. Specifically, the proposed algorithm reduces the energy +> consumption of XR devices by 24.29% to 56.62% compared to baseline algorithms. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on energy optimization of Multi-task DNN Inference in MEC-assisted XR Devices using Reinforcement Learning, with no apparent focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs), thus failing to meet the 'MUST' criteria. + +--- + +## [TAPAS: Thermal- and Power-Aware Scheduling for LLM Inference in Cloud + Platforms](https://arxiv.org/abs/http://arxiv.org/abs/2501.02600v1) +**arXiv ID:** http://arxiv.org/abs/2501.02600v1 + +**Abstract:** +> The rising demand for generative large language models (LLMs) poses +> challenges for thermal and power management in cloud datacenters. Traditional +> techniques often are inadequate for LLM inference due to the fine-grained, +> millisecond-scale execution phases, each with distinct performance, thermal, +> and power profiles. Additionally, LLM inference workloads are sensitive to +> various configuration parameters (e.g., model parallelism, size, and +> quantization) that involve trade-offs between performance, temperature, power, +> and output quality. Moreover, clouds often co-locate SaaS and IaaS workloads, +> each with different levels of visibility and flexibility. We propose TAPAS, a +> thermal- and power-aware framework designed for LLM inference clusters in the +> cloud. TAPAS enhances cooling and power oversubscription capabilities, reducing +> the total cost of ownership (TCO) while effectively handling emergencies (e.g., +> cooling and power failures). The system leverages historical temperature and +> power data, along with the adaptability of SaaS workloads, to: (1) efficiently +> place new GPU workload VMs within cooling and power constraints, (2) route LLM +> inference requests across SaaS VMs, and (3) reconfigure SaaS VMs to manage load +> spikes and emergency situations. Our evaluation on a large GPU cluster +> demonstrates significant reductions in thermal and power throttling events, +> boosting system efficiency. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on thermal- and power-aware scheduling for LLM inference in cloud platforms, addressing infrastructure management rather than prompt engineering for text-based interactions with LLMs, thus failing to meet the core subject requirement. + +--- + +## [Layer-Level Self-Exposure and Patch: Affirmative Token Mitigation for + Jailbreak Attack Defense](https://arxiv.org/abs/http://arxiv.org/abs/2501.02629v1) +**arXiv ID:** http://arxiv.org/abs/2501.02629v1 + +**Abstract:** +> As large language models (LLMs) are increasingly deployed in diverse +> applications, including chatbot assistants and code generation, aligning their +> behavior with safety and ethical standards has become paramount. However, +> jailbreak attacks, which exploit vulnerabilities to elicit unintended or +> harmful outputs, threaten LLMs' safety significantly. In this paper, we +> introduce Layer-AdvPatcher, a novel methodology designed to defend against +> jailbreak attacks by utilizing an unlearning strategy to patch specific layers +> within LLMs through self-augmented datasets. Our insight is that certain +> layer(s), tend to produce affirmative tokens when faced with harmful prompts. +> By identifying these layers and adversarially exposing them to generate more +> harmful data, one can understand their inherent and diverse vulnerabilities to +> attacks. With these exposures, we then "unlearn" these issues, reducing the +> impact of affirmative tokens and hence minimizing jailbreak risks while keeping +> the model's responses to safe queries intact. We conduct extensive experiments +> on two models, four benchmark datasets, and multiple state-of-the-art jailbreak +> benchmarks to demonstrate the efficacy of our approach. Results indicate that +> our framework reduces the harmfulness and attack success rate of jailbreak +> attacks without compromising utility for benign queries compared to recent +> defense methods. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on defending against jailbreak attacks by modifying LLM architecture (Layer-AdvPatcher) and utilizing an unlearning strategy, rather than primarily focusing on the engineering, design, or optimization of prompts for improving LLM performance through textual input manipulation. + +--- + +## [Representation Learning of Lab Values via Masked AutoEncoder](https://arxiv.org/abs/http://arxiv.org/abs/2501.02648v2) +**arXiv ID:** http://arxiv.org/abs/2501.02648v2 + +**Abstract:** +> Accurate imputation of missing laboratory values in electronic health records +> (EHRs) is critical to enable robust clinical predictions and reduce biases in +> AI systems in healthcare. Existing methods, such as variational autoencoders +> (VAEs) and decision tree-based approaches such as XGBoost, struggle to model +> the complex temporal and contextual dependencies in EHR data, mainly in +> underrepresented groups. In this work, we propose Lab-MAE, a novel +> transformer-based masked autoencoder framework that leverages self-supervised +> learning for the imputation of continuous sequential lab values. Lab-MAE +> introduces a structured encoding scheme that jointly models laboratory test +> values and their corresponding timestamps, enabling explicit capturing temporal +> dependencies. Empirical evaluation on the MIMIC-IV dataset demonstrates that +> Lab-MAE significantly outperforms the state-of-the-art baselines such as +> XGBoost across multiple metrics, including root mean square error (RMSE), +> R-squared (R2), and Wasserstein distance (WD). Notably, Lab-MAE achieves +> equitable performance across demographic groups of patients, advancing fairness +> in clinical predictions. We further investigate the role of follow-up +> laboratory values as potential shortcut features, revealing Lab-MAE's +> robustness in scenarios where such data is unavailable. The findings suggest +> that our transformer-based architecture, adapted to the characteristics of the +> EHR data, offers a foundation model for more accurate and fair clinical +> imputation models. In addition, we measure and compare the carbon footprint of +> Lab-MAE with the baseline XGBoost model, highlighting its environmental +> requirements. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a novel transformer-based masked autoencoder for imputing missing laboratory values in electronic health records, which does not meet the 'MUST' criteria of primarily focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts for improving LLM performance. + +--- + +## [From thermodynamics to protein design: Diffusion models for biomolecule + generation towards autonomous protein engineering](https://arxiv.org/abs/http://arxiv.org/abs/2501.02680v1) +**arXiv ID:** http://arxiv.org/abs/2501.02680v1 + +**Abstract:** +> Protein design with desirable properties has been a significant challenge for +> many decades. Generative artificial intelligence is a promising approach and +> has achieved great success in various protein generation tasks. Notably, +> diffusion models stand out for their robust mathematical foundations and +> impressive generative capabilities, offering unique advantages in certain +> applications such as protein design. In this review, we first give the +> definition and characteristics of diffusion models and then focus on two +> strategies: Denoising Diffusion Probabilistic Models and Score-based Generative +> Models, where DDPM is the discrete form of SGM. Furthermore, we discuss their +> applications in protein design, peptide generation, drug discovery, and +> protein-ligand interaction. Finally, we outline the future perspectives of +> diffusion models to advance autonomous protein design and engineering. The E(3) +> group consists of all rotations, reflections, and translations in +> three-dimensions. The equivariance on the E(3) group can keep the physical +> stability of the frame of each amino acid as much as possible, and we reflect +> on how to keep the diffusion model E(3) equivariant for protein generation. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on the application of diffusion models for biomolecule generation in protein design, not on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). It also concerns a medical/biological subject (protein design) and generative AI application (protein generation) outside the specified text generation driven by LLMs. + +--- + +## [EAGLE: Enhanced Visual Grounding Minimizes Hallucinations in + Instructional Multimodal Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02699v1) +**arXiv ID:** http://arxiv.org/abs/2501.02699v1 + +**Abstract:** +> Large language models and vision transformers have demonstrated impressive +> zero-shot capabilities, enabling significant transferability in downstream +> tasks. The fusion of these models has resulted in multi-modal architectures +> with enhanced instructional capabilities. Despite incorporating vast image and +> language pre-training, these multi-modal architectures often generate responses +> that deviate from the ground truth in the image data. These failure cases are +> known as hallucinations. Current methods for mitigating hallucinations +> generally focus on regularizing the language component, improving the fusion +> module, or ensembling multiple visual encoders to improve visual +> representation. In this paper, we address the hallucination issue by directly +> enhancing the capabilities of the visual component. Our approach, named EAGLE, +> is fully agnostic to the LLM or fusion module and works as a post-pretraining +> approach that improves the grounding and language alignment of the visual +> encoder. We show that a straightforward reformulation of the original +> contrastive pre-training task results in an improved visual encoder that can be +> incorporated into the instructional multi-modal architecture without additional +> instructional training. As a result, EAGLE achieves a significant reduction in +> hallucinations across multiple challenging benchmarks and tasks. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on enhancing visual grounding in multimodal models to reduce hallucinations, rather than engineering or optimizing prompts specifically for Large Language Models (LLMs). It meets none of the 'MUST' criteria for prompt engineering, focusing instead on improving the visual component of multimodal architectures. + +--- + +## [QuIM-RAG: Advancing Retrieval-Augmented Generation with Inverted + Question Matching for Enhanced QA Performance](https://arxiv.org/abs/http://arxiv.org/abs/2501.02702v1) +**arXiv ID:** http://arxiv.org/abs/2501.02702v1 + +**Abstract:** +> This work presents a novel architecture for building Retrieval-Augmented +> Generation (RAG) systems to improve Question Answering (QA) tasks from a target +> corpus. Large Language Models (LLMs) have revolutionized the analyzing and +> generation of human-like text. These models rely on pre-trained data and lack +> real-time updates unless integrated with live data tools. RAG enhances LLMs by +> integrating online resources and databases to generate contextually appropriate +> responses. However, traditional RAG still encounters challenges like +> information dilution and hallucinations when handling vast amounts of data. Our +> approach addresses these challenges by converting corpora into a +> domain-specific dataset and RAG architecture is constructed to generate +> responses from the target document. We introduce QuIM-RAG (Question-to-question +> Inverted Index Matching), a novel approach for the retrieval mechanism in our +> system. This strategy generates potential questions from document chunks and +> matches these with user queries to identify the most relevant text chunks for +> generating accurate answers. We have implemented our RAG system on top of the +> open-source Meta-LLaMA3-8B-instruct model by Meta Inc. that is available on +> Hugging Face. We constructed a custom corpus of 500+ pages from a high-traffic +> website accessed thousands of times daily for answering complex questions, +> along with manually prepared ground truth QA for evaluation. We compared our +> approach with traditional RAG models using BERT-Score and RAGAS, +> state-of-the-art metrics for evaluating LLM applications. Our evaluation +> demonstrates that our approach outperforms traditional RAG architectures on +> both metrics. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on advancing Retrieval-Augmented Generation (RAG) system architecture for Question Answering tasks, rather than specifically on the engineering, design, or optimization of prompts for Large Language Models (LLMs), as required. While LLMs are utilized, the core subject is the RAG system's enhancement, not prompt engineering for text-based LLM interactions. + +--- + +## [OpenGU: A Comprehensive Benchmark for Graph Unlearning](https://arxiv.org/abs/http://arxiv.org/abs/2501.02728v1) +**arXiv ID:** http://arxiv.org/abs/2501.02728v1 + +**Abstract:** +> Graph Machine Learning is essential for understanding and analyzing +> relational data. However, privacy-sensitive applications demand the ability to +> efficiently remove sensitive information from trained graph neural networks +> (GNNs), avoiding the unnecessary time and space overhead caused by retraining +> models from scratch. To address this issue, Graph Unlearning (GU) has emerged +> as a critical solution, with the potential to support dynamic graph updates in +> data management systems and enable scalable unlearning in distributed data +> systems while ensuring privacy compliance. Unlike machine unlearning in +> computer vision or other fields, GU faces unique difficulties due to the +> non-Euclidean nature of graph data and the recursive message-passing mechanism +> of GNNs. Additionally, the diversity of downstream tasks and the complexity of +> unlearning requests further amplify these challenges. Despite the proliferation +> of diverse GU strategies, the absence of a benchmark providing fair comparisons +> for GU, and the limited flexibility in combining downstream tasks and +> unlearning requests, have yielded inconsistencies in evaluations, hindering the +> development of this domain. To fill this gap, we present OpenGU, the first GU +> benchmark, where 16 SOTA GU algorithms and 37 multi-domain datasets are +> integrated, enabling various downstream tasks with 13 GNN backbones when +> responding to flexible unlearning requests. Based on this unified benchmark +> framework, we are able to provide a comprehensive and fair evaluation for GU. +> Through extensive experimentation, we have drawn $8$ crucial conclusions about +> existing GU methods, while also gaining valuable insights into their +> limitations, shedding light on potential avenues for future research. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. Instead, it focuses on Graph Unlearning (GU) for Graph Neural Networks (GNNs), which falls outside the specified criteria. + +--- + +## [GLoG-CSUnet: Enhancing Vision Transformers with Adaptable Radiomic + Features for Medical Image Segmentation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02788v2) +**arXiv ID:** http://arxiv.org/abs/2501.02788v2 + +**Abstract:** +> Vision Transformers (ViTs) have shown promise in medical image semantic +> segmentation (MISS) by capturing long-range correlations. However, ViTs often +> struggle to model local spatial information effectively, which is essential for +> accurately segmenting fine anatomical details, particularly when applied to +> small datasets without extensive pre-training. We introduce Gabor and Laplacian +> of Gaussian Convolutional Swin Network (GLoG-CSUnet), a novel architecture +> enhancing Transformer-based models by incorporating learnable radiomic +> features. This approach integrates dynamically adaptive Gabor and Laplacian of +> Gaussian (LoG) filters to capture texture, edge, and boundary information, +> enhancing the feature representation processed by the Transformer model. Our +> method uniquely combines the long-range dependency modeling of Transformers +> with the texture analysis capabilities of Gabor and LoG features. Evaluated on +> the Synapse multi-organ and ACDC cardiac segmentation datasets, GLoG-CSUnet +> demonstrates significant improvements over state-of-the-art models, achieving a +> 1.14% increase in Dice score for Synapse and 0.99% for ACDC, with minimal +> computational overhead (only 15 and 30 additional parameters, respectively). +> GLoG-CSUnet's flexible design allows integration with various base models, +> offering a promising approach for incorporating radiomics-inspired feature +> extraction in Transformer architectures for medical image analysis. The code +> implementation is available on GitHub at: https://github.com/HAAIL/GLoG-CSUnet. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing Vision Transformers with adaptable radiomic features for Medical Image Segmentation, which falls under the excluded categories of 'medical subjects' and 'applications of generative AI other than text generation driven by LLMs', and does not meet the core subject requirement of prompt engineering for text-based interactions with LLMs. + +--- + +## [RDD4D: 4D Attention-Guided Road Damage Detection And Classification](https://arxiv.org/abs/http://arxiv.org/abs/2501.02822v1) +**arXiv ID:** http://arxiv.org/abs/2501.02822v1 + +**Abstract:** +> Road damage detection and assessment are crucial components of infrastructure +> maintenance. However, current methods often struggle with detecting multiple +> types of road damage in a single image, particularly at varying scales. This is +> due to the lack of road datasets with various damage types having varying +> scales. To overcome this deficiency, first, we present a novel dataset called +> Diverse Road Damage Dataset (DRDD) for road damage detection that captures the +> diverse road damage types in individual images, addressing a crucial gap in +> existing datasets. Then, we provide our model, RDD4D, that exploits Attention4D +> blocks, enabling better feature refinement across multiple scales. The +> Attention4D module processes feature maps through an attention mechanism +> combining positional encoding and "Talking Head" components to capture local +> and global contextual information. In our comprehensive experimental analysis +> comparing various state-of-the-art models on our proposed, our enhanced model +> demonstrated superior performance in detecting large-sized road cracks with an +> Average Precision (AP) of 0.458 and maintained competitive performance with an +> overall AP of 0.445. Moreover, we also provide results on the CrackTinyNet +> dataset; our model achieved around a 0.21 increase in performance. The code, +> model weights, dataset, and our results are available on +> \href{https://github.com/msaqib17/Road_Damage_Detection}{https://github.com/msaqib17/Road\_Damage\_Detection}. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on road damage detection using computer vision and deep learning techniques, with no mention of Large Language Models (LLMs), prompt engineering, or textual input prompts, thereby failing to meet all the 'MUST' criteria. + +--- + +## [Forward Once for All: Structural Parameterized Adaptation for Efficient + Cloud-coordinated On-device Recommendation](https://arxiv.org/abs/http://arxiv.org/abs/2501.02837v1) +**arXiv ID:** http://arxiv.org/abs/2501.02837v1 + +**Abstract:** +> In cloud-centric recommender system, regular data exchanges between user +> devices and cloud could potentially elevate bandwidth demands and privacy +> risks. On-device recommendation emerges as a viable solution by performing +> reranking locally to alleviate these concerns. Existing methods primarily focus +> on developing local adaptive parameters, while potentially neglecting the +> critical role of tailor-made model architecture. Insights from broader research +> domains suggest that varying data distributions might favor distinct +> architectures for better fitting. In addition, imposing a uniform model +> structure across heterogeneous devices may result in risking inefficacy on less +> capable devices or sub-optimal performance on those with sufficient +> capabilities. In response to these gaps, our paper introduces Forward-OFA, a +> novel approach for the dynamic construction of device-specific networks (both +> structure and parameters). Forward-OFA employs a structure controller to +> selectively determine whether each block needs to be assembled for a given +> device. However, during the training of the structure controller, these +> assembled heterogeneous structures are jointly optimized, where the co-adaption +> among blocks might encounter gradient conflicts. To mitigate this, Forward-OFA +> is designed to establish a structure-guided mapping of real-time behaviors to +> the parameters of assembled networks. Structure-related parameters and parallel +> components within the mapper prevent each part from receiving heterogeneous +> gradients from others, thus bypassing the gradient conflicts for coupled +> optimization. Besides, direct mapping enables Forward-OFA to achieve adaptation +> through only one forward pass, allowing for swift adaptation to changing +> interests and eliminating the requirement for on-device backpropagation. +> Experiments on real-world datasets demonstrate the effectiveness and efficiency +> of Forward-OFA. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a novel approach for dynamic construction of device-specific networks for on-device recommendation, with no clear emphasis on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Explaining Humour Style Classifications: An XAI Approach to + Understanding Computational Humour Analysis](https://arxiv.org/abs/http://arxiv.org/abs/2501.02891v1) +**arXiv ID:** http://arxiv.org/abs/2501.02891v1 + +**Abstract:** +> Humour styles can have either a negative or a positive impact on well-being. +> Given the importance of these styles to mental health, significant research has +> been conducted on their automatic identification. However, the automated +> machine learning models used for this purpose are black boxes, making their +> prediction decisions opaque. Clarity and transparency are vital in the field of +> mental health. This paper presents an explainable AI (XAI) framework for +> understanding humour style classification, building upon previous work in +> computational humour analysis. Using the best-performing single model +> (ALI+XGBoost) from prior research, we apply comprehensive XAI techniques to +> analyse how linguistic, emotional, and semantic features contribute to humour +> style classification decisions. Our analysis reveals distinct patterns in how +> different humour styles are characterised and misclassified, with particular +> emphasis on the challenges in distinguishing affiliative humour from other +> styles. Through detailed examination of feature importance, error patterns, and +> misclassification cases, we identify key factors influencing model decisions, +> including emotional ambiguity, context misinterpretation, and target +> identification. The framework demonstrates significant utility in understanding +> model behaviour, achieving interpretable insights into the complex interplay of +> features that define different humour styles. Our findings contribute to both +> the theoretical understanding of computational humour analysis and practical +> applications in mental health, content moderation, and digital humanities +> research. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on explainable AI (XAI) for humour style classification in computational humour analysis, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), failing to meet the first 'MUST' criterion. + +--- + +## [Label-free Concept Based Multiple Instance Learning for Gigapixel + Histopathology](https://arxiv.org/abs/http://arxiv.org/abs/2501.02922v1) +**arXiv ID:** http://arxiv.org/abs/2501.02922v1 + +**Abstract:** +> Multiple Instance Learning (MIL) methods allow for gigapixel Whole-Slide +> Image (WSI) analysis with only slide-level annotations. Interpretability is +> crucial for safely deploying such algorithms in high-stakes medical domains. +> Traditional MIL methods offer explanations by highlighting salient regions. +> However, such spatial heatmaps provide limited insights for end users. To +> address this, we propose a novel inherently interpretable WSI-classification +> approach that uses human-understandable pathology concepts to generate +> explanations. Our proposed Concept MIL model leverages recent advances in +> vision-language models to directly predict pathology concepts based on image +> features. The model's predictions are obtained through a linear combination of +> the concepts identified on the top-K patches of a WSI, enabling inherent +> explanations by tracing each concept's influence on the prediction. In contrast +> to traditional concept-based interpretable models, our approach eliminates the +> need for costly human annotations by leveraging the vision-language model. We +> validate our method on two widely used pathology datasets: Camelyon16 and +> PANDA. On both datasets, Concept MIL achieves AUC and accuracy scores over 0.9, +> putting it on par with state-of-the-art models. We further find that 87.1\% +> (Camelyon16) and 85.3\% (PANDA) of the top 20 patches fall within the tumor +> region. A user study shows that the concepts identified by our model align with +> the concepts used by pathologists, making it a promising strategy for +> human-interpretable WSI classification. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on medical image analysis (histopathology) using vision-language models, not on engineering, design, or optimization of textual input prompts for Large Language Models (LLMs), and does not meet the core subject requirement of prompt engineering for text-based interactions with LLMs. + +--- + +## [Socratic Questioning: Learn to Self-guide Multimodal Reasoning in the + Wild](https://arxiv.org/abs/http://arxiv.org/abs/2501.02964v2) +**arXiv ID:** http://arxiv.org/abs/2501.02964v2 + +**Abstract:** +> Complex visual reasoning remains a key challenge today. Typically, the +> challenge is tackled using methodologies such as Chain of Thought (COT) and +> visual instruction tuning. However, how to organically combine these two +> methodologies for greater success remains unexplored. Also, issues like +> hallucinations and high training cost still need to be addressed. In this work, +> we devise an innovative multi-round training and reasoning framework suitable +> for lightweight Multimodal Large Language Models (MLLMs). Our self-questioning +> approach heuristically guides MLLMs to focus on visual clues relevant to the +> target problem, reducing hallucinations and enhancing the model's ability to +> describe fine-grained image details. This ultimately enables the model to +> perform well in complex visual reasoning and question-answering tasks. We have +> named this framework Socratic Questioning(SQ). To facilitate future research, +> we create a multimodal mini-dataset named CapQA, which includes 1k images of +> fine-grained activities, for visual instruction tuning and evaluation, our +> proposed SQ method leads to a 31.2% improvement in the hallucination score. Our +> extensive experiments on various benchmarks demonstrate SQ's remarkable +> capabilities in heuristic self-questioning, zero-shot visual reasoning and +> hallucination mitigation. Our model and code will be publicly available. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multimodal reasoning, visual question-answering, and reducing hallucinations in Multimodal Large Language Models (MLLMs) through a self-questioning framework, rather than primarily on the engineering, design, or optimization of textual input prompts for Large Language Models (LLMs) as required. + +--- + +## [Proof-of-Data: A Consensus Protocol for Collaborative Intelligence](https://arxiv.org/abs/http://arxiv.org/abs/2501.02971v1) +**arXiv ID:** http://arxiv.org/abs/2501.02971v1 + +**Abstract:** +> Existing research on federated learning has been focused on the setting where +> learning is coordinated by a centralized entity. Yet the greatest potential of +> future collaborative intelligence would be unleashed in a more open and +> democratized setting with no central entity in a dominant role, referred to as +> "decentralized federated learning". New challenges arise accordingly in +> achieving both correct model training and fair reward allocation with +> collective effort among all participating nodes, especially with the threat of +> the Byzantine node jeopardising both tasks. +> In this paper, we propose a blockchain-based decentralized Byzantine +> fault-tolerant federated learning framework based on a novel Proof-of-Data +> (PoD) consensus protocol to resolve both the "trust" and "incentive" +> components. By decoupling model training and contribution accounting, PoD is +> able to enjoy not only the benefit of learning efficiency and system liveliness +> from asynchronous societal-scale PoW-style learning but also the finality of +> consensus and reward allocation from epoch-based BFT-style voting. To mitigate +> false reward claims by data forgery from Byzantine attacks, a privacy-aware +> data verification and contribution-based reward allocation mechanism is +> designed to complete the framework. Our evaluation results show that PoD +> demonstrates performance in model training close to that of the centralized +> counterpart while achieving trust in consensus and fairness for reward +> allocation with a fault tolerance ratio of 1/3. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on decentralized federated learning, blockchain, and a consensus protocol, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Fuzzy Granule Density-Based Outlier Detection with Multi-Scale Granular + Balls](https://arxiv.org/abs/http://arxiv.org/abs/2501.02975v1) +**arXiv ID:** http://arxiv.org/abs/2501.02975v1 + +**Abstract:** +> Outlier detection refers to the identification of anomalous samples that +> deviate significantly from the distribution of normal data and has been +> extensively studied and used in a variety of practical tasks. However, most +> unsupervised outlier detection methods are carefully designed to detect +> specified outliers, while real-world data may be entangled with different types +> of outliers. In this study, we propose a fuzzy rough sets-based multi-scale +> outlier detection method to identify various types of outliers. Specifically, a +> novel fuzzy rough sets-based method that integrates relative fuzzy granule +> density is first introduced to improve the capability of detecting local +> outliers. Then, a multi-scale view generation method based on granular-ball +> computing is proposed to collaboratively identify group outliers at different +> levels of granularity. Moreover, reliable outliers and inliers determined by +> the three-way decision are used to train a weighted support vector machine to +> further improve the performance of outlier detection. The proposed method +> innovatively transforms unsupervised outlier detection into a semi-supervised +> classification problem and for the first time explores the fuzzy rough +> sets-based outlier detection from the perspective of multi-scale granular +> balls, allowing for high adaptability to different types of outliers. Extensive +> experiments carried out on both artificial and UCI datasets demonstrate that +> the proposed outlier detection method significantly outperforms the +> state-of-the-art methods, improving the results by at least 8.48% in terms of +> the Area Under the ROC Curve (AUROC) index. { The source codes are released at +> \url{https://github.com/Xiaofeng-Tan/MGBOD}. } + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary criteria as it focuses on outlier detection using fuzzy rough sets and multi-scale granular balls, with no mention of Large Language Models (LLMs), prompt engineering, or text generation, thereby failing to align with the specified requirements. + +--- + +## [CAMP: Collaborative Attention Model with Profiles for Vehicle Routing + Problems](https://arxiv.org/abs/http://arxiv.org/abs/2501.02977v1) +**arXiv ID:** http://arxiv.org/abs/2501.02977v1 + +**Abstract:** +> The profiled vehicle routing problem (PVRP) is a generalization of the +> heterogeneous capacitated vehicle routing problem (HCVRP) in which the +> objective is to optimize the routes of vehicles to serve client demands subject +> to different vehicle profiles, with each having a preference or constraint on a +> per-client basis. While existing learning methods have shown promise for +> solving the HCVRP in real-time, no learning method exists to solve the more +> practical and challenging PVRP. In this paper, we propose a Collaborative +> Attention Model with Profiles (CAMP), a novel approach that learns efficient +> solvers for PVRP using multi-agent reinforcement learning. CAMP employs a +> specialized attention-based encoder architecture to embed profiled client +> embeddings in parallel for each vehicle profile. We design a communication +> layer between agents for collaborative decision-making across profiled +> embeddings at each decoding step and a batched pointer mechanism to attend to +> the profiled embeddings to evaluate the likelihood of the next actions. We +> evaluate CAMP on two variants of PVRPs: PVRP with preferences, which explicitly +> influence the reward function, and PVRP with zone constraints with different +> numbers of agents and clients, demonstrating that our learned solvers achieve +> competitive results compared to both classical state-of-the-art neural +> multi-agent models in terms of solution quality and computational efficiency. +> We make our code openly available at https://github.com/ai4co/camp. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on developing a multi-agent reinforcement learning model (CAMP) for solving vehicle routing problems, with no emphasis on prompt engineering, Large Language Models (LLMs), or the manipulation of textual input prompts for LLM performance. + +--- + +## [CONTINUUM: Detecting APT Attacks through Spatial-Temporal Graph Neural + Networks](https://arxiv.org/abs/http://arxiv.org/abs/2501.02981v2) +**arXiv ID:** http://arxiv.org/abs/2501.02981v2 + +**Abstract:** +> Advanced Persistent Threats (APTs) represent a significant challenge in +> cybersecurity due to their sophisticated and stealthy nature. Traditional +> Intrusion Detection Systems (IDS) often fall short in detecting these +> multi-stage attacks. Recently, Graph Neural Networks (GNNs) have been employed +> to enhance IDS capabilities by analyzing the complex relationships within +> networked data. However, existing GNN-based solutions are hampered by high +> false positive rates and substantial resource consumption. In this paper, we +> present a novel IDS designed to detect APTs using a Spatio-Temporal Graph +> Neural Network Autoencoder. Our approach leverages spatial information to +> understand the interactions between entities within a graph and temporal +> information to capture the evolution of the graph over time. This dual +> perspective is crucial for identifying the sequential stages of APTs. +> Furthermore, to address privacy and scalability concerns, we deploy our +> architecture in a federated learning environment. This setup ensures that local +> data remains on-premise while encrypted model-weights are shared and aggregated +> using homomorphic encryption, maintaining data privacy and security. Our +> evaluation shows that this system effectively detects APTs with lower false +> positive rates and optimized resource usage compared to existing methods, +> highlighting the potential of spatio-temporal analysis and federated learning +> in enhancing cybersecurity defenses. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the 'MUST' criteria as it primarily focuses on detecting APT attacks using Spatio-Temporal Graph Neural Networks in cybersecurity, with no emphasis on engineering, design, or optimization of prompts for Large Language Models (LLMs)." +} + +--- + +## [A Bio-Inspired Research Paradigm of Collision Perception Neurons + Enabling Neuro-Robotic Integration: The LGMD Case](https://arxiv.org/abs/http://arxiv.org/abs/2501.02982v1) +**arXiv ID:** http://arxiv.org/abs/2501.02982v1 + +**Abstract:** +> Compared to human vision, insect visual systems excel at rapid and precise +> collision detection, despite relying on only tens of thousands of neurons +> organized through a few neuropils. This efficiency makes them an attractive +> model system for developing artificial collision-detecting systems. +> Specifically, researchers have identified collision-selective neurons in the +> locust's optic lobe, called lobula giant movement detectors (LGMDs), which +> respond specifically to approaching objects. Research upon LGMD neurons began +> in the early 1970s. Initially, due to their large size, these neurons were +> identified as motion detectors, but their role as looming detectors was +> recognized over time. Since then, progress in neuroscience, computational +> modeling of LGMD's visual neural circuits, and LGMD-based robotics has advanced +> in tandem, each field supporting and driving the others. Today, with a deeper +> understanding of LGMD neurons, LGMD-based models have significantly improved +> collision-free navigation in mobile robots including ground and aerial robots. +> This review highlights recent developments in LGMD research from the +> perspectives of neuroscience, computational modeling, and robotics. It +> emphasizes a biologically plausible research paradigm, where insights from +> neuroscience inform real-world applications, which would in turn validate and +> advance neuroscience. With strong support from extensive research and growing +> application demand, this paradigm has reached a mature stage and demonstrates +> versatility across different areas of neuroscience research, thereby enhancing +> our understanding of the interconnections between neuroscience, computational +> modeling, and robotics. Furthermore, other motion-sensitive neurons have also +> shown promising potential for adopting this research paradigm. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance, instead focusing on bio-inspired robotics, neuroscience, and collision detection using LGMD neurons. + +--- + +## [To Analyze and Regulate Human-in-the-loop Learning for Congestion Games](https://arxiv.org/abs/http://arxiv.org/abs/2501.03055v2) +**arXiv ID:** http://arxiv.org/abs/2501.03055v2 + +**Abstract:** +> In congestion games, selfish users behave myopically to crowd to the shortest +> paths, and the social planner designs mechanisms to regulate such selfish +> routing through information or payment incentives. However, such mechanism +> design requires the knowledge of time-varying traffic conditions and it is the +> users themselves to learn and report past road experiences to the social +> planner (e.g., Waze or Google Maps). When congestion games meet mobile +> crowdsourcing, it is critical to incentivize selfish users to explore +> non-shortest paths in the best exploitation-exploration trade-off. First, we +> consider a simple but fundamental parallel routing network with one +> deterministic path and multiple stochastic paths for users with an average +> arrival probability $\lambda$. We prove that the current myopic routing policy +> (widely used in Waze and Google Maps) misses both exploration (when strong +> hazard belief) and exploitation (when weak hazard belief) as compared to the +> social optimum. Due to the myopic policy's under-exploration, we prove that the +> caused price of anarchy (PoA) is larger than +> \(\frac{1}{1-\rho^{\frac{1}{\lambda}}}\), which can be arbitrarily large as +> discount factor \(\rho\rightarrow1\). To mitigate such huge efficiency loss, we +> propose a novel selective information disclosure (SID) mechanism: we only +> reveal the latest traffic information to users when they intend to over-explore +> stochastic paths upon arrival, while hiding such information when they want to +> under-explore. We prove that our mechanism successfully reduces PoA to be less +> than~\(2\). Besides the parallel routing network, we further extend our +> mechanism and PoA results to any linear path graphs with multiple intermediate +> nodes. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper does not meet the 'MUST' criteria as it does not focus primarily on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Instead, it addresses mechanism design for regulating human-in-the-loop learning in congestion games, with no apparent connection to LLMs or prompt engineering for text-based interactions." +} + +--- + +## [Survival Analysis Revisited: Understanding and Unifying Poisson, + Exponential, and Cox Models in Fall Risk Analysis](https://arxiv.org/abs/http://arxiv.org/abs/2501.03058v1) +**arXiv ID:** http://arxiv.org/abs/2501.03058v1 + +**Abstract:** +> This paper explores foundational and applied aspects of survival analysis, +> using fall risk assessment as a case study. It revisits key time-related +> probability distributions and statistical methods, including logistic +> regression, Poisson regression, Exponential regression, and the Cox +> Proportional Hazards model, offering a unified perspective on their +> relationships within the survival analysis framework. A contribution of this +> work is the step-by-step derivation and clarification of the relationships +> among these models, particularly demonstrating that Poisson regression in the +> survival context is a specific case of the Cox model. These insights address +> gaps in understanding and reinforce the simplicity and interpretability of +> survival models. The paper also emphasizes the practical utility of survival +> analysis by connecting theoretical insights with real-world applications. In +> the context of fall detection, it demonstrates how these models can +> simultaneously predict fall risk, analyze contributing factors, and estimate +> time-to-event outcomes within a single streamlined framework. In contrast, +> advanced deep learning methods often require complex post-hoc interpretation +> and separate training for different tasks particularly when working with +> structured numerical data. This highlights the enduring relevance of classical +> statistical frameworks and makes survival models especially valuable in +> healthcare settings, where explainability and robustness are critical. By +> unifying foundational concepts and offering a cohesive perspective on +> time-to-event analysis, this work serves as an accessible resource for +> understanding survival models and applying them effectively to diverse +> analytical challenges. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it primarily focuses on survival analysis, statistical methods, and their application in healthcare, with no apparent emphasis on prompt engineering, Large Language Models (LLMs), or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Through-The-Mask: Mask-based Motion Trajectories for Image-to-Video + Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.03059v1) +**arXiv ID:** http://arxiv.org/abs/2501.03059v1 + +**Abstract:** +> We consider the task of Image-to-Video (I2V) generation, which involves +> transforming static images into realistic video sequences based on a textual +> description. While recent advancements produce photorealistic outputs, they +> frequently struggle to create videos with accurate and consistent object +> motion, especially in multi-object scenarios. To address these limitations, we +> propose a two-stage compositional framework that decomposes I2V generation +> into: (i) An explicit intermediate representation generation stage, followed by +> (ii) A video generation stage that is conditioned on this representation. Our +> key innovation is the introduction of a mask-based motion trajectory as an +> intermediate representation, that captures both semantic object information and +> motion, enabling an expressive but compact representation of motion and +> semantics. To incorporate the learned representation in the second stage, we +> utilize object-level attention objectives. Specifically, we consider a spatial, +> per-object, masked-cross attention objective, integrating object-specific +> prompts into corresponding latent space regions and a masked spatio-temporal +> self-attention objective, ensuring frame-to-frame consistency for each object. +> We evaluate our method on challenging benchmarks with multi-object and +> high-motion scenarios and empirically demonstrate that the proposed method +> achieves state-of-the-art results in temporal coherence, motion realism, and +> text-prompt faithfulness. Additionally, we introduce \benchmark, a new +> challenging benchmark for single-object and multi-object I2V generation, and +> demonstrate our method's superiority on this benchmark. Project page is +> available at https://guyyariv.github.io/TTM/. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on Image-to-Video (I2V) generation, a generative AI application other than text generation driven by LLMs, and does not centrally concern prompt engineering for text-based interactions with LLMs, despite mentioning object-specific prompts as a component of the methodology. + +--- + +## [The Scaling Law for LoRA Base on Mutual Information Upper Bound](https://arxiv.org/abs/http://arxiv.org/abs/2501.03152v1) +**arXiv ID:** http://arxiv.org/abs/2501.03152v1 + +**Abstract:** +> LoRA (Low-Rank Adaptation) is a widely used model fine-tuning method. In +> fine-tuning, the law among model performance, model parameters, and data +> complexity has been a focal issue in the field. Existing methods often leverage +> external metrics (such as cross-entropy or perplexity) to evaluate model +> performance. In the fine-tuning process for large models, two types of +> knowledge are typically involved: the frozen, general knowledge acquired by the +> model during pre-training and the new knowledge learned through the LoRA module +> from the current data. Generally, the less LoRA's learned knowledge relies on +> the large model, the more it captures the specific knowledge of new data, +> thereby enhancing its adaptability to new tasks. However, external metrics do +> not readily capture the dependency relationship between these two types of +> knowledge. Therefore, we designed an internal metric based on the Mutual +> Information Upper Bound (MIUB) theory to investigate the scaling law of +> large-model LoRA fine-tuning. In our experiments, we validated this approach on +> benchmark datasets, using the Llama3-8B and Phi3-3B models. The results show +> that the proposed MIUB metric aligns more accurately and stably with the +> scaling law of LoRA fine-tuning compared to cross-entropy and perplexity. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on model fine-tuning (LoRA) and evaluating its performance using an internal metric (Mutual Information Upper Bound), rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [Detecting AI-Generated Text in Educational Content: Leveraging Machine + Learning and Explainable AI for Academic Integrity](https://arxiv.org/abs/http://arxiv.org/abs/2501.03203v1) +**arXiv ID:** http://arxiv.org/abs/2501.03203v1 + +**Abstract:** +> This study seeks to enhance academic integrity by providing tools to detect +> AI-generated content in student work using advanced technologies. The findings +> promote transparency and accountability, helping educators maintain ethical +> standards and supporting the responsible integration of AI in education. A key +> contribution of this work is the generation of the CyberHumanAI dataset, which +> has 1000 observations, 500 of which are written by humans and the other 500 +> produced by ChatGPT. We evaluate various machine learning (ML) and deep +> learning (DL) algorithms on the CyberHumanAI dataset comparing human-written +> and AI-generated content from Large Language Models (LLMs) (i.e., ChatGPT). +> Results demonstrate that traditional ML algorithms, specifically XGBoost and +> Random Forest, achieve high performance (83% and 81% accuracies respectively). +> Results also show that classifying shorter content seems to be more challenging +> than classifying longer content. Further, using Explainable Artificial +> Intelligence (XAI) we identify discriminative features influencing the ML +> model's predictions, where human-written content tends to use a practical +> language (e.g., use and allow). Meanwhile AI-generated text is characterized by +> more abstract and formal terms (e.g., realm and employ). Finally, a comparative +> analysis with GPTZero show that our narrowly focused, simple, and fine-tuned +> model can outperform generalized systems like GPTZero. The proposed model +> achieved approximately 77.5% accuracy compared to GPTZero's 48.5% accuracy when +> tasked to classify Pure AI, Pure Human, and mixed class. GPTZero showed a +> tendency to classify challenging and small-content cases as either mixed or +> unrecognized while our proposed model showed a more balanced performance across +> the three classes. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on detecting AI-generated text in educational content using machine learning and Explainable AI, rather than on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), as required by the criteria. + +--- + +## [BoostStep: Boosting mathematical capability of Large Language Models via + improved single-step reasoning](https://arxiv.org/abs/http://arxiv.org/abs/2501.03226v2) +**arXiv ID:** http://arxiv.org/abs/2501.03226v2 + +**Abstract:** +> Cutting-edge large language models (LLMs) demonstrate promising performance +> in solving complex math problems with a divide-and-conquer pipeline and the +> assistance of in-context learning (ICL) examples. However, their potential for +> improvement is limited by two critical problems within their ICL examples: +> granularity-mismatch and the ensuing negative-effect noise problem. +> Specifically, the LLMs are capable of the dividing process yet mostly failed by +> inaccurate reasoning within a few conquer steps, while the ICL examples +> retrieved in question-grained sometimes lack relevant steps for a specific +> challenging reasoning step. Further, this disconnect may hinder the correct +> reasoning due to its irrelevance. To this end, we focus on improving the +> reasoning quality within each step and present BoostStep. BoostStep aligns the +> granularity between the retrieving and reasoning on step grained, and provides +> highly related ICL examples for each reasoning step with a novel `first-try' +> strategy. BoostStep provides more relevant examples than the coarse +> question-grained strategy, enhancing the model reasoning quality within each +> step steadily. BoostStep is a general and robust reasoning-enhancing method +> that not only improves standalone reasoning performance but also integrates +> seamlessly with Monte Carlo Tree Search methods (MCTS) to refine both candidate +> generation and decision-making. Quantitatively, it improves GPT-4o and +> Qwen2.5-Math-72B by 3.6\% and 2.0\% respectively on various mathematical +> benchmarks, and 7.5\% gain combined with MCTS. + +**Decision Explanation:** +Original decision: REJECT +Although the paper improves LLM performance through refined in-context learning (ICL) examples, its primary focus is on enhancing mathematical reasoning capabilities within LLMs via a novel strategy, rather than prompt engineering techniques specifically for text-based interactions with LLMs. + +--- + +## [DPO Kernels: A Semantically-Aware, Kernel-Enhanced, and Divergence-Rich + Paradigm for Direct Preference Optimization](https://arxiv.org/abs/http://arxiv.org/abs/2501.03271v3) +**arXiv ID:** http://arxiv.org/abs/2501.03271v3 + +**Abstract:** +> The rapid rise of large language models (LLMs) has unlocked many applications +> but also underscores the challenge of aligning them with diverse values and +> preferences. Direct Preference Optimization (DPO) is central to alignment but +> constrained by fixed divergences and limited feature transformations. We +> propose DPO-Kernels, which integrates kernel methods to address these issues +> through four key contributions: (i) Kernelized Representations with polynomial, +> RBF, Mahalanobis, and spectral kernels for richer transformations, plus a +> hybrid loss combining embedding-based and probability-based objectives; (ii) +> Divergence Alternatives (Jensen-Shannon, Hellinger, Renyi, Bhattacharyya, +> Wasserstein, and f-divergences) for greater stability; (iii) Data-Driven +> Selection metrics that automatically choose the best kernel-divergence pair; +> and (iv) a Hierarchical Mixture of Kernels for both local precision and global +> modeling. Evaluations on 12 datasets demonstrate state-of-the-art performance +> in factuality, safety, reasoning, and instruction following. Grounded in +> Heavy-Tailed Self-Regularization, DPO-Kernels maintains robust generalization +> for LLMs, offering a comprehensive resource for further alignment research. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on Direct Preference Optimization (DPO) for aligning Large Language Models (LLMs) with diverse values and preferences, rather than prompt engineering specifically for LLMs. While LLMs are mentioned, the core subject is DPO methodology, not the design, optimization, or manipulation of textual input prompts for improving LLM performance." +} + +--- + +## [Revolutionizing Encrypted Traffic Classification with MH-Net: A + Multi-View Heterogeneous Graph Model](https://arxiv.org/abs/http://arxiv.org/abs/2501.03279v1) +**arXiv ID:** http://arxiv.org/abs/2501.03279v1 + +**Abstract:** +> With the growing significance of network security, the classification of +> encrypted traffic has emerged as an urgent challenge. Traditional byte-based +> traffic analysis methods are constrained by the rigid granularity of +> information and fail to fully exploit the diverse correlations between bytes. +> To address these limitations, this paper introduces MH-Net, a novel approach +> for classifying network traffic that leverages multi-view heterogeneous traffic +> graphs to model the intricate relationships between traffic bytes. The essence +> of MH-Net lies in aggregating varying numbers of traffic bits into multiple +> types of traffic units, thereby constructing multi-view traffic graphs with +> diverse information granularities. By accounting for different types of byte +> correlations, such as header-payload relationships, MH-Net further endows the +> traffic graph with heterogeneity, significantly enhancing model performance. +> Notably, we employ contrastive learning in a multi-task manner to strengthen +> the robustness of the learned traffic unit representations. Experiments +> conducted on the ISCX and CIC-IoT datasets for both the packet-level and +> flow-level traffic classification tasks demonstrate that MH-Net achieves the +> best overall performance compared to dozens of SOTA methods. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the primary focus criteria, as it revolves around network traffic classification using a multi-view heterogeneous graph model (MH-Net), with no apparent connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to satisfy the mandatory 'MUST' criteria. + +--- + +## [A Decision-Based Heterogenous Graph Attention Network for Multi-Class + Fake News Detection](https://arxiv.org/abs/http://arxiv.org/abs/2501.03290v1) +**arXiv ID:** http://arxiv.org/abs/2501.03290v1 + +**Abstract:** +> A promising tool for addressing fake news detection is Graph Neural Networks +> (GNNs). However, most existing GNN-based methods rely on binary classification, +> categorizing news as either real or fake. Additionally, traditional GNN models +> use a static neighborhood for each node, making them susceptible to issues like +> over-squashing. In this paper, we introduce a novel model named Decision-based +> Heterogeneous Graph Attention Network (DHGAT) for fake news detection in a +> semi-supervised setting. DHGAT effectively addresses the limitations of +> traditional GNNs by dynamically optimizing and selecting the neighborhood type +> for each node in every layer. It represents news data as a heterogeneous graph +> where nodes (news items) are connected by various types of edges. The +> architecture of DHGAT consists of a decision network that determines the +> optimal neighborhood type and a representation network that updates node +> embeddings based on this selection. As a result, each node learns an optimal +> and task-specific computational graph, enhancing both the accuracy and +> efficiency of the fake news detection process. We evaluate DHGAT on the LIAR +> dataset, a large and challenging dataset for multi-class fake news detection, +> which includes news items categorized into six classes. Our results demonstrate +> that DHGAT outperforms existing methods, improving accuracy by approximately 4% +> and showing robustness with limited labeled data. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a new Graph Neural Network (GNN) architecture for multi-class fake news detection, which does not meet the 'MUST' criteria of primarily focusing on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs) or investigating methods for improving LLM performance through textual input prompt manipulation. + +--- + +## [Multi-Modal One-Shot Federated Ensemble Learning for Medical Data with + Vision Large Language Model](https://arxiv.org/abs/http://arxiv.org/abs/2501.03292v1) +**arXiv ID:** http://arxiv.org/abs/2501.03292v1 + +**Abstract:** +> Federated learning (FL) has attracted considerable interest in the medical +> domain due to its capacity to facilitate collaborative model training while +> maintaining data privacy. However, conventional FL methods typically +> necessitate multiple communication rounds, leading to significant communication +> overhead and delays, especially in environments with limited bandwidth. +> One-shot federated learning addresses these issues by conducting model training +> and aggregation in a single communication round, thereby reducing communication +> costs while preserving privacy. Among these, one-shot federated ensemble +> learning combines independently trained client models using ensemble techniques +> such as voting, further boosting performance in non-IID data scenarios. On the +> other hand, existing machine learning methods in healthcare predominantly use +> unimodal data (e.g., medical images or textual reports), which restricts their +> diagnostic accuracy and comprehensiveness. Therefore, the integration of +> multi-modal data is proposed to address these shortcomings. In this paper, we +> introduce FedMME, an innovative one-shot multi-modal federated ensemble +> learning framework that utilizes multi-modal data for medical image analysis. +> Specifically, FedMME capitalizes on vision large language models to produce +> textual reports from medical images, employs a BERT model to extract textual +> features from these reports, and amalgamates these features with visual +> features to improve diagnostic accuracy. Experimental results show that our +> method demonstrated superior performance compared to existing one-shot +> federated learning methods in healthcare scenarios across four datasets with +> various data distributions. For instance, it surpasses existing one-shot +> federated learning approaches by more than 17.5% in accuracy on the RSNA +> dataset when applying a Dirichlet distribution with ($\alpha$ = 0.3). + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the development of a one-shot federated ensemble learning framework for medical data, using a Vision Large Language Model as a component, rather than focusing specifically on the engineering, design, or optimization of prompts for Large Language Models. + +--- + +## [Rethinking Byzantine Robustness in Federated Recommendation from Sparse + Aggregation Perspective](https://arxiv.org/abs/http://arxiv.org/abs/2501.03301v2) +**arXiv ID:** http://arxiv.org/abs/2501.03301v2 + +**Abstract:** +> To preserve user privacy in recommender systems, federated recommendation +> (FR) based on federated learning (FL) emerges, keeping the personal data on the +> local client and updating a model collaboratively. Unlike FL, FR has a unique +> sparse aggregation mechanism, where the embedding of each item is updated by +> only partial clients, instead of full clients in a dense aggregation of general +> FL. Recently, as an essential principle of FL, model security has received +> increasing attention, especially for Byzantine attacks, where malicious clients +> can send arbitrary updates. The problem of exploring the Byzantine robustness +> of FR is particularly critical since in the domains applying FR, e.g., +> e-commerce, malicious clients can be injected easily by registering new +> accounts. However, existing Byzantine works neglect the unique sparse +> aggregation of FR, making them unsuitable for our problem. Thus, we make the +> first effort to investigate Byzantine attacks on FR from the perspective of +> sparse aggregation, which is non-trivial: it is not clear how to define +> Byzantine robustness under sparse aggregations and design Byzantine attacks +> under limited knowledge/capability. In this paper, we reformulate the Byzantine +> robustness under sparse aggregation by defining the aggregation for a single +> item as the smallest execution unit. Then we propose a family of effective +> attack strategies, named Spattack, which exploit the vulnerability in sparse +> aggregation and are categorized along the adversary's knowledge and capability. +> Extensive experimental results demonstrate that Spattack can effectively +> prevent convergence and even break down defenses under a few malicious clients, +> raising alarms for securing FR systems. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on Byzantine robustness in Federated Recommendation systems, specifically addressing sparse aggregation and security against malicious clients, with no primary concern or investigation into prompt engineering for Large Language Models (LLMs) or their textual input prompts." +} + +--- + +## [Advanced Machine Learning Techniques for Social Support Detection on + Social Media](https://arxiv.org/abs/http://arxiv.org/abs/2501.03370v1) +**arXiv ID:** http://arxiv.org/abs/2501.03370v1 + +**Abstract:** +> The widespread use of social media highlights the need to understand its +> impact, particularly the role of online social support. This study uses a +> dataset focused on online social support, which includes binary and multiclass +> classifications of social support content on social media. The classification +> of social support is divided into three tasks. The first task focuses on +> distinguishing between supportive and non-supportive. The second task aims to +> identify whether the support is directed toward an individual or a group. The +> third task categorizes the specific type of social support, grouping it into +> categories such as Nation, LGBTQ, Black people, Women, Religion, and Other (if +> it does not fit into the previously mentioned categories). To address data +> imbalances in these tasks, we employed K-means clustering for balancing the +> dataset and compared the results with the original unbalanced data. Using +> advanced machine learning techniques, including transformers and zero-shot +> learning approaches with GPT3, GPT4, and GPT4-o, we predict social support +> levels in various contexts. The effectiveness of the dataset is evaluated using +> baseline models across different learning approaches, with transformer-based +> methods demonstrating superior performance. Additionally, we achieved a 0.4\% +> increase in the macro F1 score for the second task and a 0.7\% increase for the +> third task, compared to previous work utilizing traditional machine learning +> with psycholinguistic and unigram-based TF-IDF values. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus primarily on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Instead, it utilizes LLMs (GPT3, GPT4) as components for social support detection on social media, with the primary focus being on advanced machine learning techniques for classification tasks, not prompt engineering. + +--- + +## [License Plate Images Generation with Diffusion Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.03374v1) +**arXiv ID:** http://arxiv.org/abs/2501.03374v1 + +**Abstract:** +> Despite the evident practical importance of license plate recognition (LPR), +> corresponding research is limited by the volume of publicly available datasets +> due to privacy regulations such as the General Data Protection Regulation +> (GDPR). To address this challenge, synthetic data generation has emerged as a +> promising approach. In this paper, we propose to synthesize realistic license +> plates (LPs) using diffusion models, inspired by recent advances in image and +> video generation. In our experiments a diffusion model was successfully trained +> on a Ukrainian LP dataset, and 1000 synthetic images were generated for +> detailed analysis. Through manual classification and annotation of the +> generated images, we performed a thorough study of the model output, such as +> success rate, character distributions, and type of failures. Our contributions +> include experimental validation of the efficacy of diffusion models for LP +> synthesis, along with insights into the characteristics of the generated data. +> Furthermore, we have prepared a synthetic dataset consisting of 10,000 LP +> images, publicly available at https://zenodo.org/doi/10.5281/zenodo.13342102. +> Conducted experiments empirically confirm the usefulness of synthetic data for +> the LPR task. Despite the initial performance gap between the model trained +> with real and synthetic data, the expansion of the training data set with +> pseudolabeled synthetic data leads to an improvement in LPR accuracy by 3% +> compared to baseline. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on generating license plate images using diffusion models, which is an application of generative AI in image generation, not text generation driven by Large Language Models (LLMs), and does not investigate prompt engineering for LLMs. + +--- + +## [Activating Associative Disease-Aware Vision Token Memory for LLM-Based + X-ray Report Generation](https://arxiv.org/abs/http://arxiv.org/abs/2501.03458v1) +**arXiv ID:** http://arxiv.org/abs/2501.03458v1 + +**Abstract:** +> X-ray image based medical report generation achieves significant progress in +> recent years with the help of the large language model, however, these models +> have not fully exploited the effective information in visual image regions, +> resulting in reports that are linguistically sound but insufficient in +> describing key diseases. In this paper, we propose a novel associative +> memory-enhanced X-ray report generation model that effectively mimics the +> process of professional doctors writing medical reports. It considers both the +> mining of global and local visual information and associates historical report +> information to better complete the writing of the current report. Specifically, +> given an X-ray image, we first utilize a classification model along with its +> activation maps to accomplish the mining of visual regions highly associated +> with diseases and the learning of disease query tokens. Then, we employ a +> visual Hopfield network to establish memory associations for disease-related +> tokens, and a report Hopfield network to retrieve report memory information. +> This process facilitates the generation of high-quality reports based on a +> large language model and achieves state-of-the-art performance on multiple +> benchmark datasets, including the IU X-ray, MIMIC-CXR, and Chexpert Plus. The +> source code of this work is released on +> \url{https://github.com/Event-AHU/Medical_Image_Analysis}. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a novel model for X-ray report generation in a medical context, which violates the 'MUST NOT' criteria of being primarily concerned with medical subjects and not focusing on prompt engineering for text-based interactions with LLMs as the core subject. + +--- + +## [Radar Signal Recognition through Self-Supervised Learning and Domain + Adaptation](https://arxiv.org/abs/http://arxiv.org/abs/2501.03461v2) +**arXiv ID:** http://arxiv.org/abs/2501.03461v2 + +**Abstract:** +> Automatic radar signal recognition (RSR) plays a pivotal role in electronic +> warfare (EW), as accurately classifying radar signals is critical for informing +> decision-making processes. Recent advances in deep learning have shown +> significant potential in improving RSR performance in domains with ample +> annotated data. However, these methods fall short in EW scenarios where +> annotated RF data are scarce or impractical to obtain. To address these +> challenges, we introduce a self-supervised learning (SSL) method which utilises +> masked signal modelling and RF domain adaption to enhance RSR performance in +> environments with limited RF samples and labels. Specifically, we investigate +> pre-training masked autoencoders (MAE) on baseband in-phase and quadrature +> (I/Q) signals from various RF domains and subsequently transfer the learned +> representation to the radar domain, where annotated data are limited. Empirical +> results show that our lightweight self-supervised ResNet model with domain +> adaptation achieves up to a 17.5% improvement in 1-shot classification accuracy +> when pre-trained on in-domain signals (i.e., radar signals) and up to a 16.31% +> improvement when pre-trained on out-of-domain signals (i.e., comm signals), +> compared to its baseline without SSL. We also provide reference results for +> several MAE designs and pre-training strategies, establishing a new benchmark +> for few-shot radar signal classification. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on self-supervised learning and domain adaptation for radar signal recognition, with no discussion on prompt engineering, Large Language Models (LLMs), or textual input prompts, thus failing to meet all 'MUST' criteria. + +--- + +## [KG-TRICK: Unifying Textual and Relational Information Completion of + Knowledge for Multilingual Knowledge Graphs](https://arxiv.org/abs/http://arxiv.org/abs/2501.03560v1) +**arXiv ID:** http://arxiv.org/abs/2501.03560v1 + +**Abstract:** +> Multilingual knowledge graphs (KGs) provide high-quality relational and +> textual information for various NLP applications, but they are often +> incomplete, especially in non-English languages. Previous research has shown +> that combining information from KGs in different languages aids either +> Knowledge Graph Completion (KGC), the task of predicting missing relations +> between entities, or Knowledge Graph Enhancement (KGE), the task of predicting +> missing textual information for entities. Although previous efforts have +> considered KGC and KGE as independent tasks, we hypothesize that they are +> interdependent and mutually beneficial. To this end, we introduce KG-TRICK, a +> novel sequence-to-sequence framework that unifies the tasks of textual and +> relational information completion for multilingual KGs. KG-TRICK demonstrates +> that: i) it is possible to unify the tasks of KGC and KGE into a single +> framework, and ii) combining textual information from multiple languages is +> beneficial to improve the completeness of a KG. As part of our contributions, +> we also introduce WikiKGE10++, the largest manually-curated benchmark for +> textual information completion of KGs, which features over 25,000 entities +> across 10 diverse languages. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on unifying Knowledge Graph Completion and Enhancement tasks for multilingual Knowledge Graphs, utilizing a sequence-to-sequence framework, without explicit concentration on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [From Code to Compliance: Assessing ChatGPT's Utility in Designing an + Accessible Webpage -- A Case Study](https://arxiv.org/abs/http://arxiv.org/abs/2501.03572v1) +**arXiv ID:** http://arxiv.org/abs/2501.03572v1 + +**Abstract:** +> Web accessibility ensures that individuals with disabilities can access and +> interact with digital content without barriers, yet a significant majority of +> most used websites fail to meet accessibility standards. This study evaluates +> ChatGPT's (GPT-4o) ability to generate and improve web pages in line with Web +> Content Accessibility Guidelines (WCAG). While ChatGPT can effectively address +> accessibility issues when prompted, its default code often lacks compliance, +> reflecting limitations in its training data and prevailing inaccessible web +> practices. Automated and manual testing revealed strengths in resolving simple +> issues but challenges with complex tasks, requiring human oversight and +> additional iterations. Unlike prior studies, we incorporate manual evaluation, +> dynamic elements, and use the visual reasoning capability of ChatGPT along with +> the prompts to fix accessibility issues. Providing screenshots alongside +> prompts enhances the LLM's ability to address accessibility issues by allowing +> it to analyze surrounding components, such as determining appropriate contrast +> colors. We found that effective prompt engineering, such as providing concise, +> structured feedback and incorporating visual aids, significantly enhances +> ChatGPT's performance. These findings highlight the potential and limitations +> of large language models for accessible web development, offering practical +> guidance for developers to create more inclusive websites. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper's primary focus is leveraging ChatGPT for accessible web development, not prompt engineering for Large Language Models. While it mentions effective prompt engineering techniques, this is secondary to its main objective of improving web accessibility." +} + +--- + +## [Action Quality Assessment via Hierarchical Pose-guided Multi-stage + Contrastive Regression](https://arxiv.org/abs/http://arxiv.org/abs/2501.03674v1) +**arXiv ID:** http://arxiv.org/abs/2501.03674v1 + +**Abstract:** +> Action Quality Assessment (AQA), which aims at automatic and fair evaluation +> of athletic performance, has gained increasing attention in recent years. +> However, athletes are often in rapid movement and the corresponding visual +> appearance variances are subtle, making it challenging to capture fine-grained +> pose differences and leading to poor estimation performance. Furthermore, most +> common AQA tasks, such as diving in sports, are usually divided into multiple +> sub-actions, each of which contains different durations. However, existing +> methods focus on segmenting the video into fixed frames, which disrupts the +> temporal continuity of sub-actions resulting in unavoidable prediction errors. +> To address these challenges, we propose a novel action quality assessment +> method through hierarchically pose-guided multi-stage contrastive regression. +> Firstly, we introduce a multi-scale dynamic visual-skeleton encoder to capture +> fine-grained spatio-temporal visual and skeletal features. Then, a procedure +> segmentation network is introduced to separate different sub-actions and obtain +> segmented features. Afterwards, the segmented visual and skeletal features are +> both fed into a multi-modal fusion module as physics structural priors, to +> guide the model in learning refined activity similarities and variances. +> Finally, a multi-stage contrastive learning regression approach is employed to +> learn discriminative representations and output prediction results. In +> addition, we introduce a newly-annotated FineDiving-Pose Dataset to improve the +> current low-quality human pose labels. In experiments, the results on +> FineDiving and MTL-AQA datasets demonstrate the effectiveness and superiority +> of our proposed approach. Our source code and dataset are available at +> https://github.com/Lumos0507/HP-MCoRe. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on Action Quality Assessment using computer vision and pose-guided contrastive regression, with no mention of Large Language Models (LLMs), prompt engineering, or text-based interactions, thus failing to meet all 'MUST' criteria. + +--- + +## [MAJL: A Model-Agnostic Joint Learning Framework for Music Source + Separation and Pitch Estimation](https://arxiv.org/abs/http://arxiv.org/abs/2501.03689v1) +**arXiv ID:** http://arxiv.org/abs/2501.03689v1 + +**Abstract:** +> Music source separation and pitch estimation are two vital tasks in music +> information retrieval. Typically, the input of pitch estimation is obtained +> from the output of music source separation. Therefore, existing methods have +> tried to perform these two tasks simultaneously, so as to leverage the mutually +> beneficial relationship between both tasks. However, these methods still face +> two critical challenges that limit the improvement of both tasks: the lack of +> labeled data and joint learning optimization. To address these challenges, we +> propose a Model-Agnostic Joint Learning (MAJL) framework for both tasks. MAJL +> is a generic framework and can use variant models for each task. It includes a +> two-stage training method and a dynamic weighting method named Dynamic Weights +> on Hard Samples (DWHS), which addresses the lack of labeled data and joint +> learning optimization, respectively. Experimental results on public music +> datasets show that MAJL outperforms state-of-the-art methods on both tasks, +> with significant improvements of 0.92 in Signal-to-Distortion Ratio (SDR) for +> music source separation and 2.71% in Raw Pitch Accuracy (RPA) for pitch +> estimation. Furthermore, comprehensive studies not only validate the +> effectiveness of each component of MAJL, but also indicate the great generality +> of MAJL in adapting to different model architectures. + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it focuses on music source separation and pitch estimation, leveraging joint learning frameworks, without any mention of Large Language Models (LLMs), prompt engineering, or text-based interactions, thus falling outside the specified scope. + +--- + +## [AuxDepthNet: Real-Time Monocular 3D Object Detection with + Depth-Sensitive Features](https://arxiv.org/abs/http://arxiv.org/abs/2501.03700v1) +**arXiv ID:** http://arxiv.org/abs/2501.03700v1 + +**Abstract:** +> Monocular 3D object detection is a challenging task in autonomous systems due +> to the lack of explicit depth information in single-view images. Existing +> methods often depend on external depth estimators or expensive sensors, which +> increase computational complexity and hinder real-time performance. To overcome +> these limitations, we propose AuxDepthNet, an efficient framework for real-time +> monocular 3D object detection that eliminates the reliance on external depth +> maps or pre-trained depth models. AuxDepthNet introduces two key components: +> the Auxiliary Depth Feature (ADF) module, which implicitly learns +> depth-sensitive features to improve spatial reasoning and computational +> efficiency, and the Depth Position Mapping (DPM) module, which embeds depth +> positional information directly into the detection process to enable accurate +> object localization and 3D bounding box regression. Leveraging the DepthFusion +> Transformer architecture, AuxDepthNet globally integrates visual and +> depth-sensitive features through depth-guided interactions, ensuring robust and +> efficient detection. Extensive experiments on the KITTI dataset show that +> AuxDepthNet achieves state-of-the-art performance, with $\text{AP}_{3D}$ scores +> of 24.72\% (Easy), 18.63\% (Moderate), and 15.31\% (Hard), and +> $\text{AP}_{\text{BEV}}$ scores of 34.11\% (Easy), 25.18\% (Moderate), and +> 21.90\% (Hard) at an IoU threshold of 0.7. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on monocular 3D object detection for autonomous systems, involving depth estimation and sensor technologies, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions, thus failing all 'MUST' criteria. + +--- + +## [Self-adaptive vision-language model for 3D segmentation of pulmonary + artery and vein](https://arxiv.org/abs/http://arxiv.org/abs/2501.03722v1) +**arXiv ID:** http://arxiv.org/abs/2501.03722v1 + +**Abstract:** +> Accurate segmentation of pulmonary structures iscrucial in clinical +> diagnosis, disease study, and treatment planning. Significant progress has been +> made in deep learning-based segmentation techniques, but most require much +> labeled data for training. Consequently, developing precise segmentation +> methods that demand fewer labeled datasets is paramount in medical image +> analysis. The emergence of pre-trained vision-language foundation models, such +> as CLIP, recently opened the door for universal computer vision tasks. +> Exploiting the generalization ability of these pre-trained foundation models on +> downstream tasks, such as segmentation, leads to unexpected performance with a +> relatively small amount of labeled data. However, exploring these models for +> pulmonary artery-vein segmentation is still limited. This paper proposes a +> novel framework called Language-guided self-adaptive Cross-Attention Fusion +> Framework. Our method adopts pre-trained CLIP as a strong feature extractor for +> generating the segmentation of 3D CT scans, while adaptively aggregating the +> cross-modality of text and image representations. We propose a s pecially +> designed adapter module to fine-tune pre-trained CLIP with a self-adaptive +> learning strategy to effectively fuse the two modalities of embeddings. We +> extensively validate our method on a local dataset, which is the largest +> pulmonary artery-vein CT dataset to date and consists of 718 labeled data in +> total. The experiments show that our method outperformed other state-of-the-art +> methods by a large margin. Our data and code will be made publicly available +> upon acceptance. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on medical image analysis (3D segmentation of pulmonary artery and vein) and the development of a vision-language model, rather than the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs), and does not demonstrate the impact of textual input prompts on LLM output. + +--- + +## [Deep Sylvester Posterior Inference for Adaptive Compressed Sensing in + Ultrasound Imaging](https://arxiv.org/abs/http://arxiv.org/abs/2501.03825v1) +**arXiv ID:** http://arxiv.org/abs/2501.03825v1 + +**Abstract:** +> Ultrasound images are commonly formed by sequential acquisition of +> beam-steered scan-lines. Minimizing the number of required scan-lines can +> significantly enhance frame rate, field of view, energy efficiency, and data +> transfer speeds. Existing approaches typically use static subsampling schemes +> in combination with sparsity-based or, more recently, deep-learning-based +> recovery. In this work, we introduce an adaptive subsampling method that +> maximizes intrinsic information gain in-situ, employing a Sylvester Normalizing +> Flow encoder to infer an approximate Bayesian posterior under partial +> observation in real-time. Using the Bayesian posterior and a deep generative +> model for future observations, we determine the subsampling scheme that +> maximizes the mutual information between the subsampled observations, and the +> next frame of the video. We evaluate our approach using the EchoNet cardiac +> ultrasound video dataset and demonstrate that our active sampling method +> outperforms competitive baselines, including uniform and variable-density +> random sampling, as well as equidistantly spaced scan-lines, improving mean +> absolute reconstruction error by 15%. Moreover, posterior inference and the +> sampling scheme generation are performed in just 0.015 seconds (66Hz), making +> it fast enough for real-time 2D ultrasound imaging applications. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on adaptive compressed sensing in ultrasound imaging, employing deep learning for image reconstruction, and does not investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through prompt engineering, thus failing to meet all 'MUST' criteria. + +--- + +## [TACLR: A Scalable and Efficient Retrieval-based Method for Industrial + Product Attribute Value Identification](https://arxiv.org/abs/http://arxiv.org/abs/2501.03835v1) +**arXiv ID:** http://arxiv.org/abs/2501.03835v1 + +**Abstract:** +> Product Attribute Value Identification (PAVI) involves identifying attribute +> values from product profiles, a key task for improving product search, +> recommendations, and business analytics on e-commerce platforms. However, +> existing PAVI methods face critical challenges, such as inferring implicit +> values, handling out-of-distribution (OOD) values, and producing normalized +> outputs. To address these limitations, we introduce Taxonomy-Aware Contrastive +> Learning Retrieval (TACLR), the first retrieval-based method for PAVI. TACLR +> formulates PAVI as an information retrieval task by encoding product profiles +> and candidate values into embeddings and retrieving values based on their +> similarity to the item embedding. It leverages contrastive training with +> taxonomy-aware hard negative sampling and employs adaptive inference with +> dynamic thresholds. TACLR offers three key advantages: (1) it effectively +> handles implicit and OOD values while producing normalized outputs; (2) it +> scales to thousands of categories, tens of thousands of attributes, and +> millions of values; and (3) it supports efficient inference for high-load +> industrial scenarios. Extensive experiments on proprietary and public datasets +> validate the effectiveness and efficiency of TACLR. Moreover, it has been +> successfully deployed in a real-world e-commerce platform, processing millions +> of product listings daily while supporting dynamic, large-scale attribute +> taxonomies. + +**Decision Explanation:** +Original decision: REJECT +The paper does not focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs). Instead, it introduces a retrieval-based method (TACLR) for Product Attribute Value Identification, primarily concerned with information retrieval and taxonomy-aware contrastive learning, without any central focus on prompt engineering for LLMs. + +--- + +## [Diffusion as Shader: 3D-aware Video Diffusion for Versatile Video + Generation Control](https://arxiv.org/abs/http://arxiv.org/abs/2501.03847v2) +**arXiv ID:** http://arxiv.org/abs/2501.03847v2 + +**Abstract:** +> Diffusion models have demonstrated impressive performance in generating +> high-quality videos from text prompts or images. However, precise control over +> the video generation process, such as camera manipulation or content editing, +> remains a significant challenge. Existing methods for controlled video +> generation are typically limited to a single control type, lacking the +> flexibility to handle diverse control demands. In this paper, we introduce +> Diffusion as Shader (DaS), a novel approach that supports multiple video +> control tasks within a unified architecture. Our key insight is that achieving +> versatile video control necessitates leveraging 3D control signals, as videos +> are fundamentally 2D renderings of dynamic 3D content. Unlike prior methods +> limited to 2D control signals, DaS leverages 3D tracking videos as control +> inputs, making the video diffusion process inherently 3D-aware. This innovation +> allows DaS to achieve a wide range of video controls by simply manipulating the +> 3D tracking videos. A further advantage of using 3D tracking videos is their +> ability to effectively link frames, significantly enhancing the temporal +> consistency of the generated videos. With just 3 days of fine-tuning on 8 H800 +> GPUs using less than 10k videos, DaS demonstrates strong control capabilities +> across diverse tasks, including mesh-to-video generation, camera control, +> motion transfer, and object manipulation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on video generation control using diffusion models, which falls under image/video generation rather than text generation driven by Large Language Models (LLMs), violating the 'MUST NOT' criteria related to applications of generative AI. + +--- + +## [CL3DOR: Contrastive Learning for 3D Large Multimodal Models via Odds + Ratio on High-Resolution Point Clouds](https://arxiv.org/abs/http://arxiv.org/abs/2501.03879v1) +**arXiv ID:** http://arxiv.org/abs/2501.03879v1 + +**Abstract:** +> Recent research has demonstrated that Large Language Models (LLMs) are not +> limited to text-only tasks but can also function as multimodal models across +> various modalities, including audio, images, and videos. In particular, +> research on 3D Large Multimodal Models (3D LMMs) is making notable strides, +> driven by the potential of processing higher-dimensional data like point +> clouds. However, upon closer examination, we find that the visual and textual +> content within each sample of existing training datasets lacks both high +> informational granularity and clarity, which serve as a bottleneck for precise +> cross-modal understanding. To address these issues, we propose CL3DOR, +> Contrastive Learning for 3D large multimodal models via Odds ratio on +> high-Resolution point clouds, designed to ensure greater specificity and +> clarity in both visual and textual content. Specifically, we increase the +> density of point clouds per object and construct informative hard negative +> responses in the training dataset to penalize unwanted responses. To leverage +> hard negative responses, we incorporate the odds ratio as an auxiliary term for +> contrastive learning into the conventional language modeling loss. CL3DOR +> achieves state-of-the-art performance in 3D scene understanding and reasoning +> benchmarks. Additionally, we demonstrate the effectiveness of CL3DOR's key +> components through extensive experiments. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a new multimodal learning method (CL3DOR) for 3D Large Multimodal Models, which includes but is not centered on prompt engineering for Large Language Models (LLMs). The core contribution is the contrastive learning approach for multimodal understanding, not the manipulation of textual input prompts to improve LLM performance. + +--- + +## [LLaVA-Mini: Efficient Image and Video Large Multimodal Models with One + Vision Token](https://arxiv.org/abs/http://arxiv.org/abs/2501.03895v1) +**arXiv ID:** http://arxiv.org/abs/2501.03895v1 + +**Abstract:** +> The advent of real-time large multimodal models (LMMs) like GPT-4o has +> sparked considerable interest in efficient LMMs. LMM frameworks typically +> encode visual inputs into vision tokens (continuous representations) and +> integrate them and textual instructions into the context of large language +> models (LLMs), where large-scale parameters and numerous context tokens +> (predominantly vision tokens) result in substantial computational overhead. +> Previous efforts towards efficient LMMs always focus on replacing the LLM +> backbone with smaller models, while neglecting the crucial issue of token +> quantity. In this paper, we introduce LLaVA-Mini, an efficient LMM with minimal +> vision tokens. To achieve a high compression ratio of vision tokens while +> preserving visual information, we first analyze how LMMs understand vision +> tokens and find that most vision tokens only play a crucial role in the early +> layers of LLM backbone, where they mainly fuse visual information into text +> tokens. Building on this finding, LLaVA-Mini introduces modality pre-fusion to +> fuse visual information into text tokens in advance, thereby facilitating the +> extreme compression of vision tokens fed to LLM backbone into one token. +> LLaVA-Mini is a unified large multimodal model that can support the +> understanding of images, high-resolution images, and videos in an efficient +> manner. Experiments across 11 image-based and 7 video-based benchmarks +> demonstrate that LLaVA-Mini outperforms LLaVA-v1.5 with just 1 vision token +> instead of 576. Efficiency analyses reveal that LLaVA-Mini can reduce FLOPs by +> 77%, deliver low-latency responses within 40 milliseconds, and process over +> 10,000 frames of video on the GPU hardware with 24GB of memory. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on developing an efficient Large Multimodal Model (LMM) for image and video processing, optimizing vision tokens, and reducing computational overhead, rather than engineering or optimizing prompts specifically for Large Language Models (LLMs) and their text-based interactions." +} + +--- + +## [Explainable Time Series Prediction of Tyre Energy in Formula One Race + Strategy](https://arxiv.org/abs/http://arxiv.org/abs/2501.04067v1) +**arXiv ID:** http://arxiv.org/abs/2501.04067v1 + +**Abstract:** +> Formula One (F1) race strategy takes place in a high-pressure and fast-paced +> environment where split-second decisions can drastically affect race results. +> Two of the core decisions of race strategy are when to make pit stops (i.e. +> replace the cars' tyres) and which tyre compounds (hard, medium or soft, in +> normal conditions) to select. The optimal pit stop decisions can be determined +> by estimating the tyre degradation of these compounds, which in turn can be +> computed from the energy applied to each tyre, i.e. the tyre energy. In this +> work, we trained deep learning models, using the Mercedes-AMG PETRONAS F1 +> team's historic race data consisting of telemetry, to forecast tyre energies +> during races. Additionally, we fitted XGBoost, a decision tree-based machine +> learning algorithm, to the same dataset and compared the results, with both +> giving impressive performance. Furthermore, we incorporated two different +> explainable AI methods, namely feature importance and counterfactual +> explanations, to gain insights into the reasoning behind the forecasts. Our +> contributions thus result in an explainable, automated method which could +> assist F1 teams in optimising their race strategy. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing an explainable AI method for time series prediction in Formula One racing, utilizing deep learning models and XGBoost, with no primary emphasis on the engineering, design, or optimization of prompts specifically for Large Language Models (LLMs). + +--- + +## [Explainable Reinforcement Learning for Formula One Race Strategy](https://arxiv.org/abs/http://arxiv.org/abs/2501.04068v1) +**arXiv ID:** http://arxiv.org/abs/2501.04068v1 + +**Abstract:** +> In Formula One, teams compete to develop their cars and achieve the highest +> possible finishing position in each race. During a race, however, teams are +> unable to alter the car, so they must improve their cars' finishing positions +> via race strategy, i.e. optimising their selection of which tyre compounds to +> put on the car and when to do so. In this work, we introduce a reinforcement +> learning model, RSRL (Race Strategy Reinforcement Learning), to control race +> strategies in simulations, offering a faster alternative to the industry +> standard of hard-coded and Monte Carlo-based race strategies. Controlling cars +> with a pace equating to an expected finishing position of P5.5 (where P1 +> represents first place and P20 is last place), RSRL achieves an average +> finishing position of P5.33 on our test race, the 2023 Bahrain Grand Prix, +> outperforming the best baseline of P5.63. We then demonstrate, in a +> generalisability study, how performance for one track or multiple tracks can be +> prioritised via training. Further, we supplement model predictions with feature +> importance, decision tree-based surrogate models, and decision tree +> counterfactuals towards improving user trust in the model. Finally, we provide +> illustrations which exemplify our approach in real-world situations, drawing +> parallels between simulations and reality. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a reinforcement learning model for optimizing Formula One race strategies, with no apparent connection to Large Language Models (LLMs), prompt engineering, or text generation, thus failing to meet the primary 'MUST' criteria. + +--- + +## [Multi-armed Bandit and Backbone boost Lin-Kernighan-Helsgaun Algorithm + for the Traveling Salesman Problems](https://arxiv.org/abs/http://arxiv.org/abs/2501.04072v1) +**arXiv ID:** http://arxiv.org/abs/2501.04072v1 + +**Abstract:** +> The Lin-Kernighan-Helsguan (LKH) heuristic is a classic local search +> algorithm for the Traveling Salesman Problem (TSP). LKH introduces an +> $\alpha$-value to replace the traditional distance metric for evaluating the +> edge quality, which leads to a significant improvement. However, we observe +> that the $\alpha$-value does not make full use of the historical information +> during the search, and single guiding information often makes LKH hard to +> escape from some local optima. To address the above issues, we propose a novel +> way to extract backbone information during the TSP local search process, which +> is dynamic and can be updated once a local optimal solution is found. We +> further propose to combine backbone information, $\alpha$-value, and distance +> to evaluate the edge quality so as to guide the search. Moreover, we abstract +> their different combinations to arms in a multi-armed bandit (MAB) and use an +> MAB model to help the algorithm select an appropriate evaluation metric +> dynamically. Both the backbone information and MAB can provide diverse guiding +> information and learn from the search history to suggest the best metric. We +> apply our methods to LKH and LKH-3, which is an extension version of LKH that +> can be used to solve about 40 variant problems of TSP and Vehicle Routing +> Problem (VRP). Extensive experiments show the excellent performance and +> generalization capability of our proposed method, significantly improving LKH +> for TSP and LKH-3 for two representative TSP and VRP variants, the Colored TSP +> (CTSP) and Capacitated VRP with Time Windows (CVRPTW). + +**Decision Explanation:** +Original decision: REJECT +The paper does not meet the 'MUST' criteria as it focuses on optimizing the Traveling Salesman Problem algorithm using multi-armed bandit and backbone boost, with no primary focus on the engineering, design, or optimization of prompts for Large Language Models (LLMs), nor does it investigate the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Enhancing Distribution and Label Consistency for Graph + Out-of-Distribution Generalization](https://arxiv.org/abs/http://arxiv.org/abs/2501.04102v1) +**arXiv ID:** http://arxiv.org/abs/2501.04102v1 + +**Abstract:** +> To deal with distribution shifts in graph data, various graph +> out-of-distribution (OOD) generalization techniques have been recently +> proposed. These methods often employ a two-step strategy that first creates +> augmented environments and subsequently identifies invariant subgraphs to +> improve generalizability. Nevertheless, this approach could be suboptimal from +> the perspective of consistency. First, the process of augmenting environments +> by altering the graphs while preserving labels may lead to graphs that are not +> realistic or meaningfully related to the origin distribution, thus lacking +> distribution consistency. Second, the extracted subgraphs are obtained from +> directly modifying graphs, and may not necessarily maintain a consistent +> predictive relationship with their labels, thereby impacting label consistency. +> In response to these challenges, we introduce an innovative approach that aims +> to enhance these two types of consistency for graph OOD generalization. We +> propose a modifier to obtain both augmented and invariant graphs in a unified +> manner. With the augmented graphs, we enrich the training data without +> compromising the integrity of label-graph relationships. The label consistency +> enhancement in our framework further preserves the supervision information in +> the invariant graph. We conduct extensive experiments on real-world datasets to +> demonstrate the superiority of our framework over other state-of-the-art +> baselines. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on enhancing distribution and label consistency for graph out-of-distribution generalization, with no apparent connection to Large Language Models (LLMs), prompt engineering, or text-based interactions, thus failing to meet the 'MUST' criteria. + +--- + +## [Multimodal Multihop Source Retrieval for Web Question Answering](https://arxiv.org/abs/http://arxiv.org/abs/2501.04173v1) +**arXiv ID:** http://arxiv.org/abs/2501.04173v1 + +**Abstract:** +> This work deals with the challenge of learning and reasoning over multi-modal +> multi-hop question answering (QA). We propose a graph reasoning network based +> on the semantic structure of the sentences to learn multi-source reasoning +> paths and find the supporting facts across both image and text modalities for +> answering the question. In this paper, we investigate the importance of graph +> structure for multi-modal multi-hop question answering. Our analysis is +> centered on WebQA. We construct a strong baseline model, that finds relevant +> sources using a pairwise classification task. We establish that, with the +> proper use of feature representations from pre-trained models, graph structure +> helps in improving multi-modal multi-hop question answering. We point out that +> both graph structure and adjacency matrix are task-related prior knowledge, and +> graph structure can be leveraged to improve the retrieval performance for the +> task. Experiments and visualized analysis demonstrate that message propagation +> over graph networks or the entire graph structure can replace massive +> multimodal transformers with token-wise cross-attention. We demonstrated the +> applicability of our method and show a performance gain of \textbf{4.6$\%$} +> retrieval F1score over the transformer baselines, despite being a very light +> model. We further demonstrated the applicability of our model to a large scale +> retrieval setting. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on multimodal multihop question answering using graph reasoning networks, with emphasis on leveraging image and text modalities, rather than specifically engineering prompts for Large Language Models (LLMs). It does not investigate, analyze, or propose methods for improving LLM performance through the manipulation of textual input prompts as required. + +--- + +## [HIVEX: A High-Impact Environment Suite for Multi-Agent Research + (extended version)](https://arxiv.org/abs/http://arxiv.org/abs/2501.04180v2) +**arXiv ID:** http://arxiv.org/abs/2501.04180v2 + +**Abstract:** +> Games have been vital test beds for the rapid development of Agent-based +> research. Remarkable progress has been achieved in the past, but it is unclear +> if the findings equip for real-world problems. While pressure grows, some of +> the most critical ecological challenges can find mitigation and prevention +> solutions through technology and its applications. Most real-world domains +> include multi-agent scenarios and require machine-machine and human-machine +> collaboration. Open-source environments have not advanced and are often toy +> scenarios, too abstract or not suitable for multi-agent research. By mimicking +> real-world problems and increasing the complexity of environments, we hope to +> advance state-of-the-art multi-agent research and inspire researchers to work +> on immediate real-world problems. Here, we present HIVEX, an environment suite +> to benchmark multi-agent research focusing on ecological challenges. HIVEX +> includes the following environments: Wind Farm Control, Wildfire Resource +> Management, Drone-Based Reforestation, Ocean Plastic Collection, and Aerial +> Wildfire Suppression. We provide environments, training examples, and baselines +> for the main and sub-tasks. All trained models resulting from the experiments +> of this work are hosted on Hugging Face. We also provide a leaderboard on +> Hugging Face and encourage the community to submit models trained on our +> environment suite. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on multi-agent research, ecological challenges, and the development of an environment suite (HIVEX), with no clear emphasis on prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance. + +--- + +## [Fixed Points of Deep Neural Networks: Emergence, Stability, and + Applications](https://arxiv.org/abs/http://arxiv.org/abs/2501.04182v1) +**arXiv ID:** http://arxiv.org/abs/2501.04182v1 + +**Abstract:** +> We present numerical and analytical results on the formation and stability of +> a family of fixed points of deep neural networks (DNNs). Such fixed points +> appear in a class of DNNs when dimensions of input and output vectors are the +> same. We demonstrate examples of applications of such networks in supervised, +> semi-supervised and unsupervised learning such as encoding/decoding of images, +> restoration of damaged images among others. +> We present several numerical and analytical results. First, we show that for +> untrained DNN's with weights and biases initialized by normally distributed +> random variables the only one fixed point exists. This result holds for DNN +> with any depth (number of layers) $L$, any layer width $N$, and sigmoid-type +> activation functions. Second, it has been shown that for a DNN whose parameters +> (weights and biases) are initialized by ``light-tailed'' distribution of +> weights (e.g. normal distribution), after training the distribution of these +> parameters become ``heavy-tailed''. This motivates our study of DNNs with +> ``heavy-tailed'' initialization. For such DNNs we show numerically %existence +> and stability that training leads to emergence of $Q(N,L)$ fixed points, where +> $Q(N,L)$ is a positive integer which depends on the number of layers $L$ and +> layer width $N$. We further observe numerically that for fixed $N = N_0$ the +> function $Q(N_0, L)$ is non-monotone, that is it initially grows as $L$ +> increases and then decreases to 1. +> This non-monotone behavior of $Q(N_0, L)$ is also obtained by analytical +> derivation of equation for Empirical Spectral Distribution (ESD) of +> input-output Jacobian followed by numerical solution of this equation. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on the analysis of fixed points in deep neural networks (DNNs), including their emergence, stability, and applications in image processing, but does not address prompt engineering for Large Language Models (LLMs) or the manipulation of textual input prompts to improve LLM performance, thus failing to meet all 'MUST' criteria. + +--- + +## [Generative Style Transfer for MRI Image Segmentation: A Case of Glioma + Segmentation in Sub-Saharan Africa](https://arxiv.org/abs/http://arxiv.org/abs/2501.04734v1) +**arXiv ID:** http://arxiv.org/abs/2501.04734v1 + +**Abstract:** +> In Sub-Saharan Africa (SSA), the utilization of lower-quality Magnetic +> Resonance Imaging (MRI) technology raises questions about the applicability of +> machine learning methods for clinical tasks. This study aims to provide a +> robust deep learning-based brain tumor segmentation (BraTS) method tailored for +> the SSA population using a threefold approach. Firstly, the impact of domain +> shift from the SSA training data on model efficacy was examined, revealing no +> significant effect. Secondly, a comparative analysis of 3D and 2D +> full-resolution models using the nnU-Net framework indicates similar +> performance of both the models trained for 300 epochs achieving a five-fold +> cross-validation score of 0.93. Lastly, addressing the performance gap observed +> in SSA validation as opposed to the relatively larger BraTS glioma (GLI) +> validation set, two strategies are proposed: fine-tuning SSA cases using the +> GLI+SSA best-pretrained 2D fullres model at 300 epochs, and introducing a novel +> neural style transfer-based data augmentation technique for the SSA cases. This +> investigation underscores the potential of enhancing brain tumor prediction +> within SSA's unique healthcare landscape. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on MRI image segmentation, a non-text generation task, using deep learning methods, and does not investigate, analyze, or propose methods for improving Large Language Model (LLM) performance through prompt engineering, violating multiple 'MUST NOT' criteria. + +--- + +## [Retrieval-Augmented Generation by Evidence Retroactivity in LLMs](https://arxiv.org/abs/http://arxiv.org/abs/2501.05475v1) +**arXiv ID:** http://arxiv.org/abs/2501.05475v1 + +**Abstract:** +> Retrieval-augmented generation has gained significant attention due to its +> ability to integrate relevant external knowledge, enhancing the accuracy and +> reliability of the LLMs' responses. Most of the existing methods apply a +> dynamic multiple retrieval-generating process, to address multi-hop complex +> questions by decomposing them into sub-problems. However, these methods rely on +> an unidirectional forward reasoning paradigm, where errors from insufficient +> reasoning steps or inherent flaws in current retrieval systems are +> irreversible, potentially derailing the entire reasoning chain. For the first +> time, this work introduces Retroactive Retrieval-Augmented Generation +> (RetroRAG), a novel framework to build a retroactive reasoning paradigm. +> RetroRAG revises and updates the evidence, redirecting the reasoning chain to +> the correct direction. RetroRAG constructs an evidence-collation-discovery +> framework to search, generate, and refine credible evidence. It synthesizes +> inferential evidence related to the key entities in the question from the +> existing source knowledge and formulates search queries to uncover additional +> information. As new evidence is found, RetroRAG continually updates and +> organizes this information, enhancing its ability to locate further necessary +> evidence. Paired with an Answerer to generate and evaluate outputs, RetroRAG is +> capable of refining its reasoning process iteratively until a reliable answer +> is obtained. Empirical evaluations show that RetroRAG significantly outperforms +> existing methods. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on introducing a novel framework for retrieval-augmented generation with a retroactive reasoning paradigm, rather than specifically engineering or optimizing prompts for Large Language Models (LLMs). While LLMs are utilized, the core subject is the enhancement of their reliability through external knowledge integration, not prompt engineering for text-based interactions. + +--- + +## [Language and Planning in Robotic Navigation: A Multilingual Evaluation + of State-of-the-Art Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.05478v1) +**arXiv ID:** http://arxiv.org/abs/2501.05478v1 + +**Abstract:** +> Large Language Models (LLMs) such as GPT-4, trained on huge amount of +> datasets spanning multiple domains, exhibit significant reasoning, +> understanding, and planning capabilities across various tasks. This study +> presents the first-ever work in Arabic language integration within the +> Vision-and-Language Navigation (VLN) domain in robotics, an area that has been +> notably underexplored in existing research. We perform a comprehensive +> evaluation of state-of-the-art multi-lingual Small Language Models (SLMs), +> including GPT-4o mini, Llama 3 8B, and Phi-3 medium 14B, alongside the +> Arabic-centric LLM, Jais. Our approach utilizes the NavGPT framework, a pure +> LLM-based instruction-following navigation agent, to assess the impact of +> language on navigation reasoning through zero-shot sequential action prediction +> using the R2R dataset. Through comprehensive experiments, we demonstrate that +> our framework is capable of high-level planning for navigation tasks when +> provided with instructions in both English and Arabic. However, certain models +> struggled with reasoning and planning in the Arabic language due to inherent +> limitations in their capabilities, sub-optimal performance, and parsing issues. +> These findings highlight the importance of enhancing planning and reasoning +> capabilities in language models for effective navigation, emphasizing this as a +> key area for further development while also unlocking the potential of +> Arabic-language models for impactful real-world applications. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on the evaluation of Large Language Models (LLMs) for multilingual robotic navigation, rather than on engineering, design, or optimization of prompts specifically for LLMs. While LLMs are utilized, the core subject is not prompt engineering for text-based interactions, but rather the application of LLMs in navigation tasks across different languages. + +--- + +## [Detection, Retrieval, and Explanation Unified: A Violence Detection + System Based on Knowledge Graphs and GAT](https://arxiv.org/abs/http://arxiv.org/abs/2501.06224v1) +**arXiv ID:** http://arxiv.org/abs/2501.06224v1 + +**Abstract:** +> Recently, violence detection systems developed using unified multimodal +> models have achieved significant success and attracted widespread attention. +> However, most of these systems face two critical challenges: the lack of +> interpretability as black-box models and limited functionality, offering only +> classification or retrieval capabilities. To address these challenges, this +> paper proposes a novel interpretable violence detection system, termed the +> Three-in-One (TIO) System. The TIO system integrates knowledge graphs (KG) and +> graph attention networks (GAT) to provide three core functionalities: +> detection, retrieval, and explanation. Specifically, the system processes each +> video frame along with text descriptions generated by a large language model +> (LLM) for videos containing potential violent behavior. It employs ImageBind to +> generate high-dimensional embeddings for constructing a knowledge graph, uses +> GAT for reasoning, and applies lightweight time series modules to extract video +> embedding features. The final step connects a classifier and retriever for +> multi-functional outputs. The interpretability of KG enables the system to +> verify the reasoning process behind each output. Additionally, the paper +> introduces several lightweight methods to reduce the resource consumption of +> the TIO system and enhance its efficiency. Extensive experiments conducted on +> the XD-Violence and UCF-Crime datasets validate the effectiveness of the +> proposed system. A case study further reveals an intriguing phenomenon: as the +> number of bystanders increases, the occurrence of violent behavior tends to +> decrease. + +**Decision Explanation:** +Original decision: REJECT +The paper's primary focus is on developing a unified violence detection system using knowledge graphs and GAT, with LLMs being used only as a component for generating text descriptions, rather than the central focus being on prompt engineering for text-based interactions with LLMs. + +--- + +## [asanAI: In-Browser, No-Code, Offline-First Machine Learning Toolkit](https://arxiv.org/abs/http://arxiv.org/abs/2501.06226v1) +**arXiv ID:** http://arxiv.org/abs/2501.06226v1 + +**Abstract:** +> Machine learning (ML) has become crucial in modern life, with growing +> interest from researchers and the public. Despite its potential, a significant +> entry barrier prevents widespread adoption, making it challenging for +> non-experts to understand and implement ML techniques. The increasing desire to +> leverage ML is counterbalanced by its technical complexity, creating a gap +> between potential and practical application. This work introduces asanAI, an +> offline-first, open-source, no-code machine learning toolkit designed for users +> of all skill levels. It allows individuals to design, debug, train, and test ML +> models directly in a web browser, eliminating the need for software +> installations and coding. The toolkit runs on any device with a modern web +> browser, including smartphones, and ensures user privacy through local +> computations while utilizing WebGL for enhanced GPU performance. Users can +> quickly experiment with neural networks and train custom models using various +> data sources, supported by intuitive visualizations of network structures and +> data flows. asanAI simplifies the teaching of ML concepts in educational +> settings and is released under an open-source MIT license, encouraging +> modifications. It also supports exporting models in industry-ready formats, +> empowering a diverse range of users to effectively learn and apply machine +> learning in their projects. The proposed toolkit is successfully utilized by +> researchers of ScaDS.AI to swiftly draft and test machine learning ideas, by +> trainers to effectively educate enthusiasts, and by teachers to introduce +> contemporary ML topics in classrooms with minimal effort and high clarity. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a no-code machine learning toolkit (asanAI) that happens to utilize ML techniques, rather than engineering prompts specifically for Large Language Models (LLMs), thus not meeting the core subject criterion. + +--- + +## [Balanced Multi-view Clustering](https://arxiv.org/abs/http://arxiv.org/abs/2501.02564v2) +**arXiv ID:** http://arxiv.org/abs/2501.02564v2 + +**Abstract:** +> Multi-view clustering (MvC) aims to integrate information from different +> views to enhance the capability of the model in capturing the underlying data +> structures. The widely used joint training paradigm in MvC is potentially not +> fully leverage the multi-view information, since the imbalanced and +> under-optimized view-specific features caused by the uniform learning objective +> for all views. For instance, particular views with more discriminative +> information could dominate the learning process in the joint training paradigm, +> leading to other views being under-optimized. To alleviate this issue, we first +> analyze the imbalanced phenomenon in the joint-training paradigm of multi-view +> clustering from the perspective of gradient descent for each view-specific +> feature extractor. Then, we propose a novel balanced multi-view clustering +> (BMvC) method, which introduces a view-specific contrastive regularization +> (VCR) to modulate the optimization of each view. Concretely, VCR preserves the +> sample similarities captured from the joint features and view-specific ones +> into the clustering distributions corresponding to view-specific features to +> enhance the learning process of view-specific feature extractors. Additionally, +> a theoretical analysis is provided to illustrate that VCR adaptively modulates +> the magnitudes of gradients for updating the parameters of view-specific +> feature extractors to achieve a balanced multi-view learning procedure. In such +> a manner, BMvC achieves a better trade-off between the exploitation of +> view-specific patterns and the exploration of view-invariance patterns to fully +> learn the multi-view information for the clustering task. Finally, a set of +> experiments are conducted to verify the superiority of the proposed method +> compared with state-of-the-art approaches both on eight benchmark MvC datasets +> and two spatially resolved transcriptomics datasets. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on multi-view clustering, a general machine learning technique, and does not meet the 'MUST' criteria for prompt engineering specifically tailored for Large Language Models (LLMs), nor does it provide examples of textual input prompts impacting LLM output. + +--- + +## [Cracks in The Stack: Hidden Vulnerabilities and Licensing Risks in LLM + Pre-Training Datasets](https://arxiv.org/abs/http://arxiv.org/abs/2501.02628v1) +**arXiv ID:** http://arxiv.org/abs/2501.02628v1 + +**Abstract:** +> A critical part of creating code suggestion systems is the pre-training of +> Large Language Models on vast amounts of source code and natural language text, +> often of questionable origin or quality. This may contribute to the presence of +> bugs and vulnerabilities in code generated by LLMs. While efforts to identify +> bugs at or after code generation exist, it is preferable to pre-train or +> fine-tune LLMs on curated, high-quality, and compliant datasets. The need for +> vast amounts of training data necessitates that such curation be automated, +> minimizing human intervention. +> We propose an automated source code autocuration technique that leverages the +> complete version history of open-source software projects to improve the +> quality of training data. This approach leverages the version history of all +> OSS projects to identify training data samples that have been modified or have +> undergone changes in at least one OSS project, and pinpoint a subset of samples +> that include fixes for bugs or vulnerabilities. We evaluate this method using +> The Stack v2 dataset, and find that 17% of the code versions in the dataset +> have newer versions, with 17% of those representing bug fixes, including 2.36% +> addressing known CVEs. The deduplicated version of Stack v2 still includes +> blobs vulnerable to 6,947 known CVEs. Furthermore, 58% of the blobs in the +> dataset were never modified after creation, suggesting they likely represent +> software with minimal or no use. Misidentified blob origins present an +> additional challenge, as they lead to the inclusion of non-permissively +> licensed code, raising serious compliance concerns. +> By addressing these issues, the training of new models can avoid perpetuating +> buggy code patterns or license violations. We expect our results to inspire +> process improvements for automated data curation, with the potential to enhance +> the reliability of outputs generated by AI tools. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses primarily on the curation of pre-training datasets for LLMs to reduce bugs and licensing risks, rather than the engineering, design, or optimization of prompts specifically for Large Language Models, failing to meet the first 'MUST' criteria. + +--- + +## [Samba-ASR: State-Of-The-Art Speech Recognition Leveraging Structured + State-Space Models](https://arxiv.org/abs/http://arxiv.org/abs/2501.02832v3) +**arXiv ID:** http://arxiv.org/abs/2501.02832v3 + +**Abstract:** +> We propose Samba ASR,the first state of the art Automatic Speech +> Recognition(ASR)model leveraging the novel Mamba architecture as both encoder +> and decoder,built on the foundation of state space models(SSMs).Unlike +> transformerbased ASR models,which rely on self-attention mechanisms to capture +> dependencies,Samba ASR effectively models both local and global temporal +> dependencies using efficient statespace dynamics,achieving remarkable +> performance gains.By addressing the limitations of transformers,such as +> quadratic scaling with input length and difficulty in handling longrange +> dependencies,Samba ASR achieves superior accuracy and efficiency.Experimental +> results demonstrate that Samba ASR surpasses existing opensource +> transformerbased ASR models across various standard benchmarks,establishing it +> as the new state of theart in ASR.Extensive evaluations on the benchmark +> dataset show significant improvements in Word Error Rate(WER),with competitive +> performance even in lowresource scenarios.Furthermore,the inherent +> computational efficiency and parameter optimization of the Mamba architecture +> make Samba ASR a scalable and robust solution for diverse ASR tasks.Our +> contributions include the development of a new Samba ASR architecture for +> automatic speech recognition(ASR),demonstrating the superiority of structured +> statespace models(SSMs)over transformer based models for speech sequence +> processing.We provide a comprehensive evaluation on public +> benchmarks,showcasing stateoftheart(SOTA)performance,and present an indepth +> analysis of computational efficiency,robustness to noise,and sequence +> generalization.This work highlights the viability of Mamba SSMs as a +> transformerfree alternative for efficient and accurate ASR.By leveraging the +> advancements of statespace modeling,Samba ASR redefines ASR performance +> standards and sets a new benchmark for future research in this field. + +**Decision Explanation:** +Original decision: REJECT +The paper focuses on developing a new architecture (Mamba) for Automatic Speech Recognition (ASR), which is unrelated to Large Language Models (LLMs) and prompt engineering for text-based interactions, thus failing to meet the primary criteria. + +--- + +## [A Soft Sensor Method with Uncertainty-Awareness and Self-Explanation + Based on Large Language Models Enhanced by Domain Knowledge Retrieval](https://arxiv.org/abs/http://arxiv.org/abs/2501.03295v2) +**arXiv ID:** http://arxiv.org/abs/2501.03295v2 + +**Abstract:** +> Data-driven soft sensors are crucial in predicting key performance indicators +> in industrial systems. However, current methods predominantly rely on the +> supervised learning paradigms of parameter updating, which inherently faces +> challenges such as high development costs, poor robustness, training +> instability, and lack of interpretability. Recently, large language models +> (LLMs) have demonstrated significant potential across various domains, notably +> through In-Context Learning (ICL), which enables high-performance task +> execution with minimal input-label demonstrations and no prior training. This +> paper aims to replace supervised learning with the emerging ICL paradigm for +> soft sensor modeling to address existing challenges and explore new avenues for +> advancement. To achieve this, we propose a novel framework called the Few-shot +> Uncertainty-aware and self-Explaining Soft Sensor (LLM-FUESS), which includes +> the Zero-shot Auxiliary Variable Selector (LLM-ZAVS) and the Uncertainty-aware +> Few-shot Soft Sensor (LLM-UFSS). The LLM-ZAVS retrieves from the Industrial +> Knowledge Vector Storage to enhance LLMs' domain-specific knowledge, enabling +> zero-shot auxiliary variable selection. In the LLM-UFSS, we utilize text-based +> context demonstrations of structured data to prompt LLMs to execute ICL for +> predicting and propose a context sample retrieval augmentation strategy to +> improve performance. Additionally, we explored LLMs' AIGC and probabilistic +> characteristics to propose self-explanation and uncertainty quantification +> methods for constructing a trustworthy soft sensor. Extensive experiments +> demonstrate that our method achieved state-of-the-art predictive performance, +> strong robustness, and flexibility, effectively mitigates training instability +> found in traditional methods. To the best of our knowledge, this is the first +> work to establish soft sensor utilizing LLMs. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on replacing supervised learning with In-Context Learning for soft sensor modeling in industrial systems, utilizing LLMs as a core component within a larger system. While it mentions prompt enhancements (e.g., text-based context demonstrations, context sample retrieval augmentation), prompt engineering for text-based interactions with LLMs is not the central focus, but rather a means to achieve the paper's main objective of advancing soft sensor technology. + +--- + +## [FTA-FTL: A Fine-Tuned Aggregation Federated Transfer Learning Scheme for + Lithology Microscopic Image Classification](https://arxiv.org/abs/http://arxiv.org/abs/2501.03349v1) +**arXiv ID:** http://arxiv.org/abs/2501.03349v1 + +**Abstract:** +> Lithology discrimination is a crucial activity in characterizing oil +> reservoirs, and processing lithology microscopic images is an essential +> technique for investigating fossils and minerals and geological assessment of +> shale oil exploration. In this way, Deep Learning (DL) technique is a powerful +> approach for building robust classifier models. However, there is still a +> considerable challenge to collect and produce a large dataset. +> Transfer-learning and data augmentation techniques have emerged as popular +> approaches to tackle this problem. Furthermore, due to different reasons, +> especially data privacy, individuals, organizations, and industry companies +> often are not willing to share their sensitive data and information. Federated +> Learning (FL) has emerged to train a highly accurate central model across +> multiple decentralized edge servers without transferring sensitive data, +> preserving sensitive data, and enhancing security. This study involves two +> phases; the first phase is to conduct Lithology microscopic image +> classification on a small dataset using transfer learning. In doing so, various +> pre-trained DL model architectures are comprehensively compared for the +> classification task. In the second phase, we formulated the classification task +> to a Federated Transfer Learning (FTL) scheme and proposed a Fine-Tuned +> Aggregation strategy for Federated Learning (FTA-FTL). In order to perform a +> comprehensive experimental study, several metrics such as accuracy, f1 score, +> precision, specificity, sensitivity (recall), and confusion matrix are taken +> into account. The results are in excellent agreement and confirm the efficiency +> of the proposed scheme, and show that the proposed FTA-FTL algorithm is capable +> enough to achieve approximately the same results obtained by the centralized +> implementation for Lithology microscopic images classification task. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper focuses on Federated Transfer Learning for image classification, primarily dealing with Deep Learning model architectures and data privacy in the context of Lithology microscopic image classification, with no mention of Large Language Models (LLMs) or prompt engineering for text-based interactions." +} + +--- + +## [Not all tokens are created equal: Perplexity Attention Weighted Networks + for AI generated text detection](https://arxiv.org/abs/http://arxiv.org/abs/2501.03940v2) +**arXiv ID:** http://arxiv.org/abs/2501.03940v2 + +**Abstract:** +> The rapid advancement in large language models (LLMs) has significantly +> enhanced their ability to generate coherent and contextually relevant text, +> raising concerns about the misuse of AI-generated content and making it +> critical to detect it. However, the task remains challenging, particularly in +> unseen domains or with unfamiliar LLMs. Leveraging LLM next-token distribution +> outputs offers a theoretically appealing approach for detection, as they +> encapsulate insights from the models' extensive pre-training on diverse +> corpora. Despite its promise, zero-shot methods that attempt to operationalize +> these outputs have met with limited success. We hypothesize that one of the +> problems is that they use the mean to aggregate next-token distribution metrics +> across tokens, when some tokens are naturally easier or harder to predict and +> should be weighted differently. Based on this idea, we propose the Perplexity +> Attention Weighted Network (PAWN), which uses the last hidden states of the LLM +> and positions to weight the sum of a series of features based on metrics from +> the next-token distribution across the sequence length. Although not zero-shot, +> our method allows us to cache the last hidden states and next-token +> distribution metrics on disk, greatly reducing the training resource +> requirements. PAWN shows competitive and even better performance +> in-distribution than the strongest baselines (fine-tuned LMs) with a fraction +> of their trainable parameters. Our model also generalizes better to unseen +> domains and source models, with smaller variability in the decision boundary +> across distribution shifts. It is also more robust to adversarial attacks, and +> if the backbone has multilingual capabilities, it presents decent +> generalization to languages not seen during supervised training, with LLaMA3-1B +> reaching a mean macro-averaged F1 score of 81.46% in cross-validation with nine +> languages. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on developing a novel method (PAWN) for detecting AI-generated text, leveraging LLM outputs, rather than engineering or optimizing prompts for Large Language Models. Prompt manipulation for improving LLM performance is not the central concern. + +--- + +## [More is not always better? Enhancing Many-Shot In-Context Learning with + Differentiated and Reweighting Objectives](https://arxiv.org/abs/http://arxiv.org/abs/2501.04070v2) +**arXiv ID:** http://arxiv.org/abs/2501.04070v2 + +**Abstract:** +> Large language models (LLMs) excel at few-shot in-context learning (ICL) +> without requiring parameter updates. However, as the number of ICL +> demonstrations increases from a few to many, performance tends to plateau and +> eventually decline. We identify two primary causes for this trend: the +> suboptimal negative log-likelihood (NLL) optimization objective and the +> incremental data noise. To address these issues, we introduce DrICL, a novel +> optimization method that enhances model performance through Differentiated +> Learning and advantage-based Reweighting objectives. Globally, DrICL utilizes +> differentiated learning to optimize the NLL objective, ensuring that many-shot +> performance surpasses zero-shot levels. Locally, it dynamically adjusts the +> weighting of many-shot demonstrations by leveraging cumulative advantages +> inspired by reinforcement learning, thereby improving generalization. This +> approach allows the model to handle varying numbers of shots effectively, +> mitigating the impact of noisy data. Recognizing the lack of multi-task +> datasets with diverse many-shot distributions, we develop the Many-Shot ICL +> Benchmark (ICL-50)-a large-scale benchmark of 50 tasks that cover shot numbers +> from 1 to 350 within sequences of up to 8,000 tokens-for fine-tuning purposes. +> ICL-50 facilitates the evaluation of many-shot ICL strategies across seven +> prominent NLP tasks and 50 distinct datasets. Experimental results demonstrate +> that LLMs enhanced with DrICL achieve significant improvements in many-shot +> setups across various tasks, including both in-domain and out-of-domain +> scenarios. We release the code and benchmark dataset hoping to facilitate +> further research in many-shot ICL. + +**Decision Explanation:** +Original response: +{ + "decision": "REJECT", + "explanation": "The paper primarily focuses on enhancing many-shot in-context learning through novel optimization methods (Differentiated Learning and Reweighting objectives), rather than specifically on the engineering, design, or optimization of textual input prompts for Large Language Models (LLMs). The core subject is optimization of LLM performance through training methods, not prompt engineering." +} + +--- + +## [Cosmos World Foundation Model Platform for Physical AI](https://arxiv.org/abs/http://arxiv.org/abs/2501.03575v1) +**arXiv ID:** http://arxiv.org/abs/2501.03575v1 + +**Abstract:** +> Physical AI needs to be trained digitally first. It needs a digital twin of +> itself, the policy model, and a digital twin of the world, the world model. In +> this paper, we present the Cosmos World Foundation Model Platform to help +> developers build customized world models for their Physical AI setups. We +> position a world foundation model as a general-purpose world model that can be +> fine-tuned into customized world models for downstream applications. Our +> platform covers a video curation pipeline, pre-trained world foundation models, +> examples of post-training of pre-trained world foundation models, and video +> tokenizers. To help Physical AI builders solve the most critical problems of +> our society, we make our platform open-source and our models open-weight with +> permissive licenses available via https://github.com/NVIDIA/Cosmos. + +**Decision Explanation:** +Original decision: REJECT +The paper primarily focuses on a platform for Physical AI and world model development, with emphasis on video curation, model fine-tuning, and open-source resources. It does not meet the core criteria of focusing primarily on prompt engineering for Large Language Models (LLMs), manipulating textual input prompts to improve LLM performance, or providing concrete examples of prompts and their impact on LLM output. + +--- +